Nanostructured Silicon-based Powders and Composites

Citation preview

1

Objectives and state-of-the-art of nanocomposites Michel Cauchetier and André Pierre Legrand

1.1 Introduction Processes to make new ceramic materials are many and depend on the final desirable structure and macroscopic shape to be obtained. So silicon carbide can be used as an abrasive, refractory, structural material or semiconductor with a negative temperature coefficient. Processes developed to produce such materials are extremely diversified. A process is selected based on the required properties and structural shape of the parts to be obtained. As an example, the preparation of carbon fibres, which needs specific polymer precursors for spinning followed by heat and oxidation treatments, had been adapted to the preparation of silicon carbide fibres. Other processes of transformation, descended from physics or chemistry domains are chemical vapour deposition, chemical liquid deposition, etc. (Figure 1.1). They are able to present improved mechanical, thermal or electrical properties, not solely depending on a peculiar chemical composition but on the arrangement of the crystalline phases or the size of the grains. Nanocomposites constitute a new class of extremely diversified materials which have appeared in the last decade.

Monomer unit polymer precursor

Liquid or gaseous precursor molecules

Chemical vapour or liquid deposition

Solid state thermolysis

Laser pyrolysis

Sin

Mechanical, electrical and thermal properties

Analys is

ing ter

Synthesis

Characterisation by XPS X-ray and neutron diffraction TEM, SEM, optic solid state NMR

Figure 1.1 Schematic representation of ceramic preparation showing the feedback process leading to a new ceramic material involving three main steps: synthesis of a precursor; analytical methods; and sintering and validation of the physical and chemical properties.

©2003 Taylor & Francis

Vitreous grain boundary (YSiAION)

Si3N4 (SiAION-
)

1 µm Micro–nano composite

SiC

80 nm

Nano–nano composite

Figure 1.2 Ceramic/ceramic micro–nano and nano–nano composites’ aspects (from Figures 7.8 and 7.20(b)).

Sajgalik et al. (1996) proposed a classification into two main systems: • •

micro–nano composites in which the matrix is constituted by micrometric crystals embedded into a nanometric phase. Moreover, Niihara (1991) distinguishes three different configurations: intragranular, intergranular, intra/intergranular (Figure 1.2); nano–nano composites in which the nanometric grains of the matrix and the second phase are spread uniformly (Figure 1.2).

What makes a nanocomposite especially interesting is that at least one of its phases has dimensions in the nanometer range (10–100 nm). In this range, chemical and physical interactions have critical length scales and, if the nanoscale building block is made smaller than this one, the corresponding fundamental properties can be to changed. An example of this is light scattering, where by controlling the size of pores and nanocrystals in the range of 8 nm, Nanophase Technologies Corporation was able to produce transparent ceramics in the visible domain. Other enhanced properties have been developed (superplasticity, magnetoresistance, low temperature densification, enhanced and finer homogeneity, etc.).

1.2 Nanostructured ceramic elaboration Two main routes are generally used, depending on the densification rate and the properties to be obtained. 1.2.1

Amorphous ceramics obtained by solid state thermolysis of polymer precursors

Pyrolysis, under a controlled atmosphere, of polymers, such as polysilazane, polycarbosilane, etc. is an efficient process for preparation of ceramics. According to Bill et al. (1999) ‘The in-situ crystallisation of these materials permits the preparation of nanocrystalline materials by a completely powder-free process. The structure and composition of the grain boundaries that originate from these crystallisation processes strongly depend on the molecular structure

©2003 Taylor & Francis

of the initially used precursors.’ Such processes are particularly interesting in the preparation of fibres. 1.2.2

The powder-making processes

To obtain nanocomposites it is necessary to prepare nanopowders so as to increase the reactivity of the different phases during the sintering process. This is due to a high surface/volume ratio (generally diameter < 100 nm). To obtain such powders, which have to present ideal characteristics (nanometric in size, spherical in shape, monodisperse, low agglomerated, high or controlled purity), it is necessary to develop new processes. These new methods must be innovative and need to be adapted to industrial production. Such recently developed processes deal with reactions such as • • •

Solid phase: Evaporation–condensation, laser ablation, ball milling (or mechanical alloying), self-propagating high-temperature synthesis (SHS); Liquid phase: Sol–gel, spray-drying of solutions, aerosol pyrolysis; Gaseous phase: Chemical vapour condensation at low pressure, plasma and laser synthesis. Some nanosized powders are now commercially available in the USA and in Japan:

• • • • •

Nanophase Technologies Corp. has developed a gas condensation process to produce different oxide powders (aluminium, titanium, zirconium oxides). Nanodyne is focused on the production of cobalt–tungsten carbide, a ceramic–metal nanocomposite used to make tools for cutting and wear-resistant devices. Ultram International produces ceramic and composite powders by a super-highfrequency plasma chemical process. MarkeTech International proposes nanosized silicon carbide, silicon nitride, silicon carbonitride, boron carbide and silica powders obtained by a CO2 -laser pyrolysis process. Mitsubishi Mining & Cement Co. Ltd has prepared silicon/carbon/nitrogen nanocomposite powders from pyrolysis in a furnace of an organosilicon compound.

A second problem which needs to be solved concerns the sintering for the preparation of complex shaped components so as to keep the initial size of the crystals. Similarly, the identification of the most efficient methods of characterisation, all along the different steps of preparation, is of significant importance.

1.3 Nanostructured Si3 N4 /SiC materials Different methods have been developed for the preparation of Si3 N4 /SiC nanocomposites by hot pressing sintering using: •

Micrometric silicon nitride, previously recovered with carbon, so as to produce the carboreduction of silicon oxide present at the surface of the initial material: SiO2 + 3C → SiC + 2CO ↑ (Ishizaki and Yanai 1995; Watari 1989; Bahloul et al. 1993).

©2003 Taylor & Francis

•

•

Micrometric silicon nitride with nanometric Si/C/N or Si/C powders and sintering additives (Al2 O3 , Y2 O3 ) (Niihara 1990). Nevertheless, the mixture, in a liquid medium, of such different sized powders containing so many different chemical groups on the surface, is difficult to obtain. This produces agglomerates and a heterogeneous distribution of the second phase in the matrix. Nanometric Si/C/N or Si/C powders with sintering additives (Al2 O3 , Y2 O3 ) (Sasaki 1993; Kaiser 1996; Kennedy 1996).

1.4 Analytical methods Different methods can be used for the characterisation of initial to heat-treated and/or pressurised precursors such as: • • • • • • • • • •

Chemical analysis (CA) Electron paramagnetic resonance (EPR) Extended X-ray absorption fine structure (EXAFS) Infrared spectroscopy (IR) Soft X-ray spectroscopy (SXS) Solid state nuclear magnetic resonance (NMR) Specific surface area determination (SA) Transmission electron microscopy and electron diffraction (TEM) X-ray (XD) and neutron (ND) diffraction X-ray photoelectron spectroscopy (XPS).

Such methods are able to provide information of varied nature, depending on the origins of the physical phenomena involved (Table 1.1). Another important remark concerns the method of preparation of the sample. From the as-formed, heat treated and sintered sample, it is necessary to arrive at a condition adapted to the method of measurement. Crushing the material and filling, dispersing, spreading it in/on the sample holder are able to influence the results. This is why cross measurements are necessary. Table 1.1 Available information depending on the method used Method CA EPR EXAFS IR ND NMR SXS TEM XD XPS SA

Surface

× ×

Bulk × × × × × × × × ×

Crystalline

× × × × ×

Amorphous

× × × × ×

Note × Indicates the dominant domain of expertise of the method.

©2003 Taylor & Francis

Local or short order

×

Average or long order

Sample type

× × × × × × ×

Powder Powder Powder Powder Powder Powder Any Nanopowder Powder Any Any

× × ×

Some examples shown in this book demonstrate such phenomena: • • •

X-ray, TEM and NMR. Figure 2.3, section on Si/C-based powders in Chapter 4 and Figure 5.1.6: α-β-SiC and amorphous detected in different proportions; NMR and TEM. Figures 4.65, 4.66 and 5.1.27: Si3 N4 detected in the bulk and not into grains; XPS and NMR. Figures 5.3.10 and 5.1.10: SiO2 detected on the surface and not in the bulk;

where the different phases are not evaluated with the same efficiency.

1.5 Conclusion Although till now many studies on that subject have been published, results are contradictory and fragmentary. No interdisciplinary studies, till now, have been done on the silicon nitride– silicon carbide system. A better knowledge of the behaviour and potential of such materials is needed. Elaboration of high-temperature ductile nanocomposites is one of the objectives of this work. Consequently, systematic studies and analyses of the preparation proceed from step by step. Chemical bonding, nanostructure, microstructure and monolith material properties are determined. Such an approach allows us to obtain original results for each step of the preparation and to establish relations between the synthesis of powders, their sintering aptitude, the development of the microstructure and the properties obtained. The spiral in Figure 1.1 represents this symbolically.

References Bahloul, D., Pereira, M. and Goursat, P. (1993) J. Am. Ceram. Soc., 76(5), 1156–68. Bill, J., Wakai, F. and Aldinger, F. (1999) Precursor-Derived Ceramics. Wiley-VCH. Ishizaki, K. and Yanai, T. (1995) Silic. Indus., 7–8, 215. Kaiser, A. (1996) Silic. Indus., 5–6, 215. Kennedy, T. (1996) Silic. Indus., 9–10, 201. Niihara, K. (1990) J. Mater. Sci. Lett., 10, 112. Niihara, K. (1991) The Centennial Memorial Issue of the Ceramic Society of Japan, 99(10), 974. Sajgalik, P., Dusza, J., Hofer, F., Warbichler, P., Reece, M., Boden, G. and Kozankova, J. (1996) J. Mater. Sci. Lett., 15, 72. Sasaki, G. (1993) Mater. Res. Soc. Symp., 287, 335. Watari, K. (1989) Mater. Sci. Eng., A109, 89.

©2003 Taylor & Francis

2

Laser synthesis of nanosized powders Michel Cauchetier, Emmanuel Musset, Michel Luce, Nathalie Herlin, Xavier Armand and Martine Mayne

2.1 Generalities on the synthesis of nanosized powders by laser pyrolysis 2.1.1

The laser pyrolysis process

Infrared laser pyrolysis (IRLP) is a highly versatile method for the production of a wide range of nanopowders including SiC, SiCN, SiCO, BN and carbon (Haggerty and Cannon 1981; Cauchetier et al. 1988, 1994; Herlin et al. 1996; Boulanger et al. 1995; Ehbrecht et al. 1993; Voicu et al. 1996). A summary of the syntheses performed so far for materials applications is given in Table 2.1 and completes the review appearing in Knudsen (1997). In the IRLP process, the reactants are heated by IR laser radiation and decompose, causing clusters to nucleate and grow rapidly. This process is inherently very clean because homogeneous nucleation occurs in a well-defined reaction zone without interaction with chamber walls. The small reaction volume and the ability to maintain steep temperature gradients (106 ◦ C/s) allow precise control of the nucleation rate, growth rate and residence times. The physical and chemical properties of the particles can be controlled by changing the molecular precursors and the synthesis parameters (laser power, pressure, etc.). The resulting powders are very fine, spherical, extremely pure, more or less agglomerated and nearly monodispersed in size. The mean particle size can be adjusted from 10 to 100 nm. All these characteristics (i.e. nanometric size (1 ton/year) even if energy and laser equipment costs are low compared to the raw material costs. In the case of Si3 N4 laser synthesis, a study of the economics shows that a price for SiH4 less than one-fourth of its current price is needed to compete with a classic method of synthesis (Schoenung 1991). Alternative methods have been proposed with the substitution

©2003 Taylor & Francis

of chlorosilanes (Bauer et al. 1989; Suzuki et al. 1992) or organosilicon compounds (Rice 1986; Gonsalves et al. 1992; Cauchetier et al. 1994) for silane. Technical improvements in two different fields have enhanced the versatility of the method. First, new introduction systems for the silicon (Si) precursors have been tested. Liquid organosilicon compounds are introduced into the laser beam in the aerosol form, fine liquid droplets being obtained by ultrasonic nebulisation (Gonsalves et al. 1992; Cauchetier et al. 1994). Second, the use of high-power tunable continuous wave (c.w.) CO2 lasers, now available, has increased the area of applications of the process (Luce et al. 1994). In this chapter, the synthesis of Si-based powders by two methods, using gaseous or liquid precursors, will be presented: • •

SiH4 as gaseous precursor for the synthesis of Si, SiC, Si3 N4 and SiC–Si3 N4 mixture powders and also methylsilane, SiH3 CH3 , for the synthesis of SiC powder, hexamethyldisilazane [(CH3 )3 Si]2 NH or HMDS as liquid precursor for the synthesis of Si/C/N (silicon carbonitride) powders; mixture of organosilicon precursors for the synthesis of Si/C/N powders containing the elements of sintering additives.

2.2 Synthesis of Si, SiC, Si3 N4 and SiC Si3 N4 mixtures with SiH4 and SiH3 CH3 as silicon precursors 2.2.1

Experimental setup

The experimental device is presented in Figure 2.1 (Cauchetier et al. 1991). The unfocused Gaussian beam (diameter = 12 mm) of a high-power c.w. CO2 laser (CILAS CI1000, Marcoussis, France) enters the reaction cell through a KCl window and crosses the path of the gaseous flow of the reactants injected through an inlet capillary (inner diameter = 2 mm). Pumping IR spectrometer powder collector

CO2 laser beam (600 W)

Ar SiH4 + CH2NH2 (+C2H2 + NH3)

Figure 2.1 Schematic laser irradiation cell for gaseous precursors. (Reprinted from Cauchetier, M., Croix, O., Luce, M., Baraton, M. I., Merle, T. and Quintard, P., (1991) Journal of the European Ceramic Society, Nanometric Si/C/N Composite Powder: Laser Synthesis and IR Characterisation, 8, 215, with permission from Elsevier Science.)

©2003 Taylor & Francis

A resonance effect between the emission line of the laser at 10.6 µm and an infrared absorption band of one of the reactant gases (silane, SiH4 or methylsilane, CH3 SiH3 ) causes the reaction to occur. This laser-driven reaction leads to high temperatures (up to 1800◦ C) and a bright flame. An argon flow prevents powder deposition on the windows and guides the products into the collection chamber. The powders are collected and then stored in a glove-box under argon atmosphere to avoid contamination by air or water vapour. 2.2.2

Synthesis results

Table 2.2 summarises the synthesis conditions of the different powder samples used in different parts of this book (Si, SiC, Si3 N4 and mixtures of SiC/Si3 N4 ) and also presents first characterisation results. For all the experiments reported here, the cell pressure was kept constant at 1 atmosphere (105 Pa). In Table 2.2, most of the results concern SiC and Si/C/N synthesis, but some results shows the synthesis of Si from pure SiH4 (Si50), Si3 N4 from a mixture of SiH4 and NH3 (SiN7). Structural investigations on both these powders are presented in Chapter 5. SiC powders have been obtained from methylsilane (SiC151, SiC152) or from mixtures of SiH4 and C2 H2 . In the run, SiC171 diborane was added in the silane/acetylene mixture, as a boron source used for sintering aid in densification tests (Croix et al. 1991), and after pressureless sintering at 2000◦ C, densities up to 0.95 were achieved. The last samples presented in Table 2.2 are Si/C/N samples, obtained from a mixture of SiH4 , CH3 NH2 , and, in some cases, NH3 has been added to the reactive mixture. In two runs (Si50, SiC163), an inert gas (Ar, He) has been added to the reactive gases in order to increase the linear velocity. After comparing the chemical composition of the gas phase with the chemical composition of the powder, the C/Si and C/N ratios were obtained (see Table 2.2). In some cases, the flame temperature was measured using an optical pyrometer. Specific surface area (S) was determined from Brunauer–Emmet–Teller (BET) measurements and the equivalent diameter (D) was calculated from these measurements. The time of residence in the laser beam is calculated from the gas flow rates (2 mm nozzle diameter, 12 mm laser beam diameter) assuming room temperature. From the results presented in Table 2.2, some information can be obtained on the effect of experimental parameters. Laser power Increasing the laser power, and keeping all the other parameters constant, leads to an increase in temperature, which is expected to have an influence on the powders produced. In experiments SiC151 and SiC152, the laser power has been decreased by a factor of three (600/220 W) and the measured temperature decreases from 1530◦ C to 1150◦ C. The same crystalline structure is identified by X-ray diffraction (XRD). Therefore, in this temperature range, the flame temperature does not influence the structure of the powder. In contrast, BET measurements indicate a decrease in the powder size as the temperature decreases. Thus, the temperature has a direct influence on the kinetics of the powder growth. Flow rates In runs SiC163–177, the flow rates have been changed keeping the laser power constant. Increasing the flow rates leads to a decrease in the residence time as shown in Table 2.2. Figure 2.2 presents the variation in the particle size of SiC powders with the residence time for samples SiC163–177. Figure 2.2 also presents results (broken line) obtained five years before

©2003 Taylor & Francis

Table 2.2 Gas phase synthesis conditions and characterisation results of the different samples Samples

Flow rates (cm3 / min) SiH 4

Si50 SiC151

120

SiC152 SiC163 SiC171a SiC173 SiC174 SiC177 SiC212 SiN7 SiCN12

540 1080 200 350 120 596 120 340

SiCN27 SiCN29 SiCN35

280 360 350

C2 H2

Atomic ratio (gas phase)

He, Ar

0 480 (Ar) CH3 SiH3 : 200 CH3 SiH3 : 200 300 1000 (He) 600 0 110 0 192 0 66 0 306 0 0 0 0 0 0 0 0

0 0 0

CH 3 NH 2 0

NH 3

C/N

0 0 0 0 0 0 480 0

70 90 200

250 320 0

Temperature (◦ C)

1.1 1.1 1.1 1.1 1.1 1.0

0.22 0.25 1 0.57

BET results

XRD

S (m2 /g) D (nm) 600 600

930 1530

4.3 13.0

29 34

89 55

Si β-SiC

220

1150

13.0

52

36

β-SiC

1250 1550

1.2 1.3 7.3 4.2 12.1 2.5 4.3 4.8

117 135 61 74 36 70 71 47

16 14 31 25 52 27 26 40

4.3 2.9 4.1

26 40 48

72 49 39

600 600 600 600 600 640 600 600 600 400 600

1615

Note a 40 cm3 /min of diborane B2 H6 added as boron source which acts as sintering aid for the densification of SiC.

©2003 Taylor & Francis

Residence time (ms)

C/Si

0

0 0 0 0 0 0 0 200

Laser power (W)

α, β-Si3 N4 , β-SiC α, β-Si3 N4

BET diameter (nm)

50

SiC177

SiC173 25

SiC174 SiC163

0

5

10

15

Residence time (ms)

Figure 2.2 SiC particle size v. the residence time in the laser beam (laser power, 600 W; cell pressure, 1 atm). Similar results have already been observed previously (– – –) with a laser power of 640 W. (Reprinted from Tougne et al. (1993) Diamond and Related Materials, Evolution of the Structure of Ultrafine SiC-Laser-formed Powders with Synthesis Conditions, 2, 486–490, with permission from Elsevier Science.)

the present results, with a higher laser power (640 W) (Tougne et al. 1993). It must be noted that the value obtained for the SiC212 (640 W) experiment (not plotted) is in good agreement with these early results. Figure 2.2 clearly shows a linear decrease in the size of the particles as the residence time decreases. Figure 2.3 presents the XRD diagrams of the SiC163–177 samples. The lines become sharper and the full width at half maximum (FWHM) decreases as residence time increases, which indicates an increasing size of crystallites, parallel to the increasing particle size (BET measurements). TEM observations also confirm the BET and XRD results. Sample SiC163 (Figure 2.4(a)) presents fine spherical particles with a mean diameter of 20 nm (BET measurement: 16 nm) and with a unimodal distribution. They are not fully crystallised and appear as fine crystallites in an amorphous matrix. It clearly shows that the size of the particles (BET) and the size of the crystallites (XRD) must be distinguished. Sample SiC177 (Figure 2.4(b)) also presents fine spherical particles with an increased mean diameter of 60 nm (BET measurement: 52 nm). They are well crystallised with some black streaks. In correlation with the chemical composition C/Si = 1.1, some fine filaments of turbostratic carbon (C) can be noticed. In conclusion, it must be noted that both laser power and flow rates have a noticeable effect on the size of the particles. A good control of these experimental parameters allows a good control of the size and the structure of powders. Some other points which are relevant for materials applications must be noted:

•

The amount of synthesised powders is significant. For example, in the run SiC171, a production rate of about 100 g/h indicates that the process can be easily scalable. The

©2003 Taylor & Francis

3C

Intensity (a.u.)

Polytypes

3C

SiC177

3C

3C

3C

SiC173

SiC174 SiC163 20

40

60

80

2 (CuKα)

Figure 2.3 XRD patterns of the as-synthesised SiC powders (arbitrary units (a.u.)). (Reprinted from Tougne et al. (1993) Diamond and Related Materials, Evolution of the Structure of Ultrafine SiC-Laser-formed Powders with Synthesis Conditions, 2, 486–490, with permission from Elsevier Science.)

• •

duration of the experiment can be 3 h or more in order to collect powder quantities in the 150–200 g range for the elaboration of materials. Also, a very good reproducibility of the experiments can be achieved; for example, Table 2.2 shows BET results for the runs SiCN12 and SiCN35 undertaken after a delay of 4 years. Another interesting point is the good yield of the reaction. For example, in the SiC212 run, and supposing 100% conversion of SiH4 and C2 H2 to the solid phase, the maximum calculated production rate is 64 g of SiC per hour and the experimental result is 59 g/h – this reaction yield (almost 100%) is very good.

2.2.3

Chemical analyses of the as-formed powders

The chemical composition in stoichiometric compounds (Si3 N4 , SiC, SiO2 , Si, C) has been calculated from elemental analysis results, with the classical assumption that all O is in the form of SiO2 , then all N forms Si3 N4 and the remaining Si is in the form of SiC. Free C (or Si) is the difference between total C (or Si) and C (or Si) bonded to Si (or C) in SiC. Such a calculation is useful to evaluate the composition of future sintered materials, or to compare several powders but it is obvious that it is only an approximation, especially when the powders are not fully crystallised.

©2003 Taylor & Francis

(a)

100 nm

(b)

100 nm

Figure 2.4 TEMs of two as-formed powders: (a) SiC163; (b) SiC177.

SiC powders In order to avoid the presence of free Si in the powders which makes the powder very sensitive to oxygen contamination, most of the SiC samples were synthesised with a gaseous mixture containing excess C (C/Si = 1.1 in the gaseous mixture). In Table 2.3, two examples show that this ratio is also found in the powders produced. In agreement with the chemical composition, XRD patterns (similar to Figure 2.3) correspond to the β-SiC phase. The O and free C contents are low and correspond to the values usually encountered for commercial powders. Table 2.3 also presents the chemical composition calculated in stoichiometric compounds which indicates, in good agreement with the XRD pattern, that the powders are mostly composed of SiC. Si/C/N powders For the synthesis of Si/C/N composite powders, two gaseous mixtures were investigated: a binary mixture SiH4 + CH3 NH2 (SiCN35) and a ternary mixture: SiH4 + CH3 NH2 + NH3

©2003 Taylor & Francis

Table 2.3 Chemical characterisations of gas phase synthesised powders Powder

SiC173 SiC212 SiCN29 SiCN35

Samples Chemical analysis (wt%)

Atomic ratio Chemical composition (wt%)

Si

C

N

O

C/Si

68.3 68.3 58.3 66.9

30.2 30.3 6.5 13.6

0 0 34.6 18.2

0.9 1.4 0.6 2.3

1.03 1.03 0.27 0.47

Si3 N4

SiO2

0.22 0.87

0 0 86.6 45.1

1.5 2.7 1.1 4.3

SiC 96 95.7 8.3 45.1

C 2.0 1.6 4.0 0

Si 0 0 0 5.5


SiC Si

Intensity (a.u.)

 Si3N4
Si3N4

C/N

SiCN35

SiCN29

10

20

30

40 50 60 Angle (2 degrees)

70

80

90

Figure 2.5 XRD patterns of as-formed SiCN nanopowders used in material elaboration. (Reprinted from Mayne 1997.)

(SiCN29) (Table 2.2). In the case of SiCN29, the presence of NH3 is known to make the nitriding introduce an increase in the nitrogen (N) content. For SiCN35 a similar content in C and N is expected in the resulting powder due to the C/N initial atomic ratio of 1 in the formula of monomethylamine. Tables 2.2 and 2.3 show a very good correlation between C/Si and C/N ratios in the gas phase and in the powder for SiCN29, but the correlation is not so good for SiCN35 probably due to evolution of carbonaceous gas of C in the gas phase during the synthesis. Chemical composition calculations show that the powder SiCN35, containing free Si is the most sensitive to oxygen contamination. XRD patterns (Mayne 1997) are reported in Figure 2.5 and confirm the calculated compositions. The SiCN29 sample contains mainly α- and β-Si3 N4 ; the quantity of SiC is too small and is not detectable. The SiCN35 sample presents the characteristic lines of Si and β-SiC and a broad peak between 2θ = 30◦ and 40◦ , which can be attributed to an amorphous phase of silicon carbonitride. From the results presented in this section, it appears that the decomposition in stoichiometric compounds is a good representation of the powder and also that the chemical composition of the powders produced is most often well controlled by the chemical composition of the gaseous reactants. Together with the possibility of controlling the size and the structure of

©2003 Taylor & Francis

the powder, it makes this synthesis technique attractive. But, as explained before, in order to increase the safety and decrease the cost, it has been applied to the liquid precursors containing Si, C and N elements.

2.3 Synthesis of silicon carbonitride (Si/C/N) powders using hexamethyldisilazane as Si precursor 2.3.1

Experimental

Choice of the precursor As noted before, investigations based on costs of various silicon precursors for the delivery of large quantities have shown that silane remains too expensive for use in ceramic powder synthesis (Cauchetier et al. 1994). In our case [(CH3 )3 Si]2 NH or HMDS with a Si cost very low compared to the Si cost of SiH4 (43 US$/kgSi in HMDS and 343 US$/kgSi in SiH4 ) was chosen as a model of the liquid precursor in order to demonstrate the possibility of using liquid wastes of the organosilicon compound chemistry for the synthesis of nanometric Si-based powders. Indeed, HMDS has a strong IR absorption band at 10.6 µm due to the Si N bond, which was, therefore selected. It is a common liquid Si compound (boiling point of 125◦ C, density of 0.76, and a viscosity of 0.69 cP at 25◦ C) largely used for the silylation of a wide range of functional groups and it was the precursor used by Niihara and co-workers in the synthesis of Si/C/N powders leading to non-oxide superplastic ceramics (Wakai et al. 1990). The first experiments using HMDS in the vapour phase as Si precursor for Si/C/N powder synthesis by CO2 -laser pyrolysis were performed by Rice (1986). A conversion efficiency (liquid → powder) of about 60% has been obtained but with low flow rates of HMDS (7 g/h) and with a low-powered CO2 laser (135 W). Higher production rates have been attained (80–120 g/h) when the laser power has been increased to the 500–1000 W range (Li et al. 1994b). The experimental device The apparatus is shown in Figure 2.6. Liquid HMDS is placed in a special glass jar containing a piezoelectric transducer (Pyrosol 7901 type from RBI, Meylan, France, apparatus developed in collaboration with the CEA (Spitz and Viguié 1970)). The focusing of the intense beam of ultrasonic energy delivered by the transducer near the surface of the liquid yields uniform droplets whose diameter d is given by d=


3

πσ 4ρf 2

where σ and ρ are the surface tension and the density of the liquid, respectively, and f the frequency of the transducer (here 850 kHz). In order to increase the mass of the displaced liquid, the glass jar is heated near 90–100◦ C with a heating ribbon. The aerosol droplets and vapour are injected into an irradiation cell very similar to those presented in Figure 2.1. The reaction cell is maintained at a regulated pressure of 105 Pa using a flow of argon or an argon–ammonia mixture through a glass inlet tubing (inner diameter of 13 mm). The laser beam diameter is increased from 12 to 24 mm with a beam expander in order to cover the reactant flow entirely. The initial laser power is in the range 480–520 W.

©2003 Taylor & Francis

Filter

Powder collector

Heating wires Vacuum pump or IR spectrometer

KCl window

Laser CO2

Argon

Argon Chimney gas

Gas for the propulsion of the aerosol

Piezoelectric ceramic

Precursor RF power supply PYROSOL system

Figure 2.6 Schematic of the aerosol generator with the irradiation cell and the powder collector.

2.3.2

Synthesis results

Table 2.4 compares the influence of two parameters, the nature of the carrier gas (argon and nitrogen) and heating of the liquid precursor (near 100◦ C), on the synthesis results. It can be seen immediately that the yield of the reaction (expressed as the ratio between the mass of powders produced and the mass of displaced liquid), does not appear to be as good as in the gas phase. Both the amount of displaced liquid and production rate increase when the precursor is heated. The effect of the heating is to combine the enrichment of the gaseous phase in HMDS vapour and decrease the viscosity of the liquid. In order to work with significant amounts of powders, most of the experiments were carried out with liquid heating. The yield seems independent of the heating of the precursor, but is strongly dependent on the nature of the gas. It is 46–49% for Ar carrier gas flow and 38–39% for N2 carrier gas flow.

©2003 Taylor & Francis

2.3.3

Effect of experimental parameters

Effect of carrier gas (Ar or N2 ) Table 2.4 shows that the chemical composition of the powder does not depend on the nature of the carrier gas flow (Ar or N2 ) and the C/N atomic ratio remains stable in the 2.5–2.9 range. The activity of atomic N (N2 ↔ 2N) is low at the measured flame temperature: 1260 ± 20◦ C. Table 2.4 also shows that all powders are sensitive to pollution by oxygen, but the powders obtained from the heated precursor seem the most sensitive to pollution. It must be noted that the chemical analysis presented in this section is not complete due to the presence of hydrogen in the powder and due to the uncertainty of the measurements. Effect of chemical evolution of the reactive mixture Table 2.5 presents the synthesis parameters of all the samples studied in the different parts of this book. Only samples HMDS40–45 where the effect of chemical composition has been

Table 2.4 Effect of the carrier gas nature and precursor heating Run

Carrier gas flow Displaced rate (cm3 /min) liquid (cm3 /h) Ar N 2

HMDS28a 2430 — HMDS36b 2500 — HMDS38b — 2500 HMDS39a — 2500

108 164 166 95

Powder production rate (g/h) 38 61 48 28

Yield SBET Chemical analysis (wt%) (m2 /g) (wt%)

46 49 38 39

88 82 125 79

O

C/N (Atomic ratio)

Si

C

N

49.8 45.8 47.2 49.9

30.3 28.0 28.4 29.7

12.3 5.6 2.9 11.7 9.7 2.8 12.8 11.6 2.6 14.1 5.2 2.5

Notes a Liquid precursor at room temperature. b Liquid precursor heated near 100◦ C.

Table 2.5 Synthesis conditions and chemical analysis of HMDS samples (chemical analyses were always performed after several weeks of exposure to air) Run

Ar HMDS40 HMDS41 HMDS42 HMDS43 HMDS44 HMDS45 HMDS35 HMDS66 HMDS67a HMDS73a

Displaced Powder Yield SBET Chemical analysis liquid production (wt%) (m2 /g) (wt%) rate (g/h) (cm3 /h) Total Si C N O

Carrier gas flow rate (cm3 /mn) N2

1935 205 2140 182 1770 410 2180 160 1570 600 2170 175 1370 800 2170 179 1150 1040 2190 210 830 1346 2180 167 1600 1000 2600 87 1920 1920 114 1440 1920 115 862 207 1920 56

Note a H2 in the carrier gas.

©2003 Taylor & Francis

59 61 75 74 81 66 21 32 31 14

43 50 56 53 51 52 32 36 35 32

110 93 98 95 95 106 144 66 66 115

46.7 46.1 44.4 48.3 47.1 47.6 47.3 51.2 52.0 51.9

25.3 20.5 18.5 15.5 13.1 8.3 6.6 31.9 30.8 12.6

C/N (Atomic ratio)

21.8 6.1 1.35 25.2 6.7 0.95 29.7 7.7 0.73 26.8 8.0 0.67 26.7 10.9 0.57 28.9 13.4 0.34 33.4 9.8 0.23 15.1 1.8 14.8 2.4 25.3 10.1

3.0

HMDS28

C/N atomic ratio (powder)

2.5

22

2.0

1.5 40 1.0 41

43 0.5

42 44

35

45

1

2 3 4 5 C/N atomic ratio (gas + aerosol)

6

Figure 2.7 C/N atomic ratio in as-formed HMDS powders v. C/N atomic ratio in the precursors (liquid + ammonia). (Reprinted from Musset et al. 1997 with the permission of the Belgian Ceramic Society, Mons, Belgium.)

closely studied will be commented on in this section; the main part of the results appear in the PhD thesis of Emmanuel Musset (Musset 1995). For this set of six experiments, the liquid temperature is about 100◦ C, the total flow rate of the carrier gas mixture remains constant and the corresponding residence time of the precursor in the laser beam is about 0.1 ms. Table 2.5 shows the influence of the addition of ammonia in the argon used as carrier gas, namely, on the chemical composition of the resulting powders (samples HMDS40–45). Figure 2.7 shows a good correlation between the C/N ratio in the as-formed powder and in the reactive mixture; the C/N atomic ratio decreases from 1.35 to 0.36 when the ammonia volume ratio increases from 10% to 60%. This correlation confirms the possibility of controlling the chemical composition of the powders. Table 2.5 shows that the contamination by oxygen becomes more and more important when the N content in the powder increases, because NH bonds are easily hydrolysed.

2.3.4

First characterisations

Specific surface area The specific surface area is always high, about 100 m2 /g and taking into account the mean value of the density i.e. 2.00 (pycnometry measurement), the corresponding mean value for the diameter is 30 nm, in agreement with TEM observations.

©2003 Taylor & Francis

Si9N9Si N9H

Si9C

N9H

O9H C9H (CH3) HMDS45

Si9CH3 KBr

Si9N

Si9H

Absorbance (a.u.)

HMDS44

HMDS43 HMDS42

HMDS41

HMDS40

3500 3000 2500 2000 1500 1000 500 Wavenumber (cm–1)

Figure 2.8 IR spectra of the as-formed HMDS powders. (Reprinted from Musset 1995.)

Infrared spectroscopy IR spectra of the samples HMDS40–45 are shown in Figure 2.8. They present a broad absorption band between 750 and 1250 cm−1 with a maximum at 920–940 cm−1 due to a Si/C/N amorphous phase, corresponding to a flat XRD diagram. A weak band at 550 cm−1 is due to the Si N bond. Other bands are related to hydrogenous species: N H at 1180 cm−1 , Si H at 2050 cm−1 and Si CH3 at 1250 cm−1 , indicating an incomplete pyrolysis of the precursor or absorption of volatile species obtained from the pyrolysis. The presence of hydrogen is in good agreement with the chemical analysis, which is often less than 100% (Table 2.4). The results presented in this section show that it is possible to produce Si/C/N nanopowders with controlled chemical composition from a liquid precursor. The powders are obtained in significant quantities for materials applications. The main differences compared to powders obtained from the gas phase are the amorphous character and high hydrogen content of the powders obtained from the liquid phase.

2.4 Silicon carbonitride (Si/C/N) powders containing in situ the elements of the sintering aids (Al, Y): pre-mixed powders 2.4.1

Introduction: elaboration of SiC–Si3 N4 nanocomposite materials

Materials in the SiC/Si3 N4 system are elaborated by heat treatment of powders in the presence of liquid-forming sintering aids such as Y2 O3 , MgO, Al2 O3 , etc. singly or in combination. Such processes need a previous mixing step of all powders having different natures and sometimes different sizes. Recently, some studies report the incorporation of metal elements

©2003 Taylor & Francis

(Al, Y, etc.), necessary for the sintering step, into silicon nitride or carbonitride systems, most of them being focused on chemical processing routes using organometallic precursors (Soraru et al. 1992; Iwamoto et al. 1998; Koyama et al. 1998). A few other publications concern physico-chemical processes: laser irradiation of gaseous mixtures (SiH4 , NH3 , (CH3 )3 Al) allows the formation of nanosized powders with Al in a Si3 N4 matrix (Borsella et al. 1993b). Aerosol of Si/C/N particles with yttrium and aluminium compounds were prepared under dual irradiation of CO2 and excimer laser beams of a SiH4 , C2 H4 and NH3 mixture, aluminium being incorporated in the form of alumina powder in the reactant flow and yttrium being introduced as a low volatile organometallic compound in the reaction zone (Yamada et al. 1997). 2.4.2

Synthesis of pre-mixed powders

In order to prepare Si/C/N powders containing uniformly dispersed sintering aids investigations were undertaken to set up a laser spray synthesis process of Si/C/N/O/Al(Y) nanopowders using solutions of a liquid organosilicon precursor – HMDS – with solid metallic alcoxides, namely aluminium and yttrium isopropoxides [(CH3 )2 CHO]3 Al and [(CH3 )2 CHO]3 Y. Table 2.6 summarises the synthesis conditions of 14 runs which can be divided into 3 groups: • • •

Group I includes experiments carried out with solutions having different Al (+Y)/Si atomic ratio. Group II includes experiments carried out with solutions having the same Al (+Y)/Si atomic ratio. All runs are realised with a binary mixture of carrier gases (Ar + NH3 ). The third is composed of experiments made with the same solutions as in group II, but the carrier gas is composed of Ar + NH3 + H2 in order to see if the chemical composition of the powders is modified by the presence of H2 .

Some changes occur compared to the experiments described above in Section 2.3. The initial laser power is in the range 300–330 W, but, here, with an unfocused laser beam (diameter = 13 mm) which covers the aerosol flow sufficiently. For comparison, an experiment with pure HMDS carried out under similar conditions is also reported (HMDS66). 2.4.3

Effect of experimental parameters

Effect of the solutions of precursors In Table 2.6, most of the experiments were carried out with metallic isopropoxides in HMDS but, in two cases (HSAl04 and HSAlY05), isopropanol was added to the reactive solution in order to increase the solubility of the alcoxides and to decrease the viscosity of the solutions. In these experiments, more liquid is displaced than in other experiments but the production rate remains at the same level (≈30 g/h). Also, the amount of oxygen incorporated in the as-formed powders is very high (>10%). In conclusion, isopropanol does not improve the production or quality of the powders and this solution was not developed. Precursor solutions are characterised by their atomic ratios: [Al(+Y)]/Si. For the first experiments, the Al/Si ratio in the precursor solution was adjusted to 0.069 (e.g. HSAl03); this value corresponds to 6 wt% Al2 O3 content in a final SiC/Si3 N4 material according to the different oxide contents reported in the literature from about 10 (Kennedy et al. 1996) to 1 wt%

©2003 Taylor & Francis

Table 2.6 Synthesis conditions and results of pre-mixed Si/C/N powders Run

HSAl03 HSAl04 HSAl07 HSAl08 HSAlY05 HSAl09 HSAl12 HSAl13 HSAl11 HSAl14 HSAl15 HSAlY16 HSAlY17 HSAlY18

Carrier gas flow rates Displaced liquid (g/h) (cm3 /min) Ar

NH 3

H2

1900 1900 1900 0 1900 1140 1520 1710 475 855 665 1140 1520 1710

0 0 0 0 0 760 380 190 570 190 380 760 380 190

0 0 0 1900 0 0 0 0 855 855 855 0 0 0

©2003 Taylor & Francis

87 119 88 49 115 71 87 87 64 74 75 81 89 79

Powder production rate (g/h)

Chemical analysis (wt%) SBET (m2 /g)

34 32 24 11 30 23 28 30 27 26 22 26 25 28

66 81 110 280 79 112 96 83 134 89 99 100 91 77

Atomic ratio

Si

C

N

O

Al

46.5 43.5 47.1 48.7 43.7 49.5 47.2 49.8 49.8 48.8 50.9 53.0 50.7 49.3

33.9 33.7 29.1 26.6 33.4 6.8 14.8 23.0 5.8 17.3 11.7 7.5 16.9 21.9

12.9 10.7 12.8 14.1 11.2 36.0 29.0 19.2 37.2 24.5 26.8 30.3 25.4 22.4

5.7 13.2 8.6 8.9 11.5 6.1 7.5 5.9 5.7 7.7 8.8 8.2 5.4 5.1

2.5 2.4 2.5 1.6 1.7 1.5 1.5 2.0 1.5 1.6 1.8 0.4 0.9 0.5

Y 0 0 0 0 1.4 0 0 0 0 0 0 0.4 0.7 0.7

Al + Y /Si

Al/Y

0.05 0.03 0.03 0.04

0.01 0.02 0.01

3.0 4.2 2.5

alumina and yttria (Burger et al. 1997). The first result is that it is possible to incorporate Al and Y additives in the powders, but Table 2.6 clearly shows that the ratio Al/Si, in this range of experimental conditions, does not reach the 0.069 ratio. This is due to the preferential pulverisation of the most volatile compound of the mixture, that is, HMDS. So, in the following experiments, the initial Al/Si and (Al + Y)/Si ratios were decreased to 0.04 (HSAl09, 12, 13) and 0.02 respectively (HSAlY16, 17, 18). The similarity of the Al/Si atomic ratio in the starting solution (0.04) and in the as-formed HSAl powders (0.03–0.04) suggests that the nebulisation of the mixture of precursors is homogeneous. In the case of HSAlY powders, the Al/Y atomic ratios are different in as-formed powders (2.5–4.2) and in the starting solution (1.1), though (Al + Y)/Si atomic ratios remain similar (0.01–0.02) to the starting solution. This is due to the incomplete dissolution of yttrium isopropoxide in HMDS and probably to an inhomogeneous nebulisation of the mixture of precursors. Effect of carrier gas The nature and composition of the carrier gas mixture play a significant role in the chemical composition of powders. Whatever the powder (HSAl or HSAlY), an increase of ammonia relative flow rate induces a decrease in C content and C/N atomic ratio as obtained for powders without sintering aids (Musset 1995). Table 2.6 also shows that when hydrogen is added partially in the place of argon in argon–ammonia mixtures, the C content decreases. This seems to indicate that more volatile hydrocarbon species are formed during the synthesis. For example, the C content varies from 14.8 to 11.7 wt% in a 80% argon–20% ammonia mixture (HSAl12 and 15 samples) and from 23.0 to 17.3 wt% in a 90% argon–10% ammonia mixture (HSAl13 and 14 samples). It must be noticed that a large quantity of free C (10–14%) remains in powders when the ammonia relative flow rate increases from 0 to 20. 2.4.4

First characterisations

Specific surface area Powder production rates in the 22–34 g/h range and specific surface area values in 66–134 m2 /g range are very similar to the values obtained for powders without metallic additives. Infrared spectroscopy and X-ray diffraction As for powders without metallic additives, XRD patterns show a large background without any distinct peak indicating an amorphous structure. This is confirmed by infrared spectra showing a large band around 900 cm−1 . This one is essentially composed of Si C bonds when powders are synthesised with argon-rich carrier gas while Si N bonds in an amorphous environment (large band at 400 cm−1 ) appear with ammonia-rich carrier gas. Moreover, Si CH3 bonds are present (1250 cm−1 ) in all powders.

2.5 Conclusion In this chapter, it has been shown that laser pyrolysis is a versatile method well suited for the production of significant amounts of Si-based nanopowders. Due to their properties (purity

©2003 Taylor & Francis

limited only by the purity of the reactants, low size dispersion, etc.), the nanopowders are good candidates for applications in ceramics. Nanopowders can be obtained from gaseous or liquid precursors. The powders obtained from the gas phase are crystallised while the powders obtained from the liquid phase are amorphous. In both cases, the chemical composition of the powders is controlled by the chemical composition of the reactive mixture. Finally, an interesting result is the incorporation of the elements of sintering aids (Al, Y, O) during the synthesis.

References Alexandrescu, R., Morjan, I., Borsella, E., Botti, S., Fantoni, R., Dikonimos-Makris, T., Giorgi, R. and Enzo, S. (1991) Journal of Materials Research, 6, 2442. Alexandrescu, R., Borsella, E., Botti, S., Cesile, M. C., Martelli, S., Giorgi, R., Turtu, S. and Zappa, G. (1997) Journal of Materials Science, 32, 5629. Armand, X., Herlin, N., Martinengo, H., Musset, E. and Cauchetier, M. (1995) Fourth Euro-Ceramics, vol. 1 (Faenza Editrice S.p.A., Faenza), pp. 37–44. Baraton, M. I., Boulanger, L., Cauchetier, M., Lorenzelli, V., Luce, M., Merle, T., Quintard, P. and Zhou, Y. H. (1994) Journal of the European Ceramic Society, 13, 371. Bauer, R. A., Smulders, R., Becht, J. M. B., Van der Put, P. J. and Schoonman, J. (1989) Journal of the American Ceramic Society, 72, 1301. Borsella, E., Botti, S., Fantoni, R., Alexandrescu, R., Morjan, I., Popescu, C., Dikonimos-Makris, T., Giorgi, R. and Enzo, S. (1992) Journal of Materials Research, 7, 2257. Borsella, E., Botti, S., Giorgi, R., Martelli, S., Turtu, S. and Zappa, G. (1993a) Applied Physics Letters, 63, 1345. Borsella, E., Botti, S., Alexandrescu, R., Morjan, I., Dikonimos-Makris, T., Giorgi, R. and Martelli, S. (1993b) Materials Science and Engineering A, 168, 177. Bourgeois, L., Barbier, G., Viguié, J. C., Herlin, N. and Cauchetier, M. (1995) Fourth Euro-Ceramics, vol. 1 (Faenza Editrice S.p.A., Faenza), pp. 225–232. Buerki, P. R., Troxler, T. and Leutwyler, S. (1990) High Temperature Science, 27, 323. Burger, P., Duclos, R. and Crampon, J. (1997) Materials Science and Engineering, A222, 175. Casey, J. D. and Haggerty, J. S. (1987a) Journal of Materials Science, 22, 737. Casey, J. D. and Haggerty, J. S. (1987b) Journal of Materials Science, 22, 4307. Cauchetier, M., Croix, O., Luce, M., Michon, M., Paris, J. and Tistchenko, S. (1987) Ceramics International, 13, 13. Cauchetier, M., Croix, O. and Luce, M. (1988) Advanced Ceramic Materials, 3, 548. Cauchetier, M., Croix, O., Robert, C., Lance, M. and Luce, M. (1989) Euro-Ceramics, vol. 1, Elsevier Applied Science, London, pp. 130–134. Cauchetier, M., Croix, O., Luce, M., Baraton, M. I., Merle, T. and Quintard, P. (1991) Journal of the European Ceramic Society, 8, 215. Cauchetier, M., Croix, O., Herlin, N. and Luce, M. (1994) Journal of the American Ceramic Society, 77, 993. Cauchetier, M., Armand, X., Herlin, N., Mayne, M., Fusil, S. and Lefevre, E. (1999) Journal of Materials Science, 34, 1. Croix, O., Gounot, M., Bergez, P., Luce, M. and Cauchetier, M. (1991) Ceramics Today – Tomorrow’s Ceramics, Elsevier Science Publishers, Amsterdam, pp. 1447–1455. Curcio, F., Ghiglione, G., Musci, M. and Nanetti, C. (1989) Applied Surface Science, 36, 52. Curcio, F., Musci, M., Notaro, M. and Quattroni, G. (1991) Ceramics Today – Tomorrow’s Ceramics, Elsevier Science Publishers, Amsterdam, pp. 2569–2578. Fantoni, R., Borsella, E., Piccirillo, S., Ceccato, R. and Enzo, S. (1990) Journal of Materials Research, 5, 143. Förster, J., von Hoesslin, M., Schäfer, J. H., Uhlenbussch, J. and Viöl, W. (1991) Proceedings of the 10th International Symposium on Plasma Chemistry (FRG) vol. 1 (eds U. Ehlemann, H. G. Lergon and K. Wiesemann), pp. 1–6.

©2003 Taylor & Francis

Fusil, S., Armand, X., Herlin, N. and Cauchetier, M. (1997) Key Engineering Materials, 132–136, 141. Garifo, L. (1995) Laser Focus World, 30. Gonsalves, K. E., Strutt, P. R., Xiao, T. D. and Klemens, P. G. (1992) Journal of Material Science, 27, 3231. Haggerty, J. S. and Cannon, R. W. (1981) Laser-induced Chemical Processing, Plenum Press, New York, pp. 165–241. Herlin, N., Musset, E., Luce, M. and Cauchetier, M. (1994) Journal of the European Ceramic Society, 13, 285. Herlin, N., Armand, X., Musset, E., Martinengo, H., Luce, M. and Cauchetier, M. (1996) Journal of the European Ceramic Society, 16, 1063. Iwamoto, Y., Kikuta, K. and Hirano, S. (1998) Journal of Materials Research, 13, 353. Kennedy, T., O’Neil, J. P., Hampshire, S., Poorteman, M. and Cambier, F. (1996) Silicates Industriels, 9–10, 201. Kizaki, Y., Kandori, T. and Fujitani, Y. (1985) Japan Journal of Applied Physics, 24, 800. Knudsen, A. K. (1987a) Advances in Ceramics, 21, 237. Knudsen, A. K. (1987b) US Patent 4, 689, 129. Knudsen, A. K. (1997) Carbide, Nitride and Boride Materials: Synthesis and Processing, Chapman & Hall, London, pp. 343–358. Koyama, S., Nakashima, H., Sugahara, Y. and Kuroda, K. (1998) Chemistry Letters, 191. Li, Y., Liang, Y., Zheng, F. and Hu, Z. (1994a) Journal of the American Ceramic Society, 77, 1662. Li, Y., Liang, Y., Zheng, F. and Hu, Z. (1994b) Materials Science and Engineering, A174, L23. Luce, M., Croix, O., Zhou, Y. H., Cauchetier, M., Sapin, M. and Boulanger L. (1993) Euro-ceramics II, vol. 1, Deutsche Keramische Gesellschaft, Köln, pp. 233–238. Luce, M., Herlin, N., Musset, E. and Cauchetier, M. (1994) Nanostructured Materials, 4, 403. Martinengo, H., Musset, E., Herlin, N., Armand, X., Luce, M. and Cauchetier, M. (1996) Silicates Industriels, 61, 9. Mayne, Martine (May 1997) PhD Thesis, Limoges University, 97LIMO0021; http://thesenet.abes.fr Mayne, M., Armand, X., Cauchetier, M., Doucey, B., Bahloul-Hourlier, D. and Goursat, P. (1999) Ceramics: Getting in the 2000’s, Part B, Techna, Faenza, pp. 211–218. Musset, Emmanuel (Nov. 1995) PhD Thesis, Paris-Sud Orsay University, 95PA112537; http://thesenet.abes.fr Rice, R. W. (1986) Journal of the American Ceramic Society, 69, C183. Rice, R. W. (1987) Journal of the American Ceramic Society, 70, C117. Rice, R. W. and Woodin, R. L. (1988) Journal of the American Ceramic Society, 71, C181. Schoenung, J. M. (1991) American Ceramic Society Bulletin, 70, 112. Soraru, G. D., Ravagni, A., Dal Maschio, R. and Carturan, G. (1992) Journal of Materials Research, 7, 1266. Spitz, J. and Viguié, J. C. (1970) French Patent 2, 110, 622. Suzuki, M., Nakata, Y., Okutani, T. and Kato, A. (1992) Journal of Materials Science, 27, 6091. Suzuki, M., Maniette, Y., Nakata, Y. and Okutani, T. (1993) Journal of the American Ceramic Society, 76, 1195. Sumaya, Y., Marra, R. A., Haggerty, J. S. and Bowen, H. K. (1985) American Ceramic Society Bulletin, 1356. Symons, W. and Danforth, S. C. (1987) Advances in Ceramics, 21, 249. Tougne, P., Hommel, H., Legrand, A. P., Herlin, N., Luce, M. and Cauchetier, M. (1993) Diamond and Related Materials, 2, 486. Wakai, F., Kodama, Y., Sakaguchi, S., Murayama, N., Isaki, K. and Niihara, K. (1990) Nature, 344, 421. Willaime, F., Boulanger, L. and Cauchetier, M. (1995) Materials Research Symposium Proceedings, 359, 53. Yamada, T., Tanaka, Y., Suemasu, T. and Kohtoku, Y. (1997) Nuclear Instruments and Methods in Physics Research, B121, 378.

©2003 Taylor & Francis

3

Thermal behaviour of as-formed silicon-based nanopowders

General introduction This chapter presents the effects of thermal treatments on Si-based nanopowders obtained as described in the previous chapter from the gas phase or from the liquid phase. It is mainly focused on the behaviour of Si/C/N and Si/C/N/Al/Y/O powders (0.2 < C/N(at) < 2.6) because such nanopowders may be used to obtain Si3 N4 /SiC materials after thermal treatment. This chapter is divided into two parts. The first part is devoted to thermal treatment in a high-temperature graphite furnace in order to know the evolution of the physico-chemical characteristics. The second part reports the study of the thermal behaviour using TGA/MS in order to determine the chemical reactions and mechanisms involved during annealing.

©2003 Taylor & Francis

3.1 Thermal behaviour in a high-temperature graphite furnace Emmanuel Musset, Martine Mayne, Michel Cauchetier, Xavier Armand, Michel Luce and Nathalie Herlin-Boime

3.1.1 Introduction The aim of the first part of this study is to determine the influence of the annealing treatments performed in a high-temperature graphite furnace both on the weight of samples and on different physico-chemical properties of nanopowders such as the chemical composition, structure, reactivity and size. It is mainly focused on Si/C/N samples obtained from the liquid phase using the aerosol method. The results of annealing treatments and the consequences of these treatments on the chemical composition, size, organisation, etc. of silicon-based powders are presented with the intent of showing the similarities and the differences between the different powders, to correlate these properties to the synthesis conditions, etc., and to begin to compare the properties of these powders with the necessary properties for ceramic applications.

3.1.2 Weight evolution of SiC, Si/C/N and Si/C/N/Al/Y/O nanopowders The as-formed powders are annealed at atmospheric pressure in a high-temperature graphite furnace under argon or nitrogen (N) atmospheres. The temperature generally varies in the range 1000–1700◦ C, and the dwell time is 1 or 4 h. 3.1.2.1

Weight evolution under inert atmosphere

Thermal treatments were performed on SiC or Si/C/N nanopowders in an argon atmosphere for temperatures up to 1600◦ C. SiC powders exhibit a weight loss whatever the temperature. The maximum weight loss, observed near 1500◦ C, remains low (usually 0.75) (Musset 1995; Musset et al. 1997). Effectively, extended X-ray absorption fine structure (EXAFS) (Chapter 5.4) and nuclear magnetic resonance (NMR) (Chapter 5.1) analysis of C-rich powders synthesised from HMDS have shown that the atomic local arrangement of Si is essentially composed of mixed SiC3 N tetrahedra and of SiC4 tetrahedra which are already close to the SiC structure. This suggests the decomposition of slightly nitrogenous compounds (SiC3 N) during annealing in order to reach the SiC structure. Moreover, annealing induces a decrease in silica and in free C content because of the carbonitridation of this compound which involves CO evolution. For N-rich powders, EXAFS and NMR analysis have shown that the atomic local arrangement is essentially composed of SiN4 tetrahedra which are the local environment of crystalline Si3 N4 . Therefore, the evolution towards a major phase of Si3 N4 after annealing is not surprising. The drastic change between the evolution of Si/C/N towards SiC or Si3 N4 compounds is observed for the initial C/N atomic ratio in the as-formed powders near the 0.75 value and corresponding to a 3SiC–Si3 N4 mixture (Figure 3.1.1).

3.1.3.2

Si/C/N/Al/Y/O samples

For samples obtained from a mixture of organometallic compounds as precursors, the comments are mainly focused on the aspects specific to the presence of Al and/or Y in the powders. Table 3.1.2 presents the results in stoichiometric compounds after annealing at 1600◦ C for 1 h under a N2 atmosphere for Si/C/N/Al/Y/O samples with an initial C/N ratio in the same range as HMDS samples (0.2–2.6). The most important difference between Si/C/N samples obtained from only HMDS and Si/C/N/Al/Y samples obtained from a mixture of organometallic precursors is the incorporation of Al and Y metallic elements into the Si/C/N system, which could modify the evolution of the chemical composition of the Si/C/N system during annealing. That is why a comparison between Si/C/N and Si/C/N/Al/Y samples with a similar C/N ratio is presented below. Tables 2.6 and 3.1.3 clearly show that, in the Si/C/N/Al/(Y) systems, the Al+(Y)/Si ratios remain similar to those of the corresponding as-formed powders. This attests that the thermal treatment under a nitrogen atmosphere (1600◦ C) does not induce the degradation of all Al and Y containing species. Thus, Al and/or Y additives remain in the powder after annealing, which is important for the future sintering process, in which sintering aids (Y2 O3 , Al2 O3 ) are necessary in order to obtain dense nanomaterials. By comparing, after heat treatment, HSAl or HSAlY samples (Table 3.1.3) with HMDS samples (Table 3.1.2), one can see that the evolution of the chemical composition in Si/C/N/Al(Y) system as a function of the C/N ratio is comparable to that in HMDS samples previously discussed. In this way, C-rich Si/C/N/Al/Y samples (initial C/N ratio between 1.1 and 2.7) are converted essentially to SiC (content between 64% and 81%, see samples HSAl03, 04, 07 and HSAlY05) while N-rich samples (C/N ratio between 0.2 and 0.3) are

©2003 Taylor & Francis

100 45 80

44

Si3N4 SiC

60 Mass content (%)

43

SiO2 C

40 42 41

20

40

36

0

0.0

0.5

1.0

1.5

2.0

2.5

3.0

C/N atomic ratio (as-formed powder)

Figure 3.1.1 Change in phase content (SiC, Si3 N4 , SiO2 and C) in HMDS powders after 4 h annealing under nitrogen at 1600◦ C. (Reprinted from Musset 1995.)

converted essentially to Si3 N4 (content of 92%, see samples HSAl12 and HSAlY17). Samples exhibiting an intermediate composition (C/N ratio between 0.6 and 0.8) are converted to a mixture containing both SiC and Si3 N4 (SiC around 44 wt%, Si3 N4 between 52% and 53%, see HSAl12 and HSAlY17 samples). Note that C-rich samples also contain a noticeable quantity of free C (between 6% and 10%). Meanwhile, even if the evolution of the chemical composition is similar between Si/C/N and Si/C/N/Al(Y)O samples, noticeable differences appear where the C/N values and also the Si3 N4 and SiC contents are concerned, especially for Si/C/N and Si/C/N/Al/Y C-rich samples (C/N = 2.6). Figure 3.1.2 shows the evolution of the C/N ratio after a 1 h annealing treatment at 1600◦ C under a N2 atmosphere for some C-rich samples (HMDS66, HSAl03, 04 and HSAlY05). In this figure, the C/N ratio is around 15 for HMDS powders while it is only of 4.3–5.1 for HSAl(Y) powders. These results indicate that the loss of N is lower for powders containing Al and Y additives and exhibiting high C/N ratio. The SiC content is then lower and the Si3 N4 content is higher in these Si/C/N/Al/(Y)O samples than those in Si/C/N samples (see Tables 3.1.2 and 3.1.3). This suggests that the decomposition of N containing species (Si/C/N, Si3 N4 ) is limited during heat treatment of Si/C/N/Al/Y powders, probably due to the formation of a liquid phase in the

©2003 Taylor & Francis

C/N (at. ratio)

HMDS66 HSAl03 HSAl04 HSAlY05

10

0

0

1000 500 Temperature (°C)

1500

Figure 3.1.2 Evolution of the C/N atomic ratio with annealing temperature. (Reprinted from Cauchetier et al. 1999 with permission from Kluwer Academic/Plenum Publishers.)

quaternary Si–Al–O–N or in the quinary Y–Si–Al–O–N diagram preventing the degradation of nitrogenous compounds.

3.1.4 Structural changes XRD and infrared (IR) measurements were carried out in order to obtain information about the nature of the different phases present and the evolution of structure and of the different chemical bonding after heat treatment. They were carried out on both Si/C/N and Si/C/N/Al/Y samples obtained from the liquid phase. 3.1.4.1

Si/C/N samples

For annealing temperatures below 1500◦ C, the XRD diagrams of Si/C/N nanopowders (HMDS samples) are flat indicating that the powders are still amorphous. The IR spectra are very similar to the spectra of as-formed powders (see for example Figure 2.8), apart from the absorption bands at 1170, 2950 and 1255 cm−1 involving hydrogenated bonds that disappeared. At 1500◦ C, peaks are present on the XRD diagram (Figure 3.1.3), indicating the presence of crystallised β-SiC in HMDS40, 41 and 42 samples and α- and β-Si3 N4 phases in HMDS45. For HMDS43 and 44, the peaks are small and broad indicating a lower degree of crystallisation. Figure 3.1.4 presents IR spectra for three samples (C/N = 0.73, 0.67 and 0.57, respectively for HMDS42, 43 and 44) at the same temperature (1500◦ C). It shows two broad peaks centred near 900 and 500 cm−1 indicating the presence of an amorphous material and corresponding to Si C N and Si N bonds, respectively. These results, together with the XRD result, indicate that crystallisation is not complete at this temperature, especially for samples with an ‘intermediate’ C/N ratio (HMDS43 and 44). This late crystallisation will be discussed in more detail in Chapter 5.5. Between 1500◦ C and 1600◦ C, noticeable changes occur both in the XRD diagram and in the IR spectra (Figures 3.1.4 and 3.1.5). Figure 3.1.5 presents the XRD diagrams of samples annealed at 1600◦ C. C-rich powders (HMDS41 sample) lead preferentially to the β-SiC phase, as was already noticed at 1500◦ C. N-rich powders (HMDS45 sample) mainly show

©2003 Taylor & Francis

 -Si3N4

HMDS45 






-Si3N4








Al Support

 

Al
Al   




Al 




Al












HMDS44
-SiC
-SiC HMDS43
-SiC


-SiC


-SiC
-SiC

Al Al
-SiC

HMDS42

Al

HMDS41


-SiC


-SiC

Al


-SiC


-SiC HMDS40

20


-SiC


-SiC

30

40

50

60

70

80

Angle (2)

Figure 3.1.3 XRD patterns of HMDS powders after annealing at 1500◦ C. (Reprinted from Musset 1995.)

the presence of α- and β-Si3 N4 phases and very small β-SiC peaks. Samples exhibiting intermediate composition (HMDS43 and 44) lead to a mixture of SiC and Si3 N4 . In the IR spectra (Figure 3.1.4), the broad peak centred near 900 cm−1 becomes sharper between 1500◦ C and 1600◦ C. The shape becomes similar to the signature of crystallised SiC nanopowders obtained from the gas phase. The fine absorption bands present in the 800–400 cm−1 range indicate the crystallisation of N-rich powders only present in HMDS43 and 44 samples (α- and β-Si3 N4 phases, Luongo 1983). These results (XRD and IR) are in good agreement with the calculated chemical compositions given in Table 3.1.2.

©2003 Taylor & Francis

1700°C 1700°C 1700°C

1550°C

1500°C

Absorbance (a.u.)

1600°C Absorbance (a.u.)

Absorbance (a.u.)

1600°C

1550°C

1600°C 1550°C

1500°C 1500°C

1500

1000

500

Wavenumber (cm–1) HMDS42

1500

1000

500

Wavenumber (cm–1) HMDS43

1500

1000

500

Wavenumber (cm–1) HMDS44

Figure 3.1.4 Change in IR spectra during annealing under nitrogen between 1500◦ C and 1700◦ C. (Reprinted from Musset 1995.)

3.1.4.2

Si/C/N/Al/Y/O samples

As mentioned above for Si/C/N samples, annealing treatment under a nitrogen atmosphere also induces a drastic change in the structure of Si/C/N/Al/Y powders. From amorphous, in their as-formed state, they become crystallised or at least partially crystallised after a 1 h treatment at 1600◦ C under nitrogen atmosphere. Figure 3.1.6 shows XRD diagrams of Si/C/N samples (0.2 < C/N < 1.7) annealed under nitrogen at 1600◦ C for 1 h. The nature and the crystallographic variety of the phases that crystallise during heat treatment are very similar to those obtained from Si/C/N powders (HMDS samples). Thus, for a C-rich Si/C/N/Al powder (HSAl13), the XRD diagram shows an evolution towards α- and β-SiC compounds while for a N-rich Si/C/N/Al powder (HSAl09), α- and β-Si3 N4 are formed with a higher quantity of the α phase compared to the β phase. For samples having intermediate composition (HSAl12), αand β-Si3 N4 and β-SiC phases are formed. The IR spectra of the same samples (Figure 3.1.7) are in good agreement with the XRD diagram, that is to say that SiC bonds (around 900 cm−1 ) are detectable in a C-rich system, while Si N bonds in a Si3 N4 structure are detectable in a N-rich system. The sample with intermediate composition contains the Si N bonds in the Si3 N4 structure and also Si C bonds. IR and XRD results are very consistent with the chemical composition mentioned above. For C-rich samples (C/N = 2.6, HSAlY05 and HMDS66 samples), a comparison between the XRD diagrams of Si/C/N and Si/C/N/Al/Y powders annealed at 1600◦ C is presented (Figure 3.1.8). Moreover, the evolution of the IR spectra as a function of the temperature is presented for the same samples (Figure 3.1.9). XRD patterns of HSAlY05 and HMDS66 annealed at 1600◦ C are shown in Figure 3.1.8. The Si/C/N/Al/Y powder annealed for 1 h (HSAlY05 sample) presents broad diffraction

©2003 Taylor & Francis

 -Si3N4  





 



-SiC



Al








Al Support

Al







HMDS45



-Si3N4

Al
-SiC







 
-SiC

Al




-SiC 

HMDS44


-SiC 



-SiC







+


-SiC 








-SiC


-SiC  

 







HMDS43 




-SiC


-SiC




-SiC 

HMDS42

 -SiC


-SiC
-SiC


-SiC





-SiC

HMDS41
-SiC

-SiC

20

30


-SiC

Al
-SiC Al

40

Al

Al

50

60

70

80

Angle (2)

Figure 3.1.5 XRD patterns of HMDS powders after annealing at 1600◦ C. (Reprinted from Musset 1995.)

lines of β-SiC that are indicative of a low degree of crystallisation (Figure 3.1.8(a)). There is also evidence for the presence of the α-SiC phase (2H polytype). When annealing time is increased, well-crystallised α-SiC (2H polytype) and β-SiC are present (Figure 3.1.8(b)). In contrast, for the Si/C/N powder (HMDS66 sample) there is evidence of α- and β-SiC even after only 1 h of annealing and no difference can be noticed in the diffractogram for an annealing duration of 1 or 4 h (Figure 3.1.8(c) and (d)). These results suggest that for C-rich powders, the presence of Al and Y either limits the crystallisation or enables the formation of an amorphous phase initiated in the Y–Si–Al–O–N quinary system.

©2003 Taylor & Francis

Intensity (a.u.)

-Si3N4 >>
-Si3N4

HSAI09 ( +
) Si3N4 +
-SiC

HSAI12 -SiC +
-SiC HSAI13 10

20

30 40 50 60 70 Angle (2 degrees)

80

90

Figure 3.1.6 XRD analysis of annealed HSAl powder. (Reprinted from Mayne et al. 1999.)

Absorbance (a.u.)

Si-N in -Si3N4

HSAI09

KBr Si-N in -Si3N4 Si-C

HSAI12 HSAI13

Si-C

2000 1800 1600 1400 1200 1000 800 600 400 Wavenumber (cm–1)

Figure 3.1.7 IR spectra of annealed HSAl powder. (Reprinted from Mayne et al. 1999.)

Figure 3.1.9 presents examples of IR spectra of the as-formed and annealed Si/C/N and Si/C/N/Al/Y/O powders (high C/N ratio). Up to 1400◦ C, the spectra of Si/C/N/Al/Y/O powder are very similar to the spectra obtained for Si/C/N powder. The only difference seems to be a flat base line for the powders containing Al(+Y) additives whereas a pronounced dip appears near 1000 cm−1 in the IR spectrum of powders without additives

©2003 Taylor & Francis

Counts (a.u.) 20

(a) HSAIY05/1 h

(b) HSAIY05/4 h

(c) HSDS66/1 h

(d) HSDS66/4 h

40 60 2 (CuK)

80

40 60 2 (CuK)

20

80

Figure 3.1.8 XRD patterns of samples (a) and (b) HSAlY05 and (c) and (d) HMDS66 after 1 and 4 h annealing under N at 1600◦ C. (Reprinted from Cauchetier et al. 1999 with permission Kluwer Academic/Plenum Publishers.)

HMDS66 Si-N

HSAI03 Si-C

Si-N

Si-CH3

Si-N

1000°C/1 hr

1600°C/1 hr

As-formed Absorbance (a.u.)

Absorbance (a.u.)

As-formed

1400°C/1 hr

Si-C

Si-CH3

Si-CH3

As-formed Absorbance (a.u.)

HSAIY05 Si-C

1000°C/1 hr 1400°C/1 hr

1600°C/1 hr

1000°C/1 hr 1400°C/1 hr

1600°C/1 hr

1600°C/4 hr 1600°C/4 hr

1600°C/4 hr 1500

1000 Wavenumber (cm–1)

500

1500

1000 Wavenumber (cm–1)

500

1500

1000

500

Wavenumber (cm–1)

Figure 3.1.9 Changes in IR spectra with annealing temperature. (Reprinted from Cauchetier et al. 1999 with permission Kluwer Academic/Plenum Publishers.)

indicating crystallisation or grain growth as observed previously in the annealing of SiC powders obtained by a laser gas-phase driven reaction (Tougne et al. 1993) or from pyrolysis of a polycarbosilane (Sasaki et al. 1989). No significant change is observed when the dwell time increases from 1 to 4 h and no signature characteristic of Al or Y is observed. This study of chemical and structural changes during annealing of Si/C/N and Si/C/N/Al/Y/O samples under nitrogen atmosphere has shown a comparable evolution of Si/C/N and Si/C/N/Al/Y/O samples with comparable C/N ratio. A more detailed comparison seems to indicate that the presence of elements of sintering aids limits the degradation of the nitride phase during heat treatment of Si/C/N/Al/Y/O samples. Another point is that the elements Al, Y and O aids remain present in the powder after thermal treatment, which is encouraging for future elaboration of materials.

©2003 Taylor & Francis

3.1.5 Information about changes in the grain sizes BET measurements were carried out in order to measure the specific surface area, which enables determination of the changes in reactivity and grain size after heat treatment. These measurements should be related to transmission electron microscopy (TEM) analysis (see Chapter 4). 3.1.5.1

Si/C/N samples

Figure 3.1.10 shows the variation of only two Si/C/N samples (HMDS41 and HMDS44). Below 1000◦ C, the SBET decrease can be explained by the smoothing of the particles due to the reticulation reactions, which occur during the ceramisation of the powders. Then the SBET value remains constant up to 1400–1500◦ C and the decrease at 1600◦ C is due to the crystallisation and growth of the particles, as confirmed by TEM (Chapter 4). 3.1.5.2

Si/C/N/Al/Y/O samples

Annealing (nitrogen atmosphere, 1600◦ C, 1 h) of Si/C/N/Al/Y/O powders induces a decrease of the specific surface area compared with that of initial powders attesting to a structural and morphological change during heat treatment such as crystallisation, structural modification of free C and particle growth. The difference between the specific surface areas of as-formed and annealed powders increase when the C/N atomic ratio decreases. Thus for C-rich Si/C/N/Al/O powders (HSAl07 sample), the specific surface area varies from 110 m2 /g (as-formed) to 54 m2 /g (annealed) while for N-rich Si/C/N/Al/O powders (HSAl09 sample), the decrease of the specific surface area is more important (112–12 m2 /g). A similar evolution is also true for the series of Si/C/N/Al/Y/O samples. This phenomenon could suggest a different grain growth process in C-rich powders compared to N-rich powders, that is to say that the grain growth could be more pronounced in N-rich samples. Apart from a low grain growth,

BET surface area (m2 g–1)

100

HMDS41 HMDS44

80

60 40 20 0 0

200

400

600

800

1000

1200

1400

1600

Temperature (°C)

Figure 3.1.10 Change in specific surface area during annealing under N for a C-rich HMDS powder (HMDS41 sample) and a N-rich HMDS powder (HMDS44 sample). (Reprinted from Musset 1995.)

©2003 Taylor & Francis

this could also be related to the low crystallisation degree (see XRD patterns above) and to a structural modification of free C giving rise to some C ribbons as was already observed (Musset 1995) and as is usually formed from the heat treatment of amorphous C between 1500◦ C and 2000◦ C (Dresselhauss et al. 1988). These interpretations must be checked by TEM observations in order to determine the main phenomenon responsible for the surface area results.

3.1.6 Conclusion From the first section of this chapter devoted to thermal behaviour, several interesting points can be noted: • • •

The thermal behaviour of Si/C/N powders can be related to synthesis conditions; powders obtained from the gas phase appear more stable than powders obtained from the liquid phase using the aerosol method. Heat treatment under nitrogen allows one to obtain crystallised powders with a chemical composition varying from a major phase SiC to a major phase Si3 N4 with intermediate mixtures of SiC + Si3 N4 . The Al, Y and O elements of sintering aids remain present in the Si/C/N/Al/Y/O powders after thermal treatment under N atmosphere. The behaviour of these powders is qualitatively comparable to the behaviour of Si/C/N powders with comparable C/N ratio. The presence of these in situ sintering aids can be very helpful to obtain homogeneous dense materials – their influence is presented in Chapter 7.

References Cauchetier, M., Croix, O., Herlin, N. and Luce, M. (1994) Journal of the American Ceramic Society, 77, 993. Cauchetier, M., Armand, X., Herlin, N., Mayne, M., Fusil, S. and Lefevre, E. (1999) Journal of Materials Science, 34, 1. Croix, O., Gounot, M., Bergez, P., Luce, M. and Cauchetier, M. (1991) Ceramics Today: Tomorrow’s Ceramics, Elsevier Science Publishers, Amsterdam, p. 1447. Dresselhaus, M. S., Dresselhaus, G., Sugihara, K., Spain, I. L., Golbert, H. A., Cardona, M. (1988) Graphite Fibers and Filaments, Springer Verlag. Ekström, T. and Persson, J. (1990) Journal of the American Ceramic Society, 73, 2834. Hoffmann, M. J. and Petzow, G. (1993) Materials Research Symposium Proceedings, 287, 3–14 Luongo, J. P. (1983) Journal of the Electrochemical Society, 130, 1560. Mayne, Martine (May 1997) Ph.D. Thesis, Limoges University, 97LIMO0021; http://thesenet.abes.fr Mayne, M., Balhoul-Hourlier, D., Doucey, B., Goursat, P., Cauchetier, M. and Herlin, N. (1998) J. Eur. Ceram. Soc., 18, 1187. Mayne, M., Armand, X., Cauchetier, M., Doucey, B., Bahloul-Hourlier, D. and Goursat, P. (1999) Ceramics: Getting in the 2000’s, Part B (Ed. P. Vincezini) Techna, Faenza, pp. 211–218 Musset, Emmanuel (Nov. 1995) Ph.D. Thesis, Paris-Sud Orsay University, 95PA112537; http://thesenet.abes.fr Musset, E., Herlin, N., Cauchetier, M. and Luce, M. (1997) Silicates Industriels, 62, 111. Sasaki, Y., Nishima, Y., Sato, M. and Okamura, K. (1989) Physical Review B: Condensed Materials, 40, 1762. Tougne, P., Hommel, H., Legrand, A. P., Herlin, N., Luce, M. and Cauchetier, M. (1993) Diamond and Related Materials, 2, 486.

©2003 Taylor & Francis

3.2 Thermal reactivity of silicon-based nanopowders Djamila Bahloul-Hourlier, Benoît Doucey, Jean-Louis Besson and Paul Goursat

3.2.1 Introduction Si/C/N nanopowders may be converted into Si3 N4 /SiC bulk materials by consolidation and hot pressing. Several studies have been devoted to the synthesis of Si/C/N nanopowders and their sintering. However, only a few studies on the thermal decomposition of these preceramic nanopowders have been reported (Giorgi et al. 1996; Suzuki et al. 1995; Borsella et al. 1997; Mayne et al. 1998; Pan et al. 1999), in contrast to the situation with organometallic polymer precursors of Si/C/N ceramics (Burns et al. 1987; Blum et al. 1989; Han et al. 1992; Corriu et al. 1992; Schmidt et al. 1992; Lavedrine et al. 1991; Bahloul et al. 1993, 1997). It is well known that ultrafine powders can undergo structural changes when they are heated at specific temperatures. This is due to their high surface energy and non-equilibrium phases. A good knowledge of adsorbed species on the surface, the nature of chemical groups present in nanopowders and their reactivity during annealing may be used to further improve the colloidal processing, shaping and forming techniques and also to optimise sintering conditions and thus mechanical properties. The aim of the present work is to further clarify the effects synthesis conditions and chemical composition of the preceramic Si/C/N nanopowders have on their thermal stability while under different atmospheres. The synthesis conditions and chemical composition of the studied powders as well as their corresponding designation are reported in Table 3.2.1. The studied samples have a C/N ratio varying from 0.2 to 2.6. Various experimental techniques: thermogravimetric analysis coupled with mass spectrometry (TGA/MS), elemental analysis and X-ray diffraction (XRD) were used to follow the evolution of annealing in preceramic nanopowders. Table 3.2.1 Nanopowders chemical composition Powder

Chemical analysis (wt%) Atomic Chemical composition (at %) %NH3 ratio (cm3 /min) Si C N O Al Al2 O3 SiO2 Si3 N4 SiC Si C N C/N

SiCN29 SiCN35 HMDS35 HMDS66 HMDS73 HSAl07 HSAl09

40 0 40 0 10 0 40

©2003 Taylor & Francis

58 67 51 51 52 47 49

6 13 6 32 13 29 7

35 1 — 18 2 — 38 5 — 15 2 — 25 10 — 13 9 2 36 4 2

0.2 0.8 0.2 2.4 0.6 2.6 0.2

— — — — — 4 4

1 4 9 5 19 12 9

87 45 77 37 63 32 76

8 45 — 39 7 33 —

— 6 — — — — —

4 — 6 19 11 19 7

— — 8 — — — 6

3.2.2 Results and discussions 3.2.2.1

TG/MS analysis

Figure 3.2.1 shows a comparison of the measured weight losses during the decomposition of the as-received Si/C/N preceramic nanopowders in flowing helium (He). At first sight, the results indicate a direct dependence of the ceramic yields on the synthesis conditions. Obviously, the overall weight loss is greater for the family of nanopowders produced from liquid precursor (HMDS) than for that obtained from gas reactant SiH4 (gas). The major weight loss variations basically evolve at high temperature (T > 1300◦ C), where extensive degradation of the preceramic nanopowders occurs (Figure 3.2.1). A combined TGA/MS study was then used to determine the identity of the evolved species at each step. On account of the large number of powders produced by particular synthesis conditions, it is not possible to illustrate here all the mass spectra. Only the data corresponding to the sample that exhibited the highest weight loss (HMDS35) will be given in detail. However, when differences exist, they will be compared with the other samples. The TG/MS analysis of HMDS35 under He is given in Figure 3.2.2. On heating from room temperature to 300◦ C, the weight loss of 1% corresponds to the simultaneous escape of water (M/z = 18, 17) and ammonia (M/z = 17, 16, 14). The intensity of water progressively diminishes with increasing temperature up to approximately 300◦ C, while ammonia seems to stabilise between 400◦ C and 600◦ C and then diminishes with further heating. The appearance of two shoulders at M/z = 17 is evidence for the occurrence of at least two different reactions.

Temperature (°C) 0

300

600

900

1200

1500

Dwell time 1h00 0

–10

–15

0

50

100

150

200

–20

Time (min) SiCN29,

SiCN35,

HMDS66,

HMDS35,

HSA109

Figure 3.2.1 TG profiles of different nanopowders pyrolysed under He flow.

©2003 Taylor & Francis

∆M/M O (%)

–5

Temperature (°C) 0

300

600

900

1200

1500

Dwell time 1h00 0

–10

–15

0

50

M/z = 2, M/z = 18,

100 Time (min) M/z = 12, M/z = 27,

150

M/z = 14, M/z = 28,

M/z = 16, M/z = 41,

200

∆M/Mo (%)

Intensity (a.u.)

–5

–20

M/z = 17, ∆M/Mo (%)

Figure 3.2.2 TG–MS analyses of HMDS35 nanopowder pyrolysed under He flow.

The escape of H2 O and NH3 at low temperatures (less than 300◦ C) may be a consequence of the higher latent reactivity induced by the presence of the condensable Si NH2 and Si OH groups on the preceramic nanopowders (Equations 3.2.1, 3.2.2a,b). Si NH2 + HO Si

Si NH2 + H2 N Si Si NH2 +

→

Si NH Si

→

Si NH Si

+ H2 O↑

Si NH Si →

Si

N |

(3.2.1)

+ NH3 ↑

(3.2.2a)

Si

(3.2.2b)

+ NH3 ↑

Si While the nitrogen (N) bridging may take place at low temperatures, the reaction between silanol surface groups to form oxygen bridges is known to take place above 350◦ C (Equation 3.2.3). Si OH + HO Si

−H2 O

−−−→

(Si O Si)

(T > 300◦ C)

(3.2.3)

Nakamatsu et al. (1998) proposed the following reaction (Equation 3.2.4) to explain the simultaneous desorption of both H2 O and NH3 at low temperatures (around 200◦ C). SiONH4+ + SiOH → Si2 O + H2 O↑ + NH3 ↑

(3.2.4)

Besides NH3 and H2 O, some other fragment ions at M/z = 44, 30 are also detected below 400◦ C. The intensities of these peaks are very low and thus prevent a detailed characterisation of the origin structures. Between 300◦ C and 900◦ C, a second weight loss (3.8%) is attributed to the second escape of ammonia (M/z = 17, 16), which indicates the occurrence of the

©2003 Taylor & Francis

transamination/condensation reactions (Equation 3.2.5) which are known to occur during the thermal conversion of polysilazanes (Bahloul 1993). 2 ( Si NH Si ) +

Si NH Si

→ 2( Si

N |

Si ) + NH3 ↑

(3.2.5)

Si

Intensity (a.u.)

Concomitant with the evolution of NH3 is the loss of other gases with M/z varying from 42 to 12 atomic mass units (AMU) (Figure 3.2.3). It is noteworthy that compounds with molecular weights higher than M/z = 42 were not detected. Looking at these peak fragments, one would attempt to assign them to light hydrocarbons containing up to three carbon (C) atoms at most, like alcyne, olefin and alkane compounds. However, it should be remembered that the synthesis of this powder was made with NH3 as a reactant gas, being very reactive over Si H, Si C and C C bonds. Consequently, the probability of formation of hydrocarboned species having two or three carbons in their structure seems very weak (Van Dijen and Pluijmakers 1989). Moreover, it is seen from the comparison of the intensities of the obtained peaks that the most intense peak at M/z = 41 is followed, by descending order in intensity, by the peak at M/z = 40, then M/z = 39, and finally M/z = 42. No hydrocarbon combination with two or three carbons leads to such an evolution. It is, therefore, believed that these fragments may be assigned to R C N species such as CH3 CN (M/z = 41, 40, 39, 38), HCN (M/z = 26, 27) and also to less protonated species such as carbo-imines like CH3 CH NH (M/z = 42, 28). Further data that confirms the assignment of such species will be presented during the analysis of the exhaust gas at high temperatures (T > 900◦ C).

200 M/z = 27,

400 600 Temperature (°C)

M/z = 39,

M/z = 40,

M/z = 41,

800 M/z = 42

Figure 3.2.3 MS analysis of HMDS40 nanopowder pyrolysed under He flow.

©2003 Taylor & Francis

The maximum desorption of these species is situated at 600◦ C, whereas in the temperature range (600–900◦ C) the major volatile products detected are methane (M/z = 16, 15, 14) and H2 (M/z = 2). Methane continues to evolve up to 1000◦ C, whereas hydrogen evolution occurs in a wider temperature range up to 1500◦ C with a maximum loss at about 900◦ C. Some of these gaseous emissions are classical during the thermal conversion of polycarbosilane and polycarbosilazane polymers and were described by several authors whereas others are more specific to this type of laser-driven powders. Considering this temperature range, the reactions that describe the thermal decomposition of these materials are those that most probably involve radical mechanisms. There are several potential sources of radical species that can explain the above exhausted molecules (Equation 3.2.6–3.2.12). Si CN → N H→

Si• +•CN

•

(3.2.6)

•

N +H

Si CH2 Si

(3.2.7)

→ Si• +•CH2 Si •

•

Si CH2 CN → Si + CH2 C N Si CH2 CN →

SiCH2•

Si CH2 CH NH → Si CH3 →

Si• +•CH3

•

+C N •

•

Si + CH2 CH NH

(3.2.8) (3.2.9) (3.2.10) (3.2.11) (3.2.12)

For example, the formation of CH3 C N arises from radical cleavage of Si CH2 C N leading to the • CH2 C N radical, followed by H abstraction and/or radical combination reactions. The formation of the other species can also be explained using this approach. The C N groups do not appear in the initial structure of HMDS precursors, and seem to have been formed during the synthesis process. This is confirmed by Rice’s studies (1986) on the laser induced dissociation of the molecule of HMDS in the gas phase under argon. Analysis of gases of the processing chamber indicated the presence of, amongst other species, HCN and hydrocarbons, in particular. Thus, it is not surprising to think that, in our case, as the synthesis is being carried out in the presence of NH3 , the concentration of C N or amine groups giving HCN and RCN species, should increase at the expense of C C groups yielding to light hydrocarbon species. Indeed, the reaction of C or hydrocarbon compounds with ammonia has been well described elsewhere and the resulting product is HCN while no hydrocarbons are formed (Corriu et al. 1992; Bahloul et al. 1993). Above 900◦ C, a significant weight loss is observed. This is accompanied by the increased evolution of the shape of the TGA curve. Between 900◦ C and 1400◦ C, besides H2 , the major gaseous decomposition product is unambiguously ascribed to HCN at M/z = 26, 27. This exhaust gas gives credence to the description given above concerning the attribution to HCN and amine species of the signals at M/z between 42 and 12 AMU. This result is in good agreement with the chemical analysis of the as-received HMDS35 (Table 3.2.1) from which an excess of N and C are obtained assuming that only the equilibrium phases SiO2 , SiC, Si3 N4 and C, are present in the as-prepared powders (rule of mixture calculations). Furthermore, the short-range atomic structure description of the same Si C N nanopowders by XAS reveals the presence of C N bonds in the amorphous network (Ténégal et al. 1996). A structural model has been proposed (Ténégal et al. 1996), where mixed tetrahedral SiCx N4−x are randomly linked through C N bonds. The release of HCN at high temperature is further

©2003 Taylor & Francis

proof that CN groups are not only pendant groups X CN as observed for temperature T < 900◦ C, or radical scavenger/quencher groups under the experimental conditions employed in the synthesis process, but the branching groups to probably a mixed tetrahedra SiCx Ny . Homolytic cleavages of C N bonds between two mixed tetrahedra of silicon need more energy in comparison with those at chain-ends. Parallel to the evolution of HCN gas is the loss of N2 at M/z = 28, 14 up to the temperature limits of the experiment (1500◦ C) whereas the CO loss at M/z = 28, 12 begins at 1450◦ C in minor amounts (comparison of ions at M/z = 12 for CO and M/z = 14 for N2 ). At 1500◦ C, despite the important weight loss, the intensity of N2 diminishes rapidly, some traces of CO are still detected by MS during the isothermal holding. Interestingly, at 1500◦ C, during the holding time, we observe that the mass continues to decrease with increasing time, whereas any species cannot be detected by MS in large enough quantities to account for this continued weight loss. It is believed that this weight loss in the TGA curve is due to compounds such as SiO which condense rapidly in cold parts of the apparatus and may not be detected by MS. Evidence for this proposal comes from the deposition of SiO as a wool-like product on many parts of the apparatus, which consequently leads to the breaking of the thermocouple based on platinum. The most probable chemical reactions in the Si/C/N/O system, that involve the evolution of volatile species and which produce a significant weight loss, can be described in two steps. Primarily, decomposition of the quaternary SiCx Ny Oz phase leads to the formation of an intermediate ternary oxynitride solid phase Six Ny Oz and the simultaneous release of SiO and CO, N2 , in which N2 is the gas detected in the greatest quantity by MS. Then, in the second step, which occurs during holding at 1500◦ C, the oxynitride phase Six Ny Oz undergoes a decomposition in which gaseous N2 , Si(l, g), SiO and a solid phase are produced according to the following reactions: SiNx Oy Cz → Si3 N4 + SiO↑ + N2 ↑ + Si(1, v) ⇓

⇑

(3.2.13)

Six Ny Oz + N2 ↑ + CO↑ + SiO↑ The most known ternary oxynitride phase is Si2 N2 O. Although several investigators have reported on the thermal decomposition of Si2 N2 O, no mechanism has yet been clearly established. This is due to some of the following experimental difficulties: (a) the reproducibility of experiments, (b) the error in the determination of the weight loss, which is sometimes not representative of the real value of mass loss due to the condensation of the SiO species on the rod attaching the crucible to the balance, (c) progressive clogging of the aperture of the tube furnace and capillary for admission of gas to the mass spectrometer, by the deposit of SiO, and (d) the rupture of the thermocouple, stopping subsequent treatment. The majority of decomposition reactions of Si2 N2 O suppose the simultaneous escape of volatile species SiO(g) and N2 (g). Our analyses revealed that, after 10 min at 1500◦ C, no gas species are detected by MS, despite the continued weight loss as observed by TGA. It may be assumed that as the powder HMDS35 has been synthesised with ammonia as a reactant, a noticeable number of SiN4 sites are present and the amount of C in this powder is low. Taking into account our previous observations, it is thus hypothesised that the oxynitride SiNx Oy resulting from the decomposition of oxycarbonitride phase SiNx Oy Cz coexists with SiN4 sites initially present in the material. Hence when the duration of heating is increased, the oxynitride phase degrades to give SiO evolution in greater quantity, while the two other species

©2003 Taylor & Francis

HMDS35 He – 1500°C –1 h

Intensity (a.u.)

20%N2/He – 1500° C –1 h As-formed

HMDS66 He – 1500°C –1 h 20%N2/He – 1500°C–1 h As-formed 10

20

30

-Si3N4,

40 50 2 (degrees)


-Si3N4,

-SiC,

60

70

80


-SiC

Figure 3.2.4 XRD patterns of nanopowders pyrolysed under a reactive (20%N2 /He) and inert (He) atmosphere.

Si(l, g) and N2 interact on the SiN4 sites to give better organised Si3 N4 as supported by XRD illustrated in Figure 3.2.4. Rocabois et al. (1996) used Knudsen effusion MS to study the thermal degradation of several systems based on Si like binary, tertiary and even quaternary ones containing Si/C/N/O (Rocabois 1993). They stressed that the thermal degradation of Si-based materials is much less pronounced than what is predicted by thermodynamics and that their vaporisation is hindered and related to the different evaporation coefficients of each species. Moreover, the grain size, chemical composition, surface of contacts or interfaces during vaporisation would play an important role in the degradation of various phases. In our case, the nanopowder is initially amorphous as shown by XRD (Figure 3.2.4) and we can postulate that the number of surface contacts is high and the degradation of material is more homogeneous. By admitting, as suggested by Rocabois et al. (1996), that the evaporation coefficient of N2 is low in comparison with that of SiO, the two concomitant reactions, decomposition of oxynitride phase with release of SiO, Si(l, v), and N2 , and the reaction of the two latter gaseous species on SiN4 sites initially present leading to new Si3 N4 by epitaxial growth, is expected. 3.2.2.2

Major differences in thermal behaviour of the preceramic nanopowders

The results pointed out major differences between the nanopowders synthesised from the same HMDS precursor with NH3 (HMDS35 and HMDS73) or without (HMDS66). Elemental

©2003 Taylor & Francis

NH3

HMDS35

H2O

Intensity (a.u.)

HMDS73

HMDS66

100 200 300 400 500 600 700 800 Temperature (°C)

M/z = 17,

M/z = 18

Figure 3.2.5 MS analysis of HMDS nanopowders pyrolysed under He flow.

analysis of these powders confirmed that when NH3 is used during synthesis, the C/N ratio decreases. This diminution of C also depends on the concentration of NH3 in the reactant gas mixture. Moreover, during the heat treatment of these powders, the MS analysis also confirmed that the powders synthesised in the presence of NH3 contain more N X bonds (X = H, C, Si). Indeed, the evolution of NH3 at M/z = 17 (Figure 3.2.5) occurs over a wide temperature range and is characterised by two waves for HMDS35 and HMDS73, whereas for HMDS66, it exhibits only one narrow wave up to 300◦ C. In the case of HMDS66 between 400◦ C and 700◦ C, the MS spectra (Figure 3.2.6) reveal the loss of Si-containing species at M/z = 73, 59, 45, 44, 30, which suggests the formation of (CH3 )x≤4 Si, whereas for HMDS35 and HMDS73, species with M/z greater than 42 AMU were not detected. In view of the abundance of such organosilicon species, it seems that under the experimental conditions used during synthesis, the laser interaction with HMDS does not induce the total atomisation (bursting) of the molecular structure since NH3 is more reactive toward Si C, which implies the amination of Si and C sites. This is the first, chief origin of the formation of N-rich powders. Above 800◦ C, it can be seen that for preceramic powders prepared in the presence of NH3 during the laser synthesis process, HCN evolves in larger quantities compared to the powders prepared without NH3 .

©2003 Taylor & Francis

Intensity (a.u.)

HMDS35

HMDS66

200 400 600 800 1000 Temperature (°C) M/z = 26, M/z = 40, M/z = 43,

M/z = 27, M/z = 41, M/z = 45,

M/z = 39, M/z = 42, M/z = 73

Figure 3.2.6 MS analysis of HMDS nanopowders pyrolysed under a He flow.

It can also be observed, that as the temperature increases (T > 1200◦ C) the release of N2 and CO illustrated in Figure 3.2.6 for the C-rich powder HMDS66 and the N-rich powder HMDS35 is different. Indeed, the release of N2 and CO takes place during the entire thermal cycle for HMDS66, whereas no gas is detected after 10 min at 1500◦ C for HMDS35 Nrich powder. This suggests that the thermal degradation of these two nanopowders proceeds according to different reaction paths and depends on the chemical composition as well as on the microstructure of the materials. The CO response for C-rich powder HMDS66 was not monomodal indicating that the process involves more than one reaction. The oxynitride intermediate phase suggested for a N-rich powder in reaction (3.2.13) would not be formed, since an excess of C is present. Regarding previously discussed thermal decomposition of silicon carbonitride material containing free-carbon phase obtained from polysilazane precursors, degradation of C-rich nanopowders may be related to the following reaction (Bahloul 1993): SiNy Cx Oz + Cfree → SiC + CO↑ + N2 ↑ ⇓

⇑

(3.2.14)

Si(1, v) The loss of CO starts at about 1150◦ C, whereas the escape of N2 is observed from 1300◦ C. This can be explained in analogy to the reaction (Equation 3.2.15) observed in the solid phase

©2003 Taylor & Francis

by the carboreduction of Si O bonds by excess C. SiO2 + 3C → SiC + 2CO↑

(3.2.15)

In an analogous fashion, the large amount of N2 evolved during the holding time at 1500◦ C can be explained by the reaction of Si N and C: Si3 N4 + 3C → 3SiC + 2N2 ↑

(3.2.16)

Several authors (Rocabois et al. 1996; Lee and Jacobson 1992) have suggested that this reaction is inhibited when a SiC layer is formed on the surface of the C. This causes the system to shift to the equivalent SiC/Si3 N4 system and the excess loss of N2 would be suspected from the direct decomposition of Si3 N4 as shown in the following equation: Si3 N4 → 2N2 ↑ + 3Si

(3.2.17)

It was supposed that, as the experimental procedure of heat treatment is the same for all powders, we should expect that N-enriched powders generate more N2 than C-enriched powders, since they contain more Si3 N4 phase. This was not the case. Consequently, the decomposition of Si3 N4 is not able to offer an explanation for the evolution of N observed. Furthermore, powder samples after heat treatment showed no evidence for the presence of metallic silicon when examined by XRD (Figure 3.2.4). SiC is one of the main compounds in the samples obtained from C-rich powder HMDS66 heated at 1500◦ C under He. The low emission of CO during holding time can be attributed to the following reactions: SiO2 + SiC → SiO↑ + CO↑

SiO + C → 2SiC + CO↑

(3.2.18) (3.2.19)

We, therefore, suggest that in our system the SiC obtained from the reaction of Si3 N4 with C (3.2.16) may be consumed by SiO2 following reaction (3.2.18), which in turn releases C particles and consequently allows reaction (3.2.16) to continue producing gaseous N2 . 3.2.2.3

Thermal behaviour of nanopowders under a nitrogen atmosphere

The reason for studying the heat treatment under N was to answer the question: could the decomposition of silicon carbonitride, occurring under He, be stopped by using N environmental gas? The structural evolution of various nanopowders was thus followed as a function of heat treatment temperature and time whilst under a N atmosphere. The resulting curves in Figure 3.2.7, which represent the sole nanopowders that have the same ratio C/N = 0.2, show that below 1300◦ C the weight losses are nearly identical regardless of the nature of the atmosphere (nitrogen or He for the same kind of powder). The main gases detected by MS are attributed to the same species as those detected under He. In contrast, above 1400◦ C, comparison of the variation of mass is noticeably lower under N than He, as is shown for the HMDS35 sample. These differences are more pronounced for SiCN nanopowders obtained from only gaseous reactants (SiH4 + CH3 NH2 + NH3 + Ar), as shown in the data in Table 3.2.2. The volatile species which evolve are similar to those obtained under He with a majority of CO but of a lesser degree. Furthermore, it can be seen from Figure 3.2.8 that the variation

©2003 Taylor & Francis

Temperature (°C) 1050 1200 1350 1500 Dwell time 1h00 HMDS35

N2

HCN

Intensity (a.u.)

CO

N2

HMDS66 CO

100

120

140 160 Time (min) M/z = 17,

180

200

M/z = 18

Figure 3.2.7 MS analysis of HMDS nanopowders pyrolysed under He flow.

Table 3.2.2 Nanopowders weight loss changes according to atmosphere heat treatment Atmosphere heat treatment

Weight loss !M/Mo (%) SiCN35

SiCN29

HMDS66

HMDS35

HSAl07

HSAl09

He 20%N2 /He

−7.7 +6.1

−6.8 +1.3

−16.3 −6.2

−19.3 −15.5

−20.6 −14.6

−19.8 −13.1

of mass is remarkably decreased with increasing partial pressure of N. These results suggest that carbothermal nitridation reactions take place at high temperatures under N according to the following reactions (Mayne et al. 1998): SiNx Oy Cz + N2 → Si3 N4 + gaseous species (SiO↑ + N2 ↑ + Si(1, v))

©2003 Taylor & Francis

(3.2.20)

Temperature (°C) 0

300

600

900

1200

1500

Dwell time 1h00 0 –4

–12

∆M/Mo (%)

–8

–16 –20 0 SiCN29 (20%N2/He),

50

100 Time (min) HSA109 (20%N2/He),

150 HMDS35 (20%N2/He),

200 HMDS35 (He)

Figure 3.2.8 TG profiles of different nanopowders pyrolysed under a 20%N2 /He flow.

For C-enriched powders that contain more SiC phase, the nitriding process is also possible via the following reaction: 3SiC + 2N2 ↑ −→ Si3 N4 + 3C

(3.2.21)

The weight gain observed, particularly for SiCN synthesised from gaseous reactants, can be attributed to the nitridation reaction of excess metallic silicon detected by XRD (Mayne et al. 1998). According to the α- or β-Si3 N4 content calculated from XRD patterns of heat treated powders, the nitridation reaction leads to the preferential formation of the α phase. This could be explained by an oversaturation in nitrogen coming from the nitrogen atmosphere, which results in the crystallisation of the less ordered α phase. Meanwhile, the β silicon nitride phase that is more ordered and stable, is preferentially formed during annealing under the inert He atmosphere. 3.2.2.4

Influence of the nature of the precursors’ source of silicon on the chemical composition and thermal stability

It is observed that the aluminium doped powders (HSAl), synthesised under the same conditions as that of HMDS powders, except that only aluminium isopropoxide precursor is added, exhibit similar behaviour. Likewise, the differences observed between HMDS nanopowders and their conditions of synthesis, in particular the NH3 pressure in the reactive mixture of synthesis, seem to be reproduced for the HSAl powders. The role of aluminium isopropoxide introduced during the synthesis on the thermal behaviour is being studied. The use of a gas source of Si such as SiH4 has led to nanopowders of a more stable nature in terms of temperature, without doubt, due to their crystallisation state as revealed by XRD.

©2003 Taylor & Francis

The formation of CN bonds does not seem to be favoured when SiH4 is used. However, contrary to the HMDS and HSAl powders, the use of silane (SiH4 ) generates an excess of metallic silicon within the powders, which will be used profitably during sintering under N.

3.2.3 Conclusion These studies indicate that the relative atomic concentrations of Si, C, N and H are primarily influenced by the types of carrier gases used for the synthesis laser process. Residual functionalities C H, N H and Si H bonds will still be present as supported by the identification of volatile species by MS. The study of heat treatment of such systems based on Si indicates that thermal degradation is not an individual reaction but the net result of many reactions. The difference in thermal behaviour is attributed to variations in the preceramic structure. Taking into consideration the gases detected during our investigation and comparing them with similar results found by other research groups (9, 7), we are able to put forward the suggestion that by analysing preceramic nanopowders by thermal decomposition, it is possible to identify the processes by which they were synthesised. A potential advantage of laser-driven synthesis of refractory materials over gaseous reactants is the ability to include additives and co-reactants (e.g. sintering aids) (Cauchetier et al. 1999) in a truly homogeneous fashion.

References Bahloul, D., Pereira, M. and Goursat P. (1993) J. Am. Ceram. Soc., 76(5), 1156–68. Bahloul, D., Pereira, M. and Gerardin, C. (1997) J. Mater. Chem., 7(1),109. Blum, Y. D., Schwartz, K. B. and Laine, R. M. (1989) J. Mater. Sci., 24, 1707. Borsella, E., Botti, S., Martelli, S., Alexandrescu, R., Cesile, M. C., Nesterenko, A., Giorgi, R., Turtu, S. and Zappa, G. (1997) Silic. Indus., 1–2, 3. Burns, G. T., Angelotti, T. P., Hannemen, L. F., Chandra, G. and Moore, J. A. (1987) J. Mater. Sci., 27, 4651. Cauchetier, M., Armand, X., Herlin, N., Mayne, M., Fusil, S. and Lefevre, E. (1999) J. Mater. Sci., 34(21), 5257. Corriu, R. J. P., Leclercq, D., Mutin, P. H. and Vioux, A. (1992) Chem. Mater., 4, 711. Giorgi, R., Turtu, S., Zappa, G., Borsella, E., Botti, S., Cesile, M. C. and Martelli, S. (1996) Appl. Surf. Sci., 93, 101. Han, H. H., Lindquist, D. A., Haggerty, J. S. and Seyferth, D. (1992) Chem. Mater., 4, 705. Lavedrine, A., Bahloul, D., Goursat, P., Choong Kwet Yive, N. S., Corriu, R. J. P., Leclercq, D., Mutin, P. H. and Vioux, A. (1991) J. Eur. Ceram. Soc., 8, 221. Lee, K. N. and Jacobson, N. S. (1992) Ceram. Eng. Sci. Proc., 16th Ann. Conf. ou Comp. and Adv. Ceram. Mater., Cocoa-Beach. Mayne, M., Bahloul-Houlier, D., Doucey, B., Goursat, P., Cauchetier, M. and Herlin, N. (1998) J Eur Cer Soc, 18(9), 1187. Nakamatsu, T., Saito, N., Ishizaki, C. and Ishizaki, K. (1998) J. Eur. Ceram. Soc., 18, 1273. Pan, Z. W., Xie, S., Wang, G., Li, H. L. and Zhang, L. T. (1999) J. Mater. Sci., 34, 3047. Rocabois, P., Chatillon, C. and Bernard, C. (1996) J. Am. Ceram. Soc., 79(5), 1351. Rocabois, Philippe (1993) PhD, I.N.P.G., 93INPG; httl://thesenet.fr Rice, G. W. (1986) J. Am. Ceram. Soc., 69(8), C183. Schmidt, W. R., Marchetti, P. S., Interrante, L. V., Hurley, W. J., Lewis, R. H., Doremus R. H. and Maciel, G. E. (1992) Chem. Mater., 4, 937. Suzuki, M., Hasegawa, Y., Aizawa, M., Nakata, Y. and Okutani, T. (1995) J. Am. Ceram. Soc., 78(1), 83. Ténégal, F., Flank, A.-M. and Herlin, N. (1996) Phys. Rev. B, 54(17), 12029. Van Dijen, F. K. and Pluijmakers, J. (1989) J. Eur. Ceram. Soc., 5, 385.

©2003 Taylor & Francis

4

Structure of some nanopowders and micro/nano composites by transmission electron microscopy Marc Monthioux, Catherine Béraud and Nahida El Horr

4.1 As-prepared and heat-treated nanopowders Some of both the raw and heat-treated powders were investigated by transmission electron microscopy (TEM). Sample preparation prior to TEM investigations and TEM operating conditions are reported in Section 4.3, together with a brief reminder of the principles for the various TEM modes used in the study.

4.1.1

Synthesis from gaseous precursors

Starting from a mixture of silane (SiH4 ) and ethylene (C2 H2 ) as gaseous precursors, a serie of SiC-based powders was obtained. Likewise, still using silane as Si source but replacing ethylene by ammonia (NH3 ) and/or CH3 NH2 allowed Si/C/N-based powders to be obtained. Si/C-based powders Materials were investigated either as-prepared or after various treatments. Related preparation conditions and elemental analysis of the as-prepared materials are reported in Tables 2.2 and 2.3, respectively. Some of the samples reported in Table 2.2 were investigated after various treatments, including SiC173, SiC177 and SiC212. Various annealing in argon atmosphere were carried out, in order to check structural and textural changes in both the SiC and C phase. Also, SiC173 and SiC212/1200 were investigated after thermal treatment at 450◦ C performed in air (1–3 h isothermal), in order to eliminate the carbon (C) phase when present. Finally, further heat-treatment experiments in inert atmosphere were also performed at 1800◦ C on sample SiC173 oxidised. Special precautions were taken for the 1800◦ C heat-treatment experiments (Laanani 1997) to prevent any possible oxidation reaction due to oxygen traces in the gas phase, that is, a nondynamic, high purity argon atmosphere (N56 grade from Air Liquide) and an all-graphite furnace were used. SIC173, AS-FORMED

Daylight colour of the as-prepared powder is dark grey, somewhat related to the polyaromatic C content, as already expected from the elemental analysis (see Section 3.1.1) and as supported in the following by the direct evidence from TEM investigation. Grains exhibit round and isometric shapes (Figure 4.1).

©2003 Taylor & Francis

50 nm

Figure 4.1 TEM bright-field image from SiC173. Arrows show stripes on SiC crystals indicating polytypes as stacking faults.

C

SiC SiC 2 nm

Figure 4.2 TEM lattice fringe image from SiC173. Note how two adjacent SiC crystals are able to accommodate a slight mismatch in orientation of their respective lattices, making both the 111 lattice fringes connected to each other without interruption.

Grain sizes range according to a two-mode distribution. Large grains (10–60 nm) are prevalent, each being a single crystal, sometimes faulted (Figure 4.1, arrows). Grains are associated into aggregates, most often chain-like, due to both weak forces (electrostatic) or stronger ones, that is, structural. Figure 4.2 illustrates the latter, showing two grains whose lattices are connected despite a misorientation of 10◦ , obviously inducing a strong (covalent) bonding between grains. They are SiC crystals, β structure mainly, but some stacking faults

©2003 Taylor & Francis

(indicating the local occurrence of polytypes) are revealed in electron diffraction patterns as additional aligned spots (Figure 4.3), as a probable consequence of the rapid growth. Grains are coated with a thin (∼1 nm) layer of an amorphous phase, assumed to be a silicon oxide, as supported by X-photon energy dispersive spectroscopy (X-EDS) measurement, whose spectra show the presence of the element O, in addition to Si and C as expected (Figure 4.4).

4 nm–1

Figure 4.3 Electron diffraction of SiC crystals from SiC173. The structure is mainly β-SiC (see Figure 4.7 for diffraction ring identification). However, arrows show some aligned spots indicating stacking fault sequences. The occurrence of an amorphous coating at the grain surface (thickness ∼1 nm) is probable, though confusion with overfocusing Fresnel fringes is possible. Si

O C

I

0.50 11 CNT

1.00 1.50 0.00 keV

2.00 10 eV/ch

A EDAX

Figure 4.4 X-EDS spectrum of SiC crystals from SiC173. The presence of O attests the occurrence of an amorphous silica layer associated to the SiC grains.

©2003 Taylor & Francis

The minor occurrence of a silica phase is consistent with the results from elemental analysis (Table 2.3). On the other hand, small grains are a very minor phase, whose sizes (3–10 nm) are difficult to estimate from regular bright-field images. Dark-field images help in more easily appreciating the occurrence and size distribution of it, as compared to the major phase (Figure 4.5(c), arrows). They are not Si3 N4 but possibly SiC crystals too, since they appear in Figure 4.5(c) while not in Figures 4.5(a) or (b). However, other small, minor crystals are also found, despite that the amount of it is too small to appear in the diffraction pattern, from which highresolution images reveal fringes whose spacing is too large to match the SiC lattice distances (not illustrated). Rarity and concentration of this minor phase are such that identification was not possible. A third phase is found associated with the major (SiC) phase, more or less coating the grain surface, and identified as polyaromatic C. Its low proportion is such as it does not appear in the electron diffraction patterns when exposure times are appropriately chosen to reveal the SiC phase (Figure 4.3). However, its structure is most probably turbostratic (i.e. not having the three-dimensional structure of graphite), as assumed from its morphology and nanotextural characteristics. Lattice fringe imaging mode illustrates the latter (Figure 4.2). The C phase appears as thin, somewhat buckling and polygonal, ribbon-like graphene stacks. Using the parameters N (number of graphenes per stack) and L2 (length of the continuous, though distorted, d002 lattice fringe) to characterise the nanotexture of the C phase, as proposed by Oberlin (1989), the latter exhibits a rather low N values (N ∼ 2–5) with respect to

(a)

(b)

(c)

50 nm

Figure 4.5 TEM dark-field image sequence from SiC173. (a) and (b) correspond to ‘polyaromatic C-specific’ (DF1) dark-field imaging conditions for two selected orientations. The selected orientation direction is indicated by the ‘=’ sign, that is, only graphene stacks displayed parallel to the ‘=’ sign are imaged. Graphene stacks appear as small bright rims between SiC crystals. (c) corresponds to ‘SiC-specific’ (DF2) dark-field imaging conditions. Arrows show the occurrence of the minor phase (probably SiC) as small size grains among the larger SiC crystals.

©2003 Taylor & Francis

a rather high L2 value (up to 20–30 nm). As far as the existence of free edges of graphenes is unlikely, since inducing improbable dangling bonds, and since the so-called ribbons never appear twisted as genuine ribbons might be, the graphene stacks are believed to be walls of three-dimensional, somewhat polyhedral, polyaromatic C shells. These shells are likely to form under mechanisms related to the thermal and/or catalytic decomposition of a carbide phase for instance (Badami 1965; Oberlin and Rouchy 1971; Audier et al. 1981). Similar textures, whose nanotexture parameters are fairly consistent with those measured here, were also found in oxygen-free SiC-based materials obtained from heat-treated polycarbosilane (Delverdier et al. 1991; Laanani 1997). Low magnification, specific (i.e. using an objective aperture selecting some diffracted polyaromatic 002 beams) dark-field images (Figures 4.5(a) and (b)) illustrate the proportion and location of the C phase relative to the SiC grains, as the latter appear as round-shaped, grey to pale white objects while the former appears as bright, elongated, tiny spots (arrows) in which graphenes are oriented parallel to the elongation direction. Though not systematic, graphene stacks are often found oriented parallel to the grain surface. SIC177, 1200◦ C ANNEALED

The daylight colour of the starting powder is dark grey. The material is very similar to SiC173/as-prepared. Slight discrepancies concern the distribution of grain size (one-mode instead of two-mode), and the size range (slightly larger: 10–100 instead of 10–60 nm). Another difference lies in the probable absence of the non-identified, crystallised, minor phase of SiC173. Instead, another minor phase is found, mainly amorphous since specific darkfield images show that it does not contain any crystallised phases, neither Si3 N4 nor SiC (not illustrated). Considering the high diffusion contrast of this phase in dark-field images when the beam tilt angle corresponds to the Bragg angle for the 002 reflection of polyaromatic C, it is believed that this amorphous phase is enriched in poorly organised, turbostratic C, as expected from the transformation of a formerly amorphous C phase due to the annealing temperature (1200◦ C is actually far above the minimal carbonisation temperature – ∼500–600◦ C – i.e. the temperature sufficient to transform any kind of non-gaseous C precursor into polyaromatic solids). This phase is distinct from the better organised, polyaromatic, turbostratic C phase associated with the SiC phase (Figure 4.6). Figure 4.6 exhibits a C shell still partially filled with a SiC crystal (arrow), which strongly supports that the porous C phase originates from a carbide decomposition process. SIC212, 1200◦ C ANNEALED

The daylight colour of the starting powder is dark grey. The sample is mainly made of agglomerated SiC grains with round shapes (see Figure 4.1 for reference). Grain size range is 3–40 nm, that SiC-specific dark-field images show with a probable two-mode distribution similar to that of SiC177/as-formed (see Figure 4.5(c) for reference). The structure is strictly β since only diffraction rings corresponding to 111, 200, 220 and 311 reflections of β-SiC appear (Figure 4.7). SiC crystals are coated with amorphous silicon oxide, ∼1.5 nm thick (Figure 4.8). The coating may help the SiC crystals to link to each other (Figure 4.8). However, although confusion with an amorphous or poorly organised C coating was unlikely due to the annealing temperature, an oxidation experiment was performed to ascertain the oxide nature of the coating (see below). Unlike SiC173/as-formed or SiC177/1200◦ C, the relatively well-organised, turbostratic, polyaromatic C phase associated to the SiC phase (see Figures 4.2 or 4.6 for reference) is not

©2003 Taylor & Francis

C SiC

C

SiC

5 nm

Figure 4.6 TEM lattice fringe image from SiC177/1200. Arrow shows a small SiC crystal filling incompletely a polyaromatic C shell, supporting the hypothesis that the polyaromatic C phase between SiC grains originates from the early decomposition of some primary SiC crystals.

311

220

200

111

4 nm–1

Figure 4.7 Electron diffraction of SiC crystals from SiC212/1200. The structure is exclusively β-SiC.

found. Instead, a poorly organised, turbostratic, polyaromatic C phase is found, concentrated into a minor phase consequently exhibiting a nanoporous texture (Figure 4.9), assumed to be similar to that found in sample SiC177/1200. SIC177, 1500◦ C ANNEALED

The thermal treatment did not induce any change but the average SiC grain size, which has increased mainly due to a coalescence mechanism by impingement according to

©2003 Taylor & Francis

2 nm

Figure 4.8 TEM lattice fringe image of SiC crystals from SiC212/1200. The amorphous, about 2 nm wide silica coating is imaged, somewhat bonding crystals together.

5 nm

Figure 4.9 TEM bright-field image from SiC212/1200. The left part of the image illustrates the minor, poorly organised, polyaromatic C phase.

Le Coustumer et al. (1993), Monthioux and Delverdier (1996) and Laanani (1997). The smallest grains are now about 30–40 nm, many are in the range of ∼100 nm, and the largest diameters are difficult to measure by TEM due to the chances of overlapping for crystals whose large dimensions make them no longer electron transparent. In addition, an amorphous oxide layer about 2 nm thick at the grain surface is still found, quite similar to that of Figure 4.8. Nevertheless, the C phase associated to SiC crystals is also still found, though less abundant, with about the same textural and structural features than that found in Figures 4.2 or 4.6. Electron diffraction patterns show that β is the most frequent SiC structure, but the occurrence of

©2003 Taylor & Francis

diffraction spots located besides the regular 111, 200, 220 and 311 diffraction rings for the 3C polytype indicates that other polytypes (α structures) are present (Figure 4.10). SIC173, 1800◦ C ANNEALED

The main effect of such a high heat-treatment temperature is the development of the polyaromatic C phase, either as SiC-free porous material or shells around SiC crystals (Figure 4.11). Though still turbostratic as ascertained by electron diffraction patterns, since only 00l reflections and hk bands are present (Figure 4.12(a)), the nanotextural state of the C phase is very good, that is, exhibiting perfect graphene stacks at long range (Figure 4.13), which is consistent with a probable origin from the partial SiC decomposition (Oberlin et al. 1976; Iijima

4 nm–1

Figure 4.10 Electron diffraction of the SiC phase from SiC177/1500. The 3C (β structure) polytype is prevalent, but some other polytypes are found.

C

C SiC

10 nm

Figure 4.11 TEM bright-field image from SiC173/1800. The partial decomposition of SiC crystals induces the secondary formation of empty or crystal-filled C shells.

©2003 Taylor & Francis

(a)

(b) 11

10 002

4 nm–1

Figure 4.12 Electron diffraction patterns from SiC173/1800. (a) On the C phase: the structure is turbostratic, since only hk reflections are found in addition to the 002 reflection. (b) On SiC crystals (+C). The 3C polytype of SiC is no longer prevalent.

2 nm

Figure 4.13 TEM lattice fringe image from SiC173/1800. Walls of the C shells seen edge-on. The nanotexture (i.e. graphene perfection) is very good, though the structure is not graphitic but turbostratic according to electron diffraction patterns.

1982). Indeed, though the decomposition of SiC might sound doubtful since 1800◦ C is still early regarding the theoretical sublimation temperature of SiC (2700◦ C), the SiC crystals are the only C source available to explain the increase in the C phase amount. Beside, the average SiC grain size increased dramatically (more than two to four times larger than in the starting material), which is consistent with the similar effect observed in the SiC177/1500◦ C sample (but in a lesser extent). Meanwhile, the diffraction patterns now exhibit many diffraction spots other than that of the 3C polytype, indicating the occurrence of other polytypes (Figure 4.12(b) as compared to Figure 4.3).

©2003 Taylor & Francis

SIC173, 450◦ C OXIDISED

As a result of the oxidising treatment, the daylight colour of the powder has turned from dark grey to green-grey. The treatment was successful in removing the primary polyaromatic C phase without altering the SiC grains (Figure 4.14), as compared to the non-oxidised, as-prepared powder (Figure 4.15). Every other feature was kept unchanged. SIC173, 450◦ C OXIDISED, 1800◦ C ANNEALED

A secondary polyaromatic C phase has appeared again, abundant, as C shells systematically surrounding the SiC grains (see Figure 4.11 for reference). This experiment definitely ascertains that SiC crystals start to decompose at temperatures as low as 1800◦ C (but beyond

20 nm

Figure 4.14 TEM bright-field image from SiC173/450 O2 . The oxidation treatment was successful in removing the C phase without oxidising the SiC crystals.

20 nm

Figure 4.15 TEM bright-field image from SiC173. Arrows show the polyaromatic C phase associated to the SiC crystals. To be compared with Figure 4.16.

©2003 Taylor & Francis

1500◦ C, as shown by the 1500◦ C annealing experiment of sample SiC177, where only the primary C phase was found), although the average grain diameter increases. The fact that this temperature is far below the theoretical sublimation temperature for SiC (2700◦ C) may be explained by the small sizes of the SiC grains. Generally speaking, physical constants of a given phase are likely to be lowered when finely partitioned into nanocrystals as compared to macrocrystals, due to a higher contribution of surface atoms relative to core atoms (Buffat and Borel 1976). SIC212, 1200◦ C ANNEALED, 450◦ C OXIDISED

Like for sample SiC173, the daylight colour of the sample has turned from dark grey to green-grey. This is the only change observable. Specifically, the amorphous coating on the SiC grains is still found (not illustrated), ascertaining that the coating is silicon oxide, since amorphous C should have burnt into CO2 during the oxidation experiment. Si/C/N-based powders Two materials were investigated as-prepared, labelled SiCN29 and SiCN35. Related preparation conditions and elemental analysis of the as-prepared materials are reported in Tables 2.2 and 2.3, respectively. Though the amount of silane is similar, they mainly differ from each other both by the amount of nitrogen (N) supplied (twice for the former), and the N-containing gaseous precursors used (CH3 NH2 only for the latter, as compared to CH3 NH2 + NH3 for the former). This induces a higher supply in C for SiCN35. SICN29, AS-FORMED

Daylight colour of the material is light grey-green. As often, the sample exhibits two kinds of morphologies. The major phase is made of large grains, in the range of 20–150 nm, with frequent round shapes (Figure 4.16), sometimes polyhedral (Figure 4.17). According to the

50 nm

Figure 4.16 TEM bright-field image from SiCN29.

©2003 Taylor & Francis

20 nm

Figure 4.17 Low magnification TEM lattice fringe image of some crystals from SiCN29.

dark-field images, each grain is a single crystal (Figure 4.18(b)). Grains are associated to each other into aggregates or chains both by covalent bonding, since lattice fringe imaging suggests structural continuity between adjacent crystals through grain boundaries (Figure 4.19), or by weak forces. They are identified as α-Si3 N4 crystals by electron diffraction (Figure 4.20), as supported by the occurrence of the specific reflection 101 at 0.432 nm. However, the concomitant occurrence of the β structure is possible. Indeed, from the point of view of selected area electron diffraction mode, diffraction patterns from α- and β-Si3 N4 structures are difficult to distinguish from each other, due to the high amount of reflections many of which are present in both structures on one hand, and due to the frequent absence of either complete diffraction patterns or complete diffraction rings on the other hand. Although lattice fringe images reveal some distorted fringes at the Si3 N4 crystal surface looking like turbostratic C stacks (Figure 4.19), it is believed that, unlike in the SiC-based powders, they do not correspond to a polyaromatic C phase. Indeed, the X-EDS spectra (Figure 4.21) do not show any presence of C (either sp2 or sp3 ), neither associated to the Si3 N4 major phase nor to the SiOx Ny minor phase (see below). Also, considering the chemical balance for a SiC-based sample like SiC173 (Table 2.3) shows that 2 wt% only of calculated excess C are enough to provide obvious evidence of polyaromatic C stacks in the lattice fringe or dark-field images (Figures 4.2 or 4.5(a) and (b), respectively). On the contrary, twice a calculated amount of excess C in SiCN29 would barely correspond to the few distorted stacks (in lattice fringe mode, Figure 4.19) or bright rims (in dark-field mode, Figure 4.18(a), open arrow) shown in the TEM images. The absence of a polyaromatic C phase is therefore suggested, and the phase found at the crystal surface is more likely a silicon oxinitride phase. Actually, previous studies combining both TEM and XPS investigations and performed on various Si3 N4 powders have shown that silicon oxinitride is able to exhibit the same kind of distorted fringes as turbostratic C (Madigou et al. 1990a,b; Madigou 1991). However, since the chemical balance indicates the presence of C atoms, they could possibly be incorporated into the amorphous silicon oxinitride phase with such a low proportion that would make C nondetectable. Such amorphous silicon oxicarbonitride phases do actually exist, as synthesised from the thermal curing under gaseous oxygen or air of polycarbosilazane polymers (Okamura et al. 1984; Mocaer et al. 1993).

©2003 Taylor & Francis

(a)

(b)

50 nm

Figure 4.18 TEM dark-field image sequence from SiCN29. (a) ‘C-specific’ (DF1) imaging conditions. Actually, crystals appearing bright are not C but Si3 N4 crystals. They are imaged in the selected imaging conditions because the crystals exhibit lattice distances close to that of C002 . Likewise, the white arrow shows bright rims which are more likely an organised Si O N phase rather than a polyaromatic phase, according to X-EDS spectra (see text). The black arrow shows a polyaromatic C-rich phase. (b) ‘SiC-specific’ (DF2) imaging conditions. Crystals appearing bright are actually not SiC but Si3 N4 . They are imaged in the selected imaging conditions because the crystals exhibit lattice distances close to that of SiC111 . The fact that the crystals are thus able to be imaged in both dark-field imaging conditions allow them to be identified as Si3 N4 instead of C or SiC.

2 nm

Figure 4.19 TEM lattice fringe image on Si3 N4 crystals from SiCN29. Evidence of a coating layer, ∼1 nm thick, at the crystal surface, whose distorted fringes are oriented parallel to the grain surface.

©2003 Taylor & Francis

4 nm–1

Figure 4.20 Electron diffraction from SiCN29. Si3 N4 crystals typically exhibit a lot of lattice distances at small scattering angles.

Si N

O

I

0.50 17 CNT

1.00 1.50 0.00 keV

2.00 10 eV/ch

A EDAX

Figure 4.21 X-EDS spectrum on some Si3 N4 crystals from SiCN29. The occurrence of O and the absence of C suggests that the phase coating the crystals is actually Si O N-type.

The minor phase is made of grains whose size range is often lower than for the major phase, with round shapes. Since always faintly scattering the electron beam in dark-field imaging mode whatever the position of the objective aperture (Figures 4.18(a) and (b)) – though decreasingly as the tilt angle increases – the structure of this phase is amorphous. The chemical composition is qualitatively given by X-EDS measurements (not illustrated), which indicate an enrichment in oxygen and a depletion in N relative to the Si3 N4 phase (Figure 4.21), in addition to the presence of Si. It is therefore believed that this minor phase is amorphous silicon oxinitride or oxicarbonitride.

©2003 Taylor & Francis

SICN35, AS-FORMED

Daylight colour of the material is dark grey-green. As sample SiCN29, two morphologies are again observed. The major phase exhibits large grains with irregular and polyhedral shapes associated into compact aggregates (Figure 4.22). Size range cannot be ascertained due to the superimposition effects which make the grain contours difficult to determine. Dark-field imaging reveals that each grain is itself an aggregate containing crystals with various sizes up to about 40 nm (Figure 4.23(c)). These crystals were identified as pure β-SiC (=3C polytype only) by electron diffraction (Figure 4.24). Both dark-field (Figure 4.23(a)) and lattice fringe (Figure 4.25) images reveal that the SiC crystals are associated with an amorphous phase somewhat embedding them. According to the X-EDS analysis which shows the presence of N and O in addition to Si and C (not illustrated), this amorphous phase is probably a silicon oxinitride or oxicarbonitride. No turbostratic C phase was revealed. The minor phase exhibits much smaller grain sizes, in the range of 10–30 nm, with round shapes, associated into elongated chains (Figure 4.22, arrow). Dark-field imaging (Figure 4.23(b), arrow) supported by lattice fringe imaging (Figure 4.26) both reveal that each grain is itself an aggregate of tiny crystals, few nanometres large. They were identified as Si crystals (CFC structure) by means of electron diffraction (Figure 4.27). Correspondingly, lattice fringe images show lattice distances consistent with the (111) planes of Si (Figure 4.26). These images also reveal that Si crystals are associated with an amorphous phase, located at their outer surface as a coating, about 2 nm thick (Figure 4.26). According to the X-EDS analysis which barely shows another light element than O (in addition to the prevalent Si peak – not illustrated), this amorphous phase is probably a silicon oxide phase, which is likely considering the sensitivity of pure Si to oxidation.

50 nm

Figure 4.22 TEM bright-field image from SiCN35. Arrows show chain-like aggregate of small grains.

©2003 Taylor & Francis

(a)

(b)

(c)

50 nm

Figure 4.23 TEM dark-field image sequence from SiCN35. (a) ‘C-specific’ (DF1) imaging conditions. No crystals appear, indicating that no Si3 N4 crystals are present (see Figure 4.20 caption for further explanation). (b) Dark-field imaging conditions selected are intermediate between (a) and (c), therefore admitting both C002 and SiC111 distances. Both SiC crystals and another phase (arrowed) are imaged. (c) ‘SiC-specific’ (DF2) imaging conditions. The azimuthal angle is same as for (b). Correspondingly, the same SiC crystals as in (b) are imaged, but the crystals arrowed in (b) no longer appear, attesting that they are not SiC, but Si crystals (see Figures 4.26 and 4.27).

4 nm–1

Figure 4.24 Electron diffraction on a SiC-rich area from SiCN35. Crystals are pure β-SiC crystals.

©2003 Taylor & Francis

2 nm

Figure 4.25 TEM lattice fringe image from the SiC phase from SiCN35. SiC crystals are embedded in an amorphous phase.

2 nm

Figure 4.26 TEM lattice fringe image from the Si phase from SiCN35. Si crystals are coated with an ∼1–2 nm thick amorphous phase, very probably silica.

Conclusion regarding ceramic nanopowders from gaseous precursors SI/C-BASED MATERIALS

Though SiC173 was the only material investigated as-prepared, it is believed that comparing SiC173 with SiC177/1200 and SiC212/1200 is possible since the annealing temperature is rather low. Since the ethylene/silane ratio is about the same (∼0.5) for all the samples, the main difference in preparation conditions lies in the total flow rate. It increases from SiC177 to SiC173 to SiC212 for a similar C/Si ratio in the precursor gases, inducing a decrease in the residence time from 12.1 to 7.3 then 2.5 ms, respectively (Table 2.2). Correspondingly,

©2003 Taylor & Francis

4 nm–1

Figure 4.27 Electron diffraction on a Si phase-rich area from SiCN35. Crystals are pure Si, CFC structure.

the grain size range decreases, and SiC polytypes other than the primary form 3C (β structure), though present in the two former samples, do not have the time to develop in the latter. Likewise, the organisation of the polyaromatic C phase tends to decrease. Conditions for the preparation of SiC212 prevent the occurrence of a primary polyaromatic C phase. One explanation could be that long residence times used for SiC173 and SiC177 allow a Si C (H) solid solution enriched in C (relative to the 1 : 1 stoichiometry of SiC) to form in the reaction zone. The subsequent crystallisation of the SiC phase makes the excess C to be expelled as a polyaromatic C phase associated to the SiC crystals. Heat treatments in inert atmosphere provoke the increase in the average SiC grain size due to a coalescence mechanism (i.e. growth from coalescence) by impingements, while the primary C phase, when any, remains unchanged up to 1500◦ C. Beyond 1500◦ C (at least from 1800◦ C), the latter mechanism competes with an early decomposition mechanism, due to the nanometric size of the SiC phase. The partial decomposition of SiC induces the formation of a secondary polyaromatic C phase. Meanwhile, the SiC structure tends to increasingly change from a majority of 3C (cubic) polytype to a mixture of various polytypes (except 2H which is irreversibly unstable beyond ∼1600◦ C – Inoue and Kourachi 1983), probably mostly as stacking fault sequences within the same grains. An explanation is that the coalescence mechanism involves solid crystals, that does not allow single crystals with defect-free, homogeneous structure to be obtained. SI/C/N-BASED MATERIALS

Correspondingly to the higher N content in the precursor gas phase for SiCN29 (twice that of SiCN35), Si3 N4 crystals only develop in the former. A SiC phase develops in SiCN35 instead, further enhanced by a higher amount of C supply. The fact that no Si3 N4 phase is able to form in SiCN35 is amazing since, consistently with the theoretical phase content calculated from the elemental composition (Table 2.3), the N supply should be sufficient. This is not merely due to a lower formation kinetic for Si3 N4 relative to SiC, since Si3 N4 is able to form in SiCN29 while the residence time is lower. The main parameter is therefore believed to be the nature of the primary gases, and possibly the presence of C specifically.

©2003 Taylor & Francis

Thus, when combined with silane and ethylene, a C-free precursor like NH3 is more likely to induce the Si3 N4 formation than a C-containing precursor like CH3 NH2 . The reason is probably that the former precursor gas combination only provides C H, Si H and N H bonds which are easier to thermally break than the pre-existing C N bonds in the latter. Si N bonds are therefore allowed to form more easily and early in the former than in the latter. 4.1.2

Synthesis from liquid precursors

Nebulisation of liquid precursors such as hexamethyldisilazane (Si2 C6 NH18 ) produced other Si/C/N-based powders. Adding an increasing amount of NH3 allowed the C/N atomic ratio of the as-prepared material to be controlled. The various experiments result in a whole series of samples listed in Table 2.5, and characterised by a continuous decrease in their C/N atomic ratio. On the other hand, the same principle of using an increasing NH3 /Ar ratio for a constant total flow rate was applied to a nebulised liquid precursor made of a mixture of HMDS and aluminium isopropoxide (see Section 2.4.2) in order to prepare Si/Al/C/N-based powders. Materials obtained and relative preparation conditions are reported in Table 2.6. Si/C/N-based powders Among the whole series of samples listed in Table 2.5, HMDS42 and HMDS44 were investigated by TEM. They correspond to a slope change in the variation trend of the C/N ratio value measured in the starting materials v. the SiC/Si3 N4 content in the 1600◦ C-annealed materials (see Figure 3.9). HMDS42 and HMDS44 exhibit C/N ratios of 0.73 and 0.56, respectively. Since amorphous at the as-prepared state, they were investigated after increasing annealing heat treatments under pure N atmosphere at temperatures ranging from 1000◦ C to 1600◦ C. HMDS42, 1000◦ C ANNEALED

The bulk structural evolution of the sample with increasing annealing temperature from 1000◦ C and beyond is illustrated by electron diffraction, whose patterns are gathered on Figure 4.28. After 1000◦ C annealing, the daylight colour of the material is black. Though all the grains exhibit a rather round and isometric morphology (Figure 4.29(a)), a two-mode distribution of grain size is observed, with small grains (20–60 nm) being prevalent, and the subsidiary occurrence of large grains (80–60 nm). Small grains exhibit a nanometric porosity revealed by a Fresnel contrast effect when unfocused. Grains are associated into aggregates (often chainlike) either by strong, structural bonds (illustrated by the chain-like aggregate made of three large grains at the centre of Figure 4.29(a)), or by weak (electrostatic) forces (as illustrated by the small grain aggregate at the top of Figure 4.29(a)). It is worth noting that isolated grains look isometric, but aggregates are not. Considering one morphology (isometric) or the other (chain-like for instance) will provide very different shape factors, which is likely to explain possible mismatches between actual and calculated surface area of the powder. Electron diffraction patterns indicate an overall amorphous structural state, both for large and small grains (Figures 4.28(a) and (b)). However, patterns from small grains exhibit an additional modulation (white arrow) whose location suggests the early occurrence of

©2003 Taylor & Francis

Small grains

Large grains

(a)

(b)

1000°C (c)

(d)

1300°C (e)

4 nm–1

1400°C (f)

(g)

1500°C

Figure 4.28 Electron diffraction from HMDS42 on large grain and small grain phases respectively after increasing annealing treatment. From (a) to (g) the arrows show the occurrence of a faint C002 reflection.

©2003 Taylor & Francis

(a)

(b)

(c)

50 nm

Figure 4.29 TEM bright-field image and dark-field image sequence from HMDS42/1000. (a) illustrates the differentiation into two phases with different grain size range. (b) ‘C-specific’ dark-field imaging conditions. The polyaromatic C phase appears as tiny bright dots representing primary graphene stack entities ∼1 nm large. (c) ‘SiC-specific’ dark-field conditions. No phase has crystallised yet.

very small (1500◦ C) with low or no temperature gradient, growth mechanisms (and kinetics) are different, possibly due to a higher mobility of atoms allowing crystals to grow freely, therefore, giving rise to primary crystal morphologies instead of the round, isometric morphology (like that resulting from the flashtemperature nucleation-then-growth mechanism occurring for the gaseous-precursor-based nanopowders). Whatever, polyaromatic C is always the first phase to crystallise, though with a lowdimensionality – turbostratic – structure, according to a common behaviour of Si-based polymer-derived ceramics (Monthioux and Delverdier 1996). Consequently, the amount of it is directly related to the C content of the as-prepared material. SI/C/N/AL-BASED MATERIALS

Whether NH3 is present or not in the gas phase, all the as-prepared materials investigated are nearly all amorphous but tend to a two-morphology differentiation with small, nanoporous grains, and larger, non-porous grains, which could be related to the subsequent SiC/Si3 N4 phase differentiation. Effect of the 1600◦ C annealing • • • •

Average grain sizes decrease, revealing a weight loss subsequent to the ceramisation process. The crystallised phases mainly depend on the N content of the chemical system. Not mentioning the polyaromatic C phase, which is not systematic, the SiC phase crystallises prior to the Si3 N4 /SiAlON phase. SiC crystals are always coated with an amorphous (Si?) Al O phase which is an interesting feature for subsequent attempts of high-temperature sintering. Correspondingly to the high annealing temperature, the SiC structure is always a mixture of various polytypes (though with various proportions from a material to the other), present as stacking faults rather than as single crystals.

Effect of the presence of NH3 in the precursor gas phase • •

After annealing, N-depleted chemical system, induced by the absence of NH3 in the precursor gas phase, on one hand allows the occurrence of a polyaromatic C phase and on the other hand allows the formation of a crystallised SiC phase only (i.e. no Si3 N4 ). Adding NH3 in the precursor gas phase induces the early crystallisation of SiC. However, the nucleation step does not occur as a widespread germination but occurs locally. Such a heterogeneous behaviour has to be related to a heterogeneity in chemical composition, possibly originating in the non-homogeneous incorporation of N and/or H into the material. On the contrary, the absence of NH3 provides a quite amorphous material.

©2003 Taylor & Francis

•

Adding NH3 in the precursor gas phase allows the required stoichiometry to be reached, resulting in the rapid crystallisation of a SiAlON-type phase after annealing. Meanwhile, the high temperature at which the overall SiC crystallisation occurs induce such an atom mobility than the respective intrinsic growth rates of the various crystal faces are respected.

Effect of varying the NH 3 concentration in the precursor gas phase •

•

•

For the lowest and the highest supply in NH3 , the prevalent polytype of the resulting crystallised SiC phase after annealing is 3C, corresponding to the β-cubic structure. An intermediate NH3 supply results in a more rapid growth of the SiC phase, for which the other polytypes are correspondingly prevalent. Such a non-monotonous effect of the increasing N content of the chemical system regarding the behaviour of the material during annealing was already observed for Si-based polymer-derived ceramics and originates from the competition between antagonist mechanisms (Delverdier et al. 1992a). Increasing the N proportion slows down the kinetics of structural rearrangement; inducing the highest amount of remnant amorphous phases in the N-richest system after annealing. Another reason is also the early crystallisation of SiC prior to Si3 N4 , which induces a local excess of N atoms relative to the Si atoms left with respect to the required Si3 N4 stoichiometry, which then gather into an amorphous phase. The crystallisation of a nitride phase occurs for a 20% and 40% input of NH3 in the precursor gas phase. In both materials, the high annealing temperature tends to induce the overall crystallisation of massive SiAlON-type crystals rather than Si3 N4 . In addition, seldom SiAlON needle-like crystals are found in the latter sample (40% NH3 ), while numerous Si3 N4 needle-like crystals are found in the former (20% NH3 ). Assuming that the massive and needle-like morphologies correspond to α and β structures respectively, these data suggest that the structure change might be somehow controlled by the N content of the chemical system, independently from the extent of Al Si and O N substitution.

4.2 As-prepared sintered composites Some of the as-prepared, sintered, micro–nano-type composites, whose detailed processing conditions and mechanical properties are described in Chapter 7, were investigated by TEM to investigate further the structural and textural relationships between the composite components. Related sample preparation prior to TEM investigations and TEM operating conditions are reported in Section 4.3. Sintering temperature was 1650◦ C for all the composites (1600◦ C for the bulk Si3 N4 material). All the materials (i.e. bulk Si3 N4 and composites) are added with the regular sintering aids, Al2 O3 and Y2 O3 , added in the proportions of 3% and 6%, respectively. For technical reasons, all the materials but the bulk Si3 N4 were sectioned, thinned and observed in a direction parallel to the hot pressing direction. In the following, after describing the monolithic Si3 N4 material, composites will be presented according to the decreasing amount of theoretical SiC content (Table 7.1). 4.2.1

Bulk Si3 N4

Monolithic sintered Si3 N4 – that is, not added with any nanopowder – was studied as a reference material. The sintering temperature at 1600◦ C provides the highest densification rate

©2003 Taylor & Francis

200 nm

Figure 4.57 TEM bright-field image from the SiAlON monolithic material.

(see Figure 7.6(b)). The sample was investigated in a plane perpendicular to the hot-pressing direction. This plane also contains the direction of mechanical testing. The material appears highly densified (Figure 4.57). Due to the high sintering temperature and the presence of oxides, a dramatic chemical and structural transformation has occurred, inducing a high β-SiAlON (Si6−z Alz Oz N8−z ) crystal content (see Section 7.2), about 70% according to calculations based on XRD data (Rossignol 1995), as compared to 100% of α structure in the starting powder (see Section 7.4.1). Correspondingly, a majority of the grains are more or less elongated, some exhibiting an obvious needle-like morphology. This is consistent with previous TEM observations which demonstrated that oriented pressure sintering is able to statistically induce a preferred orientation perpendicular to the pressing direction as far as non-isometric crystals – like high-temperature β-SiAlON or β-Si3 N4 – are concerned (Champion et al. 1992). Of course, aspect ratios revealed by TEM are just apparent, since depending on the crystal orientation relative to the sectioning direction for the non-isometric grains. Actually, many of the nearly isometric crystal morphologies are likely to also derive from β-type crystals, that is, about 250–300 nm, while the maximal value observed for the elongated crystal dimension is about 2 µm. Crystals most often exhibit a uniform grey contrast, indicating that they are mainly free of stacking faults. 4.2.2

Si3 N4 /SiC173 composites

The composites investigated were initially prepared with 20% and 40% of SiC173 nanopowder, respectively (Table 7.1). Considering that the starting SiC173 nanopowder is made of quasi pure SiC (with some content of polyaromatic C, see section on Si/C-based powders), such proportions also correspond to the theoretical proportions of the SiC phase in the final composites (Table 7.1).

©2003 Taylor & Francis

No obvious discrepancy was found between composite Si3 N4 /SiC(20) and composite Si3 N4 /SiC(40) as observed in TEM. The following comments are therefore valid for both materials. Low magnification TEM images are consistent with the theoretical SiC phase proportions (Figure 4.58), and emphasise the size difference between the SiAlON and SiC phases. SiAlON crystals appear as large, polyhedral grains, quite similar to that of the monolithic material (Section 4.2.1). Although diffraction patterns indicate a majority of β structure consistent with the high-temperature sintering (Figure 4.59(a)), a reflection specific of the α phase (at 0.432 nm) is still found sometimes (Figure 4.59(b)), indicating that the structural α to β transformation rate is not 100%, which is consistent with the theoretical calculations (Table 7.3). However, considering that XRD was not able to reveal any α structure, the actual amount of α crystals is surely less than the 11% predicted from calculations (and probably less than 7%, the lowest percentage determined by X-ray in composite Si3 N4 /SiCN35(22.2)). Most of the crystal facets appear as straight lines, as for single crystals when allowed to grow freely (Figure 4.58). However, crystal facets of SiAlON crystals at the contact of SiC grains often appear notched (see Figure 4.66 for reference) due to the presence of SiC grains at the interface, confirming that SiC grains were not allowed to melt or soften during the sintering process, unlike the growing SiAlON crystals. It is worth noting that only few elongated crystals are found. Calculating an average apparent crystal size is therefore possible: ∼470 nm

100 nm

Figure 4.58 TEM bright-field image from composite Si3 N4 /SiC(20). The SiC crystals (small, round grains) are gathered into aggregates.

©2003 Taylor & Francis

(a)

(b)

4 nm–1

Figure 4.59 Electron diffraction from composite Si3 N4 /SiC(20). (a) Most of the diffraction spots are characteristic from the β-SiAlON phase. (b) However, some specific diffraction spots (arrow) reveal the minor occurrence of the α structure as well.

for composite Si3 N4 /SiC(40). The slightly lower value for Si3 N4 /SiC(20) is not significant. Comparison with the monolithic Si3 N4 material regarding the evolution of the grain size is limited since it was not investigated along the same section plane. SiC crystals appear as small, round crystals, most often gathered into agglomerates instead of being randomly dispersed all around the SiAlON crystals (Figures 4.58). Contrarily to the latter, SiC crystals exhibit numerous, non-periodical, dark contrasts revealing the occurrence of polytypes as stacking faults. Indeed, the β to α (i.e. polytypes other than β) SiC transformation has slightly progressed, as ascertained by the more frequent occurrence of specific features in the diffraction patterns, like the occurrence of diffraction spots aside the regular diffraction rings for the β structure, and of some aligned spots corresponding to the presence of stacking faults (Figure 4.60 to be compared with Figure 4.3). On the contrary, the average SiC grain size has not progressed significantly compared to the starting powder (see section on Si/C-based powders). This indicates that the growth mechanism by coalescence, though an usual process at such temperatures (Monthioux and Delverdier 1996), was prevented, probably because of the presence of the intergranular phase (Figure 4.62), as opposed to the starting nanopowder (see Section 4.1.1). Some SiC grains are also found possibly enclosed into SiAlON crystals (Figure 4.61). However, high magnification images are not sufficient to ascertain if the SiC crystals are actually enclosed into the SiAlON crystals or if they merely are superimposed to them in the projected image as allowed by the thickness value of the TEM preparation (∼100 nm). Proofs of enclosing of SiC crystals into SiAlON crystals will be brought using dark-field imaging (see Figure 4.66 to come). The polyaromatic C phase associated with SiC grains revealed in the starting powder is no longer found, suggesting that it might have burnt into CO during the sintering process. An amorphous, intergranular phase is found, gathering the whole yttrium content of the composite (in addition to some content in the other elements present, Si, Al, O, N) as revealed

©2003 Taylor & Francis

4 nm–1

Figure 4.60 Electron diffraction from a SiC grain agglomerate in composite Si3 N4 /SiC(40). Though reflections from the β structure are still prevalent, many other polytypes are present.

5 nm

Figure 4.61 High magnification TEM image from composite Si3 N4 /SiC(20). A SiC grain (faulted) seems to be enclosed in the SiAlON crystal.

by the darkness of the grey contrast associated to this phase, originating from the intense electron inelastic scattering induced by the high Z value for yttrium (Z = 39). However, most of the intergranular phase is found within SiC grain agglomerates either filling the voids between grains (Figure 4.62) or at the grain interfaces (Figure 4.62, arrows). On the contrary, the intergranular phase is absent between SiAlON crystals (Figure 4.63) or might eventually

©2003 Taylor & Francis

SiAION

SiC

SiC

SiC

5 nm

Figure 4.62 High magnification TEM image from composite Si3 N4 /SiC(20). Spaces between SiC grains or at the SiC/SiAlON interface appear dark due to the presence of the yttrium-containing amorphous intergranular phase.

be present as some very thin – 5% according to the theoretical calculations), the SiC phase tends to gather into agglomerates apart of the SiAlON phase, instead of being dispersed in intergranular position. Agglomeration could be caused by the occurrence of the surrounding minor phase (polyaromatic C in case of SiC173 nanopowder, and an amorphous silicon oxycarbide phase in case of SiCN35 nanopowder), preventing a convenient dispersion during the mixing step prior to making the green bodies (see Chapter 7). Such agglomerates, because they are made of small, round grains having limited surfaces of contact, exhibit a high intergranular porosity allowing to somehow trap most of the yttrium-containing amorphous phase, which is then no longer available to form interfacial films between SiAlON crystal facets. A poor superplastic behaviour is therefore expected for these composites. On the contrary, a low content in SiC nanograins is favourable to a better dispersion in intergranular position between SiAlON crystals. However, SiAlON crystal faces are then prevented to grow freely, the hindering presence of the SiC grains at the SiAlON crystal interfaces while developing from the liquid phase inducing indentations in the related crystal faces. SiC grains thus act as ‘keystones’ between SiAlON grains, which is also detrimental to a superplastic behaviour. Nano/nano composites are therefore expected to be preferable to micro/nano composites regarding the ability to creep without developing cracks. Further and extended comments will be found in Chapter 7.

4.3 TEM sample preparation prior to investigations and operating conditions 4.3.1

Preparation of samples

In order to be electron transparent at 120 kV acceleration voltage, materials investigated should exhibit a thickness ideally lower than 100 nm in the direction parallel to the electron beam. Ceramic nanopowders therefore require limited efforts for TEM preparation, considering the convenient initial grain sizes. On the contrary, composites require a specific thinning process. Powders Aliquots of each ceramic nanopowder were dispersed into ethanol under sonication for several minutes in order to make a suspension from which a drop was then deposited on a regular (3 mm) 200 mesh copper microgrid previously coated with a holey or a lacey amorphous C film. For samples suspected to be possibly hydrolysable, dry preparation – that is, direct deposition of the nanopowder onto the microgrid – were sometimes also made for double checking. When large powder agglomerates were formed unable to be dismantled by the wet sonication step, gentle grinding using a boron carbide mortar was previously made (use of a silica mortar should be avoided, due to its lower hardness relative to SiC, which is likely to provoke some input of silica into the sample being prepared). The main limitation of such a preparation procedure is the possibility to induce a gravimetric discrimination – the phases exhibiting much higher density, or much larger grains being likely to sink to the bottom of the flask.

©2003 Taylor & Francis

Sintered materials A slice about 100 µm thick of each sintered materials was obtained using a diamond-coated slow circular-saw (preferably to a diamond-coated tungsten-wire saw, whose efficiency was limited in this specific case). The slice dimensions was then adjusted to fit in a 3-mm wide copper aperture, then glued on a stainless steel support using a thermofusible resin. Then the thickness of the slice was decreased down to ∼50 µm using a mechanical polishing procedure involving submicrometric diamond powders. The slice was then reversed upside down in order to complete the polishing/thinning process through the other side of the slice, hopefully down to ∼20 µm. The slice was then glued onto a 3-mm copper aperture, then ion-milled both sides for several hours. Acceleration voltage for the argon ions was ∼4 kV, and ion gun direction angles with respect to the sample surface were 12–15◦ . A hole was created, whose edges were suitable for the TEM investigation. All the micro/nano composites investigated (specifically those with a high SiC content) were more or less found to specifically exhibit a very high hardness making them unusually difficult to slice during the first step of the procedure (diamond-enhanced sawing), as opposed to the behaviour of regular micro/micro-type composites (such as in Champion et al. 1992). On the contrary, the micro/nano composites were found to dismantle more easily than regular micro/micro composites at the end of the polishing/thinning procedure. A possible explanation is that the primary specifically high hardness is due to the lack of amorphous intergranular phase at the SiAlON crystal interfaces, making the crystals collectively supporting the shearing stresses developed during the sawing procedure as a monolithic single crystal would do. On the other hand, the removal of SiC grain agglomerates is easier during the next polishing/thinning step due to their high content in amorphous phase. This is likely to promote the earlier dismantling of the composite with respect to micro/micro composites, as soon as a critical limit (∼50 µm) for the slice thickness is reached. 4.3.2

TEM operating procedure

We do not intend to describe all about the principle of electron imaging. The goal of this section is merely to provide to the non-microscopist reader the keys needed to understand the comments reported in this chapter. Basics and more about the principles of TEM can be found in Williams and Carter (1996). About dark-field mode imaging Contrasted bright field and lattice fringe are two regular imaging modes involving the use of apertures with suitable opening diameters placed in the back focal plane of the objective lens. For the non-microscopist reader, considering the electron microscope as a powerful, highresolution magnifying lens able to image down to the lattice planes of crystals is enough to understand the kind of comments made in this chapter. Likewise for the selected area electron diffraction, whose principles are similar to that of XRD. However, further explanation might be useful to fully understand the data brought by the dark-field imaging mode. Although a TEM may contain seven or more electromagnetic lenses, its main quality and mode abilities depend on the objective lens. As for regular, optical glass lenses, electromagnetic lenses are characterised by three planes: the focalisation plane (where the specimen should be located for its image to be focused on the image plane), the focal plane (specifically the back focal plane, which contains the electron diffraction patterns in the case of the objective lens), and the image plane, where the enlarged image of the specimen is formed

©2003 Taylor & Francis

(a)

(b)

e–

e–

Specimen

Objective lens Back focal plane Objective aperture

Image plane

Figure 4.68 Sketch of the electron beam paths in the objective lens of a microscope. (a) Bright-field image, (b) dark-field image. B

1

A 1

SiC 200 + C 10 SiC 220 SiC 311

DF1

DF2

1/d (nm–1)

C 002 SiC 111

Figure 4.69 Examples of various positions of the objective aperture relative to the diffraction pattern of a multiphase material (β-SiC + turbostratic C) used to perform specific dark-field imaging.

(Figure 4.68(a)). Placing an aperture (so-called objective aperture) in the back focal plane of the objective, therefore, allows specific electron beams to be selected, that is, the image is built with some electron beams only (Figure 4.68(b)). For instance, selecting a specific diffracted beam with the objective aperture results in an image whose bright parts indicate the regions of the specimen from which the selected diffracted beam is emitted, while everywhere else, in or around, the specimen appears dark (or nearly). It is then already quite clear that, by successively selecting adequately various positions for the objective aperture, sequential dark-field images can be formed which can be specific of different orientations for a single crystallised phase, or of different phases within a multiphase material, or both. This principle is illustrated in Figure 4.69, with the example of a material

©2003 Taylor & Francis

containing both β-SiC and polyaromatic, turbostratic C. The related theoretical diffraction patterns are both represented. Positioning the aperture in DF1 allows part of the C002 diffracted beams to go through, and to build the image of polyaromatic entities – if any – whose (001) lattice planes are oriented both parallel to the electron beam (i.e. graphenes are seen edge-on) and parallel to the E–W direction (or nearly, since the width of the aperture is such that it admits an arc instead of a single dot). Then, positioning the aperture at 90◦ along the same diffraction ring (position DF1⊥) will image other polyaromatic entities – if any – oriented S N, that is, perpendicularly to the former. The related two images are exclusive from SiC crystals and are therefore specific of the C phase. SiC crystals will in turn be able to be imaged by using position DF2. In that case, rotating the aperture relative to the diffraction pattern all along the SiC111 diffraction ring (meanwhile also rotating along the SiC200 ring, which is very close) will successively reveal any SiC crystal oriented so that they present (111) or (100) reticular planes parallel to the electron beam. It is worth noting that the DF2 position is not really specific of a SiC phase, strictly speaking, since the aperture width is such that the 10 diffraction band of turbostratic C is also admitted, allowing the related electrons to participate in building the image (using smaller apertures would be too much detrimental to the resolution). Graphene stack entities oriented perpendicularly to the electron beam (i.e. lying flat on the specimen support grid) are thus supposed to be imaged together with the SiC crystals. Practically, due to the intensity of the SiC111 reflection, by far much higher than for the C10 , DF2 images are considered specific of a SiC phase in most cases. Because the objective aperture is placed in various positions relative to the diffraction pattern, such dark-field imaging sequences are said to correspond to a radial and azimuthal exploration of the reciprocal space. For this reason, the objective aperture apparent widths are given in nm−1 . TEM operating conditions Bright-field, dark-field, lattice fringe and selected area diffraction mode images were obtained using a Philips EM400 microscope operating at 120 kV and equipped with a high-resolution stage providing a point resolution better than 3 Å, with a tungsten wire as electron source. The width of the aperture used for the contrasted bright-field and lattice fringe imaging was 8 nm−1 , while that used for dark-field imaging was 2 nm−1 . According to the Abbe relation ρ = 0.61 × λ/a where λ is the wavelength and a the half-width of the aperture (in radians) using an objective aperture as small as 2 nm−1 reduces the resolution ρ to ∼6 Å. Dark-field imaging was carried out most often using DF1 (aperture centred at 2.2 nm−1 ) and DF2 (aperture centred at 4.3 nm−1 ) positions described in the previous section. Considering the width of the aperture opening, the radial tolerance regarding the frequency range admitted for both positions are 1.3–3.3 nm−1 and 3.3–5.3 nm−1 , respectively. Likewise, the azimuthal deviations relative to the centring position due to the aperture opening are ±24◦ and ±15◦ , respectively. Attention has to be paid that, depending on the phases present in the material investigated, the specificity of the positions may vary. For instance, the powder diffraction pattern for a Si3 N4 phase exhibits numerous diffraction rings (Figure 4.20 for instance), making Si3 N4 crystals likely to be imaged in any possible position for the objective aperture, including DF1 and DF2. Also, other aperture positions were sometimes used, for specific purposes. Finally, though they do not provide diffracted beams strictly speaking, but are anyway able to scatter electrons in some specific directions because of short-range structural periodicities (often around

©2003 Taylor & Francis

the most intense diffracted beam for the related crystallised material), so-called amorphous materials are likely to also appear in dark-field images. For instance, amorphous silica may be revealed in DF1 position, providing an intense halo at 0.44 nm−1 (Chaudhari et al. 1972), while the most intense reflection for any form of crystallised silica is located at ∼0.41 nm−1 . X-EDS spectroscopy was performed using a Philips CM20 microscope equipped with a high-resolution goniometric stage (super-twin stage) and an EDAX detector (Li-doped silicon diode), protected by an ultra-thin window allowing the detection of chemical elements including light elements down to C.

References Audier, M., Oberlin, A., Oberlin, M., Coulon, M. and Bonnetain, L. (1981) Carbon, 19, 217. Badami, D. V. (1965) Carbon, 3, 53. Buffat, P. and Borel, J.-P. (1976) Phys. Rev. A, 13, 2287. Champion, E., Goursat, P., Besson, J. L., Madigou, V., Monthioux, M. and Lespade, P. (1992) Ceram. Eng. Sci. Proc., American Ceramic Society, USA, p. 732. Chaudhari, P., Graczyk, J. F. and Herd, R. (1972) Phys. Stat. Sol., 51, 801. Delverdier, O., Monthioux, M., Mocaer, D. and Pailler, R. (1992a) Proc. 5th Eur. Conf. Compos. Mater., (eds A. R. Bunsell, J. F. Jamet and A. Massiah), Bordeaux, France, p. 691. Delverdier, O., Monthioux, M., Oberlin, A., Lavedrine, A., Bahloul, D. and Goursat, P. (1992b) High Temp. Chem. Proc., 1, 139. Delverdier, O., Monthioux, M., Mocaer, D. and Pailler, R. (1993) J. Eur. Ceram. Soc., 12, 27. Delverdier, O., Monthioux, M., Mocaer, D. and Pailler, R. (1994) J. Eur. Ceram. Soc., 14, 313. Iijima, S. (1982) Solid State Chem., 42, 101. Inoue, Z. and Kourachi, Y. (1983) Proc. Int. Symp. Ceram. Components Engines, Japan, p. 519. Laanani, Fadéla (1997) Thèse de Doctorat, no 2707, Université Paul Sabatier, Toulouse, France. Le Coustumer, P., Monthioux, M. and Oberlin, A. (1993) J. Eur. Ceram. Soc., 11, 95. Madigou, Véronique, Monthioux, M. and Oberlin, A. (1990a) Proceedings of Matériaux Composites pour Applications à Hautes Températures (eds R. Naslain et al.), AMAC-CODEMAC, Bordeaux, France, p. 221. Madigou, Véronique, Monthioux, M. and Guimon, C. (1990b) Proc. XX Int. Conf. Carbon, GFEC, Paris, France, p. 220. Madigou, Véronique (1991) Thèse de Doctorat, n◦ 116, Université de Pau et des Pays de l’Adour, France. Mocaer, D., Pailler, R., Naslain, R., Richard, C., Pillot, J.-P., Dunoguès, J., Delverdier, O. and Monthioux, M. (1993) J. Mater. Sci., 28, 2639. Monthioux, M. and Delverdier, O. (1996) J. Eur. Ceram. Soc., 16, 721. Monthioux, M., Oberlin, A. and Bouillon, E. (1990) Compos. Sci. Technol., 37, 21. Niihara, K., Suganuma, K., Nakahira, A. and Izaki, K. (1990) J. Mater. Sci. Lett., 9, 598. Oberlin, A. (1989) Chemistry and Physics of Carbon, vol. 22 (ed. P. A. Thrower), Marcel Dekker, New York, p. 1. Oberlin, A. and Rouchy, J.-P. (1971) Carbon, 9, 39. Oberlin, A., Oberlin, M. and Comte-Trotet, J. R. (1976) J. Microsc. Spectrosc. Electron., 1, 391. Oberlin, A., Boulmier, J. L. and Villey, M. (1980) Kerogen (ed. B. Durand), Technip, Paris, p. 191. Oberlin, A., Bonnamy, S., Bourrat, X., Monthioux, M. and Rouzaud, J.-N. (1986) Petroleum Derived Carbons (eds J. D. Bacha, J. W. Newman and J. L. White), ACS Symposium Series 303, USA, p. 85. Okamura, K., Sato, M., Hasegawa, Y. and Amano, T. (1984) Chem. Lett., 2059. Rossignol (1995) Thèse de Doctorat, Université de Limoges, France. Sasaki, G., Suganuma, K., Fujita, T., Hiraga, K. and Niihara, K. (1993) Mater. Res. Soc. Symp. Proc., 287, 335. Williams, D. B. and Carter, C. B. (1996) Transmission Electron Microscopy. A Textbook for Materials Science, Plenum Press, New York.

©2003 Taylor & Francis

5

From short- to long-range order Structural organisation of silicon-based nanopowders

©2003 Taylor & Francis

5.1 Chemical order studied by solid-state nuclear magnetic resonance André Pierre Legrand, Jean-Baptiste d’Espinose de la Caillerie and Youssef El Kortobi

Native or heat-treated ceramics contain different atoms, the nuclei of which can be used, if provided with a magnetic moment, as NMR probes. Among them 1 H,13 C,27 Al and 29 Si are of particular interest. Different information can be obtained, depending on the method of observation and on the nature of the nucleus under analysis. Each proton has a magnetic moment (1 H) and is relatively easy to observe. Nevertheless, it physically interacts through dipole–dipole interactions. This induces a line broadening of the spectrum, which can hide the chemical information concerning the different chemical species on, and inside ceramics (hydroxyls, water, etc.). So different approaches are possible by taking advantage of this dipolar interaction or on the contrary by suppressing it. Silicon nucleus (29 Si), on the contrary to 1 H, is an isotope with a low natural abundance of 4.7%. The same is true for 13 C (1.1%). For as-formed ceramics, these diluted species are in dipolar interaction with the abundant proton (1 H) species. To observe the nature of the different chemical species in the material, it is necessary to suppress this dipolar interaction. This is obtained through specific techniques. Aluminium nucleus (27 Al) has a natural abundance of 100%. Contrarily to the previous nuclei, 27 Al is not a spin 21 . Its quantum number is 5/2. This has positive and negative effects. A positive one is that the nucleus is sensitive to the local electric field gradient and consequently this gives an important information on the symmetry of the aluminium sites. The negative one is that if the electric field gradient from one site to another site is different, for example, tetrahedral to octahedral, the quantitative determination of the relative proportions could be questionable.

5.1.1 Nuclear magnetic methods As mentioned above, the as-formed amorphous ceramics may contain hydrogen atoms belonging to different chemical species such as hydroxyls and adsorbed water molecules. They constitute a set of 1 H spins which, by dipolar interaction, produce a line broadening hiding the chemical information. This harmful effect disappears by heat treatment, with the expulsion of protonated species.

5.1.1.1

Rare spin half spectroscopy (29 Si, 13 C)

Considering the low abundance of 29 Si,13 C isotopes, the dipolar interaction responsible for line broadening is mainly due to neighbouring protons, for as-formed ceramics. To determine

©2003 Taylor & Francis

the nature of the different functionalities of such nuclei in ceramics, it is necessary to use specific techniques. High power decoupling (HPD) To suppress the unwanted proton-rare spin dipolar interaction, the following method can be used. It requires application of a strong radio frequency field at the Larmor frequency of the abundant spins. Through this operation, the populations of spins in the low and high Zeeman energy levels are made equal and the resulting magnetisation falls to zero, leaving only the chemical shift interaction to be studied (Bloch 1958). Averaging the chemical shift anisotropy (CSA) – Magic angle spinning (MAS) The method involves putting the sample into a rotor, the axis of rotation of which is at an angle of rotation of 54.73◦ with the external magnetic field. This method is efficient not only to average the dipolar interaction, when MAS speed is high enough, but also the CSA (Engelhardt and Michel 1987). The chemical shift has a theoretical expression that is mathematically similar to that of the dipolar interaction. It is composed of an isotropic value, similar to that observed in liquid state, and an anisotropic part. In consequence, for a powder, this term produces a line broadening, due to the superposition of all possible orientations of the chemical species studied. This requires that the speed of rotation be high enough relatively to the CSA. If not, a more complex spectrum appears which is composed of the mere spectrum, flanked on either side with more or less intense replicas of this one. These satellites are separated by multiples of the rotation frequency. The different isotropic lines are characterised by a position in frequency with respect to a given reference, tetramethylsilane (TMS) in this book, and consequently can be identified. Numerous experiments in solids have demonstrated that the observed chemical shifts are near what is observed in liquids, if the environment is the same. Data already collected for species in solution can therefore be used to analyse solid-state spectra (Lippmaa et al. 1982; Lippmaa and Samoson 1982; Engelhardt and Michel 1987; Fyfe 1983; Marsmann 1981). As the technology has improved, allowing in particular to increase the MAS speed, it is possible to achieve such measurements using higher magnetic fields than previously, which enables increasing resolution and sensitivity. Cross-polarisation (CP) For as-formed ceramics containing an important amount of hydrogen, this method is jointly used with MAS to enhance the signal of carbon (C) or silicon (Si) bonded to protons. First, it involves cooling down the proton reservoir through a sequence shown in Figure 5.1.1. Second is the time to transfer the low temperature of the proton reservoir to that of rare spins. Last is to observe the signal of the rare spin while at the same time using the HPD. The method is consequently denominated CP/MAS. The complete sequence is represented in Figure 5.1.2. Model compound example Role and interest of the different methods mentioned above are summarised in Figure 5.1.3. 29 Si NMR spectra of a powdered sample of pure cubic octamer silicic trimethylsilyl ester

©2003 Taylor & Francis

x

y

Abundant spins

Time /2 Spin lock (b)

M1H

(a)

Measurement M1H

y

M1H

(c)

B1

B1

y Relaxation in the rotating frame

y

90° phase shift B1 x

x

x

Observation in the rotating frame

Figure 5.1.1 Setting of a spin temperature for the abundant spins (observation in a plane perpendicular to the director field B0 ): (a) the magnetisation is brought in the perpendicular plane after a 90◦ pulse; (b) the radio frequency field B1 is phase shifted by a 90◦ angle, bringing it parallel to the magnetisation; (c) after a certain time, the magnetisation decreases slowly in the B1 field and is measured after switching this field off, on the free induction decay. The upper part of the figure represents schematically the application of radio frequency pulses along the Ox and Oy axis in the rotating frame. (Reproduced with permission of John Wiley & Sons from Legrand et al. 1998.)

x

y

Abundant spins

Time /2 Spin lock

Decoupling

Waiting time

Time

Rare spins Contact

Measurement

Figure 5.1.2 Complete sequence of CP for abundant spins (e.g. protons 1 H) in interaction with less abundant spins (e.g. silicon 29 Si). (Reproduced with permission of John Wiley & Sons from Legrand et al. 1998.)

©2003 Taylor & Francis

(d)

a.u.

(c)

(b)

(a) 60

40

20

0

–20 –40 –60 –80 –100 –120 –140 –160 ppm

(d) Q8M8 spin rotation 5 kHz

13

12

11

10 –107 ppm

–108

–109

–110

–111

Figure 5.1.3 Rare spin 29 Si solid state NMR spectra at 59.6 MHz of a powdered pure cubic octamer silicic trimethylsilyl ester (Q8 M8 ) as model compound, showing the influence of different types of NMR measurements: (a) static or broadband spectrum with no high power decoupling of the protons. Only a broad line is observed: no chemical information is available; (b) static or broadband spectrum with high power decoupling of protons. Two bands, due to chemical anisotropy, corresponding to two Si are observed: no clear chemical identification available; (c) HPD with CP (contact time 1 ms) and MAS with a speed or rotation not high enough (567 Hz). Chemical information partially available; (d) CP/MAS + HPD. Chemical information available. (Courtesy of P. Tougne.)

(Q8 M8 ) present the following aspects: 1 2 3 4

With static or broadband method, only broad lines mainly due to the dipolar interaction 1 H 29 Si are observed. No chemical information is available. As above, with HPD of the protons, CSA of two Si types are observed. CP/MAS method, but with MAS speed of 560 Hz. Side bands with the mere spectrum are observable. CP/MAS method, with MAS speed high enough. Side bands are suppressed and the spectrum shows a triplet and a quadruplet.

©2003 Taylor & Francis

5.1.1.2

Aluminium 27 Al NMR spectroscopy

Unlike 29 Si or 13 C, 27 Al has an I = 5/2 spin and therefore a nuclear quadrupole moment. This implies that not only this nucleus is sensitive to the chemical shift and dipolar interactions but also to the quadrupolar one. The six energy levels (2I + 1) of the 27 Al due to the Zeeman effect are modified, assuming a first-order perturbation effect, which results in five different transition frequencies. The corresponding Hamiltonian is: Hˆ Q =

 eQ eq  ˆ2 η  ˆ2 3Iz − I (I + 1) + I+ + Iˆ−2 4I (2I − 1) 2

(5.1)

where eQ is the nuclear quadrupolar moment and Vzz = eq, the main electric field gradient (|Vzz | > |Vyy | > |Vxx |); η = (Vyy − Vxx )/Vzz is the asymmetrical parameter. Oxyz are principal axes of the tensor that describe the electric field gradient. All transitions except the central one ( 21 ↔ − 21 ) are subject to a first-order quadrupolar interaction depending on the orientation of the electric field gradient tensor with respect to the external magnetic field. Since in powder samples there is a random distribution of such orientations, it results in a large spreading of the line which prevents its observation. On the contrary, the central line is only sensitive to the second-order perturbation effect, which causes a line spreading proportional  2 to e2 Qq/ h /νL = νQ , where νL is the Larmor frequency of the 27 Al nucleus. Contrarily to spin half, the observation is not obtained with a 90◦ pulse, although this one is obtained with a solution of aluminium salt, in which the Brownian motion averages the quadrupolar effect. Samoson and Lippmaa (1983) proved that the signal intensities are independent of the magnitude of νQ only if the pulse length is short enough. For an error less than 5% in the signal intensity for the centre band of the central transition ( 21 , − 21 ), the spin flip angle ω1 τ must satisfy the condition ω1 τ (I + 1/2) ≤ π/6, with I = 5/2 for 27 Al. An example of the influence of the pulse duration on the relative intensities of the lines is given in Figure 5.1.4 for 23 Na (spin 3/2). Such quadrupolar and line shifts render more difficult the interpretation of the line structure but are able to provide additional information of the charge distribution (Figure 5.1.5). 5.1.1.3

Interest of the method

The chemical shift of the nuclear resonance provides an incredibly delicate fingerprint of the chemical functionality of an element in a compound. NMR has provided chemists with one of the most powerful qualitative and quantitative analytical tools available to the chemist, material scientist and medical technician. The advent of high-resolution techniques appropriate to nuclei in solids supplied further dimensions to do qualitative analysis of chemical functionalities. Moreover, some information concerning structure and local organisation of clusters can be obtained. Such a method had been largely used for ceramic characterisation (Müller 1999; Templin et al. 1999).

5.1.2 SiC powders analysis In Figure 5.1.6, the 29 Si spectrum for sample as-formed SiC163 is shown. Like previously published spectra of silicon carbide powders (Table 5.1.1), it consists of a superposition of two kinds of line: a single line at −16.0 ppm can be attributed to the β-SiC type, whereas the triplet consisting of lines at −15.6, −20.5 and −24.9 ppm can be attributed to α-SiC polytypes.

©2003 Taylor & Francis

(a) NaCl NaNO2

10

5

0

–5

–10 –15 –20 –25 –30 –35 –40 –45 –50 ppm

(b)

2,925

Integral 10

6

2

–2

17,590 –6 –10 –14 –18 –22 –26 –30 –34 –38 ppm

Figure 5.1.4 2D solid state 23 Na MAS–NMR spectra at 79.39 MHz of sodium chlorine and nitrite salts. (a) Influence of the pulse length (from 0.25 to 5 µs, by step of 0.25 µs) on the relative intensities of the two sodium sites; (b) Relative proportion has to be evaluated at short pulse length (here 1 : 6). (Courtesy of P. Tougne.)

As the experimental high-resolution solid-state NMR technique involves only MAS, the spectra can be decomposed and the amount of each type, α-SiC or β-SiC, is proportional to the respective areas of the different lines, if the repetition time for accumulation is long enough relatively to the spin-lattice relaxation times. The 29 Si spectrum of SiC174 is very similar to the previous spectrum except that the amount of α-SiC has decreased a little.

©2003 Taylor & Francis

Altet

Aloct

a.u.

Gibbsite Saponite

90

80

70

60

50

10 ppm

5

0

–5

–10

Figure 5.1.5 Solid state 27 Al MAS–NMR spectrum at 130.32 MHz of gibbsite and saponite. MAS speed: 12 kHz. Clear distinction between four- and sixfold coordinated aluminium in AlO4 and AlO6 units is observed. 27 Al chemical shifts are between +50 and +80 ppm for AlO4 and about −10 to +20 ppm for AlO6 .

SiC163

0

SiC174

– 40 ppm

0

SiC173

– 40 ppm

0

SiC177

– 40 ppm

0

– 40 ppm

Figure 5.1.6 Solid state 29 Si MAS–NMR spectra at 59.6 MHz of as-formed silicon carbide (duty cycle, 500 s; 90◦ pulse, 5 µs; number of scans, 161, 100, 164 and 60). Spectra are classified as function of decreasing residence time into the laser irradiation cell (Table 2.1). (Reprinted from Tougne et al. (1993) Diamond Relat. Mater., Evolution of the structure of ultrafine SiC-laser-formed powders with synthesis conditions, 2, 486–490, with permission from Elsevier Science.)

©2003 Taylor & Francis

Table 5.1.1

29

Si chemical shift in ppm/TMS for silicon carbides

Authors

Finlay et al. 1985 Guth et al. 1987 Hartman et al. 1987 Wagner et al. 1989 Dando and Tadayyoui 1990 Apperley et al. 1991 Schmidt et al. 1991–2 Richarson et al. 1992 Dybowski et al. 1996 Tateyama et al. 1997

Polytype 3C(C)

6H (A B C)

15R (A B C)

−18.3

−13.9 −20.2 −24.2

4H(B C)

2H (D)

−18.4 −16.1 −14.3 −20.4 −24.9 −14.6 −20.5 −24.1 −18.3 −13.9 −2.02 −24.5 −14.9 −20.8 −24.4 −18.4 −16.1a −16

−14.4 −20.5 −25

−18.4

−14.7 −20.9 −25.4

−18

−14

−20

−15.2 −20.2 −24 −19.7 −22.5 −20

−25

−14.5 −20.4 −25.1 −16 −18.4

−22

−27

−14.7 −19.7 −25.4

−20.9 −22.5 −20

Note a Depends on the crystalline structure of the sample (−16.1 for small single crystals).

No quantitative accurate decomposition has been performed but it is qualitatively apparent from the spectra. Moreover, on comparing the figure which shows the spectrum for sample asformed SiC173 with the figure which shows the spectrum for sample SiC177 the trend follows the same line. Increasing the residence time into the hot region of the cell where the synthesis takes place results in a decreasing amount of α-SiC phase. To assess this evolution, samples asformed SiC163 and SiC177 were annealed in a furnace at 1600◦ C for 24 h. Figure 5.1.7 shows the spectra of the annealed samples SiC163 and SiC177. One single line at −16.0 ppm now appears which can unambiguously be attributed to a pure β-SiC phase, it appears therefore rather surprisingly that the β-SiC phase is the high temperature stable phase. By comparing the evolution of samples SiC163 and SiC171, one can study the effect of heat treatment. Table 5.1.2 shows that the grain growth occurs at a higher temperature for sample SiC171 than for sample SiC163. At 2000◦ C, the densification of sample SiC171 is 94% but the grain size has become comparable to the grain size of a commercial powder. The problem is then to achieve good densification and conserving a fine structure. A good knowledge of the local structure evolution would be very helpful. Figure 5.1.8 presents the NMR spectra of samples SiC163 and SiC171 after heat treatment between 1600◦ C and 2050◦ C. One can see that the α-phase is always present in sample SiC171, while it almost disappears in sample SiC163 between 1700◦ C and 1900◦ C. At 2050◦ C, β-phase is transformed into α polytypes. 6H, 15R and 4H polytypes are unambiguously identified in sample SiC171 while only the 6 H environment is in sample SiC163. It seems then that there is a relation between the annealing and the presence of the α-SiC polytypes.

©2003 Taylor & Francis

SiC163

SiC177

0

ppm

–40

0

ppm

–40

Figure 5.1.7 Solid state 29 Si MAS–NMR spectra at 59.6 MHz of 1600◦ C (24 h) annealed silicon carbide (duty cycle, 500 s; 90 ◦ pulse, 5 µs; number of scans, 69 and 57). (Reprinted from Tougne et al. (1993) Diamond Relat. Mater., Evolution of the structure of ultrafine SiC-laser-formed powders with synthesis conditions, 2, 486–490, with permission from Elsevier Science.)

Table 5.1.2 Surface of different silicon carbide batches as a function of temperature dwell (dwell duration: 1 h) Temperature (◦ C)

1600 1800 2000

SBET (m2 /g) SiC171

SiC163

SiC177

24.9 3.7 0.09

13.2 7.3 5.3

8.1 3.3 2.3

5.1.3 Si/C/N powders analysis 5.1.3.1

29 Si

and 13 C solid-state MAS–NMR measurements

Well-crystallised silicon nitrides have been analysed and 29 Si chemical shifts of the two main α and β structures had been determined (Table 5.1.3). Average chemical shift identification of different Si environments in Si/C/N compounds has been done by comparison between pure silanes and pure silicon carbide or silicon nitride. Analysis of the results is given in Table 5.1.4. Besides, different authors have estimated the presence in less organised compounds of Si environments as SiCN3 , SiC2 N2 and SiC3 N (Seitz et al. 1996; Gérardin et al. 1997; Verdecia

©2003 Taylor & Francis

SiC171
9SiC

SiC163
9SiC

6H

15R

15R 15R 6H 4H 6H 2050°C

6H

2000°C

2050°C

1900°C 1900°C

1800°C 1700°C

1700°C –10

–20 ppm

1600°C

–30

–10

–20 ppm

–30

Figure 5.1.8 Evolution of 29 Si MAS–NMR spectra of sample 163 and 171 as a function of the temperature (dwell duration 1 h). Top of β-SiC peak had been cut to observe polytypes signal.

Table 5.1.3 Chemical shift of 29 Si in ppm/TMS for crystallised amorphous silicon nitrides. Values within parentheses are full-width at half-height of each peak Authors

β-Si3 N4 (singlet)

α-Si3 N4 (doublet) equal intensities

Amorphous

Carduner et al. 1987, 1990 Harris et al. 1990

−48.7 (2.5) −48.5 (1.3)

−46.8 (1.8) −47.1

−46.4 (27)

Table 5.1.4

−48.9 (1.8) −49

29 Si average chemical shift of different silicon nitrides. Chemical shift measurements done on silane series (upper part) exhibit an average shift of −20 ppm comparatively to pure silicon carbide and silicon nitride. Consequently interpolation is done for intermediary environment of the Si (lower part)

Silane

Chemical shift (ppm/TMS)

(CH3 )4 Si (CH3 )3 SiN(CH3 )2 (CH3 )2 Si(N(CH3 )2 )2 (CH3 )Si(N(CH3 )2 )3 Si(N(CH3 )2 )4 Silicon environment in solid state SiC4 SiC3 N SiC2 N2 SiCN3 SiN4

0 6.9 to 5.0 −1.3 to 1.85 −16.3 to −17.5 −28.1 to −28.6 Average chemical shift (ppm/TMS) −20 −13 to −14 −21 to −22 −36 to −37 −46 to −49

©2003 Taylor & Francis

Seitz (1996)

5 0

Gérardin (1997)

–5 –10

–20 –25 –30 –35 –40

Verdecia (1998) α

–45

Carduner (1987, 1090)

 (ppm/TMS)

–15

–50 β 200

400

600

800 1000 1200 1400 1600 1800 2000 Temperature (°C)

Figure 5.1.9 Chemical shift of 29 Si in ppm/TMS for crystallised SiCx N4−x compounds as a function of the HTT of different ceramics from Seitz et al. (1996), Gérardin et al. (1997) and Verdecia et al. (1998). Symbols are
SiN4 , • SiCN3 and  SiC2 N2 ; highest temperature is for well-crystallised silicon nitrides (Carduner et al. 1987, 1990).

et al. 1998). Chemical shifts of such Si environments are dependent on the first and second sphere of coordinence. This one is able to contain hydrogen atoms for as-formed or low heat-treated materials and consequently the chemical shift is susceptible to modifications. Figure 5.1.9 summarises different observations. Series HMDS41 to HMDS44, heat-treated from 1000◦ C to 1600◦ C in N atmosphere during 1 h, have been analysed using 29 Si MAS–NMR. This method enables us to distinguish, through previous mentioned observations given in Figure 5.1.9, the evolution of the local structure around Si. Figures 5.1.10 and 5.1.11 give the general results of the survey. They are classified as a function of the initial C/Nini composition or of the heat-treatment temperature (HTT). Some remarks have to be made:

•

At HTT 1000◦ C, no important differences are observable. A broad massif centred around −50 ppm demonstrates an amorphous silicon nitride environment. Besides, a shoulder near −35 ppm can be attributed to SiCN3 , the importance of which decreases with initial C/N ratio. A small peak at −110 ppm corresponds to silica.

©2003 Taylor & Francis

(a)

(b)

1600°C/N2/1 h

1600°C/N2/1 h

1550°C 1500°C 1500°C

1400°C 1400°C

1000°C

1000°C

40

20

0 –20

–40 –60 ppm

40

–80 –100 –120

(c)

20

0 –20

–40 –60 ppm

–80 –100 –120

(d)

1600°C/N2/1 h 1600°C/N2/1 h

1550°C 1500°C 1500°C 1400°C 1400°C

1000°C 1000°C 40

20

0 –20

–40 –60 ppm

–80 –100 –120

40

20

0 –20

–40 –60 ppm

–80 –100 –120

Figure 5.1.10 Solid state 29 Si MAS–NMR spectra at 99.36 MHz, rotation frequency of 5 kHz, of HMDS powders heat-treated during 1 h in N atmosphere, as a function of the initial C/N (atomic number) composition, at increasing temperature. (a) HMDS41 (duty cycle, 500 s; 90◦ pulse, 5 µs; number of scans, 160, 160, 160, 12, 84); (C/N)ini = 0.95; (b) HMDS42 (duty cycle, 500 s; 90◦ pulse, 5 µs; number of scans, 160, 103, 160, 75); (C/N)ini = 0.73; (c) HMDS43 (duty cycle, 500 s; 90◦ pulse, 5 µs; number of scans, 144, 125, 140, 120, 81); (C/N)ini = 0.67; (d) HMDS44 (duty cycle, 500 s; 90◦ pulse, 5 µs; number of scans, 150, 160, 102, 40); (C/N)ini = 0.57.

©2003 Taylor & Francis

(a)

(b)

HMDS44

HMDS44

HMDS43

HMDS43

HMDS42

HMDS42

HMDS41

HMDS41

40

20

0 –20

–40 –60 ppm

–80 –100 –120

40

20

0 –20

–40 –60 ppm

–80 –100 –120

0 –20

–40 –60 ppm

–80 –100 –120

(d)

(c)

HMDS44

HMDS44

HMDS43

HMDS43

HMDS42

HMDS42

HMDS41

HMDS41

40

20

0 –20

–40 –60 ppm

–80 –100 –120

40

20

Figure 5.1.11 Solid state 29 Si MAS–NMR spectra of HMDS powders. Experimental conditions are given in Figure 5.1.10 caption. (a) 1000◦ C/N2 /1 h; (b) 1400◦ C/N2 /1 h; (c) 1500◦ C/N2 /1 h; (d) 1600◦ C/N2 /1 h.

©2003 Taylor & Francis

• • •

At HTT 1400◦ C, a silicon nitride broad peak (SiN4 and SiCN3 ) emerges from the massif. It appears as a shoulder near −20 ppm (SiC2 N2 ). X-ray diffraction patterns do not provide similar information, owing to the amorphous organisation of the material. At HTT 1500◦ C, the structure of the massif is more evident. Three more or less well-resolved peaks are observed which are respectively attributed to SiN4 , SiCN3 and SiC2 N2 . A shoulder near −18 ppm seems to appear which can be silicon carbide. At HTT 1600◦ C, crystallised silicon nitride (α and β) are observed, the proportion of which increases as initial C/N decreases. A silicon carbide massif (SiC4 , SiC3 N) is clearly separated from this one.

Two significant samples, HMDS42 and HMDS44, are analysed in more detail using least square decomposition method, assuming different occupations sites as given in Figure 5.1.9. Results are shown on Figures 5.1.12 and 5.1.13. Table 5.1.5 gives the relative proportions of the different sites. Neither SiC4 nor SiC2 N2 are observed till HTT is 1000◦ C. A mixture of such sites is observed at a higher temperature. From HTT 1550◦ C, crystallised SiN4 (α and β) sites are clearly identified. In the silicon carbide broad line, SiC4 and SiC3 N sites are present. The relative proportions of crystallised silicon nitride clearly depend on the initial C/Nini composition (C/Nini = 0.73 and 0.57, respectively, for HMDS42 and HMDS44). Such different evolutions are shown on the chemical composition (Figure 5.1.14). An important remark concerns the duty cycle, that is, the time between two accumulations, to obtain reliable spectra. This time has to be of the order of five times that of the spin-lattice

(a)

(b)

SiN4

SiN4 SiCN3

SiCN3

1000°C

1000°C

SiC2N2

SiCN3

SiN4

SiCN3

SiN4

SiC2N2 1400°C

1400°C SiCN3 SiC2N2 & SiC4

SiCN3 SiN4

SiN4

SiC2N2 & SiC4

1500°C 1500°C

20

0

–20 –40 ppm (TMS)

–60

–80

20

0

–20 –40 ppm (TMS)

–60

–80

Figure 5.1.12 Solid state 29 Si MAS–NMR spectra at 99.36 MHz, rotation frequency of 5 kHz, of (a) HMDS42 and (b) HMDS44 heat-treated at 1000◦ C, 1400◦ C and 1500◦ C during 1 h in N atmosphere and decomposition, using a least square method, into different assumed Si environments (El Kortobi 1998).

©2003 Taylor & Francis

(a)

(b) SiC

SiC2N2

SiC SiC3N

SiC3N

SiC2N2

1550°C

1550°C

Si3N4 (,
)

SiC

SiC

Si3N4 (,
)

SiC3N

SiC2N2

SiC3N

1600°C

1600°C

SiC SiC

SiC3N 1700°C 20

SiC2N2

SiC3N 1700°C

0

–20 ppm (TMS)

–40

–60

20

0

–20

–40

–60

ppm (TMS)

Figure 5.1.13 Solid state 29 Si MAS–NMR spectra at 99.36 MHz, rotation frequency of 5 kHz, of (a) HMDS42 and (b) HMDS44 heat-treated at 1550◦ C, 1600◦ C and 1700◦ C during 1 h in N atmosphere and decomposition, using a least square method, into different assumed Si environments (El Kortobi 1998).

relaxation times of the species to observe. Figure 5.1.15 shows spectra obtained for different values of such time. One notices that the relative apparent proportion of silicon nitride is only significant for temperatures higher than 1550◦ C, which shows that the two massifs are associated with relatively short and long spin-lattice relaxation times. Different authors, using MAS–NMR, had similarly used duty cycle as long as 30 mn to obtain reliable spectra (Dybowski et al. 1996). In the case considered here, it is necessary to reach 8000 s to estimate the relative proportion of silicon nitride in the different specimens. Amorphous silicon carbide massif relaxes more rapidly as shown in Figure 5.1.16. In Chapter 5.2, electron paramagnetic resonance (EPR) measurements show that the materials contain an important amount of paramagnetic centres. Besides, the study of the magnetisation as a function of the temperature as in the example on HMDS44 heat-treated at 1550◦ C, shows a Curie law dependence. This is in favour of localised paramagnetic centre organisation. As the above observation concerning the difference of spin-lattice relaxation times between silicon nitride and other species, it is obvious to consider that such centres are inside the amorphous silicon carbide structure. Same samples, HMDS42 and HMDS44, have been analysed using 13 C MAS–NMR (Figure 5.1.17). The useful CP/MAS–NMR is useless in that temperature range, due to the lack of proton in the material. The difficulty in acquiring a good signal over noise spectra is related to the low-frequency measurement. This had been done to prevent side bands coming from a large CSA of the aromatic carbons. Different comments have to be made: •

At HTT 1000◦ C, some differences are observable. A broad massif centred around −130 ppm is attributed to aromatic sp2 and oxygenated carbons. This is generally

©2003 Taylor & Francis

Table 5.1.5 H T T (◦ C)

HMDS42 1000 1400 1500 1550 1600 1700 HMDS44 1000 1400 1500 1550 1600 1700

29

Si chemical shift δ in ppm/TMS, obtained using a least square fitting method for HMDS42 and HMDS44 samples. Gaussian fit shape assumed for a theoretical decomposition of the experimental spectra. ! is the Gaussian quadratic square deviation of each specie assumed SiN4

SiCN3

SiC2 N2

SiC3 N

SiC4

δ (ppm)

! (ppm)

%

δ (ppm)

! (ppm)

%

δ (ppm)

! (ppm)

%

δ (ppm)

! (ppm)

%

δ (ppm)

! (ppm)

%

−53.6 −50 −48 / / /

18 14 14 / / /

67 54 34 9 13 10

−35 −34.6 −32 / / /

20 13 15 / / /

33 30 31 / / /

/ −21 / −21 / /

/ 14 / 8 / /

/ 16 / 26 / /

/ / /

/ / / 10 13 5

/ / / 34.5 31.1 12.6

/ / −14.5 −17 −18 −17

/ / 15 8 10 6

/ / 35 31.5 56.5 74.4

−53.6 −50 −53 / / /

17 14 12 / / /

80 67 40 48 40 42

−37 −36 −37 / / /

17 14 14 / / /

20 16 36 / / /

/ −23 / −22 −23 −21

/ 14 / 7 5 6

/ 17 / 5.5 6 3.2

/ / /

/ / / 9 8.5 8

/ / / 12.3 11.4 14.1

/ / −20 −17 −17 −17

/ / 13 10 8 6

/ / 24 34.2 42.6 40.8

©2003 Taylor & Francis

−9 −6 −11

−7 −10 −9

HMDS40 HMDS42 HMDS44 HMDS45 Wakai (1990) −−−−> HTT increase

N 0.0 1.0 0.1 0.9 0.2

0.8

Ni t

rog en

0.3

0.7

0.4

0.6

0.5

0.5

0.6 0.7

0.4 45

44

42

40

0.3

0.8

0.2

0.9 1.0 0.0 C

0.1 0.0 0.1

0.2

0.3

0.4

0.5

0.6

Carbon

0.7

0.8

0.9

1.0

Si

Figure 5.1.14 Atomic chemical composition Siy Cx Nz (x + y + z = 1) of HMDS42 and HMDS44 as a function of heat treatment temperature. Arrows indicate evolution from as-formed to highest treatment temperature.

• •

considered as ‘free carbon’. Besides, for HMDS42, a small peak near −20 ppm can be attributed to sp3 Si bonded C. From HTT 1400◦ C, the amount of Si bonded C increases, which is in accordance with the 29 Si MAS–NMR observation. Besides, the existence of other environments such as CSix Ny are possible. At HTT 1550◦ C, the line around −20 ppm, related to silicon carbide, is more important for both samples. Although the difference is noticeable concerning the ‘free carbon’ still present in the richer C HMDS42 sample than in HMDS44, were it is now absent.

More generally, similar determinations are of interest to characterise silicon carbide fibres and consequently their mechanical properties (Hommel et al. 1990; Laffon et al. 1989). Some comparisons can be done between the atomic chemical compositions and 29 Si NMR determinations. From Table 5.1.5, with the relative proportions of SiN4 (p), SiCN3 (q), SiC2 N2 (r), SiC3 N(s) and SiC4 (t), it is possible to calculate the average atomic compositions Siy Cx Nz (x + y + z = 1) as a function of HTT and to plot Figure 5.1.18, using the following formula: 1 2 3 4 x =q +r +s +t y 4 4 4 4 z 4 3 2 1 =p +q +r +s y 3 3 3 3

(5.2)

The above values are listed in Table 5.1.6. Such obtained plots show a convergence of the two types of analysis at high heat temperature treatment of both samples. The divergence observed at low HTT originates from

©2003 Taylor & Francis

HMDS44 1550°C

HMDS44 1700°C

HMDS44 1600°C

D1 = 8000 s

D1 = 8000 s

D1 = 8000 s

D1 = 500 s

D1 = 500 s

D1 = 500 s

20 10

0 –10 –20 –30 –40 –50 –60 –70 –80

20

10

0

–10 –20

–30 –40 –50 –60

20

10

0 –10 –20 –30 –40 –50 –60 ppm

ppm

ppm HMDS42 1600°C 20 10

0 –10 –20 –30 –40 –50 –60

D1 = 8000 s

D1 = 500 s

20 10

0 –10 –20 –30 –40 –50 –60 ppm

HMDS42 1700°C 20 10

0 –10 –20 –30 –40 –50 –60

D1 = 8000 s

D1 = 500 s

20 10

0 –10 –20 –30 –40 –50 –60 ppm

Figure 5.1.15 Solid state 29 Si MAS–NMR spectra obtained at different duty cycle (D1) values on HMDS44 and HMDS42, at different heat treatment temperatures. Spectra are adjusted so that the maximum of each spectrum is similar to the others. Signal over noise depends on the number of scans and consequently of the total time of accumulation, which is different for some spectra (El Kortobi 1998).

an underestimation of C content. Those not being bonded to silicon C atoms are set in the amorphous ‘free carbon’ as it had been discussed above (Figure 5.1.17). This effect is not important for sample heat-treated at temperature higher than 1550◦ C, as it can be observed on the points corresponding to NMR measurements. They are in the vicinity of those in the chemical analysis. From the C content values obtained for chemical analysis and NMR, it is possible to estimate the proportion of ‘free carbon’ in total C. This calculation is only valid, within the experimental errors, in the temperature range 1000–1550◦ C (Figure 5.1.19). 5.1.3.2

Thermal evolution from NMR measurements

This evolution can be roughly decomposed in three steps as follows: First step: HT T inferior to 1000◦ C This domain, from as-prepared powder till almost end of mineralisation, is mainly discussed in Chapter 2. The most efficient methods of

©2003 Taylor & Francis

HMDS44 1600°C

D1 = 8000 s

D1 = 4000 s

D1 = 1000 s

20

10

0

–10 –20 –30 –40 –50 –60 ppm

Figure 5.1.16 Solid state 29 Si MAS–NMR spectra obtained at different duty cycle (D1) values on HMDS44 heat-treated at 1600◦ C (LB = 30 Hz, NS = 16) (El Kortobi 1998).

characterisation are in particular TGA and IR. It is characterised by emission of H2 O, NH3 and CH4 . It results in the formation of Si N with ammoniac (Gérardin et al. 1997; Bowen 1980) and Si C bonds with methane (Gérardin et al. 1997, Hasegawa and Okumura 1983) emissions. Second step: 1000–1500◦ C As seen from 29 Si MAS–NMR, it starts with a rich nitrogen (N) to Si environment (SiN4 , SiCN3 ) depending on the initial C/Nini concentration. Besides, 13 C MAS–NMR demonstrate the presence of C not bonded to Si, named ‘free carbon’. The proportion of such C decreases considerably till 1500◦ C. Parallelly, it is observed as an increase of C to Si environment. Third step: 1550–1700◦ C The crystallisation of the material is observed, with important loss in weight due to the departure of N, consecutive to a carbo-reduction. Demixtion occurs with the appearance of well-crystallised α − β SiN4 and of amorphous silicon carbides (SiC4 , SiC3 N and SiC2 N2 ).

5.1.4 Pre-alloyed Si/C/N/Al/Y nanopowders In Chapter 7 explanations are given on the interest to use additives for the sintering of silicon nitride based materials. The objective here is to characterise nanosized ready-tosinter powders including additives. Two types of additives have been incorporated: alumina (Al2 O3 ) and yttria (Y2 O3 ).

©2003 Taylor & Francis

HMDS42

HMDS44 CSi4

CSi4

Carbon sp2

1550°C

1550°C

1500°C

1500°C

1400°C

1400°C

1000°C

1000°C

250 200 150 100 50 ppm (TMS)

0

250 200 150 100 50 ppm (TMS)

0

Figure 5.1.17 Solid state 13 C MAS–NMR spectra at 25.15 MHz, rotation frequency of 5 kHz, of HMDS42 and HMDS44 at different heat treatment temperatures.

0.0 0.1 0.2

N

1.0 0.9 0.8 0.7

0.4

0.6

0.5

0.6

on

0.5

lic Si

rog en

0.3 Nit

HMDS44 HMDS42 HTT increase NMR HMDS44 NMR HMDS42

0.4

0.7

0.3

0.8

0.2

0.9 1.0 0.0

C

0.1 0.0 0.1

0.2

0.3

0.4

0.5 0.6 Carbon

0.7

0.8

0.9

1.0

Si

Figure 5.1.18 Atomic NMR and chemical compositions Siy Cx Nz (x + y + z = 1) of HMDS42 and HMDS44 as a function of heat-treated temperature. Arrows indicate evolution from as-formed to highest treatment temperature.

©2003 Taylor & Francis

Table 5.1.6 Atomic NMR composition Cx Siy Nz (x + y + z = 1) of HMDS42 and HMDS44 as a function of heat treatment temperature H T T (◦ C)

C

Si

N

HMDS42 1000 1400 1500 1550 1600 1700 HMDS44 1000 1400 1500 1550 1600 1700

0.03896 0.06987 0.15315 0.33175 0.38462 0.42157

0.4329 0.43668 0.45045 0.47393 0.48077 0.4902

0.52814 0.49345 0.3964 0.19431 0.13462 0.08824

0.02155 0.05652 0.12796 0.20354 0.24658 0.23767

0.43103 0.43478 0.47393 0.44248 0.45662 0.44843

0.54741 0.5087 0.3981 0.35398 0.2968 0.3139

1.0

Relative proportion of ‘free carbon’

0.8

0.6

0.4 HMDS42 HMDS44

0.2 0.0

1000

1100

1200

1300

1400

1500

1600

HTT (°C)

Figure 5.1.19 ‘Free carbon’ evolution as a function of heat treatment temperature. Values are obtained owing to the formula: (xCA − xNMR )/xCA where xCA , xNMR are those of the chemical analysis and NMR, respectively.

5.1.4.1

Free carbon effect

Samples HSAl03 and HSAl05, heat-treated from 1000◦ C to 1600◦ C in N atmosphere during 1 h, have been analysed using 29 Si MAS–NMR (Figure 5.1.20). The following observations can be made: • •

From as-formed to HTT 1000◦ C, no important differences are observed. A broad peak around −20 ppm shows different environments of Si (SiC3 N, SiC2 N2 and/or SiC4 ). At HTT 1400◦ C, a line narrowing is observed. Only SiC2 N2 and/or SiC4 environments remain.

©2003 Taylor & Francis

(a)

(b) SiC

SiC

1600°C/N2 /1h

1600°C/N2 /1h

1400°C

1400°C

1000°C

1000°C

SiC3N As-formed

100

50

SiCN3

0 –50 ppm (TMS)

SiC3N As-formed

SiC4 SiC2N2 –100

100

50

SiCN3 SiC4 SiC2N2 0 –50 ppm (TMS)

–100

Figure 5.1.20 Solid state 29 Si MAS–NMR spectra at 99.36 MHz, MAS speed of 5 kHz, of (a) HSAl03 and (b) HSAl05 at different heat treatment temperatures. At 1600◦ C, no silicon nitride remains; compare to Table 3.1.3 (El Kortobi 1998).

•

At HTT 1600◦ C, the narrowing is more important. A structure is perceptible, which can be attributed to α and β silicon carbide. 13 C

MAS–NMR is able to improve such results. Only as-formed samples can be analysed using CP and MAS. The sites are CH3 Si and Si CH2 Si in the range 0–5 ppm. • •

At HTT 1000◦ C, as hydrogen is expulsed with heat treatment, 13 C MAS–NMR sp3 and sp2 C are observed. From HTT 1400◦ C to 1600◦ C, the ‘free carbon’ massif disappears progressively. The line around 24 ppm is proportionally improved. This confirms the presence of partially crystallised silicon carbide (Figure 5.1.21). 27 Al

MAS–NMR shows three aluminium environments for as-formed materials. Tetra, penta and octahedral environments are observed (Figure 5.1.22). • •

At HTT 1000◦ C, such three sites are still observable, with a significant increase of the octahedral environment. From HTT 1400◦ C to 1600◦ C, besides tetra and octahedral sites, there is a broad massif at lower field, around 90 ppm. This can be attributed, owing to measurements done on defined samples (AlON and SiAlON) (Figure 5.1.23), to aluminium oxynitrides AlOx N4−x , with 0 < x < 4 (Smith 1992; Mackenzie et al. 1994; Fitzgerald et al. 1994). At HTT 1400◦ C, this massif contains AlO3 N and AlO2 N2 (96 ppm), AlON3 (106 ppm) and AlN4 (114 ppm). At HTT 1600◦ C, the peak corresponding to AlN4 is easier to observe (Figure 5.1.24).

©2003 Taylor & Francis

sp3 carbon

sp2 carbon (free carbon)

1600°C

MAS

1400°C

24

MAS

1000°C

MAS 2

130

As-formed

CP/MAS

250 200 150 100

50

0

–50 –100

ppm (TMS)

Figure 5.1.21 Solid state 13 C MAS–NMR spectra at 25.15 MHz, rotation frequency of 5 kHz, of HSAl03 at different heat treatment temperatures (El Kortobi 1998). (a)

(b)

AlO6

AlN3O

AlN3O

AlO6

AlN4

AlN4

1600°C/N2 /1 h

AlNO3 AlN2O2

AlNO3 1400°C/N2 /1 h

Al04 1000°C/N2 /1 h AlO5

As-formed

200 100 ppm

0 –100 –200

(Al(H2O)63+)

200 100

0 –100 –200

ppm (Al(H2O)63+)

Figure 5.1.22 Solid state 27 Al MAS–NMR spectrum at 130.32 MHz of (a) HSAl03 and (b) HSAl05. MAS speed 12 kHz, 1 µs pulse, duty cycle 1 s, number of scans 600 (El Kortobi 1998).

©2003 Taylor & Francis

60–70

YSiAlON glass

YSiAlO glass

AlN4105

Conventional sintered sample

γ-AlON

300

200

100

0

–100 –200

–300

ppm (Al(H2O)63+)

Figure 5.1.23 Solid state 27 Al MAS–NMR spectrum at 130.32 MHz of different reference samples. MAS speed 12 kHz, 1 µs pulse, duty cycle 1s, number of scans 600. At 1600◦ C, N is bonded to aluminium; compare to Table 3.1.3.

So, the following remarks can be made: • •

Al O and A N bonds are observed. N, initially bonded to Si as SiC3 N and SiC2 N2, is finally bonded to aluminium oxynitrides AlOx N4−x .

In conclusion, the study of the local environment of 13 C,27 Al and 29 Si nuclei, enables us to understand why, within the proportions of reactants used for the synthesis, such powders did not produce a compound suitable for sintering (Chapter 7). The large amount of ‘free carbon’ favours the nitridation of aluminium owing to the reaction: Al2 O3 + (3 − 2x)Cfree + 2 Si/C/N4−x → 2 AlOx N4−x + (3 − 2x)CO + 2 SiC This aluminium nitridation is responsible for the N loss in the Si environment. A new series of preparation is done with hydrogen and ammonia added on. This is to obtain the trapping of reactive free radical • CH3 . The 27 Al spectra show that the aluminium nitridation is noticeably decreased (Figure 5.1.24). It presents an intense line at 60 ppm, attributed to AlO4 , sharper than that observed in a glass or conventional sintered sample. This line corresponds to a well-organised tetrahedral specie, only surrounded by oxygen atoms. This should be an aluminium solid solution in β-SiAlON. Similarly, the 29 Si spectra demonstrate the presence of silicon nitride (Figure 5.1.25). The comparison between samples prepared without (HSAl07) and with (HSAl09) ammonia (Figure 5.1.26) shows clearly this

©2003 Taylor & Francis

AlO6

AlO4

1600°C/N2 /1 h

HSAlY17

HSAl15

1600°C AlN4

HSAl14

1600°C

HSAl12

1600°C

HSAl09

1600°C

HSAl12

As-formed

200 150

100 ppm

50

0

–50 –100

(Al(H2O)63+)

Figure 5.1.24 Solid state 27 Al MAS–NMR spectrum at 130.32 MHz of HSAl12 to HSAlY17. MAS speed 12 kHz, 1 µs pulse, duty cycle 1 s, number of scans 600.

phenomenon not only for Si but also for aluminium. This support the analysis done above on the role of ‘free carbon’ effect which is clearly observable on 13 C spectra (Figure 5.1.26).

5.1.4.2

Influence of hot pressing

A comparison can be done between hot pressed and heat-treated nanopowders. As examples HSAl09 29 Si and 27 Al NMR measurements are shown in Figures 5.1.27 and 5.1.28. As the hot pressing had been done at a higher temperature than that of the nanopowder, silicon carbide has almost disappeared. The structured peak, corresponds to different silicon nitrides. As mentioned in Table 5.1.3, α- and β-Si3 N4 respectively present a doublet and a singlet lines. Carduner et al. (1987) measured line-width between 1.5 and 2.5 ppm (fullwidth at half-height) for well-crystallised material. Although the spectra are in accordance with these measurements, the decomposition, in order to have the relative proportion of β/(α + β), is questionable. The powder heat treated at 1600◦ C is able to contain some amorphous part, which explains the higher intensity at −47 ppm. For the sintered powder, the β proportion estimation by X-ray diffraction is 30% (Table 7.6), which is qualitatively in accordance with the 29 Si NMR observation. 27 Al NMR spectra (Figure 5.1.27) show the Al O N broad line with the typical peaks or x y z shoulder characteristics of AlN4 , AlO4 and AlO6 environments. A significant difference is observed with the relative proportions of the different species from one sample to another. This has to be compared with Figure 5.1.22 where alumina is bonded to variable environments containing N.

©2003 Taylor & Francis

Si3N4 Free

carbon

SiC

SiC HSAl15

HSAl15

HSAl14

20

10

HSAl14

0 –10 –20 –30 –40 –50 –60 –70 –80 ppm

200 180 160 140 120 100 80 60 40 20 ppm

0 –20

Figure 5.1.25 Solid state 29 Si and 13 C MAS–NMR spectra at 99.36 MHz and 125.76 MHz of HSAl14, HSAl15 heat treated at 1600◦ C/N2 /1 h. 29 Si MAS speed 5 kHz, 5.2 µs pulse, duty cycle 500 s, number of scans 38, 4; 13 C MAS speed 5 kHz, 5 µs pulse, duty cycle 500 s, number of scans 244. 1600°C/N2 /1 h

1600°C/N2 /1 h

Si3N4

HSAl09 D1 = 8000 s

AlO4 HSAl09

HSAl09 D1 = 500 s

SiC

AlO6

AlN4

HSAl07

HSAl07 D1 = 500 s

0

–20

–40 ppm

–60

–80

250 200 150 100 50

0 –50 –100 –150 –200

ppm

Figure 5.1.26 Solid state 29 Si and 27 Al MAS–NMR spectra at 99.36 MHz and 130.32 MHz of HSAl07, HSAl09 heat treated at 1600◦ C/N2 /1 h. 29 Si MAS speed 5 kHz, 5.2 µs pulse, duty cycle 500, 8000 s, number of scans 38, 121 and 4; 27 Al MAS speed 14 kHz. 1.2 and 1 µs pulse, duty cycle 1 s, number of scans 3600. An underestimation of the amount silicon nitride is observed when a too short duty cycle is used (see Figure 5.1.15).

©2003 Taylor & Francis

HSAl09 D1 = 8000 s

Si3N4

Sintered powder 1700°C/35 Mpa/20°mn/N2

Powder heat treated 1600°C/N2 /1 h − and
-Si3N4

Sintered powder 1700°C/35 MPa/20°mn/N2

-and
-Si3N4

α-Si3N4 + amorphous phase

Si3N4

Powder heat treated 1600°C/N2 /1h SiC

–40 –42 –44 –46 –48 –50 –52 –54 ppm

20

10

α-Si3N4 + amorphous phase

–40 –42 –44 –46 –48 –50 –52 –54 ppm

0 –10 –20 –30 –40 –50 –60 –70 –80 ppm

Figure 5.1.27 Solid state 29 Si MAS–NMR spectra at 99.36 MHz. MAS speed 5 kHz, 5.5 µs, duty cycle 8000 s, number of scans 10. Comparison between sintered and heat treated HSAl09 nanopowder. HSAl09

Sintered powder 1700°C/35 MPa/20°mn/N2

AlO4

AlN4

AlO6 Powder heat treated 1600°C/N2 /1 h

200

100

0 ppm

–100

–200

Figure 5.1.28 Solid state 27 Al MAS–NMR spectrum at 130.32 MHz. MAS speed 14 kHz, 1.2 µs pulse, duty cycle 1 s, number of scans 600. Comparison between sintered and heat treated HSAl09 nanopowder.

5.1.5 Conclusion The analyses of different nanopowders, as-prepared and heat treated at different temperature, show the efficiency of the solid-state NMR, particularly for not well-crystallised samples. The possibility of obtaining cross informations from different nuclei, as 13 C,27 Al and 29 Si and eventually proton via CP will improve the structural analysis. This enables us to modify synthesis conditions as demonstrated on pre-alloyed Si/C/N/Al/Y nanopowders.

©2003 Taylor & Francis

References Apperley, D. C. Marshall, G. L. and Thompson, D. P. (1991) J. Am. Ceram. Soc., 74(4), 777. Bloch, F. (1958) Phys. Rev. 111, 841. Bowen, H. K. (1980) Mater. Sci., 44, 1. Carduner, K. R., Blackwell, C. S., Hammond, W. B., Reidinger, F. and Hatfield, G. R. (1990) J. Am. Chem. Soc., 112, 4676. Carduner, K. R., Carter III, R. O., Milberg, M. E. and Crosbie, G. M. (1987) Anal. Chem., 59, 2794. Dando, N. R. and Tadayyoni, M. A. (1990) J. Am. Ceram. Soc., 73(8), 2242. Dybowsky, C., Gaffney, E. J., Sayir, A. and Rabinowitz, M. (1996) Colloids. Surf., A118, 171. El Kortobi, Youssef (1998) Thesis, Université P. M. Curie, Paris, France, http://corail.sudoc.abes.fr80 & Sudoc-Catalogue no 053764552. Engelhardt, G. and Michel, D. (1987) High-resolution Solid-state NMR of Silicates and Zeolites, J. Wiley & Sons, New York. Finlay, G. R., Harman, J. S., Ricardson, M. F. and Williams, B. L. (1985) J. Chem. Soc., Chem. Commun., 159. Fitzgerald, J. J., Kohl, S. D. and Piedra, G. (1994) Chem. Mater., 6, 1915. Fyfe, C. A. (1983) Solid-state NMR for Chemists, C.F.C. Press Guelph. Gérardin, C., Taulelle, F. and Bahloul, D. (1997) J. Mater. Chem., 7(1), 117. Guth, J. R. and Petuskey, W. T. (1987) J. Phys. Chem., 91(20), 5361. Harris, R. K., Leach, M. J. and Thompson, D. P. (1990) Chem. Mater., 2, 320. Hartman, J. S., Richardson, M. F., Sherriff, B. and Winsborrow, B. G. (1987) J. Am. Soc., 109, 6059. Hasegawa, Y. and Okamura, K. (1983) J. Mater. Sci., 18, 3633. Hockey, J. A. (1965) Chem. Ind. (London) 57. Hommel, H., Miquel, J. L. and Legrand, A. P. (1990) L’Industrie Céramique, 849, 5. Laffon, C., Flank, A. M., Lagarde, P., Laridjani, M., Hagege, R., Olry, P., Cotteret, J., Dixmier, J., Miquel, J. L., Hommel, H. and Legrand, A. P. (1989) J. Mater. Sci., 24, 1503. Lippmaa, E. and Samoson, A. (1982) Bruker Rep., 1, 6. Lippmaa, E., Magi, M., Samoson, A., Engelhardt, G. and Grimmer, A. R. (1982) J. Am. Chem. Soc., 102, 4899. Mackenzie, K. J. D., Meinhold, R. H., White, G. V., Sheppard, C. M. and Sherriff, B. L. (1994) J. Mater. Sci., 29, 2611. Marsman H. (1981) NMR Basic Principles and Progress, 17 (eds P. Diehl, E. Fluck and R. Kosfeld), Springer-Verlag Berlin, p. 65. Müller, K. (1999) Precursor-derived Ceramics (eds J. Bill, F. Wakai and F. Aldinger), Wiley-VCH. Richardson, M. F., Hartman, J. S., Guo, D. and Winsborrow, B. G. (1992) Chem. Mater., 4, 318. Samoson, A. and Lippmaa, E. (1983) Phys. Rev., B28, 6567. Schmidt, W. R., Interrante, L. V., Doremus, R. H., Trout, T. K., Marchetti, P. S. and Maciel, G. E. (1991) Chem. Mater., 3, 257. Seitz, J., Bill, J., Egger, N. and Aldinger, F. (1996) J. Eur. Ceram. Soc., 16, 885. Smith, M. E. (1992) J. Phys. Chem., 96, 1444. Tateyama, H., Noma, H., Adashi, Y. and Komatsu, M. (1997) Chem. Mater., 9, 766. Templin, M., Friedrich, U., Wiesner, U. and Spiess, H. W. (1999) Precursor-derived Ceramics (eds J. Bill, F. Wakai and F. Aldinger), Wiley-VCH. Tougne, P., Hommel, H., Legrand, A. P., Herlin, N., Luce, M. and Cauchetier, M. (1993) Diamond Relat. Mater., 2, 486. Tuel, A., Hommel, H., Legrand, A. P., Chevallier, Y. and Morawski, J.C. (1990) Colloids. Surf., 45, 413. Vercedia, G., O’Brien, K. L., Schmidt, W. R. and Apple, T. M. (1998) Chem. Mater., 10, 1003. Wagner, G. W., Na, B. -K. and Vannice, M. A. (1989) J. Phys. Chem., 93, 5061. Wakai, F., Kodama, Y., Sakaguchi, S., Murayama, N., Izaki, K. and Niihara, K. (1990) Lett. Nature, 344, 421.

©2003 Taylor & Francis

5.2 Paramagnetic defect states in silicon-based nanopowders Abdelhadi Kassiba and Stéphane Charpentier

5.2.1 Introduction Electron paramagnetic resonance (EPR) technique is a suitable tool in characterising the paramagnetic defects in the Si/C/N nanopowders. The EPR experiments are carried out both following the annealing temperature (realised separately up to 1800◦ C) and the specimen temperature (in situ from 4 to 1000 K). Different EPR frequency bands (4–95 GHz) and numerical EPR spectrum analyses are needed for an understanding of the defect nature and properties. However, when mixed compositions and structures coexist as in (Si/C/N) networks, the overlapping of the resonance lines leads to structureless EPR spectra. In such materials, it is possible to assess the concentration of paramagnetic defects and their properties such as being localised or delocalised. More quantitative analyses can be achieved on more homogeneous materials (Si/C) where structured EPR spectra are obtained. Precise insights on the paramagnetic defects are made and correlated with the material composition, structure and physical properties. The first set of investigations is devoted to SiC nanopowders synthesised by a laser pyrolysis of a gaseous mixture (SiH4 , C2 H2 ). The batches, referred as SiC242, SiC163 and SiC177, differ by the residence time of the reactants in the laser beam and by the gaseous precursor flux rate [SiH4 ]/[C2 H2 ]. The defect concentration is relatively high in all the samples (NS ≈ 1020 spin g−1 ) and their nature seems correlated with the coexisting SiC polytypes. A consistent numerical analysis of the EPR spectra, over several EPR frequencies, was developed in order to identify the paramagnetic centres. Whatever the annealing stage and ˜ tensors coexist the EPR frequency band, three paramagnetic defects with well-defined (g, ˜ A) with contributions dependent on the annealing stages. Delocalised centres, thermally induced by varying the sample temperature, are also evidenced in these nanopowders. They would contribute to the significant electric conduction induced by annealing at Ta ≥ 1500◦ C. The second deals with Si/C/N materials (referred as HMDS40–45) where the coexisting structures (SiC, Si3 N4 ) and the free carbon (C) content give rise to a large unpaired spin concentration brought by two types of paramagnetic centres. Their contributions depend on the synthesis conditions and on the annealing stages. We have shown that the rate of C on nitrogen (N) monitors the defect concentration and the annealing under N partial pressure (nitriding process) favours the creation of unpaired spins in the Si3 N4 -rich powders. The forthcoming EPR investigations focus on these two classes of silicon-based nanomaterials.

©2003 Taylor & Francis

5.2.2 Theoretical background Paramagnetic defects in these Si/C/N covalent networks, consist mainly in unpaired electrons. Under an external magnetic field B0 , each paramagnetic defect is described by an effective spin Hamiltonian which reads:  S · A˜ k · Ik Hˆ spin = β S · g˜ · B0 + k

where β represents the Bohr magneton, k refers to an interacting nucleus (nuclear spin Ik ) with hyperfine diagonal tensor A˜ k in a reference frame (xk , yk , zk ) centred on the (k) nuclei. The g˜ tensor, has a diagonal form in its principal (magnetic) frame (XY Z), with components in the order of ge = 2.0023 value; that is, the free electron Landé factor. In crystalline powders, the magnetic field B0 is defined by angular variables (θ0 , ϕ0 ) with respect to the magnetic axes (X, Y, Z). The local resonance fields Bres ; that is, the EPR line position, is expressed as follows: geff βBres = hνEPR = ge βBe  2 sin2 θ cos2 ϕ + g 2 sin2 θ sin2 ϕ geff = gZ2 cos2 θ0 + gX 0 0 0 0 Y where h is the Planck’s constant and νEPR = 9.5 GHz in X-band EPR measurements. Be is the resonance position of a free electron. The EPR signal of paramagnetic defects would have the shape of a continuous absorption background with singularities at the positions

ge Be Bres (i) = gi i=X,Y,Z However, in current EPR experiments the derivative of the absorption line is recorded. The small departure !gi = gi − ge , in a given (i = X, Y, Z) direction, results from a weak  through the perturbation formula:  · S) spin–orbit coupling (λ · L  |ψ0 |Li |ψn |2 !gi = 2λ − n E(0) − E(n) where |ψ0  and |ψn  represent respectively the ground and the excited electronic states of energies E(0) and E(n) . Quantitative analysis of the g˜ tensor components cannot be undertaken safely in these heterogeneous Si/C/N media. However, the evolution of !g shifts with particular thermal treatments that can be used to characterise the changes of the local environments around the involved paramagnetic centres. With regard to the hyperfine tensor components Aki , they are determined experimentally from the hyperfine line positions. A brief theoretical justification of these parameters can be established in the following simple case. When the electronic spin density on the nuclei exhibits an axial symmetry around the zk -axis, the components of the A˜ k tensor are: A(k)  = Aiso + 2Aaniso

and

A(k) ⊥ = Aiso − Aaniso

with Aiso =

  8  (k) 2 (0) πge βgn βn ϕ(s) 3

©2003 Taylor & Francis

and

Aaniso = ge gn ββn



3 cos2 θ − 1 r3

(k)  |ϕ(p)

Aiso represents the contact interaction parameter and Aaniso accounts for a dipolar interaction. (k) (k) ϕ(s) (r ) and ϕ(p) (r ) represent the orbital wave function of (s) and (p) types related to the (k) ion. (r, θ) are the radial and azimuthal coordinates of the unpaired electron in the reference frame centred on the nucleus. gn and βn represent respectively the nuclear g-factor and the nuclear Bohr magneton. The hyperfine lines are observed when the unpaired electron interacts with a nuclear spin in its near environment. For Si/C/N materials, the nuclear spin (I ) is brought by 13 C (I = 21 , natural abundance (na) = 1.1%), by 29 Si (I = 21 , na = 4.7%) or by 14 N (I = 1, na = 99.6%). The intensity and the number of the hyperfine lines can be used to determine the nature of the interacting nuclei. Moreover, the integrated intensity of the EPR signal is proportional to the paramagnetic defect concentration (Ns ) and to their magnetic susceptibility (χ ). The temperature dependence of χ informs on the localised or delocalised nature of the paramagnetic species. Indeed, when χ varies with temperature following the Curie law (χ = C/T ), the unpaired electrons are localised in the host material sites. In the case of delocalised unpaired electrons, χ is temperature independent and ‘Pauli-like’ paramagnetism is expected.

5.2.3 Experimental details EPR experiments are carried out on Brucker spectrometer with different frequency bands [S (4 GHz), X (10 GHz), K (24 GHz), Q (34 GHz) and W (95 GHz)]. Investigations at different temperatures (4–300 K) are conducted on an Oxford cryostat. The high unpaired spin concentration in the samples limits the microwave power below 2 mW and the modulation field to 0.5 G. Saturation phenomena and EPR line shape distortions are thus avoided. The absolute concentration (Ns ) of paramagnetic species is determined by a comparison with CuSO4 · 5H2 O single crystal (≈3·1021 spin g−1 ). This absolute Ns estimate is within a range of four times more or less than the real value.

5.2.4 Si/C powders Three Si/C batches (SiC163, SiC177 and SiC212) synthesised from gaseous precursors (Table 2.2) are investigated. Whatever the nanopowder series are, the EPR spectra have common features and similar electrical properties. So, the forthcoming report is limited to the SiC212 batch referred below as SiC212 1200–1800◦ C according to annealing temperature T = 1200–1800◦ C. The structural modifications, the powder composition and grain morphology are well known (Charpentier 1997, 1999 and 1999a). Figure 5.2.1(a) shows the room temperature EPR spectra at X-band in the SiC212 1200◦ C and 1650◦ C samples. High temperature measurements in the SiC212 1200◦ C sample up to 600◦ C under air did not show any change on the EPR spectrum features except for the obvious intensity decrease. The reversible behaviour of the EPR spectra, that is, before and after the oxidation rules out any influence of the oxidation process or thermal removal of the paramagnetic species. Furthermore, it seems that the defects are not connected with hydrogen which may be released by annealing at high temperatures. The defects are intrinsic to the material mainly caused by vacancies in the C or silicon (Si) sites of the coexisting C arrangements and SiC polytypes. Low temperature measurements are needed in order to assess the localised or delocalised nature of the defects. In the SiC212 1200◦ C sample, the X-band EPR spectrum is drastically

©2003 Taylor & Francis

SiC212 T = 300 K

------ 1200°C 1650°C

Intensity (a.u.)

(a)

3380

SiC212 T = 7 K

3440

3460

------ 1200°C 1650°C

Intensity (a.u.)

(b)

3400 3420 Magnetic field (G)

3380

(c)

3400 3420 Magnetic field (G)

3440

3460

Intensity (a.u.)

Experimental

DI

Fit

DII

DIII

33,460

33,480 33,500 33,520 Magnetic field (G)

33,540

Figure 5.2.1 (a) Typical X-band EPR spectra in SiC212 nanomaterials annealed at 1200◦ C and 1650◦ C. (b) Evolution of the SiC212 1200◦ C EPR spectrum with the temperature. (c) Experimental and calculated W-band EPR spectra in the SiC212 1200◦ C sample. The contributions from the different paramagnetic centres are also shown (relative contributions to the EPR spectrum DI: 21%; DII 49%; DIII 30%).

©2003 Taylor & Francis

1.00

(a)

SiC212 1200°C 0.95

I ×T (a.u.)

0.90 0.85 0.80 0.75 0.70

0

50

100

150

200

250

300

Temperature (K) 10

(b)

SiC212 1600°C

I × T (a.u.)

8

6

4

2

0

50

100

150

200

250

300

Temperature (K)

Figure 5.2.2 Evolution with the temperature of the EPR signal integrated intensity weighted by T : (a) in the SiC212 1200◦ C sample and (b) in the SiC212 1600◦ C sample.

broadened by lowering the temperature (Figure 5.2.1(b)). This is understood by the possible defect motion which is frozen at low temperature. A double integration of the EPR signal allows the determination of the overall intensity which is proportional to the spin susceptibility χ (Figure 5.2.2). A departure of χ (T ) from the Curie law above 150 K, also observed in the SiC212 1650◦ C sample (Figure 5.2.2), indicates the presence of delocalised paramagnetic species. Furthermore, the peak-shaped behaviour of the product χ (T ) × T around 30 K is consistent with a localisation of the defects and correlated unpaired spins. Consistent numerical analyses of the EPR spectra are hereafter conducted in order to determine the paramagnetic defect features in these SiC nanopowders. The paramagnetic centres are S = 1/2-defects assumed to interact with p equivalent nuclei (nuclear spin

©2003 Taylor & Francis

I = 21 ). Under an applied magnetic field B0 , the spin Hamiltonian is given by: Hˆ spin = β B0 g˜ S + SA˜ I A˜ represents the hyperfine coupling tensor between the electronic and nuclear spins. The magnetic dipole transition intensities and their resonance positions are determined numerically from Hˆ spin . The basis is limited to |mS , mI  states with mS , mI = ± 21 due to the multiplicity of electronic and nuclear spins (S = I = 21 ) in SiC materials. On the other hand, the na (=0.047) of the 29 Si and (na = 0.01) for 13 C, giving rise to a hyperfine coupling, impose the EPR spectrum of a given paramagnetic defect to be a superposition of the two contributions. One is related to a spectrum with a hyperfine interaction, weighted by the p × na, superimposed to a spectrum without the hyperfine interaction as for 28 Si and 12 C nuclei weighted by (100 − p × na). A homemade program for EPR spectrum adjustments in crystalline powders was developed. All the magnetic field orientations (θ0 , ϕ0 ) are allowed (0 ≤ θ0 ≤ π/2, 0 ≤ ϕ0 ≤ π ) due to the random orientations of the crystalline sites in the SiC powders. The calculated resonance positions are used with suitable line widths in order to adjust the experimental EPR spectra. Whatever the annealing temperature or the operating νEPR frequency bands are, the EPR ˜ spectra are accounted by using three paramagnetic defects with the spectroscopic data (g, ˜ A) summarised below at room temperature: • • •

DI centres: gxI = 2.0044(2); gyI = gzI = 2.0024(1). Ax = Ay = Az = 1110−4 cm−1 DII centres: gxII = 2.0036(1); gyII = 2.0026(1); gzII = 2.0030(1); Ax = Ay = Az = 11·10−4 cm−1 DIII centres: gxIII = gyIII = gzIII = 2.0031(2), no hyperfine coupling.

Furthermore, the EPR spectra in X-band are characterised by individual resonance line widths less than 2 G for DI and DII species and close to 8 G for DIII centres. The relevance of the EPR signal adjustments is shown for the SiC212/1200◦ C sample in the W-band, highly sensitive to the anisotropy of the defect g-tensors ˜ (Figure 5.2.1(c)). The hyperfine constants are determined from the S-band measurements where the hyperfine lines are well resolved.

5.2.5 Assignment of the paramagnetic species in Si/C nanopowders The components of g˜ and A˜ tensors, the behaviour with the sample temperature of the EPR spectrum line width and integrated intensity are the relevant information in identifying the paramagnetic species. The anisotropic DI centres are characterised by a g˜ tensor with an axial symmetry along X-direction (!gx = gx − ge = 0.0021 and !gy = !gz = 0). Qualitatively, such shifts are not consistent with the assignment of the unpaired spins to dangling bonds (DB). Indeed, taking the DB along X-direction, a symmetry argument implies that the shifts would be !gx = 0 and !gy , !gz = 0. This conclusion is supported by the hyperfine coupling constant (A = 11·10−4 cm−1 ) which is proportional to the residence probability of the unpaired electron on the Si or C nucleus. The theoretical hyperfine constants for free ions are

©2003 Taylor & Francis

(Luchsinger et al. 1997): 29 −4 Afree cm−1 iso ( Si) ≈ −1510·10

29 −4 Afree cm−1 aniso ( Si) ≈ −37·10

and 13 −4 cm−1 Afree iso ( C) ≈ 1260·10

and

13 −4 cm−1 Afree aniso ( C) ≈ 33·10

From the above EPR spectrum analysis, the experimental isotropic hyperfine tensor for DI exp and DII centres gives Aiso = 11·10−4 cm−1 . Then, the probability of the unpaired electron being in the s orbital of Si or C is:  exp Aiso  2 29 ηs ( Si) = free  ≈ 0.006 ≈ ηs2 (13 C) Aiso 29 Si

No more than 1% of the spin density is accounted for by the localisation on one Si or C nucleus. So, the paramagnetic defects DI and DII in these SiC nanopowders can be associated to unpaired electrons delocalised inside the cubic and hexagonal crystalline sites of the coexisting SiC polytypes (3C SiC and α-SiC). The possible origin of these paramagnetic centres consists in charged vacancies in the SiC sites favoured by the initial C in excess in these materials (Kityk 2000). The elimination of the DI centres by annealing above 1500◦ C (Figure 5.2.3) seems correlated with the structural modification α-SiC → 3C SiC. The paramagnetic centre DIII is characterised by an isotropic g˜ tensor with components close to 2.0031(2), that is, close to the average value ((gx + gy + gz )/3) of DI or DII defects. The EPR spectrum contribution from DIII centres consists of a broad and unstructured resonance line (Figure 5.2.1(c)). Such features can result from defects involved in amorphous or disordered regions. However, by varying νEPR , we have not observed any line broadening

Unpaired spin concentration (spin g–1)

1e20 DI DII DIII

1e19

1e18

1e17

1200

1300

1400

1500

1600

1700

Annealing temperature

Figure 5.2.3 Variation with annealing of the absolute contributions from DI, DII and DIII paramagnetic centres. The vanishing of the DI contribution in the SiC212 1600◦ C sample coincides with the structural change α → β-SiC.

©2003 Taylor & Francis

Relative contribution (%)

70

DI DII DIII

60 50 40 30 20 10 0

20

40 60 Frequency (GHz)

80

Figure 5.2.4 Evolution with the EPR frequency band of the relative contributions of the different paramagnetic species. DI exhibits a stationary variation while that of DII and DIII indicates a conversion between these centres.

(!Hpp ) connected with a spread of the g˜ III components which may be caused by a structural X = 8 G in X-band changes only to !H W = 13 G disorder around DIII defects. Indeed, !Hpp pp in W-band. If a spread of g-values works, the ratio between the line widths will be close to W !Hpp X !Hpp

≈

W νEPR ≈ 10 X νEPR

Such a criterion rules out the location of DIII defects in disordered regions. On the other hand, when νEPR increases, that is, for short EPR characteristic times, the relative proportion of DII and DIII exhibits an opposite evolution (Figure 5.2.4). Similar behaviours are observed by varying the SiC212 1200◦ C and 1650◦ C sample temperature. As shown in Figure 5.2.5, the relative contributions to the EPR spectrum from the paramagnetic centres exhibit an anti-coincidence behaviour between the anisotropic (DI, DII) and the isotropic DIII centres (Figure 5.2.5). These results are understood by a conversion between the localised paramagnetic DII species and the DIII ones. The departure from the Curie law of the overall spin susceptibility would result from the DIII centres being delocalised in the material. This can be partly responsible for the significant macroscopic electric conduction which appears in these nanopowders annealed above 1500◦ C (Kassiba 2000) as discussed in Chapter 6. To sum up, EPR gives relevant information on the paramagnetic defects in SiC materials. Consistent analyses can be achieved when the EPR signal exhibits resolved details depending on the operating conditions (νEPR , T , etc.). A correlation can be made between the delocalised paramagnetic centres and the electrical properties in SiC nanopowders as reported in Chapter 6.

5.2.6 Silicon carbonitride (Si/C/N) powders X-band EPR spectra in HMDS series (Table 2.5) exhibit similar features in all the as-formed powders. The spectra are composed of a broad Gaussian line (peak to peak line width

©2003 Taylor & Francis

80

(a)

DI + DII DIII

Relative contribution (%)

70 60 50 40 30 20

0

50

100

150

200

250

300

Temperature (K) (b)

80 DII DIII

Relative contribution (%)

70 60 50 40 30 20 0

50

100

150

200

250

300

Temperature (K)

Figure 5.2.5 Variation with the sample temperature of the relative contributions of the anisotropic (DI, DII) and the isotropic (DIII) centres. A conversion between defects is evidenced in the (a) SiC212 1200◦ C sample and in the (b) SiC212 1600◦ C sample.

!Hpp = 9 G) superimposed on a narrow Lorentzian line (!Hpp = 4 G). The resonance position at g = 2.0030(2), is reported in a wide class of (Si, C, N) based materials where dangling bonds (DBs) are attached to C or N atoms (Warren 1990; Demichelis 1991; Suzuki 1995 and Xingguo Li 1996). Measurements down to 50 K in the as-formed HMDS40 powder did not show any resolved hyperfine lines and any effect due to a slowing down of the paramagnetic centre motions. The integrated EPR spectrum intensity varies with the temperature according to I (T ) = C/T +Ip (Figure 5.2.6). This is supported by a coexistence of localised defects giving rise to the (C/T ) contribution and delocalised ones with a temperature-independent contribution (Ip ). The latter is connected with Pauli-like paramagnetism which can originate from delocalised electrons on the free C arrangement present in a high concentration in the as-formed powders.

©2003 Taylor & Francis

1200 1150

HMDS45 HMDS40

I × T (a.u.)

1100 1050 1000 950 900 850 800 100

120

140

160 180 Temperature (K)

200

220

240

Figure 5.2.6 Evolution v. the temperature of the product I ∗ T , I represents the EPR signal integrated intensity. (•) HMDS45 annealed at 1500◦ C in 50% Ar and 50% N partial pressure. (
) HMDS40 as-formed powder. The slope of the curves is connected with the importance of delocalised paramagnetic species.

HMDS45

Intensity (a.u.)

HMDS44 HMDS43 HMDS42 HMDS41 HMDS40

3310

3320

3330

3340

3350

3360

3370

Magnetic field (G)

Figure 5.2.7 Typical X-band EPR spectra in HMDS40–45 annealed at 1500◦ C in 50% Ar and 50% N partial pressure. In HMDS45, Si3 N4 -rich material, a creation of a large defect concentration is induced by the annealing and nitriding process.

With regard to the behaviour with annealing of the paramagnetic centres in the HMDS series, 1500◦ C is a critical annealing temperature. Indeed, while the EPR broad Gaussian component exhibits an enhanced intensity only in the HMDS43–45 batches (Figure 5.2.7), the sharp Lorentzian line is dramatically reduced in all the batches after annealing at 1500◦ C

©2003 Taylor & Francis

4.0×1020

(a)

Broad line Sharp line

3.5×1020

As-formed

Number (spin g–1)

3.0×1020 2.5×1020 2.0×1020 1.5×1020 1.0×1020 5.0×1019 0.0

(b)

40

41

42 43 HMDS

44

45

44

45

4.0×1020

Broad line Sharp line

3.5×1020

Number (spin g–1)

3.0×1020

HTT 1500°C/N2/1 h

2.5×1020 2.0×1020 1.5×1020 1.0×1020 5.0×1019 0.0 40

41

42 43 HMDS

Figure 5.2.8 Paramagnetic centre concentrations: (a) in the as-formed HMDS40–45 powders and (b) in the annealed powders at 1500◦ C. Black and white bars refer respectively to the contribution from two types of defects involved in these materials.

©2003 Taylor & Francis

HMDS42

HMDS44 1700°C

1700°C

1600°C

1600°C

1550°C

1550°C

1500°C

1500°C

1400°C

1400°C

1000°C

1000°C

As-formed

As-formed

3300

3320 3340 3360 Magnetic field (G)

3380

3300

3320 3340 3360 Magnetic field (G)

3380

Figure 5.2.9 EPR spectra in HMDS42–44 v. the annealing temperature.

(Figure 5.2.8(b)). A careful analysis of the EPR spectra in HMDS44–45, reveals that the spectra consist of a sharp line and shoulders which undergo a better resolution after annealing above 1500◦ C (Figure 5.2.9, El Kortobi 1998). Since these materials are N rich (a low C/N atomic ratio), the observed EPR spectrum shoulders would originate from hyperfine interactions between unpaired spins and 14 N nuclei. A crude EPR signal deconvolution in HMDS44, annealed at 1600◦ C, is made in terms of a sharp central line and three equivalent and equidistant lines with a separation of about 5 G (Figure 5.2.10). According to EPR investigations of DB centres in amorphous Si nitride (Warren et al. 1991), Si DBs interacting with the nearest N neighbours are characterised by g˜ tensor (g = 2.0028, g⊥ = 2.0032) and by 4.6 G as a hyperfine coupling constant. Furthermore, (SiN3 )− are the most thermally stable DBs in SiC Si3 N4 networks heat treated above 1673 K under partial N pressure (Li et al. 1996). So, in HMDS44–45 annealed above 1500◦ C, DBs such as (SiN3 )− would probably bring the large concentration of the paramagnetic centres. Their magnetic susceptibility follows the Curie law in the temperature range (4–100 K) (Figure 5.2.9, El Kortobi 1998) and indicates that the involved DBs are localised. Structural reorganisations and composition changes of the HMDS networks after annealing at 1500◦ C were noticed by NMR investigations (Chapter 5.1). The behaviour of the paramagnetic

©2003 Taylor & Francis

Intensity (a.u.)

Experimental

Fit C1 N1 N2 N3

3310

3320

3330

3340

3350

3360

3370

Magnetic field (G)

Figure 5.2.10 Comparison between the experimental EPR spectrum in HMDS44 annealed at 1600◦ C and the calculated ones by a superposition of a central line (C1) and triplet lines (N1, N2, N3). (C1) is associated to paramagnetic centres implying C atoms. (N1, N2, N3) lines are induced by hyperfine coupling of unpaired electrons on Si atoms with equivalent N nuclei in centres such as (SiN3 )− .

centres seems intimately connected with these modifications. What can be addressed below is the correlation between the characteristic EPR lines, that is, the paramagnetic species and the sample composition. On one hand, the narrow Lorentzian EPR line can be partially caused by delocalised paramagnetic defects as suggested by the temperature variation of the EPR integrated intensity (Figure 5.2.11). Motional narrowing and exchange between localised and delocalised spins by a thermal activation were reported in C black particles embedded in a polymer matrix (Adriaanse et al. 1997). Similarly, in C rich material (HMDS40–41–42), the importance of the narrow Lorentzian lines would result from DBs and delocalised unpaired electrons on C atoms. On the other hand, the broad EPR line is consistent with paramagnetic species implying N atoms. Indeed, when the material composition is 90% of Si3 N4 type (HMDS44 annealed at 1600◦ C), the experimental spectrum is characterised by an isotropic g tensor (2.0030(2)) and hyperfine triplet lines due to the coupling of S = 21 defects with I (14 N) nuclear spins. The large spin concentration created by annealing under N partial pressure can be explained by the large rate of structural defects induced by the nitriding process and structured reorganisation. These features are similar to those reported in ultrafine SiC Si3 N4 composite powders (Li et al. 1996). According to these authors, annealing above 1500◦ C gives rise to different behaviours of the EPR signal depending on the atmosphere used. Under argon, the defect concentration decreases with annealing while the opposite behaviour is obtained for annealing under Ar and 50% N2 . We have shown that more homogeneous materials are obtained by annealing at 1500◦ C and above, HMDS40–41–42 (SiC-rich materials) and HMDS43–44–45 (N-rich materials).

©2003 Taylor & Francis

Integrated intensity (a.u.)

25

20

HMDS44 1600°C

15

10

5

0 0.0

0.1

1/T (°K–1)

0.2

0.3

Figure 5.2.11 Integrated EPR line intensity in HMDS44 annealed at 1600◦ C v. the inverse sample temperature (1/T ). The Curie law is revealed and indicates the localised nature of the coexisting paramagnetic centres.

A large concentration of paramagnetic defects is created in HMDS45–44–43 consecutive to the annealing and the nitriding process. The opposite behaviour is observed in HMDS40– 41–42 where a dominant SiC structure is realised as well as the removal of the free C content.

References Adriaanse, L. J., Brom, H. B., Michels, M. A. J. and Brokken-Zijp, J. C. M. (1997) Phys. Rev. B55, 15, 9383. Charpentier, S., Kassiba, A., Emery, J. and Cauchetier, M. (1999) J. Phys.: Condens. Matter, 11, 4887. Charpentier, S., Kassiba, A., Bulou, A., Monthioux, M. and Cauchetier, M. (1999a) EPJ: AP, 8, 111. Charpentier, S., Kassiba, A., Fusil, S., Armand, X., Cauchetier, M., Fayet, J. C. and Emery, J. (1997) Appl. Magn. Reson., 12, 255. Demichelis, F., Pirri, C. F., Tresso, E., Rigato, V. and DellaMea, G. (1991) J. Non-Cryst. Solids, 128, 133. El Kortobi, Youssef (1998) Thesis, Université P. M. Curie, Paris, France. http://corail.sudoc.abes.fr80 & Sudoc-Catalogue no 053764552. Kassiba, A., Tabellout, M., Charpentier, S., Herlin, N. and Emery, J. R. (2000) Solid State Commun., 115, 389. Kityk, I., Kassiba, A., Tuesu, K., Charpentier, S., Ling, Y. and Makowska-Janusik, M. (2000) Mater. Sci. Eng. B, 77(2), 147. Li, X., Chiba, A., Nakata, Y., Nagai, H. and Suzuki, M. (1996) Mater. Sci. Eng., A219 95. Luchsinger, R. H., Yu Zhou and Meier, P. F. (1997) Phys. Rev. B55, 1, 6927. Suzuki, M., Hasegawa, Y., Aizawa, M., Nakat, Y. and Okutani, T. (1995) J. Am. Ceram. Soc. 8, 1. Warren, W. L., Lenahan, P. M. and Curry, S. E. (1990) Phys. Rev. Lett. 65, 2, 207.

©2003 Taylor & Francis

5.3 Electronic structure studied by X-ray photoemission (XPS) and soft-X-ray (SXS) spectroscopies Adriana Gheorghiu de La Rocque, Georges Dufour, François Ténégal and Christiane Sénémaud

5.3.1 Principles X-ray Photoemission Spectroscopy (XPS) provides the electronic distributions of the core levels and valence band (VB) of solids. As concerned the VB states, energy distribution curves (EDCs) correspond to the total density of states (DOS) modulated by appropriate photoionisation cross-sections: I (E) ∝ σS NS (E) + σP NP (E) + · · · σx being the cross-section and Nx (E) the partial DOS of x state. Thus, in some cases, information concerning particular symmetry states can be arbitrarily enhanced. Due to the small escape depth of emitted photoelectrons (0.5–10 nm), the experiments are sensitive to the superficial layer of the samples. X-ray spectroscopy probes separately the valence and the conduction DOS. In X-ray Emission (XES), radiative transitions between the VB and a localised inner vacancy are investigated. Initial core holes are created by means of either electron impact or X-ray photons, possibly synchrotron beam. The radiative de-excitations of the inner vacancy are governed by dipole selection rules which imply that only valence states with particular angular momentum are involved. Moreover, as the core level concerns a specific atomic site, the DOS around each atomic species of a compound are investigated separately. The information obtained by XES concerns the volume. The spectral intensity of an X-ray emission is, in first approximation:  1±1 I (hν) ∝ ν 3 Mif2 (E)Nocc (E)L(hν–E) dE where Mif (E) is the matrix element which does not affect the main features of an XES 1±1 (E) is the DOS with (1±1) symmetry; L(hν–E) spectrum over a limited spectral range; Nocc is the Lorentzian distribution of the core hole. By X-ray absorption spectroscopy, transitions between a core level and the conduction band states are analysed. In the spectral region close to an absorption edge: X-ray Absorption Near Edge Structure (XANES), the photoabsorption coefficient µ is in first approximation proportional to the density of the empty states Nunocc above the Fermi level (EF ), with particular angular momentum around a specific site:  1±1 µ(hν) ∝ ν Mif2 (E)Nunocc (E)L(hν–E) dE By XANES either the volume or only the surface of the sample is concerned according to the technique chosen (Total Electron Yield (TEY), transmission or Fluorescence Yield (FY)).

©2003 Taylor & Francis

For both XES and XANES spectra, the contribution of the Lorentzian profile of the core hole induces a broadening of the spectra, which limits the resolution of the method and implies to study only transitions involving a sufficiently sharp core level. For a valuable comparison of XPS and SXS data, it is particularly important to set both XPS–XES (or XPS–XANES) experimental curves in a common energy scale referring to the Fermi level. This is directly obtained in XPS as EDCs are referred to EF . In the case of XES and XANES spectra, the spectral curves are obtained in a transition energy scale (EVB –EX ) or (ECB –EX ) and the energy position of the Fermi level can be obtained by determining the core level binding energy (BE) EX , provided that relaxation effects are neglected.

5.3.2 Experiments 5.3.2.1

X-ray Photoemission Spectroscopy (XPS)

XPS spectra were obtained by using Mg Kα incident radiation (hν = 1253.6 eV). An electrostatic hemispherical analyser was used in the fixed analyser transmission mode (FAT) with a pass energy equal to 40 eV. Under these conditions, the full-width at half-maximum (FWHM) of the Ag 3d5/2 line is 1 eV. The core level spectra were scanned with a 0.1 eV step and the XPS VB with a 0.2 eV step. In some cases, an energy shift of the photoelectron peak due to a charging effect was observed (Shirley 1972; Jentoft et al. 1999). In order to take this effect into account, we calibrate the BE scale with reference to the C 1s line observed from hydrocarbonated species, which are always present at the surface of the sample. The C 1s line is set at 285.0 eV. In our case, the corrections are more complicated because carbon (C) is also present in the studied materials. The energy scale was then adjusted so that O 1s peak, which corresponds essentially to silicon dioxide, coincides approximately for all spectra. In this procedure, it is assumed that oxygen atoms are essentially present in silicon dioxide bonding. Let us note that in our experimental work the oxygen proportion measured by XPS is only a few percent in as-formed (AF) samples and slightly higher, 10–15%, in annealed samples, due to surface contamination (Gheorghiu et al. 1997). The interpretation of the core level lines is based on a decomposition of each spectral distribution into Voigt functions; the spectra are superposed to a background due to the inelastic electron scattering. For a mathematical treatment of core level spectra, a subtraction of this background is performed by assuming that its variation is linear or polynomial. No correction of background was made in the case of the VB spectra. 5.3.2.2

Soft X-ray Spectroscopies (XES–XANES)

In the case of Si/C/N powders, we measured the Si Kβ (3p → 1s) emission line which probes the Si 3p VB states and Si K absorption line (1s → np) which provides the Si p conduction band (CB) states. The X-ray emission spectra Si Kβ were analysed with a bent-crystal vacuum spectro¯ crystal (2d = meter (Sénémaud et al. 1989) using as a monochromator either a quartz 1010 850.766 pm) in the first order of reflection, or a gypsum crystal (020) (2d = 1519.84 pm) in the second order of reflection bent to a radius of 250 mm. The detector was a proportional counter (500 mb Ar/CH4 90/10) with an entrance slit fixed to 50 µm. The spectra were scanned with a 0.2 eV wide step. The total instrumental broadening, mainly due to the monochromator itself, is estimated to be about 0.3 eV (Sénémaud and Ardelean 1990). The emissive target was prepared by depositing a thin layer of laser synthetised powder onto

©2003 Taylor & Francis

a copper anode. The spectra were excited by electron bombardment with exciting conditions 5 kV, 5 mA. The thickness sampled is determined by the penetration of the incident electron beam. In our experimental conditions, we estimate the thickness of the emitting layer to about 0.5 µm. Thus in this case the whole volume of the powder particles is concerned. The near edge photoabsorption (XANES) measurements were carried out at Laboratoire pour l’Utilisation du Rayonnement Electromagnétique – Orsay (LURE). The Super-Aco storage ring was operating at 800 MeV, and the average current was approximately 200 mA. A two-crystal monochromator equipped with InSb (111) crystals (2d = 748.06 pm) was used. The photoabsorption spectra corresponding to the Si K-edge were obtained in the transmission geometry and scanned with a 0.2 eV-wide step. The detector was a proportional counter filled with a 300 mb Ar/CH4 (90/10) mixture. The absorbing screens were obtained by depositing powders onto 5-µm-thick millipore polycarbonate membranes. The spectra of the incident beam were obtained through a membrane without powder. The instrumental resolution is evaluated to be approximately 0.4 eV (Bouisset et al. 1991).

5.3.3 Study by XPS–SXS of Si, Si/C, Si/N and Si/C/N laser synthetised nanopowders produced from gaseous precursors The samples were prepared by laser pyrolysis of appropriate precursors in the same conditions as reported previously (Gheorghiu et al. 1992), summarised in Table 2.2. 5.3.3.1

Core levels

The Si 2p spectrum of Si50 powder (Figure 5.3.1) exhibits a strong asymmetric line accompanied on the high-BE side by a low intensity component due to SiO2 . This oxide contribution has approximately the same relative intensity as that of a native oxide layer at the surface of a monocrystalline Si wafer (Himpsel et al. 1988). The main peak, located at 99.3 eV,

SiO2

Si 2p1/2 Si 2p1/2

Intensity (au)

Si50 Si 2p

106

104 102 100 Binding energy (eV)

98

Figure 5.3.1 Si 2p core levels for Si50 sample.

©2003 Taylor & Francis

(a)

(b)

SiO2 SiC

SiO2 SiC SiC152a

SiC151

Si 2p Intensity (au)

Intensity (au)

Si 2p

106

104 102 100 Binding energy (eV)

98

106

98 104 102 100 Binding energy (eV)

Figure 5.3.2 Si 2p core levels for (a) SiC151 and (b) SiC152a samples.

corresponds to the 2p1/2,3/2 doublet from Si Si bonds and is well fitted by a sum of two Voigt lines separated by 0.6 eV, with an intensity ratio 1/2 which accounts for the spin–orbit splitting. The total width at half-maximum of the line is 1.2 eV, close to the value obtained with a single-crystal wafer and in the same experimental conditions: 1.1 eV (Rochet et al. 1986). The remaining intensity in between Si and SiO2 lines reveals probably the existence of an intermediate oxide SiOX (Hollinger et al. 1988). A tailing could result from inhomogeneities in the sample. For pure Si powders, the XPS results show that the sample is a well-organised silicon (Si), which is in agreement with X-ray diffraction (XRD) results (Table 2.2). In Si/C samples, SiC151 and SiC152a (Figures 5.3.2(a) and (b), the Si 2p spectra have identical shapes: quasi-symmetric with a high BE tail. The decomposition achieved on the basis of two components: one at the energy position of the maximum intensity, the second at higher BE, corresponding to SiO2 (102.9 ± 0.1 eV), shows that the energy of the main line is located at 101.3 and 101.2 eV for SiC151 and SiC152a, respectively. The Si 2p line is slightly sharper in SiC151 than in SiC152a. Finally, a low proportion of SiO2 is present at the surface for both samples. It is noteworthy that the intensity is negligible at the BE corresponding to Si Si bonds. Thus in Si/C, heteropolar Si C bonds are present, without any segregation of Si atoms. In SiC151, the local order is slightly better than in SiC152a and commercial β-SiC (Driss-Khodja et al. 1992) (fewer bond angle and bond length fluctuations). Consequently, XRD measurements led to the same conclusions (Cauchetier et al. 1988). The C 1s BE from powders is close to the value that characterised the C Si bonds in β-SiC (Driss-Khodja et al. 1992). The C 1s line is shifted towards high BE in powder samples as compared to β-SiC. As the energy distance E(Si 2p) − E(C 1s) varies, this effect cannot be interpreted in terms of a Fermi-level shift but rather by the presence of H atoms in the network, in agreement with previous measurements (Fang and Ley 1989). This is in agreement with the identification of Si Hn groups by infrared (IR) spectroscopy in samples prepared under the same conditions (Cauchetier et al. 1990). For Si/C powder materials, the XPS measurements reveal that only Si C bonds are present in Si C4 and C Si4 groups and that there is no segregation of Si atoms in the network. They reveal, moreover, that the samples contain a very low proportion of oxide. The atomic

©2003 Taylor & Francis

(a)

SiO2

Si3N4

(b)

Si

SiO2 Si3N4

SiN7 Si 2p

α-Si3N4

Intensity (au)

Intensity (au)

Si 2p

106

104 102 100 Binding energy (eV)

98

106

104 102 100 Binding energy (eV)

98

Figure 5.3.3 Si 2p core levels for (a) SiN7 sample and (b) α-Si3 N4 powder.

structure and the long-range order of the material depend on the preparation conditions. Thus, an increase of the laser power used for the powder synthesis from 220 to 600 W (Table 2.2) induces an increase of the flame temperature and consequently of the local order; the Si 2p becomes slightly sharper. The Si 2p peaks observed from the SiN7 sample (Figure 5.3.3(a)) compared to commercial α-Si3 N4 (Figure 5.3.3(b)) show that, in both cases, the spectrum consists of a wide, asymmetric line, the high-binding energy tail resulting from the presence of SiO2 . The decomposition of the α-Si3 N4 spectrum in two components presented in Figure 5.3.3(b) shows a main Si 2p component, 1.7 eV wide at 101.8 eV BE, accompanied by a contribution from SiO2 at 103.1 eV having an intensity of about 11%. In this case the spin–orbit splitting of Si 2p is not observed due to the increase of Si 2p FWHM as compared to pure Si. The Si 2p spectrum from SiN7 sample has a similar shape but is significantly broader than the commercial α-Si3 N4 one. The peak decomposition gives a main peak, 2.0 eV wide, located at 101.7 eV, and a SiO2 contribution with a relative intensity of about 12% (Figure 5.3.3(a)). Moreover, a low BE feature is present at 99.0 eV which we interpret as due to the existence of a low proportion (about 2%) of Si Si clusters. Thus, in the SiN7 powder, most Si atoms are bonded to N atoms in Si N4 units which is characteristic of the silicon nitride network. The broadening of the corresponding Si 2p component as compared to the α-Si3 N4 one (2.0 instead of 1.7 eV) can be attributed to the existence of structural disorder in the network, introduced by fluctuations in the bond lengths and/or bond angles as compared to their values in the crystalline phase. The presence of H atoms in the material is possible and could contribute to such a disorder. The N 1s line is observed at approximately the same BE value (397.7 eV) in both the powders and the stoichiometric α-Si3 N4 samples indicating the N Si3 configuration, but the width is significantly broader in the powder than in crystalline α-Si3 N4 (1.7 eV). This result confirms the existence of a chemical order in the powder sample, the broadening of the line being attributed to the local topological disorder, in agreement with the above results concerning the Si 2p line. The existence of N H bonds can contribute to a broadening of the N 1s line as previously observed in Si NX : H samples prepared by photo-chemical-vapour deposition (PCVD) (Driss-Khodja et al. 1989).

©2003 Taylor & Francis

Si3N4

(a)

SiO2

Si3N4

(b)

SiC Si

SiO2

SiCN12 Si 2p

SiC

Si

Intensity (au)

Intensity (au)

SiCN27 Si 2p

106

104 102 100 Binding energy (eV)

98

106

104 102 100 Binding energy (eV)

98

Figure 5.3.4 Si 2p core levels for (a) SiCN12 and (b) SiCN27 samples.

We conclude that the SiN7 powder contains mainly Si N4 groups, N atoms being surrounded by three Si atoms. The local bonding is similar to that of silicon nitride α-Si3 N4 . A few Si atoms are involved in Si Si bonds; the proportion of oxide is very low (12% in the powder). The presence of Si Si bonds in the network, as well as the possible existence of hydrogen in the sample, can contribute to the introduction of short-range disorder around both Si and N atoms. The Si 2p spectra from the two Si/C/N samples, SiCN12 and SiCN27 (Figures 5.3.4(a) and (b)) consist of a rather broad asymmetric line with different FWHM: 2.2 and 1.7 eV, respectively. According to the conditions of decomposition, we infer that, besides the small SiO2 contribution around 103.3 eV, the SiCN12 spectrum results from the contribution of two main lines, while SiCN27 spectrum corresponds to a single Voigt line. In the case of SiCN12 spectrum, the two Voigt curves, 1.5 and 1.7 eV wide, are located at 101.1 and 101.9 eV, respectively; the intensity ratio of the components is about 0.7. These two components are close to Si 2p lines from β-SiC and α-Si3 N4 , respectively. An additional line, lower in intensity, at 99.6 eV can be attributed to Si Si. In contrast, the main peak of SiCN27 is 1.7 eV wide and is located at 101.6 eV. It is accompanied by two smaller components at 99.4 and 103.3 eV that we attribute to Si Si bonds and SiO2 , respectively. This result clearly shows that the local bonding is quite different in the two Si/C/N samples under consideration. In one case, SiCN27, for which C/N = 0.22, most Si atoms have the same local structure SiCx N4−x , Si being mainly bonded to N atoms. The value of the FWHM is the same as in the crystalline α-Si3 N4 and reveals that the network is well ordered. It is noteworthy that the Si 2p line from this sample is sharper than that from the Si/N7 sample. Consequently, the modifications of the gas mixture (addition of CH3 NH2 ), which leads to an increase of temperature, contributes to give a better ordered material. In the SiCN12 sample, with C/N = 0.93, our results show clearly that both Si N4 and Si C4 groups are present, with approximately the same proportion. The C 1s BE corresponding to C Si bonds is practically the same in both samples. However, the FWHM increases significantly in mixed compounds as compared to Si/C, indicating a loss of local order around C atoms induced by the presence of N atoms.

©2003 Taylor & Francis

The N 1s BE value remains the same for all samples containing N atoms, revealing a preferential N Si3 arrangement in Si/C/N samples. A decrease of its FWHM is observed for mixed compounds as compared to Si/N, which confirms the increase of local order induced by C atoms. In the case of composite Si/C/N, our results show that different materials can be prepared depending on the preparation conditions, especially the initial gases concentration. The powders are formed either by a mixture of Si C4 and Si N4 units or by a single type of local Si environment. For a C/N ratio of about 1 in the initial gas mixture, the existence of two types of local Si bonding, Si C4 and Si N4 , with a proportion of about 0.7–1 is revealed by our XPS measurements; a few Si atoms are involved in Si Si bonds. For an initial C/N ratio in the gas mixture equal to 0.22, the XPS spectra reveal that most Si atoms are involved in a single type of chemical environment; the Si 2p core level has a BE close to the value corresponding to α-Si3 N4 ; the Si C bonds are also present, as revealed by the C 1s line. Consequently our results suggest the existence of local C Si N3 arrangement around Si atoms.

5.3.3.2

XPS valence band results

The VB spectra of powders Si/C (SiC151), Si/N (SiN7) and Si/C/N (SiCN12, SiCN27) exhibit a broadband (Figure 5.3.5) noted A extending between 0 and about 13 eV, followed by peaks B and C, at about 15 and 19 eV, respectively, which have different relative intensities in the considered samples. The distances from these features to EF are given in Table 5.3.1. The SiC151 VB spectrum is almost identical to a β-SiC spectrum, as already reported (Driss-Khodja et al. 1992). Part A exhibits two peaks labelled A2 and A4 located at 5.7 and 9.5 eV, respectively. From a comparison with theoretical densities of states (Robertson 1992) these peaks are attributed to C 2p states mixed with Si 3p and Si 3s, respectively; a wide minimum separates A4 from peak B, located at 14.4 eV BE, which corresponds to C 2s states mixed to Si 3p and Si 3s. The SiN7 spectrum is very similar to the silicon nitride (Johnson Mattey, 99.99% powder) XPS VB spectrum previously reported (Sénémaud et al. 1993). The upper part A exhibits

C

B

A

Intensity (au)

A⬙

A5 A4

SiCN27

A⬘

SiCN12

A3 A1 A2

SiN7 SiC151

20

15

10

5

EF

Binding energy (eV)

Figure 5.3.5 XPS VB spectra of laser-synthesised SiC151, SiN7, SiCN12 and SiCN27 samples.

©2003 Taylor & Francis

Table 5.3.1 Energy positions of the XPS VB features (eV) for laser synthesised samples A1 SiC151 SiN7 SiCN12 SiCN27

A2

A′

A3

5.7

A′′

A5

9.5

3.3

11.4 10.8

7.6

B

C

14.4

7.5 7.0

3.9

A4

14.9 11.4

18.9 19.3 19.0

three structures labelled A1 , A3 and A5 located at 3.3, 7.5 and 11.4 eV, respectively. From the previous study of Si3 N4 XPS and SXS spectra (Sénémaud et al. 1993), in agreement with theoretical DOS (Robertson 1992), peak A1 is unambiguously attributed to N 2pπ lone pair, peaks A3 , A5 are associated respectively to N 2p–Si 3p states and N 2p–Si 3s states. Peak C, well separated from the A band, is due to N 2s states mixed with Si 3s–Si 3p states. The VB spectra from SiCN12 and SiCN27 are quite different in shape (Driss-Khodja 1993). The spectrum from sample (SiCN12) with C/N = 0.93 exhibits only two structures A′ and A′′ at 7.0 and 10.8 eV. The energy position of A′ is intermediate between that of A2 from SiC151 and A3 from SiN7. Similarly, A′′ is located between A4 (SiC151) and A5 (SiN7) (Table 5.3.1). The low BE edge of the spectrum is more contrasted than in the SiN7 spectrum; however, it is less abrupt than in SiC151. It is noteworthy that both peaks B and C are observed in this spectrum at 14.9 and 19.3 eV. Their energy positions correspond to those of peaks B from SiC151 and C from SiN7 suggesting that both C 2s and N 2s mixed to Si 3s, 3p states are present. For SiN27, which corresponds to C/N = 0.22, the spectrum is rather similar to the SiN7 one. The upper band A exhibits three structures (see Table 5.3.1); it is followed by a sharp peak C located at 19.0 eV, well separated from band A; consequently, the VB states correspond essentially to those of Si N bonds found in silicon nitride. Let us note that the minimum between A and C is not so marked as in SiN7 spectrum and that a shoulder is clearly observed around 15 eV. 5.3.3.3

SXS-XPS results

In Figures 5.3.6–5.3.9, we have plotted, in the same BE scale with EF as origin, the XES Si Kβ, the XPS VB and the XANES spectra for powders SiC151, SiN7 and SiCN12 and SiCN27. Let us recall that the XES Si Kβ curves give the Si 3p VB states distribution and the XANES curves the unoccupied Si p states distribution. Si Kβ from SiC151 powder (Figure 5.3.6) is very similar to silicon carbide Si Kβ spectrum observed previously (Driss-Khodja et al. 1992; Zahorowski et al. 1989). The maximum intensity located at 5.3 eV from EF corresponds to peak A2 of XPS VB spectrum, the decrease of intensity towards EF is rather abrupt; its inflection point coincides with that of the low BE edge of XPS VB spectrum. The high BE edge is less steep and shows a shoulder at about half-maximum, that is at an energy corresponding approximately to the energy of peak A4 . Finally a small intensity peak, well resolved from the main band, is observed at 15 eV from EF , showing that Si 3p states are present deeper in the VB. This small peak coincides with peak B of XPS VB spectrum. The XANES spectrum of SiC151 presents a two-step edge extending from about 0 to 5.3 eV, where a sharp maximum occurs. A secondary maximum occurs at 11.5–12 eV and a strong peak at 19 eV from EF .

©2003 Taylor & Francis

B

Intensity (au)

XPS

SiC151

A4 A2

XAS

XES

30 20 10

EF

Binding energy (eV)

Figure 5.3.6 XPS VB, XES and XAS spectra from SiC151 sample adjusted in a common energy scale with EF as origin.

SiN7

C A5

Intensity (au)

XPS

A3 A1 XAS

XES

30 20 10

EF

Binding energy (eV)

Figure 5.3.7 XPS VB, XES and XAS spectra from SiN7 sample adjusted in a common energy scale with EF as origin.

Si Kβ from SiN7 (Figure 5.3.7) is a quasi-symetric band; its overall shape is close to that of silicon nitride Kβ emission band previously reported (Sénémaud et al. 1993); however, its FWHM is wider, 6.5 instead of 5.5 eV. The energy position of the maximum, 8.4 eV from EF gives the mean BE of the Si 3p states; it is the same as in the nitride. This maximum coincides approximately with peak A3 of XPS VB spectrum. Consequently, this result confirms that the peak A1 of XPS VB spectrum corresponds to pure N states. We note that the decrease of intensity on the low BE tail of Si Kβ is not steep but exhibits some tailing; we attribute this effect to the presence of a few Si Si bonds in this sample. Let us recall that these like-atom bonds had already been revealed by the study of Si 2p core level from the same sample (Figure 5.3.3(a)). The low intensity peak located at 19.5 eV is in coincidence with peak C of XPS VB spectrum and reflects the presence of Si 3p states mixed to N 2s ones in this part

©2003 Taylor & Francis

of the VB. The XANES spectrum of SiN7 consists of a steep edge whose inflection point is located 1 eV above EF , followed by a maximum at 3.8 eV from EF . A second maximum is observed around 22 eV from EF . In SiCN12 and SiCN27, the Si Kβ emission band is a quasi-symmetric band with FWHM 7.7 and 6.5 eV, respectively (Figures 5.3.8 and 5.3.9) (Driss-Khodja 1993). In SiCN12 sample, the main peak is accompanied by a shoulder at 8.5 eV, in coincidence with SiN7 Kβ maximum; a faint feature around 5 eV could be a signature of SiC151 Si Kβ maximum. The distance from EF of the low BE edge, measured at half amplitude, is about the same as in SiC151. The Si Kβ spectrum is broader than in binary compounds and can be considered approximately as a sum of both the SiC151 and the SiN7 spectral distributions. It is noteworthy that, in the high BE range, a small feature is observed at the energy positions of B. The photoabsorption edge of SiCN12 has a continuous slope, slightly less abrupt than in SiN7 spectrum. Beyond the first maximum, two peaks are observed at about 3.8 and 5.3 eV from EF , that is, at approximately

Intensity (au)

XPS

SiCN12

C A4 B A2

XAS XES

30 20 10 EF Binding energy (eV)

Figure 5.3.8 XPS VB, XES and XAS spectra from SiCN12 sample adjusted in a common energy scale with EF as origin.

SiCN27

B A5 A3 A1

Intensity (au)

XPS

XAS XES

30

20

10

EF

Binding energy (eV)

Figure 5.3.9 XPS VB, XES and XAS spectra from SiCN27 sample adjusted in a common energy scale with EF as origin.

©2003 Taylor & Francis

the same position as in SiC151 and SiN7 photoabsorption curves, respectively. Beyond this energy range, a slight feature is noted at 11.5 eV from EF and a strong maximum at 19.6 eV from EF . In SiCN27, the Si Kβ spectrum is narrower than in SiCN12; its shape is rather similar to SiN7 Si Kβ. The maximum is approximately in coincidence with peak A3 of the XPS VB curve, while no feature appears at the energy position of A1 . The photoabsorption curve exhibits a strong maximum at 3.6 eV from EF , that is, close to that observed in SiN7 photoabsorption, giving the appearance of a ‘white line’. A shoulder is observed at 7 eV and a wide maximum at 21 eV from EF . 5.3.3.4

Discussion

Complementary information on the valence and conduction states of solids can be obtained by using both SXS and XPS spectroscopies. XES and XANES spectra correspond in first approximation to the convolution of the Si p occupied or unoccupied state distributions by the Lorentzian distribution of the Si 1s core-level distribution, which is about 0.5 eV wide (Sénémaud et al. 1993). The XPS spectra correspond to the total valence distribution modulated by the photoionisation cross-sections which, in our case, emphasises s states as compared to p states. For binary systems Si/C and Si/N, the experimental XPS and XES–XAS spectral curves are qualitatively well reproduced by the DOS calculations performed by J. Robertson in the tight-binding formalism for crystalline β-SiC (Robertson 1992) and α-Si3 N4 crystalline phases (Robertson 1991). Differences in relative intensities of the peaks are due to differences in photoionisation cross-sections (Driss-Khodja et al. 1996). These conclusions are in total agreement with those obtained for the same sample from the XPS core levels and EXAFS studies (Gheorghiu et al. 1992). In Si/C, our results confirm unambiguously the existence in the powder material of chemically ordered sp3 bonded network with Si C heteropolar bonds. The Si 3p states whose position corresponds to Si Kβ emission are mainly located at the top of the VB, in coincidence with peak A2 of XPS. The sharp XPS peak A4 at 9.5 eV, at the same energy as a low feature of Si Kβ, can be associated to sp states. From Robertson’s calculations, we have deduced that this peak is due to mixed Si 3s and C 2p states; from our results Si 3p states are also involved in this energy range. A sharp minimum, corresponding to the ionicity gap, separates the maximum A4 from peak B; this energy gap is characteristic of Si C bonds in a crystalline network. The minimum is expected to be partially filled in the case of amorphous SiC, as predicted from a theoretical study of α-SiC density of states performed by Finocchi et al. (1993). Peak B is largely C 2s in character, with a small, but non-negligible Si 3p contribution as clearly revealed by our XES data, in excellent agreement with theoretical predictions. As concerned the conduction band, an overall agreement is found between the photoabsorption curve and the theoretical DOS, for which the absorption edge is followed by two peaks attributed to antibonding Si 3p–C 2s and Si 3p–C 2p states distant by about 3 eV ; the non-resolution of these features in the experimental curve is probably due to the core-level broadening. Beyond the first peak, the experimental photoabsorption curve shows well-resolved peaks that reveal the existence of a well-organised network. This is in agreement with the EXAFS results (Gheorghiu et al. 1992) we have obtained for this powder, which clearly revealed the existence of Si C4 groups. The influence of hydrogen incorporation in SiC matrix has been recently studied in a theoretical simulation by Finocchi and Gattia (1994). For a-SiC : H alloys at stoichiometric

©2003 Taylor & Francis

composition, containing 20% hydrogen, these authors predict that hydrogen atoms are bonded to both C and Si atoms and that the presence of H favours the formation of a tetrahedral network and thus enhances the chemical order in the material. This is well supported by our results. In Si/N powder, the VB distribution observed experimentally is very close to the Si3 N4 one. Our results confirm that the top of the VB corresponds to an N 2pπ lone pair orbital, in agreement with theoretical results (Robertson 1991). The lone pair orbital is characteristic of the planar configuration of N atoms surrounded by three Si atoms. This feature is followed by mixed Si 3p–N 2p and Si 3s–N 2p states; N 2s states, mixed to Si 3s–3p states appear as a well-resolved peak. A broadening of the Si 3p distribution, when going from Si3 N4 (Sénémaud et al. 1993) to Si/N powders is noted. Two effects can contribute to this broadening. The first effect is due to a loss of long-range order as compared to well-crystallised silicon nitride; a topological disorder attributed to fluctuation of bond lengths and bond angles has been evidenced in this material from a previous study (Gheorghiu et al. 1992). Let us note that this effect does not modify the overall shape of the VB distribution, which is essentially determined by the short-range order (Kärcher et al. 1984). The second effect is the presence of like-atoms bonds Si Si in a low, but not negligible proportion, as already revealed by the Si 2p core level study. Thus Si/N powders are characterised by both structural and chemical disorder. The presence of hydrogen in the network is possible, according to the preparation conditions. In a previous study of a-SiNX : H alloys, we had shown that hydrogen atoms are preferentially bonded to N atoms (Driss-Khodja et al. 1989). In this case, the Si states distribution should not be significantly modified. From the results obtained by Ley for hydrogenated a-Si NX samples (Kärcher et al. 1984), hydrogen incorporation involves a recession of the VB maximum, which remains probably small for the low H concentration corresponding to our samples. Moreover, due to the low value of H 1s cross-section under our experimental conditions (hν = 1253 6 eV), the effect of hydrogen on XPS VB spectrum remains probably quite negligible. The first photoabsorption peak gives the distribution of the unoccupied Si p states; its shape corresponds to a disordered network and supports the previous EXAFS results on the same samples (Gheorghiu et al. 1992). In this energy range, the DOS calculations for silicon nitride (Robertson 1991), which covers an energy range of about 12 eV from EF , predicts the existence of a broad maximum between 5 and 10 eV from EF ; the shift between the experimental and the calculated data that we have already mentioned (Gheorghiu et al. 1992) can be explained by the experimental gap value introduced in the calculation. It is interesting to note that the presence of hydrogen in the network favours a higher chemical and structural order in the case of Si/C systems; on the contrary, it involves a tendency for more disorder in the case of Si/N powder; in both cases the laser power used for the synthesis is 600 W (Table 2.2). In Si/C/N powders, the valence and conduction band distributions are quite different according to the composition. For C/N value equal to 0.93 (SiCN12), our results indicate clearly that Si atoms are involved in both carbide and nitride configurations. This is evidenced from XPS VB spectrum which reveals that both C 2s and N 2s states, mixed to Si sp states, are present; on the XES Si Kβ spectrum, we observe a small peak at the energy position of B (Si 3p–C 2s states). Moreover, these results are well supported by the photoabsorption spectrum of SiCN12, which shows two features close to the edge in coincidence with the first absorption maximum of SiN7 and SiC151, respectively, next features being close to those observed in SiC151 spectrum.

©2003 Taylor & Francis

In SiCN27 with C/N = 0.22, the spectral distributions are rather close to SiN7 ones, but slight differences are noted which clearly indicate that the network is not identical to silicon nitride: (i) the XES Si Kβ spectrum (Figure 5.3.9) shows that the Si 3p maximum is slightly shifted towards low BE as compared to SiN7 Si 3p, without broadening; this result shows that additional Si C bonds are present in the network, a few Si N bonds being replaced by Si C bonds; (ii) the minimum between A5 and B on the XPS VB spectrum is slightly enhanced as compared to SiN7 one. The energy position of this minimum corresponds to that of peak B of Si151C spectrum and consequently this additional shoulder reveals that C 2s states due to a few Si C bonds are present in the network. On the photoabsorption curve, the sharp maximum following the photoabsorption edge corresponds to the SiN7 first maximum; it is characteristic of a crystalline Si3 N4 network. A maximum is located at about 20 eV, close to both SiN7 and SiC151 maxima. The presence of a shoulder at about 7 eV from EF could result from Si C bonds. Consequently, for sample SiCN27, our experimental results are consistent with the existence of a single type of chemical environment of Si atoms with both Si N and a few Si C bonds in the network. This is in quite good agreement with our previous results obtained from XPS core levels and EXAFS studies (Gheorghiu et al. 1992). Therefore, our results reveal that in ternary systems the local environment of Si atoms depends essentially on the preparation conditions of the powder. For C/N ratio close to 1, Si atoms are involved in both Si C4 and Si N4 configurations. EXAFS data for these samples (Gheorghiu et al. 1992) indicated that atomic distances Si C and Si N are close to the corresponding crystalline values. Moreover, they clearly revealed the presence of β-SiC microcrystallites in the material; we suggested that the microcrystals could be contained in an amorphous phase of SiN4 units, H atoms being present in NHX , CHX or SiHX groups. For an initial ratio C/N equal to 0.22, we confirm a local atomic arrangement with both C and N atoms surrounding the same Si atom, in agreement with XPS core levels and EXAFS results (Gheorghiu et al. 1992).

5.3.4 Effect of heat treatment on the electronic structure of nanometric Si/C/N ex-HMDS and pre-alloyed Si/C/N/Al/Y powders studied by XPS spectroscopy 5.3.4.1

HMDS nanopowders

The samples were prepared by laser pyrolysis of hexamethyldisilazane. The experimental method, already described (Cauchetier et al. 1994; Martinengo et al. 1996), is based on the resonance between the emission of a continuous-wave CO2 laser, at 10.6 µm, and the IR absorption of the liquid precursor, in the form of an aerosol produced by an ultrasonic generator. The Si/C/N ex-HMDS powders have an average grain size in the nanometric range (20–40 nm). Their composition, determined by chemical analysis, is adjusted by varying the partial pressure of the carrier gas (NH3 ). Two series of powders in the intermediate range were studied, namely HMDS43 and HMDS44, which correspond to chemical composition C/N equal to 0.67 and 0.58, respectively, for AF samples. The XPS measurements were performed on AF powders and after heat treatment at Ta = 1500◦ C and 1600◦ C during 4 h under N2 . The AF powders were stored under inert atmosphere. The analysed samples were pressed pellets prepared from pure laser synthetised powders.

©2003 Taylor & Francis

Core levels The Si 2p, C 1s and N 1s core levels spectra from HMDS43 and HMDS44 powders are reported in Figures 5.3.10–5.3.12 for AF samples and for powders after heat treatment at 1500◦ C and 1600◦ C for 4 h. The Si 2p spectra (Figures 5.3.10(a) and (b)) are broad and asymmetric peaks which reveal the presence of several contributions in each spectrum. A decomposition of the spectra into Voigt functions is achieved, by assuming that each contribution corresponds to stoichiometric species: pure Si (99.5 eV), SiC (101.1 eV), silicon nitride Si3 N4 (101.8 eV), or silicon dioxide SiO2 (103.2 eV) and no intermediate bonds such as Si CX N4−X are considered (Sénémaud et al. 1998). For AF samples (Figures 5.3.10(a) and (b)), the maximum of the peak is close to the BE of Si3 N4 . The overall shape of the spectrum is significantly wider in HMDS43 than in HMDS44, for which a marked asymmetry towards low BE is observed. The Si 2p peaks can be fitted with a sum of two Voigt functions, the energy separation of which is close to the energy distance SiC Si3 N4 and, consequently, are attributed to a sum of silicon carbide and silicon nitride contributions, respectively. In agreement with the results of chemical analysis of the samples, the relative proportion Si3 N4 /SiC (noted G) is much higher in HMDS44 than in HMDS43; G is respectively 2.2 and 4.2 for HMDS43 and HMDS44. The two components are broad (about 2 eV) and their energy position is not well defined (within ± 0.5 eV). This result reveals that the AF powders exhibit a high degree of chemical disorder. The existence of mixed tetrahedra with C and N atoms around a same Si atom, which would correspond to peaks located at intermediate energy positions, is likely (Gheorghiu et al. 1997, 1998b). Hydrogen is present in AF powders (Musset et al. 1994) and the presence of H bonds in the network can contribute to increase the chemical disorder and consequently the width of the lines. After a heat treatment of powders at 1500◦ C, the Si 2p line from both samples exhibits a high BE tail which can be attributed to the presence of silicon dioxide. Besides this SiO2 (a)

SiO2 Si 2p

Si3N4

(b)

HMDS43

Si 2p

AF

Intensity (au)

1500°C

Si3N4

SiC

HMDS44

1600°C

1600°C Intensity (au)

SiO2

SiC

1500°C

AF 104 100 Binding energy (eV)

104 100 Binding energy (eV)

Figure 5.3.10 Si 2p core levels for AF samples and for 1500 ◦ C and 1600◦ C annealed samples (a) HMDS43 and (b) HMDS44.

©2003 Taylor & Francis

contribution, a shift of the maximum of the line towards higher BE is observed in HMDS44 as compared to HMDS43. The decomposition of the lines into peaks from silicon carbide and silicon nitride shows that the apparent shift is due to the variation of the relative proportion of the Si3 N4 to SiC contributions (G ratio) which is respectively 1.1 and 2.1. In HMDS43, the relative proportion of nitride and carbide is approximately the same whereas in HMDS44, Si N bonds are dominant. Moreover, the widths of the components are significantly lower in HMDS43 than in HMDS44 (about 1.6–1.7 instead of 1.9 eV). In 1600◦ C annealed samples, the presence of SiO2 is still noted with about the same proportion as in 1500◦ C annealed sample. It is noteworthy that the factor G decreases considerably from AF samples to 1500◦ C and 1600◦ C samples: G is respectively 0.4 and 0.8 for 1600◦ C annealed HMDS43 and HMDS44 samples. The contribution from SiC is dominant for both samples; it is higher for HMDS43 than for HMDS44; once again, this result is in good agreement with the chemical analysis results. The energy positions of SiC and Si3 N4 contributions are obtained, in this case, close to the reference binding energy values; their energy widths are significantly narrower than in AF samples. It is noteworthy that Si 2p lines from SiC are only 1.6 and 1.5 eV wide respectively in 1600◦ C annealed HMDS43 and HMDS44 samples; these values correspond to a perfectly well-ordered SiC network. The nitride contribution is narrower in HMDS43 (1.6 eV) than in HMDS44 (1.8 eV); in both cases the result reveals a well organised network. C 1s spectra observed from HMDS43 and HMDS44 samples (Figures 5.3.11(a) and (b)) exhibit quite different shapes according to the composition and to the heat treatment. In HMDS43 AF sample, C 1s consists of a quasi-symmetric broad line; in HMDS44, the peak is split into two components. The decomposition into Voigt functions reveals in both cases the existence of two peaks. In HMDS43 and HMDS44 the high BE component corresponds to C N (285.9 eV) (Dufour et al. 1994), possibly mixed to C H (285.0 eV) bonds; the low BE component is clearly resolved from the high BE one in HMDS44 and thus it can be associated

(a)

C9N C9C C9Si C9O C 1s

C9H

(b)

HMDS43

C 1s

Intensity (au)

1500°C

C9H

HMDS44

1600°C

1600°C Intensity (au)

C9N C9C C9Si C9O

1500°C

AF AF 286 282 Binding energy (eV)

286 282 Binding energy (eV)

Figure 5.3.11 C 1s core levels for AF samples and for 1500◦ C and 1600◦ C annealed samples (a) HMDS43 and (b) HMDS44.

©2003 Taylor & Francis

to C Si (283.2 eV) (Driss-Khodja et al. 1992). In HMDS43, the low BE component is closer to the high BE component and reveals more probably the presence of C C bonds (sp2 ). After a 1500◦ C annealing of sample HMDS43, the C 1s line is strikingly sharp and the energy maximum of the narrow peak corresponds to C C bonds (284.3 eV) (Diaz et al. 1996). The small width of this peak (only 1.2 eV) reveals unambiguously the presence of small crystallites. This sharp peak is accompanied by a low BE shoulder corresponding to C Si bonds in small proportion and by a tailing towards high BE which reveals the presence of C N bonds (or C H). A 1500◦ C annealing of HMDS44 sample gives quite a different shape for C 1s spectra. The line is well fitted by both C C and C N contributions; a low proportion of C Si bonds is present as revealed by the low BE component. The high BE tailing of the spectrum can be interpreted by the existence of a few C O species at the surface of the sample. After a 1600◦ C annealing, the most intense line is attributed to C Si (carbide) for both HMDS43 and HMDS44 samples. Towards higher BE, the tailing is interpreted by the presence of C N bonds. The presence of C C bonds, the energy position of which is intermediate between those of C N and C Si contributions, is not excluded. N 1s spectra from HMDS43 and HMDS44 samples are reported in Figures 5.3.12(a) and (b). In AF HMDS43 powder, a broad line is observed which can be fitted with a sum of two Voigt functions having approximately the same intensity. The low BE peak can be associated to N Si3 group (398.0 eV) (Gheorghiu et al. 1992); the higher energy component can be associated to N H and N C bonds; these two contributions are only separated by 0.5 eV and located at respectively 400.1 and 399.5 eV (Bischoff et al. 1987; Rossi et al. 1994). In fact, AF powders contain a non-negligible proportion of H atoms (Musset et al. 1994) which are essentially present in N Hn groups. For HMDS44, the width of the observed line is noticeably lower than in HMDS43. A tailing of the line on the low BE side is noted. The decomposition reveals the existence of a strong peak attributed to N H and N C bonds and a lower intensity peak associated to N Si bonds. Consequently, the significant decrease of the total width of the line is due to an appreciable decrease of the low BE contribution.

N9C N9H N9Si

(a)

(b)

HMDS43

N 1s

N9Si

1600°C Intensity (au)

Intensity (au)

AF

N9C HMDS44

N 1s

1600°C

1500°C

N9H

1500°C

AF

400 396 Binding energy (eV)

400 396 Binding energy (eV)

Figure 5.3.12 N 1s core levels for AF samples and for 1500 ◦ C and 1600◦ C annealed samples (a) HMDS43 and (b) HMDS44.

©2003 Taylor & Francis

After a 1500◦ C annealing of the sample, most H atoms are eliminated and, consequently, the proportion of N H bonds is much lower. In this case, the N 1s peak is asymmetric and reveals the existence, in both samples, of two components that we attribute to N Si and N C bonds, respectively. In HMDS43, the low BE (N Si) peak is more intense than the high BE (N C) one. On the contrary, in HMDS44, the high BE component dominates. After a 1600◦ C annealing, both HMDS43 and HMDS44 spectra exhibit contributions at approximately the same energies. In HMDS43, the relative intensities are about the same as at 1500◦ C. In HMDS44, the relative intensities are completely modified, as compared to 1500◦ C, and the N Si contribution becomes the most important. The splitting of the peak is well observed in 1600◦ C annealed samples. This effect is principally due to a narrowing of N Si component in both samples.

XPS valence band spectra The VB distributions from HMDS43 and HMDS44 powders in AF samples and in samples annealed at 1500◦ C and 1600◦ C are reported in Figures 5.3.13(a) and (b). The VB of the AF samples consists, for both spectra, of a wide band A in the range 0–14 eV; the maximum of this band is located at about 12 eV BE. This first band is followed by a deep minimum around 15 eV; an intense peak D is observed at 20 eV, and a lower structure E at about 26 eV. For 1500◦ C annealed samples, the overall shape of the spectral curves remains approximately the same and only small modifications are observed as compared to AF powders. The maximum A is not significantly shifted but it becomes slightly broader, especially in HMDS43. The minimum between A and D is broader and slightly filled up; in HMDS43, the sharp minimum is replaced by a 1 eV wide step. Peak D is noticeably narrower; the energy position of features D and E are not modified. The minimum between D and E is more pronounced. These modifications are approximately the same in both samples. Striking changes of the VB distributions are observed after the annealing at 1600◦ C, in both HMDS43 and HMDS44 samples. The maximum of peak A is shifted by 2 eV towards (a)

E

D B

A HMDS43

(b)

1500°C

D B

A HMDS44

1600°C Intensity (au)

Intensity (au)

1600°C

E

1500°C

AF

AF EF 30 20 10 Binding energy (eV)

30 20 10 EF Binding energy (eV)

Figure 5.3.13 XPS VB spectra for AF, 1500◦ C and 1600◦ C annealed samples (a) HMDS43 and (b) HMDS44.

©2003 Taylor & Francis

lower BE for both samples; it becomes sharper and a shoulder at about 7.5 eV BE is clearly observed. The most striking modification, as going from Ta = 1500–1600◦ C, consists of a dramatic decrease of peak D intensity and the appearance of an additional peak labelled B at about 15 eV. It is noteworthy that peak B is well resolved from A and from D, and that wide minima occur between A, B and D. The intensity of B is slightly higher than that of D for both HMDS43 and HMDS44 samples. As far as peak E is concerned, its intensity is strongly enhanced for 1600◦ C annealed samples as compared to 1500◦ C annealed ones. The modifications of the spectral distributions observed as Ta increases from 1500◦ C to 1600◦ C can be discussed by reference to VB distributions of the stoichiometric compounds SiC (Driss-Khodja et al. 1992) and Si3 N4 (Sénémaud et al. 1993), which are reported in Figures 5.3.14 and 5.3.15. The energy positions of features A, B and D are given in Table 5.3.2 for HMDS43 and HMDS44 powders (AF, Ta = 1500◦ C and 1600◦ C), comparatively to stoichiometric compounds SiC and Si3 N4 . Let us recall that in SiC, peak A corresponds to C 2p states mixed to Si 3s and Si 3p states and peak B to C 2s states mixed to Si 3s and Si 3p states (Driss-Khodja et al. 1992) In Si3 N4 , peak A is attributed to N 2p states mixed to Si 3s and Si 3p states, the lower BE feature being attributed to the N 2pπ lone pair; peak D is associated to N 2s states mixed to Si 3s and Si 3p states (Sénémaud et al. 1993). For HMDS43 and HMDS44 nanopowders annealed at 1600◦ C, the shape of band A is intermediate between those of SiC and Si3 N4 : the maximum of the peak is located between the maximum of A in SiC and in Si3 N4 . Both peaks B and C are observed for the samples annealed at 1600◦ C; they correspond to Si 3s, 3p states mixed to C 2s states (B) and respectively N 2s states (D); they reveal unambiguously the existence of both Si C4 and Si N4 groups in these systems. The results obtained from the analysis of the N 1s core level have shown (see Section 4.1.1) that in 1500◦ C annealed samples, N atoms are bonded to both Si (N Si bonds) and to C atoms (N C bonds). The contributions of these bonds in the VB are located in the energy range corresponding to peak D (about 20 eV BE). For these samples annealed at 1500◦ C, the study of C 1s core level has shown that C atoms are essentially bonded to C atoms (C C bonds) or to N atoms (C N bonds); the contribution of C Si bonds is very low (even negligible), as revealed by the very low intensity in the energy range concerned. The presence of C C bonds (either sp2 or sp3 ) induces VB states around 17 eV BE, accompanied by a sharp peak at 11.5 eV in diamond (Mac Feely et al. 1974). These states are located well apart from A

SiC

B Intensity (au)

E

30

20

10

EF

Binding energy (eV)

Figure 5.3.14 XPS VB spectrum for stoichiometric SiC.

©2003 Taylor & Francis

D E

A

Intensity (au)

30

Si3N4

20

10

EF

Binding energy (eV)

Figure 5.3.15 XPS VB spectrum for stoichiometric Si3 N4 .

Table 5.3.2 BE referred to the Fermi level EF for HMDS43 and HMDS44 HMDS43 and HMDS44

A (eV)

B (eV)

D (eV)

AF Ta = 1500◦ C Ta = 1600◦ C SiC Si3 N4

12.0 12.0 10.0 (5.7) 9.5 (3.9 7.6) 11.4

— — 14.5 14.4 —

20.0 20.0 20.0 — 18.9

C Si(Si C4 ) contribution which occurs at 14.4 eV. As a consequence, the contribution of C C bonds induces in the VB a broadening of peak D and a filling up of the minimum between peak D and peak A; no separate peak is expected. Consequently, the great intensity of peak D, in AF and 1500◦ C annealed samples, is consistent with an associate contribution of N C and C C bonds in the VB. The slight filling up of the VB distribution in the energy range of peak B reveals the low contribution of C Si bonds to the VB. After annealing at 1600◦ C, there is a complete reorganisation of bonds around C and N atoms which are involved mainly in Si C4 and Si N4 groups. It is to be noted that some effects may be due to rehybridisation of orbitals (e.g. C sp2 –C sp3 ). The data obtained for the VB distribution, as a function of heat treatment, are in good agreement with the results observed for the Si 2p, N 1s and C 1s core levels. In particular, after a 1500◦ C annealing, the presence of the strong peak at about 20 eV BE observed in the VB spectrum can be interpreted by the fact that N atoms are involved in both N Si and N C bonds; the existence of both types of bonds involving a broadening of the corresponding N 2s peak of the VB around BE = 20 eV. In these samples, C atoms are essentially involved in C N and C C bonds, as revealed by the analysis of core levels spectra, and this can explain why the VB distribution exhibits no contribution in the BE range corresponding to C 2s states from C Si bonds.

©2003 Taylor & Francis

Our results are in quite good agreement with a structural model composed of SiN3 C tetrahedra linked through C N bonds obtained from EXAFS measurements on the AF SiCNHMDS samples (Ténégal et al. 1996). 5.3.4.2

Si/C/N/Al/Y nanopowders

The samples were prepared by laser pyrolysis of a liquid mixture of hexamethyldisilazane (HMDS) and dissolved solid Al,Y-isopropoxides. This method is already described in Section 5.3.4.1 (HMDS powders). Amorphous ultrafine (20–40 nm) nanopowders are produced by laser pyrolysis of a liquid precursor. As for the SiCN-HMDS powders, the Si/C/N/Al/Y nanopowders may have variable N contents depending on the proportion of NH3 in the carrier gas (which is a mixture of argon and ammoniac). The powders were annealed at 1600◦ C for 1 h under N2 atmosphere. The atomic compositions of the studied powders (amorphous and annealed) as well as their designation are reported Table 5.3.3. The Al + Y content does not exceed 1 at% in the AF powders. Evolution of the Al and Y content in the superficial layers with annealing treatment CORE LEVELS

In order to get information about atomic concentrations in the superficial layers of the powder grains and about their evolutions with annealing treatment, the intensities of the photoemission peaks (core levels) were compared. The core levels were normalised with respect to the number of measured scans. Figure 5.3.16 gives the Al 2p and Si 2p core levels for the intermediate composition (C/N = 0.78) for the AF and annealed sample and one can observe a strong increase of the Al 2p/Si 2p intensities ratio with annealing. From the measurements of the peak intensities and from the values given by Scofield (Scofield 1976) for the photoionisation cross-sections, we deduced the Al/Si atomic ratio in the superficial layers: IAl2p σSi2p Al = Si ISi2p σAl2p

(5.3.1)

where IAl2p and ISi2p are the intensities of the Al 2p and Si 2p photoemission peaks, σAl2p and σSi2p the photoionisation cross-sections given by Scofield (Scofield 1976). Table 5.3.4 reports the results of the intermediate composition. Al/Si atomic bulk (global) ratios are also given. The comparison between bulk and superficial ratio (from superficial layers) underline a strong increase of the Al/Si superficial ratio with annealing, whereas the bulk ratio tends Table 5.3.3 Atomic composition of the studied Si/C/N/Al/Y/O nanopowders Powders

Si (at%)

N (at%)

C (at%)

O (at%)

C/N

Al (at%)

Y (at%)

HSAlY17, AF HSAlY16, 1600◦ C/1 h/N2 HSAlY17, 1600◦ C/1 h/N2 HSAlY18, 1600◦ C/1 h/N2

33.5 41.3 44.4 47.3

33.5 52.1 29.5 8.3

26 4.9 24.1 41.9

6.2 1.5 1.6 1.9

0.78 0.09 0.82 5.05

0.57 0.14 0.34 0.41

0.14 0.02 0.11 0.18

©2003 Taylor & Francis

(a)

1 eV

AF (C/N = 0.78)

HSAIY17

1600°C (C/N = 0.82) Intensity (au)

Al 2p 60

0 81

78

75 Binding energy (eV)

(b) AF (C/N = 0.78) 4000

72

0.5 eV

HSA117

1600°C (C/N = 0.82)

Intensity (au)

Si 2p

2000

0 108

105

102

99

Binding energy (eV)

Figure 5.3.16 Al 2p and Si 2p core levels for a powder with an intermediate C/N ratio (HSAlY17). (a) For the AF sample (C/N = 0.78) and (b) for the annealed one (1600◦ C, C/N = 0.82).

Table 5.3.4 Surface ratios deduced from XPS analysis compared to bulk ratios from atomic global compositions for HSAlY17

AF 1600◦ C/1 h/N2

IAI2p /ISi2p

AP/Si surface ratio (XPS)

AP/Si bulk ratio

0.004 0.022

0.01 0.03

0.02 0.01

to decrease. This result shows that Al atoms move towards the surface of the grains, when annealing is performed, and chemical shifts of the Al 2p and Si 2p peak by 1 and 0.5 eV (Figure 5.3.16) reveal that the chemical environment of Si and Al atoms changes as they reach the surface.

©2003 Taylor & Francis

As we want to evaluate in which proportions aluminium content increases in the superficial area of the grains, the atomic ratio of Al (1600◦ C/AF) is deduced from the measurements of the Al 2p peak intensities for the AF sample and for the annealed one (1600◦ C). As we compare ratio of the same atomic species (Al or Si) in two different samples, then equation (5.3.1) is no longer valid and differences of densities, textures (amorphous or crystalline powders) which modify λ (the elastic mean free path for the measured electron kinetic energies), and of the size of the grains, must be taken into account in the evaluation of the atomic surface ratio. For the crystalline sample (1600◦ C), the density is about 1.5 times more than the one found in the AF sample (amorphous powder) and then the elastic mean free path is reduced in the crystalline sample. In a first approximation, the number of atoms probed by XPS for the two samples is almost the same: in the annealed sample, the distances covered by the electrons are shorter than in the AF sample but the number of atoms per unit of volume is increased. In the crystalline sample (1600◦ C), the increase of the average size of the grains (by 30–40%) induces a reduction of the atomic fraction seen by XPS: sensitivity to superficial layers is increased for the annealed sample. Table 5.3.5 reports the intensities ratio for Al 2p and Si 2p core levels as well as the atomic ratios given for the bulk. Collectively, Si content increases slightly (ratio of 0.75) when annealing is performed whereas the Al content (ratio of 1.68) decreases. The analysis of superficial ratios (relative to superficial layers) peak intensities confirms a strong increase of the Al content in the superficial layers after annealing: for the annealed powder (Figure 5.3.16(a), Table 5.3.6), the intensity for the Al 2p peak is five times greater than that found in the AF sample. The strong difference between atomic compositions and measured

Table 5.3.5 Ratios of the atomic concentration between the AF sample and the annealed one (1600◦ C) for Al and Si. Measurements were performed in the same experimental conditions for both samples (same preparation, same detection angle). Samples HSAl16-17-18

A1 2p Si 2p

IAF/I 1600◦ C

AF/1600◦ C surface (XPS)

AF/1600◦ C bulk

0.17 1

0.17 1

1.68 0.75

Table 5.3.6 Results of the simulations with Voigt functions of the Al 2p and Y 3d core levels Powders

HSAlY17, AF HSALY16, 1600◦ C/1 h/N2 HSALY17, 1600◦ C/1 h/N2 HSALY18, 1600◦ C/1 h/N2 References: YSiAlON Sintered sample

©2003 Taylor & Francis

Al 2p

Y 3d

AlO4 + AlO6

AlN4

E(eV)

Ŵ(eV)

E(eV)

Ŵ(eV)

E(eV)

Ŵ(eV)

E(eV)

Ŵ(eV)

75.7 75 74.9 74.8

2.6 1.8 1.9 1.5

— 73.8 73.8 73.8

— 1.8 1.6 1.5

— 160.9 160.9 160.7

— 1.7 2.0 1.9

— 159.0 159.0 158.6

— 1.8 1.8 1.8

74.9 74.9

1.8 2.2

73.8 73.8

1.7 2.1

160.7 160.0

2.1 1.8

158.6 158.2

2.2 1.7

3/2

5/2

AF (C/N = 0.78) 1600°C (C/N = 0.82) Intensity (au)

HSAl17 Si 2s Y 3d

170

165

160

155

Binding energy (eV)

Figure 5.3.17 Y 3d core levels (3/2 and 5/2) for the intermediate powder (HSAlY16, HSAlY17, HSAlY18) (AF and the annealed sample).

AIO4+ (1600˚C)

AIO6

AlN4 AI 2p

Intensity (au)

HSAIY18 (C/N = 5.05)

HSAIY17 (C/N = 0.82)

HSAIY16 (C/N = 0.09)

78

75

72

Binding energy (eV)

Figure 5.3.18 Al 2p core levels for the annealed samples (1600◦ C (HSAlY16, HSAlY17, HSAlY18) and results of the simulations with Voigt functions.

values for Al 2p confirms the increase of Al concentration at the surface of the grain. The same result is obtained for yttrium. Measurements performed on the as-prepared sample show that Y 3d core levels are not detected (Figure 5.3.17). On the other hand, XPS spectra display a clear Y 3d signal for the annealed sample (Figure 5.3.17), whereas yttrium global content does not change. As the temperature increases until 1600◦ C, SiC and Si3 N4 crystallites appear and Al and Y atoms are totally or partially rejected at the surface of the crystallites.

©2003 Taylor & Francis

AIO4

+AIO6 AlN4 AI 2p

Intensity (au)

Sintered sample

YSiAION

78

75 72 Binding energy (eV)

Figure 5.3.19 Al 2p core levels for the references compounds (sintered sample and YsiAlON) as well as the reconstructions.

Figure 5.3.16 shows that the Si 2p and Al 2p XPS peaks move towards lower BE when annealing is performed, indicating that Al and Si atoms are bonded to less electronegative atoms at 1600◦ C than in the AF powder. The Al 2p peak is located at 75.8 eV in the AF sample and its large width at half maximum (2.6 eV) indicates that Al atoms are embedded in a disordered network. This energetic position indicates the presence of voids in the network. After annealing, the peak position is between 75 and 74 eV depending on the C/N atomic ratio of the sample (Figure 5.3.18). Figure 5.3.18 presents the results of the simulations performed on Al 2p core levels for the annealed samples. Three compositions were analysed: a rich C powder (C/N = 5.05), an intermediate one (C/N = 0.82) and a rich N powder (C/N = 0.09). The results obtained for the reference compounds (YSiAlON vitreous phase and a sintered sample) are also given (Figure 5.3.19). The solid sintered sample is a mixture of silicon nitride (Si3 N4 ), SiC, β-SiAlON (crystalline phase) and YSiAlON. First of all, the width of the Al 2p peaks is quite similar in the nanopowders (1600◦ C) and in the reference compounds. The results of the simulations show that Al 2p peaks can be simulated using two Voigt functions, the energetic position of which is that found for aluminium in AlN and Al2 O3 (Muilenberg 1979; Briggs and Seah 1996): AlN4 and AlO4 + AlO6 sites are detected in the annealed powders. The proportion of Al N bonds increases with the C/N atomic ratio. Concerning reference compounds, one finds the same contributions to Al 2p core levels but the proportion of AlN4 sites is lower than that found in the powders. These results are in excellent agreement with those found by XAS analysis by Flank et al. (1999) on the same powders, (see part XAS): EXAFS results show that crystalline AlN is present in the powders at 1600◦ C. The XANES study revealed also that Al atoms are embedded in an amorphous phase in the as-prepared powders. The results of our simulations are given in Table 5.3.6. The FWHM of the contributions is quite similar for all the compounds. However, we observed the lowest width for the rich C powder, which is in agreement with a preferential formation of crystalline AlN in that powder.

©2003 Taylor & Francis

Y 3d core levels (3/2 and 5/2) for the rich C powder are presented in Figure 5.3.20 (annealed powder) as well as those from the YSiAlON phase. The position and width of the peaks are similar for both compounds. The results of the simulations, given in Table 5.3.4 for the nanopowders and for the references, show that the chemical environment for yttrium atoms in the nanopowders is close to that found in the YSiAlON vitreous phase. The comparison of the simulations for the YSiAlON reference phase and for the sintered compound reveals a slight difference of the final parameters (E and Ŵ), due to differences in composition between the YSiAlON reference and the YSiAlON phase inside the sintered composite. The results of the decompositions (annealed samples) are given in Figure 5.3.21 for Si 2p core levels, in Figure 5.3.22 for N 1s core levels and in Figure 5.3.23 for C 1s ones. Si 2p core levels are modelised with four components relative to Si Si4 , Si C4 , Si N4 and Si O4 sites in crystalline compounds (Si, SiC, Si3 N4 and SiO2 ). As the content of C increases, the proportion of crystalline SiC increases and the proportion of Si C4 sites increases (Figure 5.3.21). A weak contribution from Si Si bonds is observed for the rich N powder. The study of Si 2p core levels confirms the results of XRD analysis, which revealed the existence of Si3 N4 and SiC crystalline phases with an increasing proportion of SiC when the C content of the powder increases. These results are in agreement with those of the previous structural studies performed on SiCN nanopowders (Sénémaud et al. 1998; Musset et al. 1994; El Kortobi et al. 1996; Ténégal et al. 1996, 1998) with different C/N ratios. The study of the N1s core levels (Figure 5.3.22) confirms the existence of N Si bonds (near 398 eV), N Al bonds (396.8 eV) and N C bonds (399–399.5 eV) in the annealed powders. The main contribution to the N 1s core level comes from N Si bonds (even for the rich C powder). The proportion of N C bonds is more important in the powder with intermediate C/N ratio. Contribution at higher BE (near 401 eV) is attributed to N O bonds. The existence of crystalline AlN at the surface of the grains is confirmed by analysis of the N 1s core levels.

3/2

(1600°C)

5/2 Y 3d

Intensity (au)

YSiAION

HSAIY18

HSAIY17

HSAIY16 165

162

159

156

Binding energy (eV)

Figure 5.3.20 Y 3d core levels for annealed HSAlY16, HSAlY17, HSAlY18 as well as the results of the simulations with Voigt functions.

©2003 Taylor & Francis

Si9O4

Si9C4 Si9N4 Si9Si4

Intensity (au)

Si 2p C/N = 5.05

HSAIY18

C/N = 0.82

HSAIY17

C/N = 0.09

HSAIY16

105

102 Binding energy (eV)

99

Figure 5.3.21 Si 2p core levels for the annealed samples HSAlY16, HSAlY17, HSAlY18. One can observe the increase of the SiC content with the increase of the C/N ratio.

N9O

N9C

N9Si3

N9Al4

N 1s

Intensity (au)

HSAIY18 C/N = 5.05

HSAIY17 C/N = 0.82

HSAIY16 C/N = 0.09 402

399 396 Binding energy (eV)

Figure 5.3.22 N 1s core levels for annealed HSAlY16, HSAlY17, HSAlY18. The simulations confirm the existence of an AlN crystalline phase at the surface of the grains. N C bonds are also detected.

©2003 Taylor & Francis

The study of C 1s core levels confirms the existence of C N bonds (286 eV), while C H bonds from surface contamination are also detected as well as contributions from C Si bonds in silicon carbide. Contrary to the case of N 1s core levels, the N and C rich powders display very different shapes for C 1s core levels. Indeed, for the C rich powder, the main contribution to the core level comes from C Si bonds in silicon carbide, the shoulder at 285–286 eV is expected to come from C H and C N bonds, but for the rich N powder, the intensity of the C Si contribution strongly decreases and the main contribution to the C 1s core level is that from C N bonds. These results show that C N bonds are stable through annealing treatments. These results are in good agreement with the previously detailed structural study of Si/C/N nanopowders produced by laser pyrolysis of hexamethyldisilazane (Ténégal et al. 1996, 1998, see the XAS part) which underline the existence of homogeneous amorphous C substituted Si3 N4 -type structures (1–2 nm) for annealing temperatures below 1500◦ C. Mixed tetrahedra SiCX N4−X and C N bonds were detected by XPS, RMN and XAS (Sénémaud et al. 1998; El Kortobi et al. 1996; Ténégal et al. 1996). When the annealing temperature increases beyond 1500◦ C, the N departure allows the local formation of SiC germs and powders crystallise in a mixture of Si3 N4 and SiC, the proportions of which are dependent on the initial C/N ratio (in the AF powder). Our results show that when the C content decreases, the stable phase for C atoms is no longer a silicon carbide phase but a C substituted silicon nitride one. XPS VALENCE BAND SPECTRA

VB spectra are given in Figure 5.3.24 for the three powders annealed at 1600◦ C. For the intermediate composition (HSAlY17), the corresponding spectra for powders without Al and Y is that of HMDS43 or HMDS44 given in Section 4.1 (Figures 5.3.13 and 5.3.14). The overall shape of the VB spectra for HSAlY17 (Figure 5.3.24) is similar to that obtained for HMDS43 and HMDS44 (Figures 5.3.13, 5.3.14 in Section 4.1.2). In SiC, peaks A and B are C9N

C9H

C9Si4

C 1s

Intensity (au)

HSAIY18 C/N = 5.05

HSAIY17 C/N = 0.82

HSAIY16 C/N = 0.09 288

285 282 Binding energy (eV)

Figure 5.3.23 C 1s core levels for annealed HSAlY16, HSAlY17, HSAlY18. C N bonds are dominant in the rich N sample at 1600◦ C.

©2003 Taylor & Francis

HSAIY18

Intensity (au)

E D B

A HSAIY17

HSAIY16 30

20 10 EF Binding energy (eV)

Figure 5.3.24 VB distributions in the pre-alloyed HSAlY16, HSAlY17, HSAlY18 powders annealed at 1600◦ C.

respectively attributed to mixtures of C 2p states with Si 3s–Si 3p states and C 2s states with Si 3s–Si 3p states. In silicon nitride, A and D are respectively attributed to mixtures of N 2p states with Si 3s–Si 3p states and N 2s with Si 3s–Si 3p states (see Section 4.1.2). The incorporation during the laser synthesis of Al and Y atoms does not change the shape of the VB. However, a slight decrease of the D peak (N 2s states mixed with Si 3s–Si 3p states) in the pre-alloyed powder (HSAlY17) is observed. The VB spectra for the rich N powder (HSAlY16) is quite different from that found for the silicon nitride reference (Figure 5.3.15 in Section 4.1.2). Modulations from 12 eV to the Fermi level observed on the silicon nitride VB spectra are not detected on the VB of HSAlY16. We expect to explain this difference by the presence of C and Al atoms in the network of silicon nitride which leads to a smoothing of the structures near the top of the VB. The presence of Al atoms in superficial layers either in AlN or SiAlON(C) crystalline phases at 1600◦ C can explain the VB smoothing from 12 eV to the Fermi edge. The VB distribution for HSAlY18 is close to that found in crystalline SiC (Figure 5.3.14, Section 4.1.2).

References Bischoff, J. L., Kubler, L. and Bolmont, D. (1987) J. Non-Cryst. Solids, 97–98, 1407. Bouisset, E., Esteva, J. M., Karnatak, R. C., Connerade, J. P., Flank, A. M. and Lagarde, P. (1991) Phys. Rev., B24, 1609. Briggs, D. and Seah, M. P. (1996) in Practical Surface Analysis, 2nd edn, vol. 1; Auger and X-ray Photoelectron Spectroscopy, John Wiley & Sons. Cauchetier, M., Croix, O. and Luce, M. (1988) Adv. Ceram. Mater., 3, 548. Cauchetier, M., Croix, O., Herlin, N. and Luce, M. (1994) J. Am. Ceram. Soc., 77, 993. Cauchetier, M., Croix, O., Luce, M., Baraton, M. I., Merle, T. and Quintard, P. (1990) J. Eur. Ceram. Soc., 8m, 215. Diaz, J., Paolicelli, G., Ferrer, S., and Comin, F. (1996) Phys. Rev., B54, 8064. Driss-Khodja, M., Gheorghiu, A., Dufour, G., Roulet, H., Sénémaud, C. and Cauchetier, M. (1996) Phys. Rev., B53, 4287. Driss-Khodja, M., Dufour, G., Gheorghiu, A., Roulet, H., Sénémaud, C., Cauchetier, M., Croix, O. and Luce, M. (1992) Mater. Sci. Eng., B11, 97.

©2003 Taylor & Francis

Driss-Khodja, M., Le Corre, A., Sénémaud, C., Gheorghiu, A., Thèye, M. L., Allain, B. and Perrin, J. (1989) J. Non-Cryst. Solids, 114, 489. Driss-Khodja, M. (1993) Thèse de l’Université Pierre et Marie Curie, Paris, 93PA066357; http://corail.sudoc.abes.fr80. Dufour, G., Poncey, C., Rochet, F., Roulet, H., Sacchi, M., De Santis, M. and De Crescenzi, M. (1994) Surf. Sci., 319, 251. El Kortobi, Y., Sfihi, H., Legrand, A. P., Musset, E., Herlin, N. and Cauchetier, M. (1996) Coll. Surf. A: Phys. Eng. Asp., 115, 319–327. Fang, R. C. and Ley, L. (1989) Phys. Rev., B40, 3818. Finocchi, F. and Gallia, G. (1994) Phys. Rev., B50, 7393. Finocchi, F., Gallia, G., Parrinello, M. and Bertoni, C. M. (1993) Physica, B85, 379. Flank, A. M., Armand, X., Cauchetier, M. and Mayne, M. (1999) J. Synchr. Rad. 6, 512. Gheorghiu, A., Dufour, G., Sénémaud, C., Herlin, N., Musset, E., Cauchetier, M. and Armand, X. (1997) J. Phys. III, 7, 529. Gheorghiu, A., Sénémaud, C., Roulet, H., Dufour, G., Moréno, T., Bodeur, S., Reynaud, C., Cauchetier, M. and Luce, M. (1992) J. Appl. Phys., 71, 4118. Gheorghiu-de La Rocque, A., Dufour, G., Bonnefont, P. A., Sénémaud, C. and Cauchetier, M. (1998a) MRS Symp. Proc., 501, 115. Gheorghiu-de La Rocque, A., Dufour, G., Sénémaud, C., Cauchetier, M. and Legrand, A. P. (1998b) Proceedings of International Symposium of Carbon Science and Technology for New Carbons, Tokyo, p. 424. Himpsel, F. J., Mac Feely, F. R., Taleb-Ibrahimi, A., Yarmoff, J. A. and Hollinger, G. (1988) Physics and Chemistry of SiO2 and Si/SiO2 Interface (eds C. R. Helms and B. E. Deal), Plenum Press, New York. Hollinger, G., Sferco, S. J. and Lannoo, M. (1988) Phys. Rev., B37, 7149. Jentoft, C., Weinberg, G., Wild, U. and Schölgl, R. (1999) Precursor-derived Ceramics (eds J. Bill, F. Wakai and F. Aldinger), Wiley-VCH. p. 175. Kärcher, R., Ley, L. and Johnson, R. L. (1984) Phys. Rev., B30, 1896. Mac Feely, R. F., Kowalczyk, S. P., Ley, L., Cavell, R. G., Pollak, R. A. and Shirley, D. A. (1974) Phys. Rev., B9, 5268. Martinengo, H., Musset, E., Herlin, N., Armand, X., Luce, M., Cauchetier, M., Roulet, H., Gheorghiu, A., Dufour, G. and Sénémaud, C. (1996) Silicates Industriel, 1–2, 9. Muilenberg, G. E. (1979) Handbook of X-ray Photoelectron Spectroscopy, Perkin Elmer. Musset, E., Cauchetier, M., Herlin, N., Luce, M., Gheorghiu, A., Dufour, G., Roulet, H. and Sénémaud, C. (1994) High Temp. Chem. Proc., 3, 535. Robertson, J. (1991) Philos., Mag., B63, 47. Robertson, J. (1992) Philos. Mag., B66, 615. Rochet, F., Rigo, S., Froment, M., D’Anterroches, C., Maillot, C., Roulet, H. and Dufour, G. (1986) Adv. Phys., 35, 237. Rossi, F., Andre, B., van Veen, A., Mijnarends, P. E., Schut, H., Labohm, F., Dunlop, H., Delplancke, M. P. and Hubbard, K. (1994) J. Mater. Res., 9, 2440. Scofield, J. H. (1976) J. Elect. Spectro. Rel. Phen., 8, 129–137. Sénémaud, C. and Ardelean, I. (1990) J. Phys. Condens. Matter., 2, 8741. Sénémaud, C., Driss-Khodja, M., Gheorghiu, A., Harel, S., Dufour, G. and Roulet, H. (1993) J. Appl. Phys., 74, 5042. Sénémaud, C., Gheorghiu de La Rocque, A., Dufour, G. and Herlin, N. (1998) J. Appl. Phys., 84, 4945. Sénémaud, C., Laporte, D., André, J. M., Khérouf, R., Paquier, P. and Ringuenet, M. (1989) Proc. ECO 2 X-ray Instrumentation, (ed. R. Benattar), 1140, 297. Shirley, D. A. (1972) Phys. Rev., B5, 4709. Ténégal, François, Bouchet, B., Bellissent, R., Herlin, N., Cauchetier, M. and Dixmier, J. (1998) Philos. Mag., A78, 4, 803. Ténégal, F., Flank, A. M. and Herlin, N. (1996) Phys. Rev., B54, 12029. Zahorowski, W., Wiech, G., Mell, H. and Weiser, G. (1989) J. Phys. Condens. Matter, 1, 9571.

©2003 Taylor & Francis

5.4 Local order in Si/C/N nanopowders studied by X-ray absorption spectroscopy Study at the Si K edge François Ténégal, Anne-Marie Flank, Adriana Gheorghiu-de La Rocque and Christiane Sénémaud 5.4.1 X-ray absorption principles and data analysis 5.4.1.1

Principles

X-ray absorption spectroscopy (XAS) refers to the measurement of the X-ray absorption coefficient µ as a function of photon energy E. µ is given by the Beer–Lambert law: I0 I where I0 and I are the intensities of the incident and transmitted beams, respectively and x is the sample thickness. µ is a measurement of the absorption cross-section σ which is given by the Fermi golden rule formula. As the energy of the incident X-ray beam reaches the binding energy of a core level, one can observe a strong increase of the absorption coefficient, corresponding to the excitation of an electron from its core level to bounded vacant states or to the conduction band. The edge position depends on the chemical environment of the absorbing atom. Transitions occur within the dipolar approximation for the absorbing atom. For a K-shell absorption (1 s electron), due to the dipole selection rules, only transitions to p-states are allowed (!l = ±1) and the X-ray absorption spectrum reflects the local partial unoccupied p-density of state (DOS) around absorbing atoms. As the energy gradually increases, if the absorbing atoms are isolated, the absorption coefficient decreases monotonically and its shape can be evaluated by a Victoreen function. If the atoms are embedded in a condensed phase (liquid or solid), then one observes modulations which have been shown to depend on the local environment around the absorbing atom (Figure 5.4.1). The photoelectron which has a variable kinetic energy (depending on the photon energy) is scattered by the neighbours of the excited atom, and builds up interference patterns that contain structural information. In the early 1970s, Lytle et al. (1975) showed that a subsequent Fourier transform (FT) of the extended X-ray absorption fine structure (EXAFS) oscillations yields a pseudo-radial distribution function. As the mean free path of the photoelectron depends on its kinetic energy (Figure 5.4.2), one will distinguish two regimes. In the EXAFS regime (Ec > 50 eV) (Figure 5.4.1), the interaction is in a first approximation calculated using the single scattering approach. The initial state for the electron is the localised core level corresponding to the absorption edge (K, LI, LII, LIII, . . .). The final state is that of the ejected electron, which can be represented as an outgoing spherical wave originating from the X-ray absorbing atom which is backscattered by the neighbouring atoms. The final state is the sum of the outgoing and incoming waves that gives µx = ln

©2003 Taylor & Francis

XANES

(au)

EXAFS

E0 1800

2000

2200 E (eV)

2400

CB VB

.. .

h

LII (2p1/2), LIII (2P3/2) LI (2s) K (1s)

Core levels

Figure 5.4.1 Absorption coefficient (background was subtracted) for β-SiC (cubic). When the energy of the X-ray beam reaches the binding energy of a core level (E0 = 1841 eV), then an abrupt increase of µ is observed. As the energy increases (E > 1841 eV), modulations of the coefficient are clearly evidenced.

Au

100

Al Escape depth (Å)

Ag

Au

Au

Au Ag

10

Fe

1

10

C

Be W

Ag

1

Mo

Ag

100

W C

Be

Mo

Ag Be Mo

Ni

1000

10000

Electron energy (eV)

Figure 5.4.2 Mean free path for the electron as a function of its kinetic energy.

rise to the sinusoidal variation of µ known as EXAFS. For an unoriented sample with small or approximately Gaussian disorder, the EXAFS oscillations are given by (Teo 1986) µ(k) − µ0 (k) µ0 (k)  Fj (π, k, Rj ) (−2Rj / λj (k)) −2σj2 k 2  = S02 (k)(N)j e sin[2kRj + φj (k, Rj ) + 2δc ] e 2 kR j j

χ (k) =

(5.4.1)

©2003 Taylor & Francis

k is given by k = 4π 2

√ 2m(E − E0 ) h2

where E0 is the threshold energy, µ0 (k) is the atomic absorption (for isolated atoms) and µ(k) the absorption coefficient. The summation is made over the concentric coordination shell j around the absorbing atom at a distance Rj . Nj is the number of atoms of the same kind inside the coordination shell j . Nj and Rj are structural parameters. The electronic parameters Fj (π, k, Rj ) (the backscattering amplitude for the shell j ) and φj (k, Rj ) + 2δc (the phase function) were calculated for each kind of atomic pairs and tabulated by Teo and Lee (1979), McKale et al. (1986) and Rehr and Albers (1990). They can be extracted from model compounds. S02 (k) is an amplitude reduction factor which accounts for the intrinsic inelastic loss (concerning the absorbing atom), exp(−2Rj /λj (k)) is a reduction factor accounting for extrinsic inelastic process (concerning neighbouring atoms) and exp(−2σj2 k 2 ) is a reduction factor accounting for thermal and structural disorder, where σj is the Debye–Waller factor. The EXAFS formula (5.4.1) neglects multiple scattering events which were introduced in the EXAFS theory by Lee and Pendry (1975). Later on, Rehr and Albers (1990) gave a formulation of the EXAFS signal as a sum of the contributions from each possible scattering path Ŵ (single or multiple scattering path) for the photoelectron:  χ (k) = χŴ Ŵ

 S 2 (k)   0 f Ŵ (k) sin[2kR + φ Ŵ (k) + 2δc (k)]e−2σŴ2 k 2 e−2R/λ(k) = eff 2 kR

(5.4.2)

Ŵ

R is the effective path radius (the total length of the path Ŵ is 2R). σŴ is the DW factor Ŵ (k) is its effective scattering amplitude and φ Ŵ (k) + 2δ (k) its effective for the Ŵ path, feff c phase shift. The second formulation of the EXAFS signal accounts for multiple scattering events which must sometimes be included in the EXAFS analysis (Figure 5.4.3). The FEFF code developed by Rehr and Albers calculates the EXAFS signal using the second expression (5.4.2) (Rehr and Albers 1990). The calculation requires an atomic coordinate file from which a muffin-tin potential is calculated. The paths are numbered and selected using amplitude

1

2

Figure 5.4.3 Representation of a single scattering process (1) for the photoelectron and a multiple scattering process (2) for the photoelectron.

©2003 Taylor & Francis

filter paths importance criteria (Zabinsky et al. 1995). All the contributions Ŵ are individually calculated and gathered to give the EXAFS signal. Multiple scattering events become fundamental in the ‘low kinetic energy’ regime (Ec < 50 eV) where the mean free path of the photoelectron can be large. In that region, the photoabsorption spectrum is very sensitive to the medium-range order, as it is illustrated in Figure 5.4.4, where the XANES part of the XAS spectra for SiC in the crystalline beta form and in the amorphous phase are compared. The loss of medium-range order in the amorphous material washes out the structures in the near edge area. As well, some features of the near-edge structures may appear as fingerprints of the crystalline phases present in the sample (Figure 5.4.5).

E0 XANES

 (au)

EXAFS

c-SiC a-SixCy

1830

1860

1890

1920

E (eV)

Figure 5.4.4 XANES part for β-SiC and amorphous silicon carbide (a-SiC) prepared by RF sputtering. XANES features are strongly smoothed in the case of a-SiC.


-Si3N4

Absorption (au)

-Si3N4
-SiC B

C

1 0

1840

1850

1860

1820

E (eV)

Figure 5.4.5 Comparison of Si K edge XANES spectra for β-SiC, α-Si3 N4 , β-Si3 N4 .

©2003 Taylor & Francis

5.4.1.2

Experimental

An X-ray absorption experiment requires a white X-ray source of reasonable intensity to collect data. One fundamental reason for the development of the technique, in the 1970s, was the advent of the synchrotron radiation X-ray sources of high brilliance and intensity, which were improved since then. The photoabsorption measurements (HMDS powders) were carried out on the SA32 beamline (Super-ACO, LURE-Orsay) equipped with a double-crystal (InSb (111)) monochromator which allows an energy resolution of 0.7 eV. The incident beam was monitored by measuring the total electron drain current of an aluminium foil located downstream the monochromator. The photoabsorption EXAFS spectra at the Si K edge were recorded at 298 K, were collected in the transmission mode as a function of the incident photon energy, in the 1800–2500 eV energy range and with a 1 eV step. The transmitted beam was measured with an ionisation chamber filled by a low pressure of air (200 Torr). Measurements for the powders produced from gaseous precursor were carried out on the SB3 beamline which was equipped with a two crystal InSb (111) monochromator. The photoabsorption spectra were recorded at room temperature, in the transmission mode as a function of the incident photon energy, in the 1750–2550 eV range with a 2 eV step. The transmitted beam was measured with a proportional counter filled with a 220 Torr Ar/CH4 (90/10) mixture. More accurate measurements were done in the near edge region (1830–1910 eV, 0.2 eV step), as the XANES part of the absorption spectra has been shown to be a good fingerprint of the degree of crystallinity and of the nature of the silicon phases present in the compound (Sainctavit et al. 1991). As a good approximation very near the edge, which is sufficient to draw qualitative inferences, a constant background was subtracted in the pre-edge region from the as-recorded spectra. They were then normalised far from the edge (at 1910 eV). As an example, Figure 5.4.5 presents the XANES spectra of β-SiC, α-Si3 N4 , and β-Si3 N4 . All the XANES spectra were recorded during the same shift, and references were checked before and after the measurements, as well as crystalline Si and crystalline and amorphous SiO2 already reported in the literature (Lagarde and Flank 1986; Lagarde et al. 1992, 1993). 5.4.1.3

EXAFS data analysis

The first step is the extraction of the oscillatory part χ (k) from the absorption coefficient µ. The atomic absorption µ0 is modelised by a polynomial (4th or 5th order) fit of the modulations. The spectra are normalised using the Lengeler–Eisenberger (1980) approximation. The normalisation coefficient is given by the Heitler formula:   8(E − E0 ) µ1 (E) − µ0 (E) = [µ1 (E0 ) − µ0 (E0 )] 1 − (5.4.3) 3E0 where µ1 (E0 ) − µ0 (E0 ) is the experimental value of the jump at the threshold (E0 ). µ0 (E) is the atomic absorption given by the polynomial fit. Once the oscillatory part is extracted, FT in R space of the k 2 or k 3 weighted oscillations χ (k) is performed. A Kaiser (τ = 2.5) or Hanning window is applied to the k n χ (k) oscillation. FTs are calculated over the same energy range (2–10 Å−1 ) for all the samples and references compounds. Extraction of the oscillatory part and FT calculations must be executed using the same parameters for all the compounds that will be compared. FT calculations give according to the expression (I) a pseudo-radial distribution function (uncorrected from the phase shift) of the coordination shells surrounding the absorbing Si atoms (Figure 5.4.6). The sorted and back transformed

©2003 Taylor & Francis

12 Si


-SiC

2

|FT| (au)

4C

1

0

A

0

2

4

6

R (Å)

Figure 5.4.6 Modulus of the FT of k 2 χ (k) for β-SiC(2–10 Å−1 ). The first peak corresponds to the first coordination shell around Si atoms (4 C atoms at 1.89 Å), the second to the second coordination shell (12 Si atoms at 3.07 Å). Peak A is a multiple scattering structure known as a fingerprint of the cfc structure. It is a focusing effect due to the alignment of 3 Si atoms. Phases shift are not corrected and the peaks do not appear at the real distances.

in the k space peak, corresponding to a given coordination shell, is at first fitted using the electronic parameters (amplitude and phase functions) extracted from model compounds in which bonding configurations similar to those being investigated are present: β-SiC for Si C pairs (4 at 1.89 Å) and Si Si second neighbour pairs (12 at 3.08 Å), α-Si3 N4 for Si N pairs (4 atoms at an average distance of 1.74 Å), and α-SiO2 for Si O pairs (4 at 1.61 Å) (Wyckoff 1963). The experimental signal extracted for the first shell is fitted using the single scattering formula (5.4.1) for EXAFS. For the first shell, given the available energy range of the EXAFS spectra !k equal to about 8 Å−1 , and the !R equal to 1.4 Å used in the back-FT, the number of free parameters (2 · !R · !k/π) turns out to be seven. The free parameters during fitting process were the coordination number and the interatomic bond length. A three-shell fit deals with six free parameters. The description of the other contributions (beyond the closest neighbours) must include multiple scattering events and formula (5.4.2) must be used.

5.4.2 Study by EXAFS at the Si K edge of Si, SiN, SiC and Si/C/N laser-synthetised nanopowders produced from gaseous precursors Six as-formed nanopowders (Si, SiC151, SiC152a, SiN7, SiCN12 and SiCN27) were measured. The Si nanopowder was formed from a SiH4 Ar gas mixture. XRD measurements revealed the existence of c-Si in the as-formed nanopowder. Two SiC (SiC151 and SiC152a) powders were synthetised from CH3 SiH3 gaseous precursor. They were differentiated by the temperature reaction which induces changes in X-ray diffraction (XRD) patterns, as

©2003 Taylor & Francis

observed previously in the case of the SiH4 C2 H2+n mixtures (Luce 1990). The XRD measurements performed on SiC151 and SiC152a revealed the existence of β-SiC crystallites in the nanopowders. The SiN7 powder was formed from a SiH4 NH3 mixture and the two SiCN powders (SiCN12 and SiCN27) from SiH4 CH3 NH2 NH3 mixture in variable proportions. Thus, NH3 was only used for the synthesis of SiCN27 and allowed the increase of the N content in SiCN27 (C/N at. = 0.22) compared to SiCN12 (C/N at. = 0.93). XRD measurements revealed the existence of α and β-Si3 N4 crystallites in SiN7, SiCN12 and SiCN27 powders and of β-SiC crystallites in SiCN12. Note that a small proportion of c-Si was also detected in the SiN7 powder. 5.4.2.1

Si, SiN7, SiC151 and SiC152a nanopowders

Figure 5.4.7 presents the FT of the Si nanopowder. The first peak (A) located near 2 Å corresponds to the first coordination shell surrounding the Si atoms. It is related to the four Si atoms which build the tetrahedral environment of the absorbing Si atom (RSi Si = 2.35 Å). Beyond this first intense peak (R > 3 Å), one finds a set of small peaks, related to the intermediate-range organisation in the material, which are characteristic of that of crystalline Si (Lagarde and Flank 1986). These results are in good agreement with those found by XRD which revealed the existence of c-Si. The FT curves obtained for SiN7 and α-Si3 N4 are displayed in Figure 5.4.8 and one can take qualitative information of their comparisons. First, one can observe a strong decrease in the amplitude of the peaks located at R > 2 Å for SiN7 compared to α-Si3 N4 . This reduction is an effect of the static disorder. A reduction in the amplitude of the first peak (A) is also observed in SiN7. An attempt was made to simulate the contribution of this peak on the basis of a one-shell model. The only parameters which were varied are the interatomic bond lengths and the coordination number. The DW factor was not allowed to vary during the fitting procedure, which means that we considered this factor equal to that found in the crystalline compound α-Si3 N4 for the Si N bond. A slight variation ( 2 Å is observed. The first shell contribution to the

©2003 Taylor & Francis

12 SiC151

|FT| (au)

9

SiC152a

6

D

3

0

0

2

R (Å)

4

6

Figure 5.4.9 FTs of k 3 χ (k) for SiC151 and SiC152a.

FT of SiC152a was simulated using the electronic parameters extracted from SiC151 (4 C at 1.89 Å). The result given in Table 5.4.1 is the decrease of the coordination number. One finds 3.4 C atoms around one Si atom. This result gives evidence of the presence of crystallite of β-SiC in SiC152a. The reduction of the coordination number may have several explanations. First, the H content is higher in the SiC152a sample, than in the SiC151 sample, and induces a decrease of the coordination numbers and so, the H effusion during the pyrolysis process is less important in SiC152a. We expect that, the existence of Si H bonds in the as-formed powder is responsible for this decrease, since the gaseous precursor contains SiH3 groups. Second, the structural disorder is more important in SiC152a than in SiC151 due to bond length and/or angular fluctuations. Smaller crystallites (1–2 nm) of silicon carbide in SiC152a would also lead to a decrease of the coordination number around Si atoms. For SiC151, the proportion and/or the size of the β-SiC clusters are more important and the features of the EXAFS spectra are those of the crystalline β-SiC phase. 5.4.2.2

SiCN12 and SiCN27

FTs of SiCN12 and SiCN27 are given in Figure 5.4.10. The two compounds exhibit the features of ordered structures. Beyond the first peaks of the FT, one finds a set of peaks related to the intermediate-medium-range organisation in the material. The presence of the focusing effect peak in SiCN12 indicates that, contrary to SiCN27 for which the peak is absent, this powder is crystallised partially in a β-SiC phase. This result is in agreement with the XRD measurements which detect α, β-Si3 N4 phases for 1 and SiCN27 and the β-SiC phase only for SiCN12. The analysis of the other peaks is difficult because of the multiplicity of the possible configurations for Si atoms, in such SiCN compounds, and because a fine analysis of this part of the pseudo-radial distribution function requires the introduction of multiple scattering contributions. Concerning the first peak related to the first coordination shell, a two-shell fit (Si N + Si C) was realised using electronic parameters extracted from model compounds (SiC151 and α-Si3 N4 ). The results are given in Table 5.4.1. One

©2003 Taylor & Francis

9 SiCN12

|F T| (au)

SiCN27 6

3 D 0

0

2

R (Å)

4

6

Figure 5.4.10 FTs of k 3 χ (k) for SiCN12 and SiCN27.

Table 5.4.2 Influence of the flame temperature T during the pyrolysis process on the proportion of Si H remaining bonds in the as-formed nanopowders assuming that the reduction of the coordination number is completely due to Si H bonds (!N /N = ±10%) T (◦ C) NSi−H

SiN7

SiC152a

SiC151

SiCN12

SiCN27

1250 0.9

1150 0.6

1530 0

1550 0.3

1615 0.1

finds that the NSi C /NSi N ratio follows the C/N ratio. For SiCN12 (C/N = 0.93), the first coordination shell of one Si atom is composed of 2 C atoms and 1.7 N atoms, whereas for SiCN27 (C/N = 0.22), it is composed of 2.9 N atoms and 1 C atom. The coordination numbers are close to 4 for the two compounds which indicate that, as for SiC151, there is not much H bonded to Si. We expect that the proportion of Si H bonds in the material is correlated to a pyrolysis reaction parameter which is the flame temperature. Indeed, for SiC151, SiCN12 and SiCN27, this temperature is about 1500–1600◦ C (Table 5.4.2) whereas for SiC152a and SiN7, in which a higher reduction of the coordination number (20–30%) in the first coordination shell is found, it is about 1100–1200◦ C. The results of the fits show that both C and N atoms are present in the first coordination shell of Si atoms in SiCN12 and SiCN27. For SiCN12, both SiC4 and SiN4 tetrahedra are present in the compounds, since XRD measurements show the existence of β-SiC and α, β-Si3 N4 crystallites in the powder. For SiCN27, SiN4 tetrahedra are present in α, β-Si3 N4 crystallites, but we cannot conclude about the existence of SiC4 or SiCx N4−x tetrahedra, since a two-shell fit of the EXAFS data provides an average picture of the local arrangement of C and N atoms in the first coordination shell of the absorbing Si atoms.

©2003 Taylor & Francis

5.4.3 Short-range order determination in amorphous SiCN-HMDS nanopowders: evolution as a function of the C/N ratio and annealing treatment This part reports a detailed structural analysis of nanosized SiCN powders produced from the liquid precursor hexamethyldisilazane (HMDS). We focused our study on the intermediate compositions, for which a delay in the crystallisation is observed. XAS was used to describe the short- and medium-range order in the amorphous powders. The amorphous → crystal transition which occurs beyond 1500◦ C was also studied by X-ray and neutron diffraction (see Section 5.4.4 ND and XRD). The goal of this study is to give a model for the structural evolution of these powders when the annealing temperature increases. The informations obtained from absorption and diffraction measurements are related to the structure of the powders at the atomic scale and are then particularly adapted to the study of amorphous compounds. The XAS experiments at the Si K edge give a radial distribution function (RDF) around the Si atoms in the material, and diffraction measurements give a total radial distribution function. The analysis of these RDF allows the determination of structural parameters, such as bond lengths or coordination numbers, and of information related to the intermediatemedium-range order, which can be modelised and tested through XAS and diffraction spectra calculations. This detailed study of the amorphous phases was performed to obtain structural information, which could explain the delay in the crystallisation and give information about the mechanism that leads to the crystallisation of the nanopowders. 5.4.3.1

Mixed tetrahedra and C N bonds

Short-range order deduced from the analysis of the first coordination shell of Si atoms Figure 5.4.11 shows the extracted EXAFS signal as a function of the powder composition for the as-formed samples. The FT curves obtained for the various C/N ratio samples are

k × (k) (Å–1)

-Si3N4


-SiC HMDS45 as-formed HMDS44 as-formed HMDS43 as-formed HMDS40 as-formed 3

6 k (Å–1)

9

Figure 5.4.11 Si K EXAFS raw spectra for β-SiC, α-Si3 N4 and Si/C/N with variable C/N. The spectra have been vertically displayed for convenience. The corresponding concentration of the samples are displayed (Table 2.5).

©2003 Taylor & Francis

(a)
-SiC

|FT| (au)

2

-Si3N4

1

0 HMDS40 as-formed

|FT| (au)

2

HMDS43 as-formed HMDS44 as-formed

Im(FT) (au)

HMDS45 as-formed

1

0

(b)

HMDS42 as-formed

0

2

R (Å)

4

6

2

HMDS40 as-formed HMDS42 as-formed HMDS43 as-formed HMDS44 as-formed HMDS45 as-formed

0

0

2

4

6

R (Å)

Figure 5.4.12 (a) FT modulus of the k 2 weighted EXAFS oscillations for the two references β-SiC, α-Si3 N4 , and for the Si/C/N samples; (b) imaginary part of the FT for the Si/C/N samples as a function of C/N.

reported in Figures 5.4.12(a) and (b), and one can take out some qualitative information of their observation: absence of long-range order (for R values higher than 4 Å), short and intermediate orders very slightly dependent on the C/N ratio, as evidenced clearly by both the magnitude and the imaginary part of the FT, which are very similar up to 4 Å. In order to determine the local order around Si atoms, a two-shell fit has been calculated on the basis of Si C and Si N pairs. The structural parameters obtained as a function of the C/N ratio are summarised in Table 5.4.3.

©2003 Taylor & Francis

Table 5.4.3 First-shell structural parameters obtained with a two-shell fit (Si C + Si N) for the different samples. !N /N = 10%, !R = 0.02 Å Sample

Si C NSi

HMDS40 HMDS43 HMDS44 HMDS45

1.4 0.9 0.9 0.7

C

Si N RSi 1.88 1.9 1.87 1.88

C (Å)

NSi 2.1 3.2 3.2 3.3

N

NSi RSi 1.73 1.73 1.72 1.73

C+Si

N

N (Å)

3.5 4.1 4.1 4.0

We first used a fitting procedure where the two interatomic distances have been kept fixed to their crystalline values, within 0.02 Å. As we used experimental electronic parameters, the values of the adjustable disorder factors are just the difference between the unknown and 2 2 2 the model compound !σ 2 = σalloy − σmodel , where σmodel is included in the experimental amplitude. This difference, in all cases, is found to be negligible by the fitting procedure. One has to evaluate the DW factor for Si C and Si N bonds by fitting the first coordination shell in model compounds with the calculated backscattering amplitudes and the phase shifts taken from McKale et al. (1988). We found a DW factor value less than 5·10−2 Å, consistent with the value found in ZnSe (Diop and Grigenti 1995), and in c-Ge (Dalba et al. 1993). Then, during the fitting procedure, the damping factors !σ 2 were kept equal to 0, which means that we considered a DW factor for the unknown structure equal to those of the reference phases. As for the values of the energy edge shift !E0 = E0 alloy − E0 model , we checked that small variations of this parameter (±2 eV) leave the fit unaltered, therefore, this parameter was not varied during the fits. Then in a second step, the only parameters which were varied during the fit procedure were the interatomic distances and the partial coordination numbers. The different interatomic distances appear to be constant over the C/N range. The values obtained for the different pairs may be well compared with the model values. Such constancy has already been observed in hydrogenated amorphous silicon carbide alloys (Pascarelli et al. 1992). The total coordination number of Si is expected to be close to 4, as there is limited bonding of Si to H. Actually, hydrogen with its single electron shows no measurable scattering, and its structural effect can only be evidenced when it is located between the absorber and the backscatterer in well-defined sites. In disordered structure as the one described here, the main effect of the presence of H in the samples would result in a reduction of the effective coordination. The results concerning the total coordination number are 4 ± 10%, an acceptable uncertainty for EXAFS measurements. The proportion of the different pairs (partial coordination numbers) follows approximately the chemical composition, as it can be seen in Figure 5.4.13, which shows the evolution of the ratio of the partial coordination numbers NC /NN as a function of C/N. One can observe that this evolution is linear with a slope equal to 0.5. As the chemical composition of the powders relies on the presence of oxygen in the powders, fits including Si O pairs were also performed. Table 5.4.4 reports the results obtained with a three-shell fit and the result of the adjustment for HMDS43 is given in Figure 5.4.14. The introduction of Si O bonds improves definitely the quality factor of the fit (from 1.3 × 10−2 to 5.5 × 10−3 ). The proportion of Si O pairs, that never exceeds 10%, increases with the O atomic content of the alloy.

©2003 Taylor & Francis

I II

HMDS40 HMDS43 HMDS42

0.6 NC/NN

HMDS44 HMDS45 Si3N4 0

0

0.4

0.8 C/N

1.2

Figure 5.4.13 Variation of the ratio of the partial coordination numbers NC /NN with C/N: I is for parameters given in Table 5.4.3 and II for those in Table 5.4.4.

Table 5.4.4 First-shell structural parameters obtained with a three-shell fit (Si C + Si N) for the different samples. !N/N = 10%, !R = 0.02 Å Sample

Si C NSi

HMDS40 HMDS43 HMDS44 HMDS45

1.6 0.9 0.9 0.5

C

Si N RSi

C (Å)

1.89 1.88 1.87 1.88

NSi 1.9 2.6 2.7 3.0

N

Si O RSi

N (Å)

1.73 1.75 1.73 1.75

NSi

NSi

O

0.2 0.7 0.8 0.8

RSi 1.61 1.62 1.58 1.63

C+Si

N+Si

O

O (Å)

3.7 4.2 4.4 4.3

Experiment

k × (k) (Å–1)

0.4

Fit (C + N + O)

0

–0.4 2

4

6 k (Å–1)

8

10

Figure 5.4.14 Best fit of the filtered first peak (0.6–2 Å) of the FT for HMDS43 (as-formed).

©2003 Taylor & Francis

The estimation of the uncertainties in the fitted parameters is a non-trivial problem, because of the correlation between the different parameters, Rij and !E0 on one hand, and Nij and σij on the other. One should also consider the role played by the anharmonicity of pair potentials, which means the deviation from a gaussian behaviour. In another strongly covalent system, amorphous Ge, the study of the local disorder using the cumulant expansion leads to corrections, to interatomic distances and to coordination numbers (Dalba et al. 1993). These corrections are fully within the error bars we have quoted here. So we did not consider any anharmonic behaviour, which should be in any case very difficult to evidence in a two-shell disordered model. Furthermore, a non-negligible uncertainty of the first-shell fit procedure comes from the fact that C and N, acting as backscatterers, are neighbours in the periodic classification table, which means that they cannot easily be distinguished. Last, a two-shell model, with partial coordination numbers NC and NN , provides an average picture of the local arrangement of the C and N atoms in the first coordination shell of the absorbing Si atoms. It does not allow the discrimination of a locally demixed phase with SiC4 and SiN4 tetrahedra arrangement (complete chemical ordering), or a random stacking of mixed SiCx N4−x tetrahedra. This point will be clarified by looking at the near edge structure in the next section. The Si K XANES studies In X-ray absorption, an electron absorbs an incident photon and is promoted to an empty state in the conduction band. Within the dipolar approximation, only transitions corresponding to !l = ±1 between the orbital quantum numbers of the initial and final states are allowed. The experimental Si K edge spectrum therefore reflects the p local DOS on a Si atom. For instance, a sharp and intense resonance is observed at the threshold of spectra of molecular species. This resonance has been associated for molecular spectra to a Si 1s electron promoted to σ ∗ orbitals with appreciable Si 3p character (Hitchcock et al. 1993) and its energy position and intensity vary with the nature of the first Si neighbours. An alternative way to interpret the experimental XANES spectra is to use the full multiple scattering formalism, where the effect of the core hole is also taken into account. This has been applied with success to reproduce the polarisation dependence of XANES of α-quartz, for which very large clusters are required (Briois et al. 1993), and of β-SiC (Sainctavit 1993) using the CONTINUUM code developed by Natoli et al. (1980). It has been shown that the structures B and C (Figure 5.4.5) in the SiC XANES spectrum can be related to the size of the microcrystals through EXAFS information and full multiple scattering calculations. As a passing note one can observe the very peculiar shape of the β-SiC XANES. It has to be compared with the one of Si (CH3 )4 (Bodeur et al. 1990), where Si is also surrounded by 4 C, the XANES of which exhibit a prominent edge resonance. The effect of the long-range order in β-SiC appears as a corner cut-off of the edge. In all Si reference compounds, the structures, present between the resonance at the edge and the broad resonance located about 20 eV above the threshold, are strongly dependent on the structural medium-range order. The different calculations described above show definitely this point. One can notice as an illustration that α- and β-Si3 N4 can be differentiated by taking a glance at the XANES: the β form, where the unit cell is half as large as that of the α phase, presents the most structured XANES spectrum. The second point to notice is the chemical shift existing when going from pure Si to silicon oxide. The threshold position is strongly dependent upon the nature of the atoms which built the tetrahedral environment of Si, and is directly related to the electronegativity of these atoms, that is, to the more or less large density of p conduction states on Si sites.

©2003 Taylor & Francis

This allows us to give qualitative conclusions concerning the XANES spectra of the Si/C/N compounds (Figure 5.4.15(a)). First of all the shape (which remains similar whatever the C/N ratio) of the near edge domain is constituted of a well-defined threshold, with a broad resonance (A) followed only by a broad structure (C) at about 15 eV above the edge. The fact that all the XANES structure between A and C is washed out indicates a medium-range disorder between the tetrahedra, and clearly confirms the amorphous state of the as-formed sample. Second, looking at the energy position of the absorption threshold, we see that it varies continuously but stays at an intermediate position between that of SiC and Si3 N4 v. the C/N ratio. We think that this is a direct evidence for a mixed environment of a SiCx N4−x type. Actually, a demixed environment would lead to a two-steps edge, as it can be observed for samples heated above the crystallisation temperature. Figure 5.4.15(b) is a clear illustration of the demixing process accompanying the beginning of the crystallisation: the XANES

(a)

HMDS40 as-formed HMDS42 as-formed HMDS43 as-formed HMDS44 as-formed HMDS45 as-formed
-SIC -SI3N4

Absorption (au)

A

C

1

0

1840

1850

1860

1870

E (eV) (b) HMDS44 1550°C/4 h

Absorption (au)


-SIC -Si3N4

1

0

1840

1850

1860

1870

E (eV)

Figure 5.4.15 (a) Evolution of the Si K XANES spectra of Si/C/N powders for different C/N ratio (b) Si K XANES spectrum for the two references β-SiC and α-Si3 N4 and for the HMDS44 sample after heating above the crystallisation temperature. The last one appears clearly as a combination of the two references ones.

©2003 Taylor & Francis

1844

Si3N4 HMDS45

E (eV)

HMDS44

HMDS43

1843

HMDS42

1842

0

0.4

0.8 C/N

HMDS40

1.2

Figure 5.4.16 Variation of the edge position with the C/N ratio.

becomes very structured near the threshold, and looks more like a linear combination of both SiC and Si3 N4 spectra: the two kinds of tetrahedral environment are now present in different proportions, depending on the annealing conditions. Therefore, concerning the as-formed samples, a local chemical ordering can be excluded by just looking at the XANES spectra. Figure 5.4.16 shows the evolution of the edge position when the C/N ratio increases, and C atoms are replaced by N atoms in the Si environment. Such a mixed configuration for Si atoms was also evidenced previously for laser synthesised Si/C/N nanoparticles, using gaseous Si precursor (Gheorghiu et al. 1992) or an aerosol precursor with a low C/N value (Driss-Khodja et al. 1996). Such a mixed environment for Si atoms has already been suggested from NMR studies by Suzuki et al. (1995), who concluded to the presence of SiCN3 tetrahedra in Si/C/N systems at high N content.

EXAFS – Analysis of the intermediate order: modelisation of the experimental data Beyond the first peak of the pseudo-radial distribution function (FT), which corresponds to the first coordination shell of the Si atoms, one can observe (see Figure 5.4.12) two small peaks in the range 2–3.5 Å (uncorrected from phase-shift). These peaks are very little dependent on the C/N ratio, and this means that the intermediate order is in average independent of the chemical composition. Focusing on one sample (HMDS43 with a C/N equal to 0.7), an attempt has first been made to fit these peaks with Si Si contributions, with a wide distance distribution, as one can expect in an amorphous covalent network where the close environment is well defined, but where there can be an enhancement of the bond angle distortions. Figure 5.4.17(a) shows the very bad quality obtained for the fit with only Si Si second environment. Fitting the experimental data by Si N pairs at a distance of 2.78 Å only, or by Si C pairs at 2.84 Å, strongly increases the fit quality (Figure 5.4.17(b)). The improvement obtained by this solution evidences the fact that most of the silicon tetrahedra are not in majority linked by their corners. Such a two-shell simulation assumes also the presence of C N bonds in the amorphous network. From these preliminary results, a model is proposed (Figure 5.4.18), where mixed tetrahedra are randomly linked through C N bonds.

©2003 Taylor & Francis

(a) Experiment k × (k) (Å–1)

Fit (Si)

(b) Experiment

k × (k) (Å–1)

Fit (C)

3

6

9 k (Å–1)

Figure 5.4.17 (a) Best fits of the filtered second peak (2–2.8 Å) of the FT (Figure 5.4.13) for HMDS43 sample with one Si Si shell; (b) One-shell fit of the filtered second peak corresponding to Si C at 2.84 Å as second neighbours for Si atoms. A fit corresponding to Si N at 2.78 Å gives also a good agreement with the experiment.

as-formed

N C

Si1 Si3

Si2

Calculation

Figure 5.4.18 Proposed model for the amorphous atomic structure of Si/C/N with C/N = 0.7.

©2003 Taylor & Francis

This model was built with a molecular mechanic software (CERIUS). An analysis of various two, three and four bodies interactions allows an evaluation of the forces on each particle. The energy expression deduced for N-body systems, such as our model, is calculated by superposition of single terms (bond, angle, torsion, inversion). The exact nature of each term depends on the particular loaded force field. Here, we used the DreidingII force field (Mayo et al. 1990), which is recommended for structures with novel combination of elements and those for which there is little or no experimental data. A computational module allows then the minimisation of the energy expression of the model and leads to an optimum geometry. This average model is composed of SiN3 C tetrahedra linked through C N bonds, in order to form six-edges polyhedra. The only constraints imposed before the relaxation process are the first neighbours distances for Si C and Si N bonds. They were fixed to those found by the experimental analysis (1.89 Å for Si C and 1.74 Å for Si N). The coordinates of the atoms in the model are then extracted and used as input to the FEFF code (Rehr et al. 1992), in order to compare the corresponding calculated EXAFS signal, which includes multiple scattering paths, to the experimental one. In that calculation, the default values were used for the input parameters. The calculation is performed with Si1 , Si2 and Si3 playing the role of the central atom (see Figure 5.4.18). The number of paths is limited to four, and the maximum distance (Rmax ) to 6 Å. The first shell gives the main contribution. Beyond it, one finds a large distribution of Si N and Si C two-legs paths between 2.6 and 2.85 Å, followed by three-legs paths corresponding to Si/C/N and Si/N/C, the distances of which are found between 2.9 and 3.15 Å. The first Si Si distances are found around 3.3 3.4 Å. It happens that the best fit is obtained when averaging the results on Si2 and Si3 as central atom. Figure 5.4.19 shows the good agreement obtained by this way. The calculated result obtained with Si1 as central atom leads to an enhancement of the second neighbours peaks. This amplitude difference, due to the fact that Si2 and Si3 do not have all their second neighbours, may have two explanations. First, this could evidence a static and/or dynamic disorder on the outer shells distance, due to distortions of the valence bond angles of C and N first neighbours. As it is well known in EXAFS and has been discussed previously,

2

|FT| and Im(FT) (au)

HMDS43 as-formed Model

1

0

–1

0

2

4

6

R (Å)

Figure 5.4.19 Comparison of the FT (modulus and imaginary part) for the experimental spectrum of the HMDS43 Si/C/N sample, with the FEFF calculation using the atomic coordinates of the model proposed in Figure 5.4.18.

©2003 Taylor & Francis

such a disorder appears as a reduction of the coordination number, then the environment of Si2 and Si3 could represent a more realistic situation than Si1 . This amplitude reduction can also be interpreted as an anharmonic effect originating in the size of the nanometric particles. But nanometric Si/C/N powders concern particles of an average diameter of about 30 nm, with only about 10% of the atoms in surface. The only disorder taken into account intrinsically in our simulation is that coming from the calculated positions of the atoms, with no further parameter: this accounts for the structural disorder and its deviation to a gaussian distance distribution. Second, this lower coordination number might be real and could come from dangling bonds of N and C, or to a bonding to H atoms. In terms of an EXAFS analysis, these two structural situations are indistinguishable, and can be both present in the material. Finally, such a limited model is not able to reproduce the XANES part of the spectrum, probably due to the fact that the muffin-tin approximation in the calculation of the potential is not appropriate.

5.4.3.2

Evolution of the local order during the pyrolysis process

1400◦ C Our purpose is then to understand the structural evolution before the crystallisation, when the sample is treated under N2 at 1400◦ C for 24 h, 1500◦ C for 4 h and up to 1600◦ C. The observed changes are rendered on the FT as modifications of the second and third neighbours in the intermediate-range order (Figure 5.4.20). The comparison of the FT of the as-formed powder and the annealed ones underlines the slight evolution of the short-range order after annealing at 1400◦ C (Figure 5.4.20), while more important change occurs in the intermediate-range order. In order to determine the local order around Si, a two-shell fit of the first peak has been performed on the basis of Si C and Si N pairs. We obtained the structural parameters presented in Table 5.4.5. As for the as-formed samples, a two-shell model with partial NSi C

2

|FT| and Im (FT) (au)

HMDS44 as-formed HMDS44 1400°C/24 h 1

0

–1

0

2

4

6

R (Å)

Figure 5.4.20 FT (modulus and imaginary part) of the k 2 weighted EXAFS oscillations for HMDS44, as-formed, and annealed 24 h at 1400◦ C.

©2003 Taylor & Francis

Table 5.4.5 First-shell structural parameters obtained with a two-shell fit (Si C + Si N) for two compositions and for two annealing temperatures (1400◦ C for 24 h and 1500◦ C for 4 h). !N /N = 10%, !R = 0.02 Å Sample

Si C NSi

HMDS42-1400 HMDS42-1500 HMDS44-1400 HMDS44-1500

1.4 2.3 1.2 1.7

C

Si N RSi

C (Å)

1.87 1.88 1.87 1.87

NSi 2.8 1.7 2.8 2.3

N

NSi RSi 1.74 1.75 1.73 1.73

C+Si

N

N (Å)

4.2 4 4 4

1400°C

Si

N C

Calculation

Figure 5.4.21 Proposed model for the amorphous atomic structure of the sample annealed at 1400◦ C (24 h).

and NSi N provides an average picture of the local arrangement of the C and N atoms in the first coordination shell of the absorbing Si atoms. Beyond the first peak of FT (Figure 5.4.20), the structures relative to intermediate-range order evolve with the annealing temperature. Unlike the results obtained for the as-formed powders, and as early as 1400◦ C, one has to introduce Si Si bonds as second neighbours. A new 3D structure based on EXAFS results (short range and intermediate order) was built and then tested through multiple scattering calculations at the Si K edge of central Si atoms of the structure. The model presented in Figure 5.4.21 is related to the 1400◦ C annealed temperature. It is composed of rigid SiN3 C tetrahedra (Si N = 1.74 ± 0.02 Å and Si C = 1.89 ± 0.02 Å linked by C N bonds with large distance distributions (1.46–1.55 Å) or partially linked by their corner. The evolution compared to the as-formed model is the possibility to link the SiN3 C tetrahedra by their corner. This model was in a second step tested on 3 Si central atoms through multiple scattering calculation using the FEFF code. The average calculated spectrum and its FT presented in Figures 5.4.22(a) and (b) are compared to the corresponding experimental parts. The agreement between experiment and calculation is good. Particulary, the evolution beyond the first peak is well reproduced. The evolution in the way in which tetrahedra are linked is confirmed by the calculations: Si atoms as second neighbour of Si

©2003 Taylor & Francis

(a) HMDA44 1400°C/24 h

1

k × (k) (au)

Model

0

–1 2

(b)

4

6 k (Å–1)

8

10

2

|FT| and Im (FT) (au)

HMDA44 1400°C/24 h Model

1

0

–1

0

2

4

6

R (Å)

Figure 5.4.22 Comparison of (a) k × χ (k) and (b) FT (modulus and imaginary part) for the experimental spectrum of the sample (HMDS44), annealed at 1400◦ C for 24 h with the FEFF calculation, using the atomic coordinates of the model proposed in Figure 5.4.21.

atoms appear at 1400◦ C. This result is consistent with the increase of the density which is also observed in the powders. Concerning the static disorder, the Si atoms taken as absorbing atoms in the FEFF calculations still have an incomplete second coordination shell, which simulates the static disorder. The increase of the coordination number compared to the asformed model simulates the decrease of the static disorder when the annealing temperature increases. 1500◦ C The same approach on the XAS spectra, corresponding to the 1500◦ C annealed sample, shows this time an evolution of the NC /NN ratio. Figure 5.4.23 presents the FT of HMDS44 annealed at 1400◦ C (24 h) and 1500◦ C (4 h). Strong changes occur in the first coordination

©2003 Taylor & Francis

2

|FT| and Im (FT) (au)

HMDS44 1400°C/24 h HMDS44 1500°C/4 h 1

HMDS44 1550°C/4 h A

0

–1

0

2

4

6

R (Å)

Figure 5.4.23 FT of the k 2 weighted EXAFS for HMDS44 annealed at 1400◦ C (24 h), 1500◦ C (4 h) and 1550◦ C (4 h). HMDS45 1500°C/4 h HMDS43 1500°C/4 h HMDS44 1500°C/4 h HMDS45 1500°C/4 h
-SiC -Si3N4

Absorption (au)

3

2

1

0

1840

1850

1860

1870

E (eV)

Figure 5.4.24 Evolution of the Si K XANES spectra at 1500◦ C with different C/N ratio.

shell around the Si atoms. The amplitude of the first peak in the FT decreases at 1500◦ C (4 h). Table 5.4.5 gives the results obtained for HMDS42 and HMDS44 with a two-shell fit (Si C and Si N) of the first peak. The evolution of the ratio NC /NN as a function of the C/N ratio is still linear. The first coordination shell of the Si atoms is enriched in C. This result is confirmed by the XANES study of the samples annealed at 1500◦ C (4 h). For HMDS42, HMDS43 and HMDS44, the spectra still exhibit a unique threshold, the energetic position of which is intermediate between those of SiC and Si3 N4 (Figure 5.4.24). A small shift towards lower energies of the threshold position is observed (i

where Sij (q) are the partial structure factors for the atomic pair i–j and Wij (q) are the weighting factor which depend on the composition of the studied samples and on the

©2003 Taylor & Francis

Neutron

XRD

0.4

Si9Si

Wij (q)

N9N

Si9N

Si9N

0.2

N9C Si9C Si9Si C9C Si9O

0 6.74 q (Å–1)

13.48 0

Si9C N9N N9C Si9O C9C 6.9 q (Å–1)

13.8

Figure 5.5.1 Weighting factors Wij (q) for the six pairs of atoms in the HMDS44 (1400◦ C for 24 h) samples, defining the profile of the X-ray and neutron high angle spectra. The values are also given for the Si O pair. Note that Wij (q) varies with q in X-ray experiments but remains constant for neutron experiments. The important fact is that the subnetwork of Si is reinforced by X-ray while the N one is preferentially seen by neutron.

energy of the incident radiation. A complete determination of the six partial structure factors (Si Si, Si N, Si C, N N, N C, C C) requires six experiments. Unfortunately, anomalous wide angle X-ray scattering measurements (AWAXS) are experimentally unreachable at Si, N and C photoabsorption edges and then the complete determination of partial structure factors is impossible. The difference in the nature of the radiation used in X-ray and neutron experiments is responsible for a strong contrast effect on N and Si amorphous subnetwork. Figure 5.5.1 compares the Faber–Ziman weighting factors (Waseda 1984) (Wij (q)) calculated for X-ray and neutron radiations for HMDS44 (annealed at 1400◦ C for 24 h). In the range of composition of the HMDS powders, the contrast is very strong between the X-ray spectra which reflects for 80% Si Si and Si N correlations, whereas the neutron spectra reflects for 80% N N, N Si and N C correlations.

5.5.2.2

Experimental and data analysis

Neutron The neutron experiments have been performed on the 7C2 spectrometer at Laboratoire Léon Brillouin (LLB) at Saclay (CEA-France). The neutrons wavelength (λ = 0.7 Å) was selected with a copper monochromator (Cu(111)). The diffracted intensity was measured with a 640 BF3 -cells detector which allows to perform measurements with a constant !θ path. The larger measurable scattering angle 2θ is 128◦ . The corresponding scattering vector q is equal to 16 Å−1 (q = 4π sin θ/λ). The experimental data are corrected for absorption effect. The contribution of the container I c(q) was subtracted from the measured diffracted intensity I (q). The Placzek correction was applied (Placzek 1952) to correct I (q) from inelastic scattering process. The spectra were then normalised by the coherent scattering cross-section

©2003 Taylor & Francis

of the sample to obtain the structural factor S(q) (small contribution of the incoherent crosssection): S(q) =

I (q) αN σcoh

(5.5.2)

where I (q) is the total elastic scattered intensity, S(q) is the total structure factor, N is the number of atoms, σ the coherent scattering cross-section (Neutron 1992) and α a constant. X-ray The wide angle X-ray diffraction experiments were performed at LURE-DCI (Orsay-France) on the wiggler DW31B beamline equipped with a 2-circles goniometer using a scintillator. A 2-crystal (Si (220)) monochromator with sagitally focalisation and a double setup of distant slits define the energy resolution. A 20.2 keV incident energy has been chosen in order to use a q window going from 1 to 16 Å−1 (the same as for neutron experiments). Due to low absorption coefficient of the studied materials, the spectra have been recorded in the reflection mode with a nearly grazing incidence (θ0 = 2.5◦ ). The Compton contribution to the total signal has been removed using the analytic Hajdu’s formula (Hajdu 1971) adapted for light elements. The reduction of the recorded data after correction for absorption and asymmetric geometry has been made by integration assuming:  qmax  qmax  2  I (q, E)q 2 dq = N f  + Icomp (q) q 2 dq (5.5.3) qmin

qmin

where Icomp (q) is the modelised Compton contribution and N the number of diffracting atoms, I (q, E) is the total experimental scattered intensity. The elastic scattered intensity for one atom i(q, E) can be written as: i(q, E) = |f 2 | + |f |2 [S(q, E) − 1] (5.5.4)   where |f |2 = i,j ci cj fi fj∗ and |f 2 | = i ci |fi |2 , S(q, E) is the total structure factor and fi are the complex atomic scattering factors given by: fi (q, E) = fi0 (q) + f ′ (q, E) + if ′′ (q, E)fi

(5.5.5)

are calculated for each element from tabulated values (Waasmaier and Kirfel 1995; Sazaki 1984). ci are the atomic fractions of each element in the powders. 5.5.2.3

Structural transition in the amorphous state

Evaluation of the hydrogen content The strong incoherent (q dependent) and inelastic scattering of neutrons by hydrogen gives a form factor to the H-rich powders neutron spectra. Figure 5.5.2 presents crude spectra of the composition HMDS44 for the as-prepared sample and for annealed samples at 1000◦ C (4 h), 1500◦ C (4 h) under N2 atmosphere. The reduction of the form factor (decrease with q) with temperature is the consequence of the hydrogen effusion during annealing treatment. At 1500◦ C, H has totally effused. The residual decrease of the intensity at high q (1500◦ C) comes from the minor contribution

©2003 Taylor & Francis

As-prepared 8000

4000 1000˚C-4 h

I (q) (au)

2000

1000 1500˚C-4 h 2000

1000 4

8 q (Å–1)

12

16

Figure 5.5.2 Evolution of the HMDS44 original spectra after the usual corrections with the annealing temperature. Note the incoherent scattering form factor of H which allows to measure its content in the as-formed powder (21.5 at%) and in the 1000◦ C sample (6 at%). At 1500◦ C, H has totally effused.

of inelastic scattering process concerning the other atoms (Si, N, C). For the sample which present an important form factor, a quantitative evaluation of the atomic proportion of hydrogen in the powder can be performed (Dixmier et al. 1994). This can be achieved by making the assumption that at high q values, the measured intensity corresponds to the coherent contribution from Si, C, N atoms without the H contribution which has drastically decreased. The theoretical asymptotic scattering cross-section σ is calculated for one average atom using the tabulated scattering cross-section of the elements and the composition. The I (q = ∞) theoretical value is equal to αN σ , where N is the number of atoms (hydrogen excluded) and α is a constant. The atomic form factor for one H atom is evaluated using the inelastic scattering theory and the H content is then estimated by a trial and error method which consists in subtracting from the spectrum a certain proportion of the background, in order to force the curve to oscillate around the constant asymptotic value I (q = ∞). The subtracted intensity corresponds to the contribution from H atoms: αNH σH , where NH is the number of H atoms and σH the incoherent scattering cross-section for one H atom (σH = 80 barn for q = 0).

©2003 Taylor & Francis

From this value and from the asymptotic value I (q = ∞), we calculated the H content. One will find 21.5 at% for the as-prepared powder and 6 at% for the 1000◦ C annealed powder. Structural evolution of Si/C/N nanopowders during the pyrolysis process: 1400◦ C → 1500◦ C SHORT-RANGE ORDER: EXISTENCE OF C N BONDS

The coupled X-ray and neutron diffraction experiments have been performed on the composition HMDS44 for two annealing temperatures (1400◦ C for 24 h and 1500◦ C for 4 h). X-ray absorption spectroscopy (XAS) measurements underlined strong modification in the short-range atomic structure around Si atoms between 1400◦ C and 1500◦ C. An increase of the NSi C /NSi N ratio is observed. The neutron (Sn (q)) and X-ray (SX (q)) structure factors for HMDS44 (1400◦ C for 24 h and 1500◦ C for 4 h) are reported in Figure 5.5.3. As expected, X-ray and neutron spectra, which respectively reflect silicon and nitrogen subnetwork, are very different. The structure factors exhibit a set of broad diffuse halo characteristics of covalent amorphous compounds. At low q values, on the neutron spectra, one can find a first sharp diffraction peak (FSDP) related to the nitrogen medium-range atomic arrangement in the powders. Dixmier (1992) showed that in a disordered network in which four coordinated atoms (Si or C) are connected to three coordinated atoms (N), the structure factor S(q) exhibits a main peak near 3 Å−1 generated by the first neighbour and two prepeaks at 2 and 1.5 Å−1 generated respectively by the mixture of four and three valencies. At 1400◦ C, we expect that the extreme sharpness of the peak located near 3 Å−1 is a signature of the stacking of disordered layers on coherent length of 2–3 nm. In contrast, the Si dominated X-ray spectrum exhibits a diffuse first peak, but small sharp diffraction structures for q > 6 Å−1 . These small peaks indicate the existence in the amorphous state of well-defined long-range correlations in the Si amorphous subnetwork. At 1500◦ C, the shape of the structure factors change. The

(a) Sn(q) Sx(q)

S (q)

1500°C-4 h (b)

1 1400°C-24 h 0

6

12 q (Å–1)

Figure 5.5.3 Neutron (Sn (q)) and X-ray (SX (q)) interference functions for HMDS44 annealed at (a) 1500◦ C for 4 h and (b) 1400◦ C for 24 h . The two spectra (Sn (q) and SX (q)) reveal the different evolutions of the Si and N subnetwork with the annealing temperature.

©2003 Taylor & Francis

prepeak of neutron spectrum is split into two components and the morphology of the successive haloes changes. For X-ray spectrum, the first halo becomes sharper. This evolution is the signature of the decrease of the chemical and topological disorder of the Si amorphous subnetwork when the annealed temperature goes from 1400◦ C to 1500◦ C. Strong structural modifications occur in the short-range and medium-range order (MRO) of the amorphous compound at 1500◦ C. These evolutions are consistent with those underlined by extended X-ray absorption fine structure (EXAFS) measurements concerning the evolution of the local order around Si atoms with annealing temperature. The EXAFS results show that the Si subnetwork is very disordered at 1400◦ C due to the mixture of the configurations for Si Si pairs (Si N(C) Si and Si N(C) C(N) Si). The reduction of the number of Si N(C) C(N) Si bonds at 1500◦ C is responsible for the decrease of the topological and chemical disorder. The contrast effect is clearly visible on the corresponding pair of correlation functions G(r) presented in Figure 5.5.4. The first peak is not much affected (hetero-atomic bonds) but the second one is definitively shifted (homo-atomic pairs such as N N or Si Si). An attempt to simulate the first peak of the RDF (= 4πρ0 r 2 G(r)) was done. The procedure starts with the modelisation of the distribution of j type atoms around an i type atom by a Gaussian distribution fij (r). This distribution is related to the partial i–j pair correlation function Gij (r) by: fij (r) = 4πρ0 r 2 cj Gij (r) 3

where ρ0 is the atomic density (at/Å ), cj the atomic fraction of the element j and Gij (r) the partial pair correlation function associated to the partial structure factor Sij (q). An FT of the r-space contribution is then performed to obtain the q-space contribution Sij (q) and the (a)

Gn(r) GX(r)

G (r)

1500°C-4 h (b)

1400°C-24 h

0 0

2

6

4

8

r (Å)

Figure 5.5.4 Neutron (Gn (r)) and X-ray (GX (r)) pair correlations functions for HMDS44 annealed at (a) 1500◦ C for 4 h and (b) 1400◦ C for 24 h. The common integration limit S(qmax) = 15 Å of the FT generate non-physical peaks before 1 Å at the same value of r in the both experiments but not with the same amplitude. We expect that due to the strong decrease of the X-ray scattering factors at high q, the neutron description of the short-range atomic structure is more accurate than the X-ray description (particularly for the shortest correlations). The contrast effect is well marked for the homo-atomic pairs N N and Si Si regrouped in the second peak near 3 Å.

©2003 Taylor & Francis

20 (Si9N + Si9C)

RDF (at/Å)

15

10

5

0 0

1

2

3

r (Å)

Figure 5.5.5 Result of the simulation (neutron) of the first peak of the RDF with only two contributions (Si C and Si N) for HMDS44 annealed at 1400◦ C (24 h). The corresponding structural parameters are given in Table 5.5.1.

Faber–Ziman weighting factor Wij (q) of the i–j pair is applied. Then, one goes back to the real space by a new FT of the weighted contribution of the i–j pair (Wij (q)Sij (q)). If several contributions have to be taken into account to simulate the first RDF’s peak, the components are added. The obtained contribution is directly comparable to the experimental RDF since the limits of integration used to calculate the FT are in both cases the same (1.5–15 Å−1 ). Figure 5.5.5 presents the result of the simulation (neutron) using only two components (Si C and Si N) for the 1400◦ C annealed sample. Note that the values found for the centroid of Gaussian distributions are those found in the crystalline compounds SiC and Si3 N4 for the Si C bond length (1.89 Å) and the Si N bond length (1.74 Å). The results of the EXAFS analysis revealed that the Si N and Si C bond lengths remain similar in the amorphous state compared to the crystalline state. These two contributions (Si C + Si N) cannot fit the experimental shape of the first RDF’s peak (Figure 5.5.5). A third shorter contribution must be added. Figure 5.5.6 shows the very good fit (neutron) obtained when a C N (1.5 Å) contribution is added. Following this procedure, two fits were realised for each spectrum taking as origin either Si or C and N atoms in the modelisation of fij (r). Table 5.5.1 gives the structural parameters obtained for the samples annealed at 1400◦ C (for 24 h) and 1500◦ C (for 4 h). One can observe that the proportion of C N bonds decrease with annealing treatment. For the sample annealed at 1400◦ C, one finds 3N + 1C around one Si atom, and 2.3Si + 0.7C around one N atom. At 1500◦ C, the proportion of C atoms around one Si atom increases (2.2N + 1.8C) while the proportion of C N bonds decreases. Let us recall that EXAFS results give 3N + 1C at 1400◦ C and 2N + 2C at 1500◦ C. The results obtained by coupling X-rays and neutron diffraction are in very good agreement with those obtained by EXAFS. C N bonds are taken into account in EXAFS local order models and are detected by X-ray photoelectron spectroscopy (XPS) measurements. C C bonds detected by XPS are not introduced in the simulation since the corresponding weighting factor WCC is lower than 5% (HMDS44). Their contribution to the experimental spectra cannot be unambiguously proved. C C bonds from free carbon phase were reported in the study by Dürr et al. (1999) for Si/C/N ceramics obtained by pyrolysis

©2003 Taylor & Francis

20

RDF (at/Å)

15

(C9N + Si9N + Si9C)

10

5

0 0

1

2

3

r (Å)

Figure 5.5.6 Result of the simulation (neutron) of the first peak of the RDF with three contributions (Si C, Si N and C N) for HMDS44 annealed at 1400◦ C (24 h). The corresponding structural parameters are given in Table 5.5.1.

Table 5.5.1 First shells structural parameters obtained with a three-shell fit of the first peak of the radial distributions functions (!R = ±0.02 Å, !N/N = ±10%, !σ = ±0.01 Å). For each compound, two set of simulations (Si N + Si C + N C or N Si + C Si + N C) were realised corresponding to two different choices for the central atom taken for the modelisation of the distribution functions R(Å)

1400◦ C (24 h) Si N (N Si) Si C (C Si) N C 1500◦ C (4 h) Si N (N Si) Si C (C Si) N C

N

σ (Å)

X-ray

Neutron

X-ray

Neutron

X-ray

Neutron

1.76

1.74

0.07

1.89

2.8 2.3 0.9 3 0.7

0.07

1.89

2.8 2.4 1 3.2

0.07

0.05

2.2 2.4 1.8 3.4 0.4

0.08

0.06

0.07

0.04

1.50 1.76

1.73

1.91

1.88 1.51

2.1 2.2 1.7 3.6

0.2

0.18

of a polymer species (C/N = 1.3). In their study, the weighting factor ωcc was of the order of 20% and the peak related to C C bonds was clearly observed in the neutron correlation pair functions. ANALYSIS OF THE MEDIUM-RANGE ORDER: THE TRANSITION 1400◦ C(α) → 1500◦ C(β)

The existence of a layered structure evidenced at 1400◦ C for N atoms must already occur in the initial mineral residue at 1000◦ C. This should be induced by the possible sp2 bonding of

©2003 Taylor & Francis

4 0r (G (r)-1)

1550°C-4 h 1500°C-4 h 1400°C-2 h 1000°C-4 h A 3

4

5

6

r (Å)

Figure 5.5.7 Reduced neutron pair correlation function (4πρ0 r(Gn (r) − 1)) in the MRO domain for HMDS44. The peak at 3.9 Å (A) is specific of the a stacking mode of α-Si3 N4 . This correlation disappears at 1500◦ C and is replaced by a new one at 3.6 Å. The stacking mode for the N atoms changes at 1500◦ C.

C atoms but also by the flat three coordinated N environment. Figure 5.5.7 presents the MRO of Gn (r) (neutron) for HMDS44 for annealing treatments ranging from 1000◦ C to 1550◦ C. At 1000◦ C, one can observe two small bumps at 3.4 and 3.9 Å which become sharper at 1400◦ C. The existence in the pair correlation function of a correlation at 3.9 Å is attributed to the alpha stacking mode for the atoms in silicon nitride (α-Si3 N4 ). Calculations using the Debye formula were performed on clusters of silicon nitride (Dixmier et al. 1994) which evidenced a correlation between the peak at 3.9 Å and the α stacking mode in silicon nitride. We performed a calculation of the atomic pair distribution in the two known crystalline forms of silicon nitride (α and β). The result is that the correlation at 3.9 Å in α-Si3 N4 can be associated with N N pairs at 3.82 Å between alternated layers. In the α form, the stacking mode (along the crystallographic c axis) is ABCDABCD. . . and N N pairs at 3.82 Å exist between N atoms of the layer A and N atoms of layer C, and N atoms of the layer B and N atoms of the layer D. In the β form, the mode is ABABAB. . . This modification in the stacking mode leads to the disappearance of the N N pairs at 3.82 Å and so to the annihilation of the correlation at 3.9 Å in Gn (r). Then, the correlation at 3.9 Å in Gn (r) is the signature of the α stacking mode for nitrogen atoms. At 1500◦ C, the correlation at 3.9 Å disappears while a new one appears at 3.6 Å on the experimental Gn (r). So, locally the increase of the NSi C /NSi N ratio at 1500◦ C is accompanied by a change in the sequential stacking mode of the layers. We expect that the α-mode is replaced by the β-mode at 1500◦ C. In the amorphous state, nanometric (2–3 nm) SiCN structures which are reminiscent of a crystalline topological order (α- and β-Si3 N4 ) exist. The amorphous structure cannot definitely be considered as a mixture of an amorphous silicon nitride phase and free C phase since mixed SiCx N4−x tetrahedra are detected by a X-ray near edge structure (XANES) study at the SiK edge and by NMR measurements. Carbon atoms induced structural and chemical disorder in α- and β-type SiNy phases. The term ‘chemical disorder’ should be used with caution here. Indeed, the XANES and RMN studies show the

©2003 Taylor & Francis

existence of mixed tetrahedra SiCx N4−x . For temperatures below 1400◦ C, the environment of Si atoms are composed of 3N + 1C (around the same Si atom) or 4N. This allows us to make the hypothesis of a relatively well chemically ordered SiCN structure in which C atoms are homogeneously distributed. The chemical order is then different from those found in crystallised SiC and Si3 N4 . At 1500◦ C, a broadening of the peak relative to Si environment measured by RMN is observed. This broadening is attributed to an increase of chemical disorder around the Si atoms: all the environments are detected for Si atoms. This increase of the chemical disorder is acccompanied by an increase of the NSi C /NSi N ratio. In short, before 1400◦ C we expect the existence of a nano-ordered α-SiCN structure composed of SiCN3 and SiN4 tetrahedra and of C N bonds. At 1500◦ C, nano-ordered β-SiCN, in which a beginning of demixion occurs, are formed. Gradients of C concentrations appear at the atomic scale in the β-SiCN structure. Locally, an early formation of six member rings of SiC occurs but the number of mixed tetrahedra remain dominant.

5.5.2.4

Nanocrystalline SiCN structural models

Building of the models and calculations From the experimental results, the hypothesis of nano-ordered SiCN structures was tested. This could be achieved by the modelisation of nanocrystalline structures involving experimental data (XAS, RMN, XPS and diffraction). In order to simulate the α → β transition, we have decorated the structure of the two polytypes α-Si3 N4 and β-Si3 N4 with C atoms substituting Si and N atoms in such a manner that mixed tetrahedra must be formed as well as C N bonds. C atoms substituted to Si form CN4 tetrahedra and C atoms substituted to N atoms form CSi3 sites (Table 5.5.2). Then, a relaxation of the mechanical stress is performed by using an algorithm which allows the potential energy minimisation of the substituted structure. We used the CERIUS software to build the models and to perform potential energy minimisation (see section modelisation). The bond distances in the models were forced to those given by the experimental analysis (1.74 Å for Si N, 1.89 Å for Si C and 1.5 Å for C N) during the relaxation process. This could be done by the addition in the energetic expression of a term which is zero if the bond distances reach the constraint values. We obtained new periodic (low symmetric) SiCN structures in which C atoms are bonded to Si and N atoms. Figure 5.5.8 presents the α-substituted model which is related to 1400◦ C. The theoretical scattered intensities of the models are calculated using the Debye formula (Debye 1915):

I (q) =



fn (q)fm (q)

n,m

sin(qrnm ) qrnm

(5.5.6)

where I (q) is the total scattered intensity, rnm the distance between the atoms n and m, fn (q) and fm (q) the scattering factor. The sum is taken over all the atomic pairs of the cluster. The theoretical structure factors are deduced and compared with the experimental ones. The calculations were performed using the INSIGHT program.

©2003 Taylor & Francis

Table 5.5.2 Construction of the α- and β-substituted models and structural parameters 1400◦ C (24 h)

1500◦ C (4 h)

1 Construction of the cluster of crystalline Si3 N4 α-Si3 N4 (P 31c; a = b = 7.753 Å, c = 5.618 Å; α = β = 90◦ , γ = 120◦ ) 16 cells (2a × 2b × 4c) 448 atoms (1.5 nm × 1.5 nm × 2 nm)

β-Si3 N4 (P 63/m; a = b = 7.606 Å, c = 2.909 Å; α = β = 90◦ , γ = 120◦ ) 32 cells (2a × 2b × 8c) 448 atoms (1.5 nm × 1.5 nm × 2 nm)

2 New description of the structures Space group → P1

Space group → P1 gathering of 2 cells: (1a × 1b × 2c) → (1a × 1b × 1c′ )

3 Periodic substitution of Si and N atoms by C atoms (Experiment – HMDS44) (Experiment – HMDS44) 160Si → 36% Si 38.9% 176Si → 39% Si 41.0% 224N → 50% N 42.8% 176N → 39% N 41.2% 64C → 14% C 15.5% 96C → 22% C 15.9% O 2.9% O 2.4% SiCN3 in the majority SiC2 N2 , SiCN3 SiN4 , SiC3 N2 in the minority a few CN4 tetrahedra CN4 tetrahedra 4 Relaxation of the geometric stresses d Si C = 1.89 Å ± 0.02 Å d Si N = 1.74 Å ± 0.02 Å d N C = 1.5 Å ± 0.02 Å N(C) Si N(C) from 100◦ to 120◦ Si N Si(C) from 105◦ to 135◦ Si(N) C Si(N) from 100◦ to 120◦

d Si C = 1.89 Å ± 0.02 Å d Si N = 1.74 Å ± 0.02 Å d N C = 1.5 Å ± 0.02 Å N(C) Si N(C) from 90◦ to 125◦ Si N Si(C) from 105◦ to 135◦ Si(N) C Si(N) from 100◦ to 130◦

Results of the calculations and discussion NEUTRON

The neutron structure factor Sn (q) of the model relative to 1400◦ C (α-substituted) and for the reference α-Si3 N4 are given in Figure 5.5.9. Figure 5.5.10 presents the calculated spectra for the model relative to 1500◦ C (β-substituted) and for the corresponding cluster of β-Si3 N4 . The hypothesis of nanocrystal of α-Si3 N4 can be dismissed. The structures of the experimental spectrum (1400◦ C) are too different from those of the α-Si3 N4 cluster. The general features of the experimental spectra are well reproduced by the models. In particular, the evolution observed in the low q domain (q < 6 Å−1 ) characterised by the splitting in two components of the FSDP at 1500◦ C is reproduced. For q > 6 Å−1 , the interference functions exhibit small sharp diffraction peaks as expected since the models are small crystallites of α- or β-SiCN. These structures which are not present in the experimental spectra simulate a too strong topological order in the models which are derived from crystal structures. The geometry obtained after relaxation is close to those of crystalline α- and β-Si3 N4 . The networks of connections between the atoms are the same as in the crystals since the C atoms are only substituted to N and Si atoms and the mixing of the valence states is the same as in silicon nitride. Nevertheless, the haloes which support the small sharp diffraction structures

©2003 Taylor & Francis



N C Si

Figure 5.5.8 Nanocrystalline α-SiCN model for the intermediate compositions (HMDS43, HMDS44). This structure is relative to 1400◦ C.

are reproduced as well as the width and position of the first sharp diffraction peak which show that the size of the nano-structures to be considered in the calculations are those of the models. To improve the experiment–theory agreement, we simulated the loss of the medium– long-range order in the model (1400◦ C) by a Fourier Transform (FT) filtering of the high r correlations in the RDF. This corresponds to the disappearance in the interference function of high frequency contributions. We performed the FT on the theoretically calculated S(q), we suppressed the high r contributions (r > 5 Å) and then we made a back FT calculation on the filtered real space contribution (0 → 5 Å). Figure 5.5.9 presents the results for the α-model. The process smooth high frequencies contributions and the agreement with the experiment at high q values is better. Unfortunately, the FSDP is broadened by this process and the agreement in the low q (q < 4 Å−1 ) domain is deteriorated. We expect that this procedure will simulate a too drastic increase of the structural disorder beyond 5 Å. The smoothing process simulates an abrupt decrease of the structural order beyond 5 Å around each of the 448 atoms of the model. From these results, we can conclude that the experimental spectrum is intermediate between those of non-smoothed model and smoothed model the sizes of which (1 nm) are close to those of local order models obtained from EXAFS measurements. It is interesting to note that for the β-model (1500◦ C), all the small diffraction peaks appear on the experimental spectrum of the sample annealed at 1550◦ C.

©2003 Taylor & Francis

-Si3N4

Sn (q)

-SiCN model FF (model) Experiment-1400°C

0

5

10

15

q (Å–1)

Figure 5.5.9 Calculated neutron interference functions (Sn (q)) for the α-model compared with the experimental spectra (1400◦ C). The Fourier filtered (FF) calculated spectrum is given as well as the crude one for the cluster of α-Si3 N4 .


-Si3N4

Sn (q)


-SiCN model

Experiment/1550°C-4h

Experiment/1500°C-4h

0

5

10

15

q (Å–1)

Figure 5.5.10 Calculated neutron interference functions (Sn (q)) for the β-model compared with the experimental spectra (1500◦ C and 1550◦ C). The calculated spectrum for the cluster of β-Si3 N4 is also given.

Figure 5.5.11 presents the calculated and experimental RDF. The evolution of the MRO is reproduced by the calculations. X-RAY

The structure factors SX (q) of the models compared to the experimental ones are presented in Figure 5.5.12. The Fourier filtered theoretical spectra are also given (the correlations at r ≥

©2003 Taylor & Francis

Experiment

RDFn

1400°C-24 h 1500°C-4 h

RDFn

Models 1400°C-24 h 1500°C-4 h

2

r (Å)

4

6

Figure 5.5.11 Calculated and experimental neutron RDF. The disappearance of the correlation at 3.9 Å is well reproduced by the modification of the stacking mode (α → β) when going from 1400◦ C to 1500◦ C.

4 Å for the α-model and r ≥ 3.8 Å for the β-model are removed). As for the neutron calculated spectra, the X-ray calculated spectra exhibit a lot of small sharp structures relative to longrange interatomic correlations in the models, but contrary to neutron, experimental spectra exhibit the signatures of long-range correlations in the high q domain (q > 5 Å−1 ). These correlations are pretty well reproduced by the α-model (1400◦ C). The experimental spectra are, on the whole q range, intermediate between the theoretical smoothed and theoretical non-smoothed spectra. The differences concerning the low q domain are also reproduced by the calculation: the X-ray spectra do not exhibit a FSDP. The main result of these calculations is the existence of nano-ordered structures α- and β-SiCN in these powders. The structure change when the annealing temperature increases and the modifications of local order are accompanied by ‘phases transition’ which occur in the amorphous state. This result underlines the importance of the existence of the β-SiCN (β-Si3 N4 substituted and relaxed) phase in the nucleation of β-SiC crystallite. The nucleation of β-SiC is beginning within the β-SiCN phase at 1500◦ C. For intermediate compositions (HMDS43 and HMDS44), the C content does not allow the local formation of β-SiC nuclei at 1500◦ C. More energy is needed to allow the diffusion of C atoms in the β-SiCN structure to reach locally the SiC stoichiometry. At 1500◦ C, for the rich C powders (HMDS40 and HMDS42), the β-SiCN structure is saturated by C atoms; locally one finds SiC4 tetrahedra and nuclei of β-SiC that are formed. The excess C forms a free C phase. For the N-rich powder

©2003 Taylor & Francis

(a) -SiCN model

SX(q)

FF (model)

Experiment – 1400°C

0

5

10

15

q (Å–1) (b)

SX(q)


-SiCN model

FF (model)

Experiment – 1500°C

0

5

10

15

q (Å–1)

Figure 5.5.12 (a) Calculated X-ray interference functions (SX (q)) for the α-model compared with the experimental spectra (1400◦ C), (b) calculated X-ray interference functions (SX (q)) for the β-model compared with the experimental spectra (1500◦ C). The Fourier filtered (FF) spectrum is also given.

(HMDS45), there is not enough C in the structure to prevent the formation of nano-crystallites of β-Si3 N4 .

References Debye, P. (1915) Ann. Physik, 46, 809. Dixmier, J. (1992) J. Phys. I France, 2, 1011. Dixmier, J., Bellissent, R., Balhoul, D. and Goursat, P. (1994) J. Eur. Ceram. Soc., 13, 293. Dürr, J., Schempp, S., Lamparter, P., Bill, J. and Aldinger, F. (1999) Precursor-derived Ceramics (eds J. Bill, F. Wakai and F. Aldinger), Wiley-VCH, p. 224. Hajdu, F. (1971) Acta Crystallogr., A27, 73.

©2003 Taylor & Francis

Mayo, S. L., Olafson, B. D. and Goddard III, W. A. (1990) J. Phys. Chem., 94, 8897; (1992) Neutron News, 3(3), 29–37. Placzeck, G. (1952) Phys. Rev., 86(3), 377. Sazaki, S. (1984) KEK report, anomalous scattering factors for synchrotron radiation users, calculated using Cromer and Liberman’s method, National laboratory for high energy physics. www-lib.KEK.jp/cgi-bin/img_index?198324022. Waasmaier, D. and Kirfel, A. (1995) Acta Crystallogr., A51, 416. Warren, B. E. (1969) X-ray Diffraction, Addison-Wesley Publishing Company. Waseda, Y. (1984) Lecture notes in physics, novel application of anomalous X-ray scattering for structural characterization of disordered materials, Springer Verlag, Berlin Heidelberg, New York, Tokyo.

©2003 Taylor & Francis

6

Conductivity and dielectric properties of Si/C nanopowders Abdelhadi Kassiba

6.1 Introduction Nano-sized powders of silicon carbide (SiC) have recently attracted a great deal of interest mainly oriented towards the improvement of mechanical and thermal properties of nanocomposite materials such as MgO SiC, Al2 O3 SiC (Niihara and Nakahira 1991) and oxide matrices reinforced by SiC fibres (Karlin and Colomban 1998; Monthioux and Delverdier 1996). The nanometric SiC grains are obtained by pyrolysis of polycarbosilane polymers in inert gas (Yajima 1980) or by laser pyrolysis of (SiH4 , C2 H2 ) gaseous mixture (Cauchetier et al. 1988). Whatever the used process, the final network contains carbon (C) in excess arranged as isolated clusters or associated with the SiC grains (Charpentier et al. 1999, Monthioux, Chapter 4). Several reports have pointed out the correlation between the structural, mechanical and electrical properties of SiC fibres with the C in excess content (Karlin and Colomban 1998; Mouchon and Colomban 1996; Monthioux and Delverdier 1996). Particularly, the enhanced thermal stability of the ceramics and their improved mechanical features as well as the electrical conductivity of SiC-based ceramics depend on C excess, which prevails in materials annealed at temperatures below 1400◦ C. Electron paramagnetic resonance (EPR) investigations (Kassiba, Chapter 5.2) on the SiC nanopowders, reveal a high content of delocalised unpaired spins (1019 spin g−1 ) thermally activated by low energies (∼10 meV) (Charpentier et al. 1999). Furthermore, a drastic damping of an EPR λ/2-resonant cavity was found in these materials when annealed above 1400◦ C. Dielectric losses by conducting paths modify the quality factor of the cavity and the microwave absorption rate. All experimental investigations (Charpentier et al. 1999; Kassiba et al. 2000) have demonstrated that annealing alters irreversibly all the physical properties including the grain morphology, size and composition as well as the crystalline structure and interface features. The latter which are dominated by the C-rich phase at the outermost particle surfaces and the disorder are expected to play an important role in the conduction process and dielectric behaviour. Moreover, we have demonstrated earlier (Kityk et al. 2000), that SiC nanocrystallites embedded within a polymer matrix show promising nonlinear optical phenomena. In that case, the interface SiC-polymer is of particular importance on the induced second harmonic generation signal. This can open new areas through the search of original electronic and non-linear optical properties in materials based on SiC nanocrystallites. The following report is dedicated to the analysis of the dielectric behaviour, electrical and transport properties in SiC nanopowders with regard to the structure, grain sizes and interface features. Different experimental approaches are conjugated for a better understanding of the electrical properties. Among them, we have used dielectric relaxation spectroscopy

©2003 Taylor & Francis

(DRS), microwave method (MWC), 4-points method and Hall effect measurements. These investigations are carried out as functions of annealing which modulate in large extent the physical properties of the network. Measurements with variable temperatures (150, 570 K) and frequencies (0.1 Hz, 10 GHz) are achieved for a quantitative understanding of the electronic, relaxation and transport phenomena in these SiC nanopowders.

6.2 Material specifications SiC nanopowders are synthesised by CO2 laser pyrolysis of gaseous mixture (SiH4 , C2 H2 ) (Cauchetier et al., 1988, Chapter 2). The rate of C in excess is monitored by the gas fluxes while the residence time of the reactants in the laser beam monitors the particle diameters in the as-formed powder. The composition, the SiC crystalline structures and the particle sizes can be modulated to a large extent by annealing under argon up to 1800◦ C (Charpentier et al. 1999). The mean particle diameter varies from ∼30 nm (annealing temperature 1200◦ C) to the micrometer range (annealing at 1800◦ C). Transmission electron microscopy (TEM) investigations (Monthioux, Chapter 4) have shown that the grains are constituted by SiC polytypes partially covered by thin layers of C in one batch SiC177 (C in excess 10%) or by an oxide in SiC212 (C in excess 4%). Structural investigations import on the involved SiC polytypes which are essentially 6H and 3C polytypes with annealing-dependent ratios. Amorphous SiC is also present in a minor concentration (≤10%) as well as sp2 C for annealing below 1500◦ C in SiC212 and remains in SiC177 annealed up to 1800◦ C. Whatever the C in excess rate (SiC163: 1.11, SiC173: 1.10 and SiC212: 1.04), the batches exhibit similar electrical behaviour at the same annealing stages. So, the following report is limited to the SiC212 batch with the powders referred below as SiC212/1200◦ C and SiC212/1800◦ C, following their annealing temperatures at 1200◦ C and 1800◦ C.

6.3 Experimental results and interpretation 6.3.1

Dielectric losses and electrical conductivity in the microwave range (MWC)

These measurements are of a particular interest since neither ohmic contacts nor a particular sample shape processing is required. A simple diagnostic of the conducting states in the 10 GHz domain can be obtained easily. This requires an EPR spectrometer with a λ/2-resonant cavity in order to evaluate the dielectric losses in the SiC nanopowders. The mechanism is based on a damping of the resonant cavity and lowering its quality factor by conducting materials. The resonant cavity design is devoted to enhance these effects by using a small amount of the SiC powder (∼10 mg) located in the cavity region where the microwave electric field is maximum. In all the investigated SiC batches, Ta = 1500◦ C corresponds to the critical temperature where the dielectric losses are manifested and alters significantly the microwave response of the materials. Figure 6.1 shows the behaviour of the reflected microwave power by the SiC nanopowders annealed between 1200◦ C and 1800◦ C. Narrow dips correspond to a minimum of the reflected power realised in the samples annealed below 1500◦ C. Conversely, flat and broad dips correspond to the materials annealed between 1500◦ C and 1650◦ C. This behaviour is consecutive to the occurrence of macroscopic conduction induced by annealing at T ≥ Ta .

©2003 Taylor & Francis

Reflected microwave power (a.u.)

1800°C

1550°C

1200°C –2 × 10–3

2 × 10–3

0.0 /( – 0)

Microwave electrical conductivity (S m–1)

Figure 6.1 Reflected microwave power by SiC212 powder located in the region of an X-band (ω0 = 9.5 GHz) EPR resonant cavity where the microwave electric field is maximum. The dip shapes represented from bottom to top are due to the samples SiC212 annealed at 1200◦ C, 1300◦ C, 1400◦ C, 1500◦ C, 1550◦ C, 1600◦ C, 1650◦ C, 1700◦ C and 1800◦ C. The maximum of dielectric losses (flat dip) is realised in the SiC212/1550◦ C sample.

0.8

SiC212

1800°C 1650°C 1500°C

0.6

0.4

0.2 100

150

200

250

300

350

400

450

500

T (K)

Figure 6.2 Electrical conductivity in the microwave range v. the sample temperature in SiC212 nanopowders annealed at 1500◦ C (
), 1650◦ C (•) and 1800◦ C (
); the lines are a guide for the eye.

Quantitative measurements of the electric conductivity by using the process developed by J. G. Theobald in Besançon University (Frand et al. 1995; Kassiba et al. 1999) are shown in Figure 6.2. The conductivity increases with the sample temperature and follows qualitatively an activated law: σmW (T ) = σ0 e−(Eg /kT ) with [σ0 (S m−1 ), Eg (meV)] ≈ [1.1, 20], [0.7, 6], [0.9, 9] respectively for annealing at 1500◦ C, 1650◦ C and 1800◦ C.

©2003 Taylor & Francis

4.0

ln( (T ))

3.5

SiC212 1650°C

3.0

2.5

2.0 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 1/T (K–1)

Figure 6.3 Electrical conductivity measurements in SiC212 by the conventional 4-points method. The evolution with the temperature shows a thermal activation of the conductivity with energies, 3 meV below 185 K and 30 meV above.

Then, it seems that a substantial electrical conductivity is induced by annealing above 1400◦ C with a thermal activation of the charge carrier motions by low energies 6–20 meV. These values are in agreement with those required to delocalise some paramagnetic species (Kassiba, Chapter 5.2). 6.3.2

Electrical conductivity by the 4-points method

The samples are prepared as pressed pellets (diameter 12 mm and 0.5 mm thick). On each pellet, four gold contact points are deposited by CVD technique. Annealing under hydrogen at 300◦ C during 5 min improves the contact resistors which change from 1000 to 40 J and remain ohmic on all the explored temperature range. Figure 6.3 shows the dc conductivity v. the sample temperature in the SiC212/1650◦ C sample. Above 185 K, a thermally activated conductivity is found, that is, σdc (T ) = σ0 exp(−Eg /kT ) with Eg ∼ 30 meV and a pre-factor σ ≈ 60 S m−1 . The activation energy is in agreement with those obtained from the above microwave characterisations (MWC). However, a discrepancy is found in the absolute conductivity values. The main origins are plausibly due to the shape of the investigated materials, the different porosity degree and the inhomogeneous conducting paths. Furthermore, transport measurements by the Hall effect are carried out in order to determine the nature and concentration of the charge carriers as well as their mobility. n-type conduction, charge carriers concentration N ∼ 1019 cm−3 and a mobility µ ∼ 1 cm2 V−1 s−1 are the main parameters of the electronic transport in these SiC nanomaterials. 6.3.3

Dielectric relaxation spectroscopy (DRS)

Theoretical background and phenomenology DRS determines the relative complex permittivity ε∗ (ω, T ) = ε ′ (ω, T ) − iε ′′ (ω, T ) v. the alternative applied voltage frequency (ω/2π ) and the sample temperature T . The overall

©2003 Taylor & Francis

complex conductivity is deduced from the relation: σ ∗ (ω, T ) = iωε0 ε∗ (ω, T ), where ε0 is the vacuum permittivity. As far as the forthcoming experimental facts are concerned, the real conductivity part can be set in the form of contributions from the steady-state (σdc ) conductivity and a frequencydependent part as follows: σ ′ (ω) = σdc + ωε0 ε′′ (ω) However, phenomenological representations are usually employed to account for the frequency behaviour of ε∗ (ω) in heterogeneous media (Hunt 1998). Among them, the Havriliak Nagami (HN) response (Havriliak and Nagami 1966), which reads: ε∗ (ω) =

σdc !ε + ε∞ + iε0 ω (1 + (iωτ )α )β

!ε represents the difference between the low frequency limit, that is, the relaxed permittivity (εs ) and the high frequency limit ε∞ with respect to the explored frequency range. The phenomenological parameters α, β monitor the asymptotic behaviour of ε∗ (ω). Indeed, when ωτ ≫ 1; ε′′ ∼ ω−αβ and for ωτ ≪ 1; ε′′ ∼ ωα , τ , represents also a phenomenological relaxation time (Hunt 1998) even if it can be associated to the experimental dielectric loss peak position. Experimental details Novocontrol Broadband Dielectric Spectrometer is well adapted for a class of materials with a conductivity lying between 1 S cm−1 to that of high insulating materials. Measurements of the dielectric function are possible over a wide frequency range (0.1 Hz–1 GHz) by using two experimental configurations. From 0.1 Hz up to 10 MHz, a solartron Si1260 combined with a broadband dielectric converter (BDC) determines the overall sample impedance. In the high frequency limit (1 MHz–1 GHz) an HP4291 radio frequency impedance analyser is used with a precision coaxial line (Kremer and Arndt 1997). For both configurations, the sample cell consists of two round golden plates electrodes filled with the material (pressed pellet with diameter 12 mm and 0.5 mm thick) to form a capacitor. In the high frequency range, the sample capacitor is used as a termination of a golden coaxial line. The corresponding impedance is deduced from the complex reflection factor at the end of the line. The measurements were performed as a function of temperature between 198 and 573 K. A gas stream of nitrogen (N) is used and the measured temperature close to the sample is controlled within the accuracy of ±0.1 K. Dielectric behaviour Due to the critical annealing temperature Ta = 1500◦ C, which alters significantly all the physical properties of the SiC nanopowders, we have examined in detail the conduction and dielectric behaviour of two representative samples, namely SiC212/1400◦ C and SiC212/1700◦ C. In the former, the absence of pertinent details on the dielectric functions, above 10 MHz, allows us to limit the measurements to the frequency range 0.1 Hz, 10 MHz. Figure 6.4(a) and (b) shows in a log–log scale, the frequency and thermal evolution of the dielectric function [ε′ (ω), ε′′ (ω)]. The continuous curves are plotted from HN response with the parameters

©2003 Taylor & Francis

(α1400 = 1; β1400 = 0.5). α and β are independent of the sample temperature (Ts ) as well as ε∞ = 8. Following the frequency range, two variation regimes of ε ′′ (ω) are evidenced. In the low frequency side, the linear variation occurs from the steady-state dc conductivity. The second regime, in the high frequency side, is marked by a dielectric loss peak centred at a frequency νmax = 1/τmax , which varies from ∼10 Hz to ∼1 MHz with increasing Ts . The real dielectric function ε′ (ω) is also marked by a relaxation process with the frequency νmax and by a low frequency dispersion (LFD). The latter behaviour depends on the experimental conditions and testifies the existence of contributions from the contacts between the sample and the electrodes. So, in the adjustment of ε ′ (ω) by the (HN) equation, we have ignored the (LFD) contribution. The relaxed parameter εs1400 ≈ 140 is one order of magnitude larger than the value in bulk crystalline SiC. This can be due to an interfacial polarisation favoured by the high specific surface of the SiC212/1400◦ C sample and by the inhomogeneous particle surface composition through the predominance of C in excess. (a)

100

⬘

573 K 513 K 468 K 423 K 378 K 333 K 288 K 243 K 198 K HN fit

10 10–1

100

101

102 103 Frequency (Hz)

104

105

(b) 108

106

573 K 513 K 468 K 423 K 378 K 333 K 288 K 243 K 198 K HN fit

107 106 105

⬙

104 103 102 101 100 10–1

10–1

100

101

102 103 Frequency (Hz)

104

105

106

Figure 6.4 Log–log representation of the (a) real and (b) imaginary dielectric function v. the frequency in the sample annealed at 1400◦ C. The continuous lines are fits from HN equation. Similarly (c) and (d) represent the dielectric function in the SiC212/1700◦ C sample.

©2003 Taylor & Francis

(c) 50 40

⬘

30

20

10

193 K 273 K 373 K 453 K 533 K HN fit 106

107

108

109

Frequency (Hz)

(d)

103 193 K 273 K 373 K 453 K 533 K HN fit

⬙

102

101

100 106

107

108 Frequency (Hz)

109

Figure 6.4 (Continued).

Similar investigations were performed on the SiC212/1700◦ C sample (Figure 6.4(c) and (d)). However, higher frequencies (up to 1 GHz) are needed in order to characterise the dielectric loss peak as in the SiC212/1400◦ C sample. The phenomenological parameters are (α1700 = 0.5; β1700 = 1) and ε∞ = 11, independent of Ts . The relaxation process is marked by a characteristic frequency (νmax ) in the order of 100 MHz and increases towards high frequencies with increasing Ts . The relaxed dielectric constant, εs1700 = 40, is quite lower than in the SiC212/1400◦ C sample. This is connected with the grain surfaces being striped from the C in excess (Charpentier et al. 1999) and with a coalescence of SiC particles, that is, mean diameter 1600 nm for annealing at 1700◦ C, limiting the interface polarisation effects which prevail in the SiC212/1400◦ C sample.

©2003 Taylor & Francis

1 × 10–4

(a)

1 × 10–5 Real conductivity ⬘ (S cm–1)

1 × 10–6 1 × 10–7 1 × 10–8

573 K 513 K 468 K 423 K 378 K 333 K 288 K 243 K 198 K HN fit

1 × 10–9 1 × 10–10 10–11 10–12 –1 10

100

101

102

103

104

106

105

(b)

1

Real conductivity ⬘ (S cm–1)

Frequency (Hz)

0.1

193 K 233 K 373 K 453 K 533 K HN fit 0.01

106

107

108

109

Frequency (Hz)

Figure 6.5 Log–log representation of the conductivity v. the frequency obtained by DRS spectroscopy: (a) SiC212/1400◦ C and (b) SiC212/1700◦ C.

Electrical conductivity In SiC212/1400◦ C and SiC212/1700◦ C samples, the frequency dependence of the conductivity are depicted in Figures 6.5(a) and (b) where the continuous lines are drawn from the (HN) equation. We may note that the behaviour at low frequencies is dominated by σdc up to crossover values about 1 kHz and 10 MHz respectively in the SiC212/1400◦ C and SiC212/1700◦ C samples. These values are temperature dependent and follow qualitatively the behaviour of the dielectric loss peak positions. In the high frequency side, no universal behaviour such as σac (ω) ∝ ωp with 0 < p < 1 was identified (Jonscher 1977; Funke 1993). This excludes any speculation on the real nature of the conduction process in these heterogeneous media. The thermal behaviour of the dc conductivity are summarised in Figures 6.6(a) and (b). In the SiC212/1400◦ C sample, an activation law σdc (Ts ) = σ0 exp(−Eg /kTs ) with Eg = 500 meV and a pre-factor σ0 = 7 S m−1 are realised above T1 = 330 K. Similarly, in the

©2003 Taylor & Francis

1 1×10–5

2

3

4

6 10–3

SiC1400 SiC1700

1×10–6 1×10–7

dc (S cm–1) (SiC1700)

dc (S cm–1) (SiC1400)

5

1×10–8 1×10–9 1×10–10 10–11 10–12

10–4 1

2

3

4

5

6

1000/Ts (K–1)

Figure 6.6 Dc conductivity v. 1000/T obtained from DRS spectroscopy measurements. Activated conductivity following Arrhenius law is realised in a limited temperature range.

SiC212/1700◦ C sample, by limiting the temperature range to 250 ≤ Ts ≤ 430 K, σ (Ts ) is also accounted by an activated process with an Eg1700 = 35 meV and a pre-factor σ0 = 0.3 S m−1 . Below 250 K, the thermal variation of the conductivity is characterised by a continuous decrease of the activation energy without any defined variation law such as that relevant in a variable range hopping process (Mott and Davis 1979) or that used in SiC fibres (Chauvet et al. 1992). Above 430 K, as shown in Figure 6.6(b), the conductivity exhibits a rounded shape behaviour (negative temperature coefficient). This process is reversible but it is no longer connected with a metallic character of the SiC212/1700◦ C sample (Kassiba et al. 2000). The observed behaviour has to be analysed with respect to the conductivity supports in the material, that is, charge carriers diffusion in the crystallite bulk, tunnelling through the interface barriers as well as the material porosity. All these parameters have to be optimised for relevant insights on the particular behaviour around 430 K observed in all the SiC nanomaterials annealed above 1400◦ C. It is worth noticing the agreement between the absolute conductivity values and the activation energies obtained by DRS and MWC methods even if different sample processing (pellet, powder) are used. With regard to the relaxation process, the parameter νmax varies with the sample temperature as the inverse of the dc conductivity (Figures 6.7(a) and (b)). From the experimental value of νmax and limiting the temperature range to that where an Arrhenius law is realised, we have established the following relations: −11 5890/T τ 1400 e (s) ≈ max = 3.10

ε0 ε∞ 1400 σdc (Ts )

and −10 450/T τ 1700 e (s) ≈ max = 2.10

©2003 Taylor & Francis

ε0 ε∞ 1700 σdc (Ts )

(a) 1e – 5 1e – 6

dc (s cm–1)

1e – 7

1e – 8

1e – 9

1e – 10

1e – 11 1e – 6

1e – 5

1e – 4

1e – 3

1e – 2

1e – 1

1e + 0

max (s) (b)

dc (s cm–1)

1e – 3

1e – 4 4e – 10

1e – 9 max (s)

2e – 9

Figure 6.7 Representation of the static conductivity v. the relaxation time of the HN response in the samples (a) SiC212/1400◦ C and (b) SiC212/1700◦ C.

This proportionality is quite general and encountered both in electronic (Hunt 1993; Pelster et al. 1994) and in ion conducting systems (Macedo et al. 1972; Moynihan 1994) as well. No physical supports have been clearly assigned to this conductivity-relaxation dependence as relevant in transport phenomena dominated by surface or bulk effects. On the other hand, EPR investigations (Charpentier et al. 1999) have evidenced charged vacancies, in the SiC crystallites, submitted to a thermal delocalisation by energies 13 meV. Qualitatively, this value can be looked as the binding energy of individual charges to their local sites and constitute a low contribution to the overall activation energy associated with the macroscopic conductivity. The main part comes from the energy barriers at the grain boundaries. The exact mechanism cannot be drawn rigorously but it is worth noting that

©2003 Taylor & Francis

the large difference in the activation energies 500 meV and 35 meV respectively in the SiC212/1400◦ C and SiC212/1700◦ C samples agrees with that between the specific surfaces (SiC212/1400◦ C : 35 m2 g−1 and SiC212/1700◦ C : 1.2 m2 g−1 ). This analysis is supported by the magnitude of the static dielectric constant being enhanced in materials with high specific surfaces due to the interface polarisation effects. Then, surface effects at the grain boundaries monitor the electrical behaviour of these powders and the conductivity-relaxation dependence can be attributed to such effects.

6.4 Conclusion We have used complementary techniques in order to quantify the electrical, dielectric and transport properties in SiC nanopowders. A thermally activated conductivity is pointed out in the materials annealed at T ≥ Ta = 1500◦ C. At this stage, a removal of the C in excess is operated as well as a coalescence of the SiC particle. These modifications favour the development of conduction channels through the increase of the mean free path of the charge carriers and the lowering of the barrier thickness between the grains. The interface effects seems crucial and explain the conductivity and dielectric behaviour in these media. The interpretation correlates the dielectric properties with the structural and morphological changes by annealing. It is worth noting also the possibility to modulate the electrical behaviour of SiC nanopowders as well as their effective dielectric constant by suitable thermal treatments.

References Cauchetier, M., Croix, O. and Luce, M. (1988) Adv. Ceram. Mater., 3(6), 147. Charpentier, S., Kassiba, A., Bulou, A., Monthioux, M. and Cauchetier, M. (1999) Europ. Phys. J: AP, 8, 111. Charpentier, S, Kassiba, A., Emery, J. and Cauchetier, M. (1999) J. Phys. Cond. Matter, 11, 4887. Chauvet, O., Stoto, T. and Zuppiroli, L. (1992) Phys. Rev. B, 46(13), 8139. Frand, G., Bohnke, O., Lacorre, P., Fourquet, J. L., Carré, A., Eid, B., Theobald, J. G. and Gire, A. (1995) J. Solid State Chem., 120, 157. Funke, K. (1993) Prog. in Solid State Chem., 22(2), 111. Havriliak, S. and Nagami, S. (1966) J. Polym. Symp., 14, 89. Jonscher, A. K. (1977) Nature, 267, 673. Karlin, S. and Colomban, Ph. (1998) Composites, 29B, 41. Kassiba, A, Tabellout, M., Charpentier, S., Herlin, N. and Emery, J. R. (2000) Solid State Comm., 115, 389. Kityk, I. V., Makowska, M., Kassiba, A. and Plucinski, K. J. (2000) Opt. Mater., 13, 449. Kremer, F. and Arndt, M. (1997) Dielectric Spectroscopy of Polymeric Materials, Ch. 2 (eds J. P. Runt and J. J. Fitzgerald), American Ceramic Society. Macedo, P. B., Moynihan, C. T. and Bose, R. (1972) Phys. Chem. Glasses, 13(6), 171. Monthioux, M. and Delverdier, O. (1996) J. Eur. Ceram. Soc., 16, 721. Mouchon, E. and Colomban, Ph. (1996) J. Mater. Sci., 31, 323. Mott, N. F. and Davis, E. A. (1979) Electronic Process in Non-Crystalline Materials, 2 edn. Clarendon Press, Oxford. Moynihan, C. T. (1994) J. Non-Cryst. Solids, 1395, 172. Niihara, K. and Nakahira, A. (1991) Ann. Chim. Fr., 16, 479. Pelster, R., Nimtz, G. and Wessling, B. (1994) 49(18), 12718. Yajima, S. (1980) Philos. Trans. R. Soc. London Ser., A 294, 419.

©2003 Taylor & Francis

7

Processing and tailoring of Si/C/N-based nanocomposites Paul Goursat, Djamila Bahloul-Hourlier, Benoît Doucey and Martine Mayne

7.1 Introduction Silicon nitride exists in two structural forms with a hexagonal unit cell α-Si3 N4 = (space group P31c ) (Lang 1977) and β-Si3 N4 (space group P63/m ) (Grün 1979). The covalent nature of Si N bonds provides very good mechanical properties, such as high strength and high toughness up to elevated temperatures. The sintering of a powdered sample corresponds to the formation of necks between grains due to the presence of evaporation–condensation and/or diffusion mechanisms with an atomic or molecular redistribution. In the case of Si3 N4 the self-diffusion for silicon (Si) or nitrogen (N) is very low in the tridimensional network of SiN covalent bonds, which explains the difficulties encountered to densify this material. Moreover, decomposition occurs at high temperature which blocks the increase in mobility of the species with rising temperature. These difficulties are usually partially solved in other systems by using nanosized powders, but in the case of Si3 N4 , it is necessary to add a driving force – mechanical and/or gas pressure, sintering aids, etc. – which explains the different types of processes and materials: reaction bonded silicon nitride (RBSN); hot-pressed silicon nitride (HPSN); hot isostatically pressed silicon nitride (HIPSN); gas pressured silicon nitride (GPSN). The additives used for the sintering of silicon nitride (MgO, Al2 O3 , Y2 O3 , etc.) react with silica always present at the surface of the grains and lead to the formation of an oxynitrided eutectic phase with the general formula MSiM′ ON, where M and M′ are metallic cations. These vitreous phases, which have been studied in quaternary or quinary systems (Figure 7.1), play a very important role during sintering and on the properties at high temperature of the sintered material (Hampshire et al. 1978). Experiments on these glasses and, in particular, the vitreous domain show that the N content changes with the metallic cations and the temperature range (Lewis 1993). Among the main characteristics of these materials it has been shown (Hampshire et al. 1994) that the viscosity increases with N content which explains that crystallisation is slower than for the corresponding oxide glasses and that after sintering, the thermodynamic equilibrium is not reached.

7.2 Sintering of silicon nitride with additives A lot of papers are devoted to the sintering of silicon nitride for modelling, the determination of mechanisms or the development of microstructures (Ziegler et al. 1987). During the heat treatment of the powdered sample (α-Si3 N4 +additives : Al2 O3 , Y2 O3 , etc.), the oxynitrided eutectic flows in the porosity and induces capillary forces, dissolution of fine particles and

©2003 Taylor & Francis

4YN

2Y2O3

4AIN

Si3N4 50

10 2AI2O3

3/2 Si2N2O

eq%N

3SiO2

Figure 7.1 YSiAlON system.

∆I I0

  +
III II

I Time

Figure 7.2 Shrinkage and α → β phase transformation v. time.

a rearrangement process. The shrinkage is usually described by three different overlapping mechanisms (Figure 7.2). 7.2.1

Particle rearrangement

The liquid phase wets the α-Si3 N4 grains and it localises mainly at the contact points. Under the forces of capillary action, a rapid densification is obtained by grain sliding. In the case of HP or HIP, the densification rate is sharply increased, the applied mechanical pressure being much more higher. The densification rate changes with the particle size, the viscosity and the volume fraction of the liquid. Different modelling emphasise the influence of these parameters. 7.2.2

Dissolution–Diffusion–Reprecipitation

The applied pressure induces a preferential dissolution at the contact points of α-Si3 N4 crystals in the liquid phase with concentration gradients. A kinetic regime is established with an intergranular migration of the dissolved species and a reprecipitation of β-Si3 N4 crystals or the β ′ -SiAlON solid solution when alumina is used as a sintering additive. The

©2003 Taylor & Francis

flow of mobile species explains the redistribution of matter and a macroscopic shrinkage. The comparison of the rate (dτ /dt) of the α → β transformation and the rate of densification (dρ/dt) shows that the ratio dτ/dt =1 dρ/dt corresponds to a diffusion regime. In the case of hot-pressing (Bowen et al. 1978), the rate of densification is given by the following equation which is derived from the Coble model established for the creep of materials: dρ 4.75 × P D b JW = dt G3 × KT

(7.1)

where P is the applied mechanical pressure; Db the diffusion coefficient in grain boundaries; J the volume of the mobile species; W the thickness of grain boundaries; G the diameter of the crystals; K is the Boltzmann constant. 7.2.3

Grain growth

When soluble particles with different particle sizes (radius r1 , r2 ) are in the presence of a liquid, the difference in solubility provides the driving force for grain growth which corresponds to the Ostwald ripening phenomena. The grain growth rate (Greenwood 1956) has been formulated as:

−2γ JD 1 1 dr1 (7.2) = C∞ − dt KT d r1 r2 C∞ is the equilibrium concentration for a planar surface, γ the surface energy, J the molecular volume of the mobile species, T the absolute temperature, K the Boltzmann constant, D the diffusion coefficient in the liquid phase, d the mean distance between particles 1 and 2. The grain growth kinetics by liquid phase sintering usually follows the equation: r¯ n − r¯on = kt

(7.3)

where r¯ is the mean particle radius, n = 3 the growth exponent, k the growth rate constant and t is time. As a matter of fact, for the sintering of silicon nitride, a lot of parameters are to be considered and the equations are more complex. The presence of acicular crystals of β-Si3 N4 in the starting powder hinders the rearrangement mechanism which overlaps the diffusion stage. It is better to use α-Si3 N4 with an equiaxed morphology from the kinetics point of view. The content and the size of β-Si3 N4 grains in the starting powder also play an important role during the 2nd and 3rd densification stage. β-Si3 N4 crystals can act as the nuclei during the reprecipitation process if their size is higher than the size of nuclei issued from the homogeneous precipitation. According to Kang and Han (1995), the growth of β grains along the c axis is due to the different roughness of the liquid–solid interface at an atomic scale of (100) and (001) planes. If the number of β-Si3 N4 crystals is low, the free grain growth of large crystals induces a microstructure with a bimodal particle size distribution. It is possible to control and adjust the microstructure of sintered materials with a given size or aspect ratio of β grains by adding β nuclei (Lee et al. 1994) or with a specific post sintering heat treatment (Rossignol 1994).

©2003 Taylor & Francis

The viscosity of the liquid phase also influencing the sintering rate is also very important. For the same content of liquid phase, impurities such as oxides of Ca, Mg, Na, Fe, etc. which lower the viscosity, facilitate the rearrangement and the diffusion processes. In return, other elements such as carbon (C) have a deleterious effect.

7.3 Sintering of silicon carbide Silicon carbide is usually densified by solid-state sintering with aluminium, boron and C as sintering additives (Greskowich et al. 1977). Recently it was shown that silicon carbide can be densified at 1800–1900◦ C with a liquid phase containing Y2 O3 and Al2 O3 (Kleebe 1992). The β → α-SiC transformation is observed with a growth of elongated α-SiC grains. It is also possible for SiC to design microstructures for specific properties.

7.4 Si3 N4 /SiC nanocomposites In recent years, much interest has been focused on nanophase materials, which, due to their small grain size, can offer improved properties compared to traditional materials (Siegel 1994). A further advantage of these materials is their ability to be hot-formed because of their fine microstructures (Waikai 1990). The main objective of this work is to elaborate Si3 N4 /SiC nanocomposites in order to fabricate ceramic parts by hot-forming. The approach consists in introducing SiC (N) nanometric powders as a second phase into a Si3 N4 matrix in order to influence the α-Si3 N4 to β-Si3 N4 phase transformation and the microstructural development of Si3 N4 /SiC nanocomposites. Usually nanocomposites are ordered in two groups (Niihara 1991): • •

micro–nano materials: a matrix with microscale crystals and an intragranular or an intergranular nanoscale second phase; nano–nano materials: a matrix and a second phase with nanoscale crystals.

These materials are fabricated according to three different routes. The sintering of mixtures of commercial silicon nitride with additives and a nanopowder of SiC. In the second route the nanoparticles are synthesised in situ during sintering by a reaction between silica and C deposited on silicon nitride crystals (Ishizaki and Yanai 1995). The last one corresponds to the sintering of a mixture of nanopowders: Si3 N4 , SiC and additives or Si/C/N and additives (Niihara et al. 1990). Till now, the number of papers devoted to the study of the sintering kinetics is very limited, in particular to determine the influence of nanoparticles on the different stages during the sintering of silicon nitride. 7.4.1

Dispersion and slip casting of Si/C/N nanopowders with Si3 N4 in organic liquids

It is well known that the method by which dispersions are prepared is a critical issue in forming high quality green bodies and reliable ceramic parts. The present study concerns the fabrication process of such materials which must be optimised because of the different sizes of powders, and the hard agglomerates in Si/C/N nanopowders. The aim is to obtain a homogeneous distribution of the nanometric second phase in order to prevent the presence of defects (essentially nanopowder agglomerates) decreasing drastically the mechanical properties.

©2003 Taylor & Francis

Si3 N4 used as the matrix is the commercially available α phase Si3 N4 powder (UBESNE 10). Si/C/N nanopowders used as second phase are synthesised by laser pyrolysis. Y2 O3 (Lonza) and Al2 O3 (Rhône Poulenc) are used as sintering aids. It has been observed by X-ray photoelectron spectroscopy (XPS), extended X-ray absorption fine structure (EXAFS) and X-ray near edge structure (XANES) that Si/C/N nanopowders obtained by laser pyrolysis are composed of mixed tetrahedra SiCx N4−x with a small amount of Si O bonds (Gheorghiu et al. 1992; Ténégal et al. 1996). The surface of such powders, analysed by IR spectroscopy, contains Si OH and Si H bonds (Cauchetier et al. 1991). Preparation of the green bodies Two successive stages are necessary to prepare green bodies: the mixing of the different powders and the forming process. In order to avoid the reagglomeration phenomenon induced by dry mixing, the mixing process in liquid medium is well adapted. Moreover, the use of organic liquids is preferable to prevent hydrolysis of silicon nitride based powders. Four different organic liquids are tested: ethanol, acetone, acetophenone and hexane. Each powder (Si3 N4 , Si/C/N) is mixed in each liquid (concentration = 50 g/l), and the evaluation of the most suitable liquid is realised by sedimentation tests. A mechanical device and a dispersing agent are chosen in order to, respectively, break down agglomerates and stabilise the suspensions. Two mechanical devices are selected: the ‘turbula’ (system Schutz) allowing powder mixing by helicoidal movement, and the ultrasonic probe (Vibra Cell 72412, Bioblock Scientific) involving the breaking of hard agglomerates. Two dispersing agents (ICI Chemicals) having different nature are tested: the basic PS2 surfactant soluble in alcohol and the acid PS4 surfactant soluble in ketones. Each powder (Si3 N4 or Si/C/N) is mixed using ‘turbula’ or ‘turbula’ + tultrasonic probe in ethanol + PS2 or acetophenone + PS4 . The evaluation of the most suitable mechanical device and dispersing agent is achieved by sedimentation tests. Viscosity measurements coupled with sedimentation tests allow one to determine the optimum concentration of the well-adapted dispersing agent. After the homogenisation and the dispersion of suspensions, the liquid must be removed in order to obtain green bodies. The forming process selected is slip casting in a plaster mould because it allows to preserve the dispersed state of the suspension. Organic liquid medium Whatever the powder (Si3 N4 or Si/C/N) and the organic liquid (ethanol, acetone, acetophenone, hexane), flocculation occurs in the suspensions. Sedimentation tests consisted of evaluating the relative height of sedimentation (hrelative = hsediment /htotal ). In the case of flocculated suspensions, a low relative height indicates a compact stacking of powders due to a beneficial effect of the liquid on particle dispersion. The evolution of relative height of sedimentation in each liquid is presented for the two silicon nitride based powders (Figure 7.3). For the two powders, an increase of the relative height is observed as follows: ethanol, acetone, acetophenone, hexane. Such an evolution corresponds to the classification of the more adapted liquid to the less adapted. The suitability of each liquid for powders can be explained by acid–base interactions between the surface sites of the particles and liquid molecules. The surface of Si3 N4 or Si/C/N powders is essentially composed of acid Si OH bonds and some basic Si2 NH and N H bonds.

©2003 Taylor & Francis

Relative sedimentation height

1.0 Si3N4 SiCN

0.8 0.6 0.4 0.2 0.0

Hexane Acetophenone Acetone Liquid

Ethanol

Figure 7.3 Evolution of the relative sedimentation height for each powder in each liquid.

The interactions liquid/powder result mainly from the attraction of basic liquid by acid Si OH groups. Thus, the more basic liquid (acetophenone) should be the best medium for the dispersion of powders. Meanwhile, the adsorption of large molecules, such as acetophenone, on all Si OH sites is difficult because of the rigidity of Si NH Si bonds and also of the repulsion between the lone electron pair of N and the π orbitals of benzene group. That’s why the linear ethanol molecule, even though it is less basic than acetophenone, is easily adsorbed on Si3 N4 and Si/C/N powders. Mechanical device for mixing and dispersion, and dispersing agent Whatever the liquid, flocculation occurs in the suspensions. So the use of a well-adapted mechanical device to disperse powders and of a dispersant to stabilise the suspensions is necessary. Sedimentation tests reveal that basic PS2 surfactant in ethanol does not have an effective action on powders, because for various dispersant contents, the suspensions remain flocculated. This is probably due to the low number of acid Si OH sites available for the adsorption of basic PS2 molecules. Effectively, these sites should be occupated by the basic ethanol molecules. The suspensions dispersed with PS4 (5% for Si3 N4 , 10% for Si/C/N, wt% of the dry matter content) are deflocculated, indicating the beneficial effect of this surfactant. The evolution of the sediment height with time for Si3 N4 and Si/C/N suspensions mixed with ‘turbula’ or ‘turbula’ + ultrasonic probe (Figure 7.4) indicates the positive effect of ultrasonic probe, suggesting that this device is necessary to break down agglomerates. The suitability of PS4 dispersant is linked to a good compatibility between acid surfactant molecules and basic surface sites of particles, non-occupated by the acetophenone molecules. Forming process For slip casting in plaster mould, suspensions should have high dry matter content (≥50 vol%) in order to eliminate all the liquid by capillarity. The studied slurry is composed

©2003 Taylor & Francis

Relative sedimentation height (mm)

7 6 5 4 3 2 1 0

1

2

3

4 5 Time (days)

6

7

8

Figure 7.4 Evolution of sedimentation height with time. T: Turbula, US: Ultrasonic Probe;
Si3 N4 ,  Si/C/N(T), × Si3 N4 (T + US), ' Si/C/N(T + US).

of a mixture containing of 71 wt% of Si3 N4 , 20 wt% of Si/C/N, 6 wt% of Y2 O3 , 3 wt% of Al2 O3 and acetophenone. The dry matter content is 60 wt%. The dispersant content adapted to this slurry is determined by both sedimentation tests and viscosity measurements. For a 2.5 wt% of PS4 , sedimentation tests indicate a sediment height near to zero during three days. This result gives evidence that the suspension is stabilised by 2.5 wt% of PS4 surfactant. Viscosity measurements as a function of dispersant content (Figure 7.5) reveal a minimum of viscosity around 2.5 wt% of PS4 which is in agreement with sedimentation tests. Slip casting of this slurry takes place over one or two days, and the density of the green bodies obtained is high (≈57% of the theoretical density) linked to a compact stacking of powder particles resulting from the well-dispersed suspension. This study has shown that compact green bodies can be produced in an organic liquid from a mixture of two silicon nitride based powders with different granulometries. Although the acetophenone is not the most convenient medium for the mixing process of these powders, the acetophenone/PS4 surfactant seems to be a good chemical couple to disperse suspensions. Thus, stable and well-dispersed slurries containing nanometric powders are obtained using an ultrasonic probe to break down hard agglomerates of nanopowders. Slip casting of such slurries leads to well-compacted green bodies associated with the homogeneous and dispersed state of the suspensions. 7.4.2

Sintering of micro–nanocomposites

Powders (Si3 N4 , (UBE SN E10), Si/C/N) and additives (Y2 O3 = 6 wt%, Al2 O3 = 3 wt%) are deagglomerated in liquid media. After drying, the green bodies are hot-pressed (35 MPa) in the presence of N (1 atm) (Table 7.1). The theoretical density of the composites is calculated from the density of each constituent (Si3 N4 , SiC, glass-phase) and by using the rule of mixtures.

©2003 Taylor & Francis

60

Viscosity  (mPa s)

55

50

45

40

35 2

4 6 8 Dispersant content (wt %)

10

Figure 7.5 Viscosity as a function of dispersant content.

Table 7.1 References of micro–nanocomposites Compounds (wt%) Si3 N4

SiCN29

SiCN35

SiC173

91 71 79.9 68.8 46.6 71 51

— 20 — — — — —

— — 11.1 22.2 44.4 — —

— — — — — 20 40

Theoretical SiC content (wt%)

Grades

0 1.7 5 10 20 20 40

Si3 N4 Si3 N4 /SiCN29 (20) Si3 N4 /SiCN35 (11.1) Si3 N4 /SiCN35 (22.2) Si3 N4 /SiCN35 (44.4) Si3 N4 /SiC (20) Si3 N4 /SiC (40)

Sintering kinetics with a linear increase of temperature The shrinkage curves for the different compositions with the same content of nanopowders of SiC show a sigmoidal shape (Figure 7.6(a)). The degree of densification at 1700◦ C is close to 98%. On this figure is also represented the curve concerning the shrinkage of Si3 N4 . In order to evidence the different behaviours, the derived curves giving the rate of densification are shown in Figure 7.6(b). For monoliths and nanocomposites containing a SiCN second phase, the beginning of the shrinkage corresponds to the formation of the oxynitrided eutectic in the YSiAlON system at about 1350◦ C. The content, the composition and the properties, particularly the viscosity of the liquid phase change strongly with temperature. Up to 1470◦ C with a high viscosity, the rate of densification corresponding to a rearrangement process and a dissolution of small particles is low. The shift to high temperatures for the Si3 N4 /SiC

©2003 Taylor & Francis

(a)

100

Relative density (%)

90

80

70

60

50

(b)

1200

1300

1200

1300

1400 1500 Temperature (°C)

1600

1700

1600

1700

0.35

Rate of densification (˚C–1)

0.3 0.25 0.2 0.15 0.1 0.05 0 1400

1500

Temperature (°C)

Figure 7.6 Isobar densification curves. (a) Relative density, (b) rate of densification.  Si3 N4 /SiCN29(20), × Si3 N4 /SiCN35(22, 2),  Si3 N4 /SiC(20).

'

Si3 N4 ,

mixtures is explained by the SiC nanopowders. In this case, free C can react during the heat treatment with oxygen at the surface of Si3 N4 grains and/or the additives. This reaction delays the formation of the liquid phase and modifies its quantity or composition and therefore the viscosity (Hampshire et al. 1978). Furthermore, SiC particles and agglomerates being insoluble in the liquid phase, the dissolution stage is suppressed. At 1470◦ C, the sharp increase in the rate of densification is explained by the dissolution of Si3 N4 and SiCN particles which initiates the α → β phase transformation. These results show that at the beginning, the rate of densification for the monolith is higher than for Si3 N4 /SiCN composites and it is the

©2003 Taylor & Francis

opposite situation above 1550◦ C. The contributions of the rearrangement and dissolution processes to the overall densification change with the size of the crystals and the thickness of the intergranular vitreous phase (see equations 7.1 and 7.2). For nanopowders, at the same liquid phase content the large specific surface area and the presence of agglomerates induce a decrease of the thickness of grain boundaries and the flow of the species which explains the lower rate of densification at the beginning of sintering for Si3 N4 /SiCN composites compared to Si3 N4 monoliths. At higher temperatures (T > 1550◦ C), the increase for composites is explained by a faster dissolution of SiCN nanoparticles (see equation 7.2). One notices that the temperature corresponding to the end of densification depends on the SiC content in the final material. The rate of densification during the dissolution– reprecipitation stage is governed by the thickness of grain boundaries (see equation 7.1). The thickness diminishes when the content of intergranular SiC particles increases. The flow of mobile species is impeded which explains the shift to higher temperatures. The same phenomenon is observed for Si3 N4 /SiC composites. Moreover, the starting SiC powder contains an excess of free C which reacts with oxygen from silica or amorphous silicon oxynitride located at the surface of silicon nitride grains or from additives and modifies the viscosity and the liquid phase content. These experiments show clearly that the source of SiC nanoparticles plays an important role on the dissolution–reprecipitation process and consequently on the rate of densification. The SiCN nanoparticles are more soluble in the oxynitrided liquid phase than SiC nanoparticles that are practically inert in this temperature range (Kleebe 1992).

Densification kinetics during isothermal treatments The densification curves (Figure 7.7(a)) for the various grades exhibit the same shape with a monotonous decrease of the rate of densification with time (Figure 7.7(b)). The differences are mainly due to the composition and the content of the second nanometric phase. These results are consistent with the previous experiments. At 1530◦ C, the oxynitrided eutectic is formed which explains the high rate of densification for the rearrangement stage. The rate increases with the content of SiCN nanoparticles. The dissolution of these particles starts at the beginning of the rearrangement when the pressure is applied. The overall rate is faster than for Si3 N4 alone. On the contrary, for Si3 N4 /SiC composites the shrinkage is low due to the inertness of SiC nanoparticles towards the eutectic. Moreover, SiC precipitates hinder the diffusion of species in grain boundaries which limits the densification whatever the starting nanopowder used. The densification for composites is uncompleted after 2 h at 1530◦ C under a pressure of 35 MPa.

Composites microstructure It is known that it is possible to modify or adjust the microstructure of ceramic materials by introducing an inert second phase but little work is published on the Si3 N4 /SiC micro– nanosystem. Due to the fact that a lot of parameters influence the microstructure of silicon nitride, it is necessary to study carefully all the phenomena. First, the α → β phase transformation for Si3 N4 and the crystallisation of β-Si3 N4 from SiCN nanopowders are analysed, then the behaviour of mixtures is studied in order to determine the interactions and to localise SiC nanoprecipitates.

©2003 Taylor & Francis

100

(a)

Relative density (%)

90

80

70

60

50

40

(b)

0

20

40

60 80 Temperature (°C)

100

120

0.03

Rate of densification (˚C–1)

0.025

0.02 0.015

0.01 0.005 0

0

2

4

6 Time (min)

8

10

Figure 7.7 Isothermal densification curves. (a) Relative density, (b) rate of densification. ' Si3 N4 ,  Si3 N4 /SiCN29(20), × Si3 N4 /SiCN35(22, 2),  Si3 N4 /SiC(20). SEM OBSERVATIONS

After sintering, the samples are polished and plasma etched (CF4 /O2 ). Then, they are analysed in a plane perpendicular to the hot-pressing direction. Si3 N4 /SiCN29–Si3 N4 /SiCN35 COMPOSITES

A fine microstructure composed of equiaxed grains, with a mean diameter close to 0.2 µm is observed when samples are heat treated at 1550◦ C. Then the aspect ratio increases with

©2003 Taylor & Francis

(a)

(b)

3 µm

3 µm

Figure 7.8 Si3 N4 /SiCN29(20) micro–nanocomposite sintered (a) 1550◦ C, (b) 1650◦ C.

3 µm

Figure 7.9 Si3 N4 /SiC35(22.2) micro–nanocomposite sintered at 1650◦ C.

the sintering temperature (Figure 7.8). The comparison of Si3 N4 /SiCN29 and Si3 N4 /SiCN35 microstructures (Figure 7.9) indicates that the growth of β-Si3 N4 crystals is hindered by higher SiC contents, but the agglomerates of nanograins are present. For these two grades the grain size is smaller than in the case of monolithic Si3 N4 densified with the same experimental conditions and with the same additives (Figure 7.10).

Si3 N4 /SiC173 COMPOSITES

For these composites the influence of SiC nanograins on the development of the microstructure and the growth of β-Si3 N4 crystals is detected, but with these nanopowders large agglomerates

©2003 Taylor & Francis

3 µm

Figure 7.10 Monolithic Si3 N4 sintered at 1650◦ C.

1 µm

Figure 7.11 Si3 N4 /SiC micro–nanocomposite sintered at 1650◦ C.

are revealed by a chemical attack (Figure 7.11). The presence of agglomerates is mainly due to the insolubility of SiC in the eutectic phase in the temperature range used for sintering. PHASE TRANSFORMATIONS

The different powders have been heat treated according to the following conditions: T: treated at 1530◦ C, 2 h, under N2 , 1 atm; TP: T + 35 MPa pressure; TPA: TP + additives(Y2 O3 6 wt%, Al2 O3 wt%). The α-, β-Si3 N4 contents were determined by XRD experiments. The results are summarised in Table 7.2. The values for the α/β ratios after T and TP heat treatments are nearly equivalent and without additives no phase transformation is observed. On the contrary, when Y2 O3 and Al2 O3

©2003 Taylor & Francis

Table 7.2 Influence of heat-treatment conditions on the densification and the rate of α → β transformation. T: treated at 1530◦ C/ 2 h/N2 ; TP: treated at 1530◦ C/2 h/N2 /35 MPa; TPA: treated at 1530◦ C/2 h/N2 /35 MPa/(+6%Y2 O3 + 4%Al2 O3 )

Si3 N4 UBE SiCN29 SiCN35

Starting α/β (wt%)

α/β ratio after heat treatment (wt%)

T

TP

TPA

T

TP

TPA

100/0 49/51 79/21

100/0 49/51 78/22

70/30 0/100 20/80

0 0 —

0 0 —

30 100 74

Relative density (%) TPA / 90 63

1.8 1.6

Ratio (β/)

1.4 1.2 1 0.8 0.6 0.4

1560

1580 1600 1620 Temperature (°C)

1640

Figure 7.12 β-Si3 N4 content v. temperature. ' Si3 N4 ,  Si3 N4 /SiCN29(20), × Si3 N4 /SiCN35(22, 2),  Si3 N4 /SiC(20).

are added, crystals or SiCN particles are dissolved in the YSiAlONC eutectic and β-Si3 N4 reprecipitates. The rate of the α-β transformation is higher for SiCN nanopowders than for Si3 N4 . Moreover, the densification decreases when SiC content is raised. In the case of powder mixtures (Si3 N4 , SiCN, additives), samples are hot-pressed under N at different temperatures during 2 h. The ratios α + β are compared with previous results corresponding to the individual constituents (Figure 7.12 and Table 7.3). The α → β transformation being thermally activated, the β-Si3 N4 content increases with the sintering temperature for all the materials. When the same heat treatment conditions are used, the curves on Figure 7.13 show an increase of the α → β transformation rate with SiC content. For sintering temperatures lower than 1600◦ C and whatever the SiCN nanopowder content the nanocomposites contain more β-Si3 N4 than the monoliths. Above 1600◦ C, the β-Si3 N4 content is always higher in nanocomposites than for monoliths but the gap is less pronounced.

©2003 Taylor & Francis

Table 7.3 α-Si3 N4 content after sintering Materials

Sintering temperatures (◦ C)

Monolith Si3 N4

1550 1600 1650 1550 1600 1650 1550 1600 1650 1550 1600 1650 1550 1600 1650 1550 1600 1650 1550 1600 1650

Si3 N4 /SiCN29(20)

Si3 N4 /SiCN35(11.1)

Si3 N4 /SiCN35(22.2)

Si3 N4 /SiCN35(44.4)

Si3 N4 /SiC(20)

Si3 N4 /SiC(40)

Experimental results (%) α/α + β 70 30 15 43 27 12 52 22 10 42 24 7 70 15 0 68 15 0 87 53 0

Theoretical ratio α/α + β

57 24 11 67 28 14 64 26 13 55 21 11 70 30 15 70 30 15

1.7

Ratio (β/)

1.65 1.6 1.55 1.5 1.45 0

5

10 15 SiC content (wt%)

20

25

Figure 7.13 β-Si3 N4 /τ v. SiC content (T = 1650◦ C).

These changes can be explained by the more important contribution to the phase transformation by nanopowders at low temperatures compared to that of microscale Si3 N4 . A very interesting result is obtained in the case of samples containing SiC nanopowders which are practically insoluble in the eutectic. An increase of the rate of the

©2003 Taylor & Francis

α → β-Si3 N4 transformation is clearly obtained and the transformation is completed at 1650◦ C. Conclusion The overall experiments show that the morphology, the composition and the content of the second phase play an important role on the α → β-Si3 N4 transformation. The transformation rate changes with the SiC content in the SiCN starting powder. This effect is more pronounced with SiC nanopowders. Similar observations are mentioned in the literature, nevertheless, they are fragmentary and much debated. Sasaki et al. (1992) suggest that SiC nanoparticles act on the precipitation process as heterogeneous nuclei. However, this explanation is questioned by Pan et al. (1996) who show that for Si3 N4 /SiC nanocomposites SiC nanoprecipitates are not nuclei but only included by β-Si3 N4 during the grain growth. The present work which concerns mixtures with SiCN or SiC shows that the nanopowders influence the germination and reprecipitation stage. Si3 N4 nuclei are formed at the surface of SiC nanoparticles which lower the energetic barrier. It corresponds to a heterogeneous nucleation induced by SiC nanoparticles. Moreover, it has been shown (Kleebe 1992) that SiC is weakly soluble at very high temperature in the presence of YSiAlONC liquid phase. This very limited solubility is in favour of the reprecipitation of β-Si3 N4 or β-SiAlON on SiC particles. The influence of SiC is more effective when its content increases from 10% to 40%, but the number of nuclei is too high for the transformation, and the main part is localised at grain boundaries. 7.4.3

Sintering of nanocomposites

The study of the rheology of these nanopowders in different organic liquids and the nature of dispersing agents is in progress. Different grades of nanopowders, synthetised by laser pyrolysis are mixed with oxides (Y2 O3 , ∅ ∼ = = 1 µm, 99.9%, Rhône-Poulenc; Al2 O3 , ∅ ∼ 50 nm, 99.9%, Marketech International) in ethanol. The starting compositions and their references are given in Table 7.4. The powders are mixed and deagglomerated by milling in a polyethylene jar with Si3 N4 balls. The mixtures are dried at 200◦ C and hot-pressed (35 MPa) in a atmosphere at various Table 7.4 References and starting compositions Grade

Starting composition (wt%) Nanopowder

HSAl09 HMDS35 + 3A SiCN35 + 6Y3A SiCN29 + 6Y3A HMDS35 + 6Y3A HSAl07 + 6Y HSAl15 + 6Y HSAl09 + 6Y SiCN35 + 9Y4.5A SiCN29 + 9Y4.5A HSAl09 + 6Y3A

©2003 Taylor & Francis

HSAl09 HMDS35 SiCN35 SiCN29 HMDS35 HSAl07 HSAl15 HSAl09 SiCN35 SiCN29 HSAl09

100 97 91 91 91 94 94 94 86.5 86.5 91

Theoretical composition (wt%)

Y2 O3

Al 2 O3

SiCNO phase

Y2 O3

Al 2 O3

— — 6 6 6 6 6 6 9 9 6

— 3 3 3 3 — — — 4.5 4.5 3

97 97 91 91 91 90.2 90.5 91.2 86.5 86.5 88.3

— — 6 6 6 6 6 6 9 9 6

3 3 3 3 3 3.8 3.5 2.8 4.5 4.5 5.7

temperatures. After sintering, the different phases are determined by X-ray diffraction and the chemical composition is obtained by elemental analysis. The rate of densification is calculated by using the rule of mixtures and the theoretical density of the crystalline phases. Up to now, the number of papers devoted to sintering of silicon carbonitrided powders is very limited compared to papers on the synthesis of nanopowders. In particular, the route of sintering aids addition, in situ or by mixing powders may influence the development of the microstructure. The thermal stability studies of different grades of nanopowders have shown the influence of the precursors used and the chemical composition. In order to fabricate dense nanocomposites, it is necessary to study sintering kinetics of various nanopowders and to establish relationships between the characteristics of the powders and the sintering behaviour.

Influence of the synthesis process NANOPOWDERS FROM LIQUID PRECURSORS

The mixtures are densified with a linear increase of temperature or during isothermal treatments. The shrinkage curves on Figure 7.14 show that powders with a high N content (C/N = 0.2) exhibit a better ability for sintering than carbon-rich powders (C/N = 2.5). The presence of additives containing aluminium (HSAl09, HMDS35 + 3A) enhances the densification rate due to a possible oxynitride liquid phase. In the case of a high N content combined with the addition of Al introduced during the synthesis (HSAl09) a positive effect is observed compared to the grade HMDS35+3A where Al is added by a mechanical mixing. The rate of densification is higher for the HSAl09 powder than for HMDS35 + 3A at the same temperature. However, the shrinkage is blocked at around 70% as it is shown by the

60

Relative density (%)

55 50 45 40 35 30 1300

1400

1500

1600

1700

Temperature (°C)

Figure 7.14 Isobar densification curves. Relative density • HMDS35 + 3A,  HMDS66,
HSAl07, × HSAl09.

©2003 Taylor & Francis

100

Relative density (%)

90 80 70 60 50 40 30 0

20

40

60

80

100

120

Time (min)

Figure 7.15 Isobar densification curves. Relative density. • HMDS35 + 6Y3A, ' HSAl09 + 6Y, × HSAl09. Table 7.5 Chemical composition of the mixtures HMDS35 + 6%Y2 O3 + 3%Al2 O3 and HSAl09 + 6%Y2 O3 Grade

91% HMDS35 6% Y2 O3 3% Al2 O3 HMDS35 + 6Y3A 94% HSAl09 6% Y2 O3 HSAl09 + 6Y

Chemical composition (wt%) Si

C

N

O

Al

Y

46.0 — — 46 46.6 — 46.6

5.4 — — 5.4 6.4 — 6.4

35.0 — — 35.0 33.8 — 33.8

4.6 1.3 1.4 7.3 5.7 1.3 7.0

— — 1.6 1.6 1.5 — 1.5

— 4.7 — 4.7 — 4.7 4.7

isothermal experiment carried out at 1650◦ C on the HSAl09 powder (Figure 7.15). A sharp increase of the shrinkage is observed at the beginning of the isotherm, then the rate of densification decreases progressively. This behaviour could be related to the decomposition of the nanopowder (see thermal stability) which hinders the sintering. In order to restrict this decomposition, 6 wt% of Y2 O3 is incorporated in the starting mixture (HMDS35 + 6Y3A, HSAl09 + 6Y) to have the same additive content (Table 7.5). Isobar curves reported in Figure 7.16(a) show clearly that an increase of the additive content favour the wetting of the surface by the liquid phase which stabilises the nanoparticles. This phenomenon explains that the beginning of the shrinkage starts at a lower temperature and the relative density is higher for all the grades. One notices that whatever the additive contents (HSAl, HSAl + 6Y), the rate of densification (Figure 7.16(b)) depends on the percentage of

©2003 Taylor & Francis

(a)

100

Relative density (%)

90 80 70 60 50 40 30 1400

1450

1500

1550

1600

1650

1700

1750

1650

1700

1750

Temperature (°C) (b)

0.5

Rate of densification (°C-1)

0.4

0.3

0.2

0.1

0.0 1400

1450

1500

1550

1600

Temperature (°C)

Figure 7.16 Isobar densification curves. (a) Relative density. (b) rate of densification.
HSAl07+ 6Y, ♦ HSAl09 + 6Y,  HSAl15 + 6Y, • HMDS35 + 6Y3A.

C in the starting nanopowders. When the C/N ratio increases, the rate of densification and the final density are lowered. Shrinkage curves with a sigmoidal shape are obtained only in the case of N rich powders. The comparison of the different powders confirms the beneficial approach of prealloyed powders when Al is introduced in situ during the synthesis (Figure 7.16). The phase contents (α-Si3 N4 , β-Si3 N4 , Si2 N2 O) of the nanocomposites are reported in Table 7.6. The incorporation of additives promotes the α- to β-Si3 N4 transformation but also the precipitation of silicon oxynitride.

©2003 Taylor & Francis

Table 7.6 Crystallised phases Grade

Si2 N2 O

Si3 N4

β/(α + β) (%)

HSAl09 HSAl09 + 6Y HSAl09 + 6Y3A

+ ++ +++

+++ ++ +

30 80 100

100

Relative density (%)

90 80 70 60 50 40 30

1300

1400

1500

1600

1700

Temperature (°C)

Figure 7.17 Isobar densification curves. Relative density.  SiCN35 + 6Y3A, ▽ SiCN35 + 6Y3A,  SiCN29 + 9Y4, 5A,  SiCN29 + 6Y3A. NANOPOWDERS FROM GASEOUS PRECURSORS

Different grades of SiCN powders mixed with the same additives are hot-pressed with a linear increase of temperature. The kinetic curves are reproduced in Figure 7.17. A better compaction at room temperature of these powders is observed, compared with the previous grades; moreover, the relative density changes with the gas used during the synthesis of the powder. This difference could be related to the nature and the quantity of the adsorbed species on the surface of the particles as it has been demonstrated by TGA–MS (see thermal stability). These species influence the behaviour of the nanoparticles during the mixing in ethanol. Carbon is detrimental to the sintering processes whatever the additives percentage. When C/N ratio increases, densification is delayed. The same relative density (80%) is reached at 1650◦ C for SiCN29 + 6Y3A and at 1740◦ C for SiCN35 + 6Y3A. XRD patterns evidence the presence of β-Si3 N4 and an amount of β-SiC which increases with the C/N atomic ratio in the starting powders. Comparative study of micro–nano and nano–nanocomposites In order to compare the sintering behaviour and to determine the experimental conditions to obtain fully dense composites, different micro or nanopowders were used. The mixtures

©2003 Taylor & Francis

(SCN29 + 6Y3A, HMDS35 + 6Y3A, HSAl09 + 6Y, Si3 N4 /SiCN35 + 6Y3A) with the same additive content were heat treated. The C/N ratio equal to 0.2 represents a compromise to obtain a composite with a final composition of Si3 N4 /8 wt% SiC which exhibits a ductile behaviour at high temperature. SiC particles stabilise the microstructure during a transitory stage for hot forming. Then this ratio allows an acicular grain growth of β-Si3 N4 crystals during a post forming treatment to obtain a good toughness. The shrinkage curves evidence the difficulties to have a high compactness with nanopowders (Figure 7.18(a)). Nevertheless, the rate of densification (Figure 7.18(b)) increases rapidly and fully dense composites are obtained at lower temperatures than for the micro–nano mixture. Isothermal curves confirm the ability for sintering of these fine powders. On the contrary, (a)

100

Relative density (%)

90 80 70 60 50 40 1300

1400

1500

1600

1700

Temperature (°C) (b)

0.5

Rate of densification (°C–1)

0.4

0.3

0.2

0.1

0.0 1400

1450

1500

1550 1600 1650 Temperature (°C)

1700

1750

Figure 7.18 Isobar densification curves. (a) Relative density, (b) rate of densification. • HMDS35 + 6Y3A, ♦ HSAl09 + 6Y,  SiCN29 + 6Y3A,  Si3 N4 /Si/C/N + 6Y3A.

©2003 Taylor & Francis

micro–nanocomposites reach 90% of the theoretical density after 2 h at 1550◦ C under 35 MPa (Figure 7.19). The α → β-Si3 N4 transformation ratio was determined by XRD for the starting powders and for the sintered composites. Si3 N4 content in the SiCN powders was calculated using the rule of mixtures and by taking into account the chemical analysis results (Table 7.7). This ratio is equal to 14% for micro–nanocomposites and 100% for the nano–nanocomposites (Isotherm 1550◦ C, 2 h, N2 ). For the two mixtures, the beginning of the shrinkage is detected at 1460◦ C which illustrates the beneficial effect of the size of the powders. For the nano– nanocomposites with a densification higher than 96% after 90 mn 1550◦ C, the shape of the curves changes with the starting powder. The SiCN29 + 6Y3A grade, nanopowders issued from gaseous precursors, exhibits a higher rate of densification than the powders issued from liquid precursors (HMDS35 and HSAl09). 100

Relative density (%)

90 80 70 60 50 40 30 0

20

40

60

80

100

120

Time (min)

Figure 7.19 Isobar densification curves (1650◦ C). Relative density • HMDS35 + 6Y3A, ♦ HSAl09 + 6Y,
SiCN29 + 6Y3A,  Si3 N4 /Si/C/N + 6Y3A. Table 7.7 α → β-Si3 N4 transformation rate α/α + β (wt%) Powders Si3 N4 UBE SiCN29 (87% Si3 N4 ) SiCN35∗ (45% Si3 N4 ) SiCN29 + 6Y 3A nano/nano Mixture Sintered material Si3 N4 /SiCN 35 + 6Y 3A micro/nano Mixture Sintered material

©2003 Taylor & Francis

Transformation rate Si3 N4 –α → β(%)

100 47 78 37 0

100

98 84

14

(a)

(b)

1 µm

80 nm

Figure 7.20 SiCN29 nanocomposite sintered at 1600◦ C. (a) SEM, (b) TEM.

Microstructure Si3 N4 /SiC NANOCOMPOSITES

Observations of SiCN nano–nanocomposites show a uniform microstructure with equiaxed grains. The mean diameter of the grains is about 150 nm and the maximum length of β-Si3 N4 crystals is lower than 450 nm (Figure 7.20). TEM micrographs confirm these observations. Small particles (∅ < 3 nm) with stacking faults correspond to SiC nanoprecipitates which are randomly dispersed within β-Si3 N4 grains or at grain boundaries. The grain size for micro–nanocomposites (Figure 7.21) is approximately twice greater than for nano–nanocomposites. The microstructure refinement and the distribution of SiC precipitates is explained by the size of the starting powders. The dissolution is faster in the case of nanopowders and the supersaturation of the liquid phase is reached rapidly, which induces a more homogeneous nucleation process. Moreover SiC nanoprecipitates formed can act as nuclei. In consequence, the number of precipitate seeds increases and crystal growth is limited.

Si3 N4 /Si2 N2 O NANOCOMPOSITES

In the case of composites issued from HMDS or HSAl powders, the grain size is wider. Equiaxed grains with 200 nm in diameter are observed but also acicular crystals which are distributed all around agglomerates of small grains. The development of this bimodal distribution of grains could be explained from results obtained by Emoto et al. (1998), on the sintering of (Si3 N4 -Mg2 Al4 Si5 O18 ) powders. The precipitation of Si2 N2 O nuclei in the first stage is due to the high oxygen content in the liquid phase. The growth of these crystals is favoured by a rapid migration of the species linked to the low viscosity of the eutectic. This process is followed by compositional changes around Si2 N2 O crystals which induce in a second stage the precipitation of β-Si3 N4 . As a result, fine Si3 N4 crystals with equiaxed morphology are confined by acicular Si2 N2 O grains.

©2003 Taylor & Francis

1 µm

Figure 7.21 Si3 N4 /SiCN29 micro–nanocomposite sintered at 1600◦ C.

Discussion According to previous works (Bellosi 1995) on the sintering of nanopowders, the compaction at room temperature is lower than 40%. This value which characterises fine powders is explained by hard agglomerates with particles bonded by Van der Waals forces but also by the nature of chemical species on the surface of the particles. When a bimodal distribution is used as in the case of the micro–nano system, the compactness is favoured. This difference explains that at the beginning of the shrinkage the rate is higher for nano–nanocomposites. The rearrangement process and the solubility of the nano particles which are inversely proportional to the size are faster. As a result of these phenomena, the α → β-Si3 N4 transformation is completed only for the nano–nanocomposites hot-pressed for 2 h at 1550◦ C. These results emphasise the influence of the process to introduce additives for the sintering of silicon nitride materials. Kinetic curves for prealloyed powders with the same additive content are faster which shows that the distribution of aluminium at an atomic scale facilitates the rearrangement process by a simultaneous wetting of the particles. On the contrary, with the classical route by mixing powders the oxynitride eutectic is localised in a limited number of grain boundaries which necessitates its migration and the impregnation of the porosity. It has been demonstrated that the sintering of silicon nitride is closely linked to the volume of liquid phase. A minimum quantity is necessary to impede its decomposition. The same behaviour is observed with nanopowders. The too low content of aluminium introduced in situ during the synthesis does not allow it to reach the theoretical density. Two processes are competing when temperature increases. The densification starts due to the rearrangement and dissolution–diffusion–reprecipitation mechanisms but also the degradation of SiCNO grains with a release of volatile species. Above 1650◦ C, the decomposition exceeds the rate of the shrinkage. For the behaviour at high temperature of these powders, an important parameter is the C content. The two different grades obtained by pyrolysis of liquid or gaseous precursors contain a high C content. The presence of C delays the shrinkage but the role of C differs according to its nature, free C or bounded to Si. In the case of HMDS and HSAl powders, it is mainly

©2003 Taylor & Francis

free C which impedes the densification. As observed by Doug et al. (1997) and explained in the study of thermal stability, at high temperature an in-situ carboreduction reaction takes place in the powder and carbon monoxide is formed. The release of CO modifies the oxygen content of the liquid phase and induces an increase in its viscosity. In consequence, these changes are followed by a higher temperature for the beginning of the shrinkage and by a lower densification rate. Concerning C bounded to Si its influence on the sintering behaviour of powders issued from gaseous precursors is less important than free C. SiC nanoprecipitates are progressively formed but they are practically inert towards the liquid phase in this range of temperature. The densification of silicon carbide with Al2 O3 and Y2 O3 additives is only efficient above 1750◦ C. This is the reason why SiC nanoparticles formed by the dissolution of SiCNO phase impede the intergranular diffusion of the species and the oversell rate of densification. These studies show also the great reactivity of nanopowders which changes with the nature of the precursors. When powders issued from liquid precursors are hot-pressed, Si2 N2 O is formed during sintering. This oxynitride precipitates in the YSiAlON system only when the eutectic is saturated by oxygen. This element comes from additives but also from reactions with the environment during the mixing process of the powder in a liquid media. Recently, Wada et al. (1999) have demonstrated that the oxygen content of carbonitrided nanopowders increases when a slurry is prepared with ethanol. In the case of the HMDS nanopowders, the amorphous state and the different surface groups could explain the specific reactions with ethanol. The experiments on the thermal stability have evidenced the presence of NH and NH2 groups bounded on the surface of the particles. It is known that these groups are easily hydrolysed or can react with ethanol which induces the formation of Si O Si bonds. The quantity of oxygen fixed at the surface varies with the surface groups which depend on the pyrolysis conditions and the nature of the precursors.

7.5 Conclusion These kinetics studies in the sintering of Si3 N4 /SiC composites from various powders evidence the role of the SiC content on the starting and the development of the microstructure. When Si/C/N nanopowders are added to a Si3 N4 micropowder, they generate an increase in the densification rate due to the faster solubility of nanograins in the liquid phase. The densification kinetics are also closely linked to the grain structure and the chemical composition. To fabricate fully densified composites, it is necessary to use crystallised powders which are stable and more able to be sintered. The shrinkage curves show that better results are obtained with powders issued from gaseous precursors than from liquids. For the different grades of powders the rate of densification decreases with a higher C/N atomic ratio. The incorporation of SiC nanoparticles hinders the shrinkage due to their inertness towards the eutectic. Moreover, free C present at the surface of the grains or as particles in the powders reacts with oxygen from additives which delays the formation of the liquid phase. On the contrary, SiC nanoparticles quicken the α/β-Si3 N4 transformation and form nuclei for β-Si3 N4 grains which induces a heterogeneous nucleation and a fine microstructure. The route to incorporate additives is also an important factor. The distribution of aluminium at an atomic scale favours the migration of the liquid, the wetting of the surface of the grains and the rearrangement process. These experiments validate the interest of prealloyed powders and their ability to fabricate Si3 N4 /SiC nanocomposites.

©2003 Taylor & Francis

References Bellosi, A. (1995) Mater. Sci. Forum, 195, 79. Bowen, L. J., Weston, R. J., Carruthers, T. G. and Brook, R. J. (1978) J. Mater Sci., 13, 350. Cauchetier, M., Croix, O., Luce, M., Baraton, M.-I., Merle, T. and Quintard, P. (1991) J. Eur. Ceram. Soc., 8, 215. Doug, S., Jiang, D., Tan, S. and Guo, J. (1997) J. Mater. Sci. Lett., 16, 1080. Emoto, H., Mitomo, M., Wang, C. M., Hirosturu, H. and Inaha, T. (1998) J. Eur. Ceram. Soc., 18, 527. Gheorghiu, A., Sénémaud, C., Roulet, H., Dufour, G., Moreno, T., Bodeur, S., Reynaud, C., Cauchetier, M. and Luce, M. (1992) J. Appl. Phys., 71 (9), 4118. Greenwood, G. W. (1956) Acta Metall., 4, 241. Greskowich, C., Prochazka, S. and Rosolowski, J. H. (1977) Sintering behaviour of covalently bonded materials, Nitrogen Ceramics (ed. F. L. Riley), Noordhoff, Leyden. Grün, R. (1979) Acta Crystallogr., B35, 800. Hampshire, S., Nestor, E., Flynn, R., Besson, J. L., Rouxel, T., Lemercier, H., Goursat, P., Sebaï, M., Thompson, D. P. and Liddell, K. (1994) J. Eur. Ceram. Soc., 14, 261. Hampshire, S., Park, H. K., Thompson, D. P. and Jack, K. H. (1978) Nature, 274, 880. Ishizaki, K. and Yanai, T. (1995) Silic. Indus., 7–8, 215. Kang, S. J. and Han, S. M. (1995) MRS Bull., 2, 33. Kleebe, H. J. (1992) J. Am. Ceram. Soc., 10, 151. Lang, J. (1977) Nitrogen Ceramics (ed. F. L. Riley), Noordhoff, Leyden. Lee, R. R., Chen, C. J. and Lin, J. T. (1994) Ceram. Trans., 42, 221. Lewis, M. H. (1993) Trans. Tech. Publ., 89–91, 334. Niihara, K. (1991) Ceram. Sci. Jpn, 99, 10, 974. Niihara, K., Suganuma, K., Nakatiera, A. and Izaki, K. (1990) J. Mater. Sci. Lett., 9, 598. Pan, X., Gu, H., Van Weeren, R., Dauforth, S. C., Cannon, R. and Rühle, M. (1996) J. Am. Ceram. Soc., 79, 2313. Rossignol, F., Goursat, P., Besson, J. L. and Lespade P. (1994) J. Eur. Ceram. Soc., 13, 299. Sasaki, G., Nakase, H., Sugaruma, K., Fujita, T. and Niihara, K. (1992) J. Ceram. Soc. Jpn, 100, 536. Siegel, R. W. (1994) Springer Ser. Mater. Sci. Phys. New Mater., 27, 65. Ténégal, F., Flank, A.-M. and Herlin, N. (1996) Phys. Rev. B, 54, 17, 12029. Wada, S., Kondo, Y., Sudo, E. and Maki, Y., (1999) J. Ceram. Soc. Jpn, 107, 611. Wakai, F., Kodama, Y., Sakaguchi, S., Murayama, N., Izaki, K. and Niihara, K. (1990) Nature, 344, 421. Wang, C. M., Emoto, H. and Mitomo, M. (1998) J. Am. Ceram. Soc., 81, 5, 1125. Ziegler, G., Heinrich, J. and Wötting, G. (1987) J. Mater. Sci., 22, 3041.

©2003 Taylor & Francis

8

Mechanical properties Jean-Louis Besson

The composites used to assess the mechanical properties are mainly those fabricated using Si/C/N nanoparticles synthesised from gaseous precursors as the source of SiC (grade SiCN35). They are referred to as SiC micro–nanocomposite when the Si/C/N powder is added to Si3 N4 powder and Si/C/N nanocomposite when the Si/C/N powder is the only source of both Si3 N4 and SiC. In the assessment of the compressive creep behaviour, composites elaborated with nanoparticles coming from the decomposition of hexamethyldisilazane (grade HMDS35) or that of a liquid mixture of hexamethyldisilazane and aluminium isopropoxyde (grade HSAl09) are also briefly considered under the respective references: HMDS and HSAl nanocomposites.

8.1 Room temperature properties 8.1.1

Introduction

Niihara and coworkers (Niihara 1991) have observed a remarkable improvement in strength and toughness by incorporating nanosized particles of SiC in various ceramics matrix (Al2 O3 , MgO, Si3 N4 ). In every case, most of the finest SiC particles (typically less than 200 nm) were predominantly dispersed within the matrix grains and some larger particles at the grain boundaries. From HRTEM investigations, it was concluded that no impurity phases were present at the interfaces between the SiC particles and the matrix grains so that the particles were directly bonded to the matrix. Moreover, the fracture mode was changed from a mixed intergranular–intragranular mode to a completely transgranular mode. Though different toughening and grain-refinement mechanisms were proposed for the different systems, the basic idea was that the intragranular nanosized particles play a major role in the increase in toughness and strength up to high temperatures, and this for two main reasons: first, the development, generated by the large thermal expansion mismatch, of high tensile hoop stresses around the particles within the matrix grains, leading to crack deflection; second, the direct bonding of the SiC particles with the matrix even in the presence of grain boundary phases coming from sintering aids or impurities. In the case of a Si3 N4 matrix (Niihara 1991), the fracture strength increases from 850 MPa to a peak value of 1550 MPa and the fracture toughness from 4.5 to 7.5 MPa m1/2 . The peak values were both obtained for 25 wt% SiC. Similar trends were observed by Niihara’s group (Sawaguchi et al. 1991; Sasaki et al. 1992; Hirano and Niihara 1995) for Si3 N4 /SiC nanocomposites processed by various routes, though the peak position differed and the improvements were less pronounced. The degradation of the properties at high SiC content was claimed not

©2003 Taylor & Francis

to be caused by the dispersed nanoparticles but related to flaws coming from aggregation of SiC particles (Sasaki et al. 1992). In Al2 O3 /SiC nanocomposites, neither Zhao et al. (1993) nor Borsa et al. (1997) or Xu et al. (1997) found an influence of the intragranular SiC nanoparticles on the intrinsic material toughness. In the case of Si3 N4 matrix composites, Pezzoti and Sakai (1994), using a material densified without any sintering agent, also excluded any strengthening effect due to the presence of intergranular SiC particles. Their results are consistent with earlier work by Greskowich and Palm (1980) who found that in Si3 N4 /SiC composites where the sintering aids (BeSiN2 and SiO2 ) dissolve in the β-Si3 N4 lattice, the toughness was independent of the SiC volume fraction up to 23 vol%. Contrary to Niihara and co-workers, most of the other researchers (Rendtel et al. 1995; Sajgalik et al. 1996; Sato et al. 1995; Liu et al. 1997; Hermann et al. 1998) have reported at best no improvement of either toughness or strength and more generally a degradation. As it seems to be now established that the intragranular location of SiC nanoparticles does not enhance the ‘intrinsic’ fracture properties of these so-called ‘nanocomposites’, changes in fracture toughness and fracture strength are to be related to the efficiency with which the ‘classical’ energy dissipating mechanisms (such as crack deflection, crack bridging or closing forces) are acting in the material under study, taking into account its actual microstructure. 8.1.2

Fracture strength

Fracture strength, σr , was evaluated by 3-point bending test using rectangular bars (4 × 3 × 25 mm3 ). The bars were bevelled parallel to their length in order to eliminate edge flaws. The tensile surfaces of the specimens were perpendicular to the hot-pressing direction and polished to a 3-µm diamond finish. The span length was 20 mm and the cross-head speed 0.2 mm/min. σr =

3F x wh2

(8.1.1)

where F is the ultimate load, w the width, h the height and x the distance between the section of rupture and the nearest fulcrum point. Changes in fracture strength with the sintering temperature are shown in Figure 8.1. For a sintering temperature of 1550◦ C, the strength increases with the SiC content in spite of the lower densification. The strength of the monolith increases sharply after sintering at 1600◦ C and remains stable for higher temperatures. The same trend is observed for the 5 wt% SiC micro–nanocomposite whereas in the case of the 10 wt% SiC micro–nanocomposite, the strength increases slowly up to a sintering temperature of 1650◦ C. As the densification of the monolith is complete after hot-pressing at 1550◦ C, the change in strength between 1550◦ C and 1600◦ C cannot be related to a change in porosity, and, as it remains constant for higher sintering temperatures, it cannot be associated with the acicular growth of the β-Si3 N4 grains. The remaining explanation is based on the existence of residual stresses. X-ray diffraction (XRD) studies have shown that after sintering at 1550◦ C, the α-and β-Si3 N4 grains are under compression, which implies that the intergranular glass phase experiences local tensile stresses (Rossignol 1995). These internal stresses decrease as the α- to β-Si3 N4 transformation goes on and stabilises when the α content becomes lower than about 50%. In the case of the 5 wt% SiC micro–nanocomposite, the α content after sintering at 1550◦ C is ∼55% instead of 70% for the monolith and the strength is higher, in spite of a lower

©2003 Taylor & Francis

1000

Flexural strength (MPa)

800

600

400

200

0

1550

1600

1650

1700

Sintering temperature (°C)

Figure 8.1 Fracture strength v. sintering temperature:
monolith;  5% SiC micro–nanocomposite; • 10% SiC micro–nanocomposite;
Si/C/N nanocomposite as-sintered;  annealed 5 h at 1600◦ C.

density, due to the lower residual tensile stress in the glass phase. The increase in strength when the sintering temperature is raised to 1600◦ C can be attributed partly to an increase in densification (from 97.5 to 99% TD) and partly to the progress in the α- to β-Si3 N4 transformation. As the α content in the 10 wt% SiC micro–nanocomposite is already lower than 50% after sintering at 1550◦ C, the increase in strength is solely due to the increase in densification from 83 to 99%. In the case of the Si/C/N nanocomposite, the strength is rather low (470 ± 30 MPa) for the as-sintered material (1600◦ C, 1 h), but it increases to 570 ± 30 MPa after a 5 h heat treatment at the sintering temperature and becomes similar to that of the 10 wt% SiC micro–nanocomposite. 8.1.3

Toughness

Fracture toughness, K1C was estimated by microhardness indentation method in a Vickers test using Evans and Charles’ formula (Evans and Charles 1976). K1C = 7.4 × 10−8

P c3/2

(8.1.2)

where P is the indentation load (N) and c the half median crack length (m) and K1C is expressed in MPa m1/2 . The toughness of the monolith increases twofold in the sintering temperature range whereas the increase in toughness of the micro–nanocomposites remains moderate

©2003 Taylor & Francis

7 6

Toughness (MPa m1/2)

5 4 3 2 1 0

1550

1600 1650 Sintering temperature (˚C)

1700

Figure 8.2 Toughness v. sintering temperature:
monolith;  5% SiC micro–nanocomposite; • 10% SiC micro–nanocomposite;
Si/C/N nanocomposite as-sintered.

(Figure 8.2). The toughness of the as-sintered Si/C/N nanocomposite is similar to that of the micro–nanocomposite though a higher fraction of the SiC precipitates is found inside the Si3 N4 grains. When similar microstructures are considered (full density and equiaxed Si3 N4 grains, i.e. sintering temperatures of 1550◦ C, 1600◦ C and 1650◦ C for the monolith, the 5 wt% SiC and the 10 wt% SiC micro–nanocomposites, respectively) the toughness is similar (Figure 8.3). However, if the comparison is made for the minimum sintering temperature leading to full density whatever the material (1600◦ C), the toughness of the monolith is much higher than those of the micro–nanocomposites that enlightens the dominating role of the aspect ratio of the Si3 N4 grains. The toughness of the monolith is enhanced by the development of elongated β-Si3 N4 (the aspect ratio rises from 3.5 to 6) as usually observed for self-reinforced silicon nitride. In contrast, the presence of a large amount of intergranular SiC nanoprecipitates, though accelerating the α- to β-Si3 N4 transformation, inhibits the acicular growth of the β-Si3 N4 grains that remain quite equiaxed in shape and this saturation in aspect ratio prevents an increase in fracture toughness. In the case of the Si/C/N nanocomposite, post-sintering heat treatments at temperatures superior to 1650◦ C, lead to a more and more acicular shape for the β-Si3 N4 grains and the toughness increases from 3.6 ± 0.3 to 5.4 ± 0.4 MPa m1/2 (Table 8.1). An additional increase in toughness is obtained after a further heat treatment at 1400◦ C leading to a partial crystallisation of the amorphous intergranular phase. So, it is confirmed that the main factor is the aspect ratio of the β-Si3 N4 grains and that the intragranular SiC particles play a minor role, if any.

©2003 Taylor & Francis

7

Toughness (MPa m1/2)

6 5 4 3 2 1 0

0

5

10

15

20

SiC (wt%)

Figure 8.3 Fracture toughness at room temperature: - - -
- - - for similar microstructures.

—
—after sintering at 1600◦ C;

Table 8.1 Room temperature toughness of the SiC nanocomposite after various post-sintering heat treatments Toughness (MPa m1/2 )

Heat treatment Temperature (◦ C)

Time (h)

As-sintered 1600 1650 1700 1750 1750 + 1400

5 5 5 5 5+3

3.6 ± 0.3 3.5 ± 0.1 3.8 ± 0.5 4.6 ± 0.4 5.4 ± 0.4 6.9 ± 0.7

8.2 High-temperature properties 8.2.1

Introduction

High-temperature properties of silicon nitride based ceramics, though influenced by the size and morphology of the Si3 N4 grains, are primarily controlled by the nature and chemistry of the intergranular phase. After sintering, the intergranular phase is amorphous but can be converted more or less completely into crystalline phases by a post-sintering heat treatment. However, there always remain thin intergranular films. The most profound influence on the properties comes from the viscosity of these grain boundary films. For instance, MgO additive

©2003 Taylor & Francis

leads to a low viscosity intergranular phase whereas Y2 O3 converts easily to N-melilite (Y2 Si3 N4 O3 ). The viscosity of the residual glass phase must be high enough to prevent cavitation and grain separation, but low enough to accommodate grain boundary sliding through diffusion of matter. The rheology of the intergranular phase can also be modified by a small amount of impurity such as fluorine, which has a detrimental effect on the hightemperature strength (Tanaka et al. 1994; Rendtel et al. 1995). Then, in order to assess the actual role of SiC nanoparticles on the high-temperature behaviour, it is necessary to compare the behaviour of the composites with that of their monolithic counterpart densified with the same additives. Moreover, the choice of the densification aids will be different depending on the high-temperature properties which are desired, high strength and high creep resistance or high ductility and even superplastic behaviour: in the former case, yttria alone will be used whereas it will be a mixture of yttria and alumina in the latter case. So, Niihara and co-workers (1990) have obtained an improvement in the flexural strength by incorporating 32 wt% SiC nanoparticles in a silicon nitride densified with 8 wt% yttria. The 4-point bending strength was still 840 MPa at 1500◦ C whereas that of the corresponding monolith had decreased to 500 MPa at 1400◦ C. This improvement was attributed to the fact that the SiC nanosized particles in the grain boundaries were directly bonded to the Si3 N4 adjacent grains, without any impurity phase. The same grade of composite was also tested in tensile creep and it was noted that, contrary to the monolith that suffered severe cavitation induced by grain boundary sliding, cavity formation was rarely observed in the composite (Hirano et al. 1996). From the study of the bending creep of silicon nitride doped by nanosized SiC particles from different origins and sintered with Y2 O3 aid only, Rendtel et al. (1995, 1998) also concluded that the addition of SiC nanoparticles is a very effective means of improving creep resistance of silicon nitride ceramics. On the contrary, if it is a high ductility that is desired, Si3 N4 /SiC nanocomposites densified with Y2 O3 or MgO are not suitable as they fracture in a brittle manner with less than 2% elongation whereas the grades sintered with Y2 O3 plus Al2 O3 exhibit a high ductility which can reach more than 100% elongation (Rouxel et al. 1992). 8.2.2

Tensile ductility

For the present materials, the high-temperature tensile behaviour was studied in the 1595–1640◦ C temperature range under nitrogen atmosphere. Dog-bone shape specimens with 10 mm long and 3 mm diameter cylindrical gauges were used. The gauges were polished finally with 0.3 µm diamond paste. Since large elongations were involved, true strain, εt and true stress, σt , were calculated on the basis of a constant volume throughout the deformation: εt = ln(1 + ε)

σt = σ (1 + ε)

(8.2.1) (8.2.2)

where ε is the nominal strain (or elongation), !l/l0 , with l0 the initial gauge length and σ the nominal stress. The experiments were conducted at constant strain rate by monitoring the cross-head speed, ε˙ , from the measurement of the elongation, ε: ε˙ t = ε˙

1 1+ε

©2003 Taylor & Francis

(8.2.3)

Table 8.2 Theoretical SiC content, sintering temperature (Ts ), mean grain size, aspect ratio of β-Si3 N4 grains and α/β-Si3 N4 ratio SiC (wt%)

Material

Ts (◦ C) Mean grain size (nm) Si3 N4

Monolith Micro–nanocomposite Si/C/N nanocomposite

1550 1600 1700 1600 1650 1600

10 20 8

250–300 350a 400a 400 100–400a 150–450

SiC

α/β-Si3 N4

30–40 30–40 x0

where x0 is the critical concentration, equal to x0 = ZB /(ZA + ZB ) In the case of a-Si1−x Cx , a currently used order parameter is τ = NCC /NSiC . One can see that for a fully disordered system τ = (ZC /ZSi )x/(1 − x), which is in general smaller than x/(1 − x), since Si has a larger average coordination number than C. For a perfectly ordered systems, τ = 0 (x < x0 ); (ZC /ZSi )x/(1 − x) − 1(x > x0 ). Apart from the fact that τ is ill defined for very high C concentration (x → 1), one can see that τ does not purely depend on the C concentration x, but also on the ratio of the average coordination numbers of C

©2003 Taylor & Francis

3.0



2.0

Full disorder, ZC = 3.5, ZSi = 4.2 Full disorder, ZC = 4, ZSi = 4 Complete order, ZC = 3.5, ZSi = 4.2 Complete order, ZC = 4, ZSi = 4

1.0

0.0 0.0

0.2

0.4 X

0.6

0.8

Figure 9.2 Order parameter τ = NCC /NCSi computed for two virtual Si1−x Cx systems, having all the C and Si coordination numbers equal to ZC = 3.5 and ZSi = 4.2, or ZC = 4.0 and ZSi = 4.0, respectively. Both cases are representative of real amorphous samples. The cases of full disorder and perfect chemical order are plotted. Note that the curve corresponding to the disordered alloy for ZC < ZSi may be close to that of the completely ordered Si1−x Cx system with both C and Si atoms fourfold coordinated.

and Si. This is not a purely academic question, since an appreciable amount of graphitic-like bonding implies ZC < ZSi . Even when other order parameters, which are free from the singularities at x = 0 or x = 1, are adopted, such as ζ = NCC /(NCC + 2NSiC + NSiSi ), the degree of order is invariably a function of both C concentration and C average coordination number, compared to that of Si. Also the critical value of the C concentration (x0 ), at which some homonuclear bonds appear even in the case of perfect chemical order, is a function of the average C and Si coordination numbers (Figure 9.2). The previous discussion is noteworthy, since many experimental data were discussed on the basis of a simplified model for the chemical order, in which ZC = ZSi = 4, even if sp2 C bonding was actually detected in the amorphous sample. Moreover, in the case of strong departure from the chemical order, the crystalline, tetrahedral phase of SiC cannot be used as a starting model for the microscopic structure of amorphous Si C alloys. Therefore, in the following, we prefer to discuss the observed trends for the chemical order, keeping in mind that the precise values of the order parameter cannot be always considered as a definitive measure of the degree of chemical order in the absence of an independent estimate of both C and Si average coordination numbers. 9.2.2

Amorphous silicon carbide

Kelires (1991) carried out a Monte Carlo simulation of amorphous SiC, by adopting an effective interatomic potential energy introduced by Tersoff (1989). The model potential

©2003 Taylor & Francis

for SiC is derived from a two- and three-body interaction originally designed for elemental semiconductors, and then modified in order to take into account the chemical specificity of C and Si through environment-dependent parameters. Kelires found a moderate trend towards chemical order (τ = 0.5 at C/Si = 1). A nonnegligible fraction of C atoms are threefold coordinated. In the following first-principles molecular dynamics simulation, in which the interatomic forces are derived on quantum mechanical grounds, Finocchi et al. (1992, 1993) obtained α-SiC at stoichiometric composition and density close to that of the β crystalline phase, through a quench from the melt. The final sample, at room temperature, has a complex structure which can be classified neither as chemically ordered nor as completely random. Although the Si Si, Si C and C C equilibrium bond lengths (characterised by the first peaks in the pair correlation functions) are very close to those in the corresponding crystals, they are statistically distributed over a different range of distance, depending on the specific bonding. For instance, the width of first peaks of the C C, Si C and Si Si pair correlation functions increase in the order, showing a greater disorder on the Si sub-lattice than on that of carbon (C). The resulting total radial distribution function, being a sum of the various contributions, may be difficult to be interpreted. As a consequence, only the access to the partial correlation functions by means of complementary and species-sensitive techniques, allows one to disentangle the specific contributions. Although the simulated system may strongly differ from the real amorphous samples, we think that such an ambiguity is characteristic of semi-conducting alloys with (i) a large degree of disorder and (ii) competition between different type of bonding (e.g. sp2 /sp3 for C). In a following paper, Tersoff (1994) focused on the influence of the enthalpy of formation of Si C bonds from Si Si and C C bonds (!HSi C ), on the degree of chemical order. In the crystal, !HSi C is equal to −16 kJ/mol/bond, according to the experimental data at normal conditions. Tersoff performed Monte Carlo simulations of amorphous Si1−x Cx , using different interatomic potentials, which correspond to distinct !HSi C . In that way, he studied the order parameter τ as a function of !HSi C . He found that both !HSi C and atomic size difference between Si and C drive the system towards the chemical order. However, the steric factor is much more effective at driving ordering in the crystal than in the amorphous alloy, since in the latter the microscopic stress resulting from the formation of homonuclear bonds can be more efficiently released than in the long-range ordered phase. All these arguments are valid at not too high temperatures, when the entropy contribution, which usually favours disorder, can be neglected. Further ab initio molecular dynamics simulations (Finocchi and Galli 1994) show that the inclusion of hydrogen in the α-SiC network, at small concentrations, favours tetrahedral bonded carbons. Since a trend of sp2 -like carbons to segregate has been observed in the ab initio simulations – as a result of the strong C C bonding – the authors concluded that the inclusion of H promotes chemical order, towards values of the order parameter (τ = 0.5) consistent with most experimental data (Figure 9.3). A recent molecular dynamics simulation (Mura et al. 1998) studied the influence of the quenching rate and the C/Si ratio on the final chemical order. The use of a modified version of the Tersoff effective potentials (instead of the more accurate, but time-consuming firstprinciple-based interatomic potential energy) allows the authors to carry out much longer simulations. Apart from quantitative differences from the results of the ab initio simulations, they also found that the Si sub-lattice can be easily distorted, a fact which attenuates the preference for a chemical order, which is far from being complete in the network. Some of the previous results can be discussed on the basis of simple arguments based upon the thermodynamic data, which, in the case of silicon carbide, drive the metastable

©2003 Taylor & Francis

2.0

Si9Si

0.0 6.0 4.0

C9C

2.0 0.0 8.0 6.0 C9Si

4.0 2.0 0.0 0.0

1.0

2.0 R (A)

3.0

4.0

Figure 9.3 Pair distribution functions for α-SiC : H, as computed by Finocchi and Galli (1994). They are defined as gij (r) = V /(Ni Nj )nij (r)/(4π r 2 !r) where nij denotes the timeaveraged number of atomic species j in a shell between r − !r/2 and r + !r/2 centred on a type i atom. V is the volume of the simulation box and Ni the number of atoms of type i inside V . From the gij , the total pair distribution function is obtained   as: G(r) = ij Ni Nj gij (r)/( i Ni )2 . For the equations connecting gij (r) with the X-ray, neutron diffracted intensities (Finocchi and Galli 1994; Waseda 1980).

amorphous phases towards the chemical order. However, three factors conspire to maximise disorder: i ii

iii

9.2.3

the entropy contribution, which is especially important at high temperatures; the presence of a remarkable disorder on the Si sub-lattice; indeed, the absolute value of the energy of formation of two stretched Si C bonds from a C C and a stretched Si Si bond, may be reduced in this case; the presence of sp2 -like carbon, which reinforces C C bonding, lowers the average C coordination number, and distorts the tetrahedral network. Although the energetic terms should prevail for small C concentrations, and drive the α-SiC towards chemical order, the C-rich systems might exhibit a complex behaviour, which depends on the macroscopic thermodynamic parameters (density, temperature, H contents, etc.) as well as on the preparation conditions and on the history of the sample. Amorphous silicon nitride

The enthalpy of formation of crystalline silicon nitride from Si and molecular nitrogen is equal to −828 kJ/mole (O’Hara et al. 1997). The degree of chemical order expected for amorphous

©2003 Taylor & Francis

Si3 N4 and SiC on the basis of purely energetic arguments can thus be compared by considering the enthalpy of formation of one heteronuclear Si N(Si C) bond, obtained by normalising the enthalpy of formation of the crystalline compounds from the respective elemental phases, by the number of heteronuclear bonds. This amounts to −0.70 eV per Si N bond, and to −0.16 eV per Si C bond. The absolute values of the two enthalpies correspond to very different temperatures, of about 8000 and 2000 K, respectively. This crude estimation shows that, according to thermodynamics, a-Si3 N4 should be almost perfectly chemically ordered. Moreover, the segregation of a Si-rich phase might be accompanied by the formation of highly diffusing N dimers, resulting in a continuous N loss. In agreement with this framework, all the simulations so far performed on a-Si3 N4 indicate a high degree of chemical order, with small amounts of homonuclear bonds. Umesaki et al. (1992) carried out molecular dynamic runs on amorphous Si3 N4 , using a simple Busing-type potential, which consists of a point charge Coulomb interaction, plus a short-range repulsive term. The parameters were fitted to reproduce the α and β crystalline phase of silicon nitride. The density was fixed at 2.6 g/cm3 , and 420 or 1260 particles were used in the simulation. The structure of the simulated samples was characterised in terms of the partial pair distribution functions. The computed Si N mean bond length is equal to 1.74 Å, and the Si N first coordination number is slightly lower than 4 (3.95). The second neighbour distances (the corresponding coordination numbers are given in brackets) are 2.65 Å (8.3) and 3.05 Å (7.1), for N N and Si Si, respectively. All these numbers are in very good agreement with the X-ray and neutron diffraction experiments. The authors conclude that the short-range order of amorphous Si3 N4 is very similar to that of the crystals. However, the presence of some dangling bonds, and the distortions of the dihedral angles between the SiN4 tetrahedra result in the loss of the medium-range order with respect to the crystal. Recently, de Brito Mota et al. (1998, 1999) simulated a-SiNx systems (for 0 ≤ x ≤ 1.5) through Monte Carlo runs in the constant-temperature constant-pressure (NPT) ensemble. They used interatomic potentials of the Tersoff type, which were fitted to crystalline silicon nitride, as well as to molecules containing Si, N and H. The inclusion of N in the a-Si network favours denser metastable phases, since the equilibrium density grows from 2.31 g/cm3 at x = 0 to about 3.5 g/cm3 at x = 1.5. This trend compares well with the available experimental results (Guraya et al. 1990). It is a consequence of the formation of Si N bonds, at the expense of the longer Si Si bonds, as N is incorporated in the amorphous sample. Correlated with such a behaviour, the angular disorder around Si sites grows as a function of the N concentration x. a-SiNx can be considered as well chemically ordered, although a non-negligible number of Si Si bonds are detected (τ = 0.15 for x = 1.33, which corresponds to the stoichiometric composition), in accord with the interpretation of the X-ray photoemission spectra for SiNx powders (Sénémaud et al. 1993). The inclusion of hydrogen reduces the number of dangling bonds, releases the microscopic stress and improves the conductivity properties. However, no major structural changes are recorded in the Si Si and the Si N pair distribution functions, between the hydrogenated and the unhydrogenated simulated systems. They are shown in Figure 9.4, for SiN1.33 . Given the relative success of the empirical potentials in describing a-Si3 N4 , Vashista and collaborators (Vashista et al. 1995; Kalia et al. 1997) performed large-scale molecular dynamics simulations on disordered silicon nitride phases, adopting a slightly more complicated interatomic potential which, in addition to the Coulomb charge–charge interaction, and to the short-range repulsion, contains a charge–dipole interaction. In a first simulation (Vashista et al. 1995), they prepared the amorphous samples (consisting of 1344 atoms) by quench from the melt, at different macroscopic densities. As the density decreases, the

©2003 Taylor & Francis

x = 1.5

x = 1.33

x = 1.0

x = 1.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 R (Å)

Figure 9.4 Total pair correlation functions for amorphous SiNx . Copyright (1998) by the American Physical Society for x = 0.5, 1.0, 1.33 and 1.5. The experimental results for a-SiN1.33 (Guraya et al. 1990) are shown in the inset for comparison. (Reprinted with permission from de Brito Mota, F., Justo, J. F. and Fazzio, A. (1998) Phys. Rev., B58, 8323–8328.)

network connectivity is reduced and floppy modes appear in the vibrational spectra. The consequences for the mechanical and thermal properties of a-Si3 N4 are discussed. In a following study, they assembled pre-formed Si3 N4 clusters to form a nanophase Si3 N4 system, containing about a million atoms. The system was characterised as composed by (1) interfacial regions with a large fraction of undercoordinated Si, which separate (2) bulk-like domains from (3) the pores (three-phase model). The dependence of the computed elastic moduli on porosity and grain size, as well as the behaviour of the material under crack propagation, can be understood on the basis of the three-phase structural model. In particular, the nanophase system is able to sustain an order-of-magnitude larger external strain than crystalline Si3 N4 .

9.3 Simulations of ternary Si/C/N systems As previously discussed in the Introduction (Section 9.1), the simulation of disordered ternary systems is a formidable task, which has limited so far the application of atomicscale simulation techniques to the study of amorphous or poly-crystalline Si C N systems. In the following, we mainly review the recent theoretical advances in the field of crystalline Si C N systems, and in the study of disordered (amorphous or nanopowders) alloys.

©2003 Taylor & Francis

9.3.1

Theoretical studies of C defects in Si3 N4 and N defects in SiC

In this section, we give a brief account of the theoretical work that has been carried out to study the structural and electronic characteristics of the inclusion of either C or N in the crystalline network of the crystal formed by the two other species. In particular, we account for the most relevant modification of the structure and the electronic properties resulting from the inclusion of the substitutional impurity at very low concentrations, as a first step towards the understanding of the formation of ternary Si C N compounds. They are not necessarily the most stable point defects in these compounds. For example, in SiC, the SiC anti-site defects is more stable than both NSi and NC (Bernholc et al. 1989). The reader who is interested to the more general issue of point defects in semiconductors can look to the available reviews (see, for instance, Lannoo 1992). Fukumoto (1997) studied the inclusion of N in 3C SiC within the density functional theory (DFT) in the local density approximation (LDA) for the exchange and correlation effects (for a review on the method, see Jones and Gunnarsson 1989). Super-cells containing 64 atoms in total were used, and the relaxation of the atomic positions were performed through damped dynamics. However, in those calculations, the volume of the super-cell is not varied as a function of the position of the impurity, so that a spurious, residual pressure might somehow affect the energetic, since finite N concentrations are considered. Fukumoto considered only substitutional impurities, that is, NSi and NC . NC gives rise to a shallow donor state, with electron density widely spread over the entire crystal, and a maximum close to off-bond sites around Si. NSi , on the other hand, forms a deep level, characterised by a very localised electron distribution around the substitutional N. He also studied the formation energy of the defects, as a function of the Si chemical potential, and found that NC is always favoured over NSi by about 7 eV, in the whole physical range of Si chemical potentials compatible with the existence of stoichiometric silicon carbide. Amadon and Finocchi (1999) recently studied, within the same theoretical framework (DFT in the LDA) CSi β-Si3 N4 , at different C concentrations (1 or 2 carbons in a supercell containing 84 atoms). Both the atomic and the external degrees of freedom (i.e. the lattice parameters) are consistently relaxed for all the structures considered. For the lowest CSi concentrations, the C N bonds are stretched with respect to those computed in β-C3 N4 . When doubling the CSi concentration, scattered configurations of CSi are favoured over geometries with second neighbours CSi . The resulting microscopic stress is long ranged and lowers the cohesive energy of the host crystal. As a consequence, the formation of CSi defects in β-Si3 N4 from diamond and silicon nitride crystals, and molecular nitrogen is thermodynamically unfavoured, showing that the progressive substitution of Si with C in silicon nitride is not a practical way towards the synthesis of ternary alloys.

9.3.2

Crystalline Si/C/N compounds

Si C N ordered systems have only recently been obtained. They have been initially prepared in nanocrystalline phases by Riedel et al. (1997) through the reaction of silicon tetrachloride with trimethylsyl-carbo-diimide (CH3 )3 Si N C N Si (CH3 )3 , followed by calcination. The resulting amorphous Si C N powder crystallises at 400◦ C to give SiC2 N4 . The latter, when heated up to 900◦ C, decomposes in crystalline Si2 CN4 , C2 N2 and N2 . These phases were studied by in situ X-ray diffraction. Si2 CN4 shows a layered structure, formed by distorted SiN4 tetrahedra as in Si3 N4 , which are linked through carbodiimide groups. It has an orthorhombic unit cell (lattice parameters: a = 5.44 Å, b = 13.58 Å and c = 4.81 Å),

©2003 Taylor & Francis

[010]

[100] [001]

Figure 9.5 Structure of Aaba2 Si2 CN4 . N atoms are dark grey, Si light grey and C white. The three directions, parallel to the lattice vectors, are drawn. Note that in the [100] and in the [001] directions, the structure is replicated, in order to get a better feeling of the connectivity of the crystal.

with space group Aaba2 (Figure 9.5). However, the internal parameters defining the precise atomic positions were only roughly determined. The structure of SiC2 N4 was obtained too, but with less accuracy. The stability of these structures has been theoretically investigated by Kroll et al. (1999) through density functional calculations. They exploited the functional analogy between the carbodiimide and the O groups to build up various structural model compatible with the X-ray diffraction data. The cubic high-temperature β phase of SiC2 N4 consists of a Si atom at the centre of a tetrahedron formed by the four NCN groups. However the authors, thinking that the cubic phase may only reflect an average geometry, searched for other less symmetric structures. Two tetragonal structures resulted to have a higher cohesive energy than the cubic phase. The most stable one can be described as two interpenetrating Si N C N Si networks. It has a c/a ratio equal to 2 and a space group P4322 . Such a structure is only ≃4 KJ/mol more stable than the ideal cubic one, an energy difference which is close to the intrinsic precision of the theoretical method employed. However, the P4322 structure was proposed as a good candidate for the low-temperature α phase of SiC2 N4 , not only for energetic reasons. Indeed, the energy plotted as a function of the bonding angle of N shows a shallow minimum around 160◦ for the tetragonal structures, at odds with its 180◦ value in the cubic structure. The tetragonal structure has thus more flexibility, with two consequences: its computed bulk modulus (about 8 GPa) is much smaller than that of the cubic phase (108 GPa);

©2003 Taylor & Francis

at higher temperatures, the tetragonal structure might appear as cubic on average, due to the easy deformation of the bonding and dihedral angles. As far as the crystalline Si2 CN4 system is concerned, two theoretical studies appeared almost simultaneously: that by Kroll et al. (1999), and another one by Amadon and Finocchi (1999). Both of them found that the orthorhombic structure with space group Aaba2 proposed by Riedel et al. (1997) is a mechanically stable phase. The computed lattice parameters agree within less than 2% with those extracted from the X-ray diffraction data: a = 5.51 Å, b = 13.75 Å, c = 4.83 Å (Amadon and Finocchi 1999); a = 5.45 Å, b = 13.81 Å, c = 4.82 Å (Kroll et al. 1999); a = 5.44 Å, b = 13.58 Å, c = 4.81 Å (Riedel et al. 1997). Apart from a small discrepancy about the b/a ratio, both calculations found a larger equilibrium volume (by 2–3%) than the experiment. This is in contrast with the outcome of most density functional calculations adopting the LDA, which usually gives a trend to over-binding and to slightly underestimated lattice parameters – by 1–2% in elemental and compound semiconductors (Jones and Gunnarsson 1989). Due to the presence of very soft modes, essentially linked to the easy deformations of the Si N C angles, a non-monotone behaviour of the phonon frequencies v. the volume cannot be excluded, which might give a non-trivial value of the thermal expansion coefficient. However, for the moment, we are still at the level of speculations. An important consequence of the coexistence of more rigid Si N layers linked with deformable N C N sticks, into a highly non-isotropic structure, is the relevance of the relaxation of the internal degrees of freedom (i.e. the atomic positions inside the unit cell) for the calculation of the mechanical properties of these phases. From the computed elastic constants, Amadon and Finocchi calculated a bulk modulus equal to 110 GPa. Experimentally, Si2 CN4 decomposes at high temperatures in silicon carbide, silicon nitride, plus molecular nitrogen (reaction 1). Both calculations give a positive reaction enthalpy at 0 K, ranging from 215 to 285 kJ/mole. However, another possible reaction is possible, giving solid C, silicon nitride and molecular N as products (reaction 2). The path from the products in reaction 1 to those in reaction 2 implies a negative enthalpy, both experimentally and theoretically, of the order of about −280 kJ/mole. This raises the question of the stability of Si2 CN4 with respect to reaction 2. While Amadon and Finocchi found a reaction enthalpy in the range −20 to +10 kJ/mol, the latter was estimated at −39 kJ/mole by Kroll et al., who concluded for the thermodynamic instability of Si2 CN4 . However, Amadon and Finocchi were less conclusive on this point, since (i) the reaction enthalpies are likely to be too low by few tens of kJ/mole in the case of reaction 2, as a result of the use of the LDA, and (ii) the presence of high barriers and entropy contributions should favour the decomposition of Si2 CN4 in silicon carbide, silicon nitride, plus molecular N, as in reaction 1. Crystalline Si C N thin films were recently grown by microwave plasma-enhanced chemical vapour deposition on the (100) surface of Si (Chen et al. 1996), and then studied by X-ray absorption spectroscopy (Chang et al. 1998). The X-ray diffraction data were interpreted on the basis on a crystalline structure analogous to that of the α phase of Si3 N4 , from which it differs for the values of the lattice parameters (those of α-Si3 N4 are given in brackets): a = b = 6.908 Å (7.753 Å; c = 5.262 Å (5.618 Å). The XANES spectra of the thin Si C N films, taken at the C, N and Si K edge, showed marked similarities with those of the β phase of SiC, the α phase of Si3 N4 , and the combination of the two latter phases, respectively. Moreover, from the EXAFS oscillations at the Si K edge, Chang et al. derived an average Si N coordination equal to 3 (with a Si N bond length equal to 1.73 Å) and Si C mean coordination number equal to 1 (Si C bond length: 1.88 Å). These facts suggest that the local bonding environments of C and N in the ternary thin films are rather similar

©2003 Taylor & Francis

to those in the binary crystals formed with Si (β-SiC and α-Si3 N4 ). Si, at variance, shows a dominant Si N bonding, but the number of Si C bonds is not negligible. All the experimental results were complemented by theoretical calculations in the muffin-tin approximation for the scattering potentials provided by the ions. From these results, the authors suggested a structural model based on mixed C Si N3 tetrahedral units in which C N bonds should represent a very small fraction of the total number of bonds. These observations raise a few interesting issues. First, the thin films are said to be crystalline, but their composition was not precisely estimated. As a rough indication, under the hypothesis that C and N atoms are fourfold and threefold coordinated to Si, respectively, and that Si atoms form mixed C Si N3 tetrahedra, one obtains a C/Si ratio equal to 1/4, and a N/Si ratio equal to 1. Thus, the progressive inclusion of C atoms in an hypothetical α-Si3 N4 network, should be made by substitution of N with C, with a much smaller proportion of Ni replaced by Si atoms, which depends on the precise concentration of the three components. As a consequence of the difference of Si N and Si C bond lengths, a microscopic stress at the atomic scale would result, which may strongly influence the ground state properties of the crystals, such as its stability, mechanical behaviour and electronic structure, as in the case of CSi Si3 N4 and NC SiC, which have been discussed in Section 9.2. This observation is also coherent with recent experimental results reported for amorphous CSix Ny thin films (x ∼ 0.2, y ∼ 0.3) which were grown on (100) oriented Si substrates, by laser deposition (Thangen et al. 1999). The films are hard and plastic, and show an internal compressive stress of the order of a few GPa. A reversed approach has been undertaken by some theorists, who searched for super hard Si C N structures (for a review, see Badzian and Badzian 1997), obtained by progressive replacement of Si by C in β-Si3 N4 . On the basis of an empirical model, Cohen (1985) suggested that the structures characterised by tetrahedral units with short (and strong) covalent bonds are good candidates for materials with a high bulk modulus. Among them, the hypothetical solid β-C3 N4 , iso-structural to the β phase of Si3 N4 , should have a very large bulk modulus, comparable to that of diamond. Afterwards, some calculations (Liu and Cohen 1989; Liu and Wentzcovitch 1994) studied, on first-principle grounds, the stability and the mechanical properties of C3 N4 solids. In particular, through a variable-shape ab initio molecular dynamics simulations, Liu and Wentzcovitch (1994) confirmed the very low compressibility of the β phase (space group P63 /m). Such a crystal would have a short C N bond length (1.45 Å) and a bulk modulus equal to 437 GPa, just smaller than that of diamond. However, the authors found a more stable rhombohedral structure (space group R3m), about 0.15 eV/unit formula lower in energy than the β-C3 N4 phase. The rhombohedral phase is more similar to graphite than to diamond, with a C N bond length equal to 1.36 Å on average, and a quite low bulk modulus (51 GPa). Both phases are thermodynamically unstable vis-à-vis of the decomposition in diamond plus molecular N (the enthalpy of reaction ranging from −75 to −62 kJ/mol). However, as in the case of diamond, they found both phases mechanically stable, and did not exclude that they might be synthesised. These seminal studies motivated the search for hypothetical Si C N tetrahedrally bound compounds, obtained through the replacement of Si with C in Si3 N4 . Among them, Wang et al. (1998) proposed two model structures of β-Si2 CN4 and β-SiC2 N4 on the basis of density functional calculations in the full potential linear augmented plane wave method. They calculated the equilibrium structural parameters and the bulk modulus of the hypothetical structures, under symmetry restrictions. However, these calculations did not provide any information on the thermodynamic and mechanical stability of these compounds. Then, Lowther (1999) compared the β and the cubic phases of hypothetical SiC2 N4 with C3 N4 . He

©2003 Taylor & Francis

found that the electron distribution along the C N bonds in SiC2 N4 is rather similar to its counterpart in β-C3 N4 , which suggests a remarkable stability of the strong C N bonding in various chemical environments. Moreover, he found the cubic (P4n2) phase more stable than the β phase by a very tiny energy. Once again, the comparison of the cohesive energies of these ternary phases with respect to elemental or binary crystals was not done.

9.3.3

Si/C/N nanopowders and amorphous alloys

It is noteworthy that a fit to the X-ray or neutron diffraction spectra performed on Si C N nanopowders at different temperatures gives results similar to the binary crystals, regarding the Si N and Si C bond lengths (see Chapter 5). In analogy with the results by Chang et al. on the Si C N crystalline thin films, Ténégal et al. (1996) found that the local bonding structure of amorphous silicon carbonitride powders, as probed by EXAFS, may be described in term of mixed C Si N3 tetrahedrons, but, at the same time, evidence for C N bonds connecting the tetrahedrons was also found. The same samples were further analysed by X-ray and neutron diffraction. The experimental spectra were interpreted (Ténégal et al. 1998) on the basis of a medium-range order based on different stacking of the Si/C/N layers, reminiscent of either the α or β phase of crystalline silicon nitride. Also in that case, the Si3 N4 crystalline phases were used as input to build up structural models through: (i) replacement of Si or N atoms by carbons and (ii) a local optimisation of the structure by using empirical force fields. One can note that all the models derived through the selective replacement of C atoms in a silicon nitride crystalline network have the virtue of fitting the experimental data remarkably, thereby providing an atomic scale picture of the configurations of ternary Si C N systems. However, they are not complemented by any thermodynamic data, or by the calculation of the electronic properties for the corresponding atomic arrangement. As a consequence, they are neither able to explain the Si N, C N and Si C bonding properties within the disordered network, nor understand the thermodynamic evolution of the Si C N systems. Moreover, they have not been showed to be coherent with the measured electronic structure of the ternary amorphous or crystalline Si C N systems. The comparison between photoemission experiments (Driss-Khodja et al. 1996) and the electronic structure computed for the proposed structural models would be thus significant to assess their validity. To our knowledge, Matsunaga et al. (1999) were the first to simulate truly amorphous Si3 Cz N4−z ceramics. They employed a Tersoff potential to describe the C C, Si S, Si N and Si C interactions. Both C N and N N interatomic potentials are purely repulsive. The justification of such a simplified model was given on the basis of the high mobility of N2 dimers – which do not contribute to the bonding in the condensed phase – as far as the N N interaction is concerned. The C N attractive interaction was not taken into account since a decomposition of C N bonds is observed during pyrolysis above 1273 K (Seher et al. 1994). The molecular dynamics simulations were performed in the canonical (NVT) ensemble, by generating the amorphous configurations through a rapid quench from a liquid-like structure at ∼4000 K. The statistic was taken on the sample equilibrated at T ∼ 1270 K for about 50 ps. Several C concentrations z were simulated. The results were characterised in term of partial pair distribution functions. For Si3 C1.5 N2.5 , both N and C atoms are undercoordinated (ZN ∼ 2.4; ZC ∼ 3.1), at odds with Si (ZSi ∼ 4.2), which are bonded to 2 C atoms, 1.4 N atoms and 0.8 Si atoms, on average. As more C is added, more and more Si C bonds are created at the expense of Si N bonds.

©2003 Taylor & Francis

The overall picture of the structure of the amorphous sample cannot in any case be characterised as purely consisting of mixed C Si N3 and C2 Si N2 tetrahedra, since both the presence of Si Si bonds, and the low coordination number of N and C reveal a bigger complexity. However, the computed total pair distribution function for Si3 C1.5 N2.5 (Figure 9.6), summed on the different bonding configurations, fairly compares to that measured by X-ray diffraction (Uhlig et al. 1996) for Si3 C1.8 N2.7 .

(a) (SiC0.50N0.83)

3

2

G (r )

1

0

(b)

(SiC0.60N0.90)

3

2

1

0 0.10

0.20

0.30 0.40 r (nm)

0.50

0.60

Figure 9.6 (a) Computed total distribution function for amorphous Si3 C1.5 N2.5 , by Matsunaga et al. (1999). (Reproduced from kind permission of Ceramic Society of Japan.) (b) The comparison with the experimental one (Uhlig et al. 1996) is also shown. The first peak marked in the figure corresponds to both Si N and Si C bonds. The second one is due to Si Si bonds, while the third peak comes from C C, N N and N C second neighbours.

©2003 Taylor & Francis

Such a simulation raises the interesting issues of the presence of (minority) Si Si and C N bonding, as well as of C and N dangling bonds, eventually saturated by hydrogen, which can be studied further. Also, the role of the formation of micro-voids may help the release of the internal stress and the measured mechanical properties. In conclusion, the available theoretical models give a still incomplete picture of Si C N systems. Apart from the theoretical and experimental investigations on carbodiimide-based Si C N crystals, some structural models are available, which fit the experimental data, without explaining why the proposed structures are (meta)stable. Some other theoretical investigations provide precise information on the energetic and the mechanical stability of hypothetical compounds, which have not yet been experimentally obtained. In very few cases the authors showed the way to a practical synthesis of these hypothetical compounds. A joint theoretical and experimental effort should be done in order to bridge the gap between the measured properties of these ternary systems and their explanation in terms of a unifying model.

References Allen, M. P. and Tildesley, D. J. (1987) Computer Simulations of Liquids. Clarendon Press, Oxford. Amadon, B. and Finocchi, F. (1999) Eur. Phys. J., B11, 207. Baskes, M. I. (1999) Curr. Opin. Solid State Mater. Sci., 4, 273. Bader, R. W. F. (1991) Chem. Rev., 91, 983. Badzian, A. and Badzian, T. (1997) Int. J. Refractory Metals Hard Mater., 15, 3. Bernholc, J., Antonelli, A., Wang, C., et al. (1989) Mater. Sci. Forum, 38/41, 713. Car, R. and Parrinello, M. (1985) Phys. Rev. Lett., 55, 2471. Car, R. and Parrinello, M. (1988) in Simple Molecular Systems at Very High Densities (eds A. Polian, P. Loubeyre and N. Boccara), Plenum Press, New York. Cargill III, G. S. and Spaepen, F. (1981) J. Non-Cryst. Solids, 43, 91. Chen, L. C., Yang, C. Y., Bhusari, D. M., et al. (1996) Diamond Relat. Mater., 5, 514. Chang, Y. K., Hsieh, H. H., Pong, W. F., et al. (1998) Phys. Rev., B58, 9018. Cohen, M. L. (1985) Phys. Rev., B32, 7988. de Brito Mota, F., Justo, J. F. and Fazzio, A. (1998) Phys. Rev., B58, 8323. de Brito Mota, F., Justo, J. F. and Fazzio, A. (1999) J. Appl. Phys., 86, 1843. Driss-Khodja, M., Gheorghiu, A., Dufour, G., et al. (1996) Phys. Rev., B55, 4287. Faraday Discussion (1997) Solid State Chemistry: New Opportunities From Computer Simulations, 106. Finocchi, F., Galli, G., Bertoni, C. M., et al. (1992) Phys. Rev. Lett., 68, 3044. Finocchi, F., Galli, G., Bertoni, C. M., et al. (1993) Physica, B185, 379. Finocchi, F. and Galli, G. (1994) Phys. Rev., B50, 7393. Fukumoto, A. (1997) Phys. Stat. Sol., 202, 125. Guraya, M. M., Ascolani, H., Zampieri, G., et al. (1990) Phys. Rev., B42, 5677. Jones, R. and Gunnarsson, O. (1989) Rev. Mod. Phys., 61, 689. Kalia, R. K., Nakano, A., Tsuruta, K., et al. (1997) Phys. Rev. Lett., 78, 689, 2144. Kelires, P. C. (1991) Europhys. Lett., 14, 43. Kroll, P., Riedel, R. and Hoffman, R. (1999) Phys. Rev., B60, 3126. Liu, A. Y. and Cohen, M. L. (1989) Science, 245, 841. Liu, A. Y. and Wentzcovitch, R. M. (1994) Phys. Rev., B50, 10362. Lannoo, M. (1992) in Handbook on Semiconductors, vol. I. (ed. T. S. Moss), Elsevier, Amsterdam. Lowther, J. E. (1998) Phys. Rev., B57, 5724. Lowther, J. E. (1999) Phys. Rev., B60, 11943. Matsunaga, K., Iwamoto, Y., Fisher, C. A. J., et al. (1999) J. Ceram. Soc. Jpn, 107, 1025.

©2003 Taylor & Francis

Meyer, M. and Pontikis, V. (eds), Proceedings of the NATO Advanced Study Institute on Computer Simulations in Material Science, Aussois, France (1991). Series E: Applied Science – vol. 205. Kluwer Academic Publishers (Dordrecht/Boston/London). Mousseau, N. and Lewis, L. J. (1997) Phys. Rev., B56, 9641. Mura, D., Colombo, L., Bertoncini, R., et al. (1998) Phys. Rev., B58, 10357. Oguch, T. and Sasaki, T. (1991) Prog. Theor. Phys., 103, 93. O’Hara, P. A. G., Tomasziewicz, I. and Seifert, H. J. (1997) J. Mater. Res., 12, 3203. Riedel, R. A., Greiner, G., Miehe, et al. (1997) Angew. Chem., 36, 603. Robertson, J. (1991) J. Non-Cryst. Solids, 137/138, 825. Rosato, V., Celino, M. and Colombo, L. (1998) Comp. Mater. Sci., 10, 67. Seher, M., Bill, J., Aldinger, F., et al. (1994) J. Cryst. Growth, 137, 452. Sénémaud, C., Driss-Khodja, M., Gheorghiu, A., et al. (1993) J. Appl. Phys., 74, 5042. Servalli, G. and Colombo, L. (1993) Europhys. Lett., 22, 107. Ténégal, F., Flank, A. M. and Herlin, N. (1996) Phys. Rev., B54, 12029. Ténégal, F., Bouchet, B., Bellissent, R., et al. (1998) Philos. Mag. A, 78, 803. Tersoff, J. (1989) Phys. Rev., B39, 5566. Tersoff, J. (1994) Phys. Rev., B49, 16349. Thangen, T., Lippold, G., Riede, V., et al. (1999) Thin Solid Films, 348, 103. Uhlig, H., Frieß, M., Durr, J., et al. (1996) Z. Naturforsh., 51a, 1179. Umesaki, N., Hirosaki, N. and Hirao, K. (1992) J. Non-Cryst. Solids, 150, 120. Vashista, P., Kalia, R. K. and Ebbsjo, I. (1995) Phys. Rev. Lett., 75, 858. Vollmayr, K., Kob, W. and Binder, K. (1996) Phys. Rev., B54, 15808. Wang, C. -Z., Ho, K. M. and Chan, C. T. (1993) Phys. Rev. Lett., 70, 611. Wang, C. -Z., Wang, E. -G. and Dai, Q. (1998) J. Appl. Phys., 83, 1975. Waseda, Y. (1980) The Structure of Non-crystalline Materials. Mc-Graw Hill, New York. Wooten, F. and Weaire, D. (1987) Solid State Phys., 40, 1.

©2003 Taylor & Francis

Nanostructured Silicon-based Powders and Composites

Edited by André P. Legrand and Christiane Sénémaud

©2003 Taylor & Francis

First published 2003 by Taylor & Francis 11 New Fetter Lane, London EC4P 4EE Simultaneously published in the USA and Canada by Taylor & Francis Inc, 29 West 35th Street, New York, NY 10001 Taylor & Francis is an imprint of the Taylor & Francis Group © 2003 Taylor & Francis Typeset in Times New Roman by Newgen Imaging Systems (P) Ltd, Chennai, India Printed and bound in Great Britain by TJ International, Padstow, Cornwall All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Every effort has been made to ensure that the advice and information in this book is true and accurate at the time of going to press. However, neither the publisher nor the authors can accept any legal responsibility or liability for any errors or omissions that may be made. In the case of drug administration, any medical procedure or the use of technical equipment mentioned within this book, you are strongly advised to consult the manufacturer’s guidelines. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data A catalog record for this book has been requested ISBN 0-415-30113-0

Cover: A laser irradiation cell observed through a KCl window

©2003 Taylor & Francis

Christiane Sénémaud was a CNRS Directeur de Recherches until she passed away in 1998. Her area of expertise was the determination of the electronic structure of complex materials by X-ray photoelectrons and soft X-ray spectroscopies. She had published more than one hundred papers and communications in international meetings, and owing to her open mind and vast knowledge, she skilfully established fruitful cooperation between foreign and French laboratories. Moreover, she directed the theses of many young students and contributed to graduate education at the university, more precisely the ‘Complex Materials’ doctoral school. Finally, she was the leader of the CNRS research group that was the origin of this book. Her merit was recognised with the attribution of the Charles-Louis de Saulces de Freycinet Award by the French Academy of Sciences. Christiane Sénémaud was above all smart, discrete, fair, active, efficient and kind. For these reasons, the authors would like to dedicate this book to her memory.

©2003 Taylor & Francis

Contents

List of contributors Foreword Preface Acknowledgements 1

Objectives and state-of-the-art of nanocomposites M I C H E L C AU C H E T I E R A N D A N D R É P I E R R E L E G R A N D

1.1 1.2 1.3 1.4 1.5

2

Introduction Nanostructured ceramic elaboration Nanostructured Si 3 N4 /SiC materials Analytical methods Conclusion References

Laser synthesis of nanosized powders M I C H E L C AU C H E T I E R , E M M A N U E L M U S S E T, M I C H E L L U C E , NAT H A L I E H E R L I N , X AV I E R A R M A N D A N D M A RT I N E M AY N E

2.1 2.2 2.3 2.4 2.5

3

Generalities on the synthesis of nanosized powders by laser pyrolysis Synthesis of Si, SiC, Si 3 N4 and SiC Si 3 N4 mixtures with SiH 4 and SiH 3 CH 3 as silicon precursors Synthesis of silicon carbonitride (Si/C/N) powders using hexamethyldisilazane as Si precursor Silicon carbonitride (Si/C/N) powders containing in situ the elements of the sintering aids (Al, Y): pre-mixed powders Conclusion References

Thermal behaviour of as-formed silicon-based nanopowders

©2003 Taylor & Francis

3.1 Thermal behaviour in a high-temperature graphite furnace E M M A N U E L M U S S E T, M A RT I N E M AY N E , M I C H E L C AU C H E T I E R , X AV I E R A R M A N D , M I C H E L L U C E A N D NAT H A L I E H E R L I N - B O I M E

3.1.1 3.1.2 3.1.3 3.1.4 3.1.5 3.1.6

Introduction Weight evolution of SiC, Si/C/N and Si/C/N/Al/Y/O nanopowders Chemical evolution during annealing in a nitrogen atmosphere Structural changes Information about changes in the grain sizes Conclusion References

3.2 Thermal reactivity of silicon-based nanopowders D JA M I L A BA H L O U L - H O U R L I E R , B E N O Î T D O U C E Y, J E A N - L O U I S B E S S O N A N D PAU L G O U R S AT

3.2.1 3.2.2 3.2.3

4

Introduction Results and discussions Conclusion References

Structure of some nanopowders and micro/nano composites by transmission electron microscopy M A R C M O N T H I O U X , C AT H E R I N E B É R AU D A N D NA H I DA E L H O R R

4.1 4.2 4.3

5

As-prepared and heat-treated nanopowders As-prepared sintered composites TEM sample preparation prior to investigations and operating conditions References

From short- to long-range order

5.1 Chemical order studied by solid-state nuclear magnetic resonance A N D R É P I E R R E L E G R A N D , J E A N - BA P T I S T E D ’ E S P I N O S E D E L A C A I L L E R I E A N D YO U S S E F E L KO RTO B I

5.1.1 5.1.2 5.1.3 5.1.4 5.1.5

Nuclear magnetic methods SiC powders analysis Si/C/N powders analysis Pre-alloyed Si/C/N/Al/Y nanopowders Conclusion References

©2003 Taylor & Francis

5.2 Paramagnetic defect states in silicon-based nanopowders A B D E L H A D I K A S S I BA A N D S T É P H A N E C H A R P E N T I E R

5.2.1 5.2.2 5.2.3 5.2.4 5.2.5 5.2.6

Introduction Theoretical background Experimental details Si/C powders Assignment of the paramagnetic species in Si/C nanopowders Silicon carbonitride (Si/C/N) powders References

5.3 Electronic structure studied by X-ray photoemission (XPS) and soft-X-ray (SXS) spectroscopies A D R I A NA G H E O R G H I U D E L A RO C QU E , G E O R G E S D U F O U R , F R A N Ç O I S T É N É G A L A N D C H R I S T I A N E S É N É M AU D

5.3.1 5.3.2 5.3.3 5.3.4

Principles Experiments Study by XPS–SXS of Si, Si/C, Si/N and Si/C/N laser synthetised nanopowders produced from gaseous precursors Effect of heat treatment on the electronic structure of nanometric Si/C/N ex-HMDS and pre-alloyed Si/C/N/Al/Y powders studied by XPS spectroscopy References

5.4 Local order in Si/C/N nanopowders studied by X-ray absorption spectroscopy: study at the Si K edge F R A N Ç O I S T É N É G A L , A N N E - M A R I E F L A N K , A D R I A NA G H E O R G H I U - D E L A RO C QU E A N D C H R I S T I A N E S É N É M AU D

5.4.1 5.4.2

X-ray absorption principles and data analysis Study by EXAFS at the Si K edge of Si, SiN, SiC and Si/C/N laser-synthetised nanopowders produced from gaseous precursors 5.4.3 Short-range order determination in amorphous SiCN-HMDS nanopowders: evolution as a function of the C/N ratio and annealing treatment 5.4.4 Local structure of pre-alloyed Si/C/N/Al/Y nanopowders studied by XAS at the Al K edge using fluorescence yield detection References

5.5 Neutron and X-ray diffraction applied to the study of the medium-range order in SiCN nanopowders F R A N Ç O I S T É N É G A L , B R I G I T T E B O U C H E T- FA B R E , J E A N D I X M I E R A N D RO B E RT B E L L I S S E N T

5.5.1

Diffraction applied to the study of disordered compounds

©2003 Taylor & Francis

5.5.2

6

Early stages of the crystallisation of Si/C/N-HMDS nanopowders studied by X-ray and neutron diffraction: Nano-ordered Si/C/N structures References

Conductivity and dielectric properties of Si/C nanopowders A B D E L H A D I K A S S I BA

6.1 6.2 6.3 6.4

7

Introduction Material specifications Experimental results and interpretation Conclusion References

Processing and tailoring of Si/C/N-based nanocomposites PAU L G O U R S AT, D JA M I L A BA H L O U L - H O U R L I E R , B E N O Î T D O U C E Y A N D M A RT I N E M AY N E

7.1 7.2 7.3 7.4 7.5

8

Introduction Sintering of silicon nitride with additives Sintering of silicon carbide Si 3 N4 /SiC nanocomposites Conclusion References

Mechanical properties JEAN-LOUIS BESSON

8.1 8.2 8.3

9

Room temperature properties High-temperature properties Conclusion References

Theoretical investigations of Si/C/N-based alloys FA B I O F I N O C C H I

9.1 9.2 9.3

Introduction Simulations of Si/C and Si/N disordered alloys Simulations of ternary Si/C/N systems References

©2003 Taylor & Francis

Contributors

Xavier Armand Commissariat à l’Energie Atomique, Service des photons, atomes et molécules, Bat 522. 91191 Gif-sur-Yvette Cedex, France Robert Bellissent Laboratoire Léon Brillouin, CEA-Saclay, batiment 563, 91191 Gif-surYvette Cedex, France Catherine Béraud Laboratoire de Physique de la Matière Condensée, INA,135 avenur de Raugueil, 31077 Toulouse Cedex, France Jean-Louis Besson SPCTS, UMR CNRS 6638, ENSCI, 47-73, avenue Albert Thomas, 87065 Limoges Cedex, France Brigitte Bouchet-Fabre Laboratoire d’Utilisation du Rayonnement Electromagnétique, Bat. 209, Centre Universitaire Paris-Sud, 91405 Orsay Cedex, France Michel Cauchetier Commissariat à l’Energie Atomique Service des photons, atomes et molécules, Bat 522. 91191 Gif-sur-Yvette Cedex, France Stéphane Charpentier Laboratoire de Physique de l’Etat Condensé, Université du Maine, Faculté des Sciences, Boulevard Olivier Messiaen, 72085 Le Mans Cedex, France Jean Dixmier Laboratoire de Physique du Solide, 1 place Aristide Briand, 92195 Meudon Cedex, France Benoît Doucey SPCTS, UMR CNRS 6638, Faculté desSciences, 123, avenue Albert Thomas, 87060 Limoges Cedex, France Georges Dufour UMR CNRS 7614, Laboratoire de Chimie-Physique, Université Pierre et Marie Curie, 11 rue P.M. Curie, 75231 Paris Cedex 05, France Nahida El Horr CEMES, BP 4347, 29 rue Jeanne Marvig, 31055 Toulouse Cedex 4, France Youssef El Kortobi FRE CNRS 2312, Laboratoire de Physique Quantique, Ecole Supérieure de Physique et de Chimie Industrielles de la Ville de Paris, 10, rue Vauquelin, 75231 Paris Cedex 05, France Fabio Finocchi UMR CNRS 7588, Groupe de Physique des Solides, UPMC, 2, place Jussieu, 75251 Paris Cedex 05, France Anne-Marie Flank Laboratoire d’Utilisation du Rayonnement Electromagnétique, Bat. 209, Centre Universitaire Paris-Sud, 91405 Orsay Cedex, France

©2003 Taylor & Francis

Adriana Gheorghiu-de La Rocque UMR CNRS 7614, Laboratoire de Chimie-Physique, Université Pierre et Marie Curie, 11, rue P.M. Curie, 75231 Paris Cedex 05, France Paul Goursat SPCTS, UMR CNRS 6638, Faculté des Sciences, 123, avenue Albert Thomas, 87060 Limoges Cedex, France Nathalie Herlin-Boime Commissariat à l’Energie Atomique, DSM/DRECAM/SPAMLaboratoire Francis Perrin URA CNRS 2453, Bat 522. 91191 Gif-sur-Yvette Cedex, France Djamila Bahloul-Hourlier SPCTS, UMR CNRS 6638, Faculté des Sciences, 123, avenue Albert Thomas, 87060 Limoges Cedex, France Abdelhadi Kassiba Laboratoire de Physique de l’Etat Condensé, Université du Maine, Faculté des Sciences, Boulevard Olivier Messiaen, 72085 Le Mans Cedex, France Jean-Baptiste d’Espinose de la Caillerie FRE CNRS 2312, Laboratoire de Physique Quantique, Ecole Supérieure de Physique et de Chimie Industrielles de la Ville de Paris, 10, rue Vauquelin, 75231 Paris Cedex 05, France André Pierre Legrand FRE CNRS 2312, Laboratoire de Physique Quantique, Ecole Supérieure de Physique et de Chimie Industrielles de la Ville de Paris, 10, rue Vauquelin, 75231 Paris Cedex 05, France Michel Luce Commissariat à l’Energie Atomique Service des photons, atomes et molécules, Bat 522. 91191 Gif-sur-Yvette Cedex, France Martine Mayne Commissariat à l’Energie Atomique, DSM/DRECAM/SPAM-Laboratoire Francis Perrin URA CNRS 2453, Bat 522. 91191 Gif-sur-Yvette Cedex, France Marc Monthioux CEMES, BP 4347, 29 rue Jeanne Marvig, 31055 Toulouse Cedex 4, France Emmanuel Musset Commissariat à l’Energie Atomique, Service des photons, atomes et molécules, Bat 522. 91191 Gif-sur-Yvette Cedex, France Christiane Sénémaud UMR CNRS 7614, Laboratoire de Chimie-Physique, Université Pierre et Marie Curie, 11, rue P.M. Curie, 75231 Paris Cedex 05, France François Ténégal Laboratoire d’Utilisation du Rayonnement Electromagnétique, Bat. 209, Centre Universitaire Paris-Sud, 91405 Orsay Cedex, France

©2003 Taylor & Francis

Foreword

The preparation of inorganic materials by thermolysis of preceramic compounds has become of substantial interest for the production of innovative materials in recent years. The general idea of this type of process route is that element-organic precursor molecules are constituted of structural units of the inorganic material aimed for by this process, thus providing novel paths of controlling the atomic array, composition and microstructure of materials. Of special interest is the manufacture of amorphous or nano-crystalline covalently bounded inorganics on the basis of silicon, boron, carbon and nitrogen, revealing a unique potential of properties not known from conventionally prepared materials. It is the intention of this book to provide a comprehensive survey on the French contributions to the synthesis of nano-sized powders and the production of nano-structured siliconbased composites by liquid-phase sintering using nano-sized powders where interfaces play a decisive role for the evolution of properties. The infrared laser thermolysis of gaseous or liquid precursors is a highly versatile method well-fitted for the production of silicon-based amorphous or crystalline nanopowders. Due to the purity level and degree of dispersion, such powders are good candidates for the application in ceramics. Apart from ‘pure’ powders consisting of silicon, carbon and/or nitrogen, such ones containing homogeneously distributed alumina and yttria are typically used as sintering aids for silicon carbide and silicon nitride based materials. It has been proven that there is a strong influence of the carrier gases used for the synthesis laser process on residual hydrogen bonded to silicon, carbon and/or nitrogen providing some functionality during further thermal treatments. For the analysis of the amorphous and nano-crystalline structure of the as-thermolysed and heat treated powders, morphological and analytical transmission electron microscopy studies are decisive. Chemical ordering of the material such as powders in particular is analysed by nuclear magnetic resonance studies. Further analytical investigations are electron paramagnetic resonance, X-ray photoelectron spectroscopy, X-ray absorption spectroscopy, X-ray and neutron diffraction. Furthermore, numerical simulations are used for the study of the chemical disorder and to propose structural models for the structure of the amorphous state of the materials. The content of the book is – even though very comprehensive – very detailed and strongly interdisciplinary. Therefore, the book is valuable for a broad readership in research and education.

©2003 Taylor & Francis

The editors are well-known experts in the field of research covered by the book. They can be congratulated for a proper selection of authors and for providing an excellent piece of work. Fritz Aldinger Max-Planck-Institut für Metallforschung, Stuttgart, Germany June 2001

©2003 Taylor & Francis

Preface

Physical properties of solids are greatly influenced by the arrangements of atoms present in the network. Nanostructured systems can offer improved and new properties compared to those of bulk materials. Consequently, in such materials, the influence of interfaces and interphases are of noticeable importance. Such interactions between the grains can lead to new physical and chemical properties. In recent years, an increasing interest has been devoted to nanostructured composites. This attention is largely due to exciting possible applications ranging from new catalysts to the preparation of nanocomposite ceramics, with significant improvements in their properties. Ceramic composites, obtained by dispersing nanoscale particles of one constituent into larger particles of a second constituent, have shown such improvements in mechanical properties (micro/nano composites). In particular, superplasticity has been obtained by mixing two equally fine constituents (nano/nano composites). Different modes of preparation have been used to make nanosized particles in large enough quantities. Among them, laser ablation of solid targets, plasma- or laser-induced reactions of gaseous mixtures have been developed. Other methods are related to evaporation– condensation, ball milling, self-propagating high temperature synthesis, sol–gel, spray drying of solutions, aerosol pyrolysis and chemical vapour decomposition at low pressure. Due to the large variety of such materials, comparative analysis is questionable. The originality of the book consists principally of focusing on a series of identical materials including preparation conditions, physical and chemical analysis, preparation and mechanical property determination of silicon-based nanocomposites. André P. Legrand ([email protected])

©2003 Taylor & Francis

Acknowledgements

We thank the Chemistry Department of the Centre National de la Recherche Scientifique (CNRS), which endorsed this work through a Groupe de Recherche (CNRS GdR 1168), for its interest and partial support, and more particularly its Director Jean-Claude Bernier and Associate Director Pierre Henri Dixneuf. The Délégation Générale pour l’Armement contributed to the development of this research through its financial support. We express here our sincere thanks.

©2003 Taylor & Francis

Part 1

NOVEL NANOSTRUCTURES AND DEVICES

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

1Nanopatterning with DiblockCopolymers P. M. Chaikin'' 2 , C. Harrison', M. Park', R. A. Register 2' 3 , D. H. Adamson 2 , D. A. Huse', M. A. Trawick', R. Li 4 and P. Dapkus 4 ' Department of Physics, PPrinceton Materials Institute, 3 Department of Chemical Engineering, Princeton University, Princeton, New Jersey 08544, Compound Semiconductor Laboratory, Department of Electrical Engineering/ Electrophysics, University of Southern California, Los Angeles, CA 90089

1.1 INTRODUCTION There has been an interest in going beyond conventional lithographic techniques in order to make features of ever smaller scale and higher density over large areas. In this paper we discuss progress that has been made over the past decade in using the self-assembly of diblock copolymer films as a template for creating two dimensional patterns (lines and dots) with a characteristic spacing of 20-30 nm. Typically trillions of dots, holes, posts of semiconductors and metals are produced on conventional semiconductor wafers. We describe the basic concept of the pattern formation and the technology of the transfer of the pattern from soft to hard materials. In order to produce and study these nanoscopic patterns we had to develop some new techniques for getting two and three dimensional images. 3D depth profiling with reactive ion etch (RIE) slices of 7 nm thickness alternating with electron microscope pictures proved very effective. We became very interested in the pattern formation and annealing necessary to control the long range order of the arrays and found new ways to follow the ordering. The coarsening was found to obey a t" 4 power law, (that is the size of the "grains" grew with time with this dependence) and at least for the striped pattern (cylinders lying down in a fingerprint like pattern) we could understand the microscopic origin of this behavior. We studied these phenomena with time lapse AFM microscopy and found that the disorder was dominated by the presence of disclinations and the annealing occurred by the annihilation of disclination multipoles rather than simple disclination antidisclination, dipole dynamics. We also found that the orientation of the patterns could be controlled by introducing alignment marks, step edges.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

4

Chaikin et al.

1.1.1 Why Nanolithography? Our interest in periodic patterns on the nanometer scale originated in a physics problem, the Hofstadter (1976) "butterfly", a problem of incommensurability between a periodic potential and flux quantization in a magnetic field. The competition of lengths-scales leads to a fascinating fractal energy spectrum. The best way to observe these effects is to take a quantum Hall device in the lowest Landau level and decorate it with a periodic potential on the scale of the cyclotron radius (Thouless, 1982). For a magnetic field of 1 Tesla the characteristic magnetic length, 1, which gives a flux quanta (4o he/e) through its area (H1 2= Oo ) is 1-20 nm. We therefore wanted to create a two dimensional lattice with unit cells on this scale and transfer a potential from this pattern to the two dimensional electron gas that resides about 20 nm below the surface of a quantum Hall device. Lithographic techniques are constantly evolving and the feature size is getting smaller. Presently large scale integrated devices (like Pentium chips) are produced by optical lithography with feature size >150 nm. Smaller features are readily produced by electron beam lithography, down to >25 nm, but it is difficult to place such features next to one another at the same scale and to produce periodic arrays of them. Moreover it is extremely time consuming to cover large areas with such patterns if each must be separately written, even once. Aside from the Hofstadter spectrum such dense periodic arrays should have interest for magnetic disk drives, for addressable memories, as optical elements, for quantum dots, for excitation and transfer between dye molecules and in biology as filters and sensors for proteins and nucleic acid sections. In many of these applications, e.g. filters, disks drives, quantum dots, it is the size and density that are of interest, while in other applications the periodicity and long range order are required. Our interest in using diblock copolymers for this work was initiated in discussions with Dr. Lew Fetters who had studied the synthesis and three dimensional structure of different diblock copolymer phases (Morton, 1975). The cross sections of his samples showed beautiful lattices with spacings on the 20 nm scale and perfect order over many microns. The idea of transferring these patterns to other organic and inorganic substrates was attractive since the copolymer selfassembly could be done over large scales simultaneously, the morphology and length scale were chemically modifiable and the materials were fairly easy to work with. The basic physical and chemical properties of the diblock copolymers were already a well developed science in the bulk and they could be readily processed by techniques used in conventional semiconducting lithography. The fact that nanoscale patterns remained suitable in thin films was demonstrated by Mansky et al. (Mansky, 1995, 1996).

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanopatterning with Diblock Copolymers

5

1.1.2 Basics of Diblock Copolymers Consider two types of monomers, A and B, which have a net repulsive interaction. When we make a polymer of each AAAAAAAA and BBBBBBB, the repulsive interaction between the segments is enhanced (by the number of monomers per segment) and a mixture of the two would phase separate like oil and water with one floating above the other (de Gennes, 1979). However, if the segments or blocks are covalently bound, AAAAAAAA BBBBBBB, making a diblock copolymer, figure ]a), they cannot macroscopically phase separate. The best that they can do is microphase separate putting all the A's together and all the B's together with an interface between them. If the blocks are of similar length then the arrangement shown in figure lb) is appropriate and a lamellar phase results. If the A segment is much smaller than ~ A~A.B~t4~H

Fe

a)

LA L Pads A With A B urthB ~'

Lamellae d)

es

Cylinders

XN L

fA 0.5 Figure 1 a) Schematic of diblock copolymer. b) Microphase separation into lamellae when A and B segments have about the same length. c) Cylindrical and Spherical phases form when segments have very different length. d) Mean field phase diagram of as a function of total repulsion between segments XN and fraction of diblock which is A monomers.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

6

Chaikin et al.

the B segment, figure 1c ) then the area on the A side of the interface is smaller than on the B side and there is a natural curvature to the interface. The result is cylinders or spheres which arrange on a Hexagonal or Body Centered Cubic lattice respectively. The mean field phase diagram as originally calculated by Leibler (1980) is shown schematically in figure ld). • is the Flory parameter (Flory, 1953) which characterizes the repulsive interaction and N is the polymerization index (number of monomers per polymer) and for a specific monomer is proportional to the polymer molecular weight. •N is then a measure of the total repulsion between polymer blocks and the microphase separation occurs when that energy overcomes the mixing entropy. f A is the fraction of diblock which is A. As suggested above we have lamellae when there are equal A and B length or f A =0.5. For A rich phases (f > .5) we have regions of B in a continuous matrix of A and conversely for B rich phases (f A < .5). The actual phase diagram depends on the actual intermolecular interactions and asymmetries and is considerably more complex with fascinating multiple interconnected phases such as the gyroid. Interested readers are referred to the excellent reviews by Bates (Bates, 1990, Bates, 1991). For our purposes the cylindrical and spherical phases are most useful.

SILICON

Figure 2 Cartoon of monolayer films of spherical and cylindrical phase diblock copolymers with rubber component wetting both surfaces.

From what we know of the three dimensional phase diagram we thought that we could make monolayers, as cartooned in figure 2 which we might use to pattern transfer and form lines or dots in inorganic materials. What controls the length scale? Clearly the stretched length of the polymer, Na, where a is a monomer size, is a limit. The actual scaling is a playoff between the interfacial energy, 6, between the A and B rich phases, and the elastic energy in stretching the polymers. Consider a particular structure, say spheres, which is set by f A the ratio of A to B. We want to know the size of a microphase separated region, a micelle, and the number of polymers, n, associated with it. If the length scale is R then the interfacial energy is the area S times the surface tension, a, or Sa/n (=47LR26/n) per polymer. The number of monomers in the micelle is proportional to the number of polymers times the number of monomers per polymer chain, nN. It is also proportional to the volume V of the micelle times the monomer density p, or n = pV/N. If the chains are Gaussian and are stretched to a length R then 2 the harmonic elastic energy per chain is (3/2)(W/Na )k B T. The Elastic plus 2 2 2 interfacial energy is SaN/pV + (3/2)(R /Na )kB T = C(aN/pR) + (3/2)(W/Na )k B T, where C = SR/V = 3 for spheres, 2 for cylinders and 1 for lamellae. Minimizing 2/s 2 2 with respect to R we have R = N (CGa /3pk n T)' .

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanopatterning with Diblock Copolymers

(Co-a2)

?

R=N 3

7

(1.1)

3pkBT

Typically for a Polystyrene-Polybutadiene, PS-PB, diblock with molecular weight PS-PB 65-10 kg/mol we have PB spheres of 10 nm diameter with a center to center spacing of 25 nm. The spherical micelles consist of about 100 PB chains with about 100 monomers per chain. The size and periodicity will scale with molecular weight to the 2/3 power. Plastic H poly(styrene)

-

H CH, I

H

-

H

k l H H_

H

H

H

Rubbers

H H-

1,4 peIy(buFadicne)

1,4 polylisuprcne)

e.g.

H

-c-c=c-c-

-c-c=c-c-

• polpsLyrenc-r.i i,opiz Jl e

rl

I

I1

~~~ ~ PS

m

H JO ~

CIµ I 'I

Figure 3 The diblocks we use are usually a plastic and a rubber as in the monomers and diblock shown here.

The polymers that we used were generally a plastic (such as polystyrene) and a rubber (such as polybutadiene or polyisoprene, PI), figure 3. For observation and pattern transfer we need some contrast between the two parts of the diblock. What proved convenient for many of our studies and for processing was to use the fact that the PB backbone had a double bond, while the PS did not. Reactions with the double bond were used both to stain and crosslink the PB and PI with Osmium tetroxide for SIMS and electron microscopy and to break the double bond and fragment the PI and PB by ozonation for processing. For AFM imaging the contrast was supplied by the different elastic properties (stiffness) of the rubbers and plastics. 1.2 IMAGING AND DEPTH PROFILING The basic idea of using a diblock copolymer for patterning a substrate is straightforward, put down a monolayer, use some contrast between the blocks as a mask and then etch through to the substrate (Harrison, 1998). However, each step required extensive new investigations not the least of which was figuring out what we had before the transfer took place.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

8

Chaikin et al.

1.2.1 Sample Preparation The diblocks we used were soluble in toluene and a conventional spin coater was used to cast a film on commercial Si wafers. The polymer concentration, volume and spin rate determine the thickness. Thickness measurements were accomplished largely by ellipsometry, interferometry and secondary ion mass spectrometry (SIMS). It is well known that ordered diblock films are quantized in thickness due to the discreet layer spacing (Coulon, 1990, Mansky, 1995b). If you spin coat a copolymer film at a thickness equal on average to 1.5 layers after annealing, the result is half the surface coated with a monolayer and the other half with two monolayers, usually in islands. For such thin films the layer thicknesses do not correspond to the bulk layer thickness since one or the other polymer preferentially wets the substrate and the air which bound the film. The cartoons which are shown in figure 2 are what we deduced from many experiments and then confirmed by performing SIMS with OsO, staining of the rubber component (Harrison, 1998). There have been several studies of the effects of surface treatment on which polymer wets the surface and the different morphologies which result (Ahagon, 1975, Jones, 1992, Hashimoto, 1992). With neutral surfaces both phases can wet and the result is cylindrical films (and lamellar films) which form perpendicular to the surface rather than lie flat (Huang, 1998, 2000, Harrison, 2000). A good bit of work went into understanding how the morphology of the films depend on their thickness (Henke, 1988, Anastasiadis, 1989, Mansky, 1995a, Park, 1997b, Radzilowski, 1996). For example, if we start with a sample with cylinders of PB in a PS matrix, a bulk layer spacing of 30 nm and PB wets the surface then: a 30 nm thickness film will not even phase separate, a 40 nm film will have a layer of PB spheres, and a 50 nm layer will have the desired cylindrical morphology cartooned in figure 2. The hand waving explanation is that the surface layers rob PB from the phase separated interior region and effectively reduce f A in the phase diagram of figure Id). 1.2.2 3D Imaging In our original studies of monolayer films we used TEM images made through a SiN window prepared on a Si wafer (Morkved, 1994, 1996, Park, 1997). This allowed us a projection view of the pattern over a region about 10 micron square. In order to further develop the science and technology necessary for this process we needed to observe the pattern over larger scale and on the actual surface that we hoped to decorate. The answer was to take SEM images of the monolayer with contrast supplied by the OS0 4 staining. Unfortunately, the wetting layers on these films were the rubbery component, which was the stained phase. SEM revealed only the uniform surface and none of the interior structure we were after. The surface had to be removed. The answer was to reactive ion etch to a depth where the microphase separated regions could be observed. A number of different gases and procedures were tried (Harrison, 1999). Some of our observation include: Ar gives a very rough etch of both PS and PB, Cl, destroys the microstruture, C1HF, and 02 give a rough etches of PB and PS, CF, gives smooth etches on PS, PB and PB with OsO, stain. The choice was the CF, at a rate of about 20 nm/minute. The etch was sufficiently smooth and nonpreferential that we could step through the

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanopatterning with Diblock Copolymers

9

monolayer and see the surface, the bulk continuous phase, the micelles etc. We would remove a slice, take a SEM image, remove another slice, take another i mage etc. What was particularly surprising was that the etch did not seem to produce additional roughness even for thick samples (Harrison, 1998a). With ten -twenty layers we could image the last layer with essentially the same flatness and resolution as the first (Harrison, 1998b). With experiments on trilayer islands, figure 4, it was possible to follow the structures and defects through three layers, to see that the patterns were interdigitated and to establish the depth resolution as - 7 nm (Harrison, 1997). This technique therefore allows three dimensional real space imaging of many soft materials and has been exploited by several groups. For example between successive RIE slice removal the imaging step can be done by AFM instead of SEM.

500 nm Figure 4 SEM images of a three layer island after RIE removal to reveal structures at different depths. Note the alternate white-black lines in the circle indicating interdigitation of disclination pattern.

1.3 PATTERN TRANSFER TECHNIQUES Having established that we could form monolayers of the appropriate hexagonal dots or parallel lines on a surface the next challenge was finding a way to transfer the patterns to a substrate. The obvious technique was RIE, but again the

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

10

Chaikin et al.

questions involved what gases would recognize the polymer as a mask and what to use for the contrast in the diblock pattern. 1.3.1 Basic Positive and Negative Techniques CF, was an ideal gas for uniformly etching PS and PB at essentially the same rate. It also provided a reasonable but rougher etch into Si and SiNx. Directly

SI;Fti--11

\III iII,

1 b) C}ra+n:uwi Sample

Silicon Nitride RIr (Cl

1,iticor`iiride

f'f,.i?-y

Silicon Nitride Rlf_

,

i ~ A L Al A } A +1 Silicon Nitride

Silicon Nitride

Silicon Nitride

Silica., Nil, ;.I. Kil

44

e#

h l

iH

Silicon -Nil ride

Silicon Nitride

Silicon Nitride

Hole

N'l;

Figure 5 Process flow chart for transferring positive or negative patterns from diblock monolayer to etched features in substrate.

etching with CF, would therefore provide no contrast or pattern transfer. The answer was to remove one of the components of the diblock and this was accomplished by again taking advantage of the double bonds in the rubbery phase. Ozone attacks these double bonds and leaves small polymer segments which are volatile and soluble. The easiest processing step to degrade the PB is to immerse the coated samples in a water bath with bubbling ozone (-5% ozone) (Lee, 1989). In the process the PS is also crosslinked. The result is a mask with voids in place of

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanopatterning with Diblock Copolymers

11

PS-Red

Patterned Silicon Nitride

Figure 6 SEM images of mask and transferred pattern.

PB and a different thickness of PS between the air and the substrate to be patterned. As illustrated in figure 5 a CF, etch can now transfer the pattern into the substrate. Since the etch rates of PS and Si are comparable, the aspect ratio of the holes transferred is -1. To make sure that polymer film is completely removed we perform an 0 2 RIE as a final cleaning step. In figure 6 we show an ozonated polymer mask and the pattern transferred by RIE into SiNx. In these SEM images the contrast comes from the height profile rather than staining. It is also possible to make a negative transfer, minority features in the diblock pattern become elevated regions on the patterned surface. In this case we use our conventional stain, OsO„ to decorate the rubbery regions. Since CF, etch PB with OsO, at about the same rate as PS, we add a mixture CF 4 /02 which preferentially etches PS over PB. (Cartoon in figure 5). The work in this section is largely found in Park, 1997a, Harrison, 1998c, 1998d, 1998e, 1999, 2001. 1.3.2 More Recent Developments, Multilevel, Multistep Processing Typically different materials require slight modifications of the basic techniques outlined above. Ge reacts unfavorably with the ozonation process and therefore a thin layer of SiNx is sputtered onto its surface before the diblock layer is spun down and processed. The RIE step is then increased to eat through the SiNx and into the Ge. A cross section of a Ge Film with etched holes is shown in figure 7. Here we see the periodic pattern on the surface with N aspect ratio of the holes about 1.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

12

Chaikin et al.

Figure 7 TEM's of mask and transferred pattern for Ge film. Note the protective layer of SiNx. Right side - cross-sectional TEM of a broken section illustrating that the aspect ratio of the etched holes is -1.

I'S al'I:

Rli: t(i,a )

4.

vI

ii iii

1. Ow -

Nitride

Suhstrate

Pulrimide satnirAft

2.

.

R1 II: l I: C11,) '

5.

Ozonated Polymer l Stlir:ua \ilride

We Suhclrair

3.

Substrate

-

I

6.

1~I s

RIF. 1r

Metal Deposition

sub

trate

+ r

IN I 1

e 7.

i i ft-off ~

Substrate

~

~

~

~

~

Suhstratc

Metal Dots Figure 8 Trilayer technique allows a polyimide layer to planarize rough surfaces. The undercut allows easy liftoff after metal depositions.

In order to use the diblock as an evaporation mask for metal deposition we needed to enhance the aspect ratio of the mask itself and to make an undercut of the mask itself. Directly evaporating onto a polymer mask with holes through to the surface produced samples with lift off problems. The metal on the substrate surface retained contact with the metal on the mask surface. A trilayer technique solved the problem (Park, 2000). The substrate is first covered with a 50 nm layer of polyimide by spin coating

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanopatterning with Diblock Copolymers

13

W

41.4

J,

0.7

Figure 9 AFM image of Au dots prepared by trilayer technique. Lattice constant -27 nm. then a SiNx layer is sputtered and finally the diblock film is spun coat. The process is shown schematically in figure 8. The positive resist techniques are followed to pattern the SiNx layer and then the RIE gas is switched to 0 2 which etches polyimide much more rapidly than SiNx and even leaves the desired undercut. The metal evaporation is then followed by dissolving the polyimide layer. An AFM image of Au dots produced by this technique is shown in figure 9. The trilayer technique is quite versatile since it allows the coating of most any surface. The Polyimide acts as a planarization layer to flatten rough surfaces. It therefore can be used to decorate a previously processed surface and create three dimensional structures. For example we could deposit a set of metal wires then apply the polyimide layer and SiNx etc. to evaporate a cross set of wires. In other applications we could use the large aspect ratio polyimide mask to electroplate or grow materials through the mask, figure 40, to make e.g. DNA mazes, Volkmuth, 1994.

Lift-off

Electroplating

Replication

jj_~ J IM - F - 1 OF i '

Substrate

Metal Deposition

Substrate

Suhslrule

Metal Growth

sub."nue Matrix Remo~• a 1 Lift-off a ~ a a a w. substrate Metal Dots

Substrate Metal Posts

Negative Replica

Figure 10 Other potential uses of trilayer technique.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

14

Chaikin et al.

Another quite successful application of the diblock lithography is in the growth of dense arrays of quantum dots (Li, 2001). In this case the compound semiconductor is covered with SiNx and the diblock. Processing progresses to the stage where there are holes in the SiNx layer. Growth of the semiconductor through the holes in the mask is accomplished by MOCVD and the mask is then removed by a wet etch. The process is schematized in figure 11 and an AFM i mage of GaAs quantum dots produced by this technique is shown in figure 12. Further TEM analysis of cross-section shows that the dots are epitaxial with the substrate. polymer SiN x (15 run)

00000 GeAs

GaAs

Figure 11 Processing for MOCVD selective area growth of GaAs quantum dots.

1.4 ORDERING, ANALYSIS AND CONTROL OF NANOPATTERNS From the mask and transferred patterns it is clear that we can readily cover large areas with periodic lines and hexagonally arranged dots. However, the ordered regions are of finite extent. Perfect crystallites had a characteristic size of 20-50 periods (squared) in our early studies. For many applications, filters, quantum dots for diode lasers, etc. the size monodispersity and the density are of paramount importance. For other applications, such as conventional memory storage, we would like an addressable array with long range order covering the entire surface. (We should note that there are schemes for complete addressability do not require perfect long range order and that there are also applications such as magnetic disk recording which benefit from ordered regions on the scale of -100 microns rather then centimeters.) Scientifically we were also interested in how perfect and long range the ordering could be made. We therefore decided to devote a considerable effort to understanding the annealing or coarsening problem and to see whether the patterns could be aligned or registered with larger scale features Harrison, 2000b, Segalman, 2001. There have been similar attempts and successes by other groups (Morkved, 1996, Kramer, 2001). The two dimensional patterns we are interested in exhibit varying degrees of both orientational and translational order. A crystalline lattice is

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanopatterning with Diblock Copolymers

15

tapping-mode atomic force microscopy TMAFM)

70

I IY1 I I I I

diameter: 23 ± 3 nm overall height: 14 ± 2 nm

0 0

2

k

8 1P 12 14 1

,

18 20 22 2k 21 28 00

Size (nm) Figure 12 GaAs quantum dots from copolymer masks.

characterized by a set of order parameters which correspond to the amplitude and phase of density waves at all of the reciprocal lattice vectors. To consider the simplest system first we focussed on the cylindrical phase which forms the fingerprint like patterns with only two degrees of broken symmetry, one orientational and one translational (one periodicity). The symmetry of this phase is the same as for smectic liquid crystals which are layered in three dimensions or striped in two dimensions.

466 K Figure 13 Cylindrical phase diblock copolymer monolayers after one hour annealing at the indicated temperatures. The size of the correlation length is indicated by the red bar. Solid circles are +1/2 disclinations and dotted circles are- 1/2 disclinations.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Chaikin et al.

16

1.4.1 Correlation Functions, Fingerprints, Disclinations In figure 13 we show two SEM images of the monolayer cylindrical phase after annealing for one hour at 395 K and 466 K. It is clear that the higher temperature anneal yields a more ordered structure. In order to make this quantitative we can define an orientation which is aligned with the stripes. As for a director field for a nematic liquid crystal the orientation is ambiguous with respect to sign or rotation by 180 degrees. It is therefore convenient to use an order parameter of the form s(x) = s0e2 i° which is the same for 0 and 0 + tt. Computationally the images are digitized and the orientation taken from gradients of the intensity. We can then calculate the orientational correlation function and fit it to a simple exponential to obtain the correlation length ~. In figure 13 we show the time dependence of the correlation function for different annealing temperatures. Since the SEM measurement is destructive the data were taken by making a master wafer, breaking it into pieces and removing pieces successively from the annealing furnace for analysis. It is clear that higher temperatures yield longer correlation lengths and that the annealing process is quite slow at long times. The actual values of the correlation length correspond to what we observe in the i mages as a "grain size" or the distance at which the orientation changes by about 90 degrees. The correlation function alone does not give us a real clue as to the annealing or coarsening dynamics. Striped patterns, like the cylindrical phase monolayers in figure 13 occur many places in nature, for us most commonly in the form of fingerprints. There is a large literature on fingerprints, mainly for identification, but there is also some work by physicists. In particular there is a paper by R. Penrose (Penrose, 1976) (which refers to the work of his father L. Penrose who did analysis of fingerprints (Penrose, 1966)) which points out the importance of the defects, loops and triradii, figure 15, which occur in fingerprints or ridged or striped phases. (Fingerprints are not completely understood, they are not genetic as evidenced by the fact that identical twins have different fingerprints.) In directed systems like nematic liquid crystals the defects are known as +1/2 and -1/2 disclinations (Kleman, 1983, Chaikin, 1995, de Gennes, 1993). For a vector field the defects are vortices with a winding number of 2n since a

10'

m E C

q

0,

10 10' Annealing Time (seconds)

0°

10'

10 10' Annealing Time (seconds)

0°

Figure 14 Correlation length and square root of inverse disclination density (separation of disclinations) for 396K and 435K annealing temperatures.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanopatterning with Diblock Copolymers

17

continuous path around a defect must rotate the vector by tic (times an integer) to match up with the starting direction. For the rods, or stripes that make up a nematic phase the winding number can be +/- rt to end up with the same orientation. Penrose pointed out that fingerprints cannot be the result of a potential or force which would act on a vector field and only allow vortex like defects. The disclinations and hence the fingerprints must be produced by a tensorial field such as a stress. The observation that disclinations and stresses are important in striped phases is important for our studies as well. First it is clear that loops and triradii are present in our patterns (figure 13). Moreover, the density decreases as the order increases. In fact when we measured the disclination density p we found that it's reciprocal root (p)-t/2 tracks the time and temperature dependence of the 112 , correlation length. In fact, (p)_ the average distance between disclinations is essentially the correlation length. This is also clear from figure 13 - compare the distance between two disclinations to the distance that it takes to "bend" the stripes by 90 degrees. What makes this interesting is that the force between disclinations is known and can be used to find the coarsening law. In the far field, several core diameters away from the disclination, and neglecting the anisotropy to find the distance dependence of the force, we can take a path around the topological defect of length 27tR. Since we must accumulate an orientation change of 180 degrees over this path independent of its length, the local strain field will vary as 1/R and the strain is proportional to the stress as is the force exerted on another disclination at this distance. If we have opposite disclinations they will attract like 1/R and the response will be that they move toward one another viscously. dR/dt « gF « WR, RdR « dt, R 2 « t, R « t' /2 ( t is a mobility). a So the distance between the disclination collapses as R t t12 . Or after time t all disclinations that were separated by R have annihilated and only ones with further separation have survived. The density of disclinations then goes down as p « 1/R2 « 1/V and the correlation length increases as ~ a t t . In fact this relationship between disclination density and correlation length and the power law has been well known and experimentally tested for nematic liquid crystals. How does it work for our system? A log-log plot of disclination density and correlation length as a function of time are shown in figure 15. While the

Figure 15 Top - human fingerprint with loop and triradius, corresponding to the topological defects: left) +1/2 disclination and right) -1/2 disclination.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

18

Chaikin et al.

correlation length indeed seems to remain the distance between disclinations t"4 2 (p « ~ ) and they obey a power law, the power law is best fit by rather than t t12 . 1.4.2 Time Lapse AFM Videos and Disclination Dynamics There have been studies of the dynamics of striped systems as well as analysis of their defect structure but usually in relation to different problems. In terms of dynamical systems fluids heated from below have been of interest for at least a century. The first instability of such a system is the formation of convective rolls carrying heat from the bottom to the top of the liquid layer. When viewed from the top the rolls form striped patterns which are dynamic and coarsen with time (at intermediate temperature differences between top and bottom, at higher differences there are further instabilities and the system goes chaotic and turbulent). This Raleigh-Benard convection problem has attracted much attention as a way of understanding pattern formation (Hou, 1997, Cross, 1995, Elder, 1992, Christensen, 1998). Along with the experiments are many theoretical treatments and especially simulations on models, such as the Swift-Hohenberg model (Cross, 1995), which hope to capture the essence of the problem. Most of the work finds that the patterns coarsen with a power law 4 « t 14 to t"S depending on the magnitude of the noise term used in the simulations. These values are similar to what we have observed. But there is no general conclusion as to what is the mechanism for this coarsening. Moreover most analytical work using several different mechanisms tends to give t 12 .

Figure 16 Cartoon of the configuration used for making time lapse AFM movies of diblock annealing. The outer blue phase is rubbery and the AFM contrast is in the elastic response.

Rather than run simulations, we decided that insight into this t/4 exponent would come only from monitoring the motion of disclinations in an experimental realization. Since our conventional imaging technique (SEM/RIE) vitrified and destroyed the sample, we were forced to develop an alternative imaging technique which was not destructive. If we invert our system so that the continuous phase is the rubber, which also preferentially wets the wafer and the air, then we can use the "tapping" mode of an AFM to sense the elastic response of the film (Hahm, 1998, figure 16. If there is a hard plastic region under the rubber surface below the tip then we will find a contrast with the region with no plastic. It is like feeling the pea under a soft pillow. The next problem is that there is no mobility at room temperature and annealing only occurs upon heating.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanopatterning with Diblock Copolymers

19

However, repeated heating in air degrades the double bond in the rubber we had been using. We therefore chemically treated the polyisoprene to saturate all of the bonds and make it polyethylpropylene (PEP) which remains rubbery but better resists degradation. We then cast the polymer on a wafer and set a piece of the wafer on a hot stage in an AFM. The movies are made by heating to 100C for one minute, cooling to room temperature taking an AFM tapping mode image, reheating, recentering the image against its small shifts, taking another image etc.

Figure 17 Quadrupole annihilation. A series of time lapse AFM images which show the most common disclination annihilation process. Sequence is across and down. We see the disclinations attract and annihilate with the creation of several dislocations which then repel. Each frame is 3 micron base.

1.4.3 Multiple Disclination Annihilation The movies are quite spectacular and show the decrease in disclinations and the growth of the correlation length. After staring at the movies for some time, and especially on running them backward it becomes clear that the dominant process which leads to ordering is the annihilation of multiple disclinations, not just plus-

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

20

Chaikin et al.

minus pairs. Most often it is quadrupoles, but occasionally triplets and fives, almost never two. A typical series of pictures is shown in figure 17. Here we see a characteristic -_+ quadrupole annihilation. Once it is clear what is going on it is clear why the original dipole annihilation doesn't work. The system we are dealing with has both orientational and translational (periodic) order, it is striped or smectic, not nematic. In figure 18 we see that in order to have two disclination move toward one another, we have to break the stripes or create dislocations (Yurke, 1993, Liu, 1997), the defect associated with periodic order. In fact for every lattice step that the disclinations come together we have to introduce two dislocations. This is not energetically favorable. If, on the other hand, we can absorb the dislocation by moving other disclinations, for example in the collapse of the opposite pair in a quadrupole annihilation, then we can get rid of disclinations without creating an excess of dislocations. We can even work out the scaling for the coarsening law. Take a characteristic length of R for the distance between all disclinations in a quadrupole. The dislocation has to move a distance R from the + disclination to the other + disclination in order for the + to move one step, dr, toward the disclination. Since the force on the disclinations is 1/R the force on the dislocation is 1/R 2. The dislocation has to move (viscously) a distance R at 2 velocity 1/R which takes time dt « R' The velocity of the disclination dr/dt is then proportional to 1/R' which integrates to R' « t, R « t t/4 . This explains the power law that we and others studying striped phases have seen for decades. Note that it doesn't really rely on 4 disclinations, rather on the fact that we are exchanging dislocations between disclinations over a distance R. Quadrupole Annihilation r

Collapse quadrupole by or (unit displace ent), energy decrease s by AE- 11r

_dr

Dislocations move a distance - r

Q0

2

2

Force on dislocations, f Al 1/r ==> v-1/r ti me for dislocation to go a distance dt- r/v-r 3

dr/dt - 1/r 3 ===> r - t,

r --t 1 !4

Figure 18 Cartoon of the quadrupole annihilation process, clockwise from upper left. Note in the second frame that for a disclination to advance one lattice constant requires the creation of two dislocations. To prevent the proliferation of dislocations from disclination annihilations there must be other disclinations present to absorb the dislocations.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanopatterning with Diblock Copolymers

21

Finally we should mention our annealing studies of the hexagonal phase, the two dimensional crystal. We have made time lapse movies and found that the coarsening law is similar. Unfortunately the explanation is not yet so clear. We see hardly any free disclinations. Rather the disclinations are tightly bound in dislocations and the dislocations form strings or low angle grain boundaries. However, we have not yet found a satisfactory way to define grains, largely because the strings of dislocations do not tend to form closed regions. It is more like a system of interacting strings. However, since chains of dislocations can also look like two disclinations at the chain ends, we may be getting back to the same explanation as for the stripes (or we may be going in circles).

Figure 19 AFM image of Annealing of grains in a monolayer film of hexagonal spheres. There is a step edge of -30 nm height from top to bottom on the right side. The step edge registers the first row of spheres and leads to an aligned crystallite. Further annealing leads to complete alignment of the area shown with the step edge. (Note that the area to the right of the edge is not polymer covered.) We have also had some success at aligning the patterns. A step edge of about the same height as the monolayer nicely aligns the first row of spheres in the hexagonal phase, figure 19. Upon annealing this aligned edge serves as the growth point which completely aligns regions of several microns. More

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

22

Chaikin et al.

spectacularly, Prof. Steve Chou's group at Princeton have shown that directly pressing on the diblock monolayer while it is heated can completely align the film over centimeter distances.

CONCLUSION We have demonstrated that the self-assembly of diblock copolymers can serve as a useful and often unique technique for forming dense nanostructured arrays over large areas. We have extended our initial patterning to many different processes which now allow work on metal, semiconductor and insulator substrates and etching, growth and evaporation. We have also made progress in registration and alignment of the patterns so that they can we addressed. And particularly fortuitously we have found that they are very interesting systems on their own for fundamental research into the ordering and dynamics of two dimensional systems.

ACKNOWLEDGEMENT We greatly acknowledge support from NASA and from NSF DMR9809483.

REFERENCES Ahagon A. and Gent A. N., 1915, J. Polym. Sci., Polym. Phys. Ed. 13, 1285. Anastasiadis S. H., Russell T. P., Satija S. K., Majkrzak C. F., 1989, Phys. Rev. Lett. 62, 1852. Bates F. S., 1991, Science 251, 898. Bates F. S. and Fredrickson G. H., 1990, Annu. Rev. Phys. Chem. 41, 525. Chaikin P. M. and Lubensky T. C., 1995, Principles of condensed matter physics, Cambridge University Press (Cambridge, 1995). Christensen J. J. and Bray A. J., 1998, Phys. Rev. E 58, 5364. Coulon G., Ausserre D. and Russell T. P., 1990, J. Phys. (Paris) 51, 777. Cross M. C. and Meiron D. I., 1995, Phys. Rev. Lett. 75, 2152. de Gennes P.-G., 1979, Scaling Concepts in Polymers (Cornell University Press, Ithaca), p. 103. de Gennes P. G. and Prost J., 1993, The Physics of Liquid Crystals (Oxford Science Publications, New York, ed. 2. Elder K. R., Vinals J. and Grant M., 1992, Phys. Rev. Lett. 68, 3024. Flory P. J., 1953, Principles of Polymer Chemistry (Cornell University Press, 1953), Ch. XII. Hahm J., Lopes W. A., Jaeger H. M. and Sibener S. J., 1998, J. Chem. Phys. 109, 10111. Harrison C., Park M., Chaikin P., Register R. A., Adamson D. H. and Yao N., 1998a, Macromolecules 31, 2185. Harrison C., Park M., Chaikin P. M., Register R. A., Adamson D. H. and Yao N., 1998b, Polymer 39, 2733-2744.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanopatterning with Diblock Copolymers

23

Harrison C., Park M., Chaikin P. M., Register R. and Adamson D., 1998c, Polymeric Materials for Micro- and Nanopatterning Science and Technology (ACS Symposium Series), H. Ito, E. Reichmanis, O. Nalamasu and T. Ueno, eds. (Washington: American Chemical Society). Harrison C., Park M., Register R. A., Adamson D. H. and Chaikin P. M., 1998d, J. Vacuum Sci. Technol. B 16, 544-552. Harrison C., Park M., Chaikin P. M., Register R. A., Adamson D. H., 1998e, Macromolecules 31, 2185. Harrison C., 1999, Ph.D. thesis. Princeton University. Harrison C., Chaikin P. M., Huse D., Register R. A., Adamson D. H., Daniel A., Huang E., Mansky P., Russell T. P., Hawker C. J., Egolf D. A., Melnikov I. V. and Bodenschatz E., 2000a, Macromolecules 33, 857-865. Harrison C., Adamson D. H., Cheng Z., Sebastian J. M., Sethuraman S., Huse D. A., Register R. A. and Chaikin P. M., 2000b, Science 290, 1558-1560. Harrison C., Chaikin P. M. and Register R. A., 2001 Book Chapter "Block Copolymer Templates for Nanolithography", section 5.5.25 in Encyclopedia of Materials: Science and Technology, K. H. J. Buschow, R. W. Cahn, M. C. Flemings, B. Ilschner, E. J. Kramer and S. Mahajan, eds. (New York: Pergamon Press). Hashimoto T. and Hasegawa H., 1992, Polymer 33, 475. Hofstadter D., 1976, Phys. Rev. B 14, 2239. Henkee C. S., Thomas E. L. and Fetters L. J., 1988, J. Mater. Sci. 23, 1685. Hou Q. and Goldenfeld N., 1997, Physica A 239, 219. Huang E., Russell T. P., Harrison C., Chaikin P. M., Register R. A., Hawker C. J. and Mays J., 1998, Macromolecules 31, 7641-7650. Huang E., Mansky P., Russell T. P., Harrison C., Chaikin P. M., Register R. A., Hawker C. J. and Mays J., 2000, Macromolecules 33, 80-88. Jones R. A. L., Norton L. J., Shull K. R., Kramer E. J., Felcher G. P., Karim A. and Fetters L. J., 1992, Macromolecules 25, 2539. Kleman M., 1983, Points, Lines and Walls ( Wiley, New York, 1983). Lee J. S., Hirao A. and Nakahama S., 1989, Macromolecules 22, 2602. Leibler L., 1980, Macromolecules 13, 1602. Li R. R., Dapkus P. D., Thompson M. E., Jeong W. G., Harrison C., Chaikin P. M., Register R. A. and Adamson D. H., 2000, Appl. Phys. Lett. 76: 13, 1689-1691. Liu C. and Muthukumar M., 1997, J. Chem. Phys. 106, 7822. Mansky P. and Russell, T. P., 1995, Macromolecules 28, 8092. Mansky P., Chaikin P. M. and Thomas E. L., 1995, J. Mat. Sci. 30, 1987. Mansky P., Harrison C. K., Chaikin P. M., Register R. A. and Yao N., 1996, Appl. Phys. Lett. 68, 2586. Morkved L., Wiltzius P., Jaeger H. M., Grier D. G. and Witten T. A., 1994, Appl. Phys. Lett. 64, 422. Morkved L., Lu M., Urbas A. M., Ehrichs E. E., Jaeger H. M., Mansky P. and Russell T. P., 1996, Science 273, 931. Morton M. and Fetters L. J., 1975, Rubber Chem. Technol 48, 359.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

24

Chaikin et al.

Park M., Chaikin P. M., Register R. A. and Adamson D. H., 2001, Appl. Phys. Lett. 79, 257-259. Park Miri, Harrison C., Chaikin P. M., Register R. A. and Adamson D. H., 1997a, Science 276, 1401. Park M., Harrison C., Chaikin P. M., Register R. A., Adamson D. H. and Yao N., 1997, Volume 461: Morphological Control in Multiphase Polymer Mixtures, Materials Research Society, Boston, MA, Dec 2-6, 1996; Materials Research Society: Pittsburgh, PA, 1997b; pp. 179-184. Penrose L. S., 1965, Nature. Penrose R., 1979, Ann. Hum. Genet, London, 42, 435. Radzilowski L. H. and Thomas, E. L., 1996, J. Polym. Sci. B: Polym. Phys. 34, 3081. Segalman R. A., Yokoyama H. and Kramer E. J., 2001, Adv. Mater. 13, 1152. Thouless D. J., Kohmoto M., Nightingale M. P. and den Nijs M., 1982, Phys. Rev. Lett. 49, 405. Volkmuth W. D., Duke T., Wu M. C., Austin R. H. and Szabo A., 1994, Phys. Rev. Lett. 72, 2117. Yurke B., Pargellis A. N., Kovacs T. and Huse D. A., 1993, Phys. Rev. E 47, 1525.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

2

Nanostructured Materials: Basic Concepts, Microstructure and Properties H. Gleiter Forschungszentrum Karlsruhe, Institut fur Nanotechnologie, Postfach 36 40, D-76021 Karlsruhe, Germany

1. INTRODUCTION Nanostructured Materials (NsM) are materials with a microstructure the characteristic length scale of which is on the order of a few (typically 1-10) nanometers. NsM may be in or far away from thermodynamic equilibrium. NsM synthesized by supramolecular chemistry are examples of NsM in thermodynamic equilibrium. NsM consisting of nanometer-sized crystallites (e.g. of Au or NaCI) with different crystallographic orientations and/or chemical compositions are far away from thermodynamic equilibrium. The properties of NsM deviate from those of single crystals (or coarse-grained polycrystals) and/or glasses with the same average chemical composition. This deviation results from the reduced size and/or dimensionality of the nanometer-sized crystallites as well as from the numerous interfaces between adjacent crystallites. An attempt is made to summarize the basic physical concepts and the microstructural features of equilibrium and nonequilibrium NsM. Concerning the properties of NsM three examples (the diffusivity, the plasticity and the ferromagnetic properties) will be considered. In fact, the properties of NsM will be shown to deviate significantly from those of the corresponding coarse-grained polycrystals if a characteristic length and/or energy scale of the microstructure of NsM (e.g. the grain size, the excess boundary energy, etc.) is comparable to a characteristic length and/or an energy scale controlling a property of a NsM (e.g. the thickness of a ferromagnetic domain wall or an activation energy).

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

26

Gleiter

2. BASIC CONCEPTS One of the very basic results of the physics and chemistry of solids in the insight that most properties of solids depend on the microstructure, i.e. the chemical composition, the arrangement of the atoms (the atomic structure) and the size of a solid in one, two or three dimensions. In other words, if one changes one or several of these parameters, the properties of a solid vary. The most well-known example of the correlation between the atomic structure and the properties of a bulk material is probably the spectacular variation in the hardness of carbon when it transforms from diamond to graphite or vice versa. The synthesis of materials and/or devices with new properties by means of the controlled manipulation of their microstructure on the atomic level has become an emerging interdisciplinary field based on solid state physics, chemistry, biology and material science. The materials and/or devices involved may be divided into the following three categories Gleiter (1995). The first category comprises materials and/or devices with reduced dimensions and/or dimensionality in the form of (isolated, substrate-support or embedded) nanometer-sized particles, thin wires or thin films. The second category comprises materials and/or devices in which the nanometer-sized microstructure is limited to a thin (nanometer-sized) surface region of a bulk material. PVD, CVD, ion implantation and laser beam treatments are the most widely applied procedures to modify the chemical composition and/or atomic structure of solid surfaces on a nanometer scale. Surfaces with enhanced corrosion resistance, hardness, wear resistance or protective coatings (e.g. by diamond) are examples taken from today's technology. In this paper we shall focus attention on the third category of bulk solids with a nanometer-scale microstructure. In fact, we shall focus on bulk solids in which the chemical composition, the atomic arrangement and/or the size of the building blocks (e.g. crystallites or atomic/molecular groups) forming the solid vary on a length scale of a few nanometers throughout the bulk. Two classes of such solids may be distinguished. In the first class, the atomic structure and/or the chemical composition varies in space continuously throughout the solid on an atomic scale. Glasses, gels, supersaturated solid solutions or some of the implanted materials are examples of this type. In the last two decades a second class of materials with a nanometer-sized microstructure has been synthesized and studied. These materials are assembled of nanometer-sized buildings blocks - mostly crystallites - as displayed in Fig. 1. These building blocks may differ in their atomic structure, their crystallographic orientation and/or their chemical composition (Fig. 2). In other words, materials assembled of nanometer-sized building blocks are microstructurally heterogeneous consisting of the building blocks (e.g. crystallites) and the regions between adjacent building blocks (e.g. grain boundaries). It is this inherently heterogeneous structure on a nanometer scale that is crucial for many of their properties and distinguishes them from glasses, gels, etc. that are microstructurally homogeneous. Materials with this kind of a nanometer-sized microstructure are called "Nanostructured Materials" or - synonymously - nanophase materials, nanocrystalline materials or supramolecular solids. In this paper we shall focus on these "Nanostructured Materials" (NsM).

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Figure 1 Schematic, two dimensional model of one kind of nanocrystalline material.

Figure 2 Synthesis of nanomaterials with different chemical microstructures by the consolidation of small, pre-fabricated, isolated nm-sized crystal. (a) All atoms have idendical, chemical composition. (b) The free surfaces of the nm-sized crystals are coated with atoms that differ chemically from the core resulting in a NsM with boundaries that are chemically different from the crystalline regions open and full circles. (c) Nm-sized crystals with different chemical compositions resulting in a nanocomposit.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

28

Gleiter

3. SYNTHESIS OF NANOSTRUCTURED MATERIALS (NSM) The methods deviced for the synthesis of NsM may be divided in the following two groups. Top-down synthesis routes. This approach involves the assembly of NsM from pre-fabricated or pre-existing structural elements (e.g. pre-fabricated nm-sized crystals, supramolecular units, etc.). These elements or building blocks are assembled into a bulk solid with a nm-scale microstructure. -

The bottom-up synthesis starts from individual atoms/molecules and assembles them into a bulk piece of material. Evaporation onto a cold substrate or crystallization from the glassy state are examples of this route of synthesis.

3.1 Top-down Synthesis of NsM One frequently used top-down route for the synthesis of nanocrystalline materials involves a two-step procedure. In the first step, isolated nanometer-sized crystallites are generated which are subsequently consolidated into solid materials. PVD, CVD, electrochemical, hydrothermal, thermolytic, pyrolytic decomposition and precipitation from solution have been used so far. The most widely applied PVD method involves inert gas condensation. Here, the material is evaporated in an inert gas atmosphere (most frequently helium at a pressure of about 1 kPa) is used. The evaporated atoms transfer their thermal energy to the (cold) helium and hence, condense in the form of small crystals in the region above the vapor source. These crystals are collected and consolidated into a bulk NsM. Instead of evaporating the material into an inert gas atmosphere, bulk nanocrystalline materials may also be obtained by depositing the material in the form of a nanometer-sized polycrystalline layer onto a suitable substrate. The methods for generating small crystallites by precipitation reactions may be divided into processes involving precipitation in nanoporous host materials and host-free precipitation. In both cases a wide range of solvents (e.g. water, alcohol, etc.) as well as different reactions (e.g. addition of complex forming ions, photochemical processes, hydrolytic reaction, etc.) have been utilized. A widely applied method for generating nanometer-sized composites is based on the sol-gel process. An interesting subgroup of sol-gel generated nanocomposits are organicinorganic nanoscale ceramics, so called ceramers, polycerms of ormocers (Schmidt, 1992). Following the ideas of Mark and Wilkes (Garrido et al., 1990), these materials are prepared by dissolving pre-formed polymers in sol-gel precursor solutions, and then allowing the tetraalkyl orthosilicates to hydrolyze and condense to form glassy Si02 phases of different morphological structures. Alternatively, both the organic and inorganic phases can be simultaneously generated through the synchronous polymerization of the organic monomer and the sol-gel precursors. The main advantages of producing nanocrystalline materials by a two-step procedure (involving the generation of isolated nanometer-sized crystals followed by a consolidation process) are as follows (Fig. 2): (i) Crystals with different chemical compositions can be co-generated, leading to "alloying" on a

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanostructured Materials: Basic Concepts and Properties

29

nanometer-scale. (ii) The free surfaces of the small crystals may be coated prior to the compaction process by evaporation, sputtering, chemical reaction (e.g. by surface oxidation) or in suspension. (iii) The interior of the crystallites may be modified by ion implantation before consolidation. Due to the small crystal size, the implantation results in materials that have the same chemical composition throughout the volume. In bulk materials, ion implantation is limited to surface regions.

3.2 Bottom-up Synthesis of NsM 3.2.1 Synthesis of NsM from glasses or sols In principle the following two routes have been used so far to generate nanocrystalline materials by means of bottom-up synthesis method. The first method to be discussed here starts from a noncrystalline structure, e.g. a glass. The nanocrystalline materials are obtained by nucleating numerous crystallites in the glass e.g. by annealing. These nuclei subsequently grow together and result in a nano-crystalline material (Fig. 3). The various modifications of this approach differ primarily in the starting material used. So far metallic glasses (e.g. produced by melt spinning, Lu et al., 1991) and (Chakravorty, 1992) have been successfully applied. The most important advantages of this approach are as follows. Low-cost mass production is possible and the material obtained exhibits little or no porosity. Obviously this approach is limited to chemical compositions which permit the generation of glasses or sols.

Figure 3 Synthesis of a nanocrystalline material (right figure) by crystallization from the glass (left).

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

30

Gleiter

3.2.2 Self-organized 1 nanostructured arrays A modified Stranski-Krastanov growth mechanism has been noticed to result in self-organized (periodic) arrays of nanometer-sized crystallites. If a thin InGaAs layer is grown on a AIGaAs substrate, the InGaAs layer disintegrates into small islands once it is thicker than a critical value (Notzel et al., 1995). These islands are spontaneously overgrown by a AIGaAs layer so that nanometer-sized InGaAs crystals buried in AnGaAs result. The observations reported indicate that the size, morphology and the periodic arrangement of the buried islands are driven by a reduction in the total free energy of the system (Pohl et al., 1999).

3.2.3 Polymeric nanostructured materials So far, the considerations have been limited to elemental or low molecular weight NsM, i.e. NsM formed by atoms/molecules that are more or less spherical in shape. A different situation arises if NsM are synthesized from high molecular weight polymers, i.e. long, flexible molecular chains.

Figure 4 Molecular folding in semicrystalline polymers resulting in stacks of lamellar crystals with a thickness of about 10-20 nm separated by "amorphous" regions.

It is one of the remarkable features of semicrystalline polymers that a nanostructured morphology is always formed (Fig. 4) if these polymers are crystallized from the melt or from solution, unless crystallization occurs under high pressure or if high pressure annealing is applied subsequent to crystallization. The disordered interfacial regions between neighboring crystals (Fig. 4) consist of macromolecules folding back into the same crystal and/or of tie molecules that meander between neighboring crystals. The typical thickness of the crystal lamellae are of the order of 10-20 nm. These relatively small crystal I

In this paper the term self-organization is used for dynamic multistable systems generating, spontaneously, a well-defined functional microstructure. It covers systems exhibiting spontaneous emergence of order in either space and/or time and also includes dissipative structures such as nonlinear chemical processes, energy flow, etc. Systems are called self-assembled in the spontaneously created structure is in equilibrium (Landauer, 1987; Haken, 1978 and 1994; Nocolis and Prigogine, 1977).

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanostructured Materials: Basic Concepts and Properties

31

thickness have been interpreted in terms of a higher nucleation rate of chainfolded crystals relative to extended-chain crystals or in terms of a frozen-in equilibrium structure: at the crystallization temperature, the excess entropy associated with the chain folds may reduce the Gibbs free energy of the chainfolded crystal below that of the extended-chain crystal. Chain folding may lead to rather complex nanometer-sized microstructures, depending on the crystallization conditions. Spherulites consisting of radially arranged twisted lamellae are preferred in unstrained melts. However, if the melt is strained during solidification, different morphologies may result, depending on the strain rate and the crystallization temperature (i.e. the undercooling). High crystallization temperatures and small strain rates favor a stacked lamellae morphology (Fig. 5a), high temperatures combined with high strain rates result in needle-like arrangements (Fig. 5b). Low temperatures and high strain rates lead to oriented micellar structures (Fig. 5c). The transition between these morphologies is continuous and mixtures of them may also be obtained under suitable conditions (Fig. 5d). The way to an additional variety of nanostructured morphologies was opened when multicomponent polymer systems, so-called polymer blends, were prepared. Polymer blends usually do not form specially homogeneous solid solutions but separate on length scales ranging from a few nanometers to many micrometers. The following types of nanostructured morphologies of polymer blends are formed in blends made up of one crystallizable and one amorphous (non-crystallizable) component: (I) The spherulites of the crystallizable component grow in a matrix consisting mainly of the non-crystallizable polymer. (II) The non-crystallizable component may be incorporated into the interlamellar regions of the spherulites of the crystallizable polymer. (III) The non-crystallizable component may be included within the spherulites of the crystallizable polymer forming domains having dimensions larger than the interlamellar spacing. For blends of two crystallizable components, the four most common morphologies are: (I) Crystals of the two components are dispersed in an amorphous matrix. (II) One component crystallizes in a spherulitic morphology while the other crystallizes in a simpler mode, e.g. in the form of stacked crystals. (III) Both components exhibit a separate spherulitic structure. (IV) The two components crystallize simultaneously resulting in so-called mixed spherulites, which contain lammelae of both polymers. Block copolymers constitute a class of self-organized nanostructured materials. The macromolecules of a block copolymer consist of two or more chemically different sections that may be periodically or randomly arranged along the central backbone of the macromolecules and/or in the form of side branches. As an example of the various self-organized nanostructured morphologies possible in such systems, Fig. 6 displays the morphologies formed in the system polystyrene/polybutadiene as a function of the relative polystyrene fraction. The large variety of nanostructured morphologies that may be obtained in polymers depending on the crystallization conditions and the chemical structure of the macromolecules causes the properties of polymers to vary dramatically depending on the processing conditions. NsM formed by block copolymers seem to represent (metastable) equilibrium structures despite the high excess energy stored in the interfaces between the structural constituents. The formation of these interfaces results from the local accumulation of the compatible segments of the macromolecules.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

a) a)

b)

c)

d)

I

shish kebab structure

I

Figure 5 a) Stacked lamellar morphology in polyethylene (TEM bright field). b) Needle-like morphology in polybutne-1 (TEM bright fiComputer-Aided Multivariate AnalysisComputer-Aided Multivariate An terephthalate (TEM dark field) phology in polyethylene terephthalate (TEM dark field micrograph). d) Shish-kebab morphology in isotactic polystyrene (TEM dark field micrograph) (Petermann,1991).

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanostructured Materials: Basic Concepts and Properties

33

Fraction of polystyrene blocks Figure 6 Electron micrographs of the morphologies of a copolymer consisting of polystyrene and polybutadiene blocks, as a function of the fraction of polystrene blocks. The spacial arrangements of the polystyrene and polybutadiene in the solidified polymer are indicated in the drawings above the micrographs (Petermann, 1991).

• two-dimensional units

• two-dimensional

•

P.

rigid rods

a r

0 • three-dimensional

I

a

N

a

0 t.

.+

• three-dimensional units

Figure 7 Schematic diagram indicating some of the (many) possible nanometer-sized molecular structures to be synthesized by supramolecular polymer chemistry (Lehn, 1993).

3.2.4 Supramolecular self-assembled structures Supramolecules are oligomolecular species that result from the intermolecular association of a few components (receptors and substrates) following an inherent assembling pattern based on the principles of molecular recognition.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

34

Gleiter

Supramolecular self-assembly' concerns the spontaneous association of either a few or a large number of components resulting in the generation of either discrete oligomolecular supermolecules or of extended polymolecular assemblies or of extended polymolecular assemblies such as molecular layers, films, membranes, etc. Self-assembly seems to open the way to nanostructures, organized and functional species of nanometer-sized dimensions that bridge the gap between molecular events and macroscopic features of bulk materials. For a detailed discussion of this development and of future perspectives, we refer to the review by Lehn (1995). This paper will be limited to outline only those aspects of the field (Lehn, 1993, 1995 and 1997) that are directly related to the synthesis of NsM. 3.2.5 Self-assembled organic architectures Self-assembly of organic architectures utilizes the following types of interaction between the components involved: electrostatic interaction, hydrogen bonding, Van der Waals or donor-acceptor effects. Self-assembly by hydrogen bonding leads to two- or three-dimensional molecular architectures which often have a typical length scale of a few nanometers. The self-assembly of structures of this type requires the presence of hydrogen-bonding subunits, the disposition of which determines the topology of the architecture. Ribbon, tape, rosette, cage-like and tubular morphologies have been synthesized. Supramolecular interactions play a crucial role in the formation of liquid crystals and in supramolecular polymer chemistry. The latter involves the designed manipulation of molecular interactions (e.g. hydrogen bonding, etc.) and recognition processes (receptor-substrate interaction) to generate main-chain or side-chain supramolecular polymers by self-assembly of complementary monomeric components. Figure 7 displays several different types of polymeric superstructures that represent supramolecular versions of various species and procedures of supramolecular polymer chemistry leading to materials with nanometer-sized microstructures. The controlled manipulation of the intermolecular interaction opens the way to the supramolecular engineering of NsM. 3.2.6 Template-assisted nanostructured materials and self-replication The basic idea of templating is to position the components into pre-determined configurations so that subsequent reactions, deliberately performed on the preassembled species or occurring spontaneously within them, will lead to the generation of the desired nanoscale structure. The templating process may become selfreplicating if spontaneous reproduction of one of the initial species takes place by binding, positioning and condensation (Dhal and Arnold, 1991; Philips and Stoddart, 1991; Benniston and Harriman, 1993). 2

Self-assembly should be distinguished from templating. Templating involves the use of a suitable substrate that causes the stepwise assembly of molecular or supramolecular structures. These structures would not assemble in the same way without the template.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanostructured Materials: Basic Concepts and Properties

35

Inorganic and organic templating has been used for the generation of nanometer-sized polymer arrangements displaying molecular recognition through i mprinting, i.e. a specific shape and size-selective mark on the surface or in the bulk of the polymer (Wulff, 1986 and 1993). Mesophase templating represents a special case that appears to be of considerable significance for the development of this area. Silica precursors when mixed with surfactants result in polymerized silica "casts" or "templates" of commonly observed surfactant-water liquid crystals. Three different mesoporous geometries have been reported (Kresge et al_, 1992; Beck et al_, 1992; McGehee et al., 1994; Monnier et al., 1993), each mirroring an underlying surfactant-water mesophase. These mesoporous materials are constructed of walls of amorphous silica, only about 1 nm thick, organized about a repetitive arrangement of pores up to 10 nm in diameter. The resulting materials are locally amorphous (on atomic length scales) and periodic on larger length scales. The availability of highly controlled pores on the 1-10 nm scale offers opportunities for creating unusual composites, with structures and properties unlike any that have been made to date. However, the effective use of mesoporous silicates requires two critical achievements: (i) controlling the mesophase pore structure; and (ii) synthesizing large monolithic and mesoporous "building blocks" for the construction of larger, viable composite materials. A special case is the reproduction of the template itself by self-replication. Reactions occurring in organized media (e.g. molecular layers, mesophases, vesicles) (Hub et al., 1980; Gros et al., 1981; Ringsdorf et al., 1988; Wegner, 1982; Paelos, 1985; Rees et al., 1993; Vlatakis et al., 1993; Morradian et al., 1989; Sellergren et al., 1989) offer an entry into this field. Supramolecular templating processes seem to provide an efficient route for the synthesis of nanoporous materials used as molecular sieves, catalysts, sensors, etc. In fact, mesoporous bulk (Kresge et al., 1992; Beck et al., 1992; Beck, 1994; Davis, 1993) and thin-film (Yang et al., 1991; Aksay, 1996) silicates with pore sizes of 2-10 nm have been synthesized by using micellar aggregates of long-chain organic surfactant molecules as templates to direct the structure of the silicate network. Potential applications of these molecular-sieve materials are catalysts, separation membranes and components of sensors.

3.2.7 Synthesis by an highly enhanced free energy

An important approach for synthesizing an organic or metallic NsM by the bottom-up route is Computer-Aided Multivariate Analysisbased on increasing the free en grained) polycrystal. The various modifications of this synthesis route differ by the procedures that are applied to increase the free energy. Ball-milling (Fig. 8), high-strain-rate deformation (Fig. 9) sliding wear, irradiation with high-energy particles, spark erosion and detonation of solid explosives have been used so far. All of these techniques are based on introducing a high density of lattice defects by means of plastic deformations or local atomic displacements. The crystal size of the final product can be varied by controlling the milling, the deformation, the irradiation or the wear conditions (e.g. the milling speed, temperature, type of mill used, etc.). This group of methods have been scaled up successfully. For example,

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

36

Gleiter

40 NComputer-Aided Multivariate

Analysis

Figure 8 Low temperature magnete ball milling. The trajectories of the balls are controlled by the external magnetic field resulting in shearing (A) or inpaet milling conditions (B). (Calka and Nikolov, 1995). P

II

die

test material

Figure 9 Generation of nanostructured materials by severe plastic deformation. Left side: Severe plastic trosion under pressure. Right side: Equal channel angular extrusion (Valiev, 2000).

cryomilling has been applied to produce commercial quantities of nanocrystalline A1/A1203 alloys. 4. STRUCTURE-CONTROLLED PROPERTIES OF NANOSTRUCTURED MATERIALS The properties of solids depends on size, atomic structure and the local chemical composition. Hence NsM may exhibit new properties as all three parameters are modified in a NsM in comparison to a single crystal with the same (average) chemical composition (cf. Fig. 2). In fact, it turns out that the most significant property variations are observed, if a characteristic length, energy, etc. scale of the NsM (e.g. the crystal size, the boundary thickness, the interfacial energy, etc.) becomes comparable to a characteristic length, energy, etc. scale (e.g. a magnetic exchange length, a coherency length, an activation energy, etc.) controlling the

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanostructured Materials: Basic Concepts and Properties

37

properties (e.g. the diffusional, the magnetic etc. properties) of solids. The enhanced diffusivity, the remarkable mechanical and magnetic properties of NsM are used in this paper as examples for the modification of properties by a nanometer-sized microstructure.

. Ag

~ Cu

i n nc-Cu, 10nm I

Figure 10 Enhanced diffusivity of Ag and Cu in nanocrystalline Cu. The line labelled SC displays the lattice self diffusion in Cu (Wurschum, 1996).

Figure 10 displays, the diffusivity of Ag and Cu in nanocrystalline Cu in comparison to the diffusivity in a Cu single crystal. A most remarkable enhancement of the diffusivity of as much as about 10 18 was noticed'. This enhancement results from the high atomic mobility in the grain boundaries between the nanometer-sized Cu crystallites (Horvath et al., 1997; Schumacher et al., 1989). The experimental observations suggest that the enhancement is primarily due to a reduced activation energy of diffusion in the boundaries relative to lattice diffusion. For a comprehensive review we refer to the paper by Wurschum (1996). The high density of grain boundary in nanocrystalline materials has been demonstrated to affect the mechanical properties significantly. If the plastic deformation of polycrystals occurs at elevated temperature, deformation processes based on grain boundaries sliding and diffusional processes become increasingly important. In other words, nanocrystalline materials are expected (due to their reduced crystal size) to become more ductile at elevated temperatures than coarsegrained polycrystals with the same chemical composition. The expected enhanced

3

For comparison, if the rate of growth of trees is compared with the speed of light in vacuum, one finds an enhancement by a comparable factor.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Gleiter

38

6 = 0% (as-deposited)

50 mm

48-200% 6-600%

a. E=5100%

Konv. Konv. (polykrist.) (polykrist.) TiO2

Nanokrist. TiO 3

2

b.

i

W -

c. \

"O2 III I

Ti02

\ Figure 11 (a) Plastic deformation of nanocrystalline Cu at ambient temperature (Lu et al., 2000). (b) Comparison of the plasticity (diamond indentation) of polycrystalline and nanocrystalline Ti02 at 20°C. (c) Schematic diagram showing schematically the response of polycrystalline and nanocrystalline Ti02 if a diamond pyramid penetrates into the material (Karch et al., 1987).

ductility has been experimentally confirmed (Fig. 11) for both metallic (Lu et al., 2000; McFadden et al., 2001), as well as ceramic nanocrystalline materials (Karch et al., 1987). Figure 11 displays the plasticity of nanocrystalline Cu as well as of nanocrystalline Ti02. At low temperatures, grain boundaries may act as slip barriers during plastic deformation of polycrystals. Hence, nanocrystalline materials are expected to be harder than a single crystal with the same chemical composition. Indeed at low temperatures, nanocrystalline materials were noted to become harder when the crystal size is reduced. A spectacular example of this kind has been reported for nanocomposits of Si3N4 and TiN or W2N, respectively. Computer-Aided Mu If the crystal size was reduced to 3 nm, the hardness became comparable to diamond (Veprek et al., 1996). High performance hard magnetic materials are based on the optimization of the intrinsic magnetic properties, the microstructure and the alloy composition. Nanocrystalline microstructures permit the tuning of the exchange and dipolar

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Nanostructured Materials: Basic Concepts and Properties

39

coupling between adjacent ferromagnetic grains, and, moreover the reduction of the grain size into the ferromagnetic single domain regime (Me Henry, Laughlin, 2000). Modern developments of computational micromagnetism (KronmUller et al., 1996) indicate that magnets with maximum coercitivity are obtained if single domain grains are magnetically decoupled by non-ferromagnetic boundary regions. Maximum remanence requires a mixture of magnetically hard and soft nanocrystals. The grain size of the magnetically soft grains (low crystal anisotropy energy) is shown to be optimized if it is about twice the wall width of the hard phase (Fig. 12). Approaches to improving intrinsic and extrinsic soft ferromagnetic properties involve (a) tailoring chemistry and (b) optimizing the microstructure. Significant in microstructural control has been the recognition that a measure of the magnetic hardness (the coercivity, He) is roughly inversely proportional to the grain size (Dg ) for grain sizes exceeding -0.1-1 tm where D g exceeds the domain (Bloch) wall thickness, &. Here grain boundaries act as impediments to domain wall motion, and thus fine-grained materials are usually magnetically harder than large grain materials. Significant recent development in the understanding of magnetic coercivity mechanisms has led to the realization that for very small grain sizes D g (1>). The lines are offset vertically for clarity.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

248

Jiang et al.

(a)

-1 5

10

t (1IQ o) Figure 3.2 Curves (a) and (b) show evolutions of the electron probabilities for dephasing rates Fd=3flo and curve (c) corresponding current through the detector.

The influence induced by detecting has been shown in Figure 3.2. It can be seen from curves (a) and (b) that the probabilities decay quickly and approach to 1/2 at sufficiently later time, revealing that the QD system lost its coherence. Comparing the curve (c) with the curve (a), we can find that when the electron probability approaches the maximum, the current also reaches the maximum in phase. This phenomena indicates that one can truly extract the information of the coupled QD system by means of measuring the current variation.

4 CONCLUSION In conclusion, using rate equations we demonstrate that this coupled QD system may perform all the operations of single qubit. By measuring the current variation we can extract the information of the coupled QD system. Also, we show that in the presence of the dephasing, the oscillating current through the detector decays drastically. For the application of the coupled QD system in the quantum computing and information, keeping an appropriate dephasing rate is necessary.

ACKNOWLEDGEMENT This work was supported by the National Natural Science Foundation of China.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Quantum Dynamics of Coupled QD Qubits

249

REFERENCES Aleiner, I. L., Wingreen, N. S. and Meir, Y., 1997, Dephasing and the orthogonality catastrophe in tunneling through a quantum dot: the "Which Path?" interferometer, Physical Review Letters, 79, pp. 3740-3743. Blick, R. H. et al., 1998, Formation of a coherent mode in a double quantum dot, Physical Review Letters, 80, pp. 4032-4035. Buks, E. et al., 1998, Dephasing in electron interference by a 'which-path' detector, Nature (London), 391, pp. 871-874. Gurvitz, S. A., 1997, Measurements with a noninvasive detector and dephasing mechanism, Physical Review B, 56, pp. 15215-15223. Gurvitz, S. A. and Prager. Ya. S., 1996, Microscopic derivation of rate equations for quantum transport, Physical Review B, 53, pp. 15932-15943. Levinson, Y., 1997, Dephasing in a quantum dot due to coupling in a quantum dot, Europhysics Letters, 39, pp. 299-304. Oosterkamp, T. H., et al., 1998, Microwave spectroscopy of a quantum-dot molecule, Nature(London), 395, pp. 873-876. Tsukada, N. et al., 1997, Dynamical control of quantum tunneling due to ac stack shift in an asymmetric coupled quantum dot, Physical Review B, 56, pp. 92319234. Wu, N. J. et al., 2000, Quantum computer using coupled-quantum-dot molecules, Jpn. Journal of A pplied Physics, 39, pp. 4642-4646.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

31 Coherent Dynamics and Quantum Information Processing in Josephson Charge Devices J. Q. You t '2 , Franco Nori t ' 3 and J. S. Tsai' 4 t Frontier Research System, The Institute of Physical and Chemical Research, (RIKEN), Wako-shi 351-0198, Japan 2 National Laboratory for Superlattices and Microstructures, Institute of Semiconductors, Chinese Academy of Sciences, Beijing 100083, China 3 Center for Theoretical Physics, Physics Department, The University of Michigan, Ann Arbor, M148109-1120, USA 4 NEC Fundamental Research Laboratories, Tsukuba 305-8051, Japan

1. INTRODUCTION Quantum information technology focuses on the quantum-state engineering of a system, with which the quantum states of the system can be prepared, manipulated and readout quantum mechanically. Recently, the "macroscopic" quantum effects in low-capacitance Josephson-junction circuits have received renewed attention because suitable Josephson devices may be used as qubits for quantum information processing (QIP) (Makhlin et al., 1999, 2001 and Mooij et al., 1999) and are expected to be scalable to large-scale circuits using modern micro fabrication techniques. Experimentally, the energy-level splitting and the related properties of state superpositions were observed in the Josephson charge (Nakamura et al., 1997 and Bouchiate et al., 1998) and phase devices (van der Wal et al., 2000 and Friedman et al., 2000). Moreover, coherent oscillations were demonstrated in the Josephson charge device prepared in a superposition of two charge states (Nakamura et al., 1999). These experimental observations reveal that the Josephson charge and phase devices are suitable for solid-state qubits in QIP. To realize QIP devices of practical use, the next immediate challenge involved is to implement two-bit coupling and then to scale up the architectures to many qubits in a feasible fashion. A subtle way of coupling Josephson charge qubits was designed in terms

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

Coherent Dynamics and QIP

.

• .

251

O 1

G ., T

x.

Figure 1: Schematic illustration of the proposed quantum device, where all Josephson charge-qubit structures are coupled by a common superconducting inductance.

of the oscillator modes in a LC circuit formed by an inductance and the qubit capacitors (Makhlin et al., 1999, 2001). In that design, interbit coupling is switchable and any two charge qubits can be coupled. However, an appropriate quantumcomputing (QC) scheme still lacks with this design and the interbit coupling terms calculated applies only to the case both when eigen-frequency of the LC circuit is much faster than the quantum manipulation times and when the phase conjugate to the total charge on the qubit capacitors fluctuates weakly. Here we propose a new QC scheme based on charge-qubit structures. In our proposal, a common inductance (but not LC circuit) is used to couple all Josephson charge qubits. Because the proposed QC architectures have appropriate Hamiltonians, we are able to formulate an efficient QC scheme by means of these Hamiltonians. Moreover, our QC scheme is also scalable because any two charge qubits (not necessarily neighbors) can be effectively coupled by an experimentally accessible inductance.

2. QUANTUM DEVICE The proposed quantum device consists of N Cooper-pair boxes coupled by a common superconducting inductance L (see Fig. 1). For the kth Cooper-pair box, a superconducting island with charge Qk = 2enk is weakly coupled by two symmetric dc SQUID's and biased by an applied voltage V xk through a gate capacitance C k . The two symmetric dc SQUID's are assumed to be identical and all junctions in them have Josephson coupling energy E °,k and capacitance C, k . Since the size of the loop is small ('- 1 µm), we ignore the self-inductance effects of each SQUID loop. The effective coupling energy produced by a SQUID (pierced with a magnetic flux 0 )& ) is given by EJk(h xk )cosO kA(B) with EJk (~P xk ) = 2Ejk ° cos(70 XV4 °), where 0 ° = h/2e is the quantum flux. The effective phase drop 0 kA (B), with subscript A(B) labelling the SQUID above (below) the island, equals the average value, [0 L, (B) + ~pRk.4(B)]/2, of the phase drops across the left and right Josephson junctions in the dc SQUID.

© 2003 Zikang Tang and Ping Sheng/CRC Press LLC

252

You, Nori and Tsai

The quantum dynamics of the Josephson charge device is governed by the Hamiltonian N

H=JHk + 1LI 2 , 2 k=1

(1)

with Hk = E,k(nk - CkVxkl2e) Here E ck

2- EJk((P

xk)(cosO

kA + cosO kB).

(2)

= 2e /(C k + 4CJk) is the charging energy of the superconducting 2

island and I =

Ik

l

I

k

is the total supercurrent through the superconducting

inductance, as contributed by all coupled Cooper-pair boxes. The phase drops kAL

0 kA L and 0

are related to the total flux 0 = cP L +LI through the inductance L by the kAL

constraint 0 kBL - O = 2TtP /' 0, where 0 L is the applied magnetic flux threading L. In order to implement QC in a feasible way, the magnetic fluxes through the two SQUID loops of each Cooper-pair box are designed to have the same values but opposite directions. Because this pair of fluxes cancel each other in any loop ~p R 0 kA', which yields 0 kR - 0 kA = 2nO /!P 0 enclosing them, there is ML - 0 kA L =0 kB for the average phase drops across the Josephson junctions in SQUID's. Here, each Cooper-pair box is operated both in the charging regime Ek >> E Ck and at low temperatures kBT