Handbook of Advanced Ceramics. Materials, Applications, Processing, and Properties 978-0-12-385469-8

Citation preview

Chapter 1.1.1

The Latest Analytical Electron Microscope and its Application to Ceramics Yoshiyasu Harada* and Yuichi Ikuharay JEOL Ltd., 3-1-2 Musashino, Akishima, Tokyo 196-8558, Japan, y Institute of Engineering Innovation, School of Engineering, University of Tokyo, 2-11-16 Yayoi, Bunkyo, Tokyo 113-8656, Japan *

Chapter Outline 1. 2. 3. 4. 5.

Introduction General Overview of Analytical Electron Microscope Transmission Electron Microscopy Scanning Transmission Electron Microscopy Analysis Method

3 3 7 8 10

1. INTRODUCTION The transmission electron microscope (TEM) has been utilized for observation of fine structures and analysis of a crystalline structure of material, since its invention in 1932 [1]. After that, in the 1970s, the functions of scanning transmission electron microscope (STEM), energy dispersive spectroscopy (EDS), electron energy loss spectroscopy (EELS), etc. were built in TEM, and its application fields have been drastically expanded to include compositional analysis and analysis of an electronic structure as well as image observation by TEM/STEM up to the present date. As such, the TEM/STEM, which can make both the image observation and the analysis, is called an analytical electron microscope (AEM) [2,3]. Then, the electron gun and objective lens, which are important elemental technologies of the electron microscope, have been greatly improved, and the performance of AEM has improved drastically. That is, as for the electron gun, in the 1990s, instead of the conventional thermionic emission electron gun (TEG), field emission electron gun (FEG) of high brightness has been put into practical use, and the current of the nano-probe focused on a specimen has been greatly increased. In addition, as for the objective lens, by commercialization of TEM equipped with a condenser objective lens (C/O lens) in the 1980s, observation of TEM image and STEM image became possible by using the identical objective lens excitation, and then

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00001-0 Copyright Ó 2013 Elsevier Inc. All rights reserved.

6. Application of Analytical Electron Microscopy to Ceramics 7. Conclusion References

12 20 20

research of material science field has been developed. Moreover, in the 2000s, the corrector of spherical aberration, which used to restrict the resolution of TEM/STEM, has been put into practical use; therefore, the image observation and the analysis of atomic levels have become possible, and AEM has been utilized as a useful evaluation instrument in the most advanced area of material development in the material science fields such as metal, semiconductor, and ceramics, etc. In this article, we will describe the general overview of the most advanced AEM first, and will especially explain the electron gun and objective lens as well as the spherical aberration corrector, which relates to the AEM performance. Next, we will describe the image formation in TEM and STEM and analysis method (EDS and EELS), which are indispensable for applying AEM to material science field. Finally, we will introduce how AEM is utilized in research of ceramics material concretely based on several application examples.

2. GENERAL OVERVIEW OF ANALYTICAL ELECTRON MICROSCOPE 2.1. Various Information Obtained by Interaction between Electron Beam and Specimen When electron beams are irradiated to a specimen, as shown in Figure 1, as a result of an interaction between the

3

4

FIGURE 1 Various information obtained by interaction between primary electrons and specimen.

electron beam and specimen, information such as the secondary electron, Auger electron, backscattered electron, absorbed electron, characteristic X-ray, cathode luminescence, etc. can be obtained. In addition, in case the specimen is a thin film, by interaction between the electron beam and the specimen, information of the transmitted electron, which suffered elastic scattering and inelastic scattering, can be obtained. By utilizing this information, AEM can obtain a great deal of knowledge about a specimen. In this chapter, we cover major image-forming methods by TEM and STEM which obtain magnified images of a specimen by using transmitted electrons; in addition, we describe the compositional analysis and chemical state analysis of a specimen using EDS and EELS.

2.2. Composition of Analytical Electron Microscope Figure 2 shows the appearance of the most advanced AEM (JEM-ARM200F) with spherical aberration corrector (Cscorrector), which enables the observation of both TEM image and STEM image [4]. Electron beams are emitted from the electron gun at the top, and the emitted electron beams are accelerated and magnified (in the case of TEM) or demagnified (in the case of STEM) by the illumination lens system, and then irradiated to the specimen. In the case of TEM, electrons passing through a specimen are formed into an image by image-forming lens system, and the image are observed on a fluorescent screen in an image observation chamber. In addition, they can be observed through a CCD camera in an image observation chamber or, at the lower part of it, on a liquid crystal monitor. In the case of STEM, by scanning a demagnified electron probe, a synchronized signal is detected by various detectors attached at the lower part of the image observation chamber, and the image can be observed on a liquid crystal monitor. Also, as analysis function, EDS detector is attached in the specimen chamber, and EELS detector is

Handbook of Advanced Ceramics

attached at the lower part of the camera chamber; thus, compositional analysis can be possible. Among many elements composing of AEM, the factors to determine the performance are the electron gun and objective lens (OL) including aberration correction technology. With the instrument shown in Figure 2, FEG is provided as an electron gun; however, there are instruments which have TEG where tungsten (W) hairpin filament and lanthanum hexaboride (LaB6), etc. are used as an electron source. We describe the characteristics comparison of these electron guns in 2.3. In addition, as for the OL and aberration correction technology, we describe in 2.4 and 2.5 respectively.

2.3. Performance Comparison of Various Electron Guns Table 1 shows the features of various electron guns. Since the invention of the electron microscope, TEG with W hairpin filament has been utilized as a standard electron gun until the 1970s, but in the 1980s, the use of TEG with LaB6 [5] of high brightness became popular. Later on, in the 1990s, FEG of higher brightness was put into practical use, and began being used as the electron gun of AEM. There are principally two types of FEGs. One is the cold-type FEG(CFEG) which uses W(310) as electron source which is kept at room temperature [6]. The other is the thermal-type FEG (TFEG) with heated ZrO/W(100) [7] as an electron source. The feature of C-FEG is that the energy spread of an emission current is small, while the features of T-FEG are that large emission currents can be taken, the current stability is high, and that the ease of operation is also good. With the most advanced AEM, though it has both advantages and disadvantages, either type of FEG has been used as a standard. Recently, Schottky-type electron gun whose feature is about same as the T-FEG is commonly used in place of T-FEG.

2.4. Objective Lens The resolution (d) of the TEM is expressed as d ¼ 0.65Cs1/4$l3/4. Here, Cs is the spherical aberration coefficient and l is the wavelength of electron. Since the invention of the electron microscope, in order to improve the resolution, efforts have been made whether making spherical aberration coefficient of OL smaller or raising the accelerating voltage, that is, making l shorter. At present, the standard accelerating voltage of AEM is 200e300 kV, that is, l is 0.00251e0.00197 nm. On the other hand, the efforts to reduce Cs of the OL have been done by many researchers since the invention of TEM. There are principally two trends. In other words, the shape of the OL pole pieces differs greatly, whether inserting a specimen from the top through the upper bore of the pole piece (top entry method) or inserting a specimen from the side through the gap (side entry method) to enter a specimen in the

Chapter | 1.1.1

The Latest Analytical Electron Microscope and its Application to Ceramics

5

FIGURE 2 Appearance of an analytical electron microscope (JEM-ARM200F). For color version of this figure, the reader is referred to the online version of this book.

gap between pole pieces . To correspond to the top entry method, it is necessary to select asymmetric pole pieces where the bore diameter of the upper pole (b1) is larger than that of the lower pole (b2). Also, to correspond to the side entry method, it is necessary to select symmetric pole pieces of b1 ¼ b2, and to make the gap length(s) of the pole pieces relatively greater. As for the asymmetric pole pieces, since the top entry specimen stage is more stable in terms of antivibration properties, it has been improved along the trend that pursues the resolution of the TEM image. However, in order to use it in the case of AEM, it is necessary to change the

excitation of the OL in order to switch from TEM image to STEM image, which is difficult in terms of ease of operation, so that its use in standard AEM was stopped in the mid-1980s. On the other hand, as for the symmetric pole pieces, in consideration that the improvement of antivibration properties of the side entry specimen stage has advanced, as the TEM image and STEM image can be observed by using an identical excitation of the OL, it is excellent in terms of ease of operation; therefore, it developed rapidly in the 1980s, and it has been used as a standard OL of AEM up to the present date. Table 2 shows the shape of pole pieces in these two types of

6

Handbook of Advanced Ceramics

TABLE 1 Comparison of Characteristics of Various Electron Guns Thermionic emission

Field emission

W-hairpin

LaB6

T-FEG ZrO/W(100)

C-FEG W(310)

Brightness (A/cm2 $ sr) at 200 kV

~5  105

~5  106

~5  108

~5  108

Optical source size

~20 mm

~10 mm

~20 nm

~10 nm

Energy spread (eV)

~3

~2

~0.8

~0.3

Operating temperature (K)

~2800

~1800

~1800

~300

Typical vacuum (Pa)

~10

Emission current (mA)

~100

4

TABLE 2 Comparison of Optical Properties of Symmetrical and Asymmetrical Objective Lenses Symmetrical b1 ¼ b2

Asymmetrical b1 > b2

b1(mm)

2.0

5.0

s(mm)

2.5

2.0

b2(mm)

2.0

1.0

Cs(mm)

0.5

0.5

Cc(mm)

0.7

1.0

f0(mm)

1.2

1.4

NI/Vr

30.0

28.0

OL to obtain 0.5 mm Cs at an accelerating voltage of 100 kV. Here, Cc is the chromatic aberration coefficient, fo is the focal pffiffiffiffiffi length, and NI= Vr is the excitation parameter (NI: ampere turn, Vr: relativistic accelerating voltage). In the case of an asymmetric lens [8], it is necessary to make b1 great, therefore, b1 ¼ 5 mm. In addition, in the case of a symmetric lens which is used as C/O lens [9], it is necessary to make s relatively large, therefore, s ¼ 2.5 mm. If we assume that Cs of approximately 0.5 mm can be realized at the accelerating voltage in the range of 100e300 kV with either OL, the resolution at that time will be 0.26 nm, 0.19 nm, and 0.16 nm, respectively, at 100 kV, 200 kV, and 300 kV. When the Cscorrector is not used, these resolutions are the best at each accelerating voltage.

2.5. Spherical Aberration Corrector In the mid-1990s, Cs-corrector was put into practical use, and, in the 2000s, AEM with Cs-corrector of TEM/STEM has rapidly become widely used, until today. Right after the electron microscope was invented, a study on the aberration of the electron lens was executed

~10 ~30

5

~10 ~100

7

~10

8

~10

actively. Scherzer demonstrated that Cs and chromatic aberration coefficient (Cc) of rotational symmetric electron lens cannot take a negative value [10], and proposed a method to correct these aberrations by using a rotational asymmetric electron lens which can take a negative value of these aberration coefficients [11]. Later on, many researchers have done theoretical and experimental studies of aberration correction by using rotational asymmetric electron lens such as the quadrupole lens, octupole lens, hexapole lens, and dodecapole lens; however, these were hardly put into practical use. The reasons are that (1) the simulation of rotational asymmetric electron lens has never been accurately done, (2) the technology to process the rotational asymmetric electron lens with accuracy was not available, (3) the electric stability of rotational asymmetric electron lens has never been improved, and (4) the electric stability of the electron microscope basic unit has not been sufficient. However, in 1995, about 50 years after the proposal by Scherzer, the above-described problems were resolved by Rose and Haider who are the successors of Scherzer’s research laboratory, and Cs-corrector using a 2-stage hexapole lens was put into practical use [12,13]. Lens structure of the corrector, which can correct spherical aberration of the third order, is shown in Figure 3. Figure 3(a) is the case of TEM which forms a magnified image of the specimen, and Cs-corrector is placed at the lower side of the OL. Figure 3(b) shows the case of STEM which forms the nano-probe on the specimen, and the Cs-corrector is placed at the upper side of OL. With this Cscorrector, along with the large three-fold astigmatism, negative spherical aberration appearing in terms of higher order from the three-fold astigmatism is utilized for the spherical aberration correction of the OL with the 1st-stage and 2nd-stage hexapole lens. The three-fold astigmatism arising from the hexapole lens of the 1st stage and the other of 2nd stage is set in reverse directions; therefore, it became

Chapter | 1.1.1

The Latest Analytical Electron Microscope and its Application to Ceramics

7

FIGURE 4 Ray path of image-forming lens system to observe image (a) and (b) and diffraction (c). FIGURE 3 TEM (a) and STEM (b) Cs-correctors with two hexapole elements.

the optical system that makes three-fold astigmatism zero as a whole. With the electron microscope which has the Cscorrector, if the accelerating voltage is 300 kV, the resolution of 0.05 nm is obtained in TEM/STEM, and the minimum probe diameter on the specimen is approximately 0.1 nm[14]. In addition, in case the accelerating voltage is 200 kV, the resolution is 0.1 nm in TEM/STEM, and approximately 0.1 nm as the minimum probe diameter on the specimen is obtained [4].

3. TRANSMISSION ELECTRON MICROSCOPY 3.1. Electron Microscope Image and Electron Diffraction Pattern In the case of an observation of a magnified image and observation of an electron diffraction pattern, which are the basic observation methods of TEM, the ray diagram of the image-forming lens system is shown in Figure 4. Observation of a wide range of magnified images is possible from low magnification of 50 times to high magnification of 1,000,000 times. Figure 4(a) shows a ray diagram to obtain low magnification of 50e1000 times, and in order to observe an image of a wide field of view with small aberration, excitation of OL is cut, objective mini (OM) lens is used instead of the OL, and the image is magnified by the intermediate lens (IL) and projector lens (PL), and can be observed on the fluorescent screen. Figure 4(b) is the ray diagram to obtain a magnified image at high magnification of 1000e1000,000 times, and the image magnified by the OL is magnified by several intermediate lens (IL1eIL3) and

PL further, and observed on the fluorescent screen. The insertion of OL aperture on the back focal plane of the OL is possible, and by selecting the size of the OL aperture, observation of magnified image, which has various contrasts, is possible. Figure 4(c) is the ray diagram to obtain an electron diffraction pattern. The electron diffraction pattern formed at the back focal plane of the OL, is magnified by IL and PL, and observed on the fluorescent screen. By selecting the size of the selected-area (SA) aperture placed on the object plane of the IL1, selection of an area on the specimen, which corresponds to the electron diffraction pattern, is possible. Electron microscope image and electron diffraction pattern can be switched on at the flip of a switch and observed. In case the specimen is crystal, a spot-shaped electron diffraction pattern is obtained by Bragg reflection on the crystal plane which is parallel to the incident electron beam, and by selecting only the transmitted spot without being diffracted by the OL aperture, a bright field image can be observed, and by selecting the diffraction spot only, a dark field image can be observed. With a bright field image and dark field image, the bright and dark contrasts are observed, and these contrasts differ depending on the specimen thickness and the local distortion accompanied by the curvature of the crystal plane, dislocation, stacking fault, etc. As for the reason why these contrasts are formed, explanation by the electron diffraction theory [15], which consists of the kinematical theory and dynamical theory, is possible, but we do not cover it here.

3.2. High-Resolution Electron Microscopy There are principally two methods of high-resolution electron microscopy (HREM) to study crystal material [16].

8

One is the observation of crystal lattice images. When the electron beam is illuminated in parallel to crystal zone axis, a two-dimensional electron diffraction pattern is obtained . If only the transmitted wave (spot) and the diffracted wave, which is the closest to the transmitted wave, are taken in the OL aperture for image formation, a two-dimensional crystal lattice image, which has information of unit cell of crystal by interference of transmitted wave and diffracted wave, can be observed. With this method, information on defects such as twins, stacking fault, dislocation, grain boundary, etc. can be obtained. The other is the observation of crystal structure images. If the transmitted wave and many diffracted waves are taken in the OL aperture, and an image is formed under the condition to be explained later, crystal structure image corresponding to its projected potential can be observed. With the most advanced TEM with Cs-corrector, a crystal lattice image and crystal structure image can be easily observed at a resolution of about 0.1 nm. In the crystal structure image, the atomic column in thin crystal (several atomic rows lined up in the direction of specimen thickness) can be observed directly. As for the thin specimen like this, amplitude of the electron wave passing through specimen hardly changes, and only phase changes; therefore, such a thin specimen is called a phase object. The image formation theory of phase object was reported by Scherzer in 1949 [17]. We do not cover the details here, but when the contrast transfer function of the OL, which is determined by the Cs of OL and defocus values (df), fell under a certain value without vibration in a wide spatial frequency area, the projected potential of the crystal, that is, the crystal structure, will be well reflected to an electron microscopic image. In other words, the resolution is the highest, and the optimal df is given as df ¼ 1.2(Csl)1/2, which is called Scherzer focus. The resolution (d) of TEM is, as explained in Section 2.4, expressed as d ¼ 0.65Cs1/4$l3/4. In addition, in case the actual HREM image is observed, it is necessary to consider the deterioration of the resolution, caused by chromatic aberration attributed to energy spread (dE) of the incident electron beam and Cc of OL, and illumination angle (a) of the electron beam on the specimen. These influences are described as an envelope function showing the attenuation of high-frequency component of a scattering wave [18,19]. Figure 5 shows the calculation example of contrast transfer function in consideration of the envelope function when an electron microscope of 300 kV accelerating voltage with Cscorrector is provided with T-FEG (dE ¼ 0.8 eV), or C-FEG (dE ¼ 0.3 eV) In the case of T-FEG, as energy spread is large, observation is only possible at a resolution of about 0.07 nm, but, in the case of C-FEG, it is estimated that there is a possibility that a resolution of 0.05 nm can be obtained.

Handbook of Advanced Ceramics

FIGURE 5 Contrast transfer function in the case of C-FEG (dE ¼ 0.3 eV) and T-FEG(dE ¼ 0.8 eV).

What to pay attention most in interpreting the image observed by this image-forming method is that the contrast of the electron microscopic image changes greatly by df, that is, by vibration of contrast transfer function. If one wrong step is taken, there is a possibility that an incorrect image interpretation will be made. In this sense, it is inevitable to conduct image observation under the Scherzer focus condition. Also, it is essential to interpret the observed image carefully, by comparing it with the computer simulation image by using the multi-slice method which was offered by Cowley & Moodie [20].

4. SCANNING TRANSMISSION ELECTRON MICROSCOPY In the 1980s, as the C/O lens was adopted as a standard OL for TEM, switching between TEM and STEM images became easy, and thus the observation method by STEM rapidly became common. With the C/O lens, a specimen is placed almost in the center of the gap of OL pole pieces, and the prefield above the specimen works as a CL and the postfield below the specimen works as an OL. Figure 6 shows a typical ray diagram of the illumination lens system in TEM mode and in STEM mode. Figure 6(a) shows the ray diagram of the illumination lens system in the TEM mode. By focusing electron beams at the front focal plane of the prefield of OL by using a CM lens, the parallel beam is irradiated to the specimen and an electron microscopic image of high coherency can be observed. Figure 6(b) and (c) show the ray diagram of the illumination lens system in STEM mode. In STEM mode, electron beams are demagnified to form a nano-probe on the specimen and the electron probe diameter is a great factor in determining the resolution of the STEM image. Figure 6(b) shows that, if excitation of the CM lens is disconnected and if the size of the CL aperture is constant, the specimen illumination angle becomes the maximum angle a1, which is determined by the size of the CL aperture. Also, by changing the excitation of the CM lens which is exited, the

Chapter | 1.1.1

The Latest Analytical Electron Microscope and its Application to Ceramics

FIGURE 6 Ray path of illumination lens system to observe TEM image (a) and STEM Image (b and c).

illumination angle (a2) can be controlled. As such, to change the illumination angle, it is not only important in changing the electron probe current on the specimen, but also important in determining what kind of transmitted electrons to be taken in STEM detectors, which will be mentioned later. In addition, under the condition of image-forming lens system to obtain an electron diffraction pattern as described in Section 3.1, if a large illumination angle (convergence angle) is set in a similar way as the STEM mode in Figure 6(b) and (c), a convergent beam electron diffraction (CBED) pattern can be obtained from the nanometer area. If CBED is used, determination of the point group and space group of the crystal is possible [21]; moreover, measurement of the lattice distortion around the grain boundary in the crystal is possible.

4.1. Various Detection Methods of Transmitted Electron Electrons passing through a thin film specimen contain transmitted electrons without being suffered by scattering, electrons that suffered elastic scattering, and electrons that suffered inelastic scattering. In general, it is known that electrons that suffered elastic scattering will distribute in a relatively wide scattering angle from 50 to 100 mrad or so, and electrons that suffered inelastic scattering which excites an electron in an atom will distribute in a relatively small scattering angle of about 1 mrad. In the case of TEM, as the parallel electron beam illuminates the specimen, the electron diffraction pattern of the crystalline specimen becomes spot shaped, but in the case of STEM, as

9

mentioned earlier, as incident electron beams are focused on the specimen, in case the Cs-corrector is not installed, the convergent angle (full angle) is about 10e20 mrad or so, and in case Cs-corrector is installed, the convergent angle (full angle) becomes larger to be 50e100 mrad or so; then, in the case of the crystalline specimen, the electron diffraction pattern made at the back focal plane of the OL will be disk shaped. The image formation of STEM is determined by taking which part of the disk-shaped electron diffraction pattern in the detector [22]. If the detector is set at the transmission disk corresponding to the convergent angle of the incident electron beam, a bright field image can be obtained. If the detector is set at the diffraction disk, a dark field image is obtained. In addition, if the detector is set where the transmission disk and the diffraction disk overlap, as in TEM’s case, a crystal lattice image which is the typical example of phase contrast can be observed. In other words, it has been reported that the reciprocity theorem can apply to the image formation by the TEM and STEM [23,24]. Here, we describe two detection methods which are extremely useful in the research of material science field among several STEM image detection methods, namely, HAADF (high-angle annular dark-field) method and ABF (annular bright field) method. The schematic diagram of those methods is shown in Figure 7. Figure 7(a) shows the schematic diagram of the HAADF method and Figure 7(b) shows the schematic diagram of the ABF method.

4.2. HAADF-STEM HAADF-STEM is a method to form an STEM image by detecting only high-angle scattered electrons by using an annular detector. The annular detector was first used by Crewe [25], in order to observe the image of a single atom which depends on the atomic number, by forming STEM image in proportion with the elastically scattered electron of 10 mrad or more and inelastically scattered electron of 1 mrad or less. Later, since the result was presented in the field of crystal material by Pennycook et al. by using HAADF-STEM with an annular detector of 50e70 mrad or so [26], many researchers have made use of HAADF-STEM. The elastically scattered electrons which were detected by the annular detector of 50e70 mrad or so are the Rutherford scattering, and as its intensity (I) is in proportion to the square of the atomic number (Z) of the atom in the specimen, the image contrast observed by using HAADF-STEM is called a Z square contrast image. As the scattering angle becomes large, the atomic scattering factor of elastic scattering reduces rapidly and the intensity of thermal diffuse scattering (TDS) by phonon excitation is dominant. As a result, TDS forms the major contrast of HAADF-STEM image [27]. As this image is basically an incoherent image, the image interpretation is easy which is

10

FIGURE 7 Schematic detector arrangement to observe HAADF-STEM image (a) and ABF-STEM image (b). For color version of this figure, the reader is referred to the online version of this book.

Handbook of Advanced Ceramics

(a)

(b)

HAADF

α

ABF

Probe-forming lens

α

Specimen

β2

a great feature and different from the HREM image in the TEM and the STEM phase contrast images that were mentioned earlier. That is, if the specimen thickness is constant, the atomic column of the large Z shows a brighter contrast than that of a small Z; therefore, it has a feature that the difference between the atomic columns can be clearly distinguished.

4.3. ABF-STEM ABF-STEM [28,29], which was recently proposed by JEOL, University of Tokyo, and Japan Fine Ceramics Center, is a method to form an STEM image by using an annular detector set at the transmission disk area corresponding to the convergent angle of the incident electron beam. It was verified that contrast of the atomic column of light element in ABF-STEM is higher than in HAADF-STEM by systematic experiments and simulation of STEM image [30]. They used several kinds of oxide as a specimen and compared the images observed by using ABF-STEM and HAADF-STEM, and showed that the atomic column of oxygen was not observed by HAADF-STEM but by ABFSTEM, in which the atomic column of oxygen can be clearly observed as a dark contrast [29]. Also, they made a simulation of an ABF-STEM image by systematically changing the specimen thickness and defocus values, made a defocus-thickness map, and made it clear that the STEM images of both the light atom column and heavy atom column can be observed with a wide range of defocus values and specimen thickness [30]. ABF-STEM is a novel observation method of the STEM image, which has been proposed recently, and it is a useful method which enables a simultaneous observation of both the column of light atom and heavy atom. It is expected that it will be utilized by many researchers in the field of ceramics.

Probe-forming lens Specimen

β1

HAADF detector

β2

β1

ABF detector

5. ANALYSIS METHOD Here, we explain the EDS and EELS analyses which have been widely used since the 1970s as the most basic compositional analysis of AEM. EDS is a method to analyze the composition by a pulse-height analysis of the energy of the characteristic X-ray coming from the specimen by using an energy dispersive X-ray detector. EELS is a method of compositional analysis by measuring energy loss of transmitted electrons which suffered inelastic scattering in the specimen.

5.1. EDS Analysis Figure 8 shows the principle of the EDS analysis. By using a semiconductor detector (SSD: solid state detector) of a high-purity silicon single crystal doped with a trace of lithium, it can make an elemental analysis by immediately converting all of the energy of a characteristic X-ray generated from the specimen into a pulse voltage according to the amplitude of the field-effect transistor (FET), and counting the pulse number by a multichannel pulse-height analyzer. The spectrum is indicated where X-ray energy is in horizontal axis and the number of photons is in the

FIGURE 8 Constitution and block diagram of EDS system.

Chapter | 1.1.1

The Latest Analytical Electron Microscope and its Application to Ceramics

vertical axis. In the EDS system, it is always necessary to cool SSD and FET to liquid nitrogen temperature in order to suppress their thermal noise. There are two kinds of EDS detectors which are used with AEM. One is a normal detector (Be-window type) which detects an X-ray through Be-window of about 7-mm thickness and the other is UTW (ultrathin window)-type detector which detects an X-ray through the window where a thin plastic film is coated with about 0.1-mm-thick aluminum. The elements that are heavier than Na (Z ¼ 11) are detectable in case of Be-window type, because Be film absorbs X-ray, while with the UTW type, the detectable elements are the ones that are heavier than C (Z ¼ 6). In other words, with the UTW type, the analysis capability of light elements such as C, N, O, etc. is the biggest feature. The feature of Be-window type is the high detection efficiency because it can make an effective area of the detector large. EDS can provide not only a qualitative analysis but also a quantitative analysis and it has the feature that elemental analysis can be done easily. However, the utmost disadvantage is that the energy resolution is as low as about 150 eV. Therefore, spectrum overlapping occurs and attention will be needed in interpretation. There are principally two methods in EDS analysis. One is the point analysis method in which the electron probe is stopped at one point on the specimen by using TEM/STEM, and an X-ray spectrum is acquired. The other is the elemental mapping method, which scans two-dimensionally the electron probe on the specimen by using STEM, modulates the brightness corresponding to the intensity of a certain characteristic X-ray, synchronizes with a scanning signal, and displays the two-dimensional image of the characteristic X-ray intensity on the liquid crystal monitor. The most advanced AEM with FEG along with a Cs-corrector can focus the minimum electron probe diameter on the specimen to be 0.1 nm or so. Therefore, the atomic level point analysis and two-dimension elemental mapping are possible and is widely used for research in the material science field.

11

FIGURE 9 Parallel detection system in EELS.

specimen, composition analysis, and chemical state analysis is possible. In addition, focusing on the fine structure of the inner-shell electron excitation spectrum, that is, the energy loss near-edge structure (ELNES), information on the electronic structure can be obtained. EELS spectrometer which measures the energy of the transmitted electron has several types. Here the most commonly used parallel detection system [31] is introduced as shown in Figure 9. The parallel detection system consists of a parallel detector of semiconductor which is composed of a fluorescent material (YAG), a fiberoptic plate, and photo diode array. As the parallel detector can read signals of each channel at the same time, it has the characteristics of high detection efficiency. Table 3 shows a comparison of the UTW/EDS and the EELS analyses. In the case of the EELS analysis, as transmitted electron energy loss is to be measured, the specimen thickness needs to be thin enough. The thickness of the specimen is said to be appropriate to be about the same degree of thickness as the mean free path(lp) of plasmon loss. If the specimen becomes thick, the signal of plasmon loss increases and it will become the background in detecting a signal of the inner-shell electron excitation. The detection element is possible from Li(3) to U(92) and its feature is the high-energy resolution compared with the EDS analysis.

5.2. EELS Analysis Among the electrons passing through the thin film specimen, transmitted electrons, elastically scattered electrons, and inelastically scattered electrons which lose energy as a result of interaction with an atom in the specimen exist. Inelastically scattered electrons arise from lattice vibration (phonon excitation), collective excitation of valence electron (plasmon excitation), interband transition, inner-shell electron excitation (core electron excitation), etc., and by measuring the energy of the inelastically scattered electrons, information such as the electronic structure and composition of the specimen can be obtained. Focusing on the energy loss peak accompanied by the inner-shell electron excitation, identification of the element in the

TABLE 3 Comparison of Characteristics of UTW/EDS and EELS UTW/ EDS

EELS

Conditions of specimen optional

thin film (~lp)

Detectable elements

C(6) ~U(92)

Li(3)~U(92)

Energy resolution (eV)

~150

~1

S/B ratio

~80

~0.3

Element mapping

easy

necessary data processing

12

Handbook of Advanced Ceramics

FIGURE 10 Spectra of EELS and UTW/EDS obtained by using BaTiO3. For color version of this figure, the reader is referred to the online version of this book.

Figure 10 shows the spectra of EELS and UTW/EDS by using barium titanate (BaTiO3) as a specimen with AEM of 200 kV accelerating voltage. In the UTW/EDS analysis, peaks of TieKa (4.51 keV) and BaeLa (4.47 keV), TieKb (4.93 keV) and BaeLb (4.83 keV), and OeKa (0.53 keV) and TieLa (0.45 keV) overlapped, and identification of Ti and Ba was not possible. On the other hand, with the EELS analysis, L-absorption edge of Ti (0.46 keV) and M-absorption edge (0.80 keV) of Ba have been clearly separated; thus, it can be understood that the resolution of EELS is superior

to the one in EDS. In addition, though oxygen is detected in either spectrum, it is demonstrated that while in the EDS analysis, the ratio of signal (S) and background (B), S/B ratio is high, while in the EELS analysis, the S/B ratio is low.

6. APPLICATION OF ANALYTICAL ELECTRON MICROSCOPY TO CERAMICS The mechanical and electromagnetic properties of a ceramic material have a close relation to microstructures. The

Chapter | 1.1.1

The Latest Analytical Electron Microscope and its Application to Ceramics

microstructure contains grain boundary, hetero-interface, dislocation, planar defect, precipitate, composition distribution, etc., and each factor determines various properties of material. Recently, as the material synthetic process becomes more accurate, the microstructure of material began being controlled in atomic levels. From such a viewpoint, characterization of the microstructure, which has a higher resolution and accuracy, is needed. As a method to observe the inner microstructures of a material, the transmission electron microscopy (TEM) is the most useful method. Moreover, it can also handle a structure analysis of the atomic level, local area compositional and chemical bonding state analysis in nanometer order, and the information obtained by using this method is quite a lot. In this chapter, in relation with the above-mentioned clause, we will explain the most advanced application examples by using high-resolution electron microscopy (HREM), CBED, scanning transmission electron microscopy (STEM), EDS, and EELS.

6.1. Application of High-Resolution Electron Microscopy When grain boundary or interface of a material is observed at a high resolution, the interface has to be edge on (meaning that the incident direction of the electron beam has to be parallel to the interface), and the Bragg condition of two crystal grains which sandwiches the interface has to be satisfied. Figure 11 shows the observation procedure for the grain boundary. This example shows the case where a thin amorphous phase exists in the ceramic grain boundary [32]. When the electron beam is irradiated at the grain boundary in the edge-on state, the electron diffraction pattern from two crystal grains and the diffuse scattering electrons from the amorphous phase are formed on the back focal plane of the OL. By taking in these waves in the OL

FIGURE 11 Observation procedure for the grain boundary.

13

aperture and form image, an HREM image of the grain boundary can be obtained. The size of the OL aperture is determined by considering the resolution of the electron microscope and image contrast. In the case of grainboundary observation of a polycrystal, as the grain boundary which satisfies the above-mentioned observation conditions are limited, it is necessary to search for the grain boundary which is observable through a trial-and-error process. In cases like when the grain boundary is curved to the thickness direction, the observation will be difficult as lattice images from two crystals that overlap in the grain boundary. In addition, if only the diffuse scattering electrons are taken in the objective aperture and observed by dark field imaging, only the amorphous phase can be observed as a white contrast, and it is convenient when the distribution is desired to be known. Figure 12 shows an HREM image of grain boundary in silicon nitride and it is recognized that the amorphous phase with a thickness of about 1 nm, which is not observable at low magnification, has been clearly observed. This photograph is observed under almost the same conditions as in Figure 11. The hightemperature strength of silicon nitride is considered to be dependent on the characteristics of the amorphous phase which exists at the grain boundary. Therefore, the information on the thickness and distribution of the phase becomes important; thus, HREM can be a useful approach. In addition, as for the evaluation of the composition and chemical bonding state in the grain boundary of amorphous phase, it can be possible to use EDS together with EELS by using a nano-probe as explained later. In order for further quantitative analysis, such as the identification of the location of an atom in the grain boundary and interface, comparison with computer simulation will be needed. The structure model of grain boundary (super cell) needs to be considered for the simulation. In constructing a grain-boundary model, various methods can be considered and the simplest way is the rigid model. This is a model to simply connect two

FIGURE 12 HREM image of Si3N4 grain boundary.

14

crystals without consideration of the atomic structure relaxation in the grain boundary. This is a simple and easy method that enables the construction of a structure which can be assumed from the high-resolution image and compares it with a simulation image. However, in general, with the rigid model, the structure cannot be uniquely determined and several assumed structure models need to be constructed. In addition, as it does not take in the atomic structure relaxation of the grain boundary, it cannot provide a quantitative analysis of grain-boundary atomic structure. For more quantitative analyses, the general method is to construct an optimal structure of the grain boundary or interface including structure relaxation by using the first principles calculation, static lattice calculation, and molecular dynamics calculation, and to determine the structure by comparing the simulation image with the observed image [33e35]. Figure 13 shows the analysis example of S9 grain boundary of YSZ (yttrium-stabilized zirconia) [36]. This specimen is a twin crystal made by adjusting two single crystals in a certain orientation, and by using thermal diffusion bonding at a high temperature. This twin crystal is made so that the common rotation axis is [110] and the grainboundary plane is {221} with S9 symmetry. Figure 13(a) shows the HREM image of the grain boundary observed by an ultra-high voltage electron microscope (accelerating voltage 1250 kV), and the black contrast directly corresponds to the atomic position. As shown in the figure, at the grain boundary, two kinds of asymmetric structure units are formed periodically. HREM image simulation for the most stable structure of the same grain boundary is shown in Figure 13(b), and the most stable structure model obtained by theoretical calculation is shown in Figure 13(c), respectively. It can be seen that, also in the theoretical model, an asymmetrical unit periodically appears and this result is consistent with the experimental image. At the grain boundary, cation sites of a different coordination number are formed as shown with an arrow in Figure 13(c), and these

FIGURE 13 Analysis example of S9 grain boundary of YSZ. (a): HREM image of the grain boundary observed by an ultra-high voltage electron microscope, (b): Simulation image by static lattice calculation, and (c): The most stable structure model obtained by theoretical calculation [36].

Handbook of Advanced Ceramics

sites would be an important influence in determining the characteristic of this grain boundary.

6.2. Application of Convergent Beam Electron Diffraction Electron diffraction has been also used to identify various crystal phases. The diffraction method from a narrow area using the nano-probe is used for various materials in general. As a method of the electron diffraction, CBED has been well known, and it is effectively used in determining the space group of the crystal phase, but it is also used in measuring lattice distortion of a narrow area. In other words, as CBED can measure the lattice constant in high accuracy from the diffraction line in higher-order Laue zone (HOLZ), it is useful in measuring small distortion of a local area such as the grain boundary and interface, etc. As the functional property of material is greatly influenced by not only the atomic structure of grain boundary, but also the lattice distortion due to the atomic structure of the grain boundary, CBED with a high spatial resolution attracts attention. Figure 14 shows the analysis results of S21 grain boundary of alumina by using CBED. In the figure, (a) shows an HREM image of the grain boundary, (b) shows a HOLZ line which is formed in the CBED pattern, and (c) shows the strain distribution in the vicinity of the grain boundary which was measured by this technique [37]. (b) shows the relation of the HOLZ line which is used in measuring the lattice constant of a-plane(ð2110Þ plane), and the incident direction of the electron beam is [0001] direction. As shown in the figure, by measuring the distance d and e, which connects the intersecting point of the HOLZ line, an estimation of the lattice constant of a-plane is possible. In this case, a probe diameter of about 0.7 nm is used, but calibration of the accelerating voltage of the electron microscope is performed by using a bulk sapphire single crystal, and its lattice constant is used for calibration in advance as the standard value. Also, the measurement of

Chapter | 1.1.1

(a)

The Latest Analytical Electron Microscope and its Application to Ceramics

15

(b)

FIGURE 15 HAADF-STEM image of Y-doped alumina grain boundary. For color version of this figure, the reader is referred to the online version of this book [39].

(c)

FIGURE 14 Analysis result of S21 grain boundary of alumina by CBED. (a): HREM image of the grain boundary, (b): HOLZ line which is formed in the CBED pattern, and (c): Distortion distribution in the vicinity of the grain boundary. For color version of this figure, the reader is referred to the online version of this book [37].

alumina polycrystal, and prevent the grain boundary sliding at high temperatures, but there are many unknown issues regarding the intrinsic mechanism. Here, the grainboundary structure of the Y-doped alumina was observed using the STEM and the result of clarification of the grainboundary strengthening mechanism by Y is shown. Figure 15 shows the HAADF-STEM image of the Y-doped alumina grain boundary [39]. The grain boundary shown here is so called S31 which can be classified in the general grain boundary [40]. As shown from this figure, in the Y-doped grain boundary, it is recognized that a very strong contrast has been observed at the particular atomic site along the grain boundary. This corresponds to the Y atom position segregated in the grain-boundary cores, and it shows that the Y atoms segregate selectively and regularly only at the center of the seven-membered ring structure in the grain-boundary cores. Figure 16 shows an electron density map of the Y-doped grain boundary which was obtained by the first principles calculation based on these observation results. The bonds without segregation have

the distance between intersecting points is performed on the image which was magnified by 20 times of the negative photographed with a 1.2-m camera length. The relation between the lattice constant of a-plane and the distance from the grain boundary, which were measured as such, is shown in (c). It is revealed that the lattice constant greatly changes in the vicinity of the grain boundary and that it reaches about 0.07% lattice distortion at the maximum.

6.3. Application of HAADF-STEM 6.3.1. The Grain Boundary of Rare Earth Element-Doped Alumina Ceramics It is known that the high-temperature strength of alumina ceramics improves drastically by doping a trace of rare earth element such as Y and Lu, etc. [38]. These doped trace elements are assumed to segregate in the grain boundary of

FIGURE 16 Charge density map for Y-doped alumina grain boundary obtained by the first principles calculation. For color version of this figure, the reader is referred to the online version of this book.

16

Handbook of Advanced Ceramics

mainly ionic bonding. However, it is recognized that the electron density distribution around Y atoms which segregated in the center of the seven-membered ring structure caused an orientation and covalent-like bonds have been formed [39]. It has been clear that rare earth elements have the effect to greatly change the surrounding chemical bonds and improve the grain-boundary strength by segregating in the grain boundary. That is, we can bring to a close that the intrinsic effect of rare earth elements doped in alumina ceramics is attributed to the local change of a chemical bonding state in the grain boundary.

6.3.2. STEM-EELS Observation of Superlattice Thin Film It is known that superlattice thin film provides a new function which is not available with bulk material. For example, the thermoelectric characteristic (performance index ZT) of strontium titanate (SrTiO3) is 0.08 or so at room temperature and it is far from ZT > 1, which is supposed to be the index for practical application. However, with the superlattice thin film where 1-unit cell of Nb-doped SrTiO3 layer is inserted (made) in the SrTiO3 without dopant, ZT goes up to 2.4, the thermoelectric characteristic which is equivalent to the ones of BieTe system and PbeTe system which are already in practical use, can be obtained [41]. In order to efficiently develop such a superlattice thin film, it is necessary to clarify the mechanism where this function appears, by combining the analysis method of a high spatial resolution with the theoretical calculation. Here, we will introduce the atomic and electronic state analysis of SrTiO3/ Nb-doped SrTiO3 /SrTiO3 superlattice, by combining STEM with EELS and first principles calculations. Figure 17(a) shows the HAADF-STEM image of SrTiO3/Nb-doped SrTiO3/SrTiO3 superlattice thin film which was fabricated by the pulsed laser deposition (PLD) method [41]. As the intensity of the HAADF-STEM image is proportional to about the square of Z, compared with Ti (Z ¼ 22) column, Sr (Z ¼ 38) column can be observed more brightly. It is known that Nb-doped SrTiO3 layer inserted by every 24-unit cell appears as a stripe-shaped contrast. Figure 17(b) and (c) show the high-resolution image around the Nb-doped SrTiO3 layer and the line profile of image intensity of the Sr atomic row and Ti atomic row in the same area. It is recognized that the image intensity does not change in Sr atomic row: however, the image intensity in Ti atomic row became stronger at the Nb-doped SrTiO3 layer. Taking into consideration that the atomic number of Nb is 41, it is considered that Nb exists in the Ti site. On the other hand, since the atomic numbers of Nb and Sr are close, whether Nb exists in the Sr site or not cannot be judged only from the HAADF-STEM image. In order to clarify that, the solution energy of Nb was estimated by

FIGURE 17 HAADF-STEM image (a) of SrTiO3/Nb-doped SrTiO3/ SrTiO3 superlattice thin film, (b): magnified image of region (a) and (c) line profile of image intensity of Ti and Sr layers [41].

using the first principles PAW (projector augmented-wave) method. It has been proven that the solution energy of Nb to Sr sites is higher by 7.6 eV compared with the one to Ti sites. This shows that Nb does not dissolve in the Sr sites and it exists in the Ti sites. Figure 18 shows the spectra of Ti-L2,3 ELNES obtained from the SrTiO3 area and Nb-doped SrTiO3 layer in SrTiO3/ Nb-doped SrTiO3/SrTiO3 superlattice. It is recognized that in the SrTiO3 layer, four peaks clearly appear; on the other hand, in the Nb-doped SrTiO3 layer, the peaks become broad. In order to interpret this change, the first principles relativistic multi-electron calculation was performed. As a result, it became clear that the change of the experimental spectrum is due to the transition from Ti4þ to Ti3þ, which accompanies Nb doping. This shows that in the Nb-doped SrTiO3 layer, the surplus electrons were introduced and the two-dimensional electron gas is being formed [41]. So far, it has been clear that the high thermoelectric characteristics of SrTiO3/Nb-doped SrTiO3/SrTiO3 superlattice are attributed to this two-dimensional electron gas.

6.4. Application of ABF-STEM The clarification of a lithium-ion battery mechanism and improvement of the reliability are related to the behavior of the lithium ion, and it is considered that if direct visualization of a lithium atom is possible, it will be a great breakthrough in the research and development. In the past,

Chapter | 1.1.1

The Latest Analytical Electron Microscope and its Application to Ceramics

FIGURE 18 (a): Ti-L2,3 ELNES spectrum obtained from SrTiO3 layer and Nb-doped SrTiO3 layer in SrTiO3/Nb-doped SrTiO3 superlattice and (b): The spectrum obtained by the first principles relativistic multi-electron calculation.

an attempt to observe the lithium atom by using a highresolution electron microscopy has been made. However, only indirect observations such as the method to reconstruct HREM image [42], and the other method to modify the lithium site with other element [43], etc., have been reported,

17

and it has been considered that the direct observation of a lithium atom which is a light element is very difficult in terms of technique. Here, it is demonstrated that the observation of a lithium atom with the atomic number 3 became possible by using ABF-STEM [29,30]. A series of observations of lithium ion have been carried out for materials such as LiMn2O4, LiCoO2, and LiFePO4, etc. so far, and direct observations of the lithium atomic row have been done successfully [44,45]. Here, the observation example of the lithium-ion row in LiCoO2 crystal with a layer structure is introduced. Figure 19(a) shows the crystal structure model of LiCoO2. It is understood from the figure that this crystal has a layer structure, and that along ½1120Š direction, each atom of Li, O, Co is allocated side by side. Figure 19(b) is the ABF-STEM image of LiCoO2 crystal observed from ½1120Š direction, and it is recognized that in addition to the Co atomic row, which has a strong black contrast, the Li atomic row can be clearly observed between oxygen atomic layers. This is the image obtained at the 8e25-mrad detection angle, and the observation of atomic row has been made for a light element and a heavy element at the same time. This observation is only possible by using STEM with a resolution of 0.1 nm or lower by the spherical aberration correction and by setting the detection angle where the lithium atom can be observed by using theoretical calculation. By further applying this method to practical materials, the diffusion mechanism of the lithium ion and the behavior of the lithium ion during charging and discharging have became clear and it is expected that future research and development of the lithium-ion battery will be achieved. It has been recently reported that hydrogen atoms can be observed by ABFSTEM imaging technique [46].

6.5. Application of EDS In a nano-probe analysis of the local area such as the grain boundary and interface, etc., it is important to fix the probe at the interface position for sure. In that case, as measurement FIGURE 19 (a): Crystal structure model of LiCoO2 and (b): ABF-STEM image of LiCoO2 observed from ½1120Š direction. For color version of this figure, the reader is referred to the online version of this book.

18

FIGURE 20 (a) to (c): Schematic diagram showing position relation between nano-probe and the grain boundary or interface during analysis and (d): Position relation between nano-probe and the grain boundary or interface during an EDS analysis.

by the probe of sub-nanometer diameter or lower is required, FEG with high brightness is normally used. Figure 20(a)e(c) show schematically the position relation of the nano-probe and the grain boundary or interface during analysis. The nano-probe is illuminated to the grain boundary or interface as the conical beam which has a certain illumination angle. Therefore, unless a probe is precisely focused within the area to be analyzed (position of disk of least confusion), it is not a correct analysis including the quantitative analysis. In case the focus position is correct, the focusing will be almost in the center of the specimen thickness (Figure 20(a)), but if the focus is misaligned (Figure 20(b)), electron beams are transmitting the specimen while spreading depends on the illumination angle; thus, the actual spatial resolution turned out to be deteriorated. On the one hand, as shown in Figure 20(c), when the interface is tilted against the probe, the interface area that occupies in the probe reduces, and then the detection sensitivity of the segregation element, etc. declines

FIGURE 21 Analysis example of the WC/Co interface of 0.5 wt% VC-doped WCeCo super hard alloy. (a): HREM image of WC/Co interface and (b): EDS spectra obtained at the position of A to E in (a) by using a nano-probe of about 0.5 nm in diameter.

Handbook of Advanced Ceramics

as well as its quantification. Figure 20(d) shows the position relation of the interface and nano-probe during an EDS analysis. In the case of an EDS analysis, as that the EDS detector is positioned on the extension of the interface will be an advantage in terms of detection efficiency (plane hkl in Figure 20(d)), it is necessary to set the tilt direction of the specimen to be an EDS detector side at least. This not only improves the quantification of the EDS analysis, but also increases the detection efficiency of X-ray; therefore, it also leads to the reduction of the specimen drift and specimen damage, etc. In addition, the position accuracy of the probe and the influence of the stray X-ray need to be investigated in advance. For example, the influence of the spurious X-ray and actual probe diameter can be confirmed from the EDS spectrum when the probe is fixed in a space of about 1 nm away from the specimen edge. Moreover, with that state, when it is changed to the diffraction mode, if we check whether a diffraction pattern is observed or not, then, we will be able to know the probe position accuracy due to the stray magnetic field of the electromagnetic lens system. Also, an attention is needed, that in order to perform a precise quantitative analysis, the analysis area also needs to be a thin area where a thin film approximation can be applied. Figure 21 is an analysis example of the WC/Co interface of 0.5 wt% VC-doped WCeCo super hard alloy [47]. WCeCo super hard alloy is widely used in cutting tool materials, etc., and it has a characteristic composition where the hard WC particle is bonded by Co phase, which is a metal phase. Figure 21(a) is the HREM image of WC/Co interface, Figure 21(b) is the EDS spectra obtained at the position of A to E in the HREM image, and these are the results of an analysis by using a nano-probe of about

Chapter | 1.1.1

The Latest Analytical Electron Microscope and its Application to Ceramics

19

0.5 nm in diameter. Judging from a comparison of B to D, clear V-Ka peak is detected at the WC/Co interface (C). However, in the WC and Co grains (B and D), which are away by only 2 nm from the interface, V is not detected; therefore, it is shown that the doped V segregates only near WC/Co interface. This segregation volume differs by a facet of the WC grain, that is, more segregation of V on the (0001) facet (C) is recognized compared with the other facet (E). On the other hand, at the intersection points of these two facets, micro-precipitates as shown by arrows in the HREM image are formed, and from the EDS analysis as well as the HREM observation, it is clear that these precipitates are (W,V)2C. From a series of these analysis results, the grain growth inhibitor mechanism of the WC/ Co super hard alloy by VC has been clarified and it can be understood that EDS with a nano-probe is useful for practical materials. On the other hand, by scanning the nano-probe in STEM, elemental mapping by EDS is possible. Figure 22 shows (a) bright field-STEM image of a trace amount of Lu2O3-doped Al2O3 polycrystal, and (b) Lu-Ka-STEM mapping image (probe diameter of about 0.5 nm). It is clearly known that the doped Lu ions segregate along the grain boundary. It has been reported that the creep properties of the Al2O3 polycrystal greatly improve by doping the trace amount of Lu2O3 [38]. However, it has been clarified by such a series of analysis that the segregations of Lu in the grain boundary is the reason for that. The STEM-EDS analysis has the feature that elemental distribution can be obtained with a wide field of view and at a high resolution, different from the above-mentioned point analysis, and its application is expected a lot in the future.

6.6. Application of EELS EELS is useful for the chemical bonding state analysis of a local area. In other words, with the spectrum in EELS, ELNES contains information such as a local atomic arrangement and chemical bonding. So far, for interpretation of ELNES, comparison with a spectrum from reference substance, the so-called fingerprint matching has been done, but along with the complication of the substance, a more quantitative method is needed. In the past, calculation of ELNES has been done by various methods such as the cluster method, etc. At present, calculation by using OLCAO (orthogonalized linear combinations of atomic orbitals) method, a kind of the first principles band calculation methods, has been attracting attention as it can reproduce an experimental spectrum of ELNES well [48]. ELNES is equivalent of transition energy of an electron from the inner-shell orbit to the unoccupied orbit, and during that transition process an inner-shell hole is formed. The calculation is made by taking in the effect

FIGURE 22 (a): Bright field-STEM image of a trace amount of Lu2O3doped Al2O3 polycrystal and (b) X-ray mapping image of LueKa.

of inner-shell hole and, by using the OLCAO method, an effective calculation is possible. Figure 23 shows the experimentally obtained O-K ELNES in the vicinity of the ZnO grain boundary, and the calculated spectrum by the OLCAO method [49]. It is widely known that the grain boundary of ZnO shows a nonlinear currentevoltage characteristic and it is assumed that the origin of this characteristic is due to the point defects segregated in the grain boundary. Figure 23 (upper) is the O-K ELNES measured in the vicinity of the high angle grain boundary of ZnO (area about 20 nm away from the grain boundary), and this spectrum is characterized by three main peaks as shown by A, B, and C. On the other hand, the lower spectrum is the theoretical O-K

20

FIGURE 23 Measured spectrum (upper) and theoretical spectrum (lower) of OK-ELNES in the vicinity of ZnO grain boundary.

ELNES from the O atom in a perfect crystal of ZnO. It is recognized that this theoretical O-K ELNES is very similar to the experimental spectrum in terms of both shape and position of peaks. This result suggests that remarkable segregation of point defects, which has an influence on the currentevoltage characteristic in the vicinity of ZnO grain boundary, does not occur. As such, for the quantitative evaluation of ELNES, interpretation by a theoretical calculation is useful.

7. CONCLUSION As FEG and Cs-corrector have been installed with the most advanced AEM, not only image observation at a resolution of about 0.1 nm but also formation of the probe diameter less than 0.1 nm have been possible. As a result, not only direct observation of the atomic column from a light element to a heavy element in the thin crystal by TEM and STEM, but also the atomic-level elemental mapping by EDS and EELS, and ELNES, have become possible. In this paper, we have explained the general overview of the most advanced AEM, as well as shown its application research of ceramic materials concretely. It would be wonderful if this paper is able to serve as some kind of help in conducting research and development of ceramic materials.

REFERENCES [1] Knoll M, Ruska E. Das Elektronenmikroskop. Z. Physik 1932;78:318e39. [2] Heren JJ, Goldstein JI, Joy DC. Introduction to analytical electron microscopy. New York: Plenum Press; 1979. [3] Sindo D, Oikawa T. Analytical electron microscopy for materials science. Tokyo: Springer-Verlag; 2002. [4] Isabell T, Brink J, Kawasaki M, Armbruster B, Ishikawa I, Okunishi E, et al. Development of a 200 kV atomic resolution analytical electron microscope. Microscopy Today May 2009:8e11. [5] Broers AN. Electron gun using long-life lanthanum hexaboride cathode. J Appl Phys 1967;38:1991e2.

Handbook of Advanced Ceramics

[6] Crewe AV, Eggenberger DN, Wall. J, Welter MJ. Electron gun using a field emission source. Rev Sci Instrum 1968;39:576e83. [7] Swanson LW, Martin NA. Field emission cathode stability studies: zirconium/tungsten thermal field cathode. J Appl Phys 1975; 46:2029e50. [8] Yanaka T, Shirota K, Yonezawa A, Arai Y. Capability of high resolution objective lens with miniatuarized lower pole piece top. Proc. 8th ICEM, Canberra 1974;1:128e9. ¨ ber die Auflo¨sungsgrenzen des Durchstrahlungs-Elec[9] Ruska E. U tronen mikroskops. Optik 1965;45:319e48. ¨ ber einige Fehler von Electronenlinsen. Z Phys [10] Scherzer O. U 1936;101:593e603. [11] Scherzer O. Spha¨rische und chromatische Korrektur von Electronen-Linsen. Optik 1947;2:114e32. [12] Rose H. Correction of aperture aberrations in magnetic systems with threefold symmetry. Nucl Instr Methods 1981;187:187e99. [13] Haider M, Braunshausen G, Schwan E. Correction of the spherical aberration of a 200 kV TEM by means of a hexapole-corrector. Optik 1995;99:167e79. [14] Sawada H, Tanishiro Y, Ohashi N, Tomita T, Hosokawa E, Kaneyama T, et al. STEM imaging of 47-pmseparated atomic columns by a spherical aberration-corrected electro microscope with a 300 kV cold field emission gun. J Electron Microsc 2009;58:357e61. [15] Hirsch PB, Howie A, Micholson RB, Pashley DW, Whelan MJ. Electron microscopy of thin crystals. London: Butterworths; 1965. [16] Sindo D, Hiraga K. High-resolution electron microscopy for materials science. Tokyo: Springer-Verlag; 1998. [17] Scherzer O. The theoretical resolution limit of the electron microscope. J Appl Phys 1949;20:20e9. [18] Hanben KJ, Trepte L. Der Einflub von Strom- und Spannungsschwankungen, sowie der Energiebreite der Strahlelektronen auf Kontrastu¨bertragung und Auflo¨sung des Electronenmikroskops. Optik 1971;32:519e38. [19] Frank J. The envelope of electron microscopic transfer functions for partially coherent illumination. Optik 1973;38:519e36. [20] Cowley JM, Moodie AF. The scattering of electrons by atoms and crystals. A new theoretical approach. Acta Cryst 1957;10:609e19. [21] Tanaka M. Convergent-beam electron diffraction. Acta Cryst 1994;A50:261e86. [22] Liu J, Cowley JM. High-resolution scanning transmission electron microscopy. Ultramicroscopy 1993;52:335e46. [23] Zeitler E, Thmson MGR. Scanning transmission electron microscopy. Optik 1970;31:258e80. [24] Zeitler E, Thmson MGR. Scanning transmission electron microscopy.II. Optik 1970;31:359e66. [25] Crewe AV, Wall J, Langmore J. Visibilty of single atoms. Science 1970;168:1338e40. [26] Pennycook SJ, Jesson DE. High-resolution incoherent imaging of crystals. Phys Rev Lett 1990;64:938e41. [27] Howie A. Image contrast and localized signal selection techniques. J Microscopy 1979;117:11e23. [28] Okunishi E, Ishikawa I, Sawada F, Hosokawa E, Hori M, Kondo Y. Visualization of light elements at ultrahigh resolution by stem annular bright field microscopy. Microsc Microanal 2009;15:164e5. [29] Findlay SD, Shibata N, Sawada H, Okunishi E, Kondo Y, Yamamoto T, et al. Robust atomic resolution imaging of light elements using scanning transmission electron microscopy. Appl Phys Lett 2009;95. 191913e1e3.

Chapter | 1.1.1

The Latest Analytical Electron Microscope and its Application to Ceramics

[30] Findlay SD, Shibata N, Sawada H, Okunishi E, Kondo Y, Ikuhara Y. Dynamics of annular bright field imaging in scanning transmission electron microscopy. Ultramicroscopy 2010;110:903e23. [31] Krivanek OL, Ahn CC, Keeney RB. Parallel detection electron spectrometer using quadrupole lenses. Ultramicroscopy 1987;22: 103e16. [32] Horiuchi S, Ikuhara Y, Hojo K. In: Japanese Society of Surface Chemistry, editor. Transmission electron microscopy. Tokyo (in Japanese): Maruzen; 1999. [33] Ikuhara Y, Yamamoto T. Denshikenbikyo 2002;37:51e6 (in Japanese). [34] Ikuhara Y, Yamamoto T. Denshikenbikyo 2002;37:129e33 (in Japanese). [35] Kohyama M. Computational studies of grain boundaries in covalent materials. Mode Simul Mater Sci Eng 2002;10:R31. [36] Shibata N, Oba F, Yamamoto T, Ikuhara Y. Structure, energy and solute segregation behaviour of [100] symmetric tilt grain boundaries in yttria-stabilized cubic zirconia. Philos Mag 2004;84:2381e415. [37] Saito T, Hirayama T, Yamamoto T, Ikuhara Y. Lattice strain and dislocations in polished surfaces on sapphire. J Am Ceram Soc 2005;88(8):2277e85. [38] Yoshida H, Ikuhara Y, Sakuma T. High-temperature creep resistance in lanthanoid ion-doped polycrystalline AL203. Phil Mag Lett 1999;79:249e55. [39] Buban JP, Matsunaga K, Chen J, Shibata N, Ching WY, Yamamoto T, Ikuhara Y. Grain boundary strengthening in alumina by rare earth impurities. Science 2006;311(5758):212e5. [40] Ikuhara Y. Physics of ceramics. in Japanese: Nikkan Kogyou Shinbunsha; 1999.

[41]

[42]

[43]

[44]

[45]

[46]

[47]

[48] [49]

21

Ohta H, Kim S, Mune Y, Mizokuchi T, Nomura K, Ohta S, et al. Giant thermoelectric Seebeck coefficient of two-dimensional electron gas in SrTiO3. Nat Mater 2007;6(2):129e34. Shao-Horn Y, Croguennec L, Delmas C, Nelson EC, O’Keefe MA. Atomic resolution of lithium ions in LiCoO2. Nat Mater 2003;2(7):464e7. Chung SY, Choi SY, Yamamoto T, IKuhara Y. Atom-scale visualization of antisite defects in LiFePO4. Phys Rev Let 2008;100(12): 125502e5. Huang R, Ikuhara Y, Mizokuchi T, Findlay SD, Kuwabara A, Fisher CAJ, et al. Oxygen-vacancy ordering at surfaces of Lithium Maganese(III,IV) oxide spinel nanoparticles. Angew Chem Int Ed 2011;50(13):3053e7. Huang R, Hitosugi T, Findlay SD, Fisher CAJ, Ikuhara YJ, Moriwake H, Oki H, Ikuhara Y. Real-time direct observation of Li in LiCoO2 cathode material. Appl. Phys. Lett 2011;98:051913. Findlay SD, Saito T, Shibata N, Sato Y, Matsuda J, Asano K, Akiba E, Hirayama T, Ikuhara Y. Direct imaging of hydrogen within a crystalline environment. Appl. Phys. Express 2010;3: 116603. Yamamoto T, Ikuhara Y, Sakuma T. High resolution transmission electron microscopy study in VC-doped WC-Co compound. Sci Tech Adv Mater 2000;1:97e104. Tanaka I, Mizoguchi T, Yoshiya M, Iwata T, Ogasawara K, Adachi H. Denshikenbikyou 2000;35:221e9 (in Japanese). Sato Y, Mizoguchi T, Oba F, Yodogawa M, Yamamoto T, IKuhara Y. Identification of native defects around grain boundary in Pr-doped ZnO bicrystal using electron energy loss spectroscopy and first-principles calculation. Appl Phys Lett 2004;84(26):5311e3.

Chapter 2.1

Advanced Carbon Materials Michio Inagaki Professor Emeritus of Hokkaido University, Japan

Chapter Outline 1. 2. 3. 4.

Carbon Materials Chemical Bonding and Carbon Families Structure Carbonization and Graphitization

25 25 27 31

1. CARBON MATERIALS Carbon materials have played important roles for human beings: charcoals as a heat source and as an adsorbent since prehistorical age, flaky natural graphite powder as pencil lead, soot in black ink for the development of communication techniques, graphite electrodes for steel production, carbon blacks for reinforcing tires in order to develop the motorization, electric conductive carbon rods and carbon blacks for supporting the development of primary batteries, a compound of graphite with fluorine (graphite fluoride) for improving the performance of primary batteries, thin graphite flakes in membrane switches for making computers and control panels thinner and lighter, etc. In relation to global-warming problems, nuclear reactors are recognized to be important, which are constructed from high-density isotropic graphite blocks, having different roles as reflectors, moderators, etc. For the storage and efficient usage of sustainable energies, which are usually unstable, lithium-ion rechargeable batteries and electric double-layer capacitors are essential devices, in which carbon materials are used as electrodes and govern their performance: in the former, lithium intercalation/deintercalation into the galleries of the anode graphite is a fundamental electrochemical reaction and, in the latter, physical adsorption of electrolyte ions onto the surfaces of the electrodes of porous carbon is a fundamental electrochemical reaction. Nanocarbons developed recently, which are represented by carbon nanotubes, fullerenes, and graphene, are promoting the development of nanotechnology in various fields of science and engineering. Carbon materials are predominantly composed of carbon atoms, only one kind of element, but they have largely diverse structures and properties. Diamond has a threedimensional structure and graphite has a two-dimensional

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00002-2 Copyright Ó 2013 Elsevier Inc. All rights reserved.

5. Various Carbon Materials 6. Importance of Textures in Carbon Materials References

36 51 56

nature, whereas carbon nanotubes are one-dimensional and buckminsterfullerene C60 is quasi-zero-dimensional. Fullerenes behave as molecules, although other carbon materials do not. Graphite is an electric conductor and its conductivity is strongly enhanced by AsF5 intercalation, higher than metallic copper, whereas diamond is completely insulating. Diamond which is the hardest material is used for cutting tools, and graphite is so soft that it can be used as a lubricant. This chapter provides fundamental information on carbon materials, before going into the details on various advanced carbon materials, nanocarbons including carbon nanotubes, high-density isotropic carbons, carbon fibers and their composites, and carbon materials for energy devices.

2. CHEMICAL BONDING AND CARBON FAMILIES The carbon atom has four electrons in the outermost 2s and 2p orbitals, namely 2s22p2. When it forms chemical bonds with other atoms, it can have three different hybridizations, sp, sp2, and sp3. The hybridized sp orbital is composed of two s electrons, associated with two p electrons, the sp2 of three s electrons with one p electron, and the sp3 of four s electrons. It is well known that various combinations of these hybrid orbitals produce a variety of organic hydrocarbons, from aliphatic to aromatic ones. This situation is illustrated in Figure 1, by showing some hydrocarbons [54,59]. The carbonecarbon bonds (CeC bonds) using sp2 hybrid orbitals give a series of aromatic hydrocarbons, such as benzene, anthracene, phenanthrene, etc. The CeC bonds using sp3 hybrid orbitals give various aliphatic hydrocarbons, such as methane, ethane, propane, adamantane, etc. Mixing the

25

26

Handbook of Advanced Ceramics

FIGURE 1 Extension from organic hydrocarbons to carbon families.

different types of hybridizations, sp, sp2, and sp3, in a molecule expands the variety of hydrocarbons. Inorganic carbon materials, such as diamond, graphite, fullerenes and carbynes, are considered as an extension of these organic molecules to giant ones, as shown in Figure 1. A simple repetition of CeC bonds with sp2 hybrid orbital gives planar hexagons of carbon atoms, as

benzene, anthracene, ovalene, etc. The extension of this series of aromatic hydrocarbons reaches graphite, in which the s bonds using the sp2 hybrid orbitals are reinforced by the delocalized p-electrons. The giant planes tend to stack on each other due to the interactions between the pelectron clouds of the stacked layers, giving rise to weak van der Waals-like interaction, i.e., graphite. The

Chapter | 2.1

Advanced Carbon Materials

27

somewhat curved corannulene molecule is formed by associating a pentagon of sp2-hybridized carbon atoms with 5 hexagons. By the polymerization of these corannulene molecules to form a closed shell, various sizes of carbon clusters are formed and the resultant carbon materials are called fullerenes, where carbon hexagons have to be located between pentagons. The smallest closed shell is the buckminsterfullerene C60, consisting of 12 pentagons and 20 hexagons of carbon atoms. As shown schematically in Figure 1, the infinite repetition of CeC bonds using sp3 hybrid orbitals results in a three-dimensional network of carbon atoms, which is known to be diamond. The carbon atoms in the organic compound named as adamantane are exactly in the same atomic positions as those in diamond crystal. The infinite repetition of CeC bonds with sp hybrid orbital gives the carbon materials called carbynes, in which the carbon atoms form linear chains either with double bonds (one s-bond with one p-bond) or with the alternative repetition of single (one s-bond) and triple bonds (one s-bond with two pbonds), the former being called cumulene type and the latter polyyne type, respectively. Based on the extension by the repetition of three kinds of CeC bonds to infinite molecules, we may define carbon families, represented by the perfect structure in each family, that is, diamond, graphite, fullerene, and carbyne [54]. However, each family is known to have characteristic

Hexagonal graphite

Space group: P63/mmc z number: 4 Equivalent points: 0,0,0; 2/3,1/3,0; 0,0,1/2; 1/3,2/3,1/2 Lattice parameters: a0=0.2462 nm, c0=0.6708 nm

diversity in structure, as will be explained briefly in the following sections.

3. STRUCTURE 3.1. Graphite Family The carbon family having a planar sp2 bonding is represented by graphite, where the layers of carbon hexagons are stacked in parallel; the stacking regularity of ABAB$$$ sequence belongs to the hexagonal crystallographic system (hexagonal graphite) and that of ABCABC$$$ to the rhombohedral system (rhombohedral graphite). The unit cells of these two allotropes are shown in Figure 2, together with their crystallographic data: space group, equivalent points, and lattice parameters. For the rhombohedral graphite, both rhombohedral and hexagonal cells are shown together in the figure. The hexagonal graphite is a thermodynamically stable phase. The rhombohedral structure of graphite was shown to form by applying shearing forces, for example, by grinding, which seems to be formed mainly due to the introduction of stacking faults in the crystallites having the hexagonal stacking regularity. In addition to these regular stacking modes, the parallel stacking of the layers without any regularity occurs mostly in the carbon materials prepared at low temperatures, below 1300  C. In this case, the layers of hexagons are usually small

Rhombohedral graphite

a) Rhombohedral system (thick lines) Space group: R3m z number: 2 Equivalent points: 1/6,1/6,1/6; 5/6,5/6,5/6 Lattice parameters: a0=0.3635 nm, α =39.49 o

b) Hexagonal system (double lines) z number: 6 Equivalent points: 0,0,0; 2/3,1/3,0; 1/3,2/3,1/3; 2/3,1/3,1/3; 1,0,2/3; 1/3,2/3,2/3 Lattice parameters: a0=0.2462 nm, c0=1.0062 nm

FIGURE 2 Crystal structures and lattice parameters of two graphite allotropes.

28

Handbook of Advanced Ceramics

FIGURE 3 X-ray diffraction patterns for two carbon materials.

in size and only few layers are stacked in parallel. This random stacking of layers is called turbostratic structure. Figure 3a shows the X-ray powder diffraction pattern (powder pattern) of natural graphite with a high crystallinity. The diffraction lines of graphite have to be classified into three groups, the lines with 00l, hk0, and hkl indices, mainly because of the strong anisotropy in structure. The lines with 00l indices are due to the reflection from basal planes (hexagonal carbon layers, graphite basal planes), where only the even l indices are allowed because of the extinction rule due to the parallel stacking of the layers. The lines with hk0 indices are due to the reflection from the crystallographic planes perpendicular to the basal planes and the lines with hkl indices come from the planes declining to the basal planes, where the three-dimensional structure with regular stacking of layers has to be established. Therefore, the hkl lines are called three-dimensional lines. On the other hand, the powder pattern which is given by low-temperature-treated carbons is quite different

from well-crystallized graphite, because it consists of random stacking of small layers, as shown on a petroleum coke heat-treated up to 1000  C in Figure 3b. The diffraction pattern is characterized by very broad 00l lines due to the small number of stacked layers, and by unsymmetrical hk lines and no hkl lines because of the random turbostratic stacking of layers. The diffraction lines due to the planes perpendicular to the hexagonal carbon layers are missing l index because of the absence of regularity in the direction along the normal to the layers (along the c-axis); therefore, they are expressed as 10 and 11 in Figure 3b. When graphite is observed by scanning tunneling microscopy (STM), the hexagonal carbon layer looks like a triangular arrangement of spots (Figure 4a). Because of the interaction with the lower-lying atoms in the B layer, only a half of the carbon atoms are detected under STM. In the turbostratic structure, however, the interactions with the lower-lying atoms is much weaker, because of the

FIGURE 4 Scanning tunneling micrographs of two structures existed in carbon materials. (Courtesy of Prof. K. Oshida of Nagano National College of Technology, Japan)

Chapter | 2.1

Advanced Carbon Materials

irregular stacking, and consequently the spots form a hexagonal arrangement (Figure 4b) [21]. Since the turbostratic structure can be partly transformed into a regular stacking of layers with a heat treatment at high temperatures, a wide range of structural diversity may result in the graphite family. The graphitization degree is defined as the probability of occurrence of the regular graphitic AB stacking in turbostratic stacking. Carbon materials that belong to the graphite family can have various textures mainly because the basic structural units (BSUs), i.e., the stacked hexagonal carbon layers, are highly anisotropic. The texture at the nanometric scale (nanotexture) will be separately discussed in the Section 4. In 2010, Nobel Prize in physics was given to Profs A. K. Geim and K. S. Novoselov for their work on graphene. Graphene is defined by an isolated single layer of carbon hexagon, two-dimensional crystal with just one atom thickness, and is expected to have high mechanical strength, high electrical and thermal conductivities, etc. On graphene, many reviews have been published [29,62,129,136].

3.2. Fullerene Family Although based on sp2-type hybridization of carbon atoms, the bonding nature in fullerenes is slightly different from graphite in the fact that the layer is curved by the introduction of pentagons, as seen in corannulene (Figure 1). The repetition of these curved layers can result in various

29

closed shells with various sizes, from the smallest closed shell of C60 to giant fullerenes. The introduction of additional hexagons among pentagons in C60 leads to giant fullerenes by keeping closed shell morphology, as shown in the left-hand side of Figure 5. In this fullerene family, the variety of structure is mainly due to the number of carbon atoms building the closed shell. Most of fullerenes behave as molecules [18], i.e., they can be dissolved into an organic solvent, giving a characteristic color, which is exceptional in the carbon materials because all other carbon materials are not dissolved into any organic solvent. As shown in the center of Figure 5, there is another way to increase the number of hexagons between two groups of 6 pentagons, which results in a single-wall carbon nanotube. The carbon nanotube seems to locate in between the graphite and fullerene families, because it has the characteristics of graphite together with those of fullerenes. The closed ends of a nanotube consist of carbon pentagons with hexagons, which is the characteristic for the fullerene family, and the tube wall consists of carbon hexagons and can be considered to be rolled from an hexagonal carbon layer, which is the fundamental structural unit composed of the graphite family. A singlewall carbon nanotube is characterized by its diameter and chiral vector [18,132]. The illustration on the right-hand side of Figure 5 presents an example to roll up to form a single-wall carbon nanotube with a chirality of (5, 2). Fullerenes and nanotubes consisting of multi-layers are also formed.

FIGURE 5 Relations among fullerene, carbon nanotube, and graphite.

30

Handbook of Advanced Ceramics

FIGURE 6 Structure of diamond in two allotropes.

3.3. Diamond Family Diamond consists of sp3-hybridized carbons, where the purely covalent chemical bonds extend three-dimensionally. Diamond is very hard and electrically insulating owing to no p electrons. In order to construct a diamond crystal, a periodical and regular repetition of the CeC bond is required in a long range. For a couple of carbon atoms indicated as A and B in Figure 6a, the carbon atom A has to be connected with four carbon atoms, including B, to make a tetrahedron because of directional sp3 bonds, and the atom B also has to be surrounded by four carbon atoms, including A. Looking down these two tetrahedra centered at the A and B atoms along their connecting line, there are two possibilities in mutual relation between two basement triangles consisting of three carbon atoms. If these two basement triangles are rotated with respect to each other by 60 , as shown in Figure 6b, the resultant diamond crystal belongs to the cubic system (cubic diamond, Figure 6b’). If there is no rotation between these two basement triangles (Figure 6c), the structure belongs to the hexagonal system (hexagonal diamond, Figure 6c’). Most of the diamond crystals, either occurring in nature or being synthesized, are cubic. In the case where a long-range periodicity is not attained between interconnected tetrahedra (Figure 6a), in other words, in the case of a random rotation between the two basement triangles of tetrahedra, an amorphous structure is formed, which is called diamond-like carbon (DLC). Because of a random rotation between tetrahedra, some carbon atoms cannot form chemical bonds with

neighboring carbon atoms, most of which are supposed to be connected with hydrogen atoms for stabilization. DLC is usually obtained as a thin film, mainly because a random repetition of tetrahedra is difficult to keep on a long distance. DLC is as hard as the diamond crystal, because it is also based on sp3-hybridized atoms, and it contains a relatively large amount of hydrogen. A metastable crystalline phase of carbon atoms, consisting of a face-centered cubic unit cell with a parameter of 0.3563 nm, was found in either thin film or nanoparticles [68].

3.4. Carbyne Family Carbynes have been supposed to be formed by sphybridized carbon atoms bound linearly, where two pelectrons have to be involved, giving two possibilities, i.e., an alternative repetition of single and triple bonds (polyyne) and a simple repetition of double bonds (cumulene) (Figure 1) [87]. The detailed structure of carbynes is not yet clarified, but some structural models have been proposed [76]. A structural model is illustrated in Figure 7, where some numbers of sp-hybridized carbon atoms form chains, which are associated together by van der Waals interaction between p-electron clouds to make a layer, and then the layers are stacked. In the carbyne family, the variety of structures seems to be mainly due to the number of carbon atoms forming a linear chain, in other words, due to the layer thickness and the density of chains in a layer.

Chapter | 2.1

Advanced Carbon Materials

FIGURE 7 Structure model of carbyne.

3.5. Acceptance of Foreign Atoms in the Structure Each carbon family also shows various characteristic possibilities for accepting foreign species. The substitution of either boron or nitrogen atom for carbon atom is possible for all carbon families, even though the amount substituted is very limited [79,96]. Since this substitution is expected to modify the properties of carbon materials without noticeable structural changes, various modifications have been investigated on different types of carbon materials for various applications in order to reinforce their functions. Effects of boron doping on accelerating the development of graphitic stacking [22,37] and on improving the electrode performance in lithium ion rechargeable batteries [13] have been reported. Nitrogen-substituted diamond crystals occur rarely in nature. Recently, nitrogen-containing carbon materials were reported to have a high capacitance in electrochemical capacitors [61,77], though the state of the nitrogen atoms in the carbon matrix was not yet well understood. Nitrogen substitution in buckminsterfullerene, C59N (heterofullerene), was also reported [41]. Intercalation between the layers of graphite and carbynes is effective and useful for the modification of their properties. The intercalation of either iron or potassium is reported to stabilize the structure of carbynes [87]. Into the carbon materials belonging to the graphite family, a wide variety of foreign species (atoms, ions, and even molecules) have been intercalated to give interesting functionalities to the materials [26,49]. Either donors (such as lithium and

31

potassium) or acceptors (such as sulfuric acid, bromine, and ferric chloride) can be intercalated in the galleries between neighboring hexagonal carbon layers. The intercalation of sulfuric acid is widely used for the production of exfoliated graphite, which is converted to flexible graphite sheets for shielding and packing [56]. Intercalation/deintercalation of lithium ions is the fundamental electrochemical reaction for the discharge/charge of lithium-ion rechargeable batteries [14]. Graphite intercalation compounds with AsF5 and SbF5 were reported to have high electrical conductivity [102,158], higher than metallic copper, but they were never used in practice mainly because of instability of the compounds in air. The buckminsterfullerene C60 crystal can be doped by alkali metals (e.g., potassium), which occupy all the tetrahedral and octahedral interstices of the closest packing of fullerene clusters, giving rise to superconductive materials [148]. In the fullerene family, besides the substitution of carbon atoms in the shell and doping in the interstices between shells described above, there are two additional possibilities to accept foreign species: the insertion into the inner space of the closed shells [162] and the covalent bonding of organic radicals to the surface of the shell [166]. Rare earth elements have been inserted into fullerene cages (endohedral metallofullerenes). In the carbyne family, doping by intercalation into the space between the carbon chains in a layer is theoretically possible, but to date experimental results have not been reported.

4. CARBONIZATION AND GRAPHITIZATION 4.1. Pyrolysis and Carbonization Carbon materials are produced by the heat treatment of some organic polymers, i.e., carbon precursors, up to a temperature of 800e1500  C in an inert atmosphere (carbonization). The general scheme of carbonization process is shown in Figure 8 for a carbon precursor, such as a pitch, by indicating the main out-gassed components and changes in residues, although its details depend strongly on the carbon precursor used. At the beginning of precursor pyrolysis, aliphatic and then aromatic hydrocarbons with low molecular weights are released as gases, mainly because some of the CeC bonds are weaker than CeH bonds. Associated with this hydrocarbon release, cyclization and aromatization proceed in the residues and then polycondensation of aromatic molecules occurs. From around 600  C, mainly foreign atoms such as oxygen and hydrogen are released as CO2, CO, and CH4. In this stage, the residues are either liquid or solid phases depending on the starting precursors. Above 800  C, the main evolved gas is H2 because of polycondensation of aromatics, and the residues are

32

Handbook of Advanced Ceramics

FIGURE 8 Scheme of out-gases, reactions, and state of residues during carbonization.

carbonaceous materials which still contain hydrogen and a small amount of foreign atoms, such as O, N, and/or S, even though they are often called carbon materials. In order to eliminate these foreign atoms, a heat treatment up to 2000  C is required, strongly depending on the chemical state of these foreign atoms in the carbonaceous solid. The principal reactions in the carbon precursor up to about 800  C are pyrolysis and those above 800  C are carbonization. However, these two processes occur continuously and often overlap with each other. Usually, the whole process from the precursor to the carbon material

is called “carbonization”. This process depends strongly on the carbon precursors and also on the heat-treatment conditions, mainly temperature, heating rate, and pressure. A simple carbonization process is exampled on a polyimide film (commercially available Kapton film with 25 mm thickness); the structure of the repeating unit in the polyimide, the evolution of gases, and the changes in weight and size of the film up to 1000  C are shown in Figure 9aec, respectively [50]. The first step of carbonization occurs in rather narrow temperature range of  500e600 C, showing an abrupt weight loss associated

FIGURE 9 Carbonization of a polyimide film.

Chapter | 2.1

Advanced Carbon Materials

33

with the evolution of a large amount of carbon monoxide and a pronounced shrinkage along the film (pyrolysis). The second step occurs in the temperature range from 600  C to 1000  C with a small weight loss, the evolution of small amounts of methane and hydrogen, and a little shrinkage. The third step above 1000  C, which is not shown in Figure 9, is accompanied by the evolution of nitrogen, a negligibly small shrinkage, and a small weight loss. The release of nitrogen in the third step was found to continue up to temperatures higher than 2000  C, leaving a large amount of pores in the film if it was heated continuously up to 2400  C [36]. In the case of polyimides, the starting molecular structure of polyimide used is known to govern the structure and nanotexture of the resultant carbons. Carbonization process of pitches, which have been used in many carbon industries, various organic resins, such as phenol and furfuryl alcohol, and various biomasses, is much more complicated, because they have much more complicated molecular compositions. In Table 1, the carbonization processes are classified on the basis of the intermediate phases (carbonaceous materials) to the final carbon materials and the representative carbon materials formed are listed with their characteristics.

As can be seen from Table 1, most of carbon materials belong to the graphite family, of which the fundamental structural unit is a stack of carbon layers with a strong anisotropy. The way how the structural units tend to agglomerate is determined during carbonization process by the schemes and degrees of their preferred orientation, and results in a variety of nanotextures in carbon materials, which have the determining effects on their properties and also on their structural changes at high temperatures. Therefore, carbonization is the most important process for producing carbon materials with various functions.

4.2. Nanotexture A classification of nanotextures in carbon materials, which are constructed by fundamental structural units with nanosize, has been proposed as shown in Figure 10 [48,54]. Firstly, random and oriented nanotextures are differentiated and, secondly, the latter is divided accordingly to the scheme of orientation, in parallel to a reference plane (planar orientation), along a reference axis (axial orientation) and around a reference point (point orientation). In Figure 10, some representative carbon materials are listed for each nanotexture.

TABLE 1 Carbonization Processes for the Production of Various Carbon Materials Carbonization process

Precursors

Carbon materials

Characteristics

Gas-phase carbonization

Hydrocarbon gases, decomposition in space

Carbon blacks

Fine particles, so-called “structure”

Hydrocarbon gases, deposition on substrate

Pyrolytic carbons

Various textures, preferred orientation

Hydrocarbon gases, with metal catalysts

Vapor-grown carbon fibers Carbon nanofibers

Fibrous morphology, various nanotextures

Hydrocarbon gases, without any catalyst

Diamond-like carbon

Thin film, sp3 bond, amorphous structure

Carbon vapor

Carbon nanotubes

Tubular, single-wall and multiwalled

Carbon vapor

Fullerenes

Spherical, molecular nature

Plants, coals, and pitches

Activated carbons

Highly porous adsorptivity

Furfuryl alcohol, phenol resin, cellulose, etc.

Glass-like carbon

Amorphous structure, conchoidal fracture, gas impermeability,

Poly(acrylonitrile), pitch, cellulose and phenol resin

Carbon fibers

Fibrous morphology, high mechanical properties,

Polyimide films

Carbon films

Film wide range of graphitizability

Pitch, coal tar

Cokes Mesocarbon microbeads

Spherical particles

Cokes with binder pitches

Polycrystalline graphite blocks, including high-density isotropic graphite

Various densities, various degrees of orientation

Solid-phase carbonization

Liquid-phase carbonization

34

Handbook of Advanced Ceramics

FIGURE 10

Nanotextures in carbon materials.

The extreme case of planar orientation is a graphite single crystal, i.e., a perfect orientation with a regular stacking of large-sized hexagonal carbon layers. Single crystals can be found in either naturally occurring graphite (natural graphite) or in precipitated graphite flakes from molten iron (kish graphite). However, finding large-size crystals (more than 10 mm) is extremely difficult. Various intermediate nanotextures between the perfectly planar and random orientations are found in pyrolytic carbons and coke particles, depending on the preparation and heat-treatment conditions. In pitch-derived cokes, the planar orientation of the layers is improved with increasing the heat-treatment temperature (HTT), as demonstrated through high-resolution transmission electron microscopy (HR-TEM) [115,117]. The extreme case of co-axial orientation is carbon nanotubes, including single-wall and multi-walled

FIGURE 11

nanotubes. An axial orientation of the layers is found in various fibrous carbon materials; in other words, a fibrous morphology is possible because of this axial orientation scheme. In carbon fibers, the co-axial and radial alignments of the layers along the reference axis (i.e., fiber axis) are possible. In the point orientation, the concentric and radial alignments have also to be differentiated (Figure 10). The extreme case of concentric point orientation is found in the fullerene family. This orientation is also found in spherical particles of carbon blacks, of which the diameter is from few tens to few hundreds nanometers, fundamental structural units consisting of minute hexagonal carbon layers being preferentially oriented along the surface (Figure 11a). A radial alignment of the layers to form a sphere is found in the carbon spherules, which are formed by pressure carbonization of a mixture

Nanotextures in three spherical carbon materials.

Chapter | 2.1

Advanced Carbon Materials

of polyethylene and poly(vinyl chloride) (Figure 11c) [47,159]. The radial point orientation scheme appears near the surface of most mesophase spheres (mesocarbon microbeads, MCMBs), but in their center the orientation of the layers is not radial (Figure 11b) [2,38]. As a consequence of the radial orientation near the surface, tangential cracks are formed in most of MCMB particles after heat treatment at high temperatures [45,46]. A randomly oriented nanotexture occurs in glass-like carbons and also in carbon materials obtained by carbonization of polymer precursors (e.g., phenol resin and sugar). The fundamental structural units composing most glass-like carbons are so small that they are difficult to be observed by TEM. Therefore, their nanotexture was often discussed on the basis of TEM observations on high-temperature-treated materials, where the layers become somewhat larger. A 002 lattice fringe image of sugar coke heat-treated at a high temperature is shown in Figure 12a [117] and a model of nanotexture in Figure 12b [135]. In this model, closed pores are formed by the concentric alignment of small structural units. By taking the large amount of closed micropores in most glass-like carbons and their gas impermeability into consideration, this model is believed to be realistic.

4.3. Graphitization Heat treatment at high temperatures, in most cases up to 3000  C, in inert atmosphere is an essential process for the production of polycrystalline graphite blocks. The principal purpose of this treatment is the improvement in structure, i.e., the increase in the graphitization degree. The development of the graphite structure is evaluated by

35

different parameters. It is exactly evaluated by the graphitization degree P1, which is the probability to have the AB stacking sequence for neighboring hexagonal carbon layers. Conventionally, however, the average interlayer spacing between neighboring layers d002 and the crystallite sizes, thickness along the c-axis Lc, and lateral size of the layers La which are determined by X-ray diffraction analysis are often used as parameters to estimate the graphitization degree. The d002 values of highly crystallized graphite and amorphous carbon are 0.3354 nm and more than 0.344 nm, respectively. Figure 13aec show the variation of d002, Lc, and La versus heat treatment temperature, HTT, for various carbon materials with different nanotextures [59]. The needle-like coke mainly with a planar orientation and vapor-grown carbon fibers with a coaxial orientation show a rapid decrease in d002 and a rapid increases in Lc and La with increasing HTT, d002 approaching 0.3354 nm and La being more than 100 nm. Carbon blacks, thermal and furnace blacks, which have a co-axial point orientation, give relatively high d002 values of 0.339e0.341 nm and small Lc and La, even after the heat treatment up to 3000  C, the former being able to reach the smaller d002 and the larger Lc and La, mainly because of larger particle size than the latter with smaller size. Glass-like carbon with a random orientation gives much larger d002 and very small La even after 3000  C treatment. It had been proposed to classify the carbon materials into two classes, graphitizing (graphitizable and sometimes soft) and nongraphitizing (nongraphitizable and hard) carbons: the former can change to graphite by high temperature treatment under atmospheric pressure but the latter cannot. However, this classification is not recommended now, because intermediate behaviors are very often

FIGURE 12 002 lattice fringe image (a) and nanotexture model (b) of the carbon materials with random orientation.

36

Handbook of Advanced Ceramics

FIGURE 13 Changes in lattice parameters of different carbon materials. Planar orientation: needle-like and regular cokes; axial orientation: vaporgrown carbon fiber; point orientation: thermal and furnace blacks; and random orientation: glass-like carbon.

found, as clearly seen in Figure 13. The difference in graphitization behavior is now known to be mainly due to the nanotexture.

however, are described in more detail later because most of them can be called advanced carbon materials.

5.1. Gas-Phase Carbonization 5. VARIOUS CARBON MATERIALS In Table 1, various carbon materials are listed up on the basis of carbonization processes: gas-, liquid-, and solid-phase carbonization. Here, various carbon materials, which have been widely used in the industries and listed in Table 1, are briefly described on their preparation process and fundamental properties by dividing into three carbonization processes. Some of carbon materials shown in Table 1,

When carbon materials are produced through the decomposition of hydrocarbon molecules in gaseous phase, the process is called gas-phase carbonization. A large variety of carbon materials over the whole carbon families were produced through this process (Table 1). For a high concentration of hydrocarbon gases, carbon blacks are obtained. When a solid substrate is placed in the carbonization system, pyrolytic carbons are produced under a low concentration of hydrocarbon gases. If some fine metallic particles, such as Fe or Ni,

Chapter | 2.1

Advanced Carbon Materials

are added into the carbonization system, various carbon materials are produced: carbon nanotubes, carbon fibers (vapor-grown carbon fibers), carbon nanofibers with various nanotextures and morphologies. From the carbon vapor produced by electric-arc discharge or laser ablation, carbon nanotubes and fullerenes are formed.

5.1.1. Carbon Blacks Carbon blacks formed through incomplete combustion of either gaseous or mist-like hydrocarbons are very important industrial products [59]. They have been used as a colorant in inks since the third century AD and are now applied in a large amount for rubber reinforcement. They are characterized by spherical primary particles, which are in different sizes and more or less coalesced into aggregates, what is called “structure” in the field of rubber reinforcement. In Figure 14, transmission electron micrographs of some carbon blacks are shown. On the basis of the reaction process and raw materials, carbon blacks are classified into furnace black, channel black, lamp black, thermal black, and acetylene black; furnace black consists of the primary particles of few tens nanometer diameters with a marked “structure” as shown in Figure 14c and d, which are used for

FIGURE 14

37

the reinforcement of tires, but lamp black and thermal black consist of larger primary particles of few hundreds nanometer diameters without noticeable “structure” (Figure 14a). Ketjenblack is produced as a by-product of heavy oil gasification and has a size of primary particles of 30e40 nm and a high surface area (ca. 1300 m2/g) [99]. Ketjenblack and also acetylene black are often used as a conductive additive in the electrodes of primary and secondary batteries, and also for electrochemical capacitors. The concentric orientation of the carbon layers is more marked near the surface, whereas the layers are randomly oriented at the center, of which 002 lattice fringe image of TEM is shown in Figure 15a. When large-sized carbon black particles are heat-treated at a high temperature as 3000  C, the particles are polygonized and a hollow core appears (Figure 15b), because of the crystallite growth.

5.1.2. Pyrolytic Carbons Carbon deposition occurs on the surface of a substrate inserted into the carbonization system from hydrocarbon gases, such as methane and propane [59]. This process is a chemical vapor deposition (CVD) and the products are called pyrolytic carbons. The deposition conditions are

Transmission electron micrographs of various carbon blacks.

38

Handbook of Advanced Ceramics

FIGURE 15

002 lattice fringe images before and after the heat treatment at 3000  C.

known to govern strongly the nanotexture of the resultant pyrolytic carbons. The deposition can occur on either static or dynamic substrates. In the former, the substrate is placed in a furnace, which is heated by direct passing either of electric currents or from the surroundings. For a solid substrate, graphite plates and also the mats composed from carbon fibers are often used. In the latter, small substrate particles are fluidized in the furnace. The factors controlling the nanotexture and properties of pyrolytic carbons are (1) the kind of precursor hydrocarbon gas, (2) the gas concentration, (3) the deposition temperature, (4) the contact time between the gas molecule and the substrate heated at a high temperature, which is governed by the gas flow rate and the size of the furnace, and finally (5) the

FIGURE 16

geometrical arrangement of the furnace, particularly the ratio of the substrate surface area and the volume occupied by the gas [8]. In Figure 16, the bulk density of pyrolytic carbons formed on the solid substrate was plotted against the deposition temperature as a function of the precursor gases and their partial pressures. Two representative crosssectional optical textures are also shown in the figure. In order to prepare pyrolytic carbon with high density and singularly-nucleated texture at 1400e2000  C, it is necessary to use a low methane pressure, as low as 2 Pa. High methane pressure as 400 Pa gives low-density pyrolytic carbon in this temperature range, and the resultant pyrolytic carbon has regenerative optical texture, which is

Density and optical texture of pyrolytic carbon as a function of deposition condition.

Chapter | 2.1

Advanced Carbon Materials

known to be mainly due to the formation of carbon black particles incorporated into the pyrolytic carbon. The plates of so-called highly-oriented pyrolytic graphite (HOPG) [7] are produced by the annealing under high-temperature and high-pressure conditions. HOPG has a very high degree of planar orientation of the hexagonal carbon layers, but the size of the layers is not so large. In other words, the structure is close to a perfect planar orientation in the cross-section of the HOPG plate, but it is polycrystalline in the plane parallel to the plate [164]. FESEM image on a fractured cross-section and a channeling contrast image on a surface of an HOPG plate are shown in Figure 17a and b, respectively. A well-ordered stacking of the hexagonal carbon layers is seen in Figure 17a, whereas Figure 17b demonstrates the polycrystalline nature, consisting of small grains of graphite basal planes in different a-axis orientations The so-called carbon/carbon composites were fabricated by the filling of the pyrolytic carbons in the spaces between carbon fibers, the process being often called chemical vapor infiltration (CVI).

39

method was developed [20], in which nanometric catalyst metal particles are formed in situ by the decomposition of its precursor, instead of the method where the catalyst is seeded on the surface of a substrate plate (seeding catalyst method). Since most VGCFs can be changed to well-graphitized carbon fibers and they can be prepared with a high purity, without the coexistence of other carbon forms like carbon blacks, several works were performed to prepare fibrous carbons through the decomposition of various gases, not only hydrocarbons but also CO, under different conditions [1,3,6,128,137]. By selecting the decomposition conditions, carbon helical microcoils with a controlled pitch were prepared [12,108,109]. Fibrous carbons with a stacking of cup-like carbon layers along the fiber axis, cup-stacked nanofibers, were also prepared in a similar process [24,25]. Carbon nanohorns, showing a high-storage capacity for methane, were also synthesized by a similar technique [44]. The addition of a small amount of VGCF into the electrode of lead-acid and lithium-ion rechargeable batteries was reported to improve their performance [23].

5.1.3. Vapor-Grown Carbon Fibers Vapor-grown carbon fibers (VGCFs) are formed on a substrate from benzene vapor in a flow of high-purity hydrogen using fine iron particles as catalyst [19,116]. The high-resolution TEM analysis using various modes (brightfield images, selected area electron diffraction, dark-field images with 002 and 10 diffractions, 002 and 10 lattice fringes) shows that the VGCFs are constituted of two parts: a hollow tube with few straight layers (carbon nanotube) (Figures 18a and c) and small layers deposited on the carbon nanotube and oriented preferentially and concentrically along the fiber axis (Figures 18a and d) [116]. The formation of similar VGCFs was reported by using different precursor hydrocarbon gases, carrier gases, catalyst metals and deposition conditions [64,65,75,151]. In order to get a high yield of VGCFs, the floating catalyst

5.1.4. Carbon Nanofibers CVD processes under the conditions similar to VGCF production was reported to give various carbon nanofibers. Some of the papers, particularly in the beginning, reported that they prepared carbon nanotubes, but most of their products were carbon nanofibers. A concentric alignment of the layers is realized in multiwalled carbon nanotubes and various carbon nanofibers, most of which are grown through CVD using fine catalyst particles. However, different orientation schemes can be found along their axis, from parallel (tubular type) to perpendicular (platelet type) through herring-bone type, as illustrated in Figure 19. Examples of herring-bone- and tubular-type nanotextures are shown in Figure 20 for both as-prepared and high-temperature-treated carbon

FIGURE 17 SEM image of cross-section and electron channeling pattern of the surface of HOPG. (Courtesy of Prof. A. Yoshida of Tokyo City University, Japan)

40

Handbook of Advanced Ceramics

FIGURE 18 Transmission electron micrographs of vapor-grown carbon fiber. (Courtesy of Prof. M. Endo of Shinshu University, Japan)

nanofibers. After high-temperature treatment, the nanotexture is recognized more clearly, mainly because of the layer growth. Impregnation of molten poly(vinyl alcohol) into the channels of anodic aluminum oxide films gives nanofibers with a platelet nanotexture [80].

nanotubes were reported. The hollow tubes thus obtained show a wide range of diameters from 1 to 50 nm and their walls consist of different numbers of carbon layers. Most of them are closed at the end with the help of pentagons, the smallest diameter being the same as the size of the smallest fullerene C60.

5.1.5. Carbon Nanotubes Carbon nanotubes were found to be formed at the beginning of VGCF formation, as shown in Figure 18 [116], and later formed in the carbon deposits on the graphite anode after arc discharge in He atmosphere [42]. The carbon nanotubes consisting of a single carbon layer were also found later [5,43]. The temperature at the graphite electrode during arc discharging was estimated to reach up to 2500  C. By selecting the arc discharge conditions between the graphite electrodes, relatively high yields of carbon

5.1.6. Diamond and Diamond-Like Carbons Diamond was firstly synthesized under a high static pressure of about 10 GPa by using a catalyst of either Fe or Ni [10]. Diamond synthesis was carried out under a high pressure with various conditions, including static and shock-wave compression. Also, diamond crystals were synthesized by CVD of hydrocarbon gases, such as CH4, C2H2, andC6H6, under a low pressure of about 1 Pa at around room temperature [73].

Chapter | 2.1

Advanced Carbon Materials

FIGURE 19 Nanotexture in carbon nanofibers.

Diamond-like carbon (DLC) is synthesized via gasphase carbonization of hydrocarbons. It is as hard as the diamond crystal, because of sp3 covalent CeC bonding, but has amorphous structure due to random orientation of carbon tetrahedra, as explained before. It contains a relatively large amount of hydrogen. A large number of DLC films were examined by using the X-ray absorption fine structure spectroscopy [131], most of them containing of sp2-hybridized carbon atoms in a range of 20e60 %.

5.1.7. Fullerenes The carbon cluster C60 was firstly found in the soot formed by laser ablation of graphite in He atmosphere and named

41

as buckminsterfullerene [85]. C60 was also found in the carbon deposits formed by arc discharge between graphite electrodes [83]. This carbon cluster C60 could be extracted using solvents, such as toluene, carbon disulfide, etc. and also be separated from the deposits by vaporization at around 400  C in vacuum [84]. Carbon clusters with different sizes, consisting of different numbers of carbon atoms, such as C70, C82, ., C960, were identified and this series of clusters was called fullerenes. Among the different gas atmospheres which were examined for the formation of fullerenes, helium atmosphere gives relatively high yield of fullerenes in most cases. Fullerene inserted by H2, H2@C60, was successfully synthesized through an organic synthesis process [78]. Also, Ar was inserted into fullerene to give Ar@C60, and, by doping K into the interstitial sites of these clusters, K3Ar@C60, was shown to improve superconductive performance [139]. The Gd-metallofullerenol Gd@C82(OH)n, where Gd atoms are inserted into the C82 cage and eOH radicals are attached on the carbon atoms of C82, has been reported to be useful as a contrast agent for magnetic resonance imaging (MRI) [74].

5.2. Solid-Phase Carbonization Thermosetting resins, such as phenol-formaldehyde and furfuryl alcohol, and also cellulose can be converted to carbon materials by solid-phase carbonization. When precursors are carbonized quickly, the resultant carbon materials become porous, as being represented by activated carbons. If the carbonization is performed so slowly that the resultant carbonaceous solids can shrink completely,

FIGURE 20 002 lattice fringe images of nanofibers with different nanotexture.

42

Handbook of Advanced Ceramics

FIGURE 21 Preparation process of activated carbon.

the so-called glass-like carbons are produced, which contain a large amount of closed pores.

5.2.1. Activated Carbons In activated carbons, a large amount of open pores are formed, which are usually evaluated by surface area and pore-size distribution [15,101]. They have been used as adsorbents since prehistoric age and are now used widely not only in our daily life but also in various industries in the processes for both production and waste treatment. For the predominant applications of activated carbons as adsorbents, the pore texture is the most important property to be controlled. Figure 21 shows a conventional process for the production of activated carbon; various carbon precursors, not only thermosetting resins, including natural biomasses, but also some thermoplastic resins, such as pitches and coals, are used. For thermoplastic precursors, the so-called stabilization, which is a slight oxidation, is required prior to carbonization. In order to develop micropores with a width less than 2 nm, mild oxidation of carbonized materials is usually carried out, which is called “activation”. For activation, different processes are employed, gas activation (e.g., diluted oxygen gas, air, steam, CO2, etc.) and chemical activation using H3PO4, ZnCl2, KOH, etc. During gas activation, the pores may grow from micropores to mesopores and finally to macropores. Figure 22 shows the changes in pore size during air activation of glass-like carbon spheres [60]. At the very beginning of oxidation, ultramicropores (50 nm width)

and mesopores are usually coexisted with micropores. These large-sized pores, macropores and some of mesopores, work as diffusion paths for oxidizing agents to create micropores during activation process and also as those for adsorbate molecules to reach the micropores during adsorption application. The pore-size distributions of activated carbons used for the adsorption of small gas molecules and decolorization of water are shown in Figure 23b. By the control of micropore size, the activated carbons may be successfully used as molecular sieves for various gases. Mesopores are effective for the adsorption of gasoline vapor composed from large hydrocarbon molecules [69]. Pore-size distribution of the activated carbon for vehicle

FIGURE 22 Pore development during gas activation on glass-like carbon sphere.

Chapter | 2.1

Advanced Carbon Materials

43

FIGURE 23 Model of pore structure and pore-size distributions in activated carbons.

application to adsorb and desorb gasoline vapor (car canister) is shown in Figure 23c.

5.2.2. Glass-Like Carbons Glass-like carbons (glassy carbons) are produced by the pyrolysis of different precursors, such as phenol-formaldehyde resin, poly(furfuryl alcohol), and cellulose, through an exact control of the heating process [28,114]. They are characterized by amorphous structure and also by properties very similar to inorganic glasses, such as high hardness, brittle conchoidal fracture, as shown in Figure 24a, and gas impermeability. The most crucial parameter for their production is a very slow heating, slower than the rate of shrinkage of carbonaceous materials on the carbonization process to compensate the formation of open pores due to the evolution of decomposition gases from the precursor block. In Figure 24b,

the changes in weight, volume, BET surface area, and adsorption of water vapor with carbonization temperature are shown for poly(furfuryl alcohol) condensates [28]. The rapid decreases in weight and volume up to 700e800  C are due to the precursor pyrolysis and correspondingly the BET surface area and adsorption of water vapor increase. Above 800  C, however, the BET surface area and water vapor adsorption decrease quickly under very slow heating rate, because of the fact that most of pores are closed. The commercially available glass-like carbons have a bulk density of 1.3e1.5 g/cm3, suggesting the existence of a large amount of pores, and show very low gas permeability as 10 12 cm2/s, almost gas impermeable, suggesting that all of the pores are closed. The formation of closed pores with rather homogeneous size of about 5 nm was confirmed on a glass-like carbon, the volume of which reached around one-third of the bulk [112].

FIGURE 24 Conchoidal fracture surface of a glass-like carbon and changes in structural parameters with carbonization temperature. (Courtesy of Dr. Ekinaga of Tokai Carbon Co., Ltd.)

44

Handbook of Advanced Ceramics

5.2.3. Carbon Fibers In Table 2, carbon fibers are classified on the basis of their precursor, and their characteristics are summarized, including vapor-grown carbon fibers (VGCFs) prepared through gas-phase carbonization for the comparison. Commercially available carbon fibers are classified into two grades, general purpose and high-performance grades on the basis of their mechanical properties. The latter is further divided into two types, high-modulus and highstrength types [16]. In order to produce carbon fibers from precursor polymers (for example, PAN) and various pitches, stabilization is essential after their spinning, which consists of chemical reaction by using different oxidizing gases, such as air, oxygen, chlorine, and hydrochloric acid vapor [16]. The stabilized fibers are then subjected to solid-state carbonization (extrinsic solidphase carbonization) in inert atmosphere. When cellulose and phenol fibers are used as precursors, the stabilization process is not necessary, and the spun fibers are subjected directly to carbonization (intrinsic solid-state carbonization). In Figure 25, the variety in nanotexture of different carbon fibers is illustrated, on the cross-sections parallel and perpendicular to the fiber axis [59]. Most of thermosetting precursors, such as phenol and cellulose, and also isotropic pitches, give carbon fibers with random nanotexture in both cross-sections. For mesophase-pitch-based carbon fibers, however, different nanotextures in the crosssection perpendicular to the fiber axis, from radial and coaxial to random, are possible (Figure 25) [51]. VGCFs have a co-axial alignment of small carbon layers on a carbon nanotube at the center (Figure 18). Figure 26

shows, on mesophase-pitch-based carbon fibers, that the cross-sectional nanotexture governs also the graphitization behavior of the fibers [51]; fibers with straight radial nanotexture in its cross-section have a relatively high graphitization degree, expressed by low d002 and high La values, but those with zigzag radial nanotexture have a poor graphitization degree. In VGCFs, the small layers coalesce with each other forming long and smooth layers after 3000  C treatment; the degree of concentric alignment is greatly improved and the perpendicular cross-section of the fiber is polygonized, mainly due to crystallite growth [165]. The difference in the nanotexture in carbon fibers becomes more pronounced after high-temperature treatment.

5.2.4. Carbon Films Polyimide films with different molecular structures have been developed as thermoresistant polymers and used in different fields, especially in the field of electronics. The heat treatment of these films at high temperatures leads to a wide variety of carbon structures, from highly crystalline graphite to amorphous glass-like carbon films [53]. This is a typical case where the molecular structure of the organic precursors and the texture of their films govern the structure and texture of the resultant carbon films, i.e., crystallinity and preferred orientation of hexagonal carbon layers. As shown on a polyimide film (commercially available film Kapton) in Figure 9b [50], the yield after carbonization is relatively high, about 60 mass%, and the linear shrinkage along the film is slightly larger than 20 %. Even if these changes in weight and size are not so small, it is noteworthy to mention that the film can keep its original form, although

TABLE 2 Classification of Carbon Fibers Based on their Precursors and their Characteristics Precursor

Carbon fibers

Characteristics

Polyacrylonitrile (PAN)

PAN-based carbon fibers

Different grades and types

Isotropic pitch

Isotropic-pitch-based carbon fibers

General purpose grade Random cross-sectional nanotexture

Mesophase pitch

Mesophase-pitch-based carbon fibers

High modulus types Various cross-sectional nanotextures

Cellulose

Cellulose-based carbon fibers

General purpose grade Random nanotexture

Phenol

Phenol-based carbon fibers

General purpose grade Random nanotexture

Hydrocarbon gases

Vapor-grown carbon fibers

High graphitizability Annual-ring texture in cross-section

Chapter | 2.1

Advanced Carbon Materials

45

FIGURE 25 Nanotexture in various carbon fibers.

a little care not to give mechanical shock is required. Figure 27 shows a crane made from a polyimide film by a technique of paper folding, the form of which is maintained after carbonization up to 1300  C and even after graphitization at about 3000  C. Carbon films were also prepared from other organic films, poly(phenylene vinylene) (PPV) [118] and poly-pphenylene-1,3,4-oxadiazole (POD) [110,111,134]. From the polyimide films with different molecular structures, either consisting of a large number of phenyl rings or containing fluorine, highly microporous films were synthesized by a simple carbonization at around 1000  C in an inert atmosphere [119,120]. The polyimide film having 31.3 mass% fluorine in its repeating unit of molecules gave a highly nanoporous carbon film having a microporous surface area of 1340 m2g 1 and consisting of micropores with a width of ca. 0.55 nm [119]. By controlling carbonization temperature, microporous carbon films were prepared, which had molecular sieving function [27,33]. Carbon films containing macropores with a diameter less than 1 mm were prepared from a polyimide by the introduction of phase inversion technique during the gelation of poly(amic acid) [32]. Co-polymers of polyimide and polyurethane gave macroporous carbons [140]. The preparation of microporous carbon films and their properties were recently reviewed [149].

5.3. Liquid-Phase Carbonization Liquid-phase carbonization occurs for some precursors, such as pitches, which become viscous fluids before carbonization. Through this process, the cokes are prepared from petroleum and coal tar pitches, which are used as the raw materials for industrial production of various polycrystalline graphite blocks as the electrodes for steel refining and electrical discharge machining, the jigs for the growth of semiconductor crystals, the structure components of the nuclear reactors, etc. Polycrystalline graphite blocks are fabricated by using a filler coke, which is prepared from pitches through liquid-phase carbonization, and a binder pitch, which is carbonized in between the particles of the filler coke in a formed block also through liquid-phase carbonization. Spheres with optically anisotropic texture are formed at the beginning of the liquid-phase carbonization of pitches and are separated from the isotropic pitch matrix, which are called mesocarbon microbeads (MCMBs).

5.3.1. Cokes Cokes are the products of liquid-phase carbonization of pitches. In the course of this process, optically anisotropic spheres are firstly formed, which are called mesophase spheres. By further heating, the spheres grow and coalesce with each other to give liquid crystals with different

46

Handbook of Advanced Ceramics

textures composed of optically anisotropic units, called bulk mesophase [9,38]. A series of polarized-light microscopy images in Figure 28 shows the sequence from the formation of small mesophase spheres to their growth and partial coalescence. By further heating, their bulk mesophase becomes solid with anisotropic textures. By applying a shear stress during liquid-phase carbonization, the so-called needle-like cokes are produced, which are now important raw materials in the production of largesized graphite electrodes for metal refining [59]. In the needle-like coke, basic structural units are well oriented along the shear direction, as shown by polarized-light micrograph in Figure 29. The structure improvement, i.e., graphitization, occurs markedly in the cokes with increasing heat-treatment temperature, as shown by plotting XRD parameters against HTT for a needle-like coke in Figure 13. In Figure 30, 002 lattice fringe image of TEM is shown on a coke with different HTTs, revealing that the size of the layer planes observed as 002 fringes grow and their parallel stacking is improved with increasing HTT.

5.3.2. Mesocarbon Microbeads

FIGURE 26 Nanotexture in the perpendicular cross-section of mesophase-pitch-based carbon fibers and their lattice parameters after 3000  C treatment.

The anisotropic mesophase spheres shown in Figure 28b can be separated from the isotropic pitch matrix by using a solvent [160]. The separated spheres are called mesocarbon microbeads (MCMBs) and have been used as anode material of lithium-ion rechargeable batteries [150]. Because of their nanotexture of radial point orientation, many cracks in radial direction are formed, accompanying certain structure improvement, after the high-temperature treatment, as shown in Figure 31, which seem to make intercalation/deintercalation of lithium ions easier.

FIGURE 27 Crane made from a polyimide film by paper folding technique and heat-treated at different temperatures. For color version of this figure the reader is referred to the online version of this book (Courtesy of Dr. A. Hatori of AIST, Japan)

Chapter | 2.1

Advanced Carbon Materials

47

FIGURE 28 Formation, growth, and coalescence of mesophase spheres in a pitch.

FIGURE 29 Appearance and cross-section of a needle-like coke. For color version of this figure, the reader is referred to the online version of this book.

FIGURE 30

(a)

(b)

(c)

(d)

002 lattice fringe images of a needle-like coke heat-treated at different temperatures. (Courtesy of Mme A. Oberlin, France)

5.3.3. Polycrystalline Graphite Blocks The production procedure of polycrystalline graphite blocks is shown in Figure 32 [59]. For the filler, cokes derived from petroleum and coal tar pitches are usually used, but natural graphite, carbon blacks, and also recycled graphite particles are sometimes added. In advance, the cokes are carbonized

in order to avoid the evolution of a large amount of volatiles, which introduces some shape distortions and cracks in the product and also reduces the density. For the binder, petroleum and coal tar pitches are used in most cases, because they have relatively high carbon yield as about 60 mass%, and also because they give a carbon similar to the filler cokes after carbonization. The particle size of the filler and the

48

Handbook of Advanced Ceramics

5.4. Novel Carbonization Processes The recent developments in science and technology require a more exact control of structure/nanotexture and properties of carbon materials. In order to meet these requirements, various novel carbonization processes have been proposed: template method, polymer blend method, defluorination of fluorinated hydrocarbons, and carbonization of organic aerogels [57].

5.4.1. Template Methods

FIGURE 31 Crack formation in mesocarbon microbeads after hightemperature treatment.

mixing ratio of the filler to the binder have to be controlled, in accordance with the requirements from the applications. The filler and the binder are mixed at a temperature higher than the softening point of the binder. The mixture thus prepared, which is usually called carbon past, is formed after warming up at a temperature around 150  C by either extrusion, molding, or cold isostatic pressing. The formed blocks are carbonized at a temperature of 700e1000  C (called calcination in industries) and then graphitized at a high temperature above 2500  C. The blocks heat-treated at a high temperature are polycrystalline, and often called “synthetic graphite” or “artificial graphite”. For producing graphite blocks, liquid-phase carbonization is involved in two steps: i) during the process of filler coke preparation and ii) during the carbonization of the binder pitch in a carbon paste. Forming is an important process for the fabrication of polycrystalline graphite blocks, because it governs the preferred orientation of the filler particles, which are often anisotropic. The extrusion gives a preferred orientation of either flaky or needle-like filler particles along the direction of extrusion. Electrodes for metal processing with a large diameter up to 80 cm, carbon rods with different sizes, and also leads for automatic pencil with a diameter as thin as 0.3 mm are fabricated by this forming process. In graphite electrodes used in steel refining, needle-like cokes are essential to attain a high thermal shock resistance. Sudden and large expansion due to abrupt increase in temperature can be absorbed into a crack formed in needle-like coke particles as shown by arrow in Figure 29b. In the molding process, which is applied for the fabrication of carbon brushes for electric motors and electric contacts, the filler particles are statistically aligned perpendicular to the compressing direction. By isostatic pressing, the filler particles are randomly oriented, which lead to producing isotropic high-density graphite blocks, which will be described in detail later.

The template carbonization technique was firstly developed for the preparation of thin oriented graphite films using two-dimensional spaces in clay minerals with a layered structure [88,138]. Template carbonization reveals a high possibility to control the dimensionality of morphology by selecting suitable template materials: onedimensional carbon nanofibers were synthesized using anodic aluminum oxide films, two-dimensional graphite layers using layered compounds, and three-dimensional nanoporous carbons using zeolites, silicas, MgO, and some surfactants [63]. Carbon nanofibers were prepared by carbon deposition from propylene gas at 800  C on the inner walls of nanosized channels in an anodic aluminum oxide film [89]. Aluminum oxide is subsequently dissolved either by HF at room temperature or by NaOH aqueous solution at 150  C in an autoclave. The procedure is schematically shown in Figure 33 with a SEM image of the resultant nanofibers. The tubular carbon deposits in the channels of aluminum oxide could be preferentially modified by fluorine and oxygen [34,92], and also easily filled by metal and metal oxide to prepare their nanowires [11,91,121,126,127]. The tubular carbons thus prepared could be easily converted to multiwalled carbon nanotubes by the heat treatment at high temperatures in inert atmosphere. In order to control the pore structure in carbon materials, various templates were developed, as summarized in Table 3 by listing template, carbon precursors, and characteristics of the pore structure in the resultant carbons. Carbons with a high BET surface area, even higher than 3000 m2/g, were prepared using the three-dimensional channels of zeolite, of which micropores had homogeneous size and were ordered [90,93,97,98,103]. By using mesoporous silicas, mesoporous carbons were formed, pores being possible to be either ordered or disordered and also to be channels [70,71,86,94,130]. By using some surfactants, mesoporous carbons could be prepared directly from furfuryl alcohol and polyacrylonitrile [81,95,147]. MgO was also used as template for obtaining mesoporous carbons [106,107]. A thermoplastic carbon precursor (such as poly(vinyl alcohol), poly(ethylene terephthalate), and pitch) is mixed with a MgO precursor

Chapter | 2.1

Advanced Carbon Materials

49

FIGURE 32 Production process of polycrystalline graphite blocks.

(MgO, Mg acetate, Mg citrate, and Mg gluconate), and heat-treated at 900  C in inert atmosphere. Then MgO formed through the pyrolysis of its precursor is dissolved by a diluted acid solution, liberating the mesoporous carbon. It was proved that the mesopore size of the resultant carbon is governed by the size of MgO thus formed. A large amount of mesopores with a size of about 5 nm is easily obtained using Mg citrate [104]. Figure 34a and b show the pore volume and size distribution, respectively, as a function of the mixing ratio between Mg citrate and PVA. From the mixture of Mg citrate and gluconate, a carbon with bi-modal mesopore sizes was obtained [105] Microporous and/or mesoporous carbons were proposed to be used as adsorbents for various molecules

with a wide range of sizes, electrodes for electrochemical capacitors, etc. [63] By impregnating a polyimide into organic foams, commercially available polyurethane and melamine foams, carbon foams were prepared after carbonization [58]. Micropores could be easily created in the walls of macropores.

5.4.2. Polymer Blend Method By spinning of a mixture of two kinds of carbon precursors, the one giving a relatively high carbonization yield and the other having a very low yield, and following carbonization, carbon nanofibers were successfully prepared

50

Handbook of Advanced Ceramics

FIGURE 33 Preparation scheme and SEM of the carbon nanofibers through template method. (Courtesy of Prof. T, Kyotani of Tohoku University, Japan)

[39,40,123,133]. The preparation procedure is schematically shown in Figure 35. The resultant nanofibers could be converted to multiwalled carbon nanotubes by the heat treatment at high temperatures in inert atmosphere (Figure 35). Poly(urethane-imide) films were easily converted to macroporous carbon films [66,140,141,142]. The films prepared by blending poly(amide acid) with phenol-terminated polyurethane prepolymers were heated up to 200  C for the phase separation to polyimide (PI) matrix and small polyurethane (PU) islands. By heat treatment up to 400  C, PU component was pyrolyzed to gases and resulted in porous polyimide films, which were able to convert easily to porous carbon films by carbonization in inert atmosphere. Pore size in these polyimide and carbon films was controlled by changing the blending ratio of PI to PU and also the molecular structure of PU. The carbon films prepared were

shown to be suitable as a medium for culturing biological cells, with macropores in the range of 0.6e3.0 mm sizes being shown to be preferable for the cells [82].

5.4.3. Defluorination of Polymers Mesoporous carbons with a high surface area are obtained through the defluorination of poly(tetrafluoroethylene) (PTFE) with alkali metals [17,67,145,146,161]. PTFE film with 100 mm thickness was pressed with lithium metal foil with 200 mm thickness under 4 MPa in Ar atmosphere for 48 h to defluorination of PTFE. Excess lithium metals and finely dispersed LiF were eliminated by washing with methanol, heating up to 700  C, and washing again by dilute HCl [161]. Defluorination of PTFE was also possible through heating a mixture of PTFE powders with alkali

TABLE 3 Template Carbonization for Pore Control in Carbon Template In-organic

Organic

Carbon precursors

Carbon prepared

Zeolites

Zeolite Y, b and L, ZSM-5

Propylene CVI þ FA impregnation

Ordered micropores

Mesoporous silicas

MCM-48, -41, SBA-1, -15

Impregnation of sucrose, FA, PF, MP. CVI of acetylene

Ordered or disordered mesopores

Colloidal silicas

Organic silicates

Mixing with RF, AN,

Disordered micro- and mesopores

MgO

MgO, Mg acetate, citrate, Mg(OH)2

Mixing with PVA, coal Tar pitch, PET, PI

Disordered mesopores

Surfactants

Diblock and triblock copoymers

Mixing with RF, EOA, PGF, AN

Ordered or disordered mesopores

Organic foams

Urethane and melamine foams

PI

Macropores

CVI: chemical vapor infiltration, FA: furfuryl alcohol, PF: phenol formaldehyde, MP: mesophase pitch, RF: resolicinol-formaldehyde, PVA: poly(vinyl alcohol), PET: poly(ethylene terefuthalate), PI: polyimide, EOA: triethyl orthoacetate, PGF: phloroglucinol-formaldehyde.

Chapter | 2.1

Advanced Carbon Materials

51

FIGURE 34 Pore volume and pore-size distribution of mesoporous carbons prepared by MgO-templated method.

metals (Na, K, or Rb) in vacuum at 200  C in a closed vessel [141,146]. Defluorination of PTFE with Na metal was found to give mesopore-rich carbon and very high SBET as 2225 m2g 1.

5.4.4. Carbon Aerogels Mesoporous carbons were also prepared by the pyrolysis of resorcinol-formaldehyde organic aerogels [4,30,31,124,125,143,144,163]. Primary carbon particles have the size of about 4e9 nm and interconnected with each other to form a three-dimensional network, which gives predominantly interparticle mesopores. Adsorption isotherms of the resultant carbon aerogels belong to type IV and have a clear hysteresis [31]. Pore structure of carbon aerogels was known to be governed by that of precursor organic aerogels, which was controlled by the mole ratios of resorcinol to formaldehyde (R/F), to water (R/W), and also to basic catalyst Na2CO3 (R/C) [143]. Instead of supercritical drying of aqueous gels, freeze-drying method

was also applied [144]. On the gels prepared through freeze drying (cryogels), much smaller shrinkage in pore size during carbonization was observed.

6. IMPORTANCE OF TEXTURES IN CARBON MATERIALS 6.1. Nanotextures Nanotexture in carbon particles, which is established in the process of carbonization, governs their structure and properties. As shown in Figure 13, structural improvement in carbon materials with the heat treatment under atmospheric pressure depends strongly on their nanotextures; needle and regular cokes with planar orientation can approach graphite structure above 2500  C, but glass-like carbon with random orientation is almost impossible to be graphitized even at very high temperature as higher than 3000  C. Thermal and furnace blacks with point orientation have an intermediate behavior between the carbon with

FIGURE 35 Preparation scheme of carbon nanofibers through polymer blend method. (Courtesy of Prof. A. Oya of Gunma University, Japan)

52

Handbook of Advanced Ceramics

FIGURE 36 Change in 002 diffraction profile and polarized-light micrograph of a glass-like carbon with heat-treatment temperature under the pressure of 0.5 GPa.

planar orientation and that with random orientation, the former being improved its structure with heat treatment more markedly than the latter. Graphitization behavior for vapor-grown carbon fibers, which has an axial orientation, is similar to the cokes, but that for other carbon fibers is governed by their axial orientation schemes along their fiber axes, as exampled by using mesophase-pitch based carbon fibers in Figure 26. In the case of intercalation reactions, the nanotexture of the host carbon materials has a strong effect. The intercalation of sulfuric acid into natural graphite can proceed at room temperature in concentrated acid. The resulting intercalation compound is commonly used in industry for the preparation of exfoliated graphite [56]. However, in order to intercalate sulfuric acid into carbon fibers, the electrolysis in the acid is needed [152,153,155]. Potassium as a vapor, on the other hand, can be easily intercalated in various carbon materials, even into lowtemperature-treated carbon fibers. The nanotexture once formed is extremely difficult to change; in other words, it needs very severe conditions to destroy the nanotexture formed. Glass-like carbons, which have random orientation, cannot change to graphite under normal pressure even above 3000  C, as shown in Figure 13. In order to change them to graphite, either melting or heat treatment under a high pressure is needed. Melting point of carbon is located at a temperature around 4000 K and a pressure above 10 MPa. When a rod of glass-like carbon

was heated by direct passing the electric current under Ar pressure, it could be melted at its middle part and quenched suddenly to a low temperature, because of its breaking. After melting, a ball was found in a crater formed by melting. It was shinny and lubricant graphite with exactly the same lattice parameters as natural graphite. However, the wall of the crater was far from graphite, of which 002 diffraction line was broad and unsymmetrical, and of which d002 spacing was about 0.34 nm, even though it was heated close to melting point [113]. Glass-like carbon spheres could be graphitized under a high pressure as high as 0.5 GPa at 1600  C [52,72]. Change in 002 diffraction profile with HTT under 0.5 GPa pressure was shown in Figure 36a. The 002 diffraction line has a composite profile, consisting of two structural components, the one located at a low angle side (T component) is due to the pristine glass-like carbon and the one at high angle side having d002 of about 0.335 nm (G component), which was found at the contact point between two spheres of glass-like carbon, supposing to be stress accumulated point, as shown by polarized-light micrographs in Figure 36b. With increasing HTT under pressure, relative intensity of the G component to the T component increases, revealing gradual increase in graphitized parts, and correspondingly optically anisotropic parts increase their areas. Finally, the single peak of the G component in X-ray diffraction and totally anisotropic area in optical micrograph were obtained at 1600  C, revealing a completion of graphitization.

Chapter | 2.1

Advanced Carbon Materials

53

(b) After 1000 oC- treatment

(a) After 110 oC curing

10 µm

(c) After 2500 oC-treatment

(d) After 2800 oC-treatment

FIGURE 37 Change in polarized-light micrograph of carbon fiber/glass-like carbon composite with heat-treatment temperature. For color version of this figure, the reader is referred to the online version of this book.

Graphitization of glass-like carbon due to stress accumulation was also observed in the composite of carbon fibers with glass-like carbon matrix [35]. Polarized-light micrographs of the composites heat-treated at different temperatures under normal pressure are shown in Figure 37. Optically anisotropic area is formed preferentially at the interfaces between fiber and glass-like carbon matrix formed from furfuryl alcohol condensate after carbonization at 1000  C (Figure 37b), although the composite before carbonization was optically isotropic (Figure 37a). This formation of anisotropic areas is reasonably supposed to be due to a large shrinkage of the matrix resin during pyrolysis and carbonization, although the carbon fibers do not shrink because they have been prepared at around 1300  C. These anisotropic areas grow with increasing HTT (Figure 37c); finally, almost all area of the composite became anisotropic (Figure 37d), and the development of graphitic structure was proved by X-ray diffraction. Carbon fibers with random orientation, such as PANbased carbon fibers, are difficult to be graphitized by the heat treatment under normal pressure, but they can be graphitized when their nanotexture was destroyed by exfoliation [154,156]. In Figure 38, changes in Raman spectrum are shown on a PAN-based carbon fiber from the pristine fiber to that after exfoliation, 900  Cannealing, and then the 3000  C-treatment. The pristine fiber shows relatively strong D-band, although it was already heat-treated at 3000  C, but the exfoliated carbon fiber shows very weak D-band after the heat treatment at 2800  C.

6.2. Microtextures Most particles with planar and axial orientation, such as cokes, carbon fibers, and carbon nanotubes, are also anisotropic and as a consequence their agglomeration in various carbon blocks can create a further variety in microtexture, mainly due to preferred orientation of anisotropic particles. Therefore, in addition to the nanotexture and also graphitization degree of each particle, it is necessary to take into consideration the microtexture formed.

FIGURE 38 Change in Raman spectrum of a PAN-based carbon fiber with exfoliation, 900  C-annealing, and 2800  C-treatment.

54

Handbook of Advanced Ceramics

FIGURE 39

Microtexture formed by carbon fibers.

The microtexture is usually formed during the forming process of bulky carbon materials. In large-sized graphite electrodes for metal refining, for example, the particles of needle-like coke tend to be oriented along the extrusion direction, i.e., the rod axis, during their forming process through extrusion with pitch binder. To prepare carbon fiber reinforced plastics (CFRPs), different microtextures based on the orientation of carbon fibers, some examples of which are shown in Figure 39, have been applied in order to get high strength and high modulus of the composites. Three methods have been employed for realizing the isotropy of carbon materials which fundamentally consist

of anisotropic structural units or crystallites: i) the random aggregation of micrometer-sized particles, even though these particles are anisotropic, ii) the aggregation of spherical particles of carbon, and iii) the random agglomeration of nanometer-sized crystallites in the bulk. The first method is realized in the so-called isotropic high-density graphite blocks where small size coke particles are formed by using isostatic pressing, often called rubber pressing. The second method was carried out by using mesocarbon microbeads. The last method, i.e., aggregation of nanosized carbon layers, was realized in glass-like carbons, which are isotropic, but have a large amount of closed pores and nongraphitizing nature.

(a)

(b)

(c)

(d)

(e)

(e)

FIGURE 40 Optical micrographs of the cross-section of isotropic high-density graphite blocks.

Chapter | 2.1

Advanced Carbon Materials

FIGURE 41

FIGURE 42 pores.

55

Dependences of mechanical properties on average pore area in isotropic high-density graphite blocks.

Appearance of the macropores in the worm-like particles of exfoliated graphite and the distribution histogram of cross-sectional area of

The presence of pores in the carbon blocks influences the bulk properties of the material. In such a case, the microtexture, including the shape and size of pores, has to be taken into account in addition to that due to the orientation of anisotropic particles. Optical micrographs on polished sections of isotropic high-density graphite blocks having different bulk densities from 1.735 to 1.848 g/cm3 are shown in Figure 40 [122]. Although the difference in bulk density looks rather small, a marked difference is easily observed in

the shape, size, and distribution of pores in the cross-section. Different pore parameters of these carbon materials, such as density, average cross-sectional area, roundness, and fractal dimension, were determined with the help of image analysis. Figure 41 shows the plots of mechanical properties (elastic modulus, bending strength, and fracture toughness KIc) versus the average cross-sectional area of pores. A dependence of the mechanical properties of these carbons on pore size is clearly seen.

56

Macropores in exfoliated graphite particles (customary called worm-like particles) can sorb a large amount of heavy oil (more than 80 g per 1 g of exfoliated graphite) [157]. An example of the fractured surface of a worm-like particle of exfoliated graphite is shown in Figure 42a. The histogram for the cross-sectional area of macropores in worm-like particles, which was measured with the aid of image analysis on more than 3000 pores in the fractured surfaces, is shown in Figure 42b [55]. The size and volume of these macropores govern the sorption capacity of heavy oil [157].

REFERENCES [1] Audier M, Coulon M, Oberlin A. Relative crystallographic orientations of carbon and metal in a filamentous catalytic car. Carbon 1980;18:73e6. [2] Augie D, Oberlin M, Oberlin A, Hyvernat P. Microtexture of mesophase spheres as studied by high resolution conventional transmission electron microscopy (CTEM). Carbon 1980;18:337e46. [3] Baird T, Fryer JR, Grant B. Carbon formation on iron and nickel foils by hydrocarbon pyrolysisdreactions at 700  C. Carbon 1974;12:591e602. [4] Bekyarova E, Kaneko K. Structure and physical properties of tailormade Ce, Zr-Doped carbon aerogels. Adv Mater 2000;12:1625e8. [5] Bethune DS, Kiang CH, deVries MS, Gorman G, Savoy R, Vazquez J, et al. Cobalt-catalysed growth of carbon nanotubes with single-atomic-layer walls. Nature 1993;363:605e7. [6] Boehm HP. Carbon from carbon monoxide disproportionation on nickel and iron catalysts: morphological studies and possible growth mechanisms. Carbon 1973;11:587e90. [7] Bokros JC. Deposition, structure, and properties of pyrolytic carbon. In: Chemistry and physics of carbon, vol. 5. Marcel Dekker; 1969. p. 70e81. [8] Bokros JC, LaGrange LD, Schoen FJ. Control of structure of carbon for use in bioengineering. In: Chemistry and physics of carbon, vol. 9. Marcel Dekker; 1972. p. 103e71. [9] Brooks JD, Taylor GH. The formation of some graphitizing carbons. In: Chemistry and physics of carbon, vol. 4. Marcel Dekker; 1968. p. 243. [10] Bundy FP, Hall HT, Strong HM, Wentorf Jr RH. Man-made diamonds. Nature 1955;176:51e5. [11] Che G, Lakshml BB, Martin CR, Fisher ER. Metal-nanoclusterfilled carbon nanotubes: catalytic properties and possible application in electrochemical energy storage and production. Langmuir 1999;15:750e8. [12] Chen X, In-Hwang W, Shimada S, Fujii M, Iwanaga H, Motojima S. High-temperature heat treatment of carbon microcoils obtained by chemical vapor deposition process and their properties. J Mater Res 2000;15:808e14. [13] Dahn JR, Reimers JN, Sleigh AK, Tiedje T. Density of states in graphite from electrochemical measurements on Lix(C1-zBz)6. Phys Rev B 1992;45:3773e7. [14] Dahn JR, Sleigh AK, Shi H, Way BM, Weydanz WJ, Reimers JN, et al. Carbons and graphites as substitutes for the lithium anode. In: Lithium batteries, Industrial chemistry library, vol. 5. Elsevier; 1994. p. 1e47.

Handbook of Advanced Ceramics

[15] Derbyshire F, Jagtoyen M, Thwaites M. Activated carbons production and applications. In: Porosity in carbons. Edward Arnold,; 1995. p. 227e52. [16] Donnet JB, Wang TK, Rebouillat S, Peng JCM. Carbon fibers. 3rd ed. Marcel Dekker; 1998. [17] Dousek FP, Jansta J, Baldrian J. Reactivity of polymeric carbon chains reduced from poly(tetrafluoroethylene). Carbon 1980;18:13e20. [18] Dresselhaus MS, Dresselhaus G, Eklund PC. Science of fullerenes and carbon nanotubes. Academic Press; 1996. [19] Endo M, Koyama T, Hishiyama Y. Structure and properties of graphitized carbon fiber. Jpn J Appl Phys 1974;13:1933e9. [20] Endo M, Shikata M. Growth of vapor-grown carbon fibers using ultra-fine particles of metals. Ouyou Butsuri 1985b;54:507e10 [in Japanese]. [21] Endo M, Oshida K, Kobori K, Takeuchi K, Takahashi K, Dresselhaus MS. Evidence for glide and rotation defects observed in well-ordered graphite fibers. J Mater Res 1995;10:1461e8. [22] Endo M, Kim C, Karaki T, Tamaki T, Nishimura Y, Matthewa S, et al. Structure, Raman scattering, and transport properties of boron-doped graphite. Phys Rev B 1998;58:8991e6. [23] Endo M, Kim YA, Hayashi T, Nishimura K, Matusita T, Miyashita K, et al. Vapor-grown carbon fibers (VGCFs): basic properties and their battery applications. Carbon 2001;39:1287e97. [24] Endo M, Kim YA, Hayashi T, Fukai T, Oshida K, Yanagisawa T, et al. Structural characterization of cup-stacked-type nanofibers with an entirely hollow core. Appl Phys Lett 2002;80:1267e9. [25] Endo M, Kim YA, Hayashi T, Yanagisawa T, Muramatsu H, Ezaka M, et al. Microstructural changes induced in “stacked cup” carbon nanofibers by heat treatment. Carbon 2003;41:1941e7. [26] Enoki T, Suzuki M, Endo M. Graphite intercalation compounds and applications. Oxford Univ. Press; 2003. [27] Feurtes AB, Nevskaia DM, Centeno TA. Carbon composite membranes from Matrimid and Kapton polyimides for gas separation. Micropor Mesopor Mater 1999;33:115e25. [28] Fitzer E, Schaffer W, Yamada S. The formation of glasslike carbon by pyrolysis of polyfurfuryl alcohol and phenolic resin. Carbon 1969;7:643e6. [29] Geim AK, Novoselov KS. The rise of graphene. Nature Mater 2007;6:183e91. [30] Hanzawa Y, Kaneko K, Pekala RW, Dresselhaus MS. Activated carbon aerogels. Langumir 1996;26:6167e9. [31] Hanzawa Y, Kaneko K, Yoshizawa N, Pekala RW, Dresselhaus MS. The pore structure determination of carbon aerogels. Adsorption 1998;4:187e95. [32] Hatori H, Yamada Y, Shiraishi M. Preparation of macroporous carbon films from polyimide by phase inversion method. Carbon 1992;30:303e4. [33] Hatori H, Takagi H, Yamada Y. Gas separation properties of molecular sieving carbon membrane with nanopore channels. Carbon 2004;42:1169e73. [34] Hattori Y, Watanabe Y, Kawasaki S, Okino F, Pradhan BK, Kyotani T, et al. Carbon-alloying of the rear surface of nanotubes by direct fluorination. Carbon 1999;37:1033e8. [35] Hishiyama Y, Inagaki M, Kimura S, Yamada S. Graphitization of carbon fiber/glassy carbon composites. Carbon 1974;12:249e58. [36] Hishiyama Y, Yoshida A, Kaburagi Y, Inagaki M. Formation of pores in carbonized polyimide film Kapton by high temperature heat treatment. Carbon 1992;30:517e9.

Chapter | 2.1

Advanced Carbon Materials

[37] Hishiyama Y, Irumano H, Kaburagi Y. Structure, Raman scattering, and transport properties of boron-doped graphite. Phys Rev B 2001;63:245406. [38] Honda H. Advanced coal liquefaction and pitch-based carbon fiber. TANSO 1983;113:66e78 [in Japanese]. [39] Hulicova D, Sato F, Okabe K, Koishi M, Oya A. An attempt to prepare carbon nanotubes by the spinning of microcapsules. Carbon 2001;39:1438e42. [40] Hulicova D, Hosoi K, Kuroda S-I, Abe H, Oya A. Carbon nanotubes prepared by spinning and carbonizing fine core-shell polymer microspheres. Adv Mater 2002;14:452e5. [41] Hummelen JC, Knight B, Pavlovich J, Gonzalez R, Wudl F. Isolation of the Heterofullerene C59N and its dimer (C59N)2. Science 1995;269:1554e6. [42] Iijima S. Helical microtubules of graphitic carbon. Nature 1991;354:56e8. [43] Iijima S, Ichihashi T. Single-shell carbon nanotubes of 1-nm diameter. Nature 1993;363:603e5. [44] Iijima S, Yudasaka M, Yamada R, Bandow S, Suenaga K, Kokai F, et al. Nano-aggregates of single-walled graphitic carbon nano-horns. Chem Phys Lett 1999;309:165e70. [45] Inagaki M, Ishihara M, Naka S. Mesophase formation during carbonization under pressure. High Temp High Press 1976;8:279e91. [46] Inagaki M, Ishihara M, Naka S. Crack formation in the separated mesophase spherules. Carbon 1976;14:88e9. [47] Inagaki M, Kuroda K, Sakai M. Pressure carbonization of polyethylene-polyvinylchloride mixtures. Carbon 1983;21:231e5. [48] Inagaki M. Microtexture in carbon materials. TANSO 1985; 122:114e21. [49] Inagaki M. Applications of graphite intercalation compounds. J Mater Res 1989;4:1560e8. [50] Inagaki M, Harada S, Sato T, Nakajima T, Horino Y, Morita K. Carbonization of polyimide film “Kapton”. Carbon 1989;27:253e7. [51] Inagaki M, Iwashita N, Hishiyama Y, Kaburagi Y, Yoshida A, Oberlin A, et al. Studies on “mesophase”-pitch-based carbon fibers: part I. Structure and textures. TANSO 1991;147:57e65. [52] Inagaki M, Meyer RA. Stress graphitization. In: Chemistry and physics of carbon, vol. 26. Marcel Dekker; 1999. p. 149e244. [53] Inagaki M, Takeichi T, Hishiyama Y, Oberlin A. High quality graphite films from aromatic polyimides. In: Chemistry and physics of carbon, vol. 26. Marcel Dekker; 1999. p. 245e333. [54] Inagaki M. New carbons e control of structure and functions-. Elsevier; 2000. [55] Inagaki M, Toyoda M, Kang F, Zheng YP, Shen W. Pore structure of exfoliated graphite. A report on a joint research project under the scientific cooperation program between NSFC and JSPS. New Carbon Mater 2003;18:241e9. [56] Inagaki M, Kang F, Toyoda M. Exfoliation of graphite via intercalation compounds. In: Chemistry and physics of carbon, vol. 29. Marcel Dekker; 2004. p. 1e69. [57] Inagaki M, Kaneko K, Nishizawa T. Nanocarbons -recent research in Japan. Carbon 2004;42:1401e17. [58] Inagaki M, Morishita T, Kuno A, Kito T, Hirano M, Suwa T, et al. Carbon foams prepared from polyimide using urethane foam template. Carbon 2004;42:497e502. [59] Inagaki M, Kang F. Carbon materials science and engineering. Tsinghua Univ. Press; 2006.

57

[60]

[61] [62] [63] [64]

[65]

[66]

[67]

[68] [69]

[70]

[71]

[72]

[73]

[74]

[75]

[76]

[77]

[78]

[79]

Inagaki M, Nishikawa T, Oshida K, Fukuyama K, Hatakeyama Y, Nishikawa K. Air oxidation of carbon spheres II. Micropore development. Ads Sci Tech 2006;24:55e64. Inagaki M, Konno H, Tanaike O. Carbon materials for electrochemical capacitors. J Power Source 2010;195:7880e903. Inagaki M, Kim YA, Endo M. Graphene: Preparation and structural perfection. J Mater Chem 2011;21:3280e94. Inagaki M, Orikasa H, Morishita T. Morphology and pore control in carbon materials via templating. RSC Adv 2011;1:1620e40. Ishioka M, Okada T, Matsubara K, Endo M. Formation of vaporgrown carbon fibers in CO-CO2-H2 mixtures, II. Influence of catalyst. Carbon 1992;30:865e6. Ishioka M, Okada T, Matsubara K. Preparation of vapor-grown carbon fibers by floating catalyst method in Linz-Donawitz converter gas: Influence of catalyst size. Carbon 1993;31:699e703. Jang J, Bae J. Fabrication of polymer nanofibers and carbon nanofibers by using a slat-assisted microemulsion polymerization. Angew Chem Int Ed 2004;43:3803e6. Jansta J, Dousek FP, Patzelova V. Low temperature electrochemical preparation of carbon with a high surface area from polytetrafluoroethylene. Carbon 1975;13:377e80. Jarkov SM, Tctarenko YN, Churilov GN. Electron microscopy studies of FCC carbon particles. Carbon 1998;36:595e7. Johnson PJ, Setsuda DJ, Williams RS. Activated carbon for automotive application. In: Carbon materials for advanced technologies. Pergamon; 1999. p. 235e68. Joo SH, Choi SJ, Oh I, Kwak J, Kiu Z, Terasaki O, et al. Ordered nanoporous arrays of carbon supporting high dispersions of platinum nanoparticles. Nature 2001;412:169e72. Jun S, Joo SH, Ryoo R, Kruk M, Jaroniec M, Liu Z, et al. Synthesis of new, nanoporous carbon with hexagonally ordered mesostructure. J Am Chem Soc 2000;122:10712e3. Kamiya K, Mizutani M, Noda T, Inagaki M. Effect of pressure on graphitization of carbon IV. Abrupt graphitization of hard carbon under 5 kbar. Bull Chem Soc Jpn 1968;41:2169e72. Kamo M, Chawanya H, Tanaka T, Sato Y, Setaka N. Growth of diamond on single crystals of tungsten carbide in microwave plasma. Mater Sci Eng A 1988;105/106:535e41. Kato H, Kanazawa Y, Okumura M, Taninaka A, Yokawa T, Shinohara H. Lanthanoid endohedral metallofullerenols for MRI contrast agents. J Am Chem Soc 2003;125:4391e7. Katsuki H, Matsunaga K, Egashira M, Kawasumi S. Formation of carbon fibers from naphthalene on some sulfur-containing substrates. Carbon 1981;19:148e50. Kavan L. Electrochemical carbonization of fluoropolymers. In: Chemistry and physics of carbon, vol. 23. Marcel Dekker; 1991. 69e171. Kodama M, Yamashita J, Soneda Y, Hatori H, Nishimura S, Kamegawa K. Structural characterization and electric double layer capacitance of template carbons. Mater Sci Eng B 2004;108:156e61. Komatsu K, Murata M, Murata Y. Encapsulation of molecular hydrogen in fullerene C60 by organic synthesis. Science 2005;307:238e40. Konno H, Nakahashi T, Inagaki M. State analysis of nitrogen in carbon film derived from polyimide Kapton. Carbon 1997; 35:669e74.

58

[80] Konno H, Sato S, Habazaki H, Inagaki M. Formation of platelet structure carbon nanofilaments by a template method. Carbon 2004;42:2756e9. [81] Kosonen H, Valkama S, Nykaenen A, Toivanen M, Brinke GT, Ruokolainen J, et al. Functional porous structures based on the pyrolysis of cured templates of block copolymer and phenolic resin. Adv Mater 2006;18:201e5. [82] Koyama S, Haniu H, Tanaka S, Takeichi T. Cellgrowing with activated carbon fivers -Significances of porous size. Mol Cryst Liq Cryst 2002;388:173e6. [83] Kraetschmer W, Fostiropoulos K, Huffman DR. The infrared and ultraviolet absorption spectra of laboratory-produced carbon dust: evidence for the presence of the C60 molecule. Chem Phys Lett 1990;170:167e70. [84] Kraetschmer W, Lowell D, Lamb K, Fostiropoulos K, Hoffman DR. Solid C60: a new form of carbon. Nature 1990;347:354e8. [85] Kroto RF, Hearth JR, O’Brien SCO, Curl RF, Smalley RE. C60:Buckminsterfullerene. Nature 1985;318:162e3. [86] Kruk M, Jaroniec M, Ryoo R, Joo SH. Characterization of ordered mesoporous carbons synthesized using MCM-48 silicas as templates. J Phys Chem B 2000;104:7960e8. [87] Kudryavtsev YP, Evsyukov S, Guseva M, Babaev V, Khvostov V. Carbyne e A linear chainlike carbon allotrope. In: Chemistry and physics of carbon, vol. 25. Marcel Dekker; 1996. p. 1e69. [88] Kyotani T, Sonobe N, Tomita A. Formation of highly orientated graphite from polyacrylonitrile by using a two-dimensional space between montmorillonite lamellae. Nature 1988;331:331e3. [89] Kyotani T, Tsai L, Tomita A. Formation of ultrafine carbon tubes by using an anodic aluminum oxide film as a template. Chem Mater 1995;7:1427e8. [90] Kyotani T, Nagai T, Inoue S, Tomita A. Formation of new type of porous carbon by carbonization in zeolite nanochannels. Chem Mater 1997;9:609e15. [91] Kyotani T, Tsai L, Tomita A. Formation of platinum nanorodes and nanoparticles in uniform carbon nanotubes prepared by a template carbonization method. Chem Commun 1997;1997:701e2. [92] Kyotani T, Nakazaki S, Xu WH, Tomita A. Chemical modification of the inner walls of carbon nanotubes by HNO3 oxidation. Carbon 2001;39:782e5. [93] Kyotani T, Ma Z, Tomita A. Template synthesis of novel porous carbons using various types of zeolites. Carbon 2003;41:1451e9. [94] Lee J, Yoon S, Hyeon T, Oh SM, Kim KB. Synthesis of a new mesoporous carbon and its application to electrochemical doublelayer capacitors. Chem Commun 1999:2177e8. [95] Liang C, Hong K, Guiochon GA, Mays JW, Dai S. Synthesis of a large-scale highly ordered porous carbon films by self-assembly of block copolymers. Angew Chem Int Ed 2004;43:5785e9. [96] Lowell CE. Solid solution of boron in graphite. J Am Ceram Soc 1967;50:142e5. [97] Ma ZX, Kyotani T, Tomita A. Preparation of a high surface area microporous carbon having the structural regularity of Y zeolite. Chem Commun 2000:2365e6. [98] Ma ZX, Kyotani T, Liu Z, Terasaki O, Tomita A. Very high surface area microporous carbon with a three-dimensional nanoarray structure: Synthesis and its molecular structure. Chem Mater 2001;13:4413e5. [99] Maeno S. The structure and characteristics of conductive carbon black “KETJENBLACK”. TANSO 2006;222:140 [in Japanese].

Handbook of Advanced Ceramics

[100] Marsh H, Yan DS, O’Grady TM, Wennerberg A. Formation of active carbons from cokes using potassium hydroxide. Carbon 1984;22:603e11. [101] Marsh H, Rodriguez-Reinoso F. Activated carbon. Elsevier; 2006. [102] Matsubara H, Yamaguchi Y, Shioya J, Murakami S. Preparation and properties of graphite grown in vapor phase. Synth Met 1987;18:503e7. [103] Matsuoka K, Yamagishi Y, Yamazaki T, Setoyama N, Tomita A, Kyotani T. Extremely high microporosity and sharp pore size distribution of a large surface area carbon prepared in the nanochannels of zeolite Y. Carbon 2005;43:876e9. [104] Morishita T, Ishihara K, Kato M, Tsumura T, Inagaki M. Mesoporous carbons prepared from mixtures of magnesium citrate with poly(vinyl alcohol). TANSO 2007;226:19e24 [in Japanese]. [105] Morishita T, Ishihara K, Kato M, Tsumura T, Inagaki M. Preparation of a carbon with a 2 nm pore size and of a carbon with a bimodal pore size distribution. Carbon 2007;45:209e11. [106] Morishita T, Wang L, Tsumura T, Toyada M, Konno H, Inagaki M. Pore structure and application of MgO-templated carbons. TANSO 2010;242:60e8. [107] Morishita T, Tsumura T, Toyoda M, Przepio´rski J, Morawski AW, Konno H, et al. A review of the control of pore structure in MgO-templated nanoporous carbons. Carbon 2010; 48:2690e707. [108] Motojima S, Kawaguchi M, Nozaki K, Iwanaga H. Growth of regularly coiled carbon filaments by Ni catalyzed pyrolysis of acetylene, and their morphology and extension characteristics. Appl Phys Lett 1990;56:321e3. [109] Motojima S, Hasegawa I, Asakura S, Ando K, Iwanaga H. Vapor phase preparation of micro-coiled carbon fibers by metal powder catalyzed pyrolysis of acetylene containing a small amount of phosphorus impurity. Carbon 1995;33:1167e73. [110] Murakami M, Watanabe K, Yoshimura S. High-quality pyrographite films. Appl Phys Lett 1986;48:1594e6. [111] Murakami M, Yoshimura S. Highly conductive pyropolymer and high-quality graphite from polyoxadiazole. Synth Met 1987; 18:509e14. [112] Nishikawa K, Fukuyama K, Nishizawa T. Structure change of glass-like carbon with heat treatment, studied by small angle X-Ray scattering: I. Glass-like carbon prepared from phenolic resin. Jpn J Appl Phys 1998;37:6486e91. [113] Noda T, Inagaki M. The melting of glassy carbon. Bull Chem Soc Jpn 1964;37:1709e10. [114] Noda T, Inagaki M, Yamada S. Glass-like carbons. J Non-Cryst Solids 1969;1:285e302. [115] Oberlin A, Endo M, Koyama T. High resolution electron microscope observations of graphitized carbon fibers. Carbon 1976;14:133e5. [116] Oberlin A, Endo M, Koyama T. Filamentous growth of carbon through benzene decomposition. J Cryst Growth 1976;32: 335e49. [117] Oberlin A. High-resolution TEM studies of carbonization and graphitization. In: Chemistry and physics of carbon, vol. 22. Marcel Dekker; 1989. p. 1e143. [118] Ohnishi T, Murase I, Noguchi T, Hirooka M. Highly conductive graphite film prepared from pyrolysis of poly(p-phenylene vinylene). Synth Met 1986;14:207e13.

Chapter | 2.1

Advanced Carbon Materials

[119] Ohta N, Nishi Y, Morishita T, Tojo T, Inagaki M. Carbonization and pore development of carbon films derived from aromatic polyimides. TANSO 2008;233:174e80. [120] Ohta N, Nishi Y, Morishita T, Tojo T, Inagaki M. Preparation of microporous carbon films from fluorinated aromatic polyimides. Carbon 2008;46:1350e7. [121] Orikasa H, Inokuma N, Ittisanronnachai S, Wang X-H, Kitakami O, Kyotani T. Template synthesis of water-dispersible and magnetically responsive carbon nano test tubes. Chem Commun 2008:2215e7. [122] Oshida K, Ekinaga N, Endo M, Inagaki M. Pore analysis of isotropic graphite using image processing of optical micrographs. TANSO 1996;173:142e7 [in Japanese]. [123] Ozaki J, Endo N, Ohizumi W, Igarashi K, Nakahara M, Oya A, et al. Novel preparation method for the production of mesoporous carbon fiber from a polymer blend. Carbon 1997;35:1031e3. [124] Pekala RW, Kong FM. Resorcinol-formaldehyde aerogels and their carbonized derivatives. Polym Preprints 1989;30:221e3. [125] Pekala RW, Alviso CT, Kong FM, Hulsey SS. Aerogels derived from multifunctional organic monomers. J Non-Cryst Solids 1992;145:90e6. [126] Pradhan BK, Toba T, Kyotani T, Tomita. A. Inclusion of crystalline iron oxide nanoparticles in uniform carbon nanotubes prepared by a template carbonization method. Chem Mater 1998;10:2510e5. [127] Pradhan BK, Kyotani T, Tomita A. Nickel nanowires of 4 nm diameter in the cavity of carbon nanotubes. Chem Commun 1999:1317e8. [128] Rodriguez NM. A review of catalytically grown carbon nanofibers. J Mater Res 1993;8:3233e50. [129] Ruoff R. Graphene: Calling all chemists. Nature Nanotech 2008;3:10e1. [130] Ryoo R, Joo SH, Jun S. Synthesis of highly ordered carbon molecular sieves via template-mediated structural transformation. J Phys Chem B 1999;103:7743e6. [131] Saikubo A, Yamada N, Kanda K, Matsui S, Suzuki T, Niihara K, et al. Comprehensive classification of DLC films formed by various methods using NEXAFS measurement. Diam Relat Mater 2008;17:1743e5. [132] Saito R, Dresselhaus G, Dresselhaus MS. Physical properties of carbon nanotubes. Imperial College Press; 1998. [133] Sandou T, Ozaki J, Oya A. Preparation of nanocarbons using polymer blend technique. TANSO 2008;234:244e50. [134] Shioya M, Shinotani K, Takaku A. Carbonization behavior of polyoxadiazole fibers and films. J Mater Sci 1999;34: 6015e25. [135] Shiraishi M. Graphitization of carbon. In: Introduction to carbon materials (revised). Carbon Society of Japan; 1984. p. 29e40 [in Japanese]. [136] Soldano C, Mahmood A, Dujardin E. Production, properties and potential of graphene. Carbon 2010;48:2127e50. [137] Soneda Y, Makino M. Formation and texture of carbon nanofilaments by the catalytic decomposition of CO on stainless-steel plate. Carbon 2000;38:478e80. [138] Sonobe N, Kyotani T, Hishiyama Y, Shiraishi M, Tomita A. Formation of highly oriented graphite from poly(acrylonitrile) prepared between the lamellae of montmorillonite. J Phys Chem 1988;92:7029e34.

59

[139] Takeda A, Yokoyama Y, Ito S, Miyazaki T, Shimotani H, Yakigaya K, et al. Superconductivity of doped Ar@C60. Chem Commun 2006:912e4. [140] Takeichi T, Yamazaki Y, Ito A, Zuo M. Preparation and properties of porous polyimide films prepared by the pyrolysis of poly(urethane-imide) films. J Photopolym Sci Technol 1999;12:203e8. [141] Takeichi T, Yamazaki Y, Fukui T, Matsumoto A, Inagaki M. Prepqaration and characterization of porous carbonized films by the pyrolysis of poly(urethane-imide) films. TANSO 2000:388e94 [in Japanese]. [142] Takeichi T, Yamazaki Y, Zuo M, Ito A, Matsumoto A, Inagaki M. Preparation of porous carbon films by the pyrolysis of poly(urethane-imide) films and their pore characteristics. Carbon 2001;39:257e65. [143] Tamon H, Ishizaki H, Mikami M, Okazaki M. Porous structure of organic and carbon aerogels synthesized by sol-gel polycondensation of resorcinol with formaldehyde. Carbon 1997;35:791e6. [144] Tamon H, Ishizaki H, Yamamoto T, Suzuki T. Preparation of mesoporous carbon by freeze drying. Carbon 1999;37:2049e55. [145] Tanaike O, Yoshizawa N, Hatori H, Yamada Y, Shiraishi S, Oya A. Mesoporous carbon from poly(tetrafluoroethylene) defluorinated by sodium metal. Carbon 2002;40:457e9. [146] Tanaike O, Hatori H, Yamada Y, Shiraishi S, Oya A. Preparation and pore control of highly mesoporous carbon from defluorinated PTFE. Carbon 2003;41:1759e64. [147] Tanaka S, Nishiyama N, Egashira Y, Ueyama K. Synthesis of ordered mesoporous carbons with channel structures from an organic-organic nanocomposite. Chem Commun 2005:2125e7. [148] Tanigaki K, Hirosawa I, Ebbesen TW, Mizuki J, Rsai J-S. Structure and superconductivity of C60 fullerides. J Phys Chem Solids 1993;54:1645e53. [149] Tao Y, Endo M, Inagaki M, Kaneko K. Recent progress in synthesis and applications of nanoporous carbon films. J Mater Chem 2011;21:313e23. [150] Tatsumi K, Iwashita N, Sakaebe H, Shioyama H, Higuchi S, Mabuchi A, et al. The influence of the graphitic structure on the electrochemical characteristics for the anode of secondary lithium batteries. J Electrochem Soc 1995;142:716e20. [151] Tibbetts GG. Lengths of carbon fibers grown from iron catalyst particles in natural gas. J Cryst Growth 1985;73:431e8. [152] Toyoda M, Shimizu A, Iwata H, Inagaki M. Exfoliation of carbon fibers through intercalation compounds synthesized electrochemically. Carbon 2001;39:1697e707. [153] Toyoda M, Katoh H, Inagaki M. Intercalation of nitric acid into carbon fibers. Carbon 2001;39:2231e4. [154] Toyoda M, Kaburagi Y, Yoshida A, Inagaki M. Accelerated graphitization of exfoliated carbon fibers. Carbon 2002;40:628e9. [155] Toyoda M, Sedlacik J, Inagaki M. Intercalation of formic acid into carbon fibers and their exfoliation. Synth Met 2003;130:39e43. [156] Toyoda M, Kaburagi Y, Yoshida A, Inagaki M. Acceleration of graphitization in carbon fibers through exfoliation. Carbon 2004;42:2567e72. [157] Toyoda M, Iwashita N, Inagaki M. Sorption of heavy oils into carbon materials. In: Chemistry and physics of carbon, vol. 30. Taylor & Francis; 2007. p. 177e234. [158] Vogel FL. The electrical conductivity of graphite intercalated with superacid fluorides: experiments with antimony pentafluoride. J Mater Sci 1977;12:982e6.

60

[159] Washiyama M, Sakai M, Inagaki M. Formation of carbon spherules by pressure carbonization -relation to molecular structure of precursor. Carbon 1988;26:303e7. [160] Yamada Y, Imamura K, Kakiyama H, Honda H, Oi S, Fukuda K. Characteristics of meso-carbon microbeads separated from pitch. Carbon 1974;12:307e19. [161] Yamada Y, Tanaike O, Shiraishi S. Preparation of porous carbon by deluorination of PTFE and its application to electric double layer capacitor. TANSO 2004;215:285e94. [162] Yamamoto E, Tansho M, Tomiyama T, Shinohara H. 13C-NMR study on the structure of isolated Sc2@C84 metallofullerene. J Am Chem Soc 1996;118:2293e4.

Handbook of Advanced Ceramics

[163] Yamamoto T, Nishimura T, Suzuki T, Tamon H. Effect of drying conditions on mesoporosity of carbon precursors prepared by solegel polycondensation and freeze drying. Carbon 2001;39:2374e6. [164] Yoshida A, Hishiyama Y. Electron channeling effect on highly oriented graphites e size evaluation and oriented mapping of crystals. J Mater Res 1992;7:1400e5. [165] Yoshida A, Hishiyama Y, Ishioka M, Inagaki M. Polygonization of vapor-grown carbon fibers. TANSO 1995;168:169e75 [in Japanese]. [166] Zhou S, Burger C, Chu B, Sawamura M, Nagahama N, Toganoh M, et al. Spherical bilayer vesicles of fullerene-based surfactants in water: a Laser light scattering study. Science 2001;291:1944e7.

Chapter 2.2

Novel Carbon-Based Nanomaterials: Graphene and Graphitic Nanoribbons E. Gracia-Espino,x F. Lo´pez-Urı´as,* Y.A. Kim,y T. Hayashi,y H. Muramatsu,y M. Endo,y H. Terrones,Z Mauricio Terrones** and Mildred S. Dresselhausyy x * y

Department of Chemistry, and Department of Physics, Umea˚ University, SE-901 87 Umea˚, Sweden, Former Affiliate: Advanced Materials Department, IPICYT (Instituto Potosino de Investigacio´n Cientı´fica y Tecnolo´gica, A.C.), Camino a la Presa San Jose´ 2055, Colonia Loası´, San Luis Potosı´, Mexico, Faculty of Engineering, Shinshu University, 4-17-1 Wakasato, Nagano 380-8553, Japan,

Oak Ridge, Tennessee 37831-6367, United States,

**

Z

Oak Ridge National Laboratory, One Bethel Valley Road,

Department of Physics, Department of Materials Science and Engineering and Materials Research

Institute, The Pennsylvania State University, 104 Davey Lab., University Park, Pennsylvania 16802-6300, United States,

yy

Department of Physics and

Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139-4307, United States

Chapter Outline 1. Introduction 2. Theoretical Studies on PhysicaleChemical Properties of Graphene Nanoribbons 3. Defects in Graphene and Graphene Nanoribbons 4. Synthesis Methods of Graphene and Graphitic Nanoribbons 5. Role of Chemical Doping in Graphene Nanoribbons

61 62 65 68 71

1. INTRODUCTION In 1962, Hanns-Peter Boehm et al. [1] reported the isolation and identification of single graphene sheets, “Thinnest Carbon Films”. Recently, the two-dimensional monolayer crystalline allotrope of carbon, known as graphene, was isolated using the so-called “scotch tape method”, in which an ingenious technique for its observation under an optical microscope was also described [2,3]. The relative ease of the sample preparation using the Novoselov’s method and the fascinating properties of this 2D atomic crystal have stimulated extensive experimental and theoretical studies on the novel linear E(k) graphene carbon electronic structures in monolayer graphene exhibiting sp2 hybridization in bilayer and few-layer graphene. Their related graphitic nanoribbons have subsequently emerged, each with novel and distinguishable characteristic properties. At present, there has been a surge in the number of reviews and dedicated journal issues on the synthesis and properties of graphene [4,5]. Concurrently, new developing routes for the

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00003-4 Copyright Ó 2013 Elsevier Inc. All rights reserved.

6. Experimental Detection of Edge-States in Graphene Nanoribbons 7. Graphene Applications 8. Future Work Acknowledgments References

73 77 83 83 83

synthesis of few graphene layers have emerged. The first one consists of the growth of ultrathin epitaxial graphite films by thermal decomposition on the (0001) surface of hexagonal faces of SiC single crystal [6,7]. The second one synthesizes films with single- to few-layer graphene (from 1 to 12 layers) using an ambient pressure chemical vapor deposition (CVD) method on a polycrystalline Ni substrate [8]. This last method was subsequently improved by Sun Z. et al. [9]; here the graphene layers were grown from different carbon sources, such as polymer films and small molecules deposited on a metal catalysis substrate. Graphitic nanoribbons, which are quasi-1D graphene nanostructures, can exhibit a band gap between the valence and conduction band states. The band gap depends on both the edge morphology and ribbon width [10,11], which is typically a few nanometers, making graphene nanoribbons a very interesting material for potential electronics applications [12]. Graphite nanoribbons less than 10 nanometers wide are expected to be semiconductors, independent of

61

62

their edge patterns [12]. Narrow nanoribbons are thus excellent candidates for use in electronic devices, such as field-effect transistors (FETs), which form the basis of microchips in computers. A thorough exploration of the chemical and mechanical properties of nanoribbons will undoubtedly suggest other applications for these structures, perhaps as sensors, catalysts, and scaffolds for tissue regeneration or components of composite materials. This chapter reviews studies of graphene with special attention given to their related graphitic nanoribbons. Different production methods, characterization techniques, and theoretical calculations for nanoribbons are discussed. In particular, we emphasize the role of defects and chemical doping in the electronic properties of graphene nanoribbons. Results of the conductance of un-doped and doped armchair nanoribbons are also described. Finally, applications and future work on graphene and graphene nanoribbon materials are discussed.

2. THEORETICAL STUDIES ON PHYSICALeCHEMICAL PROPERTIES OF GRAPHENE NANORIBBONS The graphene structure consists of a basic two-dimensional monolayer formed by sp2 hybridized carbon atoms (three coordinated neighboring carbon atoms), which belongs to one of the five 2-D Bravais lattices called the hexagonal (triangular or honeycomb) lattice. It is noteworthy that by piling up graphene layers, in an ordered way with AB interlayer stacking order [13], one can form hexagonal 3-D graphite, which belongs to the P63/mmc (194) space group. Graphene was initially considered as a theoretical building block used to describe the graphite crystal, and to study the formation of carbon nanotubes (rolled graphene sheets), and to predict the electronic properties of sp2 carbons generally. This 2-D atomic (one atom thick) crystal of sp2 hybridized carbon has one dominant fingerprint: a unique electronic structure with a linear E(k) dispersion close to the Fermi level. Such an electronic structure describes massless fermions [3], which result in new physical phenomena. Graphene nanoribbons consist of finite size graphene crystals, thus forming doubly coordinated atoms at the edges, in contrast to the triply coordinated interior atoms. If the width of the ribbon is of the order of nanometers, we have a ribbon-like nanostructure that exhibits properties different from those observed in an infinite size graphene sheet.

2.1. Electronic Properties of Zigzag and Armchair Graphene Nanoribbons Theoretical studies on graphene nanoribbons started with the work of Fujita et al. [10,14]. These authors studied the

Handbook of Advanced Ceramics

electronic states of graphene ribbons with armchair and zigzag edges, and found that graphene ribbons presented different properties depending on the edge shape. The zigzag ribbon shows a remarkably sharp peak in the electronic density of states at the Fermi level, which is not present in infinite 2-D graphene. The authors also found that the singular electronic states arise from the partly flat bands at the Fermi level, whose wave functions are mainly localized on the zigzag edge. They also applied the Hubbard model within the mean-field approximation, finding a stable magnetic order in zigzag nanoribbons. The atomic morphology of zigzag and armchair nanoribbons is depicted in Fig. 1a-b. In order to show the environmental dependence of the local density of states, non-spin-polarized density functional calculations were carried out on a zigzag and armchair nanoribbon (see Figure 1c and d). These calculations do not yield electronic states at the Fermi level for armchair nanoribbons (Figure 1c). However, zigzag nanoribbons do show a high density of electronic states at the Fermi level, with the main contribution coming from atoms located at the zigzag borders (see Figure 1d), as previously predicted by Fujita et al. [10].

2.2. Width and Length Dependence of the Electronic Properties of Graphene Nanoribbons From the first studies on the electronic properties of graphene nanoribbons, within the limitations of the tight binding (TB) approximation, it has been shown that armchair graphene nanoribbons (A-GNR) exhibit semiconducting or semimetallic behavior depending on the ribbon’s width. It was also determined that they do exhibit a metallic behavior for N ¼ 3p 1, where p is an integer (N is the number of zigzag chains perpendicular to the grown axis), and semiconducting behavior otherwise [10]. DFT calculations preformed by Son et al. [15] revealed that hydrogen passivated armchair and zigzag GNRs always have nonzero and direct band gaps. They also found that A-GNR can be grouped into three different families: NA ¼ 3p, NA ¼ 3p þ 1 and NA ¼ 3p þ 2, exhibiting in all cases different electronic band gaps (Eg3p > Eg3pþ1> Eg3pþ2). The band gaps for each family decrease with increasing p and eventually tend to zero as p / N. These results have been proved to have the same trend by using first-principles calculations based on many-body perturbation theory, such as the GW approximation [16]. Compared with previous TB and DFT studies, the quasiparticle band gaps calculated with the GW approximation show significant self-energy corrections for both armchair and zigzag nanoribbons; in all cases the band gaps were increased by self-energy corrections.

Chapter | 2.2

Novel Carbon-Based Nanomaterials: Graphene and Graphitic Nanoribbons

63

FIGURE 1 (a) Armchair and (b) zigzag chiralities in graphene nanoribbons. The armchair and zigzag nanoribbons are obtained by cutting an infinite graphene sheet along the chiral angle directions 0 and 30 , respectively. The geometric parameters are defined by the width and the growth directions. In both cases, the unit cell is indicated. The different borders (edges) of the nanoribbons play a crucial role in determining the ribbon properties, especially the electronic properties for GNRs of small width. Density functional calculations of the local density of states (LDOS) of three different atoms (A, B, and C) along the structure of the armchair and zigzag nanoribbon are shown in (c) and (d), respectively. The position of the chosen atoms is indicated in the corresponding structures. The atoms type A (double-coordinated) and B (triple-coordinated) are edge atoms, and C is an interior atom. For color version of this figure, the reader is referred to the online version of this book.

For Z-GNRs, local spin density approximation (LSDA) and GW calculations have both shown an antiferromagnetic insulating ground state with ferromagnetic coupling at each zigzag edge, and antiferromagnetic coupling across the ribbon. Two notable characteristics are found in the electronic structure of Z-GNRs: (1) the top of the valence band and the bottom of the conduction band are composed of mainly edge states and (2) the spineorbit interaction introduces a finite band gap in the Z-GNRs [16]. A band gap engineering investigation by M. Y. Han et al. [17] in lithographically patterned graphene nanoribbons demonstrated that the energy gap scales inversely with the ribbon width. Inspired by this work, Sols et al. [18] have revealed that the existence of an energy gap in graphene nanoribbons may be understood in terms of Coulomb blockade effects. A systematic study on the electronic properties of graphene nanoribbons was carried out by V. Barone et al. [19]. Based on DFT calculations, these authors analyzed the band gaps and the relative stabilities of semiconducting carbon nanoribbons with and without hydrogen termination, and widths up to 3 nm. They are able to predict the nanoribbon properties using an extrapolated inverse power law, that is, nanoribbons with band gaps on the order of the band gaps of bulk Ge and InN should have a width between 2 and 3 nm. If a larger band gap material is needed, with a gap comparable to that of Si, InP, or GaAs, the width of the carbon nanoribbons should be between 1 and 2 nm. Zero-dimensional graphitic systems (i.e. finite width and length) have also been studied by different groups.

Using DFT calculations, Shemella et al. [20] found that in addition to quantum confinement along the width of the ribbon, finite size effects emerge along the length of the ribbon, which modifies the electronic states of A-GNRs with zigzag terminated (length confined) edges. The finite size Z-GNRs do not exhibit length dependences since the electronic states near the Fermi level come from the states located along the widths of the ribbon. On the same theory level, O. Hod et al. [21] also determined the effect of finite size on the electronic properties of long graphene nanoribbons, finding that most of the physical features appearing in the density of states of an infinite GNR are recovered at a length of 40 nm.

2.3. Multilayered Graphene Nanoribbons Developing a fully general tight binding Hamiltonian from first principles, Finkenstadt et al. [22] studied the effect of the thickness (i.e., more than one layer) on the electronic structure of graphene nanoribbons. The authors found that layered nanoribbons of graphene approach the familiar band structure of graphite with increasing ribbon width, except for zigzag edge states near the Fermi level. In this context, K-T. Lam et al. [23] investigated armchair bilayer graphene nanoribbons in the DFT framework. Their results show that the band gap (Eg) of a bilayer A-GNR, in general, is smaller than that of a monolayer A-GNR and they show that bilayer A-GNRs exhibit two distinct groups, metals and semiconductors, whereas monolayer A-GNRs display

64

purely semiconducting behavior. Moreover, the band gap Eg of bilayer A-GNRs is highly sensitive to the interplanar distance. Inspired by recent experimental reports, indicating that joule heating can atomically sharpen the edges of CVD grown graphitic nanoribbons [24,25], Cruz-Silva et al. [26] using quantum molecular dynamics (MD) calculations, studied the role of defects and interstitial atoms in the coalescence of graphene edges. The authors addressed the issue of the effect of electron irradiation on graphene and explained why loops (or coalesced adjacent graphene edges) do not form when fast electrons first irradiate the sample before joule heating is introduced. These authors simulated the electron irradiation by putting several types of defects in a bilayer graphene nanoribbon and determined that both vacancies and interstitials are key for keeping graphene layers parallel and for avoiding bilayer edge coalescence (or loop formation).

2.4. Magnetism and Half-Metallicity on Zigzag Nanoribbons Since the first studies by Nakada and Fujita et al. [10,14], it was realized that edge states in Z-GNRs could exhibit interesting magnetic properties. By studying the spin structure at edges, Lee et al. [27] found that magnetic order could be established at edges of graphitic fragments. These authors concluded that zigzag nanoribbons exhibit

FIGURE 2 Spin-polarized electronic density of states (DOS) and the two different spin configurations in zigzag graphene nanoribbons (Z-GNRs). (a) The most stable spin configuration with ferromagnetically ordered spins along each edge but with an opposite spin direction between the spins on opposite edges (FM-A) and (b) the corresponding DOS relation to (a). (c) Spin configuration with ferromagnetically ordered spins at both edges with the same spin direction (FM-F) and (d) the corresponding DOS. In both cases, the maximal magnitude of the magnetic moment is 1.2 mB, where mB is the Bohr magneton. For this case, the energy difference between the two FM-A and FM-F edge structures is small (DE ¼ 4.5 meV), as found in results obtained by DFT calculations. For color version of this figure, the reader is referred to the online version of this book.

Handbook of Advanced Ceramics

ferromagnetically ordered spins at each edge but with the opposite spin direction (FM-A). However, the spin configuration corresponding to the ferromagnetically ordered spins at each edge with the same spin direction (FM-F) is quasi degenerate with an FM-A spin configuration at the opposite edges. Figure 2 illustrate the spin-polarized density of states and the FM-A and FM-F spin configurations of zigzag carbon nanoribbons. When the width of the nanoribbon is increased, the FM-A and FM-F spin configurations become energetically degenerate. Wimmer et al. found that an ideal GNR (with symmetric edges) has zero spin conductance but has a nonzero spin Hall conductance. Moreover, only GNRs with imperfect edges (asymmetrically shaped edges) exhibit a nonzero spin conductance [28]. It has been demonstrated that applying external electric fields across the zigzag nanoribbons results in the appearance of half-metallicity [29]. In this case, states with only one type of spin (up or down) are present at the Fermi level, and such a handle to control spin ordering could be important for the design and fabrication of spin-valve devices. T. Nikolaos et al. [30] reported the observation of spin transport on a micrometerscale distance in a single graphene layer. They measured the spin transport in a four-terminal spin valve device, finding that the spin transport is relatively insensitive to the temperature and that the observed spin relaxation length, and relaxation times in their device, are limited by intrinsic impurity scattering and not by the intrinsic properties of graphene.

Chapter | 2.2

Novel Carbon-Based Nanomaterials: Graphene and Graphitic Nanoribbons

65

3. DEFECTS IN GRAPHENE AND GRAPHENE NANORIBBONS 3.1. Classification of Defects In general, defects within graphene-like structures can now be classified, isolated, and investigated in detail. Defects play a crucial role in the properties of crystals and diverse nanostructures, and the sp2 carbon nanostructures are no exception. In 1996 a systematic compilation of defects was made by Terrones and Terrones [31]. Graphene-like systems are so versatile that they can accommodate different kinds of defects, thereby altering the basic structure of the host material and its physico-chemical properties. In particular, defects could change the atomic arrangement of carbon atoms in different ways: preserving the topology (not changing the total curvature), introducing curvature, altering the detailed sp2 carbon connectivity, and accepting foreign atoms into the lattice (doping). According to this classification, defects can be categorized into five groups: (1) structural defects [32e35], (2) pentagoneheptagon pairs and bond rotations generating grain boundaries or extended lines [33e40], (3) doping-induced defects [41e44], (4) non-sp2-carbon defects [45,46], and (5) high-strain folding of graphene sheets [47e51]. It is still a challenge to study defects at the atomic scale, control their location, and pinpoint their particular properties in order to tailor the electronic properties of graphene-like systems (see Fig. 3). Recent experimental evidence, at the atomic scale, of some types of defects in graphene obtained by different research groups [52e55] is depicted in Fig. 4. 1. Structural defects are related to imperfections that distort the hexagonal carbon lattice by introducing curvature, but preserving the sp2 connectivity of the carbon atoms (carbon atoms connected to three neighbors). These defects are caused by the presence of non-hexagonal rings (e.g., pentagons, heptagons, or octagons) surrounded by hexagonal rings. For example, if a single pentagon or a few pentagons are embedded into the graphene lattice, nanocones with different apex angles are obtained [56e60]. A 30 angle in a single-walled carbon nanotube (SWCNT) could also be explained by the presence of a pentagon on one side of the tube and a heptagon on the opposite side [61]. The reactivity of pentagons, heptagons, or octagons with specific acceptor or donor molecules still has to be determined from theoretical and experimental standpoints. As will be explained in the following section, the presence of adjacent pentagons and heptagons in the same proportion could produce local alterations without changing the total curvature, since the positive curvature from the pentagons is canceled by the negative curvature of the heptagons, and the morphological perturbation of a 5e7 pair is only local (see Fig. 3).

FIGURE 3 Defects in graphene and graphene nanoribbons. (a) Vacancies (single and double vacancies), (b) StoneeThrowereWales 5e7 pair defects, (c) two examples of graphene antidots, (d) a zigzag-armchair grain boundary by 5e7 carbon rings, (e) vacancies and bond rotation, and (f) hexagonal Haeckelite structure with 5-6-7 carbons rings. For color version of this figure, the reader is referred to the online version of this book.

2. PentagoneHeptagon pairs, bond rotations (grain boundaries and extended lines of defects). The presence of pentagoneheptagon pairs in graphene distorts the lattice locally, not changing the total curvature of the sheet. In particular, these defects could be 5-7-7-5 pairs embedded in the hexagonal network (also known as StoneeThrowereWales (STW-type) defects [35,36]) that could be created by rotating a carbonecarbon bond by 90 within four neighboring hexagons, thus resulting in the generation of two pentagons and two heptagons while preserving sp2 connectivity [61,62]. 2-D planar graphene-like systems containing pentagons, hexagons, and heptagons, called pentaheptite or Haeckelites, have been proposed and found to be metallic in theoretical studies [63,64]. Nanoribbons constructed from Haeckelites could be considered as a new hypothetical nano-architecture with fascinating properties that could be applied in electronics. Isolated pentagoneheptagon pairs could also be introduced to form a grain boundary or an extended line of defects in graphene [37e40,65] or in a graphitic nanoribbon, thus changing their edge termination and electronic properties and forming

66

Handbook of Advanced Ceramics

FIGURE 4 (a) TEM image of a Stonee ThrowereWales defect (55-77), formed by rotating a carbonecarbon bond by 90 and (b) a single vacancy as seen in an experimental TEM image [52]. (c) Top image: Defect structure and superimposed defect model of 5-8-5 defects. Bottom image: Line defect with image profile in the direction perpendicular to the wire (inset). The brighter area surrounding the defect originates from the states with wave functions localized at the defect [53]. (d) Top images: TEM image of a di-vacancy (V2(5-8-5)) and (e) top image V2(5555-6-7777) di-vacancy; in the bottom images of the defect models are superimposed [54]. (f) Two grains (bottom left, top right images) intersect with a 27 relative rotation. An aperiodic line of defects stitches the two grains together. The image from (g) with the pentagons (blue), heptagons (red), and distorted hexagons (green) of the grain boundary outlined [55]. For interpretation of the references to color in this figure legend, the reader is referred to the online version of this book.

a hybrid graphitic nanoribbon. These particular hybrid nanoribbons exhibit half-metallicity in the absence of an electric field, and could be used to transport electrons with one type of spin; such nanoribbons could be a step forward in designing new spintronic devices [37]. It is noteworthy that the electronic and chemical properties of these 5e7 or 5-7-7-5 pairs are different from the structural defects (see above), and their reactivity and detection need to be investigated theoretically and experimentally as specific defect types. Zettl and co-workers were able to observe 5e7 and STW (StoneeThrowereWales) defects directly on isolated graphene surfaces using aberration-corrected transmission electron microscopy (see supplementary material of Reference [65]). Figure 4f shows such a defect.

3. Doping-induced defects, arising from substitutional non-carbon atoms, can also be embedded into a sp2 carbon lattice. In this case, it has been demonstrated that nitrogen and boron atoms can be introduced into the hexagonal sp2 hybridized carbon lattice. With both N and B dopants, the chemical reactivity of the graphene surface increases, in one case due to the fact that N has one electron more than C (making the material n-type), and in the other because B has one electron less than C and the material becomes p-type. Therefore, these types of defects could be used to tune the type of conduction in graphene-like materials, ranging from n-type transport (substitutional nitrogen doping) to p-type conduction (substitutional boron atoms in the lattice) [41]. Recent studies have demonstrated that other elements such as P, S, and Si and paired dopants such as P-N

Chapter | 2.2

Novel Carbon-Based Nanomaterials: Graphene and Graphitic Nanoribbons

could also be introduced in the hexagonal lattice of carbon nanotubes [42e44,66]. Therefore, the insertion of non-carbon atoms into graphene-like materials could tailor the chemical reactivity and electronic transport of the layers and further work along this line needs to be carried out [67]. 4. Non-sp2 carbon defects caused by the presence of highly reactive carbons, such as dangling bonds, carbon chains, interstitials (free atoms trapped between SWCNTs or between graphene sheets), edges (in open nanotubes or in graphene nanoribbons), adatoms, and vacancies. These defects are usually observed in an HRTEM, when the adsorbed atoms on these reactive sites are removed by means of the energy of the electron beam. It has been demonstrated that the creation of such defects could promote the formation of covalent nanotube junctions [45] and could also trigger the coalescence of nanotubes [46]. However, for nanoribbons, the dynamics of these defects has not been fully investigated either as a class nor have the differences between members of the class been investigated for nanoribbons. Further research is required to move forward in both of these directions. 5. High-strain folding of graphene sheets (loop formation) can be induced by the thermal annealing of two adjacent graphene layers. These types of “loops” have often been observed through TEM measurements after open edges are thermally annealed at constant temperatures above ~1500  C [47e50], but the chemical reactivity and electronic properties of the loops or the loop-formation process have not yet been investigated in detail. In fullerene (or carbon cage) chemistry, the main type of defect is structural, which deals with the presence of pentagonal rings that induce a highly curved fragment caused by the mixture of sp2 and sp3 orbitals. In nanotube chemistry, these defects mainly occur on the caps (where pentagons are located). Narrow or highly curved nanotubes (usually 0.1 MeV) a,b: Transient irradiation creep parameters K: Steady irradiation creep coefficient E0: Longitudinal elastic modulus before irradiation (GPa). Figure 10 shows the steady-state irradiation creep coefficient of IG-110 as a function of irradiation temperature [14].

Handbook of Advanced Ceramics

2.0

2.0

(a)

Rel. tensile strength change

Rel. compressive strength change

120

1.8 1.6 1.4 1.2 IG-110

1.0 0.8

0

1

2

3

Fast Neutron Fluence (1025n/m2, E>0.18MeV)

(b) 1.8 1.6 1.4 1.2 1.0 0.8

IG-110 0

1

2

Fast Neutron Fluence (1025n/m2, E>0.18MeV)

Relative strength = (Strength after irradiation)/(Mean strength before irradiation) FIGURE 9 Changes of (a) compressive and (b) tensile strengths of IG-110 graphite by fast neutron irradiation [14].

FIGURE 10 Steady-state irradiation creep coefficient of IG-110 as a function of irradiation temperature [14].

3. R&DS FOR VHTR DEVELOPMENTS 3.1. Use of Graphite Components for Long Service Life Mechanical and thermo stability of in-core graphite components is important during their service life in reactor operation. The service life of the core graphite components in the HTTR is about 1.5  1025 n/m2 (E > 0.18 MeV) in the fast neutron fluence. It is possible to extend the service life for commercial reactors based on the in-core design. The service life of the core graphite components in the GTHTR300C [13] is evaluated as the fast neutron fluence of about 5  1025 n/m2 (E > 0.18 MeV). For the prismatic block type reactor, graphite components are replaceable. The fuel graphite blocks in the GTHTR-300C will be replaced at

every four-year interval during the reactor operation. For Pebble bed-type reactors, the maximum fast neutron fluence may be in the order of 1026 n/m2 (E > 0.1 MeV) in case that core graphite components are not replaced throughout the reactor lifetime. It is important to design the graphite components to bear the severe fast neutron irradiation. Since the in-core graphite components of the VHTR are used subject to severe condition, it is important to assess their structural integrity during the service life. Residual stress is accumulated in the graphite components by irradiation-induced dimensional change due to thermal gradient and neutron flux distribution within the component during the reactor operation. The lifetime of the graphite component is determined by the comparison of the residual stress and the stress limit. The allowable service life is determined so that the irradiation-induced residual stress does not exceed the limit [25]. For the stress limit of the graphite components in the design standard of the graphite components in the HTTR, it is set as one-third of the specified ultimate minimum strength Su for the core graphite components, which are replaceable components [15]. The graphite grade with high strength will give the high stress limit. It is important to use a graphite grade with high strength to use in the long service life. In this point, IG-110 graphite used in the HTTR has enough high strength. On the other hand, the residual stress is evaluated by viscoelastic analysis taking uncertainty of in-core temperature and neutron irradiation condition into account [26]. The evaluated residual stress, hence, contains some safety margin. It is possible to extend the allowable components lifetime by reducing the margin with reasonable way, if we can measure the residual stress directly. JAEA is developing the nondestructive evaluation technique for the residual stress of graphite components by using a microindentation method. It uses the change of the indentation behavior loading/unloading at the residual stress condition

Chapter | 2.5

Nuclear Graphite

of graphite. Since the indentation behavior is affected by the residual stress condition, it is possible to evaluate the stress from microindentation behavior. The details of this study are shown in references [27,28].

3.2. Standard for the Graphite Components of VHTR For the design of the graphite components in the HTTR, the graphite structural design code for the HTTR [14] and inspection standard of graphite for the HTTR [21] were applied. It was so because the general standard system for the HTGR had not been established at that time. Recently, to prepare the general standard system, a “special committee on research on preparation for codes for graphite components in HTGR” at Atomic Energy Society of Japan (AESJ) had established draft of standard for graphite components in HTGR (including VHTR) in March 2009 [29]. In the United States, a subgroup in the American Society of Mechanical Engineers (ASME) is working to establish a new design code for graphite core components, which can be applicable to gas-cooled reactors including VHTR. For the establishment of the standard of the AESJ, the standards for the graphite components of HTTR and ASME were reviewed. The established standard is based on the HTTR graphite design code [14] and the inspection standard of the graphite components for the HTTR [21]. The established “draft of standard for graphite core components in high temperature gas-cooled reactor” is composed of (1) general rules, (2) design standard, (3) material and product standards, (4) in-service inspection and maintenance standard, and (5) appendix and explanation. The graphite components are classified into three categories (A, B, and C) in the standpoints of safety functions and replaceability. In the design section, the stress limits are determined by deterministic evaluation in principle. Alternatively, it is possible to determine the stress limits by probabilistic evaluation with an appropriate fracture probability. The general requirements for the in-service inspection and maintenance standard are determined. When a defect larger than a design assumption is found in graphite components by the in-service inspection, the cause shall be investigated, and the propriety of continuous use shall be judged by application of fracture mechanics, etc. As an appendix of the design standard, the graphical expressions of material property data of IG-110 graphite as a function of fast neutron fluence are expressed. The graphical expressions were determined through the intrapolation and extrapolation of the irradiated data. The irradiation data for historical grades H-451 and ATR-2E were used for the evaluation of the intrapolation and extrapolation of irradiation data for IG-110. The details of the intrapolation and extrapolation method were reported in references [24,30].

121

3.3. Development of Control Rod Made of C/C Composite The control rods of HTGRs are used at severe temperature condition. At the reactor scram, control rods are immediately installed into the reactor core with high temperature. For the HTTR, ferritic superalloy Alloy 800H is used for the metal parts of the control rod [8]. Its maximum allowable temperature to be used repeatedly after the scrams is 900  C. On the other hand, for the VHTR, thermo condition of the core will be more severe than that of the HTTR. It is important to develop the control rod made of heat-resistant ceramics composite-material substitute for metallic materials. It is one of the major subjects for the VHTR development. The heat-resistant ceramic composite, C/C composite, and SiC/SiC composite are major candidate materials for this purpose. Figure 11 shows the scheme of R&Ds at JAEA for the control rod made of C/C and SiC/SiC composites [31]. The R&D items are categorized in the following phases: (1) database establishment, (2) design and fabrication, and (3) demonstration test in the HTTR. For the database establishment in phase (1), the data on material properties at un-irradiated and irradiated conditions are necessary. JAEA focuses especially on a twodimensional C/C composite (2D-C/C composite) that has the layer structure of laminae composed of fibers and matrix [32]. Since the 2D-C/C composite has great anisotropy in thermal and mechanical properties in parallel and vertical to lamina directions, it is important to take the anisotropy into account for design of the components. Neutron irradiation effects on the properties are also anisotropic and important. A PAN-based and plain-woven 2D-C/C composite, CX-270G grade, is one of the candidate materials for the control rod application. It is graphitized at

(1) Database (C/C and SiC/SiC) Un-irradiation/ Irradiation experiments

International collaboration

strength, elastic modulus, CTE, thermal conductivity, etc.

JAEA (2) Design Control rod structure Design methodology

(3) Demonstration by HTTR

Graphite irradiation database

Evaluation on irradiation effects for graphitized C/C composite

Large component irradiation at primary helium-gas coolant with temperature gradient

International collaboration

FIGURE 11 Scheme of R&Ds on application of C/C composite to control rod [31]. For color version of this figure, the reader is referred to the online version of this book.

122

Handbook of Advanced Ceramics

TABLE 3 Typical Material Properties of CX-270G and IG-110 Graphite [32]. Bulk density (Mg/m3) CX-270G (2D-C/C) IG-110 (graphite)

(pl) (pp)

Bending strength (MPa)

1.63

Tensile strength (MPa)

Elastic modulus (GPa)

Thermal expansion coefficient (10 6/K)

167

81

0.2 10.8

25

10

4.6

133 1.77

39

pl: parallel lamina direction, pp: perpendicular lamina direction

2800  C in its manufacturing process. Table 3 shows its typical material properties in comparison with fine-grained isotropic IG-110 graphite [32]. For the design and fabrication in phase (2), it is necessary to determine the control rod structure. Connecting method for the component parts is one of the key technologies for the fabrication of the control rod. Appropriate design methodology, e.g. fracture theory, for the fiber-reinforced ceramics has not been fully developed yet. It is also an important issue for the C/C and also SiC/SiC composite application to the control rod for the VHTR. For the demonstration test by the HTTR in phase (3), it is the final stage for the composite control rod development. The large scale component of C/C and SiC/SiC composite can be irradiated in the HTTR, and the performance of it can be confirmed.

4. CONCLUDING REMARKS HTGR/VHTR can provide high temperature of helium gas about 950  C at the reactor outlet, which is applicable not only for power generation by gas and steam turbine systems but also for the process heat source in a hydrogen production system. Graphite and carbon materials are used for in-core structural components of the reactors, such as fuel blocks, reflectors, and core support structures. The in-core components are categorized based on their functions and replaceability. Development of general design code for the in-core components are been carrying out in Japan and the United States for the commercial HTGR/VHTR. Since neutron irradiation effect, depending on graphite grade, is one of the important subjects to be considered in the reactor design, it is necessary to establish the irradiation database for the candidate graphite grades. For the control rod of the VHTR, it is important to develop the control rod made of heat-resistant ceramics-composite materials, C/C and SiC/SiC composites, to substitute for metallic materials in order to use high temperature condition in the core. JAEA is carrying out the R&Ds for the ceramics-

composite control rod, and it is possible to demonstrate it by using the HTTR in the future.

REFERENCES [1] Burchell TD. Fission reactor application of carbon. In: Carbon materials for advanced technologies. Pergamon; 1999. p. 429e84. [2] Matsui K, Shiozawa S, Ogawa M, Yan XL. Present status of HTGR development in Japan. In: Proc. 16th Pacific Basin Nuclear Conference (16PBNC), Aomori, Japan, Oct. 13e18, 2008, PaperID P16P1355. [3] Snead LL. Fusion energy application. In: Carbon Materials for Advanced Technologies. Pergamon; 1999. p. 389e424. [4] Fujita I, Amemiya A, Hino T, Yamashina T, Akiba M, Ando T, et al. Surface analysis of JT-60 graphite divertor tiles. J Nucl Mater 1992;196e198:168e73. [5] Ise H, Satoh S, Yamazaki S, Akiba M, Nakamura K, Suzuki S, et al. Development of fabrication technologies for the ITER divertor. Fusion Eng Des 1998;39e40:513e9. [6] Sakurai S, Masaki K, Shibama YK, Tamai H, Matsukawa M. Design study of plasma-facing components for JT-60SA. Fusion Eng Des 2007;82:1767e73. [7] Raffray AR, Nygren R, Whyte DG, Abdel-Khalik S, Doerner R, Escourbiac F, et al. High heat flux components e readiness to proceed from near term fusion systems to power plants. Fusion Eng Des 2010;85:93e108. [8] Saito S, Tanaka T, Sudo Y, Baba O, Shindo M, Shiozawa S, et al. Design of high temperature engineering test reactor (HTTR). JAERI 1332. Japan Atomic Energy Research Institute; 1994. [9] Fujikawa S, Hayashi H, Nakazawa T, Kawasaki K, Iyoku T, Nakagawa S, et al. Achievement of reactor outlet coolant temperature of 950  C in HTTR. J Nucl Sci Technol 2004;41:1245. [10] Takamatsu K, Ueta S, Sumita J, Goto M, Hamamoto S, Tochio D, et al. High temperature continuous operation in the HTTR (HP-11) e summary of the test results in the high temperature operation mode e, JAEA-Research 2010-038. Japan Atomic Energy Agency; 2010 [in Japanese]. [11] Hoffelner W, et al. International R&D Roadmap for Generation IVVHTR Materials Research. In: Proc. of ICAPP’05, Paper 5206, 2005, Seoul, Korea. [12] OECD Nuclear Energy Agency. GEN IV International Forum 2009 Annual report, 2009. [13] Kunitomi K, Yan X, Nishihara T, Sakaba N, Mori T. JAEA’s VHTR for hydrogen and electricity cogeneration: GTHTR300C. Nucl Eng Technol 2007;39(1):9e20.

Chapter | 2.5

Nuclear Graphite

[14] HTTR Designing Laboratory, Department of High-Temperature Engineering, Department of Fuel and Materials Research. Graphite structural design code for the high temperature engineering test reactor, JAERI-M 89e006. Japan Atomic Energy Research Institute; 1989 [in Japanese]. [15] Ishihara M, Sumita J, Shibata T, Iyoku T, Oku T. Principle design and data of graphite components. Nucl Eng Des 2004;233(1e3): 251e60. [16] Ishihara M, Iyoku T, Toyota J, Sato S, Shiozawa S. An explanation of design data of the graphite structural design code for core components of high temperature engineering test reactor. JAERI-M 91e153. Japan Atomic Energy Research Institute; 1991 [in Japanese]. [17] Snead LL, Burchell TD, Katoh Y. Swelling of nuclear graphite and high quality carbon fiber composite under very high irradiation temperature. J Nucl Mater 2008;381:55e61. [18] Burchell TD. Irradiation induced creep behavior of H-451 graphite. J Nucl Mater 2008;381:46e54. [19] Haag G. Properties of ATR-2E graphite and property changes due to fast neutron irradiation. Ju¨l-4183. Forschungszentrum Ju¨lich; 2005. [20] Vreeling JA, Wouters O, van der Laan JG. Graphite irradiation testing for HTR technology at the high flux reactor in Petten. J Nucl Mater 2008;381:68e75. [21] Toyota J, Iyoku T, Ishihara M, Takikawa N, Shiozawa S. An inspection standard of graphite for the high temperature engineering test reactor. JAERI-M 91e102. Japan Atomic Energy Research Institute; 1991 [in Japanese]. [22] Simmons JHW. Radiation damage in graphite. Pergamon Press; 1965. [23] Kelly BT. Physics of graphite. Applied Science Publisher; 1981. [24] Kunimoto E, Shibata T, Shimazaki Y, Eto M, Shiozawa S, Sawa K, et al. Expansion of irradiation data by interpolation and extrapolation for design of graphite components in high temperature gas-cooled

123

[25] [26] [27]

[28]

[29]

[30]

[31]

[32]

reactor e evaluation on IG-110 graphite irradiation data for component design e, JAEA-Research 2009-008. Japan Atomic Energy Agency; 2009 [in Japanese]. Oku T, Ishihara M. Lifetime evaluation of graphite components for HTGRs. Nucl Eng Des 2004;227(2):209e17. Iyoku T, Ueta S, Sumita J, Umeda M, Ishihara M. Design of core components. Nucl Eng Des 2004;233:71e9. Shibata T, Hanawa S, Sumita J, Tada T, Sawa K, Iyoku T. Nondestructive evaluation on mechanical properties of nuclear graphite with porous structure. In: Proc. GLOBAL 2005, Tsukuba, Japan, Paper No. 360, 2005. Shibata T, Sumita J, Tada T, Sawa K. Development of nondestructive evaluation method for degradation of HTGR graphite components. J Nucl Mater 2008;381:204e9. Shibata T, Eto M, Kunimoto E, Shiozawa S, Sawa K, Oku T, et al. Draft of standard of graphite core components in high temperature gas-cooled reactor, JAEA-Research 2009-042. Japan Atomic Energy Agency; 2009. Shibata T, Kunimoto E, Eto M, Shiozawa S, Sawa K, Oku T, et al. Interpolation and extrapolation method to analyze irradiationinduced dimensional change data of graphite for design of core components in very high temperature reactor (VHTR). J Nucl Sci Technol 2010;47(7):591e8. Shibata T, Sumita J, Makita T, Takagi T, Kunimoto E, Sawa K. Research and development on C/C composite for very high temperature reactor (VHTR) application. Ceram Eng Sci Proc (CESP) 2009;30(10):19e32. Shibata T, Sumita J, Kunimoto E, Sawa K. Characterization of 2DC/C composite for application to in-core structure of very high temperature reactor (VHTR). In: Proc. of Carbon 2008, Nagano, Japan, 2008, P0370.

Chapter 2.6

Carbon Materials for Si Semiconductor Manufacturing Osamu Okada Product Planning Department, Toyo Tanso Co., Ltd., 2181-2 Nakahime, Ohnohara-cho, Kanonji, Kagawa 769-1612, Japan

Chapter Outline 1. 2. 3. 4. 5.

Introduction Silicon Semiconductor Manufacturing Processes Manufacturing Processes for Polycrystalline Silicon Manufacturing Process for Monocrystalline Silicon Processes of Machining a Silicon Single-Crystal Ingot

125 125 128 128 129

1. INTRODUCTION We use a number of electronic devices in our daily life, benefiting from them, and these electronic devices use a great many number of semiconductors; in other words, it may safely be said that they support our life. In manufacturing processes of semiconductors that exist close to us, as we have just mentioned above, a number of carbon parts are used. In this article, carbon parts used in various manufacturing processes of silicon semiconductors are described. The carbon mainly used in current silicon semiconductors manufacturing is isotropic graphite. The designation of isotropic graphite comes from the fact that the properties of a graphite block manufactured is about the same in all directions, and isotropic graphite was produced on a commercial basis for the first time in the world by Toyo Tanso (Japan) in 1974. Having excellent properties of high density and high strength in comparison with conventional graphite, isotropic graphite is used extensively not only in the manufacture of silicon semiconductors but also in electric discharge machining electrodes used in manufacturing molds, dies for casting copper alloys and the like continuously, metal melting crucibles, hot-pressing molds, core materials for high-temperature gas-cooled reactors, and other similar uses. Figure 1 shows the manufacturing processes of this isotropic graphite.

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00007-1 Copyright Ó 2013 Elsevier Inc. All rights reserved.

6. 7. 8. 9. 10.

Manufacturing Process of Epitaxial Wafers CVD Membrane Forming Process Dry-Etching Process Ion Implantation Process Thermal Diffusion Process

129 130 131 132 132

Isotropic graphite is produced through these processes, with a time of five to six months taken for the raw materials to be processed into graphite blocks. Table 1 shows the characteristics of our typical materials, often used in the manufacture of silicon semiconductors, among a number of kinds of isotropic graphite.

2. SILICON SEMICONDUCTOR MANUFACTURING PROCESSES Silicon semiconductors are manufactured roughly in the processes shown in Figure 2; the processes underlined in black use graphite. The reason why graphite is used in these processes is that it generally has the following properties meeting the requirements for parts used in these processes: 1. Lightweight 2. Resistant to heat and not melting under normal pressure (sublimation temperature: about 3650  C) 3. Low vapor pressure (3) (about 3.8  10 2 mmHg at 2500  C) 4. High emissivity (about 0.7 to 0.8) 5. Suitable electric resistivity (an important factor for use in heater applications) 6. An increase in strength proportional to temperature (flexural strength at 2500  C in a nonoxidizing atmosphere about 1.5 times that at room temperature)

125

126

Handbook of Advanced Ceramics

FIGURE 1 Manufacturing processes of isotropic graphite. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 2.6

Carbon Materials for Si Semiconductor Manufacturing

127

TABLE 1 Typical characteristics of some grades by Toyo Tanso

Grade

Apparent density Mg/m3

Hardness HSD

Electrical resistivity mU$m

Flexural strength MPa

IG-110

1.77

51

11.0

39

78

4.5

IG-710

1.83

58

10.0

47

103

4.6

Compressive strength MPa

Coefficient of linear thermal expansion  10 6/K

Temperature range over which the coefficient linear thermal expansion was measured: 623w723 K.

7. High thermal conductivity and low coefficient of thermal expansion make it resistant to thermal shock and stable in dimensions. 8. Easy machining 9. Capable of being purified (ash contents can be reduced to the order of ppm.) Silicon semiconductors are made from silicon purified to an ultra-high purity. That purity is called eleven nines, that is, the purity of silicon is 99.999999999%. The use of ultra-high-purity silicon as raw material to manufacture semiconductors in this way naturally leads to the requirement for a high purity in graphite material

used in the manufacturing processes of the silicon raw material. Ordinary graphite contains impurities of the order of 400 mass ppm or so. Purifying this raw material at temperatures over 2000  C in the stream of halogen gas causes metal impurity contents to volatilize in the form of halogen compounds with low boiling points and nonmetal impurities to evaporate; this makes it possible to obtain high-purity graphite with an impurity content of less than 20 mass ppm. In addition, improved processing methods will make it possible to manufacture ultra-high-purity graphite with an impurity content of less than 5 mass ppm.

FIGURE 2 Manufacturing processes for silicon semiconductors. For color version of this figure, the reader is referred to the online version of this book.

128

3. MANUFACTURING PROCESSES FOR POLYCRYSTALLINE SILICON The manufacture of silicon as raw material for silicon semiconductors begins with the manufacture of metallic silicon by means of reducing relatively pure silica. After this process, metallic silicon is subjected to the reaction with hydrogen chloride to remove impurities in metallic silicon and to form trichlorosilane, which is distilled repeatedly to produce high-purity trichlorosilane. Using this highly purified trichlorosilane gas, polycrystalline silicon is grown around a silicon seed rod in a reaction vessel heated up to about 1200  C by means of the CVD method, and a seed rod with a diameter of about 10 mm is grown to a polycrystalline silicon rod with a diameter of about 100e150 mm in the final process. This method is called the Siemens method, with this device schematically shown in Figure 3. In this process, it is necessary to pass current through the seed rod for the rod itself to be a heating element, with the graphite part serving both as the electrode and as the seed rod holder to allow current to pass through the silicon seed rod. With the CVD process requiring a high temperature between 1000 and 1200  C, the characteristics of graphite of the absence of reduction in strength and the possession of adequate electrical conductivity under high temperatures are used; pieces of graphite used in this way are commonly called seed chucks. In actual manufacturing processes, the graphite seed chuck undergoes an action to break it to collect polycrystalline silicon manufactured by the Siemens method, and, for such a use, a less expensive graphite material made of relatively coarse grains called an extruded graphite is used in the main.

FIGURE 3 Schematic diagram of a Siemens method reduction furnace. For color version of this figure, the reader is referred to the online version of this book.

Handbook of Advanced Ceramics

In addition to these uses, graphite is used as the preheater to heat the silicon seed rod and the nozzle introducing gas into the furnace, and the like.

4. MANUFACTURING PROCESS FOR MONOCRYSTALLINE SILICON This is the process by which monocrystalline silicon rods are produced with ultra-high-purity polycrystalline silicon manufactured in the above method as the raw material. Two methods are available to produce a silicon single crystal, the CZ (Czochralski) and the FZ (floating zone) methods. In both methods, a single crystal is grown by contacting polycrystalline silicon with a monocrystalline seed. However, the techniques are different from each other, with the FZ method using few carbon parts. In the CZ method, to grow a single crystal, on the other hand, a monocrystalline seed is brought into contact with melted polycrystalline silicon, and the seed is pulled upward gradually while being rotated slowly. Almost all of furnace internal components are made of purified graphite material or carbon heat insulator; this process is the one using the largest number of graphite parts in the silicon semiconductor manufacturing processes. Figure 4 shows a silicon single-crystal manufacturing furnace based on the CZ method schematically. In the illustration, typical designations are used for different parts because they are called differently depending on the user and device manufacturer. By means of such growing furnaces, huge silicon single crystals with a diameter of 300 mm and a length of over 2 m are manufactured. Graphite parts are usually machined from a cylindrical block. However, the larger the diameter of a silicon single crystal becomes, the larger that of the graphite used; for this reason, the diameter of a crucible to make a 300-mm-dia. silicon single crystal exceeds 800 mm. As a result, the graphite heater to heat such a crucible and the heat shield member installed outside the heater to intercept heat from the heater have diameters as large as 1000e1300 mm. In connection with this, the upper parts exposed to Si vapor and SiO gas at high temperature tend to suffer from cracking and chipping as a result of the reaction between carbon and silicon. In addition, carbon contents mixed in manufactured single-crystal silicon as a result of the occurrence of the above cracking and chipping are quite capable of leading to problems such as abnormal product quality. As a means of resolving such a problem, graphite parts are used whose surfaces are coated with SiC or pyrolytic carbon. Lightweight and high-strength C/C composite materials, which are manufactured by laminating carbon fiber in response to an increase in the diameter of parts in the

Chapter | 2.6

Carbon Materials for Si Semiconductor Manufacturing

129

FIGURE 4 Schematic diagram of a CZ method single-crystal manufacturing furnace. For color version of this figure, the reader is referred to the online version of this book.

furnace, are finding their applications in crucibles and shielding materials.

5. PROCESSES OF MACHINING A SILICON SINGLE-CRYSTAL INGOT After its perimeter has been machined to the desired diameter, a silicon single-crystal rod prepared by the CZ or the FZ method is provided with a plane called an orientation flat or a small notch to show the crystal orientation. After this, the single-crystal ingot is sliced to a thickness of 1 mm with a slicing device shown in Figure 5. In the past, an inside perimeter edge blade was used to slice off disks one by one. As a result of the difficulty in manufacturing an inside perimeter edge blade with a large

diameter to match an increase in the diameter of a singlecrystal ingot and an increase in the production efficiency, the saw slicing with a wire saw capable of producing a number of slices simultaneously has become the mainstream production method. In carrying out these machining operations, the singlecrystal ingot is bonded to the slicing beam that is made of graphite. Being porous, graphite is easy to bond and cut and allows a wafer to be cut and separated easily; this is the reason why graphite is often used.

6. MANUFACTURING PROCESS OF EPITAXIAL WAFERS In some cases, as needs arise, the CVD method is used to grow thin single-crystal silicon film epitaxially on the

130

Handbook of Advanced Ceramics

FIGURE 5 Diagrammatic sketch of a silicon single-crystal slicing device. For color version of this figure, the reader is referred to the online version of this book.

FIGURE 6 Schematic diagram of epitaxial manufacturing device.

surface of a mirror wafer, finished in a mirror-like smooth surface by grinding a sliced wafer; this is called an epi-wafer. Epitaxial manufacturing devices are roughly classified into the pan-cake type and the barrel type, which are shown schematically in Figure 6. In this process, SiC-coated high-purity graphite is used for the platform, called the susceptor, on which to place a mirror wafer. The reason why the platform is called the susceptor is that the platform, made of SiC-coated high-purity graphite and caused to generate heat by induction heating, serves also as a heater to heat the mirror wafer placed on it. The reasons for providing SiC coating are: 1. The purity of SiC is higher than that of graphite; 2. It does not easily react with the silicon wafer with which it comes in contact; and 3. Being porous, graphite adsorbs much gas, having the property of releasing the gas adsorbed when heated. The prevention of gas release by means of coating reduces the adverse effect on the silicon wafer grown epitaxially.

The difficulty of processing two or more wafers placed on the platform simultaneously due to upsized wafer diameters and the requirement for the quality improvement such as membrane thickness uniformity have led to the frequent use of single-wafer-processing susceptors that process a single large-diameter wafer placed on them.

7. CVD MEMBRANE FORMING PROCESS With semiconductor manufacturing processes requiring many layers of circuits to be stacked, securing insulation between layers becomes necessary, and membranes made of SiO2, SiON, SiN, and the like are used as insulating membranes. Used as devices to form these insulating membranes are atmospheric pressure CVD systems (Figure 7) and plasma enhanced CVD systems (Figure 8). High-purity graphite and SiC-coated graphite are used also in platforms to carry wafers in the furnace and plates to hold wafers in the furnace, for example.

Chapter | 2.6

Carbon Materials for Si Semiconductor Manufacturing

131

FIGURE 7 Schematic diagram of an atmospheric pressure CVD system.

FIGURE 8 Schematic diagram of plasma enhanced CVD system.

8. DRY-ETCHING PROCESS SiO2 is used for interlayer insulating membranes in the circuit formed on the surface layer of a wafer; however, the unnecessary portion of the SiO2 membranes must be removed after they were formed by means of the CVD method. For this purpose, the etching gas is turned into plasma, and, using its energy, the unnecessary portions of the insulating membranes are removed in a dry-etching process. As shown in Figure 9, high-frequency voltage is applied between the upper and the lower electrode to generate discharge, which turns etching gas into plasma to etch only the desired surface portion of the wafer. The graphite is used as the cathode electrode in the upper part, and is shaped into a disk with a number of holes provided to blow out etching gas.

Graphite is used as the cathode electrode for the excellence of its good electrical conductivity, sputter resistance, gas-corrosion resistance, and the like. With

FIGURE 9 Schematic diagram of a dry-etching system.

132

Handbook of Advanced Ceramics

FIGURE 10 Schematic diagram of an ion implanter system.

FIGURE 11

Schematic diagram of a thermal diffusion furnace.

the cathode electrode positioned above the wafer, if particles like carbon atoms sputtered by the collision of machining powder and ions fall on a wafer, this will affect the circuit on the wafer adversely. For this reason, graphite is very rarely used alone, but it is used with SiC coating on it, or with its surface converted to SiC, or in the form of amorphous carbon (glassy carbon), and so on.

9. ION IMPLANTATION PROCESS In manufacturing semiconductors from silicon, it is impurity elements, introduced in very small quantities into silicon (a fourth group element), of the third group, B, P, As, and others that allow the performance as semiconductors to be displayed.

In this process, the element introduced is evaporated and its vapor is hit by electrons and ionized, and then the ions generated in this way are drawn out by the force of the electric field, being further accelerated by the electric field to accumulate their energy, and finally they are shot into the target position on the wafer at a high speed. Figure 10 schematically shows an ion implanter, in which several tens of high-purity graphite parts are used in the ion extraction electrodes, accelerating electrodes, shaping slits to adjust the beam shape, and the like.

10. THERMAL DIFFUSION PROCESS This process also is intended to introduce impurity elements into a wafer, using the self-diffusion of atoms by heat.

Chapter | 2.6

Carbon Materials for Si Semiconductor Manufacturing

Figure 11 shows the system schematically, in which graphite with its surface converted into SiC is used on the jig called a wafer boat used to vertically arrange wafers to be processed. Although there are wafer boats made of quartz, those made of SiC-converted graphite are more resistant to high temperatures and allow the diffusion treatment

133

to be carried out at higher temperatures, thereby providing the advantage of shortening the treatment time. This concludes an explanation of carbon materials used in the silicon semiconductor manufacturing processes; we hope it would be a help for the readers to understand the roles of carbon.

Chapter 2.7

Isotropic Graphite for Electric Discharge Machining Hiroaki Itami Product Planning Department, Toyo Tanso Co., Ltd., 2181-2 Nakahime, Ohnohara-cho, Kanonji, Kagawa 769-1612, Japan

Chapter Outline 1. What is Electric Discharge Machining? 135 2. What is a Mold? 135 3. Choice Between Electric Discharge Machining and Cutting to Best Suit the Purpose 136 4. Principle of Electric Discharge Machining 137

1. WHAT IS ELECTRIC DISCHARGE MACHINING? Electric discharge machining is a method of machining in which, as shown in Figure 1, when an electrode and a workpiece are brought closer to each other with pulse voltage being applied between them, dielectric breakdown

5. Kinds of Electrode used in Electric Discharge Machining 6. Kinds of Graphite Electrode Material for Electrical Discharge Machining 7. High-performance Graphite 8. Concluding Remarks

137 139 139 140

occurring at a certain distance between them triggers discharge, whose high temperature and pressure in turn cause part of the workpiece to be melted and blown. Only if current flows through a workpiece, even one made of a hard metal that cannot be machined by cutting can be machined; this is what electric discharge machining can do. Figure 2 shows a piece of graphite, the electrode, and the workpiece produced by transferring a shape by means of electric discharge from the electrode. This electric discharge technology is used to manufacture molds.

2. WHAT IS A MOLD?

FIGURE 1 EDM outline. For color version of this figure, the reader is referred to the online version of this book.

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00008-3 Copyright Ó 2013 Elsevier Inc. All rights reserved.

Today, most of the products and their parts surrounding us such as automobiles, home appliances, and everyday items are manufactured by means of molds. If you make and retain the shape of a product in a mold, simply closing the mold and pouring resin or molten metal into the mold enables one to make a great number of products of the same shape economically in a matter of a short time. Molds are now indispensable items to maintaining our comfortable life. Molds are classified according to raw materials used to make products, manufacturing methods, and other factors. Table 1 shows the kinds of molds and methods of molding products.

135

136

Handbook of Advanced Ceramics

FIGURE 2 EDMing Sample. For color version of this figure, the reader is referred to the online version of this book.

TABLE 1 Kinds of Mold and Methods of Molding Kind of mold

Method of molding

Plastics mold

Pressure is applied to melted resin to inject it into a mold for forming.

Casting mold

Melted metal is poured into a mold for forming.

Die-cast mold

Pressure is applied to melted metal to inject it into a mold for forming.

Forging mold

Metal that has become soft due to heating is formed by hitting it with a mold.

Press mold

A metal plate is held between molds and pressed for forming.

Rubber mold

Heated rubber raw material is subjected to blow of compressed air from the inside and pressed against the inside of a mold for forming.

3. CHOICE BETWEEN ELECTRIC DISCHARGE MACHINING AND CUTTING TO BEST SUIT THE PURPOSE To make a mold, a desired shape is formed out of a metal lump in one of the two widely used methods that are described below: 1. Mechanical machining: Edged tools are used to cut metal directly. 2. Electric discharge machining: An electrode having the shape of a desired product is made, and electric discharge is generated between the electrode and the workpiece to melt the metal. In the following, the criteria as to the choice between electric discharge machining and mechanical machining are described.

Metals with hardness of 55HRC (Rockwell hardness) or more are called difficult-to-machine materials because machining mold materials with hardness of 55HRC or more results in much wear in end mill cutting edges. When a difficult-to-machine material is handled, a material with good machinability (graphite and copper, for example) is machined into the product shape, which in turn is used as the electrode installed on an electric discharge machine to carry out electric discharge machining (shape transfer). In a case where the material is not difficult to machine, too, mechanical machining to depth requires the protrusion of the end mill to be longer, which increases the probability that vibration occurs during mechanical machining. If L/D ¼ (machining depth) O (end mill diameter) exceeds 10, mechanical machining will be accompanied by

Chapter | 2.7

Isotropic Graphite for Electric Discharge Machining

FIGURE 3 Choice between cutting and electric discharge machining.

the risk of vibration, and, to avoid this, electric discharge machining is chosen. Figure 3 shows the criteria for choosing a machining method on the basis of the “hardness of mold material” and the “machining depth.”

137

With these phenomena, described in items 1 to 6, occurring several thousands to several tens of thousands of times per second, electric discharge machining proceeds. The quantity to be blown away is determined by the electric conditions. Advancing machining steps to the closest point to the ultimate shape is called rough machining, and finalizing machining steps from the rough machining to the ultimate shape is called finish machining. In some cases, the electrical conditions are changed through a gradual reduction in discharge power in several steps. Two important parameters determining the electrical conditions are the current value and the ON-time duration. In rough machining, in which the importance is placed on the machining speed, the current value is made large as shown in Figure 4(a), with a current waveform with a longer ON-time duration appearing. In finish machining, in which the importance is placed on the surface roughness, the current waveform is defined as one with a smaller current value and a shorter ON-time duration (Figure 5).

4. PRINCIPLE OF ELECTRIC DISCHARGE MACHINING

5. KINDS OF ELECTRODE USED IN ELECTRIC DISCHARGE MACHINING

1. Voltage increase: Pulsed high voltage is applied between the workpiece and the electrode. 2. Discharge column formation: When the electrode and the workpiece come closer to each other to a distance (several tens to hundreds microns), a discharge column, a kind of plasma, is formed 3. Melting: Heat generated by discharge melts the electrode and the workpiece at the same time. 4. Evaporation: the surrounding working fluid is evaporated instantly. 5. Explosion: The pressure generated by the evaporation blows the melted electrode and workpiece away. 6. Cooling: The working fluid cools off the discharge point, with a hollow left on the workpiece surface.

Electrode materials used in electric discharge machining are graphite (isotropic graphite), copper, copper tungsten, and the like. Copper is used in cases in which finer surface roughness is required. However, copper has a high bulk density of 8.9 Mg/m3, and, when the electrode area is large, the machining accuracy may be affected adversely as a result of the surpassing of the allowable suspension load of the electric discharge machine or of the bending of the electric discharge machine, at the time of jumping, caused by the electrode mass. Copper is not suitable to high-speed electric discharge machining, because allowing large current to flow through the copper piece causes the discharge surface to soften due

FIGURE 4 How EDM Works. For color version of this figure, the reader is referred to the online version of this book.

138

Handbook of Advanced Ceramics

TABLE 2 Advantages of Graphite (as Compared with Copper)

FIGURE 5 Waveform of Current. For color version of this figure, the reader is referred to the online version of this book.

Advantage of graphite in mechanical machining

Advantage of graphite in electric discharge machining

No burrs occurring

Shorter time before the start of discharge

Low cutting resistance, hence easy to machine

Fast electric discharge machining

Lightweight, hence easy to handle

Lightweight, hence less burden on the main shaft

Integrated electrode construction possible Smaller stress deformation

Smaller thermal deformation

to heat and the surface becomes rough as a result of explosion. With a bulk density of about 1.8e1.9 Mg/m3, about one-fifth of that of copper, graphite is a light substance used for large electrodes (Figure 6). In addition, graphite has high strength at elevated temperatures, and the advantage of faster machining speed because it allows larger current to flow through itself than copper does. However, since holes called pores are present in graphite, graphite is inferior to copper in terms of threshold discharge surface roughness. Extremely resistant to heat, copper tungsten is used in machining ultra-difficult-to-machine materials such as carbide. However, the extremely poor machinability makes it difficult to manufacture electrodes, and the price is high. Table 2 shows the advantages of graphite as compared with copper in mechanical machining and electric discharge machining. Having smaller cutting resistance, graphite is machined mechanically in a shorter time than copper. Mechanical

machining of copper is accompanied by burrs and its removal needs much labor, but graphite is not accompanied by burrs and does not need the time to remove them. After mechanical machining, graphite can be passed to electrical discharge machining, the immediately following process. In other words, graphite is the material indispensable to implement the automatic changeover from mechanical machining to electrical discharge machining. In electrical discharge machining, the short time needed to start electrical discharge and the high machining speed allow the machining time to be reduced. The coefficient of thermal expansion of graphite is as low as less than one-third of that of copper and its deformation due to heat in mechanical and electrical discharge machining is small; this contributes to high accuracy machining (Figure 7).

FIGURE 6 Bulk Density. For color version of this figure, the reader is referred to the online version of this book.

FIGURE 7 Coefficient of Thermal Expansion. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 2.7

Isotropic Graphite for Electric Discharge Machining

139

TABLE 3 Graphite Grade for EDM Hardness HSD

Specific resistivity microOhm・m

Flexural strength MPa

Class

Grade

Recommendation

Bulk density Mg/m3

Medium-Grade

ISEM-8

for Rough

1.78

65

13.5

52

ISO-63

for Rough and Finish

1.78

76

15.0

65

TTK-5

for Finish

1.78

80

15.5

80

Excellent-Grade

Source: (Toyo Tanso catalog)

6. KINDS OF GRAPHITE ELECTRODE MATERIAL FOR ELECTRICAL DISCHARGE MACHINING Graphite is black as ebony and all kinds of it look the same. However, differences in the parameters specified in manufacturing processes such as the grain size distribution of raw material, sintering temperature, and graphitization temperature yield kinds of graphite having general physical properties and discharge characteristics different from those of other kinds. Both materials suitable for rough machining accompanied by fast electrical discharge machining speed and materials suitable to machining with less electrode wear and finer discharge surface roughness are available; it is necessary to choose the graphite one uses depending on the purpose. Table 3 shows typical graphite materials for electrical discharge machining supplied by Toyo Tanso.

graphite, mechanical machinability was traded off for electrical discharge machinability, that is, a decrease in the electrode wear as the top priority with the aim of controlling manufacturing process conditions to increase “flexural strength.” This sometimes posed problems of chipping during mechanical machining. Against this background, the newly developed TTK material manufactured by Toyo Tanso is a product with inventions in materials design; the material is produced on the basis of a technique of homogenizing grain sizes in a small dimension as shown in Figure 8. As a result, a material was developed successfully that meets both mechanical machinability and electrical discharge machinability that had been in a trade-off relationship. In developing this new material (Figure 9), we reduced and homogenized grain sizes so that the load on the end mill may be constant during mechanical machining; this leads to the realization of more accurate shapes (Figures 10 and 11).

7. HIGH-PERFORMANCE GRAPHITE The demand for high-speed machining coupled with high accuracy has become strong also in areas of industry that have been using copper electrodes, with the request for using graphite having increased. However, the electrical condition for electrical discharge machining in parts with smaller areas and finer surface roughness tends to lead to larger electrode wear with graphite electrodes. For the reason mentioned above, a graphite electrode with satisfactory surface roughness and low electrode wear has been looked for eagerly. In the past, a kind of graphite with low electrode wear was available, but although it exhibited excellent performance in electrical discharge machining, its performance in mechanical machinability with an electrode was poor. Generally speaking, “mechanical machinability” and “electrical discharge machinability” of graphite are in a trade-off relationship, and satisfying both simultaneously is difficult. In manufacturing conventional high-grade

FIGURE 8 Strength of Graphite depends on Temperature. For color version of this figure, the reader is referred to the online version of this book.

140

Handbook of Advanced Ceramics

FIGURE 9 Image and metallographic pictures of a high-grade material. For color version of this figure, the reader is referred to the online version of this book.

FIGURE 10 Straight pin with a diameter of 50 mm. For color version of this figure, the reader is referred to the online version of this book.

In electrical discharge machining, a smoother electrode surface contributed to improving the surface roughness and waviness substantially. Normally, however, the grain size may be reduced; the presence of air holes called pores is unavoidable because of the method of manufacturing graphite. For this reason, the new material cannot excel copper or copper tungsten in terms of the threshold discharge surface roughness. However, TTK material excels both of the conventional materials in that it allows ultra-precision and minute shapes to be realized. When copper or copper tungsten is machined mechanically, cutting resistance causes thermal deformation. Graphite also is subject to thermal deformation; however, since its coefficient of thermal expansion is very small, onethird of that of copper, thermal deformation is slight, which allows high-speed feed and thereby reduces the machining time substantially. In a typical case, a PCD (polycrystalline diamond) tool was used to machine a column with a diameter of 30 mm. When undergoing electrical discharge machining, copper and copper tungsten cannot maintain the shape as a result of the softening and thermal expansion due to the heat from discharge. On the other hand, highly resistant to heat and having a small coefficient of thermal expansion, graphite is capable of keeping the shape while being machined by means of electrical discharge.

8. CONCLUDING REMARKS FIGURE 11 Rib with a thickness of 50 mm. For color version of this figure, the reader is referred to the online version of this book.

The current trend is reducing the number of electrical discharge machining processes. As described above, however, there are a number of shapes that cannot be

Chapter | 2.7

Isotropic Graphite for Electric Discharge Machining

handled with mechanical machining, and electrical discharge machining is very excellent in terms of accuracy and machine service life. In addition, there are also cases of machining in which shapes cannot be realized even by means of a copper or copper tungsten electrode. Graphite material is

141

extremely suitable to such shapes and is compatible with shapes that have not been considered (high aspect ratio shapes). The use of graphite material in electrical discharge machining of ultra-precision shapes promises a very great possibility.

Chapter 2.8

Carbon Fibers Young-Pyo Jeon, Rebecca Alway-Cooper, Marlon Morales and Amod A. Ogale Center for Advanced Engineering Fibers and Films, and Department of Chemical and Biomolecular Engineering, Clemson University, Clemson, South Carolina 29634, United States

Chapter Outline 1. Background 2. PAN-based Carbon Fibers 3. Pitch-based Carbon Fibers

143 144 147

1. BACKGROUND Carbon fibers are the strongest and stiffest fibrous reinforcement known to date [1]. Carbon fibers were first used by Edison in 1870 as filaments for incandescent bulbs. Carbon fibers for structural applications became available starting only in the early 1960s and improvement of their properties took place steadily over the next three decades. Many grades possess a tensile strength of about 5 GPa, which make them almost five times stronger than steel. Other grades possess a thermal conductivity of 1,000 W/m$K, which is almost three times that of copper [2]. Coupled with their low density of 1.7e2.0 g/cm3, such fibers have found extensive use in high-performance aerospace and defense applications [3]. In the 1990s and later, significant efforts have been devoted to improving the production processes, which has led to not only improvement of properties but also a general reduction in their cost. Since then, carbon fibers have been used in industrial components, nuclear reactor components, rehabilitation of bridges and buildings [4], high-performance sporting goods, and packaging of electronics [5]. In addition to strength and stiffness, other carbon fiber characteristics that have led to above applications include [6]: l

l

l

high dimensional stability, low thermal expansibility, and low abrasion for missile cone and aircraft brakes; good vibration damping and outstanding stiffness for audio equipment, high-performance hi-fi equipment, and robotic arms; high electrical conductivity for EMI and RF shielding (electromagnetic shielding blocks radio frequency electromagnetic radiation); and

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00009-5 Copyright Ó 2013 Elsevier Inc. All rights reserved.

4. Rayon and Bio-based Precursors Acknowledgments References

l

151 152 152

chemical inertness and high corrosion resistance for chemical and nuclear industries.

Their outstanding strength to weight ratio makes carbon fibers ideal reinforcements for military and civilian aircraft applications. F-16 and Stealth bombers have been made of carboneepoxy composites for almost twenty-five years now. Recently, Boeing 787 “Dreamliner” commercial airplane has been designed to use about 50% by weight of carbon fiber composites for its primary structure [7]. In landbased transportation applications, the high cost of carbon fibers (US$ 10e1500) somewhat limits its use to highperformance sporting goods (golf clubs and tennis rackets), and racing cars and high-performance automobiles such as BMW “E-series” electric vehicles [8]. Current, worldwide production of carbon fibers is estimated at 50,000 metric tons and the overall market is US$ 1.5 billion, which is expected to go up to about US$ 2 billion by 2015 [9,10]. Although commercial carbon fiber processing has been around for almost fifty years now, it is still not a mature technology due to the fact that carbon fibers are intractable. Therefore, thin fibers (nominally 10 mm) must be first obtained from an organic or hydrocarbon precursor. Then, the precursor fibers have to be converted to carbon fibers by thermal treatment to temperatures ranging from 1000 to 3000  C [11]. The complex reactions lead to removal of noncarbon constituents, leaving behind carbon fibers. The chemical composition of the precursor plays a profound role on the microstructure and properties of the resulting carbon fibers. Therefore, the remainder of the chapter is divided into three main sections based on the precursors: (i) PAN-based, (ii) pitch-based, and (iii) rayon- and bio-based.

143

144

Handbook of Advanced Ceramics

2. PAN-BASED CARBON FIBERS 2.1. Conventional PAN Precursors Over 90% of commercial carbon fibers are produced from polyacrylonitrile (PAN) precursors [12,13]. The primary reason for this widespread use of PAN is the fact that PAN polymer does not melt, but can be dissolved in suitable solvents. Consequently, PAN precursors can be solution-spun into thin fibers that can be thermally processed into intractable fibers because PAN does not melt. It undergoes crosslinking and ladder-formation reactions before melting, and leads to the formation of intractable fibers, which can be further heat treated to temperatures in the 1000e1500  C range to obtain carbon fibers [14]. A schematic diagram of the solution-spinning process is displayed in Figure 1. A solution of PAN (15e25 wt%) is first prepared using dimethyl sulfoxide (DMSO), dimethyl acetate (DMAc), dimethyl formamide (DMF), or another suitable solvent. The solution is wet-spun into a coagulation bath where the solvent diffuses out, leaving behind PAN fibers. Due to limited viscoelastic properties (strain hardening) displayed by PAN solutions, wet spinning cannot be performed at a fast rate or a large draw-down ratio. Also, the solvent must be allowed to diffuse slowly to prevent the formation of a hard skin that can lead to entrapment of solvent, which ultimately results in large cavities and voids in the fiber that result in a weak fiber. In any event, the out-diffusion of solvent leaves behind micropores that must be collapsed in a post-draw of the solidified PAN fibers. As described in Figure 1, PAN-based carbon fiber can also be specified into three categories according to the heat treatment conditions and their properties: low-temperature heat-treated (300e1000  C) type 1 fiber of low modulus and strength, intermediate-temperature

heat-treated (1000e2000  C) type 2 fiber of high-strength, and highetemperature-treated (2000e3000  C) graphitized type 3 fiber of high modulus. Next, the fibers are thermo-oxidatively reacted in a step, often called “stabilization,” to make them intractable and hold them together in such a way as to avoid extensive relaxation and chain scission during the final carbonization step. Stabilization is typically conducted in air at temperatures ranging from 200 to 300  C, and leads to crosslinking and/or cyclization of PAN [15e18]. The cyclization is an exothermic reaction during which the pendant nitrile groups react with each other, transforming PAN backbone into a ladder-type structure, which is intractable. The reaction mechanisms for stabilization depend on the experimental conditions and type of copolymer used [19]. Numerous reactions can take place during heating of PAN, and many are still not well understood [20]. Some studies, such as of Burland and Parsons [21], showed that the first step of the stabilization is cyclization through reaction of the nitrile groups, with dehydrogenation being significant only above 300  C. Others, such as Grassie and McGuchan [18], have proposed that dehydrogenation and cyclization reactions take place simultaneously, the former occurring both within the noncyclized polymer chain as well as within the condensed heterocyclic rings. Cyclization reactions are highly exothermic and can lead to runaway reactions. So, this behavior can be considerably reduced if a comonomer such as methyl acrylate (MA), vinyl acetate (VA), or itaconic acid (IA) is introduced into the polymer chain. Also, the activation energy of the cyclization reaction is smaller for the copolymer, so the cyclization reaction sets in at a lower temperature and is completed over a broader range over a longer duration. Thus, the amount of comonomer in the

FIGURE 1 Schematic diagram of PAN-based carbon fiber processing.

Chapter | 2.8

Carbon Fibers

precursor affects the rate of oxidative stabilization [15,19] , and thence the applied tension requirements [22]. At the molecular structure level, several studies have discussed the stereospecificity of the cyclization reaction [23e25]. It has been argued that the cyclization reaction should occur preferentially in isotactic sequences to form the ladder-like structure. Generally, it has been observed that intramolecular cyclization reactions occur at lower temperatures (175e230  C) in the amorphous phase of the polymer [12,26]. The crystalline regions act as “bridge” points between the amorphous regions, holding the structure together. Above 320  C, oxidation and intermolecular crosslinking take place, and that oxidative degradation reactions occur above 380  C [12,26]. As noted earlier, a range of copolymer compositions are used to facilitate the dehydrogenation and ladder-formation reactions. This thermo-chemical conversion process is extremely complicated; so, precursor compositions and precise processing condition remain proprietary. However, this topic has also stimulated fundamental studies, which are largely cited in this chapter. In addition to the chemical structure of PAN precursors, processing conditions play a key role in the transformation of organic precursors into carbon fibers and their microstructural characteristics. For instance, molecular orientation of PAN precursor significantly affects the properties of the polymer fibers, and orientation must be maintained as much as possible during stabilization if the final properties of the carbon fibers are to be maximized. Finally, heat treatment conditions play a pivotal role in obtaining fibers with a range of graphitic content and texture. “Graphitized” carbon fibers can be obtained from mesophase pitch precursors using heat treatment temperatures (HTT) of 3000  C or higher, which yield a high modulus and high

145

FIGURE 2 SEM image of a typical PAN-based T300 carbon fiber.

thermal conductivity. For high strength and intermediate modulus, HTT range of 1000e1500  C is typically used for PAN precursors. A representative scanning electron micrograph (SEM) of T300 carbon fiber is displayed in Figure 2. It is interesting to note that the precursor fibers and the resulting carbon fibers derived from PAN precursors are typically not circular, but are “kidney” shaped. The crease and crenulations along the length of the fibers are a consequence of the solvent out-diffusion during wet spinning. A typical wide-angle X-ray diffractogram for T300 fibers is displayed next in Figure 3. The results were obtained from a Rigaku-MSC unit using an X-ray ˚ (Cu target, 45 kV and 0.65 mA). wavelength of 1.5406A The integrated azimuthal distribution is shown in Figure 3(a), which indicates a two-theta peak associated with (002) layer planes at about 25.4 . This is typical of

FIGURE 3 Wide-angle X-ray diffractogram for a typical T300 PAN-based carbon fiber: (a) two-theta plot showing positive interference from (002) and (004) planes; and (b) azimuthal distribution of the (002) planes centered along the fiber axis (phi ¼ 90 ).

146

Handbook of Advanced Ceramics

the “turbostratic” structure where the interplanar spacing is 0.344 nm, significantly greater than the 0.3354 nm found in crystalline graphite. Also, the registry of the planes is not the abab. sequence found in graphite. Figure 3(b) shows the azimuthal distribution of the 002 planes; the full-width at half of the maximum height (FWHM) is almost 30 , typical of ordinary PAN-based carbon fibers. Later we will see that in highly oriented and graphitic carbon fibers, this distribution can be significantly narrower. Typical mechanical properties of PAN-based carbon fibers are listed in Table 1. The largest producer of carbon fibers, Toray (based in Japan) and Cytec (based in US), both use fiber classifications such as T300. Historically, it represented a tenacity (tensile strength) of 300 ksi, which translated to about 2.1 GPa. With precursor and process improvements, T300 grade now possesses a tenacity of almost 3.7 GPa. The modulus of such fibers is typically 200e400 GPa, whereas the strain-to-failure is 1.5e2%. Another mechanical property of great relevance in structural applications is the compressive strength of fibers, since composite structures can be subjected to compressive loads during actual use. Due to the microfibrillar orientation along the fiber axis, carbon fibers generally display much lower compressive strengths, typically less than about 2 GPa. It is also noted that the crease and surface defects present in carbon fibers cause the strength of such fibers to be much lower than the theoretical, in-plane strength of graphite that is predicted at almost 100 GPa, but one that has never been achieved. As for transport properties of PAN-based carbon fibers, electrical resistivity is typically 16e20 mU-m, whereas the thermal conductivity is about 10e20 W/m$K depending upon the HTT, with higher temperatures resulting in higher electrical and thermal conductivity. It should be noted,

however, that the thermal conductivity of PAN-based carbon fibers is at least one order of magnitude lower than those derived from mesophase pitch (as will be discussed in the next major section).

2.2. Alternative PAN-based Precursors and Processing Routes As noted earlier, the vast majority of commercial carbon fibers are produced from wet-spun polyacrylonitrile (PAN) precursor fibers [12,13]. Current research on PAN-based carbon fibers is directed towards either reducing the cost significantly or improving the properties greatly. To get away from high costs associated with the wetspinning process, the high-volume melt-spun PAN fiber route has been attempted by external plasticization of PAN and internal plasticization by making a copolymer of 10e15% comonomer content [27,28]. PAN-based precursor fibers, however, must be thermally stabilized to obtain an intractable structure before carbonization. Therefore, a suitable stabilization route is necessary for melt-spun PAN (before oxidation) to avoid melting of fibers. In studies conducted at Clemson University, we have reported earlier that UV-assisted stabilization of PAN copolymer is an alternative; however, due to lack of UV sensitivity the process took prolonged stabilization time [30,31]. Subsequently, we used a PAN terpolymer containing a UV-sensitive monomer, acryloyl benzophenone (ABP) [32], which was synthesized by researchers at Virginia Tech. Fibers were melt-spun at 220  C and UV irradiated under a high-power (4.5 kW) mercury arc lamp for 1e5 min durations. After UV irradiation, the terpolymer was sufficiently crosslinked such that the fibers could be successfully oxidatively stabilized at 320  C and then carbonized at 1500  C. The modulus of the precursor fibers

TABLE 1 Properties of PAN- and Pitch-based Carbon Fibers [1,43] Classification

Carbon fiber

Modulus [GPa]

Tensile strength [GPa]

Electrical resistivity [mU$m]

Thermal conductivity [W/m$K]

PAN-based carbon fiber

T300

230

3.5

18.0

14

T650

255

4.3

15.2

15

T800

294

5.9

14

16

M35 J

343

4.7

11

45

P25

160

1.56

12.1

36

P30

207

1.4

10.2

62

P55

220

0.9

9.4

74

P100

470

1.0

4.6

321

K1100

565

3.1

1.1e1.3

900e1000

Pitch-based carbon fiber

Chapter | 2.8

Carbon Fibers

was about 6 GPa and the tensile strength was about 230 MPa and strain-to-failure was about 16%. The carbonized fibers displayed tensile strengths and moduli as high as 730 MPa and 140 GPa, respectively. Although the modulus of the carbonized fibers was acceptable, the tensile strengths were lower than those of commercial carbon fibers due to the presence of occlusions and other defects. Nevertheless, the study demonstrated that melt spinning, combined with UV-assisted stabilization, is a viable solvent-free route for producing PAN-based carbon fibers [32]. An interesting characteristic of these carbon fibers derived from melt-spun PAN terpolymer is that their two-theta peak associated with (002) layer planes in is observed at a low value of 24.9 , when typical T300 type carbon fibers display it at 25.6 . The low two-theta value indicates low graphitic crystallinity and suggests that such fibers may be used in high-temperature, insulation applications that require thermal stability. Another recent study describes a continuous, plasmabased process for faster stabilization of precursor fibers [33]. PAN, pitch, or any other suitable organic/polymeric precursor fiber can be used as a feedstock in the microwave-plasma system. The authors expect about 20% reduction in processing costs. Finally, in other studies, researchers are attempting to increase the properties of carbon fibers significantly. In a recent study [34], three different strategies were used simultaneously: a significant reduction in the diameter of carbon fibers, from 6 mm to 1 mm, was achieved using an island-in-sea bicomponent fiber spinning of a highmolecular-weight PAN precursor to which 1 wt% singlewall carbon nanotubes were added. The tensile strength was reported to be 4.5 GPa and modulus was over 450 GPa for the SWNT-modified fibers in comparison with 3.7 GPa tensile strength and 337 GPa modulus for pure PAN-based, ultrathin carbon fibers. Ultimately, the translation of these enhanced fiber properties into properties of structural composites needs to be determined. If successful, such fibers can lead to nextgeneration composites that are either significantly stronger than the current ones (at a higher cost) or much less expensive than current composites (at similar property level).

3. PITCH-BASED CARBON FIBERS 3.1. Conventional Pitch Precursors The second most common precursor in current use for production of structural carbon fibers is mesophase pitch (MP), a thermotropic liquid crystalline material. Thus, it can be melt-spun using processing temperatures above the softening point. Mesophase pitches are disk-like, polynuclear aromatic compounds [35]. MP-based carbon fibers have been recognized for their superior lattice-based

147

properties, viz. tensile modulus and thermal conductivity [12,13,27]. Mesophase pitch is typically obtained from either thermal polymerization of isotropic pitches [36,37] or chemical synthesis of monomers such as naphthalene or anthracene [38]. Perhaps, the most easily processed MP pitch is one derived from polymerization of naphthalene. The reaction is catalyzed by a BF3 Lewis acid, and produces an MP pitch with a low softening point of about 270  C due to a low molecular weight centered around 500e600 amu [39,40]. Corrosion issues associated with the catalysts lead to significant process costs, and such precursors have typically been sold for about US$ 50e 100/kg. The other grade of MP pitches can be produced from coal-tar residues or from reaction of decant oils obtained during cracking of petroleum [41]. Coal-tar pitches are heavier in the polynuclear aromatics hydrocarbons (PAHs) and can lead to a much higher mesophase content; they also suffer from toxicity of the PAHs. Petroleum pitches offer a much safer and less toxic composition. In a typical process, MP pitch is fed into continuous extruders fitted with gear pumps, as illustrated in the schematic of Figure 4. Extrusion temperatures typically range from 300 to 360  C and pitch fibers are produced with diameters ranging from about 8 to 15 mm [42,43]. As with other precursors, these MP fibers are “stabilized” using oxidative crosslinking of the hydrocarbon. Aliphatic pendant groups attached to the discotic PAH facilitate this thermo-oxidative crosslinking reaction. The stabilized fibers are then carbonized in an inert environment to temperatures of ~ 1000  C and ultimately graphitized at temperatures exceeding 3000  C. The properties of resulting carbon fibers are a consequence of the high degree of graphene plane orientation [28e31], which in turn is generated by flow-induced orientation of discotic mesophase precursor. Circular symmetry associated with round-hole spinnerets typically results in a radial structure of the carbon fiber [44,45] as displayed in Figure 5. As heat treatment temperature increases, stack height and plane orientation increase. Crystallites are axially well orientated in the outer region of the die where shear rates are high [46]. In contrast, crystallites in the fiber are less orientated due to low shear rates at the centerline of the circular die. Figure 6(a) and (b) display wide-angle X-ray diffraction results for MP-based carbon fibers. Figure 6(a) indicates a two-theta peak associated with (002) planes at values between 26 and 26.4 . This corresponds to a large fraction of the (002) planes being stacked with an interplanar spacing of 0.3354 nm and the registry of the planes being abab sequence found in crystalline graphite. The highest two-theta value of about 26.4 for K1100 grade is consistent with the highest extent of interplanar packing and three-dimensional crystalline order resulting from an

148

Handbook of Advanced Ceramics

FIGURE 4 Schematic diagram of pitch-based carbon fiber processing.

(a)

(b)

(c)

(d)

FIGURE 5 Radial structure of MP pitch-based carbon fibers at various heat treatment temperatures.

extremely high HTT of 3000e3300  C. For P25 grade MP-based carbon fibers, the two-theta value is about 26.1 , consistent with other lattice-dominated properties that indicate a lower HTT.

The (002) peak is very sharp for K1100 fibers with an FWHM of about 0.4 . For P25 fibers, the peak is less sharp with a higher FWHM of about 0.8 . These FWHM values of ~1 for MP-based fibers are significantly smaller than

Chapter | 2.8

Carbon Fibers

149

FIGURE 6 WAXD diffractograms for P25 and K1100 MP pitch-based carbon fibers as compared with that for T300 PAN-based fibers: (a) integrated azimuthal distribution as a function of two-theta; and (b) azimuthal distribution of the (002) planes centered along the fiber axis.

that of about 4.6 for PAN-based fibers. The stacking height, LC, for K1100, P25, and T300 is of the order of 200, 20, and 2 nm. This clearly illustrates the larger threedimensional stacking developed in carbon fibers derived from disk-like MP precursors versus the chain-like PAN precursors. From the azimuthal orientation distribution of the (002) graphene layer planes within the fibers, displayed in Figure 6(b), the FWHM values were about 4 for K1100 and 14 for P25. These FWHM values for MP-based fibers are much smaller than those for PAN-based fibers (FWHM greater than 30 for T300). Typical mechanical properties of commercial MP pitch-based carbon fibers are summarized in Table 1, in conjunction with those discussed earlier for PAN-based carbon fibers. The most significant difference is that the tensile moduli of MP-based fibers are significantly higher than those of PAN-based ones, but tensile strength and strain-to-failure values are lower. This trend is a consequence of the high degree of graphitic development, which enhances lattice properties such as the modulus. In contrast, strength is a defect-controlled property, governed by the “weakest link”. The formation of large crystallite within MP-based carbon fibers also results in significant weak areas or critical defects in the inter-granular region. This region is prone to rapid crack propagation and catastrophic failure within the fiber. Consequently, MP-based fibers display significantly lower strain-to-failure values of less than 1%. For very high crystalline development, such as for K1100, strain values are less than 0.5%. However, thermal conductivity, a lattice-controlled property, is highest in such fibers. For K1100, axial fiber conductivity is reported at about 1100 W/m$K, a value two orders of magnitude higher than that of 10 W/m$K for PAN-based carbon fibers. Thermal conductivity is the ease of phonon transfer within a material, which is greatly enhanced by

the high graphitic crystallinity within MP-based carbon fibers. Overall, fiber properties not only depend upon properties of precursor but also heavily depend upon the processing conditions, heat treatment temperature, holding time at that temperature, and the defect density generated during processing. For carbon fibers produced from three MP precursors (AR-MP from Mitsubishi Gas and Chemicals, experimental MP-1, and experimental MP-2), lattice-dominated properties are displayed in Table 2. Experimental MP-1 had a mesophase content of about 75% and displayed, as expected, the lowest modulus and highest electrical resistivity. From the LavineIssi correlation [2] proposed for mesophase pitch-based carbon fiber, k ¼ [440,000/(r þ 258) ] 295 where k [W/m$K] is the thermal conductivity and r [mU$cm] is the resistivity. The estimated thermal conductivity is about 250 W/m$K, lowest among the three. For experimental MP-2, under two different spinning conditions (draw-down ratios), thicker fibers display a higher predicted conductivity (400e500 W/ m$K), and also display a “Pac-man” splitting of the fiber cross section as, illustrated in Figure 7. The hoop stresses generated during heat treatment and the highly anisotropic texture of the fiber section result in the splitting. Mochida suggested that the radial open wedge occurs due to the anisotropic manner of shrinkage along the circumference of the fiber in the outer area at high temperature [47]. The v-notches generated are also deleterious to fiber strain-tofailure values. However, the splitting of the fibers enables the graphene layers to pack and register more perfectly, which results in a higher modulus and conductivity. Another feature of MP-based fibers that is different from the one for PAN-based fibers is their compressive strength. For MP-based carbon fibers with the highest tensile modulus and thermal conductivity, the compressive

150

Handbook of Advanced Ceramics

TABLE 2 Tensile, Electrical, and Thermal (Predicted) Properties of Mesophase Pitch-based Carbon Fibers Heat-Treated at 2500  C Carbon fiber

Diameter [mm]

Apparent modulus [GPa]

Strain-to-failure [%]

Electrical resistivity [mU$m]

Thermal conductivity [W/m$K]

AR _828

10.2  0.6

414  186

0.3  0.2

2.5

570

L1_806

10.6  0.5

301  33

0.4  0.2

5.4

254

L2_828

9.4  0.7

502  66

0.3  0.1

4.6

316

L2_444

17.2  1.7

426  101

0.2  0.1

3.0

495

strength is only about 0.7 GPa, a small fraction of its tensile strength of about 2 GPa, and also a similar low fraction compared with compressive strength of about 2 GPa for PAN-based carbon fibers. From a structural composite perspective, this is not desirable because most structures must endure not only tensile stress but also compressive and bending stresses.

3.2. Alternative Mesophase Pitch Precursors and Strategies As discussed above, flow-induced orientation of discotic mesophase precursor typically results in a radial structure of the carbon fibers. However, this structure also results in flaw sensitivity, and reduces tensile and compressive strength of such fibers relative to those obtained from polyacrylonitrile precursors. Therefore, process-based control has been studied to determine its control of the microstructure of mesophase pitch [48e51]. Mochida was able to establish a strong correlation between FIGURE 7 SEM micrographs of carbon fiber heattreated at 2500  C and derived from: (a) AR synthetic mesophase; (b) experimental petroleum pitch MP-1; (c) experimental petroleum pitch MP-2 melt-spun at a high draw-down ratio; and (d) MP-2 melt-spun at a low draw-down ratio.

(a)

(c)

microstructure and the selection of pitch feedstock and spinning temperature, observing a range of textures from planar radial to quasi-onion [48]. Hamada established that mesophase domain size correlated strongly with shear stresses which, in turn, depend directly upon viscosity and inversely upon diameter cubed [49]. Also, placement of a mixer directly before the spinning nozzle resulted in microstructures deviating from planar radial, namely random, onion, and quasi-onion textures [50]. Use of a fine mesh, effectively a static mixer, also cut domain sizes and produced a distorted radial microstructure [51]. In a composition-control-based study, we have demonstrated the modification of morphology of mesophase pitch-based carbon fibers by the incorporation of carbon nanotubes [52]. Carbon multiwall nanotubes (MWNTs) with various aspect ratios were used. “Carpet” type of MWNTs, obtained from the CVD process, provided a long aspect ratio of over 1000, whereas shorter MWNTs had an aspect ratio of about 30. A range of concentrations between 0.1 and 1.0 wt% MWNTs were melt-dispersed into molten

(b)

(d)

Chapter | 2.8

Carbon Fibers

151

AR-MP mesophase pitch. Breaking of filaments during extrusion was observed at MWNT concentrations of 0.5 and 1.0 wt%. However, at 0.1 and 0.3 wt%, the spinnability was good. Additionally, a batch of mesophase pitch was compounded without the addition of nanotubes (0 wt%) as a control, in order to verify that the processing step did not impact the finally fiber microstructure. Figure 8 displays SEM micrographs of the unmodified (0 wt%), 0.1 wt% and 0.3 wt% long aspect ratio, and 0.3 wt % short aspect ratio MWNT-modified carbon fibers in cross section. The unmodified fiber exhibits a strong radial orientation of the graphite planes around the fiber axis, which has resulted in the radial splitting typical of these types of fibers. Cross section of carbon fibers containing 0.1 and 0.3 wt % long aspect ratio MWNTs exhibited a random texture. The fiber containing 0.3 wt% short aspect ratio MWNTs appears to have more of a PanAm structure. Additionally, the graphite sheets of the nanomodified fiber show significantly more folding and shorter length of continuity, compared to the unmodified fiber. In conclusion, the addition of only 0.3 wt% MWNTs to the mesophase pitch precursor allowed for the suppression of radial splitting, which can adversely affect fiber mechanical properties. These results clearly indicate that carbon nanotubes homogenize or randomize the texture of the carbon fibers and prevent them from splitting. This nested texture also prevents the fiber from splitting. For carbon fibers containing 0 wt% nanotubes (control samples), the compressive and tensile strengths were 0.5 and 2.3 GPa, for a ratio of about 0.2. For carbon fibers containing 0.1 wt% MWNTs, the compressive and tensile strengths were 1.0 and 2.5 GPa, for a ratio of about 0.4, whereas for carbon fibers containing 0.3 wt% MWNTs, the

(a)

(c)

(b)

(d)

compressive and tensile strengths were 0.4 and 1.3 GPa, for a ratio of about 0.3 [53]. The estimated axial thermal conductivities for 0, 0.1, and 0.3 wt% MWNT-modified carbon fibers (all heat treated to only 2400  C) were 600  50, 700  230, and 630  30 W/m$K, with there being no statistically significant difference in these values. Thus, compressive strength was improved in the nanomodified fibers without sacrificing the axial thermal conductivity. It is hypothesized that the addition of nanotubes to mesophase pitch nucleates a larger number of smaller-size domains in the cross section, and resulted in different morphology and texture of the resulting fibers. However, along the longitudinal axis, the crystallite length is largely unaffected and helps in retention of axial thermal conductivity of the fiber. Although synthetic AR-MP mesophase precursors are chemically most stable at spinning temperatures and can be thermo-oxidatively stabilized at moderate temperatures, the production of such grades has been suspended in recent years. Consequently, petroleum pitches will find more routine use in the future. However, like any naturally occurring substance, it is prone to compositional variation due to variability in petroleum composition obtained from different parts of the world. However, if suitable analytical techniques can be devised to characterize and standardize the raw materials, petroleum pitch is potentially a very inexpensive precursor that can significantly change the increased use of carbon fibers.

4. RAYON AND BIO-BASED PRECURSORS As noted above, PAN and MP pitch precursors produce HCN and other toxic gases during oxidative stabilization FIGURE 8 Scanning electron micrographs of ARmesophase pitch-based carbon fibers: (a) unmodified (0 wt%) and (b) 0.1 wt% long aspect ratio MWNT, (c) 0.3 wt% long aspect ratio MWNT, (d) 0.3 wt% short aspect ratio MWNT.

152

and carbonization steps. Therefore, current research studies are investigating the replacement of PAN or pitch by environmentally friendly/sustainable precursor alternatives. Although numerous hydrocarbons can be converted to carbon, very few have been actually converted to structural carbon fibers. Besides PAN, mesophase pitch and rayon are the only other precursors that have produced carbon fibers. Mesophase pitch produces highly graphitic carbon fibers with high modulus, high thermal conductivity, but low strength. Rayon, on the other hand, uses naturally occurring cellulose, but its conversion to carbon is mechanistically limited to carbon fibers with poor strength and cannot be used in primary reinforcement applications. Rayon fibers are produced from wet spinning of cellulose that is dissolved in CS2 and NaOH [42,43]. Typically, rayon fibers do not have to be oxidized, but have to be stretched significantly during carbonization. Typically, the carbon fibers have a modulus of under 100 GPa and a tensile strength of under 1 GPa. The overall carbon yield from rayon precursors is only about 10e30%, when it is 40e50% for PAN, and about 70e75% for mesophase pitch. In contrast to the presence of significant polynuclear aromatic hydrocarbons in mesophase pitch that leads to a high carbon yield, the lack of aromatic structure within rayon results in this low carbon yield. The lack of aromatic structure in rayon also leads to a low graphitic crystallinity in rayon-based carbon fibers. As determined by wide-angle X-ray diffraction, integrated azimuthal profiles display a two-theta peak associated with the stacking of (002) layer planes at 24.5 , when PANbased fibers display it at about 25.6 . This indicates a low degree of graphitic content in rayon-based carbon fibers, which in turn leads to a fairly low thermal conductivity, at least within the family of carbon fibers. The low thermal conductivity has led to their use as thermal insulation materials in ultra-high-temperature environment such as rocket nozzles. Rayon-based carbon fibers are typically used with a phenoleformaldehyde matrix as a composite in contact with hot exhaust gases. Such materials ultimately ablate, but due to their insulative properties protect the inside metal components from melting or degrading. Even though rayon is derived from naturally occurring cellulose, ironically the chemical process used to make rayon utilizes solvents (xanthates) that are not environmentally friendly. Among other bio-derived feedstocks, lignin is an aromatic material that holds potential as a carbon precursor. Lignin encapsulates cellulose fibers in wood, and provides rigidity and strength to wood. The use of lignin as a precursor for carbon fibers dates back to 1960s. Carbon fibers have been produced from a mixture of lignin and a polymer plasticizer dissolved in an alkaline solution [54]. Since then, carbon fibers have also been produced without a polymer additive. Different types of lignin used for the purpose include steam-exploded

Handbook of Advanced Ceramics

lignin [55], organosolv lignin [56,57], and hardwood kraft lignin [58]. More recent interest in lignin is derived from the fact that it possesses an aromatic structure, which facilitates conversion to carbon. However, lignin also possesses a three-dimensional molecular architecture, which makes it intractable in its unmodified state, and must be broken down into smaller-molecular-weight components. Softwood kraft lignin is commercially available, and has been chemically modified into an acetylated form that renders it melt processable. To date, lignin-based carbon fibers have displayed tensile strengths limited to about 1.6 GPa and tensile modulus limited to about 100 GPa, significantly lower than their MP-pitch or PAN-based counterparts [59]. In summary, the recent trend in the use of bio-based, sustainable materials has revived interest in the production of carbon fibers from lignin-derived precursors, which in turn are derived from wood and other biomass. Because lignin is abundantly available in the biosphere, it holds promise for producing cost-competitive carbon fibers, whose applications will extend to larger industrial and transportation sectors to achieve energy efficiency from the use of lightweight carbon-based composite materials.

ACKNOWLEDGMENTS This material is based upon work partly supported by The Engineering Research Centers Program of National Science Foundation under NSF EEC-9731680, NSF 0823012, UTC/AFRL 08-S568-053-01-C1 / FA8650-05-D5807, and DOE-EPSCoR DE-FG02-07ER46364. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of Department of Energy, National Science Foundation, and other sponsors.

REFERENCES [1] Morgen P. Carbon fibers and their composites. CRC Press; May 2005. [2] Lavin JG, Boyington DR, Lahijani J, Nystem B, Issi JP. The correlation of thermal conductivity with electrical resistivity in mesophase pitch-based carbon fiber. Carbon 1993;31:1001e2. [3] Edie DD, McHugh JJ. Carbon materials for advanced technologies: chapter 4 e high performance carbon fibers. Elsevier Ltd.; 1999. pp. 119e138. [4] Crasto AS, Kim RY, Mistretta JP. The application of composites for the rehabilitation of concrete bridge infrastructure. Adv Comp Mater 2001;10:147e57. [5] Chung DDL. Electromagnetic interference shielding effectiveness of carbon materials. Carbon 2001;39:279e85. [6] Bahl OP, Shen Z, Lavin JG, Ross RA. Carbon Fibers. 3rd ed.; 1998. New York. [7] “Paris Air Show Report 2005,” High-Performance Composites, pp.24e28; September 2005. [8] Kurylko DT. BMW electric car using carbon fiber. Plastics News April 12, 2010.

Chapter | 2.8

Carbon Fibers

[9] Del Pero MS, “Market trends”, High Performance Composites, p. 9; May 2010. [10] Sloan J, “Carbon fiber market: cautious optimism,” High-Performance Composite; March 2011. [11] Toyoda M, Kaburagi Y, Yoshida A, Inagaki M. Acceleration of graphitization in carbon fibers through exfoliation. Carbon 2004; 42:2567e72. [12] Gupta A, Harrison IR. New aspects in the oxidative stabilization of PAN-based carbon fibers. Carbon 1996;34:1427e45. [13] Edie DD. The effect of processing on the structure and properties of carbon fibers. Carbon 1998;36:345e62. [14] Lavin JG. Fracture of carbon fibers. Fiber Fracture 2002:157e79. [15] Grassie N, Hay JN, McNeill IC. Thermal coloration and insolubilization in polyacrylonitrile. J Polym Sci 1962;56:189e202. [16] Takata T, Hiroi I, Taniyama M. Coloration in acrylonitrile polymers. J Polym Sci Part A 1964;2:1567e85. [17] Brandrup J, Peebles Jr LH. On the cromophore of polyacrylonitrile. IV. Thermal oxidation of polyacrylonitrile and other nitrile-containing compounds. Macromolecules 1968;1: 64e72. [18] Grassie N, McGuchan R. Pyrolysis of polyacrylonitrile and related polymers-III; thermal analysis of preheated polymers. Eur Polym J 1971;7:1357e71. [19] Fitzer E, Mu¨ller DJ. The influence of oxygen on the chemical reactions during stabilization of PAN as carbon fiber precursor. Carbon 1975;13:63e9. [20] Bashir Z. A critical review of the stabilization of polyacrylonitrile. Carbon 1991;29:1081e90. [21] Burland WJ, Parsons JL. Pyrolysis of polyacrylonitrile. J Polym Sci 1956;22:249e56. [22] Fitzer E, Mu¨ller DJ. Zur Bildung von gewinkelten leiterpolymeren in polyacrynitril-fasern. Makromol Chem 1971;144:117e33. [23] Olive´ GH, Olive S. Inter-versus intramolecular oligomerization of nitrile groups in polyacrylonitrile. Polym Bull 1981;5:457e61. [24] Coleman MM, Sivy GT, Painter PC, Snyder RW, Gordon III B. Studies of the degradation of acrylonitrile/acrylamide copolymers as a function of composition and temperature. Carbon 1983;21: 255e67. [25] Chen SS, Herms J, Peebles LH, Uhlmann DR. Oxidative stabilization of acrylic fibres. J Mater Sci 1981;16:1490e510. [26] Gupta A, Harrison IR. New aspects in the oxidative stabilization of PAN-based carbon fibers: II. Carbon 1997;35:809e18. [27] Grove D, Desai P, Abhiraman AS. Exploratory experiments in the conversion of plasticized melt spun PAN-based precursors to carbon fibers. Carbon 1988;26:403e11. [28] Daumit GP, Ko YS, Slater CR, Venner JG, Young CC, Zwick MM. US Patent 4,933,128; 1990. [29] Smierciak RC, Wardlow Jr E, Ball LE. US Patent 5,618,901; 1997. [30] Paiva MC, Kotasthane P, Edie DD, Ogale AA. UV stabilization route for melt-processible PAN-based carbon fibers. Carbon 2003;41:1399e409. [31] Mukundan T, Bhanu VA, Wiles KB, Bortner M, Baird DG, McGrath JE. Photocrosslinkable acrylonitrile terpolymers as carbon fiber precursors. Polym Prepr 2003;44:651652. [32] Naskar AK, Walker RA, Proulx S, Edie DD, Ogale AA. UV assisted stabilization routes for carbon fiber precursors produced from melt-processible polyacrylonitrile terpolymer. Carbon 2005; 43:1065e72.

153

[33]

[34]

[35] [36]

[37]

[38]

[39]

[40]

[41]

[42] [43] [44] [45]

[46]

[47]

[48]

[49]

[50]

[51]

[52]

White TL, Paulauskas FL, Bigelow TS. “System to continuously produce carbon fiber via microwave assisted plasma processing.” United States Patent 7824495; November 2, 2010. Chae HG, Choi YH, Minus ML, Kumar S. Carbon nanotube reinforced small diameter PAN based carbon fiber. Comp Sci Technol 2009;69:406e13. Diefendorf RJ. Polymers for fibers and elastomers. Chapter 12. ACS Symposium Series 1984;260:209e18. Mochida IY, Shimizu K, Korai Y, Sakai Y, Fujiyama S. Mesophase pitch derived from isotropic anthracene pitch produced catalytically with HF/BF3. Bull Chem Soc Jpn 1990;63:2945e50. Korai Y, Yoon SH, Oka H, Mochida I, Nakamura T, Kato I, Sakai Y. The properties of co-oligomerized mesophase pitch from methylnaphthalene and naphthalene catalyzed by HF/BF3. Carbon 1998;36:369e75. Dumont M, Chollon G, Dourges MA, Pailler R, Bourrat X, Naslain R, Bruneel JL, Couzi M. Chemical, microstructural and thermal analyses of a naphthalene-derived mesophase pitch. Carbon 2002;40:1475e86. Mochida IY, Korai Y, Ku CH, Watanabe F, Sakai Y. Chemistry of synthesis, structure, preparation and application of aromaticderived mesophase pitch. Carbon 2000;38:305e28. Kundu S, Ogale AA. Rheostructural studies on a synthetic mesophase pitch during transient shear flow. Carbon 2006;44: 2224e35. Park YD, Mochida IY. A two-stage preparation of mesophase pitch from the vacuum residue of FCC decant oil. Carbon 1989;27: 925e9. Fitzer E, Manocha LM. Carbon reinforcements and carbon /carbon composites. Springer-Verlag; 1998. Buckler JD, Edie DD. Carbonecarbon materials and composites. NASA Reference Pub.; 1992. McHugh JJ, Edie DD. The orientation of mesophase pitch during fully developed channel flow. Carbon 1996;34:1315e22. Barnes AB, Dauche FM, Gallego NC, Fain CC, Thies MC. As-spun orientation as an indication of graphitized properties of mesophasebased carbon fiber. Carbon 1998;36:855e60. Kundu S, Ogale AA. Microstructural effects on the dynamic rheology of a discotic mesophase pitch. Rheol Acta 2007;46: 1211e22. Mochida I, Yoon SH, Takano N, Fortin F, Korai Y, Yokogawa K. Microstructure of mesophase pitch-based carbon fiber and its control. Carbon 1996;34:941e56. Hamada T, Nishida T, Sajiki Y, Matsumoto M, Endo M. Structures and physical properties of carbon fibers from coal tar mesophase pitch. J Mater Res 1987;2:850e7. Hamada T, Nishida T, Furuyama Y, Tomioka T. Transverse structure of pitch fiber from coal tar mesophase pitch. Carbon 1988;26: 837e41. Matsumoto M, Iwashita T, Arai Y, Tomioka T. Effect of spinning conditions on the structures of pitch-based carbon fiber. Carbon 1993;31:715e20. Fathollahi B, White JL. Stabilization of mesophase matrix composites at oxidation temperatures below 270  C. In: Proc. Carbon’99; 1999. p. 362e3. Cho T, Lee YS, Rao R, Rao AM, Edie DD, Ogale AA. Structure of carbon fiber obtained from nanotube-reinforced mesophase pitch. Carbon 2003;41:1419e24.

154

[53] Ahn YR, Lee YS, Ogale YS, Yun CH, Park CR. Compressional behavior of carbon nanotube reinforced mesophase pitch-based carbon fibers. Fibers and Polymers 2006;7:85e7. [54] Charsley EL, Warrington SB, Cammenga HK. In: Thermal analysis: techniques and applications. Cambridge: Royal Society of Chemistry; 1992. [55] Sudo K, Shimizu K. A new carbon fiber from lignin. J Appl Polym Sci 1992;44:127e34. [56] Uraki Y, Kubo S, Nigo N, Sano Y, Sasaya T. Preparation of carbon fibers from organosolv lignin obtained by aqueous acetic acid pulping. Holzforschung 1995;49:343e50.

Handbook of Advanced Ceramics

[57] Kubo S, Uraki Y, Sano Y. Preparation of carbon fibers from softwood lignin by atmospheric acetic acid pulping. Carbon 1998;36: 1119e24. [58] Kadla J, Kubo S, Venditti R, Gilbert R, Compere A, Griffith W. Lignin-based carbon fibers for composite fiber applications. Carbon 2002;40:2913e20. [59] Baker FS, Baker DA, Gallego NC, “Utilization of sustainable resource materials for production of carbon fiber materials for structural and energy efficiency applications,” Abstract no. 421, Proc. Carbon 2010, Clemson, SC, July, 2010.

Chapter 2.9

Activated Carbon Fibers Angel Linares-Solano and Diego Cazorla-Amoro´s Dpto. de Quı´mica Inorga´nica e Instituto de Materiales (Department of Inorganic Chemistry and Institute of Materials), Universidad de Alicante (University of Alicante), Aptdo 99, 03080 Alicante, Spain

Chapter Outline 1. Introduction 2. Preparation of ACFS 3. Activated Carbon Nanofibers (ACNFs) and Activated Multiwalled Carbon Nanotubes (AMWCNTS)

155 156 160

1. INTRODUCTION Activated carbon fibers (ACFs) are porous carbons with a fiber shape and a well-defined porous structure which can be prepared with a high adsorption capacity. Although the ACFs are very promising materials, they do not still have a market as important as the activated carbons due to their difference in production costs. The main characteristics and advantages of the ACFs are the following [1e4]: (i) They have both high apparent surface area and adsorption capacity. (ii) They have fiber shape with small diameters, typically between 10 and 40 mm, which is a very important characteristic for applications requiring higher packing density (i.e. gas storage) [3,5]. More recently, similar materials with much smaller diameters (in a nm scale) have been developed, as it will be commented later on. (iii) ACFs are light materials and can be easily woven into different fabrics (i.e. cloths, felts, etc.). (iv) The poresize distribution of the ACFs can be narrow and uniform, being essentially microporous materials, although mesoporous ACFs can also be prepared. (v) The narrow diameter essentially eliminates mass transfer limitations, with the adsorptionedesorption rates being very rapid. Since the ACFs are fibrous materials that can be easily molded and woven, filters can be designed which do not have the settling and channeling problems of the conventional granular and powder activated carbons [1]. Due to their low hydrodynamic resistance, they can be used as thin cloths for the treatment of high flow of gases, which is very useful for control of gas-phase pollution [4]. The development of ACFs and activated carbon cloths is closely related to that of carbon fibers (CFs). This makes

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00010-1 Copyright Ó 2013 Elsevier Inc. All rights reserved.

4. Some Examples on ACF Applications 5. Conclusions Acknowledgments References

161 166 166 166

that the raw materials used for the preparation of ACFs be, chronologically, the same as for CFs. Thus, in 1966, viscose and acetate cloths were, like for CFs, the first materials used to obtain ACFs [6,7]. The low yield of the ACFs, and CFs, obtained from the above precursors, oriented the research toward the seeking of other raw materials for the preparation of cheaper CFs and ACFs and with higher yields. In this way, ACFs were prepared from 1970 using lignin (with the brand of Kayacarbon ALF), polyvinyl-chloride [8] (i.e. Saran polymer, already used to obtain activated carbons), and phenolic precursors [9]. The high yield and the good mechanical properties of the ACFs obtained make these precursors very useful for this application. In fact, Economy and Lin [10] developed ACFs from a phenol formaldehyde precursor which are commercialized since 1976 with the name of Novolak. In 1980, Kuray Chemical Co. Ltd. commercialized ACF from phenolic resin with the name of Kynol [11]. The preparation of PAN-based ACFs was initiated in 1976 by Toho Rayon Co. Ltd. [12] and the use of pitch to obtain ACFs started in 1985, and were commercialized by Osaka Gas Co. Ltd (Ad’all) [13]. Due to the low price of the pitch (petroleum and or coal tar) and high yield of the ACFs obtained, the manufacture of these pitch-based ACFs has increased considerably, nowadays being one of the main precursors of ACFs [1]. The research on ACFs is not different from what is usual for other activated carbon materials. It focuses on understanding the preparation process, on the characterization of the materials, and the analysis of their performance in given applications. A literature search on this topic gives us more

155

156

than eight hundred contributions, taking into account only the papers published in journals. The research on ACFs started more than thirty years ago regarding its preparation (as it has been detailed above). However, most of the work done concentrates mainly in the last twenty years and it is essentially focused on their characterization and applications. More recently, other forms of ACFs having diameters in a nm scale have been developed, such as activated carbon nanofibers (ACNFs) or activated carbon nanotubes (ACNTs) [5,14e20]. Thus, this chapter on ACFs will cover the most important aspects concerning the preparation routes and the most important applications of these materials (mainly ACFs and also ACNFs and ACNTs).

2. PREPARATION OF ACFS Once the precursor (i.e. pitch, polymer, etc.) is transformed into a fiber shape, by a suitable spinning process and is carbonized after a proper stabilization stage, the activation of the resulting CFs is needed to increase its adsorption capacity. The starting points for the activation of the CFs precursors are not different compared to those for conventional granular or powder carbon precursors. As it happens in all of them, the starting porosity of the carbon material to be activated has a relevant role in its subsequent activation. To prepare ACFs, the precursor and the method of preparation need to be conveniently selected. These two factors have great importance as they determine the final porous structure of the ACFs. For a given precursor, the main stage determining the porous structure is the method of activation. The objective during the activation is both to increase the number of pores and to increase the size of the existing ones, so that the resulting porous carbon has a high adsorption capacity. The preparation of ACFs can be achieved by first selecting the appropriate precursor and controlling its pyrolysis step and then controlling their subsequent activation process by any of the following two methods: (i) using a reactant gas in the so-called physical activation and (ii) using a chemical agent in the so-called chemical activation. The preparation of ACFs by physical activation includes a controlled gasification of the CFs at temperatures between 800 and 1000  C with an oxidant gas, so that carbon atoms are being removed selectively. The removal of the outer and less ordered carbon atoms leads to the creation of micropores and/or the widening of their size, what results in an increase of its pore volume. Thus, for a given precursor, the pore-size distribution in the ACFs depends on the preparation conditions (mainly temperature, time, and gas flow), the activating agent used, and the presence of catalysts. Some representative

Handbook of Advanced Ceramics

examples of the influence of the experimental conditions mentioned before can be found elsewhere [4,21e34]. In order to have an efficient activation process, the reaction must take place inside the carbon fibers, at least predominantly, compared with the reaction occurring outside. If only external reaction takes place, the obtained material does not develop porosity. However, if the reaction occurs inside the fiber, there is porosity development: the higher the amount of carbon removed, the higher the porosity development. Carbon dioxide and steam are the activating agents most commonly used, which reactions with carbon are endothermic. These gases react with the carbon atoms in the precursor according to the following reactions: C þ CO2 4 2CO C þ H2 O 4 CO þ H2

DH ¼ 159:0 kJ=mol DH ¼ 118:5 kJ=mol

Although the activation with carbon dioxide or steam produces essentially microporous ACFs, strong differences have been found between these two activating agents regarding the porous texture and the mechanical properties of the ACFs [21,32]. The addition of metals such as cobalt, silver, rare earth metals, or platinum either to the starting pitch (followed by spinning, stabilization, and carbonization) or to the carbon fiber, followed by gasification with steam, allows the preparation of ACFs with a significant content of mesoporosity [24,28e31,33,34]. The chemical activation process consists of mixing a carbonaceous precursor with a chemical activating agent, followed by a pyrolysis stage [35e39]. The material after this stage is richer in carbon content and presents a much ordered structure and, after the thermal treatment and the removal of the activating agent, has a well-developed porous structure. Different compounds can be used for the activation; among them, KOH, NaOH, H3PO4, and ZnCl2 have been reported in the literature [26,27,35e39]. The chemical activation presents advantages over the physical one, that can be summarized as follows: (i) the chemical activation uses lower temperatures and pyrolysis time, (ii) it usually consists of one stage, (iii) the yields obtained are higher, (iv) it produces highly microporous activated carbons, and (v) it is a suitable method for applying it to materials with a high ash content [36e39]. On the other hand, the chemical activation presents disadvantages such as the need of a washing stage after the pyrolysis and the corrosiveness of the chemical agents used. Although the work done on physical activation of carbon fibers is wide, the research on chemical activation of carbon fibers is scarce and is mainly restricted to the use of some alkaline hydroxides as activating agents [40,41]. The chemical activation must be done under wellcontrolled experimental conditions in order to avoid

Chapter | 2.9

Activated Carbon Fibers

157

FIGURE 1 N2 adsorption isotherms at 77 K of ACFs prepared by (a) KOH activation and (b) CO2 activation.

destruction of the fiber shape. The resulting ACFs are essentially microporous materials (i.e. pore size below 2 nm), although differences exist depending on the activating agent used and the starting CFs. In the activation by metal hydroxides, the main variables affecting the final porous texture are hydroxide/ carbon ratio, heating rate, temperature, and time of pyrolysis. Moreover, there are two additional parameters that have recently been reported [37e39]: nitrogen flow rate and the washing stage (washing with water or washing with hydrochloric acid), which have an important role in the porosity development.

An example of the porosity development upon activation of a given CF precursor (Donacarbo S-241, provided by Osaka Gas, Japan) can be seen in Fig. 1 and Table 1 [42]. The figure presents the N2 adsorption isotherm of the ACFs obtained by KOH activation at 750  C using three KOH/CFs ratios (Table 1 and Fig. 1a) and ACFs obtained by CO2 activation at 825  C at three activation times (Table 1 and Fig. 1b). It has to be pointed out that: (i) the starting CF has no adsorption of N2 at 77 K due to well-known diffusion problems [43e47] and (ii) the CF precursor used develops very well its porosity upon activation. All the isotherms are of type I, typical of microporous solids in which most of the

158

Handbook of Advanced Ceramics

TABLE 1 Porous Texture Data and Yield of ACFs Prepared by KOH and CO2 Activations Sample

SBET/m2/g

VDR(N2)/cm3/g

VDR(CO2)/cm3/g

Yield/%

KOH 1:1

683

0.31

0.32

83

KOH 3:1

1288

0.59

0.57

74

KOH 6:1

1894

0.89

0.65

56

CO2 4 h

351

0.16

0.20

85

CO2 24 h

1118

0.52

0.48

52

CO2 48 h

1659

0.71

0.54

40

pores have a size below 2 nm. At high activation yields, the porosity is mainly due to narrow micropores ( dSL > dISO), consistent with crystallographic parameters. Table 5 gives comparison of the different processes used for making C/Cs. The type of hydrocarbon has been observed to influence the texture of carbon in CVD carbon composites. Kimura et al. [33] have observed cone-like columnar structures in a propane-derived pressure-controlled carbon matrix, the dependence of which on processing parameters is shown in Figure 7.

2.2.1. Macroscopic Features With respect to minimechanical features, C/Cs are a buildup of fiber bundles containing filaments, thin interfilament matrix, interbundle-matrix-rich regions, and macropores. The latter includes the voids formed as a result of imperfect fiber stacking or processing and the bubble pores due to evaluation of gaseous by-products during

179

TABLE 4 Crystallite Dimensions of Carbon Matrix in As-Deposited and Heat-Treated C/Cs Made by CVD Routes Condition Material (nm)

Density (g/cc) d002 (nm)

Lc (nm)

RL-PAN

As-deposited

1.83

0.344e0.345

8.2e9.2

Heat-treated

1.86

0.337

31e46

As-deposited

1.70

0.346e0.347

5.8e6.3

Heat-treated

1.58

0.337e0.342

9.1e15.2

ISO-PAN As-deposited

1.47

0.344e0.345

5.4e9.1

Heat-treated

1.45

0.341e0.345

7.0e11

SL-PAN

RL, rough laminar; SL, smooth laminar; ISO, isotropic; PAN, Polyacrylonitrile; d002, interlayer spacing; Lc, stack height.

carbonization, dry zones, and shrinkage cracks, etc. [6,42,51,52]. Figure 8 is a schematic representation of different types of matrixes present in C/Cs [52]. Interfilament spacing in a bundle may also affect the microstructure of the matrix in that region. In C/Cs made with pitches as matrix precursor by lowpressure processing, voids, bubble pores, and the matrix macrosequences are observed. Voids are features of the microstructure of C/Cs, which are mainly generated during the initial process of the prepreg and during the subsequent steps of carbonization. Polarized light microscopy shows distinct three types of voids in undensified composites: elongated cracks between fiber bundles in both lateral (parallel to the plane of the fiber bundles) and transverse (perpendicular to the plane of the fiber bundles) directions and isolated devolatization pores between or within the bundles (Figure 9). Optical micrographs of undensified C/Cs showed different types of voids. It showed elongated transverse cracks, which are produced by carbon matrix shrinkage during the carbonization treatment and by misalignment of the fibers during initial stacking of the coated fiber bundles. This crack runs parallel to the plane of the fiber bundles. It also showed lateral cracks, which can be devolatilization pores joined together by cracks of a single pore. These cracks run perpendicular to the plane of fiber bundles. The size of all these voids ranged from as low as 30 mm  30 mm to as high as 1400 mm  700 mm. Densification exhibits region of distinct primary and secondary carbon matrixes (Figure 9b). The type of reinforcement and pitch-matrix precursor used is found to be influencing the structure of C/Cs (Figure 10). The high quinoline content pitch as the matrix displayed mosaic-type structure throughout the pyrolysis stage of the composite, while the low quinoline-content

180

Handbook of Advanced Ceramics

TABLE 5 Comparison of Characteristic Features and Properties of Carbon Matrixes from Vapor-Phase, Pitch, and Resin Precursors Type of Carbon Matrix Characteristic Feature or property

CVD 3

Pitch

Resin 5

Density

High, approx. 2000 kg m ,

High, increases with carbonization pressure

Low 1300e1600 kg m

Carbon yield

C directly deposited; no further thermal degradation

Varies according to pitch composition 50e80% w/w

About 50% for phenolics, increasing to 85% for polyphenylenes.

Porosity

Low, expect for isotropic form; laminar fissures

Macrosized gas entrapment pores plus shrinkage and thermal stress fissures.

High microporosity (pore diameter 200 nm after 3000  C HTT

Highly graphitizable La and Lc usually less than 100 nm

Normally nongraphitizing

Purity/ composition control

Controlled by gas composition; enables other elements to be incorporated into the deposit (e.g. Si, B)

Controlled by source; can be of high purity; not easy to incorporate other elements expect as powders

As for pitch

Reactivity to oxidizing gases

Very low reactivity for highly oriented Pyrographite

Low reactivity decreases with HTT

Usually high reactivity due to micropore network; decreases with HTT but still relatively high

Thermal expansion

Depends on preferred orientation; can be highly anisotropic approaching values for the crystal in the two major directions

Depends on domain orientation; expansion partly accommodated by lamellar cracks; 1e5  10 6 K 1

Isotropic in bulk; ~3  10

Thermal and electrical conductivity

Determined by referred orientation, approaching single-crystal graphite values

Depends on domain orientation, HTT, and internal porosity, increasing with HTT

Isotropic in bulk

Young’s modulus

7e40 GPa depending on the structure

5e10 GPa depending on grain size, porosity, and degree of graphitization but up to 14 GPa for very fine-(1 mm) grain size materials

10e30 GPa

Strength

Depends on microstructure and degree of graphitization; 10e500 mPa

Depends on porosity and pore geometry’ 10e50 mPa for most polycrystalline carbon/graphites rising to 120 mPa for very fine-grain (1 mm) size materials

Approximately 8e150 mPa for glassy carbons, depending on porosity

Failure strain

0.3e20% depending on structure; Higher values at highest deposition/ HTT

Up to about 0.3% depending on grain size and degree of graphitization

Up to about 0.4% (largest values for glassy carbons)

Lc, stack height; La stack width; HTT, heat treatment temperature; BSUs, basic structural units.

6

K

1

Chapter | 2.10

CarboneCarbon Composites

181

FIGURE 7 Development of CVD structure under different processing parameters.

pitch and AR-mesophase pitch show flow-type structure. The matrixes in both bulk as well as interfilament show a well oriented lamellar-like structure. Although matrixes in all these composites exhibit good orientation of the carbon matrix, the extent of orientation is smaller in the case of high-strength PAN-based carbon fiber-reinforced composites than that seen in composites with intermediate modulus Dona F-180 pitch carbon fiber and ultrahigh modulus P-100 fiber-reinforced composites. C/Cs developed with AR-mesophase pitch, containing milled intermediate modulus-pitch fibers, exhibit disturbed structure of the matrix. In graphitized C/Cs with HT-PAN-based carbon fibers comparatively good bonding with the matrix is observed as against ultrahigh modulus-carbon fiber-reinforced composites. 2D C/Cs made with woven cloth as reinforcement exhibit additional significant microstructural features due to the yarns running in two directions (warp and fill), passing over and under each other and imparting a wavy

nature to them. Although these composites are made by prepregging and compaction technique similar to UD composites, a significant portion of the composite remains as unreinforced matrix-rich region such as interyarn pockets, bundle cross-over region, interplay regions, etc. The configurations of these regions will depend on the weave type (plain, twil, or satin weave), stacking, compaction technique, etc. [52e55]. In 2D composites, especially in thick ones, the heating rate has to be chosen carefully to avoid delamination of the composites during heat treatment. Similar to UD and 2D composites, MD composites also contain shrinkage cracks and thermal stress cracks. Since the latter composites are composed of a rigid fiber architecture that restrains the overall shrinkage, cracks are likely to be more pronounced in them than in UD and 2D composites. Another striking feature of the MD carbonecarbon structures is the billowing of the fiber bundles, as shown in Figure 11 [56].

FIGURE 8 Schematic presentation of different type of carbon matrix in C/Cs.

182

Handbook of Advanced Ceramics

FIGURE 9 Optical micrographs of C/Cs (a) skeletons and (b) densified composites.

Thermosetting polymer route results in an amorphous carbon matrix. However, in composites made with high-strength carbon fibers or surface-treated carbon fibers due to strong fiber/matrix bonding at polymer stage and differential pyrolysis shrinkage in fibers and matrix, thermal stresses are generated at the fiber/matrix interface during carbonization of the composites, causing stress graphitization of the matrix in the composites. The graphitic structure starts developing at a temperature of 2000  C [42,57] (Figure 12). The interbundle matrix and the matrix far away from the fibers remain isotropic in nature. However, in the case of high fiber-volume content composites, the shrinkage stresses generated at the interfaces propagate further into the matrix, which on heat treatment to 2000  C and above cause stress graphitization of the complete matrix [52,53,58]. This subject has been of great scientific interest, especially correlating fiber/ matrix bonding and resulting orientation of graphite basic structural units in the matrix. The technological importance of this phenomenon has been to control the fiber/matrix bonding so that low-priced carbon fibers can be used for making C/Cs for mechanical and general purpose application.

FIGURE 10 Optical micrographs of the composites made with different types of fibers and composites.

Chapter | 2.10

CarboneCarbon Composites

183

FIGURE 11 Billowing of fibers during processing of MD composites.

Another prominent feature of the matrix microstructure, especially in the lamellar regions, is the extensive kinking of the layer planes. Kinking is typically observed in the lamellar matrix around fibers with uneven surfaces with the kinks conforming to the depression and asperities on the fiber surface. Kinked matrixes are also observed in composites made with excessively surface-treated carbon fibers. In contrast to the microstructure of thermosettingresin-derived carbon, that of thermoplastic-pitch-derived carbon is controlled by nematic-liquid crystal (mesophase)

formation during liquid-phase pyrolysis of the pitches. The flow of the oriented molecules of the liquid crystals during their formation and subsequent deformation results in the introduction of disclination in the structure. During pyrolysis of the pitch precursor in the fabrication of C/Cs, the carbon filament forms a circular sheath several microns thick around each filament. The parallel alignment of the filaments and mesophase sheath force the disclination lines to be parallel to the fibers. However, the strength and type of disclination is determined by the local arrangement or

FIGURE 12 Optical micrographs of C/Cs made with resin-derived carbons and heat-treated to different temperatures.

184

Handbook of Advanced Ceramics

FIGURE 13 SEM micrographs of oriented carbon matrix around carbon fibers in C/Cs.

pattern of the arrays of filaments [59,60]. The sheath-like orientation of the pitch-carbon matrix has been observed in all types of C/Cs made with round-carbon fibers irrespective of PAN or pitch (Figure 13) [6,48]. However, due to the relatively less active surface in pitch-carbon fibers, the matrix debonds from the fibers and the majority of shrinking cracks are developed at the fiber/matrix interface, instead of in the bulk matrix.

3. PROPERTIES OF C/CS C/Cs, in true sense, cover a large range of materials with targetted properties. The properties of interest are strength and stiffness, fracture toughness, frictional properties, thermal conductivity, and resistance to oxidation at high temperatures. The operating mechanisms for these properties are quite different especially in such multiphase composite materials. The mechanical properties of the constituents and their volume fraction, bonding, and crack propagation mechanism control the mechanical properties of the composites, whereas the thermal properties are governed by thermal transport phenomena. Moreover, the constituents, both reinforcement and matrix, are likely to undergo a change in properties during processing as influenced by heat-treatment temperature, differential dimentional changes, thermal stresses, etc. All these factors influence the ultimate properties of the composites. It is difficult to cover completely all these aspects in this article. However, a general trend of the most fascinating properties of C/Cs are given below.

3.1. Mechanical Properties of C/Cs The mechanical properties of advanced composites are generally related to the properties, volume fraction, and direction of the fibers relative to the direction of the applied stress. Considering these factors, the mechanical properties of PMCs or, in a broader sense, of ductile matrix composites are calculated by a simple rule of mixture.

However, in C/Cs, the matrix has a failure strain of less than 0.5%, which is lower than that of the fibers [2,6]. Consequently, theories of reinforcement that have been developed for commercial composites with high-strain matrix systems and strong fiber/matrix bonding do not apply to C/Cs. Moreover, matrix structure (both at mini- and microscales) and fiber/matrix bonding in these composites differ depending on the type of reinforcements and processing routes. Accordingly, different principles will prevail for the mechanical properties of C/Cs processed through various routes. Furthermore, the data concerning mechanical properties, which are available in the literature, are based on the constituents and the processing condition adopted by individuals.

3.1.1. Strength of the Somposite The prediction of the strength matrix composites is rather complicated since it is governed by the mechanism responsible for crack initiation and crack propagation. However, in brittle matrix composites, with the matrix having a lower failure strain than the fibers, crack initiation takes place in the matrix at much lower stress levels, and the fiber strength is not fully utilized. Therefore, in order to have maximum utilization of the fiber strength, the failure strain of the matrix must be increased. Further, the strength and fracture of C/Cs are governed by CookeGorden [61] theory for strengthening of brittle solids, which states that if the ratio of the adhesive strength of the interface to the cohesive strength of the solid is in the right range, a large increase in strength and toughness of otherwise brittle material is achieved. Extensive work has been done in achieving the translation of highest possible fiber properties in C/Cs. Composites with strong fiber/matrix bonding fail catastrophically without fiber pullout, while those with controlled interface fail in the mixed tensile-cumshear mode exhibiting high strength. A generalized view of the effect of processing temperature, the fracture mechanism, and the strength of different types of C/Cs is given in Figure 14.

Chapter | 2.10

CarboneCarbon Composites

185

FIGURE 14 Stressestrain curves and fracture of C/Cs made with surface treated- and nonsurface- treated carbon fibers and heat-treated to different temperatures.

Like with any composite material with a polymer matrix, in C/Cs too, the fiber layup architecture and size of the carbon control the ultimate strength of the composites. This is seen in Figure 15.

3.1.2. Mechanical Properties of Composites with Pyrolytic Carbon Matrix In C/Cs with pyrolytic carbon, tensile strengths in the fiber direction of 1000e1400 MN/m2 have been reported, which correspond to 80e90% of the fiber properties’ translation [62,63]. At lower heat treatment temperature (HTT < 1500  C), composites with high strength have

been obtained, while the opposite trend is observed at HTT > 1500  C. The type of CVD technique chosen for composite fabrication has also been found to influence the mechanical properties. Composites with 50% fiber volume contents and made by the isothermal process are reported to exhibit much higher strength than those made with the temperature- or pressure-gradient technique. CVD-derived ISO and SL carbon matrixes are strongly bonded to the reinforcing carbon fibers. As-deposited RL and SL carbons do not contain microcracks. However, after heat treatment of the composites, circumferential microcracks develop in the SL matrix as well as at the

FIGURE 15 General trends in mechanical properties of C/C composites made with different fiber layup.

186

Handbook of Advanced Ceramics

FIGURE 16 StresseStrain curves of different C/Cs made with different CVI C/Cs.

interface between RL and SL due to anisotropic thermal contraction [63]. Both the strong fiber/matrix bonding and microcracks in the matrix result in low-strength C/Cs. Because of this, RL structures are preferred for making high-performance C/Cs. As-deposited RL composites posses a similar or somewhat higher flexural strength and modulus than SL composites. Upon heat treatment to graphitization temperature, SL matrixes exhibit a higher degree of graphitization with the others following in the order RL > ISO. With heat treatment temperature, there is only marginal change in properties of the latter composites. Kimura et al. [33] have studied the fracture behavior of CVD multiphase C/Cs. As seen in Figure 16, composites with ISO þ Rough Columnar (RC) matrix fracture predominantly in tensile mode without fiber pullout. A crack initiated in RC carbon matrix propagates perpendicular to the fiber axis without

FIGURE 17

any deflection owing to the strong interface between RC and ISO structures as well as between the carbon fiber and ISO matrix. These composites also exhibit a high compressive strength and the crack once initiated, encounters multiple deflections due to microcracks in the Smooth Columnar (SC) matrix. These composites posses a somewhat lower interlaminar shear strength but higher fracture energy. The conclusions derived from UD composites have been found to hold true also for CVD felt, 2D, and MD C/Cs [6,8e11,24]. Moreover, the strength of the composites made with highemodulus nonsurface-treated carbon fibers increases with further densification of the composites, but it decreases for those made with high-strength carbon fibers. Similar effects are observed when carbon fibers of the same high modulus but with different surface functionalities are used as reinforcement [6]. Contrasting differences are observed also in their fracture behaviors. Composites made with strong fiber/matrix bonding fail catastrophically with no fiber pullout (Figure 17), while others fail in mixed tensile-cum-shear mode with fiber pullout depending on the extent of bonding [42,52,57]. However, upon further heating of the composites to graphitization temperature, the strengths of the two composites exhibit different trends. Composite processing strong fiber/matrix bonding and low strength at the carbonized stage exhibit higher strengths at the graphitized stage and vice versa. Based on the experience with different fiber/matrix systems, a general trend can be derived as shown in Figure 17. The fracture behavior of the composites also changes with heat treatment. This is evident from stressestrain diagrams and scanning electron micrographs of the fractured composites. The change in the strength pattern on heat treatment is a consequence of the change in failure modes. In the temperature range of 1200e2200  C, a mixed-mode failure prevails. The

Generalized trend of crack propagation in brittle matrix composites.

Chapter | 2.10

CarboneCarbon Composites

187

heat treatment in this temperature range weakens the fiber/matrix interface resulting in crack blunting. This continues until an optimum level is reached where the interfacial bond strength is weak enough for deflection of the crack and strong enough to effectively utilize the stresstransfer capabilities of the matrix. As a net result, the composites fail in mixed mode with low fiber pullout. On further heating to 2500  C and above, the graphitization of the interfilament matrix increases. Multidimensional C/Cs exhibit pseudoplastic fracture behavior at room temperature with strains up to 5%, which is characteristically different to that of UD and 2D reinforced C/Cs. This difference attributes to the presence of reinforcing fibers in different direction and the continuous crack structure of the matrix whose geometry is dependent on the fiber array.

3.1.3. Sandwich Composites Sandwich composites are mix-form composites made of predetermined alternating layers of UD and felt- or clothbased prepregs. The purpose of making sandwich composites is to increase the transverse strength of C/Cs while retaining as much axial strength as possible, and the exact strength values, however, will depend on the type of reinforcement used and the ordering adopted for making the sandwich [6].

3.1.4. Fracture Toughness of C/Cs In addition to strength and stiffness, another property of considerable importance from a structural point of view is toughness, i.e. resistance to crack propagation. Carbon as such, both in bulk as well as in fibrous form, is brittle with a fracture toughness (energy required to propagate a crack) in the range of 0e200 J/m2. However, if carbon fibers and carbon matrix are combined to form composite materials with optimum fiber/matrix bonding, the composites will exhibit a higher toughness and work of fracture than carbon

TABLE 6 Fracture Toughness of Some C/Cs

Material System

Vf%

Sample Size (mm)

Felt/pitch

25

1.85.0

83  8

Felt/pitch

33

1.85.0

100  8

Felt/CVD

9

9.59.5

34  4

UD/pitch

50

4.84.8

15  2

Fabric/pitch

35

210

15  2

Angle ply/CVD

NA

NA

20  3

K (Nm m3/2)

in monolithic form (Table 6). C/Cs exhibit good fatigue and creep resistance too. [64]

3.2. Mechanical Properties at High Temperature Figure 18a and b shows typical cases of mechanical properties of various C/Cs at elevated test temperatures [6]. In all of the composites, the strength as well as the modulus at elevated temperatures is higher than that at room temperature. At high temperatures, several characteristic phenomena responsible for the mechanical properties of the composites take place simultaneously. Annealing of gross defects result in an improvement of mechanical properties. The closer of stress-distributed microcracks due to thermal expansion of fibers and matrix too change the fracture behavior of the composites [9,10]. As a combined result of all of these phenomena, the C/Cs exhibit increase in strength and stiffness by 10e 60% above 1000  C. The compressive strength has been observed to increase by as much as 100%. The tensile and comprehensive modulus of the composites decreases with temperature above 1600e2000  C. The fall in the modulus can be explained on the basis of the symmetric potential curve of graphite crystals. With an increase in temperature, the average atomic distance is increased, resulting in a decrease in the modulus. Like strength, the work of fracture of C/Cs also increases with test temperature up to 1800e2000  C. As mentioned above, the structural changes at high temperature influence the stress transfer mechanism in the composites. Senet et al. [65] have studied the fracture behavior of 2D discontinuous carbon-fiber-felt-reinforcement phenolic resin and pitch precursor C/Cs at elevated temperature using both chevron-notch and straight-notch single-edge beam specimens. They observed that Kic values of the chevron-notch specimens increased slightly with the test temperature through 1650  C, while the straight-notch values remained almost constant throughout the temperature range up to 1650  C. They suggest that fiber debonding, matrix damage, and fractional sliding during pullout are the critical constituents of an effective wakezone process responsible for the toughening of such composites. Composites made with PAN-carbon fibers exhibit a much higher increase in fracture toughness than other composites [66]. The mechanical behavior of C/Cs at high temperature discussed above holds true for UD, bidirectional, and multidirectional reinforced C/Cs made with continuous fibers or felt and liquid- as well as gas-phase-derived carbon matrix. The absolute strength values may vary depending upon the configuration. C/Cs are susceptible to oxidation at high temperature if not protected. Therefore, for the reliable performance of

188

Handbook of Advanced Ceramics

FIGURE 18 Change in mechanical properties of C/Cs with temperature.

a carbonecarbon component, it is essential to know the effect of oxidation on both the mechanical and thermal properties of the composite. Although this is an important aspect, not much information is available in the literature. Oxygen attacks preferentially the less-ordered carbon at matrix-rich region, bundle/bundle interfaces, or fiber/ matrix interfaces. Accordingly, the fracture behavior and the mechanical and thermal properties will be altered. Thermal conductivity will be reduced due to the porosity generated in the composites. The change in mechanical properties are governed by the preferential attack by oxygen as stated above [65,67]. This will have a severe effect if the weight loss is due to attack on the fiber/matrix interface and is more than 5% [65], which will result in a drastic decrease in the strength of the composites. If oxidation takes place at low temperatures (400e500  C) and is less than 2%, it may alter the fiber/matrix or matrix/ matrix interfaces, resulting in change of fracture to mixedmode type and, hence, an improvement of mechanical properties [68]. However, in practice, the application temperatures are much higher than 500  C, and the oxidation weight loss is also higher than 5%. A general trend of the effect of air oxidation on the mechanical properties of C/Cs is shown in Figure 19 [65]. Similar trends are observed for CVI-based C/Cs [65,69]. CVD pyrocarbons-based C/Cs too exhibit a range of high fracture toughness (20e100 Nm m3/2) and good creep and fatigue resistance. When subjected to high temperature testing, C/Cs have been found to exhibit about 10e20% increase in mechanical properties at 2000  C under inert atmosphere. However, in air, the properties drop down to 10e20%, depending on the temperature and time i.e. the

FIGURE 19 Decrease in mechanical properties of the composites with oxidation.

weight loss. Carbon is known to be a stable and well-sought material for nuclear application; so are carbonecarbon compositeC/Cs. Under neutron irradiation at low fluence level of 1021, these composites exhibit an increase in strength and fracture toughness by 20e30% and Young’s modulus by about 30% [9,70]

3.3. Thermal Properties of C/Cs 3.3.1. Thermal Conductivity C/Cs being heterogeneous structure consisting of fibers, matrix, and pores, with the former two having varieties of microstructure, the estimation of thermal transport

Chapter | 2.10

CarboneCarbon Composites

189

FIGURE 20 Schematic presentation of factors contributing to transport mechanism in C/Cs. (For color version of this figure, the reader is referred to the online version of this book.)

properties of these composites becomes complex, as schematically represented in Figure 20. Carbon fibers and the matrix may exhibit a variety of microstructures depending on the precursor and processing condition. Obviously, these microstructures will posses varying thermal properties. The thermal transport properties in the direction perpendicular to the fibers are difficult to predict. This is because the heat flow does not follow

a simple path. The fibers and the pores serve to disrupt heat flow in the carbon matrix, thus decreasing thermal conductivity. The thermal conductivity data available in the literature generally pertain to products made by the manufacturers. Very few groups have carried out a systematic evaluation of the thermal conductivity of C/Cs. A general comparison of thermal conductivity of C/Cs with different fiber/matrix combination is given in Figure 21 [6]. In

FIGURE 21 General trends in thermal conductivity of different C/Cs. For color version of this figure, the reader is referred to the online version of this book.

190

Handbook of Advanced Ceramics

900 Thermal Conductivity W/m-K

800 700

TYPE FIBER

600

– UHM

500

O – HM

400 300 200 100 0 0

200

400

600 800 1000 Temperature °C

1200

1400

FIGURE 23 Thermal conductivity of carbonecarbon composite made with different pitch-based carbon fibers. For color version of this figure, the reader is referred to the online version of this book.

FIGURE 22 Thermal conductivity of CVD C/Cs.

contrast to thermosetting-resin-derived C/Cs, pitchprecursor-derived C/Cs exhibit very high thermal conductivity. This is also evident from Figure 21. The conductivity in the fiber direction increases markedly with the number of densification cycle. The absolute value is further increased on graphitization of the composites at 3000  C [71,72]. Similar effects of the matrix precursor on the thermal conductivity of C/Cs have been observed in 2D composites [72]. A CVD-derived carbon matrix may exhibit a wide range of microstructures from isotropic to highly anisotropic RL structure. Accordingly, the thermal conductivity of

FIGURE 24

C/Cs with a CVD matrix will also vary with the matrix microstructure and heat treatment temperature (Figures 22 and 23) [71e74]. The Thermal conductivity of heat-treated composites are found to be higher than that of composites, which have not been heat treated and is highest for RL structure. For most of the carbonecarbon products made from CVD carbon matrix, RL-type microstructure is preferred. In CVD-derived C/Cs too, irrespective of the matrix precursor, the thermal conductivity is observed to increase with increasing heat-treatment temperature and decrease with increasing measurement temperature. The decrease in thermal conductivity with increasing test temperature is observed in all C/Cs preheat treated to that temperature irrespective of the fiber type and architecture (Figure 24) [74]. The extant of variation depends on the matrix precursor and fiber/matrix combinations. The effect is maximum for very high conducting C/Cs.

Change in thermal conductivity of C/Cs with test temperature.

Chapter | 2.10

CarboneCarbon Composites

FIGURE 25

191

Thermal conductivity of mesophase pitch-fiber/pitch-carbon composites.

As discussed above, high thermal conductivity C/Cs can be produced by choosing the proper combination of reinforcing fibers and the matrix precursor. C/Cs with very high thermal conductivity in the order of 850 w/mK at room temperature have been developed using the ultrahighmodulus pitch-based carbon fiber P-130 and mesophase pitch as matrix precursor [74] (Figure 25). By using ribbonshaped mesophase pitch carbon fibers in pitch precursor carbon matrix, C/Cs have been developed with a high thermal conductivity in the directions parallel and perpendicular to the fiber direction [48,75,76]. The higher thermal conductivity of the composites transversing to the fiber direction suggests a considerable contribution by the matrix to the transverse thermal conductivity of ribbonshaped carbon-fiber-reinforced carbon matrix composites. Current interest has been to use hybrid matrix derived from highly aromatic resins and pitches to develop high density, high conductivity C/Cs [77]. Using highly graphitic vapor-grown carbon fibers with resin/pitch-based carbon matrix followed by CVI, a thermal conductivity of 824 w/mK in the direction of the fibers has been reported. Though these composites exhibit highly anisotropic character and low conductivity in transverse directions, there is always scope of improvement by varying the fiber architecture and their addition in different forms and ways.

3.4. Thermal Expansion of Composites Graphite is known to be highly anisotropic with pyrographite exhibiting a thermal expansion anisotropy of about 20. Accordingly, carbon fibers also exhibit a Coefficient of thermal expansion (CTE) range that depends on the precursor and heat treatment temperature. In C/Cs, thermal

expansion in the direction of reinforcement is mainly controlled by that of the reinforcing fibers, whereas in the direction perpendicular to it, the type and content of matrix and porosity have a major influence. Though coefficient of thermal expansion of the composites is more dictated by the fiber orientation, typically, it is 1 ppm/oC in the fiber direction and 6e8 ppm/ C in a direction perpendicular to the fibers. Further, composites heat-treated to graphitization temperatures exhibit a lower thermal expansion in both directions. The ratio in the two directions increases with the crystalline order of the fibers. The CTE in 3D C/Cs is lower than in UD or 2D C/Cs [6]. Moreover, composites with a high fiber volume content and low porosity will exhibit a higher thermal expansion in the perpendicular direction than those with a low fiber volume content and a higher porosity.

3.5. Thermal Shock Resistance of C/Cs Thermal shock resistant of a material is an additional required thermal property of material intended for structural application at high temperature, specifically under cyclic varying temperature condition. Graphite, having a high thermal conductivity and a low CTE, exhibits very high thermal shock resistance. C/Cs, with a much higher tensile strength and thermal conductivity than polycrystalline graphite, exhibit a two to three times higher thermal shock-fracture toughness than polycrystalline graphite [78].

3.6. Frictional Properties of the Composites Carbon-based materials are widely used in wear-related applications, e.g. bearings, seals, and electrical brushes.

192

Handbook of Advanced Ceramics

The development of C/Cs has further widened the scope of application of carbon-based materials in wear-related applications from bearing seals and electrical brushes to brake pads for heavy duty vehicles such as military, supersonic, and civilian aircrafts to trucks and railways. The function of the brake is essentially controlled by its ability to absorb energy in the form of heat. Therefore, the thermal characteristics of candidate brake materials are of major importanceda potential high-performance brake material must have a high specific heat, a high melting point, and adequate mechanical properties at elevated temperature. Carbon fibers along their length are expected to have low wear and good frictional characteristics. Therefore, the early carbonecarbon disks were made with layers of carbon fabric or with random fibers arrayed parallel to the braking surface. C/Cs exhibit low coefficient of friction m 0.3e0.5 in the fiber direction and 0.5e0.8 in perpendicular direction. Wear rates also follow similar trend (0.05e0.1 and 0.1e0.3 mm). The friction and wear mechanism of C/Cs on application of brakes is quite complex, and various factors like peak temperature and formation of debris and films on sliding surfaces further affect the coefficient of friction, etc. which are too vast to cover here. This results in the formation of groundwear debris, containing flake-like turbostratic, lessordered carbon than the base composite. The friction surfaces are characterized by dull, unpolished wear surface. The wear debris produced under these conditions has a higher degree of order than the taxiing debris. The particulate debris does not lower the coefficient of friction, whereas film-type debris acts as a solid lubricant, lowers

the coefficient of friction, and reduces the wear rate [79e83]. Resin-derived CVD and pitch-derived carbon matrixes heated to same high temperature not only posses different graphitic contents and crystalline ordering, leading to variation in thermal conductivity, but also vary in structural characteristics, i.e. fiber/matrix bounding. Accordingly, during a braking operation these composites are heated up to different temperatures, which are highest for resin-derived low-conducting carbon and lower for CVD and pitch-derived highly ordered, high thermalconducting carbon matrixes. The average friction coefficient is to increase with decreasing initial velocity. The friction coefficient decreases with increasing pressure. According to Kimura et al. [84], the coefficient decreases with increasing Young’s modulus of the composites. This implies that the composites should be heat-treated at higher temperatures than usual. The frictional characteristics are also found to be influenced by the environment. In a wet environment, the rise in friction surface temperature is low and so is the coefficient of friction. Not only the carbon fiber and composite characteristics but also the inorganic additive, if added to the matrix of the composites, influences the friction characteristics of the ultimate composites. In commercial carbonecarbon brake pads, carbon fiber-fabric plies are used with fiber tows inserted in the third direction and pitch/CVD route for densification. Figure 26 shows comprehensive data on tribological properties of C/Cs [6]. However, the choice of the constituents and processing parameters is propriety of the manufacturers.

FIGURE 26 Coefficient of friction and wear of C/Cs.

Chapter | 2.10

CarboneCarbon Composites

4. OXIDATION PROTECTION OF C/CS Carbon, depending on graphitic contents, begins to oxidize in air at temperature around 400  C, the rate of oxidation increasing rapidly with temperature. The rate-controlling factors for oxidation of C/Cs at high temperatures (600e700  C) are the diffusion of the oxygen through the boundary layers surrounding the components and outward diffusion of reaction products (CO or CO2). For long-time application of these composites at elevated temperature under normal environment, it is essential that these composites, in the true sense, all components of the composites: fibers, matrix, and interface, be protected against oxidation. Therefore, studies on oxidation protection of C/Cs is as important as the development of the composites themselves. Therefore, oxidation protection is an important aspect of carbonecarbon technology. Studies on oxidation of C/Cs and their protection started concurrently with the development of these composites. Many types of coatings have been developed for protection of C/Cs. Various approaches that have been described in the open literature for protection of C/Cs are summarized in Table 7. Oxidation protection systems for C/Cs are based on (i) to block the carbon active sites and slow down the oxidation rate through modification of the matrix by addition of oxidation inhibitors (like P, B. Si, Zr or their compounds) and/or (ii) application of diffusion barrier coatings and overcoats to prevent oxygen ingress and carbon egress. However, most of the current systems of oxidation protection use several of these methods in a compatible or complimentary manner. Needless to say, the type of protection system required is dictated by the end application of the C/C. The rate of carbon gasification at low temperature can be effectively retarded by decreasing the porosity and blocking or poisoning the active sites by elements, which have an inhibiting effect on oxidation, such as boron, phosphorus, and silicon. Phosphorus compounds alone or in combination with halogens are found to have very strong

TABLE 7 Methods for Oxidation Protection of C/Cs Inhibitors and Sealants

Barrier Coatings and Overcoats

Halogens

Noble metals

Phosphorus compounds

Borides

Boron compounds

Carbides/nitrides

Polysiloxanes

Silicides

Silicon

Intermetallic compounds

Borate and silicate glasses

Engel/beewer compounds

193

inhibiting effects. The inhibition effect of halogens on the oxidation behavior of carbon appears to be confined to temperature below about 900  C. At higher temperatures, the halogens atoms are desorbed from the carbon surface, and their protective effect is thereby lost [84e86]. Of all the materials used as oxidation inhibitors for C/Cs, boron in one form or another, i.e. either elemental boron or boron compounds, have been found to be the most appropriate materials for oxidation protection at temperatures less than 1500  C. Phosphates in combination with other inhibitors such as silica, silicon carbide, and boron oxide have been found to reduce the oxidation of carbon at 1500  C to one tenth. Mixtures of metallic borides such as those of hafnium, zirconium, elemental boron, silicon carbide, and other glassforming particulates are also often added as inhibitors in the matrixes during the formation of C/Cs. On exposure to oxygen, these additives form mobile borates or borosilicate glasses and protect the C/Cs from oxidation. Silicon-based ceramic coatings, due to their suitable thermophysical properties, have been quite attractive for carbonecarbon oxidation protection at temperature exceeding 1000  C for extended periods. Various techniques have been developed to coat C/C with silicon-based materials [87]. However, excessive cracking, especially during cooling, may become a path for the ingress of oxygen. Therefore, multilayer approaches have been adopted wherein the glass layers between the outer layer and the substrates heal the cracks and protects the carbon fibers and the matrix. Numerous patent reports describe multilayer coatings consisting of Mo, Zr, Ta, Cr, Al2O3, silicon, boron, etc. as such or in the form of their inorganic compounds [88]. Figure 27 shows a general scheme of such a multilayer coating system. A permutation/combination of the elements and their compounds mentioned in Table 8 can be selected.

5. APPLICATIONS OF C/CS As is evident from the previous sections, C/Cs of the desired shape with properties required for particular application can be produced by meticulously choosing the type, architecture, and amount of carbon fiber and matrix precursor, and the processing conditions. Therefore, C/Cs are considered to be a class of materials with a wide spectrum of properties and applications. Moreover, the ongoing development of high-performance carbon fibers, high char-yielding synthetic resins, and pitches continuously add to the spectrum of carbonecarbon properties and products. It is with these milestones, C/Cs, which were developed in 1958 and nurtured under the US Air Force space plane program, Dyna-Soar, and NASA’s Apollo projects to meet the anticipated needs of the emerging space shuttle program, are regarded nowadays as a logical addition to the

194

Handbook of Advanced Ceramics

FIGURE 27 Generalized schematics of multilayer coatings for oxidation protection of C/Cs.

member of fine-grained carbon and graphites. C/Cs offer a large potential as high-performance engineering material. Therefore, in addition to the special defense, aircraft, and spacecraft applications, a steady interest is also growing in civil market segments. Based on the

TABLE 8 Most Commonly Used Elements and Their Compounds Major Function/ Protection Role

Element

Compound

Carbon

Pyrolytic graphite

Substrate adhesion/ protection from chemical attack

Phosphorus

Phosphate

Low-temperature oxidation protection

Boron

Alone

Substrate adhesion

Boron carbide

Substrate adhesion

Boron nitride

Oxidation-inhibiting matrix addition

B2O3

Sealants

Borates

Oxidation inhibition

Silicon carbide

Substrate adhesion

Silicon nitride

Oxidation protection

SiO2

Oxygen barrier

Silicides

Protective layer

Alone

Carbon-egression barrier

Silicon

Iridium

worldwide consumption, a general trend of carbone carbon consumption can be discerned as shown in Figure 28. In terms of mass consumption and money, the main applications of C/Cs today are still in military, space, and aircraft industry, while those in the field of general mechanical engineering are picking up steadily. In the following sections, general applications of C/Cs are arranged according to the various specific properties of the composites. There is great potential for the wider application of these materials.

5.1. C/Cs as Brake Discs Concerted research and development efforts in the design and manufacture of carbonecarbon brake systems led to

FIGURE 28

General market trends for C/Cs.

Chapter | 2.10

CarboneCarbon Composites

FIGURE 29

195

Carbonecarbon brake pads.

the widespread use of commercial C/Cs aircraft wheel brakes from military aircrafts [89] to civilian and commercial aircrafts with improved performance. The potential for growth in carbonecarbon brakes market is further underlined by their uses in racing cars, heavy-duty surface transport systems, and high- speed rail system [90]. For surface transport systems, the temperature required for brake operation is quite low (400e600  C), and therefore, relatively cheap designs and normal processing methods of 2D C/Cs can be utilized for such applications. Figure 29 shows a photograph of carbonecarbon disk for an automotive brake.

5.2. CarboneCarbon Materials for the Aerospace Industry The aerospace field continues to be one of the primary areas of application of C/Cs. The classic examples are the thermal protection system of space shuttles, rocket nozzles, exit cones, nose tips of reentry vehicle, etc. These components provide thermal protection to the instrumentation from the searing heat of reentry and maintain structural integrity over multiple missions, especially under condition of subzero temperatures in the outer space to close to 1700  C during entry back into the atmosphere. C/Cs for such applications are required to process high thermal conductivity and high specific heat so that the component operates as a heat sink, absorbing the heat flux without any problem. Advanced C/Cs with silicon carbide and ceramic multilayer coating for oxidation protection are utilized

in the European space plane, Hermes and National Aerospace Plane (NASP), with anticipated temperature of 2750  C on the nose cap and approximately 1950  C on the wing and tail-leading edges. In Rocket motors with a solid propellant system, the throat and exit cone are made of C/Cs. Although these materials are only for short time in operation, the property requirements are extremely stringent.

5.3. Aeroengine and Turbine Components The choice of C/Cs in jet engine rotors and stators offers the possibility of operating at temperature well in excess of 500e600  C higher than those used in conventional engines with high-temperature alloys. The basis of this development is the utilization of the high tensile strength of the fiber materials, which allows the compensation of the centrifugal stress on the blades.

5.4. C/Cs as Refractory Materials With carbon-fiber reinforcement, the mechanical properties of these carbon refractories can be further improved, thereby opening up new fields of industrial applications.

5.5. High-Temperature Fasteners The strength and stiffness at high temperature of C/Cs guarantee high self-fastening stability at high

196

Handbook of Advanced Ceramics

temperature. Therefore, carbonecarbon bolts, screws, nuts, and washers (Figure 30) are used where high temperature and severe chemical condition are present [91,92].

5.6. C/Cs in the Glass Industry

FIGURE 30 Carbonecarbon nuts and bolts.

Carbon and Graphite are not wetted by molten glass. Therefore, carbonecarbon materials are used as nonwetting crucibles for molten metals and in various parts of glass-container-forming machines [6]. They are used in channeling systems to carry the gobbets of molten glass, as molds for crystal glass products, and as an asbestos replacement for hot-end glass contact elements for moving hot glassware articles (Figures 31 and 32).

5.7. Emerging Applications

FIGURE 31

FIGURE 32 glasses (b).

Carbonecarbon heating elements.

The application and market for C/Cs are continually developing for different sectors. The latest applications of C/C structural materials are in advanced thermal management systems. These materials have potential for mechanical engineering applications in hot-load-bearing structure and corrosive surroundings, as substrates for electronics applications. The high electrical conductivity of airborne graphite particles creates an unhealthy environment for electrical equipment near machining areas. This aspect will be always kept in mind while processing, machining, and selecting application of C/Cs for general industrial applications.

Carbonecarbon channeling for carrying gobbets of molten glass (a,c) and molds for crystal and lead crystal drinking

Chapter | 2.10

CarboneCarbon Composites

REFERENCES [1] Inagaki M. New carbons: control of structures and functions. Elsevier; 2000. [2] Fitzer E. Carbon 1987;25:163e90. [3] Pierson HO. Handbook of carbon, graphite, diamond and fullerenes. Noyes Publication; 1993. [4] Meyyappan M. Carbon nanotubes: science and applications. CRC Press; 2004. [5] Dresselhouse MS, et al. In: Carbon nanotubes. Springer Pub; 2001. [6] Fitzer E, Manocha LM. Carbon reinforcements and carbonecarbon composites. Berlin: Springer-Verlag; 1998. [7] Delmonte J. Technology of carbon and graphite fiber composites. New York: Van Nostrand Reinholt; 1981. [8] Thomas CR. In: in Thomas CR, editor. Essentials of carbonecarbon composites. Royal Society of Chemistry; 1993. p. 1e36. [9] Burchell TD. Carbon materials for advanced technologies. Pergamon; 1999. [10] Buckley JD, Edie DD. Carbonecarbon materials and composites. Noyes Pub; 1993. [11] Savage G. Carbonecarbon composites. New York: Chapman & Hall; 1993. [12] Mazdiyasni KS, editor. Fiber reinforced ceramic composites. Park Ridge, NY: Noyes; 1990. [13] Fitzer E. Carbon fibersdpresent state and future applications in carbon fibers filaments and composites. In: Figueiredo JL, et al., editors. Kluwer Academic Publishers; 1989. p. 3e41. [14] Manocha LM. Carbon fibers in encyclopaedia of materials: science and technology. In: Jurgen Buschow KJ, et al., editors. Elsevier; 2001. p. 906e16. [15] Edie DD. The effect of processing on the structure and properties of carbon fibers,. Carbon 1998;36:345e62. [16] Paris O. Carbon 2002;40:551. [17] Chung DDL. Carbon 2001;39:1119. [18] Buckley JD. Ceramic Bulletin 1988;67:364. [19] Heym M, Fitzer E. Carbon fiber reinforced carbons in processing and uses of carbon fiber reinforced plastics. VDI Dusseldorf; 1981. 85. [20] Ko F. Ceramic Bulletin 1909;68:401. [21] Lachman WL, Crawford JA, McAllister LE. Multidirectionally reinforced C/C composites. In: Norton B, et al., editors. Prof. international conference on composite materials. NY: Metallurgy Society of AIME; 1978. [22] McAllister LE, Lachman WL. Multidirectionally carbonecarbon composites. In: Kelly A, Milieko ST, editors. Handbook of composites, vol. 4. Elsevier; 1983. [23] Lackey WJ. Carbon carbon composites in encyclopaedia of materials: science and technology. In: Jurgen Buschow KJ, et al., editors. Elsevier; 2001. p. 906e16. [24] Delhaes P. Carbon 2002;40:641e57. [25] Rand B. Matrix precursor for carbonecarbon composites. In: Thomas CR, editor. Essentials of carbonecarbon composites. London: Royal Society of Chemistry; 1993. [26] Manocha LM. Sadhna [27] Kochendorfer R. In: Singh M, Jessen T, editors. Ceramic matrix composites in ceramic engineering and science proceedings, vol. 22. American Ceramic Society Publication; 2001. p. 11e22. No.3. [28] Liebermann ML, Pierson HO. Carbon 1974;12:233. [29] Pierson HO, Liebermann ML. Carbon 1975;13:159. [30] Bauer DW, Kotlensky WV. SAMPE Quarterly 1972;4:24.

197

[31] [32] [33] [34]

[35]

[36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47]

[48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64]

[65] [66] [67] [68] [69] [70] [71]

Bokros J. In: Walker PL, editor. Chemsitry and physics of carbons, vol. 5; 1969. p. 1e118. Huttinger KJ. Adv Mater CVD; vol. 4; 1998. p. 151e158. Kimura S, Yasuda E, Takase N, Kayusa S. High temperature high pressure, vol. 13; 1980. 193. Lackeye WJ, Starr TI. Fabrication of fiber reinforced ceramic composites. In: Mazdiyasni KS, editor. Fiber reinforced ceramic composites. Park Ridge, NY: Noyes; 1990. p. 397. Lackey WJ, Vaidyaramans. Fabrication of C/C composites by forced flow thermal gradient chemical vapour infiltration. US Pat. 5916633 1999. Vaidyaraman S, et al. Carbon 1996;34:609e1123. Gebhard JJ, Stover ER, Mueller W. In: Petroleum derived carbons. ACS symposium series, vol. 21; 1976. p. 212. Oh SM, Lee JY. Carbon 1988;26:763. Dupel P, Bourrat X, Pailler R, Naslain R. Carbon 1995;33:1193. Jeong HJ, Park HD, Lee JD, Park JO. Carbon 1996;34:417. Yudin VE, et al. Carbon 2002;40:1427. Manocha LM. Carbon 1994;32:213. Pierson HO. J Compos Mater 1975;9:118. Fitzer E, Huettener W, Manocha LM. Carbon 1980;18:291. Bahl OP, Manocha LM, et al. J Sci Ind Res 1991;50:533. Modification of resins with graphite powder (Manocha Kimura). Yasuda E, et al. Application of iodine treatment on fabrication of pitch based C/C composites. In: Palmer KR, Wright MA, editors. Applications of carbon and carbonaceous composites. World Scientific, 1996. Klett JW, Edie DD. Carbon 1995;33:1485. Manocha LM. Mater Sci Eng A 2005;412:27e30. Endo M, et al. Appl Phys 2008;111:13. Jortener J. MTC annual conference. Carbondale: Southern Illinois Uni; 1990. p. 51. Manocha LM, Yasuda E, Kimura S, Tanabe Y. Carbon 1988;26:333. Manocha LM. Composites 1988;19:311. Manocha LM, Bahl OP, Singh YK. Carbon 1989;27:381. Jortener J. Carbon 1990;30:153. Jortener J. Carbon 1986;24:603. Murdie N, et al. Seventh annual MTC conference. Carbondale: Southern Illinois Uni; 1991. p. 81. Zaldivar RJ, Rellick GS. Carbon 1991;29:1155. White JL, Shaeffer PM. Carbon 1989;27:697. Zimmer JE, White JL. Disclinations and fracture, fourth annual conference. Carbondale: Southern Illinois Uni; 1987. p. 143. Cook J, Gorden JE. Proc Roy Soc Lond 1964:508. A 2. Oh SM, Lee JY. Carbon 1988;26:760. Granoff B, Pierson HO, Schuster DM. Carbon 1973;11:177. Huettener W. Potential of carbonecarbon composites for structural applications. In: Figueireds JL, editor. Carbon fibers, filaments and composites. Kluwar Pub; 1990. p. 275. Crocker P, McEnnaney B. Carbon 1991;29:881. Sato A, Karumada A, Iwaki H, Komatzu Y. Carbon 1989;27:791. Zao JX,, Bradt RC, WalkerJr PL. Carbon 1985;23:9. Ahearn C, Rand B. Carbon 1996;34:239. Labruquere S, et al. Carbon 2001;39:97. Hamada K, Sato S, Kohama A. J Nuc mat 1994;212:1228. Kimura S, Yasuda E, Tanabe Y. Microstructure and related properties in C/C composites. Proc Int Symp On Ceramic Composites for Engines. Japan; 1983. p. 783.

198

[72] Tanamura T, et al. Tanso 1991;149:220. [73] Zimmer J. Extended abstracts 20th biennial carbon conf. 1991, p. 390. [74] Burchell TD, Oku T. Materials property data for fusion reactor plasma facing C/Cs, preprints (personal communication) [75] Manocha LM, et al. Carbon Science 2005;6(1):6e14. [76] Manocha M L, et al. Carbon 2006;44(3):480e95. [77] Chollon G, et al. Carbon 2001;39:2065. [78] Sato S, Kurumada A, Kawamata K, Ishide R. Int Carbon Sympo Tsukuba; 1990. p. 214. [79] Murdie N, Ju CP, Don J, Fortuanto FA. Carbon 1991;29:335. [80] Ju CP, Lee KJ, Wu HD. Carbon 1994;32:971. [81] Shin HK, Lee HB, Kim KS. Carbon 2001;39:959. [82] Lei B, et al. Carbon 2011;49:4554.

Handbook of Advanced Ceramics

[83] Reitsch C J, et al. Carbon 2009;47:85. [84] Kimura S, Yasuda E, Narita N. Friction and wear of C/C composites. J Japanese Society of Lubricating Engineers 1983;28:185. [85] McKee DW, Spiro CI, Lamby EJ. Carbon 1984;22:285e507. [86] McKee DW. Oxidation protection of carbon materials in chemistry and physics of carbons, vol. 23; 1991. p. 173. [87] Manocha S, Manocha LM. Carbon 1995;33:435. [88] Li H- J. Carbon 2010;48:1636. [89] Stimson IL, Fischer R. Trans Roy Soc Lond 1980;A294:583. [90] Klein AJ. Mater process 1986;130:64. [91] Carbon fiber reinforced composites Product Brochure, Schunk Kohlenstoff, Germany. [92] Carbonecarbon composites Product Brochure, Sigri Kohlenstoff, Germany.

Chapter 2.11

Carbon Materials used for Polymer Electrolyte Fuel Cells Hiroshi Shioyama National Institute of Advanced Industrial Science and Technology, 1-8-31 Midorigaoka, Ikeda, Osaka 563-8577, Japan

Chapter Outline 1. 2. 3. 4.

Introduction Operating Principles and Characteristics of PEFCs Separators Current Collectors

199 199 200 201

1. INTRODUCTION Fuel-cell power technology is considered to be one of the most promising new power-generation technologies owing to its various merits, such as high power efficiency and excellent compatibility with the environment [1]. Fuel cells can be broadly classified into phosphoric acid fuel cells (PAFCs), molten carbonate fuel cells, polymer electrolyte fuel cells (PEFCs), and solid oxide fuel cells. Car manufacturers and electronics companies are competing to develop PEFCs because of their advantages such as small size and weight, high power density for normal temperature operation, and excellent start-up capability. Carbon is an essential component material in PEFCs. This chapter, while focusing on the separator, electrocatalyst, and proton conductor, reviews how the research on carbon materials is involved in improving the performance of PEFCs.

2. OPERATING PRINCIPLES AND CHARACTERISTICS OF PEFCS A PEFC comprising a proton-conductive polymer electrolyte membrane of thickness 50e180 mm gives an all solid-state structure consisting of gas diffusion electrodes made from carbon supporting platinum on both sides of the membrane (Figure 1). Hydrogen supplied to the anode side is oxidized, and produced Hþ moves through the membrane to the cathode side. This Hþ is used in the

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00012-5 Copyright Ó 2013 Elsevier Inc. All rights reserved.

5. Electrocatalyst 6. Proton Conductor 7. Conclusion References

201 205 208 208

reaction in which water is formed by the reduction of oxygen supplied to the cathode side. As the electrons generated at the anode flow through the external circuit to arrive at the cathode, an electric current can be retrieved. The reactions at the anode and cathode, and the overall reaction, are given as follows. Anode: H2 / 2Hþ þ 2e Cathode: 1/2O2 þ 2Hþ þ 2e / H2O Overall reaction: H2 þ 1/2O2 / H2O Pure hydrogen or hydrogen reformed from methanol or natural gas is used as fuel. In addition, the direct methanol fuel cell (DMFC) is considered where methanol is supplied directly to the anode instead of hydrogen to produce power. The most typical polymer electrolyte membrane is the perfluorosulfonic-acid-type ion exchange membrane such as NafionÒ , which has a molecular structure based on a main polytetrafluoroethylene chain with side chains containing sulfonic acid. PEFCs operate at temperatures below ca. 90  C. The high power density of PEFCs is largely attributable to the high ionic conductivity of the polymer electrolyte membrane at these temperatures. However, because the polymer electrolyte membrane is a strong acid, the materials that can be used in the vicinity of the electrolyte are limited. A polymer electrolyte membrane containing water has acidity comparable to a solution with 10 wt% sulfuric acid. The anode side is a hydrogen ambient atmosphere, while the cathode side is

199

200

Handbook of Advanced Ceramics

FIGURE 1 Schematic diagram of PEFCs.

air. Because of humidification and water produced by the reaction, the reactive gas contains saturated water vapor. In addition, because the incorporation of impurity ions in the polymer electrolyte membrane lowers the performance, the use of acid-soluble materials should be avoided. With these strong constraints, fine particles of precious metals such as platinum are used for the electrocatalyst, whereas for the catalyst supports, current collectors, and separators, carbon materials with high electrical conductivity and high chemical stability are used.

3. SEPARATORS The PEFC stack mainly consists of the membrane-electrode assembly (MEA) and separators, as shown in Figure 2. There are grooves on the surface of the separators that allow for the supply of the reactive gases of hydrogen and oxygen. This member is called a separator because of its role as a partition; thus, the reactive gases supplied to neighboring single cells in the stack do not mix. It is sometimes called an interconnector because of its function of electrically connecting neighboring cells in series. It is also called a bipolar plate or bulkhead plate. (1) Separator Classification Separators can broadly be classified into metallic and carbonaceous. When considering cost, weight, corrosion resistance, and machinability, the metallic separator is inferior to the carbonaceous separator.

Given the carbonaceous separator, glasslike carbon was the material of the first commercialized separators. It was used as a separator for the PAFC. Although its gas impermeability was highly sufficient, there were difficulties in the electrical conductivity and machinability. They are expensive and not currently used. Consequently, separators made of resin-impregnated isotropic graphite were used. They are still used as single cells in research. Molded carbon resin products are now the mainstream of the separator material in the PEFC stack. This material can be divided by classification of the forming method (compression molding and injection molding) or a variety of resins (thermosetting resin and thermostable resin). In any case, the separator costs approximately over 90% of entire PEFC costs, and it is thus desirable to develop highperformance separators that can be mass produced at low cost. (2) Molded Carbon Resin Separators For the serpentine flow field of the reactive gas engraved on the separator, both the groove width and depth are about 1 mm. Furthermore, the gas inlet and outlet construction and the flow channel for the cooling water must be set up. After ensuring the machining accuracy of these fine structures, it must withstand repeated expansion and contraction due to the difference between the operating temperature of ca.90  C and room temperature, the swelling shrinking stress of the electrolyte membrane attributable to operation stoppages and a certain amount of expected vibration when mounted on a vehicle.

Chapter | 2.11

Carbon Materials used for Polymer Electrolyte Fuel Cells

201

FIGURE 2 Construction of a PEFC stack.

Based on the study of the manufacture of the molded carbon resin separator that overcomes such conditions, the following methods are currently recommended: the ‘compression molding using thermosetting resin’ and the ‘injection molding using thermoplastic resin’. The former has the disadvantage of time requirements as high as 10 min to produce a single separator. On the other hand, the latter has disadvantages such as high electrical resistance and inadequate gas permeability, although it is possible to produce a single separator in about 1 min. Incidentally, the details of the resin used in molded carbon resin separators are not disclosed by the manufacturer. (3) Carbon Coating for Metal Separators Metallic separators are effective for slimline modeling and are excellent in terms of conductivity and reliability. Accordingly, carbon coating has been investigated in order to increase the corrosion resistance under the operating conditions of a fuel cell. As methods of carbon coating, the following have been tested: press molding onto a metallic substrate of materials similar to those used in the abovementioned molded carbon resin separator [2], carbonizing an applied polymer by spin coating, amorphous carbon coating by plasma sputtering [3], and so on. In all cases, the coating is carried out after removal of the oxide film on the metal. The evaluation findings for the corrosion resistance and conductivity of the finished product are favorable, although it is still in the prototype stage.

4. CURRENT COLLECTORS To make electrical contact between the above-mentioned separator and the electrocatalyst described later, carbon fiber has been used as the current collector. Carbon fibers are broadly categorized as carbon paper and carbon cloth, and are selected depending on the user’s choice. For example, from the E-TEK Division of PEMEAS USA Inc. Fuel Cell Technologies, which sells fuel cell parts, various products of carbon paper or cloth treated with the desired degree of water repellent can be purchased. Academic interest related to carbon materials is scarce with respect to the dependence of current collectors on the power-generation characteristics; therefore, it will not be discussed further.

5. ELECTROCATALYST (1) The Role of the Carbon Support The electrode reaction takes place at the so-called threephase interface (platinumeelectrolyteereactive gas). In the case where the electrolyte is a solid-state membrane, the reaction location is limited to the contact interface between the platinum electrocatalyst and the electrolyte. Therefore, the performance of PEFCs is affected by not only the amount and surface area of the platinum used as catalyst but also the availability of the platinum actually taking part in the reaction. For example, in order to increase the availability of the expensive platinum, there

202

Handbook of Advanced Ceramics

is a need to decrease platinum particles that cannot contribute to the reaction because of no contact with the electrolyte or reactive gas. By choosing a carbon support, the supported platinum amount, surface area, and availability can be controlled. Therefore, it is no exaggeration to say that the PEFC performance is determined by the carbon support. If a superior carbon support can be developed with characteristics that can activate the electrocatalyst, it is possible to reduce the amount of platinum that should be used. PEFCs can become cheaper when the amount of platinum used is small, and the problem of depletion of platinum resources can be resolved if its use becomes widespread. Because of the high specific surface area of carbon black, the platinum particles can be supported on carbon black by using various methods, such as the impregnation method, preparation using colloidal particles, or preparation by ion exchange. With carbon black as a support, the microparticulation of platinum is made possible. Accordingly, the activity per unit mass of platinum can be increased. In addition, given the diffusion of the reactive gas within the electrode or the distance of ionic conduction, it is better for the electrocatalyst layer to be thinner. In such a case, the supported density of platinum must be increased with maintaining the highly dispersed state. For example, the loaded amount and particle diameter of platinum supported on carbon black, VulcanÒ XC-72, as well as the platinum surface area are summarized in Table 1 [4]. Here, XC-72 is the carbon black which is widely used in the research of electrocatalysts. There have been many reports on carbon supports with better performance compared to the reference standard XC-72. Because the specific surface area of ca. 250 m2/g is not so high, Ketjen Black with an increased specific surface area is sometimes used. In Figure 3 [5], it is shown that the mass activity (MA)

TABLE 1 Platinum Catalyst Loaded on the Carbon Black XC-72 [4] Loaded amount of platinum

Mean diameter of platinum (nm)

Specific surface area of platinum (m2/g)

10 wt%

2.0

140

20 wt%

2.5

112

30 wt%

3.2

88

40 wt%

3.9

72

60 wt%

8.8

32

80 wt%

25.0

11

Platinum black

10.0

28

FIGURE 3 MA of 0.9 V of Pt catalysts versus surface area of carbon support [5].

normalized by weight of platinum is dependent on the specific surface area of the support, and this shows the effect of increased surface area of platinum as the particle diameter became smaller. The basic idea behind the carbon support is roughly the same, whether it is the anode or the cathode. However, since the chemical species of the supported fine particles of precious metal are sometimes different, and, naturally, the electrode reactions, oxidation, and reduction are completely different, these are explained separately below. (2) Cathode With respect to the fine particles of platinum, which act as the catalyst, the increase in the MA of platinum with the decrease in the particle diameter was referred to above. In contrast, there are many reported cases that the activity per surface area of platinum (specific activity) decreases as the particle size decreases. Consequently, there would be an optimum value for the platinum particle diameter considering the activity for both per unit weight and per unit surface area of platinum, but to clarify the effect of particle size would require the collection of further data. In addition, the particle diameters 3e10 nm are sufficiently accessible when the platinum is supported 30e70% using carbon black as a support. Again, catalytic activity is determined by the particle diameter and surface area of the platinum catalyst, and the platinum particle diameter and surface area are highly dependent on the carbon black support. Electrocatalytic activity is also affected by the reactive site and mass transfer inside the porous gas diffusion electrode that is determined by the form of the carbon black. First, as shown in Figure 4, for the case where dibutylphthalate (DBP) absorption is small, the cell voltage decreases [5]. The pores inside the electrode have the function of diffusing oxygen and expelling water, but they are absent when DBP absorption is small and the performance has deteriorated. However, from performance assessments of cells using various kinds of carbon black

Chapter | 2.11

Carbon Materials used for Polymer Electrolyte Fuel Cells

FIGURE 4 Cell voltage at 0.5 A/cm2 of Pt catalysts versus DBP adsorption of carbon support [5].

(Figure 5), it is shown that primary-pore-rich carbon black is not suitable as a support [6], for polymer electrolyte cannot intrude into the pores. In this manner, the effect of the structure of carbon black on the electrocatalytic activity cannot be explained easily, and further studies are awaited. (3) Anode There is no problem when pure hydrogen is used as fuel, but when hydrogen is obtained by reforming methanol or natural gas, the residual carbon monoxide (CO) is strongly adsorbed onto the platinum catalyst surface, and this leads to a large drop in catalytic activity. To prevent this poisoning, the method of adding metals of group VIII, such as ruthenium, has been widely used. The particle diameter and specific surface area of the platinum catalyst, which contains ruthenium, are similarly dependent on specific surface area of the support and show

FIGURE 5 Polarization curves showing the dependence of carbon black support. Loaded platinum: 0.5 mg/cm2. Electrolyte: Flemion [6].

203

a trend similar to platinum alone. It is therefore advantageous for the specific surface area of the support to be large for the electrode oxidation activity of pure hydrogen. In addition, results of a previous study [5] show that it is advantageous for CO poisoning tolerance to increase the specific surface area of the carbon black support, which allows the platinumeruthenium specific surface area to also increase, while there are also opinions that if the specific surface area of the support is high, the platinume ruthenium alloying cannot be done well, and the poisoning tolerance is not sufficient. It is therefore necessary to comprehensively weigh up these properties to determine the desired carbon black. (4) Durability In PAFCs, aggregation of the platinum catalyst particles occurs because of long periods of power generation; hence, a drop in performance is said to occur. This happens probably because of the comparatively high operating temperature of ca. 200  C. Although PEFCs are operated at most below 100  C, similar catalyst particle growth has occurred [5], and catalyst particle growth is detected particularly at the cathode side. Furthermore, at the anode side, the degradation of CO poisoning tolerance has also been observed. However, the details of the relationship between durability and the carbon black support have not yet been clarified, and, at present, research is proceeding [7]. (5) Development of a New Carbon Support As mentioned above, carbon black is used as a support for the electrocatalyst used in PEFCs, but research is also underway using other carbon materials aimed at trying to make performance improvements. Carbon nano-horns discovered by Iijima et al. [8] have a conical structure with a single layer of graphene with a diameter of 2e4 nm and a length ca. 50 nm. This carbon nano-horn is obtained in the form of numerous spherical aggregates, with the platinum particle becoming as tiny as about 2 nm when the platinum is loaded on the aggregates. It has also been found that a uniform dispersion on the carbon nano-horn aggregate is possible; therefore, a superior electrocatalyst is being developed for DMFC use. On the other hand, many studies of new electrocatalyst supports have reported using carbon nanofibers manufactured by the arc discharge, chemical vapour deposition (CVD), and template methods [9e22]. All of these reports conclude that when carbon nanofiber is used as a support, a superior electrocatalyst is obtained. For example, Joo et al. [13] showed that the electrocatalyst with nano-carbon prepared by the template method provides a good oxygen reduction MA (Figure 6). One reason for this is that the supported platinum particles are small and have uniform particle diameter in the same way as the above-mentioned carbon nano-horns. Furthermore, other studies demonstrate

204

Handbook of Advanced Ceramics

FIGURE 6 Electrocatalytic MA of Pt/carbon catalysts for the O2 reduction, measured at a potential of þ0.900 V vs. NHE and at a rotating speed of 10,000 rpm in 0.1 M H2ClO4 saturated with O2 [13].

FIGURE 7 A Tafel plot of IK for O2 reduction at catalyst layers formed from Pt/activated carbon. Concentrations of CF3SO3K solutions used for its adsorption are indicated inside the figure [23].

other reasons why carbon nanofiber is a superior electrocatalyst support: (1) the high electrical conductivity of carbon nanofiber [9,14]; (2) the presence of only few impurities, such as sulfur, compared to carbon black [9]; (3) the fact that the morphology of supported platinum is different for carbon black [9,15]; and (4) some kind of interaction between platinum and nano-carbon [14,15]. Detailed explanations could be considered during a future task. Despite having a high specific surface area similar to carbon black, active carbon is not used as a catalyst support for PEFC. This is due to the low electrical conductivity of carbon. Furthermore, even if the platinum is supported in the interior of the micropores and sub-micropores of the active carbon, ions cannot reach there via the polymer electrolyte. Maruyama et al. have reported that if trifluoromethane sulfonic acid is adsorbed onto the micropores, it facilitates ion transfer and hence catalytic activity increases (Figure 7) [23]. Many papers have elucidated the carbon support dependence on the catalytic activity. When the catalytic activity was compared for the vapor-phase oxidation of hydrogen, methanol, and formic acid, marked dependence on the carbon support was noted [24]. In the case of electrochemical oxidation of methanol with platinumeruthenium particles supported on mesocarbon microbeads, a higher activity than an electrocatalyst with carbon black support was reported [25,26]. There are also reports that hard carbon spheres [27], graphite intercalation compounds [28], and graphene sheets obtained by reducing oxidized graphite [29] are excellent supports. There are only few cases where the causes of these support effects have been investigated. However, in reports relating to the electron spin resonance (ESR) measurement [30] and the analysis of the electrical double layer [31], the degree of electron transfer between platinum and carbon is different, depending on the type of carbon. Consequently, the electron density on platinum is also different.

Additionally, through the pretreatment of carbon used as a support, attempts are being made to improve the electrocatalytic activity. Activity improvements for methanol oxidization were reported when acid functional groups were added to the surface of carbon fiber as catalyst support [32]. Activity improvement of platinum has also been noted by adding fluorine functional groups [33]. (6) Carbon with Its Own Electrocatalytic Activity An important research field for PEFCs is the development of alternative materials to platinum. Instead of enhancing the electrocatalytic activity by choosing the kind of carbon supports and reducing the amount of platinum which have been stated so far, this research aims to develop other active substances without using platinum. In the past, the use of every kind of metal complex and metal oxide has been attempted. Regarding carbon materials, heat treatment products of cobalt macrocyclic complex were found to have a catalytic activity in the oxygen reduction reaction [34], and research is underway with carbonized products of similar complexes. Of the central metal, the use of group VIII transition metals, such as iron, cobalt, or nickel, is very common. For the cobalt complex, carbonized products of phthalocyanine, porphyrin, or other types of complex have been reported. For the onset of oxygen reduction, both cobalt metal and nitrogen are important factors. For the iron complex ligands, there are many reported cases of not just the phthalocyanine and porphyrin macrocyclic ligand but also ligands containing the phenanthroline [35] or cyano group [36]. Interestingly, the ligands in these metal complexes contain nitrogen atoms as well. Furthermore, with regard to the heat treatment temperature during carbonization, it is experimentally known that catalytic activity of treated samples is high in the region from 600e900  C. It has been anticipated that, due to the heat treatment, some kind of arrangement of iron and nitrogen atoms will form an active site. If the heat

Chapter | 2.11

Carbon Materials used for Polymer Electrolyte Fuel Cells

205

FIGURE 9 Plots of AC impedance spectroscopy of fullerenol pellet [41]. FIGURE 8 Polarization and power density curves at 80  C for a nanoshell carbon as the cathode catalyst. Anode: Pt (0.5 mg-Pt/cm2), H2 (200 sccm); cathode: (3.8 mg-carbon/cm2), O2 (200 sccm); membrane: Nafion 112; cell temperature: 80  C; back-pressure: 0.35 MPa [40].

treatment temperature is too high, aggregation of the iron takes place and the activity will probably drop. Furthermore, Maruyama et al. showed that, in light of the fact that the carbonized products of hemoglobin and catalase possess high catalytic activity in oxygen reduction, the carbonized products of red blood cells are also promising as an electrocatalyst in oxygen reduction [37]. These can be classified in the category of complexes that contain iron and nitrogen atoms. It is also interesting from the standpoint of waste utilization during meat manufacture. In recent years, it has been reported that some carbon materials with nitrogen atoms possess high oxygen reduction activity even in the absence of metal atoms. When nitrogen is introduced to the surface of carbon black with the ammonia treatment at ca. 600  C [38], oxygen reduction activity appears in the carbon itself. Further improvement in oxygen reduction activity has been reported by doping with boron and nitrogen [39]. Ozaki et al. assumed that the onset of oxygen reduction activity is dependent on the shell structure of the carbon surface. Unusual shell structures are produced when macrocyclic complexes of iron, cobalt, and nickel are carbonized, and this structure is assumed to be effective in oxygen reduction activity. This activity seems to be significant if nitrogen and boron atoms exist at an edge site. The results of the power-generation performance assessment of an MEA made without using platinum are shown in Figure 8 [40]. The performance is low compared to the case where platinum is used, but it is understood that reasonable values have been obtained.

directed at their application for PEFCs has been started. Proton conduction of fullerene derivatives was reported for the first time in 2001 [41]. C60(OH)n (n ¼ 10e12) is pressed at 500 MPa to manufacture a pellet of 15-mm diameter, giving the impedance characteristic of Figure 9. From the facts that the sample electrical resistance value was 1014 Ucm and the time constant in the high-frequency region was approximately 10 5 s, proton conduction can be concluded. In the absence of moisture, the proton conductivity of C60(OH)n is 7  10 6 S/cm (295 K), which shows that the estimated proton dissociation constant pKa is close to the value for phenol. On the other hand, C60(OH)24 hardly shows any proton conduction due to its weak electron affinity. In order to ensure that protons pass through as a form of Hþ and that H3Oþ is not involved, the time dependence of the proton conductivity was observed using samples in dry conditions. From the fact that the proton conductivity was constant, as shown in Figure 10, Hþ is proved to be responsible for proton conduction. From the temperature dependence of proton conduction shown in Figure 11, activation energy of 0.33 eV was obtained. This value is an order of magnitude higher than the proton conductivity for Nafion, which predicts that the basis of proton conduction is hopping conduction. Powder X-ray diffraction shows that the crystal structure of C60(OH)n (n ¼ 10e12) is similar to that of C60. In other words, the OH group is considered to exist in the gap

6. PROTON CONDUCTOR (1) Fullerene-based Proton Conductor In recent years, it has been understood that the fullerene derivatives behave as proton conductors, and research

FIGURE 10 Time dependence of conductivity of fullerenol during DC measurement in a dry hydrogen gas [41].

206

Handbook of Advanced Ceramics

the conditions of relative humidity (RH) at 95% and at 25  C, was 0.01 S/cm [49]. As shown in Figure 12, the proton conductivity is strongly dependent on the RH. Under dry conditions, C60(OSO3H)x(OH)y (x ~ 6, y ~ 6) and C60(OSO2O)6 z(OSO3H)z(OH)z (Z < 1) show high proton conductivity, and especially the former is reported to have a proton conductivity of 0.01 S/cm [42]. (2) FullereneeNafion Composites

FIGURE 11 Arrhenius plots of ionic conductivity of fullerenol pellets [41].

FIGURE 12 Plots of proton conductivity vs. RH (n) and the water content n vs. RH (:) [49].

where the neighboring C60 framework is not connected. With such a crystal structure, it is not easy to explain the hopping conduction, and we are eager to anticipate the findings of further research. As we know that there is proton conductivity in fullerene hydroxide, the hunt for fullerene derivatives possessing even higher conductivity has started. For example, Li et al. showed that the proton conductivity of C60 > C[PO(OH)2]2 (in the formula, > means a C crosslinking bond between adjacent carbon atoms in C60), under

One trend in attempting to use fullerene-based proton conductors as proton conduction membrane is fullerene composites with Nafion. Tasaki et al. formed the composite by immersing Nafion 117 in C60 toluene solution or C60(OH)12 tetrahydrofuran (THF) solution, and reported its characteristics [43]. As shown in Table 2, the water content of the composite is greater than simple Nafion, and the water retention characteristics were found to be excellent (Figure 13). Because of this retained moisture, the proton conductivity was improved in conditions where the RH is low, as shown in Figure 14. The interaction between fullerene derivatives and Nafion is not yet fully understood, but by computer simulation it has been possible to confirm the affinity between the sulfonic group in Nafion and the OH group in PHF: C60(OH)12 [43]. The simulation concluded that the fullerene derivative is present in the proton channel of Nafion. The existence of fullerene derivatives with proton conductivity even under dry conditions in the proton channel of Nafion gives a high proton conductivity of the composite in conditions of low RH. While the affinity between fullerene derivatives and Nafion was discussed, the aggregate of the fullerene derivative was observed with an optical microscope in the fullerene derivativeeNafion composite [43,44]. To avoid the aggregation, a special fullerene derivative C60[CH2C6H4(OCH2CH2O)3OCH3]n, which gives an extremely improved dispersion characteristic, was used [44]. A Nafion composite with either HC60(CN)3 or C60(TEO)5, which has the cyano or ethylene oxide group, gives improved dispersion characteristic and proton conductivity [45]. Some of the results are shown in Table 3. As mentioned above, it was found that the proton conductivity for the fullerene derivativeeNafion composite was high even under conditions of low RH, but its application for the fuel cell has not been reported.

TABLE 2 Wet and Dry Water Uptake for FullereneeNafion Composite Membrane (%) [43] Water uptake

Nafion 117

C60/Nafion117

PHF/Nafion 117

Wet

25.5

27.4

27.1

Dry

4.3

5.5

4.7

Chapter | 2.11

Carbon Materials used for Polymer Electrolyte Fuel Cells

FIGURE 13 The weight loss of the fullerene composites and Nafion 117 at room temperature under dry N2 gas flow as a function of time [43].

FIGURE 14 Proton conductivities of the fullereneeNafion composite membranes: C60/Nafion 117, PHF/Nafion 117, and Nafion 117 measured at 20  C as a function of RH. PHF: C60(OH)12 [43].

When preparing the MEA, Shioyama et al. added C60(OH)10 or C60F36 to the Nafion ionomer in order to increase the performance at the electrode reactive site on the interface between the electrocatalyst and the Nafion 117 membrane [46]. As shown in Figure 15 and Table 4, an improvement in power-generation characteristics was acknowledged in all cases. Here, the effect of C60(OH)10 was meager, for the sites where C60(OH)10 exist are restricted to the vicinity of the electrocatalyst, and was not able to affect the proton conductivity of the whole membrane of Nafion 117.

207

FIGURE 15 Comparison of polarization curves. Pt loading and additive for the cathode: 0.6 mg(Pt)/cm2 and additive free (A); 0.05 mg(Pt)/cm2 and additive-free (,); 0.05 mg(Pt)/cm2 and 10% C60(OH)10 (n); and 0.06 mg(Pt)/cm2 and 5% C60F36 (6) [46].

On the other hand, there is a marked improvement in power-generation characteristics with the addition of C60F36. As proton conduction could not be expected with the addition of C60F36, the improvement in power-generation characteristics leads to the conclusion that there is a structural change in the proton channel. It is known that Nafion comprises a perfluoro backbone and the side chains a sulfonic group. As a result, the rod-like structures of the sulfonic group pointing outwards are taken off when Nafion is dispersed with a polar solvent such as water or alcohol. When C60F36 is added to this Nafion ionomer solution, the C60F36 molecule is found to exist not in ethanol or in the vicinity of the sulfonic group but rather in the Nafion aggregate, which has been revealed by analysis using photochemical methods [47]. These results can arise from the fact that C60F36 is not able to interact with the [Ru(bpy)3]2þ which exists in the vicinity of the sulfonic group as shown in Figure 16. In other words, in the case where the MEA is made by adding C60F36 to Nafion ethanol solution, an interaction between Nafion perfluoro backbone and C60F36 is thought to have affected the Nafion proton channel structure in the vicinity of the electrocatalyst and to have improved the power-generation characteristics. The details of the change in the proton channel structure are not understood, and it merits further research.

TABLE 3 Proton Conductivity of HC60(CN)3eNafion Composites (mS/cm) [45] RH(%)

Recast Nafion

1%HC60(CN)3Nafion

1%HC60(CN)3-0.5% C60(TEO)5-Nafion

0.5%C60(TEO)5Nafion

25

1.6

3.4

4.6

1.5

80

31.4

32.8

34.9

31.8

208

Handbook of Advanced Ceramics

TABLE 4 Comparison of the MEA Performance Prepared with and without Additive to Nafion ionomer [46] Additive to Nafion ionomer

Pt loading mg(Pt)/cm2

Observed current at 0.75 V mA/cm2

Additive-free

0.63

325

Additive-free

0.051

55

Additive-free

0.062

52

5% C60(OH)10

0.050

75

10% C60(OH)10

0.054

97

20% C60(OH)10

0.058

73

5% C60F36

0.061

140

10% C60F36

0.064

75 FIGURE 18 Amount of methanol permeation through the fullerene membrane and a perfluorinated sulfonic acid membrane. The methanol permeation is adjusted to the membrane thickness [48].

(3) Fullerene-based MEA without Nafion

FIGURE 16 (I0／I)e[C60F36] relation for the luminescence quenching of [Ru(bpy)3]2þeC60F36 systems [47].

In the preceding section, proton conduction of fullerene derivativeeNafion composites was reviewed, whereas fullerene-based MEA without Nafion was attempted in fuel cell generation. Hirakimoto et al. have synthesized a proton conductor where seven eC2F4eOeC2F4eSO3H groups are appended to each C60, and the neighboring fullerene derivatives are bonded to each other by e((CF2)8)ne. The MEA with this fullerene derivative gives good performance as shown in Figure 17 [48]. This performance is the same as when using Nafion as a proton conductor. With the low amount of methanol crossover, which is inferred from Figure 18 in view, future developments can be expected.

7. CONCLUSION This chapter reviewed carbon materials involved in studies of the dissemination of the practical use of PEFCs. The association between PEFCs and carbon materials is profound. Developments are expected in a complementary manner between these two study fields in the future.

REFERENCES

FIGURE 17 DMFC performance using membrane and electrode with fullerene-based proton conductor under ambient conditions [48].

[1] Vielstich W, Lamm A, Gasteiger HA. Handbook of fuel cells. John Wiley and Sons Ltd; 2003. [2] Kitta S, Uchida H, Watanabe M. Metal separators coated with carbon/resin composite layers for PEFCs. Electrochim Acta 2007;53:2025e33.

Chapter | 2.11

Carbon Materials used for Polymer Electrolyte Fuel Cells

[3] Mori Y, Ueda M, Hashimoto M, Aoi Y, Tanase S, Sakai T. Amorphous carbon coated stainless separator for PEFCs. Surf Coat Tech 2008;202:4094e101. [4] E-TEC, Inc. Catalogue “Gas Diffusion Electrodes and Catalyst Materials”. [5] Tada T. High dispersion catalysts including novel carbon supports. In: Vielstich, et al., editors. Handbook of fuel cells, vol. 3. John Wiley and Sons Ltd; 2003. p. 481e8. [6] Uchida M, Ohara H, Sugawara Y. Polymer electrolyte fuel cells. Matsusita Tech J 1998;44:467e76 [in Japanese]. [7] Stevens DA, Dahn JR. Thermal degradation of the support in carbon-supported platinum electrocatalysts for PEM fuel cells. Carbon 2005;43:179e88. [8] Iijima S, Yudasaka R, Yamada R, Bandow S, Suenaga K, Kokai F, et al. Nano-aggregates of single-walled graphitic carbon nano-horns. Chem Phys Lett 1999;309:165e70. [9] Bessel CA, Laubernds K, Rodriguez NM, Baker RTK. J Phys Chem B 2001;105:1115e8. [10] Che G, Lakshmi BB, Fisher ER, Martin CR. Carbon nanotube membranes for electrochemical energy storage and production. Nature 1998;393:346e9. [11] Che G, Lakshmi BB, Martin CR, Fisher ER. Metal-nanoclusterfilled carbon nanotubes: catalytic properties and possible applications in electrochemical energy storage and production. Langmuir 1999;15:750e8. [12] Guo D-J, Qiu X-P, Chen L-Q, Zhu W-T. Multi-walled carbon nanotubes modified by sulfated TiO2: a promising support for Pt catalyst in a direct ethanol fuel cell. Carbon 2009;47:1680e5. [13] Joo SH, Choi SJ, Oh I, Kwak J, Liu J, Terasaki O, et al. Ordered nanoporous arrays of carbon supporting high dispersions of platinum nanoparticles. Nature 2001;412:169e72. [14] Li W, Ling C, Qiu J, Zhou W, Han H, Wei Z, et al. Carbon nanotubes as support for cathode catalyst of a direct methanol fuel cell. Carbon 2002;40:791e4. [15] Li W, Liang C, Zhou W, Qiu J, Zhou Z, Sun G, et al. Preparation and characterization of multi walled carbon nanotube-supported platinum for cathode catalysts of direct methanol fuel cells. J Phys Chem B 2003;107:6292e9. [16] Li W, Liang C, Zhou W, Qiu J, Li H, Sun G, et al. Homogeneous and controllable Pt particles deposited on multi-wall carbon nanotubes as cathode catalyst for direct methanol fuel cells. Carbon 2004;42:436e9. [17] Li W, Waje M, Chen Z, Larsen P, Yan Y. Platinum nanoparticles supported on stacked-cup carbon nanofibers as electrocatalysts for proton exchange membrane fuel cell. Carbon 2010;48: 995e1003. [18] Rajesh B, Thampi KR, Bonard J-M, Xanthopoulos N, Mathiew HJ, Viswanathan B. Carbon nanotubes generated from template carbonization of polyphenyl acetylene as the support for electrooxidation of methanol. J Phys Chem B 2003;107:2701e8. [19] Reddy ALM, Rajalakshmi N, Ramaprabhu S. Cobalt-polypyrrolemultiwalled carbon nanotube catalysts for hydrogen and alcohol fuel cells. Carbon 2008;46:2e11. [20] Shi Y, Yang R, Yuet PK. Easy decoration of carbon nanotubes with well dispersed gold nanoparticles and the use of the materials as an electrocatalyst. Carbon 2009;47:1146e51. [21] Steigerwalt ES, Deluga GA, Cliffel DE, Lukehart C. A PteRu/ graphitic carbon nanofiber nenocomposite exhibiting high relative

[22]

[23]

[24]

[25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

[33]

[34]

[35]

[36]

[37]

[38]

209

performance as a direct-methanol fuel cell anode catalyst. J Phys Chem B 2001;105:8097e101. Steigerwalt ES, Deluga GA, Lukehart CM. PteRu/Carbon fiber nanocomposites: synthesis, characterization and performance as anode catalysts of direct methanol fuel cells, A search for exceptional performance. J Phys Chem B 2002;106:760e6. Maruyama J, Abe I. Enhancement effect of an adsorbed organic acid on oxygen reduction at various types of activated carbon loaded with platinum. J Power Sources 2005;148:1e8. Shioyama H, Yamada Y, Ueda A, Kobayashi T. Screening of carbon supports for DMFC electrode catalysts by infrared thermography. Carbon 2003;41:607e9. Liu Y-C, Qiu X-C, Huang Y-Q, Zhu W-T, Wu GS. Influence of preparation process of MEA with mesocarbon microbeads supported Pt-Ru catalysts on methanol electrooxidation. Appl Electrochem 2002;32:1279e85. Liu Y-C, Qiu X-P, Zhu W-T, Wu G-S. Impedance studies on mesocarbon microbeads supported PteRu catalytic anode. J Power Sources 2003;114:10e4. Yang R, Qiu X, Zhang H, Li J, Zhu W, Wong Z, et al. Monodispersed hard carbon spherules as catalyst support for the electrooxidation of methanol. Carbon 2005;43:11e6. Shioyama H, Yamada Y, Ueda A, Kobayashi T. Graphite intercalation compounds as PEMFC electrocatalyst supports. Carbon 2005;43:2374e8. Dong L, Gari RRS, Li Z, Craig MM, Hou S. Graphene-supported platinum and platinumeruthenium nanoparticles with high electrocatalytic activity for methanol and ethanol oxidation. Carbon 2010;48:781e7. Hillenbrand LJ, Lacksonen JW. The platinum-on-carbon catalyst system for hydrogen anodes. J Electrochem Soc 1965;112: 249e52. Bagotzky VS, Skundin AM. Electrocatalysts on supports I. Electrochemical and adsorptive properties of platinum microdeposits on inert supports. Electrochim Acta 1985;29:757e65. Attwood PA, McNicol BD, Short RT. The electrocatalytic oxidation of methanol in acid electrolyte: preparation and characterization of noble metal electrocatalysts supported on pre-treated carbon-fiber papers. J Appl Electrochem 1980;10:213e22. Shioyama H, Honjo K, Kiuchi M, Yamada Y, Ueda A, Kuriyama N, et al. C2F6 plasma treatment of a carbon support for a PEM fuel cell electrocatalyst. J Power Sources 2006;161:836e8. Bagotzky VR, Tarasevich MR, Radyushkina KA, Levina OA, Andrusyova SI. Electrocatalysis of the oxygen reduction process on metal chelates in acid electrolyte. J Power Sources 1977;2: 233e40. Bron M, Fiechter S, Hilgendorff M, Bogdanoff P. Catalysts for oxygen reduction from heat-treated carbon-supported iron phenanthroline. J Appl Electrochem 2002;32:211e6. Sawai K, Suzuki N. Heat-treated transition metal hexacyanometallates as electrocatalysts for oxygen reduction insensitive tomethanol. J Electrochem Soc 2004;151:A682e8. Maruyama J, Okamura J, Miyazaki K, Abe I. Two-step carbonization as a method of enhancing catalytic properties of hemoglobin at the fuel cell cathode. J Phys Chem C 2007;111: 6597e600. Niwa H, Horiba K, Harada Y, Oshima M, Ikeda T, Terazuka K, et al. X-ray absorption analysis of nitrogen contribution to oxygen

210

[39]

[40]

[41] [42]

[43]

Handbook of Advanced Ceramics

reduction reaction in carbon alloy cathode catalysts for polymer electrolyte fuel cells. J Power Sources 2009;187:93e7. Ozaki JI, Anahara T, Kimura N, Oya A. Simultaneous doping of boron and nitrogen into a carbon to enhance its oxygen reduction activity in proton exchange membrane fuel cells. Carbon 2006;44:3358e61. Ozaki J-I, Tanifuji S-I, Furuichi A, Yabutsuka K. Enhancement of oxygen reduction activity of nanoshell carbons by introducing nitrogen atoms from metal phthalocyanines. Electrochim Acta 2010;55:1864e71. Hinokuma K, Ata M. Fullerene proton conductors. Chem Phys Lett 2001;341:442e6. Hinokuma K, Ata M. Proton conduction in polyhydroxy hydrogensulfated fullerenes. J Electrochem Soc 2003;150: A112e6. Tasaki K, DeSousa R, Wang H, Gasa J, Venkatesan A, Pugazhendhi P, et al. Fullerene composite proton conducting membranes for polymer electrolyte fuel cells operating under low humidity conditions. J Memb Sci 2006;281:570e80.

[44] Tasaki K, Gasa J, Wang H, DeSousa R. Fabrication and characterization of fullerene-Nafion composite membranes. Polymer 2007;48:4438e48. [45] Wang H, DeSousa R, Gasa J, Tasaki K, Stucky G, Jousselme B, et al. Fabrication of new fullerene composite membranes and their application in proton exchange membrane fuel cells. J Memb Sci 2007;289:277e83. [46] Shioyama H, Ueda A, Kuriyama N. Fullerene derivatives in PEMFC electrocatalyst. Extended Abstract of 2008 International Conference on Carbon, Nagano, P0005; 2008. [47] Shioyama H. Quenching of excited Ru(bpy)2þ 3 in Nafion ionomer solution and in Nafion membrane. Mol Cryst Liq Cryst 2009;515:230e8. [48] Hirakimoto T, Fukushima K, Li Y, Takizawa S, Hinokuma K, Senoo T. Fullerene-based proton-conductive materials for the electrolyte membrane and electrode of a direct methanol fuel cell. Extended Abstract of 214th ECS Meeting, Honolulu, #1108; 2008. [49] Li YM, Hinokuma K. Proton conductivity of phosphoric acid derivative of fullerene. Solid State Ionics 2002;150:309e15.

Chapter 2.12

Carbons for Supercapacitors Yasushi Soneda Energy Storage Materials Group, Energy Technology Research Institute, Tsukuba West Office, National Institute of Advanced Industrial Science and Technology, 16-1 Onogawa, Tsukuba, Ibaraki 305-8569, Japan

Chapter Outline 1. Electric Double-Layer Capacitors 2. General Aspects of Carbons for EDLCs 3. Template Method for Pore Structure Control of Electrode Carbons 4. Nitrogen-Enriched Carbons 5. Exfoliated Carbon Fibers

211 211 212 213 216

1. ELECTRIC DOUBLE-LAYER CAPACITORS Electric double-layer capacitors (EDLCs) using porous carbon electrodes have a long history [1], with the first patent appearing in 1957. In the 1970s, EDLCs were commercialized in memory-backup devices and the market was established with small-size devices having capacitances approximately up to the 10 F range. Conversely, medium- and large-size EDLC cells with a capacitance of 100e1000 F are anticipated for future applications such as power assistance in hybrid electric vehicles, and load leveling for pulsed current change in power generators using renewable energy, in which the storage device is required to work with a large and instantaneously fluctuating electric current flow. In comparison with secondary batteries, which work via electrochemical reactions between electrode materials and electrolyte, EDLCs are superior in such applications due to intrinsic characteristics such as high power output and long durability.

2. GENERAL ASPECTS OF CARBONS FOR EDLCs EDLCs are electric energy storage devices in which the electric double layer between a polarizable electrode surface and an electrolyte stores electric charge. Since the amount of electric charge stored in the electric double layer increases with increasing electrode surface area, activated carbons with high porosity are employed for the realization of EDLCs. However, overall performance of EDLCs is

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00013-7 Copyright Ó 2013 Elsevier Inc. All rights reserved.

6. Carbon-Coated Transition-Metal Carbides 7. Carbon Xerogels with Conductive Polymer Nano-Coating 8. Summary References

217 218 221 221

governed not only by the surface area of the activated carbon, but also by the pore structure, size, and distribution. The surface area of activated carbon is usually determined by the measurement of nitrogen gas adsorption isotherms and influenced mainly by micropores in the materials, which are classified as less than 2 nm in diameter. Conversely, the capacitance has been found to depend mainly on the surface area of carbon electrodes, even though high surface area carbons do not always give a large capacitance on carbons derived from the same precursor. This is attributed mainly to the fact that the micropores are not useful in forming the electric double layer of solvated ions, and do not offer sufficient accessibility to the electrolyte. Since the disadvantage of micropores is also notable under high-rate chargeedischarge conditions, the incorporation of mesopores (2 ~ 50 nm in diameter) with smaller resistance for ion transfer is considered to be necessary. The control of pore-size distribution in electrode materials is also important for the energy and power density of EDLCs, since these characteristics are strongly dependent on the properties of the electrode material, and there is usually a trade-off relationship. The pore structure of carbons in electrode materials has a definitive and complicated influence on the performance of EDLCs and is still under discussion [2,3]. The specific surface area of carbon materials is essentially finite. This fact implies that the improvement of specific capacitance by means of the enlargement of specific surface area has a finite limitation. Accordingly, the introduction of redox (Faradic) reactions at the

211

212

electrode surface to store electric charge has been widely attempted to enhance the specific capacitance of carbon materials. Since the capacitance due to these redox reactions differs from the electric double-layer capacitance, this kind of charge storage is called pseudocapacitance. The redox reactions that contribute to pseudocapacitance are a variety of electrochemical reactions and generally result in a certain degree of deterioration of durability. The energy density of a capacitor cell is determined by the product of electrode capacitance and the square of the working voltage, which is restricted by decomposition of electrolyte solutions or electrode materials. Electrolytes are aqueous and nonaqueous solutions of different types of acids, bases, or salts. The working voltage window of aqueous solutions is usually restricted to under ~1 V, due to the decomposition of water. On the other hand, commonly used organic solutions have a tolerance of around 2.5 V. This difference in the working voltage window results in about a 6-fold larger energy density for nonaqueous electrolytes compared with aqueous electrolytes. Recently, EDLCs using organic electrolytes have attracted great attention from industrial and fundamental points of view, because of the potential application in high-energy density devices [4,5]. In fact, most recent capacitor cells with high capacity have been realized by adoption of an organic electrolyte. Nonaqueous EDLCs, however, have several barriers to widespread commercialization, including cost, and, in some cases, safety of the components. On the other hand, the aqueous electrolyte for EDLCs is mainly dilute sulfuric acid, which has a long history of use in leadeacid secondary batteries, and is superior from the viewpoints of low cost, safety, and physical properties. Aqueous electrolytes, such as dilute sulfuric acid solutions, have higher electrical conductivity than the commonly used organic electrolytes. This is an important parameter in achieving a high-rate chargeedischarge performance. The operating temperature range of aqueous electrolyte solutions is also suitable for capacitor cells, when compared with the most organic solutions (except ionic liquids). Therefore, the development of new types of carbon electrodes, which can be used in aqueous electrolytes while attaining high energy density, is eagerly awaited. Single-walled carbon nanotubes (SWCNTs) are one of the most anticipated materials for EDLC electrode materials because of their large surface area, monotonic poresize distribution, and high electrical conductivity [6]. However, most SWCNT products are bundled, resulting in lower surface area than the ideal single carbon nanotube. In recent years, the so-called super growth method of SWCNT production (SG-SWCNT) was developed, producing highpurity and bundle-free SWCNTs at mass scale. The advantages of using SG-SWCNTs as EDLC electrodes have been reported by Tanaike et al. [7]. As mentioned above, the investigation of novel carbon materials is directly connected to the improvement of

Handbook of Advanced Ceramics

EDLC performance, since carbon is essential for capacitor electrode materials. The following section presents recent developments in carbon materials for EDLC electrodes, in which the enhancement of various characteristics was achieved by various techniques such as chemical and structural control, and hybridization with heteroatoms. Recently, capacitors using these novel carbon materials, having Faradic reactions and novel mechanisms to enhance the total performance, have become known as supercapacitors.

3. TEMPLATE METHOD FOR PORE STRUCTURE CONTROL OF ELECTRODE CARBONS A template method for the synthesis of various carbons has been developed for control of morphology and pore structure of the resulting products. For capacitor electrodes, the carbon materials should have sufficiently high electrical conductivity and well-developed pore structure. For this purpose, various template materials, such as mesoporous silica [8e12], zeolite [13], colloidal silica [14], and organic nanocomposite [15], have been examined for the carbonization of different carbon precursors. Among these template materials, mesoporous silica has been studied the most extensively. Hyeon et al. first developed porous carbons from mesoporous silica templates and reported their application as electrode materials for supercapacitors in an early paper [16]. The synthesis process comprises the incorporation of carbon precursor from vapor or liquid/solution phase into the narrow spaces throughout the template structure, followed by carbonization at elevated temperature under inert atmosphere. The template structure is then eliminated by using strong chemicals, such as hydrofluoric acid. Mesoporous carbons produced from a mesoporous silica template have attracted attention from various international groups, due to the homogeneous and monodisperse mesoporous structure. The resulting materials appear to be suitable for use as capacitor electrode materials. Although there are different types of mesoporous silica, two kinds of structure have been revealed to be most suitable as templates to produce porous carbons, namely, MCM-48 with three-dimensional continuous mesopores, and SBA-15 with a hexagonal arrangement of rod-like mesopores. The mesoporous carbons derived from these templates have a relatively high surface area from about 700 to 2000 m2/g, with highly developed mesopores. By choosing the template and controlling the synthesis conditions, the porous character of the product carbons can be changed, as shown in Table 1. The specific surface area and the pore-size distribution in the resulting carbons are strongly affected by the precursor compounds when

Chapter | 2.12

Carbons for Supercapacitors

213

TABLE 1 Specific Surface Area of Meso- and Microporous Carbons Synthesized by Template Method with Different Precursors BET surface area (m2/g)

Template material

Carbon precursor

SBA-15

Propylene (CVD)

713

Pitch

923

MCM-48

Sucrose

1470

Divinylbenzene

1233

Propylene (CVD)

850

Ref. [9]

[9]

Pitch

1300

Sucrose

2000

Phenol

1257

Divinylbenzene

1171

Zeolite

Furfuryl alcohol/ propylene

4100

[13]

Colloidal silica

Phenol

1217

[14]

[8]

SBA-15 and MCM-48 are used. The largest specific area in the resulting carbons was achieved by a type of zeolite template, reaching to ~4000 m2/g. In most cases, the electric double-layer capacitance of carbon materials is closely related with the BET (BrunauereEmmetteTeller theory) surface area. This is also true of the mesoporous carbons; Figure 1 shows the strong correlation of the gravimetric capacitance of porous carbons with the specific surface area. In other words, all the carbons have the areal specific capacitance in the range from 0.1 to 0.2 F/m2. Since the areal specific capacitances of mesoporous carbons apparently coincide with those of usual activated carbons, the advantage of using mesoporous carbons from the template method is ambiguous from the

FIGURE 1 Relation between the specific surface area and the gravimetric capacitance of mesoporous carbons and activated carbons in a three-electrode cell with 1 mol/dm3 H2SO4 electrolyte.

viewpoint of volumetric capacitance, due to the higher porosity of mesoporous carbon. Conversely, mesoporous carbons are considered to be suitable for high-rate chargeedischarge conditions (high current density), because the developed mesopores with homogeneous and narrow pore-size distribution allow fast transfer of solvated ions in the electrolyte solution. Since the formation of the electric double layer is essentially accompanied by ion transfer from the bulk solution to the solidesolution interface, larger pores are favorable for this phenomena due to smaller ion-transfer resistance. If the carbon material possesses a large amount of meso- or macropores, the porosity increases and the volumetric capacitance decreases. Therefore, the balance of pore-size distribution is certainly an important factor in obtaining optimized capacitive behavior. Microporous carbons formed using a zeolite template have a narrow distribution of micropore size, and extraordinarily large surface areas as high as 4000 m2/g [13]. The gravimetric capacitance values of such microporous carbons are smaller than expected from the high specific surface area, when compared to other high surface area activated carbons. However, significantly higher rate performance, even in organic electrolytes (in which the solvated ions have large diameter), has been reported. The periodicity of micropores with monotonic diameter is considered to be the reason for the small resistance of ion transfer at high current density polarization.

4. NITROGEN-ENRICHED CARBONS Some layered silicate materials, such as clay and mica, are known to accommodate small nitrogen-containing organic solvent molecules in their chemical structures. These can be carbonized in nano-space, in between the adjacent silicate layers. The resulting carbons have a morphology reflecting the template of the layered silicate materials, resulting in very thin films that are transparent in electron micrographs, and that are circular with a diameter of a few tens of nanometers. Since the intercalation of solvent molecules inside the layered structure is mainly due to interaction of the nitrogen-containing molecules with cations between the silicate layers, organic molecules such as quinoline or pyridine are preferably employed. The carbons obtained from this process contain substitutional nitrogen heteroatoms originating from the organic precursor molecules, and are generally referred to as nitrogen-enriched carbons. Nitrogen-enriched carbons derived from expandable fluorinated mica have a specific surface area of about 80e100 m2/g and a gravimetric capacitance of about 100e160 F/g in sulfuric acid electrolyte (Figure 2) [17]. This results in the areal specific capacitances ranging from 1.2 to 2.2 F/m2, which are 10-fold larger than those of conventional

214

Handbook of Advanced Ceramics

FIGURE 2 Cyclic voltammogram of the nitrogen-enriched carbons derived from pyridine with mica template in a three-electrode cell with 1 mol/dm3 H2SO4 electrolyte at 1 mV/s of scan rate.

activated carbons. Such large capacitance values per unit surface area are not explainable by simple charge accumulation in the electric double layer, and are considered to be due to the pseudocapacitance related to nitrogen functionality in the carbon skeleton. Stimulated by such nitrogen-enriched carbons derived from mica templates, a variety of carbon materials synthesized from nitrogen-containing organic precursors have been extensively studied for EDLC electrodes [18e22]. One representative material is carbon synthesized from polymerized melamine (C3H6N6), using a mica template. This material contains more than 20 wt% nitrogen after heat treatment at 800  C [23]. Electrochemical testing in sulfuric acid electrolyte indicated that the melamine-derived nitrogen-enriched carbons have 150 ~ 200 F/g gravimetric capacitance without an apparent redox reaction, even though the materials have only 15 ~ 400 m2/g specific surface area [19]. The simple calculation of areal specific capacitance using the above results gives extraordinary values which are probably not realistic. Nitrogen atoms in carbon materials have various types of chemical bonds, which can be elucidated by X-ray photoelectron spectroscopy (XPS). Figure 3 shows that nitrogen in melamine-derived carbons exists mainly in the pyridinic, quaternary, and oxidized states, and that the population of those chemical states changes with heat treatment temperature. By comparing materials heattreated from 850 to 1000  C, quaternary bonding (substitutional nitrogen bonded to three neighboring carbon atoms) increases with increasing temperature, and pyridinic bonding (nitrogen located at the outside edges of carbon rings and bonded to two neighboring carbon atoms) decreases with increasing temperature. The role of nitrogen in carbon materials for electrochemical capacitors is considered to be the fast faradic reaction between cations and lone pair electrons in the pyridinic

FIGURE 3 X-Ray photoelectron spectroscopy of N1s core level on the nitrogen-enriched carbon derived from melamine: N-p, N-q, and N-o denote pyridinic, quaternary, and oxidized chemical state of nitrogen bonds, respectively.

state. However, the nitrogen-enriched carbons usually do not show any redox reaction in electrochemical measurements such as cyclic voltammograms (CV). The high capacitance of these materials has been observed not only in acidic media but also in basic solutions [24]. Although the mechanism behind large gravimetric capacitance in carbons with low surface area derived from melamine resin is thought to be some kind of electrochemical reaction, cyclability has been confirmed up to few million cycles while maintaining the initial capacitance value. Other materials in this category are carbonized melamine foam, and mesoporous nitrogen-enriched carbons derived from mesoporous silica templates. The carbonization of melamine foam (used as commercially available cleaning sponges) results in the formation of a unique carbon foam which has resiliency and a similar shape to the original sponge [25]. The gravimetric capacitance of carbonized foam in sulfuric acid electrolyte, measured by galvanostatic chargeedischarge cycling, reached 240 and 220 F/g at 0.2 A/g and 1.0 A/g discharge current density, respectively (Figure 4). By using mesoporous silica templates and nitrogencontaining organic compounds, nitrogen-enriched mesoporous carbons have been produced to accomplish high capacitance in tandem with nitrogen-associated pseudocapacitance, and highly developed pore structure

Chapter | 2.12

Carbons for Supercapacitors

215

FIGURE 4 Galvanostatic chargeedischarge curves of the nitrogen-enriched carbon foams derived from commercial melamine foam (washing sponge) in a three-electrode cell with 1 mol/dm3 H2SO4 electrolyte at current densities of 1 and 2 A/g.

FIGURE 5 SEM (left) and TEM (right) images of the nitrogen-enriched mesoporous carbon. White bar corresponds to 200 nm.

[26]. Figure 5 shows scanning electron microscope (SEM) and transmission electron microscope (TEM) images of nitrogen-enriched mesoporous carbons synthesized from quinoline-polymerized pitch with an SBA-15 template. In the micrographs, a series of straight, aligned mesopores are clearly observed. The wall of the mesopore structure is composed of nitrogen-enriched carbon containing ~6 wt% nitrogen, and possessing a significant amount of micropores, revealed by the adsorption isotherm measurement. The gravimetric capacitance of related materials is listed

in Table 2. Clearly, nitrogen-containing mesoporous carbons show higher capacitances than nitrogen-free materials using the same template. The higher capacitance of nitrogen-enriched mesoporous carbons was also confirmed in organic electrolyte (1 mol/dm3 Et4NBF4 in propylene carbonate) in which specific capacitances of 100e140 F/g were obtained, exceeding that of activated carbons used in commercial capacitor devices (70 ~ 90 F/g). Similar enhancement by nitrogen enrichment in carbon materials was also recognized for microporous carbons

TABLE 2 Specific Surface Area and Gravimetric Capacitance of Nitrogen-Enriched Mesoporous Carbons BET surface area (m2/g)

Gravimetric capacitance * (F/g)

Template material

Carbon precursor

SBA-15

Nitrogen-containing quinoline pitch

615

221

Nitrogen-free pitch

595

143

MCM-48

Nitrogen-free polymer

1233

132

Nitrogen-containing quinoline pitch (A)

1710

290

Nitrogen-containing quinoline pitch (B)

1296

250

Nitrogen-free polymer

1171

131

*1 mol/dm3 H2SO4 electrolyte, current density 200 mA/g

216

derived from zeolite templates using an acetonitrile precursor [27].

5. EXFOLIATED CARBON FIBERS In the case of activated carbon for EDLC electrodes, the enhancement of specific capacitance was pursued mainly by increasing the specific surface area and controlling the pore-size distribution. Modification of surface functionality, introduction of heteroatoms, and the addition of metal compounds have also been extensively studied. Most of these studies, however, were performed on carbons with low crystallinity because of the moderate preparation temperature. There were no experiments on carbons with high crystallinity, for example, graphite. This is because it is very difficult to increase the specific surface area of graphite, which is considered to be essential for capacitor electrodes. Recently, exfoliated carbon fibers (ExCFs) which posses both high crystallinity and large surface area were developed, and their characteristic performance as capacitor electrodes was determined [28,29]. Carbon fibers (mesophase-pitch- or PAN-based) were used as starting materials. Their intercalation compounds were synthesized by electrolysis in 13 mol/dm3 HNO3. The intercalated carbon fibers thus obtained were rapidly inserted into a tubular furnace at 1000  C for 5 s for exfoliation. The detailed exfoliation behavior of carbon fibers is reported in several papers [30e32]. X-Ray diffraction patterns of pristine, electrolyzed, and exfoliated carbon fibers are shown in Figure 6. Conventional

FIGURE 6 X-Ray diffraction patterns of carbon fiber; (a) pristine, (b) electrolyzed in HNO3, (3) exfoliated (heat-treated at 1000  C), and (d) conventional activated carbon.

Handbook of Advanced Ceramics

activated carbon with low crystallinity is also shown for comparison. The pristine carbon fiber has high crystallinity, showing a sharp diffraction peak with a d spacing of 0.337 nm (Figure 6a). After electro-oxidation in HNO3, the diffraction peaks of the graphite structure disappeared and a broad peak at a d-value of 0.78 nm appeared (Figure 6b). According to the chemical analysis of this material, the product of this treatment coincided with graphite oxide. Subsequently, graphite oxide was completely decomposed after heat treatment at 1000  C (Figure 6c) and the carbon structure was recovered, although the carbon layer stacking along the c-axis is smaller than that in pristine carbon fibers, because of the exfoliation process. SEM observations revealed that individual fibers were converted to bundles of thin filaments by exfoliation, reflecting the fibril texture along the fiber axis. The nitrogen adsorption isotherm for ExCFs at 77 K showed a welldeveloped mesoporous character, with about 300 m2/g BET surface area. Mercury porosimetry indicated that ExCFs have total pore volume of about 20 cm3/g, which can be reasonably attributed to meso- and macropore regions among the thin filaments. Extremely large capacitance values reaching 555 and 450 F/g were obtained by CV and chargeedischarge measurements, respectively, when 18 mol/dm3 H2SO4 electrolyte was used. The CV curve slightly deviated from the ideal rectangular shape, but no redox peak was observed. The galvanostatic experiment at low current density gave a triangular chargeedischarge curve with linear change in potential, reflecting the CV curve, without peaks. Galvanostatic cycling with high current density (500 mA/g) in 18 mol/dm3 H2SO4 was measured to be stable for more than 7000 chargeedischarge cycles, although the specific capacitance was almost half that observed in the former experiment because of the high current density. When 1 mol/dm3 electrolyte was used, the capacitance was 160 and 117 F/g measured by CV and the chargeedischarge curve, respectively. Figure 7 shows the dependence of areal specific capacitance for ExCFs and activated carbon fibers (ACFs), obtained from the 50th discharge curve of galvanostatic cycling, plotted against sulfuric acid electrolyte concentration. The capacitance value for ExCFs increased gradually with increasing concentration up to 10 mol/dm3, then suddenly jumped to an extremely large value in the higher concentration region, reaching about 1.4 F/m2 (450 F/g). However, the capacitance for ACFs does not show any trend with changing electrolyte concentration, with an approximately constant, small value of about 0.1 F/m2. Since the capacitance of ExCFs in concentrated sulfuric acid is considerably large compared with the reported value of conventional carbon materials, the value for ExCFs obtained here cannot be explained by only the electric double layer. The pseudocapacitance also has to be taken into account.

Chapter | 2.12

Carbons for Supercapacitors

217

FIGURE 7 Dependence of areal specific capacitance of ExCF and activated carbon fiber on the concentration of H2SO4 electrolyte.

Exfoliated natural graphite (ExNG) with the highest crystallinity showed a similar trend with electrolyte concentration, but abnormally large capacitances were obtained in 12 and 18 mol/dm3 solutions, and a distinct intercalation reaction of H2SO4 molecules into graphite interlayer was confirmed. Considering the above observation, the enhancement of the specific capacitance of ExCFs with concentrated electrolyte is not associated with the formation of stage structure caused by the intercalation reaction, although there may be some strong interaction between the electrode and the H2SO4 molecules. The large capacitance of ExCF in concentrated electrolyte implies that the charge transfer reaction (Faradic reaction) occurs at the electrode surface, since the capacitance values were unreasonably large and the intercalation of H2SO4 is obvious. The lack of stage structure formation in ExCF may due to lower crystallinity and a more distorted structure compared to graphite. The behavior of ExCF is better than ExNG, since the formation of stage structures introduces a volume change in the electrode, resulting in electrode collapse at short cycling times.

6. CARBON-COATED TRANSITION-METAL CARBIDES Several types of transition-metal oxides are known to work as capacitor electrode materials, since the metal undergoes a change in oxidation state and then a hydrated state over a wide range working window. The apparent CV curves from these compounds show a pseudo-rectangular box shape resulting in the origin of the term pseudocapacitance. The most well-known example is ruthenium oxide, which has the highest capacitance in acidic media. Ruthenium oxide, however, is not considered suitable for use in

FIGURE 8 X-Ray powder patterns of the carbon-coated tungsten carbide fine particles with different synthesis temperature.

commercial device applications because of its extremely high cost. Inagaki et al. developed a simple process to coat carbon onto various metallic particles and compounds, and studied their functionality in different applications [33]. A similar procedure was applied for tungsten and molybdenum compounds to exploit their large change in oxidation state, as a replacement for ruthenium [34,35]. It was found that the thus-obtained carbon-coated tungsten and molybdenum carbides were easily converted to their corresponding oxides in sulfuric acid electrolyte, and that they showed stable capacitive behavior. The starting materials used were hydroxyl propyl cellulose (HPC) as the carbon precursor, and reagent-grade K2WO4 and K2MoO4 powders as the precursors for tungsten and molybdenum carbide, respectively. The HPC was dissolved in water at a concentration of 10 mol/dm3, and the carbide precursor (either K2WO4 or K2MoO4) was dissolved into this HPC solution in a mass ratio of 1/1. The mixed precursor solutions thus prepared became gels at room temperature, were dried at 100  C, and then heattreated at a temperature between 800 and 1050  C in Ar gas flow. After heat treatment, the obtained powders were washed by 1 mol/dm3 H2SO4 and then distilled water, in order to remove any potassium metal which might remain. Figure 8 shows the change in X-ray diffraction pattern with heat treatment temperature for mixtures of HPC and K2WO4. Above 850  C, two types of tungsten carbide were observed (W2C and WC), as well as metallic W. With increasing temperature, the relative intensities of the

218

diffraction peaks for WC increased, suggesting that the HPC carbon precursor decomposes to carbonaceous materials, which coat K2WO4 particles. K2WO4 particles covered by the carbonaceous layer probably then decompose to amorphous tungsten oxide below 850  C. This amorphous tungsten oxide reacts with the coated carbonaceous materials to form W2C, and finally WC. From TEM observation, opaque particles of 30e50 nm in size with irregular shape were observed, and reasonably assumed to be WC. They appear to be coated by a carbon layer, assumed to be porous. In the case of K2MoO4, the formation of Mo2C was confirmed above 800  C. Figure 9 shows the CV curve of carbon-coated tungsten components e a mixture of WC, W2C, and W in 1 mol/dm3 sulfuric acid electrolyte. No apparent redox peaks were observed, and the gravimetric capacitance was calculated to be 104 F/g from this voltammogram. It was revealed by chemical analysis that only 13 wt% of carbon was included in the composite material, indicating the important contribution of tungsten components on the capacitive behavior. The X-ray powder pattern of the composite materials after one voltammetry cycle showed diffraction lines from tungsten oxyhydrate (H0.10WO3$1.06H2O), and all tungsten components found in the raw composite had disappeared. This implies that the first polarization of tungsten carbide in sulfuric acid causes an electrochemical reaction and the formation of oxyhydrate from the parent carbide. It is known that tungsten oxyhydrate undergoes a change in oxidation state associated with the hydration composition along with the polarization potential. Therefore, these changes might be responsible for the pseudocapacitive behavior in sulfuric acid electrolytes. Since

FIGURE 9 Cyclic voltammogram of the carbon-coated tungsten carbide fine particles in a three-electrode cell with 1 mol/dm3 H2SO4 electrolyte at 1 mV/s of scan rate.

Handbook of Advanced Ceramics

the tungsten compounds were coated with carbon and remained inside the carbon layer, the electrode fabricated from this composite material showed stable chargee discharge cycling for several hundred cycles. Similar behavior was also found for carbon-coated molybdenum carbide particles.

7. CARBON XEROGELS WITH CONDUCTIVE POLYMER NANO-COATING Poly(p-fluorophenylthiophene) (PFPT), a conductive polymer, is one of the most promising candidates for electrode materials because it has a large charge storage capacity for both p- and n-doping, and a large potential difference between these two processes [36]. However, the rate performance of polymer-based capacitors is inferior to that of EDLC, because the rate of diffusion of ions incorporated in the polymer is much slower than that of accumulation of ions in an electric double layer. In order to improve the rate performance, it is necessary to shorten the diffusion path of ions by forming the polymers into thin layers, on porous substrates with high electrical conductivity. Carbon xerogels (CXGs) are derived via solegel polycondensation, subcritical drying, and high-temperature heat treatment. They have properties making them suitable for use as electrode materials, including high electrical conductivity, porosity, and corrosion resistivity. These properties are also desirable in substrates for PFPT coating by polymerization. In order to study PFPT/CXG composite electrodes, the relationship between electrochemical properties and layer thickness was first investigated on a PFPT layer formed on a nonporous carbon film, with the aim of determining the optimal layer thickness. Then, carbon xerogels were coated with a PFPT layer having appropriate thickness, and the electrochemical properties of the resulting electrodes were characterized. The monomer p-fluorophenyl thiophene (FPT) was synthesized by a nickel(II)-catalyzed coupling reaction, according to the literature (Figure 10) [37]. By using electrochemical polymerization of dissolved FPT in a propylene carbonate (PC) solution of 1 mol/dm3 tetraethylammonium tetrafluoroborate (Et4NBF4), a PFPT layer with uniform thickness was formed on the carbon film. The thickness of the PFPT layer increased in proportion to the

FIGURE 10 Synthesis scheme of p-fluorophenyl thiophene (p-FPT) monomer.

Chapter | 2.12

Carbons for Supercapacitors

polymerization charge per unit surface area (Qpoly), and reached about 8 mm at Qpoly ¼ 0.28 mAh/cm2 (Figure 11). In the cyclic voltammogram for the PFPT layer, two peak pairs, assigned to p- and n-doping and de-doping, appear in the potential ranges of 0.2e0.8 V and e2.1 to e1.6 V, respectively. The capacity of p- and n- de-doping at various sweep rates is shown in Figure 12, as a function of Qpoly. In both p- and n-doping, the maximum capacitance exceeds 210 F/g. Above Qpoly ¼ 0.01 mAh/cm2, the capacity decreases with increasing Qpoly, and the decrease is made larger with increasing sweep rate. This is because ionic species cannot be inserted inside the PFPT layer, due to the insufficiently slow diffusion of ionic species in the polymer matrix. Conversely, below Qpoly ¼ 0.01 mAh/ cm2, the thickness of the PFPT layer is so small that the ionic species can be inserted across the whole layer even under fast potential sweeping. Thus, the capacity is determined by the structure of the PFPT layer. The change of capacity with Qpoly suggests that there is a structural difference between PFPT layers with different thickness. In order to investigate the change in reaction mechanism during polymerization, the stoichiometric number (N) was calculated from the neutralization charge, the polymerization charge, and the molecular weight of FPT. As shown in Figure 13, the value of N is considerably larger than the ideal case of 2 [38] at Qpoly < 0.01 mAh/cm2. This indicates that some preferential side reaction takes place. It is therefore considered that the small capacity at Qpoly < 0.01 mAh/cm2 results from the existence of structural defects introduced by side reactions. From the results shown above, it is concluded that a layer thickness of about 200 nm, achieved by polymerization with Qpoly ¼ 0.01 mAh/cm2, is optimal from the viewpoint of maximizing capacity.

FIGURE 11 Dependence of deposited mass of poly p-fluorophenyl thiophene (PFPT) on carbon film electrode against polymerization charge.

219

The pores in the CXG substrate have to be large enough so as not to destroy contact with the external surface even after the PFPT coating is applied. At the same time, the walls of CXG should have the minimum volume in order to maximize the PFPT content in the coated composite material, while maintaining high electrical conductivity. By taking into account the estimated densities of the CXG skeleton and the PFPT (1.80 and 0.94 g/cm3, respectively) the thickness of the CXG walls must be below 100 nm to give 90 wt% PFPT in the coated composite. CXGs were prepared from aqueous solutions of resorcinol, formaldehyde, and sodium carbonate, using

(a)

(b)

FIGURE 12 Capacity of (a) p- and (b) n-de-doping processes for PFPT layers formed on carbon film electrodes vs. polymerization charge density. Rates of potential change are shown in the figure.

220

Handbook of Advanced Ceramics

(a)

(b) FIGURE 13 Change in stoichiometric number of electron during the electrolytic polymerization of p-FPT with polymerization charge density.

a procedure described in the literature [39]. The solution was gelated by a two-step heat treatment at 20  C for 7 days then 50  C for 4 days. In order to minimize shrinkage during drying, water in the wet gel was first exchanged with acetone, and then acetone was removed from the gel by evaporation. The organic xerogels were carbonized by heating at up to 1500  C under nitrogen gas flow. By optimizing the preparation conditions (such as gelation period), the above-mentioned structural parameters required of the pore diameter and wall thickness were achieved. Figure 14a indicates that the thus-obtained CXG has a uniform distribution of primary carbon spheres, which are connected in a three-dimensional network. The size of intra-particle mesopores increased with gelation period, exceeding the target size of 400 nm after 7 days (Table 3). To form PFPT on CXG, the PC solution saturated with FPT was first impregnated into the CXG electrode under

FIGURE 14 SEM images of carbon xerogel electrodes (a) before and (b) after applying PFPT coating.

reduced pressure. Then, the electrode was placed in a 0.5 mol/dm3 Et4NBF4 PC solution, and a constant potential of 1.05 V was applied. The current decreased as polymerization progressed, due to the reduction of the amount of residual monomer. After the current reached 1 mA, the PFPT on the CXG was neutralized by sweeping the potential to 0 V. Figure 14b shows that the PFPT uniformly coated the primary carbon spheres in CXG, with

TABLE 3 Preparation Conditions and Characteristics of Pore Structure in Carbon Xerogels Sample

t20 (day)

rbulk (g cm 3)

rapp (g cm 3)

Vtotal (cm3 g 1)

Sext (m2 g 1)

Dave (nm)

A

0

0.40

1.37

1.74

61

113

B

1

0.36

1.34

1.97

56

139

C

2

0.33

1.34

2.26

55

164

D

3

0.35

1.34

2.07

42

193

E

4

0.32

1.34

2.37

39

237

F

7

0.26

1.33

3.08

27

446

t20: Gelation period at 20  C, rbulk: Bulk density, rapp: Apparent density, Vtotal: Total pore volume, Sext: External surface area, Dave: Average pore diameter

Chapter | 2.12

Carbons for Supercapacitors

FIGURE 15 Cyclic voltammograms of PFPT-coated carbon xerogel electrodes with different scan rate.

a thickness of about 100 nm. The cyclic voltammograms of PFPT-coated CXG gave the same electrochemical response for p- and n-doping and de-doping as PFPT on a carbon film, over a wide range of scan rates (Figure 15).

8. SUMMARY Carbon materials with various structural characteristics have been developed and employed as capacitor electrode materials. The strategy for the enhancement of capacitance is also based on a broad array of mechanisms. The variation in structural diversity of carbon materials and combination with heteroatoms and other materials will enable them to be applied successfully in supercapacitors for future applications.

REFERENCES [1] Conway BE. Electrochemical supercapacitors. New York: Kluwer Academic; 1999. [2] Mochida I, Lee S-I, Mitani S, Yoon S-H, Korai Y. Performances, working mechanism and future development of active carbons in super capacitor. TANSO 2003;210:250e7. [3] Kim YJ, Tantrakarn K, Abe Y, Yanagiura T, Takeuchi K, Oshida K, et al. Correlation between the capacitor performance and pore structure. TANSO 2006;221:31e9. [4] Takeda T, Endo M. Development and application of electric double-layer capacitor with sulfuric acid electrolyte. TANSO 1999;189:179e87. [5] Morimoto T. Electric double-layer capacitor using organic electrolyte. TANSO 1999;189:188e96.

221

[6] Shiraishi S, Oya A. Electric double-layer capacitance properties of carbon nanotubes and carbon nanofibers, e the present state as capacitor electrode. Electrochemistry 2003;71(10):887e93. [7] Tanaike O, Imoto S, Hatori H, Futaba DN, Hiraoka T, Hata K. EDLC properties of carbon nanotubes prepared by “super growth” method. 32th Annual meeting abstracts of the Carbon Society of Japan, 2005;2A12, 330e1. [8] Yoon S-H, Lee J, Hyeon T, Oh SM. Electric double-layer capacitor performance of a new mesoporous carbon. J Electrochem Soc 2002;147:2507e12. [9] Vix-Guterl C, Frackowiak E, Jurewicz K, Friebe M, Parmentier J, Beguin F. Electrochemical energy storage in ordered porous carbon materials. Carbon 2005;43:1293e302. [10] Fuertes AB, Lota G, Centeno TA, Frackowiak E. Templated mesoporous carbons for supercapacitor application. Electrochim Acta 2005;50:2799e805. [11] Li L, Song H, Chen X. Pore characteristics and electrochemical performance of ordered mesoporous carbons for electric doublelayer capacitors. Electrochim Acta 2006;51:5715e20. [12] Liu HY, Wang KP, Teng H. A simplified preparation of mesoporous carbon and the examination of the carbon accessibility for electric double layer formation. Carbon 2005;43:559e66. [13] Yao A, Gouki Y, Kawasaki S, Okino F, Ma Z, Kyotani T, et al. Electrochemical double layer capacitor performance of microporous carbons using zeolite template. 29th Annual Meeting Abstracts of the Carbon Society of Japan, 2002, P12, 222e3. [14] Moriguchi I, Nakahara F, Furukawa H, Yamada H, Kudo T. Colloidal crystal-templated porous carbon as a high performance electrical double-Layer capacitor material. Electrochem Solid State Lett 2004;7:A221e3. [15] Tanaka S, Nishiyama N, Egashira Y, Ueyama K. Synthesis of ordered mesoporous carbons with channel structure from an organiceorganic nanocomposite. Chem Commun 2005;(16): 2125e7. [16] Lee J, Yoon S, Hyeon T, Oh SM, Kim KB. Synthesis of a new mesoporous carbon and its application to electrochemical doublelayer capacitors. Chem Commun 1999;(21):2177e8. [17] Kodama M, Yamashita J, Soneda Y, Hatori H, Nishimura S, Kamegawa K. Structural characterization and electric double layer capacitance of template carbons. Mat Sci Eng B 2004;108: 156e61. [18] Lota G, Grzyb B, Machnikowska H, Machnikowski J, Frackowiak E. Effect of nitrogen in carbon electrode on the supercapacitor performance. Chem Phys Lett 2005;404:53e8. [19] Kawaguchi M, Yagi S, Yamanaka T, Ito H. Electric double layer capacitances of carbonaceous materials containing nitrogen prepared from nitrogen-rich organic precursors. Carbon 2006. Extended Abstracts, 3B3. [20] Beguin F, Raymundo-Pinero E, Lota G, Leroux F, Frackowiak E. High performance pseudo-capacitors based on nitrogen and oxygen enriched carbons. Carbon 2006. 2006 Extended Abstracts, 3B2. [21] Kijima M, Oda T, Yamazaki T, Tazaki Y, Nakamura J. Efficient thermal conversion of poly(pyridinediylbutadiynylene)s to nitrogencontaining microporous carbon. Chem Lett 2006;35:844e5. [22] Kibe M, Shiraishi S, Oya A. Effect of heteroatom doping on double layer capacitance and electrochemical properties of porous carbon electrode. 31th Annual Meeting Abstracts of the Carbon Society of Japan 2004; 1A12, 26e7.

222

[23] Hulicova D, Yamashita J, Soneda Y, Hatori H, Kodama M. Supercapacitors prepared from melamine-based carbon. Chem Mater 2005;17:1241e7. [24] Hulicova D, Kodama M, Hatori H. Electrochemical performance of nitrogen-enriched carbons in aqueous and non-aqueous supercapacitors. Chem Mater 2006;18:2318e26. [25] Kodama M, Yamashita J, Soneda Y, Hatori H, Kamegawa K. Preparation and electrochemical characteristics of N-enriched carbon foam. Carbon 2006. 2006 Extended Abstracts, 2P49. [26] Kodama M, Yamashita J, Soneda Y, Hatori H, Kamegawa K, Moriguchi I. Structure and electrochemical capacitance of nitrogen-enriched mesoporous carbon. Chem Lett 2006;35:680e1. [27] Ichikawa J, Sugiura S, Yotsui T, Okino F, Hou PX, Yamazaki T. et al., Effect of nitrogen doping and carbon precursor on electric double-layer capacitor performance using ordered microporous carbon. 32nd Annual Meeting Abstracts of the Carbon Society of Japan 2005; P33, 176e7. [28] Soneda Y, Toyoda M, Hashiya K, Yamashita J, Kodama M, Hatori H, et al. Huge electrochemical capacitance of exfoliated carbon fibers. Carbon 2003;41:2680e2. [29] Soneda Y, Yamashita J, Kodama M, Hatori H, Toyoda M, Inagaki M. Pseudo-capacitance on exfoliated carbon fiber in sulfuric acid electrolyte. Appl Phys A 2006;82:575e8. [30] Toyoda M, Shimizu A, Iwata H, Inagaki M. Exfoliation of carbon fibers through intercalation compounds synthesized electrochemically. Carbon 2001;39:1697e707.

Handbook of Advanced Ceramics

[31] Toyoda M, Katoh H, Inagaki M. Intercalation of nitric acid into carbon fibers. Carbon 2001;39:2231e4. [32] Toyoda M, Sedlacik J, Inagaki M. Intercalation of formic acid into carbon fibers and their exfoliation. Synthetic Met 2002;130: 39e43. [33] Inagaki M, Hirose Y, Matsunaga T, Tsumura T, Toyoda M. Carbon coating of anatase-type TiO2 through their precipitation in PVA aqueous solution. Carbon 2003;41:2619e24. [34] Morishita T, Soneda Y, Tsumura T, Inagaki M. Preparation of porous carbons from thermoplastic precursors and their performance for electric double layer capacitors. Carbon 2006;44:2360e7. [35] Liu W, Soneda Y, Hatori H. A novel carbothermal method for the preparation of nano-sized WC on high surface area carbon. Chem Lett 2006;35:1148e9. [36] Rudge A, Raistrick I, Gottesfeld S, Ferraris JP. A study of the electrochemical properties of conducting polymers for application in electrochemical capacitors. Electrochim Acta 1994; 39:273e87. [37] Sarker H, Gofer Y, Killian JK, Poehler TO, Searson PC. Synthesis and characterization of fluoro-substituted polyphenylthiophenes for charge storage applications. Synthetic Met 1997;88:179e85. [38] Roncali J. Conjugated poly(thiophenes): synthesis, functionalization, and applications. Chem Rev 1992;92:711e38. [39] Saliger R, Bock V, Petricevic R, Tillotson T, Geis S, Fricke J. Carbon aerogels from dilute catalysis of resorcinol with formaldehyde. J Non Cryst Solids 1987;221:144e50.

Chapter 3.1

Silicon Carbide and Other Carbides: From Stars to the Advanced Ceramics Branko Matovic* and Toyohiko Yano** *

Vinca Institute of Nuclear Sciences, University of Belgrade, 11001 Belgrade, Serbia, Former Affiliate: Research Laboratory of Nuclear Reactor, Tokyo Institute of Technology, ** Research Laboratory for Nuclear Reactors, Tokyo Institute of Technology, 2-12-1 Ookayama, Meguro, Tokyo 152-8550, Japan

Chapter Outline 1. 2. 3. 4. 5.

Introduction Carbides in Nature Transition Metal Carbides Covalent Carbides Synthesis

225 226 228 233 235

1. INTRODUCTION Carbides are a class of ceramic materials that have drawn great interest to the scientific community due to their exceptional profile of mechanical, physical, chemical, and even biological properties. Therefore, carbides are the basis of advanced ceramics with numerous applications and with a future prospect. Although they have been known for over one hundred years, and being important materials in many technological fields, such as tool industry, mechanical engineering, engine industry, optic, electronic, catalytic, nuclear technology, and chemical industry, most of their applications are recent. Although an extensive compilation of the carbides exists in textbooks, almost no one treats the natural carbides. The poorly known data are discussed and pointed. Therefore, in this chapter, carbides are described and discussed in the context of their natural occurrences as well as their same synthetic phases. Since native carbides are refractory compounds, special attention is given to the carbides that possess high-melting point, particularly on their synthesis, structure, and some high-temperature properties. From the chemistry point of view, carbides are compounds in which carbon is combined with less electronegative elements of a metal or a semimetal. Carbon is the chemical element (symbol C) with atomic number 6. It belongs to group 14 on the periodic table, and, thus, it is nonmetallic. Carbon is tetravalent, enabling 4 electrons available to form covalent chemical bonds. Since carbon is the one of the most abundant chemical elements in the universe by mass (after hydrogen, helium, and oxygen) as

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00014-9 Copyright Ó 2013 Elsevier Inc. All rights reserved.

6. Extreme Environment Application 7. Importance of Natural and Synthetic Carbides References

240 241 241

well as in the Earth’s crust, it forms more compounds than any other element. Among them, carbides are of great significance as engineering materials with numerous applications. Because of numerous carbon compounds, sometimes it becomes extremely cumbersome to define to which group the corresponding compound belongs. By convention, the term carbide is only used for compounds formed by carbon with other elements of lower or nearly equal electronegativity. They may be classified on the basis of chemical composition, chemical bond, method of manufacture, physical form, or according to their applications. However, the interrelated atomic features play an essential role in the formation of carbides, i.e., the difference of the electronegativities (DEN) of carbon and the metal or the semimetal, the atomic size of the respective elements, and their corresponding bonding type. According these criteria, several classes of carbides are usually distinguished: (i) salt like, with high DEN and ionic behavior, (ii) covalent compounds, with small, almost vanishing DEN and strong covalent bonding, (iii) interstitial compounds, and (iv) “intermediate” transition metal carbides with intermediate DEN and metallic properties.

1.1. Salt-like Carbides (SCs) Salt-like carbides are formed by the elements of groups I, II, and III. Because of the highly electropositive elements of alkali, alkaline-earth metals, and group III metals,

225

226

Handbook of Advanced Ceramics

carbon can be assumed to be anion and thus the corresponding bonding is anionic. According to the type of the carbon anion, salt-like carbides can be subdivided into isolated C anions, and C2 and C3 dumbbells, namely methanides with C4 anions (Be2C and Al4C3), acetylides with C22 anions (Na2C2 and CaC2), and allylenides with C43 anions (Li4C3 and Mg2C3).

1.1.1. Methanides Methanides group or methides get the name because of methane gas created during their reaction with water. The most important methanides are beryllium carbide (Be2C) and aluminum carbide (Al4C3). However, only Be2C meets the criteria as high-melting-point carbide:

approximately 0.135 nm) of the groups IV, V, and VI and 4th, 5th, and 6th periods, which act as a host lattice for the small carbon atom that occupies the interstices of the close-packed metal atoms. They are often described as interstitial carbides. According to the mode of metal atom arrangement, interstitial carbides possess 1:1 or 2:1 stoichiometry. When the metal atoms are cubic close packed (ccp), all octahedral interstices are filled by carbon, resulting in 1:1 stoichiometry, such as rock salt structure. On the other hand, when the metal atoms are hexagonal close packed (hcp), the octahedral interstices lay directly opposite each other side of the layer of metal atoms and only one is filled by carbon achieving 2:1 stoichiometry.

C4 þ 4Hþ / CH4

(1)

1.4. Intermediate Transition (IM) Carbides

Al4 C3 þ 12H2 O / 3CH4 ðgÞ þ 4AlðOHÞ3

(2)

Intermediate carbides are formed from transition elements of groups VII and VIII. The size of metal ion of the intermediate transition carbide is smaller than the critical 0.135 nm and accommodation of carbon atom in interstitial positions is impossible without distortion of the crystal lattice. They are characterized by multiple stoichiometries and from the view of structure they are not interstitial but are more complex.

1.1.2. Acetylides Acetylides are carbides which can be described as salt-like derivates of C22 . Anion C22 has a triple bond between two carbon atoms. Alkali metals and alkaline-earth metals form acetylides as well as lanthanoids with formula M2C3. Actinide compounds, with stoichiometry MC2 and M2C3, also belong to this subgroup. They react with water, giving acetylene and the corresponding hydroxide: CaC2 þ 2H2 O / C2 H2 ðgÞ þ CaðOHÞ2

(3)

1.1.3. Sesquicarbides Sesquicarbides possess anion C43 , which yields methylacetylene (propyne), CH3CCH, and hydroxide on hydrolysis: Mg2 C3 þ H2 O / CH3 C ¼ CH þ MgðOHÞ2

(4)

1.2. Covalent Carbides There are only two elements (Si and B) that are most similar to carbon in size and electronegativity and therefore form carbides with a high degree of covalent bonding (SiC and B4C). The fundamental structural unit in SiC is a covalently bonded primary co-ordination tetrahedron (either SiC4 or CSi4) and in B4C is boron icosahedrons (B12) bridged by carbon atoms.

1.3. Interstitial Carbides (IC) Interstitial carbides are formed from relatively large transition metals (metal atom radius is greater than

2. CARBIDES IN NATURE 2.1. Carbides and Natural Occurrence It is thought that the total amount of carbon in the atmosphere, oceans, and other near-surface reservoirs is negligible compared to that stored in the Earth’s mantle (highly viscous layer between the crust and outer core of the Earth) [1,2]. However, the mode of carbon storage in the mantle is unknown. The observation of micro-bubbles on dislocations in minerals from mantle xenoliths (an individual foreign crystal included within an igneous body) indicates that carbon could be soluble in silicates at high pressure [3,4]. Carbon solubility in silicates is crucial for understanding the dynamics of the carbon exchange between the mantle and near-surface reservoirs. Even the solubility of the order of a few hundred parts per million (ppm) by weight would be sufficient to incorporate all carbon in the upper mantle [5]. Since carbon is abundant in the Earth and, therefore, carbides’ occurrence in nature is not surprising. The first natural occurrence of carbides was mineral moissanite (SiC), which was found in 1893 as a small component of the Canyon Diablo meteorite in Arizona by Dr. Ferdinand Henri Moisson [6], after whom the material was named in 1905. Until the 1950’s, no other source, apart from meteorites, had been encountered. Later, moissanite was found as

Chapter | 3.1

Silicon Carbide and Other Carbides: From Stars to the Advanced Ceramics

227

TABLE 1 Some of Natural Transition Element (Fe, Cr, Co, Ni) Carbides [21] Mineral formula

Cohenite (Fe,Ni,Co)3C

Isovite (Cr,Fe)23C6

Haxonite (Fe,Ni)23C6

Tongbaite Cr3C2

Yarlongite (Cr,Fe,Ni)9C4

Chemical Composition C

6.56

5.77

5.70

13.34

9.22

Fe

54.92

26.82

89.50

e

40.60

Ni

26.86

e

4.91

e

8.54

Cr

e

67.41

e

86.66

e

Co

9.66

e

0.18

e

e

S

100

100

100.29

100

99.74

inclusion in kimberlite rock and lamproites that are formed deep within the mantle from a diamond mine in Yakutia [7e10], Shandong Province, China [11], Green River Formation in Wyoming [12], West Kimberly Province, western Australia [13], Monastery, South Africa [14], Sloan, USA kimberlites [15], and Argyle, Australia [16]. Discoveries have shown that moissanite occurs naturally as inclusions in diamonds, xenoliths, and ultramafic rocks such as kimberlite and lamproite [17,18]. In addition, the other SiC polymorphs a- and b-SiC are found in kimberlite from Fuxian, China [19]. The discovery of a cluster of SiC entrapped in diamond indicates diamondeSiC paragenesis. As SiC can be synthesized only at high temperatures, and because of its association with diamond-bearing kimberlite, a source of region of Earth’s mantle is inferred. Thus, SiC is probably generated in a deep Earth crust and its exhumation in many rocks comes later with geological activities [20].

SiC occurs naturally and that is a widespread, albeit rare, phase in diamond-bearing rocks. Recently, several natural carbides were found in Luobusha ophiolite in southern Tibet [21]. Most of them are carbides of transition elements (Table 1). From metallurgy it has been shown that WeCoeC phases could dissolve a substantial amount of metals such as V, Cr, and Ta, which are known to positively influence the microstructure of hard metals [22]. This is the nature approach. The minerals listed in Table 2 show the high degree of their solubility. Thus, carbide minerals are more solid solutions than individual carbides. It is considered that transition carbides are xenocrysts from the mantle, transported to shallow depths by a rising plume and then captured by the melts from which the chromite magma is crystallized [23]. The other evidence of the carbide minerals (Table 2) is discussed in the following.

TABLE 2 Natural Carbides Khamrabaevite: (Ti,V,Fe)C; (Bowman, 1961 [24], Novgorodova et. al., 1984 [25]) Niobocarbide: NbTaC; (Jedwab, 1990 [26], Novgorodova et al., 1997 [27]) Tantalcarbide: TaC; (Novgorodova and Generalov, 1997 [27]) Jedwabite: Fe7(Ta,Nb)C3; (Novgorodova et al., 1997 [28]) Qusongite: WC; (Zhang et al. (1989) [29]) Isovite: (Cr,Fe)23C6; (Generalov et al., 1998 [30]) Tongbaite: Cr3C2; (Zheng et al., 2004 [31]) Yarlongite: (Cr,Fe,Ni)9C4; (Nicheng et al., (2005) [21], 2008 [32]) Cohenite: Fe3C; (Sharp, (1966) [33], Stachel et al., (1998) [34], Goodrich and Berkley, (1998) [35], Jacob et al., (2004) [36])

2.2. Stellar Sources of Carbon and Carbides While rare on the Earth, carbides are common in space, especially silicon carbide. Carbides have been found as stardust around carbon-rich stars [37]. Their evidence comes also from pristine (unalterated) meteorites. They are isotopically different from the fine-grained matrix of primitive meteorites, such as hondrites, whose differences from the surrounding meteorite suggest that they are older than the solar system and usually called presolar grains [38]. Crystallinity in these presolar compounds ranges from that of micrometer-sized silicon carbide (SiC) crystals down to that of diamond and unlayered graphene crystals with fewer than 100 atoms. These grains were probably

228

formed in supernovae or the stellar outflows of red giant stars, and then incorporated into the molecular cloud from which the solar nebula separated to form our solar system. Presolar grains identified so far consist of refractory minerals, which survived the collapse of the solar nebula, and the subsequent formation of planetesimals (solid objects form out of dust grains that collide and stick to form larger bodies). There are different types of presolar grains. The most abundant is silicon carbide, which usually appears in submicrometer- to micrometer-sized grains. Presolar SiC occurs as a single-polytype grain or polytype intergrowths. The atomic structures observed contain the two lowest order polytypes: hexagonal 2H and cubic 3C with varying degrees of stacking fault disorder as well as 1-dimensionally disordered SiC grains. In comparison, terrestrial laboratory-synthesized SiC is known to form over two hundred different polytypes. This limited polytype distribution in presolar SiC can be explained by condensation of SiC at a relatively low total pressure in circumstellar shells [39]. Presolar grains consisting of the following carbide minerals have so far been identified: SiC, TiC, RuC, Fe3C, ZrC, and MoC [40e42]. Crystals of Ti-, Zr-, and Mobearing carbides are found within the spherules which served as nucleation centers for the graphite growth. These carbides have remained cloaked in a protective covering of graphite since their formation and therefore provide direct information on the chemical and physical properties of carbon-bearing circumstellar condensate [40]. Energy dispersive X-ray spectroscopy (EDS) analysis of the crystals shows that they are generally carbides of the refractory elements Ti, Zr, and Mo, although some crystals rich in Ni, Fe, and Ru were also detected. The most abundant element is Ti, which appears mostly as pure Ti carbide. Although the variability in element ratio points to a significant degree of solid solution carbides, which is in good agreement with numerous papers from the materials science studies. The atomic ratios relative to the dominant element Ti are enhanced by a factor of 10 to several thousands compared with material from solar origin, which clearly depicted their presolar origins as stellar condensates. Enrichments of Zr and Mo by a factor of 10e100 are predicted to occur in the envelopes of an AGB (asymptotic giant branch stars appear as a red giant star of low or intermediate mass in a late phase of stellar evolution with a central and inert core of carbon and oxygen and a shell where helium is undergoing fusion to form carbonehelium burning, another shell where hydrogen is undergoing fusion forming heliumehydrogen burning) star through s-process nucleosynthesis in their He-burning shells. Such enrichments are astronomically observed in carbon stars that have s-process enrichments of Zr by a factor of 10e100 relative to Ti [43].

Handbook of Advanced Ceramics

Equilibrium thermodynamics give us a possibility to infer the relative condensation sequence of the various carbides observed in the meteorites. The types of minerals condensing from a gas with solar abundances of all the elements depend critically on the C/O ratio. In the case of C/O < 1, the formed phases are oxides and silicates, and opposite, when C/O > 1, most of oxygen is tied up in CO, allowing graphite and carbides to condense. The studies that explored the parameter space of pressure, temperature, and C/O ratio show the order of condensation: ZrC condensed first, followed by MoC or Mo2C and TiC [44]. This is consistent with the relative condensation temperatures of these carbides, i.e., ZrC, MoC, and TiC. Abundant carbideemagnetite assemblages occur in the matrix of chondrite-type meteorites. Carbides, cohenite (Fe,Ni)3C, and haxonite (Fe,Ni)23C6 depict compositional variations between different meteorites and appreciable ranges within meteorites. Metal associated with carbides and magnetite consists of a high amount of Ni and Co. Textural observations indicate that carbideemagnetite assemblages formed by replacement of metal-sulfide nodules. The high Co contents of residual kamacite (natural alloys of iron and nickel) in association with carbides indicate that Co is not incorporated into carbides (i.e., Fe/ Co is much higher in the carbides than in kamacite). In Fe carbides rich with Ni, Krut et al. [45] suggested that carbideemagnetite assemblages are formed by hydrothermal alteration of metallic Fe in metal-triolite (iron sulfide, FeS) nodules by a CeOeH-bearing fluid on their parent bodies. This alteration resulted in the carbodization of FeeNi metal by CO gas: 15FeðsÞ þ 4COðgÞ ¼ 4Fe3 CðsÞ þ Fe3 O4 ðsÞ

(5)

3FeðsÞ þ 2COðgÞ ¼ Fe3 CðsÞ þ CO2 ðgÞ

(6)

while the magnetite is formed by the oxidation by H2O gas: 3FeðsÞ þ 4H2 OðgÞ ¼ Fe3 O4 ðsÞ þ 4H2

(7)

The CeOeH bearing fluids are derived from ice, adsorbed gases, or hydrated minerals, while CO is the result of the reaction of carbon compounds (e.g., hydrocarbons) with water vapor or magnetite. Iron carbide (cohenite) has been reported as inclusions in diamond, either isolated [34] or intergrown with graphite [46].

3. TRANSITION METAL CARBIDES 3.1. Parameters Influencing the Carbide Structure and Properties The important parameters that can explain the variety and similarity of carbides are given in Table 3. Although the Hf carbide is not found in nature, it is considered that Hf is

Chapter | 3.1

Silicon Carbide and Other Carbides: From Stars to the Advanced Ceramics

229

TABLE 3 Atomic Radius and Electronegativity of Elements That Form Natural Carbides Element

Atomic radius (nm)

Carbon/metal atomic ratio

Electronegativity

Type of carbide*

Carbon

0.078

e

2.5

e

Silicon

0.117

0.666

1.8

CC

Titanium

0.1467

0.526

1.5

IC

Zirconium

0.1597

0.483

1.4

IC

Hafnium

0.1585

0.486

1.3

IC

Vanadium

0.1338

0.576

1.6

IC

Niobium

0.1456

0.530

1.6

IC

Tantalum

0.1457

0.529

1.5

IC

Chromium

0.1267

0.609

1.6

IM

Molybdenum

0.1386

0.556

1.8

IC

Tungsten

0.1394

0.553

1.7

IC

Manganese

0.1261

0.618

1.5

IM

Iron

0.1260

0.619

1.8

IM

Cobalt

0.1252

0.623

1.8

IM

Nickel

0.1244

0.627

1.8

IM

*CC: Covalent carbide, IC: Interstitial carbide, IM: Intermediate carbide.

always present in Zr compounds. In addition, it has a potential for structural components that are able to serve in extreme high-temperature conditions. The electronegativity considerations suggest simple metal(Me) / C electron donation. Electropositive alkali and alkaline-earth metals form ionic-type compounds with negatively charged nonmetals. These carbides are characterized by lack of electrical conductivity, transparency, and hydrolysis. On the other hand, these properties are not found in the alloys of carbon atoms with transition metals. The interstitial carbides are excellent electrical conductors and hydrolytically stable. Thus, simple electronegativity charge transfer concepts do not work when carbon atoms are interstitially dissolved in metal lattices. It is pointed out that the amount and direction of charge transfer depend on the property considered [47]. When considering band occupation, electron transfer is from C / Me, but this reverses when considering integrated charges around atoms. However, some trends exist. For example, the difference in electronegativity between carbon and the elements forming carbide is more pronounced for the elements from the groups IV, V, and VI of the periodic table. In general, a large difference in electronegativity leads to interstitial carbides (ICs), while a smaller difference favors covalent carbides (CCs) or intermediate carbides (IMs).

In interstitial carbides, carbon atoms take the interstitial space within the framework of the metal atoms and the spacing between carbon atoms is too large for atomic interaction; there is no carbon-to-carbon bond resulting from overlapping electron shells. The overall bonding energy is the sum of metal-to-metal and metalto-carbon bonds. Pierson [48] reviewed the atomic bonding of interstitial carbides and he gave some of the major differences between the host metal and the carbide: 1. Carbides are hard and brittle, while the host metals are malleable and much softer. 2. Carbides have a high bond strength, which exceeds by far the strength characteristics of the host metal. 3. The melting point of carbides is generally much higher than that of the parent metal. 4. A switch to a more stable structure occurs in every case when going from the metal to the carbide. The best qualitative view of the bond’s energy can be obtained from the comparison of the melting points of carbides. The higher melting point of carbide means the higher MeeC bond strength. The carbides from group IV have much higher melting points than their host metals. The difference is less pronounced in group V, while the metals have higher

230

Handbook of Advanced Ceramics

3.2. Crystallochemistry

FIGURE 1 Melting points of interstitial carbides and their host metals of the groups IVeVI.

melting points than the corresponding carbides in group VI. A crude interpretation of this trend is that the maxima are associated with the formation of a half-filled d-shell and, to achieve this, carbon donates one electron to the metal. According to the melting points of interstitial carbides and their host metals [49], Storms [50] has established the following order (Figure 1):

Group IV

MeeMe bond weak

MeeC bond very strong

Group V

MeeMe bond strong

Me-C bond strong

Group VI

MeeMe bond very strong

MeeC bond weak

An extensive compilation of the crystal structures of transition metal carbides is found in many handbooks [48,51e55]. On the other hand, the atomic-radii ratio of carbon atom and host metals determines the suitability of the elemental components to form interstitial structures.

The crystallochemistry of the interstitial carbides is similar to that of pure metals. They often possess a simple crystal structure, with the transition metal atoms forming lattices of face-centered cubic (fcc, or, the other term, cubic closepacked, ccp), hexagonal close-packed (hcp), simple hexagonal (hex), or body-centered cubic (bcc) structures. Their crystal structure can also be explained by the EngeleBrewer valence bond theory [56]. According to this theory, the crystal structure is determined by the number of outer-shell sp electrons per atom (e/a). These are itinerant and control the range order. For up to 1.5 e/a, the bcc structure is obtained. A value of 1.7e2.1 e/a produces the hcp structure, and for 2.5e3 e/a attains the ccp structure. The carbon atoms enter into the interstitial sites between metal atoms showing the origin of the name for this class of materials. The available sites can be octahedral in fcc and hcp, or trigonal prismatic in hex lattices. The carbon atoms in these alloys occupy the octahedral interstitial sites. Thus, the crystal structure is determined by the geometric factor, which is based on an empirical rule given by Ha¨gg. The rule states that simple structures are formed when the ratio of the radii of nonmetal to metal (rn/rm) is less than 0.59. This criterion fits the metals of the early transition elements (group IVeVI), with chromium being the exception (Table 1). Carbon/chromium atomic ratio is 0.61 and, strictly speaking, belongs to the intermediate class of carbides. However, the physical and chemical properties of chromium carbide are similar to the other transition carbides. The structure of transition carbides can be described by means of a model for the simple ionic structure by two coordination numbers e one for each type of ion. They possess the rock salt structure (NaCl) or calcium fluoride structure (CaF2). In the NaCl structure (Figure 2), the number of cations and anions are equal, and both the coordination numbers are six, so the structure is (6,6).

FIGURE 2 Crystal structure of the interstitial carbides.

Chapter | 3.1

Silicon Carbide and Other Carbides: From Stars to the Advanced Ceramics

231

FIGURE 3 Example of phase diagram of the Group IV transition metalecarbon systems. Titaniumecarbon system.

The other structure type (CaF2) is (8,4), meaning that eight anion neighbors surround each cation, and each anion by four cations. Their elementary cell differs only due to either (i) atom radius of metal or (ii) stoichiometry of carbon atom in the structure. Many binary transition metal carbides, especially carbides from group IV elements, exist over a broad range of compositions with an upper limit of the carbon to metal ratio near 1. The chemical formula is usually written as MeC1ex, where Me represents a transition metal, C is carbon, and x is the C/Me ratio. Ideal stoichiometry does not exist in these compounds. The values of x range from approximately 0.5 to 0.97. Thus, these materials are carbon-deficient, nonstoichiometric compounds with a wide range of vacancy concentrations [57]. Thus, interstitial carbides can be considered to be essentially nonstoichiometric materials in which stoichiometry is rarely complete. This is particularly often in the case of the TieC system. Figure 3 shows an example of the group IV of transition metalecarbon systems. They are generally very similar to each other concerning their properties: extensive homogeneity ranges and high congruent melting point. In addition, they possess NaCl-type cubic basis crystal structure. Their structures thus consist of identically stacked, closely packed {111} planes of the metal atoms alternating with {111} planes containing the C atoms. Monocarbides from group V have the same crystal

structure, as well as comparable melting points (Figure 4). However, they contain the structurally identical subcarbide phases Me2C and Me4C3 x. Also, the phases Me2C have low- and high-temperature modifications, which differ only by arrangement of the carbon atoms on the interstitial positions [58]. The compositions of the phase Me4C3ex deviate from the ideal crystallochemical C/Me ratio 3/4 [51]. The other carbides with different carbon content such as Nb3C2, Ta3C2, Nb4C3, Ta4C3, V6C5, and V8C7 have been observed [59]. Deviation of stoichiometry is the largest for V4C3ex in which carbon content is below 40 at%. The group VI of transition metalecarbon system, the stoichiometric monocarbides WC and MoC are hexagonal. Also, cubic monocarbides MeC1ex in the MoeC and WeC system exist but at high temperatures and with stoichiometric compositions. In addition, the other phases with different stoichiometries are observed (W2C, Mo2C, and Mo3C2). The CreC system differs from the other transition metalecarbon systems (Figure 5). This system is characterized by the absence of a monocarbide phase. The other phases are Cr3C2, Cr7C3, and Cr23C6. In the phase Cr23C6, each carbon has 8 Cr neighbors in the form of a quadratic antiprism and this structure is basically different from all other early transition metal carbides. Intermediate transition carbides in the MneC and FeeC have similarities with the CreC system with respect to occurring phases (Figure 6). The phase Mn3C is isotypic

232

Handbook of Advanced Ceramics

FIGURE 4 Example of phase diagram of the Group V transition metalecarbon systems. Niobiumecarbon system.

FIGURE 5 Example of phase diagram of the Group VI transition metalecarbon systems. Chromiumecarbon system.

Chapter | 3.1

Silicon Carbide and Other Carbides: From Stars to the Advanced Ceramics

233

FIGURE 6 Example of phase diagram of intermediate transition metalecarbon systems. Manganeseecarbon system.

with Fe3C; also, Mn23C6 and Mn7C3 are isotypic with the chromium carbide phases of the same compositions. In general, these phases are not stable. Their phase diagrams are of eutectic type. They are obtained in rapidly quenched NieC and CoeC alloys.

3.3. Properties The carbides of the transition metals have extremely high melting points and sometimes they are collectively referred to as refractory carbides. Also, these carbides are extremely hard, retaining high hardness at very high temperature and therefore finding industrial use in cutting tools and wear-resistant parts. In addition, they possess good thermal conductivity as well as good thermal shock resistance. Although the FeeNieCoeC systems are not high-temperature systems as carbides of the group IVeVI, they are the more important systems in the metallurgy of steels. As a solid solution they can persist for a very long time at temperatures up to 700  C. Their decomposition is extremely slow. WC is also used for fabrication of as cemented carbide tools for cutting steel. The addition of other carbides such as TiC, TaC, and NbC improves the working temperature by increasing the oxidation resistance of the tool. Sintered carbides of Co, Mo, and W are known as superalloys, which are used in high-temperature applications such as rocket nozzles and jet engine parts.

4. COVALENT CARBIDES As mentioned in Section 1, there are only two covalent carbides: boron carbide (B4C) and silicon carbide (SiC). However, a low abundance of boron in both the solar system and the Earth’s crust is a reason why the B4C is not found naturally. On the other hand, occurrence of SiC is widespread. Silicon and carbon atoms are located one above the other in the periodic table having similar electronic structure (C e 2s2 2p2, Si e 3s2 3p2); both form four sp3 hybridized orbital with a tetrahedral symmetry and the angle between the edges form 109 280 identical to the structure of diamond (Figure 7). The difference in the electronegativity of atoms C and Si (2.5 e 1.8 ¼ 0.7) indicates that the ionic character of the relationship is 10%. The bond strength of SiC is about 300 kJ/mol, in diamond it is 356 kJ/mol. Silicon atom is only slightly larger than carbon atom. Some general characteristics of SiC are: 1. 2. 3. 4. 5. 6.

It is a nonmetallic compound, its elemental constituents have low atomic weight, it has a low density, its bonding is essentially covalent, it is an extremely hard and very strong material, it fully meets the refractory criteria (high melting point, and thermal and chemical stability), and 7. it has semiconductor properties.

234

Handbook of Advanced Ceramics

FIGURE 7 The tetrahedral bonding of a carbon atom to four nearest silicon neighbors.

Silicon carbide is the only stable compound in the siliconecarbon system [60]. It is binary phase with a composition of 70.05 wt% Si and 29.05 wt% C (Figure 8). SiC does not have a melting point, it readily decomposes at 2545  40  C into carbon and a silicon-rich melt in a closed system at a total pressure of 0.1 MPa. This is the highest

temperature at which SiC is formed. In the open system, SiC starts to decompose at approximately 2300  C with generated species such as Si, SiC2, and Si2C. Also, SiC and Si degenerate into an eutectic at 1404  C and 0.02at% C. The solubility of carbon in liquid silicon is about 13at% C at the peritectic temperature. Silicon carbide exhibits a pronounced tendency to crystallize in a multitude of different modifications named polytypes [61]. All polytypes consist of closepacked layers of carbon (C) and silicon (Si) atoms, where the C atoms are situated above the centers of triangles of Si atoms and underneath the Si atoms belonging to the next layer (Figure 9). The distance between the neighboring silicon or carbon atoms (a) is approximately 0.307 nm for all polytypes. The carbon atom is positioned at the center of mass of the tetrahedron outlined by the four neighboring Si atoms so that the distance between the C atoms to each of the Si atoms is the same. Geometrical considerations give that this distance, CeSi is approximately equal to 0.189 nm and the distance between the carbon atoms is approximately 0.307 nm [62]. The distance between two silicon planes is approximately 0.252 nm. The short-range structure of SiC consists of tetrahedra connected together in a two-dimensional space with lattice spaces arbitrarily occupied by Si or C atoms resulting in a planar SiC4 or CSi4 configuration. These planes are arranged in a parallel or antiparallel manner for the longrange structure of SiC.

FIGURE 8 Equilibrium phase diagram of the SieC system.

Chapter | 3.1

Silicon Carbide and Other Carbides: From Stars to the Advanced Ceramics

235

FIGURE 9 Ramseld diagram showing layer alternation of the most abundant SiC polytypes. Carbon atom (C); silicon atom (O).

Usually, the ABC notation, known as Ramsdel notation, is used as a more descriptive of the position of atomic layers (Figure 9). For example, the number 3 in 3CeSiC refers to the three-bilayer periodicity of the stacking ABC and the letter C denotes the cubic symmetry of the crystal. The stacking sequence ABAB shows the wurtzite crystal structure for 2HeSiC reflecting its two-bilayer stacking periodicity and hexagonal symmetry. This periodicity doubles and triples in 4H and 6H polytypes. All the rhombohedral polytypes are labeled by R (15ReSiC and 21ReSiC) (Table 4). The parallel arrangement makes an ABCABC sequence of either Si or C layers, which has the cubic, zinc blende structure. According to a definition, the zinc blende structure is an fcc package with half of tetrahedral sites filled. Since SiC is a covalent compound it can be said that the SiC structure consists of two identical interpenetrating close packings, one of Si and the other of C. This structure is denoted as b-SiC. The antiparallel arrangement makes an ABAB sequence of layers, which has a hexagonal, wurtzite structure. It is denoted as a-SiC and it has several polytypes. The difference between the polytypes is the stacking order between the succeeding planes of the tetrahedra. As far as mechanical properties are concerned, the polytypes do not play an important role and from this point of view only a-SiC and b-SiC are important. The typical polytypes are shown in Figure 10. The cubic polytype 3C-SiC is assumed to be more stable than the hexagonal polytypes up to 2100  C [65]. The 2H polytype is the unstable one. In presolar grains only these two SiC polytypes occurred. This is explained by low pressure in the circumstellar shells during SiC formation [39]. Also, it is noticed that under some conditions 6HeSiC can transform to 3CeSiC. This does not commonly occur

at high temperatures. During the hot pressing of 6HeSiC and TiN mixture in a flowing Ar atmosphere, all the hexagonal SiC is converted to cubic SiC above 2200  C, although 3C polytype is not stable at this temperature. It is found that N has an essential influence on 6HeSiC transformation [66].

5. SYNTHESIS Natural sources of transition metal carbides as well as silicon carbide are extremely rare and are a mineralogical curiosity that has no significance as raw materials. All carbide-based ceramics are derived from the synthetic materials, exclusively. The majority of carbide ceramics is made from powders and therefore depends, to a greater extent, on the quality of the starting powders. All the carbides can be prepared by heating almost any source of corresponding metal and carbon at high temperatures [67]. Several different synthetic routes for obtaining carbide powders are available; some of them are mentioned below: 1. solid combustion synthesis by direct combination of the elements at elevated temperatures, 2. carbothermal reduction reaction of metal oxide compounds, 3. vapor phase synthesis, 4. pyrolyzes of metal-organic compounds, 5. solegel route, 6. laser-induced reactions, and 7. plasmachemical synthesis. The synthesis of pure carbide powder is one of the important factors for obtaining dense ceramics. Although the quality and performance are important, the cost of mass

236

Handbook of Advanced Ceramics

TABLE 4 Some SiC Polytypes and their Lattice Parameters [63,64] Polytype

No. of space group

Lattice constant a0 (nm)

Lattice constant c0 (nm)

3C

216

0.43596

0.43596

2H

186

0.30730

0.50480

4H

186

0.30730

1.0053

6H

186

0.30730

1.5110

8H

186

0.30730

2.0147

19 H

156

0.30730

4.7849

21 H

156

0.30730

5.2870

27 H

156

0.30730

6.7996

36 H

156

0.30730

9.0650

15 R

160

0.30730

3.7700

21 R

160

0.30730

5.2890

24 R

160

0.30730

6.0490

27 R

160

0.30730

6.7996

33 R

160

0.30730

8.3110

45 R

160

0.30730

11.333

51 R

160

0.30730

12.843

105 R

160

0.30730

26.439

141 R

160

0.30730

35.505

189 R

160

0.30730

47.628

393 R

160

0.30730

98.760

production is the key factor in the commercialization of carbide ceramics. With this point of view, many of the above-mentioned fabrication methods would never meet the commercialization.

5.1. Solid Combustion Synthesis In the solid combustion synthesis of carbides, the mixed powders are compacted beforehand to increase an intimate contact between the individual metal and carbon particles. The reactant mixture is degassed and ignited in a vacuum or in an inert atmosphere [68]. Some examples of this route are: Si þ C / SiC; T ¼ 1400 C;

(8)



W þ C / WC; T ¼ 1500 C; Ta þ C / TaC; T ¼ 2200

(9) 

2400 C:

(10)

Combustion synthesis can be conducted in a self-propagating mode (SHS, self-propagating high-temperature

synthesis). Compared with the conventional ceramic processing, the main advantages of SHS are: 1. it needs less energy than conventional methods, 2. it takes a short reaction time resulting in less processing cost and time saving, 3. shaped and unshaped products can be obtained in a single step, and 4. expensive and complicated operating equipments are not necessary. The SHS reactions can be characterized by the adiabatic combustion temperature, Tad [69]. The combustion temperature can be calculated by assuming that the enthalpy of the reaction heats up the product and no loss of energy occurs to the surrounding environment. As a rule of thumb, if Tad is less than 1200  C, combustion does not occur. In case Tad is higher than 2200  C, self-propagating combustion occurs. In the range between 1200  C and 2200  C, a combustion wave cannot propagate but it is possible by special techniques such as preheating of the

Chapter | 3.1

Silicon Carbide and Other Carbides: From Stars to the Advanced Ceramics

237

FIGURE 10 Crystal structure of typical SiC polytypes.

reactants, or by introducing mechanical energy in the system. For example, the reaction between Si and C that results in SiC is not self-sustaining because of Tad (1527  C) and the ignition temperature is 1300  C. In the case of Ti and C, which give TiC, reaction is self-sustaining (Tad ¼ 2937  C) and ignition temperature is 1027  C. Also, the SHS reaction can be accomplished by introducing a mechanical energy in the system by ball

milling of metal powders and carbon. During the milling process, powder particles are repeatedly fragmented and cold-welded, which results in the intimate mixing of the reactant particles on the nano scale. The milling process also leads to the reduction in the crystallite size and the accumulation of defect in the powder particles, which introduces an additional energy to the reactant system and thus effectively lowers the activation barrier for

238

the reaction. For some highly exothermic reaction systems, spontaneous compound formation can occur by mechanically induced self-propagating reactions, the mechanism of which is analogous to that of the thermally ignited self-propagating high-temperature synthesis. Mechanochemical synthesis is usually carried out under room temperature conditions and therefore it is particularly attractive for the synthesis of high-melting-point carbides [70]. For example, TiC as well as b-SiC can be synthesized by a high-energy ball milling at room temperature without thermal treatment [71]. In general, combustion synthesis is difficult to control due to the high reaction rate. Some improvement of reaction control is possible by the addition of diluents, which do not take part in the reaction, but increase the thermal mass of the system and lower Tad. The other way is to increase the particle size of the reactants resulting in the decrease of the combustion waves.

5.2. Carbothermal Reduction Commercially, most of the carbides are obtained by the Acheson process, which is also known as the carbothermal reduction reaction. The production process can be carried out in a variety of ways in various types of reactors (electric arc furnaces, rotary tube reactors, moving bad furnaces, fluidized bed reactors, etc.). However, the basic design of the first furnace (Acheson) remained unchanged, despite better efficiency. Generally, graphite electrodes connected to a graphite core are laid in a mixture of carbon and metal oxide. When an electric current is passed through the graphite core, the reaction takes place resulting in the formation of designated carbide. Also, it is the most promising candidate for obtaining a large variety for all types of nonoxide products with important technical uses. This reaction involves reduction of oxygenated materials, such as silica (SiO2) usually by mixing with a reducing agent (carbon) in excess at a temperature higher than 1600  C for several hours under an inert atmosphere. As carbon sources can be used graphite, charcoal, carbon black, pyrolyzed organic polymers, however, the most common carbon source is the final residue from the crude oil refining known as the petroleum coke. This procedure offers the possibility of an economically attractive production route from naturally occurring materials that possess high specific surface area, high metal content, and low price [72]. The carboreduction of oxides is the most developed production method for SiC. Many authors have studied the formation of SiC powders from the raw materials, such as high-purity quartz sand [73]. Sepiolite, diatomaceous earth, and mounting leader have been used recently [74e76].

Handbook of Advanced Ceramics

During the thermal treatment of the silica/carbon mixture, Si or SiO is liberated in gaseous form, which further reacts with excess carbon to form SiC according to the following reaction: SiO2 þ 3C / SiC þ 2CO

(11)

The formation of the final product is more complex than the above equation. There are many intermediate stages, since the formation of SiC requires a series of solidesolid, solidegas, and gasegas reactions [77]. It is usually accepted that SiC forms as a result of four main subreactions (Eqns 12e15), which provide mass transport via vapor phases: SiO2 ðsÞ þ CðsÞ / SiOðgÞ þ COðgÞ

(12)

SiO2 þ COðgÞ / SiOðgÞ þ CO2 ðgÞ

(13)

2CðsÞ þ SiOðgÞ / SiCðsÞ þ COðgÞ

(14)

CðsÞ þ CO2 ðgÞ / 2COðgÞ

(15)

Silica initially reacts at the contact point with carbon particles in the solid-state reaction liberating CO and SiO gases (Eqn 12). The reaction between CO and remaining SiO2 results in the formation of additional SiO and CO2 gases (Eqn 13). SiC is formed by the reaction of gaseous SiO directly at the surface of the carbon particles (Eqn 14). The intermediate gaseous phase CO2 is re-converted to CO by Eqn 15. Typical examples for the carbothermal reduction and the corresponding lowest occurring temperatures at atmospheric pressure are given below [69]: TiO2 þ 3C/TiC þ 2CO; 1300 C 

2MoO3 þ 7C/Mo2 C þ 6CO; 500 C; WO3 þ 4C/WC þ WC þ 3CO; 700 C;

(16) (17) (18)

The conversion of TiO2 to TiC takes place via the intermediate phases such as Ti3O5, Ti2O3, and TiO. Also, the reduction of the other oxides of transition metals proceeds via the intermediates. Generally, carbides are prepared by the reduction of oxides with carbon; only Mo2C and WC are manufactured by the reaction of the metal powders with graphite or carbon black [60]. Thermite reaction is a special type of combustion synthesis involving the reduction of a metallic compound by strongly reducing metals, such as Mg and Al. The more desirable reducing agent is easily leachable with hydrochloric acid. Reaction proceeds in two steps. The first involves the reduction of an oxide to an element and, the second, the reduced element reacts with another element to form carbide. For example, reducing silica by Mg, followed

Chapter | 3.1

Silicon Carbide and Other Carbides: From Stars to the Advanced Ceramics

239

by a reaction between Si and C for obtaining SiC by thermite-type reaction [78]:

The vapor-phase thermal decomposition of some organometallic compounds can form SiC, for instance,

SiO2 þ 2Mg þ C / SiC þ 2MgO

CH3 SiCl3 / SiC þ 3HCl

(30)

ðCH3 Þ4 Si2 H2 / 2SiC þ 2CH4 þ 3H2

(31)

(19)

A metal oxide is formed as a by-product. For the synthesis of pure carbide powders, the oxide has to be chemically leached. Another example is obtaining chromium carbide by reducing Cr2O3 with Al in the presence of carbon: 3Cr2 O3 þ 6Al þ 4C / 2Cr2 C3 þ 3Al2 O3

(20)

This principle has a practical usage in the synthesis of WC powders from a naturally occurring tungsten mineral wolframite (FeWO4): FeWO4 þ CaC2 þ 2Al / WC þ Fe þ CaO þ Al2 O3 þ C (21) In principle, this technique is used when the cost of elemental powder is too high (Ta, Hf ), or when the direct reaction of the element does not generate sufficient amount of heat for the self-sustaining reaction.

5.3. Vapor-Phase Synthesis The vapor-phase synthesis of the carbides can be performed by the reaction of the metal or chlorides with carbon-containing gases such as methane or benzene:

5.4. Pyrolysis of Organic Compounds Organometallic polymer precursors offer unique opportunities for manufacturing of ceramic components using versatile plastic shaping techniques established in the engineering plastics industry such as casting, pressing, injection molding, or extrusion [79]. Pyrolysis of metalcontaining polymers has been effective in the preparation of ceramic materials at temperatures significantly lower than the conventionally sintered materials. However, only SiC has attracted significant attention for this method. The main reason is only a few compounds from the group IV and VI metals that are stable have a low carbon/metal ratio and do not contain oxygen. Also, the process suffered from the inability to scale up ceramic yields. In addition, there are still several drawbacks such as high cost and toxicity. The gas-phase decomposition of organosilanes or polycarbosilanes leads to the formation of nanocrystalline SiC [80]. Typically, gas-phase reaction of the SiH4 or SiCl4 with hydrocarbons such as CH4 results in the ultrafine b-SiC powders [81].

Ti þ CH4 / TiC þ 2H2

(22)

5.5. SoleGel Route

ZrCl4 þ CH4 / ZrC þ 4HCl

(23)

Solegel route has been developed by mixing the reactants in the liquid phase. This route mainly focuses on SiC. For example, ethylsilicate liquid and liquid phenolic resin as the sources of silica and carbon, respectively, and toluensulfonic acid as the catalyst were used [82]. The powders obtained after the carbothermal reduction of the gel were sinterable b-SiC, containing no residual SiO2 and carbon at a proper starting resin content. Similar solegel route has also been used to synthesize cubic SiC fibers with hydrochloric acid as the catalyst [83] and tetraethoxysilane (TEOS) and novolac phenolic resin as the starting materials, and oxalic acid and hexamethylenetetramine as the catalysts [84]. Also, the other carbides were synthesized by the solegel route. For example, ZrC and HfC powders were prepared using Zr-n-butoxide and Hf-isopropoxide and polyhyclric alcohol as carbon sources [85]. TaC is synthesized by the liquid precursor route starting from Ta-ethylate solution and activated carbon powders [86]. This method has become very attractive for manufacturing bioamorphous ceramics [87,88] due to an eco-friendly route for advanced ceramic materials which can be produced with renewable natural resources, such as

When chlorides are used, the reactions proceed in the gas phase above 600  C. Since the transition metal halogenides are very volatile, with this route it is possible to obtain nanopowders or to use it for chemical vapor deposition of carbide layers on the solid substrate. Also, gasegas reaction is developed to manufacture carbides, especially SiC: CH3 þ SiH3 / SiC þ 3H2

(24)

CH4 þ SiH4 / SiC þ 4H2

(25)

SiCl4 þ CH4 / SiC þ 4HCl

(26)

7SiCl4 þ C7 H8 þ 10H2 / 7SiC þ 28HCl

(27)

SiCl4 þ CCl4 þ 4H2 / SiC þ 8HCl

(28)

3SiH4 þ C3 H8 / 3SiC þ 10H2

(29)

The vapor-phase reactions offer the advantages of synthesizing high-purity nanopowders with a good control over size, shape, and crystal structure as well as easy control of the reaction rate. However, the method does not yield appropriate amount of product for commercialization.

240

wood, with tailorable properties and behavior like ceramics manufactured by conventional approaches [89]. Although most research is focused on biotemplating SiC ceramics [90e92], the transition metal carbides have been successfully manufactured by this process [93].

5.6. Others The gas-phase process can be also achieved in plasma or laser-induced reaction. The result of this process is an amorphous or only partially crystallized powder with a high specific surface area. The gas-phase processes and the solegel process offer very high purity powders with high sinterability. However, these powders have a very poor shaping behavior and a high cost in comparison to the traditional routes. The above-mentioned classification of the synthetic routes is very conventional because, in most cases, the preparation method is based on more than one basic principle [94e96]. The synthesis of pure carbide powder is one of the important factors for obtaining dense ceramics. Although the quality and performance are important, the cost of mass production should be the key factor in the commercialization of carbide ceramics.

6. EXTREME ENVIRONMENT APPLICATION The stability that chemical compounds ordinarily have is eventually limited when exposed to extreme environmental conditions such as heat, radiation, pressure, etc. Recently, there is request for many engineering materials that can survive harsh conditions such as temperatures higher than 3000  C, or conditions in nuclear reactors. The current increasing interest in hypersonic vehicles with new propulsion system components, and wing leading edges and nose tips needs ultra-high-temperature materials. There are more than 300 compounds with melting temperatures over 2000  C, including refractory metals (Hf, Nb, Re, Ir, Ta, and W), oxides (HfO2, ZrO2, UO2, and ThO2), and a variety of transition metal borides, carbides, and nitrides. However, the list of materials with melting temperatures above 3000  C is limited to less than 15 compounds, all nonoxides. Among them, the most numerous compounds are the carbides (Table 5). However, in the real engineering application, melting temperature is only one of the many properties used in the material selection process [97], and other properties include high-temperature strength, thermal conductivity, thermal expansion coefficient, density, and cost. However, the oxidation behavior is the second-ranking property associated with the material selection for the extreme

Handbook of Advanced Ceramics

TABLE 5 Compound With Melting Temperature Above 3000  C [99] Compound Melting temperature ( C) Existing in nature TaC

3980

yes

HfC

3928

?

ZrC

3420

yes

NbC

3600

yes

TiC

3067

yes

TaN

3097

e

HfN

3387

e

TaB2

>3000

e

HB2

3380

e

ZrB2

3220

e

BN

3000

e

environmental situation. The main reason is that the materials applied will involve exposure to oxidizing fuels or aeroheating [98]. All nonoxide compounds will undergo oxidation and undergo a significant weakening of the mechanical properties. There are some oxides with high melting point; however, their poor thermal shock resistance, because of the low thermal conductivity as well as high thermal expansion, eliminates them in extreme environment applications. On the other hand, SiC exhibits excellent oxidation resistance up to 1700  C because of the formation of a layer of silica glass that inhibits oxygen diffusion to the parent material [99,100]. This is the case of passive oxidation according to the reaction (Eqn 32). Longterm usage leads to the evolution of a protective oxide layer mainly consisting of silica [101]: 2SiC þ 3O2 / 2SiO2 þ 2CO

(32)

However, active oxidation, which causes corrosion of the substrate materials, does not form a protective layer due to the formation and volatilization of the silicon monoxide [102]: SiC þ O2 / SiO þ CO

(33)

Under reduced pressure environments, degradation of SiC and decomposition of already formed silica layer limited the application of SiC as an ultra-high temperature ceramic material. The maximum usable temperature of SiC is limited to 1700  C, which cannot satisfy the new engineering requests. Transition-metal carbides from the group IV and V elements have superior melting points; however, not all of

Chapter | 3.1

Silicon Carbide and Other Carbides: From Stars to the Advanced Ceramics

them can serve in harsh environments. For example, TiC and NbC form oxides TiO2 and Nb2O3, which possess low melting points, 1840 and 1485  C, respectively. The potential candidates are TaC and HfC compounds. They possess the highest melting temperatures of any compounds (3980 and 3928  C, respectively). In addition, they have very high hardness and elastic modulus. The main difference between HfC and TaC is the melting temperature of the oxide formed. HfO2 has a melting point of 2758  C, while Ta2O5 has a much lower melting point, 1872  C. HfC is a better choice and it is the reason why HfC is studied in more detail. It was shown that HfC has excellent oxidation behavior at temperatures above 1800  C due to the densified oxide formed. However, at temperatures below 1500  C, formed oxide grains are not able to sinter, causing them to spontaneously spall of the parent material.

7. IMPORTANCE OF NATURAL AND SYNTHETIC CARBIDES Both terrestrial and solar carbides do not have technological significance. However, the importance is reflected in the fact that carbides may be a link with the theory of the Earth’s carbon cycle as well as the theory of galactic chemical evolution. For example, it is found that SiC is richer in 12C than relative to the standard. Isotope fractionation might have occurred through an isotope exchange reaction in a common carbon reservoir [19]. Mineral SiC may thus ultimately provide information on carbon cycling in the Earth’s mantle. Since SiC can be formed in highly reduced conditions, its occurrence is very important for understanding the existence of regions in the deep Earth, influence of the oxidation state on the composition of vapor, the geochemistry of the chalcophile and siderophile elements, and the manner of partial melting of the mantle in highly reduced environments [103,104]. On the other hand, carbides as more resistant parts in meteorite are essentially a frozen piece of a single star that ended their life before the formation of the solar system [105]. Therefore, carbides among other resistive minerals (alumina, diamond, spinels, etc.) provide a unique insight into the evolution of and nucleosynthesis within these environments. Some of the isotopic variations also reflect the chemical evolution of the galaxy and can be used to constrain corresponding models. Presolar grain microstructures provide information about the physical and chemical conditions of dust formation in stellar environments [106]. The detailed characterization of these grains can yield information on the pressure, temperature, and chemical composition in the stellar atmospheres and ejecta. Therefore, the microscopic grains reveal more detail about

241

the evolution of galaxy than by the astronomical observation with the currently possible best telescopes. Man-made carbides that have the same composition as natural carbides are considered to be the main focus of research because of the superior melting temperatures and formation of stable high-melting temperature oxides. The combination of their properties such as high hardness, good chemical stability, and strength at high temperatures makes these materials potential candidates for a variety of hightemperature structural applications such as plasma arc electrodes, cutting tools, furnace elements, hypersonic vehicles, and high-temperature shielding [98].

REFERENCES [1] Javoy M, Pineau F, Alle’gre CJ. Carbon geodynamic cycle. Nature 1982;300:171e3. [2] Jambon A. Earth Degassing and large-scale geochemical cycling of volatile elements. Rev Mineral 1994;30:479e417. [3] Green HV. A CO2-charged asthenosphere. Nature 1972;238:2e5. [4] Green HV, Radcliffe SV. Fluid precipitates in rocks from the Earth’s mantle. Geol Soc Am Bull 1975;86:846e52. [5] Keppler H, Wiedenbeck M, Shcheka SS. Carbon solubility in olivine and the mode of carbon storage in the Earth’s mantle. Nature 2003;424:414e6. [6] Moissan H. Nouvelles recherches sur la meteorite de Canon Diablo. Comptes rendus 1904;139:773e86. [7] Bobrievich AP, Bondarenko MN, Gnevushev MA, et al. Diamond deposites of Yakuta. State scientific technical printing office for literature related to geology and conversation of resources, Moscow (in Russian); 1959. [8] Marshintsev VK, Shchelchkova SG, Zolnikov GV, Voskresenskaya VB. New data on moissonite from the Yakutian kimberlites. Geologiya Geojzika Novosibirsk 1967;12:22e31 (in Russian). [9] Kaminskiy FV, Bukin VI, Potapov SV, et al. Occurrences of silicon carbide under natural conditions and their genetics significance. Akad Nauk SSSR Zzv Ser Geol 1968;6:57e66. [10] Marshintsev VK. Natural silicon carbide in Yakutian kimberlites. Min Mug 1990;12:17e26 (in Russian). [11] Guan-Zhi H. Kimberlites in China and their major components: A discussion on the Physico-chemical properties of the upper mantle. In: Komprobst J, editor. Kimberlites I: kimberlites and related Rocks. Elsevier; 1984. p. 181e94. [12] Bauer JF, Hrichova´ R. Natural heSilicon Carbide. American Mineralogist 1963;48:620e34. [13] Jaques AL, Lewis JD, Smith CB. The kimberlites and lamproites of western Australia. Bull Geol Survey Western Australia 1986;132. [14] Moore RO, Otter ML, Rickard RS, et al. The occurrence of moissanite and ferro-peticlase as inclusions in diamond. In 4th International Kimberlite Conference, Perth, Extended Abstract. Abstr Geol Soc Australia 1986;16:409e11. [15] Otter ML, Gurney JJ. Mineral inclusions in diamonds from the Sloan diatreme, Colorado-Wyoming State Line kimberlite district, North America. In: Kimberlites and related rocks. Their mantle/

242

[16]

[17] [18] [19]

[20] [21]

[22]

[23]

[24] [25]

[26] [27]

[28]

[29] [30]

[31]

[32] [33] [34]

Handbook of Advanced Ceramics

Crust Setting, Diamonds and Diamond Exploration, vol. 2. Geol Soc Australia Spec Publ 14; 1989. p. 1042e53. Jaquest AL, Hall AE, Sheraton JW. Composition of crystalline inclusions and C-isotopic composition of Argyle and Ellendale diamonds. In: Kimberlites and related rocks. Their mantle/Crust Setting, Diamonds and Diamond Exploration, vol. 2. Geol Soc Australia Spec Publ 14; 1989. p. 966e89. Di Pierro S, Gnos E, Grobety BH, et al. Rock-forming moissanite (natural a-silicon carbide). Am Mineral 2003;88:1817e21. Burlakov EV. The puiva deposit, subpolar Urals, Russia. Mineral Record 1999;30:451e65. Leung IS, Guo W, Freidman I, Gleason J. Natural occurrence of silicon carbide in a diamondiferous kimberlite from Fuxian. Nature 1990;346:352e4. Xu S, Wu W, Xiao W, et al. Moissanite in serpentinite from the Dabie Mountains in China. Am Mineral 2007;92:403e7. Nicheng S, Zhesheng M, Ming X. The crystal structure of (Fe4Cr4Ni)9C4. Science in China, Series D (Earth Sciences) 2005;48:338e45. Reichel B, Wagner K, Janish DS, Lengauer W. Alloyed W-(Co, Ni, Fe)-C phases for reaction sintering of hardmetals. Int J Refract Metals and Hard Mater 2010;28:638e45. Rombison PT, Paul T, Wenji B. Ultra-high pressure mineral in the Luobusha ophiolite, Tibet, and their tectonic implications. In: Malpas J, John LjN, Flecher JR, Aitchison JC, editors. Aspects of the tectonic Evolution of China, vol. 226. Geological Society London Special Publication; 2004. p. 247e71. Bowman AL. The variation of lattice parameter with carbon content of tantalum carbide. J Phys Chem 1961;65:1596e602. Novgorodova MIRG, Yusupov MT, Dmitrieva AI, et al. Khamrabaevite, (Ti, V, Fe)C, a new mineral. Zapiski Vserossiyskogo Mineralogicheskogo Obshchestva 1984;113(6). 697e603. Jedwab J. Native ruthenium in tantalum carbide concentrates from the Ural mountains. USSR Miner Petrol 1990;43:137e46. Novgorodova MI, Generalov ME, Trubkin NV. A new TaC-NbC isomorphous series and Niobocrbide, a new mineral from platinum-bearing placers of the Urals. Zapiski Vserossiyskogo Mineralogicheskogo Obshchestva 1997;126(1):76e95 (in Russian). Novgorodova MI, Generalov ME, Trubkin NV. Jedwabite, Fe7(Ta, Nb)C3, a new mineral in paragenesis with carbides of tantalum and niobium from platinum-bearing placers. Zapiski Vserossiyskogo Mineralogicheskogo Obshchestva. 1997;126(2):100e4 (in Russian). Zhang J, Yang G, Li Z. Discovery of natural tungsten carbide (WC) in China. Am Mineral 1989;74:948e53. Generalov ME, Naumov VA, Mokhov AV, Trubkin NV. Isovite (Cr, Fe)23C6-a new mineral from the gold-platinum-bearing placiers of the Urals. Zapiski Vserossiyskogo Mineralogicheskogo Obshchestva 1998;127(5):26e37 (in Russian). Zheng J, Griffin WL, O’Really, et al. 3.6 Ga lower crust in central China: new evidence on the assembly of the North China craton. Geology 2004:229e32. Nicheng S, Wenji B, Guowu L. Yarlongite: a new metallic carbide mineral. Acta Geologica Sinica 2008;83:52e6. Sharp WE. Pyrrhotite: a common inclusion in South African diamonds. Nature 1966;211:402e3. Stachel T, Harris JW, Brey GP. Rare and unusual mineral inclusions in diamonds from Mwadui, Tanzania. Contrib Mineral Petrol 1998;132:34e47.

[35] Goodrich CA, Berkley JL. Primary magmatic carbon in ureilites; evidence from cohenite-bearing metallic spherules. Geochim Cosmochim Acta 1986;50(5):681e91. [36] Jacob DE, Kronz A, Viljoen KS. Cohenite, native iron and troilite inclusions in garnets from polycrystalline diamond aggregates. Contrib Mineral Petrol 2004;146:566e76. [37] Scho¨nba¨chler et al. Nucleosynthetic Os Isotropic Anomalies in Carbonaceous Chondrites. 38th Lunar and Planetary Science Conference; 2007. [38] Lugaro M. Stardust from meteorites: an introduction to presolar grains. World Scientific Publishng Co. Pte. Ltd; 2005. [39] Daulton L, BernatowiczLewis TJ, et al. Polytype distribution in circumstellar silicon carbide. Science 2002;296:1852e5. [40] Bernatowicz TJ, Cowsik TJ, Gibbons PC, et al. Constrainson stellar grain formation from presolar graphite in the Murchison meteorite. Astrophys J 1996;472:760e82. [41] Hus G. The survival of presolar grains during the formation of the solar system. Conf Proc 402. In: Bernatowicz TJ, Zinner E, editors. Astrophysical implications of the study of the laboratory study of presolar materials. Woodbury: AIP; 1997. p. 721e48. [42] Helden G, Tielens AGGM, Heijnsbergen D. Titanium carbide nanocrystals in circumstellar environments. Science 2000;288: 313e6. [43] Gaustafsson B. Chemical analyses of cool stars. In: Annual review of astronomy and astrophysics, vol. 27; 1989. 701e756. [44] Sharp C, Wasserburg G. Molecular equilibria and condensation temperatures in carbon-rich gases. Geochim Cosmochim Acta 1995;59:1633e52. [45] Krut AN, Zolensk ME, Wasson JT, et al. Carbide-magnetite assemblages in type-3 ordinary chondrites. Geochim et Cosmoch Acta 1997;61:219e37. [46] Bulanova GP. The formation of diamond. J Geochem Explor 1995;53:1e23. [47] Gellat CD, Williams AR, Moruzza VL, in: Conference on the physics of transition metals, Leeds, England; 1980. [48] Pierson OH. Handbook of refractory carbides and nitrides: properties, characteristics, processing and applications. Noyes Publications; 1996. [49] Toth LE. Transition metal carbides and nitrides. New York: Academic Press; 1971. [50] Storms EK. The refractory metal carbides. New York: Academic Press; 1967. [51] Lengauer W, Eder A. Carbides: transition metal solid-state chemistry, Carbides: transition metal solid-state chemistry, 2000, p. 2e16. [52] Andrievski RA. Review-nanocrystalline high-melting point compounds-based materials. J Mater Sci 1994;29:614e31. [53] Andrievski RA. Strength of high-melting point compounds and materials on their base. Metallurgia, Cheliabinsk 1989. [54] Wassen R, Stover D. Processing and properties of nanophase nonoxide ceramics. Mater Sci Eng A 2001;301:59e68. [55] Andrievski RA. Nanocrystalline high-melting point carbides, borides, and nitrides. In: Gogotski Y, editor. Nanomaterials handbook. Taylor and Francis; 2006. p. 415e83. [56] Brewer L. Most striking confirmation of the Engel metallic correlation. Acta Metal 1967;15:553e6. [57] Davis RF. Nonstoichiometry and its effect on mass transport, orderdisorder phenomena, and deformation behavior in transition-metal

Chapter | 3.1

[58]

[59]

[60] [61] [62] [63] [64]

[65]

[66]

[67] [68] [69] [70]

[71]

[72] [73] [74]

[75]

[76]

[77] [78]

Silicon Carbide and Other Carbides: From Stars to the Advanced Ceramics

carbides. In: Catlow C, Mackrodt W, editors. Advances in ceramics. Nonstoichiometric compounds, vol. 23. The American Ceramic Society 1987. p. 529e57. Parthe´ E, Yvon K. On the crystal chemistry of the close packed transition metal carbides. II. A proposal for the notation of the different crystal structures. Acta Crystallogr B 1970;26: 153e63. Lengauer W. Transition metal carbides, nitrides, and carbonitrides. In: Riedel R, editor. Handbook of ceramic hard materials. Weinheim, Germany: Wiley-VCH Verlag GmbH; 2008. Lee WE, Rainforthe M. Ceramic microstructures property control by processing. London: Chapman and Hall; 1994. 415e451. Shaffer PT. A review of the structure of silicon carbide. Acta Cryst B 1969;25:477e82. Saddow SE, Agarwal A. Advances in silicon carbide: processing and applications. London: Artech House Publishers; 2004. Cheung R. Silicon carbide microelectromechanical system for harsh environments. Imperial College Press London; 2006. Kelly JF, Fisher GR, Barnes P. Correlation between thickness and periodicity of long polytypes in silicon carbide. Mater Research Bull 2005;40:249e55. Snead LL, Nozawa T, Katoh Y, et al. Handbook of SiC properties for fuel performance modeling. J Nucl Mater 2007;371: 329e77. Takekawa TH, Tsutsumi M. 6H/3C-SiC transformation in SiCTiN powder mixture at high temperature. J Ceram Soc Jpn 1991;99:376e9. Pampuch R, Stobierski L. Solid combustion synthesis of refractory carbides: (a review). Ceram Inter 1991;17:69e77. Pampuch R. Some fundamental and applied aspects of solid combustion. Solid State Ionics 1997;101e103. 899e807. Schubert U, Hu¨sing N. Synthesis of inorganic materials. Weinheim, Germany: Wiley-VCH Verlag GmbH; 2000. Yen BK. X-ray diffraction study of mechanochemical synthesis and formation mechanisms of zirconium carbide and zirconium silicides. J Alloy Comp 1998;268:266e9. El-Eskandarany MS. Mechanical alloying for fabrication of an advanced engineering materials. Norwich. N.Y: Noyes Publications; 2001. Krstic DV. Production of fine, high-purity beta silicon carbide powders. J Am Ceram Soc 1992;75:170e4. Schwetz AK. Silicon carbide and its high-technology ceramics. Radex-Rundschau, Heft 1989;1:26e37. Devecerski A, Radosavljevic-Mihajlovic A, Egelja A, et al. Fabrication of SiC by carbothermal-reduction reactions of sepiolite. Materials Science Forum 2007;555:261e5. Matovic B, Saponjic A, Devecerski A. Fabrication of SiC by carbothermal-reduction reactions of diatomaceous earth. J Mater Sci 2007;42:5448e51. Devecerski A, Posarac M, Egelja A, et al. Fabrication of SiC by charbothermal-reduction reactions of mountain lather asbestos. J Alloy Compound 2008;464:270e6. Filsinger DH, Bourrie DB. Silica to silicon: key carbothermic reaction and kinetics. J Am Ceram Soc 1990;73:1726e32. Shi Y, Zhang F, Hu YS, et al. Low-temperature pseudomorphic transformation of ordered hierarchical macro-mesoporous SiO2/C nanocomposites to SiC via magnesiothermic reduction. J Am Chem Soc 2010;132:5552e3.

[79]

[80]

[81] [82] [83]

[84]

[85]

[86] [87] [88]

[89] [90]

[91]

[92]

[93]

[94]

[95]

[96]

[97]

[98]

243

Greil P. Pyrolysis of active and passive fillereloadded preceramic polymers. In: Somiya S, editor. Handbook of advances ceramics. Academic Press./ /-SIC can; 2003. p. 369e90. Dressler W, Riedel R. Progress in silicon-based non-oxide structural ceramics. Int J Refractory Metals and Hard Materials 1997;15:13e47. Bo¨cker WDG. Covalent high-performance ceramics. Adv Mater 1992;4:169e78. Tanaka H, Kurachi Y. Synthesis of b-SiC powder from organic precursor and its sinterability. Ceram Int 1988;14:109e15. Hasegawa I, Nakamura T, Motojima S, Kajiwara M. Silica gelphenolic resin hybrid fibres; new precursor for continuous bsilicon carbide fibres. J Mater Chem 1995;5:193e4. Li J, Tian J, Dong L. Synthesis of SiC by a two-step sol-gel process and their conversion to SiC powders. J Eur Ceram Soc 2000;77:1853e7. Sacks MD, Wang CA, Yang Z, Jain A. Carbothermal reduction synthesis of nanocrystalline zirconium carbide and hafnium carbide powders using solution-derived precursors. J Mater Sci 2004;39:6057e66. Ma M, Shen W, Zhang P, et al. Synthesis of TaC nanopowders by liquid precursor route. Matter Lett 2010 (Accepted). Greil P. Biomorphous ceramic from lignocellulosic. J Eur Ceram Soc 2001;21:105e18. Vogli E, Sieber H, Greil P. Biomorphic SiC-ceramic prepared by Si-vapor phaseinfiltration of wood. J Eur Ceram Soc 2002;22: 2663e8. Matovic B, Boskovic S. Synthesis oxide ekoceramics. Roman J Mater 2008;38:329e34. Arrelano-Lopez AR, Martinez-Fernadez J, Gonzales P, et al. Biomorphic SiC: new engineering ceramic materials. Int J Appl Ceram Technol 2004;1:56e67. Matovic B, Egelja A, Babic B, et al. The synthesis and microstructure of morphogenetic SiC ceramics. J Optoelectronic Adv Mater 2008;10:549e53. Egelja A, Gulicovski J, Devecerski A, et al. Preparation of biomorphic SiC ceramics. Science of Sintering 2008;40: 141e5. Rambo CR, Cao J, Russina O, Sieber H. Manufacturing of biomorphic (Si, Ti, Zr)-carbide ceramics by sol-gel processing. Carbon 2005;43:1174e83. Khongwong W, Imai M, Yoshida K, Yano T. Synthesis of b-SiC/ SiO2 core-shell nanowires by simple thermal evaporation. J Ceram Soc Jpn 2009;117:194e7. Khongwong W, Imai M, Yoshida K, Yano T. Influence of raw powder size, reaction temperature, and soaking time on synthesis of SiC/SiO2 coaxial nanovires via thermal evaporation. J Ceram Soc Jpn 2009;117:439e44. Khongwong M, Yoshida K, Imai M, Yano T. Simple approach to fabricate SiC/SiO2 composite nanowires and their oxidation resistance. Mater Sci Eng B 2010;173:117e21. Opeka M, Talmy I, Zaykoski J. Oxidation-based materials selection for 2000  C hypersonic þ aerosurfaces: theoretical considerations and historical experience. J Mater Sci 2004;39:5887e904. Wuchina E, Opila E, Opeka M, Fahrenholtz W, Talmy I. UHITCs: ultra-high temperature ceramic materials for extreme environment applications. The Electrochem Soc Interface Winter 2007;16: 30e6.

244

[99] Ogbuji LUJT, Opila EJ. A comparison of the oxidation kinetics of SiC and Si3N4. J Electrochem Soc 1995;142:925e30. [100] Ogbuji LUJT. Effect of oxide diversification on oxidation kinetics of SiC. J Am Ceram Soc 1997;80:1544e50. [101] Weidenmann KA, Rixecker G, Aldinger F. Liquid phase sintered silicon carbide (LPS-SiC) ceramics having remarkably high oxidation resistance in wet air. J Eur Ceram Soc 2006;26: 2453e7. [102] Opila EJ, Hann RE. Paralinear oxidation of CVD SiC in water vapor. J Am Ceram Soc 1997;80:197e205.

Handbook of Advanced Ceramics

[103] Eggler DH, Baker DR. Reduced volatiles in the system C-O-H: Implications to mantle melting, fluid formation and diamond genesis. In: Akimoto S, Manghani MH, editors. High pressure research in geophysics. Academic Publications; 1982. p. 237e50. Ctr. [104] Enggler DH, Lorand JP. In: Meyer HOA, Leonardos OH, editors. Diamonds: characterization, genesis and exploration. Camp Pesq Rec Miner Spec Publ IB 1994. p. 160e9. [105] Russell S. Stars in stones. Nature 2004;428:903e4. [106] Nittler LR. Presolar stardust in meteorites: recent advances and scientific frontiers. Earth and Planetary Sci Lett 2003;209:259e73.

Chapter 3.2

Review and Overview of Silicon Nitride and SiAlON, Including their Applications Hideki Kita,* Kiyoshi Hirao,y Hideki Hyuga,y Mikinori Hottay and Naoki Kondoy *

Department of Molecular Design and Engineering, Graduate School of Engineering, Nagoya University, Furo-cho, Chikusa, Nagoya, Aichi 464-8603, Japan, Former Affiliate: National Institute of Advanced Industrial Science and Technology, y National Institute of Advanced Industrial Science and Technology, 2266-98 Anagahora, Shimoshidami, Moriyama, Nagoya, Aichi 463-8550, Japan

Chapter Outline Introduction High Thermal Conductivity Silicon Nitride Ceramics References Development of Silicon Nitride for High-Temperature Use References Low-Friction Si3N4 Ceramics with Carbon Fiber References Frictional Properties and Microstructure of Si3N4 Containing Mo and Fe Compounds Prepared by Hot Pressing 4.1. Introduction 4.2. Experimental Method 4.3. Experimental Results and Discussion References Low-cost Fabrication of Silicon Nitride Ceramics References

245 245 248 248 249 250 251 252 252 252 252 254 254 256

INTRODUCTION The applications of silicon nitrides mainly utilize their mechanical properties such as high strength, high durability, high wear resistance, as well as chemical stability at room and high temperatures. Recently, nitride ceramics have received much more attention in the field of electronics, in the context of the following social background. In this paper, highly functionalized silicon nitride and SiAlON and their applications are reviewed. Sub Chapter 1

High Thermal Conductivity Silicon Nitride Ceramics Energy and environment-related problems are serious social issues. In order to save energy as well as to reduce the

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00015-0 Copyright Ó 2013 Elsevier Inc. All rights reserved.

High-Strength and High-Toughness Silicon Nitride Ceramics References Development of Fine-Grained Silicon Nitride Ceramics with a Small Amount of Sintering Additive Development of Advanced a-SiAlON Ceramics 8.1. Synthesis of Spherical Hollow SiAlON Powders Composed of Nanosized Particles 8.2. Fabrication of Dense SiAlON Nanoceramics 8.3. Fabrication of Porous SiAlON Ceramics References Applications 9.1. Automotive Applications 9.2. Industrial Applications

256 259 259 261 261 262 263 263 264 264 265

emission of carbon dioxide, energy sources tend to shift from fossil fuel to electric power; hence, highly efficient use of electric power becomes extremely important. Power devices are key technologies for this purpose. Power devices conduct direct conversion and control of electric power by means of semiconductors, which can drastically save energy compared to traditional systems. Power supply and packing density of power modules are increasing more and more as their application field expands, in particular in the automobile industry. In order to guarantee the stable operation of the power module, heat release technology in the module becomes of great importance. So far, AlN has been used as a substrate for power devices. In general, a thick metal plate as an electrode (about 0.3e0.6 mm in thickness) is directly bonded to a ceramic substrate via high-temperature brazing as illustrated in Figure 1.1. With an increase of the power density in the module, micro-cracks in the ceramic caused by the large difference in thermal expansion coefficient of metal and ceramic become a serious problem. Such a situation leads to strong demands for insulating substrates to

245

246

Handbook of Advanced Ceramics

FIGURE 1.1 Structure of power module using a ceramic substrate.

FIGURE 1.3 Microstructure parameters affecting thermal conductivity of sintered b-silicon nitrides.

have higher mechanical properties, and thereby silicon nitrides are receiving a lot of attention. Figure 1.2 summarizes the properties of commercial ceramic substrates. The thermal conductivities of commercially available silicon nitrides are less than 90 W/(mK), though their strength is more than twice as high as those of aluminum nitrides. An important issue in silicon nitride is, therefore, to increase thermal conductivity without degradation of mechanical properties. So far, some researchers have estimated the intrinsic thermal conductivity of b-Si3N4 crystal based on solid-state physics [1e3]. From these fundamental simulations, it could be anticipated that the intrinsic thermal conductivity of silicon nitride is over 200 W/(mK) [4]. Being different from the ideal single crystal, a sintered silicon nitride is composed of b-Si3N4 grains as well as two grain boundaries, and secondary phases with low thermal conductivities. In addition, inside the grains, there are several kinds of crystal imperfections such as point, line, and planer defects as shown in Figure 1.3 [5,6]. These defects scatter phonons to decrease thermal conductivity of the b-Si3N4 crystal. Similar to the case of high thermal conductivity AlN, lattice oxygen, or oxygen dissolving in b-Si3N4 lattice, drastically decreases the thermal conductivity of silicon nitrides as shown in Figure 1.4 [5,7].

In order to reduce these negative parameters, high thermal conductivity Si3N4 is generally fabricated with the following points in mind [4]. (1) High-purity fine Si3N4 is employed in order to reduce impurities. (2) Concurrent addition of rare earth oxide and alkaline earth oxides is conducted; the former plays a role of decreasing lattice oxygen and latter is needed for assisting densification. With respect to alkaline earth element, the use of MgSiN2 as one of the sintering additives was found to be effective for improving thermal conductivity [8]. (3) Enhancement of grain growth is one of the important factors for improving thermal conductivity because it can decrease the number of two grain junctions as well as the lattice oxygen via solution re-precipitation process [4]. In practice, gas-pressure sintering (GPS) with long sintering time over several tens of hours is necessary in order to achieve a high thermal conductivity over 100 W/(mK). Figure 1.5 shows a relation between thermal conductivity and bending strength of GPS silicon nitrides. Considerable grain growth is needed in order to improve thermal conductivity over 100 W/(mK), which results in a drastic decrease in strength. Even in the high-purity powder it

FIGURE 1.2 Properties of commercially available ceramic substrates. Data were collected through trade catalogs or websites of Japanese ceramic manufacturing companies.

FIGURE 1.4 Effect of lattice oxygen content on the thermal resistivity of Si3N4 fabricated by gas-pressure sintering (closed circle) and hot pressing (HP; square). The thermal resistivities for the hot-pressed specimens were measured in two directions parallel to and perpendicular to the hotpressing direction Refs [5,6].

Chapter | 3.2

Review and Overview of Silicon Nitride and SiAlON, Including their Applications

247

TABLE 1.1 Characteristics of Raw Si Powder. Ref. [11]

FIGURE 1.5 Relation between thermal conductivity and bending strength of gas-pressure sintered silicon nitrides with Yb2O3eMgO or Y2O3eMgO as sintering additives.

contains about 1.2 mass% of impurity oxygen, one-third of which is lattice oxygen, which makes it difficult to fabricate high thermal conductivity Si3N4 with high strength via the GPS method. In order to overcome this problem, a research group at AIST, Japan, has taken notice of reaction-bonding process [9e11]. When a Si powder compact is heated around 1400  C in nitrogen atmosphere, it is nitrided to form a silicon nitride compact. As the nitridation reaction proceeds mainly via gasesolid reaction system, the external dimensions of the compact retains while each of the individual Si particles expands by about 22%. Consequently, the silicon nitride compact with higher density can be obtained without dimensional change. Well-known advantages of reaction-bonding (RB) process include the following: (1) cheaper Si powder can be employed as starting powder rather than the more expensive Si3N4 powder and (2) high dimensional accuracy due to low shrinkage after post sintering [12]. In addition, it is possible to carry out the whole process from nitridation to post-sintering process without exposing the compacts to the air, which is favorable for controlling the oxygen content in Si3N4. Zhou et al. have fabricated Si3N4 ceramics by nitriding Si compacts with Y2O3 and MgO as sintering additive at 1400  C for 8 h followed by post-sintering at 1900  C for 1e48 h [11]. For the sake of comparison, GPS Si3N4 ceramic with same composition was fabricated in the same sintering conditions. High-purity Si powder with an average particle size of about 7 micron was used (see Table 1.1). It includes impurity oxygen of about 0.34 mass%, and, after milling the powder in the same condition for mixing of the starting powder, it was slightly oxidized such that impurity oxygen increases to about 0.7 mass%. After complete nitridation, oxygen content in converted Si3N4 is estimated to be 0.42 mass%. It should be noticed that oxygen content in the nitride compact is expected to be less than half as low as that of commercial Si3N4 powder with high purity. Figure 1.6 shows a variation of thermal conductivity for sintered reaction-bonded Si3N4 (SRBSN) and gas-pressure sintered Si3N4 (GPS) with sintering time. The SRBSNs

Si powder (for SRBSN)

Si3N4 powder1 (for GPS)

Particle size

7 mm

0.1 mm

Purity (except oxygen)

>99.9 %

>99.9 %

Oxygen content (as received)

0.34 mass%

e

Oxygen content (after milling)

0.70 mass%

e

Oxygen content in Si3N4 (after nitridation)2

0.42 mass%2

1.2 mass%

1 2

E-10 grade powder, UBE industry Co., Japan. Estimated oxygen content in a specimen after complete nitridation.

exhibit higher thermal conductivity compared to the GPS materials. After 48 h of heating, the SRBSN exhibited high thermal conductivity over 140 W/(mK). Another important result for SRBSN is that it exhibits high fracture toughness value (KIC). Figure 1.7 shows microstructures of SRBSNs post-sintered at 1900  C for 6 and 24 hrs, along with their mechanical properties [11]. Specimen sintered for 6 h exhibits good mechanical properties combing high strength over 700 MPa and high fracture toughness of 8.4 MPa$m1/2. The specimen sintered for 24 hr has a moderate strength of about 600 MPa and extreme high fracture toughness over 11 MPa$m1/2. Well-developed rodlike grains can be seen in the fractured surface of SRBSNs, which is one reason for high KIC of the materials. In addition, no use of Al2O3 as a sintering additive might be another factor for high KIC of the specimens. It is well known that Al2O3 dissolves into b-Si3N4 lattice to form b-SiAlON, which results in strong interfacial strength. The situation in this study is opposite to the case including Al2O3.

FIGURE 1.6 Thermal conductivity of sintered reaction-bonded Si3N4 (SRBSN) and gas-pressure sintered Si3N4 (GPS) sintered at 1900  C under 0.9 MPa N2.

248

Handbook of Advanced Ceramics

[11] Zhou Y, Zhu X, Hirao K, Lences Z. Sintered reaction-bonded silicon nitride with high thermal conductivity and high strength. Int J Appl Ceram Technol 2008;5:119e26. [12] Moulson AJ. Reaction-bonded silicon nitride: its formation and properties. J Mater Sci 1979;14:1017e51.

Sub Chapter 2 FIGURE 1.7 Fractured surfaces of sintered reaction-bonded Si3N4 sintered at 1900  C for (a) 6 h and (b) for 24 h under 0.9 MPa N2. Thermal conductivity, 4-point bending strength, and fracture toughness are 105 W/ (mK), 736 MPa, and 8.4 MPa$m1/2 for the specimen (a), and 132 W/(mK), 516 MPa, and 11 MPa$m1/2 for the specimen (b).

In summary, compared with the GPS materials prepared from a high-purity Si3N4 powder, the SRBSN materials from Si powder with low oxygen impurity exhibited a better balance between thermal conductivity and bending strength. The reaction bonding of silicon powder compacts followed by post-sintering is a promising way of the fabrication of Si3N4 with both high thermal conductivity and high strength.

REFERENCES [1] Haggerty JS, Lightfoot A. Opportunities for enhancing the thermal conductivities of SiC and Si3N4 ceramics through improved processing. Ceram Eng Sci Proc 1995;16:475e87. [2] Watari K, Li B-C, Pottier L, Fournier D, Toriyama M. Thermal conductivity of b-Si3N4 single crystal. Key Eng Mater CSJ Series 2000;181/82:239e49. [3] Hirosaki N, Ogata S, Kocer C. Molecular dynamics caluculation of the ideal thermal conductivity of single-cryastal a- and b-Si3N4. Phys Rev 2002;B65:134110/1e13411/11. [4] Hirao K, Zhou Y. Session 16: thermal conductivity. In: Riedel R, Chen IW, editors. Ceramics science and technology, Vol. 2. WileyVCH; 2010. p. 667e96. [5] Hirao K, Watari K, Hayashi H, Kitayama M. High thermal conductivity silicon nitride ceramic. Mat Res Bull 2001;26: 451e5. [6] Kitayama M, Hirao K, Tsuge A, Watari K, Toriyama M, Kanzaki S. Thermal conductivity of b-Si3N4 I: microstructural factors. J Am Ceram Soc 1999;82(11):3105e12. [7] Kitayama M, Hirao K, Tsuge A, Watari K, Toriyama M, Kanzaki S. Thermal conductivity of b-Si3N4 II: effect of lattice oxygen. J Am Ceram Soc 2000;83:1985e92. [8] Hayashi H, Hirao K, Toriyama M, Kanzaki S, Itatani K. MgSiN2 addition as a means of increasing the thermal conductivity of b-silicon nitride. J Am Ceram Soc 2001;84:3060e2. [9] Zhu X, Zhou Y, Hirao K, Lences Z. Processing and thermal conductivity of sintered reaction-bonded silicon nitride. I: effect of Si powder characteristics. J Am Ceram Soc 2006;89(11): 3331e9. [10] Zhu X, Zhou Y, Hirao K, Lences Z. Processing and thermal conductivity of sintered reaction-bonded silicon nitride. (II) effect of magnesium compound and yttria additives. J Am Ceram Soc 2007;90(6):1684e92.

Development of Silicon Nitride for High-Temperature Use There are some candidate ceramics for high-temperature use. Silicon nitride is one of them, which shows higher fracture toughness than the other ceramics. (Table 2.1) This is why silicon nitride has been developed for hightemperature use. A well-known silicon nitride, SN282 by Kyocera Co. Ltd., was developed for ceramic gas turbine components. [1,2] This silicon nitride has been improved for practical requirement. SN282 contains lutetium oxide, Lu2O3, as sintering additive. Microstructure is shown in Figure 2.1, and properties are summarized in Table 2.2. Many research works were done following SN282 to reveal why SN282 shows excellent high-temperature properties. Figure 2.2 shows a phase diagram of Si3N4eLu2O3eSiO2 system. Some research focused on the ratio of Lu2O3 and SiO2. As this system has grass formation region in the Si3N4eLu2SiO5eLu2Si2O7 triangle, this region is mentioned “G” in the triangle. At an elevated temperature above 1400  C, the glass is softened to degrade hightemperature strength. Difference between the composition “A” and “B” is small. However, microstructure, strength at room and elevated temperatures, and fracture toughness of the gas-pressure sintered specimens are quite different [4]. Properties are summarized in Table 2.2. SN282 has

TABLE 2.1 Properties of Major Ceramics for High-Temperature Use Oxidation resistance

Creep resistance

Fracture toughness

Silicon Nitride

Moderate

Good

Good

Silicon Carbide

Good

Very Good

Moderate

Alumina

Very Good

Moderate

Moderate

Chapter | 3.2

Review and Overview of Silicon Nitride and SiAlON, Including their Applications

249

TABLE 2.3 Properties of Gas-pressure Sintered Silicon Nitrides with Composition “A” and “B” Secondary Strength Strength Fracture at 1500 toughness Grain crystalline at RT  C MPa MPa$m1/2 MPa Composition size phases A

Small Lu2Si2O7

550

400

4

B

Large Lu2Si2O7, Lu2SiO5

640

330

10*

*Fracture toughness of 10 might be higher than the accurate value.

FIGURE 2.1 Microstructure of SN282 silicon nitride ceramic.

TABLE 2.2 Properties of SN282 Silicon Nitride Ceramic [MHI] Density

Strength at RT MPa

Strength at 1500  C MPa

Thermal conductivity W/mK

3.4

634

366

64

a composition similar to “A”. In composition “A” located near the line Si3N4eLu2Si2O7, therefore, a small amount of glassy phase forms during sintering. This leads to the smaller grain size. Strength and fracture toughness at room temperature are low due to the small amount of grain boundary glassy phase. At elevated temperatures, however, a smaller amount of glassy phase is advantageous to maintain the strength. Composition “B” is located inside the Si3N4eLu2SiO5eLu2Si2O7 triangle. As a larger amount of glassy phase formed, properties were contrastive to “A”. To improve the properties of composition “A”, two techniques were tried. One is adding a small amount of Yb2O3 [5] and another is making anisotropic microstructure. [6] By adding 0.2 wt% Yb2O3, the strength of gas-pressure sintered specimen at RT was same as in “A” but strength at

FIGURE 2.2 Phase diagram of Si3N4eLu2O3eSiO2 system.

1500  C and fracture toughness at RT are improved to 480 MPa and 6 MPa$m1/2, respectively. By making anisotropic microstructure using forging technique, strength along the alignment direction at RT and 1500  C is improved to 950 MPa and 700 MPa, respectively (Table 2.3). Another work to change the composition far from “A” and “B” was done. Composition “C”, which is located near the line Si3N4eLu4Si2O7N4, is chosen. [7,8] This composition is expected to form a small amount of glassy phase at elevated temperatures. Therefore, the hot-pressed specimen shows a very high strength of 800 MPa at 1500  C as well as creep resistance. On May 21, 2010, Japanese Venus Climate Orbiter “AKATSUKI (PLANET-C)” was launched. [3,9], “AKATSUKI” has a ceramic thruster, made of the silicon nitride, SN282, for the first time in the world. On June 28, 2010, the thruster jetted for 13 seconds, and on-orbit verification was successfully performed for the first time in the world. Ceramic gas turbine is far from practical use, but the silicon nitride developed for the turbine is successfully transferred to aerospace use as a thruster.

REFERENCES [1] Tanaka K, Tsuruzono S, Terazono H. Characteristics and application of new silicon nitride materials, SN281 and SN282, for ceramic gas turbine components. In: Niihara K, Hirano S, Kanzaki S, Komeya K, Morinaga K, editors. Proc. of 6th international symposium on ceramic materials and components for engines; 1997. p. 248e52. [2] Nakashima T, Tatsumi T, Takehara I, Ichikawa Y, Kobayashi H. Research & development of CGT302. In: Niihara K, Hirano S, Kanzaki S, Komeya K, Morinaga K, editors. Proc. of 6th international symposium on ceramic materials and components for engines; 1997. p. 233e6. [3] Mishima H, Morishima K, Nonaka Y, Nishino H, Sawai S. Development of ceramic thruster for space propulsion system. Mitsubishi Heavy Industries Technical Review 2005;42(5):250e3. [4] Kondo N, Ishizaki M, Ohji T. Fabrication of silicon nitride with lutetia additive. Silicate Industries 2004;69(7-8):183e6. [5] Kondo N, Hyuga H, Yoshida K, Kita H, Ohji T. Effect of Yb2O3 addition on Si3N4-Lu2O3-SiO2 ceramics. J Ceram Soc Jpn 2006;114: 1097e9.

250

Handbook of Advanced Ceramics

[6] Kondo N, Asayama M, Suzuki Y, Ohji T. High-temperature strength of sinter-forged silicon nitride with lutetia additive. J Am Ceram Soc 2003;86:1430e2. [7] Guo S-Q, Hirosaki N, Nishimura T, Yamamoto Y, Mitomo M. Hotpressed silicon nitride with Lu2O3 additives: oxidation and its effect on strength. J Am Ceram Soc 2003;86:1900e5. [8] Nishimura T, Guo S, Hirosaki N, Mitomo M. Improving heat resistance of silicon nitride ceramics with rare-earth silicon oxynitride. J Ceram Soc Jpn 2006;114:880e7. [9] Venus Climate Orbiter “AKATSUKI / PLANET-C”, http://www.stp. isas.jaxa.jp/venus/top_english.html

Sub Chapter 3

Low-Friction Si3N4 Ceramics with Carbon Fiber Ceramic materials exhibit properties that make them suitable candidates for a number of industrial applications. In the field of machinery industries, they are expected to be used where the tribological environment is severe, such as under high normal pressures and corrosive environments [1,2]. In particular, there is the possibility of using ceramics in tribological applications under nonlubricated or marginally lubricated sliding contacts. Compared with other engineering ceramics, silicon nitride has superior mechanical properties, with high fracture strength and fracture toughness. In tribological applications, Si3N4 ceramics have been used as the ball component of bearing systems under lubricated conditions. However, the sliding contact of Si3N4eSi3N4 self-mated tribopairs under dry conditions produces a high friction coefficient and high wear rate because the abrasive wear is affected by the intrinsic brittle nature of ceramics [3]. Therefore, many attempts have been carried out to produce Si3N4-based ceramics with a solid lubrication function, and many researchers have reported self-lubricating materials. Among these materials, graphite possesses one of the highest solid lubricating functions. However, technical problems arise during the fabrication of ceramic graphite composites, since the fine graphite powder employed is difficult to mix with the raw ceramic powder due to its hydrophobic nature, and results in adhesion to the balls and pot used for the ball milling. In graphite materials, there is little work reporting on the solid lubrication properties of ceramic/carbon fiber composites. It has been reported that the solid-lubrication effect of ceramic/carbon composites is not sufficient to allow their use in tribological applications under dry sliding conditions. It has been reported that under dry sliding conditions against a stainless steel counterbody, a Si3N4 ceramic/carbon fiber composite did not produce any beneficial lubricating effect when compared

with the results of the matrix material alone [4]. It was suggested that this was because the formation of a graphitic lubricating film on the surface of the counter-body did not occur, due to the inherent properties of the fiber employed. The properties of carbon fibers, in particular the elastic modulus, are determined by the degree of texturing of the graphite bonding layers and by the amount of graphitic bonding in the carbon fiber composition. Our group has recently shown that fibers with a high degree of orientation and high graphite content were effective for decreasing the friction coefficient under dry sliding conditions [5,6]. We introduced this Si3N4/carbon short fiber composite. Figure 3.1 shows SEM images of the cross section of the carbon fibers with high elastic modulus. It can be seen that the layering is more defined in the fiber. These results indicated that the graphite content was higher, the size of graphite crystals was larger, and the graphite layer structure had a higher radial orientation in the fibers with higher tensile modulus. For fabrication of carbon fiber-dispersed Si3N4-based composite, the hot-press method was chosen. The composition of the starting powders was 93 mass % Si3N4, 2 mass % Al2O3, and 5 mass % Y2O3. The raw powder was homogeneously mixed by ball milling and then dried. After sieving the dried powders, they were mixed with 5 vol % of the respective carbon fibers using planetary ball milling. Prior to mixing, the fibers were cut to lengths of 3e5 mm. The composite powders were hot-pressed at 1950  C for 2 h at 30 MPa pressure in a 900-kPa nitrogen atmosphere. The microstructure of the obtained the composite is shown in Figure 3.2. Homogeneously dispersed carbon fibers of a few hundred microns in length were observed in the composite. The bending strength of the composite was around half that of the monolithic Si3N4, indicating that the carbon fibers act as defects. On the other hand, fracture toughness was increased up to about 10 MPa$m1/2. Fiber pull-out was observed on the fractured surface of this composite as reported by many researchers [7,8] for ceramics/fiber composites, and this mechanism was responsible for the toughening effects of

FIGURE 3.1 SEM images of the cross section of the carbon fibers with high elastic modulus.

Chapter | 3.2

Review and Overview of Silicon Nitride and SiAlON, Including their Applications

251

FIGURE 3.3 Worn tracks of disks of this composite and a conventional Si3N4 ceramics under dry sliding test.

FIGURE 3.2 The microstructure of the obtained the composite.

the Si3N4/carbon short fiber composite. Table 3.1 shows the tribological test results of this composite and a conventional Si3N4 ceramics under dry sliding test. In the case of monolithic Si3N4, the initial friction coefficient indicated higher values, which were maintained in the steady state. On the other hand, although the initial friction coefficient of the Si3N4/carbon fiber composites also indicated high values, it gradually decreased and was in the range of 0.25. Figure 3.3 shows worn tracks of disks of this composite and a conventional Si3N4 ceramics under dry sliding test. The worn track of monolithic Si3N4 showed a rough surface with fine crushed debris. On the other hand, for the composite specimens, the graphite fibers were preferentially worn in all cases. In addition, the worn tracks were covered with adhesive debris with a smooth appearance, which may have consisted mainly of SiO2 formed by tribochemical reaction [9] and graphite. These results indicated that the graphite fibers inside the composite were supplied to the worn surface as a result of partial fracture followed by grain dropping during the wear test and acted as solid lubrication with adhesion on the surface depleting the graphite layer structure of the carbon fibers. The bonding energy between crystal layers of graphite is weak, such that slip between crystal layers is assumed to be one of the dominant mechanisms of low-friction sliding. The friction coefficient of graphite has been reported to be 835 nm wavelength) was sent to a monochromator through an optical fiber and analyzed by a CCD detector. Figure 8 shows the resonant Raman spectra from terrace-microsphere (of the BaOeSiO2eTiO2 glass) with various irradiation points. Corresponding laser pumping

FIGURE 8 Resonant Raman spectra from terrace microsphere with various irradiation points. Corresponding laser pumping spots (from A to B) are shown in the inset figure. For color version of this figure, the reader is referred to the online version of this book.

407

spots (from A to E) are illustrated in the inset figure. Pumping laser power was 10 mW. Strong resonant peaks due to WGMs were observed, especially when the terrace portion (the spot B) was irradiated. Emission intensities of a terrace microsphere are plotted against the pumping power of a CW Arþ laser (l ¼ 514.5 nm) and the threshold was estimated to be 2.5 mW. Emission spectrum of a terrace microsphere pumped by a tunable CW Ti:sapphire laser (l ¼ 750e820 nm) are shown in Figure 9. Pumping power was 15 mW. The upper emission spectrum is that from a terrace microsphere. A terrace microsphere performed strong resonant peaks due to WGMs in the wavelength region of 835e940 nm. When an uncoated microsphere was pumped, however, the resonant peaks appeared only in the 860e940 nm regions (not shown in the figure). The bottommost spectra showed Raman scattering due to the BaOeSiO2eTiO2 glass (l ¼ 835e880 nm) and fluorescence due to Nd3þ (l ¼ 880e940 nm), where the pumping wavelength was 810 nm. In Figure 10, the emission intensities of a terrace microsphere are plotted against pumping power. Blue and Red lines in Figure 10 correspond to blue and red circles in the upper spectrum in Figure 9: the resonant peaks due to Raman scattering (l ¼ 870.9 nm) and fluorescence (l ¼ 901.1 nm). The slope of Nd3þ fluorescence line changed at about 4 mW. In the Raman scattering plot, the slope of emission intensity vs. pumping power line increased clearly above 4 mW, the threshold. As shown in the figure (dashed

FIGURE 9 Emission spectrum of a terrace microsphere pumped by Ti:sapphire laser. For color version of this figure, the reader is referred to the online version of this book.

408

Handbook of Advanced Ceramics

energy transfer from fluorescence to Raman scattering support the enhancement effect in the glass sphere [40]. Second, the threshold of the Nd3þ-stimulated emission itself was of remarkably low value, less than 1 mW. These results reflect the strong interaction between laser light and glass matrix in spherical-cavity glass microspheres.

3.2. Pumping Demonstration Using a Half-Polished Fiber Coupler

FIGURE 10 Emission intensities against pumping power. Blue and red lines correspond to the emission peaks highlighted with the same color in Figure 9. For color version of this figure, the reader is referred to the online version of this book.

line), the threshold of stimulated Raman scattering and the change point in the Nd3þ fluorescence intensity was correlated with each other. The two physical meanings should be noted from Figures 9 and 10. First, stimulated Raman scattering (SRS) was observed and the intensities of the emission peaks showed the maximum at the overlap region of spontaneous Raman scattering band with fluorescence band due to Nd3þ. It seemed that the SRS enhancement effect caused by the fluorescence was induced in the solid-state glass spheres. The fluorescence intensity showed the saturation of intensity against pumping power and subsequently SRS emission started in Figure 10; The FIGURE 11 Setup for the pumping experiments using a half-polished fiber coupler. For color version of this figure, the reader is referred to the online version of this book.

The setup for the pumping experiments of the high-index glass microspheres is shown in Figure 11. A glass microsphere was placed on a coated half-polished fiber coupler. A tunable CW Ti:sapphire laser was used as a pumping light source. Emission from a glass microsphere was collimated by a lens with the similar technique described above in the terrace-microsphere experiment. The fiber coupler was made from a single-mode optical fiber with a core diameter of 3.7 mm (refractive index of the core glass is nD ¼ 1.470, cut-off wavelength lc ¼ 820 nm) and outer diameter of 125 mm (refractive index of the cladding glass, nD ¼ 1.461). The bending fiber with a certain curvature was buried into plastic resin and was polished from surface to center of the fiber core. To improve optical coupling efficiency, high-refractive-index film (nD ¼ 1.73, 500 nm in thickness) made from organiceinorganic hybrid material was coated on the polished core. Typical emission spectra from high-index glass microspheres are shown in Figure 12. In the upper and lower spectra, the sphere diameters were 36 mm and 33 mm, respectively. The similar spectra were obtained with those of Figure 9. Using the uncoated fiber coupler, only the resonant emissions in 880e940 nm wavelength regions were observed (not shown in the figure). Emission was originated from Raman scattering (l ¼ 840e880 nm) and Nd3þ fluorescence (l ¼ 880e940 nm), and the calculated mode spacing agreed well with the measured values, for example, 4.35 nm vs. 4.39

Chapter | 5.3

Optical Resonators and Amplifiers: Fiber, Waveguide, and Spherical Lasers

409

FIGURE 13 Emission intensities against pumping power in the pumping experiment using a coated half-polished fiber coupler.

FIGURE 12 Typical emission spectra from high-index glass spheres pumped by Ti:sapphire laser through a coated half-polished fiber coupler. For color version of this figure, the reader is referred to the online version of this book.

nm in the upper spectrum. Emission intensities of 860.1 nm (due to Raman scattering) and 902.1 nm wavelength (due to fluorescence) are plotted as a function of pumping power in Figure 13. With increasing pumping intensity, the resonant

emission due to fluorescence was observed at early stage (below 9.5 mW), and then that of Raman scattering was evolved at relatively high power pumping (above 12.5 mW). The thresholds of Nd3þ emission and Raman were 12 and 13 mW, respectively. Glass superspheres of 0.5 mol% Nd3þ-doped 15.5Na2Oe12.8CaOe71.7SiO2 were also prepared by the surface tension mold (StM) technique as standard samples, which was previously reported by our groups [41]. Pumping experiments of the Nd3þ-doped glass superspheres were carried out using a half-polished fiber coupler. Emission spectra in the 1060 nm wavelength region are shown in Figure 14. Laser emission due to well-known Nd3þ active ion was observed at three wavelengths originated from WGMs. The threshold intensities were estimated as 6.5 mW (1060.0 nm wavelength), 9 mW (1062.4 nm), and 12 mW (1063.3 nm). Additional pumping experiments, direct laser pumping in free space of Nd3þdoped glass superspheres was carried out, resulting in threshold intensities of 40e100 mW [41]. Many authors have investigated spherical resonators with WGMs. Relatively early investigations from theoretical and experimental views were published in the literatures [8,42]. Furthermore, transmission characteristics of optical fibers including taper fibers have also investigated theoretically [43]. Silica glass microspheres were successfully pumped by silica-glass fiber tapers and showed high-Q factors (quality factors) >109, where the matching between WGMs of microspheres and transmission modes of fiber-tapers was inevitable for laser

410

FIGURE 14 Laser emission spectra of 0.5 mol% Nd3þ-doped soda-lime silica glass spheres at around 1060 nm wavelength.

performance [44]. Nowadays, [fiber tapers]/[silica-glass spheres] system was well established experimentally in a laboratory scale and theoretical background was understood satisfactory. As pointed out in the literature [45], however, from the theoretical point of views, silica glass microspheres and optical coupling devices such as a fiber taper should have nearly equal refractive index for satisfying the mode matching. Moreover, fiber tapers, unclad and unsupported waveguides, are very fragile [44]. Therefore, if we have plans to open up a way to use the wide possibilities of glass microspheres of various compositions, development of the pumping techniques that is conventional and applicable for commercial uses is an urgent issue. Similarly, tailoring the glass composition is requested for obtaining desirable physical and chemical properties from their wide-spreading properties. Our works on the high-index glass spheres aim at giving one of the answers to the issues: two pumping techniques are applicable to the practical uses and the multicomponent high-index glass is chosen to explore the possibilities. In terrace microspheres, for obtaining high resonant emission, refractive index difference between the sphere and the terrace portion (nsphere/nterrace ¼ nr > 1.3) is inevitable [45]. The pumping experiment of high-index spheres using a half-polisher fiber coupler showed the improvement of optical coupling by

Handbook of Advanced Ceramics

high-refractive-index film (nD ¼ 1.73) coating on the polished core [11]. These results were not explained by the above-mentioned theoretical consideration of the [fiber tapers]/[silica-glass spheres] system. Theoretical support for our new experimental techniques is the next important issue to make advance in practical applications. To overview the efficiency of various resonators discussed in these sections, thresholds in various resonators such as fiber lasers, waveguide lasers, and micrometer-size spherical lasers are summarized in Figure 15. Designation (R) means the thresholds of Raman laser action, and (Nd3þ) means the thresholds of Nd3þ-doped waveguide and spherical lasers. The data in direct pumping (Nd3þ), fiber couplers (R and Nd3þ), and terrace (R) were obtained in my laboratory. An arrow in the figure shows the historical progress of small-size resonators along with optical coupling techniques, resulting in lowering the thresholds of laser emission. Historically speaking, there were a number of earlier contributions of Nd3þ-diped fiber lasers during the 1960se1970s [18,46], and the 1980s is the era of the rapid rise of Er3þ-doped fiber lasers and fiber amplifiers [47]. Raman lasers were also investigated in the 1970s, but the gain of Raman is smaller than that of rare-earth ion-doped fibers [48]; 1e10 W power was needed for laser action at an early stage. Waveguide-type lasers made from Nd3þ and Er3þ-doped glasses have been investigated since the early1970s; several to several tens of mW power was

FIGURE 15 Thresholds of laser emission in various small-sized resonators: fiber lasers (R); [4,5] waveguide lasers (Nd3þ); [7] glass-sphere lasers pumped by direct pumping [41], half-polished fiber coupler (R and Nd3þ) [11], terrace (R) [9,45] and taper fibers (R) [32]. (R) and (Nd3þ) represent stimulated Raman emission and laser emission due to Nd3þ. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 5.3

Optical Resonators and Amplifiers: Fiber, Waveguide, and Spherical Lasers

necessary for laser demonstration [7] and is not still so widely used in the industrial-scale applications. Investigation of spherical lasers is in the midway from the stage of ‘conventional fiber and waveguide lasers’ to the ultimate stage of ‘[High-Q silica glass spheres]/[a taper fiber]’. Theoretically and experimentally, high-Q silica glass spheres pumped by taper fiber showed microwatt (mW)level threshold for Raman laser action [32]. Such a microcavity Raman laser is highly attractive for extending the available wavelength range of conventional laser sources. Therefore, the fragile situation of taper fibers should be improved or replaced by other optical coupling techniques to achieve the commercial-scale applicability.

4. CONCLUDING REMARKS In this review, I have tried to cover briefly the development of fiber lasers, amplifier, and spherical lasers as small-sized resonators. As the spherical cavity glass lasers are one of the promising optical devices for the next optical system, I described them in detail with including the data in my laboratory. Although the fiber lasers and amplifiers have been developed rapidly, the spherical cavity lasers are still under investigation to overcome the optical coupling issues to satisfy both the high efficiency and the sufficient mechanical strength for the practical applications. We described two coupling techniques developed in our laboratory for pumping multicomponent glass spheres. The techniques show the possibility of milliwatt order thresholds for laser demonstration, along with multiwavelength emissions.

REFERENCES [1] Agrawal G. Nonlinear fiber optics, Chap 12, Fiber lasers. Academic Press; 1995. [2] Mears RJ, Reekie L, Jauncey IM, Payne DN. Low-noise erbiumdoped fibre amplifier operating at 1.54 mm. Electron Lett 1987;23:1026e8. [3] Dianov EM, Prokhorov AM. Medium-power CW Raman fiber lasers. IEEE J Selected Topics in QE 2000;6:1022e8. [4] Stolen RH, Ippn EP, Tynes AR. Raman oscillation in glass optical waveguide. Appl Phys Lett 1972;20:62e4. [5] Hill KO, Kawasaki BS, Johson DC. Low-threshold CW Raman laser. Appl Phys Lett 1976;29:181e3. [6] Saruwatari M, Izawa T. Nd-glass laser with three-dimensional optical waveguide. Appl Phys Lett 1974;24:603e5. [7] Kitagawa T, Hattori K, Shimizu M, Ohmori Y, Kobayashi M. Guided-wave laser based on erbium-doped silica planar lightwave circuit. Elect Lett 1991;27:334e5. [8] Chang RK, Campillo AJ. Optical processes in microcavities. Singapore: World Scientific; 1996. [9] Shibata S, Yano T, Segawa H. Sol-gel-derived spheres for spherical microcavity. Acc Chem Res 2007;40:913e20. [10] Uehara H, Yano T, Shibata S. Terrace-microsphere lasers: spherical cavity lasers for multi-wavelength emission. Proc SPIE 2010;7598: 75981E1e9.

411

[11] Ohkawa S, Segawa H, Yano T, Shibata S. Pumping of optical glass microspheres by optical fiber coupler. Proceeding of the 48th Symposium on Glass and Photonics Materials; 2007. p. 148e149. [12] Historical introduction of glass fibers is presented. In: Kapany NS, editor. Fiber Optics Principle and Application. Academic Press; 1967. [13] Kao KC, Hockham GA. Dielectric-fiber surface wave-guides for optical frequencies. Proc Inst Elect Eng 1966;113:1151e8. [14] Kapron FK, Keck DB, Maurer RD. Radiation losses in glass optical waveguides. App Phys Lett 1970;17:423e5. [15] French WG, MacChesney JB, O’Connor PB, Tasker GW. Optical waveguides with very low loss. Bell Syst Tech J 1974;53:951e4. [16] Izawa T, Kobayashi S, Sudo S, Hanawa F. Continuous Fabrication of high silica fiber preform. Proc. Int. Conf. Integrated Optics Optical Communication, Tokyo, Japan, 1977. [17] Miya T, Terunuma Y, Hosaka T, Miyashita T. Ultimate low-loss single-mode fiber at 1.55 mm. Electron Lett 1979;15:106e8. [18] Snitzer E. Optical maser action of Nd3þ in a barium crown glass. Phys Rev Lett 1961;7:444e6. [19] Koester CJ, Snitzer E. Amplification in a fiber laser. Appl Opt 1964;3:1182e6. [20] Po H, Cao JD, Laliberte BM, Minns RA, Robinson RF, Rockney BH, et al. High power neodymium-doped single transverse mode fiber laser. Electron Lett 1993;29:1500e1. [21] Llmpert J, Roser F, Klingebiel S, Schreiber T, Wirth C, Peschel T, et al. The rising power of fiber lasers and amplifiers. IEEE J Select Topics in QE 2007;13:537e45. [22] Nakazawa M, Suzuki K, Yamada E. Femtosecond optical pulse generation using a distributed-feedback laser diode. Electron Lett 1990;26:2038e40. [23] Stolen RH, Lin C, Shah J, Leheny F. A fiber raman ring laser. IEEE J Quant Electron QE 1978;14:860e2. [24] Kawachi M. Silica waveguides on silicon and their application to integrated-optic components. Opt Quan Elect 1990;22:391e416. [25] T Miya. Silica-based planar lightwave circuits: passive and thermally active devices. IEEE J Select Topics in QE 2000;6:38e45. [26] Suzuki S, Shuto K, Hibino Y. Integrated-optic ring resonators with two stacked layers of silica waveguide on Si. IEEE PT4 IEEE Photo Tech 1992;4:1256e528. [27] Ohishi Y, Kanamori T, Kitgawa T, Takahashi S, Snitzer E, Sigel Jr GH. Pr3þ-doped fluoride fiber amplifier operating 1.31 mm. Opt Lett 1991;16:1747e9. [28] Percival RM, Williams JR. Highly efficient 1.064 mm upconversion pumped 1.47 mm thulium doped fluoride fibre amplifier. Electron Lett 1994;30:1684e5. [29] Hansen PB, Stentz AJ, Eskilden L, Grubb SG, Strasser TA, Pedrazzani JR. High sensitivity 1.3 mm optically preamplified receiver using Raman amplification. Electron Lett 1996;32:2164e5. [30] Kani J, Jinno M, Oguhi K. Fiber Raman amplifier for 1520 nm band WDM transmission. Electron Lett 1998;34:1745e7. [31] Masuda H. Recent progress on optical fiber amplifiers and their applications. Proc SPIE 2006;6389(638902):1e10. [32] Spillane SM, Kippenberg TJ, Vahala KJ. Ultralow-threshold Raman laser using a spherical dielectric microcavity. Nature 2002;415:621e3. [33] Garrett CGB, Kaiser W, Bond WL. Stimulated emission into optical whispering modes of sphere. Phys Rev 1961;124:1807e9.

412

[34] Qian SX, Chang RK. Multiorder stokes emission from micrometersize droplets. Phys Rev Lett 1986;56:926e9. [35] Sandoghdar V, Treussart F, Hare J, Lefevire-Seguin V, Raimondm JM, Haroche S. Very low threshold whispering-gallerymode microsphere laser. Phy Rev A 1996;54:R1777e80. [36] Serpenguzel A, Arnold S, Griffel G. Excitation of resonances of microspheres on an optical fiber. Opt Lett 1995;20:654e6. [37] Knight JC, Cheung G, Jacques F, Birks TA. Phase-matched excitation of whispering-gallery-mode resonances by a fiber taper. Opt Lett 1997;22:1129e31. [38] Gorodetsky ML, Savchenkov AA, Ilchenko VS. Ultimate Q of optical microsphere resonators. Opt Lett 1996;21:453e5. [39] Saitou M, Yano T, Segawa H, Shibata S. Site-selective excitation and fluorescence of Nd3þ-doped glasses for lasing at 900 nm band. 3rd Inter. Conf. on Sci. & Tech. for Adv. Ceram., Yokohama, Japan; 2009. [40] Uehara H, Kishi T, Yano T, Shibata S. Enhancement of stimulated Raman scattering in terrace-microsphere. J. Opt. Soc. Am. B 2011; 28:2436e43. [41] Yano T, Sato T, Segawa H, Shibata S, Kishi T, Yasumori A. Q factor of super-spherical optical cavity of Nd3þ-doped soda-lime

Handbook of Advanced Ceramics

[42] [43]

[44]

[45]

[46] [47]

[48]

glass prepared by StM method. Inter. Conf. on the Physics of NonCryst. Solids 2009;0284. Barber PW, Chang RK. Optical effects associated with small particles. Singapore (: World Scientific; 1988. Bures J. Guided optics, optical fibers and all-fiber components, Chap. 7, Tapered fibers. Weinheim: Wily-VCH Verlag GmbH & Co. KGaA; 2009. Matsko AB, Ilchenko VS. Optical resonators with whisperinggallery mode-part I: basics. IEEE J Selected Topics in Quant Elec 2006;12:3e14. Shibata S, Ashida S, Segawa H, Yano T. Low-threshold spherical Raman laser; 2007. XIVth International Sol-Gel Conference, Montpellier, France. Stone J, Burrus CA. Neodymium-doped silica lasers in end-pumped fiber geometry. Appl Phys Lett 1973;23:388e9. Morishige Y, Kishida S, Washio K, Toratani H, Nakazawa M. Output stabilized high-repetition-rate 1.545-mm Q-switched Er:glass laser. Opt Lett 1984;9:147e9. Stolen RH, Ippen EP. Raman gain in glass optical waveguides. Appl Phys Lett 1973;22:276e8.

Chapter 6.1

Multi-layered Ceramic Capacitors Takaaki Tsurumi and Takuya Hoshina Graduate School of Science and Engineering, Tokyo Institute of Technology, 2-12-1 Ookayama, Meguro, Tokyo 152-8550, Japan

Chapter Outline 1. High-Capacitance MLCCs With Nickel Internal Electrodes 2. Nonlinear Dielectricity of MLCCs 3. Capacitance Aging in MLCCs 4. Size Effect of BaTiO3 Ceramics

415 417 418 421

1. HIGH-CAPACITANCE MLCCs WITH NICKEL INTERNAL ELECTRODES Chip-type capacitors are the main products of electronic ceramic components due to their superior frequency performance and volume efficiency, which meets the requirement from the advanced electronic devices. A schematic view of multilayer ceramic capacitor (MLCC) is shown in Figure 1. The capacitor is composed of many thin plate capacitors in parallel connection. The capacitance of MLCC with electrode area S, dielectric layer thickness t, and the number of dielectric layers n is given by C ¼

n3r S 3r ST ¼ 2 t t

FIGURE 1 Schematic view of MLCC. For color version of this figure, the reader is referred to the online version of this book.

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00024-1 Copyright Ó 2013 Elsevier Inc. All rights reserved.

5. Reliability of MLCCS e Lifetime in HALT 6. Computer Simulation and Further Prospective of MLCC Technology References

423 425 426

where 3r is dielectric constant and T is the chip thickness (Figure 1). The above relationship indicates that the reduction of dielectric layer thickness t is very effective to increase the volume efficiency. The increase of capacitance volumeedensity of MLCCs has been therefore realized by reducing dielectric layer thickness; the latest commercialized products have the layer thickness less than 1 mm as shown in Figure 2 [1]. Since drastic increase of price during 1995e1997, MLCC manufacturers tried to replace silverepalladium inner electrodes with nickel electrodes. As nickel is easily oxidized in air at elevated temperatures, MLCCs with nickel inner electrodes should be fired in an atmosphere with a low oxygen partial pressure. However, the problem was that properties of BaTiO3-based dielectrics were easily reduced in such sintering conditions. Figure 3 shows a famous result of Sakabe at Murata Manufact., indicating the possibility to produce BaTiO3 ceramics with anti-reducing properties [2,3]. Resistivity r of BaTiO3 ceramics sintered in reducing atmosphere (oxygen partial pressure of 2  10 12 MPa) was not degraded by doping CaO when A/B ratio m was over unity. Essential points in the composition designing are 1) control of the molecular ratio (Ba þ Ca)O/(Ti Zr)O2 larger than unity, 2) incorporation of Ca ions into Ti lattice site (B site) of BaTiO3, and 3) doping acceptor as Mn ions. Mg ions have the same effect as Ca ions. The formation of oxygen vacancies is described as 1 0 OO /V,, O þ O2 þ 2e ; 2

1

2 2 K ¼ ½V,, O Šn PðO2 Þ

415

416

Handbook of Advanced Ceramics

FIGURE 2 Dielectric layers and Ni internal electrodes in MLCCs [1]. For color version of this figure, the reader is referred to the online version of this book.

where K is the equilibrium constant, and the electron concentration is 1

n ¼

K2 1

1

2 4 ½V,, O Š PðO2 Þ

The substitution of Mg ions to Ti ions can be described as MgO/OO þ Mg00Ti þ V,, O

FIGURE 3 Resistivity of [(Ba1 xCax)]mTiO3 ceramics sintered in P(O2) ¼ 2  10 12 MPa at 1400  C for 2 h.

(a)

From the above relationships, Mg ion substitution to Ti sites increases the oxygen vacancy concentration ½V,, O Š, resulting in the decrease of electron concentration n. Because of the decrease of charge carrier, the resistivity of BaTiO3 ceramics does not degrade as shown in Figure 3, enabling us to use inexpensive nickel internal electrodes in MLCCs. In modern electronic circuits, the multilayer ceramic capacitor (MLCC) is an indispensable passive component for manufacturing miniaturized electronic circuits. The production of MLCC increases every year, especially highcapacitance MLCCs are gradually replacing the market of tantalum capacitors. Most of high-capacitance MLCCs use Ni internal electrodes and satisfy the EIA X7R specifications, where dielectric layers have the ‘coreeshell’ structure (Figure 4) , in which one grain of BT-based ceramics consists of inner core of almost pure BT and outer shell of BT with

(b)

FIGURE 4 Transmission electron microphotographs of dielectric layers in X7R-MLCCs.

Chapter | 6.1

Multi-layered Ceramic Capacitors

417

FIGURE 5 AC-field dependence of permittivity of BZT ceramics. For color version of this figure, the reader is referred to the online version of this book.

additives [4e7]. However, several problems such as nonlinear dielectricity, capacitance aging, size effect, and reliability have arisen to prevent further increase of the capacitance density.

2. NONLINEAR DIELECTRICITY OF MLCCs Nonlinear dielectricity means the electric field dependence of permittivity. It becomes important in high-capacitance MLCCs because the electric field applied on one dielectric layer increases with decreasing layer thickness. Nonlinear dielectricity causes capacitance degradation under high DC-bias voltage and temperature stability degradation under high AC voltage. Nonlinear dielectric responses of high-capacitance MLCCs are attributable to the ferroelectricity of BaTiO3 [8]. AC-field dependence of permittivity of (Ba0.79Zr0.21) TiO3 (BZT), dielectrics in MLCCs with EIA Y5V specification, is shown in Figure 5. The permittivity exhibits a maximum at a certain electric field. This behavior can be explained by the model of relaxors; the increase of permittivity is due to the enhancement of dipole polarization of thermally fluctuating dipoles in polar nano-regions (PNRs) by the application of high fields. This enhancement saturates at a certain field, giving rise to the maximum of permittivity [9, 10]. The temperature variation of permittivity of BZT measured with different AC fields is shown in Figure 6. As the AC field increases, the temperature at permittivity peak (Tm) decreases and the peak becomes

asymmetric. In relaxors, it is known that no phase transition occurs at Tm. The decrease of permittivity at low temperatures is caused by the freezing of thermally fluctuating dipoles. Those freezing dipoles were forced to move by high AC fields, contributing dipole polarization; therefore, permittivity measured by high AC field does not decrease even at low temperatures. Thus, the shift of Tm and the peak broadening to low temperature side with increasing AC field can be explained. To elucidate the dielectric properties of the core and the shell parts in X7R-MLCCs, we have prepared three types of MLCCs with only core grains, only shell grains, and coreeshell grains. The composition of core grains was BaTiO3 while that of shell grains was (Ba0.95Ho0.05) (Ti0.975Mg0.015Mn0.01)O3. Figure 7 shows dielectric properties of the three kinds of MLCCs measured with various AC fields [7]. Comparison between Figure 7(a) and (c) revealed that the dielectric properties of X7R-MLCCs were dominated by the core part. The formation of coree shell structure depressed the AC-field dependence of permittivity at low temperatures. Relatively flat temperature dependence was observed for MLCCs with only core grains (Figure 7(a)), although the core grains consisted of pure BaTiO3 which shows a sharp permittivity peak at 130  C. This is because the ferroelectricity of BaTiO3 was depressed in core grains of 200e300 nm by the size effect. On the other hand, the dielectric property of shell parts (Figure 7(b)) is similar to that of BZT (Figure 6), indicating that the dielectric response of the shell grains is explained by the mechanism of relaxors. Consequently, the dielectric layers of X7R-MLCCs consist of BaTiO3 core parts

418

Handbook of Advanced Ceramics

FIGURE 6 Temperature dependence of permittivity of BZT ceramics measured with different AC fields. For color version of this figure, the reader is referred to the online version of this book.

showing depressed ferroelectricity by the size effect and the shell parts of BaTiO3-based solid solutions showing relaxor-like behaviors. Grains in the dielectric layers vary in the dopant distribution at the shell parts, giving rise to the distribution of Tm in Figure 7(b). The role of shell parts is to increase the permittivity at low temperatures; the shell parts contribute the temperature stability of MLCCs. However, in the case of ultra-thin dielectric layers, the electric field applied to the dielectric layers increases and the temperature stability of permittivity will be achieved without the shell parts. It should be also noted that the permittivity of the shell parts is lower than that of the core parts (Figures 7(a) and (b)), which means that the shell parts degrade the capacitance of MLCCs.

3. CAPACITANCE AGING IN MLCCS Capacitance aging is a phenomenon wherein the capacitance of MLCCs somehow decreases with time under DCbias voltages at relatively high temperatures [6,11]. The capacitance-aging problem of BaTiO3 has been studied for a long time [12e14] but these studies mostly dealt with aging without DC fields. MLCCs with high capacitances are mainly used as bypass capacitors in power lines of electric circuits to stabilize the DC voltage of the lines, which means that the capacitors are used under DC fields for long durations. One of the most important applications of high-capacitance MLCCs is bypass capacitors connecting to power lines of CPU. These MLCCs are used under DC-bias voltages at relatively high temperatures. The decrease of capacitance sometimes makes it difficult to

supply enough current to CPU, which is not predicted by electronic circuit engineers. Figure 8 shows a typical capacitance-aging curve of X7R-MLCCs measured under a DC electric field. dC (%) is defined as dC ¼ (C C0)/C0  100, where C0 is the capacitance measured without a DC field at the temperature shown in the figure. It is possible to see that the reduction of the capacitance proceeds in two stages as pointed out by Nomura et al. [11]. The decrease of the capacitance in the first stage takes place immediately after the application of a DC field. The first stage is therefore attributable to the nonlinear dielectric permittivity of dielectric materials and it should not be included in the capacitance-aging phenomenon. The second stage starts after the first stage and continues up to 105 s and more. The dC in the second stage depends on temperature, indicating that a thermal activation process is involved in the aging phenomenon. Figure 9 shows the change in the aging behavior with MnO content (x) into the dielectric with the composition of 100BaTiO3e0.5Ho2O3e0.5MgOexMnOe1.5BaSiO3. The relaxation time of the aging obviously decreased with increasing MnO content and the dC of MLCCs with different Mn contents tended to saturate to a certain value after a long aging time. The dC after a long aging time is independent of the Mn content, which indicates that the internal field or space charges are less related with the aging phenomenon. The aging behavior measured at different temperatures is shown in Figure 10. The relaxation time was reduced with temperature, indicating that a thermally activated process is involved in the aging phenomenon. It must be

Chapter | 6.1

Multi-layered Ceramic Capacitors

419

FIGURE 8 Typical examples of capacitance aging in X7R-MLCCs measured at room temperature and at 110  C. For color version of this figure, the reader is referred to the online version of this book.

FIGURE 9 Effect of MnO addition on capacitance aging at 110  C. For color version of this figure, the reader is referred to the online version of this book.

FIGURE 7 Temperature dependence of permittivity of MLCCs with (a) core, (b) shell, and (c) coreeshell grains.

noted that the capacitance aging was markedly depressed at 150  C, which is above the Curie temperature of BaTiO3, because ferroelectric domains deeply concern with the aging phenomenon. In Figure 11, the aging behavior of MLCCs fired in reducing or air atmosphere was compared. The aging in the second stage is markedly depressed in the air-fired specimen. It is also notable that the dC after the aging is almost consistent for the two specimens. In other words, the aging in the second stage could be moved to the first stage by firing in air atmosphere. The Mn ions in

FIGURE 10 Capacitance aging measured at different temperatures.

the core of the reduced specimen have a trivalent (Mn3þ) or divalent (Mn2þ) state, which clamp the domain-wall motion. It is well known that Mn is an effective additive for making the ‘hard’ PZT ceramics, in which the domain-wall motion is clamped by the defect dipoles associated with

420

FIGURE 11 Effect of firing condition on the capacitance aging. For color version of this figure, the reader is referred to the online version of this book.

Mn3þ and/or Mn2þ substituted with B-site ions in the perovskite structure [15,16]. In the MLCCs fired in air atmosphere, Mn3þ and/or Mn2þ are possibly changed to Mn4þ, which is not effective in clamping the domain-wall motion. This change results in a relatively free movement of domain walls by the first application of a DC field; therefore, the dC in the second stage moved to the first stage in the air-fired specimen. The dC after the aging was consistent for the two specimens because the firing

Handbook of Advanced Ceramics

conditions did not affect considerably the nonlinear permittivity and domain structure of MLCCs. The capacitance aging in the second stage is caused by the 90 domain switching in BT in the core. The mechanism of the capacitance aging in X7R-MLCCs is shown in Figure 12. Most of the domain walls in the core are clamped by the defect dipoles formed by negatively charged Mn sites and positively charged oxygen vacancies. The movement of unclamped domain walls may be involved in the dC in the first stage, although the dC in the first stage is due mainly to the nonlinear dielectric permittivity of the relaxor-like shell. In the second stage, the continuous application of DC fields moves the electrons on Mn sites to eliminate the defect dipoles with time. This enhances the domain-wall motion to align the c-axis of BT to the electric field direction, resulting in the dielectric permittivity due to the permittivity difference between 333 and 311. The movement of electrons is a thermally activated hopping process. Because the increase in Mn sites in the core enhances the hopping of electrons, relaxation time is reduced with increasing MnO content (x) as shown in Figure 9. The difference between 333 and 311 is almost independent of Mn content; so, the dC after the aging tends to saturate to a certain value as shown in Figure 9. Figure 13 shows the change in the aging behavior with the DC field. It should be noted that the first stage was

FIGURE 12 Mechanism of capacitance aging. Each hexagon shows one grain in the dielectric layer. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 6.1

Multi-layered Ceramic Capacitors

421

FIGURE 13 Capacitance aging curves of specimen with x ¼ 0.8 at 110  C measured under different DC fields. The effects of the nonlinear permittivity and the domain switching on the DC are shown as arrows.

enhanced but the second stage was depressed to disappear when a high DC field was applied. This is because a high DC field moves the domain walls quickly in the first stage in spite of the clamping of domain walls by defect dipoles. The same phenomenon is normally observed in ferroelectrics when an applied DC field exceeds the coercive field. The difference in dC after the aging is purely due to the effect of nonlinear permittivity because the domain switching is completed after the second stage. Figure 14 shows capacitance (C) vs voltage (V) and electric displacement (D) vs electric field (E) curves measured for the specimen in Figure 13. The clear hysteresis observed in the DeE curve indicates that domain switching definitely takes place in the dielectrics. By assuming that the difference in dC after the aging in Figure 13 was caused only by the nonlinear permittivity, it was possible to separate the effect of the nonlinear permittivity and that of the domain switching from the experimental results. The two additional curves in Figure 14 (b) show both effects. The DeE curve observed could be explained by the sum of the nonlinear permittivity and domain switching. The effects of the nonlinear permittivity and domain switching were also separated and are shown in Figure 13. It is important that the domain switching effect on the dC is consistent for all curves in the figure and the dC due to the nonlinear permittivity increases with DC field.

FIGURE 14 (a) Capacitance vs voltage curve and (b) electric displacement vs electric field curve of specimen in Figure 12. For color version of this figure, the reader is referred to the online version of this book.

depend on internal stress and impurities. The two-step sintering method [18e20] is effective to prepare finegrained BT ceramics without internal stress or impurity. Figure 15 shows the grain size dependence of permittivity for the BT ceramics prepared using conventional and

4. SIZE EFFECT OF BATIO3 CERAMICS It is now recognized that the permittivity shows a maximum at about 1 mm, i.e., the permittivity of BT ceramics increases with decreasing grain size in the micron level while that decreases in the submicron level. Arlt et al. [17] proposed that the 90 domain walls were the origin of the enhancement of permittivity. To study the intrinsic size effect, it is important to prepare fine-grained BaTiO3 ceramics without internal stresses or impurities because the crystal structure and the dielectric property of BT ceramics

FIGURE 15 Grain size dependence of permittivity of BT ceramics. For color version of this figure, the reader is referred to the online version of this book.

422

Handbook of Advanced Ceramics

(a)

(b)

FIGURE 16

(c)

90 domain configurations in BaTiO3 ceramics with different grain sizes.

FIGURE 18 Grain size dependence of permittivity determined by dipole and ionic polarizations. For color version of this figure, the reader is referred to the online version of this book.

FIGURE 17 Wideband dielectric spectra of BT ceramics with average grain sizes of (a) 13 and (b) 1.4 mm. For color version of this figure, the reader is referred to the online version of this book.

two-step sintering methods [21]. The permittivity of pure BaTiO3 ceramics increases with decreasing grain size when the grain size is above 1.1 mm, whereas it decreases when the grain size is below 1 mm. The maximum permittivity of the BT ceramics is 8000 for a grain size of 1.1 mm. Figure 16 shows scanning electron microscope (SEM) images of the BT sample surface after chemical etching with various grain sizes [21]. The observations reveal that most of the grains had simple lamellar structures of 90 domains. The 90 domain width decreases with decreasing grain size. To understand the domain contribution to the permittivity, the measurement of ultrawide range dielectric

spectra of the BaTiO3 ceramics is very effective [22, 23]. Figure 17 shows the dielectric spectra of the BaTiO3 ceramics with various grain sizes. In this figure, the Debyetype dielectric relaxations observed below 109 Hz are attributable to the dipole polarization due to the domainwall vibration induced by the electric fields. On the other hand, the dielectric dispersions observed over 1010 Hz are attributable to ionic polarization. Figure 18 shows the grain size dependence of the permittivity due to dipole or ionic polarization. For grain sizes over 1 mm, the dipole polarizability due to domain-wall vibration increases with decreasing grain size. It indicates that a high 90 domainwall density enhanced the dipole polarizability due to the domain-wall vibration. Moreover, the ionic polarizability also increases with decreasing grain size. The 90 domain structure induces the lattice strain into the BaTiO3 lattices around the domain wall. It is believed that the lattice strain region around the 90 domain wall has high ionic polarizability [24, 25]. Therefore, high domain-wall density enhances the ionic polarizability as well as the dipole polarizability. The increase in permittivity with decreasing grain size to 1 mm is due to the 90 domain contribution. In contrast, for grain sizes below 1 mm, the dipole polarizability and the ionic polarizability decrease with decreasing grain size. It suggests that the domain contribution

Chapter | 6.1

Multi-layered Ceramic Capacitors

FIGURE 19 Grain size dependence of permittivity in BT-based dielectrics in MLCCs. For color version of this figure, the reader is referred to the online version of this book.

decreases when grains became smaller than 1 mm. However, it is still difficult to explain the mechanism of permittivity decrease below grain sizes in the submicron range. For further studies, it is necessary to clarify theoretically the weakening mechanism of dipole and ionic polarizability for the fine-grained BaTiO3 ceramics with sizes below 1 mm. Grain size dependence of the permittivity of dielectric layers in X7R-MLCCs (Taiyo Yuden, Co., Japan) is shown in Figure 19 [26]. The permittivity simply decreases with decreasing grain size. To understand the size effect of BaTiO3-based dielectric layers in X7R-MLCCs, it should be recalled that dielectric layers of the MLCCs consist of the “coreeshell” grains (Figure 4). The coreeshell structure is a complex structure in one grain, where the shell part is heavy-doped BaTiO3 showing relaxor-like behavior, while the core part is almost pure BaTiO3 showing depressed ferroelectric behavior due to the size effect of BaTiO3. Figure 20 shows the permittivity of dielectric layers with different grain sizes obtained by the SPLINE analysis of experimental data taken for MLCCs with different grain sizes [25]. It should be noted that the temperature stability of the permittivity was markedly degraded when the grain size became smaller than 120 nm. In the MLCCs with the coreeshell structure, the shell dominated the permittivity at low temperatures, while the core dominated the permittivity at high temperatures within the X7R specification. The degradation of temperature stability observed in small grains indicated that the portion of shell part increased in small grains. The shell parts are formed by controlling the diffusion of additive ions from the grain boundary, which means that the formation of thin shell parts becomes difficult as the grain size becomes small, giving rise to the increase of the shell part in small grains. In the size effect of dielectric layers in MLCCs, the coreeshell structure plays an important role.

423

FIGURE 20 Relative permittivity of dielectric layers with different grain sizes in MLCCs. For color version of this figure, the reader is referred to the online version of this book.

5. RELIABILITY OF MLCCS e LIFETIME IN HALT The increase of capacitance density of MLCCs has been realized by the reduction of dielectric layer thickness but enough reliability (a long lifetime in HALT) has been another criterion to be achieved in the development of MLCCs. It is worth to describe the difference in breakdown, capacitance aging, and lifetime in the HALT. A typical relation between DC voltage and leakage current of X7RMLCCs is shown in Figure 21 (a) [27]. The application of a high voltage shown by X-point brings about insulation break of MLCCs, which is called breakdown and the voltage of X-point is called breakdown voltage. By applying a DC voltage far below the breakdown voltage, the capacitance of MLCCs decreases with time down to a few tens of percents, which is called capacitance aging caused by the domainwall motion in the core parts as mentioned in the section of capacitance aging. Figure 21 (b) shows the leakage current variation with time in the HALT of MLCCs where a DC voltage below the breakdown voltage is applied at relatively high temperatures of 120e200  C. The leakage current gradually increases with time, and, at a certain time, the current increases steeply to the breakdown of MLCCs. This time is called a lifetime in the HALT. The breakdown in the HALT is caused by the accumulation of oxygen vacancies at the cathode. The model of oxygen vacancy distribution before and after the HALT is shown in Figure 22 [28]. Plenty of oxygen vacancies are formed in the core parts of almost pure BaTiO3 after sintering in reducing atmosphere. However, the number of oxygen vacancies in the shell parts should be smaller than those of the core parts because some additives in the shell parts work as anti-reducing agents. Positively charged oxygen vacancies ðV,, O Þ migrating to the cathode under a DC field are blocked at the shell parts and

424

Handbook of Advanced Ceramics

FIGURE 21 (a) A typical relation between DC voltage and leakage current of X7R-MLCCs and (b) the leakage current variation with time in the HALT of MLCCs [27]. For color version of this figure, the reader is referred to the online version of this book.

FIGURE 22 Distribution of oxygen vacancies in the coreeshell structure before and after HALT [28]. For color version of this figure, the reader is referred to the online version of this book.

grain boundaries. In this sense, grain boundaries and the shell parts contribute to the reliability of X7R-MLCCs. An interesting question is the role of rare earth elements such as Dy, Ho, and Y to improve the lifetime in the HALT. The relation between Ho content and the lifetime in HALT is shown in Figure 23. The lifetime rapidly increases from the Ho content of 0.2 atom% and tends to saturate with increasing Ho content. Analysis of local Ho distribution indicated that Ho segregated at the shell parts and grain boundaries; similar results were obtained for Dy and Y. Thus, rare earth elements such as Dy, Ho, and Y change the properties of shell parts and grain boundaries to depress the migration of oxygen vacancies. These rare earth elements are incorporated into the A-sites of perovskite structure as a donor when the content is low. The addition of donor additives decrease the number of oxygen vacancies and act as a block of oxygen vacancy migration. The similar

FIGURE 23 Relation between lifetime in HALT and Ho content in the dielectrics of X7R-MLCCs [27].

Chapter | 6.1

Multi-layered Ceramic Capacitors

425

FIGURE 25 MLCCs.

FIGURE 24 Current vs electric field curve of MLCCs with different amount of Ho. For color version of this figure, the reader is referred to the online version of this book.

situation is valid for grain boundaries but the change of electric properties at grain boundary is more effective for the improvement of lifetime in HALT. Figure 24 shows the relation between DC voltage and leakage current of X7R-MLCCs added with different amounts of Ho. Solid circles in the figure indicate that the leakage current was not stable but increased to breakdown during measurements. MLCCs added with 1.5 mol% of Ho stably passed a relatively large current under high field, which is possibly due to a tunneling current. The resistance of core parts, shell parts, grain boundaries, and electrode interfaces was determined by the equivalent circuit analysis of MLCCs as shown in Figure 25. The electrode interface has the highest resistance and the grain boundaries have the second highest resistance. As the resistance of electrode interfaces should be independent of Ho contents, the resistance change, giving rise to the change of leakage current shown in Figure 23, is due to the grain boundaries, which means that the addition of Ho decreases the resistance of grain boundaries to stably pass the current under DC voltage. The resistance of grain boundaries is higher than that of core parts and shell parts. The DC voltage applied to MLCCs is divided to core parts and shell parts according to their resistance. If the resistance of grain boundaries is

The relation between resistance and voltage for each part in

extremely high, DC voltage is mainly applied to grain boundaries, giving rise to the enhancement of oxygen vacancy migrations across grain boundaries followed by the breakdown of grain boundaries by heat generation. The enhancement of oxygen vacancy migration across the grain boundary leads to the breakdown of the whole device by the accumulation of oxygen vacancies at the electrode interfaces. On the contrary, if grain boundaries stably pass a relatively high current, the oxygen migration across grain boundaries is depressed because the DC voltage does concentrate to grain boundaries, giving rise to the improvement of the lifetime in HALT. The contribution of Dy, Ho, and Y to the improvement of the lifetime is interpreted as follows: 1) these rare earth elements are possibly incorporated into both A-sites at the shell parts and decrease the oxygen vacancy concentration to depress the migration of oxygen vacancies, 2) rare earth elements incorporated into grain boundaries decrease the resistance of grain boundaries to avoid the voltage concentration leading to the breakdown at grain boundaries, and 3) the decrease of DC voltage applied on the grain boundaries depresses the oxygen vacancy migration across the grain boundaries and their accumulation at electrode interfaces.

6. COMPUTER SIMULATION AND FURTHER PROSPECTIVE OF MLCC TECHNOLOGY The reduction of dielectric layer thickness demands the simultaneous reduction of BaTiO3 grain size to keep the several grain boundaries in one dielectric layer. However, the dielectric permittivity reduced with BaTiO3 grain size because of the size effect. Consequently, there must be a limit of capacitance density in the current technologies of MLCC. Computer simulation software (MLCC-simulator)

426

FIGURE 26 Predicted property of 47 mF MLCC with 1608 chip size. For color version of this figure, the reader is referred to the online version of this book.

was developed to optimize the MLCC structure and to predict the limit of capacitance density of MLCCs [25]. The highest capacitance density predicted by the simulation was 47 mF in 1608 mm chip size where the electrode thickness was 0.4 mm and the BaTiO3 grain size was 93 nm. Figure 26 shows the dielectric properties of this MLCC. The temperature stability was markedly degraded when grains smaller than 120 nm were used. The third simulation seemed to be close to the technical limit of the current technology. The results of simulations gave the following important information: 1) there is no advantage to use BaTiO3 grains smaller than 80 nm, 2) the reduction of Ni internal electrode thickness is effective to obtain high capacitance, 3) enough reliability should be obtained with the small number of grains in one dielectric layer, 4) the reduction of dielectric layer thickness is effective to obtain 22e47 mF in 1608 chip size, and 5) the limit of capacitance in 1608 chip size is about 47 mF as long as we use the current technology. Finally, we have to stress that the technical limit was predicted by assuming there is no progress of the current technology. The fact that several grain boundaries are necessary to achieve a long lifetime in HALT indicates that the grain size decreases with decreasing dielectric layer thickness, giving rise to the degradation of permittivity of dielectric layers. The simulation indicated that the highest capacitance achieved by the current technology is 47 mF in 1608 chip size, which is close to the density of latest products. This limit is mainly determined by the number of grains needed in one dielectric layer and the thickness of Ni internal electrodes. We will discuss on the issues concerning dielectrics. Microstructure control of dielectrics is the key issue to increase the capacitance density. The ideal structure is that

Handbook of Advanced Ceramics

the shell part is as thin as possible to obtain the high permittivity, and the distribution of additives at grain boundaries is precisely controlled to achieve the high reliability. To realize them, one of the most important issues is the selection of raw BaTiO3 powders. Hoshina et al. [29] analyzed the internal structure of BaTiO3 nanoparticles and revealed that the BT nanoparticles consisted of three regions: surface cubic layers, internal tetragonal layers, and gradient lattice strain layers (GLSLs) between them. The c/a ratio determined by the XRD analysis decreases with increasing thickness of cubic layers, which means that the c/a ratio is the index of defect concentration in BaTiO3 powders. In the formation of coreeshell structures, additives diffuse from the surface to the inner part of BaTiO3 particles. The diffusion should be as slow as possible to form the coreeshell structure in fine BaTiO3 grains but the defects in crystals usually increase the diffusion constant. It is therefore difficult to form the coreeshell structure using BT powders with low c/a ratio. The collapse of coreeshell structure degrades the permittivity and its temperature stability. The size effect is an intrinsic effect of BT and it may be difficult to completely overcome it as long as BT-based ceramics are used in dielectric layers. We cannot expect to double the capacitance if we decrease the dielectric layer thickness to half any more. The increase of capacitance density should be preceded by the microstructure control which implies the formation of coreeshell structure with thin shells and proper distribution of additives at grain boundaries. The formation of ideal coreeshell structure is achieved by the selection of BT powders with fewer defects as well as the optimization of temperature profiles in the sintering.

REFERENCES [1] Photograph was provided by Taiyo Yuden Co. ltd., Japan. [2] Sakabe Y. Calcium-Doped Barium Titanate Ceramics for Nickel Electrode Multilayer Capacitors. Ph.D thesis. Kyoto Univ; 2002. [3] Sakabe Y. Dielectric materials for base-metal multilayer ceramic capacitors. Ceram Bull 1987;66:1338e41. [4] Kishi H, Mozuno Y, Chazono H. Base-metal electrode-multilayer ceramic capacitors: past, present and future perspectives. Jpn J Appl Phys 2003;2:1e15. [5] Mizuno Y, Hagiwara T, Kisshi H. Microstructural design of dielectrics for Ni-MLCC with ultra-thin active layers. J Ceram Soc Jpn 2007;115:360e4. [6] Tsurumi T, Shono M, Kakemoto H, Wada S, Saito K, Chazono H. Mechanism of capacitance aging under DC electric fields in multilayer ceramic capacitors with X7R characteristics. Jpn J Appl Phys 2005;44:6989e94. [7] Tsurumi T, Adachi H, Kakemoto H, Wada S, Mizuno Y, Chazono H, et al. Dielectric properties of BaTiO3-based ceramics under high electric field. Jpn J Appl Phys 2002;41:6929e33.

Chapter | 6.1

Multi-layered Ceramic Capacitors

[8] Tsurumi T. Non-linear Piezoelectric and Dielectric Behaviors in Perovskite Ferroelectrics. J Ceram Soc Jpn 2007;115:17e22. [9] Tsurumi T, Yamamoto Y, Kakemoto H, Wada S, Chazono H, Kishi H. Dielectric properties of BaTiO3eBaZrO3 ceramics under a high electric field. J Mater Res 2002;17:755e9. [10] Tsurumi T, Soejima K, Kamiya T, Daimon M. Mechanism of diffuse phase transition in relaxor ferroelectrics. Jpn J Appl Phys 1994;33:1959e64. [11] Nomura T, Kawano N, Yamamatsu J, Arashi T, Nakano Y, Sato A. Aging behavior of Ni-electrode multilayer ceramic capacitors with X7R characteristics. Jpn J Appl Phys 1995;34:5389e95. [12] Mason WP. Aging of the properties of barium titanate and related ferroelectric ceramics. J Acoust Soc Am 1955;27:173e85. [13] Bradt RC, Ansell GS. Aging in tetragonal ferroelectric barium titanate. J Am Ceram Soc 1969;52:192e9. [14] Arlt G, Dederiches H. Complex elastic, dielectricd and piezoelectric constants by domain wall damping in. ferroelectric ceramics. Ferroelectrics 1980;29:47e50. [15] Tsurumi T, Kil Y-B, Nagatoh K, Kakemoto H, Wada S, Takahashi S. Intrinsic elastic, dielectric, and piezoelectric losses in lead zirconate titanate ceramics determined by an immittancefitting method. J Am Ceram Soc 2002;85:1993e6. [16] Robels U, Schneider-Stro¨rmann L, Arlt G. Domain wall trapping as a result of internal bias fields ferroelectrics. Ferroelectrics 1992;133:223e8. [17] Arlt G, Hennings D, De With G. Dielectric properties of finegrained barium titanate ceramics. J Appl Phys 1985;58:1619e25. [18] Chen I-W, Wang X-H. Sintering dense nanocrystalline ceramics without final-stage grain growth. Nature 2000;404:168e71. [19] Polotai A, Breece K, Dickey E, Randall C, Ragulya A. A novel approach to sintering nanocrystalline barium titanate ceramics. J Am Ceram Soc 2005;88:3008e12.

427

[20]

[21]

[22]

[23]

[24]

[25]

[26]

[27]

[28]

[29]

Wang X-H, Deng X-Y, Bai H-L, Zhou H, Qu W-G, Li L-T, Chen I-W. Two-step sintering of ceramics with constant grain-size, II: BaTiO3 and NieCueZn ferrite. J Am Ceram Soc 2006;89:438e43. Hoshina T, Kigoshi Y, Hatta S, Takeda H, Tsurumi T. Domain contribution to dielectric properties of fine-grained BaTiO3 ceramics. Jpn J Appl Phys 2009;48:09KC01. Tsurumi T, Li J, Hoshina T, Kakemoto H, Nakada M, Akedo J. Ultrawide range dielectric spectroscopy of BaTiO3-based perovskite dielectrics. Appl Phys Lett 2007;91:182905. Teranishi T, Hoshina T, Tsurumi T. Wide range dielectric spectroscopy on perovskite dielectrics. Mater. Sci Engineer B 2007;161:55e60. Hoshina T, Takizawa K, Li J, Kasama T, Kakemoto H, Tsurumi T. Domain size effect on dielectric properties of barium titanate ceramics. Jpn J Appl Phys 2008;47:7607e11. Tsurumi T, Hoshina T, Takeda H, Mizuno Y, Chazono H. Size effect of barium titanate and computer-aided design of multilayered ceramic capacitors. IEEE Trans Ultrason Ferroelectr Freq Contr 2009;56:1513e22. Mizuno Y, Hagiwara T, Kishi H. Microstructural design of dielectrics for Ni-MLCC with ultra thin active layers. J Ceram Soc Jpn 2007;115:360e4. Chazono H. Structure-Property Relationship in BaTiO3-Based Dielectrics for Multiilayer Ceramic Capacitor. Ph.D thesis. Keio Univ; 2004. Chazono H, Kishi H. dc-Electrical degradation of the BT-based material for multilayer ceramic capacitor with Ni internal electrode: impedance analysis and microstructure. Jpn J Appl Phys 2001;40:5624e9. Hoshina T, Wada S, Kuroiwa Y, Tsurumi T. Composite structure and size effect of barium titanate nanoparticles. Appl Phys Lett 2008;93:192914.

Chapter 6.2

Lead-Free Piezoelectric Ceramics Tadashi Takenaka Faculty of Science and Technology, Tokyo University of Science, 2641 Yamazaki, Noda, Chiba 278-8510, Japan.

Chapter Outline 1. Introduction 2. Perovskite-structured Piezoelectric Ceramics

429 430

3. Summary References

443 444

1. INTRODUCTION Piezoelectric materials play an important role for electronic devices such as actuators, sensors, accelerators, ultrasonic motors, transducers, filters and resonators, and micro electromechanical systems (MEMS). Important ferroelectric oxides for piezoelectric ceramics are restricted to perovskite-type, tungsten bronze-type, and bismuth layer-structured compounds. The most widely used piezoelectric materials are perovskite-type PbTiO3PbZrO3 (PZT)-based multi-component systems (PZT systems) [1e3] because of their excellent piezoelectric properties. However, it has been recently desired to use lead-free materials for environmental protection during the waste disposal of products. For example, legislation has been enforced in the EU as the draft Directives on Waste from Electrical and Electronic Equipment (WEEE) since January 1st, 2004, Restriction of Hazardous Substances (RoHS) since July 1st, 2006, and End-of-Life Vehicles (ELV) since July 1st, 2003. Therefore, lead-free piezoelectric materials have been attracting attention worldwide [4e6] as new materials in place of PZT-based piezoelectric ceramics. Lead-free piezoelectric materials, such as piezoelectric single crystals, e.g., langasite [7], and ferroelectric ceramics with a perovskite structure [8e59], a tungsten bronze structure [60,61], and bismuth layer-structured ferroelectrics (BLSF) [62e68] have been extensively reported for mainly in the past 10 years. Recently, various perovskite-structured ferroelectrics such as BaTiO3 [BT], (Bi1/2Na1/2)TiO3 [BNT], (Bi1/2K1/2)TiO3 [BKT], KNbO3 [KN], (K,Na)NbO3 [KNN], and their solid solutions have

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00025-3 Copyright Ó 2013 Elsevier Inc. All rights reserved.

been actively studied [16e59] as candidates for lead-free piezoelectric ceramics. These ceramics have been widely studied and are suitable for actuator and high-power applications because of their relatively large piezoelectric constants, d, among lead-free piezoelectrics. However, there are some problems such as low Curie temperatures, Tc, or low depolarization temperatures, Td, difficulties in poling treatments, and/or low relative densities. The ferroelectric strengths of perovskite-structured ABO3-type oxides are representative by PbTiO3 [PT] in 2e4 type, KNbO3 [KN] in 1e5 type, and BiFeO3 [BF] in 3e3 type, respectively. However, all of them, except PT, do not show their strong ferroelectrics as high as those of PZT. Therefore, no leadfree materials displays better piezoelectric properties than those of PZT-based systems. To replace PZT-based systems, it is necessary that the required piezoelectric properties for various applications are classified and developed for each application. For example, the perovskite-type ceramics seem to be suitable for actuator and high-power applications. On the other hand, BLSF ceramics seem to be candidate materials for ceramic resonator applications. In this section, a brief survey of non-lead-based piezoelectric ceramics is given and dielectric, ferroelectric, and piezoelectric properties of typical lead-free perovskite ferroelectric ceramics such as BaTiO3 [BT], Bi1/2Na1/2TiO3 [BNT], Bi1/2K1/2TiO3 [BKT], and KNbO3 [KN]-based systems for piezoelectric applications are described as superior candidates for environmental friendly lead-free piezoelectric ceramics to reduce some damage to the earth.

429

430

2. PEROVSKITE-STRUCTURED PIEZOELECTRIC CERAMICS 2.1. BaTiO3 [BT]-based Ceramics Barium titanate, BaTiO3 [BT] [8,9], is the firstdiscovered ferroelectric oxide with perovskite structure. This ceramic has a relatively high electromechanical coupling factor, k33 (~0.50), and a piezoelectric strain constant, d33 (~190 pC/N), and has been partially used for piezoelectric applications such as sonars. However, the working temperature range of the BT is narrow for actual piezoelectric applications because the BT has a low Curie temperature, Tc (¼120e135  C) [8,9,54,55]. On the other hand, bismuth potassium titanate, (Bi1/2K1/2) TiO3 [BKT], is a typical lead-free ferroelectric with a perovskite structure of tetragonal symmetry at room temperature and a relatively higher Tc of 380  C. [13] Hiruma et al. [42] reported the electrical properties of BKT ceramics prepared by the hot-pressing (HP) method. To expand the working temperature range, that is, to elevate the Tc of BT ceramic, dielectric and piezoelectric properties of (1 x)BT-xBKT solid solution (BTBKT100x) system were also investigated [39,94]. The Tc increased linearly with increasing the BKT content (x) and the Tc of BT-BKT20 (x ¼ 0.2) reaches higher than 200  C. Recently, large piezoelectric strain constants, d33 (¼350 pC/N) and d33 (¼450 pC/N), of BT ceramics were reported, respectively, in the microwave sintered BT by using hydrothermally prepared fine particles [54] and in the BT prepared by the two-step sintering method [55]. These large d33 values may be caused by the elevated very high relative free permittivity ð3T33 =30 Þ higher than 5000.

2.2. KNbO3 [KN]eNaNbO3 [NN]eLiNbO3 [LN] System Potassium niobate, KNbO3 [KN], ceramics have attracted much attention as a candidate material for lead-free piezoelectric applications, because the single-crystal KN has a large piezoelectricity and a high Curie point [10,30,32]. KN has an orthorhombic symmetry at room temperature, and has phase transitions at 10, 225 and 425  C corresponding to rhombohedral / orthorhombic / tetragonal / cubic, respectively. The electromechanical coupling factor of the thickness-extensional mode, kt, in a KN crystal reaches as high as 0.69 for the 49.5 rotated X-cut about the Y-axis, which is the highest among current lead-free piezoelectrics [49]. KN single crystals are known to have high piezoelectric activities. However, it is difficult to obtain a dense ceramic body of KN by the ordinary firing process. To obtain the dense KN-based ceramic, the hot-press (HP) method or liquid-phase sintering by additive dopants was investigated

Handbook of Advanced Ceramics

[31,33]. On the other hand, electrical properties of potassiumesodium niobates, KNbO3eNaNbO3 [KNN] system, were reported by Egerton et al. [11,14] Their works on ceramics in the system indicated that relatively low dielectric constants and high electromechanical coupling factors could be obtained over a wide range of compositions. However, it is difficult to realize the desired structure in a ceramic form because the sintering of these materials in an air requires long soaking periods to achieve sufficient densification. There have been many reports on KN solid-solution systems such as KNbO3eNaNbO3 [KNN]. Good piezoelectric properties, such as a large planar coupling factor, kp ¼ 0.56, and a large remanent polarization, Pr up to 30 mC/cm2, have been observed for KNN ceramics [11]. An excellent piezoelectric property of 416 pC/N in textured (K0.5Na0.5)NbO3-based ceramics has recently been reported by Saito et al. [35] Figure 1 shows (a) the piezoelectric strain constant, d31, as a function of the content of Ta and Li ions and (b) the d33, as a function of the Curie temperature [35]. Table 1 summarizes piezoelectric properties of the textured LF4T compared with those of PZT4 [35]. Recently, Guo et al. [36] also reported that (1 x)(K,Na)NbO3exLiNbO3 solid solution [(1 x) KNN-xLN] shows the MPB composition between the orthorhombic and the tetragonal symmetries near 0.05< x S (tetragonal) > S (rhombohedral) for their compositions. Table 4 [6] summarizes the piezoelectric properties of BNBK2:1(x) (x ¼ 0.78, 0.88 and 0.98) including the d and the S  defined by obtained from the results of Figure 17. The d33 the equation,  d33 ½pm=VŠ ¼

FIGURE 12 Phase relation of the (Bi1/2Na1/2)TiO3 (BNT)e(Bi1/2K1/2) TiO3 (BKT)eBaTiO3 (BT) system around the MPBs.

Strain; S  106 Ea ½kV=mmŠ

(1)

for example, is 188 pm/V with relatively high Td (206  C) in the tetragonal composition (x ¼ 0.78). Finally, Table 5 summarizes the depolarization temperatures, Td, and piezoelectric properties of rhombohedral, MPB, and tetragonal compositions of the BNBK2:1(x) ternary system.

Chapter | 6.2

Lead-Free Piezoelectric Ceramics

437

FIGURE 15 The depolarization temperature, Td and the Tm, determined by the maximum dielectric constant, 3r, from the measurement of temperature dependence of 3r.

FIGURE 16 The relationship between the Td and the d33 of the tetragonal side (Tetr.) and the rhombohedral side (Rhomb.).

FIGURE 14 Temperature dependence of electromechanical coupling factor, k33, of (a) rhombohedral side of x ¼ 0.92e0.98, (b) MPB composition of x ¼ 0.88e0.90, and (c) tetragonal side of x ¼ 0.30e0.84 in BNBK2:1(x).

2.6. (Bi1/2Na1/2)TiO3 [BNT]e(Bi1/2Li1/2)TiO3 [BLT]e(Bi1/2K1/2)TiO3 [BKT] system The BNT presents the low depolarization temperature Td of 185  C, and the MPBs of BNT-based solid solutions show a particularly low Td of approximately 100  C [51,53,56,76,77], even though the MPBs have excellent piezoelectric properties. The piezoelectric working temperature range is limited to below the Td because piezoelectric properties disappear at the Td and higher temperatures. Therefore, it is important to increase the Td of the MPB composition to enable its practical use in

piezoelectric applications. From the effects [53] of Li- and K-substitution on the Td in the A-site of BNT, it is seen that the Td of BNT increases when the small amounts of Li and K are substituted. Phase transition temperatures, Td, TReT, and Tm, of a solid solution, (1 x)(Bi1/2Na1/2)TiO3-x(Bi1/2 K1/2)TiO3 [BNKT100x] are summarized in Figure 18. Field-induced strains of BNKT100x under unipolar driving at 0.1 Hz are shown in Figure 19, and the piezoelectric constant, d33, and the normalized strain,  ð ¼S d33 max =Emax Þ, at 80 kV/cm are summarized in  values were higher than the d , because Figure 20. All d33 33  the d33 includes the domain contribution due to the high applied voltage and the low measuring frequency [80]. The highest value was obtained near the MPB composition for  to the d are BNKT100x. Moreover, the ratios of the d33 33 larger for the tetragonal compositions than for the rhombohedral compositions due to the difference in the domain structures. Double substitution of Li and K is thought to be effective in increasing the Td of BNT-based solid solutions. Figure 21 shows the phase relation of x(Bi1/2Na1/2)

438

Handbook of Advanced Ceramics

FIGURE 17 Strain of BNBK2:1(0.88), BNBK2:1(0.78), BNBK2:1(0.98) as a function of the applied electric field, Ea.

and

TiO3-y(Bi1/2Li1/2)TiO3-z(Bi1/2K1/2)TiO3 (x þ y þ z ¼ 1) (abbreviated to BNLKT100y-100z) the three-component system [57,78]. In order to increase the Td, a small amount of Li-substitution is probably optimal. In addition, it is important to use the MPB composition to enhance the piezoelectric properties. X-ray powder diffraction patterns show co-existence of rhombohedral and tetragonal phases at z ¼ 0.18e0.20 for BNLKT4-100z and at z ¼ 0.18e0.22 for BNLKT8-100z. Therefore, the MPBs exist in these compositions. Figure 22 shows the variation in the depolarization temperature, Td, as a function of the content (z) of BKT, for BNLKT0-100z, BNLKT4-100z, and BNLKT8-100z. In order to determine the Td accurately, it was measured from the temperature dependence of piezoelectric properties using fully poled 33-mode specimens and dielectric loss tangent, tan d, using fully poled specimens [53]. The Td of BNLKT4-100z was higher than those of BNLKT0-100z and BNLKT8-100z. On the rhombohedral side (0 < z 0.6 m/s. Moreover, the Qm(31) of BNLKT4-8Mn0.6 was higher than 400, even at v0 p¼ 1.5 m/s. It is considered that the small decay of Qm(31) under a high-amplitude vibration for BNLKT4-8 and BNLKT4-

8Mn0.6 is attributed to the high coercive field Ec. It is known that the Ec at the rhombohedral composition is larger than that at the MPB and on the tetragonal side for BNT-based solid solutions [51]. The BNLKT4-8 has high Ec of 60 kV/cm, which is 4 times larger than that of PZT-H. Moreover, domain-wall motion is suppressed by oxygen vacancies associated with the doping of acceptor ions [82]. Therefore, the acceptor-ion-doped rhombohedral composition of BNT-based solid solutions is stable even at a vibration velocity higher than PZT-H.

2.7. (Bi1/2K1/2)TiO3 [BKT]-based Ceramics Bismuth potassium titanate, (Bi1/2K1/2)TiO3 [BKT], is a typical lead-free ferroelectric with a perovskite structure of tetragonal symmetry at room temperature and has a relatively high Curie temperature, Tc, of 380  C [13]. This indicates that BKT has a certain promise as a candidate for lead-free piezoelectrics in a wide working temperature range. However, there are few reports about this material owing to its poor sinterability, except for our reports [42]

Chapter | 6.2

Lead-Free Piezoelectric Ceramics

441

FIGURE 24 (a) Temperature dependence of the coupling factor, k33, for BNLKT4-8Mnw (w ¼ 0 and 0.6) and (b) the Td as a function of w.

FIGURE 26 Temperature dependences of dielectric constant, 3s, and dielectric loss tangent, tan d, for the hot-pressed BKT ceramics sintered at (a) 1060  C and (b) 1080  C.

FIGURE 25 Variations in Qm(31) as a function of the vibration velocity v0 p for BNLKT4-8, BNLKT4-8Mn0.6, and PZT-H.

[93]. This problem has restricted extensive activities of researchers in investigations for the BKT-based solidsolution systems. Recently, Hiruma et al. [42,93] reported the electrical properties of BKT ceramics prepared by the hot-pressing (HP) method. The optimum sintering temperature seems to be 1060e1080  C. Figure 26 shows the temperature dependences of dielectric constants, 3s, and dielectric loss tangents, tan d, for BKT-HP1060  C and BKT-HP1080  C, in the temperature range from room temperature to 600  C, measured at frequencies of 10 kHz, 100 kHz, and 1 MHz. The tan d curves in Figure 26 show two peaks. Hightemperature peaks show frequency dispersions; therefore, it is thought that these peaks are related to the Tc. On the other hand, low-temperature peaks are almost independent of frequencies; therefore, it is considered to indicate the second phase transition, T2, between tetragonal and pseudo-cubic symmetries, whose temperatures of BKT-HP1060  C and BKT-HP1080  C are about 340 and 315  C, respectively. The Tc of Bi2O3, La2O3, or MnCO3-doped BKT ceramics

showed a tendency to decrease with an increase in the content of dopants. Figure 27 shows DeE hysteresis loops of BKTHP1060  C and BKT-HP1080  C. Well-saturated DeE hysteresis loops with low leakage current were obtained for all specimens at RT. The Pr and the coercive fields Ec were 22.2 mC/cm2 and 52.5 kV/cm for BKT-HP1080  C, and 14.2 mC/cm2 and 47.3 kV/cm for BKT-HP1060  C, respectively. The different tendencies seem to be the difference in grain size 0.2 mm for BKT-HP1060  C and 0.4 mm for BKT-HP1080  C. A solid-solution system, (1-x) (Bi1/2K1/2)TiO3e xBaTiO3 [BKTeBT100x] [6,95], was investigated for evaluating their characterizations and electrical properties by using randomly oriented and grain-oriented samples. Especially, the compositions near the BKT (x ¼ 0e0.4) were focused as the lead-free piezoelectric ceramics with the wide working temperature. X-Ray diffraction patterns of BKTeBT100x ceramics with 0 & x & 1 show a single phase of perovskite structure with tetragonal symmetry at room temperature. The sintered ceramics of BKTeBT indicated higher relative densities than 95%, even in the compositions of the BKT side as shown in Table 6. Figure 28 shows the temperature dependence of electromechanical coupling factor, k33, and the phase, q, in the impedanceefrequency characteristics of the (33)-mode for BKT [42] ceramic. This figure indicates that the BKT and BKTeBT ceramics seem to be attractive for

442

Handbook of Advanced Ceramics

FIGURE 27 DeE hysteresis loops of the HP-BKT at 1060  C and 1080  C.

FIGURE 28 Temperature dependence of electromechanical coupling factor, k33, and the phase, q, in the impedanceefrequency characteristics of the (33)-mode for BKT ceramic [42].

higher-temperature applications compared with the BT ceramic. Figure 29 shows lattice constants, a and c, and lattice anisotropy, c/a, as a function of the amount (x) of BT in BKTeBT100x ceramics. Both BKT (x ¼ 0) and BT (x ¼ 1) have the same crystal structure of tetragonal symmetry; however, the c/a indicated nonlinear tendency. The c/a shows the maximum (~1.025) at the composition around x ¼ 0.2, which is a very large value among the lead-free piezoelectric materials. This result corresponds almost to that shown in Buhrer’s report [13]. From temperature dependences of dielectric constant, 3r and loss tangent, tan d, for the BKTeBT100x ceramics, the Curie temperature, Tc, linearly shifted to lower temperatures with an increase in the amount of BT content (x), as shown in Figure 30. The Tc of BKTeBT80 (x ¼ 0.8) shows still a higher temperature than 200  C. However, both the 3r at RT and at the Tc decrease with increasing x. The T2 in Figure 30 shows the second phase transition temperature from tetragonal to pseudo-cubic phases, existing near 270  C in BKT. The T2 of BKT ceramic was reported by Ivanova [96]. Furthermore, grain orientation effects for piezoelectric properties were investigated in BKTeBT100x using a reactive template grain growth (RTGG) method [97]. Piezoelectric strain constant, d33, for randomly oriented BKTeBT10, 20, and 30 are 73.4, 69.1, and 67.6 pC/N, respectively. These values are relatively small for practical use to actuators. So, we tried to prepare the textured samples by the RTGG method to enhance their piezoelectric properties. Textured specimens were prepared using the RTGG method with matrix and templates of plate-like Bi4Ti3O12 (BIT) particles for BKTeBT. Calcination and sintering temperatures were 900e1000  C and

TABLE 6 Physical and Piezoelectric Properties for Nonoriented (OF) and Grain-Oriented (RTGG) BKTeBT10, 20, and 30 Prepared by Ordinary Firing (OF) and RTGG Methods [97] BKTeBT10

BKTeBT20

BKTeBT30

OF

RTGG

OF

RTGG

OF

RTGG

r0 /rx (%)

98.5

97.8

99.3

95.4

98.0

94.3

F (%)

e

35

e

72

e

61

k33

0.35

0.37

0.36

0.33

0.38

0.38

3T33 =30

602

560

532

501

461

426

E ðpm2 =NÞ s33

8.2

10.5

8.2

10.1

7.8

13.0

d33 ðpC=NÞ

73.4

84.5

69.3

70.7

67.6

83.3

 ðpm=VÞ d33

103

168

116

103

134

rx: relative density ratio, F: orientation factor, k33: electromechanical coupling factor,  constant, d33 : calculated d33 by the Eqn 1.

143 3T33 :

free permittivity,

E s33 :

elastic constant, d33: piezoelectric strain

Chapter | 6.2

Lead-Free Piezoelectric Ceramics

FIGURE 29 Lattice anisotropy, c/a, and lattice constants, a and c, as a function of the amount (x) of BT in BKTeBT100x ceramics.

FIGURE 30 Curie temperature, Tc, and secondary phase transition temperature, T2, of BKTeBT100x ceramics, as a function of the amount (x) of BT, measure at 1 MHz.

1100e1400  C, respectively. The piezoelectric properties of the textured and nonetextured BKTeBT10, 20, and 30 are summarized in Table 6. The measured direction of textured specimens is parallel [//] to the tape stacking direction. The d33 values in textured specimens were improved as compared with those of nontextured specimens; for example, the d33 in BKTeBT10 was improved from 73.4 to 84.5 pC/N. However, the increment was relatively small because the orientation factor, F, was still low, of about 35 %. Further systematic studies for improving the F may be necessary as a future work.

3. SUMMARY The dielectric, ferroelectric, and piezoelectric properties of lead-free perovskite-type ceramics were investigated, being candidates for lead-free piezoelectric materials, to reduce environmental damage during the waste disposal of piezoelectric products. Perovskite ferroelectric ceramics seem to be suitable for high-power applications such as piezoelectric actuators requiring large piezoelectric constants, d33, and high Curie temperatures, Tc, or high depolarization temperatures, Td.

443

The dense KN-based ceramics, KNbO3þMnCO3 x wt% [KN-Mn x], can be obtained by using the modified conventional ceramic firing technique with good dielectric, ferroelectric, and piezoelectric properties. The electromechanical coupling factor, k15 ¼ 0.55, and the piezoelectric strain constant, d15 ¼ 207 pC/N, are obtained at Mn-doped KNeMn0.1. The Bi-excess bismuth sodium titanate, (Bi1/2Na1/2)TiO3 (BNT) þ Bi2O3 x wt% [BNT-x], and hot-pressing BNT [HP-BNT] ceramics were prepared and their piezoelectric properties were investigated. All of BNT-x ceramics sintered at 1225  C showed a high density ratio, more than 97 % to the theoretical density. The BNT0.3 seems to be a stoichiometry because of the measurement results from the resistivity, r, Curie temperature, Tc, and microstructure. BNT-0.3 ceramic showed a relatively large electromechanical coupling factor, k33 (¼0.47), and piezoelectric constant, d33 (¼93 pC/N). The large piezoelectricity, k33 (¼0.48) and d33 (¼98 pC/N), with the high density ratio (98 %), could be obtained for the first time on the HP-BNT ceramic sintered at 1100  C for 30 min and pressed at 200 kg/cm2 as lead-free piezoelectric materials. In the case of the ternary system x(Bi1/2Na1/2)TiO3y(Bi1/2K1/2)TiO3-zBaTiO3 (x þ y þ z ¼ 1), [BNBKy:z (x)], enhanced piezoelectric properties were obtained near the MPB composition, and the highest electromechanical coupling factor, k33, and d33 were 0.58 for BNBK2:1(0.89) and 181 pC/N for BNBK2:1(0.88). Nevertheless, the Td shift to lower temperatures (about 100  C) around the MPB compositions corresponds to BNBK2:1(0.88e0.90). On the tetragonal side, Td shifts to a value higher than 200  C for x < 0.78, with k33 ¼ 0.452 and d33 ¼ 126 pC/N for BNBK2:1(0.78). This ternary system shows high potential (with properties such as large d33> 250 pC/N and high Td > 250  C) for actuator applications. A small amount of Li-substitution was very effective for increasing the Td because the Td is related to the variation of lattice distortion. The Td of x(Bi1/2Na1/2)TiO3-y(Bi1/2Li1/2) TiO3-z(Bi1/2K1/2)TiO3 (x þ y þ z ¼ 1) (BNLKT100y-100z) three-component system increased from 185  C to 221  C in BNLKT4-100z at the rhombohedral composition. Although excess Li-substitution enhanced the piezoelectric properties, Td drastically decreased with increasing amount of Li-substitution. Considering both high Td and high d33, the tetragonal composition of BNLKT4-100z is optimal for piezoelectric actuator applications. The rhombohedral composition of BNLKT4-8 seems to be suitable for highpower applications. In addition, a small amount of Mndoping is very effective for improving Qm. The high-power characteristics of BNLKT4-8Mn0.6 are superior to those of PZT-H at approximately vp-p > 0.6 m/s. Therefore, a Mndoped BNT-based solid solution with rhombohedral symmetry is a promising candidate for lead-free highpower applications.

444

The (1 x)(Bi1/2K1/2)TiO3-xBaTiO3 [BKTeBT100x] system was selected to elevate the Tc of BT to higher temperatures than 200  C. The Tc of BKTeBT80 was about 230  C. From these results, it is clear that the BKTeBT system is a superior candidate for lead-free piezoelectric materials for to use in high-temperature applications. The piezoelectric properties of the textured and nontextured BKTeBT10, 20, and 30 are studied. The d33 in textured specimen of BKTeBT10 was improved as compared with that of nontextured specimen from 73.4 to 84.5 pC/N. To replace PZT-based systems, it is necessary that special features of each lead-free material correspond to the required piezoelectric properties for each application such as high-temperature sensors, high-frequency applications, and temperature-stable ceramic resonators. Future trends in the research and development of environmentally gentle lead-free piezoelectric ceramics seem to be focused on textured grain orientations realized using the templated grain growth (TGG) and seeded polycrystal conversion (SPC) methods. In addition, domain controls including engineered domain of lead-free ferroelectric ceramics will be very important to enhance the piezoelectric activities. The trend toward thick and thin films with lead-free piezoelectric ceramics seems to be a necessity for FBAW and MEMS applications.

REFERENCES [1a] Sawaguchi E. Ferroelectricity versus antiferroelectricity in the solid solutions of PbZrO3 and PbTiO3. J Phys Soc Jpn 1953;8:615e29. [1b] Jaffe B, Roth RS, Marzullo S. Piezoelectric properties of lead zirconate-lead titanate solid-solution ceramics. J Appl Phys 1954; 25:809e10. [2] Jaffe B, Cook WR, Jaffe H. Piezoelectric ceramics. New York: Academic; 1971. p. 142. [3] Yamamoto T. Ferroelectric properties of the PbZrO3-PbTiO3 system. Jpn J Appl Phys 1996;35(9B):5104e8. [4] Takenaka T. Piezoelectric properties of some lead-free ferroelectric ceramics. Ferroelectrics 1999;230:87e98. [5] Takenaka T, Nagata H. Current status and prospects of lead-free piezoelectric ceramics. J European Cer Soc 2005;25(12): 2693e700. [6] Takenaka T, Nagata H, Hiruma Y. Current developments and prospective of lead-free piezoelectric ceramics. Jpn J Appl Phys 2008;47:3787e801. [7] Shimamura K, Takeda H, Kohno T, Fukuda T. Growth and characterization of lanthanum gallium silicate La3Ga5SiO14 single crystals for piezoelectric applications. J Cryst Growth 1996; 163:388e92. [8] Von Hippel A. J Ind Eng Chem 1946;28:1097. [9] Robert S. Dielectric and piezoelectric properties of barium titanate. Phys Rev 1947;71:890e5. [10] Mathias BT, Remeika JP. Dielectric properties of sodium and potassium Niobates. Phys Rev 1951;82:727e9.

Handbook of Advanced Ceramics

[11] Egerton L, Dillon DM. Piezoelectric and dielectric properties of ceramics in the system potassium-sodium niobate. J Am Ceram Soc 1959;42:438e42. [12] Smolensky GA, Isupov VA, Agranovskaya AI, Krainic NN. New ferroelectrics of complex composition. IV. Soviet Phys Solid State 1961;2(11):2651e4. [13] Buhrer CF. Some properties of bismuth perovskites. J Chem Phys 1962;36(3):798e803. [14] Jaeger RE, Egerton L. Hot pressing of potassium-sodium niobates. J Am Ceram Soc 1962;45(5):209e13. [15] Sakata K, Masuda Y. Ferroelectric and antiferroelectric properties of (Bi1/2Na1/2)TiO3-SrTiO3 solid solution ceramics. Ferroelectrics 1974;7:347e9. [16] Suchanicz J, Roleder K, Kania A, Handerek J. Electrostrictive strain and pyroeffect in the region of phase coexistence in Na0.5Bi0.5TiO3. Ferroelectrics 1988;77:107e10. [17] Takenaka T, Sakata K. New piezo- and pyroelectric sensor materials of (BiNa)1/2TiO3-based ceramics. Sensor Mater 1988;1:123e31. [18] Takenaka T, Sakata K. Dielectric, piezoelectric and pyroelectric properties of (BiNa)1/2TiO3-Based ceramics. Ferroelectrics 1989; 95:153e6. [19] Takenaka T, Sakata K, Toda K. Piezoelectric properties of (Bi1/2Na1/2)TiO3-based ceramics. Ferroelectrics 1990;106: 375e80. [20] Takenaka T, Maruyama K, Sakata K. (Bi, Na)TiO3-BaTiO3 system for lead-free piezoelectric ceramics. Jpn J Appl Phys 1991;30(9B): 2236e9. [21] Takenaka T, Hozumi A, Hata T, Sakata K. Mechanical properties of (Bi1/2Na1/2)TiO3-based piezoelectric ceramics. Silicates Industries -Ceramic Science and Technology 1993;58(7e8):136e42. [22] Nagata H, Takenaka T. Lead-free piezoelectric ceramics of (Bi1/2Na1/2)TiO3-1/2(Bi2O3$ Sc2O3) system. Jpn J Appl Phys 1997;36:6055e7. [23] Takenaka T, Okuda T, Takegahara K. Lead-free piezoelectric ceramics based on (Bi1/2Na1/2)TiO3-NaNbO3. Ferroelectrics 1997;196:175e8. [24] Chiang Y-M, Farrey GW, Soukhojak AN. Lead-free high-strain single-crystal piezoelectrics in the alkalineebismuthetitanate perovskite family. Appl Phys Lett 1998;73:3683e5. [25] Nagata H, Takenaka T. Lead-free piezoelectric ceramics of (Bi1/2Na1/2)TiO3-KNbO3-1/2(Bi2O3$Sc2O3) system. Jpn J Appl Phys 1998;37:5311e4. [26] Nagata H, Takenaka T. (Bi1/2Na1/2)TiO3-based non-lead piezoelectric ceramics. J Korean Phys Soc 1998;32:S1298e300. [27] Takenaka T, Nagata H. Ferroelectric and piezoelectric properties of lead-free (Bi1/2Na1/2)TiO3-KNbO3-1/2(Bi2O3,Sc2O3) ceramics. Proc. the 11th IEEE International Symposium on the Applications of Ferroelectrics (ISAF XI ’98) (IEEE Catalog No.98CH36245); 1999, pp. 559e562. [28] Nagata H, Koizumi N, Takenaka T. Lead-free piezoelectric ceramics of (Bi1/2Na1/2)TiO3-BaTiO3-BiFeO3 system. Ferroelectrics 1999;229:273e8. [29] Sasaki A, Chiba T, Mamiya Y, Otsuki E. Dielectric and piezoelectric properties of (Bi, Na)TiO3-(Bi, K)TiO3 systems. Jpn J Appl Phys 1999;38(9B):5564e7. [30] Nakamura K, Kawamura Y. Orientation dependence of electromechanical coupling factors in KNbO3. IEEE Trans Ultrason 2000;47:750e5.

Chapter | 6.2

Lead-Free Piezoelectric Ceramics

[31] Ishii H, Nagata H, Takenaka T. Morphotropic phase boundary and electrical properties of bismuth sodium titanate-potassium niobate solid-solution ceramics. Jpn J Appl Phys 2001;40(9B):5660e3. [32] Wada S, Seike A, Tsurumi T. Poling treatment and piezoelectric properties of potassium niobate ferroelectric single crystals. Jpn J Appl Phys 2001;40:5690e7. [33] Kakimoto K, Masuda I, Osato H. Ferroelectric and piezoelectric properties of KNbO3 ceramics containing small amounts of LaFeO3. Jpn J Appl Phys 2003;42:6102e5. [34] Nagata H, Yoshida M, Makiuchi Y, Takenaka T. Large piezoelectric constant and high curie temperature of lead-free piezoelectric ceramic ternary system based on bismuth sodium titanate-bismuth potassium titanate-barium titanate near the morphotropic phase boundary. Jpn J Appl Phys 2003;42:7401e3. [35] Saito Y, Takao H, Tani T, Nonoyama T, Takatori K, Homma T, et al. Lead free piezoceramics. Nature Nov. 2004;42:84e7. [36] Guo YP, Kakimoto K, Ohsato H. Phase transitional behavior and piezoelectric properties of NKN-LN ceramics. Appl Phys Lett 2004;85:4121e3. [37] Guo Y, Kakimoto K, Ohsato H. Structure and electrical properties of lead-free (Na0.5K0.5)NbO3-BaTiO3 ceramics. Jpn J Appl Phys 2004;43:6662e6. [38] Kakimoto K, Masuda I, Osato H. Solid-solution structure and piezoelectric property of KNbO3 ceramics doped with small amounts of elements. Jpn J Appl Phys 2004;43:6706e10. [39] Hiruma Y, Aoyagi R, Nagata H, Takenka T. Piezoelectric properties of BaTiO3-(Bi1/2K1/2)TiO3 ferroelectric ceramics. Jpn J Appl Phys 2004;43(11A):7556e9. [40] Birol H, Damjanovic D, Setter N. Preparation and characterization of KNbO3 ceramic. J Am Ceram Soc 2005;88:1754e9. [41] Makiuchi Y, Aoyagi R, Hiruma Y, Nagata H, Takenaka T. (Bi1/2Na1/2)TiO3-(Bi1/2K1/2)TiO3-BaTiO3ebased lead-free piezoelectric ceramics. Jpn J Appl Phys 2005;44(6B):4350e3. [42] Hiruma Y, Aoyagi R, Nagata H, Takenaka T. Ferroelectric and piezoelectric properties of (Bi1/2K1/2)TiO3 ceramics. Jpn J Appl Phys 2005;44(7A):5040e4. [43] Nagata H, Shinya T, Hiruma Y, Takenaka T, Sakaguchi I, Haneda H. Piezoelectric properties of bismuth sodium titanate ceramics. Ceram Trans 2005;167:213e22. [44] Yoshida T, Hiruma Y, Aoyagi R, Nagata H, Takenaka T. Processing and electrical properties of KNbO3 ferroelectric dense ceramics added with small amount of Bi2O3 and MnCO3. Key Eng Materials 2005;301:19e22. [45] Matsumoto K, Hiruma Y, Nagata H, Takenaka T. Piezoelectric properties of pure and Mn-doped potassium niobate ferroelectric ceramics. Jpn J Appl Phys 2006;45(5B):4479e83. [46] Nagata H, Matsumoto K, Hirosue T, Hiruma Y, Takenaka T. Fabrication and electrical properties of potassium niobate ferroelectric ceramics. Jpn J Appl Phys 2007;46(10B):7084e8. [47] Zgonik M, Schlesser R, Biaggio I, Voit E, Tscheny J, Gunter P. J Appl Phys 1993;74(2):1287. [48] Onoe M, Tiersten HF, Meitzler AH. J Acoust Soe Amer 1963;35:36. [49] Nakamura K, Tokiwa T, Kawamura Y. J Appl Phys 2002;91:9272. [50] Yoshii K, Hiruma Y, Nagata H, Takenaka T. Electrical properties and depolarization temperature of (Bi1/2Na1/2)TiO3-(Bi1/2K1/2)TiO3 lead-free piezoelectric ceramics. Jpn J Appl Phys 2006; 45(5B):4493e6.

445

[51]

[52]

[53]

[54]

[55]

[56]

[57]

[58]

[59]

[60]

[61]

[62]

[63]

[64]

[65]

[66]

[67]

[68]

Hiruma Y, Makiuchi Y, Aoyagi R, Nagata H, Takenaka T. Leadfree piezoelectric ceramics based on (Bi1/2Na1/2)TiO3-(Bi1/2K1/2) TiO3-BaTiO3 solid solution. Ceram Trans 2005;174:139e46. Hiruma Y, Yoshii K, Aoyagi R, Nagata H, Takenaka T. Piezoelectric properties and depolarization temperatures of (Bi1/ lead-free piezoelectric 2Na1/2)TiO3-(Bi1/2K1/2)TiO3-BaTiO3 ceramics. Key Eng Mater 2006;320:23e6. Hiruma Y, Nagata H, Takenaka T. Phase transition temperatures and piezoelectric properties of (Bi1/2Na1/2)TiO3-(Bi1/2K1/2)TiO3-BaTiO3 lead-free piezoelectric ceramics. Jpn J Appl Phys 2006;45(9B): 7409e12. Takahashi H, Numamoto Y, Tani J, Tsurekawa S. Piezoelectric properties of BaTiO3 ceramics with high performance fabricated by microwave sintering. Jpn J Appl Phys 2006;45(9B):7405e8. Karaki T, Yan K, Miyamoto T, Adachi M. Lead-free piezoelectric ceramics with large dielectric and piezoelectric constants manufactured from BaTiO3 nano-powder. Jpn J Appl Phys 2007;46(4):L97e8. Zhang S, Shrout TR, Nagata H, Hiruma Y, Takenaka T. Piezoelectric properties in (K0.5Bi0.5)TiO3-(Na0.5Bi0.5)TiO3-BaTiO3 lead-free ceramics. IEEE Trans Ultrason Ferroelect Freq Contr 2007; 54(5):910e7. Hiruma Yuji, Nagata Hajime, Takenaka Tadashi. Phase-transition temperatures and piezoelectric properties of (Bi1/2Na1/2)TiO3(Bi1/2Li1/2)TiO3-(Bi1/2K1/2)TiO3 lead-free ferroelectric ceramics. IEEE Trans Ultrason Ferroelect Freq Contr 2007;54(12):2493e9. Takenaka T, Nagata H, Hiruma Y, Yoshii K, Matsumoto K. LeadFree piezoelectric ceramics based on pervskite structures. J Electroceramics 2007;19:259e65. Hiruma Y, Watanabe Y, Nagata H, Takenaka T. Phase Transition temperatures of divalent and trivalent ions substituted (Bi1/2Na1/2) TiO3 ceramics. Key Eng Mater 2007;350:93e6. Kimura M, Minamikawa T, Ando A, Sakabe Y. Temperature characteristics of (Ba1 x Srx)2NaNb5O15 ceramics. Jpn J Appl Phys 1997;36:6051e4. Doshida Y, Kishimoto S, Ishii K, Kishi H, Tamura H, Tomikawa Y, et al. Miniature cantilever-type ultrasonic motor using Pb-free multilayer piezoelectric ceramics. Jpn J Appl Phys 2007;46:4921e5. Ikegami S, Ueda I. Piezoelectricity in ceramics of ferroelectric bismuth compound with layer structure. Jpn J Appl Phys 1974;13(10):1572e9. Takenaka T, Sakata K. Grain orientation and electrical properties of hot-forged Bi4Ti3O12 ceramics. Jpn J Appl Phys 1980;19(1): 31e9. Takenaka T, Sakata K. Grain orientation effects on electrical properties of bismuth layer-structured ferroelectric Pb(1-x)(NaCe)x/2Bi4Ti4O15 solid solution. J Appl Phys 1984;55(4):1092e9. Takenaka T, Sakata K. Piezoelectric and pyroelectric properties of calcium-modified and grain-oriented (NaBi)1/2Bi4Ti4O15 ceramics. Ferroelectrics 1989;94:175e81. Takenaka T. Grain orientation effects on electrical properties of bismuth layer- structured ferroelectric ceramics. J Ceramic Soc Jpn 2002;110(4):215e24. Takenaka T, Nagata H. Grain orientation and electrical properties of some bismuth layer-structured ferroelectrics for lead-free piezoelectric applications. Ferroelectrics 2006;336:119e36. (Part VII Novel processing and new materials, No. 27 “Grain orientation and electrical properties of bismuth layer-structure ferroelectrics”). In: Ye Z-G, editor. Handbook of advanced

446

[69]

[70]

[71]

[72]

[73]

[74]

[75]

[76]

[77]

[78]

[79]

[80]

[81]

[82]

Handbook of Advanced Ceramics

dielectric, piezoelectric and ferroelectric materials. Woodhead Publishing Limited and CRC Press LLC; 2008. p. 818e51. Park S-E, Chung S-J. Nonstoichiometry and the long-range cation ordering in crystals of (Bi1/2Na1/2)TiO3. J Amer Cer Soc 1994;77(10):2641e7. Pronin IP, Syrnikov PP, Isupov VA, Egorov VM, Zaitseva NV, Ioffe AF. Peculiarities of phase transitions in sodium-bismuth titanate. Ferroelectrics 1980;25:395e7. Zvirgzds JA, Kapostis PP, Zvirgzde JV. X-ray study of phase transition in ferroelectric (Bi0.5Na0.5)TiO3. Ferroelectrics 1982; 40:75e7. Yi JY, Lee J-K, Hong K-S. Dependence of the microstructure and the electrical properties of lanthanum-substituted (Bi1/2Na1/2)TiO3 on cation vacancies. J Amer Cer Soc 2003;85(12):3004e10. Herabut A, Safari A. Processing and electrochemical properties of (Bi1/2Na1/2)1 1.5xLaxTiO3 ceramics. J Amer Cer Soc 1997;80(11): 2954e8. Watanabe Y, Hiruma Y, Nagata H, Takenaka T. Phase transition temperatures and electrical properties of divalent Ions (Ca2þ, Sr2þ and Ba2þ) substituted (Bi1/2Na1/2)TiO3 ceramics. Ceram Int 2008;34(4):761e4. Wang XX, Tang XG, Chan HLW. Electromechanical and ferroelectric properties of (Bi1/2Na1/2)TiO3e(Bi1/2K1/2)TiO3eBaTiO3 lead-free piezoelectric ceramics. Appl Phys Lett 2004;85(1): 91e3. Lin D, Xiao D, Zhu J, Yu P. Piezoelectric and ferroelectric properties of [Bi(Na K Li)]TiO3 lead free piezoelectric ceramics. Appl Phys Lett 2006;88(062901). Hiruma Y, Yoshii K, Nagata H, Takenaka T. Investigation of phase transition temperatures on (Bi1/2Na1/2)TiO3-(Bi1/2K1/2)TiO3 and (Bi1/2Na1/2)TiO3-BaTiO3 lead-free piezoelectric ceramics by electrical measurements. Ferroelectrics 2007;346:114e9. Hiruma Y, Yoshii K, Nagata H, Takenaka T. Phase transition temperature and electrical properties of (Bi1/2Na1/2)TiO3-(Bi1/2A1/2) TiO3 (A ¼ Li and K) lead-free ferroelectric ceramics. J Appl Phys 2008;103. pp. 084121e1~7. Siny IG, Tu C-S, Schmidt VH. Critical acoustic behavior of the relaxor ferroelectric Na1/2Bi1/2TiO3 in the intertransition region. Phys Rev B 1995;51:5659e65. Tsurumi T, Sasaki T, Kakemoto H, Harigai T, Wada S. Domain contribution to direct and converse piezoelectric effects of PZT ceramics. Jpn J Appl Phys 2004;43:7618e22. Hiruma Y, Watanabe T, Nagata H, Takenaka T. Piezoelectric properties of (Bi1/2Na1/2)TiO3-based solid solution for lead-free high-power applications. Jpn J Appl Phys 2008;47(9B):7659e63. Takahashi S, Hirose S. Vibration-level characteristics of leadzirconate-titanate ceramics. Jpn J Appl Phys 1992;31:3055e7.

[83] Hirose S, Magami N, Takahashi S. Piezoelectric ceramic transformer using piezoelectric lateral effect on input and on output. Jpn J Appl Phys 1996;35:3038e41. [84] Tashiro S, Ikehiro M, Igarashi H. Influence of temperature rise and vibration level on electromechanical properties of high-power piezoelectric ceramics. Jpn J Appl Phys 1997;36:3004e9. [85] Umeda M, Nakamura K, Ueha S. The measurement of high-power characteristics for a piezoelectric transducer based on the electrical transient response. Jpn J Appl Phys 1998;37:5322e5. [86] Kawada S, Ogawa H, Kimura M, Shiratsuyu K, Niimi H. Highpower piezoelectric vibration characteristics of textured SrBi2Nb2O9 ceramics. Jpn J Appl Phys 2006;45:7455e9. [87] Kawada S, Ogawa H, Kimura M, Shiratsuyu K, Higuchi Y. Relationship between vibration direction and high-power characteristics of -textured SrBi2Nb2O9 ceramics. Jpn J Appl Phys 2007; 46:7079e83. [88] Nagata H, Takenaka T, Effects of Substitution on Electrical Properties of (Bi1/2Na1/2)TiO3eBased Lead-Free Ferroelectrics. Proc. the 12th IEEE International Symposium on the Applications of Ferroelectrics (ISAF XII 2000) (IEEE Catalog No.00CH37076) pp. 45e51, 2001. [89] Zhu M, Liu L, Hou Y, Wang H, Yan H. Microstructure and electrical properties of MnO-doped (Na0.5Bi0.5)0.92Ba0.08TiO3 lead-free piezoceramics. J Am Ceram Soc 2007;90:120e4. [90] Gerthsen P, Ha¨rdtl KH, Schmidt NA. Correlation of mechanical and electrical losses in ferroelectric ceramics. J Appl Phys 1980;51:1131e4. [91] Takahashi S. Effects of impurity doping in lead zirconate-titanate ceramics. Ferroelectrics 1982;41:143e56. [92] Hayashi K, Ando A, Hamaji Y, Sakabe Y. Study of the valence state of the manganese ions in PbTiO3 ceramics by means of ESR. Jpn J Appl Phys 1998;37:5237e40. [93] Hiruma Y, Nagata H, Takenaka T. Grain-size effect on electrical properties of (Bi1/2K1/2)TiO3 ceramics. Jpn J Appl Phys 2007; 46:1081e4. [94] Hiruma Y, Nagata H, Takenaka T. Dielectric, ferroelectric and piezoelectric properties of barium titanate and bismuth potassium titanate solid-solution ceramics. J Ceramic Soc Jpn, Supplement 2004;112(5):S1125e8. [95] Takenaka T, Hiruma Y, Nemoto M, Nagata H. Piezoelectric properties of (Bi1/2K1/2)TiO3 e BaTiO3 ceramics with wide working temperatures. Extended Abstract of the 17th International Symposium on the Applications of Ferroelectrics (ISAF 2008), PL002, 2008. [96] Ivanova VV, Kapyshev AG, Veenevtsev YN, Zhdanov GS. Akad Nauk SSSR 1962;26:354. [97] Nemoto M, Hiruma Y, Nagata H, Takenaka T. Fabrication and piezoelectric properties of grain-oriented (Bi1/2K1/2)TiO3-BaTiO3 ceramics. Jpn J Appl Phys 2008;47:3829e32.

Chapter 6.3

Heat Capacity Study of Functional Ceramics Hitoshi Kawaji Materials and Structures Laboratory, Tokyo Institute of Technology, 4259 Nagatsuta-cho, Midori, Yokohama, Kanagawa 226-8503, Japan

Chapter Outline 1. Introduction 2. Thermodynamic Properties of Formation and Growth of Ferroelectric Nanoregion in Relaxors

447 448

1. INTRODUCTION The stable state of materials is determined by the enthalpy of the materials at zero kelvin, which means that the most energetically favorable state is realized at that temperature. However, the stable state at a finite temperature cannot be determined by the enthalpy alone, but the Gibbs energy, which includes both enthalpy and the entropy term, should be considered to discuss the stability of the materials. This means that some disorder, which increases the entropy of the materials, will be excited with increasing the temperature. On the other hand, the large amount of disorder may be obtained abruptly at a temperature, i.e. a phase transition. The phase transition changes the physical properties drastically and thus influences the functionality of the material. For example, the superconductivity of a copper oxide superconductor disappears above the critical temperature and the finite resistivity appears above the temperature. The situation that the functionality of materials can be obtained below a phase transition is also seen in the ferromagnetism, ferroelectricity, and so on. There are some cases that the functionality of the materials appears above the phase transition temperature as the superionic conductivity of silver iodide. These indicate the importance of the investigation of phase transitions for understanding of the functionality of materials. The phase transition is governed by the Gibbs energy at constant pressure and thus the study of Gibbs energy is essential for the understanding the phase transition properties. The heat capacity measure ments inform us the temperature dependence of the

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.000026-5 Copyright
2013 Elsevier Inc. All rights reserved.

3. Thermodynamic Studies of Giant Particle-Size Effect on the Phase Transition in Dielectric Crystals 4. Conclusion References

451 453 453

enthalpy and the entropy, thus the Gibbs energy, of materials and are important for the investigation of phase transitions. In addition, the entropy change due to the phase transition can be obtained only by the heat capacity measurements. To obtain the data of enthalpy and entropy for the determination of stability of materials, the absolute value of the heat capacity should be measured. The precise date can be obtained only by the adiabatic calorimetry, which had been used to confirm the third law of thermodynamics. The principle of the measurements is very simple, in which the temperature increase by the Joule heating is precisely measured under the adiabatic condition. First, the sample is held under the adiabatic condition, which can be obtained by vacuum insulation and by an adiabatic shield which is kept at the same temperature with the sample; thus, the temperature of the sample is maintained at constant. Next, the known energy is added to the sample, keeping the adiabatic condition. After the addition of the energy, the sample reaches at a temperature. The heat capacity of the sample is obtained by dividing the Joule energy by the temperature increment. In general, the increment of the each heating is about 1e2 K below room temperature. Actually, the heat capacity of the empty sample vessel with the thermometer and the heater is measured before the sample measurement and subtracted from the total heat capacity including sample to obtain the heat capacity of the sample. The precision and the accuracy of the measurements are determined by the accuracy of the adiabatic

447

448

control and of the temperature scale of the thermometer. The precision is estimated about 0.1% of the value at room temperature, if about 1 g of the sample is used for the measurements. The precision and accuracy of the measurements decrease with decreasing the amount of the sample. Thus, in the case that the only small amount of the sample is available, a relaxation method is applicable. The measurements with the precision of about 1% are possible using the sample of about several tens of milligrams [1]. However, a special care is needed to determine the phase transition enthalpy and the entropy of a first-order phase transitions using the relaxation method. Above room temperature, DSC is also available to determine the heat capacity with the precision of about 1% using about several tens of milligrams of sample [2].

2. THERMODYNAMIC PROPERTIES OF FORMATION AND GROWTH OF FERROELECTRIC NANOREGION IN RELAXORS In this section, the results of the heat-capacity study of the mechanism of relaxor behavior in perovskite oxides are discussed. The perovskite oxides are well known as the dielectric functional materials, especially lead complex oxides show superier electromechanical properties and are studied from the application points of view. In the group of the lead complex oxides Pb(B0 ,B00 )O3, there are some compounds which show a large and frequency-dependent peak in the dielectric constant. These compounds are called relaxor due to the relaxation properties in the dielectric constant [3]. Some compounds of relaxor show no macroscopic change in the crystal structure, meaning that the cubic symmetry above room temperature is maintained till zero kelvin. The lead complex oxides Pb(B0 ,B00 )O3 are classified into several groups. In the group of compounds with the B5þ ÞO3 , the typical relaxors, Pb(Mg1/3 formula PbðB2þ 1=3 2=3

Handbook of Advanced Ceramics

analysis by X-ray diffraction and neutron diffraction indicates that lead ions occupy the off-centered position by 0.03 nm from the original perovskite A site, and oxide ions also occupy in the displaced position as shown in Figure 1 [13]. The averaged crystal structure belongs to the cubic space group Pm3m in the whole temperature range and no structural phase transition is observed [14,15]. On the other hand, very small polarized region, which is called as polar nanoregion (PNR), is reported in the cubic matrix of the compounds [16,17]. The local symmetry of the PNR is of rhombohedral R3m and the polar axis of each PNR orientated randomly to eight directions of the cubic matrix. Thus, the polarizations are canceled and thus no averaged polarization appears. The volume fraction of PNR increases with decreasing the temperature, and reaches about 20% at about 200 K. The growth of PNR freezes at about 200 K and the size of PNR is limited to about 10 nm [13,17]. The formation of PNR is reported to start below about 600 K in PMN by Burns et al. [18]. In addition to the formation of PNR, the short-range ordering of B-site cations is reported in PMN by TEM observation [19e25]. The superlattice reflection indicates the presence of locally 1:1 ordered region in the compound. The region with about 2 nm size is called chemically ordered domain (COD) and is distributed randomly in the disordered matrix keeping the cubic symmetry on the average [25,26]. The heat capacity study is carried out for the typical relaxor to detect the formation and growth of PNR and COD in relaxors and to understand the mechanism of the relaxor properties from the thermodynamic point of view [28]. Figure 2 shows the dielectric constant of PMN and PMT measured at various frequencies. The broad dielectric peaks are observed

Nb2/3)O3 (PMN) and Pb(Mg1/3Ta2/3)O3 (PMT), are known, in which the transition metal ions are distributed almost B5þ ÞO3 , the randomly in B site [4]. In the case of PbðB3þ 1=2 1=2 long-range ordering of B site can occur and the dielectric properties are influenced by the degree of the ordering [5]. The sample with highly ordered B site shows normal ferroelectric or antiferroelectrics property and the sample with less ordered shows relaxor behavior. The relaxor property has been investigated vigorously and some microscopic models have been proposed [6e12]. However, the precise mechanism is still an open question and the thermodynamic investigation has been required to clarify the mechanism of the relaxor properties. PMN (Pb(Mg1/3Nb2/3)O3) is the relaxor found very early and studied most extensively. The crystal structural

FIGURE 1 Schematic crystal structure model of PMN.

Chapter | 6.3

Heat Capacity Study of Functional Ceramics

FIGURE 2 Real and imaginary parts of dielectric constant of PMN and PMT.

449

FIGURE 4 Debye characteristic temperature calculated from the heat capacity of PMN and PMT.

around room temperature in PMN and reache about 15000. The peak temperature in the real part decreases with decreasing the measurement frequency and the temperature of the corresponding peak in the imaginary parts also decreases, indicating the relaxation properties. The relaxor properties in PMT are similar to PMN except is the ones observed below about 100 K. Figure 3 shows the heat capacity measured using an adiabatic calorimeter. No remarkable heat capacity anomaly is observed in the figure around the temperature where the dielectric peaks are observed. However, the heat capacity curves of PMN and PMT are crossing at 250 K: the heat capacity of PMN is smaller than that of PMT below 250 K and vice versa above 250 K. It is very curious and indicates the existence of excess heat capacity in PMN above 250 K. Figure 4 shows the Debye characteristic temperature calculated from the heat capacity data using the equation
3 Z QD =T T x4 ex dx: Cv ¼ 3nR QD ðex 1Þ2 0

The material which completely follows the Debye lattice vibration model shows constant Debye characteristic temperature in the whole temperature range. However, the Debye characteristic temperature depends on the temperature in the real materials, and the typical example is that of BMT (Ba(Mg1/3Ta2/3)O3) shown in Figure 4. At very low temperatures, the Debye temperature is determined by acoustic phonons and scarcely depends on the temperature. With increasing the temperature, the Debye characteristic temperature shows a minimum around at QD /20 due to the effect of the zone boundary phonons and then approaches a certain value. However, the each curve of Debye characteristic temperature of PMN and PMT shows another minimum at 250 K and 320 K, respectively, as in Figure 4, which corresponds to the excess heat capacity in the materials. The excess heat capacity separated from the total heat capacity is shown in Figure 5. The excess heat capacity spreads over a wide temperature range from about 100 K to above 500 K in PMN. The onset temperature corresponds

FIGURE 3 Molar heat capacity of PMN and PMT.

FIGURE 5 Excess heat capacity due to the growth of PNR and dielectric constant in PMN and PMT.

450

Handbook of Advanced Ceramics

FIGURE 6 Heat capacity and dielectric constant of highly ordered PST.

to the Burns’ temperature [18] below which PNR is considered to begin to form in PNM. The peaks are observed around the temperatures where the dielectric peaks appear in both materials. These results indicate that the excess heat capacity corresponds to the formation of PNR in the relaxors. The entropy calculated from the excess heat capacity is 3.3 J K 1 mol 1 for PMN and 2.9 J K 1 mol 1 for PMT. The large values show that the formation of PNR relates to orderedisorder mechanism in the materials as expected from the disordered crystal structure. Assuming the orderedisorder mechanism, the expected entropy is calculated to DS ¼ Rln8 ¼ 17.3 J K 1 mol 1 for the perfect ordering, which is 5 times larger than that of the measured value. The small entropy value obtained form the excess heat capacity corresponds the fact that the growth of PNR does not fully progress in the crystal.

PST (Pb(Sc1/2Ta1/2)O3) is another type of relaxor, in which the ordering of B site is possible and influences the dielectric properties. The degree of ordering can be controlled by quenching and annealing at elevated temperature [28e32]. The highly ordered sample shows ferroelectric property and the less ordered shows relaxor behavior. The heat capacity and the dielectric constant of the highly ordered sample, which is grown by a flux method around 1200 K, are shown in Figure 6. The heat capacity shows a sharp peak due to the ferroelectric phase transition at 310 K and the corresponding sharp dielelectric peak is observed at the same temperature independently of the measuring frequency. Figure 7 shows the heat capacity and the dielectric constant of the less ordered sample which is quenched from 1830 K. The heat capacity peak moved to lower temperature and is smaller than those of the highly ordered sample. The

FIGURE 7 Heat capacity and dielectric constant of less ordered PST.

Chapter | 6.3

Heat Capacity Study of Functional Ceramics

451

FIGURE 8 Debye characteristic temperature calculated from the heat capacity of highly ordered and less ordered PST.

dielectric peak depends on the measuring frequency: the peak temperature decreases with decreasing the frequency, which is a characteristic feature of relaxor. Figure 8 shows the Debye characteristic temperature of both samples of PST. In the less-ordered sample, a broad anomaly like PMN is superimposed to the small dip due the remaining phase transition, indicating that the broad heat capacity anomaly is essential for the relaxor behavior. It is very interesting that the broad heat capacity anomaly is observed above the phase transition temperature. The entropy calculated from the excess heat capacity is 4.1 J K 1 mol 1 and 3.3 J K 1 mol 1. The values are still very small compared to the theoretical value and indicate the incompleteness of ordering in the both samples [27,33]. The detailed analysis of the mechanism is still needed.

3. THERMODYNAMIC STUDIES OF GIANT PARTICLE-SIZE EFFECT ON THE PHASE TRANSITION IN DIELECTRIC CRYSTALS It has been known that the crystal structure and the physical properties of materials are affected from its particles size. The typical example of creamics is of ferroelectric properties of BaTiO3. The large crystal of BaTiO3 shows the ferroelectric phase transition around 400 K. However, the phase transition temperature decreases with decreasing the particle size and the phase transition disappears at around 10 nm. The other phase transitions in BaTiO3 are also considered to behave similarly. Figure 9 shows the heat capacity of the BaTiO3 powders heated at various temperatures [34]. The fine powders are used as starting materials and the averaged particle size increases with increasing the heated temperature. The observed heat capacity peaks correspond to orthorhombic to tetragonal phase transition

FIGURE 9 Heat-capacity anomaly due the orthorhombic-tetragonal phase transition in BaTiO3 samples with different sintering temperature.

in BaTiO3 and grow with the heated temperature, thus the particle size. In general, these particle-size effects can be analyzed in terms of the influence of the surface of the fine particles. The relative population of the atom near surface increases drastically with decreasing the particle size especially in nanometer scale, and the surface effect becomes considerable. On the other hand, a peculiar particle-size effect on the phase transition is found in some dielectric substances, in which the boundary of phase transition behavior locates around 1 mm not at nanometers or micrometers. BZG (BaZnGeO4) is a ferroelectrics and shows an incommensurate phase transition. Figure 10 shows the heat capacity of BZG crystals [35]. Two heat capacity peaks are clearly seen in the crystal with large size: the higher-temperature peak corresponds to the known phase transition from phase VeIV and the lower phase transition is newly found (the lowest temperature phase is phase VI). However, the sample with less than 0.1 mm shows only one heat capacity peak and the phase VI does not appear. The crystal structure change due to the newly found phase transition is confirmed by the X-ray diffraction [36,37]. The surface effect should be ignored in millimeter size and another mechanism should be considered. More recently, CZP (CsZnPO4) is found to be another material showing similar giant particle size effect on

452

Handbook of Advanced Ceramics

FIGURE 10

Heat capacity of BZG with different particle size.

phase transition [38e40]. Figure 11 shows the heat capacity of CZP with different particle size. The sample with large size above 0.1 mm shows a heat capacity anomaly due to the phase transition around 220 K, but no heat capacity anomaly is observed in the samples below 0.1 mm. This behavior is confirmed by X-ray diffraction as shown in Figure 12. The single crystal of CZP shows a jump in the lattice parameter at 220 K in the heating direction and at 200 K in the cooling direction. The jump of the lattice parameter and the thermal

FIGURE 11 Heat capacity of CZP with different particle size.

FIGURE 12 Lattice parameters of single crystal and powder of CZP.

hysteresis clearly indicates the first-order transition nature. On the other hand, the powdered sample shows no jump in the lattice parameter and the temperature dependence indicates that the high temperature phase is supercooled until the lowest temperature. In addition to the giant particle-size effect on the phase transition, the phase transition shows a relaxation behavior. Figure 13 shows the heat capacity anomaly of the samples annealed below the phase transition temperature for various time intervals. The heat capacity anomaly grows with increasing the annealing time. The kinetics of the growth of the heat capacity anomaly can be analyzed using the Avrami model and the results are shown in Figure 14, indicating a heterogeneous nucleation and growth. The mechanism of this curious giant particle-size effect and the relaxation properties of the phase transition have not yet been understood clearly. However, the high-temperature phase is ferroelastic and the ferroelastic domain may influence the phase transition property. The detailed studies are required.

Chapter | 6.3

Heat Capacity Study of Functional Ceramics

453

is the measure of disorder in the materials and is directly determined only by the precise heat capacity measurements. The heat-capacity studies of the formation and growth of ferroelectric nanoregion in relaxors indicate that the entropy change due to the formation of the nanoregion is limited only one-fifth of the perfect ordering, and the disorder remains and freezes at low temperature. Such information can be obtained only by the low-temperature heat-capacity measurements, and the usefulness of the thermodynamic is clearly shown in the first topic. In addition, all the possible types of thermal agitation contribute to the heat capacity of materials; thus, the heat-capacity measurements have an advantage to find novel phenomena occurring with the temperature change as shown in second topic. The thermodynamic study of materials is classic but still effective in the investigations of the functional materials.

REFERENCES

FIGURE 13 Heat capacity anomaly due to the phase transition of CZP annealed below the phase transition temperature for various times.

FIGURE 14 Annealing time dependence of the phase transition enthalpy annealed at 195 K (O) and 200 K (C).

4. CONCLUSION Heat-capacity studies of functional materials have the importance that not only the temperature dependence of thermodynamic functions, which govern the satiability of the materials, can be determined but also the entropy which

[1] Lashley JC, Hundley MF, Migliori A, Sarrao JL, Pagliuso PG, Darling TW, et al. Critical examination of heat capacity measurements made on a Quantum Design physical property measurement system. Cryogenics 2003;43:369. [2] Mraw SC, Naas DF. The heat capacity of stoichiometric titanium disulfide from 100 to 700 K: absence of the previously reported anomaly at 420 K. J Chem Thermodyn 1979;11:567. [3] Cross LE. Ferroelectrics 1987;76:241. [4] Smolenskii GA, Isupov VA, Agranovskaya AI, Popov SN. Ferroelectrics with diffuse phase transitions. Sov PhyseSolid State 1961;2:2584. [5] Randall CA, Bhalla AS. Nanostructural-property relations in complex lead perovskites. Jpn J Appl Phys 1990;29:327. [6] Cross LE. Relaxor ferroelectrics: an overview. Ferroelectrics 1994;151:305. [7] Viehland D, Jang SJ, Cross LE, Wuttig M. Freezing of the polarization fluctuations in lead magnesium niobate relaxors. J Appl Phys 1990;68:2916. [8] Viehland D, Li JF, Jang SJ, Cross LE, Wuttig M. Dipolar-glass model for lead magnesium niobate. Phys Rev B 1991;43:8316. [9] Viehland D, Wuttng M, Cross LE. The glassy behavior of relaxor ferroelectrics. Ferroelectrics 1991;120:71. [10] Blinc R, Dolinsek J, Gregorovic A, Zalar B, Filipic C, Kutnjak Z, et al. Local Polarization distribution and EdwardsAnderson order parameter of relaxor ferroelectrics. Phys Rev Lett 1999;83:424. [11] Pirc R, Blinc R. Spherical random-bonderandom-field model of relaxor ferroelectrics. Phys Rev B 1999;60:13470. [12] Westphal V, Kleemann W, Glinchuk MD. Diffuse phase transitions and random-field-induced domain states of the ‘‘relaxor’’ ferroelectric PbMg1/3Nb2/3O3. Phys Rev Lett 1992;68:847. [13] Uesu Y, Tazawa H, Fujishiro K, Yamada Y. Neutron scattering and nonlinear outical studies on the phase transition of ferroelectric relaxor Pb(MgNb)O3. J Korean Phys Soc 1996;29:S703. [14] Bonneau P, Garnier P, Husson E, Morell A. Structural study of PMNceramics by X-ray diffraction between 297 and 1 023 K. Mater Res Bull 1989;24:201.

454

[15] Bonneau P, Garnier P, Calvarin G, Husson E, Gavarri J, Hewat AW, et al. X-ray and neutron diffraction studies of the diffuse phase transition in PbMg(1/3Nb2/3)O3 ceramics. J Solid State Chem 1991;91:350. [16] Mathan N, Husson E, Calvarin G, Cavarri JR, Hewat AW, Morell A. A structural model for the relaxor PbMg1/3Nb2/3O3 at 5 K. J Phys Condens Matter 1991;3:8159. [17] Fujishiro K, Uesu Y, Yamada Y, Dkhil B, Kiat JM, Yamashita Y. Bierefrigence and field dependence in lead magnesium niobate. J Korean Phys Soc 1998;32:S964. [18] Burns G, Dacol FH. Crystalline ferroelectrics with glassy polarization behavior. Phys Rev B 1983;28:2527. [19] Hilton AD, Barber DJ, Randoll CA, Shrout TR. On short range ordering in the perovskite lead magnesium niobate. J Mater Sci 1990;25:3461. [20] Bursill LA, Qian H, Peng JL, Fan XD. Observation and analysis of nanodomain textures in the dielectric relaxor lead magnesium niobate. Physica B 1995;216:1. [21] Yoshida M, Mori S, Yamamoto N, Uesu Y, Kiat JM. TEM observation of Polar domains in Relaxor ferroelectric Pb(Mg1/3Nb2/3)O3. Ferroelectrics 1998;217:327. [22] Husson E, Abello L, Morell A. Superstructure in Pb(Mg1/3Nb2/3)O3 ceramics revealed by high resolution electron microscopy. Mat Res Bull 1988;23:357. [23] Chen J, Chan HM, Harmer P. Ordering Structure and Dielectric Properties of. Undoped and La/Na-doped Pb(Mg1/3Nb2/3)O3. J Am Ceram Soc 1989;72:593. [24] Shrout TR, Huebner W, Randall CA, Hilton AD. Aging mechanisms in Pb (Mg1/3Nb2/3)O3-based relaxor ferroelectrics. Ferroelectrics 1989;93:361. [25] Boulesteix C, Varnier F, Llebaria A, Husson E. Numerical Determination of the local ordering of PbMg1/3Nb2/3O3 (PMN) from high resolution electron microscopy images. J Solid State Chem 1994;108:141. [26] Newnham RE. Chemistry of electronic ceramic materials. NIST Spec Publ 804, Chem Electronic Ceram Materials, Proc Inter Conf 1991;39. [27] Moriya Y, Kawaji H, Tojo T, Atake T. Specific-heat anomaly caused by ferroelectric nanoregions in Pb(Mg1/3Nb2/3)O3 and Pb(Mg1/3Ta2/3)O3 relaxors. Phys Rew Lett 2003;90:205901.

Handbook of Advanced Ceramics

[28] Setter N, Cross LE. The role of B-site cation disorder in diffuse phase transition behavior of perovskite ferroelectrics. J Appl Phys 1980;51:4356. [29] Setter N, Cross LE. The contribution of structural disorder to diffuse phase transitions in ferroelectrics. J Mater Sci 1980;15:2478. [30] Setter N, Cross LE. An optical study of the ferroelectric relaxors Pb(Mg1/3Nb2/3)O3, Pb(Sc1/2Ta1/2)O3, and Pb(Sc1/2Nb1/2)O3. Ferroelectrics 1981;37:551. [31] Stenger CGF, Scholten FL, Burggraaf AJ. Ordering and diffuse phase transitions in Pb(Sc0.5Ta0.5)O3 ceramics. Solid State Commun 1979;32:989. [32] Stenger CGF, Burggraaf AJ. Orderedisorder reactions in the ferroelectric perovskites Pb(Sc1/2Nb1/2)O3 and Pb(Sc1/2Ta1/2)O3. Physica Stat Sol 1980;A61(275):1980. ibid. A61, 653. [33] Moriya Y, Kawaji H, Tojo T, Atake T. Heat capacity anomaly due to the ferroelectric transition in Pb(Sc1/2Ta1/2)O3. Trans Mater Res Soc Jpn 2002;27:287. [34] Fukui Y, Izumisawa S, Atake T, Hamano A, Shirakam i T, Ikawa H. Effects of heat treatment on the phase transitions in BaTiO3 fine particles. Ferroelectrics 1997;203:227. [35] Hamano A, Atake T, Saito Y. Thermodynamic studies of successive phase transitions in BaZnGeO4 and a new solid phase below 186.1 K. Appl Phys A 1989;49:91. [36] Yamamoto N, Kikuchi M, Atake T, Hamano A, Saito. Incommensurate-commensurate phase transition of BaZnGeO4 studied by electron microscopy and diffraction, Y. J Phys Soc Jpn 1992;61:3178. [37] Takai S, Hamano A, Atake T, Ishizawa N, Marumo F. X-ray diffraction study on phase relation of BaZnGeO4. Jpn J Appl Phys 1993;32:4635. [38] Yamashita I, Tojo T, Kawaji H, Atake T. Thermal studies of the IV-III phase transition in CsZnPO4. Phys Rev B 2002;65:212104. [39] Yamashita I, Kawaji H, Atake T, Kuroiwa Y, Sawada. Orderdisorder mechanism of the I-II phase transition in CsZnPO4, A. Phys Rev B 2003;67:014104. [40] Yamashita I, Kawaji H, Atake T, Kuroiwa Y, Sawada A. Particlesize effect on the III-IV phase transition in CsZnPO4. Phys Rev B 2003;68:092104.

Chapter 6.4

New Frontiers Opened Up Through Function Cultivation in Transparent Oxides Hideo Hosono Frontier Collaborative Research Center and Materials and Structures Laboratory, Tokyo Institute of Technology, Mail Box R3-1, 4259 Nagatsuta-cho, Midori, Yokohama, Kanagawa 226-8503, Japan

Chapter Outline 1. Introduction 2. Research Concept and Strategy 3. Light metal TCO: 12CaO$7Al2O3 with Built-in Nano-Porous Structure 4. Transparent Amorphous Oxide Semiconductors

455 456 457 468

1. INTRODUCTION Top 10 of the Clerk number showing the order in natural abundance of elements in the earth crust are listed as oxygen, silicon, aluminum, iron, calcium, sodium, potassium, magnesium, hydrogen, and titanium. This means that our territorial environment consists of oxides of these abundant metals and water. Human beings have developed civilization utilizing these oxides to date, i.e., Al2O3, SiO2, CaO, and MgO for the Stone Age, Fe for the Iron Age, and the present Information Age is supported by Si semiconductors and SiO2 glass fibers. It is no doubt that the coming age will be created through utilizing these abundant materials intelligently. Oxide ceramics is probably among the oldest of manmade materials owing to the abundance and easy availability of the ingredients. Although most oxides are optically transparent, which is important for optical applications, it has been believed that active functions based on “excited electrons”, such as in crystalline semiconductor materials, are not possible. For example, alumina and glasses, which are representative oxides, are optically transparent but electrically insulating. However, during the past two decades techniques to purify oxides have drastically advanced. It is now possible to obtain highly pure materials, resulting from concentrated studied on fine ceramics in the 1980s. For instance,

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00027-7 Copyright Ó 2013 Elsevier Inc. All rights reserved.

5. Iron-Pnictide Superconductors 6. Future Challenge: Ubiquitous Element Strategy Acknowledgments References

477 484 485 485

metal oxide chemicals containing sub-ppm impurities are now commercially available. Furthermore, the thin film deposition technique of oxides was much advanced through intensive study on high Tc superconductors for electronic applications in the past decade. Taking advantage of this situation, our group has explored transparent conducting oxides (TCOs) while following a working hypothesis established based on a consideration of chemical bonding and point defects, resulting in more than ten new TCOs. Important among them are p-type TCO, CuAlO2, reported in Nature (1997) [1] and a series of amorphous TCOs in which the Fermi level is controllable by intentional doping (1996) [2]. The former is particularly important because most active functions in semiconductors originate from PN-junctions [3]. The absence of practical application of transparent oxide semiconductors is due primarily to the missing of P-type TCO. It is therefore now possible to examine the possibility of invisible circuits based on transparent oxides as shown in Fig. 1. As for optical materials, Hosono et al. reported novel photosensitive or photorefractive glasses [4], SiO2 glasses [5], and a CaF2 crystal [6] for excimer laser applications. These materials were found through accumulated fundamental research on defects in glass and dielectric crystals. This study started from October 1, 1999 under such a situation.

455

456

FIGURE 1 Impact of discovery of a p-type transparent conductive oxide. For color version of this figure, the reader is referred to the online version of this book.

2. RESEARCH CONCEPT AND STRATEGY A research concept “Transparent Electro-Active Materials (TEAM)” was born in a dream “we wish to convert alumina to transparent semiconductor”. Transparent oxides are most abundant and stable on earth, and environmentally friendly. Although they have been used as ingredients of traditional materials such as porcelain, cement, and glass, few active functions have been found in them. In fact, these materials are described as typical insulators in college textbooks. A widely accepted view “a transparent oxide cannot be a platform for electro-active materials” comes from phenomenological observation. I think it possible to realize a variety of active functionalities for transparent oxides by appropriate approach based on an original view for these materials. What is unique for transparent oxides? I think there are 2 characteristic features: a wide variety of crystal structures and that constituting elements are characteristic of oxides. No such vast variety is seen in typical semiconductor or compound semiconductors. The crystal structure of the latter is limited to the diamond- type. This striking difference arises from that in the chemical bonding nature between oxides and compound semiconductors, i.e., although the

Handbook of Advanced Ceramics

ionic nature controlling a long range ordering is primary in oxides, the covalent nature determining the local coordination structure contributes to the bonding significantly, depending on the type of a metal cation. On the other hand, bonds in compound semiconductors are almost covalent and, as a consequence, local coordination configuration is tetrahedral. This feature results in the striking difference in nature between hole and electron pathways. The conduction band minimum (CBM) that works as an electron pathway is primarily composed of metal vacant orbitals, whereas 2porbitals of electro-negative oxygen ions dominantly contribute to the valence band maximum (VBM) which functions as a hole pathway. This electronic nature is the basis for materials designing in transparent oxide semiconductors. Figure 2 summarizes the chemical bonding in oxides in comparison with a pure covalent compound and a pure ionic compound. Here we focus on transparent oxides with low-dimensional crystal structure, i.e., nano-porous structure and lowered structure. If an electron could be injected to the nano-sized cage, the resulting cages may be regarded as coupled quantum dot system for which one may expect a novel electronic structure. For layered structures composed of a narrow semi-conducting layer sandwiched by wide gap oxides, we may anticipate a unique electronic state of a 2D-electronic nature. Another is a variety of excited states. Although transparent oxides composed of abundant elements appear less interesting in the ground state, there are so many interesting excited states. Representative examples are a self-trapped exciton that works as energy localization center, leading to persistent defect formation and a polaron composed of a charge carrier and its associated polarization and host lattice deformation field. This feature provides valuable opportunities to modify transparent dielectric crystalline and amorphous oxides toward novel functions and devices by electronic excitation. Several representative articles on this subject are described in the literature [7e11]

FIGURE 2 Characteristics of transparent oxides in comparison with covalent semiconductor and ionic crystal. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 6.4

New Frontiers Opened Up Through Function Cultivation in Transparent Oxides

3. LIGHT METAL TCO: 12CAO$7AL2O3 WITH BUILT-IN NANO-POROUS STRUCTURE A wide variety of materials and crystal structure is a unique feature of oxides. Electro-conductive oxides are, however, restricted to transition metal or heavy metal cation-bearing materials. Representative light metal oxides that are most abundant in our territorial crust are classified as typical transparent insulators as described in textbooks. They are the main constituents of traditional ceramics such as porcelain, cement, and glass, and no one has succeeded in converting them to persistent electronic conductors until a report [12] in 2002 utilizing the nano-structures built-in crystal structure of 12CaO$7Al2O3 (C12A7). This section describes the electronic structure of C12A7 and various approaches to electron doping along with unique properties of electron-doped C12A7.

3.1. Crystal Structure of 12CaO$7Al2O3 This compound can be easily prepared by a conventional solid-state reaction in air, but is not formed in O2 or moisture-free atmosphere [13]. The crystal lattice of 12CaO$7Al2O3 (C12A7) with a microporous structure has a cubic lattice constant of 1.199 nm, and the unit cell includes two molecules [14]. It is characterized by a positively charged lattice framework [Ca24Al28O64]4þ having 12 crystallographic cages with a per unit cell as shown in Figure 3. Each cage with an inner diameter of ~0.44 nm and a formal charge of þ 1/3 is coordinated by 8 cages to form a 3-dimensional structure by sharing a monomolecular open mouse. This structure is a similar to a zeolite but the excess charge of the cage wall is reverse. Free oxygen ion O2 accommodates in 1/6 out of 12 cages in the unit cell to reserve electroneutrality in the stoichiometric state. The remaining two oxide ions (O2 ) are entrapped as counter anions in the cages in the stoichiometric composition and called free oxygen ions because the cage diameter is larger by ~50% than the diameter of

457

O2 . This fact is associated with the fast oxygen ion conduction in this material [15]. Monovalent anions, such as hydroxyl ions (OHe) or halogen anions, X (X ¼ F and Cl ), are known to substitute for a part or all of the free oxygen ions to form derivatives [Ca24Al28O64]4þ$4(X ). The replacement of a free O2 ion by two monovalent anions likely stabilizes the structure through charge delocalization [16]. C12A7 crystals can be easily synthesized by conventional solid-state reactions using CaCO3 and Al(OH)3 or Al2O3. Single crystals can be grown from the melt by using a floating-zone (FZ) method [17] or the Czochralski (CZ) technique [18].

3.2. Electronic Structure and Our Approach Figure 4 shows the band structure of stoichiometric C12A7 calculated by density functional method with GGA approximation using VASP code. Two major features are evident from the figure [19]. (1) The O2p levels of the free oxygen ions are located ~1 eV above the valence bands. (2) There are two types of conduction bands, i.e., one is a conventional conduction band that is primarily composed of Ca 4s levels, and is called the framework conduction band (FCB). Another is a conduction band composed of 3dimensionally connected sub-nanometer-sized empty cages, which is not comprised of orbitals of specific ions constituting the cage wall but is constituted of a coupled “particle-in-a box”, and is called the ‘cage conduction band (CCB)’ [19e21]. CCB, which is unique to C12A7, is located ~2 eV below from the bottom of FCB and has moderate energy dispersion [22]. These features led us to an idea that if the free oxygen ions are replaced by electrons, the resulting state may exhibit high electronic conduction; the energy level of the FCB bottom is too high for electron doping, whereas the location of CCB is marginal for successful electron doping compared in the doping limit [23,24] for other n-type semiconductors. Various methods to inject electrons to the CCB are described in the next section.

3.3. Electron Doping 3.3.1. Light-Induced Doping for C12A7:H

FIGURE 3 Crystal structure of C12A7. Free O2 ions randomly distributed in the cages are omitted for simplicity. For color version of this figure, the reader is referred to the online version of this book.

Hayashi et al. [12] demonstrated a process by which the transparent insulating oxide C12A7 can be converted into an electronic conductor. H ions are incorporated into the sub-nanometer-sized cages of the oxides by a thermal treatment in a hydrogen atmosphere at 1300  C; subsequent illumination of the resulting C12A7:H with ultraviolet light (l < 350 nm) results in a conductive state that persists after irradiation ceases. The photo-activated material exhibits moderate electrical conductivity (~0.5 S cm 1) at room temperature, with visible light absorption losses of

458

Handbook of Advanced Ceramics

FIGURE 4 Band structure of C12A7. Note that there are two types of conduction bands, and 2p levels of free O2 ions are located above the framework valence band top. The origin of cage conduction band is electron tunneling through the thin cage wall. For color version of this figure, the reader is referred to the online version of this book.

only one percent for 200-nm thick films, as shown in Figure 5. The incorporation of H anions in the cage may be understood through the following chemical reactions: O2 ðcageÞ þ H2 ðatmosphereÞ/OH ðcageÞ þ H ðcageÞ; (1) or O2 ðcageÞ þ H2 ðatmosphereÞ/1=2O2 ðatmosphereÞ þ 2H ðcageÞ:

(2)

Specimens C12A7:H look white as polycrystalline powder or colorless and transparent for single crystals and is insulating (conductivity < 10 10 S cm 1 at RT). However, upon illumination with ultraviolet light (wavelength < 330 nm), two absorption bands centered at 2.8 eV and 0.4 eV are induced, and simultaneously, a drastic increase in the conductivity was observed along with the coloration. The conductivity increased from Nth (inset of Figure 9). These observations demonstrate that insulating C12A7 is converted to metallic state by replacing more than half of the free oxygen ions by electrons. This IM transition accompanies a sharp increase in the drift mobility (me) from ~0.1 to 4 cm2 V 1 s 1 as Ne increases. It should be noted that the doping and conduction mechanisms in C12A7 are much

460

FIGURE 9 Conductivity extrapolated to 0 K and mobility as a function of carrier electron concentration. When each O2 ion in the cages is replaced by 2 electrons, Ne corresponds to 2.3  1021 cm 3. For color version of this figure, the reader is referred to the online version of this book.

different from those in conventional semiconductors due to the unique crystal and electronic structures of C12A7. In typical semiconductors such as phosphorus-doped Si [29], IM transition occurs due to the percolative overlap of

Handbook of Advanced Ceramics

the donor wave functions as explained by Mott in terms of the critical donor concentration Nc, which follows the empirical relation N1/3 c $Rc ~ 0.25 (Rc denotes the radius of the electron orbital) [30]. When the Nth value of ~1  1021 cm 3 is put into this relation, the obtained Rc value is of ~0.26 nm, which is close to the radius of the crystallographic cage embedded in C12A7 (~0.25 nm). Thus, it is feasible that the IM transition in C12A7 occurs via the percolation of the ‘donor’ orbitals. However, this view cannot be directly applicable to C12A7 because even an ‘undoped’ stoichiometric C12A7 crystal has a ‘donor wave function’ in each empty cage as a component of the CCB. Although the ‘donor wave function’ intrinsically exists, the electrons doped to the CCB do not contribute to the band conduction at Ne < Nth at low temperatures because the electron is localized in a cage forming an Fþ-like center [12,25] due to the lattice relaxation, as described above. However, a high electron doping may average out the deformations of the cage geometry throughout the crystal and it converts the localized F þ-like state to a delocalized state, invoking an IM transition. This model for the IM transition is validated by MEM/ Rietveld analysis and ab-initio density functional theory (DFT) calculations [28]. Figure 10 displays the electron density maps in the (001) plane of stoichiometric C12A7

FIGURE 10 Electron density map of cages in insulating and metallic conducting C12A7 determined by MEM-Rietvelt analysis of X-ray diffraction pattern. This density is on the (001) plane. Note that two Ca2þ ions at the axial position are split into 2 sites corresponding to an O2e entrapped cage and untrapped cage. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 6.4

New Frontiers Opened Up Through Function Cultivation in Transparent Oxides

and metallic C12A7 obtained by the MEM/Rietveld analysis [31]. The electron densities of the Ca2þ ions on the cage wall of stoichiometric C12A7 have long tails like a dumbbell shape, indicating that the map is a superposition of the two deformed O2 -encaging cages and the ten undeformed empty cages. The distances between the Ca2þ ions (DCaeCa) in the deformed and ˚ empty cages obtained by the Rietveld analysis are 4.22 A ˚ and 5.65 A, respectively. Thus, the map in Figure 10 provides solid evidence that the O2 -encaging cage shrinks largely from the empty cage. On the other hand, the map in Figure 12c indicates that the cage structure of [Ca24Al28O64]4þ(4e ) clearly differs from that of

461

stoichiometric C12A7. Each atom has a spherical electron density, indicating that the local structural deformation is removed in [Ca24Al28O64]4þ(4e ). That is, all ˚) the twelve cages are homogenized and DCaeCa (5.64 A becomes close to that of the empty cage, suggesting the electrons are distributed uniformly over the twelve cages within the accuracy of the MEM/Rietveld analysis. Figure 11 shows relaxed lattice structures calculated by DFT with the isosurfaces of the wave function (jJj2) of 2p level of the free oxygen ions and those of the CCB bottom (isosurface state density) for (i) stoichiometric C12A7, (ii) [Ca24Al28O64]4þ(O2 )$(2e ) and (iii) metallic C12A7. The DFT results are summarized in

FIGURE 11 Isosurface state density by DFT calculation on relaxed structure and schematic band structure of C12A7. For color version. For color version of this figure, the reader is referred to the online version of this book.

462

Handbook of Advanced Ceramics

FIGURE 12 Direct synthesis of electron-doped C12A7 from the melt and glass. Process and photos of the resulting samples are shown. For color version of this figure, the reader is referred to the online version of this book.

Figure 13 as a schematic energy level diagram. Figure 11a(i) demonstrates that two free oxygen ions respectively occupy the cages at the center and the corners of the unit cell and the other cages are empty in the stoichiometric C12A7. The calculated DCaeCa of the ˚ , in good agreement with O2 -encaging cage is 4.27 A ˚ . Further, each empty cage the observed value of 4.22 A has an s-like state, and the connection of these states forms the CCB even in the stoichiometric C12A7. On the other hand, the O2 -encaging cage has no such CCB state because of the entrapping of O2 , whose 2p energy level is located slightly above the valence band maximum of the cage framework (Figure 11c(i)) Thus, stoichiometric C12A7 has no electron in the CCB and is a band insulator [22]. FIGURE 13 Electrical conductivity in polycrystalline C12A7 thin films as a function of implanted Arþ fluence. Right panel shows infrared absorption spectra of OH in C12A7 thin films subjected to various treatments. For color version of this figure, the reader is referred to the online version of this book.

The DFT result for [Ca24Al28O64]4þ(O2 )$(2e ), where a free oxygen ion is extracted from a cage and two electrons occupy two of the other cages in the unit cell, ˚ ) of the electron-encagdemonstrates that DCaeCa (5.57 A ing cage becomes larger compared to that of the O2 ˚ ), but it is still smaller than that of encaging cage (4.27 A ˚ ). The wave function at the CCB the empty cage (5.77 A minimum (Figure 11b(ii)) implies that the electron is still localized in a specific cage, forming an Fþ-like center (Figure 11c(ii)). Figure 11b(iii) is the relaxed lattice structure with the isosurface state density at the CCB minimum of ˚) [Ca24Al28O64]4þ(4e ), revealing that the DCaeCa (5.63 A of the electron-encaging cage further expands from that of [Ca24Al28O64]4þ(O2 )$(2e ) and it becomes very close to

Chapter | 6.4

New Frontiers Opened Up Through Function Cultivation in Transparent Oxides

that of the empty cage. It also indicates the isosurface state density in each cage is equivalent, indicating that the electrons are delocalized over the cages and consequently the cage conduction band has a rather large dispersion of ~2 eV. Thus, the DFT calculation, consistent with the MEM/ Rietveld analysis, qualitatively explains why C12A7 shows IM transition upon electron doping. That is, the successive reduction of deformation of the electron-encaging cage with increasing the electron concentration induces the IM transition, which accompanies the sharp increase in the electron mobility. It is noteworthy that simple ionic oxides such as MgO have not exhibited such a high conductive state although the F/Fþ-like centers can be incorporated at high concentrations similar to the present case [32]. Thus, the three-dimensionally connected sub-nanometer-sized cage structure in C12A7 plays an essential role in the appearance of the metallic state.

3.3.3. Doping via Melt Processes The development of a simple synthesis process, which is compatible for mass production of C12A7:e is strongly required for practical applications. Possible candidate to satisfy the requirement is a direct synthesis of C12A7:e from the melt or glass state. Extensive studies regarding the relationship between the melting condition and corresponding precipitated phases in the CaO-Al2O3 system [33] verify that the presence of template anions to compensate the positive charge of the C12A7 framework is inevitable for the precipitation of the C12A7 phase. More specifically, the C12A7 phase is reproducibly prepared from the stoichiometric C12A7 melt in wet air [13], involving O2 and/or OH ions that act as the template. On the contrary, the thermal treatment in inert gas atmospheres, where template anions are seemingly absent, hinders the formation of the C12A7 phase and only a mixture phase of C3A (3CaO$Al2O3) and CA (CaO$Al2O3) is formed. If we could precipitate the crystalline C12A7 under an appropriate reducing atmosphere, there is a possibility of direct synthesis of C12A7:e from the melt or glass. The critical issue for the direct synthesis is to find a novel template ion that can exist in the melt or glassy phase even when they are exposed to reducing atmospheres. With efforts in searching for such templates, successful results were obtained for the process via the reduced melt [34] or glass state [34], where the C22 ion most likely acts as a template for the C12A7 phase formation [35]. Figure 12(a) shows schematically a direct solidification process to prepare C12A7:e from the C12A7 melt. C12A7 powders of the stoichiometric mixture of CaO and Al2O3 are melted in a semi-airtight carbon crucible with a carbon cap at 1600  C for 1 hr in air. Heating the carbon crucible at 1600  C produces a strongly reducing atmosphere

463

(PO2 ¼ ~10 16 atm) inside the crucible. The crucible is slowly cooled down at a rate of ~400  C/hr. X-ray diffraction measurements clarify that solidified products are mixtures of C3A and CA. Then, they are re-melted and solidified again according to the same temperature profile as that of the first run. The second run results in the formation of the C12A7 crystalline phase with green color and electrical conduction. Raman spectra of the products after the first run and thermal desorption measurement in the cooling process in the second run suggest that C22 ions are dissolved into the melt from the carbon crucible to compensate for the oxygen deficiency caused by the strongly reducing atmosphere. ˚ ) is similar to Furthermore, the ionic radius of C22 (1.2 A 2 2 ˚ O (1.4 A), which supports that the C2 anions serve as the template in a reducing atmosphere, instead of the free oxygen ions in an oxidizing atmosphere. Moreover, the C22 Raman band is not observed in the C12A7:e and CO gas is evolved in the cooling process, which implies that the encaged C22 ions are only stable in the nucleation and/or initial stage of the crystallization and they are released from the cages, leaving electrons in the lattice during the cooling process. Thus, the electron formation processes in the cages may be expressed as follows: C22 ðmeltÞ / C22 ðcageÞ / 2CðsolidÞ þ 2e ðcageÞ; (3) and/or C22 ðcageÞ þ 2O2 ðcageÞ / 2COðatmosphereÞ þ 6e ðcageÞ

(4)

Figure 12b shows schematically the glass-ceramics process for the preparation of C12A7:e . Oxygen-deficient glass, which is obtained by a rapid quench of the C12A7 melt in a carbon crucible placed in air, is crystallized in an evacuated silica tube by heating to 1000  C, which is higher than the crystallization temperature (>900  C). The resultant C12A7 samples were green in color and exhibited rather high electrical conductivity, indicating that electrons are incorporated into the cages instead of free oxygen ions. Similar to the direct solidification process, C22 ions, which are likely incorporated in the glass, are considered to act as the template. The resulting C12A7:e having 1019e1020 cm 3 exhibits 0.1e1Scme1 at RT.

3.4. Thin Film Fabrication of Electron-Doped C12A7 Thin film fabrication is needed for device application of C12A7:e. Bulk formation processes described above are not applicable for thin film fabrication because the thickness of reaction layers formed by metal treatment is comparable to the conventional film thickness. Thus, different processes

464

Handbook of Advanced Ceramics

are required to be invented. Two processes to be described below have been reported for this purpose.

3.4.1. Ion Implantation Ion implantation, a nonequilibrium physical process, has such advantages over conventional chemical doping processes that high concentration of ions can be introduced in a designated local area with an excellent controllability [29,30], and thus it is widely used for the modification of thin films. Miyakawa et al. [36,37] conducted Arþ ion implantation into polycrystalline C12A7 thin films and found that the electron anions are incorporated into the cages as a result of kicking out the free oxygen ions from the cages through the nuclear collisions. The polycrystalline C12A7 thin films (~800 nm in thickness) were prepared on MgO(100) single-crystal substrates by a pulsed laser deposition technique at RT and a subsequent annealing at 1100  C for 1 hr in air to crystallize the deposited amorphous films. Ion implantation was carried out at temperatures from RT to 600  C in a vacuum less than ~2  10 8 torr. The Arþ fluence (accumulated dose) set to vary from 1  1016 to 1  1018 cm 2. Moreover, the C12A7 framework is stable for hot implantation, although a crystalline film is converted to amorphous phase if the implantation is performed at RT. Figure 13 shows the electrical conductivities of 300 keV Arþ-implanted films subjected to the implantation at high temperature of 600  C as a function of Arþ-ion fluence. The electrical conductivity of the film increases with implanted fluence: films with fluences less than 1  1017 cm 2 are insulators, while those with 5  1017 to 1  1018 cm 2 exhibit electrical conductivity of ~0.5 and 1 S cm 1, respectively. The enhancement of the electrical conductivity is accompanied by the appearance of optical absorption bands at 0.4 and 2.8 eV, indicating the F þ-like centers are simultaneously formed by the hot implantation process. These results suggests that the free oxygen ions are kicked out from the cages by the collision between the free oxygen ions and Arþ ions, leaving electrons in the cages. As a result, the electrons are directly produced by Arþ-implantation via the extrusion of free oxygen ions through the following reaction: O2 ðcageÞ / 1=2O2 ðatmosphereÞ þ 2e ðcageÞ

introduced during the post-annealing process for the crystallization of amorphous films in air. That is, electrons are generated by the extrusion of free oxygen ions through reaction (5) and simultaneously protons (Hþ) are created by the decomposition of preexisting OH groups in the asprepared film through reaction (6): OH ðcageÞ / O2 ðcageÞ þ Hþ ðattached to cage wallÞ (6) Then, Hþ ions immediately react with the electrons in the cages, leading to the formation of H ions as follows: 2e ðcageÞ þ Hþ ðattached to cage wallÞ / H ðcageÞ (7) Thus, H ions are formed in lower Arþ-implanted samples, leading to light-induced electronic conduction and coloration. Figure 14 shows the schematic illustration of the electron and H formations by Arþ-implantation as mentioned above. Arþ-implantation produces the electrons and protons in the cages by nuclear collisions and electronic excitation effects through the reactions (5) and (6). When the number of the electrons produced by reaction (5) is less than that of protons, all produced electrons are captured by Hþ ions through reaction (7) and H ions are formed in the cages. However, when the generated electron concentration

(5)

It is noted in samples implanted with less than 1  1017 cm 2 Arþ ions that the conductivity and the optical absorption intensities increase after UV-light irradiation. Since these phenomena are commonly observed in H incorporated C12A7, it suggests that H ions are created into the cages by the hot implantation of Arþ ions. It was considered that the formation of the H ions by Arþimplantation is caused by preexisting OH groups that are

FIGURE 14 Mechanism for electron carrier generation in C12A7 thin films by inert gas implantation. There are two independent processes, direct knock-on of free O2 and photo-ionization of H that was derived from OH in the cage by action of energetic ion.

Chapter | 6.4

New Frontiers Opened Up Through Function Cultivation in Transparent Oxides

465

FIGURE 15 Fabrication process of C12A7:e and S12A7:e thin films. For color version of this figure, the reader is referred to the online version of this book.

exceeds the proton concentration, the excess electrons themselves are directly introduced into the cages. Since higher concentration of the electrons is incorporated three-dimensionally in a designated narrow space, the synthesis of the C12A7:e thin films by ion implantation paves a way to realize novel applications such as patterned electric wiring, electron injection electrode, and cold electron emitter.

3.4.2. Chemical Reduction

obtained S12A7:e thin films by tuning the temperature and atmosphere for the post-annealing of a-S12A7. Figure 16 shows the DC conductivities of resulting C12A7:e and S12A7:e thin films as a function of temperature. It is evident that both samples exhibit metallic conduction. The Hall mobility and carrier concentrations in the C12A7:e film were evaluated to be 2.5 cm2(Vs) 1 and 1.3  1021 cm 3, respectively. Similar values (1.0 cm2(Vs) 1 and 1.4  1021 cm 3) were obtained for the S12A7:e thin films. No carrier doping was observed in the same conditions without

Miyakawa et al. [38] found a method for fabrication of C12A7:e thin films by using a deposition of oxygen-deficient amorphous C12A7 layer on crystalline C12A7 thin films. Figure 15 shows the processes. First, amorphous thin films of ~400 nm thick were deposited at RT by the pulsed laser deposition (PLD) method using ceramic C12A7 targets. The partial O2 pressure was ~1  10e3 Pa during the depositions, and the substrates employed were MgO(100). Polycrystalline(p-) C12A7 thin films were obtained by heating the resulting amorphous (a-) C12A7 thin films to 1100  C in an ambient atmosphere. Next, a-C12A7 (~30 nm thick) layers were deposited on the p-C12A7 thin films at 700  C in an O2 pressure of ~1  10e3 Pa by PLD. Last, the over-layered a-C12A7 layer was removed by chemicalemechanical polishing (CMP). By tuning the duration and temperature of post-annealing temperature, the electron concentration in C12A7 thin film can be increased up to ~2  1021 cm 3 (the theoretical maximum). 12SrO$7Al2 O3 (S12A7) with the same crystal structure as C12A7 is known as a meta-stable phase [39]. Miyakawa et al. [40] applied this method to S12A7 and successfully

FIGURE 16 Electrical conductivity in electron-doped 12SrO$7Al2O3 thin films. Numbers in the panel denote Carrier electron concentrations measured by the Hall effect. Data on electron-doped 12CaO$7Al2O3 thin film are shown for comparison. For color version of this figure, the reader is referred to the online version of this book.

466

deposition of a-C12A7 on pc-C12A7 or S12A7. Deposition of Au layers or Ti layers was examined for carrier generation. The former did not work, whereas the latter did. These observations imply that the role of a-C12A7 deposition on p-C12A7(S12A7) is to exact free O2 accommodated in the cages and inject electrons into the cages.

3.5. Unique Properties of Electron-Doped C12A7 and Applications C12A7:e is an exceptional material that exhibits metale insulator transition in light metal oxides. Their maximum conductivity at RT is 1500 S cm 1, which is higher by a factor of 2 than that of graphite and is comparable to metal manganese. The optical transmittance of thin films (100 nm thick) is >60% in the visible range. The conductivity and visible transmittance of metallic C12A7:e is obviously inferior to that of ITO compared as TCO, but C12A7:e has totally unique physical properties, i.e., very low work function and superconductivity.

3.5.1. Low Work Function Toda et al. [21] evaluated work function (fWF) of C12A7:e by photoelectron yield spectroscopy (PYS) using bulk C12A7:e single crystals. A fresh surface of the sample was prepared by filing the surface with a diamond file in a preparation chamber attached to the measurement chamber under a vacuum of ~5  10e7 Pa before each photo-electron measurement. Figure 17a shows photoemission current (IPE) as a function of VBIAS at different excitation photon energies (hn). IPE rapidly increases with VBIAS up to 10 V and levels off. IPE is related to hn according to Eqn (8): 
hn fWF þ Z; (8) lnIPE T 2 ¼ lnF kB T

Handbook of Advanced Ceramics

where FðxÞ ¼

ZN

y 1 þ expðy

xÞ

dy

0

Here kB is the Boltzmann constant, T the temperature, and Z a constant. Figure 17b shows (IPE)1/2ehn plots (the approximated Fowler plot) for samples A and B with different Ne at VBIAS values of 10 V, 20 V, and 30 V. The fWF value was determined to be 2.4 eV by obtaining the best-fit result to Equation (8). It is seen that this fWF value gives fairly good agreement with the theory for all the VBIAS values especially at hn around the fWF value. These results indicate that the vacuum level (EVAC) locates at ~2.4 eV above EF, ~0.4 eV above FCBM, and ~7.9 eV above VBM (see Figure 18). Here we consider the reason why C12A7:e has a lower fWF than those of alkali and alkaline earth metals (Na ¼ ~2.8 eV, K ¼ ~2.3 eV, Rb ¼ ~2.2 eV, Cs ¼ ~2.1 eV, and Ca ¼ ~2.9 eV) and why C12A7:e is chemically stable regardless of the low fWF value. Figure 18 compares the energy structures of the Fþ center (an oxygen vacancy trapping an electron) in CaO and CCB in C12A7:e . CCB in C12A7:e is composed of 3-dimensionally connected sub-nanometer-sized cages, in some of which electrons are entrapped. An electron in the cage is octahedrally coordinated with 6 Ca2þ ions constituting the cage wall. Thus, the electron-trapped cage is an analog to the Fþ center in CaO except for two points: (1) the separation between the electron and the Ca2þ ion is longer by ~50 % than that in the Fþ center in CaO, and (2) an oxygen ion directly bonding with the Ca2þ ion in the lattice framework of C12A7:e is connected one or two tetrahedrally coordinated Al3þ ions. As a consequence, the net negative charge affecting at the cage center is smaller than that at the Fþ center in CaO. These differences cause a smaller Madelung potential for C12A7:e than for CaO,

FIGURE 17 Evaluation of work function (FWF) in C12A7:e single crystals by photoelectron yield spectroscopy. Two samples with different carrier electron concentrations (1  1019 or 2  1021 cm 3) were used for the measurements. Photocurrent vs. bias voltage for different excited photon energy. (b) I1/2 PE ehn plots. The intercept yields fWF of 2.4 eV. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 6.4

New Frontiers Opened Up Through Function Cultivation in Transparent Oxides

467

FIGURE 18 Comparison of electronic energy levels of C12A7:e with a Fþ center in CaO. Note that the energy level of the trapped electron at the site of oxygen vacancy is different by 3.6 eV between the Fþ center in CaO and CCB in C12A7, whereas the energy gap and the location of conduction band minimum (CBM) and valence band maximum (VBM) are almost the same. For color version of this figure, the reader is referred to the online version of this book.

resulting in the ~3.6 eV shallower electronic levels in the cages of C12A7:e than at the oxygen vacancy site in CaO, as shown in the figure. The chemical stability of C12A7:e compared with the alkali and alkaline earth metal oxides is also understood from its crystal structure. When C12A7:e is exposed to an ambient atmosphere, the rigid structure of the lattice framework made of CaeO and AleO bonds prohibits in-diffusion of H2O and O2 molecules into a cage to react with the low work function electrons at around room temperature. This geometrical restriction is the primary cause for chemical and thermal stability of C12A7:e , providing an opportunity for practical application. Recently, several applications of C12A7:e have been reported., i.e., cold [41] and thermoionic [42] electron emission sources, cathode materials in organic light-emitting diode(OLED) [43], and efficient electrode material in fluorescent lamps [44]. Better performance is obtained by using C12A7:e for each application compared with the currently used materials. These are examples of straightforward applications based on unique properties of C12A7:e. An intriguing application has been found for a chemical reducing reagent available for various organic reactions in aqueous media [45].

3.5.2. Superconducting Transition [46] In order to investigate whether the superconducting transition occurs in the metallic C12A7:e or not, two samples exhibiting the highest conductivity in each crystal growth method (FZ and CZ) were used for measurements of

electrical properties at low temperatures. Here, the C12A7 electride denotes that C12A7 in which most of free O2 ions were replaced by electrons, i.e. an electron serves as an anion in this crystal. The electrical conductivities at 300 K of the FZ and CZ single-crystal electrides were ~810 S cm 1 and ~1500 S cm 1, respectively. Figure 19 shows the temperature (T) dependence of the electrical resistivity (r) of two samples. A sharp drop in resistivity is observed in the FZ single-crystal electride with an onset temperature of ~0.2 K. However, electrical resistivity of the metallic C12A7 electride prepared from the CZ single crystal never becomes zero down to 90 mK, as shown in the inset of the figure. Figure 19 (inset) shows that the onset temperature in the FZ sample decreases with the application of magnetic field, which strongly suggests that the FZ single-crystal C12A7 electride is a superconductor below a transition temperature of ~0.2 K. The superconductivity disappears by the magnetic field above a magnetic field of 30 mT. Bulk superconductivity was confirmed by the magnetic measurement, as shown in the right panel of Figure 19. It is confirmed that the FZ single-crystal electride with a conductivity of ~770 S cm 1 (sample B) also becomes a superconductor with the onset temperature of ~0.19 K. The superconducting transition was further observed with the onset temperature of ~0.16 K and ~0.4 K in epitaxial thin film electrides with Ne of 1.6  1021 cm 3 and 2.0  1021 cm 3, respectively. The variation in Tc in both single-crystal and thin film electrides suggests that the increase in Ne leads to an increase in Tc in the C12A7 electride. The critical magnetic field Hc(0) is estimated to be ~19 mT for the bulk and ~33 mT for the thin film, using

468

Handbook of Advanced Ceramics

FIGURE 19 Conductivity of heavy electron-doped C12A7 bulk single crystals. Zero resistivity at 0.2 K and its shift to lower temperature by applying magnetic field are clearly observed. Right panel shows complete diamagnetism (Meissner effect). For color version of this figure, the reader is referred to the online version of this book.

a relation of Hc(T) ¼ Hc(0)[1 (T/Tc)2]. The low transition temperature and the small critical magnetic field imply that C12A7:e is a type-I superconductor. Perfect diamagnetism (c ~ 1) below ~0.20 K in the c T curve was also observed, confirming the bulk superconductivity in the C12A7 electride and 100% volume fraction of the superconducting phase. Here, the mechanism [47] for superconducting transition in the C12A7 electride is described. In the BardeeneCoopereSchrieffer theory, Tc is expressed by McMillan’s formula Tc ¼ [QD/1.45]$exp[ 1.04$(1 þ l)/ (l m*$(1 þ 0.62$l))], where QD is the Debye temperature, l is the phononeelectron coupling constant, and m* is the Coulomb pseudopotential. l is provided by l ¼ N(0)$V, where N(0) is the electron concentration at the Fermi energy, and V is the phonon-mediated coupling constant for a single electron. Using the experimentally obtained value of QD ¼ 604 K [48] from heat capacity and the assumption of m* as 0.2 in the low electron density system, we obtain l ¼ ~0.45, which is much larger than those of alkali metals, as mentioned below. The observed increase in Tc with Ne is consistent with McMillan’s formula, provided that Ne is correlated with N(0). Although the feature that the conduction electron in the electride is that the s-electron is similar to that of the alkali metal, superconducting transition is observed in C12A7 electride but not in alkali metals at ambient pressure except Tc ¼ 0.4 mK of Li [49]. The difference is attributable to the magnitude of the electronephonon interaction. The anionic electron may interact strongly with the rigid lattice framework that is formed due to the strong covalent and ionic bonds among Ca2þ, Al3þ, and oxygen ions. These facts possibly lead to the high values of l (~0.45) and QD (604 K), which are much higher than those of alkali metals such as Li (l ¼ 0.38, QD ¼ 420 K), Na (l ¼ 0.20, QD ¼ 150 K), and K (l ¼ 0.14, QD ¼ 100

K). That is, the strong electronephonon interactions likely resulted from the covalent and ionic bond nature of the oxide framework is responsible for the emergence of superconductivity in the C12A7 electride. Stable electrides with a rigid framework structure and with electrons populated in free space is a new category of superconductors.

4. TRANSPARENT AMORPHOUS OXIDE SEMICONDUCTORS 4.1. History of Oxide Semiconductors and Oxide TFTs Figure 20 shows the history of oxide semiconductors and TFTs based on them. The research on oxide TFTs started in the 1960s but almost disappeared in the open-accessible literature after that. The research reappeared in the 1990s and became rather active in the 2000s. The primary engine for accelerating the research on the oxide semiconductors and oxide TFTs is strong need in rapidly expanding flat panel displays. High performance and easily processable TFTs are needed to meet strong demand for backplane to drive organic LED panels as well as next-generation LCD panels workable at higher frame rates. An oxide semiconductor is N-type except for several materials. This is a natural consequence of chemical bonding nature in oxides. Figure 21 shows the schematic band structure of ionic oxides and silicon. The valence band maximum (VBM) in oxides is primarily composed of 2porbitals of oxygen ions with strong electro-negative nature. So, the location of VBM is so deep that positive holes are difficult to be doped. On the other hand, the situation is reverse for the conduction band minimum (CBM), which is composed of metal cation’s orbitals and is located at

Chapter | 6.4

New Frontiers Opened Up Through Function Cultivation in Transparent Oxides

469

FIGURE 20 History of oxide TFTs and flat panel displays using a TAOS-TFT backplane. For color version of this figure, the reader is referred to the online version of this book.

a rather lower level. This is the reason why P-type doping is extremely difficult but not for N-type doping in oxides [50]. Research on oxide TFTs was restricted on polycrystalline samples until late 2004. Polycrystalline N-type oxide semiconductors represented by ZnO and SnO2 are sensitive to the environmental atmosphere because adsorption of oxygen-related species on the surface/grain boundaries traps a carrier electron and desorption induces the reverse process. The instability of polycrystalline oxide TFTs, which is the major obstacle, comes from the intrinsic nature of surfaces and grain boundaries in oxide semiconductors. This obstacle was practically resolved by using amorphous oxide semiconductors (AOSs) in place of polycrystalline forms in 2004 [51]. Amorphous semiconductors have outstanding advantages in processing over crystalline semiconductors. They have no grain boundaries

and their thin films can be deposited at low temperatures. These features make them very favorable for large-sized thin film application such as TFTs. It was widely believed that the electron transport properties of crystalline semiconductors are largely degraded when they become amorphous on the basis of several examples such as silicon, chalcogenides, and transition metal oxides. However, this was not true for ionic oxide semiconductors, as will be described. Ionic amorphous oxide semiconductors, which are called ‘transparent amorphous oxide semiconductors (TAOS)’ named after their band gaps, have large electron mobilities comparable to those in the corresponding crystalline analogs. TAOSTFTs have mobilities of 10e50 cm(Vs) 1, which are larger by 1e2 orders of magnitude than that in a-Si:H TFTs and can be fabricated by the conventional sputtering process.

FIGURE 21 Schematic energy bands (left) Ionic nontransition metal oxides. (right) Covalent compounds. ex Si. For color version of this figure, the reader is referred to the online version of this book.

470

Handbook of Advanced Ceramics

4.2. Material Design Concept of TAOS with High Mobility In ionic oxides, the nature of CBM which works as an electron pathway totally differs from that of VBM which act as a hole pathway. CBM in ionic oxides is primarily composed of vacant s-orbitals of a cation, and the contribution of oxygen 2p-orbitals, which are dominant at VBM, is rather small. The spatial spread of this vacant s-orbital is so large that direct overlap between the s-orbitals of the neighboring cations is possible in post-transition metal oxides, and therefore an effective mass of electron is small in these oxides. In most amorphous materials, structural randomness appears prominently as the bond angle distribution. When the bond angle has a large distribution, how is the effective mass (in other words, the transfer rate between neighboring cation s-orbitals) modified for carrier electrons? We considered two cases for this question: covalent semiconductors and ionic semiconductors. In the former case, the magnitude of the overlap between the vacant orbitals of the neighboring atoms is very sensitive to the variation in the bond angle. As a consequence, rather deep localized states would be created at somewhat high concentrations and thereby the drift mobility would be largely degraded due to carrier scattering with these defects. On the other hand, the magnitude of the overlap in the latter case is critically varied by the choice of metal cations: when the spatial spread of the s-orbital is larger than the inter-cation distance, the magnitude should be insensitive to the bond angle distribution because the s-orbitals are isotropic in shape. As a consequence, we may anticipate that these ionic amorphous materials have large mobility comparable to that in the corresponding crystalline form. In the case that the spatial spread of the metal s-orbital is small, such a favorable situation cannot be expected. The spatial spread of the s-orbital of a metal cation is primarily determined by the principal quantum number (n) and is modified by the charge state of the cation, as shown in Figure 22 [2,52]. Thus candidates for a transparent amorphous semiconductor having large electron mobilities comparable to those of the corresponding crystals are transparent oxides constituted by post-transition metal cations with an electronic configuration (n 1)d10ns0, where n  5. Figure 23 illustrates the comparison in orbital drawings of Si and a post-transition metal (PTM) oxide between crystalline and amorphous states. The drastic reduction of the electron mobility in the amorphous state from c-Si may be understood intuitively from the figure, whereas medium mobility in c-PTM oxides is reserved even in the amorphous state. In a sense, the situation of CBM in PTM oxides is similar to that in amorphous metal alloys in the aspect that metal orbitals dominantly constitute the electron

FIGURE 22 Calculated values of the overlap integral between the ns orbital functions of metals with (n 1) d10ns0 electronic configuration versus the interatomic separations in the crystals of simple metal oxides. Slater-type orbital functions were used for the calculation. The crystals exhibiting electronic conductivity have an overlap integral larger than the threshold of ~0.4, which is indicated by a dotted line. A dotted circle encloses s-orbitals of n ¼ 5.

pathways. The conductivity of amorphous/liquid metal alloys remains at a slightly lower level compared with the corresponding crystalline phases. The structure of amorphous metal is modeled by dense random packing of metal spheres and occupation of a metalloid at the interstitial position. We may consider that in a-PTM oxides the vacant ns-orbitals of the metal cations work as metal atoms in a metal.

4.3. Electron Transport in Amorphous Oxide Semiconductors Ionic amorphous oxide semiconductors have several common properties that are not seen in conventional amorphous semiconductors. First is their large electron mobility such as 10e40 cm2(Vs) 1, which is higher by 1e2 orders of magnitude than that in a-Si:H or by several orders of magnitude than that in glassy semiconductors or amorphous chalcogenides. Second is that a degenerate state can be realized by doping. This is very different from the other amorphous semiconductors. For instance, c-Si is easily changed to the degenerate state by carrier doping (~several 100 ppm), but no such a state has been attained in a-Si:H [9] to date. That is, conduction takes place by hopping or percolation between tail states in conventional amorphous semiconductors. This is the reason why drift mobility in the amorphous state is so small compared with that in the crystalline state. On the other hand, in a post transition oxides Fermi-level can exceed the mobility gap easily by carrier doping, leading to band conduction. It is considered that this striking difference originates from that in chemical bonding nature between the materials, i.e., strong ionic bonding with spherical potential is much favorable to form a shallow tail state having small density of state.

Chapter | 6.4

New Frontiers Opened Up Through Function Cultivation in Transparent Oxides

471

FIGURE 23 Schematic orbital drawing of electron pathway (conduction band bottom) in covalent compound semiconductors and ionic oxide semiconductors. For color version of this figure, the reader is referred to the online version of this book.

The last unique feature is that electron carriers are primarily injected by chemical doping. Substitutional doping, which is effective in crystalline semiconductors and a-Si:H, is very inefficient. Each ion is primarily connected by ionic bonding and has a large flexibility in a coordination structure compared with the tetrahedral unit in a-Si:H. Thus, doping (or substitution) of aliovalent ions does not yield charge carriers if neutrality of the total formal charge of the constituent ions is maintained. Therefore, an effective doping way is to alter the

stoichiometry of oxygen ion by controlling the oxygen gas pressure during deposition processes or ion implantation of cations with a low electron affinity. Among them amorphous In2O3eGa2O3eZnO (IGZO) has been extensively studied as the semi-conducting channel layer of transparent TFTs since the first report [51] in late 2004. Figure 24 summarizes electrical properties (Hall mobility mHall and carrier concentration Ne) for the films in the In2O3eGa2O3eZnO system [53]. All the films were deposited at the same condition: i.e., on SiO2 glass

FIGURE 24 The amorphous formation region (right) and Hall mobility and carrier electron concentrations (left) in the In2O3eGa2O3eZnO system. The thin film was deposited on a glass substrate by a pulsed laser deposition (PO2 ¼ 1 Pa). Numbers in the parenthesis denote carrier electron concentration (1018 cm 3). For color version of this figure, the reader is referred to the online version of this book.

472

FIGURE 25 Schematic illustration of near the conduction band bottom. Arrows denote conduction paths of electron for the cases of EF > Eth or EF < Eth. For color version of this figure, the reader is referred to the online version of this book.

substrate at RT and oxygen partial pressure of 1.0 Pa. Although pure In2O3 and ZnO films exhibited large Hall mobilities of ~34 and ~19 cm2 V 1 s 1, respectively, they were crystalline even if the films were deposited at RT. Moreover, it is not easy to control carrier concentration down to 1020 cm 3) became very difficult in the larger Ga content films. This result indicates that large mobility is not easily obtained in the a-IGZO films with large Ga contents if one tries to dope carriers by impurity doping or introducing oxygen vacancies. This feature is unfavorable for TCO application, but it would not be disadvantage for semiconductor device applications because the difficulty in carrier doping by oxygen vacancy suggests better controllability and stability of carrier concentration, especially at

FIGURE 26 Tail state density in a-IGZO estimated from TFT characteristics. Estimation is valid energy range below ~1 eV from the CBM. For comparison, data on a-Si:H are shown. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 6.4

New Frontiers Opened Up Through Function Cultivation in Transparent Oxides

low concentrations. Even if high-density doping is difficult by choosing deposition condition, it is still possible to induce high-density carriers by an external electric field if TFT structures are employed, which may make it possible to utilize the potential large mobilities that may be available at large carrier concentrations. Figure 26 compares tail state density in a-IGZO with a-Si:H [54]. The tail state density around the CBM controlling N-channel mobility in a-IGZO is lower by 3-orders of magnitude than that of aSi:H. Such a low tail state density makes it possible to up shift the Fermi-level to the CBM, which induces band conduction giving a high mobility comparable to that in the crystal. So far, no amorphous semiconductor in which band conduction occurs has been found except for TAOS. Hall mobilities larger than 30 cm2 V 1 s 1 are obtained in the In2O3eZnO (a-IZO) systems [55], which are used as amorphous TCOs. However, controllability of carrier concentration and stability of low carrier state are not satisfactory in a-IZO films.

473

4.4. Unique Features of TAOS-TFTs TAOS-TFTs have three unique characteristics compared with other TFTs. First is high field mobility >10 cm2(Vs) 1. Second is easy fabrication at low temperature using conventional DC sputtering. The last is a large process allowance. The TFTs fabricated at unoptimized conditions exhibits poor performance, but the TFT performance can be much improved to that prepared under the optimized condition just by annealing at an appropriate temperature far below the crystallization temperature of TAOS. Figure 27 is an example of a-IZGO -TFTs showing effectiveness of post-annealing to improve TFT performance. The annealing temperature is 250e300  C which is much lower than the crystallization temperature (>500  C). No distinct structural change around each metal cation was noted before and after annealing. Pronounced annealing effects are observed commonly for TAOS-TFTs [56,57]. It is worth noting that TFT performance is practically

FIGURE 27 Effects of post-annealing on the performance of a-IGZO-TFTs. Top: transfer curve of TFTs fabricated under optimized and unoptimized conditions. Annealed in ambient atmosphere. Bottom: effect of various atmosphere during annealing. For color version of this figure, the reader is referred to the online version of this book.

474

Handbook of Advanced Ceramics

FIGURE 28 Histogram of device performance in 97 a-IGZO-TFTs fabricated on a 1 cm2 area. For color version of this figure, the reader is referred to the online version of this book.

determined by this post-annealing treatment. This is similar to the case of amorphous metal and provides a large process window for deposition. Humidity in annealing atmosphere enhances the improvement of the TFT devices [58]. Figure 28 shows the performance histograms of aIGZO-TFTs that were fabricated on a glass substrate by conventional sputtering and subsequently annealed [59]. About 100 TFTs were fabricated from a 1 cm  1 cm area of a-IGZO thin film. The TFT exhibits excellent uniformity and high average performance. The value of msat resides within the range of 0.5 cm2(Vs) 1 and m is 0.11 cm2(Vs) 1 (0.76% of the average value), demonstrating the excellent uniformity of the a-IGZO-TFTs. It strongly suggests that the a-IGZO-TFTs essentially have a good short-range uniformity and are advantageous in integrated circuits and large area. Short channel TFTs are needed for integrated circuits. It was reported by the Samsung group in 2008 that a-IGZOTFTs keep the almost same mobilities, ON/OFF ratios, threshold voltages, and subthreshold slopes when the channel length is reduced down to 50 nm [60]. These results indicate that a-IGZO-TFT is a candidate for selection transistor in 3D-cross-point stacking memory. Table 1 shows comparison of TAOS-TFTs with a-Si:H and poly-Si-TFTs summarized by Jeong et al. in 2008 [61]. It is obvious that a-IGZO-TFTs meet the high mobility, homogeneity, and low cost requirements. The major technical issue was instability under biasing. TAOS-TFTs are sensitive to the light corresponding to the sub-band gap and two performance changes are induced, positive shift of the

threshold voltage (DVth) and increase in off-current (mobility and subthreshold slope remain unchanged). For example, such a change is induced by illumination of light shorter than 460 nm [62]. The magnitude of DVth depends on intensity and wavelength of light. The extent is rather smaller than that of a-Si:H but is larger than that in poly-SiTFTs.

4.5. Progress in TAOS-TFT as the FPD Backplane TAOS-TFTs are several unique features as follows: (1) high electron mobility ranging 10e30 cm2(Vs) 1 depending on the material systems, (2) being capable by conventional sputtering at low temperatures, (3) controllable performance by post-annealing, (4) optically transparent, (5) only N-channel operation, no inversion occurs due to high tail and defect state density above the VBM, (6) a large material selection rage for gate insulator due to (4), and (7) no short channel effect down to 50 nm. Utilizing these advantages, extensive studies are being continued on the application of TAOS-TFTs s to the backplane for FPDs as well as novel display structures.

4.5.1. Novel Display Structure An innovative e-paper display structure called the “front drive” type was recently proposed by Ito et al. of Toppan Printing [63]. Alignment of the TFT array to a color filter array is a troublesome process in the display assembling

Chapter | 6.4

New Frontiers Opened Up Through Function Cultivation in Transparent Oxides

475

TABLE 1 Comparison of TFT Technology a-Si:HTFT

Poly-Si TFT

Amorphous oxide TFT

Generation

8e10G

4G (LTPS)/8G(HTPS)?

IG/8G(?)

Channel

a-Si:H

ELA LTPS/SPC HTPS

a-InGaZnO4

4e5 masks

5e9 (11) masks

4e5 masks

Mobility (cm /Vs)

30 V

0.1 cm2(Vs) 1 has been realized to

477

2007. It is considered that instability and/or high gap state density is the primary reason. For example, Cu2O is a wellknown p-type semiconductor and has attracted as the active layer since the first TFT parent proposed by Heil in 1935. Matsuzaki et al. fabricated epitaxial thin films and obtained a Hall mobility of ~100 cm2(Vs) 1 at the hole concentration of ~1013 cm 3, which is comparable to that in single crystalline Cu2O. However, Cu2O-based TFT did not operate enough and the estimated field-effect mobility remained ~0.1 [64]. This striking difference comes from large tail state densities. Such a situation appears to be similar to other p-type oxide semiconductors. In 2008, Ogo et al. [65] reported a p-channel TFT with the mobility of 1.4 cm2(Vs) 1 employing SnO (not SnO2) as the active layer. This is a first demonstration of a p-channel oxide TFT with a mobility >1 cm2(Vs) 1, which was a long-standing target in this area. The next goal is fabrication of complementary (C) MOS by combining p-channel and n-channel oxide TFTs. Although a monopolar channel is enough for TFTs for the backplane of displays, C-MOS is applicable for logic circuits. Exploiting bipolar semiconductive oxides with low tail state densities which can be fabricated at low temperatures is leading as the most essential work for this goal. Oxide semiconductors are easy to fabricate by conventional sputtering and are robust to oxygen and radiation, in general. If oxide-based C-MOS structure can be fabricated on various types of substrates including plastics, flexible electronic circuits would be promising. Of course, the formation of a hetero-junction between a TOS and an organic semiconductor is a practical and promising way to active device application [66] such as a photosensor, a C-MOS, and a solar cell.

5. IRON-PNICTIDE SUPERCONDUCTORS 5.1. Discovery of Iron-Based Superconductors 5.1.1. From p-Type Transparent Semiconductor to Magnetic Semiconductors A layered compound composed of alternatively stacked active layers is an attractive platform for exploration of wide gap semiconductors and/or functions due to the following reasons: (1) carrier doping layer can be separated spatially from the carrier transport layer, which may enhance carrier mobility through suppression of impurity scattering; (2) electron and/or holes are confined in the active layers which serve as wells, stabilizing exciton at room temperature; (3) magnetic interaction tends to be weakened because of intrinsic magnetic instability in 2 dimensions; and (4) the band gap is increased due to suppression of 3-dimensional band dispersion. We chose

478

FIGURE 32 From p-type transparent semiconductor LaCuOCh (Ch ¼ chalcogenide) to LaTMPnO (TM ¼ þ2 charged state 3d transition metal cation, Pn ¼ pnictgen anion) as a candidate for magnetic semiconductors. Both materials belong to the ZrCuSiAs-type crystal structure composed of an insulating LaO layer with a positive charge and CuCh or TMPn with a negative charge. For color version of this figure, the reader is referred to the online version of this book.

a Cuþ-based mixed oxyanion layered material as the candidate for p-type transparent oxide semiconductors (pTOS). In general, a hole at the valence band top tends to localize at oxygen 2p-orbitals primarily constituting the VBM, due to its strong ionic nature. For p-type TCOs, how to delocalize the positive hole at the VBM of oxides works as a strategy. We focused a Cuþ with [Ar](3d10) configuration which has 3d electron levels comparable to those of O 2p, leading to the discovery of p-TOS, CuAlO2 with a layered structure (delafossite). Further improved performance was obtained in a mixed oxychalcogenide LaCuChO (Ch:chalcogenide) [67]. We found the unique optoelectronic properties for epitaxial thin films of LaCuChO, i.e., large hole mobility, RT-stable exciton [68], a large 3rd order optical nonlinearity [69], 2D-electronic structure [70], and blue light-emitting diode utilized PN junction [71]and efficient excitonic emission. Intriguing properties of transition metals primarily come from the freedom combined with spin and charge of 3d-electrons. Although Cuþ with a closed 3d-shell structure is appropriate for keeping optical transparency (a large gap), no such properties can be expected for its compounds. We thus determined to start the exploration of new magnetic semiconductors in the LaTMPnO system (where TM is a 3d transition cation with þ2 charge state) with the same crystal structure as LaCuChO as shown in Figure 32.

5.1.2. Electromagnetic Properties of LaTMPnO The binary compounds of transition metal pnictides exhibit a diversity of electromagnetic properties by changing the combination of TM and Pn elements. However, few studies

Handbook of Advanced Ceramics

on electromagnetic properties of layered compounds LaTMPnO had been performed notwithstanding that the existence of many compounds were already reported. We tried to synthesize the bulk compounds LaTMPnO (TM:3d transition metal, Pn:P, As, Sb) by solid-state reactions in evacuated silica glass tube and measured electromagnetic properties on the resulting materials. For device application, thin film fabrication by pulsed laser deposition method was examined. Figure 33 summarizes the electromagnetic properties elucidated for LaFeAsPnO. The properties drastically vary with the 3d electron number in the TM: an anti-ferromagnetic semiconductor with Neel temperature (TN) >400 K for the Mn system, a Pauli para metal with a superconducting critical temperature (Tc) at ~4 K for Fe, a ferromagnetic metal with a Curie temperature of ~60 K for Co, a Pauli para metal with Tc ¼ 3 K for Ni, and a diamagnetic semiconductor for Zn. It is of interest to note that metalesuperconducting transition is observed for T2þ M with even number of 3d-electrons, i.e., Fe2þ with 3d6 and Ni2þ with 3d8. The optical properties of LaTMPnO rather differ from those of binary TMPn, but the magnetic properties are similar to each other: anti-ferromagnetic (AFM) for Mn and FeAs, ferromagnetic (FM) for Co, and Pauli para for Fe and Ni. Noteworthy is the striking difference between FeP and FeAs. FeP is a Pauli para metal, while FeAs is an AF metal.

5.1.3. LaFePnO: Striking Difference Between LaFePO and LaFeAsO Figure 34 shows reT curves between LaFePO and LaFeAsO1 xFx. The resistivity of the former simply decreases with temperature and suddenly drops to zero around 4 K [72], while the latter has higher resistivity by an order of magnitude than the former but does not change monotonically with temperature, showing a distinct break around 150 K and r increases with T. When electrons are doped via partial replacement of the oxygen sites by fluorine, this break disappears and instead the bulk superconductivity of Tc(onset) of 32 K emerges [73]. This break in the reT curve is not seen for LaFePnO and LaNiPnO with Tc ¼ 3K [74]. After we submitted a communication reporting LaFeAsO1 xFx with Tc(mid-point) ¼ 26 K, we concentrated to examine the superconductivity in the compounds with multi Fe(Ni)Pn layers following a drastic jump of Tc in cuprates by the discovery of YBCO with bilayers of CuO2. The first material we chose was BaNi2P2 [75]with 122 structures. We found a Tc but it remained ~4 K. The next compound we chose was BaFe2As2 and tried to dope electrons by replacing oxygen sites by fluorine but the result was totally unsuccessful (it took almost 2 years to succeed in synthesizing superconducting Sr122 compounds by electron doping via

Chapter | 6.4

New Frontiers Opened Up Through Function Cultivation in Transparent Oxides

479

FIGURE 33 Electrical and magnetic properties of LaTMPnO. For color version of this figure, the reader is referred to the online version of this book.

aliovalent substitution of alkaline earth ion with a help of high pressure [76]). At this time, a paper reporting Ba1 xKxFe2As2 with a maximum Tc ¼ 38 K by the Jahrendt group of Munich [77] was posted on the cond-mat preprint-server. It is evident that BaFe2As2, the parent compound of Tc ¼ 38 K, exhibits a break at ~140 K like LaFeAsO but BaNi2P2 has no such anomaly as LaFePO and LaNiAsO (Tc ¼ 3 K). So, it is natural to postulate that

the presence of a break in the reT curve is requisite for the parent material for high Tc; but the invalidity of this idea will be shown later. What happens around this break temperature? First, we suspected crystallographic transition and measured powder XRD data by using the SR-ring as a function of temperature. As a consequence, a phase transition from tetragonal to orthorhombic (not monoclinic from the extinction rule)

FIGURE 34 Temperature dependence of resistivity in LaFePO and LaFeAsO1-xFx. Right figure shows electron doping to FeAs-layers by replacement of oxygen with fluorine. For color version of this figure, the reader is referred to the online version of this book.

480

Handbook of Advanced Ceramics

FIGURE 35 Phase diagrams for 1111 and 122 Iron-pnictide superconductors. Ts: crystallographic transition temperature, TN: magnetic transition (Neel) temperature. QCP: quantum critical point (phase transition at 0 K). For color version of this figure, the reader is referred to the online version of this book.

symmetry was clearly observed at ~155K [78]. Measurements of heat capacity and the internal field at the Fe-site monitored by Mossbauer spectroscopy revealed that crystallographic transition and anti-magnetic ordering occur at ~150 K and ~140 K, respectively. The most decisive information on the magnetic structure was elucidated by neutron diffraction. The spin configuration and the magnitude of local moment at a Fe ion at the AFM state in the parent compound were determined by a group of Oak Ridge National Lab (ORNL) [79]. Figure 35 shows electronic phase diagrams for the 1111-type and 122-type iron-based compounds. Superconductivity emerges by introducing carriers to the parent compounds which are anti-ferromagnetic metals. The major differences between these 2 systems are the presence of overlapped region between the AF phase and the superconducting phase and convergence of structural transition and magnetic transition temperature for the 122 system.

5.2. Advances in Materials [80e82] When we discovered the high Tc in LaFeAsO1 xFx in October, 2007, the target was switched from LaFeAsO1 xFx and found a raise of Tc to 43 K under a pressure of ~4 GPa. This temperature exceeded the Tc of MgB2 and is next to the high Tc cuprates. We submitted this paper on February 26, 2008 and started to synthesize LuFeAsO1-xFx at the same time. It was evident that high pressure effectively works to raise the Tc in this system. We thought if the La ion with the largest radius among rare earth ions could be replaced by nonmagnetic rare earth ions with the smallest radius, the highest Tc would have been realized. This trial was very unsuccessful as well. The compound LuFeAsO itself was not obtained (even to date). During struggle with the synthesis of LuFeAsO, higher Tcs were reported on a condmat-server by several Chinese groups at USTC and IOP in

early March. They obtained the high Tc by replacing the La ion with Ce, Pr, or Sm ion with an open 4f-shell. These reports surprised us because we intentionally avoided the use of magnetic rare earth ion based on the experimental result (presented at the 15th Conference on Intercalation at Korea in May, 2007) that superconductivity disappears when the La ion in LaFePO (Tc ¼ 4 K) was replaced by Ce. That is, we misunderstood without careful checking whether or not the 4f/5d orbitals of rare earth ions participate to the Fermi-level which are primarily composed of Fe 3d orbitals. Figure 36 summarizes the progress in Fe(Ni)-based superconductors viewed from material aspects. The abscissa is the date of receipt in a journal or posted though the preprint-server. The Tc reached a maximum of 56 K in a short period by replacing the La ion in LaFeAsO with an appropriate rare earth ion. Two types of new materials were reported in June: one is Ba1 xKxFe2As2, 122-type, by Johredt’s group and the other FeSe1 x, 11-type, by Wu’s group of Taiwan. Subsequently, Na(Li)FeAs, 111type was reported by Chu’s group of Houston and Jin’s group of IOP, and in March 2009, Shimoyama’s group of Tokyo found Tc in FePn compounds with perovskite-like blocking layers. These 5 types of compounds summarized in Figure 37 have a common structural unit, a square lattice of Fe2þ ions (Figure 38). The Fermi level is primarily composed of Fe 3d orbitals and the contribution of pnicogen/chalcogen orbitals is ~10% at the Fermi level. This situation is rather different from high Tc cuprates in which both orbitals of Cu 3d and O 2p participate to the Fermi level. In June 2008, a unique phenomenon was posted by a group of ORNL. They found that superconductivity is induced by partially replacing the Fe-sites in the 1111 compounds by Co ions. The similar results were observed in the 122 phases. This result strikingly differs from that in

Chapter | 6.4

New Frontiers Opened Up Through Function Cultivation in Transparent Oxides

481

FIGURE 36 Progress in iron (nickel)-based superconductors. The ordinate is the date of the paper accepted by the journal or posted on cond-mat preprint-server. For color version of this figure, the reader is referred to the online version of this book.

FIGURE 37 Crystal structure of 5 representative types of parent materials. For color version of this figure, the reader is referred to the online version of this book.

482

Handbook of Advanced Ceramics

FIGURE 38 Common structural unit in 5 types of parent materials and contribution of Fe 3d and As 4p bands (calculated) to the Fermi-level of LaFeAsO. The observed density of state measured by photoemission was shown. For color version of this figure, the reader is referred to the online version of this book.

high Tc cuprates in which partial substitution of Cu sites by other ions diminishes Tc. The group led by Clearfield at Iowa found that Tc in Fe-substituted 122 compounds with a 3d/4d transition metal could be scaled by the excess number of electrons per Fe ion.

5.3. Current Status 5.3.1. What is the Primary Factor Controlling Tc? It was implied from striking difference in Tc between LaFePO and LaFeAsO systems or BaFe2As2 and BaNi2As2 systems that materials with higher TN exhibit high Tc. However, this idea is discounted by a plot in Figure 39 showing the anti-correlation between Tc and TN

for each material system. If this anti-correlation is valid, highest Tc should be derived from the parent material with higher TN, but the fact is reverse. Thus, this implication is invalid. Lee et al. [83] found a good correlation between Tc and the bond angle of Pn(Ch)-Fe-Pn(Ch), i.e., high Tc is observed as this bond angle approaches that (109.28’) of the regular tetrahedron. Kuroki et al. [84] proposed a height (hPn) of pnictgen (chalcogen) from the iron plane as a measure of Tc on the basis of calculation on spin-fluctuation-mediated superconductivity using five 3d orbitals of Fe. Since the energy level of a band derived from Fe 3dx2 y2 is most sensitive to hPn, the shape of the Fermi surface is rather changed with hPn as illustrated in Figure 40. The g-pocket at (p,p) that is hidden for low hPn appears as hPn is increased. As a consequence, nesting becomes satisfactory and Tc is increased as hPn. Figure 41 shows a plot of Tc vs. hPn. Although there are some exceptions such as the 11 system, a good correlation is observed as a whole.

5.3.2. Comparison with Cuprates and MgB2 Table 2 summarizes the features of iron pnictide along with cuprates and MgB2 (the representative of BCS superconductors). Three unique properties are evident for iron pnictides: robustness to impurity doping, very high upper critical magnetic fields, and low crystallographic anisotropy in physical performances. These features are favorable for application to superconducting wires.

5.4. Perspective FIGURE 39 Tc vs. magnetic(TN)/structural (Ts)transition temperature. For color version of this figure, the reader is referred to the online version of this book.

Iron, a representative magnetic element, was believed to the last constituent for emergence of superconductivity because long range magnetic ordering competes with the

Chapter | 6.4

New Frontiers Opened Up Through Function Cultivation in Transparent Oxides

483

FIGURE 40 Change in Fe 3d energy level and Fermi surface with the height of pnictgen from the iron plane (hPn). When hPn increased, nesting becomes satisfactory via appearance of a g-pocket. This result well explains the striking difference in Tc between LaFePO and LaFeAsO1 xFx. For color version of this figure, the reader is referred to the online version of this book.

formation of a Cooper pair requisite for superconductivity. However, once LaFeAs(O, F) with Tc ¼ 26 K was discovered, many iron pnictide (chalcogenide) superconducting materials have been found and the maximum Tc reached 56 K, which is next to the high Tc cuprates exceeding MgB2. I think there are two significances to the discovery of iron-based superconductors. First, we realized that a magnetic element is not a hateful enemy but a powerful friend to realize high Tc superconductors. Second, it provides a large opportunity to find new high Tc materials because there exist several hundreds of layered compounds containing square lattice of transition metal cations taking tetrahedral coordination with nonoxide anions. We expect materials with higher Tc and/or novel class of superconductors would be hidden among these. To our interest, the crystal structure of 122 is the same as that of a representative heavy fermion superconductor CeCu2T2(T ¼ Si, Ge). One may expect some clue to bridge these two superconducting systems would be found. What we should not forget is a historical fact that most groundbreaking materials including high Tc superconductors have been discovered by serendipity in the course of a concentrated exploration effort. I am anticipating new material functions would be discovered as a result of concentrated material exploration with the help of

theoretical modeling and advanced characterization. Iron is the most important element that led to a leap of civilization. I hope iron would serve the same role in the history of superconductivity. Strike while the iron is hot. I think this saying is still true for superconductivity research.

FIGURE 41 Correlation between Tc and hPn. For color version of this figure, the reader is referred to the online version of this book.

484

Handbook of Advanced Ceramics

TABLE 2 Comparison Between 3 Representative Superconductors Fe-pnictides

MgB2

Cuprates

Parent Material

(bad) metal (TN ~ 150 K)

metal

Mott Insulator (TN ~ 400 K)

Fermi Level

3d 5-bands

2-bands

3d single band

Max Tc

56 K

40 K

~140 K

Impurity

robust

sensitive

sensitive

Sc gap symmetry

sign inverted s-wave(?)

s-wave

d-wave

Hc (0)

100e200T>

~40T

~100T

Anisotropy

2e4(122)

~3.5

5e7 (YBCO) 50e90 (Bi system)

2

6. FUTURE CHALLENGE: UBIQUITOUS ELEMENT STRATEGY Although more than 100 elements are known to exist to date, the number of elements which are practically available for materials are restricted to 60e70 due to toxicity, abundance, and radioactivity. Our territorial crust is primarily composed of oxides of representative metal, i.e., Si, Al, Ca, Na, Mg, and so on. Iron and titanium are only two transition metals included in the top 10 elements in the Clarke number. Human beings have created cultures by utilizing these abundant elements as represented by the Iron Age which was opened by creation of the plow leading to the development of agriculture. The present information era is supported by silicon (for memory and the central processing unit in the computer) and silicon dioxide (for optical fibers). It is natural to consider that the coming era, which is facing social difficulties such as energy, resource, and environmental issues, should be supported by innovative materials based on abundant elements as well. We call the approach to realize the important material function using abundant elements with high Clarke number, the ubiquitous element strategy. The major fraction of elements with high Clarke number belong to the light representative element group; this strategy may be regarded as innovative materials science based on light representative metal oxides which are widely used as the main ingredient for traditional ceramics such as cement, glass, and porcelain. The function of materials is determined by combination of elements and structures. Since the number of abundant elements is limited to 20e30, the key to realizing the ubiquitous element strategy is how to find or create a structural element in Figure 42. What we can now raise as the structural element are nanostructure, surface and interface, and defects including nonconventional

valence state. Materials scientists have an image for each element based on their accumulated experience and knowledge. How to realize a functionality which has been essential for a specific element using a totally different element by utilizing an innovative structural element and specific internal freedom of element (charge, orbital, spin, etc.) is the most successful example of this strategy. Prerequisite is a deep understanding of role of a key element in the material functionality in question. Novel information which can be probed by advanced characterization and theoretical modeling will be the platform for this understanding. In the last decade science and technology of nano-sized materials have been much advanced at the global scale under the extensive sponsorship of the nation. Ubiquitous element strategy is thus ranked as a real challenge of nanoscience and technology accumulated to date.

FIGURE 42 Schematic relationship between functionality, structure, and element. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 6.4

New Frontiers Opened Up Through Function Cultivation in Transparent Oxides

ACKNOWLEDGMENTS This work was supported by the ERATO-SORST Project (October 1999eMarch 2010) sponsored by the Japan Science and Technology. A part of the research described herein was obtained by the Element Science and Technology Initiative Project sponsored by MEXT, JAPAN, and the FIRST Project by JSPS. The author appreciates the collaborators who participated in this research.

REFERENCES [1] Kawazoe H, Yasukawa M, Hyodo H, Kurita M, Yanagi H, Hosono H. P-type electrical conduction in transparent thin films of CuAlO2. Nature 1997;389:939. [2] Hosono H, Yasukawa M, Kawazoe H. Novel oxide amorphous semiconductors: transparent conducting amorphous oxides. J Non Cryst Sol 1996;203:334. [3] Ohta H, Kawamura K, Orita M, Hirano M, Sarukura N, Hosono H. Current injection emission from a transparent p-n junction composed of p-SrCu2O2/n-ZnO. Appl Phys Lett 2000;77:475. [4] Hosono H, Asada N, Abe Y. Properties and mechanism of photochromism in reduced calcium aluminate glasses. J Appl Phys 1990;67:2840. [5] Hosono H, Mizuguchi M, Kawazoe H, Ogawa T. Effects of fluorine dimer excimer laser irradiation on the optical transmission and defect formation of various types of synthetic SiO2 glasses. Appl Phys Lett 1999;74:2755. [6] Mizuguchi M, Hosono H, Kawazoe H. Time-resolved photoluminescence for diagnosis of ArF excimer laser damage toughness in CaF2 single crystals. J Opt Soc Am B Optical Physics 1999;16:1153. [7] Kajihara K, Skuja L, Hirano M, Hosono H. Diffusion and reaction of hydrogen in F2-laser-irradiated SiO2 glass. Phys Rev Lett 2002;89:135507. [8] Kawamura K, Otsuka T, Hirano M, Kamiya T, Hosono H. Development of latent images due to transient free carrier electrons by femto second laser pulses and its application to grating shape trimming. Appl Phys Lett 2007;90:011107. [9] Hosono H, Kamiya T, Hirano M. Function cultivation of transparent oxides utilizing built-in nanostructure. Bull Chem Soc Jpn 2006;79:1. [10] Hosono H, Hirano M. In: Hosono, Mishima, Takezoe, Mackenzie, editors. Chap.1 in nanomaterials. Elsevier; 2006. [11] Hosono H, Kawazoe H, Matsunami N. Experimental evidence for Frenkel defect formation in amorphous SiO2 by electronic excitation. Phys Rev Lett 1998;80:317. [12] Hayashi K, Matsuishi S, Kamiya T, Hirano M, Hosono H. Lightinduced conversion of refractory oxide into persistent electronic conductor. Nature 2002;419:462. [13] Taylor HFW. Cement Chemistry. 2nd ed. London: Thomas Telford; 1997. [14] Bartl H, Scheller T. Zur Struktur des12CaO$7Al2O3. N Jb Miner Mh 1970;35:547. [15] Lacerda M, Irvine JTS, Glasser FP, West AR. High oxide ion conductivity in Ca12Al14O33. Nature 1988;332:525. [16] Jeevaratnam, Glasser FP, Dent Glasser LS. Anion substitution and structure of 12CaO$7Al2O3. J Am Ceram Soc 1964;47:105.

[17]

485

Watauchi S, Tanaka I, Hayashi K, Hirano M, Hosono H. Crystal growth of Ca12Al14O33 by the floating zone method. J Cryst Growth 2002;237:801. [18] Kurashige K, Toda Y, Matsuishi S, Hayashi K, Hirano M, Hosono H. Czochralski growth of 12CaO$7Al2O3, crystals. Cryst Growth Des 2006;6:1602. [19] Sushiko PV, Shluger AL, Hayashi K, Hirano M, Hosono H. Electron localization and confined electron gas in nanoporous inorganic electrides. Phys Rev Lett 2003;91:126401. [20] Sushiko PV, Shluger AL, Hayashi K, Hirano M, Hosono H. Role of hydrogen atoms in the photo induced formation of stable electron centers in H-doped 12CaO$7Al2O3. Phys Rev B 2006;73:014101. [21] Toda Y, Yanagi H, Ikenaga E, Kim J, Kubota M, Ueda S, et al. Work function of a room-temperature, stable electride [Ca24Al28O64]4þ(ee)4. Adv Mater 2007;19:3564. [22] Sushiko PV, Shluger AL, Hirano M, Hosono H. From insulator to electride: a theoretical model of nanoporous oxide 12CaO$7Al2O3. J Am Chem Soc 2007;129:942. [23] Zunger A. Practical doping principles. Appl Phys Lett 2003;83:57. [24] Hosono H. Transparent oxide conductors. In: Facchetti A, Marks TJ, editors. Transparent Electronics. Springer; 2010. in Chap 2. [25] Matsuishi S, Hayashi K, Hirano M, Hosono H. Hideride ion as photoelectron donor in microporous crystal. J Am Chem Soc 2005;127:12454. [26] Hayashi K, Toda Y, Kamiya T, Hirano M, Yamanaka M, Tanaka I, et al. Electronic insulator-conductor conversion in hydride iondoped 12CaO$7Al2O3 by electron-beam irradiation. Appl Phys Lett 2005;86:022109. [27] Matsuishi S, Toda Y, Miyakawa M, Hayashi K, Kamiya T, Hirano M, et al. High-density electron anions in a nanoporous single crystal: [Ca24Al28O64]4þ(4e ). Science 2003;301:626. [28] Kim SW, Matsuishi S, Nomura T, Kubota Y, Takata M, Hayashi K, et al. Metallic state in a lime-alumina compound with nanoporous structure. Nano Lett 2007;7:1138. [29] Mott NF. Metal-insulator transition. London: Taylor & Francis; 1990. [30] Edwards PP, Sienko MJJ. Phase separation in metal solutions and expanded fluid metals. J Am Chem Soc 1981;103:2967. [31] Nomura T, Hayashi K, Kubota Y, Kamiya T, Hirano M, Takata M, et al. Anion incorporation-induced cage deformation in 12CaO$7Al2O3 crystal. Chem Lett 2007;36:902. [32] Tardio M, Ramirez R, Gonzalez R, Chen Y, Alves E. Electrical conductivity of MgO crystals implanted with lithium ions. Nucl Instr Meth 2001;B191:191. [33] Taylor HFM. Cement chemistry. 2nd ed. London: Thomas Telford Publishing; 1997. [34] Kim SW, Miyakawa M, Hayashi K, Sakai T, Hirano M, Hosono H. Simple and efficient fabrication of room temperature stable electride: melt-solidification and glass ceramics. J Am Chem Soc 2005;127:1370. [35] Kim SW, Toda Y, Hayashi K, Sakai T, Hirano M, Hosono H. Synthesis of a room temperature stable 12CaO$7Al2O3 electride from the melt and its application as an electron field emitter. Chem Mater 2006;18:1938. [36] Miyakawa M, Hayashi K, Hirano M, Toda Y, Kamiya T, Hosono H. Fabrication of highly conductive 12CaO$7Al2O3 thin films

486

[37]

[38]

[39] [40]

[41]

[42]

[43]

[44]

[45]

[46]

[47]

[48]

[49] [50]

[51]

[52]

Handbook of Advanced Ceramics

encaging hydride ions by proton implantation. Adv Mater 2003;15:1100. Miyakawa M, Toda Y, Hyashi K, Kamiya T, Hirano M, Matsunami N, et al. Formation of inorganic electride thin films via site-selective extrusion by energetic inert gas ions. J Appl Phys 2005;97:023510. Miyakawa M, Hirano M, Kamiya T, Hosono H. High electron doping to a wide band gap semiconductor 12CaO$7Al2O3 thin film. Appl Phys Lett 2007;90:182105. Yamaguchi O, Narai A, Shimizu K. New compound in the system SrO-Al2O3. J Am Ceram Soc 1986;69:C36. Miyakawa M, Ueda N, Kamiya T, Hirano M, Hosono H. Novel room temperature stable electride 12SrO$7Al2O3 thin films: fabrication, optical and electron transport properties. J Ceram Soc Jpn 2007;115:567. Toda Y, Matsuishi S, Hayashi K, Ueda K, Kamiya T, Hirano M, et al. Field emission of electron anions clathrated in subnanometer-sized cages of [Ca24Al28O64]4þ(4e ). Adv Mater 2004;16:685. Toda Y, Kim SW, Hayashi K, Hirano M, Kamiya T, Hosono H, et al. Intense thermal field emission from room-temperature stable electride. Appl Phys Lett 2005;87:254104. Kim K, Kikuchi M, Miyawaka M, Yanagi H, Kamiya T, Hirano M, et al. Photoelectron spectroscopic study of C12A7:e and Alq3 interface: the formation of a low electron-injection barrier. J Phys Chem 2007;C111:8403. Watanabe S, Watanabe T, Ito K, Miyakawa N, Ito S, Hosono H, et al. Secondary electron emission and glow discharge property of 12CaO$7Al2O3 electride for fluorescent lamp. Sci Tech Adv Mater 2012;12:034410. Buchammagari M, Toda Y, Hirano M, Hosono H, Takeuchi D, Osakada K. Room temperature-stable electride as a synthetic organic reagent: application to pinacol coupling reaction in aqueous media. OrgLett 2007;9:4287. Miyakawa M, Kim SW, Hirano M, Kohama Y, Kawaji H, Atake T, et al. Superconductivity in an inorganic electride 12CaO$7Al2O3:e . J Am Chem Soc 2007;129:7270. Kohama Y, Kim SW, Tojo T, Kawaji H, Atake T, Matsuishi S, et al. Evidence for Bardeen-Cooper-Schrieffer-type superconducting behavior in the electride (CaO)12(Al2O3)7:e from heat capacity measurements. Phys Rev B 2008;77:092505. Kohama Y, Tojo T, kawaji H, Atake T, Matsuichi S, Hosono H. Disorder of O2 ion and incorporation of O and O2 radicals in nanoporous crystal 12CaO$7Al2O3 studied by low-temperature heat capacity measurements. Chem Phys Lett 2006;421:558. Tuoriniemi J, Juntunen-Nurmilaukas K, Uusvuori J, Pentti E, Salmela A, Sebedash A. Nature 2007;447:187. Hosono H. Recent progress in transparent oxide semiconductors: materials and device application. Thin Solid Films 2007;515:6000. Nomura K, Ohta H, Takagi A, Kamiya T, Hirano M, Hosono H. Room-temperature fabrication of transparent flexible thin film transistors using amorphous oxide semiconductors. Nature (London) 2004;432:488. Hosono H, Kikuchi N, Ueda N, Kawazoe H. Working hypothesis to explore novel wide band gap electricaly conducting amorphous oxides and examples. J Non Cryst Sol 1996;198-200:165.

[53] Takagi A, Nomura K, Ohta H, Yanagi H, Kamiya T, Hirano M, et al. Carrier transport and electronic structure in amorphous oxide semiconductor, a-InGaZnO4. Thin Solid Films 2005;486:38. [54] Hsich HH, et al. Modeling of amorphous InGaZnO4 thin film transistors and their subgap density of states. Appl Phys Lett 2008;92:133512. [55] Nomura K, Takagi A, Kamiya T, Ohta H, Hirano M, Hosono H. Amorphous oxide semiconductors towards high-performance flexible thin-film transistors. Jpn J Appl Phys 2006;45:4303. [56] Nomura K, Kamiya T, Ohta H, Uruga T, Hirano M, Hosono H. Local coordination structure and electronic structure of the large electron mobility amorphous oxide semiconductor In-Ga-Zn-O. Phys. Rev. B 2007:75. 035212. [57] Hosono H, Nomura K, Ogo Y, Uruga T, Kamiya T. Local coordination structure and electronic structure of the large electron mobility amorphous oxide semiconductor IneGaeZneO: experiment and ab initio calculations. J Non Cryst Sol 2008;354:2796. [58] Nomura K, Kamiya T, Hirano M, Hosono H. Origins of threshold voltage shifts in room-temperature deposited and annealed aIneGaeZneO thin-film transistors. Appl Phys Lett 2009;95:013502. [59] Hayashi R, Ofuji M, Kaji M, Takahashi K, Abe K, Yabuta H, et al. Circuits using uniform TFTs based on amorphous IneGaeZneO. JSID 2007;15:915. [60] Song I, Kim S, Yin H, Kim CJ, Park J, Kim S, Choi HS, et al. Short channel characteristics of galliumeindiumezinceoxide thin film transistors for three-dimensional stacking memory. IEEE Elect Dev Lett 2008;29:549. [61] Jeong JK. A New Era of Oxide Thin-Film Transistors for LargeSized AMOLED Displays. Information Displays 2008;24:20. [62] Fung T-C, Chuang C-S, Nomura K, Shieh H-P, Hosono H, Kanicki J. Photofield-effect in amorphous IneGaeZneO (a-IGZO) thin-film transistors. J Infom Display 2009;9:21. [63] Ito M, Kon M, Miyazaki C, Ikeda N, Ishizaki M, Matsubara R, et al. Amorphous oxide TFT and their applications in electrophoretic displays. Phys Stat Sol (a) 2008;205:1885. [64] Matsuzaki K, Nomura N, Yanagi H, Kamiya T, Hirano M, Hosono H. Epitaxial growth of high mobility Cu2O thin films and application to p-channel thin film transistor. Appl Phys Lett 2009;93:202107. [65] Ogo Y, Hiramatsu H, Nomura K, Yanagi H, Kamiya T, Hirano M, Hosono H. p-Channel thin-film transistor using p-type oxide semiconductor, SnO. Appl Phys Lett 2008;93:032113. [66] Nomura K, Kamiya T, Kikuchi Y, Hirano M, Hosono H. Comprehensive studies on the stabilities of a-IneGaeZneO based thin film transistor by constant current stress. Thin Solid Films 2010;518:3012. [67] Hiramatsu H, Ueda K, Ohta H, Hirano M, Kamiya T, Hosono H. Intrinsic excitonic photoluminescence and band-gap engineering of wide-gap p-type oxychalcogenide epitaxial films of LnCuOCh (Ln ¼ La, Pr, and Nd; Ch ¼S or Se) semiconductor alloy. Appl Phys Lett 2003;82:1048. [68] Ueda K, Inoue S, Hosono H, Sarukura N, Hirano M. Roomtemperature excitons in wide-gap layered-oxysulfide semiconductor. Appl Phys Lett 2001;78:2333. [69] Kamioka H, Hiramatsu H, Ohta H, Ueda K, Hirano M, Kamiya T, et al. Third-order optical nonlinearity originating from roomtemperature exciton in layered compounds LaCuOS and LaCuOSe. Appl Phys Lett 2004;84:879.

Chapter | 6.4

New Frontiers Opened Up Through Function Cultivation in Transparent Oxides

[70] Ueda K, Hiramatsu H, Ohta H, Hirano M, Kamiya T, Hosono H. Phys Rev B 2004;69:155305. [71] Hiramatsu H, Ueda K, Ohta H, Kamiya T, Hirano M, Hosono H. Single-atomic-layered quantum wells built in wide-gap semiconductors LnCuOCh (Ln ¼ lanthanide, Ch ¼ chalcogen). Appl Phys Lett 2005;87:211107. [72] Kamihara Y, Hiramatsu H, Hirano M, Kawamura R, Yanagi H, Kamiya T, et al. Iron-based layered superconductor: LaOFeP. J Am Chem Soc 2006;128:10012. [73] Kamihara Y, Watanabe T, Hirano M, Hosono H. Iron-based layered superconductor La[O1 xFx]FeAs (x ¼ 0.05e0.12) with Tc ¼ 26 K. J Am Chem Soc 2008;130:3296. [74] Watanabe T, Yanagi H, Kamiya T, Kamihara Y, Hiramatsu H, Hirano M, et al. Nickel-based oxyphosphide superconductor with a layered crystal structure, LaNiOP. Inorg Chem 2007;46:7719. [75] Mine T, Yanagi H, Kamiya T, Kamihara Y, Hirano M, Hosono H. Nickel-based phosphide superconductor with infinite-layer structure, BaNi2P2. Solid State Commun 2008;147:111. [76] Muraba Y, Matsuishi S, Kim SW, Atou T, Fukunaga O, Hosono H. High-pressure synthesis of the indirectly electron-doped iron pnictide superconductor Sr1 xLaxFe2As2 with maximum Tc ¼ 22 K. Phys Rev B 2010;82:180512R.

[77]

487

Rotter R, Tagel M, Johrendt D. Superconductivity at 38 K in the iron arsenide Ba1 xKxFe2As2. Phys Rev Lett 2008;101:107006. [78] Nomura T, Kim S-W, Kamihara Y, Hirano M, Sushko PV, Kato K, et al. Crystallographic phase transition and high-Tc superconductivity in LaFeAsO: F. Supercond Sci Technol 2008;21:125028. [79] de la Cruz C, Huang Q, Lynn J, Li J, Ratcliff II W, Zarestky JL, et al. Magnetic order close to superconductivity in the iron-based layered LaO1e2xFxFeAs systems. Nature 2008;453:889. [80] Wen HH. Developments and perspectives of iron-based hightemperature superconductors. Adv Mat 2008;20:3764. [81] Ishida K, Nakai Y, Hosono H. To what extent iron-pnictide new superconductors have been clarified: a progress report. J Phys Soc Jpn 2009;78:062001. [82] Johnson DC. The puzzle of high temperature superconductivity in layered iron pnictides and chalcogenides. Adv Phys 2010;59:803. [83] Lee C-H, Iyo A, Eisaki H, Kito H, Teresa H, Diaz F, et al. Effect of structural parameters on superconductivity in fluorine-free LnFeAsO1 y (Ln ¼ La, Nd). J Phys Soc Jpn 2008;77:083704. [84] Kuroki K, Usui H, Onari S, Arita R, Aoki H. Pnictogen height as a possible switch between high-Tc nodeless and low-Tc nodal pairings in the iron-based superconductors. Phys Rev 2009;B79:224511.

Chapter 6.5

Rapid Prototyping of Ceramics Thierry Chartier and Alexander Badev SPCTS (Science des Proce´de´s Ce´ramiques et de Traitements de Surface; Science of Ceramic Processing and Surface Treatments) - UMR 7315 CNRS (Centre National de la Recherche Scientifique; National Center for Scientific Research), European Ceramic Centre, 12, rue Atlantis, 87068 Limoges, France

Chapter Outline 1. Introduction 2. Process Chain in Ceramic Solid Freeform Fabrication 3. 3D Additive Manufacturing Techniques

489 494 498

1. INTRODUCTION Rapid prototyping (RP) is a revolutionary and powerful technology with wide range of applications. The process of prototyping involves quick building up of a prototype or working model for the purpose of testing the various design features, ideas, concepts, functionality, output and performance. The user is able to give immediate feedback of the prototype and its performance. RP is an essential part of the process of system designing and it is quite beneficial as far as reduction of project cost and risk are concerned. Depending on the technologies involved, it is referred as Solid Freeform Fabrication (SFF) or freeform fabrication, digital fabrication, automated freeform fabrication, 3D printing, solid imaging, layer-based manufacturing, laser prototyping and additive manufacturing.

1.1. History in RP Techniques The development of RP involves the development of computer applications in the industry. The increase in the use of computers has engaged the advancement in many computer-related areas including Computer-Aided Design (CAD), Computer-Aided Manufacturing (CAM) and Computer Numerical Control (CNC) machine tools. In particular, the apparition of RP systems is directly related to the CAD technology. However, many other technologies and advancements in other fields such as manufacturing systems and materials have played an important part in the development of RP systems. CAD technology (Computer-Aided Design TechnologydTCAD) is a branch of electronic design automation that mostly models semiconductor fabrication and

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00028-9 Copyright Ó 2013 Elsevier Inc. All rights reserved.

4. Data File Formats 5. Applications of RP Techniques References

516 519 522

semiconductor device operation [1]. The modeling of the fabrication is termed Process TCAD, while the modeling of the device operation is termed Device TCAD. CAM is the use of computer software to control machine tools and related machinery in the manufacturing of work pieces [2]. The CAM technology involves planning, managing, and control manufacturing operations. CAM systems have become possible through the development of computers, electronically operational, accurate, versatile, and productive machine tools, and through the development of sophisticated software technologies. Through the use of CAM systems it is possible to generate numerical control (NC) instructions to control a machine based on geometric information from a CAD database. Computer-aided engineering (CAE) is the broad usage of computer software to aid in engineering tasks [3]. CAE technology is related to the use of computer systems in order to establish CAD geometry, and simulating the behavior of the product. Through a CAE system, the CAD geometry can be redefined and optimized if necessary. CAE tools are available for a wide range of analyses, such as: stress and strain evaluation, heat transfer analyses, magnetic field distribution, fluid dynamics, vibration and kinematics analyses, etc. Finite element analysis tools usually solve each of these types of problems, by transforming a physical file into a simplified model made up of interconnected elements. The historical development [4] of the technologies related to RP is as follows: l

Sixties: The first RP techniques became accessible in the later eighties and they were used for production of prototype and model parts. The history emerged in the

489

490

l

l

l

Handbook of Advanced Ceramics

late sixties, when Professor Herbert Voelcker questioned himself about the possibilities of doing interesting things with the computer controlled and automatic machine tools. These machine tools had just started to appear on the factory floors then. Voelcker was trying to find a way in which the automated machine tools could be programmed by using the output of a design program of a computer. Seventies: Voelcker developed mathematical algorithms that clearly describe the three-dimensional aspects and resulted in the earliest theories of mathematical theories for solid modeling. These theories range form the basis of modern computer programs used to design almost all mechanical objects [5]. They changed the designing methods in the seventies, but, the old methods for designing were still very much in use. The old method revolves on a computer-controlled machine. Eighties: In 1987, Carl Deckard, a researcher form the University of Texas, came up with a good revolutionary idea. He pioneered the layer-based manufacturing, wherein he thought of building up the model layer by layer [6]. He printed 3D models by utilizing laser light for fusing metal powder in solid prototypes, single layer at a time. Deckard developed this idea into a technique called Selective Laser Sintering (SLS). The results of this technique were extremely promising. The history of RP is quite new and recent. As this technology of RP has such wide ranging scope and applications with amazing results, it has been tremendously developing over the years. Present-day RP: Today, the computer engineer has to simply sketch the ideas on the computer screen with the help of a design program that is computer aided. Computer-aided designing allows to make modification as required and it is therefore possible to create a physical prototype that is a precise, as well as proper a 3D object.

Voelcker and Deckard researches and innovations have given extreme development to this significant new industry known as RP or freeform fabrication. It has revolutionized the designing and manufacturing processes. There are many references of people contributing to the development of the RP technology such as Charles Hull’s patent of Apparatus for Production of 3D Objects by Stereolithography (SL) (Table 1).

1.2. Main Notions on Prototype 1.2.1. Definition of a Prototype Prototypes are created in order to realize the conceptualization of a design. It is often used as part of the product design process to allow engineers and designers the ability to explore alternatives, test theories and confirm performance prior to starting production of a new product.

TABLE 1 Historical Development of RP and Related Technologies [7] Year

Event

1770

Mechanization

1946

First Computer

1952

First NC Machine Tool

1960

First commercial Laser

1961

First commercial Robot

1963

First Interactive Graphics System

1988

First commercial RP System

Engineers use their experience to tailor the prototype according to the specific unknowns still present in the intended design. Some prototypes are used to confirm and verify consumer interest in a proposed design; other prototypes will attempt to verify the performance or suitability of a specific design approach [8]. The fabrication of prototypes involves several stepsdmaterial removal, castings, molds, joining with adhesives etc. and with many material typesdaluminum, zinc, urethanes, wood, etc. [9]. Prototyping processes have gone through three phases of development, the last two of which have emerged only in the last 20 years. Like the modeling process in computer graphics, the prototyping of physical models is growing through its third phase. In general, an iterative series of prototypes are designed, constructed and tested as the final design is prepared for production. Multiple iterations of prototypes are used to progressively refine the design. A common strategy is to design, test, evaluate and then modify the design based on analysis of the prototype. In many products it is common to assign the prototype iterations Greek letters. For example, a first iteration prototype is called an “Alpha” prototype. Often this iteration is not expected to perform as intended and some amount of failures or issues are anticipated. Subsequent prototyping iterations (Beta, Gamma, etc.) will be expected to resolve issues and perform closer to the final production intent. In many product development organizations, prototyping specialists are employeddindividuals with specialized skills and training in general fabrication techniques that can help bridge between theoretical designs and the fabrication of prototypes. An example of a resonator prototype destined for high frequency applications is shown on the Fig. 1 below:

1.2.2. Types of Prototypes The implementation of a prototype covers the range of prototyping the complete product (or system) to

Chapter | 6.5

Rapid Prototyping of Ceramics

491

l

FIGURE 1 Ceramic resonator prototype made by stereolithography [10]. For color version of this figure, the reader is referred to the online version of this book.

prototyping part of, or a sub-assembly or a component of the product. The complete prototype, as its name suggests, models most, if not all, the characteristics of the product. It is usually implemented full-scale as well as being fully functional. The main prototype categories are: l

l

l

Proof-of-Principle Prototype (Model) (in electronics sometimes built on a breadboard). A Proof of concept prototype is used to test some aspects of the intended design without attempting to exactly simulate the visual appearance, choice of materials or intended manufacturing process [11]. Such prototypes can be used to “prove” out a potential design approach such as range of motion, mechanics, sensors, architecture, etc. These types of models are often used to identify which design options will not work, or where further development and testing is necessary. Form Study Prototype (Model). This type of prototype will allow designers to explore the basic size and the look of a product without simulating the actual function or exact visual appearance of the product. They can help assess ergonomic factors and provide insight into visual aspects of the product’s final form. Form Study Prototypes are often hand-carved or machined models from easily sculpted, inexpensive materials (e.g. urethane foam), without representing the intended color, finish, or texture. Due to the materials used, these models are intended for internal decision making and are generally not durable enough or suitable for use by representative users or consumers. User Experience Prototype (Model). A User Experience Model invites active human interaction and is primarily used to support user focused research. While intentionally not addressing possible esthetic treatments, this type of model does more accurately represent the overall size, proportions, interfaces, and articulation of a promising concept [12]. This type of model allows early assessment of how a potential user

l

interacts with various elements, motions, and actions of a concept which define the initial use scenario and overall user experience. As these models are fully intended to be used and handled, more robust construction is necessary. Materials typically include plywood, REN shape, RP processes and CNC machined components. Construction of user experience models is typically driven by preliminary CAID/CAD which may be constructed from scratch or with methods such as industrial CT scanning. Visual Prototype (Model) will capture the intended design esthetic and simulate the appearance, color and surface textures of the intended product but will not actually embody the function of the final product [13]. These models will be suitable for use in market research, executive reviews and approval, packaging mock-ups, and photo shoots for sales literature. Functional Prototype (Model) (also called a working prototype) will, to the greatest extent practical, attempt to simulate the final design, esthetics, materials and functionality of the intended design. The functional prototype may be reduced in size (scaled down) in order to reduce costs [14]. The construction of a fully working full-scale prototype and the ultimate test of concept is the engineers’ final check for design flaws and allow last-minute improvements to be made before larger production runs are ordered.

1.3. Classification of RP Techniques RP systems are mainly classified on the nature of the material that the prototype or part is built with. In this manner, all RP systems are categorized into liquid-based, solid-based and powder-based. We will present in detail the most common 3D RP techniques used in the manufacturing of ceramic materials.

1.3.1. Liquid-based In liquid-based RP systems, the initial form of its material is in liquid state [15]. Through a process of photopolymerization or curing, the liquid is converted into the solid state. The following RP systems fall into this category: l l l l l l l

3D Systems’ Stereolithography Apparatus (SLA) Solid Ground Curing (SGC) 3D Systems Microstereolithography Apparatus (mSLA) Fused Deposition Modeling Solid Object Ultraviolet-Laser Printer (SOUP) Soliform System Two Laser Beams

As is illustrated in Fig. 2, two methods are possible under the liquid polymerization processes: the SGC and the microfabrication (SLA/mSLA processes) which are most

492

Handbook of Advanced Ceramics

FIGURE 2 Classification of rapid manufacturing processes.

commonly used due to a better spatial resolution of the processes. SL/mSL processes will be presented in Chapter 3.

l l l

1.3.2. Solid-based

l l

Solid-based RP systems enclose all materials in the solid state and the shapes could be in the form of a wire, a roll, laminates and pellets [17]. The following RP techniques fall into this definition: l l l l l l l

Laminated Object Manufacturing (LOM) Fused Deposition Modeling (FDM) Paper Lamination Technology (PLT) 3D Systems’ MultiJet Modeling System (MJM) ModelMaker and PatternMaster Slicing Solid Manufacturing (SSM) Extrusion Modeling and Multifunctional RPM Systems

1.3.3. Powder-based This category of technologies uses powder in a solid state. The following RP systems fall into this definition: l l

SLS Three-Dimensional Printing (3DP)

l

Laser Engineered Net Shaping (LENS) Direct Shell Production Casting (DSPC) Multiphase Jet Solidification (MJS) Electron Beam Melting (EBM) Direct Metal Deposition 3D Printing Process

All the above RP systems employ the Joining/Binding method. The method of joining/binding differs for the above systems in that some employ a laser while others use a binder/glue to achieve the joining effect [18].

1.4. Advantages and Problems Encountered in RP Processes RP has various advantages relative to regular, traditional prototyping and also to the technologies which have taken its concept further. The revolutionary capability which it introduced is the automatic creation of physical 3D forms from computer images [19]. Moreover, the additive fabrication technique which it uses has the ability to produce almost any geometric shape or feature with absolute fidelity to the original design. With suitable material, shapes can be highly convoluted and even include parts nested within

Chapter | 6.5

Rapid Prototyping of Ceramics

493

parts. The technique has had a significant impact on certain fields of production engineering (Fig. 3). The ability to produce an accurate model of a product or component direct from the design greatly facilitates communication between designer, manufacturer, marketer, purchaser and user, all of whom can visualize the final product and, where relevant, make sure they are entirely satisfied before manufacturing commences. This in turn decreases development time because changes can be made, and errors detected and rectified, before production, thus eliminating or decreasing costly mistakes. Projects are less risky, take less time, and their overall cost is less so output can be higher. Once the product has been manufactured and gone into use, the same time and cost savings apply when it becomes desirable to make adjustments to improve its effectiveness. Thus necessary features can be added and redundant features eliminated comparatively early in the life of the design. Additional advantages of RP include the nonindustry-specific nature of the technique and the facility for creating a large number of variants on a single model. Certain other processes have been developed taking the

concept of RP a step further like rapid tooling and rapid injection molding, both of which have advantages and disadvantages compared with RP itself. RP is faster than the other two procedures mentioned and can produce complex shapes. If only a few parts are needed it is relatively inexpensive (in a production engineering context), but for a quantity of parts there is no saving because each one takes as long and costs as much as the first. The finish on the product tends to be rough, and there are only a few materials which can be used. If strength or heat-resistance need to be tested, for example, this can make the process a poor choice for assessing a part’s function. For rapid tooling there is still a limited choice of materials, and the need to make molds restricts the complexity of the designs which can be produced. Also the time taken is a little longer. However the quality of the parts is significantly superior to that produced by RP. Rapid injection molding also provides good quality components, and additionally, because of the wider variety of materials which can be used in the metal molds, the functionality of the parts can be more readily tested. The time it takes is comparable with

full production

100%

Production design

Tool design

Part drawing

Tool manufacturing

Function testing

Assembly/ test

Special tools documents

Small sized production

full production

RP Model

Production design/CAO

Fabrication Time and costsaving

Assembly Function testing

RP patterns

Documents/ brochure

10 to 50% FIGURE 3 RP versus full production model [7]. For color version of this figure, the reader is referred to the online version of this book.

494

that of rapid tooling, and can be contrasted very favorably with that of traditional injection molding which, while providing the ultimate in quality, takes weeks or months as opposed to days. This technique requires a specialist input at every stage, starting with the engineer who creates the CAD design. The costs are high. Then, the conversion of the CAD design into the STL format, and the creation of the physical prototype itself, which is likely made of resin, is of limited use for anything other than design and fit. The relative economies of these different styles of additive manufacturing are related. In one case the machine may cost more, while each prototype costs less; in another the material used to make the prototype may be more expensive, but the lead time required to finish the article is reduced thus saving worker time. Despite all of the benefits derived from this technology, RP has several disadvantages: it sometimes fails in replication of the real product or system. It could happen that some important developmental steps could be omitted to get a quick and cheap working model. This can be one of the greatest disadvantages of RP. Another disadvantage of RP is one in which many problems are overlooked resulting in endless rectifications and revisions. One more disadvantage of RP is that it may not be suitable for large sized applications [20]. The user may have very high expectations about the prototype’s performance and the designer is unable to deliver these. The system could be left unfinished due to various reasons or the system may be implemented before it is completely ready. The producer may produce an inadequate system that is unable to meet the overall demands of the organization. Too much involvement of the user might hamper the optimization of the program. The producer may be too attached to the program of RP, thus it may lead to legal involvement.

2. PROCESS CHAIN IN CERAMIC SOLID FREEFORM FABRICATION SFF is an important developing technology that enables the fabrication of functional objects with complex properties directly from computer data. The basic operation of any SFF system consists of slicing a 3D computer model into thin cross sections, translating the result into 2D position information, and using this information flow to control the placement of solid material [21]. This process is repeated for each cross section and the object is built up one layer at a time. SFF has historically been associated with manufacturing technologies, and used for the rapid production of visual models, lowrun tooling, and functional objects. The impact of SFF goes far beyond these applications, however; the additive nature of SFF techniques offers great promise for producing objects with unique material combinations and geometries which

Handbook of Advanced Ceramics

could not be obtained by traditional manufacturing methods. Through the years, the use of SFF techniques has increased in the fields of biomedical engineering, electronics, aerospace, architecture, and archeology. In the last decades the development of RP techniques has led to the production of a diversity of mechanical, optical and fluidic components and systems. As almost all of the patterning techniques are appropriate only for selected materials, as a consequence, material available for microsystems technology is inevitably limited. Examples are the LIGA technique (German acronym for X-ray deep lithography, electro deposition, and polymer molding) for the production of parts of polymers and selected metals, silicon technologies, micromachining of metals or patterning of UV-sensitive glasses. For some applications, however, the specific properties of ceramics, such as high hardness, high thermal and chemical resistance or dielectric properties are very important and selective to the used SFF technique. Unfortunately, the required properties of the materials used make the use of the established patterning techniques impossible for most ceramic materials. Therefore the development and application of ceramic microcomponents have been retarded.

2.1. Main Processes Ceramic components are mostly formed in the green state by consolidating a ceramic powder with the help of more or less organic additives into the desired shape. A variety of processes were developed or known techniques adapted to the special requirements of the shaping of ceramic components with patterning details in the sub millimeter range. These techniques differ in design, and costs, but have in common that a patterned mold is needed [22]. As most of these techniques require metallic molds at least in one-step of the replication process due to the applied pressure and/or temperature, these metal molds are mainly produced by more or less expensive and time-consuming techniques like erosion methods, mechanical micromachining, or even the LIGA technique. For faster supply of models and prototypes, a large number of RP methods have been developed in the last 15 years. In 1986 the stereolithography technique of polymers (Hull 1986) has improved the resolution and precision of these manufacturing technologies, as reported by Halloran et al. [23]. However, SFF techniques like MJS, SLS, LOM or fused deposition of ceramics are not well suited for the production of high resolution ceramic components, as they still do not possess sufficient accuracy or do not provide prototypes with the properties of conventionally shaped ceramics. In order to obtain a better resolution of the final product, stereolithography technique is mostly used

Chapter | 6.5

Rapid Prototyping of Ceramics

495

[24,25]. Fabrications of patterned ceramics or ceramic parts make high demands on the precision and resolution of the molding process. As the finishing process of miniaturized ceramic components is nearly impossible, shaping has to be done by a replication step in the unfired state. To avoid high tooling costs in product development, a RPPC has been established that enables rapid manufacturing of ceramic microcomponents from functional models to small lot series within a short time. This process chain combines the fast and inexpensive supply of master models by RP with accurate and flexible ceramic manufacturing by low-pressure injection molding (LPIM). Besides proper feedstock preparation and sufficient small grain size, the quality of the final components is mainly influenced by the quality of the master model. Hence, the RP method must be carefully selected to meet the requirements of the component to be fabricated. The manufacturing of ceramic parts involves the building of 3DeCAD model of the desired part. The data are transferred to the RP equipment, where a model is fabricated out of polymers. The polymer master model is then molded with liquid silicon rubber, which is subsequently used as a tool in the LPIM. The final ceramic part is obtained after dewaxing and sintering.

2.2. Process Chain There are three fundamental fabrication processes: subtractive, additive and formative processes [26]. In the subtractive process the material is removed from the final product until the desired shape is reached. The additive process is used to create a product larger than the starting material. A material is manipulated so that its successive portions combine to form the desired object. The formative process is a building sequence where mechanical forces are applied on a material to form it into the desired shape. The main parts of the process chain are 3D modeling, data conversion and transmission, checking and preparing, building and postprocessing (Fig. 4) [27]. Depending on the quality of the model, the process may be iterated until a satisfactory model or part is achieved. However, like other fabrication processes, process planning is important before the RP commences. In process planning, the steps of the RP process chain are listed. The first step is 3D geometric modeling. In this instance, the requirement would be a workstation and a CAD modeling system. The various factors and parameters which influence the performance of each operation are examined and decided upon. For example, if a SLA is used to build the part, the orientation of the part is an

Slice n°36

CAD File

• •

STL File (triangulation)

Slice n°130

Slicing (160 layers)

Directly from CAD file. No break of the digital chain.

Debinding - Sintering

Cleaning

FIGURE 4 Process chain in ceramic freeform fabrication. For color version of this figure, the reader is referred to the online version of this book.

496

Handbook of Advanced Ceramics

important factor which would, amongst other things, influence the quality of the part and the speed of the process. An operation sheet used in this manner requires proper documentation and guidelines. Good documentation, such as a process logbook, allows future examination and evaluation, and subsequent improvements can be implemented to process planning.

2.3. Master Model Fabrication The advantage of a RPPC consists in the flexibility of the model preparation (Fig. 5). In principle the full-range of RP methods could be used if it leads to accurate models. However, due to high processing time or costs, a careful choice of the RP method is strongly recommended to obtain the desired model properties. For most ceramic components a standard stereolithography, MJM and the RMPD technique (Rapid Micro Product Development) were used. With these three methods all applications could be realized with sufficient accuracy [29]. Each route starts with the generation of a three-dimensional CAD model of the ceramic component to be produced. The 3D model is subjected to triangulation, i.e. it is approximated by a structure consisting of triangles (Fig. 6). By varying the number of these triangles, the amount of data and the resolution of the component are influenced. If the sintering shrinkage of the ceramic is uniform in all dimensions it can be compensated by a simple rescaling of the model size. However even an inhomogeneous shrinkage

Model manufacturing by Rapid Prototyping (STL, MJM, ...)

may be compensated by an anisotropic scaling. For the preparation of models with a lower resolution a commercial stereolithography machine could be used. Due to the large spot size, parts with a relatively large volume can be fabricated within short times. For the production of small-sized holes by stereolithography, problems arise from the epoxides or acrylates, which are used as precursor resins. After removing the part from the resin bath these details cannot be cleaned sufficiently from the adherent resin residues, therefore the parts are replicated with poor accuracy. For the fabrication of such items an extrusion based RP method like FDM [31] or a ballistic method like MJM [32] is used. Although these methods are less suited for microfabrication because of the inherent limited accuracy, it has been demonstrated that they meet the requirements for applications where the high surface roughness can be tolerated. For parts with fine details or high resolution the models could be made of acrylates using microstereolithography technique, which is suited for micro dimensioning, and allows reaching a precision of about 5 mm.

2.4. File Format The information chain used in all additive processes encloses the following steps [33]: l l

Creation of a CAD model of the design Conversion of the CAD model to STL format

Mold in Silicone LP-Injection Molding (Gel-Casting)

Sintering

CAD

Re-Design

FIGURE 5 Process steps in RP process chain [28]. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 6.5

Rapid Prototyping of Ceramics

497

FIGURE 6 Chordal error and definition of a layer through the intersection between the 3D STL Model and a slicing plane [30].

l l l

Slicing the STL file into thin cross-sectional layers Construction of the model by stacked layers Cleaning and finishing the model

CAD Model Creation: First, the object to be built is modeled using a CAD software package. Solid modelers, such as Pro/ENGINEER, tend to represent 3D objects more accurately than wire-frame modelers such as AutoCAD, and will therefore yield better results. The designer can use a preexisting CAD file or may wish to create one expressly for prototyping purposes. This process is identical for all of the RP build techniques. Conversion to STL Format: The various CAD packages use a number of different algorithms to represent solid objects. To establish consistency, the STL (stereolithography, the first RP technique) format has been adopted as the standard of the RP industry. The second step, therefore, is to convert the CAD file into STL format. An STL file is a triangular representation of a 3D surface geometry [34]. The surface is tessellated logically into a set of oriented triangles (facets). Each facet is described by the unit outward normal and three points listed in counterclockwise order representing the vertices of the triangle. While the aspect ratio and orientation of individual facets is governed by the surface curvature, the size of the facets is driven by the tolerance controlling the quality of the surface representation in terms of the distance of the facets from the surface. The choice of the tolerance is strongly dependent on the target application of the produced STL file. In industrial processing, where stereolithography machines perform a computer controlled layer by layer laser curing of a photosensitive resin, the tolerance may be in order of 0.1 mm to make the produced 3D part precise with highly worked out details. The native STL format needs to fulfill the following specifications: the normal and each vertex of every facet are specified by three coordinates each, so there is a total of 12 numbers stored for each facet. Each facet is part of the boundary between the interior and the exterior of

the object. The orientation of the facets is specified redundantly in two ways which must be consistent. First, the direction of the normal is outward. Second, the vertices are listed in counterclockwise order when looking at the object from the outside (right-hand rule). Each triangle must share two vertices with each of its adjacent triangles. This is known as vertex-to-vertex rule. The object represented must be located in the all-positive octant (all vertex coordinates must be positive). However, for nonnative STL applications, the STL format can be generalized. The normal, if not specified (three zeros might be used instead), can be easily computed from the coordinates of the vertices using the right-hand rule. Moreover, the vertices can be located in any octant. And finally, the facet can even be on the interface between two objects (or two parts of the same object). This makes the generalized STL format suitable for modeling of 3D nonmanifold objects. The STL standard includes two data formatsdASCII and binary. While the ASCII form is more descriptive, the binary form is far more common due to the very large resulting size of the CAD data when saved in the ASCII format. The first line in the ASCII format is a description line that must start with the word “solid” in lower case, followed eventually by additional information as the file name, author, date etc. The last line should be the keyword “endsolid”. The lines in between contain descriptions of individual facets as follows: facet normal 0:0 0:0 1:0 outerloop vertex 1:0 1:0 0:0 vertex 1:0 1:0 0:0 vertex 0:0 1:0 0:0 endloop endfacet Slice the STL File: Once the STL files are verified to be error-free, the RP system’s computer analyzes the STL files

498

Handbook of Advanced Ceramics

that define the model to be fabricated and slices the model into cross sections. The cross-sections are systematically recreated through the solidification of liquids or binding of powders, or fusing of solids, to form a 3D model. In a SLA, for example, each output file is sliced into cross sections, between 0.12 mm (minimum) and 0.50 mm (maximum) in thickness. Generally, the model is sliced into the thinnest layer (approx. 0.12 mm) as they have to be very accurate. The supports can be created using coarser settings. An internal cross hatch structure is generated between the inner and the outer surface boundaries of the part. This serves to hold up the walls and entrap liquid that is later solidified with the presence of UV light. Preparing building parameters for positioning and stepwise manufacturing in the light of many available possibilities can be difficult if not accompanied by proper documentation. These possibilities include determination of the geometrical objects, the building orientation, spatial assortments, arrangement with other parts, necessary support structures and slice parameters. They also include the determination of technological parameters such as cure depth, laser power and other physical parameters as in the case of SLA. It means that user-friendly software for ease of use and handling, user support in terms of user manuals, dialog mode and online graphical aids will be very helpful to users of the RP system.

removal of supports. Similarly, for SLS parts, the excess powder has to be removed. Likewise for LOM, pieces of excess wood-like blocks of paper which acted as supports have to be removed [36]. As shown in Table 2, the SLA procedures require the highest number of post-processing tasks. More importantly, for safety reason, specific recommendations for post-processing tasks have to be prepared, especially for cleaning of SLA parts. It was reported that accuracy is related to the posttreatment process. Most defects in SLAbuilt parts occur with the use of inadequate cleaning solvents. Parts are typically cleaned with low viscosity solvent to remove unreacted photosensitive resin. Depending upon the build style and the extent of crosslinking in the resin, the part can be distorted during the cleaning process. This effect was particularly pronounced with the more open build styles and aggressive solvents. With the build styles approaching a solid fill and more solvent-resistant materials, damage with the cleaning solvent can be minimized. With newer cleaning solvents, like TDA (tridecyl acrylate monomer) introduced by 3D Systems, part damage due to the cleaning solvent can be reduced or even eliminated.

3. 3D ADDITIVE MANUFACTURING TECHNIQUES

2.5. Object Setup Using one of several techniques, RP machines build one layer at a time from polymers, paper, or powdered metal. Most machines are fairly autonomous, needing little human intervention. The building time can take from minutes up to several hours depending on the size of the object, the precision and number of layers for 3D processes [35].

2.6. Finishing Processes The cleaning task involves the removal of excess parts which may have remained. Thus, for SLA parts, this refers to excess resin located in entrapped portion, as well as the

Very sophisticated techniques have been developed in the past 10 years in order to create 3D objects. These techniques have been integrated into a new technology known as “Additive manufacturing techniques” that use different materials as listed in the Table 3 below: All these techniques use a common principle, the transformation of a CAD model into a physical one produced layer by layer. They build complex 3D parts in one sequence without human intervention. This chapter will present these techniques in general and most attention will be focused on one of the most sophisticated ones, the stereolithography and mSLA.

TABLE 2 Finishing Processes for Different RP Systems [7] RP Technologies Finishing processes

Selective laser sintering

Stereolithography

Fused deposition modeling

Laminated object manufacturing

Cleaning

Yes

Yes

No

Yes

Posttreatment

No

Yes

No

No

Setting

Yes

Yes

Yes

Yes

Chapter | 6.5

Rapid Prototyping of Ceramics

TABLE 3 3D Additive Manufacturing Techniques 3D prototyping techniques

Base materials

SLS

Thermoplastics, metals, powders

InkJet printing

Almost any alloy, metal

FDM

Thermoplastics, eutectic metals

Stereo/microstereolithography (SLA/mSLA)

Photopolymer

LOM

Paper

3D printing (3DP)

Various materials

3.1. Selective Laser Sintering 3.1.1. History SLS was created in the University of Texas, and is a popular process used in RP and product development. The SLS technology was brought commercialized by DTM Corporation which is now called 3D Systems. It is a layer additive production process that creates three-dimensional objects using a CO2 laser to melt, or sinter, and fuse selective powder molecules based on information supplied by a CAD file (Fig. 7).

499

Since being patented in 1989, the SLS technology has become one of the most used processes for prototyping and product development in all industries. SLS equipment and machinery have also developed and expanded to meet the criteria of users. Currently, there are five different types of machines including high-speed systems. Some of them have round platforms while others have square and rectangular build areas. The build platforms range in size from small (12" in diameter), to medium (12  14” rectangle), all the way up to the largest platform (20  20” square). The different size build areas allow users to choose how parts are built and oriented depending on how large or small the parts are designed. If parts exceed the parameters of the build area it is not difficult to build parts in the SLS process in multiple sections and then bond them in postproduction. The variety of properties that the final materials offer makes it quite easy to bond sectioned parts together with an adhesive that is strong and tough.

3.1.2. Principle The technique, shown in Fig. 8 below, uses a laser beam to selectively fuse powdered materials, such as nylon, elastomer, and metal, into a solid object [37]. Parts are built upon a platform which sits just below the surface in a bin of the heat-fusible powder. A laser traces the pattern of the first layer, sintering it together. The platform is lowered by

FIGURE 7 CAD file of a craniofacial implant (3D Ceram, Limoges, France).

500

Handbook of Advanced Ceramics

FIGURE 8 (a) Laser wise building method. (b) Fabrication sequence in SLS technique. For color version of this figure, the reader is referred to the online version of this book.

the height of the next layer and powder is reapplied. This process continues until the part is complete. Excess powder in each layer helps to support the part during the build. The powder material that is fused during the process is commonly called thermoplastic material or, in some cases, thermoplastic binders for use in metals [38]. This technology allows for these materials to be fused together in thin layers ranging between 0.003” and 0.006”. This allows SLS to create parts with accurate details and tolerances comparable to stereolithography. However, it has an added benefit in that the strength and durability of the parts it creates is much better. Additionally, the SLS process makes parts that have longer stability than stereolithography.

3.1.3. Materials There are a variety of different types of materials available for use in the SLS process [39]. The most beneficial characteristic is how durable and functional the materials are. They are made of nylon-based materials and create plastic prototypes. Other types of materials are used for investment casting patterns, rubber-like parts, and

materials that make metal prototypes. Additionally, there is continual research and development going on to bring new materials to market. Each of these materials require little to no post build processing to be ready to use, which cuts out several steps in post-processing of parts as compared to stereolithography. However, all of the selective laser sintering materials can be finished in multiple ways to meet the desire or needs of users [38]. Among other types of post-processing, parts can be sanded, painted, plated, tapped, or even machined. This allows for a higher grade of smoothness and appearance to parts and assemblies and also gives users an unlimited number of ways to use these parts. Depending on the material, up to 100% density can be achieved with material properties comparable to those found with traditional manufacturing methods.

3.1.4. Applications The SLS process is able to produce parts and/or prototypes that fit a wide range of different applications. The most widely used application of the process is functional

Chapter | 6.5

Rapid Prototyping of Ceramics

prototypes [40]. Some thermoplastics provide a choice of a durable and flexible plastic or a stiff and rigid plastic that have properties very close to an injection molded part. Parts created with these materials are ideal for end users to test for form, fit, and/or function. Other than functional prototypes, parts can be used as models, whether it is a presentation model for marketing purposes, or a showpiece for display. Yet another regular application that uses this technology is casting patterns i.e. investment casting because it is capable of making patterns with high accuracy and intricate detail. The SLS material could be a polystyrene that is coated with wax to solidify the pattern after being built in the machine. Once at the foundry, the pattern is melted away prior to pouring the molten metal. The parts made usually require a highly esthetic finish which is often called a master finish. Master finishing is a postproduction process that calls for the parts to be sanded extra smoothly and can even be primed, plated or painted to the end users desire.

3.1.5. Advantages The advantages of SLS parts are [41]: l

l

l

l

l

l

l

l

Fast efficient technique to build complex geometries due to the layer wise building method; SLS is produced faster than conventional tooling, taking off as much as 80e90% of the time it takes to create first parts; parts can be created in days as opposed to weeks; It is typically delivered at a lower cost compared to conventional toolingdas much as 90e95% less; Parts and/or assemblies that move and work that have a good surface finish and feature detail; This technique gives the capability of flexible snaps and living hinges as well as high stress and heat tolerance; Wide variety of materials such as flexible and rigid plastics, elastomeric materials, fully dense metals and casting patterns; Tight dimensional tolerances all the way down to thousandths of an inch; Finishing capabilities that include painting for presentations, tapping or threading for use and inserts for assemblies;

3.1.6. Limitations The limitations are [42]: l l

l

Surface roughness caused by the grain like structure; Lower dimensional accuracy than (SLA/mSLA), two sided tolerance >0.1 mm; Low surface hardness of sintered materials (HV z 170).

501

3.2. Fused Deposition Modeling 3.2.1. History FDM is an additive manufacturing technology commonly used for modeling, prototyping, and production applications [43]. The technology was developed by S. Scott Crump in the late 1980s and was commercialized in 1990. FDM uses the extrusion process to build 3D models. Stratasys introduced its first RP machine, the 3D modeler in early 1992 and started shipping the units later that year. Over the past decade, Stratasys has grown progressively, seeing her RP machine sales increase from six units in the beginning to a total of 1582 units in the year 2000.

3.2.2. Principle The principle of FDM revolves in surface chemistry, heat energy and layer deposition. The process constructs threedimensional objects directly from 3D CAD data. A temperature-controlled head extrudes thermoplastic material layer by layer [44]. The FDM process starts with importing an STL file of a model into preprocessing software. This model is oriented and mathematically sliced into horizontal layers varying from 0.127 to 0.254 mm thickness. A support structure is created where needed, based on the parts position and geometry. After reviewing the path data and generating the tool paths, the data is downloaded to the FDM machine. This machine operates in X, Y and Z axes, drawing the model one layer at a time. In this process, a plastic or wax material is extruded through a nozzle that traces the parts cross-sectional geometry layer by layer. The build material is usually supplied in filament form, but some setups utilize plastic pellets fed from a hopper instead. The nozzle contains resistive heaters that keep the plastic at a temperature just above its melting point so that it flows easily through the nozzle and forms the layer. The plastic hardens immediately after flowing from the nozzle and bonds to the layer below. Once a layer is built, the platform lowers, and the extrusion nozzle deposits another layer. The layer thickness and vertical dimensional accuracy is determined by the extruder die diameter, which ranges from 0.013 to 0.005 inches. In the XeY plane, 0.00100 resolution is achievable. Once the part is completed the support columns are removed and the surface is finished [45] (Fig. 9). The build material is usually supplied in filament form, but some setups use plastic pellets fed from a hopper instead. Several materials are available with different compromise between strength and temperature properties. One of the main materials used are acrylonitrile butadiene styrene (ABS) polymer, polycarbonates, polycaprolactone, polyphenylsulfones and waxes. A “water-soluble” material can be used for making temporary supports while

502

Handbook of Advanced Ceramics

FIGURE 9 (a) Movable table with nozzle ejecting molten material; (b) FDM process materials [46]. For color version of this figure, the reader is referred to the online version of this book.

manufacturing is in progress, this soluble support material is quickly dissolved with specialized mechanical agitation equipment using a precisely heated sodium hydroxide solution. As thermoplastic materials, PC/ABS (Polycarbonate-ABS) is one of the most widely used industrial thermoplastics. It offers the most desirable properties of both materialsdthe superior mechanical properties and heat-resistance of PC and the excellent property of ABS. PC/ABS blends are commonly used in automotive, electronics and telecommunications applications. It provides a good strength and rigidity in conjunction with toughness and temperature tolerance. The produced parts present a good surface appeal.

l

l

l

3.2.3. Applications FDM technique is used for the following application areas [47]: l

The creation of concept models used in early stages of product development. FDM models reduce costs and shorten development timelines;

The creation of functional prototypes for testing purposes. These prototypes allow to test in real world environments and make decisions that have a dramatic effect on the cost to manufacture to the product; Fabrication of end-use parts. Without the expense and lead time of traditional tooling or machining, FDM produces end-use parts tough enough for integration into the final product. Ideal for building small quantities of parts while waiting for tooling, FDM Technology makes it possible to get the products to market faster; Fabrication of manufacturing tools. FDM reduces the time it takes to create manufacturing tools by up to 85%. It produces manufacturing tools such as jigs and fixtures, tooling masters and production tooling in hoursdwithout expensive machining or tooling.

3.2.4. Advantages l

Minimal wastage. The FDM process build parts directly by extruding semiliquid melt onto the model. Only those material needed to build the part and its support

Chapter | 6.5

l

l

Rapid Prototyping of Ceramics

are needed, and material wastages are kept to a minimum. There is also little need for cleaning up the model after it has been built [48]; Ease of support removal. With the use of Break Away Support System and Water Works Soluble Support System, support structures generated during the FDM building process can be easily broken off or simply washed away. This makes it very convenient for users to get to their prototypes very quickly and there is very little or no post processing necessary; Ease of material change. Build materials, supplied in spool form (or cartridge form in the case of the Dimension or Prodigy Plus), are easy to handle and can be changed readily when the materials in the system are running low.

3.2.5. Limitations l

l

l

Limited accuracy. Parts built with the FDM process have limited accuracy due to the shape of the material used, i.e. the filament form. The filament usually used has a diameter of 1.27 mm and limits the accuracy of the built part; Slow building process. The building process is slow, as the whole cross-sectional area needs to be filled with building materials. Building speed is restricted by the extrusion rate or the flow rate of the build material from the extrusion head. As the build material used are plastics and their viscosities are relatively high, the build process could not be sped up [49]; Unknown shrinkage. As the FDM process extrudes the build material from its extrusion head and cools them rapidly on deposition, stresses induced by such rapid cooling invariably are introduced into the model. As such, shrinkages and distortions caused to the model built are a common occurrence and are usually difficult to predict.

503

Example: Elaboration of composites of piezoelectric ceramics and polymers with specific spatial arrangement of phases (connectivity) for the desired electromechanical performances (Fig. 10):

3.3. 3D Printing 3.3.1. History 3D printing is the process of creating three-dimensional objects from digital file using a materials printer, in a manner similar to printing images on paper. The term is most closely associated with additive manufacturing technology, where an object is created layer by layer. Since 2003, there has been large growth in the sale of 3D printers. Additionally, the cost of 3D printers has decreased. The technology is mainly used in the fields of jewelry, footwear, industrial design, architecture, engineering and construction (AEC), automotive, aerospace, dental and medical industries, education, geographic information systems, civil engineering, and many others.

3.3.2. Principle Three-dimensional printing operates by building parts in layers. From a computer (CAD) model of the desired part, a slicing algorithm draws detailed information for every layer [51]. Each layer begins with a thin distribution of powder spread over the surface of a powder bed. Using a technology similar to inkjet printing, a binder material selectively joins particles where the object formed. A piston that supports the powder bed and the part-inprogress lowers so that the next powder layer can be spread and selectively joined. This layer by layer process repeats until the part is completed [52]. Following a heat treatment, unbound powder is removed, leaving the fabricated part. Resolution is given in layer thickness and XeY resolution

FIGURE 10 Electronic devices made by FDM technique [50]. For color version of this figure, the reader is referred to the online version of this book.

504

Handbook of Advanced Ceramics

FIGURE 11

Three-dimensional printingdprinciple [53].

in dpi. Typical layer thickness is around 100 mm (0.1 mm), although some machines such as the Object Connex can print layers as thin as 16 mm. XeY resolution is comparable to that of laser printers. The particles (3D dots) are around 50e100 mm (0.05e0.1 mm) in diameter (Fig. 11). The building sequence is shown below:

3.3.3. Materials The 3DP process combines powders and binders with high geometric flexibility. The support gained from the powder bed means that overhangs, undercuts and internal volumes can be created (as long as there is a hole for the loose powder to escape). Material options, which include metal or ceramic powders, are somewhat limited but are inexpensive related to other additive processes [54]. Further, because different materials can be dispensed by different print heads, 3D Printing can exercise control over local material composition. Material can be in a liquid carrier, or it can be applied as molten matter. The proper placement of droplets can be used to create surfaces of controlled texture and to control the internal microstructure of the printed part.

3.3.4. Applications Standard applications include design visualization, prototyping/CAD, metal casting, architecture, education, geospatial, healthcare and entertainment/retail [55]. Other applications would include reconstructing fossils in paleontology, replicating ancient and priceless artifacts in archeology, reconstructing bones and body parts in forensic pathology and reconstructing heavily damaged evidence acquired from crime scene investigations. More recently, the use of 3D printing technology for artistic expression has been suggested.

3.3.5. Advantages Compared to other 3D techniques, 3D printing is optimized for speed, low cost, and ease of use, making it suitable for visualizing during the conceptual stages of engineering design through to early-stage functional testing [56]. No toxic chemicals like those used in stereolithography are required, and minimal postprinting finish work is needed; one need only to use the printer itself to blow off surrounding powder after the printing process. Bonded powder prints can be further strengthened by wax or thermoset polymer impregnation. The main advantages of the technology are stated below: l

l

l

l

l

Increases Innovation. Print prototypes in hours, obtain feedback, refine designs and repeat the cycle until designs are perfect; Improves Communication. Creates a full color, realistic 3D model to impart infinitely more information than a computer image; create physical 3D models quickly, easily and affordably for a wide variety of applications Speeds Time to Market. Compress design cycles by 3D printing multiple prototypes on demand, right in your office Reduces Development Costs. Cuts traditional prototyping and tooling costs. Identifies design errors earlier. Reduces travel to production facilities Wins Business. Brings realistic 3D models to prospective accounts, sponsors and focus groups

3.3.6. Limitations l l

l

Lack of material strength for large sized objects [57]; Generated objects possess rough and ribbed surface finish due to plastic beads or large size powder particles; 3D printers are expensive.

Chapter | 6.5

Rapid Prototyping of Ceramics

505

3.4. Inkjet Printing

Function testing

3.4.1. History The additive fabrication technique of inkjet printing is based on the 2D printer technique of using a jet to deposit tiny drops of ink onto paper. In the additive process, the ink is replaced with thermoplastic and wax materials, which are held in a melted state. When printed, liquid drops of these materials instantly cool and solidify to form a layer of the part. For this reason, the process if often referred to as thermal phase change inkjet printing. Several manufacturers have developed different inkjet printing devices that use the basic technique described above. Inkjet printers from Solidscape Inc., such as the ModelMaker (MM), use a single jet for the build material and another jet for support material. 3D Systems has implemented their MJM technology into their ThermoJet Modeler machines that utilize several hundred nozzles to enable faster build times. Inkjet printing offers the advantages of excellent accuracy and surface finishes. However, the limitations include slow build speeds, few material options, and fragile parts. As a result, the most common application of inkjet printing is prototypes used for form and fit testing. Other applications include jewelry, medical devices, and highprecision products.

3.4.2. Principle The inkjet printing process, as implemented by Solidscape Inc., begins with the build material (thermoplastic) and support material (wax) being held in a melted state inside two heated reservoirs. These materials are each fed to an inkjet print head which moves in the XeY plane and shoots tiny droplets to the required locations to form one layer of the part. Both the build material and support material instantly cool and solidify. After a layer has been completed, a milling head moves across the layer to smooth the surface. The particles resulting from this cutting operation are vacuumed away by the particle collector. The elevator then lowers the build platform and part so that the next layer can be built. After this process is repeated for each layer and the part is complete, the part can be removed and the wax support material can be melted away [58]. The technology is schematized below (Fig. 12): There are three types of Inkjet methods: thermal phase change inkjet, suspension inkjet and UV-curable inkjet. l

Thermal Phase Change Inkjet

During this process, the inkjet machines hold the build and support materials at elevated temperatures in a reservoir until the fabrication of the part beings. Once the process has begun, the liquid material moves through

Assembly

FIGURE 12 Inkjet printing process (Ceradrop). For color version of this figure, the reader is referred to the online version of this book.

thermally insulated tubing to individual jetting heads. The jetting heads then disperses the material in the form of tiny droplets to create part geometry. As the jetting heads continue to lay the material droplets, layers are formed. The droplets begin to cool and harden immediately after leaving the jetting head. When one layer is complete, a milling head passes over the previously created layer to produce a uniform thickness. The material particles created from this step are removed by a vacuum as the milling process is taking place. Before the next layer is created, the nozzles are checked to assure that the flow path is clear. If the nozzles do not need to be cleaned, the table will move down a set distance for the next layer to begin. Once the part is complete, the support structures are melted away. This type of machines is capable of a fast production of fine parts when using multiple jet heads. However, the accuracy is lower when using multiple jet heads, and usually additional operations are required in order to obtain a uniform layer thickness. l

Suspension inkjet [59]:

The main components of these inks are volatile organic compounds (VOCs), organic chemical compounds that have high vapor pressures. Inkjet printing is performed by the deposition of ceramic suspensions which dry by evaporation of the solvent that uses systems initially devoted to ink-jet printing on paper. The high print speed of many solvent printers demands special drying equipment, usually a combination of heaters and blowers. The substrate is usually heated immediately before and after the print heads apply ink. l

UV-curable inkjet:

These inks consist mainly of acrylic monomers with an initiator package. After printing, the ink is cured by exposure to strong UV-light. The advantage of UV-curable inks

506

Handbook of Advanced Ceramics

is that they “dry” as soon as they are cured, they can be applied to a wide range of uncoated substrates, and they produce a very robust image. Disadvantages are that they are expensive, require expensive curing modules in the printer, and the cured ink has a significant volume and so gives a slight relief on the surface. Though improvements are being made in the technology, UV-curable inks, because of their volume, are somewhat susceptible to cracking if applied to a flexible substrate. As such, they are often used in large “flatbed” printers, which print directly to rigid substrates such as plastic, wood or aluminum where flexibility is not a concern. UV Curable Ink Properties and Functions: l

l

l

l

Photoinitiators: Absorb the UV energy from the light source on the print head. Chemical reaction occurs that converts the liquid ink into a solid film. Monomers: Used as solvents because of their ability to reduce viscosity (thickness) and combine with other ink components. Hundred percent solids and do not release VOCs (volatile organic compounds). Monomers also add improved film hardness and resistance properties. Oligomers: Determine the final properties of the cured ink film, including its elasticity, outdoor performance characteristics and chemical resistance. Colorants: Can be dye-based or pigment-based. Usually, pigment-based because of the greater light fastness and durability of pigments compared with dyes. Pigments used in outdoor advertising and display applications have similar requirements to those used in automotive paints. Consequently, there is some crossover of use. While a pigment is selected on the basis of the required application, size control

and reduction along with dispersion technique are major components of ink formulation.

3.4.3. Applications l

l

l

l

l

Three-dimensional printing constructs a prototype by “printing” appreciably thick cross-sections of material on top of one another [60]; They are used in the production of Microelectromechanical systems (MEMS) Inkjet printers are used to form conductive traces for circuits, and color filters in LCD and plasma displays; This technique is in fairly common use in many labs around the world for developing alternative deposition methods that reduce consumption of expensive, rare, or problematic materials. These printers have been used in the printing of polymer, macromolecular, quantum dot, metallic nanoparticles, carbon nanotubes etc. The applications of such printing methods include organic thin-film transistors, organic light emitting diodes, organic solar cells, sensors, etc; Inkjet technology is used in the emerging field of bioprinting (Fig. 13).

Examples:

3.4.4. Advantages l l

l l

l

Multimaterial process; Deposition of different materials on successive layers via a multinozzle system; No limitation relative to the material; High flexibility as the 3D part is defined by CAD files [61]; High definition (z50 mm) controlled by the aperture of the printing head.

FIGURE 13 Microelectronic devices using inkjet technology (Ceradrop, Limoges, France). For color version of this figure, the reader is referred to the online version of this book.

Chapter | 6.5

Rapid Prototyping of Ceramics

3.4.5. Limitations l l

The ink is very expensive; The lifetime of inkjet prints produced by inkjets using aqueous inks is limited; they will eventually fade and the color balance may change. On the other hand, prints produced from solvent-based inkjets may last several years before fading, even in direct sunlight, and so-called “archival inks” have been produced for use in aqueous-based machines which offer extended life;

3.5. Stereolithography 3.5.1. History The term “stereolithography” was introduced in 1986 by Charles W. Hull. It was defined as a method for making solid objects by successively “printing” thin layers of the ultraviolet curable material one on top of the other [62]. Hull described a concentrated beam of ultraviolet light focused onto the surface of a vat filled with liquid photopolymer. The light beam draws the object onto the surface of the liquid layer by layer, causing polymerization or cross linking to give a solid. Because of the complexity of the process, it must be computer-controlled. In 1986 Chuck Hull founded the first company to generalize and commercialize this procedure, which is currently based in Rock Hill, SC. More recently, attempts have been made at constructing mathematical models of the stereolithography process, and designing algorithms that will automatically determine whether or not a proposed object may be constructed by this process.

3.5.2. Principle SL is a RP technique that allows the fabrication of threedimensional ceramic pieces with final properties (mechanical, thermal.) close to those obtained by classical processing techniques. Its spatial resolution is z50 mm. SLA machines have been made since 1988 by 3D Systems of Valencia, CA. To this day, 3D Systems is the industry leader, selling more RP machines than any other company. Because it was the first technique, it is regarded as a benchmark by which other technologies are judged. Early made prototypes were fairly brittle and prone to curing-induced distortion, but recent modifications have largely corrected these problems. This technology uses a UV laser beam that induces photopolymerization of a reactive system, consisting in the dispersion of ceramic particles into a green shape solid. The polymerization of patterns in stacked layers leads to the creation of complex 3D objects, as presented in the research of Halloran and Chartier [63,64]. The data of the object to build is transferred from a three-dimensional

507

CAD file to the automated equipment which physically builds the green part. The interest in this fabrication technique appeared in different domains of applications: microelectronics, with the elaboration of devices (filters, resonators) with high dimensional resolution and density, biomedical implants of hydroxyapatite with controlled porosities, structural complex alumina 3D parts. The process begins with the vat filled with the photocurable liquid resin and the elevator table set just below the surface of the liquid resin. The operator loads a three-dimensional CAD solid model file into the system. Supports are designed to stabilize the part during building. The translator converts the CAD data into a STL (* bmp) file. The control unit slices the model and support into a series of cross sections from 0.025 to 0.5 mm (0.001e0.02000 ) thick. The computer-controlled optical scanning system then directs and focuses the laser beam so that it solidifies a two-dimensional cross-section corresponding to the slice on the surface of the photocurable liquid resin to a depth greater than one layer thickness. The elevator table then drops enough to cover the solid polymer with another layer of the liquid curable suspension. A leveling wiper (or vacuum blade recoating system) moves across the surfaces to recoat the next layer of resin on the surface. The laser then draws the next layer. This process continues building the part from bottom up, until the system completes the part [65]. The part is then raised out of the vat and cleaned of excess polymer. The main components of the SLA system are a control computer, a control panel, a laser, an optical system and a process chamber as shown in the figure below (Fig. 14): Parts are built from a photocurable liquid resin that cures when exposed to a laser beam (basically, undergoing the photopolymerization process) which scans across the surface of the resin [66]. The building is done layer by layer, each layer being scanned by the optical scanning system and controlled by an elevation mechanism which lowers at the completion of each layer. The building sequence is presented below (Fig. 15): The layer thickness is controlled by a precision elevation mechanism. It will correspond directly to the slice thickness of the computer model and the cured thickness of the resin. The limiting aspect of the RP system tends to be the curing thickness rather than the resolution of the elevation mechanism. The important component of the building process is the laser and its optical scanning system. The key to the strength of the SLA is its ability to rapidly direct focused radiation of appropriate power and wavelength onto the surface of the liquid photopolymer resin, forming patterns of solidified photopolymer according to the cross-sectional data generated by the computer. In the SLA, a laser beam with a specified power and wavelength is sent through a beam expanding telescope to fill the optical aperture of a pair of

508

Handbook of Advanced Ceramics

FIGURE 14 SL process (successive layers production). For color version of this figure, the reader is referred to the online version of this book.

FIGURE 15 Building sequence in SL. For color version of this figure, the reader is referred to the online version of this book.

cross axis, galvanometer driven, and beam scanning mirrors. These form the optical scanning system of the SLA. The beam comes to a focus on the surface of a liquid photopolymer, curing a predetermined depth of the resin after a controlled time of exposure (inversely proportional to the laser scanning speed). The solidification of the liquid resin depends on the energy per unit area (or exposure) deposited during the motion of the focused spot on the surface of the photopolymer. There is a threshold exposure that must be exceeded for the photopolymer to solidify. To maintain accuracy and consistency during part building using the SLA, the polymerized thickness (Ep) and the cured line width (Lp) must be controlled. As such, accurate exposure and focused spot size become essential. Parameters which influence performance and functionality of the parts are the physical and chemical properties of the resin, the speed and resolution of the optical scanning system, the power, wavelength and type of the laser used the spot size of the laser, the recoating system, and the post-curing process. The projected density of energy Ei (mJ$cm 2) is related to the nominal laser power (P0),

scanning speed (ns) and beam radius (u0) through the following equation: Ei ¼

2P0 pu0 vs

(1)

3.5.3. Materials and Photopolymerization Process There are many types of liquid photopolymers that can be solidified by exposure to electro-magnetic radiation, including wavelengths in the gamma rays, X-rays, UV and visible range, or electron-beam. The vast majority of photopolymers used in the commercial RP systems, including 3D Systems are low viscosity acrylate oligomers and monomers. SLA machines are curable in the UV range. UV-curable photopolymers are resins which are formulated from photoinitiators and reactive liquid monomers/oligomers. There are a large variety of them and some may contain fillers and other chemical modifiers to meet specified chemical and

Chapter | 6.5

Rapid Prototyping of Ceramics

509

medium must be lower than 100 mPa.s at low shear rates, in order to hold powder concentrations above 55 vol% and insure good weatherability of the ceramic particles. In presence of a volume fraction F of ceramic powder, the increase of the viscosity of the reactive suspension can be evaluated by the KriegereDougherty equation: h ¼ h0

FIGURE 16 Components of a UV curable medium. For color version of this figure, the reader is referred to the online version of this book.

mechanical requirements. They are therefore called reactive suspensions. The schematic illustration of a reactive suspension is presented Fig. 16. The choice of formulation in SL is driven by the physical and chemical properties of the reactive monomer/ oligomer. The main criteria of choice are: l

l

Polymerization rate (reactivity). The threshold or critical energy of polymerization (Ec) should be sufficient enough to generate the photopolymerization process according to BeereLambert law [67]:   E0 (2) Ep ¼ Dpln Ec where Ec is the critical energy for photopolymerization, which is the minimum input energy necessary to trigger the curing process, Dp is the cure depth or beam penetration depth. The rheology of the monomer/oligomer blend and of the suspension. The viscosity of the organic

l

 1

b$F F0



½hŠF0

(3)

where h/h0 is the relative viscosity of the resin, F0 the volume fraction of the filler for a close-packed particles corresponding to an infinite viscosity (no flow) and [h] the intrinsic viscosity which depends on the shape of particles (h ¼ 2.5 for spheres). The mechanical properties of the polymerized layers (flexibility, planarity). The polymerized layer should exhibit high mechanical strength and planarity (Table 4).

Figure 17 and Table 4 show the main resins used in SL process. The UV curable system is composed of a photoinitiator dissolved in a reactive oligomer. The photoinitiator is chosen as a function of the laser wavelength. A reactive diluent may be added in order to improve the flow behavior of the suspensions, due to its low viscosity. A ceramic material used is then chosen depending on its specific area, refractive index, and mean particle size. In order to improve its dispersion in the UV curable suspension and the homogeneity of the microstructure of the green part, the ceramic powder is first deagglomerated by attrition milling. This deagglomeration is performed during several hours in solvent (ethanol) with the addition of a dispersant on the dry powder basis. It is then introduced in the reactive medium. The main steps involved in the elaboration of a reactive suspension are presented in the Fig. 18. The process through which photopolymers are cured is referred to as the photopolymerization process. It is a chain

TABLE 4 Physical Properties of Different Oligomer/Monomers used in SLA Viscosity (mPa s) at 25  C

Functionality

Mechanical properties ¼ f (functionality)

Name of product

HDDA

12

2

Strong

Hexanediol diacrylate

TDA

5

1

Flexible

Tridecyl acrylate

UA

25,000

2

Strong

Aliphatic urethane acrylate

PEAAM

70

3

Strong

Amine modified polyether acrylate

DTPTA

700

4

Strong

Di-Trimethylolpropane tetraacrylate

510

Handbook of Advanced Ceramics

Epoxy acrylates

Urethane Acrylates Ex: Aliphatic UA good weatherability of ceramic filler

High reactivity

Mechanical resistance ACRYLATE MONOMERS

Amine modified acrylates High reactivity and polymerization rate

Polyester acrylates Flexibility and weatherability

Special oligomers Adhesion promoters Resistant to agressive medium. FIGURE 17 Main resins used in SLA. For color version of this figure, the reader is referred to the online version of this book.

Powder deaglomeration

Dispersant added (1 wt% / filler content) + ethanol

Stage 2 :

Mixture :

Powder mixing with monomer (s) and reactive diluent

Monomer + powder + diluant ; manual mixing

Stage 1 :

Milling during 3 hrs in planetary mills V = 250trs/min

Addition of Photoinitiator (PI) and homogenization

Drying at 70°C Powder + adsorbed dispersant

Dispersed suspension

Stage 3 : Viscosity and reactivity tests Ec, Ep, Dp)

FIGURE 18 Protocol to make a reactive suspension. For color version of this figure, the reader is referred to the online version of this book.

of reactions and its initiation stage is of photochemical nature. Multifunctional monomers/oligomers lead to a dense cross-linked 3D network upon curing. Radical photopolymerization is the most used reaction for stereolithography applications. It takes place within the curing of

high UV reactivity monomers/oligomers having a vinyl unsaturation, such as acrylates and methacrylates. Their nature impacts the final properties of the reticulated polymer: elastic and opaque in case of acrylates; hard and transparent in case of methacrylates. Due to the faster cure

Chapter | 6.5

Rapid Prototyping of Ceramics

511

response of acrylates, the liquid to solid phase change is immediate (~0.2 s) upon intense radiation. The group of polyester acrylates encloses a large variety of low viscosity oligomers with different functionalities and molar masses (such as PEAAM resin used in this study). The radical photopolymerization process of polyester acrylates occurs in three stages: initiation, propagation and termination. l

Initiation

The first stage of the reaction corresponds to the initiation, i.e. the activation of a monomer/oligomer molecule by the photoinitiator (PA). The photopolymerization undergoes its decomposition (PA*) under the influence of UV irradiation in order to generate two free radicals R+1 and R+2 following the reaction: hv

PA ! PA / R+1 þ R+2

(4)

One or both radicals will combine with a molecule of the monomer/oligomer (M), and therefore generates a new radical RM, able to propagate the polymerization reaction as follows: 

R þ M / RM



(5)

The irradiation of the monomer/oligomer instantly leads to an increase of np that reaches its maximum in a few seconds when most of the active double bonds are consumed [68,69]. This phenomenon results in an increase of the viscosity of the medium, leading to a gel formation at the beginning of the polymerization. This effect is called self-acceleration or gel effect. During this stage, the mobility of free radicals is limited in the reactive medium. The self-acceleration is followed by a rapid decrease of the polymerization rate. The viscosity continues to grow and the mobility of reactive species greatly decreases. This stage is called self-deceleration or glass stage and marks the vitrification state of the polymer. At the end of the photopolymerization reaction, the conversion rate is lower than 100%. The technique used to determine the degree of the reaction, respectively the rate of conversion, of an acrylate double bond is Real-Time Infrared Spectroscopy (RTIR). This technique is based on the variation in absorption of reactive functions at each time of the reaction. For one of the most used oligomer-amine modified polyether acrylate (PEAAM) in stereolithography formulations, the rate of conversion at each point of the photopolymerization reaction has been estimated using the following formula:

The rate of initiation na is directly related to the quantum yield of initiation fa and to the absorbed radiation intensity Iabs. va ¼ fa $Iabs l

(6)

Propagation

This stage corresponds to the successive addition of monomer/oligomer units on the growing polymer chain, in order to generate larger radicals: 

kp



RMn þ M / RMnþ1

(7)

The rate of propagation yb, reflecting the variation of the monomer/oligomer concentration [M], along the reaction is:   (8) vp ¼ kp RMn $½MŠ where kp is the propagation rate constant.

The kinetics of the photopolymerization reaction of multifunctional acrylates shows a particular behavior: selfacceleration and self-deceleration phenomena accompanied by an incomplete conversion of functional groups. l

Termination

The termination corresponds to the end of the photopolymerization process, i.e. the end of the chain growth. The rate of termination nt is expressed by:   nt ¼ kt RMn (9) where kt is the termination rate constant.

acrylate

%CðtÞ

A810or1636 A810or1636 t 0 A1720 A1720 t 0  100 ¼ 810or1636 At A1720 t

(10)

where A0 is the initial absorbance with no radiation and At is the absorbance at a time t of the radiation. The two bands B810 and B1636 were used to follow the absorption decrease of the acrylate function (Fig. 19). The rate of polymerization (Rp) has been obtained by derivation of the rate of conversion versus time.

3.5.4. Applications The SLA technology provides manufacturers with cost justifiable methods for reducing time to market, lowering product development costs, gaining greater control of their design process and improving product design. It offers a variety of real world applications in diverse fields from manufacturing to biomedicine [70]. Before taking a design into full-scale production, stereolithography allows businesses to evaluate their design for feasibility, manufacturability, ergonomics, and esthetics without slow and costly prototyping. Models can also be tested using Optical Stress Analysis to study the effects of external loadings, torsion, tension, pressure and placed in wind tunnels to measure aerodynamics. Manufacturing firms also use SLA models for casting, tooling and production line design. Businesses also use these models as a marketing tool since it enables

512

Handbook of Advanced Ceramics

Absorbance

(a)

0,55 0,5 0,45 0,4 0,35 0,3 0,25 0,2 0,15 0,1 0,05 0

C=O O

C= C=H H

C=C

1700

1500

1300

1100

900

700

Conversion rate (%)

(b)

100

9

90

8

80

7

70

6

60

5

50

4

40

3

30 20

2

10

1

0

Rp (%.s-1)

Wavenumber (cm-1)

0 0

6

12

18

24

30

36

42

48

54

Time (s)

FIGURE 19 PEAAM resin: (a) FTIR spectrum. (b) Rate of conversion. For color version of this figure, the reader is referred to the online version of this book.

them to present a physical representation of the product before production is complete. SL prototyping lowers the cost of production while increasing product durability, benefiting both industry and consumers. SLA is also becoming prevalent in the field of biomedical engineering. Surgeons are able to make models of patients and implants to practice delicate routines in advance. These models are also being used to educate families, patients, and students. Examples: SL is used to fabricate different structural 3D parts used in different areas as shown Fig. 20: It is particularly suited for the fabrication of complex 3D microdevices such as resonators and RF filters used in

High frequency operation circuits, due the fabrication precision and spatial resolution of the technique [71]. SLA is widely used in biomedicine for bone scaffold tissue engineering and the fabrication of biocompatible implants used as bone substitute in the human body (3DCeram, Limoges, France) (Fig. 21).

3.5.5. Advantages One of the appealing aspects about SL is that a functional part can be created within one day [72]. The length of time it takes to produce any one part depends on the size and complexity of the project and can take anywhere from a

FIGURE 20 Different parts made by SLA (3D Ceram, Limoges, France). For color version of this figure, the reader is referred to the online version of this book.

Chapter | 6.5

Rapid Prototyping of Ceramics

513

FIGURE 21 Biomedical applications of SLA process (3D Ceram, Limoges, France). For color version of this figure, the reader is referred to the online version of this book.

few hours to more than a day. Most SL machines can produce parts with a maximum size of approximately 50 cm  50 cm  60 cm (2000  2000  2400 ) and some are capable of producing parts of more than 2 m in a single piece (the Mammoth has a build platform of 210 cm  70 cm  80 cm). Prototypes made by stereolithography can be very beneficial as they are strong enough to be machined and can be used as master patterns for injection molding, thermoforming, blow molding, and also in various metal casting processes. Moreover this technology has one of the best surface finishes amongst RP technologies.

3.5.6 Limitations The main disadvantage of stereolithography is that the process is often expensivedthe photocurable resin costs anywhere from $80 to $210 per liter. The parts made by stereolithography often require post-curing and postprocessing [73]. The last one includes the removal of the support and residual noncured suspension. This could damage the part and is time consuming.

3.6. Microstereolithography (mSLA)

for the microfabrication of complex three-dimensional parts, based on layer by layer polymerization of a photosensitive resin [74]. This technique has significantly improved both lateral and vertical resolution of stereolithography. As the fabrication method of this technique relies on the space-resolved and light induced polymerization of a resin, the improvement of its resolution is mostly related to the reduction of the interaction volume between light and matter. This objects made are destined in the fields of microrobotics, microfluidics or microsystems. Although most research teams involved in the microstereolithography field have investigated its use as a microfabrication technique, the most promising application field of this technology is nevertheless the RP domain that faces an increasing demand of small-size high-resolution prototype parts. This unique technology provides building of complex microstructures used in electromechanical systems (MEMS). In principle, this technique uses an UV light beam (365 nm) that scans the surface of a digital mask (DMD) and by reflection, comes into contact with the surface of a photosensitive resin and induces a process of photopolymerization.

3.6.1. History

3.6.2. Principle

Microstereolithography, a technique derived from the stereolithography process, has emerged as a new technique

The fabrication process of a part by microstereolithography firstly consists in the creation of a CAD

514

file of the object to build. Then the CAD file is converted to STL file. This one contains the closed surfaces of the 3D model of the object, broken down into elementary triangles. Then the object is digitally sliced, allowing defining the forms for each layer to cure. The final file (with the process parameters-beam intensity, scanning speed) is finally sent to the automated machine to begin the process technology. The high dimensional accuracy of this technique has been achieved using a dynamic mask composed of a matrix of 1024  768 micromirrors (DigitalMicromirror device, Texas Instruments) with a size of 14 microns [75]. Each individual micromirror can be rotated 12 . The source of radiation consists of a UV light source (Hammamatsu) including an optical waveguideanda365 nm filter. An integrated shutter allows the control of the exposure time and the projected energy density. The beam scans the surface of the DMD onto which is loaded the section of the object to build. The reflected image of the object passes through an optical system comprising a focusing lens that provides a transmission above 83%, and is then projected onto the work surface onto which a photosensitive resin is spread out. For a magnification of 1, the dimensions of the object vary between 5 mm and 30 m (lateral resolution of the system). This is due to the light scattering contribution of the ceramic particles in a curable suspension. A spreading and deposit systems control the uniformity and thickness of the thin layers. The resolution of the technique (at SPCTS laboratory, Limoges, France) is determined by the size of the micromirrors (14 mm). There are two types of microstereolithography processes: laser and integral. Integral microstereolithography processes (also named projection microstereolithography) are based on the projection of an image created by a dynamic mask on the surface of the photosensitive resin, and consequently, the object to be built is described by a series of black and white bitmap files rather than by vectors as it is the case in scanning methods. These bitmap files are then used to shape the light beam using a dynamic mask, such that the image of each layer can be projected on the surface of the liquid resin after being reduced and focused by an appropriate optical system. The projected image induces a solidification of the irradiated areas, and a shutter controls the duration of the irradiation step. The superposition of many layers of different shapes allows the fabrication of a complex object in a similar way as for conventional RP techniques. Integral microstereolithography processes are inherently faster than scanning techniques, as the irradiation of a complete layer is done in one-step, regardless to its complexity. Additionally, the light flux density on the surface of the photopolymerizable resin is much smaller when projecting an image than when focusing a light beam in one point, which allows avoiding problems related to unwanted thermal polymerization.

Handbook of Advanced Ceramics

The schematic diagram of integral microstereolithography (SPCTS Laboratory, Limoges, France) is shown on Fig. 22. l

Digital Micromirror Device (DMD)

The Digital Micromirror Device made by Texas Instruments, is widely used in video projection applications and has been used as dynamic mask in integral microstereolithography machines. This component is in fact an array of micromirrors actuated by electrostatic forces, and is used as a light switch in the mSLA [76] Each 14 mm square mirror can be independently actuated either to reflect the incident light beam either into or out of the pupil of the optical system, such that the corresponding pixel of the projected image appears bright or dark. For video projection applications, gray levels and colored images can be produced by combining the rapid on/off movement of the mirrors with a rotating color wheel, but these features are not used in the mSLA apparatus. The first microstereolithography machine using the DMD component as dynamic mask was developed by Bertsch et al. and used a metal halide lamp combined with optical filters to select a band of visible wavelength for the irradiation of the resin. Different photopolymerizable resins were developed for this apparatus. In a first step, a high resolution resin reacting at 530 nm was used. It allowed the manufacture many components, but showed important curl deformations. In a second step, resins reacting at 410 nm were developed. Their composition is a lot closer to conventional acrylate-based stereolithography resins, even if they do not react with UV light, but in the blue part of the visible spectrum. A second machine was later built by the same research team with an improved resolution (XGA: 1024  768 pixels), and an irradiation in the UV. An acrylate-based resin was formulated specifically for this machine, which was used in an industrial context to produce high-resolution prototypes.

3.6.3. Materials The fabrication sequence, materials used and photopolymerization process are the same as in stereolithography ones. The curable suspension contains a reactive oligomer/ monomer, a photoinitiator and a ceramic powder in concentration above 50 vol% in order to obtain a high density and good mechanical properties of the sintered part. Here, the viscosity of the suspensions should not exceed 5 Pa.s. Therefore it is possible to improve the quality and resolution of the polymerized layers by adding a high viscosity oligomer such as amine modified polyether acrylates.

3.6.4. Applications As the market for miniaturized products grows rapidly, there is also an increasing need for high-resolution

Chapter | 6.5

Rapid Prototyping of Ceramics

515

Labview program DMD (Texas Instruments) Digital Micro mirror Device

OFF

SPCTS Laboratory

ON

European Ceramic Center Parc Technopole d’Ester 12, rue Atlantis 87068 Limoges

Convergent lens

Suspension Deposition system FIGURE 22 Schematic presentation of integral mSLA apparatus. For color version of this figure, the reader is referred to the online version of this book.

prototype parts. When small size objects have to be built with dimensions of only a few millimeters or less, current RP technologies are limited with respect to the feasibility of small features: openings and small holes are difficult to make and have to be cleaned once the prototype is built, which is particularly difficult. Another aspect, which has to be taken into consideration for small components, is that manual surface finishing can be of a great challenge. Microstereolithography is used for the manufacturing of small size high-resolution components [77]. There is a growing demand of high-resolution prototype objects in particular in the medico-technical industry. For example, medical devices in which optical and chemical sensors could be embedded require very high precision and definition that microstereolithography has to offer. Hearing-aid manufacturers try to design lightweight products small enough not to be detected, comfortable, with rounded shapes to be close from the natural geometry of the ear canal. For such applications it is necessary to prototype small mechanical components with very specific details. Most 3D models made by microstereolithography can also be used as a simulation tool for medical teams before exercising an operation.

In the electro-technical industry, microstereolithography can be used for the prototyping of small connectors, and other microdevices. This technique has also been used for prototyping MEMS components and optimizing their geometry, before investing in photomasks and using more conventional technologies for mass production. In the future, the manufacturing of hybrid polymer structures by combining various types of polymers, such as conductive polymers, polymers of various refractive index or flexible polymers could also lead to the creation of new optical, chemical and biochemical microsystems with the microstereolithography technology. Example: An example of a complex 3Dresonator microstructure produced by the process of microstereolithography is shown in Fig. 23 [78]. This resonant structure is composed of three different parts. The first part represents stacks of rods with a missing row in the middle. The other two parts are metallized alumina plates and placed above and below the stacks of rods, forming a sandwich structure. The dimensions of the cavity created are 1.33 mm  1.33 mm  0.65 mm and operate between 140e145 GHz in

516

Handbook of Advanced Ceramics

FIGURE 23 Resonator structure: stack of rods between two parallel plates (SPCTS Laboratory, Limoges, France). For color version of this figure, the reader is referred to the online version of this book.

Rod

Cavity

y

500 µm

0

the fundamental mode TE101 (the electric field is polarized along the axis of the rods). This low-loss alumina compact structure can be easily integrated into microwave circuits for signal filtering, while providing are latively high quality factor (2500) and a reflection coefficient z100%.

3.6.5. Advantages Unlike stereolithography, microstereolithography uses an integral irradiation of the layer section to be built thus creating less fabrication defects (the density of energy is evenly distributed along the surface of the represented section) [79]. By operating with higher volume rate of oligomer in the compositions, it is possible to improve the definition of the created objects and to reduce the lateral overcure phenomenon.

3.6.6. Limitations One of the major limitations of the microstereolithography technology is related to the materials that can be used in this manufacturing technique: only a few polymers can be used, acrylates in general or eventually epoxies. Because of their three-dimensional geometry, most objects produced by microstereolithography cannot be molded, which implies that, for some applications, microstereolithography is no longer a RP technology but a manufacturing technique [80]. In this case, the produced objects need to have adequate mechanical, chemical or physical characteristics. As a result, studies on the use of new materials for microstereolithography have been started, and in particular on the use of composite materials made of ceramic particles embedded in a polymer matrix as reactive medium in this technology. The ultimate goal is the production of microcomponents in ceramic materials, which can be obtained by sintering the composite. However, for sintering successfully such composite components, the load of particles embedded in the resin has to be sufficiently high, which increases significantly the viscosity of the chemical media.

x

4. DATA FILE FORMATS 4.1. STL File Format The natural file format for RP techniques such as stereolithography would be a series of closed polygons corresponding to different Z-values. However, since it’s possible to vary the layer thicknesses for a faster though less precise build, it seemed easier to define the model to be built as a closed polyhedron that could be sliced at the necessary horizontal levels [81]. The STL file format is capable of defining a polyhedron with any polygonal facet, but in practice it’s only ever used for triangles, which means that much of the syntax of the ASCII protocol is superfluous. STL files are supposed to be closed and connected like a combinatorial surface, where every edge is part of exactly two triangles, and not self-intersecting. Since the syntax does not enforce this property, it can be ignored for applications where the closeness doesn’t matter. The closeness only matters insofar as the software which slices the triangles requires it to ensure that the resulting 2D polygons are closed. Sometimes such software can be written to clean up small discrepancies by moving endpoints of edges that are close together so that they coincide. STL is a file format native to the stereolithography CAD software created by 3D Systems. This file format is supported by many other software packages; it is widely used for RP and CAM. STL files describe only the surface geometry of a three-dimensional object without any representation of color, texture or other common CAD model attributes. The STL format specifies both ASCII and binary representations. Binary files are more common, since they are more compact. In 2011, ASTM replaced the STL format with the Additive Manufacturing File Format (AMF), which has native support for color, multiple materials, and constellations. An STL file describes a raw unstructured triangulated surface by the unit normal and vertices (ordered by the right-hand rule) of the triangles using a three-dimensional Cartesian coordinate system.

Chapter | 6.5

l

Rapid Prototyping of Ceramics

ASCII STL file.

An ASCII STL file begins with the line: solid name where name is an optional string (though if name is omitted there must still be a space after solid). The file continues with any number of triangles, each represented as follows: facet normal ni nj nk outerloop vertex v1x v1y v1z vertex v2x v2y v2z vertex v3x v3y v3z endloop endfacet where each n or v is a floating point number in signmantissa ‘e’-sign-exponent format, e.g. “ 2.648000e002”. The file concludes with: endsolid name The structure of the format suggests that other possibilities exist (e.g. facets with more than one ‘loop’, or loops with more than three vertices) but in practice, all facets are simple triangles. White space (spaces, tabs, new lines) may be used anywhere in the file except within numbers or words. The spaces between ‘facet’ and ‘normal’ and between ‘outer’ and ‘loop’ are required l

Binary STL file.

Because ASCII STL files can become very large, a binary version of STL exists. A binary STL file has an 80 character header (which is generally ignoreddbut which should never begin with ‘solid’ because that will lead most software to assume that this is an ASCII STL file). Following the header is a 4 byte unsigned integer indicating the number of triangular facets in the file. Following that is data describing each triangle in turn. The file simply ends after the last triangle. Each triangle is described by twelve 32-bit-floating point numbers: three for the normal and then three for the X/Y/Z coordinate of each vertexdjust as with the ASCII version of STL. After the twelve floats there is a two byte unsigned ‘short’ integer that is the ‘attribute byte count’din the standard format, this should be zero because most software does not understand anything else. UINT8 [80] e Header UINT32 e Number of triangles foreach triangle REAL32 [3] e Normal vector REAL32 [3] e Vertex 1 REAL32 [3] e Vertex 2 REAL32 [3] e Vertex 3 UINT16 e Attribute byte count end

517

There are at least two variations on the binary STL format for adding color information: l

l

The VisCAM and SolidView software packages use the two ‘attribute byte count’ bytes at the end of every triangle to store a 15 bit RGB color: l bit 0e4 are the intensity level for blue (0e31) l bits 5e9 are the intensity level for green (0e31) l bits 10e14 are the intensity level for red (0e31) n bit 15 is 1 if the color is valid n bit 15 is 0 if the color is not valid (as with normal STL files) The Materialize Magics software does things a little differently. It uses the 80 byte header at the top of the file to represent the overall color of the entire part. If color is used, then somewhere in the header should be the ASCII string “COLOR¼” followed by four bytes representing red, green, blue and alpha channel (transparency) in the range 0e255. This is the color of the entire object unless overridden at each facet. Magics also recognize a material description; a more detailed surface characteristic. Just after “COLOR ¼ RGBA” specification should be another ASCII string, “MATERIAL¼” followed by three colors (3  4 bytes): first is a color of diffuse reflection, second is a color of specular highlight, and third is an ambient light. Material setting wags are preferred over color. The per-facet color is represented in the two ‘attribute byte count’ bytes as follows: l bit 0e4 are the intensity level for red (0e31) l bits 5e9 are the intensity level for green (0e31) l bits 10e14 are the intensity level for blue (0e31) n bit 15 is 0 if this facet has its own unique color n bit 15 is 1 if the per-object color is to be used

The red/green/blue ordering within those two bytes is reversed in these two approachesdso whilst these formats could easily have been compatible the reversal of the order of the colors means that they are notdand worse still, a generic STL file reader cannot automatically distinguish between them. There is also no way to have facets be selectively transparent because there is no per-facet alpha valuedalthough in the context of current RP machinery, this is not important. In both ASCII and binary versions of STL, the facet normal should be a unit vector pointing outwards from the solid object. In most software this may be set to (0,0,0) and the software will automatically calculate a normal based on the order of the triangle vertices using the ‘right-hand rule’. Some STL loaders check that the normal in the file agrees with the normal they calculate using the right-hand rule and warn you when it does not. Other software may ignore the facet normal entirely and use only the right-hand rule. Although it is rare to specify a normal that cannot be

518

Handbook of Advanced Ceramics

calculated using the right-hand rule, in order to be entirely portable, a file should both provide the facet normal and order the vertices appropriately. A notable exception is SolidWorks which uses the normal for shading effects.

4.2. Tessellated Model Tessellation is the process of creating a two-dimensional plane using the repetition of a geometric shape with no overlaps and no gaps. Generalizations to higher dimensions are also possible. Tessellations frequently appeared in the art of M. C. Escher, who was inspired by studying the Moorish use of symmetry in the Alhambra tiles during a visit in 1922. Tessellations are seen throughout art history, from ancient architecture to modern art. In Latin, tessella is a small cubical piece of clay, stone or glass used to make mosaics [82]. The word “tessella” means “small square” (from “tessera”, square, which in its turn is from the Greek word for “four”). It corresponds with the everyday term tiling which refers to applications of tessellations, often made of glazed clay. Examples of tessellations in the real world include honeycombs and pavement tiling. In the subject of computer graphics, tessellation techniques are often used to manage datasets of polygons and divide them into suitable structures for rendering. Normally, at least for real-time rendering, the data is tessellated into a triangle, which is sometimes referred to as triangulation. In computer-aided design the constructed design is represented by a boundary representation topological model, where analytical 3D surfaces and curves, limited to faces and edges constitute a continuous boundary of a 3D body. Arbitrary 3D bodies are often too complicated to analyze directly. So they are approximated (tessellated) with a mesh of small, easy-to-analyze pieces of 3D volumedusually either irregular tetrahedrons, or irregular hexahedrons. The mesh is used for finite element analysis as shown on the Fig. 24. The mesh of a surface is usually generated per individual faces and edges (approximated to polylines) so that original limit vertices are included into mesh. To ensure that approximation of the original surface suits the needs of the further processing, three basic parameters are usually defined for the surface mesh generator: l

l

The maximum allowed distance between the planar approximation polygon and the surface (aka “sag”). This parameter ensures that mesh is similar enough to the original analytical surface (or the polyline is similar to the original curve). The maximum allowed size of the approximation polygon (for triangulations it can be maximum allowed length of triangle sides). This parameter ensures enough detail for further analysis.

FIGURE 24 A tessellation of a disk used to solve a finite element problem. For color version of this figure, the reader is referred to the online version of this book.

l

The maximum allowed angle between two adjacent approximation polygons (on the same face). This parameter ensures that even very small humps or hollows that can have significant effect to analysis will not disappear in mesh.

Algorithm generating mesh is driven by the parameters. Some computer analyses require adaptive mesh, which is made finer (using stronger parameters), in regions where the analysis needs more detail.

4.3. Other File Formats 4.3.1. IGES Format IGES (Initial Graphics Exchange Specification) is a format used to exchange graphics information between commercial CAD systems. It was setup as an American National Standard in 1981. The IGES file can precisely represent CAD models [83]. It includes not only the geometry information (Parameter Data Section) but also topological information (Directory Entry Section). In the IGES, surface modeling, constructive solid geometry and boundary representation are introduced. Especially, the ways of representing the regularized operations for union, intersection, and difference have also been defined. The advantages of the IGES standard are its wide adoption and comprehensive coverage. Since IGES was setup as American National Standard, virtually every commercial CAD/CAM system has adopted IGES implementations. Furthermore, it provides the entities of points, lines, arcs, splines, NURBS surfaces and solid elements. IGES is a generally used data transfer medium which interfaces with various CAD systems. Advantages of using IGES over current approximate methods include

Chapter | 6.5

Rapid Prototyping of Ceramics

precise geometry representations, few data conversions, smaller data files and simpler control strategies. However, the problems are the lack of transfer standards for a variety of CAD systems and system complexities.

4.3.2. HPGL Format HPGL, sometimes hyphenated as HPeGL, was the primary printer control language used by HewlettePackard plotters. The name is a start for HewlettePackard Graphics Language. It later became a standard for almost all plotters [84]. HewlettePackard’s printers also usually support HPGL in addition to PCL. The language is formed from a series of two letter codes, followed by optional parameters. For instance an arc can be drawn on a page by sending the string: AA100; 100; 50; This means Arc Absolute, and the parameters place the center of the arc at absolute coordinates 100,100 on the page, with a starting angle of 50 measured counterclockwise. A fourth optional parameter (not used here) specifies how far the arc continues, and defaults to 5 . Typical HPGL files start with a few setup commands, followed by a long string of graphics commands. The original HP/GL-Language did not support definition of line width, as this parameter was determined by the pens loaded into the plotter. With the advent of the first inkjet plotters, line width for the “pens” specified within the HP/ GL-files had to be set at the printer so it would know what line width to print for each pen, a cumbersome and errorprone process. With HP/GL-2, definition of line width was introduced into the language and allowed for elimination of this step. Also, among other improvements a binary file format was defined that allowed for smaller files and shorter file transfer times, and the minimal resolution was reduced.

519

surfaces and curves. NURBS represent in compact form geometrical shapes, and can be handled efficiently by programs while still allowing human interaction. The SLC file extension is also used by SolidTools for its solid imaging. This RP component uses a database in the translation of 3D geometry, to parts or physical models that uses a variety of materials and resins. Files in the .slc file extension are the format used for output files by StudioTools. A NURBS surface is translated first to a file in the .slc file extension, before it can be read by the software for the solid imaging machine. Furthermore, geometry from StudioTools can also be converted to a file in the .slc file extension. Files in this format are able to cut contours in 2D of the three-dimensional base. An advantage to using the SLC file extension is that the geometry description of NURBS in StudioTools is sliced directly, and so less iteration is needed between the primary geometry and the sent data to be built in the Solid Imaging machine.

4.3.4. LEAF File The LEAF or Layer Exchange ASCII Format is generated by Helsinki University of Technology. To describe this data model, concepts from the object-oriented paradigm are borrowed. At the top level, there is an object called LMTfile (Layer Manufacture Technology file) that can contain parts which in turn are composed of other parts or by layers [86]. Ultimately, layers are composed of 2D primitives and currently the only ones which are planned for implementation are polylines. For example, an object of a given class is created. The object classes are organized in a simple tree. Attached to each object class is a collection of properties. A particular instance of an object specifies the values for each property. Objects inherit properties from their parents. In LEAF, the geometry of an object is simply one among several other properties.

4.3.3. SLC File The SLC file extension is a CAD slice file that is used with CAD software such as Rhinoceros 3D. A file in the .slc file extension can be created using the Model works command in Rhinoceros 3D and sent to the Rhino model [85]. The software has an edit box for slice files, from which you can also enter the name for the file in the .slc file extension that you want stored. This can be sent to other 3D modeling machines such as Solidscape. The thickness of CAD slice files in the .slc file extension can be edited in Rhinoceros 3D, using a drop down menu that allow you to set the thickness of your build. Files in the .slc file extension can be generated from slice created by meshes from a NURBS model. NURBS is also known as Nonuniform rational B-spline and is a mathematical model, which is utilized in computer graphics for the representation and generation of

5. APPLICATIONS OF RP TECHNIQUES RP is widely used in the automotive, aerospace, medical, and consumer products industries. Although the possible applications are virtually limitless, nearly all fall into one of the following categories: prototyping, rapid tooling, or rapid manufacturing. By exchanging prototypes early in the design stage, manufacturing can start tooling up for production while the art division starts planning the packaging, all before the design is finalized. Prototypes are also useful for testing a design, to see if it performs as desired or needs improvement. Engineers have always tested prototypes, but RP expands their capabilities. First, it is now easy to perform iterative testing: build a prototype, test it, redesign, build and test, etc. Such an approach would be far

520

Handbook of Advanced Ceramics

too time-consuming using traditional prototyping techniques, but it is easy using RP [87]. In addition to being fast, RP models can do a few things metal prototypes cannot. For example, Porsche used a transparent stereolithography model of the 911 GTI transmission housing to visually study oil flow. Snecma, a French turbomachinery producer, performed photoelastic stress analysis on a SLA model of a fan wheel to determine stresses in the blades.

l

l

5.1. Rapid Tooling A much-anticipated application of RP is rapid tooling, the automatic fabrication of production quality machine tools. Tooling is one of the slowest and most expensive steps in the manufacturing process, because of the extremely high quality required. Tools often have complex geometries, yet must be dimensionally accurate to within a hundredth of a millimeter. In addition, tools must be hard, wear-resistant, and have very low surface roughness (about 0.5 mm root mean square). To meet these requirements, molds and dies are traditionally made by CNC-machining, electrodischarge machining, or by hand. All are expensive and time consuming, so manufacturers try to incorporate RP techniques to speed the process. Peter Hilton, president of Technology Strategy Consulting in Concord, MA, believes that “tooling costs and development times can be reduced by 75% or more” by using rapid tooling and related technologies. Rapid tooling can be divided into two categories, indirect and direct [88].

l

Sand Casting: a RP model is used as the positive pattern around which the sand mold is built. LOM models, which resemble the wooden models traditionally used for this purpose, are often used. If sealed and finished, a LOM pattern can produce about 100 sand molds. Investment Casting: some RP prototypes can be used as investment casting patterns. The pattern must not expand when heated, or it will crack the ceramic shell during autoclaving. Both Stratasys and Cubital make investment casting wax for their machines. Paper LOM prototypes may also be used, as they are dimensionally stable with temperature. The paper shells burn out, leaving some ash to be removed. To counter thermal expansion in stereolithography parts, 3D Systems introduced QuickCast, a build style featuring a solid outer skin and mostly hollow inner structure. The part collapses inward when heated. Likewise, DTM sells Trueform polymer, a porous substance that expands little with temperature rise, for use in its SLS machines. Injection molding: CEMCOM Research Associates, Inc. has developed the NCC Tooling System to make metal/ceramic composite molds for the injection molding of plastics. First, a stereolithography machine is used to make a match-plate positive pattern of the desired molding. To form the mold, the SLA pattern is plated with nickel, which is then reinforced with a stiff ceramic material. The two mold halves are separated to remove the pattern, leaving a matched die set that can produce tens of thousands of injection moldings.

5.1.1. Indirect Tooling

5.1.2. Direct Tooling

Most rapid tooling today is indirect: RP parts are used as patterns for making molds and dies. RP models can be indirectly used in a number of manufacturing processes:

To directly make hard tooling from CAD data is the biggest achievement of rapid tooling. Realization of this objective is still several years away, but some strong strides are being made:

l

Vacuum Casting: in the simplest and oldest rapid tooling technique, a RP positive pattern is suspended in a vat of liquid silicone or room temperature vulcanizing rubber. When the rubber hardens, it is cut into two halves and the RP pattern is removed. The resulting rubber mold can be used to cast up to 20 polyurethane replicas of the original RP pattern. A more useful variant, known as the Keltool powder metal sintering process, uses the rubber molds to produce metal tools. Developed by 3 M and now owned by 3D Systems, the Keltool process involves filling the rubber molds with powdered tool steel and epoxy binder. When the binder cures, the “green” metal tool is removed from the rubber mold and then sintered. At this stage the metal is only 70% dense, so it is infiltrated with copper to bring it close to its theoretical maximum density. The tools have fairly good accuracy, but their size is limited to less than 25 centimeters.

l

RapidTool: a DTM process that selectively sinters polymer-coated steel pellets together to produce a metal mold. The mold is then placed in a furnace where the polymer binder is burned off and the part is infiltrated with copper (as in the Keltool process). The resulting mold can produce up to 50,000 injection moldings.

In 1996 Rubbermaid produced 30,000 plastic desk organizers from a SLS-built mold. This was the first widely sold consumer product to be produced from direct rapid tooling. l

LENS is a process developed at Sandia National Laboratories and Stanford University that can create metal tools from CAD data. Materials include 316 stainless steel, Inconel 625, H13 tool steel, tungsten, and titanium carbide cermets. A laser beam melts the top layer of the part in areas where material is to be added. Powder metal is injected into the molten pool, which

Chapter | 6.5

l

l

l

Rapid Prototyping of Ceramics

then solidifies. Layer after layer is added until the part is complete. Unlike traditional powder metal processing, LENS produces fully dense parts, since the metal is melted, not merely sintered. The resulting parts have exceptional mechanical properties, but the process currently works only for parts with simple, uniform cross sections. The system has been commercialized by MTS corporation. Direct AIM (ACES Injection Molding): a technique from 3D Systems in which stereolithography-produced cores are used with traditional metal molds for injection molding of high and low density polyethylene, polystyrene, polypropylene and ABS plastic. Very good accuracy is achieved for about 200 moldings. Long cycle times (~5 min) are required to allow the molding to cool enough that it will not stick to the SLA core. In another variation, cores are made from thin SLA shells filled with epoxy and aluminum shot. Aluminum’s high conductivity helps the molding cool faster, thus shortening cycle time. The outer surface can also be plated with metal to improve wear resistance. Production runs of 1000e5000 moldings are envisioned to make the process economically viable. LOM Composite: Helysis and the University of Dayton are working to develop ceramic composite materials for LOM. LOM Composite parts would be very strong and durable, and could be used as tooling in a variety of manufacturing processes. Sand Molding: at least two RP techniques can construct sand molds directly from CAD data. DTM sells sandlike material that can be sintered into molds. Soligen http://www.3dprinting.com/uses 3DP to produce ceramic molds and cores for investment casting, (Direct Shell Production Casting).

5.2. Rapid Manufacturing A natural extension of RP is rapid manufacturing (RM), the automated production of market products directly from CAD data [89]. Currently only a few final products are produced by RP machines, but the number will increase as metals and other materials become more widely available. RM will never completely replace other manufacturing techniques, especially in large production runs where mass-production is more economical. For short production runs, however, RM is much cheaper, since it does not require tooling. It is also ideal for producing custom parts tailored to the user’s exact specifications. A University of Delaware research project uses a digitized 3D model of a person’s head to construct a custom-fitted helmet. NASA is experimenting using RP machines to produce spacesuit gloves fitted to each astronaut’s hands. From tailored golf club grips to custom dinnerware, the possibilities are endless.

521

The other major use of RM is for products that simply cannot be made by subtractive (machining, grinding) or compressive (forging, etc.) processes. This includes objects with complex features, internal voids, and layered structures.

5.3. Future Developments RP is starting to change the way companies design and build products. There are several developments that improve manufacturing. One such improvement is increased speed [90]. RP machines are still slow by some standards. By using faster computers, more complex control systems, and improved materials, RP manufacturers are dramatically reducing build time. For example, Stratasys (since January 1998) has introduced FDM Quantum machine, which can produce ABS plastic models 2.5e5 times faster than previous FDM machines. Continued reductions in build time make it possible rapid manufacturing economical for a wider variety of products. Another future development is improved accuracy and surface finish. Today’s commercially available machines are accurate to ~0.08 mm in the xey plane, but less in the z (vertical) direction. Improvements in laser optics and motor control should increase accuracy in all three directions. In addition, RP companies are developing new polymers with better mechanical properties and. The introduction of nonpolymeric materials, including metals, ceramics, and composites, represents another much anticipated development. These materials allow RP users to produce functional parts. Today’s plastic prototypes work well for visualization and fit tests, but they are often too weak for function testing. More rugged materials would yield prototypes that could be subjected to actual service conditions. In addition, metal and composite materials will greatly expand the range of products that can be made by rapid manufacturing. Many RP companies and research labs are developing new materials. For example, the University of Dayton is working with Helisys to produce ceramic matrix composites by LOM. An Advanced Research Projects Agency/ Office of Naval Research sponsored project is investigating ways to make ceramics using FDM. Sandia/Stanford’s LENS system can create solid metal parts. Another important development is increased size capacity. Currently most RP machines are limited to objects 0.125 m3 or less. Larger parts must be built in sections and joined by hand. To remedy this situation, several “large prototype” techniques are in the works. The most fully developed is Topographic Shell Fabrication from Formus in San Jose, CA. In this process, a temporary mold is built from layers of silica powder (high quality sand) bound together with paraffin wax. The mold is then used to produce fiberglass, epoxy, foam, or concrete models up to 3.3 m  2 m  1.2 m in size.

522

At the University of Utah, a team is developing systems to cut intricate shapes into 1.2 m  2.4 m sections of foam or paper. Researchers at Penn State’s Applied Research Lab are aiming even higher: to directly build large metal parts such as tank turrets using robotically guided lasers. Group leader Henry Watson states that product size is limited only by the size of the robot holding the laser. All the above improvements are the key to the development of the RP industry worldwide One future application is Distance Manufacturing on Demand, a combination of RP and the Internet that will allow designers to remotely submit designs for immediate manufacture. Finally, the rise of RP has spurred progress in traditional subtractive methods as well. Advances in computerized path planning, numeric control, and machine dynamics have increased the speed and accuracy of machining. Modern CNC machining centers can have spindle speeds of up to 100,000 rpm, with correspondingly fast feed rates. Such high material removal rates translate into short build times. For certain applications, particularly metals, machining will continue to be a useful manufacturing process.

REFERENCES [1] Luciano L, Lou S, Grant M. Electronic design automation for integrated circuits handbook; ISBN 9780849330964; 2006. [2] U.S. Congress, Office of Technology Assessment. Computerized manufacturing automation. DIANE Publishing; 1984. p. 48. ISBN 9781428923645. [3] Meguid SA. Integrated computer-aided design of mechanical systems. Springer; 1987. p. 7. ISBN 9781851660216. [4] Chua CK, Leong KF, Lim CS. Rapid prototyping e principles and applications. 2nd ed. World Scientific Publishing Co. Pte. Ltd; 2005. p. 7e10. [5] Voelcker Herbert. A current perspective on tolerancing and metrology. Manuf Rev 1993;6(4):258e68. [6] Deckard C. “Method and apparatus for producing parts by selective sintering”. U.S. Patent 4,863,538, filed October 17, 1986, published September 5, 1989. [7] Chua CK, Leong KF, Lim CS. Rapid prototyping e principles and applications. 2nd ed. World Scientific Publishing Co. Pte.Ltd; 2005. p. 15. [8] Wohlers Report. State of the industry annual worldwide progress report on additive manufacturing. 2009. Wohlers Associates, http:// www.wohlersassociates.com/; 2009. ISBN 0-9754429-5-3. [9] Pham DT, Dimov SS. Rapid manufacturing. Springer-Verlag, ISBN 1-85233-360-X; 2001. p. 6. [10] Delhote N, Bila S, Baillargeat D, Chartier T, Verdeyme S. Advanced design and fabrication of microwave components based on shape optimization and 3D ceramic stereolithography process (Advances in ceramics-synthesis and characterization, processing and specific applications), Intech; August 2011. [11] Bruce C. Carsten’s corner “Power conversion and intelligent motion”; November 1989. p. 38.

Handbook of Advanced Ceramics

[12] Houde S, Hill C. What do prototypes prototype? In: Helander M, Landauer T, Prabhu P, editors. Handbook of human-computer interaction. 2nd ed. Amsterdam: Elsevier Science B. V.; 1997. [13] Wagner A. Prototyping: a day in the life of an interface designer. In: Laurel B, editor. The art of human computer interface design. Reading, MA: Addison-Wesley; 1990. p. 79e84. [14] Wong YY. Rough and ready prototypes: lessons from graphic design. In: Proceedings of CHI ’92 posters and short talks. ACM Press; May 1992. p. 83e4. [15] Yang S, Leong KF, Du Z, Chua CK. Tissue Engineering. The design of scaffolds for use in tissue engineering. Part II. Rapid prototyping techniques 2002;8(1):1e11, Mary Ann Liebert Inc pub. http://dx.doi.org/10.1089/107632702753503009. [16] Paulo Bartolo. Stereolithography: materials, processes and applications. Springer; 2011. p. 8. [17] Ang BY, Chua CK, Du ZH. Study of trapped material in rapid prototyping parts. Int J Adv Manuf Technol 2000:120e30, http://dx.doi.org/10.1007/s001700050017. [18] Jain Prashant K, Senthilkumaran K, Pandey Pulak, M, Rao PVM. Advances in materials for powder based rapid prototyping. In: Proceeding of international conference on recent advances in materials and processing. Coimbatore, India: PSG-tech.; Dec 15e16 2006. [19] Rounds R. Prototype disadvantages and rapid prototyping disadvantages. Computers and Technology: Software; 2008. [20] Chua CK, Chou SM. In: Computer aided and integrated manufacturing systems, vol. 3; 2003. p. 165e85. eISBN: 9789812796790. [21] Leong KF, Leah CM, Chua CK. Solid freeform fabrication of three-dimensional scaffoldsfor engineering replacement tissues and organs. Biomaterials 2003;24:2363e78. [22] Knitter R, Bauer W. Ceramic microfabrication by rapid prototyping process chains, vol. 28; 2003. parts 1&2308. [23] Halloran JW, Tomeckova V, Gentry S, Das S, Cilino P, Yuan P, et al. Photopolymerization of powder suspensions for shaping ceramics. J Eur Ceram Soc November 2011;31(14):2613e9. [24] Griffith ML, Halloran JW. free form fabrication of ceramics via stereolithography. J Am Ceram Soc 1996;79(10):2601e8. [25] Jardini ALM, Maciel R, Scarparo MAF, Andrade SR, Moura LFM. Improvement of the spatial resolution of prototypes using infrared laser stereolithography on thermosensitive resins. J Mater Process Technol 20 February 2006;172(1):104e9. [26] Chua CK, Leong KF, Lim CS. Rapid prototyping e principles and applications. 2nd ed. World Scientific Publishing Co. Pte. Ltd; 2005. 25. [27] Gebhardt A. Rapid prototyping. Munich: Carl HanserVerlag; 2003. ISBN: 3446242590. [28] Knitter R, Bauer W. Ceramic microfabrication by rapid prototyping process chain. Sadhana Academy Proceedings in Engineering Sciences February/April 2003;28(1 & 2):307e18. [29] Balc N, Campbell R. From CAD and RP to innovative manufacturing. Brasov, Romania: Computing and Solutions in Manufacturing Engineering; 2004. [30] Bartolo Paulo. Stereolithography: materials, processes and applications. Springer; 2011. p. 6. [31] Galantucci LM, Perocco G, Angelelli G, Lopez C, Introna F, Liuzzi C, et al. Reverse engineering techniques applied to a human skull, for CAD 3D reconstruction and physical replication by rapid prototyping. J Med EngTechnol 2006;30(2):102e11, http://dx.doi. org/10.1080/03091900500131714.

Chapter | 6.5

Rapid Prototyping of Ceramics

[32] Bauer W, Knitter R, Emde A, Bartelt G, Gohring D, Hansjosten E. Replication techniques for ceramic microcomponents with high aspect ratios. Microsys Technol 2002;9(1-2):81e6, http://dx.doi. org/10.1007/s00542-002-0200-z. [33] Marshall Burns. “Automated fabrication”. Prentice Hall. ISBN 978e0131194625. [34] Alpha prototypes: rapid prototyping &stereolithography services. http://www.alphaprototypes.com/how-to-create-stereolithographyfiles.aspx. [35] Yeong WY, Chua CK, Leong KF, Chandrasekaran M. Rapid prototyping in tissue engineering: challenges and potential. Trends Biotechnol 2004;22(12):643e52. [36] Karunakaran K, Vivekananda Shanmuganathan P, Koch KU, RothKock S. Direct rapid prototyping of tools solid freeform fabrication symposium 1998. Austin: Univ. of Texas; 1998. 105e112. [37] Berry E, Brown JM. Preliminary experience with medical applications of rapid prototyping by selective laser sintering. Med Eng Phys 1997;19(1):90e6. [38] Kruth JP, Wang X, Laoui T, Froyen L. Lasers and materials in selective laser sintering, vol. 23, Number 4. Emerald, Group Publishing Ltd.; 2003. pp. 357e371(15). [39] Gibson I, Shi D. Material properties and fabrication parameters in selective laser sintering process. Rap Prototyp J 1997;3(4): 129e36. [40] Williams JM, Adewunmi A. Bone tissue engineering using polycaprolactone scaffolds fabricated via selective laser sintering. Biomaterials 2005;26(23):4817e27. [41] Sallador V. Selective laser sintering e its description and benefits. EzineArticle, http://ezinearticles.com/?Selective-Laser-Sinteringd Its-Description-and-Benefits&id¼4628681; 2010. [42] Kruth JP, Mercelis P, Van Vaerenbergh J, Froyen L, Rombouts M. Binding mechanisms in selective laser sintering and selective laser melting. Rap Prototyp J 2005;11(1):26e36. [43] Hopkinson N, Hague JM. Rapid manufacturing: an industrial revolution for the digital age. Wiley Online Library, http://dx.doi. org/10.1002/0470033991; 2006. [44] Xue Y, Gu P. A review of rapid prototyping technologies and principles. Comput-Aided Des 1996;28(4):307e18. [45] Chen Z, Li D, Lu B, Tang Y, Sun M, Xu S. Fabrication of osteostructure analogous scaffolds via fused deposition modeling. Scr Mater 2005;52(2):157e61. [46] Safari Ahmad, et al. Powder processing, rheology, and mechanical properties of feedstock for fused deposition of Si3N4 ceramics. J Am Ceram Soc 2000;83(7):1663e9. [47] Kalita S, Bose S. Development of controlled porosity polymerceramic composite scaffolds via fused deposition modeling. Mater Sci Eng C 2003;23(5):611e20. [48] Crump SS, Comb JW, Priedeman WR, Zinniel RL. Process of support removal for fused deposition modeling. Patent number 5503785, issue date 1996. [49] Winder J, Richard B. Medical rapid prototyping technologies: state of the art and current limitations for application in oral and maxillofacial surgery. J Oral Maxillofac Surg 2005;63(7): 1006e15. [50] Lous GM, Cornejo IA, McNulty TF, Safari A, Danforth SC. Fabrication of piezoelectric ceramic/polymer composite transducers using Fused Deposition of ceramics. J Am Ceram Soc 2000;83(1):124e8.

523

[51]

[52]

[53]

[54]

[55] [56]

[57]

[58] [59]

[60]

[61]

[62]

[63]

[64]

[65]

[66] [67] [68]

[69]

Cima M, Sachs E, Fan T, Bredt JF, Michaels SP, Kanuja S, et al. “Three-dimensional printing techniques”. Patent number: 5204055, issue date 1993. Sachs E, Cima M. Three dimensional printing: rapid tooling and prototypes directly from a CAD model. CIRP Annals e Manuf Technol 1990;39(1):201e4. Borland SW, Giordano RA, Cima LG, Sachs EM, Cima MJ. Solid freeform fabrication of drug delivery devices. J of Controlled Release 1996:77e87. Dimitrov D, Schreve K, De Beer N. Advances in three dimensional printing e state of the art and future perspectives. Rap Prototyp J 2006;12(3):136e47. Kastra WE. Oral dosage forms fabricated by three dimensional printing. J Control Release 2000;66(1):1e9. Seitz H, Rieder W, Irsen S, Leukers B, Tille C. Three-dimensional printing of porous ceramic scaffolds for bone tissue engineering. J Biomed Mater Res B Appl Biomater August 2005; 74B(2):782e8. Anderson TC, Bredt JF, Russell DB. Three dimensional printing material system and method. Patent number: 6610429, Filing date 10 Apr 2001. Issue date: 26 Aug 2003. Yoshimura K, Kishimoto M, Suemune T. InkJet printing technology. OKI Tech Rev 1998;64:41e4. Lejeune M, Dossou-Yovo C, Chartier T, Noguera R. Ink-jet printing of ceramic micro-pillar arrays. J Eur Ceram Soc 2009; 29:905e11. Calvert P. Inkjet printing for materials and devices. Chem Mater 2001;13(10):3299e305. Department of Materials Science and Engineering, University of Arizona, Tucson Arizona 85721. Sirringhaus H, Kawase T, Friend RH. High-resolution inkjet printing of all-polymer transistor circuits. Science 2000;290(5499): 2123e6. Charles W. Hull. “Method for production of three-dimensional bodies by stereolithography. Patent number: 5174943, Filing date Aug 23, 1991. Issue date: Dec 29, 1992. Halloran John W. Tomeckova Vladislava. Predictive models for the photopolymerization of ceramic suspensions. J Eur Ceram Soc October 2010;30(14):2833e40. Chartier T, Corbel S, Hinczewski C. Ceramic suspensions suitable for stereolithography. J Eur Ceram Soc 1998;18(6): 583e90. Jacobs P. In: Rapid prototyping and manufacturing: fundamentals of stereolithography. 1st ed. Society of Manufacturing Engineers; 1992. ISBN: 0872634256. Paulo Bartolo. Stereolithography: materials, Springer. processes and applications; 2011. p. 81. Jacobs PF. Fundamentals of stereolithography. 3D Systems Inc 1992. Badev A, Abouliatim Y, Chartier T, Lebaudy P, Lecamp L, Chaput C, et al. Photopolymerization kinetics of a polyether acrylate in the presence of ceramic fillers used in stereolithography. J Photochem Photobiol A Chem 5 July 2011;222(1): 117e22. Chartier T, Badev A, Abouliatim Y, Lebaudy P, Lecamp L. Stereolithography process: influence of the rheology of silica suspensions and of the medium on polymerization kinetics e cured depth and width. J Eur Ceram 2012;32(8):1625e34.

524

[70] Melchels FP, Feijen J, Grijpma DW. A review of stereolithography and its applications in biomedical engineering. Biomaterials 2010; 31(24):6121e30. [71] Liu B, Gong X, Chappell WJ. Applications of layer-by-layer polymer stereolithography for three-dimensional high-frequency components. Microwave Theory Tech, IEEE Trans 2004;52(11):2567e75. [72] Pham DT, Gault RS. A comparison of rapid prototyping technologies. Int J Mach Tool Manu 1999;38(10-11):1257e87. [73] Santler G, Ka¨rcher H, Gaggl A, Kern R. Stereolithography models vs milled models. Production, indications, accuracy. Mund Kiefer Gesichtschir 1998 March;2(2):91e5. [74] Bertsch A, Jiguet S. Microfabrication of ceramic components by microstereolithography. J Micromech Microeng 2004;14:197, http://dx.doi.org/http://dx.doi.org/10.1088/0960-1317/14/2/005. [75] Sun C, Fang N. Projection micro-stereolithography using digital micro-mirror dynamic mask. Sens Actuators A Phys 2005;121(1): 113e20. [76] Paulo Bartolo. Stereolithography: materials. Springer: processes and applications; 2011. 107. [77] Bertsch A, Bernhard P. Microstereolithography: concepts and applications emerging technologies and factory automation. In: Proceedings, 8th IEEE International Conference, vol. 2; 2001. p. 289e98. [78] Chartier T, Duterte C, Delhote N, Baillargeat D, Verdeyme S, Delage C, et al. Fabrication of millimeter wave components via ceramic stereo- and microstereo-lithographie processes. J Am Ceram Soc 2008;91(8):2469e74.

Handbook of Advanced Ceramics

[79] Maruo S, Ikuta K. Movable microstructures made by two-photon three-dimensional microfabrication. Micromechatronics Hum Sci 1999:173e8. [80] Maruo S, Ikuta K. Light-driven MEMS made by high-speed twophoton microstereolithography. Micro Electro Mech Sys 2001:594e7. [81] Daniel Rypl, http://mech.fsv.cvut.cz/~dr/papers/Lisbon04/node2. html; 2005-11-05. [82] Armstrong MA. Groups and symmetry. New York: Springer-Verlag; 1988. ISBN 978e3540966753. [83] IGES 5.3 (ANSI-1996)”. US Product Data Association; 1996-09-23. [84] Hewlett-Packard Company. HP/GL-2 and HP RTL reference guide: a handbook for program developpers. 2nd ed; 1996. [85] Paul Bourke. Alphabetical list of 3D API specifications and data formats, http://paulbourke.net/dataformats/slc/ [86] Jeffrey Mogul. The leaf file access protocol. Department of Computer Science, Stanford University; 1986. [87] Berce P, Chezan H, Balc N. ESAFORM. The application of rapid prototyping technologies for manufacturing the custom implants; 2005. Conference, Cluj-Napoca, Romania. [88] Ed Tackett. “Rapid tooling”, Saddleback College Advanced Technology Center. [89] Wang Carol Y. Rapid manufacturing. CSA, Technology Research Database; 2002. [90] Dickens PM. Research developments in rapid prototyping Proceedings of the Institution of Mechanical Engineers, Part B. J Eng Manuf August 1995;209(4):261e6.

Chapter 7.1

Biomorphous Ceramics from Lignocellulosic Preforms Peter Greil, Tobias Fey and Cordt Zollfrank Department of Materials Science and Engineering (Glass and Ceramics), University of Erlangen-Nuernberg, Martensstrasse 5, 91058 Erlangen, Germany

Chapter Outline 1. 2. 3. 4.

Introduction Lignocellulosic Cellular Preforms Processing Microstructure and Mechanical Properties

527 528 531 541

1. INTRODUCTION Inspired by biological concepts of design, microstructure, and property optimization, new biomorphous inorganic materials with advanced functions and structures have attained increasing interest in materials science at the frontier between biology and chemistry [1,2]. The driving force to develop novel approaches for producing synthetic materials by use of biological processes and principles is for a great need for materials of enhanced performance and reliability, which provide better efficiency, specialization, and optimization of their properties [3,4]. Inorganic materials synthesized by biologically controlled growth processes are hierarchically organized microcomposite materials, which are characterized by an inorganic biomineral phase and a biopolymer phase. Prominent biominerals include the following: carbonates in mollusc shells, nacre, enamel, and corals; phosphates in bone and teeth; iron oxides in magnetotactic bacteria; or silica in diatoms and sponges [5,6]. Almost 70 different mineral types are known to be formed by living organisms. A bioorganic phase including various proteins and polysaccharides is a key element needed to control the directed growth of the biomineral in time and space (hierarchical and cellular structure anatomy) and morphogenesis (crystal habitus). Since biogenic inorganic composites materials are synthesized at mild conditions (ambient temperature), their limited temperature stability only allows applications at low temperatures due to the included bioorganic phase.

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00029-0 Copyright Ó 2013 Elsevier Inc. All rights reserved.

5. Applications 6. Conclusions References

548 551 552

Advanced functional and engineering materials, however, are often subjected to harsh environments such as elevated temperature, corrosion, and high mechanical loading. Thus, even small fractions of organic constituents that often form thin interface layers between the inorganic particles may limit the application of these materials at harsh environmental and application conditions. Furthermore, the low synthesis temperatures and often low concentrations of reaction solutions result in very slow (mm/h) growth rates of the biomineral so that an economical manufacturing process of bulk components could not be realized yet. Therefore, only low-dimensional products such as magnetic powders, size- and shape-specific silica for chromatography, and SiC whiskers from rice husks for ceramic cutting tools were fabricated using biological preform materials [7,8]. A significant increase of synthesis rate can be achieved when cellular biological structures with open porosity are used as a microstructural template for physicochemical infiltration and reaction techniques [9]. Thus, conversion of cellular tissue of naturally grown plants into inorganic materials (biomorphous ceramics) has gained an increasing interest in recent years [10e20]. Basic principles of conversion of plant-derived preforms (Lignocellulosics) into ceramic materials mimicking the initial template structure at various hierarchical micro- and macrostructural levels were discussed for example in [14]. Carbon derived from the natural cellular preform (biocarbon) may also serve as a template for a variety of infiltration and reaction processes. A comprehensive survey of the transformation

527

528

Handbook of Advanced Ceramics

TABLE 1 Examples of Biomorphous Ceramics Derived from Natural and Preprocessed Plant Preforms Template

Plant

Material

ceramics processing, microstructure development, and property optimization will be given.

Ref.

Fibers Sisal

Al2O3, TiO2

[23]

Cotton

SiC

[24]

Pine

SiC, Al2O3, SrAl2O4: Eu

[9,25e27]

Maple

SiC

[28]

Walnut

SiC/Si

[29]

Beech

SiC, Al/SiC

[30]

Oak

SiC

[31]

Balsa, Cypress

SiC

[32]

Rattan

Si/SiC, Ca5(PO4)3 OH

[26,33]

Sponge

Ca5(PO4)3OH

[34]

Wood fibers

SieCu/SiC

[35]

Wood fiber boards

SiC/Si/C

[36]

Fiber papers

Al2O3, SiC

[37]

Wood

Other plants

Wood products

of Lignocellulosics into biocarbon can be found in [21,22]. Furthermore, plant tissue-derived products such as fiber paper and fiber boards may serve as preforms with a more uniform microstructure compared with naturally grown tissue, which may vary with seasonal and local growth conditions significantly (e.g. annual growth ring patterns, reaction tissue). A variety of biomorphous ceramics were fabricated from Lignocellulosic templates including fibers, wood tissue, and wood pulp paper and fiber board preforms, Table 1. Mimicking the cellular design of natural tissue anatomy by material compositions relevant for functional and engineering applications (e.g. multicomponent oxides and nonoxides) offers a highly attractive approach for creating a novel class of biomorphous materials. The mechanical, thermal, or optical properties of the biomorphous ceramics are governed by the hierarchical cellular structure and the phase composition, which generally result in highly anisotropic materials. These materials have a high potential of achieving an excellent density-to-strength relationship and outstanding thermomechanical performance. In the following, a review on the current status of biomorphous

2. LIGNOCELLULOSIC CELLULAR PREFORMS Lignocellulosics form the class of organic matter (biomass) produced by land-growing plants in the form of trees, shrubs, and agricultural crops. It is the essential carbon sink of the planet, which is formed by catalytic conversion of carbon dioxide to an organic mass mainly consisting of the elements CeOeH (eN, S, P) [14,38]. Lignocellulosic biomass is mainly composed of polysaccharides (cellulose, polyoses) and lignin biopolymers. A mature vascular plant contains several differentiated cell types grouped together in tissues. The tissues of a plant are organized into three tissue systems: the dermal, the ground, and the vascular tissue system, respectively. Phloem and xylem are the major cellular components of the vascular tissue system. The xylem conducts water and dissolved minerals from the roots to all the other parts of the plant. Two types of waterconducting cells of xylem are usually discerned: tracheids and vessel elements. Tracheids are individual cells tapered at each end, so the tapered end of one cell overlaps that of the adjacent cell. They have thick, lignified walls, which are perforated so that water can flow from one tracheid to the next. Vessels are thick-walled tubes that can extend vertically through several meters of xylem tissue. Their diameter may be as large as 0.7 mm. The inherent vascular transportation system provides a facile access to infiltration fluids for transforming biological template structures into inorganic materials. Wood is a natural composite with partially crystalline cellulose fibrils embedded in an amorphous matrix of polyoses and lignin as the major biomolecular constituents. It exhibits a unique hierarchical cellular architecture with a remarkable combination of strength, stiffness, and toughness [39]. Wood cell walls provide skeletal support (mechanical stability), but they also play a dominant role in cell growth and morphogenesis, in cell recognition and signaling, in digestibility, and in herbivore nutrition. A single longitudinal tracheidal cell exhibits a layered wall structure, a thin primary wall (P) and a thicker secondary wall composed of sublayers (S1, S2), which vary in cellulose microfibril orientation and which play a key role for mechanical behavior of cellular tissue, Figure 1. Adjacent tracheids are joined together by a highly lignified layer (middle lamella, ML). Because of seasonal variations of growth conditions, growth ring patterns develop with low density/high porosity in the earlywood regions and high density/low porosity in the latewood regions. Cellulose (C6H10O5)n is the main load-bearing component in fibers. It is a linear polymer composed of

Chapter | 7.1

Biomorphous Ceramics from Lignocellulosic Preforms

529

FIGURE 1 Hierarchical organization of Lignocellulosic cellular preforms.

glucose units joined together by b-1,4 glycosidic bonds. Cellulose is able to attain a high stiffness and tensile strength due to intra- and intermolecular hydrogen bonding. The degree of polymerization n in a cellulose chain of a native softwood fiber is above 10, 000 which is decreased after chemical pulping to 500e2000. The cellulose crystal is distinguished by a Young’s modulus of 145 GPa and a theoretical tensile strength estimated at 7500 MPa [40]. The orientation angle of the cellulose microfibrils with respect to the fiber cylinder axis therefore plays a crucial role in determining the mechanical properties of fibers.

2.1. Wood Preforms In angiosperms (hardwoods), most of the water is conducted in the xylem vessels. Their walls are thickened with secondary deposits of cellulose and are usually further strengthened by impregnation with lignin. The secondary walls of the xylem vessels are deposited in spirals and rings and are usually perforated by pits. Xylem also contains tracheids. In ring-porous wood, such as oak and basswood, the spring vessels (earlywood) are much larger and more porous than the smaller, summer tracheids (latewood). This difference in cell size and density produces the conspicuous, concentric annual growth rings in these woods. Softwoods are gymnosperms, such as pine, spruce, and fir. Gymnosperms generally do not have vessels but produce only tracheids in their xylem. The tracheidal tissue of gymnosperms shows a lower variation of cell pore sizes and hence a more uniform cellular microstructure compared

with angiosperms. Table 2 summarizes some of wood templates used for processing of biomorphous ceramics. The shape of vascular cells may vary from spherical (for example Meranti, Shorea spp.) over oval (walnut, Juglans regia.) to rectangular (pine e Pinus sylvestris), and the thickness of cell walls typically ranges from 4 mm (Okoume´, Aucoumea klaineana) to 18 mm (Ebony, Diospyros spp.). The density can be as low as 0.05 g/cm3 for balsa and as high as 1.03 g/cm3 for ebony. While small pores with diameters 0.028, residual Si will always remain in pores larger than approximately 0.1e0.3 mm (for ds z 5e10 mm), resulting in a two-phase SiCeSi composite material. Pores larger than rmax z 0.4 mm, however, will remain unfilled since gravity exceeds capillary suction. Though the complex three-dimensional cellular structure of biocarbon is characterized by a high degree of heterogeneity (variation of cell shape elliptical vs. rectangular, cell size, cell orientation longitudinal vs. radial, interconnectivity between cells), experimental observations generally confirm that pores below 1 mm could not be deeply infiltrated and pores with a diameter exceeding 100e300 mm remained free of residual Si. [9]. Depending on the template porosity and pore size distribution, the biomorphous SiC ceramics processed by liquid Si infiltration contained a residual Si content ranging from 20 wt.%

536

Handbook of Advanced Ceramics

FIGURE 8 Infiltration time vs. pore radius for liquid-Si infiltration into a porous biocarbon template assuming parallel pore channels calculated from Eqn. (3).

FIGURE 9 Microstructures of liquid Si-infiltrated biomorphous SiC, derived from pine (Pinus sylvestris) (a) and from oak (Quercus) (b) (bright Si, light gray: SiC, dark gray: C, black: pores).

(maple) to 70 wt.% (balsa) [70]. Figure 9 shows two typical microstructures of biomorphous b-SiC formed by liquid Si infiltration process of softwood (pine) and hardwood (oak), which display pronounced transition from early- to latewood regions. Figure 10 shows the variation of pore size distribution of pine and oak from the native template, the biocarbon template pyrolyzed at 800  C and 1600  C, and for Si-melt infiltrated SiC. At low pyrolysis temperature of 800  C the biocarbon template retains the pore structure of the native template, increasing the temperature (1600  C) leads to close the small pore fraction. All pores smaller than approximately 1 mm disappeared in the Si-infiltrated SiC reaction product and only pores larger than approximately 10 mm remain.

Instead of liquid infiltration, vapor-phase infiltration was applied to form single-phase SiC ceramics [25,29,71]. The gas transport fluxes, featuring ordinary diffusion, viscous flow (from Darcy’s law), and Knudsen diffusion, are evaluated using the DustyeGas Model [72]. Two parameters have been shown to pilot the vapor infiltration and reaction process, namely the Thiele modulus, F, rffiffiffiffiffiffiffiffiffiffi kSv (4) F ¼ L εDeff and the characteristic infiltration time, tinf, tinf ¼ ½UC0 Sv kŠ

1

(5)

Chapter | 7.1

FIGURE 10

Biomorphous Ceramics from Lignocellulosic Preforms

537

Pore size distributions determined by Hg-porosimetry and SEM image analyzes (axial cross-section) of pine and oak templates.

L is the characteristic diffusion length, e.g. half width of component thickness, Sv is the internal surface area, ε is the porosity, Deff is the effective diffusion coefficient, U is the deposit molar volume, C0 is a reference precursor concentration, and k(T) is a heterogeneous rate constant and follows an Arrhenius-like law. For low values of F 800  C; inert amosphere Cellulose þ ðReSieX1:5 Þn / C þ SieOeCðeNÞ þ Gas with R ¼ H; CH3 ; C6 H5 ; . The SieH functional group is able to react with OH groups of cellulose and lignin biopolymers to form SieO ether bonds [43]. Annealing in air above 600  C oxidizes the biocarbon residue leaving a highly porous ceramic residue. Figure 15 shows the cellular microstructures of various biomorphous oxide ceramics derived from pine wood. Pine wood was vacuum infiltrated with different low-viscous alumina-, titania- , and zirconia-sol. Subsequent pyrolysis in inert atmosphere at 800  C and annealing in air up to 1550  C resulted in the formation of porous, microcellular a-Al2O3, TiO2 (rutile), and c-ZrO2 (stabilized by 8 mol% Y2O3), respectively [90]. The initial cellular anatomy was reproduced in the ceramic products. Mean particle size of the Al2O3-, TiO2- or ZrO2 grains is about 3e5 mm. The skeleton density of biomorphous Al2O3- and TiO2 ceramics are very close to the theoretical density of Al2O3 (3.97 g/ cm3) and TiO2 (4.26 g/cm3), respectively, indicating only a small amount of residual strut porosity. Ota et al. [91] produced biomorphous TiO2 ceramics by infiltration of wood species with titanium isopropoxide.

Chapter | 7.1

Biomorphous Ceramics from Lignocellulosic Preforms

541

FIGURE 15 Biomorphous oxide ceramics derived from pine wood after sintering at different temperatures in air: Al2O3 (1550  C/1 h), TiO2 (1200  C/ 1 h), ZrO2 (1500  C/1 h) (SEM micrographs of cross sections perpendicular to stem axis).

Shin et al. [12] used a surfactant-templated solegel process. Rattan palm preforms were converted to biomorphous Al2O3, ZrO2, and 3Al2O32SiO2 ceramics by a solegel process with metal alkoxides, metal oxide chlorides, and silica nanopowders [90]. Biomorphous ZrO2 ceramics were manufactured by sol infiltration of wood [92e94]. Hydrothermal synthesis was applied to produce biomorphous NiO and ZnO from pine and fir wood, respectively [93e95]. Their porous structures were found to be hierarchical from 1 up to 25 mm (in micrometer scale) and from 2 to 60 nm (in nanometer scale). Porous silica (SiO2) ceramic with a wood-like structure was prepared by wet impregnation tetraethyl orthosilicate into biological template that was derived from linden wood (Tilia amurensis) [96]. After repeated pressure impregnations, the subsequent annealing in air atmosphere at 800  C resulted in oxidative removal of the template and consolidation of the oxide layers. It was found that the bioorganic structure

was converted into oxide ceramics (SiO2). At low temperature (800  C), pore radius varied between 2 and 10 nm, indicating that the samples were mostly mesoporous. Samples treated at higher temperature (1300  C) lost the mesoporous character; however, they were still porous, having the microstructural features of the biological preform.

4. MICROSTRUCTURE AND MECHANICAL PROPERTIES The mechanical properties of biomorphous ceramics are dominated by the hierarchical and ansiotropic structure of biological tissue. The representative volume element over which the average of mechanical properties are representative of the whole can be quite large (103 mm3 and even equal to the component size) compared with conventional

542

FIGURE 16

Handbook of Advanced Ceramics

Hierarchical microstructure of wood tissue. Expressions for the fractional density r*of left- and middle-cell shapes were taken from [100].

engineering materials (down to 10 12 mm3) [97]. Generally, averaging techniques from solid-mechanics theories (rule of mixtures for composites, laminate theories) are applied, which, however, is associated with the loss of some information of the subsystem in the process. This is particularly relevant in relation to mechanical events occurring at local levels (cellular level), such as damage initiation and fracture propagation, for example, as opposed to global events such as elastic behavior or structural instabilities of the kind associated with buckling and fracture [98]. The two main characteristics relevant for mechanical property validation are the density distribution of the biomorphous ceramic manufactured from the wood templates and the level of microstructural anisotropy between directions parallel (axial) and perpendicular (radial and tangential) to the growth direction that will correlate with anisotropy in the mechanical properties of biomorphous ceramics [99]. Furthermore, the cell shape and dimensions may vary significantly with growth conditions, Figure 16. Periodical undulations of density generally occur in wood-derived ceramics due to seasonal variations of growth conditions. Figure 17 shows the radial distribution of fractional density r* as determined experimentally by

scanning electron microscope (SEM) image analysis. While in the earlywood region r* < 0.4, a pronounced increase of r* up to 0.9 was found to occur in the latewood regions. As a consequence, nonlinear transitions of material properties (Young’s modulus, toughness, and strength) are expected to affect the fracture behavior in radial direction [98,99,101]. Local variation of fracture toughness is expected to occur when the cell size, strut thickness, and cell morphology are changed near the growth ring patterns. Figure 18 shows a typical fracture surface perpendicular to the longitudinal (axial) direction (L) of a Si-infiltrated SiC, derived from beech with longitudinal tensile loading.

4.1. Mechanical Properties of Models Young’s modulus as well as strength models of porous ceramic materials available in literature mainly refer to two approaches: (1) Models for low-porosity materials (70%) where the mechanical behavior is mostly related with the bending and brittle failure of the cell walls, and E* is expressed by a power law of the fractional density r*(¼rcell/rstrut) [100]: E ¼

FIGURE 17 Radial undulations of fractional density at the late-/earlywood transitions (growth ring patterns) of biomorphous SiC derived from pine wood.

modulus, strength or toughness, respectively, by an exponential scaling law of fractional density r* [102]:

E ¼

Ecell ¼ exp½ Estrut

bð1

rފ

(9)

where b was shown to be primarily a function of pore shape and alignment with respect to the stress axis but is independent on pore size, as long as that size is small with respect to the dimensions of the body. b was found to equal

Ecell ¼ Cðfr Þn þ C0 ð1 Estrut

fÞr

(10)

where f represents the fraction of open porosity (interconnected) to closed porosity and defines the range of validity for the open cell and wood modification of the cellular model at f / 1. C and C0 are constants which depend on the loading direction with respect to the cellular pore orientation. Depending on the direction of loading, n may attain characteristic values of 1e3, Table 3.

4.2. Strength Brittle cellular ceramics of high porosity fail abruptly at a stress that is usually lower than the strength of the bulk material. The propagation of cracks will depend on, among other factors, the direction of loading to the cellular microstructure. In compression, the cells suffer progressive crushing; in tension, the cellular structure fails by fast brittle fracture. As with any brittle solid, fracture in tension is controlled by the largest defect (a crack, notch, pore, or cluster of damaged cells). Experimental values of FIGURE 18 Typical fracture surface of SieOeC derived from beech (MS-PMS infiltration) showing a nonplanar fracture propagation, triggered by periodical undulations of density and cell structure. (For color version of this figure, the reader is referred to the online version of this book.)

544

Handbook of Advanced Ceramics

TABLE 3 Power Law Coefficients n of the Fractional Density of Selected Mechanical Properties of Cellular Materials According to [100] Direction of loading

Out-of-plane (axial)

In-plane (radial, tangential)

Young’s modulus

n¼1

n¼3

Strength

n¼1

n¼2

Toughness

n ¼ 1.5

n ¼ 1.5

strength of biomorphous ceramics were measured by four-point-bending (uniaxial loading) with longitudinal and radial loading directions or by ball-on-ring bending (biaxial loading), Figure 19. Applying central load in the longitudinal direction results in an in-plane biaxial tensilestress state at the bottom of the cylindrical sample surface. Tensile loading in the radial and tangential directions is more sensitive to the cell microstructure, e.g. geometry, size, and packing arrangement, compared with the “strong” longitudinal cell-elongation direction. Generally, theoretical considerations predict that the in-plane strength, in every instance, is lower than that for out-of-plane loading because bending of the cell walls dominates in the first case and axial deformation in the second direction [100].

Due to the unidirected pore morphology, the highly porous, biomorphous SiC ceramics prepared by Si-vaporphase conversion exhibit an anisotropic mechanical behavior with fracture stress in axial direction being 20 times higher compared with strength in radial direction [10,98,99]. The strength in axial direction mainly depends on the material fraction on the axial cross section similar to the strength of honeycombs [104]. In contrast, strength in radial direction strongly depends on cell morphology features, which include shape, size, and topology of pores and struts. Generally, the in-plane stiffness and strength (stress acting perpendicular to cell elongation) are the lowest, because in-plane loading makes the cell walls bend. The out-of-plane stiffness and strength (stress acting parallel to cell elongation) are much larger, because they require axial extension or compression of the cell walls. With increasing fractional density, cell walls (e.g. the struts) increase in average thickness, but cell shapes can change significantly and hence stiffness and strength increase in a complex manner. Figure 20 summarizes strength data of biomorphous ceramics measured by four-point-bending and ball-on-ring bending. Examinations of the fracture behavior of biomorphous SiC filled with high fractions of Si (>30%) have shown that the experimentally observed behavior can be modeled more realistically as a porous material rather than a cellular material, and that application of cellular model to strength of biomorphous ceramics is limited [66,99]. The minimum

FIGURE 19 Uniaxial and biaxial bending loading for strength measurements of biomorphous ceramics.

Chapter | 7.1

Biomorphous Ceramics from Lignocellulosic Preforms

545

FIGURE 20 Bending strength of biocarbon template and biomorphous SiC, derived from various kinds of wood for axial and radial/tangential loading directions.

solid-area model for close porosity is straightforward, and for connected porosity, it was calculated considering pores with truncated shapes [99]. The relative strength of the porous microstructure is equal to the relative minimum solid area in the direction perpendicular to the stress. The fracture stress of biomorphous Si prepared by liquid Si infiltration and a fractional density r* < 0.5 can be approximated by Eqn (1) with b varying from 3e7 (sstrut z 340 MPa). Vapor-phase- reacted SiC with a low fractional density r* > 0.5 shows a clear conformity with the model for cellular solids Eqn ((2)) with n ¼ 2 found for the strength values measured by biaxial bending. In a material with random distribution of pores, a sudden decrease of strength is observed, when the solid struts lose interconnection. Nature optimizes solid matter distribution by cellular anatomy and strength is maintained even at low fractional density. At low densities, the distribution of wall material available is optimized, and periodically packed structures like the one from pine wood are found. In pine wood, the number of neighbors is reduced to four and the cells have square shape. Thus, as the density of the structures is reduced, the microstructures of the biomorphous SiC materials change to maintain a good strength value (e.g. strut microstructure becomes more and more dominating the strength). In the case of axial compression of longitudinal tubular pores, the type of stacking and shape of pores do not have influence on the theoretical strength (the strength is simply equal to the volume fraction). For compression in radial direction, the models predict a strong dependence with the type of stacking of the tubular pores and a slight dependence with their shape.

4.3. FE-Calculations of Stress and Strain Distribution Periodical undulations of fractional density due to variation of seasonal growth conditions give rise to a complex threedimensional stress and strain distribution upon loading. Furthermore, transitions of cell shape such as for example from quadratic cells to elliptical cells, and cell arrangement may occur at the late-wood to earlywood transition. Numerical finite element methods (FEMs) were applied to simulate the stress distribution in ceramic honeycomb structures [105] and the effect of pore shape on Young’s moduli and Poisson’s ratio [106,107]. An overview about the use of FEM simulations for ceramics is given by Mackerle [108]. The FEM was selected as a vehicle for modeling the stress distribution of a representative volume segment of pine wood-derived SiC subjected under transverse (perpendicular to the axial direction) tensile loading [109]. Using a bottom-up method, the transitionregion of wood-cell system was first divided into substructure I (square pores), and substructure II (elliptical pores), and then the complex structure with the earlywood to late-wood transition areas (combined structure III) was derived by combination of substructure I and substructure II. A 3D mesh was generated using Marc 7.3.2 and Mentat 3.3.2 (MSC.Software). A total number of 4  105-three dimensional 20-node brick elements were applied in the calculations. As boundary conditions, a uniaxial tensile stress of 10 MPa was applied to the surface, and the strut material was represented by the material properties of b-SiC (Young’s modulus 410 GPa;

546

Handbook of Advanced Ceramics

FIGURE 21 Distribution of the normal stress component sxx parallel to the applied tensile stress for the cell morphology transition area in pine-derived SiC, calculated by FEM [98]. For color version of this figure, the reader is referred to the online version of this book.

Possion’s ratio 0,22; density 3,1 g/cm3). Figure 21 shows the distribution of the normal stress component (sxx) parallel to the applied tensile stress for cell morphology transition, where a square cell geometry dominates in the earlywood area and an ellipsoidal cell geometry in the late-wood area. Strut thickness, cell size, and shape parameters were taken from experimental data. Square cells loaded in transverse direction exhibit a significantly lower tensile stress at the same level of fractional density compared with ellipsoidal pores. At a fractional density of r* ¼ 0.35, the maximum tensile stress between the nodes in the square cell pattern is less than 20 MPa, whereas tensile stresses exceeding 35 MPa were calculated at the sharp edges of the ellipsoidal cells. Despite the higher fractional density and strut thickness, the area at the transition from early- to latewood tissue anatomy may give rise to increased probability of crack propagation under the loading conditions considered. Crack advance is likely to be localized on the highly stressed areas, whereas regions of high density should yield higher fracture resistant segments in the loaded structure, Figure 22. Thus, the transition regions at the growth rings might act as crack deflection interfaces, which induce sequential crack propagation and hence a step-like stressestrain behavior. It would be of particular interest therefore to manipulate the growth of natural plants by chemical, physical, or genetic methods in order to tailor tissue structures with desired cellular pore shape and pore arrangement for improved fracture resistance and flaw tolerance.

FIGURE 22 Crack propagation model for biomorphous ceramics dominated by late-/earlywood transitions (density and cell morphology) under longitudinal loading direction and cell-wall crack propagation under in-plane loading direction.

While fracture propagation for longitudinal loading conditions is likely to be dominated by periodical density fluctuations, the strut microstructure of the cell walls was found to play a key role for in-plane loading situation. Figure 23 shows the stress vs. strain curves of vapor phaseprocessed biomorphous SiC with three different strut microstructures: a highly porous, single-layer SiC strut was formed by infiltration with SiO (2C þ SiO(v) / SiC þ CO(v)); a dense, bilayer SiC strut was formed by infiltration with Si (C þ Si(v) / SiC) with a small interface layer of residual carbon often found as ML; and a three-layer strut structure was obtained after infiltration with CH3SiCl3/H2 with an inner core of dense SiC and an outer layer of porous SiC [98]. All specimens exhibit a noncatastrophic failure behavior with a maximum failure strain of 0.6%.

Chapter | 7.1

Biomorphous Ceramics from Lignocellulosic Preforms

547

FIGURE 23 Stress vs. strain diagrams of gas phase-synthesized biomorphous SiC, derived from pine under biaxial in-plane loading conditions [98].

Fracture strength optimization of the strut material ss involves minimization of flaws and porosity within the struts and maximize toughness (fracture-mechanics approach). An increase in strut strength would be expected as a result of reduced probability of finding a critical flaw in a smaller volume of material as predicted by Weibull’s weakest link hypothesis for strength variability. For the case of brittle materials, variation of mean strength, sc, and toughness, Kc, for a given loading direction were found to depend not only on the fractional density but on the Weibull modulus, m, by [100]: sc;1 ¼ sc;2

 2  1 l 2 m r2  m l1 r1 

(11)

and Kc;1 ¼ Kc;2

 1 2  ð2 l 1 2 m r1  l2 r2 

4.4. Hardness and Erosion Vickers hardness (HV) measurements have been performed in SiC samples fabricated from various wood precursors (pine, beech, and eucalyptus) [110]. It was found that hardness was highly dependent on the starting wood structure: a fully dense area corresponding to the wood growth rings and another that was mostly porous (porosity volume fraction between 30% and 50%). Hardness was found to increase with decreasing SiC/Si ratio. Solid-particle erosion studies were conducted on biomorphous SiSiC [111,112]. It was found that material loss occurred by brittle fracture involving the formation of both lateral and radial cracks [113]. Erosion rates in terms of volume (ERV) were predicted from particle impact parameters and target material properties, based on the model of Wiederhorn and Lawn [114]:

1 Þ m

(12)

Thus, the fracture toughness can either increase or decrease with cell size (l) depending on the value of the Weibull modulusdfor m > 4 Kc increases with increasing cell size, while if m < 4 it decreases. Evaluation of the ceramic microstructures indicates a reduction of the cell size l by a factor of three and an increase of strut thickness t by a factor of three at the early- to latewood transition of pine-derived SiC. Thus, for a reasonable Weibull modulus of SiC of 8, a variation of fracture toughness by 30% and of strength by 25% is estimated to occur periodically in the late- to earlywood cellular microstructure transitions.

1:2 ERV fVp2:4 R3:7 p rp

H 0:11 Kc1:3

(13)

where Vp, Rp, rp are the particle velocity, particle radius, and particle density, respectively. Kc is the target toughness and H is the target hardness. From experimental results, it was concluded that radial cracks able to penetrate through thin strut walls of the channel morphology of pores in the biomorphous SiC are likely to contribute to material loss by brittle fracture [99,111]. The erosion-induced surface damage causes a pronounced strength degradation. Most importantly, the erosion rates of the biomorphous SiC ceramics were found to be significantly higher than for the monolithic SiC ceramics.

548

Handbook of Advanced Ceramics

4.5. Creep Behavior

5. APPLICATIONS

Compressive creep experiments for Si-infiltrated SiC derived from a variety of wood template structures including eucalyptus (0.67 g/cm3), beech (0.49 g/cm3), and pine wood (0.33 g/cm3) were conducted at 1300  C for constant pressures ranging from 400e900 MPa [99,115]. Creep strain rate was modeled in a similar way to systems, where creep is due to viscous grain boundary phase. The deformation of the intergranular Si allows limited grain boundary sliding between the SiC grains. For wood-derived SiC (63%) filled with high fractions of Si, an expression of the creep strain rate was derived [115]:  3 dεz Oz h ¼ g εz (14) dt h d

5.1. Wood Tissue

where sz is the applied (compressive stress), h is the effective viscosity of the Si (intergranular layer), g is a shape factor for SiC grains (cubic g ¼ 2.38), h/d is the ratio of initial Si-intergranular layer h width to the grain size d, and εz is the creep strain. At 1300  C and stresses of 250e290 MPa, strain rates smaller than 10 10 1/s were observed for 1% strains. Despite the high Si fraction, the creep resistance of biomorphous SiC compressed in the axial direction was found to be superior to that of conventional Si-infiltrated SiC materials, which was explained by an optimized distribution of SiC in the biomorphous material at least in the axial direction for better mechanical properties compared with conventional materials with a uniform distribution of SiC and Si. Table 4 summarizes selected mechanical properties of a variety of biomorphous ceramics derived from different wood templates.

TABLE 4 Mechanical Properties (Longitudinal Loading Direction) of Biomorphous SiC Ceramics Processed by Si-melt Infiltration [9,116] Wood species

Si content (vol.%)

Fractional density

Oak

20

0.75

120

210

Maple

20

0.92

180e300

285

Beech

35

0.94

210

275

Pine

40

0.86

180

190

Paulowina

55

0.92

80

Balsa

50

0.72

70

Strength (MPa)

Young’s modulus (GPa)

160

The diversity of natural wood template structures offers a number of options in terms of material selections and applications. Potential applications of biomorphous ceramics derived from natural plant tissue are stimulated by a huge diversity of natural template structures with respect to pore volume (30e90%), pore diameter (800 m/s) was demonstrated. Laminated cellulose fibre ceramics were manufactured from beech veneer by impregnation with phenol formaldehyde resin and airtight sintering [121]. The resulting laminated biocarbon material exhibited a layered structure and partially preserved microstructural characteristics of normal wood. The fracture toughness was reported to increase to 0.6e1.2 MPam1/2 because of the laminated structure and the material exhibits a progressive failure behavior. Medium density fiber (MDF) board and wood powder were used to manufacture cellulose fibre derived ceramics [122,123]. Metal ceramic composites were fabricated by SieCu infiltration into porous biocarbon preforms derived from MDF board [35]. b-SiC and SieCu3Si were the major reaction products forming a biomorphous ceramicemetal composite. A high coefficient of friction of m z 0.6 and a good wear resistance were obtained with an optimized Si/Cu ratio of 3:1.

5.3. Nanopowders Biomorphous b-SiC ceramics were produced at 1400  C from pine wood impregnated with silica [124]. This onestep carbothermal reduction process decreases the cost of manufacturing of SiC ceramics compared with siliconization of carbonized wood in silicon vapor. After ultrasonic milling, the powdered sample showed an average particle size of 140 V/K) was determined which increases with elevating temperature and demonstrated a certain correlation with the anisotropic cellular structure of the biomorphous SiC.

5.4. Medicine

5.6. Catalysis

Biomorphous silicon carbide ceramics were considered as a promising option for dental and orthopedic implants due to their unique mechanical and microstructure properties. The three dimensional, highly oriented pore anatomy of native rattan (Calamus rotang) was used as a template to fabricate biomorphous hydroxyapatite ceramics design for bone regeneration scaffolds [33]. The porosity structure of theses scaffolds is similar to the anatomy of cortical bone with typical cell diameters ranging from 50 nm to 50 mm and long continuous pores up to 400 mm in diameter with a total porosity of 55e70%. In another study, it was concluded that the controlled porosity that biomorphous SiC can present is potentially excellent for osseointegration [125]. Enhanced bioactivity for accelerated osseointegration was achieved by coating the biomorphous SiC ceramic with silica-based bioactive glasses [126]. The biocompatibility and bioactivity were assessed by in vitro simulated body fluid (SBF) release tests and cytotoxicity studies. Bioactivation of highly porous SiC ceramics by a multistep chemical surface activation process was reported by Will et al. [19] After creating a carbon-rich surface oxidation with mineral acids, (HNO3) causes the formation of negatively charged carboxylic (eCOO ) groups, which are able to effectively absorb Ca2þ ions from the body fluid. Absorbed Ca2þ provides ability to trigger in vitro formation of a bonelike apatite reaction layer as proofed by SBF exposure experiments. Variation of cellular anatomy of the native template with respect to porosity structure and fractional density is an advantage for adopting the biomechanical behavior of a biomorphous SiC implant to the surrounding bone tissue.

Hierarchical porous structures are of major significance in catalysis, separation, and adsorption, where long-range and short-range material transports via liquid and vapor phases play a key role. Thus, biomorphous zeolite-based macrostructures with novel shapes, complex functional patterns, and hierarchical porosity are of interest for adsorbents, membranes, sensors, and catalysts of improved performance to be applied in reactor devices [128]. Slurry coating was applied to form zeolite coatings of various thicknesses on ceramic support structures derived from natural tissue templates [129]. A self-supporting MFI-type (ZSM-5) zeolite monolith derived from Luffa was manufactured by a hydrothermal surface crystallization process and tested in a fixedbed catalytic reactor [130]. Despite the low density and high complexity Figure 25, the biomorphous zeolite reactor insert was demonstrated to achieve a reasonable mechanical stability and a high efficiency in n-hexane cracking reaction with a low deactivation tendency even at 550  C. Photocatalytic TiO2 (anatase) presents a high photocatalytic activity for l < 390 nm as well as an important chemical stability. Biomorphous TiO2 processed from natural template, such as for example cedar wood [131] or leave surfaces [132,133], might be used in filtration and flow reactor devices. The photocatalytic activity of the biotemplated TiO2 was demonstrated by degradation of methylene blue or rhodamine dye under UV and visible light irradiation.

5.5. Energy The use of biomorphous high-temperature resistant silicon carbide is an alternative to metals for the fabrication of channeled as well as fibrillar solar radiation absorbers in the high-temperature solar-tower power generation [125]. The service temperatures of the absorber exceeds 800  C and a specific energy absorption rate 1000 kW/m2 is required. Porous SiC with biomorphous structure was investigated for the thermoelectric properties [127]. The dielectric effect offers direct conversion of thermal energy to electricity, which has gained an interest for an ecological process for power conversion. Biomorphous SiC/Si demonstrated a negative thermal electromotoric force

5.7. Optics Biologically inspired optical science is a relatively new and expanding field [134]. Bioinspired optical materials with a microstructure adopted from a natural cellular template are intriguing since they possibly show increased sensitivity, improved spatial resolution, and accelerated signal processing in short-time domains [135]. Biomorphous phosphor materials with a unidirectional cellular microstructure may gain increasing interest for high-resolution screen and imaging devices. The hierarchical cellular tissue anatomy of the three-dimensionally structured biotemplate provides the microenvironment to achieve size, shape, and orientation control of the inorganic phosphor optical material. Phosphor materials exhibiting long-lasting phosphorescence or photostimulated luminescence (PSL) are interesting materials for accumulation and storage of photon-derived energy [136]. These materials have

Chapter | 7.1

Biomorphous Ceramics from Lignocellulosic Preforms

551

FIGURE 25 Biomorphous SiC derived from Luffa and coated with MFI zeolit catalyst (left) and model of reactor insert (right) [130]. For color version of this figure, the reader is referred to the online version of this book.

(a)

(b)

FIGURE 26 Biomorphous BaFBr:Eu2þ (left [135],) and Y2O3:Eu3þ (right, panchromatic PL image [89],) fabricated from Pinus sylvestris.

potential applications in road markings, electrical powerfree illumination, information storage devices, and safety labeling applications in the case of public-light blackouts. Patterning of PSL phosphor materials is attractive for highresolution screen and imaging devices. Biomorphous ceramics with a hierarchical cellular structure were processed by liquid infiltration of inorganic or metal-organic sols into the cellular template followed by a controlled heat treatment (calcinations) at 400e1200  C to obtain oxide phases such as ZnO [95], Eu3þ-doped Y2O3 [89], Eu2þ-doped SrAl2O4 [27], or nonoxide phases like Eu2þ-doped BaFBr [137]. The biomorphous optical ceramics are characterized by a cellular pore structure with pore channel morphologies and dimensions mimicking the cellular anatomy of the template and a submicron

(nanoscale) crystallite microstructure in the inorganic struts, Figure 26.

6. CONCLUSIONS As the result of evolutionary processes, nature displays a huge number of biological materials with a large variety of cellular microstructures. Mimicking the cellular design of natural tissue anatomy by material compositions relevant for functional and engineering applications offers a highly attractive approach for creating a novel class of biomorphous ceramic materials. The mechanical, thermal, or optical properties of the biomorphous ceramics are governed by the hierarchical cellular structure and the phase composition. These materials have a high potential of

552

achieving an excellent density-to-strength relationships and outstanding thermomechanical performance. Biomorphous ceramics may become one of increasing interest for emerging fields of application including catalysis, energy, photonics, and medicine.

Handbook of Advanced Ceramics

[20] [21]

[22]

REFERENCES [1] Dujardin E, Mann S. Bio-inspired materials chemistry. Adv Eng Mat 2002;4:461. [2] Fratzl P. Biomimetic materials research: what can we really learn from nature’s structural materials? J Royal Soc Interface Sci 2007;4:637. [3] Calvert P. Biomimetic ceramics and composites, 17. MRS Bull; 1992. 36. [4] Sarikaya M, Aksay IA. In: Krishnan KM, editor. An introduction to biomimetics: a structural viewpoint, in microstructure of materials. CA: San Francisco Press; 1993. [5] Mann S, Webb J, Williams RP, editors. Biomineralization e chemical and bio-chemical perspectives. VCH Weinheim, D; 1989. [6] Lowenstam HA, Weiner S. On biomineralization. UK: Oxford Univ. Press; 1989. [7] Sharma NK, Williams WS, Zangvil A. Formation and structure of silicon carbide whiskers from rice hulls. J Am Ceram Soc 1984;67:715. [8] Frankel RB, Bazylinski DA. Structure and function of magnetotactic bacteria, in biomimetics e design and processing of materials. Woodbury, NY: AIP Press; 1995. p. 328. [9] Greil P, Lifka T, Kaindl A. Biomoprhic cellular silicon carbide ceramics from wood: I. processing and microstructure, II. Mechanical properties. J Eur Ceram Soc 1998;18. 1961 and 1975. [10] Ota T, Takahashi M, Hibi T, Ozawa M, Suzuki H. Biomimetic process for producing SiC wood. J Am Ceram Soc 1995;78:3409. [11] Byrne CE, Nagle DC. Cellulose-derived composites e a new method for materials processing. Mater Res Innovations 1997;1:137. [12] Shin DW, Park SS, Choa YH, Niihara K. Silicon/silicon carbide composites fabricated by infiltration of a silicon melt into charcoal. J Am Ceram Soc 1999;82:3251. [13] Singh M. Environment conscious ceramics (ecoceramics). Ceram Sci Eng Proc 2000;21:39. [14] Greil P. Biomorphous ceramics from lignocellulosics. J Eur Ceram Soc 2001;21:105. [15] Sotiropoulou S, Sierra-Sastre Y, Mark SS, Batt CA. Biotemplated nano structured materials. Chem Mater 2008;20:821. [16] Hall SR. Biotemplated syntheses of anisotropic nanoparticles. Proc Roy Soc A 2009;465:335. [17] Fan TX, Chow SK, Zhang D. Biomorphic mineralization: from biology to materials. Prog Mater Sci 2009;54:542. [18] Zollfrank C. Biotemplating: polysaccharides in materials engineering. In: Brebbia CA, Carpi A, editors. Design and nature V, comparing design in nature with science and engineering. Southampton, UK: WIT Press; 2010. p. 441. [19] Will J, Hoppe A, Wendt S, Mu¨ller FA, Raya CT, Ferna´ndez JM, et al. Bio-activation of biomorphous silicon carbide bone

[23] [24] [25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

[33]

[34]

[35] [36]

[37] [38] [39] [40]

implants. Acta Biom 2010;6. http://dx.doi.org/10.1016/ j.actbio.2010.06.036. Paris O, Burgert I, Fratzl P. Biomimetics and biotemplating of natural materials, 35. MRS Bulletin; 2010. p. 219. Paris O, Zollfrank C, Zickler GA. Decomposition and carbonisation of wood bio-polymers e a microstructural study of softwood pyrolysis. Carbon 2005;43:53. Zollfrank C, Fromm J. Ultrastructural development of the softwood cell wall during pyrolysis. Holzforschung 2009;63:248. Patel M, Padhi BK. Production of alumina fibre through jute fibre substrate. J Mat Sci 1990;25:1335. Krishnaro RV, Mahajan YR. Preparation of silicon carbide fibres from cotton fibre and silicon nitride. J Mat Sci Lett 1996;15:232. Vogli E, Sieber H, Greil P. Biomorphic SiC-ceramic prepared by Si-vapor phase infiltration of wood. J Eur Ceram Soc 2002;22:2663. Rambo CR, Andrade T, Fey T, Sieber H, Martinelli AE, Greil P. Microcellular Al2O3 ceramics from wood for filter applications. J Am Ceram Soc 2008;91:852. Kostova M, Zollfrank C, Batentschuk M, Goetz-Neunhoeffer F, Winnacker A, Greil P. Bioinspired design of SrAl2O4:Eu3þ phosphor. Adv Funct Mat 2009;19:599. Singh M, Salem JA. Mechanical properties and microstructure of biomorphic siliconcarbide ceramics fabricated from wood precursors. J Eur Ceram Soc 2002;22:2709. Kiselov VS, Kalabukhova EN, Sitnikov AA, Lytvyn PM, Poludin VI, Yukhymchuk VO, et al. Effect of Si infiltration method on the properties of bio-morphous SiC. Semicond Phy Quantum Electron Optoelectron 2009;12:68. Wilkes TE, Stock SR, De Carlo F, Xiao X, Faber KT. X-ray micro-computed tomography of beech wood and biomorphic C, SiC and Al/SiC composites. Phil Mag 2009;89:1373. Vogli E, Mukerji J, Hoffman C, Kladny R, Sieber H, Greil P. Conversion of oak to cellular silicon carbide ceramic by gas-phase reaction with silicon monoxide. J Am Ceram Soc 2001;84:1236. Yamane H, Kawamura F, Yamada T. Low temperature synthesis of biomorphic cellular SiC ceramics from wood by using a Na flux. J Ceram Soc Jpn 2008;116:163. Eichenseer C, Will J, Rampf M, Wend S, Greil P. Biomorphous porous hydroxyl-apatite-ceramics from rattan (Calamus rotang). J Mater Sci Mater Med 2010;21:131. Rambo CR, Mu¨ller FA, Mu¨ller L, Sieber H, Hoffmann L, Greil P. Biomimetic apatite coating on biomorphous alumina scaffolds. Mat Sci Eng, C 2006;26:92. Haas D, Fey T, Greil P. Biomorphous metal-ceramic-composites with high coefficient of friction. Adv Eng Mat 2007;9:892. Herzog A, Vogt U, Kaczmarek O, Klingner R, Richter K, Thoemen H. Porous SiC ceramics derived from tailored woodbased fiberboards. J Am Ceram Soc 2006;89:1499. Travitzky N, Windsheimer H, Fey T, Greil P. Preceramic paperderived ceramics (feature). J Am Ceram Soc 2008;91:3477. Fengel D, Wegener G. Wood: chemistry, ultrastructure, reactions. Berlin: de Gruyter; 1984. Gibson EJ. Wood: a natural fibre reinforced composite. Metals Mater 1992;6:333. Sturcova A, Davies GR, Eichhorn SJ. Elastic modulus and stresstransfer properties of tunicate cellulose whiskers. Biomacromolecules 2005;6:1055.

Chapter | 7.1

Biomorphous Ceramics from Lignocellulosic Preforms

[41] Wagenfu¨hr R. Holzatlas. Germany: Hanser Verlag Leipzig; 2007. [42] Fisher JB, Tan HTW, Toh LPL. Xylem of rattans: vessel dimensions in climbing palms. Am J Bot 2002;89:196. [43] Zollfrank C, Kladny R, Sieber H, Greil P. Biomorphous SiOC/Cceramic composites from chemically modified wood templates. J Eur Ceram Soc 2004;24:479. [44] Shebani AN, van Reenen AJ, Meincken M. The effect of wood extractives on the thermal stability of different wood-LLDPE composites, thermo-chim. Acta 2008;471:43. [45] Byrne CE, Nagle DC. Carbonization of wood for advanced materials applications. Carbon 1997;35:259. [46] Kercher AK, Nagle DC. Monolithic activated carbon sheets from carbonized medium-density fiberboard. Carbon 2003;41:3. [47] Mochida I, Ku C, Korai Y. Anodic performance and insertion mechanism of hard carbons prepared from synthetic isotropic pitches. Carbon 2001;39:399. [48] Tang MM, Bacon R. Carbonization of cellulose fibers I. low temperature pyrolysis. Carbon 1964;2:211e20. and; Carbonization of cellulose fibers II. Physical property study. Carbon 1964; 2:221. [49] Peng SJ, Shao HL, Hu XC. Lyocell fibers as the precursor of carbon fibers. J Appl Polym Sci 2003;90:1941. [50] Kim DY, Nishiyama Y, Wada M, Kuga D. Graphitization of highly crystalline cellulose. Carbon 2001;39:1051. [51] Kim DY, Nishiyama Y, Wada M, Kuga S, Okano T. Thermal decomposition of cellulose crystallites in wood. Holzforschung 2001;55:521. [52] Ishimaru K, Hata T, Bronsveld P, Imamura Y. Microstructural study of carbonized wood after cell wall sectioning. J Mater Sci 2007;42:2662. [53] Hata T. Microstructural analysis of biocarbon from wood. Cellulose Comm 1999;6:27. [54] Hata T, Vystavel T, Bronsveld P, DeHosson J, Kikuchi H, Nishimiya K, et al. Catalytic carbonization of wood charcoal: graphite or diamond? Carbon 2004;42:961. [55] Zickler GA, Smarsly B, Gierlinger N, Peterlik H, Paris O. A reconsideration of the relationship between the crystallite size La of carbons determined by X-ray diffraction and Raman spectroscopy. Carbon 2006;44:3239. [56] Bourgois JP, Doat J. Torrefied wood from temperate and tropical species. Advantages and prospects. In: Egne´us H, Ellega˚rd A, editors. Bioenergy, vol. 84. London: Elsevier; 1985. p. 153. [57] Gosselink RJA, Krosse AMA, van der Putten JC, van der Kolk JC, de Klerk-Engels B, van Dam JEG. Wood preservation by lowtemperature carbonization. Ind Crops Prod 2004;19:3. [58] Crawford RL. Lignin biodegradation and transformation. New York: John Wiley and Sons; 1981. [59] Zickler GA, Wagmaier W, Funari SS, Burghammer M, Paris O. In-situ investigation of thermal decomposition of wood cellulose. J Anal Appl Pyrolysis 2007;80:134. [60] Zollfrank C, Paris O. Thermal degradation of wood and microstructure of biocarbon, Proc. 12th Int. Symp. Wood pulp. Chem, vol. 1. Madison: University of Wisconsin; 2003. 349. [61] Nishimiya K, Hata T, Imamura Y, Ishihara S. Analysis of chemical structure of wood charcoal by X-ray photoelectron spectroscopy. J Wood Sci 1998;44:56.

[62]

[63]

[64] [65] [66]

[67]

[68] [69]

[70] [71]

[72]

[73] [74] [75] [76]

[77] [78] [79]

[80]

[81]

[82]

553

Hata T, Imamura Y, Kobayashi E, Yamane K, Kikuchi K. Onionlike graphitic particles observed in wood charcoal. J Wood Sci 2000;46:89. Cheng HM, Endo H, Okabe T, Saito K, Zheng G. Graphitization behaviour of wood ceramics and bamboo ceramics as determined by X-ray diffraction. J Porous Mater 1999;6:233. Hillig WB. Melt infiltration approach to ceramic matrix composites. Comm Am Ceram Soc 1988;71:C96. Yang J, Ilegbusi OJ. Kinetics of silicon-metal alloy infiltration into porous carbon. Composites: Part A 2000;31:617. Varela Feria FM, Martinez-Fernandez J, de ArellanoLopez AR, Singh M. Low density biomorphic silicon carbide: microstructure and mechanical properties. J Eur Ceram Soc 2002;22:2719. Gern FH, Kochendo¨rfer R. Liquid silicon infiltration: description of infiltration dynamics and silicon carbide formation. Composites A 1997;28:355. Fitzer E, Gadow R. Fibre reinforced silicon carbide. Am Ceram Soc Bull 1986;65:326. Messner RP, Chiang YM. Liquid-phase reaction bonding of silicon carbide using alloyed Si-Mo melts. J Am Ceram Soc 1990;73:1193. Kaindl A. Cellular SiC ceramics from wood. PhD thesis (in German), Univ. of Erlangen: Germany; 1999. Zollfrank C, Sieber H. Microstructure evolution and reaction mechanism of bio-morphous SiSiC ceramics. J Am Ceram Soc 2005;88:51. Mason EA, Malinauskas MM. Gas transport in porous media: the dusty-gas model, chemical engineering monographs, 17. Amsterdam, NE: Elsevier; 1983. Vignoles GL. Modelling of the CVI process. Adv Sci Technol 2006;50:97. Trans Tech Publ, Switzerland. Skamser DJ, Jennings HM, Johnson DL. Model of chemical vapour infiltration temperature gradients. J Mat Res 1997;12:724. Greil P. Templating approaches using natural cellular plant tissue, 35(2). MRS Bull; 2010. p. 145. Zhou H, Singh RN. Kinetics model for the growth of silicon carbide by the reaction of liquid silicon with carbon. J Am Ceram Soc 1995;78:2456. Chiang YM, Messner RP, Terwilliger CD. Reaction formed silicon carbide. Mater Sci Eng, A 1991;114:63. Pampuch R, Walasek E, Bialoskorski J. Mechanism of reactions in Sil þCf system. Ceram Int 1987;13:63. Pancholi V, Mallick D, AppaRao Ch, Samajdar I, Chakrabarti OP, Maiti HS, et al. Microstructural characterization using orientation imaging microscopy of cellular Si/SiC ceramics synthesized by replication of Indian dicotyledonous plants. J Eur Ceram Soc 2007;27:367. Sieber H, Zollfrank C, Popovska N, Almeida D, Gerhard H. Gas phase processing of porous, biomorphous TiC ceramics, in key engineering materials, vols. 264e268. Switzerland: TransTechPubl; 2004. p. 2227. Li JZ, Furuno T, Katoh S, Uehara T. Chemical modification of wood by anhydrides without solvents or catalysts. J Wood Sci 2000;46:215. Li JZ, Furuno T, Katoh S. Preparation and properties of acetylated and propionylated wood-silicate composites. Holzforschung 2001;55:93.

554

[83] Mai C, Militz H. Modification of wood with silicon compounds. Treatment systems based on organic silicon compounds e a review. Wood Sci Technol 2004;37:453. [84] Newman RH, MacKay AL. Silica spicules in canary grasses. Ann Bot 1983;52:927. [85] Sapei L, No¨ske R, Strauch P, Paris O. Isolation of mesoporous biogenic silica from the perennial plant equisetum hyemale. Chem Mater 2008;20:2020. [86] Currie HA, Perry CC. Silica in plants: biological, biochemical and chemical studies. Ann Bot 2007;100:1383. [87] Webb MA. Cell-mediated crystallization of calcium oxalate in plants. Plant Cell 1999;11:751. [88] Brinker CJ, Scherer GW. Solegel science. London (: Acad. Press Inc.; 1990. [89] van Opdenbosch D, Kostova MH, Gruber S, Krolikowski S, Greil P, Zollfrank C. Replication of wood biomorphous nanocrystalline Y2O3:Eu3þ phosphor materials. Wood Sci Technol 2010;44:547. [90] Cao J, Rambo CR, Sieber H. Manufacturing of microcellular, biomorphous oxide ceramics from native pine wood. Ceramics Int 2004;30:1967. [91] Ota T, Imaeda M, Takase H, Kobayashi M, Kinoshita N, Hirashita T, et al. Porous titania ceramic prepared by mimicking silicified wood. J Am Ceram Soc 2000;83:1521. [92] Singh M, Yee BM. Light weight biomorphous cellular ceramics from cellulose templates. Ceram Eng Sci Proc 2002;23:719. [93] Bosch Ojeda C, Sanchez Rojas F, Cano Pavon JM. Experimental design in the optimization of a microwave acid digestion procedure of biomorphic ceramic based on ZrO2. J Wood Chem Technol 2010;30:140. [94] Liu Z, Fan T, Zhang D. Synthesis of biomorphous nickel oxide from a pinewood template and investigation on a hierarchical porous structure. J Am Ceram Soc 2006;89:662. [95] Liu ZT, Fan TX, Ding J, Zhang D, Guo QX, Ogawa H. Synthesis and cathodo-luminescence properties of porous wood (fir)templated zinc oxide. Ceram Int 2008;34:69. [96] Matovic B, Babic B, Egelja A, Radosavljevic-Mihajlovic A, Logar V, Saponjic A, et al. Preparation of porous silica ceramics using the wood template. Mater Manuf Proc 2009;24:1109. [97] Jeronimidis G. Structureeproperty relationships in biological materials, and design and function of structural biological materials. In: Elices M, editor. Structural biological materials. Amsterdam, NE: Pergamon Press; 2000. p. 3. [98] Greil P, Vogli E, Fey T, Bezold A, Popovska N, Gerhard H, et al. Effect of microstructure on the fracture behavior of biomorphic silicon carbide ceramics. J Eur Ceram Soc 2002;22:2697. [99] Martinez-Fernandez J, Munoz A, de Arellano Lopez AR, Valera Feria FM, Dominguez-Rodriguez A, Singh M. Microstructureemechanical properties correlation in siliconized silicon carbide ceramics. Acta Mater 2003;51:3259. [100] Gibson LJ, Ashby MF. Cellular solids: structure and properties. NY: Pergamon Press; 1988. [101] Vogli E, Sieber H, Greil P. Wood derived porous and cellular ceramics. Ceramic Eng Sci Proc 2002;23:211. [102] Rice RW. Porosity of ceramics. NY: Marcel Dekker; 2000. [103] Andersson CA. Derivation of the exponential relation for the effect of ellipsoidal porosity on elastic modulus. J Am Ceram Soc 1996;79:2181.

Handbook of Advanced Ceramics

[104] Gibson LJ, Ashby MF. Hierarchical cellular materials, vol. 295. MRS Proceedings; 1992. [105] Winterstein G, Stahn M, Voigt M, Kuhn G. Ceramic honeycumb structures. CFI, Ceram Forum Int 1998;75:8. [106] Roberts AP, Garboczi EJ. Elastic properties of model porous ceramics. J Am Ceram Soc 2000;83:3041. [107] Ramakrishnan N, Arunachalam VS. Effective elastic moduli of porous ceramic materials. J Am Ceram Soc 1993;76:2745. [108] Mackerle J. Ceramics and ceramic matrix composites: finite element and boundary element analyses. A bibliography (1998e2000). Finite Elem Anal Des 2002;38:567. [109] Fey T, Sieber H, Greil P. Stress distribution in biomorphous SiCceramics under radial tensile loading. J Eur Ceram Soc 2005;25:1015. [110] Gutierrez-Mora F, Goretta KC, Varela-Feria FM, Arellano Lo´pez AR, Martı´nez Ferna´ndez J. Indentation hardness of biomorphic SiC. Int J Refract Met Hard Mat 2005;23:369. [111] de Arellano-Lo´pez AR, Martı´nez-Ferna´ndez J, Varela-Feria FM, Orlova TS, Goretta KC, Gutierrez-Mora F, et al. Erosion and strength degradation of biomorphic SiC. J Eur Ceram Soc 2004;24:861. [112] Amirthan G, Udayakumar A, Bhanu Prasad VV, Balasubramanian M. Solid particle erosion studies on biomorphic Si/SiC ceramic composites. Wear 2010;268:145. [113] Evans AG. In: Bradt RC, Hasselmann DPH, Lange FF, editors. Impact damage in ceramics, in fracture mechanics of ceramics, vol. 3. NY: Plenum; 1978. p. 303. [114] Wiederhorn SM, Lawn BR. Strength degradation of glass impacted with sharp particles. J Am Ceram Soc 1979;62:66.0. [115] Martı´nez Ferna´ndez J, Valera Feria FM, Singh M. High temperature compressive mechanical behavior of biomorphic silicon carbide ceramics. Scr Mater 2000;43:813. [116] Park HS, Jang JJ, Lee KH, Lim KH, Park SB, Kim YC, et al. Effects of microstructure on flexural strength of biomorphic C/SiC composites. Int J Fract 2008;15:233. [117] Mark RE. Mechanical properties of fibres. In: Mark RE, et al. editors. Handbook of physical testing of paper, vol. 1. NY: Marcel Dekker; 2001. p. 727. [118] Thiermann KH, Scha¨fer W. Ceramic components for environmentally friendly internal combustion engines. Ceram Forum Int 2000;80:E15. [119] Kormann M. Synthesis of porous carbon from biomorphous carbide based ceramics. PhD thesis, University of ErlangenNuernberg; 2010. [120] Heidenreich B, Krenkel W, Lexow B. Development of CMCmaterials for light-weight armor. In: Ceram Eng Sci Proc, vol. 24(3). Am Ceram Soc; 2003. p. 375. [121] Sun DL, Yu XC, Liu WJ, Sun DB. Laminated wood-ceramics prepared from beech veneer and phenol formaldehyde resin. Wood Fiber Sci 2010;42:1. [122] Akagaki T, Hokkirigawa K, Okabe T, Saito K. Friction and wear of wood ceramics under oil and water lubricated sliding contacts. J Porous Mat 1999;6:197. [123] Li J, Li SJ. Pyrolysis of medium density fiber board impregnated with phenol form-aldehyde resin. Wood Sci Technol 2006;52:331. [124] Vyshnyakova K, Yushin G, Pereselentseva L, Gogotsi Y. Formation of porous SiC ceramics by pyrolysis of wood impregnated with silica. Int J Appl Ceram Technol 2006;3:485.

Chapter | 7.1

Biomorphous Ceramics from Lignocellulosic Preforms

[125] Eichenseer C, Will J, Rampf M, Wend S, Greil P. Biomorphous porous hydroxyl-apatite-ceramics from rattan (Calamus rotang). J Mater Sci Mater Med 2010;27:131. [126] deArellano-Lopez AR, Martinez-Fernandez J, Gonzalez P, Dominguez C, Fernandez-Quero V, Singh M. Biomorphic SiC: a new engineering ceramic material. Int J Appl Ceram Technol 2004;1:56. [127] Gonzalez P, Serra J, Lista S, Chiussi S, Leon B, Perez-Amor M, et al. New biomorphic SiC ceramics coated with bioactive glass for biomedical applications. Biomaterials 2003;24:4827. [128] Kiselov VS, Poludin VI, Kiselyuk MP, Kryskov TA, Belyaev AE. Effect of macrostructure on the thermoelectric properties of biomorphous SiC/Si ceramics. Semicond Phys Quantum Electron Optoelectron 2009;12:64. [129] Dong A, Wang Y, Tang Y, Ren N, Zhang Y, Yue Y, et al. Zeolitic tissue through wood cell templating. Adv Mater 2002;14:926. [130] Zampieri A, Schwieger W, Zollfrank C, Greil P. Organic preforms of biological origin: natural plant tissues as templates for inorganic and zeolitic macrostructures. In: Behrens P, Baeuerlein E, editors. Handbook of biomineralization. Wiley-VCH; 2007. p. 255. [131] Zampieri A, Mabande GTP, Selvam T, Schwieger W, Rudolph A, Herman R, et al. Biotemplating of luffa cylindrica sponges to

[132]

[133]

[134]

[135] [136] [137] [138]

555

self-supporting hierarchical zeolite macrostructures for bioinspired structured catalytic reactors. Mat Sci Eng C 2006;26:130. Zhu S, Zhang D, Chen Z, Zhou G, Jiang H, Li J. Sonochemical fabrication of morphogenetic TiO2 with hierarchical structures for photocatalyst. J Nanopart Res 2010;12:2445. Li X, Fan T, Zhou H, Chow SK, Zhang W, Zhang D, et al. Enhanced light-harvesting and photocatalytic properties in morph-TiO2 from green-leaf biotemplates. Adv Funct Mater 2009;19:45. Hana T, Fan T, Chow SK, Zhang D. Biogenic NeP-codoped TiO2: synthesis, characterization and photocatalytic properties. Bioresour Technol 2010;101:6829. Lee LP, Szema R. Review, inspirations from biological optics for advanced photonic systems. Science 2005;310:1148e50. Sinzinger S, Jahns J. Microoptics. 2nd ed. Weinheim: Wiley-VCH Verlag; 2003. Goldberg P. Luminescence of inorganic solids. London: Academic Press; 1966. [Chapter 4] 1. Kostova MH, Batentschuk M, Goetz-Neunhoeffer F, Gruber S, Winnacker A, Greil P, et al. Biotemplating of BaFBr:Eu2þ for X-ray storage phosphor applications. Mat Chem Phys 2010;123:166.

Chapter 7.2

Application of a Quartz Crystal Microbalance with Dissipation for In Situ Monitoring of Interfacial Phenomena between Bioceramics and Cells Motohiro Tagaya,* Cross J. Scott,y Toshiyuki Ikomay and Junzo Tanakay *

Department of Materials Science and Technology, Nagaoka University of Technology, 1603-1 Kamitomioka-machi, Nagaoka, Niigata 940-2188, Japan,

Former Affiliate: Department of Metallurgy and Ceramics Science, Tokyo Institute of Technology, y Department of Metallurgy and Ceramics Science, Tokyo Institute of Technology, 2-12-1 Ookayama, Meguro, Tokyo 152-8550, Japan

Chapter Outline 1. Introduction 2. Interface Between Bioceramics and Cells 3. In Situ Monitoring of Interfacial Phenomena

557 557 559

1. INTRODUCTION When bioceramics are implanted into the body, proteins in the body fluid adsorb and subsequently cells adhere on the surface. Thus, investigation of the interfacial layers formed between the bioceramics and cells is of significant importance for controlling cell functions and understanding biocompatible phenomena. The interfacial layers can be divided into three layers, which include the ion/water adlayer, protein adlayer, and adherent cell layer. In order to in situ monitor the interfacial layers, a quartz crystal microbalance with dissipation (QCM-D) technique is one of the excellent analytical methods. The hydroxyapatite sensor surface deposited by an electrophoretic method applicable for the QCM-D technique has been recently studied for understanding the interfacial phenomena. In this chapter, the QCM-D technique that can monitor the interfacial phenomena between bioceramics and cells is described and suggests the importance of monitoring the interfacial phenomena.

2. INTERFACE BETWEEN BIOCERAMICS AND CELLS 2.1. Bioceramics: Current Status and Challenges Ceramics used for the repair and reconstruction of diseased or damaged parts of the body are termed as “bioceramics.”

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00030-7 Copyright Ó 2013 Elsevier Inc. All rights reserved.

4. Summary References

571 571

Bioceramics are a class of advanced ceramics, which are defined as ceramic products or components used for the repair and replacement of diseased and damaged parts of the musculoskeletal system, employed in medical and dental applications, mainly as implants and replacements. They are biocompatible, and can be inert, bioactive, and degradable in a physiological environment that makes it an ideal biomaterial. Bioceramics have become a diverse class of biomaterials presently including three basic types, i.e., bio-inert highstrength ceramics (alumina, zirconia, etc.), bioactive ceramics (bioactive glasses, glass-ceramics, hydroxyapatite (HAp), etc.) which form direct chemical bonds with bone or even with the soft tissue of a living organism, and various bioresorbable ceramics (tricalcium phosphate, etc.) that actively participate in the metabolic processes of an organism with predictable results. These characteristic materials applicable for biocompatible materials, such as artificial bone, dental roots, joints, blood vessels and trachea, and percutaneous devices and bone filling, have been widely studied. As future challenges, the surface property put on the biotissue should be investigated in order to elucidate the material design which promotes organizational cures. It has been known that ceramics made of calcium phosphate salts can be used for replacing and augmenting bone tissue. The most widely used calcium phosphate based bioceramics is HAp. HAp has the chemical formula of

557

558

Ca10(PO4)6(OH)2 and possesses a Ca/P ratio of 1.67, a hexagonal structure, and a space group of P63/m with the cell dimensions of a ¼ 9.423 A and c ¼ 6.875 A [1]. The Ca(II) atom is surrounded by six O atoms belonging to the PO4 groups and an OH group, whereas the Ca(I) atom is nearly octahedrally surrounded by nine O atoms belonging to the PO4 groups. The Ca(II) atoms form a triangle normal to the c-axis, and the Ca(II) triangles stack along the c-axis,  rotating mutually 60 from each other. The HAp surface has both positive and negative charges due to the charged Ca ion and the PO4 groups, indicating the high biocompatibility [2] and characteristic property of protein absorption [3,4]. Thus, the HAp has been widely studied for application use as a bone-filling material [5e8], column-filling material [9e11], drug delivery carrier [12e15], and novel inorganiceorganic composite material [16,17]. Although the applications of bioceramics have been widely studied, there is little information on the interfacial phenomena between the surfaces and biomolecules. Thus, the interfacial phenomena with the dynamic biomolecular behavior on the surfaces should be elucidated, and the future toward potential use of bioceramics should be studied for improving the biocompatibility and designing superior surfaces.

2.2. Importance of Interface Phenomena Figure 1 shows the scheme of successive events after the implantation. When biomaterials are implanted into the body, three successive processes occur as follows [18]. The first substances to reach the surface are ions and water molecules, which are known to interact and bind depending on the surface properties. The water layers formed on the surface are an important factor influencing proteins, and then other

FIGURE 1 Scheme of successive events after implantation of biomaterials.

Handbook of Advanced Ceramics

molecules arrive a little later. Second, water-soluble proteins have hydration shells, and the interactions between the surface water layers with the protein water shells strongly influence the fundamental kinetics and thermodynamics at the interface to form protein adlayers. The interactions would determine the structures of the protein adlayers such as being denatured or not, orientation, and coverage, indicating that the surface properties such as topography, wettability, and charge are of great importance. Third, when cells arrive at the surface, they realize the structure of protein adlayers, which are initially determined by the water layers, to adhere, spread, and form the interface on the surface. Thus, the initial cell adhesion behavior is strongly affected by both the surface properties and structures of the protein adlayers, and the interface layers would be the dominant factors affecting the biocompatibility [18e21]. Figure 2 shows the scheme of a typical interface between the protein adlayer and cell on the biomaterial surface. After implantation into the body, multiple proteins in the body fluid immediately and competitively adsorb on the surface [19e21]. Albumin (Ab) and immunoglobulin (IgG) are the larger mass fractional protein components in the fluid, whereas extracellular matrix (ECM) proteins, which obligate the cell activities, are relatively minor components. In the protein adlayer, the structure of the extracellular matrix (ECM) containing the arginineeglycineeasparagine (RGD) sequence plays a role at the interface between the

FIGURE 2 Typical scheme of the interface between protein adlayer and cell on the biomaterial surface. An overview of the structure at the interface in which cell membrane containing integrin binds to the ECMproteins on the surface is shown.

Chapter | 7.2

Application of a Quartz Crystal Microbalance with Dissipation

surface and cell [24,30,31]. The conformation and denaturation of protein adlayers and the orientation of the RGD sequence strongly governs the biocompatibility [22,23], which is attributed to the cell activities and functions, such as adhesion, proliferation, migration, differentiation, expression, and survival. The ECM is classified into collagen, noncollagen glycoprotein, elastin, and proteoglycan groups [35e37]. The noncollagen glycoprotein includes fibrinogen (Fgn), fibronectin (Fn), laminin, vitronectin, thrombospondins, and tenascins, which significantly affect the initial cell adhesion behavior. On the other hand, integrins in the cell membrane directly mediate attachment between a cell and the RGD sequence of the protein adlayer, which may also be other cells or the ECM [25,26]. The cytoskeleton firmly combined with the integrin inside the cell membrane is changed, and subsequently forms adhesion points at the interface. Along with forming the points, signal transduction molecules through the cytoskeleton are successively communicated to determine the cell functions. Simultaneously, the cell produces ECM at the interface to consolidate the interfacial junction. The interfacial region above the material surface including the ECMeintegrinecytoskeleton significantly affects the cell activities and functions. Therefore, in situ monitoring technique of the interface between the material surface and cell is important for controlling the cell functions. The surface morphology and protein structure are well known to be attractive parameters for controlling the cell functions. The surface morphology at the nano- and micrometer levels and subsequent protein adsorption are known to affect the later cellular activities (proliferation, survival, and gene expression) [27e29]. These later activities would be significantly affected by the initial cellular activities (adhesion and spreading), but there has been no clarification of the relationship between the physicochemical property and the initial behavior. The initial cell behavior on the adsorbed Fn on the self-assembled monolayer [32,33] and on poly (tetrafluoroethylene) surfaces [34] has been investigated, but the higher-order structure of the ECM proteins has not been significantly related to the initial behavior. It appears that techniques for in situ monitoring the interfacial phenomenon in an aqueous solution have not been established. Thus, it is indispensable to clarify the phenomena between the material surface and cell during the initial stage by new techniques for understanding the biocompatibility.

3. IN SITU MONITORING OF INTERFACIAL PHENOMENA 3.1. Various Current Techniques Among the various means to investigate the biomolecular interactions on the surfaces, in situ monitoring and highlysensitive detecting techniques for the adsorption and

559

desorption on material surfaces in an aqueous solution have been widely developed. The in situ monitoring of the dynamic interaction changes by the calorific capacity [44,45], Fourier transform infrared spectroscopy with attenuated total internal reflectance (FTIR/ATR) [48,49], optical interferometry [42,43], ellipsometry [46,47], surface plasmon resonance (SPR) [38,39], total internal reflection fluorescence (TIRF) [40,41], and a quartz crystal microbalance (QCM) [50] have already been investigated. The evaluation of the absolute amount of the adlayer with the calorific capacity and FTIR/ ATR techniques is difficult, and the refraction index parameters for the ellipsometry and optical interferometry and the refraction index parameter and density for the TIRF and SPR are needed for the evaluation, indicating the difficulty of monitoring an unknown adlayer. Biomolecules have various functional groups that include hydrophobic groups (aromatic rings, alkyl chains, etc.), hydrophilic groups (eOH, eNH2, etc.), and dissociation groups ( COO ; NHþ 3 ; ¼ PO4 ). The hydrophilic and dissociation groups interact to form water molecules with hydrogen bonds in an aqueous solution. Thus, the monitoring of biomolecules with a hydration structure is of great importance for investigating the interfacial structure. The QCM technique provides the mass containing the hydrated water molecules [51] to show the higher values as compared with other techniques, such as ellipsometry and SPR [52,53]. The QCM technique is also a reliable and suitable method with a high sensitivity due to its high response to mass changes of several nanogram order and for forming multilayered films, and can be used for evaluating the interface between the materials and cells.

3.2. Quartz Crystal Microbalance with Dissipation (Qcm-D) Technique 3.2.1. Fundamental Principles of Qcm-D The mechanism of signal transduction in the QCM technique, which relies upon the piezoelectric effect in quartz crystals, was first discovered in 1880 by the Curie brothers via a pressure effect on quartz [54]. A change in inertia of a vibrating crystal has been shown by Lord Rayleigh to alter its resonant frequency (f) [55]. The subsequent developments for crystal stability were achieved through the use of electric resonators [56] and room-temperature stable AT-cut crystals [57]. For quartz crystals used in typical QCM applications, a few tenths of a mm in thickness at a 35 100 angle from the Z-axis are cut as the AT form [58e60]. This geometry provides a stable oscillation with almost no temperature fluctuation in the f at room temperature. A valuable advantage of the QCM technique for monitoring solid thin films is its capability to detect an adsorbed mass. For the pure elastic mass on the sensor surface, the well-known linear equation was first developed

560

Handbook of Advanced Ceramics

by Sauerbrey [61]. Based on the assumption of a homogeneous foreign mass (Dm) with a spatially uniform density, the equation can be written including harmonic resonances as follows: Df ¼

2Dmnf02 pffiffiffiffiffiffiffiffiffiffi A mq rq

(1)

where Df is the frequency shift of the quartz, f0 is the fundamental frequency, n is the overtone number, elasticity of quartz, and density of quartz, and A is metallic electrode area on the quartz. mq and rq are the shear modulus (2.95  1011 dyn/cm2) and density (2.65 g/cm3) of the quartz, respectively. Dm, in the nanogram range, uses a linear relationship between the added mass and the change in the resonant frequency, and an increase in the mass bound to the quartz surface causes the crystal’s oscillation frequency to decrease. However, the equation does not sufficiently fit for “viscoelastic adlayers” due to the strong damping of oscillation in liquid. Konash [62] and Nomura [49] modified the oscillation circuit and covered the nonsensitive sensor surface to provide QCM monitoring with oscillation in liquid. Landau et al. [63] showed that the effective detecting height l in liquid is rffiffiffiffiffiffiffiffi hl (2) l ¼ rl f 0

a 10-mm sensor which contains a thin disk of crystalline AT-cut quartz sandwiched between two gold electrodes, and the top one is the sensor surface. Any layer of molecules absorbing on the surface of the quartz crystal creates subtle changes in the frequency of oscillation and its duration. Figure 3 shows the experimental setup for the QCM-D, and the typical rapid excitation of the QCM near resonance and subsequent damping. The measurement cell is connected to computer-controlled electronics. A control program tracks the f and excites the crystal through a signal generator with an applied AC voltage for 10 ms. When the signal generator is switched off for 2 ms, an oscilloscope records the exponentially damped response of the oscillation from the quartz crystal in Figure 3(a). Thus, the computer repeatedly excites the crystal and records the changes in the f and decay time (s) as shown in Figure 3(b), and the damping data are transformed from analog to digital. Based on Eqn (4), a numerical fit of an exponentially damped sinusoidal function to the recorded signal provides the f and s: AðtÞ ¼ A0 et=s sin 2pft

(4)

where A0 is the amplitude at t ¼ 0, and s is defined as the time when A0 reaches 1/e. These values are stored as one data point as a function of time.

where rl and hl are the liquid density and viscosity of the liquid, respectively. Since the liquid is dragged by the oscillation of the sensor, there is a problem of an increase in the resonance series resistance. Assuming the adsorbed liquid on the sensor surface and considering l dependent on rl and hl of the liquid and Dm (¼ rl l A), the relationship between the Df and the viscosityedensity product from Eqns (1) and (2) was represented by Kanazawa and Gordon as follows [64]: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3=2 (3) hl rl =pmq rq Df ¼ f0

When the interface between the solution and the crystal surface is nonslip and the region at l is a uniform viscoelastic layer, the relation between Df and Dm in Eqn (3) is established. In other words, the equation for the locally different energy damping at l on the surface has not been established. The bimolecular adlayer in an aqueous solution is complex with a hydration structure, and can be considered to be a viscoelastic layer. Thus, the accurate evaluation of the mass and viscoelasticity of the adlayer is needed. Based on these points, Kasemo et al. suggested a quartz crystal microbalance with dissipation (QCM-D) technique for the in situ monitoring of the energy dissipation process with bimolecular adsorption [65e68], which provides a rapid evaluation of the mass and viscoelasticity of the hydrated bimolecular adlayers. The heart of the QCM-D is

FIGURE 3 (a) Experimental setup for QCM-D measurement and (b) the typical rapid excitation of the QCM near resonance, followed by an exponentially damped sinusoidal wave after the rapid disconnection to obtain the f and s values.

Chapter | 7.2

Application of a Quartz Crystal Microbalance with Dissipation

561

The energy dissipation (D) is a dimensionless quantity defined by Eqn (5) [65e68]:

Dm, and DD corresponds to DEdissipation due to the constant Estored during each period as follows [69]:

D ¼ 1=Q ¼ Edissipation =2pEstored

DD ¼ DEdissipation =2pEstored

(5)

where Q is the quality factor of oscillation in the equivalent electric circuit (Butterworth Van-Dyke element), and Estored and Edissipation are the energy stored and energy dissipation in the oscillation circuit during one period, respectively. The Q value is represented as Q ¼ R1 =2pfL1

(6)

where R1 and L1 are an inductance and resistance, respectively. The equivalent electric circuit corresponds to a mechanical vibration model, and s equals 2L1/R1. Thus, based on Eqns (5) and (6), D is represented by s and f as follows: D ¼ 1=pf s

(7)

Figure 4 shows the schemes of the damping with the biomolecular adsorption on the QCM-D sensor surface, and the Df and DD curves in situ measured with the changes in f and D. When the changes from f0 and D0 in the time at t ¼ 0 to fi and Di in the time at t ¼ i are observed with the formation of an adlayer, Df (¼ fi e f0) and DD (¼ DieD0) were calculated. The Df based on Eqn (1) corresponds to

FIGURE 4 Schemes of (a) the damping with the biomolecular adsorption on the sensor surface, and (b) the Df and DD curves in situ measured with the changes in f and D.

(8)

The Df and DD curves can be obtained by measuring fi and Di in the time at t ¼ i for each time. Viscous adlayers effectively lose the oscillation energy by friction among the molecules, and then the oscillations are rapidly damped to produce higher DD values. Elastic adlayers effectively retain the oscillation energy, and the oscillations are slowly damped to produce lower DD values. Therefore, in situ monitoring of the mass and viscoelastic changes due to biomolecular adsorption can be successfully measured.

3.2.2. Fabrication of Hydroxyapatite (Hap) Sensor Surface and its Reusability To understand the interaction between a biomaterial surface and biomolecules, the use of a biomaterial as the sensor surface has been investigated. The fabrication of a sensor surface with controlling the film thickness and adhesion is an important factor for the sensing response ability. The QCM-D sensor surfaces of gold, titanium, and poly(styrene) (PS) have been developed by q-sense Co. Ltd., but a bioceramic sensor has yet not been developed. To clarify the interactions between bioceramics and biomolecules, the development of a fabrication technique of the HAp surface similar to that in our body is of great importance. Furthermore, a reusability technique of the QCM-D sensor should be developed due to its expense. Surface modifications with HAp formations on bio-inert materials have been widely studied to provide a biocompatibility. For example, magnetron sputtering [70,71], beam sputtering [72], plasma spray [73,74], laser ablation [75], and electrophoretic deposition (EPD) [76,77] have been developed. However, sputtering and laser ablation have several issues such as target fabrication and expensive equipment costs. The films formed with the sputtering, laser ablation, and plasma spray have different chemical compositions and morphologies as compared with the HAp in our body because they are high-energy and high-temperature processes. On the other hand, the EPD of synthetic HAp nanocrystals on a gold surface has been reported [73,78e81], since the synthesized chemical composition and morphology can be retained during the formation. Figure 5(aec) show the EPD system setup, the film formation process of HAp nanocrystals, and the HAp sensor. HAp surfaces with the EPD method applicable for the QCM-D technique have already been fabricated [83e87]. As the specific fabrication method, a suspension of HAp nanocrystals synthesized using a wet chemical method [86] was centrifuged at 2000 g for 15 min, washed three times with ethanol, and ultrasonically dispersed in ethanol at 1 wt%. Before the EPD, the gold sensor surface

562

Handbook of Advanced Ceramics

FIGURE 5 (a) EPD system setup, (b) the film formation process of HAp nanocrystals (EPD and ultrasonication), and (c) the HAp sensor. (d, e): the AFM topographic images of (d) gold and (e) HAp sensors in a 1  1 mm2 area.

was cleaned by immersing it for 10 min in an ammonia and hydrogen peroxide mixture (APM) solution (weight ratio of 5:1:1 mixture of Milli-Q quality distilled water, H2O2, and NH3) at 70  C and then dried by blowing N2 gas. The cleaned surface was irradiated with UV light (lirr ¼ 254 and 185 nm; UV/Ozone, Bioforce Nanoscience Co., Ltd.) in air for 10 min. A direct current voltage was applied at 100 V/cm for 1 min as shown in Figure 5(a). Any surplus nanocrystals were removed by a 1-min ultrasonic treatment (28 kHz, 100 W) in ethanol as shown in Figure 5(b), and

the Hap sensor was successfully obtained as shown in Figure 5(c). As a result, the AFM topographic images of gold and HAp sensors in a 1  1 mm2 area were observed as shown in Figure 5(d) and (e). The weight change of the HAp nanocrystals under air was 4.0  0.2 mg/cm2 by QCM-D, and the thickness was calculated at 12.9  0.5 nm (based on the density of HAp as 3.14 g/cm3). The HAp nanocrystals deposited on the gold sensor had an RMS value of 4.2  0.8 nm, which was almost the same as the gold surface with the values of 0.8  0.4 nm.

Chapter | 7.2

Application of a Quartz Crystal Microbalance with Dissipation

The EPD mechanism has been described as follows. Alcohols are known to behave as a proton donor in the presence of bases, and pure alcohol can ionize in the following way: RCH2OH þ RCH2OH ¼ RCH2O þ RCH2OHþ 2 . Thus, the protonated alcohol is decomposed and leaves a proton on the surface, while the dissociated alcohol and alkoxide ion are desorbed into the solution. In ethanol, the ethoxide ion behaves as the counter anion, and then the HAp surface has a positive charge [80e82]. Based on the EPD method, the HAp nanocrystals with different crystallinities synthesized at 4  C, 24  C and 80  C were precipitated, resulting in a deposited HAp nanolayer with a thickness of 10e20 nm consisting of the different crystal sizes [85]. The crystalline sizes of d100 and d002 were 22 nm and 67 nm at 4  C, 28 nm and 88 nm at 24  C, and 54 nm and 158 nm at 80  C, respectively. The application of a DC bias of 10 and 50 V/cm resulted in inhomogeneous depositions with partly bare gold surfaces, and 100 V/cm of the DC bias ensured a homogeneous HAp covering over the gold surface. The density of the nanocrystals on the gold surface varied depending on the applied DC voltage, indicating the importance of the acceleration of the positively charged HAp nanocrystals to the cathode for controlling the deposition ability. Furthermore, the QCM-D measurement can reveal the real-time protein adsorption on the HAp sensor surfaces, and the adlayers were found to exhibit a flexible property [78,83,85,87,88]. The QCM-D sensor is very expensive, and its reusability should also be investigated for practical applications. The cleaning methods of the sensor for organic chemicals [89] or proteins [90] adsorbed on metal sensors involve detergents and monitoring by the QCM technique. For example, the amount of tripalmitin or dotriacontane desorbed from the gold sensors was 88 or 78 wt%, respectively, when using octaethylene glycol mono-n-dodecyl ether as the detergent [89], and the removal of Ab adsorbed on the titanium and chromium sensors is 90 wt% when using an ionic surfactant of sodium dodecyl sulfate (SDS) at 37 or 90  C [90]. On another HAp sensor (HAp/phosphate-terminated polymer/gold), the repeated adsorption and removal of a tea stain were studied using a flowing aqueous solution of sodium tripolyphosphate and QCM technique. A decrease in the frequency shift was observed during repeated testing [71]. Ionic surfactants, such as SDS, have been considered as candidates for removing chemicals adsorbed on bioceramics because of their strong exchange reactions; however, these surfactants were easily adsorbed on HAp [91e94]. Effective removal methods for HAp that do not damage or change the sensor surface properties and retain the reusability of the HAp sensors should be developed. Q-Sense Co. Ltd. recommends a combined APM and UV treatment for cleaning the QCM-D gold sensor surface and its reusability. The treatment with APM removes dust and nanoparticles from the SiO2/Si surface; it is a standard

563

treatment by the Radio Corporation of America [95]. The APM generates the hydroxyl radical (OH) through the reaction of H2O2 with H2O; the hydrogen peroxide radical (OOH) is then produced by the reaction of the OH radical with H2O2 [96e98], and these radical species have been used as a SiO2/Si etchant. The free-radical-mediated oxidation also leads to the hydroxylation of the amino groups of proteins, the conversion of amino acid residues to carbonyl derivatives, and the cleavage of polypeptide chains [99]. On the other hand, Bolon and Kunz [100] reported that exposure to ultraviolet light (UV) in an oxygen atmosphere leads to the depolymerization of photoresist polymers. The UV treatment has been used as an effective method of decomposing hydrocarbons on substrates [101,102]. The common light source has intense emission lines at 185 and 254 nm. The 185-nm photons induce the dissociation of O2 by generating singlet oxygen (O(1D)), which collides with inert N2 to generate triplet oxygen (O(3P)) and reacts with O2 to produce ozone (O3). The UV also decomposes O3 to O(1D), which cleaves organic bonds such CeC, CeH, and OeH [103,104]. The repeatability of the adsorption and removal of Fgn and fetal bovine serum (FBS) on HAp sensors using the APM and UV methods was investigated with an FT-IR spectroscopy and QCM-D monitoring [83]. Proteins adsorbed on the HAp sensors were removed by an ammonia/hydrogen peroxide mixture (APM), ultraviolet light (UV), UV/APM, APM/UV, and sodium dodecyl sulfate (SDS) treatments. The FTIR spectra of the reused surfaces revealed that the APM and SDS treatments left peptide fragments or proteins adsorbed on the surfaces, whereas the other methods successfully removed the proteins. The QCM-D measurements indicated that in the removal treatments, Fgn was slowly adsorbed in the first cycle because of the change in surface wettability revealed by contact angle measurements. The SDS treatment was not effective in removing proteins. The APM or UV treatment decreased the Df for the reused HAp sensors. The UV/APM treatment did not induce the Df, but decreased the DD. Therefore, the APM/UV treatment was found to be the most useful method for reproducing the protein adsorption behavior on HAp sensors.

3.2.3. Protein Adsorption Studies on the Hap 3.2.3.1. Mono-Component Adsorption It is well known that protein adsorption depends on the surface properties such as wettability, free energy, charge, and roughness [105,106]. The mono-component protein adsorption process and any conformational changes by the QCM-D technique have been vigorously investigated [110,111]. The ions in the solvent are known to adsorb on surfaces to form hydration structures, which influence the fundamental protein adsorption kinetics [18]. Thus, the studies analyzed for the adsorption depending on the

564

solvent have been described by the batch method to clarify the Ab adsorption on HAp in phosphate-buffered saline (PBS) [107], Ab adsorption on silicaetitania in PBS and 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid [108], and Ab, IgG, Fgn, and lysozyme adsorption on germanium in PBS or TriseHCl [109]. The conformation of the Ab adlayer on gold [112] and Fn adlayer on HAp [113] involving the binding of a monoclonal antibody were also discussed. However, in situ monitoring of the adsorption behavior and conformational change by the QCM-D technique on the bioceramics have not yet been reported. To understand the interfacial phenomena in vivo and in vitro, the various protein adsorption behaviors in body fluids should be precisely investigated. Only a few studies of the saturated DD/Df value (DDsat/ Dfsat) in the DDeDf plot from the measured Df and DD curves have been used for evaluating the adsorption behavior and the conformation [69,78,114,115]. Figure 6 shows the schemes of the DDeDf plots of different protein adlayers and the structures. Based on Eqn (8), the DDsat/ Dfsat value corresponds to the Edisspation per unit adsorption mass, indicating the viscoelastic property of the adlayer. The higher or lower DDsat/Dfsat value donates the viscous or

FIGURE 6 Schemes of (a) the DDeDf plots (plot 1, 2) with the different protein adlayers, and (b, c) the structures of the adlayers. The DD/Df value at the saturated stage (DDsat/Dfsat) in plot 2 is lower than that in plot 1, indicating the possible (b) viscous and (c) elastic structures for plots 1 and 2, respectively.

Handbook of Advanced Ceramics

elastic property of the adlayer, respectively. Plot 1 showed a higher DDsat/Dfsat value when compared with plot 2 in Figure 6(a), indicating the relatively loose structure and viscous property of plot 1 in Figure 6(b). Ozeki et al. revealed a significant relationship between the DDsat/Dfsat value and the hydrated amount [69]. Rodahl et al. revealed by defining the protein adlayer as a Newtonian liquid that the DDsat/Dfsat value was significantly related to the inverse of the friction coefficient between the adlayer and the sensor surface [114]. Monkawa et al. reported that the DDsat/Dfsat value of the Fgn adlayer successfully corresponds to the conformation with an additional FT-IR analysis [78]. Based on the evaluation, Yoshioka et al. discussed the conformation of various acidic and basic protein adlayers on HAp using the DDsat/Dfsat values [115]. Therefore, the DDsat/Dfsat value is of great importance for evaluating the structure of the protein adlayer. The detectable height (l) of D in the QCM-D system can be represented by Eqn (9) as follows [116]. rffiffiffiffiffiffiffiffi h (9) l ¼ prf where the h and r values are the viscosity and density on the sensor surface, respectively. Thus, the viscous liquid and adlayer on the sensor surface showed the higher l value. When 1/tan d is defined as c, the Df and DD values can be represented by Eqn (10) and Eqn (11), respectively [116]:   2d 2 c 1 (10) dad rad f 1 þ 2 ad 2 Df y rq d q 3l ð1 þ c Þ  3  2rad fdad 1 DDy (11) 3pf0 rq dq l2 ð1 þ c2 Þ where dad is the thickness of the adlayer. Thus, the DD/Df value is affected by only dad and tan d, and the adlayer of plot 1 also indicates the higher dad and/or tan d values as compared with plot 2. Based on the viewpoints and the previous reports on hydration [69] and friction [78], the model structure of the adlayer of plot 1 and 2 can be represented as shown in Figure 6(b) and (c), respectively. Therefore, the DDsat/Dfsat value is one of the excellent tools to evaluate the viscoelasticity and structure of the adlayer. Figure 7 shows the scheme of the secondary structure of Fgn. The Fgn, a structural glycoprotein in blood plasma with an isoelectric point at 5.5, is often employed for QCM experiments of protein adsorption [52,117e119] due to its moderate molecular weight (340 kDa) and size (45 nm length [120]). A central hydrophobic E domain is connected to two hydrophobic D domains with a coiled-coil chain. These hydrophobic domains are charged negative under neutral pH condition. The aC domains, with Arg and Lys residues, are charged positive and are substantially more hydrophilic than the E and D domains. On the other

Chapter | 7.2

Application of a Quartz Crystal Microbalance with Dissipation

FIGURE 7 Scheme of the secondary structure of Fgn. The detailed characteristics of the labeled domains are described in the text.

hand, Ab is globular with a pI at 4.7 and molecular weight (66.5 kDa), and has an asymmetric heart-like structure in which three main domains are divided into six subunit domains [121]. The protein surface with many carboxyl groups and 19 imidazole groups affects the hydrophilic property. The adsorption behavior depending on the characteristic protein morphology (fiber or globular) can be also detected by the QCM-D technique. Figure 8 shows the DD-Df plots and possible schemes of the conformational changes with the Fgn and Ab adsorptions on the HAp in PBS. The DDeDf plots in Figure 8(a, c) show the 2-step change for Fgn and linear change for Ab, indicating the two-step conformational change and monomolecular adsorption behavior on the HAp surface. The DDsat/Dfsat value of Ab on the HAp sensors was 1.0  10 8, which is much lower than that of

565

Fgn (34.5  10 8) [78,85]. Thus, the Ab was hardly absorbed on the HAp surface compared with the Fgn. The adsorption behavior of Fgn on HAp with its dumbbell-like structure has already been described [78,85]. One of the aC domains of Fgn, charged positive, is bonded to the negative sites of the HAp surface similar to phosphate and/or hydroxyl ions like in an “end-on” model as shown in Figure 8(b). On the contrary, the Ab has an asymmetric heart-like structure in which three main domains are divided into the six subunit domains already mentioned. The low adsorption amount of Ab as compared with that of Fgn indicates that the adsorption model of Ab could be “side-on” at the monolayer as shown in Figure 8(d), and the dissociated carboxyl and imidazole groups interact with the positively charged calcium ions on the HAp surface. Therefore, the different adsorption behavior of the Fgn and Ab with an almost similar pI value can be attributed to their secondary structure and realization of different adsorption models such as ‘‘side-on’’ and “end-on”. 3.2.3.2. Multicomponent Adsorption When biomaterials are implanted into the body, multiple proteins in the body fluids immediately and competitively adsorb on the surfaces [19e21]. Since the protein adlayers govern the biocompatibility, the proteinesurface interaction is important for designing biocompatible materials. From these points, the protein adsorption behaviors have been

FIGURE 8 (a, c) DD-Df plots and (b, d) the possible schemes of the conformational changes with the (a, b) Fgn and (c, d) Ab adsorption on the HAp surface. The Fgn adsorbs on the HAp with the changes from “side-on” to “end-on” vs. time, whereas the Ab adsorbs at the monolayer. (Reprinted with permission from [85], T. Ikoma et al, J Am Ceram Soc 2009; 92:1125e8. Ó 2009 The American Ceramic Society.)

566

widely investigated [127e130], but the complex reaction at the interface including multiple proteins is still controversial. The “Vroman effect” is a phenomenon in which the adsorbed Fgn on surfaces is replaced with other abundant proteins in serum [124]. Multiple-protein adsorption for Ab and IgG labeled with radioactive 125I and 131I [131,132] and surface plasmon resonance for blood plasma solution [133] have been described in recent studies. Protein adsorption mechanism on a biomaterial surface can be divided into the (i)e(iii) processes vs. time: (i) The hydrated protein in the liquid interacts with the hydration layer on the surface, and the adsorption on the surface occurs with disruption of the hydration structure [18]. (ii) The protein repeatedly adsorbs and desorbs on the surface to thermodynamically stabilize the structure [122]. During the adsorption of multicomponent proteins, the adsorbed proteins simultaneously exchange with other proteins [123,124,126]. (iii) The adsorbed proteins change their conformation to become stable higher-order structures on the surface, and the stabilization indicates an equilibrium state [125]. During the adsorption of multicomponent proteins, other proteins may subsequently co-adsorb on the stabilized protein surfaces. These interfacial changes are very important for understanding the phenomena in body fluids, but only a few studies of the adsorption behavior and conformational changes have already been reported [127e130]. Therefore, the QCM-D technique is one of the suitable techniques for in situ analyzing the interfacial phenomena with the multicomponent protein adsorption. In culture of cells in vitro, a culture medium is added for the cell’s nature. The medium contains FBS of the basic components, such as hormones and growth factors. The FBS is blood containing many kinds of proteins without cells, platelets, and clotting factors. The two major groups of proteins are Ab and globulin. Ab is the largest mass fractional protein component in the blood and is known to eliminate cell attachment and block nonspecific binding [134,135]. The other proteins in the blood are Fgn, Fn, collagen (Col), and other subtle trace proteins (osteopontin, laminin, and vitronectin). These proteins are obligate adhesive proteins for integrin-receptor-based cell attachment and proliferation on surfaces, and are important components of the ECM. Grainger et al. investigated the Fn conformation adsorbed on polytetrafluoroethylene from a double-component FneAb mixture solution using an Fn monoclonal antibody [34]. The antibody binding to the compound adlayer was suppressed by the existence of Ab to clarify Ab masking of the adsorbed Fn. Thus, the ratio of the nonadhesive (e.g. Ab) to adhesive (e.g. Fn) proteins from a multicomponent solution (e.g. FBS) selectively adsorbed on the bioceramics is the key parameter to understand and improve cell attachment on the surfaces. The adsorption behavior of FBS on HAp was analyzed in situ in PBS, minimum essential media (aMEM or DMEM),

Handbook of Advanced Ceramics

and different carbonate-concentrated solutions using the QCM-D technique [87,136]. Although the adsorption behaviors of FBS on gold showed no dependence on the solvents, those on HAp clearly depended on the adsorbed ions in the solvents. Figure 9(a) shows the DDeDf plots of the FBS adsorption on the HAp sensor in PBS, aMEM, and DMEM. The adsorption amounts of FBS in PBS, aMEM, and DMEM after 60 min were 1.57  0.05, 0.79  0.06, and 0.73  0.06 mg/cm2 on the HAp, respectively. The DDsat/Dfsat values after 60 min were e8.8  3.3  10 8, e4.1  1.7 10 8, and e3.2  1.8 10 8 on the HAp, respectively. The adsorption amount and viscoelastic property on the HAp in PBS are approximately two times larger than those in the MEMs. The pre-adsorptive ions in the solvents on the HAp and components of the surfaces induce different adsorption behaviors. Acidic proteins with a negative charge at neutral pH of the buffers would preferentially adsorb on the HAp surface. Figure 9(b) shows the scheme of the FBS adlayers on the HAp sensor in the different solvents. The structure of FBS adlayer on the HAp depended on the solvent. The adsorption of carbonate ions from the MEMs on HAp caused a decrease in the adsorption amount of FBS and degradation of the viscoelastic property of the FBS adlayer. Thus, the carbonate ions adsorbed on the HAp surface strongly affect the FBS adsorption. As a consequence, the adsorption amounts of carbonate ions on the HAp in the carbonate solutions increased with increasing carbonate concentrations, which inhibited the FBS adsorption and decreased the viscoelastic property of the adlayers. Vroman et al. described the exchange adsorption phenomenon of proteins on a biomaterial: the order of the exchange adsorption was related with the molecular weight of the proteins [124]. The exchange adsorption from the multiple proteins has not been clearly detected based on a change in the DDeDf plots by the QCM-D technique; however, the DDsat/Dfsat values of the FBS adsorption were apparently different from those of a single-protein adsorption. It is speculated that the multiple proteins, such as Ab, IgG, etc., are simultaneously adsorbed on the HAp and the protein species adsorbed on the HAp found to depend on the preadsorbed ions and the solvent components. 3.2.3.3. Voigt-Based Viscoelastic Property of Protein Adlayers Figure 10 shows the schemes of the geometry and parameters used to simulate the quartz crystal covered with a viscoelastic protein adlayer, and the Voigt-based viscoelastic model (infinite spring and dashpots model). Voinova et al. reported that the measured Df and DD curves by the QCM-D technique have been fitted by a Voigt-based viscoelastic model to characterize the viscoelastic properties of the adlayers as a Newtonian fluid as shown in Figure 10(a) [116,126,137]. The Voigt-model is assumed to be a linear spring in parallel to

Chapter | 7.2

Application of a Quartz Crystal Microbalance with Dissipation

567

FIGURE 9 (a) DDeDf plots of FBS adsorption on HAp sensors in different solvents (PBS, aMEM and DMEM) for 1 h. The arrows indicate the saturation direction in the plots. (b) possible schemes of FBS adlayer on HAp at 1 h. The adsorption amount of the proteins is in the order 10 mM carbonate solution > PBS > MEMs (aMEM and DMEM) and 120 mM carbonate solution. The DDsat/Dfsat value is in the order PBS > MEMs > 10 mM carbonate solution and 120 mM carbonate solution. (Reprinted with permission from [136], M. Tagaya et al, Mater Sci Eng B 2010;173:176e81. Ó 2010 Elsevier B.V.)

(a)

(b)

FIGURE 10 (a) Scheme of the geometry and parameters used to simulate the quartz crystal covered with a viscoelastic protein adlayer having the thickness dad in contact between the sensor surface and a semi-infinite Newtonian liquid. The adlayer is represented by mad, had, and rad. The bulk liquid is represented by rl , hl. (b) Scheme of Voigt-based viscoelastic model with a parallel connection between the dashpot and spring. (Reprinted with permission from [126], F. Ho¨o¨k, et al, Anal Chem 2001;73:5796e804. Ó 2001 American Chemical Society.)

the dashpot for the load shaft to simulate the viscoelastic behavior as shown in Figure 10(b). The viscoelastic property is represented by a complex shear modulus G* given by Eqn (12) based on the Voigt-based model: G ¼ G0 þ iG00 ¼ mad þ i2pf had

(12)

where G0 is the real part of G* (storage modulus), G00 is the imaginary part of G* (loss modulus), f is the oscillation frequency, mad is the elastic shear modulus, and had is the shear viscosity. The viscoelastic parameters, such as mad, had, density (rad), and thickness (dad) of the adlayer, were determined using the density (rl) and viscosity (hl) of the

568

Handbook of Advanced Ceramics

bulk liquid. The rl and hl values are fixed. The adlayer is situated between the QCM electrode and a semi-infinite Newtonian liquid under no-slip conditions. The changes in the Df and DD are fitted by the b function as follows: Df ¼ ImðbÞ=2ptq rq DD ¼

ReðbÞ=pftq rq

(13) (14)

where b is represented by Eqn (15) with a, x1, mad, had and dad and x2, and x1 represented by Eqn (16) with mad, had, and rad as follows: b ¼ x1

2pf had imad 1 a expð2x1 dad Þ 2pf 1 þ a expð2x1 dad Þ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð2pf Þ2 rad x1 ¼ mad þ i2phad

(15)

(16)

where a is represented by Eqn (17) with x1, x2, mad, had, and hl, and x2 is represented by Eqn (18) with rl and hl, as follows: x1 2pf had imad þ1 x 2pf hl a ¼ 2 x1 2pf had imad 1 x2 2pf hl sffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2pf rl x2 ¼ i hl

(17)

(18)

The ratio of G00 and G0 can be calculated as a loss tangent delta (G00 / G0 ) to evaluate the viscoelasticity of the adlayer as shown in Eqn (19): tan d ¼ G00 =G0

(19)

The tan d < 1, tan d ¼ 1, and tan d > 1 of the adlayer can be defined as elastic, viscoelastic, and viscous properties, respectively. In the energy dissipation process, the molecular structure of the adlayer changes with the oscillation (mechanical energy) to enhance the potential energy, which is an energy storage process. Next, the intramolecular and/ or intermolecular interactions are not fixed to generate the sliding friction from the adlayer. This indicates that the elastic energy converts to molecular kinetic energy (thermal energy), which is a loss energy process. The elastic adlayer (tan d < 1) induces the changed molecular structures with the mechanical energy that effectively returns to the original structure with a lower energy state. On the other hand, the viscous adlayer (tan d > 1) induces the resistance force with molecular movements that effectively generates and converts to thermal energy. Based on the Voigt-model, only a few studies of the viscoelastic property of the adlayer have been investigated, and the Ab [138] and laminin [139] adlayers on gold and Col (type I from calf skin) adlayer on poly(styrene)

(PS) and oxidized PS (PSox) [140] have been discussed. The Ab had a mad of 100 kPa, had of 6 mPa$s, and tan d of 1.8 [138], whereas those of the laminin were a mad of 7.6  1.9 kPa, had of 1.83  0.11 mPa$s, and tan d of 7.7 [139]. Thus, the laminin, which is the ECM protein, clearly shows a more viscous structure on the gold based on the tan d values. The Col shows a mad of 125  25 kPa, had of 4.0  1.0 mPa$s, and tan d of 1.15  0.05 on PS and a mad of 20  10 kPa, had of 2.5  0.5 mPa$s, and tan d of 4.7  1.5 on PSox [140]. Thus, the surface property also affects the viscoelastic property of the adlayer. The viscoelasticity of the adlayer on the bioceramics have not yet been reported. As compared with the viscoelastic property on the bioceramics, the evaluation of interfacial phenomena based on the Voigt-based viscoelastic model would be an effective tool.

3.2.4. Cell Adhesion Studies on the Hap The adhesion behaviors of cells depend on the surface properties, such as topography, wettability, and charge, and on the protein adlayers [19e21]. The conformational change, denaturation, and RGD sequence of the adsorbed proteins on the surfaces govern the biocompatibility, which is attributed to the adhesion, proliferation, migration, and differentiation of cells [18,23]. The ECM with the RGD sequence affects cell adhesion [24]. The integrin of a cell binds to the RGD sequence in the ECM [25], and subsequent actin cytoskeletons are produced and the associated proteins form focal adhesion points on the surfaces [26]. With the interfacial reactions and morphological changes in Figure 1, the cells adhere and spread on the material surfaces. Thus, in situ monitoring of these interfacial phenomena with the initial cell behaviors on the surfaces is of great importance for controlling cell functions. The protein adsorption and subsequent cell adhesion on the surfaces were reviewed by Anselme [21]. The adsorption of different amounts of FBS on nanophase ceramics caused different cell adhesion behavior [141]. The proliferation and mineralization of osteoblasts on a HAp sintered body with and without pre-adsorbed type I Col were investigated, revealing various phenotype and gene expression patterns that were different from those on PS dishes [28,29]. If the cell behavior is determined by the adlayers of the type I Col, the gene expression should be the same for coatings of the type I Col on the HAp and PS. Therefore, these results suggested that the cell functions were determined not only by the interfacial proteins but also by the substrate surface properties. The understanding of cell adhesion and spreading process on the interfacial protein layers adsorbed on a substrate surface was also of great importance. Various physical parameters were measured to evaluate the interfacial phenomena with the initial cell adhesion,

Chapter | 7.2

Application of a Quartz Crystal Microbalance with Dissipation

569

FIGURE 11 Scheme of the interfacial detectable height (l) between the protein-modified surface and cell with the QCM-D technique.

such as resistance [68,142e147], impedance [148,149], transient decay time, maximum oscillation amplitude [150], and rheometer [163]. Li et al. recently described the viscoelastic properties of a fibroblast cell monolayer on gold using a thickness shear mode quartz crystal resonator with a transfer matrix model [162]. They calculated the mad values of 21e39 kPa, had values of 0.92e1.56 mPa$s, and tan d values of 1.2e2.3 at a 5-MHz resonance frequency. Ferna´ndez et al. investigated the viscoelastic properties of a fibroblast cell monolayer with a rheometer and obtained G0 , G00 , and tan d values at 10 Hz of 400 Pa, 150 Pa, and 0.3, respectively, indicating that the cell monolayer was an elastic body [163]. Palmer et al. suggested that the elastic shear modulus dominates the viscoelastic property of cells due to their rigid cytoskeletons at lower frequencies, but that the viscous modulus dominates the property due to the cytoplasmic fluid at higher frequencies [164]. The actin networks at the high frequency of 1 MHz showed a liquidlike property and those at the low frequency of 10 Hz showed an elastic property. The detected viscoelasticity was found to be attributed to the measurable surface region that depends on the excitation frequency. Thus, the QCM’s high frequency caused the viscosity behavior of the adherent cells due to the dependence of the actin network property on the frequency. The different viscoelastic properties of the adherent cells on the material surfaces could be attributed to the different actin network structures close to the surfaces. The QCM-D technique based on the energy dissipation shift [65,67,126,151] has allowed the monitoring of the cell adhesion and spreading behaviors on PSox [152e154,156,157], tantalum (Ta) [155e157], chromium (Cr) [155], titanium (Ti), and steel [158]. The QCM-D technique has also been used to understand the effect of the pre-adsorption of proteins on the cell adhesion [155e158], for example, on Ta and Cr with and without the pre-adsorption of FBS [155] and on Ti, TiO2, and steel with the pre-adsorption of Fn or Fgn [158]. The viscoelastic property of the interface with the cell adhesion on the HAp using the QCM-D technique should be investigated. Eqn (9) indicates that a higher h value for the adlayer can detect a higher l region. The l of a resonating wave in a culture medium at 37  C with a r of 0.993 g/cm3 and an h of 0.692 mPa$s [148] can be calculated to be ca. 100 nm at 15 MHz. The FBS adlayer with the viscoelastic values

measured shows an l of 100e200 nm at 15 MHz. The l of the adherent cell layers with the measured viscoelastic values is also 100e200 nm at 15 MHz. The values of l were extremely lower than the actual height of cells, which was on the order of micrometers as shown in Figure 11. This suggested that the QCM-D measured the lower parts of the cells close to the protein adlayer on the sensor surfaces. Therefore, the interface between the material surface and cell by the QCM-D effectively causes the differences in the viscoelastic properties of the adherent cells. In situ monitoring of the pre-adsorption of proteins and subsequent cell adhesion on HAp as compared with that on other surfaces should be investigated. HAp has a good biocompatibility for fibroblasts; downgrowth of the percutaneous device [165] and catheter [166] is improved by using an HAp sintered body and coatings. The biocompatible features are attributed to the surface property of HAp. The pre-adsorption of proteins and initial subsequent cell adhesion process on HAp using QCM-D and confocal laser scanning microscopy (CLSM) techniques have been reported [159e161]. The adhesion and morphological changes of fibroblast NIH3T3 cells on the HAp and PSox sensors with pre-adsorbed FBS have been analyzed [159]. The Df and DD curves on the HAp showed a decrease in Df with increasing DD for 80 min and the subsequent increase in Df with decreasing DD, while those on PSox showed a decrease in Df for 120 min with increasing DD for 50 min and then with subsequent decreasing DD. The cells on HAp had rough fibrous pseudopods and those on PSox had dense particulate pseudopods. These different adhesion behaviors dependent on the surface property were attributed to the cellesurface interactions. The pre-adsorption of three proteins (Ab, Fn, or Col) and subsequent adsorption of FBS and adhesion of fibroblasts have been analyzed [161]. Figure 12 shows the DDeDf plots of the fibroblasts on the HAp modified with the proteins for 2 h, and the tan d values and the aspect ratio of the adherent cells on the surfaces at 2 h. Figure 12(aed) show the DDeDf plots of the cells on the HAp modified with the FBS, FBS-Ab, FBS-Fn, and FBS-Col for 2 h. The Col-modified surface clearly showed an increase in Df with a decrease in DD, while the other surfaces showed a decrease in Df with an increase in DD and a subsequent increase in Df with a decrease in DD that occurred at 75 min for FBS, at 100 min for FBS-Ab and at 60 min for FBS-Fn.

570

Handbook of Advanced Ceramics

FIGURE 12 DDeDf plots of the cell adhesion on the HAp modified with (a) FBS, (b) FBS-Ab, (c) FBS-Fn and (d) FBS-Col for 2 h, and (e) the tan d values and (f) aspect ratio of the adherent cells on the surfaces cultured at 2 h. The interfacial tan d of the cells was evaluated by the Void-based viscoelastic model. The aspect ratio was observed by a CLSM, and was obtained by the wide width divided by the narrow width. The statistical analysis was examined by the Student’s t test. The N. S., * and ** denote no significant difference, P < 0.05 and P < 0.01. (Reprinted with permission from [161], M. Tagaya et al, IOP Conference Series: Mater Sci Eng 2011;18:192009e12. Ó 2011 IOP Publishing.)

The changes on the modified surfaces with ECM proteins (FBS-Fn and FBS-Col) indicate the interfacial cellesurface reaction such as binding to adhesion points during the fast stage. The Df and DD values at 2 h were e10.3 Hz and þ13  10 6 on FBS, e31.7 Hz and þ21  10 6 on FBS-Ab, e3.4 Hz and þ1.2  10 6 on FBS-Fn, and þ54.8 Hz and e13  10 6 on FBS-Col. The saturated DDsat/Dfsat values were 0.51, 0.14, 0.82, and 0.54 on the FBS, FBS-Ab, FBS-Fn, and FBS-Col, respectively, indicating the higher value of the ECM proteins (FBS-Fn and FBS-Col) with the

binding between the modified surfaces and the cells. These results indicated that the different adhesion processes occur depending on the modified surfaces for 2 h, and the pre-adsorption of Col and Fn effectively affects the celle surface interactions. Furthermore, the different interfacial tan d and the morphology observed by the CLSM also depended on the surfaces as shown in Figure 12(e, f). The tan d values were calculated from the viscoelastic parameters based on the Voigt-based viscoelastic model, and were evaluated within

Chapter | 7.2

Application of a Quartz Crystal Microbalance with Dissipation

the detectable height of the D value [137]. The viscoelasticity (tan d) and the spreading degree (aspect ratio) of the adherent cells on the surfaces modified with the ECM proteins showed higher values. The adherent cells on the surfaces modified with Col had expanded pseudopods, while those modified with Ab had round shape. This indicated that the interfacial viscoelasticity and the morphology clearly depended on the consisted proteins: the interfacial layer on the pre-adsorbed ECM proteins showed loose and flexible structures and uniaxially expanded morphologies. Therefore, the rearrangements of ECM and cytoskeleton changes at the interfaces would cause the different cellesurface interactions. The different cell adhesion process, viscoelasticity, and morphology depending on the surfaces were successfully in situ monitored and evaluated by the QCM-D and CLSM techniques.

[7]

[8]

[9]

[10]

[11]

[12]

4. SUMMARY [13]

In situ monitoring of the interfacial layers formed between bioceramics and cells is crucial for controlling cell functions and understanding biocompatible phenomena. To detect and clarify the structural characteristics at the interface, a QCM-D technique is one of excellent in situ analytical methods. The QCM-D detecting the interfacial phenomena with protein adsorption and initial cell adhesion has been investigated. In the future perspectives for scientific challenges and opportunities, the exploration of the mutual interaction should be studied by developing novel detection techniques for the interface. Furthermore, superior bioceramics applicable for medical implants, biosensors and biochips for diagnostics, tissue engineering, bioelectronics, and biomimetic materials will be designed.

REFERENCES [1] Kay MI, Young RA, Posner AS. Crystal structure of hydroxyapatite. Nature 1964;204:1050e2. [2] Suchanek W, Yoshimura M. Processing and properties of hydroxyapatite-based biomaterials for use as hard tissue replacement implants. J Mater Res 1998;13:94e117. [3] Yongli C, Xiufang Z, Yandao G, Nanming Z, Tingying Z, Xinqi S. Conformational changes of fibrinogen adsorption onto hydroxyapatite and titanium oxide nanoparticles. J Colloid Interface Sci 1999;214:38e45. [4] Kandori K, Miyagawa K, Ishikawa T. Adsorption of immunogamma globulin onto various synthetic calcium hydroxyapatite particles. J Colloid Interface Sci 2004;273:406e13. [5] Kikuchi M, Itoh S, Ichinose S, Shinomiya K, Tanaka J. Selforganization mechanism in a bone-like hydroxyapatite/collagen nanocomposite synthesized in vivo and its biological reaction in vivo. Biomaterials 2001;22:1705e11. [6] Letic-Gavrilovic L, Piattelli A, Abe K. Nerve growth factor beta (NGF beta) delivery via a collagen/hydroxyapatite (Col/HAp)

[14]

[15]

[16]

[17]

[18] [19]

[20]

[21] [22]

[23]

[24]

571

composite and its effects on new bone in growth. J Mater Sci Mater Med 2003;14:95e102. Liao SS, Cui FZ. In vitro and in vivo degradation of mineralized collagen-based composite scaffold: nanohydroxyapatite/collagen/ poly(L-lactide). Tissue Eng 2004;10:73e80. Yunoki S, Ikoma T, Monkawa A, Ohta K, Kikuchi M, Sotome S, et al. Control of pore structure and mechanical property in hydroxyapatite/collagen composite using unidirectional ice growth. Mater Lett 2006;60:999e1002. Kadoya T, Ogawa T, Kuwahara H, Okuyama T. High performance liquid chromatography of proteins on a hydroxyapatite column. J Liq Chromatogr 1988;11:2951e67. Kadoya T. High-performance liquid chromatography of proteins on a ceramic hydroxyapatite with volatile buffers. J Chromatogr 1990;515:521e5. Itokazu M, Sugiyama T, Ohno T, Wada E, Katagiri Y. Development of porous apatite ceramic for local delivery of chemotherapeutic agents. J Biomed Mater Res 1998;39:536e8. Paul W, Sharma CP. Ceramic drug delivery: a perspective. J Biomater Appl 2003;17:253e64. Kim HW, Knowles JC, Kim HE. Porous scaffolds of gelatine hydroxyapatite nanocomposites obtained by biomimetic approach: characterization and antibiotic drug release. J Biomed Mater Res B 2005;74B:686e98. Mizushima Y, Ikoma T, Tanaka J, Hoshi K, Ishihara T, Ogawa Y, et al. Injectable porous hydroxyapatite microparticles as a new carrier for protein and lipophilic drugs. J Control Release 2006;110:260e5. Ikoma T, Tonegawa T, Watanabe H, Chen GP, Tanaka J, Mizushima Y. Drug-supported microparticle of calcium carbonate nanocrystals and its covering with hydroxyapatite. J Nanosci Nanotech 2007;7:822e7. Nemoto R, Wang L, Ikoma T, Tanaka J, Senna M. Preferential alignment of hydroxyapatite crystallites in nanocomposites with chemically disintegrated silk fibroin. J Nanoparticle Res 2004;6:259e65. Matsuda A, Ikoma T, Kobayashi H, Tanaka J. Preparation and mechanical property of core-shell type chitosan/calcium phosphate composite fiber. Mater Sci Eng C 2004;24:723e8. Kasemo B. Biological surface science. Surf Sci 2002;500: 656e77. El-Ghannam A, Ducheyne P, Shapiro IM. The effect of serum protein adsorption on osteoblast adhesion to bioactive glass and hydroxyapatite. J Orthop Res 1999;17:340e5. Rouahi M, Champion E, Gallet O, Jada A, Anselme K. Physicochemical characteristics and protein adsorption potential of hydroxyapatite particles: influence on in vitro biocompatibility of ceramics after sintering. Colloids Surf B: Biointerfaces 2006;47:10e9. Aneslme K. Osteoblast adhesion on biomaterials. Biomaterials 2000;21:667e81. Pierschbacher MD, Ruoslahti E. Cell attachment activity of fibronectin can be duplicated by small synthetic fragments of the molecule. Nature 1984;309:30e3. Barrere F, Mahmood T, Degroot K, Vanblitterswijk C. Advanced biomaterials for skeletal tissue regeneration: instructive and smart functions. Mater Sci Eng R 2008;59:38e71. Nath N, Hyun J, Ma H, Chilkoti A. Surface engineering strategies for control of protein and cell interactions. Surf Sci 2004;570:98e110.

572

[25] Steele JG, Johnson G, Underwood PA. Role of serum vitronectin and fibronectin in adhesion of fibroblasts following seeding onto tissue culture polystyrene. J Biomed Mater Res 1992;26:681e4. [26] Huang SJ, Ingber DE. Shape-dependent control of cell growth, differentiation, and apoptosis: switching between attractors in cell regulatory networks. Exp Cell Res 2000;261:91e103. [27] Chen CS, Mrksich M, Huang S, Whitesides GM, Ingber DE. Geometric control of cell life and death. Science 1997; 276:1425e8. [28] Hanagata N, Takemura T, Monkawa A, Ikoma T, Tanaka J. Pre-adsorbed type-I collagen structure-dependent changes in osteoblastic phenotype. Biochem Biophy Res Commn 2006; 344:1234e40. [29] Hanagata N, Takemura T, Monkawa A, Ikoma T, Tanaka J. Phenotype and gene expression pattern of osteoblast-like cells cultured on polystyrene and hydroxyapatite with pre-adsorbed type-I collagen. J Biomed Mater Res A 2007;83A:362e71. [30] Ruoslahti E, Pierschbacher MD. New perspectives in cell adhesion: RGD and integrins. Science 1987;238:491e7. [31] Arnold M, Cavalcanti-Adam EA, Glass R, Bluemmel J, Eck W, Kantlehner M, et al. Activation of integrin function by nanopatterned adhesive interface. Chem Phys Chem 2004;5:383e8. [32] Michael KE, Vernekar VN, Keselowsky BG, Meredith JC, Latour RA, Garcia AJ. Adsorption-induced conformational changes in fibronectin due to interactions with well-defined surface chemistries. Langmuir 2003;19:8033e40. [33] Keselowsky BG, Collard DM, Garcia AJ. Surface chemistry modulates fibronectin conformation and directs integrin binding and specificity to control cell adhesion. J Biomed Mater Res Part A 2003;66A:247e59. [34] Grainger DW, Pavon-Djavid G, Migonney V, Josefowicz M. Assessment of fibronectin conformation adsorbed to polytetrafluoroethylene surfaces from serum protein mixtures and correlation to support of cell attachment in culture. J Biomater Sci Polym Ed 2003;14:973e88. [35] Kefalides NA, Winzler RJ. The chemistry of glomerular basement membrane and its relation to collagen. Biochem 1966;5:702e13. [36] Kefalides NA, Denduchis B. Structural components of epithelial and endothelial basement membranes. Biochem 1969;8:4613e21. [37] Spiro RJ. Studies on the renal glomerular basement membrane: preparation and chemical composition. J Biol Chem 1967;242: 1915e22. J Biol Chem 242:1915e1922. [38] Pei RJ, Cui XQ, Yang XR, Wang EKR. Real-time immunoassay of antibody activity in serum by surface plasmon resonance biosensor. Talanta 2000;53:481e8. [39] Asanov AN, Wilson WW, Odham PB. Regenerable biosensor platform: A total internal reflection fluorescence cell with electrochemical control. Anal Chem 1998;70:1156e63. [40] Lassen B, Malmsten M. Competitive protein adsorption studied with TIRF and ellipsometry. J Colloid Interface Sci 1996; 179:470e7. [41] Cross GH, Reeves A, Brand S, Swann MJ, Peel LL, Freeman NJ, et al. The metrics of surface adsorbed small molecules on the Young’s fringe dual-slab waveguide interferometer more options. J Phys D: Appl Phys 2004;37:74e80. [42] Freeman NJ, Peel LL, Swann MJ, Cross GH, Reeves A, Brand S, et al. Real time, high resolution studies of protein

Handbook of Advanced Ceramics

[43]

[44]

[45]

[46]

[47]

[48]

[49]

[50] [51]

[52]

[53]

[54] [55]

[56] [57] [58]

[59] [60]

[61]

adsorption and structure at the solid-liquid interface using dual polarization interferometry. J Phys Cond Matt 2004;16: S2493e6. Goobes R, Goobes G, Campbell CT, Stayton PS. Thermodynamics of statherin adsorption onto hydroxyapatite more options. Biochemistry 2006;45:5576e86. Pollitt MJ, Buckton G, Brocchini S, Alpar HO. Calorimetric study of bovine serum albumin dilution and adsorption onto polystyrene particles. Int J Pharm 2005;298:333e8. Azzam RMA, Rigby PG, Krueger JA. Kinetics of protein adsorption and immunological reactions at a liquid-solid interface by ellipsometry. Phys Med Biol 1977;22:422e30. Seitz R, Brings R, Geiger R. Protein adsorption on solid-liquid interfaces monitored by laser-ellipsometry. Appl Surf Sci 2005;252:154e7. Ong JL, Chittur KK, Lucas LC. Dissolution/repricipitation and protein adsorption studies of calcium phosphate coatings by FTIR/ATR techniques. J Biomed Mater Res 1994;28:1337e46. Zeng H, Chittur KK, Lacefield WR. Analysis of bovine serum albumin adsorption on calcium phosphate and titanium surfaces. Mater Sci 1999;20:377e84. Nomura T, Okuhara M. Frequency shifts of piezoelectric quartz crystals immersed in organic liquids. Anal Chim Acta 1982;142:281e4. Ebersole RC, Foss RP, Ward MD. Piezoelectric cell growth sensor. Bio-Technol 1991;9:450e4. Edvardsson M, Svedhem S, Wang G, Richter R, Rodahl M, Kasemo B. QCM-D and reflectometry instrument: applications to supported lipid structures and their biomolecular interactions. Anal Chem 2009;81:349e61. Ho¨o¨k F, Voros J, Rodahl M, Kurrat R, Boni P, Ramsden JJ, et al. A comparative study of protein adsorption on titanium oxide surfaces using in situ elipsometry, optical waveguide lightmode spectroscopy, and quartz crystal microbalance/dissipation. Colloid Surf B: Biointerfaces 2002;24:155e70. Berglin M, Pinori E, Sellborn A, Andersson M, Hulander M, Elwing H. Fibrinogen adsorption and conformational change on model polymers: novel aspects of mutual molecular rearrangement. Langmuir 2009;25:5602e8. Curie J, Curie P. Piezoelectric and allied phenomena in Rochelle salt. Comput Rend Acad Sci Paris 1880;91:294e7. Rayleigh L. On the free vibrations of an infinite plate of homogeneous isotropic elastic matter. Proc London Math Soc 1889;20:225e6. Cady WG. The piezoelectric resonator. Phys Rev 1921; 17:531e3. Lack FR, Willard GW, Fair IE. Some improvements in quartz crystal circuit elements. Bell Syst Technol 1934;13:453e5. O’Sullivan CK, Guilbault GG. Commercial quartz crystal microbalances: theory and applications. Biosens Bioelectron 1999;14:663e70. Janshoff A, Galla HJ, Steinem C. The quartz- crystal microbalance in life science. Angew Chem Int Ed 2000;39:4004e32. Pavey KD. Quartz crystal analytical sensors: the future of label-free, real-time diagnostics. Expert Rev Mol Diagn 2002; 2:173e86. Sauerbrey G. Quartz crystal microbalance theory and calibration. Z Phys 1959;155:206e22.

Chapter | 7.2

Application of a Quartz Crystal Microbalance with Dissipation

[62] Konash PL, Bastiaans GJ. Piezoelectric crystals as detectors in liquid chromatography. Anal Chem 1980;52:1929e31. [63] Landau LD, Lifshtz EM. Fluid mechanics. Oxford, England: Pergamon; 1959. p. 88e90. [64] Kanazawa KK, Gordon IIJG. The oscillation frequency of a quartz resonator in contact with a liquid. Anal Chim Acta 1985;175:99e105. [65] Rodahl M, Ho¨o¨k F, Krozer A, Kasemo B, Breszinsky P. A quartz crystal microbalance setup for frequency and Q-factor measurements in gaseous and liquid environments. Rev Sci Instrum 1995;66:3924e30. [66] Ho¨o¨k F, Rodahl M, Brzezinski P, Kasemo B. Measurements using the quartz crystal microbalance technique of ferritin monolayers on methyl-thiolated gold: dependence of energy dissipation and saturation coverage on salt concentration. J Colloid Interface Sci 1998;208:63e7. [67] Rodahl M, Kasemo B. A simple setup to simultaneously measure the resonant frequency and the absolute dissipation factor of a quartz crystal microbalance. Rev Sci Instrum 1996;67:3238e41. [68] Muramatsu H, Tamiya E, Karube I. Computation of equivalent circuit parameters of quartz crystals in contact with liquids and study of liquid properties. Anal Chem 1988;60:2142e6. [69] Ozeki T, Morita M, Yoshimine H, Furusawa H, Okahata Y. Hydration and energy dissipation measurements of biomolecules on a piezoelectric quartz oscillator by admittance analyses. Anal Chem 2007;79:79e88. [70] Tsumagari A, Nakamura A, Hashimoto K, Tada Y, Yamashita Y. Protein adsorption onto hydroxyapatite thin films with QCM measurements. Bioceramics 1999;12:361e4. [71] Langford J, Pavey KD, Olliff CJ, Cragg PJ, Hanlon GW, et al. Real-time monitoring of stain formation and removal on calcium hydroxyapatite surfaces using quartz crystal sensor technology. Analyst 2002;127:360e7. [72] Zhitomirsky L, Galor L. Electrophoretic deposition of hydroxyapatite. J Mater Sci-Mater Med 1997;8:213e9. [73] Wei M, Ruys AJ, Milthorpe BK, Sorrell CC. Precipitation of hydroxyapatite nanoparticles: effects of precipitation method on electrophoretic deposition. J Mater Sci-Mater M 2005; 16:319e3124. [74] Sun L, Berndt CC, Gross KA, Kucuk A. Material fundamentals and clinical performance of plasma-sprayed hydroxyapatite coatings: a review. J Biomed Mater Res Appl Biomater 2001;58:570e92. [75] Cheang P, Khor KA. Influence of powder characteristics on plasma sprayed hydroxyapatite coatings. Therm Spray Technol 1996;5:310e6. [76] Wang C, Chen Z, Guan L, Liu Z, Wang P, Zheng S, et al. Structural characterization of ion beam sputter deposited calcium phosphate coatings. Surf Coat Technol 2000;130:39e45. [77] Cleries L, Martinez E, Fernandez-Pradas JM, Sardin G, Esteve J, Morenza JL. Mechanical properties of calcium phosphate coatings deposited by laser ablation. Biomaterials 2000;21:967e71. [78] Monkawa A, Ikoma T, Yunoki S, Yoshioka T, Tanaka J, Chakarov D, et al. Fabrication of hydroxyapatite ultra-thin layer on gold surface and its application for quartz crystal microbalance technique. Biomaterials 2006;27:5748e54. [79] Ikoma T, Tagaya M, Tonegawa T, Takemura T, Okuda M, Hanagata N, et al. The surface property of hydroxyapatite:

[80]

[81]

[82]

[83]

[84]

[85]

[86]

[87]

[88]

[89]

[90]

[91]

[92]

[93]

[94]

[95]

573

sensing with quartz crystal microbalance. Key Eng Mater 2009;396e398:89e92. Javidi M, Bahrololoom ME, Javadpour S, Ma J. Studying surface charge and suspension stability of hydroxyapatite powder in isopropyl alcohol to prepare stable suspension for electrophoretic deposition. Adv in Appl Ceram 2009;108:241e8. Zhitomirsky I. Cathodic electrodeposition of ceramic and organoceramic materials. Fundamental aspects. Adv Colloid Interf Sci 2002;97:279e317. Xiao XR, Liu RF. Effect of suspension stability on electrophoretic deposition of hydroxyapatite coatings. Matter Lett 2006;60: 2627e32. Tagaya M, Ikoma T, Chakarov D, Kasemo B, Tanaka J. Cleaning methods of proteins adsorbed on hydroxyapatite nanocrystal sensor. Sci Technol Adv Mater 2010;11:045002e11. Yoshioka T, lkoma T, Monkawa A, Tonegawa T, Chakarov D, Kasemo B, et al. Proteins adsorption on hydroxyapatite nanocrystals with quartz crystal microbalance technique. Key Eng Mater 2008;361:1119e22. Ikoma T, Tagayam M, Hanagata N, Yoshioka T, Chakarov D, Kasemo B, et al. Protein adsorption on hydroxyapatite nanosensors with different crystal sizes studied in situ by quartz crystal microbalance with dissipation method. J Am Ceram Soc 2009;92:1125e8. Ikoma T, Yamazaki A, Nakamura S, Akao M. Preparation and structure refinement of monoclinic hydroxyapatite. J Solid State Chem 1999;144:272e6. Tagaya M, Ikoma T, Takemura T, Okuda M, Hanagat N, Yoshioka T, et al. Adsorption of proteins derived from fetal bovine serum onto hydroxyapatite nanocrystals with quartz crystal microbalance technique. Key Eng Mater 2009;396e398: 47e50. Redepenning J, Schlesinger TK, Mechalke EJ, Puleo DA, Bizios R. Osteoblast attachment monitored with a quartz crystal microbalance. Anal Chem 1993;65:3378e81. Weerawardena A, Drummond CJ, Caruso F, McCormick M. Real time monitoring of the detergency process by using a quartz crystal microbalance. Langmuir 1998;14:575e7. Payet V, Brunner S, Galtayries A, Frateur I, Marcus P. Cleaning of albumin-contaminated Ti and Cr surfaces: an XPS and QCM study. Surf Interface Anal 2008;40:215e9. Bose S, Saha SK. synthesis and characterization of hydroxyapatite nanopowders by emulsion technique. Chem Mater 2003;15:4464e9. Koumoulidis GC, Katsoulidis AP, Ladavos AK, Pomonis PJ, Trapalis CC, Sdoukos AT, et al. Preparation of hydroxyapatite via microemulsion route. J Colloid Interface Sci 2003; 259:254e60. Lim GK, Wang J, Ng SC, Gan LM. Nanosized hydroxyapatite powders from microemulsions and emulsions stabilized by a biodegradable surfactant. J Mater Chem 1999;9:1635e9. Shimabayashi S, Tanaka H, Nakagaki M. Adsorption of dodecyl sulfate ion on hydroxyapatite and concurrent release of phosphate and calcium ions from the surface of hydroxyapatite. Chem Pharm Bull 1986;34:4474e8. Kern W, Puotinen DA. Clean solutions based on hydrogen peroxide for use in silicon semiconductor technology. RCA Rev 1970;31:187e206.

574

[96] van der Meerakker JEAM, van der Straaten MHM. A mechanistic study of silicon etching in NH3/H2O2 cleaning solutions. J Electrochem Soc 1990;137:1239e43. [97] Yamamoto K, Nakamura A, Hase U. Control of cleaning performance of an ammonia and hydrogen peroxide mixture (APM) on the basis of a kinetic reaction model. IEEE Trans Semicond Manuf 1999;12:288e94. [98] Hiroki A, LaVerne JA. Decomposition of hydrogen peroxide at watereceramic oxide interfaces. J Phys Chem B 2005;109: 3364e70. [99] Stadtman ER, Levine RL. Free radical-mediated oxidation of free amino acids and amino acid residues in proteins. Amino Acids 2003;25:207e18. [100] Bolon DA, Kunz CO. Ultraviolet depolymerization of photoresist polymers. Polym Eng Sci 1972;12:109e11. [101] Sowell RR, Cuthrell RE, Mattox DM, Bland RD. Surface cleaning by ultraviolet radiation. J Vac Sci Technol 1974;11:474e5. [102] Vig JR. UV ozone cleaning of surfaces. J Vac Sci Technol 1985;A3:1027e34. [103] Ye T, McArthur EA, Borguet E. Mechanism of UV photoreactivity of alkylsiloxane self-assembled monolayers. J Phys Chem B 2005;109:9927e38. [104] Lunts AC. Chemical dynamics of the reactions of O(1D) with saturated hydrocarbons. J Chem Phys 1980;72:1143e52. [105] Norde W. Adsorption of proteins at solid-liquid interfaces. Cell Mater 1995;5:97e112. [106] Brash JL, Horbett TA. In: Proteins at interfaces II: fundamentals and applications. ACS symposium series, vol. 602. Washington, DC: American Chemical Society; 1995. p. 1e23. [107] Yin G, Liu Z, Zhan J, Ding F, Yuan N. Impacts of the surface charge property on protein adsorption on hydroxyapatite. Chem Eng J 2002;87:181e6. [108] Kurrat R, Prenosil JE, Ramsden JJ. Kinetics of human and. bovine serum albumin adsorption at silicaetitania surfaces. J Colloid Interface Sci 1997;185:1e8. [109] Wei T, Kaewtathip S, Shing K. Buffer effect on protein adsorption at liquid/solid interface. J Phys Chem C 2009;113:2053e62. [110] Lubarsky GV, Davidson MR, Bradley RH. Hydrationedehydration of adsorbed protein films studied by AFM and QCM-D. Biosensors Bioelectron 2007;22:1275e81. [111] Hemmersama AG, Rechendorffa K, Foss M, Sutherlanda DS, Besenbachera F. Fibronectin adsorption on gold, Tie, and Tae oxide investigated by QCM-D and RSA modelling. J Colloid Interface Sci 2008;320:110e6. [112] Ho¨o¨k F, Rodahl M, Brzezinski P, Kasemo B. Energy dissipation kinetics for protein and antibody-antigen adsorption under shear oscillation on a quartz crystal microbalance. Langmuir 1998;14:729e34. [113] Dolatshahi-Pirouz A, Jensen T, Foss M, Chevallier J, Besenbacher F. Enhanced surface activation of fibronectin upon adsorption on hydroxyapatite. Langmuir 2009;25:2971e8. [114] Rodahl M, Kasemo B. On the measurement of thin liquid overlayers with the quartz-crystal microbalance. Sensors Actuators A 1996;54:448e56. [115] Bannasch H, Fohn M, Unterberg T, Bach AD, Weyand B, Stark GB. Skin tissue engineering. Clin Plast Surg 2003;30: 573e89.

Handbook of Advanced Ceramics

[116] Cho NJ, Kanazawa KK, Glenn JS, Frank CW. Employing two different quartz crystal microbalance models to study changes in viscoelastic behavior upon transformation of lipid vesicles to a bilayer on a gold surface. Anal Chem 2007;79:7027e35. [117] Hemmersam AG, Foss F, Chevallier J, Besenbacher F. Adsorption of fibrinogen on tantalum oxide, titanium oxide and gold studied by the QCM-D technique. Colloids Surf B:Biointerfaces 2005;43:208e15. [118] Roach P, Farrar D, Perry CC. Interpretation of protein adsorption: surface-induced conformational changes. J Am Chem Soc 2005;127:8168e73. [119] Choi KH, Friedt JM, Frederix F, Campitelli A, Borghs G. Simultaneous atomic force microscope and quartz crystal microbalance measurements: investigation of human plasma fibrinogen adsorption. Appl Phys Lett 2002;81:1335e7. [120] Hall CE, Slayter HS. The fibrinogen molecule: its size, shape, and mode of polymerization. J Biophys Biochem Cytol 1959;5:11e7. [121] Serro AP, Bastos M, Pessoa JC, Saramago B. Bovine serum albumin conformational changes upon adsorption on titania and on hydroxyapatite and their relation with biomineralization. J Biomed Mater Res 2004;70A:420e7. [122] Beissinger RL, Leonard EF. Plasma protein adsorption and desorption rates on quartz: approach to multi-component systems. Trans Am Soc Artif Intern Organs 1981;27:225e30. [123] Lundstrom I, Elwing H. Simple kinetic models for protein exchange reactions on solid surfaces. J Colloid Interface Sci 1990;136:68e84. [124] Vroman L, Adams AL. Findings with the recording ellipsometer suggesting rapid exchange of specific plasma proteins at liquid/ solid interfaces. Surf Sci 1969;16:438e46. [125] Soderquist ME, Walton AG. Structural changes in proteins adsorbed on polymer surfaces. J Colloid Interface Sci 1980;75:386e97. [126] Ho¨o¨k F, Kasemo B, Nylander T, Fant C, Sott K, Elwing H. Variations in coupled water, viscoelastic properties, and film thickness of a Mefp-1 protein film during adsorption and crosslinking: a monitoring, ellipsometry, and surface plasmon resonance study. Anal Chem 2001;73:5796e804. [127] Horbett TA. Mass action effects on the adsorption of fibrinogen from hemoglobin solutions and from plasma. Thromb Haemost 1984;51:174e81. [128] Lutanie E, Voegel JC, Schaaf P, Freund M, Cazenave JP, Schmitt A. Competitive adsorption of human immunoglobulin G and albumin: consequences for structure and reactivity of the adsorbed layer. Proc Natl Acad Sci 1992;89:9890e4. [129] Haynes CA, Norde W. Structures and stabilities of adsorbed proteins. J Colloid Interface Sci 1995;169:313e28. [130] De´jardin P, ten Hove P, Yu XJ, Brash JL. Competitive adsorption of high molecular weight kininogen and fibrinogen from binary mixtures to glass surface. Langmuir 1995;11:4001e7. [131] Holmberg M, Stibius KB, Larsen NB, Hou X. Competitive protein adsorption to polymer surfaces from human serum. J Mater Sci 2008;19:2179e85. [132] Holmberg M, Hou X. Competitive protein adsorption: multilayer adsorption and surface induced protein aggregation. Langmuir 2009;25:2081e9. [133] Green RJ, Davies MC, Roberts CJ, Tendler SJB. Competitive adsorption as observed by surface plasmon resonance. Biomaterials 1999;20:385e91.

Chapter | 7.2

Application of a Quartz Crystal Microbalance with Dissipation

[134] McClary K, Ugarova TP, Grainger DW. Modulating fibroblast adhesion, spreading, and proliferation using self-assembled monolayer films of alkylthiolates on gold. J Biomed Mater Res 2000;50:428e39. [135] Koenig AL, Gambillara V, Grainger DW. Correlating fibronectin adsorption with endothelial cell adhesion and signaling on polymer substrates. J Biomed Mater Res 2003;64A:20e37. [136] Tagaya M, Ikoma T, Migita S, Okuda M, Takemura T, Hanagata N, et al. Fetal bovine serum adsorption onto hydroxyapatite sensor monitoring by quartz crystal microbalance with dissipation technique. Mater Sci Eng B 2010;173:176e81. [137] Voinova MV, Rodahl M, Jonson M, Kasemo B. Viscoelastic acoustic response of layered polymer films at fluid-solid interfaces: continuum mechanics approach. Phys Scrip 1999;59:391e6. [138] Rodahl M, Ho¨o¨k F, Fredriksson C, Keller CA, Krozer A, Brzezinski P, et al. Simultaneous frequency and dissipation factor QCM measurements of biomolecular adsorption and cell adhesion. Faraday Discuss 1997;107:229e46. [139] Jenny M, Hossein A, Peter K, Duncan SS. Viscoelastic modeling of highly hydrated laminin layers at homogeneous and nanostructured surfaces: quantification of protein layer properties using QCM-D and SPR. Langmuir 2007;23:9760e8. [140] Gurak E, Dupont-Gillain CC, Booth J, Roberts CJ, Rouxhet PG. Resolution of the vertical and horizontal heterogeneity of adsorbed collagen layers by combination of QCM-D and AFM. Langmuir 2005;212:10684e92. [141] Webster TJ, Siegel RW, Bizios R. Osteoblast adhesion on nanophase ceramics. Biomaterials 1999;20:1221e7. [142] Zhou T, Marx KA, Warren M, Schulze H, Braunhut SJ. The quartz crystal microbalance as a continuous monitoring tool for the study of endothelial cell surface attachment and growth. Biotech Prog 2000;16:268e77. [143] Marx KA, Zhou T, Montrone A, Schulze H, Braunhut SJ. A quartz crystal microbalance cell biosensor: detection of microtubule alterations in living cells at nM nocodazole concentrations. Biosensors Bioelectron 2001;16:773e82. [144] Marx KA, Zhou T, Warren M, Braunhut SJ. Quartz crystal microlbalance study of endothelial cell number dependent differences in initial adhesion and steady-state behavior: evidence for cell-cell cooperativity in initial adhesion and spreading. Biotech Prog 2003;19:987e99. [145] Marx KA, Zhou T, Montrone A, McIntosh D, Braunhut SJ. Quartz crystal microbalance biosensor study of endothelial cells and their extracellular matrix following cell removal: evidence for transient cellular stress and viscoelastic changes during detachment and the elastic behavior of the pure matrix. Anal Biochem 2005;343:23e34. [146] Le Guillou-Buffello D, Bareille R, Gindre M, Sewing A, Laugier P, Amedee J. Additive effect of RGD coating to functionalized titanium surfaces on human oseoprogenitor cell adhesion and spreading. Tissue Eng A 2008;14:1445e55. [147] Fohlerova´ Z, Skla´dal P, Tura´nek J. Adhesion of eukaryotic cell lines on the gold surface modified with extracellular matrix proteins monitored by the piezoelectric sensor. Biosensors Bioelectronics 2007;22:1896e901. [148] Wegener J, Seebach J, Janshoff A, Galla HJ. Analysis of the composite response of shear wave resonators to the attachment of mammalian cells. Eur Biophys J 2000;78:2821e33. [149] Janshoff A, Wegener J, Sieber M, Galla HJ. Double-mode impedance analysis of epithelial cell monolayers cultured on shear wave resonators. Eur Biophys J 1996;25:93e103.

575

[150] Marxer CM, Coen MC, Greber T, Greber UF, Schlapbach L. Cell spreading on quartz crystal microbalance elicits positive frequency shifts indicative of viscosity changes. Anal Bioanal Chem 2003;377:578e86. [151] Rodahl M, Ho¨o¨k F, Kasemo B. QCM operation in liquids: explanation of measured variations in frequency and Q factor with liquid conductivity. Anal Chem 1996;68:2219e27. [152] Fredriksson C, Khilman S, Kasemo B, Steel DM. In vitro realtime characterization of cell attachment and spreading. J Mater Sci: Mater Med 1998;9:785e8. [153] Fredriksson C, Khilman S, Rodahl M, Kasemo B. The piezoelectric quartz crystal mass and dissipation sensor: a means of studying cell adhesion. Langmuir 1998;14:248e51. [154] Cans AS, Ho¨o¨k F, Shupliakov O, Ewing AG, Eriksson PS, Brodin L, et al. Measurement of the dynamics of exocytosis and vesicle retrieval at cell populations using a quartz crystal microbalance. Anal Chem 2001;73:5805e11. [155] Modin C, Stranne AL, Foss M, Duch M, Justesen J, Chevallier J, et al. QCM-D studies of attachment and differential spreading of pre-osteoblastic cells on Ta and Cr surfaces. Biomaterials 2006;27:1346e54. [156] Lord MS, Modin C, Foss M, Duch M, Simmons A, Pedersen FS, et al. Monitoring cell adhesion on tantalum and oxidized polystyrene using a quartz crystal microbalance with dissipation. Biomaterials 2006;27:4529e37. [157] Lord MS, Modin C, Foss M, Duch M, Simmons A, Pedersen FS, et al. Extracellular matrix remodelling during cell adhesion monitored by the quartz crystal microbalance. Biomaterials 2008;29:2581e7. [158] Moseke C, Ewald A. Cell and protein adsorption studies using quartz microgravimetry with dissipation monitoring. Mater Wiss Werkstof 2009;40:36e42. [159] Tagaya M, Ikoma T, Takemura T, Migita S, Okuda M, Yoshioka T, et al. Initial adhesion behavior of fibroblasts onto hydroxyapatite nano-crystals. 22nd international symposium on ceramics in medicine. Bioceramics 2009;22:439e42. [160] Tagaya M, Yamazaki T, Migita S, Hanagata N, Ikoma T. Hepatocyte adhesion behavior on modified hydroxyapatite nanocrystals with quartz crystal microbalance. 22nd international symposium on ceramics in medicine. Bioceramics 2009; 22:407e10. [161] Tagaya M, Ikoma T, Takemura T, Yoshioka T, Hanagata N, Tanaka J. Protein Adsorption and Subsequent Fibroblasts Adhesion on Hydroxyapatite Nanocrystals. IOP Conference Series: Mater Sci Eng 2011;18:192009e12. [162] Li F, Wang HC, Wang QM. Thickness shear mode acoustic wave sensors for characterizing the viscoelastic properties of cell monolayer. Sensors Actuators B 2008;128:399e406. [163] Ferna´ndez P, Heymann L, Ott A, Aksel N, Pullarkat PA. Shear rheology of a cell monolayer. New J Phy 2007;9:419e46. [164] Palmer A, Mason TG, Xu J, Kuo SC, Wirtz D. Diffusing wave spectroscopy microrheology of actin filament networks. Biophy J 1999;76:1063e71. [165] Aoki H, Akao M, Shin Y, Tsuzi T, Togawa T. Sintered hydroxyapatite for a percutaneous device and its clinical application. Med Prog Technol 1987;12:213e20. [166] Furuzono F, Ueki M, Kitamura H, Oka K, Imai E. Histological reaction of sintered nano-hydroxyapatite-coated cuff and its fibroblast-like cell hybrid for an indwelling catheter. J Biomed Mater Res Part B Appl Biomater 2009;89B:77e85.

Chapter 7.3

Anticancer Diagnoses and Treatments Using Ferrite Nanoparticles and Bulk Masanori Abe,* Tomoaki Ueda,y Takashi Nakagawa,** Mamoru Hatakeyamay and Hiroshi Handayy Professor Emeritus of Tokyo Institute of Technology, Japan, y Solutions Research Laboratory, Tokyo Institute of Technology, 4259 Nagatsuta-cho, Midori, Yokohama, Kanagawa 226-8503, Japan, ** Management of Industry and Technology, Graduate School of Engineering, Osaka University, *

2-1 Yamadaoka, Suita, Osaka 565-0871, Japan,

yy

Department of Biological Information, Graduate School of Bioscience and Biotechnology,

Tokyo Institute of Technology, Nagatsuta-cho, Midori, Yokohama, Kanagawa 226-8503, Japan

Chapter Outline 1. Introduction 2. High-Performance MRI Contrast Agent 3. Mediators for Self-controlled Induction Heating

577 578 578

1. INTRODUCTION Ferrite nanoparticles (FNPs) have a high biocompatibility and exhibit unique functionality when exposed to magnetic fields and electromagnetic radiations. Applying these features, some anticancer treatments have been developed using FNPs in vivo [1,2]. Among them, magnetic resonance imaging (MRI) contrast enhancement has reached a stage of standard clinical procedure [3]. It uses colloidal fluids, called ResovistÒ, FeridexÒ, etc., of superparamagnetic FNPs which are approved by U.S. Food and Drug Administration (FDA) for intravenous administration into human bodies. Other anticancer diagnoses and treatments using FNPs are in clinical test stage at highest. They include the following. 1. Magnetic drug delivery: Anticancer drugs are immobilized to FNPs, which are delivered to target tumors selectively using focused external magnetic field and sustained release is made. This allows reducing dosage of the drugs and thus suppressing the drug side effects [4]. However, high-performance magnetic drug delivery has not yet been realized, mainly because magnetic drug forces acting on FNPs become very weak when they are located in deep position. 2. Magnetic hyperthermia: As shown in Figure 1, FNPs are selectively targeted to tumors, either immobilized onto the surfaces of the tumors or uptaken by the cancer cells. In addition, they are induction-heated by an external alternating magnetic field to cause heat death

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00031-9 Copyright Ó 2013 Elsevier Inc. All rights reserved.

4. Sentinel Lymph Node Mapping for Monitoring Cancer Metastasis References

580 581

of cancer cells selectively [5]. This suppresses unwanted heat damage to normal cells and allows noninvasive anticancer treatments. To realize the magnetic hyperthermia, however, we must resolve the problems remaining in attaining not only strong but also controlled heat rise by the FNPs. More expectation is placed on the magnetic hyperthermia at present, since it was found lately that anticancer immunity is enhanced by the hyperthermia; in animal experiments, tumors without injection of heat mediators were cured as well by the hyperthermia treatment [6]. It is suggested that the physiologically active substances (e.g., heat shock

FIGURE 1 Anticancer magnetic hyperthermia utilizing ferrite nanoparticles which generate heat under alternating magnetic field to kill cancer cells selectively. For color version of this figure, the reader is referred to the online version of this book.

577

578

Handbook of Advanced Ceramics

proteins) released from the cancer cells which underwent heat death enhance tumor immunity. Utilizing particles and bulk ferrite, we have developed high-performance MRI contrast agent, fabricated thermal mediators for self-temperature-controlled hyperthermia, and proposed a novel type of “sentinel lymph node mapping” for staging metastasis of cancers. The following sections review our studies, explaining the background.

2. HIGH-PERFORMANCE MRI CONTRAST AGENT The FDA-approved MRI contrast agents (e.g., ResovistÒ) are made of superparamagnetic particles (several nanometers in size) of spinel ferrites, Fe3O4, g-Fe2O3, or the intermediate [3]. They enhance MRI contrast by shortening T2 relaxation time of protons nearby. They are coated with carboxyl dextran to form spheres of 18e65-nm diameter, which are uptaken by Kupffer cells in livers. Because cancer cells do not contain Kupffer cells, tumors are contrasted against normal cells containing Kupffer cells. Thus, the contrast agents are used to identify liver cancers. Since brighter and clearer images are obtained by T1 shortening signals than by T2 shortening signals in MRI, we [7] have developed a novel T1-weighted contrast agent of FNPs. They were synthesized through a thermolysis of the ironeoleate complex organic solvent to spheres of 4.6  0.5 nm diameter with narrow distribution. Then they were coated with citric acid by a two-step ligand exchange reaction to attain a high dispersibility in water. Figure 2 shows that the T1-weighted images obtained by the citrate-coated FNPs were as bright as those obtained by MagnevistÒ (a T1-weighted MRI contrast agent of superparamagnetic gadolinium complex), and were much brighter than those obtained by ResovistÒ. This is because

FIGURE 2 T1 weighted MRI images of water solutions of oleic acid. They were obtained using as contrast agent (a) 4 nm ferrite nanoparticles, (b) MagnevistÒ, and (c) ResovistÒ, at various concentrations (normalized with molecular weight of Fe or Gd) which reduce in the order from left to right [7].

our novel FNPs have magnetization much stronger than MagnevistÒ, and have smaller particle size (4.6 nm) than ResovistÒ (18e65 nm). Therefore, we have successfully fabricated T1 shortening contrast agent free from gadolinium, an element which may cause strong toxicity when solved out from the complex. Although small in number, deaths by dissolved gadolinium from the contrast agent have been reported. Developing gadolinium-free T1-weighted MRI contrast agent has been longed for. Our novel contrast agent of FNPs is promising to meet the needs. Also, the FNPs have an advantage that functional molecules can be easily immobilized onto their surfaces via citric acid to attain tissue specificity.

3. MEDIATORS FOR SELF-CONTROLLED INDUCTION HEATING Usually, for magnetic hyperthermia FNPs smaller than several tens nm having single magnetic domains are used as heating mediators. Heat is generated by Ne´el relaxation losses (in spin dynamics) and Brownian relaxation losses (in mechanical dynamics) and secondarily by magnetic hysteresis losses [8]. We revealed that the FNPs exhibit the highest heating ability when their size falls in the range of 15e20 nm [9], as expected from calculation based on Ne´el and Brownian relaxations, as shown in Figures 3 and 4. One of the major difficulties in developing magnetic hyperthermia is to strictly control the temperature to rise above the fatal temperature (ca. 42.5 C ) of cancer cells and yet keep it low enough to avoid thermal damages to normal cells. To solve this problem, we have fabricated smart heat mediators of a bulk ferrite in which the induced temperature rise is automatically stabilized by

FIGURE 3 Temperature rise profiles obtained under alternating magnetic field (120 kHz, 120 Oe) for ferrite nanoparticles with different average diameter d [9]. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 7.3

Anticancer Diagnoses and Treatments Using Ferrite Nanoparticles and Bulk

FIGURE 4 Heat power calculated to be induced by alternating magnetic field (900 kHz, 40 Oe) in ferrite nanoparticles. Results are plotted as a function of average particle size, assuming that the size is distributed with various values of standard deviation s [10]. For color version of this figure, the reader is referred to the online version of this book.

specific magnetic characteristics of the ferrite [10]. They are millimeter in size and of implant type which are inserted into targets. To stabilize the induction heat rise, we chose a perovskite La2/3Sr1/3Mn1 xCuxO3 d (LSMC) whose thermomagnetization curves sharply decrease to zero around their Curie temperature [11]. We prepared bulk samples by a polymerized complex method. As shown in Figure 5,

FIGURE 5 Thermomagnetization curves obtained under 100 Oe static magnetic field for La2/3Sr1/3Mn1 xCuxO3 d bulk samples [11]. Inset shows for example how Curie temperature Tc of Cu044 (x = 0.44) sample was determined.

579

FIGURE 6 Temperature rise profiles obtained for La2/3Sr1/3 Mn1 xCuxO3 d (e.g. Cu041 means x = 0.41) ferrite bulk samples under alternating magnetic field (1 MHz, 40 Oe) [11].

their magnetization decreased sharply around the Curie temperature. Therefore, magnetic induction heating will be automatically balanced at the Curie temperatures; in the LSMC bulk samples, induction heating is caused mainly by the magnetic hysteresis losses and slightly by eddy current losses due to the conductivity accompanying the magnetization. Figures 6 and 7 show that under an alternating magnetic field (1 MHz, 40 Oe) the temperatures of the samples were raised and balanced at temperatures roughly equal to the

FIGURE 7 Relation between balanced temperature Tb and Curie temperature Tc obtained for La2/3Sr1/3Mn1 xCuxO3 d (e.g. Cu041 means x = 0.41) bulk samples [11]. Dashed line indicates Tb = Tc relation.

580

FIGURE 8 Concept of sentinel lymph node (SLN): the lymph node to which fluid (contains isolated cancer cells) reaches first from a tumor. In SLN mapping (for breast cancer in this case) a marker is injected into tumor, accumulated at SLN’s, and then detected. For color version of this figure, the reader is referred to the online version of this book.

Curie temperature [11]. Therefore, LSMC perovskite is a promising material for preparing heat mediators having desired self-controlled temperature.

4. SENTINEL LYMPH NODE MAPPING FOR MONITORING CANCER METASTASIS Cancers metastasize to lymph nodes because cancer cells isolated from primary tumors drift along lymph flow. The

Handbook of Advanced Ceramics

first lymph node that filters the fluid draining away from the primary tumor, as shown in Figure 8, is called “sentinel lymph node (SLN).” If an SLN is not metastasized, cancer is not spread to downstream lymph nodes. Therefore, SLN diagnoses and treatments based on “SLN mapping” were established as follows [12]. A radioisotope marker is injected into a tumor and the isotope marker accumulated at SLNs is detected by a gamma ray counter. Then, the identified SLNs are surgically removed for biopsy to monitor metastasis of the cancer. If no metastasis is found in the SLNs, removal of the downstream lymph nodes is avoided and the patient’s QOL is improved much. The SLN mapping has become standard treatment at present for breast cancers and melanomas and gaining popularity in other cancers of the stomach, colon, lung, etc. However, the isotope marker has a problem of strict law regulation, especially in Japan and Europe, and small medical facilities cannot afford. To solve the problem, SLN mapping by detecting FNPs by SQUID sensors has been proposed [13]. But SQUID sensors are expensive and complicated. Therefore, we are developing a novel type of SLN mapping based on “sonicwave emission by magnetically stimulated particles,” a phenomenon which we found and called SEMP in short [14]. As shown in Figure 9, FNPs are injected into the tumor as a marker. The FNPs accumulated at the SLN are excited by an alternating magnetic field to generate sonic waves, which are detected by a microphone mounted on the bottom of the probe. We succeeded in detecting an “artificial SLN,” a polymer gel sphere (c.a. 3 mm in radius) which contained FNP fluid of Resovist®, at an amount similar to that is accumulated in human SLNs.

FIGURE 9 Sentinel lymph node mapping by detecting ferrite nanoparticles accumulated there. Acoustic waves excited by the ferrite nanoparticles under alternating magnetic field are measured by a microphone attached to the bottom of a probe. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 7.3

Anticancer Diagnoses and Treatments Using Ferrite Nanoparticles and Bulk

REFERENCES [1] Pankhurst QA, Connolly J, Jones SK, Dobson J. Applications of magnetic nanoparticles in biomedicine. J Phys D Appl Phys 2003;36:R167e81. [2] Sandhu A, Handa H, Abe M. Synthesis and applications of magnetic nanoparticles for biorecognition and point of care medical diagnostics. Nanotechnology: 442001 2010;21(22pp). Online: http://iopscience. iop.org/0957-4484/21/44/442001. [3] Lawaczeck R, Menzel M, Pietsch H. Superparamagnetic iron oxide particles: contrast media for magnetic resonance imaging. Appl Organomet Chem 2004;18:506e13. [4] Misra RDK. Magnetic nanoparticle carrier for targeted drug delivery: perspective, outlook and design. Mater Sci Tech 2008;24:1011e9. [5] Jordan A, Wust P, Fahling H, Johns W, Hinz A, Felix R. Inductive heating of ferrimagnetic particles and magnetic fluids: physical evaluation of their potential for hyperthermia. Int J Hyperthermia 1993;9:51e68. [6] Ito A, Shinkai M, Honda H, Kobayashi T. Medical application of functionalized magnetic nanoparticles. J Biosci Bioeng 2005;100:1e11. [7] Hatakeyama M, Kishi H, Kita Y, Imai K, Nishio K, Karasawa S, et al. J Mater Chem 2011;21:5959e66. [8] Rosensweig RE. Heating magnetic fluid with alternating magnetic field. J Magn Magn Materl 2002;252:370e4.

581

[9] Okawa K, Sekine M, Maeda M, Tada M, Abe M, Matsushita N, et al. Heating ability of magnetite nanobeads with various sizes for magnetic hyperthermia at 120 kHz, a noninvasive frequency. J Appl Phys 2006;99:08H102e08H104. [10] Horiki M, Nakagawa T, Yoshioka T, Seino S, Yamamoto T, Abe M. Preparation of La2/3Sr1/3Mn1 xCuxO3 d perovskite ceramics for temperature-self-controlled magnetic hyperthermia mediators and estimates of their heating ability under alternating magnetic fields. J Magn Soc Jpn 2011;35:22e6 (in Japanese). [11] Nguyen C, Pham QN, Hoang NN, Nguyen HL, Nguyen DT. Preparation and magnetocaloric effect of (La0.67 xGdx)Sr0.33MnO3 (X ¼ 0.1, 0.15) nanoparticles. Physica B 2003;327:214e7. [12] Jakub JW, Pendasa S, Reintgen DS. Current status of sentinel lymph node mapping and biopsy: facts and controversies. Oncologist 2003;8:59e68. [13] Song G. Development of a SQUID (Superconducting Quantum Interference Device) Detection System of Magnetic Nanoparticles for Cancer Imaging. A dissertation submitted for the degree of Doctor of Philosophy (Physics) in The University of Michigan; 2009 [14] Abe M, Kakegawa K, Ueda T, Nakagawa T, Tada M, Handa H. Anticancer sentinel-lymph-node mapping by sonic waves generated by magnetic beads under ac magnetic field. Digests Intermag Conf., Sacramento, CA, U.S.A., CB-03; 2009.

Chapter 8.1

Diesel Particulate Filters Joerg Adler and Uwe Petasch Fraunhofer-Institut fuer Keramische Technologien und Systeme (Fraunhofer IKTS; Fraunhofer Institute for Ceramic Technologies and Systems), Winterbergstrasse 28, 01277 Dresden, Germany

Chapter Outline 1. 2. 3. 4. 5.

Introduction Diesel Particulate Matter (DPM) Limits Working Principle of DPFS Design and Properties of DPFS

585 585 586 588 590

1. INTRODUCTION Diesel engines have the highest efficiency of all combustion engines and are widely used in heavy-duty vehicles, agricultural machinery and stationary equipment. Their use in passenger cars has also risen greatly in Western Europe over the last few decades, already making up 50% of the market (approx. 7 M cars) in terms of new registrations in 2006 [1]. Rising concerns regarding fuel consumption and greenhouse gases should drive this trend further. Even in the US, where the share of diesel cars was less than 4% in 2005, a strong increase is expected in the future [2]. Extensive discussions revolving around particulate matter and particularly the harmful aerosol and particulate emissions from incomplete combustion processes (‘soot’) have made purification of diesel engine exhaust a priority in environmental laws and created public awareness of these problems. Factory equipping of diesel cars with particulate filters became technically and economically feasible in 2002. In 2007 the absolute number of ceramic filters in use was approximately 11.7 M. Introduction of the EURO 5/6 emission standards in the EU made equipping of new vehicles (cars, light-duty (LD) and heavy-duty (HD) vehicles) with high-efficiency diesel particulate filters (DPFs) in combination with NOx reduction systems mandatory. However, based on the environmental regulations already in place experts realize that nearly all diesel enginesdfrom forklift trucks to diesel locomotives, from road sweepers to excavators and mobile cranesdin Western Europe and the US will have to be equipped with filters in the next few years.

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00032-0 Copyright Ó 2013 Elsevier Inc. All rights reserved.

6. Ceramic DPFS 7. Market Situation 8. Outlook References

596 601 602 603

Due to their commercial importance DPFs with a very wide range of materials (powders, fibers; ceramics, metals) and designs (deep-bed filters, surface filters) are now being developed and offered, with a large part of the alternative filter and material designs utilizing different regeneration principles. Typically monolith ceramic filters with parallel channels plugged at alternate ends are factory-installed in vehicles.

2. DIESEL PARTICULATE MATTER (DPM) 2.1. Origin and Structure of DPM DPM is formed as a result of incomplete combustion of the diesel fuel in the engine and is composed of carbon and various accumulated compounds including water, inorganic oxides, sulfur-containing compounds and hydrocarbons such as the highly toxic polycyclic aromatic hydrocarbons (PAH). The exact composition depends on the fuel composition, the quality of the intake air for combustion, engine wear and the specific motor design and operating point. The DPM is emitted from the engine as primary particles of size 5e20 nm and rapidly agglomerates to chains or clusters with typical sizes of 50e150 nm. Hence particles in the exhaust system are different regarding time and places. Specifically both the size and the composition of the DPM is influenced by the given engine, fuel and fuel additives [3e5]. After exiting the exhaust system the particles change again. Ultrafine particles of size 2200  C) in protective gas atmosphere. The fine SiC particles sublimate and recondense on the contact points of the coarse particles,

Chapter | 8.1

Diesel Particulate Filters

599

6.1.4.2. Reaction-Bonded SiC

FIGURE 8 Segmented SiC filter. For color version of this figure, the reader is referred to the online version of this book. (Courtesy: IBIDEN).

resulting in autogenously bonding of the coarse grain framework and producing a porous material. The pore size is controlled through selection of the grain size of the coarse fraction, use of special additives and the heat treatment regime. RSiC was originally developed for use as kiln furniture (beams, shelves and capsules) for fast firing of porcelain. The company IBIDEN began developing high-porosity RSiC for filtration applications in 1985 [75]; the first DPF honeycomb filters were reported in 1993 [76] and began being used standard in passenger cars in 2000 [77]. The pore size was 10 mm and the pore volume was 50% for regeneration with a fuel-borne additive. Now coarser and more porous variants are used for catalytic coatings [78]. The binder phase can be modified with fine particles or special additives to improve the material strength and toughness [79,80]. Almost at the same time in Denmark relatively coarse and thick-walled RSiC DPF honeycombs were developed with a relatively wide range of properties and geometries mainly for nonroad and retrofitting applications [81,82]. These honeycombs feature larger and circular segments. Special processes for plugging and the adhesive putty were also developed and employed. Recrystallized SiC can also be produced by means of special microscopic surface structuring for better anchoring of catalytic coatings [83]. Another interesting material is a RSiC material in which, unlike in the above manufacturing process, silicon and a carbon additives are presintered to produce SiC, which then recrystallizes in a subsequent high-temperature heat treatment [84]. Boron and other additives were used to adjust the electrical resistance to regenerate the filter through direct electrical resistance heating [85]. The target application areas were stationary equipment and large engines on ships and locomotives, but development/fabrication was discontinued in 2003.

Secondary bonding of SiC from the reaction of Si with C or bonding by pure silicon represents an alternative to RSiC. In 2000 a DPF material in which the SiC framework was bonded with silicon was introduced [86]. Because the thermomechanical properties (thermal expansion and conductivity) are similar for silicon and SiC (see Table 1) a porous SieSiC produces similar properties to those of RSiC [87]. The silicon bonding yields lower thermal and chemical stabilities than pure SiC and the thermal conductivity of SieSiC is lower than that of RSiC because of the conductivity barrier effect of the SieSiC grain boundaries [88], but extensive applications have thus far demonstrated adequate performance data for standard use in passenger cars [89,90]. At approx. 1600  C the sintering temperature is lower than for RSiC, but sintering must be done in a protective gas. If Si3N4 and carbon additives are used the secondary SiC bonding of a SiC particle framework can be achieved at lower temperatures and pore parameters typical to DPFs (e.g. pore size of 12 mm and pore volume of 43%) can be obtained [91,92]. SiC DPFs can also be manufactured by reaction of a mixture of Si powder and Al and carbon additives, although reaction of the relatively coarse silicon powder results in coarse pores. Various starting powders can be used to adjust the pore size between 7 mm and 22 mm and obtain a comparatively high pore volume [93,94]. 6.1.4.3. Clay-Bonded SiC Ceramics Clay-bonded SiC ceramics can be sintered in air at temperatures of up to 1350  C through addition (approx. 20%) of clay or synthetic SiO2 formers and a flux and are hence much less expensive than the materials requiring the expensive protective gas technology. During sintering the SiC surface oxidizes and a glassy silicate binder phase interspersed with mullite or other crystalline phases according to the additives used forms between the particles. There has been a great deal of testing of the corresponding commercial filter materials for hot gas filtration, particularly in test power plants with pressurized fluidized bed combustion at operating temperatures of up to 850  C [95]. DPFs made from these materials were also developed [96] and used mainly for nonroad applications [97]. The materials have a low thermal conductivity and may have to be segmented depending on the loading conditions. 6.1.4.4. Liquid Phase-Sintered SiC (LPS-SiC) The relatively new variant of SiC known as liquid phasesintered silicon carbide (LPSeSiC) can be sintered using a small amount (approx. 4%) of special additives (e.g. alumina-yttria). This results in doped SiCeSiC bonds and thin grain boundary phases, e.g. composed of the very stable

600

yttria-alumina garnet (YAG) material; dense LPSeSiC grades are characterized by high strengths and toughness levels. With coarse SiC grains and special sintering techniques (95% and a service interval of 80,000 km due to the above-mentioned ash problems were specified. A special service company was established to remove the ash from the filters using a special water vapor cleaning technique. The practicality and reliability were verified even in the extreme conditions of Paris taxi operation [123]. Through lowering of the catalyst concentration the ash content in the second generation could be lowered to such a point that the service interval could be extended to 150,000 km. With the introduction of special channel geometry in 2004 to increase the ash capacity the service interval was extended to approx. 250,000 km. In June 2003 shipment of the 500,000th DPF-equipped vehicle was announced; the one million mark was reached at the beginning of 2005 [124]. All remaining major carmakers followed with DPFs in Europe, creating temporary bottlenecks in filter supply. At the end of 2004 worldwide manufacturing capacity for SiC filters was about 1 M per year; by the end of 2006 it reached approx. 6e9 M (RSiC and SiSiC) filters/year thanks to new plants being opened in Poland and Hungary [125e128]. In early 2006 the first aluminum titanate filter as a standard passenger car feature was announced. The production capacity for cordierite filters was increased between 2004 and 2006, mainly to meet the demand for medium and heavy-duty vehicles [129,130]. A market leader forecast its turnover for DPFs to reach the same level as that of its substrate business for catalytic exhaust aftertreatment in 2010. A study of market trends from 2004 forecast an annual turnover of 1.47 B Euros for passenger car DPFs in 2012

602

and a volume of 4.2e4.4 M per year in Western Europe [131], with the system price taken to be approx. 340 Euros (in 2006 approx. 450 Euros). In 2006 the price of the SiC ceramic filter made up approximately 30% of the cost of the overall system, i.e. between 130 Euros and 160 Euros depending on size and number of units. The oxidation catalyst accounts for a much higher share of the overall system costs due to the high price of the noble metal. The actual developments greatly surpassed all predictions. A more recent study carried out at the end of 2009 [132] assumed 8.5 M DPFs per annum, albeit at a system price of approx. 520 Euros, corresponding to turnover of 4.4 B Euros. The year 2016 is expected to see a volume of 13 M units (system unit price: 480 Euros) and turnover running to 6.2 B Euros, approx. 2.45 B Euros of which would be attributable to the ceramic filter substrates. A 2006 study covering passenger cars and vans, pickups, jeeps etc. [133] was carried out for the North American market (US, Canada and Mexico). The share of diesel engines was expected to increase from 3.6% (2005) to 9.8% in 2012, with about 50% of the vehicles considered being passenger cars. Starting in 2007 equipping of all the considered diesel vehicles (only those manufactured in the US) with DPFs was expected to be mandatory. Turnover was supposed to be US$2.05 B in 2012. The considerably higher unit price in comparison with that in the Western European study was associated with the larger vehicle sizes in the US. However, these predictions haven’t been met yet; in the US the diesel engine market has been stagnating over the last few years. A study from 2011 [134] for ‘light vehicles’ forecast an increase from about 160,000 units (1.3%) in 2009 to about 3% (545,000 units) in 2016, with all units assumed to be fitted with DPFs. Approximately 3 M new commercial vehicles were registered in Europe in 2007. The lion’s share, 75%, was made up of commercial vehicles up to 3.5 T. The majority of the first standard applications for EURO IV relied on DeNOx systems (SCR) to reach the prescribed particle limits without DPFs. Only in 2012 will coupled (DPF þ DeNOx) systems be introduced on a large scale. In the US and Japan, however, other prescribed limits already made DPFs mandatory in 2007; these filters were made mainly from cordierite and to a lesser extent from SiC. With the introduction of environmental zones in some cities in Western Europe in about 2006 the introduction of retrofitting systems for passenger cars also became a topic of discussion. Due to the technical difficulties posed by regeneration only open filter systems, usually the two above-mentioned metal filter types, were developed to the stage allowing them to be deemed marketable. Various studies were based on a volume of approx. 0.7e2 M vehicles in Germany alone. Because of (frequently politically motivated) delays and public discussions about the

Handbook of Advanced Ceramics

effectiveness of such filters the actual volume was only a fraction of the predicted numbers. Greater durability and life of larger engines has transformed the topic of retrofitting lorries and nonroad systems. Commercial vehicles can be retrofitted with metal and ceramic DPFs [135,136]. There are currently approx. 3.2 M commercial vehicles in Germany. The share of retrofits depends on the legal situation as well as tax incentives and is hence difficult to estimate. The oldest market for DPFs is in the retrofitting of nonroad vehicles. This can mainly be traced back to Swiss authorities who passed regulations for tunnel construction sites early on; assessment criteria and test procedures for complex approval testing (so-called VERT tests) were developed based on them. However, there are a very large number of different requirements and a correspondingly large number of systems and filters for this market, making it practically impossible for a comparative overview to be gained and predictions to be made. Test results and recommendations for filter systems are currently being updated and published [137]. In contrast to commercial vehicles, these machines are not surveyed statistically. The starting point for a market estimate is hence formed by the production data supplied by the manufacturers. The number of newly manufactured nonroad vehicles in Europe was estimated to be approx. 120,000 to 150,000 in 2007. Given an average system price of 3500 Euros the annual market volume could reach approx. 0.6 B Euros in 2012/2013.

8. OUTLOOK The PMeNOx trade-off has driven the development of numerous technical alternatives for restricting emissions of harmful substances from diesel engines, especially through the use of DPFs, DeNOx systems, exhaust gas recirculation and internal engine alterations. However, trends in regulation of limits already point to use of DPFs, at least in combination with DeNOx systems, becoming obligatory in the future for nearly all diesel engines in industrialized nations. The introduction of particulate number limits in EURO 6/VI can cause a redesign of some filter strategies, although most of the current wall flow filter systems reach the limit of 6 * 1011 particles/km. Ceramic DPFs are a great success from the ceramist’s point of view, comparable to the oxidation catalyst carrier business. The number and variety of ceramic DPFs have been increasing tremendously since their commercial introduction in cars. Wall flow filters made of silicon carbide ceramics are at an advantage in cars because of their robustness. Aluminum titanate, cordierite and mullite are of interest mainly because they can be produced more cheaply. Higher cell density, asymmetric cell design and catalyzed filters have been developed and introduced in the market.

Chapter | 8.1

Diesel Particulate Filters

The trend in materials is marked by strong diversification allowing optimization of properties, manufacturing technologies and costs, but also by competition and patentrelated issues. Development of the filter design and system is also focused on further optimization in relation to pressure drop, costs (filter, coating and additional fuel consumption), and the increase in ash tolerance and especially the combined DPFeDeNOx systems. Current filter system development activities also concentrate on lifetime guarantees, fast catalyst light-off and on-board diagnostics. The extent to which the necessity of exhaust aftertreatment and the associated costs will cancel out the advantages of the diesel engine in comparison with the Otto-cycle engine is an obvious question. However, expected limits for CO2 emissions and also possibly introduction of particle count-based limits for Otto-cycle engines will accelerate the trend toward diesel engines in the future. Thus, the market share for diesel engines should still continue to rise. Over the long term, however, at least for passenger cars, alternative drive technologies (hybrid drives, electric vehicles, fuel cells etc.) will have to be considered.

REFERENCES [1] European Union Economic Report. Brussels: ACEA; February 2008. [2] Global markets for diesel powered light vehicles to 2017. J.D. Power and Ass.; Dec. 2007. [3] Kittelson DB. Engines and nanoparticles: a review. J Aerosol Sci 1998;29(5/6):575e88. [4] Liati A, Eggenschwiler PD. Characterization of particulate matter deposited in diesel particulate filters: visual and analytical approach in macro-, micro and nano-scales. Combust Flame 2010;157:1658e70. [5] Wal RLV, et al. HRTEM Study of diesel soot collected from diesel particulate filters. Carbon 2007;45:70e7. [6] Peters A, et al. Respiratory effects are associated with the number of ultrafine particles. Am J Respir Crit Care Med 1997;155:1376e83. [7] Samaras Z, et al. Characterisation of exhaust particulate emissions from road vehicles, final publishable report under EC-Contract 2000-RD.11091, April 2005. [8] Modelling and Assessment of the Health Impact of Particulate Matter and Ozone, Summary report by the Joint Task Force on the Health Aspects of Air Pollution of the World Health Organisation/ European Centre for Environment and Health and the Executive Body. United Nations Economic Commission for Europe. EB.AIR/WG 1/2004/11. [9] Calderon-Garciduenas L. PM and the central nervous system: brain inflammation and neuro-degeneration in exposed children and young adults, 10th ETH-Conference on Combustion Generated Nanoparticles, 8/2006. [10] Hansen J, et al. Climate change and trace gases. Philos Trans R Soc A 2007;365:1925e54.

603

[11] [12]

[13] [14]

[15] [16] [17] [18]

[19]

[20]

[21]

[22]

[23] [24] [25] [26] [27] [28] [29]

[30] [31]

[32]

Johnson TV. Review of diesel emissions and control. SAE 201001-0301; 2010. Regulation (EC) No 715/2007 of the European Parliament and of the Council of 20 June 2007 on type approval of motor vehicles with respect to emissions from light passenger and commercial vehicles (Euro 5 and Euro 6) and on access to vehicle repair and maintenance information. UN/ECE Reg 83(Suppl. 7). Regulation (EC) No 595/2009 of the European Parliament and of the Council of 18 June 2009 on type approval of motor vehicles with respect to emissions from heavy duty vehicles (Euro VI) and on access to vehicle repair and maintenance information, Official Journal L188. Johnson TV. Diesel emissions in review. SAE 2011-01-0304; 2011. EU Official Gazette No. L257 from 10.10.1996, p. 26. Olsen III RA, Martins LCB. Cellular ceramics in metal filtration. Adv Eng Mater 2005;7(4):187e92. Gainer G, Rudely J, Damson B, Huber J. Eberspa¨cher Active Exhaust Technology e funktionsintegrierte motorunabha¨ngige Strategie fu¨r die aktive Abgasnachbehandlung. In: Herausforderung Abgasnachbehandlung fu¨r Dieselmotoren. Dresden: 3. Konferenz FAD; 2005. p. 115e27. Bremser K, Kroner P, Ranalli M. Eine alternative Regenerationsstrategie: Diesel Kraftstoffverdampfer. In: Herausforderung Abgasnachbehandlung fu¨r Dieselmotoren. Dresden: 3. Konferenz FAD; 2005. p. 181e6. Terres F, Michelin J, Weltens H. Beladungs- und Regenerationsverhalten von Partikelfiltern fu¨r Diesel-Pkw; MTZ 7e8/ 2002. Sappok A, et al. Characteristics and effects of ash accumulation on diesel particulate filter performance: rapidly aged and field aged test results. SAE 2009-01-1086; 2009. Rocher L, et al. New generation fuel borne catalyst for reliable DPF operation in globally diverse fuels. SAE 2011-01-0297; 2011. Cutler WA. Overview of ceramic materials for diesel particulate filter applications. Ceram Eng Sci Proc 2004;25(3):421e30. Adler J. Ceramic diesel particulate filters. Int J Appl Ceram Technol 2005;2(6):429e39. Masoudi M. Pressure drop of segmented diesel particulate filters. SAE 2005-01-0971; 2005. Ohno K, et al. Characterization of SiCeDPF for passenger car. SAE 2000-01-0185; 2000. Sato H. Soot mass limit analysis of SiC DPF. Ceram Eng Sci Proc 2004;25(3):431e6. Mercedes-Benz Press Release, October 09, 2006. Paule M, Mackensen A, Binz R, Enderle C. Challenges for the next generation of BlueTEC emission technology. SAE 201101-0294; 2011. Rohart E, et al. Acidic zirconia mixed oxides for NH3-SCR catalysts for PC and HD applications. SAE 2011-01-1327; 2011. Xu L, McCabe R, Tennison P, Jen HW. Laboratory and vehicle demonstration of “2nd-Generation” LNT þ in-situ SCR diesel emission control systems. SAE 2011-01-0308; 2011. Macaudiere P. SCR as a technical response to automotive environmental challenges. Berlin: 3th International MinNOx Conference, Session1 Presentation 3; June 2010.

604

[33] Toyota Press Release, October 18, 2003. [34] Shoji A. An improvement of diesel PM and NOx reduction system. Detroit: Diesel Engine-Efficiency and Emissions Research (DEER) Conference; August 2007. [35] Hirabayashi H. Development of new diesel particulate active reduction system for both NOx and PM reduction. SAE 201101-1277; 2011. [36] Cavataio C, Girard JW, Lambert CK. Cu/Zeolite SCR on high porosity filters: laboratory and engine performance evaluations. SAE 2009-01-0897; 2009. [37] Tan J, Solbrig C, Schmieg SJ. The development of advanced 2-way SCR/DPF systems to meet future heavy-duty diesel emissions. SAE 2011-01-1140; 2011. [38] Naseri M, et al. Development of SCR on diesel particulate filter system for heavy duty applications. SAE 2011-01-1312; 2011. [39] Stewart M, Rector D, Muntean G, Maupin G. A mechanistic model for particle deposition in diesel particulate filters using the Lattice-Boltzmann technique. Ceram Eng Sci Proc 2004; 25(3):437e46. [40] Konstandopoulos A, Kostoglou M, Vlachos N, Zarvalis D. Modelling of diesel particulate filters. In: Steinmetz E, editor. Minimierung der Partikelemissionen von Verbrennungsmotoren. Essen: Expert-Verlag, ISBN 3-8169-2430-1; 2004. p. 94e107. [41] Koltsakis C, et al. Performance of catalyzed particulate filters without upstream oxidation catalyst. SAE 2005-01-0952; 2005. [42] Tang W, et al. Modelling aspects of asymmetric channel configuration DPFs. SAE 2009-01-1272; 2009. [43] Chiatti G, Chiavola O, Falcucci G. Soot morphology effects on DPF performance. SAE 2009-01-1279; 2009. [44] Surenhalli HS, et al. Modeling study of active regeneration of a catalyzed particulate filter using one-dimensional DOC and CPF models. SAE 2011-01-1242; 2011. [45] Koltsakis GC, et al. Flow maldistribution effects on DPF performance. SAE 2009-01-1280; 2009. [46] Kelm EF. Manufacture of multiple flow path body. US 4041592, published 16.8.1977. [47] Schmidtgen T, Adler J. Flexibles DPF-Design gepaart mit innovativer SiC-Keramik. Dresden: 5. Konferenz FAD; 2007. 139e149. [48] Ogyu K, Ohno K, Hong S, Komori T. Ash storage capacity enhancement of diesel particulate filter. SAE 2004-01-0949. [49] Bardon S, et al. Asymmetrical channels to increase DPF lifetime. SAE 2004-01-0950. [50] Aravelli K, Heibel A. Improved lifetime pressure drop management for Robust Cordierite (RC) filters with Asymmetric Cell Technology (ACT). SAE 2007-01-0920; 2007. [51] Kriegesmann J, Strate P, Buechler R. Flachmembranstapel und Verfahren zur Herstellung eines solchen. DE 10329229; 28.06.2003. [52] Mizuno Y, et al. Study on wall pore structure for next generation diesel particulate filter. SAE 2008-01-0618; 2008. [53] Furuta Y, et al. Study on next generation diesel particulate filter. SAE 2009-01-0292; 2009. [54] Ogyu K, Oya T, Ohno K, Konstandopoulos AG. Improving of the filtration and regeneration performance by the SiC-DPF with the layer coating of PM oxidation catalyst. SAE 2008-01-0621; 2008. [55] Karin P, Hanamura K. Particulate trapping and oxidation on a catalyst membrane. SAE 2010-01-0808; 2010.

Handbook of Advanced Ceramics

[56] Oki H, Karin P, Hanamura K. Visualization of oxidation of soot nanoparticles trapped on a diesel particulate membrane filter. SAE 2011-01-0602; 2011. [57] Adler J, Rehak P. Keramischer Partikelfilter und Verfahren zu seiner Herstellung. DE 10343438; 15.09.2003. [58] Mizutani T. Performance verification of next generation diesel particulate filter. SAE 2010-01-0531; 2010. [59] Iwasaki S, et al. New design concept for diesel particulate filter. SAE 2011-01-0603; 2011. [60] Howitt JS, Montierith MR. Cellular diesel particulate filter. SAE 810114; 1981. [61] Bachiorrini A. New hypotheses on the mechanism of the detoriation of cordierite diesel filters in the presence of metal oxides. Ceram Int 1996;22:73e7. [62] Maier NM, Nickel KG, Engel c, Mattern A. Mechanisms and orientation dependence of the corrosion of single crystal Cordierite by model diesel particulate ashes. J Eur Ceram Soc 2010;30(7):1629e40. [63] Locker RL, et al. Diesel particulate filter operational characterization. SAE 2004-01-0958; 2004. [64] Boger T, et al. A next generation cordierite diesel particle filter with significantly reduced pressure drop. SAE 2011-01-0813; 2011. [65] Cutler WA, Merkel GA. A new high temperature ceramic material for diesel particulate filter applications. SAE 2000-01-2844; 2000. [66] CORNING Press Release, 27.04.2005. [67] Ogunwumi SB, et al. Aluminium titanate compositions for diesel particulate filters. SAE 2005-01-0583; 2005. [68] Bogner T. et al. Untersuchung der Eigenschaften neuer Dieselpartikelfilter, MTZ 09/2005, 660e669, 2005. [69] Ingram-Ogunwumi RS, et al. Performance evaluations of aluminium titanate diesel particulate filters. SAE 2007-01-0656; 2007. [70] Rose D, et al. On road durability and field experience obtained with an aluminium titanate diesel particulate filter. SAE 2007-011269; 2007. [71] Boger T. Next generation aluminium titanate filter for light duty diesel applications. SAE 2011-01-0816; 2011. [72] Franz A, Klupski S, Hofmeister M. Papierwickel alsHalbzeug fu¨r die Rußfilterproduktion. ATZ Autotechnology 2010;Vol. 10:52e5. 03-2010. [73] Ohno K, Shimato K, Tsuji M. Honeycomb filter and ceramic filter assembly. US 6.669.751, filed 26.09.2000. [74] Briot A. Minimizing filter volume by design optimization. SAE 2007-01-0657; 2007. [75] Tsukada K. Porous silicon carbide sinter and its production. US 4777152, Filed 29.2.1988 (Japanese Priority 29.04.1985). [76] Itoh K, et al. Study of SiC application to diesel particulate filter (Part 1): material development. SAE 930360; 1993. [77] Salvat O, Marez P, Belot G. Passenger car serial application of a particulate filter system on a common rail direct injection diesel engine. SAE 2000-01-0473; 2000. [78] Gege I, Ohno K, Hong S, Sato H. Dieselpartikelfilter-Filtermaterialien, innovation und design. FAD Konferenz Dresden; 2003. p. 698. [79] Oya T, et al. Performance evaluation of SiCeDPF sintered with sintering additive. SAE 2005-01-0579; 2005. [80] Ohno K, Yamayose K, Sato H, Ogyu K. Further durability enhancement of re-crystallized SiC-DPF. SAE 2004-01-0954; 2004.

Chapter | 8.1

Diesel Particulate Filters

[81] Stobbe P, Petersen HG, Hoj JW, Sorenson SC. SiC as a substrate for diesel particulate filter. SAE 932495; 1993. [82] Young DM, et al. Silicon carbide for diesel particulate filter applications: material development and thermal design. SAE 2002-01-0324; 2002. [83] Mey D, et al. Improved DPF substrate for washcoat accommodation. SAE 2009-01-0288; 2009. [84] Maier HR, Schumacher U, Best W, Schaefer W. Porous permeable moulded body. US 5.853.444, filed 29.12.1998 (EP Priority 23.03.1996). [85] Maier HR, Best W, Schumacher U, Schaefer W. Characteristics and design of diesel soot filters based on direct electrical regeneration. cfi/Ber. DKG 1998;75(5):25e9. [86] Takahiro T, Yuichiro T, Shuichi I, Takashi H. Silicon carbide based porous article and method of preparation thereof. PCT WO 02/081406, filed 17.10.2002. [87] Miwa S, et al. Diesel particulate filters made of newly developed SiC. SAE 2001-01-0192; 2001. [88] Ichikawa S, Uchida Y, Kaneda A, Hamanaka T. Durability study on SieSiC material for DPF (2). SAE 2004-01-0951; 2004. [89] Harada M, et al. Durability study on SieSiC material for DPF (3). SAE 2005-01-0582; 2005. [90] Schaefer-Sindlinger A, et al. Engine bench and vehicle durability tests of Si bonded SiC particulate filters. SAE 2004-01-0952; 2004. [91] Tsuneyoshi K, Hayama S. A new SiC-based DPF for the automotive industry. Dearborn: Diesel Engine-Efficiency and Emissions Research (DEER) Conference; August 2008. [92] Tsuneyoshi K, Tagaki O, Yamamoto K. Effects of washcoat on initial PM filtration efficiency and pressure drop in SIC DPF. SAE 2011-01-0817; 2011. [93] Hajireza S, Lundorf P, Wolff T. XP-SiC: an innovative substrate for future applications with low weight and high porosity. Dearborn: Diesel Engine-Efficiency and Emissions Research (DEER) Conference; August 2009. [94] Wolff T, Friedrich H, Johannesen LT, Hajireza S. A new approach to design high porosity silicon carbide substrates. SAE 201001-0539; 2010. [95] Westerheide R, Adler J. Advances in hot gas filtration technique. In: Heinrich JG, Aldinger F, editors. Ceramic materials and components for engines. Wiley-VCH Weinheim; 2001. p. 73e7. [96] Joulin JP, Pourchet F, Courty P, Dementhon J-B. Monolithic honeycomb structure made of porous ceramic and use as a particle filter. US 6.582.796, filed 21.7.2000. [97] CTI begins production of diesel particulate filters. Bull Am Ceram Soc 2005;84(12):7. [98] Adler J, Teichgra¨ber M, Klose T, Sto¨ver H. Poro¨se Siliciumcarbidkeramik und Verfahren zu ihrer Herstellung. DE 197.27.115; 01.07.1996. [99] Rembor HJ, Rahn T. A new approach for a diesel particle filter material with liquid phase sintered silicon carbide and an innovative segmented geometry. SAE 2010-01-0532; 2010. [100] Miyakawa N, Sato H, Maeno H, Takahashi H. Characteristics of reaction-bonded porous silicon nitride honeycomb for DPF substrate. JSAE Rev 2003;24:269e76. [101] Wallin SA, Prunier AR, Moyer JR. Mullite bodies and methods of forming mullite bodies, US 6.306.335, filed 27.8.1999.

605

[102] Li CG, et al. Properties and performance of diesel particulate filters of an advanced ceramic material. SAE 2004-01-0955; 2004. [103] Li Ch, Mao F, Zhan R, Eakle S. Durability performance of advanced ceramic material DPFs. SAE 2007-01-0918; 2007. [104] DOW AUTOMOTIVE Press Release, June 2006. [105] DOW Press Release, December 19, 2007. [106] Mayer A, Buck A. Knitted ceramic fibers e a new concept for particulate traps. SAE 920146; 1992. [107] Miller RK, et al. Design, development and performance of a composite diesel particulate filter. SAE 2002-01-0323; 2002. [108] Ido T, Kunieda M, Ogyu K, Ohno K. Fundamental study and possible application of new concept honeycomb substrate for emission control. SAE 2007-01-0658; 2007. [109] Zuberi B, et al. Advanced high porosity ceramic honeycomb wall flow filter. In: Diesel Engine-Efficiency and Emissions Research (DEER) Conference, Detroit, August 2007. [110] Zuberi B, et al. Advanced high porosity ceramic honeycomb wall flow filters. SAE 2008-01-0623; 2008. [111] Gabathuler JP, et al. New developments of ceramic foam as a diesel particulate filter. SAE 910325; 1991. [112] Reichle M. Partikelverminderung im Abgas von Stadtlinienbussen e Grundlagenentwicklung. BMFT TV8436 A5; 1987. [113] Claussen M. Katalytische Russfilter fu¨r Dieselaggregate und Rapso¨lmotoren. Nr. 29. Clausthal-Zellerfeld: Cutec-Schriftenreihe; 1997. [114] Labhsetwar N, et al. Application of catalytic materials for diesel exhaust emission control. Curr Sci 2004;87(12):1700e4. [115] Pontikakis GN, Koltsakis GC, Stamatelos AM. Dynamic filtration modelling in foam filters for diesel exhaust. Chem Eng Commun 2001;188:21e46. [116] Kolke R. Partikelfilter und deren Beitrag zur Emissionsminderung von Dieselfahrzeugen. In: Herausforderung Abgasnachbehandlung fu¨r Dieselmotoren. Dresden: 2. Konferenz FAD; 2004. p. 189e200. [117] Frank W, Himmen M, Huethwohl G, Neumann P. Sintered metal particulate filter to meet life cycle requirements of commercial vehicles. In: Steinmetz E, editor. Minimierung der Partikelemissionen von Verbrennungsmotoren. Essen: Expert-Verlag, ISBN 3-8169-2430-1; 2004. p. 293e310. [118] Bru¨ck R, et al. Metal supported flow-through particulate trap; a non-blocking solution. SAE 2001-01-1950; 2001. [119] Brillant S, Zikoridse G. Metal fibre diesel particulate filter: function and technology. SAE 2005-01-0580; 2005. [120] Miebach R, Pelzer W, Kolbeck C, Figoutz S. Vorrichtung zur Entfernung von Russpartikeln aus einem Abgasstrom von Verbrennungsmotoren. EP 1.515.012; 11.09.2003. [121] Naumann D, Bo¨hm A, Saberi S, Timberg L. Application of hightemperature-stable and corrosion-resistant INCOFOAM in diesel exhaust aftertreatment systems. In: Herausforderung Abgasnachbehandlung fu¨r Dieselmotoren. Dresden: 3. Konferenz FAD; 2005. p. 171e8. [122] Koltsakis GC, et al. Metal foam substrate for DOC and DPF applications. SAE 2007-01-0659; 2007. [123] Jeuland N, et al. Performances and durability of DPF tested on a fleet of Peugeot 607 taxis e first and second test phases results. SAE 2002-01-2790; 2002. [124] PSA Peugeot Citroen Press Release, 18.02.2005.

606

[125] [126] [127] [128] [129] [130] [131]

IBIDEN Press Release, 18.10.2002. IBIDEN Press Release, 22.07.2004. IBIDEN Press Release, 02.03.2006. NGK Press Release, 26.11.2003. CORNING Press Release, 08.04.2004. CORNING Press Release, 23.04.2005. Frost&Sullivan. European market for next generation diesel engine technologies, B389e18, December 2004. [132] Frost&Sullivan. Executive assessment of the European market for diesel exhaust after-treatment technologies, M455e18, December 2009.

Handbook of Advanced Ceramics

[133] Frost&Sullivan. North America light vehicle exhaust emission control system markets, F492e18, 2006. [134] Frost&Sullivan. Analysis of diesel powertrain outlook and technology roadmap in North America, N721e18, February 2011. [135] EMITEC Press Release, 21.06.2006. [136] Vonarb R, Pichon G. EXOCLEAN DPF system: review of performance of an active DPF system on Renault Refuse Trucks. In: Herausforderung Abgasnachbehandlung fu¨r Dieselmotoren. Dresden: 3. Konferenz FAD; 2005. p. 151e7. [137] http://www.bafu.admin.ch/partikelfilterliste/index.html (last updated 28.09.2011).

Chapter 9.1

Mechanical Properties of Ceramics Robert Danzer,* Tanja Lube,* R. Morrellz and P. Supancicy *

Institut fuer Struktur- und Funktionskeramik (Institute for Structural and Functional Ceramics), Montanuniversitaet Leoben (University of Leoben), Peter-Tunner-Strasse 5, 8700 Leoben, Austria, y Materials Center, Institut fuer Struktur- und Funktionskeramik, (Institute for Structural and Functional Ceramics), Montanuniversitaet Leoben (University of Leoben), Peter-Tunner-Strasse 5, 8700 Leoben, Austria, z Materials Division, National Physical Laboratory, Teddington, Middlesex, TW11 0LW, United Kingdom

Chapter Outline 1. Important Properties 2. Elastic Properties 3. Plasticity, Compressive Strength, Yield Strength, and Hardness 4. Thermal Strains and Thermal Stresses 5. Thermal Diffusion

609 610 613 614 617

1. IMPORTANT PROPERTIES Ceramic materials behavedin generaldas linear elastic, that is, there exists a linear (but tensor) relationship between the components of the stress tensor and the components of the strain tensor (Hooke’s law). For a one-dimensional bar, the relationship between the stress s and the strain ε is [1] s ¼ E ε:

(1)

The elastic constant E is Young’s modulus. Although inelastic (plastic) deformations occur at very high stresses (i.e. compressive stresses in the order of some E/100), plasticity generally does not have to be considered in the stress analysis of components, because the strength limit is exceeded before the yield stress is reached [2,3]. The yield strength of ceramic materials is strongly related to the hardness, which is an important property if wear becomes important. Any successful application of ceramics needs a careful design against occurrence and propagation of cracks, which can be described in the framework of linear elastic fracture mechanics [4,5]. Failure is caused by the spontaneous extension of cracks, which can be described by the Griffith/ Irwin criterion. It occurs if the stress intensity factor (SIF) pffiffiffiffiffiffi K ¼ sY pa (2) reaches or exceeds the fracture toughness Kc: K  Kc :

(3)

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00033-2 Copyright Ó 2013 Elsevier Inc. All rights reserved.

6. Toughness 7. Strength 8. Strength Degradation with Time 9. Final Remarks References

619 622 625 628 629

In Eqn (2), s is the stress in the uncracked body,1 Y is a geometric factor, which describes the geometry of the crack, the specimen and the stress field and a is the crack length. The SIF describes the loading of the crack (more precisely the stress field at the crack tip), and the fracture toughness is the resistance of the material against crack extension. Note that the energy necessary to create 1 m2 of newly cracked area is Gc ¼ Kc2 =E;

(4)

which is called the specific fracture energy. In service, temperature changes often occur in components. This causes thermal strains and stresses, which may nucleate and extend cracks in the component [6,7]. Therefore, all properties having any influence on these strains and stresses are of highest relevance for the safe operation of ceramic components. If a material undergoes a temperature change DT ¼ T0 T (from the reference temperature T0 to the temperature T ), a thermal strain εth ¼ a$DT

(5)

occurs, where a is the coefficient of thermal expansion (CTE). If thermal strains are mechanically constrained, thermal stresses come into existence, which (at the surface

1. For simplicity, a uniaxial stress state is assumed if not mentioned otherwise. For general stress fields, s is a parameter, which characterizes the amplitude of the field.

609

610

Handbook of Advanced Ceramics

of a body and for complete constraint) reach a level as high as sth ¼

a E$DT : 1 y

(6)

The parameter E=ð1 yÞ is the elastic constant for plain stress and y is Poisson’s ratio. Even if the thermal strains are not completely constrained, the thermal stresses scale with the stress given in Eqn (6). Therefore, the material parameter a E=ð1 yÞ is relevant for the thermal stresses in ceramic components,2 because it determines the maximum stress per degree Celsius temperature difference. These thermal stresses disappear in bulk materials if the temperature differences in the component become zero (Eqn (6)).3 This is promoted by a high thermal diffusivity. Therefore, ceramic materials with a high thermal diffusivity are less prone to thermal shock damage than are those with a low thermal diffusivity. Even though temperature changes are the most frequent reason, strain and stress mismatches can also occur for many other reasons. Examples are phase transformations, the switching of twins or domains, and dimensional changes due to diffusion processes. Although these processes may become very important for functionality and even for structural integrity of ceramic components, they are very special and will not be discussed in this overview. The most important property of ceramics used in design is strength. While the compressive strength of a ceramic may be much higher than the tensile strength, from a factor of about five for porous or microcracked materials to as much as 20 for very strong materials with uniform finegrained microstructures [3], it is always difficult to use a ceramic entirely in compression to take advantage of this. Typically, compressive loading involves also shear and tensile stresses in some regions of a component, and thus, one always has to keep an eye on these aspects in designing an application. Under completely constrained conditions, it is possible to induce stresses high enough to cause plastic deformation and yield. An example is the loading situation under a hardness indenter. Properties under these conditions are of interest, for example, in ballistic performance of ceramic used for armor applications. Such inelastic behavior is dealt with in the section on hardness.

2. This is true for thermal stresses caused by thermal strains in a single material component. If temperature changes occur in systems made of a mix of materials, the situation is more complex. Here, thermal stresses scaledin generaldwith the difference (mismatch) of the CTEs of the involved materials. 3. Note that commonly temperature gradients developed during rapid cooling from the usual high-temperature sintering process can also cause the development of thermal strains and stresses, especially in larger parts.

Tensile strength is limited by the brittle extension of structural discontinuities, often termed ‘flaws’, which behave like cracks and which exist within the component and on its surface. From the fracture criterion (Eqns (2) and (3)), it follows that the tensile strength of a component is sf ¼

Kc pffiffiffiffiffiffiffiffiffi; Y p ac

(7)

where ac is the size of the fracture origin (the flaw where fracture starts). Because the size of the fracture origin differs from component to component, the strength of individual components is also different. This makes a statistical description of strength necessary (e.g. using the so-called Weibull theory [8,9]). It also causes the wellknown size effect of strength. The probability of finding a large flaw in a large component is higher than in a small component. Therefore, the mean strength of large components is generally lower than that of small components when loaded in the same way. Of course, there also exist damaging mechanisms, which degrade the strength with time. Very important is the so-called subcritical crack growth (SCCG) [10], where a crack slowly grows, even under a constant load. SCCG always precedes brittle fracture and may cause a significant loss of strength with time. The reason is that crack growth is a thermally activated process that can be enhanced, especially in oxide materials, by the presence of water at the crack tip, which reduces the energy barrier to crack growth [11]. A second mechanism is caused by repeated (cyclic) loading (fatigue crack growth) and is thought to result from the local enhancement of stresses at the crack tip because crack faces do not completely separate or close. Fatigue effects are even possible high under compressive loading [12].

2. ELASTIC PROPERTIES The linear elastic behavior of materials is described by Hooke’s law (Eqn (1)), which, in its general form, linearly relates the components of the stress tensor sij with the components of the strain tensor εkl [1]: sij ¼ Cijkl εkl ;

(8)

where Cijkl is the tensor of the elastic constants [13]. It holds the Einstein convention, that is, a summation has to be done over the same indices. To give an example, a stress in the z-direction ðszz Þ isdin generaldrelated to nine strain components ðεxx ; εyy ; εzz ; εxy ¼ εyx ; εyz ¼ εzy ; εzx ¼ εxz Þ and to nine elastic constants ðCzzkl Þ. Symmetry can be used to reduce the number of independent constants. In cubic crystals, only three independent constants exist. For single

Chapter | 9.1

Mechanical Properties of Ceramics

611

crystals, the tensor behavior described above has to be accounted for. Many polycrystalline ceramic materials and glasses behave macroscopically isotropic. Then, the number of independent elastic constants is only two. Some important elastic constants are Young’s modulus (E), which relates the stress with the strain in the stress direction ðsii ¼ E$εii Þ, and the shear modulus (G), which relates the shear stress with the shear strain ðsij ¼ G$εij ; isjÞ. Alternatively, Poisson’s ratio ðyÞ which is the ratio of the strain components perpendicular to the stress to the strain components in stress direction: y ¼ εjj =εii ¼ εkk =εii 4 can be used. Because in homogeneous materials only two constants are independent, there exists a relationship between these constants: G ¼

E : 2 ð1 þ yÞ

(9)

Some ceramic materials have a texture, and therefore, the tensorial elastic behavior has to be appropriately taken into account. For example, elongated grains may be oriented by slip or tape casting [14,15]. Other examples are piezoceramics, where the texture changes with the applied electrical field [16,17]. It should be noted that the effective symmetry of the polycrystalline ceramic may differ from the symmetry of the underlying single crystals. The elastic properties of ceramics are directly related to the bonding between the atoms and to the placement of the atoms in the body [16,17]. Today, elastic constants can be determined from first principle quantum mechanical calculations using density functional theory. The interaction between neighboring atoms can be described by a potential P (Figure 1). The minimum of the potential corresponds to the equilibrium distance (d0) between the atoms (at zero Kelvin). The first derivative of the potential (P) with the distance (d) is the force and the second derivative the curvature, whichdin the neighborhood of the equilibrium distancedis almost constant. This value corresponds to the stiffness of a spring and is called bond stiffness ðv2 P=vd2 Þd¼d0 ¼ S0 [18]. Let us model the elastic behavior of a solid by neighboring atoms connected by springs (Einstein model) [19]. Then, stretching or compressing two neighboring atoms would require the force F ¼ S0 ðd

d0 Þ:

(10)

In the cross-section of a body, there are N ¼ ð1=d0 Þ2 atoms per square meter and N “springs” per square meter are bridging each cross-section. The force per square

4. Note that a stress in the z-direction causes strains in the z-direction as well as in the x- and y-directions. Typical values of the Poisson ratio are between 0.2 and 0.3.

FIGURE 1 Schematic sketch of the binding potential between two atoms (full line). d0 is the equilibrium distance at zero Kelvin. The curvature at the equilibrium distance is the bond stiffness. The dotted lines indicate energy levels, and the dashed line marks the mean positions of these states.

meter (i.e. the stress) needed to pull these “springs” apart is N F ¼ s ¼ S0 ðd d0 Þ=d02 . The elongation ðd d0 Þ divided by the starting distance is the strain: ðd d0 Þ=d0 ¼ ε. Therefore, we get s ¼ ðS0 =d0 Þ ε ¼ E ε, and thus Young’s modulus is E ¼

S0 : d0

(11)

Note that a stronger binding between the atoms results in a deeper potential well, which has a larger bond stiffness (curvature) S0. Therefore, Young’s modulus increases with bond strength. Of course, this model only applies if the application of a stress causes stretching or compressing of bonds. This generally happens if the atoms are surrounded by four or more neighbors. If the number of neighbors is lower, as in the case of oxygen in silica glass, elongations can by made by changing the bond angle [20] and Eqn (11) may overestimate the modulus. It is obvious that this simple model is only valid for very small strains ðd d0 Þ=d0 , that is, at distances where the potential can by modeled by a quadratic function P z S0 ðd d0 Þ2 . At higher distances, the potential becomes asymmetric and nonlinear terms become important. Therefore, the linear elastic regime is restricted to small strains, that is, strains less than a few percent. Because the elastic strain at brittle fracture (which occurs at tensile loading) is typically a few per mille, this limit can hardly be reached in tensile loading.5 But in compression, strains of a few percent are possible, and nonelastic (plastic) strains can also occur in ceramic materials. This point will be addressed in the next section. 5. In some ceramic materials, nonelastic (plastic) deformations may also occur at relatively small strains due to the switching of twins or domains (e.g. in piezoelectric ceramic materials, lead zirconate titanate, PZT) or due to phase transformations (e.g. in partially stabilized zirconia PSZ).

612

In crystals, the bond stiffness as well as the equilibrium distance change with the direction. Therefore, the elastic modulus (the elastic constants) is not isotropic, and it depends on the crystal orientation. In principle, the approximation of Eqn (11) is made for zero Kelvin, where the ground state near the energy minimum (Figure 1) is occupied. At higher temperatures, higher energy states are also occupied. Here, the asymmetry of the potential curve becomes important and causes a slight increase of the distance d0, which should cause a slight decrease of the modulus with increasing temperature. Such behavior is also observed experimentally [21]. For ceramic phases, which have mixed ionic/covalent bonding, the bond stiffness is relatively high, between 15 N/m and 200 N/m [18]. The distance between bonded atoms typically is 2$10 10 m to 3$10 10 m [18]. Therefore, Young’s modulus of ceramic phases is expected to be between 50 GPa and 1000 GPa. In ceramic materials, the microstructure will also have some influence on the modulus. In polycrystalline ceramics, the net properties are an average over the orientation of the individual crystals. For multiphase materials, the average over the microstructural elements lies between the bonds given by the Voigt- (equal strains) and the Reuss- (equal stresses) rules of mixture. Porosity, which for some ceramic materials is an important microstructural constituent, has a significant influence on elastic modulus. The influence of porosity on the modulus of course depends on the shape and the distribution of the pores. For sintered ceramics, a simple estimate gives E z E0 ð1 b PÞ, where E0 is the Modulus of the pore free ceramic, P is the porosity volume fraction, and b is an empirical constant in the range of b ¼ 2 [22]. There are many routes to measure elastic modulus [23e27]. Static bending of a thin bar can be used, but it requires precision in loading and measurement of very small displacements. It is difficult to achieve an accuracy of Young’s modulus to better than 5% [28]. Using strain gauges can achieve greater accuracy, and lateral strain measurement can be used to determine Poisson’s ratio. Ultrasonic wave velocity measurements can be used, and the moduli calculated from the density and the compression and shear wave velocities. This method is the best for coping with anisotropy to determine the tensor stiffness of a uniformly anisotropic material. Perhaps the most reliable method of determining the properties of isotropic materials is based on the natural vibration frequencies (eigenfrequencies) of a bar or a disc of very regular shape, which can be excited by either driving in resonance or impacting the test piece and determining the decaying vibration frequencies. These methods are readily extended to hightemperature measurement. An accuracy of better than 0.1% is achievable provided the shape is prepared to small dimensional tolerances. It should be noted that the dynamic methods of measuring modulus yield slightly higher values

Handbook of Advanced Ceramics

than in the quasistatic loading method, because the latter is essentially isothermal, and the others are adiabatic [29]. Some elastic properties are listed in Table 1. Covalently bonded materials have high bond strength and high bond stiffness. Therefore, their modulus is also high (Eqn (11)). Engineering ceramics have a very high Young’s modulus (150 GPae1100 GPa), a little higher than metals and alloys (40 GPae500 GPa) and much higher than engineering polymers (0.1 GPae6 GPa). The stiffest material known is diamond. Our measurements on a slab of polycrystalline diamond give E ¼ 1080  10 GPa and y ¼ 0.09  0.01. Porous ceramic materials (rocks, pottery, etc.) have a modulus between 30 GPa and 200 GPa. The elastic properties are design parameters for applications, which require a certain flexibility (e.g. springs) or stiffness (e.g. beams). In many cases, a combination of several material parameters can be defined to be a figure of merit, whichdto optimize the performance in a particular applicationdshould be as high as possible. Some very instructive examples on this topic can be found in the book “Materials selection in mechanical design” by Ashby [34]. Examples for figures of merit are E1=2 =r for bent beams with minimum weight, s2f =E for springs with minimized volume TABLE 1 Young’s Modulus and Poisson Ratio for Some Materials at Room Temperature [2,30e33] Young’s modulus GPa

Poisson ratio

Diamond-based materials

900e1100

0.08e0.12

Cemented carbides

400e600

0.17e0.20

Silicon carbides

300e400

0.8e0.9

Silicon nitrides

300e350

0.22e0.275

Alumina

250e400

0.21e0.27

Zirconia

150e250 3Y-TZP: 208

0.2 3Y-TZP: 0.32

Glasses

40e80

0.17e0.30

Steels

180e250

0.29e0.31

Aluminum alloys

60e90

0.33e0.35

Magnesium alloys

40e50

e

Nylons

2e5

e

Epoxies

1e3

0.33

High-density polyethylene

0.4e0.9

~0.4

Low-density polyethylene

0.1e0.3

~0.4

Material

Chapter | 9.1

Mechanical Properties of Ceramics

613

(ceramic springs are used at ambient and high temperatures and for operation in chemical aggressive environments), s2f =Er for springs with minimized weight, sf =E for best performing elastic hinges and seals, s3f =E2 and E, respectively, for knife edges and pivots, which can bear a maximized load or have a minimized contact area respectively, where sf is the strength of the material and r is the mass density. Note that depending on the application, very high as well as very low values of Young’s modulus can be desirable.

3. PLASTICITY, COMPRESSIVE STRENGTH, YIELD STRENGTH, AND HARDNESS Plastic deformation caused by the movement of dislocations is well recognized for metals. The yield strength of pure metals is very low, and strengthening mechanisms have to be tailored to give metals a usable strength [35]. Ceramic materials have a hybrid (mixed ionic/covalent) bond, so that binding orbitals of electrons are localized around the corresponding ion cores, that is, the electrons are immobile. This is the main reason why the generation and movement of dislocations in ceramics needs much more energy than in metals, where the valence electrons are mobile. Plastic deformation of ceramics is possible but only at very high stresses (or at relatively high temperatures), and furthermore often a result of there being a glassy second phase in the microstructure. In tensile loading of ceramics, brittle fracture in general occurs before severe plastic deformation takes place and the (plastic) fracture strain is (almost) zero. But in compressive loading, brittle fracture is suppressed, and some plastic deformation of ceramic materials becomes possible [3]. The compressive strength is connected to the formation of glide bands and the pile-up of dislocations at grain boundaries, which may locally cause large tensile or shear stresses, and the generation of microcracks, which initiates the failure. Although other mechanisms may also be important, the yield strength ðsy Þ is roughly equal to the compressive strength ðsc Þ of most ceramics. Therefore, in general, it holds that sy zsc [36]. One example where plasticity is commonly present is in the testing of hardness. Hardness is a measure of the resistance of the material against plastic deformation. There exist numerous variants for hardness testing, but for ceramic materials, the Vickers or Knoop hardness test is generally used. Here, a diamond pyramid is pressed with the force F onto the surface of the material, which produces a square (kite-shaped in the case of Knoop indenters) indentation in the surface (experimental details will be discussed later). The so-called true hardness (or pressure hardness) is defined by H ¼

F A

(12)

where A is the projected area of the impression. Around and under the indentation, the material is plastically deformed. There exists a relationship between the hardness and the yield strength [36]: H z 3 sy :

(13)

In principle, all strengthening mechanisms known from metals also apply for ceramics. Therefore, they can be used to increase the hardness of ceramics. There are some other reasons for inelastic deformations of ceramics than the movement of dislocations. A prominent example is the family of piezoceramic perovskites, where a cubic high-temperature phase transforms to a tetragonal and/or a rhombohedric low-temperature phase (which is ferroelastic). Let us restrict the discussion to the tetragonal phase. Compared with the cubic phase, the tetragonal crystals are elongated in the direction of one of the three cubic orientations. During cooling from the sintering temperature, tetragonal crystals of all three possible orientations are built in the cubic mother crystals. If a stress is applied, inelastic (plastic) strains can arise by switching the orientation of tetragonal subcrystals, corresponding to a movement of domain walls. It is interesting to note that because the tetragonal crystals have a dipole moment, a switching of the crystal orientation can also be caused by the application of an electric field (i.e. ferroelectric effect). This causes the high piezoelectric effect (length change due to the application of an electric field) of this class of materials [37]. Hardness tests for ceramics have been standardized, based on experimental experience. There are two principal issues of concern: one is an apparent indentation force dependence of hardness, especially at low force levels, and the other is the damage (cracking, spallation) associated with indentation, which makes the measurement of indentation size unreliable or impossible (Figure 2). For Vickers

FIGURE 2 HV5 indentation in alumina. Typical problems arising with hardness testing of ceramics become apparent: radial cracks emanating from the indentation corners, secondary cracks, lateral cracking (indicated by the halo around the indent), unclear corners. For color version of this figure, the reader is referred to the online version of this book.

614

Handbook of Advanced Ceramics

indentation, 1 kgf indentation force is preferred [38e41] to minimize damage around indentations while maximizing the size of indentation for measurement accuracy. The hardness value is conventionally computed as force divided by the actual area of the indentation surface, which is related to the diagonal size d through HV ¼ 1:8455 F=d2 with no units (kgf/mm2 implied), or in gigapascals. For Knoop indentation, the test force is recommended to be 1 kgf or 2 kgf [38,40,42], and the hardness value is computed as HK ¼ 14:22 F=d2 with, in this case, d being the long diagonal length, and with similar units. Less frequently, hardness from other types of tests may also be reported, such as Rockwell tests [40], or depth sensing hardness (e.g. Martens hardness). Hardness values of ceramic materials in comparison to those of other materials are listed in Table 2. Compared to polymers and even to metals, ceramic materials are very hard. The hardest material ever tested is diamond. Hardness is a key property in all applications related to wear. The wear mechanisms of ceramic materials can be classifieddwith increasing wear ratedin wear caused by (i) atomic adhesion, (ii) chemical reactions, (iii) grooving and (iv) material fragmentation [44]. A high hardness especially reduces the grooving mechanism, sometimes tested via a ‘scratch hardness’ test. Because ceramic materials aredin generaldalso chemically very stable

TABLE 2 Yield Strength and True Hardness of Some Materials at Room Temperature. Note that Hz3 sy [30,32,43] Material

Yield Strength (MPa)

True Hardness (MPa)

Diamond-based materials

35,000

20,000e70,000

Cemented carbides

1000e3000

3000e20,000

Silicon carbides

5000e9000

15,000e20,000

Silicon nitrides

4000e7000

12,000e21,000

Alumina

3000e5000

9000e15,000

Zirconia

1000e5000

3000e15,000

Glasses

600e3000

1800e5000

Mild Steels

300e600

900e1800

High-speed steels

500e1000

1500e3000

7075 Aluminum alloys

300e700

900e2100

Nylons

60e120

180e360

Epoxies

40e90

120e270

High-density polyethylene

15e50

45e150

Low-density polyethylene

8e20

24e60

(which restricts adhesion and chemical reactions), they are an ideal solution for wear protection either as a bulk product or as a coating on a softer substrate. Prominent ceramic materials used in wear protection are silicon nitride ceramics, alumina ceramics, silicon carbides, cemented carbides, and diamond-based materials. Typical applications, where wear reduction is of the highest interest are knives for industrial cutting, indexable inserts for metal cutting, ceramic tiles for wear protection of paper-making machines, or yarn guides in looms. Ceramic materials have also been used for many years for knife edges and pivots, for example, in mechanical watches, where the high yield strength causes an excellent load bearing capacity. The corresponding figure of merit is s3y =E2 . Ceramic materials are used for face seals ðsy =EÞ and bearings, where the high yield strength of the material is used. Such seals, for example, based on silicon carbides, can also be used at very high temperatures (up to 1400  C).

4. THERMAL STRAINS AND THERMAL STRESSES 4.1. Thermal Strains Generally, a change of temperature induces a dimensional change (i.e. expansion or shrinkage) of solids [6] generally described by a displacement or strain field. In crystals, the thermal strains εth are described by a (symmetric) secondorder tensor, εth;ij ¼ aij $DT;

(14)

and therefore, the shape of an initially spherically shaped (unconstrained) solid will change to an ellipsoid due to a temperature change. Note that a change of the coordinate system into the principal axes leads to a set of three fundamental heat expansion coefficients aii with respect to the principal axes in the solid. For isotropic materials, this will result in a spherical tensor, where the essential parameter of the heat expansion is given by one single heat expansion coefficient a 0 1 1 0 0 aij ¼ a$@ 0 1 0 A: (15) 0 0 1 The trace of aij gives the volume (bulk) heat expansion coefficient [13]. Care has to be taken on the definition of the coefficient of linear thermal expansion (CTE), because two different ones are used. The technical (or mean) CTE is defined as the relative change in length Dl ¼ l l0 =l0 (l: length at temperature T; l0: initial length at the reference temperature T0) resulting from a temperature change: atech ðT; T0 Þ ¼

1 Dl $ l0 ðT T0 Þ

(16)

Chapter | 9.1

Mechanical Properties of Ceramics

615

FIGURE 3 (a) A Schematic sketch of the thermal strain versus temperature and (b) of the technical and differential CTE. In most cases, the curve in (a) can nicely be fitted by a second-order polynomial. Then, the slope of the differential CTE is twice that of the technical CTE.

This CTE is suitable for engineering purposes, because it is the mean CTE in a given temperature interval (i.e. secant coefficient). Additionally, this quantity is the natural result of thermal expansion measurements, which are performed on bar-shaped specimens by instruments such as dilatometers, recording the length change with respect to the temperature change. The linear (or differential) CTE, sometimes termed expansivity, is defined by the relative length change due to an infinitesimal temperature increase [45,46]: adiff ðTÞ ¼ Lim

T0 /T

1 Dl 1 dl ¼ $ $ $ l0 ðT T0 Þ l dT

(17)

Using the differential CTE the thermal strain within linear thermoelasticity is ZT aðT 0 Þ$dT 0 : (18) εth ðT; T0 Þ ¼ T0

Figure 3 shows a typical shape of the thermal strain with the temperature and the difference between the differential and the technical CTEs. In an analysis of thermal strains and stresses, the accurate description of the temperature dependence of the CTE can be of high relevance and a simple approach, which ignores this dependence, can be misleading. In that respect, a useful approximation is presented in the following. Let us restrict our analysis to homogeneous and isotropic solids. Remember that the length change DlðT; T0 Þ ¼ lðTÞ l0 is caused by the temperature change DT ¼ T T0 . In most cases, the length change can nicely be fitted by a simple second-order polynomial6: DlðT; T0 Þ ¼ ath $DTðT; T0 Þ þ bth $DTðT; T0 Þ2 : l0

(19)

Then, the technical CTE simply is atech ðT; T0 Þ ¼ εth =DTðT; T0 Þ ¼ ath þ bth $DTðT; T0 Þ; (20) 6. In some cases, a quadratic fit is not sufficient, and a higher order (e.g. third order) or other nonlinear ansatz function is necessary.

and for the differential CTE ðT / T0 Þ we get adiff ðTÞ ¼ dεth =dT ¼ ath þ 2bth $DTðT; T0 Þ:

(21)

Of course, the coefficients for the power series and the technical CTE depend on the reference temperature, but the differential CTE does not. In this approximation, the CTE linearly increases with the temperature difference, and the slope of the differential CTE is twice that of the slope of the technical CTE. Similarly to Young’s modulus, the CTE is strongly related to the shape of the binding potential between the neighboring atoms. At temperatures above zero Kelvin, higher energy states above the ground state are occupied. Because the potential is asymmetric (nonparabolic) and wider at higher energy states (compared with the ground state), the mean position is shifted to a larger distance between the atoms (see the dashed line in Figure 1). For atoms with four or more neighbors, most of this increase in distance is translated into a length increase of the solid body. If atoms in a crystal are lower coordinated, the unit cell is not perfectly filled and contains a large fraction of empty space. Then, bond angles can be changed, atoms can be shifted into the empty spaces, and the CTE can be smaller than expected from the binding potential asymmetry. Of course, the shape of the potential and the arrangement of the atoms in the crystal depend on the direction in the crystal. Therefore, the CTE of noncubic crystals is anisotropic. But in polycrystalline ceramic materials without overall texture and in glasses, an average over all direction occurs, and the CTE appears macroscopically isotropic. The influence of microstructure on the CTE is small, and simple rules of mixture can, in general, be used to account for the influence of several microstructural constituents. If the grains in polycrystalline ceramics have some texture, this has to be considered to describe the anisotropy of heat expansion correctly, and dependent on the orientation, the CTE has three different values. The measurement techniques for thermal elongations and CTEs using dilatometry are well established since a long time [47e50]. A good description can be found in

616

Handbook of Advanced Ceramics

[51]. Although the CTEs of stiff materials may be low, they still generate significant thermal stresses when a component is subjected to temperature gradients and are thus highly relevant for thermal stress analysis. Therefore, the measurements generally have to be performed with very high precision. To give an example, for some designs, a precision in CTE of 0.1 ppm/K is necessary. If the temperature range of interest is DT ¼ 300 K and the specimen length is l0 ¼ 25 mm (if specimens are cut out of components, they can even be shorter), this corresponds to a length change of the specimen during the experiment as small as 0.75 mm. This is of the order of magnitude of the roughness of machined surfaces. Therefore, a key problem of high precision dilatometer measurements is the proper preparation of the specimens, which has to be done with highest precision. The parallel ends of the specimens should be very smooth (polished) and must be exactly perpendicular to the specimen’s axis. Some CTE values of engineering materials are listed in Table 3. It can be recognized that ceramic materials and glasses have small CTEs, smaller than those of most metals and alloys and even much smaller than those of polymers, for which the CTE is around two orders of magnitude greater than for ceramics. For some zirconia ceramics and

TABLE 3 CTE and the Thermal Stress Parameter of Some Materials at Room Temperature [2,30,32] Material

Differential CTE10 6/K

TSF MPa/K

Diamond-based materials

1e2

1e1.5

Cemented carbides

5e6

3.5e4.5

Silicon carbides

4e5

1.5e3

Silicon nitrides

2.5e4

0.8e1.2

Alumina

7e9

2e3

Zirconia

6e12

1,5e3

Glasses

3e5

0.2e0.6

Steels

12e18

3e4

Aluminum alloys

20e30

1.8e3

Copper alloys

20e30

2e3

Nylons

80e150

0.3e0.5

Epoxies

60e100

0.1e0.3

High-density polyethylene

130e200

0.07e0.15

Low-density polyethylene

150e250

0.03e0.07

some steels, the CTE is (almost) equal, which makes a joining with a reasonably small thermal strain mismatch possible.

4.2. Thermal Stresses Thermal stresses in a body come into existence, if thermal strains are constrained. They are proportional to the thermal strains, to the appropriate elastic constant E0 , and to a factor Sth , which is between zero and one and which depends on the constraint [7]. Ifdto give an exampledthe ends of a one-dimensional bar are free to move, a temperature increase causes an elongation and a thermal strain but no thermal stress (there is no constraint) in the bar. If the ends are clamped, a temperature increase causes no elongation but a thermal strain and a thermal stress (there is full constraint). In general, the thermal stress is sth ¼ εth ðT; T0 Þ$E0 $Sth ¼ atech ðT; T0 Þ$E0 $DT$Sth ¼ sth;0 $Sth ;

(22)

For a uniaxial stress state, as it occurs in the one-dimensional bar discussed above, it holds that E0 ¼ E. For a plane stress state, as occurs during quenching of components on their surfaces, it holds that E0 ¼ E=ð1 yÞ. For the bar with free ends, there is no constraint (Sth ¼ 0), and for the bar with clamped ends, the constraint is complete (Sth ¼ 1). The sth;0 is the thermal stress, if the strains are completely constraint (for Sth ¼ 1). From Eqn (22), it follows that thermal stresses scale with the material parameter atech E=ð1 yÞ ¼ fs;th, which we call thermal stress factor (TSF). It is interesting to note that for dense (non-porous) metallic and ceramic materials, in which atoms are highly coordinated ðK  4Þ, the values of the TFS are in a narrow range [18]: E (23) z 2 3 MPa K 1 : fs;th ¼ atech $ 1 y This rule of thumb can also be related to the shape of the binding potential. There is a tendency that the CTEs of strongly bonded materials are low and vice versa. As stated in the paragraph on the elastic constants, a stronger binding needs a deeper potential well, which causes a larger Young’s modulus. A deep well is also narrower and less asymmetric. Therefore, the CTE is smaller, and the product of CTE and modulus is almost constant. Some data of the TSF are listed in Table 3. In solids, which contain atoms with a small coordination number, strains between atoms can be accommodated by changing of bond angles and the TFS may become smaller than defined in Eqn (23). Important examples are diamond materials and silicon nitride, which also have an excellent thermoshock resistance, and glasses. A similar effect can

Chapter | 9.1

Mechanical Properties of Ceramics

617

result from microcracks, which are developed in some materials during cooling down from the consolidation temperature. This may happen, if the individual grains have largely anisotropic expansion/contraction characteristics or in multiphase materials. In this case, opening and closing of crack borders accommodate the strains and reduce the constraint. Although their tensile strengths are generally low, these materials may have an extremely high thermal shock resistance. Prominent examples are aluminum titanate and some plasma sprayed ceramics [52,53]. Let us discuss the consequences of the rule of thumb (Eqn (23)) in more detail. It tells us that for most ceramic and metallic materials, very small temperature changes can cause very high mechanical stresses. Temperature differences of only 100  C can cause a stress of about 300 MPa, which is about the yield strength of simple structural steel. Temperature differences of about 500  C can cause stresses even beyond the strength of high speed steels. Creating inhomogeneous temperature distributions in a constrained body is probably the easiest way to generate very large tensile stresses. Brittle failure of ceramics occurs if the stress exceeds the tensile strength7: sth  sf :

(24)

Using Eqns (22) and (24), we can determine the critical temperature, DTc where failing of ceramic components (under plain stress loading) will occur under full constraint [54]: R ¼ DTc ¼

sf fs;th

¼

sf ð1 yÞ : aE

(25)

Many ceramics have a low tensile strength. For example, for many electro ceramics, the strength is about 100 MPa. Then, the critical temperature can be as low as 30  C. It should be noted that, even if the difference in the TSF between different materials is relatively small, it can be of the highest relevance for the mechanical reliability of the components. Temperature changes can occur during the production of the ceramic component or in service. Typical reasons for temperature changes during production are heat treatments, soldering of components, or the application of layers on ceramic components. Sintering temperatures aredin generaldhigher or even much higher than 1000  C. Therefore, during cooling down from sintering temperature, thermal expansion mismatch may cause significant thermal stresses, which may even destroy the component. In operation, for example, in heat engines, temperature changes of several hundred degrees Celsius are also

7. For the definition of the tensile strength, see the following sections.

possible. Then, very high thermal stresses can occur, which have to be considered in the design of the component. In multimaterial components, thermal stresses may also occur if the materials have a different CTE (which is the common case), such as in rigid joints. This CTE mismatch causesdto give a common exampledthe bending of a bimetal strip with temperature. In ceramic multilayers, different CTEs can be used to produce layers with compressive stresses, which can act as crack stoppers [55,56]. In modern multilayer ceramic systems such as those used in electronic systemsdthe CTE mismatch between the ceramic and the metallic electrodes can cause severe thermal stress. In principle, the stresses scale as follows: sth ¼ Da$E DT:

(26)

where Da is the difference between the CTEs of the materials and E is a “mean” elastic modulus of the two materials. The CTE mismatch between silicon nitride ceramics and steel is around 10 ppm/K, and a temperature change of 100  C can therefore cause stresses of E =1000, which can be a few hundred MPa. If in a multimaterial component copper or aluminum alloys are in combination with ceramic materials (as in the case of electroceramic components), the CTE mismatch is between 20 and 30 ppm/K. This can be the reason for crack formation in the ceramic after sintering or after some thermal cycling. It may even occur as a result of the temperature differences between night and day [57]. Thermal strains are also of great relevance, if components are joined together. Changing the temperature from that at which the joint is stress-free will cause significant mismatch strains (Eqn (26)) to develop, leading to fracture if these are excessive. An example is shrink fitting, in which initial dimensional differences have to be very carefully calculated to avoid overstressing the ceramic. The guiding rule is always to place the ceramic into compression.

5. THERMAL DIFFUSION Thermal strains and stresses in monolithic components are triggered by temperature differences developed in the components, which are determined by the rates of heat fluxes and heat sinks or sources (i.e. energy transformations of heat). The transfer of heat within a material is mainly determined by thermal diffusion and the corresponding change in the temperature field Tð! r ; tÞ. It can be described with the well-known equation of heat conduction, based on Fourier’s law [6]. In a simplified approach (i.e. neglecting phase transformations, anisotropy), this equation can be expressed as dTð! r ; tÞ ¼ aV2 Tð! r ; tÞ: dt

(27)

618

where ! r is the position vector, t is the time, V2 is the Laplace operator, and a is the thermal diffusivity, which is taken here as constant. The heat transport in ceramics is dominated by the movement of phonons [58]. The thermal diffusivity is proportional to the heat capacity, the phonon velocity, and the mean free path of the phonons. In the classical limit, the heat capacity is three times the Boltzmann constant (cp ¼ 3k), but quantum mechanical arguments show that this is only true at temperatures well above the Debye temperature (which is several hundred degrees Celsius for most ceramic materials). At low temperatures, the quantum states are not completely filled (“they freeze out”) and they cannot consume energy. Therefore, at lower temperatures, the heat capacity is much lower than described in the classical limit and even reaches zero at zero Kelvin. For details, see Ref. [45]. The phonon velocity is equal to the speed of sound, which is pffiffiffiffiffiffiffiffi vs ¼ E=r. This property only weakly depends on temperature. The mean free path of phonons in a perfect crystal is restricted by the phononephonon scattering, which increases with increasing nonparabolicity (asymmetry) of the binding potential. Therefore, the phononephonon scattering is low if the binding partners have the same or a similar mass, but it is high if they have a very different mass. In diamond, where C-atoms are bonded to C-atoms, the nonparabolicity is small, and the mean free path is large. Diamond is the material with the highest thermal diffusivity known. In zirconia, where the mass of the binding partners is very different, the nonparabolicity is large and the mean free path is small (about one lattice vector). The phononephonon scattering largely increases with temperature, and the mean free path decreases. In summary, the thermal diffusivity has a maximum (which is at temperatures much lower than zero degree Celsius) and then decreases (for several orders of magnitude in some cases). Therefore, a proper analysis of temperature fields and of their evolution with time needs the knowledge of the thermal diffusivity in the temperature range of interest. The microstructure has a strong influence on the thermal diffusivity. Alloying increases the phononephonon scattering and drastically reduces the thermal diffusivity. The arrangement of the constituents is of great importance in multiphase materials. If heat conducting channels exist, they dominate the diffusivity. If the heat conducting phases are enclosed by insulating layers, the insulating phase dominates the conduction rate. This may happen in liquidphase sintered ceramics, in which glassy films exist at the grain boundaries. Thermal diffusivity is most readily measured by the socalled ‘laser flash’ method [59e61]. A parallel-faced disc (thickness, t) is subjected to a laser flash on one face, and

Handbook of Advanced Ceramics

the time taken for the temperature of the other face to increase to half its maximum ðs1=2 Þ is determined. Thermal diffusivity a (units m2/s) is then computed from the expression: a ¼ 0:138 t2 =s1=2 . In practice, depending on the equipment, a number of corrections may have to be made for factors such as finite pulse time effects and radiation losses. Compared with more traditional ways of measuring thermal conductivity under steady-state temperature gradients, thermal diffusivity measurements employ much smaller temperature differentials and thus tend to be more representative of true material behavior. The method is also readily extended to very high temperatures. Thermal conductivity l can be computed from thermal diffusivity by multiplying the latter by density r and specific heat Cp: l ¼ a$r$Cp Example data are shown in Table 4. High values of thermal diffusivity or thermal conductivity implicate that localized hot or cold spots are quickly removed and the temperature is homogenized faster than in materials with low values of these properties. Therefore, thermally induced internal strains are reduced and the propensity to cracking also. High thermal conductivity

TABLE 4 Thermal Diffusivity and Thermal Conductivity of Some Materials at Room Temperature [3,30,33] Thermal diffusivity 10 6 (m2/s)

Thermal conductivity (W/m K)

Diamond-based materials

400e1300

600e2000

Cemented carbides

20e30

60e80

Silicon carbides

30e100

100e200

Silicon nitrides

4e12

7e40

Alumina

6e11

15e30

Zirconia

0.7e1.2

1e2

Aluminum nitride

30e80

80e200

Glasses

0.6e0.8

0.7e1.3

Steels

3e15

11e55

Aluminum alloys

30e90

75e230

Copper alloys

50e120

160e390

Nylons

0.09e0.15

0.24

Epoxies

0.85e2.5

0.2e0.5

Polyethylene

0.15e0.3

0.42

Material

Chapter | 9.1

Mechanical Properties of Ceramics

619

FIGURE 4 Crack in a body with the definition of the local (crack tip) coordinates.

is therefore of advantage in most thermal shock situations.8

6. TOUGHNESS Toughness of ceramic materials can be defined in the framework of linear elastic fracture mechanics [4,5], where it is assumed that the material can be approximated by a homogeneous and isotropic elastic continuum, which contains a crack. Via the crack borders, no loads can be transferred. Then, the local stresspfield ffiffiffiffiffiffi sij around the crack tip depends on the SIF K ¼ sY pa (Eqn (2)): sij ðr; qÞ ¼

K pffiffiffiffiffiffiffiffi$fij ðqÞ; 2pr

(28)

where r is the distance of the volume element from the crack tip, where the stress is analyzed, and q is the angle between the crack plane and the line from the tip to the volume element (Figure 4). The tensor fij ðqÞ is dimensionless and only depends on the angle, q. Note that the pffiffiffiffiffiffiffi stress field scales with K ¼ s Y p a, that is, a large stress applied in a body containing a small crack causes the same crack tip field as a small stress on a large crack. Different crack opening modes, that is, opening, sliding and tearing of the crack, are taken into account by defining different SIFs ðKI ; KII ; KIII Þ and toughness values ðKIc ; KIIc ; KIIIc Þ. Because the opening mode I is by far the most pertinent to crack growth in ceramics, and because measured values for KIIc ; KIIIc do virtually not exist, the following discussion is limited to this loading case. Details can be found in textbooks on fracture mechanics [5,62]. 8. An exception to this rule may arise if the temperatures in the surrounding of a specimen are cyclically changed (thermal fatigue). Then, during the heating phase, heat flows into the specimen and some fraction of this heat flows out of the specimen during the cooling phase. This results in a fluctuating temperature field, where, during some part of the cooling phase, the temperature inside the specimen is higher than on its surface. Depending on the boundary conditions, larger temperature differences may arise in materials having a thermal conductivity as compared to those with a low thermal conductivity.

The Griffith/Irwin criterion (Eqn (3)) predicts that brittle fracture occurs, if the SIF (the stress field around the crack tip) reaches a critical value. This critical value is called fracture toughness, Kc. The energy release rate G describes the energy per newly cracked fracture surface, which is released by crack propagation (energy contributions are the elastic energy stored in the body and the work done by crack propagation). It relates to the SIF: G ¼ K 2 =E. The critical strain energy release rate is the energy needed to produce 1 m2 of new fractured area (specific fracture energy). It holds (Eqn (4)) that Gc ¼ Kc2 =E. It can be expected that this energy is the surface energy, which has to be invested to generate the new surface (Gc;0 ¼ 2g; the factor two arises because two new fracture surfaces are created in each new fracture; typical surface energies are in the order of magnitude of 1 J/m2), but it can easily be shown that the specific fracture energy of technical materials is much higher. As will be discussed in the following, there exist also other contributions to the specific fracture energy, which may become important. Two main processes are known to contribute to the specific fracture energy: process zone mechanisms and crack bridging mechanisms. Process zones are built ahead of the crack tip, if the material showsdat least a littled nonlinear behavior. Equating the yield strength ðsy Þ with the stress given in Eqn (28) (for simplicity, we neglect the tensor behavior of the stress field and set fij ðqÞ ¼ 1) and if the SIF reaches the fracture toughness ðK ¼ Kc Þ, we get an estimate for the diameter of the inelastically deformed zone dPZ (process zone):   1 Kc 2 : (29) dPZ z $ p sy Kc is the fracture toughness without the action of the mechanisms increasing the fracture energy (i.e. it is the fracture toughness arising from the surface energy). This is shown in Figure 5a. Examples for reasons of inelastic strains ðεin Þ can be plasticity (in this case, the process zone is called plastic zone), the switching of domains (examples are piezoceramics) or phase transformations (examples are partially stabilized zirconia, PSZ and zirconia toughened alumina, ZTA). If a crack propagates, the process zone moves with the crack tip through the body, leaving behind a wake of nonelastically deformed material at the crack borders (see Figure 5b). The energy dissipated per deformed volume (assuming “ideal plastic” behavior) is zsy εin . If the crack extension ðvaÞ is large compared with the zone diameter and if we assume a straight through crack in a plate of thickness t, the volume of the deformed material can be estimated to be dPZ t$va z dPZ $vA, where vA is the newly cracked area. Multiplying the dissipated energy density

620

Handbook of Advanced Ceramics

FIGURE 5 Process zone ahead of the crack tip: (a) the extension of the zone is limited by the yielding of the material: sij ¼ sy . (b) The process zone moves with the crack through the body. This causes deformed crack borders in the wake of the crack.

with the deformed volume we get for the dissipated energy per newly cracked area, vGPZ z dPZ sPZ εin :

(30)

If we consider, for example, the relationships of typical steel (where inelastic deformation is plastic deformation), typical values are dPZ y 10 3 m, sPZ ¼ sy z 500 MPa and εin z 5$10 2 . The specific fracture energy is 2.5$104 J/m2. If we set, for the elastic modulus E ¼ 200 GPa, and if the other contributions to the fracture energypcan ffiffiffiffiffiffiffiffiffi be neglected, the fracture toughness is Kc ¼ EGc z 70 MPaOm. This is a typical value of mild steels [34]. In the case of ceramics, the yield strength is very high (see Section 3), and the diameter of the plastic zone is very small. To give an example, for sy z 3000 MPa and Kc z 3 MPaOm, the diameter of the plastic zone diameter is only one-third of a micron, much too small to be recognized by a light microscope. The contribution of plasticity to the specific fracture energy is then negligible. In the case of transformation toughened ceramics, inelastic strains are produced by the phase transformation of metastable particles. To improve strength, the volume of the particles after transformation has to be larger than the volume before transformation. A prominent example is the transformation of tetragonal into monoclinic zirconia, which is associated with a volume increase of the unit cell of about 5%. The transformation is triggered by the tensile stresses ahead of the crack tip, that is, it occurs only in the so-called “transformation zone.” The transformation (yield) stress ðsT Þ can be adjusted by the particle size and the process zone diameter can reach several tens to more than hundred micrometers. Then, the transformation zone can be even seen with the naked eye [63e65]. Transformation-toughened ceramics have a relatively high toughness. For dPZ z 5$10 5 m, sPZ ¼ sT z 1000 MPa, and εin z 1$10 2 (if, e.g. ~20 vol.% of particles transform and the volume increase of each particle is ~5%, the resulting inelastic strain of the ceramic is ~1%), 200 GPa, the specific fracture energy is 500 J/m2. With E ¼ p ffiffiffiffiffiffiffiffi ffi we get for the fracture toughness Kc ¼ EGc z 10 MPaOm.

A third kind of toughening mechanism related to process zones is the so-called microcrack toughening, which may occur in ceramics being composed of two or more phases, for example, in ZTA [66], SiCeTiB2 [67], and B4CeTiB2 [68]. During cooling down from the sintering temperature, the difference between CTEs of different phases causes a strain mismatch and thermal stresses. At sintering temperature, the atomic mobility is very high and internal stresses can quickly be relaxed by diffusion; therefore, the body is assumed to be stress free. But at lower temperatures, the atomic mobility is reduced and diffusion can no longer relax internal stresses. These stresses between the phases vary at the scale of the grains. They can be added to the external stress field. If the sum of external and internal stresses is large enough, microcracks can come into existence (at a grain boundary, in the matrix or in particles), which relax the strain mismatch. The cracking is therefore related to some inelastic strains of the ceramic (which correspond to the mismatch strains between the grains) and the “yield strength” is the stress, where cracking occurs. Again, the contribution to the specific fracture energy can be estimated using Eqn (30). In the case of bridging mechanisms, load can be transferred across the crack borders. Bridging forces act behind the crack tip (Figure 6). Typical examples for crack bridges are fibers, which connect the borders of a crack, but bridges also exist in any material that has a rough fracture surface (which causes a geometric interlocking of the surfaces). Prominent examples are silicon nitride ceramics, which have deliberately elongated grains (the aspect ratio can be up to 1:10), but bridging is also very relevant for ceramics with coarse equiaxed grains (e.g. alumina) [69e72]. The dissipated energy can be estimated to be vGBr z 2

Zumax

pðuÞ$du;

(31)

0

where pðuÞ is the stress in a “typical bridge,” which depends on the crack opening u and umax is the opening, which causes the failing of the bridge. The bridges cause a relieve of stresses at the crack tip (a reduction of the SIF), which can be described as an increase of the fracture

Chapter | 9.1

Mechanical Properties of Ceramics

621

FIGURE 6 Bridging stresses can be transferred across the crack borders; (a) schematic sketch and (b) crack bridges in a silicon nitride material.

toughness (in a similar way as shown for process zone mechanisms). The bridging stresses may become very high but in monolithic ceramics, umax is in general very small. To give a simple example, if the bridging stress increases linearly with the crack opening and if it reaches 100 MPa at umax ¼ 1 mm, the integral Eqn (31) gives 100 J/m2, and the toughness of such a material would be around 4.5 MPaOm (for a material with E ¼ 200 GPa). In fiber-reinforced materials, the critical crack opening may be much larger (e.g. hundred times and even more), and very high contributions to the fracture energy may result from bridging mechanisms [69]. In general, several toughening mechanisms act simultaneously, and the specific fracture energy has several additive contributions [73,74]: X X vGBr þ . (32) vGPZ þ Gc ¼ 2g þ i

i

There exists a strong influence of the microstructure on the fracture toughness, but this influence depends on the analyzed mechanism and a short summary is in danger of oversimplifying this aspect. To give some ideas, in the case of process zone mechanisms, the increase of specific fracture energy is vGPZ z dPZ sPZ εin (Eqn (30)), whichdby the use of Eqn (29)d becomes vGPZ f εin =sPZ . The property to be optimized is therefore εin =sPZ . In the case of toughening by plasticity, many influences on the yield strength are known (see the standard textbooks on physical metallurgy) [35]; some are solid solution hardening, precipitates, duplex structures, grain size, and so on. Of course, the hardening also has an influence on the plastic strain. In the case of transformation-toughened ceramics, the particle size has an influence on the strength where the transformation occurs (the transformation stress decreases with increasing particle size up to a critical size, where the transformation still happens without any stress). The toughening is maximum if the transforming particles are a little smaller than the critical size, but there is no toughening at all, if they are larger [75,76]. In the case of bridging mechanisms, coarse microstructures in general cause higher toughening than fine

microstructures (Eqn (31)), because the maximum possible crack opening ðumax Þ scales with the size of the relevant microstructural feature [77,78]. As is evident from Figures 5 and 6, toughening mechanisms become only active if a crack propagates. This leads to an increase of the fracture toughness Kc with crack extension of typically several times the size of the process zone from a lower starting value to a higher saturation value. The crack-length-dependent fracture toughness is also referred to as the R-curve. Toughness measurements are based on the Griffith/ Irwin criterion (Eqn (3)). A well-defined crack is made into a specimen, which is then loaded. The loading causes an SIF (Eqn (2)) that can be increased by increasing the load until it reaches the fracture toughness. Then, the specimen fails. If the fracture stress, the geometry and size of the fracture initiating crack and the geometric factor (Eqn (2)) are known, the critical SIF (the fracture toughness) can be determined. The production of a well-defined starter crack is the most relevant problem in the fracture toughness measurement of ceramics. In the single-edge precracked beam (SEPB) method [79e81], a sharp planar crack is ‘popped in’ to a beam specimen from a row of indentations using a stiff jig system. This can be a difficult task for the inexperienced. A modified version is the single edge V-notch method (SEVNB) [82,83] in which a straight through notch is machined into a bending specimen using diamond paste on a razor blade to reach a sharp tip radius of 3 mm, approaching a sharp crack. An investigation showed [84] that the SEVNB method is applicable if the grain size of the material is larger than the notch tip radius. In the case of the surface-crack-in-flexure (SCF) method [81,85], a sharp crack is introduced by intending a bending specimen with a Knoop indenter. To obtain a residual-stress free starting crack, a surface layer of the specimen, which contains the plastically deformed material, has to be removed (e.g. by grinding). Then, the specimen can be fractured in bending, and the size of the indentation crack can be determined fractographically.

622

Another standardized method, the chevron notch beam (CNB) method [81,86], requires a chevron notch to be machined into bending specimens. During loading the chevron notch, a sharp crack pops into the specimen, which acts as a starter crack for the fracture toughness measurement. In a European round robin concerning the practical applicability of this method, it turned out that the method is relatively expensive and difficult to perform properly [87]. In particular, a stiff testing machine is needed to obtain controlled fracture, a requirement of the analysis. If only a limited amount of material is available, the indentation fracture (IF) method [88] is very commonly used. However, there exists a myriad of proposed evaluation formulas, all relying on empirical ‘calibrations’ that are often erroneous correlations with fast fracture toughness data from other methods. The computed values are usually termed the ‘IF resistance’ and are calculated from the length of the cracks emanating from the corners of Vickers indentations (Figure 2). Normally, no two formulas give the same result for a given material. Because the measurement is performed under a condition of crack arrest (instead of the situation of the moving crack becoming critical as in all other methods), the measured quantity may be somehow related to but is never equal to fast fracture toughness [89]. Measurement of crack length is also subjective. The method should never be used for material comparison or specification purposes. In summary, to the authors’ point of view, the SEVNB method is a testing procedure that is easy to perform and is applicable to all types of ceramics having grain diameters of several micrometers or larger. The SCF method may be a little more precise, but the experimental procedure is more complex and time consuming, and does not always work with coarser-grained materials. Both methods employ effective crack lengths of 50

300e2000

Aluminim alloys

22e35

60e550

Copper alloys

30e90

100e550

Nylons

2e5

90e160

Epoxies

0.4e2

45e90

Polyethylene

1.5

20e45

Material

which are distributed in the specimen or on its surface. The critical flaw, ac, ispthe ffiffiffiffiffiffiffiflaw, which is loaded with the highest SIF, K ¼ s Y p a.9,10 Brittle fracture occurs, if the SIF of the critical flaw reaches or exceeds the fracture toughness of the material: K  Kc (Eqn (3)). The tensile strength is the stress (Eqn (7)) necessary to fulfill this condition. In the daily routine, brittle fracture is the most important factor limiting the successful use of ceramics. The main reasons for unexpected failures are hidden or unexpected tensile stresses, which have not been accounted for in the design. The important reasons for such stresses are temperature changes or temperature gradients and contacts between the ceramic component and other objects. An important consequence of Eqn (7) is that the tensile strength of ceramic specimens (components) is not given

9. Becausedin generaldthe tensile stress (s) in the uncracked specimen and the geometric factor (Y ) can be different from position to position (the geometric factor also depends on the orientation of the crack in relation to the tensile stress components), the critical flaw must not be the crack with the largest size. 10. For simplicity and without loss of generality, we restrict the following discussion to a homogeneous and uniaxial stress field (as occurs in a tensile specimen) and to a homogeneous and isotropic flaw population in a homogeneous and isotropic elastic continuum.

Chapter | 9.1

Mechanical Properties of Ceramics

623

by a simple number but by a distribution function. There is always some risk that a specimen fails at a very low stress (if it contains a very large flaw). It is important to understand that the distribution of strength data is caused by the size distribution of flaws. There is a direct correlation between the size of the critical flaw, the fracture toughness, and the strength   1 Kc ; (33) ac ¼ $ p sf which follows from Eqn (7) and is a direct consequence of the Griffith/Irwin criterion (Eqn (3)). A number of articles [8,9,90e92] have been written in the past to develop a theory of brittle fracture, which accounts for the scatter of the strength. It has been shown that for specimens containing noninteracting flaws (i.e. the flaws should be sparsely distributed), which can be described as cracks and which fail due to the weakest link criterion, the probability of failure is given by the simple relationship [93] as follows:      (34) F s; S ¼ 1 exp Nc;S ;

where hNc;S i is the mean number of critical flaws. It can be shown that, using a failure criterion of the Griffith/Irwin type (Eqn (3)), hNc;S i depends on the applied stress, s and the size of the specimen S. The mean is taken over a large sample of identical specimens. At very high stresses, it is possible that a specimen contains “on average” several critical flaws (their size is small at high stresses, see Eqn (34)), and the frequency for small flaws will in general be much higher than for large flaws. But there remains a (small) probability to find no critical flaw in one specimen, and the probability of failure is therefore always a little smaller than 1.11 At a very low number of critical flaws ðhNc;S i 70 years ago by the Swedish engineer Weibull on the basis of empirical arguments [8,9]. Note that the Weibull parameters are related to the shape of the flaw population. The Weibull distribution is defined by the shape parameter m and a scale parameter 1=m s0 V0 . The normalizing parameter V0 can be freely selected, but the characteristic strength s0 depends on the definition of V0. It holds that m ¼ 2ðr 1Þ and m V0 sm 0 ¼ ðm=2g0 a0 Þ$ðKc =pa0 YÞ . These equations demonstrate that the shape of the Weibull distribution depends on the shape of the frequency distribution density of the flaws in the material. The Weibull distribution describes the experimentally observed behavior of ceramics: the strength scatters and the mean strength decreases with the specimen size (Figure 8). Note that that ‘size’ refers to the volume of the specimen if volume defects cause failure and to the surface, if surface defects are responsible for fracture. The preceding analysis is specific and related to many model assumptions, but the results are much more general and can be extended to many other cases, for example, to surface or edge flaw distribution or to inhomogeneous and multiaxial stress fields [94e97]. In the latter case, the volume in Eqn (38) has to be replaced by an effective volume and the stress by an equivalent stress. This is described in an other chapter of this book. The Weibull distribution is a special case out of a more general class of distribution functions for flaws whose size frequency distribution decreases with increasing flaw size corresponding to a power law (Eqn (35)). Although this occurs in many cases, it needs not to be true in any case. Therefore, the reader should bear in mind that the Weibull distribution is only a model description, which is true in many but not in all cases. The microstructure has an influence on the strength (Eqn (7)) via the fracture toughness and the flaw population. The first aspect has been discussed in the last

section. The characteristic strength can be increased by reducing the frequency of large flaws; the scatter can be reduced by a narrower flaw size distribution. In fact, both have successfully been achieved in the last 30 years to improve strength and reliability of modern ceramic materials. Large flaws that act as fracture origins can be identified on the fracture surface of specimens. An example is shown in Figure 9. The fracture origin is an agglomerate produced by spray drying of powders, which has not been completely destroyed during the uniaxial pressing of the component. Such type of defects could be avoided by changing the spray drying conditions of the powders. This resulted in an increase of strength of nearly 20% [98]. In fact, the characteristic strength of commercial ceramic materials could be increased significantly over the last 30 years and the scatter of data could be reduced. Thirty years ago, the Weibull modulus of a state-of-the-art ceramic was between 5 and 10, and nowadays, it is between 15 and 25. This causes an enormous increase in the reliability of components. To give a simple example, we compare two samples of ceramic specimens, having Weibull moduli of m ¼ 5 and m ¼ 25. The characteristic strength of both samples is 500 MPa. If the specimens are loaded with a stress of 250 MPa, 3% of the specimens fail in the first, but only 3$10 6% in the second sample. In other words, the risk of failure at half of the characteristic strength is reduced by a factor 106 by increasing the Weibull modulus from 5 to 25. As shown earlier, the strength of brittle materials is related to the defect population(s) present in the specimen or component (cf., Eqn (7)). Due to the distribution of defect sizes, it exhibits an intrinsical scatter. Following a statistical approach, a strength distribution describes adequately the strength of brittle materials, which is typically defined by a Weibull distribution and the corresponding Weibull parameters (i.e. shape and scale parameter).

FIGURE 8 Characteristic strength versus (effective) surface for an alumina ceramic. Test results on bending specimens (:) as well as on biaxially tested discs (B) and rectangular plate specimens (n) are shown [99].

FIGURE 9 Fracture origin in a standard flexure specimen. Sickle-shaped pores indicate the location in a powder agglomerate that was not completely destroyed during pressing.

Chapter | 9.1

Mechanical Properties of Ceramics

The basic goal of strength measurements is to determine the previously unknown strength distribution parameters of a material as accurately as possible. The measurement of a sample containing a finite number of specimens reflects this intrinsic scatter, and due to statistical reasons, the result of a such strength measurements can only deliver the distribution parameters within a certain confidence range according to a given confidence level (i.e. confidence intervals for both Weibull parameters). Consequently, meaningful results of strength measurements are not unique numbers for the distribution parameter but certain confidence intervals, whose widths depend on the confidence level and the sample size. Strength tests for ceramics should be so simple to perform that a sufficient number of tests can be easily accomplished to obtain a sound basis for the statistical evaluation. The standards suggest a minimum of 30 specimens per condition as a tradeoff between experimental effort and a sensible small confidence bandwidth [100e103]. Tensile tests are in principle possible but require expensive specimens and sophisticated testing equipment to provide true uniaxial stressing, that is, without excessive superimposed bending [104]. The strength of ceramics is thus preferentially measured in bending. Bending tests can be performed under either uniaxial [105e109] or biaxial loading conditions [110]. Uniaxial bending tests are usually performed on long, thin (length over thickness ratio > 10) prismatic beams or cylinders. For such specimens, simple expressions from the linear-elastic theory can be used for the calculation of the strength from the fracture force and the geometry. The possible measurement error is well investigated [111,112] and is in the range of 2e10% for specimens with a length > 10 mm. The lower limit of this error budget can be achieved when tests are performed with utmost care. The upper limit may be too much to determine high Weibull moduli above m > 20. The experimental scatter then masks the inherent strength scatter [113]. Biaxial strength tests may have advantages over uniaxial bending tests if, for instance, specimen machining can be omitted because components already have a suitable geometry. Furthermore, biaxial stresses are often highly relevant for applications, as in the case of (cyclic) thermal loading. Most of the suggested biaxial bend tests, that is, ball-on-ring, ring-on-ring, or punch-on-ring tests suffer from the discrepancy between the actual loading situation and the idealized descriptions, which are assumed to derive the evaluation formulas. The maximal stresses occurring in the specimen strongly depend on the contact situation (at the load application points). Analytical solutions are usually only approximations, which may differ significantly from the real situation [114]. An exception is the ball-on-threeballs test, which is evaluated using a numerical fit on parametric finite element method (FEM) calculations [114].

625

(A tool for the evaluation of this test is available on http:// www.isfk.at/de/960/). This test is also extremely tolerant against imprecisions of specimens or testing equipment. With this test, a measurement error around 1% can even be achieved with specimens with diameters of approximately 2 mm and as thin as 500 mm [115]. For design purposes, not only the magnitude of a materials’ strength but also its scattering property play a significant role. Using, for example, the Weibull distribution as the basis, it is possible to calculate a tolerable load level to achieve a certain short-time reliability or vice versa. For such calculations, it is indispensable to use representative values for the strength distribution, that is, those that are caused by the same defect population that causes fracture during operation [116]. In the standards for bending tests, requirements for the quality of the tensile surface of the bend bars are given to test the failure strength (distribution) as a material property and not as a consequence of a certain machining procedure. The someasured strength distribution may be the relevant one for failure during operation, but it can also be an irrelevant one, if, for example, other parts of the components that are in the as-sintered state and certainly have a different defect population are under sufficient high stress. Another issue that complicates reliability assessment is the fact that strength is usually tested on different (effective) volumes (or surfaces) than the components have. The volume dependence of strength may be another reason why critical defect populations are different in specimens and components [117].

8. STRENGTH DEGRADATION WITH TIME In the last section, we focused on fast brittle fracture, that is, on failures in which the crack velocity almost reaches the speed of the sound. Fracture could therefore be described as instantaneous. But in practice, brittle fracture has also been observed to occur some time after load application. This can arise for several reasons (e.g. slow plastic deformation (i.e. creep), corrosion, cyclic fatigue, etc.). In this section, we will focus on the so-called slow or subcritical crack growth (SCCG), which is the most relevant time-dependent damage mechanism in ceramics. It not only occurs at low and intermediate temperatures under the action of a constant load but also under varying loads and causes a progressive decrease of remaining strength with time. The first direct observations of SCCG were made in glass [118]. It was recognized that SCCG is thermally activated and also strongly depends on the environment. Tests in a humid atmosphere show (up to some orders of magnitudes) larger crack growth rates than in vacuum. Similar observations have also been made in the presence of other polar molecules [11]. The crack growth rate also strongly depends on the SIF (KI; Eqn (2)). Very slow growth rates

626

Handbook of Advanced Ceramics

FIGURE 10 Crack growth rate versus SIF for (a) an alumina replotted from Ref. [121] and (b) a silicon nitride ceramic, replotted from Ref. [122]. Paris laws are indicated by straight lines.

(103 m/s for most ceramic materials). So the observed SCGG rates vary over the enormous range of 17 orders of magnitude. This is an extremely wide range for a parameter (for comparison, the age of the world is about 1017 s). Several crack growth mechanisms have been proposed to occur in the SCCG range, but a clear understanding is still missing. Therefore, the statements in the next paragraphs are the current state of understanding. To describe SCCG in vacuum, thermally activated breaking of bonds at the crack tip has been proposed [119,120]. At high crack propagation rates, this is also the dominating mechanism. In polycrystalline ceramics, this often leads to transgranular fracture. In chemically inert environments or in vacuum, thermally activated bond breaking may even be responsible for SCCG at very low and intermediate growth rates. At very slow crack extension rates, there exists a strong influence of the environment. Many ceramics contain (glassy) silicate grain boundary phases. Here, water molecules can hydrolyze silicate bonds at the crack tip, and this process can be many orders of magnitude faster than the thermally activated breaking of the silicate bonds. The chemical reaction at the crack tip and the subsequent breaking of the hydrolyzed bonds is the rate-controlling factor. At intermediate growth rates, the transport of water molecules at the crack tip may become rate limiting. In polycrystalline ceramics, this SCCG mechanism promotes intergranular crack propagation. A consequence of SCCG is that the lifetime of components could lie somewhere between some hours and some tens of years. This lifetime range is caused by SCCGrates in the environmentally influenced low crack growth rate regime, that is, at crack growth rates between 10 13 and 10 5 m/s. In this regime, the crack growth rate (v) is

often empirically described by the so-called Paris type crack growth law:  n K ; (39) v ¼ v0 Kc with the (temperature- and environment-dependent) material parameters v0 and the SCCG exponent (Paris exponent) n. 12 For the following discussion, it is important to realize that the SCCG exponent in general has a very high value. Some SCCG data for a commercial alumina ceramic [121] and a commercial silicon nitride [122] are shown in Figure 10. At room temperature, the SCCG exponent is n z 33 for alumina and n z 50 for silicon nitride, respectively. These data are typical for materials with glassy silicate grain boundary phases. For ceramics that do not have a glassy grain boundary phase, values around n z 200 are reported, for example, for silicon carbide ceramics [123]. For the silicon nitride ceramic, the crack growth rate increases significantly if the temperature reaches or exceeds the softening temperature of the glassy grain boundary phase, and the exponent decreases markedly. Very little is known about the influence of the microstructure on the SCCG behavior of ceramics apart from the fact that silicate grain boundary phases can accelerate the crack growth rate significantly. If the strong dependence of SCCG on stress, or more precisely, on the SIF, is considered, it seems obvious that local internal stresses may have a tremendous influence on the growth rate. For the same reason, it is obvious that the size of the flaw that extends under the action of SCCG also has a significant influence on the growth rate. This point will be discussed in more detail in the paragraphs on testing procedures. In any case, reducing the size of the internal flaws through improved 12. It should be emphasized that this is only a model that is broadly matched by experiment, but there is no physical reason why n should be a constant over the entire crack growth velocity range.

Chapter | 9.1

Mechanical Properties of Ceramics

627

processing is one of the best strategies to increase the lifetime of components under stress. Another strategy is to avoid grain boundary phases. Crack growth data obtained on test pieces can be used to predict the lifetime of components. Under the action of a stress ðsðtÞÞ, which may be time dependent, a crack grows from its initial length ai to the critical final (Griffith) length ac ¼ af , at which point failure occurs almost immediately. The time to failure (lifetime) can be determined in the following way: the crack growth rate is the differential of the crack length with time v ¼ da=dt. Then, the differential of the time is dt ¼ ð1=vÞ$da. By separation of variables and using Eqns (2) and (39) we get: Ztf ti

1 sðtÞ dt ¼ v0 n

Zac 

ai

n Kc pffiffiffiffiffiffiffi da: Y pa

(40)

If we assume that small cracks already exist in the specimen (or component), we can set ti ¼ 0. For such small cracks, which are commonly present in most ceramic materials, the geometric factor Y is a constant, and the integration can be made analytically. For n > 2, it holds that Ztf

sðtÞn dt ¼

2 n

0

 1

n  ai Kc $ $ pffiffiffiffiffiffiffiffiffi 2 v 0 Y p ai  ðn ai ac

2Þ=2

:

(41)

The right-hand side of Eqn (41) is a constant value. Therefore, the stress dependence of lifetime results from the left-hand side, so the lifetime strongly depends on the applied stress and its history. It should be noted that if significant crack growth occurs, the second term in the square bracket can be neglected because n >> 1. This shows that the most prominent material properties influencing the lifetime are the initial crack length (ai) and the Paris exponent (n). Measurements of SCCG have been made using specimens with artificial cracks [124]. The crack advance after progressive time intervals can be determined using microscopic techniques. If a crack propagation distance of, for example, 10 mm is recorded after one day, the determined crack velocity is v z 10 5 m/105 s ¼ 10 10 m/s. This demonstrates that measuring very slow crack growth rates on artificial (long) cracks needs very long testing times (at least three months) under difficult conditions and is therefore not practicable. In addition, the tests usually required long cracks that can be reliably measured, but because components generally do not have long cracks in them while still serviceable, it is questionable whether such macrocrack growth data are equivalent to microcrack

growth that is the usual source of delayed failure in components. Instead, to simulate the latter, measurements are normally based indirectly on the growth of natural flaws. Two different testing procedures are frequently used: specimens are either loaded at a constant stress and the time to failure is recorded or they are loaded with a linearly increasing stress (constant stress rate) until failure occurs. These situations will now be discussed in more detail. Experiments under constant load ðsðtÞ ¼ sstat Þ are made in most cases in bending. Very simple testing equipment can be used, for example, the stress can be applied by simply loading the specimens with dead weights. However, gaining the data for a proper veK-curve may still require three months or even more. For constant load testing, the solution of Eqn (41) is straightforward, giving n   ðn 2Þ=2

2 ai Kc ai $ $ tf ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi $ 1 n 2 v0 Y p ai sstat ac

2 ai ac v i $ $ 1 $ : ¼ ai v f n 2 vi (42) The indices i and c refer to the initial and final (Griffith) crack length, respectively, vf is the final crack velocity. In experimental practice, the scatter of the defect sizes has to be taken into account. For example, at a fixed stress level, a distribution of failure times can be determined, which reflects the distribution of initial defect sizes, ai. This distribution is a Weibull distribution. If the Weibull distribution of the strength (which should be determined in such a way that SCCG has only negligible influence, i.e. it should be the inert strength distribution) is known, the Paris parameters can be determined by comparison of both distributions. For example, the modulus of the distribution of lifetimes is m0 ¼ m=ðn 2Þ. The lifetime distribution is very wide, because typical values of m z 10e20 and of n z 30; therefore, m0 z 1/3e2/3. In the case of ceramics without glassy grain boundary phases, the SCCG exponent is much larger, the modulus of the distribution of lifetimes is smaller, and the distribution of times to failure even wider. Therefore, some early failures can arise if the experimental design is not properly made. It follows from Eqn (42) that the time to failure is proportional to ðn 2Þ=2 $sstatn . If the high value of the SCCG-expot f f ai nent is considered, it is immediately clear that a reduction of the stress level is a very effective measure to increase the lifetime of a component and to reduce the probability of early failure. An alternative approach to determining the SCCG exponent is to test the subsets of test specimens at different constant stress levels. It can be expected that for every subset at the 50% percentile value of the lifetime, the initial

628

crack (ai) had almost the same length, and the SCCGexponent can be determined from the relationship: tf f sstatn . As a result of the high value of the SCCG exponent, the development of the crack length with time under constant load may appear a little surprising. This is shown in Figure 11 for three hypothetical materials having different SCCG exponents. The increase in crack size is very slow at the beginning of the experiment. The crack needs a large fraction of its lifetime to grow just a few percent in length. Then, it quickly accelerates, and brittle failure occurs. This explains why failure due to SCCG occurs quite unexpectedly. Recognized cracks seem to be stable for a long time, and changes in crack length are so small that they can hardly be determined. But after some “incubation time,” almost instantaneous brittle fracture occurs. The second frequently used type of test for the indirect determination of veK curves is the constant stress rate test: _ [125]. In this case too, tests are in general cons ¼ s$t ducted in bending. For load application, a universal testing machine is necessary, which can precisely be regulated even at slow loading rates. The left-hand side of Eqn (41) can be integrated analytically. After some rearrangement, we get  n 2

2ðn þ 1Þ ai n sf _ ; (43) $ ¼ $s $ s 1 snþ1 f si n 2 v0 i pffiffiffiffiffiffiffi where si ¼ Kc =ðY pai Þ is the inert strength (i.e. the strength of the specimen without any SCCG) and sf is the determined fracture strength of the test specimen. Of course, sf  si . Again, if any significant SCCG occurs, the second term in the square bracket is very small compared with unity and can be neglected. In this region, the strength depends on the stress rate according to sf f s_ 1=ðnþ1Þ (see also Figure 12; region s_ < 102 MPa/s). At higher stress rates, the test duration is too short for significant SCCG to

FIGURE 11 Crack length versus time for materials having a different SCCG exponents n (using typical data for alumina, silicon nitride and silicon carbide, respectively, and setting sstat =si ¼ 0:95) in relative units. The cracks need a large fraction of their lifetime to grow the first few percent. They then accelerate, and fracture occurs very quickly. This behavior is strongly dependent on the SCCG exponent.

Handbook of Advanced Ceramics

FIGURE 12 (Mean) strength versus stress rate for constant stress rate tests performed on an alumina ceramic [126].

occur, and the strength almost equals the inert strength. Data as shown in Figure 12 can be used to derive the veK curve of the material. Compared with tests with constant load, constant stress rate tests offer some advantages. In particular, the time of individual tests can be estimated much easier. But in this type of test too, very slow crack growth rates can only be assessed with very time consuming and expensive experiments. The determination of a veK curve (in the technically relevant parameter regime) may need about 3 months. Due to the expensive and long-lasting experiments necessary to measure a veK curve, data on SCCG behavior are very rare and determined in most cases at room temperature and in laboratory air. For these conditions, a typical SCCG exponent for materials having silicate grain boundary phases is n ¼ 30e50. Recent measurements on low temperature co-fired ceramics (LTCCs), which are nowadays used for integrated circuit boards, and which can be used at elevated temperatures and under severe mechanical conditions (vibrations), gave similar values [127,128]. The same happens for the SCCG exponent of glasses. In SiC ceramics, which have no glassy grain boundary phases, an SCCG exponent of about n z 200 has been determined using this technique [123]. The technical significance of SCCG is the fact that the strength degrades with time. A desired reliability cannot be guaranteed for infinite time, and the reliability decreases with time. These aspects, which are highly relevant for a successful use of ceramics, have to be considered in the design [129]. Allowable stress levels have to be modified to account for SCCG, and the lower the value of the exponent, the bigger the required reduction in allowable stress.

9. FINAL REMARKS The mechanical properties discussed so far are used to analyze the response of ceramic components to stress and to predict their reliability and lifetime. But fractographic evidence tells us that there are three, often hidden reasons

Chapter | 9.1

Mechanical Properties of Ceramics

for additional causes of stresses in components, which are often ignored in stress analyses and which are responsible for unexpected failures: internal residual stresses, imposed thermal stresses, and stresses arising in contact situations. We will conclude this chapter with some comments on the effects of these and on the need to guarantee properties in critical components. Generally distributed internal residual stresses can arise inside components as a consequence of either thermal expansion mismatches from one area to another, or because some regions of the component are cooled much faster than others during the final stages of the sintering or densification cycle. They are more prevalent in larger monolithic items than in smaller ones, and are best removed by the use of appropriate annealing conditions. In some cases, they are deliberately introduced, such as in multilayered materials, but they have a strong influence on strength and reliability. In addition, machining can introduce surface residual compressive stresses. Often, these are beneficial in reducing the ability of small surface cracks to propagate, but they strongly depend on the manner in which the machining is undertaken. For predictive purposes, it is crucial to match the machining conditions of specimens for experimental determinations of strength with those that are employed in manufacturing components. Thermal stresses can also arise from temperature differences between one part of the component and another. Unlike in metal component designs where the consequences of such stresses are often ignored, in ceramic components, they are of much more importance, adding to imposed mechanically derived stresses. Often, the cooler parts of components are subjected to tensile stress, while the hotter parts are in balancing compression, assuming a positive expansion coefficient. Stresses are clearly reduced in lower thermal expansion materials, which tend to drive material choice for thermal applications. A special case of thermal stress arises when temperatures are changed very rapidly, particularly during rapid cooling. Then, the surface of the ceramic is subjected to high tensile stress, which if too large is relieved by surface crazing. Again, such effects can be mitigated by using materials possessing low thermal expansion characteristics. The third type of additional stresses are those due to mechanical contact. Any form of load transfer through localized contacts produces highly localized stress concentrations. Usually associated with such contacts are high levels of local tensile stress, which can by themselves introduce strength-reducing cracks. It should also be noted that machined ceramics have rough surfaces, so proper modeling of contacts is difficult. Often, some means of soft load spreading through metal or polymer layers is needed. Designing ‘benign’ load transfer is essential to minimize such effects.

629

A final issue that needs discussion is the means that can be taken to ensure that any individual component has the desired strength properties. The only practical solution to this problem is to undertake mechanical proof testing, that is, to apply a stress level to the component, which will ensure that it will survive in the application. There are two strategies for deciding what stress to use; this can be done by analysis of subcritical growth and of probabilistic risks of failure under the stress distribution applied (using the approaches described in the preceding sections) or this can be done in a pragmatic manner with the intention of removing the weak end of the distribution of component strengths. The former requires detailed relevant input data concerning strength statistics and SCCG and is seldom undertaken. The latter approach, combined with design stress minimization, is more commonly employed and successfully reduces the risks of failure in service. By assuming that proof testing will be employed, an appropriate maximum design stress level can be estimated for the ceramic by a simple estimate of Weibull modulus and the SCCG parameter. In this way, the advantageous properties of ceramics, such as hardness, dimensional stability, wear resistance, electrical insulation, can be exploited while minimizing the risks of brittle fracture.

REFERENCES [1] Timoshenko SP, Goodier JN. Theory of elasticity. McGraw-Hill; 1970. [2] Wachtman JB. Mechanical properties of ceramics. Wiley-Interscience; 1996. [3] Munz D, Fett T. In: Hull R, Osgood RM, Sakaki H, Zunger A, editors. Ceramics. Springer series in materials science, Vol. 36. Springer; 1999. [4] Gross D, Seelig T. Fracture mechanics. Mechanical engineering series. Springer; 2006. [5] Anderson TL. Fracture mechanics: fundamentals and applications. 3rd ed. CRC Press Inc.; 2005. [6] Carslaw HS, Jaeger JC. Conduction of heat in solids. Clarendon Press; 1993. [7] Boley BA, Weiner JH. Theory of thermal stresses. Krieger Publishing Company; 1983. [8] Weibull W. A statistical distribution function of wide applicability. J Appl Mech 1951;18:293e8. [9] Weibull W. A statistical theory of the strength of materials. Ingenio¨rsvetenskapsakademiens Handlingar 151, Vol. 151. Generalstabens Litografiska Anstalts Fo¨rlag; 1939. [10] Wiederhorn SM. Subcritical crack growth in ceramics. In: Bradt RC, Hasselman DPH, Lange FF, editors. Fracture mechanics of ceramics. New York: Plenum Press; 1974. p. 613e46. [11] Michalske TA, Freiman SW. A molecular mechanism for stress corrosion in vitreous silica. J Am Ceram Soc 1983;66:284e8. [12] Gilbert CJ, Dauskardt RH, Ritchie RO. Microstructural mechanisms of cyclic fatigue-crack propagation in grain-bridging ceramics. Ceram Int 1997;23:413e8.

630

[13] Nye JF. Physical properties of crystals. Clarendon Press; 2009. [14] McNamee M, Morrell R. Textural effects in the microstructure of a 95% alumina and its relationship to strength. Sci Ceram 1983;12:629e34. [15] Pavlacka RJ, L. MG. Processing and mechanical properties of highly textured Al2O3. J Eur Ceam Soc 2010;30:2917e25. [16] Hall DA. Review: nonlinearity in piezoelectric ceramics. J Mater Sci 2001;36:4575e601. [17] Jones JL, Iverson BJ, Bowman KJ. Texture and anisotropy of polycrystalline piezoelectrics. J Am Ceram Soc 2007;90: 2297e314. [18] Ashby MF, Jones DRH. Engineering materials 1. International series on materials science and technology. Pergamon Press; 1980. [19] Kittel C. Introduction to solid state physics. Wiley; 1996. [20] Chiang Y-M, Birnie DI, Kingery DW. Physical ceramics. John Wiley & Sons, Inc.; 1997. [21] Richerson DW. Modern ceramic engineering. Marcel Dekker Inc 1992. [22] Dean EA, Lopez JA. Empirical dependence of elastic moduli on porosity for ceramic materials. J Am Ceram Soc 1983;66:366e70. [23] Lins W, Kaindl G, Peterlik H, Kromp K. A novel resonant beam technique to determine the elastic moduli in dependence on orientation and temperature up to 2000  C. Rev Sci Instrum 1999;70:3052e8. [24] ASTMC1259. Standard test method for dynamic young’s modulus, shear modulus, and poisson’s ratio for advanced ceramics by impulse excitation of vibration; 2001. [25] ASTMC1198. Standard test method for dynamic young’s modulus, shear modulus, and poisson’s ratio for advanced ceramics by sonic resonance; 2008. [26] EN 843-2. Advanced technical ceramics - mechanical properties of monolithic ceramics at room temperature - Part 2. Determination of Young’s modulus, shear modulus and Poisson’s ratio 2006. [27] JIS R 1602. Testing methods for elastic modulus of fine ceramics; 1995. [28] Lord DJ, Morrell R. Elastic modulus measurement. Measurement good practice guide, Vol. 98. National Physical Laboratory; 2006. [29] Landau LD, Lifschitz EM. Elastizita¨tstheorie. 5. Auflage ed. In: Lehrbuch der Theoretischen Physik, Vol. VII. Akademie-Verlag; 1983. [30] Field JE, editor. The properties of natural and synthetic diamond. London: Academic Press; 1992, 710. [31] Czichos H. Hu¨tte - Die Grundlagen der Ingenieurswissenschaften. Springer-Verlag; 1989. 29. Auflage ed. [32] Shackelford JF, Alexander W, Park JS, editors. CRC materials science and engineering handbook. Boca Raton: CRC Press; 1994. [33] Ashby MF. Materials selection in mechanical design. Pergamon Press; 1992. [34] Ashby MF. Materials selection in mechanical design. Elsevier; 2007. [35] Haasen P. Physikalische metallkunde. Springer; 1984. [36] McColm IJ. Ceramic hardness. Plenum Press; 1990. [37] Jona F, Shirane G. Ferroelectric crystals. International series of monographs on solid state physics. Pergamon Press; 1962. [38] ISO 14705. Fine ceramics (advanced ceramics, advanced technical ceramics) e test method for hardness of monolithic ceramics at room temperature; 2008.

Handbook of Advanced Ceramics

[39] ASTMC1327. Standard test method for vickers indentation hardness of advanced ceramics; 2003. [40] E.N 843-4. Advanced technical ceramics - mechanical properties of monolithic ceramics at room temperature - Part 4: Vickers. Knoop and Rockwell superficial hardness 2005. [41] JIS R 1610. Test methods for hardness of fine ceramics; 2003. [42] ASTMC1326. Standard test method for knoop indentation hardness of advanced ceramics; 2003. [43] Morrell R. Handbook of properties of technical & engineering ceramics, part 1: an introduction for the engineer and designer. handbook of properties of technical & engineering ceramics, Vol. 1. Her Majesty’s Stationary Office; 1989. [44] Warren R. Ceramic-matrix composites. Blackie Son Ltd; 1992. [45] Ashcroft NW, Mermin DN. Solid state physics. Saunders College; 1976. [46] Tipler PA, Mosca G. Physics for scientists and engineers. W.H. Freeman; 2008. [47] ISO 17562. Fine ceramics (advanced ceramics, advanced technical ceramics) e test method for linear thermal expansion of monolithic ceramics by push-rod technique; 2001. [48] ASTME228. Standard test method for linear thermal expansion of solid materials with a push-rod dilatometer; 2011. [49] EN 821-1. Advanced technical ceramics - monolithic ceramics thermo-physical properties - part 1. Determination of thermal expansion 1995. [50] JIS R 1618. Measuring method of thermal expansion of fine ceramics by thermomechanical analysis; 2002. [51] Valentich J. Tube type dilatometers. Instrument Society of America; 1981. [52] Damani R, Rubesa D, Danzer R. Fracture toughness thermal shock behaviour of bulk plasma sprayed alumina - effects of heat treatment. J Eur Ceram Soc 2000;20:1439e52. [53] Kriegesmann J, editor. Technische keramische Werkstoffe. Ellerau: HvB Verlag GbR.; 1989. [54] Hasselman DPH. Unified theory of thermal shock fracture initiation and crack growth. J Am Ceram Soc 1969;52:600e4. [55] Bermejo R, Pascual Herrero J, Lube T, Danzer R. Optimal strength and toughness of Al2O3-ZrO2-laminates designed with external or internal compressive layers. J Eur Ceram Soc 2008;28:1575e83. [56] Sestakova L, Bermejo R, Chlup Z, Danzer R. Strategies for fracture toughness, strength and reliability optimisation of ceramic-ceramic laminates. Int J Mater Res 2011;102:613e26. [57] Bermejo R, Supancic P, Kraleva I, Morrell R, Danzer R. Strength reliability of 3D low temperature co-fired multilayer ceramics under biaxial loading. J Eur Ceram Soc 2011;31: 745e53. [58] Bejan A. Heat transfer. John Wiley & Sons, Inc.; 1993. [59] EN 820-2. Advanced technical ceramics - methods of testing monolithic ceramics - thermomechanical properties - part 2. Determination of self-loaded deformation 2003. [60] ASTME1461. Standard test method for thermal diffusivity by the flash method; 2007. [61] ISO 18755. Fine ceramics (advanced ceramics, advanced technical ceramics) e determination of thermal diffusivity of monolithic ceramics by laser flash method; 2005. [62] Lawn BR. Fracture of brittle solids. In: Davis EA, Ward IM, editors. Cambridge solid state science series. 2nd ed. Cambridge University Press; 1993.

Chapter | 9.1

Mechanical Properties of Ceramics

[63] Claussen N. Fracture toughness of Al2O3 with unstabilized ZrO2 dispersed phase. J Am Ceram Soc 1976;59:49e51. [64] Hannik RHJ, Kelly PM, Muddle BC. Transformation toughening in zirconia-containing ceramics. J Am Ceram Soc 2000;83:461e87. [65] Liu T. Characterization of three-dimensional through-tickness transformation zone in 9-mol%-Ce-TZP ceramics. J Am Ceram Soc 1994;77:2203e6. [66] Ru¨hle M, Claussen N, Heuer AH. Transformation and microcrack toughening as complementary process in ZrO2-toughened Al2O3. J Am Ceram Soc 1986;69:195e7. [67] Gu W-H, Faber KT, Steinbrech R,W. Microcracking and R-curve behaviour in SiC-TiB2 composites. Acta Metall Mater 1992;40: 3121e8. [68] Sigl LS. Microcrack toughening in brittle materials containing weak and strong interfaces. Acta Mater 1996;44:3599e609. [69] Becher PF, Hsueh C-H, Angelini P, Tiegs TN. Toughening behavior in whisker-reiforced ceramic matrix composites. J Am Ceram Soc 1988;71:1050e61. [70] Swanson PL, Fairbanks CJ, Lawn BR, Mai Y-M, Hockey BJ. Crack-interface grain bridging as a fracture resistance mechanism in ceramics: I, experimental study on alumina. J Am Ceram Soc 1987;70:279e89. [71] Reichl A, Steinbrech RW. Determination of crack-bridging forces in alumina. J Am Ceram Soc 1988;71:C299e301. [72] Becher PF, Sun EY, Kevin PP, Alexander KB, Hsueh C-H, Lin H-T, et al. Microstructural design of silicon nitride with improved fracture toughness: I, effect of grain shape and size. J Am Ceram Soc 1998;81:2811e30. [73] Becher PF. Microstructural design of toughened ceramics. J Am Ceram Soc 1991;74:222e69. [74] Danzer R, Sigl L. Influence of the microstructure on the toughness of ceramics. Veitsch Radex Rundschau 1994;1994:511e22. [75] Heuer AH, Claussen N, Kriven WM, Ru¨hle M. Stability of tetragonal ZrO2 particles in ceramic matrices. J Am Ceram Soc 1982;65:642e50. [76] Becher PF, Swain MV. Grain-size-dependent transformation behaviour in polycrystalline tetragonal zirconia. J Am Ceram Soc 1992;75:493e502. [77] Tomaszewski H, Boniecki M, Weglarz H. Effect of grain size on R-curve behaviour of alumina ceramics. J Eur Ceam Soc 2000;20: 2569e74. [78] Fu¨nfschilling S, Fett T, Hoffmann MJ, Oberacker R, Schwind T, Wippler J, et al. Mechanisms of toughnening in silicon nitrides: The roles of crack mbrindging and microstructure. Acta Mater 2011;59:3978e89. [79] ISO 15732. Fine ceramics (advanced ceramics, advanced technical ceramics) e test method for fracture toughness of monolithic ceramics at room temperature by single edge precracked beam (SEPB) method; 2003. [80] JIS R 1607. Testing methods for fracture toughness of fine ceramics at room temperature; 2010. [81] ASTMC1421. Standard test methods for determination of fracture. Toughness of advanced ceramics at ambient temperature 2007. [82] ISO 23146. Fine ceramics (advanced ceramics, advanced technical ceramics) e test methods for fracture toughness of monolithic ceramics e single-edge V-notch beam (SEVNB) method; 2008.

631

[83]

[84]

[85]

[86]

[87] [88]

[89] [90] [91] [92] [93] [94]

[95]

[96]

[97]

[98]

[99]

[100]

[101] [102] [103]

CEN/TS 14425-5. Advanced technical ceramics - test methods for determination of fracture toughness of monolitic ceramics - part 5: single-edge vee-notch beam (SEVNB) method; 2004. Damani R, Schuster C, Danzer R. Polished notch modification of SENB-S fracture toughness testing. J Eur Ceram Soc 1997;17: 1685e9. ISO 18756. Fine ceramics (advanced ceramics, advanced technical ceramics) e determination of fracture toughness of monolithic ceramics at room temperature by the surface crack in flexure (SCF) method; 2003. ISO 24370. Fine ceramics (advanced ceramics, advanced technical ceramics) e test method for fracture toughness of monolithic ceramics at room temperature by chevron-notched beam (CNB) method; 2005. Primas RJ, Gstrein R. ESIS TC6 round robin on fracture toughness. Fatigue Fract Eng Mater Struct 1997;20:513e32. Anstis GR, Chantikul P, Lawn BR, Marshall DB. A critical evaluation of indentation techniques for measuring fracture toughness: I, direct crack measurements. J Am Ceram Soc 1981;64:533e8. Quinn GD, Bradt RC. On the vickers intentation fracture test. J Am Ceram Soc 2007;90:673e80. Jayatilaka AdS, Trustrum K. Statistical approach to brittle fracture. J Mater Sci 1977;12:1426e30. Hunt RA, McCartney LN. A new approach to fracture. Int J Fract 1979:15. McCartney LN. Extensions of a statistical approach to fracture. Int J Fract 1979;15:477e87. Danzer RA. General strength distribution function for brittle materials. J Eur Ceram Soc 1992;10:461e72. Freudenthal AM. Statistical approach to fracture, in fracture. In: Liebowitz H, editor. New York, London: Academic Press; 1968. 591e619. Lamon J. Ceramics reliability: statistical analysis of multiaxial failure using the weibull approach and the multiaxial elemental strength. J Am Ceram Soc 1990;73:2204e12. Thiemeier T, Bru¨ckner-Foit A, Ko¨lker H. Influence of fracture criterion on the failure prediction of ceramics loaded in biaxial flexure. J Am Ceram Soc 1991;74:48e52. Danzer R, Supancic P, Pascual J, Lube T. Fracture statistics of ceramics - weibull statistics and deviations from weibull statistics. Eng Fract Mech 2007;74:2919e32. Hangl M. Processing related defects in green and sintered ceramics, in fractography of glasses and ceramics IV. In: Varner JR, Quinn GD, Fre`chette VD, editors. The American Ceramic Society. Westerville; 2001. p. 133e44. Danzer R, Supancic P, Harrer W, Wang Z, Bo¨rger A. Biaxial strength testing on mini specimens. In: Gdoutos EE, editor. Fracture of nano and engineering materials and structures. Dordrecht: Springer; 2006. p. 589e90. EN 843-5. Advanced technical ceramics - mechanical properties of monolithic ceramics at room temperature - part 5. Statistical analysis 1996. JIS R 1625. Weibull statistics of strength data for fine ceramics; 2010. ASTMC1683. Standard practice for size scaling of tensile strengths using weibull statistics for advanced ceramics; 2010. ISO 20501. Fine ceramics (advanced ceramics, advanced technical ceramics) e Weibull statistics for strength data; 2003.

632

[104] Cranmer DC, Richerson DW, editors. Mechanical testing methodology for ceramic design and reliability. New York: Marcel Dekker; 1998. 434. [105] EN 843-1. Advanced technical ceramics - mechanical properties of monolithic ceramics at room temperature - part 1. Determination of flexural strength 2006. [106] ASTMC1161. Standard test method for flexural strength of advanced ceramics at ambient temperature; 2008. [107] ASTMC1684. Standard test method for flexural strength of advanced ceramics at ambient temperature-cylindrical rod strength; 2008. [108] ISO 14704. Fine ceramics (advanced ceramics, advanced technical ceramics) e test method for flexural strength of monolithic ceramics at room temperature; 2008. [109] JIS R 1601. Testing method for flexural strength (modulus of rupture) of fine ceramics at room temperature; 2008. [110] ASTMC1499. Standard test method for monotonic equibiaxial flexural strength of advanced ceramics at ambient temperature; 2009. [111] Baratta FI, Matthews WT, Quinn GD. Errors associated with flexure testing of brittle materials. Watertown: U.S. Army Materials Technology Laboratory; 1987, 43. [112] Lube T, Manner M, Danzer R. The miniaturisation of the 4-point bend-test. Fatigue Fract Eng Mater Struct 1997;20:1605e16. [113] Bermejo R, Supancic P, Danzer R. Influence of measurement uncertainties on the determination of the Weibull modulus. J Eur Ceram Soc 2012;32:251e5. [114] Bo¨rger A, Supancic P, Danzer R. The ball on three balls test for strength testing of brittle discs - stress distribution in the disc. J Eur Ceram Soc 2002;22:1425e36. [115] Bo¨rger A, Supancic P, Danzer R. The ball on three balls test for strength testing of brittle discs - part II: analysis of possible errors in the strength determination. J Eur Ceram Soc 2004;24:2917e28. [116] Quinn GD, Morrell R. Design data for engineering ceramics: a review of the flexure test. J Am Ceram Soc 1991;74:2037e66.

Handbook of Advanced Ceramics

[117] Danzer R, Lube T. New fracture statistics for brittle materials. In: Bradt RC, Hasselmann DPH, Munz D, Sakai M, Shevchenkov VY, editors. Fracture mechanics of ceramics. New York: Plenum Publishing Corp.; 1996. p. 425e39. [118] Wiederhorn SM. Influence of water vapour on crack propagation in soda-lime glass. J Am Ceram Soc 1967;50:407e14. [119] Lawn BR. An atomistic model of kinetic crack growth in brittle solids. J Mater Sci 1975;10:469e80. [120] Schoeck G. Thermally activated crack propagation in brittle materials. Int J Fract 1990;44:1e14. [121] Danzer R, Lube T, Supancic P, Damani R. Fracture of ceramics. Adv Eng Mater 2008;10:275e98. [122] Lube T, Dusza J. A silicon nitride reference material - a testing program of ESIS TC6. J Eur Ceram Soc 2007;27:1203e9. [123] Richter H, Kleer G, Heider W, Ro¨ttenbacher R. Comparative study of the strength properties of slip-cast and of extruded silicon-infiltrated SiC. Mater Sci Eng 1985;71:203e8. [124] Creyke WEC, Sainsbury IEJ, Morrell R. Design with non-ductile materials. Applied Science Publishers; 1982. [125] EN 843-3. Advanced technical ceramics - mechanical properties of monolithic ceramics at room temperature - part 3. Determination of subcritical crack growth parameters from constant stressing rate flexural strength tests 2005. [126] Lube T, Baierl R. Sub-critical crack growth in alumina -a comparison of different measurement and evaluation methods. Berg- und Hu¨ttenma¨nnische Monatshefte 2011;156:450e8. [127] Tandon R, Newton CS, Monroe SL, Glass SJ, Roth CJ. Sub-critical crack growth behavior of a low-temperature co-fired ceramic. J Am Ceram Soc 2007;90:1527e33. [128] Bermejo R, Supancic P, Morrell R, Danzer R. Subcritical crack growth on low temperature co-fired ceramics under biaxial loading. Eng Fract Mech, in press. [129] Davidge RW. Mechanical behaviour of ceramics. Cambridge University Press; 1979.

Chapter 9.2

Testing and Evaluation of Mechanical Properties Tatsuki Ohji National Institute of Advanced Industrial Science and Technology, 2266-98 Anagahora, Shimoshidami, Moriyama, Nagoya, Aichi 463-8560, Japan

Chapter Outline 1. Introduction 2. Fracture Strength 3. Fatigue and Slow Crack Growth

633 633 638

4. Creep and Creep Rupture References

646 650

1. INTRODUCTION

2. FRACTURE STRENGTH

Because of their brittle nature, ceramics show totally different fracture behavior from that of metals and polymers. Ceramic engineers should properly understand this unique behavior when designing ceramic components for safe and reliable use. Fracture behavior of ceramics can be clarified through proper mechanical tests suitable to materials and their applications; so mechanical tests comprise a significant part of ceramic technology. Over the years, substantial progress has been made in our understanding of mechanical properties of ceramics as well as test methods needed to determine those properties. This chapter addresses mechanical properties of advanced ceramics; particular interest is given to fast fracture strength and time-dependent failure properties such as slow crack growth, fatigue, and creep. The testing and evaluation procedures are reviewed for establishing the reliability specific to each of these properties, together with some related research topics. Numerous international and domestic standards related to these procedures have been issued in recent years, which are also listed in this chapter. More detailed information on the testing and evaluation procedures is available in these standards or other literature [1]. Fracture toughness, which is one of the most important parameters for expressing brittle fracture of ceramics, is not discussed here in detail however, since it is dealt with in the Chapter, “Fracture Mechanics Measurements.”

Fracture strength is one of the most common and widely used mechanical properties for structural ceramics because it is the most simple and important material reliability parameter. Ceramic component design demands critical investigation of fracture strength and its distribution due to the nature of probabilistic fracture of ceramics. A number of methodologies and techniques have been developed for determining ceramics fracture strength [2e32]. Most of these techniques equate fracture strength to maximum stress (tensile or compressive) at fracture. Consequently, in order for a particular load and specimen geometry to be useful for determination of fracture strength, stress distribution must be well established. A complicating factor in the determination of fracture strength is that ceramic materials strength is quite sensitive to size, shape, and surface finish. Such sensitivity is largely responsible for wide variation in strength values reported for a given material. Viable test methodology must therefore account for these effects. This section describes the commonly used methods, typically divided into four types: flexural tests, tensile tests, compressive tests, and biaxial tests, followed by the related standards.

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00034-4 Copyright Ó 2013 Elsevier Inc. All rights reserved.

2.1. Testing Methods 2.1.1. Tensile Tests Uniaxial tensile tests have been widely used for many years in order to obtain information on mechanical strength

633

634

Handbook of Advanced Ceramics

properties of industrial materials. However, in the field of ceramics, flexural tests are commonly used instead of tensile tests. This is because stress concentration occurring during tests cannot be relaxed by local plastic deformation in ceramics; hence, even slight eccentric loading in tensile tests causes parasitic bending stress. Susceptibility of stress concentration also causes improper fractures inside the gauge section. Especially in high-temperature measurements, the degree of difficulty increases because material and structure of the testing fixture are restricted. On the other hand, ceramic elasticity is usually maintained until fracture occurs, enabling researchers to determine strength in flexural tests by elementary stress elastic beam equations. In addition, flexural tests are superior to tensile tests due to the simple specimen geometry, low cost of specimen preparation, and adaptability to high-temperature measurements. Nevertheless, tensile tests have the following advantages. First, fracture strength and its distribution can be determined from the stress field within large effective volumes of the material. This is most useful for reliable design of large mechanical components. Second, fractures in tensile tests more frequently originate from internal natural fracture defects than those in flexural ones, resulting in more intrinsic strength. Third, even in the high-temperature region where the material shows nonelastic behavior and validity of stress calculation based on the elastic body in flexural tests is questionable, tensile tests can determine stress precisely and simply. Also, in such a region, tensile tests are most advantageous for determining stressestrain curves, and creep curves, because deformation amounts can be detected properly with an appropriate extensometer. Examples of nonlinear

stressestrain curves of silicon nitride at high temperatures in tension are shown in Figure 1. As already stated, there are two major barriers to ceramic tensile tests. One is eccentric loading which results in unexpected parasitic bending strain. The other is stress concentration inside the gauge section. In general cases, the latter barrier can be overcome by employing the long gauge section, which is distant from the loading points with moderate change of cross section (which is due to St. Vennant’s effect). For the first barrier, there are several sources of eccentricity including off-center loading, end bending moment, end torque moment, and an imperfect or curved specimen. Difficulty in solving the first barrier exists in these sources being involved with it in the complex. Sources include an improperly aligned load train, improper specimen/grip interface, imperfect (e.g. curved or nonsymmetric) specimen geometry, and so on. Examples of successful tensile testing ceramics are available in Refs. [2e15]. The geometry of tensile test specimen depends on the purpose of the tensile test itself. When a tensile test is performed for knowing the tensile strength of an as e fabricated component, its dimension limits the dimensions of the specimen. Gripping devices also restrict the design of the test specimen geometry. The range of possible tensile test specimen geometries is illustrated in ISO 15490, “Fine ceramics (advanced ceramics, advanced technical ceramics) e Test method for tensile strength of monolithic ceramics at room temperature,” and ASTM C 1273, “Standard test method for tensile strength of monolithic advanced ceramics at ambient temperatures.”

FIGURE 1 Stressedisplacement curves showing effect of strain rate for a hot-pressed silicon nitride at 1260  C.

Chapter | 9.2

Testing and Evaluation of Mechanical Properties

2.1.2. Compressive Tests Although the compressive strength of ceramic materials is generally much higher than tensile strength, the dominant loading mode in many actual applications is compression. Reliable component design in these applications requires precise compressive strength measurements. Compressive strength measurements are also necessary for development of fundamental insights into micromechanical and statistical aspects of compressive failure. Compressive failure is thought to involve coalescence of damage in the form of microcracks and microvoids which propagate by alternating growth and arrest mechanisms [16]. For ceramic materials tested at ambient temperatures, this damage is thought to be generated by localized microplasticity arising from twinning and slip [17]. Failure occurs through structural collapse when density of damage in a zone reaches a critical level. Because this size appears to depend only upon intrinsic material characteristics, compressive strength is not generally affected by specimen size. For uniaxial testing, compressive stress is given simply by the applied load divided by the cross-sectional area. However, high stresses required for failure under compressive loading have created a number of problems in test fixture design and specimen geometries required for successful compression strength measurement. Particularly, problems related to improper alignment and load block stress concentrations can lead to the generation of tensile stresses in the specimen sufficient to cause failure. Therefore, it should be noted that measured compressive strength tends to underestimate true values. Examples for successful measurements of compressive strength in ceramics are available in References [16e21]. In recent years, compressive tests have been most frequently used for measuring strength of porous ceramics, which are now expected to be adopted for a wide variety of industrial applications from filtration, absorption, catalysts, and catalyst supports to lightweight structural components [21e24]. It has been known that the strength widely scatters, by one to two orders of magnitude, even for the same porosity, depending on the porous structure, etc. [24].

635

volume of specimen, and other reasons. When resultant flexural strength data are used for component design, one must assume that strength limiting flaw distribution (pores, surface flaws, agglomerates, inclusions, etc.) is equivalent to other methods. Therefore, surface machining processes for test specimens are extremely important; when machining processing causes severe surface or subsurface damage, it becomes difficult to obtain intrinsic strength. This tendency increases with higher material strength. Such materials have limited fracture origin size and require damage-less surface treatment. Two kinds of flexure loading are commonly used: threepoint loading and four-point loading (Figure 2). In fourpoint flexure, pure flexure moment without shear stress is generated within an inner span and beam theory can be applied more properly for strength calculation than in three-point flexure. Also, four-point flexure has larger effective volume or effective surface than three-point flexure for identical specimen dimensions. Still, three-point loading has simpler geometry and simplifies testing fixture design. This is particularly important when adapted to hightemperature measurements or with limited available specimen size. The fracture strength, sf, is equated to the maximum tensile stress generated at the failure load, P, as follows: 1:5PðLo Li Þ (1) bh2 where Lo and Li are the outer and inner load spans, respectively (Li is zero for three-point loading), and b and h are the width and height of the rectangular specimen. sf ¼

2.1.3. Flexural Tests Due to advantages including simple specimen geometry, low-cost specimen fabrication, and adaptability to hightemperature measurements, flexural tests are most widely used for measuring ceramic materials strength [25e41]. Compared to other testing methods such as tension and compression, specimens and testing fixtures for flexural tests are much simpler and, consequently, less expensive. Flexural tests, however, are disadvantageous by vulnerability of strength to surface machining, small effective

FIGURE 2 Three-point and four-point loading geometries.

636

Handbook of Advanced Ceramics

In recent years, application of ceramic materials has been expanded widely and flexural strength measurements have been reported for various types of ceramics including lead zirconate titanate (PZT) ceramics [33e35], armor ceramics [36], ultra-high-temperature ceramics [37], and dental ceramics [38e40]. Flexure tests have been also used for determining strength of ceramic tubes specifically for circumstances when flaws located at the tube’s outer diameter are the strength limiters (e.g., heat exchangers) [41].

2.1.4. Biaxial Tests Strength measurements are usually performed in uniaxial loading. Most ceramic components are subjected to biaxial (or multiaxial) loading in actual practice. In addition, biaxial loading conditions are much more severe in a stress state than uniaxial conditions, leading to more reliable component design. One of the most frequently used biaxial loading techniques is biaxial flexure using thin plates as specimens. Biaxial flexure is free from the specimen edge effects which substantially affect strength measurements in three- or four-point flexural tests. This technique also can reflect flaws of any direction. In biaxial stress fields, a larger number of flaws are subjected to tensile stress than in uniaxial stress fields, resulting in more conservative strength evaluation. In addition, simplified loading geometry eases specimen preparation and lowers its cost. Several types of loading methods have been employed to evaluate biaxial flexural strength: ball-on-ring, pistonon-ring, and ring-on-ring. [42e57]. The ring can be substituted for balls placed at the circumference of the same radius since applied stresses are independent of angles between and the number of supports for this loading geometry. The ball-on-three-ball and the piston-on-threeball formations are therefore very frequently used instead, as shown in Figure 3. One advantage of the ball-on-threeball is its tolerance of rough specimen surfaces, simplifying surface machining. The fracture strength, sf, can be calculated from the failure load, P, as follows: sf ¼

   3Pð1 þ nÞ a 1 n 1 þ 2 ln þ 1 4pt2 b 1þn

b2 2a2



a2 R2

FIGURE 3 Wachtman-typed piston-on-three-ball test fixture is schematically illustrated.

devices, information storage media, and biomedical devices, and biaxial flexural tests have been used for the strength evaluation of ceramic components in such applications [50,51]. In recent years, this test method also has been employed most frequently to determine strength of dental materials [52e56].

2.2. Related Standards In these past few decades, considerable efforts have been made to standardize fracture strength measurements for both room and high temperatures. Applicable general standards for strength measurement tests of advanced ceramics are as follows (some additional standards for specific uses not included): ISO (International Organization for Standardization), TC206 “Fine ceramics (advanced ceramics, advanced technical ceramics)” l

l



(2)

where a is the radius of the support ring, R and t are the specimen radius and thickness, respectively, and n is the Poisson’s ratio. b is the radius of the piston for the piston-onring, that of the contact area of the loading ball for the ballon-ring, or that of the loading ring for the ring-on-ring. Similar to other strength measurements of ceramics, biaxial strength of ceramics is affected by the effective volume or area of the stressed portion of the specimen [49]. Thin plates and minute components are now used in many technical applications including semiconductor

l

l

l

l l

ISO 13124: Test method for interfacial bond strength of ceramic materials ISO 14704: Test method for flexural strength of monolithic ceramics at room temperature ISO 15490: Test method for tensile strength of monolithic ceramics at room temperature ISO 15733: Test method for tensile stressestrain behavior of continuous, fiber-reinforced composites at room temperature ISO 17565: Test method for flexural strength of monolithic ceramics at elevated temperature ISO 20501: Weibull statistics for strength data ISO 20504: Test method for compressive behavior of continuous-fiber-reinforced composites at room temperature

Chapter | 9.2

l

l

Testing and Evaluation of Mechanical Properties

ISO 20505: Determination of the interlaminar shear strength of continuous-fiber-reinforced composites at ambient temperature by the compression of doublenotched test pieces and by the Iosipescu test ISO 20506: Determination of the in-plane shear strength of continuous-fiber-reinforced composites at ambient temperature by the Iosipescu test

ASTM International (formerly known as the American Society for Testing and Materials), C28 “Advanced Ceramics” l

l

l

l

l

l

l

l

l

l

l

l

l

l

C1161: Standard Test Method for Flexural Strength of Advanced Ceramics at Ambient Temperature C1211: Standard Test Method for Flexural Strength of Advanced Ceramics at Elevated Temperatures C1239: Standard Practice for Reporting Uniaxial Strength Data and Estimating Weibull Distribution Parameters for Advanced Ceramics C1273: Standard Test Method for Tensile Strength of Monolithic Advanced Ceramics at Ambient Temperatures C1275: Standard Test Method for Monotonic Tensile Behavior of Continuous Fiber Reinforced Advanced Ceramics with Solid Rectangular Cross Section Test Specimens at Ambient Temperature C1292: Standard Test Method for Shear Strength of Continuous Fiber Reinforced Advanced Ceramics at Ambient Temperatures C1322: Standard Practice for Fractography and Characterization of Fracture Origins in Advanced Ceramics C1323: Standard Test Method for Ultimate Strength of Advanced Ceramics with Diametrally Compressed C-Ring Specimens at Ambient Temperature C1341: Standard Test Method for Flexural Properties of Continuous Fiber Reinforced Advanced Ceramic Composites C1358: Standard Test Method for Monotonic Compressive Strength Testing of Continuous Fiber Reinforced Advanced Ceramics with Solid Rectangular Cross Section Test Specimens at Ambient Temperatures C1359: Standard Test Method for Monotonic Tensile Strength Testing of Continuous Fiber Reinforced Advanced Ceramics With Solid Rectangular Cross Section Test Specimens at Elevated Temperatures C1366: Standard Test Method for Tensile Strength of Monolithic Advanced Ceramics at Elevated Temperatures C1424: Standard Test Method for Monotonic Compressive Strength of Advanced Ceramics at Ambient Temperature C1425: Standard Test Method for Interlaminar Shear Strength of 1 D and 2 D Continuous Fiber Reinforced Advanced Ceramics at Elevated Temperatures

l

l

l

l

l

l

l

l

637

C1468: Standard Test Method for Transthickness Tensile Strength of Continuous Fiber Reinforced Advanced Ceramics at Ambient Temperature C1469: Standard Test Method for Shear Strength of Joints of Advanced Ceramics at Ambient Temperature C1495: Standard Test Method for Effect of Surface Grinding on Flexure Strength of Advanced Ceramics C1499: Standard Test Method for Monotonic Equibiaxial Flexural Strength of Advanced Ceramics at Ambient Temperature C1557: Standard Test Method for Tensile Strength and Young’s Modulus of Fibers C1674: Standard Test Method for Flexural Strength of Advanced Ceramics with Engineered Porosity (Honeycomb Cellular Channels) at Ambient Temperatures C1683: Standard Practice for Size Scaling of Tensile Strengths Using Weibull Statistics for Advanced Ceramics C1684: Standard Test Method for Flexural Strength of Advanced Ceramics at Ambient Temperature Cylindrical Rod Strength

CEN (European Committee for Standardization), TC184 “Advanced Technical Ceramics” l

l

l

l

l

l

l

l

EN 658-1: Advanced technical ceramics e Mechanical properties of ceramic composites at room temperature e Part 1: Determination of tensile properties EN 658-2: Advanced technical ceramics e Mechanical properties of ceramic composites at room temperature e Part 2: Determination of compression properties EN 658-3: Advanced technical ceramics e Mechanical properties of ceramic composites at room temperature e Part 3: Determination of flexural strength EN 658-4: Advanced technical ceramics e Mechanical properties of ceramic composites at room temperature e Part 4: Determination of interlaminar shear strength by compression loading of notched test specimens EN 658-5: Advanced technical ceramics e Mechanical properties of ceramic composites at room temperature e Part 5: Determination of interlaminar shear strength by short span bend test (three points) EN 820-1: Advanced technical ceramics e Methods of testing monolithic ceramics e Thermomechanical properties e Part 1: Determination of flexural strength at elevated temperatures EN 843-1: Advanced technical ceramics e Mechanical properties of monolithic ceramics at room temperature e Part 1: Determination of flexural strength EN 843-5: Advanced technical ceramics e Mechanical properties of monolithic ceramics at room temperature e Part 5: Statistical analysis

638

l

l

l

l

l

l

l

l

l

l

Handbook of Advanced Ceramics

EN 843-6: Advanced technical ceramics e Mechanical properties of monolithic ceramics at room temperature e Part 6: Guidance for fractographic investigation EN 843-7: Advanced technical ceramics e Mechanical properties of monolithic ceramics at room temperature e Part 7: C-ring tests EN 843-8: Advanced technical ceramics e Mechanical properties of monolithic ceramics at room temperature e Part 8: Guidelines for conducting proof tests EN 1007-4: Advanced technical ceramics e Ceramic composites e Methods of test for reinforcement e Part 4: Determination of tensile properties of filaments at ambient temperature EN 1007-5: Advanced technical ceramics e Ceramic composites e Methods of test for reinforcements e Part 5: Determination of distribution of tensile strength and of tensile strain to failure of filaments within a multifilament tow at ambient temperature EN 1007-6: Advanced technical ceramic e Ceramic composites e Methods of test for reinforcements e Part 6: Determination of tensile properties of filaments at high temperature EN 1007-7: Advanced technical ceramics e Ceramic composites e Methods of test for reinforcements e Part 7: Determination of the distribution of tensile strength and of tensile strain to failure of filaments within a multifilament tow at high temperature EN 1894: Advanced technical ceramics e Mechanical properties of ceramic composites at high temperature under inert atmosphere e Determination of shear strength by compression loading of notched specimens EN 12788: Advanced technical ceramics e Mechanical properties of ceramic composites at high temperature under inert atmosphere e Determination of flexural strength EN 12789: Advanced technical ceramics e Mechanical properties of ceramic composites at high temperature under air at atmospheric pressure e Determination of flexural strength

JISC (Japanese Industrial Standards Committee), Divisional Council on Ceramics l

l

l

l

l

JIS R 1601: Testing method for flexural strength (modulus of rupture) of fine ceramics at room temperature JIS R 1604: Testing method for flexural strength (modulus of rupture) of fine ceramics at elevated temperature JIS R 1606: Testing methods for tensile strength of fine ceramics at room and elevated temperature JIS R 1608: Testing methods for compressive strength of fine ceramics JIS R 1624: Testing method for bending strength of fine ceramic joint

l

l

l

l

l

l

l

l

l

JIS R 1625: Weibull statistics of strength data for fine ceramics JIS R 1630: Test methods for tensile strength of fine ceramic joint JIS R 1643: Testing methods for interlaminar shear strength of continuous fiber-reinforced ceramic composites JIS R 1656: Testing methods for tensile stressestrain behavior of continuous fiber-reinforced ceramic matrix composites at room temperature JIS R 1663: Testing method for flexural strength of continuous fiber-reinforced ceramic matrix composites JIS R 1664: Testing method for bending strength of porous fine ceramics JIS R 1673: Test method for compressive behavior of continuous fiber-reinforced ceramic matrix composites at room temperature JIS R 1674: Statistical analysis for influence of machining damage on strength of fine ceramics JIS R 1687: Testing method for tensile behavior of continuous fiber-reinforced ceramic matrix composites at elevated temperatures

3. FATIGUE AND SLOW CRACK GROWTH Fatigue, or fracture after loading of some duration at a stress level less than that required for immediate fracture, is caused by several mechanisms. (While the term fatigue in the metallurgy community indicates solely cyclic fatigue, cyclic mechanical loading of a specimen, or a component, fatigue for ceramic and glass includes static and dynamic fatigue as well.) One well-known mechanism is environmentally enhanced crack growth, very frequently referred to as slow crack growth (SCG), which occurs in nearly all structural ceramics [58]. This is crack growth enhanced by some chemical reaction, which substantially weakens the atomic bond at the crack tip even at stresses well below intrinsic strength. The SCG behavior of ceramics and glass is generally expressed by a relationship between crack velocity, v, and an applied stress intensity factor, K, to show the so-called veK diagram. These veK diagrams of ceramics and glasses typically contain stages I, II, and III with threshold stress intensity below which SCG does not occur (Kth), as shown in Figure 4. Crack velocity at each stage of the diagram is governed by a different mechanism; and the diagram varies substantially depending on materials and environments. Some cases show no threshold stress intensity while in other cases stages II and III are negligible. Once the veK diagram is established for a particular material and condition, fracture mechanics can predict its lifetime. Since stage

Chapter | 9.2

Testing and Evaluation of Mechanical Properties

FIGURE 4 Schematic relationship between the crack velocity and applied stress intensity K during SCG.

I crack propagation occupies a large part of total lifetime, this stage is essential for critical lifetime prediction. A typical example of veK diagram exists in the classic work by Wiederhorn [59]. They measured crack growth of soda lime silica glass in inert nitrogen environments with different humidities and obtained veK diagrams which all consist of stages I, II, and III. It is generally considered that stages I, II, and III correspond to impurity controlled activity, impurity diffusion limited, and bulk environment controlled regions, respectively. In the above crack growth study, Wiederhorn considered that moisture activity controls crack velocity at stage I, while stage II is limited by moisture diffusivity. Thus, moisture concentration is decisive and insensitive to KI, leading to the so-called plateau region. Here, KI is the applied stress intensity factor in mode I. Great effort has been made to understand fundamental mechanisms determining the relationship between SCG and stress intensity factor. Although a number of mathematical expressions to describe this relationship have been proposed, the following forms are used presently: v ¼ v0 ðKI =KIC ÞN

(3)

v ¼ v0 expðKI =KIC ÞB

(4)

Here, v0, N, and B are constant, and KIC is the critical applied stress intensity factor, or fracture toughness, of the material. Neither expression has obtained general acceptance as a “fundamental” form. Eqn (3) is most commonly employed in component design due to its ease in mathematical manipulation; log (v) is linear in log (K), with

639

N slope. Parameter N is recognized as an indicator of susceptibility to SCG; the larger the N value, the smaller the SCG. However, it is frequently observed that obtained N values do not agree between two different measurement techniques or two different specimen sizes for identical materials and environments [60]. This indicates that the power law form does not always adequately express crack velocity at small stress intensity factors. This shortcoming significantly limits potential of extrapolating the smaller velocity range. Several techniques have been proposed for measuring a veK diagram. Many researchers have used direct measurement of propagating cracks in a fracture mechanics test specimen. [59e61]. An advantage of this method is that a veK diagram can be drawn easily from one specimen or one test. The fracture mechanics test specimen usually has a long artificial crack for which the K value can be determined easily. However, such long artificial cracks do not always behave similarly to short natural cracks. Ceramic components are usually inspected through nondestructive tests before service so that they do not contain intolerant long cracks. Thus, it is short cracks that govern the lifetime of ceramic components in service. Another way to establish a veK diagram is to estimate it indirectly from strength measurement results of static or dynamic fatigue tests [62e68]. Static fatigue tests involve static loading of tensile stresses lower than the fast fracture strength until failure occurs: this shows the relationship between time to failure and stress. In dynamic fatigue tests, tensile stress is monotonically increased at various loading rates until there is a failure to obtain a relationship between loading rate and strength. Both obtained relations lead to a veK relationship using fracture mechanics. One advantage of these methods is that observed crack propagation behavior can be known from short flaws. This has practical importance, as mentioned above. Short cracks can occur in ceramic components as pores, inclusions, and machining surface flaws. When KIC depends on crack length a (or R-curve behavior), SCG behavior differences between short and long cracks become clear. As shown schematically in Figure 5, short crack extension starts at considerably lower stress intensities than long cracks for materials with Rcurves; and plateau fracture toughness is not attained before catastrophic failure [67,68]. Unlike tests where long cracks are forced to initiate from a notch tip in a controlled fashion, short cracks originate from local natural flaws that are frequently located at the surface. The short crack with the shallowest R-curve slope becomes unstable first (the fracture stress at this instability corresponds to flexure stress in this instance). This indicates that under equivalent loading conditions, specimens or components cannot take advantage of the full R-curve potential. In addition, crack resistance measurements generated from long cracks might

640

FIGURE 5 In a material with a characteristic R-curve behavior, short or natural cracks show a different behavior from that of long cracks.

therefore overestimate ceramic component toughness in real, load-bearing applications. The relationship between material microstructure and its properties has been addressed; also, the veK relationship is often characterized by information on interactions between the moving crack and the microstructure [69,70]. Importance of fractographic analysis on specimens after the test has been recognized with strength measuring techniques because this can show whether cracks propagate trans- or intergranularly, thereby clarifying effects of grain boundary phases on microstructural crack propagation. For specimens tested at high temperatures, fractography can determine whether strength degradation arises from SCG or other mechanisms such as creep [71,72]. Furthermore, degree of moving crack velocity can be estimated in some cases from fractography, as cracks grow trans- or intergranularly depending on their velocity [66]. Finally, direct observation of indentation cracks has been proposed recently for investigating both the veK relationships and fundamental aspects of flaw shape evolution [73,74]. This section describes fatigue and slow crack growth tests including: (1) static and dynamic fatigue for short (or natural) cracks, (2) cyclic fatigue for long cracks, (3) static and dynamic fatigue for long cracks, and (4) cyclic fatigue for long cracks, followed by the related standards. In addition, we will discuss “Cyclic Fatigue of Toughened Ceramics” for a related research topic.

3.1. Testing Methods 3.1.1. Static and Dynamic Fatigue Tests Using Short Cracks Fatigue tests for specimens containing short natural flaws can provide data closest to the real service application. The static test involves applying a mode I, plane strain static

Handbook of Advanced Ceramics

stress in an appropriate environment and measuring the time to failure. Conducting measurements at different applied stresses, the veK curve can be established even though natural flaw sizes are unknown. The methods are essentially identical to simple strength measurement methods. Principally, the fatigue parameter N can be measured without any assumptions about flaw size or shape. However, this method is very time consuming to obtain even one data point and a number of specimens are needed for each condition to account satisfactorily for statistical variability in strength. Dynamic fatigue generates principally the same results, but testing time is shortened through testing at several constant stress rates. Any specimen configuration which is generally used for any strength test can be used for both fatigue tests. Examples of successful measurements on static and dynamic fatigue of ceramics are available for short artificial cracks [71e77] and natural cracks [78e83].

3.1.2. Cyclic Fatigue Tests Using Short Cracks Structural ceramic components in service very often are subjected to cyclic mechanical loading. Cyclic fatigue performance of structural ceramics determines the life of components in many cases, including gas turbine rotors or blades and diesel engine valves. Therefore, characterizing and properly understanding ceramic cyclic fatigue is crucial for confident use of components subjected to cyclic loading. Many reports explain cyclic fatigue of ceramics using short or natural flaws [84e94]. Several mechanisms for cyclic fatigue of ceramics have been proposed. Most of them are related strongly to toughening mechanisms of ceramics at room temperature. They include degradation of transformation toughening zone at the crack tip (such as in some zirconia); damage to bridging reinforcements (grains, whiskers, platelets, and brittle and/or ductile particles) over the bridged interface; cyclic loading-induced accumulation of microcracks around the crack tip; crack tip blunting/resharpening; and relaxation of any mechanically beneficial residual stress due to microcracking, as shown in Figure 6. It should be noted that these degradation mechanisms are not active in static or dynamic fatigue. In some cases, at elevated temperatures where softening of the grain boundary glassy phase can occur, other mechanisms become active. These include oxidationinduced deformation and viscoelastic effects [84,85]. Lead zirconate titanate (PZT) ceramics have potential for increased use in advanced industrial applications such as piezoelectric actuators and ultrasonic vibrators because of their quick response, compactness, and good power efficiency. When they are subjected to a powerful alternating electric field, cyclic stresses are introduced into these piezoelectric ceramics which can lead to fracture

Chapter | 9.2

Testing and Evaluation of Mechanical Properties

641

FIGURE 6 Schematic classification of primary microstructural toughening mechanisms, which are strongly involved with property degradation during cyclic loading.

and/or microcracking. Moreover, degradation of piezoelectric properties is a serious practical concern, as this is a major cause of reduced device lifetime. Therefore, in recent years, many studies have been devoted for clarifying cyclic fatigue behavior and mechanical reliability of this kind of material [95e110]. Ceramic materials have been also extensively applied in prosthodontics in recent years, since all-ceramic crowns are superior to traditional porcelain-fused-to-metal crowns in esthetics, wear resistance, and chemical inertness. In such applications, the materials are susceptible to cyclic fatigue that substantially degrades their mechanical reliability and structural durability. Therefore, considerable efforts have been made to investigate cyclic fatigue behavior of various ceramics that can be used for these applications [55,111e115].

3.1.3. Static and Dynamic Fatigue Tests Using Long Cracks Crack propagation testing using long cracks (i.e., fracture mechanics specimens) is fundamentally different from crack propagation testing with short or natural flaws (i.e., strength measurement specimens) for several reasons. Crack length being propagated may be on the order of millimeters, while crack lengths dealt with in short fatigue crack propagation were on the order of micrometers or tens or hundreds of micrometers. “Long” cracks are not easily influenced by microstructural residual stresses while “short” cracks are. By using fracture mechanics specimens (i.e., specimens which are often used for measuring fracture

FIGURE 7 Frequently used long-crack test geometries: (a) compact tension (CT); (b) single-edge notched beam (SENB); (c) chevron notch (CN); (d) double torsion (DT); (e) double cantilever beam (DCB). Grayed area is crack.

toughness), one can measure directly the crack velocity v as a function of applied stress intensity K [59e61]. However, test specimens and the experimental facility are more complicated; it is probably difficult to operate at high temperature or in an extremely corrosive environment. Specimens may be loaded in load control or in displacement control and relaxation techniques may also be used. Figure 7 shows examples of frequently used longcrack test geometries. Test specimens must have a relatively large sharp crack which will propagate during testing. Ideally, the crack tip should be atomically sharp. However, a notch produced by a diamond saw has been used frequently, assuming the sharp crack is produced from tip of the notch. Several studies have been conducted to investigate slow-crack growth behavior and mechanical properties of inert bioceramics, such as alumina- and yttria-stabilized zirconia, which are used as components for orthopedic and dental applications [116e119].

3.1.4. Cyclic Fatigue Tests Using Long Cracks Also, there is substantial difference in cyclic loading between crack propagation with long cracks and that with

642

Handbook of Advanced Ceramics

short or natural flaws. The so-called S-N (stress-number of cycles) curves represent results of fatigue crack propagation testing of short cracks, while crack propagation per cycle as a function of stress intensity range or maximum stress intensity represents results of crack propagation testing with long cracks. However, toughening mechanisms, which are related to enhanced crack propagation in cyclic fatigue with short cracks, can be similarly related to that in cyclic fatigue with long cracks. There are examples of successful measurements on cyclic fatigue of ceramics with long cracks [120e129]. Similar to the veK relationships described above, crack propagation rates or crack growth extension per unit cycle (da/dN) for cyclic fatigue testing with long cracks are also typically represented by a power-law relationship. The commonly known Paris law [130] relates da/dN to the stress intensity range (DK ¼ KImax KImin, with KImax and KImin being maximum and minimum applied stress intensities, respectively) as da=dN ¼ CðDKÞm

(5)

Figure 8 shows comparison of growth rate data for some structural ceramics and metal alloys.

3.2. Related Standards Applicable general standards for fatigue, slow crack growth, and fracture toughness tests of advanced ceramics are as follows (some additional standards for specific uses not included). It should be noted that most standards for simple strength measurements can be applied to fatigue tests with natural cracks, by changing loading rates, modes, etc. ISO (International Organization for Standardization), TC206 “Fine ceramics (advanced ceramics, advanced technical ceramics)” l

l

where C is a constant and m is a cyclic fatigue exponent. Crack propagation rate (da/dt) or v is related to da/dN as da=dt ¼ v ¼ ðda=dNÞf

(6)

where f is the cyclic frequency. For determining the cyclic fatigue exponent, m, the crack growth increment per cycle da/dN, or the crack propagation rate da/dt, as a function of stress intensity range, DK. While typical m values of most metals range between 2 and 4, those of ceramics are as high as 50 or larger in mid-growth rate regime [120,122]. FIGURE 8 Cyclic fatigue crack growth rate, da/dN, as a function of the applied stress intensity range, DK, for a range of ceramics, including, Mg-PSZ, pyrolytic carbon, coarse-grained Al2O3, SiC whisker-reinforced Al2O3, and columnar-grained b-Si3N4. Data are compared with two high-strength metallic alloys [120]. (Reprinted with permission of Elsevier. All rights reserved)

l

l

ISO 15732: Fine ceramics (advanced ceramics, advanced technical ceramics) e Test method for fracture toughness of monolithic ceramics at room temperature by single-edge precracked beam (SEPB) method ISO 18756: Fine ceramics (advanced ceramics, advanced technical ceramics) e Determination of fracture toughness of monolithic ceramics at room temperature by the surface crack in flexure (SCF) method ISO 22214: Fine ceramics (advanced ceramics, advanced technical ceramics) e Test method for cyclic bending fatigue of monolithic ceramics at room temperature ISO 23146: Fine ceramics (advanced ceramics, advanced technical ceramics) e Test methods for

Chapter | 9.2

l

Testing and Evaluation of Mechanical Properties

fracture toughness of monolithic ceramics e Singleedge V-notch beam (SEVNB) method ISO 24370: Fine ceramics (advanced ceramics, advanced technical ceramics) e Test method for fracture toughness of monolithic ceramics at room temperature by chevron-notched beam (CNB) method

ASTM International (formerly known as the American Society for Testing and Materials), C28 “Advanced Ceramics” l

l

l

l

l

l

C1360: Standard Practice for Constant Amplitude, Axial, TensioneTension Cyclic Fatigue of Continuous Fiber-Reinforced Advanced Ceramics at Ambient Temperatures C1361: Standard Practice for Constant Amplitude, Axial, TensioneTension Cyclic Fatigue of Advanced Ceramics at Ambient Temperatures C1368: Standard Test Method for Determination of Slow Crack Growth Parameters of Advanced Ceramics by Constant Stress Rate Strength Testing at Ambient Temperature C1421: Standard Test Methods for Determination of Fracture Toughness of Advanced Ceramics at Ambient Temperature C1465: Standard Test Method for Determination of Slow Crack Growth Parameters of Advanced Ceramics by Constant Stress Rate Flexural Testing at Elevated Temperatures C1576: Standard Test Method for Determination of Slow Crack Growth Parameters of Advanced Ceramics by Constant Stress Flexural Testing (Stress Rupture) at Ambient Temperature

CEN (European Committee for Standardization), TC184 “Advanced Technical Ceramics” l

l

l

l

l

EN 843-3: Advanced technical ceramics e Mechanical properties of monolithic ceramics at room temperature e Part 3: Determination of subcritical crack growth parameters from constant stressing rate flexural strength tests EN 13234: Advanced technical ceramics e Mechanical properties of ceramic composites at ambient temperature e Evaluation of the resistance to crack propagation by notch sensitivity testing CEN/TS 14425-1: Advanced technical ceramics e Test methods for determination of fracture toughness of monolithic ceramics e Part 1:Guide to test method selection EN 14425-3: Advanced technical ceramics e Test methods for determination of fracture toughness of monolithic ceramics e Part 3: Chevron-notched beam (CNB) method CEN/TS 14425-5: Advanced technical ceramics e Test methods for determination of fracture toughness of

l

l

l

l

l

643

monolithic ceramics e Part 5: Single-edge V-notch beam (SEVNB) method EN 15156: Advanced technical ceramics e Mechanical properties of ceramic composites at room temperature e Determination of fatigue properties at constant amplitude EN 15157: Advanced technical ceramics e Mechanical properties of ceramic composites at high temperature in air at atmospheric pressure e Determination of fatigue properties at constant amplitude EN 15158: Advanced technical ceramics e Mechanical properties of ceramic composites at high temperature under inert atmosphere e Determination of fatigue properties at constant amplitude EN ISO 15732: Fine ceramics (advanced ceramics, advanced technical ceramics) e Test method for fracture toughness of monolithic ceramics at room temperature by single-edge precracked beam (SEPB) method EN ISO 18756: Fine ceramics (advanced ceramics, advanced technical ceramics) e Determination of fracture toughness of monolithic ceramics at room temperature by the surface crack in flexure (SCF) method

JISC (Japanese Industrial Standards Committee), Divisional Council on Ceramics l

l

l

l

l

l

l

l

JIS R 1607: Testing methods for fracture toughness of fine ceramics at room temperature JIS R 1617: Testing method for fracture toughness of fine ceramics at elevated temperature JIS R 1621: Testing method for bending fatigue of fine ceramics at room temperature JIS R 1632: Test methods for static bending fatigue of fine ceramics JIS R 1658: Testing method for bending fatigue of fine ceramics at elevated temperature JIS R 1668: Testing method for fracture toughness of porous fine ceramics JIS R 1677: Testing method for bending fatigue of porous fine ceramics at room temperature TS R 0002: Testing method for crack growth resistance (R-curve) of fine ceramics

3.3. Research Topic: Cyclic Fatigue of Toughened Ceramics Most ceramics subcritical crack growth rates under cyclic loads exceed corresponding growth rates under sustained, quasi-static loads by many orders of magnitude at equivalent stress intensity levels. This contrast clearly indicates that mechanical properties of ceramics are degraded by

644

cyclic fatigue. A number of tough ceramics show crack shielding effects which substantially influence cyclic fatigue behaviors. Mechanisms for cyclic fatigue crack growth in ceramic materials are generally based on suppression of crack-tip shielding in the crack wake under cyclic loading. Gilbert et al. [129,131] characterized cyclic fatigue crack propagation behavior of “in situ toughened” monolithic silicon nitride, in light of fatigue crack growth models for ceramics toughened by grain-bridging mechanisms. They used a model based on diminished crack-tip shielding in the crack wake under cyclic loads due to frictional wear degradation of the grain-bridging zone [120], as shown in Figure 9. The notion of cyclic crack growth promoted by diminished shielding was consistent with measured (long crack) growth rates, fractography, in situ crack profile analyses, and measurements of back face strain compliance. Growth rates depended on maximum applied stress intensity, K-max, much more than on the applied stress intensity range, DK. Fatigue thresholds exhibited a marked dependence on the load ratio, R ¼ K-min/K-max. These effects were inconsistent with traditional models of fatigue crack closure. In particular, when characterized in terms of K-max, growth rates below about 10 9 m/cycle exhibit an

Handbook of Advanced Ceramics

inverse dependence on load ratio. This observation was consistent with the grain-bridging phenomenon. Specifically, with increasing R, sliding distance between the grain bridges decreased, leading to less frictional wear and less degradation in shielding per loading cycle. The in situ toughening has been adapted also into silicon carbide recently. Similar crack growth resistance curves (R-curve) and cyclic fatigue behaviors have been found in such silicon carbide materials [132e137]. Gilbert et al. [132] investigated R-curve behavior in two silicon carbide (SiC) ceramics with sharply contrasting fracture properties, using small (3 mm) through-thickness cracks. The first, an in situ toughened material designated ABC-SiC, fails by intergranular fracture, whereas the second, a commercial SiC (Hexoloy SA), fails by transgranular cleavage. The former material, hot-pressed with aluminum, boron, and carbon additives, contained a network of plate-shaped grains [133]. An amorphous grain boundary film of ~1-nm thickness promoted debonding and crack deflection. Resultant grain bridging generated R-curve toughening. In contrast, no evidence of crack-tip shielding is observed in the latter SiC, Hexoloy SA. Although the Hexoloy SA fails catastrophically at 3 mm) and small (11 MPa m1/2 (determined by SEPB methods) as well as high fracture strength of >1.1 GPa (measured by the four-point flexure tests) were obtained at 5 vol.% dispersion of the seed crystals (see Figure 4 (C)). Another feature of this silicon nitride is a narrow strength distribution. The Weibull modulus of the distribution is >45, which is substantially high compared with those for conventional silicon nitrides, typically 10e25. Thus, the seeded and tape-cast silicon nitride shows synergistic improvement in all of the fracture strength, fracture toughness, and strength stability. Imamura et al. [22] used three different-sized seed crystals to investigate the effects of the seed size on the microstructure and mechanical properties of the seeded and tape-cast silicon nitrides, and found that small seed crystals (a mean diameter of ~0.5 mm and a mean length of ~2 mm) lead to highly anisotropic and fine microstructure with a monomodal grain-diameter distribution and improved

FIGURE 6 Microstructure of the silicon nitride fabricated with seeding (2 vol%) and tape-casting technique: (a) parallel and (b) perpendicular to the casting plane. The fibrous grains elongated from the seed crystals are somewhat randomly oriented, but always stay within the casting plane. (Courtesy of Kiyoshi Hirao et al.)

mechanical properties. The fracture strength determined by the four-point flexure tests was ~1.4 GPa and the fracture toughness determined by SEPB methods was 12 MPa.m1/2 in stress application parallel to the grain alignment (see Figure 4 (d)). Teshima et al. [23] employed extrusion process, in place of tape-casting, and succeeded in unidirectional alignment of fibrous seed crystals due to the high shear stress exerted during the extrusion process. Small matrix grains were almost consumed by selective grain growth into larger grains grown from seed particles, and consequently, large elongated grains with high aspect ratio were dominant in the microstructure of this material. Such microstructure development is related to the fact that the unidirectional alignment of seeds along a precise direction inhibits mutual impingement of elongated grains during growth along their length direction. Both high strength (four-point flexure tests) of ~1.4 GPa and high fracture toughness (SEPB methods) of ~14 MPa m1/2 could be

Chapter | 9.3

Microstructural Control and Mechanical Properties

achieved by testing in the direction parallel to grain alignment (see Figure 4 (e)). In the silicon nitrides with aligned fibrous grains, a large number of fibrous grains are involved with the crack-wake toughening mechanism when a crack propagates perpendicularly to the grain alignment, and then the toughening works more quickly in a short crack extension. Ohji et al. [24] investigated the fracture resistance curves (R-curves) of the seeded and tape-cast silicon nitrides, by measurements of initial (as indented) and fracture instability (indentation-strength measurement) crack lengths by Vickers indentation. The results are compared with that of a conventional self-reinforced silicon nitride as shown in Figure 7. The fracture resistance of the self-reinforced material increases continuously as a crack propagates and reaches 11 MPa m1/2 at about 500 mm, similarly to other measurements for self-reinforced silicon nitrides [25,26]. The seeded and tape-cast material, however, exhibited high fracture resistance above 10 MPa m1/2 from the beginning of the measured crack length range and remained almost constant in the following crack extension, when a crack propagated in the direction normal to the fiber axis (transverse crack propagation). Generally the fibrous grains, which span the crack wakes behind the crack tip, induce two crack-wake toughening mechanisms: elastic grain bridging and frictional grain pullout. The fibrous grain alignment gives at least a couple of benefits to these

661

toughening mechanisms: one is that a greater number of the fibrous grains are involved with the toughening and the other is that the toughening works effectively because the fibrous grains stand normal to the crack plane. If the fibrous grains are inclined to the crack plane, additional bending is applied to the grains and the grains are readily fractured during crack opening, reducing substantially the toughening effects [27]. These benefits give rise to the above unique R-curve behavior of the seeded and tape-cast material in the transverse crack propagation. It is generally known that the crack bridging (or elastic bridging) is responsible for short crack toughening while the pullout is for the long crack one. It is suggested that the former toughening occurs very effectively in the seeded and tapecast material due to the above benefits from the fibrous grain alignment. The rising R-curve behavior generally observed in the conventional self-reinforced material has little contribution on catastrophic fracture, since long crack extension is required to obtain a substantial increase of fracture resistance. On the contrary, the steep rise of fracture resistance in the seeded and tape-cast material should be related to an increase in the catastrophic failure strength. In other words, high strength and high fracture toughness are compatible as revealed by Hirao et al. [20]. Another benefit brought by this steep R-curve is that the strength distribution can be narrower resulting in a higher Weibull modulus. It is known that if the flaws involved in strength determination exhibit rising R-curve behavior, the Weibull modulus substantially increases [28e30]. This leads to the high Weibull modulus of >45 as above stated.

3.2. Fine Grain Control via Forging Techniques

FIGURE 7 R-curve behaviors of the seeded and tape-cast (closed symbol; a crack propagates perpendicularly to the grain alignment) and self-reinforced (open symbol) silicon nitrides, determined by as-indented crack lengths (circle) and indentation-strength measurements (square). (Reprinted from Ref. [24] with permission of John Wiley and Sons. All rights reserved)

Superplastic forging and superplastic sinter-forging techniques are approaches to align relatively fine fibrous grains in the process of deformation. In the former the deformation stress is externally applied to a sintered body, while in the latter it is done during sintering. Schematic illustrations of these techniques are shown in Figure 8. In either approach, the fibrous grains are aligned along the tensile direction, or the direction perpendicular to the compression. Superplastic deformation of silicon nitride occurs due to the grain boundary sliding when the grain size is small even if the grain shape is fibrous. In the case of fibrous grains, they are aligned along the tensile direction in the tensile deformation [31]. Using this phenomenon, Kondo et al. [32,33] realized superplastic plane-strain compressive deformation on silicon nitrides of fine and fibrous grains and investigated their strength and fracture toughness in comparison to those of the original material. Silicon nitride

662

Handbook of Advanced Ceramics

FIGURE 8 Schematic illustration of superplastic forging and superplastic sinter-forging techniques for fabricating fibrous-grain-aligned silicon nitrides.

prepared by gas-pressure sintering with additives of Y2O3 and Al2O3 was plane-strain deformed superplastically in a graphite channel die, with a pressure of 49 kN for 3 h at 1750  C. This deformation resulted in 50% height reduction. The X-ray diffraction analysis confirmed that both the original and deformed specimens consisted of b-silicon nitride grains without a trace of a-phase. The microstructure of the deformed specimen is shown in Figure 9 (a); the fibrous grains were aligned perpendicularly to the pressing direction. The fracture strength determined by three-point flexure tests for the original specimen was ~1.1 GPa, while it increased to ~1.7 GPa for the deformed specimen when the stress was applied along the extruding direction. The fracture toughness, measured by a single-edge-V-notchedbeam (SEVNB) method, also increased from 8.5 MPa m1/2 to 12 MPa m1/2. It has also been known that this silicon nitride showed remarkably high fracture energies, 200e630 J/m2, particularly at high temperatures [34]. In addition, the superplastically deformed silicon nitride showed substantially improved creep resistance at high temperatures, when the stress was applied along the extruding direction. The creep rates of the deformed specimen in tensile creep tests at 1200  C was found to be about one order of magnitude lower than that of the original one [35]. Superplastic sinter-forging is a process where sintering and forging (or deformation) are carried out concurrently [36]. Compacted powder mixture of a-Si3N4, 5 wt% Y2O3, and 3 wt% Al2O3 was sinter-forged with a pressure of 49 kN at 1750  C for 3 h in 0.1 MPa of nitrogen. The

FIGURE 9 Microstructure of the silicon nitrides fabricated through (a) superplastic forging and (b) superplastic sinter-forging techniques. The fibrous grains are aligned perpendicularly to the pressing direction (vertical). (Reprinted from Refs. [32] and [36] with permission of John Wiley and Sons. All rights reserved)

microscopic study revealed that the obtained samples have similar microstructures to the superplastically deformed ones, but the grain size is relatively small compared to them (see Figure 9 (b)). When the stress was applied perpendicular to the compressive direction, the fracture strength determined by three-point flexure tests was as high as 2.1 GPa with the SEVNB fracture toughness of 8.3 MPa m1/2. As already stated, the fracture toughness of ceramic materials generally varies inversely with the fracture strength. The fracture toughness and strength values for a variety of silicon nitrides are plotted in Figure 10. It can be known that the antagonistic relation exists in the silicon nitrides fabricated by the two techniques shown here, though both the properties are substantially improved compared to other silicon nitrides. The superplastically sinter-forged specimen shows higher fracture strength but

Chapter | 9.3

Microstructural Control and Mechanical Properties

FIGURE 10 Fracture strength and fracture toughness for various silicon nitrides. The closed circular (a) denotes the self-reinforced material [4,24,38]; the closed triangle (b) does the seeded and tape-cast [20,22]; the open triangle (c) does the seeded and extruded [23]; the open circular (d) does the nano-sized [37]; the open square (e) does the superplastically forged [32]; the closed square (f) does the superplastically sinter-forged [36].

lower fracture toughness than those of the superplastically deformed one [32], due to substantially small grain size. The former also exhibits higher strength than that with smaller grain size (~50 nm) [37], which is very likely attributable to the steep R-curve behavior and improved fracture toughness caused by the grain alignment, as well as to the reduced sizes of flaws during the deformation.

4. GRAIN BOUNDARY PHASE CONTROL 4.1. Fracture Resistance Significant improvements in the fracture toughness of silicon nitrides are realized also by tailoring the chemistry of the intergranular amorphous phase. The interface between the b-silicon nitride grains and the intergranular glassy phase should be debonded to some appropriate extent so that the fibrous b-silicon nitride grains can effectively contribute to the toughening mechanisms such as crack bridging and crack deflection. The interfacial debonding in silicon nitrides can be adjusted by the composition of sintering additives which ultimately altered the composition of the intergranular amorphous phase [21,39e41]. Chemical bonding across the interface determines the strength of the interface while the internal residual stress arising from thermal expansion mismatch imposed on the interface alters the debonding length. Sun et al. [21] described improvements in the fracture resistance of the seeded and tape-cast silicon nitride ceramics by

663

FIGURE 11 R-curve behaviors determined by applied moment double cantilever beam techniques for the seeded and tape-cast silicon nitrides with additives of 6.25 wt% Y2O3e1.0 wt% Al2O3, 5 wt% Y2O3e2.0 wt% Al2O3, and 4.0 wt% Y2O3e2.8 wt% Al2O3. (Reprinted from Ref. [21] with permission of John Wiley and Sons. All rights reserved)

controlling the chemistry of the intergranular glassy phase. They prepared the seeded and tape-cast silicon nitrides with different yttriumealuminum ratios and investigated their R-curve behavior in relation to the microstructural features. Although the different sintering additives generally result in microstructures with different grain morphologies and sizes, the seeding method is effective in regulating the grain morphology and size, and then is an idealistic approach for investigating the effect of sintering additives on the fracture behavior. The different yttriumealuminum ratio in the sintering additives leads to the different composition of the intergranular glass. The fracture resistance also varied as a function of the yttriumealuminum ratio in the sintering additives, as shown in Figure 11. The stress was applied in the direction parallel to the fibrous grain alignment. All the behaviors obtained for different compositions exhibited steeply rising R-curves; i.e., the fracture resistance of the materials increased rapidly with initial crack extension. It should be noted, however, that the long crack (or plateau) toughness values increased systematically with increasing the yttriumealuminum ratio. The in situ scanning electron microscopy (SEM) observation of cracks interacting with the microstructures also revealed that the interfacial debonding behavior between the large fibrous grains and the intergranular glass depended on the sintering additives employed. As shown in Figure 12, despite the similar angles of incidence, the fibrous grain in the sample of 4.0 wt% Y2O3e2.8 wt% Al2O3 failed transgranularly while

664

Handbook of Advanced Ceramics

ratio, being about 60, 70, and 75 degrees at the ratios of 4:2.8, 5:2, and 6.25:1, respectively. This indicates that, with the higher yttriumealuminum ratio, the interface between the fibrous grains and the surrounding glass has a lower interfacial debonding energy and is more readily debonded, leading to the stronger R-curve behavior and the higher plateau toughness. Becher and co-workers [42e46] further investigated the influence of intergranular glass on the interface between the prismatic faces of b-silicon nitride grains and the glass and exhibited that, for example, the interfacial strength increases with increasing the aluminum and oxygen contents of the epitaxial b-SiAlON layer that forms on the Si3N4 grains. This was to the formation of a network of strong bonds (cross bonds) that span the glassecrystalline interface. The in situ high-resolution electron microscopy observations revealed that the debonding path can occur at the interface between the grains and the intergranular glass film or within the film, depending on the film’s composition [46]. Furthermore, the recent studies [47e50] clarified more precisely rare-earth adsorption behaviors at intergranular interfaces in silicon nitride ceramics, and its effects on the phase transformation, microstructure evolution, and mechanical properties in silicon nitrides. For example, Shibata et al. [47,48] showed that the rare earths have different tendencies to segregate to the grain surfaces in silicon nitrides using atomic-resolution scanning transmission electron microscopy and first-principles calculations, which is essential for forming elongated grains and a toughened microstructure.

4.2. Heat Resistance

FIGURE 12 In situ observations of crack interacting with the large fibrous grains. In spite of the similar angles of incidence, the fibrous grain in the sample of 4.0 wt% Y2O3e2.8 wt% Al2O3 failed transgranularly while crack deflection and interfacial debonding occurred in that of 5.0 wt% Y2O3e2.0 wt% Al2O3. (Reprinted from Ref. [21] with permission of John Wiley and Sons. All rights reserved)

crack deflection and interfacial debonding occurred in that of 5.0 wt% Y2O3e2.0 wt% Al2O3. For the smaller angles of incidence, interfacial debonding occurred in both the samples; however, the latter sample showed longer debonding lengths. They also estimated the critical debonding angle, beyond which no crack deflection could occur at the interface, at the different yttriumealuminum ratio and found the angle to increase with increasing the

It is essentially important to control the chemistry of the intergranular phase in silicon nitrides, not only to improve the fracture toughness but also to give the supreme heat or creep resistance. The glassy phases, which are present at grain boundaries of many ceramics including silicon nitrides, are softened at high temperatures, leading to severe degradation of the mechanical reliabilities. Then, the highly refractory phase is required to the interface when ceramics are used for high-temperature applications. In recent decades, great deal of efforts to strengthen the grain boundary phase for improving heat resistance of silicon nitride have been devoted, leading to significant improvements of the high-temperature mechanical reliability. For example, some grades of silicon nitrides sintered with lutetia sintering additive showed excellent heat and creep resistances even at temperatures above 1400  C [51e54]. The X-ray diffraction (XRD) analyses for these materials revealed Lu2Si2O7 and Lu4Si2N 2O7 as secondary phases. Kondo et al. [55] fabricated highly heat resistant silicon nitride by using the sinter-forging technique with a lutetia

Chapter | 9.3

Microstructural Control and Mechanical Properties

sintering additive. The specimen exhibited the fracture strength of ~700 MPa at 1500  C, which is remarkably high compared with high-temperature strength reported so far for silicon nitride. Such superior high-temperature strength was attributed to grain alignment as well as to the refractory grain-boundary glassy phase and the existence of glass-free grain boundaries. Zeng et al. [56] tried to make highly refractory phase at grain boundaries in the seeded and tape-cast silicon nitrides, by using Lu2O3eSiO2 sintering additive systems. The slurry of a-Si3N4 powder, 3 wt% b Si3N4 seed crystals, 1 wt% SiO2, and 9 wt% Lu2O3 was tape-cast; sintering was carried out in 0.9 MPa N2 atmosphere at 1950  C for 6 h. Typical examples of the microstructure and grain boundary phase are shown in Figure 13. Because of the higher sintering temperature than the previous works [20e23], very large fibrous grains, whose length and diameter were typically several tens of mm and ~5 mm, respectively, were grown from fibrous b-Si3N4 seed crystals and tend to align along the casting direction. Again, it should be noted that many of them are not completely aligned but substantially inclined in the tape planes. Highly refractory Lu2Si2O7 phase was detected as secondary phase at many grain boundaries. Figure 14 shows the fracture strength determined by three-point flexure tests and fracture energies measured by chevron-notched beam (CNB) tests [57] at room temperature and 1500  C, in comparison with those of the unseeded material, which was obtained in the same procedures except for not adding b-Si3N4 seed crystal. The stress was applied in two directions parallel (strong direction) and perpendicular (weak direction) to the fibrous grain alignment. The strength measured in the strong direction at 1500  C was equivalent to the room-temperature one, while the unseeded material shows substantial strength degradation. The strength in the weak direction is low; however, it is comparable to that of the unseeded one at 1500  C. The excellent fracture strength at high temperatures is very likely attributable to the improved facture resistance. The fracture energy of the unseeded material was about 100 and 450 J/m2 while that of the seeded one in the strong direction was about 300 and 800 J/m2, at room temperature and 1500 C, respectively. The large fracture energies apparently were attributable to enhanced crack shielding effects of bridging and frictional pullout of the very large fibrous grains. It should be noted, however, that even in the weak direction the seeded material showed substantially high fracture energies particularly at the high temperature. This is presumably due to the large number of the substantially inclined fibrous grains, which give rise to grain bridging and frictional pull out effects, even in the weak direction. The inclined fibrous grains are also beneficial to toughening in the strong direction due to the so-called wedge effects [58], as shown in Figure 15. Compared to the completely

665

FIGURE 13 (a) Microstructure of the silicon nitride fabricated through dispersing 3 wt.% of b-seed crystals into the raw powder slurry of a-phase and Lu2O3eSiO2 sintering additives and tape-casting it (parallel to the casting plane). Note the large fibrous grains elongated from the seed crystals. (b) Grain boundary of this material. Lu2Si2O7 phase was detected as secondary phase. (Courtesy of Yu-Ping Zeng et al.)

aligned grains, further energy is required for the grains to be pulled out, owing to the additional bending applied to the grains. This mechanism, however, works effectively when the grains are strong enough, or the surrounding phases are weak enough, for the inclined grains not to be fractured during crack opening by such additional bending force. Thus, the large fracture energy can be obtained in the seeded materials which contain the large fibrous grains and at high temperatures where the grain boundary phases are softened.

666

Handbook of Advanced Ceramics

This section shows how the mechanical properties change when the size, shapes, and orientation of pores as well as grains are controlled, taking examples of porous silicon nitrides. Particular emphasis is placed in unique or improved properties including fracture energy, fracture toughness, strain tolerance (fracture strain), and thermal shock resistance through such porous structure control.

5.1. Isotropic Porous Structure

FIGURE 14 Fracture strength and fracture energies at room temperature and for the seeded and tape-cast silicon nitride with Lu2O3eSiO2 sintering additives and reference material (unseeded but otherwise fabricated in the same procedures). The stress was applied in two directions parallel (strong direction) and perpendicular (weak direction) to the fibrous grain alignment. (Courtesy of Yu-Ping Zeng et al.)

FIGURE 15 Schematic of alignment with tilted fibrous grains. Because of the wedge effects, tilted fibrous grains give rise to improved properties both in the strong (parallel to the alignment) and weak (perpendicular to the alignment) directions.

5. POROUS STRUCTURE CONTROL In structural materials, pores are generally believed to deteriorate the mechanical reliability; however, this is not always true whenever the microstructure is carefully controlled.

A variety of processes have been developed to produce porous silicon nitride ceramics. Sintering powder compacts to a fixed degree of densification, so-called partial sintering, is one of the most frequently employed approaches to fabricate porous ceramic materials. In the case of oxide materials, due to their good sinterability, the density or porosity can be adjusted by the heating characteristics, such as temperature and holding time. The sintering of silicon nitride ceramics is difficult, however, because of strong covalent bonding between silicon and nitrogen atoms, so that sintering additives are necessary for consolidating silicon nitride ceramics by liquid-phase sintering. On the other hand, the difficulty of sintering silicon nitride ceramics is beneficial for controlling density or porosity through adjusting the additives and the sintering process. A fibrous silicon nitride microstructure is developed during sintering at adequately high temperatures in porous material, and excellent mechanical properties can be obtained [59e61], compared with those of porous oxide ceramics obtained by partial sintering. Yang et al. [61] fabricated porous silicon nitride ceramics using Yb2O3 sintering additives and investigated the microstructures and mechanical properties. Sintering was performed at various temperatures between 1600 and 1850  C under a nitrogen gas pressure of 0.6 MPa. A liquidphase sintering technique and powder compaction similar to the fabrication of dense silicon nitride were used to make the fabrication process simple and cost-effective. The sintering additive, Yb2O3, is known to form crystalline Yb4Si2O7N2 at grain boundaries or triple junctions [62,63] which presumably enhances the high-temperature mechanical properties of silicon nitride ceramics. The addition of Yb2O3 also is known to accelerate the fibrous grain growth of b-Si3N4 [64], which is advantageous for the crack shielding effects of grain bridging and pullout. Furthermore, because of the high melting point of Yb2O3 and the high viscosity of Yb2O3related glass, this additive is unsuitable for the densification of silicon nitride, a characteristic that is conversely beneficial for the fabrication of porous silicon nitride. The microstructures of the porous silicon nitrides sintered at different temperatures are shown in Figure 16. The microstructure of the sample sintered at 1600  C consisted solely of fine, equiaxed grains. The average grain size was almost the same as that of the starting powder, indicating

Chapter | 9.3

Microstructural Control and Mechanical Properties

FIGURE 16 Microstructures and mechanical properties (flexural strength and fracture toughness) of porous silicon nitrides sintered with 5 wt.% Yb2O3 at 1600  C, 1700  C, 1800  C and 1850  C. “P” denotes porosity. (Reprinted from Ref. [61] with permission of Elsevier. All rights reserved)

little phase transformation, and the XRD analysis identified almost only a-Si3N4 peaks. When the sintering temperature increased above 1700  C, however, the formation and development of b-Si3N4 grains were observed, indicating enhanced phase transformation and grain growth. At 1700  C, very fine, fibrous b-Si3N4 grains were obtained, and with increasing the sintering temperature, the microstructure becomes coarse while the porosity changes relatively little. The fracture strength, sf, and fracture toughness (critical stress intensity factor), KIC, are shown also in Figure 16 as a function of sintering temperature. Each data point marks an average of five or six measurements. The fracture strength determined by three-point flexure and the fracture toughness was converted from the fracture energy measured by chevron-notched beam (CNB) tests using the following relationship [65] 2 ð1 geff ¼ KIC

n2 Þ=2E

(1)

where E is the Young’s modulus and n is the Poisson’s ratio. The obtained fracture toughness is an averaged value

667

FIGURE 17 Young’s modulus, E, fracture strength, sf, fracture energy, geff, fracture toughness, KIC, and fracture strain, or strain tolerance, ef as a function of porosity for (a) fine-grained and (b) coarse-grained porous silicon nitrides. The values are normalized by the respective values of the dense material: sf0 ¼ 1.1 GPa, E0 ¼ 330 GPa, KIC0 ¼ 6.3 MPa m1/2, and geff0 ¼ 70 J/m2 for the fine-grained material and sf0 ¼ 1.0 GPa, E0 ¼ 330 GPa, KIC0 ¼ 7.0 MPa m1/2, and geff0 ¼ 90 J/m2 for the coarse-grained one.

when the toughness varies with crack extension (R-curve). When the sintering temperature is 1600  C and the microstructure is equiaxed, the mechanical properties are low; however, the properties are markedly improved along with the fibrous grain formation above 1700  C. For example, the sample sintered at 1700  C has a strength of 380 MPa, which is about 10 times larger than that at 1600  C. As the sintering temperature further increased and the microstructure became coarser, the strength decreased while the fracture toughness increased. Nonetheless, the toughness values are still lower than 4 MPa m1/2. Mechanical properties of porous silicon nitrides have also been investigated as a function of porosities by using partial hot-pressing process, where the porosity is determined by the configuration of the carbon mold and the powder amount, leaving other parameters such as characteristics of a powder mixture and sintering additives fixed [59]. Two series of porous material with different porosities were prepared: one was fine-grained material series sintered at 1700  C and the other was coarse-grained at 1850  C. Figures 17 (a) and (b) showed porosity dependences of Young’s modulus, E, fracture strength, sf, fracture energy, geff, and fracture toughness, KIC, for the

668

fine- and coarse-grained material series, respectively. The values are normalized by the respective values of the dense material: sf 0 ¼ 1.1 GPa, E0 ¼ 330 GPa, KIC 0 ¼ 6.3 MPa m1/2, and geff0 ¼ 70 J/m2 for the fine-grained material and sf 0 ¼ 1.0 GPa, E0 ¼ 330 GPa, KIC 0 ¼ 7.0 MPa m1/2, and geff0 ¼ 90 J/m2 for the coarse-grained one. For the fine-grained material, the fracture energy first increases somewhat with increasing the porosity up to 10e20% and then decreases while all the other properties continuously decrease. The degree of decrease in the Young’s modulus is larger than that in the strength in the porosity range of 20e30%; therefore, the fracture strain, or strain tolerance, εf, given by εf ¼ sf /E is substantially high in this range. On the other hand, for the coarse-grained material, the fracture energy shows similar tendency, but is relatively high compared to that of the fine-grained material. While all the other properties continuously decreased, the degree of decrease in the Young’s modulus is smaller than that in the strength, resulting in the low fracture strain. The existence of pores lowers the strength, Young’s modulus, and fracture toughness, but may increase the facture energy and fracture strain depending on the porosity etc. in the isotropic porous silicon nitrides.

5.2. Anisotropic Porous Structure with Fibrous Grain Alignment Taking the seeding and tape-casting technique, Inagaki et al. [66,67] developed a porous silicon nitride, where the fibrous grains were uniaxially aligned (hereafter denoted by “anisotropic porous silicon nitride”) and investigated the effects of pores on the mechanical properties. Since the pores exist around the aligned fibrous grains, enhanced operations of the grain bridging and pullout can be anticipated when the crack propagates perpendicularly to the direction of grain alignment. Furthermore, the presence of pores surrounding the silicon nitride grains causes cracks to tilt or twist, namely crack deflections [68,69]. As starting materials they used b-Si3N4 seed crystals, which were obtained by heating the powder mixture of a-Si3N4, 5 wt% Y2O3, and 10wt% SiO2 in a silicon nitride crucible at 1850  C for 2 h [19]. The slurry containing the fibrous seed crystals mixed with 5 wt% Y2O3 and 2 wt % Al2O3 was tape-cast so that the seed crystals in the sheets were aligned along the casting direction. After the green sheets were stacked and bonded under pressure, sintering was performed at 1850  C under a nitrogen pressure of 1 MPa. The texture of the anisotropic porous silicon nitride with porosity of 14% was shown in Figure 18; the material consists only of fibrous grains, pores, and grain boundary phase. The fibrous grains of silicon nitride tend to be aligned along the casting direction, but many of the grains are substantially tilted. The pores, whose shapes are

Handbook of Advanced Ceramics

FIGURE 18 Microstructures of anisotropic porous silicon nitride prepared by tape-casting fibrous seed crystals (parallel to the casting plane). The porosity is 14%. (Reprinted from Ref. [66] with permission of John Wiley and Sons. All rights reserved)

mostly plate-like along the same direction, exist among the grains. The effect of porosity on the mechanical properties of the anisotropic porous silicon nitride is investigated in both directions parallel to and perpendicular to the grain alignment. Figure 19 shows the porosity dependencies of fracture strength and fracture energy when the stress is applied in the parallel direction to the grain alignment of the porous silicon nitride, in comparison with those of the isotropic coarse materials shown in Figure 17 (b). The strength was measured by three-point flexural tests while the fracture energy was determined by the CNB technique. The numbers of measurements for the former and latter are six and three, respectively, and the plots in the figure are their averages. The fracture strength became larger as the porosity decreased. In the porosity ranging below 5%, the

FIGURE 19 Porosity dependence of fracture strength and fracture energy of anisotropic porous silicon nitrides, in comparison with those of the isotropic materials shown in Figure 17 (b). Stress is applied parallel to the alignment direction. Bar is the standard deviation.

Chapter | 9.3

Microstructural Control and Mechanical Properties

strength attained above 1.5 GPa. This value was almost comparable to that of fibrous-grain-aligned dense silicon nitride fabricated through superplastic forging [32]. When fibrous grains were aligned, pores around the grains promoted debonding between interlocking fibrous grains without fracture. Therefore, even small amount of pores enhanced the grain bridging improving the strength. On the other hand, fracture energy of porous silicon nitride ranged from 300 to 500 J/m2 in the porosity range below 20%. These values were considerably high compared with other studies even for dense silicon nitrides [34,57]. Large fracture energy was mainly due to the effect of bridging crack by aligned fibrous grains and/or pullout of the grains. Debonding of grain boundaries by the existence of pores and aligned grains bridging the crack enhanced these crack shielding effects. Figure 20 shows the micrograph of the ligament area of fractured surface; the protruding grains and holes from which the fibrous grains have been pulled out clearly support these speculations. As porosity increased nearly 15%, fracture energy became larger; however, fracture energy decreased monotonously in the porosity range over 15% with increase of the porosity. This drop of fracture energy is presumably due to the reduction of substantial bridging and/or pullout area. The porosity dependences of the fracture toughness estimated from the fracture energy using Eqn 1 as well as the fracture strength are shown in Figure 21, in comparison with those of the isotropic coarse materials. High fracture toughness above 17 MPa m1/2 as well as high strength

669

FIGURE 21 Porosity dependence of fracture strength and fracture toughness, KIC, of anisotropic porous silicon nitrides, in comparison with those of the isotropic materials shown in Figure 17 (b). Stress is applied parallel to the alignment direction.

above 1.5 GPa were attained in the porosity range below 5%. Especially, fracture toughness of a specimen with a very small porosity less than 5% exceeded that of a fully dense specimen and became close to 20 MPa m1/2. Figure 22 shows the dependencies of fracture strength and fracture energy on porosity in the perpendicular direction to the grain alignment of the porous silicon nitride (weak direction), in comparison with those of the isotropic porous silicon nitrides. While the fracture strength is lower than that of the isotropic material, the fracture energy is substantially higher in the whole porosity range, due to the wedge effects caused by tilted large fibrous grains as shown in Figure 15.

5.3. Thermal Shock Resistances When ceramic materials are subjected to rapid change of temperature, substantial thermal stress arises, sometimes

FIGURE 20 Fracture surface of anisotropic porous silicon nitride (porosity: 14%). The protruding grains and the holes from which grains have been pulled out are arrowed. (Reprinted from Ref. [66] with permission of John Wiley and Sons. All rights reserved)

FIGURE 22 Porosity dependence of fracture strength and fracture energy of anisotropic porous silicon nitrides in comparison with those of the isotropic materials shown in Figure 17 (b). Stress is applied perpendicularly to the alignment direction. Bar is the standard deviation.

670

Handbook of Advanced Ceramics

resulting in cracking or complete failure. This failure is called as thermal shock fracture. Particularly porous ceramic components used at high-temperature such as hot gas filters are susceptible to this fracture. Thus, to designers and end users of ceramic components, it is essentially important to know properly the resistance properties against thermal shock for the material. This section intends to describe how thermal shock resistances vary when the porous structures change in porous silicon nitrides. Generally mechanical strength of ceramics after thermal shock remains constant until the temperature difference, DT, reaches a critical value, DTc, which is determined by the shape and sizes of specimen and heat transfer conditions, in addition to the material properties. When DT > DTc, a crack propagates catastrophically or quasi-statically depending on the initial crack size, leading to abrupt or gradual decrease of strength. Therefore, two parameters, “thermal shock fracture resistance”, R, and “thermal shock damage resistance”, R””, are often used to characterize the resistance against thermal shock crack initiation and propagation, respectively. As defined by Hasselman [70,71], R and R”” are given as follows: R ¼ sf ð1

vÞ=EaNDTc

R’’’’ ¼ Egeff =s2f ð1

vÞ

(2) (3)

where a is the coefficient of thermal expansion. It should be noted that the effects of fracture strength and Young’s modulus on R and R”” are completely opposite; when the fracture strength is high and the Young’s modulus is low, R is high but R”” is low. From the properties including the fracture strength, Young’s modulus, and fracture energy, it is possible to estimate R and R”” of the porous silicon nitrides as a function of porosity. Figure 23 (a) and (b) show the estimated R and R”” for the isotropic fine- and coarse-grained porous silicon nitrides which were sintered at 1700  C and 1850  C, respectively. For the fine-grained materials, the thermal shock fracture resistance, R, first increases with increasing the porosity up to 20e30%, and then decreases while the thermal shock damage resistance, R””, remains almost constant. For the coarse-grained materials, R decreases and R”” increases monotonously with increasing the porosity. It should be noted that completely different tendencies are obtained for both the R and R”” between the fine- and coarse-grained materials. In order to experimentally characterize the R and R””, thermal shock tests were performed by dropping the heated specimens from a resistance furnace into a container of water at 20  C [72]. The specimen was covered with mullite blocks on all sides except one face exposed to the water quenching. After heating in air at a rate of 5  C/min to a preset temperature and holding at this temperature for 30 min, the specimen was quenched into a water bath at 20  C. The fracture strengths of the quenched specimens

FIGURE 23 Estimated thermal shock fracture resistance, R, and thermal shock damage resistance, R””, of the isotropic (a) fine- and (b) coarsegrained porous silicon nitrides, as a function of porosity.

were determined at room temperature by three-point flexure tests. Figure 24 shows the results of the thermal shock tests for the isotropic fine- and coarse-grained porous silicon nitrides, whose porosity is 30% and 28%, respectively. The fine-grained material shows strength retention up to 1200  C and sharp drop above this temperature, indicating relatively good R and poor R””. On the contrary, in the coarse-grained material, the strength starts to gradually decrease around 600  C, indicative of relatively poor

FIGURE 24 Results of water-quenching thermal shock tests for the isotropic fine- and coarse-grained porous silicon nitrides.

Chapter | 9.3

Microstructural Control and Mechanical Properties

671

about 1200  C, indicating good agreement with the results of estimated R shown in Figure 25.

REFERENCES

FIGURE 25 Estimated thermal shock fracture resistance, R, and thermal shock damage resistance, R””, of the anisotropic porous silicon nitride as a function of porosity, in comparison with those of isotropic dense finegrained material.

R and good R””. This is consistent with the results of estimated R and R”” shown in Figure 23. It should be noted again that this notable difference in thermal shock fracture behavior arises solely from the different sintering temperatures of 1700  C and 1850  C. Figure 25 shows the estimated R and R”” of the anisotropic porous silicon nitride as a function of porosity, in comparison with those estimated for dense fine-grained silicon nitride whose property is shown in Figure 17 (a). The thermal shock fracture resistance, R, remains high in the porosity region up to 20% and then starts to decrease. On the other hand, the thermal shock damage resistance, R””, increases with increasing porosity up to 30% followed by the plateau region afterward. It should be noted that high values of both R and R”” are realized in the porosity range of 15~25%, which are much higher than those of the dense material. Figure 26 shows the results of the thermal shock tests for the anisotropic porous silicon nitride (14% porosity) in comparison with the above dense material. The anisotropic porous material shows the excellent R with DTc larger than 1200  C, while that of the dense material is

FIGURE 26 Results of water-quenching thermal shock tests for anisotropic porous (porosity: 14%) and isotropic dense fine-grained silicon nitrides.

[1] Kanzaki S, Brito ME, Valecillos MC, Hirao K, Toriyama M. Microstructure designing of silicon nitride. J Eur Ceram Soc 1997;17:1841e7. [2] Li CW, Yamanis J. Super-tough silicon nitride with R-curve behavior. Ceram Eng Sci Proc 1989;10:632e45. [3] Matsuhiro K, Takahashi T. The effect of grain size on the toughness of sintered Si3N4. Ceram Eng Sci Proc 1989;10:807e16. [4] Kawashima T, Okamoto H, Yamamoto H, Kitamura A. Grain-size dependence of the fracture-toughness of silicon-nitride ceramics. J Ceram Soc Japan 1991;99:320e3. [5] Mitomo M, Uenosono S. Microstructural development during gaspressure sintering of alpha-silicon nitride. J Am Ceram Soc 1992;75:103e8. [6] Becher PF. Microstructural design of toughened ceramics. J Am Ceram Soc 1991;74:255e69. [7] Steinbrech RW. Toughening mechanisms for ceramic materials. J Eur Ceram Soc 1992;12:131e42. [8] Lee DD, Kang SJL, Petzow G, Yoon DN. Effect of a to b (b0 ) phase transition on the sintering of silicon-nitride ceramics. J Am Ceram Soc 1990;73:767e9. [9] Kra¨mer M, Hoffmann MJ, Petzow G. Grain growth studies of silicon-nitride dispersed in an oxynitride glass. J Am Ceram Soc 1993;76:2778e84. [10] Han S-M, Kang SJL. Grain growth in Si3N4-based materials. MRS Bull 1995;20:33e7. [11] Dressler W, Kleebe HJ, Hoffmann MJ, Ru¨hle M, Petzow G. Model experiments concerning abnormal grain growth in silicon nitride. J Eur Ceram Soc 1996;16:3e14. [12] Tajima Y, Urashima K. Improvement of strength and toughness of silicon nitride ceramics. In: Hoffmann MJ, Petzow G, editors. Tailoring of mechanical properties of Si3N4 ceramics. Dordrecht, Netherlands: Kluwer Academic Publishers; 1994. p. 101e9. [13] Becher PF, Lin HT, Hwang SL, Hoffmann MJ, Chen IW. The influence of microstructure on the mechanical behavior of silicon nitride ceramics. In: Chen IW, Becher PF, Mitomo M, Petzow G, Yen TS, editors. Silicon nitride ceramics. Pittsburgh, PA: Materials Research Society; 1993. p. 147e58. [14] Hirosaki N, Akimune Y, Mitomo M. Effect of grain-growth of betasilicon nitride on strength, weibull modulus, and fracture-toughness. J Am Ceram Soc 1993;76:1892e4. [15] Wittmer DE, Doshi D, Paulson TE. Development of self-reinforced composites. Ceram Eng Sci Proc 1992;13:907e17. [16] Hirao K, Nagaoka T, Brito ME, Kanzaki S. Microstructure control of silicon-nitride by seeding with rodlike beta-silicon nitride particles. J Am Ceram Soc 1994;77:1857e62. [17] Emoto H, Mitomo M. Control and characterization of abnormally grown grains in silicon nitride ceramics. J Eur Ceram Soc 1997;17:797e804. [18] Hirao K, Tsuge A, Brito ME, Kanzaki S. Preparation of rod-like beta-Si3N4 single-crystal particles. J Ceram Soc Japan 1993;101:1071e3. [19] Inagaki Y, Ando M, Ohji T. Synthesis of porous silicon nitride by using seed crystals. J Ceram Soc Japan 2001;109:978e80.

672

[20] Hirao K, Ohashi M, Brito ME, Kanzaki S. Processing strategy for producing highly anisotropic silicon-nitride. J Am Ceram Soc 1995;78:1687e90. [21] Becher PF, Sun EY, Plucknett KP, Alexander KB, Hsueh CH, Lin HT, Waters SB, Westmoreland CG, Kang ES, Hirao K, Brito ME. Microstructural design of silicon nitride with improved fracture toughness: II, effects of yttria and alumina additives. J Am Ceram Soc 1998;81:2031e40. [22] Imamura H, Hirao K, Brito ME, Toriyama M, Kanzaki S. Further improvement in mechanical properties of highly anisotropic silicon nitride ceramics. J Am Ceram Soc 2000;83:495e500. [23] Teshima H, Hirao K, Toriyama M, Kanzaki S. Fabrication and mechanical properties of silicon nitride ceramics with unidirectionally oriented rodlike grains. J Ceram Soc Japan 1999;107:1216e20. [24] Ohji T, Hirao K, Kanzaki S. Fracture-resistance behavior of highly anisotropic silicon-nitride. J Am Ceram Soc 1995;78:3125e8. [25] Ramachandran N, Shetty DK. Rising crack-growth-resistance (Rcurve) behavior of toughened alumina and silicon-nitride. J Am Ceram Soc 1991;74:2634e41. [26] Li C-W, Lee D-J, Lui S-C. R-curve behavior and strength for insitu reinforced silicon nitrides with different microstructures. J Am Ceram Soc 1992;75:1777e85. [27] Becher PF, Hwang SL, Lin HT, Tiegs TN. Microstructural contribution to the fracture resistance of silicon nitride ceramics. In: Hoffmann MJ, Petzow G, editors. Tailoring of mechanical properties of Si3N4 ceramics. Dordrecht, Netherlands: Kluwer Academic Publishers; 1994. p. 87e100. [28] Kendall K, Alford N McN, Tan SR, Birchall JD. Influence of toughness on Weibull modulus of ceramic bending strength. J Mater Res (journal of materials research) 1986;1:120e3. [29] Cook RF, Clarke DR. Fracture stability, R-curves and strength variability. Acta Metall 1988;36:555e62. [30] Shetty DK, Wang J-S. Crack stability and strength distribution of ceramics that exhibit rising crack-growth-resistance (r-curve) behavior. J Am Ceram Soc 1989;72:1158e62. [31] Kondo N, Sato E, Wakai F. Geometrical microstructural development in superplastic silicon nitride with rod-shaped grains. J Am Ceram Soc 1998;81:3221e7. [32] Kondo N, Ohji T, Wakai F. Strengthening and toughening of silicon nitride by superplastic deformation. J Am Ceram Soc 1998;81:713e6. [33] Kondo N, Ohji T, Wakai F. Superplastic forging of silicon nitride ceramics with anisotropic microstructure control. J Mater Sci Lett 1998;17:45e7. [34] Kondo N, Inagaki Y, Suzuki Y, Ohji T. High-temperature fracture energy of superplastically forged silicon nitride. J Am Ceram Soc 2001;84:1791e6. [35] Kondo N, Suzuki Y, Brito ME, Ohji T. Improved creep resistance in anisotropic silicon nitride. J Mater Res 2001;16:2182e5. [36] Kondo N, Suzuki Y, Ohji T. Superplastic sinter-forging of silicon nitride with anisotropic microstructure formation. J Am Ceram Soc 1999;82:1067e9. [37] Yoshimura M, Nishioka T, Yamakawa A, Miyake M. Grain size. controlled high-strength silicon nitride ceramics. J Ceram Soc Japan 1995;103:407e8. [38] Kawashima T, Okamoto H, Yamamoto H, Kitamura A. Characteristic variety of silicon nitride made through microstructural control. In: Mitomo M, Somiya S, editors. Silicon nitride ceramics II. Tokyo, Japan: Uchida Rokakuho Publishing; 1990. p. 135e46.

Handbook of Advanced Ceramics

[39] Tajima Y. Development of high performance silicon nitride ceramics and their applications. Mater Res Soc Symp Proc 1993;287:189e96. [40] Urashima K, Tajima Y, Watanabe M. R-curve and fatigue behavior of gas-pressure sintered silicon-nitride. In: Bradt RC, editor. Fracture Mechanics of Ceramics, vol. 9. New York, USA: Plenum Press; 1992. p. 235e49. [41] Wo¨tting G, Ziegler G. Influence of powder properties and processing conditions on microstructure and mechanical properties of sintered Si3N4. Ceramic International 1984;10:18e22. [42] Becher PF, Sun EY, Hsueh CH, Alexander KB, Hwang SL, Waters SB, Westmoreland CG. Debonding of interfaces between beta-silicon nitride whiskers and SieAleY oxynitride glasses. Acta Mater 1996;44:3881e93. [43] Sun EY, Becher PF, Hsueh CH, Painter GS, Waters SB, Hwang SL, Hoffmann MJ. Debonding behavior between beta-Si3N4 whiskers and oxynitride glasses with or without an epitaxial beta-SiAlON interfacial layer. Acta Mater 1999;47:2777e85. [44] Becher PF, Painter GS, Sun EY, Hsueh CH, Lance MJ. The importance of amorphous intergranular films in self-reinforced Si3N4 ceramics. Acta Mater 2000;48:4493e9. [45] Painter GS, Becher PF, Sun EY. Bond energetics at intergranular interfaces in alumina-doped silicon nitride. J Am Ceram Soc 2002;85:65e7. [46] Becher PF, Painter GS, Lance MJ, Ii S, Ikuhara Y. Direct observations of debonding of reinforcing grains in silicon nitride ceramics sintered with yttria plus alumina additives. J Am Ceram Soc 2005;88:1222e6. [47] Shibata N, Pennycook SJ, Gosnell TR, Painter GS, Shelton WA, Becher PF. Observation of rare-earth segregation in silicon nitride ceramics at subnanometre dimensions. Nature 2004; 428:730e3. [48] Shibata N, Painter GS, Satet RL, Hoffmann MJ, Pennycook SJ, Becher PF. Rare-earth adsorption at intergranular interfaces in silicon nitride ceramics: subnanometer observations and theory. Physical Review B 2005;72:140101. [49] Becher PF, Painter GS, Shibata N, Waters SB, Lin HT. Effects of rare-earth (RE) intergranular adsorption on the phase transformation, microstructure evolution, and mechanical properties in silicon nitride with RE2O3 þ MgO additives: RE ¼ La, Gd, and Lu. J Am Ceram Soc 2008;91:2328e36. [50] Becher PF, Shibata N, Painter GS, Averill F, van Benthem K, Lin HT, Waters SB. Observations on the Influence of secondary Me oxide additives (Me ¼ Si, Al, Mg) on the microstructural evolution and mechanical behavior of silicon nitride ceramics containing RE2O3 (RE ¼ La, Gd, Lu). J Am Ceram Soc 2010;93:570e80. [51] Ohji T. Long-term tensile creep behavior of highly heat-resistant silicon nitride for ceramic gas turbines. Ceram Sci Eng Proc 2001;22:159e66. [52] Lofaj F, Wiederhorn SM, Long GG, Jemian PR, Ferber MK. Tensile creep in the next generation silicon nitride. Ceram Sci Eng Proc 2001;22:166e73. [53] Lofaj F, Wiederhorn SM, Long GG, Hockey BJ, Jemian PR, Browder L, Andreason J, Taffner U. Non-cavitation tensile creep in Lu-doped silicon nitride. J Eur Ceram Soc 2002;22(14e15):2479e87. [54] Kondo N, Asayama M, Suzuki Y, Ohji T. High-temperature strength of sinter-forged silicon nitride with lutetia additive. J Am Ceram Soc 2003;86:1430e2.

Chapter | 9.3

Microstructural Control and Mechanical Properties

[55] Nishimura T, Hirosaki N, Yamamoto Y, Takigawa Y, Cao JW. Tensile creep behavior in lutetia-doped silicon nitride ceramics. J Mater Res 2005;20:2213e7. [56] Zeng YP, Yang JF, Kondo N, Ohji T, Kita H, Kanzaki S. Fracture energies of tape-cast silicon nitride with beta-Si3N4 seed addition. J Am Ceram Soc 2005;88:1622e4. [57] Ohji T, Goto Y, Tsuge A. High-temperature toughness and tensilestrength of whisker-reinforced silicon-nitride. J Am Ceram Soc 1991;74:739e45. [58] Kovalev S, Miyajima T, Yamauchi Y, Sakai M. Numerical evaluation of toughening by crack-face grain interlocking in self-reinforced ceramics. J Am Ceram Soc 2000;83:817e24. [59] Yang JF, Zhang GJ, Ohji T. Porosity and microstructure control of porous ceramics by partial hot pressing. J Mater Res 2001;16:1916e8. [60] Yang JF, Zhang GJ, Ohji T. Fabrication of low-shrinkage, porous silicon nitride ceramics by addition of a small amount of carbon. J Am Ceram Soc 2001;84:1639e41. [61] Yang JF, Deng ZY, Ohji T. Fabrication and characterisation of porous silicon nitride ceramics using Yb2O3 as sintering additive. J Euro Ceram Soc 2003;23:371e8. [62] Wang CM, Pan X, Hoffmann MJ, Cannon RM, Ruhle M. Grain boundary films in rare-earth-glass-based silicon nitride. J Am Ceram Soc 1996;79:788e92.

[63] [64]

[65] [66]

[67]

[68] [69] [70]

[71] [72]

673

Nishimura T, Mitomo M. Phase-relationships in the system Si3N4eSiO2eYb2O3. J Mater Res 1995;10:240e2. Park HJ, Kim HE, Niihara K. Microstructural evolution and mechanical properties of Si3N4 with Yb2O3 as a sintering additive. J Am Ceram Soc 1997;80:750e6. Ohji T. Microstructural design and mechanical properties of porous silicon nitride ceramics. Mater Sci Eng A 2008;498:5e11. Inagaki Y, Ohji T, Kanzaki S, Shigegaki Y. Fracture energy of an aligned porous silicon nitride. J Am Ceram Soc 2000; 83:1807e9. Inagaki Y, Shigegaki Y, Ando M, Ohji T. Synthesis and evaluation of anisotropic porous silicon nitride. J Eur Ceram Soc 2004;24:197e200. Faber KT, Evans AG. Crack deflection processes .1. theory. Acta Metall 1983;31:565e76. Faber KT, Evans AG. Crack deflection processes .2. experiment. Acta Metall 1983;31:577e84. Hasselman DPH. Unified theory of thermal shock fracture initiation and crack propagation in brittle ceramics. J Am Ceram Soc 1969;52:600e4. Hasselman DPH. Strength behavior of polycrystalline alumina subjected to thermal shock. J Am Ceram Soc 1970;53:490e5. She J-H and Ohji T. Unpublished work.

Chapter 9.4

Determination of the Mechanical Reliability of Brittle Materials* Stephen W. Freimany and Jeffrey T. Fong** y

Freiman Consulting, 8212 Inverness, Hollow Terrace, Potomac, Maryland 20854, United States, ** Applied and Computational Mathematics Division, National Institute of Standards and Technology, Gaithersburg, Maryland 20899, United States

Chapter Outline 1. 2. 3. 4.

Introduction Lifetime Prediction Expressions Measurement Procedures Uncertainty Calculations

675 676 676 677

1. INTRODUCTION Were it not for the existence of subcritical crack growth, designing ceramic components to withstand known stresses would be relatively straightforward, requiring only a statistical estimate of the strength of the material. However, most ceramics are subject to slow crack growth under stress in moisture-containing environments, a phenomenon that can lead to mechanical failure after a period of time at stresses lower than the short-term fracture strength [1]. Examples of components that are vulnerable to delayed failure include optical fibers (SiO2), multilayer capacitors (BaTiO3), infrared transmitting windows (ZnS, ZnSe), and piezoelectric transducers, to name but a few. The stresses that could lead to failure can arise, not just due to mechanical loads, but also as a result of thermal gradients, phase transformations, or the presence of electric fields. What is needed is a method to assure that the most severe flaw in a component is not subjected to a stress such that it will grow to a critical size before the desired lifetime of the part. Three methods for assuring such reliability are available. The most direct is nondestructive evaluation of a part prior to it being placed into service, which in principle would allow the identification of a flaw that could lead to failure at short times. Unfortunately, * Contribution of the National Institute of Standards and Technology. Not subject to copyright

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00036-8 Copyright Ó 2013 Elsevier Inc. All rights reserved.

5. Summary Acknowledgments References

679 679 679

because of the small sizes of the flaws in most ceramics, no procedures with the necessary resolution to distinguish one critical flaw from the many others in the distribution are available today. A second alternative is proof testing in which the part is loaded to a stress exceeding that expected in service, and then rapidly unloaded. While this procedure can be effective if carried out correctly, it is expensive to conduct on every part, and a significant loss of parts during the proof test should be expected. In addition, the stresses must be applied in a dry environment, exactly as would be expected in service, and rapid unloading from the proof stress is essential [2]. We are therefore left with the necessity of using fracture mechanics expressions relating crack growth rates to material parameters and known loads, coupled with a statistical analysis of the experimental data as a means of determining reliability. An essential point is that each of the parameters needed for the reliability assessment has a measurement uncertainty associated with it. It is crucial to determine not only the predicted lifetime but also to know to what degree of confidence it can be specified. Previous studies have addressed this issue in various ways [3,4]. This chapter provides a summary of suggested testing methodology, analysis techniques, and statistics needed to assure safe operation of brittle components over specific times, including a discussion of the methods by which the results of laboratory-scale specimens can be used to predict the reliability of a system of parts.

675

676

Handbook of Advanced Ceramics

2. LIFETIME PREDICTION EXPRESSIONS Our ability to predict the reliability or lifetime of a ceramic part rests on the use of a mathematical expression to represent the crack growth data. Following chemical rate theory, which has been shown to be applicable to the environmentally enhanced crack growth process [1], the following exponential relationship between crack velocity, V, and the stress intensity factor, KI, is derived: V ¼ V0 expðbKI Þ:

(1)

V0 is a material-dependent constant containing terms such as the relative humidity of water in the particular environment and the temperature, and b is a constant proportional to the activation volume for the process [1]. In principle, Eqn (1) can be integrated to yield a time for a flaw to grow from its initial size to criticality. However, there are two problems: First, a closed-form solution for the integration is difficult to achieve. While such integration has been reported [5], the integration limits are in terms of flaw size, which from a practical point of view cannot be determined with any degree of accuracy. Second, the effect of a residual stress term associated with a flaw has not been incorporated into such an expression [6]. An empirical power law is used instead:
N KI : (2) V ¼ V0 KIC The exponent N represents the slope of the crack growth curve. In water, N ranges from as low as 15 for oxide glasses to 50 for ceramics such as silicon nitride and silicon carbide. KIC is the critical stress intensity factor, used here as a scaling parameter. However, most flaws in actual components will experience not only a far-field stress but are also surrounded by a stress field generated by the impact-type process by which they were introduced, for example, machining and finishing. Eqn (2) is then modified to the following [7]:
N 0 KI ; (3) V ¼ V0 KIC where N0 ¼ (3N þ 2)/4 for “point flaws,” that is, elliptical cracks typically created by machining or other impact processes; N0 is smaller than N and so provides a more conservative estimate of flaw behavior in a part. It is useful to put KI in terms of the stress, s, in the component through 1

KI ¼ Ysa2 ;

(4)

where a is the flaw depth in the surface, and Y is a geometric constant dependent on loading, flaw geometry, and location. Here, s is the stress in the specimen or part.

Fuller et al. [7] showed that a time to failure, tf, due to cracks emanating from “point flaws”:  2  0 0 0 4N 3 2KIC SN s N ; (5) tf ¼ N 0 2 Y 2 V0 3 where S is the strength of the as-finished material. The other terms have been previously defined. However, because of the need to separately measure KIC and V0 using large cracks in fracture mechanics specimens, this form of the expression is unwieldy to apply. Not only might these parameters be difficult to obtain in this way, for example, because of amounts of material required but the results also could differ from those applicable to the small flaws of interest. Fuller et al. [7] also showed that the term in the parentheses in Eqn (5) can be expressed as   0 2 2KIC 0 3N 2 l Sv2 N ; ¼ (6) 0 3 2 0 N Y V0 ðN þ 1Þ4 where l is obtained from a plot of stressing rate versus strength measurements made on indented specimens, that is, dynamic fatigue tests, and Sv, which we call the reference strength, is the strength of a set of indented specimens in which the as-created cracks are not allowed to grow subcritically. The right-hand side of Eqn 6 contains only parameters that can be obtained from strength measurements. We now combine Eqn (5) with Eqn (6):  N 0 2 0 l S s N: (7) tf ¼ 0 1 Sv N Both S and Sv will be taken as the minimum values of strength within the measured distributions.

3. MEASUREMENT PROCEDURES 3.1. Initial Strength Distribution The initial strength, S, is based on the flaw distribution in the material just before a part is put into service. In carrying out a reliability analysis, a key assumption is that no flaws more severe than those in the initial flaw distribution are introduced during the lifetime of the part. In addition, it is assumed that flaws grow only due to environmentally enhanced crack growth. It would be advantageous if the strength distribution of actual parts could be obtained under loading conditions that simulate the service loading conditions. However, this can be difficult and costly. Consequently, tests are usually conducted on small pieces of the same material. A key assumption is that the processing procedures and surface treatments, for example, machining and polishing of these specimens, are identical to that seen by the component. Because of the propensity of brittle materials to fail from surface flaws, flexural tests are the primary test

Chapter | 9.4

Determination of the Mechanical Reliability of Brittle Materials

method. Such tests can be conducted in either uniaxial or biaxial loading. Because failure from sharp edges is typically a problem with brittle materials, the decision of whether to conduct uniaxial or biaxial tests is usually based on whether edges in the part will be subject to significant stresses. If uniaxial testing, for example, four-point flexure, is chosen, edges should be rounded or chamfered to minimize failure from them.Testing must be conducted at a high loading rate in an inert environment, for example, dry gaseous nitrogen, in order to eliminate slow crack growth effects [8]. In engineering practice, about thirty specimens are recommended to assure that an adequate statistical distribution can be established. However, this number is rooted in the statistics literature to be compatible with the rigorous estimate of confidence levels [9]. The actual number of specimens that are needed will depend on the scatter in the data and the standard deviation (sd) desired [10]. At times, a component may experience a stress state that could cause it to fail from internal flaws such as pores or inclusions. In this case, flexural tests will not be effective, and direct tensile tests will be required. In the simplest case, all the flaws leading to failure are of the same type and come from one population, that is, one source. In many instances, however, there are multiple flaw populations involved leading to more complex strength distributions. Fractographic analysis of the broken specimens, particularly those in the low-strength region, is important to determine the actual cause of failure and to separate unusually low strengths from those in the general population [11].

677

wide a range of stressing rates as possible, but not less than three orders of magnitude, in order to obtain the required accuracy. ASTM Standard C-1368-06 [13] provides more details of the testing and analysis procedures for this test. The Standard also describes conditions by which specimens can be preloaded to positive tensile stresses before the stressing rate test itself to reduce the time needed to carry out tests at the slowest rates. A crucial assumption in obtaining the data in this way is that the strength versus stressing rate curve can be extrapolated to lower stressing rates over a number of orders of magnitude, implying that the mechanism of crack extension does not vary at low stressing rates. We know that for certain materials, there is a downturn in crack growth rates at low stress intensity factors [1]; this would then lead to a conservative prediction of failure time. In a very few cases, however, crack growth deviates to a plateau at small stress intensity factors [1]. The procedure outlined here should not be used for these materials.

3.3. Reference Strength Because indented specimens were used to determine N0 , measurements are needed to calculate Sv, the upper limit to the indented strength. These strength measurements must also be made at a high loading rate in an inert environment.

3.4. Stresses in the Component The stress distribution in the component is typically obtained through a finite element analysis.

3.2. Crack Growth Parameters While values of the crack growth parameters, N and l, can be obtained from direct measurements of crack growth rates, dynamic fatigue tests employing cracks that are of comparable size to actual flaws in the component are both more convenient, as well as more relevant to the practical failure issue. As noted earlier, indentations surrounded by a stress distribution lead to a more conservative value of the environmental exponent, that is, N0 rather than N. In the dynamic fatigue test, indented flexural bars are loaded to failure at a constant stressing rate. The relevant expression from which the crack growth parameters are extracted is [12]: ds : (8) dt A linear regression of the log of the failure strength, sf, on _ yields the parameters N0 the log of the stressing rate, s, and l. The stressing rate tests should be carried out at the harshest temperature and environment that will be experienced by the part. It is necessary to take the data over as 0

s Nf þ1 ¼ l

4. UNCERTAINTY CALCULATIONS It is not sufficient to simply calculate a time to failure; one must also know the uncertainty in this calculation, particularly its lower bound. Each of the measured parameters described in the previous section will contain some degree of uncertainty. These uncertainties, which can be calculated from standard statistical models [11], must be combined to yield an uncertainty in the predicted time to failure. Here, we describe a three-step strategy to estimate the uncertainty of the failure time in the full-scale structure. The first and most important step is to determine the uncertainty in the minimum strength, S, of the specimens. This is the most critical parameter in the data set because it is a measure of the most severe flaw. Second, by combining uncertainties, we can determine the uncertainty in the minimum lifetime of the specimen set as determined from Eqn (7). Third, we then determine the uncertainty in the lifetime of the part itself.

678

Handbook of Advanced Ceramics

To determine the uncertainty in the minimum S, one fits the measured strength distribution to a particular mathematical expression. The Weibull function, which is typically used in fracture analysis, is only one of many possible expressions that could be employed but has been used extensively because it seems to represent the underlying physics governing brittle fracture, namely, weakest link theory [14,15]. Historically, a two-parameter Weibull distribution, which allows for failure at zero applied stress, has been used to fit strength data and is the basis for the ASTM standard [16]:
Z  m  s dV : (9) P ¼ 1 exp s0 In this expression, P is the probability of failure, s is the applied stress, and s0 the Weibull scale parameter. However, this approach is unduly conservative; flaws in components that would fail at extreme stresses could be visually observed, or possibly the component would break before being placed into service. Other, more complex, probability distributions have been put forward as well [17,18]. It is suggested that the three-parameter Weibull distribution be used. The expression is given by:  
Z  s su m dV : (10) P ¼ 1 exp s0 Here, su is the minimum in the strength distribution, or the location parameter in the statistical literature. To some degree, assuming that the Weibull model is the only possible choice impedes our ability to fit the experimental data with a minimum of uncertainty. It is advisable to use goodness-offit tests to establish the best statistical fit to the data. One can then choose the acceptable level of confidence desired for the lower bound in this calculation [19e21].

4.2. Step 2 From Eqn (7), we can see that the time to failure, tf, is a function of five parameters, namely, l, N 0 , S, Sv , and s. Provided that the coefficient of variation of each of the parameters is 0.55W, where W is the full width of the specimen, and (L a) > .65W. As noted by Shyam and Lara-Curzio [13], this means that for L/S ¼ 3, KI is constant over the middle 60% of the test specimen. Too short a starting crack leads to an overestimate of the toughness, while too long a crack underestimates it, which has been confirmed experimentally (Figure 5) [16]. As with a number of the other fracture mechanics specimen configurations, the DT specimen must be precracked in order to ensure that an accurate value of fracture toughness is obtained. Some of the same procedures applicable to the DCB test can be used here as well. Another method of starting a crack is to taper a notch at the frontend of the specimen. If the notched specimen is loaded slowly, the pop-in of a crack can be observed through a drop in the load. A second method is to introduce a series of Knoop indentations along the starting notch. The presence of the cracks emanating from the indentation sites will help produce a sharp crack during preloading. Also,

FIGURE 5 KIC determined at various crack length to specimen length ratios for a soda-lime glass specimen. (After Pletka et al. [16]).

Chapter | 9.6

Fracture Mechanics Measurements

723

a Knoop indenter can be used to form a straight scratch on the center of one end of the specimen from which a crack will emanate.

4.1. Fracture Toughness Determination Critical fracture toughness can be determined using the DT specimen by loading a precracked specimen rapidly until the crack grows the length of the specimen. KIC is calculated from Eqn (18) using the largest recorded load, P. The same issues of loading rate and moisture as discussed with respect to the DCB test apply here as well. While this method yields accurate values of fracture toughness, the quantity of material required for each specimen means that other techniques to determine KIC may be better choices. The exception would be if the material is produced in large sheets or can be cast to shape such as with concrete.

to a predetermined load, Pi, at a relatively high rate of loading so that no crack growth occurs prior to reaching the fixed load. The crosshead is stopped, and the relaxation of the load is followed. The data obtained by Evans on polycrystalline alumina in order to verify this procedure are shown in Figure 6. Caution should be exercised in carrying out this technique to very low crack velocities, i.e. below 10 7 m/s, because of possible interferences due to compliance variations in the test machine, temperature variations, or electrical perturbations. Evans [20] also suggested a second method of acquiring crack growth data on the DT specimen. He observed that if one holds the rate of displacement, dy/dt, constant then dy ¼ BPV dt

Since KI ¼ AP, A is a constant involving specimen geometry and elastic properties: dy ¼ B0 KI V dt

4.2. Crack Growth Measurements

(23)

(24)

Although it is possible to collect crack growth data in a DT specimen by placing it under constant load while monitoring crack length optically [19], two other procedures, a load relaxation and a constant displacement rate testing method, have been shown to be effective in determining crack growth rates without a need to view the crack directly [20]. Because the driving force on the crack is independent of crack length, the change in specimen compliance can be used to determine crack extension. The deflection of the specimen at the loading points, y, is given by y ¼

2 3PaWm 3 Gt W

(19)

The change in crack length, Da, can therefore be calculated from a corresponding Dy through Da ¼

Gt3 W Dy 2 3pWm

(20)

Evans [20] noted that, as shown above, the displacement of the loading points, y, is directly related to the applied load, P, and the crack length, a, through y ¼ PðBa þ CÞ

(21)

where B and C are constants related to the elastic properties of the material and the dimensions of the specimen and the test device. For large initial crack lengths, ai, one can show that
 ai Pi dP V ¼ (22) P2 dt Tests are conducted in a machine capable of constant crosshead displacement by loading a precracked specimen

FIGURE 6 Crack growth data for a polycrystalline alumina using the load-relaxation procedure. (Evans [20]).

724

Handbook of Advanced Ceramics

FIGURE 8 SENB specimen.

FIGURE 7 Schematic showing the formation of a load plateau at constant displacement rates.

A plateau in load, P, is evidence that the crack is growing in a stable manner (Figure 7). The constant loading rate technique is particularly useful in obtaining data under severe environmental conditions in which either viewing the crack directly or the stability of the loading system is questionable. One drawback in the use of the constant displacement rate method is that the range of crack velocities that can be obtained is limited to those >~10 5 m/s. The DT technique is an excellent choice for obtaining crack growth data, particularly under conditions in which observing a crack optically is difficult. It is less likely to be the choice for a critical fracture toughness determination for most brittle materials because of the amount of material required. However, one of the significant users of the DT technique has been the geological community, which has made use of the ability to scale the specimen to large sizes to investigate mineral and rock fracture.

5. TESTS BASED ON FLEXURAL LOADING A number of fracture mechanics test geometries and procedures are based on a rectangular bar in which a notch or crack has been introduced, and which is subsequently loaded in either three-point or four-point flexure. These tests include the single-edge-notch-bend (SENB) test, the work of fracture (WOF) test, the single-edge-V-notch-bend (SEVNB) test, the surface crack in flexure (SCF) test, and the single-edge-precracked-beam (SEPB) method. Except for the WOF test, these procedures differ essentially through the method by which a starting crack is introduced into the flexural bar. Only some of these are described in this chapter. The reader is referred to Ref. [2] for descriptions of the others. All of the tests in this section require that a specimen be broken in flexure. Reproducible, accurate results of

FIGURE 9 CNBT specimen. (Munz et al. [26]).

a flexural test depend on the specimen having the correct dimensions and machining, and it being loaded correctly. The particulars of how to conduct a flexural test for ceramics, including detailed descriptions of loading fixtures, are contained in both an ASTM [21] and an ISO standard [22]. In addition, an excellent review of flexural strength procedures was published by Quinn and Morrell [23].

5.1. Single-Edge-Notch-Bend Test The first fracture mechanics test for brittle materials involving a flexural specimen was the SENB test (also called the notch-beam test) suggested by Davidge and Tappin [24]. As its name suggests, it consists of a bar with a notch cut into one of the narrow sides (Figure 8). While there are extensive data in the literature on results of this test, because of the absence of a true crack, SENB data should not be trusted. It is typically assumed that in the case of brittle materials, a machining-induced crack exists at the root of the notch. This assumption is frequently incorrect, and leads to overestimates of the fracture toughness.

5.2. Chevron-Notch-Bend Tests Munz et al. [25] suggested an improved version of the SENB test, which makes use of a flexural bar containing a symmetrical chevron notch, and is termed the chevron notch bend test (CNBT) (Figure 9). The specimen can be loaded in either three- or four-point bending in a universal test machine. Critical fracture toughness is calculated from the maximum load reached before failure using knowledge of the specimen compliance, thereby avoiding the need to

Chapter | 9.6

Fracture Mechanics Measurements

725

that can act as cracks. This test has become known as the SEVNB test. Ku¨bler [28] expanded the discussion of the SEVNB test providing valuable details on the best procedures for use of the razor blade to sharpen the notch. He reported that notch widths as narrow as two micrometers could be created in a hot-pressed silicon nitride. The ISO standard [29] for the chevron notch test (ISO 24370) is relatively similar to ASTM Standard C-1421. However, it calls for only two specimen geometries, and does not provide for the possibility of three-point bending. FIGURE 10

Typical loading curves seen in chevron notch testing.

5.3. Single-Edge-Precracked-Beam determine an initial crack size. This specimen is one of three methods that has been adopted as both an ASTM Standard [6] and an ISO Standard [26]. Equation (25) can be used to calculate fracture toughness, designated as KIvb. The subscript reflects ASTM’s position that, because of factors such as a different response to environmentally enhanced crack growth, each of these tests may lead to a different value of fracture toughness: 

Pmax ½So Si Š10 6  (25) KIvb ¼ Ymin BW 3=2 where SO and Si are the outer and inner loading points, respectively, W is the specimen height, and B is the spec is defined as the minimum stress intensity imen width. Ymin factor coefficient which depends on the specific specimen geometry. In order for a test to be valid, the forceedisplacement curve must exhibit a smooth maximum as seen in Figure 10a. Curves such as shown in Figure 10b are indicative of a valid test as well, but care must be taken to use the Pmax point on the diagram in the calculation. The observation of a sharp drop (Figure 10c) warns of an overestimate of the fracture toughness. The narrower is the notch the more likely it is that a stable crack growth will ensue, and the less will be the scatter in data. It has been observed that some skill is needed in order to properly machine such a specimen. Valid tests are more likely if cracks can be initiated at the chevron tip prior to loading in the test itself. The standard suggests that one way of accomplishing this is to turn the specimen upside down and load it in compression over a few loading cycles. As a way of better ensuring that a crack grows smoothly from the tip of the chevron, Nishida et al. [27] introduced the idea of sharpening the notch in an SENB specimen through the use of a razor blade impregnated with 1-micrometer diamond paste. The razor blade is drawn back and forth through the notch to produce V-shaped notches

The SEPB test uses a specially designed system to introduce a controlled crack into the surface of a bend specimen. It was put forward by Nose and Fujii [30] based on the so-called “bridge-indentation” procedure [31] for stably extending a crack either from a hardness indentation or from a machined notch. This method now exists as both an ASTM [6] and an ISO standard [32]. The crack starter can be a machined notch, one or more Vickers indentations, or a Knoop indentation with its long axis perpendicular to the longitudinal axis of the specimen. The precrack is then loaded in compression in a specially designed fixture. A sound is created at popin, but it may be necessary to use a stethoscope to detect it. The use of an ultra-penetrating fluorescent dye penetrant helps to determine the final crack length before testing. As with other flexural tests, it is necessary to ensure that stable crack growth occurred, as evidenced from the shape of the forceedisplacement curve (Figure 10) in order to obtain an accurate value of KIC. ASTM Standard C-1421 [6] gives expressions which can be used to calculate KIpb. Because the method and the fixture are relatively complex, this technique is not used extensively.

5.4. Surface Crack in Flexure The SCF technique was first suggested by Petrovic and Mendiratta [33]. It has since become incorporated as an established procedure in ASTM Standard C-1421 [6] as well as into an ISO Standard [34]. The test involves making a Knoop hardness indentation in the surface of a flexural bar, thereby forming a crack underneath the indentation site, removing the residual stressed zone surrounding the indent, and fracturing the bar in four-point bending. A schematic of the indented bar is given in Figure 11. The standard points out that the indentation can be made in either long face of the bar.

726

Handbook of Advanced Ceramics

of fracture toughness. The potential effect of such a phenomenon can be ascertained by testing at two different loading rates, or can be avoided by testing in an inert environment.

6. INDENTATION METHODS FIGURE 11 SCF specimen.

In ASTM Standard C-1421 [6] guidelines for the indentation force are given which depend on the estimated toughness of the material being tested. It is suggested that forces of 10e20 N are suitable for low toughness materials, e.g. glasses, 25e50 N for intermediate toughness materials, e.g. alumina, and 50e100 N for higher toughness materials, e.g. silicon nitride. In order to obtain an accurate value of fracture toughness by this technique, it is important that all of the stressed material is removed during polishing. The Standard recommends that material be removed to a depth of 4.5e5 times the depth of the indentation itself. Quinn and Salem [35] recommend that if lateral cracks, which can interfere with the test accuracy, are observed or suspected, then 7e10 times the indentation depth should be removed. After fracturing in four-point bending, the initial crack size is measured on the fracture surface. The ease and accuracy with which the dimensions of the crack can be determined depend on the microstructure of the material and, to some degree, on the experience of the individual in fractographic analysis. The Standard makes some suggestions regarding this part of the procedure, but the reader is referred to the National Institute of Standards and Technology (NIST) Recommended Practice Guide SP 960e16 “Fractography of Ceramics and Glasses” [36] for more details. The fracture toughness is calculated from the following expression:

 3Pmax ðSo Si Þ pffiffiffi a (26) KIC ¼ Y 2BW 2

where a is the crack depth, So and Si are the outer and inner spans of the 4-point flexure system, Pmax is the maximum load recorded during the test, B and W are the width and depth of the specimen, and Y is the stress intensity coefficient that accounts for the flaw shape. Y varies around the perimeter of the crack. Typically, Y is calculated at the surface and the maximum depth of the crack, and the largest value of Y is used in the calculation. As with the other flexural-type fracture toughness tests, consideration must be given to the possibility of environmentally enhanced crack growth prior to reaching criticality, which would yield an underestimate

By indentation methods is meant those procedures that involve using a hardness indentation coupled with the ensuing crack system and the stress field introduced by the indentation as the primary source of the starting crack in a fracture mechanics test. There are two basic indentation test methods. In the first, known as indentation fracture (IF) [37], a Vickers indenter is used to create a hardness impression. The length of the cracks emanating from the corners of the indentation is measured and combined with knowledge of the elastic properties and hardness to calculate fracture toughness. In the second, called indentation strength (IS) [38], the indentation is placed in a bar of the material, and the strength measured in flexure. This test differs from the surface crack-in-flexure method discussed previously in that the residual stress system associated with the indentation itself is not removed and is considered as part of the driving force on the crack.

6.1. Indentation Fracture In the IF technique, a Vickers hardness indentation is placed in the surface of the material producing cracks emanating from the corners of the hardness impression (Figure 12); the indenter is then removed, and the length of the cracks measured. One of the serious complications and ultimate shortcomings of this test is the empirical nature of the expressions used to relate “KIC” to the measurable parameters, load, hardness, elastic modulus, and crack length. Numerous expressions relating fracture toughness to indentation parameters have been suggested to calculate fracture toughness by this procedure [39e41]. One of the more popular versions of such an expression is given in Eqn (27) [37]: !
1 E 2 P (27) KC ¼ k 3=2 H C 0

where E is Young’s modulus, c0 is the measured crack length, P is the indentation load, k is claimed to be a material-independent constant, and H is the hardness determined from the expression: H ¼

P a0 a2

(28)

Chapter | 9.6

Fracture Mechanics Measurements

FIGURE 12 Crack pattern after indentation. (After Anstis et al. [37]).

The term KC rather than KIC is used to emphasize that the procedure should not be considered to lead to “standard” fracture toughness. In this expression, a is the size of the indentation impression and a0 is a numerical constant. Anstis et al. obtained a value of k ¼ 0.016  0.004 by averaging fracture toughness data obtained by conventional techniques over a range of materials. There are three significant problems associated with this test procedure [39e41]: l

l

l

The method cannot yield a value of fracture toughness that is consistent with the definition of KIC given previously, namely the stress intensity at which unstable crack growth occurs. Rather than accelerating, the crack decelerates and stops. There is no basis for assuming that this crack arrest value corresponds to a fracture toughness value obtained by accepted methods in which a single crack is grown to failure. The crack system itself can be extremely complex depending on the microstructure of the material. More than one crack can be present, and the presence of lateral cracks, which extend more or less parallel to the surface below the indentation, can further interfere with the growth of the radial cracks. None of the expressions used to calculate KC are fundamental in nature.

The question arises of whether this test could be used to rank materials without depending on the numerical

727

accuracy of KC. This too is questionable. Ponton and Rawlings [39,40] compared materials using a number of the expressions proposed to calculate fracture toughness, and concluded that an accurate ranking of materials with respect to fracture toughness depended on the set of materials chosen and the expression used to calculate KC. Nonetheless, there are situations in which there are no other techniques available to determine fracture toughness. One example is the determination of the fracture resistance of small single crystals, where the size of the available crystals (of the order of a few millimeters or less) precludes the use of standard fracture mechanics tests. Determination of the fracture toughness of single crystals poses an additional problem, however, due to the anisotropic nature of the elastic properties. Another use of the IF test is in the investigation of the possible role of environment, e.g. water, on crack extension. If environmentally enhanced crack growth takes place, then the cracks will extend to a greater distance than c0, the crack length measured in an inert environment. This factor must be taken into account in any calculation of fracture toughness, but it has been shown that it also possible to use the ratio of crack lengths in various environments to obtain at least a semi-quantitative measure of the material’s susceptibility to subcritical crack growth [42e44]. If one assumes that crack growth follows the relationship
N KI (29) V ¼ V0 KIC then V0 and N are empirical material parameters. Based on the analysis by Gupta and Jubb [42], one can derive the following expression for the ratio of crack lengths in two different environments:

2  c V0 t ð3Nþ2Þ ¼ ð3N þ 2Þ c0 2c0

(30)

where c0 is the initial crack length after indentation and c is the crack length measured at time, t. This expression is valid when ð3N þ 2ÞV0 t=2c0 [1. Despite the fact that some data are lost because of the difficulty in obtaining measurements at very short times, White et al. [43] and Cook and Liniger [44] have demonstrated that this procedure can be used to determine values of N that agree quite well with those obtained from classical fracture mechanics specimens. This indentation procedure has also been employed as a way of determining effects of residual stresses [45] and applied electric fields [46] on crack growth. It has become a primary technique for studying fracture in piezoelectric materials.

728

Handbook of Advanced Ceramics

6.2. Indentation Strength The IS technique to determine fracture toughness was suggested by Chantikul et al. [38], and is similar to the SCF test described earlier. The difference is that instead of removing the stress field due to the indentation, it is combined in a mathematical expression with the far-field stress as part of the driving force on the crack. Fracture strength can be measured in either uniaxial or biaxial flexure. Chantikul et al. [38] derived the following expression for Kc: Kc ¼ hRv


1=8  3=4 E sP1=3 H

(31)

where E is Young’s modulus, H is the hardness, s is the fracture strength, and P is the indentation load. Chantikul et al. [38] determined an average value for the constant hRv ¼ 0.59. The authors claim an accuracy of 30e40% for KC; however, one should note that some materials, glasses in particular, deviate from the trend line, and caution must be exercised in assuming that fracture toughness values for glasses are within this range. There are other factors which can lead to errors in the determination of KC by this method. These include interactions between the radial and lateral cracks and microstructural effects, including anisotropy. At small indentation loads, one must make sure that failure began from the indentation site and not a preexisting flaw in the surface or edge of the specimen. One test to determine whether this method is applicable to a particular material is to logarithmically plot fracture strength as a function of indentation load. The slope of the line should be e1/3. However, as shown in Figure 13, deviations from the behavior predicted by Eqn (31) can occur. Such deviations indicate that there are additional stresses in the material that have not been accounted for. Under no circumstances should a value of KC be taken from this regime. On the other hand, such deviations can be valuable in obtaining information regarding the influence of microstructural-level stresses on fracture. One must also be aware, as with other tests, of the possible role of test environment in reducing the apparent value of KC. Testing under inert conditions, e.g. in previously dried vacuum pump oil or nitrogen gas, and at fairly rapid loading rates will reduce the influence of environment. The essentially same IS procedure is followed in the socalled “dynamic fatigue” test in which the strength of an indented flexural bar is measured over a range of loading rates. The slope of the strength versus stressing rate curve yields a value of N, the measure of the materials’ sensitivity to environmentally enhanced crack growth. This test is

FIGURE 13 Strength as a function of the logarithm of the load used to produce the hardness indentation. Equation 45 indicates that the slope should be a e1/3, but deviations occur at small loads corresponding to small flaw sizes due to microstructural effects. (After Lawn et al. [47]).

FIGURE 14 Schematic of crack orientation in a flexural specimen which can be used to determine crack growth under mixed-mode I and -mode II conditions.

discussed in more detail in the chapter on Mechanical Reliability in this Handbook. The IS procedure is also useful for investigating mixedmode crack growth. Mixed-mode fracture can occur if the face of the flaw is oriented at an angle to the tensile stress. A schematic of such a situation is shown in Figure 14. Both KII and KIII components of the stress field can act on the flaw, with KII being greatest at the tensile surface, and KIII being largest at the maximum flaw depth. Effects of mode III loading generally will not play a major role in brittle fracture, and are generally ignored.

7. DOUBLE CLEAVAGE DRILLED COMPRESSION The double cleavage drilled compression (DCDC) specimen was first suggested by Janssen [48], and consists of a square compression rod, usually having approximate dimensions of 150  15  15 mm with a circular hole drilled through the center (as shown in Figure 15). When

Chapter | 9.6

Fracture Mechanics Measurements

729

FIGURE 16 Fracture toughness of ZnSe as a function of the flaw size to grain size ratio. (Rice et al. [53]).

FIGURE 15 DCDC specimen.

loaded in compression, the presence of the hole modifies the applied compressive stress to create tensile stresses leading to the formation of vertical cracks emanating from the hole. There have been a few analyses performed to determine the stress intensity factor as a function of the geometric parameters in the specimen [49,50]. The simplest of these obtained through a finite element analysis is given as follows [49]: KI ¼

sr 1=2 1:595 þ 0:3530

 a

(32)

r where r is the hole radius, a is the crack length, and s is the stress applied to the ends of the specimen. Michalske et al. [49] found that over the crack lengths considered the results were within 10% of the values of fracture toughness obtained through other tests. Because the crack propagates into a decreasing stress field, what is measured is a crack arrest value rather than KIC. The stability of the crack in terms of both its tendency to arrest and the fact that it grows uniformly on the centerline of the specimen make this configuration attractive for some crack growth studies such as the investigation of interfaces [51].

8. INTERPRETATION AND USE OF FRACTURE MECHANICS DATA Because of the low fracture toughness of ceramics, the sizes of microstructural features, e.g. grain size, can be nearly equivalent to the flaw sizes which limit the strength of actual components. Such features can become

a significant factor in the measurement and interpretation of crack growth. Therefore, a primary consideration in interpreting test results is the fact that many fracture mechanics tests require cracks much larger than the microstructure, e.g. DCB, DT, and SENB. Others such as indentation test methods employ much smaller cracks, of the order of the microstructure. The former tests are likely to yield significantly higher values of crack growth resistance due to microstructural mechanisms that apply only to the larger cracks. Therefore, fracture toughness values obtained with large cracks may not predict the behavior of small flaws, i.e. toughness may not translate into strength. The presence of microstructure can lead to what are called rising R (crack resistance) curves, that is, the resistance to crack growth increases with increasing crack size. R-curves can arise due to a number of factors: (1) A transition from single crystal to polycrystallinecontrolled fracture. This type of behavior is frequently experienced in relative large grain size material. Flaws within a grain will be governed by the fracture toughness of a single crystal of that material. This type of behavior is demonstrated quite clearly in Figure 16 for ZnSe [52,53]. The transition from single crystal to polycrystalline-controlled fracture can be very sharp [53] or broad depending on the crystal structure of the material and grain orientation. (2) A number of toughening mechanisms will apply to large cracks and not to small ones. For example, toughening mechanisms such as crack deflection [54,55], microcracking [56], transformation toughening [57], and grain bridging [58] depend on forces acting on the crack flanks. Therefore, until a steady state is reached, the longer the crack, the greater is the restraining force on it. However, even if R-curve behavior occurs in a material, increasing values of fracture toughness may not translate into higher strength [59].

730

FIGURE 17 Scanning electron micrograph of fracture MgF2 flexural specimen. (a) Critical flaw, (b) view of mist and hackle regions on fracture surface, (c) magnification of area C in B, showing primarily intergranular crack growth, and (d) magnification of D showing transgranular fracture.

Handbook of Advanced Ceramics

(a)

(c)

(b)

(d)

(3) In addition, the mode of fracture can vary with crack velocity. For example, it has been observed that crack growth in MgF2 changes from intergranular to transgranular when the crack reaches a particular velocity (Figure 17) [60].

9. SUMMARY This chapter summarizes the most important features of each of the leading fracture mechanics test procedures currently in use. The effect of the microstructure in ceramics is demonstrated to play an important role in fracture, and must be taken into account in the selection of a test procedure.

ACKNOWLEDGMENTS The many helpful discussions with George Quinn and Jack Mecholsky are gratefully acknowledged. The aid of Nick Mecholsky in preparing many of the figures is also appreciated.

REFERENCES [1] Chapter, this volume. [2] Freiman SW, Mecholsky JJ. The fracture of brittle materials. Wiley; 2012. [3] Griffith AA. The phenomena of rupture and flow in solids. Phil Trans R Soc Lond 1920;A221:163e98. [4] Irwin GR, Paris PC. In: Liebowitz H, editor. Chapter 1 in fracture 3. New York: Academic Press; 1971. [5] Irwin GR. Encyclopedia of physics VI. Heidelberg: Springer; 1958.

[6] ASTM Standard C 1421e01b “Standard test methods for determination of fracture toughness of advanced ceramics at ambient temperature”. [7] Obreimoff JW. The splitting strength of mica. Proc Roy Soc 1930;A127:290e7. [8] Gillis PP, Gilman JJ. Double cantilever cleavage mode of crack propagation. J Appl Phys 1964;35:647e58. [9] Wu CCm, McKinney KR, Lewis D. Grooving and off-center crack effects on applied-monent double-cantilever-beam tests. J Am Ceram Soc 1984;67:C-166e8. [10] Wiederhorn SM, Shorb AM, Moses RL. Critical analysis of the theory of the double cantilever method of measuring fracturesurface energies. J Appl Phys 1968;39:1569e72. [11] Freiman SW, Mulville DR, Mast PW. Crack propagation studies in brittle materials. J Mater Sci 1973;8:1527e33. [12] Outwater JO, Murphy MC, Kumble RG, Berry JT. In: Fracture toughness and slow stable cracking. STP 559 American Society for Testing and Materials; 1974. p. 127e38. [13] Shyam A, Lara-Curzio E. The double-torsion testing technique for determination of fracture toughness and slow crack growth behavior of materials: a review. J Mater Sci 2006;41: 4093e104. [14] Tait RB, Fry PR, Garrett GG. Review and evaluation of the doubletorsion technique for fracture toughness and fatigue testing of brittle materials. Exp Mech 1987;27:14e22. [15] Fuller Jr ER. In: Fracture mechanics applied to brittle materials. STP 678 American Society for Testing and Materials; 1979. p. 3e18. [16] Pletka BJ, Fuller Jr ER, Koepke BG. In: Fracture mechanics applied to brittle materials. STP 678, American Society for Testing and Materials; 1979. p. 19e37. [17] Trantina GG. Stress analysis of the double-torsion specimen. J Am Ceram Soc 1977;60:338e41.

Chapter | 9.6

Fracture Mechanics Measurements

[18] Ciccotti M. “Realistic finite-element model for the double-torsion loading configuration. J Am Ceram Soc 2000;83:2737e44. [19] McKinney KR, Smith HL. Method of studying subcritical cracking of opaque materials. J Am Ceram Soc 1973;56:30e2. [20] Evans AG. Method for evaluating the time-dependent failure characteristics of brittle materials and its application to polycrystalline alumina. J Mater Sci 1972;7:1137e46. [21] ASTM Standard C1161. Flexural strength of advanced ceramics at ambient temperature. Philadelphia, PA: ASTM; 1991. [22] ISO Standard 14704:2008, “Fine ceramics (advanced ceramics, advanced technical ceramics) e test method for flexural strength of monolithic ceramics at room temperature”; 2008. [23] Quinn GD, Morrell R. Design data for engineering ceramics: A review of the flexure test. J Am Ceram Soc 1991;74:2037e66. [24] Davidge RW, Tappin G. The effective surface energy of brittle materials. J Mater Sci 1968;3:165e73. [25] Munz D, Bubsey RT, Srawley JE. Int J Fract 1980;16:359e74. [26] ISO Standard 24370, Fine ceramics (advanced ceramics, advanced technical ceramics) e test method for fracture toughness of monolithic ceramics at room temperature by chevron-notched beam (CNB method). [27] Nishida T, Yoshikazu H, Pezzotti G. Effect of notch-root radius on the fracture toughness of a fine-grained alumina. J Am Ceram Soc 1994;77:606e8. [28] Kubler J. Fracture toughness of ceramics using the SEVNB method: preliminary results. Cer Eng Sci Proc 1997;18:155e62. [29] ISO Standard 23146:2008, “Fine ceramics (advanced ceramics, advanced technical ceramics) e test method for flexural strength of monolithic ceramics e single-edge V-notch beam (SEVNB) method”; 2008. [30] Nose T, Fujii T. Evaluation of fracture toughness for ceramic materials by a single-edge-precracked-beam method. J Am Ceram Soc 1988;71:328e33. [31] Warren R, Johanneson B. Creation of stable cracks in hard metals using ‘bridge indentation’. Powder Metall 1984;27:25e9. [32] ISO Standard 15732:2003, ISO Standard 18756:2003. “Fine ceramics (advanced ceramics, advanced technical ceramics) e test method for flexural strength of monolithic ceramics at room temperature by single edge precracked beam (SEPB) method”; 2003. [33] Petrovic JJ, Mendiratta PK. Fracture from controlled surface flaws. In: Fracture mechanics applied to brittle materials. STP 678, American Society for Testing and Materials 73e82; 1979. [34] ISO Standard 18756:2003, “Fine ceramics (advanced ceramics, advanced technical ceramics) e test method for flexural strength of monolithic ceramics at room temperature by the surface crack in flexure (SCF) method”; 2003. [35] Quinn GD, Salem JA. Effect of lateral cracks on fracture toughness determined by the surface-crack-in-flexure method. J Am Ceram Soc 2002;85:873e80. [36] Quinn GD. Fractography of ceramics and glasses. NIST Special Publication; 2007; 960e16. [37] Anstis GR, Chantikul P, Lawn BR, Marshall DB. A critical evaluation of indentation techniques for measuring fracture toughness: i, direct crack measurements. J Am Ceram Soc 1981;64:533e8. [38] Chantikul P, Anstis GR, Lawn BR, Marshall DB. A critical evaluation of indentation techniques for measuring fracture toughness: ii, strength method. J Am Ceram Soc 1981;64:539e43.

731

[39]

[40]

[41] [42] [43]

[44] [45] [46]

[47]

[48]

[49]

[50]

[51]

[52]

[53]

[54] [55] [56] [57] [58]

[59] [60]

Ponton CB, Rawlings RD. Vickers indentation fracture toughness test part 1, review of literature and formulation of standardized indentation toughness equations. Mat Sci Eng 1989;5: 865e72. Ponton CB, Rawlings RD. Vickers indentation fracture toughness test part 2, application and critical evaluation of standardized indentation toughness equations. Mat Sci Eng 1989;5:961e76. Quinn GD, Bradt RC. On the vickers indentation fracture toughness test. J Am Ceram Soc 2007;90:673e80. Gupta PK, Jubb NJ. Post-indentation slow growth of radial cracks in glasses. J Am Ceram Soc 1981;64:C-112e4. White GS, Freiman SW, Wilson AM. Indentation determination of crack growth parameters in gallium arsenide. J Am Ceram Soc 1991;74:419e21. Cook RF, Liniger EG. Kinetics of indentation cracking in glass. J Am Ceram Soc 1993;76:1096e105. Kese K, Rowcliffe DJ. Nanoendentation method for measuring residual stress in brittle materials. J Am Ceram Soc 2003;86:811e6. Raynes AS, White GS, Freiman SW, Rawal BA. Electrical field effects on crack growth in a lead magnesium niobate ceramic. Ceram Trans 1990;15:129e32. Lawn BR, Freiman SW, Baker TA, Cobb DD, Gonzalez AC. Study of microstructural effects in the strength of alumina using controlled flaws. Comm Am Ceram Soc 1984;67:C-67e9. Janssen C. Specimen for fracture mechanics studies on glass. In: Proc. 10th Int. Cong. on glass. Ceramic Society of Japan, Tokyo; 1974. pp. 10.23e10.30. Michalske TA, Smith WL, Chen EP. Stress intensity calibration for the double cleavage drilled compression specimen. Eng Fract Mech 1993;45:637e42. He MY, Turner MR, Evans AG. Analysis of the double cleavage drilled compression specimen for interface fracture energy measurements over a range of mode mixities. Acta Metal Mater 1995;43:3453e8. Turner MR, Evans AG. An experimental study of the mechanisms of crack extension along an oxide/metal interface. Acta Mater 1996;44:863e71. Freiman SW, Mecholsky Jr JJ, Rice RW, Wurst JC. Influence of microstructure on crack propagation in ZnSe. J Am Ceram Soc 1975;58:406e9. Rice RW, Freiman SW, Mecholsky Jr JJ. Effect of flaw to grain size ration on fracture energy of ceramics. J Am Ceram Soc 1980;63:129e36. Faber KT, Evans AG. Crack deflection processes-I. Theory. Acta Metall 1983;31:565e76. Faber KT, Evans AG. Crack deflection processes-II. Experiment. Acta Metall 1983;31:577e84. Fu Y, Evans AG. Microcrack zone formation in single phase polycrystals. Acta Metall 1982;30:1619e25. McMeeking RM, Evans AG. Mechanics of transformationtoughening in brittle materials. J Am Ceram Soc 1982;65:242e6. Freiman SW, Swanson PL. In: Barber DJ, Meredith PG, editors. Deformation processes in mineral, ceramics and rocks. Boston: Unwin Hyman; 1990. p. 72e83. Swain MV, Rose LRF. Strength limitations of transformationtoughened zirconia alloys. J Am Ceram Soc 1986;69:511e8. Mecholsky JJ. Intergranular slow crack growth in MgF2. J Am Ceram Soc 1981;64:563e6.

Chapter 9.7

Layered Ceramics Raul Bermejo and Marco Deluca Institut fuer Struktur- und Funktionskeramik (Institute for Structural and Functional Ceramics), Montanuniversitaet Leoben (University of Leoben), Franz-Josef-Strasse 18, 8700 Leoben, Austria

Chapter Outline 1. Introduction 2. Residual Stresses in Layered Ceramics 3. Mechanical Behavior

733 735 739

1. INTRODUCTION The interest for the mechanical behavior of ceramic materials has been always motivated by their possible application as structural components, especially in the cases where properties such as high hardness, chemical stability, low density, and high strength, among others, are sought. Due to their brittleness, ceramics have been used for many decades as structural elements, but usually under compressive loading conditions. Nowadays, most of the new engineering designs need to withstand tensile stresses, which imply potential limitations for ceramics due to their low fracture toughness and the sensitivity of their strength to the presence of defects [1e3]. This is very well known for glass, which is one of the strongest manufactured materials when the surfaces are free from flaws (as for the case of glass fibers), whereas under ordinary conditions “glass” is a synonym for fragility. The brittle fracture of glasses and ceramics is a consequence of the material defects located either within the bulk or at the surface, resulting from the processing and/ or machining procedures [4,5]. Under external applied stress, the stress concentration associated with those defects is the common source of failure for ceramic components. If each defect is considered as a crack or a potential source for crack initiation, then it becomes clear that the size and type of these defects determine the mechanical strength of the material [6]. The distribution of defects within a ceramic component yields a statistically variable strength which can be described by the Weibull theory [7,8]. Since flaws are intrinsic to processing and in most cases unavoidable, the reliability of ceramic components in terms of strength is associated with such flaw distribution.

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00039-3 Copyright Ó 2013 Elsevier Inc. All rights reserved.

4. Design Guidelines to Optimize Strength and Toughness 5. Outlook References

742 746 747

In order to reduce both the defect population and the defect size, many studies have been devoted in the past to improve ceramic processing. The use of colloidal routes has, to some extent, enabled the reduction of the critical size of the flaw causing the failure of the material, thus allowing fabrication of advanced ceramics with relatively high strength [9]. In an attempt to reduce the level of uncertainty in mechanical strength and to overcome the lack of toughness of structural and functional ceramics, several processing routes have arose in the last two decades which do not utilize the conventional “flaw elimination” approach, but rather use the implication of energy release mechanisms to obtain “flaw tolerant” (strength reliable) materials, with improved fracture toughness [9e26]. In this regard, the outstanding mechanical behavior of hierarchical layered structures found in nature has inspired material scientists to reproduce or mimic some of these architectures for improving engineering designs [10,15,16,27e32], for instance, the extraordinary toughness and strength of mollusk shells (Figure 1), which are related to their finescale structure, namely a laminate of thin calcium carbonate crystallite layers consisting of 99% calcium carbonate (CaCO3) and tough biopolymers, arranged in an energy-absorbing hierarchical microstructure [33]. The strength and toughness of such layered structures are significantly higher than those of their constituents [30,34]. In addition to the improvement of mechanical strength, there is another motivation for the production of layered ceramics. The demanding requirements for advanced devices involve in many cases the combination of several material classes (such as metals and ceramics) to fulfill the

733

734

FIGURE 1 Fracture of a mollusk shell consisting of CaCO3 brittle layers and tough biopolymers, arranged in a hierarchical microstructure. (Courtesy of Dr. Deville [33]).

performance in a given system. The unique mechanical and functional properties offered by ceramic materials make them good candidates for many advanced engineering applications (e.g. solid oxide fuel cells, piezoactuator and sensor devices, thermal barrier coatings, and conducting plates for wireless communications). The fabrication of components having two different materials is here a challenge from the viewpoint of not only structural integrity (i.e. mechanical resistance) but also functionality. Multilayer piezoelectric actuators, for instance, are constituted by thin piezoceramic layers with interdigitated metal electrodes [35,36]. In order to ensure the device functionality, the propagation of cracks between adjacent electrodes must be avoided [37,38]. Another example is the case of low temperature co-fired ceramics (LTTCs), where the concept of co-sintering multilayer ceramic substrates, metal electrodes, and vias at relatively low temperatures (i.e. ca. 900  C) has enabled the improvement of wireless communication systems (e.g. mobile phones and GPS technology) at relative low costs [39]. In LTCCs the mechanical reliability of the ceramic substrate (subjected to thermo-, mechanical-, and electrical loads) relies on avoiding crack propagation which would reduce or even completely hinder the performance of the microelectronic device. In recent years, as a result of remarkable progress in terms of microstructural design and advanced processing [9,31,40,41], toughness and strength of structural ceramics have been increasingly enhanced by crack shielding from microstructure-related mechanisms [10,42e48]. A direct

Handbook of Advanced Ceramics

consequence of these energy-dissipating toughening mechanisms, which aim to reduce the crack driving force at the crack tip, is the development of an increasing crack growth resistance as the crack extends, i.e. R-curve behavior. This concept first arose from studies in metals and alloys in the 1960s and was later applied to single-phase, duplex, and laminar ceramics by numerous authors [11,48e50]. Particular attention has been paid to fiber, platelets, and layer-reinforced ceramics, where the better mechanical performance is associated with the second phase or layer addition as well as with the arrangement of the fibers or the layer assemblage, respectively [32,45, 51e54]. As an extension of this laminar ceramic/fiberreinforced concept, multilayer designs have also been attempted in many ways aiming to improve both the resistance to crack propagation and the mechanical reliability of ceramic components [10,11,15,18,20,23,55e58]. This approach has been demonstrated to be more cost-effective than the former and more accurate in terms of tailoring mechanical requirements. The design of composite materials using such multilayer architectures (e.g. ceramic composites such as aluminaezirconia and mulliteealumina among others) has been reported to exhibit increased fracture toughness, higher energy absorption capability, and/or noncatastrophic fracture behavior in comparison with their constituent (monolithic) materials. Among the various laminate designs reported in literature, two main approaches regarding the fracture energy of the layer interfaces must be highlighted. On the one hand, laminates designed with weak interfaces have yielded significant enhanced fracture energy (failure resistance) through interface delamination [10,14,25,30,59e68]. The fracture of the first layer would be followed by crack propagation along the weak interface or within the weaker interlayer. This is the so-called “graceful failure,” which prevents the material from catastrophic failure e a fracture scenario where maximum applied load does not lead to unstable crack extension. In this case, the reinforcement mechanisms during fracture remind those from natural systems such as mollusk shells (cf. Figure 1). On the other hand, laminates designed with strong interfaces present crack growth resistance (R-curve) behavior through microstructural design (e.g. grain size and layer composition) and/or due to the presence of compressive residual stresses, acting as a barrier to crack propagation [15,16,20,21,24,55,58,69e71]. The increase in fracture energy in these laminates is associated with energy-dissipating mechanisms such as crack deflection/ bifurcation phenomena. Within this context, a commonly used structural design is that associated with the presence of compressive residual stresses and/or phase transformations. Compressive stresses could develop in the laminate during cooling from sintering due to differences in elastic or thermal properties (Young’s modulus, thermal

Chapter | 9.7

Layered Ceramics

expansion coefficient, etc.) between the layers [46,56,72]. The specific location of the compressive layers, either at the surface or internal, is associated with the attempted design approach, based on either mechanical resistance or damage tolerance, respectively. In the former case, the effect of the compressive residual stresses results in a higher, but singlevalue, apparent fracture toughness together with enhanced strength (the main goal) and some improved reliability [16,49,55]. On the other hand, in the latter case, the internal compressive layers are designed to rather act as a stopper to any potential crack growing from processing and/or machining flaws, at or near the surface layers such that failure tends to take place under conditions of maximum crack growth resistance [15,20,24,58]. The utilization of tailored compressive residual stresses acting as physical barriers to crack propagation has succeeded in many ceramic systems, yielding in some cases a so-called “threshold strength,” i.e. a minimum stress level below which the material does not fail [15,24,26,58,70,73,74]. In such layered ceramics, the strength variability of the ceramic material due to the flaw size distribution in the component is reduced, leading to a constant value of strength. The selection of multilayer systems with tailored compressive stresses either at the surface or in the bulk is based on the end application, and is determined by the loading scenarios where the material will work. The understanding of the conditions under which such reinforcement and energy-dissipating mechanisms occur and the influence of the layered architecture on the crack propagation must be pursued in order to improve modern structural and functional multilayer devices. The motivation of this chapter is to review the main approaches attempted in the field of layered ceramics to improve the mechanical properties of ceramic materials. Among the different possibilities to increase strength and toughness in multilayers, the use of residual stresses as key feature will be addressed aiming to provide some guidelines to design more reliable advanced ceramics.

2. RESIDUAL STRESSES IN LAYERED CERAMICS In every case where dissimilar materials are sealed together through a relative strong interface and subsequently undergo differential dimensional change, stresses arise between the materials [75]. Hence, a particular challenge in the processing of ceramic laminates is to understand the nature of these residual stresses, especially if they are to be used to enhance their mechanical properties. In ceramic laminates, residual stresses can be due to different factors: some of them are due to intrinsic causes such as epitaxial, variations of density or volume, densification, and oxidation at the surface. Others are extrinsic such as thermal or

735

thermoplastic strains developed during cooling or by external forces and momentums. Among them, the aspect most commonly referred to is the difference in the coefficients of thermal expansion (CTEs) between adjacent layers. During sintering of the laminate, it is considered that stresses between layers are negligible. However, as the temperature decreases, the differences in the CTE (ai) may promote a differential strain between layers. In addition to this differential strain, other strain differences, mainly due to phase transformations (Dεt) [46,57] or reactions (Dεr) [72] inside one layer, should also be considered. As a result, the final differential strain between two given layers A and B after cooling may be expressed as aB ÞDT þ Dεt þ Dεr

Dε ¼ ðaA

(1)

where DT ¼ Tref T0 is the difference between the reference temperature, i.e. the temperature at sintering where residual stresses are negligible, Tref, and the room temperature, T0.

2.1. Analytical Solution For ideal elastic materials, neglecting the influence of the external surfaces (where stresses may relax) and considering the laminate as an infinite plate the stress field can be determined analytically [12,75,76]. In each layer a homogeneous and biaxial residual stress state exists far from the free edges. The stress magnitude in each layer, sres;i , can be defined as sres;i ¼

Ei 1

ni

ða

ai Þ DT ¼

Ei 1

ni

Dεi ;

(2)

where Ei ; ni , and ai are material properties of the ith layer (i.e. Young’s modulus, Poisson’s ratio, and coefficient of thermal expansion, respectively). Dεi ¼ ða ai ÞDT is the mismatch strain of the ith layer, where the coefficient a is given as an averaged expansion coefficient of the laminate: , N N X X Ei ti ai Ei t i a ¼ ; (3) 1 ni 1 ni i¼1 i¼1

with ti being the thickness of the ith layer and N the number of layers. Note that the reference temperature Tref is, in practice, not easy to determine. It is always lower than the sintering temperature, since e if temperatures are reduced after sintering e the diffusion, which reduces the strain mismatch, does not stop abruptly but it becomes slower and slower. Generally, a normalization based on Eqn (2) and additional residual stress measurements have to be performed to determine Tref. For typical alumina-based or silicon-based ceramics, this stress-free empirical temperature is general between 1200  C and 1300  C [76].

736

Handbook of Advanced Ceramics

FIGURE 2 Residual stress distribution in two symmetrical layered systems with different combination of layer thicknesses but the same volume ratio. The magnitude of residual stresses in each layer, i.e. sA and sB, is the same in both cases.

When only two types of layer materials (A and B) are represented in the laminate (this is the case we will consider in the following), Eqn (3) can be written in the form !, nB nA EA aA X EB aB X tB;i tA;i þ $ $ a ¼ 1 nB i ¼ 1 1 nA i ¼ 1 ! nA nB EA X EB X tA;i þ $ tB;i : $ (4) 1 nA i ¼ 1 1 nB i ¼ 1

The magnitude of the residual stresses depends on the properties of A and B and on the ratio between P the total tA;i Þ and thickness Pof the layers of type AðTA ¼ tB;i Þ. The ratio of total thickness of the layer BðTB ¼ materials equals to their volume ratio: TB =TA ¼ VB =VA . It is interesting to note that the magnitude of residual stresses only depends on this volume ratio and not on the thickness of the individual layers i [77]: ! , nA nB X X tA; j ¼ fi ðTB =TA Þ tB; j sres;i ¼ fi j¼1

j¼1

¼ fi ðVB =VA Þ:

(5)

Finite element simulations of residual stress distributions in two symmetrical layered systems with the same volume ratio show the same magnitude of stresses in both cases, regardless of the combination of layer thicknesses of each material (see Figure 2). This is a very important aspect, which can provide the designer with more flexibility in order to tailor the disposition and thickness of layers when searching for an optimal design for a given level of residual stresses. FIGURE 3 (a) Tunneling crack through the tensile layer in a laminate with high tensile residual stresses, (b) edge crack in a laminate due to high stresses in the compressive layers [80].

(a)

2.2. Limitations for Design Although residual stresses can be a key feature to enhance the mechanical properties of many layered systems, some negative effects of high residual stresses should be considered. For instance, while compressive stresses are beneficial in acting as “shielding” mechanism against crack advance, tensile stresses will cause the cracking of the layer if they overcome its strength. A typical example is “tunneling cracks” that may appear at the surface of the tensile layers [78e81], and which can affect the structural integrity of the laminate (see Figure 3a). They can be avoided by lowering the level of tensile stresses in the corresponding layers. Another important aspect is the free surface of the material. It is well known that stresses at the free surface of layered materials are different from those within the bulk. In the region far from the free edges (i.e. in an infinite plate), biaxial residual stresses parallel to the layer plane exist, and the stresses perpendicular to the layer plane are negligible [82]. Near free edges, however, the residual stress state is no longer biaxial since the edge surface must be traction-free. As a result, a stress component perpendicular to the layer plane appears at the free edge [82e87]. This stress has a sign opposite to that of the biaxial stresses in the interior. Hence, for a compressive layer sandwiched between two tensile layers, a tensile residual stress perpendicular to the layer plane exists near the free surface of the compressive layer. The amplitude of maximum tensile residual stress at the free surface is related to the biaxial compressive residual stress in the interior. Although such tensile residual stress decreases rapidly from the edge surface to become negligible at

(b)

Layered Ceramics

737

VB sA ¼ s0 $ V A þ VB sB ¼

s0 $

VA VA þ VB

(6a) (6b)

In order to avoid edge cracks the strength of the layer, sc , seems to be a reasonable limit for jsA j, i.e. jsA j  sc . If tunneling cracks are to be avoided jsB j  sc =2 is a reasonable limit, as derived in Ref. [77]. Using Eqns (6a) and (6b) a limiting s0 to design laminates free of cracks can be derived:   1 (7) js0 j  sc $ 1 þ VB =VA   sc VB (8) js0 j  $ 1 þ 2 VA These two limits (Eqns (7) and (8)) give restrictions on the magnitude of js0 j to avoid edge cracks and tunneling cracks, respectively, and should be considered for design purposes. In Figure 4 both conditions are plotted as a function of VB =VA for a typical characteristic strength value of sc ¼ 400 MPa. Therefore, a region free of surface cracks can be estimated by tailoring the volume ratio between material A and B.

2.3. Experimental Determination of Residual Stresses The magnitude of residual stresses within a material or component can be experimentally determined by either

1000

tunnelling cracks

| (MPa)

800

0

a distance of the order of the compressive layer thickness, defects at the surface may be activated so that cracks may be initiated. An example are the so-called “edge cracks”, initiating from preexisting flaws, which can be encountered at the free edges of compressive layers (see Figure 3b). Preventing edge cracking is important for the structural integrity of the component and should be considered in the multilayer design, as it has been attempted by several authors [23,86,88]. A combined action to avoid both tunneling and edge cracks may be undertaken by selecting the level of residual stresses in the design so that the strength of the layers is not overcome. Such a limit for the magnitude of the residual stresses depends on each particular design and can be selected according to the properties of the employed materials. An example of such approach has been derived for aluminae zirconia-based multilayer ceramics [77]. It is assumed for both A-layers and B-layers the same elastic constants, fracture toughness, and strength, i.e. EA ¼ EB, nA ¼ nB, KIc,A ¼ KIc,B, and sc,A ¼ sc,B ¼ sc. The residual stress difference between adjacent layers is defined as s0 ¼ sA sB, with sA and sB being the residual stress in A-layers (compressive) and B-layers (tensile), respectively, which can be expressed as

edge cracks 600

σc

400

|

Chapter | 9.7

200

0 0

2

4

6

8

10

VB /VA (-) FIGURE 4 Design limits to prevent tunneling and edge cracks for a given volume ratio VB/VA between two materials in a layered structure. js0 j is the difference between the residual stresses in adjacent layers, sA sB. The mean strength of the layers is given by sc.

destructive or nondestructive methods. Mechanical techniques based on the use of a strain gauge are destructive in the sense that they require the component to be drilled or cut. Another method which is less invasive but still regarded as destructive is the use of micro-indentations (in general, Vickers imprints) [89,90]. This procedure is based on the measurement of the crack lengths arising from the tip of the imprint in both stressed and nonstressed materials. The magnitude and profile of the residual stresses are directly associated with the measured crack length difference. In Figure 5a a scheme of indentation profiles in inner and outer layers of an aluminaezirconia laminate is reported, whereas the results are displayed in Figure 5b. It becomes clear that the level of stresses involved in the outer layers is considerably lower. This method proves to be a quick and low-cost alternative to more complicated analytical methods (see below). However, since the indentations are performed at the surface of the specimen, the use of this technique is limited only to the evaluation of surface stresses. This effect is displayed in Figure 6 as the result of FE calculations of the average residual stresses near surface and at the center of the sample. As could clearly be seen, notable differences are present between stresses in the outer and inner layers. Nondestructive techniques for the measurement of residual stress in ceramic laminates generally are noncontact methods, namely they make use of an incident radiation (being X-rays, neutrons, or light) to probe the material surface in a position-resolved fashion. The penetration depth and the spatial resolution of the technique depend on the radiation wavelength and energy. In addition, the physical mechanism (especially the associated ease or difficulty in data interpretation) and the availability of the radiation source are very important aspects for the choice of the analytical technique.

738

Handbook of Advanced Ceramics

(a)

(b) Residual Stress [MPa]

FIGURE 5 (a) Scheme of the indentation profile in both inner and outer alumina layers of an aluminaezirconia multilayer, (b) plot of the experimentally measured residual stress profile.

B

Indentation imprints

A

B A

140 120

A inner layer

100 80 60 40

A outer layer

20

300 m

0 0

100

200

300

400

500

600

Distance to any A/B interface [ m]

The most common methods for residual stress measurements in multilayer ceramics are X-ray [91e93] and neutron diffraction [94e100]. Both methods are based on Bragg’s equation, which allows the accurate determination of atomic spacing and, consequently, elastic strains in crystalline materials. The main difference between these two methods (apart from the radiation source) is the fact that laboratory X-rays have a much shallower penetration depth (a few microns) and therefore allow only surface measurements, whereas neutron diffraction enables indepth measurements up to some millimeters [92,97]. Both techniques present significant challenges in case a positionresolved analysis is needed. In fact, generally, a neutron beam cannot be focused to a very small size (highest possible spatial resolution: ~200 mm on the material surface); on the other hand, traditional X-ray diffraction (XRD) methods involve rotation or tilt of the specimen [92], leading to a cumbersome experimental procedure. As a result, only macroscopic residual stresses are accessible by these techniques.

200

B

B A

A

B A

A

B

In the recent years, synchrotron XRD has emerged as a powerful analytical method for the study of microscopic residual stresses in ceramic laminates [101e104]. Due to the high energy of synchrotron X-rays and the advanced collimation and focusing equipment available at common beamlines, this technique allows noncontact measurements with a very narrow spot size (micron to submicron), whereas the wavelength tuneability of radiation allows choosing the penetration depth inside the material. 2D and 3D microscale residual stress measurements are made possible using special procedures such as the 3D diffraction microscopy (3DXRD), based on bidimensional detectors and sample rotation, or Laue microdiffraction, based on wavelength tuning [93]. An example of residual stress analysis in layered ceramics is constituted by the work of Leoni et al. [92] on aluminaezirconia laminates. Figure 7 shows a throughthickness residual stress profile obtained with synchrotron XRD and the Laue microdiffraction method on two alumina and one aluminaezirconia layer. In the central

A

Residual Stress [MPa]

100 0 -100 -200 -300 -400 -500

Interior

-600

Surface

-700 0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Distance in thickness direction [mm] FIGURE 6 FE calculation of residual stresses after sintering in the center and at the surface of an aluminaezirconia laminate. A difference in the average residual stresses between near-surface areas and the center of the laminate body can be observed.

FIGURE 7 Through-thickness residual stress profile in an aluminae zirconia laminate: design values (line) and measurement results for alumina (open dots) and zirconia (squares). The average stress is also shown (stars). (Courtesy of Dr. Sglavo [92]).

Chapter | 9.7

Layered Ceramics

alumina-zirconia layer, stress values belonging to both phases (calculated with isotropic elastic constants) are reported. The measurement was performed by varying the depth of the diffraction gauge volume. Examples of residual stress analyses in nanometer-sized layers also appeared recently in the literature [104], which prove the high suitability of synchrotron XRD also for the new trends in microelectronics and materials science. The need for a synchrotron X-ray or a neutron source, however, makes these techniques seldom available and not always cost-effective. These difficulties could be overcome by another method capable of position-resolved residual stress measurement on the microscale: Raman spectroscopy. This technique is based on the Raman effect, namely the inelastic scattering of monochromatic light (a laser) by a crystal. Raman spectroscopy is noncontact and nondestructive, and common spectrometers are laboratory-size equipment [105]. The spatial resolution (spot size and penetration depth) depends on the wavelength of the incident laser, on the choice of the objective lens, and on the absorption properties of the investigated material [106e109]. In common ceramics such as alumina, a spot size of ~1 mm and a penetration depth of ~15 mm can be achieved [108], which makes the technique surfacesensitive in comparison to the aforementioned XRD and neutron diffraction. The interpretation of the relationship between the Raman spectral shift and stress (piezospectroscopy) is nontrivial [110,111], but has been well established in the past for a wide range of materials (silicon [112,113], zirconia [114], and silicon nitride [115e118], among others). In the case of alumina, the influence of residual stress on the luminescence signal of intrinsic Cr3þ impurities in the ceramic body, induced by the laser and detected in the Raman spectrometer (photoluminescence, PL, henceforth), is used [87,111,119e122]. The physical mechanism in this case is different, in the sense that it is not based on the modification of the force constants of lattice vibrations (Raman piezo-spectroscopy), but rather on the shift in the position of the energy levels of Cr3þ impurities under the influence of a macroscopic stress [111]. Generally, a convolution of the calculated stress values by

739

a suitable stress model and the probe response function is done within the probe size, which leads then to a good approximation of the measured stress values [121]. An example of residual stress measurement performed with this method is provided in Figure 8 [122]; here the PL probe was scanned through alumina and aluminaezirconia layers, allowing to obtain different values of measured stresses for different layer configurations and thicknesses. It has to be noted that the stress values measured with the PL technique are generally lower than those predicted inside the layers [87]. This is associated with the free surface effect and also with the fact that the PL probe gives information only on the sum of principal stresses. Therefore, only mean stresses (trace of the stress tensor) can be measured at the surface. In the recent years, alternative techniques emerged in the scientific community, which could be useful in the future for analyses in ceramic laminates. They are electron back-scattered diffraction (EBSD) [123] and positron annihilation lifetime spectroscopy (PALS) [124]. The first technique can be carried out in an electron microscope, and provides very local (nanometer-scale) strain measurements. It is however only a surface-based technique and the methodology to avoid charging effects in ceramic materials is currently under development. PALS has already been used together with nanoindentation in aluminaezirconia laminates to study the influence of stress on localized defects. The technique has a relatively high penetration depth (~100 mm), and thus could be a valid alternative to neutron diffraction measurements.

3. MECHANICAL BEHAVIOR The response of a monolithic ceramic material to an external applied load can be characterized by its mechanical strength and its crack growth resistance. Both strength and toughness have proved to be enhanced using layered architectures with tailored properties (i.e. interface toughness, strength and toughness of each layer, composition and disposition of the layers, residual stresses, etc). Following the two main approaches in the design of ceramic

FIGURE 8 Stress profiles and micrographs of the cross section in multilayered AeAZ specimens with different thickness ratios among the layers. Each profile refers to the specimen showed on the right. Only half of the stress profile is plotted. (Courtesy of Dr. de Portu [121]).

Handbook of Advanced Ceramics

0

0

Deflection

FIGURE 9 Typical loadedeflection diagram of a layered structure with weak interfaces. The steps show the failure of individual layers and the crack propagation along the interface. A schematic side view of a broken specimen showing crack deflection into the weak layers is shown in the inset. The failure of a monolithic is also shown for comparison.

laminates, systems with weak or with strong interfaces can show a different response to fracture.

3.1. Laminates with Weak Interfaces Multilayer systems designed with weak interfaces or interlayers are aimed to prevent the component from catastrophic failure by guiding the crack along the interface or propagating the crack within the interlayer, and thus dissipating energy during the fracture process. The fracture process is based on the failure of the (stiffer) layers, and the strength of the system is associated with the mechanical resistance of such layers. Key works on this kind of multilayer material are those from Clegg et al. [10,14,125]. As an example, the fracture toughness of a silicon carbide monolithic ceramic is compared with that of a SiC FIGURE 10 Diagram to predict crack penetration into or deflection along the interface in a multilayer structure. When the crack propagates straight toward the interface (left) crack penetration occurs. If crack bifurcation takes place, the low angle with which the crack approaches the interface favors interface delamination (right).

multilayer containing graphite interlayers [125]. The socalled “graceful failure” of the material can be seen in the loadedeflection curve in Figure 9. The failure of one layer does not imply the catastrophic failure of the entire component, raising the apparent fracture toughness from 3.6 to 17.7 Mpa m1/2. A special case is that of porous interlayers, where crack deflection within the layer can occur under bending if the layer has a minimum volume fraction of porosity [14,63,65,126]. The crack path (i.e. deflection within the interlayer or along the interface) is dependent on the composition of the interlayer [59]. Models to predict the deflection of cracks within weak interlayers of laminates loaded under bending have been developed based on energy criteria. This has enabled optimizing the fracture energy of such multilayer systems as a function of interface toughness, strength of the layers, and Young’s modulus [127e130]. The particular case of a crack deflecting along a weak interface between two materials (having different elastic and/or mechanical properties) was first studied by He, Hutchinson, and Evans based on the type of loading and on the properties of the layers and interfaces (i.e. fracture energy of interface and layers, elastic constants of the individual layers, etc.) [131e133]. The tendency of a crack to deflect along or penetrate into the next layer depends on the relations between the involved fracture energies (of the material layers A and B, Glayer, and of the interface Gi) and the relevant energy release rates (of deflecting and penetrating cracks, Gd and Gp, respectively). This is also influenced by the combination of elastic properties of layers A and B, as described by the Dundurs parameters [134]. In Figure 10 a diagram showing the regions prone to crack deflection and to crack penetration is represented depending on the angle of crack propagation. The curves given by the ratio Gd/Gp limit both regions. When a crack propagates normal to the interface, it tends to

A

A1)

Gi /G layer

m on ol ith ic

Load

la m in at e

740

B

1.5

B2)

A

B

B1)

A2)

Gd /Gp

1.0 30° 45° penetration

60° 0.5

90° deflection

-1.0

-0.5

0.0

0.5

= ( E2' - E1' ) / ( E2' + E1' )

1.0

Chapter | 9.7

Layered Ceramics

741

penetrate, whereas cracks propagating with an inclined angle may favor interface delamination, as shown in Figure 10. It has been derived that the type of applied load (i.e. tensile or bending), the presence of residual stresses, and/or the angle of the approaching crack are key features conditioning the crack path. This has also been validated experimentally in different systems [14,25,60,62,68,135], even for multilayers with relatively strong interfaces [136,137].

3.2. Laminates with Strong Interfaces Multilayer architectures designed with strong interfaces can be subdivided into those with and without residual stresses in the layers. The mechanical behavior of the latter is enhanced by the different microstructures in the adjacent layers (i.e. composition, fine or large grain size, etc.), yielding crack branching or crack bridging as energydissipating mechanisms [50,138]. More significant, however, is the mechanical response of laminates with residual stresses. In this case, two multilayered designs, regarding the location of the compressive stresses either at the surface or within the bulk material, have been extensively investigated in order to tailor particular structural applications. Laminates with compressive stresses on the surface have proved to be useful for improving fracture strength as well as increasing wear resistance to contact damage [19,46,49,55,73,139e143]. On the other hand, if the compressive layers are internal, mechanical behavior enhancement is rather achieved in terms of flaw tolerance

(a)

and energy-dissipation mechanisms occurring at fracture [11,15,20,21,24,70,78,79,144e149]. From this perspective, attainment of a threshold strength, i.e. a failure stress that is independent of the original processing or machining flaw size, is a reliable evidence of the potential effectiveness of this approach [15,20,24]. An example of the mechanical response of both configurations is shown in Figures 11 and 12, respectively. In Figure 11a the bending strength distribution of an aluminaezirconia multilayer with external compressive stresses is plotted in a Weibull diagram and compared with the strength corresponding to a monolithic alumina-based ceramic. Due to the compressive stresses in the outer layer (this layer is subjected to tensile stresses under bending), the strength is enhanced with respect to the monolithic material (without compressive stresses). A slight improvement in the Weibull modulus can also be observed. The fracture path in these specimens is shown in Figure 11b. The presence of compressive stresses slightly deviates the crack from propagating straight. However, the propagation takes place under unstable conditions. In Figure 12a indentation strength results of similar aluminaezirconia multilayers are plotted together with an alumina monolithic material. For the latter the typical strength dependence with critical flaw size (indentation crack) can be observed as for brittle materials; the larger is the indentation crack introduced, the lower is the failure stress. On the other hand, the strength on the multilayer remains constant regardless of the initial indentation flaw size. A minimum failure stress is obtained for the different crack lengths introduced (threshold FIGURE 11 (a) Bending strength distribution of aluminaezirconia (A/AZ) laminates designed with compressive stresses in the external layers. The strength is compared to monolithic material (A) of the external layers. (b) Crack propagation through the layered structure; slight crack deflection is observed due to the compressive stresses in the layers.

(b) 2

99,9

A m = 10,4 = 492 MPa 0

99 90 80

1

0

63

-1

30 20

-2 10

ln ln 1/ (1-F)

probability of failure %

50

-3

-4 1

A/AZ -5 m = 18,1 = 650 MPa 0 300

386

472

558

strength [MPa]

644

730

-6

100 μm

742

Handbook of Advanced Ceramics

(a)

(b)

180

crack

160

Failure stress,

f

[MPa]

200

140 120 100

200 μm

Monolith: A Laminate: A/B

80 10

15

20

25

30

Square root of initial flaw size, c 1/2 [ m1/2 ]

(c)

FIGURE 12 (a) Indentation strength results in aluminaezirconia laminates (A/B) designed with internal compressive stresses. The strength is compared to monolithic material (A) of the external layers. (b) Arrest of an indentation crack due to the high compressive stresses in the approached layer. (c) Fracture pattern through bifurcation mechanisms, enhancing the toughness of the system.

strength). This is associated with the initial growth of cracks during bending until they reach the compressive (stopper) layer as can be seen in Figure 12b. The failure of the structure occurs when the crack propagates through the compressive layer. In such a case, the fracture takes place through deflection and/or bifurcation of the crack (see Figure 12c), which enhances significantly the fracture energy of the system. The design and use of multilayer systems with either external or internal compressive stresses should be motivated by the end application, as described in the Introduction. The design optimization of such structures in terms of strength and fracture resistance should also be regarded according to the property which is to be improved, as we will see below.

4. DESIGN GUIDELINES TO OPTIMIZE STRENGTH AND TOUGHNESS Layered ceramics designed with strong interfaces have proved to increase the strength and toughness of monolithic ceramics by tailoring residual stresses in the layers. Much effort has been focused on the optimization of symmetrical laminates with a given layer thickness ratio between

adjacent layers [26]. Recent research has shown that the distribution of layers of different materials can enhance significantly the fracture response of the laminate while maintaining a constant level of residual stresses [77]. Some design guidelines based on fracture mechanics criteria and experiments will be addressed in this section.

4.1. Fracture Mechanics Approach The fracture criterion of brittle materials is described by the linear elastic fracture mechanics (LEFM) based on the well-known Griffith/Irwin equation [1]: K  KIc

(9)

where K is the stress intensity factor and KIc is the fracture toughness of the monolithic material which can be experimentally determined using the standardized single-edge Vnotch beam (SEVNB) method [150]. For an external applied stress, sappl, an applied stress intensity factor, Kappl, can be defined as pffiffiffiffiffiffi (10) Kappl ðaÞ ¼ sappl Y pa where Y is a geometric factor depending on the crack shape and loading configuration and a is the crack length. The

Chapter | 9.7

Layered Ceramics

743

crack propagation is possible if the stress intensity factor at the crack tip, Ktip, equals or exceeds the intrinsic material toughness. The resistance to crack propagation in monolithic materials is measured normal to the direction of applied stress (mode I). However, in multilayer ceramics, especially those designed with residual stresses, the stress intensity factor at the crack tip as a function of the crack length, Ktip (a), can be given as the externally applied stress intensity factor Kappl (a) plus the contribution of the residual stresses, given by Kres (a). In these cases, an “apparent fracture toughness” as a function of the position of the crack within the layered structure is evaluated, which requires alternative approaches. In general, the effect caused by the residual stresses can be described by considering these stresses to be an additional external stress, thus adding a term Kres, to the stress intensity factor at the crack tip, Ktip, which reads

additional defect in the material (besides the crack) inducing an additional contribution to the crack driving force. This contribution is called the material inhomogeneity term Cinh. The thermodynamic force at the crack tip, denominated as the local, near-tip crack driving force Jtip, is the sum of the nominally applied far-field crack driving force Jfar (classical J-integral of fracture mechanics) and the material inhomogeneity term Cinh [153e155]. This method allows, for instance, taking into account the contribution of the different compliance of the layers to the crack propagation resistance. For the case of ceramic laminates with layers having relatively similar elastic constants, an alternative method to account for the term Kres(a) is the weight function (WF) approach. This allows us to calculate the apparent toughness KR(a) for an edge crack of length a for an arbitrary stress distribution acting normal to the prospective fracture path as [156]

Ktip ðaÞ ¼ Kappl ðaÞ þ Kres ðaÞ

Za

(11) KR ðaÞ ¼ K0

Thus, solving Eqn (11) for Kappl, the Griffith/Irwin criterion described by Eqn (9) now becomes (12)

where KR (a) is the “apparent fracture toughness”. For Kres < 0, as it holds for the action of compressive stresses, KR ðaÞ  KIc , what is called an increasing crack growth resistance curve (R-curve). This describes the “shielding” effect associated with the compressive stresses. If tensile residual stresses are acting, “anti-shielding” occurs and KR decreases with increasing crack extension. The evaluation of the fracture toughness (or fracture energy) of layered structures has been attempted by theoretical means using different approaches. One general procedure for the determination of Kres(a) is the finite element (FE) method. An FE model has to be created in order to compute the stress intensity factor at the crack tip under applied external stress for every position of the crack within the multilayer. Although there is, in general, a singularity problem at the interface, methods such as the J-integral can approximate the solution to the problem up to certain distance from the interface [151]. In order to describe the propagation of the crack along or through the interface, methods based on the LEFM have been developed such as the “finite fracture mechanics” approach, where the crack path near an interface between two different materials can be described [152]. Such methods can be used to describe the crack propagation resistance in linear elastic monolithic materials. However, the application to composites or layered structures combining different classes of materials (e.g. metaleceramic and ceramicepolymer) with different elastic constants should be carried out with care. In this regard, a new approach is the configurational forces (CF), which considers a material inhomogeneity (e.g. second phase and layer) as an

where K0 is the intrinsic fracture toughness of each individual layer, x is the distance along the crack length measured from the surface, a is the crack length, and h(a,x) is the weight function, which has been developed for an edge crack in a bar under different loading configurations [157]. The methods described above give very similar results and correlate with experimental findings for layered ceramics with residual stresses [77] (see Figure 13). Therefore, any of them can be used for K-value or J-integral calculations. Through comparison of the computational effort factors of the FE method, the CF method, and the WF method, it can be stated that the WF approach is much less time-consuming and thus is recommended when optimization processes of laminates are to be performed. 240

Weight function (Bermejo et al.) FEM (Sestakova et al.) Configurational forces (Chen et al.) SEVNB-Experiments (Pascual et al.)

200 160

2

Kres ðaÞ ¼ KR ðaÞ

(13)

0

Jappt (J/m )

Kappl ðaÞ  KIc

hðx; aÞsres ðxÞdx

A-layer

B-layer

A-layer

120 80 40 0 0.0

0.1

0.2

0.3

0.4

0.5

0.6

a (mm) FIGURE 13 Comparison of the apparent J-integral, Jappt, as a function of the crack length a, calculated using experimental as well as analytical and numerical methods, for an aluminaezirconia laminate with compressive residual stresses in the external layers [26,77,158,159].

744

Handbook of Advanced Ceramics

This will be used below to compare and optimize the mechanical properties of layered ceramics with external or internal compressive stresses.

4.2. Optimal Laminates with External Compressive Stresses

21

tA + tB

18

4

3

18

2

K R,peak

)

12

1 K appl

9

appl

ECS2 K Ic

6 3

tA,ECS2 0 0.00

tA + tB = 1000 μm

15

ECS1

1/2

15

Note that s0 is a negative number. Therefore, for a given s0 , pffiffiffiffiffi the peak toughness is maximum, if tB tA =ðtA þ tB Þ is maximum. For ðtA þ tB Þ ¼ constant, this value has its maximum for tB =tA ¼ 2. Figure 15 shows the peak toughness KR;peak versus tA =ðtA þ tB Þ for laminates with different values of ðtA þ tB Þ. It can be recognized that the maximum always occurs for tB =tA ¼ 2, i.e. tA =ðtA þ tB Þ ¼ 1=3, with the toughness increasing with tA þ tB . Additionally to the optimization of the apparent toughness, the strength of the laminate must also be regarded. The strength depends on the size of the fracture origins (the flaws where fracture starts), considered as the depth of the through-thickness edge crack. The typical range of flaw sizes occurring in a ceramic material depends on the processing conditions. For technical state-of-the-art materials, the typical size of a volume flaw is about 30 mm (about 10 mm for surface flaws) [8]. A typical range of fracture origins is indicated in Figure 14 (shaded bar). However, larger flaws may also occur, causing the strength scatter in that type of ceramics [3,160]. Indeed, in a laminate system, no processing flaws larger than the layer thickness ðtA Þ can occur. In order to interpret the

0.01

0.02

tA,ECS1 0.03

K R,peak (MPa.m

Apparent toughness, K R (MPa.m1/2 )

In this section, the design optimization in terms of toughness and strength of an aluminaezirconia laminate with external compressive stress (ECS) is illustrated. Material properties used in this study are based on typical values for aluminaezirconia periodic laminates (i.e. all layers of material A have the same thickness; the same holds for material B) [23]. Hence, elastic properties are chosen as E ¼ 390 GPa, n ¼ 0.22, and coefficients of thermal expansion between layers as Da ¼ 1  10 6 K 1. In Figure 14 the qualitative behavior of ECS laminates is shown (the data correspond to laminates with s0 ¼ 590 MPa). For the determination of the apparent toughness, the WF method has been employed. Plotted pffiffiffiffiffiffi are the R-curves versus the crack length parameter Y pa for the first two layers of two laminates having a different material volume ratio VB =VA (i.e. for ECS1: VB =VA ¼ 7=3 and for ECS2: VB =VA ¼ 9=1). The geometric factor Y is taken as 1.12 for an edge crack. Therefore, the stress magnitudes in the layers of the two laminates are also different (i.e. for ECS1: sA ¼ 413 MPa, sB ¼ þ177 MPa and for ECS2: sA ¼ 531 MPa, sB ¼ þ59 MPa). It is assumed that the thickness of the first two layers is constant ðtA þ tB ¼ 1 mmÞ for both laminates. In the laminate ECS1, the compressive outer A-layer is thicker ðtA Þ than in laminate ECS2 but the magnitude of the compressive stresses is lower. The second (tensile) B-layer is thinner ðtB Þ in ECS1 but the stress magnitude is higher than in ECS2. The shielding effect of the compressive residual stresses causing a rise of the

R-curve within the first layer can clearly be recognized. In the analyzed region (the first two layers) maximum shielding is always achieved at the first A/B interface of the external compressive A-layer, i.e. the length of the crack having a maximum shielding is a ¼ tA. For this case, the influence of the compressive residual stresses in the outer A layer can easilypffiffiffiffiffiffi be determined. It simply holds: KR ðaÞjatA ¼ 1:12$sA pa with sA ¼ s0 $VB =ðVA þ VB Þzs0 $tB =ðtA þ tB Þ. Maximum shielding is reached if a ¼ tA . Therefore, the peak toughness is pffiffiffiffi pffiffiffiffiffi (14) KR;peak ¼ K0 1:12 p$s0 $tB tA =ðtA þ tB Þ

(tA + tB) increases

12

tA + tB = 200 μm

9 6

tA + tB = 100 μm 0.04

0.05 1/2

Crack length parameter, 1.12( a)

0.06

1/2

(m )

FIGURE 14 Apparent fracture toughness of laminates designed with external compressive stresses. The peak toughness, KR, peak, depends on the thickness of the first layer. The range of typical defect sizes is represented as a shaded bar.

1/3

3 0.0

0.2

0.4 0.6 t A/(t A+t B) (-)

0.8

1.0

FIGURE 15 Peak toughness as a function of the thickness ratio tA/tA þ tB for different total thicknesses of the first two layers tA þ tB. A maximum toughness value is obtained independent of the tA þ tB selected.

Layered Ceramics

The first term of Eqn (15) increases as tA is reduced. It is obvious that this threshold stress can be increased by decreasing the layer thickness, but, for technological reasons, laminates with layers thinner than 5e10 mm can, today, hardly be processed. Then the contribution of the first term to the threshold can reach the characteristic strength of the material if the cracks are as large as the thickness of the first layer. The second term is positive and, for tA2/3Tm (where Tm is the absolute melting point), when the grain size is very small (less than several micrometers for metals and less than 1 mm for ceramics) and stable during deformation [1,2]. Micrograins apparently move past one another like sand particles flowing, as long as the fracture is suppressed by accommodation processes such as diffusion and dislocation motion. The tensile specimens can be elongated uniformly without necking at relatively low stresses (Figure 1). This property has been applied to manufacture net-shaped components of some superplastic alloys. While ceramics are brittle materials that show almost no plastic deformation at ambient temperatures, a very wide range of superplastic ceramics had been developed since the superplasticity of zirconia (ZrO2) and its composite was

4. Application of Superplasticity References

768 769

found in the mid-1980s [3,4]. The ceramic superplasticity is common to fine-grained ceramics that are produced by controlling the microstructures to sub-micrometer scales, for example, oxides (ZrO2, alumina, and their composites), nonoxides (silicon nitride [5], SiC), bioceramics, and superconductors, as has been reviewed extensively [6e8]. The ceramic superplasticity is of industrial interest, as it forms the basis of a fabrication method that can be used to produce components having complex shapes from materials that are hard to machine. The superplasticity of ZrO2 has been applied to superplastic forging, sinter forging, sheet forming, gas-pressure forming, extrusion, deep drawing, stretch forming, and superplastic joining. The superplastic sinter forging is also used to produce a textured silicon nitride with improved strength and fracture toughness. The use of superplastic forming may become even more widespread if large deformation can be achieved at higher strain rates, lower flow stresses, and lower temperatures. The better superplastic ceramics have been developed by improving processing methods and by controlling the grain boundary structure and its chemistry by adding dopant atoms [9]. Especially, high-strain-rate superplasticity in ZrO2-based composite is promising for an efficient shapeforming technology [10,11].

2. MECHANICAL PROPERTIES AND MECHANISM OF SUPERPLASTICITY FIGURE 1 Superplastic elongation of an SiAlON [57]. Undeformed (top) and deformed (bottom) specimens. The elongation of 470% has been achieved.

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00041-1 Copyright Ó 2013 Elsevier Inc. All rights reserved.

In superplastic alloys, the flow stress is particularly sensitive to the rate of deformation s ¼ K ε_ m

(1)

765

766

where s is the stress, ε_ is the strain rate, m is the strain-rate sensitivity index, and K is a constant. The very large elongation occurs at m  0:3, because large m is a necessary condition for the stability of uniform elongation without local necking. Alternatively, the constitutive equation is described as [12]     ADGb b p s s0 n (2) ε_ ¼ G kT d where D is the diffusion coefficient ( ¼ D0 expð Q=kTÞ,Q is the activation energy), G is the shear modulus, b is the Burgers vector, k is the Boltzmann constant, T is the absolute temperature, n is the stress exponent, p is an exponent of the inverse grain size, s0 is the threshold stress [13], and A is a dimensionless constant. Equation (2) is equivalent to Eqn (1) at s0 ¼ 0, and n is the inverse of m. The grain structure remains almost equiaxed after the elongation of hundred percent in superplasticity. This behavior is quite different from that observed in diffusional creep (Figure 2(a)) where grains distort and elongate. Ashby and Verral [14] proposed the grain switching model for superplasticity (Figure 2(b)) from the observation of deformation of oil grains or soap froth. The four grains exchange their neighbors at the final state so that the equiaxed shape is restored. Gifkins [15] considered that the boundary “mantle” of the grain behaved differently from the central “core” of the grain. Grains change their shape by deformation of the “mantle” in his core-mantle model (Figure 2(c)). The shift in the marker line shows grain boundary sliding. The grain compatibility during grain boundary sliding is maintained by a concurrent accommodation process, which involves diffusion, dislocation motion, and grain boundary migration. The grain switching event in the two-dimensional model consists of two

Handbook of Advanced Ceramics

independent processes, formation and disappearance of a grain boundary, in three dimensions [16]. When grain boundary sliding is accommodated by grain boundary diffusion and bulk diffusion, the exponents of inverse grain size are p ¼ 3 and p ¼ 2, respectively. With reference to Eqn (2), at constant temperature, decreasing the value of the grain size d will increase the strain rate. But if the grain size is very small and the grain boundary diffusion is very fast, the strain rate is controlled by interface reaction, which is the creation and annihilation of vacancy, rather than by the kinetics of long-range diffusion [17]. At elevated temperatures, failure of ceramics commonly occurs intergranularly by the cavities growing to coalesce to form cracks [18]. Thus, the suppression of cavity formation is important to allow superplastic elongation. A fast grain boundary diffusion, a small grain size, and a small ratio of grain boundary energy to surface energy are necessary for suppressing the nucleation of cavities [19]. To ensure the reliability of superplastically formed components, it is necessary to characterize the cavitation damage induced by the superplastic deformation. The cavity size increases exponentially with strain in superplasticity of ZrO2 [20]. The cavity formation and the maximum elongation to failure are influenced by processing-dependent microstructural factors such as the size distribution of matrix grains, second-phase particles, and residual defects [21]. Superplastic deformation is often accompanied by grain growth, the rate of which depends on either strain [22] or stress [23] and is usually well in excess of that found in the absence of deformation (static grain growth). The enhanced grain growth in superplasticity is termed the dynamic grain growth.

3. SUPERPLASTIC CERAMICS 3.1. Zirconia

FIGURE 2 Shear deformation in a regular array of grains [18]. (a) Diffusional creep; (b) Soap froth model; (c) Core-mantle model. (The gray circles show the “core”.)

The fracture toughness of ZrO2-based ceramics is enhanced by the stress-activated tetragonal to monoclinic transformation [24,25], in other words, by the transformation toughening. The fine grain size, which was required to retain the metastable tetragonal phase, facilitated also the superplasticity of Y2O3-stabilized tetragonal ZrO2 polycrystals (Y-TZP) [3]. The extensive studies on superplasticity of Y-TZP have been summarized in a review [7]. The relation between the strain rate and stress of high purity Y-TZP is schematically plotted as a solid line in Figure 3 on logarithmic scales. The apparent stress exponent n, which is given by the slope of the curve, is approximately n ¼ 2 at the high stress region, but it increases to n > 3 at the intermediate stress region. Several mechanisms have been proposed to explain this transition in stress exponent: 1)

Chapter | 9.9

Development of Superplastic Ceramics

767

3.2. Composites

FIGURE 3 Relation between strain rate ð_εÞ and stress ðsÞ in arbitrary units. The solid line schematically shows the experimental result for high purity Y-TZP [29], and the broken line shows this for low purity Y-TZP.

Transition of the accommodation process of grain boundary sliding from diffusion controlled one at high stresses to interface controlled one at the intermediate stresses [17,27,28], 2) Threshold stress [7,13]. It has been also reported that the curve shows n ¼ 1 at stresses lower than the threshold stress [29]. The deformation of Y-TZP is significantly affected by small amounts of impurities (broken line in Figure 3) [30]. The strain rate at the low stress region is enhanced by the addition of 0.12 wt% Al2O3, and n z 2 is observed in a wide range of stresses [31]. A similar effect is also observed in a material containing 10 2 s 1, it is conveniently called high-strain-rate superplasticity [1]. High-strain-rate superplasticity of the ZrO2eAl2O3espinel composite [10,11] was achieved at a temperature (1650  C) 200  C higher than the deformation temperature for Y-TZP (1450  C), mainly because the grain growth was suppressed by dispersion of the secondand the third-phase particles. The grain boundary sliding was partly accommodated by dislocation motion in spinel grains [43], and the threshold stress was observed in the deformation of the composite [44].

3.3. Silicon Nitride Silicon nitride (Si3N4) is a light, hard, and strong engineering ceramic that has been developed mainly as a structural material for high-temperature applications. Although creep resistance and superplasticity are incompatible functions, superplastic forming of silicon nitride can be applied to make wear-resistant components that are used at intermediate temperatures. The liquid-phase sintered Si3N4 has residual glass phase pockets and a thin glass film with a thickness of approximately 1 nm at grain boundaries. The superplasticity and creep of Si3N4 are affected by properties of the intergranular glass phase: chemical composition, quantity, glass transition temperature, viscosity, and solubility of Si3N4 to the liquid. A wide variety of superplastic silicon nitrides have been developed during 1982e2012: (1) Si3N4/SiC nanocomposite, (2) b-silicon nitride, (3) b-SiAlON, (4) equiaxed a/b-silicon nitride, (5) equiaxed a/b-SiAlON, and (6) equiaxed b-silicon nitride. The superplasticity of silicon nitride was found for the first time in the deformation of fine-grained Si3N4-based composite (material (1)), which was developed by sintering

768

amorphous SieCeN powder [5,45]. Self-reinforced b-silicon nitrides are analogs to whisker-reinforced ceramics relying on the formation of elongated rodlike b-grains in sintering of a-powder. When rodlike grains are very small, silicon nitride (material 2) can still exhibit superplastically large elongations despite this peculiar microstructure [46,47]. SiAlON is a solid solution of Si3N4, in its structure, some Si and N atoms are replaced by Al and O atoms. The strain rate of b-SiAlON (material 3) is enhanced significantly by the liquid phase transiently formed during sintering [48,49]. Because equiaxed grains are preferable for grain boundary sliding and thus, for superplasticity, materials consisting of equiaxed a/b-grains have been developed (materials 4 [50], 5 [51,52]). However, the microstructure of these materials is unstable due to the a- to b-phase transition and grain growth during deformation. This problem was solved by using b-grains as the starting powder in sintering, to get stable fine microstructures with equiaxed grains (material 6) [53e55]. When superplastic forming is conducted on silicon nitride with an equiaxed microstructure, the fracture toughness and the strength at elevated temperatures can be improved by a heat treatment to form elongated rodlike grains for the self-reinforcement. On the other hand, in the superplastic deformation of silicon nitride with rodlike grains, the grains align parallelly to the tensile direction in tension tests [54], and in a plane vertical to the compression direction in compression tests [47]. The formation of anisotropic texture can improve the fracture toughness and strength of components in a specific direction [47]. The superplastic forging of silicon nitride has been achieved at high-strain rates by using a spark plasma sintering (SPS) furnace [56]. The superplasticity of silicon nitride takes place by grain boundary sliding accommodated by viscous flow of intergranular glass phase and solution-precipitation creep [57]. In the superplasticity of equiaxed silicon nitride [52], the solution-precipitation process is controlled by diffusion at high stress region [58], and by interface reaction at low stress region [59]. A peculiar phenomenon of shear thickening is also observed in the compressive deformation of SiAlON [51].

3.4. Silicon Carbide Silicon carbide is a covalent material that adopts network structures featuring vertex sharing of [SiC4]. Although the diffusion coefficient in SiC is very slow, the superplasticity can be achieved by doping with a small amount of boron [60]. The doped boron segregates at grain boundaries and takes the place of silicon, forming bonds in a local environment that is similar to that in the B4C structure [61]. It is supposed that the segregated boron enhanced the grain

Handbook of Advanced Ceramics

boundary diffusion, because the diffusion of Si is faster by three orders of magnitude in B4C than in SiC. The superplasticity of SiC is also affected by oxygen impurity atoms [62] and the doping of Al atoms [63].

3.5. Functional Materials Hydroxyapatite (Ca10(PO4)6(OH)2) is biocompatible with bone and can be used for orthopedic and dental implants. When hydroxyapatite is used as a bone replacement, nearnet-shape forming is required to fit for each patients. Hydroxyapatite, with a grain size of sub-micrometer, is translucent, and it can be superplastically deformed at a relatively low temperature of 1000  C due to its fast diffusivity [64e66]. Carbonate apatite (CAP) is resorbed by osteoclasts and is biocompatible similar to hydroxyapatite. Fine-grained CAP can also be superplastically formed [67]. Oxide superconductor is brittle, and it is usually hard to make wires and ribbons by plastic deformation. A finegrained YBa2Cu3O7 x could undergo very large deformation at elevated temperatures, and superplastic elongation was demonstrated recently [68].

4. APPLICATION OF SUPERPLASTICITY In the ceramic industry, complex-shaped components with accurate dimensions have been usually fabricated by sintering the shaped powder compact. Therefore, in comparison to the sintering, two factors are important for the practical application of superplastic forming: 1) reliability and 2) efficiency. First, it is necessary to ensure the reliability of the products. The source of variability in strength is related to flaws, particularly for ceramics due to their inherent brittleness. The cavitation must be suppressed by selecting appropriate forming conditions. From the viewpoint of efficiency, the combination of sintering and forming, for example, sinter forging [69], is most promising, because both densification and net shaping are achieved simultaneously. Low temperature and high-strainrate superplastic sinter forging of alumina/spinel composites is actually successfully demonstrated. Porous preforms of nanoceramic composites that were partially densified at low temperatures were superplastically deformed by SPS at 1000e1050  C at 10 2 s 1 [70]. Spark plasma sintering is similar to conventional hot pressing, and the powder compact (green ceramics) is heated by a direct-pulsed DC current in the graphite die. The compressive deformation of SiAlON also exceeded 10 2 s 1 at 1500  C in the SPS apparatus [56]. This result suggests a favorable relationship between high-strain-rate superplasticity and ultrafast sintering. Superplastic forming of ZrO2/alumina/spinel

Chapter | 9.9

Development of Superplastic Ceramics

composite could be performed at 1,150  C by using SPS [71]. Superplastic forming has been applied mainly to ZrO2, for example, sheet forming [72], gas-pressure forming [73], extrusion [74], deep drawing [75], stretch forming [76]. The superplastic forming concurrent with the diffusion bonding process, which is used in the production of titanium alloys in the aerospace industry, is also applicable to ceramics. Superplasticity has been used for the bonding or joining of dissimilar ceramic materials, for example, ZrO2/alumina composites [77e79], ZrO2/hydroxyapatite composites [80], and functionally gradient materials [81]. The superplastic joining of ceramics and ceramic composites has been summarized in a review [82].

769

[17] [18] [19]

[20]

[21]

[22]

REFERENCES [1] Nieh TG, Wadsworth J, Sherby OD. Superplasticity in metals and ceramics. Cambridge University Press 1996. [2] Chokshi AH, Mukherjee AK, Langdon TG. Superplasticity in fine grained ceramics and ceramic composites: current understanding and future prospects. Mater Sci Eng R 1993;10:237e74. [3] Wakai F, Sakaguchi S, Matsuno Y. Superplasticity of yttria-stabilized tetragonal ZrO2 polycrystals. Adv Ceram Mater 1986;1: 259e63. [4] Wakai F, Kato H. Superplasticity of TZP/Al2O3 composite. Adv Ceram Mater 1988;3:71e6. [5] Wakai F, Kodama Y, Sakaguchi S, Murayama N, Izaki K, Niihara K. Superplasticity of hot isostatically pressed hydroxyapatite. Nature 1990;344:421e60. [6] Chen I-W, Xue LA. Development of superplastic structural ceramics. J Am Ceram Soc 1990;73:2585e609. [7] Jime´nez-Melendo M, Domı´nguez-Rodrı´guez A, Bravo-Leo´n A. Superplastic flow of fine-grained yttria-stabilized zirconia polycrystals: constitutive equation and deformation mechanisms. J Am Ceram Soc 1998;81:2761e7. [8] Wakai F, Kondo N, Shinoda Y. Ceramics superplasticity. Curr Opin Solid State Mater Sci 1999;4:461e5. [9] Sakuma T, Yoshida H. High temperature grain boundary plasticity in ceramics. Mater Trans 2009;50:229e35. [10] Kim BN, Hiraga K, Morita K, Sakka Y. A high-strain-rate superplastic ceramic. Nature 2001;413:288e91. [11] Hiraga K, Kim B-N, Morita K, Yoshida H, Suzuki TS, Sakka Y. High-strain-rate superplasticity in oxide ceramics. Sci Tech Adv Mater 2007;8:578e87. [12] Mukherjee AK. An examination of the constitutive equation for elevated temperature plasticity. Mater Sci Eng A 2002;322:1e22. [13] Mohamed FA. Interpretation of superplastic flow in terms of a threshold stress. J Mater Sci 1983;18:582e92. [14] Ashby MF, Verral RA. Diffusion-accommodated flow and superplasticity. Acta Mater 1973;21:149e63. [15] Gifkins RC. Grain rearrangements during superplastic deformation. J Mater Sci 1978;13:1926e36. [16] Wakai F, Shinoda Y, Ishihara S, Dominguez-Rodriguez A. Topological transformation of grains in superplasticity-like deformation. Acta Mater 2002;50:1177e86.

[23]

[24] [25]

[26] [27]

[28]

[29]

[30]

[31]

[32]

[33]

[34]

[35]

[36]

Arzt E, Ashby MF, Verral RA. Interface controlled diffusional creep. Acta Metall 1983;31:1977e89. Chan KS, Page RA. Creep damage development in structural ceramics. J Am Ceram Soc 1993;76:803e26. Evans AG, Rice JR, Hirth JP. Suppression of cavity formation in ceramics: prospects for superplasticity. J Am Ceram Soc 1980;63:368e75. Schissler DJ, Chokshi AH, Nieh TG, Wadsworth J. Microstructural aspects of superplastic tensile deformation and cavitation failure in a fine-grained yttria stabilized tetragonal zirconia. Acta Metall Mater 1991;39:3227e36. Hiraga K, Nakano K, Suzuki TS, Sakka Y. Processing-dependent microstructural factors affecting cavitation damage and tensile ductility in a superplastic alumina dispersed with zirconia. J Am Ceram Soc 2002;85:2763e70. Wilkinson DS, Caceres CH. On the mechanism of strain-enhanced grain growth during superplastic deformation. Acta Metall 1984;32:1335e45. Kim B-N, Hiraga K, Sakka Y, Ahn B-W. A grain-boundary diffusion model of dynamic grain growth during superplastic deformation. Acta Mater 1999;47:3433e9. Garvie RC, Hannink RH, Pascoe RT. Ceramic steel? Nature 1975;258:703e4. Gupta TK, Bechtold JH, Kuznicki RC, Cadoff LH, Rossing BR. Stabilization of tetragonal phase in polycrystalline zirconia. J Mater Sci 1977;12:2421e6. Wakai F, Nagano T. Effects of solute ion and grain size on superplasticity of ZrO2 polycrystals. J Mater Sci 1991;26:241e7. Owen DM, Chokshi AH. The high temperature mechanical characteristics of superplastic 3 mol% yttria stabilized zirconia. Acta Mater 1998;46:667e79. Berbon MZ, Langdon TG. An examination of the flow process in superplastic yttria-stabilized tetragonal zirconia. Acta Mater 1999;47:2485e95. Morita K, Hiraga K. Critical assessment of high-temperature deformation and deformed microstructure in high-purity tetragonal zirconia containing 3 mol.% yttria. Acta Mater 2002;50:1075e85. Lakki A, Schaller R, Nauer M, Carry C. High temperature superplastic creep and internal friction of yttria doped zirconia polycrystals. Acta Mater 1993;41:2845e53. Sato E, Morioka H, Kuribayashi K, Sundararaman D. Effect of small amount of alumina doping on superplastic behavior of tetragonal zirconia. J Mater Sci 1999;34:4511e8. Morita K, Hiraga K, Kim B-N. Effect of minor SiO2 addition on the creep behavior of superplastic tetragonal ZrO2. Acta Mater 2004;52:3355e64. Yoshizawa Y, Sakuma T. Role of grain-boundary glass phase on the superplastic deformation of tetragonal zirconia polycrystals. J Am Ceram Soc 1990;73:3069e73. Kajihara K, Yoshizawa Y, Sakuma T. The enhancement of superplastic flow in tetragonal zirconia polycrystals with SiO2-doping. Acta Mater 1995;43:1235e42. Ikuhara Y, Thavorniti P, Sakuma T. Solute segregation at grain boundaries in superplastic SiO2-doped TZP. Acta Mater 1997;45:5275e84. Mimurada J, Nakano M, Sasaki K, Ikuhara Y, Sakuma T. Effect of cation doping on the superplastic flow in yttria-stabilized tetragonal zirconia polycrystals. J Am Ceram Soc 2001;84:1817e21.

770

[37] Kuwabara A, Nakano M, Yoshida. H, Ikuhara, Sakuma T. Superplastic flow stress and electronic structure in yttiria-stabilized tetragonal zirconia polycrystals doped with GeO2 and TiO2. Acta Mater 2004;52:5563e9. [38] Yoshida H, Morita K, Kim BN, Hiraga K, Yamamoto T. Doping amount and temperature dependence of superplastic flow in tetragonal ZrO2 polycrystal doped with TiO2 and/or GeO2. Acta Mater 2009;57:3029e38. [39] French JD, Zhao J, Harmer MP, Chan HM, Miller GA. Creep of duplex microstructure. J Am Ceram Soc 1994;77:2857e65. [40] Clarisse L, Baddi R, Batille A, Crampon J, Duclos R, Vicens J. Superplastic deformation mechanisms during creep of aluminae zirconia composites. Acta Mater 1997:3843e53. [41] Suzuki TS, Sakka Y, Nakano K, Hiraga K. Effect of ultrasonication on the microstructure and tensile elongation of zirconia-dispersed alumina ceramics prepared by colloidal processing. J Am Ceram Soc 2001;84:2132e4. [42] Yoon CK, Chen IW. Superplastic flow of two-phase ceramics containing rigid inclusions- zirconia/mullite composite. J Am Ceram Soc 1990;73:1555e65. [43] Morita K, Hiraga K, Kim B-N. High-strain-rate superplastic flow in tetragonal ZrO2 polycrystal enhanced by the dispersion of 30 vol.% MgAl2O4 spinel particles. Acta Mater 2007;55:4517e26. [44] Chen T, Mohamed FA, Mecartney ML. Threshold stress superplastic behavior and dislocation activity in a three-phase aluminazirconia-mullite composite. Acta Mater 2006:4415e26. [45] Rouxel T, Wakai F, Izaki K. Tensile ductility of superplastic Al2O3eY2O3eSi3N4/SiC composites. J Am Ceram Soc 1992;75: 2363e72. [46] Burger P, Duclos R, Crampon J. Microstructure characterization in superplastically deformed silicon nitride. J Am Ceram Soc 1997; 80:879e85. [47] Kondo N, Ohji T, Wakai F. Strengthening and toughening of silicon nitride by superplastic deformation. J Am Ceram Soc 1998;81: 713e6. [48] Wu X, Chen IW. Exaggerated texture and grain growth in a superplastic SiAlON. J Am Ceram Soc 1992;75:2733e41. [49] Kondo N, Ohji T, Wakai F. Deformation conditions of b-SiAlON to achieve large superplastic elongation. J Ceram Soc Jpn 1998;106: 1040e2. [50] Rouxel T, Rossignol F, Besson JL, Goursat P. Superplastic forming of an a-phase rich silicon nitride. J Mater Res 1997;12:480e92. [51] Chen IW, Hwang SL. Shear thickening creep in superplastic silicon nitride. J Am Ceram Soc 1992;75:1073e9. [52] Rosenflanz A, Chen I-W. “Classical” superplasticity of SiAlON ceramics. J Am Ceram Soc 1997;80:1341e52. [53] Xie RJ, Mitomo M, Zhan GD. Superplasticity in a fine-grained beta-silicon nitride ceramic containing a transient liquid. Acta Mater 2000;48:2049e58. [54] Xu X, Nishimura T, Hirosaki N, Xie RJ, Yamamoto Y, Tanaka H. Superplastic deformation of nano-sized silicon nitride ceramics. Acta Mater 2006;54:255e62. [55] Nishimura T, Xu X, Kimoto K, Hirosaki N, Tanaka H. Fabrication of silicon nitride nanoceramics e powder preparation and sintering: A reiview. Sci Tech Adv Mater 2007;8:635e43. [56] Shen Z, Peng H, Nygren M. Formidable increase in the superplasticity of ceramics in the presence of an electric field. Adv Mater 2003;15:1006e9.

Handbook of Advanced Ceramics

[57] Mele´ndez-Martinez JJ, Domı´nguez-Rodrı´guez A. Creep of silicon nitride. Prog Mater Sci 2004;49:19e107. [58] Raj R, Chyung CK. Solution-precipitation creep in glass ceramics. Acta Mater 1981;29:159e66. [59] Wakai F. Step model of solution-precipitation creep. Acta Metall Mater 1994;42:1163e72. [60] Shinoda Y, Nagano T, Gu H, Wakai F. Superplasticity of silicon carbide. J Am Ceram Soc 1999;82:2916e8. [61] Gu H, Shinoda Y, Wakai F. Detection of boron segregation to grain boundaries in silicon carbide by spatially resolved electron energyloss spectroscopy. J Am Ceram Soc 1999;82:469e72. [62] Ohtsuka S, Shinoda Y, Akatsu T, Wakai F. Effect of oxygen segregation at grain boundaries on deformation of B, C- doped silicon carbides at elevated temperatures. J Am Ceram Soc 2005;88:1558e63. [63] Tokiyama T, Shinoda Y, Akatsu T, Wakai F. Enhancement of hightemperature deformation in fine-graine silicon carbide with Al doping. Mater Sci Eng B 2008;148:261e4. [64] Wakai F, Kodama Y, Sakaguchi S, Nonami T. Superplasticity of hot isostatically pressed hydroxyapatite. J Am Ceram Soc 1990;73:457e60. [65] Singh D, De La Cinta Lorenzo-Martin M, Routbort JL, Gutie´rrezMora F, Case ED. Plastic deformation of hydroxyapatites and its application to joining. Appl Ceram Tech 2005;2:247e55. [66] Tago K, Itatani K, Suzuki TS, Sakka Y, Koda S. Densification and superplasticity of hydroxy apatite ceramics. J Ceram Soc Jpn 2005;113:669e73. [67] Adachi M, Wakamatsu N, Doi Y. Superplastic deformation in carbonate apatite ceramics under constant compressive loading for near-net-shape production of bioresorbable bone substitutes. Dent Mater J 2008;27:93e8. [68] Albuquerque JM, Harmer MP, Chou YT. Tensile superplastic deformation of YBa2Cu3O7 x high-Tc superconductor. Acta Mater 2001;49:2277e84. [69] Panda PC, Wang J, Raj R. Sinter-forging characteristics of finegrained zirconia. J Am Ceram Soc 1988;71:C507e9. [70] G-Zhan D, Garay JE, Mukherjee AK. Ultralow-temperature superplasticity in nanoceramic composites. Nano Lett 2005;5: 2593e7. [71] Hulbert DM, Jiang D, Kuntz JD, Kodera Y, Mukherjee AK. A lowtemperature high-strain-rate formable nanocrystalline superplastic ceramic. Scripta Mater 2007;56:1103e6. [72] Lesuer DR, Wadsworth J, Nieh TG. Forming of superplastic ceramics. Ceram Int 1996;22:381e8. [73] Wittenauer J, Nieh TG, Wadsworth J. Superplastic gas-pressure deformation of YTZ sheet. J Am Ceram Soc 1993;76:1665e72. [74] Wu X, Chen I-W. Hot extrusion of ceramics. J Am Ceram Soc 1992;75:1846e53. [75] Winnubst AJA, Boutz MMR. Superplastic deep drawing of tetragonal ceramics at 1160 C. J Eur Ceram Soc 1998;18:2101e6. [76] Wu X, Chen I-W. Superplastic bulging of fine-grained zirconia. J Am Ceram Soc 1990;73:746e9. [77] Nagano T, Kato H, Wakai F. Diffusion bonding of zirconia/alumina composites. J Am Ceram Soc 1990;73:3476e80. [78] Gutie´rrez-Mora F, Goretta KC, Majumdar S, Routbort JL, Grimdisch M, Domı´nguez-Rodrı´guez A. Influence of internal stresses in superplastic joining of irconia-toughened alumina. Acta Mater 2002;50:3475e85.

Chapter | 9.9

Development of Superplastic Ceramics

[79] Boniecki M, Kalinski D, Librant Z, Wesolowski W. Superplastic joining of alumina and zirconia ceramics. J Eur Ceram Soc 2007;27:1351e5. [80] Singh D, de la Cinta Lorenzo-Martin M, Gutie´rrez-Mora F, Routbort JL, Casa ED. Self-joining of zirconia/hydroxyapatite composites using plastic deformation process. Acta Biomater 2006;2:669e75.

771

[81]

[82]

Nagano T, Wakai F. Fabrication of zirconia-alumina functionally gradient material by superplastic diffusion bonding. J Mater Sci 1993;28:5793e9. Goretta KC, Singh D, Chen N, Gutierrez-Mora F, de la Cinta Lorenzo-Martin M, Dominguez-rodriguez A, et al. Microstructure and properties of ceramics and composites joined by plastic deformation. Mater Sci Eng A 2008;498:12e8.

Chapter 10.1

Joining Ceramics and Metals Katsuaki Suganuma Institute for Scientific and Industrial Research, Osaka University, 8-1 Mihogaoka, Ibaraki, Osaka 567-0047, Japan

Chapter Outline 1. Introduction 2. Interface Chemistries 3. Physical Contact at Interface 4. Surface Roughness and Damage of Bond Face of Ceramics

775 776 779

5. Thermal Stress 6. Joining Process 7. Summary References

780 782 787 788

780

1. INTRODUCTION In many applications of ceramics in electronics and structural engineering fields, ceramics are required to be joined to dissimilar materials, especially metals. For the electronics applications, for instance, alumina (Al2O3) ceramics have been widely used as insulating materials for many decades, represented by printed circuit boards, on which metallic wirings and electrodes are patterned with tight interfaces. Nowadays, functional substrates, such as aluminum nitride (AlN) and low-temperature co-fired ceramics (LTCC), have been also used widely. On the other hand, a variety of ceramic sensors for gas, pressure, magnetic force, temperature, fire, and others have been applied connected with metallic electrodes. As structural materials, oxide ceramics such as partially stabilized zirconia (psz, for instance, ZrO2eY2O3) and nonoxide ceramics such as Si3N4 have been successfully applied to cutting tools and to automotive engine components replacing the conventional metallic materials since the beginning of the 1980s. Figure 1 shows typical ceramic parts established in the commercial fields. Although many methods have been already established for joining ceramics and metals, one has to understand the fact that most of ceramic/metal joint structures are unstable because of the large gaps both in chemical and in physical nature between two materials. There is always some sort of defect structures in an interfacial region. They are, for example, a brittle reaction layer, voids, unjoined area, microcracks from thermal stress, and thermal residual stress. Therefore, in a practical application of ceramic/metal system using

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00042-3 Copyright Ó 2013 Elsevier Inc. All rights reserved.

joining technology, one has to take the influence of all defect structures in a joint into account and should design an optimized joint structure which can compensate for the negative influence from these defects. Ceramic/metal joints usually fracture in a brittle manner similar to ceramics and, then, the scatter in strength can be treated by the Weibull statistics. The statistic treatment of strength data can provide us useful information when one tries to grasp suitable joining parameters. There are many defect categories that may cause scatter in strength. From a microscopic view, interface contact formed by wetting, and chemical and physical reaction at interfaces should be of concern in the first place. The nature of atomic bonding, which is on a nanometer scale, may not influence scatter in strength directly but does on the absolute strength because the defect size determining the scatter in strength seems to be beyond a few mm in practical cases. From the macroscopic view, when a reaction layer grows thick, cracking in the layer frequently reduces joint strength. Thermal or residual stress in a joint becomes the other important factor. Large thermal stress both in joining process and in services induces flaws into joints. The final goal for joining research will be in establishing a technique that produces a reliable strong interface by eliminating these defects and by accommodating thermal stress. The present chapter focuses on reviewing the factors affecting the structural integrity of a ceramic/metal joint. Especially, processing parameters influencing structural reliability are discussed. Processing parameters such as interface reaction are considered initially. The next topic is

775

776

Handbook of Advanced Ceramics

FIGURE 1 Si3N4 products for automobile engine components (Courtesy to Nissan Motor, and NTK Co. Ltd.). The arrows point to Si3N4 brazed to steel. For color version of this figure, the reader is referred to the online version of this book.

the physical contact and the bond face damage effect. Then the effect of thermal stress on strength is discussed. In the last section, practical joining methods, from the conventional to relatively new, are introduced.

2. INTERFACE CHEMISTRIES Interface formation of a ceramic/metal system has been discussed for many decades on the basis of reaction chemistry [1]. Interface characterization has been carried out with various microstructural observation methods and, especially, with the wetting experiment with the assistance of thermodynamic considerations. Successful results have been achieved from such semi-empirical works and the active metal brazing method is one of the fruitful establishments. In recent years, the research has directed toward the nanometer scale phenomena, i.e., lattice structure observed by highresolution transmission microscopy [2], atomic binding at dissimilar materials interface evaluation with ESCA/Auger analysis assisted with the first principle/molecular dynamics simulation [3], and even wetting phenomena at ceramic/ metal interfaces [4]. These works provide the scientific understanding of ceramic/metal interfaces and the comprehensive explanations for controlling process parameters. The interface chemistry has a great influence on the reliability of a ceramic/metal joint by changing the uniformity of microstructure resulting in increasing scatter in strength. If homogeneous bonding is achieved over the entire

bond face, the scatter in strength is expected to be small. Then, from the viewpoint of the scatter in strength, one should think about the homogeneity of interface bonding and this is influenced primarily by reaction and wetting behavior. In the metal/metal brazing systems, it is relatively easy to get homogeneous wetting between metal substrates and braze metals if materials and atmosphere are selected suitably. In most of the ceramic/metal brazing systems, it is possible to obtain good wetting by using the active metal brazes [5]. One of the typical active metal brazes is AgeCu eutectic braze metal with a small amount of Ti. The addition of Ti is between 1 and 4 wt%. Even though an increased amount of Ti makes good wetting on ceramics, the brazing metals become very brittle resulting in degradation of joint strength. The brazing atmosphere is also important to prevent severe oxidation of the active metals. The vacuum under 10 2 Pa order is required to achieve good wetting. Figure 2 shows a typical interface microstructure of Si3N4/AgeCueTi. At the interface, thin TiN/M(Si, Cu, Ti)6N reaction layers are formed [6]. If ceramics are oxides or carbides, they are TiO/ M(Si, Cu, Ti)6O or TiC/M(Si, Cu, Ti)6C, respectively. The Al-based alloys are also regarded as one of the active metals since Al itself is quite reactive on most of ceramics. However, since Al can form a stable oxide skin, it sometimes prevents brazes from achieving good wetting on substrates. Figure 3 shows Weibull plots of the Si3N4 joints brazed with pure Al and Al-X (X ¼ Si, Mg) binary dilute alloys [7]. The scatter in strength of the Si3N4/Al/Si3N4

Chapter | 10.1

Joining Ceramics and Metals

777

FIGURE 2 Interface of active metal brazed Si3N4 (SEM).

FIGURE 3 Weibull plots of Si3N4 joints brazed with Al joint strength [7].

joint is small and the addition of a small amount of Si does not influence it so much. However, a similar amount of Mg addition to the Al braze makes the scatter very large. These changes in strength are closely related to the change in fracture mode. The addition of Mg increases interface

fracture. From TEM observation, it was found that the Mg addition changes the interface microstructure. Figure 4 shows a schematic illustration of the Si3N4/Al alloy interface and the fracture path. When Mg is not in the braze alloy, a nano-crystalline b’-sialon layer and an AleSieO

778

Handbook of Advanced Ceramics

FIGURE 5 Weibull plots of bending strength of Si3N4 joints with or without additive brazed with Al [8].

FIGURE 4 Schematic of Si3N4 joint interfaces brazed with Al dilute alloys [7].

amorphous layer are uniformly formed at the interface. In the case of the Si3N4/AleMg interface, on the other hand, Mg concentrates at the interface and oxide products, i.e., a-Al2O3 and g-Al2O3 with Si, are discretely formed along the interface. The fracture path primarily runs along the oxide layer. Thus, the addition of a small amount of Mg to the Al braze promotes the formation of crystalline Al2O3. This microstructural change increases the scatter in strength of the joint drastically. The reason for the increased scatter in the presence of the stable crystalline oxide products can be ascribed primarily to the weak bonding between the Al oxides and Al/Si3N4. The presence of sintering additives in ceramics may also influence the interface chemistry and then the reliability. For instance, Si3N4 with or without the sintering additives Al2O3 and Y2O3 was brazed with pure Al and the strength of the joint was found to depend on the presence of the additives [8]. TEM observation revealed that the Si3N4 with the additives

formed a thick AleSieO layer reaching 1000 nm, while that without any additive formed a 400-nm-thick layer. Figure 5 shows the Weibull plots of strength. The joint of Si3N4 with the additive has higher strength and smaller scatter in strength than the joint without any additive. XPS analysis showed that fracture of the additive-free Si3N4 joint occurred at the Al/reaction layer interface, while the joint with the additive fractured in the Al layer. These results imply that the thick AleSieO layer can cover the interface between the substrate Si3N4 and the Al braze layer much more uniformly than the thin formation of the layer. This microstructural difference seems to result in changing the scatter in strength between the two cases. A similar result has been obtained by a pre-oxidation treatment of Si3N4 before brazing [9]. The oxidized treatment of Si3N4 promotes wetting between Si3N4 and Al resulting in decreasing the unjoined area. The oxidized interface has a thick AleSieO layer as one of the two reaction layers [10]. It has been known that the ion mixing of a ceramic/ metal interface can provide good interface adhesion [11,12]. Peteves reported the influence of ion beam mixing on scatter in strength of the Cr-coated Si3N4/NieCr alloy joint [12]. By Xe ion irradiation on the Cr-coated surface, a two-fold improvement in Weibull modulus was achieved with the improvement of the absolute strength. The improvement of the adhesion of Cr film to Si3N4 can be attributed to the chemistry change due to the enhanced diffusion, and the annealing out of some of the residual surface damage of Si3N4.

Chapter | 10.1

Joining Ceramics and Metals

3. PHYSICAL CONTACT AT INTERFACE In the practical joining sequences, a perfect interface adhesion over the entire interface is hardly achieved within a limited joining period and temperature. Then the initial surface roughness and the applied pressure have two of the critical parameters which have great influences on achieving interfacial contact not only in solid-state bonding but also in brazing. In solid-state bonding, interfacial contact is promoted by plastic deformation in the early stage followed by creep deformation and diffusion in the later stage. Pressure across the joint interfaces primarily influences achieving contact by plastic deformation in the first stage. Unjoined islands are inevitably formed on the interface under a limited pressure. It is dependent on the amplitude of pressure, period, temperature, and various material factors such as flow/ creep stress. Figure 6 shows the relationship between the fracture stress and the unjoined area of the solid-state bonded Al2O3/Nb joint [13]. Apparently, the increase in unjoined area decreases the joining strength. The importance of the contact pressure during joining on the scatter in strength has been reported by Gottslig et al. [14] They showed that the high-pressure joining promotes the substantial reduction of the scatter in strength. If interface reaction releases gas as the reaction product, the pores filled with the gas may be left on the interface resulting in the inhibition of contact. The Si3N4/ Ni interface is one of the cases. This interface is weak due to the presence of pores along the interface [15]. When Ni contains nitride-forming elements such as Cr, no pore is formed at an interface and the strength is improved [12,16]. Unjoined area is frequently formed at the edge of a joint. This edge defect weakens the joint extremely as it

FIGURE 6 Bending strength of individual Al2O3/Nb joints as a function of unjoined area formed on interface [13].

779

works as a notch induced on the interface and one must control the formation of edge unjoined band. The inhomogeneity in deformation of the metal layer will also reflect the strength. In the case of the reaction gas-releasing system, the reaction in the outer region may be promoted by continuous evacuation [16]. This will cause excess thinning of the ceramic at the edge region. In brazing, pressure has a great effect on strength and its scatter. The changes in the joint layer thickness and its uniformity seem to be responsible for the influence. Figure 7 shows the effect of brazing pressure on the strength of the Si3N4 joint with an Al braze [17]. Only slight applied pressure is enough to produce a sound joint with little scatter in strength. Johnson reported the influence of the braze layer thickness effect on the strength of the Si3N4 joint brazed with the AgeCueTi braze [18]. The strength of the joint was proportional to the inverse of the square root of joint thickness when the thickness was below 50 mm. This indicates that the joint layer becomes one of the defects of which size is taken as the thickness. Thus, from the processing viewpoint, an appropriate pressure both for solid-state bonding and for brazing should be determined together with the other processing and material parameters. To remove the edge unjoined region, hydrostatic pressing such as HIPing becomes one of the powerful methods. The rough bond face of a metal also has some effect.

FIGURE 7 Effect of applied pressure on bending strength of Si3N4 joints brazed with Al [17].

780

Handbook of Advanced Ceramics

4. SURFACE ROUGHNESS AND DAMAGE OF BOND FACE OF CERAMICS Surface roughness has three basic effects on the strength of a ceramic/metal joint. Rough bond face will prevent completing contact at an interface under a limited pressure. Rough bond face may have the damaged surface layer of ceramics which has deep scratches and residual stress. On the other hand, an irregular bond face may have an anchoring effect which promotes joining by mechanical interlocking. In the actual joining cases, these effects influence on the mechanical properties of a joint in a competing way. Figure 8 shows the Weibull plots of the strength of the Si3N4 joint brazed with pure Al varying the surface grinding condition [19]. Clearly, the rougher bond face made the joint weaker. This degradation can be attributed to the fact that the roughly ground bond face has a damage layer remaining in the joint even after the joining treatment. The fracture of the joint occurs in the damaged layers along the joining interface. Thus, a roughly ground bond face can weaken the ceramic/metal joint if the damaged layer remains in the joint. The surface finishing methods also have important influence on the roughness effects. To make a certain rough surface by polishing is one of the promising conditions for metaleceramic interface [20].

FIGURE 9 Effect of size and shape of bond face on residual stress of Si3N4/invar alloy joints [21]. The residual stress was vertical to the interface on the Si3N4 surface.

The distribution of thermal (or residual) stress is not uniform in a ceramic/metal joint even along the interface.

Concentration of thermal stress becomes more severe with proximity to the interface and to the free surface. The most harmful effect on joint strength is caused by tensile stress at an interface or in a ceramic. The maximum tensile stress appears near the edge of an interface and on the free surface. Since this stress acts almost vertical to the interface, the apparent bonding strength measured by tensile or bending tests is substantially reduced. The amplitude of residual stress depends on the shape and dimension of the interface [21]. Figure 9 shows the diameter dependence of the thermal stress of the Si3N4/invar alloy joint measured on the surface near the interface. The larger diameter makes the larger residual stress. It is also noteworthy that stress concentration at the corner of the rectangular bond face joint is more serious.

FIGURE 8 Influence of surface roughness of a bond face on bending strength of Si3N4 joints brazed with Al [19].

FIGURE 10 Weibull plots of bending strengths of Si3N4/invar and Si3N4/kovar joints brazed with Al [22].

5. THERMAL STRESS

Chapter | 10.1

Joining Ceramics and Metals

TABLE 1 Interlayer Structures for Ceramic/Metal Joining Systems [1] Interlayers

Examples

Soft metal

ceramics/(Al, Cu, or Ni)/metals

Soft metal/hard & low expansion metal laminate layer

Nb/Mo for Al2O3/steel Fe/W for Si3N4/steel Ni/W/Ni for Si3N4/steel

Grading layer(FGM)

Al2O3/Al2O3eFe grading/Fe

The joint with large thermal expansion mismatch decreases strength. However, it occasionally happens that some specimen is strong but the other is weak even if they are the same kind. This depends on the presence and distribution of internal flaws induced by thermal stress during joining treatment. For example, the strengths of the Si3N4/invar and Si3N4/kovar joints are shown in Figure 10 [21]. The latter joint had larger thermal stress than the former as seen in Figure 9. The Si3N4/invar joint exhibits good strength with small scatter. On the other hand, the Si3N4/kovar joint shows large scatter in strength. The distribution of the strength is divided into two parts, i.e., high strength with small scatter and low strength with large scatter. The latter joint always has a large interfacial flaw formed on cooling from the joining temperature. Thus,

781

large thermal stress occasionally induces serious flaws into a joint, which not only weaken the joint but also may make the scatter in strength large. In other words, it is very important to evaluate the scatter in strength, especially for joints with large expansion mismatch. Thus, it is important to relax thermal stress effect. A key point is in reducing the tensile part of residual stress in a ceramic and on an interface because ceramics are weak against tensile stress while the tensile stress in a metal could be consumed by elastic and plastic deformation. A wide variety of thermal stress-relaxation methods have been developed by using certain interlayers for many ceramic/metal systems, for example, soft metal interlayers, composite interlayers, laminate interlayers (soft metal/low expansion and hard metal), and fine crack interlayers. Table 1 is the summary of those interlayers [1]. The soft metal interlayer such as Al and Cu reduces the residual stress by elastic, plastic, and creep deformation. However, such a single layer cannot remove the residual stress effectively, especially for silicon ceramics/metal systems. In addition, it cannot resist to sudden temperature change or severe heat cycle because plastic/creep deformation cannot follow sudden temperature change. Suganuma et al. first proposed the effectiveness of the laminate interlayer that improves resistance to sudden temperature change [23]. Since the soft metal is restricted with a ceramic and a low expansion/hard metal from both sides, it can easily deform following the shrinkage of the ceramic. In addition, the harmful effect of the large expansion/ shrinkage of a base metal is blocked by the hard metal layer. Figure 11 shows the effectiveness of Nb/Mo

FIGURE 11 Heat cycle resistance of Al2O3/Steel joint with various interlayers [23].

782

Handbook of Advanced Ceramics

TABLE 2 Comparison of Various Joining Methods Category

Methods

Benefits

Drawbacks

Mechanical joining

Bolting, Shrunk fit, Cast molding

Easiness Cheapness

Low strength Poor reliability

Adhesive joining

Organic adhesives, Inorganic adhesives (Cements)

Easiness Cheapness

Low strength Poor reliability Vacuum tightness

Brazing Soldering

Brazing (active metal brazing, brazing after metallization) Lead-free soldering

Excellent reliability Easiness

Heat-resistivity

Welding

Gas welding, Arc welding, Electron beam welding, Laser welding, resistance welding

High strength

Grain growth Heat-shock damage

Solid-state bonding

Hot-pressing, HIPing, Eutectic bonding, Room-temperature bonding, Friction welding, Field-assisted bonding

High strength Excellent heat resistance

Limited shape High cost

Others

SQ bonding

Excellent strength Irregular interface application Easiness Cheapness

Al only

interlayer to thermal cycle resistance of the laminate interlayer applied for the Al2O3/stainless steel joint. A single interlayer cannot maintain initial strength after only one cycle between room temperature and 500  C. The Nb/ Mo laminate layer keeps the initial strength even after 100 cycles. The laminate interlayer method, however, still has a limit in bonding size. A large area bonding is still one of the serious matters because the thermal stress increases as increasing bond area. A composite interlayer, which has a long history socalled as “the powder-graded seal,” “the composite interlayer,” or “the functionally gradient materials,” is one of the effective methods to compensate thermal stress. However, the main problem is strength and reliability of the interlayer itself. Cermet-type composites are generally very brittle, resulting in poor joining strength.

6. JOINING PROCESS Table 2 summarizes a brief comparison of variety of joining methods for ceramics/metals. Mechanical joining and adhesives with organics and cements have been widely used because of their easiness and inexpensiveness. Mechanical joining has a variety of processes. For example, bolting and clamping are the simplest methods. The shrunk-in inserts were applied for the part of production of the turbo-charger rotor [24]. The rocker arm chip made of

Si3N4 was inserted in an Al alloy die-cast arm (Figure 1) [25]. The mechanical joints can have heat resistance up to about 500  C. For the adoption of these methods, it is important to note that the mechanical joining sometimes accompanies a severe stress concentration and that ceramics are weak especially for tensile stress. Although the conventional organic adhesives can produce low strength below 20 MPa generally, the organic adhesive, which has an adhesive tensile strength of about 80 MPa, is commercially available. Fusion welding can produce stable interfacial structures up to elevated temperatures, i.e., melting temperature. However, the problems of grain growth, formation of pores, and thermal stress, in addition to thermal shock, become serious. Friction welding is also one of the simple processes, but effective [26]. Joints can have a tensile strength beyond 100 MPa. However, because the joining torque is limited to be low in order to prevent damage in a ceramic, only soft metals, such as Al, can be selected as a metallic constituent.

6.1. Active Metal Brazing and MoeMn Method Brazing and solid-state joining have excellent potential for structural purposes because they can supply both good strength and heat resistance. Research, however, has been focused on oxide ceramics, especially for

Chapter | 10.1

Joining Ceramics and Metals

application to electrical and electronic components for many years. Metallizations, such as the so-called MoeMn method only for Al2O3, are one of the most common processes in industries and have been widely used in the actual insulator productions. Although the MoeMn method consists of somewhat complex steps as shown in Figure 12, it is well established. The bonding strength is high depending on brazing alloys. Recent structural applications of ceramics require different types of joining techniques not only for oxide ceramics but also for nonoxide ceramics in order to produce the stronger and more reliable ceramic/metal joints, which are sometimes expected to have heat resistance superior than those obtained by the conventional metallization processes. Active metal brazing, which is mentioned in the previous sections, is one of the best choices as a brazing method. There are several

FIGURE 12

783

process choices for the active metal brazing as shown in Figure 13. In each process, vacuum or reduction joining atmosphere is required to achieve tight interface bonding in order to form a semi-metallic layer such as TiN, TiC, or TiO, which is essential in making an ideal bridge layer between ceramic and metal, at the interface.

6.2. Eutectic Liquid Bonding The Cu/Al2O3 joint structure has been successfully applied as shear-resistant electronic circuits. One of the hottest applications is the direct bonded copper (DBC) substrate, which has been used for power semiconductors. Cu and Al2O3 can be bonded in the temperature range just beyond 1063  C, at which Cu forms a thin eutectic liquid film on its surface in the presence of controlled oxygen atmosphere [27]. The eutectic CueO liquid promotes wetting. The CuxOeAg system also provides a good wetting for oxide

Mo-Mn metallization process. For color version of this figure, the reader is referred to the online version of this book.

784

Handbook of Advanced Ceramics

FIGURE 13 Active metal brazing processes.

ceramics beyond 930  C resulting in the formation of the intimate interface [28]. Some of transition metals such as Fe, Ni, and Co have eutectic reaction between Si3N4 and SiC. Figure 14 shows the interface between Si3N4 and Ni joined at 1300  C [29]. At the interface, Si3N4 contacts Fe directly after decomposition of Si3N4 into Si and N. Si atoms diffuse into Fe during joining, while N atoms go into atmosphere.

6.3. Room-Temperature Bonding (SAB: Surface-activated bonding) If fresh surfaces of metals and ceramics mate without any contamination, they can form tight bonding even at room temperature. Room-temperature bonding utilizes ion bombardment to remove surface contamination and oxidation followed by intimate contact in low vacuum. Al, Au, and Cu can be successfully bonded to various

FIGURE 14

kinds of ceramics [30]. In most cases, direct atomic contacts are formed at interfaces and the bonding strength reaches the parent body. Si and glass are joined for MEMS applications by using the SAB method as shown in Figure 15.

6.4. SQ Process The author has developed a cast bonding process for joining ceramics and metals [31]. This method utilizes a pressure casting of Al and its alloys and is called as the SQ process (squeeze casting). The basic concept of the SQ process is schematically shown in Figure 16. In this process, ceramic components are heated up to 500  Ce800  C in a mold. Liquid Al in the temperature range between 700  C and 800  C is poured into the mold and pressure is applied immediately. Al liquid rapidly solidifies under a pressure and tight bonding at the interface is accomplished within a few

Eutectic liquid bonded Si3N4/Ni interface [29].

Chapter | 10.1

Joining Ceramics and Metals

785

FIGURE 15 Glass/Si joint sensor processed by room temperature bonding. For color version of this figure, the reader is referred to the online version of this book.

seconds. This process has many advantages for the production of ceramic/metal joint structures such as no void formation, no brittle IMC formation, low cost without vacuum atmosphere, short tact time, applicability to irregular interfaces, etc. The successful application is the DBA substrate shown in Figure 17 [32]. The strength of the joint interfaces reaches the strengths of Al and alloys.

6.5. Soldering For joining glasses and metals, soldering has been widely used with high-Pb solder with the aid of ultrasonic vibration [33]. One of the applications is the hermetic edge seal of a pair of glass substrates for buildings and houses. Glazing with glasses of low-temperature melting points,

FIGURE 16 Schematic illustration of SQ process and Si3N4/stainless steel joint with Al interlayer [32]. For color version of this figure, the reader is referred to the online version of this book.

786

Handbook of Advanced Ceramics

FIGURE 17 DBA substrate fabricated by SQ method and thermal shock test result [32].

i.e., Pb-based glasses, has been also widely used by filling and joining very small gaps with large sealing lengths [34]. In the current movement of lead-free technologies in industries, joining glasses also required to be lead free. The author has developed a novel process for hermetic seal of paired glass substrates [35]. Ti-doped SneZn eutectic solder was developed for sealing glasses. The sealing process is based on friction-activated capillary

action. The sealing machine is equipped with a solder container and a solder outlet with a friction plate as shown in Figure 18. Sealing speed is as high as 10 m/s.

6.6. Field-Assisted Bonding Field-assisted bonding has a long history for glass or b-Al2O3 to metal bonding [36,37]. Bonding is accomplished

Chapter | 10.1

Joining Ceramics and Metals

787

FIGURE 18 Friction-activated capillary soldering [35]. For color version of this figure, the reader is referred to the online version of this book.

beyond 300  C in about 1 min by applying a DC voltage in 100 to a few hundred volts between the glass and the metals. Under this condition, cations become quite mobile and can form a tight interface. One of the limiting factors is the flatness of the surfaces to be bonded. They must be polished to a roughness below 100 nm. The field-assisted bonding method has been already applied to MEMS applications.

7. SUMMARY This chapter has focused on joining ceramics and metals. The influences of several important processing factors on the reliability of joints are summarized. The interface formation methods and the relaxation of stress inside the joints are two of the essential parameters in order to obtain

high-quality joints. In practical application, a ceramic/metal joined component is used not only under an external stress but also under the internal stress originating from the elastic and expansion mismatch. Since most of engineering ceramics such as Si3N4 and SiC usually suffer from severe thermal stress during operation because of their extremely small thermal expansion coefficients against metal components, thermal stress is one of the most dominant factors that restricts the lifetime of joints. There are other important factors in determining the lifetime, i.e., environmentally induced degradation such as oxidation, corrosion, stressassisted corrosion, and also fatigue. Further intensive works are required for individual applications of ceramics by understanding of the mechanisms by which joint strength decreases in certain environments and also for establishing

788

the design technology for a ceramic/metal joining system. Such efforts are expected to accompany both the development of new techniques and the refinements of existing techniques.

REFERENCES [1] For examples Nicholas MG, et al. Ceramic/metal joining for structural applications. Mater. Sci. Technol., 1 1985;657. K.Suganuma, et al, “Joining of ceramics and metals”, Ann. Rev. Mater. Sci., 18(1988), 47. [2] For example LeLay G. Physics and electronics of the noble-metal/ elemental-semiconductor interface formation: a status report. Surface Sci. 1983;132:169. M. Ruhe and A.G. Evans, "Structure and chemistry of metal/ceramic interfaces", Mater. Sci. Engineer., A107, 187(1989). [3] For example Ohuchi FS, Kohyama M. Electronic structure and chemical reactions at metal-alumina and metal-aluminum nitride interfaces. J. Am. Ceram. Soc. 1991;74:1163. [4] Saiz E, Tomsia AP. Kinetics of high-temperature spreading. Current Opinion in Solid State and Materials Science 2005;9(4e5):167. [5] McDonald JE, Eberhart JG. Adhesion in aluminum oxide-metal systems. Trans. AIME 1965;233:512. [6] Carim AH. Transitional phases at ceramicemetal interfaces: orthorhombic, cubic, and hexagonal TieSieCueN compounds. J. Am. Ceram. Soc. 1990;73:2764. [7] Ning XS, et al. Bond strength and interfacial structure of silicon nitride joints brazed with aluminum-silicon and aluminummagnesium alloys. J. Mater. Sci. 1991;26:2050. [8] Ning XS, et al. Effect of an oxide-additive in silicon nitride on interfacial structure and strength of silicon nitride/Al/silicon nitride joints. J. Mater. Sci. 1989;24:2865. [9] Morita M, et al. Effect of pre-heat-treatment on silicon nitride joining with aluminium braze. J. Mater. Sci. Letters 1987;6:474. [10] Ning XS, et al. Interfacial strength and chemistry of additive-free silicon nitride ceramics brazed with aluminium. J. Mater. Sci. 1989;24:2879. [11] Noda S, et al. Improvement for adhesion of thin metal films on ceramics by ion bombardment and the application to metaldceramic joining. J. Mater. Sci. Letters 1986;5:381. [12] Peteves SD. Effects of ion implantation on the mechanical properties of silicon nitride ceramics. Mater. Sci. Mongr. 1991; 66B:1469. [13] Suganuma K, et al. Effect of surface grinding condition on strength of alumina/niobium joint. Ceram. Eng. Sci. Proc. 1989;10:1919. [14] Gottselig B, et al. Joining of ceramics demonstrated by the example of SiC/Ti. J. European Ceram. Soc. 1990;6:153. [15] Brito ME, et al. Si3N4eNi bonding and its reaction. In: Doyama M, Somiya S, editors. Advanced Materials, vol. 8; 1989. p. 23e8. Mater. Res. Soc. International Meeting, Tokyo.

Handbook of Advanced Ceramics

[16] Nakamura M, Peteves SD. Solid-state bonding of silicon nitride ceramics with nickelechromium alloy interlayers. J. Am. Ceram. Soc. 1990;73:1221. [17] Suganuma K, et al. Joining of silicon nitride to silicon nitride and to invar alloy using aluminium interlayer. J. Mater. Sci. 1987; 22:1359. [18] Johnson SM. Mechanical behaviour of brazed silicon nitride. Ceram. Eng. Sci. Proc. 1989;10:1846. [19] Suganuma K, et al. Effect of surface damage on strengths of silicon nitride bonded with aluminum. Adv. Ceram. Mater. 1986;1:356. [20] Beraud C, et al. Study of copper-alumina bonding. J. Mater. Sci. 1989;24:4545. [21] Suganuma K, et al. Influence of shape and size on residual stress in ceramic/metal joining. J. Mater. Sci. 1987;22:2702. [22] Suganuma K, et al. Acoustic emission from ceramic/metal joints on cooling. Am. Ceram. Soc. Bull. 1986;65:1060. [23] Suganuma K, et al. Solid-state bonding of oxide ceramic to steel. J. Nucl. Mater. 1985;133&144:773. [24] Katayama K, et al. Development of Nissan high response ceramic turbocharger rotor. SAE Technical Paper Series 1986. 861128. [25] Ogawa Y, et al. Ceramic rocker arm insert for internal combustion engines. SAE Technical Paper Series 1986. 860397. [26] Suzumura A, et al. Friction welding of aluminum to steel. KouonnGakkai-Shi 1987;13:43. [27] J.F.Burgess, C.A. Neugebauer, “The direct bonding of metals to ceramics and application in electronics”, U.S. Patent No.3744120, (1973), or Electrocomponent Science and Technology, 233(1976). [28] Kim JY, et al. Effects of CuO content on the wetting behavior and mechanical properties of a AgeCuO braze for ceramic joining. J. Am. Ceram. Soc. 2005;88(9):2521. [29] Suganuma K, et al. Solid-state bonding of silicon nitride with nickel interlayer. J.Less.Common.Metals 1989;158:59. [30] Shigetou A, et al. Bumpless interconnect through ultrafine Cu electrodes by means of surface-activated bonding (SAB) method. IEEE Trans. Adv. Packag. 2006;29(2):218. [31] Suganuma K. New process for brazing ceramics utilizing squeeze casting. J.Mater.Sci. 1991;26:6144. [32] Ning XS, et al. Interface of aluminum/ceramic power substrates manufactured by casting-bonding process. Mat. Res. Soc. Symp. Proc. 1997;vol. 445:101. [33] Nagano K. Reports Res. Lab. Asahi Glass Co., Ltd. 1971; 21(2):177. [34] Collins RE, et al. Vacuum glazingeA new component for insulating windows. Building and Environment 1995;30(4):459. [35] Sakaguchi K, et al. New glass sealing method with lead-free solders. Advances in Science and Technology 2006;45:1604. [36] Wallis G. Field assisted glass-metal sealing. J. Appl. Phys. 1969;40(10):3946. [37] Dunn B. Field-assisted bonding of beta-alumina to metals. J. Am. Ceram. Soc. 1979;62:545.

Chapter 10.2

Heat-Resistant Coating Technology for Gas Turbines Yoshiyasu Ito Tocalo Co. Ltd., 4-13-4 Fukaekita-machi, Higashinada, Kobe, Hyogo 658-0013, Japan, Former Affiliate: R&D Center, Power and Industrial Systems, Toshiba Corp.

Chapter Outline 1. Introduction 2. Development of Heat-Resistant Superalloys for Gas Turbines 3. Development of Heat-Resistant Coatings for Gas Turbines

789 790 792

1. INTRODUCTION Because Japan relies on imports from abroad for much of its coal, petroleum oil, natural gas, and uranium, the country’s self-sufficiency in terms of energy supply, including hydroelectric power, is only 4%, which is significantly lower than other countries. On the other hand, considering that energy consumption is increasing worldwide, Japan, having such low-energy self-sufficiency, will be greatly affected by world energy consumption, which may threaten Japan’s independency as a nation. Figure 1 shows a long-term outlook of energy demand up to the year 2030, provided by the International Energy Agency (IEA). According to Figure 1, the world energy consumption in 2009 temporarily decreased for the first time since 1981, and the energy consumption is on the increase again as the economy recovers. World power consumption will steadily increase at the rate of 2.5% per year up to the year 2030, and non-OECD countries, including China, will account for more than 80% of this increase. Fossil fuel will still serve as the main energy source up to the year 2030. In particular, coal dependency will grow so much that coal-fired power generation will become the largest electrical power source, accounting for 44% of the total. On the other hand, although nuclear power generation will become popular in regions other than Europe, the proportion of nuclear power will decrease, accounting for only about 20% of the total. Renewable energy other than hydroelectric power (wind power,

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00043-5 Copyright Ó 2013 Elsevier Inc. All rights reserved.

4. Development of Thermal Barrier Coatings 5. Damage Modes Observed in Thermal Barrier Coatings 6. Heat-Resistant Evaluation Technology 7. Conclusions References

794 796 799 805 805

photovoltaics, geo-thermal generation, wave power, and biomass energy) is expected to increase, with wind power and photovoltaics being the major sources, although it will account for as low as 8.6% of the total power generation [1]. As described above, thermal power-generation systems will continue to be the major power-generating energy facilities. Furthermore, nuclear power-generation systems, serving as base load stations, will be enhanced, and the development of environment-friendly renewable energy and new energy, such as fuel cells, will accelerate up to the year 2030. Figure 2 shows the relationship between powergeneration efficiency and operating temperature for various types of power-generating facilities. As is apparent from the figure, high operating temperatures directly lead to enhancements in power-generation efficiency for any type of facility, and hence, power-generation systems that operate at high-temperature are being studied and developed. Thermal power-generation systems, compared with fuel cells, which are capable of converting chemical energy directly into electrical energy, suffer from much energy loss and therefore require operating temperatures of 1000  C or more to achieve power-generation efficiency equivalent to that of fuel cells. In recent years, most thermal power-generation systems using liquefied natural gas fuel are likely to employ a combined cycle (LNG C/C) to achieve efficient energy use, environmental friendliness, and improved economical

789

790

FIGURE 1 Prospects for energy demand in the world up to the year 2030.

FIGURE 2 Progress of various power generation systems for the relationship between power generation efficiency and operating temperature.

efficiency. In more detail, the operating temperature of gas turbines in the 1990s and later has been notably high in order to achieve high-efficiency power-generating plants by combining these gas turbines and steam turbines. Such high operating temperatures have been made possible with the development of heat-resistant superalloys forming turbine hot parts, as well as advances made in heat-resistant coating technology and cooling technology. More specifically, for rotor blades with an inlet gas temperature of about 1100  C in a gas turbine for power generation, the inner surface of the substrate, formed of polycrystalline Ni-based superalloy, is air-cooled, whereas the outer surface of the substrate is covered with a corrosion-resistant and oxidation-resistant coating. Furthermore, gas turbine rotor blades of first stage with a turbine inlet temperature of about 1300  C, which are more common at present, are formed of a unidirectionally solidified Ni-based superalloy, and the inner surface of the substrate is air-cooled, whereas

Handbook of Advanced Ceramics

the surface of the substrate and the inner surfaces of the cooling holes are covered with a corrosion-resistant and oxidation-resistant coating. In addition, gas turbine rotor blades of first stage with a turbine inlet temperature of about 1500  C, whose commercial operation has already begun, are formed of single-crystal Ni-based superalloys. For these gas turbine rotor blades, the inner surface of the substrate is steam-cooled, the inner surfaces of the cooling holes are covered with a corrosion-resistant and oxidationresistant coating, and the surface of the substrate is covered with a thermal barrier coating (TBC) [2]. Heat-resistant superalloys and heat-resistant coating technology greatly contributing to high-temperature gas turbines, as described above, were first applied to aircraft engines and then to large-scale gas turbines for power generation. Particularly for metal-based corrosion-resistant and oxidation-resistant coatings, processes such as diffusion coating, vapor deposition, and thermal spraying have been realized in consideration of economic efficiency and the characteristics required for hot parts. The development of these corrosion-resistant and oxidation-resistant coating technologies has greatly influenced the development of various types of facilities exposed to a high-temperature atmosphere, such as boilers and industrial furnaces. On the other hand, ceramic thermal barrier coatings, which have recently attracted attention, are metal-based corrosionresistant and oxidation-resistant coatings covered with ceramic, having superior heat shielding characteristics, which are realized by processes such as thermal spraying or electron beam-physical vapor deposition (EB-PVD) [2e4]. The development of heat-resistant coating technology will be continuously advanced with the aim of further enhancing heat resistance and durability of gas turbine hot parts. This paper reviews the trend of development of heatresistant coating technology for gas turbines. The paper also reviews the trend of development and standardization of heat resistance evaluation test methods for coatings, because such evaluation test methods are indispensable for the development of heat-resistant coating technology.

2. DEVELOPMENT OF HEAT-RESISTANT SUPERALLOYS FOR GAS TURBINES Figure 3 shows one example of the cross-sectional structure of a gas turbine for power generation. The air taken in from the left side in the figure is compressed to a pressure ratio of 15 to 25 by the compressor and is used as combustion air or cooling air. The combustion air is introduced into the combustor in the middle, where it is mixed with fuel and combusted, thus rapidly increasing in volume. The flow direction of the resultant high-pressure combustion air supplied to the turbine section is controlled by the stator vanes, and the high-pressure combustion air is blown onto

Chapter | 10.2

Heat-Resistant Coating Technology for Gas Turbines

791

FIGURE 3 Typical structure of large-scale gas turbine for power generation. For color version of this figure, the reader is referred to the online version of this book.

the rotor blades secured to the disk to transmit rotational energy to the rotor. This turbine section is composed of normally 3e5 stages of blade rows and vane rows. Complicated cooling flow passages are machined in these rotor blades and stator vanes and are cooled with compressed air, for instance. However, the air-cooling method has already reached its limits. For gas turbines with an inlet temperature of about 1500  C, steam, with a thermal conductivity 1.5 times higher than that of air, is also employed as a cooling medium to cool the rotor blades and stator vanes. The key to high-efficiency gas turbines is hightemperature. For this purpose, it is necessary to increase the heat resistance of turbine hot parts. To enhance the heat resistance of hot parts, it is essential to develop heatresistant superalloys for the rotor blades and stator vanes of turbine. Figure 4 shows year-by-year changes of inlet gas temperature of aircraft engines and gas turbines for power generation [4]. This figure demonstrates that the increases in the inlet gas temperature in the gas turbines for power generation are about 10e20 years behind those of aircraft engines. The figure also shows that the inlet gas temperature in gas turbines for power generation is increasing at a rate of about 20  C/year. In fact, Ni-based and Co-based superalloys, serving as heat-resistant superalloys, have been developed for the combustor and the turbine sections exposed to the hottest combustion atmosphere. In

particular, attempts have been made to enhance the hightemperature strength of Ni-based superalloys by crystal control technology, such as unidirectional solidification and single-crystal structures, to put Ni-based superalloys into practical use as rotor blade superalloys. Compounding processes, such as eutectic alloying, oxide-dispersion strengthening, and fiber reinforcing, are also being developed to increase the high-temperature strength of these superalloys. Turbine rotor blades, which rotate at high speed, are exposed to a high-temperature combustion atmosphere and undergo compound damage, such as creep due to centrifugal force, low-cycle fatigue due to thermal stress resulting from repeated start/stop operation, and high-temperature oxidation and high-temperature corrosion due to fuel elements. For gas turbines whose operating temperature has been increased to the limit, it is very difficult to realize both high-temperature strength characteristics and environment resistance with the use of existing heat-resistant superalloys. For this reason, gas turbine hot parts are being developed based on the concept that the substrate is responsible for the high-temperature strength characteristics, whereas the coating is responsible for environment resistance and functions. As a result, a wide variety of high-temperature corrosion-resistant and oxidation-resistant coatings have been developed and put into practical use depending on the combustion atmosphere in gas turbines [5,6].

792

Handbook of Advanced Ceramics

FIGURE 4 Rise of inlet gas temperature for aircraft engines and gas turbines for power generation.

3. DEVELOPMENT OF HEAT-RESISTANT COATINGS FOR GAS TURBINES Figure 5 shows the progress in development of corrosionresistant and oxidation-resistant coatings (environmental barrier coatings) and thermal barrier coatings for gas turbine

hot parts. The severity of the operating conditions of gas turbines differs greatly among the combustor, rotor blades, and stator vanes, and the instances where thermal barrier coatings have been applied greatly differ accordingly. The inner surface of gas turbine combustors has been covered with a ceramic thermal barrier coating from

FIGURE 5 Progress of coating technologies for gas turbine hot parts at high temperature. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 10.2

Heat-Resistant Coating Technology for Gas Turbines

relatively early times. These coatings are based on atmospheric plasma spraying (APS), and the present two-layer thermal barrier coating has been practically realized through the changes to various layered structures. More specifically, in early phases of development, a three-layer coating was developed, i.e., a three-layer coating including a NiAl bond coat (also referred to as undercoat), a CaO or MgO partially stabilized ZrO2 top coat, and a cermet layer, serving as an intermediate layer, formed by mixing a bond coat and top coat for thermal stress reduction. After that, CoCrAlY bond coat with superior corrosion resistance and oxidation resistance at higher temperature was developed and utilized as a bond coat. To further enhance the thermal stress reduction effect, a graded coating in which the ratio of the CoCrAlY bond coat and the MgO partially stabilized ZrO2 top coat changes gradually between the two coats was developed. However, superior oxidation resistance and thermal cycle performance were demanded, and in response, a NiCrAlY or NiCoCrAlY bond coat and a Y2O3 partially stabilized ZrO2 top coat were developed. At present, the inner surface of combustors is covered with a two-layer thermal barrier coating by the atmospheric plasma-spraying process. On the other hand, applying a ceramic thermal barrier coating to gas turbine blades and vanes that require high reliability is carefully performed because hot parts may be fatally damaged if TBC is failed. Initially, rotor blades were subjected to diffusion treatment with Al or Cr and compound treatment involving Pt plating and Al diffusion treatment for the purpose of corrosion-resistant and oxidation-resistant coatings at high-temperature. After that,

793

coatings of, for example, CoCrAlY and NiCoCrAlY formed by electron beam-physical vapor deposition (EBPVD) were employed. Next, NiCoCrAlY coatings formed by low-pressure plasma spraying (LPPS), which is superior for mass production, or compound treatment of Al diffusion treatment applied to a CoCrAlY coating surface were realized. At present, the mainstream is a two-layer thermal barrier coating composed of a NiCoCrAlY bond coat formed by low-pressure plasma spraying and a Y2O3 partially stabilized ZrO2 top coat formed by atmospheric plasma spraying. Figure 6 shows examples of 1300  C-class gas turbine rotor blades (unidirectionally solidified Ni-based superalloy), 1500  C-class gas turbine rotor blades (single-crystal Ni-based superalloy substrate), and a two-layer thermal barrier coating applied to singlecrystal Ni-based superalloy blades. High-velocity oxygen fuel (HVOF) is also used for the bond coat of the two-layer thermal barrier coating. HVOF is a process involving spraying particles onto the substrate surface at a high velocity of 600 m/s or more with the high-pressure combustion energy of fuel and air (or oxygen) serving as a heat source. Although performed in an atmosphere, HVOF is less affected by oxidation during thermal spraying and is capable of producing a dense coating with superior bonding properties. Thus, it is now employed as an alternative to low-pressure plasma spraying. In addition, hot parts of some aircraft engines and gas turbines are covered with a two-layer thermal barrier coating formed by electron beam-physical vapor deposition (EB-PVD), i.e., a two-layer thermal barrier coating composed of an MCrAlY bond coat and a Y2O3 partially

FIGURE 6 Progress of gas turbine blade for high-temperature use. For color version of this figure, the reader is referred to the online version of this book.

794

Handbook of Advanced Ceramics

stabilized ZrO2 top coat with superior erosion resistance, thermal shock resistance, and thermal cycle resistance. The progress of development of heat-resistant coating technology for gas turbines is described in detail in published reports [5e17]. Furthermore, recent trends are summarized in the Journal of the Ceramic Society of Japan, Special Issue (2008) “Recent progress of ceramic heat-resistant coatings” [18e26].

4. DEVELOPMENT OF THERMAL BARRIER COATINGS 4.1. Effect of Thermal Barrier Coatings As shown in Figure 7, thermal barrier coating (TBC) is a technology for applying ceramic having low thermal conductivity and high radiation factor to the front surface of a heat-resistant superalloy substrate whose rear surface is forcibly cooled to decrease the front surface temperature of the substrate. Without a thermal barrier coating, the front surface of the substrate is exposed to high-temperature despite the rear surface being forcibly cooled. On the other hand, with a thermal barrier coating, the temperature decreases in the ceramic layer having lower thermal conductivity than that of the substrate, and the front surface temperature of the heat-resistant alloy can be controlled to a low level. Furthermore, with ceramic having a high radiation factor, the front surface temperature of the substrate can be decreased. This thermal barrier effect can be evaluated quantitatively based on the thermal resistance R, as shown in the expression as follows [15,16]: R ¼ 1=ag þ tc =lc þ tb =lb þ ts =ls þ 1=ac

(1)

where ag and ac are heat transfer coefficient of the hotside surface and the cold-side surface, respectively, ls , lb, and lc are thermal conductivities of the heat-resistant superalloy, bond coat, and the ceramic top coat, respectively, and ts, tb, and tc are the thicknesses of the heatresistant superalloy, bond coat, and the ceramic top coat, respectively. Eqn (1) demonstrates that when the heat transfer coefficient on the high-temperature-gas side is low, the ceramic layer is thick, the ceramic thermal conductivity is low, the thermal resistance R becomes large, and the thermal barrier effect becomes enhanced accordingly. Furthermore, because thermal resistance is the ratio of coating thickness to thermal conductivity, a ceramic layer that is thinner than the heat-resistant superalloy substrate is expected to exhibit a sufficient thermal barrier effect. However, a thermal barrier coating is formed on the heat-resistant superalloy surface and is used under a steep temperature gradient. For this reason, important characteristics of TBC materials include a thermal expansion coefficient close to that of the heat-resistant alloy, superior heat resistance, superior heat cycle resistance for repeated thermal load, thermal shock resistance, and so on. Figure 8(a) shows thermal conductivities measured by the laser flash method for a Y2O3 partially stabilized ZrO2 (8YSZ) top coat formed by atmospheric plasma spraying, a NiCoCrAlY bond coat formed by low-pressure plasma spraying, and the Ni-based superalloy IN738LC. The thermal conductivity of the top coat formed by atmospheric plasma spraying is 1e2 W/mK, which is an order of magnitude smaller than those of the bond coat and the substrate. In addition, unlike the bond coat and the substrate, the thermal conductivity of the top coat changes only

FIGURE 7 Principle of thermal barrier effect for thermal barrier coating formed on substrate.

Chapter | 10.2

Heat-Resistant Coating Technology for Gas Turbines

795

4.2. Cross-Sectional View of Thermal Barrier Coatings

FIGURE 8 Calculated results of thermal barrier effect formed by IN738LC, NiCoCrAlY and 8mass%Y2O3-ZrO2 (a) Temperature dependency of thermal conductivity (b) Effect of top coat thickness on temperature.

slightly over the entire temperature range, including the high-temperature region, demonstrating that the higher the temperature, the greater the thermal barrier effect. Thermal barrier effects calculated using these measurement results are shown in Figure 8(b). The calculation conditions are shown in the figure. The heat transfer coefficients of the hot-side surface and the cold-side surface were set to 2750 W/m2K and 2150 W/m2K, respectively. The temperature distribution was calculated by one-dimensional steady-state thermal conduction analysis. The figure demonstrates that the thicker the top coat, the larger the temperature difference, that is, the thermal barrier effect, between top coat surface A and substrate surface C. For example, with a top coat thickness of 300 mm, a temperature difference of nearly 200 K occurs in the top coat layer, showing the effectiveness of the TBC [2].

Figure 9(a) shows one example of a cross-sectional structure of a thermal barrier coating formed by atmospheric plasma spraying on a Ni-based superalloy. This is a normal TBC applied to the inner surface of a gas turbine combustor and is a two-layer thermal barrier coating composed of a NiCrAlY bond coat and an 8 mass % Y2O3-ZrO2 (8YSZ) top coat. Oxide formation is observed in the NiCrAlY bond coat. The porosity of a mechanically cut piece measured by the Archimedean method is 3.6%. Irregularities on the surface of bond coat are noticeable, and mechanical bonding with the top coat is superior. Furthermore, the 8YSZ top coat is porous, with a porosity of 15.9%. Figure 9(b) shows the observed microstructure of the 8YSZ top coat. The images are cross-sectional views of a TBC plated with Ni as observed with a scanning electron microscope. The elemental Ni distribution was observed based on a secondary electron image and an X-ray microanalyzer (EPMA). Microstructural evaluation of the complicated sprayed layer is easily performed as a result of Ni being plated in open pores of the 8YSZ top coat formed by atmospheric plasma spraying. As is apparent from the figure, a large number of crack-like narrow open pores can be observed on the 8YSZ top coat. These crack-like open pores can be classified into horizontal cracks along the stack direction (horizontal direction in the figure) of the sprayed layer and vertical cracks along the direction orthogonal to the horizontal cracks. The horizontal cracks are disbonded regions that are formed when flying molten particles adhere to the formed coating. The vertical cracks are solidification/ shrinkage cracks formed when the adhering molten particles are cooled and have smaller length and width than disbonded regions. In this manner, there are two types of microcracks on the 8YSZ top coat formed by atmospheric plasma spraying. The disbondings in the stack direction of the sprayed layer increase the thermal barrier effect, whereas the solidification cracks in the vertical direction are effective for thermal stress reduction [8]. Recently, it has been noticed that segmentation of the top coat of thermal barrier coatings is effective in enhancing the delamination resistance of the coatings. Thus, segmented (vertical-crack-introduced) 8YSZ top coats have been realized for thick TBCs formed by atmospheric plasma spraying. The segmentation structure of a top coat was also observed on a TBC formed by electron beam-physical vapor deposition; one example of the cross-sectional structure is shown in Figure 10. In general, electron beam-physical vapor deposition is the process of irradiating coating materials in a vacuum container by using a high-output electron beam to ablate the materials, thereby depositing the

796

Handbook of Advanced Ceramics

(a)

(b)

FIGURE 9 Microstructure of thermal barrier coating formed by atmospheric plasma spray process. (a) Cross-section view of top coat and bond coat (b) SEM observation for top coat of 8mass%Y2O3-ZrO2. For color version of this figure, the reader is referred to the online version of this book.

materials on a substrate surface heated at high-temperature. A segmented TBC formed of columnar grains with a width of several micrometers, as shown in the figure, is obtained by performing coating under appropriate deposition conditions. There is a gap of 1 mm or less between individual columnar grains, and superior thermal cycle resistance characteristics are exhibited as a result of thermal stress being reduced. Furthermore, it has been demonstrated that columnar grains branch to form feather-like structures, exhibiting superior thermal barrier characteristics [22].

5. DAMAGE MODES OBSERVED IN THERMAL BARRIER COATINGS Figure 11 summarizes the damage modes of thermal barrier coatings applied to gas turbine hot parts [12]. First, chemical damage includes high-temperature corrosion due to high-temperature oxidation or corrosion factors in

combustion gas. Evaluation of the high-temperature oxidation characteristics of thermal barrier coatings is carried out, in many cases, by continuous heating tests at a constant temperature in a static atmosphere. For this evaluation, a combustion flame heating test and a corrosive ash deposition heating test in a burner rig have been carried out. A large amount of data have been accumulated from these tests [10,19]. Pure ZrO2 has a certain degree of resistance to sulfuric acid corrosion and vanadium attack. However, ZrO2 degrades when Y2O3, MgO, CaO, and so on added to stabilize the ZrO2 crystal structure react with SO2, SO3, Na2SO4, V2O5, PbSO4, and so on acting as corrosion components in the combustion gas. More specifically, ZrO2 stabilized in the form of a tetragonal or cubic structure becomes unstable due to a corrosive reaction, causing cracks or delamination to occur on the top coat due to a change in volume at the time of a phase change into monoclinic crystal.

Chapter | 10.2

Heat-Resistant Coating Technology for Gas Turbines

797

FIGURE 10 Cross-section view of thermal barrier coating formed by electron beam physical vapor deposition process. For color version of this figure, the reader is referred to the online version of this book.

interface between the 8YSZ top coat and the MCrAlY bond coat, as shown in Figure 12. This is because oxygen passing through the top coat causes the MCrAlY bond coat surface to be oxidized. Thus, the thermally grown oxide increases in volume and produces cracks or delamination on the top coat layer. The thickness u of this TGO layer, which is given by the expression below, becomes larger as the heating temperature becomes higher and the heating time becomes longer [27]: u ¼ k$tn ðn ¼ 0:45Þ

FIGURE 11 Various damage modes observed in thermal barrier coatings during operation.

In addition, the delamination life characteristics of thermal barrier coatings prominently decrease through thermal cycle tests, compared with high-temperature continuous oxidization tests. From this point of view, many studies on delamination life evaluation of thermal barrier coatings using thermal cycle tests have been carried out [10,19]. As a result, it has been clarified that 6e8 mass % Y2O3 partially stabilized ZrO2 top coats exhibit significantly superior thermal cycle resistance compared with other ceramic top coats. Next, when thermal barrier coatings are exposed to a high-temperature atmosphere, thermally grown oxide (TGO) with Al2O3 as a main component is formed at the

(2)

In short, the thickness u of the TGO layer follows an n-th power law of the heating time t, where the velocity constant k represents the Arrhenius temperature dependency. As a simplified method, an attempt has been made to define the limit thickness of the TGO layer at which the 8YSZ top coat starts to be delaminated and to predict the delamination service life of the TBC based on quantification of the growth behavior of the thickness of the TGO layer. The growth of an embrittled layer that is formed on the substrate side due to reaction diffusion at the interface between the bond coat and the substrate is also important as a damage mode [28]. On the other hand, mechanical damage includes cracks and delamination on the 8YSZ top coat due to repeated thermal stress. If compressive stress acts upon the TBC, interface cracks occur due to irregularities at the interface between the 8YSZ top coat and the MCrAlY bond coat. These interface cracks develop through thermal cycling

798

Handbook of Advanced Ceramics

FIGURE 12 SEM and EPMA analysis for cross-section of thermal barrier coating at 1073 K for 5000 h. For color version of this figure, the reader is referred to the online version of this book.

and, at their limit size or larger, lead to TBC delamination. In addition, when very high compressive stress occurs, the TBC undergoes buckling, which then leads to spalling (delamination resulting from buckling). In contrast, when tensile stress acts upon the TBC, the TBC exhibits vertical cracks, that is, cracks orthogonal to the coating. When the TBC undergoes high tensile stress, vertical cracks initially generated on the TBC develop, resulting in delamination in the TBC layer or delamination at the interface between the 8YSZ top coat and the MCrAlY bond coat. However, many experiments have confirmed that delamination damage observed in thermal cycle tests of TBCs occurs in the course of cooling, and major delamination is assumed to occur in the compressive stress field. While it is true that delamination is greatly affected by the thermal load acting on the TBC, residual stress occurring in the TBC at the fabrication stage, as well as oxidation/inflation of the MCrAlY bond coat, that is, residual stress due to growth of the TGO layer, also contributes to delamination. In addition, sites where delamination occurs differ depending on the type of TBC. More specifically, for a TBC formed of an MgO partially stabilized ZrO2 top coat, delamination occurs in the non-stabilized ZrO2 layer. For both a 7 mass % Y2O3eZrO2 top coat and NiCoCrAlY bond coat formed by atmospheric plasma spraying,

delamination occurs in the bond coat. If only the NiCoCrAlY bond coat is formed by low-pressure plasma spraying, delamination occurs in the top coat close to the interface with the bond coat. If the 7 mass % Y2O3eZrO2 top coat is formed by EB-PVD, and the NiCoCrAlY bond coat is formed by low-pressure plasma spraying, delamination occurs in the TGO layer formed at the interface between the top coat and the bond coat or occurs at the interface between the TGO layer and the bond coat. These delamination sites are assumed to be the weakest points of the respective TBCs. Recently, sintering/shrinkage cracks of the top coat due to the high-temperature in gas turbines and resultant delamination have been observed [29]. Figure 13 shows a TBC surface “as-sprayed”, as well as TBC surfaces heated for one hour at 1373 K, 1473 K, and 1573 K, respectively. Cracks on the 8YSZ top coat surface are seen in the assprayed state, and the 8YSZ top coat can be seen to start sintering at a heating temperature of 1473 K or more. This sintering shrinkage of the 8YSZ top coat causes not only a decrease in the thermal barrier effect due to densification of the coating but also delamination. More specifically, if the top coat surface layer undergoes delamination damage due to sintering shrinkage, the underlying surface is exposed to high-temperature and is subjected to another sintering

Chapter | 10.2

Heat-Resistant Coating Technology for Gas Turbines

799

FIGURE 13 SEM observation of thermal barrier coating heated for one hour at high temperature. For color version of this figure, the reader is referred to the online version of this book.

shrinkage that leads to secondary delamination damage, thus decreasing the thickness. The sintering shrinkage rate (DL/L0) of this 8YSZ top coat is given by DL=L0 ¼ k0 $tn ðn ¼ 0:4Þ

(3)

More specifically, the sintering shrinkage rate of the 8YSZ top coat follows an n-th power law of the heating time t, where the sintering rate constant k0 represents the Arrhenius temperature dependency. Based on this sintering shrinkage behavior of the 8YSZ top coat, the occurrence of TBC vertical cracks can be evaluated under a temperature gradient. Under a temperature gradient, vertical cracks generated on the 8YSZ top coat are bowed, causing delamination on the surface layer. Last, although depending on the type of fuel used in the gas turbine, erosion damage of the 8YSZ top coat due to solid particles in the combustion gas is also a critical damage mode. As described above, thermal barrier coatings include various damage modes, and damage phenomena appear in diverse patterns depending on the TBC operating environment and type. Figure 14 schematically illustrates systematization of coating engineering for heat-resistant coatings of gas turbines. In summary, coating technology is composed of (1) coating materials and process technology, (2) coating design technology, and (3) coating evaluation technology for service life and characteristics. Development of conventional heat-resistant coating technology has been aimed at enhancing characteristics, with emphasis on materials and process technology. To achieve higher functions, performance and higher reliability, however, development of coating design technology is essential, and enrichment of databases will be indispensable in doing so. Furthermore, standardization of test methods is important for evaluating service life and characteristics, which has so far been carried out in various ways.

FIGURE 14

Is it possible to organize the coating engineering?

6. HEAT-RESISTANT EVALUATION TECHNOLOGY 6.1. Trends in Development of Coating Evaluation Tests In the history of development for practical realization of gas turbine hot parts, development of test methods for evaluating the service life and characteristics of heat-resistant coatings has long played an important role. Initially, tests originally designed for heat-resistant superalloys were adapted to heatresistant coatings. At present, however, damage evaluation tests exclusively designed for heat-resistant coatings are employed. Table 1 summarizes heat resistance evaluation test methods used for gas turbine hot parts, classified into several types, as well as the development process of each type. As is apparent from the figure, heat resistance evaluation tests can be classified into (1) burner heating tests,

800

Handbook of Advanced Ceramics

TABLE 1 Progress of Heat Resistance Evaluation Test for Gas Turbine Hot Parts

(2) furnace þ burner heating tests, and (3) furnace heating tests, based on the heating method. First, the burner heating test has been conducted as a high-flow-rate combustion test (burner rig test) to evaluate heat resistance under the operating conditions used in actual facilities. This is a test method primarily for verifying the heat resistance of hot parts by approximating the combustion conditions and high-temperature corrosion environment in the test to those in actual facilities. The burner heating test includes an atmospheric pressure combustion test and an actual-pressure combustion pressure test [9]. Next, the furnace þ burner heating test is a hybrid test in which furnace heating is combined with high-temperature corrosive gas generation by burner combustion. Compared with the burner heating test, this hybrid test can control the test temperature with high precision. Furthermore, because of its ability to simulate a high-temperature corrosion environment with combustion gas, the hybrid test is an effective method for carrying out long-term high-temperature corrosion evaluation [9]. The furnace heating test is the most basic heat resistance evaluation test method and has been used at the material development stage. Because uniform heating is easy in a furnace, many actual examples of hightemperature continuous oxidization tests, high-temperature cyclic oxidation tests, and high-temperature corrosion tests for heat-resistant superalloy and heat-resistant coatings have been reported [10]. All the heating tests described above were initially developed for evaluating the heat resistance of noncooled blades (1980s and before). As shown in Table 1, however,

these tests have evolved into heat resistance evaluation test methods taking into consideration the mechanical load exerted on the rotor blades and stator vanes, and further into heat resistance evaluation test methods with a temperature gradient, in consideration of the wide adoption of cooled blades. The evolution of such heat resistance evaluation test methods has enabled the evaluation of gas turbine hot parts under conditions more closely approximating the actual conditions and has accelerated the development of hot parts. In more detail, the gas turbine rotor blades rotate at high speed in a high-temperature combustion gas atmosphere and are subjected to centrifugal force (tensile stress) load. On the other hand, the stator vanes are exposed to high-temperature combustion gas with constrained thermal expansion and are subjected to compressive stress load. First, in the burner heating test, heat resistance evaluation tests reflecting the above-described load conditions are carried out. In short, this is a test method in which a test specimen is subjected to both mechanical load and a temperature gradient together. In fact, a high-temperature axial-strain control test device with a temperature gradient has been developed, and long-term tests in a high-temperature atmosphere have mainly been carried out [15e17]. Next, also in the furnace heating test, many heat resistance evaluation tests for high-temperature corrosionresistant and oxidation-resistant coatings reflecting the load conditions of actual blades and vanes have been carried out. In the furnace heating test, a thermal mechanical fatigue test (TMF) in which a thermal barrier coating test specimen is subjected to mechanical load and a temperature gradient has been developed. With this test, both the temperature and stress history at the time of startup/stopping of the gas

Chapter | 10.2

Heat-Resistant Coating Technology for Gas Turbines

turbine were simulated to investigate the effect on TBC delamination damage [15e17]. Also for the hybrid (furnace þ burner) heating test, on the other hand, a test composed of a thermal mechanical fatigue test plus the influence of high-temperature corrosion is possible, although no actual examples have been reported at present. Such a test will be carried out if longterm and high-temperature corrosion evaluation becomes necessary in future due to the diversity of fuels.

6.2. Standardization of Evaluation Test Methods Normally, heat resistance evaluation for gas turbine hot parts can be classified into four steps IeIV, as shown in Figure 15. More specifically, step I corresponds to coating design, step II to the furnace heating test, step III to the burner heating test, and step IV corresponds to a rainbow rotor test (Verification test) with hot parts installed in a gas turbine. Based on this classification, a standardization road map, as shown in Table 2, has been formulated for test methods for

801

evaluating the characteristics of gas turbine heat-resistant coatings [18]. Based on this road map, world-leading standardization activities are being carried out in the New Material Center affiliated with Osaka Science & Technology Center. The progress of this standardization is described below. First, “Testing Methods for Heat Resistance Under Temperature Gradient (JIS H7851)”, involving burner heating, was formalized as a Japanese Industrial Standard in 2005. More specifically, as shown in Figure 16, this test is a combustion test method in which the surface of a diskshaped test specimen brazed on a copper holder is heated with a burner, and the rear surface of the copper holder is cooled to produce one-dimensional heat flow, thus carrying out heat resistance evaluation. This is an evaluation test method mainly intended to elucidate the damage mechanism of heat-resistant coatings, and the thermal flow rate, equivalent thermal conductivity, and so on can be calculated using the expressions shown in the figure based on temperature measurements. At present, this testing method is under discussion for international standardization (ISO 13123) by ISO/TC107.

FIGURE 15 Four steps for practical use of gas turbine hot parts. For color version of this figure, the reader is referred to the online version of this book.

802

Handbook of Advanced Ceramics

TABLE 2 Roadmap for Standardization of Heat-Resistant Coating Tests

FIGURE 16

Testing method for heat resistance under temperature gradient (JIS H 7851/ISO 13123:2011).

Chapter | 10.2

Heat-Resistant Coating Technology for Gas Turbines

Next, “Testing Methods for Thermal Cycle and Thermal Shock Resistance of Thermal Barrier Coatings (JIS H8451)”, involving furnace heating, and “Testing Methods for Thermal Cycle and Thermal Shock Resistance of Oxidation-Resistant Metal Coatings (JIS H7852)” were formalized as Japanese Industrial Standards in 2008. JIS H8451 specifies a thermal cycle test method in which a TBC test specimen is repeatedly subjected to thermal cycling by furnace heating and air cooling, as well as a thermal shock test method in which a TBC test specimen is subjected to thermal shock by furnace heating and then water cooling, as shown in Figure 17. Figure 18 shows one example of an 8 mass % Y2O3eZrO2 thermal barrier coating compared with an Al2O3 coating formed by atmospheric plasma spraying in order to elucidate the thermal shock resistance characteristics of the 8 mass % Y2O3eZrO2 coating. It should be noted, however, that SUS304 is used for the substrate. The 8 mass % Y2O3eZrO2 thermal barrier coating exhibits a thermal shock temperature difference, DT, of 900  C or more, whereas the Al2O3 coating exhibits a thermal shock temperature difference, DT, of 500  C, when DT is measured

FIGURE 17

803

just before the residual bonding strengths of the thermally shocked coatings (JIS H8402) decrease by 30%. At present, JIS H8451 is also under discussion for international standardization (ISO 14188) by ISO/TC107. For heat-resistant coatings, on the other hand, it will be indispensable to introduce coating design technology, as shown in Figure 15. For this purpose, it is important to construct a database necessary for coating design. From this standpoint, a project sponsored by the New Energy and Industrial Technology Development Organization (NEDO), “Standardization Investigation Project on Testing Method for Characteristic Evaluation of Heat-Resistant Coating”, was conducted in 2006e2008, and a JIS draft regarding evaluation of mechanical characteristics and thermal characteristics of thermal barrier coatings was prepared. Figure 19 shows an example of residual stress measurements for thermal barrier coatings. The figure shows calculated results of residual stress distribution in the thickness direction of the thermal barrier coatings, based on the substrate deformation when a strip-shaped test specimen is

Testing methods for thermal cycle of thermal barrier coatings (JIS H 8451/ISO 14188:2012).

804

Handbook of Advanced Ceramics

FIGURE 18 Testing results for thermal shock resistance of thermal barrier coatings (JIS H 8451/ISO 14188:2012).

FIGURE 19

sprayed or the substrate deformation resulting from coating removal. The CoNiCrAlY coating exhibited higher tensile residual stress on the surface than the 8 mass % Y2O3eZrO2 thermal barrier coating, showing relatively good agreement with the X-ray surface residual stress measurements. Figure 20 shows thermal diffusivity obtained by measuring the temperature history curve of coatings by the laser flash method and then calculating based on the differences from values measured with the substrate alone. The thermal diffusivity of the CoNiCrAlY coating is in good agreement with the results measured with a single CoNiCrAlY layer if the thickness of the coating is 0.25 mm or more. On the other hand, the thermal diffusivity of the 8 mass % Y2O3eZrO2 thermal barrier coating is slightly smaller than the measurements of a single 8 mass % Y2O3eZrO2 layer but does not differ depending on the top coat thickness. At present, a JIS draft related to methods for measuring the elastic constant, linear expansion coefficient,

Measurement results for residual stress distribution of thermal barrier coating.

FIGURE 20 Measurement results for thermal diffusivity of thermal barrier coating.

Chapter | 10.2

Heat-Resistant Coating Technology for Gas Turbines

805

drastically increased in number. The reliability of this technology is also increasing accordingly. Among others, studies on characterization of thermal barrier coatings and delamination life evaluation have advanced remarkably. This is because researchers at universities and research institutes have actively taken part in development of technology for evaluating the service life and characteristics of thermal barrier coatings, while manufacturers of gas turbines have primarily been involved in process development. In near future, standardization of heat-resistant coating technology will become increasingly important to improve the reliability of gas turbine hot parts. For this purpose, acceleration of standardization through industryeacademia collaboration is essential. From these efforts, we expect rapid advancement of heat-resistant coating technology in the future. The names of the products referred to in this paper may or may not be trade names of respective corporations. FIGURE 21 Non-destructive evaluation of thermal barrier coating on gas turbine blade using infrared thermograph (Delaminaton A: 3 mm in diameter, Delamination B: 5 mm in diameter, Delamination C: 10 mm in diameter). For color version of this figure, the reader is referred to the online version of this book.

and thermal conductivity of thermal barrier coatings is being produced. On the other hand, one important technology related to heat resistance evaluation tests for heat-resistant coatings is a technology for nondestructive detecting delamination of coatings. The ultrasound method or holography method has already been investigated, and in addition, technology for detecting delamination of coatings by infrared image analysis is being used in practice [30]. In short, this technology is an approach for detecting delamination sites by infrared thermography, by drawing upon the phenomenon that only the delamination sites are overheated when the TBC surface is heated, as shown in Figure 21. With this technology, not only delamination of heat-resistant coatings can be detected, but also changes in thermal barrier characteristics of coatings due to erosion or local sintering phenomena can be measured. Future improvements in this technology include enhancement in detection accuracy, early application to complicated structures such as actual facilities, and standardization.

7. CONCLUSIONS For 1500  C-class gas turbines, the adoption of singlecrystal Ni-based superalloy blades and ceramic thermal barrier coatings is indispensable, and additionally, steamcooled technology should be employed. In particular, thermal barrier coating (TBC) technology is recognized as important, and research papers on this technology have

REFERENCES [1] IEA. World energy outlook 2009 executive summary; 2009. [2] Ito Y. Recent progress of heat-resistant coating technology for gas turbine. Bull Ceram Soc Jpn 2006;41(10):821e39. [3] Harada Y. Coating technology for gas turbine. Bull Gas Turb Soc Jpn 2003;31(2):94e107. [4] Yoshida T. Cooling technology. Bull Gas Turb Soc Jpn 1997;25(97):29e34. [5] Ito Y. High-temperature corrosion resistant coating for gas turbine (1). In: Science of Machine, vol. 44e2. Yokendo; 1992. p. 257e263. [6] Ito Y. High-temperature corrosion resistant coating for gas turbine (1). In: Science of Machine, vol. 44e3. Yokendo; 1992. p. 369e376. [7] Ito Y. Thermal barrier coatings for land-based gas turbine (1). In: Science of Machine, vol. 47e7. Yokendo; 1995. p. 746e751. [8] Ito Y. Thermal barrier coatings for land-based gas turbine (2). In: Science of Machine, vol. 47e8. Yokendo; 1995. p. 848e852. [9] Ito Y. Performance evaluation of thermal barrier coatings for land-based gas turbine (1). In: Science of Machine, vol. 47e9. Yokendo; 1995. p. 947e951. [10] Ito Y. Performance evaluation of thermal barrier coatings for landbased gas turbine (2). In: Science of Machine, vol. 47e10. Yokendo; 1995. p. 1040e1048. [11] Ito Y. Recent progress of thermal barrier coatings for the mechanical properties (1). In: Science of Machine, vol. 47e11. Yokendo; 1995. p. 1150e1154. [12] Ito Y. Recent progress of thermal barrier coatings for the mechanical properties (2). In: Science of Machine, vol. 47e12. Yokendo; 1995. p. 1244e1249. [13] Ito Y. Afterwards of thermal barrier coating technology for gas turbine (1). In: Science of Machine, vol. 52e10. Yokendo; 2000. p. 1049e1054. [14] Ito Y. Afterwards of thermal barrier coating technology for gas turbine (2). In: Science of Machine, vol. 52e110. Yokendo; 2000. p. 1152e1158. [15] Ito Y. Performance evaluation of heat-resistant coatings (1). In: Science of Machine, vol. 57e6. Yokendo; 2005. p. 633e640.

806

[16] Ito Y. Performance evaluation of heat-resistant coatings (2). In: Science of Machine, vol. 57e7. Yokendo; 2005. p. 750e758. [17] Ito Y. Performance evaluation of heat-resistant coatings (3). In: Science of Machine, vol. 57e8. Yokendo; 2005. p. 864e871. [18] Ito Y. Recent progress of ceramic heat-resistant coatings. Bull Ceram Soc Jpn 2008;43(5):358e62. [19] Kojima Y. Thermal barrier coating technologies for heavy duty gas turbine. Bull Ceram Soc Jpn 2008;43(5):363e6. [20] Sodeoka S. Trends in thermal barrier coating for jet engines. Bull Ceram Soc Jpn 2008;43(5):367e70. [21] Vassen R, Cernuschi F, Rizzi G, Markocsan N, Ostergren L, Kloosterman, et al. Overview in the field of thermal barrier coatings including burner rig testing in the European Union. Bull Ceram Soc Jpn 2008;43(5):371e82. [22] Wada K, Fuse T, Ishiwata Y, Ito Y. Trend of the development on thermal barrier coatings produced by electron physical vapor deposition. Bull Ceram Soc Jpn 2008;43(5):383e8. [23] Wilson S, Dorfman M, Sporer D. Recent development of high temperature thermal spray abradable coatings. Bull Ceram Soc Jpn 2008;43(5):371e82.

Handbook of Advanced Ceramics

[24] Harada Y. Microstructures and hardness of sprayed Cr3c2eNiCr cermet coatings and their blast erosion. Bull Ceram Soc Jpn 2008;43(5):396e400. [25] Kawasaki A. Progress of evaluation technology on thermal resistance of TBCs under controlled temperature gradient. Bull Ceram Soc Jpn 2008;43(5):401e4. [26] Takahashi S, Yoshiba M. Standardization of testing methods for spalling resistance of high-temperature coating systems. Bull Ceram Soc Jpn 2008;43(5):405e8. [27] Ito Y. Basic study of oxidation behavior of zirconia thermal barrier coating at high-temperature. J Soc Mat Sci Jpn 1998;47(7):665e71. [28] Ito Y, Tamura M. Reaction diffusion behaviors for interface between Ni-based superalloy and vacuum plasma sprayed MCrAlY coatings. Trans ASME J Eng Turb Power 1999;121:476e83. [29] Ito Y, Kameda T, Okamura T, Nagata K. Sintering behavior of zirconia thermal barrier coating under gradient temperature. J Soc Mat Sci Jpn 1999;48(7):740e5. [30] Ito Y, Saitoh M, Kashiwaya H, Oishi M, Kaneko T. Infrared thermography analysis of an interfacial crack and its evaluation. J Ceram Soc Jpn 1989;97:1358e64.

Chapter 10.3

Application of High-Temperature Corrosion-Resistant Ceramics and Coatings under Aggressive Corrosion Environment in Waste-To-Energy Boilers Yuuzou Kawahara Dai-Ichi High Frequency Co., Ltd., 1-45 Mizue-cho, Kawasaki, Kawasaki, Kanagawa 210-0866, Japan, Former Affiliate: Yokohama R&D Center, Mitsubishi Heavy Industries, Co., Ltd.

Chapter Outline 1. 2. 3. 4.

Introduction Corrosion Environments of Waste-To-Energy Boilers Application of Advanced Refractory Materials Advances in High-Temperature Corrosion-resistant Materials and Coatings

807 808 810 817

1. INTRODUCTION Promotion of high-level recycling of waste and reduction in environmental load such as CO2 and dioxins are required for waste treatment worldwide. The concepts of higher efficiency waste-to-energy (WTE) plants and waste gasification and ash melting power generation plants are positioned in the center of thermal recycling in Japan. The high-efficiency WTE boilers have been progressed recently with steam conditions of more than 400  C/3.9 MPa. On the other hand, the fluidized bed boilers burning biomass such as woods, straw, scrap tires, RDF, RPF etc. have recently been operated in 440e540  C and 6e30 MPa electric power generation plants. Nowadays a high total cost performance is required in the storm surrounding the society such as the globalization of economy, the strict pollution regulations, and carbon neutral power plants. Therefore, the development of low-cost and highly durable materials and application processes has become important issues that are essential for a high thermal efficiency and economy of WTE plants. In addition, in order to improve the durability of high-temperature, high-pressure boiler materials, it is considered necessary to apply

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00044-7 Copyright Ó 2013 Elsevier Inc. All rights reserved.

5. Field Experiences of Cermet and Ceramic Coatings for Superheaters 6. Deterioration Mechanisms of Coatings 7. Summary References

821 825 833 835

optimum boiler designs that prevent high temperature corrosion (HTC) and erosionecorrosion of superheater tubes (SHTs) and waterwall tubes (WWTs). There are high expectations for the development of the high-temperature corrosion-resistant materials (CRMs) and the corrosionresistant coatings (CRCs) that are highly durable and workable against aggressive corrosion environments. Advanced refractory materials (RMs) that realized excellent furnace performance have been developed and applied for combustion furnace to prevent high-temperature damages and slugging of ash. Also, CRMs have been progressively used for SHTs, and the application of CRCs such as ceramic tile, metal spray coatings, and weld overlay of Ni-base alloys has been widespread for the WWTs in order to reduce maintenance cost and keep stable operation for long duration. There is a strong need for low-cost materials to improve the durability in severe erosionecorrosion damages caused by the soot blower attack. This report describes 1) the formation of severe corrosion environment of WTE boilers, 2) advanced RMs used in combustion furnace as porous ceramics mainly manufactured by using artificial raw materials, 3) application trends

807

808

Handbook of Advanced Ceramics

and field experiences of advanced ceramic and cermet coatings that have good durability in WWTs and SHTs, and 4) corrosion and deterioration mechanisms of alloys, cermet, and ceramic coatings formed by thermal spraying.

2. CORROSION ENVIRONMENTS OF WASTE-TO-ENERGY BOILERS In the recent trends of most WTE plants, the steam condition of 300e450  C/2.9e5.8 MPa shown in Figure 1 is adopted to avoid the corrosion damages of boilers. Due to a large change in life styles of people and in production styles by industries brought about by technological and economic development, the properties of municipal and industrial waste and the WTE plant operations have changed as follows during recent forty years [1]: 1. Operational change for strict pollution regulation: low O2 operation, high-temperature combustion, rise of plant operating rate, etc. to prevent generation of NOx and dioxins. 2. Energy-saving and effective utilization of heat recovery: severe corrosion environment has been formed due to improvement of electric power-generation efficiency such as high-temperature and high-pressure steam condition (400e500  C/3.9e9.8 MPa) of boilers. 3. Needs to reduce volume of combustion ash and produce nonpolluting ash have led to the development of new incineration processes, such as waste pyrolysis gasification and ash melting plants and oxygen-enriched combustion [2]. 4. Improvement of cost performance: Reducing of operational cost by advances in maintenance-free operation, such as combustion control and monitoring technologies.

Various substances including incombustibles as well as combustibles such as woods, paper, and plastics are inhomogeneously mixed in the waste. Therefore, the fluctuation of gas temperature and composition containing much HCl, SOx, O2, H2O etc. is larger than that in fossil fuel boilers; also, many low-melting-point deposits containing high concentration of chlorides is generated in the combustion gas. Corrosive constituents influenced to HTC in different fuel boilers are listed in Table 1 [3]. Also, Figure 2 shows the explanatory drawing of the corrosion factors formed by combustion gas in WTE boilers [4]. Chlorides and sulfate dusts containing high concentration of alkaline metals (Na, K, etc.), heavy metals (Pb, Zn, etc.), deposit on the material surfaces exposed in the combustion gas, and severe HTC is caused in boiler components such as WWTs and SHTs by the deposits melting at 300e550  C. In SHTs, the metal temperature is as high as 300e550  C, then deposits are easily molten by eutectic reaction and corrosion rate sometimes shows as high as several mm/year or more. Figure 3 shows the increase in corrosion rate of SHT materials with increase in the combustion gas temperature at 450 and 550  C [5]. Furthermore, on the WWT surfaces where the gas temperature is very high (850  C or higher), high concentration of chlorides with lower melting point is promoted, and the corrosion damage is caused mainly by chlorination reactions. The corrosion form of materials is commonly general corrosion, although intergranular and/or localized corrosion occur by the existence of molten deposits and the breakdown of the protective oxide layer on even highly CRMs [6]. Both the adherence of strongly corrosive dust constituents, such as chlorides and sulfates, and the gas temperature fluctuation can be reduced by reducing the combustion gas temperature flowed into the SHTs to less than ~650  C. Then the reduction in gas

FIGURE 1 Trends in steam temperature of WTE boilers.

Chapter | 10.3

TABLE 1 Corrosive Matter and Deposits Condition Influencing to High Temperature Corrosion in Typical Different Fuel Boilers

Corrosive matter

Combustor

Fuels

Wasterto-energy Boiler

Municipal and Industrial waste, Biomass etc.

Sodium Recavary, Biomass Boiler

Black Liquor

Coal Fired Boiler

Coal, Wood etc.

Cl

S

Metals

O

Na

K

Low melting point deposits

Others

Melting point /  C

Zn, Pb 290 ~ 516

e

Fe

D

Corrosion phenomena

Molten phase content

High temperature gaseous corrosion

Chlorination

D

300 ~ 500

Medium

D

550 ~ 620

Samll

D

D

e

D



D

Al, Ca Si, e C

D

Heavy Oil Boiler

Heavy oil, Asphart etc.

D

e

V

480 ~ 725

Medium

D

Heavy Oil Gas Turbine (G/T)

Heavy oil

D

e

V

480 ~ 725

Large

D

Remarks

Degree of influence

D

Oxidation

Small & 20%

Decomposition Coal gas etc. Gasfication, Singas Cooler

D

Sulfidation

Molten salt related corrosion

Application of High-Temperature Corrosion-Resistant Ceramics

Condition of corrosive deposits

: Main corrosion reaction : Existence of reaction D: According to condition

809

810

Handbook of Advanced Ceramics

FIGURE 2 Image diagram of combustion and corrosion factors in stoker-type waste incinerator.

temperature is effective for stabilization of protective oxide layer formed on the metal surface. However, the soot blowing used for the purpose of removing deposits is known to result in severe thermal cycling condition and even the breakdown of protective oxides layer.

RMs used in combustion furnace of boilers are summarized schematically in Figure 4. Advanced refractory linings (RLs) and RMs mainly applied to the corrosive environment of stoker combustion furnaces are shown in Figure 2 and described in this chapter.

3. APPLICATION OF ADVANCED REFRACTORY MATERIALS

3.1. Firebrick Linings

Progression of combustion technologies in the WTE boiler, the high durability of furnace structure, and RMs become the key issue to maintain high-performance and continuous operation of furnaces. Therefore, raw materials of RMs change from natural materials to artificial materials such as high-purity silicon carbides (SiCs) and alumina (Al2O3). RMs are constructed by relatively coarse and inhomogeneous grains compared with metallic materials and fine ceramics, then those properties porous and somewhat scatter. These are the reasons for difficulties of materials evaluation and selection. Influenced factors for damages of

Furnace walls (FWs) in relatively small incinerators, without boilers are constructed by brick linings. Required properties and mainly applied RMs along the gas flow in stoker and bubbling fluidized bed furnaces are shown in Table 2 [7]. Also, Figure 5 shows typical RMs and installation methods used for waste incinerator and industrial boiler. High-SiC brick linings that have high corrosion resistance for molten ash (mp: 1200e1300  C) are used to prevent large slugging of molten ash (so-called “clinker”) in high-temperature zone and to easily remove clinker without damaging RMs. One example of clinker growth in

Chapter | 10.3

Application of High-Temperature Corrosion-Resistant Ceramics

811

FWs is shown in Figure 6 [8]. SiC bricks are also very hard and have high wear resistance, than that installed near and along the stoker grate to prevent wear by hard components including in waste. SiC bricks have high thermal shock resistance due to their high thermal conductivity, but are easily oxidized by H2O vapor etc. at a high temperature of more than approximately 900  C, as shown in Figure 7 [9]. SiC particles oxidize to SiO2 ones that have higher volume, and as a result of oxidation, swelling and fracture damages occur as listed in Figure 8 [10]. As high Al2O3 bricks are also hard and highly oxidation and corrosion resistant, they are therefore sometimes installed in the high gas temperature zone. However, thermal shock resistance is relatively lower than that of SiC bricks. Common fireclay bricks made by alumina and silica are installed at a relatively low-gastemperature zone in furnace because of lower heat and corrosion resistances. Table 3 shows comparison of physical properties in typical RMs used in WTE boilers [11]. Also, Table 4 shows comparison of main damage resistances in typical RMs required for boiler applications [12]. FIGURE 3 Change in maximum corrosion thickness loss of superheater materials with gas temperature (Metal temperature: 450  C, 550  C).

FIGURE 4 Damage factors for refractory materials used in actual combustion furnace.

812

Handbook of Advanced Ceramics

TABLE 2 Required Performance and Applied Refractories for the Furnace of Stoker and Fluidized Bed Combustor in Waste Incineration Boilers Combustor

Stoker type Dry stage

Combustion stage

After

Positions

Feeder

Up

Bottom

Up

Bottom

Up

Temperature /  C

500 ~ 900

900 ~ 1100

900 ~ 1100

900 ~ 1300

900 ~ 1100

800 ~ 1000

e

e

Needed performance

Thermal shock resistance

e

Oxidation resistance

e

e

Stream oxidation resistance e

Corrosion resistance for alkarine

e

e

e

e

e

e

Abresion resistance

Applied refractories

e

e

Removability of clinker

e

e

e

e

High strength

e

e

e

e

e

e

High thermal conductivity

e

e

e

e

e

e

High-SiC

B

High-alumina (SK35~)

B, C 35 ~

B, C 35 ~

e

e

B, C 35 ~

e

Firecray (~SK34)

B, C 32 ~ 34

B, C 32 ~ 34

e

e

B, C 34

e

e

e

High-strength, Low-abrasion

e

B

e

e

B, C, P

B, C, P

e

B, C, P

e

Remarks: Mainly used refractories (B: Bricks, C: Castable, P: Plasic)

3.2. Cooled Furnace Wall 3.2.1. Water Cooled Lining Increase in through-put and calorific value of waste and reduction in furnace wall temperature have been required to prevent slugging problem of clinker and damages of RMs. Also in large-scale WTE plants, efficient heat recovery from combustion gas becomes important, subject to increase in boiler thermal efficiency. Therefore, water-cooled FWs with boiler WWTs become major furnace structures at this time. In water-cooled refractory linings (RLs), reduction in furnace wall

temperature, strength of RMs, and highly durable support system for linings are important subjects. The following support systems for unshaped RLs shown in Figure 9 are commonly applied to maintain the long lifetime of FWs [13,14]: 1. Studs with SiC ceramic cap system 2. Anchor system made by heat-resistant alloys or ceramics In castable and plastic refractories, grain structures cannot be sintered completly, because rear-side of RLs is cooled by WWTs that is low surface temperature of approximately

Chapter | 10.3

Application of High-Temperature Corrosion-Resistant Ceramics

813

TABLE 2 Required Performance and Applied Refractories for the Furnace of Stoker and Fluidized Bed Combustor in Waste Incineration Boilers Stoker type

Fluidized bed type

After combustion

Boiler

Bottom

Gas Mixing

Inlet

Air dispersion Outlet plate

Combustor Transition Burner

Freeboard

Furnace roof

Boiler

800 ~ 1000

800 ~ 1000

800 ~ 1000

500 ~ 600

500 ~ 800

900 ~ 700 ~ 800 1000

1000 ~ 1300

1000 ~ 1100

1000 ~ 1100

350 ~ 950

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e B, C 35 ~

e

C

e

e

e

e

e

e

e

e

e

C, P

C, P e

e

e

e

e

e

e

C, P

C, P e

e

e

e

e

B, C 34 ~ 35 e

e

e

e

e

e

e C

e

C

B

B

e

C

C

e

e

C C

e

FIGURE 5 Classification of refractory materials used for waste incinerators and industrial boilers.

C

e

C, P

e

814

Handbook of Advanced Ceramics

FIGURE 6 Clinker growth problem due to high temperature combustion in waste incinerator.

FIGURE 7 High temperature oxidation rate of SiC refractory at 1000  C in typical oxidized gases.

FIGURE 8 Classification of main damage modes and factors in high-SiC bricks.

Chapter | 10.3

Application of High-Temperature Corrosion-Resistant Ceramics

815

TABLE 3 Typical Properties of Refractory Bricks Applied in Waste Incineration Boilers Crush strength / kg,cm 2

Thermal conductivity/ kcal,m 1,hr 1,  C 1 at 1000  C

Chemical composition/ mass%

Bricks

Apparent density

Bulk density

Apparent porosity (%)

Firecray

2.67

2.27

15.1

600

1.5

Al2O3

43

High-alumina

3.19

2.48

14.2

800

2.0

Al2O3

55

High-SiC (Silicate bonded)

3.12

2.63

15.8

817

13.5

SiC

90

-

2.65

11.5

1200

7.5

SiC

68

High-SiC (Si3N4 bonded)

3.10

2.65

14.6

1500

14.1

SiC

73

Si3N4

22

230e330  C (against steam pressure: 2.9e9.8 MPa). Therefore, deterioration of RMs, also studs and anchors occurs by penetrated corrosive gas components such as HCl, SO2 etc. and then RLs are sometimes peeled off. Corrosion-resistant materials such as CreSieAleFe-base alloys and stainless steels are used for studs and anchor.

3.2.2. Ceramic Tile System Ceramic tiles with longer lifetime than conventional castable and plastic RMs have been developed and applied. Tile is manufactured by high-SiC and sometimes high-Al2O3 RMs in many cases [15]. As shown in Figure 10, the tiles are installed by using special hanging hook or fixing nut etc. with refractory mortar onto the outer surface of WWTs. The surface temperature of the tile linings reaches approximately 900  C or lower by cooling of WWTs. SiC tiles are chemically stable and have a size of 200e300 mm and a thickness of 25e60 mm to prevent damages due to thermal stresses during operation. The RMs, production process, shape, and installation methods are determined based on the evaluation of the deterioration resistance, thermal shock resistance, etc.

3.2.3. Air-Cooled Structure In order to prevent clinker trouble, air-cooled FWs shown in Figure 11 have been developed [16]. Surface temperature of high-SiC brick (with the thickness of 80e150 mm) is controlled by air cooling from rear side of FWs. The following performances for air-cooled FWs are required in furnace design. 1. Thermal shock resistance of FWs is required to prevent fracture damages against rapid temperature change such

as start-up/shut-down and gas temperature fluctuation etc. Applied RMs, size, and support system are also determined considering individual operating conditions. 2. Heat- and corrosion-resistant structures are designed to keep a long lifetime for dead weight loading, thermal stress, corrosion reaction, etc. 3. Good workability of FWs is required against easy installation, repair of RMs, and using of refractory mortar.

3.3. Unshaped Refractory Linings Unshaped (monolithic) RMs such as castable and plastic refractories are used mainly for hanging wall, WWTs, and curved parts etc. Application forms of each RLs were described in detail as follows.

3.3.1. Castable Refractories Commonly, fireclay, high-Al2O3, and high-SiC castable RLs have been used with a thickness of more than 100e150 mm. Casting molds are installed along FWs, and blended castable refractories are poured into moldings. Compressed or vibration casting methods and gunning are sometimes applied to disperse large particles homogeneously. The terms important for designing and operation of furnace are as follows: 1. RLs are divided by adequate cut score lines to prevent cracks and peeling off of RLs due to loading of thermal stress. 2. Anchor systems supporting RLs at high temperature are important to maintain long lifetime. Therefore, heat-resistant alloys and sometimes ceramics shown in Figure 12 [17] are selected for anchor materials.

816

Handbook of Advanced Ceramics

TABLE 4 Main Resistance Properties of Typical Refractories for Damages in Waste Incineration Boiler Highalumina

Firecray

Thermal shock resistance

D



Cooling effect

D



Item

High-SiC

Alkarine resistance



Corrosion resistance for CaO

D

Abrasion resistance

D

Oxidation resistance : Exellent,

: good, D: Common, : Bad

FIGURE 10

Example of anchor systems in refractory title.

etc. due to its good workability, and can be installed as thinner linings than castable RMs. The design points important for application of plastic refractories are as follows:

FIGURE 9 Installation profile of unshaped refractory applied to waterwall of waste incineration boiler.

3. Slow drying rate at start-up is needed to prevent initial damages by quick water vaporization. Also, water penetration should be prevented at water cleaning in shut-down.

3.3.2. Plastic Refractories Plastic RMs can be applied for boiler tube surface, curved parts such as among boiler tubes, narrow parts

1. Dense microstructure of plastic RMs is favorable for corrosion protection, but thermal shock resistance somewhat reduces. Also, homogeneous installation by using hammer and rammer etc. is required to maintain mechanically strong properties for prevention of erosion/corrosion damage etc. 2. Support systems as same as castables are important to keep high durability of RLs. Recently, spray lining (gunning) methods have been commonly carried out to minimize working duration and to form homogeneous coated layer. Also, it is effective to give special properties such as corrosion resistance, antislugging etc. onto the surface of RLs.

Chapter | 10.3

Application of High-Temperature Corrosion-Resistant Ceramics

817

corrosion resistance, have progressed as shown in Figure 13. Details are described in the following.

4.1. Corrosion-resistant Metal Coatings for Waterwall Tubes

FIGURE 11 Example of estimated temperature distribution in air-cooled furnace structure.

The metal temperatures of the WWTs are relatively low depending upon the steam pressure as mentioned above. CRCs such as ceramic tile, metal spray coatings, weld overlay, or the cladding are applied on the tube surfaces of carbon steels or CreMo steels as shown in Figure 14. Table 5 shows the examples of application and durability evaluation of various CRCs. 1. Spray Coatings The metal spray coatings can be applied on site quickly with the thickness of approximately 150e500 mm. The actual application of Al/80Ni20Cr alloy flame spray coating on WWTs has started in 1985 [18]. Many tests in actual plants were conducted for WWTs and SHTs [19e24], and recently NiCrSiB alloy coatings by the high-velocity oxygen fuel (HVOF) process with high durability have been installed in many cases [25e27]. The coating systems (applied both onshop and on-site) are selected in accordance with the workability, the severity of corrosion environment, the required lifetime, and the cost. Recently fused NiCrSiB alloy spray-coating layer by induction heating has been successfully applied to WWTs in South Asian countries. This coating has a homogeneous, dense, and chemically bonded thick layer of 0.8e1.5 mm. Lifetime of this coating shows five times or more compared to that of above assprayed coatings due to no permeability of corrosive gases and completely chemically bonded interface [28]. 2. Weld Overlay and Composite Tubing

FIGURE 12 Typical refractory anchor systems for unshaped refractory materials applied in furnace wall.

4. ADVANCES IN HIGH-TEMPERATURE CORROSION-RESISTANT MATERIALS AND COATINGS In order to combat corrosion in above-mentioned severe environments, the development and application of advanced CRMs and CRCs, and also evaluation methods of the

A dense and thick (2e3 mm) coating layer chemically bonded with the base metal can be obtained by the weld overlay. The durability of alloy 625 weld overlays, that is 10 years or longer, is higher than that of as-sprayed coatings [29]. In many cases for the WWTs, automatic MIG and shielded arc welding are adopted for both on-shop and onsite installations [30]. For the SHTs, also PPW (plasma powder welding) are applied [31], and the tube bending can be done even after weld overlay. For the alloy 625 weld overlay applied in WWTs of 500  C/9.8 MPa highefficiency boiler, the maximum corrosion rates of approximately 0.1e0.2 mm/year have been observed by the 2-year field tests as shown in Figure 15 [32]. Mainly in Europe, composite tubings with cladding of CRMs (outer tube) to carbon steels and low-alloy steels (inner tube) are used for WWTs and SHTs. There are many application experiences of high Crehigh MoeNi base alloys including alloy 625 (Sanicro 63) [33].

818

Handbook of Advanced Ceramics

FIGURE 13 Trends in development and application of WTE material technologies.

FIGURE 14 Applications of corrosion resistant coatings to WTE boiler waterwall.

4.2. Corrosion-resistant Alloys for Superheaters The alloy design of seamless tubes for SHTs takes into consideration both corrosion-resistant and heat-resistant characteristics. Furthermore, excellent weldability, plastic workability etc. are required for fabrication of the

boilers. Generally, at steam temperatures of approximately 350  C or lower, carbon steel is used, while in the high-temperature region of 400  C or higher, CrMo steels stainless steels and Ni-base austenitic materials are used. The addition of alloying elements such as Mo, Nb, and Si to NieCreFe alloys is considered to be effective for corrosion resistance from the results of field corrosion

Chapter | 10.3

Applied Condition ( C ) Coating process Spray Coating

Chemical composition (Materials)

Parts

Metal temp.

Gas temp.

Durability (years)

Flame

Al /80Ni20Cr

WW

230 ~ 300

700 ~ 900

>3

Flame/ Fused

10Cr-Si, B-Ni base (12C)

SH

370 ~ 540

d

>1

Plasma

Plasma/ Fused

15Cr-Si,B, Fe-75Ni base

WW

230

700 ~ 800

4

HVOF

Hybride Cap

18Cr-5Fe-5Nb-6Mo -Ni base (Diamalloy)

SH

370 ~ 540

d

>0.4

DJ-1000

17Cr-4Fe-3.5B-4Si-Ni base (No-Fused)

WW SH

230 ~ 330

700 ~ 900

>3

D-Gun DJ-1000

50TiO2-50 Alloy625 (Cermet)

SH

430 ~ 460

500

>3

JP-5000

17Cr-4Fe-3.5B-4Si-Ni base

WW

230

700 ~ 900

>3

Ex-4000

Modified Alloy625

WW

d

d

>1

Modified PJ

ZrO2 / Alloy 625, NiCrSiB Alloy

SH

430 ~ 500

510 ~ 650

>3

MIG

21Cr-9Mo-3.5Nb-AI, Ti-Ni base (Alloy 625)

WW SH

270 ~ 330

700 ~ 1000

>10

PPW

18Cr-14Mo-4W-Ni base (C-276M, 625M)

SH

470 ~ 510

510 ~ 650

(z625)

MIG

23Cr-16Mo-1.6Cu-Ni base (HC-2000)

WW SH

260 ~ 480

870 ~ 1730

(0.52mm/y) (>625)

PAW

Alloy 625

SH

470 ~ 530

510 ~ 650

>1

Laser Cladding

625, HC-22, NiCr 309L, 686 etc.

WW

(Coal Fired Black Liquor)

Flame

Plasma Jet Weld Overlay

Application of High-Temperature Corrosion-Resistant Ceramics

TABLE 5 Example of Durability for Corrosion Resistant Coatings in WTE Boilers

d

Remarks: WW (Waterwall), SH (Superheater)

819

820

Handbook of Advanced Ceramics

because the difference of corrosion resistance in each NieCreFe alloy is small under the steam conditions below 400  C in relatively weak corrosion environments, as shown in Figure 16 [5]. Also alloy 825 and Sanicro 28 (composite tube) of 20e30% Cr 30e40% Ni alloys are sometimes used in a severely corrosive environment [36]. 2. High Crehigh MoeNi Base Alloys Alloy 625 and Sanicro 63 (composite tube) have good corrosion resistance under high-temperature steam conditions of higher than 450  C as shown in Figure 16. Also, it is known that HC-22 and JHN24 alloys, whose Mo content is higher than alloy 625, show excellent corrosion resistance. For high temperature applications of these high-Mo containing alloys, consideration of age deterioration of microstructures and mechanical properties is required. 3. High Siehigh Crehigh Ni Alloys FIGURE 15 Maximum corrosion thickness loss of alloy 625 weld overlay and NiCrSiB alloy spray coatings in 500  C /9. 8 MPa boiler.

tests [34]. The CRMs currently developed and applied are describedas follows [35]. 1. High Crehigh NieFe Base Alloys Austenitic stainless steels of 25Cre14e20Ni steels such as 309S, 310S, and 310HCbN, NF709 which are modified 310S alloys, are mainly used for 400  C/3.9 MPa boilers,

QSX3 and QSX5 are modified 310S alloys with added Si of approximately 3%. Also, it is known that Nicrofer 45TM demonstrates good corrosion resistance. MAC-F and MAC-N alloys containing 21e28%Cr, 3.2e4.2%Si, Ni, and Fe have been applied for four years in the 500  C/ 9.8 MPa SHTs [37]. The corrosion resistance varies according to the locations in the SHTs where corrosion environments are different. Therefore, it is necessary to make best use of the materials in accordance with the corrosion environment.

FIGURE 16 Change in maximum corrosion thickness loss of corrosion resistant superheater alloys with [Cr+Ni+Mo] concentration after 6000 hours exposure in metal temperature of 450  C and 550  C.

Chapter | 10.3

Application of High-Temperature Corrosion-Resistant Ceramics

5. FIELD EXPERIENCES OF CERMET AND CERAMIC COATINGS FOR SUPERHEATERS In order to improve the lifetime of SHTs at metal temperature of more than 430  C, challenging for application of cermet and ceramics spray coatings have been carried out in commercial plant that operates under 500  C/9.8 MPa steam conditions shown in Figure 17 [38]. TiO2$625 alloy cermet coating by HVOF, Cr3C2$NiCr alloy cermet, and ZrO2/Ni-base alloy dual layer coatings by the plasma jet process were exposed in SHTs that have metal temperature of 432e448  C for 1.3e2 years. Figure 18 shows three spray processes used in this investigations. In this chapter, the results of lifetime evaluation of coatings in actual boiler are described in detail [39,40].

5.1. TiO2$625 Cermet HVOF Coating Before field evaluations, 50 hours laboratory corrosion tests were carried out to select the adequate composition of cermet coatings. D-gun and HVOF are both a kind of supersonic frame spray process that forms dense layer. Addition of 30e50%TiO2 ceramic powder to alloy 625 improves largely corrosion resistance of coating in both spray processes as shown in Figure 19 and Figure 20. This effect is considered

FIGURE 17

821

to be due to increase in corrosion resistance and coating density with less than 0.5% porosity [39]. In order to prevent corrosion in tertiary (alloy 625) and secondary SHTs (310HCbN) affected by soot blowing, 50% TiO2$50%625 cermet coating that is selected in laboratory has been installed. Specimens sprayed on shop and on-site were both installed to the following three positions. In each position, on-site spraying was applied from the upstream side of the gas in the first row of SHTs where accelerated corrosion had been observed due to soot blowing attack [3]: l

l

Steam inlet and outlet portions of the tertiary SHTs (alloy 625 tube, OD.38.1  thickness 8 mm) Steam outlet portion of the secondary SHTs (310HCbN, OD.38.1  thickness 8 mm)

Erosion/corrosion resistance of the TiO2$625 cermet exposed for 2 years was clarified to be high compared to typical CRMs as shown in Figure 21. Also, corrosion thickness loss of layer reduced linearly as shown in Figure 22. Therefore, lifetime of coating was estimated to be approximately 4 years (27,000 hours) for a coating thickness of 200 mm. This mechanism was considered to be due to the formation of protective oxide layer, mainly including TiO2 and Cr2O3 from the EDX analysis results shown in Figure 23.

Test positions of cermet and ceramic coatings in superheater of WTE boiler.

822

Handbook of Advanced Ceramics

FIGURE 18 Three kinds of spray coating equipments applied in field tests.

5.2. Cr2O3/NiCr Cermet and ZrO2/Ni-Base Alloy Plasma-jet Coating 5.2.1 Test Procedure Figure 24 shows a system flow of a field-applicable ultrasonic plasma-jet spray (UPJ) equipment. This system, the Plazjet-250 (250 kW), is characterized by a higher output (100 kW), higher voltage, and lower current (200 VDC/500 A) compared with general atmospheric thermal spray (APS) equipment (40 kW, 50 VDC/800 A). The system can generate a supersonic N2/H2/Ar plasma jet whose velocity is 2,000 m/s by a water-cooled spray gun [40]. Two types of dual-layer coating with alloy 625 and NiCrSiB alloy as the under-coat and yttria-stabilized zirconia (ZrO2e8%Y2O3, hereinafter abbreviated as YSZ) as the top-coat were applied to the tertiary SHTs on site. In addition, a cermet single-layer of Cr3C2$75Ni25Cr (CrC$NiCr), were applied to the secondary SHTs. This cermet coating has been successfully used for prevention of erosion damage in commercial power generation boiler SHTs burning coals including much Cl [41].

The existing tube surface has been corroded with large pits mostly on the upstream side from the combustion gas. Then, the surface of those tubes was slightly ground and grid-blasted before field spraying, while a part of the corroded surface was blasted without any grinding in order to examine the effect of surface condition. Superheater tubes were sprayed for a quarter of the circumference (approx. 90 ) on the surface facing the gas upstream. Furthermore, new tubes were sprayed on shop for the entire surface of each tube, and were welded onto the front row SHs in each test portion. An inorganic sealant was only applied to a half area in length of the shop-sprayed tubes to examine the effect of the sealant. Table 6 shows a list of the temperature conditions and sprayed materials. Figure 25 shows typical microstructures of the UPJ coating layers tested. The application specifications for each UPJ layer are shown below: a. 625 (150 mm)/YSZ (150 mm) dual layer; b. NiCrSiB (150 mm)/YSZ (150 mm) dual layer: An inorganic sealant was applied on the YSZ layer; and c. CrCeNiCr (200 mm) single layer.

Chapter | 10.3

Application of High-Temperature Corrosion-Resistant Ceramics

823

FIGURE 19 Corrosion weight loss of various D-gun coatings in laboratory tests.

The following results were obtained regarding durability of the coatings in the field tests.

5.2.2. Field Examination Results Figure 26 shows the external appearance of an on-site sprayed 625/YSZ layer in the steam inlet portion of the tertiary SHT after 1.3 years of exposure. Both 625/YSZ and NiCrSiB/YSZ layers showed an almost sound condition in the upstream side of the gas. In sealant-coated positions, the sticking between the YSZ layer and deposits was weak and the deposits could be easily removed. Therefore, the sealant is thought to have an effect of preventing slugging of ash. Also in the steam outlet

portion of the tertiary SHTs, deterioration such as peeling, etc. was not found even after 1.3 years of exposure. The on-site sprayed layer was applied satisfactory even in the presence of many corrosion pits on the older surface, as shown in Figure 27. On the other hand, peeling occurred in the CrC$NiCr layer in the secondary SHT, particularly in the soot bloweraffected position after four months’ exposure as shown in Figure 28. Many hexagonal patterned cracks were found near the area of peeling, while in the same portion, limited erosion/ corrosion damage of the YSZ layer occurred in a part of both the 625 and NiCrSiB/YSZ coatings after 8 months’ exposure. The on-shop sprayed Ni base alloy/YSZ layer in the steam inlet portion of the tertiary SHTs maintains

824

Handbook of Advanced Ceramics

FIGURE 20

Corrosion weight loss of D-gun and HVOF coatings in laboratory tests.

FIGURE 21 Maximum corrosion thickness loss of D-gun coating and tube materials exposed for 13800 hours in secondary superheater. * Stem outlet of secondary SH: Tg ¼ 761 K, Tm ¼ 724 K, Stem inlet of secondary SH: Tg ¼ 734 K, Tm ¼ 702 K.

Chapter | 10.3

Application of High-Temperature Corrosion-Resistant Ceramics

825

FIGURE 22 Time dependence of maximum corrosion thickness loss of 50% alloy 625/50% TiO2 coating sprayed by D-gun (200 mm) in secondary superheater outlet.

a satisfactory sound condition even after one-year exposure as shown in Figure 29. The coatings showed a lower deterioration tendency in compared with the on-site sprayed layer.

5.2.3. Results of Lifetime Evaluation After one year and 1.3 years’ exposure, some on-site applied layers were partially cut off, and a detail examination was carried out. Figure 30 shows typical microstructures of the on-site sprayed 625/YSZ layer in the steam inlet portion of tertiary SHTs. The formation of small pores was found within the coating layer, but no corrosion occurred in the Ni base alloy/ YSZ interfaces. On the other hand, in the left and right side surfaces of the tube, porosity and cracking inside the coating layer were formed with coating thickness reduction. By the results of EPMA shown in Figure 31, a slight amount of Cl was found within the under-coated layer and interface in the sound coating layer. However, alkali and heavy metals that form low melting point components of the deposits were not found. It was clarified that the sealant and YSZ layer effectively act as a penetration barrier of corrosive components into the interface. Since the corrosive environments in the steam outlet areas of both tertiary and secondary SHTs were weaker than that in the steam inlet area of tertiary SHTs, the degree of deterioration and penetration of the corrosive components was small. Figure 32 shows the hardness changes of 625 and NiCrSiB under-coated layers with exposure time. The hardness of the NiCrSiB alloy shows an increase with the metal temperature, but the hardness of the alloy 625 remains almost a constant value. Table 7 shows a list of the microstructural examination results for each part. It is

thought that the lifetime of the Ni-based alloy/YSZ in the steam inlet part of the tertiary SHTs where corrosion conditions are most severe is 1.5 years or longer. From the many experiences that the durability of the conventional single NiCrSiB HVOF layers are several months in the same superheaters, it became clear that the durability of the Ni-base alloy/YSZ/sealant dual coating was considered to be remarkably improved. The durability of each spray coating is large in the order shown below including the results of TiO2$625 cermet HVOF coating. 625=YSZ  NiCrSiB=YSZ > TiO2 $625 Cermet HVOF > CrC$NiCr Table 8 shows the comprehensive evaluation results of the durability of the coatings after 1.3 years’ exposure. These coatings can prevent largely tube thickness losses in each part of the SHTs. It could be confirmed that the 625, and NiCrSiB/YSZ spray coatings have a good durability of 3 years or longer in a aggressive corrosion environment near a soot blower, which cannot be achieved by single layer of as-sprayed metal spray coating. Practical applications of these spray coatings are expected to contribute to continuous plant operation, and reductions in maintenance costs.

6. DETERIORATION MECHANISMS OF COATINGS 6.1. Metallic Materials Forming Protective Oxides Film Main corrosion factors influencing the corrosion rate are shown schematically in Figure 33. The gas temperature, especially the temperature gradient DT (gas temperature e metal

826

Handbook of Advanced Ceramics

FIGURE 23 Cross-sectional BSE images and EDX analysis of corrosion products in 50% alloy 625/50% TiO2 coating sprayed by D-gun exposed for 13800 hours in second superheater outlet position.

temperature), is considered to be the driving force for condensation of the vapor components in the gas. The chlorides concentration in the deposits shows a high level at locations where DT is large, and there is a tendency to form low melting deposits [42]. In addition, it is known that the amounts of Cl, SO4 and alkaline and heavy metals affect the physical properties of the deposits such as, molten phase amount, permeability, etc. [43]. The corrosion peak appears around 20% in the molten phase amount of the NaCl/KCl/Na2SO4 mixture as shown in Figure 34. The penetration of corrosive gas

components through the deposits and the presence of oxidizing constituents such as O2 are considered necessary for this kind of corrosion reaction. Moreover, severe thermal cycling acts on the tube surface due to gas temperature fluctuation and the use of soot blowers in the actual plants. Then, peeling-off and the regeneration of deposits and oxides layer are repeated. Figure 35 shows the EPMA results in the crack generation site of alloy 625 oxides layer at the soot blow-affected zone. The partial pressures of Cl and S

Chapter | 10.3

Application of High-Temperature Corrosion-Resistant Ceramics

827

FIGURE 24 Schematic illustration of 100 kW plasma-jet spray coating equipment. For color version of this figure, the reader is referred to the online version of this book.

TABLE 6 Test Conditions and Installed Positions of Ceramic Coatings in WTE Boiler Superheater Test positions

3rd superheater steam inlet

3rd superheater steam outlet

2nd superheater steam outlet

Tube Materials

Alloy625

Alloy625

310HCbN

Steam Temp.

445  C

500  C

450  C





Gas Temp.

621 C

510 C

488  C

Under / Top Coating Materials

, 625 / YSZ , NiCrSiB / YSZ

, 625 / YSZ , NiCrSiB / YSZ

, 625 / YSZ , NiCrSiB / YSZ , Cr3C2 , 75Ni25Cr

Test Bundles

, On-site (for used tube) : First tube in No. 2,3,4 bundles , In-shop (for new tube) : First tube in No. 9 bundle

FIGURE 25 Typical microstructures of spray coating layers before corrosion test. For color version of this figure, the reader is referred to the online version of this book.

FIGURE 26 General view of alloy 625/YSZ layer sprayed on site near soot blower in each superheater position after 1.3 years exposure. For color version of this figure, the reader is referred to the online version of this book.

828

Handbook of Advanced Ceramics

FIGURE 27 Microstructure of alloy 625/YSZ dual layer sprayed to corrosion pit of alloy 625 tube on site in tertiary superheater steam inlet position after 1.3 years exposure. For color version of this figure, the reader is referred to the online version of this book.

FIGURE 28 Deterioration of Cr3C2$NiCr layer sprayed on site near soot blower of secondary superheater after one year exposure. For color version of this figure, the reader is referred to the online version of this book.

FIGURE 29 Typical surface condition of dual coating layer sprayed on shop in tertiary superheater steam inlet position after one year exposure. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 10.3

Application of High-Temperature Corrosion-Resistant Ceramics

829

FIGURE 30 Typical circumferential microstructures of alloy 625/YSZ layer sprayed on site in tertiary superheater after 1.3 years exposure. For color version of this figure, the reader is referred to the online version of this book.

FIGURE 31 EPMA images of alloy 625/YSZ layer sprayed in tertiary superheater steam inlet position after one year exposure. For color version of this figure, the reader is referred to the online version of this book.

830

Handbook of Advanced Ceramics

FIGURE 32 Change in hardness values of under-coated layer with test duration. For color version of this figure, the reader is referred to the online version of this book.

TABLE 7 Microstructural Examination and Evaluation Results of Coatings in Each Position Circumferencial position of tube On-site sprayed layer under / top

Positions (tube materials)

Bundle no.

Secondary Superheater Stream Outlet (310HCbN)

4

Tertiary Superheater Steam Outlet (Alloy 625)

4

NiCrSiB / YSZ

3

625 / YSZ

4

NiCrSiB / YSZ

Tertiary Superheater Steam Inlet (Alloy 625)

3 4

Gas Flow Direction Left 35° Right 35°

Cr3C2,NiCr Single Layer

625 / YSZ

Effect of soot blower

Left 35

Gas upstream Right 35

Remarks

Large







Weak

~D

~D

~D

, Partial peeling-off near soot blower , Many londitudinal cracking

Large

D~,

,~

D~,

Weak

,

Large

D~,

Weak

,

Weak

, ,~

D~, ,

, Partial peeling-off of YSZ layer in side of tube , No corrosion scale formation of tube

: Totally sound condition ,: Deterioration of sealant and top coat (YSZ) D: Consumption of top coat and deterioration of under coat : Consumption of under coat

gases on the corrosion interface are considered to rise under the deposits due to penetration of chlorides and sulfates through the cracks. This type of corrosion is so called “molten salt induced corrosion”, because the corrosion reaction becomes active, when the amount of deposits increases and a part of deposits melts [44]. If the protective oxides layer breaks down, the direct reaction with molten chlorides may occur as shown in the schematic drawing of the alloy 625 corrosion scale in

Figure 36 [34]. The formation and self-healing of stable protective oxides layer, that can be considered to be like a kind of ceramic coating film, are important properties for CRMs applied to aggressive corrosion environments. From the configuration and properties of corrosion products distributing as chlorides, sulfides, and oxides from the nearest side to the interface, the corrosion in steady state is considered to be kept as a high temperature gaseous reaction such that chlorination/sulfidation/

Chapter | 10.3

Application of High-Temperature Corrosion-Resistant Ceramics

831

TABLE 8 Results of Lifetime Prediction for Ni Base Alloys/YSZ Coatings after 1.3 Years Exposure Position (materials)

2ry SH Steam Outlet (310HCbN)

Tertiary SH Steam Outlet (Alloy625) Tertiary SH Steam Inlet (Alloy625)

Gas temp./ steam temp. ( C)

488/450

510/500

Coating materials

Corrosion environment Corrosivity

Effect of Soot Blower

Position affected by SB

Position not affected by SB

Weak

Strong

>3

S1.5

YSZ/NiCrSiB

>3

S1.5

Cr3C2-NiCr

3

ee

>3

S1.5

>3

S1.5

YSZ/625

YSZ/625

Medium

Weak

YSZ/NiCrSiB 621/445

Lifetime of coatings

YSZ/625 YSZ/NiCrSiB

Strong

Strong

FIGURE 33 Schematic illustration for the role of main corrosion factors influencing to corrosion rate.

832

Handbook of Advanced Ceramics

FIGURE 34 Change in corrosion weight loss with molten phase content of deposits between coating test and crucible test in labolatory.

FIGURE 35 Breakdown of protective oxide layer and penetration of corrosive species in alloy 625 superheater tube influenced by soot blowing.

Chapter | 10.3

Application of High-Temperature Corrosion-Resistant Ceramics

833

FIGURE 36 Schematic illustration of corrosion mechanisms in NieCreMoe (Nb, Fe) alloy.

oxidation occur simultaneously. It is presumed that corrosion by the direct reaction with molten salt changes gradually to the gas reaction as the oxides layer grows. The influence of major factors such as temperature gradient, temperature fluctuation, molten ash amount, etc. to the corrosion rate was examined quantitatively by recent research [45]. Furthermore, the factors of characteristic intergranular corrosion should be investigated considering both the material (impurities, grain boundary precipitates, etc.) and the environment (stress, molten salt, etc.) [46].

6.2. Metal and Ceramic Spray Coatings Figure 37 shows the deterioration mechanisms and bonding strength reduction of the metal spray coating layers used for a long period in a boiler. Corrosion of the base material and deterioration of the coating layer proceed due to the corrosive gases penetration (HCl, etc.) onto the base material/coating interface. Then, “swelling” of the coating layer occurs, the reduction of adhesive strength is accelerated, and finally breaking down of the layer occurs [18]. Accordingly, dense coating is indispensable for improvement of lifetime, and HVOF is preferable process. The material factors that govern

durability are the corrosion resistance and open porosity of coating materials, the adhesive strength with base materials, the thermal properties such as thermal expansion coefficient, and the residual stress. The physical properties of the coating are largely dependent upon the spraying conditions. Recently, a quantitative evaluation method of lifetime by using small electric resistance methods [47] has been developed and applied, based on these deterioration mechanisms. In the case of dual ceramic coating of Ni base alloys/ YSZ, the same mechanisms are clarified in the field tests. The YSZ top-coated layer acts as a diffusion barrier for penetration of corrosive gas components shown in Figure 38 and is able to improve the lifetime of under coating and also base materials.

7. SUMMARY Recently, the WTE plants are required to satisfy many additional performances such as the suppression of pollutants, high electric power generation efficiency, material recycling, etc. Various combustion methods and plant systems have been developed, therefore a kind of such HTC environment increased more wide and diversified. Although the main factor influenced to HTC is considered to be almost

834

Handbook of Advanced Ceramics

FIGURE 37 Deterioration process of spray coating layer and reduce in bonding strength in severe corrosive environment.

FIGURE 38 Schematic illustration of deterioration mechanisms in Ni base alloy/YSZ dual layers.

Chapter | 10.3

Application of High-Temperature Corrosion-Resistant Ceramics

same for many corrosion environments for many kinds of WTE boilers. It is believed that the development and application of CRMs and CRCs aim strongly at the right material in the right place with a reasonable total cost. The improvement in performances and lifetime of the high temperature parts has been supported by the development of CRMs and CRCs technologies including both metallic and ceramic materials. It is also required to advance the application of ceramics under adequate corporation with metals and plant engineering. There are many subjects for future research for this kind of aggressive corrosion environment; it is expected that engineers and researchers in the field of CRMs and CRCs will take up the challenge.

REFERENCES [1] Kawahara Y. Recent trends and future subjects on high-temperature corrosion-prevention technologies in high-efficiency waste-to energy plants. Corros Eng 2005;54:213e35. [2] Kawahara Y, Kaihara Y. recent trends in corrosion-resistant tube materials and improvement of corrosion environments in the WTE plants. CORROSION /2001, Paper No.1173. Houston TX: NACE; 2001. [3] Kawahara Y. Formation and break-down of protective oxides layer in high temperature corrosion resistant materials used in environmental and energy equipments, Seminar Text of The Japan Institute of Metals 2006;99:46e55. [4] Kawahara Y. Recent trends in development and application of high-temperature corrosion-resistant materials and coatings in high efficiency waste incineration boiler. MTERE2 2002;41(3):190. [5] Kawahara Y, Nakamura M, Tsuboi H, Yukawa K. Evaluation of new corrosion resistant superheater tubings in high efficiency waste-to-energy plants. Corrosion 1998;54(7):576. [6] Kawahara Y, Orita N, Nakamura M, Ayukawa S, Hosoda T. Laboratory corrosion tests for simulating fireside wasteage of, superheater materials in waste incinerators. CORROSION /99, Paper No.89. San Antonio TX: NACE; 1999. [7] The Technical Association of Refractories, Japan. Refractories handbook, vol. 81; 1998. 310e11. [8] Kawahara Y. Improvement of durability and evaluation technologies for refractories in waste incinerator and industrial boiler. Taikabutu 2009;61(7):320e34. [9] Lea AC. J Soc Glass Technol 1949;33:127e507. [10] Iemura I. Shinagawa Giho 1979;23:73. [11] Horio T, Ikeda M, Hiragushi K, Fujiwara S, Aoyama K. Taikabutu Overseas 1981;2(1):3e14. [12] Shinohara Y. Advances of refractory materials. Japan: The Technical Association of Refractories; 1987. p.8e9. [13] Haurnek P, Mogensen E. Technical Report, Ho¨gana¨s AB, Sweden. [14] Kuchta HD. Interceram 1992;41(6):415e8. [15] Strach L, Lane L, Wasyluk DT. CORROSION 93, Paper No.219. Houston: NACE; 1993. [16] Kawamura K, Akagi K, Kaketa K. Fuel Combust 1988;53(12):9e17. [17] Plibrico Japan Co. Ltd. PLIBRICO monolithic refractories e New edition; 1994. p. 86e92.

[18]

[19] [20]

[21] [22] [23]

[24]

[25] [26]

[27]

[28]

[29]

[30] [31]

[32]

[33]

[34]

[35]

[36]

835

Kawahara Y, Kira M. Corrosion prevention of waterwall tube by field metal spraying in municipal waste incineration plants. Corrosion 1994;53:241. Kubin PZ. CORROSION/99, Paper No.90. San Antonio TX: NACE; 1999. Burnham DE, Gibbs DR, Gitlinger JS, Johnson HE. In: Proceedings of 1989 Conference on municipal solid waste as a utility fuel. Springfield; 1989. Oct. p. 10e12. Wright IG, Krause HH, Dooley RB. CORROSION/95, Paper No. 562. Orlando FL: NACE; 1995. Krause HH. CORROSION/92, Paper No.140. Nashville TN: NACE; 1992. Yoshikawa T. Spray coating and weld overlay in Circulating Fluidized Bed boiler (CFB) and Waste-To-Energy (WTF) boilers. In: Proceedings of International Symposium on High Temperature Oxidation and Corrosion/2005, Nara, Japan; 2005.10, p. 27. Rademakers PLF, Wetzel RWA, Kokmeijer E. In: Proceedings of the 6th Liege conference Part II, 5 Materials for Advanced Power Engineering, 1998, p. 751. Yamada Y, Koyama M, Otuka T, Osawa M, Toyama K. YosyaGijyutu 1994;14(3):30. Kawahara Y, Sasaki K, Nakagawa Y. Application of flame spray coatings of ni base alloy for waste incineration boilers. In: Proceedings of 43th Conference of JSCE Materials and Environments, vol. B-114; 1996. p. 195. Fukuda Y, Kawahara Y, Hosoda T. Application of high velocity flame sprayings for superheater tubes in waste incinerators. CORROSION/ 2000, Paper No. 264. Orlando FL: NACE; 2000. Matsubara Y, Sochi Y, Tanabe M, Takeya A. Advanced coatings on furnace wall tubes. J Therm Spray Technol 2007;16(2):195e201. June. Lai G, Hulsizer P. Performance of corrosion-resistant weld overlays in power boilers & kraft recovery boilers. In: Proceedings of Stainless Steel World/99. Netherlands: Den Haag; November 1999. p. 681. Brennan MS, Gassmann RC. CORROSION/2000, Paper No. 235. Orlando FL: NACE; 2000. Takeuchi Y, Kawahara Y, Hosoda T. Application of composite tubes plasma-weld-overlayed with NieCreMo alloy for superheater in waste incinerators. CORROSION /2000, Paper No. 261. Orlando FL: NACE; 2000. Kawahara Y, Orita N, Takahashi K, Nakagawa Y. Demonstration test of new corrosion-resistant boiler tube materials in high efficiency waste incineration plant. Tetsu-To-Hagane 2001; 87(No.8):24. Wilson A, Forsberg U, Lunderg M, Nylo¨f L. Composite tubes in waste incineration boilers. In: Proceedings of Stainless Steel World/99. Netherlands: Den Haag; November 1999. p. 669. Kawahara Y. High temperature corrosion mechanisms and effect of alloying elements for materials used in waste incineration environment. Corros Sci 2002;44:223e45. Kawahara Y. Application of high temperature corrosion-resistant materials and coatings under severe corrosive environment in wasteto-energy boilers. J Therm Spray Technol 2007;16(2):202e13. Haggblom E, Nylo¨f L. Materials for advanced power engineering, part II, 1994. Netherlands: Kluwer Academic Publishers; 1994. p. 1597.

836

[37] Kawahara Y, Sasaki K, Nakagawa Y. Development and application of high Cr-high SieFeeNi alloys to high efficiency waste-toenergy boilers. J Jpn Inst Met 2007;71(1):76e84. [38] Kawahara Y, Sasaki K, Yamada Y, Xiangyang Li, Hosoda T, Mizuko T. Demonstration test of corrosion-resistant tube materials in high efficiency waste-to-energy boiler, Hightemperature corrosion and protection 2000, In: Proceedings of the International Symposium on High Temperature Corrosion and Protection/2000, Hokkaido, Japan, Science Review; 2000 p. 591e9. [39] Fukuda Y, Kawahara Y. Evaluation of corrosion resistance of waste incinerator superheater tubes coated by high velocity thermal spray. J Jpn Inst Met 2002;66(4):563e8. [40] Kawahara Y, Nakagawa Y, Kira M, T.SakuradaKamakura H, Imaizumi Y. Life evaluation of ZrO2/Ni base alloy plasma-jet coating systems in high efficiency waste-to-energy boiler superheaters. CORROSION/2004, Paper No.04535. New Orleans LA: NACE; 2004. [41] Imaizumi Y, Kamakura H, Sakurada I. Denki-Genba-Gijyutu 2001,8;40(471):17.

Handbook of Advanced Ceramics

[42] Kawahara Y. Basic science and application of high temperature corrosion and protection in energy conversion and environmental equipments. JIM Seminar Text ISBN: 4-88903-129-4, C3057, p. 39; 2001. [43] Kawahara Y, Kira M. Effect of physical properties of molten deposits on high temperature corrosion of alloys in waste incineration environment. Zairyo-to-Kankyo 1997;46(1):8. [44] Spiegel M, Grabke HJ. High temperature corrosion of 2.25Cr-1Mo in simulated waste incineration environments. Mater Corros 1995; 46:121e31. [45] Kawahara Y. Evaluation of high-temperature corrosion life using temperature gradient corrosion test with thermal cycle component in waste combustion environment. Mater Corros 2006;57(1):60e72. [46] Uekado M, Kamei Y, Hujii K. Effect of static stress on high temperature corrosion behavior of boiler tubes in waste incineration environment. J Jpn Ins Met 2002;66(6):576. [47] Kawahara Y. Advanced Coating Technologies in Aggressive Corrosion Environment of WTE and Power Plants, In: Proceedings of International Symposium on high temperature oxidation and corrosion (ISHOC2010), Zushi, Japan, Nov. 2010, to be published.

Chapter 10.4

A New Thick Film Coating Technology-Laser Chemical Vapor Deposition Takashi Goto Institute for Materials Research, Tohoku University, 2-1-1 Katahira, Aoba, Sendai, Miyagi 980-8577, Japan

Chapter Outline 1. 2. 3. 4.

Introduction CVD for High-Speed Deposition YSZ Thermal Barrier Coating a-Al2O3 Coating for Cutting Tools

837 838 840 841

1. INTRODUCTION Nowadays, highly functional materials are used under severe conditions such as high-temperature, high-stress, and corrosive environments. Metallic materials have often been used as structural materials because of their superior properties such as high strength and ductility. Ceramic materials, in contrast, have inferior mechanical properties than metallic materials do. However, they have various functionalities such as high thermal/chemical stability and higher hardness than metallic materials have. Therefore, ceramic-coated metallic materials can function as good high-performance multifunctional materials. In general, coating technology can be divided into dry (vacuum) and wet (solution) processes [1]. The dry process is further categorized into two subtypes: physical vapor deposition (PVD) and chemical vapor deposition (CVD). In PVD, a source material (target) is transported to a substrate forming film by a physical process, such as ablation and sputtering, without an accompanying chemical reaction. The composition may be readily conserved from the target to the film, where the thickness of the coating can range from nanometers to several millimeters. Because PVD would encompass the sequential processes of collision, adsorption, and deposition of particles and/or clusters, it is fundamentally

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00045-9 Copyright Ó 2013 Elsevier Inc. All rights reserved.

5. HAp Coating for Dental Implants 6. Summary References

843 845 846

a “line-of-sight” process. Thus, the backside or inside of holes can hardly be coated by PVD. CVD, in contrast, accompanies the chemical reactions of source gases and takes place where the gases can arrive. CVD is then not a line-of-sight process and has good adherence and conformal coverage. However, it is a thermally activated process; thus, it requires a high temperature. As a result, it is difficult to use CVD for coating on nonrefractory substrates such as low melting point metals and polymers, but it can be used to prepare thermally stable refractory ceramic films even on substrates with complex shapes [2]. CVD has many deposition parameters, such as temperature, total pressure, partial pressure of each source gas, and the geometry of the CVD chamber (how to heat the substrate, etc.). By controlling these parameters, various microstructures, that is, amorphous fine-grained polycrystals and epitaxial single crystals, can be prepared [3]. CVD has wide-ranging applications, typically in semiconductor thin film devices and for coating on cutting tools. The deposition rate of CVD is generally a few to several tens of micrometer per hour. Thus, it is mainly applied to the thin film coatings. However, thick film coatings also have many useful industrial applications, and CVD can be applicable for such coatings. This chapter discusses the

837

838

Handbook of Advanced Ceramics

high-speed CVD used in the preparation of thick film coatings.

2. CVD FOR HIGH-SPEED DEPOSITION The main deposition process in CVD is the diffusion of gases in a gas phase and their chemical reactions on a heated substrate. A schematic diagram of the CVD deposition process is depicted in Figure 1 [4]. The source gases diffuse from the main gas stream to the substrate through the gas boundary layer and are absorbed/reacted on the substrate. Then, the byproduct gases desorb from the substrate and diffuse out to the main gas stream through the gas boundary layer. These sequential steps are all thermally activated processes; therefore, CVD is often termed thermal CVD. Figure 2 depicts the general relationship between the logarithmic deposition rate and the reciprocal deposition temperature [5]. In a low-deposition-temperature region, the rate-controlling step could be a chemical reaction on the substrate, where the deposition rate exponentially increases with increasing deposition temperature. The activation energy in this temperature region is generally several tens to hundreds of kilojoules per mol. In a higher-depositiontemperature region, the deposition rate still increases, while the activation energy becomes rather small, generally less than tens of kilojoules per mol, because the rate-controlling step could be mass transport (diffusion) of gases (source gas or byproduct gas) through the gas boundary layer. Therefore, to increase the deposition rate of the CVD so that a thick film is obtained, as depicted in Figure 2 (broken line), the region in which the chemical reaction occurs should be expanded to a high-temperature region while keeping source gases easily accessible to the substrate.

For the highest deposition rate, CVD can be conducted at a high deposition temperature under the mass transport controlled region. The source gases should be appropriately thermally stable in the high-temperature region without becoming depressed via a homogeneous reaction in a gas phase. Halide gas is commonly used because of its moderate stability at a high temperature. This reaction is called halide CVD. The chemical/thermal stability of substrate materials is also crucial to high-temperature deposition. Graphite substrate is often used for preparing nonoxide films in a nonoxidizing atmosphere. Nonoxide thick films, typically Si3N4 and SiC, are prepared using halide precursors (such as SiCl4 and CH3SiCl3) and graphite substrates at high deposition rates of 1e2 mm/h at 1700e1800 K [6,7]. Metalorganic (MO) compound precursors can also be used in CVD, termed Metalorganic Chemical Vapor Deposition (MOCVD), but MO precursors are easily decomposed and product films (mainly oxides) are reactive with the substrate material at a high temperature. Thus, the deposition rate of MOCVD is generally lower than that of halide CVD. The precursor gases are heated before arriving at the substrate, and the deposition reaction occurs in various places inside the CVD chamberdpartially in a gas phase to form powder and partially on a CVD chamber wall to form premature deposits. The deposition rate of CVD is therefore significantly affected by the manner in which the precursor gases and the substrate are heated, that is, whether the substrate is indirectly heated by the chamber wall or is directly heated without heating the CVD chamber wall. CVD is also categorized by the type of chamber wall, that is, the cold wall type CVD and the hot wall type CVD. Figure 3 depicts the schematics of cold and hot wall CVD

FIGURE 1 A schematic diagram of the deposition process of CVD.

Chapter | 10.4

A New Thick Film Coating Technology-Laser Chemical Vapor Deposition

839

FIGURE 2 Relationship between deposition rate and temperature.

chambers. In the cold wall type CVD, the substrate is directly heated by either radiofrequency or microwave induction heating and electric current joule heating. In these heating methods, the gas phase reaction (homogeneous reaction) can be minimal; only radiation from the substrate may cause the homogeneous reaction. The deposition rate and deposition efficiency are generally high, whereas the uniform temperature zone in the CVD chamber is rather narrow. In the hot wall type CVD, a wide area in the CVD chamber can be uniformly heated. Thus, a large number of substrate pieces or wafers can be simultaneously coated; however, the deposition rate is usually low, around several micrometers per hour. Lasers are versatile energy sources and have been applied to CVD, termed laser CVD. Laser CVD is a typical cold wall type CVD in which a laser directly irradiates the substrate (or only gas) but not the chamber wall [8]. Because a laser has high-energy photons, it can directly debond molecules photochemically, thereby causing deposition reactions. This process is called photolytic laser CVD. Because the deposition occurs without heating the substrate, CVD can be conducted at almost the room temperature. By using photolytic laser CVD, semiconductor device films have been prepared on n- or p-type Si single crystal substrates without affecting the substrate’s dopant distribution [8]. In contrast, a laser such as a CO2 laser or an Nd:YAG laser (infrared wavelength laser) has high thermal energy as that of a heat source. By focusing a laser beam on a specific area of a substrate, thermally activated chemical reactions and grain growth can occur in a localized area without a homogeneous reaction in a gas phase (lasers usually do not interact with gases), as

FIGURE 3 Schematics of cold (a) and hot wall CVD chambers (b).

depicted in Figure 4. In common thermal CVD, the deposition rate would be limited by diffusion in a gas phase in a high-temperature region, as shown in Figure 2. In pyrolytic laser CVD, the deposition zone (i.e. the laser beam size) is usually small; thus, the source gas can easily access the deposition zone. As a result, the deposition rate is not limited by diffusion and can be enormously high, more than several hundreds of meters per hour. Figure 5 demonstrates

FIGURE 4 Thermally activated chemical reactions that result from focusing a laser beam on the substrate surface.

840

FIGURE 5 Deposition rate of pyrolytic laser CVD as a function of laser power.

the deposition rate of pyrolytic laser CVD as a function of laser power [9]. However, it is believed that pyrolytic laser CVD can be used to prepare only thin rods, whiskers, and nanotubes (i.e. one-dimensional deposits) at a high deposition rate. Therefore, wide-area films or plates cannot be prepared at a high deposition rate by pyrolytic laser CVD. We have developed a new laser CVD to prepare thick films at deposition rates 100e1000 times greater than that of conventional CVD on wide-area and complex-shaped substrates [10]. The laser is broadly irradiated to a substrate and only the substrate surface is heated, that is, the entire substrate body is not heated. This methodology enables thick ceramic film coatings to be prepared on metal substrates at a high deposition rate. In this chapter, we discuss refractory thick film coatings on metal substrates using laser CVD for applying a thermal barrier yttriastabilized zirconia (YSZ) coating on an Ni-based superalloy, an antiabrasive a-Al2O3 coating on cutting tools, and a hydroxyapatite (HAp) coating on a Ti implant.

3. YSZ THERMAL BARRIER COATING YSZ is widely adopted for thermal barrier coatings (TBCs) on gas turbine blades made of an Ni-based superalloy because of its high thermal stability, low thermal conductivity, and a relatively large thermal expansion, which is close to that of the metal substrate. An Ni-based superalloy can withstand a temperature of around 1400 K, while the gas turbine can endure a temperature of around 1800 K. The temperature difference (several hundred degrees Celsius) between the working temperature of the gas turbine and the refractory temperature of the metal substrate is partially attributable to TBC. Atmospheric plasma spray or electron beam physical vapor deposition (EBPVD) has been applied to TBC [11]; however, the working temperature of TBC should be raised, and the

Handbook of Advanced Ceramics

lifetime should be extended by improving the quality of the YSZ coating. Thus, a new and more advanced coating process should be developed. CVD is advantageous for preparing a YSZ coating with good conformal coverage and adherence and is also a promising technique for TBC [12]. However, the precursor and oxygen source gases are generally reactive in gas phase to form powder (homogeneous reaction) at a high gas concentration, and the deposition rate of oxide films is generally too low (usually several tens of micrometers per hour) to prepare a TBC coating with a thickness of around several hundreds of micrometers. Laser CVD can be used for preparing YSZ thick film coatings at high deposition rates without causing the homogeneous reaction in the gas phase because the laser would not generally interact with the source gases. Figure 6 depicts the cross-section of the YSZ coating prepared by laser CVD [13]. A typical columnar texture, similar to that of EBPVD, can be prepared. Although the thermal expansion coefficient of YSZ is close to that of an Ni-based superalloy, a small difference still exists between the two, which might have caused thermal stress and delamination at the interface. The columnar texture is vertically aligned. Furthermore, the crack formation resulting from severe thermal stress would extend vertically as well. This tendency would prevent catastrophic delamination of the entire coating. Figure 6 (b) shows a magnified view of the columnar grains. Many voids exist at the columnar boundary. Similar voids are also observed in the YSZ coatings by EBPVD, and they are known as being effective in the relaxation of thermal stress at the interface. The CVD of YSZ films has been widely investigated because YSZ is used for oxide ion conductor films and as a buffer layer between a metal tape and a YeBaeCueO superconductor film. Figure 7 demonstrates the deposition rates of YSZ films as a function of deposition temperature by conventional CVD [14e18]. Attempts have been made for preparing a YSZ coating using CVD so that the coating can be applied as TBC. A deposition rate of 50e100 mm/h has been attained by employing a cold wall type CVD. However, the conventional CVD still does not have a suitable deposition rate for practical application. In contrast, the deposition rate of YSZ film by laser CVD is several to hundred times greater than that of conventional CVD. The chemical reaction around the substrate can be highly activated by laser irradiation. Therefore, the rate-controlling step of the deposition process is not a chemical reaction but rather a diffusion reaction (mass transfer) in a gas phase with low activation energy. Figure 8 demonstrates the nanostructure of the columnar grains of the YSZ coating by laser CVD [5]. A large number of nanopores of about a few nanometers in diameter can be observed in the grains. Figure 9 depicts the relationship between the deposition rate and the thermal

Chapter | 10.4

A New Thick Film Coating Technology-Laser Chemical Vapor Deposition

841

FIGURE 6 Cross-section of the YSZ coating prepared by laser CVD (b) shows the magnified view of the columnar grains of (a).

FIGURE 9 Relationship between deposition rate and thermal conductivity of YSZ coating by laser CVD. FIGURE 7 Deposition rates of YSZ films as a function of deposition temperature by conventional CVD.

conductivity of the YSZ coating by laser CVD [12]. The number of nanopores (i.e. density) is almost proportional to the deposition rate; in contrast, thermal conductivity is inversely proportional to the deposition rate. The thermal conductivity of the YSZ coating was one-third to onefourth that of YSZ bulk ceramics. Thus, this coating is a promising candidate for TBC.

4. a-Al2O3 COATING FOR CUTTING TOOLS

FIGURE 8 Nanostructure of the columnar grains of the YSZ coating by laser CVD.

WCeCo cemented carbide has been widely used as a cutting tool because of its high hardness and ductility. It is commonly coated with a-Al2O3 thick film to protect the substrate from thermal and mechanical degradation. CVD is most commonly used for the a-Al2O3 coating [19]. Al2O3 has many poly types, mainly a, g, k, q, d; a-Al2O3 is the

842

Handbook of Advanced Ceramics

FIGURE 10 Deposition rates of Al2O3 coating as a function of deposition temperature by conventional and laser CVD.

most stable and hardest among them. The volume change by phase transformation at a high temperature during service will cause delamination of the Al2O3 coating on WCeCo. Therefore, the top coat on WCeCo should comprise the most stable a-Al2O3 coating [20]. Because a-Al2O3 is a high-temperature form of Al2O3, the deposition temperature for preparing its coating should be generally high, that is, >1200e1300 K. Such a high deposition temperature will often cause degradation of the WCeCo substrate; thus, the deposition temperature should be lowered. W is also a rare resource, and the substitution of WCeCo into TiNeNi cermet has been investigated. The deposition temperature should be further lowered to

prepare the a-Al2O3 coating on TiNeNi cermet because the outward diffusion of Ni and many interface reactions become significant at high temperatures. Therefore, a lowtemperature and high-speed a-Al2O3 coating process should be developed. We have applied laser CVD to prepare the a-Al2O3 coating for enabling low-temperature and high-speed deposition [21e23]. Figure 10 depicts the deposition rates of the Al2O3 coating as a function of deposition temperature by both conventional CVD and laser CVD [24e28]. The a-Al2O3 coating has been commercially prepared by halide CVD using mainly AlCl3 as the Al source and CO2 gas as an oxidant. This coating has commonly been prepared by halide CVD at >1300 K. A strong temperature dependence of deposition rate whose activation energy is 90e170 kJ/mol is characteristic of halide CVD, as depicted in Figure 10. The MOCVD has also been applied to prepare the Al2O3 coating, and amorphous and the g-Al2O3 coating has often been prepared. The deposition rates by MOCVD are higher than those by halide CVD at low temperatures 1014 U cm

Breakdown voltage

150 V/mm

Vickers hardness

9.2 Gpa

Adhesion force

80 MPa

Chapter | 10.5

Aerosol Deposition Method for Room-Temperature Ceramic Coating and Its Applications

855

FIGURE 8 Before and after exposed to plasma, the surface morphology of yttrium oxide films produced by various methods.

that produced by the other processes, i.e. the AD-yttrium oxide layer was free from pores, although noticeable pores were found in the yttrium oxide samples fabricated by the other processes. This result demonstrates that the AD method considerably improved the antiplasma corrosion and the surface smoothness of yttrium oxide. Owing to the attainment about the AD method, we achieved the antiplasma corrosion, the wear resistance, and the reduction of the adsorption gas and the particle emission that are required for the next-generation semiconductor manufacturing industries. Figure 9 shows the large area (50  50 cm2)

AD-yttrium oxide coating and the electrostatic chuck produced by TOTO Co. Ltd [18]. The success of the development is due to the distinct feature of the AD method that can form good-performance ceramic layers for the antiplasma corrosion and the wear resistance at low cost.

7. APPLICATION TO MEMS DEVICES Because piezoelectric materials act both as sensors and as actuators, they have a wide range of potential applications,

FIGURE 9 The large area (50  50 cm2) AD-yttrium oxide films and the electrostatic chuck produced by the AD method. For color version of this figure, the reader is referred to the online version of this book.

856

such as ink-jet printers, high-speed actuators for nanopositioning, and micro-, ultrasonic devices. To realize integrated microdevices with piezoelectric materials, thinlayer deposition technologies and associated microfabrication methods are being actively studied in the research fields of micro-electro-mechanical systems (MEMSs) and micro-total analysis systems (m-TASs). The required thickness of piezoelectric layers for these applications is from 1 to about 100 mm (this thickness range is called “thick layers”). High-speed resonance-type optical microscanners are expected to be a key component in future laser displays and retinal projection-type displays. For such applications the requirements are scanning frequency greater than about 30 kHz, scanning angle greater than 20 , mirrors sufficiently large to prevent distortion, and reduction of drive voltage. It is thought that piezoelectric layers will be desirable for such applications because of their simple structure and high actuation force. However, these devices are not easy to make using conventional thin-layer deposition technologies combined with MEMS fabrication processes, because of low deposition rates and complicated etching processes.

Handbook of Advanced Ceramics

Using the AD process, a high-performance optical microscanner with a scanning speed at a resonance frequency over 30 kHz and a scan angle (peak-to-peak value) over 30 in atmospheric environment was successfully fabricated by the deposition of the piezoelectric materials at a high rate onto the scanner structure fabricated by Si-micromachining, as shown in Figure 10 [19]. Neither mirror bending nor deformation of the scanning laser beam was observed by the thick structure of the microscanner device. This scanner performance was higher than that of scanners made with conventional methods. However, MEMS process technology has several problems: processes are complicated, products are costly, and many facilities are needed. In addition, Si-based optical microscanners are made from silicon and can be easily broken due to mechanical shock and high stress concentration near the mirror. It is very attractive to use a metal substrate instead of silicon in optical microscanner devices because of a reduction of the device cost and an improvement of the ambient durability by changing brittle silicon to ductile stainless steel. Figure 11 (a), (b) shows a schematic and an optical image of

FIGURE 10 Optical microscanner driven with PZT thick layer deposited on Si-MEMS structure by AD method. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 10.5

Aerosol Deposition Method for Room-Temperature Ceramic Coating and Its Applications

857

FIGURE 11 (a) Principle of Lamb wave driving mechanism; (b) Sample image; and (c) projection image of laser TV with Lamb wave optical MEMS scanner. For color version of this figure, the reader is referred to the online version of this book.

a metal-based optical microscanner. An optical mirror was attached using two narrow torsion hinges from both sides at the center of gravity of the mirror (symmetrical mirror structure). The other sides of these hinges were connected to the scanner frame and one edge of the scanner frame was cramped and fixed with a heavy block. A piezoelectric layer was directly deposited on the stainless-steel scanner frame beside the mirror. The d31 actuation mode of the piezoelectric layer bends the scanner frame and induces the vibration of a ‘‘Lamb wave’’ in the scanner frame [20]. Lamb waves are two-dimensionally guided and concentrate at the narrow torsion hinges holding the mirror through the specially designed metal scanner structure. The waves change with the torsion vibrations of the mirror caused by the vertical difference in the direction between progressing waves and the torsion hinge. The torsion vibration vertically scans the mirror axis. A large optical scanning angle (80 ) was achieved at a high resonance frequency (30 kHz) at a driving voltage of 15 V (peak-to-peak) in ambient air using the prototype scanner device. Then low-power consumption laser projection system using this optical microscanner was obtained in Figure 11 (c). In addition, diaphragm actuators [21] used for micropumps can be made from thin Si membranes deposited with the AD method and micromachined. These have flat frequency response and good symmetrical elastic deformation over a wide frequency range. During fatigue testing (stability of long time performance) under high driving

electrical fields, there was neither de-polarization nor peeling of the AD deposited layer from the substrate [13]. The amplitude of this actuator was 25 mm at a resonance frequency of 22.4 kHz and a driving voltage of 8 V. This performance is suitable for applications using micromixers and micropumps.

8. POTENTIAL OF ENERGY APPLICATION AD method allows room-temperature coating of ceramics materials; therefore, the thermal diffusion at the interface between ceramic and metal/polymer in the material integration can be reduced. The realization of low-cost and high-throughput wide area coating for energy related devices can be expected. Figure 12 (a) shows the all solidstate Li-ion battery with LCO/LATP/LTO structure, which was fabricated by AD method at room temperature. The charge and discharge properties were confirmed [22]. Figure 12 (b) shows the superconductive material MgB2 thick layer deposited at room temperature by AD method. The phase transition of superconductivity was observed at 25e33 K [23]. These results indicate the potential of AD method application to energy-related devices. In the ceramic fuel cell (SOFC), the interconnected electrode AD layers assisted to improve the device lifetime, and the ionic conductive electrolyte AD layer reduces the internal resistance. Now, other applications such as CIGS solar cell, TiO2 dye-sensitized solar cell (DSSC), and thermoelectric conversion devices are going to be fabricated.

858

Handbook of Advanced Ceramics

FIGURE 12 Energy-related applications of AD method: (a) all solid-state Li-ion battery with LCO/LATP/LTO and (b) superconductive AD sheet (MgB2). For color version of this figure, the reader is referred to the online version of this book.

9. FUTURE PROSPECTS FOR USING AD METHODS IN MATERIAL INTEGRATION TECHNOLOGY Figure 13 shows the application load map of AD method toward the future. The application in left-lower (small size and high cost) area is microelectric devices, which have been researched and developed since 2000e2008 in NEDO national project “Nano structure forming for advanced integration ceramic technology” [24]. Now application developments are going on the right-higher area toward the future, which indicates the large-size and low-cost devices such as energy-related devices. The AD method is a nonthermal equilibrium process, which solidifies feed particles at room temperature to form layers. This differs from thermal spray-coating methods, which use higher temperatures for layer deposition. There are two important features of the AD method. One is that the AD deposition rate is higher than that of conventional thin-layer processes, because the deposition material refers to particles,

which carry substantially more mass than the molecules used in conventional methods such as sputter deposition. In addition, it is easy to form complicated oxide thin layers with the AD method, because the crystal structure of the starting particles is preserved during deposition. Therefore, the AD method is useful for developing composite or integrated materials with various ceramics, metals, and polymers. However, because many defects are introduced during deposition, a process such as annealing is needed to achieve acceptable electrical properties of the deposited layers. One of the origins for these defects may be that the crystals in the starting particles are crushed during collision with the substrate. Or it may be that surface defects of the starting particles are introduced into the deposited layer. Because integration of ceramics with low-melting-point materials is important for a broad range of applications, reduction of process temperature is a requirement of future fabrication techniques. To achieve these goals with the AD method, the deposition mechanism, formation of defects in

Chapter | 10.5

Aerosol Deposition Method for Room-Temperature Ceramic Coating and Its Applications

859

FIGURE 13 Application load map of AD method. For color version of this figure, the reader is referred to the online version of this book.

the layer, and recovery of electrical and mechanical properties by post-deposition annealing will be investigated in detail. Establishment of methods for controlling the particle size and degree of aggregation of the starting particles is also important.

REFERENCES [1] Akedo J, Lebedev M. Ceramics coating technology based on impact adhesion phenomenon with ultrafine particles e aerosol deposition method for high speed coating at low temperature. Materia 2002;41(7):459e66 (in Japanese). [2] Akedo J, Lebedev M. Microstructure and electrical properties of lead zirconate titanate (Pb(Zr52Ti48)O3) thick film deposited with aerosol deposition method. Jpn J Appl Phys 1999;38(9B):5397e401. [3] Akedo J. Aerosol deposition of ceramic thick films at room temperature -densification mechanism of ceramic layer. J Am Ceram Soc 2006;89(6):1834e9. [4] Kiyohara M, Tsujimichi Y, Mori K, Hatono H, Migita J, Kusunoki T, et al. Anti-wear coating fabricated by aerosol deposition method for magnetic card reader. In: Proc. of 15th Ceram. Soc. Jpn Autumn Symp, 228; 2002.

[5] Levedev M, Akedo J, Mori K, Eiju T. Simple self-selective method of velocity measurement for particles in impact-based deposition. J Vac Sci & Tech A 2000;18(2):563e6. [6] Akedo J, Lebedev M. Influence of carrier gas conditions on electrical and optical properties of Pb(Zr, Ti)O3 thin films prepared by aerosol deposition method. Jpn J Appl Phys 2001;40:5528e32. [7] Akedo J, Lebedev M. Powder preparation for lead zirconate titanate thick films in aerosol deposition method. Jpn J Appl Phys 2002;41:6980e4. [8] Lebedev M, Akedo J. Patterning properties of lead zirconate titanate (PZT) thick films made by aerosol deposition. IEEJ Trans. 2000;120-E(12):600e1. [9] Akedo J. Study on rapid micro-structuring using jet-molding: -Present status and structuring subjects toward HARMST-. Microsystem Technologies 2000;6(11):205e9. [10] Akedo J, Park J-H, Tsuda H. Aerosol deposition e Film formation, fine patterning of piezoelectric materials and its applications. J Electroceram 2009;22(1e3):319e26. [11] Ide T, Ito T, Hatono H, Kiyohara M, Lebedev M, Akedo J. Alumina layer fabricated by aerosol deposition method for electrostatic chuck. In: Proc of 15th Ceram Soc Jpn Autumn Symp, 229; 2002. [12] Akedo J, Lebedev M. Piezoelectric properties and poling effect of Pb(Zr, Ti)O3 thick films prepared for microactuators by aerosol deposition. Appl Phys Lett 2000;77:1710e2.

860

[13] Akedo J, Lebedev M. Effects of annealing and poling conditions on piezoelectric properties of Pb(Zr0.52, Ti0.48)O3 thick films formed by aerosol deposition method. J Cryst Growth 2002;235: 397e402. [14] Oh S-W, Akedo J, Park J-H, Kawakami Y. Fabrication and evaluation of lead-free piezoelectric ceramic LF4 thick film deposited by aerosol deposition method. Jpn J Appl Phys 2006;45(9B):7465e70. [15] Nam S-M, Yabe H, Kakemoto H, Wada S, Tsurumi T, Akedo J. Low temperature fabrication of BaTiO3 thick films by aerosol deposition method and their electric properties. Tran of MRS Jpn 2004;29(4):1215e8. [16] Ryu J, Choi J-Jin, Hahn B-Dong, Park D-S, Yoon W-Ha, Kim KHoon. Fabrication and ferroelectric properties of highly dense lead-free piezoelectric (K0.5Na0.5)NbO3 thick films by aerosol deposition. Appl Phys Lett 2007;90:152901e3. [17] Suzuki M, Noguchi Y, Miyayama M, Akedo J. Polarization and leakage current properties of bismuth sodium titanate ceramic films deposited by aerosol deposition method. J Ceram Soc, Japan 2010;118(1382):899e902. [18] Iwasawa J, Nishimizu R, Tokita M, Kiyohara M, Uematsu K. Dense yttrium oxide film prepared by aerosol deposition procee at room temperature. J Ceram Soc Japan 2006;114(3):272e6 (in Japanease). [19] Asai N, Matsuda R, Watanabe M, Takayama H, Yamada S, Mase A, et al. A novel high resolution optical scanner actuated by

Handbook of Advanced Ceramics

[20]

[21]

[22]

[23]

[24]

aerosol deposited PZT Films; 2003. Proc. of MEMS 2003, Kyoto, Japan, p. 247e250. Akedo J, Lebedev M, Sato H, Park J- H. High-Speed optical microscanner driven with resonation of lamb waves using Pb(Zr, Ti)O3 thick films formed by aerosol deposition. Jpn J Appl Phys 2005;44(9B):7072e7. Lebedev M, Akedo J, Akiyama Y. Actuation properties of lead zirconate titanate thick films structured on Si membrane by the aerosol deposition method. Jpn J Appl Phys 2000;39:5600e3. Akedo J, Popovici D. Fabrication of an all-solid-state thin-film lithium-ion battery prototype using a room temperature process e Towards the full-scale development of a novel technique for producing secondary batteries using aerosol deposition. AIST http://www.aist.go.jp/aist_e/latest_research/2010/ press-release, 20101224/20101224.html; 5, November (2010). Akedo J, Sakata H, Lebedev M. Microstructure and superconductivity of dense MgB2 thick layer on metal and Si substrate formed at room temperature with novel method: aerosol deposition method; 2003. The 5th International Meeting of Pacific Rim Ceramic Societies Incorporating the 16th Fall Meeting of the Ceramic Society of Japan, Nagoya, Japan. Akedo J. Room temperature impact consolidation (RTIC) of fine ceramic powder by aerosol deposition method and applications to microdevices. J Therm Spray Tech 2008;17(2):181e98.

Chapter 11.1.1

Ceramic Powders for Advanced Ceramics: What are Ideal Ceramic Powders for Advanced Ceramics?* Shigeyuki Somiya,y Sridhar Komarneni** and Rustum Roy**, z y

Professor Emeritus of Tokyo Institute of Technology, Japan, ** Materials Research Institute, Materials Research Laboratory building, University Park, Pennsylvania 16802, United States, z Passed away on August 26, 2010

Chapter Outline 1. 2. 3. 4.

Introduction Characteristics of Powders Methods to Produce Fine Ceramic Powders Hydrothermal Syntheses and Characteristics of Hydrothermal Powders

863 864 866

5. Summary Acknowledgments References

879 879 879

871

1. INTRODUCTION Powders play a key role in many materials, such as ceramics, catalysts, cosmetics, food, and medicine. Many books and articles have reported on the particles and powders used for sintered ceramic bodies as final products: for example, there are articles described or edited by C. R. Veale [1]; A. Kato and T. Yamaguchi [2]; P. Vincenzini [3]; D. W. Johnson, Jr. [4]; D. L. Segal [5a,5b]; S. Saito [6]; S. Shirasaki and A. Makishima [7]; G. Jimbo [8]; R. J. Brook [9]; S. Somiya, T. Hayashi, and K. Asaga [10]; D. Ganguli and M. Chatterjee [11]; TIC Editorial Office [12]; and P. Vincenzini and J.-F. Baumard [13]. There are many scholars who mentioned important points of ceramic powders for ceramic-sintered materials. According to A. I. Aksay of the University of Washington [10], presently at Princeton University, and who is well known for work on ceramics before firing and mullite

* A part of this article was originally published in Transactions of the Materials Research Society of Japan, 35(3), 473 (2010). This article is dedicated to Professor Rustum Roy.

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00047-2 Copyright Ó 2013 Elsevier Inc. All rights reserved.

ceramics, the important powder characteristics are as follows: 1. 2. 3. 4.

Pore distribution. Particle size distribution. Shape of the particles. Size of the particles.

As for H. K. Bowen of the Massachusetts Institute of Technology (MIT) [10], presently at Harvard University, and who is an author of “Introduction to Ceramics, 2nd Edition,” Wiley, New York, the following are important: 1. 2. 3. 4. 5.

Use of fine powders. Narrow particle size distribution. No agglomeration. As much sphericity as possible. Uniformity of chemical composition.

According to J. A. Pask of the University of California, Berkeley, California [10], who was one of the pioneers of modern ceramics, the important characteristics are as follows: 1. Absence of coagulation of particles. 2. Particle size.

863

864

Handbook of Advanced Ceramics

R. Roy of The Pennsylvania State University [14,10] developed the solegel process in 1949e1956, which has been cited formally and informally >30,000 times for making fine and pure powders of virtually any composition in ceramics. He stated the following:

5. Ease of producing powders, such as oxides, carbides, and nitrides because it is easy to control the atmospheric condition for producing powders.

1. Control of chemical composition. 2. Particle size. 3. Simplicity (of processing).

1. Both single crystal and polycrystal powders with crystallinity and crystal structure that are easily controlled. 2. Multicomponent powders. 3. Both homogenous powders (i.e. powders with uniform chemical composition) and coated powders (i.e. powders coated with various components). 4. Both nonaggregate and aggregate particles with different porosities including nonporosity.

According to O. J. Whittemore, Jr. of the University of Washington [10], who studied ceramic products, the important characteristics are as listed below. 1. Size of the pores. 2. Shape of the pores. 3. Distribution of the pores. A. Krauth who worked at Rosenthal, Germany (a famous company for Christmas decorative dishes) [10], stated the following: 1. Preferable to be able to sinter green bodies 100  C. Hydrothermal reaction occurs in water (liquid H2O). In addition, there is a possibility that water vapor (gas) reacts with substances by hydrothermal reaction. For typical hydrothermal research [4,25,36e46], one needs a high-temperature and high-pressure apparatus called an autoclave or a bomb. A great deal of experimental work has been carried out using Morey bombs [25,36,37a] and Tuttle-Roy test tube bombs [20,25,37a] (made by TemPres Division, LECO Corp. [47]) shown in Figure 1 [37a] and Figure 2 [37a]. There are many articles related to autoclaves and bombs, and these vessels are extremely important in hydrothermal experiments. Thus, we explain these vessels in Section 4.1. There are two types of systems that use the autoclave or the bomb: the batch system and the continuous system. In general, people in the industry prefer the continuous system while those in academic research prefer the batch system. These systems are explained in Section 4.2. As for heating, there are also two types: an external system and an internal system. These systems are explained in Section 4.3. Relating to pressuring, namely, the increase of the water vapor pressure in the autoclave or the bomb, there are also two methods: one involves heating and the other uses a pump or an intensifier. Water becomes vapor at 100  C under atmospheric pressure, but it still remains liquid at a high temperature under high pressure unless the conditions exceed the critical point. Therefore, solubility in water at high temperatures under high pressures is different from that in ordinary water. The solubility of almost all materials increases as temperature increases. This is one of the reasons for using the hydrothermal synthesis method.

FIGURE 1 Autoclave with a flat plate closure (Morey type). Laudise and Nielsen (1961). Solid State Physics Vol. 12, p. 163. [37a]).

In addition, the rate of reaction becomes high at high temperature. The action of water depends on its surrounding conditions as described below: 1. Water functions as a solvent in a reaction. 2. Water functions as a starting material, namely, a substance that reacts with other starting materials. 3. Water accelerates the rate of a reaction as a catalyst or a mineralizer (a reaction promoter) to rearrange or recrystallize the atoms and ions of products. 4. Water functions as the medium and regulates temperature and pressure. In the case of water vapor or steam, steam acts a medium to regulate temperature and pressure. In addition, there are cases in which steam functions as a starting material, or a catalyst, or a mineralizer. In other words, in some cases, hydrothermal synthesis involves H2O as a component of reacted products (i.e. solid phases) in the synthesis or a catalyst at elevated temperatures (>100  C (or 60  C)) and pressure (more than a few atmospheres) [44c]. The hydrothermal synthesis method is used for the synthesis of materials and the crystal growth of materials under high pressure and temperature of water. Section 4.4 deals with starting materials of hydrothermal reactions. As

872

Handbook of Advanced Ceramics

FIGURE 2 Reaction vessel with a cold cone seat closure (TuttleeRoy type). (Laudise and Nielsen (1961). Solid State Physics Vol. 12, p. 164. [37a]).

shown in Table 3 [20,21], there are many types of reactions, and these reactions are explained in Section 5.7 of the first edition of this handbook. Moreover, example reactions are briefly described in Section 4.5. Characteristics of hydrothermal processing are shown in Table 4 [5,19,20]. In Section 4.6, zirconia-based powders are explained as example products of the hydrothermal reaction, hydrothermal homogenous precipitation. In addition, Section 4.7 describes the characteristics of hydrothermal powders and comparisons between hydrothermal and ideal powders.

4.1. Autoclaves or Bombs According to Laudise and Nielsen [48], the most important apparatus for hydrothermal crystal growth experiments are autoclaves and closures. The ideal hydrothermal autoclave should have the following characteristics: (a) inertness to acids, bases, and oxidizing agents; (b) ease of assembly and disassembly; (c) ruggedness in order to require no treatments between runs; (d) enough length to produce large internal temperature differences within the autoclave and

a diameter that is large enough to grow crystals of a sufficient size; (e) no leaks; and (f) unlimited pressure and temperature in the autoclave. Obviously, no such vessel exists even though vessels to fulfill one or some of the above criteria are available. Table 5 [37a,44b] lists the limits of pressure and temperature of autoclaves, and Table 6 [25,36,37a,44b,46,49] lists sealing methods for autoclaves. One needs an autoclave or a bomb in order to do typical hydrothermal research [4,25,36e46]. There are Morey bombs and Tuttle bombs or Tuttle-Roy bombs, and their names have originated from the names of researchers of hydrothermal work and inventors of autoclaves, Morey, Tuttle, and Roy. The TuttleeRoy bomb is also called the TuttleeRoy test tube reactor or the TuttleeRoy test tube bomb. Early studies related to hydrothermal systems at high temperature and pressure were carried out by Morey and coresearchers at the Geophysical Laboratory of the Carnegie Institution of Washington, Washington, DC, since 1913 [50].

Chapter | 11.1.1

Ceramic Powders for Advanced Ceramics

TABLE 3 Hydrothermal Reactions [20,21] Hydrothermal synthesis

873

TABLE 5 Limits of Pressure and Temperature of Autoclaves Pressure (atm)

Temperature (oC)

Pyrex (5 mm as inner diameter, 9 mm as outer diameter)

0.6

250

0.6

300

6. Hydrothermal sintering

Pyrex (5 mm as inner diameter, 9 mm as outer diameter)

7. Hydrothermal reaction sintering

Morey type (flat plate seal)

400

400

8. Corrosion reaction

Welded WalkereBuehler

2000

480

9. Hydrothermal oxidation

Delta ring

2300

400

10. Hydrothermal precipitation

Bridgeman seal

370

500

11. Hydrothermal decomposition

Modified Bridgeman seal

3700

500

12. Hydrothermal electrochemical reaction

Cold test tube

1. Hydrothermal crystal growth 2. Hydrothermal treatment 3. Hydrothermal alteration 4. Hydrothermal dehydration 5. Hydrothermal extraction

13. Hydrothermal mechanochemical reaction

Stellite 25

4000

800

14. Hydrothermal reaction under microwave conditions

Rene 41

10000

740

15. Hydrothermal reaction under ultrasonic conditions

TZM

3000

1100

16. RESA; spark discharge in solution

(After Laudise and Nielsen [37a] and Somiya [44b])

TABLE 4 Characteristics of Hydrothermal Processing Based on Characteristics of Powders [5,19,20]

TABLE 6 Sealing Methods [44b]

1. Powders formed directly from solution 2. Crystalline and amorphous powders formed depending on hydrothermal temperature

Sealing methods Flat plate closure (Morey type) [25,36,37a] Cold-cone seat closure (Tuttle type) [46,37a]

3. Particle size controlled by hydrothermal temperature

Delta and “D” ring closure [37a]

4. Particle shape controlled by starting materials

Modified Bridgeman closure [37a]

5. Easy to control chemical composition

Gray Loc closure [49]

6. In many cases, powders formed without water (i.e. processes to remove water are not necessary) 7. In many cases, powders formed without agglomeration because these powders are fine and anhydrous (i.e. the milling of powders is not necessary) 8. Highly reactive powders formed (i.e. calcination or preheating is not necessary before sintering)

The Morey type autoclave is quite easily assembled, and the gasket plate requires minimum mechanical precision. The first version of the Morey bomb was sealed off with a fixed volume of water, and hence, the pressure was fixed at a specific temperature. Therefore, the most important modification was made to introduce a drilled

plunger in order to pressure the vessel outside it. Moreover, there were two additional problems inherent in design [51]. First, the thrust of the closure nut (“plunger” in Figure 1) on the soft metal sealing gasket produced leaks at a high pressure or with rapid temperature change. In addition, the first problem made the operation of the vessel time consuming because of not only leaks but also difficulties in removing the closure at the end of a run. The second problem was pressure measurement because the entire Morey bomb and closure lay within the furnace. At that time, the pressure was calculated from the pressuree volumeetemperature (PeVeT) relationship assuming that

874

Handbook of Advanced Ceramics

FIGURE 3 PeT limits for stellite vessel. (Laudise and Nielsen (1961). Solid State Physics Vol. 12, p. 165. [37a]).

nothing was lost during an experimental run after sealing a known volume of water in the vessel at the beginning of the run. In other words, the pressure was calculated based on a known ratio of the water to the vessel. Later, these problems were solved in a modified design, and the above types of bombs have been used up to 3 kb and up to about 600  C. Vessels made out of Inconel may be used under a pressure as high as 4 kb and at 600  C. Morey vessels for academic experiments are used in order to obtain a lot of samples because of a large volume is a requirement for the vessel. Later, a vessel different from the Morey bomb was designed by Tuttle, Roy, and other researchers. This type is called a Tuttle bomb or a TuttleeRoy bomb. The Tuttle bomb is also called a cold seal pressure vessel. This type of vessel has the basic advantages of having no threaded parts that are heated and having sealing made by pressure on the cone-in-cone seal. The principal advantages of this type of apparatus are its simple constitution and a small size of the pressure vessel, which permits more rapid heating and quenching than externally heating vessels do (see Heating Methods in Section 4.3). A Morey vessel and the closure are within the furnace, while the closure of a Tuttle bomb vessel or a cold seal pressure vessel is placed outside the furnace. This type of vessel is simply a metal cylinder with an axial hole drilled from one end to within about 100 of the other end. The closed end is placed upward in the furnace with the pressure connection below and outside the furnace. Roy et al. modified the Tuttle vessel to suspend a vessel in the furnace. They then named it a “test tube” pressure vessel for this unit. The first type required that a capsule with the starting powders be supported by a rod within the vessel, but an improved type does not require this capsule.

The TuttleeRoy bomb may be made cheaply out of ordinary stainless steel and bombs made out of Stellite 25 are suitable for experiments at 1 kb and 950  C. When the closed end of a TuttleeRoy bomb is lowered, quenching becomes very efficient when one immerses the closed end in a bath of water. The use of a close-fitting rod in order to decrease the volume of the vessel enhances quick safety and greatly diminishes the effect of convection. These types of vessels have been used at a pressure as high as 6 kb to 7 kb and at 700  C. At present, one is able to get many kinds of autoclaves to cover different pressureetemperature ranges and volumes [44a]. In the USA, there are three companies: Tem-Pres, a division of LECO Corp. [47], Autoclave Engineers, a division of Snap-tite Inc. [52], and Parr Instrument Co. [53]. Each autoclave has limitations of pressure and temperature. These limitations depend on autoclave material, sealing methods, operating temperature, and time duration as shown in Figure 3 [37a] and Figure 4 [49], and Tables 5 and 6. In Japan, there are regulations for high pressure gas operations. In addition, safety is one of the important points for the operation of autoclaves.

4.2. A Batch System and a Continuous System [41] Processes entail two techniques: One is a batch system, and the other is a continuous system. In the ordinary case, batch systems are common in experimental rooms. Some researchers, however, prefer continuous systems. Sakai Chemical Industry Co. Ltd. has produced barium titanate (BaTiO3) using a hydrothermal continuous system [54].

Chapter | 11.1.1

Ceramic Powders for Advanced Ceramics

875

FIGURE 4 Comparison for self- and pressure-energized sealing systems applied to large autoclaves. (Modified Bridgeman seal and Gray-Loc seal; after Asahara and Nagai, and Harada (1983). Proceedings of the First International Symposium on Hydrothermal Reactions, p. 439. [49]).

4.3. Heating Methods There are two kinds of heating methods, the internal and the external heating method. The internal heating system contains heaters in a reactor, while the external heating system has an external furnace outside a reactor. The external heating system is simpler than the internal heating system, and hence, the external heating system is very commonly used around the world. Heating methods are selected depending on operating pressure and temperature, the length of operating time, and the size of the reactor. The internal heating system is preferably used under conditions such as high pressure and temperature, long operating time, and large-volume autoclaves. A requirement of reactors for industrial products is large volume because of large-sized products, many products, or both. Thus the internal heating system is mainly used. Concerning academic research, the internal heating method may be used in order to use autoclaves under conditions up to 10 kb and 1500  C.

4.4. Starting Materials [44c] For hydrothermal experiments, the starting materials are very important. Requirements for the starting materials are as follows: 1. The use of high-purity starting materials. 2. The use of fine starting materials, namely, fine powders. 3. Accurate ratio of the starting materials for the chemical composition of a product. 4. Homogenous mixing of the starting materials. There are six types of starting materials that are commonly used for research and industrial production: (a) glasses; (b) gels; (c) dry mixtures of oxides; (d) chemical salts, such as carbonates, hydroxides, nitrates, and sulfates; (e) natural minerals and rocks; and (f) metal (with regard to metal,

there are cases in which small pieces and wires may be used instead of powders). Some starting materials, such as carbonates, easily absorb water, and hence, it would be better to use starting materials without water in order to correctly weigh the required substances. For example, one may heat starting materials in order to remove water before weighing them.

4.5. Hydrothermal Reactions Table 3 lists the various hydrothermal reactions and some example reactions are introduced here. Among them, Reactions 1, 4, 6, 7 as listed in Table 3 are used to produce nonpowders (i.e. bulks) as final products. We also mention these reactions because some are industrially and scientifically important, and it would be valuable to introduce them.. Important reactions under hydrothermal conditions are Reactions 1, 7, 10. From the industrial viewpoint, the first two reactions are important. An example of Reaction 1 is the crystal growth of artificial quartz (SiO2) [49,55]. Hydrothermal crystal growth is not a process for preparing powders. Reaction 2 is used to remove the impurities from objective products under hydrothermal conditions. An example is the removal of glassy materials from fluorphlogopite rock [55]. An example of Reaction 3 is the exchange reaction in rock, fluorine ion 4 hydroxide ion (F 4 OH ). Kaolinite is produced from feldspar [55]. An example for Reaction 4 is to prepare potassium hexatitanate (K2Ti6O13; fiber) from hydrous titanium dioxide (TiO2$nH2O), potassium hydroxide (KOH), and water (H2O) along with either magnesium (Mg metal) or zinc (Zn metal) as a dehydration reagent [56]. Example products were not powders but fibers as described above. Examples of Reaction 5 are to obtain aluminum oxide or alumina (Al2O3) from bauxite (ore containing Al2O3) in

876

Handbook of Advanced Ceramics

alkali solution [57], and magnetite (Fe3-xO4) from ilmenite (FeTiO3) in KOH solution [58,59]. Reaction 6 [60] may be used to remove water from hydrated materials during hot pressing by using hydrothermal apparatus and retaining the water in the apparatus. In other words, hydrothermal sintering may be used for hot pressing of hydrates in order to obtain anhydrous sintered products. Examples of such products are porous silicon dioxide or silica (SiO2) and dental materials, such as quartz (SiO2). Hydrothermal sintering is not a process for preparing powders. Final products of Reaction 7 are not powders, either. Reaction 7 is used to change metal powders or thin pieces of metal into oxide powders and sinter the oxide powders [61]. In other words, powders are intermediate products and sintered products are not powders. This sintering under certain conditions allows one to produce smaller grain sizes of sintered bodies than that of the starting metal [61]. Grain size increases as the sintering temperature increases. From the scientific viewpoint, this is one of the most important hydrothermal reactions. R. L. Coble who was a professor of the MIT, an authority on the sintering of ceramics, and an inventor of transparent alumina “Lucalox,” had mentioned that hydrothermal reaction sintering became a scientific research issue because particle size reduction from starting powders (chromium (Cr metal)) to intermediate powders (chromium oxide (Cr2O3)) was observed [62]. Examples of Reaction 8 are as follows [63]: SiC þ 2H2O / SiO2 (amorphous) þ CH4 Si3N4 þ 6H2O / 3SiO2 (amorphous) þ 4NH3 In the case of Reaction 9, the starting powders are oxidized and oxidized powders are produced. This means that these powders are not necessarily dense because they are not sintered. Examples of “hydrothermal oxidation” are shown below: 2Cr þ 3H2O / Cr2O3 þ 3H2 [61] 2Al þ 6H2O / 2Al(OH)3 þ 3H2; 2Al(OH)3 / Al2O3 þ 3H2O [64] In addition, zirconium (Zr metal) is oxidized under water to become zirconium dioxide or zirconia (ZrO2) via the metastable material zirconium hydride (ZrHx) [65,66]. In other words, zirconium reacts with water to form zirconium hydride and then zirconium dioxide. Reaction 10 may be hydrothermal hydrolysis [25]. An example is a reaction to produce yttrium-stabilized zirconium dioxide (Y-ZrO2) from zirconium oxychloride (ZrOCl2) and yttrium chloridehexahydrate (YCl3$6H2O) [67].

Examples for Reaction 11 are shown below: 3FeTiO3 þ H2O / Fe3 xO4 þ 3TiO2 þ H2 [59] ZrSiO4 þ 2NaOH / Na2ZrSiO5 þ H2O [68] (Sodium hydroxide (NaOH) is used as a catalyst in order to decompose zirconium silicate (zircon); (ZrSiO4)) ZrSiO4 þ Ca(OH)2 / ZrO2 þ CaO$SiO2 þ H2O [68] An example of Reaction 12 is to produce BaTiO3 by reaction between Ti (metal) and barium nitrate (Ba(NO3)2) [69]. An example of Reaction 13 is to produce barium hexaferrite (BaO$6Fe2O3) using chemicals, such as barium hydroxide (Ba(OH)2) and iron (III) chloride (ferric chloride; FeCl3) [70]. Special characteristics of Reaction 14 [71] are to save time, energy, and cost because of (a) rapid heating to reaction temperature, (b) low reaction temperature and pressure compared with conventional hydrothermal synthesis conditions, (c) the increase of reaction speed by one or two orders of magnitude compared with conventional hydrothermal synthesis, namely, short duration time, (d) formation of a novel phase of a material, and eliminating the metastable phase processes, and (e) environmentally safe and benign processes. Three examples are explained below [71]. Products under microwave hydrothermal conditions with titanium tetrachloride (TiCl4) in hydrochloric acid (HCl) at 164  C for both 0.5 and 2 h were rutile (TiO2). Meanwhile, products under conventional hydrothermal conditions at 164  C for both 2 and 24 h were rutile and a small amount of anatase (TiO2). Monoclinic ZrO2 was produced under microwave hydrothermal conditions with ZrOCl2 solution at 194  C for 2 h, but ZrO2 was not produced under conventional hydrothermal conditions at 195  C for 2 h. Titanium oxychloride (TiOCl2) reacts with Ba(OH)2 to produce BaTiO3 and a small amount of impurity under microwave hydrothermal conditions at 194  C for 2 h, but BaTiO3 is not produced under conventional hydrothermal conditions for 24 h even though the same sources with Ti and Ba were used. Examples for Reaction 15 are shown below [72]: Al2O3 (manufactured by Rhone-Poulene) þH2O /Al(OH)3 (bayerite) In addition, Al2O3 reacts with phosphoric acid (H3PO4) in water to produce hydrated aluminum phosphate (AlPO4$xH2O (e.g. x ¼ 1.67)). Example products of Reaction 16 [28] are oxides, carbides, and nitrides. For example, part of a Zr metal rod

Chapter | 11.1.1

Ceramic Powders for Advanced Ceramics

in water becomes fine ZrO2 powder in water by the RESA method. Reaction 16 (the IwatanieIshibashi method) is similar to the RESA method as described in Section 3.2. A difference between the two methods is the type of solution used and material produced [29,30]. An example product of Reaction 16 [29,30] is aluminum hydroxide (Al(OH)3) produced from aluminum (Al metal) and H2O. Al(OH)3 may be calcined to produce Al2O3. When Al which does not contain sodium (Na) is used as a starting metal to produce Al(OH)3, Na-free Al(OH)3 and then Na-free Al2O3 are produced.

4.6. Zirconia and Yttrium-Doped Zirconia Powders Manufactured by Hydrothermal Homogenous Precipitation We explain zirconia (ZrO2) and yttrium-doped zirconia (YZrO2) as sample commercial powders manufactured by one of the hydrothermal precipitation methods, namely, homogenous precipitation. There are many articles related to hydrothermal zirconia and zirconia-related materials [73e78]. Chichibu Cement Co. Ltd. (presently Taiheiyo Cement Corp. through Chichibu-Onoda Corp.) and one of the authors (S. S.) produced zirconia (ZP 20 as in Table 7 and Figure 6), and partially stabilized zirconia (ZY 30) and stabilized zirconia (ZY 80) using yttrium by adopting the hydrothermal homogenous precipitation method with financial support from the Research Development Corp. of Japan (presently Japan Science and Technology Agencythrough the Japan Science and Technology Corp.) [79,80]. Figure 5 [79,80] shows the processes for preparing hydrothermal homogenous precipitation powders of ZrO2 and Y-ZrO2. Properties of ZrO2 and Y-ZrO2 prepared by using this method are shown in Table 7 [79,80]. Transmission electron microsopy (TEM) pictures of ZrO2(ZP 20) and YZrO2(ZY 30, ZY 80) are shown in Figure 6 [79,80]. A picture of a sintered body of Y-ZrO2 (ZY 30) at 1400  C for 2 h is shown in Figure 7 [79,80]. These have very fine powders and grains.

4.7. Characteristics of Hydrothermal Powders [19,20] Hydrothermal powders are directly formed from solution. These powders may be single crystals, polycrystals, amorphous forms, and metastable crystals depending on hydrothermal preparation temperature as the main condition, and on hydrothermal preparation time. In addition, anhydrous powders may be prepared. Particle size and shape are controlled by hydrothermal temperature and starting materials, respectively. Hydrothermal powders generally do not

877

TABLE 7 Typical Characteristics of Zirconia (ZP 20) and Yttrium-doped Zirconia (ZY 30 & ZY 80) Powders Prepared by Hydrothermal Homogenous Precipitation and their Sintered Bodies Manufactured by Chichibu Cement Co. Ltd. [79,80] ZY 30*

ZY 80*

ZP 20*

ZrO2

94.7

86.0

>99.9

Y2O3

5.2

13.9

e

Al2O3

0.01

0.01

0.005

SiO2

0.01

0.01

0.005

Fe2O3

0.005

0.005

0.005

Na2O

0.001

0.001

0.001

Cl

OC2 H5 > OC3 H7 > OC4 H9 > OC6 H13

(12)

The rate is smaller for branched alkyl groups than for nonbranching ones: n

C4 H9 > sec

C4 H9

(13)

This can be explained based on the steric hindrance effect.

2.1.4. Effect of the Type of Alkylalkoxysilane For alkylalkoxysilanes (CH3)xSi(OC2H5)4ex, the following results are obtained [7]. In HCl-catalyzed solution, the hydrolysis rate increases with increasing number of CH3 groups x bonded to silicon directly. SiðOC2 H5 Þ4 < CH3 SiðOC2 H5 Þ3 < ðCH3 Þ2 SiðOC2 H5 Þ2 < ðCH3 Þ3 SiðOC2 H5 Þ

(14)

On the other hand, in NH3-catalyzed solution, the rate of hydrolysis decreases as the number of CH3 bonded to Si directly increases. These effects can be explained by the inductive effect.

2.2. Gelation in the Phase-Separable Solution It was found that macroporous silica containing macropores with controlled size and size distribution are prepared through the solegel reaction accompanied by phase separation of the starting solution [23e25]. For this purpose, starting solutions containing water-soluble polymers or organic substances, such as polyacrylic acid, together with alkoxysilane as silica source are used. When polycondensation of silica components progresses in these solutions, phase separation due to spinodal decomposition into silica-rich phase and organic molecule-rich phase occurs. Each phase shows connected irregular shape. When the organic molecule-containing phase is dissolved away, silica-gel body with micron-size continuous pores (0.5e5 mm) results. Silica frameworks contain nanometer-size (5e50 nm) open pores. Accordingly, the silica bodies thus prepared are so called a double pore system, that is, bigger pores are used for transport of liquid substrate and fine pores are used for absorptionedesorption of molecules, for instance, when used as column for chromatography. It is suggested that the volume and size of both macropores and micropores can be adjusted by controlling gelation and growth of phase separation.

Chapter | 11.1.2

SoleGel Process and Applications

2.3. Reaction for Nonsilica Oxides Solegel method is useful for fabrication of functional materials, such as photocatalyst, nonlinear optical materials, ferroelectrics, and superconductors. For this purpose, solegel preparation of simple and complex nonsilica oxides, including TiO2, ZrO2, Al2O3, ZnO, WO3, Nb2O5, rare earth oxides, and so on, have to be studied. Metal alkoxides are often employed for those oxides. Most of them, however, are unstable, that is, they are rapidly hydrolyzed and are easily precipitated, which makes it difficult to form homogeneous multicomponent oxide products. Methods of controlling reactivity of the transition metal alkoxides are introduced below.

2.4. Hydrolysis and Condensation of Transition Metal Alkoxides According to Livage [26] and Sanchez [27], the rate of both hydrolysis and condensation of transition metal alkoxides is very large compared to that of silicon alkoxides. As an example, the hydrolysis rate of Ti(OEt)4 is about five orders of magnitude greater than that of Si(OEt)4. The high reactivity of transition metal alkoxide may lead to the oligomer formation such as formation of [Ti(OEt)4]n (n ¼ 2 or 3) and [Ti(OMe)4]n. In order to avoid the precipitation by suppressing the rate of hydrolysis and polycondensation, chemical modification is applied to the transition metal alkoxides [28,29]. An example of the chemical modification is that 2-methoxyethanol HOC2H4OCH3, diethanolamine (HOCH2CH2)2NH, or triethanolamine is used as solvent in place of simple alcohols. For example, zirconium isopropoxide [Zr(OiC3H7)4(iC3H7OH)]2 reacts with 2-methoxyethanol solvent and forms Zr(OC2H4OCH3)3(OiC3H7), which suppresses the hydrolysis. Another example is the modification of the coordination state of the metal [26]. It is known that carboxylic acid reacts with the titanium isopropoxide, modifying Ti(O-iC3H7)4 to Ti(O-iC3H7)3(OCOCH3). Another method is the use of b-diketones, such as acetylacetone and ethylacetate [30] and b-diols [31] as a part of the solvent. These solvents are often used to prevent the precipitation by complexing the ligand of metal alkoxides with high reactivity. Besides control of reactivity, the modification makes the otherwise insoluble metal alkoxide soluble in alcohol. Copper alkoxides such as copper methoxide and propoxide are solids which are insoluble in alcohol. Modification of these alkoxides by b-diketone makes them soluble in alcohol, which makes possible the use of copper alkoxides as starting material for solegel processing. Kakihana [31] introduce a great number of water-soluble titanium complexes which can be used as precursors for synthesizing titania and titanium-containing oxides through solegel method. Synthesis, coordination structure,

887

TABLE 3 Heterometallic Alkoxides for the Synthesis of Complex Oxides [Taken from Kessler’s Article [29]] Complex oxides to be synthesized

Heterometallic alkoxides as starting material

BaTiO3

BaTiO(OiPr)4(ROH)x

BaZrO3

BaZr(OH)(OiPr)5$3ROH

MgAl2O4

MgAl(acac)(OiPr)2

NaxWO3

NaW(OR)6 R ¼ Me, Et, iPr, iBu

PbMg1/3Nb2/3O3

MgNb2(OEt)12$2EtOH PbNb4O4(OEt)24

PbMg1/3Ta2/3O3

MgTa2(OEt)12$2EtOH

(Pb, La)(Ti, Zr)O3 (PLZT)

Pb2Ti2O(OR)8(OAc)2

SrBi2Nb2O6

SrNb2(OiPr)12$2ROH

SrBi2Ta2O6

SrTa2(OiPr)12$2ROH

solubility, stability, and application of the soluble titanium complexes are also discussed. Many of the functional oxides such as ferroelectric materials contain two or more kinds of metals as ion. Alkoxide of each metal may not be so stable, undergoing a rapid hydrolysis and polycondensation. Therefore, it is not easy to obtain homogeneous products from the mixture of different metal alkoxides. It is considered that the use of heterometallic alkoxides might dissolve this problem. Heterometallic alkoxide molecule holds two or more different metal atoms in it and is quite useful for preparation of homogeneous gel which contains different metallic species. Then many heterometallic alkoxides have been synthesized. Kessler [29] lists up a large number of heterometallic alkoxides, a part of which is shown in Table 3.

2.5. Formation of OrganiceInorganic Hybrid Organiceinorganic hybrids have microstructures in which inorganic species like eSieOeTieOe are chemically connected with radicals like CH3 and C6H5 or organic polymers like poly(methylmethacrylate). Philipp [33] prepared a transparent, homogeneous organiceinorganic hybrid material, which may be used as hard contact lens, by hydrolysis and condensation of a mixture of epoxysilane, methacryloxysilane, silicon alkoxide, titanium alkoxide, and monomeric methacrylate with a small amount of peroxide polymerization catalyst. Since then, the organice inorganic hybrids have been attracting much attention, becoming a very important area of solegel technology.

888

Handbook of Advanced Ceramics

Solegel processing of organiceinorganic hybrids can be carried out by the following ways.

possible to prepare silica hybrids with a large amount of hydrophobic organic molecules.

2.5.1. Use of Alkylalkoxysilanes Such as CH3Si(OC2H5)3

2.5.5. Synthesis of Hybrids Containing Characteristic Polymers

This is the simplest method of preparing hybrids [34]. An alkyl group or groups bonding directly to a silicon atom of the silicon alkoxide molecule remain bonded in the process of hydrolysis and polycondensation of alkoxy groups in the alkoxide-alcohol-water-catalyst solution, until they are decomposed around 200  C or higher temperatures. When dimethyldiethoxysilane [34] with two methyl groups per Si atom is employed for solegel synthesis, tetraethoxysilane has to be introduced into the starting solution to form threedimensional network structure.

Organiceinorganic hybrids with interesting properties are synthesized by solegel processing. For example, polyvinylpyrrolidone (PVP)eSiO2 hybrid [38], polyoxazoline (POZO)eSiO2 hybrid, and diphenyldimethoxysilaneeSiO2 hybrid [39] are synthesized.

2.5.2. Use of Polydimethylsilane Polydimethylsilaneetetraethoxysilane solutions subjected to hydrolysisecondensation reaction give rise to soft and hard organiceinorganic solid materials [35], depending on the content of polydimethylsilane.

2.5.3. Radical Polymerization of Alkyltrimethoxysilane Followed by Hydrolysis It is shown [36] that transparent flexible or hard organice inorganic hybrid gels are prepared by radical polymerization of vinyltrimethoxysilane followed by hydrolytic polycondensation of polyvinyltrimethoxysilane. The gels contain both siloxane and carbonecarbon bonds and show mechanical properties depending on the degree of condensation of the siloxane network.

2.5.4. Use of Polysilazane Perhydropolysilazane (PHPS) composed of SieN bonds with SieH terminal groups has been employed as source material for SiO2 since coating film on silicon wafers made from PHPS solution was found to show density and refractive index close to those of silica glass. According to Kozuka [37], coating films prepared from xylene solution of PHPS are converted to dense, hard silica films at room temperature by exposing to the vapor from aqueous ammonia. It is demonstrated [37] that crack-free, transparent PMMA (polymethylmethacrylate)esilica hybrid coatings are prepared from the PHPSePMMAexylene solution at room temperature through the similar procedure. These hybrid coatings exhibit higher hardness and chemical durability than alkoxide-derived ones, which indicates the great advantage of PHPS over alkoxide. It is also suggested that owing to the hydrophobic nature, PHPS makes it

2.5.6. Synthesis of Hybrids Containing Characteristic Silicate Anion This is a method of preparing organiceinorganic hybrids by bonding characteristic silicate anions, such as cage-like structured cubic octamer Si8 O20 8 , with functional organic molecules [40]. This procedure may be called a buildingblock approach. An example is shown [41]. A methanolic solution, in which cubic octamers are present as tetramethylammonium silicate, is added drop by drop to a 2,2-dimethoxypropane [(CH3O)2C(CH3)2] solution of dimethyldichlorosilane [(CH3)2SiCl2], being kept under stirring at room temperature. Then, organiceinorganic hybrid, in which octamers are bonded with Si(CH3)3(OSi) units, results.

2.5.7. OrderedeMesostructured OrganiceInorganic Hybrid In the 1990s, ordered mesoporous thin silica films were prepared by calcining orderedemesostructured thin organicesilica hybrid fabricated from the silicate solution containing surfactants or liquid crystals as template [42]. Recently, orderedemesostructured organiceinorganic hybrid materials themselves also attract a considerable attention. Many works on the formation of organicesilica hybrids with ordered mesostructure, using surfactants as template, are reported. Shimojima [43,44] prepared ordered, nanostructured silica-based hybrid materials by solegel synthesis from designed organosiloxane oligomer having a terminal phenyl group without the use of any special structureinducing agent. The formation of ordered mesostructure was attributed to the formation of amphiphilic species during hydrolysis of the alkoxide. With similar idea, Shimojima [45] synthesized siloxaneeorganic hybrids with well-ordered mesostructures through the self-assembly of amphiphilic molecules that consist of cubic siloxane heads and hydrophobic alkyl tails. Besides silica-based hybrids, for instance, ZnO-based hybrids with aligned ZnO nanocrystals are formed with the assistance of a surfactant.

Chapter | 11.1.2

SoleGel Process and Applications

889

2.6. Nonhydrolytic SoleGel Reaction

2.7. Formation of Nanopores in the Gel

Most of the solegel preparation of metal oxides is based on the hydrolysis and subsequent polycondensation of metal alkoxides. In the early 1990s, the group of Vioux [46,47] developed the nonhydrolytic solegel process, in which MeOeM bridges are formed as a result of the condensation between metal halide, usually metal chloride, and metal alkoxide with elimination of alkyl halide:

2.7.1. Noncontrolled Pore Formation

MX þ MOR / MeOeM þ RX

(15)

Here M is a metal, X is a halogen, and R is an alkyl group. As an example, monolithic gel is produced through nonhydrolytic reaction between AlCl3 and Al(OiPr)3 in CCl4e(C2H5)2O solvent by heating at 110  C for 24 h. The resultant gel remains amorphous after calcination at 750  C for 5 h. Stability of the gel and the presence of pentacoordinated Al ions in resulting monolithic gel show the difference between hydrolysisecondensation process and nonhydrolytic process. In his article, Vioux [47] indicate that a great majority of nonhydrolytic reactions for oxide formation so far studied are based on the reaction between chloride precursors and alkoxide, ether or alcohol, as expressed by the following equations, respectively: MCln þ MðORÞn / 2MOn=2 þ nRCl

(16)

MCln þ ðn=2ÞROR / MOn=2 þ nRCl

(17)

MCln þ ðn=2ÞROH / MOn=2 þ nRCl þ nHCl

(18)

Many kinds of metal oxides, such as SiO2, Al2O3, TiO2, ZrO2, SnO2, and Nb2O5, and mixed oxides, such as SiO2eTiO2, SiO2eAl2O3, TiO2eAl2O3, TiO2eZrO2, and SnO2eAl2O3, have been prepared by the nonhydrolytic process. It is reported [48] that this method is suitable for the preparation of organiceinorganic hybrids. As an example, fabrication of low-melting hybrid glass will be cited. Niida [49] synthesized such glass by reacting the mixture of dimethylchlorosilane Me2SiCl and anhydrous orthophosphoric acid H3PO4 in flask at temperatures higher than 40  C under flow of nitrogen. The reaction follows Eqn (19): H3 PO4 þ R2 SiCl3

/ þ ðOHÞ2 OPeOeSiClR2 þ HCl[

Inorganic dried gel products are usually porous materials. Examples of the porosity are given in Table 4 for three kinds of silica gels, which have different starting compositions. It should be noted that all the three gels consist of granular particles and so, pores are connective. It is observed that the gel contains several tens volume% of porosity when silica gels are formed without any special procedure. Gel no.1 is transparent due to the small pores [50]. The gel doped with fluorescent and laser molecules is applied as photo-functional gel. Gel no.2 is translucent [51], due to larger pores of 16 nm and can be applied to the fabrication of transparent silica bodies. The particles of gel no.3 are very big with 5 mm diameter [52]. It is observed that these big particles are secondary particles consisting of very small primary particles.

2.7.2. Controlled Formation of Porous Materials with Ordered Mesopores Porous materials with controlled mesopores have been attracting much attention. In these materials, uniform mesopores are orderly arranged like an array of atoms or ions in crystals [53]. The outline of this technology is introduced below, according to the early work on this matter [42]. The amphiphilic surfactant, such as C16H33(CH3)3NOH/Cl, is employed to make ordered structure. The process starts with tetramethylammonium

TABLE 4 Three Kinds of Silica Gels with Different Pore Diameters and Porosities No. 1

R ¼ CH3 Network structure consisting of PeOeSi bondings is developed as the reaction proceeds with escape of HCl, resulting in transparent solid glassy materials with a low glass transition temperature.

No. 3

Composition of starting solution (molar ratio) Si(OCH3)4

1

1

1

H2O

4

12

1.53

CH3OH

2

2

2

(CH3)3NCHO

e

1

e

HCl

0.4

e

NH4OH

(19)

No. 2

0.4 4

e

5  10

Transparent

Translucent

Opaque

2.5 nm

16 nm

5e10 mm

46

73

60

e

Property Appearance Average pore diameter Porosity (Vol.%)

890

Handbook of Advanced Ceramics

silicate solution containing surfactants. It is assumed that circular micelles with the center part consist of hydrophobic ends of surfactants and hydrophilic silicon containing periphery. These circles are assembled face to face, producing micellar rods or cylinders. Since the surface of the cylinder attracts tetramethylammonium ions and becomes rich in silica, the rods gather together, forming hexagonal arrays. When these hexagonal arrays are calcined to decompose internal organic parts, 2-dimensional hexagonal arrays of mesopores arranged in silica networks are obtained in coating film. After this work an extensive research has been directed to fabrication of ordered mesoporous materials. At present, this process is classified into two types [44]. One is the use of surfactants as structure creating agents that enables the formation of lamellar, hexagonal, or cubic mesostructures. Another type is the formation of self-directed assembly of certain organoalkoxysilanes during hydrolysis and polycondensation, providing a one-step surfactant-free route to mesostructured hybrid. In this case, starting materials have to be selected so that amphiphilic species might be formed during the reactions [43].

3. FORMATION OF SHAPES AND MICROSTRUCTURES In the solegel technology, the sol is formed into bulk, fiber, or coating film in the course of solegel reaction. Possibility of forming specific-shaped gels largely depends on the composition of the starting solution. This section describes design of the starting solution for forming fibers and bulk gels [54].

3.1. Formation of Bulk Gel and Glass The bulk gel is formed by molding the sol and completing solegel reaction. The resulting wet gel is dried by heating at temperatures lower than or around 100  C. The most serious problem is possible occurrence of cracks or fracture on drying wet gel. For inorganic oxides, preparation of bulk bodies such as rods larger than 10 mm  10 cm in area and thicker than 5e20 mm is not easy. It was suggested by Zarzycki [55] that fracture in the course of drying, if any, is caused by the occurrence of the tensile stress in the surface layer of the gel due to the capillary force DP: DP ¼ 4 g cos q=D

(20)

where g is the surface tension of the liquid filling the micropores, q is the contact angle, and D is the pore diameter. Equation (20) indicates that cracking may be reduced by slower drying, larger pores, and smaller surface tension of the liquid in gel pores. It is easily understood that supercritical drying is very effective in obtaining bulk gel without cracking. Bulk silica gel and glass were obtained

by enlarging pores [25,56,57] and reducing surface tension of the liquid. It should be mentioned that large silica preforms of 25 kg for the clad of optical fiber for optical telecommunication were prepared [58].

3.2. Fiber Drawing Solegel method facilitates fabrication of glass and ceramic fibers. In making silica fibers by drawing from solegel solution, we start from a solution containing Si(OC2H5)4, water, ethanol, and hydrochloric acid. When the solution is kept standing under reflux at temperatures from room temperature to 80  C, the hydrolysis and polycondensation reactions proceed, leading to an increase in viscosity. When the composition of the starting Si(OC2H5)4eH2OeCH3OHeHCl solution is appropriate, fibers can be drawn from the viscous sol having a viscosity higher than 10 poises. Gel fibers drawn from the solution can be converted to silica glass fibers by heating at 800e850  C. It is noted that the maximum heating temperature throughout the processes, that is, 800e850  C, is much lower than the temperature at which silica fibers are drawn from molten quartz or silica glass rod. As to the spinnability of the solution, Sakka [59] found that the starting solution of low water content that is catalyzed with acid becomes spinnable. This condition corresponds to those for the starting solutions in which long-shaped linear particles are formed [60]. Figure 2 compares the rheological properties of spinnable and nonspinnable sols [61]. It is observed that the spinnable sol shows Newtonian behavior at high viscosities, while nonspinnable sols show structural viscosity or thixotropy characteristic of granular particles at such high viscosities where fibers can be drawn if the composition is appropriate. These conditions for the occurrence of spinnability and rheological behavior of spinnable sols can be applied to many sols of different compositions from silica, that is, alumina, titania, superconducting oxide sols. Fibers of various compositions were prepared by the solegel method [62].

3.3. SoleGel-Coating Films Coating of glass, ceramic, metal, and plastic substrates by the solegel method is very useful for modifying properties of substrates with a large or small surface area or providing substrates with new active properties, which are needed for developing optical, electronic, chemical, and biological devices. The solegel method allows application of the film to nonheat-resistant substrates.

3.3.1. Formation of Coating Films Dip coating [63,64], spin coating, laminar flow coating [65], printing, and spray coating are employed in solegel coating.

Chapter | 11.1.2

SoleGel Process and Applications

891

FIGURE 2 The viscosity shear rate relation at various reaction times for the spinnable solution reacted in an open beaker OP-1 (a) and in a closed beaker CL-1 (b). Composition: 1Si(OC2H5)4$2H2O$1C2H5OH$0.01 HCl is used.

Dip and spray coatings are two representative coating methods that are used in solegel coating. Here, discussion will be made mainly with dip coating; however, most of the discussion on the relation of thickness with crack formation will be valid also for other coating methods.

3.3.2. Adherence and Thickness of the Film In dip coating, a coating procedure consists of dipping of the substrate in the solution, drawing up of the substrate from the solution and heating at 200e500  C for the adherence of the film to the substrate. The adherence is attributed to the formation of bonding eMeOeM0 e (M, M0 are metal atoms in the film and substrate, respectively), when both the film and substrate are inorganic oxides. Roughly speaking, the thickness of the film is proportional to the amount of the solution dragged by the drawn-up substrate, which increases with increasing solution viscosity and substrate drawing speed. Based on this idea and detailed discussion [63,64], the thickness is expressed by the following simplified form [50]: tfhm vn

(21)

where h is the viscosity of the sol and v is the substrate drawing speed. It is shown that m has a value between 0.5 and 0.6 [50,64], and n is close to 0.5 in most cases [66].

3.3.3. Cracking of Film and Critical Thickness It is known [67] that for one coating run, the maximum thickness of the coating films of SiO2 and other oxides achieved without peeling off and crack formation, that is, the critical thickness is 0.2e0.5 mm. Occurrence of cracks is assumed to be caused by the stress formation upon heating of the film. According to Kozuka [68], in situ observation of gel films under heating indicates that macroscopic cracking occurs in the heating-up stage and

that the cracking onset temperature is affected by heating rate, film thickness, water-to-alkoxide ratio, and humidity. There are two methods for increasing film thickness without cracking. One is the repetition of the coating procedure. Examples can be found for BaTiO3 [50] and TiO2eGeO2 films [69]. It is assumed that for every subsequent coating procedure, one uses a new substrate consisting of combination of the original substrate and previously applied films. Another method of avoiding crack formation is to add organic polymers to starting components. The critical thickness is larger than 0.7 mm when silica films are prepared from a solution containing CH3Si(OC2H5)3 [70]. BaTiO3 of 2.1 mm thickness is obtained without cracking by using the starting solution containing poly-pyrrolidone by the amount corresponding to PVP=TiðOC3 H7 i Þ4 ratio ¼ 0.5 [68,71].

3.3.4. Striations in the Coating Films Radiative striations, which consist of radially extending neighbored ridges and valleys, are often observed in solegel spin-coating films. In general, striations are regarded as defects causing light scattering. Kozuka [68] examined the effect of coating condition on the formation and properties of striations, using the sol prepared from a solution of molar composition, Si(OC2H5)4:H2O:HNO3: C2H5OH ¼ 1:4:0.01:2. It was shown that striations are formed by local Marangoni convection, irrespective of the presence or absence of rotation of substrate, and that use of less volatile alcohols is effective in avoiding striation, because striations once formed diminish in the course of gelation. Uchiyama [72] demonstrated that striations are formed also in dip-coating films when starting solutions of molar compositions, Ti(OC3Hi7)4:H2O:HNO3:i-C3H7OH: PVP ¼ 1:2:0.2:60:0.10.7, are used. The addition of PVP

892

Handbook of Advanced Ceramics

(polyvinylpyrrolidone) is necessary to cause striations. In the case of dip coating, linear arrangements of striations are realized, because the sol on the glass substrate unidirectionally flows downward. The formation of the arrangement of linear striations on titania films might be applied as a method of making surface patternings.

3.3.5. Microstructure Formation of Coating Films Various microstructures as shown in Figure 1 are realized in solegel-coating films. For inorganic oxide-coating films, the presence or absence of crystallization and the type of crystallization are important microstructure characterizing coating films. It was found with Li2B4O7-coating film on Si wafer prepared from metal alkoxides LiOCH3 and B(OC4H9)3 that three kinds of microstructures, that is, amorphous, polycrystalline, and crystal-oriented microstructures, are made by changing the acid content of the starting solution [73]. Electric properties of ferroelectric thin films largely depend on the microstructure, especially orientation of crystals, even if the same crystalline species are precipitated. In solegel fabrication of ferroelectric coating films, many factors, such as kind of source materials, composition of starting solution, chemical reaction in the solution, kind of substrate, temperature and time of heat treatment, atmosphere in heating, and so on, have to be considered. As an example, Schneller [74] demonstrated that crystal orientation of Pb(Zr,Ti)O3 films on platinum depends upon the atmosphere at heat treatment. For instance, higher oxygen pressure hinders formation of PtxPb phase at Pt-film boundary, leading to decrease in (111) orientation and, consequently, polarization of the film is very much reduced.

4. APPLICATIONS Solegel method is characterized by the low-temperature processing, homogeneous products, fabrication of nanostructured materials, and organiceinorganic hybrid formation. Accordingly, this method is applied to fabrication of a great many kinds of functional materials [75,76]. Tables 5e7 show some of the important solegel-derived materials with photonic and electronic functions, thermal and mechanical functions, and chemical, biochemical, and biomedical functions, respectively. It is observed from those tables that various functional materials are fabricated by solegel method. In this section, description will be made on selected applications: phosphors, lasers, antireflecting coating films, coating films for automobile windows, photocatalyst, transparent conductors, ferroelectric materials, aerogels, and biochemical and biomedical functions.

4.1. Phosphors Intensive emission and stability of emission are two important luminescent properties for good phosphors. To achieve improvement, many kinds of phosphors have been prepared by solegel method. As example, white LED (light-emitting diode) CaAl2Si2O8: Eu2þ, Mn2þ phosphors [78] were prepared by solegel method. Stoichiometric amounts of Ca(NO3)2$4H2O, Al(NO3)3$9H2O, Eu(NO3)2$6H2O and Mn(NO3)2$xH2O were dissolved in ethanol and Si(OC2H5)4 was added. The resultant mixtures were stirred for 2 h and then heated at 70  C until homogeneous gels were formed. The gels crystallized at 1000  C showed emission bands excited by ultraviolet light. In a similar manner, a red-emitting long afterglow

TABLE 5 SoleGel-Derived Materials with Photonic and Electronic Functions Function

Specific function or microstructure

Examples

Photonic

Generation of light

Lasers [77]; phosphors [78].

Transmission of light

Optical fiber preform [58]; planar optical waveguide [79,80].

Optical absorption

Colored automobile window [81].

Electrochromism

WO3 Electrochromic film [82].

Sensing

H2S sensor [83].

Antireflection

Antireflective coating films [84].

Photovoltaic effect

Graetzel type solar cell [85].

Photocatalysis

Photocatalyst [86,87].

Transparent conductor

Transparent electrode of In2O3 [88,89].

Ferroelectric

Ferroelectric materials (Capacitors, non-volatile memories) [90,91].

Electronic

Chapter | 11.1.2

SoleGel Process and Applications

893

TABLE 6 SoleGel-Derived Thermal and Mechanical Materials Function

Specific function or microstructure

Examples

Thermal

Heat resistance

Silica-alumina-boria fibers [92].

Thermal insulation

Aerogel [1,93].

Mechanical protection

Hard coat on plastic lens [94].

Nanopatterning

Nanopatterned film [95].

Spacer

SiO2 spherical particles for LCD panel spacer [96].

Mechanical

TABLE 7 SoleGel-Derived Materials with Chemical, Biochemical, and Biomedical Functions Function

Specific function or microstructure

Examples

Chemical

Catalysts

CeO2eZrO2 catalyst for oxidation [97].

Membrane

Asymmetric membrane for filtering [98].

Macropores

Porous silica monolith for HPLC [99].

Water repellent

Fluoroalkylsilane hybrid coating for automobile window [100].

Passivation

Organiceinorganic hybrid coating [101].

Barrier coating

Organiceinorganic hybrid for gas barrier film [102,103].

Biochemical and biomedical

Glucose biosensor with glucose oxidase in hybrid gel [104].

Artificial tissue

Encapsulation of pancreatic cells [105]. Artificial bone [106].

Diagnosis

Antibodies in gel for immunoassay [107].

Lab-on-a-chip device

DNA arrays [108].

Sr3Al2O6: Eu2þ, Pr3þ phosphor with single phase was synthesized by solegel methods. In this case, heating at 1200  C for 2 h in the reducing atmosphere improved the light intensity and the light-lasting time (15 min) of these materials [109]. There are many kinds of organic fluorescent molecules. In order to improve the stability of such molecules, the use of inorganic materials such as silica as matrix was proposed. Reisfeld [110] prepared a sheet of silica doped with rhodamine 6G, applying to fluorescent solar collector. Afterwards, organiceinorganic hybrid materials are used as matrix in place of silica, since the organic molecules are well dispersed in the hybrids as seen in the following Section 4.2.

4.2. Lasers Nd glass laser is the first laser to which solegel processing was applied [111]. It is known that in preparing Nd-doped silica glass by sputtering or melting, homogeneous doping is difficult, because rare-earth elements tend to form clusters in

glass. This results in low threshold concentration of laser element for lasing. One may assume that the formation of the domains can be avoided by the use of solegel processing. However, this is not the case. Even with the solegel processing, addition of Al2O3 to glass is necessary to reduce clustering of Nd ions. The advantageous use of solegel method for rare earth doped silica glass is found in fabrication of planar waveguide for amplification [79] in WDM device for optical communication systems. When erbium-doped planar optical wave amplifier prepared by solegel method is used in place of the fiber amplifier, the amplification can be carried out within a smaller volume. Laser properties of planar waveguides are affected by the host material, and erbium-doped planar waveguides are fabricated.on the host of GeO2eSiO2 [112] and aluminosilicate [113]. It is well known that incorporation of functional organic molecules in silica matrix [8] opened new fields for solegel method around 1984. This idea was applied to the development of lasers containing organic laser dyes [77]. First, solid-state tunable lasers were prepared instead of liquid

894

Handbook of Advanced Ceramics

tunable lasers. The next development of dye lasers was to use organiceinorganic materials, especially organically modified silicates (ormosil). Those materials can dissolve higher amount of organic laser dyes. For this purpose, various organic laser dyes, such as perylimide, pyrromethene, rhodamine, and cyanine dyes, were synthesized and used for making red, blue, and UV lasers. A problem with dye lasers is their weak photostability. To improve the photostability of dye lasers, photostable matrix as well as photostable laser dyes have to be explored. Reisfeld [114] prepared photostable berylimide dyes. Yagi [39] indicated that incorporation of phenyl groups in organiceinorganic hybrid matrix improves photostability of dye lasers. Transparent spherical particles containing organic laser dyes become micro-lasers [115]. The diameter of the micro-sphere lasers is several to several tens of microns. The spheres are made from organiceinorganic hybrids. Introduced lights go around by reflecting at the sphere surface within the sphere crust, emitting laser light.

4.3. Antireflective Coating The reflection of light takes place at the surface of glass. The reflectivity is about 4% for one surface of soda-lime-silica glass having refractive index of about 1.52. In other words, about 4% of the incident light is lost at every surface of glass. The reflection of light causes decrease of measuring accuracy in optical systems, lowering of the oscillation threshold and increased damage of glass laser lenses in the laser fusion system, and difficulty in clear sight for glass-protected paintings in museum and showcases. When the refractive index nf and the thickness df of a film satisfy the following equations, antireflection is achieved for the light of wavelength l: nf ¼

pffiffiffiffi ng

nf $df ¼ l=4 þ m=2$l

(22)

Substrate=first intermediate index layer=second higher index layer=third lower index layer (25) So far, various antireflective films have been prepared by solegel method. Mukherjee [116] prepared single-layered antireflective film of silica-rich porous glass on glass lenses for laser fusion system by the use of phase separation of deposited borosilicate glass film. A small reflectivity of 0.15e0.7% at the wavelength of 1.06 mm and an increase in the damage threshold of the fusion laser from 4 J/cm2 to 22 J/cm2 were achieved. Yoldas [117] deposited a porous antireflective film of pure silica on fused silica surface by solegel reaction involving Si(OC2H5)4. The pore size and morphology were tailored by controlled heat treatment. Antireflection was achieved over the entire spectral range of silica with better than 99% transmission. Floch [118] developed a scratchresistant one-layer antireflective coating film with a thickness corresponding to m ¼ 0 in Eqn (22). The resulting coated glass can be applied to general uses, such as household articles, architectures and ophthalmic lenses, as well as fusion laser glasses. Yamazaki [84] developed single-layered antireflective coating film from the starting solution containing methyltriethoxysilane (MTES) as well as tetraethoxysilane (TEOS) as the source of silica. The low refractive index is achieved by heating at 500  C, which creates pores or voids as a result of decomposition of clusters of methyl groups remaining in the gel. This simple method can be applied to automobile windows. Three-layered antireflective coating films were developed in Schott of Germay [119]. The film has a constitution given by Formula (24), corresponding to glass substrate/ SiO2eTiO2 layer/TiO2 layer/SiO2 layer. Based on this constitution, Schott fabricated commercial products “Amiran” for show windows and “Conturan” for a filter to increase CRT contrast [120].

(23)

where ng is the refractive index of the glass substrate. Since the refractive index of conventional soda-lime glass is around 1.52, the refractive index nf desirable for the antireflective coating should be 1.23. It is difficult to achieve such a low value for the practically useful film, and so what to do is to make the refractive index of the coating film as low as possible. When a two-layered film is designed, the first higher refractive index layer is deposited on the glass, and on this layer the second lower refractive index layer is deposited to achieve antireflection, as shown below: Substrate=first higher index layer=second lower index layer

For a three-layered antireflective film, the film is constructed in the following way:

(24)

4.4. Coating Films for Automobile Windows Representative solegel-coating films which are applied to automotive windows are listed in Table 8

4.4.1. Reflective Film The reflective coating film as combiner of reflective screen in the head-up-display (HUD) system on the inner surface of windshield is used to display automobile information [121,122]. The coating was made by Central Glass Company in Japan and became a trigger for application of solegel coating for preparing various functional coatings on automobile windows. In the HUD system, the display light is reflected by the combiner film, reaching the

Chapter | 11.1.2

SoleGel Process and Applications

895

TABLE 8 Examples of SoleGel-Prepared Coatings for Automobile Windows Kind of film

Example of the composition

Application

Reflective film

SiO2eTiO2 film on glass

Combiner of the head-up-display system [121,122].

Antireflective film

(Glass)/SiO2eTiO2/TiO2/SiO2

Antiglare mirror [84].

Colored film

(Glass)/CueMneCoeCreO spinel/ CueMneCoeCreO containing SiO2

Gray-colored coating for passenger comfort [81,123].

UV shielding film

(Glass)/SiO2eTiO2/CeO2eTiO2

Protection of passengers and inner commodities from the effect of UV rays [124].

IR-shielding film

SiO2eITO

Suppression of temperature rise within automobile [125].

Water-repellent film

Fluoroalkylsilane

Assurance of sight [126,127].

Hydrophilic film

TiO2eSiO2

Assurance of clear sight for side mirror [128].

Antifogging film

Polyurethane

Antifogging for automobile windows [84].

driver’s eye. It is designed so that the displayed information, such as speed of the car as image, appears in front of the driver. Complying with regulations concerning automobiles, TiO2eSiO2-coating films with refractive index of 2.05, thickness of 220 nm, and size of 4.500  300 were prepared.

4.4.2. Antireflective Film For the antireflective film, refer to previous Section 4.3.

conventionally, UV absorbing agents were introduced into glass. However, application of UV shielding film on window glass is more convenient. Introduction of TiO2 and CeO2 is known to provide highly UV-absorbent coatings, which is much more stable than the UV shielding film based on organic UV absorbing dyes. Tomonaga [124] noticed that the film containing Ce and Ti has a high refractive index reaching 2.1, showing marked interference colors of reflected light. This intense coloring was suppressed by inserting an intermediate layer consisting of TiO2eSiO2.

4.4.3. Colored Film Neutral-colored glasses are important for car windows. Some passengers request having dark tinted windows for the privacy, comfort, and distinguished design. In order to accomplish this color by coating window glasses with gray color film, films consisting of two layers are applied to the glass; the inner layer consists of CueMneCoeCreO spinel-type oxide pigment and the outer layer consists of SiO2 matrix containing oxides of copper, manganese, cobalt, and chromium [123]. After coating with 200 nm thickness inner layer and 30 nm thickness outer layer, the coating film is fired at 700  C for a few minutes. After the treatment glass shows lower transmittance in the whole wavelength region, effectively shielding ultraviolet and infrared rays and darkening the interior of automobiles for privacy.

4.4.4. UV Shielding Film UV shielding glass is needed for automobile windows, in order to protect passengers and internal commodities from the detrimental effect of UV rays. For this purpose,

4.4.5. IR-Shielding Film IR (infrared ray) shielding film for automobile windows are needed to suppress the temperature rise within cars and keep comfort of passengers. For windshield, IR shielding is carried out through the intermediate film in glass, while IR shielding of front-door windows are carried out by applying a film to the surface of glass. At present, IR shielding for front doors [125] is based on ITO (In2O3-doped with SnO2) fine particles in SiO2 film. ITO particles absorb the IR of 0.3e2 mm wavelengths, while they do not absorb the visible lights so much. Therefore, the transmittance of the film over 70% required by the regulation on automobiles is obtained. To coat sodalime glass with SiO2eITO infrared-shielding film, a solution containing 3-glycidoxypropyltrimethoxysilane and tetraethoxysilane as source of SiO2 is used. The coating film heat-treated at 160  C shows high strength and high pencil hardness of 9H, besides high IR-shielding property, that is, low transmittance in the infrared region of wavelengths longer than 1.2 mm.

896

4.4.6. Water Repellent Film In order to drive safely and comfortably on rainy days without wipers, water-repellent surface of windshield of cars is required. Originally, water-repellent reagent was applied to the surface of glass, and this worked for a while. In the early 1990s, sustainable water-repellent coating films on glass substrate was developed, based on the solegel method [127]. The coating film is prepared by a two-step process. In the first step, porous SiO2 film is formed by heating alkoxy-derived SiO2 gel layer at 290  C and in the second step, partially hydrolyzed fluoroalkyltrimethoxysilane is applied on the precoated porous SiO2 layer. The coating film showed excellent water repellency expressed by the contact angle of water as high as 100 . Around the year 2000, it was discussed that for the clear sight the glass surface should have high water drop slidability as well as high water repellency. Asahi Glass Company [100] developed abrasion-resistant and chemically durable new water repellent film, which shows high water drop slidability as well and applied the film to automobile windows. Such films are also developed in other glass companies [126].

4.4.7. Hydrophilic Coating Hydrophilic coatings are applied to the outside mirror of cars to enhance rearward visibility on rainy days [128]. When the mirror is covered with rain droplets, the mirror does not work, due to a cloudy surface. The hydrophilic coating spreads out rain drops on the mirror surface to thin water film. Presently, manufactured hydrophilic film consists of SiO2 matrix containing TiO2 fine particles, which show photocatalytic hydrophilicity. There is another type of film, in which the first layer on glass is TiO2 photocatalyst and the second layer on the first layer is very porous SiO2 hydrophilic layer.

Handbook of Advanced Ceramics

Among polymorphs of TiO2, anatase phase shows the most effective photocatalytic effect. The effect of photocatalysts is explained by the following example of photocatalytic decomposition of surface stains, which occurs in photocatalytic self-cleaning coatings [129]. Pairs of electron and hole are excited by UV lights. The photo-generated electrons produce anion radicals like O2 by reacting with O2. The holes produce active OH radicals by reacting with water. Those active species decompose organic compounds, such as stains of oil and fat. Thus, due to its highly oxidizing and reducing power, photocatalyst is employed for decomposition of organic stains, elimination of bad smells, decomposition of nitrogen oxide (NOx), antibacteria, and so on.

4.5.2. Fabrication of Photocatalyst Film: Self-Cleaning of Windows Fabrication of photocatalyst film will be discussed with self-cleaning window glass as example [87,130]. Starting solution containing Si(OC2H5)4, TiO2 fine particles, and SiO2 fine particles is applied onto plate glass by spray coating. TiO2 fine particles used are anatase-phase particles of 20 nm in diameter. The thickness of the film is 200e300 nm. Since a large size plate glass cannot be heated to 400e500  C without breaking, the anatase particles prepared separately by solegel method are used. The addition of silica to TiO2 is made, in order to keep hydrophilicity of the coating film. The prepared film becomes hydrophilic on exposure to UV rays in solar radiation, working as self-cleaning film [131]. Photocatalyst decomposes organic contaminants on the coating under sunshine, and rain enters between the nondecomposed contaminants and coating, eliminating the contaminant.

4.5.3. Improvement of Photocatalytic Activity 4.4.8. Antifogging Film An antifogging film [84] was developed based on polyurethane, which absorbs much water. The film is applied to the inner surface of windshield. The antifogging film of 20 mm thickness shows an outstanding antifogging performance even when the outside temperature is as low as the freezing temperatures.

4.5. Photocatalyst Film 4.5.1. Performance, Preparation and Application of Photocatalyst Photocatalysts are defined as materials which decompose detrimental substances under the sun lights containing UV rays. Mainly, TiO2 is used as photocatalyst at present.

A large number of papers have been published on improvement of the activity of TiO2 photocatalysts. One of the methods is an increase in surface area of TiO2 photocatalyst. Luo [132] accomplished some improvement of catalytic activity by preparing TiO2 anatase nanocrystals with a diameter of 8e10 nm using a mixed template of polyethylene glycol and cetyltrimethoxylammonium bromide. The resultant TiO2 crystals showed high photocatalytic activity determined by reaction 3I /I3 þ 2e . Another method of improving photocatalytic activity is to dope TiO2 photocatalyst. As an example, Raileanu [133] prepared coatings of TiO2 doped with S and Ag for water purification, showing that the photocatalytic activity of the doped samples is higher than that of undoped sample in the test using degradation of organic chloride compounds from aqueous solution.

Chapter | 11.1.2

SoleGel Process and Applications

897

4.5.4. Visible-Light Photocatalyst In TiO2 anatase photocatalyst, which is currently available for practical use, only UV lights contained in the sun lights work for photocatalysis. On the other hand, visible-light photocatalysts [134] make use of visible lights, working inside of houses and buildings as well as outside, and so, their development is highly desired. Visible lights occupy most of the sun lights. At present, however, the efficiency of visible-light photocatalyst is relatively low, and exploration of highly efficient materials is needed. In addition, the visible-light photocatalyst films are necessarily colored and cannot be applied to windows which are basically colorless and transparent. There are visible-light photocatalysts doped with transition element, such as Cr and V [135], and those doped with anion, such as N [136]. It appears that TiO2exNy, that is, TiO2 doped with N, is more efficient. Also, TiO2 doped with two kinds of anions Br and N is proposed [86]. Besides these TiO2-based materials, microcrystalline CuAl2O4 spinel [137] and In2O3eCaIn2O4 [138] are reported to show visible-light photocatalytic activity.

4.6. Transparent Conductive Film Transparent conductive coating films are employed as transparent electrodes of solar cell, liquid crystal display panel, electrochromic devices [82], and so on. Most important ones are In2O3:Sn film (ITO film) and SnO2:Sb (ATO film). Low resistivity is prerequisite for practical application of transparent conductive films. At present

many of the transparent conductive films are fabricated by PVD methods such as sputtering and vacuum vapor deposition. It is known that the resistivity of ITO film prepared by the optimum condition for low resistivity is 1  10 4 ohm cm. It is generally believed that ITO shows lower resistivity than ATO. Solegel preparation of ATO film was carried out in the mid-1980s [139]. The ATO film was prepared from a starting solution containing Sn(OC4H9)4 and Sb(OC4H9)3 as source of SnO2 and Sb2O3, respectively. Resistivity of the film was 2.5  10 3 ohm cm, lower than that (7.5  10 3 ohm cm) of coating prepared on the high temperature substrate by spraying. Kodaira [140] used dipcoating method. Solegel preparation of ITO film was published in 1982 [89]. Soda-lime glass substrate was coated with the solution of acetylacetonates of indium and tin by dip coating. Heating of the gel film at 500  C produced ITO film thinner than about 10 nm. The lowest resistivity of 1.3  10 3 ohm cm was attained when 3e7% SnO2 was added to In2O3. Furusaki [141] prepared ITO films from solution of In2(SO4)3$nH2O and SnSO4 and from colloidal particles synthesized from In(NO3)4$3H2O and SnCl4. It is found that the resistivity of the film decreases upon heating in vacuum. The lowest resistivity obtained by [142] is 6e8  10 4 ohm cm. Lower resistivity of 2.7  10 4 ohm cm by dip coating [143] and 9.5  10 5 ohm cm [144] are reported. Examples of reported resistivities of ITO and ATO are shown in Table 9. It is desirable to continue research on this subject so that low resistivity can be obtained constantly for glass substrates of large size.

TABLE 9 Electrical Resistivity of Transparent Conducting Coatings Prepared by SoleGel Method No. (1) (2) (3) (4)

(5)

Kind of film ATO ATO ITO ITO

ITO

Thickness (nm)

Resistivity (ohm cm)

247

2.5  10

140

3

2 nm) and superporous silica with a lamellar texture (pore size 1e2 nm). Ionic liquids are efficient solvents for the synthesis of ambient pressure dried aerogels, which have a specific pore volume similar to supercritically dried aerogels. It is interesting to note that ionic liquids can act as a catalyst for both hydrolysis and polycondensation, because their anions are good Lewis base and their cations work as Broensted acid. This means that hydrolysis and polycondensation of silica gel precursors such as tetramethoxy silane (TMOS) and methyltrimethoxysilane (MTMS) can be made. Karout [223] prepared aerogels from ionic liquideSiO2 precursor (tetramethoxysilane, methyltrimethoxysilane)e water solution to compare the effect of the kind of ionic liquid. CO2 supercritical drying of gels was applied. The ionic liquids used were 1-butyl-3-methylimidazolium

Chapter | 11.1.2

SoleGel Process and Applications

tetrafluroborate ½ðBuMeImþ ÞBF4 Š, N-butyl-3-methylimidazolium tetrafluoroborate ½ðBuMePyÞþ BF4 Š, 1-hexyl3-methylimidazolium chloride [(HeMeImþ)Cl ], and 1-butyl-3-methylimidazolium chloride [(BuMeImþ)Cl ]. The resulting materials presented a new type of mesoporous aerogel textures, indicating that ionic liquids can be used as template to produce a regular array of pores in the mesoporous range. Donati [224] prepared silica xerogels by the solegel technique using tetraethoxysilane as precursor and hydrofluoric acid as catalyst, together with an imidazolium as ionic liquid. Three kinds of imidazolium-type ionic liquids were used. The three different anions of the ionic liquids led to different silica products of a compact lamellar monolith, a free-flowing powder of aggregated spheric particles, and solid aggregates with honeycomb structure. In view of the microstructure and forms of silica products mentioned above, processing with ionic liquids may be applied to formation of catalyst, catalyst support, chromatographic material, immobilizer, optics and electronics, and template for the formation of advanced materials.

5.5. SoleGel Nanotubes Carbon nanotubes have attracted much attention owing to their versatility in the field of advanced technology. In the area of inorganic materials, imogolites that are naturally occurring aluminosilicate hydrate nanotubes are familiar and their artificial synthesis is developed for practical application. Imogolites are nanotubes which have the diameter of 1.8e2.2 nm and the length of several to several tens of nm. Triggered by these nanotubes, many other nanotubes have been prepared and some of them have been prepared by solegel technique. Table 12 shows examples of inorganic nanotubes prepared by solegel method. Silica nanotubes with coiled channels in the walls were prepared [225], using self-assemblies of a gelator

903

L-PHePyBr as template from the solution tetraethoxysilane, ammonia, and hydrochloric acid. It is noted that the structure of silica nanotubes obtained is not simple but the wall has a helical structure. TiO2 nanotubes were prepared by Kasuga [226]. TiO2 nanotubes are made by treating solegel-derived fine TiO2eSiO2 powders of the size less than 45 mm prepared with 5e10 M NaOH aqueous solution for 20 h at 110  C. It was shown that TiO2 nanotubes have high photocatalytic activity due to their large surface area. B-doped TiO2 nanotubes [227] prepared by the combination of solegel process with hydrothermal treatment showed significantly enhanced photocatalytic activity compared with pure TiO2 nanotubes. ZnO hexagonal microtubes were fabricated [228] by microemulsion-based hydrothermal method using aqueous reverse micelles as template. Vanadium oxide nanotubes were prepared [229] by mixing hexadodecylamine with V2O5$nH2O gels followed by hydrothermal treatment. Lead zirconate nanotubes were prepared [230] using anodic porous alumina template. The nanotubes obtained have heights of 8e30 mm and average pore diameters in the range 110e200 nm. DC electrophoresis served in growing PbZrO3 nanotubes. SrBi2Ta2O9 (SBT) nanotubes were fabricated [158] by dipping the anodic porous alumina template into an SBT sol and calcining at 700  C. Homogeneous SBT sol was synthesized using ethoxytantalum, strontium acetate, and bismuth subnitrate as starting materials, ethylene glycol as solvent, and glacial acetic acid as catalyst. Resultant nanotubes had smooth morphologies and well-defined diameters corresponding to the diameter of the applied template. Nanotubes of WS2 were prepared by Li [231]. They were obtained by heating (800  C in Ar) tungsten sulfide composites prepared based on template self-assembly of WS4 2 and cationic surfactants in solution.

TABLE 12 Nanotubes of Inorganic Compounds Prepared by SoleGel Technique Compound

Diameter

Length

Remark

Reference

SiO2

25 nm

0.2 mm

Gelator template

[225]

TiO2

8 nm

0.1 mm

Anatase phase

[226]

TiO2: B

30 nm

>0.5 mm

Anatase phase

[227]

ZnO

500 nm

1 mm

Hexagonal microtube

[228]

V2O5

50e130 nm

1e10 mm

Hydrothermal treatment

[229]

PbZrO3

150 nm

>5 mm

Anodic alumina template

[230]

SrBi2Ta2O9

250 nm

>1 mm

Anodic alumina template

[158]

WS2

5e37.5 nm

0.2e0.5 mm

Template self-assembly

[231]

904

Handbook of Advanced Ceramics

5.6. Fuel Cells Fuel cells attract much attention as environmentallyfriendly power source replacing fossil fuels such as petroleum and natural gas. Oxide fuel cells (OFCs) using oxide ion-conducting zirconia as electrolyte work at high temperatures of 800e1000  C and are employed for largescale power generation, while proton-exchange membrane fuel cells (PEMFCs) work at low temperatures below or just above 100  C, being aimed at the use in automobiles and houses. Solegel technology finds the application to both types of fuel cells [232]. At present, however, great efforts are directed to improvement of the electrolyte of PEMFC in the field of solegel processing. A PEMFC consists of an electrolyte membrane and carbon paper electrodes contacting its both sides. At the anode, hydrogen molecules are oxidized into protons through the reaction caused by Pt catalyst carried by the electrode and present at the interface between the electrode and the electrolyte membrane: H2 / 2Hþ þ 2e

at anode

(28)

Electrons are taken out of the anode as electric current. Protons diffuse through the membrane, reaching the cathode, and react with oxygen molecules and electrons supplied via the cathode on the catalyst, being reduced to water: 2Hþ þ 1=2O2 þ 2e / H2 O

at cathode

(29)

Commercially available PEMFCs use ion-exchange membrane of fluororesin represented by NafionÒ which shows high proton conductivity at low temperatures near room temperature, but at temperatures higher than 80  C the conductivity and chemical durability of the film decrease. The problem with conventional PEMFCs which are operated at temperatures lower than 80  C is CO poisoning of Pt catalyst. In order to improve conversion efficiency and avoid CO poisoning of catalyst, PEFMC has to be operated at higher temperatures of 130e150  C, retaining some water for high proton conductivity. Porous inorganic or organiceinorganic hybrid membranes prepared by solegel processing are assumed to satisfy these conditions. Tadanaga [233] prepared proton conductive inorganice organic hybrid films, which show the ionic conductivity of about 1  10 2 S cm 1 under 25% relative humidity at 130  C, using epoxycyclohexylethyltrimethoxysilane and orthophosphoric acid. A PEMFC fuel cell using these films shows a power of about 40 mW cm 2 with current density of 110 mA m 2. Daiko [234] prepared highly conductive P2O5eSiO2 gels by solegel method for PEMFC. Xion [235] prepared highly conductive P2O5eSiO2 gels with ordered mesopores of 5e7 nm using template. Very interesting electrolytes which show high ion conductivity without water

were fabricated [236]. They are P2O5eSiO2 gels prepared by the use of ionic liquid of the imidazolium system. Their conductivity at 250  C is 1 mS/cm in the order of magnitude, which is shown without any humidity.

6. CONCLUDING REMARKS The solegel method supports advanced technologies by supplying advanced nanomaterials. It was shown that solegel processing not only creates novel materials but also supplies better methods of fabricating novel materials explored by conventional sintering and sputtering.

REFERENCES [1] Kistler SS. Coherent expanded aerogels. J Phys Chem 1932; 36:52e64. [2] Roy R. Aids in hydrothermal experimentation: II, Methods of making mixtures for both ‘dry’ and ‘wet’ phase equilibrium studies. J Am Ceram Soc 1956;39:145e6. [3] Schroeder H. Oxide layers deposited from organic solutions. In: Haas G, Thun RE, editors. Phys. thin films, vol. 6. Academic Press; 1969. p. 87e141. [4] Mazdiyasni KS, Dollof RT, Smith II JS. Preparation of highpurity submicron barium titanate powders. J Am Ceram Soc 1969;52:523e6. [5] Dislich H. New routes to multicomponent oxide glasses. Angew Chem Int Ed Engl 1971;10:363e70. [6] Sakka S, Kamiya K, Yamanaka I. Non-crystalline solids of the TiO2eSiO2 and Al2O3eSiO2 systems formed from alkoxides. In: Kunugi M, Tashiro M, Soga N, editors. Proc. Xth Int. Congress on Glass, Kyoto, Japan, vol. 13. The Ceramic Society of Japan; 1974. p. 44e8. [7] Schmidt. Organically modified silicate by the solegel process. In: Brinker CJ, Clark DE, Ulrich DR, editors. Mat. Res. Soc. Symp. Proc. Better ceramics through chemistry, vol. 32; 1984. p. 327e35. [8] Avnir D, Levy D, Reisfeld R. The nature of the silica cage as reflected by spectral changes and enhanced photostability of trapped rhodamin 6G. J Phys Chem 1984;88:5956e9. [9] Beck JS, Vartuli JC, Roth WJ, Leonowitcz ME, Kresge CT, Schmitt KD, et al. A new family of mesoporous molecular sieves prepared with liquid crystal template. J Am Chem Soc 1992;114:10834e43. [10] Nakanishi K, Takahashi R, Nagakane T, Kitayama K, Koheiya N, Shibata H, et al. Synthesis and characterization of microporous silica prepared with sodium silicate and organosilane compounds. J SoleGel Sci Technol 2000;17:191e8. [11] Brinker CJ, Scherer GW. SoleGel science. Academic Press; 1990. [12] Sakka S. Science of solegel method. Tokyo: Agne-Shofu-Sha; 1987 [in Japanese]. [13] Sakka S. Applications of solegel method. Tokyo: Agne-ShofuSha; 1997 [in Japanese]. [14] Liu R, Xu Y, Wu D, Sun Y, Gao H, Yuan H, et al. Comparative study on hydrolysis kinetics of substituted ethoxysilanes by liquid state 29Si-NMR. J Non-Cryst Solids 2004;343:61e70.

Chapter | 11.1.2

SoleGel Process and Applications

[15] Schmidt H, Scholze AH, Kaiser A. Principles of hydrolysis and condensation reaction of alkoxy silanes. J Non-Cryst Solids 1984;63:1e11. [16] Aelion R, Loebel A, Erich F. Hydrolysis of ethyl silicate. J Am Chem Soc 1950;72:5705e12. [17] Brinker CJ, Keefer KD, Schaefer DW, Assink TA, Kay BD, Ashley CS. Solegel transition in simple silicates II. J Non-Cryst Solids 1984;343:61e70. [18] Sakka S, Kozuka H, Kim S-H. Various factors affecting the conversion of silicon alkoxide solutions to gels. In: Mackenzie JD, Ulrich DR, editors. Ultrastructure processing of advanced ceramics. John Wiley and Sons; 1988. p. 159e71. [19] Yamane M, Inoue S, Yamamori A. Solegel transition in hydrolysis of silicon methoxide. J Non-Cryst Solids 1984;63:13e21. [20] Keefer KD. The effect of hydrolysis conditions on the structure and growth of silicate polymers. In: Brinker CJ, Clark DE, Ulrich DR, editors. Better ceramics through chemistry. NorthHolland; 1984. p. 15e24. [21] Sakka S, Kamiya K. The solegel transition in the hydrolysis of metal alkoxides in relation to the formation of glass fibers and films. J Non-Cryst Solids 1986;83:353e74. [22] Pope EJA, Mackenzie JD. Solegel processing of silica II. The role of the catalyst. J Non-Cryst Solids 1986;87:185e98. [23] Nakanishi K. Solegel process of oxides accompanied by phase separating. Bull Chem Soc Jpn 2006;79:673e91. [24] Nakanishi K, Soga N. Phase separation in gelling silica-organic polymer solution: systems containing poly[sodiumstyrenesulfonate]. J Am Ceram Soc 1991;74:2518e30. [25] Nakanishi K, Soga N. Phase separation in silica solegel system containing polyacrylic acid I. Gel formation behavior and effect of solvent composition. J Non-Cryst Solids 1992;139:1e13. [26] Livage J. Molecular design of transition metal alkoxide precursors. In: Lee BT, Pope EJA, editors. Chemical processing of ceramics. Marcel Dekker; 1994. p. 3e21. [27] Sanchez C, Livage J, Henry M, Babonneau F. Chemical modification of alkoxide precursors. J Non-Cryst Solids 1988;100:65e76. [28] Hubert-Pfalzgraph LG. Metal alkoxides for electrooptic ceramics. In: Lee BT, Pope EJA, editors. Chemical processing of ceramics. Marcel Dekker. Inc; 1994. p. 23e57. [29] Kessler VG. The synthesis and solution stability of alkoxide precursor. In: Kozuka H, editor. Solegel science and technology. Solegel processing, vol. I. Kluwer Academic Publishers; 2004. p. 3e40. [30] Tohge N, Shimmou K, Minami T. Effects of UV-irradiation on the formation of oxide thin films from chemically modified metal alkoxides. J SoleGel Sci Technol 1994;2:581e5. [31] Tu YL, Milne SJ. A study on the effects of process variables on the properties of PT films produced by a single-layer sol-gel technique. J Sol-Gel Sci Technol 1995;30:2507e16. [32] Kakihana M, Kobayashi M, Tomita K, Petrykin V. Application of water-soluble titanium complexes as precursors for synthesis of titanium-containing oxides via aqueous solution process. Bull Chem Soc Jpn 2010;83:1323e32. [33] Philipp G, Schmidt H. New materials for contact lenses prepared from Si- and Ti alkoxides by the solegel process. J Non-Cryst Solids 1984;63:283e92. [34] Sakka S, Tanaka Y, Kokubo T. Hydrolysis and polycondensation of dimethyldiethoxysilane and methyltriethoxysilane as materials for the solegel process. J Non-Cryst Solids 1986;82:24e30.

905

[35]

[36]

[37]

[38]

[39]

[40]

[41] [42]

[43]

[44]

[45]

[46]

[47]

[48]

[49]

[50]

[51]

[52]

Huang Q, Mackenzie JD. A study on reinforcement of rubber-like ormosils. In: Pope EJA, Sakka S, Klein LC, editors. Ceramic Transactions. Solegel science and technology, vol. 55. The American Ceramic Society; 1995. p. 291e8. Abe Y, Namiki T, Tsuchida K, Nagao Y, Misono T. Preparation and properties of silicon-containing hybrid gels from vinyltrimethoxysilane. J Non-Cryst Solids 1992;147/148:47e51. Kozuka H, Fujita M, Tamoto S. Polysilazane as the source of silica: the formation of dense silica coatings at room temperature and the new route to organic-inorganic hybrids. J SoleGel Sci Technol 2008;48:148e55. Toki M, Chow TY, Ohnaka T, Sanuma H, Saegusa T. Structure of poly(vinylpyrrolidone)-silica hybrid. Polym Bull 1992;29: 653e60. Yagi K, Shibata S, Yano T, Yasumori A, Yamane M, Dunn B. Photostability of the laser dye DCM in various inorganic-organic host. J SoleGel Sci Technol 1995;4:67e73. Hoebbel D, Reinert T, Endres K, Schmidt H. Synthesis and reactions of unsaturated ligands containing Al and Zr alkoxide precursors. In: Sanchez C, Ribot F, editors. Proceedings of first European workshop on hybrid organiceinorganic materials, France, Nov. 8e10, 1993; 1993. p. 319e23. Hasegawa I. Building block approach to organic/silica hybrid materials. J SoleGel Sci Technol 1995;5:93e100. Kresge CT, Leonowicz ME, Roth WJ, Vartuli J, Beck JS. Ordered mesoporous molecular sieves synthesized by a liquid-crystal template mechanism. Nature 1992;359:710e2. Shimojima A, Liu Z, Ohsuna T, Kuroda K. Self-assembly of designed oligomeric siloxane with alkyl chains into silica-based hybrid mesostructures. J Am Chem Soc 2005;127:14108e16. Shimojima A, Kuroda K. Designed synthesis of nanostructured siloxane-organic hybrids from amphiphilic silicon based precursors. Chem Rec 2006;6:53e63. Shimojima A, Goto R, Atsumi N, Kuroda K. Self-assembly of alkyl-substituted cubic siloxane cages into ordered hybrid materials. Chem Eur J 2008;14:8500e6. Acosta S, Corriu R, Leclercq D, Mutin PH, Vioux A. Novel nonhydrolytic solegel route to metal oxides. J SoleGel Sci Technol 1994;2:25e8. Vioux A, Mutin PH. Non-hydrolitic solegel technology. In: Kozuka H, editor. Handbook of solegel science and technology. Solegel processing, vol. 1. Kluwer Academic Publishers; 2004. p. 621e37. Hay JN, Ravel HM. Synthesis of organiceinorganic hybrids via the non-hydrolytic solegel process. Chem Mater 2001;13: 3396e403. Niida H, Takahashi M, Uchino T, Yoko T. Preparation of organicinorganic hybrid precursors O ¼ P(OSiMe3)x(OH)3 x for lowmelting glasses. J Ceram Soc Jpn 2003;111:171e5. Sakka S, Kokubo T. Preparaion of glasses and ceramics for electrical use based on alkoxide and unidirectional solidification method. Jpn J Appl Phys 1983;22(22-2):3e7. Adachi T, Sakka S. Preparation of monolithic silica gel and glass by the solegel method using N, N-dimethylformamide. J Mater Sci 1987;22:4407e10. Kozuka H, Yamaguchi J, Sakka S. A new method for making opaque silica gel monolith from silicon alkoxide solutions. In: Mazurin OV, editor. Proceedings of the XVth International

906

[53]

[54]

[55] [56]

[57]

[58]

[59]

[60]

[61]

[62]

[63]

[64]

[65]

[66] [67]

[68]

[69]

[70]

[71]

Handbook of Advanced Ceramics

Congress on Glass, Leningrad, USSR, vol. 2a. Nauka; 1989a. p. 32e7. Edler KJ. Formation of ordered mesoporous thin films through templating. In: Kozuka H, editor. Solegel science and technology, vol. 1. Kluwer Academic Publishers; 2004. p. 541e97. Sakka S. Solegel technology as representative processing for nanomaterials; Case studies on the starting solution. J SoleGel Sci Technol 2008;46:241e9. Zarzyski J, Prassas M, Phalippou J. Synthesis of glasses from gels: the problems of monolithic gels. J Mater Sci 1982;17:3371e9. Kozuka H, Sakka S. Formation of particulate opaque silica gels from highly acidic solutions of tetramethoxysilane. Chem Mater 1989b;1:398e404. Yamane M, Aso S, Okano S, Sakaino T. Low temperature synthesis of a monolithic silica glass by the hydrolysis of a silica gel. J Mater Sci 1979;14:607e11. Trevor DJ. Fabrication of large near net shapes of fiber optic quality silica. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 27e64. Sakka S, Kamiya K. The solegel transition in the hydrolysis of metal alkoxides in relation to the formation of glass fibers and films. J Non-Cryst Solids 1982;48:31e46. Kamiya K. Processing of fibers. In: Kozuka H, editor. Handbook of solegel science and technology, vol. 1. Kluwer Academic Publishers; 2004a. p. 401e15. Sakka S. Viscosity and spinnability of gelling solutions. In: Almeida RM, editor. Solegel science and technology, vol. 2. Kluwer Academic Publishers; 2004a. p. 237e63. Kamiya K. Solegel prepared glass and ceramic fibers. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004b. p. 207e19. Brinker CJ, Hurd AJ, Shunk PR, Frye GC, Ashley CS. Review of solegel thin film formation. J Non-Cryst Solids 1992;147/ 148:424e36. Strawbridge I, James PF. The factors affecting the thickness of solegel derived silica coatings prepared by dipping. J Non-Cryst Solids 1986;86:381e93. Belleville PF, Floch HG. A new room temperature deposition technique for optical coatings. J SoleGel Sci Technol 1994;3:23e9. Guglielmi M, Zenezini S. The thickness of solegel silica coatings obtained by dipping. J Non-Cryst Solids 1990;121:303e9. Yamamoto Y, Kamiya K, Sakka S. Study on properties of coating films prepared by solegel method. J Ceram Soc Jpn 1982;91:328e33 [in Japanese]. Kozuka H. Fundamental issues on solegel coatings e stress evolution, cracking and radiative striation. In: Kozuka H, editor. Handbook of solegel science and technology, vol. 1. Kluwer Academic Publishers; 2004. p. 247e307. Makishima A, Asami M, Wada K. Preparation and propertiesof TiO2eCeO2 coatings by the solegel process. J Non-Cryst Solids 1990;121:310e4. Innocenzi P, Kozuka H, Sakka S. Preparation of coating films doped with gold metal particles from methyltriethoxysilane-tetraethoxysilane solutions. J SoleGel Sci Technol 1994;1:305e18. Kozuka H. On ceramic thin film formation: evolution of stress, cracks and radiative striations. J Ceram Soc Jpn 2003; 111:624e32.

[72] Uchiyama H, Namba W, Kozuka H. Spontaneous formation of linear striations and cell-like patterns on dip-coating titania film prepared from alkoxide solutions. Langmuir 2010;26: 11479e84. [73] Yamashita H, Yoko T, Sakka S. Preparation of Li2B4O7 thin films with preferential orientation by solegel method. J Am Ceram Soc 1991;74:1668e74. [74] Schneller T, Waser R. Chemical modifications of Pb(Zr0.3, Ti0.7) O3 precursor solutions and their influence on the morphological and electrical properties of the resulting thin films. J SoleGel Sci Technol 2007;42:337e52. [75] Sakka S. Applications of solegel technology. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004b. p. 3e26. [76] Sakka S. Current and future possibilities of solegel process. Trans Ind Ceram Soc 2005;64(1):13e9. [77] Reisfeld R. Solegel processed lasers. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 239e61. [78] Wang B, Sun L, Ju H, Zhao S, Deng D, Wang H, et al. Single-phased white-light emitting CaAl2Si2O8: Eu2þ, Mn2þ phosphors prepared by a solegel method. J SoleGel Sci Technol 2009;50:368e71. [79] Kik PG, Polman A. Erbium-doped optical-waveguide amplifiers on silicon. MRS Bull 1998;23:48e54. [80] Uhlmann DR, Motakef S, Suratwara T, Wade R, Teowee G, Boulton JM. Ceramic films for optical applications. J SoleGel Sci Technol 1994;2:335e40. [81] Yoneda T, Sanada Y, Morimoto T. Solegel coatings applied to automotive windows. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 691e702. [82] Heusing S, Aegerter MA. Solegel coatings for electrochromic devices. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 719e60. [83] Gaspera ED, Guglielmi M, Agnoli S, Granozzi G, Post ML, Bello V, et al. Au nanoparticles in nanocrystalline TiO2eNiO films for SPRbased, selective H2S gas sensing. Chem Mater 2010;22:3407. [84] Yamazaki S, Itakura N, Makita K, Tsutsumi K. 2004. In: Proceedings of the 5th ICCG. p. 1e6. [85] Graetzel M. Solar cells based on solegel films. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 259e310. [86] Shen Y, Xiong T, Du H, Shang J, Yang K. Investigation of Br, N codoped TiO2 photocatalysts: preparation and photocatalytic activities under visible light. J SoleGel Sci Technol 2009;52:41e8. [87] Watanabe T, Takami K, Nakajima A, Hashimoto H, Adachi T. Preparation and properties of inorganic-organic graded films. Applications for intermediate layer between TiO2 coatings and polymer products. In: International conference on organiceinorganic hybrids science, technology and applications, June 12e14, 2000. University of Surrey; 2000. Paper 1. [88] Aegerter MA, Al-Dahoudi N. Wet chemical processing of transparent and antiglare conducting ITO coating on plastic substrate. J SoleGel Sci Technol 2003;27:81e9. [89] Ogiwara S, Kinugawa K. Characteristics of transparent conductive In2O3 films prepared by thermal decomposition of indium nitrate acetylacetonate. J Ceram Soc Jpn 1982;90:157e63.

Chapter | 11.1.2

SoleGel Process and Applications

[90] Kato K, Fu D, Suzuki K, Tanaka K, Nishizawa K, Miki T. Ferroand piezoelectric properties of polar-axis-oriented CaBi4Ti4O15 films. Appl Phys Lett 2004;84:3771e3. [91] Van Bael MK, Van den Rub H, Vangenechten, Mullens J, D’Haen J, Goux L, Wouters DJ. Effect of La3þ content on the properties of ferroelectric Bi4 xLaxTi3O12 (BLT) thin films prepared by aqueous chemical solution deposition. J SoleGel Sci Technol 2007;42:239e45. [92] Sowman HG. Aluminaeboriaesilica ceramic fibers from the solegel process. In: Klein LC, editor. Solegel technology. Noyes Publications; 1988. p. 162e83. [93] Yokogawa H. Hydrophobic silica aerogel. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 73e84. [94] Schottner G, Rose K, Posset U. Scratch and abrasion resistant coatings on plastic lenses e state of the art, current developments and perspectives. J SoleGel Sci Technol 2003;27:71e9. [95] Matsuda A. Solegel micropatterning process. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 637e59. [96] Adachi T, Kawashima J, Shosi M, Matsubara M, Sakai K, Nakasone T. Solegel production of silica microparticles. In: Schmidt H, editor. Solegel production. Trans Tech Publications; 1998. p. 1e6. [97] Wang JA, Valenzuela MA, Castiblo S, Salmones J, MoranPineda M. Studies of zirconia-doped ceria nanomaterials for CO and C3H8 oxidation. J SoleGel Sci Technol 2003;26:879e82. [98] Guizard C, Ayral A, Barboin M, Julbe A. Solegel processed membranes. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 139e78. [99] Nakanishi K. Monolithic porous silica for high speed HPLC. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 65e72. [100] Morimoto T, Sanada Y, Tomonaga H. Functional wet chemical coatings for automotive glasses and CRTs. In: Proc. 3rd ICCG 2000; 2000. p. 223e49. [101] Popali M, Kappel J, Pilz M, Schulz J, Feyder G. A new inorganicorganic polymer for the passivation of thin film capacitors. J SoleGel Sci Technol 1994;2:157e60. [102] Amberg-Schwab S. Inorganiceorganic polymers with barrier properties against water vapor, oxygen and migrating monomers. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 455e78. [103] Amberg-Schwab A, Katschorek H, Weber U, Hoffmann M, Burger A. Barrier properties of inorganiceorganic polymers; influence of starting compounds, curing conditions and storage scaling up to industrial application. J SoleGel Sci Technol 2000;19:125e9. [104] Sampath S, Pankratov I, Gun J, Lev O. Solegel derived ceramic carbon enzyme electrodes: glucose oxidase as a test case. J SoleGel Sci Technol 1996;7:123e8. [105] Sakai S, Ono T, Ijima H, Kawakami K. In vitro and in vivo evaluation of alginate/solegel synthesized aminopropyl-silicate alginate membrane for bioartificial pancreas. Biomaterials 2002;23:4177e83. [106] Oveisi H, Rahighi S, Jiang X, Nemoto Y, Beitollahi A, Wakatsuki S, et al. Unusual antibacterial property of mesoporous titania films: drastic improvement by controlling surface area and crystallinity. Chem Asian J 2010;5:1978e83.

907

[107] Livage J, Roux C, DaCosta JM, Desportes I, Quinson JF. Immunoassays in solegel matrices. J SoleGel Sci Technol 1996;7:45e51. [108] Durucan C, Pantano CG. Hybrid sol/gels for DNA arrays and other lab-on-a-chip applications. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 551e75. [109] Zhang P, Xu M, Liu L, Li L. Luminescent properties of Sr3Al2O6: Eu, Pr prepared by solegel method. J SoleGel Sci Technol 2009;50:267e70. [110] Reisfeld R. Spectroscopy and application of molecules in glasses. J Non-Cryst Solids 1990;121:254e66. [111] Thomas IM, Payne SA, Wilkes GD. Optical properties and laser demonstration of Nd-doped solegel silica glasses. J Non-Cryst Solids 1992;151:183e94. [112] Martucci A, Brusatin G, Guglielmi M, Strohhoefer C, Fick J, Pelli S, et al. Fabrication and characterization of solegel GeO2e SiO2 erbium-doped planar waveguides. J SoleGel Sci Technol 1998;13:535e9. [113] Benatsou M, Capoen B, Bouazaoui M, Tchana W, Vilcot JP. Structural and optical properties of solegel derived aluminosilicate planar waveguides doped with Er3þ ions. J SoleGel Sci Technol 1998;13:529e33. [114] Reisfeld R, Gvishi R, Burshtein Z. Photostability of the laser action of the red perylimide dye in composite glass. J SoleGel Sci Technol 1955;4:49e55. [115] Shibata S, Yamane M, Kamada K, Sasaki K. Laser emission from dye-doped organic-inorganic particles of microcavity structure. J Sol-Gel Sci Technol 1995;8:950e64. [116] Mukherjee SP, Lowdermilk WH. Gel-derived single layer antireflection films. J Non-Cryst Solids 1982;48:177e84. [117] Yoldas BE, Partlow DP. Formation of broad band antireflective coatings on fused silica for high power laser applications. Thin Solid Films 1985;129:1e14. [118] Floch HG, Belleville PF. A scratch-resistant single-layer antireflective coating by a low temperature solegel route. J SoleGel Sci Technol 1994;1:293e304. [119] Hinz R, Dislich H. Anti-reflecting light scattering coatings via the solegel procedure. J Non-Cryst Solids 1986;82:411e6. [120] Dislich H. 1994. Private communication. [121] Hattori A, Makita K, Okabayashi S. Development of HUD combiner for automotive windshield application. In: ’89 SPIE international congress and exposition proceedings, vol. 1168; 1989. 272e282. [122] Makita K. Solegel preparation of reflective coatings. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 663e75. [123] Morimoto T. Solegel products news: CueMneCo spinel colored coating for automobile window. J SoleGel Sci Technol 2002;23:88e90. [124] Tomonaga H, Morimoto T. Novel functioning of glass by wet coating process: part 2. Ultraviolet rays shielding glass. New Glass 1997;12(1):61e7 [in Japanese]. [125] Kumon S. Infrared rays cutting glass. In: Sakka S, editor, New Technology of Sol-Gel Processing. CMC Publishing 2010;142e146. [in Japanese]. [126] Kamitani K, Teranishi T. Development of water-repellent glass improved water-sliding property and durability. J SoleGel Sci Technol 2003;26:823e5.

908

[127] Yamazaki S, Inaba H, Sakai H, Tatsumisago M, Tohge N, Minami T. Water-repellent coatings on glass substrates by the solegel process. In: Proceedings of XVI International Congress on Glass (Boletin de la Sociedad de Ceramica y Vidrio 31-C), vol. 7; 1992. 291e295. [128] Yamazaki S, Honjo H, Hata S, Kai Y. SoleGel products news: development of hydrophilic outside mirror coated with titania photocatalyst. J SoleGel Sci Technol 2002;23:267e70. [129] Yoshida N, Watanabe T. Solegel processed photocatalytic titania films. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 355e83. [130] Tanaka H, Obana S. Coatings with photocatalyst on architectural glass. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 385e97. [131] Nakajima A, Koizumi S, Watanabe T, Hashimoto K. Effect of repeated photo-illumination on the wettability conversion of titanium dioxide. J Photochem Photobiol Chem 2001;146:129e32. [132] Luo S, Wang F, Shi Z, Xin F. Preparation of highly active photocatalyst anatase TiO2 by mixed template method. J SoleGel Sci Technol 2009;52:1e7. [133] Raileanu M, Crisan M, Dragan N, Crisan D, Galtayries A, Braileanu A, et al. Solegel doped TiO2 nanomaterials: a comparative study. J SoleGel Sci Technol 2009;51:315e29. [134] Taga Y, editor. Visible-light photocatalyst. CMC Publishing Company; 2005 [in Japanese]. [135] Anpo M, Takeuti M. The design and development of highly reactive titanium oxide photocatalysts operating under visible light irradiation. J Catal 2003;216:505e16. [136] Asahi R, Morikawa T, Ohwaki T, Aoki K, Taga Y. Visible-light photocatalysis in nitrogen-doped titanium oxides. Science 2001;293:269e71. [137] Jiang Y, Li J, Sui X, Ning G, Wang C, Gu X. CuAl2O4 powder synthesis by solegel method and its potodegradation property under visible light irradiation. J SoleGel Sci Technol 2007;42:41e5. [138] Ge L. Preparation of novel visible-light-driven In2O3eCaIn2O4 composite photocatalyst by solegel method. J SoleGel Sci Technol 2007;44:263e8. [139] Gonzales-Oliver CJR, Kato I. Sn(Sb)-oxide solegel coatings on glass. J Non-Cryst Solids 1986;82:400e1. [140] Kodaira K, Sohma M, Furusaki T. Preparation and properties of SnO2 thin films by dip-coating method. In: Ceramic trans. Ceramic thin and thick films, vol. II. The American Ceramic Society; 1990. p. 301e6. [141] Furusaki T, Kodaira K. Preparation and properties of In2O3 thin films by the solegel method. In: Vincenzini P, editor. High performance ceramic films and coatings. Elsevier; 1991. p. 241e7. [142] Furusaki T, Takahashi J, Kodaira K. Preparation of ITO thin films by solegel method. J Ceram Soc Jpn 1994;102:200e5. [143] Seki S, Sawada Y, Nishida T. Preparation of indium-tin-oxide transparent conducting films: influence of the concentration of dip-coating solutions and the withdrawal rate on the film thickness and resistivity. Trans Mater Res Soc Jpn 2001;26(4):1231e4. [144] Kobayashi C, Seki S, Sawada Y. Indium-tin-oxide transparent conducting films deposited by spray method using ethanol solution of indium chloride and tin chloride. Trans Mater Res Soc Jpn 2001;26(4):1227e30.

Handbook of Advanced Ceramics

[145] Xu Y, Chen CJ, Desimone N, Mackenzie JD. Photo-voltaic effect and optical properties of ferroelectric thin films made by solegel processing. In: SPIE Proc. vol. 1328; 1990. 40. [146] Hirano S, Kato K. Preparation of crystalline LiNbO3 films with preferred orientation by hydrolysis of metal alkoxides. Adv Ceram Mater 1988;3:503e6. [147] Nashimoto K, Nakamura S. Preparation and characterization of solegel derived epitaxial and oriented Pb(Zr0.52Ti0.48)O3 thin films. Jpn J Appl Phys 1994;33:5147e50. [148] Yamashita H, Yoko T, Sakka S. Preparation of highly oriented Li2B4O7 films by solegel method. J Ceram Soc Jpn 1990;98:913e6. [149] Ohya Y, Ito T, Takahashi Y. Dielectric properties of multilayered ferroelectric thin films fabricated by solegel method. Jpn J Appl Phys 1994;33:5272e6. [150] Doi H, Atsuki T, Soyama N, Sasaki G. Influence of buffer layers on microstructural and ferroelectric characteristics of solegel derived PbZrxTi1 xO3 thin films. Jpn J Appl Phys 1994;33:5159e66. [151] Xu R, Xu Y, Mackenzie JD. Amorphous thin films of ferroelectric oxides. In: Mackenzie JD, editor. Solegel optics II. SPIE Proc. vol. 1758; 1992a. p. 261e73. [152] Soyama N, Sasaki G, Atsuki T, Yonezawa T, Ogi K. Preparation of dielectric thin films from photosensitive solegel solution. Jpn J Appl Phys 1994;33:5268e71. [153] Payne DA. Powderless processing of ceramics: thin layer electroceramics. J SoleGel Sci Technol 1994;2:311e5. [154] Okuwada K, Imai M, Kakuno K. Preparation of Pb(Mg1/3Nb2/3)O3 thin film by solegel method. Jpn J Appl Phys 1989;28:L1271e3. [155] Kang J, Yoko T, Kozuka H, Sakka S. Preparation of Pb-based comlex perovskite coating films by solegel method. In: Mackenzie JD, editor. Solegel optics II. SPIE Proc., vol. 1758; 1992. p. 249e60. [156] Xu Y, Mackenzie JD. Ferroelectric thin films prepared by solegel processing. Integr Ferroelectrics 1992b;1:17e42. [157] Xu R, Xu Y, Chen CJ, Mackenzie JD. Solegel processing of strontium-barium niobate ferroelectric thin film. J Mater Res 1990b;5:916e8. [158] Feng M, Wang W, Zhou Y, Jia D. Synthesis and characterization of ferroelectric SrBi2Ta2O9 nanotubes arrays. J SoleGel Sci Technol 2009;52:120e3. [159] Sun H, Mao X, Wang W, Chen X. Ferroelectric properties of La and V co-substituted SrBi4Ti4O15 films prepared by solegel method. J SoleGel Sci Technol 2009;52:197e201. [160] Xu J, Shen B, Zhai, J. 2010. Dielectric, ferroelectric and optical properties of BaZr0.2Ti0.8O3 thin films prepared by solegelhydrothermal process. [161] Liu H, Wang X. Large electric polarization in BiFeO3 film prepared via a simple solegel process. J SoleGel Sci Technol 2008;47:154e7. [162] Jo H, Kim S, Do D. Fabrication and properties of xBiFeO3-(1 x) Bi4Ti3O12 system. J SoleGel Sci Technol 2009;49:336e40. [163] Fu D, Suzuki K, Kato K. High piezoelectric response in polaraxis-oriented CaBi4Ti4O15 ferroelectric thin films. Appl Phys Lett 2004;85:3519e21. [164] Kistler SS. Coherent expanded aerogels and jellies. Nature 1931;127:741. [165] Nicolaon GA, Techener SJ. On a new method of preparation of xerogels and aerogels of silica and their textural property. Bull Soc Chim Fr 1968:1900e6.

Chapter | 11.1.2

SoleGel Process and Applications

[166] Fricke J. Aerogels and their applications. J Non-Cryst Solids 1992;147&148:356e62. [167] Tillotson TM, Hrubush LW. Transparent ultralow-density silica aerogels prepared by a two-step solegel process. J Non-Cryst Solids 1992;145:44e50. [168] Yokogawa H, Yokoyama M. Hydrophobic silica aerogels. J NonCryst Solids 1995;186:23e9. [169] Anderson AM, Carroll MK, Green EC, Melville JT, Bono MS. Hydrophobic silica aeregels prepared via rapid supercitical extraction. J SoleGel Sci Technol 2010;53:199e207. [170] Jones SM. Aerogel: space exploration applications. J SoleGel Sci Technol 2006;40:351e7. [171] Yokoyama M. SoleGel products news: hydrophobic silica aerogel. J SoleGel Sci Technol 2000;18:287e8. [172] Aravind PR, Shajesh P, Soraru GD, Warrier KGK. Ambient pressure drying: a successful approach for the preparation of silica and silica based mixed oxide aerogels. J SoleGel Sci Technol 2010;54:105e17. [173] Einarsrud M-A, Haereid S. Preparation of transparent, monolithic silica xerogels with low density. J SoleGel Sci Technol 1994;2:903e6. [174] Parvathy Rao P, Venketeswara Rao A. Hydrophobic and physical properties of the two step processed ambient pressure dried silica aerogels with various exchanging solvents. J SoleGel Sci Technol 2005;36:285e92. [175] Aravind PR, Shajesh P, Mukunden P, Warrier, KGK. 2009. Silicatitania aerogel monoliths with large pore volume and surface aera by ambient pressure drying. [176] Schwertfeger F, Frank D, Schmidt M. Hydrophobic water glass based aerogels without solvent exchange or supercritical drying. J Non-Cryst Solids 1987;225:24e9. [177] Uzma KH, Bangi A, Parvathy Rao H, Hirashima A, Venkateswara Rao A. Physico-chemical properties of ambiently dried sodium silicate based aerogels catalyzed with various acids. J SoleGel Sci Technol 2009;50:87e97. [178] Cao S, Yao N, Yeung K. Synthesis of free-standing silica and titania-silica aeogels with ordered and disordered mesopores. J SoleGel Sci Technol 2008;46:323e33. [179] Luo H, Churu C, Fabrizo EF, Schnohrica J, Hobb A, Pass A, et al. Synthesis and characterization of the physical, chemical and mechanical properties of isocyanate-crosslinked vanadia aerogels. J SoleGel Sci Technol 2008;48:113e34. [180] Bono MS, Anderson AM, Caroll MK. Alumina aerogels prepared via rapid supercritical extraction. J SoleGel Sci Technol 2010;53:216e26. [181] Brown P, Hope-Weeks LJ. The synthesis and characterization of zinc ferrite aerogels prepared by epoxide addition. J SoleGel Sci Technol 2009;51:238e43. [182] Mohanan JL, Brock SL. CdS aerogels: Effect of concentration and primary particle size on surface area and optoelectronic properties. J SoleGel Sci Technol 2006;40:341e50. [183] Lee JK, Gould GL. Polydicyclopentadiene based aerogel: a new insulation material. J SoleGel Sci Technol 2007;44:29e40. [184] Lorjai P, Chaisuwan T, Wanghasemjit S. Porous structure of polybenzoxazine-based organic aerogel prepared by solegel process and their carbon aerogels. J SoleGel Sci Technol 2009;52:56e64. [185] Digambar Y, Nadargi SS, Latthe A, Venkateswara R. Effect of posttreatment (gel aging) on the properties of methyltrimethoxysilane

909

[186]

[187]

[188] [189]

[190]

[191]

[192] [193]

[194]

[195]

[196] [197]

[198]

[199] [200]

[201]

[202]

[203]

[204]

based silica aerogels prepared by two-step solegel process. J Sole Gel Sci Technol 2009;48:172e81. Kanamori K, Aizawa M, Nakanishi K, Hanada T. Elastic organicinorganic hybrid aerogels and xerogels. J SoleGel Sci Technol 2008;48:172e81. Osaka A, Tsuru K, Hayakawa S. Inorganic-organic hybrids for biomedical applications. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 527e50. Besanger TR, Brennan JD. Entrapment of membrane proteins in solegel derived silica. J SoleGel Sci Technol 2006;40:209e25. Juergen-Lohmann DL, Nacke C, Simon LC, Legge RL. Effect of solegel hydrophobicity on the distribution and structure of different proteins in organically modified solegel thin films. J SoleGel Sci Technol 2009;52:370e81. Soltmann U, Boettcher H. Utilization of solegel ceramic for the immobilization of living microorganisms. J SoleGel Sci Technol 2008;48:66e72. Guan C, Wang G, Ji J, Wang J, Tan M. Bioencapsulation of living yeast (Pichia pastoris) with silica after transformation with lysozyme gene. J SoleGel Sci Technol 2008;48:66e72. Braun S, Rappoport S, Zusman R, Avnir D. Biochemically active solegel glass; the trapping of enzymes. Mater Lett 1990;10:1e5. Ellerby LM, Nishida CR, Yamanaka SA, Dunn B. Encapsulation of proteins in transparent porous silicate glasses prepared by the solegel method. Science 1992;255:1113e5. Karout A, Pierre C. Partial transesterification of sunflower oil with ethanol by a silica fiber reinforced aerogel encapsulated lipase. J SoleGel Sci Technol 2009;52:276e86. Pope EJA. Gel encapsulated microorganisms; Saccharomyces cerevisiae-silica gel biocomposites. J SoleGel Sci Technol 1995;4:225e9. Nassif N, Caiffier A, Coradin T, Roux C, Livage J, Bovet O. 2003. Viability of bacteria in hybrid aqueous silica gels. Compostrini R, Carturan G, Caniato R, Piovan A, Filippini R, Innocenzi P, et al. Immobilization of plant cells in hybrid solegel materials. J SoleGel Sci Technol 1996;7:87e97. Bonansegna S, Dal Monte R, Carturan G. Alginate microspheres loaded with animal cells and coated by a siliceous layer. J Sole Gel Sci Technol 2003;26:1151e7. Dave C, Dunn B, Valentine JS, Zink JL. Solegel encapsulation methods for biosensor. Anal Chem 1994;66:1120Ae7A. Chen Z, Kaplan DL, Yang K, Kumar J, Karx KA, Tripathy SK. Phycobiliproteins encapsulated in solegel glass. J SoleGel Sci Technol 1996;7:99e108. Dosoretz C, Arman R, Starosvetsky J, Rothchild N. Entrapment of parathion hydrate from Pseudomonas spp in solegel glass. J SoleGel Sci Technol 1996;7:7e11. Rietti-Shati M, Ronan D, Mandelbaum RT. Atrazine degradation by Pseudomonas strain ADP entrapped in solegel glass. J Sole Gel Sci Technol 1996;7:77e9. Mahltig B, Gutmann E, Reihold M, Meyer DC, Boettcher H. Synthesis of Ag and Ag/SiO2 sols by solvothermal method and their bacteriocidal activity. J SoleGel Sci Technol 2009;51: 204e14. Kokubo T, Ito S, Yamamuro T. Formation of a high-strength bioactive glass-ceramics in the system MgOeCaOeSiO2eP2O5. J Mater Sci 1986;21:536e40.

910

[205] Kim I-Y, Ohtsuki C, Kawakami G, Kamitakahara M, Cho S-B. Preparation of bioactive microspheres of organic modified calcium silicates through solegel processing. J SoleGel Sci Technol 2008;45:43e9. [206] Ma J, Chen CZ, Wang DG, Meng XG, Shi JZ. In vitro degradability and bioactivity of mesoporous CaOeMgOeP2O5eSiO2 glasses synthesized by solegel method. J SoleGel Sci Technol 2010;54:69e76. [207] Roux C, Livage J, Farhati K, Monjour I. Antibody-antigen reactions in porous solegel matrices. J SoleGel Sci Technol 1997;8:663e6. [208] Turnianski A, Avnir D, Bronshtein A, Aharonson N, Altstein M. Solegel entrapment of monoclonal anti-atrazine antibodies. J SoleGel Sci Technol 1996;7:135e43. [209] Fidalgo A, Lopez TM, Ilharco LM. Wet solegel silica matrices as delivery devices for phenytoin. J SoleGel Sci Technol 2009;49: 320e8. [210] Finnie KS, Walkes DJ, Perret FL, Krause-Hener AM, Lin HQ, Hanna JV, et al. J SoleGel Sci Technol 2009;49:12e8. [211] Ito Y, Nogawa M, Takeda M, Shibuya T. Photo-reactive polyvinylalcohol for photo-immobilized microarray. Biomaterials 2005;26:211e6. [212] Usui K, Takahashi M, Nokihara K, Mihara H. Peptide arrays with designed a-helical structures for characterization of proteins from FRET fingerprint patterns. Mol Divers 2004;8:209e18. [213] Takami K, Nakajima A, Fujishima A, Hashimoto H, Watanabe T. Dependence of structure on the process condition of the selforganized inorganic/organic graded thin films utilizing polymer adsorption. In: International conference on organiceinorganic hybrids science, technology and applications, June 12e14, 2000. University of Surrey; 2000. Paper 4. [214] Mizuta Y, Daiko Y, Mineshige A, Kobune M, Yazawa T. Phaseseparation and distribution of phenyl groups for PHTES-TEOS coatings prepared on polycarbonate substrate. J SoleGel Sci Technol 2011;58:80e84. [215] Jitianu A, Amatucci G, Klein LC. Phenyl-substituted siloxane hybrid gels that soften below 140  C. J Am Ceram Soc 2009;92:36e40. [216] Jitianu A, Doyle J, Amatucci G, Klein LC. Methyl modified siloxane melting gels for hydrophobic films. J SoleGel Sci Technol 2010;53:272e9. [217] Takahashi K, Tadanaga K, Matsuda A, Hayashi A, Tatsumisago M. Effects of phenyltriethoxysilane concentration in starting solutions on thermal properties of polyphenylsilsesquioxane particles prepared by a two-step acid-base catalyzed solegel process. J Ceram Soc Jpn 2007a;115:131e5. [218] Takahashi K, Tadanaga K, Matsuda A, Hayashi A, Tatsumisago M. Thermoplastic and thermosetting properties of polyphenylsilsesquioxane particles prepared by a two-step acid-base catalyzed solegel process. J SoleGel Sci Technol 2007b;41:217e22. [219] Matsuda A, Sasaki T, Hasegawa K, Tatsumisago M, Minami T. Thermal softening behavior and application to transparent thick films of poly(benzylsilsesquioxane) particles prepared by the solegel process. J Am Ceram Soc 2001;84:775e80. [220] Klein LC, Jitianu A. Organiceinorganic hybrid melting gels. J SoleGel Sci Technol 2010;55:86e93.

Handbook of Advanced Ceramics

[221] Jeong S, Ahn S-J, Moon J. Fabrication of patterned inorganicorganic hybrid film for the optical waveguide by microfluidic lithography for the optical waveguide by microfluidic lithography. J Am Ceram Soc 2005;88:1033e6. [222] Karout A, Pierre C. Silica xerogels and aerogels synthesized with ionic liquids. J Non-Cryst Solids 2007;353:2900e9. [223] Karout A, Pierre C. Porous texture of silica aerogels made with ionic liquids as gelation catalyst. J SoleGel Sci Technol 2009;49:364e72. [224] Donati RK, Migliorini MV, Benvegnu MA, Stracke MP, Gelesky MA, Pavan, Schrekke CML, et al. Synthesis of silica xerogels with highly distinct morphologies in the presence of imidazolium ionic liquids. J SoleGel Sci Technol 2009; 49:71e7. [225] Pei X, Zhang J, Wang S, Chen Y, Wu X, Li Y, et al. Organization of helical mesoporous silica nanotubes. J SoleGel Sci Technol 2009;50:397e402. [226] Kasuga T, Hirasmatsu M, Hoson A, Sekino T, Niihara K. Formation of titanium oxide nanotube. Langmuir 1998;14:3160e3. [227] Deng L, Chen Y, Yao M, Wang S, Zhu B, Huang W, et al. Synthesis, characterization of B-doped TiO2 nanotubes with high photocatalytic activity. J SoleGel Sci Technol 2010;53:535e41. [228] Wang J, Shi N, Qi Y, Liu M. Reverse micelles template assisted fabrication of ZnO hollow nanospheres and hexagonal microtubes by a novel fast microemulsion-based hydrothermal method. J SoleGel Sci Technol 2010;53:101e6. [229] Chandrappa GT, Steunou N, Cassaignon S, Bauvais C. Vanadium oxide: from gels to nanotubes. J SoleGel Sci Technol 2003;26: 593e6. [230] Nourmohammandi A, Hietschold M. Template-based electrophoretic growth of PbZrO3 nanotubes. J SoleGel Sci Technol 2010;53:342e6. [231] Li YD, Li XL, He RR, Zhu J, Deng ZX. Artificial lamellar mesostructures to WS2 nanotubes. J Am Chem Soc 2002;124: 1411e6. [232] Klein LC, Aparicio M, Damay F. Solegel processing for battery and fuel cell applications. In: Sakka S, editor. Handbook of solegel science and technology, vol. 3. Kluwer Academic Publishers; 2004. p. 311e28. [233] Tadanaga K, Yoshida H, Matsuda A, Minami T, Tatsumisago M. Inorganic-organic hybrid films using epoxycyclohexylethyltrimethoxysilane and orthophosphoric acid for PEFC operated at medium temperatures. Solid State Ionics 2005;175:2997e9. [234] Daiko Y, Akai T, Kasuga T, Nogami M. Remarkable high proton conducting P2O5eSiO2 glass as a fuel cell electrolyte working at sub-zero to 120  C. J Ceram Soc Jpn 2001;109:815e7. [235] Xiong L, Shi J, Zhang L, Nogami M. Facile one-step synthesis of highly ordered bimodal mesoporous phosphosilicate monoliths. J Am Chem Soc 2007;129:11878e9. [236] Lakshminarayana G, Nogami M. 2010. Inorganic-organic hybrid membranes with anhydrous proton conduction prepared from tetramethoxysilane/methyl-trimethoxysilane/trimethylphosphate and 1-ethyl-3methylimidazolium-bis(trifluoromethanesulfonyl) imide for H2/O2 fuel cells.

Chapter 11.1.3

Colloidal Processing Fundamentals David McKinneyy and Wolfgang Sigmundyy y

MM Virtuoso, Inc., Gainesville, Florida 32606, United States, yy 225 Rhines Hall, Department of Materials Science and Engineering, University of Florida, P.O. Box 116400, Gainesville, Florida 32611-6400, United States

Chapter Outline 1. 2. 3. 4.

Introduction of Powder Processing Colloidal Processing Concepts and Theories Rheological Properties of Ceramic Slurries Processing Additives

911 911 915 918

5. Forming Techniques 6. Summary and Outlook References

921 925 925

1. INTRODUCTION OF POWDER PROCESSING

2. COLLOIDAL PROCESSING CONCEPTS AND THEORIES

Ceramics have prohibitively high melting temperatures, far exceeding most other materials. Hence, they typically cannot be processed via melts such as metals and polymers. Instead, 90% of ceramic products begin in powder form. To get to the finished product, practitioners process the powder through several steps. Put simply, ceramic powders are added to liquids, dispersed, formed or cast, dried, and sintered. Advanced ceramics require optimization throughout the entire process. As in baking bread, the final product’s integrity and quality depend on precise execution of process steps. This is vitally important, because with knowledge and understanding of ceramic processing, we can identify the exact step where something went wrong and fix it the next time. Here, we present colloid processing fundamentals, rheological properties of ceramic slurries, information on processing additives, and forming techniques. As we lay out the background science of ceramic processing, we also present general applications for key formulas and techniques. Moreover, we define and establish exact meanings for words common to several disciplines, but which mean different things to scientists in different fields. We hope that interdisciplinary researchers adopt these and standardize their use.

Nearly all ceramics begin in powder form, and today manufacturers can produce them nearly infinitely fine. Indeed, advanced ceramic powders contain submicron- and even nanometer-sized particles. Usually, the first processing step is to add the powder to a liquid to form slurry. Chemically, slurries are akin to mixtures in that the particles do not dissolve into the liquid. Physically, the ceramic particles are ideally dispersed evenly throughout the liquid. In such systems, particles and liquid remain separate entities. By definition, small particles dispersed in liquid form either suspensions or colloids. Suspensions contain particles larger than 1 mm and colloids contain particles 1 nm to 1 mm in size. Also, technically, particles in a suspension can settle and those in a colloid should not.1 Understanding colloid and surface science is important to ceramic processing because this provides the key to optimizing a ceramic product’s characteristics and controlling for flaws later on in the process. The term colloid stems from the Greek word kolla, meaning glue.

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00049-6 Copyright Ó 2013 Elsevier Inc. All rights reserved.

1. One often sees the terms ceramic suspension used synonymously with ceramic colloid. For precision’s sake, we will be consistent and use ceramic colloid throughout.

911

912

Handbook of Advanced Ceramics

When particles are smaller than 100 mm in diameter, surface forces govern their behavior. Unlike gravitational forces, surface forces make particles stick together, hence the term colloid. In ceramic processing, these sticky surface forces must be controlled.2 Stability is the most desired property in a ceramic colloid, and interparticle forces play an important role in stabilization [1]. Tiny particles dispersed in a liquid medium spontaneously clump together due to attractive surface forces. Thus, particle surfaces must be modified to produce repulsive potentials that counteract attractive potentials. Below we introduce attractive and repulsive surface forces that ceramic processors use to optimize ceramic colloids.

2.1. Attractive Surface Forces Dutch scientist and Nobel prize winner Johannes Diderik van der Waals (1837e1923) discovered attractive forces between atoms and molecules in the 1870s. He made changes to Boyle’s ideal gas law to make it work for real gases with polar behavior, e.g. H2O, over a broader temperature range and first proposed attractive intermolecular forces. Since then, several physicists have explained their origins. Typically, scientists discuss three types of van der Waals forces, all electromagnetic and dipole interactive: Keesom forces, Debye forces, and London or dispersion forces. The first two require at least one permanent dipole, while the third and strongest depends only on transient dipoles inducing dipoles in neighboring atoms or molecules. Another Dutch scientist, Hugo Christiaan Hamaker (1905e1993), described these energies more quantitatively in 1937. He showed that the energy of van der Waals attraction EvdW equals a materials constant A times distance d and body geometry-dependent function G(d): EvdW ¼ A  GðdÞ

(1)

We now know A as Hamaker’s constant [2], and though this material-specific constant is fundamental, theoretical and experimental access to Hamaker constants still proves difficult even today. As such, we know Hamaker constants for a limited number of materials only. For more detailed information on Hamaker constants and van der Waals forces in ceramic processing, refer to two excellent review papers by Bergstro¨m and French [3,4].

2.2. Repulsive Surface Forces As mentioned earlier, stability is the most desired property in a liquid ceramic colloid. To achieve stabilization, one 2. Force means different things to physicists, surface scientists, and ceramists. To be clear, this chapter deals with variations in electromagnetic force only, i.e. longer-range forces that are not gravitational.

needs to create mutually repulsive potentials on particle surfaces to counterbalance attractive van der Waals potentials. There are four ways to do this: charge ions on particle surfaces (electric double-layer stabilization), adsorb polymer chains onto particles (steric stabilization), combine the two mechanisms (electrosteric stabilization), or add nonadsorbing nanosize objects to the dispersing media (depletion stabilization).

2.2.1. Electric Double-Layer Forces A colloidal particle’s surface charge forms the basis of electric double-layer (EDL) stabilization. Surface charge magnitude determines repulsion magnitude between particles, thereby influencing the stability of the colloid. A perfect stoichiometric crystal is neutral, but as soon as the particle surface oxidizes and hydrolyzes, hydroxyl and other groups form that develop charges when immersed in protic liquid. Surface group ionization or dissolution, specific ion adsorption, and ion exchange on the surface are major sources of these charges for any ceramic [5]. After oxidation, counterions (oppositely charged ions) in the electrolyte containing liquid attract to a particle’s surface charges. This forms a fixed, or Stern, layer on the particle surface and adds a diffuse second layer in solution. The Debye length Eqn (2) describes the ionic double layer’s (EDL’s) dimension, thickness. For a particle surrounded by a diffuse double layer in a shear field, shear slippage occurs at some distance away from the surface, which is where electric potential, or zeta potential, can be measured. Zeta potential is believed to be close to the Stern layer’s potential, and when zeta potential is zero, the corresponding pH level is called the point of zero charge (PZC), or iso-electric point (IEP) (Figure 1). Point of zero charge refers to having neither negative nor positive charges on the surface. IEP applies to complex ceramic surfaces, and the sum of negative and positive charges yields net zero charge. This means locally charged regions are still available at the IEP. Moreover, EDLstabilized ceramic colloids become unstable at the PZC or IEP. Usually, this causes rapid coagulation; however, some systems, like silica in aqueous media f0 < 50mV , may require days for coagulation to happen at the IEP. We discuss rates of coagulation later in this chapter. Table 1 gives PZCs and IEPs of various compounds. Theoretically, the relationship between surface charge density and a dispersed particle’s surface potential is given by [6] 
1 zFf0 2 (2) s0 ¼ ð8εε0 RTcÞ sin h 2RT where ε is the medium’s dielectric constant, ε0 the permittivity of free space, R the gas constant, F the Faraday constant, c the electrolyte ions’ concentration, and z the

Chapter | 11.1.3

Colloidal Processing Fundamentals

913

Another important relationship to note is that ion surface concentration positively correlates to repulsive pressure P (i.e. ion surface concentration is proportional to repulsive pressure P). Therefore, distances between particles become very small, and both ion surface concentration and repulsive pressure approach infinity with high solids loading [7,8]: ps kT D 2s ps ¼ ze

P ¼

(5) (6)

However, this infinite pressure assumes that surface charge remains constant. This does not actually occur. When repulsive pressure P forces two particles close to each other, counterions adsorb onto their original sites. Hence, as D approaches 0, s also falls, and surface potential remains constant. This effect is known as charge regulation. Furthermore, if no binding occurs, surface charge density remains constant, and counterion concentration approaches a constant value 2s/eD at the smallest possible separation distance. In summary: FIGURE 1 Electric double layer. The electric potential as a function of distance from the particle surface. Distance equals 1/k, or the Debye length. Js ¼ surface potential, Jz ¼ zeta potential, J0 ¼ zero potential. (Reprinted from Handbook of Advanced Ceramics, Vol. 1, Eds. Somiya, Aldinger, Claussen, Spriggs, Uchino, Koumoto, Kaneno, Chapter 3, p. 131e185, 2003, First Edition)

electrolyte ions’ valence. For low potentials (i.e. when f0 < 50 mV), the above equation becomes s0 ¼ εε0 kf0

(3)

where 1/k is the Debye length and defined as follows: k2 ¼

2cz2 F 2 εε0 RT

(4)

TABLE 1 Nominal Iso-Electric Points Of Oxides Material

Iso-electric point

a-Alumina (Al2O3)

9 9.5

Silica (SiO2)

3 4

Zinc Oxide (ZnO)

9

Titania (TiO2)

4 5

Calcium Carbonate (CaCO3)

9 10

Lead Oxide (PbO)

10

Magnesia (MgO)

12 13

Zirconia (ZrO2)

4 5

Stannic Oxide (SnO2)

4 7

l

l

constant charge EDL repulsion is always higher than constant potential EDL repulsion. real surfaces show intermediate behavior between constant charge and constant potential [8].

2.2.2. Repulsion between Double Layers Two Russian scientists, Derjaguin and Landau, and two Dutch scientists, Verwey and Overbeek, simultaneously calculated the total forces acting between two particles in the 1940s; hence, the theory of interaction between electric double layers is referred to as DLVO theory. DLVO theory states that the sum of attractive potential energy VA and repulsive potential energy VR gives total interaction energy VT (Figure 2): V T ¼ VA þ V R

(7)

Osmosis causes EDLs to repel particles and overcome attractive van der Waals forces. When two particles with like charges approach one another, their diffuse double layers overlap. This ionic cloud overlap causes local increase in ionic strength, which depletes liquid in the region, yielding an osmotic pressure. Finally, liquid diffuses into the region to equilibrate the concentration, pushing particles apart. If repulsive potential is large enough to overcome attractive van der Waals potential, the colloid has EDL stabilization (Figure 3).

2.2.3. Steric and Electrosteric Forces Sometimes EDLs do not create a large-enough repulsive potential to overcome van der Waals potential. In such

914

Handbook of Advanced Ceramics

other, their layers overlap and penetrate. As they cannot move about freely, conformational entropy decreases. The particles, essentially, cannot come closer than roughly twice the layer thickness. If this distance exceeds the range of van der Waals forces, particles will not stick. Furthermore, a combination of electrostatic and steric forces can stabilize colloids e electrosteric stabilization. Some polymer molecules have groups that ionize in polar solvents to create polyelectrolytes (charged polymers). Solvent conditions decide the fraction of dissociated functional groups f, and as f increases, the polymer charge can vary from zero to highly positive or negative. Figure 4 shows a schematic of the dissociation of poly acrylic acid (PAA) segments. In such cases, repulsion is greater due to the strong long-range EDL repulsion between charged molecules. If the particle itself is charged, it produces longrange repulsive potential when particles approach one another; if the polymer is charged, it produces an additional short-range repulsive potential (Figure 5).

2.2.4. Depletion Stabilization FIGURE 2 Potential energy between two particles in liquid medium due to van der Waals attraction and electric double-layer repulsion. (Reprinted from Handbook of Advanced Ceramics, Vol. 1, Eds. Somiya, Aldinger, Claussen, Spriggs, Uchino, Koumoto, Kaneno, Chapter 3, p. 131e185, 2003, First Edition)

cases, other techniques can be used to stabilize the ceramic colloid: steric stabilization and electrosteric stabilization. Steric forces depend on added polymers (organic macromolecules). They adsorb onto ceramic particles’ surfaces. When two particles with adsorbed polymers approach each FIGURE 3 Overlap of electric double layers due to approach of two parallel surfaces results in ionic concentration increase. This causes repulsion between EDLs. (Reprinted from Handbook of Advanced Ceramics, Vol. 1, Eds. Somiya, Aldinger, Claussen, Spriggs, Uchino, Koumoto, Kaneno, Chapter 3, p. 131e185, 2003, First Edition)

FIGURE 4 Dissociation of poly acrylic acid. (Reprinted from Handbook of Advanced Ceramics, Vol. 1, Eds. Somiya, Aldinger, Claussen, Spriggs, Uchino, Koumoto, Kaneno, Chapter 3, p. 131e185, 2003, First Edition)

Lastly, osmotic repulsion concepts described above for ions in solution or adsorbed polymers apply to dispersed nanoparticles or nonadsorbed, dissolved polymers. For depletion stabilization to work, nonadsorbed nano-objects have to fit in the space between two particles. For recent advances, refer to Rahaman’s Ceramic Processing [9].

2.3. Coagulation Kinetics In the preceding section, we described forces that contribute to colloidal stability. This section covers the

Chapter | 11.1.3

Colloidal Processing Fundamentals

915

FIGURE 5 Schematic of electrosteric stabilization. (Reprinted from Handbook of Advanced Ceramics, Vol. 1, Eds. Somiya, Aldinger, Claussen, Spriggs, Uchino, Koumoto, Kaneno, Chapter 3, p. 131e185, 2003, First Edition)

kinetics of unstable colloids. We repeat, stability is the most desired property, yet most colloids are thermally unstable. Particles remain dispersed only when the potential barrier is [kT. If the potential barrier is kT, particles stick. Thus, it is important to understand a colloid’s rate of coagulation. Technically, a reversible adhesion is called coagulation; by definition, a coagulate forms when particles in an aqueous slurry adhere to each other due to van der Waals attraction. If this association is irreversible, the effect is called flocculation. Colloid particles do not exhibit prescribed kinetic energy. At any given point in time, some particles may have higher energy than others. Napper and Hunter extensively studied the effect of DLVO theory on kinetic behavior of colloidal particles [10]. Stability ratio W is given by W ¼

Number of collisions between particles (8) Number of collisions leading to coagulation

Rf (9) W where Rs is the rate of slow coagulation (when VT [ kT) and Rf the rate of fast coagulation (when VT  kT). Rs ¼

3h 4kTv0

(11)

Thus, smaller particles have faster coagulation rates than larger particles.

2.3.2. Rate of Slow Coagulation With a repulsive potential energy barrier [kT, the rate of coagulation becomes orders of magnitude smaller than in Eqn (10), i.e. slow coagulation. Consider the energy barrier as the activation energy needed for a coagulation event to take place. In this case, the equation for the rate of slow coagulation becomes Rs ¼ RN

4pDv0 
VT =kT dr exp r2

(12)

where VT is the magnitude of the potential energy barrier and r the distance from the particle center. Substituting the values of Rs and Rf in Eqn (9), the value of stability ratio becomes Rf ¼ 2a W ¼ Rs

2.3.1. Rate of Fast Coagulation

ZN


VT =kT dr exp r2

(13)

2a

Smoluchowski addressed the rate of coagulation without any repulsive force, i.e. fast coagulation [11]. He calculated a particle’s flux when it comes in contact with a fixed particle, and the rate of fast coagulation is given by

2

t1=2 ¼

2a

The rate of slow coagulation is given by

Rf ¼ 8pDav20

Time for bulk concentration of particles to reduce by half is given by

(10)

where D ¼ kT/6pah (m /s) (assuming uniform spheres), a is the particle radius (m), v0 is the initial particle concentration (number/m3), and h is the viscosity of dispersing medium (Pas).

This shows that the stability ratio is proportional to the particle radius, indicating that larger particles have a stronger tendency to coagulate.

3. RHEOLOGICAL PROPERTIES OF CERAMIC SLURRIES Rheology is the science of deformation and flow of matter [8]. Practitioners study it to help characterize and optimize particle dispersion, especially in high solids loading and

916

Handbook of Advanced Ceramics

low-viscosity colloids, which describes ideal ceramic colloids. Solids loading refers to volume percent of particles in the slurry.3 Viscosity describes flow behavior of liquids and slurries, suspensions, and colloids. As solids loading increases, the viscosity also increases. When a fluid’s viscosity is independent of the shear rate, the fluid exhibits Newtonian behavior (e.g. water). However, colloids are more often non-Newtonian. This formula applies to viscous fluids4 in which viscosity changes with shear rate at a constant temperature. Viscosity can be defined as [12] s (14) h ¼ g_ where s is the shear stress, g_ the shear rate, and h the shear viscosity. If viscosity and shear rate positively correlate, a fluid is shear thickening. Colloids exhibit this behavior at very high solids loadings. If viscosity and shear rate negatively correlate, a fluid is shear thinning, or pseudoplastic. Finally, other curves are formed when the particle network creates internal stress (Figure 6). Such viscosity measurements can be used to analyze and determine the optimum solids loading for a colloidal system, i.e. achieve maximum solids loading with minimum viscosity, and thereby obtain optimum green-body5 properties after forming. Two major approaches attempt to characterize behavior of colloid systems: hard sphere and soft sphere. The hardsphere approach is simplistic and assumes the sole interaction among particles is repulsion. The soft-sphere approach assumes a potential with a barrier, a primary minimum like in DLVO theory, and applies better to real systems [5].

3.1. Hard-Sphere System and Solids LoadingDependent Viscosity For dispersions of noninteracting particles, viscosity depends on solids loading. Increased solids loading increases viscosity. Einstein proposed a method to describe this phenomenon. It is given by h ¼ hs ð1 þ k1 fÞ

(15)

where h is the slurry viscosity, hs the suspending medium viscosity, k1 the Einstein coefficient (2.5 for spheres), and f the volume fraction of solids in slurry. However, this equation does not account for particle size influence. Hence, it holds only when dispersed

FIGURE 6 Typical rheological behavior of colloidal suspensions. (a) Newtonian, (b) shear thickening, (c) shear thinning/pseudoplastic, (d) Bingham, and (e) shear thickening with yield stress. (Reprinted from Handbook of Advanced Ceramics, Vol. 1, Eds. Somiya, Aldinger, Claussen, Spriggs, Uchino, Koumoto, Kaneno, Chapter 3, p. 131e185, 2003, First Edition)

particles do not interact, e.g. dilute systems. When the system involves interaction between particles, the situation becomes more complex, and the contributions of particle interactions must also be included. If we include higher-order terms of f, the extended Einstein equation becomes h ¼ hs ð1 þ k1 f þ k2 f þ .:Þ

A convenient way of expressing viscosity is by dividing slurry viscosity hs by suspending media viscosity h1. This is a dimensionless quantity called relative viscosity hr. Thus, the effect of volume fraction of solids on viscosity behavior is best studied in relation to the maximum packing fraction. With the addition of more and more particles, at some point they become jam-packed, resulting in a continuous, three-dimensional contact throughout the system. This immobilizes the fluid. At this point, when no flow takes place, the colloid has maximum packing fraction fm . Krieger and Dougherty introduced a model for hard spheres [13,14]:
hr ¼ 1

3. Authors often synonymously use concentration, but this term technically refers to solution molarity. 4. Sigmund et al. illustrate similar curves for viscoelastic fluids [4]. 5. Green body refers to a ceramic product after forming but before drying.

(16)

f fm



½hŠfm

(17)

where [h] is the intrinsic viscosity. However, this model applies only to monosize spheres and does not consider particle-size distribution of ceramic powders.

Chapter | 11.1.3

Colloidal Processing Fundamentals

917

3.2. Effects of Particle-Size Distributions Most commercial ceramic powders have a continuous particle-size distribution. The effect of particle-size distribution on viscosity becomes pronounced at very high solids loading, close to the colloid’s maximum packing fraction. Typically, a polydisperse system’s viscosity is lower than a monodisperse system’s at identical volume fraction, as interstitials between larger particles can be occupied by smaller particles [15]. Also, as the KriegereDougherty Eqn (17) suggests, viscosity of monodisperse suspensions reaches a very high value at low solids loading. For efficient processing, however, high solids loadings are preferable. Colloids consisting of mostly smaller particles exhibit a lower maximum solids loading, thixotropy, or plasticity. This may be caused by the increase in specific surface area binding the dispersing medium. Colloids with a higher percentage of large particles tend to be dilatant. Dilatancy limits processing and, thus, maximum solids loading. If medium-size particles dominate, distribution resembles more and more a monodisperse system. Thus, limited numbers of particles can be packed into the colloid with a limit of about 64 vol% for monosize randomly packed spheres [16].

FIGURE 7 Schematic of pH effect on alumina colloid viscosity. (Reprinted from Handbook of Advanced Ceramics, Vol. 1, Eds. Somiya, Aldinger, Claussen, Spriggs, Uchino, Koumoto, Kaneno, Chapter 3, p. 131e185, 2003, First Edition)

3.5. Electrosteric System Rheology 3.3. pH Variations of Viscosity A colloid’s pH significantly affects the system’s viscosity. In the case of oxides, the viscosity also negatively correlates with absolute value of the zeta potential [12]. Viscosity of alumina colloids increases with increases in pH, up to a certain value. Beyond a certain pH, viscosity becomes inversely proportional to pH. The pH at which viscosity reaches maximum value corresponds with the colloid’s IEP (Figure 7).

3.4. Steric System Rheology In cases of dispersions stabilized by the addition of polymers, their rheology depends on three factors: l

l

l

FloryeHuggins parameter c. If this parameter is 50 vol%). They combine principles of EDL and steric stabilization, or electrosteric stabilization, and they depend on pH and ionic strength, as Naito et al. discuss [17]. At low solids loading (~20 vol%), viscosity is relatively low, and it is affected very little by pH changes. As solids loading increases, however, pH affects viscosity significantly. The amount of added polyelectrolyte also has a profound effect on colloidal rheology. It should be optimized to just saturate the surface. Additional polyelectrolytes result in excess amounts of polymer in the system, and excess polymer can cause depletion flocculation in high solids loaded systems. Conformation of adsorbed polyelectrolyte also plays an important role in rheological behavior of electrosterically stabilized colloids, and, in turn, polyelectrolyte conformation depends on the system’s pH. A detailed study of adsorption behavior on Al2O3 shows that polyelectrolyte adsorption on particles increases as pH decreases. Typically, a 10-fold increase of adsorbed amount is observed from the uncharged to the charged state [18,19]. When pH increases or decreases beyond zero charge, the fraction of the polyelectrolyte dissociated moves toward 1. Hence, charges in the polyelectrolyte repel each other and the molecule stretches. At this moment, two models exist: the charged polyelectrolyte adsorbs flat on the surface or the polyelectrolyte adsorbs in a tail-like brush structure.

918

Handbook of Advanced Ceramics

Conformation shape of the adsorbed polyelectrolyte highly influences dispersion quality. Which types of structures e flat, pancake-like, or brush-like e are achieved depends on adsorption conditions and the materials involved. For pancake-like adsorption, the polymer only contributes short-range repulsive force, and EDL forces of the charged polyelectrolyte mainly contribute to stabilization via long-range interactions. For brush-style structures, the repulsion is much stronger, and true electrosteric contributions are present. Polyelectrolytes can also be used as dispersants when they are uncharged, i.e. at their PZC. However, they will favor coil-like conformations. Hence, much higher molecular weights will be needed to achieve thicker layers of adsorbed polymer coils, and steric forces predominantly contribute to stabilization [1].

4. PROCESSING ADDITIVES Almost all powders start agglomerated in the process. To overcome attractive van der Waals forces, additives must be used to disperse them. Additives come in a wide variety and impact the particle-dispersing medium or particleeparticle interfaces. Both organic and inorganic additives can be used to improve the properties of final ceramic green bodies and sintered products [1]. For example, ceramists use polymeric additives as they provide superior mechanical properties for handling. However, at the same time, such compounds introduce specific problems related to their size. In one such case, the dry powder direct compaction process, ceramists add organic binders to control packing uniformity as fine ceramic raw powders are processed into granules. Each granule should consist of hundreds of homogeneous distributions of particles and binder. However, in reality, granules might be hollow or show strong gradients of binder and fine particles on the surface. Additionally, when dispersing ceramic powders in solvent, dispersants make a high solids loaded colloid stable by adjusting interparticle forces. For certain casting processes, e.g. tape casting or injection molding, selection of suitable additives is vital. The following section describes commonly used additives.

4.1. Dispersing Media/Solvents Dispersing medium is a liquid that determines the slurry’s flow characteristics. It also acts as solvent for organic additives. However, it sometimes behaves disfavorably by dissolving the ceramic powders also. A good dispersing medium must: l l l l

dissolve organic additives, have low viscosity, not foam, have a low boiling point,

l l

not react with the powder, and be nonvolatile and nontoxic.

The right dispersing medium for slurry has to meet all these characteristics for the given powder [20,21]. For example, either water or ethanol is adequate for an alumina system, whereas a BaTiO3 system may need a nonaqueous mixture of trichloroethylene ethanol to avoid dissolution of Ba2þ ions in water. Generally, blends of alcohol and organic solvent, based on powder and dispersant colloidal chemistry, make up the system. This allows other additives to dissolve, and it keeps viscosity low during milling. It is important to note that dispersing medium choice defines both attractive and repulsive forces. High-dielectric versus low-dielectric solvents considerably affect van der Waals interactions [1,22]. Some media allow steric stabilization only, while others allow any choice, including combinations. Moreover, if capillary forces are elected for compaction of the particle system, then the dispersing medium’s properties for capillarity are utmost important [1,23e25].

4.2. Dispersants Dispersants, molecules or polymer chains of typically low-molecular-weight organic chemicals, adsorb onto ceramic particles. Dispersants carry functional groups for anchoring on surfaces and have buoy moiety. The buoy performs best if additional highly polar groups are available, especially carboxylic acid groups, as these develop negative surface charges in protic dispersing media above pH 3. Typical examples include citric or maleic acids. Especially useful in tape-casting processes (covered later in this chapter), dispersants keep a slurry uniform and homogeneous. Equal dispersion helps provide uniform packing in a green body, and it yields uniform shrinkage during drying, binder burnout, and sintering for equiaxed particles. Complete surface coverage of powder with dispersant optimizes dispersion. This avoids bridging and achieves uniform surface properties defined by the dispersant. Complete surface coverage does not require large amounts; for submicron powders, typically less than 1 wt% suffices. As with other additives, dispersants are system specific; however, Menhaden fish oil can be used for a variety of ceramic materials in nonaqueous dispersing media. It is a highly purified, nondrying oil derived from Brevoortia tyrannus, a herring-like fish found off United States’ Eastern coast. When auto-oxidized it transforms into a useful dispersant for ceramic processing. Auto-oxidation causes a poorly controlled polymerization of highly unsaturated triglycerides. The process is stopped at a certain viscosity. Ceramists, technically, describe Menhaden fish oil as a polydisperse polymeric mixture (dispersity above 40,

Chapter | 11.1.3

Colloidal Processing Fundamentals

whereas standard synthetic polymers are around 2) having a molecular weight (Mw) above 100,000. Unfortunately, even after oxidizing, the oil still contains fatty acids, nonreacted triglycerides, and an intermediate molecular weight fraction. This causes multitudinous problems due to the nonreproducibility of the oil and its colloidal stabilizing ability. Also, though not endangered, Menhaden is overfished and populations dwindle sizably each year. Synthetic substitutes are now available. Other organic dispersants include polymers and polyelectrolytes with Mw’s from 2,000 for highly charged polyelectrolytes and 100,000 for polymers. Particle size and dispersing medium (main parameters controlling van der Waals forces) determine the size and required charge of a successful dispersant. In general, the repulsive barrier must be e1.5 kT to have enough Brownian energy to move particles out of attraction and achieve free flow. Typical polar or dissociable functional groups for dispersant include hydroxyl (eOH), carboxyl (eCOOH), sulfonate ð SO3 Þ, sulfate ð OSO3 Þ, ammoniumð NHþ 4 Þ, amino (eHN2), imino (eNHe), and polyoxyethylene (eCH2 CH2Oe) groups. Such nonionic macromolecules in solid/solvent systems mainly contribute to steric stabilization [25]. Table 2 lists commonly used dispersants. A ceramic material’s solubility also strongly influences a dispersant’s effect. Solubility is specific and covers many orders of magnitude. It depends on the material and the dispersing medium’s pH. Higher solubility increases counterions, which compress the EDL. Under these conditions, a purely EDL-stabilized system will collapse. Steric-stabilized systems will also change because solubility and polymer conformation depend on ionic strength. Polymers may salt out, i.e. coagulate. Such actions sometimes occur unwanted, or they may be especially designed, e.g. in direct casting or constant volume processes.

919

Mesmer’s book, The Hydrolysis of Cations, details the solubility of ceramic materials [26].

4.3. Binders Dry ceramic particle networks are brittle and exhibit low compression strengths, typically less than megapascals. To improve handling, ceramists add binders to toughen and strengthen the ceramic, but which burnout in the process. No other material group except organic polymers yields such properties. Polymer strengths derive from their size, i.e. molecular weight. Polymer solubility in the dispersing medium and slurry viscosity limit the polymer’s size; hence, binders are polymers with typically Mw ~ 100,000. In tape-casting processes, a binder adds strength to the dried ceramic tape. This is important especially for storing and machining dry tape [20]. Since the binder is yet another polymer in the slip, surface chemistry must be tailored and processing hierarchy maintained to achieve the final product’s desired, well-defined nanostructure. Ceramists add dispersants first, then the binder. Any other sequence yields unfavorable nanostructures on particle surfaces, which lead to high viscosities and coagulation. The binder’s polymer properties allow the ceramic to maintain its strength and shape until firing [27]. However, binders increase slurry viscosity and change flow characteristics. Thus, binderedispersant and other additives in the ceramic mixture must be compatible to avoid phase separation as the dispersing medium evaporates. A high molecular weight and a low glass transition temperature are advantageous. Binders should: l l l l

TABLE 2 Common Dispersants Used in Ceramic Processing Low molecular weight

Large molecular weight

Sodium pyrophosphate

Poly(acrylic acid) (PAA)

Ammonium citrate

Poly(methacrylic acid) (PMAA)

Sodium citrate

Ammonium polyacrylate

Sodium tartrate

Sodium polyacrylate

Sodium succinate

Sodium polysulfonate

Glyceryl trioleate

Poly(ethylene imine)

Phosphate ester

Menhaden fish oil Random copolymers Comb polymers

form a strong, cohesive tape when dry, be miscible or emulsible in the solvent, be inexpensive, and readily burnout without leaving carbon residue [28].

In fine powder granulation, e.g. spray drying, binders bond hundreds of fine particles to form larger (>50 mm) granules [23]. The binder temporarily bonds the powder particles and gives the green body good handling and storage characteristics. This reduces cracks, pinholes, and other defects in the unfired and fired parts. Table 3 gives typical recipes for spray granule additives, and Table 4 lists binders commonly used in ceramic processing.

4.4. Plasticizers Solvents called plasticizers lower the glass transition temperature of polymers. Such plasticizers have low vapor pressure and typically exhibit a high boiling point (above 250  C). Also, they are typically low-molecular-weight organic compounds. They function to increase chain mobility of binder molecules. With an optimal concentration

920

Handbook of Advanced Ceramics

TABLE 3 Typical Spray Granule Additive Recipes Additives

wt.% of additives

Function

Polyvinyl alcohol (PVA) 83.6

Binder

Glycerin

8

Plasticizer

Ethylene glycol

4

Plasticizer/Lubricant

Tri ammonium citrate

4

Dispersant

Ammonium stearate

0.2

Wetting agent

Octanol

0.2

Defoamer

of plasticizer, intermolecular forces between polymers weaken, and interchain distances between polymers lengthen, thereby increasing intermolecular mobility.

Compatibility, efficiency, and stability determine plasticizer choice. Most are either glycols or phthalates, e.g. polyethylene glycol or dioctylphthalate [20]. Often, binders require a plasticizer to improve flexibility and workability of the dried, green ceramic tape because glass’s brittle behavior at room temperature is not conducive to later processing steps. External plasticizing is most common. In this way, binder polymers mix with low-molecular-weight plasticizers (Mw ~ 300e400) or with oligomers (Mw ~ 3000e4000) [8]. Too much plasticizer degrades the binder’s strength, which will also cause problems later during machining. Therefore, the binder/plasticizer ratio must be carefully monitored to provide optimal properties. Furthermore, the plasticizer must be emulsible in the same solvent used to process the binder. Traditionally, two different plasticizers are added to slurry, although one suffices. This tradition might have a positive impact on drying and burnout.

TABLE 4 Common Binders Used in Ceramic Processing Binders

Soluble in water

Soluble in organic solvents

Poly(vinyl alcohol)

O

Lowemedium

Poly(acrylic acid)

O

Mediumehigh

Polyacrylamide

O

High

Poly(ethylene glycol)

O

Lowemedium

Poly(ethylene imine)

O

Lowemedium

Poly(methylacrylic acid)

O

Mediumehigh

Polyvinylpyrrolidone

O

Low

Methylcellulose

O

Lowehigh

O

Hydroxypropylmethylcellulose

O

Lowehigh

O

Hydroxyethylcellulose

O

Lowehigh

O

Sodium carboxymethylcellulose O

Lowehigh

O

O

Ethyl cellulose

Viscosity grade (effect to change the solvent viscosity)

Biodegradable

Mediumehigh

Starches

O

Lowemedium

O

Dextrins

O

Very low

O

Sodium alginate

O

High

O

Ammonium alginate

O

Very high

O

Poly(vinyl butyral)

O

Mediumehigh

Poly(vinyl formol)

O

Mediumehigh

Poly(methyl methacrylate)

O

Mediumehigh

Gum Arabic

O

Very low

O

Guar gum

O

Very high

O

Chapter | 11.1.3

Colloidal Processing Fundamentals

4.5. Lubricants Lubricants reduce friction between particles or between particles and the die wall in extrusion and injectionmolding processes. Stearic acid, stearates, and various waxy substances are commonly used [6,12,20,29e31].

4.6. Other Additives Other additives can be used to enhance a final product. Wetting agents, or homogenizers, increase emulsibility of many liquids in slurry. They also contribute to highquality tape surfaces. Cyclohexanone is a common wetting agent. Defoaming agents, such as dimethylsilicones, prevent foaming and allow trapped air bubbles to escape easier on degassing. They commonly have to be used in aqueous slurries or with submicron powders. Defoaming agents work to remove surplus surfactants from bubble surfaces, increase bubble interfacial energies, and make bubbles unstable. However, having several organic additives makes predicting a defoaming agent’s effectiveness complex. Table 5 shows examples

921

of organic additive combinations used in industrial pressing powders.

5. FORMING TECHNIQUES Wet powder processes disperse ceramic powders in solvent to form a stable, high solids loaded slurry for consolidation. Drained casting, tape casting, and constant volume techniques all consolidate colloids into designed, functional shapes. Each technique differs from the other primarily by the type of mold. In all cases, after a wet green body is formed, it is demolded and dried to form a dry green body. Drained casting techniques use porous molds (e.g. slip casting) to consolidate the slip, tape-casting techniques do not use a mold, and constant volume techniques use nonporous molds.

5.1. Drained Casting Techniques Drained casting techniques rely on phenomena that separate a ceramic colloid’s liquid and solid components

TABLE 5 Examples of Organic Additive Combinations Used in Industrial Powder Pressing Product

Dispersing medium Binder

Plasticizer

Others

Alumina

Acetone

Polyvinyl alcohol (low viscosity)

Polyethylene glycol (PEG) (Mw ~400)

Mg Stearate (lubricant)

Trichloroethylene (TCE)/Ethanol

Polyvinyl butyral (PVB)

Poly (alkylene glycols)

Menhaden fish oil (dispersant)

Alumina substrates

Water

Polyethylene glycol (Mw ~20,000)

None

Talc, clay (lubricant)

Alumina spark plug insulation

Acetone

Microcrystalline wax emulsion KOH þ tannic acid

TiO2

Toluene

Ethyl cellulose Methylabietate Diethyl oxalate Triethylene glycol Dihydroabietate

Al2O3/ZrO2

TCE/Ethanol

Polyvinyl butyral

Polyethylene glycol/phthalates mixture

Mullite/SiCw

Methyl ethyl ketone (MEK)/TCE

Poly(ethylmethacrylate) Poly(methylacrylate) Poly(methacrylic acid)

Benzyl butyl phthalate (BBP)

MnZn Ferrites

MEK/TCE

Polyvinyl alcohol (low viscosity)

Polyethyleneglycol (Mw ~400)

Zn Stearate (lubricant)

Ba Titanate

MEK/Ethanol

Acrylic binder

PEG/BBP

Phosphate ester (dispersant)

Ceramic tile

Water

Clay

Water

Talc, clay (lubricant)

Hotel china

Water

Clay, polysaccharide

Water

Clay (lubricant)

Refractories

Water

Ca/Na, lignosulfonate

Water

Stearate (lubricant)

LaxSrl xMnO3

Ethanol/Toluol

PVB

PEG/octyl phthalate

Glyceryltrioleate

Wax, talc, clay (lubricant)

Glyceryl trioleate (dispersant)

922

Handbook of Advanced Ceramics

at the macroscopic level. Various techniques for liquidphase removal have been developed, and all seek to consolidate particles into a green body of desired form with sufficient mechanical strength for mold removal and handling.

5.1.1. Slip Casting In slip casting, the solvent drains through the mold’s porous walls. Capillary forces depend on pore size, a, and the surface tension at the liquidevapor interface, gLV, which cause a pressure gradient in the liquid and drive its flow: 2gLV a

Pc ¼

(18)

Often, the growth rate of the slip-cast layer perpendicular to the mold surface exhibits a parabolic decrease over time. This makes slip casting a comparably slow technique and less suitable if thick, consolidated layers are required.

5.1.2. Osmotic Consolidation Osmotic consolidation exploits the osmotic pressure of a colloid in contact with a polymer solution via a semipermeable membrane. Here, differences in chemical potentials force the solvent to flow through the membrane into the polymer solution. The resulting osmotic pressure can be defined as P ¼

ms

mP Vm

(19)

where m0 is the chemical potential of the solvent in the ceramic suspension, mP the chemical potential in the polymer solution, and Vm the molar volume of the solvent. Osmotic pressures are significantly higher than capillary pressures in slip casting.

5.1.3. Filter Pressing Alternatively, an external, mechanical pressure can be applied to the colloid, which drives the solvent through a filter membranedfilter pressing. Consolidation rates are comparably high for filter pressing because employed pressures are usually about one or two orders of magnitude more than slip casting. Liquid flux through the porous particle layer is a function of pressure gradient VP permeability D, and liquid viscosity h, according to Darcy’s law: J ¼

D VP h

(20)

5.1.4. Centrifugal Casting and Electrophoretic Deposition Macroscopic separation of colloidal solid and liquid phases can also be achieved with the help of body forces acting on dispersed particles. Analogous to particle sedimentation caused by gravitational forces, centrifugal casting consolidates particles in a rotating mold. The settling velocity of a single spherical particle in the centrifugal force field can be approximated by a modified Stokes equation for particle movement in dilute colloids: v ¼

2DrR2 u2 z 9h

(21)

where Dr is the density difference between solid and liquid, R the particle radius, u the angular velocity, and z the distance from the rotation axis. Since liquid is not forced through a porous layer, this method offers the advantage of a linear relation between cast layer growth and casting time. As in slip casting, centrifugally cast bodies may suffer from inherent density gradients. An electric field can be used in a similar manner to accelerate particles suspended in a liquid if the particles carry a surface charge. A corresponding technique used to consolidate particles into dense layers is called electrophoretic deposition. Similar to centrifugal casting, the accelerating force is balanced by viscous forces, resulting in a steady-state particle velocity: v ¼

εr ε0 zE fH h

(22)

where εr is the relative dielectrics constant, ε0 the permittivity of free space, z the particle’s zeta-potential, E the applied electric field strength, and fH the Henry constant (ranges from 1 to 1.5 depending on particle size and ionic strength).

5.1.5. Attraction and Repulsion Formulas Applied consolidation pressure must exceed yield stress of the space-filling particle network in colloids with primarily attractive interparticle forces. The relation between yield stress, sY, and solids volume fraction can often be described by an empirical power law of the type sY fFn

(23)

where n is a constant fitting parameter. Consolidation pressure must exceed osmotic pressure to achieve dense particle packing in colloids with primarily repulsive interparticle forces. Hard-sphere system osmotic pressure, P, is a function of solids volume fraction, F, namely   RTF 1 þ F þ F2 F3 (24) PðFÞ ¼ VðFmax FÞ3

Chapter | 11.1.3

Colloidal Processing Fundamentals

923

where R is the gas constant, T the temperature, V the molar volume of the colloid phase, and Fmax the maximum solids loading [32].

splashing and other nonuniformities. This technique produces porous electrodes for fuel cells and multilayer ceramic capacitors [33].

5.2. Tape-Casting Techniques

5.2.4. Paper Casting

Liquid casting techniques allow easy manipulation of ceramic shapes before sintering. Tape casting uses wet casting principles originally used for drained casting to make flat, thin ceramic tapes as small as a few microns (1 mm to 1 mm). Such parts could not otherwise be fabricated until recently with new extrusion techniques. In tape casting, a ceramic slurry deposits on a flat carrier substrate, and the solvent simply evaporates. Solvent removal rate mainly depends on temperature, convective and diffusive solvent transport at the liquidevapor interface, and internal capillary pressures in porous layers of cast tape. Tape cast products include dielectric capacitors, piezoelectric actuators, and thick and thin films. The chief market today remains electronics and radio applications.

In this process, low-ash paper is passed through ceramic slurry. Slurry sticks to the paper, passes through a dryer, and rolls onto a reel. The paper is removed during binder burnout. This process is used for honeycomb structures such as heat exchangers and catalytic converters [33].

5.2.1. Slurry Composition Although shaping, binder burnout, and sintering are important stages in determining the final properties of a tape cast item, original slurry composition influences the final product most. Slurry variables for a tape-casting run must be optimized for viscosity, particle size, and miscibility. In general, desirable viscosity is 3e5 Pas with powder particle size of 0.5e2 mm. However, these properties vary depending on the main powder, casting speed, thickness, and final application. Common additives to tape cast processes include dispersants, binders, plasticizers, etc.

5.2.2. Common Problems and Defects Variables involved in tape casting must be carefully monitored and controlled to prevent defects (cracks, nonuniformity, delamination, surface roughness, etc.) in the final product. Most problems originate with slurry composition. For instance, a common tape-casting problem is excessive sticking of cast tape to the belt. Adding more dispersant and plasticizer alleviates this problem. However, adding too much causes the cast tape to release from the belt early and curl up. Other signals include weak tape (caused by too little binder) and brittle tape (caused by too little plasticizer). Small deviations in additive composition greatly change the cast tape and final product.

5.2.3. Waterfall Technique Analogous to glazing ceramic tile, the waterfall casting technique utilizes a pump to pour a curtain of slurry onto substrate, then the slurry curtain dries. The slurry curtain, or waterfall, must remain thin and uniform to avoid

5.3. Constant Volume Techniques 5.3.1. Gel Casting Gel casting developed in Canada in the 1960s. Since then, it evolved into an attractive forming process for manufacturing near-net-shape, very large, high-quality, complex ceramic parts with defined threshold strength [34]. The process does not remove any component or shrink the powder compact in mold. Rather, dispersing medium either is solidified via polymerization or via trapping molecules or voids within the particle system. Numerous gel casting processes developed over the past 20 years to achieve solidification, starting with polymerization of monomers/oligomers or crosslinking polymers. Some use internal chemical reactions that form salt or change the system’s pH. Others use solubility, adsorption, or dispersant conformation changes of the ceramic material itself. Also, the dispersing medium can be frozen [1]. Differences in dispersing and gelation methods cause various challenges later in process. As examples: if more than 1 wt% organics are needed, the burnout may take days to weeks (depending on the size); if a chemical reaction facilitates consolidation, chemical gradients must be controlled to achieve uniform packing, bubble formation, or additive precipitation suppression. Gel casting’s foremost challenge is time needed to consolidate slurry in mold. Compared to pressing techniques, direct casting can take twice to thrice as long. However, the final products form at near-net-shape and with utmost quality. This saves substantial machining time and costs. Lastly, monomer toxicity remains problematic and hinders universal adoption. However, research into organic gelation compounds promises success [35].

5.3.2. Extrusion Extrusion as a manufacturing technique produces continuously long bodies of materials (metals, plastics, ceramics, etc.) with a common cross section by pressing material through shaped nozzles. Highly ductile metals (aluminum) are extruded cold (room temperature) to form hollow bars and tubes for low-strength applications such as window

924

frames. Other metals (copper and brass) are extruded hot to form water pipes and pipe fittings. Thermoplastics become viscous with heat and are extruded in many forms. Ceramic materials are extruded to form square-section bars, which can be cut into bricks; some (viscous alumina strips) are extruded into complex, thin-walled honeycombs and used as support media for catalysts [6,12]. In a conventional extrusion process for ceramic green bodies, ceramic paste is prepared by first adding polymers and other additives to ceramic powder. Then, the paste is heated, and a rotating screw (auger or piston) pushes it through a shaped die (nozzle) to form a green ceramic string, which takes the cross-sectional shape of the extrusion die. It can be no smaller than the die’s orifice. Next, air or water cools the profiled string, and it gets cut into desired lengths. All sorts of profiles can be made, and production volumes are normally high. Water (traditional ceramics) or organic materials (advanced ceramics) provide adequate plasticity for forming. The difficulty with extrusion is removing the organic material prior to firing without causing cracks or distortions. During extrusion, the augers’ (or pistons’) driving force must exceed the paste’s resistive force and the die’s wall friction. Pressure is highest in the barrel and decreases along the extruder’s axis. Specific flow behavior depends on the die’s geometry and the material’s viscosity. Co-extrusion in microfabrication (MFCX), a new method to produce axis-symmetric ceramic objects with micrometer-size features in two dimensions, can produce several shapes in alumina and PbO-containing ferric ceramics with carbon black as fugitive (material removed later in process). MFCX processes use simple round or square dies for size reduction. Shaping is done by using a shaped object of two or more materials as the extrudate. The primary material is the desired ceramic compound. The second material is a fugitive substance, which fills space and has similar viscosity as the primary ceramic. The two are co-extruded, cross-sectional dimensions are reduced, and the fugitive material is later removed. Composites with layers as fine as 8 mm have been made with this process [36e38].

5.3.3. Injection Molding Injection-molding processes produce ceramic parts with complex shapes that other forming techniques cannot. Essentially, feed material consists of a ceramic colloid with binder, dispersant, and plasticizer additives. It is fed into a heated chamber to soften or liquefy the binder (necessary for optimum mold filling). A plunger or screw forces the feed material through a nozzle into the mold cavity and fills it. Next, the material cools until the binder solidifies. Finally, the mold opens, and the ceramic part is removed.

Handbook of Advanced Ceramics

Suitable rheological properties of the feed material are required during mixing and injection. This means that injected material must have sufficiently low viscosity at relevant shear rates and temperatures. Accordingly, not only must the binder viscosity be low, but solids fraction of the feedstock also needs to be well dispersed in the binder matrix phase. Furthermore, viscoelastic properties of the feed material have to be considered; e.g. time and temperature-dependent relaxation of shear stresses after mold filling might lead to unwanted dimensional changes. Binder properties are of considerable importance in subsequent processing stages; e.g. burnout behavior of the binder can pose limitations on the geometry and the dimensions of the ceramic part [39]. Although it is comparably easy to switch production to a different product simply by exchanging the mold and feedstock, limiting factors of injection molding include cost competitiveness and difficulty producing large parts. An injection-molding machine’s design must account for high pressures. Moreover, precise control of injected material’s mass and temperature as well as position of the plunger/screw are essential for manufacturing high-quality parts. Finally, surfaces that contact the highly abrasive feed material wear, thus increasing costs [6].

5.3.4. Pressing Dry powder processing involves direct compaction of either raw fine particles or granular powders in a rigid die. Common methods include die pressing, cold isostatic pressing (CIP), hot pressing (HP), and hot isostatic pressing (HIP). HP and HIP enable the green body to be formed and sintered simultaneously, but they require special apparatus. Die pressing produces simple-shaped parts and can easily be automated. This process applies vertical stresses directly to ceramic powder granules in mold to generate loose bonding between particles. This gives the compact some mechanical strength for handling. During die pressing, the ceramic powder granules’ shapes change over three steps: (a) packing of granules, (b) deformation of granules, and (c) sliding between granules to decrease void size between deformed granules. When stress is applied from one direction, e.g. the upper side of the mold, this is referred to as uniaxial pressing. The process automates fast and easy, quite an advantage, but sundry disadvantages also exist. Stress gradients result in density gradients in the compact green body. Sometimes, lamination may also occur because of inhomogeneous distribution of applied stress. Biaxial pressing, applied stress from above and below, minimizes the above problems. However, for very large parts, density gradients still exist, especially in areas along the mold’s walls due to friction. CIP utilizes liquid media hydrodynamic forces to apply stress to compacted ceramic powder in all

Chapter | 11.1.3

Colloidal Processing Fundamentals

directions. Therefore, it forms parts with an elongated dimension, relatively complex parts, or parts that may not be easily die-pressed. CIP processes are either wet-bag or dry-bag. In wet-bag processes, powder is filled into a mold, then stress is applied omnidirectionally. Wet-bag presses usually have operating pressures up to 200 MPa, and wet-bag presses operating up to 500 MPa are used for pressing large blocks. Such systems require that the ceramic-powder-filled rubber mold be immersed in a liquid medium. Then, force applied to the liquid medium passes on to the rubber mold and is equal in all directions. This method produces compact ceramic green bodies with fewer density gradients. However, it encounters problems when the part is too complicated and/or the surface finish is not good. In dry-bag processes, forces apply radially through a liquid medium between a flexible mold and a rigid shell.

925

[6] [7]

[8] [9] [10] [11] [12] [13] [14]

6. SUMMARY AND OUTLOOK

[15]

The push toward nanotechnology in recent years changed the landscape of ceramic processing. Designed nanostructures’ fabrication methods were the focal point of research, and less attention was afforded to nanopowder processing. Still, tremendous challenges exist for processing nanoparticles that need to be addressed. With nanotechnology’s transition from research to industry and production, a new wave of research to enhance nanoparticle processing knowledge is expected. Future efforts will require more computational approaches since slurries are so complex. Specifically, inadequacies of current measuring techniques limit our understanding of what happens when particles come within close proximity of a few nanometers. Furthermore, simplification of averages, which works for long distances, no longer applies. Thus, studying nanoparticles with proximities of a few nanometers will help achieve dense structures with acceptable shrinkage.

[16] [17]

[18]

[19]

[20]

[21] [22] [23] [24]

REFERENCES [1] Sigmund WM, Bell NS, Bergstro¨m L. Novel powder-processing methods for advanced ceramics. J Am Ceram Soc 2000; 83(7):1557e74. [2] Lee S, Sigmund WM. AFM study of repulsive van der Waals forces between teflon AF thin film and silica or alumina. Colloid Surface Physicochem Eng Aspect 2002;204, 1e3(23):43e50. [3] Bergstro¨m L. Hamaker constants of inorganic materials. Adv Colloid Interface Sci 1997;70:125. [4] French RH. Origins and applications of London dispersion forces and Hamaker constants in ceramics. J Am Ceram Soc 2000; 83(9):2117e46. [5] Sigmund WM, Pyrgiotakis G, Daga A. Theory and applications of colloidal processing. In: Lee B, Komarneni S, editors.

[25] [26] [27] [28] [29]

[30] [31]

Chemical processing of ceramics. 2nd ed. Boca Raton, FL: CRC Press; 2005. Rahaman MN, NetLibrary Inc. Ceramic processing and sintering. New York: Marcel Dekker, Inc; 1995. Hunter RJ, White LR. Foundations of colloid science. Oxford, Oxfordshire; New York: Clarendon Press; Oxford University Press; 1987. Israelachvili JN. Intermolecular and surface forces. London: Academic Press; 1985. Rahaman MN. Ceramic processing. Boca Raton, FL: CRC Press; 2006. Napper DH, Hunter RJ. Hydrosols in M.T.P. International Review of Science. Physical Chemistry Series I 1972;7:241. Smoluchowski von M. Z Phys 1916;17:557, 585. Reed JS. Principles of ceramics processing. 2nd ed. , New York: Wiley; 1995. Krieger IM, Dougherty TJ. A mechanism for non-newtonian flow in suspensions of rigid spheres. Trans Soc Rheol 1959;3:137e52. Krieger IM. Rheology of monodisperse lattices. Adv Colloid Interface Sci 1972;3:111e36. Conley JRF. Practical dispersion: a guide to understanding and formulating slurries. New York: VCH; 1996. Dobia´a B, Qui X, Rybinski WV. Solideliquid dispersions. In: Surfactant science series, vol. 81. New York: Marcel Dekker; 1999. Naito M, Shinohara N, Uematsu K. Raw materials. In: Somiya S, et al., editors. Handbook of advanced ceramics. Kidlington, Oxford: Elsevier Academic Press; 2003:81e129. Cesarano III J, Aksay IA. Processing of highly concentrated aqueous a-alumina suspensions stabilized with polyelectrolytes. J Am Ceram Soc 1988;71:1062e7. Cesarono III J, Aksay IA, Bleir A. Stability of aqueous alphaAl2O3 suspensions with poly (methacrylic acid) poly-electrolyte. J Am Ceram Soc 1988;71:250e5. Bo¨hnlein-Maub J, Sigmund W, Wegner W, Meyer WH, Hebel F, Seitz K, et al. The function of polymers in the tape casting of alumina. Adv Mater 1992;4:73e81. Mistler RE. Tape casting: past, present, future. Am Ceram Soc Bull 1998:77. Myers D. Surfaces, interfaces, and colloidsdprinciples and applications. New York: VCH; 1989:286e296. Lange FF. Powder processing science and technology for increasing reliability. J Am Ceram Soc 1989;72:3e15. Horn RG. Surface forces and their action in ceramic materials. J Am Ceram Soc 1990;73:1117e35. Lewis JA. Colloidal processing of ceramics. J Am Ceram Soc 2000;83:2341e59. Baes CF, Mesmer R. The hydrolysis of cations. Malabar, Florida: Krieger Publishing Comp; 1986. Onoda Jr GY, Hench LL. Ceramic processing before firing. New York: John Wiley and Sons; 1978. Roosen A. Ceram Trans I; Part B., Ceramic Powder Science 1988:675. Mistler RE, Runk R, Shanefield D. Tapecasting of ceramics. In: Onoda G, Hench L, editors. Ceramic processing before firing. New York: Wiley; 1978:441e8. Paul CW, Cotts PM. Macromolecules 1986;19:692e9. Napper DH. Polymeric stabilization of colloidal dispersions. London: Academic press; 1983.

926

[32] Guo J, Lewis JA. Aggregation effects on compressive flow properties and drying behavior of colloidal silica suspensions. J Am Ceram Soc 1999;82:2345e58. [33] Richerson DW. Modern ceramic engineering. New York: Marcel Dekker; 1992. [34] Si W, Graule TJ, Baader FH, Gauckler LJ. Direct coagulation casting of silicon carbide components. J Am Ceram Soc 1999;82(5):1129e36. [35] Vandeperre LJ, De Wilde AM, Luyten J. Gelatin gelcasting of ceramic components. J Mater Process Tech 2003;135:312e6.

Handbook of Advanced Ceramics

[36] Greenwood R, Kendall K, Bellon O. A method for making alumina fibres by co-extrusion of an alumina and starch paste. J Eur Ceram Soc 2001;21:507e13. [37] Hoy CV, Barda A, Griffith M, Halloran JW. Microfabrication of ceramics by co-extrusion. J Am Ceram Soc 1998;81: 152e8. [38] Wright Jr JF, Reed JS. Polymer-plasticized ceramic extrusion. Part 1. Am Ceram Soc Bull 2001;80:31e5. [39] Mutsuddy BC, Ford RC. Ceramic injection molding: materials technology series. London: Chapman & Hall; 1995.

Chapter 11.1.4

Solvothermal Synthesis of Metal Oxides Masashi Inoue Professor Emeritus of Kyoto University, Japan

Chapter Outline 1. Introduction 2. Safety Consideration 3. Solvothermal Synthesis of Metal Oxides e Case Studies

927 928 930

1. INTRODUCTION Syntheses of metal oxides are usually performed by calcination of suitable precursors, such as hydroxides, nitrates, carbonates, carboxylates, etc. Since this process requires large thermal energies, sintering of the product particles takes place and aggregated particles are usually obtained. Liquid-phase syntheses allow the formation of metal oxides at relatively lower temperatures, and products with welldefined morphologies can be obtained. The solegel method (or alkoxide method) is one of these methods [1,2], which enables the formation of metal oxides at low temperatures. However, this method usually gives amorphous products, and calcination of the products is required to obtain crystallized products. This chapter covers solvothermal methods. Various compounds have been prepared by solvothermal reactions: metals [3,4], metal oxides [5e11], halides [7], chalcogenides [12,13], nitrides [12,14,15], phosphides [16], openframework structures [17,18], oxometalate clusters [19,20], organiceinorganic hybrid materials [18,21,22], metal organic frameworks (MOF) [23], and carbon nanotubes [24,25]. Most solvothermally synthesized metal oxides are composed of nano- or microparticles with well-defined morphologies. The distribution of the particle size of the product is usually quite narrow, and the formation of monodispersed particles is frequently reported [26,27]. When the solvent molecules or additives are preferentially adsorbed on (or have a specific interaction with) a certain surface of the products, growth of the surface is prohibited and therefore products with unique morphologies may be formed by the solvothermal reaction [12,28,29]. Thus, nanorods [29], nanowires [30], tubes [31], and sheets [32]

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00050-2 Copyright Ó 2013 Elsevier Inc. All rights reserved.

4. Concluding Remarks References

942 943

of various types of products have been obtained solvothermally. Since 1984, we have been exploring the synthesis of inorganic materials in organic media at elevated temperatures (200e300  C) under autogenous pressure of the organics using an apparatus shown in Figure 1 [33]. This technique is now generally called the “solvothermal” method [14]. The term “solvothermal” nominally means reactions in liquid or supercritical media at temperatures

FIGURE 1 Geometry of our apparatus for solvothermal reaction. a) For the hydrolysis of alkoxide by water dissolved from the gas phase (Section 3.2.5), water was added in the gap between the autoclave wall and test tube.

927

928

Handbook of Advanced Ceramics

higher than the boiling point of the medium. Some researches define “solvothermal reaction” as a reaction that is carried out in a closed vessel. Reactions in a closed system using autoclaves or sealed ampoules and in an open system using a flask equipped with a reflux condenser sometimes give completely different results, especially when low-boiling-point products such as water are formed as by-products and affect the reaction. However, most researchers use the word “solvothermal” without showing evidence that the reaction behavior in the closed system is different from that in the open system. The density of the liquid solvent is essentially unchanged at above or below the boiling point because the compressibility of the liquid is quite small. The effect of pressure on the reaction kinetics is determined by the activation volume, relative volume of the activated complex at the transition state to the volume of the starting molecule(s). However, in order to measure the effect of reaction pressure, the GPa-scale pressure is required. This means that the autogenous pressure created by the vapor pressure of the solvent has only a minor effect on the reaction rate. Therefore, there is no need to differentiate the reactions at the temperature above and below the boiling point. We previously proposed a broad definition of “solvothermal” reaction as a reaction in a liquid (or supercritical) medium at high temperatures [34]. However, this definition is so broad that flux methods are also included in this category, and a review paper discussed the flux method as a solvothermal method [35]. However, flux rather reflects the nature of melt, not solution, and, therefore, in this article, solvothermal reaction is defined as a solution-based reaction at high temperatures. The prefix “solvo-” means any kind of solvent; various solvents such as water [36e40] (Tc, 374; Pc, 218 atm), ammonia [41e43] (bp, 78  C; Tc, 132  C, Pc, 113 atm), sulfur dioxide (bp, 10  C, Tc, 157.5  C, Pc, 78 atm), hydrofluoric acid [44] (bp, 19.5  C; Tc, 188  C, Pc, 188 atm), and many organic compounds have been used for the solvothermal reaction. Hydrothermal reactions are one type of solvothermal reaction; in this chapter, however, “solvothermal” is used to refer to reactions in organic solvents. When alcohols and glycols are used as the reaction media, the reactions are called “alcohothermal” and “glycothermal” reactions, respectively. In 2004, Cooper et al. [45] reported a new type of solvothermal synthesis in which an ionic liquid or eutectic mixture (they used the mixture of choline chloride and urea

CH3 H3C

N

(CH2)2OH

+ Cl¯

CH3 Choline chloride; mp = 300 oC

in a 1:2 ratio, which has a melting point of 12  C and exhibits unusual solvent properties that are very similar to those exhibited by the ionic liquid [46]) was used as both the solvent and the structure directing agent in the synthesis of zeolites. This methodology was termed ionothermal synthesis and has received growing interest for synthesizing open-framework structures such as zeolites, aluminophosphates, and organiceinorganic hybrid materials, many of which have novel structures [47,48].

2. SAFETY CONSIDERATION 2.1. High-Pressure Experiments To carry out reactions at temperatures higher than the boiling point of the reaction medium, pressure vessels (autoclaves) are usually required. Some researchers favor the use of sealed ampoules of glass or silica, but these experiments should be carried out with great care because the ampoules are easily broken by the internal pressure of the reaction medium. To avoid explosion of the ampoules, they may be placed in an autoclave together with a suitable medium to create a vapor pressure to balance the internal pressure of the ampoule. Thermal expansion of liquid is an important factor that should be taken into account to carry out experiments safely. The volume of liquid increases with the increase in temperature, and thermal expansion is drastic at near the critical point. When the volume of the liquid reaches the volume of the vessel, a further increase in the reaction temperature resulted in a drastic increase in the pressure of liquid, because the compressibility of liquid is quite small. Therefore, the percentage of fill of the reaction medium in the autoclave is an important factor. For the hydrothermal system, the Kennedy diagram [49] can be used for estimation of the reaction pressure. For organic solvents, however, such diagrams are not available. Therefore, one should estimate the density of the reaction medium at the desired reaction temperature by using a suitable method (for example, the Hanson diagram [50]). From the density of the liquid, one can calculate the maximum volume that can be charged to the reaction vessel. However, the volume of the medium should be smaller than the maximum volume, so that a reaction does not create unexpected high pressure.

2.2. Organic Medium Various organic solvents have been applied for the syntheses of inorganic materials [34]. As ceramists may not be familiar with organic solvents, some important features are summarized here. From a thermodynamic point of view, most organic compounds have an inherent tendency to decompose into carbon, hydrogen, nitrogen, and so forth, at high temperatures,

Chapter | 11.1.4

Solvothermal Synthesis of Metal Oxides

although the reaction is endothermic. Methane is the most stable aliphatic hydrocarbon; however, it can decompose into carbon and hydrogen at temperatures >360  C in the presence of a suitable catalyst such as nickel or iron [51]. Therefore, most of the organic compounds, even though they may be stable at room temperature, act as reducing agents at high temperatures, Therefore, the presence of any oxidants in the reaction system of the solvothermal reaction using organic solvents cannot be recommended. The gas phase in the reaction vessel must be completely purged with an inert gas such as nitrogen. Since organic compounds easily react with oxygen with highly exothermic nature (i.e., combustion of organics), the reaction continues until completion. Note that even though one could conduct an experiment safely, this fact does not guarantee the safety of another experiment. When an organic molecule contains oxygen atoms in the structure, it may spontaneously decompose into carbon dioxide, water, nitrogen, and so forth, and the reaction becomes exothermic. To assess the intrinsic nature for the exothermic decomposition of organics, oxygen balance may be used. Oxygen balance (OB) is defined as the weight of oxygen that is excess or deficient for the complete decomposition of 100 g of a compound and is given by the following equation: z=2ÞO2 / xCO2 Cx Hy Oz Nn þ ðx þ y=4 þ y=2 H2 O þ n=2N2 OB ¼

1600ð2x þ y=2

zÞ=Mw

(1) (2)

where Mw is the molecular weight of the compound. A compound having an OB value near zero has a high tendency to explode. One should also consider the oxygen balance of the reaction system. For the reaction of a nitrate of a divalent metal ion with ethylene glycol, for example (Eqn 3), oxygen balance is zero, indicating that this reaction system can explode: MðNO3 Þ2 þ HOCH2 CH2 OH / MO þ 2CO2 þ 3H2 O þ N2 OB ¼ 0%

(3)

For the choice of the solvent, a less toxic solvent should be selected, for example, toluene instead of benzene. Benzene is highly toxic and it causes fatal damage to hematopoietic organs. n-Hexane is one of the most common aliphatic solvents; however, this compound has a specific toxicity that is not shown by branched C6 hydrocarbons or by other straight-chain paraffins such as pentane and heptane. The origin of this toxicity is well established. Through the metabolic system of human beings, this compound is oxidized by oxygenase to 2,5hexanedione (Eqn 5), which attacks the peripheral nervous system [52]. Therefore, heptane should be used in place of hexane:

O

Cytochrome P-450

O

H3C

CH3 (5)

Thermal stabilities of the solvents in terms of decomposition and polymerization should also be taken into account. Spontaneous decomposition of organic compounds was discussed earlier. Cyclic ethers such as tetrahydrofuran (THF) may cause ring-opening polymerization which has a severe exothermic nature. Dimethyl sulfoxide (DMSO) is an aprotic polar solvent and is widely used in organic chemistry. However, the thermal stability of DMSO is slightly lower than that of other aprotic polar solvents such as N,N-dimethylformamide (DMF) and hexamethylphosphoramide (HMPA). Moreover, this compound may cause an explosive reaction resulting in the formation of dimethyl sulfide when it contacts compounds such as acid halides.

O CH3SCH3

H3C N-CHO H3C

Dimethyl sulfoxide (DMSO); bp = 189 oC

Dimethylformamide (DMF); bp = 153 oC O

The reaction shown in Eqn (3) may take place via the following equation: MðNO3 Þ2 þ HOCH2 CH2 OH / MðOHÞ2 þ O2 NOCH2 CH2 ONO2

929

(4)

The last compound of the equation is a nitrate ester of ethylene glycol and is known as nitroglycol, a highly powerful explosive that was used as an ingredient of dynamite. Even though excess ethylene glycol is used, where the OB value of the whole reaction system is highly negative, the reaction system should be carefully treated as a mixture of nitroglycol and ethylene glycol.

H3C

N

H3C

P

N

CH3

N H3C

CH3

CH3

Hexamethylphosphoamide (HPMA); bp = 233 oC Before performing solvothermal reactions in any type of solvent, ceramists or inorganic chemists who are not familiar with the organic compounds are recommended to consult with organic chemists and refer to the material safety data sheets available from the supplier of the chemicals.

930

Handbook of Advanced Ceramics

2.3. Surface Acidity of the Products Some mixed oxides such as amorphous silicaealumina have strong surface acidity. Since the particles of solvothermal products are usually covered with organic moieties, the surface acidity does not work well. However, when the reaction system contains a small amount of water, it may hydrolyze the surface organic moieties, resulting in exposure of an acidic surface. An acidic surface possibly decomposes an organic solvent such as alcohol resulting in the formation of water. The thusformed water further hydrolyzes surface organic moieties, resulting in facilitating the decomposition of the organic solvent. Therefore, the reaction pressure may suddenly begin to increase. This event has occurred, and, therefore, reaction pressure should be monitored during the course of the solvothermal experiments. The reaction pressure is usually determined by the vapor pressure of the solvent. However, under the solvothermal conditions, the reaction pressure may be much higher than expected from the vapor pressure of the solvent because of the decomposition of the solvent.

3. SOLVOTHERMAL SYNTHESIS OF METAL OXIDES e CASE STUDIES 3.1. Glycothermal Synthesis of a-Alumina 3.1.1. Reaction of Microcrystalline Gibbsite In 1989, Inoue et al. [5] briefly reported that the reaction of microcrystalline gibbsite (a modification of Al(OH)3) in 1,4-butanediol or 1,6-hexanediol at ~300  C for 2 h (glycothermal reaction) yields a-alumina through the glycol derivative of boehmite as an intermediate (Figure 2) [21,53,54]. This intermediate is a class of intercalation compound [55], in which some of the hydroxyl groups in the layers of boehmite (a modification of aluminum

FIGURE 3 Morphologies of a-alumina particles obtained by the glycothermal method.

hydroxide oxide, AlOOH) are substituted with u-hydroxylalkoxyl groups [21,33]. In other words, the alkyl groups are bound to the boehmite layers through covalent bonding. The particles of starting gibbsite are hexagonal palates, while the intermediate phase has a characteristic thin wrinkled sheet morphology. The produced a-alumina particles are hexagonal plates with a relatively large surface area (Figure 3). This indicates that the dissolutionerecrystallization mechanism takes place.

3.1.2. Thermodynamical Consideration Hydrothermal reactions of aluminum compounds have been well studied, and this reaction easily yields boehmite at relatively low temperatures (~200  C) [56]. The equilibrium point between diaspore (another polymorph of AlOOH) and a-alumina þ H2O under the autogenous vapor pressure of water was reported to be 360  C [57]. However, near the equilibrium point, the transformation rate would be very sluggish, and only a small conversion of diaspore would be attained. Therefore, a much higher temperature is required to achieve complete conversion of diaspore into a-alumina. Since boehmite is slightly less stable than diaspore (i.e., boehmite is a metastable phase), the hypothetical equilibrium point between boehmite and a-alumina under the hydrothermal conditions would be slightly lower than that for the diasporeea-alumina system. However, a-alumina

FIGURE 2 Reaction of microcrystalline gibbsite in 1,4-butanediol at 285  C.

Chapter | 11.1.4

Solvothermal Synthesis of Metal Oxides

would not be formed by a hydrothermal reaction at such a low temperature. Actually, Yanagida et al. reported that more than 10 h are required for the complete conversion of boehmite into a-alumina, even with a reaction at 445  C in a 0.1 N NaOH solution and in the presence of seed crystals [58]. On the other hand, under glycothermal conditions, complete conversion of gibbsite into a-alumina is attained by the reaction at 285  C for 4 h. To the best of our knowledge, this reaction is the first example of crystallization of a thermodynamically stable metal oxide in an organic solvent at a temperature lower than that required by the hydrothermal reaction. In other words, this example showed that dense metal oxides can crystallize in organic solvents at temperatures lower than those required for crystallization in water. The difference between these two reactions may be attributed to the activities of water present in the reaction systems, since the overall reaction is dehydration of aluminum hydroxide. However, intentional addition of a small amount of water caused the enhancement of a-alumina formation rather than the retardation expected from the equilibrium point of view [59]. Therefore, facile formation of a-alumina under glycothermal conditions is attributed to the thermodynamical instability of the intermediate, the glycol derivative of boehmite, as compared with the well-crystallized boehmite formed under hydrothermal conditions (Figure 4). The glycol derivative of boehmite has AleOeC bonds and therefore is more unstable with respect to boehmite itself. Thus, conversion of the former compound into a-alumina has a much larger driving force. These arguments lead to an interesting prediction that if the formation of well-crystallized boehmite from gibbsite is somehow prevented under the hydrothermal conditions, the formation of a-alumina would take place at much lower temperatures. The glycothermal reaction of gibbsite is strongly affected by the particle size of the starting material [59]. When gibbsite with a particle size less than 0.2 mm is used, complete conversion into a-alumina is attained. When coarse gibbsite is used for the reaction, a-alumina was not formed at all but the product is c-alumina together with wellcrystallized boehmite [59]. The latter compound is partially formed by thermal dehydration of coarse gibbsite via

931

intraparticle hydrothermal reaction [60,61]. Since the reaction temperature is close to the temperature for thermal dehydration of gibbsite, the formation of the intermediate, the glycol derivative of boehmite via dissolution of the starting material competes with the thermal dehydration of gibbsite yielding c-alumina and well-crystallized boehmite. Therefore, the small particle size of gibbsite, which facilitates the dissolution rate, is essential for the formation of a-alumina. When thermal dehydration of gibbsite into c-alumina takes place, the thus-formed c-alumina facilitates epitaxial decomposition of the glycol derivative of boehmite into the transition alumina. Therefore, coarser gibbsite as the starting material of the glycothermal reaction does not give a-alumina, even at higher reaction temperatures.

3.1.3. Occlusion of Foreign Ions When precipitates are formed from alkaline solutions, a significant amount of alkali cations is incorporated into the precipitates. Since commercial gibbsite is produced by crystallization from the sodium aluminate solution, the gibbsite samples are always contaminated with sodium ions. This can be easily recognized by the fact that the mother liquor after the hydrothermal conversion of gibbsite yielding boehmite is highly alkaline. On the other hand, the mother glycol solution after the glycothermal conversion of gibbsite into a-alumina is not alkaline. This result indicates that a significant amount of sodium ions is occluded in the glycothermal a-alumina. Some researchers claimed that one of the specificities of the solvothermal reaction is the formation of contaminationfree products. However, this is not a general characteristic of the solvothermal reactions but rather specific to the hydrothermal reaction. Solvothermal products obtained in organic solvents easily occlude various foreign ions because the solubilities of these ions in organic solvents are rather low. This is a drawback of the solvothermal reactions but is a merit when preparing composite oxides. When one of the component ions has high solubility in water, these ions are easily expelled from the products under hydrothermal conditions. On the other hand, the solvothermal method gives a convenient route for preparing composite oxides that are difficult to prepare by the hydrothermal method.

3.1.4. Effects of Seeds

FIGURE 4 Comparison between glycothermal and hydrothermal conversions of gibbsite into a-alumina.

We reported that the reaction of aluminum alkoxide in glycols also yields the glycol derivatives of boehmite [54,55]. The products have the same morphology as that of the products obtained from gibbsite. However, the reaction of aluminum alkoxides in glycols does not yield a-alumina even at elevated reaction temperatures. To account for these results, two other factors must be considered [59]. First, water formed as a by-product in the formation of the

932

glycol derivative of boehmite facilitates crystallization of a-alumina by increasing the dissolution rate. Intentional addition of a small amount of water actually lowers the aalumina crystallization temperature; however, excess water causes the formation of boehmite, which is fairly stable under the reaction conditions and contaminates the final product. Another point is the presence of a-alumina nuclei in the starting gibbsite sample. The starting gibbsite is commercially produced by milling coarser gibbsite. Prolonged grinding or milling of gibbsite particles causes mechanochemical transformation into a-alumina [62]. Therefore, a-alumina-like domains are formed at this stage and act as the nuclei of a-alumina in the glycothermal reaction. Kaiser et al. [63] and Bell et al. [64] also addressed the importance of seed particles on the formation of a-alumina and on the size of glycothermally prepared hexagonal a-alumina platelets. Although the glycol derivative of boehmite prepared from aluminum alkoxides cannot transform into a-alumina because of the absence of a-alumina nuclei and water, the reaction of a mixture of gibbsite (microcrystalline) and aluminum alkoxide in the presence of a small amount of water gives a-alumina with larger crystallite sizes than those obtained from gibbsite alone [59]. Kaiser et al. [63] reported that the reaction of a 0.6-mmsized gibbsite in 1,5-pentanediol did not yield a-alumina but that tohdite (5Al2O3$H2O) was obtained. This is presumably because the crystal size of the starting material is insufficiently small for glycothermal synthesis of a-alumina. They also noted that the presence of a-alumina seed crystals in the reaction system yields a-alumina. Cho et al. [65] reported that glycothermal treatment of gibbsite prepared by precipitation of aluminum nitrate with potassium hydroxide yields a-alumina, even though the particle size of the gibbsite is much larger than 0.2 mm. However, their starting material seems to be contaminated with potassium ions, which seem to play an important role in the nucleation of a-alumina. They noted that the produced a-alumina contained 0.09 wt% potassium.

3.1.5. Morphology of Glycothermal a-Alumina Bell and Adair [66] examined the effects of various additives on the morphology of the glycothermally obtained a-alumina. They concluded that alcohols specifically adsorb on f01112g habit planes (see Figure 3) to promote a platelet morphology, while the addition of glacial acetic acid causes specific adsorption-stabilizing f0112g habit planes resulting in the formation of acicular morphology. They also reported that hydroxide ions and pyridine also affect the morphology of the a-alumina particles because of their preferential adsorption of a specific surface. Adair et al. [67] also reported the effects of process conditions such as the stirring rate and seed loading on the morphology of a-alumina formed by the glycothermal

Handbook of Advanced Ceramics

reaction. Other researchers also discussed the morphology of glycothermally prepared a-alumina [68e70]. Li and Choi [71] discussed the effect of adsorption of 1,4-butanediol on the morphology of the a-alumina particles on the basis of molecular dynamics calculation. However, these discussions neglect the effect of alkali ions remaining in the a-alumina and further study is required to completely understand the morphology of a-alumina formed under glycothermal conditions.

3.1.6. GaseLiquid Equilibrium It was reported that the addition of THF (a by-product formed from 1,4-butanediol; vide infra) and nitric acid had essentially no effect on the morphology of the product [66]. In the solvothermal reaction, the gaseliquid equilibrium should be considered as well as the adsorption of solvent and additive molecules on the surface of the starting material or the product. Since THF has a much lower boiling point than 1,4-butanediol, most of the THF molecules are present in the gas phase, thus having negligible effect on the morphology of the product. As for the effect of nitric acid, the authors attributed the result to the nonadsorbing electrolyte property of nitrate ions. However, nitrato complex of various metal cations is known and, therefore, this explanation is questionable. Nitric acid is a strong oxidant, and, therefore, is easily reduced by 1,4-butanediol, resulting in the formation of NO and N2O. These by-products easily escape to the gas phase. The nitric acid oxidation of organic compounds is a severely exothermic reaction and it is not recommended to carry out such an experiment in a closed vessel, although the authors carried out the reaction with only a small amount of nitric acid.

3.2. Titania 3.2.1. Thermal Decomposition of Titanium Alkoxide in an Inert Organic Solvent We examined the solvothermal reaction of titanium isopropoxide (TIP) and titanium oxyacetylacetonate (TiO(acac)2) in various organic solvents, including toluene, 2-butanol, and glycols, and found that some of the reactions yield anatase crystals [72]. Thermal decomposition of metal alkoxides is usually applied for synthesizing oxide films or aerosols by the chemical vapor deposition (CVD) method [73]. Whereas CVD reactions of metal alkoxides under reduced pressure usually give amorphous products, solvothermal reactions of alkoxides in inert organic solvents, such as toluene, in some cases, yield crystalline products. For example, thermal decomposition of aluminum isopropoxide, TiO(acac)2, and zirconium isopropoxide in toluene at 300  C yields c-alumina [74], anatase [72], and tetragonal zirconia [75],

Chapter | 11.1.4

Solvothermal Synthesis of Metal Oxides

respectively. This means that solvent molecules present in the reaction system facilitate the crystallization of the product. This reaction is strongly affected by the structure of the alkyl groups of the alkoxides. For aluminum and zirconium alkoxides, reactivity of the alkoxides increases according to the following order of alkyl structure: primary < secondary < tertiary. This result indicates that heterolytic cleavage of the CeO bond yielding carbocation and metaloxo anion (hMeO ; Eqn 6) is the key step, and stability of the carbocation determines the reactivity of the metal alkoxides [74]: Since the solvent molecules cannot stabilize these charged species, however, the formation of an ion pair seems to be most probable. The anion (hMeO ) attacks another molecule of the alkoxide yielding hMeOeMh bond and alkoxide anion (RO ). On the other hand, the carbocation gives olefin and proton via E1 reaction. It is known that dissociative reaction (E1 reaction) is favored by the high temperature rather than the associative route (SN1 reaction) required by the formation of ether from two molecules of metal alkoxides. The RO anion gives alcohol by reaction with the proton: M(OR)n

M(OR)n

O

ROH

C H

CH3(CH2)7

P

(CH2)7CH3

(CH2)7CH3

Olefin + H+ O

RO

amorphous product, which then crystallizes into anatase. Kim et al. [82] reported that the thermal decomposition of TIP in toluene is affected by the TIP/toluene ratio, and found that anatase crystals with 10e30 nm size can be obtained at 250  C for 3 h, when TIP/toluene ratios are in the range of 10/100e30/100. We reported that TIP was not decomposed in toluene even at 300  C, although the reaction was carried out with a low TIP/toluene ratio [72]. Since this procedure is severely affected by the presence of a trace of water, further work is desired to determine why high or low loading of TIP does not give anatase crystals. Investigation on the effect of the structure of alkyl groups on the reactivity of titanium alkoxides is also desired, because the reactivity tendency may be different from those found in aluminum and zirconium alkoxides. Parala et al. [83] synthesized anatase nanoparticles with 3e5-nm particle size by thermal decomposition of titanium tri-isopropoxidedimethylaminoethoxide (Ti(OiPr)3(OCH2 CH2NMe2)) and diethylaminoethoxide (Ti(OiPr)3(OCH2 CH2NEt2)) in trioctylphosphine oxide (TOPO) at 325  C. They reported that the product showed a broad emission

M-O-M + RO¯

M-O¯ ll R+ M

933

C

(6) Shulman et al. [76] examined the kinetics of the thermal decomposition of aluminum alkoxides and found that the activation entropy is positive, suggesting a transition state less organized than the initial state. Based on these findings, they ruled out the possibility that the reaction takes place via a cyclic mechanism (Eqn 6) and proposed a carbocation mechanism, which is consistent with our conclusion. However, this reaction is strongly affected by the metal cation of the alkoxide, and TIP was found not to decompose in toluene at 300  C [72]. The factors of metal cations controlling the decomposition of alkoxides have not yet been fully elucidated; electronegativity and the oligomeric structure of the alkoxides seem to determine the reactivity of metal alkoxides. Following the preliminary work [72], Kominami et al. [77] further examined the solvothermal decomposition of titanium tert-butoxide in toluene at 300  C, and found that this reaction yields nanocrystalline anatase. Note that the lowest temperature required for the formation of crystalline titania by CVD synthesis was reported to be 400e450  C [78e80]. Payakgul et al. [81] found that the reaction of titanium n-butoxide in toluene at 300  C initially forms an

Trioctylphosphine oxide (TOPO); mp = 50 °C; bp = 201 °C/2 mmHg Toxicological properties of this compound have not been fully investigated centered at 370 nm and that large spherical submicron particles are obtained by slow evaporation of a dispersion of titania nanocrystals in toluene [83].

3.2.2. Hydrolysis of Titanium Alkoxides by Water, Homogeneously Formed in Organic Media Hydrolysis of metal alkoxide with water formed by the reaction of the solvent is a variation of the solegel synthesis of metal oxide. Water formed homogeneously in organic media under the solvothermal conditions hydrolyzes metal alkoxides and facilitates solvothermal crystallization of metal oxides. This method can be regarded as the solegel version of the homogeneous precipitation method, where precipitants are formed homogeneously in the reaction system avoiding the heterogeneity caused by the addition of precipitants to the reaction system. To the best of our knowledge, this strategy was first applied by Fanelli and Burlew [84], who examined the reaction of aluminum sec-butoxide in 2-butanol at 250  C and reported the formation of alumina with quite a large surface area. In this

934

Handbook of Advanced Ceramics

reaction, 2-butanol decomposes yielding butenes and water. Before the work of Fanelli and Burlew [84], Takahashi et al. [85] noted that the presence of 2-propanol in the CVD reaction facilitates the decomposition of TIP because of the formation of water from alcohol. Following the work of Fanelli and Burlew [84], Inoue et al. examined the reaction of TIP in 2-butanol at 250  C and found that anatase with 22 nm crystallite size was formed [72]. Coutecuisse et al. [86] reported a flow reactor for the formation of titania by the decomposition of TIP in 2-propanol at 260e300  C. They reported that the ratedetermining step is the thermal dehydration of titanium hydroxide formed by the hydrolysis of the precursor alkoxide [87]. They also found that an increase in the supercritical fluid density decreased the overall reaction rate, but this was not adequately explained. Kominami et al. [88] examined the reaction of various titanium alkoxides in secondary alcohols and concluded that the reaction takes place by dehydration of the alcohols followed by hydrolysis of the alkoxides. They also found that the products had high activities for photocatalytic mineralization of acetic acid. They explained the results by the high crystallinity and high surface area of the products [88]. Besides dehydration of alcohols, esterification can be used for this strategy. Thus, Invanda et al. [89] synthesized titania (anatase þ amorphous phase) by the reaction of TIP in mixtures of carboxylic acid and alcohol. Water formed by esterification of the solvents hydrolyzed the alkoxide yielding metal oxide. Cozzoli et al. obtained nearly spherical nanocrystals by the reaction of TIP in oleic acid and ethylene glycol [90].

titanium-oxo cluster [Ti3O(OiPr)7(O3C9H15)] was identified by the reaction of TIP with acetone at room temperature for 24 h [96]. The moiety, O3C9H15, in the cluster is formed by the aldol condensation at both methyl groups of acetone. Garnweitner et al. [97] reported the formation of anatase by the solvothermal reaction of TIP in acetone at 130  C, and quite complex mixtures of the aldol condensation products were obtained. iPr

O O

iPrO

iPrO

Ti

OiPr

Ti

O iPr

O C

H3C

C

Ti

O

OiPr

iPr

O C

O C

C

CH3

CH3

CH3

Li et al. [98] reported a unique method: they found that anatase nanoparticles were formed by the reaction of titanium n-butoxide with linoleic acid and triethylamine in the presence of NH4HCO3, which spontaneously decomposes at 150  C to produce water.

CH3(CH2)4 H

(CH2)7COOH

CH2 C

C

C HH

C H

Linoleic acid; bp = 365 oC CH3(CH2)7

(CH2)7COOH C

C

H H Oleic acid; bp = 360 oC Zhong et al. [91] synthesized anatase nanorods by calcination of the precursor obtained by solvothermal treatment of a mixture of TIP, acetic acid, 2-propanol, and aniline, in the presence of a drop of H2SO4 at 100  C. H2SO4 was added to accelerate the esterification reaction. Because of the amorphous nature of the solvothermal product, mechanisms for the formation of nanostructure and role of amines were not elucidated. Titanium alkoxides are highly active homogeneous catalysts for esterification and transesterification [92e95], and the catalyst activity of TIP was reported to be 100 times higher than that of H2SO4 when activity was compared based on mole of the catalyst [94]. Aldol condensation may also be used for the supply of water. However, actual chemistry is quite complex and

3.2.3. Glycothermal Reaction of Titanium Alkoxides The reaction of TIP and TiO(acac)2 in 1,4-butanediol at 300  C yields anatase with relatively small crystallite sizes [72]. This is a simple extension of glycothermal reaction of aluminum compounds, and the first step of this reaction is an alkoxyl exchange reaction that takes place at lower temperatures [99]: MðORÞn þ nR0 OH % MðOR0 Þn þ nROH

(7)

This reaction is in equilibrium, and therefore the composition of alkyl groups in the coordination sites of the metal is determined by the relative amount of the two alkyl groups in the reaction system. Since the alcohol derived from the alkoxides has a lower boiling point than glycol, the alcohol will escape from the reaction system to the gas phase (see Section 3.1.6), and therefore the corresponding alkyl groups are completely expelled from the product.

Chapter | 11.1.4

Solvothermal Synthesis of Metal Oxides

The crystallite size of the product obtained from aluminum alkoxide increased in the following order: HO(CH2)2OH < HO(CH2)3OH < HO(CH2)6OH < HO(CH2)4OH. The physical properties of the products and the alumina derived by calcination thereof varied according to this order [53,100]. These results suggest that the development of the product structure is controlled by the heterolytic cleavage of the OeC bond of the glycoxide intermediate >AleOe(CH2)nOH formed by alkoxyl exchange between aluminum alkoxide and the glycol used as the medium. In the solvolytic reaction of >AleOe(CH2)nOH, aluminate ion (>AleO ) is considered to be a leaving group, and because of the electronwithdrawing effect of the intramolecular hydroxyl group the cleavage of the OeC bond proceeds more facilely with increasing the carbon number of the glycol. When the carbon number is 4 the cleavage of the OeC bond is accelerated by participation of the intramolecular hydroxyl group [101], yielding protonated tetrahydrofuran, which was actually detected in the supernatant of the reaction mixture (see, Figure 5). Similarly, reaction of titanium alkoxides proceeds yielding anatase. Kang et al. [102] reported that the anatase sample prepared from TIP using 1,4-butanediol has a high photocatalytic activity for the decomposition of CHCl3. Klongdee et al. [103] also reported that the glycothermally prepared anatase has a high activity for photocatalytic degradation of diuron [3-(3,4-dichlorophenyl)-1,1-dimethylurea], a herbicide. Although the reaction in 1,4-butanediol proceeds facilely yielding anatase, the reaction product obtained by the reaction of TIP in ethylene glycol was not identified at that time [72], and this product seems to be identical with the product obtained by Wang et al. [104]. They prepared a one-dimensional titanium glycolate complex, Ti(OCH2CH2O)2, by autoclaving a solution of titanium tetraethoxide, butylamine, and ethylene glycol in the molar ratio of 1:1:20 at 160e180  C for 5 days. In this structure, each ethylene glycolato ligand coordinates to a titanium ion

935

while one of the oxygen atoms also binds to another titanium ion, thus forming a one-dimensional structure. They also reported that needle-like anatase is prepared by topological decomposition of the glycolate precursor [105]. By the addition of titanium alkoxide to ethylene glycol (EG/Ti > 1000) preheated at 170  C, Jiang et al. [106] prepared titanium glycolate nanowires with an average diameter of 50 nm, which can easily be transformed into polycrystalline anatase nanowire by calcination. They noted that when precursor concentration is higher, much thicker and shorter nanowires were formed [106]. Kominami et al. [107] reported that solvothermal treatment of TiO(acac)2 in ethylene glycol at 300  C in the presence of sodium laurate [CH3(CH2)10COONa] and a small amount of water yielded microcrystalline brookite having an average size of 14 nm  67 nm without contamination of other TiO2 phases.

3.2.4. Reaction of Titanium Chloride Since this method is closely related to the nonhydrolytic solegel reaction [11,108,109], the basic chemistry of the reaction is discussed first. This method was initially used for the synthesis amorphous gels, but nowadays many papers report on the formation of crystalline products by the solvothermal treatment of the nonhydrolytic solegel systems. The initial step of the reaction of metal chloride with alcohol or ether is the formation of alkoxy groups by the alcoholysis or the etherolysis of metal halides: hMeX þ ROH / hMeOR þ HXðalcoholysisÞ

(8)

hMeX þ ROR / hMeOR þ RXðetherolysisÞ

(9)

The subsequent step is the formation of an oxo bridge between two metal atoms by eliminating a small organic molecule as shown in the following equations: hMeOR þ ROeMh / hMeOeMh þ RORðether eliminationÞ

(10)

hMeOCOR þ R0 OeMh / hMeOeMh þ R0 OCORðester eliminationÞ

(11)

hMeCl þ ROeMh / hMeOeMh þ RClðalkyl halide eliminationÞ

(12)

In the ordinary solegel reactions, this step is performed by the reaction of two of hMeOH species yielding hMeOeMh and water and is known as the condensation reaction. 1. Alkyl halide elimination FIGURE 5 Mechanism for the formation of metal oxides under glycothermal conditions.

Arnal et al. [110] examined the reactions of TiCl4 with TIP (Eqn 12), ethers (Eqn 9 / Eqn 12), and alcohols (Eqn 8 / Eqn 12) in a sealed glass tube at 110  C for 7 days.

936

Handbook of Advanced Ceramics

Titania samples prepared by these methods strongly varied in structure depending on the reaction used. For example, the reaction of TiCl4 with diethyl ether affords anatase, and the reaction of TiCl4 with ethanol led to rutile, whereas the reaction with tert-butyl alcohol resulted in the formation of a mixture of rutile and brookite. They concluded that the effect of the alcoholic medium is predominant for low-temperature crystallization of titania. However, we believe that hydrochloric acid liberated from the reaction facilitates the crystallization of titania through the bond breaking and forming equilibrium (as shown in Eqn 13): hTieOeTi þ HX % hTieOH þ XeTi

(13)

Arnal et al. [111] subsequently examined the solution chemistry of the alkyl halide elimination (Eqn 12) and the etherolysis (Eqn 9) by means of NMR spectroscopy. They found that the etherolysis (Eqn 9) of TiCl4 with diisopropyl ether (iPr2O) yielding Ti(OiPr)Cl3 is quite fast even at room temperature, while the second alkoxylation step is much slower and is not completed when condensation (Eqn 12) starts. The main reaction that takes place at room temperature in a mixture of TiCl4 and TIP is the redistribution of OiPr and Cl ligands. In both cases (etherolysis or direct condensation), the condensation reaction is slow at room temperature, and it takes place at 100  C after an induction period. Thus, the true precursors in both nonhydrolytic routes (alkoxide or ether) are titanium chloropropoxide species TiCl4 n(OiPr)n in equilibrium. Another interesting point is the autocatalytic nature of the condensation reaction. Usually, such behavior is explained by the reaction catalyzed by the reaction products. However, the condensation rate was not modified when the reaction was performed in the presence of titania or isopropyl chloride, the final products of the condensation. Therefore, Arnal et al. concluded that the intermediate oxochloropropoxide species (Cl3 n(OiPr)nTieOeTiCl3 n(OiPr)n) are much more reactive toward condensation than the starting chloropropoxide, possibly because they are better Lewis acids [111]. These results clearly show that complex chemistry exists in these reactions, although the overall reactions are simple. As for the ether elimination reaction (Eqn 10), Mutin and Vioux [11] wrote in their recent review that ether elimination has been postulated in the formation of metal oxoalkoxides from alkoxides. They also noted that methyl and benzyl groups favor the ether elimination route. Niederberger and Garnweitner [112] found that the reaction of TIP in benzyl alcohol [C6H5CH2OH; bp ¼ 205  C; causes severe eye irritation and may cause skin irritation] at 200  C proceeds by ether elimination mechanism, yielding dibenzyl ether together with crystalline anatase. As mentioned in Section 3.2.1, we proposed a heterolytic-cleavage mechanism for thermal decomposition of metal alkoxide (Eqn 6). In alcoholic media, this reaction proceeds more easily because the intermediates, carbocation

and metaloxo anion, are stabilized by salvation. The overall reaction of TIP can be written as follows: 2hTieOeCHðCH3 Þ2 / hTieOeTih þ ðCH3 Þ2 CHOH þ CH3 CH ¼ CH2 (14) Since the benzyl group cannot give olefinic products, the cation formed from this group reacts with alkoxide anion (RO ) yielding ether molecules. Following intensive studies on the formation of oxide gel by the direct reaction between metal halides and alcohols or alkoxides [110,111], Trentler et al. [113] in 1999 reported synthesis of “hydroxyl-free” anatase nanocrystals based on the alkyl halide elimination process (Eqn 12). Titanium halide was mixed with TOPO in heptadecane [CH3(CH2)15CH3; bp ¼ 301.8  C] and heated to 300  C under dry nitrogen and a titanium alkoxide was then rapidly injected into the hot solution. Anatase with a crystallite size of less than 10 nm in diameter was obtained, although there was considerable size distribution. However, the yields of the products were less than 50%. The reaction rate dramatically increased with greater branching of the alkyl group of the alkoxide, which is in good agreement with the reactivity of metal alkoxides for thermal decomposition as discussed in Section 2.1. Moreover, the boiling points of titanium halides are relatively low (TiCl4, 130  C; TiBr4, 231  C), and the reaction rate was not affected by the nature of X. The thermal decomposition of titanium alkoxide probably took place under their reaction conditions, and hydrogen halides liberated from TiX4 is facilitate the decomposition of titanium alkoxide, while Ti atoms from TiX4 seem not to be incorporated in formation of anatase. Trentler et al. made three important points: (1) Use of a high-boiling-point organic solvent such as heptadecane, which permits the high-temperature reaction under ambient pressure, (2) use of a capping (passivating) agent such as TOPO, which prevents the coagulation of nanoparticles during the synthesis, and (3) injection of the reactant to the preheated organic solvent, so that a huge number of nuclei of the product are formed. As mentioned in Section 1, the effect of pressure is negligible, and use of an open vessel (glass apparatus) or closed vessel (autoclave) is not particularly important unless the low-boiling-point byproducts affect the reaction: In this case, the reaction in an open vessel seems to facilitate the elimination of surface hydroxyl groups. For the second point, the use of phosphine-based capping agents possibly limits the application of the product. For the third point, injection of a large quantity of the reactant is impossible; therefore, this method cannot be applied for large-scale synthesis. Nevertheless, this paper had an enormous impact upon researchers in material chemistry, and this methodology is

Chapter | 11.1.4

Solvothermal Synthesis of Metal Oxides

now called an “injection method” and has been applied for synthesizing various metal oxides. Jun et al. [114] used a similar method for preparing anatase nanocrystals: Dioctyl ether [CH3(CH2)7O(CH2)7 CH3; bp ¼ 286e287  C] was used as a solvent and lauric acid [CH3(CH2)10COOH; mp ¼ 44.8  C; bp ¼ 298.9  C] and TOPO as selective and nonselective surface capping agents, since the former capping agent selectively binds to {001} faces through a bridge-bonding mode. The effect of the relative ratio of two types of capping agents was examined and it was found that truncated octagonal bypyramids were formed when lauric acid was added to the reaction system. Long rods and branched rods were also formed (Figure 6). Koo et al. [115] found that when a solution of TiCl4 in 1-octadecene [CH3(CH2)15CH]CH2; bp ~ 315  C] and a solution of TIP in the same solvent were simultaneously injected into preheated oleylamine using two syringe pumps, the injection rate of the two solutions affected the morphology as well as the crystal structure: The decreased injection rate preferred the formation of rutile.

CH3(CH2)7

(CH2)8NH2

C C H H Oleylamine; bp = 196 oC/15 mmHg Causes skin burns and may be harmful In contact with the skin Following the procedure of Trentler et al. [113], Vukicevic et al. [116] prepared an aqueous suspension of anatase nanorods (1-dodecylamine [CH3(CH2)11NH2; bp ¼ 247e249  C; causes skin burns] was also used as the capping agent together with TOPO). After the reaction, the products were dispersed in CHCl3 and irradiated with UV light in the presence of water. Photocatalytic decomposition of the capping agents causes migration of anatase nanorods to the aqueous layer. 2. Reaction of titanium halide in alcohols Wang et al. [117] reported the reaction of TiCl4 in various alcohols at 100  C and 160  C. Methanol, ethanol, 1propanol, and 2-propanol gave anatase, and butanol and octanol yielded rutile, while ethylene glycol yielded a mixture of rutile and anatase. The products consisted of spherical particles, but agglomerated tenuous fibers were obtained in octanol (rutile) and 2-propanol (anatase). The quite low crystallization temperature of titania and the low decomposition temperature of the alkoxides derived from TiCl4 are mediated by HCl liberated from TiCl4. They also noted that the phase formed by the reaction is affected by the concentration of HCl, a lower concentration of HCl favoring the formation of anatase.

937

FIGURE 6 Morphologies of anatase nanoparticles proposed by Jun et al. [144] (upper) and by Niederberger et al. [122] (lower).

Niederberger et al. developed a nonhydrolytic route by the use of benzyl alcohol [118e121], and titania can be crystallized at temperatures as low as 40  C [112,118]. Benzyl cation is stabilized by delocalization of the positive charge to the phenyl ring, and therefore the cleavage of CeO bond of this alcohol proceeds facilely. They also functionalized the surface of the nanoparticles of titania by the addition of catechol ligands such as dopamine and 4-tert-butylcatechol to the reaction mixture. This was achieved because the TiCl4/benzyl alcohol system proceeds at low reaction temperatures. This reaction provides a simple route to surface-functionalized nanoparticles that are dispersible either in water or in organic solvents [120]. They also found that when the same reaction is performed in the presence of 2-amino-2-(hydroxymethyl)-1,3-propanediol [(HOCH2)3CNH2] or 2-amino1,3-propanediol [(HOCH2)2CHNH2], the as-synthesized titania nanocrystals assemble into nanowires upon redispersion in water. These nanowires are composed of a continuous string of precisely ordered nanoparticles assembling along the [001] direction through oriented attachment [121]. They concluded that the anisotropic assembly is a consequence of the water-promoted desorption of the organic ligands selectively from the {001} faces of the crystalline building blocks together with the dissociative adsorption of water on these crystal faces. These processes induce the preferred attachment of the titania nanoparticles along the [001] direction (Figure 6) [122]. One shortcoming of this method is that the product was contaminated with chloride ions. Li et al. [123] obtained phase-pure anatase nanocrystals by mixing of TiCl4 with ethanol (>97%; both ethanolysis and hydrolysis of TiCl4 take place) at 0  C followed by gelation at AleO ) is expected to be formed by the decomposition of the glycoxide intermediate derived from aluminum alkoxide and 1,4-butanediol (see Figure 5), the presence of metal cation that gives basic oxides would give MeOeM0 bonds. According to this working hypothesis, Inoue et al. examined the reaction of aluminum isopropoxide (AIP) with yttrium acetate (Y(OAc)3; Ac ¼ CH3CO) in 1,4-butanediol at 300  C and found the formation of crystalline yttrium aluminum garnet (YAG) [6]. The hydrothermal reaction of pseudoboehmite (microcrystalline boehmite, the hydrolyzed product of AIP) with yttrium acetate at 300  C yielded boehmite together with a small amount of YAG. Single-phase YAG was not obtained, even with prolonged reaction time [6,153]. The difference between glycothermal and hydrothermal reaction can be attributed to the different stabilities of the intermediate phases, that is, the glycol derivative of boehmite vs. well-crystallized boehmite (see, Figure 4). Similarly, the reaction of the stoichiometric mixture of AIP and RE (GdeLu) acetate (RE(OAc)3) gives the corresponding rare earth aluminum garnet (REAG) in single phase [153]. Therefore, all the thermodynamically stable REAGs together with metastable GdAG were prepared by the glycothermal method. The glycothermal reaction of RE(OAc)3 alone yields RE(OH)2(OAc), REO(OAc) (two polymorphs), and RE(OH)(OAc)2, depending on the ionic size of the RE ion [154,155]. The acetate ions are not completely expelled from the coordination sites of the RE ion. However, in the presence of aluminum alkoxide as the starting material, acetate ions are fully eliminated from the product.

Handbook of Advanced Ceramics

Therefore, the aluminate ion facilitates the cleavage of the bond between acetate and RE ions. It is interesting to note that the reaction of Sm(OAc)3 or Eu(OAc)3 with AIP produced SmAG or EuAG, although the product was contaminated with RE acetate oxide (REO(OAc)) [153]. EuAG was recently prepared by a solegel method followed by calcination at 800e850  C [156], but SmAG has never been prepared by other methods. These results can be interpreted by the instability of the intermediate phase(s); it is possible that all the metastable phases having free energies lower than the intermediate can be formed from the thermodynamical point of view. The reaction of RE(OAc)3 (NdeLu) with gallium acetylacetonate (Ga(acac)3) in 1,4-butanediol at 300  C yielded the corresponding RE gallium garnets (REGGs) [26,157]. The garnet phases were reported to be thermodynamically stable for all the RE elements from Sm to Lu [158], and all of the stable REGGs were prepared by the glycothermal method (Figure 7). The reaction at 270  C also gave phase pure REGGs for SmeLu, but the reaction at 250  C gave amorphous products except for the reaction of yttrium, which gave yttrium gallium garnet (YGG). The hydrothermal reaction of Ga(acac)3 with RE acetate under conditions identical to the glycothermal reaction (except reaction medium) gave g-Ga2O3 together with a small amount of the garnet phase [26]. The reaction of Pr(OAc)3 and Ce(OAc)3 with Ga(acac)3 gave carbonate hydroxide (RECO3(OH)) as the sole crystalline product. When the reaction was carried out in the presence of gadolinium garnet (GGG) seed crystals, garnets (PrGG and CeGG) were crystallized in spite of the fact that cell parameters of these garnets are much larger than those of GGG. [26]. These results suggest that under glycothermal conditions, nucleation is the most difficult process and that, once nucleation takes place, crystal growth proceeds easily.

FIGURE 7 Synthesis of monodispersed particle of rare earth gallium garnets.

Chapter | 11.1.4

Solvothermal Synthesis of Metal Oxides

The particles of gallium garnets with large RE ions are spherical (0.5e2 mm) and the surface of the particles is smooth. On the other hand, the surface of the particles (0.1e0.3 mm) of the gallium garnets of Tb and RE elements having ionic sizes smaller than Tb is rough, with apparent polycrystalline outlines. However, high-resolution images of the latter type of particles show that a whole particle is covered with a single lattice fringe, indicating that each particle is a single crystal grown from only one nucleus. Inoue et al. concluded that the latter type of morphology is formed by quite rapid crystal growth [26]. For these garnets, monodispersed particles were obtained (Figure 7). Monodispersed particles can be prepared if a burst of nuclei is formed at the early stage of the reaction and if nucleation does take place during the crystal growth stage [159]. Once nucleation occurs in the glycothermal synthesis of the latter class of gallium garnets, the concentration of the intermediate in solution decreases, which is determined by the balance between the dissolution rate and the rate of consumption of the reactants in the solution by crystal growth. Since the crystal growth of the garnet with small RE ions proceeds rapidly, the concentration of the reactants in the solution becomes low, prohibiting nucleation during the crystal growth stage, thus leading to monodispersed particles. The morphology of the products suggests that defects are frequently incorporated into the products. This point seems to be closely connected with an important feature of solvothermal products. In hydrothermal reactions, dissolutionedeposition equilibrium takes place. Dissolution of adatoms on the surface occurs selectively, while preferential adsorption of ions at the vacancies of the surface proceeds. Therefore, a nearly perfect surface is formed by the hydrothermal reaction. On the other hand, under solvothermal conditions, dissolution of oxide materials into the organic solvent barely takes place, and therefore the product usually contains various types of crystal defects. The defect structure of the glycothermally prepared REAG was examined recently [160]. The unit cell parameters of REAGs were significantly larger than the values reported in JCPDS cards. Rietveld analysis of the products indicates the presence of Al vacancies in the 24d sites and oxygen vacancies in the 96h sites. Partial substitution of Al ions in the 16a sites with RE ions (antisite defects) was also suggested, which contributes to the enlargement of lattice parameter. The occupancy of RE ions in the 16a sites increased with the decrease in the ionic size of the RE element [160]. Allison et al. [161] prepared YAG:Ce (30 nm size) by the glycothermal method: Yttrium acetate, cerium acetate, and aluminum tris(2-methoxyethoxide) prepared by the reaction of AIP with 2-methoxyethanol were used as the starting materials, and the reaction was carried out in 1,4-butanediol in the presence of oleic acid and diphenyl

941

ether at 305  C for 12 h. The effects of the starting material and additives on the physical properties of the products were not clarified; however, Allison et al. reported that the fluorescence decay lifetime of the thus-prepared nanophosphor varied with temperature [161]. Nishi et al. [162] prepared YAG:Er3þ nanocrystals using the glycothermal method and examined the fluorescence properties in the optical-telecommunication band. The emission intensity and lifetime increased with the increase in the heat treatment temperature and time, which is attributed to the removal of organic matter remaining in the glycothermal product and the decrease in the specific surface area. They concluded that Er3þ ions in the surface region with a thickness of about 10 nm do not work well as luminescence centers due to the defects in the host material [162]. Kasuya et al. [163] prepared YAG:Ce3þ by the glycothermal method and examined the photoluminescence properties of the product. Contrary to the case of YAG:Er3þ, the photoluminescence intensity of this product decreased by calcination [163,164]. The internal quantum efficiency significantly increased when poly(ethylene glycol) (PEG) was added to the glycothermal system and PEG was found to be preferentially adsorbed on the phosphor particles preventing the oxidation of Ce3þ to Ce4þ. Kasuya et al. also showed that the addition of PEG increased the tetrahedral/octahedral ratio of Al3þ ions, which also contributes to the enhancement of photoluminescence intensity [163]. Photoluminescence intensity is also increased by prolonging the reaction time, and this result is attributed to the homogeneous incorporation of Ce3þ ions in the lattice of YAG, since the atomic composition of the product was not altered [165]. Interestingly, Kasuya et al. [165] showed that tetrahedral Al3þ ions are not present near the surface of the YAG:Ce3þ phosphor by comparison of single-pulse and cross-polarization MAS 27 Al NMR spectra. The same group also reported that the addition of citric acid to the glycothermal system significantly decreased the particle size of YAG:Ce3þ phosphor and increased its photoluminescence intensity [166]. Kasuya et al. also reported the synthesis of a transparent color-conversion thick film of YAG:Ce3þ prepared by the glycothermal method. The colloidal suspension of the product was placed in a refrigerator for 2 weeks to settle the coarse particles. To the supernatant, ethanol/methanol (95/5) was added and the mixture was centrifuged yielding a transparent yellow paste, which was placed inside the gap between the slide and cover glasses [167]. This group also used repeated glycothermal treatment so that the nanoparticles grow with the increase in the crystallinity. In this procedure, the particles formed by the first glycothermal reaction act as the seed crystals for the second glycothermal reaction as mentioned previously. Thus, they succeeded in suppressing the photobleaching of the YAG:Ce3þ phosphor caused by oxidation of Ce3þ ions [168]. Biological

942

application of the glycothermally prepared YAG:Ce3þ phosphor was also reported [169,170]. Nyman et al. [171] found that the YAG:Ce3þ phosphor prepared in 1,4-butanediol at 225  C has a higher photoluminescence efficiency and explained the result by suggesting that the phosphor is surface-passivated with the glycol derivative of boehmite formed as the intermediate. They also explored various glycols for synthesizing the phosphor at that temperature: 1,4-Pentanediol and 1,5pentanediol give amorphous products, and the product obtained in 1,3-butanediol does not emit the yellow-green luminescence, while diethylene glycol results in nonluminescence precipitates [171]. The Ce3þ emission of (Y,Tb)AG:Ce and (Y,Gd)AG:Ce prepared by a similar glycothermal method was reported to be red-shifted due to the increase in crystal field strength around Ce3þ ion [172].

3.3.2. Solvothermal Crystallization of Precipitated Gel Chinese groups use the co-precipitationesolvothermal route for the preparation of garnets. A stoichiometric mixture of rare earth nitrate and aluminum nitrates was precipitated with a suitable precipitant such as aqueous ammonia [173,174] and ammonium hydrogen carbonate [175,176], and the precipitates were washed. Then, the precipitates were dispersed in aqueous ethanol [173,177] or aqueous ethylenediamine [174,178] and then solvothermally treated. It was reported that YAG was crystallized by solvothermal treatment at 280  C, while the hydrothermal treatment of the same starting material at 300  C resulted in the formation of a mixture of YAG, g-Al O , 2 3 YOOH, and YAlO3 [173]. Using these techniques, these groups reported the formation of YAG [173,179], YAG:Ce [175], YAG:Eu [176], YAG:Tb [178], and LuAG [174]. In this method, the synthesis of homogeneous precipitate seems to be the most important point, as the authors noted that the reverse-strike method (i.e., addition of the metal salt solution to the precipitant solution so that the pH of the solution is kept alkaline during the course of the precipitation) has the advantage of higher cation homogeneity [178]. Based on close examination of the TEM images of the products reported [175e178], however, I think that the reproducibility of this method would be rather low.

3.3.3. NonHydrolytic SoleGel-Like Method Su et al. [180] report the synthesis of YAG:Ce nanophosphor with a size of 250 nm by the thermal decomposition of the solution of mixture of metal isopropoxides at 250  C in the presence of TOPO and oleic acid stabilizers. However, the details of the experimental procedure and characterization of the product were not reported. The same group also reported the synthesis of the phosphor of 5 nm size by the thermal

Handbook of Advanced Ceramics

decomposition of metal carbonateeoleate complex at 300  C; thus, a mixture of yttrium, cerium, and aluminum carbonates prepared by precipitating from the mixed chloride solution with ammonium carbonate was dissolved in oleic acid with vigorous stirring. To the solution, oleylamine was added and the mixture was solvothermally treated for 6 days [181].

3.3.4. Thermal Decomposition of Metal Nitrates in Alcohol Metal nitrates (Gd, Y, and Ce) and AIP dissolved in 2-propanol were thermally treated in an autoclave at 230  C for 5 h. This procedure gave an amorphous product, and (Gd0.9Y0.1)3Al5O12:Ce was prepared by calcination (>1000  C) of the product [182]. Using a similar method, YAG:Eu [183], GdAG:Dy [184], (Gd0.9Y0.1)3Al5O12:Tb [185], and GdAG:Tb [186] have been prepared. In this method, metal alkoxide may be formed according to the following equation: MðNO3 Þ3 þ 3ðCH3 Þ2 CHOH / MðOCHðCH3 Þ2 Þ3 þ 3HNO3 Alkoxides may be hydrolyzed by water formed by thermal dehydration of the secondary alcohol, 2-propanol. Because of the difference of the hydrolysis rate of the alkoxides, the homogeneity of metal cations in the product is not attained, and high calcination temperature is required, while an undesired phase such as orthorhombic phase may be contaminated in the products [182,184,186]. Although metal nitrate itself decomposes endothermically, the presence of organic compound in the reaction system may cause an exothermic reaction (see, Section 2.2). Therefore, the reaction of metal nitrates should not be carried out with an organic compound in a sealed vessel.

4. CONCLUDING REMARKS Metal oxides can be synthesized by various solvothermal techniques: solvothermal dehydration of metal hydroxides, solvothermal decomposition of metal alkoxides, solvothermal synthesis of mixed oxides, solvothermal crystallization of amorphous oxides, solvothermal ion exchange or intercalation, and solvothermal oxidation of metals. Although some of the reaction products may not be oxides, but hydroxides, oxyhydroxides, acetates, or organiceinorganic hybrid materials, oxides can be prepared by calcination of these products. Most of the solvothermal products have fine particle sizes and therefore the oxides derived from these products have physical properties significantly different from those of metal oxides obtained by conventional methods. In this article, solvothermal routes for synthesizing oxides are reviewed, addressing the differences and similarities between solvothermal and hydrothermal reactions.

Chapter | 11.1.4

Solvothermal Synthesis of Metal Oxides

As the hydrolysis of metal alkoxides proceeds exothermically, the reaction would appear to be irreversible. However, metal alkoxide and water are in equilibrium with hydroxide and alcohol [99], and this equilibrium can be attained at high temperature with high-boiling-point alcohols: MðOHÞn þ nROH % MðORÞn þ nH2 O The glycothermal reaction of aluminum alkoxide yields aluminum glycoxide by alkoxyl exchange reaction and the thermal decomposition of the intermediate yields the final product, i.e., the glycol derivative of boehmite. The structure and physical properties of the product are essentially identical to those obtained by the reaction of aluminum hydroxide in the same glycol. This is achieved by the above-mentioned equilibrium. This means that partially hydrolyzed metal alkoxide may be used in the glycothermal reaction. However, fully hydrolyzed alkoxide may give completely different results. Solvothermal reaction usually gives nanoparticles, but the toxicity of the nanoparticles is unknown. Since nanoparticles may pass through the cell membranes and become internalized within the cell, impairing cellular functions, the toxicity of nanoparticles is completely different from micrometer-size particles or bulk materials [187,188]. If nanoparticles are stabilized by a capping agent, the agent may cause the cytotoxicity, because of its presence at a high level on the surface of nanoparticles [188].

REFERENCES [1] Jones RW. Fundamental principles of solegel technology. Brookfield, VT: Institute of Metals; 1989. [2] Jeffrey CB. Solegel science: the physics and chemistry of solegel processing. Boston: Academic Press; 1990. [3] Figlarz M, Fievet F, Lagier JP. Preparation of controlled size metal powders in the micronic and sub-micronic range: a new process from polyol solutions. In: Presented at the international meeting on advanced materials, Tokyo, Japan, May 30eJune 3, 1988. Tokyo, Japan: Nikkan Kogyo Shinbun, Ltd; 1988. [4] Silvert P-Y, Herrera-Urbina R, Tekaia-Elhsissen K. Preparation of colloidal silver dispersions by the polyol process part 2. d mechanism of particle formation. J Mater Chem 1997;7:293e9. [5] Inoue M, Tanino H, Kondo Y, Inui T. Formation of microcrystalline a-alumina by glycothermal treatment of gibbsite. J Am Ceram Soc 1989;72:352e3. [6] Inoue M, Otsu H, Kominami H, Inui T. Synthesis of yttrium aluminum garnet by the glycothermal method. J Am Ceram Soc 1991;74:1452e4. [7] Feng W, Sun L-D, Zhang Y-W, Yan C-H. Synthesis and assembly of rare earth nanostructures directed by the principle of coordination chemistry in solution-based process. Coord Chem Rev 2010;254:1038e53. [8] Pinna N, Niederberger M. Surfactant-free nonaqueous synthesis of metal oxide nanostructures. Angew Chem Int Ed 2008;47:5292e304.

943

[9] Yan Z-G, Yan C-H. Controlled synthesis of rare earth nanostructures. J Mater Chem 2008;18:5046e59. [10] Takesue M, Hayashi H, Smith Jr RL. Thermal and chemical methods for producing zinc silicate (willemite): a review. Prog Cryst Growth Character Mater 2009;55:98e124. [11] Mutin PH, Vioux A. Nonhydrolytic processing of oxide-based materials: simple routes to control homogeneity, morphology, and nanostructure. Chem Mater 2009;21:582e96. [12] Yu S-H. Hydrothermal/solvothermal processing of advanced ceramic materials. J Ceram Soc Jpn 2001;109:S65e75. [13] Rajamathi M, Seshadri R. Oxide and chalcogenide nanoparticles from hydrothermal/solvothermal reactions. Curr Opin Solid State Mater Sci 2002;6:337e45. [14] Demazeau G. High pressures: recent trends in materials science. High Press Res 2000;18:203e12. [15] Demazeau G. Solvothermal processes: a route to the stabilization of new materials. J Mater Chem 1999;9:15e8. [16] Qian Y. Solvothermal synthesis of nanocrystalline IIIeV semiconductors. Adv Mater 1999;11:1101e2. [17] Bibby DM, Dale MP. Synthesis of silica sodalite from nonaqueous systems. Nature (London) 1985;317:157e8. [18] Bowes CL, Ozin GA. Self-assembling frameworks: beyond microporous oxides. Adv Mater 1996;6:13e28. [19] Day VW, Klemperer WG, Yaghi OM. Synthesis and characterization of a soluble oxide inclusion complex, ½CH3 CN3ðV12 O432 ފ. J Am Chem Soc. 1989;111:5959e61. [20] Spandl J, Bru¨dgam I, Hartl H. Solvothermal synthesis of a 24nuclear, cube-shaped squarato-oxovanadium(IV) framework: [N(nBu)4]8[V24O24(C4O4)12(OCH3)32]. Angew Chem Int Ed 2001;40:4018e20. [21] Inoue M, Kondo Y, Inui T. An ethylene glycol derivative of boehmite. Inorg Chem 1988;27:215e21. [22] Yu S-H, Yoshimura M. Shape and phase control of ZnS nanocrystals: template fabrication of wurtzite ZnS single-crystal nanosheets and ZnO flake-like dendrites from lamellar molecular precursor ZnS-(NH2CH2CH2NH2)0.5. Adv Mater 2002;14:296e300. [23] Zacher D, Shekhah O, Woll C, Fischer RA. Thin films of metaleorganic frameworks. Chem Soc Rev 2009;38:1418e29. [24] Jiang Y, Wu Y, Zhang, Xu C, Yu W, Xie Y, et al. Catalyticassembly solvothermal route to multiwall carbon nanotubes at a moderate temperature. J Am Chem Soc 2000;122:12283e384. [25] Hu G, Cheng M, Ma D, Bao X. Synthesis of carbon nanotube bundles with mesoporous structure by a self-assembly solvothermal route. Chem Mater 2003;15:1470e3. [26] Inoue M, Nishikawa T, Otsu H, Kominami H, Inui T. Synthesis of rare earth gallium garnets by the glycothermal method. J Am Ceram Soc 1998;81:1173e83. [27] Chen S-J, Chen X-T, Xue Z, Li L-H, You X-Z. Solvothermak preparation of Cu2O crystalline particles. J Cryst Growth 2002;246:169e75. [28] Stupp SI, Braun PV. Molecular manipulation of microstructures: biomaterials, ceramics, and semiconductors. Science 1997;277: 1242e8. [29] Patzke GR, Krumeich F, Nesper R. Oxidic nanotubes and nanorodsdanisotropic modules for a future nanotechnology. Angew Chem Int Ed 2002;41:2446e61. [30] Rao CNR, Deepak FL, Gundiah G, Govindaraj A. Inorganic nanowires. Prog Solid State Chem 2003;31:5e147.

944

[31] Wei C, Deng Y, Lin Y-H, Nan C-W. Solvothermal synthesis of porous tellurium nanotubes. Chem Phys Lett 1980;372:590e4. [32] Yu S-H, Yang J, Qian Y-T, Yoshimura M. Optical properties of ZnS nanosheets, ZnO dendrites, and their lamellar precursors ZnS (NH2CH2CH2NH2)0.5. Chem Phys Lett 2002;361:362e6. [33] Inoue M, Kondo Y, Inui T. The reaction of crystalline aluminum hydroxide in ethylene glycol. Chem Lett 1986:1421e4. [34] Inoue M. Solvothermal synthesis. In: Lee B, Komarneni S, editors. Chemical processing of ceramics. Boca Raton: Taylor & Francis; 2005. Chap 2. [35] Ehrentraut D, Sato H, Kagamitani Y, Sato H, Yoshikawa A, Fukuda T. Solvothermal growth of ZnO. Prog Cryst Growth Character Mater 2006;52:280e335. [36] Yoshimura M, Somiya S. Fabrication of dense nonstabilized ZrO2 ceramics by hydrothermal reaction sintering. Am Cream Soc Bull 1980;59:246. [37] Hirano S. Hydrothermal processing of ceramics. Am Cream Soc Bull 1987;67:1342e4. [38] Dawson WJ. Hydrothermal synthesis of advanced ceramic powders. Am Cream Soc Bull 1988;67:1673e8. [39] Walton RI. Subcritical solvothermal synthesis of condensed inorganic materials. Chem Soc Rev 2002;31:230e8. [40] Yamasaki N. Development of recycling technologies, and functional material formation by hydrothermal processes. J Ceram Soc Jpn 2003;111:709e15. [41] Jacobs H, Stu¨ve C. Hochdrucksynthese der h-phase im system MneN: Mn3N2. J Less Common Met 1984;96:323e9. [42] Su HL, Xie Y, Li B, Liu XM, Qian YT. A Simple, convenient, mild solvothermal route to nanocrystalline Cu3P and Ni2P. Solid State Ionics 1999;122:157e60. [43] Demazeau G, Goglio G, Denis A, Largeteau A. Solvothermal synthesis: a new route for preparing nitrides. J Phys Condens Matter 2002;14:11085e8. [44] Bialowons H, Mu¨ller M, Mu¨ller BG. Titanium tetrafluorideda surprising simple column structure. Anorg Allg Chem 1995;621:1227e31. [45] Cooper ER, Andrews CD, Wheatley PS, Webb PB, Wormald P, Morris RE. Ionic liquids and eutectic mixtures as solvent and template in synthesis of zeolite analogues. Nature 2004;430: 1012e6. [46] Abbott AP, Capper G, Davies DL, Rasheed RK, Tambyrajah V. Novel solvent properties of choline chloride/urea mixtures. Chem Commun 2003:70e1. [47] Parnham ER, Morris RE. Ionothermal synthesis of zeolites, metaleorganic frameworks, and inorganiceorganic hybrids. Acc Chem Res 2007;40:1005e13. [48] Tarascon J-M, Recham N, Armand M, Chotard J-N, Barpanda P, Walker W, et al. Hunting for better Li-based electrode materials via low temperature inorganic synthesis. Chem Mater 2010;22:739e47. [49] Kennedy GC. Pressureevolumeetemperature relations in water at elevated temperatures and pressures. Am J Sci 1958;248: 540e64. [50] Hanson ES. Calculation of liquid densities at elevated temperatures. Ind Eng Chem 1949;41:96e9. [51] Koerts T, Deelen MJAG, van Santen RA. Hydrocarbon formation from methane by a low-temperature 2-step reaction sequence. J Catal 1992;138:101e14.

Handbook of Advanced Ceramics

[52] Spencer PS, Kim MS, Sabri MI. Aromatic as well as aliphatic hydrocarbon solvent axonopathy. Int J Hyg Environ Health 2002;205:131e6. [53] Inoue M, Kominami H, Inui T. Reaction of aluminium alkoxides with various glycols and the layer structure of their products. J Chem Soc Dalton Trans 1991:3331e6. [54] Inoue M, Kominami H, Inui T. Thermal reaction of aluminum alkoxides in glycols. J Am Ceram Soc 1990;73:1100e2. [55] Inoue M, Kimura M, Inui T. Alkoxyalumoxanes. Chem Mater 2000;12:55e61. [56] Ginsberg H, Hu¨ttig W, Strunk-Llichtenberg G. The Influence of the starting material on the crystalline forms arising in the thermal decomposition and conversion of aluminum hydroxides. II. influence of the starting material on the course of the decomposition of the g-hydroxides. Z Anorg Allgem Chem 1957;293: 204e13. [57] Fyfe WS, Hollander MA. Equilibrium dehydration of diaspore at low temperatures. Am J Sci 1964;262:709e12. [58] Yamaguchi G, Yanagida H, Fujimaru T. Corundum a-Al2O3 formation from the dehydration of boehmite g-AlOOH under hydrothermal conditions. II. the reaction rates and the mechanism. Bull Chem Soc Jpn 1965;38:54e8. [59] Inoue M, Kimura M, Kominami H, Tanino H, Inui T. Mechanisms for the Formation of a-Alumina from Aluminum Hydroxide by Glycothermal Reaction. In: 72nd National Meeting of the Chemical Society of Japan, Tokyo, March 1997, abstract no. 1BY09; 1997. [60] de Boer JH, Fortuin JMH, Steggerda JJ. The dehydration of alumina hydrates. Konikl Ned Acad Wetenschap Proc 1954;57B:170e80. [61] de Boer JH, Fortuin JMH, Steggerda JJ. The dehydration of alumina hydrates. II. Konikl Ned Acad Wetenschap Proc 1954;57B:435e43. [62] Arai Y, Yasue T, Yamaguchi I. Mechanochemical phenomena of aluminum hydroxide by grinding. Nippon Kagaku Kaishi (J Chem Soc Jpn) 1972:1395e404. [63] Kaiser A, Sporn D, Bertagnolli H. Phase transformations and control of habit in lyothermal synthesis of a-Al2O3. J Eur Ceram Soc 1994;14:77e83. [64] Bell NS, Cho S-B, Adair JH. Size control of a-alumina particles synthesized in 1,4-butanediol solution by a-alumina and a-hematite seeding. J Am Ceram Soc 1998;81:1411e20. [65] Cho SB, Venigalla S, Adair J,H. Morphological forms of a-alumina particles synthesized in 1,4-butanediol solution. J Am Ceram Soc 1996;79:88e96. [66] Bell NS, Adair JH. Adsorbate effects on glycothermally produced a-alumina particle morphology. J Cryst Growth 1999;203: 213e26. [67] Adair JH, Cho S-B, Bell NS, Perrotta AJ. Recent developments in morphological control of a-Al2O3 particles synthesized in 1,4butanediol solution. J Dispersion Sci Technol 2001;22:143e65. [68] Tian M, Shi E, Wang B, Yuan R, Li W, Zhong W, Guo J. Solvothermal preparation of a-alumina and boehmite crystallites and investigations on the formation of their morphologies. Sci Chin: E 1998;41:288e94. [69] Li W-J, Shi E-W, Yin Z-W. Growth habit of rutile and a-Al2O3 crystals. J Cryst Growth 2000;208:546e54. [70] Song KX, Wu SY, Chen XM, Qin N. Glycothermal synthesis of a-Al2O3 in 1,4-butadioal organic solvent. J Electroceram 2008;21:805e9.

Chapter | 11.1.4

Solvothermal Synthesis of Metal Oxides

[71] Li C, Choi P. Molecular dynamics study on the effect of solvent adsorption on the morphology of glycothermally produced a-alumina particles. J Phys Chem C 2008;112:10145e52. [72] Inoue M, Kominami H, Otsu H, Inui T. Synthesis of microcrystalline titania in organic media. Nippon Kagaku Kaishi (J Chem Soc Jpn) 1991:1364e6. [73] Hitchman ML, Jensen KF. Chemical vapor deposition: principles and applications. London: Academic Press; 1993. [74] Inoue M, Kominami H, Inui T. Thermal transformation of c-alumina formed by thermal decomposition of aluminum alkoxide in organic media. J Am Ceram Soc 1992;75:2597e8. [75] Inoue M, Kominami H, Inui T. Novel synthetic method for the catalytic use of thermally stable zirconia: thermal decomposition of zirconium alkoxides in organic media. Appl Catal A: General 1993;97:L25e30. [76] Shulman GP, Trusty M, Vickers JH. Thermal decomposition of aluminum alkoxides. J Org Chem 1963;28:907e10. [77] Kominami H, Kato J, Takada Y, Doushi Y, Ohtani B, Nishimoto S, Inoue M, Inui T, Kera Y. Novel synthesis of microcrystalline titanium(IV) oxide having high thermal stability and ultra-high photocatalytic activity: thermal decomposition of titanium(IV) alkoxide in organic solvents. Catal Lett 1997;46:235e40. [78] Yokozawa M, Iwasa H, Teramoto I. Vapor deposition of TiO2. Jpn J Appl Phys 1968;7:96e7. [79] Okuyama K, Kousaka Y, Tohge N, Ymamamoto S, Wu JJ, Flagan RC, Seinfeld JH. Production of ultrafine metal oxide aerosol particles by thermal decomposition of metal alkoxide vapors. AIChE J 1986;32:2010e9. [80] Komiyama H, Kanai T, Inoue H. Preparation of porous, amorphous, and ultrafine TiO2 particles by chemical vapor deposition. Chem Lett 1984:1283e6. [81] Payakgul W, Mekasuwandumrong O, Pavarajarn V, Praserthdam P. Effects of reaction medium on the synthesis of TiO2 nanocrystals by thermal decomposition of titanium(IV) n-butoxide. Ceram Internat 2005;31:391e7. [82] Kim C-S, Moon BK, Park J-H, Chung ST, Son S-M. Synthesis of nanocrystalline TiO2 in toluene by a solvothermal route. J Cryst Growth 2003;254:405e10. [83] Parala H, Devi A, Bhakta R, Fishcher RF. Synthesis of nano-scale TiO2 particles by a nonhydrolytic approach. J Mater Chem 2002;12:1625. [84] Fanelli AJ, Burlew JV. Preparation of fine alumina powder in alcohol. J Am Ceram Soc 1986;69:174eC175. [85] Takahashi Y, Suzuki H, Nasu N. Rutile growth at the surface of TiO2 films deposited by vapour-phase decomposition of isopropyl titanate. J Chem Soc Faraday Trans I 1985;81:3117e25. [86] Courtecuisse VG, Bocquet JF, Chhor K, Pommier C. Modeling of a continuous reactor for TiO2 powder synthesis in a supercritical fluiddexperimental validation. J Supercrit Fluids 1996;9:222e6. [87] Courtecuisse VG, Chhor K, Bocquet J-F, Pommier C. Kinetics of the titanium isopropoxide decomposition in supercritical isopropyl alcohol. Ind Eng Chem Res 1996;35:2539e45. [88] Kominami H, Kato J, Murakami S, Kera Y, Inoue M, Inui T, Ohtani B. Synthesis of titanium(IV) oxide of ultra-high photocatalytic activity: high-temperature hydrolysis of titanium alkoxides with water liberated homogeneously from solvent alkohols. J Mol Catal A: Chem 1999;144:165e71. [89] Ivanda M, Music S, Popovic S, Golic M. XRD, Raman and FT-IR spectroscopic observations of nanosized TiO2 synthesized by the

945

[90]

[91]

[92]

[93] [94]

[95]

[96]

[97]

[98] [99] [100]

[101]

[102]

[103]

[104]

[105]

[106]

[107]

solegel method based on an esterification reaction. J Mol Struct 1999;480:645e9. Cozzoli PD, Kornowski A, Weller H. Low-temperature synthesis of soluble and processable organic-capped anatase TiO2 nanorods. J Am Chem Soc 2003;125:14539e48. Zhong Z, Ang T-P, Luo J, Gan H-C, Gedanken A. Synthesis of one-dimensional and porous TiO2 nanostructures by controlled hydrolysis of titanium alkoxide via coupling with an esterification reaction. Chem Mater 2005;17:6814e8. Marsi M. Mechanistic study of titanate-catalyzed transesterification: influence of aggregation on reactivity. Inorg Chem 1988;27:3064e6. Blandy C, Pellegatta J-L, Gilot B. Catalytic properties of supported titanates in transesterification. J Catal 1994;150:150e4. Blandy C, Pellegatta J-L, Cassoux P. Homogeneous and supported titanates as catalysts for transesterification of acrylic esters. Catal Lett 1997;43:139e42. Ahmadnian F, Velasquez F, Reichert K-H. Screening of different titanium(IV) catalysts in the synthesis of poly(ethylene terephthalate). Macromol React Eng 2008;2:513e21. Steunou N, Ribot F, Boubekeur K, Maquet J, Sanchez C. Ketones as an oxolation source for the synthesis of titanium-oxo-organo cluster. New J Chem 1999;23:1079e86. Garnweitner G, Antonietti M, Niederberger M. Nonaqueous synthesis of crystalline anatase nanoparticles in simple ketones and aldehydes as oxygen-supplying agents. Chem Commun 2005:397e9. Li X-L, Peng Q, Yi J-X, Wang X, Li Y. Near monodisperse TiO2 nanoparticles and nanorods. Chem Eur J 2006;12:2383e91. Bradly DC. Metal alkoxides. Prog Inorg Chem 1960;2:303e61. Inoue M, Kominami H, Inui T. Synthesis of large pore-size and large pore-volume aluminas by glycothermal treatment of aluminium alkoxide and subsequent calcination. J Mater Sci 1994;29:2459e66. Winstein S, Allred E, Heck R, Glick R. Neighboring methoxyl participation in solvolytic nucleophilic substitution. Tetrahedon 1958;3:1e13. Kang M, Lee S-Y, Chung C-H, Cho SM, Han GY, Kim B-W, et al. Characterization of a TiO2 photocatalyst synthesized by the solvothermal method and its catalytic performance for CHCl3 decomposition. J Photochem Photobiol A: Chem 2001;144:185e91. Klongdee J, Petchkroh W, Phuempoonsathapron K, Praserthdam P, Vangnai AS, Pavaranjarn V. Activity of nanosized titania synthesized from thermal decomposition of titanium(IV) nbutoxide for the photocatalytic degradation of diuron. Sci Technol Adv Mater 2005;6:290e5. Wang D, Yu R, Kumada N, Kinomura N. Hydrothermal synthesis and characterization of novel one-dimensional titanium glycolate complex single crystal: Ti(OCH2CH2O)2. Chem Mater 1999;11: 2008e12. Wang D, Yu R, Chen Y, Kumada N, Kinomura N, Takano M. Photocatalysis property of needle-like TiO2 prepared from a novel titanium glycolate precursor. Solid State Ionics 2004;172:101e4. Jiang X, Wang Y, Herricks T, Xia Y. Ethylene glycol-mediated synthesis of metal oxide nanowires. J Mater Chem 2004; 14:695e703. Kominami H, Kohno M, Kera Y. Synthesis of brookite-type titanium oxide nano-crystals in organic media. J Mater Chem 2000;10:1151e6.

946

[108] Vioux A, Leclercq D. Non-aqueous routes to solegel. Heterogen Chem Chem Rev 1996;3:65e73. [109] Vioux A. Nonhydrolytic solegel routes to oxides. Chem Mater 1997;9:2292e9. [110] Arnal P, Corriu RJP, Leclercq D, Mutin PH, Vioux A. Preparation of anatase, brookite and rutile at low temperature by non-hydrolytic solegel methods. J Mater Chem 1996;6:1925e32. [111] Arnal P, Corriu RJP, Leclercq D, Mutin PH, Vioux A. A solution chemistry study of nonhydrolytic solegel routes to titanis. Chem Mater 1997;9:694e8. [112] Niederberger M, Garnweitner G. Organic reaction pathway in the nonaqueous synthesis of metal oxide nanoparticles. Chem Eur J 2006;12:7282e302. [113] Trentler TJ, Denler TE, Bertone JF, Agrawal A, Colvin VL. Synthesis of TiO2 nanocrystals by nonhydrolytic solution-based reactions. J Am Chem Soc 1999;121:1613e4. [114] Jun Y-W, Casula MF, Sim J-H, Kim SY, Cheon J, Alivisatos AP. Surfactant-assisted elimination of a high energy facet as a means of controlling the shape of TiO2 nanocrystals. J Am Chem Soc 2003;125:15981e5. [115] Koo B, Park J, Kim Y, Choi S-H, Sung Y-E, Hyeon T. Simultaneous phase- and size-controlled synthesis of TiO2 nanorods via non-hydrolytic solegel reaction of syringe pump delivered precursors. J Phys Chem B 2006;110:24318e23. [116] Vukicevic U, Ziemian S, Bismarck A, Shaffer MSP. Self-cleaning anatase nanorods: photocatalytic removal of structure directing agents and subsequent surface modification. J Mater Chem 2008;18:3448e53. [117] Wang C, Deng Z-X, Zhang G, Fan S, Li Y. Synthesis of nanocrystalline TiO2 in alcohols. Powder Technol 2002;125:39e44. [118] Niederberger M, Bartl MH, Stucky GD. Benzyl alcohol and titanium tetrachlorideda versatile reaction system for the nonaqueous and low-temperature preparation of crystalline and luminescent titania nanoparticles. Chem Mater 2002;14: 4364e70. [119] Niederberger M, Bartl MH, Stucky GD. Benzyl alcohol and transition metal chlorides as a versatile reaction system for the nonaqueous and low-temperature synthesis of crystalline nanoobjects with controlled dimensionality. J Am Chem Soc 2002;124:13642e3. [120] Niederberger M, Garnweitner G, Krumeich F, Nesper R, Co¨lfen H, Antonietti M. Tailoring the surface and solubility properties of nanocrystalline titania by a nonaqueous in situ functionalization process. Chem Mater 2004;16:1202e8. [121] Polleux J, Pinna N, Antonietti M, Niederberger M. Liganddirected assembly of preformed titania nanocrystals into highly anisotropic nanostructures. Adv Mater 2004;16:436e9. [122] Polleux J, Pinna N, Antonietti M, Hess C, Wild U, Schlo¨gl R, et al. Ligand functionality as a versatile tool to control the assembly behavior of preformed titania nanocrystals. Chem Eur J 2004;10:925e3551. [123] Li G, Li L, Boerio-Goates J, Woodfield BF. High purity anatase TiO2 nanocrystals: near room-temperature synthesis, grain growth kinetics, and surface hydration chemistry. J Am Chem Soc 2005;127:8659e66. [124] Wang C, Deng Z-X, Li Y. The Synthesis of nanocrystalline anatase and rutile titania in mixed organic media. Inorg Chem 2001;40:5210e4.

Handbook of Advanced Ceramics

[125] Doeuff S, Dromzee Y, Taulelle F, Sanchez C. Synthesis and solidand liquid-state characterization of a hexameric cluster of titanium(IV): Ti6(m2-O)2(m3-O)2(m2-OC4H9)6(OCOCH3)8. Inorg Chem 1989;28:4439e45. [126] Kim C-S, Moon BK, Park J-H, Choi B-C, Seo H-J. Solvothermal synthesis of nanocrystalline TiO2 in toluene with surfactant. J Cryst Growth 2003;257:309e15. [127] Joo J, Kwon SG, Yu T, Cho M, Lee J, Yoon J, et al. Large-scale synthesis of TiO2 nanorods via nonhydrolytic sol-gel ester elimination reaction and their application to photocatalytic inactivation of E. coli. J Phys Chem B 2005;109:15297e302. [128] De Marco L, Manca M, Giannuzzi R, Malara F, Melcarne G, Ciccarella G, et al. Novel preparation method of TiO2-nanorod-based photoelectrodes for dye-sensitized solar cells with improved light-harvesting efficiency. J Phys Chem C 2010;114: 4228e36. [129] Zhang Z, Zhong X, Liu S, Li D, Han M. Aminolysis route to monodisperse titania nanorods with tunable aspect ratio. Angew Chem Int Ed 2005;44:3466e70. [130] Fujiwara M, Wessel H, Park HS, Roesky HW. A Solegel method using tetraethoxysilane and acetic anhydride: immobilization of cubic m-Oxo SieTi complex in a silica matrix. Chem Mater 2002;14:4975e81. [131] Guo G, Whitesell JK, Fox MA. Synthesis of TiO2 photocatalysts in supercritical CO2 via a non-hydrolytic route. J Phys Chem B 2005;109:18781e5. [132] Inoue M, Kominami H, Inui T. Novel synthesis method for thermally stable monoclinic zirconia: hydrolysis of zirconium alkoxides at high temperatures with a limited amount of water dissolved in inert organic solvent from the gas phase. Appl Catal A 1995;121:L1e5. [133] Kominami H, Takada Y, Yamagiwa H, Kera Y, Inoue M, Inui T. Synthesis of thermally stable nanocrystalline anatase by hightemperature hydrolysis of titanium alkoxide with water dissolved in organic solvent from gas phase. J Mater Sci Lett 1996;15:197e200. [134] Kominami H, Murakami S, Kera Y, Ohtani B. Titanium(IV) oxide photocatalyst of ultra-high activity: a new preparation process allowing compatibility of high adsorptivity and low electronehole recombination probability. Catal Lett 1998;56:125e9. [135] Kominami H, Kohno M, Takada Y, Inoue M, Inui T, Kera Y. Hydrolysis of titanium alkoxide in organic solvent at high temperatures: a new synthetic method for nanosized, thermally stable titanium(IV) oxide. Ind Eng Chem Res 1999;38:3925e31. [136] Kominami H, Oki K, Kohno M, Onoue S, Kera Y, Ohtani B. Novel solvothermal synthesis of niobium(V) oxide powders and their photocatalytic activity in aqueous suspensions. J Mater Chem 2001;11:604e9. [137] Kominami H, Miyakawa M, Murayama S, Yoshida T, Kohno M, Onoue S, et al. Solvothermal synthesis of tantalum(V) oxide nanoparticles and their photocatalytic activities in aqueous suspension system. Phys Chem Chem Phys 2001;3:2697e703. [138] Dinh C-T, Nguyen T-D, Kleitz F, Do T-O. Shape-controlled synthesis of highly crystalline titania nanocrystals. ACS Nano 2009;3:3734e7. [139] Wahi RK, Liu Y, Falkner JC, Colvin VL. Solvothermal synthesis and characterization of anatase TiO2 nanocrystals with ultrahigh surface area. J Colloid Interface Sci 2006;302:530e6.

Chapter | 11.1.4

Solvothermal Synthesis of Metal Oxides

[140] Lee S-H, Kang M, Cho SM, Han Y, Kim B-W, Yoon KJ, et al. Synthesis of TiO2 photocatalyst thin film by solvothermal method with a small amount of water and its photocatalytic performance. J Photochem Photobiol, A 2001;146:121e8. [141] Lin H, de Oliveira PW, Grobelsek I, Haettich A, Veith M. The synthesis of anatase TiO2 nanoparticles by solvothermal method using ionic liquid as additive. Z Anorg Allg Chem 2010;636:1947e54. [142] Wu J, Lu¨ X, Zhang L, Huang F, Xu F. Dielectric constant controlled solvothermal synthesis of a TiO2 photocatalyst with tunable crystallinity: a strategy for solvent selection. Eur J Inorg Chem 2009:2789e95. [143] Inoue M, Kitamura K, Tanino H, Nakayama H, Inui T. Alcohothermal treatments of gibbsite: mechanisms for the formation of boehmite. Clays Clay Miner 1989;37:71e80. [144] Inoue M. Glycothermal synthesis of metal oxides. J Phys Cond Matter 2004;16:S1291e303. [145] Cockayne B. The use and enigmas of the Al2O3eY2O3 phase system. J Less-Common Met 1985;114:199e206. [146] Abell JS, Harris IR, Cockayne B, Lent B. An investigation of phase stability in the Y2O3eAl2O3 system. J Mater Sci 1974;9:527e37. [147] Messier DR, Gazza GE. Systems of MgAl2O4 and Y3Al5O12 by thermal decomposition of hydrated nitrate mixture. Ceram Bull 1972;51:692e7. [148] Yamaguchi O, Takeoka K, Hirota K, Takano H, Hayashida A. Formation of alkoxy-derived yttrium aluminum-oxides. J Mater Sci 1992;27:1261e4. [149] Cinibulk MK. Synthesis of yttrium aluminum garnet from a mixed-metal citrate precursor. J Am Ceram Soc 2000;83: 1276e8. [150] Kakade MB, Ramanathan S, Ravindran PV. Yttrium aluminum garnet powders by nitrate decomposition and nitrateeurea solution combustion reactionsda comparative study. J Alloys Compd 2003;350:123e9. [151] Mill’ BV. Hydrothermal synthesis of aluminum and gallium garnets. Soviet Phys Crystallogr 1967;12:137e9. [152] Kolb ED, Laudise RA. Phase equilibria of Y3Al5O12 hydrothermal growth of Gd3Ga5O12 and hydrothermal epitaxy of magnetic garnets. J Cryst Growth 1975;29:29e39. [153] Inoue M, Otsu H, Kominami H, Inui T. Glycothermal synthesis of rare earth aluminum garnets. J Alloys Comp 1995;226:146e51. [154] Inoue M, Kominami H, Otsu H, Inui T. Formation of La(CH3COO)2(OH) by glycothermal treatment of lanthanum acetate sesquihydrate and its thermal decomposition behavior. Nippon Kagaku Kaishi (J Chem Soc Jpn) 1991:1254e60. [155] Kominami H, Inoue M, Inui T. Formation of lanthanide acetate hydroxide and acetate oxide by glycothermal treatment of lanthanide acetate hydrate and their thermal decomposition behavior. Nippon Kagaku Kaishi (J Chem Soc Jpn) 1993: 605e11. [156] Garskaite E, Sakirzanovas S, Kareiva A, Glaser J, Meyer H-J. Synthesis and structure of europium aluminum garnet (EAG). Z Anorg Allg Chem 2007;633:990e3. [157] Inoue M, Otsu H, Kominami H, Inui T. Synthesis of submicron spherical crystals of gadolinium gallium garnets by the glycothermal method. J Mater Sci Lett 1995;14:1303e5. [158] Mizuno M, Yamada T. Phase diagram of the system Ga2O3eSm2O3 at high temperatures. J Ceram Soc Jpn 1989;97:1334e8.

947

[159] LaMer VK, Dinegar RH. Theory, production and mechanism of formation of monodispersed hydrosol. J Am Chem Soc 1950;72:4847e54. [160] Hosokawa S, Tanaka Y, Iwamoto S, Inoue M. Defect structure of rare earth aluminium garnets obtained by the glycothermal method. J Alloys Compd 2008;451:309e13. [161] Allison SW, Gillies GT, Rondinone AJ, Cates MR. Nanoscale thermometry via the fluorescence of YAG:Ce phosphor particles: measurements from 7 to 77  C. Nanotechnol 2003;14:859e63. [162] Nishi M, Tanabe S, Inoue M, Takahashi M, Fujita K, Hirao K. Optical-telecommunication-Bband fluorescence properties of Er3þ-doped YAG nanocrystals synthesized by glycothermal method. Opt Mater 2005;27:655e62. [163] Kasuya R, Isobe T, Kuma H, Katano J. Photoluminescence enhancement of PEG-modified YAG:Ce3þ nanocrystal phosphor prepared by glycothermal method. J Phys Chem B 2005;109:22126e30. [164] Isobe T. Low-temperature wet chemical syntheses of nanocrystal phosphors with surface modification and their characterization. Phys Status Solidi A 2006;203:2686e93. [165] Kasuya R, Isobe T, Kuma H. Glycothermal synthesis and photoluminescence of YAG:Ce3þ nanophosphors. J Alloys Compd 2006;408-412:820e3. [166] Asakura R, Isobe T, Kurokawa K, Takagi T, Aizawa H, Ohkubo M. Effects of citric acid additive on photoluminescence properties of YAG:Ce3þ nanoparticles synthesized by glycothermal reaction. J Lumin 2007;127:416e22. [167] Kasuya R, Kawano A, Isobe T, Kuma H, Katano J. Characteristic optical properties of transparent color conversion film prepared from YAG:Ce3þ nanoparticles. Appl Phys Lett 2007;91:111916. [168] Kamiyama Y, Hiroshima T, Isobe T, Koizuka T, Takashima S. Photostability of YAG:Ce3þ nanophosphors synthesized by glycothermal method. J Electrochem Soc 2010;157:J149e54. [169] Asakura R, Isobe T, Kurokawa K, Aizawa H, Ohkubo M. Tagging of avidin immobilized beads with biotinylated YAG:Ce3þ nanocrystal phosphor. Anal Bioanal Chem 2006;386:1641e7. [170] Asakura R, Kusayama I, Saito D, Isobe T, Kurokawa K, Hirayama Y, et al. Preparation of fluorescent poly(methyl methacrylate) beads hybrized with Y3Al5O12:Ce3þ nanophosphor for biological application. Jpn J Appl Phys 2007;46:5193e5. [171] Nyman M, Shea-Rohwer LE, Martin JE, Provencio P. Nano-YAG: Ce mechanism of growth and epoxy-encapsulation. Chem Mater 2009;21:1536e42. [172] Byun H-J, Song W-S, Kim Y-S, Yang H. Solvothermally grown Ce3þ-doped Y3Al5O12 colloidal nanocrystals: spectral variations and white LED characteristics. J Phys D: Appl Phys 2010;43:195401. [173] Zhang X, Liu H, He W, Wang J, Li X, Boughton RI. Synthesis of monodisperse and spherical YAG nanopowder by a mixed solvothermal method. J Alloys Compd 2004;372:300e3. [174] Xing L, Peng L, Gu M, Tang G. Solvothermal synthesis of lutetium aluminum garnet nanopowders: determination of the optimum synthesis conditions. J Alloys Compd 2010;491:599e604. [175] Li X, Liu H, Wang J, Cui H, Han F. YAG:Ce nano-sized phosphor particles prepared by a solvothermal method. Mater Res Bull 2004;39:1923e30. [176] Li X, Li Q, Wang J, Yang S. Synthesis of YAG:Eu phosphors with spherical morphology by solvo-thermal method and their luminescent properties. Mater Sci Eng B 2006;131:32e5.

948

[177] Li X, Liu H, Wang J, Cui H, Han F, Boughton RI. Production of nanosized YAG powders with spherical morphology and nanoaggregation via a solvothermal method. J Am Ceram Soc 2004;87:2288e90. [178] Li X, Liu H, Wang J, Cui H, Yang S, Boughton RI. Solvothermal synthesis and luminescent properties of YAG:Tb nano-sized phosphors. J Phys Chem Solids 2005;66:201e5. [179] Wu Z, Zhang X, He W, Du Y, Jia N, Liu P, et al. Solvothermal synthesis of spherical YAG powders via different precipitants. J Alloys Compd 2009;472:576e80. [180] Su LT, Tok AIY, Zhao Y, Ng N, Boey FYC, Woodhead JL, et al. Electronephonon interactions in Ce3þ-doped yttrium aluminum garnet nanophosphors. J Phys Chem B 2008;112:10830e2. [181] Su LT, Tok AIY, Zhao Y, Ng N, Boey FYC. Synthesis and electronephonon interactions of Ce3þ-doped YAG nanoparticles. J Phys Chem B 2009;113:5974e9. [182] Park JY, Jung HC, Raju GSR, Moon BK, Jeong JH, Son S-M, et al. Sintering temperature effect on structural and luminescence properties of 10 mol% Y substituted Gd3Al5O12:Ce phosphors. Opt Mater 2009;32:239e96. [183] Jung HC, Park JY, Raju GSR, Jeong JH, Moon BK, Kim JH, et al. Crystalline structure dependence of luminescent properties of

Handbook of Advanced Ceramics

[184]

[185]

[186]

[187]

[188]

[189]

Eu3þ-activated Y2O3eAl2O3 system phosphors. Curr Appl Phys 2009;9:S217e21. Raju GSR, Jung HC, Park JY, Kanamadi CM, Moon BK, Jeong JH, et al. Synthesis and luminescent properties of Dy3þ:GAG nanophosphors. J Alloys Compd 2009;481:730e4. Park JY, Jung HC, Raju GSR, Moon BK, Jeong JH, Kim JH. Tunable luminescence and energy transfer between Tb3þ and Eu3þ in GYAG:Bi3þ, Tb3þ, Eu3þ phosphors. Solid State Sci 2010;12:719e24. Park JY, Jung HC, Raju GSR, Moon BK, Jeong JH, Kim JH. Solvothermal synthesis and luminescence properties of Tb3þ-doped gadolinium aluminum garnet. J Lumin 2010;130:478e82. Verma A, Uzun O, Hu Y, Hu Y, Han H-S, Watson N, et al. Surface-structure-regulated cell-membrane penetration by monolayer-protected nanoparticles. Nature Mater 2008;7:588e95. Andelman T, Gordonov S, Busto G, Moghe PV, Riman RE. Synthesis and cytotoxicity of Y2O3 nanoparticles of various morphologies. Nanoscale Res Lett 2010;5:263e73. Kominami H, Onoue S, Matsuo K, Kera Y. Synthesis of microcrystalline hematite and magnetite in organic solvents and effect of a small amount of water in the solvents. J Am Ceram Soc 1999;82:1937e40.

Chapter 11.1.5

Supercritical Hydrothermal Synthesis Tadafumi Adschiri,* Seiichi Takami,** Toshihiko Arita,** Daisuke Hojo,* Kimitaka Minami,y Nobuaki Aoki* and Takanari Togashiyy *

World Premier International Research Center-Advanced Institute for Materials Research (WPI-AIMR), Tohoku University, 2-1-1 Katahira, Aoba,

Sendai, Miyagi 980-8577, Japan, ** Institute of Multidisciplinary Research for Advanced Materials, Tohoku University, 2-1-1 Katahira, Aoba, Sendai, Miyagi 980-8577, Japan, y New Industry Creation Hatchery Center, Tohoku University, 6-6-10 Aoba, Aramaki, Aoba, Sendai, Miyagi 780-8577, Japan, yy

Department of Material and Biological Chemistry, Faculty of Science, Yamagata University 1-4-12 Kojirakawa-machi, Yamagata, Yamagata 990-8560,

Japan, Former Affiliate: WPI-AIMR, Tohoku University

Chapter Outline 1. 2. 3. 4. 5.

Introduction Basic Principles of Supercritical Hydrothermal Synthesis Apparatus NP Synthesis by Supercritical Hydrothermal Synthesis Supercritical Hydrothermal Synthesis of Organice Inorganic Hybrid Nanoparticles

949 950 953 956

6. Self-assembly of Hybrid OrganiceInorganic Nanoparticles 7. Hybrid Nanomaterials 8. Summary References

965 972 974 974

957

1. INTRODUCTION The nanomaterials are known for their unique mechanical, chemical, physical, thermal, electrical, optical, electronic, magnetic, and specific surface area properties. Various inorganic and organic compounds have been fabricated in nanoscales as tubes, ribbons, spheres, wires, belts, dots, fibers, dendrites, mushrooms like, etc. Most commonly used methods are physical vapor deposition, colloidal chemistry approach, mechanical alloying, solegel, mechanical grinding, hydrothermal method, laser ablation, chemical vapor deposition (CVD), electrodeposition, plasma synthesis, microwave techniques, etc. In nanotechnology, hydrothermal processing has an edge over the other conventional processes, because it facilitates the issues like simplicity, cost effectiveness, energy saving, use of large volume of equipment (scale-up process), better nucleation control, pollution-free (since the reaction is carried out in a closed system and in “green” water), higher dispersion, higher rate of reaction, better shape control, and lower temperature operation in the presence of an appropriate solvent, etc. The hydrothermal technique has a lot of other advantages such as the following: it accelerates interactions between solid and fluid species; phase pure and homogeneous materials can be achieved; reaction kinetics can be enhanced; the hydrothermal fluids offer higher diffusivity,

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00051-4 Copyright Ó 2013 Elsevier Inc. All rights reserved.

low viscosity, and facilitate mass transport and higher dissolving power. Most important is that the chemical environment can be suitably tailored. However, the high dissolving power is a drawback for the nanoparticle (NP) synthesis, because not only to obtain higher supersaturation degree but also to suppress crystal growth, lower solubility is preferred. To solve the problems for the synthesis of NPs, Adschiri and Arai have proposed to use supercritical state for the hydrothermal synthesis [1e4]. A fluid is supercritical when its temperature and pressure are higher than their critical point values (Tc, Pc). Most of the interesting applications of supercritical fluids occur at 1 < T/Tc < 1.1 and 1 < P/ Pc < 2. Under these conditions, the fluid has some of the advantageous properties of both liquid and gas phases. Further, supercritical hydrothermal synthesis is known for making new materials under mild conditions, since they increase the chemical reactions, improve the mass transfer, and greatly assist in the stability of selected or desired products. The supercritical hydrothermal method with water as a solvent is highly suitable for the NP fabrication of a wide range of high melting-point inorganic compounds and also for hybrid organiceinorganic NPs. One of the promising applications of NPs is for hybrid materials with polymers. The organiceinorganic hybrid

949

950

Handbook of Advanced Ceramics

FIGURE 1 Density of water around the critical point. For color version of this figure, the reader is referred to the online version of this book.

materials have a tremendous application potential as sensors, biological tags, catalysts, bioelectronics, solar energy conversion, mechanics, membranes, protective coatings, transparent pigments, electromagnets, cosmetics, etc. The possibility of combining properties of organic and inorganic materials is an old challenge, which began from the time of industrial era. Some of the earliest and most wellknown inorganiceorganic representatives are those where inorganic pigments or fillers are dispersed in the organic components such as solvents, surfactants, plastics, polymers, etc., to improve optical or mechanical properties of materials. It is to be noted that a major challenge in the nanomaterial fabrication of whether organic, inorganic, or composite NPs is the preparation of highly dispersible unagglomerated NPs with an accurate control over the size and shape, which in turn is directly linked with the nanomaterial processing method. Owing to the higher surface energy of the particles in nanoscale, there is a greater tendency for NPs’ agglomeration and aggregation. At the same time, it is difficult to break down the aggregation of NPs into individual particles size. To disperse the NPs into polymers, surface energy should be reduced by increasing the affinity between NPs and polymers. Thus, surface modification of NPs is of great importance. Therefore, a new strategy for an effective solution to this problem has to be worked out. In this context, the supercritical hydrothermal method can be very effective to fabricate the nanomaterials, especially for hybrid materials, since organic modified inorganic NPs can be synthesized in situ nucleation in supercritical hydrothermal synthesis [3,4]. In this chapter, first the basic principles of supercritical hydrothermal synthesis are introduced. Designing the reactor, especially for mixing method, is explained. After summarizing some specific features of supercritical hydrothermal synthesis method, commercialization of the process is introduced. Then, organiceinorganic hybrid NP synthesis by using supercritical condition is explained. This provides the fabrication of “superhybrid nanomaterials”.

2. BASIC PRINCIPLES OF SUPERCRITICAL HYDROTHERMAL SYNTHESIS Figure 1 shows the density of water around the critical point [5]. The critical temperature and pressure of water are respectively 374  C and 22.1 MPa. Above the critical point, the density of water varies greatly with a little change in temperature and pressure. Because of the drastic change in density, all the fluid properties change greatly around the critical point, including the dielectric constant, which is a controlling factor of reaction rate, equilibrium, and solubility of metal oxides. Therefore, it is necessary to

FIGURE 2 Dielectric constant of water around the critical point. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 11.1.5

Supercritical Hydrothermal Synthesis

understand the basic principles, which insist upon the understanding of the properties of water, including density, dielectric constant (Figure 2) [6], and ion product, varying greatly around the critical point of water and result in a specific reaction atmosphere. The dielectric constant of water at room temperature is 78, and with a raise in temperature at a constant pressure, this value decreases greatly, and around critical point of water, it is nearly equal to the dielectric constant of polar organic solvents. Due to the variation in the properties of water, phase behavior changes greatly around the critical point. Since supercritical water is of high-density steam, light gases like oxygen or hydrogen from a homogenous phase with supercritical water, as shown in Figure 3, Figure 4 shows the critical loci for binary systems of organic compound and water [7]. In the right-hand side of these curves, a homogeneous phase is formed at any composition. The organic compounds and water are not miscible at a lowtemperature range but forms a homogeneous phase at higher temperatures, which is because of the reduced dielectric constant of water. It should be noted that when the pressure is very high (i.e. the density of fluid is as high as of liquid), phase separation occurs, just like liquid watereliquid oil-phase separation. The reaction rate, equilibrium, phase behavior, solubility of metal oxides, and distribution of soluble chemical species change greatly at the critical point range. Various models have been proposed to describe the variation of

951

FIGURE 4 Critical loci for organic moleculeewater binary systems. For color version of this figure, the reader is referred to the online version of this book.

reaction rate or equilibrium over a range of supercritical state [8,9]. The knowledge on the solubility of metal oxides formed and the identification of the ionic species are essential in order to understand the supercritical hydrothermal processes. For example, the main equilibrium reactions for AlOOH dissolution in aqueous solution are given in Table 1. For the estimation of metal oxides, the equilibrium constant K for each reaction is required. There

TABLE 1 Reactions Related to AlOOH Dissolution [3,4] AlOOH þ H2O ¼ Al3þ þ 3OH AlOOH þ H2O ¼ Al(OH)2þ þ 2OH AlOOH þ H2O ¼ Al(OH)þ þ OH AlOOH þ H2O ¼ Al(OH)4 þ OHþ Al3þ þ OH ¼ Al(OH)2þ Al(OH)2þ þ OH ¼ Al(OH)2þ Al(OH)2þ þ OH ¼ Al(OH)3 Al(OH)3 þ OH ¼ Al(OH)4 Al(NO3)3 ¼ Al3þ þ 3(NO3) NaNO3 ¼ Na3þ þ NO NaOH ¼ Naþ þ OH HNO3 ¼ Hþ þ NO3 FIGURE 3 Critical loci for gasewater binary systems. For color version of this figure, the reader is referred to the online version of this book.

H2O ¼ Hþ þ OH

952

Handbook of Advanced Ceramics

are many models to estimate this Helgeson-KirkhamFlowers (HKF) model is very widely used for K at high temperature/pressure, including supercritical state. The following equation is a concept for the estimation of K: ln KT;r ¼ ln KTr ;rr

DH 0T;rr R



1 T

1 Tr



 r) Þ2=3 þ aDuTr ;rr Tr 1 þ R T    DuTr ;rr DuT;r 1 1 1 þ RT RT ε εTr ;rr bð1:0

 1 Tr  1

T0 ¼ 298.15 K a ¼ 6.385  10 5/K r0 ¼ 0.997 g/cm3   0 b ¼ l1 DCr;T þ l Du Tr ;rr þ l3 2 r ;rr where K is the equilibrium constant, T is the absolute temperature, To is the room temperature, R is the gas constant, ε is the dielectric constant, u is a parameter determined by the reaction system, r is the density, and ro is the density at the ambient conditions. The solubility and the distribution of ionic species in solution are also important for the understanding of supercritical reaction atmosphere. Solubility and distribution of ionic species can be predicted by calculating K for each reaction (Table 1), mass balance, and charge balance. There are several models available today to estimate the solubility of the species of interest, based on the dissociation reaction in a given system. Even by using the value estimated from the ionic radius reported in the literature, the solubility behavior can be precisely estimated. Unlike the conventional hydrothermal process, the solubility around the critical point shows unique behavior around the critical point [10]. Figures 5 and 6 show how significantly the solubility of metal oxides and kinetics of hydrothermal synthesis change around the critical point due to the variation in the properties of water. The solubility of CuO in water over a wide range of temperature is shown in Figure 5. The solubility first increases with an increase in temperature and then decreases. This decrease in solubility is due to the lowering of the density and dielectric constant values. An estimation method of metal oxide solubility is described in detail by Sue et al. [11]. Figure 6 shows the Arrhenius plot of the first-order reaction rate constant of hydrothermal synthesis, which was evaluated under supercritical hydrothermal conditions, using a flow-type of apparatus [4,12]. The Arrhenius plot of kinetic constant shows a straight line below the critical point but, above the critical point, increased two orders of magnitude. This is because of the decrease of dielectric constant, and the details of such a mechanism has been discussed elsewhere [13].

FIGURE 5 Solubility of CuO around the critical point. For color version of this figure, the reader is referred to the online version of this book.

FIGURE 6 Kinetics constant for hydrothermal reactions.

These two results imply an important concept of NPs’ synthesis. If the aqueous solution is rapidly heated up to the supercritical state, rapid hydrothermal reaction takes place to form metal oxides; since the solubility of metal oxide under the condition is very low, extremely high supersaturation degree is obtained, which leads to nanoparticle synthesis. However, the rapid heating cannot be achieved by conventional autoclave reactors. To solve the problem, Adschiri and Arai proposed to use a flow-type system in which two flows are mixed to achieve rapid heating. Thus, Adschiri and Arai have pioneered the supercritical hydrothermal synthesis of metal oxides in detail during 1990s, based on this concept [1e4,12e15].

Chapter | 11.1.5

Supercritical Hydrothermal Synthesis

3. APPARATUS 3.1. Basic Concept of Flow-type Reaction System for Rapid Heating Figure 7 shows the typical experimental setup used in this study. An aqueous metal salt solution is prepared and fed into the apparatus in one stream. In another stream, distilled water is pressurized and then heated to a temperature above the desired reaction temperature. The pressurized metalesalt solution stream and the pure supercritical water stream are then combined at a mixing point, at which rapid heating of the metalesalt solution and subsequent reactions in the reactor occur. After the solution leaves the reactor, it is rapidly quenched. In-line filters are used to remove larger particles. The pressure is controlled with a back-pressure regulator. Fine particles are collected in the effluent. The supercritical hydrothermal flow reactor provides NPs with desired shape, size, and composition in the shortest possible residence time. Figure 7b shows the semipilot plant flow reactor (ITEC Co., LTD, Japan).

3.2. Microreactor For the rapid mixing/heating followed by the reactions, nucleation and particle growth, designing the geometry of the mixing point is of great importance [16]; some researchers have developed microreactor reaction systems for the supercritical hydrothermal synthesis. Microreactors are miniaturized reactors that contain microchannels of characteristic dimensions in the order of submillimeters to a few millimeters [17]. Some micromixers, which are mixers containing submillimeter mixing chambers, can also be used as a reactor especially for reactive mixing [18,19]. The reactor miniaturization provides improved mass- and heat-transfer rates owing to a high

953

surface-area-to-volume ratio and thus enables us to proceed with reactions under conditions controlled more precisely as compared with conventional macroscale reactors. This advantage results in improved yields and selectivities of desired products [20]. Enhanced mixing performance in microreactors is also effective to produce desired products in high yield and selectivity. The selectivities of desired products for very rapid multiple reactions have been improved using microreactors [21]. Thus, microreactors enable syntheses with higher efficiency and less by-products than those of conventional macroscale reactors. The efficiency of microreactor can lead to the development of more compact and environmentally benign processes. In this section, hydrothermal syntheses using microreactors are reviewed. The combination of straight tubes and T-shaped mixers such as union tee is an example of simple microreactor system. Dr. Ikushima’s group performed organic syntheses using supercritical water (scH2O), based on such microreactor system [22e24]. The schematic diagram of the system and temperature profile in the system is illustrated in Figure 8. The tubes and the union tee are made of HastelloyC-276. The inner diameter (i.d.) of the tubes and the tee is 300 mm, and the outer diameter (o.d.) of the tubes is 1/16 inch (1.59 mm). In organic syntheses, a desired product is often unstable under the reaction condition. The products tend to change into by-products in quite a short time. For increasing the selectivity of the desired product, the reaction should be quenched immediately after the formation of it. To realize such reaction operation, a reactor that enables rapid mixing and heat transfer and short residence time is required, and the features of microreactor are suitable for this purpose. As shown in Figure 8(b), the scH2O microreactor system can heat up instantaneously an ambient substrate solution to the supercritical state in a molten salt bath and can then quench rapidly to sufficiently low

FIGURE 7 Flow reactor apparatus for supercritical hydrothermal synthesis (ITEC Co., LTD, Japan). For color version of this figure, the reader is referred to the online version of this book.

954

Handbook of Advanced Ceramics

(a) Substrate

Heating bath (Molten salt) Rapid quench (Water/ice bath)

scH2O

Rapid Reaction Heating Time 80%. However, a remarkable disadvantage is the decreased solubility of these precursors compared with the methyl derivative. Consequently, the processing is difficult and polymer yields are low. In the case of the highly crosslinked precursor with R ¼ (NH)0.5, the polymer yield is even 1400  C compared to 1150  C for that of NicalonÔ fibers. For further details, see Section 5.5. TGA of the thermolysis step, which released ceramics in 60% yield, indicated that the thermal decomposition of the polymer proceeds gradually >300  C and is completed at 800  C. The pyrolytic degradation of the polymer does not take place “stoichiometrically” by elimination of methane and hydrogen according to CH 3 1/n

ΔT

Si CH 2 H

SiC + H2 + CH4

(39)

n

since almost 50% of carbon, which is present in the precursor as methyl groups, is retained in the ceramic fibers. Consequently, SiC/C composite fibers without defined microstructures rather than phase-pure SiC fibers are obtained. To yield phase-pure SiC, alternative improved precursors were developed. The idea behind these concepts was to design polymers in which the final elemental composition of the ceramic, that is, the 1:1 stoichiometry of the constituting elements silicon and carbon is already preformed. Moreover, the synthesis of precursors was aspired, which does not produce gaseous byproducts upon thermolysis other than hydrogen. Polymers, which (at least theoretically) fulfill these requirements, are polymethylsilane, [SiMeH]n [24], and polysilaethylene,

[H2SiCH2]n [27,29], which can be obtained according to procedures described in Section 2.1. Although the precursors exhibit a 1:1 Si:C ratio, it is difficult to obtain stoichiometric or near-stoichiometric SiC. The reasons are (i) that the Kumada rearrangement is a radical reaction, which takes place at elevated temperature with low selectivity and (ii) that depolymerization of the precursors during thermolysis occurs with the volatilization of carbon and/or silicon species, which cause changes in the chemical composition of the precursors. IR and NMR investigations of the thermolysis intermediates provide information on chemical reactions that occur during the heat treatment. In the case of [Si(Me)(H)CH2]n, at least four different steps can be distinguished. Below 400  C, low weight oligomers evaporate. Up to 550  C, molecular weight increases due to the formation of new SiC bonds that can be traced back to a dehydrocoupling of SieH and CeH units [166,191]. In contrast, there is no evidence for the formation of SieSi linking. A Further heating to 800  C results in the degradation of residual SieH, CH3, and SieCH2eSi units with evaporation of low molecular weight organic species, that is, molecular hydrogen and methane. This stage corresponds to the organic-to-ceramic transition and releases an inorganic covalent SiC network. Additionally, graphitic carbon forms, which is evident from 13C MAS NMR investigations [191]. Above 800  C, only very small weight loss is observed, which is due to the evaporation of residual hydrogen. Heating to temperatures exceeding 1100  C finally results in the crystallization of the predominantly amorphous ceramic. Details on this topic are presented in Section 4.

Chapter | 11.1.10

Precursor-Derived Ceramics

3.2. Thermolysis of Precursors to Silicon Carbide/Nitride Ceramics Due to the structural diversity of SieCeN precursors compared to SiC polymers, thermolysis seems to be much more complex. Although for SiC usually polysilanes or PCSs serve as preceramics, polysilazanes or poly (silylcarbodiimide)s are mostly used as precursors for ternary SieCeN ceramics. The ternary phase diagram in Figure 8 distinguishes different types of compositions in the SieCeN system. Poly(silsesquiazane)s [192] or poly(silylcarbodiimide)s [13,89,91,92], which are both comparably rich in nitrogen, usually deliver ceramics with compositions located on the tie-line CeSi3N4 (A). In contrast, ceramics derived from polysilazanes with the general structure [(R1)(R2) SieNHe]n (R1, R2 ¼ single-bonded organic unit). [41,51,55,56,167,170,193e197] have compositions located in the three-phase field SiC/Si3N4/C (B). Laine et al. published an exception from this general rule. They obtained Si3N4/C (type A) compositions by thermolysis of [H2SieN(CH3)e]n rather than silicon nitride and silicon carbide-based composites [177]. Compositions located in the three-phase field SiC/Si3N4/Si (C) were obtained by copyrolysis of PCS and PHPS [198e200]. Even though a comparison of the elemental composition of the precursors and their derived ceramics gives first hints for possible reactions that occur during the heat treatment, it is inevitable to structurally characterize thermolysis intermediates accurately. In general, methods such as NMR and IR spectroscopies, elemental analysis, and (mass-spectra-coupled) TGA are considered for this purpose.

3.2.1. Thermolysis of Polysilazanes Most of the work in this field concentrated on thermolysis of polysilazanes. It turned out that the molecular architecture of the polymer backbone, the pendent functional

1053

groups, and the crosslinking of the precursors significantly influence the polymer-to-ceramic conversion. As mentioned in Section 2.3, ammonolysis of dichlorosilanes usually delivers six- or eight-membered cyclosilazanes, [SiR1R2eNH]n (n ¼ 3, 4). Because of their low molecular weight, as-obtained polymers are mostly not qualified as precursors for ceramics. In the worst case, thermolysis of such volatile molecules results in a 100% mass loss. Moreover, cyclosilazanes are generally liquids, which have poor rheological properties, disqualifying them as precursors for ceramic fibers or bulk materials. A procedure for increasing ceramic yields of lowweight polysilazanes is a catalyst-induced crosslinking by dehydrocoupling of SieH and NeH units. For example, Seyferth and Wiseman described crosslinking reactions of cyclotetrasilazane, yielding a highly branched polymer, which is soluble in tetrahydrofuran [52,53]. Details on the procedure including a proposed mechanism are presented in Section 2.3. The increased crosslinking was directly mirrored in increased ceramic yields from 900  C and Si2CN4 (see also Sections 2.3 and 3.2), which is stable up to around 1000  C, have been reported [95]. Crystalline, B SiBn SiB6

B4+δC

SiB3 BN

B2CN2

N

BC2N BC3N BC4N

Si3N4

“C3N4“ SiC2N4

Si2CN4

SiC C

Si

FIGURE 18 SieBeCeN concentration tetrahedron including stable and metastable binary and ternary solid phases in the quaternary SieBeCeN system [244,245]. Note that the compositions BCxN are stoichiometric but not crystalline; C3N4 was calculated in detail [246] but so far not isolated in crystalline form.

phase-pure ternary BeNeC phases are to our knowledge not known so far. However, because of the almost similar lattice parameters of h-BN and graphite (both are layered materials with sheets of trigonally coordinated atoms) they build solid solutions. Stoichiometric (but not phase-pure) compositions of BCN shown in Figure 18 have usually been obtained by the reactions of trichloroborane with N-containing hydrocarbons such as acetonitrile, polyacrylnitrile, CVD processes using boranes, ammonia, and hydrocarbons or by thermolysis of boraneepyridine complexes [247e255]. Moreover, several amorphous compositions have been disclosed in as-thermolyzed PDCs, for example, amorphous carbon, amorphous silicon nitride, and amorphous silicon carbonitride [57,256e260]. A set of consistent thermodynamic data of the stable phases of the quaternary SieBeCeN system has been developed [244,245,261], which enables a detailed description of phase equilibria. With respect to possible applications of precursorderived nonoxide ceramics, SieCeN is the most important subsystem in the quaternary system. It was stated above that such ternary compositions can be obtained by the thermolysis of polysilazanes or poly(silylcarbodiimide)s at 1000e1400  C. As-obtained SieCeN materials are usually amorphous. In other words, they are in a nonequilibrium state, where phase configuration cannot be reproduced by equilibrium calculations. However, spectroscopic methods such as solid-state MAS NMR, and IR spectroscopy (IR) in combination with wide-angle scattering reveal a partial segregation of the amorphous ceramics into structural units that refer to thermodynamically stable phases. At higher temperature, segregation proceeds and the amorphous state is transformed into crystalline composites, consisting of Si3N4, SiC, and/or C, depending on the molecular structure and composition of the precursor and the conditions applied during thermolysis. An important tool for predicting thermally induced decomposition reactions are calculations of phase equilibria by the CalPhaD (Calculation of Phase Diagrams) approach using the set of thermodynamic data mentioned above. A number of thermodynamic calculations in the SieCeN system have already been published [244,245,262e265]. It turned out that the decomposition of such ceramics is to a significant extent controlled by the decomposition reactions of silicon nitride either by a carbothermal reduction according to the following. Si3 N4 þ 3C

1484  C

!

3SiC þ 2N2

ð1013 mbar N2 Þ (50)

or by decomposition into the elements: Si3 N4

1841  C

/

3Si þ 2N2

ð1013 mbar N2 Þ

(51)

Chapter | 11.1.10

Precursor-Derived Ceramics

1067

FIGURE 19 Partial pressure diagram for C/SiC/Si3N4 composites. Of major interest is the dependency of the decomposition temperature of Si3N4 and C with the formation of SiC and N2. At 1 bar p(N2), this temperature is 1484  C (1757 K), whereas at 10 bar p(N2), this value is shifted to 1700  C (1973 K). Assuming 10 4 bar in argon atmosphere, the decomposition of such a composite material should occur at 954  C (1227 K) [265].

It can be concluded from Figure 19 that the onset of the above decomposition reactions strongly depends on the nitrogen partial pressure. In 1 bar N2 atmosphere, the above reactions take place at 1484  C (1757 K) and 1841  C (2114 K), respectively. Increasing the nitrogen partial pressure to 10 bar results in a shift of the decomposition temperature due to carbothermal reduction to as much as 1700  C (1973 K), whereas at 1 mbar, one woulddconsider the results of the CalPhaD computationsdexpect decomposition already at around 1050  C. Thermochemistry studies have also been performed on polymer-derived SieOeC ceramics to assess their thermodynamic stability relative to crystalline phases. Calorimetric measurements of heats of dissolution in a molten oxide solvent indicated that these ceramics possess a negative enthalpy of formation ( 15 to 72 kJ/mol)

relative to their crystalline constituents (Table 1) [266,267]. Therefore, their crystallization resistance does not only rely on kinetic reasons but also on the thermodynamic characteristics. In contrast, calorimetric measurements performed on poly(silylcarbodiimide)-based ceramics provided evidence that amorphous carbon-rich SieCeN PDCs have slightly positive or close to zero heats of formation relative to graphite and the binary crystalline components Si3N4 and SiC (Table 1). Therefore, SieCeN PDCs are less stable than SieOeC ceramics. However, amorphous ceramics have a higher entropy than the corresponding crystalline phases, which may stabilize the amorphous PDCs at high temperatures considering the relatively small (positive) enthalpies of formation in Table 1. Therefore, the experimentally observed thermal stability of the SieCeN ceramics may still reflect at least in part thermodynamic stability.

TABLE 1 Enthalpy of formation from elements as well as from carbon and crystalline binary systems (SiO2 and SiC for SieOeC; Si3N4 and SiC for SieCeN) for one SieCeNe and two SieOeC-based ceramics pyrolyzed at 1100  C in inert atmosphere [268e271] DHf from elements at 25  C [kJ/mol]

Sample

Composition

Si1O1.1C0.73

SiO2 [mol%]

SiC [mol%]

C [mol%]

43.10

34.98

21.92

SiO2 [mol%]

SiC [mol%]

C [mol%]

14.73

25.97

59.30

Si3N4 [mol%]

SiC [mol%]

C [mol%]

Si1O0.72C2.10

Si1C2.80N1.20

14.30

4.80

80.90

DH From binaries and carbon at 25  C [kJ/mol]

242  3.6

53.9  3.6

149.9  4.4

51.9  4.4

50.6  6.4

2.8  6.4

1068

Handbook of Advanced Ceramics

In conclusion, the unique thermal stability of the PDCs and their exceptional crystallization resistance relies on (i) kinetic stabilization by the “free” carbon, which suppresses diffusion processes and (ii) thermodynamics that is, negative enthalpy formation of (amorphous) PDCs compared to their crystalline binary counterparts.

(a)

N VT50

1414°C

NCP 200

1484°C Gas + Graphite + Si3N4

Si3N4

4.3. High Temperature Properties of SieCeN Ceramics As an example to illustrate the influence of chemical composition on high temperature properties, the decomposition behavior of two different SieCeN ceramics obtained by the thermolysis of commercially available PVS VT50 [57] and polymer-hydridosilazane NCP200 [59], which possess different chemical compositions, is briefly discussed. In Figure 20, calculated isothermal sections in the ternary SieCeN system are shown. Figure 20a shows phase equilibria between 1414 and 1484  C. In this temperature range, three three-phase fields exist: C/Si3N4/ Gas, C/Si3N4/SiC, and Si3N4/SiC/Liquid. Below 1414  C (melting point of silicon), the latter is composed of Si3N4/ SiC/Si. The composition of VT50 ceramic (symbolized by a triangle) is located on the tie-line Si3N4eC. Presuming quantitative segregation into the pure (crystalline) phases, the sample is composed of Si3N4 and graphite. In contrast, the composition of NCP200 ceramic is located in the three phase fields C/Si3N4/SiC. Accordingly, a fully crystalline sample would be composed of SiC, Si3N4 and graphite. CalPhaD predicts that raising the temperatures to >1484  C (Figure 20b) initiates the carbothermal reduction of Si3N4 according to Eqn (50). The expected quantitative loss of nitrogen in the case of VT50-derived ceramics should result in the formation of a SiC/C composite. In contrast, thermally induced degradation of ceramics obtained from NCP200 (symbolized by a rhombus) should deliver a composition, which is located on the tie-line Si3N4/SiC. The calculations are confirmed by XRD of annealed ceramic samples. A comparison of XRD patterns of VT50 and NCP200-derived ceramics after annealing at 1800  C in a nitrogen atmosphere (5 h, 1 bar) is shown in Figure 21. Whereas in the XRD pattern of annealed VT50 ceramic only b-SiC reflections appear, the pattern of annealed NCP200-derived ceramic exhibits both a-Si3N4 and b-SiC reflections. As expected, an increase of the temperature to 2000  C (not shown in the figure) results in the dissociation of a-Si3N4 into the elements. After cooling to room temperature, XRD patterns of the respective sample now clearly show a-Si reflections besides those of b-SiC. The contrasting crystallization of VT50 and NCP200 ceramics is also reflected in their different mass-loss

Si3N4 + SiC + Liquid

SiC + Si3N4 + Graphite

C

Si

SiC

(b)

N

1484°C Gas + SiC + Si3N4

1841°C

Si 3N 4

Si3N4 + SiC + Liquid

SiC + Gas + Graphite

C

Si

SiC

N

(c) 1841°C

SiC + Gas + Graphite

C

SiC + Gas + Liquid

SiC

Si

FIGURE 20 Ternary Si/C/N phase diagrams including compositions of ceramics derived from VT50 (triangles) and NCP200 (rhombs) at 1 atm N2 and various temperatures: (a) 1414  C < T < 1484  C; (b) 1484  C < T < 1841  C; (c) T > 1841  C. Decomposition of as-obtained VT50 ceramic >1484  C occurs in one step delivering SiC/C composites, whereas decomposition of NCP200-derived ceramic proceeds in two steps [57].

Chapter | 11.1.10

Precursor-Derived Ceramics

1069

Si3N4. Due to the carbothermal reduction of Si3N4 at 1484  C, graphite is quantitatively consumed and bonded as SiC. Correspondingly, it is expected that the amount of Si3N4 is reduced to 38 mass-% with the formation of 12 mass-% of gas (N2). The second decomposition step at 1841  C consumes residual Si3N4. As a consequence, additional 15 mass-% of gas form along with 22% of a liquid (Si). As expected, the amount of SiC is not affected by this reaction.

4.4. High Temperature Properties of SieBeCeN Ceramics FIGURE 21 XRD patterns of VT50 and NCP200-derived ceramics after annealing to 1800  C in 0.1 MPa N2 for 5 h [269].

behavior. A suitable method for investigating mass lossaccompanied degradation is high temperature TGA. Figure 22 shows a comparison of the TGA of VT50 (top) and NCP200 ceramics (bottom). It is evident from the TGA curves that VT50 decomposes in a one-step reaction with the loss of 30% of its original sample weight. The determined value exactly fits with the nitrogen content of the ceramic determined by elemental analysis and confirms the results of the CalPhaD calculations shown in Figure 20. The onset of decomposition, which is at around 1500  C, is in accordance with the thermodynamic calculations. The corresponding DTA curve displays a strong endothermic peak, centered at around 1665  C. In contrast, NCP200-derived ceramics decompose in a two clearly separated steps. The onset of the first decomposition step caused by a carbothermal reduction of Si3N4 is at around 1550  C and thus shifted by 60  C to higher temperature compared with the result of the calculations. The second step is caused by Si3N4 dissociation into the elements and starts at around 1800  C. The mass loss of the individual decomposition steps is 12 and 15 mass-%, respectively. The total of 27 mass-% exactly matches the nitrogen content of the ceramic material. In addition, the values of the mass-loss of the single steps are purported by the amount of “free carbon.” It is interesting to note that indeed all as-thermolyzed ceramics with an overall composition within the three-phase field C/Si3N4/SiC decompose like VT50 or NCP200-derived ceramics in one- or two-step reactions, only depending on whether their C:Si ratio is >1 or 3c > 4c > 5c. This conclusion is confirmed by high temperature TGA as shown in Figure 26.

0 5c

weight change/%

-5

4c

-10

3c

-15

-20

2c 1c

-25 1000

1200

1400

1600

1800

2000

2200

temperature/°C FIGURE 26 Comparison of the thermal behavior of quaternary SieBeCeN ceramics with different boron concentration by high temperature TGA. Heating rate: 5  C/min, 1 bar argon. For details concerning the molecular structure of the polymeric precursors and the elemental and phase composition of the derived ceramics, consider Scheme 27 and Tables 2 and 3, respectively.

Chapter | 11.1.10

Precursor-Derived Ceramics

1073

1800°C α/β-SiC

+ 2c

β-Si3N4

+

+ +

++

++ + + + ++ + +

+

3c 4c 5c

10

20

30

40

50

60

70

80

2Θ FIGURE 27 Partial pressure diagram for C/SiC/Si3N4/BN composites. In contrast to the ternary C/SiC/Si3N4 system (cf. Figure 19), additional carbothermal reduction of boron nitride must be considered in the high temperature chemistry of the ceramic composites [133,269].

2000°C α/β-SiC

+ As expected, boron-free 1c starts to decompose at around 1550  C and loses 24% (¼Dmo) of its original sample weight. This is in accordance with the thermodynamically calculated (predicted) value of 25% (Dmp) given in Table 3. The incorporation of even small amounts of boron such as in 2c results in a shift of the onset of decomposition by 80  C to 1630  C. However, the total mass loss of 19% corresponds to the expected value (Dmp ¼ 20%). Further increased boron contents in 3c and 4c result in an additional shift of the decomposition temperature to around 1700  C. Moreover, a significant deviation of found (3c: Dmo ¼ 13%, 4c: Dmo ¼ 8%) and calculated (3c: Dmp ¼ 18%, 4c: Dmp ¼ 17%) mass losses is observed. Compound 5c, which possesses the highest boron content, shows no signs of decomposition in the expected temperature range. Thermal degradation is significantly retarded and starts at around 2000  C. At this temperature, carbothermal reduction of boron nitride is expected: 4BN þ C / B4 C þ 2N2

(52)

In analogy to the carbothermal reduction of silicon nitride (cf. Eqn (50)), the onset of BN decomposition also depends on the nitrogen partial pressure (Figure 27). The different behavior in the high temperature TGA shown in Figure 26 is mirrored in the crystallization/phase composition of the materials after an additional heat treatment at elevated temperature. For example, after heating to 1800  C (1 bar argon) for 5 h, XRD patterns of 2c and 3c correspond to the diffraction pattern of silicon carbide (Figure 28). This is in agreement with the high temperature TGA investigation that indicated quantitative (2c) or almost quantitative thermal degradation (3c) of silicon nitride at this temperature.

β- Si 3N 4 C

2c

+

+ +

++

+ ++ + + + ++ + +

3c 4c 5c

10

20

30

40

50

60

70

80

2Θ FIGURE 28 XRD patterns of the ceramics obtained from 2ce5c (cf. Scheme 27) after annealing to 1800  C (top) and 2000  C (bottom) in 1 bar argon for 5 h each [138]. It is evident that after heating to 1800  C in the case of less thermally stable materials 2c and 3c (cf. Figure 26), only reflections of silicon carbide are observed, whereas in the case of ceramics derived from 4c and 5c, additionally silicon nitride reflections are found. In the case of 4c, the latter vanish after annealing the material at 2000  C.

In the XRD patterns of the boron-richer materials 4c and 5c, b-Si3N4 reflections are distinctly detected besides SiC peaks. Their line widths suggest a nanocrystalline arrangement. At 2000  C, the thermodynamically expected cleavage of SieN bonds in 4c proceeds, resulting in the absence of Si3N4 reflections in the XRD diagram. In contrast, the diffraction pattern of 5c, which was shown by high temperature TGA to be the only thermally stable composite at this temperature, still represents b-Si3N4 reflections. In addition, all samples heat treated to 2000  C exhibit an additional broad reflection at 26.6 , which is a characteristic of graphitic carbon. Investigations on the influence of the nitrogen content on the thermal stability on SieBeCeN ceramics were also published [150]. Therefore, boron-modified poly (silylcarbodiimide)s with adjustable nitrogen content were

1074

Handbook of Advanced Ceramics

synthesized by a dehydrocoupling reaction of tris (hydridosilylethyl)boranes, B(C2H4SiRH2)3 (R ¼ H, CH3), and cyanamide (cf. Scheme 16) [151]. Investigations of the polymer-to-ceramic conversion of the precursors in an Ar atmosphere revealed that they deliver SieBeCeN ceramics in 65e85% yield, depending on the elemental composition and the molecular structure, that is, the crosslinking of the polymers. For example, TGA of a series of polymers obtained by the reaction of B(C2H4SiH3)3 [140] with different amounts of H2NeCN is shown (Figure 29). The calculated chemical composition of the precursors (idealized sum formula) is given in Table 4. Thermolysis of HeN1 e HeN6 releases ceramics in a 75e84% yield. The polymers with the lowest degree of crosslinking HeN2 and HeN3 have the highest ceramic yields of approximately 84%. The more highly crosslinked precursors HeN4 and HeN5 deliver ceramics in an 81%

TABLE 4 Idealized sum formulas of boron-modified silylcarbodiimide HeN1 and poly(silylcarbodiimide)s HeN1 e HeN6 obtained from B(C2H4eSiH3)3 and H2NeC^N [151] Silane: cyanamide

Sum formula (monomer unit)

HeN1

1:0.5

C6.5H20NSi3B

HeN2

1:1

C7H19N2Si3B

HeN3

1:1.5

C7.5H18N3Si3B

HeN4

1:2

C8H17N4Si3B

HeN5

1:2.5

C8.5H16N5Si3B

HeN6

1:3

C9H15N6Si3B

H-N3

Weight Change [%]

0 -5 -10

H-N2 H-N3.5

-15

H-N2 H-N1

-20

H-N5 H-N6

-25

250

500

750

1000

H-N4

1250

Temperature [°C] FIGURE 29 TGA of HeN1 e HeN6 (cf. Scheme 15). Heating rate 2  C/min, argon atmosphere [158].

and 80% yield, respectively, whereas the most highly crosslinked precursor HeN6, delivers ceramics in a 75% yield. Mass-spectra-coupled TGA revealed that neither silicon-, boron-, nor nitrogen-containing species evaporated during thermolysis. It can thus be concluded that the Si:B:N ratio which was adjusted in the precursors was maintained in the ceramics. High temperature properties of ceramics derived from HeN1 e HeN6 can clearly be correlated with their nitrogen content. Ceramics obtained from HeN5 and HeN6, which possess the highest nitrogen content among the materials investigated, decompose earliest at around 1670  C and finally loose 25% and 30% of their original sample weight. The decomposition temperature of HeN4 is shifted by 50  C to 1720  C, the final mass loss of 10% is significantly lower than that of HeN5 and HeN6. Compound HeN2, HeN3, and HeN3.5 behave like 5c ceramics (cf. Figure 26). Neither of these materials decomposes below 1950  C. The difference in the thermal mass loss behavior is again reflected in the different crystallization behavior of the ceramic materials. For example, the phase evolution observed by XRD of HeN2 and HeN5, which were annealed under similar conditions, is shown in Figure 31. As-obtained ceramics (1400  C) are predominantly amorphous. However, in the patterns of HeN2 ceramics, broadened reflections at 2Q ¼ 36 and 61 point to the presence of nanocrystalline silicon carbide. The appearance TABLE 5 Composition of ceramic materials HeN2 e HeN6 (a) weight-% and (b) calculated phase fraction in mol-% [mass-%]

Si

B

C

N

HeN2

37.9

6.0

37.2

18.9

HeN3

37.8

5.6

36.1

20.5

HeN3,5

37.2

4.8

35.7

22.3

HeN4

35.5

4.8

35.1

24.6

HeN5

33.1

4.5

35.3

27.1

HeN6

31.5

4.9

32.2

31.4

SiC

[mol-%]

BN

Si3N4

HeN2

17.5

21.9

23.7

36.9

HeN3

16.7

23.4

23.2

36.7

HeN3,5

14.0

31.8

14.6

39.6

HeN4

13.9

36.0

8.7

41.4

HeN5

12.9

41.1

1.2

44.8

HeN6

14.0

48.2

L6.8

44.6

C

Chapter | 11.1.10

Precursor-Derived Ceramics

1075

of this XRD pattern remains unchanged, even after heating the sample to 1800  C indicating that there is no further crystallization up to this temperature. After annealing to 1900  C, reflections of both a-SiC and b-Si3N4 appear. Remarkably, the latter do not decrease in intensity even after heating the samples to 2000  C. The crystallization behavior of HeN5 ceramic is significantly different and reflects the decomposition behavior observed by TGA shown in Figure 30. No crystallization of phase formation is observed 1200 2  C/min, 1 bar argon atmosphere. For details concerning the molecular structure of the polymeric precursors and the elemental and phase composition of the derived ceramics, consider Scheme 16 and Table 5, respectively [150].

FIGURE 31 XRD patterns of the ceramics obtained from HeN2 and HeN5 (cf. Scheme 16) annealed at 1400e2000  C (100  C steps, 1 bar N2) each for 3 h [150].

1076

Handbook of Advanced Ceramics

N Thermally stable Thermally less stable

G + Si3N4 + SiC + BN C + Si3N4 + SiC + BN

Si3N4 + C + BN SiC + C + BN

Si + Si3N4 + SiC + BN SiC + C + BN + B4C

SiC + BN + B4C + L

Si

C

FIGURE 32 Calculated phase equilibria in the quaternary SieBeCeN system at 10 atom-% boron and a total pressure of 1 bar. Thermally stable SieBeCeN ceramics with compositions in the four-phase system C/Si3N4/SiC/BN are usually poor in nitrogen, whereas thermally less stable materials are often located close to the three-phase field system C/Si3N4/BN (on the tie-line CeSi3N4).

crystallization of such composites delivers both silicon nitride and silicon carbide. Detailed studies on the phase evolution of amorphous SieBeCeN ceramics by solid-state NMR using MAS were published by Schuhmacher et al. [203,204]. They compared the crystallization behavior of high temperature stable SieBeCeN ceramic, which was obtained from [B(C2H4SiHeNH)3]n (MW33 [133]) with that of a less stable material, derived from nitrogen-rich boron-modified poly(silsesquiazane) [B(C2H4Si(NH)1.5)3]n (MW60 [133]). The representative 29Si MAS NMR spectra given in Figure 33 cover the temperature range from 1050 to 2000  C. After thermolysis to 1050  C, 29Si NMR spectra

of both ceramics appear almost similarly (the NMR spectrum of the 1050  C MW60 sample is not shown). They are each characterized by a broad resonance centered at around 40 ppm. Apart from a line narrowing in the spectrum of MW60-derived ceramic, further heat treatment to 1400  C causes only minor changes in the signal appearance. The chemical shift range in combination with the broadened resonance signals at and below this temperature point to the presence of a mixture of SiCxN4 x sites (x ¼ 1e4), suggesting a fully amorphous structure. After heating to 1600  C, significant progress in the phase evolution of both materials occurs. In the case of MW33-derived material, two new resonance signals centered at 19 and 49 ppm MW60

MW33 SiC4

SiN4

SiC4

FIGURE 33 Experimental 29Si NMR spectra of ceramics derived from boron-modified polysilazanes [B(C2H4SiHeNH)3]n MW33 (left) [133] and boron-modified poly(silsesquiazane) [B(C2H4Si(NH)1.5)3]n MW60 (right) [133] after annealing up to finally 2000  C (atmosphere: 1 bar nitrogen). All spectra were obtained under MAS conditions and single pulse excitation [202,204].

Chapter | 11.1.10

Precursor-Derived Ceramics

evolve, which indicate the formation of SiC4 and SiN4 sites, respectively. However, a broad signal, which appears in between, still points to a remarkable amount of SiCxN4 x sites. In contrast, annealing of MW60 ceramic to 1600  C causes the disappearance of high field shifted signals. The only resonance signal in the 29Si NMR spectrum of MW60 ceramic is found at 19 ppm and suggests quantitative degradation of SieN motifs due to a carbothermal reduction with the formation of silicon carbide. Heating MW33 ceramics further results in additional demixing. As a consequence, only two, well-separated signals remain after heating to 1800  C, which exhibit considerably smaller line widths than those obtained from ceramics annealed at lower temperatures. It was claimed that this is a consequence of the presence of SiC and Si3N4 crystallites. Obviously, all structural components with silicon in mixed coordination, such as SiCxN4 x sites, are disintegrated at this temperature, as also found by X-ray studies on the same samples. Even though obtained from structurally different precursors, the crystallization behavior of ceramics obtained from MW33 and MW60 shown in Figure 34 is comparable with that obtained from boron-modified polysilylcarbodiimides depicted in Figure 31. Ceramics with “high” nitrogen contents decompose around 1550e1600  C with crystallization of silicon carbide, whereas in the case of ceramics with “low” nitrogen content, neither thermal degradation nor crystallization is observed 1100  C the SieOeC matrix starts to phase separate to silica, silicon carbide, and excess carbon, whereas the metal oxide nanoparticles crystallize to t-ZrO2 and t-HfO2; (iv) >1400  C a solid-state reaction between silica and the metal oxide phase occurs leading to the formation of a crystalline MSiO4 (M ¼ Zr, Hf) phase in the form of well-dispersed nanoparticles (20e30 nm) throughout the amorphous PDC matrix (Figure 37b) [163,273]. Systematic studies on the thermal stability of SieOeC/ ZrO2 and SieOeC/HfO2 ceramic nanocomposites upon annealing at T >> 1100  C have shown that the incorporation of the metal oxide phase within SieOeC strongly suppresses decomposition processes (e.g. carbothermal decomposition) even at temperatures as high as 1600  C. In contrast to the SieOeC sample, which exhibits a mass loss of 48.8% if annealed for 5 h in Argon atmosphere at 1600  C, significantly lower mass losses are recorded for the Hf-modified samples (Table 6). The modification of the polysiloxane with 30 vol% of zirconium or hafnium alkoxide leads to SieOeC/ZrO2 and SieOeC/HfO2

TABLE 6 Mass loss of SieOeC, SieOeC/HfO2 (two different contents of hafnia, as 10 vol% and 30 vol% hafnium alkoxide has been used for the chemical modification PMS) and SieOeC/ZrO2 (two different contents of zirconia, similar to SieOeC/HfO2) upon annealing at 1300, 1400, and 1600  C for 5 h in an argon atmosphere Mass Loss [%] Sample

1300  C

1400  C

1600  C

SieOeC

0.60

1.00

48.80

SieOeC/HfO2 (10 vol%)

1.19

1.88

18.28

SieOeC/HfO2 (30 vol%)

0.22

1.33

8.56

SieOeC/ZrO2 (10 vol%)

e

2.2

65.7

SieOeC/ZrO2 (30 vol%)

e

0.1

4.7

ceramic nanocomposites with mass losses of 4.7 wt% and 8.6 wt%, respectively, if annealed under similar conditions [163,273]. The strong improvement of the thermal stability of the ceramic nanocomposites by the incorporation of zirconia or hafnia into the SieOeC matrix relies on the suppression of the carbothermal reaction of silica with carbon. Instead, a reaction between silica and metal oxide occurs and leads to the formation of MSiO4 (see Scheme 28, as for SieOeC/ HfO2) [273]. SCHEME 28 Possible processes occurring in SieOeC/HfO2 ceramic nanocomposites upon annealing at 1600  C.

1080

Handbook of Advanced Ceramics

Preceramic Compound

- Ceramic/ Metal Powder + Preceramic Compound

- Milling Sieving Shaping

- Milling Sieving - Precursor aerosols - Gas phase Pyrolysis

- Melt Spinning - Dry Spinning

- Dip Coating - CVD

Infiltration

Ceramization

Monoliths

Sintering

Powders

Fibers

Infiltrated Material

Coatings

FIGURE 38 Preparation of ceramic materials by precursor processing.

The carbon contents measured in the samples obtained upon annealing at temperatures >1300  C (as for SieOeC/HfO2, see Table 7) support the fact that the carbothermal reaction of silica is suppressed [273]. Thus, the carbon content was shown to remain practically constant upon annealing, indicating that it is not consumed on reaction with silica.

5. APPLICATIONS

TABLE 7 Carbon and oxygen content of hafniamodified silicon oxycarbide (SieOeC/HfO2) annealed at 1300, 1400, and 1600  C Temperature

C [wt%]

O [wt%]

1300  C

9.78

41.60

1400  C

10.66

40.71

1600  C

9.70

41.35

5.1. General Comments The wide variety of available organometallic polymers, suitable as preceramic materials, offers exceptional opportunities for the development of PDCs. According to Figure 38, bulk parts, fibers, coatings, infiltrated media, fiber matrix composites, and near net shape manufactured ceramic components are available [3,243,274e276]. The preformsdalso referred to as green partsdcan be produced using techniques well known from polymer process engineering. The different approaches and techniques illustrated in Figure 38 will be discussed in the following sections topologically.

5.2. Monolithic Ceramics Bulk ceramics can be obtained by polymer powder compaction and subsequent ceramization according to the flow scheme shown in Figure 39. In contrast to conventional sintering processes, no sinter additives are required for the densification. Powder compaction is usually achieved by cold, warm, or hot isostatic or uniaxial pressing,

Polymer

Crosslinking Crosslinked Polymer

Homogenization, Powder Compaction Polymeric Monolith

Thermolysis Ceramic Monolith FIGURE 39 Flow scheme for the preparation of precursor-derived bulk ceramics.

Chapter | 11.1.10

Precursor-Derived Ceramics

depending on the physical/chemical properties of the precursor. A key step is the initial transformation of volatile or liquid precursors into highly crosslinked preceramic infusible networks. PDCs suffer from a high volumetric shrinkage (>50 vol percent) and residual porosity that evolves during pyrolysis of the preceramic polymers. This is a consequence of the loss of organic substituents and an increasing density from 1 g/cm3 (precursor) to approximately 3 g/cm3 (ceramic). Thus, the conversion into ceramic is accompanied by the generation of mechanical stresses that lead to defects and cracking of the ceramic parts [277]. Many studies in the last two decades were directed to reduce shrinkage and porosity by optimizing precursor compositions and finding suitable processing conditions, for example, with respect to crosslinking and ceramization. Fillers can be used to minimize the overall shrinkage (Figure 40). There are two types of fillers for this purpose. Passive fillers serve as a space holders, they are not involved in reactions during pyrolysis (their size and composition remain constant). In contrast, active fillers react during pyrolysis either with the polymer matrix, the decomposition volatiles, or the pyrolysis atmosphere, thus changing/tailoring the shrinkage and the composition of the final ceramic [184,185]. Inert fillers are, for example, metal oxides, carbides, or nitrides (e.g. Al2O3, SiO2, Y2O3, SiC, B4C, Si3N4, BN). The volumetric shrinkage of a system consisting of a preceramic polymer and an inert filler (uV(IF)) can be described as uV ðIFÞ ¼ ð1 VF =VF* ÞuV ðPÞ, where VF is the volume fraction of the inert filler, VF is the critical volume fraction of the inert filler and uV(P) represents the volumetric

1081

shrinkage of the preceramic polymer. The critical volume fraction of the inert filler is the volume fraction at which the filler particles start to form a rigid network, that is, further shrinkage of the system is not possible. Hence, at volume fractions VF, the total system shrinkage is zero. Active fillers are mainly used to produce near net shape parts because they nearly compensate the shrinkage of the precursors during pyrolysis. Usually, reactive powders are used, such as pure metals, intermetallics, metal hydrides or metal carbonyl complexes. The active filler usually reacts with the decomposition products or with the reactive pyrolysis atmosphere forming new phases that expand in volume. Therefore, the volumetric shrinkage of a system consisting of a preceramic polymer and an active filler, uV(AF), can be described as: uV ðAFÞ ¼ ½ð1 VF =VF* ÞuV ðPފ þ VF uV ðFÞ, where uV(F) is the shrinkage/expansion value resulting from the filler transformation reaction. Thus, the critical volume fraction of an active filler, VAF, to achieve net zero shrinkage in the bulk composite part is defined * ¼ u ðPÞ=ðu ðPÞ=V * Þ uV ðAFÞ [184,185]. as VAF V V F The near net shape processing of PDC parts using active fillers has been denoted by Greil as Active Filler Controlled Pyrolysis [185]. Composite systems with near zero shrinkage can be designed with both passive and active fillers. However, in the case of active fillers, lower filler volume fractions are necessary to achieve zero shrinkage. The use of filler particles not only reduces shrinkage during pyrolysis but may also endow tailored properties to the composite. In this way, special properties such as mechanical strength, thermal and electrical conductivity, magnetism, or surface features can be adjusted.

FIGURE 40 Dense and crack-free green body (left) and ceramic part (right) prepared by crosslinking and pyrolysis of an alumina-filled polysilsesquioxane (reprinted with the permission of Elsevier) [278].

1082

Handbook of Advanced Ceramics

FIGURE 41 TMA of cold isostatically pressed green bodies obtained from a PVS [281].

20 Open Porosity [%]

The polymer powder compaction process for the preparation of PDC bulk parts by cold isostatic pressing was developed by Riedel et al. [60,279,280] in the early 1990s. The final density of ceramic materials obtained by this procedure was in the range of 90% maximum. For obtaining dense monolithic bodies, however, warm pressing of suitable (nonvolatile) precursors is the state of the art. Warm uniaxial pressing as described by Seitz and Bill of PVS VT50 [281] delivered glassy green bodies with 7.5% of open porosity. Subsequent thermolysis released ceramic bulk parts with relative densities of around 97%. Optimized conditions for this process were determined by means of thermomechanical analysis (TMA). Figure 41 shows such a diagram obtained from a cold-isostatically pressed PVS body (precursor: VT 50, Hoechst AG) [281]. TMA investigations give information on the shrinkage (or expansion) of materials upon heat treatment. If it is performed in an oscillating modus, it additionally allows for the determination of the softening temperature of the polymeric precursor (cf. inset in Figure 41). This value is of major importance because an increasing temperature results in a decrease in viscosity of the polymeric precursor, which enables a facilitated powder compaction. On the other hand, the maximum temperature applied for the powder compaction must be considered carefully. Exceeding this value (which correlates to the maximum of the TMA curve) results in the shrinkage of the green part due to a beginning ceramization during compression, causing the formation of shrinkage cracks which usually appear parallel to the pressing direction. As a consequence, shape integrity is lost, and in the worst case, the monoliths break into pieces. Haug et al. investigated in detail plastic forming of boron-modified poly(silylcarbodiimide)s for the production of SieBeCeN ceramic monoliths and correlated the conditions applied for the powder compaction with the open porosity of the derived bulk parts [282]. It turned out that the open porosity of bulk ceramics obtained from green

100°C

15

10 120°C

5

0 30 35 40 45 Pressure applied during Warm Pressing [MPa] FIGURE 42 Open porosity of ceramics derived from boron-modified poly(silylcarbodiimide) [B(C2H4Si(CH3)NCN)3]n [146,147]. Powder compaction was performed by plastic forming of the polymer powder at different temperatures and pressures and subsequent pyrolysis at 1400  C for 2 h [282]. For color version of this figure, the reader is referred to the online version of this book.

bodies prepared at 100  C/20 MPa could be decreased from about 20% to 12% by increasing the applied pressure to 43 MPa. If the same polymer was plastically formed at 120 , the porosity decreased from 6.6% at 20 MPa to 5% if 43 MPa were applied for the densification of the polymeric powder (Figure 42). It is obvious that the temperature applied during plastic forming of the precursors has more influence on the open porosity than the pressure applied. However, in the above case, further raising the temperature did not result in an additional reduction of the open porosity, because at and above 150  C, shape integrity was totally lost as is evident from Figure 43 [282]. The absence of low-melting grain boundary phases in precursor-derived materials leads to extraordinarily high temperature mechanical properties that can be very different from those of conventionally sintered ceramics.

Chapter | 11.1.10

140°C

Precursor-Derived Ceramics

150°C

FIGURE 43 Limits in the densification of bulk ceramics derived from boron-modified poly(silylcarbodiimide) [B(C2H4Si(CH3)NCN)3]n [146] by means of plastic forming. After polymer powder compaction at 140  C, sufficient open porosity is provided for releasing gaseous thermolysis byproducts, whereas after compaction at 150  C, the green body is too dense to efficiently release such gases [282].

Recent studies of the high temperature mechanical properties of as-thermolyzed SieCeN and SieBeCeN materials have shown that the amorphous state reveals an outstanding mechanical stability, even at rather high temperatures [283e285]. However, the deformation mechanisms are not yet well understood. Christ et al. characterized amorphous SieBeCeN ceramics using isothermal compression creep testing in the temperature range of 1200e1500  C [283,284]. It was found that the deformation rate contains a stress-independent section as well as a stress-dependent component that is proportional to the applied stress, indicating that this portion of the deformation mechanism is based on viscous flow. As shown in Figure 44 for SieBeCeN ceramics derived from [B(C2H4SiHeNH)3]n (MW33) and [B(C2H4SiMeeNH)3]n (T2-1), a continuous decrease in the deformation rate from approximately 10 6 to 10 8 s 1 with time was found, which was mainly explained with a reduction of free volume in the amorphous material. Stationary creep was not observed, even after 450 h, indicating that the materials improve their creep stability during the creep tests: It is obvious and interesting to note that the curves of both materials can be divided into two ranges. The first one referred to as range I up to approximately 210 5 is similar

FIGURE 44 Deformation rates of amorphous ceramics derived from [B(C2H4SiHeNH)3]n (MW33) [133] and [B(C2H4SiMeeNH)3]n (T2-1) [124] during isothermal creep testing at 1400  C under a load of 100 MPa [283]. Atmosphere: air.

1083

to primary creep in conventional materials, which can often be described using the Norton power law [286]. Range II is characterized by an accelerated decrease of the deformation rate as compared to range I. Christ et al. [283,284] as well as Thurn et al. [287,288] also observed that the temperature dependence of the creep behavior of both SieCeN and SieBeCeN ceramics is comparably low and qualitatively the same for all temperatures. For example, the influence of the temperature on the deformation rate of T2-1-derived SieBeCeN ceramic is shown in Figure 45. Among the investigated samples, deformation rate was the lowest at 1350  C and the highest at 1500  C. The range I to range II transition occurred earlier at elevated temperatures. This shift was explained by a stress-independent reduction of free volume, which proceeds diffusioncontrolled and thus accelerates with increasing temperature. All creep tests mentioned above were performed in air and thus represent long time oxidation experiments. Once tested, the specimen usually possessed porous oxide surface layers consisting of silicon and oxygen as determined by energy-dispersive X-ray spectroscopy (EDX) and wavelength dispersive X-ray spectroscopy (WDX). Although the thickness of such oxide layers increased with testing time, a saturation value was reached. Although after short time exposures, even up to 1700  C, the thickness of the surface layers was 0.3 nm such as CO, CO2, or CH4 is suppressed. This emphasizes the great potential of these membranes in applications related to the hydrogen purification. An amorphous-silica-based membrane was also synthesized by pyrolysis in air of a polysilazane deposited on a silicon nitride porous support. It exhibited a hydrogen permeability of 1.3  10 8 mol/m2 s Pa at 300  C and a H2/N2 selectivity of 141, which is comparable with the selectivity of other amorphous-silica- or silicon oxycarbide-based membranes [319]. Amorphous ceramic membranes were also prepared in nonoxide systems, such as SieC, SieN, SieCeN, and SieBeCeN. The possibility of a molecular sieve amorphous SiC ceramic membrane was first demonstrated for a polysilastyrene derived composite membrane on a porous Vycor glass [320]. Amorphous silicon carbide ceramic membranes were prepared by thermal [317], e-beam [321], or chemical [322] curing of PCSs followed by pyrolysis in inert atmosphere. Hydrogen-selective SieOeC-based membranes were prepared via air curing of PCSs and subsequent pyrolysis in argon [323e325].

1086

FIGURE 49 Polymer-derived SieBeCeN ceramic membrane deposited on a porous alumina substrate [327].

Amorphous silicon-nitride-based ceramic membranes were prepared by the pyrolysis of a polysilazane in ammonia atmosphere at 650  C. The as-synthesized membrane showed a hydrogen permeability of 1.3  10 8 mol/m2 s Pa at 200  C and a H2/N2 selectivity of 165, whereas after hydrothermal treatment at 300  C, the permeability was >1.0  10 7 mol/m2 s Pa at 300  C with H2/N2 selectivity beyond 100. Recently, novel polymer precursors have been synthesized for the preparation of PDC membranes in the systems SieCeN and SieBeCeN. The SieCeN-based ceramic membrane was prepared via pyrolysis of a novel polysilylcarbodiimide precursor, which was synthesized via a nonoxide solegel process based on reactions of bis (trimethylsilyl)carbodiimide with chlorosilanes. Nitrogen sorption isotherm analysis of a 500 nm SieCeN layer deposited on a porous support surface indicated the existence of pores with sizes in the range 2e5 nm [326]. Amorphous SieBeCeN-based ceramic membranes with a thickness of approximately 1.75 mm were prepared by dip coating of a polyborosilazane on a macroporous alumina support, followed by pyrolysis in inert atmosphere (Figure 49) [327]. Pore analysis revealed a trimodal distribution of the pore sizes with maxima at 0.6, 2.7, and 6 nm. Materials based on SieCeN or SieBeCeN are highly promising amorphous nonoxide ceramics that can withstand extremely high temperatures and are, therefore, of great interest for hot gas separation or filtration applications.

5.5. Ceramic Fibers The fabrication of high-modulus refractory ceramic fibers requires controllable polymer rheology and adjustable polymer reactivity. Precursor fibers are obtained by either melt or dry spinning. Different principles for fiber processes such as extrusion, downdrawing from preforms,

Handbook of Advanced Ceramics

or updrawing from precursor melt or solutions are known [328]. Once spun, the precursor fiber is rendered infusible by subsequent curing, which represents an additional chemical surface reaction. Without this step, fiber integrity is lost because of melting and/or creep of the green fiber before full transformation into the ceramic state. The first nonoxide silicon-based ceramic fibers were obtained from poly(dimethylsilane) by the Yajima process [26]. Synthesis of polydimethylsilane through alkali-metalpromoted dehalocoupling of dichlorodimethlysilane and its thermal (Kumada) rearrangement into processable poly (methylsilylene-methylene), [HSi(CH3)eCH2]n, also referred to as polysilapropene, were already discussed in detail in Section 3.1. Polysilapropene fibers were obtained by melt spinning in an inert gas atmosphere such as nitrogen or argon at around 300  C. According to Figure 50, curing of the green fibers was performed by controlled surface oxidation in air at approximately 150  C, which resulted in the oxidation of SieH units accompanied with the formation of SieOeSi linkages on the fiber surface. Heating of such partially oxidized fibers in an inert gas atmosphere to 1200  C delivered SieCeO fibers, which are commercially available under the trademark NicalonÔ [329]. As evidenced from HR TEM and X-ray photoelectron spectra, such fibers are composed of free carbon, nanocrystalline b-silicon carbide and an amorphous nonstoichiometric SieCeO matrix. Heating of the fibers to temperatures exceeding the final thermolysis temperature results in their decomposition caused by the elimination of SiO and CO. A further crucial subject is the hydrogen content of the ceramic fibers which is progressively reduced when heating the fibers to 1200e1300  C or higher. All these effects are accompanied by a coarsening of the fiber morphology and result in H

PCS

Si CH2 Me

n

Spinning

PCS

Green Fiber Electron Beam Irradiation

Oxidation Cured Fiber Thermolysis Si-C-O Fiber NicalonTM Fibers

Si-C Fiber Hi-Nicalon Fibers

TM

FIGURE 50 Flow scheme of the preparation of SiC NicalonÔ and Hi-NicalonÔ fibers [26].

Chapter | 11.1.10

Precursor-Derived Ceramics

1087

a dramatic deterioration of the high temperature mechanical properties. Alternatively, electron beam (2 MeV) or g-ray (60Co) irradiation served for the chemical modification of the fiber surfaces, by which contamination of the fibers with oxygen could almost be avoided. The mechanism behind theses processes is based on radical reactions. Noticeably, free radicals, which remain in the green fibers, have to be removed by additional annealing steps under inert atmosphere to reduce the reactivity of the extremely sensitive radical species in order to avoid oxidation and/or hydrolysis when exposing the fibers to air. Commercially available silicon carbide fibers, which are obtained by melt spinning and subsequent electron beam curing, are, for example, HiNicalonÔ fibers [329]. Even though oxygen-free, these fibers also crystallize at comparably low temperatures ( 0.6, the out-of-plane lattice constant abruptly decreased at x ¼ 0.6 and remained almost constant close to the BaTiO3 end of the composition spread. A gradual increase of the inplane lattice parameter shows that the gradual substitution of Ba for Sr resulted in an expansion of the unit-cell volume. For x < 0.6, coherent epitaxy took place despite the cell volume increase, resulting in an in-plane lattice constant of the film that was equal to that of the substrate and the lattice was thus compressively strained. At x ¼ 0.6, the increasing in-plane lattice mismatch between the film and the substrate

FIGURE 6 (a) Concurrent XRD mapping of a BaxSr1 xTiO3 composition spread library. (b) Out-of-plane and in-plane lattice constants derived from the concurrent XRD mapping.

1109

generated misfit dislocations at the interface. Hence, the compressive stress was abruptly relaxed, as seen by the abrupt change of the in-plane and out-of-plane lattice constants and the broadening of the film diffraction peak. This binary composition spread experiment was used to obtain a growth mode map of the alloy phases and was used to map the strain-dependent dielectric properties. Similar high-throughput characterization methods have recently been used to study (La,Sr)MnO3 [13], Sr2(Rh,Ru)O4 [17], and (Zn,Co)Fe2O4 [31] systems.

3.2. Ternary Phase Diagrams Formation of addressable ternary composition spreads with predetermined composition distributions has been achieved by controlling the film deposition time or thickness at each position on the substrate with the aid of physical shadow mask movement and substrate rotation [15,32e34]. Figure 7 schematically illustrates the preparation process for an addressable ternary composition spread library. A ternary composition spread library can be grown by superimposing three linear gradients that follow the three axes of an equilateral triangle, as illustrated in Figure 7(a)e(d). This can be easily achieved by using a single shadow mask and 120 substrate rotations. The total thickness of the three gradient films deposited in a single cycle is kept constant at one unit cell. The elemental diffusion length should exceed the vertical film thickness of each deposition cycle on a heated substrate, whereas it should be short enough to maintain lateral spatial resolution of the library. This condition is always satisfied in thin films. The sequence of growing linear gradients and library rotations is repeated until the desired total film thickness is obtained, as shown in Figure 7(e), yielding an addressable continuous ternary composition spread thin film library. A variety of shadow mask designs have been used in combinatorial thin film growth systems. In many cases, the type of masks used is dictated by the vacuum system geometry. In general, shadow masks used for preparing thickness gradient films use either a single edge or two edges, which pass over the substrate surface once or twice in each deposition cycle. Figure 8 shows a schematic illustration of the preparation procedure of a ternary composition spread film by using a single-edge mask that can be used without rotating the substrate [34]. Three separate masks, shown in Figure 8(a)e(c), are used, with working edges oriented at 30 , 90 , and 30 relative to the mask movement direction. The three masks are used to grow three film layers with the thickness gradients of the layers separated by 120 rotations, as shown in Figure 8(d)e(f). The actual mask shape used in a deposition chamber is shown in Figure 8(h). In principle, the whole ternary phase space shown in Figure 8(g) can be

1110

Handbook of Advanced Ceramics

FIGURE 7 Schematic diagram showing the procedure of fabricating a ternary composition spread film. (aec) The shades of gray denote the film thickness on a substrate. By combining submonolayer gradients (aec), a ternary composition spread film is formed, corresponding to a ternary phase diagram (d). (e) Steps (aec) are repeated several times until the desired film thickness is achieved, yielding an addressable ternary composition spread library.

covered, although typically a single library covers only a subset of the full phase diagram. Figure 9 shows a schematic illustration of a masking technique that also produces a ternary composition spread library but uses double-edge masks [34]. An arbitrary angle a of the double edge strongly affects the shadow mask motion speed and the film deposition speed. The characteristic point in the double edge is that a linear thickness gradient is available along the 90 rotated direction with respect to the mask movement direction, whereas this gradient direction is impossible when depositing a library with a single edge mask. When the double edge mask is employed for ternary composition spread fabrication, the shadow masks move as shown

FIGURE 8 Preparation procedure of a ternary composition spread film by using three single-edge masks (aec). The shifting edges in (aec) result in thickness gradients that are rotated by 120 with respect to each other (def), mapping a ternary phase space (g). A ternary composition spread can be fabricated using a single linear motion mask (h).

Figure 9(aec). The composition gradients are rotated by 120 with respect to each other, as shown in Figure 9(def). Figure 9(h) illustrates the shadow mask pattern when double edges are used, suggesting that the total mask length is shorter but the deposition time is longer than with a single edge mask in Figure 8(h). The choice of the mask depends partly on the desired deposition time and deposition rate. This is particularly important for the PLD growth technique, which may set limits on the number of discrete deposition pulses and pulse rates that give optimal film growth but may be less than optimal for operating masks. The ternary composition spread technique has been demonstrated for ReCa4O(BO3)3 system where Re is a rare

Chapter | 11.1.11

Combinatorial Nanoscience and Technology for Solid-State Materials

1111

FIGURE 9 Preparation procedure of a ternary composition spread library by using double-edge masks (aec). The shifting active edge, marked with thick lines in (aec) defines three film thickness gradient directions, rotated by 120 with respect to each other (def), covering a ternary phase space (g). (h) Schematic illustration of the most efficient mask pattern for creating an addressable continuous composition spread thin film library.

earth metal, Tb, Sc, or Pr. GdCa4O(BO3)3 and YCa4O(BO3)3 are known as nonlinear optical materials with transmittance in the range of 210e2600 nm, a high damage threshold, and good chemical stability [32]. Tb atoms at the Re site are known to induce luminescence in the green region of the spectrum, at 542 nm as a result of a 5D4 / 7F5 transition under ultraviolet excitation. A ternary composition spread library of epitaxial 300-nm Tb1 x-yScxPryCa4O(BO3)3 films grown on a YCa4O(BO3)3 (100) substrate is shown in Figure 10. The photoluminescence intensity of the composition spread film was investigated by taking a color photograph of the library under ultraviolet excitation at 254 nm. The intensity of white light emission is shown in Figure 10. A composition region with the highest light emission intensity was identified in the luminescence intensity map. The Tb0.6Sc0.4PryCa4O(BO3)3 region had the brightest emission, and there was no positive effect of Pr in this system. A similar photoluminescence intensity dependence on the composition was confirmed in the conventional photoluminescence spectra of 5D4 / 7F5 emission (542 nm) excited with

a frequency-doubled Ar laser (488 nm). The ternary composition spread technique has also been used for highthroughput screening of many materials systems, including dielectrics HfO2eY2O3eAl2O3 [15], photocatalysts Sr(Ti,Cr,V)O3 [35], phosphors Eu1 x yScxPryCa4O(BO3)3 [36], (Eu,Tm,Tb)-doped Y2O3 [33], high-mobility oxide semiconductors SrTiO3eLaAlO3eLaTiO3 [37], and flux materials [38e40].

3.3. Superlattice Libraries To carry out parallel fabrication of layered structures by an atomic layer-by-layer process, combinatorial laser molecular beam epitaxy (CLMBE) was developed [12]. The CLMBE allows concurrent fabrication of a number of artificial lattices or heterojunctions where the composition and thickness of each layer and the deposition sequence are manipulated at an atomic scale. This represents a considerable improvement in the ability to design new materials systems: libraries can now be constructed where

FIGURE 10 (Left) shows a schematic illustration of highthroughput screening for the photoluminescence properties under ultraviolet 254-nm lamp excitation. The white light emission intensity can be detected with a digital camera for rapid screening. (Right) The observed composition dependence in a ternary composition spread library can be verified by comparing with standard single-point photoluminescence measurement results.

1112

Handbook of Advanced Ceramics

FIGURE 11 CLMBE. (a) Fabrication scheme of a thin film superlattice library. (b) RHEED intensities plotted against time during simultaneous deposition of SrTiO3/BaTiO3 superlattices on a single substrate. The layering sequence at different times is schematically illustrated at the bottom.

arrangements of atoms are essentially controlled in three dimensions. Three key features of a CLMBE system are illustrated in Figure 11. It is a unique combinatorial synthesis tool, as it includes a set of physical masks for defining the film growth site, a scanning RHEED system [14] for in situ diagnostics of film growth mode at various sites during the growth, and a fiber-guided Nd:YAG laser for substrate heating. By monitoring the growth mode with RHEED and by synchronizing the target switching with mask movements, the synthesis of a number of atomiclayer controlled materials can be coordinated on a single substrate. This form of lattice engineering was first verified FIGURE 12 Intensity mapping for low angle (a) and high angle (b) diffraction for a combinatorial superlattice chip containing ten superlattices of [(SrTiO3)n/(BaTiO3)n]30 (n ¼ 12, 14, 16, . , 30). The superlattice satellite peaks appeared at angles corresponding to the superlattice periodicities.

in the simultaneous fabrication of perovskite oxide (ABO3) superlattices. In such an experiment, clear intensity oscillation in scanning RHEED, each of which corresponds to the growth of a unit-cell layer, is observed on the surfaces defined by the mask pattern in Figure 11(b). The RHEED beam is scanned across the substrate surface by a pair of coils, and the scanning and image acquisition are synchronized with the moving of shadow deposition masks to control the growth of as many as ten thin film strips in parallel. Figure 12 (a) and (b) show concurrent XRD patterns for a library of ten [(SrTiO3)n/(BaTiO3)n]30 superlattices with

Chapter | 11.1.11

Combinatorial Nanoscience and Technology for Solid-State Materials

n ¼ 12, ., 30. In addition to the specular reflection band near 0.70 and the SrTiO3(001) reflection near 22.75 , many satellite peaks are visible due to the superstructure. The positions of the reflection peaks agree well with the simulated positions and show that the actual periodicities of the superlattices match the design.

3.4. Temperature-Gradient Epitaxy Besides mapping phase diagrams, combinatorial thin film growth can also be used for rapidly optimizing the growth of known phases. The obvious advantage of using parallel growth techniques is the increased experimental throughput. However, another very important advantage is the increased reliability of the data from a growth parameter optimization experiment. For example, in a typical one-by-one experiment procedure, a sequence of samples may be grown to study the effects of film microstructure on electronic properties, such as breakdown fields of thin film capacitors. Such measurements generally have large error margins, largely because the electronic parameters are very sensitive to minute changes in the film microstructure. As each deposition run is slightly different, a sequence of individual samples often results in a large scatter of data

1113

points. In a combinatorial experiment, all samples within a single library are grown in a single deposition run on a single substrate. For growth parameter mapping, such as growth temperature, only that single parameter is intentionally varied. This can lead to a significant improvement in the reliability of the characterization data, which means that a smaller number of samples are sufficient, and all samples can be prepared in a single experiment. Temperature is the most common parameter used for controlling crystal growth, either in bulk or thin film form. Mapping film microstructures, growth modes, and physical properties as a function of the growth temperature is therefore a very common and important task. A parallel technique for generating a lateral temperature gradient over a library surface during thin film growth has been achieved by using a special sample holder, illustrated in Figure 13 [16]. A continuous wave Nd:YAG laser is used to heat the sample holder, with the laser beam focused on the backside of the free end of the cantilever-like center part of the sample holder. A 5  15-mm2 substrate is attached to the central part of the holder with silver paste. The heating laser beam is focused to a 3-mm diameter spot on the free-standing end of the sample holder while heat is lost mainly by conduction through the other end, which is attached to the main sample

FIGURE 13 (a) Schematic illustration of a temperature-gradient sample holder for mapping growth temperature effects on film properties. (b) Variation of local temperature on the sample surface as a function of position. (cef) AFM images before (c) and after the growth of (La,Sr)MnO3 epitaxial thin films at 650  C (d), 750  C (e), and 900  C (f).

1114

holder and the sample stage in the chamber. A steady heat flow in the sample holder results in a stable temperature gradient in this way, covering a range of temperatures from about 400 to 900  C over a distance of 12 mm in an oxygen back pressure of 1 mTorr. The temperature profile on the substrate surface can be measured by scanning the optical pyrometer across the sample surface, as shown in Figure 13(b). The measurement spot size was about 2 mm. Heteroepitaxial (La,Sr)MnO3 films have been grown using this temperature-gradient system on a step-and-terrace SrTiO3(001) substrate. The film surface morphologies, measured at different parts of the temperature-gradient library, corresponding to local substrate temperatures of 650  C, 750  C, and 900  C are shown in Figure 13(def). The original substrate surface is shown for comparison in Figure 13(c). The film grown in layer-by-layer mode at 650  C in Figure 13(d) exhibits step lines and islands with uniform single unit-cell step heights. In contrast, the surface of step-flow growth region in Figure 13(f) is hardly distinguishable from the initial substrate surface with atomically smooth terraces and straight steps. The mixed growth region in Figure 13(e) shows small islands on atomically smooth terraces and relatively straight steps. The same method has been used to optimize the growth of Sm2Mo2O7 [41], Sr2(Rh,Ru)O4 [17], and many other thin films. Recently, the temperature-gradient technique has been used without laser heating as well [42]. A special heat-flow substrate holder is the key point to obtain stable temperature gradients in a library chip.

3.5. Combinatorial Screening of Flux Materials for Single-Crystal Film Growth Flux materials are often used for the synthesis of compounds that have high melting temperatures and/or decompose incongruently. A flux is needed mainly to improve the crystallinity by reducing the growth temperature and suppressing thermodynamic phase separation in bulk crystal syntheses [36]. The presence of flux materials have been found useful in improving the crystallinity of many different films, for example, thin films grown by

FIGURE 14 Schematic illustration of conventional thin film growth (a) and the flux-mediated epitaxy (bed). In flux-mediated epitaxy, a liquid flux layer is deposited on the substrate (b). Subsequently, the film material is deposited through the liquid flux layer. The film nucleates and grows at the substrate interface, resulting in better crystallinity than in direct gas-phase growth (c). The film growth continues at the fluxefilm interface until the desired film thickness has been reached (d).

Handbook of Advanced Ceramics

vapor phase epitaxy, such as the chemical vapor deposition growth of Si whiskers in the liquid phase by the vapore liquidesolid (VLS) method [43]. The VLS process has recently been extended to triphase epitaxy (TPE) [44,45] of oxides, using PLD to deposit a liquid layer and film precursors from gas phase by laser ablation of stoichiometric solid targets, as shown in Figure 14. The advantage of using liquid phase growth for oxide thin films has been demonstrated by the growth of near-perfect single-crystal thin films of cuprate superconductors, NdBa2Cu3O7 d [44] and Ca-doped NdBa2Cu3O7 d [45]. The main advantage of TPE for superconductor growth is the lack of grain boundaries and improved crystallinity due to the reduction of misfit dislocations. Flux-mediated epitaxy, both VLS and TPE, are therefore receiving more attention as methods of fabricating very high-quality oxide thin films. The growth of Y-type magnetoplumbite (Ba2Co2Fe12O22:Co2Y) thin films is a good demonstration of the capabilities of flux-mediated epitaxy [46e48]. Magnetic oxides can be used as core materials for thin film inductors and other microwave devices operated in the GHz frequency range. The synthesis of magnetoplumbite thin films is complicated by incongruent melting that leads to a decomposition into BaFe2O4 and Co2Y. If magnetoplumbite films are deposited by conventional methods from a stoichiometric target, a segregated BaFe2O4 impurity phase grows on the substrate. To suppress the formation of the BaFe2O4 impurity phase, flux-mediated epitaxy has been used with the aid of CoO self-flux. The CoO self-flux layer is first grown directly on an MgAl2O4(111) substrate. Subsequently, a stoichiometric magnetoplumbite film can be grown through the predeposited flux layer. Figure 15(a) schematically depicts the cross-sectional library structure for combinatorial optimization of the CoO flux layer ˚ . Concurrent XRD patterns thickness from 800 to 1,000 A for the peaks of the impurity phase and Co2Y are shown in Figures 15(b) and (c), respectively. The impurity peaks completely disappeared, and the Co2Y peak intensity increased when the thickness of the CoO layer exceeded ˚ , reaching a maximum for a 950 A ˚ CoO flux layer 850 A thickness. By optimizing the thickness, a single-crystal

Chapter | 11.1.11

Combinatorial Nanoscience and Technology for Solid-State Materials

FIGURE 15 (a) Cross-section of a flux library, showing the variation of the CoO flux layer thickness on a substrate. Concurrent XRD patterns of BaFe2O4 (2q ¼ 41.3 ) and an unknown segregated phase (2q ¼ 43 ) (b), and Co2Y (2q ¼ 31.8 ) (c).

magnetoplumbite film could be grown on an MgAl2O4(111) substrate, as shown by transmission electron microscopy and XRD analysis [46e48].

4. DISCOVERIES MADE BY COMBINATORIAL TECHNOLOGY 4.1. Quantum Size Effect in Photocatalytic TiO2 Photocatalysis has attracted considerable attention since the discovery of the HondaeFujishima effect [49] due to the possibility of hydrogen fuel production by water splitting and photodeposition of various materials. To optimize the catalytic properties, it is essential to understand the surface reaction on a nanometer scale. For this purpose, photocatalytic properties of epitaxial TiO2/Nb:TiO2 heterostructures were investigated [50]. The use of heterostructures in catalysts has the advantage of eliminating most of the surface morphology effects on catalytic activity, because the uniform atomically flat film surfaces are nearly identical to the original single-crystal substrates. Another advantage of thin films over powder catalysts is the

1115

ability to accurately control the layer thickness, which corresponds to the particle size of a conventional powder catalyst. Most of the structural parameters of heterostructures, in particular layer thicknesses, can be easily mapped in combinatorial experiments. To study the process of photocatalysis on a titania surface, TiO2 thin films have been fabricated on atomically smooth Nb:TiO2(110) (Nb:0.5 wt %) single-crystal substrates by PLD [51], which allows for accurate in situ growth condition optimization by real-time RHEED intensity oscillation monitoring. Films with flat surfaces normally show sharp RHEED patterns and a step and terrace surface morphology that is equivalent to that of the substrate. Thickness gradient films can be used to investigate the effect of the surface layer thickness on the photoactivity. A convenient way for measuring the absolute activity and spatial variation of catalytic activity on an oxide surface is to use the photochemical deposition of Ag particles. When thin film samples are immersed in a 0.01 N AgNO3 electrolyte under ultraviolet illumination, as illustrated in Figure 16(a), the amount of photodeposited Ag particles can be evaluated by scanning X-ray fluorescence (XRF), giving an accurate measure of the catalytic activity with a spatial resolution of the XRF probe. A photograph of the catalyst surface after a photochemical reaction is shown in Figure 16(b). The bright white line close to the left edge of the library shows that the photodeposited Ag particles accumulate in a narrow thickness region of the TiO2 cap layer. Figure 16(c) shows the Ag concentration on the surface as a function of the TiO2 layer thickness. A strong photoactivity peak can be seen at a film thickness of about 5 nm, with a full width at half maximum of just 3.9 nm. By conventional one-by-one process, it is difficult to find such a narrow thickness range, which is the only region where strong photoactivity occurs in this material system. The combinatorial technique is very efficient at finding such narrow regions of interest in broad phase diagrams and is therefore suitable for the detection of photocatalytic size effects in single-crystal heterostructures. The same technique has also been used to study (Ti,V)O2 [50], (Ba,Sr) TiO3 [52], and Sr(V, Cr, Ti)O3 [35] systems and confirmed to be useful for the investigation of the film thickness effects as well as the film composition dependences.

4.2. Room-Temperature Transparent Ferromagnetism in Co-Doped TiO2 A class of ferromagnetic semiconductors that are obtained by doping magnetic impurities into host semiconductors is one of the key materials for spintronics in which the correlation between charge and spin of electrons is used to bring about spin-dependent electronic functionality such as giant magnetoresistance and a spin field effect transistor switching. Among the materials that had been reported

1116

Handbook of Advanced Ceramics

FIGURE 16 Schematic illustrations of the combinatorial screening process for a photocatalytic material. (a) Photochemical deposition process for Ag particles on a catalyst surface. (b) Photograph of photodeposited Ag particles on a TiO2/Nb:TiO2 heterostructure surface. (c) Film thickness dependence of photodeposited Ag particle density. Strong photoactivity is visible only in a narrow thickness region of around 5-nm.

until 2001, Mn-doped GaAs had the highest Curie temperature, Tc z 100 K, and is still presumed to be a promising candidate, with a gradual increase of Tc up to 250 K [53], for practical applications. However, there had been no report on any ferromagnetic semiconductor in oxides at that time. An exploration of a novel ferromagnetic oxide semiconductor was thus a challenge to verify the capability of our combinatorial thin film method. ZnO [54], rutile, and anatase TiO2 [21,25,55], the latter two phases of which could be controlled by using a different latticematched substrate of sapphire (0001) or SrTiO3(001), were selected as a host semiconductor, and each was systematically doped with all the 3d transition metal ions. The “one-month combinatorial doping experiment” yielded 243 distinct samples as shown in Figure 17, and the solubility limits of each different transition metal ion in anatase, rutile and ZnO films were quickly determined from XRD measurements [54,55]. For a quick screening of magnetic properties of these 3d-transition metal-doped TiO2 and ZnO film libraries, a scanning SQUID microscope technique was employed, leading to a serendipitous discovery of room-temperature transparent ferromagnetism in both of Co-doped anatase and rutile TiO2 films [21,22]. Figure 18 shows a series of scanning SQUID microscope images taken at 3 K for anatase thin films with different Co contents of 0% (a), 3% (b), and 6% (c) of Co. In all of the Co-doped TiO2 films, magnetic domain structures with an approximately 20-mm size can be seen, while none of the nondoped TiO2 films show magnetic structure within an experimental error of around 0.5 mT. With an increase of the Co content in the film, the magnitude of the observed magnetic field is also enhanced as a result of the increased spontaneous magnetization. These observations are evidence for the emergence of a ferromagnetic long-range order in the Codoped TiO2, irrespective of the crystal phase being rutile or anatase. The Tc was estimated to be >400 K at that

FIGURE 17 Combinatorial fabrications of TiO2 and ZnO films doped with all 3d transition metal ions.

Chapter | 11.1.11

Combinatorial Nanoscience and Technology for Solid-State Materials

1117

FIGURE 18 A series of scanning SQUID microscope images (200  200 mm2) taken at 3 K for anatase thin films with different Co contents on a combinatorial chip: (a) 0%, (b) 3%, and (c) 6%. Magnetic domain structures were observed in all doped films, suggesting the presence of long-range ordering of magnetic moments induced by Co doping in the anatase thin film. (d) Transmittance spectrum of a Co0.06Ti0.94O2 sample. The inset shows a sample photograph.

time. Figure 18(d) is an UltravioleteVisible spectrum for a Co-doped anatase thin film, together with a photograph of the sample. The Co-doped TiO2 film is transparent in the visible and near-infrared regions, exhibiting a band gap at 400 nm (3.1 eV). On the other hand, no signs of such a magnetic domain structure were observed in any other combinatorial TiO2 and ZnO film libraries, though some research groups claimed that Co-doped ZnO also became ferromagnetic, which is still somewhat controversial. In fact, to date, there have been reports of many types of ferromagnetic oxide semiconductors including Co-doped TiO2 and ZnO, none of which has, however, been commonly accepted as a ferromagnetic semiconductor. In this controversial situation, it is quite recently that hydrogen was found to play a key role in the emergence of ferromagnetism in Co-doped ZnO [56] and the origin of ferromagnetism in Co-doped TiO2 was verified to be intrinsic and rationalized by a scheme of carrier-induced ferromagnetism. Fukumura and coauthors have successfully demonstrated that the magnitude of ferromagnetism in Co-doped TiO2 can be well scalled by the carrier density, which can be chemically controlled by the degree of oxygen deficiency and also physically modulated by the field effect even at room temperature [57e61]. Therefore, Co-doped has now become one of the most

promising and representative ferromagnetic semiconductors for a possible realization of transparent semiconductor spintronics devices operable at room temperature.

4.3. Growth of Single-Crystal Oxide Films by Flux-mediated Epitaxy The key point in the successful application of flux-mediated epitaxy is the selection of an appropriate flux material. It is not sufficient to examine the well-known fluxes predicted from the bulk phase diagrams, because thin film growth conditions are definitely different from those of the bulk process. In fact, the volatile Bi2O3 self-flux of Bi4Ti3O12, which is predicted from the phase diagram of Bi2O3 and TiO2, has proven to be useless for the Bi4Ti3O12 thin film growth [38]. It is therefore necessary to find a suitable impurity flux individually for each case and without the ability to rely on obvious guiding principles that are known for flux composition selection in the case of corresponding bulk crystal growth. Owing to the large diversity of possible flux materials, this is another area where combinatorial parallel synthesis can greatly speed up the optimization process. It is inevitably laborious and time consuming to optimize the composition of the flux materials in a conventional one-by-one procedure.

1118

The combinatorial approach offers a novel and unique way to overcome the throughput problem in flux optimization. To explore a flux material for Bi4Ti3O12, addressable ternary composition spread libraries were fabricated by the use of a combinatorial PLD system. The libraries were composed of Bi4þxTi3 xOd (0 < x < 3) self-flux and impurity fluxes. Candidates of impurity fluxes were oxides of V, W, Cu, Mo, BieP, and Ba. All these oxides are wellknown flux materials for the growth of various oxide bulk crystals. Figure 19 schematically illustrates the combinatorial process of the flux material search for Bi4Ti3O12 thin film growth [38,39]. At first, a ternary composition spread of the flux layers was deposited on a SrTiO3(001) substrate, followed by a Bi4Ti3O12 film grown through the flux layer. After growth, the crystallinity of the Bi4Ti3O12 film in the library was characterized by concurrent XRD [19], by taking several cuts through the ternary phase diagram library and combining the XRD data in a crystallinity phase diagram plot, as shown in Figure 19(b). Figure 20(a) shows an XRD pattern of a Bi4Ti3O12 thin film prepared in a conventional one-by-one process. The Bi4Ti3O12 thin film on an SrTiO3(001) c-axis oriented. Using concurrent XRD, all the c-axis lattice parameters and the Bi4Ti3O12 (0014) peak intensities were mapped over the entire library area for 500-nm-thick Bi4Ti3O12 films, including a Bi4Ti3O12 thin film grown without a flux. By plotting the c-axis parameter for each measurement point

FIGURE 19 Schematic illustration of the combinatorial screening process used for finding a suitable flux material to assist the growth of ferroelectric Bi4Ti3O12 single-crystal films. (a) A flux composition spread is deposited on the SrTiO3 substrate. (b) A Bi4Ti3O12 film is grown through the predeposited flux layer. (c) Characterization of the spatial variation of the crystallinity of the Bi4Ti3O12 film in the ternary flux composition spread library by concurrent XRD.

Handbook of Advanced Ceramics

vs. the (0014) peak intensity, it is possible to evaluate the crystallinity of the film as a function of the flux composition. The results of such a mapping are shown in Figure 20(beh). Each ternary library contained z5,000 measurement points in a triangle with 8 mm sides. A total of 35,000 data points were collected to find the best flux composition for Bi4Ti3O12 film growth. Although the c-axis parameters and peak intensities showed slight fluctuations from one position to another even in the case of a Bi4Ti3O12 film grown without a flux, detailed statistical analysis of the flux library made it possible to establish a clear tendency: the series of Bi4Ti3O12 films grown by using a CuO-containing flux showed higher X-ray peak intensities than any other library, while the c-axis lattice parameter values approached the bulk Bi4Ti3O12 value of 3.283 nm. This combinatorial screening suggested that a CuO-containing flux would be the best candidate to assist Bi4Ti3O12 film growth. After further optimization of several growth parameters, stoichiometric and very high-quality Bi4Ti3O12 single-crystal films could finally be grown in a CuOeBiOx flux [38]. Figure 21 shows a comparison of transmission electron microscope (TEM) images of Bi4Ti3O12 films grown without and with a CuO-containing flux. Without using the CuO flux, many out-of-phase boundaries originating from the step structure of the substrates could be observed. In contrast, no boundaries were seen in a high-quality Bi4Ti3O12 film grown with a flux containing CuO. It appeared that the CuO flux suppressed boundary growth at the substrate steps by maintaining thermodynamic equilibrium conditions across substrate step edges. Figure 21(c) shows an AFM image of a Bi4Ti3O12 film grown with the CuO flux. Atomically smooth terraces can be seen with all steps having a uniform 1.6-nm height, corresponding to half a unit of the Bi4Ti3O12 crystal structure. The large terrace width is a direct result of the equilibrium conditions prevailing at the liquidesolid interface in the flux [62]. The uniformity of the step height can be attributed to the uniformity of the surface termination layer, which was determined to be the (Bi2O2)2þ layer from cross-sectional high-resolution TEM analysis of the surface. AFM images of flux-grown Bi4Ti3O12 films showed no 0.4-nm step structures originating from out-of-phase boundaries, supporting the TEM analysis, which also failed to find such boundaries. The effect of decreased structural defect density can be seen in the leak currents of the two types of films. Figure 21(d) shows a comparison of the leak current density, and it is clear that flux-mediated growth resulted in a drop of the leak current density from 10 3e10 4 A/cm2 to 10 7e10 8 A/cm2. The leak current density improvement is also related to reduced Bi deficiency, based on secondary ion mass spectrometry (SIMS) analysis. Without using a CuO flux, SIMS measurements showed a high Bi deficiency in the films. In contrast, the composition of flux-grown Bi4Ti3O12 single-crystal films was very close to stoichiometric: the

Chapter | 11.1.11

Combinatorial Nanoscience and Technology for Solid-State Materials

1119

FIGURE 20 (a) XRD pattern of a Bi4Ti3O12 thin film fabricated by a conventional one-by-one process. Closed and open circles mark the SrTiO3 substrate and Bi4Ti3O12 film, respectively. (beh) Mapping of the c-axis parameter vs. the Bi4Ti3O12 (0014) peak intensity for 35,000 (5,000  7) different flux compositions. From a statistical point of view, Bi4Ti3O12 films grown in the presence of a CuO-containing flux had the highest X-ray intensities and the closest lattice parameter to the bulk value.

FIGURE 21 Cross-sectional TEM images of the Bi4Ti3O12 film grown with no flux (a) and in the presence of a CuOcontaining flux (b). Many out-of-phase boundaries running across the Bi4Ti3O12 film can be seen in (a), while none are visible in (b). (c) AFM image of a Bi4Ti3O12 film grown by flux-mediated epitaxy. (d) Leak current density vs. applied electric field for Bi4Ti3O12 films prepared with and without a flux.

1120

films had Bi concentrations corresponding to a Bi4Ti3O12 powder. Combinatorial screening sometimes yields serendipitous discoveries [40]. As a result of accumulating 35,000 flux composition data points, a CuO-based flux was found to assist Bi4Ti3O12 thin film growth, with the crystallinity of the film being the driving parameter for the optimization process. However, it is known that different flux materials can also affect other aspects of crystal growth, such as crystal orientation, besides crystallinity. Because the XRD data sets included more information than just the crystallinity data, the data set of 6 ternary composition spread libraries was reanalyzed. The concurrent XRD mappings showed that only the V oxide flux library showed a shoulder peak around the SrTiO3(002) substrate reflection, suggesting the growth of an unexpected phase. Moreover, the same library color was visually yellowish in the V oxide rich region. To understand these observations, a conventional powder XRD pattern was taken from an entire region of the ternary composition spread library containing V oxide, as shown in Figure 22(a). In addition to the c-axis oriented phase, (110)-oriented Bi4Ti3O12 was found to be present in the V oxide flux library but not in any other libraries. This was very surprising, because Bi4Ti3O12 films grown directly on the SrTiO3(001) surface are known to have a unique c-axis orientation. In order to look further into this intriguing phenomenon, the (220) peak intensity was mapped over the entire library by concurrent XRD, as shown in Figure 22(beg). For most flux compositions, the (220) peak was not detected, with the (00l) peaks

FIGURE 22 (a) XRD pattern of a Bi4Ti3O12 thin film grown in the presence of a Voxide flux. Both c-axis and (110) oriented peaks are visible. (ceg) Concurrent XRD mappings of the Bi4Ti3O12 (220) peak intensity. Only the library containing a V oxide showed a diffraction signal, with the highest intensity in the pure V oxide flux region (b).

Handbook of Advanced Ceramics

from the growth of c-axis oriented Bi4Ti3O12 being dominant. Only the V oxide flux library exhibited elevated (220) peak intensity, with the highest intensity in the pure V oxide flux region in the entire library shown in Figure 22(b). From this first screening, it was concluded that V oxide may be a promising flux to assist orientation control from the common (001) direction to the (110) direction during the growth of Bi4Ti3O12 on the SrTiO3(001). To understand the relationship between a flux containing V oxide and Bi4Ti3O12(110)-oriented growth, the V oxide film thickness dependence was investigated by preparing a combinatorial library illustrated in Figure 23(a). A pure V oxide flux film was deposited with a thickness gradient of 0e50 nm directly on a SrTiO3(001) substrate by moving a shadow mask during deposition. Subsequently, a 500-nmthick Bi4Ti3O12 film was deposited on the Voxide thickness gradient layer. Scanning X-ray microdiffraction analysis was used for mapping the X-ray peak intensity as a function of the 2q angle in the 37 e49 range as a function of the V oxide thickness flux layer thickness (0e50 nm). It was found that even a small amount of V oxide was sufficient to promote the growth of a (110)-oriented Bi4Ti3O12 film, and the (220) peak intensity gradually increased in accordance with the increase of the V oxide flux layer thickness. The preferential growth of the (110)-oriented Bi4Ti3O12 film was therefore attributed to the V oxide flux effect. Here, the (0014) and (220) peak intensities were evaluated from a Gaussian fitting of the XRD mapping data, as shown in Figure 23(b), and plotted as a function of the sample position in Figure 23(c). It was found that a 30-nm V oxide layer is

Chapter | 11.1.11

Combinatorial Nanoscience and Technology for Solid-State Materials

1121

A cross-sectional scanning electron microscope (SEM) image of a (110)-oriented Bi4Ti3O12 film grown on a Nbdoped SrTiO3(001) substrate is shown in Figure 24. Many rectangular skyscraper-like Bi4Ti3O12 platelets with a characteristic size of 1,000  1,000  100 nm3 grew all over the substrate. Each platelet has two equivalent in-plane orientations along the (100) direction of the substrate. The in-plane growth orientation of the Bi4Ti3O12 nanoplatelets was investigated by the use of transmission electron microscopy. Figure 24(b) shows diffraction patterns of an isolated Bi4Ti3O12 nanoplatelet, observed along directions normal to the narrower or wider sides. From these observations, the in-plane relationship between the Bi4Ti3O12 platelets and the substrate was determined as follows: Bi4 Ti3 O12 ½001Š k SrTiO3 ½010Šð½100ŠÞ; Bi4 Ti3 O12 ½1-10Š k SrTiO3 ½100Šð½010ŠÞ

FIGURE 23 The sample position dependence: (a) V oxide containing flux layer thickness, (b) XRD profile, (c) the integrated XRD intensities of Bi4Ti3O12 (0014) and (220) peaks.

sufficient to promote the formation of an almost phase-pure (110)-oriented Bi4Ti3O12 film. The growth of the (110)-oriented Bi4Ti3O12 film was duplicated by a conventional one-by-one process.

That is, the Bi4Ti3O12 nanoplatelet crystal was found to grow epitaxially on the SrTiO3 substrate, as illustrated in Figure 24(b). The epitaxial growth of the (110)-oriented Bi4Ti3O12 film was also found on other substrates, such as LaAlO3(001) and (LaAlO3)0.3(Sr2AlTaO6)0.7(001). The piezoelectric property of an isolated epitaxial nanoplatelet was examined by scanning probe piezomicroscopy. The lattice displacement and the effective d value have been measured along the [110] direction by applying a voltage. Figure 24(c) and (d) show a typical butterfly curve and a hysteresis loop, implying that the oriented platelets are ferroelectric.

FIGURE 24 (a) Cross-sectional SEM image of Bi4Ti3O12(110) orientation nanoplatelets on a conductive Nb-doped SrTiO3(001) substrate. (b) Schematic drawing of the epitaxial orientation of a Bi4Ti3O12 nanoplatelet, and diffraction patterns of wide and narrow planes. (c) and (d) Piezoelectric displacement and deff as a function of the bias voltage.

1122

Handbook of Advanced Ceramics

5. HIGH-TECH VENTURE: COMBINATORIAL MATERIAL BUSINESS Combinatorial technology is a powerful tool to bring about innovation in many kinds of materials development, not only oxides typically represented and described above but also nonoxide ceramics, metals and alloys, semiconductors, polymers, and composites. It has also been applied to highthroughput research and development of layer-structured materials such as superlattices and heterojunction devices as well as for interface phenomena. Conventional sputtering, PLD, vacuum evaporation, and MBE can be adapted to a combinatorial model by inserting a scheme of shadow masking to enable the fabrication of a material chip on which a variety of different materials are grown in a singlerun experiment. Although the spatial resolution of library cells is not as high as in a high-vacuum process, combinatorial technology can further be extended to other dry (chemical vapor deposition, etc.) and wet (inkjet, sol-gel, spray-pyrolysis, melt-quench, etc.) material processes. Matching of machine performance and experimental design can improve the research throughput by a factor of 100 as compared with the conventional one-by-one process. The introduction of combinatorial function does not necessarily require special machines and/or expensive attachment to existing equipment. Nevertheless, increasing interest in advanced materials of complex composition and nanostructures inevitably ask us to screen a wider space of

compositions and preparation conditions in a time span that is as short as possible. To fulfill the increasing needs for higher performance combinatorial material technology, it is desirable to have specially designed and computercontrolled easy-operation hardware, navigators, and skilled operators who can extract the full power of the machine. This relationship is comparable to that of a racing team composed of a tuned-up racing car, talented driver, and professional mechanics. Venture companies using combinatorial methods as a business tool emerged in the US and Europe at the end of the 1990s. Represented by Symyx and Wildcat, they were focusing on the discovery of new catalysts and on the informatics to supply software for handling large quantities of data acquired from combinatorial synthesis and characterization. In 2004, another type of venture company, Intermolecular Inc., started business, focusing solely on electronic materials. The main customers targeted are electronic companies to substitute their R & D division for Intermolecular in developing new industrial products by using a 300-mm wafers as standard substrates because modern large-scale integrated (LSI) circuit industries use 300-mm wafers for LSI manufacturing. In Japan, on the basis of fundamental combinatorial material research supported by Japan Science and Technology Agency for 1996e2001 and jointly by National Institute for Material Science (NIMS) and Tokyo Institute of Technology (TITECH) for 1999e2005, COMET Inc. was

FIGURE 25 Combinatorial materials R & D Venture (Comet Inc.).

Chapter | 11.1.11

Combinatorial Nanoscience and Technology for Solid-State Materials

founded in 2007 as a NIMS authorized venture company [63]. COMET, which stands for Combinatorial Materials Exploration and Technology has the main office inside NIMS, located in Tsukuba, Japan, and a branch in Kashiwa campus of Tokyo University, about 20 km south of Tsukuba. According to the business model, schematically illustrated in Figure 25, COMET provides a service of new material discovery on the request mainly to industry and also academia in the public sector. The laboratory is equipped with originally designed combinatorial thin film deposition systems including several types of pulsed laser MBE and multitarget (~6) sputtering for inorganic solid-state materials and an infrared laser MBE for organic and soft matter. Recently installed combinatorial sputtering equipment handles a 4-inch wafer with 6 cathodes:3 cathodes for 200 ternary matrix elements and the other 3 cathodes for 100 sputtering guns for doping or radical sources. Combinatorial cluster chambers were designed for all-in-one process from library fabrication to characterization without taking out (exposure to air) the sample, which is useful for device development. These original and patented combinatorial machines are on sale from COMET. In addition to the basic electronic property measurement in COMET, advanced structural analyses and property characterization are done in cooperative research laboratories in NIMS, Tokyo University, and TITECH. The focus of combinatorial materials research targeted by current activities at COMET covers energy saving, high-efficiency energy conversion, and energy storage technologies. From this point of view, we hope that combinatorial high-throughput technology not only opens new business opportunities but also serves the needs of all humanity by rapidly bringing new materials to the level of commercial use.

6. CONCLUSION Combinatorial materials technology can be used to develop a wide variety of functional materials and for device process optimization. With increasing interest in nanostructured and smart (e.g. multiferroic, spintronic) materials, the combinatorial library design has been expanded from one-dimensional (binary composition spreads, temperature gradients) to two-dimensional (ternary phase diagrams, composition þ preparation parameters) and three-dimensional (superlattices, heterojunction device, etc.) cases. Ceramic materials show a variety of characteristics, such as superconductivity, ferromagnetism, ferroelectricity, and light emission. Combinatorial techniques have contributed to the exploration of new ceramic materials as well as the optimization of the process parameters and discovery of new material properties. Combinatorial strategy with flexible consideration and ingenuity will continue to spread and is becoming a powerful tool for quickly solving urgent materials-related

1123

tasks that humanity is facing, such as energy efficiency, atmospheric pollution, and the emission of greenhouse gases. Combinatorial research can contribute to solving many of these problems, for example, by helping with the design of catalysts and energy harvesting devices.

ACKNOWLEDGMENTS The research in this article was supported in part by JST (Japan Science and Technology agency)-CREST (Core Research for Evolution of Science and Technology) program and by PNU (Pusan National University)-WCU (World Class University) program on Cogno-Mechatronics Engineering through the National Research Foundation of Korea (#R31-20004).

REFERENCES [1] Koinuma H. Solid State Ionics 1998;108:1. [2] Koinuma H, Aiyer NH, Matsumoto Y. Sci Tech Adv Mater 2000;1:1. [3] Merrifield RB. J Am Chem Soc 1963;85:2149. [4] Merrifield RB. J Org Chem 1964;29:3100. [5] Lebl M. J Comb Chem 1999;1:3. [6] Xiang X-D, Sun X, Briceno G, Lou Y, Wang K, Chang H, et al. Science 1995;268:1738. [7] Briceno G, Chang H, Sun X, Schultz PG, Xiang X-D. Science 1995;270:273. [8] Wang J, Yoo Y, Gao C, Takeuchi I, Sun Z, Chang H, et al. Science 1998;279:1712. [9] Koinuma H, Takeuchi I. Nat Mater 2004;3:429. [10] Takeuchi I, Lippmaa M, Matsumoto Y. MRS Bull 2006;31:999. [11] Matsumoto Y, Murakami M, Jin Z, Ohtomo A, Lippmaa M, Kawasaki M, et al. Jpn J Appl Phys 1999;38:L603. Part 2. [12] Ohnishi T, Komiyama D, Koida T, Ohashi S, Stauter C, Koinuma H, et al. Appl Phys Lett 2001;79:536. [13] Fukumura T, Ohtani M, Kawasaki M, Okimoto Y, Kageyama T, Koida T, et al. Appl Phys Lett 2000;77:3426. [14] Koinuma H, Koida T, Ohnishi T, Komiyama D, Lippmaa M, Kawasaki M. Appl Phys A 1999;69:S29. [15] Hasegawa K, Ahmet P, Okazaki N, Hasegawa T, Fujimoto K, Watanabe M, et al. Appl Surf Sci 2004;223:229. [16] Koida T, Komiyama D, Koinuma H, Ohtani M, Lippmaa M, Kawasaki M. Appl Phys Lett 2002;80:565. [17] Koida T, Wakisaki T, Itaka K, Koinuma H, Matsumoto Y. Appl Phys Lett 2002;81:4955. [18] Ohtani M, Kishida H, Hirota K, Sakurada H, Kawamoto N, Murakami Y, et al. Adv Mater 2006;18:2541. [19] Ohtani M, Fukumura T, Kawasaki M, Omote K, Kikuchi T, Harada J, et al. Appl Phys Lett 2001;79:3594. [20] Ohtani M, Fukumura T, Kawasaki M, Omote K, Kikuchi T, Harada J, et al. Appl Phys Lett 2002;80:2066. [21] Matsumoto Y, Murakami M, Shono T, Hasegawa T, Fukumura T, Kawasaki M, et al. Science 2001;291:854. [22] Matsumoto Y, Takahashi R, Murakami M, Koida T, Fan X-J, Hasegawa T, et al. Jpn J Appl Phys 2001;40:L1204. Part 2. [23] Takeuchi I, Famodu OO, Read JC, Aronova MA, Chang K-S, Craciunescu C, et al. Nat Mater 2003;2 180.

1124

[24] Kennedy K, Stefansky T, Davy G, Zackay VF, Parker EP. J Appl Phys 1965;36:3808. [25] Miller NC, Shirn GA. Appl Phys Lett 1967;10:86. [26] Hanak JJ, Gittleman JI, Pellicane JP, Bozowski S. Phys Lett 1969;30A:201. [27] Hanak JJ. J Mater Sci 1970;5:964. [28] Koinuma H, Kawasaki M, Nagata S, Takeuchi K, Fueki K. Jpn J Appl Phys 1988;27:L376. [29] van Dover RB, Schneemeyer LF, Fleming RM. Nature 1998;392:162. [30] Ohtani M, Lippmaa M, Ohnishi T, Kawasaki M. Rev Sci Instrum 2005;76:062218. [31] Iwasaki Y, Fukumura T, Kimura H, Ohkubo A, Hasegawa T, Hirose Y, et al. Appl Phys Exp 2010;3:103001. [32] Takahashi R, Kubota H, Murakami M, Yamamoto Y, Matsumoto Y, Koinuma H. J Comb Chem 2004;6:50. [33] Takahashi R, Kubota H, Tanigawa T, Murakami M, Yamamoto Y, Matsumoto Y, et al. Appl Surf Sci 2004;223:249. [34] Yamamoto Y, Takahashi R, Matsuoto Y, Chikyow T, Koinuma H. Appl Surf Sci 2004;223:9. [35] Ohsawa T, Nakajima K, Matsumoto Y, Koinuma H. Appl Surf Sci 2006;252:2603. [36] Kubota H, Takahashi R, Kim T-W, Kawazoe T, Ohtsu M, Arai N, et al. Appl Surf Sci 2004;223:241. [37] Okazaki S, Okazaki N, Hirose Y, Nishimura J, Ueno K, Ohtomo A, et al. Appl Phys Exp 2008;1:055003. [38] Takahashi R, Yonezawa Y, Ohtani M, Kawasaki M, Nakajima K, Chikyow T, et al. Adv Funct Mater 2006;16:485. [39] Takahashi R, Yonezawa Y, Ohtani M, Kawasaki M, Koinuma H, Matsumoto Y. Appl Surf Sci 2006;252:2477. [40] Takahashi R, Yonezawa Y, Nakajima K, Chikyow T, Koinuma H, Matsumoto Y. Appl Phys Lett 2006;88:152904. [41] Nishimura J, Fukumura T, Ohtani M, Taguchi Y, Kawasaki M, Ohkubo I, et al. Appl Phys Lett 2003;82:1571. [42] Ohkubo I, Christen HM, Kalinin SV, Jellison GE, Rouleau JCM, Lowndes DH. Appl Phys Lett 2004;84:1350. [43] Wagner RS, Ellis WC. Appl Phys Lett 1964;4:89.

Handbook of Advanced Ceramics

[44] Yun KS, Choi BD, Matsumoto Y, Song JH, Kanda N, Itoh T, et al. Appl Phys Lett 2002;80:61. [45] Takahashi R, Matsumoto Y, Kohno T, Kawasaki M, Koinuma H. J Cryst Growth 2004;262:308. [46] Ohkubo I, Matsumoto Y, Hasegawa T, Ueno K, Itaka K, Ahmet P, et al. Jpn J Appl Phys 2001;40:L1343. Part 2. [47] Ohkubo I, Matsumoto Y, Hasegawa T, Ueno K, Ohtani M, Kawasaki M, et al. Appl Surf Sci 2002;197e198:312. [48] Ohkubo I, Matsumoto Y, Ueno K, Chikyow T, Kawasaki M, Koinuma H. J Cryst Growth 2003;247:105. [49] Fujishima A, Honda K. Nature (London) 1972;238:37. [50] Abe T, Ohsawa T, Katayama M, Koinuma H, Matsumoto Y. Appl Phys Lett 2007;91:061928. [51] Yamamoto Y, Nakajima K, Ohsawa T, Matsumoto Y, Koinuma H. Jpn J Appl Phys 2005;44:L511. [52] Ohara K, Ohsawa T, Koinuma H, Matsumoto Y. Jpn J Appl Phys 2006;45:L339. [53] Ohno H. Science 2001;291:840. [54] Jin Z, Fukumura T, Kawasaki M, Ando K, Saito H, Sekiguchi T, et al. Appl Phys Lett 2001;78:3824. [55] Matsumoto Y, Murakami M, Hasegawa T, Fukumura T, Kawasaki M, Ahmet P, et al. Appl Surf Sci 2002;189:344. [56] Cho YC, Kim S-J, Lee S, Kim SJ, Cho CR, Nahm H-H, et al. Appl Phys Lett 2009;95:172514. [57] Toyosaki H, Fukumura T, Yamada Y, Nakajima K, Chikyow T, Hasegawa T, et al. Nat Mater 2004;3:221. [58] Toyosaki H, Fukumura T, Yamada Y, Kawasaki M. Appl Phys Lett 2005;86:182503. [59] Ueno K, Fukumura T, Toyosaki H, Nakano M, Yamasaki T, Yamada Y, et al. J Appl Phys 2008;103:07D114. [60] Yamada Y, Ueno K, Fukumura T, Yuan HT, Shimotani H, Iwasa Y, et al. Science 2011;332:1065. [61] Zutic I, Cerne J. Science 2011;332:1040. [62] Takahashi R, Tsuruta Y, Yonezawa Y, Ohsawa T, Koinuma H, Matsumoto Y. J Appl Phys 2007;101:033511. [63] http://www.comet-nht.com/index.html.

Chapter 11.2.1

Stereo Fabric Modeling Technology in Manufacturing Ceramics Hideki Kita Department of Molecular Design and Engineering, Graduate School of Engineering, Nagoya University, Furo-cho, Chikusa, Nagoya, Aichi 464-8603, Japan, Former Affiliate: National Institute of Advanced Industrial Science and Technology

Chapter Outline References

1129

Advanced ceramics are superior to conventional materials and, for this reason, have been adopted throughout the materials processing industry, including for steel, aluminum, chemicals, and semiconductors. There has been a strong and continuous growth in user confidence due to wide confirmation of the energy savings, increased productivity, and reduced maintenance requirements with the use of these materials. Thus, to match the growing demand, there is also a strong thrust toward scaling up ceramic parts while continuing to improve upon performance and simultaneously lowering costs. Take, for example, ceramic parts used in the nonferrous casting industry, where low heat loss and reduced wettability are key performance parameters. For such parts, the use of a lightweight design can greatly improve productivity. In addition, to increase panel size in the liquid crystal display industry and to increase wafer size, and thus throughput,

in the semiconductor industry, it is necessary to realize a corresponding increase in the sizes and specific rigidity of the ceramic members used for microfabrication. Generally, in the case of structural parts, the performance of the part is determined by its geometrical design. This implies that greater flexibility in the possible geometry of parts will allow to greater functionality. The techniques currently used in ceramic modeling include extrusion molding, injection molding, and press molding, depending on the shape, size, and required precision of the intended components. It is, however, difficult to fabricate a large, complex-shaped component with a very high precision, and this restricts the scope of possible designs for production. For these reasons, we have been developing a new modeling concept to simplify the fabrication of complexshaped and large ceramic components. The concept involves the integration of multiple compact, highly

FIGURE 1 Stereo fabric concept. For color version of this figure, the reader is referred to the online version of this book.

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00058-7 Copyright Ó 2013 Elsevier Inc. All rights reserved.

1127

1128

Handbook of Advanced Ceramics

FIGURE 2 Prototype on the basis of stereo fabric concept. For color version of this figure, the reader is referred to the online version of this book.

functional, and hollow units e such as LEGO blocks e to improve the energy efficiency of manufacturing plants and also product quality (Figure 1). The key feature is that it increases the degree of freedom in the structural design of ceramics, particularly for large ceramic components. The concept is referred to as stereo fabric modeling [1]. It is possible to produce a wide variety of large components by integrating small units having complex shapes. The expected effects are as follows: 1. 2. 3. 4.

Hollow structure / Weight reduction Decreased material and energy consumption Downsizing of manufacturing facilities Scope for novel shapes and structures

Several such units have so far been designed. Figure 2 shows a few prototype models that can be built using the stereo fabric concept. Several kinds of ceramic members used in manufacturing systems were considered and model parts were fabricated with the proposed concept.

FIGURE 3 Model parts for large, precise, and complicated parts. For color version of this figure, the reader is referred to the online version of this book.

Figure 3 presents an example of a large, yet light, board with small surface protrusions of diameter 0.5 mm and height 0.68 mm. Each unit is hollow and has a truss structure to reduce weight and ensure rigidity. The protrusions reduce the contact area with the work piece, which in this case will be molten or semisolid metal. The units have interlocking structures on adjacent faces, formed during the green stage, that allow them to be assembled very easily. After assembly during the green stage, the units are dewaxed and sintered in nitrogen gas. Adhesion tests were conducted using the sessile drop method on the final board. A barrel-shaped copper sample was placed near the center of the ceramic substrate with surface protrusions, and the sampleesubstrate system was inserted into the apparatus. As the temperature was gradually increased, the copper sample melted at 1100  C, became spherical, and then stabilized. The sample was maintained at this temperature for approximately 10 min before the temperature was increased to the maximum temperature of 1185  C. During this final stage, images of the sample were acquired using a CCD camera (Figure 4). The stereo fabric modeling concepts constitute the core technology of the new Japanese National Project called “Innovative Development of Ceramics Production

FIGURE 4 A molten sample placed on the substrate.

Chapter | 11.2.1

Stereo Fabric Modeling Technology in Manufacturing Ceramics

Technology for Energy Saving project,” which was initiated in 2009 and will continue through to 2014. This project involves the development of joining techniques and nearnet forming processes, aiming at reducing energy consumption throughout both the use and manufacture of ceramics. It is hoped that through this project this stereo fabric concept will be used, in collaboration with commercial

1129

enterprises, to design a novel large-scale ceramic member for application to various fields and thus contribute to efficiency improvements in the manufacturing industry.

REFERENCES [1] Kita H, Hyuga H, Kondo N. Stereo fabric modeling technology in ceramics manufacture. J European Ceram Soc 2008;28: 1079e83.

Chapter 11.2.2

Porous Ceramic Materials Tatsuki Ohji National Institute of Advanced Industrial Science and Technology, 2266-98 Anagahora, Shimoshidami, Moriyama, Nagoya, Aichi 463-8560, Japan

Chapter Outline 1. 2. 3. 4.

Introduction Partial Sintering Sacrificial Fugitives Replica Templates

1131 1133 1135 1139

1. INTRODUCTION Porous ceramics are now used for wide variety of industrial applications from filtration, absorption, catalysts, and catalyst supports to lightweight structural components. In these decades, a great deal of research efforts has been devoted for tailoring deliberately sizes, amounts, shapes, locations, and connectivity of distributed pores, which have brought improved or unique properties and functions of porous ceramics [1e12]. The merits in using porous ceramics for these applications are generally a combination of intrinsic properties of ceramics themselves and advantages of dispersing pores into them. The former include heat and corrosion resistances, wear and erosion resistance, unique electronic properties, good bioaffinity, low density, and high specific strength, and the latter are low density, low thermal conductivity, controlled permeability, high surface area, low dielectric constant, and improved piezoelectric properties [13,14]. This chapter intends to deals with the recent progress of porous ceramics. Porous materials are classified into three classes depending on the pore diameter, d: macroporous (d > 50 nm), mesoporous (50 nm > d > 2 nm), and microporous (d < 2 nm), according to the nomenclature of IUPAC (International Union of Pure and Applied Chemistry). Figure 1 shows this classification along with typical applications and fabrication processes specific to the pore diameters. One of the most representative applications of porous materials is filtration or separation of matters in fluids. Filtration is roughly classified into several classes depending on pore diameter, d, and molecular weight

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00059-9 Copyright Ó 2013 Elsevier Inc. All rights reserved.

5. Direct Foaming 6. Gas Permeability 7. Summary References

1140 1141 1142 1143

cut-off of the matters (MWCO); filtration (typically d > 10 mm), microfiltration (10 mm > d > 100 nm), ultrafiltration (100 nm > d > 1 nm, MWCO ¼ 103e106), nanofiltration (d z 1e2 nm, MWCO ¼ 200e103), and reverse osmosis (d < 1 nm, MWCO z 100). In filtration and microfiltration where pore size is relatively large, the separation is principally made by the sieving effect where matters whose size is larger than the pore size is trapped. On the other hand, in ultrafiltration, nanofiltration, and reverse osmosis where pore size is relatively small, fluid permeability depends on the affinity of solute and solvent to the porous materials as well. Because of the great deal of research work reported in this field these days, this chapter mainly focuses on macroporous ceramics; micro- and mesoporous ceramics whose pore size is below 50 nm are not included here. Representative applications of macroporous ceramics are briefly described. Ceramic filters are now widely loaded in diesel engines to trap particulate matters in the exhaust gas stream, so called, diesel particulate filters (DPFs). Since the high combustion efficiency and low carbon dioxide emission of diesel engines, the demand of DPF is also expected to further increase the world over [15e17]. Ceramic water purification filters are used for eliminating bacteria and suspension from wastewater, because of their higher flux capability, sharper pore-size distribution, better durability, and higher damage tolerance than those of organic hollow fibers [18]. Ceramic foam filters have been employed for removing metallic inclusions from molten metals such as cast iron, steel, and aluminum, as well as rectifying flow of

1131

1132

Handbook of Advanced Ceramics

devices. For example, porous piezoelectric ceramics show good piezoelectric property and are expected to be used for ultrasonic transducers, etc. [23] A variety of porous ceramics have been applied as materials for refractory bricks of kilns and furnaces in various industrial fields, due to their low thermal conductivity and high thermal shock resistance [24,25]. On the other hand, some porous materials of conductive ceramics like zirconia and silicon carbide have been utilized in heat exchangers and heaters [13]. As is known from Figure 1, the representative processes for making macroporous ceramics are (1) partial sintering, (2) sacrificial fugitives, (3) replica templates, and (4) direct foaming. A number of innovative techniques which have been developed recently for critical control of pores are introduced, divided into these four categories, together with some important properties of porous ceramics obtained in these processes. It should be noted however, that many new approaches for macroporous ceramics such as phase separations [26e30] have been developed other than the processes shown here. Figure 2 shows schematic illustrations of these processes, each of which will be interpreted in its section. We then discuss gas permeability of these porous ceramics with different pore sizes and structures in FIGURE 1 Classification of porous materials by the pore size and corresponding typical applications and fabrication processes.

the molten metals [19]. Since the metallic inclusions results in defects in cast metals, this filtration process substantially improves the performance of the products. Porous ceramics with high specific surface area are adopted for absorptive and catalytic applications, where larger contact area with reactants is preferred, particularly in high temperature or corrosive atmospheres. Bioreactors are devices or systems that provide a biologically active environment, where microorganisms and enzymes are immobilized, and biochemical reactions are performed in porous beds, and porous ceramics are often used as such bioreactor beds due to chemical stability of ceramics and accommodative function of porous structure [20]. Recently, porous bioceramics with open-pore structures have attracted great attention for bioimplant applications including of bone regeneration [21]. Bone cells are impregnated through the open pores and grow on their biocompatible walls resulting in bone ingrowth. Many electrodes used in electro-chemical devices including gas purifiers, gas sensors, fuel cells, and chemical analyzers are porous ceramics [22]. Some porous electrodes require two mode distributions of pore sizes; small pores are for the electro-chemical reactions while large ones are for flow paths of reactants. Properties of electroceramics also depend substantially on the porosity content and morphology and therefore porous ceramics are also applied or expected to be used in various electro-

FIGURE 2 Representative fabrication processes of macroporous ceramics.

Chapter | 11.2.2

Porous Ceramic Materials

1133

Section 6. Finally, a summary will be given with the issues to be solved for further realizing the potential of porous ceramics and for expanding their applicability.

2. PARTIAL SINTERING Partial sintering of powder compact is one of the most conventional and frequently employed approaches to fabricate porous ceramic materials. Particles of powder compact are bonded due to surface diffusion or evaporationecondensation processes enhanced by heat treatments, and a homogeneous porous structure forms when sintering is terminated before being fully densified (see Figure 2 (a)). Pore size and porosity can be controlled by the size of starting powders and degree of partial sintering, respectively. Generally, in order to provide the desired pore size, the size of raw powder should be geometrically in the range two to five times larger than that of the pore. Porosity decreases with increased forming pressure, sintering temperature, and time. In addition, processing factors such as the type and amount of additives, green densities, and sintering conditions (temperature, atmosphere, pressure, etc.) also greatly affect the microstructures of porous ceramics [31]. The mechanical properties depend significantly on degree of neck growth between grains, as well as porosity and pore size. For example, the formation of necks between touching particles by surface diffusion without densification can increase the elastic modulus to 10% of the fully dense value [32,33]. The porosities of porous materials obtained by partial sintering are usually below 50%. In industry, this method has been utilized for various applications including molten metal filters, aeration filters (gas bubble generation in wastewater treatment plants) [13], and water purification membranes [18]. Several processing approaches have been developed to enhance neck growth between grains and improve strength of porous ceramics. Oh et al. [34], Jayaseelan et al. [35], and Yang et al. [36] fabricated porous Al2O3 and Al2O3-based composites by the pulse electric current sintering (PECS) technique and found that the strength was substantially improved due to the formation of thick and strong necks. During sintering, the discharge between the particles is thought to promote the bridging of particles by neck growth in the initial stages of sintering. This strong neck growth leads to substantially high strength compared to those of the conventional porous materials. For example, the flexural strength of porous alumina-based composites via PECS reached 250 and 177 MPa, with 30% and 42% porosity, respectively, which are considerably high compared to those of porous alumina fabricated by conventional partial sintering, e.g., ~100 MPa at 30% porosity [35] (see Figure 3). Using PECS, Akhtar et al. [37] also fabricated porous ceramic monoliths from diatomite powders, which are known as a cheap and renewable, natural resource.

FIGURE 3 Flexural strength as a function of porosity for alumina/ 3 vol.% zirconia (AZ) fabricated via PECS and conventionally sintered alumina (top) and microstructure of AZ (bottom) [35]. Strong neck growth of the AZ results in substantially high strength compared to those of the conventional porous materials. (Reprinted with permission of John Wiley and Sons. All rights reserved)

PECS that rapidly heats diatomite powder successfully bonds the particles together into relatively strong porous bodies, without significantly destroying the internal pores of the diatomite powder. The microstructural studies revealed that consolidation proceeds by the formation of necks at temperatures around 700e750  C, which is followed by significant melt phase formation around 850  C, resulting in porous ceramics with a relatively high strength. Deng et al. [38,39] tried to obtain strong grain bonding through the combination of partial sintering and powder decomposition. A mixture of a-Al2O3 and Al(OH)3 was used as the starting powder to make porous Al2O3 ceramics, and because Al(OH)3 experiences a 60% volume contraction during decomposition and produces fine Al2O3 grains, the fracture strength of obtained porous Al2O3 was substantially higher than that of the pure Al2O3 sintered specimens because of strong grain bonding that resulted from the fine Al2O3 grains produced by the decomposition of Al(OH)3. Similar improvement of mechanical properties was also identified for ZrO2 porous ceramics fabricated by adding Zr(OH)4 [40]. Partial sintering through reaction bonding techniques have been frequently used for making porous ceramics,

1134

FIGURE 4 Microstructure of porous CaZrO3/MgO composite fabricated via in-situ reaction synthesis, exhibiting 3-dimensional network structure with strong grain necking [43]. (Reprinted with permission of John Wiley and Sons. All rights reserved)

where reaction products form or precipitate epitaxially on grains, resulting in well-developed neck growth between grains [41,42]. In combination with the reactive sintering process, Suzuki et al. [43,44] synthesized a CaZrO3/MgO porous ceramic with three-dimensional grain network structure; using reactive sintering of highly pure mixtures of natural dolomite (CaMg(CO3)2) and synthesized zirconia powders. CaMg(CO3)2) decomposes into CaCO3, MgO, and CO2 (g) at ~500  C, and CaCO3 then reacts with

FIGURE 5 Microstructures of two porous silicon carbides fabricated via an oxidation-bonding process, and their pore-size distributions determined by mercury porosimetry (V: intrusion volume) [47]. (a) From fine (0.6 mm) powder. Porosity is 31%. (b) From coarse (2.3 mm) powder. Porosity is 27%. (Reprinted with permission of John Wiley and Sons. All rights reserved)

Handbook of Advanced Ceramics

ZrO2 to form CaZrO3 and CO2 (g) at ~700  C. Through liquid formation via LiF doping, these reactions and liberated CO2 gas result in formation of a homogeneous open-pore structure with strong grain bonding as shown in Figure 4. The pore-size distribution is very narrow (with typical pore size: ~1 mm), and the porosity was controllable (~30e60%) by changing the sintering temperature. The relatively high flexural strength (~40 MPa for 47% porosity) was observed over the temperature range of R.T. 1300  C. The similar approach has been applied other materials systems such as CaAl4O7/CaZrO3 and CaZrO3/MgAl2O4 composite systems [45,46]. She et al. [47] used an oxidation-bonding process for the low-temperature fabrication of porous SiC ceramics with superior resistance against oxidation. In such a process, the powder compacts are heated in air instead of an inert atmosphere. The heating temperature was kept below 1300  C in order to suppress formation of cristobalite. Figure 5 shows microstructures of two porous silicon carbides fabricated through the oxidation-bonding process (1300  C, 1 h), using fine (0.6 mm) and coarse (2.3 mm) a-SiC powders, together with their pore-size distributions. Because of the occurrence of surface oxidation at the heating stage, SiC particles are bonded to each other by the oxidation-derived SiO2 glass. The difference of the poresize distributions arises from the different starting powders. Mechanical strength is strongly affected by particle size; the flexural strength attained as high as 185 MPa at a porosity of 31%, when using the fine powder (Figure 5 (a)),

Chapter | 11.2.2

Porous Ceramic Materials

while it was 88 MPa at 27% porosity for the coarse powder (Figure 5 (b)). The oxidation-bonding technique has been applied to other materials including silicon nitride [48], SiC/mullite composites [49], and SiC/cordierite composites [50]. The partial sintering technique has been also applied for making porous nonoxide ceramics such as porous silicon nitride with fibrous grains of high aspect ratios [51e58]. Compared with oxide ceramics, the densification of silicon nitride ceramics is difficult because of strong covalent bonding between silicon and nitrogen atoms. This difficulty of sintering silicon nitride ceramics is beneficial for controlling density or porosity through adjusting the additives and the sintering process. In order to suppress densification, oxides with high melting point and high viscosity such as Yb2O3 are frequently used as sintering additives [52]. Depending on the sintering temperature, the addition of Yb2O3 also accelerates the fibrous grain growth of b-Si3N4, which substantially affects the mechanical properties of porous silicon nitrides [52,53]. Due to differences in the melting points and the preferential absorption sites of cations, the porous structure is substantially affected by the types of sintering additives as well [54]. Yang et al. [55] directly synthesized porous b-Si3N4 ceramics by carbothermal nitridation of silica, using carbon black as the carbon source and a-Si3N4 as seeds. The complete reaction results in a large weight loss and high porosities (about 65e70%) after sintering. Fine elongated fibrous b-Si3N4 grains were developed in the seeded samples while only large equiaxial grains were observed in the seed-free samples, as shown in Figure 6. The former sample exhibited relatively high flexure strength close to 40 MPa for the high porosity of ~65%, which is five times higher than that of the latter one. Tuyen et al. [56] fabricated porous reaction-bonded silicon nitride (RBSN) by nitridation process at 1350  C and post-sintering at 1550e1850  C, which provides similar fibrous microstructure and high porosity. The sintering time had a significant effect on the microstructure and grain morphology, and porous structure with fibrous grains of high aspect ratios was obtained by adjusting the time even at the comparatively low temperature (1550  C). These techniques offer the possibility of synthesizing highly porous and strong Si3N4 materials at considerably lower cost. One of the unique processing routes for porous ceramics with anisotropic microstructure is tape-casting fibrous seed crystals or whiskers. For porous silicon nitrides, b-Si3N4 seed crystals were mixed with sintering additives as starting powders, and the green sheets formed by tape casting were stacked and bonded under pressure, followed by sintering at 1850  C under a nitrogen pressure of 1 MPa [57,58]. The microstructural observation for the porous silicon nitrides revealed that the fibrous grains are aligned toward the

1135

FIGURE 6 Microstructures of two porous silicon nitrides fabricated via carbothermal nitridation of silica [55]. (a) Seeded sample with fine elongated fibrous grains. (b) Seed-free sample with large equiaxial grains. (Reprinted with permission of Elsevier. All rights reserved)

casting direction, and the plate-like pores exist among the grains. Because of the enhanced crack shielding effects of aligned fibrous grains, the anisotropic porous silicon nitrides showed excellent mechanical behavior, when a stress is applied in the alignment direction. More detailed discussion on the microstructure changes and mechanical properties (strength, fracture toughness, and thermal shock resistances) for the isotropic and anisotropic porous silicon nitrides will be given in the Chapter “Microstructural Control and Mechanical Properties.”

3. SACRIFICIAL FUGITIVES Porous ceramics are often fabricated by mixing appropriate amounts of sacrificial fugitives as pore-forming agents with ceramic raw powder and evaporating or burning out them before or during sintering to create pores (see Figure 2 (b)). Frequently used pore-forming agents are polymer beads, organic fibers, potato starch, graphite, charcoal, salicylic

1136

acid, carbonyl, coal, and liquid paraffin. The pore-forming agents are generally classified into synthetic organic matters (polymer beads, organic fibers, etc.) [59e89], natural organic matters (potato starch, cellulose, cotton, etc.) [67,68,90e105], metallic and inorganic matters (nickel, carbon, fly ash, glass particles, etc.) [49,78,106e114], and liquid (water, gel, emulsions, etc.) [115e155]. Porosity is controllable by the amount of the agents and pore shape and size are also affected by the shape and size of the agents, respectively, when their sizes are large in comparison with those of starting powders or matrix grains. The agents, however, need to be mixed with ceramic raw powder homogeneously for obtaining uniform and regular distribution of pores. Solid fugitives like organic materials are usually removed through pyrolysis, which often requires long-term heat-treatment and generates a great deal of vaporized, sometimes harmful, by-products. Polymethylmethacrylate (PMMA) beads and microbeads have been frequently employed for sacrificial fugitives [8,59e64,77,79,82e85]. Colombo and his co-workers [8,59e61] fabricated SiOC ceramic foam by dry mixing the silicon resin powder with a sacrificial template constituted by PMMA microbeads, and subsequent heat treatments (Figure 7). Descamps et al. [63,64] produced macroporous b-tricalcium phosphate (TCP) ceramics using PMMA. An organic skeleton was formed by interconnecting the PMMA balls through a chemical superficial dissolution, and was impregnated by the TCP slurry. PMMA was then eliminated by a thermal treatment at low temperature, followed by sintering for final porous structure. This process allows a total control of the porous architecture; the porous volume can vary from 70 to 80%

FIGURE 7 SiOC ceramic foam using a sacrificial template constituted by PMMA microbeads [8]. (Reprinted with permission of Elsevier. All rights reserved)

Handbook of Advanced Ceramics

and the interconnection size from 0.2 to 0.6 times the average macro-pore size. Andersson et al. [82] used expandable microspheres as a sacrificial template to produce macroporous ceramic materials by a gel-casting process. The microspheres consist of a co-polymer shell and are filled with a blowing agent (isobutane), which allows rapid and facile burn out. By controlling the amount and size of the expandable microspheres, it is possible to tune the porosity up to 86% and the pore-size distribution from 15 to 150 mm. Up to 1e2 wt.% of the microspheres leads to a final porosity above 80 vol%. Expandable microspheres as sacrificial templates, rather than other templates such as PMMA microbeads, are advantageous because of lower levels of gaseous by-products generated during pyrolysis, and lower cost of the overall materials. Kim and his co-workers [83,84] used hollow microspheres as sacrificial templates to make porous silicon carbide ceramics synthesized from carbon-filled polysiloxane and others. Using preceramic polymer and organic microspheres for fabricating porous ceramics allows use of the low-cost and/or near-net-shaped processing techniques like extrusion and direct casting [84]. Song et al. [85] produced microcellular silicon carbide ceramics with a duplex pore structure using expandable microspheres and PMMA spheres, which resulted in the large pores and the small windows in the strut area, respectively. This porous ceramics showed excellent air permeability as shown in Section 6. Diaz et al. [93,94] fabricated porous silicon nitride ceramics using a fugitive additive, cornstarch (particle size: 5e18 mm). In order to obtain homogeneous dispersion of the fugitives, the mixture slurry was kept in agitation using a magnetic stirrer for a while, and then was frozen and dried under vacuum for sieving. Kim et al. [95] mixed various amounts of cornstarch to (Ba, Sr) TiO3 powder to obtain (Ba, Sr) TiO3 porous ceramics. They found that depending on the porosity, the PTCR effect was 1e2 orders of magnitude improved in comparison with the dense reference. Chen et al. [66] produced porous silicon nitride of equiaxed a-grains by using phosphoric acid (H3PO4) as the pore-forming agent and relatively low temperature (1000e1200 C) sintering. On the other hand, Li et al. [80] fabricated porous silicon nitride with fibrous b-grain structure, using naphthalene powder as the pore-forming agent and gas-pressure sintering of high temperatures above 1700  C. The bending strength of the former materials was 50e120 MPa in porosity range of 42e63%, while that of the latter was 160e220 MPa in porosity range of 50e54%. This substantial difference in strength is attributable to the microstructural difference (equiaxed vs. fibrous), similar to the case of Figure 5. Ding et al. [49] used graphite as the pore-former to fabricate mullite-bonded porous silicon carbide ceramics in air from SiC and a-Al2O3 through an in-situ reaction

Chapter | 11.2.2

Porous Ceramic Materials

1137

bonding technique. Graphite is burned out to form pores and SiC is oxidized at high temperatures to SiO2, which further reacts with a-Al2O3 to form mullite (3Al2O3$ 2SiO2). SiC particles are bonded by mullite and oxidationderived SiO2. Long fibers such as cotton thread [96], natural tropical fiber [97], and metal wires [109] are often used as poreforming agents for obtaining porous ceramics of unidirectional through channels. Zhang et al. [96] fabricated porous alumina ceramics with unidirectionally aligned continuous pores (diameter: ~160 mm) via the slurry coating of mercerized cotton threads. The pore size can be adjusted, using cotton threads of different diameter, and the porosity can be controlled by changing the solids’ concentration of the slurry. In this case, excellent permeability can be achieved for porous ceramics with unidirectional through channel pores, because gas can flow directly through the pores. However, the preparation of such ceramics is complex because handling long fibers such as thin wire or cotton thread is difficult. Using short fibers or whiskers as the pore-forming agent is an alternative that combines the advantages of partially sintered porous ceramic and those of unidirectional pores. Yang et al. [65] demonstrated formation of rod-shaped pores in silicon nitride ceramics, using slip casting of aqueous slurries of silicon nitride powder and sintering additives with 0e60 vol% fugitive organic whiskers. Rheological properties of slurries were optimized to achieve a high degree of dispersion with a high solid-volume fraction. Samples were heated at 800  C in air to remove the whiskers and sintered at 1850  C in nitrogen atmosphere to consolidate the matrix. Porosity was adjusted in 0e45% by changing the whisker content in 0e60 vol%. The obtained porous silicon nitride contained uniform rod-shaped pores with random directions as shown in Figure 8, and therefore exhibited relatively high gas permeability in comparison to porous silicon nitride containing equiaxed pores [156]. Isobe et al. [81,110] and Okada et al. [86,87] used carbon fibers (14 mm

diameter and 600 mm length) or Nylon 66 fibers (9.5e43 mm diameter and 800 mm length) as a pore-forming agent, and tried to align them by an extrusion technique to produce porous alumina [81,110] and mullite [86,87] ceramics with unidirectionally oriented pores. The pore sizes and porosities can be controlled by varying the fiber diameter and fiber content, and the obtained samples showed better air permeability than the conventional porous materials used for filter applications [110]. This technique can allow the production of highly oriented porous ceramics by an industrially favored extrusion method. Liquids such as water and oils, which are readily sublimated or evaporated, are often used as pore-forming agents [115e156]. One of the most frequently employed approaches in recent years is freeze-drying the water- or liquid-based slurry to produce porous ceramics of unique structure [118e154]. Figure 9 shows a schematic

FIGURE 8 Porous silicon nitride with rod-shaped pores, prepared using 50 vol% fugitive organic whiskers of 33 mm diameter. The obtained porosity is ~40% and the mean pore diameter is ~22 mm.

FIGURE 9 Schematic illustration of the freeze-dry process for macroporous ceramics and a porous silicon nitride body obtained thereby [120]. (Reprinted with permission of John Wiley and Sons. All rights reserved)

1138

illustration of the procedures which Fukasawa et al. [118e120] employed, and a porous silicon nitride body obtained thereby. When the bottom part of the slurry is frozen, ice grows macroscopically in the vertical direction, and pores are generated subsequently by sublimation of the ice. Sintering this green body results in a porous ceramic with unidirectionally aligned channels, which contain smaller pores in the internal walls (Al2O3) [118,119] or fibrous grains protruding from them (Si3N4) [120]. The advantages of this approach include a simple process without materials to be burnt out, a wide range of porosity (30e99%) controllable by the slurry concentration, applicability to various types of ceramics, and environmental friendliness without emitting harmful products. Particularly porous scaffolds with ice-designed channel-like porosity fabricated by this method have been intensively studied for a wide variety of applications including biomedical implants and catalysis supports. The porosity of the porous materials obtained using this technique is a replica of the original ice structure. The porous channels run from the bottom to the top of the samples (when the bottom part is first frozen), and the pores most frequently exhibit an anisotropic morphology in the solidification plane. Deville et al. [127e130] investigated the relationships between the freezing conditions and the final porous structures in freeze casting of ceramic slurries. It has been clarified that the morphology of the porous structures including the dimensions, shape, and orientation of porosity are adjustable by varying the initial slurry compositions and the freezing conditions. For highly concentrated solutions, the particleeparticle interactions lead to the formation of ceramic bridges between two adjacent lamellae. Using this technique, they fabricated sophisticated porous and layered-hybrid materials such as nacre-like structure with lamellar dendrites and high compressive strength (four times higher than those of materials currently used for implantation). Munch et al. [131] emulated nature’s toughening mechanisms by combining two ordinary compounds, alumina and polymethyl methacrylate, into an ice-templated structure, and succeeded in obtaining toughness more than 300 times (in energy terms) that of their constituents. Araki and Halloran [132e134] used camphene, C10H16, as a vehicle for producing porous ceramics via a freezedrying process, to realize a freezing process at room temperature. Slurries containing ceramic powder in the molten camphene were prepared at 55  C, and were quickly solidified (frozen) when they were poured into polyurethane molds at room temperature. The obtained porous ceramics have pore channels of nearly circular cross sections (unlike ellipsoidal ones obtained via conventional aqueous freeze casting). Koh and his co-workers used a similar camphene-based freeze-casting approach to fabricate highly porous Al2O3 [135e137], SiC [138,139],

Handbook of Advanced Ceramics

PZT-based ceramics [140,141], hydroxyapatite [142,143], glass-ceramics [144], and ZrO2 [145,146], etc., having interconnected pore without noticeable defects. Many researchers have tried to combine the freeze-dry process and the gel-casting technique to produce porous ceramics with refined microstructure [147e155]. It has been shown that the use of an organic polymer in the freeze-casting route affects the pore size and morphology by controlling ice crystal growth during freezing. Chen et al. [147,148] used alumina slurries containing tert-butyl alcohol (TBA) and acrylamide (AM) for the freeze-dry process. TBA freezes below 25 C and volatilizes rapidly above 30  C, acting as the freezing vehicle and template for forming pores, while AM is polymerized in the slurry as the gelation agent, strengthening the green bodies substantially. The sintered porous ceramics have high compression strength (~150 MPa at 60% porosity) because the pore channels formed by the TBA template are surrounded by almost fully dense walls without any noticeable defects. Ding et al. [149] also employed a gel freeze-drying process to fabricate porous mullite ceramics with porosity up to 93%. Alumina gel mixed with ultrafine silica was frozen isotropically, followed by sublimation of ice crystals. Porous mullite ceramics were prepared in air at 1400e1600  C due to the mullitization between Al2O3 and SiO2. Porous yttria-stabilized zirconia (YSZ) [150] and porous alumina [151] were fabricated by the freeze-drying process with addition of polyvinyl alcohol (PVA), which suppresses ice crystal growth and reduces the pore sizes substantially. Porous alumina with oriented pore structures has been also fabricated by the freeze-casting technique with a water-soluble polymer such as polyethylene glycol (PEG) [152]. Using precursor silica hydrogels, Nishihara et al. [153] fabricated ordered macroporous silica (silica gel micro-honeycomb) using freeze-dry methods, where micrometer-sized ice crystals are used as a template. The pore sizes can be controlled by changing the immersion rate into a cold bath and the freezing temperature. The average pore size can be as small as 4.7 mm with the rate of 20 cm/h at 77 K, and the thickness of the honeycomb walls can be adjusted by the SiO2 concentration. Fukushima et al. [154,155] fabricated porous cordierite or silicon carbide ceramics with porosity from 80 to 95% using a gel-freezing method; unidirectionally oriented cylindrical channels are uniformly distributed over relatively large bulk samples (typically several centimeters). Gelatin was used as the gelation agent, which was mixed with water for the freezing vehicle and raw powder. The gel was frozen at e10 to e70  C, dried under vacuum, and degreased before sintering was carried out (at 1200e1400  C for cordierite and at 1800  C for silicon carbide). The cell size and cell wall thickness both decreased with decreasing the freezing temperature, from 200 to 20 mm and from 20 to 3 mm, respectively. The

Chapter | 11.2.2

Porous Ceramic Materials

FIGURE 10 Microstructures of porous silicon carbide prepared by the freeze-dry technique using gelatin as the gelation agent frozen at e10  C.

number of cells for porous cordierite frozen at e50  C and sintered at 1400  C was 1500 cells/mm2 in the cross section; this is remarkably large in comparison to those of samples obtained by the extrusion method: 1e2 cells/mm2. The dense cell walls (Figure 10) lead to relatively high compressive strength, for example, 17 MPa for 86% porosity sample of silicon carbide.

4. REPLICA TEMPLATES Macroporous ceramics having interconnected large pores, or channels, of high porosity have been frequently fabricated by the replica techniques (Figure 2 (c)). The first step is typically impregnation of a porous or cellular template with ceramic suspension, precursor solution, etc. Various synthetic and natural cellular structures are used as the templates. The most frequently used synthetic template is porous polymeric sponge such as polyurethane. They are soaked into a ceramic slurry or precursor solution to impregnate the templates with them, and the surplus is drained and removed by centrifugation, roller compression, etc. In this process, appropriate viscosity and fluidity depending on the cell size, etc., are required in order to obtain uniform ceramic layer over the sponge walls. The ceramic-impregnated templates are dried and then heattreated to decompose the organic sponges, followed by sintering the ceramic layers at higher temperatures. Porosity higher than 90% can be obtained with cell sizes ranging from a few hundred micron meters to several millimeters. The open cells are interconnected, which allows fluid to pass through the foams with a relatively low pressure drop. Figure 11 shows a typical example of alumina foam prepared by slurry infiltration of polyurethane templates, which has been reported by Faure et al.

1139

[157]. However, due to cracking the struts during the pyrolysis, the mechanical properties of ceramic reticulated foams are generally poor. In order to avoid strut crack formation, various approaches have been made [157e163]. For examples, Vogt et al. [159] employed vacuum infiltration of ceramic slurry to fill up the struts in the presintered foam. The hollow struts caused by burnout of the polyurethane template were completely filled up, resulting in a considerable increase of compressive strength. Jun et al. [161,162] produced hydroxyapatite scaffolds coated with bioactive glass-ceramics using the polymer foam replication method, to enhance their mechanical properties and bioactivities. Highly porous ceramics can be also obtained from preceramic polymers after pyrolysis above 800  C in inert atmosphere [8]. One of the typical methods is dissolving the silicone resin preceramic polymer into a suitable solvent and adding appropriate surfactants and catalysts, followed by pyrolysis. The advantages are a wide control of pore sizes (typically 1 mm to 2 mm), well-defined open-cell structures, and macro-defect-free struts [164e167]. Travitzky et al. [168] also succeeded in fabricating single-sheet, corrugated structures, and multilayer ceramics by using various paper replica templates. Natural resources of porous structures such as woods, corals, sea sponge, etc., can be also used as replica templates. The woods are transformed to carbonaceous preforms by heat-treatment in inert atmosphere. They are then infiltrated with molten metals [169e179], gaseous metals [174,180e185], alkoxide solutions [186e188], and others [189,190]. The advantages include a wide variety of obtained porous structures (depending on the type of wood selected), low-cost starting materials, nearnet and complex shape capabilities, and a relatively lowtemperature fabrication process. Locs et al. prepared porous SiC ceramics from pyrolyzed pine wood samples via impregnation with SiO2 sol and heat-treatment at 1600  C for 4 hours, as shown in Figure 12 [190]. The longitudinal pore size in SiC is 10e20 mm and the wall thickness is 3e5 mm. The oriented vessels of the woods provide unique anisotropic porous structure of aligned unidirectional through channels, which is suitable to applications such as filtration and catalysis supports. Porous biomimetic silicon carbides obtained through this approach have been also studied for medical implant materials [191]. Biomorphic porous silicon nitride was produced from natural sea sponge via replication method. The sponges were impregnated with silicon-containing slurry via dip coating, and were heat-treated to delete the bio-polymers, leading to a Si-skeleton. Subsequent thermal treatment under flowing nitrogen promoted the nitridation of the silicon, porous a/b-silicon nitride with a porosity of 88%, and the original morphology of the sea sponge [192].

1140

Handbook of Advanced Ceramics

FIGURE 11 Alumina foam preparation by slurry infiltration of polyurethane templates [157]. (a) Polyurethane foam; (b) impregnated and dried foam; and (c) dense alumina foam. For color version of this figure, the reader is referred to the online version of this book. (Reprinted with permission of Elsevier. All rights reserved)

5. DIRECT FOAMING In direct foaming techniques, the ceramics suspension is foamed by incorporating air or gas and stabilized and dried, followed by sintering to obtain a consolidated structure (Figure 2 (d)). The advantage of this technique is low-cost and easy fabrication of highly porous ceramic materials (95% or higher porosity). Porous ceramics with unidirectional channels can be produced using continuous bubble formation in ceramic slurry [193,194]. However, due to the thermodynamic instability, the gas bubbles easily coalesce in order to reduce the total Gibbs free energy of the system, which results in undesirable large pores. It is, therefore, critically needed to stabilize the air or gas bubbles in ceramic suspension. One of the most frequently approaches for the stabilization is to use surfactants reducing the interfacial energy of the gaseliquid boundaries. Surfactants used for stabilization are classified into several types including nonionic, anionic, cationic, and protein, and the pore size of the produced porous body

ranges from below 50 mm up to the millimeter scale, depending on the used surfactants. A variety of effective surfactants have been developed for direct foaming of porous ceramics [195e201]. Barg et al. [198e200] developed a novel direct foaming process by emulsifying a homogeneously dispersed alkane or airealkane phase in the stabilized aqueous powder suspension. Foaming is made by evaporation of the emulsified alkane droplet, leading to high performance ceramic foams with porosities up to 90% and cell sizes ranging from 3 to 200 mm. This autonomous foaming process also allows high flexibility in the production of ceramic parts with gradient structures and complex shaping. Foaming proceeds as a consequence of the evaporation of the alkane phase resulting in the growth of the stabilized alkane bubbles and in a volume increase of the foam. Figure 13 shows an example of the alumina foam which was produced through sintering (1550 C/2 h) from high alkane phase emulsified suspensions containing 45 vol% particle content [200].

Chapter | 11.2.2

Porous Ceramic Materials

1141

FIGURE 13 Microstructure of sintered alumina foams (1550  C/2 h) produced from high alkane phase emulsified suspensions containing 45 vol% particle content. (a) Open interconnected porous cells and (b) Dense strut constituted by a monolayer of alumina particle [200]. (Reprinted with permission of Elsevier. All rights reserved)

FIGURE 12 Porous SiC ceramics prepared from pyrolyzed pine wood samples via impregnation with SiO2 sol and heat-treatment at 1600  C for 4 hours. (a) Cross-sectional view and (b) The walls’ outer layer is denser than the inner zone [190]. (Reprinted with permission of Elsevier. All rights reserved)

Preceramic polymer solution has been also used instead of ceramic suspension for direct foaming. Colombo et al. [202] produced porous ceramics by dissolving preceramic polymers (silicone resins) into a suitable solvent with blowing agent, surfactant, catalyst, etc., and heat-treating them at 1000e1200  C in inert atmosphere. Due to the suppressed defect formations in the struts, the strength of the obtained porous ceramics was relatively high in comparison to those of conventional reticulated foams [203]. Kim et al. [204,205] fabricated porous ceramics with a fine microcellular structure from preceramic polymers using CO2 as a blowing agent. A mixture of polycarbosilane and polysiloxane was saturated with gaseous CO2 under a high pressure and bubbles were introduced using a thermodynamic instability via a rapid pressure drop, followed by pyrolysis and sintering. It has been shown that particles with tailored surface chemistry can also be used efficiently to stabilize gas bubbles

for producing stable wet foams [206e214]. Gonzenbach et al. [210e214] have developed a novel direct foaming method that uses colloidal particles as foam stabilizers. The method is based on the in-situ hydrophobization of initially hydrophilic particles to enable their adsorption on the surface of air bubbles. In-situ hydrophobization is accomplished through the adsorption of short-chain amphiphiles on the particle surface. The obtained ultra-stable wet foams show neither bubble coalescence nor disproportionation over several days, as opposed to the several minutes typically required for the collapse of the surfactant-based foams. Because of their remarkable stability, the particlestabilized foams can be dried directly in air without crack formation. The macroporous ceramics obtained after sintering have porosities from 45% to 95% and cell sizes between 10 and 300 mm. The compressive strength of the sintered foams with closed cells is relatively high in comparison with those of foams prepared with other conventional techniques (for example, 16 MPa at a porosity of 88% in alumina foams). It has been observed that the surface-modified particles which originally cover the air bubble in wet foams become a thin surface layer of single grains after sintering. Macroporous ceramics with open porosity can be also fabricated using this technique when decreasing the concentration of stabilizing particles.

6. GAS PERMEABILITY Gas permeability is one of the most important properties of porous ceramics which are expected to be used for gasfilters such as diesel particulate filter (DPF), since large pressure drops should be avoided in such applications.

1142

Handbook of Advanced Ceramics

Highly porous ceramics with aligned unidirectionally through pore channels are expected to provide excellent permeability, and as already stated, the freeze-dry technique is one of the most representative processes for producing such porous ceramics. This section deals with Darcian air permeability of porous ceramics with different pore sizes and structures. Figure 14 shows the Darcian permeability as a function of pore size for porous ceramics fabricated by freeze-dry processes [151,155], organic spherical fugitives [61,85], graphite fugitives [215,216], extruded organic fibrous fugitives [81,87], direct foaming [195,217], and replica templates [217,218]. The pore structures are classified into three categories of “Spherical (Connected)” [61,85,195,215e218], “Cylindrical (Connected)” [81,87] and “Cylindrical” [151,155], as schematically shown in the figure. The Darcian permeability, K, is determined from pressure drop and flow rate of air by the Darcy’s law [219]. Based on the capillary model, K is expressed by K ¼ fD2p =C

(1)

where f is the porosity, Dp is the pore diameter, and C is a constant depending on the pore structure [219,220].

FIGURE 14 Darcian air permeability as a function of pore size for porous ceramics fabricated by the freeze-dry processes [151,155], in comparison to those of other processes including organic spherical fugitives [61,85], graphite fugitives [215,216], extruded organic fibrous fugitives [81,87], direct foaming [195,217], and replica templates [217,218]. The solid line indicates theoretical permeability K ¼ fD2p =32ðf ¼ 0:85Þ for the case of unidirectional cylindrical pores penetrating in parallel.

K values of Figure 14 are those adjusted from the reported values at the porosity of 0.85 using Eqn (1) for comparison. The inertial contribution (non-Darcian permeability) was considered in addition to the viscous one (Darcian permeability) in Ref. [61,152,217e219], which results in high values of K in comparison to the cases of neglected inertial effect [81,85,87,155,195,215] (the ratio of viscous contribution in total is 60e90% [151]). The solid line of the figure shows the case that the fluid flows through the unidirectional cylindrical pores penetrating in parallel (C is 32) [220]. The porous ceramics fabricated by the freeze-dry processes [151,155] showed permeability very close to this solid line, indicating the unidirectional alignment of the cylindrical pores. The permeability required for a commercially available DPF is 10 11 to 10 12 m2 [221], and most of the freeze-dry-processed materials exceed this criterion. The porous ceramics prepared with extruded organic fibrous fugitives [81,87] showed lower permeability values than those of the freeze-dry-processed ones, most likely because of limited contact area among the short fibers.

7. SUMMARY During the last decade, tremendous efforts have been devoted to research on porous ceramics, resulting in better control of the porous structures and substantial improvements of the properties. This chapter reviewed these recent progresses of porous ceramics. Because of the vast amount of research works reported in this field these days, the chapter mainly focused on macroporous ceramics whose pore size is larger than 50 nm. Followed by giving a general classification of porous ceramics, a number of innovative processing routes for critically controlling pores were described, along with some important properties. They were divided into four categories including (1) partial sintering, (2) sacrificial fugitives, (3) replica templates, and (4) direct foaming. The partial sintering, which is one of the most common techniques for making porous ceramics, has been substantially sophisticated in recent years. Very homogeneous porous ceramics with extremely narrow size distribution has been successfully prepared through sintering combined with in-situ chemical synthesis. Porous silicon nitrides with aligned fibrous grains and pores have demonstrated excellent mechanical properties, which are equivalent, or sometimes superior, to those of the dense materials. The sacrificial fugitives have an advantage that pore shape and size are controllable by the shape and size of the agents, respectively. The fugitives are generally removed through pyrolysis, generating a great deal of vaporized, sometimes harmful, by-products, and a lot of research has been conducted to reduce or eliminate them. The freeze-dry processes using water or liquid as fugitive agents are advantageous in this viewpoint and have been very intensively studied in recent years. Careful control of

Chapter | 11.2.2

Porous Ceramic Materials

ice growth leads to unique porous structures and excellent performances of porous ceramics. The replica template techniques have been widely used to fabricate highly porous ceramics with interconnected large pores. Porous polymeric sponge such as polyurethane is the most typical synthetic template used for this process. However, due to cracking struts during pyrolysis of the sponge, the mechanical property is generally low; a variety of approaches have been used to avoid strut crack formation. Natural template approaches using wood, for example, as positive replica, have been frequently studied in these years and have realized highly oriented porous open-porous structure with a wide range of porosity. The direct foaming technique is low-cost and an easy fabrication process of porous ceramics with high porosity volume. In order to suppress coalescence of gas bubbles in ceramic suspension that results in large pores in the final porous bodies, various methods which stabilize the bubbles have been developed; they include use of effective surfactants, evaporation of emulsified alkane droplets, and use of surface-modified particles. Finally, we discussed gas permeability (Darcian air permeability) of porous ceramics with different pore sizes and structures. It has been demonstrated that the freeze-dry-processed porous ceramics with cylindrical through channels have excellent permeability, which is close to the ideal case that the fluid flows through the unidirectional cylindrical pores penetrating in parallel.

1143

[12]

[13]

[14]

[15] [16]

[17]

[18] [19]

[20]

[21] [22]

[23]

REFERENCES [1] Messing GL, Stevenson AJ. Toward pore-free ceramics. Science 2008;322:383e4. [2] Colombo P. In praise of pores. Science 2008;322:381e3. [3] Kelly A. Why engineer porous materials? Phil Trans R Soc A 2006;364:5e14. [4] Greil P. Advanced engineering ceramics. Adv Mater 2002; 14(10):709e16. [5] Schuth F. Engineered porous catalytic materials. Annu Rev Mater Res 2005;35:209e38. [6] Yuan Z-Y, Su B-L. Insights into hierarchically mesoemacroporous structured materials. J Mater Chem 2006;16:663e77. [7] Colombo P. Conventional and novel processing methods for cellular ceramics. Phil Trans R Soc A 2006;364:109e24. [8] Colombo P. Engineering porosity in polymer-derived ceramics. J Eur Ceram Soc 2008;28:1389e95. [9] Studart AR, Gonzenbach UT, Tervoort E, Gauckler LJ. Processing routes to macroporous ceramics: a review. J Am Ceram Soc 2006;89(6):1771e89. [10] Takahashi M, Menchavez RL, Fuji M, Takegami H. Opportunities of porous ceramics fabricated by gelcasting in mitigating environmental issues. J Eur Ceram Soc 2009;29:823e8. [11] Kumar BVM, Kim Y-W. Processing of polysiloxane-derived porous ceramics: a review. Sci Technol Adv Mater 2010; 11:044303.

[24]

[25]

[26]

[27]

[28] [29]

[30]

[31]

Colombo P, Vakifahmetoglu C, Costacurta S. Fabrication of ceramic components with hierarchical porosity. J Mater Sci 2010; 45:5425e55. Scheffler M, Colombo P. Cellular ceramics, structure, manufacturing, properties and applications. WILEY-VCH Verlag GmbH; 2005. pp. 645. Green DJ, Colombo P. Cellular ceramics: intriguing structures, novel properties, and innovative applications. MRS Bull 2003; 28(4):296e300. Adler J. Ceramic diesel particulate filters. Int J Appl Ceram Technol 2005;2(6):429e39. Shyam A, Lara-Curzio E, Watkins TR, Parten RJ. Mechanical characterization of diesel particulate filter substrates. J Am Ceram Soc 2008;91(6):1995e2001. Pyzik AJ, Li CG. New design of a ceramic filter for diesel emission control applications. Int J Appl Ceram Technol 2005;2(6):440e51. Wakita M. Application and development in the future of ceramic membrane. Bull Ceram Soc Jpn 2010;45:796e800 [in Japanese]. Taslicukur Z, Balaban C, Kuskonmaz N. Production of ceramic foam filters for molten metal filtration using expanded polystyrene. J Eur Ceram Soc 2007;27:637e40. Zhang Y, Yu J, Chen S, Wan S. Wastewater treatment using bioreactor with dual functional ceramic membrane. Int J Environ Pollut 2009;38(3):318e27. Le Guehennec L, Layrolle P, Daculsi G. A review of bioceramics and fibrin sealant. Eur Cell Mater 2004;8:1e11. Suzuki T, Zahir H, Funabashi Y, Yamaguchi T, Fujishiro Y, Awano M. Impact of anode microstructure on solid oxide fuel cells. Science 2009;325:852e5. Roncari E, Galassi C, Craciun F, Capiani C, Piancastelli A. A microstructural study of porous piezoelectric ceramics obtained by different methods. J Eur Ceram Soc 2001;21(3):409e17. Litovsky EY, Shapiro M. Gas pressure and temperature dependences of thermal conductivity of porous ceramic materials: part 1, refractories and ceramics with porosity below 30%. J Am Ceram Soc 1992;75(12):3425e39. Litovsky E, Shapiro M, Shavit A. Gas pressure and temperature dependences of thermal conductivity of porous ceramic materials: part 2, refractories and ceramics with porosity exceeding 30%. J Am Ceram Soc 1996;79(5):1366e76. Ramı´rez -Rico J, de Arellano -Lo´pez AR, Martı´nez -Ferna´ndez J, Larrea A, Orera VM. High-temperature mechanical properties of porous NaMgF3 derived from directionally solidified NaMgF3eNaF eutectics. J Eur Ceram Soc 2008;28:2451e7. Larrea A, Orera VM. Porous crystal structures obtained from directionally solidified eutectic precursors. J Crystal Growth 2007;300:387e93. Wang B. Fabrication of La0.8Sr0.2FeO3 d porous ceramics using a phase-leaching method. J Am Ceram Soc 2008;91(12):4118e20. Suzuki Y, Morgan PED. Meso- and macroporous ceramics by phase separation and reactive sintering methods. MRS Bull 2009;34(8):587e91. Ueno S, Lin LM, Nakajima H. Formation mechanism of porous alumina with oriented cylindrical pores fabricated by unidirectional solidification. J Am Ceram Soc 2008;91(1):223e6. (For example) Fukushima M, Zhou Y, Miyazaki H, Yoshizawa Y, Hirao K, Iwamoto Y, et al. Microstructural characterization of

1144

[32]

[33] [34]

[35]

[36]

[37]

[38]

[39]

[40]

[41]

[42] [43]

[44]

[45]

[46]

[47]

[48]

[49]

Handbook of Advanced Ceramics

porous silicon carbide membrane support with and without alumina additive. J Am Ceram Soc 2006;89(5):1523e9. Nanjangud SC, Brezny R, Green DJ. Strength and Young’s modulus behavior of a partially sintered porous alumina. J Am Ceram Soc 1995;78(1):266e8. 1995. Hardy D, Green DJ. Mechanical properties of a partially sintered alumina. J Eur Ceram Soc 1995;15:769e75. Oh ST, Tajima KI, Ando M, Ohji T. Strengthening of porous alumina by pulse electric current sintering and nanocomposite processing. J Am Ceram Soc 2000;83(5):1314e6. Jayaseelan DD, Kondo N, Brito ME, Ohji T. High-strength porous alumina ceramics by the pulse electric current sintering technique. J Am Ceram Soc 2002;85(1):267e9. Yang Y, Wang Y, Tian W, Wang Z, Li C-G, Zhao Y, et al. In situ porous alumina/aluminum titanate ceramic composite prepared by spark plasma sintering from nanostructured powders. Scripta Mater 2009;60:578e81. Akhtar F, Vasiliev PO, Bergstro¨m L. Hierarchically porous ceramics from diatomite powders by pulsed current processing. J Am Ceram Soc 2009;92(2):338e43. Deng Z-Y, Fukasawa T, Ando M, Zhang GJ, Ohji T. High-surfacearea alumina ceramics fabricated by the decomposition of Al(OH)3. J Am Ceram Soc 2001;84(3):485e91. Deng Z-Y, Fukasawa T, Ando M, Zhang GJ, Ohji T. Microstructure and mechanical properties of porous alumina ceramics fabricated by the decomposition of aluminum hydroxide. J Am Ceram Soc 2001;84(11):2638e44. Deng ZY, Zhou Y, Inagaki Y, Ohji T. Role of Zr(OH)4 hard agglomerates in fabricating porous ZrO2 ceramics and the reinforcing mechanisms. Acta Mater 2003;51:731e9. Claussen N, Wu S, Holz D. Reaction bonding of aluminum oxide (RBAO) composites: processing, reaction mechanisms and properties. J Eur Ceram Soc 1994;14:97e109. She JH, Ohji T. Fabrication and characterization of highly porous mullite ceramics. Mater Chem Phys 2003;80(3):610e4. Suzuki Y, Morgan PED, Ohji T. New uniformly porous CaZrO3/MgO composites with three-dimensional network structure from natural dolomite. J Am Ceram Soc 2000;83(8):2091e3. Suzuki Y, Kondo N, Ohji T, Morgan PED. Uniformly porous composites with 3-D network structure (UPC-3D) for high-temperature filter applications. Int J Appl Ceram Technol 2004;1(1):76e85. Suzuki Y, Kondo N, Ohji T. In situ synthesis and microstructure of porous CaAl4O7 monolith and CaAl4O7/CaZrO3 composite. J Ceram Soc Jpn 2001;109(3):205e9. Suzuki Y, Kondo N, Ohji T. Reactive synthesis of a porous calcium zirconate/spinel composite with idiomorphic spinel grains. J Am Ceram Soc 2003;86(7):1128e31. She JH, Yang JF, Kondo N, Ohji T, Kanzaki S, Deng ZY. Highstrength porous silicon carbide ceramics by an oxidation-bonding technique. J Am Ceram Soc 2002;85(11):2852e4. Ding S, Zeng Y-P, Jiang D. Oxidation bonding of porous silicon nitride ceramics with high strength and low dielectric constant. Mater Lett 2007;61:2277e80. Ding S, Zhu S, Zeng Y-P, Jiang D. Fabrication of mullite-bonded porous silicon carbide ceramics by in situ reaction bonding. J Eur Ceram Soc 2007;27:2095e102.

[50] Liu S, Zeng Y-P, Jiang D. Fabrication and characterization of cordierite-bonded porous SiC ceramics. Ceram Int 2009; 35:597e602. [51] Plucknett KP, Quinlan M, Garrido L, Genova L. Microstructural development in porous b-Si3N4 ceramics prepared with low volume RE2O3eMgOe(CaO) additions (RE ¼ La, Nd, Y, Yb). Mater Sci Eng A 2008;489:337e50. [52] Yang JF, Deng ZY, Ohji T. Fabrication and characterisation of porous silicon nitride ceramics using Yb2O3 as sintering additive. J Eur Ceram Soc 2003;23:371e8. [53] Yang JF, Ohji T, Kanzaki S, Diaz A, Hampshire S. Microstructure and mechanical properties of silicon nitride ceramics with controlled porosity. J Am Ceram Soc 2002;85(6):1512e6. [54] Yang J, Yang JF, Shan SY, Gao JQ, Ohji T. Effect of sintering additives on microstructure and mechanical properties of porous silicon nitride ceramics. J Am Ceram Soc 2006;89(12):3843e5. [55] Yang JF, Shan SY, Janssen R, Schneider G, Ohji T, Kanzaki S. Synthesis of Fibrous Beta-Si3N4 Structured Porous Ceramics Using Carbothermal Nitridation of Silica. Acta Mater 2005;53(10):2981e90. [56] Tuyen D-V, Park Y-J, Kim H-D, Lee B-T. Formation of rod-like Si3N4 grains in porous SRBSN bodies using 6Y2O3e2MgO sintering additives. Ceram Int 2009;35:2305e10. [57] Inagaki Y, Ohji T, Kanzaki S, Shigegaki Y. Fracture energy of an aligned porous silicon nitride. J Am Ceram Soc 2000; 83(7):1807e9. [58] Inagaki Y, Shigegaki Y, Ando M, Ohji T. Synthesis and evaluation of anisotropic porous silicon nitride. J Eur Ceram Soc 2004;24:197e200. [59] Colombo P, Bernardo E, Biasetto L. Novel microcellular ceramics from a silicone resin. J Am Ceram Soc 2004;87(1):152e4. [60] Colombo P, Bernardo E. Macro- and micro-cellular porous ceramics from preceramic polymers. Compos Sci Technol 2003;63(16):2353e9. [61] Biasetto L, Colombo P, Innocentini MDM, Mullens S. Gas permeability of microcellular ceramic foams. Ind Eng Chem Res 2007;46:3366e72. [62] Cruz HS, Spino J, Grathwohl G. Nanocrystalline ZrO2 ceramics with idealized macropores. J Eur Ceram Soc 2008;28:1783e91. [63] Descamps M, Duhoo T, Monchau F, Lu J, Hardouin P, Hornez JC, et al. Manufacture of macroporous b-tricalcium phosphate bioceramics. J Eur Ceram Soc 2008;28:149e57. [64] Descamps M, Richart O, Hardouin P, Hornez JC, Leriche A. Synthesis of macroporous b-tricalcium phosphate with controlled porous architectural. Ceram Int 2008;34:1131e7. [65] Yang JF, Zhang GJ, Kondo N, Ohji T, Kanzaki S. Synthesis of porous Si3N4 ceramics with rod-shaped pore structure. J Am Ceram Soc 2005;88(4):1030e2. [66] Chen F, Shen Q, Yan F, Zhang L. Pressureless sintering of alphaSi3N4 porous ceramics using a H3PO4 pore-forming agent. J Am Ceram Soc 2007;90(8):2379e83. [67] Luyten J, Mullens S, Cooymans J, De Wilde AM, Thijs I. New processing techniques of ceramic foams. Adv Eng Mater 2003; 5(10):715e8. [68] Thijs I, Luyten J, Mullens S. Producing ceramic foams with hollow spheres. J Am Ceram Soc 2004;87(1):170e2. [69] Kumar BP, Kumar HH, Kharat DK. Study on pore-forming agents in processing of porous piezoceramics. J Mater Sci Mater Electron 2005;16(10):681e6.

Chapter | 11.2.2

Porous Ceramic Materials

[70] Hotta Y, Alberius PCA, Bergstrom L. Coated polystyrene particles as templates for ordered macroporous silica structures with controlled wall thickness. J Mater Chem 2003;13(3):496e501. [71] Wang DY, Caruso RA, Caruso F. Synthesis of macroporous titania and inorganic composite materials from coated colloidal spheres e a novel route to tune pore morphology. Chem Mater 2001;13(2):364e71. [72] Davis SA, Breulmann M, Rhodes KH, Zhang B, Mann S. Template-directed assembly using nanoparticle building blocks: a nanotectonic approach to organized materials. Chem Mater 2001;13(10):3218e26. [73] Kim YW, Kim SH, Park CB. Processing of closed-cell silicon oxycarbide foams from a preceramic polymer. J Mater Sci 2004;39(18):5647e52. [74] Kim YW, Jin YJ, Chun YS, Song IH, Kim HD. A simple pressing route to closed-cell microcellular ceramics. Scripta Mater 2005;53(8):921e5. [75] Li SH, de Wijn JR, Layrolle P, de Groot K. Novel method to manufacture porous hydroxyapatite by dual-phase mixing. J Am Ceram Soc 2003;86(1):65e72. [76] Wang H, Sung IY, Li XD, Kim D. Fabrication of porous SiC ceramics with special morphologies by sacrificing template method. J Porous Mater 2004;11(4):265e71. [77] Chae S-H, Kim Y-W, Song I-H, Kim H-D, Narisawa M. Porosity control of porous silicon carbide ceramics. J Eur Ceram Soc 2009;29:2867e72. [78] Yin X, Li X, Zhang L, Cheng L, Liu Y, Pan T. Microstructure and mechanical properties of Lu2O3-doped porous silicon nitride ceramics using phenolic resin as pore-forming agent. Int J Appl Ceram Technol 2010;7:391e9. [79] Kamitani K, Hyodo T, Shimizu Y, Egashira M. Fabrication of porous alumina ceramics having cell windows with controlled size by PMMA template method. J Mater Sci 2010;45:3602e9. [80] Li Y, Chen F, Li L, Zhang W, Yu H, Shan Y, et al. Gas pressure sintering of arbitrary porous silicon nitride ceramics with high mechanical strength. J Am Ceram Soc 2010;93:1565e8. [81] Isobe T, Kameshima Y, Nakajima A, Okada K, Hotta Y. Gas permeability and mechanical properties of porous alumina ceramics with unidirectionally aligned pores. J Eur Ceram Soc 2007;27:53e9. [82] Andersson L, Bergstro¨m L. Gas-filled microspheres as an expandable sacrificial template for direct casting of complexshaped macroporous ceramics. J Eur Ceram Soc 2008;28:2815e21. [83] Kim Y-W, Eom J-H, Wang C, Park CB. Processing of porous silicon carbide ceramics from carbon-filled polysiloxane by extrusion and carbothermal reduction. J Am Ceram Soc 2008;91(4):1361e4. [84] Eom J-H, Kim Y-W, Song I-H, Kim H-D. Processing and properties of polysiloxane-derived porous silicon carbide ceramics using hollow microspheres as templates. J Eur Ceram Soc 2008;28:1029e35. [85] Song I-H, Kwon I-M, Kim H-D, Kim Y-W. Processing of microcellular silicon carbide ceramics with a duplex pore structure. J Eur Ceram Soc 2010;30:2671e6. [86] Okada K, Uchiyama S, Isobe T, Kameshima Y, Nakajima A, Kurata T. Capillary rise properties of porous mullite ceramics prepared by an extrusion method using organic fibers as the pore former. J Eur Ceram Soc 2009;29:2491e7.

1145

[87]

[88]

[89]

[90]

[91]

[92]

[93]

[94]

[95] [96]

[97]

[98]

[99]

[100] [101]

[102] [103] [104]

[105]

Okada K, Shimizu M, Isobe T, Kameshima Y, Sakai M, Nakajima A, et al. Characteristics of microbubbles generated by porous mullite ceramics prepared by an extrusion method using organic fibers as the pore for mer. J Eur Ceram Soc 2010;30:1245e51. Gain AK, Song HY, Lee BT. Microstructure and mechanical properties of porous yttria stabilized zirconia ceramic using poly methyl methacrylate powder. Scripta Mater 2006;54(12):2081e5. Eom J-H, Kim Y-W. Effect of template size on microstructure and strength of porous silicon carbide ceramics. J Ceram Soc Japan 2008;116(1358):1159e63. Jiang G-P, Yang J-F, Gao J-Q, Niihara K. Porous silicon nitride ceramics prepared by extrusion using starch as binder. J Am Ceram Soc 2008;91(11):3510e6. Kitamura M, Ohtsuki C, Ogata S, Kamitakahara M, Tanihara M. Microstructure and bioresorbable properties of alpha-TCP ceramic porous body fabricated by direct casting method. Mater Trans 2004;45(4):983e8. Kamitakahara M, Ohtsuki C, Kawachi G, Wang D, Ioku K. Preparation of hydroxyapatite porous ceramics with different porous structures using a hydrothermal treatment with different aqueous solutions. J Ceram Soc Japan 2008;116(1349):6e9. Diaz A, Hampshire S. Characterisation of porous silicon nitride materials produced with starch. J Eur Ceram Soc 2004; 24(2):413e9. Diaz A, Hampshire S, Yang JF, Ohji T, Kanzaki S. Comparison of mechanical properties of silicon nitrides with controlled porosities produced by different fabrication routes. J Am Ceram Soc 2005;88(3):698e706. Kim J-G, Sim J-H, Cho W-S. Preparation of porous (Ba, Sr)TiO3 by adding corn-starch. J Phys Chem Solids 2002;11:2079e84. Zhang G-J, Yang J-F, Ohji T. Fabrication of porous ceramics with unidirectionally aligned continuous pores. J Am Ceram Soc 2001;84(6):1395e7. Gaydardzhiev S, Gusovius H, Wilker V, Ay P. Gel-casted porous Al2O3 ceramics by use of natural fibres as pore developers. J Porous Mater 2008;15(4):475e80. Naga SM, El-Maghraby AA, El-Rafei AM, Greil P, Khalifa T, Ibrahim NA. Porous fibrous mullite bodies. Am Ceram Soc Bull 2006;85(2):21e4. Wang C, Kasuga T, Nogami M. Macroporous calcium phosphate glass-ceramic prepared by two-step pressing technique and using sucrose as a pore former. J Mater Sci Mater Med 2005;16(8):739e44. Sun Y, Tan SH, Jiang DL. Synthesis of porous silicon carbide and its catalysis. J Inorganic Mater 2003;18(4):830e6. da Silva MHP, Lemos AF, Gibson IR, Ferreira JMF, Santos JD. Porous glass reinforced hydroxyapatite materials produced with different organic additives. J Non Cryst Solids 2002;304(1e3):286e92. Galassi C, Capiani C, Craciun F, Roncari E. Water-based technique to produce porous PZT materials. J Phys IV 2005;128:27e31. Barea R, Osendi MI, Miranzo P, Ferreira JMF. Fabrication of highly porous mullite materials. J Am Ceram Soc 2005;88(3):777e9. Vasconcelos PV, Labrincha JA, Ferreira JMF. Permeability of diatomite layers processed by different colloidal techniques. J Eur Ceram Soc 2000;20(2):201e7. Zivcova´ Z, Cerny M, Pabst W, Gregorova´ E. Elastic properties of porous oxide ceramics prepared using starch as a pore-forming agent. J Eur Ceram Soc 2009;29:2765e71.

1146

[106] Shao Y, Jia D, Zhou Y, Liu B. Novel method for fabrication of silicon nitride/silicon oxynitride composite ceramic foams using fly ash cenosphere as a pore-forming agent. J Am Ceram Soc 2008;91(11):3781e5. [107] Okada K, Ikawa F, Isobe T, Kameshima Y, Nakajima A. Low temperature preparation and machinability of porous ceramics from talc and foamed glass particles. J Eur Ceram Soc 2009;29:1047e52. [108] Paul RK, Gain AK, Lee BT, Jang HD. Effect of addition of silicon on the microstructures and bending strength of continuous porous SiC-Si3N4 composites. J Am Ceram Soc 2006;89(6):2057e62. [109] Miyagawa N, Shinohara N. Fabrication of porous alumina ceramics with uni-directionally-arranged continuous pores using a magnetic field. J Ceram Soc Japan 1999;107(7):673e7. [110] Isobe T, Tomita T, Kameshima Y, Nakajima A, Okada K. Preparation and properties of porous alumina ceramics with oriented cylindrical pores produced by an extrusion method. J Eur Ceram Soc 2006;26:957e60. [111] Kim H, da Rosa C, Boaro M, Vohs JM, Gorte RJ. Fabrication of highly porous yttria-stabilized zirconia by acid leaching nickel from a nickel-yttria-stabilized zirconia cermet. J Am Ceram Soc 2002;85(6):1473e6. [112] Rajamathi M, Thimmaiah S, Morgan PED, Seshadri R. Macroporous materials from crystalline single-source precursors through decomposition followed by selective leaching. J Mater Chem 2001;11(10):2489e92. [113] She JH, Deng ZY, Jayaseelan DD, Ohji T. Oxidation bonding of porous silicon carbide ceramics. J Mater Sci 2002;37(17):3615e22. [114] Shao Y, Jia D, Liu B. Characterization of porous silicon nitride ceramics by pressureless sintering using fly ash cenosphere as a pore-forming agent. J Eur Ceram Soc 2009;29:1529e34. [115] Jia D, Kim D-K, Kriven WM. Sintering behavior of gehlenite. Part I: self-forming, macro-/mesoporous gehlenite e pore-forming mechanism, microstructure, mechanical, and physical properties. J Am Ceram Soc 2007;90(6):1760e73. [116] Neirinck B, Fransaer J, der Biest OV, Vleugels J. A novel route to produce porous ceramics. J Eur Ceram Soc 2009;29:833e6. [117] Imhof A, Pine DJ. Ordered macroporous materials by emulsion templating. Nature 1997;389(6654):948e51. [118] Fukasawa T, Ando M, Ohji T, Kanzaki S. Synthesis of porous ceramics with complex pore structure by freeze-dry processing. J Am Ceram Soc 2001;84(1):230e2. [119] Fukasawa T, Deng ZY, Ando M, Ohji T, Goto Y. Pore structure of porous ceramics synthesized from water-based slurry by freezedry process. J Mater Sci 2001;36(10):2523e7. [120] Fukasawa T, Deng ZY, Ando M, Ohji T, Kanzaki S. Synthesis of porous silicon nitride with unidirectionally aligned channels using freeze-drying process. J Am Ceram Soc 2002;85(9):2151e5. [121] Koch D, Andresen L, Schmedders T, Grathwohl G. Evolution of porosity by freeze casting and sintering of solegel derived ceramics. J SoleGel Sci Technol 2003;26(1e3):149e52. [122] Mukai SR, Nishihara H, Tamon H. Formation of monolithic silica gel microhoneycombs (SMHs) using pseudosteady state growth of microstructural ice crystals. Chem Commun 2004;7:874e5. [123] Pek YS, Gao S, Arshad MSM, Leck K-J, Ying JY. Porous collagen-apatite nanocomposite foams as bone regeneration scaffolds. Biomaterials 2008;29:4300e5.

Handbook of Advanced Ceramics

[124] Schoof H, Apel J, Heschel I, Rau G. Control of pore structure and size in freeze-dried collagen sponges. J Biomed Mater Res 2001;58:352e7. [125] Ren L, Zeng Y-P, Jiang D. Preparation of porous TiO2 by a novel freeze casting. Ceram Int 2009;35:1267e70. [126] Landi E, Valentini F, Tampieri A. Porous hydroxyapatite/gelatine scaffolds with ice-designed channel-like porosity for biomedical applications. Acta Biomater 2008;4(6):1620e6. [127] Deville S, Saiz E, Nalla RK, Tomsia AP. Freezing as a path to build complex composites. Science 2006;311:515e8. [128] Deville S, Saiz E, Tomsia AP. Ice-templated porous alumina structures. Acta Mater 2007;55:1965e74. [129] Deville S. Freeze-casting of porous ceramics: a review of current achievements and issues. Adv Eng Mater 2008;10(3):155e69. [130] Deville S, Maire E, Bernard-Granger G, Lasalle A, Bogner A, Gauthier C, et al. Metastable and unstable cellular solidification of colloidal suspensions. Nature Mater 2009;8:966e72. [131] Munch E, Launey ME, Alsem DH, Saiz E, Tomsia AP, Ritchie RO. Tough, bio-inspired hybrid materials. Science 2008;322:1516e20. [132] Araki K, Halloran JW. New freeze-casting technique for ceramics with sublimable vehicles. J Am Ceram Soc 2004;87(10):1859e63. [133] Araki K, Halloran JW. Room-temperature freeze casting for ceramics with nonaqueous sublimable vehicles in the naphthalenecamphor eutectic system. J Am Ceram Soc 2004;87(11):2014e9. [134] Araki K, Halloran JW. Porous ceramic bodies with interconnected pore channels by a novel freeze casting technique. J Am Ceram Soc 2005;88(5):1108e14. [135] Koh Y-H, Song J-H, Lee E-J, Kim H-E. Freezing dilute ceramic/camphene slurry for ultra-high porosity ceramics with completely interconnected pore networks. J Am Ceram Soc 2006; 89(10):3089e93. [136] Koh Y-H, Lee E-J, Yoon B-H, Song J-H, Kim H-E, Kim H-W. Effect of polystyrene addition on freeze casting of ceramic/ camphene slurry for ultra-high porosity ceramics with aligned pore channels. J Am Ceram Soc 2006;89(12):3646e53. [137] Yoon B-H, Choi W-Y, Kim H-E, Kim J-H, Koh Y-H. Aligned porous alumina ceramics with high compressive strengths for bone tissue engineering. Scripta Mater 2008;58(7):537e40. [138] Yoon B-H, Koh Y-H, Park C-S, Kim H-E. Generation of large pore channels for bone tissue engineering using camphene-based freeze casting. J Am Ceram Soc 2007;90(6):1744e52. [139] Yoon B-H, Park C-S, Kim H-E, Koh Y-H. In situ synthesis of porous silicon carbide (SiC) ceramics decorated with SiC nanowires. J Am Ceram Soc 2007;90(12):3759e66. [140] Lee S-H, Jun S-H, Kim H-E, Koh Y-H. Fabrication of porous PZT-PZN piezoelectric ceramics with high hydrostatic figure of merits using camphene-based freeze casting. J Am Ceram Soc 2007;90(9):2807e13. [141] Lee S-H, Jun S-H, Kim H-E, Koh Y-H. Piezoelectric properties of PZT-based ceramic with highly aligned pores. J Am Ceram Soc 2008;91(6):1912e5. [142] Yoon B-H, Lee E-J, Kim H-E, Koh Y-H. Highly aligned porous silicon carbide ceramics by freezing polycarbosilane/camphene solution. J Am Ceram Soc 2007;90(6):1753e9. [143] Yoon B-H, Park C-S, Kim H-E, Koh Y-H. In-situ fabrication of porous hydroxyapatite (HA) scaffolds with dense shells by freezing HA/camphene slurry. Mater Lett 2008;62:1700e3.

Chapter | 11.2.2

Porous Ceramic Materials

[144] Song J-H, Koh Y-H, Kim H-E, Li L-H, Bahn H-J. Fabrication of a porous bioactive glass-ceramic using room-temperature freeze casting. J Am Ceram Soc 2006;89(8):2649e53. [145] Hong C, Zhang X, Han J, Du J, Han W. Ultra-high-porosity zirconia ceramics fabricated by novel room-temperature freezecasting. Scripta Mater 2009;60:563e6. [146] Han J, Hong C, Zhang X, Du J, Zhang W. Highly porous ZrO2 ceramics fabricated by a camphene-based freeze-casting route: microstructure and properties. J Eur Ceram Soc 2010;30:53e60. [147] Chen R-F, Huang Y, Wang C-A, Qi J. Ceramics with ultra-low density fabricated by gelcasting: an unconventional view. J Am Ceram Soc 2007;90(11):3424e9. [148] Chen R-F, Wang C-A, Huang Y, Ma L, Lin W. Ceramics with special porous structures fabricated by freeze-gelcasting: using tert-butyl alcohol as a template. J Am Ceram Soc 2007;90(11):3478e84. [149] Ding S, Zeng Y-P, Jiang D. Fabrication of mullite ceramics with ultrahigh porosity by gel freeze drying. J Am Ceram Soc 2007;90(7):2276e9. [150] Zuo KH, Zeng Y-P, Jiang D-L. Properties of microstructurecontrollable porous yttria-stabilized ziroconia ceramics fabricated by freeze casting. Int J Appl Ceram Technol 2008;5(2):198e203. [151] Pekor CM, Groth B, Nettleship I. The effect of polyvinyl alcohol on the microstructure and permeability of freeze-cast alumina. J Am Ceram Soc 2010;93(1):115e20. [152] Pekor CM, Kisa P, Nettleship I. Effect of polyethylene glycol on the microstructure of freeze-cast alumina. J Am Ceram Soc 2008;91(10):3185e90. [153] Nishihara H, Mukai SR, Yamashita D, Tamon H. Ordered macroporous silica by ice templating. Chem Mater 2005;17(3): 683e9. [154] Fukushima M, Nakata M, Yoshizawa Y. Fabrication and properties of ultra highly porous cordierite with oriented micrometersized cylindrical pores by gelation and freezing method. J Ceram Soc Japan 2008;116(1360):1322e5. [155] Fukushima M, Nakata M, Zhou Y, Ohji T, Yoshizawa Y. Fabrication and properties of ultra highly porous silicon carbide by the gelationefreezing method. J Eur Ceram Soc 2010;30:2889e96. [156] Ohji T, Suzuki Y, Yang J-F, Hayashi I. Development of high performance filter materials. Materials Integration 2004;17(4): 5e13. [157] Faure R, Rossignol F, Chartier T, Bonhomme C, Maitrea A, Etchegoyen G, et al. Alumina foam catalyst supports for industrial steam reforming processes. J Eur Ceram Soc 2011;31:303e12. [158] Zhu X, Jiang D, Tan S, Zhang Z. Improvement in the strut thickness of reticulated porous ceramics. J Am Ceram Soc 2001;84(7):1654e6. [159] Vogt UF, Gorbar M, Dimopoulos P, Broenstrup A, Wagner G, Colombo P. Improving the properties of ceramic foams by a vacuum infiltration process. J Eur Ceram Soc 2010;30:3005e11. [160] Luyten J, Thijs I, Vandermeulen W, Mullens S, Wallaeys B, Mortelmans R. Strong ceramic foams from polyurethane templates. Adv Appl Ceram 2005;104(1):4e8. [161] Jun I-K, Koh Y-H, Kim H-E. Fabrication of a highly porous bioactive glass-ceramic scaffold with a high surface area and strength. J Am Ceram Soc 2006;89(1):391e4. [162] Jun I-K, Song J-H, Choi W-Y, Koh Y-H, Kim H-E, Kim H-W. Porous hydroxyapatite scaffolds coated with bioactive apatitewollastonite glass-ceramics. J Am Ceram Soc 2007;90(9):2703e8.

1147

[163] Plesch G, Gorbar M, Vogt UF, Jesenak K, Vargova M. Reticulated macroporous ceramic foam supported TiO2 for photocatalytic applications. Mater Lett 2009;63:461e3. [164] Nangrejo MR, Edirisinghe MJ. Porosity and strength of silicon carbide foams prepared using preceramic polymers. J Porous Mater 2002;9(2):131e40. [165] Bao X, Nangrejo MR, Edirisinghe MJ. Preparation of silicon carbide foams using polymeric precursor solutions. J Mater Sci 2000;35(17):4365e72. [166] Nangrejo MR, Bao X, Edirisinghe MJ. The structure of ceramic foams produced using polymeric precursors. J Mater Sci Lett 2000;19(9):787e9. [167] Nangrejo MR, Bao XJ, Edirisinghe MJ. Preparation of silicon carbideesilicon nitride composite foams from pre-ceramic polymers. J Eur Ceram Soc 2000;20(11):1777e85. [168] Travitzky N, Windsheimer H, Fey T, Greil P. Preceramic paperderived ceramics. J Am Ceram Soc 2008;91(11):3477e92. [169] Singh M, Martinez-Fernandez J, de A-Lopez AR. Environmentally conscious ceramics (ecoceramics) from natural wood precursors. Curr Opin Solid State Mater Sci 2003;7:247e54. [170] Singh M, Salem JA. Mechanical properties and microstructure of biomorphic silicon carbide ceramics fabricated from wood precursors. J Eur Ceram Soc 2002;22:2709e17. [171] Singh M, Yee B-M. Reactive processing of environmentally conscious, biomorphic ceramics from natural wood precursors. J Eur Ceram Soc 2004;24:209e17. [172] Sieber H, Hoffmann C, Kaindl A, Greil P. Biomorphic cellular ceramics. Adv Eng Mater 2000;2(3):105e9. [173] Qian JM, Wang JP, Jin ZH. Preparation of biomorphic SiC-ceramics by the reactive infiltration on of Si into carbon template derived from basswood. Rare Metal Mater Eng 2004;33(10):1065e8. [174] Varela-Feria FM, Martinez-Fernandez J, de Arellano-Lopez AR, Singh M. Low density biomorphic silicon carbide: microstructure and mechanical properties. J Eur Ceram Soc 2002; 22(14e15):2719e25. [175] Greil P, Lifka T, Kaindl A. Biomorphic cellular silicon carbide ceramics from wood: I. Processing and microstructure. J Eur Ceram Soc 1998;18(14):1961e73. [176] Zampieri A, Sieber H, Selvam T, Mabande GTP, Schwieger W, Scheffler F, et al. Biomorphic cellular SiSiC/zeolite ceramic composites: from Rattan palm to bioinspired structured monoliths for catalysis and sorption. Adv Mater 2005;17(3):344. [177] Calderon NR, Escandell MM, Narciso J, Reinoso FR. The role of carbon biotemplate density in mechanical properties of biomorphic SiC. J Eur Ceram Soc 2009;29:465e72. [178] Kaul VS, Faber KT. Nanoindentation analysis of the elastic properties of porous SiC derived from wood. Scripta Mater 2008;58:886e9. [179] Pappacena KE, Faber KT, Wang H, Porter WD. Thermal conductivity of porous silicon carbide derived from wood precursors. J Am Ceram Soc 2007;90(9):2855e62. [180] Vogli E, Sieber H, Greil P. Biomorphic SiC-ceramic prepared by Sivapor phase infiltration of wood. J Eur Ceram Soc 2002;22:2663e8. [181] Rambo CR, Sieber H. Novel synthetic route to biomorphic Al2O3 ceramics. Adv Mater 2005;17(8):1088. [182] Streitwieser DA, Popovska N, Gerhard H, Emig G. Application of the chemical vapor infiltration and reaction (M-R) technique for the preparation of highly porous biomorphic SiC ceramics derived from paper. J Eur Ceram Soc 2005;25:817e28.

1148

[183] Popovska N, Streitwieser DA, Xu C, Gerhard H. Paper derived biomorphic porous titanium carbide and titanium oxide ceramics produced by chemical vapor infiltration and reaction (CVI-R). J Eur Ceram Soc 2005;25:829e36. [184] Greil P. Biomorphous ceramics from lignocellulosics. J Eur Ceram Soc 2001;21:105e18. [185] Greil P, Vogli E, Feya T, Bezold A, Popovska N, Gerhard H, et al. Effect of microstructure on the fracture behavior of biomorphous silicon carbide ceramics. J Eur Ceram Soc 2002;22:2697e707. [186] Ota T, Imaeda M, Takase H, Kobayashi M, Kinoshita N, Hirashita T, et al. Porous titania ceramic prepared by mimicking silicified wood. J Am Ceram Soc 2000;83(6):1521e3. [187] Mizutani M, Takase H, Adachi N, Ota T, Daimon K, Hikichi Y. Porous ceramics prepared by mimicking silicified wood. Sci Technol Adv Mater 2005;6(1):76e83. [188] Ota T, Takahashi M, Hibi T, Ozawa M, Suzuki S, Hikichi Y, et al. Biomimetic process for producing SiC ’wood’. J Am Ceram Soc 1995;78(12):3409e11. [189] Qian JM, Jin ZH. eparation and characterization of porous, biomorphic SiC ceramic with hybrid pore structure. J Eur Ceram Soc 2006;26:1311e6. [190] Locs J, Berzina-Cimdina L, Zhurinsh A, Loca D. Effect of processing on the microstructure and crystalline phase composition of wood derived porous SiC ceramics. J Eur Ceram Soc 2011;31:183e8. [191] Torres-Raya C, Hernandez-Maldonado D, Ramirez-Rico J, Garcia-Ganan C, de Arellano-Lopez AR, Martinez-Fernandez J. Fabrication, chemical etching, and compressive strength of porous biomimetic SiC for medical implants. J Mater Res 2008;23(12):3247e54. [192] Rambo CR, Sieber H, Genova LA. Synthesis of porous biomorphic alpha/beta-Si3N4 composite from sea sponge. J Porous Mater 2008;15(4):419e25. [193] Song H-Y, Islam S, Lee B-T. A novel method to fabricate unidirectional porous hydroxyapatite body using ethanol bubbles in a viscous slurry. J Am Ceram Soc 2008;91(9):3125e7. [194] Banno T, Yamada Y, Nagae H. Fabrication of porous alumina ceramics by simultaneous thermal gas generating and thermal slurry solidification. J Ceram Soc Japan 2009;117:713e6. [195] Tomita T, Kawasaki S, Okada K. Effect of viscosity on preparation of foamed silica ceramics by a rapid gelation foaming method. J Porous Mater 2005;12:123e9. [196] Fuji M, Kato T, Zhang F, Takahashi M. Effects of surfactants on the microstructure and some intrinsic properties of porous building ceramics fabricated by gelcasting. Ceram Int 2006;32:797e802. [197] Tan SN, Fornasiero D, Sedev R, Ralston J. The role of surfactant structure on foam behaviour. Colloids Surf A 2005;263:233e8. [198] Barg S, Soltmann C, Andrade M, Koch D, Grathwohl G. Cellular ceramics by direct foaming of emulsified ceramic powder suspensions. J Am Ceram Soc 2008;91(9):2823e9. [199] Barg S, Koch D, Grathwohl G. Processing and properties of graded ceramic filters. J Am Ceram Soc 2009;92(12):2854e60. [200] Barg S, Moraes E, Koch D, Grathwohl G. New cellular ceramics from high alkane phase emulsified suspensions (HAPES). J Eur Ceram Soc 2009;29:2439e46. [201] Kim H, Lee S, Han Y, Park J-K. Control of pore size in ceramic foams: influence of surfactant concentration. Mater Chem Phys 2009;113:441e4.

Handbook of Advanced Ceramics

[202] Colombo P, Modesti M. Silicon oxycarbide ceramic foams from a preceramic polymer. J Am Ceram Soc 1999;82:573e8. [203] Colombo P, Hellmann JR, Shelleman DL. Mechanical properties of silicon oxycarbide ceramic foams. J Am Ceram Soc 2001; 84:2245e51. [204] Kim Y-W, Kim S-H, Wang C, Park CB. Fabrication of microcellular ceramics using gaseous carbon dioxide. J Am Ceram Soc 2003;86:2231e3. [205] Kim Y-W, Wang C, Park CB. Processing of porous silicon oxycarbide ceramics from extruded blends of polysiloxane and polymer microbead. J Ceram Soc Japan 2007;115:419e42. [206] Du ZP, Bilbao-Montoya MP, Binks BP, Dickinson E, Ettelaie R, Murray BS. Outstanding stability of particle-stabilized bubbles. Langmuir 2003;19(8):3106e8. [207] Dickinson E, Ettelaie R, Kostakis T, Murray BS. Factors controlling the formation and stability of air bubbles stabilized by partially hydrophobic silica nanoparticles. Langmuir 2004;20(20). 8517e8125. [208] Binks BP, Horozov TS. Aqueous foams stabilized solely by silica nanoparticles. Angew Chem Int Ed 2005;44(24):3722e5. [209] Studart AR, Gonzenbach UT, Akartuna I, Tervoort E, Gauckler LJ. Materials from foams and emulsions stabilized by colloidal particles. J Mater Chem 2007;17:3283e9. [210] Gonzenbach UT, Studart AR, Tervoort E, Gauckler LJ. Stabilization of foams with inorganic colloidal particles. Langmuir 2006;22(26):10983e8. [211] Gonzenbach UT, Studart AR, Tervoort E, Gauckler LJ. Tailoring the microstructure of particle-stabilized wet foams. Langmuir 2007;23(3):1025e32. [212] Gonzenbach UT, Studart AR, Tervoort E, Gauckler LJ. Macroporous ceramics from particle-stabilized wet foams. J Am Ceram Soc 2007;90(1):16e22. [213] Gonzenbach UT, Studart AR, Steinlin D, Tervoort E, Gauckler LJ. Processing of particle-stabilized wet foams into porous ceramics. J Am Ceram Soc 2007;90(11):3407e14. [214] Akartuna I, Studart AR, Tervoort E, Gonzenbach UT, Gauckler LJ. Stabilization of oil-in-water emulsions by colloidal particles modified with short amphiphiles. Langmuir 2008;24(14):7161e8. [215] Latella BA, Henkel L, Mehrtens EG. Permeability and high temperature strength of porous mullite-alumina ceramics for hot gas filtration. J Mater Sci 2006;41:423e30. [216] Ding S, Zeng Y, Jiang D. Gas permeability behavior of mullite-bonded porous silicon carbide ceramics. J Mater Sci 2007;42: 7171e5. [217] Innocentini MDM, Sepulveda P, Salvini VR, Pandolfelli VC, Coury JR. Permeability and structure of cellular ceramics: a comparison between two preparation techniques. J Am Ceram Soc 1998;81:3349e52. [218] Moreira EA, Innocentini MDM, Coury JR. Permeability of ceramic foams to compressible and incompressible flow. J Eur Ceram Soc 2004;24:3209e18. [219] Ishizaki K, Komarneni S, Nanko K. Porous materials; process technology and applications. Kluwer Academic Publisher; 1998. [220] Dullien FAL. Porous Media Fluid Transport and Pore Structure. Academic Press; 1991. [221] Tomita T, Kawasaki S, Okada K. A novel preparation method for foamed silica ceramics by sol-gel reaction and mechanical foaming. J Porous Mater 2004;11:107e15.

Chapter 11.2.3

Spark Plasma Sintering (SPS) Method, Systems, and Applications Masao Tokita NJS Co., Ltd. 301 Office Shinyokohama, 2-14-8, Shinyokohama, Kohoku, Yokohama, Kanagawa 222-0033, Japan

Chapter Outline 1. 2. 3. 4.

Introduction Historical Background Suitable Materials For SPS Process Principles of the SPS Process

1149 1149 1152 1152

1. INTRODUCTION Spark plasma sintering (SPS), also called the pressureassisted pulse energizing process or the pulsed electric current sintering (PECS) process, is a promising technology for innovative processing in the field of new materials fabrication in the 21st century. SPS is a synthesis and processing technique which makes possible sintering and sinter bonding at low temperatures and in short periods by charging the intervals between powder particles with electrical energy and effectively applying a high-temperature spark plasma generated at an initial stage of energizing momentarily, and an electro-magnetic field and/or joule heating by continuous ONeOFF DC pulsed high electric current with a low voltage. As shown in Figure 1, the method is a solid compressive and a large pulse electric current energizing sintering technique that has recently drawn considerable attention as one of the newest rapid sintering methods with accurate energy density control. It is a novel sintering process featuring energy saving and high-speed consolidation and has a low power consumption of between 1/5 and 1/3 compared with conventional sintering techniques such as pressureless sintering (PLS), hot press (HP) sintering, and hot isostatic pressing (HIP) [1e6]. The system outlook is similar to a conventional hot press apparatus without outer heating element. However, the SPS has demonstrated the different superior sintering results for instance a structurally tailoring effect, minimizing grain growth, enhancement of electro-migration and strong

Handbook of Advanced Ceramics. http://dx.doi.org/10.1016/B978-0-12-385469-8.00060-5 Copyright Ó 2013 Elsevier Inc. All rights reserved.

5. Examples of SPS Process Applications 6. Summary and Outlook References

1160 1174 1175

preferential orientation effect in SPS processing [4]. This paper introduces the recent SPS technology, method, development of SPS systems, and its applications.

2. HISTORICAL BACKGROUND Since two decades ago, spark plasma sintering (SPS) method is of great interest to the powder and powder metallurgy industry and to material researchers of academia for both product manufacturing and advanced material research and development. It is generally well known that the SPS is an advanced processing technology to produce a homogeneous highly dense nanostructural sintered compacts, functionally graded materials (FGMs), fine ceramics, composite materials, new wear-resistant materials, thermo-electric semiconductors, and biomaterials. Today in Japan, a number of SPSed products for different industries have already been realized. The SPS is now moving from the scientific academia and/or R&D prototype materials level to practical industry-use product stage in the field of mold and die industry, cutting tools industry, electronics industry, and automotive industry. A technique similar to SPS was first studied in Germany around 1910 that was an electric energizing applied technique to consolidate a powder material. In the USA, G.F. Tayler patented the first resistance sintering method for sheet metals in 1933 [7]. Thereafter, G.D. Cremer obtained US patent for the method of sintering for copper, brass, or aluminum powder materials [8]. They were

1149

1150

Handbook of Advanced Ceramics

FIGURE 1 Typical classification of sintering methods.

considered as the origin of a current hot pressing (HP) technique that commonly applies a high-frequency induction heating method. The SPS was originally invented in Japan as “spark sintering (SS)” in 1962 by Dr. Kiyoshi Inoue of Japax Inc. [9,10]. In 1989, the present SPS was introduced by Sumitomo Coal Mining Co., Ltd (Japan) [11], and was developed as the thirdgeneration sintering technique to advance the first generation of spark sintering and the second generation of plasma-activated sintering (PAS) from Inoue-Japax Research Inc. As shown in Figure 2, the historical progress

of SPS technology is indicated by the relationship between size effect and shape effect containing functionality, reproducibility, and productivity. Research and development for the implementation of advanced SPS methods and systems were initiated to design practical hardware and software for industrial applications. Following the development of a box-type experimental-use SPS systems for new materials preparation and a single-head open-type introductory limited-production sintering systems, from 2001 to 2009 as the fourth-generation technology to accommodate for the product manufacturing field, five

FIGURE 2 Historical progress on spark plasma sintering (SPS) method and technology. For color version of this figure, the reader is referred to the online version of this book.

Chapter | 11.2.3

Spark Plasma Sintering (SPS) Method, Systems, and Applications

basic styles of SPS production systems from the medium to mass-production scale were developed [12,13]. As a result, the fabrication of various advanced new materials and industrial products has been presented. The 4thgeneration system performed to replace existing traditional fabrication processes to powder used processes, and to divide the SPS technology into four kinds of SPS processing and methods, namely the confirmed availability, were not only a sintering, but also a solid-phase diffusion bonding and joining [14], a surface modification (treatment) [15e17], and a synthesis technique for an

1151

example of a single-crystal fabrication as the SPS field. After 2010, it is to be called the start of a more practical manufacturing era, the progress of SPS technology is now getting into the 5th generation of “advanced SPS” with customized SPS apparatus. Figure 3 shows an example of an SPS job-shop center facility in Japan, a large ceramic sample, and total number of SPS machine systems produced and installed from 1990 to as of early 2010. Due to the increase of SPS manufacturers internationally in recent years, more than 550 units of SPS machines have already been working in the world approximately.

FIGURE 3 Increasing use of the SPS machines in the world (SPS units produced and installed, 1990e2010).

1152

Handbook of Advanced Ceramics

TABLE 1 Suitable Materials for SPS Process Classification

Materials for SPS processing

Metals

Fe, Cu, Al, Au, Ag, Ni, Cr, Mo, Sn, Ti, W, Be virtually any metal possible

Ceramics Oxides

Al2O3, mulite, ZrO2, MgO, SiO2, TiO2, HfO2

Carbides SiC, B4C, TaC, TiC, WC, ZrC, VC Nitrides

Si3N4, TaN, TiN, AlN, ZrN, VN

Borides

TiB2, HfB2, LaB6, ZrB2, VB2, MgB2

Fluorides LiF, CaF2, MgF2 Cermets

Si3N4 þ Ni, Al2O3 þ Ni, ZrO2 þ Ni Al2O3 þ Ti, ZrO2 þ SUS, Al2O3 þ SUS WC þ Co, WC þ Ni, TiC þ TiN þ Ni, BN þ Fe,

Intermetallic compounds

TiAl, MoSi2, Si3Zr5, NiAl

Other materials

Organic materials (polyimide, etc.), FRM, FRC, CNT composite materials

NbCo, Nb3Al, LaBaCuSO4, Sm2Co17

3. SUITABLE MATERIALS FOR SPS PROCESS Table 1 represents an example of suitable materials for SPS processing. Figure 4 is a typical example of SPS sintering effect on nano-SiC ceramic material in grain growth and Al2O3 in terms of Hv hardness. Table 2 shows a comparison of characteristics on SPS method and conventional hot press (HP) sintering. The SPS process features very high thermal efficiency because of the direct heating of the sintering graphite mold and compressed powder materials by a large DC pulsed current. It can easily consolidate a homogeneous, high-dense high-quality sintered compact because of the uniform heating, surface purification, and activation made possible by dispersing the spark points (early stage) and/or joule heat points during sintering.

4. PRINCIPLES OF THE SPS PROCESS 4.1. Basic Configuration of the SPS System Figure 5 shows the basic configuration of a typical SPS system. The system consists of an SPS sintering press machine with a vertical single-axis pressurization mechanism, specially designed punch electrodes incorporating a water cooler, a water-cooled vacuum chamber, a vacuum/ air/argon-gas atmosphere control mechanism, a special DC pulse sintering power generator, a cooling-water control unit, Z-axis position measuring and control unit, temperature measuring and control units, an applied pressure

FIGURE 4 Nano-SiC compact by SPS and Al2O3 Hv hardness distribution behavior sintered by various sintering methods.

TABLE 2 Comparison of Characteristics on SPS Vs HP SPS sintering

HP sintering

Temperature gradient sintering



Grain boundary controlled sintering



Fine crystalline structure controlled sintering



Temperature rise rate



Sintering time

Temperature rise time

Fast

Slow

Holding time

Short

Long

B

B

Homogeneous sintering Expandability

O

Productivity

O

Investment in equipment Running cost (

excellent, Bgood, Ofair, difficult)

B

O O

Chapter | 11.2.3

Spark Plasma Sintering (SPS) Method, Systems, and Applications

1153

FIGURE 5 Basic SPS System configuration.

display unit, and various safety interlock devices. Figures 6 and 7 show an example of a medium-sized box-structure experimental-use SPS apparatus and the inside of the water-cooling chamber during SPS sintering.

4.2. Mechanism of SPS Process Despite many years of research work concerning the SPS mechanism by many material researchers, the SPS effect,

in other words, the effect of pulsed high current on the generation of spark plasma, peculiar properties in consolidated materials, still remains unclear [1,18e24]. The SPS process is a dynamical nonequilibrium processing phenomenon that varies from early stage to later stage of sintering along with reacted material characteristics. However, following is one of the most basic proposed ideas on the mechanism of SPS processing. The ONeOFF DC pulse energizing method generates (1)

FIGURE 6 Medium-sized SPS apparatus. (Left: Sumitomo Coal Mining Co., Ltd; Right: Sinter Land Inc.) For color version of this figure, the reader is referred to the online version of this book.

1154

Handbook of Advanced Ceramics

FIGURE 7 Sintering stage and graphite mold in SPS water-cooling vacuum chamber at 1273 K.

spark plasma, (2) spark impact pressure, (3) joule heating, and (4) an electrical field diffusion effect. In the SPS process, the powder particle surfaces are more easily purified and activated than in conventional electrical sintering processes, and material transfers at both the micro- and macro-levels are promoted; so a high-quality sintered compact is obtained at a lower temperature and in a shorter time compared with conventional processes. Figure 8 illustrates a typical ONeOFF DC current path and how pulse current flows through powder particles inside the SPS sintering using conductive powder material, die, and punches made of graphite material. Conventional electrical hot press processes use DC or

commercial AC power, and the main factors promoting sintering in these processes are the joule heat generated by the power supply (I2 R) or high-frequency induction heating elements and the plastic flow of materials caused by the hydraulically or mechanically driving pressure. The SPS process is an electrical sintering technique which applies an ONeOFF DC pulse voltage and high current from a special pulse generator to a powder of particles, and, in addition to the factors prompting sintering described above, it effectively discharges between particles of powder occurring at the initial stage of the pulse energizing for sintering. When a sparking occurs, hightemperature field with sputtering phenomenon generated by spark plasma and spark impact pressure eliminates adsorptive gases and oxide films and impurities existing on the surface of the powder particles. The action of the electro-magnetic field enhances high-speed diffusion due to the high-speed migration of ions. The application of the pulse voltage induces various phenomena as shown in Figure 9 in terms of effects of ONeOFF DC pulse energizing method. In case of WITHOUT sparking due to non-conductive material used or the pulse energizing conditions, however, it is considered that there still exists the effect of ON-OFF DC pulse current energizing which provides enhancement of sinterability and densification rate on the material. The large pulsed energy generates an electro-magnetic field effect such as an electro-migration and preferential orientation of crystalline [3,4,25].

4.3. Neck Formation and Densification by SPS When a spark discharge appears in a gap or at the contact point between the particles of a material at early stage of

FIGURE 8 ONeOFF DC pulsed current path and pulsed current flow through powder particles.

Chapter | 11.2.3

Spark Plasma Sintering (SPS) Method, Systems, and Applications

1155

FIGURE 9 Effects of ONeOFF DC pulse energizing.

sintering, a local high-temperature state (discharge column) of several to ten thousands of degrees centigrade is generated momentarily. This causes evaporation and melting on the surface of powder particles in the SPS process, and “necks” are formed around the area of contact point between particles. Figure 10 shows a basic mechanism of neck formation by spark plasma sintering. Figure 11 is an SEM micrograph showing the result of SPS experiments performed at normal atmospheric pressure (no-load) and 29 MPa applied in vacuum, with a sintering die and punches made of a graphite and a spherical bronze alloy powder (Cu90/Sn10 wt%, particle size 45 mm under). Figure 11(left) shows the behavior in the initial stage of neck formation due to sparks in the plasma. The heat is transferred immediately from the center of the spark discharge column to the sphere surface and diffused so that intergranular bonding portion is quickly cooled. As seen in Figure 11 (center) photo, which shows several necks, the pulse energizing method causes spark discharge one after another between particles. Even with a single particle, the number of portions where necks are formed between adjacent particles increases as the discharges and/ or joule heating are repeated. The right of Figure 11 shows the condition of an SPS sintered grain boundary which is plastic-deformed after the sintering has progressed further. This state is the result of processing conditions in which the applied pressure was 29 MPa, the SPS sintering

temperature was 773 K (measured in the wall of graphite mold), the holding time was 120 sec., the SPS current 850 A, and the voltage was 3.9 V. Figure 12(a) (b) (c) and 13 indicate ultra-high-temperature field existence between particles as collateral evidence examples of SPS effects on metallic and ceramic materials locally. It is suggested that the rippled surfaces observed in Figure 12 (a) were caused by the high-temperature state resulting from thermal and sputtering effects of spark plasma and impact pressure by SPS on atomized cast iron at 973 K. Figure 12 (b) shows sintering by pressureless normal sintering at the same temperature affecting only thermal effect. However, as shown in Figure 12(c) and Figure 13, the local hightemperature state can be observed as a bridging, evaporation, solidification and/or recrystallization phenomena at SPS sintering temperature of Fe/973K, Ni/673K, Al2O3/ 1173K, SiC/1973K respectively. These temperatures are 1/2-1/3 of much lower temperature than each material’s melting point temperature. Thus, an existence of hightemperature field is suggested [26e29] locally. It is noticeable that a design of sintering die and punch assembly made of graphite is extremely important subject to joule heating according to the sintering progress, interaction of powder material, system resistivity, and the function as direct heating elements in order to assume the role of maintaining the homogeneity of sintering temperature [28,30e32].

1156

Handbook of Advanced Ceramics

FIGURE 10 Basic mechanism of neck formation by spark plasma sintering.

Chapter | 11.2.3

Spark Plasma Sintering (SPS) Method, Systems, and Applications

1157

FIGURE 11 Neck formation example of a spherical bronze alloy powder by SPS.

FIGURE 12

Collateral evidence examples of SPS effects on metallic materials [26,28].

4.4. DC Pulse Generator Presently, there are two basic types of DC pulse generator for SPS apparatus which are thyristor-type and invertortype power supply. The waveform, max./min. ON-time/ OFF-time pulse width, peak current, frequency, duty factor settings, control system, and energy consumption are different; however, each system has different advantages individually. Therefore, pulse generator should preferably be chosen to meet the estimated purpose and

usage of SPS processing. Majority of SPS systems working in universities, national institutes, and private companies use the thyristor-type pulse generator due to its rich reference database on SPS and higher reliability of power supply hardware. The invertor type with PWM (pulse width modulation) control provides a low power consumption and compact space-saving advantages, so that it has a high possibility as economical and of lowcost production. The development on pulse generator is still in progress. Both thyristor and invertor types,

1158

Handbook of Advanced Ceramics

FIGURE 13 Collateral evidence examples of SPS effects on ceramic materials [27,29].

FIGURE 14 Typical waveform of thyristor-type SPS pulse generator (ONeOFF ratio in 50 Hz).

FIGURE 15 Typical waveform of invertor-type SPS pulse generator (ONeOFF ratio in 50 Hz).

ONeOFF pulse width, duty factor, and frequency are designed selectively. Figures 14 and 15 show examples of typical ONeOFF pulse waveforms and different pulse widths and ratios.

SPS vacuum exhausting system can also employ a diffusion pump for 5e10  10 3 Pa high vacuum or a dry pump against a sticky gas.

4.5. Vacuum Exhausting System

4.6. SPS Sintering Temperature Measurement and Difference

SPS processing is generally under a vacuum sintering condition. The system allows operating with an inert gas such as argon gas or nitrogen gas, and atmospheric air conditions. By using a rotary vacuum pump with or without mechanical booster pump, pressure ranging from 1105 Pa (air pressure) to 5e6 Pa within 15 minutes will normally be applied. Depending on the sintering purpose and usage,

In SPS process, the sintering temperature is generally measured at the inside of the graphite die wall by thermocouple or surface by pyrometer, and not in the powder directly. Thus, there exist temperature differences between the measured value and an actual filled powder temperature. Figure 16 is an experimental example on nickel powder material indicating 110e120 K difference at 300 sec. The

Chapter | 11.2.3

Spark Plasma Sintering (SPS) Method, Systems, and Applications

1159

4.7. SPS System Machine Layout and Installation

FIGURE 16 Temperature difference between specimen and graphite die.

heating rate was 2.4 K/sec and the open circle in the graph shows sliding surfaces of the graphite punch and die. In accordance with various experiments, it was found that the measured sintering temperatures in both of metals and ceramics by SPS were about 50e250 K lower than the inside of filled powder material. As to nonequilibrium rapid processing, therefore, SPS sintering temperature in SPS data implies this measured temperature [28,30].

FIGURE 17

Figures 17 and 18 show a medium-size SPS system machine overview and 1-MN large-size SPS apparatus machine dimensions and layout drawings. As observed, it does not require a large space and special foundations for installation. Approximately 5.5 m (width) by 5.5 m (depth) in length and 3.5 m in height are sufficient. Cooing water, compressed air, and electric facilities are necessary in operation. Cooling water requires city water pressure level of 0.15e0.3 MPa with a flow rate higher than 20e35 lit./ min. For a DC pulse current output max. 10,000 A, the AC input power requires 175 or 250 kVA (3-phase, 200/220 V, 600 A). Compressed air supply facility requires a pressure level of 0.3e0.5 MPa [{3.0e5.0 kgf/cm2}]. It is usually recommended to make every effort to prevent the generation of electro-magnetic noise around SPS systems that could result in faulty operation such that images on CRT computer screens and electron microscopes located nearby may be affected by interference. Approximate weights of the 1-MN system units are sintering machine 6500 kg and DC pulse generator 2300 kg. The system was designed with special consideration complying with the specifications required for ease of use and safety in comprehensive research on new materials.

200 KN-type R&D and production-use medium-size SPS.

1160

Handbook of Advanced Ceramics

FIGURE 18

Machine system dimensions and layout of 1-MN large-size SPS apparatus.

4.8. SPS Production Systems The development of production-type SPS apparatus involves various considerations regarding production strategies which are cycle time, cost and development of optimum systems, technologies to support scale expansion, automation, mass production and NC (numerical control) system, development of process technologies for high functionality, reproducibility, and uniformity, and structure control and development of 3D near-net shape forming. Figure 19 shows the large-size SPS apparatus for the fabrication of sputtering target materials. As shown in Figure 20, in order to verify SPS productivity, five basic styles of production-type SPS machine systems have been developed: (1) multi-head SPS systems, (2) batch-type SPS systems, (3) tunnel-type SPS systems, (4) rotary-type SPS systems, and (5) shuttle-type SPS systems. In addition to

FIGURE 19 Batch-type large-size SPS System (max. pressure 3 MN/ pulse current 30000A).

these, scaling-up process, automatic handling and powder stacking equipment for materials, and processing optimizer with LCD were also developed. As a result, SPS has become an applicable technique for use as an influential means of industrial manufacturing with high-value-added products [33e38].

5. EXAMPLES OF SPS PROCESS APPLICATIONS 5.1. SiC with Al2O3/Yb2O3 Consolidated by SPS and its Mechanical Properties Rapid sintering is one of the remarkable features of SPS. In order to clarify the advantages and mechanism of SPS method, it was examined on a monolithic ceramic material containing additives. As starting powder materials, 0.28 mm b-SiC with 3 mass%Yb2O3e5 mass%Al2O3 powder was used. Densification behavior and mechanical properties of SPS-consolidated SiC were compared with those of hotpressed SiC. Figure 21 shows that nearly full densification was attained at about 2023 K for SPS-consolidated SiC and at about 2173 K for hot-pressed SiC, respectively. In Figures 22, 23 and 24, the optimal bending strength, Vickers hardness, and fracture toughness (not shown) of SPS-consolidated SiC were 720 MPa, 25 GPa, and 4.0 MPa･m1/2, respectively, and these values were higher than those (640 MPa, 23.5 GPa, and 3.5 MPa･m1/2) of hotpressed SiC, indicating that SPS consolidation can improve the bending strength without degradation of fracture toughness. Pressurization shows a good effect on the densification enhancement remarkably as given in Figure 24. From XRD and SEM (Figures 25, 26) and Raman scattering analysis (not shown), it is suggested that

Chapter | 11.2.3

Spark Plasma Sintering (SPS) Method, Systems, and Applications

1161

FIGURE 20 Typical automated SPS production systems.

FIGURE 21 Dependence of relative density on the sintering temperature in the SPS- and HP-consolidated SiCs.

FIGURE 22 Dependence of bending strength on the sintering temperature in the SPS- and HP-consolidated SiCs.

1162

Handbook of Advanced Ceramics

FIGURE 23 Dependence of Vickers hardness on the sintering temperature in the SPS- and HP-consolidated SiCs.

FIGURE 24 Dependence of relative density on applied pressure in the SPS- and HP-consolidated SiCs.

3C-type disordered structure is preserved in the SPSconsolidated SiC and the improvement of the bending strength without degradation of fracture toughness is attained by the preservation of the 3C-type disordered structure as the SPS effect; even enhancement of grain growth of SPSed SiC was larger [39]. When different additives of Al2O3/Y2O3 were used, the SPSed SiC compact had shown a similar tendency as the SPS effect [40].

5.2. Consolidation of Nano-Al2o3, Phase Transformation, and Grain Growth It is known that the SPS features provide a microstructurecontrolled sintering. Structurally tailoring effect in SPS processing was verified in the consolidation of alumina. A capability of SPS processing for the generation of dense nanostructure was investigated by consolidating g-alumina

FIGURE 25 XRD patterns obtained from raw powder, SPS- and HP- SiCs.

Chapter | 11.2.3

Spark Plasma Sintering (SPS) Method, Systems, and Applications

(a)

1163

TABLE 3 SPS Sintering Conditions of Alumina Al2O3 starting powder materials

SPS sintering SPS sintering Sintering Sintering pressure temperature keeping heating-up (MPa) (K) time (sec) (sec)

g-Al2O3

49

1173

0

60

1273

0

60

1373

0

60

1473

0

60

60

540

600

540

0

60

60

540

600

540

0

60

60

540

600

540

0

60

60

540

600

540

873

180

600

923

180

600

973

180

600

1073

180

600

1173

180

600

1473

60

540

1573

60

540

1673

60

540

1773

60

540

(b) 1573

1673

1773

FIGURE 26 SEM photos of etched surface (a) SPS-consolidated SiC (2073 K, 30 MPa) and (b) HP-consolidated SiC (2273 K, 30 MPa).

nano-powder and a-alumina powders. Up to the present time, SPS consolidation of a-alumina has been performed for various purposes such as the attainment of high bending strength [41], the estimation of microstructure inhomogeneity [42], and of the relation between process factors and densification behavior [20]. Further, generation of nanostructure due to suppression of grain growth has been indicated in a-alumina SPS-consolidated at very rapid heating conditions [43,44]. Here, it is shown that the formation of nanostructured dense alumina is possible in conventional SPS conditions by using nano-a-alumina powder. In this study, first, g-Al2O3 powder particle size of 37 nm, and a-alumina of 0.2 and 0.5 mm were consolidated by the SPS method at a heating rate of 160e197 K/min under the pressure conditions of 49 MPa or 690 MPa. Relative density and powder grain size against the consolidation temperature are shown in Table 3. In Figure 27, the a-alumina was fully densified at 1673 K under the pressure condition of 49 MPa, while g-alumina full densification was attained at 1773 K. Figure 29 shows the phase transformation of g-Al2O3 prepared at 1173e 1473 K under 49 MPa. When the pressure increased to 690 MPa (Figure 28), nanostructured 98% dense alumina with the grain size less than 500 nm was obtained under the conditions of 1173 K.

690

a-Al2O3 (0.5 mm) (0.2 mm)

49

As shown in Figures 28 and 30, it was found that density decreased at 973 K due to phase change to d- and q-phase. Under a high pressure of 690 MPa, the sintering was done at nearly full densification of g-alumina at 600 K lower temperature [45]. Second, a-alumina powder with a grain size of 100 nm used was consolidated by SPS method at a heating rate of 123 K/min under the pressure conditions of 30 MPa or 100 MPa. Nanostructured dense alumina with the grain size less than 300 nm, 99.18%TD, was synthesized under the conditions of 1423 K and 100 MPa. At the same 1423 K with 30 MPa, the sintered one was reached at 86.9% TD. To obtain 98.3%TD dense alumina compact

1164

Handbook of Advanced Ceramics

FIGURE 27 Change of density as a function of temperature under 49 MPa.

FIGURE 28

Change of density as a function of temperature at 690 MPa.

FIGURE 29 XRD patterns of g-Al2O3 prepared at 1173e1473 K under 49 MPa.

FIGURE 30 XRD patterns of g-Al2O3 prepared at 973e1173 K under 690 MPa.

with 30 MPa, a higher temperature of 1473 K was required. In Figure 31 (fractured surfaces of FE-SEM), it is indicated that the difference in the applied pressure at a consolidation temperature of 1423 K drastically changes the microstructure in the SPS-consolidated a-Al2O3; that is, full densification and nanostructure can be attained by SPS consolidation when the preferable pressure condition is selected without selecting a very high heating rate. The heating rate is more important for the synthesis of dense nanostructured and highly transparent alumina, although applied pressure is also correlated with generation of the structural tailoring. The findings suggest that it is important to select an optimum combination between SPS parameters and heating rate, which is delicately effective for generating new micro/nanostructures in the SPS-consolidated ceramics [46]. High-temperature short-period SPS sintering is expected to provide almost all ceramic materials with new characteristics and sintered effects which are different from those obtained by the HP and HIP processes.

Chapter | 11.2.3

Spark Plasma Sintering (SPS) Method, Systems, and Applications

1165

(stainless steel), ZrO2/TiAl, ZrO2/Ni, Al2O3/SUS, Al2O3/ Ti, Al2O3/Ti-6Al-4 V, WC/SUS, WC/Co, WC/Ni, Cu/SUS, SiO2glass/SUS, Al/polyimide resin, Cu/phenol resin, Cu/ polyimide resin materials, etc. [47,48]. All were sintered in ten to twenty-five minutes heating-up and keeping time for a diameter of 20e50 mm after the start of the process. Figure 32 shows a schematic illustration of SPS temperature-gradient die assembly. Figure 33 shows typical examples of bulk FGM compacts synthesized by SPS: from the left, a ZrO2(3Y)/SUS stainless steel compact with 6 interlayers, a ZrO2(3Y)/nickel compact with 7 interlayers, a copper/stainless steel (SUS) compact with 5 interlayers, and an aluminum/polyimide compact with 3 interlayers; on the right, an Al2O3/titanium compact with 3 interlayers. Full dense and no microcracks were detected in the sintered compacts obtained by this process. An example of ZrO2(3Y)/SUS stainless steel compact, a 3 mol% yttrium partially stabilized zirconia (PSZ) powder, a SUS410L stainless steel (SUS) powder, and mixed powders of the two as intermediate layers were stacked on a graphite temperature-gradient die of 20-mm internal diameter. The SUS powder has an average particle size of 9 mm, while the PSZ powder is of granulated particles with an average ˚ ). particle size of 50 mm (its crystalline size is 350 A Sintering pressures used were from 20 to 40 MPa, SPS temperatures of 1243e1293 K, with a temperature rise rate of 50 K/min. Temperatures were measured near the SUS layer [49e53].

FIGURE 31 FE-SEM micrograph of a-Al2O3 prepared at 1423/1473 K under 30 and 100 MPa.

5.4. Demands from Industries on SPS Processing Techniques

5.3. Examples of Synthesized Functionally Graded Materials (FGMs) by SPS

5.4.1. Sputtering Target Material and Fabrication of Large-size Metal and Ceramic Compact

SPS has been used successfully to synthesize a wide range of bulk FGM materials with three to eight or up to 19 intermediate mixed layers, including systems of ZrO2/SUS

With the SPS process, highly dense sintered products can be fabricated at a much lower temperature and with shorter heating-up and holding times compared with hot pressing.

FIGURE 32 Die and punch assembly for SPS temperature-gradient sintering.

1166

Handbook of Advanced Ceramics

FIGURE 33 Typical examples of bulk FGMs.

FIGURE 34 Example of 4 350-mm large-size metallic sputtering target material.

Figure 34 shows typical examples of large-sized metallic sputtering target material with a diameter of 4 350 mm produced by large-size SPS apparatus. The productivity is approximately 7e8 times higher than conventional sintering of HP and HIP processes. Because of the obtained finer grain size of the SPSed sputtering target materials, it can provide superior performance such as no splashing phenomenon in the sputtering process. Large-sized oxide (Al2O3, ZrO2, and SiO2), carbide (WC, SiC, and B4C), nitride (Si3N4), and boride (TiB2) ceramic materials were also fabricated homogeneously with finer grain size and almost full density [54e58]. The samples shown in Figure 35 were investigated by SEM observation that nearly no residual micropores or cracks were detected in the large-size disk-shaped sintered compacts.

5.4.2. Si3N4/Al2O3 Composite Compacts for Homogenizer Component Figure 36 is an example of a fine ceramic sintered compact for homogenizer’s wear-resistant parts. Rapid sintering process can minimize grain growth and manufacture a harder sintered compact with a high relative density of

99e100% and a hardness of over 20 GPa. Pressurized sintering allows for the direct sintering from powders to near-net shape dimensions of ring or cylindrical shape, and eliminating the green body compaction process of normal pressureless sintering. The sintering conditions were an SPS temperature of 1673e1873 K with heating-up and holding time approximately 15e20 minutes totally and an applied pressure of 30e50 MPa. The tolerances of as sintered parts ensure within 0.2e0.3 mm, so that it is easy to obtain a final finishing accuracy by conventional mechanical grinding as post-process.

5.4.3. Pure WC (tungsten carbide) Aspheric Glass Lens Mold Figure 37 is an example of SPSed component in an actual commercial use in the optic industry which is aspheric glass lens molds made of a binder-less pure-tungsten carbide (WC single phase, Hv2600) material without additives and not containing the W2C phase. They were homogeneously consolidated in nanostructured fine grain size. By using ultra-fine grinding machine, the super finishing of mirror surface roughness of Ra5e10 nm can be obtained to fit

Chapter | 11.2.3

Spark Plasma Sintering (SPS) Method, Systems, and Applications

1167

FIGURE 35 Examples of large-size ceramics by SPS.

produced other binder-less WC materials. Under 973 K in atmospheric furnace with 10 hours running test, it resulted in a 30e60% better oxidization in volume (g/cm2).

5.5. 3-Dimensional Complex Near-NetShape Forming of Al2O3 Blasting Nozzle

FIGURE 36 Si3N4/Al2O3 composite parts for homogenizer.

with digital camera lens application. The aspheric glass lens mold consists of three pieces: upper punch, lower punch, and sleeve die part. The advantages of SPSed pure WC material are solid-phase sintering, attaining finer grain, and higher oxidation resistance compared to conventionally

Figure 38 shows an example of an Al2O3 ceramic nozzle for sandblasting machine. The actual lifetime was investigated under the same blast operating conditions. The SPSed Al2O3 nozzle achieved 10 times longer life than a conventionally sintered one made by atmospheric pressureless sintering furnace. As sintered by SPS, the Vickers hardness of 2100e2200, and 3-D net-shape accuracy with Ra0.64mm surface roughness can be obtained. In order to eliminate the usual mechanical grinding process by SPS, the production method has been developed.

1168

Handbook of Advanced Ceramics

FIGURE 37 Examples of pure WC aspheric glass lens mold materials.

FIGURE 38

Comparison of SPSed nozzle and other commercial products.

Chapter | 11.2.3

Spark Plasma Sintering (SPS) Method, Systems, and Applications

1169

FIGURE 39 Appearance of super-plasticity on automotive component.

5.1. Improvement of the Super-plasticity and Specific Strength of Alehigh-Si Alloy Materials The use of SPS process can easily attain nano-structured dense high-silicon/aluminum alloys. The rapidly solidified Al-Si powder (average particle size 120e150 mm, silicon 12e17% or higher and containing a nano-crystalline structure) produced by water atomization process, was successfully consolidated into large sized cylindrical compact with a diameter of 60e120 mm and a thickness of 40e60 mm. By using a steel die for high-pressure SPS, the consolidation process is rapidly performed and maintains nanometric-sized crystals. A sintered compact with a relative density of almost 100% was obtained under atmospheric SPS sintering conditions at SPS temperatures of 723e773 K, applied pressures of 100e150 MPa, and a temperature rise and holding time of about 20 minutes. The TEM micrograph of nanostructure compact in Figure 39 shows that the grain diameter of the sintered compact was between 600 and 800 nm. When the size of 60 mm diameter and 40 mm in thickness nano-structured bulk sintered compact was formed into three dimensional(3-D) shape of automotive engine piston component (Figure 39) by a high-speed forging press machine, it was possible to perform a full forming for lengths of 75 mm within 15 to 20 seconds per stroke and the strain rate of 10 2S 1 or higher. These results indicate that it performs the phenomenon of super-plasticity in the material, and improving the ductility and elongation of the sintered compact. The SPSed material attained a tensile strength of about 350 MPa at normal room temperature, and it is approximately 1.5 times stronger than a conventionally forged component. This AleSi alloy with nanostructural material is expected to find wide applications in electrics, electronics, automotive, and other industrial fields [59].

5.6. Porous Materials by SPS SPS also demonstrates good features on a porous material fabrication in both of metallic and ceramics. Figure 40

shows pure titanium beads and ZrO2(3Y) beads used in porous samples prepared by SPS. They are expected in commercial applications for a bioreactor, a filter, an artificial bone joint, a vent core material for plastic molding die, EV (electric vehicle)/HEV (hybrid electric vehicle) cars, SOFC (solid oxide fuel cell) batteries, thermo-electric semiconductors, a heat spreader of thermal conductive component, and others. Applying lower sintering pressure range of 0e5 MPa, it is possible to attain high porosity with higher bonding strength at neck portion even on an active material of Ti or Al wearing TiO2 or Al2O3 oxide films on the surface of particles compared with other processes [60,61].

5.7. Development of Large-size Functionally Graded Materials (FGMs) 5.7.1. Fabrication of 4 150 mm ZrO2(3Y)/ Stainless Steel Large-size FGMs A ZrO2(3Y) submicron-size partially stabilized zirconia powder containing 3 mol% Y2O3 and SUS410L stainless steel powder with average particle sizes of 9 mm and 60 mm were used as starting materials in this study. The composition of the stainless steel powder was 0.03 wt% C, 0.89 wt% Si, 0.13 wt% Mn, 0.011 wt% P, 0.012 wt% S, 0.09 wt% Ni, and 13.06 wt% Cr. By stacking nine kinds of mixedcomposition powders with ZrO2(3Y)/stainless steel ratios of 90/10, 80/20, 70/30, 60/40, 50/50, 40/60, 30/70, 20/80, and 10/90 vol.% as interlayers between the 100% front and back layers, a total of 11 gradient layers were laminated (Figure 41). As a result of the optimization of sintering conditions, with applied pressure of 30 MPa, SPS temperature at 1253 K, and holding time of 1 minute, the healthy SPSed compact was obtained within 58 minutes, including heating-up time, holding time, and cooling time. A largesized ZrO2(3Y)/stainless steel system FGM with a diameter of 150 mm and thickness of 15 mm having 9 interlayers was synthesized free of cracks, no distortion, and no delamination, as shown in Figure 42. In Figure 43, it is suggested that the optimized SPS conditions of higher sintering pressure

1170

Handbook of Advanced Ceramics

FIGURE 40 Examples of porous material.

and temperature are necessary to prepare homogeneous dense large-sized bulk FGM from the center (00%) to the near perimeter (75%) of the specimen [12,49].

In order to fabricate cobalt content compositionally graded cemented carbide materials, initial research focused on

obtaining homogeneous large-size monolayer WC/Co hard alloys. Although fabrication of hard alloys usually takes many hours, with the development of a new automated SPS system, it is now possible to sinter such materials in a significantly shorter period of time taking advantage of the characteristics inherent in SPS rapid sintering technology. Figure 44 shows the system configuration and an outside view of the automated SPS system. Generally, dimension size and shape effects of SPS tend to fluctuate in

FIGURE 41 Semi-sintered stacked 11 layers of 4 150-mm FGMs by automatic FGM powder stacking equipment.

FIGURE 42 4 150-mm ZrO2(3Y) stainless steel (SUS410L) FGM/LB sintered compact.

5.7.2. Development of Fine-WC/Co Hard-alloy FGMs

Chapter | 11.2.3

Spark Plasma Sintering (SPS) Method, Systems, and Applications

1171

FIGURE 43 Vickers hardness distribution behavior of 4 150-mm ZrO2(3Y)/stainless steel FGMs by SPS.

FIGURE 44 System configuration and outside view of tunnel-type SPS and large-size 100  70 mm WC/Co cemented carbide hard-alloy compacts fabricated in 10 times continuous operation.

hardness around the outer portion. This degradation increases as the size of the sintered compact becomes larger. Utilizing this tunnel-type automated SPS system and optimized SPS conditions, square-shaped large-size WC/Co cemented carbide hard alloy with dimensions of 70 mm(width)  100 mm (length)  5e20 mm (thickness) were homogeneously fabricated in a shorter sintering time and with a finer grain size than with conventional sintering

methods while maintaining high quality and repeatability every 10 minutes or less per piece, as shown in Figure 44 [62]. Table 4 shows the typical mechanical properties of WC/Co tungsten carbide hard alloys sintered by SPS. A fine tungsten carbide powder of an average particle size less than 0.5 mm containing 0e6 weight percent cobalt was used as a starting powder material. Figure 45 is a comparison with other commercial products. The fine-WC/Co hard

1172

Handbook of Advanced Ceramics

TABLE 4 Typical mechanical properties of WC/Co hard alloys sintered by SPS Product cord name

Co content (wt%)

Grain size (mm)

Density g/cm3

Hardness mHv

Transverse rupture strength kg/mm2

Fracture toughness K1c

TC10