Encyclopedia of Nanoscience and Nanotechnology 9781588830579, 1588830578

Citation preview

Ingenta Select: Electronic Publishing Solutions

1 of 3

file:///C:/01/01.files/m_cp1-1.htm

Encyclopedia of Nanoscience and Nanotechnology Volume 1 Number 1 2004 Aligned Carbon Nanotubes

1

Xianbao Wang; Yunqi Liu; Daoben Zhu Alkanethiol Self-Assembled Monolayers

17

J. Justin Gooding Analytical Characterization of Single Wall Carbon Nanotubes

51

Sivaram Arepalli; Pasha Nikolaev; Olga Gorelik Analytical Ultracentrifugation of Nanoparticles

67

Helmut Cölfen Applications of Electrodeposited Nanostructures

89

G. Palumbo; J. L. McCrea; U. Erb Applications of Focused Ion Beam in Nanotechnology

101

V. J. Gadgil; F. Morrissey Atomic Manipulation by Scanning Tunneling Microscopy

111

K.-F. Braun; G. Meyer; F. Moresco; S.-W. Hla; K. Morgenstern; S. Fölsch; J. Repp; K.-H. Rieder Atomic Structure of Defects in Semiconductors

135

I. Yonenaga Atomic Structure of Diamond Surfaces

147

J. M. Perez; R. E. Stallcup II Atomic Wires

169

I. N. Yakovkin Augmented Waves for Nanomaterials

191

Pavel N. D'yachkov Bimetallic Ferrofluids

213

Michael Hilgendorff Biocompatible Core-Shell Nanoparticles for Biomedicine

235

Xiaoxiao He; Xia Lin; Kemin Wang; Liang Chen; Ping Wu; Yin Yuan; Weihong Tan Bioconjugated Silica Nanoparticles for Bioanalysis

255

Xiaojun Zhao; Lisa R. Hilliard; Kemin Wang; Weihong Tan Biodoped Sol-Gel Polymer Nanocomposites

269

Iqbal Gill

4/21/2007 8:22 PM

Ingenta Select: Electronic Publishing Solutions

2 of 3

file:///C:/01/01.files/m_cp1-1.htm

Biogenic Nanoparticles

293

Joseph M. Slocik; Marc R. Knecht; David W. Wright Biological Molecules in Nanodevices

309

Stephen C. Lee; Mark A. Ruegsegger; Mauro Ferrari Bionanodevices

329

Thomas Schalkhammer Bionanomotors

361

Dmitri Grigoriev; Dieter Moll; Jeremy Hall; Peixuan Guo Bismuth Nanostructured Materials

375

Joshua T. Moore; Charles M. Lukehart Boron-Carbon Nitride Nanohybrids

383

Yoke Khin Yap Boron Nitride Nanotubes

395

Dmitri Golberg; Yoshio Bando C60-Based Materials

409

J. G. Hou; A. D. Zhao; Tian Huang; Shan Lu Calixarenes

475

Chebrolu P. Rao; Mishtu Dey Cantilever-Based Sensors

499

D. Then; C. Ziegler Carbon Nanomaterials

517

Maheshwar Sharon Carbon Nanostructures for Cold Electron Sources

547

P. Gröning; L. Nilsson; P. Ruffieux; R. Clergereaux; O. Gröning Carbon Nanotube Growth by Chemical Vapor Deposition

581

M. Meyyappan Carbon Nanotube Sensors

591

Jun Li; Hou Tee Ng Carbon Nanotubes: Synthesis by Arc Discharge Technique

603

Yoshinori Ando Carbon Nanotube-Based Field Emitters

611

Seong Chu Lim; Hee Jin Jeong; Kay Hyeok An; Dong Jae Bae; Young Hee Lee; Young Min Shin; Young Chul Choi Carbon Nanotube-Based Supercapacitors

625

Young Hee Lee; Kay Hyeok An; Ji Young Lee; Seong Chu Lim

4/21/2007 8:22 PM

Ingenta Select: Electronic Publishing Solutions

3 of 3

file:///C:/01/01.files/m_cp1-1.htm

Carbon Nanotubes in Composite Materials

635

Ch. Laurent; A. Peigney Catalysis by Gold Nanoparticles

655

Masatake Haruta Catalytic Synthesis of Carbon Nanotubes and Nanofibers

665

Kenneth B. K. Teo; Charanjeet Singh; Manish Chhowalla; William I. Milne Ceramic Nanoparticle Synthesis

687

Xiangdong Feng; Michael Z. Hu Ceramic Nanopowders

727

Yong S. Cho; Howard D. Glicksman; Vasantha R. W. Amarakoon Charge Carrier Dynamics in Nanoparticles

745

Christian D. Grant; Thaddeus J. Norman Jr.; Jin Z. Zhang Chemical Derivatization of Carbon Nanotube Tips

761

Vladimir A. Basiuk; Elena V. Basiuk (Golovataya-Dzhymbeeva) Chemical Synthesis of Nanoparticles

777

H. Bönnemann; K. S. Nagabhushana Chemically Prepared Magnetic Nanoparticles

815

M. A. Willard; L. K. Kurihara; E. E. Carpenter; S. Calvin; V. G. Harris Chemistry of Carbon Nanotubes

849

V. N. Khabashesku; J. L. Margrave Chiral Macrocycles

863

Suk Joong Lee; Wenbin Lin Copyright © 2004 American Scientific Publishers

4/21/2007 8:22 PM

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Aligned Carbon Nanotubes Xianbao Wang, Yunqi Liu, Daoben Zhu Chinese Academy of Sciences, Beijing, People’s Republic of China

CONTENTS 1. Introduction 2. Carbon Nanotubes and Their Preparation Methods 3. Alignments and Patterns of Multiwalled Carbon Nanotubes 4. Alignments of Single-walled Carbon Nanotubes 5. Properties and Applications of Aligned Carbon Nanotubes 6. Summary Glossary References

1. INTRODUCTION Significant progress has been made in the science of carbon nanotubes (CNTs) since the publication of Iijima’s milestone paper in 1991 [1]. If there is one thing which has characterized fullerene and nanotube sciences, it is serendipity [2]. The discovery of buckminsterfullerene itself was a wonderful accident, and nanotubes were an unanticipated by-product of the bulk synthesis of C60 . CNTs have captured the imagination of physicists, chemists, and materials scientists alike. Their intriguing electronic, magnetic, optical, and mechanical properties, coupled with their unusual molecular shape and size, have made CNTs very promising as functional components for building molecular-based electronics, nanomachines, and nanoscale biomedical devices as well as composites as structural materials. Physicists have been attracted to their extraordinary electronic properties, chemists to their potential as “nanotest-tubes,” and materials scientists to their amazing stiffness, strength, and resilience. Nanotechnologists, mostly from the three aforementioned areas, have discussed and explored possible nanotube-based nanodevices such as nanogears, nanobearings, single electron transistors, quantum computers, and so on. Many review articles [3–9], books [2, 10–14], and special issues of journals [15–17] have been devoted to this topic to collect the ISBN: 1-58883-057-8/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

new developments and concepts in this rapidly developing interdisciplinary field. Following the discovery of CNTs, single-walled carbon nanotubes (SWNTs) reported simultaneously by Iijima and Ichihashi [18] and Bethune et al. [19] in 1993, which are ideal models of one-dimensional materials with unique structural and electronic properties, have generated great interest for use in a broad range of potential nanodevices. Perhaps the largest volume of research into nanotubes has been devoted to the electronic properties, of which the field emission study was extensive, profound, and highly near to a practicable product. CNTs are known to be very good electron emitters. This is why since their discovery there has been a lot of speculation and experiments about the use of nanotubes in the construction of flat panel devices. The first study on CNT field emission (FE) properties was carried out by Rinzler et al. [20] from Rice University. They found that field emission of electrons from individually mounted CNTs has been dramatically enhanced when the nanotube tips are opened by laser evaporation or oxidative etching. A short time-hereafter, the field emission from a film of postsynthesized reduced CNT alignments was made by de Heer and co-workers [21]. However, a prerequisite for a major breakthrough in the area would be the perfect alignments of CNTs on a suitable surface. This chapter is mainly focused on the synthesis and applications of aligned carbon nanotubes, including SWNTs and multiwalled carbon nanotubes (MWNTs). The organization of this chapter is as follows: Section 2 first considers three classical methods for synthesizing nanotubes, including arc discharge, laser ablation, and chemical vapor deposition (CVD), especially comparing the advantages and disadvantages of each synthesis method, and then presents the new preparation techniques developed in the last two years. In Section 3, the fabrications of various alignments and patterns are discussed in detail for MWNTs, and the emphasis in this section is on the controllable fabrication of nanotube alignments. Alignments of SWNTs are described in Section 4. Section 5 gives a discussion of properties and application of aligned CNTs, and the FE properties are of great concern. Finally, future research consideration and challenges are presented in Section 6.

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 1: Pages (1–15)

2

2. CARBON NANOTUBES AND THEIR PREPARATION METHODS A carbon nanotube consists of one or more seamless cylindrical shells of graphitic sheets. The nanotube is typically closed at each end, according to Euler’s theorem [22], by the introduction of pentagons in the hexagonal network. A theoretical model of a CNT with two closed ends was simulated by Endo and Kroto [23]. However, an experimental observation of the tube with both ends closed has seldom been reported for two main reasons. One is that most CNTs, regardless of the synthesis methods (such as arc discharge, laser ablation, and catalyst decomposition of hydrocarbon), are attached by either open ends, or one closed carbon cap, or ends capped with metal particles. The other is that the large length/diameter ratio (>1000) prevents one from simultaneously observing two ends of the tubes in the same field of vision under a transmission electron microscope (TEM), though tubes with two caps exist in the product. In an earlier report [24], it was also noticed that nanotubes produced using the arc discharge technique can be fairly short, and these structures with two capped ends can be observed under high resolution transmission electron microscopy. However, the structure is so small and asymmetric, with a diameter of at most 40 nm, that it should be precisely defined as an elongated nanoparticle of graphitic carbon rather than a well-defined CNT with capped ends. We reported a perfect carbon nanotube with two graphitic caps observed in the products by pyrolysis of metal phthalocyanine [25]. Such a large CNT with capped ends is of great interest for the formation mechanism of nanotubes. CNTs are most attractive because of their fascinating features. What makes nanotubes so special is the combination of dimension, structure, and topology that translates into a whole range of superior properties. The basic constitution of the nanotube lattice is the C–C covalent bond (as in graphite planes) which is one of the strongest in nature. The perfect alignment of the lattice along the tube axis and the closed topology endow nanotubes with in-plane properties of graphite such as high conductivity, excellent strength and stiffness, chemical specificity, and inertness together with some unusual properties such as the electronic structure, which is dependent on lattice helicity and elasticity. In addition, the nanodimensions provide a large surface area that could be useful in mechanical and chemical applications. The surface area of MWNTs has been determined by BET techniques and is ∼10–20 m2 /g, which is higher than that of graphite but small compared to activated porous carbons. This value for SWNT is expected to be an order of magnitude higher. Similarly, due to the relatively large hollow channels in the center of nanotubes, their density should be very low compared to graphite. Rough estimates suggest that SWNT density could be as small as 0.36 g/cm3 , and MWNT density could range between 1 and 2 g/cm3 depending on the constitution of the samples [7].

2.1. Classical Synthesis Methods Nanotubes are not always perfect seamless shells of graphite. Their quality depends on the method used to generate them and the exact conditions of the particular method. Making

Aligned Carbon Nanotubes

nanotubes is simple, but making good quality samples with high yields and highly graphitized shells is not trivial. There are several methods to produce CNTs, but three classical methods including arc discharge, laser ablation, and a catalytic technique remain the most practical for scientific purposes and realistic applications. The arc method [26, 27] remains by far the best technique for the synthesis of highquality nanotubes simply because the process has a very high temperature of 4000 K. However, this method suffers from a number of disadvantages [14]. First, it is labor intensive and requires some skill to achieve a satisfactory level of reproducibility. Second, the yield is rather low, since most of the evaporated carbon is deposited on the walls of the vessel rather than on the cathode, and the materials that are formed in the deposit contain substantial amounts of nanoparticle and other graphitic debris. Third, it is a batch rather than a continuous process, and it does not easily lend itself to scale-up. The amount of nanotubes that can be produced is limited. Progress in this direction has been hampered somewhat by a lack of understanding of the growth mechanism of tubes in the arc. The laser evaporation technique developed by Smalley’s group [28–30] appears to produce the highest yield and best quality materials, but the high powered lasers required for this method will obviously not be available in every laboratory. The synthesis of MWNTs in this way has been carried out by a pure graphite target [29]. In 1995, they reported the development of the laser synthesis technique which enabled them to prepare SWNTs with a target of a metal–graphite composite instead of pure graphite [28]. Subsequent refinements to this methd led to the production of SWNTs with unusually uniform diameter [30]. CVD of hydrocarbon over metal catalyst has been another classical method to produce carbon materials. Various forms of carbon fibers, filaments, and MWNTs have been synthesized by this technique in the past [31–34]. In general, metal catalytic particles are exposed to a medium containing hydrocarbon gaseous species, and the formation of nanotubes is catalyzed [35]. During growth, good uniformity in size of the tube is achieved by controlling the size of the seeded catalyst particles, and the process can be easily scaled up to produce large amounts of materials. In some cases, when the catalysts are prefabricated into patterned arrays, well-aligned nanotube assemblies are produced, which will be discussed in detail (Section 3). This seems a promising direction for further research. However, the quality of tubes produced in this way has been rather poor compared with the arc and laser methods. As shown in Figure 1, a nanotube prepared by a catalytic pyrolysis of iron phthalocyanines is a structure with lots of topological defects in a wall [36]. There is a further serious weakness of all these techniques for preparing nanotubes; they produce a wide range of the size and structures. This may not be a problem for some applications, but it could be a drawback in areas where specific tube structures with uniform properties are needed, such as in nanoelectronics. Progress in this direction has been hampered somewhat by a lack of understanding of the growth mechanism of tubes. Although a number of theories [37–40] have been put forward, none of which has gained universal acceptance, most questions remain unanswered and the uncertainty surrounding nanotube growth

Aligned Carbon Nanotubes

3 advantages favoring the scaling-up of this new method. However, this method suffers from harsh synthesis conditions (a high temperature of 800  C, a high pressure of 100 MPa, and a long time of 48 h), and electron microscopy observations revealed that very short multiwalled tubes (hundreds of nanometers) exist together with the aggregates of needlelike and polygonal nanoparticles. Compared with this method, a solvothermal route to MWNTs has been developed under a lower temperature (350  C) and shorter time (8 h) [46]. This catalyst-assembly benzene-thermal route was carried out to produce CNTs using reduction of hexachlorobenzene by metallic potassium in the presence of Co/Ni catalyst. The synthesis temperature was low, but the quantity of the nanotubes was poor.

2.2.2. Supercritical CO2 Technique

Figure 1. A carbon nanotube with many topological defects synthesized by chemical vapor deposition. Reprinted with permission from [37], M. Endo and H. W. Kroto, J. Phys. Chem. 96, 6941 (1992). © 1992, Elsevier Science.

mechanisms has impeded progress in the development of more controlled synthesis techniques.

2.2. Preparation Methods of Nanotubes In the last two years, a few new synthesis methods were developed to produce CNTs to solve one or more problems during the conventional preparation process. These ways take respective advantages over the conventional methods to a certain extent and were of important potential for basical research as a new route to produce CNTs. However, the samples of the nanotubes produced by all those new methods existed in lower quality and less perfect structures than the classical ones. A number of experimental conditions should be optimized before these synthesis methods are developed to produce nanotubes with high quantity and desirable structures to meet scientific research and practical applications.

Carbon dioxide is nonflammable, essentially nontoxic, and environmentally benign [47]. Motiei et al. [48] reported the structures of graphitic concentric shells grown by supercritical CO2 . It is found that MWNTs and nested fullerenes could be prepared from dry ice in the presence of Mg by heating the precursors in a closed vessel at the autogenic pressure of the mixture. This method avoids the complexities of using a flowing gas at controlled pressures and high temperature and requires no technically complex equipment.

2.2.3. Solid-State Metathesis Reaction The solid-state metathesis reaction has been developed over the past few years into a simple and effective route to materials that are difficult to synthesize by conventional methods [49–51]. These highly exothermic and self-propagating reactions initiated by a heated filament often use molecular precursors to produce crystalline products. An exchange reaction between carbon halides and lithium acetylide catalyzed by cobalt dichloride enables the rapid synthesis of CNTs by the following equation [52]: C2 Cl6 + 3 Li2 C2 → 8 C (nanotubes) + 6 LiCl. Without a catalyst, only graphitic and amorphous carbon forms. These reactions, once optimized, will likely require cheaper precursors, more preparation, and less expensive equipment than existing methods. However, along with multi- and single walled nanotubes, graphite-encapsulated cobalt nanoparticles, free carbons, and cobalt metals were found. These shortages should be overcome before the extensive application of this method.

2.2.1. Hydrothermal or Solvothermal Synthesis

2.2.4. Flame Synthesis

The oxidative effect of hot water on amorphous carbon is currently used for purification of CNTs by hydrothermal treatment (immersion in water at moderate and high temperatures and pressures) [41, 42]. It was found that the hydrothermal processing method could be developed to synthesize nanostructural carbon [43, 44]. Recently, MWNTs have been synthesized in the absence of metal catalyst by hydrothermal treatment of amorphous carbon in pure water at 800  C and 100 MPa [45]. The hydrothermal nanotubes are free of amorphous carbon after treatment. The homogeneity of hydrothermal processes and availability of amorphous carbon materials, without the need for catalyst, are

Combustion is widely used in industrial processes for largescale materials synthesis, such as for carbon black and metal oxides [53, 54]. These flame processes are well known for their many desirable features including continuous processing and energy efficiency. Recent investigations have sought to take advantages of these advantages to synthesize SWNTs in aerosol form [55, 56]. It was found that SWNTs could be produced by a binary or ternary gas mixture of CO/C2 H2 /H2 with a metal catalyst at 700  C in a two-stage flame. This flame system afforded the advantage that the catalyst formation step could be separated from the nanotube growth steps, which allowed investigations of catalyst particle size

Aligned Carbon Nanotubes

4 dependencies upon nanotube growth unlike methods that employ in-situ generation and concurrent growth. Apart from the methods presented previously, there are many other methods such as a gas phase synthesis [57], a reduction catalysis route at a low temperature of 200  C [58], and a solar approach [59], although these methods to produce nanotubes suffer from different drawbacks.

3. ALIGNMENTS AND PATTERNS OF MULTIWALLED CARBON NANOTUBES CNTs are the most promising materials anticipated to impact future nanoscience and nanotechnology. Their unique structural and electronic properties have generated great interest for use in a broad range of potential nanodevices [60, 61]. Most of these applications will require a fabrication method capable of producing CNT alignments or patterns with uniform structures and periodic arrangements to meet device requirements. Therefore, the ability to controllably obtain ordered or pattering CNT architectures is important to both fundamental characterizations and potential applications [62]. Controlled synthesis involving CVD has been studied as an effective strategy to order or pattern CNTs on a variety of surfaces [63–67]. This has been stimulated by its simplicity, its ability to yield structures that are two- or three-dimensional (2D or 3D) periodic over large area, and its potential to be much less expensive than other synthesis methods such as an arc discharge and a laser ablation. In these early attempts, most of the works to date have focused on fabricating 2D CNT alignments or patterns over large areas because of their paramount importance for obtaining scale-up functional devices. However, in contrast to these efforts, little research on the preparation of 3D CNT alignments or patterns has been reported because of many technical difficulties, although the 3D structures show more unique and potential applications such as photonic devices [68, 69], data storage [70, 71], and ultrahydrophobic materials [72, 73]. In this section, various fabricating methods are discussed for alignments and patterns of MWNTs, including 2D or 3D structures.

3.1. Preparation of 2D Nanotube Alignments on Different Substrates Alignments or patterns of CNTs are particularly important for fabricating functional devices such as field emitters and nanoelectronics. Earlier attempt to manipulate nanotubes for these application have been made by postgrowth methods such as cutting a polymer resin–nanotube composite [74] or drawing a nanotube–ethanol suspension through a ceramic filter [75]. In the past few years, a great quantity of research on the fabrication of 2D CNT alignments has been reported [63–67, 76–78]. Although all these 2D CNT alignments were synthesized by CVD, different technical routes were adopted by different research groups. The main differences of these synthesis methods may be summarized as the following factors: a substrate and a precursor. The substrates that are used to support the CNTs alignments may be put into two categories: porous templates (mesoporous

silica [63], nanochannel alumina [76], etc.) and plain plates (quartz glass [77], single crystal silicon plate [78], glass [65], etc.). As far as the precursor is concerned, there are two main materials: hydrocarbons [66, 76] and metal organic compounds [77]. When hydrocarbons are used as precursors, metal catalysts (such as Fe, Co, Ni) should be added to grow CNTs. However, as for metal organic compounds no additional metal catalyst is necessary since these precursors contain both the metal catalyst and carbon source required for CNT growth. The following discussion is mainly focused on the substrates where the nanotube alignments grow.

3.1.1. Planar Substrate Quartz glass is a good planar substrate for growing wellaligned nanotubes [64, 77, 79]. Terrones et al. [64] described a method for generating aligned nanotubes by pyrolysis of 2-amino-4, 6-dichloro-triazine over thin films of a cobalt catalyst patterned on a silica substrate by laser etching. We have grown well-aligned nanotubes with large diameters on quartz glass substrate by pyrolysis of a metal phthalocyanine under Ar/H2 flow at a temperature of 950  C [77]. We investigated the influence of the growth time and precursor level on the structures of MWNTs. It was found that both the outer diameter and length of aligned nanotubes increased with increasing growth time and precursor level. The outer diameters of well-aligned nanotubes range from 25 to 250 nm, and the length ranges from 1 to 500 m. However, the microstructures of aligned naotubes are independent of the growth temperature and the Ar/H2 flow rate in our case. Figure 2a is a scanning electron microscopy (SEM) image showing a large area of well-aligned CNTs with uniform diameter and length. Figure 2b is a typical TEM image of the bamboo-shaped nanotubes [80] produced by pyrolysis of 0.5 g iron phthalocyanine in 12 min. The graphite sheath sliding out from the Fe particle surface accounts for the formation of the compartments of bamboolike CNTs. The driving force of the sliding was caused by the stress accumulated

100nm -100µm

(a)

(b)

Figure 2. A SEM micrograph showing large area well-aligned CNTs perpendicular to the surface of the substrate. (b) A typical TEM image of the CNTs produced by pyrolysis of 0.5 g FePc in 12 min. Reprinted with permission from [77], X. B. Wang et al., Chem. Phys. Lett. 340, 419 (2001). © 2001, Elsevier Science.

Aligned Carbon Nanotubes

in the graphite sheath due to the segregation of carbon atoms from the inside of the sheath [80]. Apart from quartz glass, we can produce aligned nanotubes on various substrates made of other materials such as single crystalline silicon, iron plate, nickel plate, cobalt plate, and so on. Glass is the most common material for panel display. The ability to prepare aligned nanotubes over a glass substrate makes them more suitable for electron emission applications. Recently, Ren et al. [65, 81] reported the growth of large scale well-aligned CNT arrays on nickel-coated glass at temperatures below 666  C by plasma-enhanced hot filament CVD. Acetylene gas was used as the carbon source and ammonia gas was used as a catalyst and dilution gas. It was found that NH3 plays a crucial catalytic role together with the nickel layer to promote the growth of the nanotubes, and nickel thickness plays a very important role in determining the diameters. Nickel wafers were found to be another good substrate for producing nanotube alignments. Chen et al. have grown aligned graphitic nanofibers on single crystalline Ni (100) [82] and polycrystalline Ni substrate [83] via plasma assisted hot filament CVD using a gas mixture of nitrogen and methane. Small Ni particles on the substrate surface generated by the plasma acted as a catalyst for the growth of nanofibers.

3.1.2. Porous Template Mesoporous silica containing iron nanoparticles was used as a substrate when well-aligned CNTs were first synthesized by Xie et al. [63]. The mesoporous silica was prepared by a sol–gel process from tetraethoxysilane hydrolysis in iron nitrate aqueous solution [84]. This method to produce large areas of highly ordered, isolate long nanotubes is based on CVD. The tubes are up to about 50 m long and well graphitized. The growth direction of the nanotubes can be controlled by the pores from which the nanotubes grow. Porous silicon, a light-emitting material [85, 86], is an ideal substrate for growing organized nanotubes. Fan et al. [60] have prepared porous silicon patterned with Fe film by electron beam evaporation through shadow masks and then on this substrate carried out the synthesis of self-oriented regular arrays of CNTs by catalyst decomposition of ethylene at 700  C. Porous silicon substrate exhibits an important advantage over plain silicon for synthesizing nanotubes. The investigation showed that the nanotubes grew at a higher rate on porous silicon than on silicon. The well-ordered nanotubes can be used as electron FE arrays. Scaling-up of the synthesis process should be compatible with the existing semiconductor processes and allows the development of nanotubes devices integrated into silicon technology. Highly ordered arrays of CNTs can be grown by pyrolysis of acetylene on cobalt within a hexagonal close-packed nanochannel alumina template at 650  C [76]. The method is based on template growth. An illustration of a typical fabrication process flow is shown in Figure 3a. The process begins with the anodization of high purity (99.999%) alumina on a desired substrate. The next step is to deposit electrochemically a small amount of cobalt catalyst into the bottom of the template channels. The ordered arrays of nanotubes (Fig. 3b)

5 (a)

Aluminium

NCA template Al anodization

Cobalt catalyst

Carbon nanotubes

Co deposition

C2H2 pyrolysis

(b)

200 nm

Figure 3. (a) Schematic of fabrication process. (b) SEM image of the resulting ordered arrays of CNTs synthesized using the nanochannel alumina template. Reprinted with permission from [76], J. Li et al., Appl. Phys. Lett. 75, 367 (1999). © 1999, American Institute of Physics.

are grown in a flow of a mixture of 10% acetylene in nitrogen. There are several features in this fabrication technique [87, 88]. First, each pore of the template is filled with one nanotube, which defines the tube diameter, and the tube diameter distribution throughout the arrays is narrow. Second, the controlled variation of the nanotube size, density, and array spacing depends on easily adjustable parameters such as the anodizing voltage, electrolyte composition, and temperature. Tube lengths of up to 100 m can be obtained by varying the length of the pores in the alumina template, which can be achieved by varying the time of anodization. Finally, the method allows inexpensive production of large arrays of ordered nanotubes.

3.2. Controllable Fabrication of CNT Alignments 3.2.1. Selective Positioning Growth for Patterns Selective positioning growth of CNT patterns is arousing a wide range of research interest [89–91]. In the past few years, several methods have been developed to site-selectively grow nanotubes. Photolithographic Approach Wei et al. [89, 90] have used a CVD method with gas-phase catalyst delivery to direct the assembly of CNTs in a variety of predetermined orientations onto silicon/silica substrates, of which patterning was generated by photolithography followed by a combination of wet and/or dry etching. Figure 4 shows some striking examples of organized nanotube patterns grown on preselected substrate sites by the photolithographic approach. There is no nanotube grown on silicon, but the aligned nanotubes grow readily on silica in a direction that is normal to the substrate surface [90]. The preference of nanotubes to grow

Aligned Carbon Nanotubes

6

d = 10 µm

d = 2 µm

d = 5 µm

Si

Figure 4. SEM images of organized nanotube patterns grown on preselected substrate sites by a photolithographic approach. Scale bars, 100 m. Reprinted with permission from [89], B. Q. Wei et al., Nature 416, 495 (2002). © 2002, Nature Publishing Group.

selectively on and normal to silica surfaces allows the sites of nucleation and the direction of growth to be controlled. Yang et al. [67] have also reported the fabrication of patterns of perpendicularly aligned nanotubes with resolution down to m scale by pyrolysis of iron phthalocyanines onto a quartz substrate prepatterned with a photoresist film. Electron Beam Lithographic Approach With a electron beam lithographic technique, patterned growth of freestanding nanotubes on nickel dots on silicon can be achieved by plasma-enhanced hot filament CVD [92]. The thin film nickel pattern as a catalyst for growing nanotubes was fabricated on a silicon wafer by a standard microlithographic method and metal evaporators. Well-separated, single MWNTs grew on each dot of an array of ∼100 nm nickel dots under an acetylene/ammonia mixture at below 660  C. The diameter and height depend on the nickel dot size and growth time, respectively. Using this method, devices requiring freestanding vertical CNTs such as scanning probe microscopy, FE flat panel displays, etc. can be fabricated. Soft Lithographic Approach Most microfabrication methods mentioned start with photolithography to form a pattern in a photoresist on the substrate. Although these techniques are very widely used, they are incompatible for solution such as gels, some polymers, some organic and organometallic species, and biological molecules. To pattern these materials successfully, the patterned photoresist must be impermeable to the reagents used and the deposited materials should not be compromised by solvents used for the liftoff. Methods other than photolithography

often involve a shadow mask formed from a rigid metal. The air gap between the mask and substrate makes the use of rigid shadow masks to pattern materials from solution impossible [60]. One micropatterning technique that circumvents some of these drawbacks is soft lithography [93–95], which uses a patterned elastomer fabricated from poly(dimethylsiloxane) as the mask, stamp, or mold. Because the elastomer can conform to and seal reversibly against the contours of a surface, it can be used as a mask or a stamp. Soft lithography has become a very promising technique for micro/nanostructuring a wide range of materials. Various strategies, including microcontact printing (CP) [96–98] and micromolding [99], have been developed for nanoscale patterning. Kind et al. [96] elucidated some important aspects of using CP to pattern silicon substrate with catalysts followed by the growth of CNTs on the activated regions. In brief, a patterned and inked elastomeric stamp is used to print a catalyst as a pattern onto a substrate (Fig. 5). The growth of MWNTs follows from the catalytic decomposition of acetylene on the printed pattern of the catalyst. Changing the concentration of the catalyst in the ink solution allows one to tune the density of the nanotubes from single, randomly oriented nanotubes to densely packed arrays of nanotubes oriented normal to the substrate. This approach relied on rigid solid-state substrates such as silicon and alumina to achieve aligned nanotubes, which SiO2

Si

Printing

Inked stamp

Catalyst

Catalyst

CVD

Carbon nanotubes

Figure 5. Procedure for the patterned growth of carbon nanotubes by microcontact printing a Fe(III)-based catalyst precursor onto silicon wafers. The stamp is inked with an ethanolic solution of Fe(III) and then printed onto the substrate. The growth of carbon nanotubes proceeds by the catalytic decomposition of acetylene. Reprinted with permission [96], H. Kind et al., Langmuir 16, 6877 (2000). © 2000, American Chemical Society.

Aligned Carbon Nanotubes

may limit its wider scope of application. Hence, it would be interesting and challenging to investigate the possibility of performing in-situ selective growth on a novel substrate such as polymers and its direct integration with elastomeric polymers to fabricate practical devices. Ng et al. [98] have recently reported the combination of soft-lithographymediated growth and surface wettability manipulation of an elastomeric polymer to achieve in-situ highly site-selective growth of multiwalled nanotubes arrays on elastomeric substrates patterned with catalyst precursors. The realization of this approach may provide new methodologies for flexible FE display, implantable sensors, etc. Biasing Growth Approach The investigations by Avigal and Kalish [100] showed that patterns of MWNTs could be obtained by positively biasing the substrate during growth. Growth was performed in a flowing mixture of 7% CH4 in Ar onto Co covered Si held at 800  C with and without the presence of an electric field. It was found that the tube alignment occurs only when a positive bias is applied to the substrate whereas no aligned growth occurs under negative bias and no tube growth is observed with no field. Therefore, selective area biasing may permit selected area growth of vertically aligned CNTs, a process that may find many applications. Fabrication of micropatterned nanotubes remains both scientific and technically challenging. Many methods have been presented in preceding parts of this section to produce nanotube patterns. However, these synthetic methods suffer from complex pre- or postsynthesis manipulation. In 2000, we have developed a simple method [101] for large-scale synthesis of CNTs (up to several square centimeters) aligned in a direction normal to the substrate surface (typically, quartz glass plates). Unexpectedly, we found honeycomblike aligned CNTs by pyrolysis of nickel–cobalt phthalocyanine (designated as Ni–CoPc), which was synthesized by roasting the mixtures of metal salts, phthalanlione, urea, and ammonium molybdate. A typical experimental procedure is as follows: a clean quartz glass plate (4 × 2 × 01 cm) was placed in a flow reactor consisting of a quartz glass tube and a furnace fitted with an independent temperature controller: A flow of Ar–H2 (1:1, v/v, 20 cm3 min−1 ) was then introduced into the quartz tube during heating. After the central region of the furnace reached 950  C, a quartz boat with 0.5 g Ni–CoPc was placed in the region where the temperature was about 500–600  C. After 5 min heating, CNTs grew in a direction normal to the substrate surface. Figure 6 is a set of SEM micrographs showing the honeycomblike shape of CNTs. As can be seen in Figure 6a, the honeycomblike CNTs are close together, and the diameters of the honeycombs range from 15 to 80 m. Figure 6b is a higher magnification image of an area in Figure 6a and clearly shows the size and distribution uniformity of honeycombs. Figure 6c is a magnification image of a typical honeycomb shown in Figure 6a (square frame). The external diameter of the honeycomb is 80 m, which contains a hollow inner cavity with diameter of 30 m. The CNTs, forming the honeycombs, grow out from the inner cavity perpendicular uniformly to the substrate surface and then extend all around from the opening of the inner cavity, and finally twin with those of other honeycombs along its outside line.

7

5005 25kv 50µm (a)

(b)

(c)

Figure 6. (a) SEM micrograph of honeycomblike aligned CNTs. (b) A higher magnification image of honeycomblike aligned CNTs. (c) Magnification image of a typical honeycomb shown in Figure 2a. Reprinted with permission from [101], X. B. Wang et al., Appl. Phys. A 71, 347 (2000). © 2000, Springer-Verlag.

It is interesting that the same patterns of honeycomblike CNTs, two years later, were synthesized by prolysis of ferrocene and xylene on thermally oxidized silicon wafers [102].

3.2.2. Controllable Growth of Structures and Density of Aligned Nanotubes Controllable Growth of Diameters and Lengths Since nanotubes were first discovered in 1991, several advances in synthesis have led to the production of tubes in larger quanlities [63, 103] and higher purities [104–106]. Different diameter, length, and chairality of nanotubes give rise to diverse physical and mechanical properties [10, 107, 108]. However, controlling diameter, length, and charity has never been easy. Recently, several approaches for controlling diameters of SWNTs have been made. As far as the arc discharge method is concerned, the diameter of SWNTs can be changed by selecting metal catalysts [109] and ranging the pressure of helium gas [110]. For the laser ablation method, the diameters of SWNTs can be controlled by varying the furnace temperature [111] and laser pulse power [112]. However, the research on controlled growth of MWNTs, especially aligned CNTs, falls behind in contrast with that of SWNTs. Choi et al. [113] have synthesized aligned CNTs on Ni-deposited Si substrate using microwave plasma-enhanced CVD. They found that the diameter, growth rate (length), and density of CNTs could be controlled systematically by the grain size of Ni thin films. With decreasing the gain size of Ni thin films, the diameter of the nanotubes decreased, whereas the growth rate and density increased. The gain size of Ni films varied with the radio frequency (rf) power density during the rf magnetron sputtering process. As mentioned in Section 3.1.1, we can control the diameters and lengths of aligned CNTs by varying the growth time and the precursor level. Research results show that

Aligned Carbon Nanotubes

8 the length of the aligned nanotubes increases with the increase of the growth time, and the mean diameters does sharply, too. The diameter distribution width (a difference of between the largest and smallest diameter of a sample) shifts systematically to larger region with increasing growth time, which indicates that the outer diameters of nanotubes become inhomogenous in longer growth time. The template approach, shown in Section 3.1.2, can control effectively the diameter and length of aligned nanotubes by changing the nanochannel size and the thickness of template substrate, respectively. Recently, Jeong et al. [114] reported nanotube alignments with a narrow length distribution by etching the aluminum oxide template away. Sonication of aligned nanotubes on the template in an acetone solution cut the overgrown tubes effectively, resulting in short MWNTs. Site Density Controlling of Aligned CNTs Although the diameter and the length of aligned nanotubes can be easily controlled by changing the catalyst particle size, the growth time, etc., control of the site density is still challenging. For well-aligned nanotubes, tuning of site density is very important for certain applications, such as FE, nanoelectronic arrays, etc., because of the shielding effect of the dense arrays. Previous methods used to induce the site density include electron beam lithography [92], photolithography [115], microcontract printing [96], shadow mask [60], etc. However, all these methods either require expensive equipment and intensive labor or cannot control the site density in large area. Recently, research efforts to control the site density have been successful. Pulse-current electrochemical deposition has been used to prepare Ni nanoparticles that are used as the catalysts for the growth of aligned CNTs [116]. The nucleation site density of the Ni nanoparticles was controlled by changing the magnitude and duration of the pulse current. The site density of the aligned nanotubes varied from 105 to 108 cm−2 (Fig. 7).

(a)

(d)

10 µm

10 µm (b)

(e)

10 µm

10 µm (c)

(f)

1 µm

10 µm

Figure 7. SEM images of aligned CNTs with site densities of (a) 75 × 105 (b) 2 × 106 , (c) 6 × 106 , (d) 2 × 107 , and (e) 3 × 108 cm−2 , and (f) a single standing CNT. Reprinted with permission from [116], Y. Tu et al., Appl. Phys. Lett. 80, 4018 (2002). © 2002, American Institute of Physics.

substrate surface. A few pillar-shaped structures of CNTs grow out from the 2D alignments in a well-distributed mode, which characterizes the 3D CNT alignments. The CNT posts with a diameter of about 3.4 m are 7.8 m higher - 50µm

- 50µm

3.3. 3D Alignments and Patterns of CNTs In the preceding part of this section, we gave a wide coverage on the fabrication of alignments and patterns of 2D nanotubes with selective positioning and controlled growth. However, in sharp contrast to the research enthusiasm for 2D aligned CNTs, the research on 3D alignment and patterns of nanotubes is sparse. We have developed a simple method for the large-scale synthesis of 3D aligned CNT patterns on quartz glass substrate [62, 117]. A typical experimental procedure is as follows: a flow reactor, consisting of a quartz tube and a furnace fitted with an independent temperature controller, was heated to 950  C under a flow of Ar/H2 (1:1, v/v, 20 cm3 min−1 . Almost immediately after the transfer of a quartz glass plate (4 × 2 × 01 cm) from acetone solution to the central region (950  C) of the furnace, a quartz boat with 0.5 g of FePc was placed in the region where the temperature was 550  C. After 5 min CNTs grew in a direction normal to the substrate surface. Figure 8a shows an SEM image of 3D regular arrays of nanotubes aligned along the direction perpendicular to the

(a)

(b) - 200µm

(c)

Figure 8. (a) SEM images of 3D regular arrays of nanotubes aligned along the direction perpendecular to the substrate surface. SEM images of 3D nanotube patterns: (b) ringlike castles and (c) a 490-m-long crucian carp without a tail and fins. Reprinted with permission from [117], X. B. Wang et al., Chem. Commun. 8, 751 (2001). © 2001, The Royal Society of Chemistry.

Aligned Carbon Nanotubes

9

than the 2D nanotube alignments, whose height is mainly 6 m from the quartz substrate. In addition to the pillarshaped 3D nanotube alignments, most interesting patterns made of nanotubes arrays, such as ringlike castles (Fig. 8b) and a 490-m-long crucian carp without a tail and fins (Fig. 8c), were observed under similar experimental conditions. Although the growth mechanism for these patterns is incomplete at present, we think that the substrate should be responsible for their formation. At the original stage, acetone soaked on the quartz glass substrate was rapidly carbonized at high temperature before volatilization to form the patterns. The resulting carbon, together with iron atom by pyrolysis of FePc, generates metal carbide, which is a more activated catalyst to grow CNTs than metal [80]. Thus, the CNTs in the pattern region formed by metal carbide grow more rapidly than those by metal. This accounts for the formation of alignments and patterns of 3D CNTs. Apart from this, both the strong van der Waals interactions between the tubes and the high surface density of the growing nanotubes serve as additional advantages for the constituent nanotube to be “uncoiled” and allow the aligned nanotubes to develop on the quartz substrate. In spite of the interesting patterns of 3D nanotubes, the results were unexpected. On the basis of the previous experiments, we have recently realized the controllable fabrication of 3D nanotubes patterns with features of high resolution by a vacuum deposition technique through shadow masks [118]. Using SEM, we observed 3D aligned CNTs, consisting of two streaked ribbons with widths of 6.5 and 8.4 m, respectively (Fig. 9), by pyrolysis of FePc for 5 min. The nanotubes in the ribbon area are about 1.4 m higher than those of 2D alignments (beyond the ribbon region). The growth rates of the nanotubes in and beyond the ribbon region were about 25 and 20.3 nm/s, respectively. It is only the difference of growth rates that results in the formation of 3D nanotube alignment. These 3D micropatterns of well-aligned

NONE

SEI

3.0kV

X2,200

10µm

WD 8.1mm

Figure 9. SEM images of 3D nanotube alignments consisting of two ribbonlike structures (6.5 and 8.5 m). The nanotubes in the ribbon region are 1.4 m higher than those elsewhere. Reprinted with permission from [118], X. B. Wang et al., Adv. Mater. 14, 1557 (2002). © 2002, Wiley–VCH.

nanotubes were also prepared on photolithographically prepatterned substrates by the same method [119]. This opened the way for fabricating 3D nanotube micropatterns and thus developing novel nanoelectronic devices.

4. ALIGNMENTS OF SINGLE-WALLED CARBON NANOTUBES Alignments of SWNTs have been envisioned to enhance performance of various technologically important devices such as sensors [120], field emitters, and organic light-emitting diodes [121]. Immobilizing the random SWNTs into a controlled orientation would be an extremely important step for these real device application of the nanotubes. However, the creation of SWNT alignments has fallen far behind MWNT arrays because of the technical difficulty in handling or aligning individual SWNTs to ideal locations.

4.1. Chemical Alignments Chemical alignment is an effective method to order SWNTs perpendicular to the substrate by shortening SWNTs in an oxidizing environment, followed by chemical modification of the carboxyl end groups. Short and long SWNTs have exhibited considerable affinity for amine-functionalized substrates, although they tend to orient parallel to the substrate [122, 123]. Thiol functionalization of SWNTs resulted in better alignment on gold substrate; nevertheless, this system was plagued by low surface coverage and long adsorption time [124]. Recently, Wu et al. developed an alternative method for the assembly of oxidatively shortened SWNTs on silver surfaces [125]. This technique is based on the spontaneous adsorption (self-assembly) of the COOH groups at the open ends of CNTs onto silver surfaces. Their studies revealed that most of the SWNTs (ca. 80%) assembled on silver surfaces have a bundle size of 65 ± 05 nm, possibly suggesting the selective adsorption of SWNTs on silver. More recently, dense arrays of monolayer and multilayer assemblies of shortened SWNTs have been demonstrated using a metal assisted organization process from nonaqueous media [126]. The self-assembly, which was performed on a substrate such as glass, silicon (110) wafers with native oxide and quartz crystal microbalance resonator, consisted of sequential dipping in an aqueous solution of FeCl3 followed by immersion in DMF dispersed shortened SWNTs and separated by intermedia washing in DMF. A monolayer of densely packed, needlelike domains was obtained after 30 min immersion in nonaqueous dispersions of shortened SWNTs. This geometry is believed to be the result of a high concentration of carboxy groups on the severed edges of shortened SWNTs, hydroxy functionalization of Fe3+ -decorated surfaces, and strong hydrophobic interactions between adjacent shortened SWNTs.

4.2. Alignments by an Electrical or Magnetic Field Applying an electrical or magnetic field may be a simple and alternative method to align CNTs. Bubke et al. [127] reported that MWNTs dispersed in ethanol can be aligned by an electrical field. Due to the orientation of these elongated particles, the nanotube alignments exhibited optical

10 anisotropy. Thick films of aligned SWNTs and ropes have been produced by filtration/deposition from suspension in strong magnetic field [128, 129]. Recently, Chen et al. [130] aligned SWNTs on Si/SiO2 substrate with an alternating current electrical field. Highly oriented SWNT samples were prepared when an electric field with a frequency of 5 MHz was applied. The alignment of the SWNTs demonstrated significant dependencies on the frequency and the magnitude of the electric field. The results suggested that the alignment degree of CNTs was reduced gradually with the decrease of the frequency of the electric field, and the concentration of the aligned SWNTs decreased with the decrease of magnitude. However, SWNT samples oriented by the dc electric field did not demonstrate the apparent orientation. Zhang et al. [131] also demonstrated electric field directed growth of SWNTs over the quartz substrate with prepatterned polysilicon films by CVD. Highly aligned suspended SWNTs can be fabricated along the electric field direction in the range of 0.5–2 V/m.

4.3. In-Situ Self-Assembly of Aligned SWNTs Both chemical alignment and alignment by an electrical/ magnetic field were obtained by a postprocessing technique, which suffered from the difficulty of local position and complex postsynthesis manipulation. So in-situ preparation of aligned SWNTs is of paramount importance for basic research and many potential applications. Schlitter et al. [132] reported the self-assembly of single crystal SWNTs using thermolysis of alternate layers of C60 and Ni precursors. It is exciting that each crystal is composed of an ordered array of tubes with identical diameters and chirality, although these properties vary between crystals. The structures produced are almost perfect rodlike crystals of SWNTs preferentially oriented normal to the surface of a molybdenum substrate. The perfection of the crystals of SWNTs and the observation that they are all physically identical within any given crystal containing up to several thousand individual nanotubes are unexpected from the point of view of previous results and synthetic approaches in the field. Zhu et al. [133] directly synthesized the long strands of ordered SWNTs by catalyst CVD with a floating catalyst method in a vertical furnace, where n-hexane in combination with thiophene, ferrocene, and hydrogen is catalytically pyrolyzed. The long nanotube strands up to several centimeters in length, consisting of aligned SWNTs, are an alternative to the fibers and filaments spun from nanotube slurries. The research on mechanical and electrical properties indicated that the aligned nanotube strand would be a candidate for practically useful nanotube-based macroscale cables. However, well-defined alignments of SWNTs synthesized in-situ on a substrate were not fabricated until Botti et al. reported the latest results [134]. It was found that wellaligned SWNTs on heated Si substrates could be selfassembled from the carbon nanosized particles without a catalyst by a simple spraying technique. By increasing the substrate temperature, the density of the SWNTs increases and the nanotubes with a uniform length are oriented perpendicularly to the substrate. Although the aligned tubes are bundles of tightly interlaced SWNTs with different chirality

Aligned Carbon Nanotubes

and mean diameter 11 ± 03 nm, this report opens a route to prepare aligned SWNTs.

5. PROPERTIES AND APPLICATIONS OF ALIGNED CARBON NANOTUBES Carbon nanotubes are known for their superior mechanical strength and low weight [135], good heat conductance [136], varying electronic properties depending on their helicity and diameter [12], large surface area useful for adsorption of hydrogen or other gas [137], and their ability to emit a cold electron at relatively low voltages due to high aspect ratios and nanometer size tips [20]. Therefore, many future applications have been found in a wide range of fields for being used as field emitters for flat panel displays, vacuum microelectronic devices like microwave power amplifier tubes, nanofield effect transistors [138, 139], nanotube circuits [140], ultrasensitive electrometers [120, 141, 142], nanotweezers [143], nanothermometers [144], and so on. For aligned CNTs, considerable attention is mainly focused on their excellent FE properties. In this section, apart from the FE properties, we will present other potential properties and applications such as the anisotropic electrical transport property, the superamphiphobic property, and energy storage.

5.1. Field Emission Properties Earlier than FE, thermionic emission of electron guns had been a key concept in the electron-beam technique. In thermionic electron emission, the solid electron source (i.e., the cathode) is heated above 2000  C to allow free electrons to escape from the surface [145]. The greatest advantage of this so-called “hot cathode,” usually a heated tungsten (W) filament, is that it works even in non-ultra-high-vacuum (non-UHV) ambiences, which contain vast numbers of gaseous molecules. However, hot cathodes are prone to chemically react with residual water and oxygen to produce tungsten oxides and get thinner and thinner over a long duration through the sublimation of the oxides. In addition, hot cathodes require a power supply for heating, thus making it difficult to construct a compact electron-beam tool. In the mid 1950s, these disadvantage of hot cathodes were be overcome by replacing them with a FE or “cold cathode” [146]. Unfortunately, the electron emission from a FE cathode is exponentially affected by the chemical and morphological states of the electron emitting area, resulting in instability of emitted currents in non-UHV ambiences. This is particularly true of metallic cathodes, which strongly interacted with residual gaseous molecules. Chemically, carbon is far more stable, and hence more robust in non-UHV, than metals. Indeed, great effort has been made to develop field emitters based on carboncontaining materials such as diamond, diamondlike carbon, and tetrahedral-amorphous carbon [147, 148]. The high expectations and promise held by these materials, however, have not yet been matched by their performances. CNTs have emerged as one of the most promising electron field emitters. The power of CNTs as electron sources for displays and lighting devices was amply demonstrated in

Aligned Carbon Nanotubes

the last few years. The superior aspect ratio, good chemical stability, high thermal conductivity, and mechanical stiffness of CNTs are advantageous over conventional semiconductor and metal emitters [149–151]. The investigation of FE properties of individual CNTs and aligned CNT films was originally reported by Rinzler et al. [20] and de Heer et al. [21], respectively. From then on, many research techniques, including improving the method to fabricate CNT alignments, introducing gaseous adsorbates, and controlling the CNT density, have been developed to reduce the operating voltage, enhance the emission current, decrease the fieldscreening effect, lengthen the device lifetimes, increase the stability of electron emission, etc. It is worth noticing that the Samsung CNT FE displays represent an impressive feat and an important milestone toward a fully functional device [152, 153]. They work up to now in diode configuration, which implies that the brightness of a pixel is controlled by varying the potential between emitter and phosphor screen, which is on the order of several kilovolts. Conversely, a triode configuration incorporates a control electrode located near the emitter, and the brightness of the pixel is then controlled by adjusting the potential between cathode and control electrode. Many good FE properties of aligned CNTs have been obtained by different groups although they are still far from practical. Thong et al. [154] reported the emission current density up to 130 mA/cm2 could be reached at an average field strength of 1.925 V/m. This emission current was found to be very stable, with short-term fluctuations of no more than ±1.5%, while the current drifted less than 1.5%. A turn-on field of 1.2 V/m and emission currents of 1 mA/cm2 at 3V/m were achieved on well-aligned CNT emitters [115]. A test of cathode-ray tube lighting elements underway suggested a lifetime of exceeding 10,000 h.

5.1.1. Effect of Gas Adsorbates on FE Properties The FE properties are correlated with the electronic structure of the CNTs, and intrinsic properties such as electrical resistivity, thermoelectric power, and thermal conductivity can be modified by adsorption and desorption of gas adsorbates [120, 142, 155]. Therefore, one way to achieve good emission properties is to introduce adsorbates that might effectively lower the ionization potential and facilitate the extraction of electrons. Many studies have reported the effects of gases on the FE of CNTs [156–159]. Recent experiments done at Motorola [160] indicated that water molecules adsorbed on CNT tips significantly enhance FE current, while O2 and H2 not affect the FE behavior appreciably. The first principles density functional theory calculations [161] showed that the water– nanotube interaction under emission conditions at the tube tip can increase the binding energy appreciably, thereby stabilizing the adsorbate, and lower the ionization potential, thereby making it easier to extract electrons. Wadhawan et al. [162] compared the effects of O2 , Ar, and H2 gases on the FE properties of SWNTs and MWNTs. They found that H2 and Ar gases do not significantly affect the FE properties of SWNTs or MWNTs. O2 exposure temporarily increased the turn-on field of SWNTs by 22% and decreased

11 the FE current by two orders of magnitude. However, the FE properties completely recover after 40 h of FE operation in UHV. For MWNTs, the higher voltage O2 exposure leads to a 43% increase of the turn-on field and reduction of FE current by three orders of magnitude. The recovery in UHV is only partial, indicating that the MWNTs suffer from permanent structural degradation [163]. Kung and Huang [164] demonstrated that the emission current of the CNT arrays treated by O2 and O3 was increased ∼800% along with a decrease of the onset FE voltage from 0.8 to 0.6 V/m. However, a contrary experimental result has been reported by Lee et al. [165] due to different measurement methods of FE properties. Their investigations showed that the turn-on voltages of O2 and N2 gases first decreased and saturated at large gas exposure times, whereas that of H2 gas decreased initially and increased to saturation at large gas exposures. They considered this behavior to be being strongly correlated with the difference of electronegativity of the adsorbed gases. FE is a highly selective process and is extremely sensitive to small variations in the chemical nature and shape and or surroundings of the emitter [166, 167]. This makes a comparison of the results obtained by different groups delicate since the growth, purification, film preparation techniques, and experimental setups differ significantly.

5.1.2. Field-Screening Effect It is known that the FE properties of the tubes could be affected by a field screening effect provoked by the proximity of neighboring tubes. The field screening affect is determined by various factors relative to the nanotube arrays, including the density, the height, and the diameter. According to the prediction of Nilsson et al. [168], the FE will become maximum when the height of the tubes is about onehalf the intertube distance, and arrays with medium densities (∼107 emitters/cm2  show the highest emitted current densities. The experiments by Suh et al. [169] suggested that the FE was optimal when the tube height was similar to the intertube distance. This is a little deviation from the prediction. Teo et al. [170] have demonstrated that the density of carbon nanofibers can be decreased to enhance the FE properties of vertically aligned carbon nanofiber emitters, and the arrays of individual, vertically standing nanofibers spaced twice their height apart had the most desirable FE characteristics and the highest apparent field enhancement factor. Therefore, we can tune the FE properties of CNT film emitters by varying the nanotube density [171]. The optimized field emitter can also be achieved by changing the tube diameter according to the result of the first principle calculations [172].

5.2. Anisotropic Electrical Transport Properties The electrical resistivities of aligned CNTs are anisotropic, being smaller along the tubes than normal to them, because of corresponding differences in the electrical transport [173]. Chauvet et al. [174] observed magnetic anisotropies of the aligned CNTs as well. Recently, electrical, thermal, and structural anisotropies of magnetically aligned single wall CNT films have been obtained by Fischer’s et al. [128, 129].

Aligned Carbon Nanotubes

12 However, all of these aligned CNT films are not as-grown, either they were transferred to a Teflon surface [173, 174] or deposited on a nylon filter membrane [128, 129], which results in a deviation of experimental results. Moreover, these CNT films are imperfectly aligned with some entanglement or curvature of individual tubes (ropes). The disadvantageous factors make the observed results no real reflections of intrinsic properties of those partly aligned CNT films. However, to date the measurements of anisotropic electrical transport properties have not been carried out on as-grown aligned CNT films due to the difficulty in obtaining electrical resistance parallel to the tube axis. Recently, we have developed a simple technique to measure anisotropic electrical transport properties of as-aligned carbon nanotube films [175]. The temperature dependence of relative electrical resistance suggests that most of the well-aligned carbon nanotubes are semiconductive in both directions parallel and perpendicular to the tube axis. The anisotropy R⊥ /R  of electrical resistance increases with decreasing temperature T , reflecting the difference in the longitudinal and transverse hopping rates. The differences of the electrical properties in both directions could be explained by a difference degree of localization of charge carries. The plot of the logarithm of relative resistance against powers of the reciprocal temperature 1/T is closely fitted by three-dimensional variable range conduction. After annealing and Br2 -doping treatments the resistivities of the aligned carbon nanotube films decreased by two orders of magnitude, resulting from the fewer defects and the greater carrier density, respectively.

5.3. Super-“amphiphobic” Properties Wettability is an important factor for a material. Dujardin et al. have studied the wettability of CNTs in detail and found they could be wet and filled by different substances [176, 177]. In general, wettability of solid surfaces is controlled by the chemical composition and the geometrical structures of the surfaces, and it is usually enhanced by surface roughness [178], especially by fractal structures [181]. Recently, superhydrophobic or superlipophobic surfaces, that is, those with a contact angle of water or oil, respectively, that is higher than 150 [178, 179], have attracted much interest due to practical applications. These surfaces have been commonly prepared through the combination of surface roughening and lowering of the surface energy. However, few reports were concerned with super“amphiphobic” surfaces [180], which appear to have both superhydrophobic and superlipophobic properties. As the structure of aligned CNT films is similar to that of these superhydrophobic surfaces, the films are expected to show special wettability features. We reported that the aligned CNT films appear to have super-“amphiphobic” properties; namely, the contact angles for both water and oil are larger than 160 [72]. We demonstrated that the as-grown CNT films are superhydrophobic and superoileophilic. The contact angles for water and rapeseed oil are 1585 ± 15 and 0 ± 10 , respectively. However, the contact angle for water on a film of CNTs lying flat on a surface is 1365 ± 70 , which shows that the aligned

structure of the CNT films is responsible for the superhydrophobic properties. A low free energy surface is required for a superoileophobic surface to be obtained, and this can be realized by modifying the aligned CNT films with fluorinated compounds. Since the surface of CNTs is rather inert, it is very difficult to modify the surface directly. The hot concentrated H2 SO4 and HNO3 mixture (1:1 v/v) was employed to oxidize the aligned CNTs for further chemical modification. The contact angle for water on the aligned CNT film was 128 ± 30 after oxidation. It has been reported that some functional groups, such as hydroxyl and carboxyl, can be realized on the surface of CNTs through chemical oxidation treatments. Hence, the oxidized films were modified through immersion in a methanolic solution of hydrolyzed fluoroalkylsilane. After this modification the aligned CNT films contained fluoroalkylsilane groups, which repel both water and oil. The contact angles for water and rapeseed oil on the films were 171 ± 05 and 161 ± 10 respectively. Hence the film shows both superhydrophobic and superoileophobic properties; namely, it is a super-“amphiphobic” surface. The water droplets move spontaneously and do not come to rest, even when there is little or no apparent tilt of the surface ( Ag > Pt > Au > GaAs. The higher surface coverage on silver relative to gold is a consequence of the SAM structure. On Ag, the SAMs are very similar structurally to on Au(111) but with a lower tilt angle (12
compared to 30
for Au) and a smaller chain-chain distance (4.1 Å for Ag and 5.0 Å for Au) [177]. The SAMs on copper are structurally more complex and very sensitive to the surface preparation of the metal [177]. On platinum, in contrast to Au, Ag, and Cu, the SAM shows a number of gauche transformations which can reversibly be eliminated at both positive and negative potentials [166]. Surfaces where underpotential deposition is employed to deposit a monolayer of one metal onto a gold surface have also been modified with organosulfur SAMs [7, 178]. The underpotential deposition of silver onto gold has been shown to give a more stable SAM than on gold alone [7]. Chain Length The order and stability of alkanethiol monolayers are very sensitive to the length of the alkyl chain. Porter et al. [128] investigated the influence of alkyl chain length using electrochemistry, optical ellipsometry, and infrared spectroscopy. For carbon chains over 11 carbon atoms, the monolayer is densely packed and crystal-like [128]. As the chain length decreases, the structure becomes increasingly disordered with lower packing density and coverage. The transition between crystal-like and disordered occurs between 5 and 11 carbon atoms depending on the measurement technique and surface used. Electrochemical measurements, with the long-chain SAMs, show that the monolayers provide substantial barrier properties to electron transfer and are strongly resistant to ion penetration [128]. An additional influence of chain length on monolayer properties in highly ordered SAMs comes from the so-called “odd-even” effect where the orientation of the terminal group, and the structure of the SAM [179], depends on the number of carbons in the alkyl chain. The orientation of the terminal group has been shown to influence the electrochemical stability of biphenyl terminated SAMs [180], the reactivity of the SAM [181, 182], wetting [183, 184], and its tribological [184] properties. As the densely packed monolayers of long-chain alkanethiols (10 or more alkyl carbons) passivate electrode surfaces, the choice of alkyl chain length is dictated by the application of the SAM. If the SAM is required to prevent access to the underlying gold, then a long-chain alkanethiol is required [185]. However, in many sensing applications, where electrochemistry must occur at the metal below the SAM, either short-chain alkanethiols are required [12] or the order of a long-chain SAM must be disrupted [139, 186, 187]. The consequence of a shorter chain SAM, however, is poorer electrochemical and thermal stability. Recently, Markovich and Mandler [188] have shown that

Alkanethiol Self-Assembled Monolayers

long-chain SAMs can be disrupted through the incorporation of amphiphilic molecules during deposition of the SAM. The Influence of the Terminal Group If only one terminal group was available with alkanethiols, they would be of limited utility. Part of the power of alkanethiol chemistry is the ability to relatively easily tailor the chemistry of the selfassembly molecules to a particular application [189]. Altering the terminal group is perhaps the most important aspect of this ability as it allows the building onto the surface of additional components and controls the surface properties of the SAM-modified interface. The influence of the terminal group on SAM structure depends very much on the size of the terminal group. Terminal groups such as CH3 , OH, CO2 H, NH2 CONH2 , CO2 CH3 , where the cross-sectional area is smaller than that of the alkyl chain (∼20 Å) [98], do not interfere with the packing of the hydrocarbon chains. Hence, these monolayers present a single homogeneous functionality at the exposed surface. In contrast, SAMs with bulky terminal moieties, such as sulfonates and oligo(ethyleneoxides) [190], show significant deviation from √ √ the ( 3 × 3R30
hexagonal lattice due to disruption of the hydrocarbon packing [191, 192]. A consequence of disruption of the chain-chain packing is that no longer does the surface present homogeneous functionality. In the case of oligo(ethyleneoxide)-terminated surfaces their ability to be wetted is affected as a result [189]. Similarity of structure does not necessarily imply a similar function. An example of this relates to how the terminal influences the interactions of the terminal surface of the SAM with the surroundings. Methyl-terminated SAMs present low-surface free energy, hydrophobic surfaces while the high-surface free energy, alcohol-terminated surface is hydrophilic. Furthermore, the properties of the surface can be altered, depending on the terminal group, through changing solution conditions. The effect of solution conditions is demonstrated by a carboxylic acid-terminated SAM, which can be either negatively charged or neutral depending on the pH [193]. The change in the charge state of the interface will then greatly influence the blocking ability of the SAM on a surface. The variation in blocking ability with pH was used by Zhao et al. [194] to measure the surface pKa of (3-mercaptopropionic acid). A pKa titration was performed by electrochemically measuring the decreased ability of ferricyanide to penetrate the carboxylic acid-terminated SAM with increasing pH (as the surface changing from neutral to negatively charged). It is important to note that the surface pKa can be significantly different than the pKa of the molecule in bulk solution [195, 196]. Similar tailoring surface properties have been very important in using SAM-modified surfaces for tasks such as restricting protein adsorption [43, 197], orientating the approach of biomolecules [198, 199], controlled wetting [41, 200], controlling crystal growth [201], and influencing the surface friction [193, 202].

2.2.3. Stability of Alkanethiol Monolayers The application of SAMs to anything other than fundamental studies requires the resultant modified surface to be stable. Self-assembled monolayers have good stability over a

Alkanethiol Self-Assembled Monolayers 0.12 A 0.1

Fraction of pinholes, 1 –Θ

wide range of conditions. Variation in stability between different SAMs is a function of the chain-chain interactions as the gold-thiolate bond is the same. Therefore, any factor that increases the integrity of the SAM will ultimately also improve the stability of the SAM. Thermal stability of aliphatic alkanethiol SAMs has been the subject of several studies. The SAMs have been reported to begin desorbing from gold at temperatures from 75
C for butanethiol [203] to above 100
C for dodecanethiol [204]. There is some inconsistency in the reported desorption temperatures and rates for a given alkanethiol, because of the influence of parameters, such as the smoothness of the gold surface and the method of determining the loss of the alkanethiol. For example, Nuzzo et al. [17] have reported hexadecanethiol desorbs in a range between 170–230
C using X-ray photoelectron spectroscopy and infrared spectroscopy, while Schlenoff et al. [105] use radiolabelling to show complete loss of the SAM at 210
C which began at 100
C. The greater the forces of interaction between the tails of the SAM-forming molecules, the greater the thermal stability of the SAM. Therefore, monolayers where there is hydrogen bonding [205] or - bonding [206] between the alkyl chains exhibit better thermal stability than aliphatic hydrocarbons. Similarly, repulsive forces between terminal groups such as carboxylic acids and sulfonates will result in a lower stability than the equivalent methyl-terminated SAM. The impact of these repulsive terminal interactions will again be influenced by the alkyl chain length, where the influence of the terminal group is diluted as the alkyl chain increases in length. For many applications the electrochemical stability of alkanethiol SAMs is particularly important. Self-assembled monolayers have been shown to be stable to potentials between approximately +1.0 and −1.0 V versus SCE, although this potential window depends on chain length, terminal group, and the quality of the underlying gold surface [38, 40, 207]. Such a potential window is compatible with most electrochemical applications. Outside this range, the thiols are either oxidatively or reductively desorbed [33, 39, 124, 146, 150, 180, 208–210]. The instability at high potentials, however, can be used to control the order of the SAM during formation [124]. Self-assembled monolayers with high integrity and few defects are more stable, requiring more positive or negative potentials to desorb, because ordered regions, where there is a high level of bonding between alkyl chains, require whole sheets of the SAM to be removed at a time rather than individual alkanethiols [109, 150]. The influence of the gold surface on the stability of SAMs is demonstrated in Figure 5 with hexadecanethiol SAMs assembled on gold surfaces with different roughness [109]. The three gold surfaces are gold evaporated onto mica with no annealing (A), polycrystalline bulk gold (B), and atomically smooth gold prepared by template stripping (C) [155]. The roughness of the gold surfaces decreases from surface A to C. The higher the roughness of the surface, the poorer the packing between the alkyl chains and hence the greater the number of defect sites where pinholes, where bare gold is exposed to solution. Figure 5 shows the number of pinholes in the SAM formed on each surface. The pinhole fraction is measured by performing a cyclic voltammogram between −0.5 and 1.5 V versus Ag/AgCl in 0.1 M

23

0.08

0.06

0.04 0.02

B C

0 0

1

2

3

4

5

6

7

8

9

10

Number of scans Figure 5. The variation in the fraction of the SAM surface covered in pinholes as a function of the number of cycles in a cyclic voltammogram on three gold surfaces. The surfaces are (A) gold evaporated onto mica with no annealing, (B) polycrystalline bulk gold, and (C) atomically smooth gold prepared by template stripping. The roughness of the gold surfaces decreases from surface A to C. The number of pinholes in each SAM is measured by performing a cyclic voltammogram between −0.5 and 1.5 V versus Ag/AgCl in 0.1 M H2 SO4 . During this measurement there is a minor amount of further disruption of the SAM. The increase in pinhole fraction with the number of scans gives an indication of the SAM robustness with SAMs formed on smoother surfaces being more robust.

H2 SO4 , the same conditions used for the electrochemical cleaning of gold. In this experiment, as the potential is swept positive, any exposed gold from pinholes will be oxidized. Upon sweeping negative, this gold oxide is reduced and the area under the reduction peak allows the quantification of the amount of pinholes. During this measurement, there is a minor amount of further disruption of the SAM. This is evident in Figure 5 where the pinhole fraction increases as the number of scans increases. Note, however, with the smoother surfaces, where there are larger sheets of the SAM with more chain-chain interactions, the SAM is far more robust to this repeated cycling. The long-term storage stability of SAMs is less well understood. Horn et al. [211] have used reflection-absorption IR spectroscopy (RAIRS) to monitor changes in structure of aliphatic alkanethiols over a 6-month period. Changes in the symmetric and asymmetric CH2 stretching modes observed with the RAIRS are attributed to chemical oxidation of the thiolate species at the point of attachment to the gold. Upon oxidation of the thiolate to either sulfinates ( SO− 2 ) or sulfonates ( SO− 3 ), the SAM is less strongly bound to the gold as determined by their ability to desorb from the surface in solution [212]. The oxidative loss of alkanethiols is particularly problematic for applications that employ short chain alkanethiols, where oxygen and other oxidants can easily access the gold-thiolate bond [213]. The rate of oxidation of the thiol to a sulfonate has been shown to be dependent on the alkyl chain length [213, 214], as well as the terminal group of the SAM [214]. In the case of the terminal group, the rate constant for photo-oxidation has a ratio of 4:2:1 for CH3 :OH:COOH for both long and short chain SAMs [214].

24 In practice, the problem of oxidation can be reduced by storing devices in oxygen-free environments such as vacuum packs. With the depiction of alkanethiol SAMs as ordered, stable, and static monolayers on surfaces, it is easy to forget that SAMs are dynamic systems in which the alkanethiols are capable of moving about on the surface, as is evident from the kinetic studies where after adsorption the SAM slowly reorganizes itself. The dynamic behavior of SAMs is demonstrated by two key observations. First, when a SAM modified gold surface is placed in a different alkanethiol solution, exchange occurs at the grain boundaries of the underlying metal surface [40, 42, 77, 105]. This is called a place exchange reaction. Second, the alkanethiol in a SAM have been shown to surface diffuse to heal gaps of exposed gold [210]. This healing process occurs over the time span of several hours [9, 215–217]. The rate to which defects are healed is independent of the chain length of the alkanethiol, which suggests the head group dominates the adsorption and desorption steps [215] and that the first layer of gold atoms are involved in the monolayer movement [9].

2.2.4. Mixed Monolayers To use SAM surfaces as the base upon which nanofabrication can be performed requires more than one component to be incorporated within the SAM. The formation of homogeneous SAMs from mixtures of alkanethiols would allow the construction of integrated molecular systems where several components are incorporated within a single monolayer. For example, mixed monolayers can allow either large recognition elements or nanoscale building blocks to be spaced apart from each other if necessary. Furthermore, through varying the composition of a mixed SAM the density of attachment points, and hence the surface loading of added components, can be controlled. Mixed monolayers have commonly been produced through the use of a mixture of alkanethiols in the solution to which the substrate is immersed, although place exchange reactions have also been used [40, 42, 105, 169, 218–226]. In a place exchange reaction the exchange is greater in SAMs with more defects [42]. A mixed SAM formed from two components in solution has been regarded as being a reasonably homogeneous mixture of the components [41, 227–229] rather than phase segregation into “islands” as has been observed in Langmuir–Blodgett films [15]. There is some controversy as to whether phase segregation occurs or not. Homogeneity was determined via wetting studies [41], which is supported by STM measurements that suggested no segregation occurs on the scale of greater than 50 nm [230]. However, on a smaller scale some phase segregation has been reported to occur [231]. Scanning tunneling microscopy studies indicate domain-rich regions of individual components of nanometer dimensions can be observed [232, 233]. This phase separation is believed to occur by adsorption/desorption followed by movement. Differences in length and chemical functionality certainly play a key role in any phase separation [231]. For example, Atre et al. [231] showed using wetting studies, FTIR and XPS, with a mixture of HS(CH2 )15 CH2 OH and HS(CH2 )15+m CH3 with m increasing systematically from

Alkanethiol Self-Assembled Monolayers

−6 to +6, that the monolayer underwent structural change from randomly placed protruding OH-terminated chains at low m to phase segregated at high m. However, in contrast to studies that show phase separation STM studies by Takami et al. [234] do not show phase separation even at the nanometer level for mixtures of azobenzene-terminated SAMs and dodecanethiol. This contradictory evidence highlights the many differences in SAM studies such as the surface upon which assembly occurs, the alkanethiol used, and the measurement technique, which play a role in the conclusions reached. With regards to phase separation in mixed SAMs, it appears that generally mixed SAMs form a 2-dimensional alloy, which can decompose into a heterogeneous distribution particularly when the two components are significantly different [118]. When forming mixed SAMs by solution assembly, the ratio of the components in the SAM can be adjusted by varying the ratio of the components in solution. Note, however, that the mole fraction of components in solution is not necessarily the same as the two components on the surface. In the two-component mixture, the more alike the two components the closer the mole fraction in the SAM will be to the mole fraction in solution. If the two components are quite different, the surface mole fraction can be very different to the solution mole fraction. For example, Offord et al. [18, 235] have looked at mixed SAM of octadecyl mercaptan 1 with either butyl mercaptan 2 or tert-butyl mercaptan 3 and a variety of different ratios in solution. Time-of-flight secondary ion mass spectrometry showed that the adsorption efficiency of 1–3 varied in the order 1 > 2 ≫ 3. In the case of a 1:1 mixture of butyl to octadecyl adsorbate in solution, there was a 3–4 fold excess of 1 to 2 and a 40–100 fold excess of 1 to 3 [18]. These excesses are dependent on the solvent used [235]. Therefore, it is clear that the relative solubility of the two adsorbates has an influence on the surface mole fractions. Similar results have also been observed with thermal desorption mass spectrometry [236]. A novel exploitation of the phase separation of mixed SAMs to control the surface loading of a recognition element has recently been developed by Satjapipat et al. [210] (see Fig. 6). They formed a mixed SAM of 3-mercaptopropionic acid (MPA) and mercaptohexanol (MCH). As the two alkanethiols are reasonably different chemically, they tended to separate into domains rich in each of the two components. As MPA desorbs at a lower potential than MCH, MPA was selectively removed through reductive desorption. The result was a surface that contained some bare regions. Before the remaining adsorbed MCH could diffuse across the surface to cover the bare regions, the substrate was placed in a solution of thiolated DNA which assembled onto these regions. This approach represents one method of fabricating interfaces with discrete domains of different components; in essence, this is patterning a SAM. More controlled methods of patterning SAMs are discussed in Section 2.3.

2.3. Patterning of SAMs Many applications in nanotechnology require patterning of the surface chemistry, and hence the controlled positioning of the different components used to modify the surface.

Alkanethiol Self-Assembled Monolayers

25

Selective reductive desorption

Gold Electrode Thiolated DNA

Figure 6. Schematic representation of the procedure used by Satjapitat et al. [210] to fabricate a DNA recognition interface. A mixed SAM of MCH and MPA is formed. The electrode potential is scanned until the MPA is reductively desorbed to leave bare gold. The electrode is then placed in thiolated DNA which adsorbs onto the bare regions.

Applications include biosensors, microelectronics, and optical devices such as displays and waveguides. Many of these applications require the patterns to be as small as possible. Controlled positioning of different components can be achieved using a variety of techniques for patterning SAMs, which have recently begun being developed where resolution can vary from centimeters down to as low as 20 nm [237]. Approaches to patterning of SAMs can be divided into composition methods where the SAM is patterned during its assembly and decomposition methods, where parts of a complete SAM are removed to allow a second component to be added [238]. Decomposition methods, such as photolithography, electron beam writing, and micromachining, have traditionally been more popular for patterning SAMs. New decomposition methods like nanoshaving [237, 239] and a proximity printing technique using focused ion beams developed by Golzhauser et al. [240] have also been developed. However, in recent times with the advent of microcontact printing [43, 81, 83, 84, 86, 107–111, 171, 172, 197, 201, 241–254] and dip-pen lithography [73, 255–261], composition methods are moving to the fore.

2.3.1. Decomposition Methods Photolithography The wide applicability of photolithography in creating features and devices on the microscale has inevitably seen it also used for patterning SAMs. The photooxidation of alkanethiolate can be achieved directly upon exposure of the SAM to UV radiation in the presence of oxygen [262, 263]. Using a chrome mask patterned to allow selective exposure of the surface to UV radiation, patterning of the SAM can be achieved [264, 265]. The oxidized thiols can easily be desorbed from the surface and then a second alkanethiol can be assembled onto the bare surface. Alternatively, the remaining SAM can act as a chemical resist that blocks the access of etchants to the underlying surface: Therefore, exposure of the patterned surface to an etchant will result in the bare surface being etched away, with the SAM-covered regions remaining, to give a microstructure.

Patterns down to 100 nm have been achieved with photolithography by Behm et al. [266] using a microscope to focus the light in projection lithography. Although photolithography is an efficient method of patterning substrates to make microstructures, it has a number of limitations in nanofabrication, especially with regards to developing new nanodevices. First, it requires expensive specialist equipment and clean room facilities. Second, the shallow depth of field requires that only planar substrates can be patterned that limits the range of materials upon which nanostructures can be built. Finally, and perhaps most importantly, the ultimate resolution is dependent on the wavelength of light. This final limitation appears to restrict photolithographical patterning to defining regions on a surface where nanofabrication can occur rather than being involved in the precise positioning of molecular components in nanofabrication. Beam Lithography Beam lithography is a decomposition method that uses beams such as atoms, ions, or electrons to remove the alkanethiolates. Selectivity is again achieved with a mask. The best resolutions appear to be achieved using neutral beams [267, 268]. Berggren and Younkin et al. [267, 269] have patterned features smaller than 100 nm on gold and silicon using neutral cesium beams. Berggren et al. [268] have also shown nanometer resolution in patterning alkanethiols using argon beams. In common with photolithography, different alkanethiols show different susceptibility to the decomposition source, and therefore some tailoring of the technique is required for individual systems. Nanoshaving Truly nanometer-size features can be patterned onto SAMs using one of several scanning probe microscope (SPM) techniques [239, 255] collectively called scanning probe lithography. “Nanoshaving,” developed by Xu and Liu et al. [237, 239] (Fig. 7) involves displacing the adsorbate on a SAM-modified surface using an AFM tip. The tip is scanned at a load higher than the displacement threshold. In this way, patterns of bare metal with feature sizes down to 20 nm can be produced. If the nanoshaving is performed in a solution containing a second alkanethiol, the exposed metal will be modified with the second component; this is called nanografting. An equivalent nanofabrication a)

b)

Molecular transport AFM Tip Scan

0

0.1

µm

Figure 7. The principles of nanoshaving where a) AFM tip is dragged across an alkanethiol modified surface with sufficient force to remove the alkanethiol from the surface. If this nanoshaving is performed in a solution containing a second alkanethiol, then a surface patterned with two alkanethiols can be produced. Feature sizes as small as 20 nm have been fabricate; b) shows a 50 nm wide island of octadecanethiol patterned into a decanethiol SAM. Reprinted with permission from [237], S. Xu and G. Y. Liu, Langmuir 13, 127 (1997). © 1997, American Chemical Society.

Alkanethiol Self-Assembled Monolayers

26 method using STM has also been developed, where the tip potential is sufficiently high to cause thiol desorption. Nanoshaving has also been used for fabricating semiconducting nanowires [270] and to characterize the thickness of novel alkanethiol SAMs [271].

2.3.2. Composition Methods of Patterning Microcontact Printing Microcontact printing is one of a suite of soft-lithography methods for patterning surfaces developed by Xia and Whitesides [86]. The basic principles of microcontact printing are shown in Figure 8. A stamp with relief structure is prepared from a master. The master can be any structure with the appropriate relief. In fact, Deng et al. have shown how a master can simply be made using a normal computer printer and transparencies [253]. The stamp is coated with alkanethiol ink which is evaporated onto the stamp so the printing process is essentially dry. Upon contact of the inked stamp with the surface, a SAM formed at the points of contact. As the SAM only forms where the stamp contacts the surface, the pattern of the stamp is transferred to the solid substrate. In this way, features as small as 100 nm have been fabricated [245]. Placing the patterned surface into a solution containing a different alkanethiol allows a two-component system to be fabricated, which could be used to give either controlled wetting/dewetting [243] locations where the underlying electrode is passivated and others where access is maintained, provide specific sites for crystal nucleation [201] or allow the selective deposition of other materials such as polymers [247] and cells [107, 197]. Rather than deposit a second alkanethiol onto the exposed metal surface, it could be used either as a microelectrode array [252], or the printed SAM could be used as an ultra-thin chemical resist for selective etching to remove the exposed metal surface [244, 250]. The essential advantages of microcontact printing are that it is cheap and easy to set up. Existing fabricated devices can be used as the masters for new stamps and, with some PDMS Master

(a)

Master

PDMS

(b)

PDMS

(c) Thiol

PDMS

(d)

Substrate

(e)

Substrate

(f)

Substrate

Metal (g)

(h)

creativity, quite sophisticated devices can be fabricated. The low cost is because no special clean room facilities or expensive masks are required. One of the reasons that clean room facilities are not required relates to the stamp itself. Typically, it is made of polydimethylsiloxane (PDMS) which is an elastomer with a low surface energy. If the surface to be patterned is contaminated by a speck of dust (which would be a big problem in photolithography), the stamp will conform around the dust particle and the mobility of the alkanethiols may even allow assembly onto the part of the surface covered by the dust particle. Microcontact printing has been performed on nonplanar substrates [242] and even inside tubes [246]. With this emerging technique, issues that have been addressed include how to decrease the size of the pattern [245], what dimensions give a stable stamp that does not collapse upon itself [86, 248], the order of the printed SAM relative to solution assembled monolayers [108], methods of inking the stamp [251], how the ink is dispersed from the stamp to the substrate [249], the influence of the topography of the gold surface [109], and the amount of alkanethiol applied to the stamp [110] on the integrity of the printed SAM. “Dip-Pen” Lithography “Dip-Pen” lithography is an ingenious approach to nanofabricating SAM-based structures with an AFM developed by Piner and Hong et al. (Fig. 9) [255, 272]. Dip-pen lithography exploits the water meniscus that travels with an AFM tip when contacting a surface under ambient conditions. By coating the tip in an alkanethiol, the water meniscus solubilizes some of the alkanethiol and transfers it to the solution substrate upon which it is traveling. In this way, line widths of 30 nm resolution have been achieved. Dip-pen generated templates of DNA have been used to orthogonally assemble nanoparticle building blocks, thus allowing nanofabrication in a third dimension [73]. This was achieved by patterning a gold surface with 16-MCH acid to which a sequence of DNA was attached. Gold nanoparticles were modified with thiolated complementary DNA. Exposure of the patterned surface to the complementary DNA allowed ordered arrays of nanoparticles to be fabricated onto the original gold surface. A similar approach has been used to pattern amine modified polystyrene spheres, which were bound to the MCH-patterned surface via electrostatic interactions [257]. Dip-pen lithography has now been extended beyond alkanethiols to sol–gel based inks [258]. AFM Tip

Substrate

Substrate

Figure 8. Schematic of the process of microcontact printing where: (a) a master which contains a relief structure is coated with the elastomer PDMS, (b) the PDMS is removed from the master and becomes the stamp, (c) the stamp is inked with an alkanethiol, and (d) contacted with a the surface to be patterned, (e) the alkanethiol is transferred to the metal surface at the points of contact. The result is a patterned surface with exposed metal and metal covered in alkanethiol. Either (f) the patterned surface is placed in another alkanethiol solution to give a surface patterned with two different alkanethiols or alternatively in (g) the exposed metal is etched away, using the alkanethiol as a resist, followed by (h) the removal of the remaining alkanethiol.

Molecular transport

Scan Water

Figure 9. Dip-pen lithography where an AFM tip coated in alkanethiol is scanning across a surface under ambient conditions. The meniscus of water that travels with the tip solubilizes the alkanethiol, transferring it to the metal surface to produce patterns.

Alkanethiol Self-Assembled Monolayers

An earlier variant of dip-pen lithography, which produced patterned features of size greater than 10 m, is microwriting [273, 274]. In microwriting, neat alkanethiol is dispensed from the tip of a fine capillary directly onto the gold surface. Control over the distance between the tip and the surface gives some control over line widths that are achieved.

2.3.3. Controlled Positions of Alkanethiol Deposition Controlling the location where a particular alkanethiolate SAM is deposited has also been achieved electrochemically. This is done using an electrode array where the elements of the array are held at different potentials, depending on whether they are to be modified by the SAM in solution or not. Such an approach is important in controlling the modification of different elements of an electrode array with different recognition species. The first demonstration of using electrode potential to control the deposition of alkanethiols was by Tender et al. [275], where interdigitated electrodes were first all coated with HO(CH2 CH2 O)6 (CH2 )11 SH. Sufficient potential was applied to one of the interdigitated electrodes to selectively desorb the SAM. The microelectrodes were then placed into 16-mercaptohexadecane (MHD) to coat the now bare electrode. The MHD-coated electrode allowed selective adsorption of proteins. The crucial aspect of this technology is that different elements in an electrode array can be selectively coated without requiring any alignment [276]. Wang et al. [125] have used the complete opposite approach to selectively coating individual electrodes in an array. By placing a negative potential on the electrode to be coated, the adsorption of the SAM from ethanolic solution onto that electrode is accelerated. The other electrodes, held at open-circuit potential, are modified much more slowly and therefore are only partially coated. Hsueh et al. [277] have also used potential to control which elements in an electrode array are modified with a SAM. The primary difference to the approach of Tender is, rather than use alkanethiols, the electrochemical oxidation of alkylthiosulfates (Bunte salts) to certain electrodes is used. The oxidation of the Bunte salts produces either a disulfide or an alkylsulfide radical at the electrode which is then trapped on the electrode surface to produce a gold-thiolate bond.

3. APPLICATIONS OF SELF-ASSEMBLED MONOLAYERS As stated above there are a host of applications to which SAM modified surfaces have been applied including surface coatings, sensing, molecular electronics, and nanofabrication with nanoscale building blocks. We shall discuss each of these applications in turn.

3.1. Surface Coatings The affinity of alkanethiols for coinage metal surfaces and their subsequent self-assembly onto the surface made use of alkanethiols as surface coatings for metal surfaces an obvious application. One of the attractive features of using alkanethiol chemistry is that the surface coating can be achieved with nanometer control. The alkanethiol surface coatings have been used to give metal surfaces corrosion resistance,

27 changing the frictional properties of surfaces, making them hydrophobic or hydrophilic, giving controlled environments for crystal growth, making surfaces that resist fouling and preparing surfaces for controlled protein and cell growth. Many of these surface-coating technologies lead to further applications, such as surfaces for controlled protein adsorption can then be used for the development of biosensors.

3.1.1. Corrosion Resistance Since metal surfaces can be passivated using close-packed alkanethiol monolayers, several workers have used these monolayers to protect copper [7, 8, 160, 161, 164, 278–284], iron [285–288], and gold [216, 242, 289] surfaces from species in solution. It is the ability of alkanethiol monolayers to protect an underlying metal surface, which is used in microcontact printing to allow etching of exposed gold and not gold coated by the SAM [86, 242, 244, 246]. Laibinis and Whitesides [290] first attempted to prevent the oxidation of copper surfaces coated with alkanethiols when in the presence of atmospheric oxygen. XPS measurements showed that the rate of oxidation was inversely proportional to the alkanethiol chain length. The resistance to corrosion was much greater with alkanethiols where the chain length was greater than 12 carbons due to more closely packed SAMs being formed [281]. In aqueous solutions of NaSO4 , however, even octadecanethiol did not prevent corrosion [291]. As a consequence, further modification of the alkanethiol was required to reduce the coatings permeability [160, 164, 207, 278–280, 283–285]. As example of this approach is by Itoh et al. [160], where copper surfaces were modified with 11-mercapto-1-undecanol. The alkanethiol was then reacted with an alkyltrichlorosilane to form an alkylsiloxane polymer. The formation of a polymer on the copper surface improved the corrosion resistance of the coating.

3.1.2. Surfaces for Controlled Crystal Growth The terminal moiety of a SAM, and the ease to which this can be modified, provides a method of controlling the chemistry of a surface upon which crystallization occurs [201, 292–300]. The ability to easily pattern the SAM using microcontact printing allows the controlled position on a surface where nucleation of crystals occurs. This was elegantly demonstrated by Aizenberg et al. [201], where a SAM of an alkanethiol that promoted crystal growth (symbolized as HS(CH2 )15 X and which was mercaptohexanoic acid X is COOH) was patterned into dots using microcontact printing onto a metal surface (gold, silver, or palladium). The rest of the surface was coated with MCH. Placing this patterned surface into a CaCl2 solution exposed to carbon dioxide and ammonia vapor resulted in calcite crystals forming predominantly on the HS(CH2 )15 X parts of the SAM. Not only can the position of crystal growth on a surface be dictated but also control over whether many or only a single crystal grew at each location was afforded by judicious choice of the feature size, as well as the density and concentration of the crystallization solution. Even more exciting was the demonstration that different X groups in HS(CH2 )15 X resulted in crystal growth from different crystal faces. The ability to control the position of the growth of single crystals on a

28 surface, and also to be able to control the type of crystal, could be exceedingly useful for fabricating ordered arrays of nanoparticles. This concept has already been realized by Qin et al. [301] with arrays of a variety of crystals being grown including CdCl2 and Nile red.

Alkanethiol Self-Assembled Monolayers OH

O

O

3.1.3. Biological Surfaces Surfaces that interact with biological environments have enormous numbers of applications in a large number of fields ranging from biological implants, drug development, to biosensors. An important aspect of these applications is an understanding of how biological molecules and organisms interact with the man-made surfaces to which they are exposed. There is a huge amount of research in these fields which fills many volumes. In this section, the role alkanethiol SAMs have played in these fields will be discussed. Proteins Adsorption The adsorption to proteins to surfaces is exceedingly complex, and although there has been considerable research into the field there is yet to be a complete molecular level understanding [302–306]. Selfassembled monolayers are therefore important in understanding and using proteins adsorption, because they are model systems where the surface properties are well defined and can easily be altered in a known way [43, 49, 238, 307]. Furthermore, protein adsorption mediates subsequent cell adhesion which is important in implants [308, 309] and in controlled growth of cells on surfaces [49, 107, 307]. The adsorption of proteins while retaining the protein’s configuration and activity is also important in the development of biosensors [11, 44, 81, 310]. As direct contact between a metal surface and proteins will cause denaturing of the protein, and hence in the fabrication of protein surfaces, SAMs play a role in not only allowing molecular level control over the immobilization of the protein but also insulate the protein from the metal. For controlled adsorption of proteins onto surfaces, however, nonspecific adsorption needs to be eliminated. Nonspecific adsorption of proteins occurs on many different SAM end groups [190, 238, 309, 311–314] and appears to correlate with the hydrophobicity of the surface [311, 314]. Some end groups, however, have been shown to resist nonspecific adsorption of proteins onto surfaces [190]. The most widely used [107, 197, 315–320] and successful terminal group that resists protein adsorption has been poly(ethylene glycol) with between two and seven ethylene glycol units (see Fig. 10). The polyethylene glycol head group has been shown to resist nonspecific adsorption of a number of proteins with a range of molecular weights and isoelectric points under a wide range of solution conditions [321, 322]. Self-assembled monolayers that resist protein adsorption then allow specific adsorption of a protein of interest, as well as patterning a surface so that the protein adsorbs at well-defined locations. Specific adsorption of a protein for the development of biosensors has most commonly been performed via covalent attachment where either the protein is first derivatized with an alkanethiol and attached to the surface, or alternatively, the surface is modified with a SAM and covalently linking the protein to the SAM (see Section 3.2.1 and Figure 14 for an example of this strategy). An alternative approach is to use biospecific recognition of the protein for a ligand.

O

SH

Figure 10. Structure of an ethylene glycol-terminated alkanethiol that resists protein adsorption. The number of ethylene glycol repeat units typically ranges from 2 to 7.

To achieve the biospecific recognition requires the substrate to restrict nonspecific binding to the surface so that the ligand can interact with the desired protein. The classic example of such an approach in nature is the biotin-avidin affinity reaction. There have been, however, some elegant examples developed in recent years. The Whitesides group prepared a mixed SAM of ethylene glycol end groups and a benzosulfonamide ligand for the specific adsorption of carbonic anhydrase. In a similar strategy, Sigal et al. [315] prepared a mixed SAM terminated with ethylene glycols and nitrilotriacetic acid (NTA). The NTA chelates with Ni(II) to form a complex with two vacant sites on the Ni(II) center. This surface can now selectively bind proteins prepared with a histidine tag as confirmed via surface plasmon resonance. Schlereth et al. [323] have also used a biospecific binding to immobilize the enzyme lactate dehydrogenase onto electrodes where the ligand is a triazine dye. One of the attractive features of the specific adsorption of proteins using biospecific ligands is that the protein is immobilized with control over its orientation. The specific adsorption of proteins onto surfaces not only opens up the opportunity for well-controlled immobilization for biosensors but also provides a surface upon which cell adhesion can occur. Cell Adhesion The analysis of individual cultured cells, and how they interact with target molecules, is important in screening target drugs in genetic engineering and in assessing the toxicological effects of compounds. To use individual cells for any of these applications requires the isolation of large numbers of cells, controlling their distribution, and position in space. The obvious way of achieving ordered arrays of individual cells is to immobilize them on the surface. The adhesion of cells onto surfaces is mediated by proteins of the extracellular matrix (ECM) such as fibronectin, laminin, heparin, vitronectin, and collagen. Therefore, to

Alkanethiol Self-Assembled Monolayers

29

promote controlled cell adhesion, one approach is to first adsorb one or more of these proteins to the surface [81]. Both hydrophilic [324, 325] and ionic [326] SAMs have been used for this purpose; they afford poor control over the adsorption process. A more nanotechnological approach to controlled cell adhesion is to exploit the fact that the attachment of the cells to the extracellular matrix is controlled by specific interactions between the integrin receptors on the cell membranes and peptide sequences of the ECM proteins [327]. As specific peptide motifs are involved in the cell adhesion, surfaces terminated with the appropriate motif, such as ArgGly-Asp, can be used to promote adhesion. Forming a mixed SAM with an ethylene glycol-terminated component and a peptide-terminated component can promote cell adhesion while limiting nonspecific adsorption [316, 328]. Achieving the desired outcome of arrays of individual cells of defined shape, size, and distribution has been achieved using patterned SAMs [329] (see Fig. 11). The substrates, upon which individual cells of a defined shape were cultured, were prepared using microcontact printing. The metal surface was stamped with hexadecanethiol (HS(CH2 )CH3 ) to form the islands upon which the cells were grown. The rest of the gold surface was coated with HS(CH2 )11 (OCH2 )3 OH

30 µm

20 µm

5 µm

10 µm 30 µm

40 µm

to prevent protein adsorption. The islands were coated with appropriate ECM proteins. Cells cultured on the surfaces spread to the size and shape of the island. The size of the islands were shown to influence cell life or death with apoptosis increasing with decreasing island size.

3.2. Sensors The vast majority of the applications of SAMs, which involve nanofabrication, have been in diagnostic devices [10, 11, 44, 51, 81, 87, 310, 330], in particular electrochemical sensors, where a number of elegant nanostructures or integrated molecular systems have been fabricated. The focus on electrochemical sensing is partly related to organosulfur SAMs being formed on an electrode material—gold. Typically, a sensor requires a recognition molecule integrated with a signal transducer. Usually the recognition molecule, which gives the sensor its selectivity, is immobilized over the signal transducer. This arrangement is depicted in Figure 12. The signal transducer determines the extent of the recognition reaction and outputs the signal to an end user. Examples of recognition species include macrocyclic or other types of ligands, enzymes, antibodies, DNA, and even whole cells or tissues. An array of approaches to transduction have been employed including electrochemical [10, 47, 331], acoustic wave [332], optical [333], calorimetric [334], changes in surface force [335–338] or stress [330, 339, 340]. The immobilization of the recognition molecule is the critical step in the fabrication of a sensor. The most common immobilization strategy is to either entrap or covalently attach the recognition molecule within a polymer membrane. Although highly versatile as an approach, the control over the location and density of recognition molecules is poor [331]. Furthermore, if the analyte to be detected is large, then only recognition molecules at the surface of the polymer will interact with the analyte. In contrast, the modification of gold with a SAM can be achieved with molecule level control over the interface, and hence the position of the recognition molecules in space can also be controlled with molecular level precision. That is, the sensing interface can be fabricated using the bottom-up principle of nanotechnology. Recognition Layer

Sample

Analyte

Figure 11. Demonstration of the ability to grow ordered arrays of individual cells of a defined size and shape by forming islands upon which cells will adhere using microcontacting printing. Initially, the surface is patterned with a CH3 -terminated SAM with the rest of the surface coated in an ethylene glycol-terminated SAM. As extracellular matrix CH3 surface, these are the only locaproteins will only adsorb to the tions to which the cells will adhere. The size of the cell islands was shown to influence whether the cells proliferated or underwent apoptosis. Adapted with permission from [329], C. S. Chen et al., Science 276, 1425 (1997). © 1997, American Association for the Advancement of Science.

Transducer

Electrical Signal to user

Other Compounds

Figure 12. Schematic of a sensor showing the recognition layer that gives the sensor its selectivity for the target analyte despite the presence of other molecules in the sample. The transducer determines the extent of the reaction between the recognition component and the analyte and converts this to an electronic signal which can be outputted to the end user.

Alkanethiol Self-Assembled Monolayers

30 The simplest sensors based on alkanethiols are monolayers that either provide selective access to the underlying transducer surface [341, 342] or incorporate a recognition molecule attached to the SAM-modified transducer [10– 13]. Slightly more complicated examples include controlling the spacing of the recognition molecule on the surface [343, 344] or using mixed monolayers to control the microenvironment of the recognition molecule [343]. Controlling the micro-environment may serve the purpose of orientating the recognition molecule [70], preventing adsorption of electrode fouling species [43, 201, 345], or providing a charge exclusion layer to prevent interferences interacting with the electrode [52, 216, 252, 346, 347]. Using the initial monolayer as the base for layer-upon-layer fabrication allows this molecular level control to be extended into a third dimension. This extension into the third dimension may be with a single recognition element [61, 62, 348–357] or with more than one type of recognition molecule that operate cooperatively [348, 358]. The number of roles that the SAM can play in the development of a sensing interface, in many ways represents the level of control afforded by SAMs in nanofabrication as several molecular components are integrated in a controlled way to fulfill a variety of functions. The sophistication to which nanofabrication, using SAMs, is perhaps best demonstrated by enzyme biosensors.

3.2.1. Catalytic (Enzyme) Biosensors Basics of Enzyme Electrodes Enzyme biosensors employ an enzyme as the recognition element. Typically, the enzyme is a redox enzyme, which either oxidize or reduce its substrate. As the enzyme reaction is a catalytic reaction, enzyme biosensors are often classed as catalytic sensors. In an enzyme biosensor, the substrate is the analyte of interest. The analyte reacts with the enzyme and produces a product. Transduction is achieved by monitoring either a molecule consumed in the enzyme reaction or one that is produced. The classic example is a glucose biosensor which uses the enzyme glucose oxidase (GOD) to oxidize glucose in the presence of a mediator to produce gluconolactone and a reduced form of the mediator (see Fig. 13). The role of the mediating species is to complete the catalytic cycle. The enzyme in its oxidized form is reduced as the glucose is oxidized. Therefore, the mediator oxidizes the reduced form of the enzyme and in the process is itself reduced. In nature, the mediator is oxygen with hydrogen peroxide being produced while many of the commercial glucose biosensors use redox species such as ferrocene or ferricyanide [359]. As there are changes in oxidation state in the recognition reaction, it is common for the transduction of enzyme reactions to be electrochemical. In a conventional enzyme biosensor, the reduced form of the mediator is detected amperometrically at the electrode with the current being proportional to the amount of glucose in the sample. These principles can be extended to many other enzymes, where all that is required is an enzyme specific for the analyte and a product of the enzyme reaction that can be transducer. The commercial glucose meters that arise from these ideas usually immobilize the enzyme in a polymer layer over the electrode [360]. Immobilization of enzymes in polymers, however, gives poor control over where the enzyme

Glucose

Gluconolactone

Glucose Oxidase

Mred

Mox

2e-

Electrode Figure 13. Schematic of an electrochemical glucose biosensor, where glucose is oxidized by the enzyme glucose oxidase to produce gluconolactone. In the process, the enzyme is reduced. The reduced enzyme is reoxidized by a mediator. In nature, this mediator is oxygen with hydrogen peroxide being produced as the reduced form of the mediator. The reduced form of the mediator is oxidized at the electrode to give a current proportional to the amount of glucose in the sample.

is in space, which is partly responsible for reproducibility problems associated with enzyme electrodes [361]. Fabricating the Biorecognition Interface Enzyme electrodes fabricated using SAMs have been studied extensively with a view to providing more controlled enzyme immobilization [12, 13, 62, 70, 140, 155, 348–350, 354, 358, 361– 379]. The simplest enzyme electrodes fabricated using SAMs involve the covalent attachment of a single enzyme on a SAM as represented schematically in Figure 14. In this case, the SAM is prepared from 3-MPA and the enzyme is GOD [361, 376]. The reasons for using MPA are two-fold. First, the short chain alkanethiol produces a disordered SAM, which makes the underlying gold electrode electrochemically accessible. Second, the carboxylic acid functionality allows covalent attachment of the enzyme via the formation of a peptide bond after activation. In Figure 14, the activation is achieved using 1-ethyl-3(3-dimethylaminopropyl) carbodiimide hydrochloride (EDC) and N -hydroxysuccinimide (NHS), which yields a succinimide ester intermediate that NHS

EDC

H 3C N CH2CH2CH2N H3C

C

O

NCH2CH3 H 3C

HO N

H O

H 3C

H2CH2CH2C H3CH2CN

S

O

N

O NH

O O

O

S

E

NH2 O

O

O

O

NH

C

E

N

S

S

Figure 14. Reaction scheme for the immobilization of enzymes and other biomolecules onto a carboxylic acid-terminated SAM. In this case, the SAM is composed of 3-mercaptopropionic acid. The carboxylic acid terminated-SAM is activated using EDC and NHS. The succinimide ester-terminated SAM is now susceptible to nucleophilic attack from amines such as found on the surface of biomolecules. The result of the nucleophilic attack is a peptide bond.

Alkanethiol Self-Assembled Monolayers

Examples of SAM-Based Enzyme Electrodes That Employ Nanofabrication From the platform of a single enzyme-modified SAM, quite complicated molecular constructs have been fabricated with multiple functionality integrated into a single layer or multilayer. Multilayers of enzyme systems can be fabricated using a variety of strategies with molecular level control [61, 62, 348, 349, 382]. Riklin and Willner [348] have shown that multiple enzymes can be incorporated into an enzyme electrode with a layer-bylayer approach. Importantly, however, Gooding et al. [358] have shown that in a bienzyme amplification system, using glucose oxidase and glucose dehydrogenase, the response of the final enzyme electrode is sensitive to the relative spatial organization of the two enzymes (see Fig. 15). In this bienzyme system, the glucose is converted into gluconolactone by the glucose oxidase, and the glucose dehydrogenase reduces the gluconolactone back to glucose to create an amplification cycle. In each cycle where the glucose is oxidized to gluconolactone, the mediator a p-benzoquinone is reduced to hydroquinone. The hydroquinone is then detected at the electrode. When the enzyme electrode was made in two layers, one containing glucose oxidase and the other gluconolactone, a different linear range and gain was observed depending on whether GOD was in the inner layer directly adjacent to the electrode or on the outer layer. Improving and/or controlling the transduction of the biorecognition reaction has been the objective of considerable attention with SAM-modified enzyme electrodes [140, 370, 372, 373, 388, 392, 393]. In many examples of SAMbased enzyme electrodes, a redox mediator in used either in the sample solution or attached to the enzyme [12]. Blonder and Willner et al. [372, 388] modulated the access of the mediator to the enzyme interface using nitrospiropyran. Reversible photoisomerism of nitrospiropyran to the cationic nitromerocyanine can be achieved by irradiating with UV light, switching back to nitrospiropyran using light of wavelengths greater than 475 nm. By covalently attaching nitrospiropyran to the enzyme, the access of the charged mediator—ferrocene monocarboxylic acid—can be modulated through charge exclusion by switching between the

a) NAD+ G

GDH NADH

H2Q GOD

GL

Q

b) 60

(iii)

50

i /µA cm-2

is susceptible to nucleophilic attack from amines such as those found on proteins. This is perhaps the most common method of attaching enzymes, although other methods of immobilizing enzymes onto SAMs include other methods of covalent attachment [380, 381], electrostatic adsorption [62], using biotinylated enzymes bound to streptavidin [350, 382–384], using antibody-antigen binding [349, 363–368] and cross-linking [139, 351, 385–391]. Apart from the glutaraldehyde approach, which results in the uncontrolled formation of multilayers of enzyme and hence loses the advantage of controlled immobilization using SAMs, the other methods give a monolayer of enzyme immobilized on the SAM surface that exhibit good reproducibility [376, 378]. Many of the important issues in molecular level fabrication of enzyme electrodes have been reviewed elsewhere [12, 13, 376, 377]. Importantly, for this work on simple SAM-based enzyme electrodes are that the SAM provides a generic base upon which many different enzymes, and other biomolecules, can be immobilized.

31

(iv)

40

(ii)

30 20 10

(i)

0 0

25

50

75

100

[Glucose] /mM Figure 15. The bienzyme electrode fabricated by Gooding et al. [358], which employs the two enzymes GOD and glucose dehydrogenase (GDH) to create an enzyme amplification cycle as shown in a) where the analyte glucose (G) is oxidized to gluconolactone (Gl) by GOD in the presence of the mediator benzoquinone (Q). During the oxidation, the B is reduced to hydroquinone (H2 Q). The Gl is reduced back to G by GDH, in the presence of the enzyme cofactor, reduced nicotinamide adenine dinucleotide (NADH), to complete the amplification cycle; b) shows the influence of the enzyme electrode geometry on the amplification gain where (i) is an enzyme electrode fabricated with glucose alone, (ii) contains both enzymes in the same layer, (iii) the GOD is in the inner layer and the GDH is immobilized in a second layer over the GOD, and (iv) GDH is in the inner layer and GOD in the outer layer. Reprinted with permission from [358], J. J. Gooding et al., Electrochem. Commun. 2, 217 (2000). © 2000, Elsevier Science.

two photoisomers [388]. Improved exclusion of the mediator, and hence improved switching, was achieved by attaching the nitrospiropyran directly to the redox active center of GOD—flavin adenine dinucleotide (FAD) [372]. As the enzyme reaction involves a change in oxidation state, the ultimate goal of enzyme electrode research is to obviate the need for a mediator by oxidizing and reducing the enzyme directly at the electrode. The attractiveness of SAM technology for achieving direct electron transfer is that the enzymes can be immobilized close to the electrode in a highly controlled manner. Furthermore, with judicious choices of self-assembling molecules, the SAM-forming molecule can be used to facilitate the electron transfer [55, 56, 394–397]. Direct electron transfer has been achieved with SAMbased enzyme electrodes where peroxidase enzymes are used [198, 373, 393, 398]. Lötzbeyer et al. [371, 373] have investigated the distance dependence of long-range electron transfer by using hydrogen peroxide-reducing enzymes of different size which are a covalent attachment to short chain alkanethiol SAM-modified electrodes. The focus on peroxidase enzymes is a consequence of their redox active sites being located close to the enzyme surface [398]. With most enzymes, however, the redox active centers are located a sufficient distance from the surface of the glycoprotein to prevent direct electron transfer [399]. For example, in the case of glucose oxidase, the closest approach of the redox active center—FAD—to the enzyme surface is 13 Å

Alkanethiol Self-Assembled Monolayers

32 [400]. Willner et al. [392] have spanned this 13 Å gap with an electrically wired enzyme electrode where a mediator— pyrroloquinoline quinone (PQQ)—was covalently attached to a cystamine-modified gold electrode using EDC. The PQQ-terminated SAM was then bonded to a derivatized version of the FAD cofactor for GOD, N 6 -(2-aminoethyl)FAD, to form a diad. The PQQ/FAD diad was treated with apo-glucose oxidase to provide a glucose enzyme electrode mediated by PQQ. This enzyme electrode had an extremely large linear range—up to 80 mM—consistent with efficient turnover of enzyme [358, 361], and an exceedingly high sensitivity of 300 ± 100 A cm−2 . The elegant-wired enzyme electrode of Willner et al. still represents a mediated enzyme electrode where the ultimate intimate relationship between the enzyme and the mediator is achieved via molecule level fabrication. The transport of electrons between the enzyme and the electrode is still achieved via hopping from the FAD to the PQQ and then to the electrode. Liu et al. [401] are extending this principle to achieve electron tunneling directly to the electrode from the FAD (see Fig. 16). This electrode construct requires the synthesis of norbornylogous bridge-based SAMs as the connector between the electrode and the FAD. Norbornylogous bridges have been shown to allow efficient electron transfer over long distances via the super-exchange mechanism [55, 402, 403]. N 6 -(2-aminoethyl)-FAD is attached to the end of the bridge and the apo-GOD reconstituted over FAD to give active enzyme. In this case, despite the enzyme active on the FAD being embedded within the protein, and being 108
from the electrode, direct electron transfer between the FAD and the electrode is still observed.

3.2.2. Affinity Biosensors Basics of Affinity Biosensors Affinity biosensors rely on a binding reaction between the biorecognition molecule and the analyte. Typical biorecognition molecules used a)

b) Apo-GOD 6 NH C=O

After reconstitution

4

Current (nA)

HOOC

FAD

2 0 -2 -4 -6 -8 -650

Before reconstitution -450

-250

-50

Potential (mV) vs Ag/AgCl S S

Figure 16. Electrode construct designed to achieve direct electron transfer to enzymes such as glucose oxidase. In a) an 18 Å long norbornylogous bridge SAM is formed with the redox active center of glucose oxidase FAD, attached to the end of the bridge, which serves as a conduit for electron transfer between the electrode and the redox active center of the enzyme. Subsequently, the apo-glucose oxidase (apo-GOD) is reconstituted over the FAD to form the active enzyme; b) shows the electrochemistry of the FAD before and after reconstitution. The characteristic FAD electrochemistry shows that FAD 18 Å from the electrode can still be interrogated before and after the active enzyme is reconstituted. This illustrates direct electron transfer to the redox active center of glucose oxidase.

in affinity biosensors are antibodies, sequences of DNA (oligonucleotides) or peptides. Frequently, the analyte to be bound is also a protein or sequence of DNA, although the molecule to be bound can also be a small molecule. Once the binding reaction has occurred, there is still the need to transduce the biorecognition event. Unlike catalytic biosensors, where the biorecognition event produces a molecule that can then be detected, in affinity sensors the analyte simply binds. To transduce such biorecognition events either requires labels, so familiar in the myriad of immunoassay formats, or a transduction method that can detect the change that occurs at the interface. The two most popular methods that allow label-free transduction of affinity biosensors are evanescent wave techniques such as surface plasmon resonance (SPR) [81, 220, 315, 318, 404–425] and piezoelectric acoustic wave devices such as the quartz crystal microbalance (QCM) [31, 320, 332, 422, 425–451], although a variety of other transduction methods that may involve labels have been employed including electrochemical [433, 452–468], and microcantilever [330, 340, 469, 470]. Surface plasmon resonance is an optical method where a thin film of gold or silver (∼50 nm thick) is deposited over one face of a prism. It is this metal film that forms the sensing surface and will be modified with a SAM. When light is launched into the film, such that it is reflected off the prism face coated with the metal layer, the light is coupled into the surface plasmon-polaritons. The surface plasmon-polaritons occur at the outer surface of the metal. The adsorption of molecules onto this surface causes a change in the refractive index of the interface, which affects the amount of light coupled into the surface plasmon mode, and hence a change in the intensity of light reflected into the detector. To use this phenomenon for affinity sensing, the metal surface is modified with the biorecognition molecule. The binding of the analyte causes a change in the surface plasmons and hence transduction is achieved without any labels [359, 471]. Examples of simple SPR transduction for monitoring bioaffinity reactions include peptides for recognizing antibodies [407– 409] or lipoproteins [318, 410, 411], DNA to detect specific sequences of DNA [220, 412, 413, 415, 416, 419–421], and proteins to detect other proteins and small molecules as in immunosensors [315, 320, 406, 408, 409, 414, 422, 424, 425, 450, 451, 472]. Transduction using a QCM is achieved using an oscillating piezoelectric AT cut quartz crystal. The deposition of electrodes, gold in most cases, onto both faces of the quartz slice allows an alternating current to be applied to the crystal. The alternating current causes the crystal to oscillate at a frequency characteristic of the thickness of the crystal slice. Adsorption or desorption of molecules from the crystal surface cause a change in the frequency of oscillation with picogram sensitivity. Thus, by immobilizing the biorecognition molecule onto the surface of one of the gold electrodes, the bioaffinity reaction is transduced by a decrease in the frequency of the QCM [11, 332]. In a similar manner to SPR, affinity biosensors employing QCM transduction have been developed for DNA recognition [426–428, 430, 431, 435–439, 442] and for monitoring antibody-antigen binding [422, 425, 429, 432, 433, 440, 441, 443–449].

Alkanethiol Self-Assembled Monolayers

Fabricating the Biorecognition Interface In a bioaffinity sensor, the recognition molecule must be free to bind with the analyte. Therefore, immobilizing the biorecognition molecule such that it is accessible for binding is exceedingly important. The molecular level control afforded over, not only the immobilization of biomolecules but also the environment in which they are immobilized, makes SAMs very attractive for fabricating bioaffinity interfaces. An example of the control afforded over biomolecule immobilization is the SAM interface used for the immobilization of DNA by Levicky and Steel et al. [70, 473–475] (see Fig. 17). In a DNA sensor, typically a single strand of DNA (ssDNA) is immobilized onto the transducer surface. This immobilized ss-DNA is referred to as the probe strand and is the biorecognition molecule. Hybridization of the immobilized probe DNA with the complementary sequence in solution (the target strand) is the binding reaction that is to be transduced. Levicky et al. [70] synthesized probe DNA with a mercaptohexyl linker at the 5′ end of the DNA. With the thiol linker only located at one end, end point immobilization of the probe DNA is achieved via self-assembly onto a gold surface. The idea of the end-point immobilization was to minimize the decrease in configurational freedom caused by the hybridization. Such minimization is particularly important in DNA biosensors, as the probe and target must be free to rotate around each other for hybridization to ss-DNA, thiolated

Au = HS-(CH2)n-X

Figure 17. Steps involved in the fabrication of the DNA recognition interface prepared by Levicky et al. [70], which allows efficient hybridization of DNA where one of the strands is immobilized onto a gold surface using alkanethiol chemistry. Efficient hybridization is achieved by having endpoint immobilization of the DNA and preventing nonspecific binding of the DNA to the gold electrode by using a diluent layer of 6-MCH.

33 occur. Neutron reflectivity studies, however, indicated that the ss-DNA was lying flat on the gold surface with multiple adsorption points due to the DNA bases complexing with the gold [70, 210, 476, 477]. The resultant hybridization efficiency was low—10% or less. Hybridization efficiency was improved to almost 100% by preventing the nonspecific adsorption of the DNA bases. This was achieved by exposing the DNA-modified surface to 6-mercoptohexanol (MCH). The MCH not only lifted the nonspecifically adsorbed DNA off the surface but the net negative dipole of the alcohol terminus repelled the negatively charged DNA backbone, thus helping to project the probe strand out into solution. Synthesizing a biorecognition molecule with a thiol linker is not always the simplest strategy for immobilizing biorecognition molecules. However, there are a number of reasonably generic methods of fabricating recognition interfaces for affinity sensors which provide favorable conditions for the recognition reaction to occur. One is to use a similar strategy to that used for enzyme biosensors (Fig. 14), where the gold surface is modified with a carboxylic acidterminated SAM and carbodiimide coupling is used to attach the biorecognition molecule. Such an approach has been used for oligonucleotides [379, 478, 479], antibodies [405, 440], and peptides [480–482]. An alternative generic interfacial structure first developed by Häussling, Schmitt, and Spinke et al. [483–485] involves using a biotin-terminated SAM. Biotin has an exceedingly high affinity for avidins, such as strepavidin (Kaff = 1015 M), making the binding virtually irreversible. As each avidin has four binding sites, the ‘plug-and-socket’ aspect of the avidinbiotin system makes it a very important molecular building block for nanofabrication. The usefulness of the avidinbiotin system is enhanced further by the ease of modifying other molecules with biotin. In the case of SAM-based affinity sensors the avidin is bound to the biotinylated monolayer and subsequently biotinylated biorecognition molecules are bound. Such a strategy has been used for the immobilization of antibodies [319, 485], DNA [319, 413, 430], peptides [319, 486] and even enzymes [382–384]. The accessibility of the biorecognition molecule can be controlled by forming mixed SAMs, where a long-chain alkanethiol acts as a spacer to keep the biotin group away from the monolayer surface, and the shorter alkanethiol spaces the biotin coupling points apart [485]. Examples of Affinity Biosensors Relevant to Nanotechnology There have been many examples where SAMs have been employed to make affinity biosensors, primarily because of the ability to control the accessibility for binding of the biorecognition molecule and the compatibility with the SPR and QCM transduction systems. Among these examples are some of the most exquisite examples of nanofabrication that have thus far been realized. We shall discuss a few of these examples in turn. The first example is the fabrication of a simple immunosensor for Salmonella paratyphi, where transduction is achieved using a quartz crystal microbalance [447]. The biorecognition interface is fabricated as in Figure 14 by modifying the gold electrodes on the 10 MHz quartz crystal with 3-MPA followed by activation with EDC and NHS to allow by covalent attachment of antibodies specific for the

Alkanethiol Self-Assembled Monolayers

34 S. paratyphi bacteria. After immobilization of the antibody, the rest of the crystal was coated in bovine serum albumin to prevent nonspecific binding to the crystal surface. The resultant sensor could differentiate between S. paratyphi and other bacteria, including E. coli and other serogroups of Salmonella with a detection limit of 1 7 × 102 cells/mL. A sensor for low-density lipoproteins (LDLs) by Gaus and Hall et al. [318, 410, 411] provides an example where multiple molecule components are incorporated in the one interface to allow differentiation between LDLs and oxidized LDLs. Initial work used a heparin-modified surface to detect LDL adsorption to the interface via surface plasmon resonance, but this lacked the specificity to differentiate between LDLs and the more harmful oxidized LDLs. To improve the ability to differentiate between LDLs and oxidized LDLs, the fifth ligand repeat unit (LR5) of the LDL receptor was genetically engineered and attached to a mercaptoundecanoic acid SAM using EDC/NHS chemistry [410]. Unfolded LR5 was ineffective as an affinity ligand, but refolded LR5 showed a high affinity for native LDLs but little affinity for oxidized LDLs (see Fig. 18). This study showed the potential for using small peptide sequences— 40 amino acids in this case—as highly selective affinity ligands. The one drawback was the instability of the refolded LR5 ligand. To overcome the instability of the affinity ligand, Gaus and Hall investigated a library of short peptide ligands (three to five amino acids long) related to conserved peptide sequences of the LR5 receptor [318, 411]. A more sophisticated recognition interface was used comprised of a mixed SAM of mercaptoundecanoic acid for coupling the amino acids and 1-mercapto-octyl-hexa(ethylene glycol) to resist nonspecific adsorption. Amino acids were coupled

to the SAM on the SPR chip one amino acid at a time to enable in-situ synthesis of peptide sequences by step-wise elongation. Using this combinatorial approach, a library of peptides was assessed for their affinity to LDLs and oxidized LDLs. The short peptide sequences GlyCystineSerAspGlu and GlyLysLys-OH were found to be the most effective for the selective binding of LDLs and oxidized LDLs, respectively, although detection limits were not quite as low as the LR5 ligand. The variation in LDL binding with oxidation levels of the peptide sequences provides a powerful approach to detecting the LDL oxidation levels in an unknown sample which could provide a sensing system to assess a patient’s atheriosclerosis risk [411]. The above example illustrates a recognition interface involving combinatorial synthesis of the recognition element and multiple functionalities incorporated within the recognition interface. The DNA sensor developed by clinical microsensors (CMS) [345], which is illustrated in Figure 19, shows a recognition interface where nanofabrication using SAMs enables the recognition interface to also fulfill several functions. The SAM recognition interface contains two other components apart from thiolated DNA. The majority of the SAM is composed of polyethylene glycol terminated alkanethiol, designed not only to resist nonspecific binding of DNA and proteins, but also to insulate the electrode [107, 190, 316]. Oligophenylethynyl thiols are also present in the SAM. The function of the oligophenylethynyl thiols is to allow communication with the electrode, as these molecules are efficient molecular wires. Upon hybridization between a target nucleic acid and the capture probe, transduction is achieved by also hybridizing a reporter sequence of DNA with the target. The reporter sequence is modified with ferrocene labels. The molecular wires allow oxidation of the

0.5

LDL OxLDL

LDL adsorbed (°SPR shift)

0.4

0.3

0.2

0.1

0 0

5

10

15

20

25

LDL concentration (µg/mL) Figure 18. The surface plasmon resonance response to low-density lipoproteins and oxidized low-density lipoproteins to a SAM-modified interface with the LR5 ligand attached. The SPR response shows the good selectivity of the LR5 ligand for LDLs over oxidized LDLs. Adapted with permission from [410], K. Gaus et al., Analyst 126, 329 (2001). © 2001, The Royal Society of Chemistry.

Figure 19. Schematic of the DNA biosensor of Umek et al. [345]. The SAM is composed of polyethylene glycol-terminated alkanethiol, which resist nonspecific binding of DNA and insulates the electrode, oligophenylethynyl thiols, which act as molecular wires to allow communication with the electrode and thiolated probe DNA molecules. Upon hybridization between a target nucleic acid and the capture probe, transduction is achieved by also hybridizing a reporter sequence of DNA with the target. The reporter sequence is modified with ferrocene labels. The molecular wires allow oxidation of the ferrocene labels. Therefore, recognition of the target strand is transduced by the occurrence of a current signal.

Alkanethiol Self-Assembled Monolayers

ferrocene labels. Therefore, recognition of the target strand is transduced by the occurrence of a current signal. The final example is the ion-channel biosensor currently being commercialized by AMBRI [45, 317, 487–493]. The ion-channel biosensor can arguably be claimed to be the first real nanomachine with moving components. The recognition interface is comprised of a lipid bilayer which is anchored to an underlying gold electrode using alkanethiol chemistry (see Fig. 20). The lipid bilayer contains 10 or more components including gramicidin ion channels. The ion channels in the lower layer of the bilayer are fixed to the electrode while those in the upper layer float free. The ion channels in the upper layer are derivatized to allow transduction of a biorecognition event. In one variant of the biosensor, the upper ion channel is modified with antibody Fab’ fragments. The other half of the antibody-binding site is locked into its position in the bilayer using a thiolterminated membrane-spanning lipid. When there is no analyte present, the floating gramicidin channel moves through the top layer of the bilayer allowing a complete ion channel to form. Opening of the ion channel results in a flow of ions through the bilayer and there is a large increase in conductivity. When analyte is present, the two halves of the Fab-binding site bind to the analyte. This binding event locks the ion channel in a closed position; thus the conductivity decreases. The particularly elegant aspects of this design we first, utilize not only biological molecules for detecting but mimic cellular approaches to transduction and second, the transduction method can be applied to a wide variety of different types of analytes. The system has been used to detect small molecules such as digoxin, hormones, bacteria, and even sequences of DNA [45, 494].

3.2.3. Chemical Sensors The differentiation between a biosensor and a chemical sensor can be made according to the recognition element. In a chemical sensor, the recognition element is of nonbiological origin and therefore refers to a ligand of some sort. There have been many SAM-based chemical sensors and in this section only a few that are interesting examples of nanofabrication will be discussed. Hickman et al. [137] developed the first electrochemical SAM-based pH sensor that contained two terminals: a quinone was used as the pH indicator (having pH sensitive electrochemistry) and pH insensitive ferrocene as a reference electrode. The two redox species were coimmobilized onto a SAM. The oxidation and reduction peaks of the quinone shifted linearly with pH, which allowed accurate measurement of pH between pH 1–11. Beulen et al. [495] elegantly extended this principle to where the different components of a SAM act cooperatively. Rather than form a mixed SAM, a carboxylic acid-ferrocene sulfide was synthesized to give a bifunctional monolayer adsorbate. The normally pH independent electrochemistry of the ferrocene unit became pH dependent due to through space interactions with the adjacent carboxylic acid. Deprotonation of the adjacent carboxylic acid stabilized the oxidized state of the ferrocene resulting in a pH sensitive cathodic shift. Electrochemical sensors for metal ions have been the subject of considerable interest. A sophisticated example involving SAMs is reported by Rubinstein et al. [186], where SAMs

35 a)

Fab

S

MSL

PL MG

ST IG

TL

S S SS S S SS S S S S S SS S S S S S S S S S S S S Gold b) A

S S S S S SS S S S S S S S S S S S S S S S S S S S S Gold Figure 20. The ion-channel biosensor based on a suspended lipid bilayer developed by AMBRI. The self-assembled monolayer consists of spacer thiols (ST), which define an ionic reservoir under the bilayer composed of phospholipids (PL). The bilayer is anchored to the gold electrode using tethered lipids (TL), which span into the lower half of the bilayer. Incorporated within the lower part of the bilayer are immobilized gramicidins (IG), which are immobilized to the electrode with a thiol. In the open state, the IG are associated with mobile gramicidins with pendant biotin groups (MG) in the top part of the bilayer. Membrane spanning lipids (MSL), also with pendant biotin, span across both halves of the bilayer. Biotinylated antibody fragments are attached to both the MG and the MSL using streptavidin (S) as a molecular building block. When there is no analyte (A) present, the MG can move across the bilayer. a) When the MG and IG coincide, there is a large increase in conductivity as ions can pass across the bilayer. b) When the analyte binds to the two Fab antibody fragments, the gramicidins can no longer diffuse freely through the bilayer and the ion channels are closed. The result is a low conductivity.

were used to selectively detect Cu2+ in the presence of Fe3+ . This was achieved using the ligand 2,2′ -thiobisethyl acetoacetate (TBEA). The thiol group allows the assembly onto a gold electrode while the tetradentate ligand satisfied the four coordinate complexation preference of Cu(II). The metal ion sensors were prepared with a mixed SAM of

Alkanethiol Self-Assembled Monolayers

36 octadecylmercaptan (OM) and TBEA. The OM passivated the electrode, while the TBEA provided defect sites where the Cu2+ was bound and amperometrically detected. As the Fe3+ did not effectively bind to the TBEA and the OM prevented direct access to the electrode, Cu2+ could effectively be detected despite the presence of the other metal species. Many other metal ion sensors have been developed where ligands such as crown ethers [496–499], oligoethylene glycols [500–502], and calixarenes [503] are attached to an electrode using SAMs. Calixarenes have also been reported as selective recognition elements for organic molecules such as steroids [404, 456, 504]. The key drawback of using selective ligands is the need to design and synthesize a new ligand for each analyte. Therefore, a simple generic synthetic strategy is required if arrays of sensors are to be developed. Yang and Gooding et al. [481, 482, 505, 506] have addressed this issue using amino acids [505], oligopeptides [482], and polypeptides [481] as the selective ligands for metal ions. A high degree of selectivity for Cu2+ with very low detection limits has been achieved. An attractive feature of this approach is that similar, simple synthetic protocols can be used for most peptide sequences desired. Thus, the amino acids act as the alphabet from which a library of receptor ligands with varying selectivity can be synthesized. An interesting approach to the fabrication of chemical sensors for organic molecules using SAMs has recently been developed, which is analogous to molecular imprinting [507–510]. In molecular imprinting, an artificial receptor is prepared by forming a polymer or sol–gel in the presence of the analyte of interest. The analyte is then removed to produce a three-dimensional porous structure where the cavities match the molecular structure of the analyte. With SAMs of alkanethiols, the same principle can be applied to give two-dimensional receptor sites (Fig. 21) [51, 511–516]. Typically, the binding of the analyte into the receptor site was transduced by a change in capacitance of the interface [513] or the restricted access of an electroactive species to the underlying electrode [512]. One of the difficulties that must be overcome with imprinting receptor sites in SAMs is the mobility of the alkanethiols [216, 217]. Initial attempts to imprint receptor sites resulted in the loss of the receptor property after a single adsorption/desorption cycle [513]. To overcome this problem, a templating molecule with the same shape as the analyte was adsorbed to the surface with a long-chain alkanethiol. The templating molecule was maintained on the surface with pores defined by the long-chain alkanethiols available for binding the analyte. The binding of the analyte is transduced by a decrease in the interfacial capacitance of the electrode. The molecular imprinting approach has been demonstrated for cholesterol [511, 512, 515], barbiturates [513], quinines [516], phenothiazines [511], and adenine triphosphate (ATP) [51].

3.3. Electron Transfer and Molecular Electronics Electron transfer through alkanethiol SAMs, or even individual molecules, has potential applications in improving our fundamental understanding of electron transfer processes [46], molecular electronics [517], and in sensing [458]. Alkanethiol SAMs have a number of advantages that make

OH N

Au

N

HO

SH

1

+ HS

2 1

2

OH

+ N

N

HO

OH

3 3

Figure 21. The alkanethiol SAM-based artificial chemoreceptor. The binding sites are formed by adsorbing the template thiobarbituric acid 1 onto a gold surface in the presence of the matrix molecule 2 which was dodecanethiol. The template must be a similar shape to the analyte and must be able to bind strongly to the gold surface so that it is not displaced by the matrix molecule. When the analyte barbituric acid 3 is exposed to the surface, it can bind in the clefts in the SAM formed by the templating molecule. Reproduced, with permission from [513], V. M. Mirsky et al., Angew. Chem.-Int. Edit. 38, 1108 (1999). © 1999, Wiley-VCH.

them suitable for studying and exploiting long-range electron transfer. These advantages include those that have already been discussed, such as densely packed SAMs giving well-defined distance relationships, the formation of mixed SAMs allowing the spacing of redox centers, the low number of defects or pinholes in well-prepared SAMs ensuring that tunneling processes are probed, and the ease of altering their chemical structure allowing simple exploration of structure-function relationships. A further advantage is that the thiol functionality is ideal for connecting “molecular wires” to bulk materials and hence allowing a connection to the surroundings. Thus, the organosulfur molecules act as the “glue” between electronic materials in molecular devices. Such electronic materials could be bulk surfaces or quantum particles. The alkyl chain or unsaturated conjugated system acts as the conduit for electron transfer.

Alkanethiol Self-Assembled Monolayers

Thus, the electron movement can be through -bonding systems, which are the most frequently investigated systems so far. In the case of -bonding systems, the electron transfer occurs through a superexchange mechanism [403]. Alternatively, electron movement can be through a -bonding system via a typical electron transport mechanism. The actual nature of the chemical connection is exceedingly important and a subject of considerable interest. If the influence of the actual contact can be understood, then the potential to tailor the connection in organic-conductor/semiconductor hybrid systems is greatly expanded and the behavior of the devices will be better controlled. Our understanding of the influence of the contact is poor at this stage but improving rapidly through a number of elegant experiments involving alkanethiol molecules. Cui et al. [518] demonstrated the importance of an alkanethiol having a bond to both electronic materials. A gold surface was modified with either an octadecanedithiol or an octadecanethiol SAM. To the other end of the SAM was bound gold nanoparticles. So in the case of the dithiol, a gold-thiol bond formed with the nanoparticle while in the monothiol case the nanoparticle simply adsorbed onto the methyl terminal of the SAM. The current-voltage (I -V ) properties of these molecular constructs were measured using a conduction probe AFM. There was increase in current of up to four orders of magnitude for a given tip bias when the dithiol was used relative to the monothiol. A similar increase in conductivity was also observed by Selzer et al. [519] with the addition of a bond, in this case between a mercury and a semiconductor surface. The influence of different types of bonds on the conductivity was investigated by Holmlin et al. [520]. A gold surface modified with one alkanethiol SAM and a mercury electrode modified with a second SAM are brought into contact and the I-V behavior measured. With judicious choice of the monolayers, they compare electron transfer across bilayers that form either covalent, hydrogen, or van der Waals bonds. The conclusion was that tunneling rates increased in the order van der Waals < hydrogen bonds < covalent bonds. It is clear from these studies using alkanethiols how important the junction between the organic molecule and the electronic material is. The actual contact resistance of a gold-thiol bond has been estimated at 104  by Wold et al. [521] using a similar conducting AFM-type experiment to Cui et al. [518] but without the gold colloid. The research into tunneling through alkanethiol SAMs has been dominated, not by measuring the influence of the link between the organic molecule and the electronic material, but by the distance dependence of the rate of electron transfer through the organic molecules. Distance dependence has typically been measured using redox monolayers, where a redox active group attached to the terminus of a SAM is probed electrochemically. This approach is well covered in a review by Finklea [46] and in a number of excellent papers since (see, for example, [56, 142, 522, 523]). A wide variety of electroactive groups and redox proteins have been investigated [46]. A number of electrochemical methods allow the measurement of rates in the range of 10−2 –105 s−1 including cyclic voltammetry [524], AC impedance spectroscopy [525] or voltammetry [526], square wave voltammetry [527], and chronoamperometry [528].

37 The distance dependence of the rate of electron transfer can be described using Marcus theory [529], which relies on the overlap of donor and acceptor energy levels in the redox molecule and the electrode. The important parameters in the theory include the reorganization energy (), the energy needed to change the redox center’s structure and solvation sphere during the change in oxidation state, and tunneling parameter . The tunneling parameter is a structure dependent constant, which describes how the electronic coupling decays, as seen in the equation describing the electron transfer rate constant, kET , kET = k0 exp−d where k0 is the rate constant at zero overpotential and d is the distance between the donor and acceptor in this case (the redox species and the electrode). A remarkable consistency in the dependence of the electron transfer with rate is beginning to emerge for aliphatic hydrocarbons. Despite the different redox active species that have been investigated, the value of  is found to be between 0.8–1.1 Å−1 . Similar values for  are also found when the I-V characteristic of alkanethiol SAMs are measured using conducting AFM [521, 530] or the bilayer approach of Holmlin et al. [520, 531]. Although the presence of different coupling chemistry to a redox active species does not appear to significantly alter the value of  [523], large structural rearrangements in the molecule during the electron transfer, such as found with azobenzenes, do cause increases [532]. Conjugation on the other hand definitely causes a decrease in the value of , indicating less decay in the electron transfer rate with distance. Values for  that have been quoted for conjugated systems include for oligo(phenylethynyl) “molecular wire” bridges  of 0.36 Å−1 , around 0.4 Å−1 for oligophenylenevinylene bridges [394], 0.41 Å−1 [521] or 0.61 Å−1 [520] for oligophenylenes, and 0.67 Å−1 [520] for benzyl derivatives of oligophenylene. These lower decay constants indicate useful molecular wires could be developed with conjugated systems.

3.4. Nanofabrication with Other Nanobuilding Blocks In recent times, SAMs have been used as the foundation upon which nanofabrication into the third dimension is achieved using nanoscale building (nanobuilding) blocks such as nanoparticles, nanotubes, and dendrimers.

3.4.1. Nanoparticles Monolayer-protected gold nanoparticles have received considerable attention since a simple synthetic strategy for their production was described by Brust et al. [63]. The gold nanoparticles are typically of the order of 1 to 3 nm (which represents 50 to a few hundred gold atoms) with an alkanethiol coating that serves to stabilize the colloidal solution. The quality of the SAM on the nanoclusters appears to be very similar to SAMs formed on planar surfaces with a similar number of gauche defects [533, 534]. Self-assembled monolayers on nanoclusters are important from a nanoscience perspective as they can be regarded as three-dimensional

38 SAMs. As these three-dimensional SAMs can be investigated in bulk solution, the SAMs can now be investigated using a range of analytical techniques not available to the two-dimensional variety. The alkanethiol coating, however, also provides the opportunity to add specific chemical functionality to the clusters. The first example of functionalized monolayerprotected clusters (MPCs) used a place exchange reaction to substitute -ferrocenyl groups for simple aliphatic alkanethiols [535, 536]. Each cluster contains multiple redox sites which were not strongly electronically coupled with each other. If the clusters are probed electrochemically, under hydrodynamic control with a rotating disk electrode, the voltammograms indicate that nanoclusters are electronically charged in solution so as to equilibrate the potential of the metal-like Au cores of the clusters with the electrode. This double layer charging of the clusters means they are, in fact, nanoelectrodes. As the size of the nanoelectrodes decreases, their double layer charging behavior undergoes a transition from metal-electrolyte interfaces to exhibiting redox chemical behavior [537], which means the cluster capacitance changes from being dominated by electrostatic processes to one dominated by binding interactions. Nanoparticles certainly have the potential to be used as structural components in the fabrication of threedimensional nanostructures. Gold and silver nanoclusters have been used to produce surfaces with well-defined nanostructure and roughness [536], as well as to produce electrodes using an entirely wet process [64]. Connecting nanoclusters together in a well-defined way has been demonstrated by Mirkin et al. [69], where two batches of gold colloidal particles were modified with noncomplementary DNA. Upon addition of an oligonucleotide duplex, with ends complementary to the two sequences attached to the nanoparticles, the nanoparticles self-assemble into aggregates. An attractive feature of this assembly technique is that denaturing of the duplexes allows the aggregates to be redispersed. Oligonucleotide modified gold colloids have also been used by Alivisatos and Loweth et al. [538, 539] where with judicious choices of oligonucleotide sequence, DNA-based assembly of the colloids can be used to produce highly ordered linear assemblies of colloids (Fig. 22). Alternative approaches to nanoparticle assembly have used biotin-modified colloids connected using avidin [540] while Murray’s group have also assembled clusters using Cu2+ to connect gold colloids, which are modified with -carboxylate alkanethiols [65, 541]. These aggregates are surface bound using a carboxylate or thiol-terminated surface. Thermal decomposition of the surface bound MPCs produces films of the core metal. Nanoparticles are beginning to find utility in a variety of applications in sensing, either as the sensors or as labels. An example where the nanoparticle is used as the label is in the detection of DNA hybridization by the Mirkin group, using nanoparticle modified oligonucleotides in the transduction of hybridization [535, 536], and by the Willner group [71]. A recent development by the Mirkin group [74] employed oligonucleotide-modified gold nanoparticles and nanofabrication to transduce DNA hybridization by forming a connection between two gold electrodes. This is achieved using two types of capture probes.

Alkanethiol Self-Assembled Monolayers

(a)

(b)

(c)

(d)

Figure 22. Schematic illustrations and representative transmission electron microscopy images of DNA/gold nanoparticle structures [539]. Image (a) shows a 10-nm homodimer; (b) shows a 10-nm/5 nm heterodimer; (c) shows a 10-nm homotrimer; and (d) shows a 5-nm homotrimer. Reproduced with permission from [539], C. J. Loweth et al., Angew. Chem. Int. Ed. 38, 1808 (1999). © 1999, Wiley-VCH.

One set of capture probes is immobilized in a line in the gap between two microelectrodes. The second set of capture probes is attached to the surface of the gold nanoparticles in solution. A longer target strand is introduced in solution which is complementary to both capture probes. Therefore, hybridization of the target to the surface immobilized probes allows the oligonucleotide-tethered gold colloids to align in the gap between the two microelectrodes, thus increasing the conductivity between the two electrodes. The increase in conductivity can be amplified markedly by the addition of Ag(I) and hydroquinone to give a sensor that can detect concentrations of as low as 0.5 pM. The application of nanoparticles as the platform upon which nanosensors are developed has been demonstrated for fluorescence-based nanosensors, which are capable of detecting metal ions and molecular oxygen within mammalian cells [542–545]. Such sensors use polymer nanoparticles rather than exploit alkanethiol SAMs but they do highlight an application in which three-dimensional monolayers could be employed. A recent example of nanoparticle-based sensors that rely on alkanethiol chemistry is by Maxwell et al. [546]. In this biosensor, single strands of DNA are attached to gold nanoparticles using an alkanethiol linker. To the other end of the DNA is attached a fluorophore, which is attracted to the gold surface. The adsorption of the fluorophore onto the gold quenches the fluorescence. Hybridization with the complementary sequence forces the fluorophore away from the gold

Alkanethiol Self-Assembled Monolayers

nanoparticle with a concomitant increase in fluorescence. Alkanethiol-modified gold nanoparticles are also used by Storhoff et al. [547] in a DNA assay where hybridization causes the aggregation of oligonucleotide-modified gold nanoparticles. With the aggregation, there is a color change from the ruby red of individual nanoparticles to the blue to purple of the aggregated particles.

3.4.2. Carbon Nanotubes The discovery of buckminsterfullerene in 1985 [548] was the first example of the third form of carbon. The equally serendipitous discovery of carbon nanotubes by Ijima [549] followed in 1991. The two discoveries have lead to whole research fields focused on nanocarbon [550]. Nanocarbon structure, and in particular carbon nanotubes, serve as important building blocks in nanofabrication and possibly the fabrication of nanocircuits. Carbon nanotubes can be thought of as graphene sheets rolled up into cylinders. They can be classified into multi-walled nanotubes (MWNT) or single-walled nanotubes (SWNT) depending on the number of graphene sheets that make up the cylinder. Diameters range from 1 nm to tens of nanometers. The MWNT are generally perceived as concentric SWNT of different diameters arranged inside each other [550], although there is electron microscopy evidence that some tubes have a scroll-like structure [551]. The interest in carbon nanotubes from many diverse fields is stimulated by their unique properties, those being that they are relatively defect-free, highly conductive, and the strongest materials known. Because of their relatively structural simplicity, SWNT have been the subject of most attention with regards to an understanding and prediction of nanotube properties, although MWNT are simpler to manufacture [550]. With regards to the electrical properties, which are the source of much interest in molecule electronics, SWNT can be metallic or semiconducting depending on the helices index. The helices index refers to the way the graphene sheet is rolled up [552]. So what of the integration of carbon nanotubes with SAMs? Liu et al. [553] have controlled the deposition of individual nanotubes which is important in the fabrication of molecule circuit components such as the nanotube field effect transistor developed by Huang et al. [554, 555]. To achieve this, silicon surfaces were modified with a tetramethylsilane (TMS) SAM. The TMS is selectively removed using e-beam or AFM lithography and replaced with an amine-terminated SAM. Nanotubes are shown to selectively adsorb onto the amino-functionalized surface. This technique has been used to demonstrate a single nanotube connecting two gold electrodes. The integration of carbon nanotubes with alkanethiol SAMs is at its very beginning. Liu et al. [66–68, 556] have developed a number of methods for organizing SWNT normal to a surface. All these methods involve the shortening of the tubes in a combination of nitric and sulfuric acids. These oxidizing acids attack defects in the walls of SWNT causing a breakage. The now open end of the tube contains carboxylic acid moieties. These ends have then been modified with an alkanethiol, and self-assembled onto a gold surface [66], or aligned directly onto silver surfaces using the affinity of carboxylic acids for Ag+ [67].

39 More recently, aligned nanotubes have been covalently attached to a cysteamine-modified gold surface using carbodiimide coupling to form a peptide bond between the amine-terminated SAM and the carboxylic acid terminated tubes [68, 556]. Yang et al. [557] have formed the same molecular construct, but to the free ends of the nanotubes they have attached microperoxidase MP-11 and shown that direct electron transfer to the enzyme can be achieved. These represent initial experiments into using carbon nanotubes as molecular wires for sensing applications.

3.4.3. Dendrimers Dendrimers are regularly branched, monodispersed macromolecules that have attracted extensive scientific interest of late because of their unusual architecture and properties [558–561]. An example of a dendrimer is the poly(amidoamine) (PAMAM) dendrimer shown in Figure 23. The globular structures have three distinct regions. There is a central core, then a repetitive branching region and finally the terminal moieties [562]. The structure can be precisely controlled at the molecule level and the dendrimer properties can easily be altered by changing the type of core, the amount of branching, and the functionality at the termini. Furthermore, the dendrimers have a low mass density in the interior and are highly permeable to small molecules [562]. It is therefore clear that dendrimers have enormous promise as nanoscale building blocks in supramolecular devices. Applications proposed for dendrimers include catalysis [563], encapsulation and controlled release of drugs [564], chemical sensors [565, 566], and biomimetic materials [567]. Integrating dendrimers with SAMs to give threedimensional structures projecting from a surface has recently attracted some interest. An early example by Tokuhisa and Zhao et al. involved using the molecular gating capabilities of PAMAM dendrimers [562, 568]. The dendrimers were NH2

H2N

H2N

NH HN O

O

O

NH2

NH HN

N N NH

H 2N HN H 2N

H N

O

N

NH2

N

O HN

O

O N H

HN

O N

N H

O

NH

O

H N

H N O

O

N

O NH

O

NH O

H2N

NH2

O

NH

O

HN

N

O

N

N

HN

O

NH

O

N

H 2N

O

HN

O

O

N H

N

N

NH2

NH O

O HN

NH2

N N

NH H2 N

HN

O

O

O NH HN

NH2 H 2N

NH2

Figure 23. Chemical structure of a second generation the PAMAM dendrimer.

Alkanethiol Self-Assembled Monolayers

40 adsorbed directly onto a gold surface where they tended to spread out. If a SAM of hexadecanethiol was adsorbed onto the dendrimer-modified surface, the dendrimers were confined into a more compact structure. Apart from confining the dendrimers, the hexadecanethiol SAM had a second function of passivating the electrode not covered by the dendrimers. The dendrimers were used to prevent access of positively charged ruthenium (III) hexamine complex to the electrode at neutral pH, at which the dendrimer is positively charged. At the same pH, the access of negatively charged ferricyanide was barely impeded by the dendrimer. Altering the pH to 11.0 made the dendrimers neutral and allowed the Ru(NH3 )3+ 6 to pass through the dendrimer and react at the underlying electrode. To confirm that the gating was through the dendrimer rather than defects in the SAM, the terminal amine functionality of the PAMAM dendrimer was converted to 4-(trifluoromethyl)benzamido groups. With these modified dendrimers, the ferricyanide was now prevented from accessing to electrode but the ruthenium hexamine could pass through the dendrimers. Dendrimers have also been coated onto gold nanoparticles by converting some of the amine terminals of the PAMAM into thiols [569]. Dendrimers have been used to assemble nanoparticles onto a planar surfaces in a controlled way [570] and have been patterned onto SAM surfaces using microcontact printing [254]. Dendritic SAMs have also been synthesized and assembled on gold electrodes where the organothiol is the focal point of the dendrimer [571–573]. Early examples of the application of dendrimer-modified surfaces are again beginning to emerge in the field of sensing. Yoon et al. [457] have developed an affinity sensor for the detection of avidin using PAMAM dendrimers, which are modified with biotin as the recognition element for avidin and ferrocene for transduction of the affinity reaction. The amine-terminated dendrimers were covalently attached to a gold electrode modified with 3,3-dithiopropionic acid bis-N -hydroxysuccinimide ester. Glucose oxidase in solution is used as a diffusional tracer. The ferrocene-labeled dendrimer mediates the glucose oxidase reaction with glucose, thus providing a current. Binding of the avidin onto the electrode surface reduces the ability of the glucose oxidase to access the dendrimer, and therefore a reduction in current is observed. The purpose of using dendrimers to fabricate sensors is that they provide a spatially ordered and compact sensing surface. In the example by Yoon et al. [457], the enzyme layer had a surface coverage of 2.5 pmol cm−2 . This enzyme loading is 2.5 times higher than that achieved with enzyme attached directly to a SAM [375], which suggests that attaching enzymes to a dendrimer may allow more sensitive monolayer enzyme electrodes to be developed.

4. CONCLUDING REMARKS The application of SAMs of alkanethiols on surfaces has seen some exquisite and complicated examples of integrated molecule systems developed in pursuit of novel surface coatings, sensors, and molecular electronics. Many of these elegant examples represent the very edge of what has been achieved thus far in nanotechnology with bottom-up fabrication. What has been achieved so far clearly demonstrates that alkanethiol SAMs have the potential to enable

the fabrication of many useful nanodevices by self-assembly in sufficiently large volume and with sufficient control to be commercially viable. In fact, two of the most complicated affinity sensors discussed in this review, which are perhaps some of the best examples of molecular level fabrication of useful devices, namely, the AMBRI biosensor and the CMS DNA sensor, are close to or have already been commercialized. Regardless of the success of the AMBRI and CMS sensors, SAMs have enabled a considerable amount of fundamental nanoscience to be conducted which drives much of the current nanotechnological developments and enables many other nanotechnological dreams to be realized. Furthermore, what we learn about self-assembly and how to control molecular level fabrication from alkanethiol monolayers will be exceedingly valuable in developing other self-assembly systems. This brings us to the future. The author believes the research into SAMs will follow two key directions. First, more advanced integration of nanoscale building blocks such as the nanoparticles, nanotubes, and dendrimers, with SAMs will certainly be developed. Second, alternative self-assembly systems will be developed that have the many attractive features of alkanethiol chemistry outlined throughout this article but which solve the long-term stability problems that alkanethiol chemistry have. Certainly, some researchers are already directing their attention to the problem of stability through new alkanethiols which increase the forces of attraction between the alkyl chains [205–207] or through self-assembly systems which form more robust bonds with the underlying substrate using more attachment points [499] or different chemistry [574]. So the application of SAMs is expected to grow in the future.

GLOSSARY Alkanethiol The family of compounds of the general formula HS (CH2 )n X where the thiol headgroup selfassembles onto coinage metal surfaces to form selfassembled monolayers, the aliphatic chain contributes to the order of the monolayer, and the X terminal group defines the surface properties of the monolayer. Biosensor A portable analytical device that integrates a biorecognition molecule, such as an enzyme, an antibody, or DNA, with a signal transducer. The biorecognition molecule gives the biosensor the selectivity for the analyte while the transducer determines the extent of the biorecognition reaction and converts this to an electronic signal which is outputted to the user. Microcontact printing A method of patterning selfassembled monolayers where a stamp that is patterned with the desired relief is inked with self-assembling molecules and contacted with a substrate to give a patterned self-assembled monolayer on the substrate surface. Scanning probe lithography The suite of techniques for patterning self-assembled monolayers using atomic force microscopy, scanning tunneling microscopy or associated techniques. Self-assembled monolayer (SAM) Ordered monomolecular film that is spontaneously formed upon immersion of a solid into a solution of an amphifunctional molecule.

Alkanethiol Self-Assembled Monolayers

ACKNOWLEDGMENTS I would like to thank Dr. Jason Harper for reading this article so carefully. His contribution improved it considerably. I would also like to thank Dr. Katharina Gaus for useful discussion and assistance with the figures.

REFERENCES 1. N. C. Seeman and A. M. Belcher, Proc. Natl. Acad. Sci. U. S. A. 99, 6451 (2002). 2. IBM Research: Atomic and Nano Technology http://www.research. ibm.com/atomic 3. J. L. O’Brien, S. R. Schofield, M. Y. Simmons, R. G. Clark, A. S. Dzurak, N. J. Curson, B. E. Kane, N. S. McAlpine, M. E. Hawley, and G. W. Brown, Phys. Rev. B 64, 1401 (2001). 4. D. L. Allara, Biosens. Bioelectron. 10, 771 (1995). 5. D. S. Goodsell, in “American Scientist,” Vol. 88, p. 230. 2000. 6. A. Ulman, Chem. Rev. 96, 1533 (1996). 7. G. K. Jennings and P. E. Laibinis, Colloid Surf. A-Physicochem. Eng. Asp. 116, 105 (1996). 8. M. Ishibashi, M. Itoh, H. Nishihara, and K. Aramaki, Electrochim. Acta 41, 241 (1996). 9. G. E. Poirier and M. J. Tarlov, J. Phys. Chem. 99, 10966 (1995). 10. D. Mandler and I. Turyan, Electroanalysis 8, 207 (1996). 11. T. Wink, S. J. van Zuilen, A. Bult, and W. P. van Bennekom, Analyst 122, 43R (1997). 12. J. J. Gooding and D. B. Hibbert, TrAC 18, 525 (1999). 13. I. Willner and E. Katz, Angew. Chem.-Int. Edit. 39, 1181 (2000). 14. D. Cahen and G. Hodes, Adv. Mater. 14, 789 (2002). 15. A. Ulman, “An Introduction to Ultrathin Organic Films From Langmuir-Blodgett to Self-Assembly.” Academic Press, London, 1991. 16. R. G. Nuzzo and D. L. Allara, J. Am. Chem. Soc. 105, 4481 (1983). 17. R. G. Nuzzo, F. A. Fusco, and D. L. Allara, J. Am. Chem. Soc. 109, 2358 (1987). 18. D. A. Offord, C. M. John, and J. H. Griffin, Langmuir 10, 761 (1994). 19. C. Jung, O. Dannenberger, Y. Xu, M. Buck, and M. Grunze, Langmuir 14, 1103 (1998). 20. Y. S. Shon, C. Mazzitelli, and R. W. Murray, Langmuir 17, 7735 (2001). 21. E. B. Troughton, C. D. Bain, W. G. M., D. L. Allara, and M. D. Porter, Langmuir 4, 365 (1988). 22. E. Katz, N. Itzhak, and I. Willner, J. Electroanal. Chem. 336, 357 (1992). 23. H. Y. Lee, Z. L. He, C. L. Hussey, and D. L. Mattern, Chem. Mat. 10, 4148 (1998). 24. J. Noh, T. Murase, K. Nakajima, H. Lee, and M. Hara, J. Phys. Chem. B 104, 7411 (2000). 25. H. Schonherr and G. J. Vancso, Langmuir 15, 5541 (1999). 26. M. H. Schoenfisch and J. E. Pemberton, Langmuir 15, 509 (1999). 27. M. H. Dishner, J. C. Hemminger, and F. J. Feher, Langmuir 12, 6176 (1996). 28. T. Nakamura, H. Kondoh, M. Matsumoto, and H. Nozoye, Langmuir 12, 5977 (1996). 29. T. Ishida, N. Choi, W. Mizutani, H. Tokumoto, I. Kojima, H. Azehara, H. Hokari, U. Akiba, and M. Fujihira, Langmuir 15, 6799 (1999). 30. A. Ihs, K. Uvdal, and B. Liedberg, Langmuir 9, 733 (1993). 31. M. Niwa, M. Date, and N. Higashi, Macromolecules 29, 3681 (1996). 32. P. Talonen, G. Sundholm, S. Floate, and R. J. Nichols, PCCP Phys. Chem. Chem. Phys. 1, 3661 (1999). 33. K. V. Gothelf, J. Electroanal. Chem. 494, 147 (2000). 34. C. J. Zhong and M. D. Porter, J. Am. Chem. Soc. 116, 11616 (1994).

41 35. J. L. Trevor, K. R. Lykke, M. J. Pellin, and L. Hanley, Langmuir 14, 1664 (1998). 36. M. W. J. Beulen, B. H. Huisman, P. A. vander Heijden, F. van Veggel, M. G. Simons, E. Biemond, P. J. de Lange, and D. N. Reinhoudt, Langmuir 12, 6170 (1996). 37. C. J. Zhong, R. C. Brush, J. Anderegg, and M. D. Porter, Langmuir 15, 518 (1999). 38. H. O. Finklea, S. Avery, M. Lynch, and T. Furtsch, Langmuir 3, 409 (1987). 39. M. M. Walczak, D. D. Popenoe, R. S. Deinhammer, B. D. Lamp, C. Chung, and M. D. Porter, Langmuir 7, 2687 (1991). 40. W. R. Everett, T. L. Welch, L. Reed, and I. Fritsch-Faules, Anal. Chem. 67, 292 (1995). 41. C. D. Bain and G. M. Whitesides, J. Am. Chem. Soc. 110, 6560 (1988). 42. S. E. Creager, L. A. Hockett, and G. K. Rowe, Langmuir 8, 854 (1992). 43. M. Mrksich and G. M. Whitesides, Annual Reviews of Biophysical Biomolecular Structures 25, 55 (1996). 44. S. Ferretti, S. Paynter, D. A. Russell, K. E. Sapsford, and D. J. Richardson, Trac-Trends Anal. Chem. 19, 530 (2000). 45. B. A. Cornell, V. L. B. Braach-Maksvytis, L. G. King, P. D. J. Osman, B. Raguse, L. Wieczorek, and R. J. Pace, Nature 387, 580 (1997). 46. H. O. Finklea, Electroanal. Chem. 19, 109 (1996). 47. J. J. Gooding, F. J. Mearns, W. Yang, and J. Liu, Electroanalysis 15, 81 (2003). 48. M. Bartz, N. Weber, J. Kuther, R. Seshadri, and W. Tremel, Chem. Commun. 2085 (1999). 49. M. Mrksich, Chem. Soc. Rev. 29, 267 (2000). 50. D. Hernandez-Santos, M. B. Gonzalez-Garcia, and A. C. Garcia, Electroanalysis 14, 1225 (2002). 51. V. M. Mirsky, Trac-Trends Anal. Chem. 21, 439 (2002). 52. I. Bontidean, C. Berggren, G. Johansson, E. Csoregi, B. Mattiasson, J. A. Lloyd, K. J. Jakeman, and N. L. Brown, Anal. Chem. 70, 4162 (1998). 53. V. Braach-Maksvytis and B. Raguse, J. Am. Chem. Soc. 122, 9544 (2000). 54. M. Mrksich, Curr. Opin. Chem. Biol. 6, 794 (2002). 55. T. T. Wooster, P. R. Gamm, W. E. Geiger, A. M. Oliver, A. J. Black, D. C. Craig, and M. N. Paddon Row, Langmuir 12, 6616 (1996). 56. S. Creager, C. J. Yu, C. Bamdad, S. O’Connor, T. MacLean, E. Lam, Y. Chong, G. T. Olsen, J. Luo, M. Gozin, and J. F. Kayyem, J. Am. Chem. Soc. 121, 1059 (1999). 57. F. R. F. Fan, J. P. Yang, L. T. Cai, D. W. Price, S. M. Dirk, D. V. Kosynkin, Y. X. Yao, A. M. Rawlett, J. M. Tour, and A. J. Bard, J. Am. Chem. Soc. 124, 5550 (2002). 58. F. R. F. Fan, J. P. Yang, S. M. Dirk, D. W. Price, D. Kosynkin, J. M. Tour, and A. J. Bard, J. Am. Chem. Soc. 123, 2454 (2001). 59. A. L. Plant, M. Gueguetchkeri, and W. Yap, Biophys. J. 67, 1126 (1994). 60. J. K. Cullison, F. M. Hawkridge, N. Nakashima, and S. Yoshikawa, Langmuir 10, 877 (1994). 61. E. J. Calvo, C. Danilowicz, and A. Wolosiuk, J. Am. Chem. Soc. 124, 2452 (2002). 62. J. Hodak, R. Etchenique, E. J. Calvo, K. Singhal, and P. N. Bartlett, Langmuir 13, 2708 (1997). 63. M. Brust, M. Walker, D. Bethell, D. J. Schiffrin, and R. Whyman, J. Chem. Soc.-Chem. Commun. 801 (1994). 64. M. D. Musick, D. J. Pena, S. L. Botsko, T. M. McEvoy, J. N. Richardson, and M. J. Natan, Langmuir 15, 844 (1999). 65. W. P. Wuelfing, F. P. Zamborini, A. C. Templeton, X. G. Wen, H. Yoon, and R. W. Murray, Chem. Mat. 13, 87 (2001). 66. Z. F. Liu, Z. Y. Shen, T. Zhu, S. F. Hou, L. Z. Ying, Z. J. Shi, and Z. N. Gu, Langmuir 16, 3569 (2000).

42 67. B. Wu, J. Zhang, Z. Wei, S. M. Cai, and Z. F. Liu, J. Phys. Chem. B 105, 5075 (2001). 68. X. Nan, Z. N. Gu, and Z. F. Liu, J. Coll. Interface Sci. 245, 311 (2002). 69. C. A. Mirkin, R. L. Letsinger, R. C. Mucic, and J. J. Storhoff, Nature 382, 607 (1996). 70. R. Levicky, T. M. Herne, M. J. Tarlov, and S. K. Satija, J. Am. Chem. Soc. 120, 9787 (1998). 71. F. Patolsky, K. T. Ranjit, A. Lichtenstein, and I. Willner, Chem. Commun. 1025 (2000). 72. E. M. Boon, D. M. Ceres, T. G. Drummond, M. G. Hill, and J. K. Barton, Nat. Biotechnol. 18, 1096 (2000). 73. L. M. Demers, S. J. Park, T. A. Taton, Z. Li, and C. A. Mirkin, Angew. Chem.-Int. Edit. 40, 3071 (2001). 74. S. J. Park, T. A. Taton, and C. A. Mirkin, Science 295, 1503 (2002). 75. J. I. Siepmann and I. R. McDonald, in “Self-Assembled Monolayers of Thiols,” Vol. 24, p. 206. Academic Press, New York, 1998. 76. S. Y. Jiang, Mol. Phys. 100, 2261 (2002). 77. G. E. Poirier, Chem. Rev. 97, 1117 (1997). 78. G.-Y. Liu, S. Xu, and S. Cruchon-Dupeyrat, in “Self-Assembled Monolayers of Thiols,” Vol. 24, p. 82. Academic Press, New York, 1998. 79. D. A. Smith, C. Robinson, J. Kirkham, J. Zhang, and M. L. Wallwork, Rev. Anal. Chem. 20, 1 (2001). 80. N. K. Chaki and K. Vijayamohanan, Biosens. Bioelectron. 17, 1 (2002). 81. M. Mrksich and G. M. Whitesides, TIBTECH 13, 228 (1995). 82. N. B. Bowden, M. Weck, I. S. Choi, and G. M. Whitesides, Acc. Chem. Res. 34, 231 (2001). 83. Y. N. Xia, X. M. Zhao, and G. M. Whitesides, Microelectron. Eng. 32, 255 (1996). 84. H. A. Biebuyck, N. B. Larsen, E. Delamarche, and B. Michel, IBM J. Res. Develop. 41, 159 (1997). 85. D. Qin, Y. N. Xia, J. A. Rogers, R. J. Jackman, X. M. Zhao, and G. M. Whitesides, in “Microsystem Technology in Chemistry and Life Science,” Vol. 194, p. 1. 1998. 86. Y. Xia and G. M. Whitesides, Angew. Chem. Int. Ed. 37, 551 (1998). 87. A. N. Shipway, E. Katz, and I. Willner, Chem. Phys. Chem. 1, 18 (2000). 88. M. Brust and C. J. Kiely, Colloid Surf. A-Physicochem. Eng. Asp. 202, 175 (2002). 89. V. Chechik, R. M. Crooks, and C. J. M. Stirling, Adv. Mater. 12, 1161 (2000). 90. Y. Li, J. Huang, R. T. McIver, Jr., and J. C. Hemminger, J. Am. Chem. Soc. 114, 2428 (1992). 91. H. O. Finklea, in “Encyclopeadia of Analytical Chemistry.” John Wiley & Sons, New York, 2000. 92. D. E. King, J. Vac. Sci. Technol. A-Vac. Surf. Films 13, 1247 (1995). 93. H. Ron, S. Matlis, and I. Rubinstein, Langmuir 14, 1116 (1998). 94. J. T. Woodward, M. L. Walker, C. W. Meuse, D. J. Vanderah, G. E. Poirier, and A. L. Plant, Langmuir 16, 5347 (2000). 95. C. Yan, A. Golzhauser, M. Grunze, and C. Woll, Langmuir 15, 2414 (1999). 96. D. A. J. Rand and R. Woods, J. Electroanal. Chem. 31, 29 (1971). 97. J. C. Hoogvliet, M. Dijksma, B. Kamp, and W. P. van Bennekom, Anal. Chem. 72, 2016 (2000). 98. C. D. Bain, E. B. Troughton, Y.-T. Tao, J. Evall, G. M. Whitesides, and R. G. Nuzzo, J. Am. Chem. Soc. 111, 321 (1989). 99. H. Schessler, D. S. Karpovich, and G. J. Blanchard, J. Am. Chem. Soc. 118, 9645 (1996). 100. W. Pan, C. J. Durning, and N. J. Turro, Langmuir 12, 4469 (1996). 101. D. S. Karpovich, H. Schessler, and G. J. Blanchard, in “SelfAssembled Monolayers of Thiols,” Vol. 24, p. 44. Academic Press, New York, 1998. 102. R. C. Thomas, L. Sun, R. M. Crooks, and A. J. Ricco, Langmuir 7, 620 (1991).

Alkanethiol Self-Assembled Monolayers 103. M. Buck, M. Grunze, F. Eisert, J. Fischer, and F. Trager, J. Vac. Sci. Technol. A 10, 926 (1992). 104. D. S. Karpovich and G. J. Blanchard, Langmuir 10, 3315 (1994). 105. J. B. Schlenoff, M. Li, and H. Ly, J. Am. Chem. Soc. 117, 12528 (1995). 106. D. S. Karpovich and G. J. Blanchard, J. Chem. Educ. 72, 466 (1995). 107. R. Singhvi, A. Kumar, G. P. Lopez, G. N. Stephanopoulos, D. I. C. Wang, G. M. Whitesides, and D. E. Ingber, Science 264, 696 (1994). 108. N. B. Larsen, H. Biebuyck, E. Delamarche, and B. Michel, J. Am. Chem. Soc. 119, 3017 (1997). 109. D. Losic, J. J. Gooding, and J. G. Shapter, Langmuir 17, 3307 (2001). 110. D. Losic, J. G. Shapter, and J. J. Gooding, Electrochem. Commun. 3, 722 (2001). 111. M. Liebau, J. Huskens, and D. N. Reinhoudt, Adv. Funct. Mater. 11, 147 (2001). 112. O. Chailapakul, L. Sun, C. Xu, and R. M. Crooks, J. Am. Chem. Soc. 115, 12459 (1993). 113. M. H. Dishner, J. C. Hemminger, and F. J. Feher, Langmuir 13, 2318 (1997). 114. H. Morgner, Langmuir 13, 3990 (1997). 115. H. Kariis, E. Smela, K. Uvdal, M. Wirde, U. Gelius, and B. Liedberg, J. Phys. Chem. B 102, 6529 (1998). 116. M. H. Dishner, P. Taborek, J. C. Hemminger, and F. J. Feher, Langmuir 14, 6676 (1998). 117. T. Y. B. Leung, M. C. Gerstenberg, D. J. Lavrich, G. Scoles, F. Schreiber, and G. E. Poirier, Langmuir 16, 549 (2000). 118. G. E. Poirier and E. D. Pylant, Science 272, 1145 (1996). 119. G. E. Poirier, J. Vac. Sci. Technol. B 14, 1453 (1996). 120. Y. W. Yang and L. J. Fan, Langmuir 18, 1157 (2002). 121. R. Bilewicz and M. Majda, Langmuir 7, 2794 (1991). 122. P. Krysinski, R. V. Chamberlain II, and M. Majda, Langmuir 10, 4286 (1994). 123. R. Bilewicz, T. Sawaguchi, R. V. Chamberlain II, and M. Majda, Langmuir 11, 2256 (1995). 124. D. E. Weisshaar, B. D. Lamp, and M. D. Porter, J. Am. Chem. Soc. 114, 5860 (1992). 125. J. Wang, M. Jiang, A. M. Kawde, and R. Polsky, Langmuir 16, 9687 (2000). 126. F. Y. Ma and R. B. Lennox, Langmuir 16, 6188 (2000). 127. M. M. Walczak, C. Chung, S. M. Stole, C. A. Widrig, and M. D. Porter, J. Am. Chem. Soc. 113, 2370 (1991). 128. M. D. Porter, T. B. Bright, D. L. Allara, and C. E. D. Chidsey, J. Am. Chem. Soc. 109, 3559 (1987). 129. C. D. Bain, H. Biebuyck, and G. M. Whitesides, Langmuir 5, 723 (1989). 130. R. G. Nuzzo, B. R. Zegarski, and L. H. Dubois, J. Am. Chem. Soc. 109, 733 (1987). 131. R. G. Nuzzo, L. H. Dubois, and D. L. Allara, J. Am. Chem. Soc. 112, 3145 (1990). 132. M. A. Bryant and J. E. Pemberton, J. Am. Chem. Soc. 113, 3629 (1991). 133. M. A. Bryant and J. E. Pemberton, J. Am. Chem. Soc. 113, 8284 (1991). 134. M. Hasan, D. Bethell, and M. Brust, J. Am. Chem. Soc. 124, 1132 (2002). 135. C. E. D. Chidsey, G. Y. Liu, P. Rowntree, and G. Scoles, J. Chem. Phys. 91, 4421 (1989). 136. C. A. Alves, E. L. Smith, and M. D. Porter, J. Am. Chem. Soc. 114, 1222 (1992). 137. J. J. Hickman, D. Ofer, P. E. Laibinis, W. G. M., and M. S. Wrighton, Science 252, 688 (1991). 138. G. Wittstock, R. Hesse, and W. Schuhmann, Electroanalysis 9, 746 (1997).

Alkanethiol Self-Assembled Monolayers 139. X.-D. Dong, J. Lu, and C. Cha, Bioelectrochem. Bioenerg. 36, 73 (1995). 140. J. J. Gooding, V. Praig, and E. A. H. Hall, Anal. Chem. 70, 2396 (1998). 141. H. O. Finklea and D. D. Hanshew, J. Am. Chem. Soc. 114, 3173 (1992). 142. K. Weber, L. Hockett, and S. Creager, J. Phys. Chem. B 101, 8286 (1997). 143. D. F. Yang and M. Morin, J. Electroanal. Chem. 441, 173 (1998). 144. D. F. Yang, C. P. Wilde, and M. Morin, Langmuir 12, 6570 (1996). 145. D. F. Yang and M. Morin, J. Electroanal. Chem. 429, 1 (1997). 146. M. M. Walczak, C. A. Alves, B. D. Lamp, and M. D. Porter, J. Electroanal. Chem. 396, 103 (1995). 147. L. H. Guo, J. S. Facci, G. McLendon, and R. Mosher, Langmuir 10, 4588 (1994). 148. Z. Hou, N. L. Abbott, and P. Stroeve, Langmuir 14, 3287 (1998). 149. Z. Hou, S. Dante, N. L. Abbott, and P. Stroeve, Langmuir 15, 3011 (1999). 150. S.-S. Wong and M. D. Porter, J. Electroanal. Chem. 485, 135 (2000). 151. M. Hegner, P. Wagner, and G. Semenza, Surface Science 291, 39 (1993). 152. P. Wagner, M. Hegner, H. J. Guntherodt, and G. Semenza, Langmuir 11, 3867 (1995). 153. P. Wagner, F. Zaugg, P. Kernen, M. Hegner, and G. Semenza, J. Vac. Sci. Technol. B 14, 1466 (1996). 154. J. Mazurkiewicz, F. J. Mearns, D. Losic, C. Rogers, J. G. Shapter, and J. J. Gooding, J. Vac. Sci. Technol. B 20, 2265 (2002). 155. D. Losic, J. G. Shapter, and J. J. Gooding, Aust. J. Chem. 54, 643 (2001). 156. G. Gillen, J. Bennett, M. J. Tarlov, and F. Burgess, Anal. Chem. 66, 2170 (1994). 157. G. Gillen, S. Wight, J. Bennett, and M. J. Tarlov, Appl. Phys. Lett. 65, 534 (1994). 158. S. C. Chang, I. Chao, and Y. T. Tao, J. Am. Chem. Soc. 116, 6792 (1994). 159. A. Ulman, J. Mat. Ed. 11, 205 (1989). 160. M. Itoh, H. Nishihara, and K. Aramaki, Journal of the Electrochemical Society 141, 2018 (1994). 161. Y. Q. Feng, W. K. Teo, K. S. Siow, Z. Q. Gao, K. L. Tan, and A. K. Hsieh, Journal of the Electrochemical Society 144, 55 (1997). 162. J. Scherer, M. R. Vogt, O. M. Magnussen, and R. J. Behm, Langmuir 13, 7045 (1997). 163. H. Ron, H. Cohen, S. Matlis, M. Rappaport, and I. Rubinstein, J. Phys. Chem. B 102, 9861 (1998). 164. R. Haneda and K. Aramaki, Journal of the Electrochemical Society 145, 2786 (1998). 165. Z. Mekhalif, J. Riga, J. J. Pireaux, and J. Delhalle, Langmuir 13, 2285 (1997). 166. M. A. Hines, J. A. Todd, and P. Guyotsionnest, Langmuir 11, 493 (1995). 167. P. Lang, Z. Mekhalif, B. Rat, and F. Garnier, J. Electroanal. Chem. 441, 83 (1998). 168. H. O. Finklea, K. Yoon, E. Chamberlain, J. Allen, and R. Haddox, J. Phys. Chem. B 105, 3088 (2001). 169. F. P. Zamborini, S. M. Gross, and R. W. Murray, Langmuir 17, 481 (2001). 170. J. C. Love, D. B. Wolfe, M. L. Chabinyc, K. E. Paul, and G. M. Whitesides, J. Am. Chem. Soc. 124, 1576 (2002). 171. A. Carvalho, M. Geissler, H. Schmid, B. Michel, and E. Delamarche, Langmuir 18, 2406 (2002). 172. D. B. Wolfe, J. C. Love, K. E. Paul, M. L. Chabinyc, and G. M. Whitesides, Appl. Phys. Lett. 80, 2222 (2002). 173. C. W. Sheen, J.-X. Shi, J. Maartensson, A. N. Parikh, and D. L. Allara, J. Am. Chem. Soc. 114, 1514 (1992). 174. H. Ohno, L. A. Nagahara, S. Gwo, W. Mizutani, and H. Tokumoto, Jpn. J. Appl. Phys. Part 2—Lett. 35, L512 (1996). 175. T. Baum, S. Ye, and K. Uosaki, Langmuir 15, 8577 (1999).

43 176. Q. Zhang, H. Z. Huang, H. X. He, H. F. Chen, H. B. Shao, and Z. F. Liu, Surface Science 440, 142 (1999). 177. P. E. Laibinis, G. M. Whitesides, D. L. Allara, Y. T. Tao, A. N. Parikh, and R. G. Nuzzo, J. Am. Chem. Soc. 113, 7152 (1991). 178. S. Y. Lin, T. K. Tsai, C. M. Lin, C. H. Chen, Y. C. Chan, and H. W. Chen, Langmuir 18, 5473 (2002). 179. T. W. Li, I. Chao, and Y. T. Tao, J. Phys. Chem. B 102, 2935 (1998). 180. Y. T. Long, H. T. Rong, M. Buck, and M. Grunze, J. Electroanal. Chem. 524, 62 (2002). 181. V. J. Angelico, S. A. Mitchell, and V. H. Wysocki, Anal. Chem. 72, 2603 (2000). 182. K. V. Wolf, D. A. Cole, and S. L. Bernasek, Langmuir 17, 8254 (2001). 183. R. Colorado, R. J. Villazana, and T. R. Lee, Langmuir 14, 6337 (1998). 184. S. Lee, A. Puck, M. Graupe, R. Colorado, Y. S. Shon, T. R. Lee, and S. S. Perry, Langmuir 17, 7364 (2001). 185. M. Ohtani, T. Sunagawa, S. Kuwabata, and H. Yoneyama, J. Electroanal. Chem. 429, 75 (1997). 186. I. Rubinstein, S. Steinberg, Y. Tor, A. Shanzer, and J. Sagiv, Nature 332, 426 (1988). 187. T. Nagaoka, Z. D. Chen, H. Okuno, M. Nakayama, and K. Ogura, Anal. Sci. 15, 857 (1999). 188. I. Markovich and D. Mandler, Analyst 126, 1850 (2001). 189. P. E. Laibinis, B. J. Palmer, S.-W. Lee, and G. K. Jennings, in “Self-Assembled Monolayers of Thiols,” Vol. 24, p. 1. Academic Press, New York, 1998. 190. K. L. Prime and G. M. Whitesides, J. Am. Chem. Soc. 115, 10714 (1993). 191. S.-C. Chang, I. Chao, and Y. T. Tao, J. Am. Chem. Soc. 116, 6792 (1994). 192. S. D. Evans, U. E., A. Ulman, and N. Feris, J. Am. Chem. Soc. 113, 4121 (1991). 193. D. V. Vezenov, A. Noy, L. F. Rozsnyai, and C. M. Lieber, J. Am. Chem. Soc. 119, 2006 (1997). 194. J. W. Zhao, L. Luo, X. Yang, E. K. Wang, and S. J. Dong, Electroanalysis 11, 1108 (1999). 195. S. E. Creager and J. Clarke, Langmuir 10, 3675 (1994). 196. K. Kim and J. Kwak, J. Electroanal. Chem. 512, 83 (2001). 197. M. Mrksich, C. S. Chen, Y. N. Xia, L. E. Dike, D. E. Ingber, and G. M. Whitesides, Proc. Nat. Acad. Sci. U.S.A. 93, 10775 (1996). 198. L. Jiang, C. J. McNeil, and J. M. Cooper, Angew. Chem. Int. Ed. Engl. 34, 2409 (1995). 199. J. Wei, H. Y. Liu, H. Yamamoto, Y. He, and D. H. Waldeck, J. Am. Chem. Soc. 124, 9591 (2002). 200. C. D. Bain and G. M. Whitesides, Science 240, 62 (1988). 201. J. Aizenberg, A. J. Black, and G. M. Whitesides, Nature 398, 495 (1999). 202. C. D. Frisbie, L. F. Rozsnyai, A. Noy, M. S. Wrighton, and C. M. Lieber, Science 265, 2071 (1994). 203. N. Sandhyarani and T. Pradeep, Vacuum 49, 279 (1998). 204. E. Delamarche, B. Michel, H. Kang, and C. Gerber, Langmuir 10, 4103 (1994). 205. R. S. Clegg, S. M. Reed, and J. E. Hutchinson, J. Am. Chem. Soc. 120, 2486 (1998). 206. K. Murty, M. Venkataramanan, and T. Pradeep, Langmuir 14, 5446 (1998). 207. T. Kim, K. C. Chan, and R. M. Crooks, J. Am. Chem. Soc. 119, 189 (1997). 208. C. A. Widrig, C. Chung, and M. D. Porter, J. Electroanal. Chem. 310, 335 (1991). 209. T. Ito, J. Electroanal. Chem. 495, 87 (2001). 210. M. Satjapipat, R. Sanedrin, and F. M. Zhou, Langmuir 17, 7637 (2001). 211. A. B. Horn, D. A. Russell, L. J. Shorthouse, and T. R. E. Simpson, J. Chem. Soc. Faraday Trans. 92, 4759 (1996).

44 212. R. D. English, M. J. Van Stipdonk, R. C. Sabapathy, R. M. Crooks, and E. A. Schweikert, Anal. Chem. 72, 5973 (2000). 213. D. A. Hutt and G. J. Leggett, J. Phys. Chem. 100, 6657 (1996). 214. E. Cooper and G. J. Leggett, Langmuir 14, 4795 (1998). 215. R. L. McCarley, D. J. Dunaway, and R. J. Willicut, Langmuir 9, 2775 (1993). 216. F. P. Zamborini and R. M. Crooks, Langmuir 14, 3279 (1998). 217. J. K. Schoer and R. M. Crooks, Langmuir 13, 2323 (1997). 218. S. W. Tamchang, H. A. Biebuyck, G. M. Whitesides, N. Jeon, and R. G. Nuzzo, Langmuir 11, 4371 (1995). 219. M. S. Fleming and D. R. Walt, Langmuir 17, 4836 (2001). 220. E. A. Smith, M. J. Wanat, Y. F. Cheng, S. V. P. Barreira, A. G. Frutos, and R. M. Corn, Langmuir 17, 2502 (2001). 221. T. Kakiuchi, K. Sato, M. Iida, D. Hobara, S. Imabayashi, and K. Niki, Langmuir 16, 7238 (2000). 222. S. W. Chen, J. Phys. Chem. B 104, 663 (2000). 223. A. Ishida and T. Majima, Chem. Commun. 1299 (1999). 224. C. Chung and M. Lee, J. Electroanal. Chem. 468, 91 (1999). 225. K. Kajikawa, M. Hara, H. Sasabe, and W. Knoll, Jpn. J. Appl. Phys. Part 2—Lett. 36, L1116 (1997). 226. R. W. Zehner and L. R. Sita, Langmuir 13, 2973 (1997). 227. C. D. Bain, J. Evall, and G. M. Whitesides, J. Am. Chem. Soc. 111, 7155 (1989). 228. C. D. Bain, J. Evall, and G. M. Whitesides, J. Am. Chem. Soc. 111, 7164 (1989). 229. A. Ulman, S. D. Evans, Y. Shnidman, R. Sharma, J. E. Eilers, and J. C. Cheng, J. Am. Chem. Soc. 113, 1499 (1991). 230. E. Delamarche, B. Michel, B. H. A., and C. Gerber, Adv. Mater. 8, 719 (1996). 231. S. V. Atre, B. Liedberg, and D. L. Allara, Langmuir 11, 3882 (1995). 232. S. J. Stranick, A. N. Parikh, Y.-Y. Tao, D. L. Allara, and P. S. Weiss, J. Phys. Chem. 98, 7636 (1994). 233. D. Hobara, S. Takayuki, S.-I. Imabayashi, and T. Kaiuchi, Langmuir 15, 5073 (1999). 234. T. Takami, E. Delamarche, B. Michel, H. Wolf, and H. Ringsdorf, Langmuir 11, 3876 (1995). 235. D. A. Offord, C. M. John, M. R. Linford, and J. H. Griffin, Langmuir 10, 883 (1994). 236. T. Shibue, T. Nakanishi, T. Matsuda, T. Asahi, and T. Osaka, Langmuir 18, 1528 (2002). 237. S. Xu and G.-Y. Liu, Langmuir 13, 127 (1997). 238. S. W. Tam-Chang and I. Iverson, in “Adsorption and Its Applications in Industry and Environmental Protection. Vol I: Applications in Industry,” Vol. 120, p. 917. Elsevier, Amsterdam, 1999. 239. G. Y. Liu, S. Xu, and Y. L. Qian, Acc. Chem. Res. 33, 457 (2000). 240. A. Golzhauser, W. Geyer, V. Stadler, W. Eck, M. Grunze, K. Edinger, T. Weimann, and P. Hinze, J. Vac. Sci. Technol. B 18, 3414 (2000). 241. A. Kumar and G. M. Whitesides, Science 263, 60 (1994). 242. R. J. Jackman, J. L. Wilbur, and G. M. Whitesides, Science 269, 664 (1995). 243. Y. N. Xia and G. M. Whitesides, J. Am. Chem. Soc. 117, 3274 (1995). 244. E. Kim, A. Kumar, and G. M. Whitesides, J. Electrochem. Soc. 142, 628 (1995). 245. Y. Xia and G. M. Whitesides, Ad. Mater. 7, 471 (1995). 246. J. A. Rogers, R. J. Jackman, and G. M. Whitesides, J. Microelectromech. Sys. 6, 184 (1997). 247. Z. Huang, P.-C. Wang, A. G. Mac Diarmid, Y. Xia, and G. M. Whitesides, Langmuir 13, 6480 (1997). 248. E. Delamarche, H. Schmid, B. Michel, and H. Biebuyk, Ad. Mater. 9, 741 (1997). 249. E. Delamarche, H. Schmid, A. Bietsch, N. B. Larsen, H. Rothuizen, B. Michel, and H. Biebuyck, J. Phys. Chem. B 102, 3324 (1998).

Alkanethiol Self-Assembled Monolayers 250. A. J. Black, K. E. Paul, J. Aizenberg, and G. M. Whitesides, J. Am. Chem. Soc. 121, 8356 (1999). 251. L. Libioulle, A. Bietsch, H. Schmid, B. Michel, and E. Delamarche, Langmuir 15, 300 (1999). 252. H. X. He, Q. G. Li, Z. Y. Zhou, H. Zhang, S. F. Y. Li, and Z. F. Liu, Langmuir 16, 9683 (2000). 253. T. Deng, H. K. Wu, S. T. Brittain, and G. M. Whitesides, Anal. Chem. 72, 3176 (2000). 254. P. Ghosh, W. M. Lackowski, and R. M. Crooks, Macromolecules 34, 1230 (2001). 255. R. D. Piner, J. Zhu, F. Zu, S. Hong, and C. A. Mirkin, Science 283, 661 (1999). 256. Y. Li, B. W. Maynor, and J. Liu, J. Am. Chem. Soc. 123, 2105 (2001). 257. L. M. Demers and C. A. Mirkin, Angew. Chem.-Int. Edit. 40, 3069 (2001). 258. M. Su, X. G. Liu, S. Y. Li, V. P. Dravid, and C. A. Mirkin, J. Am. Chem. Soc. 124, 1560 (2002). 259. Y. Li, M. Ben, and J. Liu, China J. Inorg. Chem. 18, 75 (2002). 260. J. Hyun, S. J. Ahn, W. K. Lee, A. Chilkoti, and S. Zauscher, Nano Lett. 2, 1203 (2002). 261. Y. M. Jung, S. J. Ahn, E. R. Kim, and H. Lee, J. Korean Phys. Soc. 40, 712 (2002). 262. M. J. Tarlov, R. F. Donald, J. Burgess, and G. Gillen, J. Am. Chem. Soc. 115, 5305 (1993). 263. J. Huang and J. C. Hemminger, J. Am. Chem. Soc. 115, 3342 (1993). 264. L. F. Rozsnyai and M. S. Wrighton, Chem. Mater. 8, 309 (1996). 265. L. F. Rozsnyai and M. S. Wrighton, Langmuir 11, 3913 (1995). 266. J. M. Behm, K. R. Lykke, M. J. Pellin, and J. C. Hemminger, Langmuir 12, 2121 (1996). 267. K. K. Berggren, R. Younkin, E. Cheung, M. Prentiss, A. J. Black, G. M. Whitesides, D. C. Ralph, C. T. Black, and M. Tinkham, Adv. Mater. 9, 52 (1997). 268. K. K. Berggren, A. J. Bard, J. L. Wilbur, J. D. Gillaspy, A. G. Helg, J. J. Mc Clelland, S. L. Rolston, W. D. Phillips, M. Prentiss, and G. M. Whitesides, Science 269, 1255 (1995). 269. R. Younkin, K. K. Berggren, K. S. Johnson, M. Prentiss, D. C. Ralph, and G. M. Whitesides, Appl. Phys. Lett. 71, 1261 (1997). 270. A. B. Chwang, E. L. Granstrom, and C. D. Frisbie, Adv. Mater. 12, 285 (2000). 271. T. L. Brower, J. C. Garno, A. Ulman, G. Y. Liu, C. Yan, A. Golzhauser, and M. Grunze, Langmuir 18, 6207 (2002). 272. S. H. Hong, J. Zhu, and C. A. Mirkin, Langmuir 15, 7897 (1999). 273. A. Kumar, H. A. Biebuyck, N. L. Abbott, and G. M. Whitesides, J. Am. Chem. Soc. 114, 9188 (1992). 274. A. Kumar, N. L. Abbott, E. Kim, H. A. Biebuyck, and G. M. Whitesides, Acc. Chem. Res. 28, 219 (1995). 275. L. M. Tender, R. L. Worley, H. Y. Fan, and G. P. López, Langmuir 12, 5515 (1996). 276. L. M. Tender, K. A. Opperman, P. D. Hampton, and G. P. López, Adv. Mater. 10, 73 (1998). 277. C. C. Hsueh, M. T. Lee, M. S. Freund, and G. S. Ferguson, Angew. Chem.-Int. Edit. 39, 1228 (2000). 278. R. Haneda, H. Nishihara, and K. Aramaki, Journal of the Electrochemical Society 144, 1215 (1997). 279. M. Itoh, H. Nishihara, and K. Aramaki, Journal of the Electrochemical Society 142, 3696 (1995). 280. R. Haneda and K. Aramaki, Journal of the Electrochemical Society 145, 1856 (1998). 281. G. K. Jennings, J. C. Munro, T. H. Yong, and P. E. Laibinis, Langmuir 14, 6130 (1998). 282. O. Azzaroni, M. Cipollone, M. E. Vela, and R. C. Salvarezza, Langmuir 17, 1483 (2001). 283. D. Taneichi, R. Haneda, and K. Aramaki, Corrosion Sci. 43, 1589 (2001).

Alkanethiol Self-Assembled Monolayers 284. Z. L. Quan, S. H. Chen, and S. L. Li, Corrosion Sci. 43, 1071 (2001). 285. K. Nozawa, H. Nishihara, and K. Aramaki, Corrosion Sci. 39, 1625 (1997). 286. K. Nozawa and K. Aramaki, Corrosion Sci. 41, 57 (1999). 287. K. Aramaki and T. Kikuchi, Journal of the Electrochemical Society 147, 1734 (2000). 288. K. Aramaki, Corrosion Sci. 41, 1715 (1999). 289. I. Touzov and C. B. Gorman, Langmuir 13, 4850 (1997). 290. P. E. Laibinis and G. M. Whitesides, J. Am. Chem. Soc. 114, 9022 (1992). 291. Y. Yamamoto, H. Nishihara, and K. Aramaki, J. Electrochem. Soc. 143, 436 (1993). 292. M. Tanahashi and T. Matsuda, J. Biomed. Mater. Res. 34, 305 (1997). 293. A. Y. Lee, A. Ulman, and A. S. Myerson, Langmuir 18, 5886 (2002). 294. P. Jiang, Z. F. Liu, and S. M. Cai, Langmuir 18, 4495 (2002). 295. A. M. Travaille, J. Donners, J. W. Gerritsen, N. Sommerdijk, R. J. M. Nolte, and H. van Kempen, Adv. Mater. 14, 492 (2002). 296. R. A. Fischer, U. Weckenmann, C. Winter, J. Kashammer, V. Scheumann, and S. Mittler, J. Phys. IV 11, 1183 (2001). 297. J. Kuther, M. Bartz, R. Seshadri, G. B. M. Vaughan, and W. Tremel, J. Mater. Chem. 11, 503 (2001). 298. J. Aizenberg, J. Chem. Soc.-Dalton Trans. 3963 (2000). 299. C. Winter, U. Weckenmann, R. A. Fischer, J. Kashammer, V. Scheumann, and S. Mittler, Chem. Vapor Depos. 6, 199 (2000). 300. J. Aizenberg, A. J. Black, and G. M. Whitesides, J. Am. Chem. Soc. 121, 4500 (1999). 301. D. Qin, Y. Xia, B. Xu, H. Yang, C. Zhu, and G. M. Whitesides, Ad. Mater. 11, 1433 (1999). 302. J. J. Ramsden, Chem. Soc. Rev. 24, 73 (1995). 303. V. P. Zhdanov and B. Kasemo, Surf. Rev. Lett. 5, 615 (1998). 304. N. D. Spencer, Chimia 52, 598 (1998). 305. J. Talbot, G. Tarjus, P. R. Van Tassel, and P. Viot, Colloid Surf. A-Physicochem. Eng. Asp. 165, 287 (2000). 306. A. R. Halperin and D. E. Leckband, C. R. Acad. Sci. Ser. IV-Phys. Astrophys. 1, 1171 (2000). 307. K. Bhadriraju and C. S. Chen, Drug Discovery Today 7, 612 (2002). 308. B. D. Ratner, J. Biomed. Mater. Res. 27, 837 (1993). 309. M. Lestelius, B. Liedberg, and P. Tengvall, Langmuir 13, 5900 (1997). 310. N. K. Chaki and K. Vijaymohanan, Biosens. Bioelectron. 17, 1 (2002). 311. K. L. Prime and G. M. Whitesides, Science 252, 1164 (1991). 312. M. Mrksich, L. E. Dike, J. Tien, D. E. Ingber, and G. M. Whitesides Exp. Cell Res. 235, 305 (1997). 313. R. R. Seigel, P. Harder, R. Dahint, M. Grunze, F. Josse, M. Mrksich, and G. M. Whitesides, Anal. Chem. 69, 3321 (1997). 314. G. P. Lopez, H. Biebuyck, R. Harter, A. Kumar, and G. M. Whitesides, J. Am. Chem. Soc. 115, 10774 (1993). 315. G. B. Sigal, C. Bamdad, A. Barberis, J. Strominger, and G. M. Whitesides, Anal. Chem. 68, 490 (1996). 316. C. Roberts, C. S. Chen, M. Mrksich, V. Martichonok, D. E. Ingber, and G. M. Whitesides, J. Am. Chem. Soc. 120, 6548 (1998). 317. G. Krishna, J. Schulte, B. A. Cornell, R. Pace, L. Wieczorek, and P. D. Osman, Langmuir 17, 4858 (2001). 318. K. Gaus and E. A. H. Hall, Anal. Chim. Acta 470, 3 (2002). 319. M. Riepl, K. Enander, B. Liedberg, M. Schaferling, M. Kruschina, and F. Ortigao, Langmuir 18, 7016 (2002). 320. C. C. Wang, H. Wang, Z. Y. Wu, G. L. Shen, and R. Q. Yu, Anal. Bioanal. Chem. 373, 803 (2002). 321. M. Mrksich, G. B. Sigal, and G. M. Whitesides, Langmuir 11, 4383 (1995). 322. G. B. Sigal, M. Mrksich, and G. M. Whitesides, J. Am. Chem. Soc. 120, 3464 (1998).

45 323. D. D. Schlereth and K. R. P. H, J. Electroanal. Chem. 431, 285 (1997). 324. G. P. Lopez, W. M. Albers, S. L. Schreiber, R. Carroll, E. Peralta, and G. M. Whitesides, J. Am. Chem. Soc. 115, 5877 (1993). 325. S. Margel, E. A. Vogler, F. L., T. Watt, S. Haynie, and D. Y. Sogah, J. Biomed. Mater. Res. 27, 1463 (1993). 326. D. Kleinfeld, K. H. Kahler, and P. E. Hockberger, J. Neurosci. 8, 4098 (1988). 327. D. D. Schlaepfer and T. Hunter, Trends Cell Biol. 8, 151 (1998). 328. B. T. Houseman and M. Mrksich, J. Org. Chem. 63, 7552 (1998). 329. C. S. Chen, M. Mrksich, S. Huang, G. M. Whitesides, and D. E. Ingber, Science 276, 1425 (1997). 330. R. Raiteri, M. Grattarola, H. J. Butt, and P. Skladal, Sens. Actuator B-Chem. 79, 115 (2001). 331. E. A. H. Hall, J. J. Gooding, and C. E. Hall, Mikrochim. Acta 121, 119 (1995). 332. C. K. O’Sullivan and G. G. Guilbault, Biosens. Bioelectron. 14, 663 (1999). 333. “Optical Biosensors: Present and Future.” Elsevier, Amsterdam, 2002. 334. B. Danielson and K. Mosbach, in “Biosensors: Fundamentals and Applications,” p. 575. Oxford University Press, Oxford, 1987. 335. D. R. Baselt, G. U. Lee, M. Natesan, S. W. Metzger, P. E. Sheehan, and R. J. Colton, Biosens. Bioelectron. 13, 731 (1998). 336. O. Lioubashevski, F. Patolsky, and I. Willner, Langmuir 17, 5134 (2001). 337. M. M. Miller, P. E. Sheehan, R. L. Edelstein, C. R. Tamanaha, L. Zhong, S. Bounnak, L. J. Whitman, and R. J. Colton, J. Magn. Magn. Mater. 225, 138 (2001). 338. R. L. Edelstein, C. R. Tamanaha, P. E. Sheehan, M. M. Miller, D. R. Baselt, L. J. Whitman, and R. J. Colton, Biosens. Bioelectron. 14, 805 (2000). 339. R. Berger, E. Delamarche, H. P. Lang, C. Gerber, J. K. Gimzewski, E. Meyer, and H.-J. Güntherodt, Science 276, 2021 (1997). 340. A. M. Moulin, S. J. O’Shea, and M. E. Welland, Ultramicroscopy 82, 23 (2000). 341. A. Abdelghani, J. M. Chovelon, N. JaffrezicRenault, C. Ronot Trioli, C. Veillas, and H. Gagnaire, Sens. Actuator B-Chem. 39, 407 (1997). 342. V. M. Mirsky, M. Vasjari, I. Novotny, V. Rehacek, V. Tvarozek, and O. S. Wolfbeis, Nanotechnology 13, 175 (2002). 343. A. J. Guiomar, J. T. Guthrie, and S. D. Evans, Langmuir 15, 1198 (1999). 344. S. Takenaka, K. Yamashita, M. Takagi, Y. Uto, and H. Kondo, Anal. Chem. 72, 1334 (2000). 345. R. M. Umek, S. W. Lin, J. Vielmetter, R. H. Terbrueggen, B. Irvine, C. J. Yu, J. F. Kayyem, H. Yowanto, G. F. Blackburn, D. H. Farkas, and Y. P. Chen, J. Mol. Diagn. 3, 74 (2001). 346. M. J. Giz, B. Duong, and N. J. Tao, J. Electroanal. Chem. 465, 72 (1999). 347. I. Bontidean, J. R. Lloyd, J. L. Hobman, J. R. Wilson, E. Csoregi, B. Mattiasson, and N. L. Brown, J. Inorg. Biochem. 79, 225 (2000). 348. A. Riklin and I. Willner, Anal. Chem. 67, 4118 (1995). 349. C. Bourdillon, C. Demaille, J. Moiroux, and J.-M. Savéant, Acc. Chem. Res. 29, 529 (1996). 350. P. G. He, J. N. Ye, Y. Z. Fang, J. Anzai, and T. Osa, Talanta 44, 885 (1997). 351. S. Kim, S.-E. Yun, and C. Kang, Electrochem. Commun. 1, 151 (1999). 352. C. Danilowicz, E. Corton, F. Battaglini, and E. J. Calvo, Electrochim. Acta 43, 3525 (1998). 353. E. S. Forzani, M. Otero, M. A. Perez, M. L. Teijelo, and E. J. Calvo, Langmuir 18, 4020 (2002). 354. E. J. Calvo, R. Etchenique, L. Pietrasanta, A. Wolosiuk, and C. Danilowicz, Anal. Chem. 73, 1161 (2001). 355. E. J. Calvo, F. Battaglini, C. Danilowicz, A. Wolosiuk, and M. Otero, Faraday Discuss. 47 (2000).

46 356. E. S. Forzani, V. M. Solis, and E. J. Calvo, Anal. Chem. 72, 5300 (2000). 357. E. J. Calvo and R. Etchenique, J. Phys. Chem. B 103, 8944 (1999). 358. J. J. Gooding, L. Pugliano, D. B. Hibbert, and P. Erokhin, Electrochem. Commun. 2, 217 (2000). 359. R. L. Earp and R. E. Dessy, in “Commercial Biosensors,” Vol. 148, p. 99. John Wiley and Sons, New York, 1998. 360. T. P. Henning and D. D. Cunningham, in “Commercial Biosensors,” Vol. 148, p. 304. John Wiley and Sons, New York, 1998. 361. J. J. Gooding, E. A. H. Hall, and D. B. Hibbert, Electroanalysis 10, 1130 (1998). 362. I. Willner, A. Riklin, B. Shoham, D. Rivenzon, and E. Katz, Adv. Mater. 5, 912 (1993). 363. C. Bourdillon, C. Demaille, J. Gueris, J. Moiroux, and J.-M. Savéant, J. Am. Chem. Soc. 115, 12264 (1993). 364. C. Bourdillon, C. Demaille, J. Moiroux, and J.-M. Savéant, J. Am. Chem. Soc. 116, 10328 (1994). 365. C. Bourdillon, C. Demaille, J. Moiroux, and J.-M. Saveant, J. Am. Chem. Soc. 117, 11499 (1995). 366. O. Pierrat, N. Lechat, C. Bourdillon, and J.-M. Laval, Langmuir 13, 4112 (1997). 367. N. Anicet, C. Bourdillon, J. Moiroux, and J.-M. Saveant, J. Phys. Chem. 102, 9844 (1998). 368. N. Anicet, C. Bourdillon, J. Moiroux, and J. M. Saveant, Langmuir 15, 6527 (1999). 369. L. Jiang, C. J. McNeil, and J. M. Cooper, J. Chem. Soc. Chem. Commun. 1293 (1995). 370. A. Riklin, E. Katz, I. Willner, A. Stocker, and A. F. Bückmann, Nature 376, 672 (1995). 371. T. Lötzbeyer, W. Schuhmann, and H.-L. Schmidt, Sensors Actuators B 33, 50 (1996). 372. R. Blonder, E. Katz, I. Willner, V. Wray, and A. F. Bückmann, J. Am. Chem. Soc. 119, 11747 (1997). 373. T. Lötzbeyer, W. Schuhmann, and H.-L. Schmidt, Bioelectrochem. Bioenerg. 42, 1 (1997). 374. G. Wittstock and W. Schuhmann, Anal. Chem. 69, 5059 (1997). 375. J. J. Gooding, M. Situmorang, P. Erokhin, and D. B. Hibbert, Anal. Comm. 36, 225 (1999). 376. J. J. Gooding, P. Erokhin, and D. B. Hibbert, Biosensors Bioelectronics 15, 229 (2000). 377. J. J. Gooding, P. Erokhin, D. Losic, W. R. Yang, V. Policarpio, J. Q. Liu, F. M. Ho, M. Situmorang, D. B. Hibbert, and J. G. Shapter, Anal. Sci. 17, 3 (2001). 378. D. Losic, J. J. Gooding, J. G. Shapter, D. B. Hibbert, and K. Short, Electroanalysis 13, 1385 (2001). 379. K. Kerman, D. Ozkan, P. Kara, B. Meric, J. J. Gooding, and M. Ozsoz, Anal. Chim. Acta 462, 39 (2002). 380. E. Katz, A. Riklin, and I. Willner, J. Electroanal. Chem. 354, 129 (1993). 381. I. Willner, E. Katz, B. Willner, R. Blonder, V. Heleg-Shabtai, and A. F. Bückmann, Biosens. Bioelectron. 12, 337 (1997). 382. Q. Chen, Y. Kobayashi, H. Takeshita, T. Hoshi, and J. Anzai, Electroanalysis 10, 94 (1998). 383. J. Anzai, H. Takeshita, Y. Kobayashi, T. Osa, and T. Hoshi, Anal. Chem. 70, 811 (1998). 384. C. Mousty, J. L. Bergamasco, R. Wessel, H. Perrot, and S. Cosnier, Anal. Chem. 73, 2890 (2001). 385. Y. Kajiya, T. Okamoto, and H. Yoneyama, Chemistry Letters 2107 (1993). 386. S. E. Creager and K. Olsen, Anal. Chim. Acta 307, 277 (1995). 387. M. Imamura, T. Haruyama, E. Kobatake, Y. Ikariyama, and M. Aizawa, Sensors Actuators B 24–25, 113 (1995). 388. I. Willner, M. Lion-Dagan, S. Marx-Tibbon, and E. Katz, J. Am. Chem. Soc. 117, 6581 (1995). 389. T. Nakaminami, S. Ito, S. Kuwabata, and H. Yoneyama, Anal. Chem. 71, 1068 (1999).

Alkanethiol Self-Assembled Monolayers 390. I. D. Karalemas and D. S. Papastathopoulos, Anal. Lett. 31, 913 (1998). 391. C. Ruan, F. Yang, C. Lei, and J. Deng, Anal. Chem. 70, 1721 (1998). 392. I. Willner, V. HelegShabtai, R. Blonder, E. Katz, and G. L. Tao, J. Am. Chem. Soc. 118, 10321 (1996). 393. H. Zimmermann, A. Lindgren, W. Schuhmann, and L. Gorton, Chem.-Eur. J. 6, 592 (2000). 394. H. D. Sikes, J. F. Smalley, S. P. Dudek, A. R. Cook, M. D. Newton, C. E. D. Chidsey, and S. W. Feldberg, Science 291, 1519 (2001). 395. S. O. Kelley, N. M. Jackson, M. G. Hill, and J. K. Barton, Angew. Chem. Int. Ed. 38, 941 (1999). 396. R. E. Holmlin, P. J. Dandliker, and J. K. Barton, Angew. Chem. Int. Ed. Engl. 36, 2714 (1997). 397. F. Lisdat, B. Ge, and F. W. Scheller, Electrochem. Commun. 1, 65 (1999). 398. T. Ruzgas, E. Csoregi, J. Emneus, L. Gorton, and G. Marko-Varga, Anal. Chim. Acta 330, 123 (1996). 399. A. Heller, Acc. Chem. Res. 23, 128 (1990). 400. H. J. Hecht, D. Schomburg, H. Kalisz, and R. D. Schmid, Biosens. Bioelectron. 8, 197 (1993). 401. J. Liu, M. N. Paddon-Row, and J. J. Gooding, unpublished results. 402. A. J. Black, T. T. Wooster, W. E. Geiger, and M. N. Paddon-Row, J. Am. Chem. Soc. 115, 7924 (1993). 403. M. N. Paddon-Row, Acc. Chem. Res. 27, 18 (1994). 404. B. H. Huisman, R. P. H. Kooyman, F. vanVeggel, and D. N. Reinhoudt, Adv. Mater. 8, 561 (1996). 405. D. M. Disley, D. C. Cullen, H. X. You, and C. R. Lowe, Biosens. Bioelectron. 13, 1213 (1998). 406. D. M. Disley, P. R. Morrill, K. Sproule, and C. R. Lowe, Biosens. Bioelectron. 14, 481 (1999). 407. C. Duschl, A. F. Sevin Landais, and H. Vogel, Biophys. J. 70, 1985 (1996). 408. R. P. H. Kooyman, D. J. Vandenheuvel, J. W. Drijfhout, and G. W. Welling, Thin Solid Films 244, 913 (1994). 409. I. Vikholm, E. Gyorvary, and J. Peltonen, Langmuir 12, 3276 (1996). 410. K. Gaus, A. Basran, and E. A. H. Hall, Analyst 126, 329 (2001). 411. K. Gaus and E. A. H. Hall, Biosens. Bioelectron. 18, 151 (2003). 412. K. A. Peterlinz, R. M. Georgiadis, T. M. Herne, and M. J. Tarlov, J. Am. Chem. Soc. 119, 3401 (1997). 413. C. E. Jordan, A. G. Frutos, A. J. Thiel, and R. M. Corn, Anal. Chem. 69, 4939 (1997). 414. A. J. Haes and R. P. Van Duyne, J. Am. Chem. Soc. 124, 10596 (2002). 415. M. Boncheva, L. Scheibler, P. Lincoln, H. Vogel, and B. Åkerman, Langmuir 15, 4317 (1999). 416. J. M. Brockman, A. G. Frutos, and R. M. Corn, J. Am. Chem. Soc. 121, 8044 (1999). 417. P. D. Gershon and S. Khilko, J. Immunological Methods 183, 65 (1995). 418. K. Gaus and E. A. H. Hall, J. Coll. Interface Sci. 194, 364 (1997). 419. S. Sawata, E. Kai, K. Ikebukuro, T. Iida, T. Honda, and I. Karube, Biosens. Bioelectron. 14, 397 (1999). 420. A. J. Thiel, A. G. Frutos, C. E. Jordan, R. M. Corn, and L. M. Smith, Anal. Chem. 69, 4948 (1997). 421. H. J. Watts, D. Yeung, and H. Parkes, Anal. Chem. 67, 4283 (1995). 422. I. Vikholm, T. Viitala, W. M. Albers, and J. Peltonen, Biochim. Biophys. Acta-Biomembr. 1421, 39 (1999). 423. C. A. Rowe-Taitt, J. J. Cras, C. H. Patterson, J. P. Golden, and F. S. Ligler, Anal. Biochem. 281, 123 (2000). 424. W. M. Mullett, E. P. C. Lai, and J. M. Yeung, Methods 22, 77 (2000). 425. T. Liu, J. Tang, H. Q. Zhao, Y. P. Deng, and L. Jiang, Langmuir 18, 5624 (2002).

Alkanethiol Self-Assembled Monolayers 426. J. Wang, P. E. Nielsen, M. Jiang, X. Cai, J. R. Fernandes, D. H. Grant, M. Ozsoz, A. Beglieter, and M. Mowat, Anal. Chem. 69, 5200 (1997). 427. H. Zhang, R. Wang, T. H., L. Nie, and S. Yao, Talanta 46, 171 (1998). 428. H. Zhang, H. Tan, R. Wang, W. Wei, and S. Yao, Anal. Chim. Acta 374, 31 (1998). 429. B. A. Cavic, G. L. Hayward, and M. Thompson, Analyst 124, 1405 (1999). 430. F. Caruso, E. Rodda, D. N. Furlong, K. Niikura, and Y. Okahata, Anal. Chem. 69, 2043 (1997). 431. Y. Okahata, Y. Matsunobu, K. Ijiro, M. Mukae, A. Murakami, and K. Makino, J. Am. Chem. Soc. 114, 8299 (1992). 432. I. Ben Dov, I. Willner, and E. Zisman, Anal. Chem. 69, 3506 (1997). 433. Y. Cohen, S. Levi, S. Rubin, and I. Willner, J. Electroanal. Chem. 417, 65 (1996). 434. M. Minunni, M. Mascini, G. G. Guilbault, and B. Hock, Anal. Lett. 28, 749 (1995). 435. L. M. Furtado and M. Thompson, Analyst 123, 1937 (1998). 436. T. Hianik, V. Gajdos, R. Krivanek, T. Oretskaya, V. Metelev, E. Volkov, and P. Vadgama, Bioelectrochemistry 53, 199 (2001). 437. E. Huang, M. Satjapipat, S. B. Han, and F. M. Zhou, Langmuir 17, 1215 (2001). 438. Y. Okahata, M. Kawase, K. Niikura, F. Ohtake, H. Furusawa, and Y. Ebara, Anal. Chem. 70, 1288 (1998). 439. S. Storri, T. Santoni, and M. Mascini, Anal. Lett. 31, 1795 (1998). 440. R. D. Vaughan, C. K. O’Sullivan, and G. G. Guilbault, Fresenius J. Anal. Chem. 364, 54 (1999). 441. X. D. Su, F. T. Chew, and S. F. Y. Li, Anal. Biochem. 273, 66 (1999). 442. R. B. Towery, N. C. Fawcett, P. Zhang, and J. A. Evans, Biosens. Bioelectron. 16, 1 (2001). 443. J. Horacek and P. Skladal, Anal. Chim. Acta 347, 43 (1997). 444. S. Susmel, C. K. O’Sullivan, and G. G. Guilbault, Enzyme Microb. Technol. 27, 639 (2000). 445. X. C. Zhou and L. Cao, Analyst 126, 71 (2001). 446. Z. Y. Wu, C. P. He, S. P. Wang, G. L. Shen, and R. Q. Yu, Chem. J. Chin. Univ.-Chin. 22, 542 (2001). 447. Y. S. Fung and Y. Y. Wong, Anal. Chem. 73, 5302 (2001). 448. S. F. Chou, W. L. Hsu, J. M. Hwang, and C. Y. Chen, Anal. Chim. Acta 453, 181 (2002). 449. A. J. C. Eun, L. Q. Huang, F. T. Chew, S. F. Y. Li, and S. M. Wong, J. Virol. Methods 99, 71 (2002). 450. R. Saber, S. Mutlu, and E. Piskin, Biosens. Bioelectron. 17, 727 (2002). 451. C. C. Wang, H. Wang, Z. Y. Wu, G. M. Zeng, G. H. Huang, G. L. Shen, and R. Q. Yu, Acta Chim. Sin. 60, 1284 (2002). 452. E. Katz and I. Willner, J. Electroanal. Chem. 418, 67 (1996). 453. C. Berggren and G. Johansson, Anal. Chem. 69, 3651 (1997). 454. R. Blonder, I. Ben Dov, A. Dagan, I. Willner, and E. Zisman, Biosens. Bioelectron. 12, 627 (1997). 455. C. Berggren, B. Bjarnason, and G. Johansson, Biosens. Bioelectron. 13, 1061 (1998). 456. A. Friggeri, F. van Veggel, and D. N. Reinhoudt, Chem.-Eur. J. 5, 3595 (1999). 457. H. C. Yoon, M.-Y. Hong, and H.-S. Kim, Anal. Biochem. 282, 121 (2000). 458. J. Yan, L. M. Tender, P. D. Hampton, and G. P. López, J. Phys. Chem. B. 105, 8905 (2001). 459. M. Riepl, V. M. Mirsky, I. Novotny, V. Tvarozek, V. Rehacek, and O. S. Wolfbeis, Anal. Chim. Acta 392, 77 (1999). 460. H. Aoki, P. Buhlmann, and Y. Umezawa, J. Electroanal. Chem. 473, 105 (1999). 461. V. P. Y. Gadzekpo, K. P. Xiao, H. Aoki, P. Buhlmann, and Y. Umezawa, Anal. Chem. 71, 5109 (1999). 462. M. Liu, Q. X. Li, and G. A. Rechnitz, Electroanalysis 12, 21 (2000).

47 463. M. Dijksma, B. Kamp, J. C. Hoogvliet, and W. P. van Bennekom, Anal. Chem. 73, 901 (2001). 464. C. Berggren, B. Bjarnason, and G. Johansson, Electroanalysis 13, 173 (2001). 465. R. J. Pei, Z. L. Cheng, E. K. Wang, and X. R. Yang, Biosens. Bioelectron. 16, 355 (2001). 466. L. Alfonta, A. Bardea, O. Khersonsky, E. Katz, and I. Willner, Biosens. Bioelectron. 16, 675 (2001). 467. Z. Liron, L. M. Tender, J. P. Golden, and F. S. Ligler, Biosens. Bioelectron. 17, 489 (2002). 468. C. M. Ruan, L. J. Yang, and Y. B. Li, Anal. Chem. 74, 4814 (2002). 469. H. F. Ji, E. Finot, R. Dabestani, T. Thundat, G. M. Brown, and P. F. Britt, Chem. Commun. 457 (2000). 470. K. M. Hansen, H. F. Ji, G. H. Wu, R. Datar, R. Cote, A. Majumdar, and T. Thundat, Anal. Chem. 73, 1567 (2001). 471. J. Homola, S. S. Yee, and D. Myszka, in “Optical Biosensors: Present and Future,” p. 207. Elsevier, Amsterdam, 2002. 472. S. Busse, V. Scheumann, B. Menges, and S. Mittler, Biosens. Bioelectron. 17, 704 (2002). 473. A. B. Steel, T. M. Herne, and M. J. Tarlov, Anal. Chem. 79, 4670 (1998). 474. A. B. Steel, T. M. Herne, and M. J. Tarlov, Bioconjugate Chem. 10, 419 (1999). 475. A. B. Steel, R. L. Levicky, T. M. Herne, and M. J. Tarlov, Biophys. J. 79, 975 (2000). 476. C. Xu, L. Sun, L. J. Kepley, and R. M. Crooks, Anal. Chem. 65, 2102 (1993). 477. D. V. Leff, L. Brandt, and J. R. Heath, Langmuir 12, 4723 (1996). 478. E. Huang, F. M. Zhou, and L. Deng, Langmuir 16, 3272 (2000). 479. D. Ozkan, E. A., K. Kerman, P. Kara, J. J. Gooding, P. E. Nielsen, and M. Ozsoz, Electrochem. Commun. 4, 796 (2002). 480. B. L. Frey and R. M. Corn, Anal. Chem. 68, 3187 (1996). 481. W. Yang, J. J. Gooding, and D. B. Hibbert, Analyst 126, 1573 (2001). 482. W. Yang, D. Jaramillo, J. J. Gooding, D. B. Hibbert, R. Zhang, G. D. Willett, and K. J. Fisher, Chem. Commun. 1982 (2001). 483. L. Häussling, H. Ringsdorf, F. J. Schmitt, and W. Knoll, Langmuir 7, 1837 (1991). 484. F. J. Schmitt, L. Häussling, H. Ringsdorf, and W. Knoll, Thin Solid Films 201/211, 815 (1992). 485. J. Spinke, M. Liley, H. J. Guder, L. Angermaier, and W. Knoll, Langmuir 9, 1821 (1993). 486. H. Matsui, P. Porrata, and G. E. Douberly, Nano Lett. 1, 461 (2001). 487. P. D. J. Osman, V. L. B. Braach-Maksvytis, B. A. Cornell, L. G. King, R. J. Pace, B. Raguse, and L. Wieczorek, Biophys. J. 72, TH291 (1997). 488. B. A. Cornell, G. Krishna, P. D. Osman, R. D. Pace, and L. Wieczorek, Biochem. Soc. Trans. 29, 613 (2001). 489. G. Woodhouse, L. King, L. Wieczorek, P. Osman, and B. Cornell, J. Mol. Recognit. 12, 328 (1999). 490. G. E. Woodhouse, L. G. King, L. Wieczorek, and B. A. Cornell, Faraday Discuss. 247 (1998). 491. B. Raguse, V. L. B. Braach-Maksvytis, B. A. Cornell, L. G. King, P. D. J. Osman, R. J. Pace, and L. Wieczorek, Langmuir 14, 648 (1998). 492. V. L. B. Braach-Maksvytis, B. A. Cornell, L. G. King, P. D. J. Osman, R. J. Pace, B. Raguse, and L. Wieczorek, Biophys. J. 72, TH288 (1997). 493. B. A. Cornell, V. L. B. Braach-Maksvytis, L. G. King, P. D. J. Osman, R. J. Pace, B. Raguse, and L. Wieczorek, Biophys. J. 72, TH289 (1997). 494. S. W. Lucas and M. M. Harding, Anal. Biochem. 282, 70 (2000). 495. M. W. J. Beulen, F. C. J. M. van Veggel, and D. N. Reinhoudt, Chem. Commun. 503 (1999).

48 496. S. Flink, H. Schonherr, G. J. Vancso, F. A. J. Geurts, K. G. C. van Leerdam, F. van Veggel, and D. N. Reinhoudt, J. Chem. Soc.Perkin Trans. 2, 2141 (2000). 497. S. Flink, B. A. Boukamp, A. van den Berg, F. C. J. M. van Veggel, and D. N. Reinhoudt, J. Am. Chem. Soc. 120, 4652 (1998). 498. A. J. Moore, L. M. Goldenberg, M. R. Bryce, M. C. Petty, A. P. Monkman, C. Marenco, J. Yarwood, M. J. Joyce, and S. N. Port, Adv. Mater. 10, 395 (1998). 499. H. Y. Liu, S. G. Liu, and L. Echegoyen, Chem. Commun. 1493 (1999). 500. K. Bandyopadhyay, L. H. Shu, H. Y. Liu, and L. Echegoyen, Langmuir 16, 2706 (2000). 501. K. Bandyopadhyay, H. Y. Liu, S. G. Liu, and L. Echegoyen, Chem. Commun. 141 (2000). 502. K. Bandyopadhyay, S. G. Liu, H. Y. Liu, and L. Echegoyen, Chem.-Eur. J. 6, 4385 (2000). 503. T. D. Chung, J. Park, J. Kim, H. Lim, M. J. Choi, J. R. Kim, S. K. Chang, and H. Kim, Anal. Chem. 73, 3975 (2001). 504. A. Friggeri, F. van Veggel, D. N. Reinhoudt, and R. P. H. Kooyman, Langmuir 14, 5457 (1998). 505. W. Yang, J. J. Gooding, and D. B. Hibbert, J. Electroanal. Chem. 516, 10 (2001). 506. J. J. Gooding, D. B. Hibbert, and W. Yang, Sensors 1, 75 (2001). 507. B. Sellergren, Trac-Trends Anal. Chem. 16, 310 (1997). 508. S. A. Piletsky and A. P. F. Turner, Electroanalysis 14, 317 (2002). 509. H. Asanuma, T. Hishiya, and M. Komiyama, Adv. Mater. 12, 1019 (2000). 510. S. A. Piletsky, T. L. Panasyuk, E. V. Piletskaya, I. A. Nicholls, and M. Ulbricht, J. Membr. Sci. 157, 263 (1999). 511. Z. P. Yang, J. M. Kauffmann, M. I. A. Valenzuela, and S. Ozkan, Mikrochim. Acta 131, 85 (1999). 512. S. A. Piletsky, E. V. Piletskaya, T. A. Sergeyeva, T. L. Panasyuk, and A. V. El’skaya, Sens. Actuator B-Chem. 60, 216 (1999). 513. V. M. Mirsky, T. Hirsch, S. A. Piletsky, and O. S. Wolfbeis, Angew. Chem.-Int. Edit. 38, 1108 (1999). 514. M. Lahav, E. Katz, A. Doron, F. Patolsky, and I. Willner, J. Am. Chem. Soc. 121, 862 (1999). 515. Z. P. Yang, I. Engquist, B. Liedberg, and J. M. Kauffmann, J. Electroanal. Chem. 430, 189 (1997). 516. H. L. Zhang, M. Chen, and H. L. Li, J. Phys. Chem. B 104, 28 (2000). 517. D. Cahen and G. Hodes, Ad. Mater. 14, 789 (2002). 518. X. D. Cui, A. Primak, X. Zarate, J. Tom Fohr, O. F. Sankey, A. L. Moore, T. A. Moore, D. Gust, G. Harris, and L. S. M., Science 294, 571 (2001). 519. Y. Selzer, A. Salomon, and D. Cahen, J. Am. Chem. Soc. 124, 2886 (2002). 520. R. E. Holmlin, R. F. Ismagilov, R. Haag, V. Mujica, M. A. Ratner, M. A. Rampi, and G. M. Whitesides, Angew. Chem. Int. Ed. 40, 2316 (2001). 521. D. J. Wold, R. Haag, M. A. Rampi, and C. D. Frisbie, J. Phys. Chem. B 106, 2813 (2002). 522. S. E. Creager and G. K. Rowe, J. Electroanal. Chem. 420, 291 (1997). 523. J. J. Sumner, K. S. Weber, L. A. Hockett, and S. E. Creager, J. Phys. Chem. B 104, 7449 (2000). 524. E. Laviron, J. Electroanal. Chem. 101, 19 (1979). 525. E. Laviron, J. Electroanal. Chem. 97, 135 (1979). 526. S. E. Creager and T. T. Wooster, Anal. Chem. 70, 4257 (1998). 527. J. H. Reeves, S. Song, and E. F. Bowden, Anal. Chem. 65, 683 (1992). 528. C. E. D. Chidsey, Science 251, 919 (1991). 529. R. A. Marcus and N. Sutin, Biochim. Biophys. Acta 811, 265 (1985). 530. D. J. Wold and C. D. Frisbie, J. Am. Chem. Soc. 123, 5549 (2001). 531. M. A. Rampi and G. M. Whitesides, Chem. Phys. 281, 373 (2002). 532. Z. H. Yu, H. B. Shao, Y. Luo, H. L. Zhang, and Z. F. Liu, Langmuir 13, 5774 (1997).

Alkanethiol Self-Assembled Monolayers 533. M. J. Hostetler, J. J. Stokes, and R. W. Murray, Langmuir 12, 3604 (1996). 534. C. S. Weisbecker, M. V. Merritt, and G. M. Whitesides, Langmuir 12, 3763 (1996). 535. M. J. Hostetler, S. J. Green, J. J. Stokes, and R. W. Murray, J. Am. Chem. Soc. 118, 4212 (1996). 536. K. C. Grabar, K. J. Allison, B. E. Baker, R. M. Bright, K. R. Brown, R. G. Freeman, A. P. Fox, C. D. Keating, M. D. Musick, and M. J. Natan, Langmuir 12, 2353 (1996). 537. S. W. Chen, R. S. Ingram, M. J. Hostetler, J. J. Pietron, R. W. Murray, T. G. Schaaff, J. T. Khoury, M. M. Alvarez, and R. L. Whetten, Science 280, 2098 (1998). 538. A. P. Alivisatos, K. P. Johnsson, X. G. Peng, T. E. Wilson, C. J. Loweth, M. P. Bruchez, and P. G. Schultz, Nature 382, 609 (1996). 539. C. J. Loweth, W. B. Caldwell, X. Peng, A. P. Alivisatos, and P. G. Schultz, Angew. Chem. Int. Ed. 38, 1808 (1999). 540. S. Mann, W. Shenton, M. Li, S. Connolly, and D. Fitzmaurice, Adv. Mater. 12, 147 (2000). 541. F. P. Zamborini, J. F. Hicks, and R. W. Murray, J. Am. Chem. Soc. 122, 4514 (2000). 542. J. P. Sumner, J. W. Aylott, E. Monson, and R. Kopelman, Analyst 127, 11 (2002). 543. H. A. Clark, M. Hoyer, M. A. Philbert, and R. Kopelman, Anal. Chem. 71, 4831 (1999). 544. H. A. Clark, R. Kopelman, R. Tjalkens, and M. A. Philbert, Anal. Chem. 71, 4837 (1999). 545. H. A. Clark, S. L. R. Barker, M. Brasuel, M. T. Miller, E. Monson, S. Parus, Z. Y. Shi, A. Song, B. Thorsrud, R. Kopelman, A. Ade, W. Meixner, B. Athey, M. Hoyer, D. Hill, R. Lightle, and M. A. Philbert, Sens. Actuator B-Chem. 51, 12 (1998). 546. D. J. Maxwell, J. R. Taylor, and S. M. Nie, J. Am. Chem. Soc. 124, 9606 (2002). 547. J. J. Storhoff, R. Elghanian, R. C. Mucic, C. A. Mirkin, and R. L. Letsinger, J. Am. Chem. Soc. 120, 1959 (1998). 548. H. W. Kroto, J. R. Heath, S. C. O’Brien, R. F. Curl, and R. E. Smalley, Nature 318, 162 (1985). 549. S. Iijima, Nature 354, 56 (1991). 550. S. Subramoney, Ad. Mater. 10, 1157 (1998). 551. V. P. Dravid, X. Lin, Y. Wang, X. K. Wang, A. Yee, J. B. Ketterson, and R. H. P. Chang, Science 259, 1601 (1993). 552. J. Berholc, D. Brenner, M. Buongiorno Nardelli, V. Meunier, and C. Roland, Annu. Rev. Mater. Res. 32, 347 (2002). 553. J. Liu, M. J. Casavant, M. Cox, D. A. Walters, P. Boul, W. Lu, A. J. Rimberg, K. A. Smith, D. T. Colbert, and R. E. Smalley, Chem. Phys. Lett. 303, 125 (1999). 554. Y. Huang, X. F. Duan, Y. Cui, L. J. Lauhon, K. H. Kim, and C. M. Lieber, Science 294, 1313 (2001). 555. Y. Huang, X. F. Duan, Q. Q. Wei, and C. M. Lieber, Science 291, 630 (2001). 556. P. Diao, Z. F. Liu, B. Wu, X. Nan, J. Zhang, and Z. Wei, Chem. Phys. Chem. 3, 898 (2002). 557. J. J. Gooding, R. Wibowo, J. Liu, W. Yang, D. Losic, S. Orbons, F. J. Mearns, J. G. Shapter and D. B. Hibbert, J. Am. Chem. Soc. 125, 9006 (2003). 558. J. M. Fréchet, Science 263, 1710 (1994). 559. D. A. Tomalia, A. M. Naylor, and W. A. Goddard III, Angew. Chem. Int. Ed. 29, 138 (1990). 560. J. Issberner, R. Moors, and F. Voegtle, Angew. Chem. Int. Ed. 33, 2413 (1994). 561. D. A. Tomalia and H. D. Durst, Top. Curr. Chem. 165, 193 (1993). 562. H. Tokuhisa, M. Q. Zhao, L. A. Baker, V. T. Phan, D. L. Dermody, M. E. Garcia, R. F. Peez, R. M. Crooks, and T. M. Mayer, J. Am. Chem. Soc. 120, 4492 (1998).

Alkanethiol Self-Assembled Monolayers 563. J. W. J. Knapen, A. W. van der Made, J. C. D. Wilde, W. van Leeuwen, P. Wijkens, D. M. Groves, and G. van Koten, Nature 372, 659 (1994). 564. J. F. G. A. Jansen, E. M. M. de Brabander-van den Berg, and E. M. Meijer, Science 266, 1226 (1994). 565. M. Wells and R. M. Crooks, J. Am. Chem. Soc. 118, 3988 (1996). 566. H. Tokuhisa and R. M. Crooks, Langmuir 13, 5608 (1997). 567. J. Kukoska-Latallo, A. N. Bielinska, J. Johnson, R. Spindler, D. A. Tomalia, and J. R. Baker, Jr., Proc. Nat. Acad. Sci. U.S.A. 93, 4897 (1996). 568. M. Q. Zhao, H. Tokuhisa, and R. M. Crooks, Angew. Chem.-Int. Edit. 36, 2596 (1997). 569. V. Chechik and R. M. Crooks, Langmuir 15, 6364 (1999).

49 570. A. Friggeri, H. J. van Manen, T. Auletta, X.-M. Li, S. Zapotoczny, H. Schönher, G. J. Vancso, J. Huskens, F. C. J. M. van Veggel, and D. N. Reinhoudt, J. Am. Chem. Soc. 123, 6388 (2001). 571. C. B. Gorman, R. L. Miller, K. Y. Chen, A. R. Bishop, R. T. Haasch, and R. G. Nuzzo, Langmuir 14, 3312 (1998). 572. L. Zhang, F. W. Huo, Z. Q. Wang, L. X. Wu, X. Zhang, S. Hoppener, L. F. Chi, H. Fuchs, J. W. Zhao, L. Niu, and S. J. Dong, Langmuir 16, 3813 (2000). 573. A. Friggeri, H. Schonherr, H. J. van Manen, B. H. Huisman, G. J. Vancso, J. Huskens, F. van Veggel, and D. N. Reinhoudt, Langmuir 16, 7757 (2000). 574. F. Effenberger, G. Gotz, B. Bidlingmaier, and M. Wezstein, Angew. Chem. Int. Ed. 37, 2462 (1998).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Analytical Characterization of Single Wall Carbon Nanotubes Sivaram Arepalli, Pasha Nikolaev, Olga Gorelik G. B. Tech./NASA-Johnson Space Center, Houston, Texas, USA

CONTENTS 1. Introduction 2. Analytical Techniques Glossary References

1. INTRODUCTION During the last few years, interest in the study of single wall carbon nanotubes (SWCNTs or SWNTs) has grown exponentially because of their extraordinary properties [1, 2]. Some of the possible applications include flat panel displays, superstrong fibers and panels, multifunctional materials, nanoelectronics and sensors, supercapacitors, and fuel cells [3–5]. The large aspect ratio (nanometer diameter and several micrometer length) adds tremendous appeal to the SWCNTs as reinforcing materials for composites. Also, the large surface areas available from individual tubes provide possible means for energy storage. The anisotropic nature allows enhanced properties along the length of the tube such as directed thermal energy transfer. The applications depend heavily on the diameter, length, chirality, and bundle size as well as on the amount and extent of impurities existing in the nanotube material. These impurities are inherent to the production process and change in character with the processing and purification steps used. The carbon nanotubes can be produced by a variety of techniques including carbon arc, laser ablation, chemical vapor deposition, and carbon monoxide disproportionation [6–9] using metal catalysts. In addition to SWCNTs, the produced product may contain multiwall tubes with inferior properties. The as-produced material is known to be inhomogeneous and contain a variety of impurities including graphite coated metal clusters, amorphous carbon, and other carbon compounds including fullerenes. The type and extent of these impurities change with the production method used. The production process also influences the length, diameter,

ISBN: 1-58883-057-8/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

chirality, and surface moiety of the nanotubes (and nanotube bundles). It is indeed a challenge to determine all these impurity levels and characterize this material to a reasonable extent using the available analytical techniques. Also, the purification processes used to improve the percentage of SWCNTs in the material tend to alter the nanotubes by adding functional groups, damaging and cutting the tubes [10–13]. For some applications, it is indeed necessary to add functional groups at specific locations and therefore requires additional sophisticated analytical tools for characterization. The SWCNT material seems to behave similar to carbon soot and may require additional precautions while handling because of its submicrometer size dimensions. It is therefore necessary to assess all possible hazards involved in handling and address the toxicological effects of the nanotubes [14]. This chapter addresses the analytical tools and techniques used to characterize the SWCNT material. Most of the aforementioned aspects are covered in general and the reader may like to follow the cited references for further detailed information. Some of these techniques which are qualitative in nature can be used for quantitative estimates by using calibrated standards and procedures. Each technique is described briefly and its application to SWCNTs is shown by providing one or two measurements as examples.

2. ANALYTICAL TECHNIQUES The analytical techniques are divided into subgroups based on general operating principles. The first group is based on electron microscopy and encompasses techniques such as scanning electron microscopy (SEM), transmission electron microscopy (TEM), scanning tunneling microscopy (STM), and atomic force microscopy (AFM). The second group is related to spectrophotometry (absorption and emission) in the ultraviolet-visible-infrared (UV-VIS-IR) regions and includes Raman methods. The third group is focused on mass spectrometry [gas chromatography (GC) and time-offlight (TOF) mass spectrometry (MS)] and chromatography [high pressure liquid chromatography (HPLC)] that use sizeselection processes to characterize the material. The fourth

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 1: Pages (51–66)

52

Analytical Characterization of SWCNTs

group is based on X-ray excitations such as X-ray diffraction (XRD), X-ray photoelectron spectroscoy (XPS), and extended X-ray absorption fine structure (EPXAFS). Magnetic resonance methods such as electron paramagnetic resonance (EPR) and nuclear magnetic resonance (NMR) constitute the fifth group. The sixth group covers the thermal methods such as thermogravimetric analysis (TGA), temperature programmed oxidation (TPO), and differential scanning calorimetry (DSC).

2.1. Electron Microscopy 2.1.1. TEM Transmission electron microscopy remains one of the best tools to characterize nanotubes and related nanostructures. Historically, both multiwall [15] and single-wall [16] carbon nanotubes were discovered on the TEM micrographs of the cathode deposit and soot formed in the arc-discharge fullerene generator. Figures 1 and 2 show high resolution TEM images of single wall carbon nanotubes produced by two different processes. A number of other nanostructured carbons, like bucky peapods [17] and carbon nanohorns [18], were also discovered and studied in TEM. TEM works on the same principle as optical microscope, except photons (typical wavelength 200–800 nm) are replaced with electrons (typical wavelength 0.0025 nm at 200 kV accelerating voltage). Numerical aperture of TEM is much smaller than that in the optical microscope (35′ compared to ∼70
. Electron beam optics is essentially free of chromatic aberrations, while spherical aberrations are severe. Theoretical resolution is determined by electron wavelength and numerical aperture and is very high, while in practice it does not exceed ∼0.3 nm and is limited by spherical and diffraction aberrations. Imaging in TEM is strongly dependent on the type of the electron source. There are three common types: tungsten filament, lanthanum hexaboride (LaB6
crystal, and fieldemission gun (FEG). Tungsten filament creates a source of relatively large size, low brightness, and low coherence and is hardly used anymore. A LaB6 source is the most common

10 nm

Figure 1. High-resolution TEM image of a SWCNT produced by the HiPco process.

10 nm ORIG MAG

250000

NO DATA

Figure 2. High-resolution TEM image of single wall carbon nanotubes produced from carbon monoxide feedstock using Fe–Mo catalyst supported on silica particles.

these days and provides much better coherence, brightness, and size while still being a relatively inexpensive option. There are few field-emission-type TEMs available, which are significantly more expensive than other options. Source size and beam coherence in FEG systems are excellent, which significantly improves resolution. But brightness of the FEG source exceeds that of LaB6 by 1–2 orders of magnitude, which creates much higher fluence of the electron beam illuminating the sample. Therefore a TEM equipped with a FEG source quickly produces significant e-beam damage of the nanotube samples. The authors have observed nanotubes literally falling apart in a matter of minutes. FEG TEM has to be used with the understanding that nanotubes can be damaged very quickly. Another important factor in TEM imaging is the accelerating voltage. The common range is 100–200 kV, but modern TEMs can achieve up to 800 kV–1 MV. In theory higher accelerating voltage allows one to achieve higher resolution, since electron wavelength is smaller. In practice, however, TEM imaging is always a trade-off between resolution and contrast. Carbon atoms are relatively light and do not scatter high-energy electrons well. Scattering cross-section decreases quickly at higher accelerating voltages, and image contrast becomes very small. Defocusing can be used to increase contrast, but then resolution suffers. It is believed that accelerating voltage in the range of 100–160 kV is optimal for imaging nanotube samples. The most common way to prepare TEM specimens is to disperse nanotubes in organic solvent (alcohol, acetone, etc.) and then place a droplet of suspension on the TEM grid lying on tissue paper. The solvent is adsorbed by the tissue, and the nanotubes lay flat on the grid and coat it more or less uniformly. The thickness of the carbon film on the commercially available grids is around 50–100 nm. This means that individual nanotubes and small bundles (1–10 nm in

53

Analytical Characterization of SWCNTs

thickness) cannot be imaged effectively sitting directly on the support film. It is necessary to use “holey” or “lacey” grids, with plenty of holes in the support film, in order to image nanotubes bridging the holes. There are more exotic ways to prepare TEM specimens. For example, TEM can be used to measure nanotube diameters. But nanotubes exist almost exclusively as bundles, which makes it nearly impossible to distinguish walls of each nanotube in the bundle lying flat on the support film in order to measure the diameter of each nanotube. Moreover, if a triangular lattice of a well-ordered nanorope is positioned a certain way with respect to the electron beam,√line spacing observed on the TEM image represent either 23 d + a
or 1 d + a
, where d is nanotube diameter and a is van der 2 Waals interwall spacing in the nanorope [19]. Figure 3 shows the same nanorope rotated 0 and 60 degrees around its axis, and interline spacing follows this expression. However, if the bundle loops through the focal plane of the TEM so that nanotubes are parallel to the electron beam, one can image the cross-section of the bundle with nanotubes represented as circles (Fig. 4). It is easy to determine the diameters of a large number of nanotubes on such an image. There is a simple technique to prepare a specimen with a large number of cross-sectional views of SWNT bundles, which works well on dry and fluffy as-made nanotubes. One can place a small chunk of dry nanotube material on a TEM grid and wet it with a droplet of alcohol. The nanotube chunk becomes compressed by capillary forces and sticks to the support film. In the process SWNT bundles at the chunk extremities fold onto themselves and create loops, most of which are in-plane or at some angle to the e-beam. A small fraction of the loops, however, is parallel to the e-beam, which creates the aforementioned cross-sectional images. To make measurements of the diameters of a large number of circles easier, Nikoalev has written a simple program [20] which allows one to superimpose a circle over a nanotube image, adjust its size and position to achieve the best fit, and then record its diameter. About 200–300 circles can be accurately measured in an hour (Fig. 5).

1 nm ORIC MAC 400000

NO DATA

Figure 4. Cross-sectional view of SWCNT nanorope. Reprinted with permission from [63], A. G. Rinzler et al., Appl. Phys. A 67, 29 (1998). © 1998, Springer-Verlag.

Amorphous carbon is a major impurity in the SWNT. It is not well understood, and there is no single analytical technique which can reliably discriminate between nanotubes and other forms of carbon. TEM observations are often used to estimate the purity of SWNT materials, although such estimates are very subjective. TEM is also frequently used to determine dimensions of various nanostructures.

2.1.2. SEM Scanning electron microscopy is another common tool used to study nanotubes and related nanostructures (Fig. 6). Modern SEMs utilize the same electron sources as TEMs, tungsten, LaB6 , and field emission guns, FEG providing the 4" oven material 15

10

5

5 nm

0 0.8

1.0

1.2

1.4

1.6

1.8

Diameter, nm Figure 3. TEM images of crystalline SWCNT nanorope rotated around its axis by 0 and 60 . Interline spacing changes according to the particular view of the triangular lattice of the nanorope. Reprinted with permission from [7] A. Thess et al., Science 273, 483 (1996). © 1996, American Association for Advancement of Science.

Figure 5. Diameter distribution of SWCNT shown in Figure 4, measured from TEM micrographs. Reprinted with permission from [63], A. G. Rinzler et al., Appl. Phys. A 67, 29 (1998). © 1998, SpringerVerlag.

54

Analytical Characterization of SWCNTs

C u

C u C OC S l 0 i

C NC 0 i0

C u

0 – 20 keV Preset: 250 s 73 s 72% Dead 389 s

X-RAY: Live : Real:

Remaining: 177 s 10.80 kcps IN C

u

C u C C

5 KV

0C 0

1m 10 mm

Figure 6. SEM image of unpurified dry SWCNT.

best brightness and coherence. SEMs detect either backscattered or secondary electrons. There are no reports on using backscattered electron detectors in nanotube imaging, so all discussion will concentrate on secondary electrons (although backscattered electrons can be used to image nanocomposites, i.e., nanotubes embedded in polymers). Resolution of SEM depends on the beam diameter, which is usually 3–5 nm, but is not determined by it. This happens because electrons penetrate into the surface of the specimen, and observed secondary electrons are emitted from the pear-shaped volume within the sample which can extend up to a micrometer below the surface. This means that generally SEM cannot detect individual nanotubes, but only bundles of at least 3–5 nm in diameter. Sample preparation for SEM is fairly simple—a small piece of nanotube material is mounted on the double-sided carbon tape or glued to the substrate with silver paint, in order to assure good electrical contact. Nanotube samples are generally conductive enough to prevent charge buildup, but in rare cases it is necessary to coat samples with a thin layer of platinum to improve conductivity. The most common use of SEM is to estimate the purity of nanotube samples, although, just like in TEM, such estimates are very subjective.

2.1.3. Energy-Dispersive Spectroscopy EDS stands for energy-dispersive spectroscopy. TEMs and SEMs are frequently equipped with EDS detectors, which are able to detect characteristic X-rays generated by various elements when irradiated with high-energy electrons. The EDS spectra indicate the relative amounts of metals and other noncarbon elements in SWCNT material (Figs. 7 and 8). While EDS is good in determining relative amounts of heavy elements, it is notoriously unreliable in determining carbon concentration because of hydrocarbon contamination present in all SEMs and TEMs, and poor sensitivity of EDS detectors in the low-energy end of spectrum, where the carbon peak is found. In addition, EDS spectra taken in TEM usually show large copper peaks coming from copper TEM grids and apertures in the column, and TEM grids are frequently contaminated with silicone grease during manufacturing, producing a weak silicon signal. While the former

S i

C l

C

< – .1 FS= 2K MEM1 :

5.023 ch

C

NC

0

i0

u

keV 261=

10.1 > 40 cts

Figure 7. EDS spectrum of purified HiPco SWCNT obtained in TEM. Iron particles have been removed. The large Cu signal comes from the TEM grid and apertures. A small Si peak can appear due to impurities in SWCNTs as well as silicone grease contaminating the TEM grid.

is largely irrelevant since nanotubes do not normally have any copper in them, the latter is a problem, since silicon is frequently present in nanotube samples. EDS utilizes the same samples as TEM and SEM; therefore we refer reader to Sections 2.1.1 and 2.1.2 for sample preparation.

2.1.4. Microprobe An electron beam microprobe operates on the same principle as EDS, except the beam current is much higher and X-ray detectors are different. Microprobe instruments are CPS 6 Cl 5

4

3

2 Si

1

Fe Ni

Cu

0 2

4

6

8

10

12

Energy (keV)

Figure 8. EDS spectrum of the same purified HipCO SWCNT as in Figure 7, obtained in SEM.

55

Analytical Characterization of SWCNTs

better equipped to produce elemental maps with larger lateral span and better resolution. Sample preparation for microprobe measurements is similar to SEM sample preparation, described in the relevant section. The microprobe is usually calibrated before each use using standard samples. Accuracy of the microprobe measurements is sensitive to the precise focusing of the electron beam. Therefore, the sample surface has to be very smooth and level in order to produce meaningful elemental maps, which is rarely achieved with buckypaper-type samples. Concentration measurements performed with a stationary beam are therefore more accurate, but sample composition in one spot may not be representative of the whole specimen due to inhomogeneity of nanotube material. Figure 9 shows the elemental maps of a target material (graphite with 1% Co and Ni) used for production of nanotubes by laser ablation. The microprobe suffers from the same hydrocarbon contamination problems as EDS, and measured carbon concentration is prone to overestimation. The microprobe has been used in the past to measure the amount of fluorine in fluorinated nanotubes [21], but for the reasons mentioned and in the absence of independent verification, the accuracy of such measurements remains questionable.

2.1.5. AFM Atomic force microscopy is used to determine the topography of flat samples. A sharp tip is scanned over the sample surface, and either tip deflection or change in the amplitude of the tip oscillation serves as feedback in the vertical direction. Accuracy of the height measurement is tremendous, approaching 0.01 nm. Lateral resolution, on the other hand, is determined by the tip diameter and is typically 15–30 nm. Nanotubes have to be deposited onto a suitable smooth substrate for AFM imaging (Fig. 10). The easiest way is 50

50

0

0

5.00 µm 0

0 Data type z range

5.00 µm

Height 20.0 nm

Data type z range

Amplitude 5.00 nm

Figure 10. Tapping mode AFM image of a SWCNT wrapped in PVP (polyvinyl pyrrolidone) and deposited on amine-functionalized silicon oxide surface.

to disperse nanotubes in organic solvent or in water with surfactant, and spin-coat or dry on silicon, mica, graphite (HOPG, or highly oriented pyrolitic graphite), or even glass coverslip. AFM is often used to assess how well nanotubes are dispersed in the solvent, so it could be important to preserve the state of dispersion as nanotubes are transferred to the substrate and dried. For example, nanotubes can be electrodeposited onto a HOPG surface in the DMF suspension [22]. Another approach is to coat silicon surface with a positively charged amine-terminated self-assembled monolayer and dip such substrate in nanotubes dispersed in water in the presence of sodium dodecyl sulphate (SDS), an anionic surfactant giving nanotubes a net negative charge [23]. Other surfactants or polymers [24] can also be used. Arepalli et al. have reported nanotubes being deposited onto polished fused silica substrates directly in the laser oven, as they are produced, bypassing the solvents altogether [25]. Figure 11 shows tapping mode AFM image of nanotubes deposited onto quartz substrate this way. AFM is often used to measure diameters and lengths of nanotubes and nanoropes (nanotube bundles). It is important to mention that because of the poor lateral resolution AFM can only measure height, rather than diameter, of a nanostructure. Therefore it is impossible to distinguish one individual nanotube

b 200 µm Co

200 µm Ni 400

a

c 0

d 2 µm

200 µm C

200 µm SE

Figure 9. E-beam microprobe elemental maps and secondary electron image of a polished surface of carbon/Co/Ni target used for SWCNT production in a laser oven.

Figure 11. Tapping mode AFM image of a SWCNT deposited on a fused silica substrate directly in the laser oven. Reprinted with permission from [25], S. Arepalli et al., Appl. Phys. Lett. 78, 1610 (2001). © 2001, American Institute of Physics.

56

Analytical Characterization of SWCNTs

and a small bundle of nanotubes lying flat on the substrate, having the same height. Nanotube length measurement is straightforward for short tubes not exceeding a few micrometer in length. But if nanotubes are longer and exist in bundles, it becomes impossible to find both ends of a particular nanotube, and one can only estimate nanotube lengths.

A

2.1.6. STM (15,0)

B

(9,0)

dl/dV (a.u.)

Scanning tunneling microscopy is similar to AFM, except that tunneling current between the sample and the sharp metal tip is utilized as a feedback signal. Therefore STM samples have to conduct electric current, which limits substrates to HOPG, heavily doped silicon, and metals. Otherwise the sample preparation is similar to that of AFM samples. The tunneling current is limited to a few atoms on the tip closest to the sample, since its intensity drops exponentially with the distance from the sample. Because of that, lateral resolution of STM is much higher than in AFM and allows one to achieve atomic resolution, which usually calls for measurements in high vacuum and at low temperatures. While AFM probes actual topography of the specimen, STM probes local density of electronic states in the sample, which varies with the distance from the Fermi level. Therefore SEM images taken with different bias between sample and tip are different. STM also allows scanning tunneling spectroscopy measurements, in which the tip is kept stationary above the sample, and bias is varied in order to probe the density of states above and below the Fermi level. Such measurements, combined with helicity measurements from the atomically resolved images [26], have provided the first proof of the relationship between nanotube geometry (helicity and diameter) and electronic structure (Fig. 12).

(12,0)

(15,0)

-1.0

2.2. Spectroscopy 2.2.1. UV-VIS-NIR Absorption and Emission The age-old spectrophotometry is used to measure absorption and transmission of SWCNT samples in the solution phase as well as thin films. The measurements are done a double beam apparatus by comparing the transmission with a reference solvent [27–29] in the wavelength range of 200 to 2000 nm. Figure 13 shows changes of the optical spectra with oxidation and annealing of single wall carbon nanotubes. Recently, excitation spectra and fluorescence of individual SWCNTs (Fig. 14) were also recorded [30]. This measurement is used to identify the chirality and the “n m” designation of the semiconducting tubes in the raw HiPco material (Fig. 15). The procedure involves extensive sonication of the material in a surfactant (SDS) followed by very high level centrifuging (200,000 g) which yields very short (couple of hundred nm long) individual tubes wrapped with surfactant. The energy levels responsible for the sharp absorption and fluorescence features arise from “van Hove” singularities in the electronic density of states [31]. Metallic tubes can be monitored in the absorption spectrum only and no fluorescence can be observed because the conduction band provides a way for rapid nonradiative relaxation. The features from the semiconducting tubes can be used to determine the primary (S11), secondary (S22), as well as tertiary (S33) transitions. Fluorescence is only observed from individual

-0.5 0.0 0.5 Energy (eV)

1.0

Figure 12. (A) Typical atomically resolved STM image of a (15, 0) SWCNT. The image was recorded in the constant-current mode with bias voltage of 0.65 V and current 0.15 nA. Scale bar, 1 nm. (B) Tunneling conductance data, dI/dV, for different zigzag SWCNTs, with corresponding calculated densities of state shown below each experimental curve. Reprinted with permission from [81], M. Ouyang et al., Science 292, 702 (2001). © 2001, American Association for Advancement of Science.

tubes and disappears when bundling takes place [30]. However, absorption can still be noticed and it may be possible to determine the bundle size from the width of the spectral features. The absorption spectra of the SWCNTs can also be used to determine the stability of nanotube suspensions. If the nanotubes tend to bundle up, floculate, and fall out of the solution, the absorption decreases and this change in the absorptivity as function of time is used to assess the dispersion stability [32]. Atomic Absorption and Inductively Coupled Plasma These two techniques are used for qualitative as well as quantitative estimation of metals in materials. Atomic absorption (AA) uses selective absorption of “fingerprint” spectral lines either from a broadband source or from sharp hollow-cathode lamp sources that provide very sharp atomic

57

Analytical Characterization of SWCNTs

0.85 0.80

0.56

1.13

1.29

1000

1200

1400

1600

zigzag

Wavelength (nm)

0.75

Absorbance

0.80 Diameter (nm)

0.70 0.65

arm

cha

0.60

ir

0.55 0.50 0.45 0.40

(d) (c)

0.35 0.30 400

600

800

1000

1200

1400

1600

1.0 nm diameter

(b) (a)

Wavelength (nm)

Figure 13. UV-visible-near IR absorption spectra of SWCNT: (a) raw HiPco tubes, (b) after 225  C oxidation step, (c) after 325  C oxidation step, (d) after 425  C oxidation step. Each oxidation step is completed by HCl sonication, vacuum-drying, and annealing at 800  C in Ar for 1 hour. Spectra have been normalized at 925 nm and scale bar shows adjusted tube diameter. Reprinted with permission from [11], I. W. Chiang et al., J. Phys. Chem. B 105, 8297 (2001). © 2001, American Chemical Society.

line radiation emission. Hollow cathode lamps and multidetector arrays are used with spectrometers operating in “echelle” modes to provide simultaneous detection of several elements with high sensitivity [33]. High temperature furnaces are used to vaporize the material with a short heat pulse and thereby provide gas phase atoms for AA study. Current AA methods extend from VUV to IR wavelength regions and use laser sources to provide atomic radiations. This is especially true in the IR where solid state diode lasers are used with very high resolution and increased sensitivity. 1600 1500 1400 1300

1200

1100

1000

900 nm

Normalized absorbance or emission

1.0 532 nm excitation T = 296 K

Absorption 0.8

0.6

0.4

Emission

0.2

0.0 7,000

8,000

9,000

10,000

11,000

Frequency (cm-1) Figure 14. Emission spectrum (red) of individual fullerene nanotubes suspended in SDS micelles in D2 O excited by 8 ns, 532 nm laser pulses, overlaid with the absorption spectrum (blue) of the sample in this region of first van Hove bandgap transitions. The detailed correspondence of absorption and emission features indicates that the emission is band gap photoluminescence from a variety of semiconducting nanotube structures. Reprinted with permission from [30a], M. J. O’Connell et al., Science 297, 593 (2002). © 2002, American Association for Advancement of Science.

Figure 15. Results from the deduced spectral assignment. Observed nanotube structures marked as green hexagons with white (n m) labels on a graphene lattice. Other structures have blue or red labels for (n − m) mod 3 = 1 or 2, respectively. Reprinted with permission from [30b], S. M. Bachilo et al., Science 298, 2361 (2002). © 2002, American Association for Advancement of Science.

The inductively coupled plasma (ICP) technique involves injecting the material as a spray into a plasma torch which vaporizes the material. This process pumps the atoms to excited states which emit “fingerprint” spectral lines and can be used for quantitative elemental analysis of the material. This technique coupled with a mass spectrometer is useful for a complete elemental analysis for even organic compounds including polymers. Both of these techniques have been used to monitor the metal contents in SWCNT material [34]. It is to be understood that metals encapsulated in graphite may be difficult to detect by ICP and the quantification of the metal content determined by this method may be underestimated.

2.2.2. IR Absorption and Emission The absorption and emission in the infrared region (2 to 20 micrometer) arise from vibrational modes of the molecules constituting the material. The spectra arising from vibrational energy of the molecular bonds such as C C, C C, C H, C H2 , CH3 , C F, C N, O H, N H, etc. are commonly used as “fingerprints” of the material. In principle, the Kramer–Kronig relationship can be used to deduce the dielectric function of the material from the IR spectrum [35]. Data collection in the IR region normally done in double beam spectrometers can be done faster with fourier transform spectrometers and the capabilities are easily extended to very small solid samples using accessories. The IR region of the SWCNT is very very weak because of the low absorption cross-section of the carbon nanotubes (Fig. 16). In the recent work on SWCNT films and suspensions IR is used to monitor the functional groups attached to the SWCNTs (Fig. 17) as well as the effect of surfactants [35–38]. The IR spectrum is also used to analyze the condensable gases that are released when the material is heated (for example in conjunction with TGA [39]).

2.2.3. Raman Raman spectroscopy is a powerful spectroscopic tool in providing “fingerprint” spectra of the molecular vibrational modes of the material [40]. The Raman spectra can be obtained in any wavelength region using different laser

58

Analytical Characterization of SWCNTs

(b)

180 1592

(a)

1570

1320 nm, 20 W/cm2

4000

3000

2000

Wavenumbers

1000

(cm-1)

Figure 16. ATR-IR spectra of the SWCNT. (A) SWCNT fluorinated to ∼C5 F stoichiometry. (B) Cut SWCNT. Reprinted with permission from [39b], Z. Gu et al., Nano Lett. 2, 1009 (2002). © 2002, American Chemical Society.

wavelengths. The recent advances that include FT Raman and confocal micro-Raman provide rapid data acquisition of even micrometer size materials in addition to the normal solid, liquid, and gas phase studies. The Raman spectra of SWCNTs can be separated as three distinctive regions (Fig. 18), one in the small wavenumber region called radial breathing mode (RBM), the second in the intermediate wavenumber region called transverse mode (TM), and the third at higher wavenumber regions emanating from harmonics and combinations of these two fundamental modes [41, 42]. The RBM modes are resonantly enhanced because of van Hove singularities and can be used to monitor the changes in the diameter distribution of SWCNTs [43]. The TM (G-Mode) region also shows spectral features that can be attributed to amorphous carbon (D-mode). The intensity ratio of the D-band and G-band (“D/G” ratio) is used by many researchers to determine the relative purity of the SWCNT material [44, 45]. However, it is noted that the

100

1471

1660 1583

85

2848

90

80 75

2818

% Transmittance

95

70 65 4000

2000

Wavenumbers (cm-1)

Figure 17. Mid-IR spectrum of functionalized SWCNT (thin film on quartz substrate). s-SWCNT-CONH(CH2 )17 CH3 showed peaks at 2918 and 2845 [(C H) stretch modes of the alkyl chain], 1660 [(C O) stretch of the amide], 1583 [ (C C) stretch of the SWCNTs], and 1471 cm−1 [ (C H) bend of the alkyl chain]. Reprinted with permission from [82], M. A. Hamon et al., Chem. Phys. Lett. 347, 8 (2001). © 2001, Elsevier.

Raman Intensity (arbitrary units)

191

502 558

858

1258

1548

1067 1110 1593 1570

1064 nm, 5

815 860 922

359 464 157

W/cm2

1549 1278

1294

169

780 nm, 12 W/cm2

205 476

584

1568

1592

1546

861 1542

192

647.1 nm, 2 W/cm2

1563 1589

1321

236 390 427

739

861

1073

1593

1567

186

514.5 nm, 2 W/cm2 376

1347 1526

755 855

500

1550

1000

1500

Frequency (cm-1) Figure 18. Room-temperature Raman spectra of purified SWCNTs excited at five different laser frequencies. The laser frequency and power density for each spectrum are indicated, as are the vibrational frequencies. Reprinted with permission from [41b], A. M. Rao et al., Science 275, 187 (1997). © 1997, American Association for Advancement of Science.

broad “D-band” becomes sharper with increasing “purification” and may therefore contain a SWCNT signature. This also occurs when the nanotubes are functionalized. Recent Raman work is focused on monitoring isolated SWCNTs (Fig. 19) and their chiralities [46]. Polarized Raman studies are used to assess the alignment of SWCNT fibers by changing the polarization of the laser with respect to the fiber axis [47, 48]. There is a preferential enhancement when the polarization axis is parallel to the nanotube axis and this “parallel to perpendicular” ratio is used to estimate the extent of alignment. Raman is also used to monitor strain (Fig. 20) in the SWCNT fibers and their load transfer to the composite matrices [49]. Temperature variation studies of Raman spectra are also used to assess the intrinsic nature and their dependence of the vibrational modes [50].

59

Analytical Characterization of SWCNTs 7

G band

1593.5

RBM

1% SWNT RT, (17,7)

1590 1575

6

compression

5

εs= - 0.000%

1591

(17,3)

(8) 169(8)

1568 (8)

Intensity (arb. units)

(11)

145(13)

Raman intensity

(9)

782 nm εs - surface strain

4

εs= - 0.048%

εs= - 0.098% 3 εs= - 0.140% 2

(15,2)

1590

εs= - 0.182%

(8)

1

1560 (8)

120

170

220

Frequency (cm-1)

1530

εs= - 0.198%

1570

1595.4

195(5)

0

1610

1550

Frequency (cm-1)

1575

1600

1625

1650

Raman shift (cm-1)

A + h −→ A+ + e− KE = h − BE The exact binding energy of an electron varies with the oxidation state as well as chemical environments giving rise to

0.15

284eV

(A) 22K 2K 1.8K 1.6K 1.4K 1.2K 1K 800

0.15

600 400 200 0 1098

8:20

A variety of techniques are available [XPS, electron energy loss spectroscopy (EELS), etc.] to monitor the surface characterization of thin films consisting of SWCNTs. The techniques are all based on monitoring electrons or photons emitted from the material when excited with high energy electrons or X-rays. Photoelectron spectroscopy utilizes photoionization and energy-dispersive analysis of the emitted photoelectrons [51]. This is useful for the study of the composition as well as the electronic state of the surfaces. In X-ray photoelectron spectroscopy, the photon is absorbed by an atom in a molecule or solid, leading to ionization and the emission of a core (inner shell) electron. By contrast, ultraviolet photons can be used to probe valence levels. The kinetic energy (KE) distribution of the emitted photoelectrons is measured using an electron energy analyzer and allows one to deduce the binding energy (BE) of the electron:

“chemical shifts.” An XPS spectrum of a nanotube film is shown in Figure 21. Another technique that can be used to determine the atomic neighbors is extended X-ray absorption fine structure. The modulations noticed in the X-ray absorption spectrum arise from interference of the outgoing photon with the reflections from the neighboring atoms. This spectrum can then be used to determine the size and distances of the neighboring atoms from the atom of interest. Herrera et al. used EXAFS and X-ray absorption near edge structures

285.4eV

2.2.4. Surface Characterization Tools

Figure 20. Raman spectra (782 nm excitation) of composite sample (1% SWCNT embedded in epoxy). G-mode frequency increases proportionally to the applied compressive stress (s
.

Intensity (arb.units)

Figure 19. The RBM and G-band Raman spectra for three isolated semiconducting SWCNTs resonant with Elaser ; that is, Elaser = 241 eV for the spectrum in the top, labeled (17, 7), and Elaser = 254 eV for the other two spectra, labeled (17, 3) and (15, 2). The spectra are taken at three different spots on the substrate. The frequencies (linewidths) of the principal peaks for the RBM, −G and +G features, are displayed in cm−1 . The shoulder observed to the right side of the RBM spectral feature comes from the Si substrate. Reprinted with permission from [83], A. Jorio et al., Phys. Rev. B 65, 155412 (2002). © 2001, American Physical Society.

998

898

798

698

598

498

398

298

198

98 9

Binding Energy (eV)

Figure 21. XPS of as-prepared SWCNT film. Reprinted with permission from [84], T. V. Sreekumar et al., Chem. Mater. 15, 175 (2003). © 2003, American Chemical Society.

60

Analytical Characterization of SWCNTs

(XANES) to monitor the production of SWCNTs under varying catalyst concentrations [52]. The XANES spectra are shown in Figure 22. The EELS technique has also been used to assess the valence state of carbon atoms [53].

2.3. Chromatography Chromatographic based techniques are used for solubilized SWCNTs to separate different components and constituents of SWCNT material. Different mass fractions can be identified with mass spectrometry in conjunction with gas chromatography and matrix desorption. This information in turn can be utilized to establish the nature and extent of the derivatization of the SWCNT materials and matrices with SWCNTs. Early successes are achieved in “purification” of the SWCNT material by using HPLC columns [54]. Recently, HPLC has been used to separate the SWCNT tubes and ropes by their lengths [55].

2.3.1. GC MS and TOF MS The identification and quantification of trace organic constituents of the SWCNT material extractable by water or any organic solvent are done by gas chromatographic methods. In gas chromatography, a mobile phase (a carrier gas such as nitrogen or helium) and a stationary phase (capillary column coating) are used to separate individual compounds. When the sample solution is introduced into the column, the organic compounds are vaporized and moved through the column by the carrier gas. They travel through the column at different rates, depending on differences in partition coefficients between the mobile and stationary phases. The electrolytic conductivity detector is used to monitor halocarbons and nitrogen containing compounds. The electron

capture detector usually is used for the analysis of compounds that have high electron affinities such as halogens, peroxides, quinines, and nitro groups. The flame ionization detector is also widely used because of its high sensitivity to organic carbon-containing compounds. Mass analysis of the large fullerenes found in the SWCNT material is carried out using a time-of-flight mass spectrometric detector connected to a matrix assisted laser desorption/ionization (MALDI) system as shown in Figure 23 [56].

2.3.2. HPLC This method is used for identification and quantification of trace organic constituents and soluble fullerenes extractable by organic solvents like toluene, dichlorobenzene, carbon disulfide, and others. High performance liquid chromatography is an analytical technique in which a liquid mobile phase transports a sample through a column containing a stationary phase. Selection of a suitable column is very crucial and the researchers used columns including Styragel HMW7, C18 , and Plgel Mixed-A columns. The interaction of the sample with the stationary phase selectively retains individual compounds and permits separation of sample components. Of special concern for HPLC is the masking of the absorbance region of the HPLC mobile phase and its additives. This may reduce the range and sensitivity of the detector to sample components. This test method is applied for the determination of monoaromatic, diaromatic, and polyaromatic hydrocarbon contents in toluene extracted solutions from raw nanotube material produced by arc, laser, or HiPco methods [57]. The total aromatic content in % m/m is calculated from the sum of the individual aromatic hydrocarbon types (Fig. 24). This test method is calibrated for distillates containing from 4 to 40% (m/m) monoaromatic hydrocarbons, 0 to 20% (m/m) diaromatic 0.10

20 Hr fresh

0.00

10 min

0.10

10 Hr

30 min 0.00

Co foil 0.10

5 Hr a

0.00

b

0.10

RAW

c 0.00

d

0

1000

2000

3000

4000

5000

6000

m/z 7680

7700

7720

7740

7760

7780

Energy (eV) Figure 22. Near edge spectra XANES obtained at the K edge of Co (E0 = 7709 eV) on a fresh Co  Mo/SiO catalyst (Co  Mo = 1  2), pretreated in H2 (500  C) /He(700  C), and after the growth of carbon nanotubes by decomposition of 50% CO/He at 700  C for reaction periods of 10 and 30 min. The XANES of a Co foil is included for comparison. Reprinted with permission from [66], W. E. Alvarez et al., Carbon 39, 547 (2001). © 2001, Elsevier.

Figure 23. LDI-TOF mass spectral patterns recorded under identical conditions on HiPco raw samples soft baked for varying times. The raw sample showed a mass spectral span of 1800–4500 amu with the maximal intensity at 2600 amu. The masses were evenly separated at a spacing of 24 amu. Note that the mass number for maximal intensity has shifted to higher mass numbers in the oxidized samples indicating a lower threshold for low-molecular weight carbonaceous species. Reprinted with permission from [56], S. Ramesh et al., J. Phys. Chem. B 107, 1360 (2003). © 2003, American Chemical Society.

61

Analytical Characterization of SWCNTs

0.020

0.015

Starting material

0.010

C84

C76 C76-I C76-II

C70–07

C70

C60-0

C60

0.005

0.000 2.00

4.00

6.00

8.00

10.00

12.00

14.00

16.00

18.00 20.00

Minutes

Figure 24. HPLC chromatogram of toluene extract of SWCNTs produced in a laser oven. The Y -axis is absorption at 316 nm wavelength. This sample has ∼4% by weight of C60 and C70 , plus a small amount of higher soluble fullerenes (C76 , C78 , and C84
.

hydrocarbons, 0 to 6% (m/m) polycyclic aromatic hydrocarbons, and 4 to 65% (m/m) total aromatic hydrocarbons. The precision of this test method has been established for an extract of soluble fullerenes and their blending components, containing from 4 to 40% (m/m) monoaromatic hydrocarbons, 0 to 20% (m/m) diaromatic hydrocarbons, 0 to 6% (m/m) polycyclic aromatic hydrocarbons, and 4 to 65% (m/m) total aromatic hydrocarbons.

2.4. Magnetic Resonance Two of the prominent magnetic resonance techniques are electron spin resonance (ESR) and nuclear magnetic resonance. In the case of ESR, one monitors the precession of the valence electron in a strong magnetic filed. The ESR spectrum can be used to study the effect of ligands and chemical environments by monitoring the shape, symmetry, width, and splitting of the ESR lines. Figure 25 shows that ESR spectral features change during purification. Petit et al. [58] reported microwave resistivity and ESR of laser grown SWCNTs and observed that the system remained metallic even at low temperatures. Theoretical work by DeMartino et al. [59] successfully explained the double-peak nature of the ESR spectrum to spin-charge separation. The nuclear magnetic resonance methods are used to measure the hydrogen content (from 1 H NMR) and carbon content (by the 13 C NMR). Kleinhammes et al. [60] studied the hydrogen adsorption in carbon nanotubes by following the kinetics of adsorption and desorption as a function of temperature and pressure. The electronic structures of the SWCNTs are also determined by relaxation times of 13 C NMR signals [61]. The fast relaxing component is attributed to metallic tubes. Figure 26 shows the effect of spinning (MAS stands for magic angle spinning) on the NMR spectra of SWCNT samples. It was found to account for one-third of the sample which is consistent with random chirality of the SWCNTs. It is also noted that exposure to small amounts of oxygen can drastically change the relaxation behavior of the

Purified material

0

1000

2000

3000

4000

5000

6000

Magenic Field [Gauss] Figure 25. Electron spin resonance signal of unpurified and purified SWCNTs recorded at 9 GHz and at 300 K as a function of the static magnetic field. The former shows the huge ferromagnetic resonance line coming from the catalytic particles, while this signal is almost absent on the latter. Reprinted with permission from [86], Le T.-N et al., Nano Lett. 2, 1349 (2002). © 2002, American Chemical Society.

NMR signals. The proximity of paramagnetic or ferromagnetic impurities (catalysts used in SWCNT production) may cause broadening of the NMR linewidths [62].

2.5. Thermal Methods The following thermal methods rely on monitoring the weight, heat content, and reactivity of the SWCNTs as a function of elevated temperatures. These constitute one set of analytical techniques where the sampled material is destroyed or altered permanently. However, these can be used for quantitative estimation of purity of SWCNT samples and also yield very crucial information on thermal stability of these materials. The change in heat content is monitored by DSC where the phase transformations from crystalline to amorphous and/or glassy states are recorded.

2.5.1. TGA and TPO Thermogravimetric analysis is used to monitor the changes in weight of the sample with increasing temperature [63]. Normally, this is done in air and at a ramp rate of 5 to 10 degrees per minute. The effects of compacting, ramp rate of the furnace, and type of gas used for TGA are discussed

62

Analytical Characterization of SWCNTs 500 475 450

400

Standart deviation of T, °C

To, °C

425

375 350 325

15 10 5 0 0

10

20

30

Ramp rate, °C/min

Static

300 0

5

10

15

20

25

30

Ramp rate, °C/min Figure 28. Burning temperature (T0 , defined as a maximum in the weight loss rate) and its standard deviation dependence on the ramp rate in TGA. Run parameters and samples are similar to those in Figure 28. T0 increases by ∼70  C as heating rate goes from 1 to 30  C/min. Standard deviation of T0 also increases, due to uncontrolled combustion above 5  C/min.

400

200

0

-200

Frequency (ppm/TMS) Figure 26. The static and MAS (spun at 11.7 kHz) 13 C spectra of SWCNT. The isotropic shift is 124 ppm based on the MAS spectrum. The fit is the powder pattern of a chemical shift tensor with 11 = 195 ppm, 22 = 160 ppm, and 33 = 17 ppm. Reprinted with permission from [61], X.-P. Tang et al., Science 288, 492 (2000). © 2000, American Association for Advancement of Science.

in a recent presentation [64]. A faster ramp rate is observed to result in premature combustion and excessive compacting reduces the extent of oxidation (Figs. 27, 28). The peak of the derivative (dm/dT ) plotted in Figure 27 is now regarded as indicative of the temperature of oxidation. For SWCNTs, this will shift to lower temperatures for small diameter tubes and derivatized tubes (which can burn easily). At temperatures beyond 1073 K, the ash content can be attributed to metal oxides and therefore can be used to estimate the metal content of the SWCNT material. This technique is one of the few quantitative techniques that can be used to compare different purification methods (by comparing the metal contents of the samples).

1.5

Weight

1

1 0.8 0.5 0.6

0

0.4 0.2

Weight loss rate

2

1.2

-0.5

0

100

200

300

400

-1 500

Temperature (°C) Figure 27. TGA traces of unpurified HipCO nanotubes (100 sccm air, 2.5  C/min ramp rate; the same sample was measured three times). Small differences in the TGA traces are attributed to the inhomogeneity of the sample.

Also, the shape and size of the derivative signals used are indicative of the “temperature of oxidation.” The oxidation temperature is noted to decrease with smaller diameter tubes and also when the nanotubes have extensive derivatization and/or defects [65]. In the temperature programmed oxidation technique, the material is submitted to a programmed temperature increase in an oxidizing mixture of gas (small percentage of oxygen in inert gas) and the reduction or oxidation rates are continuously measured by monitoring the change in composition by chemical means or by weight. In general, TPO studies are carried out under low partial pressure of the reactive gas. In this way it is possible to observe the intermediate reactions, depending on analytical conditions such as temperature rate, flow rate, and concentration of reactive gas. Alvarez et al. used this to separate out the peaks in the TGA data (Fig. 29)

CO2 signal (a.u.)

MAS

Co: Mo = 2:1 Co: Mo = 1:1 Co: Mo = 1:2 Co: Mo = 1:4

100

200

300

400

500

600

700

800

Temperature (°C) Figure 29. TPO traces of SWCNTs produced by decomposition of CO on Co:Mo catalyst of varying composition. A continuous flow of 5% O2 in He is passed over the material while the temperature is linearly increased at 11  C/min ramp rate. The evolution of CO2 produced by the oxidation of the carbon species is monitored by a mass spectrometer. The arrow indicates the center of the peak corresponding to the oxidation of the SWCNT. Reprinted with permission from [87], B. Kitiyanan et al., Chem. Phys. Lett. 317, 497 (2000). © 2000, Elsevier Science.

63

Analytical Characterization of SWCNTs

and could identify and estimate the different types of carbon in the SWCNT sample [66].

2.5.2. DSC Differential scanning calorimetry uses an indirect method of estimating the heat content of the sample as a function of temperature. The DSC spectrum can be used to monitor exothermic and endothermic reactions, and transition temperatures between different phases. This becomes especially useful when nanotubes are used to make composite materials for strength or conductivity enhancements. Also, peaks in the derivative signals of DSC can be used to determine curing temperatures, glass transition temperatures, and crystallization temperatures [67]. A

100 µm

2.6. Miscellaneous Methods 2.6.1. XRD One of the early methods used to monitor the extent of crystallinity was X-ray diffraction of SWCNT material [68]. The sharpness of the peaks at low angles increases with increased purity and crystallinity (Fig. 30). The broad background in the XRD spectrum at larger angles is attributed to the extent of amorphous content. This technique has also been used for estimating the diameter distribution [69] and the phonon density of states of SWCNTs [70].

2.6.2. Optical Microscopy Resolution of the optical microscope is limited by numerical aperture and wavelength of light and does not exceed 400–500 nm. So it is not useful for imaging nanotubes and even large bundles. But optical microscopy has been utilized to assess nanotube dispersion in polymer films [71]. Nanotubes were dispersed in polyimide, and optical transmission micrographs of the resulting films were used to compare the degree of nanotube dispersion (Fig. 31). Dispersion of purified nanotube material in DMF is studied by drying a

DOUBLE-LASER SWnT

X-ray Counts/12 hours

400000

1200°C GRAPHITE TARGET 1.2 at.% Co/Ni 1000°C ANNEAL

300000

Co/Ni 200000

Gr(002)

0.5

1.0

1.5

2.0

2.5

3.0

3.5

200 µm

Figure 31. Optical transmission images of a SWCNT (0.05%) dispersed relatively well (A) and worse (B) in polyimide film.

drop of nanotubes suspended in DMF on a glass slide. For a quantitative comparison, one can resort to algorithms that can count the size and number of aggregates in the image. It is to be noted that the drying process may alter the state of aggregation of nanotube material. Optical microscopy is more useful for dispersability study when higher concentrations of nanotubes (>0.1 mg/mL) are used. For low concentrations (0.01 mg/mL), normal transmission study may be sufficient for this evaluation (see Section 2.2.1).

2.6.3. Surface Area and Surface Energy Measurements

(100) (200)

100000

0.0

B

4.0

4.5

2Θ Figure 30. X-ray diffraction pattern of laser oven SWCNTs annealed in vacuum at 1000  C. Reprinted with permission from [7], A. Thess et al., Science 273, 483 (1996). © 1996, American Association for Advancement of Science.

A variety of applications including supercapacitors and rechargeable batteries depend on the availability of the large surface area of the SWCNTs. It is to be noted that the available surface area varies with the functional groups attached to the nanotubes as well as the number of dangling bonds. This in turn determines the surface energy of the nanotube material that affects the extent of adhesion of the SWCNTs to the surrounding matrix material. The BET surface area analyzer estimates the specific external surface of a solid by determining the volume of a specific gas that is absorbed

64

Analytical Characterization of SWCNTs

N2 absorbed (cm3/g STP)

1200 1000

Raw Step 2 only 2 step process

800 600 400 200 0 0

0.2

0.4

0.6

0.8

1

1.2

Relative Pressure (P/Po)

Figure 32. N2 adsorption isotherms (77.6 K) for the raw SWCNTs and a sample purified using the complete two-step process. Isotherm for a sample which skipped Step 1 (DMF/EDA treatment) and was subjected only to the HCl treatment/wet oxidation (Step 2) is also shown for comparison. Reprinted with permission from [77], M. Cinke et al., Chem. Phys. Lett. 365, 69 (2002). © 2002, Elsevier Science.

under controlled conditions [72]. Surface area and pore volume can be determined using a commercial instrument such as a BET COULTER SA 3100. The method consists of onehour gas evacuation at 393 K and of determination of the pore volume introducing helium gas, followed by adsorption of nitrogen at the temperature of liquid nitrogen (77 K). The system performs automatic single-point and multipoint BET surface area, full adsorption and desorption isotherms, and pore volume distributions. One can use other gases such as krypton, hydrogen, carbon dioxide, and saturated hydrocarbons. A typical single-point surface area analysis takes a few hours, while a full adsorption/desorption isotherm can take up to 3 days to complete (Fig. 32). The surface areas of SWCNTs are reported to be in the range of 300 to 3000 m2 /g. Early work on the surface area determination of SWCNTs was prompted by the possibility of high hydrogen adsorption [73–75]. Other applications such as supercapacitors and catalysts required precise surface area and pore size measurements [76, 77]. Improved performance of supercapacitors using single walled CNT electrodes was reported by Young Hee Lee of Sung Kyun Kwan University, Korea. The heat treatment at high temperature was necessary to increase the capacitance and reduce the CNT-electrode resistance. The increased capacitance was well explained by the enhancement of the specific surface area and the abundant pore distributions at lower pore sizes of 30 to 50 angstroms estimated from the BET (N2
measurements [78]. Direct measurement of surface energy can use inverse gas chromatography (Hans Darmstadt, Univ. Laval [79]). Surface energy of the SWCNTs can also be estimated by measuring the contact angle with different solvents and surfaces. Initial experiments seem to indicate low contact angle with water and other common solvents [80].

GLOSSARY Bucky peapods Nanoscale material consisting of string of buckyballs (cage-like carbon clusters) enclosed inside carbon nanotube.

Chemical vapor deposition Chemical process based on catalytically assisted decomposition of gas-phase precursor molecules. Chirality Property of an object that is not superimposable on its mirror image. Chromatic aberrations Wavelength-dependent artifacts in imaging that occur because the refractive index of the lens material varies with wavelength. Field emission electron gun (FEG) Electron source based on extremely sharp tip. Electric field in the tip vicinity is high enough to allow electron emission without thermal assistance (or with very low thermal assistance). Nanohorns Nanoscale material made of graphene sheets rolled into capped cones. Nanorope Term used to describe a bundle of nanotubes held together by Van der Waals forces. Nanoropes are not twisted, unlike real ropes. Numerical aperture The sine of the vertex angle of the largest cone of meridional rays that can enter or leave an optical system or element, multiplied by the refractive index of the medium in which the vertex of the cone is located. Spherical aberrations Imaging artifact caused because a spherical lens does not focus parallel rays to a point, but along a line.

ACKNOWLEDGMENTS The authors acknowledge financial support from NASA contract NAS9-19100 to Lockheed Martin and encouragement from NASA-JSC carbon nanotube group members.

REFERENCES 1. M. S. Dresselhau, G. Dresselhaus, and P. C. Eklund, in “Science of Fullerenes and Carbon Nanotubes.” Academic Press, New York, 1996. 2. B. I. Yakobson and R. E. Smalley, Am. Sci. 85, 324 (1997). 3. C. Dekker, Physics Today 22 (1999). 4. L. Yowell, B. Files, C. Scott, B. Mayeaux, E. Sullivan, S. Arepalli, P. Nikolaev, O. Gorelik, W. Holmes, and V. Hadjiev, AIAA IAC02-IAA.12.1.01 (2002). 5. “Carbon Nanotubes: Synthesis, Structure, Properties and Applications” (M. S. Dresselhaus, G. Dresselhaus, and P. Avouris, Eds.), Topics in Applied Physics, Vol. 80. Springer-Verlag, Heidelberg, 2001. 6. C. Journet, W. K. Maser, P. Bernier, A. Loiseau, M. Lamy de la Chapelle, S. Lefrant, P. Deniard, R. Lee, and J. E. Fischer, Nature 388, 756 (1997). 7. A. Thess, R. Lee, P. Nikolaev, D. Hongjie, P. Petit, J. Robert, C. Xu, Y. H. Lee, S. G. Kim, A. G. Rinzler, D. T. Colbert, G. E. Scuseria, D. Tománek, J. E. Fischer, and R. E. Smalley, Science 273, 483 (1996). 8. H. Dai, J. Kong, C. Zhou, N. Franklin, T. Tombler, A. M. Cassell, S. Fan, and M. Chapline, J. Phys. Chem. 103, 11246 (1999). 9. P. Nikolaev, M. J. Bronikowski, R. K. Bradley, F. Rohmund, D. T. Colbert, K. A. Smith, and R. E. Smalley, Chem. Phys. Lett. 313, 91 (1999). 10. O. Gorelik, P. Nikolaev, and S. Arepalli, NASA contractor report, NASA/CR-2000-208926, 2001. 11. I. W. Chiang, B. E. Brinson, A. Y. Huang, P. A. Willis, M. J. Brownikowski, J. L. Margrave, R. E. Smalley, and R. H. Hauge. J. Phys. Chem. B 105, 8297 (2001).

Analytical Characterization of SWCNTs 12. J. L. Margrave, J. L. Zimmerman, R. K. Bradley, C. B. Huffman, and Robert H. Hauge, Chem. Mater. 12, 1361 (2000). 13. A. C. Dillon, T. Gennett, K. M. Jones, J. L. Alleman, P. A. Parilla, and M. J. Heben, Adv. Mater. 11, 1354 (1999). 14. R. F. Service, Science 300, 243 (2003). 15. S. Iijima, Nature 354, 56 (1991). 16. D. S. Bethune, C. H. Kiang, M. S. de Vries, G. Gorman, R. Savoy, J. Vazguey, and R. Beyer, Nature 363, 605 (1993). 17. B. W. Smith, M. Monthioux, B. W. Smith, M. Monthioux, and D. E. Luzzi, Nature 396, 323 (1998). 18. S. Iijima, M. Yudasaka, R. Yamada, S. Bandow, K. Suenaga, F. Kokai, and K. Takahashi, Chem. Phys. Lett. 309, 165 (1999). 19. A. Thess, R. Lee, P. Nikolaev, H. J. Dai, P. Petit, J. Robert, C. H. Xu, Y. H. Lee, S. G. Kim, A. G. Rinzler, D. T. Colbert, G. E. Scuseria, D. Tomanek, J. E. Fischer, and R. E. Smalley, Science 273, 483 (1996). 20. P. Nikolaev, Gas Phase Production Of Single-Walled Carbon Nanotubes, Ph.D. Thesis, Rice University, Houston, TX, 2000. 21. E. T. Mickelson, C. B. Huffman, A. G. Rinzler, R. E. Smalley, R. H. Hauge, and J. L. Margrave, Chem. Phys. Lett. 296, 188 (1998). 22. J. Liu, A. G. Rinzler, H. J. Dai, J. H. Hafner, R. K. Bradley, P. J. Boul, A. Lu, T. Iverson, K. Shelimov, C. B. Huffman, F. RodriguezMacias, Y. S. Shon, T. R. Lee, D. T. Colbert, and R. E. Smalley, Science 280, 1253 (1998). 23. J. Liu, M. J. Casavant, M. Cox, D. A. Walters, P. Boul, W. Lu, A. J. Rimberg, K. A. Smith, D. T. Colbert, and R. E. Smalley, Chem. Phys. Lett. 303, 125 (1999). 24. M. J. O’Connell, P. Boul, L. M. Ericson, C. Huffman, Y. H. Wang, E. Haroz, C. Kuper, J. Tour, K. D. Ausman, and R. E. Smalley, Chem. Phys. Lett. 342, 265 (2001). 25. S. Arepalli, P. Nikolaev, W. Holmes, and B. S. Files, Appl. Phys. Lett. 78, 1610 (2001). 26. T. W. Odom, J. L. Huang, T. W. Odom, J. Huang, P. Kim, and C. M. Lieber, Nature 391, 62 (1998). 27. W. Chiang, B. E. Brinson, R. E. Smalley, J. L. Margrave, and R. H. Hauge, J. Phys. Chem. B 105, 1157 (2001). 28. O. Jost, A. A. Gorbunov, W. Pompe, T. Pichler, R. Friedlein, M. Knupfer, M. Reibold, H.-D. Bauer, L. Dunsch, M. S. Golden, and J. Fink, Appl. Phys. Lett. 75, 2217 (1999). 29. M. E. Itkis, S. Niyogi, M. E. Meng, M. A. Hamon, H. Hu, and R. C. Haddon, Nano Lett. 2, 155 (2002). 30. (a) M. J. O’Connell, S. M. Bachilo, C. B. Huffman, V. C. Moore, M. S. Strano, E. H. Haroz, K. L. Rialon, P. J. Boul, W. H. Noon, J. Ma, R. H. Hauge, R. B. Weisman, and R. E. Smalley, Science 297, 593 (2002). (b) S. M. Bachilo, M. S. Strano, C. Kittrell, R. H. Hauge, R. E. Smalley, and R. B. Weisman, Science 298, 2361 (2002). 31. J. W. G. Wilder, L. C. Venema, A. G. Rinzler, R. E. Smalley, and C. Dekker, Nature 391, 59 (1998). 32. S. Arepalli, P. Nikolaev, O. Gorelik, V. Hadjiev, W. Holmes, B. Files, and L. Yowell, submitted for publication. 33. J. R. Dean and D. J. Ando, “Atomic Absorption and Plasma Spectroscopy,” 2nd ed. Wiley, New York, 1997. 34. K. L. Strong, D. P. Anderson, K. Lafdi, and J. N. Kuhn, Carbon 41, 1477 (2003). 35. T. Pichler, M. Knupfer, M. S. Golden, J. Fink, A. Rinzler, and R. E. Smalley, Phys Rev. Lett. 80, 4729 (1998). 36. F. Pompeo and D. E. Resasco, Nano Lett. 2, 369 (2002). 37. J. Sun, J. Colloid Interface Sci. 260, 89 (2003). 38. P. J. Boul, J. Liu, E. T. Mickelson, C. B. Huffman, L. M. Ericson, I. W. Chiang, K. A. Smith, D. T. Colbert, R. H. Hauge, J. L. Margrave, and R. E. Smalley, Chem. Phys. Lett. 310, 367 (1999). 39. (a) Y. S. Lee, T. H. Cho, B. K. Lee, J. S. Rho, K. H. An, and Y. H. Lee, J. Fluor. Chem. 120, 99 (2003). (b) Z. Gu, H. Peng, R. E. Smalley, and J. L. Margrave, Nano Lett. 2, 1009 (2002). 40. R. L. McCreery, “Raman Spectroscopy for Chemical Analysis.” Wiley Interscience, New York, 2000.

65 41. (a) A. M. Rao, E. Richter, S. Bandow, B. Chase, P. C. Eklund, K. A. Williams, S. Fang, K. R. Subbaswamy, M. Menon, A. Thess, R. E. Smalley, G. Dresselhaus, and M. S. Dresselhaus, Science 275, 187 (1997). (b) P. H. Tan, Y. Tang, Y. M. Deng, F. Li, Y. L. Wei, and H. M. Cheng, Appl. Phys. Lett. 75, 1524 (1999). 42. V. W. Brar, Ge. G. Samsonidze, M. S. Dresselhaus, G. Dresselhaus, R. Saito, A. K. Swan, M. S. Unlu, B. B. Goldberg, A. G. Souza Filho, and A. Jorio, Phys. Rev. B 66, 155418 (2002). 43. S. Farhat, M. Lamy de La Chapelle, A. Loiseau, C. D. Scott, S. Lefrant, C. Journet, and P. Bernier, J. Chem. Phys. 115, 6277 (2001). 44. Bandow, A. M. Rao, K. A. Williams, A. Thess, R. E. Smalley, and P. C. Eklund, J. Phys. Chem. B 101, 8839 (1997). 45. P. H. Tan, Y. Tang, Y. M. Deng, F. Li, Y. L. Wei, and H. M. Cheng, Appl. Phys. Lett. 75, 1524 (1999). 46. D. Chattopadhyay, I. Galeska, and F. Papadimitrakopoulos, J. Am. Chem. Soc. 125, 3371 (2003). 47. D. A. Walters, M. J. Casavant, X. C. Qin, C. B. Huffman, P. J. Boul, L. M. Ericson, E. H. Haroz, M. J. O’Connell, K. Smith, D. T. Colbert, and R. E. Smalley, Chem. Phys Lett. 338, 14 (2001). 48. S. Reich and C. Thomsen, Phys. Rev. Lett. 85, 3544 (2000). 49. V. G. Hadjiev, M. N. Iliev, S. Arepalli, P. Nikolaev, and B. S. Files, Appl. Phys. Lett. 78, 3193 (2001). 50. V. A. Karachevtsev, A. Yu. Glamazda, U. Dettlaff-Weglikowska, V. S. Kurnosov, E. D. Obraztsova, A. V. Peschanskii, V. V. Eremenko, and S. Roth, Carbon 41, 1567 (2003); M. Iliev, A. P. Litvinchuk, S. Arepalli, P. Nikolaev, and C. D. Scott, Chem. Phys. Lett. 316, 217 (2000). 51. S. Hufner, “Photoelectron Spectroscopy,” 3rd Ed. Springer-Verlag, Heidelberg, 2003. 52. J. E. Herrera, L. Balzano, A. Borgna, W. E. Alvarez, and D. E. Resasco, J. Catal. 204, 129 (2001). 53. T. Hertel, R. Fasel, and G. Moos, Appl. Phys. A 75, (2002); M. Kociak, L. Henrard, O. Stephan, K. Suenaga, and C. Colliex, Phys Rev. B 61, 13936 (2000). 54. G. S. Duesberg, J. Muster, V. Kristic, M. Burghard, and S. Roth, Appl. Phys. A 67, 117 (1998); S. Niyogi, H. Hu, M. A. Hamon, P. Bhowmik, B. Zhao, S. M. Rozenzhak, J. Chen, M. E. Itkis, M. S. Meier, and R. C. Haddon, J. Am. Chem. Soc. 123, 733 (2001). 55. C. Huffman, Ph.D. Thesis, Rice University, 2003. 56. S. Ramesh, B. Brinson, M. Pontier Johnson, Z. Gu, R. K. Saini, P. A. Willis, T. Marriott, W. E. Billups, J. L Margrave, R. H. Hauge, and R. E. Smalley, J. Phys. Chem. B 107, 1360 (2003). 57. O. Gorelik, P. Nikolaev, and S. Arepalli, NASA Contractor Report, NASA/CR-2000-208926 (2001). 58. P. Petit, E. Jouguelet, J. E. Fischer, A. G. Rinzler, and R. E. Smalley, Phys. Rev. B 56, 9275 (1997). 59. A. DeMartino, R. Egger, K. Hallberg, and C. A. Balseiro, Phys. Rev. Lett. 88, 206402 (2002). 60. A. Kleinhammes, C. Bower, O. Zhou, and Y. Wu, BAPS MAR 98, I19.05 (1998). 61. X. P. Tang, A. Kleinhammes, H. Shimoda, L. Fleming, K. Y. Bennoune, S. Sinha, C. Bower, O. Zhou, and Y. Wu, Science 288, 492 (2000). 62. S. Latil, L. Henrard, C. G. Bac, P. Bernier, and A. Rubio, Phys. Rev. Lett. 86, 3160 (2001). 63. A. G. Rinzler, J. Liu, H. Dai, P. Nikolaev, C. B. Huffman, F. J. Rodriguez-Macias, P. J. Boul, A. H. Lu, D. Heymann, D. T. Colbert, R. S. Lee, J. E. Fischer, A. M. Rao, P. C. Eklund, and R. E. Smalley, Appl. Phys. A 67, 29 (1998). 64. P. Nikolaev, O. Gorelik, and S. Arepalli, TGA Measurement uncertainities for SWCNT sample evaluation. Private Communication, 2003. 65. J. L. Bahr, J. P. Yang, D. V. Kosynkin, M. J. Bronikowski, R. E. Smalley, and J. M. Tour, J. Amer. Chem. Soc. 123, 6536 (2001).

66 66. W. E. Alvarez, B. Kitiyanan, A. Borgna, and D. E. Resasco, Carbon 39, 547 (2001). 67. B. McMorrow, R. Chartoff, and D. Klosterman, in “SAMPE 2003 Proceedings,” Long Beach, CA, May 2003. 68. A. Thess, R. Lee, P. Nikolaev, D. Hongjie, P. Petit, J. Robert, C. Xu, Y. H. Lee, S. G. Kim, A. G. Rinzler, D. T. Colbert, G. E. Scuseria, D. Tománek, J. E. Fischer, and R. E. Smalley, Science 273, 483 (1996). 69. S. Rols, A. Righi, L. Alvarez, E. Anglaret, R. Almairac, C. Journet, P. Bernier, J. L. Sauvajol, A. M. Benito, W. K. Maser, E. Munoz, M. T. Martinez, G. F. de la Fuente, A. Girard, and J. C. Ameline, Eur. Phys. J. B 18, 201 (2000). 70. J. Hone, B. Batlogg, Z. Benes, A. T. Johnson, and J. E. Fischer, Science 289, 1730 (2000). 71. C. Park, Z. Ounaies, K. A.Watson, R. E. Crooks, J. Smith, Jr., S. E. Lowther, J. W. Connell, E. J. Siochi, J. S. Harrison, and T. L. St. Clair, Chem. Phys. Lett. 364, 303 (2002). 72. B. C. Lippens and J. H. deBoer, J. Catal. 4, 319 (1965); J. P. M. Syvitski, “Principles, Methods, and Application of Particle Size Analysis.” Cambridge Univ. Press, Cambridge, UK, 1991.

Analytical Characterization of SWCNTs 73. Y. Ye, C. C. Ahn, C. Witham, B. Fultz, J. Liu, A. G. Rinzler, D. Colbert, K. A. Smith, and R. E. Smalley, Appl. Phys. Lett. 74, 2307 (1999). 74. C. C. Ahn, J. J. Vajo, B. Fultz, R. Yazami, D. W. Brown, and R. C. Brown, DOE Hydrogen Program Review, NREL/CP-61032405, 2002. 75. K. Pradhan, A. Harutyunyan, D. Stojkovic, P. Zhang, M. W.Cole, V. Crespi, H. Goto, J. Fujiwara, and P. C. Eklund, Mater. Res. Soc. Symp. Proc. 706 (2002). 76. P. Ajayan and O. Zhou, “Carbon Nanotubes: Synthesis, Structure, Properties and Applications” (M. S. Dresselhaus, G. Dresselhaus, and Ph. Avouris, Eds.), Topics in Applied Physics, Vol. 80. SpringerVerlag, Berlin, 2001. 77. M. Cinke, J. Li, B. Chen, A. Cassell, L. Delzeit, J. Han, and M. Meyyappan, Chem. Phys. Lett. 365, 69 (2002). 78. K. H. An, K. K. Jeon, J. K. Heo, S. C. Lim, D. J. Bae, and Y. H. Lee, J. Electrochem. Soc. 149, A1058 (2002). 79. H. Darmstadt, C. Roy, S. Kaliaguine, and H. Cormier, Rubber Chem. Technol. 70, 759 (1997). 80. R. Krishnamoorty, Univeristy of Houston, private communication.

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Analytical Ultracentrifugation of Nanoparticles Helmut Cölfen Max-Planck-Institute of Colloids and Interfaces, Golm, Germany

CONTENTS 1. Introduction 2. Instrumentation 3. Basic Experiments 4. Analysis of Nanoparticles 5. Current Trends in Analytical Ultracentrifugation 6. Availability of Free Evaluation Software 7. Conclusion Glossary References

1. INTRODUCTION Analytical ultracentrifugation (AUC) is a powerful fractionating technique for polymer and particle characterization and has played a significant part in the understanding of colloidal but especially macromolecular systems starting with the pioneering work of Svedberg and co-workers [1, 2] who initially invented this technique for the characterization of particle sizes [1, 3]. However, it soon turned out that the technique was also well suited to the study of macromolecules, especially biopolymers, so that interest shifted almost exclusively to the study of macromolecules. One important result of the early work was the proof that macromolecules truly exist and that they are not aggregates of small molecules as was heavily debated at the time. Since that time, analytical ultracentrifugation has gained importance and become one of the most important characterization techniques for polymers mainly in biophysics/biochemistry. By 1970, there were about 1500 analytical ultracentrifuges in operation throughout the world. However, evaluation of the experimental results was tedious and time consuming in many cases because the results had to be derived from photographic records. With the advent of faster, cheaper, and more convenient techniques such as light scattering (LS), the importance

ISBN: 1-58883-057-8/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

of the analytical ultracentrifuge decreased so that, by the end of the 1980s, only a few old instruments were still in use. The use of this technique for nanoparticle characterization appeared forgotten and only a few specialized, mainly industrial laboratories, still applied AUC for particle characterization. However, in the beginning of the 1990s, Beckman Instruments released a new analytical ultracentrifuge—the Optima XL-A—equipped with a scanning absorption optical system [4]. This modern and fully computer controlled instrument led to a resurrection of analytical ultracentrifugation as the data acquisition was much simplified so that, nowadays, modern computer technology and powerful software can be applied for fast and rigorous evaluation (e.g., fitting experimental data to a model, handling of multiple data sets, etc.). Analytical ultracentrifugation is an absolute method requiring no standards and covering a broad application range of molar masses between 200 and 1014 g/mol and particle sizes between less than 1 and 5000 nm. The power of the technique lies in the fractionation of the sample into its components either according to their molar masses/particle sizes or to their structure/density without the need for any stationary phase as required in many chromatographic methods. The fractionation enables the measurement of distributions of molar masses, particle sizes, and densities, and different types of basic experiments with the same instrument can yield complementary physicochemical information. Also, sedimentation equilibrium analysis is based on solid thermodynamics. Often, analytical ultracentrifugation is the only applicable technique if complex mixtures are to be investigated. For example, a unique feature of the technique is that interactions between molecules and/or particles can be investigated over a wide range of concentrations without perturbing the chemical equilibrium [5]. This is important for the study of biologically relevant relatively weak interactions (K ≈ 10–100 M−1 ) but can, in analogy, be adapted to any kind of interaction such as macromolecule–particle. The range of samples that can be investigated by analytical ultracentrifugation is extremely wide and includes all systems that consist of a solvent and a dispersed or dissolved

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 1: Pages (67–88)

68

Analytical Ultracentrifugation of Nanoparticles

substance spanning from gels to microgels and dispersions and emulsions to solutions. Also, there are no limitations on the choice of solvent even at extreme pH. This results in a whole variety of applications, and although the technique has already existed for about 80 years, there are still new methods and applications coming up. This review will give the basics of the technique but will also discuss the modern applications and recent trends of the last few years. Here, the methods that have been developed for polymer analysis but are, in principle, adaptable to nanoparticle analysis are also mentioned to give the reader the full array of available experimental or evaluation methods.

2. INSTRUMENTATION An analytical ultracentrifuge is an ultracentrifuge with one or several optical detection systems that allow for the observation of a sedimenting sample in a centrifugal field of up to 290,000-fold the gravitational force (see Fig. 1). The sample is placed in a single- or double-sector cell (see Fig. 1) where the sector shape prevents convection during sedimentation due to the radially sedimenting sample. In addition to the two optical systems that modern machines are equipped with (see Fig. 1) and that detect the radially changing concentration in the cell, the Schlieren optical system, which detects the refractive index/concentration gradient, can also be applied. It delivers the first derivative of the radial concentration gradient and is thus well suited to watch sedimenting boundaries during sedimentation velocity experiments. The typical Schlieren peak is probably the most well known experimental output of an analytical ultracentrifuge. All three optical systems have their special advantages. The UV-absorption optics combines sensitivity with selectivity due to the variable detection wavelength, whereas the Rayleigh interference optics yields accurate experimental data due to the acquisition of a number of interference fringes, which are then evaluated via a fast Fourier transformation. However, the interference optics can only determine relative concentration changes with respect to a fixed point, which is usually the meniscus air/solution. The Schlieren Reference Top view Sample

2) 3)

Upper Plateau Laser diode, light source Slits

Incident light detector Reflector

1)

Absorbance

Torroidal diffraction grating

optics is well suited for high concentration or density gradient work and all kinds of experiments where a derivative of the concentration gradient is needed for the evaluation (e.g., the determination of the z-average molar mass from sedimentation equilibrium). The Schlieren optics is similar to the setup of the Rayleigh interferometer in Figure 1 but has a phase plate or knife edge in the focus of the condenser lens as an additional element. The Schlieren optics was widely considered to be the least sensitive of the three detection systems, but it was shown that the sensitivity equals that of the Rayleigh interferometer [7]. An ultrasensitive Schlieren optical system has also been described [8]. However, despite the equipment of many ultracentrifuges with video cameras to save the photographic evaluation of the experimental traces, fully automated evaluation of the images was very difficult and only reported in one case [9] until the recent introduction of an on-line Schlieren optical system [10]. The typical optical patterns derived by the three detection systems are shown in Figure 2. For an extensive description of these optical systems, see [11]. For modern ultracentrifuges, only the Rayleigh interferometer (as an on-line detector) and the absorption optics are still available although the Schlieren optics was recently adapted for use in the Beckman Optima XL/XL-I [12]. Other detection systems include a fluorescence detector [13] and a turbidity detector specially designed for particle size analysis mainly of latexes [14–17]. The fluorescence optical system is extremely sensitive and allows for the selective investigation of components with concentrations as low as 10 ng/mL even in mixtures with a much larger amount of other components. A prototype

Condenser Lens

Rotor Rotor Mirror

Sample Rotor

Imaging system for radial scanning Aperture Xenon Flash lamp

0 Camera lens

Boundary region Lower Plateau Baseline

Camera Solvent meniscus

Cylinder lens

Radius

Slit Photomultiplier tube Mirror

Sample meniscus

Reflective prism

Figure 1. Optical detection systems of the Optima XL-I analytical ultracentrifuge. Left: UV/vis absorption optics. Right: Rayleigh interference optics based on the construction of Laue [6]. Reprinted with permission from G. Ralston, Introduction to Analytical Ultracentrifugation, Beckman Instruments, Fullerton, CA. (1993). © 1993, Beckman Coulter.

Figure 2. Photographically recorded optical data obtained from the different detection optics of an analytical ultracentrifuge for the same sample. (1) Schlieren optics, (2) Rayleigh interference optics, and (3) absorption optics. The lower diagram presents the output for scanning absorption optics. Reprinted with permission from [19], H. K. Schachman, Ultracentrifugation in Biochemistry, 1959, Academic Press, © 1959, Academic Press and with permission from G. Ralston, Introduction to Analytical Ultracentrifugation 1993, Beckman Instruments, © 1993, Beckman Coulter.

69

Analytical Ultracentrifugation of Nanoparticles

fluorescence detector for the Optima XL-I ultracentrifuge has recently been constructed [18] so that now a third detector is commercially available, which can be simultaneously used in a modern analytical ultracentrifuge. The turbidity optics, on the other hand, proves its virtues for the analysis of particle size distributions, even if they span the entire colloidal range; as with the appropriate correction of the particle concentration for MIE scattering, even weakly scattering small particles can be precisely characterized in the presence of strongly scattering large particles. It is generally advantageous to combine several optical systems. In particular, the combination of the Rayleigh interference optics and the ultraviolet/visible (UV/vis) absorption optics can yield important information about complex systems where, for example, an absorbing component is selectively detected with the absorption optics, whereas the Rayleigh interferometer detects all components. For examples, see [20] or Figure 18.

3. BASIC EXPERIMENTS There are four basic types of experiments that can be performed with an analytical ultracentrifuge. Each of them can deliver its own range of physicochemical information on the sample. The different experimental approaches will be treated in the following section. Table 1 outlines the characteristics of each experiment type and the most important accessible parameters.

3.1. Sedimentation Velocity Experiment A sedimentation velocity experiment is carried out at high centrifugal fields and is the most important AUC technique for nanoparticle characterization. Here, the molecules/ particles sediment according to their mass/size, density, and shape without significant back diffusion according to the generated concentration gradient. Under such conditions, a separation of mixture components takes place and one can detect a steplike concentration profile in the ultracentrifuge cell usually exhibiting an upper and a lower plateau (see Fig. 2, lower trace). Each step corresponds to one species. If one detects the radial concentration gradient in certain time intervals, the sedimentation of the

molecules/particles can be monitored. This is demonstrated in Figure 3a. From the velocity of the sedimenting boundary, one can determine the sedimentation coefficient s according to s=

lnr/rm  2 t

(1)

where r is the position of the midpoint or second-moment point of the moving boundary, rm the radial distance of the meniscus, t the time, and  the angular velocity of the rotor. The sedimentation coefficient is a concentration- and pressure-dependent quantity that can be taken into account by appropriate correction or by extrapolation of a concentration series to zero concentration. A plot of ln(r/rm  vs. 2 t is a line in which the slope s is equal to the sedimentation coefficient (see Fig. 3b). The sedimentation coefficient is measured in svedbergs (S) where 1 S = 10−13 s. If the diffusion coefficient D is known from other experiments (light scattering, ultracentrifugation, etc.), one can calculate the molar mass of the sample according to the Svedberg equation: M=

sRT ¯ D1 − v

(2)

where M is the molar mass of the sample, R the gas constant, T the thermodynamic temperature, v¯ the partial specific volume, and  the solvent density. The partial specific volume is a critical value, but it can be determined with good precision in a density oscillation tube if enough sample material is available. For hybrid particles in mixtures with different compositions, v¯ cannot be measured as only the average over all particles in the mixture is obtained. Here, the workaround of reasonable assumptions must be applied or quantitative density gradient experiments can yield the average density from the density distribution. In general, sedimentation velocity experiments offer a good possibility for the rapid determination of particle size distributions and molar mass but also of equilibrium constants of interacting systems, which is especially advantageous for unstable systems that cannot be subjected to sedimentation equilibrium experiments due to possible sample degradation during the experiment [22].

Table 1. Basic experiment types in analytical ultracentrifugation and their characteristics and basic accessible parameters.

Experiment

Operative term in the Lamm equation (8)

Characteristics of experiment

Main accessible physicochemical parameter

Sedimentation velocity (3.1)

Sedimentation term prevails over diffusion term

High rotational speed

Sedimentation coefficient

Synthetic boundary experiment (3.4)

Only diffusion term operative

Low rotational speed, synthetic boundary cell

Diffusion coefficient

Sedimentation equilibrium (3.2)

Both terms operative, equilibrium between sedimentation and diffusion

Low to moderate rotational speed

Molar mass, equilibrium constants, and stoichiometries of interacting systems

Density gradient (3.3)

Both terms operative, equilibrium between sedimentation and diffusion

Moderate to high rotational speed, establishment of a radial solvent density gradient

Density

70

Analytical Ultracentrifugation of Nanoparticles

spreading is only attributed to the diffusion spreading for polydisperse samples, the determined diffusion coefficient can be found to be significantly too high. Whereas polydispersity is often not an issue for biopolymers, it can be serious in nanoparticle analysis in that it can lead to erroneous results if the diffusion coefficient is determined by this method. Therefore, in nanoparticle analysis, the diffusion coefficient should be determined by other methods such as dynamic light scattering or flow–field–flow fractionation.

a) 1.2

0.8 0.6 0.4 0.2 0.0

3.1.2. van Holde–Weischet Method 6.2

6.4

6.6

6.8

7.0

7.2

Radius (cm) 1.92

s = 4.77 S

ln r / rm

1.90

1.88

1.86

1.84 8.00E+010

1.20E+011

1.60E+011

2.00E+011

ω2t c)

0.30

Monomer s = 4.58 S

0.25

g(s*)

0.20 0.15 0.10 0.05

3.1.3. Time Derivative Method

0.00 0

2

4

6

8

10

s*(Svedberg) Figure 3. Sedimentation velocity experiment on bovine serum albumin in 0.5 N NaCl at 50,000 rpm and 20  C. (a) Experimental raw data acquired with scanning absorption optics at 280 nm. Scan interval 10 min. Radius means the radial distance to the center of rotation. (b) Sedimentation coefficient calculated from (1). (c) sedimentation coefficient distribution from the time derivative method (3) and [21].

3.1.1. Moving-Boundary Method Many procedures for the determination of s from sedimentation velocity data are available which cannot all be named here. Some of them like the moving-boundary method [23] allow an evaluation of the diffusion coefficient from the spreading of the boundary during the proceeding experiment [24] and thus a calculation of M. However, the calculation of D from the boundary spreading suffers from the requirement of reasonably monodisperse or at least reasonably separated samples because polydispersity leads to a boundary spreading proportional to time whereas diffusion goes only with the square root of time. If the whole boundary

In many cases, particles are polydisperse or one detects a multimodal distribution. In such cases, it is of interest to determine the sedimentation coefficient distribution Gs or the differential form thereof gs. Although this is, in principle, possible by the van Holde–Weischet method, a 150

100

one scan

b)

A problem arising from the diffusion broadening of boundaries is that the boundaries of several components can be smeared so that the overall boundary appears to be that of a single component. An approach to removing the effects of boundary spreading by diffusion and thus enabling calculation of the integral diffusion-corrected s distribution Gs was introduced by van Holde and Weischet [25]. This is done by selecting a fixed number of data points from one experimental scan that are evenly spaced between the baseline and the plateau. Then, an apparent sedimentation coefficient s ∗ is calculated for each of the data points and plotted versus the inverse root of the run time, yielding the typical van Holde–Weischet plot (see Fig. 4). If a linear fit of the corresponding s ∗ (one slice) is performed, the integral diffusion-corrected sedimentation coefficient distribution Gs can be obtained from the y values at infinite time in the van Holde–Weischet plot. In the case of a single monodisperse component, the lines intersect in one point (see Fig. 4). For multiple components, the corresponding number of intersects is obtained whereas the intersection point is shifted to times less than infinity in the case of nonideality. Therefore, the van Holde–Weischet analysis is a rigorous test for sample homogeneity or nonideality [25–31].

s* [S]

Absorbance

1.0

50 0

one

slice

t-0.5 [s -0.5]

Figure 4. Typical van Holde–Weischet plot of a sedimentation velocity experiment with a monodisperse system. Reprinted with permission from [26], K. Schilling, Ph.D. thesis, Potsdam (1999).

71

Analytical Ultracentrifugation of Nanoparticles

better suited method for the determination of gs is the time derivative method [21, 32, 33], which determines the time derivative of the radial scans acquired at different times according to


cr t/c0  t




2 t 2 lnrm /r




r rm

2

(3)

where gs ∗  equals the true distribution gs in cases where diffusion can be ignored. An example for bovine serum albumin is given in Figure 3c. An important advantage of this procedure is significant improvement in the signal-to-noise ratio of the experimental data because two scans are subtracted from each other so that systematic errors in the optical patterns cancel out and the random noise decreases. This approach takes advantage of the rapid data acquisition of modern analytical ultracentrifuges (Rayleigh interference optics) where 100 or more scans per velocity experiment are no longer an experimental problem. Hence, even scans for very diluted solutions where the sedimenting boundary can hardly be seen in the raw scans can be evaluated with the time derivative method. By that means, concentrations as low as 10 g/mL can be investigated so that interacting macromolecules and particles can be addressed in a concentration range previously not accessible with the analytical ultracentrifuge. However, a drawback of the time derivative method is that only scans from a relatively narrow time interval can be used for a single evaluation so that, in fact, no full advantage is taken of the possibility to scan several hundreds of experimental scans throughout an experiment. If diffusion is significant, extrapolation of the gs ∗  curves calculated for different times to infinite time yields the true distribution. The diffusion coefficient can also be derived from the gs ∗  distribution using the maximum of the gs ∗  curve [21]: ∗

gs max =




s 2D

1/2

3




 rm  √

t 1 − e−22 st



(4)

√ and thus a plot of gs max vs. t/ 1 − e−22 st yields a line with a slope proportional to the square root of s/D so that the Svedberg equation can be applied to derive the molar mass ∗

M=

2RT 2 (slope) 6r 2 ¯ 1 − v m

2 trm 2 2t

J =

2RT t ¯ 2 trm 2 1 − vM

is the collection of all known terms in the analysis. As M has to be known, J can be calculated so that only A and s can vary in (7). When  in the Gaussian distribution function is replaced with the value of the standard error of the determination of gs ∗ max , a sort of mass spectrum is obtained where broad sedimentation coefficient distributions are replaced by sharp error-related distributions. This gives a very sensitive analysis for marginal differences in s even in the second digit for different components (see Fig. 5). The determination of gs ∗  can yield a lot of important information besides the sample homogeneity and number of components. In the case of interacting systems, for example, interaction constants can be derived [35–37]. If multiple Gauss curves with a maximum at fixed s are fitted to gs, then each of them corresponds to one component [38]. By that means, the aggregation state as well as the corresponding concentration of the different aggregates can be determined for aggregating systems so that the equilibrium constant and thus the free enthalpy of the association steps is accessible. An example of the advantageous application of this technique was reported for the precrystallization aggregation of lysozyme, which yielded the smallest oligomer capable of forming a crystal [39].

3.1.4. Fitting to Approximate or Finite-Element Solutions of the Lamm Equation One new and recently much investigated approach has proved useful for the determination of s and D and thus M as well as the concentrations of individual components from sedimentation velocity data as in Figure 3a. This is the fit 50 mM KCl 150 mM KCl

50

(5)

However, for heterogeneous systems, the molar mass can be underestimated by 10–20%. Another way of determining D from gs ∗  is to look at the standard deviation  of a Gaussian distribution fitted to gs ∗ : D=

where A is the amount of the considered species and

(6)

If a solution is analyzed where just different conformations of the same molecule or different particle shapes with the same molar mass or small amounts of aggregates are present, M and thus s/D can be fixed in the Svedberg equation for each component [Eq. (2)] to give the basis for a

CON-SPEC (s20,w)

gs ∗ t =

very sensitive analysis—the so-called conformational spectra (CON-SPEC) [34]. Combining (2) and (6) with the equation for the Gaussian distribution function yields
 −s ∗ − gs ∗ max 2 A gs ∗  =  exp (7) 2gs ∗ max J 2 2 gs ∗ max J

40 30 20 10 0 6.45

6.48

6.50

6.53

6.55

6.58

6.60

s20,w (Svedbergs) Figure 5. CON-SPEC (s20 w  for -actinin showing a modest conformational change upon an increase in salt concentration from 50 to 150 mM KCl. That the change is only marginally significant can be seen from the slight overlap of the two functions. Reprinted with permission from [34], N. Errington, et al., Biophys. Chem. 80, 189 (1999). © 1999,

72

3.2. Sedimentation Equilibrium Experiment In contrast to the sedimentation velocity run, a sedimentation equilibrium experiment is carried out at moderate centrifugal fields. Here, the sedimentation of the sample

a) 1.5 Sedimentation

Absorbance

As the Lamm equation (8) is the fundamental equation in analytical ultracentrifugation capable of describing all types of ultracentrifuge experiments, fitting of experimental data to this equation is a potentially very powerful approach. However, a drawback is that this method is clearly model dependent. Nevertheless, it is widely applicable and can quite accurately determine the sedimentation and diffusion coefficients even in mixtures of up to three components as long as the proper fitting function is used, the individual components are reasonably monodisperse (see the discussion about boundary broadening), and the individual components have differences in s of a factor of 1.5 or more. For smaller s differences, the van Holde–Weischet method is better suited. In the case of unknown samples, the van Holde–Weischet method can be used first to determine the type of system under investigation so that the correct model can be used for fitting the Lamm equation. If mixtures of more than three components have to be analyzed, the s and D values of some components have to be known in order to get reliable fitting results. The fitting of experimental data to solutions of the Lamm equation has its special merits for small macromolecules or nanoparticles that sediment so slowly (s < 1 S) that they do not clear the meniscus. Furthermore, such approaches permit the rapid determination of molar masses with an accuracy of 10% within 15–30 min after the start of a sedimentation equilibrium experiment using a modified Archibald approach [50]. This is especially important for unstable samples that have to be characterized rapidly. Very recently, the requirement of monodisperse samples for fitting of the Lamm equation was overcome so that now even the sedimentation coefficients and molar mass distributions of polydisperse samples can be investigated [51]. However, this approach suffers from the necessary prior knowledge of v¯ (no serious problem) and the frictional ratio f /f0 of the sample (f is the frictional coefficient and f0 is the frictional coefficient of the spherical particle having the same mass and v¯ as the sample under consideration). If the frictional ratio is not known, it can also be fitted, which, in turn, allows conclusions about the particle shape. The most significant advantage of this approach for nanoparticle analysis is the possibility to correct for the effects of diffusion on the broadening of the sedimenting boundary so that diffusion-corrected sedimentation coefficient distributions can be obtained reflecting the true sample polydispersity. The method was recently shown to yield reliable results for different model systems [52].

is balanced by a back diffusion according to Fick’s law caused by the established concentration gradient. After the equilibrium between these two transport processes is achieved, an exponential concentration gradient has formed in the ultracentrifuge cell (see Fig. 6a, which represents a true thermodynamic equilibrium). Therefore, the sedimentation equilibrium analysis is based on solid thermodynamics. The time of equilibrium attainment considerably depends on the column height of the solution [53] so that short column techniques with solution columns of about 1 mm are common practice nowadays where even a rapid equilibrium within 1–2 h or less is desired [54]. The concentration gradient contains information about the molar mass of the sample, the second osmotic virial coefficient, or the interaction constants in the case of interacting systems independently of the particle shape. An advantage is that the detection of the concentration gradient is possible without disturbing the chemical equilibrium even of weak interactions. Sedimentation equilibrium is of more value for polymeric systems, because the molar mass of particles can already be too high to allow a significant diffusion counteracting the particle sedimentation. Nevertheless, for smaller down to the smallest nanoparticles, or even subcritical clusters and complexes, it can also be applied successfully as shown for the example of complexes of Zr(IV) and Hf(IV) formed by hydrolytic polymerization in an acidic medium [55, 56]. Various procedures exist for the evaluation of

1.0

Diffusion

0.5

0.0 6.70 6.75 6.80 6.85 6.90 6.95 7.00 7.05 7.10 7.15

Radius [cm] b) 0.5 0.0 -0.5

ln A

of a series of radial concentration profiles to approximate solutions of the Lamm [40] differential equation of the ultracentrifuge [41–46] or finite-element solutions of the Lamm equation [47–50]:
2  
 c 1 c c c 2 + =D + 2c − s r (8) t r 2 r r r       Diffusion term Sedimentation term

Analytical Ultracentrifugation of Nanoparticles

-1.0 -1.5 -2.0 45

46

47

48

Radius2

[cm2]

49

50

51

Figure 6. Sedimentation equilibrium experiment on bovine serum albumin in 0.5 N NaCl at 20  C and 10,000 rpm. (a) Experimental raw data acquired with scanning absorption optics. (b) Plot of ln(absorbance A) vs. r 2 to evaluate Mw app from the slope.

73

Analytical Ultracentrifugation of Nanoparticles

sedimentation equilibrium experiments which are discussed in the following.

3.2.1. Classical Approaches The classical approach to the evaluation of sedimentation equilibrium concentration gradients is to plot ln c vs. r 2 to obtain the weighted-average molar mass Mw from the slope according to the following equation (see Fig. 6b): Mw app =

dln c 2RT 2 ¯ 1 − v dr 2

(9)

It must be noted that these Mw values have apparent character because they are calculated for finite sample concentrations. Consequently, they are denoted Mw app . The true Mw can be obtained by an extrapolation of a concentration series to infinite dilution. The evaluation according to (9) works well for monodisperse ideal samples because, in such cases, the ln(c) vs. r 2 plot is linear. However, in many cases, this plot is curved, which yields erroneous Mw app values from linear regression. Nevertheless, the curvature indicates whether the sample is nonideal (downward curvature) or heterogeneous or self-associating (upward curvature). One advantage of (9) is that it allows for the computation of local Mw app r values for every acquired data point. 1/Mw app r can be plotted against the corresponding concentrations to yield Mw and the second osmotic virial coefficient A2 due to 1 1 = + 2A2 cr Mw app r Mw

(10)

From the radial concentration gradient (Fig. 6a) and the local Mw app r values, the apparent z-average molar mass Mz app r can also be obtained according to (11) if the data set is of good quality (double differentiation of the experimental raw data) [57]:  d ln Mw app r 2RT Mz app r = Mw app r + 2 ¯ 1 − v dr 2 
 1 dcr d 2RT ln (11) = 2 dr 2  ¯ 1 − v r dr An alternative evaluation procedure was described by Lansing and Kraemer [58]: Mw app =

cb − c m 2RT 2 c r 2 − r 2  ¯ 1 − v 0 b m

(12)

where the indices b and m denote bottom and meniscus, respectively. This procedure also allows for the calculation of Mz app [59, 60]: Mz app =

RT 1 2 c − c  ¯ 1 − v b m
  
1 dcr 1 dcr − × rb dr rm dr b m

(13)

Equations (12) and (13) yield average molar masses for the whole cell and no local values. A significant disadvantage of this evaluation is that it completely relies on the correct determination of the concentration at the meniscus and the

cell bottom where there might be considerable uncertainty. Furthermore, the obtained molar masses can be strongly influenced by impurities being much smaller (cm influenced) or larger (cb influenced) than the sample due to the evaluation of a whole-cell average molar mass.

3.2.2. Fitting Concentration Gradients to a Model An approach for the analysis of sedimentation equilibrium data greatly facilitated by cheap and commonly available computers is the fit of the radial equilibrium concentration gradient to a theoretical model using nonlinear least squares analysis [61, 62]. This approach is highly popular today, and a whole variety of programs is available for this purpose, commercial as well as public-domain software [61, 63]. Furthermore, such fitting routines can easily be set up on commercial software platforms, offering freely definable functions for the curve-fitting process. In general, the radial equilibrium concentration gradient in the ultracentrifuge is described as the exponential form of (9) for a single-ideal component, as a modified exponential equation in the case of nonideality [see Eq. (14)], or as a sum of exponentials for different components in a mixture or in an interacting system [see Eq. (15)]. For a nonideal onecomponent system, the radial concentration gradient cr can be described by
2

  2 ¯ r 2 − rref cr = crref  exp M1 − v 2RT  − A2 Mcr − crref  (14) where rref is the chosen reference radius.
2 

2 2   ¯ r −rref cr = cmon rref exp M 1− v 2RT mon
2 

  2 ¯ r 2 −rref n2 Mmon 1− v + cmon rref n2 K2 exp 2RT + ···+cmon rref ni
2 

  2 ¯ r 2 −rref ni Mmon 1− v × Ki exp 2RT

(15)

For an ideal self-associating system of the type n monomer → n-mer with K = cn-mer)/c(monomer)n , the first two exponentials in (15) can be used as a fitting function, whereas for a more complicated self-associating system with several n-mers, the sum of the exponentials in the form of (15) can be applied. By such an approach, the quantitative determination of the association constant(s), stoichiometries, and second virial coefficients for self-associating systems is possible [64–69]. One can think of many more models for a system that can be used to fit the experimentally detected radial concentration gradient. However, the case of a polydisperse and often highly nonideal system—colloid chemists are very often confronted with—is not yet covered by a model in the commonly available programs. Therefore, in this case, the models for a monodisperse system must be applied knowing

74

Analytical Ultracentrifugation of Nanoparticles

that the obtained values for the polydisperse system are only averages. Very often, a baseline absorbance that originates from nonsedimenting but light-absorbing material is used as a fitting parameter for data derived from absorption optics just by adding a variable to the right-hand side of (14) and (15). This can have an enormous influence on the results and extreme care has to be taken when allowing the baseline absorbance to be a fitting variable. Wherever possible, one should determine the baseline absorbance using the so-called overspeeding technique where the sample is sedimented to the cell base, leaving only the nonsedimenting material behind. If the Rayleigh interference optics is applied, which yields interference fringe shifts relative to the meniscus, it is recommended to apply synthetic boundary experiments in order to determine the fringe shifts at the meniscus for low-speed equilibrium experiments [70]. Another possibility is to use the so-called high-speed equilibrium technique, which applies such high speeds that the meniscus is cleared of the sample, leaving an exponential radial concentration gradient. Whereas this technique works well for relatively monodisperse samples, in highly polydisperse systems the high-molecular-weight material has sedimented to the cell bottom already and is thus lost for detection, resulting in a molar mass that is found to be too low. The advantage of fitting the radial equilibrium concentration gradient to a model is that the experimental data can directly be analyzed without differentiation, an operation that amplifies experimental noise. Furthermore, multiple data sets can be fitted to one model, allowing for an analysis with better statistical significance. On the other hand, a possible danger of every fitting process that the least squares fit just corresponds to a side minimum can become serious even if the fitted curve well describes the experimental data. The fact that sums of exponentials have to be used for the analysis of the equilibrium concentration gradient limits the analysis of multicomponent systems considerably to three or four components [71] because the fitting parameters of the exponentials are badly determined by the data. A further serious disadvantage is that a model has to be known prior to the fitting process; that is, one must already have information about the system from other sources. However, one does not often know anything about the sample under investigation. In such cases, several models can lead to answers of similar accuracy from the statistical viewpoint. This requires either other independent techniques to seek more information or applying model-independent approaches such as the M ∗ function (see below) or an evaluation using a classical approach to learn something about the system of interest prior to fitting.

3.2.3. Model-Independent Approaches There are two main model-independent approaches to evaluate sedimentation equilibrium data. The first one is the so called M ∗ function [72] defined as M ∗ r = %=

cr − cm r %cm r 2 − rm2  + 2% rm r&cr − cm ' dr 2 ¯ 1 − v 2RT

with (16)

The most useful property of the M ∗ function is that it equals Mw app at the cell bottom. This means that one can get a cell average molar mass by extrapolating M ∗ r to rb . The M ∗ function has been combined with the classical evaluation according to (9) in evaluation algorithms [73, 74] to yield the whole-cell average Mw app via M ∗ r as well as the local Mw app r via (9) to allow detailed insight into the kind of system under investigation besides the evaluation of Mw app . Afterwards, the selection of a model for fitting the radial concentration gradient should lead to realistic results in coincidence with those already derived for Mw app , allowing more detailed insight into the system under consideration (e.g., concentration of different components in a mixture, equilibrium constants of interacting systems, etc.). The second model-independent approach—the ( function [75, 76] or (in a recently derived form) ) function [77]—is particularly well suited for interacting systems. The ( function is a transformed type of the concentration distribution at sedimentation equilibrium:   cr 1 − v¯ i 2 2 (i r = exp Mi rref − r 2 crref  2RT =

ai rref cr ai rcrref 

(17)

where Mi is the molar mass of the smallest sedimenting species i (1 in the beginning of the evaluation). Mi usually has to be determined in a separate experiment using routines like MSTAR [74]. ai is the thermodynamic activity of the smallest sedimenting species i equal to the concentration in the ideal case. If (i r is plotted vs. cr, the intercept (or, realistically, the extrapolated intercept) equals the ratio of ai rref /crref , where crref  is known from the experimental data. The activity/concentration at the reference radius then allows for the calculation of the concentration gradient using the integrated form of (9) [see also (14) or (15) in the form for an ideal noninteracting onecomponent species]. Then, a revised radial concentration gradient &cr − ci r' can be calculated, and the analysis is repeated for the next smaller sedimenting component. This unraveling of the total concentration gradient into those of the individual components is illustrated in Figure 7. From the concentration gradients of the individual species, one can calculate the equilibrium constants. The elegance of this model-independent approach is that it avoids any differentiation/integration, the former being especially sensitive to the amplification of experimental noise. Furthermore, it allows for the calculation of activity distributions, giving a rigorous thermodynamic treatment of nonideal systems. However, one difficulty is still the curvature in the (r vs. cr plots, especially for low cr, which are important for the extrapolation to infinite dilution to give the correct intercept (see Fig. 7a). Therefore, the ( function was modified giving the ) function [77], which shows a much better linearity when plotted as cr/)i r (y axis) versus )i+1 /)i on the x axis and thus allows a safer determination of the intercept and ci rref , where )i r is defined as   1 − v¯ i 2 2 2 Mi r − rref (18) )i r = exp 2RT

75

Analytical Ultracentrifugation of Nanoparticles

3.3. Density Gradient Experiment

1.8

Ω1(r)

1.6 1.4 1.2 1.0 0.8 0.6 0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

A230 (r) b) 1.4 = Total concentration = ETF concentration = TMADH concentration = Complex concentration

1.2

Absorbance

1.0 0.8

Meniscus

0.6 0.4 0.2 0.0 6.375

6.425

6.475

6.525

6.575

6.625

Radius [cm] Figure 7. Interacting system between trimethylamine dehydrogenase (TMADH) and the corresponding electron transfer flavoprotein (ETF): (a) shows the ( function for ETF in relation to the concentration in absorbance units to demonstrate a typical example; (b) shows the individual concentration gradients of all components in the mixture which allow the calculation of interaction constants for each set of corresponding data points.

Then, the calculation of the concentration distribution of i can be performed using ci r = ci rref )i r and the analysis is repeated for the next smaller sedimenting species. This evaluation was recently implemented in a computer program [78]. Such analysis is especially useful if nothing is known about the kind of interaction taking place in a complicated mixture (homogeneous or heterogeneous interaction) and in which several components are involved [79]. If the interaction is found to be homogeneous, the total concentration gradient cr can be fitted to simple polynomials of c1 r of the form cr = c1 r + K2 &c1 r'2 + K3 &c1 r'3 + · · · cr = c1 r + Kn &c1 r'

n

The second principal possibility of separation in an analytical ultracentrifuge is the separation due to the chemical structure expressed in different solute densities in a density gradient. To generate a continuous density gradient in the ultracentrifuge cell, either a high-density salt (CsCl, etc.) or a substance like sucrose is dissolved in water or a mixture of two organic solvents with very different densities is applied. Under the action of the centrifugal field, the salt or the more dense solvent sediments toward the cell bottom and thus changes the density of the solution continuously toward the cell bottom. If a sample is placed into the density gradient, it will sediment/float to a position where its density matches that of the gradient. In the case of a mixture, this leads to a banding of the components due to their chemical structure/density. The range of densities that can be covered using density gradients is limited (0.8–2.0 g/mL) [14] but sufficient for the separation of purely polymeric substances. However, inorganic or organic–inorganic hybrid colloids have a density that is, in most cases, too high for the successful application of a density gradient. Also, the high salt concentrations in aqueous density gradients may be a problem for the analysis of electrostatically stabilized colloids. Furthermore, these so-called static density gradient experiments take a long time, usually on the order of several days (see Fig. 8), due to the slow banding of the sample. Also, potential solvent binding of the sample in a density gradient may be a problem. Nevertheless, density gradients are an excellent tool for the investigation of structural differences in mixtures and have especially proven to be an invaluable tool for the analysis of polymer latexes [14]. The separation capabilities of a static density gradient are demonstrated in Figure 8 for the separation of a mixture of an acrylonitrile–vinylacetate copolymer in a DMF/CHBr3

dn dr

Abs

a) 2.0

t1

a)

t2

t3

t4

r

or

for monomer → n-mer

1

(19)

This allows for the analysis of complicated self-associations with a general polynomial fit or of monomer → n-mer interactions using the second part of (19). The ( resp. ) function has several other useful features. One is a sensitive diagnostic test for the discrimination between polydispersity and self-association. If (r is plotted for different cell-loading concentrations but a common crref , the plots will be superimposed in the case of selfassociating systems, whereas they will not be superimposed in the case of a polydisperse or impure system. This test is more sensitive than the commonly applied plot of Mw app r vs. cr for different cell-loading concentrations [80].

b)

c)

3

2

Figure 8. Static analytical density gradient. (a) Banding of a sample in a density gradient observed with Schlieren optics (upper) or absorption/Rayleigh interference optics (lower) at different times. (b) Acrylonitrile-vinylacetate copolymer in a DMF/CHBr3 (136 g/L) density gradient after 35 h at 33,450 rpm. Fractions 1 and 2 are separated already. (c) After 108 h at 33,450 rpm (equilibrium). Figure adapted with permission from [81], R. Buchdahl et al., J. Polym. Sci., Part C 1, 143 (1963). © 1963, John Wiley.

76

Analytical Ultracentrifugation of Nanoparticles

a

b

4 3 1 2 5

five components are separated; and the separation is further improved within the next 12 min. Therefore, dynamic density gradients are the fastest technique to learn about the structural heterogeneity of mixtures as long as the particles have densities in the limited range for this type of density gradient.

3.4. Synthetic Boundary Experiment In a synthetic boundary experiment, changes of a boundary between solution and solvent with time are observed at low centrifugal fields where no sedimentation of the sample occurs. Such experiments require special cells where the solvent is layered upon the solution column under the action of a certain centrifugal field. This is achieved by capillaries that connect the solvent compartment with the solution compartment and that allow a flow at a certain hydrostatic pressure. At t = 0, a sharp boundary comparable

a)

b) t1

t2

boundary

Concentration

density gradient. The sample consists of three fractions that could be successfully separated: (1) linear polymer (0.5 × 106 g/mol,  = 1,0347 g/mL), (2) highly branched polymer (25 × 106 g/mol,  = 1,0365 g/mL), and (3) weakly cross-linked polymer (75 × 106 g/moL,  = 1,0371 g/mL). If one looks at the densities of the separated components, one sees that the resolution of the separation is in the fourth digit of the density. Much of the theory for the analysis of analytical density gradient experiments was established in the classical papers of Meselson, Hermans and Ende [82–85]. However, the Hermans–Ende equation was restricted to ideal systems, which can result in serious errors in the calculation of the radial density profile. Therefore, it was recently improved for real systems [86–88]. The basic equations that allow for the calculation of the density of the gradient at every point in the ultracentrifuge cell can be found in these references. From the width  of the Gaussian concentration profile of the sample, the molar mass can be calculated [82]. The accuracy is much smaller, however, than that of the molar mass derived by sedimentation equilibrium experiments so that it can realistically only serve as an estimate. The disadvantage of the prolonged experimental times for static density gradient experiments can be overcome by using the so-called dynamic density gradients [89, 90]. Here, a layer of H2 O is usually layered upon D2 O in a synthetic boundary experiment (see the corresponding section for details of this technique) establishing the fast formation of a dynamic H2 O/D2 O density gradient within a few minutes. Although the density range of this type of gradient is rather limited (1.0–1.1 g/mL), it is well suited for the fast characterization of latexes—especially polystyrene—and can be extended to 0.85–1.25 g/mL under certain circumstances [89]. An example is given in Figure 9 where an 11component latex mixture with different densities is subjected to an H2 O/D2 O density gradient. As five latexes have a density in the density range between H2 O and D2 O, the gradient selectively separates these components, whereas the other six components are not detected although their density is partly quite close to the solvent. This example illustrates how fast information on the structural composition of a mixture can be derived. After only 2 min of centrifugation, four of the five latexes are already separated; after 4 min, all

t3 t4

cp

Radius (cm) 60

c) 50

c

40 1 234 5

(

cp

2 30

)

(dc/dr)

20 10

Figure 9. Formation of a dynamic H2 O/D2 O density gradient with a mixture of 11 different ethyl hexylacrylate/methylacrylate (EHA/MA) copolymer lattices that have been polymerized separately (w[MA] = 0/10/20/30/40/50/60/70/80/90/100 wt%). All particles had approximately the same diameter of 200 nm and exhibit the following densities: 0.980/1.000/1.021/1.043/1.066/1.089/1.114/1.167/1.196/1.225/1.140 g/mL. Just five of these are in the density range between 0.997 (pure H2 O) and 1.095 g/mL (pure D2 O). Run conditions: 40,000 rpm, 25  C. (a) 2 min, (b) 4 min, and (c) 16 min. Reprinted with permission from [12], W. Mächtle, Prog. Colloid Polym. Sci. 113, 1 (1999) © 1999, SpringerVerlag.

0

0

2000

4000

6000

8000

10000

12000

Time (s) Figure 10. Synthetic boundary experiment. (a) Optical patterns from Rayleigh interference optics (upper) and Schlieren optics (lower) derived at time t. (b) Schematical radial concentration profiles at different times where t4 > t3 > t2 > t1 . (c) Plot of cP /dc/dr2 vs. time to derive the diffusion coefficient from the slope of the regression line. Reprinted with permission from G. Ralston, Introduction to Analytical Ultracentrifugation 1993, Beckman Instruments. © 1993, Beckman Coulter.

Analytical Ultracentrifugation of Nanoparticles

to the meniscus is obtained (see Fig. 10b). Within seconds, this boundary spreads by diffusion according to Fick’s law due to the high concentration gradient at the boundary. The spreading of the boundary with time is monitored (see Fig. 10b and c). Now the plateau concentration of the upper plateau cP as well as (dc/dr) at the boundary position can be determined. A plot of (cP /(dc/dr))2 vs. time yields a line with a slope equal to 4D. It is also possible to derive the dependence of D on the polymer concentration in a single synthetic boundary experiment [92]. A pseudo synthetic boundary experiment can be constructed from a normal sedimentation velocity experiment if Gs is obtained by the van Holde–Weischet method and subtracted from the experimental scans at various times [93]. This has the advantage that any layering imperfections that can occur in a conventional synthetic boundary experiment are avoided and, furthermore, samples that sediment at even the low speeds of a conventional synthetic boundary experiment can be investigated. However, compared to dynamic light scattering, a synthetic boundary experiment is more tedious and time consuming for the determination of the particle diffusion coefficient, so that, here, dynamic light scattering is the method of choice. Nevertheless, if aggregates are present in a sample, the light-scattering results can be seriously obstructed, whereas the synthetic boundary experiment will be unaffected as the aggregates will sediment even at the low applied speeds.

4. ANALYSIS OF NANOPARTICLES There are numerous examples of information derived by one of the four basic experiments or combinations thereof using an ultracentrifuge. They cannot be treated completely here, but the selected applications will show that analytical ultracentrifugation is a universal absolute technique for the characterization of polymers or colloids, especially in mixtures. The examples given here have been chosen in such a way that the range of information obtained by AUC is as broad as possible and the emphasis of the selected examples is more on the methods of how to derive information on a nanoparticle system by AUC rather than on how precisely this information can be obtained for a specific nanoparticle system. At the end of this section, a table is presented which will give the reader an overview of the primary literature for specific systems.

4.1. Particle Size Distributions The application of analytical ultracentrifugation for the determination of particle sizes and their distributions to address problems of colloid analysis was already realized by the pioneers of this technique because sedimentation velocity experiments provide a sensitive fractionation due to particle sizes/molar masses [3, 94–96]. Nevertheless, it appears that the potential of this application is still not yet commonly recognized. It is relatively straightforward to convert a sedimentation coefficient distribution, which can be calculated using (1) for every data point (a) ri (if a radial scan has been acquired at a specified time) or (b) ti (if a concentration detection at a specified radius has been performed in dependence of time), to a particle size distribution. Assuming the

77 validity of Stokes’ law (e.g., the sample is spherical), the following derivative of the Svedberg equation (2) is obtained

18.si di = (20) 2 −  where di is the particle diameter corresponding to si , 2 is the density of the sedimenting particle (including solvent/ polymer, etc., adhering to the sample), and . is the solvent viscosity. If the particles are not spherical, only the hydrodynamically equivalent diameter is obtained unless form factors are applied if the axial ratio of the particles is known from other sources such as transmission electron microscopy (TEM). The conversion of a sedimentation coefficient distribution to a particle size distribution highly relies on the knowledge of the density of the sedimenting particle. For hybrid particles or very small nanoparticles less than 5 nm, this issue can be a severe problem, especially in the case of mixtures, as the density of the particles is usually not known. Measurements of the average particle density in the mixture can lead to erroneous results so that in such cases the correlation of the sedimentation coefficient distribution with a distribution obtained from a density-insensitive method such as flow–field–flow fractionation or dynamic light scattering is meaningful. This can, in turn, yield the particle density, which can give information about the relative amount of materials building up the hybrid particle [97]. But even an apparent particle size distribution that is calculated within the limits of reasonable particle densities can already yield very valuable information [98]. However, in the case of industrially important latexes, the particle density is usually exactly known from the chemistry of particle formation/polymerization. Thus, the determination of particle size distributions with the analytical ultracentrifuge is a rapid technique providing a high statistical accuracy (e.g., every sedimenting particle is detected) in contrast to transmission electron microscopy (TEM), which delivers information about the particle shape but often suffers from drying artifacts. Determination of a particle size distribution from TEM images requires counting hundreds/thousands of particles. This problem has only partly been diminished by the advent of commercially available picture evaluation algorithms. A recent test dedicated to particle size analysis by the Bayer group in 17 laboratories, confirmed the view that TEM and/or analytical ultracentrifugation are the best techniques for the determination of particle size distributions [99] as discussed above. A combination of TEM, AUC, and X-ray diffraction techniques can provide a complete insight into a colloidal system [100]. Analytical ultracentrifugation in combination with electron microscopy in its various forms can be considered the most powerful characterization approach for particle size distributions and particle morphologies known to date. The following examples illustrate the fractionation power of the analytical ultracentrifuge for latexes [14, 101, 102] (see Fig. 11a) and especially for inorganic colloids [103]. In the case of latexes, the accessible size range is between 10 and 5000 nm with a baseline resolution for monodisperse components differing just by 10% in diameter [105], whereas for inorganic colloids, analytical ultracentrifugation separates dispersions with an almost atomar resolution. The

78

Analytical Ultracentrifugation of Nanoparticles coupling point 1.0

a) 1=i

given D[nm] % 67 10 113 10 166 10 246 10 318 10 356 10 486 10 680 10 840 10 1220 10

Σmi i=1

0.6 0.4 0.2 0

0

300

600

900

reproduced D[nm] % 72 14 9 121 172 11 8 259 320 10 8 379 515 9 665 8 870 10 1180 13

1200

Di

1500 nm

Integral Differential

1.68 nm 1.77 nm

Sum of mass fractions

1.30 nm 1.40 nm 1.50 nm 1.59 nm

b) 0.8 0.6 0.4

1.92 nm

1.0

0.83 nm

0.2 0.0 0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

Particle diameter [nm]

20

c)

10 5

300 250 200 150 100 50

tim

8 10

ion

smallest reported species that could be detected was a ZrO2 precursor just 0.4 nm in diameter [98], which shows the potential of the analytical ultracentrifuge for the analysis of the smallest nanoparticles or subcritical complexes. This means that particle size distributions derived by ultracentrifugation can well be used to investigate particle growth mechanisms [98, 103] from its initial stages/the critical crystal nucleus, especially if the growth is slow enough to detect time-dependent particle growth as in the example of ZnO (see Fig. 11c) [103, 104].

1.0

0.8

0.6

0.4 500 400 300 200

0.2 0.0

h]

Figure 11. Examples of particle size distributions determined with the analytical ultracentrifuge. (a) 10-component mixture of polystyrene standard latexes corrected for MIE scattering. Reprinted with permission from [17], W. Mächtle, “Analytical Ultracentrifugation in Biochemistry and Polymer Science,” p. 147, Royal Society of Chemistry, Cambridge, UK, 1992. © 1992, Royal Society of Chemistry. (b) Pt colloid during particle growth. Reprinted with permission from [103], H. Cölfen and T. Pauck, Colloid Polym. Sci. 275, 175 (1997). © 1997, Springer-Verlag. (c) Particle growth of ZnO. Reprinted with permission from [104], H. Cölfen et al., Prog. Colloid Polym. Sci. 107, 136 (1997). © 1997, Springer-Verlag.

10

20

30

s
[S]

40

100 50

60

70

tim e[

0

g(s
)

6

12 14 le dia mete r [nm 16 ]

act

2 4

Partic

Re

0

e[

mi

n]

Arbitrary units

15

For very small particles, diffusion becomes significant so that it must be corrected. This is possible by extrapolation to infinite time, but, recently, another method was suggested which is based on the calculation of the diffusion coefficient from the particle size [106]. From this, the diffusion broadening of the sedimenting boundary can be calculated and subtracted from the measured concentration distribution, giving the diffusion-corrected particle size distribution. This procedure works best for monomodal and relatively narrow distributions, whereas it becomes less accurate for broad and/or multimodal distributions. An alternative is the diffusion correction of the sedimentation coefficient distribution by fitting the experimental concentration profiles to approximate solutions of the Lamm equation [51]. Another substance class rapidly gaining importance is the hybrid colloids between polymers and inorganic matter. Here, analytical ultracentrifugation shows all its merits for the investigation of transformations, aggregation processes, and so on. For example, the encapsulation of a molybdenum cluster with a surfactant could be characterized as well as the aggregation of the primary clusters in different solvents with high resolution [107]. An example of monitoring a restructuring process of a complex hybrid colloid in solution is given in Figure 12. Here, calcium phosphate was synthesized within functional polymeric aggregates so that a very unusual neuron-like crystal morphology was obtained within the polymeric aggregates. This neuron-like morphology is slowly transformed into a more compact spherical structure (see photos from left to right), which could be monitored by analytical ultracentrifugation so that the relative proportions of each species could be determined [108]. Since the

0

Figure 12. Transformation of “neuron-like” calcium phosphate into a more compact spherical form and the time-resolved observation of this process by analytical ultracentrifugation. The ultracentrifuge data reproduced with permission from [108], M. Antotietti et al., Chem. Eur. J. 4, 2493 (1998). © 1998, Wiley-VCH.

79

Analytical Ultracentrifugation of Nanoparticles

particle density is not known, only the consideration of the sedimentation coefficient distributions makes sense. From the sedimentation coefficient distributions, the whisker structure is evident in the species with the lower sedimentation coefficient, whereas the dense spherical structure can be identified by its high sedimentation coefficients. The formation of the dense species has a clear maximum before its amount decreases again due to macroscopic precipitation. The observation of such transformation in solution enhances the visual information from TEM and shows the potential of ultracentrifuge experiments combined with a TEM investigation. A combined particle size and density gradient analysis is a powerful tool for the analysis of very complex mixtures where the particle sizes as well as the densities of the components are different. This is illustrated in Figure 13. Here, a mixture of four different polybutylacrylate (PBA) latexes was grafted with a styrene/acrylonitrile (SAN) copolymer which has a higher density than PBA [14]. In the density gradient, the four grafted particles are clearly resolved, indicating that the degree of grafting is different for every particle size. From the particle size distributions of the ungrafted and grafted PBA particles, it becomes obvious that the mass fraction of the smaller particles increases after grafting. This allows the conclusion that the amount of grafting is proportional to the particle surface. Another possibility for obtaining the density of unknown particles is to run two velocity experiments in chemically similar solvents with different densities (e.g., H2 O/D2 O) [101, 109]. This allows us to simultaneously determine particle sizes and density distributions. The method works best for particles with densities less than 1.5 g/mL but requires 1.0

j=i

Σ mi

i=1

PBA

0.5

rather good data quality. Recently, a diffusion correction has been implemented into this method, which is especially important for small particles/molecules so that even complex mixtures become accessible by this method [110].

4.2. Characterization of Emulsions In analogy to dispersions, emulsions, respectively mini- or microemulsions, can be characterized by the particle size distribution of the dispersed phase if flat centerpieces are used which permit the passage of light to the detector through the turbid emulsion as the emulsion cannot be diluted in contrast to a dispersion. One such example was reported for water/AOT/heptane water in oil microemulsions [111]. Furthermore, it is possible to force the coalescence of the emulsion by the ultracentrifugal field which can easily be detected as the formation of a new phase by any optical detection system of the ultracentrifuge. This is a fast and effective measure to look at the stability of emulsions in a qualitative way by determining either the centrifugal field necessary for spontaneous coalescence or the time until coalescence occurs at a given centrifugal field [112–114]. From experimental parameters such as the rotational speed and density, a coalescence pressure can be calculated as a more quantitative measure [112].

4.3. Characterization of Microgels The analytical ultracentrifuge has proved to be very useful for characterization of the thermodynamic and elastic properties of gels and microgels as reviewed in [115, 116]. In the case of microgels in a thermodynamically good solvent, the degree of swelling Q can be determined from sedimentation velocity experiments [117]. If the microgel contains un-crosslinked polymer, two components are resolved in the sedimentation coefficient distribution calculated via (1). From Gs (see Fig. 14), the amount of each component can be

PBA/SAN – ρ=0.5

particles in water D01 D1 D02 D2

0

D03 D3

200 PBA

0

D04

Σ c1

D4 Di

400 nm PBA/SAN

s-distribution of these particles in the organic solvent

particles in an organic solvent

c0

s 10-1 100 101 102 103

not crosslinked

Svedberg

Σ c1 c0

s 10-1 100 101 102 103

low crosslinking

Svedberg

Σ c1 c0

ρK 1.03

ρ4ρ3 ρ ρ1 2

1.05

s 10-1 100 101 102 103

highly crosslinked

Svedberg

ρ Σ c1

1.07 g/cm3

c0

s

Figure 13. Particle size distribution and density gradient of a fourmodal ungrafted and a SAN-grafted PBA dispersion (40 : 30 : 20 : 10 wt% mixture). D refers to the particle diameter; the index 0i to ungrafted and i to the grafted latex. Reprinted with permission from [17], W. Mächtle, “Analytical Ultracentrifugation in Biochemistry and Polymer Science,” p. 147, Royal Society of Chemistry, Cambridge, UK, 1992. © 1992, Royal Society of Chemistry.

totally crosslinked

10-1 100 101 102 103

Svedberg

Figure 14. Sedimentation coefficient distributions for lattices with different degrees of cross-linking. Reprinted with permission from [117], H. G. Müller et al., Prog. Colloid Polym. Sci. 86, 70 (1991). © 1991, Springer-Verlag.

80

Analytical Ultracentrifugation of Nanoparticles

derived. The swelling degree Q can then be calculated using  − bd 2 − 2 s 18.

(21)

where d is the diameter of the compact, unswollen particle, 2 is the density of the compact, unswollen particle,  is the density of the dispersion medium, and . is the viscosity of the diluted dispersion. b is a factor according to mr = b · m where the mass of the particle mr reduced by the soluble part is related to the mass m of the particle consisting of soluble and insoluble components. It can be derived from interference optical traces. The particle diameter of the unswollen sample has to be determined in a nonsolvent in a separate experiment. The swelling degree can be related to the molar mass of the elastically effective network chains and thus to the elastic properties of the microgel applying the Flory–Rehner theory. Such experiments not only allow the characterization of microgels but can furthermore be used to investigate the efficiency of cross-linking reactions by specifying the amount and physicochemical properties of the un-cross-linked polymer [118]. Figure 14 shows the sedimentation coefficient distributions that are derived for different cross-linking degrees of latexes. One can now easily compare samples from different cross-linking reactions.

4.4. Observation of Chemical Reactions It is possible to perform chemical reactions in the analytical ultracentrifuge with a synthetic boundary experiment using a special cell (see Fig. 15a). In biochemistry, this kind of centrifugation is called “active enzyme centrifugation” [119], but the principle can be applied to every chemical reaction. In the synthetic boundary cell (Fig. 15a), a small amount 1.0 Enzyme

Absorbance

a)

25.000 RPM Scaninterval 10 min 340 nm Phosphate buffer pH 7, enzyme active

b)

0.8 0.6 0.4 0.2

Substrate

0.0 6.2

6.4

6.6

6.8

7.0

Radius [cm] 1.0

Absorbance

0.8 0.6

30.000 RPM Scaninterval 5 min 340 nm Tris buffer pH 8.8, enzyme inactive

c)

0.4 0.2 0.0 6.2

6.4

6.6

6.8

7.0

7.2

Radius [cm]

Figure 15. (a) Synthetic boundary cell with small sample compartments. (b) Sedimentation velocity profile of a glutamate dehydrogenase mutant enzyme which reacts with its substrate. (c) Sedimentation velocity profile of the same enzyme at a different pH. The enzyme is now inactive and no chemical reaction takes place.

reaction boundary

Q=

of a reactant (usually 10–15 L) is layered onto a column of the second reactant while the centrifuge is speeded up to about 3000 rpm. Thus, a reaction boundary is established. The detection wavelength of the absorption optics is set to a wavelength where the reaction product (here the enzyme–substrate complex) absorbs. One can then observe the product formation and sedimentation as visualized in Figure 15b. The absorbance increases with the time of sedimentation, indicating that the sedimenting enzyme is reacting continuously. Such experiments can yield sedimentation and diffusion coefficients of the enzyme in its catalytically active state which enables detection of differences in polymerization state, hydration, or conformation with the nonreacting enzyme. If the extinction coefficient of the reaction product is known, the product concentration in relation to time can be calculated. Figure 15c shows that a slight change in pH can turn the enzyme to its inactive state. An advantage of the application of synthetic boundary cells of the type shown in Figure 15a is that only a few microliters corresponding to a few micrograms/nanograms of reactant are required. This type of synthetic boundary experiment was recently adapted to carry out crystallization reactions inside the spinning cell of the ultracentrifuge and was named “synthetic boundary crystallization ultracentrifugation” [120, 121]. Here, a small amount of Na2 S was layered upon a solution of CdCl2 containing a stabilizer. A fast reaction to CdS with subsequent stabilization of the small nanoparticles takes place according to the scheme shown in Figure 16. [121]. The advantages of this technique lie in the fast chemical reaction within the very small reaction zone and the quenching of additional particle growth in lack of the second reaction partner as soon as the particles move out of the reaction zone by sedimentation or diffusion processes. These particles are then fractionated according to their size and density and a particle size distribution can be determined as discussed above. However, as the formed nanoparticles are very small, they show significant diffusion so that they can diffuse back into the reaction boundary and grow further, thus forming the second growth generation of particles. This process is repeated until the reaction partner in the reaction zone is used up (which usually occurs within a few minutes), so that the particles only sediment due to their size/density with the usual diffusion broadening of the boundary. By this technique, different growth stages of nanoparticles can be

1st generation

Sedimentation of particles according to their size, observation of the particle size distribution possible

2nd generation

n-th generation

cell radius

Figure 16. Schematic representation of synthetic boundary crystallization ultracentrifugation.

81

Analytical Ultracentrifugation of Nanoparticles

4.5. Determination of Sample Homogeneity, Efficiency of Chemical Reactions, and Determination of Extinction Coefficients In many fields of polymer or colloid chemistry, complicated complexes or hybrid colloids are synthesized. It is often desirable to have a quick and convenient check for the efficiency of the reaction as well as a check of sample homogeneity. Sedimentation velocity experiments can be used very advantageously here due to the fractionation without any stationary phase as is demonstrated for the case of iron oxyhydroxide particles partly stabilized in a self-assembling microgel of 2-carrageenan (Fig. 18) [123]. If one component selectively absorbs light (FeOOH) and the second does not (2-carrageenan), one can selectively detect the absorption of the iron oxide in the polymeric superstructures with absorption optics, whereas all components together are detected by the simultaneously used Rayleigh interference optics. Here it must be stated that most colored inorganic compounds have a high extinction coefficient and are thus almost exclusively detected in the UV/vis absorption optics, whereas

thiolactic acid thioglycerine L-cysteine

relative concentration

1.2 1.0

2.8 nm

2.4 nm

0.8

0.9 nm

0.6 0.4 0.2 0.0 0.5

1.0

1.5

2.0

2.5

3.0

3.5

dparticle [nm] Figure 17. Apparent particle size distributions of CdS in the presence of different stabilizer molecules as detected by synthetic boundary crystallization ultracentrifugation. Reprinted with permission from [121], L. Börger et al., Colloids Surf., A 163, 29 (2000). © 2000, Elsevier Science.

5

Carrageenan + iron oxyhydroxide Large aggregates

Absorption optics

Absorbance / Fringe shift

investigated. However, due to the extensive diffusion of the particles, the individual particle size distributions become extensively smeared so that they are only detected as a continuous distribution. Nevertheless, even such data can show the different stabilizer capabilities with a resolution in the Angström range as shown for CdS in Figure 17 [120, 121]. Another elegant way to observe reactions with a synthetic boundary technique was introduced by Wandrey and Bartkowiak [122]. They were able to observe the formation of polyelectrolyte complex membranes between two oppositely charged polyelectrolytes and the influence of parameters such as pH, ionic strength, component ratio, and temperature. Also, the membrane formation kinetics could be detected as overlaying of the two polymer solutions resulted in the formation of a thin membrane at the solution interface which could be precisely detected and its thickness measured.

Interference optics

4

Carrageenan + iron oxyhydroxide

3

0.55 fringes

Free Carrageenan

2 2.72 fringes 1 40000 RPM, 25 °C

0 6,0

6,2

6,4

6,6

6,8

7,0

Radius [cm] Figure 18. Iron oxyhydroxide in 2-carrageenan microgels. Detection of the different species by combination of UV/vis and refractive index detection. Reprinted with permission from [123], F. Jones et al., Colloid Polym. Sci. 278, 491 (2000). © 2000, Springer-Verlag.

the simultaneously applied interference optics almost exclusively represents the local polymer concentrations which can be converted to the real concentrations via the refractive index increment and the known refractive index of the solvent. From Figure 18, three species are clearly quantitatively detected: free unbound 2-carrageenan, 2-carrageenan + iron oxyhydroxide microgels, and larger cross-linked aggregates. The control of efficiencies of chemical reactions was also reported for complicated complexes between oppositely charged polyelectrolytes [124]. If the change in the refractive index of a sample (Rayleigh interference optics) is simultaneously detected with its absorption at a specified wavelength, extinction coefficients can be determined [125]. This is not particularly interesting if a single component is considered, but for mixtures the analytical ultracentrifuge is often the only method capable of determining the extinction coefficient of a specified component in a complicated mixture. The extinction coefficient can be calculated from a concentration reading in the Rayleigh interference scan which must be converted to the concentration unit of desire using the refractive index increment and the corresponding absorption. Figure 19 demonstrates this for ZnO prepared in polystyrene–polymethacrylic acid (PS–PMAA) micelles. An extinction coefficient based on an absorption reading taken in a spectrophotometer under the assumption that all Zn2+ have reacted to ZnO delivers a low extinction coefficient of less than 1000 L mol−1 cm−1 due to the majority of UV/vis inactive species. Therefore, it is of interest to detect which amount of the sample is the spectroscopically active ZnO. Figure 19a shows the results and confirms that only a minor amount of the mixture is responsible for the detected light absorption. From the interference scans in Figure 19a, it can be seen that the component with a similar sedimentation velocity as the ZnO-filled micelles detected by the corresponding UV-absorption scans has only a concentration of 0.18 fringes. The major part is a considerably slower sedimenting component (micelles with unreacted Zn2+ ). A third component (polymer or empty micelles) is also detected.

82

Absorbance / Fringe shift

Analytical Ultracentrifugation of Nanoparticles

3.0

2.80

2.5

2.62

Interference optics

a)

ZnO filled micelles

2.0

The current trends in analytical ultracentrifugation can be divided into three subcategories: (a) improvement of data evaluation methodologies, (b) new detection systems, and (c) application of analytical ultracentrifugation to novel systems apart from the classical biomolecules.

1 23 45 6 7 8

1.5

Absorption optics

1.0 0.74 1

0.5 0.0 6.0

6.2

2

6.4

3

4

6.6

5

6.8

6 7

8

7.0

5.1. Improvement of Data Evaluation Methodologies

7.2

Extinction coefficient [l mol-1 cm-1]

Radius (cm) 10000

b) 8000

6000

4000

Fringe shift J converted to ∆n: ∆n = J * λ / a ∆n = nSample - nSolvent

2000

λ = wavelength of laser a = optical pathlength

0 300

320

340

360

5. CURRENT TRENDS IN ANALYTICAL ULTRACENTRIFUGATION

380

400

λ [nm]

Figure 19. Determination of the extinction coefficient of ZnO in PS– PMAA micelles [104]. (a) Sedimentation velocity profiles acquired simultaneously with Rayleigh interference and absorption optics (315 nm). Run parameters: 20,000 rpm, scan interval 2 min. The numbers represent the corresponding interference and absorption scans. (b) Wavelength dependence of the extinction coefficient for ZnO prepared in PS–PMAA micelles. The particle diameter of the ZnO (from TEM) is 3.2–6.7 nm. The diameter of the micelles (from light scattering) is 64.4 ± 33.9 nm. Reprinted with permission from [104], H. Cölfen et al., Prog. Colloid Polym. Sci. 107, 136 (1997). © 1997, Springer-Verlag.

The interference fringe shift of the ZnO-filled micelles can be converted to a concentration and this can be related to a spectrum of the ZnO in the ultracentrifuge cell, giving the wavelength dependence of the ZnO extinction coefficient (Fig. 19b). The extinction coefficient is found to reach about 9000 L mol−1 cm−1 . This illustrates how important the separation is to determine the spectral properties of the optically active component in a complicated mixture. Many more variations of such experiments are possible, showing that the separation capability of the AUC can be used not only to determine the concentrations of the individual compounds but also to elicit their spectral properties if they absorb light.

4.6. Analytical Ultracentrifugation of Nanoparticles Sorted by Systems Table 2 gives some examples of nanoparticle systems that have been investigated by AUC with references to the original literature. The table is by no means exhaustive or representative and is just meant to guide the reader to the original literature if solutions for a particular nanoparticulate system are sought.

Progress in computer and electronic technology has greatly facilitated the analysis capabilities of analytical ultracentrifugation and catalyzed the rebirth of this technique. Nowadays, it is relatively simple to apply fitting functions for all kinds of experimental output and, furthermore, it is possible to acquire data fast and conveniently even when simultaneously using several different methods of detection. Therefore, a major current trend in analytical ultracentrifugation is the development of better evaluation methodologies and algorithms. One example is the simultaneous radial and wavelength analysis of sedimentation equilibrium experiments possible with the Optima XL-I ultracentrifuge [153]. A complicated interacting system exhibiting three different chromophores and consisting of the Band3 membrane protein, a small ligand, oxyhemoglobin, and the complexes between these components— an 11-component system overall—could be successfully analyzed. These investigations are also applicable to complicated colloid mixtures. This amounts to seven more components than possible before when analyzing a concentration profile acquired at a single wavelength. The increase in components that can be resolved is achieved by the acquisition of as much experimental information as possible— in this case wavelength scans at different radii as well as radial scans at different wavelengths. From this information, it is possible to create a two-dimensional data surface in a radius–wavelength–absorption space which allows a proper analysis by fitting procedures. Therefore, it is not surprising that more and more scientists are using the capability of scanning at multiple wavelengths [154, 155]. A further enhancement of the data space is achieved by the application of several rotational speeds in the experiments. As this increases the experimental time, so-called short-column techniques [156, 157], which considerably decrease the time to reach the sedimentation equilibrium by the application of a shorter solution column in the ultracentrifuge cell, are becoming popular as long as the resolution of the optical system is still sufficient for the short solution column. However, for the mainly polydisperse nanoparticles, such an approach is restricted. Significant effort has also been put into other efficient analyses of ultracentrifuge data which generally increases the sensitivity of ultracentrifuge experiments. Nowadays, components can be identified, which are almost not visible in the experimental raw data and thus could not be recognized in the old days of analytical ultracentrifugation. As the Rayleigh interference system of the XL-I ultracentrifuge allows a rapid acquisition of large data sets, the confidence intervals of fitting procedures in general can be

83

Analytical Ultracentrifugation of Nanoparticles

Table 2. Application of AUC to different nanoparticulate systems. System

Experiment type/quantity

Ref.

Latex dispersions

Sedimentation velocity/particle size distribution; static and dynamic density gradient/particle density

Polystyrene [14]; polystyrene, polybutylacrylatestyrene, polybutadiene, polyethylacrylate [15]; polystyrene [16]; polystyrene, polybutylacrylate (also styrene and acrylonitrile grafted), polybutadiene, acrylic homopolymer and copolymer dispersions [17]; polychloropropene [96]; polystyrene, polystyrene/polybutadiene, polystyrene-cobutadiene, polychloropropene [99]; polystyrene, polystyreneco-butadiene, styrene, and acrylonitrile grafted polybutadiene [101]; polystyrene [102]; polyurethane [105]; polystyrene [106]; polystyrene, polybutadiene, Polybutylacrylate-cobutadiene [109]

Inorganic nanoparticles

Sedimentation velocity/particle size distribution

Au [3]; Au [94]; ZrO2 [98]; Au55 cluster [100]; Pt, ZnO [103]; ZnO, Au, CdS, Pt [104]; CdS [105]; silica [106]; 3-FeOOH [109]; CdS [120]; CdS [121]; iron oxide [126]; surfactantencapsulated (NH4 )[H3 Mo57 V6 (NO)(6)O-183 H2 O(18)] [127]

Inorganic complexes

Sedimentation equilibrium/molar mass

Zr complexes [55]; Zr complexes [56]

Magnetic fluids

Sedimentation velocity

Fe3 O4 [128]; Fe3 O4 [129]

Microgels

Sedimentation velocity/composition and density

Styrene-butadiene [117]; acrylic acid [118]; polystyrene-poly-4vinylpyridine microgels [130]

Gels

Sedimentation velocity and equilibrium/ swelling degree and pressure, amount of soluble compounds

Gelatin, 2-carrageenan, agar, casein [116]; gelatin, 2-carrageenan, agar, casein [131]; other systems (reviews)

Emulsions

Sedimentation equilibrium

SDS-stabilized n-decane in water [112]; food emulsions [113]; Legumin-stabilized n-decane in water [114]; legumin-stabilized n-decane in water [132]

Supramolecular assemblies and polymers

Sedimentation equilibrium/molar mass, sedimentation velocity/particle size

Co-coordination arrays [133]; Co-coordination arrays [134]; Fe-coordination polymers [135]; Co-coordination arrays [136]

Polyelectrolyte complexes

Sedimentation velocity/particle size and composition, sedimentation equilibrium/ molar mass and interaction and membrane characteristics; synthetic boundary/ membrane formation

Alginate/chitosan [122]; polysyrenesulfonate/ polydiallyldimethylaminchlorid acrylamide [124]

Micelles

Sedimentation velocity/composition; dynamic density gradient/composition

Polymampholyte/fluorinated and hydrogenated dodecanoic acid [137]; lipid and detergent micelles [138]; review, different systems [139]; polystyrene-b-polyisoprene, polystyrene-bpoly(ethylene-co-propylene) [140]; polylactide-copolyethyleneglycol [141]; enzymes in reverse micelles [142]

Dendrimers

Sedimentation velocity and sedimentation equilibrium/hydrodynamics

Carbohydrate-coated polypropyleneimine dendrimers [143]; lactosylated polyamidoamine dendrimers [144]

Hybrid colloids

Sedimentation velocity/composition/ particle size

Au/polystyrenesulfonate microgels, CdS or Au in polystyreneb-poly-4-vinylpyridin micelles, Pt in polyethyleneoxidepolymethacrylic acid [104]; Calcium phosphates in alkylated polyethyleneoxide-b-polymethacrylic acid [108]; FeOOH, NiOOH, and CoOOH in 2-carrageenan microgels [123]; review of various systems [145]; CdS in reverse micelles [146]; Pt and Pd in polystyrene-b-polyethyleneoxide/ cetylpyridiniumchloride mixed micelles [147]; Au, Pt, and Pd in poly-2-vinylpyridine-b-polyethyleneoxide micelles [148]; Pt, Pd, Rh, and Cu in polyethyleneoxide-b-polyethyleneimine [149]

Organic colloids

Sedimentation equilibrium; sedimentation velocity/composition, molar mass

Cu-phtalocyanine [102]; 2-casein particles [150]; 2-casein particles [151]

Nanocapsules

Sedimentation velocity, density gradient/composition

Oil-filled polybutylcyanoacrylate nanocapsules [152]

84 significantly improved. However, the interference data contain time-invariant and radial-invariant noise components in addition to the random noise so that an efficient algebraic method was developed for the elimination of the time- and radial-invariant noise [158]. This leads to a high sensitivity of Rayleigh interference data even for such experiments where the sedimentation boundary can only barely be traced due to the noise components. The so-improved data sets can then be used for fitting to approximate analytical and numerical solutions of the Lamm equation. It is also possible to eliminate systematic noise components in the interference data of sedimentation equilibrium profiles by extraction of the timeinvariant noise during the approach to equilibrium [159]. It turned out by this approach that the sensitivity of the Rayleigh interference optics for sedimentation equilibrium experiments can be improved by an order of magnitude so that equilibrium experiments with concentrations as low as 50 g/mL are no problem for Rayleigh interference detection. Very recently, the described systematic noise decomposition method [158] has been applied for linear least squares modeling of the sedimenting boundary by a superposition of gs of ideal nondiffusing components [160]. This approach turned out to be especially advantageous for data acquired during a large time interval so that, in such cases, an improved resolution could be achieved especially for heterogeneous mixtures. A very recent advance is the determination of molar mass distributions from the fitting of finite-element solutions of the Lamm equation (8) paired with regularization algorithms as discussed before [51]. However, the range of successful application still needs to be explored.

5.2. New Detection Systems for the XL-I Ultracentrifuge Computer-based picture evaluation allows the on-line capture of Schlieren patterns, enabling the fast and efficient evaluation of these records, which previously had to be photographed [9, 10, 161, 162]. The on-line digitization and picture evaluation of these optical records rely on the proper detection of the zeroth-order Fresnel fringe, which is the detected Schlieren curve in Figure 2a. Either by means of picture manipulation performed in order to facilitate a proper detection of the Schlieren curve even with a nonuniformly illuminated picture or by additional consideration of higher order Fresnel fringes, the concentration gradient curve can be properly detected automatically. Especially with the system built at the University of Leicester/ Nottingham [10], high rates of data acquisition can be achieved with good data quality so that the on-line Schlieren system proves its virtues. This system is currently being adapted to the XL-I ultracentrifuge. Also, the Schlieren optics was set up on a preparative ultracentrifuge [12]. A potentially interesting optical system that should also be adaptable to XL ultracentrifuges is the Lebedev interferometer, which gives an interference pattern but with a Schlieren peak for a sedimenting boundary [163]. Therefore, the fast evaluation algorithms developed for the Rayleigh interferometer should be applicable here as well, resulting in highquality Schlieren images.

Analytical Ultracentrifugation of Nanoparticles

Another system that is currently commercially available on the XL-I ultracentrifuge is the fluorescence detector developed by Schmidt and Riesner [13] and adapted to the XL-I by MacGregor and Laue [18]. These systems are designed to be simultaneously applicable with the present absorption and Rayleigh interference optics with the exception of the Schlieren optics, which uses the optical path of the interference optics. Further current approaches toward the faster detection of UV/vis data are under way in the laboratories of Laue and Cölfen [164].

5.3. Application of Analytical Ultracentrifugation to Novel Systems Apart from biological systems, analytical ultracentrifugation is more and more being applied to complicated synthetic systems, for example, supramolecular assemblies and polymers [133–136], colloidal clusters [127], smallest colloids and their interactions with stabilizers [100, 120, 121, 127], synthetic polyelectrolytes [165–169] and complexes thereof [124], micelles [139–141], components inside micelles [142], dendrimers [143, 144], and hybrid colloids between organic and inorganic matter [108, 148, 149]. In most of these applications, advantage is taken of the simultaneous acquisition of Rayleigh interference and UV/vis absorption data. Insofar, it can be speculated how much a simultaneous application of additional detectors could yield further information and thus would open further applications. A visionary paper about future requirements in ultracentrifuge methodology and detection systems was published by Mächtle [170] more than 10 years ago, and it can at least be stated that the simultaneous application of as many detectors as possible as already realized in size exclusion chromatography is advantageous and will open up new applications for AUC.

6. AVAILABILITY OF FREE EVALUATION SOFTWARE In general, most of the best evaluation software for analytical ultracentrifugation experiments is still free and shared among different user groups. To download free evaluation software, one can use several offers: ftp://bbri.harvard.edu/dka200/anonymous/rasmb/spin/ at the Boston Biomedical Research Institute or http://www. cauma.uthscsa.edu/software at the University of Texas Center for Analytical Ultracentrifugation of Macromolecular Assemblies. At both sites, one can find a comprehensive collection of the freely available evaluation software. Furthermore, there is a user group called RASMB (Reversible Associations in Structural and Molecular Biology), which is a discussion platform for the study of interacting systems and analytical ultracentrifugation. However, do not hesitate to be included on this list if you are not directly working in the field of biochemistry/biophysics as many general aspects of analytical ultracentrifugation are adressed on this discussion platform. To be included on the e-mail list, check out http://rasmb-email.bbri.org/mailman/ listinfo/rasmb. On the RASMB, there are further links to all relevant Internet sites related to AUC.

85

Analytical Ultracentrifugation of Nanoparticles

7. CONCLUSION Analytical ultracentrifugation is one of the most universally applicable and variable physicochemical techniques for characterization of colloidal or polymeric systems in solution, dispersion, or emulsion. The power of analytical ultracentrifugation lies in the fractionation of the sample due to either the molar mass/size (sedimentation velocity/ sedimentation equilibrium) or the chemical structure (density gradient) in the solvent and therefore the possibility to measure distributions without the interaction with any stationary phase or solvent flows as occurs in current chromatographic techniques. Wherever information is sought for the individual components in a mixture, analytical ultracentrifugation is among the first techniques of choice. On the other hand, the sedimentation equilibrium is an equilibrium well described by thermodynamics. Thus, it is possible to derive information even on complicated or interacting systems on an absolute basis and without disturbing this equilibrium. There are many different physicochemical quantities that can be determined. However, simple investigations such as those of sample homogeneity can provide much information on a system. But a complete physicochemical characterization is also possible, often with information that cannot be obtained by other techniques. Similar to other analytical techniques, the rapid computer and electronics development of the past years was of benefit for analytical ultracentrifugation. This development provides many new applications, improved sensitivity of the experiments, and even the realization of evaluation approaches not possible so far. Briefly stated, not only has the amount of experimental data from an ultracentrifuge experiment been increased but also the amount of information. This has led to a resurgence of interest in this technique. Especially if multidetection systems can be realized on ultracentrifuge platforms, new dimensions of analytical information will be available from a single sedimentation velocity experiment. Therefore, a synergy between ultracentrifuge hardware and methodology development can be expected which will lead to an increasing number of applications. In biochemistry/biophysics, analytical ultracentrifugation is once again gaining importance after a slowdown in the 1980s. The same trend is not unlikely for the field of polymer or colloid chemistry although so far many chemists are not yet aware of the potential of analytical ultracentrifugation.

GLOSSARY Analytical ultracentrifugation Analytical observation of sedimentation processes at speeds up to 60000 RPM. Density gradient Separation of a binary mixture to a continuous concentration and thus density gradient in the ultracentrifugal field. Lamm equation The basic equation of analytical Ultracentrifugation which describes the transport process by sedimentation and diffusion in the centrifugal field. Partial specific volume The volume change in ml if one gram of pure sample is added to an infinite amount of solvent. Sedimentation coefficient Detected Sedimentation velocity normalized to gravitational force.

Sedimentation equilibrium experiment Particle sedimentation at moderate speed where sedimentation is balanced by counter diffusion to yield the absolute molar mass. Sedimentation velocity experiment Particle sedimentation at very high speeds to determine the sedimentation coefficient. Synthetic boundary experiment Overlaying of a solvent onto a solution to create a liquid interface.

REFERENCES 1. T. Svedberg and J. B. Nichols, J. Am. Chem. Soc. 45, 2910 (1923). 2. T. Svedberg and K. O. Pedersen, “The Ultracentrifuge.” Clarendon, Oxford, 1940. 3. T. Svedberg and H. Rinde, J. Am. Chem. Soc. 46, 2677 (1924). 4. R. Giebeler, in “Analytical Ultracentrifugation in Biochemistry and Polymer Science” (S. E. Harding, A. J. Rowe, and J. C. Horton, Eds.), p. 16. Royal Society of Chemistry, Cambridge, UK, 1992. 5. H. Kim, R. C. Deonier, and J. W. Williams, Chem. Rev. 77, 659 (1977). 6. T. M. Laue, in “Analytical Ultracentrifugation in Biochemistry and Polymer Science” (S. E. Harding, A. J. Rowe, and J. C. Horton, Eds.), p. 63. Royal Society of Chemistry, Cambridge, UK, 1992. 7. A. Rowe, S. Wynne Jones, D. G. Thomas, and S. E. Harding, in “Analytical Ultracentrifugation in Biochemistry and Polymer Science” (S. E. Harding, A. J. Rowe, and J. C. Horton, Eds.), p. 49. Royal Society of Chemistry, Cambridge, UK, 1992. 8. H. Cölfen and W. Borchard, Proc. SPIE 2136, 307 (1994). 9. U. Klodwig and W. Mächtle, Colloid Polym. Sci. 267, 1117 (1989). 10. A. C. Clewlow, N. Errington, and A. J. Rowe, Eur. Biophys. J. 25, 311 (1997). 11. P. H. Lloyd, “Optical Methods in Ultracentrifugation, Electrophoresis and Diffusion.” Oxford Univ. Press, 1974. 12. W. Mächtle, Prog. Colloid Polym. Sci. 113, 1 (1999). 13. B. Schmidt and D. Riesner, in “Analytical Ultracentrifugation in Biochemistry and Polymer Science” (S. E. Harding, A. J. Rowe, and J. C. Horton, Eds.), p. 176. Royal Society of Chemistry, Cambridge, UK, 1992. 14. H. J. Cantow, Makromol. Chem. 70, 130 (1964). 15. W. Scholtan and H. Lange, Kolloid Z. Z. Polym. 250, 782 (1972). 16. H. G. Müller, Colloid Polym. Sci. 267, 1113 (1989). 17. W. Mächtle, in “Analytical Ultracentrifugation in Biochemistry and Polymer Science” (S. E. Harding, A. J. Rowe, and J. C. Horton, Eds.), p. 147. Royal Society of Chemistry, Cambridge, UK, 1992. 18. L. K. MacGregor and T. M. Laue, Biophys. J. 76, A357 (1999). 19. H. K. Schachman, “Ultracentrifugation in Biochemistry.” Academic Press, New York, 1959. 20. A. Böhm, S. Kielhorn Bayer, and P. Rossmanith, Prog. Colloid Polym. Sci. 113, 121 (1999). 21. W. F. Stafford, in “Analytical Ultracentrifugation in Biochemistry and Polymer Science” (S. E. Harding, A. J. Rowe, and J. C. Horton, Eds.), p. 359. Royal Society of Chemistry, Cambridge, UK, 1992. 22. J. C. Lee and S. Rajendran, in “Modern Analytical Ultracentrifugation” (T. M. Schuster and T. M. Laue, Eds.), p. 138. Birkhäuser, Boston, 1994. 23. R. J. Goldberg, J. Phys. Chem. 57, 194 (1953). 24. N. Muramatsu and A. P. Minton, Anal. Biochem. 168, 345 (1988). 25. K. E. van Holde and W. O. Weischet, Biopolymers 17, 1387 (1978). 26. K. Schilling, Ph.D. thesis, Potsdam, 1999. 27. B. Demeler, H. Saber, and J. C. Hansen, Biophys. J. 72, 397 (1997). 28. J. Geiselmann, T. D. Yager, S. C. Gill, P. Camettes, and P. H. von Hippel, Biochemistry 31, 111 (1992).

86 29. S. C. Gill, T. D. Yager, and P. H. von Hippel, J. Mol. Biol. 220, 325 (1991). 30. J. C. Hansen and D. Lohr, J. Biol. Chem. 268, 5840 (1993). 31. J. C. Hansen, J. Ausio, V. H. Stanik, and K. E. van Holde, Biochemistry 28, 9129 (1989). 32. D. A. Yphantis, Biophys. J. 45, 324a (1984). 33. W. F. Stafford, Anal. Biochem. 203, 295 (1992). 34. N. Errington, O. Byron, and A. J. Rowe, Biophys. Chem. 80, 189 (1999). 35. W. F. Stafford, Biophys. J. 66, A280 (1994). 36. W. F. Stafford, Biophys. J. 74, A301 (1998). 37. W. F. Stafford, in “Modern Analytical Ultracentrifugation” (T. M. Schuster and T. M. Laue, Eds.), p. 119. Birkhäuser, Boston, 1994. 38. J. Behlke and O. Ristau, Eur. Biophys. J. 25, 325 (1997). 39. J. Behlke and A. Knespel, J. Cryst. Growth 158, 388 (1996). 40. O. Lamm, Ark. Math. Astron. Fysik B 21, 1 (1929). 41. L. A. Holladay, Biophys. Chem. 10, 187 (1979). 42. J. Behlke and O. Ristau, Prog. Colloid Polym. Sci. 107, 27 (1997). 43. L. A. Holladay, Biophys. Chem. 11, 303 (1980). 44. J. Philo, in “Modern Analytical Ultracentrifugation” (T. M. Schuster and T. M. Laue, Eds.), p. 156. Birkhäuser, Boston, 1994. 45. J. Behlke and O. Ristau, Biophys. J. 72, 428 (1997). 46. J. Philo, Biophys. J. 72, 435 (1997). 47. B. Demeler and H. Saber, Biophys. J. 74, 444 (1998). 48. P. Schuck, C. E. McPhee, and G. J. Howlett, Biophys. J. 74, 466 (1998). 49. P. Schuck, Biophys. J. 75, 1503 (1998). 50. P. Schuck and D. B. Millar, Anal. Biochem. 259, 48 (1998). 51. P. Schuck, Biophys. J. 78, 1606 (2000). 52. P. Schuck, M. A. Perugini, N. R. Gonzales, G. J. Howlett, and D. Schubert, Biophys. J. 82, 1096 (2002). 53. K. E. van Holde and R. L. Baldwin, J. Phys. Chem. 62, 734 (1958). 54. D. A. Yphantis, Ann. N.Y. Acad. Sci. 88, 586 (1960). 55. K. A. Kraus and J. S. Johnson, J. Am. Chem. Soc. 75, 5769 (1953). 56. J. S. Johnson and K. A. Kraus, J. Am. Chem. Soc. 78, 3937 (1956). 57. J. M. Creeth and R. H. Pain, Prog. Biophys. Mol. Biol. 17, 217 (1967). 58. W. D. Lansing and E. O. Kraemer, J. Am. Chem. Soc. 57, 1369 (1935). 59. H. Fujita, “Foundations of Ultracentrifugal Analysis.” Wiley, New York, 1975. 60. M. D. Lechner, in “Analytical Ultracentrifugation in Biochemistry and Polymer Science” (S. E. Harding, A. J. Rowe, and J. C. Horton, Eds.), p. 295. Royal Society of Chemistry, Cambridge, UK, 1992. 61. M. L. Johnson, J. J. Correia, D. A. Yphantis, and H. R. Halvorson, Biophys. J. 36, 575 (1981). 62. M. L. Johnson and S. G. Frazier, in “Methods in Enzymology” (C. H. W. Hirs and S. N. Timasheff, Eds.), Academic Press, San Diego, 1985. 63. D. K. McRorie and P. J. Voelker, “Self-Associating Systems in the Analytical Ultracentrifuge.” Beckman Instruments, Fullerton, CA, 1993. 64. D. A. Yphantis and T. Arakawa, Biochemistry 26, 5422 (1987). 65. N. G. Dolinnaya, E. H. Braswell, J. A. Fossella, H. Klump, and J. R. Fresco, Biochemistry 32, 10263 (1993). 66. P. B. Harbury, T. Zhang, P. S. Kim, and T. Alber, Science 262, 1401 (1993). 67. J. Liu, T. M. Laue, H. U. Choi, L. H. Tang, and L. Rosenberg, J. Biol. Chem. 269, 28366 (1994). 68. T. M. Laue, M. A. Starovasnick, R. E. Klevit, and H. Weintraub, Proc. Natl. Acad. Sci. U.S.A. 92, 11824 (1995). 69. C. G. Lon, E. H. Braswell, D. Zhu, J. Apigo, J. Baum, and B. Bodsky, Biochemistry 32, 11688 (1993). 70. D. R. Hall, S. E. Harding, and D. J. Winzor, Prog. Colloid Polym. Sci. 113, 62 (1999). 71. D. Schubert and P. Schuck, Prog. Colloid Polym. Sci. 86, 12 (1991).

Analytical Ultracentrifugation of Nanoparticles 72. J. M. Creeth and S. E. Harding, J. Biochem. Biophys. Methods 7, 25 (1982). 73. S. E. Harding, J. C. Horton, and P. J. Morgan, in “Analytical Ultracentrifugation in Biochemistry and Polymer Science” (S. E. Harding, A. J. Rowe, and J. C. Horton, Eds.), p. 275. Royal Society of Chemistry, Cambridge, UK, 1992. 74. H. Cölfen and S. E. Harding, Eur. Biophys. J. 25, 333 (1997). 75. B. K. Milthorpe, P. D. Jeffrey, and L. W. Nichol, Biophys. Chem. 3, 169 (1975). 76. D. J. Winzor and P. R. Wills, in “Modern Analytical Ultracentrifugation” (T. M. Schuster and T. M. Laue, Eds.), p. 66. Birkhäuser, Boston, 1994. 77. P. R. Wills, M. P. Jacobsen, and D. J. Winzor, Biopolymers 38, 119 (1996). 78. H. Cölfen and D. J. Winzor, Prog. Colloid Polym. Sci. 107, 36 (1997). 79. P. R. Wills, M. P. Jacobsen, and D. J. Winzor, Prog. Colloid Polym. Sci. 107, 1 (1997). 80. G. B. Ralston and M. B. Morris, in “Analytical Ultracentrifugation in Biochemistry and Polymer Science” (S. E. Harding, A. J. Rowe, and J. C. Horton, Eds.), p. 253. Royal Society of Chemistry, Cambridge, UK, 1992. 81. R. Buchdahl, H. A. Ende, and L. H. Peebles, J. Polym. Sci., Part C 1, 143 (1963). 82. M. Meselson, F. W. Stahl, and J. Vinograd, Proc. Natl. Acad. Sci. U.S.A. 43, 581 (1957). 83. J. J. Hermans and H. A. Ende, J. Polym. Sci., Part C 1, 161 (1963). 84. J. J. Hermans, J. Polym. Sci., Part C 1, 179 (1963). 85. H. A. Ende, Makromol. Chem. 88, 159 (1965). 86. M. D. Lechner, Macromol. Rapid Commun. 18, 781 (1997). 87. M. D. Lechner, W. Mächtle, and U. Sedlack, Prog. Colloid Polym. Sci. 107, 148 (1997). 88. M. D. Lechner and W. Borchard, Eur. Polym. J. 35, 371 (1999). 89. H. Lange, Colloid Polym. Sci. 258, 1077 (1980). 90. W. Mächtle, Colloid Polym. Sci. 262, 270 (1984). 91. L. Börger, Eur. Biophys. J., to appear. 92. P. M. Budd, Prog. Colloid Polym. Sci. 94, 107 (1994). 93. H. Cölfen and K. Schilling, Prog. Colloid Polym. Sci. 113, 44 (1999). 94. H. Rinde, The Distribution of the Sizes of Particles in Gold Sols, Ph.D. Thesis, Upsala, 1928. 95. J. B. Nichols, Physics 1, 254 (1931). 96. J. B. Nichols, E. O. Kramer, and E. D. Bailey, J. Phys. Chem. 36, 326 (1932). 97. H. Cölfen, Habilitation Thesis, Potsdam, 2001. 98. H. Cölfen, H. Schnablegger, A. Fischer, F. C. Jentoft, G. Weinberg, and R. Schlögl, Langmuir 18, 3500 (2002). 99. H. Lange, Part. Part. Syst. Charact. 12, 148 (1995). 100. D. H. Rapoport, W. Vogel, H. Cölfen, and R. Schlögl, J. Phys. Chem. B 101, 4175 (1997). 101. H. G. Müller and F. Herrmann, Prog. Colloid Polym. Sci. 99, 114 (1995). 102. W. Mächtle, Biophys. J. 76, 1080 (1999). 103. H. Cölfen and T. Pauck, Colloid Polym. Sci. 275, 175 (1997). 104. H. Cölfen, T. Pauck, and M. Antonietti, Prog. Colloid Polym. Sci. 107, 136 (1997). 105. H. G. Müller, Prog. Colloid Polym. Sci. 107, 180 (1997). 106. M. D. Lechner and W. Mächtle, Prog. Colloid Polym. Sci. 113, 37 (1999). 107. D. G. Kurth, P. Lehmann, D. Volkmer, H. Cölfen, M. J. Koop, A. Müller, and A. DuChesne, Chem. Eur. J. 6, 385 (2000). 108. M. Antonietti, M. Breulmann, C. G. Göltner, H. Cölfen, K. K. W. Wong, D. Walsh, and S. Mann, Chem. Eur. J. 4, 2493 (1998). 109. W. Mächtle, Makromol. Chem. 185, 1025 (1984). 110. K. Schilling and H. Cölfen, Prog. Colloid Polym. Sci. 113, 50 (1999). 111. P. M. Budd, R. K. Pinfield, and C. Price, Prog. Colloid Polym. Sci. 107, 189 (1997). 112. K. Strenge and A. Seifert, Prog. Colloid Polym. Sci. 86, 76 (1991).

87

Analytical Ultracentrifugation of Nanoparticles 113. A. Seifert, K. Strenge, M. Schultz, and H. Schmandtke, Nahrung 9, 989 (1991). 114. A. Seifert and K. D. Schwenke, Prog. Colloid Polym. Sci. 99, 31 (1995). 115. H. Cölfen, Colloid Polym. Sci. 273, 1101 (1995). 116. H. Cölfen, Biotechnol. Genet. Eng. 16, 87 (1999). 117. H. G. Müller, A. Schmidt, and D. Kranz, Prog. Colloid Polym. Sci. 86, 70 (1991). 118. W. Mächtle, G. Ley, and J. Streib, Prog. Colloid Polym. Sci. 99, 144 (1995). 119. R. Cohen, C. R. Acad. Sci. Paris 256, 3513 (1963). 120. L. Börger and H. Cölfen, Prog. Colloid Polym. Sci. 113, 23 (1999). 121. L. Börger, H. Cölfen, and M. Antonietti, Colloids Surf., A 163, 29 (2000). 122. C. Wandrey and A. Bartkowiak, Colloids Surf., A 180, 141 (2001). 123. F. Jones, H. Cölfen, and M. Antonietti, Colloid Polym. Sci. 278, 491 (2000). 124. N. Karibyants, H. Dautzenberg, and H. Cölfen, Macromolecules 30, 7803 (1997). 125. P. Voelker, Prog. Colloid Polym. Sci. 99, 162 (1995). 126. J. B. Nichols, E. O. Kramer, and E. D. Bailey, J. Phys. Chem. 36, 326 (1932). 127. D. G. Kurth, P. Lehmann, D. Volkmer, H. Cölfen, M. J. Koop, A. Müller, and A. Du Chesne, Chem. Eur. J. 6, 385 (2000). 128. A. Seifert, N. Buske, and K. Strenge, Colloids Surf. 57, 267 (1991). 129. A. Seifert and N. Buske, J. Magn. Magn. Mater. 122, 115 (1993). 130. E. E. Remsen, K. B. Thurmond, and K. L. Wooley, Macromolecules 32, 3685 (1999). 131. H. Cölfen, Colloid Polym. Sci. 273, 1101 (1995). 132. J. P. Krause, R. Wustneck, A. Seifert, and K. D. Schwenke, Colloids Surf., B 10, 119 (1998). 133. D. Schubert, J. A. van den Broek, B. Sell, H. Durchschlag, W. Mächtle, U. S. Schubert, and J. M. Lehn, Prog. Colloid Polym. Sci. 107, 166 (1997). 134. C. Tziatzios, H. Durchschlag, B. Sell, J. A. van den Broek, W. Mächtle, W. Haase, J. M. Lehn, C. H. Weidl, C. Eschbaumer, D. Schubert, and U. S. Schubert, Prog. Colloid Polym. Sci. 113, 114 (1999). 135. M. Schütte, D. G. Kurth, M. R. Linford, H. Cölfen, and H. Möhwald, Angew. Chem., Int. Ed. Engl. 37, 2891 (1998). 136. D. Schubert, C. Tziatzios, P. Schuck, and U. S. Schubert, Chem. Eur. J. 5, 1377 (1999). 137. A. F. Thünemann, K. Sander, W. Jaeger, and R. Dimova, Langmuir 18, 5099 (2002). 138. A. Lustig, A. Engel, and M. Zulauf, Biochim. Biophys. Acta 1115, 89 (1991). 139. R. W. Roxby, in “Analytical Ultracentrifugation in Biochemistry and Polymer Science” (S. E. Harding, A. J. Rowe, and J. C. Horton, Eds.), p. 609. Royal Society of Chemistry, Cambridge, UK, 1992. 140. M. Pacovska, K. Prochazka, Z. Tuzar, and P. Munk, Polymer 34, 4585 (1993). 141. S. A. Hagan, A. G. A. Coombes, M. C. Garnett, S. E. Dunn, M. C. Davies, L. Illum, and S. S. Davis, Langmuir 12, 2153 (1996). 142. N. A. Chebotareva, B. I. Kurganov, and A. A. Burlakova, Prog. Colloid Polym. Sci. 113, 129 (1999). 143. G. M. Pavlov, E. V. Korneeva, K. Jumel, S. E. Harding, E. W. Meijer, H. W. I. Peerlings, J. Fraser Stoddart, and S. A. Nepogodiev, Carbohydr. Polym. 38, 195 (1999). 144. G. M. Pavlov, E. V. Korneeva, R. Roy, N. A. Michailova, P. C. Ortega, and M. A. Perez, Prog. Colloid Polym. Sci. 113, 150 (1999). 145. H. Cölfen, Habilitation Thesis, Potsdam, 2001. 146. B. H. Robinson, T. F. Towey, S. Zourab, A. J. W. G. Visser, and A. Vanhoek, Colloids Surf. 61, 175 (1991).

147. L. M. Bronstein, D. M. Chernychov, G. I. Timofeeva, L. V. Dubrovina, P. M. Valetsky, E. S. Obolonkova, and A. R. Khokhlov, Langmuir 16, 3626 (2000). 148. L. H. Bronstein, S. N. Sidorov, P. M. Valetsky, J. Hartmann, H. Cölfen, and M. Antonietti, Langmuir 15, 6256 (1999). 149. S. N. Sidorov, L. M. Bronstein, P. M. Valetsky, J. Hartmann, H. Cölfen, H. Schnablegger, and M. Antonietti, J. Colloid Interface Sci. 212, 197 (1999). 150. H. M. Farrell, E. D. Wickham, H. J. Dower, E. G. Piotrowski, P. D. Hoagland, P. H. Cooke, and M. L. Groves, J. Protein Chem. 18, 637 (1999). 151. H. M. Farrell, T. F. Kumosinski, P. H. Cooke, P. D. Hoagland, E. D. Wickham, J. J. Unruh, and M. L. Groves, Int. Dairy J. 9, 193 (1999). 152. M. Wohlgemuth, W. Mächtle, and C. Mayer, J. Microencaps. 17, 437 (2000). 153. P. Schuck, Prog. Colloid Polym. Sci. 94, 1 (1994). 154. M. S. Lewis, R. I. Shrager, and S. J. Kim, in “Modern Analytical Ultracentrifugation” (T. M. Schuster and T. M. Laue, Eds.), p. 94. Birkhäuser, Boston, 1994. 155. A. P. Minton, Prog. Colloid Polym. Sci. 107, 11 (1997). 156. K. E. van Holde and R. L. Baldwin, J. Phys. Chem. 62, 734 (1958). 157. J. J. Correia and D. A. Yphantis, in “Analytical Ultracentrifugation in Biochemistry and Polymer Science” (S. E. Harding, A. J. Rowe, and J. C. Horton, Eds.), Royal Society of Chemistry, Cambridge, UK, 1992. 158. P. Schuck and B. Demeler, Biophys. J. 76, 2288 (1999). 159. P. Schuck, Anal. Biochem. 272, 199 (1999). 160. P. Schuck and P. Rossmanith, Biopolymers 54, 328 (2000). 161. H. Cölfen and W. Borchard, Prog. Colloid Polym. Sci. 94, 90 (1994). 162. D. Kisters and W. Borchard, Prog. Colloid Polym. Sci. 113, 10 (1999). 163. P. Lavrenko, V. Lavrenko, and V. Tsvetkov, Prog. Colloid Polym. Sci. 113, 14 (1999). 164. T. M. Laue and H. Cölfen, “Presentations on Advances in AUC and Hydro Meeting,” 2002. 165. P. M. Budd, in “Analytical Ultracentrifugation in Biochemistry and Polymer Science” (S. E. Harding, A. J. Rowe, and J. C. Horton, Eds.), Royal Society of Chemistry, Cambridge, UK, 1992. 166. E. Görnitz, M. Hahn, W. Jaeger, and H. Dautzenberg, Prog. Colloid Polym. Sci. 107, 127 (1997). 167. G. J. Timofejeva, S. A. Pavlova, C. Wandrey, W. Jaeger, M. Hahn, K. J. Linow, and E. Görnitz, Acta Polym. 41, 479 (1990). 168. C. Wandrey and E. Görnitz, Acta Polym. 43, 320 (1992). 169. M. Hahn, E. Görnitz, and H. Dautzenberg, Macromolecules 31, 5616 (1998). 170. W. Mächtle, Prog. Colloid Polym. Sci. 86, 111 (1991).

RELATED LITERATURE Although the primary literature cited above will give the reader an appropriate insight into the AUC analysis of nanoparticle systems, a selection of related literature follows which contains textbooks and books about recent AUC developments together with a short description. This is mainly for the reader who wants to learn more about AUC.

GENERAL TEXTBOOKS There are several general books in the English language on analytical ultracentrifugation. All of them are quite out of date, but are nevertheless still useful both for the beginner and for the more experienced user if they are still available. 171. T. Svedberg and K. O. Pedersen, “The Ultracentrifuge.” Clarendon, Oxford, 1940. The classical textbook about analytical ultracentrifugation still with much impact today.

88 172. H. K. Schachman, “Ultracentrifugation in Biochemistry.” Academic Press, New York, 1959. A compact and useful book covering experimental and theoretical aspects. 172. C. H. Chervenka, “A Manual of Methods for the Analytical Ultracentrifuge.” Spinco Division, Beckman Instruments, Palo Alto, CA, 1969. A very useful compilation of classical evaluation methods with examples. 173. T. J. Bowen and A. J. Rowe, “An Introduction to Ultracentrifugation.” Wiley, London, 1970. An introduction for the beginner. 174. J. W. Williams, “Ultracentrifugation of Macromolecules.” Academic Press, New York, 1972. A compilation of seminars covering polydisperse solute and self-associating systems in more detail. 175. H. Fujita, “Foundations of Ultracentrifugal Analysis.” Wiley, New York, 1975. An exhaustive description of the mathematical theory behind analytical ultracentrifugation, not recommended for the beginner.

NEW BOOKS WITH RECENT DEVELOPMENTS 176. S. E. Harding, A. J. Rowe, and J. C. Horton, Eds., “Analytical Ultracentrifugation in Biochemistry and Polymer Science.” Royal Society of Chemistry, Cambridge, UK, 1992. The most comprehensive modern book about analytical ultracentrifugation. A very good overview about methods and techniques of analytical ultracentrifugation and a valuable source of modern applications.

Analytical Ultracentrifugation of Nanoparticles 177. T. M. Schuster and T. M. Laue, Eds., “Modern Analytical Ultracentrifugation.” Birkhäuser, Boston, 1994. A compilation of some modern trends in analytical ultracentrifugation.

SYMPOSIUM BOOKS WITH RECENT TRENDS IN ANALYTICAL ULTRACENTRIFUGATION OF THE LAST DECADE 178. W. Borchard, Prog. Colloid Polym. Sci. 86 (1991). Papers from the Seventh Symposium on Analytical Ultracentrifugation, Duisburg, 1991. 179. M. D. Lechner, Prog. Colloid Polym. Sci. 94, (1994). Papers from the Eighth Symposium on Analytical Ultracentrifugation, Osnabrück, 1993. 180. J. Behlke, Prog. Colloid Polym. Sci. 99 (1995). Papers from the Ninth Symposium on Analytical Ultracentrifugation, Berlin, 1995. 181. R. Jaenicke and H. Durchschlag, Prog. Colloid Polym. Sci. 107 (1997). Papers from the 10th Symposium on Analytical Ultracentrifugation, Regensburg, 1997. 182. H. Cölfen, Prog. Colloid Polym. Sci. 113 (1999). Papers from the 11th Symposium on Analytical Ultracentrifugation, Potsdam, 1999. 183. W. Borchard and A. Straatmann, Prog. Colloid Polym. Sci. 119 (2002). Papers from the 12th Symposium on Analytical Ultracentrifugation, Duisburg, 2001.

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Applications of Electrodeposited Nanostructures G. Palumbo, J. L. McCrea Integran Technologies Inc., Toronto, Ontario, Canada

U. Erb University of Toronto, Toronto, Ontario, Canada

CONTENTS 1. Introduction 2. Synthesis and Structure of Electrodeposited Nanostructures 3. Properties of Electrodeposited Nanostructures 4. Applications of Electrodeposited Nanostructures 5. Summary Glossary References

1. INTRODUCTION The concept of a new class of materials with an ultrafine grain structure possessing unique properties was first introduced by Gleiter [1] about two decades ago. Originally referred to as “interfacial materials,” the main characteristic of these materials is the enhanced volume fraction of their interface component (i.e., grain boundaries, triple junctions, quadruple nodes [2, 3]) compared with their polycrystalline counterpart as a consequence of the ultrafine grain structure. With the large proportion of atoms located at interfacial regions, it was suggested that the atomic and electronic structures of such a material might be different from the structures found in compositionally equivalent material in the amorphous or polycrystalline states. Today these materials are referred to as nanocrystalline or nanostructured materials and are generally defined as materials with a characteristic length scale (e.g., grain size, particle size, film thickness, structure size) less than 100 nm. In the case of bulk nanostructured materials this length scale is the grain size. In recent years, bulk nanostructured materials have generated much excitement in

ISBN: 1-58883-057-8/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

the scientific community due to the many enhanced properties associated with the nanoscale grain structure, which have led to many new industrial applications. Many synthesis techniques have been developed over the past 20 years to produce bulk nanostructured materials, including: physical and chemical vapor phase processing, mechanical attrition, consolidation of nanophase powders, crystallization of amorphous precursors, as well as chemical and electrochemical methods (e.g., [4, 5]). A common feature of most synthesis techniques is that they operate far from equilibrium conditions to favor nucleation of new grains and reduce growth of existing grains, thus promoting a nanocrystalline grain structure. Unlike many other synthesis techniques, electrodeposition is a single-step process that produces fully dense material without the need for secondary consolidation of powders or annealing of amorphous precursors. The focus of this chapter presented here is on electrodeposited nanostructures. Over the past decade, electrodeposited nanostructures have advanced rapidly to commercial application because of the following factors: (1) an established industrial infrastructure (i.e., electroplating and electroforming industries), (2) a relatively low cost of application whereby nanomaterials can be produced by simple modification of bath chemistries and electrical parameters used in current plating and electroforming operations, (3) the capability in a single-step process to produce metals, alloys, and metal–matrix composites in various forms (i.e., coatings, freestanding complex shapes), and most importantly (4) the ability to produce fully dense nanostructures free of extraneous porosity. From the outset, the fully dense nanomaterials have displayed predictable material properties based upon their increased content of intercrystalline defects. This “predictability” in ultimate material performance has accelerated the adoption of electrodeposited nanomaterials by industry, whereby such

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 1: Pages (89–99)

90 extreme grain refinement simply represents another metallurgical tool for microstructural optimization. The development of a thorough understanding of the structure–property relationships of electrodeposited nanostructured materials over the years has enabled the “optimization” (rather than simple minimization) of the grain size of various nanostructured materials for specific industrial applications. In this chapter, current and emerging practical applications for electrodeposited nanostructured materials are presented and discussed in light of the enhanced properties relevant to the specific application.

2. SYNTHESIS AND STRUCTURE OF ELECTRODEPOSITED NANOSTRUCTURES Electrodeposition is likely the simplest method of producing nanostructured materials. However, the earliest systematic studies on the use of electrodeposition to produce nanocrystalline materials were only published in the late 1980s [6, 7]. Potentially, any metal or alloy can be modified to yield nanocrystalline deposits during the electrodeposition procedure. Important processing parameters include bath composition, pH, temperature, overpotential, bath additives, etc. To date, a large number of systems (pure metals, alloys, composites, and ceramics) have been electrodeposited with average grain sizes less than 100 nm. For example, the literature contains numerous examples giving electrochemical processing windows for the synthesis of nanocrystalline pure metals (e.g., Ni [8–11], Co [12], Pd [13], Cu [12]), binary alloys (e.g., Ni–P [6, 7], Ni–Fe [14, 15], Zn–Ni [16, 17], Pd–Fe [18], Co–W [19]) and ternary alloys (e.g., Ni–Fe– Cr [20–22], Co–Fe–P [23]). Even multilayered structures or compositionally modulated alloys (e.g., Cu–Pb [24], Cu–Ni [25, 26], Ag–Pd [27], Ni–P [28]), metal–matrix composites (e.g., Ni–SiC [10, 23, 29], Ni–Al2 O3 , Ni–MoS2 [23]), ceramics (e.g., ZrO2 [30]), and ceramic nanocomposites (e.g., Tla Pbb Oc [31]) have been successfully produced by electrodeposition methods. Electrodeposited bulk nanostructures, in the as-deposited condition, are fully dense materials, with negligible porosity (as determined by density [32] and position annihilation [33] measurements), unless deliberately produced in powder form. High-resolution electron microscopy has demonstrated that grain boundary structures in electrodeposited bulk nanostructures are similar to structures found in conventional polycrystalline materials [34]. Grain size measurements commonly show log normal grain size distributions [8] when the materials are deposited with an equiaxed grain structure. Average grain size and crystallographic texture can be controlled by various plating parameters, including but not limited to bath pH, current density, pulse parameters, and bath additions. Cross-sectional examination of thin (1 mm) coatings has shown that the nanocrystalline structure is fully established right at the interface with the substrate and that the grain size is independent of coating thickness [35], in contrast to conventionally produced electrodeposits which often show considerable grain coarsening with increasing coating thickness [36]. Binary and ternary alloys typically show extended

Applications of Electrodeposited Nanostructures

solid solubility ranges, compared to materials produced under thermodynamic equilibrium conditions (for example, in Ni–P [6, 7] Zn–Ni [16], Co–W [19]). Various coatings with thicknesses up to ∼100 m electroplated onto substrates to modify specific surface properties are probably the most widely known applications of electrodeposition technologies. It should be noted, however, that several other methods (e.g., brush plating, electrowinning, and electroforming) are used to produce bulk nanostructured materials as thick coatings (several mm thick) or as freestanding material in the form of sheet, foil, tubes, wire, mesh, plate, and foam. Table 1 lists examples of shapes and the corresponding applications of commonly produced electrodeposited and electroformed products.

3. PROPERTIES OF ELECTRODEPOSITED NANOSTRUCTURES Many detailed investigations have focused on identifying the important structure–property relationships of nanocrystalline materials, attempting to correlate the distinct structure of these materials with various physical, chemical, mechanical, electrical, and magnetic properties. For example, extensive studies of fully dense nanocrystalline nickel produced by electrodeposition have indicated that the structure– property relationships can be categorized into two basic groups: (i) those that exhibit a strong dependence on grain size and (ii) those with weak grain size dependence. A summary of the various grain size dependent and independent properties is given in Table 2 [37]. Detailed descriptions Table 1. Various shapes and applications of electrodeposited and electroformed products. Shapes Thin coatings (on substrates) Thick coatings (on substrates) Sheet, foil (freestanding)

Tubes, wire (freestanding) Mesh (freestanding) Plate (freestanding) Foam (freestanding) Molds (freestanding) Free forms

Powder

Applications surface modification for wear and corrosion resistance; catalytic surfaces electrosleeveTM ; repair of worn components gaskets; pressure control membranes; hydrogen purification membranes; thermal barriers; solar energy absorbers; microfoils; soft magnets surgical tools; missile guidance systems; miniature gamma radiation sources filters; precision sieve screens; razor foils; printing screens; centrifuge screens structural applications filters; electromagnetic shielding; battery electrodes; catalyst carriers CD stampers; embossing tools for holograms; compression, injection and pattern molds precision bellows; erosion shields for helicopters; trust chambers for rocket engines; components for micromagnetic motors, micro-optics, microactuators and microfiltration; shaped charge liners; precision reflectors and mirrors; nozzles catalysts; reinforcements

91

Applications of Electrodeposited Nanostructures Table 2. Grain size dependent and independent properties of electrodeposited nanocrystalline nickel.

are discussed along with the relevant material properties integral for the application.

Grain size dependent properties Property

Observation

4.1. Structural Applications

Solid solubility Hydrogen solubility Localized corrosion Corrosion potential

greatly enhanced greatly enhanced nearly eliminated shifted to more noble potential higher defect density

As would be expected from Hall–Petch considerations, numerous practical applications for nanocrystalline materials are based upon opportunities for high strength structural components. Table 3 displays mechanical properties of nanocrystalline nickel (99.99%) with grain sizes of 10 and 100 nm, in comparison to conventional polycrystalline (10 m) nickel. Remarkable increases in hardness, yield strength, and ultimate tensile strength are observed with decreasing grain size. It is also interesting to note that the work hardening coefficient decreases with decreasing grain size, to a negligible value at a grain size of 10 nm [38]. The ductility of the material decreases with decreasing grain size from 50% elongation to failure in tension for conventional material to 15% at 100 nm grain size and approximately 1% at 10 nm grain size. Ductility in bending was generally observed to be somewhat greater than in uniaxial tension. As is also shown in this table, the fatigue performance of the 100 nm material is fully consistent with that of the conventional Ni. Compared to conventional polycrystalline Ni, nanocrystalline Ni electrodeposits exhibit drastically reduced wear rates and lower coefficient of friction as determined in dry air pin-on-disk tests [37]. In a recent study on the tensile properties of electrodeposited nanocrystalline cobalt [39], the yield (0.2%) and ultimate tensile strengths for materials with an average grain size of approximately 12 nm were reported to be 1000–1100 and 1750–1950 MPa, respectively, while maintaining considerable ductility with elongation values ranging between 4 and 8%. For direct comparison, the yield (0.2%) and ultimate tensile strength for a coarse-grained polycrystalline cobalt counterpart with an average grain size larger than 1 m were reported to be 370–400 and 760–860 MPa, respectively, with elongation values between 7 and 18%.

Defect structure in passive layer Thermal stability Hydrogen diffusivity Coefficient of friction Wear resistance Strength Ductility Hardness Electrical resistivity

greatly reduced greatly enhanced reduced by a factor of 2 greatly enhanced increased by a factor of 3 to 10 greatly reduced increased by a factor of 5 increased by a factor of 3

Grain size independent properties Property

Observation

Bulk density Thermal expansion Young’s modulus Adhesion to substrate (for coatings) Thickness of passive layer

reduced by 900 >2000

50 — 207 140 0.4 241

>15 >40 214 300 0.15 275

1 — 204 650 0.0 —

1330

—

7.9

0.9

—

0.5

Source: “ASM Metals Handbook,” Vol. 2. ASM International, Materials Park, OH, 1993.

92

Applications of Electrodeposited Nanostructures

One of the first large-scale structural applications of a bulk nanostructured material is the use of an electrodeposited nanocrystalline nickel microalloy as an in-situ repair technique for nuclear steam generator tubing [40, 41]. This so-called electrosleeve process [42] has been successfully implemented in both Canadian and U.S. pressurized water reactors and has been incorporated as a standard procedure for pressure tubing repair [43], wherein a thick coating (0.5–1.0 mm) of nanostructured Ni microalloy (80–100 nm grain size) is electrodeposited on the inside surface of steam generator tubes to form a complete structural repair at sites where the structural integrity of the original tube has been compromised (e.g., corrosion, stress corrosion cracking, etc.). Figure 1 shows a cutaway of an alloy 600 nuclear steam generator tube repaired with the electrosleeve process. The high strength and good ductility of the 100 nm grain size microalloy allows for an appropriate thickness (0.5–1 mm) for which the impact on fluid flow and heat transfer in the steam generator tube is minimized.

4.1.2. Wear and Corrosion Resistant Coatings Also as a result of Hall–Petch strengthening, nanocrystalline materials display significant increases in hardness and strength relative to their coarser grained counterparts. Hardness increases on the order of 500% to 700% are typically observed [6, 44–49]. Deviations from Hall–Petch behavior have been noted at extremely fine grain sizes (i.e., 2 mM (1.4 mg/ml) and forms two types of fullerene–water colloidal systems: molecular-colloidal C60 solution in water (also called C60 FWS; here FWS is the abbreviation of “fullerene–water system”) [161–164] and typical monodisperse C60 hydrosol (also called ChH) [165]. C60 FWS can be produced by transferring fullerene from organic solution into the aqueous phase with the help of ultrasonic treatment [162], without using any stabilizers and chemical modification. It has been characterized as an aqueous molecular-colloidal solution having properties of both true solutions and colloidal systems simultaneously, which consists of both single hydrated fullerene molecules, C60 @{H2 O}n , and their fractal spherical clusters [161–164]. Tests of C60 FWS biological activity showed that these hydrated fullerenes appeared to be very promising in the context of their biological and therapeutical applications [166]. ChH is another monodisperse aqueous colloidal system with a particle size of 10 nm [165], it can be obtained by oxidation of the C60 anion with oxygen in “tetrahydrofuranwater” solutions. Both C60 FWS and ChH have the following similar characteristics: (i) high concentration (1 mg/ml and more); (ii) high stability, that is, preservation of their properties within a number of months even more under ambient conditions; (iii) almost identical color of solutions (from pale red to dark brown–red) that depends on C60 concentration; (iv) negative charge of colloidal particles; (v) preservation of the same properties in the case of dilution or concentration of both colloidal solutions; (vi) a similar ability to coagulate under the inorganic cation influence [166, 167]. In contrast to well-known reasons of stabilization of typical colloidal systems [168], the main mechanism of FWS stabilization is the hydration of fullerene molecules with formation of the supramolecular complex of C60 @{H2 O}n type [164] (see Fig. 9). The C60 @{H2 O}n clusters are stabilized by the weak donor–acceptor interactions of unpaired electrons of H2 O oxygen atoms with fullerene molecule to form -OH

 hydrogen bonds (such -OH

 hydrogen bonds have been found to possess about half the binding energy of -OH

O hydrogen bonds with, optimally, the -OH atoms centrally and vertically placed and the distances from the oxygen atom to the aromatic centroid of about 3.1–3.7 Å [169]) and by the formation of the ordered, H-bounded,

417

Figure 9. (a) A model of C60 {H2 O}80 cluster, obtained from an icosahedral water cluster with a C60 molecule taking the place of its inner water dodecahedron. (b) C60 molecules in aqueous solution form colloidal clusters based on 3.4-nm-sized icosahedral arrangements of 13 C60 molecules, where the C60 molecules are separated by water molecules [164]. The water network is formed by fully tessellated tetrahedral tricyclo decamer (H2 O)10 structures.

and spherelike H2 O molecular shells around the fullerene’s surface.

2.5. C60 on Surfaces 2.5.1. On Metal Surfaces C60 adsorpted on a noble-metal surface usually form strong ionic bonding via charge transfer [170–179] from the metal near-Fermi-level surface states to the fullerene 5t1u -derived LUMO. Quantities of charge transfer can be estimated indirectly from the following experimental results: (i) the shift of C60 vibrational energies in high-resolution electron-energyloss spectroscopy (HREELS) (Au(110) [171], Ni(110) [180], and Cu(111) [181]); (ii) the shift of carbon C1s corelevel absorption spectra using electron-energy-loss spectroscopy (EELS) (Au(110) [171]); (iii) using near-edge x-ray-absorption spectroscopy (NEXAFS) (Cu(111) [179]) compared to the known shift of alkalidoped compounds [182]; (iv) the anglur dependence in angle-resolved valenceband photoemission spectroscopy (PES) (Au(111) [183]); and (v) direct comparison of relative photoemission intensities (Ag(111) [175] and Cu(111) [179]). But for Al [184, 185] and Pt(111) [186] surfaces, experimental results by NEXAFS, HREELS, or PES show no charge transfer evidence and suggest the interaction of a covalent bonding. At the initial state the individual C60 molecules are readily mobile on most metal surfaces at room temperature, stabilized at the step edges on Au(110) [172], Au(111) [187], Ag(111) [188], and Cu(111) [189], and form twodimensional islands on Ag(110) [190]. On many low-index metal surfaces, C60 forms commen√ surate monolayers, such as Au(111) (23 × 3) [187, 191], Ag(111) [188], Ag(110) [190], Cu(111)-c(1 × 1) [192], and Au (100)-c(5 × 20), with uniaxial stress along the 110 direction [193]. But C60 adsorption on Au(110) [172] and Ni(110) [178] surfaces causes strong substrate reconstructions due to the strong interaction between the substrates and the C60 adsorbates. In the case of a high anisotropic Au(110)-p(1 × 2) surface, C60 molecules adsorbed on this

C60 -Based Materials

418 surface induce the Au surface atoms to form a (1 × 5) missing row reconstruction as a precursor for the p6 × 5
adsorbate structure and form a hexagonal close-packed corrugated layer [172]. Recent in-plane X-ray diffraction data suggested that a large fraction of Au surface atoms are displaced from their original positions producing microscopic pits to accommodate the C60 molecules [194]. In the case of Ni(110) surface, strong interaction between C60 and the surface causes a formation of added/missing Ni [001] rows which creates a corrugated structure resulting in formation of (100) microfacets and maximizing the C60 –Ni coordination [178]. It is suggested that the small separation between the LUMO of C60 and Ni Fermi level located in the d band leads to a strong hybridization interaction (i.e., covalent in character). Generally, the interaction between C60 and transition metal surfaces is stronger than that between C60 and noble metal surfaces [170, 178]. Images of the intramolecular features of adsorbed C60 molecules which are bias dependent have been studied by STM in C60 /Cu(111)-c4 × 4
[195] and C60 /Au(110) [196] systems. X-ray photoelectron diffraction has also been used to determine the molecular orientation of adsorbed C60 molecules [196–200]. For C60 monolayers on several common metal surfaces, the adsorptive structures as well as their interac√ tions √ are summarized in Table 5, and a typical 2 3 × 2 3
R30 model of the adsorptive structure on M(111) M = Au Ag Al
surfaces is shown in Figure 10 [185].

2.5.2. On Semiconductor Surfaces The initial state of adsorption of C60 on most common semiconductor surfaces shows individual isolated molecules, no island or specific interaction with step edge or defects on the surface. At room temperature the C60 molecules are generally immobile, showing strong interactions with the surface.

√ √ Figure 10. A model of the 2 3 × 2 3
R30 phase. The dashed hexagon encloses an overlayer-induced unit cell, while the solid hexagon encloses a unit cell of the M(111) surface (M = Au, Ag, Al). Adapted with permission from [185], A. J. Maxwell et al., Phys. Rev. B 57, 7312 (1998). © 1998, American Physical Society.

Si(111)-(7 × 7) The interaction of adsorbed C60 molecules with the Si surface is still not very clear. The Si(111)-(7 × 7) surface is (weakly) metallic due to the surface dangling bonds. Initially, it was suggested [205, 206] that the C60 – Si(111) interaction was ionic in character and similar to the C60 –noble metal interactions. HREELS data supported this

Table 5. Summary of C60 adsorption structures and interactions for several common metal surfaces. Surface Au (111)

Au (110) Ag (111) Ag (110) Cu (111) Cu (110) Ni (110) Al (111) Al (110) Pt (111)

Structure 38 × 38 in√phase’√ R14 2 3 × 2 3
R30 [188, 191] p6 × 5
[194, 172] reconstruction √ √ 2 3 × 2 3
R30 [188] c4 × 4
[190] c4 × 4
[192]

 10 0 [178] 1 3 Reconstruction [178] √ √ 6 × 6
, 2 3 × 2 3
R30 [185] Psedo-c4 × 4
[185]


√ √  13 × 13
R±13 9
 [186]

NDD (Å) ∼10 [188]

Interaction intermediate chemisorption, ionic

CT (eV) 0.8 [183]

10.04 [201] ∼10 [188]

intermediate chemisorption, ionic

0.75 [175]

intermediate chemisorption, ionic

1.6 [183] 1.5–2.0 [179]

WF of clean sample (eV)

WF with 1 ML C60 (eV)

770 [188]

5.3 [183]

4.7 [183]

800 [202]

5.37 [185]

4 82 ± 0 05 [185]

DT (K)

770 [188]

10.07, 10.23 [190] 10.2 [192, 179] 9.7–11.1 [178] 10.5, 10.0 [178]

strong mainly covalent

9.91, ∼10 04 [185]

intermediate mainly covalent [185]

2 ± 1 [180]

9.91, 11.44, 12.13 [185] 10 0 ± 0 3 [186]

mainly covalent [186]

700 cm−1
, the technique has a significant advantage over INS, since the much sharper peaks can provide very accurate data for mode frequencies [346]. By investigating the vibrational fine structure in the fluorescence [347] and phosphorescence [348] spectra of C60 molecules isolated in neon and argon matrices, the unambiguous identification of several vibrational modes of T3u , Hu , and Gu symmetry were obtained.

3.3. Phase Transition 3.3.1. Orientational Phase Transition On cooling below a temperature of about T01 ∼ 260 K, the nearly free rotation of C60 molecules in solid C60 will be frozen by losing two of their three degrees of rotational freedom; the four neighboring molecules in an fcc unit cell remain in their positions but become orientationally inequivalent. Their rotation motions are constrained along two standard favorite orientations and become hindered, resulting in an orientational ordering [235, 243, 249, 273, 349], but the random reorientation between the two possible standard orientations for each molecule in the solid still lead to a residue disorder (a so-called “merohedral disorder”) [32, 243, 274, 275, 349–352]. The two favorite orientations are corresponding to electron-rich double bonds on one molecule facing the electron deficient centers of hexagons (HF) and pentagons (PF) respectively on its adjacent molecules, obtained by rotating the molecules around the four 111 axes of the C60 molecules through about 38 and 98 (60 jumps) respectively [243, 284, 349, 353–355]. Recently, elastic diffuse neutron-scattering study, considering the optimization of nearest-neighbor orientations rather than long-range orientational order, showed that the actual angles may be closer to 42 and 102 [356]. The PF structure has been indicated to be an idealized ordered structure and the HF structure a “defect structure,” because the HF structure has a slightly higher energy than the PF structure. The lower energy configuration is about only 11.4 meV [352, 357] below the higher energy configuration, and the two are

C60 -Based Materials

separated by a potential barrier of ∼235–295 meV [243, 274, 280, 284, 349, 352, 357–361] which has been confirmed by various experiments. A schematic 3D model of a close-packed (111) plane of C60 molecules exhibiting uniaxial reorientation about 111 directions (low temperature phase, 90–260 K) is shown in Figure 20 [243]. The molecules reorientate between a number of different sites but principally perform 60 jumps about [111] residing in the two configurations corresponding to 38 and 98 (or 42 and 102 , using latest data) rotations. Although the centers of the C60 molecules remain in the same places, the four molecules of the conventional fcc unit cell become orientationally nonequivalent. The incompatibility of the icosahedral molecular symmetry of the different molecular orientations and the cubic lattice symmetry lowers the crystal symmetry from fcc into a simple cubic (sc) ¯ which is a subgroup of the structure (space group Th6 or Pa3, ¯ of an fcc structure) [235, 273, 362]. Thus space group Fm3m the C60 crystal undergoes a so-called orientational ordering phase transition. The orientational ordering transition is first order [111]. The occurrence of this transition has been confirmed by various experimental methods such as differential scanning calorimetry (DSC) [235, 273, 363], X-ray diffraction [235, 273], NMR [14, 23, 274–276, 363–366], neutron diffraction [243], dielectric spectroscopy [359]; SR [279]; specific heat measurements [352, 361, 367, 368], inelastic neutron scattering [33, 277, 369, 370], electron diffraction [371], sound velocity and ultrasonic attenuation [242, 372], and Raman spectroscopy [326, 373–375]. Figure 21 shows a comparative electron diffraction patterns of crystalline C60 at room temperature and at 100 K [371]. Additional spots appear at positions which are extinct in an fcc lattice on cooling below T01 and correspond to a primitive cubic lattice. The appearance of these spots reveals the low temperature structure in which the molecules are still situated on an fcc lattice but orientationally ordered and also shows that the space group Pa3¯ is a subgroup of the ¯ of the high temperature fcc phase. space group Fm3m

Figure 20. Close-packed (111) plane of C60 molecules depicting uniaxial reorientation about 111 directions (low temperature phase, 90– 260 K). The “dimples” correspond to the centers of electron-poor hexagonal/pentagonal faces; the equatorial protrusions depict electronrich 6:6 bonds. The constraints demanded by Pa3¯ symmetry imply that the electron-rich and electron-poor regions have well-defined loci corresponding to constant latitudes relative to [111]. These electrostatic considerations are consistent with the assumption that the preferred easy reorientation direction is [111]. Adapted with permission from [243], W. I. F. David et al., Europhys. Lett. 18, 219 (1992). © 1992, EDP Science.

425

Figure 21. Electron diffraction patterns of C60 crystallite along [001] (a, c) and along [011] (b, d) at room temperature (a, b) and at 100 K (c, d). Note the appearance of extra reflections at low temperature related to a lowering of symmetry due to orientation ordering. Adapted with permission from [371], G. Van Tendeloo and S. Amelinckx, in “Electron Microscopy of Fullerenes and Related Materials, Characterization of Nanophase Materials” (Z. L. Wang, Ed.), Wiley–VCH, 2000. © 2000, Wiley–VCH.

Figure 22 shows the 13 C NMR spectra of solid C60 at temperatures from 295 down to 77 K [275]. The lowtemperature NMR line shape for the C60 solid is quite different from that of C60 in solution. From RT to low temperature below Tg , the spectra changes from a single line peak at 298 K to a line shape at 77 K. The symmetry of the molecule and the equivalence of all C60 carbon atoms in C60 crystals are broken by two factors: the presence of the crystal lattice and the applied magnetic field. As the temperature is reduced, the “motionally narrowed” NMR line at 295 K (143 ppm from tetramethyl silane) in the 13C NMR spectrum of C60 , which proves that C60 molecules are rotating rapidly with respect to the NMR time scale at RT, is broadened and the C60 molecule’s orientation become static on the NMR time scale. At 77 K there is little evidence of the narrow line at 143 ppm, and a fit to this line shape yields an asymmetric chemical-shift tensor with components of 220, 186, and 25 ppm, reflecting the chemical-shift anisotropy of the 13 C nuclei averaged over the crystallite orientations. The phase transition affects various physical properties; the changes of some of these properties at the transition temperature T01 are listed in Table 9. It has also been indicated that both the transition temperature and the size and shape of the transition anomalies are strongly influenced by deformation, impurities, and solvent [354, 378].

3.3.2. Glass Transition Below about 90 K the molecules in C60 crystal are entirely frozen but never order perfectly; the C60 crystal undergoes a so-called glass transition. The glass transition of C60 crystal is second order [243]. The fraction of molecules in the more stable PF orientation is only about 60% near 260 K but increases to

C60 -Based Materials

426

Table 9. Physical properties at orientational phase transition temperature T01 . Quantity

about 84% near Tg ∼ 90 K, the glass transition temperature [243]. Below Tg , the thermal energy becomes too small compared with the energy barrier between the two states for further reorientation to occur; therefore the C60 molecules are frozen in their orientation and the fraction of PF orientation (∼84%) or HF orientation (∼16%) becomes invariable. The remaining orientational disorder causes an orientational glass phase. Because the nearest-neighbor distance of HF oriented C60 molecules is slightly smaller than that of the PF oriented molecules, there is an anomalous increase of volume thermal expansion coefficient in the transition temperature [243]. Evidence for this glass transition at low temperature has been provided by a number of experimental techniques, including velocity of sound and ultrasonic attenuation

Above T01

Jump

Structure

sc

fcc

—

Space group

P a3¯ Th6


¯ (O 5
Fm3m h

—

Lattice constant

14.11 Å

14.15 Å

0.04 Å [238, 243]

Reorientational correlation time of C60 molecules Volume Thermal conductivity Enthalpy change

2 × 10−9 s

9 2 × 10−12 s

∼200 times [274]

small 0.5 W/m K

large 0.4 W/m K

∼1% [243] 25% [352]

low

high

high small

low large

9000 J/kg (6500 J/mol) [368] 12,500 J/kg (9000 J/mol) [376] 8% [242] 1.2% [377]

Young’s modulus Magnetic susceptibility

Figure 22. 13 C NMR spectra of solid C60 obtained at ambient temperatures of 123, 100, and 77 K. The 13C NMR spectrum of C60 is expected to be a broad “powder pattern” resonance (width greater than ∼200 ppm) reflecting the chemical-shift anisotropy of the 13 C nuclei averaged over the crystallite orientations. The “motionally narrowed” NMR line at 295 K (143 ppm from tetramethyl silane) proves that C60 molecules are rotating rapidly with respect to the NMR time scale, which is defined by the inverse of the spectral spread (∼10−4 s). As the temperature is reduced, the powder pattern is recovered as the molecular motion decreases. At 77 K there is little evidence of the narrow line at 143 ppm, and a fit to this line shape yields an asymmetric chemicalshift tensor with components of 220, 186, and 25 ppm. Adapted with permission from [275], C. S. Yannoni et al., J. Phys. Chem. 95, 9 (1991). © 1991, American Chemical Society.

Under T01

∼4 56 10−9 in cgs)

∼4 61 10−9 in cgs)

studies [242, 372], specific heat measurements [361, 379], thermal conductivity [352, 380], elasticity [242, 358, 372], high resolution capacitance dilatometry [357], dielectric relaxation studies [359], electron microscopy observations [381], and neutron scattering measurements [243, 349]. It has been demonstrated that the measured values of transition temperature Tg varying in a wide range (∼90–165 K) rely strongly on the time scale of the measurements [358]. The temperature dependence of the relaxation time may be described by the relation "t = "t0 expEAt /kB T
, where EAt is the activation energy for the transition and "t0 is a characteristic relaxation time with values of "t0 = 4 ± 2 × 10−14 s and EAt = 300 ± 10 meV [358]. For high-frequency measurement probes, the molecules cannot reorient quickly enough to follow the probe frequency, so that the molecules seem to be frozen at a higher transition temperature and a lower transition temperature for low-frequency measurement probes. It is concluded that below Tg a quenched disorder occurs rather than an equilibrium glass phase [358].

3.3.3. Phase Transition in Low-Dimensional Structures At room temperature, the molecules on C60 (111) surfaces are hexagonal close-packed and rotating and reorientating freely corresponding to a 1 × 1
structure. Below a temperature of surface orientational ordering transition temperature Ts01 about 230 K [382–386] (about 30 K lower than that of C60 bulk), the C60 (111) surface exhibits a (2 × 2) reconstruction [382, 387] due to the four inequivalent frozen molecules in a (2 × 2) unit cell holding four different orientations. STM study clearly revealed the surface structure and indicated that the orientational configurations of the C60 molecules on the surface are similar to those in the bulk, except for some complex details. This similarity indicates that it is the

C60 -Based Materials

interaction of the neighboring C60 molecules that drives the molecules to an orientational ordered state [387]. Recently, a two-stage rotational disordering mechanism, with a new intermediate regime between a low-temperature ordered (2 × 2) state and a high-temperature (1 × 1) disordered phase, has been proposed for this surface phase transition based on Monte Carlo simulations [388]. C60 submonolayers adsorbed on a self-assembled monolayer of an alkylthiol can be considered as ideal twodimensional hexagonal arrays with a nearest-neighbor distance of ∼10 Å [20] due to the weak van der Waals force between C60 and the organic monolayer surface. At room temperature the C60 molecules at this surface are mobile and rotating freely. At 77 K, the C60 molecule appears as a hemisphere, a tilted doughnut, or an asymmetric dumb-bell, each being consistent with a rotating pattern around a fixed axis. That is, unlike in the case of bulk C60 , the two-dimensional rotationally ordered phase persists down to 77 K (already 13 K below the bulk freezing temperature) [20]. Unlike the orientation-related glass phase in bulk C60 and the bulk C60 (111) surface, the two-dimensional C60 array forms orientationally ordered domains at 5 K although the C60 molecules are not commensurate with alkylthiol monolayer [20]. The molecular rotation is frozen and the internal fine pattern of C60 with a well-known native cage structure can be clearly revealed by STM, because of the fairly weak interaction between C60 and the organic monolayer surface with a negligible influence on the C60 molecule by the substrate.

427

Figure 23. Phase diagram of C60 drawn on a logarithmic pressure scale to show the low pressure range: ( ), data from experiment; (- - -), theoretical curves or extrapolations from experiment. Adapted with permission from [389], D. M. Poirier et al., Phys. Rev. B 51, 1830 (1995). © 1995, American Physical Society.

3.4. Crystal Growth 3.4.1. Single Crystal

3.3.4. Phase Diagram Figure 23 gives an equilibrium phase diagram of C60 crystal covering a wide range of temperatures and pressures, drawn on logarithmic scales to bring out clearly the low-pressure behavior [389]. The solid lines are derived from experiment, and dashed lines are theoretical or reflect extrapolations from experiment. The solid–vapor coexistence line P T
is given by plotting the equilibrium vapor pressure against temperature. The thermal gravimetric analysis data (shown by solid circles) [247] are fitted to P T
= P0 exp−5H /kT
, which can be derived from the Clausius–Clapeyron equation assuming ideal gas behavior [390]; the low pressure experimental data (shown by a bar and a open diamand) [389, 391] are in good agreement with extrapolation of the solid–vapor coexistence line using the ideal gas relationship. It can be indicated from the diagram that solid fcc C60 sublimes on heating without forming a liquid phase at normal pressures. The fcc–sc coexistence line at the left of the phase diagram defines the pressure-dependent temperature of the solid-state orientational ordering phase transition; the data shown by filled squares are measured using DSC [244]. At high temperature or under ultrahigh pressure, the C60 molecules will collapse into amorphous carbon. Under higher pressure (>1 GPa) and high temperature, which is not shown in the phase diagram, crystal C60 will polymerize, induced by pressure. The pressure-induced polymerization will be introduced in Section 4.2.

Solvent grown C60 crystals may form various structures due to the intercalation of solvent molecules or small atmospheric molecules into C60 crystal lattices [392–395], such as hexagonal close packed (hcp) structure [2, 396], fcc structure [397], orthorhombic or monoclinic structure [398, 399], and so on, and various shapes, even decagonal crystals with pseudo tenfold symmetry, can formed [392, 400, 401]. C60 has an extraordinarily high vapor pressure of strong temperature dependence [246] (at 500  C, the vapor pressure of C60 is about 1035 times greater than that of the graphite [247]) and solid C60 sublimes at relatively low temperature (see Table 4) [247]. A vapor transport method is used to grow pure C60 single crystal free of solvent or gas in a large quantity and a perfect quality conveniently by slow condensation or deposition of the molecules from vapor [402]. See Table 10. The vapor growth technique for synthesizing single-crystal fullerenes usually uses a carefully cleaned, dynamically Table 10. Available rate of weight loss and vapor pressure of C60 at selected temperatures [247]. Temperature ( C) 400 450 500 550 600

dw/dt ( g/s)

Vapor pressure (mTorr)

0 004 0 014 0 085 0 43 1 75

0 019 0 069 0 446 2 32 9 67

428 pumped, horizontal quartz tube (usually about 50 cm long by 1 cm in diameter) placed in a furnace with a temperature gradient [403–405] (see Fig. 24). The initial C60 powder is filled in a gold boat and heated in the vacuum furnace at about 250  C under dynamic vacuum for 6–72 hours in order to remove the solvent intercalated in the powder. Subsequently the Au boat is heated above the C60 sublimation temperature to about 590–620  C, and high-purity, dry helium flow kept at a flow rate below 5 cc/min is employed to carry the C60 vapor to the colder part of the quartz tube. Crystallization starts at the part where the temperature is below 520  C. The growth rate in this method is expected to be about 15 mg per day. The vapor transport procedure is always kept for several days to yield larger C60 crystals. Several modified techniques based on the vapor transport growth method have been developed to grow larger and more perfect C60 crystals such as the vibrating temperature method [406], the double temperature gradient method [407], optimization of temperature distribution [408], use of a vectical furnace with a pulling technique [409], the Pizzarello method [410, 411], very small temperature gradient control [412], and so on.

3.4.2. Thin Films High-quality crystalline C60 thin films have been grown successfully on many substrates and studied extensively for the understanding of growth mechanisms and properties. The potential applications of C60 thin films in electronic devices have been explored, such as rectifying electronic diodes [413], photodiodes and solar cells [413], and thin-film field-effect transistors [414–416]. The results showed those fullerene electronic devices have stable, reproducible performances and excellent device characteristics. Most of the C60 thin films are grown under ultrahigh vacuum (UHV) environment using a sublimation-deposition method, such as a simple or enhanced vapor deposition technique [417–419], organic molecular beam epitaxy [420], the hot-wall diffusion method, or the ionized cluster beam deposition technique [421–425]. A schematic sketch of a universal hot-wall diffusion method setup is shown in Figure 25. The diffusion oven (lower part of Fig. 25) employed is composed of three parts. The temperature of each part can be controlled independently to get three different temperature zones (upper part of Fig. 25). The source C60 powder and substrate are placed in zone I and zone III in a quartz tube, respectively. The C60 molecules are evaporated from zone I, through zone II, then deposited on the substrate in zone III. The temperature of zone II (hot wall) is higher than the other two zones, which can generate a very low growth rate. As an example, the temperatures of zone I and zone II are

Figure 24. Schematic view of a horizontal quartz tube furnace apparatus for the vapor transport growth of C60 single crystals.

C60 -Based Materials

Figure 25. A sketch of diffusion oven composed of three independently temperature-controlled zones; T1  T2 , and T3 are the temperatures of the C60 source: “hot wall” and substrate, respectively. Adapted with permission from [426], J. G. Hou et al., Thin Solid Films 320, 179 (1998). © 1998, Elsevier Science.

set to T1 = 420  C and T2 = 440  C, respectively [426]. The substrate temperature (T3
is usually set from 100 to 200  C. Most of well-ordered C60 thin films or multilayers were grown on substrates including layered substrates, alkali halides surfaces, semiconductor surfaces, and metal surfaces. But studies showed that crystalline films can grown relatively easily on layered substrates like mica [417, 427–439], graphite [386, 440, 441], MoS2 [437–439, 442, 443], fluorophlogopite [444], and lamellar substrates such as GeS(001) [445–448], GaSe(0001) [341, 343], and Sb [449]. Because of the natural passivation of these substrate surfaces, the C60 –substrate interactions are dominated by van der Waals force which can relax the stringent lattice match condition and result in the formation of fcc (111) oriented C60 crystal grains. Well-ordered crystal C60 monolayers and thin films can also be grown on alkali halides and other ion crystals such as NaCl [417, 418, 437, 438, 450, 451], KCl [418, 438, 450, 451], KBr [418, 438, 450–453], KI [245, 246], and CaF2 [247, 248], due to the very weak surface bonding dominated by the electrostatic force that restricts the diffusion of C60 ad-species. But C60 thin films grown on metal substrates are much different from the weak-bonding substrates mentioned. It has been shown that there exist very strong interfacial interactions and charge transfer from the metal substrate to the LUMO of C60 [172, 191]. Compared with the nonmetal substrates, the metal substrates are usually less smooth and contain much higher densities of steps, so it is more difficult to grow high-quality single crystal films on metal surfaces. There have been reported high-quality C60 (111) oriented films successfully grown on Au [191, 419, 458], Ag [191, 419, 459, 460], Cu(111) [191, 461, 462], as well as Ni3 Fe(111), Ni3 Co(111), and Ni3 Fe(110) [426, 463] using a hot-wall diffusion method. Several techniques have been attempted to improve the crystallization and quality of C60 thin films. Using a self-mediated process [451] or using Sb as the surfactant or buffer layer [464, 465] and the multistep growth method [418], higher quality and larger grain-sized C60 can be obtained. Figure 26 shows a transmission electron microscopy (TEM) image as well as diffraction pattern of a high-quality single crystal C60 fcc (111) oriented film with a thickness of about 60 nm prepared by using Sb as the surfactant on a freshly cleaved (001) NaCl single crystal surface [465].

C60 -Based Materials

Figure 26. TEM image and diffraction pattern of a C60 film with nominal thickness of 60 nm. Large (111) orientated C60 grains with average size of about a few micrometers were formed. Adapted with permission from [465], J. G. Hou et al., J. Appl. Phys. 84, 2906 (1998). © 1998, American Institute of Physics.

3.5. Properties 3.5.1. Electronic Structure C60 solid is a semiconductor with a minimum of the energy gap at the X point of the Brillouin zone. The value of the energy gap is still under debate, which is in the range of 1.43–2.35 eV. In early studies, a typical and widely referenced band calculation for solid C60 is based upon the local density approximation (LDA) in density functional theory, in which all many-body effects are collected into the exchange-correlation energy, which is evaluated within a free-electron model. The LDA calculations obtained a bandgap of ∼1.5 eV for fcc C60 crystals [466] and a value of 1.18 eV for the bandgap at the X point [467]. The band structure of C60 crystals is derivated from the energy level distribution of an isolated C60 molecule. Figure 27 shows the electronic structure of the HOMO- and LUMOderived bands of C60 in the solid state obtained from LDA calculations.

Figure 27. Left-hand panel; a schematic diagram of the C60 -molecular orbit energy scheme. The HOMO (h1u
and LUMO (t1u
are shown in the expanded region. Central panel; the band structure of the HOMO- and LUMO-derived bands of C60 in the solid state. Righthand panel; the density of states of the HOMO- and LUMO-derived bands of solid C60 predicted from LDA calculations [466]. Adapted with permission from [468], M. Knupfer, Surf. Sci. Rep. 42, 1 (2001). © 2001, Elsevier Science.

429 By properly treating the electron excitations with an ab initio quasiparticle approach, a GW (G represents Green’s function and W is a dynamically screened Coulomb interaction) approach was developed. In the GW approach, the Dyson equation is solved within a Green’s function formalism using a dynamically screened Coulomb interaction W to obtain the quasiparticle energy spectrum. Because the method includes a treatment of the electron self-energy (in terms of a dynamically screened Coulomb interaction), it predicted a larger value of 2.15 eV [469] for the HOMO– LUMO gap, with bandwidths of 0.9, 0.7, and 0.8 eV for the HOMO-, LUMO-, and (LUMO + 1)-derived bands, respectively. This result is very close to the gap value of ∼2.3 eV given by high-energy spectroscopy experiments like PES and inverse photoemission spectroscopy (IPES) [470–472]. There are also many other experiments carried out by different methods showing different results [473]. A scheme in Figure 28 shows the difference of the gap values given by different methods. A detailed electronic structure of C60 thin films was studied by surface photovoltage spectroscopy [474, 483, 484]. It is illustrated in Figure 29, which includes a mobility gap of 2.25 eV and exponential Urbach tails of states determining the optical gap of 1.65 eV [481, 485]. In addition, there exist two deep gap states: one donor and one acceptor. The acceptor level lies 0.8 eV below the conduction band tail, and the donor level is 1.25 eV higher than the valence band tail. The exponential Urbach tails in the density of gap states may be due to thermal disorder rather than from static structure, topological, or compositional disorder. The donor level arises from the unbound intercalated oxygen and the acceptor level is considered to be the result of chemical reaction of oxygen with fullerene. This detailed electronic structure shows that the optical absorption edge [481] is not simply connected to the HOMO–LUMO gap.

Figure 28. Energy gap of C60 fullerene, obtained in various studies, including: surface photovoltage spectroscopy [474], photoemission [471], EELS [475], microwave conductivity [476], ac photoconductivity [477], real part of complex conductivity [478], photoconductivity [479], optical absorption [480] (the upper one), optical absorption [481] (the lower one), quasiparticle approach [482], and LDA calculations [466]. Adapted with permission from [473], T. L. Makarova, Semiconductors 35, 243 (2001). © 2001, MAIK Nauka/Interperiodika.

C60 -Based Materials

430

Figure 29. Electronic structure of C60 thin films, from which we can see exponential tails of states between conduction mobility edge and conduction band bottom, as well as tails between valence band top and valence mobility edge. Adapted with permission from [474], B. Mishori et al., Solid State Commun. 102, 489 (1997). © 1997, Elsevier Science.

To get a complete description of the electronic structure of C60 solids, a number of many-body calculations have focused on calculations of quantities of electron correlation effects. Direct in-situ comparison of C 1s absorption spectra taken in the gas phase with those from solid C60 reveals a close similarity [486]. It can be concluded that solid-state interactions are not important in this new material, and the electron correlation effects are mainly intramolecular. So the Coulomb repulsion (U
of two holes on the same molecule in solid C60 is an important quantity, which is nearly independent of the molecular orbital of the hole with a value of 1.6 eV [471]. This value of U leads to Frenkeltype molecular excitons with energy of about 1.5–2 eV corresponding to hu –t1u intramolecular excitations which correspond to the lowest-energy-observed optical transitions of electronic origin. The calculated values of such Frenkel excitons are 1.58 and 1.30, with experimentally measured values of 1.83 and 1.55 eV [487, 488].

3.5.2. Electrical Transport Properties At room temperature, pristine C60 solids are n-type semiconductors with a high resistivity [415, 489]. The electron mobility is considerably larger than the hole mobility. The electron and hole drift mobilities are 1.3 and 2 × 10−4 cm2 /V s, respectively, and the recombination lifetime is 1 7 × 10−6 s [490]. The dc conductivities in C60 follow a purely thermally activated behavior. The conductivity of the crystalline C60 is obviously higher than that of the amorphous films. The dc conductivity of C60 single crystal is 1 7 × 10−8 ( cm)−1 and the activation energy is 0.581 eV [491]. But for polycrystalline films, the conductivity and activation energy are reported to be 10−6 –10−8 ( cm)−1 and 0.3– 0.6 eV [492–496], respectively. For amorphous films, these values are within the intervals 10−7 –10−14 ( cm)−1 and 0.5– 1.1 eV [479, 497–499]. With improving crystallinity of C60 films, their conductivity increases, and the activation energy falls [500, 501]. The thermally activated conductivity in fullerene systems [493, 502] obeys the Meyer–Neldel (MN) rule [503]. The MN rule describes an exponential relation between the activation energy Ea and the pre-exponential factor 0 . The expression of the MN rule for the dc conductivity can be

written as  = 0 exp−Ea /kT), and the prefactor 0 correlates with the activation energy Ea as 0 = 00 exp−Ea /kT), where 00 and T0 are the Meyer–Neldel parameters. For example, the relationship between the conductivity prefactor 0 and activation energy Ea in C60 films under different O2 exposure can be described by the MN rule. The obtained 00 and kT0 are 15.6 ( cm)−1 and 0.11 eV, respectively [493]. The relationship between the conductivity prefactor and activation energy in C60 films at different stages of the growth process also obeys the MN rule with the MN parameters of 00 = 7 7 × 10−19 ( cm)−1 and kT0 = 0 012 eV [493]. High pressure has a strong effect on the conductivity, as well as the width of the gap between LUMO- and HOMO-derived bands of C60 solids [504, 505]. The electric conductivity changes four orders of magnitude from 2 × 10−6  cm)−1 (below 8 GPa) to 3 × 10−2 ( cm)−1 (around 20 GPa). The bandgap is as follows: 1 6 ± 0 1 eV at 3 and 6 GPa, and 1 2 ± 0 1 eV at 10 GPa. Exposed to air, upon the interaction with oxygen, the conductivity of C60 single crystals [506] and films falls [493, 507] by 3–6 orders of magnitude. Such an effect may be caused by trap levels for carriers and neutralizes defects forming localized electronic states, which are created by the intercalated oxygen. The conduction mechanism at different temperatures has been studied by measuring the ac (1 < f < 50 kHz) conductance of C60 polycrystals [508]. In the temperature range of 100–150 K, the ac conductance linearly depends on the temperature and nearly has a power law of 0 8 dependence on the frequency, which is characteristic of the model of the variable range hopping between midgap localized states around the Fermi level [509]. In this model, the ac conductance can be described as  T
∼ e2 kT 9N EF
:2 '−5 9ln;ph
/:4 . Here, N EF
is the density of states at the Fermi level, ;ph is the phonon frequency, and ' is a constant. In the temperature range of 200–350 K, the frequency dependence of conductance at a fixed temperature can also be roughly described as a power law  ∼ s s = 0 8
. However, in this temperature range (as the frequency varied from 1 to 10 kHz), the ac conductance increases rapidly with temperature and shows a thermal activated behavior with an activation energy Ea of 0.389 eV below 265 K and 0.104 eV above it. This is proposed to be caused by the hopping process among the tails of localized states, which can be described as T
∝ exp−Ea /kT). The two activation energies can be ascribed to the coexistence of the crystalline and amorphous phase of C60 . At very high temperatures approaching the sublimation temperature of about 430  C, experiments suggest that the mechanism of conductivity of C60 changes from hopping conductivity to band conductivity (charge transfer over delocalized states) [510–512]. From these results, it can be concluded that the C60 solid conductivity can be described in terms of the conduction mechanism of disordered semiconductors.

3.5.3. Optical Properties Optical Absorption and Photoluminescence Spectroscopies The optical properties of solid C60 in the visible-UV range were investigated by a variable angle spectroscopic ellipsometry (VASE) technique [513–515] and

C60 -Based Materials

near-normal-incidence reflection and transmission experiments [516]. UV data above ∼7 eV were obtained using EELS [475] by Kramers–Kronig analysis of the EELS loss function. The observed optical properties of C60 solids are usually expressed in terms of the complex optical dielectric function #
= #1 
+ i#2 
. Figure 30 shows the results for #1 
and #2 
, obtained from VASE and transmissionFTIR (Fourier transform infrared) studies on thin solid films of C60 on KBr substrate at T = 300 K [316]. The strong, sharp structure at low energy (below 0.5 eV) is identified with infrared-active optic phonons. The imaginary part of the dielectric constants of C60 begins to rise at 1.7 eV, which can be considered the absorption edge. At higher energies in the visible and UV ranges, the spectrum structure is due to allowed optical transitions at 3.5–5.6 eV and to excitons at energies below 3 eV. There are four prominent absorption peaks at ∼2.7, 3.6, 4.7, and 5.6 eV for the pristine C60 with dipole-allowed electronic transition [480, 481, 517–519]. Optical transitions between the HOMO and LUMO bands are forbidden by symmetry [520]. Luminescence spectra appear to be more sensitive to sample quality and to show wider sample to sample variation than absorption spectra [521] The temperature dependence of the photoluminescence spectra is given in [522]. At a temperature of 7 K, the absorption curve intersects the photoluminescence curve at the point 1.8 eV, which may be the optical gap at this temperature [523]. Transition Assignment An accurate study of optical transition in C60 was carried out by comparing the optical spectra of C60 films on mica with the transmission of toluene, hexane, and heptane solutions [480]. Experimental spectra were fitted with Gauss–Lorentz (GL) line shapes to determine the energy positions of individual transitions. The assignment of the electronic transitions given in Table 11 is based on the calculated transition energies of [524–526]. The best agreement of the calculated gap positions with experimental results is provided by the tightbinding models [525, 526]. An overview of the energy levels and optical transitions discussed is shown in Figure 31. The three leading GL bands D E + F , and G are assigned to the transitions hg  gg → t1u  hu → hg , and hg  gg → t2u ,

Figure 30. Real and imaginary parts of the dielectric function #) determined from VASE and FTIR measurements for C60 . Adapted with permission from [520], M. S. Dresselhaus et al., Synthetic Metals 78, 313 (1996). © 1996, Elsevier Science.

431 Table 11. Spectral features for solid C60 and C60 molecules. All energies are given in eV [480]. Band code

Solid C60

C60 molecules

Transition assignment

1 918

1 995 2 035 2 070 2 105 2 180

hu → t1u + Tu , Hu , Gu (+Hg , Ag )

=0 =1 =2 =3 =5

1 992 2 028 2 097

A B C

2 41 2 70 3 2

D1 D2

3 489 3 541

3 58 3 732

hg , gg → t1u

E F1 F2

3 99 4 36 4 546

4 21

hu → hg

G1 G2

5 500 5 77

5 437 5 73

hu → t1g

4 6 hg , gg → t2u

respectively [480, 513]. The oscillator strengths of these dipole-allowed transitions are in reasonable agreement with the calculations of [525]. The D band assigned to hg  gg → t1u transitions is strongly reduced in doped films due to the filling of the lowest state in the conduction band derived from the t1u molecular states [527]. The molecular F1 2 band, which originates from the hu → hg transition, split into F 1 and F 2 bands in the solid due to the splitting of the fivefold-degenerate hu hg
levels to the threefoldand twofold-degenerate tu tg
and eu eg
levels, respectively [467]. The weak E band possibly also originates from the hu → hg transition. The assignment of the two lowest transitions, hu → t1u and hu → t1g , is more delicate. The t1g h−1 u molecular state consists of electron–hole excited states with T1u  T2u  Hu , and Gu symmetry [528]. According to several calculations, the lowest allowed transition hu → t1g to the T1u excited state should be located near 3 eV with an oscillator strength of about 3% of that of the hg  gg → t1u band at 3.5 eV

Figure 31. Electronic energy levels in solid C60 and in C60 solution in n-hexane. Adapted with permission from [480], J. Hora et al., Phys. Rev. B 54, 5106 (1996). © 1996, American Physical Society.

432 [524, 525, 529]; the very small oscillator strength has been attributed to plasmon screening [525]. In addition to this allowed transition, phononinduced transitions of comparable strengths [529] to the T2u  Hu , and Gu excited states [528] should appear in the same energy region. The comparison of the energies and oscillator strengths of the B and D bands is essential for the following assignment of = B B ′ , and D subbands to the electronic transitions (see Table 11). The A-group is attributed to the t1g h−1 u electron–hole state parity forbidden in an isolated molecule, but becoming partially allowed because of level splitting. The =-group results from the forbidden molecular transition hu → t1u . These transitions gain nonzero strength through the excitation of an appropriate odd-parity vibrational mode [530] (Hertzberg–Teller coupling) and their upper electronic states can be influenced by Jahn–Teller dynamic distortions [518]. Photoconductivity Experiments confirmed the existence of the photoconductivity in fullerene solids [531, 532]. Photoconductivity spectra of C60 are, on the whole, similar to the absorption spectra, apart from the conduction dip at about 350 nm of wavelength. Such a photoconduction dip appears frequently in intense absorption regions in many photoconductors. The photoconduction edge is at about 730 nm (1.7 eV) [523]. The main spectra features are related to excitons rather than to interband transitions. The photoconduction spectra measured at various light-intensity modulation frequencies, light intensities, and applied voltages shows that the photocurrent is proportional to light intensity and applied voltage. Moreover, the photocurrent increases with decreasing light-intensity modulation frequency, which may arise from some deep carrier traps in the energy gap.

3.5.4. Dielectric Properties Solid C60 is a soft dielectric material [533]. The complex dielectric function #
= #1 
+ i#2 
of solid C60 was studied over a broad frequency range using a wide variety of techniques including ac impedance measurements at low frequencies and optical measurements at high frequencies [259, 391, 475, 516, 517, 534]. The complex optical dielectric function #
= #1 
+ i#2 
can be expressed in the terms of the complex refractive index N 
= 9n
+ ik
:. Here, n is the refractive index and k is the extinction coefficient. Both n and k are frequency-dependent optical constants. The expression is #
= 9N 
:2 = 9n
+ ik
:2 . The optical dielectric function #
of C60 was studied over a wide frequency range of 1013 –1016 Hz (corresponding to 0.05– 40 eV, 1 eV = 8066 cm−1
. Below 7 eV, it was studied using infrared (between 0.05 and 0.5 eV) [534] and visible-UV (1.5–5.5 eV) [516, 525] spectroscopies. UV data above 7 eV were obtained using EELS by Kramers–Kronig analysis of the EELS loss function [475]. The features of the #2 
dielectric loss spectrum in the infrared range include four strong lines at 526, 576, 1182, and 1428 cm−1 due to molecular vibrations. At higher energies in the visible and UV ranges, the imaginary #2 
part of the dielectric constants begins to rise at 1.65 eV and there are four prominent peaks at ∼2.75, 3.55, 4.50, and 5.50 eV in the #2 
spectrum, corresponding to four absorption peaks [269]. An absolute

C60 -Based Materials

value of the real #1 
part at 1.5 eV is 4.1. The whole trend of the optical dielectric function of C60 is as below. Where there is a loss process represented by a peak in the #2 
dielectric loss spectrum, there is a rise or a resonant oscillation in the #1 
spectrum. This result is consistent with Kramers–Kronig relations. With frequency decreasing to 1013 Hz, #1 
is approaching its dc value. The small difference between #1 (1013 Hz) ≈ 3 9 [517] and #1 (105 Hz) ≈ 4 4 [391] might be due to the losses associated with C60 molecules rapidly rotating (at the rate of ∼109 Hz) at room temperature (T > T01 , T01 = 260 K is the structure transition temperature). In the low frequency regime (below 105 Hz), polarization mechanisms come into play. The dielectric properties in this regime were investigated by ac impedance measurements, which are usually performed on a sample between closely spaced parallel conducting plates. These M–C60 –M (M = metal) “sandwich” structures consist of 200- m-wide aluminum down stripes as base electrodes, a known thickness of C60 , and 200- m-wide aluminum cross stripes as counterelectrode [535]. The ac equivalent capacitance C
and the dissipation factor D
of the resulting capacitor can be obtained. The real part of the relative dielectric function #1 
can be calculated since the relation between the capacitance and #1 
is known to be C
= #1 
#0 A/d. Here, A is the cross-sectional area of the capacitor, d is the separation distance between the plates, and #0 is the absolute permittivity of free space (8 85 × 10−12 F/m). The imaginary part of the dielectric function #2 
can be extracted from measurement of the dissipation factor D
= #2 
/#1 
. To ascertain the intrinsic dc #1 (0) of C60 , the dielectric measurement needs to be performed at frequencies low enough to include the intrinsic relaxation processes of the material, but high enough to escape the effects of impurities (e.g., molecular oxygen). The effects of oxygen and other impurities occupying the interstitial sites of solid C60 on the dielectric relaxation of C60 films can be ignored only above ∼5 kHz. So the intrinsic dielectric constant was obtained by performing a series of capacitance measurements at 100 kHz. By this method, a value of 4 4 ± 0 2 was obtained for the relative permittivity #1 (0) [391]. The plate spacing d of the C60 capacitor may change due to the thermal expansion or to phase transition at T01 ≈ 260 K. The interaction of oxygen or other species into the interstitial spaces of C60 solid can also lead to such problems. To avoid these problems, a microdielectrometry technique [536, 537] has been employed at low frequencies (from 10−2 to 105 Hz). In this technique, a coplanar electrode configuration was used as an alternative to the “sandwich” configuration. Both electrodes in the microdielectric measurement are placed on the same surface of an integrated circuit, and C60 film is placed over the electrodes by thermal sublimation in vacuum [391]. In comparison with the “sandwich” geometry, this geometry has the advantages as follows. First, the comb electrodes of a coplanar sensor, in contrast to parallel plates, ensure proper calibration of the device even if the material being measured undergoes structure transformations. Moreover, the open-face layout facilitates studies of diffusion of foreign species into the bulk of the material. By means of capacitance and dissipation factor measurements discussed previously, the temperature dependence of

C60 -Based Materials

433

the dielectric properties of C60 single crystals, polycrystalline C60 , and microcrystalline C60 films were studied [359, 538, 539]. On cooling below the first-order structural phase transition at 260 K, a Debye-like relaxational contribution to the dielectric response was observed. From the Debye-like relaxation, the presence of permanent electric dipoles was found, which are unexpected in the centrosymmetric structure of the pure ordered cubic phase C60 . It is suggested that the high degree of frozen orientational disorder of the C60 molecules is responsible for the existence of electric dipolar activity. In addition to the structure phase transition at 260 K, a signature of the glass transition was detected at ∼90 K. Studies on the dielectric properties of C60 films in the high-temperature region discovered that both the capacitance and dissipation factor curves as a function of temperature have a pronounced feature at 435 K, which confirms the fact that above 400 K a phase transition does occur in the crystal [539]. A small charge transfer between the C60 and O2 molecules, which occupy interstitial space of solid C60 exposed to oxygen (or ambient air), creates large electrical dipole moments due to the large size of the C60 molecules [535]. The dipole moments can be coupled to the applied ac electric field via a diffusion-controlled relaxation mechanism. This leads to a significant increase in the permittivity #1 accompanied by a broad dielectric loss peak #2 at ∼10– 100 Hz. For instance, the dielectric function #1 
increases dramatically from 4.4 at 105 Hz to 18.4 at 100 Hz due to the polarization caused by the interstitial oxygen [535]. With increasing oxygenation, the interstitial sites are nearly fully occupied. As a result, interstitial hopping is inhibited, and the loss peaks disappear.

3.5.5. Magnetic Properties Pristine C60 solids are diamagnetic materials due to the presence of ring currents in the C60 molecule [1]. In comparison with other carbon-based materials [540–543], such as graphite and diamond, C60 has an unusually small diamagnetic susceptibility of about −0 35 × 10−6 emu/g [544, 545]. This is because there is an unusual cancellation of ring currents in the C60 molecule. There are two kinds of rings in the C60 molecule: the aromatic rings and the pentagonal rings. The aromatic rings in C60 set up shielding screening currents in a magnetic field [546, 547] which contribute a diamagnetic term to susceptibility >, while the pentagonal rings contribute a paramagnetic term of almost equal magnitude. Table 12 presents the diamagnetic susceptibility for C60 and related materials. Table 12. The diamagnetic susceptibility for C60 and related materials at T = 300 K. Material

> 10−6 emu/g)

Ref.

C60 C70 Graphite Carbon nanotubes Diamond

−0 35 −0 59 −21 1 (c-axis)a −10 2 −0 49

[544, 545] [544, 545] [548] [549, 550] [551]

a The value refers to c-axis susceptibility of graphite. The H in-plane susceptibility of graphite is 0 4 × 10−6 emu/g.

The measurement performed on a C60 powder sample shows the temperature dependence of the magnetic susceptibility [544]. At temperatures greater than 180 K the susceptibility is temperature independent with a mass value of >g = −0 35 × 10−6 emu/g, where the subscript g refers to the susceptibility per gram of sample. At low temperatures, due to a relatively small number of electron spins, >g increases and becomes paramagnetic, yielding a molar Curie constant CM = >M T = 0 75 × 10−4 (>M is a molar susceptibility), which gives an electron spin value of 1 5 × 10−4 per carbon atom. It is believed that the Curie constant here is caused by a foreign paramagnetic impurity in the C60 sample. Magnetization of a large single C60 crystal was measured as functions of applied field and temperature [377]. The sample was carefully prepared and handled to avoid oxygen contamination. The result shows no low-temperature Curie term, from which it is concluded that the low-temperature Curie contribution is likely due to oxygen contamination. In addition, near T01 = 260 K, a discontinuity with a 1.2% change in susceptibility was observed, with >T01−
more negative than >T01+
for the susceptibility at either side of the phase transition. The discontinuity is attributed to the change in the intramolecular geometry at the orientational order–disorder transition. The susceptibility can be considered to be independent of the temperature below and above the structural transition temperature, T01 = 260 K. The experiment demonstrated that the cooperative phase transition affects the magnetic susceptibility of a single molecule. Whereas the unpolymerized C60 is diamagnetic, a few fullerene-based materials were found to be ferromagnetic. Ferromagnetism has previously been observed in [TDAE]C60 [TDAE = tetrakis(dimethylamino)ethylene] below 17 K [552]. It has been reported that hydrogenated fullerene [553], C60 H24 , and some palladium fullerides [554], C60 Pdn , may be ferromagnetic. Recently, the finding of ferromagnetism in pure Rh-C60 (rhombohedral phase 2D C60 polymer) samples by Makarova et al. [555] appears to be the first report on magnetic ordering in a pure molecular carbon material up to temperatures above 300 K (see Section 4.2.4).

3.5.6. Thermal Properties Specific Heat Measurement of the temperature dependence of the heat capacity at constant pressure Cp T
provides the most direct method for the study of the temperature evolution of the degrees of freedom, phase transitions, and the enthalpy change. The temperature dependence of the specific heat of C60 solids is studied mainly by DSC. The overall shape of the C60 specific-heat curve in the temperature range of 0.2–400 K is as follows. At very low temperature (below 1 K), the specific heat shows a Cp T
= C1 T + C3 T 3 behavior. The T 3 dependence component of Cp T
yields a Debye temperature ?D = 80 K [556–558]. The linear T contribution arises from the tunneling modes in the glassy systems [559]. The contribution saturates above ∼5 K and contributes little at high temperatures. In the temperature range 2 ≤ T ≤ 80 K, there is an initial rise in the specific heat, followed by a plateau near 80 K [560]. The intermolecular modes of C60 contribute

C60 -Based Materials

434 significantly in this range. The plateau is due to the saturation of these low-energy excitations. The low-energy excitations in this range consist of both acoustic (translational) modes and rotations (also called librations) involving the entire C60 cage. Because the bonds between cages are weak van der Waals bonds, these low-energy modes saturate below 80 K. With six modes (three translational and three librational) and a molecular number density (N /60), the Dulong– Petit limit for the C60 molecules is 6(N /60)kB = 69 3 mJ/g K, which matches the level of the plateau. Here, N is the number density (per unit mass) of carbon atoms. Above 100 K, there is a nearly linear rise in Cp T
curve, and this rise in Cp is caused by the higher energy intramolecular excitations on the ball, which start to be activated and contribute to the specific heat. The Dulong–Petit limit for the carbon atoms is equal to 3NkB = 2 08 J/g K. Based on the highest energy on-ball mode (196 meV), the saturation temperature is expected to be above 2500 K. Since C60 sublimes around 300  C, it could not be observed anyway. It is the large difference in bond strength between the covalent bonds on the cage and the molecular bonds between the cages that leads to the energy separation between the intraand intermolecular modes [560]. A remarkable feature in the temperature dependence of the specific heat for C60 is the large specific heat anomaly near T01 = 261 K orientational order–disorder transition [368, 561]. The structure of the Cp T
curve near the phase transition temperature T01 is strongly impurity concentration dependent. Moreover, the specific heat of C60 depends dramatically on the cooling or heating rate. Extremely slow heating and cooling runs (at rates of the order of 100 mK/h) showed that even at such low scanning rates the results are rate dependent [368]. The material is not in a thermodynamic equilibrium state and behaves as a system with a very long internal relaxation time (order of 10 h) in completing the first-order transition. The lower the scanning rate, the more the ordering transition temperature shifts to lower (higher) temperatures for the heating (cooling) runs. Measurements [368] on high-purity sublimed C60 show that for the orientational transformation occurring at 262.4 K during the heating run, the associated Cp T
peak is remarkably sharp and attains a value of 12,360 J/kg K. The corresponding heat of transition is about 9000 J/kg (6500 J/mol), a value considerably smaller than reported for single crystal of 12,500 J/kg (9000 J/mol) [376]. For the slow cooling run an even larger peak value of 16,730 J/kg K at a transition temperature of 260.6 K is observed. If the specific heat anomalies associated with phase transitions are removed from the experimental Cp T
data for C60 , then the temperature dependence of the residual specific heat contribution for C60 is in good agreement with that of graphite [367, 562] for T ≥ T01 (261 K). Thermal Expansion The temperature dependence of the lattice constant a0 for C60 shows a large discontinuity at the structure phase transition temperature T01 = 261 K, where the lattice parameter a0 jumps by +0 044 ± 0 004 Å on heating [238]. The fractional jump of the unit cell volume of 5v = 5V
/V is (9 3 ± 0 8
× 10−3 . The average isobaric volume coefficient of thermal expansion ' both above and below the transition is (6 2 ± 0 2
× 10−5 K−1 [238, 349]. It is

surprising that measurements of ' taken to high temperature (1180 K) show a lower value of 4 7 × 10−5 /K than for the lower temperature range [563]. The thermal expansion coefficient of C60 films with thickness t were investigated. As t decreases ' increases and is described well by the relation ' = 17 × 10−6 K−1 + 8 3 × 10−5 nm K−1 t −1 [564]. Thermal Conductivity Because C60 is a van der Waals– bonded molecular solid, its Debye temperature AD due to the intermolecular phonon modes is low with a value of about 70–80 K [350, 556]. Above the Debye temperature the specific heat due to propagating modes is independent of temperature. Consequently, the thermal conductivity 0 is an ideal probe of the phonon mean free path through the kinetic expression 0 = Cs B/3, where C is the specific heat, s is the sound velocity, and B is the phonon mean free path. Experiments showed that 0 ∼ 0 4 W/mK at room temperature and is approximately temperature independent down to 260 K [380, 352]. At 260 K, 0 takes a sharp jump from 0.4 to 0.5 W/mK with a transition width of ∼2 K. Because the heat capacity due to intermolecular modes is constant at 260 K, the 25% jump in 0 must be due to a sudden increase in the phonon mean free path. Upon further cooling, the thermal conductivity increases. At about 85 K, the thermal conductivity shows a shoulderlike structure. The temperature and time dependence of the thermal conductivity below 260 K can be described by a phenomenological model that involves thermally activated jumping motion between two nearly degenerate orientations, separated by an energy barrier of ∼260 meV.

3.5.7. Mechanical Properties Primary C60 is fcc molecular crystal formed by Van der Waals interaction between rigid C60 molecules. It is a soft and friable material showing low strength and high-elastic compliance. Information on the temperature dependence of the elastic moduli is incomplete and partially contradictory [242, 358, 372, 565–567]. The elastic constants of single-crystal C60 at room temperature have been established [568, 569]. Since solid C60 at room temperature is cubic, the matrix of the elastic constants has three independent components: C11 , C12 , and C44 . The elastic moduli C11 , C12 , and C44 were obtained experimentally on the basis of sound velocity and the density C by using the well-known relations between the elastic moduli and sound velocity in cubic crystal: C;T2 = C44 , where ;T is the velocity of transverse ultrasonic waves in the [100]-direction, C;T2 = C11 − C12 + C44
/3, where ;T is the velocity of transverse ultrasonic waves in the [111]-direction, and C;L2 = C11 , where ;L is the velocity of longitudinal waves in the [100]-direction. These values of the elastic moduli determined by the experiments are presented in Table 13.

Table 13. Experimental and theoretical values of the elastic moduli Cij (GPa) of the fcc phase of solid C60 (300 K). C11 14 9 ± 0 9 13.8

C12

C44

C11 –C12

Ref.

8 8 ± 1 0 6.6

6 6 ± 0 18 7.0

6 1 ± 0 45 7.2

Experiment [571] Theory [570]

C60 -Based Materials

435

For comparison, the theoretical values of the elastic stiffness constants obtained in terms of the model intermolecular interaction are also given in Table 7 [570]. In the theory approach, the interatomic Buckingham potential was used for calculating the intermolecular interaction and the electrostatic interaction was taken into account. To compare the previous results of a single crystal with the measurements of the elastic properties performed on compact samples and polycrystalline films of C60 , the isotropic moduli of polycrystalline C60 were estimated by averaging the elastic constants Cij of a single crystal. By regarding the anisotropy in polycrystalline C60 as a result of the disordered orientation of individual single-crystal grains, the bulk modulus K is determined from the equation K = C11 + 2C12
/3 and the shear modulus G from the relation G = 9C11 − C12
/2C44 :2/5 . All other elastic constants of the polycrystal material (Young’s modulus E, the longitudinal modulus CL , and the Poisson ratio ;) can be estimated from G and K using the well-known relations between the elastic moduli of an isotropic medium. Isotropic moduli obtained from singlecrystal, compacted samples and polycrystal films are presented in Table 14.

4. C60 -BASED MATERIALS 4.1. Intercalated Compounds 4.1.1. Alkali Metal Doped C60 M3 C60 (M = K, Rb, and Cs) When the alkali metals (M = K, Rb, and Cs) are doped in C60 , stable crystalline phases M1 C60 , M3 C60 , M4 C60 , and M6 C60 are formed. The M3 C60 system is the most widely studied of the doped C60 compounds because of the discovery of superconductivity [574]. K3 C60 and Rb3 C60 (i) Synthesis The usual way to produce M3 C60 compounds is the vapor transport technique which is most suitable for intercalation of heavy alkali metals into C60 lattice. A wide variety of methods based on this technique have been employed to prepare the samples with different crystalline forms (thin film [575], powder [576, 577], and single crystal [578, 579]). For preparing powder samples, the alkali metals are measured out using precision inner-diameter capillary tubes. The quantity of the metal is controlled by cutting an alkali metalfilled capillary tube to the required length in a dry box. The capillary tube is then placed in a 5–10 mm Pyrex (or quartz) tube along with the C60 powders. Then the Pyrex (or quartz) tube is sealed under high vacuum and heated to a temperature of 200 to 450  C for several days. The alkali metals Table 14. Isotropic moduli obtained from different samples. K (GPa)

G (GPa)

E (GPa)

CL (GPa)

;

10 8 ± 0 75 4 85 ± 0 18 12 6 ± 0 45 17 2 ± 0 45 0 306 ± 0 012 8 4 ± 0 5 3 75 ± 0 1 9 8 ± 0 4 13 4 ± 0 4 0 305 ± 0 02 6 4 ± 0 5 4 1 ± 0 2 12 ± 1 13 3 ± 0 8 0 18 ± 0 04

Sample a b c

Note: Sample a—Isotropic moduli of polycrystalline C60 , calculated from the values of the moduli Cij of a single crystal [568]; b—compacted samples powder [572]; c—polycrystalline films [573].

appear to be completely absorbed by the C60 . After that, the sample tube is held at 200–300  C for a period between 24 and 72 hours to ensure complete diffusion of the metal into the C60 sample. For preparing single-crystalline samples, the intercalation process can be controlled by measuring the conductivity of the samples in-situ. Electrical contacts to the pristine C60 samples are made prior. Then the conductivity apparatus is sealed together with fresh alkali metal in a Pyrex glass apparatus under high vacuum. Uniform doping is accomplished using a repeating high-temperature dope–anneal cycle. Both the sample and dopant are heated uniformly from room temperature to ∼200  C in a furnace first while the sample resistance is continuously monitored. At the point where the resistance of the sample reaches a minimum, the alkalimetal end of the tube is cooled to room temperature and the sample alone is annealed for several hours. Then the metal end is reheated and the sample was further doped until a new resistivity minimum point is reached. The sample alone is then again annealed for several hours. This doping and annealing process is repeated until the resistance reaches an equilibrium state. A method similar to this is also applied to obtain thin film M3 C60 samples [575]. In order to obtain M3 C60 samples with more certain chemical compositions, two methods based on vapor transport are used. One is the azide technique, in which the metal azides (i.e., MN3 , M = K, Rb) are used as the metal sources [580–582]. A mixture of C60 powders and stoichiometric amounts of azides is loaded into a quartz tube and heated slowly under dynamical vacuum. At 550  C for a couple of hours, the azide of any alkali metal decomposed into free alkali metal and N2 , during which C60 is doped. After completion of the reaction, the sample is cooled to room temperature and sealed to avoid exposure to air. In the other method, stoichiometric control is markedly improved by reacting pristine C60 with the saturation-doped product [577]. Two aliquots of C60 are weighed outside the drybox first. A weighed amount of C60 is treated with a large excess of alkali metal at 225  C under vacuum to produce M6 C60 . The air-sensitive product is then diluted with the C60 in the dry box. After heating and annealing under vacuum, the single-phase K3 C60 and Rb3 C60 are prepared. Metal fullerides can also be prepared in solution at low temperatures [583]. The solvents usually used are THF, liquid ammonia, or CS2 . The compounds are obtained by dissolving proportional molar amounts of the metal and the C60 powder in the solvent and then heating at ∼300  C to remove the solvents. (ii) Crystal structure In K3 C60 and Rb3 C60 compounds, the structure of K3 C60 and Rb3 C60 is still fcc although the lattice constants become larger than that of pristine C60 . Table 15 summarizes some of the structural parameters of K3 C60 and Rb3 C60 . M3 C60 have a merohedral disorder which means the individual C60 molecules have two different possible orientations in the fcc lattice. Except this disorder, the symmetry of the C60 molecule is not further reduced by the doping from C60 to M3 C60 . Vibrational properties have been studied by Raman scattering [586–590] and IR spectroscopy in the mid-IR and the far-IR spectral range [324, 591].

C60 -Based Materials

436 Table 15. Crystal structure and lattice constants of M3 C60 . Compounds

K3 C60

Cation position

occupy all the available tetrahedral and octahedral interstitial sites of the host lattice 14 24 14 43 10 06 10 20 3 27 3 33 1 93 2 16 [236] [584, 585]

Lattice constant (Å) Distance of C60 –C60 (Å) Distance of C–M (Å) C (g/cm3 ) Ref.

Rb3 C60

(iii) Normal state properties Electronic structure In M3 C60 , the alkali metal is expected to donate its electron to the t1u -derived conduction band. Hence M3 C60 is expected to be a metal, with a halffilled t1u band. In the solid, molecular orbitals on neighboring molecular sites mix to form bands. The band structure and density of states for K3 C60 using the LDA in DFT are shown in Figure 32 [592, 593]; the electronic structure of C60 is also shown in the figure for comparison. In the M3 C60 solids (M = K, Rb), the C60 molecules have binary (merohedral) orientational disorder, and the statistical occupancy of the two molecular orientations remains disordered even at low temperatures [236]. This disorder was predicted to have significant effects on the band structure of the materials [594], with a smearing out of the fine structure in the tlu -derived conduction band density of states (DOS), whose width remains unaffected by the disorder [595]. However, despite the disorder, the electronic states near EF are best described using an effective Bloch wave vector, and the states localizcd due to disorder lie only at the bottom or top of the band [596]. The calculated t1u -derived bandwidth is ∼0.5 eV for K3 C60 [597]. The LDA-DOS clearly supports a rigid-band picture. In this rigid band model, the band structure of Mx C60 (M = alkali) remains roughly the same as that of pure C60 , expect that the Fermi level is shifted due to the filling of electrons donated by the alkali atoms [598]. Many other calculations of the electronic properties of M3 C60 were performed at different levels of accuracy [594, 599–606]. Whereas the

Figure 32. Left: Hückel theory calculation of the molecular energy spectrum of neutral C60 . Right: Band structure and density of states using the LDA formalism [592]. Adapted with permission from [593], C. H. Pennington and V. A. Stenger, Rev. Mod. Phys. 68, 855 (1996). © 1996, American Physical Society.

effect of the intermolecular binding on the molecular geometry itself can be represented reasonably within LDA at any level, the band structure depends more significantly on the level of accuracy of the calculations. The band dispersion is also determined by the weak interaction between the chains, which cannot be expected to be correctly described within the LDA. PES and EELS were widely used to investigate the electronic structure of the C60 compounds [382, 607–618]. The metallic nature of M3 C60 has been reported from PES studies [182, 610, 611, 619] as well as from conductivity measurements [620, 621]. The PES result indicated the metallic ground state for K3 C60 through a considerable density of states at the Fermi level and a clear Fermi cutoff in the spectra [608, 609]. The C1s excitation spectrum results of K3 C60 and C60 implied the half-filling of the t1u -derived band [622, 623]. The bandwidth observed for the half-filled t1u -derived conduction band is about 1.2 eV and is much larger than the band structure calculations [624]. A clue to the explanation of this observation can be gained from high resolution PES studies of Rb3 C60 and K3 C60 [609, 610] at low temperature. PES profiles of M3 C60 calculated within a model [609, 625] showed that good agreement can be obtained by taking into account relatively strong coupling of the electronic system to the molecular Ag and Hg vibrational modes, together with an additional coupling to the charge carrier plasmon observed at about 0.5 eV in the loss function. Electrical resistivity The electrical resistivity (C) of M3 C60 was measured on various samples using different methods such as dc four-point measurement [578, 579, 626– 632] and other contactless techniques [627, 633–635]. Temperature dependences of the CT
of single crystal K3 C60 and Rb3 C60 have been measured under conditions of constant sample pressure [578, 579]. The measurement results are typically metallic as the sharp drop of resistivity at Tc and CT
varies as (a + bT 2
. Results on pellets (pressed from polycrystalline powders and subsequently annealed) obtained with a contactless technique such as optical excitation spectrum and microwave techniques also showed that CT
varies as (a + bT 2
[627, 635]. Early CT
measurement for underdoped granular film showed that CT
increases with decreasing temperature [626]. However, it has been confirmed with new measurements on crystalline films that the transport properties of both K3 C60 and Rb3 C60 exhibit metallic behavior up to 500 K and a T 2 -like dependence at temperatures from 50 to 250 K [631]. The granularity of the thin films is most probably the reason for the discrepancy. Although the electron–phonon (el–ph) scattering mechanism is expected to result in a linear temperature dependence of the resistivity, however, the T 2 -like dependence behavior cannot be the evidence for other scattering mechanism because the volume of sample varies along with the temperature during the constant pressure measurements and this change will affect the temperature dependence of C significantly. The measurement of C vs T under conditions of constant sample volume can help to find the real scattering mechanism. A linear temperature dependence CT
∼ T on Rb3 C60 was reported [630] and this linearity is consistent with the picture that the scattering mechanism is simply of the electron–phonon type with a coupling constant Btr ∼ 0 65–0.80. Table 16 lists some values of C.

C60 -Based Materials

437 Table 17. Tc and reduction gap of M3 C60 .

Table 16. Electrical resistivity of M3 C60 . Property C (transport, crystalline paraconductivity) C (optical) C (microwave) C (room-temperature, ambient pressure)

K3 C60

Rb3 C60

0.12 m cm

0.23 m cm

0.77 m cm 0.41 m cm —

0.8 m cm — 2.5 m cm

Ref. [579] [633, 634] [627] [630]

(iv) Superconductivity Tc and superconductivity gap It is widely believed that M3 C60 (M = K, Rb) are s-wave BCS-like superconductors, driven by the coupling to the intramolecular Hg phonons and probably with some strong-coupling effects [636]. Transition temperature (Tc
as a function of the lattice constant a has been widely studied for M3 C60 superconductors [272, 637–639]. In these studies the constant a was changed either by chemical substitution or by applied pressure. It was found that Tc is increased by substituting Rb for K and the authors speculated that an elastic response of the lattice to the large ionic size of Rb would lead to a narrowing of the conduction band and a subsequent increase in N EF
which has the general effect of raising Tc [640]. A linear relationship between Tc and pressure P was observed with dTc /dP = −6 3 ± 0 8 K GPa−1 performed up to 0.6 GPa [641] and with dTc /dP = 7 8 K GPa−1 performed up to 2 GPa respectively for K3 C60 [637]. A similar result has also been obtained in Rb3 C60 with dTc /dP = −9 7 K GPa−1 [642]. The authors considered that applied pressure causes band broadening, which leads to both a smaller N EF
and a smaller Tc . In the BCS theory for a system with only one type of ion with mass M, the transition temperature behaves as Tc ∼ M −
, where
= 0 5. The isotope effect when 12 C is substituted by 13 C in M3 C60 has been investigated in order to establish the mechanism of superconductivity. Various values of
were obtained by different groups [643–646] and it was indicated that
= 0 3 should be the best available experimental result [645, 646]. The superconductivity energy gap (5) was first measured by point-contact tunneling spectroscopy using a scanning tunneling microscope [647], which gave a reduced energy gap 25/kB Tc = 5 3 for Rb3 C60 . NMR measurements were also performed to obtain 25/kB Tc and were found to be in good agreement with the BCS value of 3.53 for the gap [648]. Moreover, the data has been obtained using other methods such as muon spin relaxation [649], optical methods, [634, 650], and photoemission [651]. It is noted that the data obtained from different experiments show substantial variation ranging from the BCS value of 3.53 up to about 4.2 (see Table 17). The value of the reduced gap that is substantially larger than the BCS value indicates that strong coupling effects are important. Upper critical field and coherence length A large number of experiments was performed to determine the upper critical field for crystal [184, 578, 651, 655–657], powder [642, 658–663], and thin film [626] samples of M3 C60 using different techniques such as magnetication [642, 659],

Property Tc 25/kB Tc 25/kB Tc 25/kB Tc 25/kB Tc 25/kB Tc 25/kB Tc 25/kB Tc

(tunneling) (tunneling) (NMR) (optical) (tunneling and optical) ( SR) (photoemission)

K3 C60

Rb3 C60

Ref.

19.5 K 5 3 ± 0 2 — 3.0 3.44 — — —

29.5 K 5 2 ± 0 3 2.0–4.0 4.1 3.45 4 2 ± 0 2 3 6 ± 0 3 4.1

[636] [647, 652] [653] [654] [634] [650] [649] [651]

ac susceptibility [660, 661], transport [626, 663], and radio frequency (rf) absorption [662]. Early measurements were usually carried on powder samples. Field-cooled curves from dc magnetization are first used to obtain the temperature dependence of the upper critical field [Hc2 T
]. Measurements on powder K3 C60 [659] and Rb3 C60 [642] samples in this way both showed that the temperature dependence of Hc2 is linear at temperatures not far below Tc . Due to the experimental limitations of the magnetic field window (5–8 T in superconductivity quantum interference device (SQUID) magnetometers), measurements of Hc2 by this method are usually performed at temperatures close to the transition temperature. Therefore a large uncertainty will occur in the extrapolations of Hc2 T
to T = 0 K which depend on the fitting scheme significantly. Using the standard theory of Werthamer–Helfand– Hohenberg (WHH) [664], the critical field extrapolated to T = 0 K is Hc2 0
= 49 T for K3 C60 [659] and Hc2 0
= 78 K for Rb3 C60 [642], respectively. Ac susceptibility [660, 661] and rf absorption [662] technique measurements also show linear temperature dependence of Hc2 near Tc . These two measurements can obtain the data under high fields (the ac susceptibility technique can be performed up to 30 T [660, 661] and the rf absorption technique can be performed up to 60 T [662]). Therefore the Hc2 T
curves can extend near 0 K and this will reduce the uncertainty in the extrapolations of Hc2 T
to 0 K. More accurate values of Hc2 (0) are obtained in these experiments (28 T for K3 C60 and 40 T for Rb3 C60 [660], 30– 38 T for K3 C60 [661], 38 T for K3 C60 , and 76 T for Rb3 C60 [662]). Good agreements of the experimental Hc2 T
curves with the WHH predictions are obtained [660, 662], demonstrating that this theory is successful in describing fullerene superconductors. However, a result of an enhancement of Hc2 T
compared to theory has also been found [661]. A likely mechanism proposed by the authors is the dirtylimit effect which showed that the high upper critical fields might partly result from a reduction of the mean free path by different types of defects and the limitation of the mean free path could be the reason for the strong scatter of the experimental data for Hc2 (0) [661]. The magnetoresistance of K3 C60 samples with different granularity has been measured in fields up to 7.5 T and showed that Hc2 T
is dependent on the grain size of the K3 C60 regions within the sample. dHc2 T
/dT varies from −2 18 to −2 8 T/K near the superconducting transition temperature [663]. Upper critical fields in single-crystalline samples have been obtained using various methods. All of these results

C60 -Based Materials

438 showed a linear temperature dependence of Hc2 near Tc [655, 656, 665, 666]. The Hc2 (0) has first been estimated from transport measurements, with the results 17.5 T for K3 C60 [655] and 62 T for Rb3 C60 [656], and later from dc magnetization measurements, with the results 28.5 T for K3 C60 [665] and 43 T for Rb3 C60 [666]. A upturn of Hc2 T
at temperatures very close to Tc has been observed in almost all experiments on all superconducting compounds. Different authors suggested different explanations for this effect and the best possible explanation [657] is that the upturn is a consequence of the anisotropy of the Fermi surface of fullerene superconductors [599]. From Hc2 (0), the coherence length H can be calculated using Ginzburg–Landau relations. Despite a large scatter (see Table 18) arising from the large scatter of Hc2 (0), it is clearly seen that the coherence length of M3 C60 is very small (a few tens of nanometers) and comparable to the short H of high-Tc superconductors. Lower critical field and penetration depth A large number of experiments were performed to determine the lower critical field using various methods [642, 659, 665, 667–671]. Results on Hc1 for powder K3 C60 [659] and Rb3 C60 [642] samples were first obtained from a dc magnetization method. In this method Hc1 was defined as the field at which a deviation from linearity in MH
first appeared. The result show that the values of Hc1 (0) lies in the range 10–16 mT and Hc1 T
follows the simple empirical formula Hc1 T
/Hc1 0
= 1 − T /Tc
2 [642, 659]. However, different from the ideal superconductor, the measurements of MH
in K3 C60 and Rb3 C60 usually do not exhibit ideal linearity and only have a smooth positive (even a negative) curvature. The deviation is so small that it is very hard to get the point of first deviation from such a curve. Therefore this method may not provide the “intrinsic” values of the Hc1 . Politis et al. [667] measured the reversible magnetization at high external fields and calculated Hc1 from Ginzburg–Landau relations. The value of Hc1 obtained by this method is slightly smaller (9–11.4 mT at zero temperature) than the data obtained by the first method. Table 18. Experimentally obtained Hc2 (0) and H of M3 C60 . K3 C60 Samples

Measurements

Powder Powder Powder Powder Powder Single crystal Single crystal Single crystal Single crystal Film

dc magnetization dc magnetization ac susceptibility ac susceptibility rf absorption transport

Hc2 0
49 T

Rb3 C60

H (nm) Hc2 (0) H (nm)

Ref.

2.6

3.4 2.9–3.3 — 3.1 76 T 2 4 4.5 —

[659] [642] [660] [661] [662] [655]

3.40

[665]

— 28 T 30–38 T 38 T 17.5 T

dc magnetization 28.5 T

— 78 T 40 T

2 3

—

transport

—

62 T

2 4

[656]

dc magnetization

—

43.0 T

2 77

[666]

transport

47 T

—

—

[626]

In the second method based on direct measurements of the reversible magnetization, the quantitative calculation is uncertain, especially in case of powder sample with a large distribution of grain sizes. Therefore a method based on measurement of the trapped magnetization was used to evaluate Hc1 [668]. This measurement was performed on powder and single crystalline K3 C60 and Rb3 C60 samples [665, 668]. Much smaller values of Hc1 , which are about 1 mT at zero temperature, were obtained in these studies. It is considered that wave functions between adjacent C3− 60 ions overlap relatively weakly leading to the smallness of Hc1 [668]. Using Ginzburg–Landau relations, the penetration depth B has been calculated and the value of the coherence length used in the calculations was obtained from independent measurements. These values of B are listed in Table 19. Moreover, the penetration depth has also been estimated from muon spin resonance, with the results 480 nm for K3 C60 and 420 nm for Rb3 C60 [672, 673], from optical conductivity, with the result 800 nm for both K3 C60 and Rb3 C60 [674], and from NMR data, which give 600 nm for K3 C60 and 460 nm for Rb3 C60 [654]. Critical current density The critical current density (Jc
in the M3 C60 system has been obtained from magnetization measurements by using Bean’s critical state model [675]. Based on this model, Jc is determined simply from the dc magnetization curve by the equation Jc = AM+ − M−
/R. In the equation, A is a coefficient which is dependent on the sample geometry, M+ and M− denote the magnetization measured in increasing and decreasing fields at a certain magnetic field, and R is the sample radius [675, 676]. In the case of powders which can be approximated as a collection of spheres, R is set by the average grain size which can be measured in various ways. Different values of Jc for M3 C60 powders are obtained from magnetization measurements which showed substantial hysteresis up to high enough fields (higher than 5 T) [642, 659, 677–679]. However, the values of Jc in the order of 1010 A m−2 indicate substantial flux pinning and high values of the critical current density. For single crystalline samples, certain assumptions about the internal microstructure must be made to obtain R. Jc of single crystal K3 C60 and Rb3 C60 obtained from magnetization measurements by using Bean’s critical state model [670] was of the order of 107 A m−2 , that is, much smaller than the value of Jc for powdered samples (∼1010 A m−2
. The qualitative character of the temperature and magnetic field dependence of Jc has been acquired. Jc shows a near logarithmic dependence on H and does not correspond Table 19. Experimentally obtained Hc1 (0) and B of M3 C60 . K3 C60 Samples Powder Powder Powder Powder Single crystal Single crystal and powder

Hc2 0


B (nm)

13.2 mT 240 — — — 4.2 mT — 1.2 mT 890

Rb3 C60 Hc2 (0)

B (nm)

Ref.

— 12 ± 3 mT 16.2 mT 9–11.4 mT 3.2 mT 1.3 mT

247 215 240–280 — 850

[659] [642] [669] [667] [670] [665, 668]

C60 -Based Materials

439

to a simple analytic functional form [670] in single crystalline K3 C60 and Rb3 C60 samples. The temperature dependence of Jc was first found to follow the empirical equation Jc ∼ Jc 0
91 − T /Tc
2 :n , with n = 3–5 [680] in Rb3 C60 powders. Subsequently, similar results for K3 C60 [681, 682] and Rb3 C60 powders [681] were obtained with n = 1 3–5. This temperature dependence was confirmed by measurements on single-crystalline K3 C60 [657] and Rb3 C60 [670]. The higher the magnetic field, the more rapidly Jc decreases with increasing temperature. Cs3 C60 Cs3 C60 cannot be synthesized by vapor transport reaction because of its thermodynamic instability above ∼200  C which will cause the phase to segregate into the more stable CsC60 and Cs4 C60 . Stable Cs3 C60 was synthesized in solution using liquid ammonia as solvent [683]. Raman measurement showed the Cs3 C60 has a single sharp Ag mode centered at 1447 cm−1 with a linewidth of 11 cm−1 [683]. This indicated that all the fullerene molecules are in the 3− state. X-ray diffraction data showed that the compound consists of two phases: a body-center tetragonal (bct) Cs3 C60 and an A15 cubic Cs3 C60 . The lattice parameters are a = 12 057 and c = 11 432 Å for the bct phase, and a = 11 77 Å for the A15 phase. It was reported that the Cs3 C60 had a Tc of 40 K under high pressure [683], but this result has not yet been reproduced. M3−x Mx′ C60 Structures (M, M ′ = K, Rb, Cs) A series of ternary alkali doped C60 have been synthesized with the general formulae of M3−x M′x C60 [272, 684, 685]. All of these compounds are fcc structures. Most of these M3−x M′x C60 compounds are superconductors. Table 20 lists some of these compounds as well as their structural parameters and Tc ’s. M3−x M′x C60 has merohedreal orientational disorder like M3 C60 . The relations of Tc and lattice constants in these compounds are similar to M3 C60 (see Fig. 34). The overall trend is an increase in Tc with increasing lattice constant, which is interpreted in terms of a decrease in the LUMOderived bandwidth and hence an enhancement of the Fermi energy density of states N EF
at constant band filling [272, 639, 684–687]. Mx C60 (M = K, Rb, and Cs, x = 1, 4, 6) MC60 MC60 can be synthesized by reacting an appropriate amount of pristine C60 with the saturation-doped product (M6 C60
using the vapor transport technique [688–690]. The existence of a crystal phase with stoichiometric MC60 was first indicated by Raman [688], photoemission [691, 692], and X-ray diffraction (XRD) [690] studies.

(i) High temperature MC60 Above 400 K the XRD pattern showed that MC60 has a fcc rocksalt structure in which C60 molecules are orientationally disordered [690]. M ion occupies an octahedral site in the fcc lattice (shown in Fig. 33). This phase is nonmetallic. The lattice parameters of MC60 are listed in Table 21. This phase is nonmetallic. (ii) (MC60 )n When the fcc MC60 is gradually cooled, new metallic polymeric phase forms [692–694]. Covalent bonding is formed between neighboring C60 ions and results in the short interfullerene distance along the distorted orthorhombic a-axis (also shown in Fig. 33). The linkage is through a 92 + 2: cycloaddition [695]. Therefore MC60 phase exhibits a quasi-one-dimensional polymer character as in the case in pristine C60 [693–697]. The (MC60
n is used to denote this phase. These polymer phases are stable in air, in contrast to the vast majority of fulleride salts which are extremely sensitive in air [696]. Polymerization is reversible; heating the sample recovers the monomer fcc phase. XRD studies indicated that (MC60
n phase has a body-centred orthorhombic (bco) lattice with space group Pnnm. The lattice parameters of (MC60
n are listed in Table 21. 13 C NMR spectroscopy gave the direct evidence for the presence of sp3 hybridized carbon atoms in (MC60
n [698, 699]. Another piece of direct spectroscopic identification of the intermolecular bridging C–C modes in (MC60
n (M = Rb, Cs) has been achieved by neutron inelastic scattering measurements [700, 701]. A first-principles quantum molecular dynamics model was developed [702] to describe the vibrational properties of polymerized fullerenes. Rigid ball– ball chain modes are predicted to occur in the energy range 10–20 meV in good agreement with the measured lowtemperature neutron spectra. Only few DFT-based calculations have been performed on the (MC60
n to obtain the band structures [703, 704]. (iii) Metastable structure MC60 Moreover, depending on the cooling rate from the high temperature rocksalt fcc phase, a variety of other metastable phases are encountered. A medium quenching rate to just below room temperature leads to the stabilization of a low-symmetry insulating structure which has been characterized by synchrotron XRD studies to comprise (C− 60
2 dimers, bridged by single C–C bonds [705, 706]. The RbC60 dimer phase crystallizes in the monoclinic space group P 21 /a with lattice parameters a = 17 141 Å, b = 9 929 Å, c = 19 277 Å, and K = 124 4 at 220 K [706] and transforms back to the polymer phase on heating. Another metastable conducting phase of RbC60 and CsC60 was reported with very rapid quenching by immersing the high-T sample in liquid nitrogen [707]. In this case,

Table 20. Structural parameters and Tc of M3−x M′x C60 . M3−x M′x C60

Dopant locationa

a (Å)

Tc (K)

Ref.

KRb2 C60 K1 5 Rb1 5 C60 K2 RbC60 K2 CsC60 RbCs2 C60 Rb2 KC60 Rb2 CsC60

K(t)Rb(t)Rb(o) K(t)K(o)Rb(t)Rb(o) K(t)K(t)Rb(o) K(t)K(t)Cs(o) Rb(t)Cs(t)Cs(o) Rb(t)K(t)Rb(o) Rb(t)Rb(t)Cs(o)

14.243 14.253 14.243 14.292 14.555 14.323 14.431

23.0 22.2 23.0 24.0 33.0 27.0 31.3

[686] [272] [686]

a

(o) = Octahedral site, (t) = tetrahedral site.

[272]

Figure 33. The left: high-T fcc phase of MC60 ; the middle: bco polymer phase stabilized by slow cooling from the fcc phase. The thick lines represent polymer chains; right: the linkage of C60 in (MC60
n .

C60 -Based Materials

440 Table 21. Crystal structure and parameters for Mx C60 x = 1, 4, 6). Mx C60

Crystal structure

Lattice constants (Å)

KC60 RbC60 CsC60 (KC60
n

fcc (rock salt) fcc (rock salt) fcc (rock salt) bco

(RbC60
n

bco

(CsC60
n

bco

K4 C60

bct

Rb4 C60

bct

Cs4 C60

bct (orthorhombic distorted) bcc bcc bcc

a = 14 07 a = 14 08 a = 14 12 a = 9 11 b = 9 95, c = 14 32 a = 9 11 b = 10 10, c = 14 21 a = 9 10 b = 10 21, c = 14 17 a = 11 867, c = 10 765 a = 11 969, c = 11 017

K6 C60 Rb6 C60 Cs6 C60

a = 12 15 b = 11 90, c = 11 45 a = 11 38 a = 11 54 a = 11 79

Ref. [690]

[693, 694]

[711]

[710] [724, 725]

the 3D isotropic character of the structure is retained while orientational ordering of the C− 60 ions leads to a simple cubic structure with a = 13 9671 Å at 4.5 K. Upon heating, it returns to the most stable polymer phase. M4 C60 K4 C60 and Rb4 C60 were synthesized by a vapor transport technique [584, 585, 708, 709]. XRD results showed that they have a bct structure with a space group I4/mmm [709]. The early experiments suggested that Cs4 C60 is isostructural with K4 C60 [584, 708]. Single phase Cs4 C60 , which has an orthorhombic distortion of the ideal bct structure, was later synthesized by the low-temperature liquid ammonia route [710]. Cs4 C60 is not as well studied as K4 C60 and Rb4 C60 due to problems in stabilizing single phase samples. The structure parameters for M4 C60 are listed in Table 21. M4 C60 compounds have an insulating, nonmagnetic ground state [620, 712] although their t1u -derived bands are only partly (2/3) filled, which would lead one to expect a metallic behavior within a one-electron description. Detailed LDA calculations predict that M4 C60 is a metal [713]; this indicated that the insulativity does not come from the gap created by tetragonal symmetry split of the conduction band. Furthermore, electron paramagnetic resonance (EPR) [714], NMR [715] and EELS [716] measurements imply that orientational disorder is not a key ingredient in the insulating behavior of the M4 C60 phases. The scenarios proposed to explain this insulativity have included the suggestion of a Mott–Hubbard ground state for M4 C60 [471, 717, 718] and a description of these compounds as Jahn–Teller insulators [719, 720]. A systematic study of the optical conductivities of M4 C60 has provided strong evidence for a Mott–Hubbard ground state of these compounds [716]. Low-temperature NMR showed the development of a Koringa-like term under pressure, indicating that a metal–insulator transition occurs above 8 kPa [721]. Either a Mott insulator or a Jahn–Teller distortion is consistent with this NMR observation. It seems that a complete understanding of the electronic properties of M4 C60 requires a unique combination of both electron

correlation and Jahn–Teller effects which may lead to a Mott insulator without a magnetic ground state. No superconductivity has been observed in M4 C60 down to 2 K. M6 C60 The alkali metal-saturated compounds M6 C60 were synthesized by a vapor transport technique [709]. Synchrotron XRD data indicated that the M6 C60 compound has a bcc structure and corresponds to the space group Th5 or Im3¯ [722]. In the M6 C60 structure, the alkali metal ions all occupy equivalent distorted tetrahedral sites, which are surrounded by four C60 anions, two with adjacent pentagonal faces and two with adjacent hexagonal faces. NMR spectra of M6 C60 are quite similar and in good agreement with the X-ray studies [723]. From the result it is apparent that the C60 molecules are ordered at all temperatures. The structural parameters for M6 C60 are given in Table 21. M6 C60 has completely filled electronic levels which results in its observed insulating nature. No superconductivity was observed in M6 C60 . Na and Li Doped C60 Na Doped C60 (i) Nax C60 The Nax C60 compounds were synthesized by a vapor transport technique [684]. All of the Nax C60 phases except Na4 C60 have intercalated fcc structures. In Na2 C60 , the dopants fill the tetrahedral sites, leaving the octahedral sites unoccupied [726]. In Na6 C60 , two Na ions occupy tetrahedral sites and four Na atoms (not ions) go into octahedral sites [684]. It was reported that Na3 C60 tends to phase separate (disproportionate) into Na2 C60 and Na6 C60 [684, 727, 728]. The small size of the Na atom also allows the intercalation of Na to high concentrations. The saturated phase is Na10 C60 in which Na atoms are located at the tetrahedral and octahedral sites, and at the general point with fractional occupancies [728]. Later a new Na4 C60 polymer phase which has a body-centered monoclinic unit cell was identified [729, 730]. Superconductivity has not been observed in any Na–C60 binary compounds. The structural parameters for Nax C60 are listed in Table 22. (ii) Na2 MC60 M = Alkali, Cs1−x Rbx , Hgx  The vacant octahedral site in Na2 C60 allows this unique phase to act as a host lattice for further intercalation of dopants into the vacant octahedral site. The crystal structure and lattice parameters of Na2 MC60 are list in Table 23. Superconductivity was observed in Na2 MC60 with M = K, Rb, Cs and Cs1−x Rbx . The Tc -lattice parameter correlation Table 22. Crystal structure and parameters of Nax C60 . Compound Na2 C60 Na3 C60

Na4 C60

Na6 C60 Na10 C60

Lattice constant

Group

Ref.

a = 14 19 Å

Pa3¯ ¯ Fm3m

[726]

T < 500 K, a = 11 235 Å, b = 11 719 Å, c = 10 276 Å, K = 96 16 T > 500 K, a = 11 731 Å, c = 11 438 Å

I2/m

[729]

T ≥ 250 K, a = 14 183 Å T < 250 K, unknown low-symmetry structure

a = 14 38 Å

a = 14 59 Å

[684]

I4/mmm [730] ¯ Fm3m ¯ Fm3m

[684] [728]

C60 -Based Materials

441

Table 23. Structural properties and Tc ’s of Na2 MC60 . Na2 MC60 Na2 KC60 Na2 RbC60

Na2 CsC60 Na2 Rbx Cs1−x C60 Na2 Hgy C60 Na2 Csx C60

Structural properties ¯ a = 14 025 Å, Tc = 2 K Pa3 ¯ a = 14 095 Å, fast cooling → Pa3 Tc = 3 5 K slow cooling → P 21 /a a = 13 711 Å, b = 14 544 Å, c = 9 373 Å, K = 133 53 , no Tc ¯ a = 14 19 Å, Tc = 12 K Pa3 Pa3¯ for 0 < x < 1 with Tm ≈ 300 K

x = 0 50 → a = 14 110 Å, Tc = 10 5 K x = 0 75 → a = 14 103 Å, Tc = 8 0 K ¯ a = 14 194 Å, no Tc y = 0 618 Pa3 ¯ Pa3, no Tc for x < 2 75

Table 24. Structural parameters of Lix C60 . Ref.

Compound

[731]

Li3 C60 Li4 C60 Li6 C60 Li12 C60

[732] [733]

[726] [726]

[734] [735]

X = 2 75 → a = 14 145 Å, Tc = 10 5 K X = 2 85 → a = 14 139 Å, Tc = 8 0 K

in this class of fulleride superconductors is different from the case in M3 C60 (see Fig. 34) [735]. Li Doped C60 (i) Lix C60 The Li–C60 system is the least investigated among all alkali doped C60 because of the difficulties in synthesis of single-phase samples. Lix C60 compounds can be synthesized by thermal decomposition of alkali metal azides or reaction of Li metal and C60 powder in liquid ammonia or by electrochemical doping [736–738]. Structural parameters for some of the Lix C60 compounds are listed in Table 24. No superconductivity was observed in Li fullerides. (ii) Lix MC60 (M = Cs and Rb) For smaller alkali metal K and Na, Li2 MC60 phase is not stable and phase separates [739]. The crystal structure and lattice parameters of Li2 MC60 (M = Cs and Rb) are listed in Table 25.

Structural parameters hexagonal, a = 9 840 Å, c = 16 10 Å hexagonal, a = 9 835 Å, c = 16 16 Å hexagonal, a = 9 918 Å, c = 16 26 Å ¯ (T > 553 K), a = 14 09 Å, no Tc Fm3m I4/mmm (T < 553 K), a = 14 21 Å, c = 13 98 Å (at T = 293 K)

Ref. [737]

[738]

Ammoniated Mx C60 The ammoniated alkali metal fullerides have been prepared by either exposing preformed metal fullerides to ammonia vapor or reaction of metal and C60 in liquid NH3 . Ammoniation induces structure expansion and often changes the symmetry. Table 26 lists some of these compounds and their structural parameters. Some of these ammoniated Mx C60 are superconductors under pressure. Table 26 also lists the Tc of them.

4.1.2. Alkaline–Earth Metal Doped C60 Alkaline–earth fullerides show complicated structures and interesting properties [234, 746–751]. They are more difficult to synthesize and are less understood compared to the alkali–metal fullerides. Their electronic properties show a departure from the simple band-filling picture of the fulleride bands [750–754]. Cax C60 The Cax C60 compounds were synthesized by solid phase reaction. The mixture of high-purity Ca and C60 powder was heated at 550  C under a vacuum of 10−6 Torr for several hours to obtain Cax C60 [234]. Ca5 C60 X-ray diffraction studies indicated the formation of a simple cubic phase at x = 5. The proposed structural model involves multiple occupancy of the octahedral sites by Ca2+ ions [234]. The lattice constant of Ca5 C60 is 14.01 Å [234, 755, 756]. Calcium intercalation into single-crystal C60 thin films was studied with scanning tunneling microscopy and their realspace images were obtained [757]. A model of the Ca5 C60 structure was deduced in which octahedral sites of the fcc lattice are occupied by four Ca ions while one of the tetrahedral sites is filled. The other tetrahedral site is filled at higher concentration, but these later-filled sites are arranged in rectangular or triangular lattices so that their distances are maximized. Measurements of microwave loss, magnetic susceptibility, and Meissner effect first showed that Ca5 C60 becomes superconducting below 8.4 K [234]. The Tc for Ca5 C60 follows the same relation between Tc and the C60 –C60 separation found in the alkali–metal compounds. However, the Table 25. Structural properties and Tc ’s of Lix MC60 .

¯ Figure 34. Tc as a function of the lattice parameter a for the Fm3m (triangles, circles, and open squares) and Pa3¯ (filled squares) superconductors. For K3 C60 and Rb3 C60 the lattice parameter was varied by changing the pressure, while for Mx M′3−x C60 (with M, M′ = K, Rb, Cs) the lattice parameter was varied by changing the composition. Adapted with permission from [735], T. Yildirim et al., Solid State Commun. 93, 269 (1995). © 1995, Elsevier Science.

Lix MC60

Structural and electronic properties

Ref.

Li2 RbC60 Li2 CsC60 Lix CsC60

Li(t)Li(t)Rb(o), a = 13 896 Å, no Tc Li(t)Li(t)Cs(o), a = 14 120 Å, Tc = 12 K X-ray, no Tc for x < 2 5 and x > 4 5 x = 3 → Fm3m Tc = 10 5 K x = 4 → Fm3m Tc = 8 0 K

[739] [685, 740] [741]

C60 -Based Materials

442 Table 26. Structural parameters and Tc of ammoniated Mx C60 . Synthesization method

Compounds (NH3 )K3 C60 (NH3 )8 K3 C60 NH3 NaK2 C60 NH3 NaRb2 C60 (NH3 )x NaK2 C60 (NH3 )x NaRb2 C60

(NH3 )4 Na2 CsC60

fco vapor

x = 0 55 x= 0.60 x= 1.01 x = 0.67 x = 0.71 x = 1.06

Lattice parameter (Å)

Structure

bct fcc fcc fcc liquid NH3

vapor

fcc

14 971 × 14 895 ×13 687

12 15 × 15 7 4.36 14.52 14.500 14.520 14.531 14.350 14.369 14.400 14.473

Tc (K)

Ref.

No Tc

[742, 743] [743]

8 8.5 17 12 8.5 13 11 8 29.6

[744]

[745]

Note: Although (NH3 ) K3 C60 exhibits no superconductivity at ambient pressure, superconductivity can be induced by hydrostatic pressure with an onset at 28 K.

pressure dependence of Tc of Ca5 C60 is quite different. Schirber et al. [758] found that opposite to that observed in the alkali fullerides, the Tc of Ca5 C60 increases with increasing pressure. Band calculations using the local density approximation [752–754] showed that the t1u level is totally filled, while the t1g level of C60 is half-filled. These half-filled t1g -hybridized levels are the ones responsible for the conductivity and superconductivity in Ca5 C60 . Experimental evidence for the partial filling of the t1g level comes from both photoemission spectra [755] and transport measurements [759, 760]. Electron energy loss [761] and inverse photoemission [755] studies showed that the completely filled t1u band for x < 3 is below EF , giving nonmetallic character. Ca275 C60 Studies by Ozdas et al. showed that Ca2 75 C60 is the only stable phase for 1 8 < x < 2 8 [737]. X-ray patterns and calculated results showed that the structure is a 2 × 2 × 2 supercell with space group Pcab, a direct subgroup of ¯ and the lattice parameters are 27 93 × 28 03 × 27 95 Å Pa3, (at 298 K). Ca cations are located in both the tetrahedral (T) and octahedral (O) interstitial sites of the cubic closepacked C60 lattice. All of the O-sites are occupied by single Ca cations, which are displaced off-center. By contrast, one out of every eight T-sites in the subcell is vacant. The Ca2 75 C60 phase is not superconducting down to 2 K under ambient pressure. Electron energy loss [761] and inverse photoemission [755] studies showed that the completely filled t1u band for x < 3 is below EF , giving nometallic character.

AEx C60 (AE = Ba, Sr) In contrast to calcium fullerides which are fcc at all compositions, Bax C60 and Sx C60 only exist in bcc-based structures at x = 3 x = 4, and x = 6 [746– 749]. All of these compounds have been synthesized by solid phase reaction [746]. Their crystal structure and lattice constant are listed in Table 27. No superconducting transition is observed down to 1.5 K in Ba3 C60 and Sr3 C60 [763, 762]. Early investigations show that both of the Ba6 C60 and Sr6 C60 are superconductors, with transition temperatures of 7 and 4 K, respectively [746, 762]. However, later studies indicate that AE4 C60 AE = Ba, Sr) is the actual superconducting phase rather than AE6 C60 AE = Ba, Sr) as originally proposed [747–749]. In contrast to Ca5 C60 , Tc of Ba4 C60 decreased with increasing pressure. High magnetic susceptibility measurements show that AE6 C60 phases are metallic, even though they are not superconducting [749]. Moreover, these magnetic susceptibility measurements and calculated [752] results combined with the different lattice constants dependence of Tc for Ba4 C60 [764] and Ca5 C60 [758] indicate that there is no clear relation between Tc and N EF
in AEx C60 compounds. Alkaline–Earth Substituted Mx C60 M3−x Bax C60 0 < x < 2 Direct synthesis of alkali–alkaline earth fullerides by vapor transport is impractical due to huge differences in vapor pressures. This can be avoided by using azides BaN6 and MN3 (M = alkali metal) as metal source. Direct reactions with BaN6 and MN3 yielded complicated mixed phases [763]. Stable compounds were eventually synthesized by reacting BaN6 powder with the fcc M1 C60 phase

Table 27. Structural parameters of AEx C60 (AE = Ba and Sr, x = 3, 4, 6). AEx C60 (AE = Ba and Sr) AE3 C60

Ba3 C60 Sr3 C60

AE6 C60

Ba6 C60 Sr6 C60

AE4 C60

Ba4 C60 Sr4 C60

Synthesis

Crystal structure

Lattice constant

Ref.

solid phase reaction

simple cubic A15 fcc and A15 coexistence body-centered cubic

11.34 Å 14.144 Å (fcc) 11.140 Å (A15) 11.171 Å 10.975 Å

[763] [762]

orthorbombic (bco)

11 25 × 11 69 × 10 90 Å —

[747, 748] [749]

[746] [762]

C60 -Based Materials

443

prepared by the azide decomposition method [582, 624,765– 767]. This process worked well for all MBa2 C60 with M = K, Rb, or Cs. But the M2 BaC60 cannot be obtained following the same process. However, starting from CsC60 ensuring octahedral Cs and then adding BaN6 and AN3 together in the second step yielded single-phase products MBaCsC60 for M = Na or K, while phase separation occurred for M = Rb or Cs. X-ray diffraction and Raman scattering showed that these compounds are fcc with pentavalent and quadrivalent C60 ’s, respectively [768]. Table 28 lists the structural parameters of these compounds. No superconductivity was observed in these compounds down to 0.5 K [624]. Electron spin resonance (ESR) indicated weakly metallic behavior [769]. 13 C NMR indicated strong dynamical disorder above room temperature, which may be related to the anisotropic crystal field induced by cations of different valence [624, 770]. M3−y M′y Ba3 C60 M, M′ = K, Rb, Cs Samples of M3−y M′y Ba3 C60 were typically synthesized by intercalation of alkali metals into preformed Ba3 C60 [765–767]. XRD for the samples show single phase bcc patterns with space group Im-3. Simulation of diffraction patterns indicates the random occupation of the interstitial sites by M and Ba. The lattice parameter in M3 Ba3 C60 is controllable by changing linearly with the alkali ionic radii. Moreover, by changing the y in Ky Rb3−y Ba3 C60 , the lattice parameter can be tuned quasi-continuously between K3 Ba3 C60 and Rb3 Ba3 C60 [765]. The structural parameters of these compounds are listed in Table 29. M3 Ba3 C60 is a half-filled metal in which the Fermi energy exists at the center of the next lowest unoccupied molecular orbital with the t1g symmetry [767]. Superconductivity was observed in all compounds except for Cs3 Ba3 C60 in which superconductivity was absent down to 0.5 K [765]. Tc decreased with increasing lattice parameter in M3 Ba3 C60 [765]. This Tc –lattice parameter correlation is in striking contrast with the positive correlation between Tc and the fcc cell parameter in M3 C60 superconductors.

4.1.3. Rare-Earth and Lanthanide Metal Doped C60 Yb275 C60 Yb2 75 C60 was successfully synthesized by solid phase reaction [771]. X-ray diffraction studies showed that Yb2 75 C60 has an orthorhombic unit cell with Pcab symmetry and lattice constants of a = 27 87 Å, b = 27 98 Å, and Table 28. Structural parameters and Tc of M3−x Bax C60 (0 < x < 2) compounds.

M3−x Bax C60

Crystal structure

Lattice constant

KBa2 C60 RbBa2 C60 CsBa2 C60 KbaCsC60

fcc fcc fcc fcc

14.173 14.185 14.206 14.215

Å Å Å Å

Atom position

Tc

Ref.

K(o)Ba(t) Rb(o)Ba(t) Cs(o)Ba(t) Cs(o),K and Ba randomly occupy (t)

no Tc

[624]

Note: (o) = Octahedral site, (t) = tetrahedral site.

c = 27 87 Å [771]. Near-edge X-ray absorption measurements show that all of the Yb cations are divalent, meaning that the average negative charge of the C60 anions is 5.5 [772, 773]. EXAFS and Raman measurements also reveal that the C60 anions are distorted in shape, despite their well-ordered local structure around the Yb cations [773]. Yb2 75 C60 becomes superconducting below 6 K. Sm Doped C60 The Smx C60 compounds have been synthesized by solid phase reactions [774]. The unit cell structure of the nominal Sm3 C60 phase is identified as orthorhombic (a = 28 17 Å, b = 28 07 Å, and c = 28 27 Å) derived from the fcc M3 C60 (M = Alkali metal) cell by doubling the cell dimension in each direction, analogous to that of Yb2 75 C60 [774]. However, the exact structure of the Sm3 C60 phase is still an open question. The quality of the diffraction data has prevented a full structure determination. Sm3 C60 is superconducting at Tc = 8 K [774]. Temperature dependent resistivity measurements of Smx C60 samples have been reported. The results show that the resistivity decreases with increasing Sm concentration x [775, 776]. The temperature coefficient of resistivity is negative for all compositions and the absolute value decreases with increasing x. Eu Doped C60 Eux C60 compounds can be synthesized by either high-temperature solid state reaction [777, 778] or through the liquid ammonia route [779]. The Eux C60 prepared by ammonia route was found to be amorphous by XRD [779]. Crystalline phases prepared by solid state reaction with the compositions of x ∼ 3 and 6 are stabilized [777, 778]. The structure of the x ∼ 3 phase is complicated. Experimental data show the coexistence of a minor A15 phase and a majority phase that has the same structure as Yb2 75 C60 . The Eu6 C60 phase has the same bcc structure of the saturation M6 C60 (M = alkali metal) compounds [777, 778]. The experiments reveal the presence of both Eu2+ and Eu3+ ions. Furthermore, it is found that at T = 65 K ferromagnetic Eu pairs form through a superexchange mechanism via the C60 molecules [777].

4.1.4. Halogen Intercalated C60 I2 Intercalated C60 Iodine-doped C60 C60 I4 was synthesized by vapor phase reaction of iodine with pure C60 under helium gas of 10 Torr at 250  C for several days in evacuated Pyrex tubes [780]. XRD results show than C60 I4 has a simple hexagonal cell with a = 9 962 and c = 9 984 Å. In this structure the C60 host material exhibits a sequence of (AAA ) layers, interleaved with guest iodine layers Table 29. Structural parameters and Tc of M3−y M′y Ba3 C60 (M, M′ = K, Rb, Cs) compounds.

M3−y M′y Ba3 C60

Crystal structure

Lattice constant

Atom position

K3 Ba3 C60 Rb3 Ba3 C60 Cs3 Ba3 C60

bcc bcc bcc

11.24 Å 11.34 Å 11.47 Å

Randomly occupy

Tc

Ref.

5.6 K 1.9 K no Tc

[765]

C60 -Based Materials

444 between the host layers, as in a graphite intercalation compound. The {111} planes of the fcc C60 become the {001} planes in the compound by relative shear motions which transform the layer stacking sequence from ABCABC

in fcc C60 to A/A/A/ in C60 I4 , where “/” denotes an iodine guest layer in the C60 I4 compound. The shortest I–I (in-plane) distance is 2.53 Å, which is close to the value of 2.72 Å for the molecular solid I2 [780]. Another intercalation compound of C60 with iodine was also prepared under fairly similar conditions [781] with a final composition of C60 I1 6 [782]. The XRD pattern was interpreted in terms of a C-centered orthorhombic cell with a = 17 33 b = 9 99 c = 9 97 Å, containing two formula units. Iodine atoms were placed in sites slightly shifted from the centers of the trigonal prisms, with near-neighbor and next-near-neighbor distances of 3.9 and 7.21 Å, respectively [781, 782]. Iodine intercalated C60 is nonmetallic and has a resistivity at 300 K of 109  cm. In addition, no superconductivity is observed down to 4 K. Photoemission [783], X-ray absorption [784], and 13 C NMR [785] results confirm the insulating (clathrate) nature (i.e., the absence of guest–host charge transfer). Halogenation of C60 Halogenation reactions of C60 Xn (X = Br, Cl, F) have been reported for various n values. In these reactions the halogen is attached to a carbon atom on the fullerene shell through a single radial covalent band that is formed by using one electron provided by the fullerene and the other by the halogen. The species C60 F36 have been reported by several groups [786, 787]. Complete fluorination of C60 (C60 F60
has also been reported [788]. A very high chlorine content has been found in a photochlorinated fullerene sample which was found to fit the average formula C60 Cl40 [789]. Bromination of C60 has also been observed in chemical reactions carried out in the 25–55  C range. X-ray crystallographic evidence indicates the existence of C60 Br6 , C60 Br8 , and C60 Br24 [790, 791]. These X-ray studies showed that the attachment of Br is accomplished by placing Br at symmetrically bonding vertices of the C60 molecule. These Halogenated fullerenes can be used in substitution reactions to attach aromatic groups sequentially to the fullerene shell.

Table 30. Synthesis and structural properties of gas intercalated C60 . Compound

Ref.

Powder and pellets samples of pure C60 were separately subjected to nitrogen and oxygen, at 0.15 kbar for 7 days at room temperature.

0 < x < 1, Tm ∼ 255 K 0 < x < 1, Tm ∼ 240 K

[792]

(H2
x C60

Pellet samples of pure C60 were separately subjected to hydrogen gas, at 750 atm for 5 h at 323 K and soon was cooled below 150 K, thereby “freezing” the interstitial H2 inside the C60 lattice.

0 < x < 1, out gas at T ∼ 260 K, nearly free quantum rotor at all T

[794] [795]

(CH4
x C60

The powdered C60 was loaded into a highpressure gas cell filled with CH4 and heated to ∼400  C up to 70 h. The cell was quenched in running cold water. The pressure at temperature in the cell is estimated to be between 1 and 2 kbar.

0 < x < 1, Tm ∼ 241 K

[796]

(N2
x C60 (O2)x C60

[792] [793]

form triangular planes as in the case of graphite sheets and the intercalants go between them. This is the case in simple systems such as C60 (P4
2 , C60 (S8
2 . Larger organic and organometallic molecules, like benzene and ferrocene, can also be intercalated to give C60 (C6 H6
4 and C60 (ferrocene)2 , respectively. In addition, reacting C60 with protonic acids, Lewis acids SbCl5 , AsF5 , InCl3 [797–799] also produce different compounds. Table 31 lists some of these compounds and their structural parameters. Table 31. Structural parameters of larger species intercalated C60 .

4.1.5. Molecules Intercalated C60

Compound

Intercalation of Gases into Solid C60 Various gases such as H2 , N2 , H2 O, CH4 , CO, and O2 can be intercalated into solid C60 . These neutral gas molecules can be intercalated into the interstitial sites of the C60 lattice without drastically altering the host structure. The intercalants occupy octahedral sites of the C60 lattice and depress the C60 ’s fcc–sc transition temperature (Tm
. These compounds have possible potential applications such as gas separation, sensors, hydrogen storage, etc. Table 30 lists some of these compounds along with a few structural properties.

C60 (P4
2

Intercalation of Larger Species into Solid C60 When the guest species are larger than the voids in the C60 solids, a change in stacking of the fullerene molecules from cubic to hexagonal usually occurs. Structures of these compounds are different from C60 . In most of these phases, C60 molecules

Structural properties

Synthesis

C60 (S8
2

C60 (ferrocene)2

C60 (AsF5)1 88 C60 (C6H6)4

Structural properties

Ref.

hexagonal, a = 10 084 Å, c = 10 105 Å monoclinic, a = 20 867 Å, b = 21 062 Å, c = 10 508 Å, K = 111 25 , four formula per cell. triclinic, a = 9 899 Å, b = 10 366 Å, c = 11 342 Å, ' = 90 96 , K = 90 96 , = = 118 33 (at 143 K) tetragonal (bct), a = 12 794 Å, c = 12 426 Å triclinic, a = 9 938 Å, b = 15 031 Å, c = 17 425 Å, ' = 65 38 , K = 88 30 , = = 74 83 (at 104 K)

[800, 801] [802]

[803]

[797, 798] [804, 805]

C60 -Based Materials

The TDAE intercalated C60 organic molecular system raised much interest owing to its ferromagnetic-like phase transition at Tc = 16 K [552]. Furthermore, the TDAEC60 material has the highest Tc of any molecular organic ferromagnet [552]. An X-ray diffraction study first established that TDAE-C60 is a stoichiometric material with a TDAE:C60 ratio of 1:1 [806]. The diffraction data were interpreted in terms of a centered monoclinic unit cell, a = 15 807 Å, b = 12 785 Å, c = 9 859 Å, and K = 94 02 [806].

4.2. Polymerized C60 4.2.1. Synthesis and Preparation Photopolymerization of C60 First evidence for polymerization [534] showed that oxygen-free thin films of C60 can be polymerized by the irradiation of intense visible or ultravoilet light with photon energies greater than the optical absorbtion edge (∼1.7 eV) [534, 807]. Because of the strong light absorption of C60 , only films thinner than 0.1 m can be photopolymerized efficiently. The photopolymerized C60 no longer dissolves in many common organic solvents compared to pristine C60 . Dioxygen diffused into C60 films seem to be inhibitors during the photopolymerization process [534, 808] and CCl4 can promote the process [809]. The photopolymerization process found to occur only in the fcc phase above 260 K, the orientational transition temperature of pristine C60 crystals [808], goes by the [2 + 2] cycloaddition mechanism [534] with the double intramolecular bonds broken and a four-atom ring formed by covalent C–C bonds (i.e., accompanied with rehybridizations from sp2 to sp3 ). Direct evidence for the covalent intermolecular bonds confirms the photopolymerization from laser desorption mass spectroscopy measurements [534, 810]. On heating above ∼400 K polymerized C60 reverts to normal monomeric C60 by breaking up the intermolecular bonds and reforming the intramolecular bonds. Recent Raman [811] and atomic force microscopy studies [812] also suggested that two different photopolymerized states might exist. Photopolymerization at temperatures above 350 K seems to produce mainly dimers, while material polymerized at 320 K and below contains larger oligomers. The photopolymerization of C60 usually forms one-dimensional linear-chain structured polymers. Pressure-Induced Polymerization of C60 Pressureinduced polymerization has been studied extensively under various temperatures and pressures in recent years [813]. Several polymerized fullerites have been synthesized from the fcc C60 by applying moderate pressures at elevated temperatures [814–816]. Three different phases possessing one- or two-dimensional C60 –C60 networks have been identified: one-dimensional orthorhombic phase (O-phase) [816], two-dimensional tetragonal phase (T-phase) [816], and twodimensional rhombohedral phases (3R-phase, also called Rh-C60
[814–816]. Recently single crystals of these three structures have been successfully synthesized (O-phase [817, 818], T-phase [819–823], and 3R-phase [824, 825]). The onedimensional and two-dimensional pressure-induced polymerization of C60 goes by the same cycloaddition mechanism of the photopolymerization. Under ultrahigh pressure and a higher temperature (∼13 GPa, 820 K), evidence for the existence of 3D

445 polymers has been observed [826]. But high purity 3D polymers with uniform structure have not been synthesized efficiently. Other Polymerization Methods Polymerization of C60 can also be accomplished by plasma treatments [827, 828], ion bombardments at a low does [829], a low-energy electron injection from STM probe tips to C60 films [830], or electron-beam bombardments from an electron gun [560]. The latter process has been successfully used for chlorinebased reactive ion beam etchng in electron beam lithography where a C60 film acts as a negative-type electon beam resist mask [831].

4.2.2. Structures and Phase Diagram Structures 1D Orthorhombic Phase The orthorhombic phase forms a one-dimensional linear molecular-chain structure along the 110 direction of the original fcc phase (see Fig. 16a) [813, 833]. The smaller value (∼9.1–9.2 Å, compared to the value of 10.02 Å of prinstine C60 ) of the nearest-neighbor distance observed by XRD experiment provides the evidence for a covalent bonding between the C60 cages. The space group has been identified as Pmnn, corrected from the earlier proposed Immm space group [818, 834]. 2D Tetragonal Phase The tetragonal phase forms a twodimensional tetragonal molecular-network layered structure via [2+2] cycloaddition in the (001) plane of the original fcc phase (see Fig. 16b) [813, 833]. The most recent study identified the “tetragonal” phase representing a pseudo-tetragonal packing model with an orthorhombic space group Immm [823] although with a very small higher energy (4 kJ mol−1
[835] than another possible packing model with a truly tetragonal space group P 42 /mmc. 2D Rhombohedral Phase The rhombohedral phase forms a two-dimensional hexagonal molecular-network layered structure via 92 + 2: cycloaddition in the (111) plane of the original fcc phase (see Fig. 16c) [813, 833]. The hexagonal layers are proposed to be stacked in a close-packed arrange¯ [816]. ment of the type ABCABC with space group R3m There are three possible packing modes of the hexagonal layers proposed [835]; recent XRD studies confirmed most stable modes, which suggest a trigonal symmetry in individual layers but not a hexagonal one [818, 825]. 3D Polymers and Superhard Phase The structure of 3D C60 polymers is still ambiguous due to the complex structure and bonding mechanism. There are several theoretically and experimentally suggested structures under nonhydrostatic or hydrostatic compression and high temperature treatment [836–839]. Recently one surperhard phase of 3D polymer treated under 13 GPa, 820 K has been characterized to be a body-centered orthorhombic structure (space group Immm), with 93 + 3: cycloaddition bonding between the 92 + 2: cycloaddition bonded tetragonal layers from a 2D polymer [826, 840]. Figure 35 gives the geometric structural diagrams of C60 polymers. The crystallographic data from different authors for the 1D and 2D C60 polymers are listed in Table 32.

C60 -Based Materials

446

Figure 35. Geometric structures of C60 polymers: (a) 1D C60 polymer (orthorhombic phase), (b) 2D C60 polymer (tetragonal phase), (c) 2D C60 polymer (rhombohedral phase).

Phase Diagram There have been several tentative phase diagrams under pressure–temperature treatments of C60 presented in some of the literature [318, 813]. Since the phase formation and purity of polymerized C60 depend strongly on the temperature and pressure as well as the time of the treatment, those phase diagrams are still under discussion. The typical detailed conditions for the formation of pressure-polymerized C60 have been reported as follows: 1D orthorhombic phase 1–1 2 GPa, 550–585 K, about 5 hours [318, 818]; 2D tetragonal phase: 2–2.5 GPa, 700–873 K 0.5– 4 hours [318, 822, 823]; 2D rhombohedral phase 5–6 GPa, 773–873 K, 0.5–1 hours [318, 825]. It has also been reported that after a 1000 s treatment under conditions of temperature of ∼1.5 GPa and pressure of ∼423 K, most of the pristine C60 will form dimer phase with a (C60
2 content of about 80 mol% [318] (to learn more about C60 donors. See Section 5.2).

4.2.3. Electronic Structures and Energetics Electronic Structures Polymerization of fullerene C60 results in the narrowing of the HOMO–LUMO gap that is caused by the splitting of the degenerated molecular orbitals

of the icosahedral C60 due to its distortion and the formation of intermolecular bonds. The electronic structures of 1D and 2D C60 polymers have been studied theoretically using tight-binding calculation and LDA in the framework of the DFT. From tight-binding calculation, a linear C60 chain has been found to be semiconductive with a finite band gap of 1.148 eV and almost pure s-type intermolecular bonding [843]. Assuming the weak van der Waals interlayer interaction in the rhombohedral and tetragonal polymers, a theoretical prediction on stability and conducting properties has been made according to the tight-binding calculation of a single layer [844]. Different numbers of intermolecular covalent bonds in the tetragonal and rhombohedral structures were found to cause the distinction of bandgap values, which were estimated to be 1.2 and 1.0 eV, respectively. The investigation of the electronic structure of the rhombohedral and tetragonal 2D C60 polymers has been carried out using the LDA [845, 846]. These phases were found to be elemental semiconductors having indirect gaps of 0.35 and 0.72 eV and three-dimensional electronic structures due to the short interlayer distance. Although the LDA generally underestimates the energy gap, its value is consecutively reduced from the C60 to the rhombohedral phase, which correlates well with the result of tight-binding calculations [844]. Recently, empirical tight-binding calculations showed that all 1D and 2D polymers are semiconductors with the smallest energy gap for the rhombohedral phase, and the greatest energy stability for the 1D polymer [847]. Energetics and Stabilities DSC study [848] showed that all polymers are more stable than monomers because the stabilization energy decreases with increasing number of intermolecular bonds per molecule. Formation of polymer bonds lowers the total energy [844]. The stability sequence of a distinct polymerized phase of C60 is dimer (C120 ) > 1D polymer > 2D polymer > Pristine C60 (monomer). All polymers depolymerize into pristine C60 on heating to 300  C and the depolymerization reaction is endothermic, indicating that the monomer form is the least stable phase [848].

Table 32. Crystallographic data for C60 polymers. Polymer phase

Space group

a (Å)

b (Å)

c (Å)

V (Å3
, Z

Orthorhombic phase

Immm Immm Pmnn Pmnn

9.26 9.29 9.098 9.14

9.88 9.81 9.831 9.90

14.22 14.08 14.72 14.66

650 641 658 663

Tetragonal phase

Immm P 42 /mmc

9.09 9.097

9.09 9.097

14.95 15.04

618, 1 1245, 2

P 42 /mmc P 42 /mmc Immm ¯ R3m ¯ R3m (60 ) ¯ (60 ) R3m

9.097 9.02 9.026

9.097 9.02 9.083

15.02 14.93 15.077

622, 1

Immm

9.19 9.204 9.175 8.67

— — — 8.81

24.50 24.61 24.568 12.60

Immm

8.69

8.74

12.71

Rhombohedral phase

3D polymer

1236, 2 597 602 1791.1,3

d (g/cm3


Ref. [816] [841] [834, 841, 842] [818]

1.92 (cal.) 1.88 (exp.)

1.936 (calcd.)

2.004 2.48 (X-ray) 2.5 (exp.) 2.43 (X-ray) 2.5 (exp.)

[816] [819] [834] [822] [823] [816] [834] [825] [840]

C60 -Based Materials

Compared to the most stable phase, graphite, the energies of diamond and C60 , reckoned from the graphite level, are, respectively, 0.020 and 0.42 eV per carbon atom. The total energy of distinct phases of C60 occur within a very narrow energy range of about 0.01 eV per carbon atom and the difference between the polymer and monomer forms of C60 is rather small [848].

4.2.4. Physical Properties Electrical Properties Generally, the conductivity of polymerized C60 increases with the polymerization degree and depends on their structure and the ratio of the number of sp2 to sp3 . Field effect transistor investigations indicate that the 1D plasma-polymerized C60 acts as a p-type semiconductor in contradiction to the n-type semiconductor behavior of pristine C60 [849]. The gas and photosensitivity of transport properties of a C60 polymer thin film are promising for a sensing device [849]. In addition, it is known that the alkalimetal doped (AC60
n 1D polymer will formed spontaneously below 400 K from the monomer AC60 salts. It was found that (KC60
n has metallic conductivity; (RbC60
n and (CsC60
n are also conducting at high temperatures but have a metal– insulator transition at 50 and 40 K, respectively [850]. 2D C60 high-oriented C60 polymer shows a gigantic conductivity anisotropy. Conductivity in the polymerized planes exhibits a metallic-like behavior and the out-of-plane one is semiconductor-like [851]. The anisotropy is temperature dependent and lies in the range of 103 –106 . A randomly oriented polymer shows a semiconductor behavior and obeys the Arrhenius law [852]. The observed in-plane conductivity has been explained by a theoretical calculation which considered the formation of distinguishable intermolecular bondings of 66/66, 56/66, and 56/65 models. The calculated results indicated that the occurrence of regions containing 65/56 bonded molecules within a 66/66 connected hexagonal layer may cause variations of in-plane conductivity from a semiconducting behavior to a metallic behavior [853]. If a 2D polymer is formed at a temperature above the collapse temperature of the C60 molecular cages (about 800  C), the conductivity increases intensively by four orders of magnitude [824]. Both semimetallic, variable range hopping (VRH) and semiconducting behaviors have been observed in the electrical resistivity measurement on samples with disordered structures synthesized from pure C60 at pressures in the range 8–12.5 GPa and temperatures of 900–1500 K [854]. Particularly, the temperature dependence of resistivity in a disordered crystalline 3D-polymerized C60 obtained after 9.5 GPa and 900 K treatment fits Mott’s VRH law C = 0 079exp1/T
1/4 − 1:  cm in the range 270 K down to 5 K for hopping conductivity [854]. Optical Properties Optical absorption and PL spectra of polymerized fullerenes have noticeably broader bands than those in pristine C60 [855], and the optical absorption ranges and PL peaks are shifted to the lower energies [522]. Oxygen-containing polymerized fullerene films show strong photoluminescence in the near-IR and visible ranges (1.50– 2.36 eV) at room temperature, and the photoluminescence

447 decrease in intensity after being annealed at different high temperatures [856]. The inverse transmission spectra for 1D and 2D polymers indicate that raising the polymerization temperature shifts the absorption edge to lower energies [857], showing that the polymers are semiconductors with energy gaps Eg ∼ 1 4 and 1.1 eV which agree well with the theoretical calculations for the orthorhombic and rhombohedral phases (Eg ∼ 1 2 and 1.0 eV, respectively) [844]. The Raman and infrared spectroscopies of polymerized has also been studied; the spectra are given as comparisons with those of pristine C60 crystals in Section 2.2.2. Magnetic Properties Among all the 1D and 2D C60 polymers, the 2D rhombohedral polymer is the only phase known that shows ferromagnetic features such as saturation magnetization, large hysteresis, and attachment to a small samarium–cobalt magnet at room temperature (shown in Figure 36) [555]. The temperature dependences of the saturation and remanent magnetization indicate a remarkably high Curie temperature about 500 K higher than that of the other known organic ferromagnets [555, 858]. The discovery of ferromagnetism in Rh-C60 has opened up the possibility of a whole new family of magnetic fullerenes. Elastic Properties Ultrasonic measurements have revealed that the elastic module of C60 polymer increases with the polymerization degree which depends on the polymerization pressure and temperature [859]. Table 33 gives the elastic properties of C60 polymers compared to pristine C60 and polycrystalline diamond [859]. Remarkably, the disordered 3D polymer and amorphous C60 synthesized at a temperature higher than the collapse temperature of C60 cages shows comparable and exceeding bulk moduli with diamond, respectively. It has been shown that amorphous C60 synthesized at very high pressure and temperature (∼13 GPa, 1850 K) can scratch the diamond surface easily [860, 861], with hardness exceeding that of diamond making the ultrahard C60 among the hardest known materials.

4.3. Metal/C60 Composites For metals other than alkali or alkali earth ones, which have much higher cohesive energy, it is energetically unfavorable for them to diffuse into the interstitial sites of the C60

Figure 36. 2D rhombohedral polymerized C60 attached to a small samarium–cobalt magnet at room temperature.

C60 -Based Materials

448 Table 33. Measured values of sound velocities and elastic moduli in 1D and 2D polymerized C60 samples. Material

p(GPa)/T (K)

C (g/cm3


Pristine C60 1D polymer

8/500

1 68 1 75

2D polymer

8/900

1 9

3D polymer Amorphous C60 Diamond

9.5/970 13/1770

2 62 3 30 3 74

cL (km/s)

cT (km/s)

K (GPa)

G (GPa)

6 0 7 4 6 8 7 4 18 5 18 4

3 70 4 8 4 1 4 2 11 4 8 7

10 8 30 42 45 50 450 790 490

4 85 24 40 30 40 340 250 350



Structure (X-ray results)

0 31 0 19 0 14 0 20

polycrystalline orthorhombic

0 20 0 29

disordered (fcc + bcc) one halo polycrystalline

rhombohedral

Note: cL and cT are longitudinal and transverse sound velocities. K and G are bulk and shear elastic moduli;  is Poisson ratio. Accuracy of ultrasonic measurements is ∼5% [859].

lattice and form a Mx C60 compound similar to Mx C60 alkali fullerides [170, 862]. Therefore M/C60 composites (phaseseparated solids of M/C60 ) were obtained and investigated to study the metal–C60 interaction. The most common composites are multilayer films [863–865] and the nanodispersion films [865–871]. Both of them can be easily obtained by the co-deposition method using metals and pure C60 (>99.9%) under high vacuum [865, 869]. In the M/C60 nanodispersion films, metal nanocrystallites are well dispersed in the C60 polycrystalline matrix and the grains of M are very small [866–869]. The volume fraction of the metal and the grain size of the metal particles can be controlled by adjusting the deposition rates of C60 and the metal, and the temperature of the substrate, respectively [869].

4.3.1. Al, Cu, Ag/C60 Composites Al/C60 Al–C60 interaction was first studied in mutilayers [863]. In-situ resistivity measurements showed that the presence of underlying layers of C60 reduced the critical thickness at which Al became conducting from ∼35 to ∼20 Å and there was a sudden increase in resistance that occurs when each Al layer was covered by a monolayer of C60 . A downward shift of 40 cm−1 in frequency of a considerably broadened Raman-active Ag (2) pentagonal-pinch mode was observed. A charge transfer model in which up to six electrons per C60 were transferred from the Al to the C60 layer across the planar Al–C60 interface was proposed to explain these experimental results [863]. However, the PES and C 1s X-ray-absorption spectra of C60 monolayers on Al(111) surfaces indicated that Al–C60 bonding is covalent [184]. Therefore Raman spectra were measured on Al/C60 nanodispersion films in order to get some insight into the Al–C60 interfacial interactions [868, 872]. The Ag (2) pentagon-pinch mode of C60 splits into two peaks, one pristine mode which still remains at 1468 cm−1 , and one softened mode with a shift of ∼6 to 9 cm−1 . The presence of the pristine mode of C60 demonstrates that Al– C60 interaction is a localized interface effect. In addition, the shifts of the softened Ag (2) mode depend weakly on the Al/C60 ratio and the microstructures of the films. The distinct relative increase in intensity of the Hg (7) and Hg (8) modes compared to Ag (2) mode has also been observed. This result as well as the local effect of Al–C60 interaction

suggest that covalent bonding may play a major role in the interaction between Al/C60 as that reported by Maxwell et al. [184]. Cu/C60 Hebard et al. found that C60 covered on the surface of a Cu layer showed a metallic characteristic [863] and a similar result was also observed by STM research on a C60 layer covered upon the surface of a Cu substrate [14]. Moreover, it was reported that the cover of C60 on a layer of metal can effectively affect the conductivity of metal films [3, 15]. Cu/C60 nanodispersion films grown by a co-deposition method were obtained and the structures were observed by TEM [869, 870, 874, 875]. The uniformly granular microstructure and the small clusters or grains in Cu–C60 films might form a kind of doped fullerene nanostructure so that electrons could be transferred from metal atoms to the surface layers of C60 at the interface [869]. The temperature dependence of the conductivities has been measured in-situ [874]. The linear relationship ln/RT
∝ 1/T
was observed when the temperature was lower than 250 K, and the relation ln/RT
∝ 1/T 1/4
was observed for temperatures higher than 260 K. The values of conductivity of these films (typically in the range of 10−2 – 102 S cm−1
are much higher than that of the pristine C60 solid (∼10−7 S cm−1
. The conduction of Cu/C60 interface compounds is dominated by the electrons hopping between the localized states [874]. A Raman backscattering spectra method was performed on Cu/C60 as-grown films [869, 875]. The peak width of the pentagon-pinch mode [Ag (2)] becomes wider and the peak position shifts to the lower frequency as the atomic ratio of Cu/C60 increases. The vibrational frequencies of the softened modes are about 6 cm−1 lower than that of the pristine mode so it can be inferred that about one electron is transferred to C60 . Ag/C60 Ag/C60 nanodispersion films have been prepared by the co-deposition method [866]. Linear relationships of ln  versus 1/T were observed in electrical conductivity measurement from ∼200 to ∼300  C and the conductivities are in a range similar to those in Cu/C60 films [869]. The shift of the soft Ag (2) mode in Raman spectra indicated that about 1–1.5 electrons are transferred from Ag to C60 . Moreover, Although the peak positions of the Hg (7) and Hg (8) modes are about the same as that of C60 , the relative intensities of them increase significantly in the Ag–C60 film and

C60 -Based Materials

this increase is difficult to interpret with the charge transfer model. A plausible explanation is that Ag atoms may diffuse into the C60 lattice to form a limited dilute interstitial solid solution under high temperatures similar to the result observed in the Al–C60 system [863]. The conductance of discontinuous granular Ag films covered by C60 was measured in-situ during C60 covering [876]. An anomalous peak of conductivity was observed when the thickness of the C60 overlayer was around 1.6 nm. This result indicates that 1– 2 monolayer(s) of C60 covered upon the granular Ag films are “metallic,” which is attributed to the charge transfer between Ag and C60 [876]. Charge transfer has also been observed in Ag deposited C60 films grown on the Si(111)-(7 × 7) surface by using scanning tunneling microscopy [877]. It is found that for individual Ag atoms and very small Ag clusters, the charge transfer is inhibited or less effective, and there is no chemical bonding between Ag and C60 . Only after the Ag atoms aggregated to form Ag clusters larger than a critical size will the charge transfer will become effective. From the lateral size distribution of the Ag/C60 clusters, the critical size is concluded to be no more than 4 nm. Furthermore, the results indicate that the growth of the Ag/C60 clusters should be limited to sizes smaller than 10 nm. This self-limiting growth process and the critical size effects provide a possible method to control the growth of metal clusters with a narrow distribution size.

4.3.2. Magnetic-Metal/C60 Composites Ni/C60 Ni/C60 granular films were also prepared by the co-deposition method [878]. High-resolution transmission electron microscopy (HRTEM) results show that Ni nanoparticles are well isolated and embedded in an amorphous C60 matrix in the films. Raman spectra indicated that the charge transferred to each C60 molecule in an N(Ni):N(C60
= 30 film is ∼4 electrons, and that in N(Ni):N(C60
= 1 5 and 6 films are ∼2 electrons. The charges transferred from Ni to C60 reduce when the Ni nanoparticles become smaller (the average sizes of Ni particles in N(Ni):N(C60
= 1 5, 6, and 30 films are 2.6, 3.3, and 6.6 nm, respectively). This is attributed to the quantum size effects of Ni nanoparticles [878]. Measurements of the surface magneto-optical Kerr effect show that the coercivities (HC
of Ni/C60 films are enhanced significantly. In other words, the critical size of superparamagnetic transition (3.3 nm) is much smaller compared to Ni nanoparticles embedded in graphite (13.6 nm) [879] or SiO2 matrix [880]. It was suggested that the enhancement of HC may be attributed to surface spin disorder of Ni particles induced by the strong interfacial interaction between Ni and C60 . Fe, Co/C60 Unlike the C60 based materials we discussed, Fe, Co/C60 composite films obtained by adding C60 as a minor component into Fe, Co matrix have also been investigated [881]. The concentrations of the as-deposited thin films were expressed in a formula of Bx C60 (M = Fe, Co). TEM images showed that the grains are extremely uniform in the film. Raman spectroscopy shows that most of the fullerenes are stable in the films and the electronic diffraction shows no

449 graphite phase; the carbon interfacial phase is mainly composed of fullerenes. Out-of-plane hysteresis loops for the Co162 C60 films show that both the remanence and the coercivity increase significantly in the films. These values are comparable to the promising perpendicular recording media, Co–Cr and Co–CrTa [882]. These C60 comprising media have the potential to achieve ultrahigh storage density because the C60 molecules limit grain growth, reside at grain boundaries, and reduce the magnetic coupling between grains which can lower media noise [883]. The in-plane major hysteresis loops and the virgin magnetization curve of the Fe73 C60 film display a fast switching property, which suggests this material had the potential to be used in magnetic memory, magnetic switches, spin valves, magnetic springs, and other magnetic devices. The uniformity of the grains and distribution of C60 may contribute to this fast switching ability.

5. LOW-DIMENSIONAL STRUCTURES 5.1. Heterofullerenes 5.1.1. Introduction Heterofullerenes (“dopeyballs,” “heterohedral” fullerenes) are fullerenes in which one or more carbon atoms in the cage structure are substituted by a noncarbon atom (i.e., a heteroatom). In this case, heterofullerene formation is a kind of “on-ball” or “in-cage” doping of the fullerene cage. First spectronscopic evidence for heterofullerenes was observed in gas phase of heterofullerene ions produced by laser vaporization of a graphite pellet containing boron nitride powder [884]. Through modification of the cage structure of heterofullerene, significant modification of geometry, chemical functionality, and electronic character of the fullerenes would occur. So it was anticipated that these heterofullerenes should exhibit a variety of properties and lead to formations of various derivatives due to the in-cage introduction of guest atoms.

5.1.2. C59 N/(C59 N)2 The aza[60]fullerene C59 N is formed by substituting a nitrogen atom for a carbon atom in the C60 cage structure. Neutral C59 N is an open shell molecule due to the trivalency of nitrogen leaving a “dangling bond” on an adjacent carbon atom in the cage (see Fig. 37a). Aza[60]fullerene in the first oxidized state is sometimes called aza[60]fulleronium, for which one resonance structure is shown in Figure 37b. C59 N radical is highly reactive and it easily bonds to form (C59 N)2 dimer, the most stable form of azafullerene (see Figure 38), or is hydrogenized to form hydroazafullerene C59 NH. Synthesis Up to now, the azafullerenes are the only known heterofullerenes which can be isolated as pure substances. Nitrogen-doped fullerenes can be synthesized basically by vaporization of graphite in nitrogen containing atmosphere [885, 886] or by laser ablation of fullerene derivatives which possess organic ligands bound to the carbon cage through

C60 -Based Materials

450

− unusually low g value of C

60 can be explained in terms of Jahn–Teller distortion that splits the triply degenerate t1u states, thus leading to quenching of angular momentum.

Figure 37. Important resonance structures of (a) aza[60]fullerene radical and (b) aza[60]fulleronium.

a nitrogen atom [887], or by condensed-phase organic synthesis. But the latter which yields aza[60]fullerene dimer is proved to be the only efficient method leading to the isolation of macroscopic quantities of azafullerenes so far. There are two different methods succeeding organic synthesis. In 1995 it was found to occur by the effective formation of azafulleronium C59N+ in the gas phase during fast atom bombardment mass spectrometry of a cluster opened N -methoxyethoxy methyl ketolactam, so Hummelen et al. applied this process in solution in the presence of strong acid and isolated C59 N as its dimer (C59 N)2 [888]. In 1996, Grösser et al. succeeded in converting a bisazahomofullerene [889] to the aza[60]fullerene dimer (C59 N)2 in a different way [890]. By both these methods high-purity (C59 N)2 samples can be obtained after high performance liquid chromatography (HPLC) purification. Properties Structure Properties It was found that the C59 dimer is formed by linking two cages in a trans-configuration by one single bond to minimize the repulsion of the nitrogen atoms and to optimize the whole structure (see Fig. 38). The link is made by the two carbon atoms (C, C′ ) each of which joins two hexagons together with the N atom in a C59 N molecule. The molecule has C2h symmetry. The C–C′ dimer bond was calculated to be 1.609 Å, (i.e., 0.05 Å longer than that between two average sp3 cabon atoms) [891]. The binding energy was calculated to be ∼18 kcal mol−1 [891]. The bonding energy is relatively low and the dimer can be cleaved both thermally and photochemically into two C59 N

radicals [892]. Light-induced ESR measurement of a solution of (C59 N)2 in 1-chloronaphthalene, using 532 nm laser pulses, yielded a spectrum with three equidistant lines of equal intensity, indicative of 14 N hyperfine interaction [893]. The hyperfine coupling constant is 3.73 G. The reported values for the g factor of C59 N are 2.0011(1) and 2.0013(2), higher than that of C60 radical anion, 1.9991 [894]. This

Figure 38. Schematic 3D structure for a (C59 N)2 molecule.

Electronic Properties Based upon the extended SSH model for C59 N, the calculated results are very consistent with those from the SCF-MO method. The lattice structure of C59 N is different from that of pure C60 due to effects of the dopant ion, but the deformations of the substituted fullerene cage for the C59 N are mainly limited in the vicinity of the dopant ions [895]. As one electon is doped into the LUMO of C60 in C59 N, it was changed so significantly that the energy level near the Fermi level and the corresponding smaller bandgap between HOMO and LUMO was about 0.3 eV [895, 896]. The calculated gap was much smaller than that for C60 . The HOMO states are mainly localized on the N–C double bonds and the excessive electron density is concentrated on the impurity sites. The excessive electron density of the N atom is 0.27, which implies that electronic charge accumulates on the doped N site and the N atom exists as an acceptor [895]. The electronic properties of (C59 N)2 in thin films have been studied by using PES and EELS [897, 898]. It was found that the HOMO of (C59 N)2 is located mainly on N atoms and the intermolecular C–C′ bond. In contrast, the LUMO has mainly a C-character. The optical gap of the dimer is ∼1.4 eV, some 0.4 eV smaller than that of C60 (1.7– 1.8 eV). It was obtained from the energy-loss functions that the static dielectric constant #1 (0) of (C59 N)2 is about 5.6, which is larger than that of C60 [#1 0
∼ 4], reflecting the smaller energy gap of the heterofullerene dimer. Properties of (C59 N)2 Solid Solid structural investigation of (C59 N)2 using X-ray powder diffraction techniques showed that samples of (C59 N)2 obtained by recrystallization from CS2 exhibit a hexagonal structure (space group P 63 /mmc or a subgroup). The hexagonal phase remains stable under hydrostatic compression to 22 GPa, signifying it is somewhat less compressible than pristine C60 [899]. The diffraction patterns point to a CS2 content of about 1/2 molecule per formula unit. The CS2 can be removed by sublimation of the material in vacuo at temperatures of 500– 600  C. This leads to a monoclinic structure with the space group of at most probably C2/m, where the C59 N dimers sit on the sites of a monoclinic c-centered Bravais lattice with lattice parameters a = 17 25 Å, b = 9 96 Å, c = 19 45 Å, and K = 124 32 , T = 295 K [900]. (C59 N)2 are observed to be unstable at temperatures higher than 200  C in UHV.

5.1.3. Other Heterofullerenes Borafullerenes Borafullerenes were the first reported heterofullerenes observed in 1991 [884]. In 1996, the macroscopic preparation of borafullerenes was realized by using the arc-evaporation method on graphite rods doped with either boron nitride, boron carbide, or boron [901]. Extraction and enrichment were used for the heterofullerene content of the soot, involving pyridine extraction and subsequent treatments of the extract with CS2 and pyridine. XPS spectra of the extract showed a peak at 188.8 eV that was assigned to boron (1s level) in borafullerene (C59 B mainly). The extracted materials appeared to be moisture sensitive, leading to the formation of boron oxide or boric acid [901].

C60 -Based Materials

There is another method to synthesize borafullerene by high-temperature laser ablation of a B4 C-graphite composite rod under an argon atmosphere. The product was analyzed with a laser desorption/ionization mass spectrometer, revealing that C59 B (m/z = 719) was formed together with the C60 molecule [902]. Azafullerenes with Multinitrogen Atoms The synthesis of C58 N2 is possibly complicated if the formations of the two N sites are not simultaneous, which may cause an open shell intermediate as found in the synthesis of (C59 N)2 , giving rise to an undesired reaction route. Synthetic route toward C58 N2 starting from C60 [903] has been proposed; recently, (1,3,5)pyridinophanes having [4.3.2]propellatriene units were synthesized as precursors to macrocyclic polyyne C58 H4 N2 and diazafullerene anion C58 N− 2 was detected in the laser desorption mass spectrum of the pyridinophanes [904]. A new fullerene-like material, consisting of cross-linked nano-onions of C and N, was reported [905]. Growth of the onion shells takes place atom by atom on a substrate surface and yields thin solid films during magnetron sputter deposition. The N content in the onions, up to 20%, is the highest measured for an azafullerene in the solid phase. Total energy calculations of C60−2n N2n molecules suggest the existence of a novel C48 N12 molecule. Azaborafullerenes Recently, it was reported that C58 BN can be produced by a BN substitution reaction of fullerene C60 upon irradiation by a KrF excimer laser at room temperature [906] and isolated by HPLC. Mass spectra and X-ray photoelectron spectroscopy analysis confirmed the substituted formation of C58 BN (m/z = 721).

Heterofullerenes with Heteroatoms Other than N, B Early theoretical calculation of Si in-cage doped heterofullerenes (C59 Si, 1,2-, 1,6-, and 1,60-C58 Si2
predicted that these structures are stable and show a steady decrease of the bandgaps with an increasing number of Si atoms [907]. First evidence for the possible existence of silicon heterofullerenes is provided by analyzing the products from pulsed-laser vaporization of silicon–carbon composite rods by time-of-flight mass spectrometry (TOF-MS) [908], which showed small peaks that would match the masses of SiC+ n clusters (61 ≥ n ≥ 56). Recently, silicon heterofullerenes have been synthesized in a laser vaporization source from targets processed as mixtures of graphite and silicon powders [909] or Six C1−x mixed composition targets [910, 911]. Another method to generate silicon in-cage doped fullerenes is laser-induced photofragmentation of parent clusters of composition C60 Six which are produced in a low pressure condensation cell, through the mixing of silicon vapor with a vapor containing C60 molecules [912]. It was indicated from the analysis of photofragmentation mass spectra that at least three Si atoms can be incorporated in fullerenes, the Si atoms are located close to each other in fullerenes, and Si doping was also predicted to increase the fullerene chemical reactivity [911]. Ab initio calculation of optimized geometrical structures of C59 Si and C58 Si2 isomers showed that the geometrical modifications of the heterofullerenes occur in the vicinity of the Si dopant atom [913]. In the case of oxygen, there is some possible evidence for the existence of O heterofullerenes in the form of

451 C59 O+ ions in gas-phase experiments [914] or C59 O−1 anions produced via collision-induced dissociation of oxy-fullerene anions C60 O− 2−4 [915]. Evidence for the existence of C59 P has been identified in mass spectroscopy. The formation of these heterofullerenes was achieved by evaporating phosphorus and carbon simultaneously but at different positions in a radio frequency furnace (i.e., at different temperatures) [916]. Existence of substitutionally doped fullerenes C59 M has also been identified by mass-spectrometric evidence, where M represents the atom of the transition metals such as Fe, Co, Ni, Rh, and Ir. These in-cage doped fullerenes were produced by the photofragmentation of precursor metal– fullerene clusters C60 Mx [917]

5.2. C120 (C60 Dimer) Since the discovery of the photopolymerization of C60 via 92 + 2: cycloaddition [534, 807], there have been a lot of theoretical studies [918–924] on the smallest subunit of a C60 polymer: C120 (C60 dimer), a dumb-bell-shaped all-carbon molecule polymerized from two individual C60 molecules (see Fig. 39). The most stable isomer of C120 is confirmed to be polymerized via the same 92 + 2: cycloaddition mechanism of C60 polymers (i.e., the double intramolecular bonds broken and a four-atom ring formed by covalent C–C bonds, accompanied with rehybridizations from sp2 to sp3 ).

5.2.1. Synthesis C120 can be synthesized by many methods, such as photopolymerization [534, 808], solid-state mechanochemical reaction [925–927], quasihydrostatic compression of C60 [928], and squeezing of a C60 compound [929]. The photopolymerization method gave the first evidence for the existences of C60 dimer and polymer [534, 808]. It was observed that solid C60 exposed to visible or ultraviolet light forms a polymerized phase which is no longer soluble in toluene. The solid state mechanochemical reaction of C60 with KCN by the use of a high-speed vibration milling (HSVM) technique is the first efficient method to synthesize bulk C120 [925, 926]. In a standard procedure, a mixture of C60 and 20 molar equivalents is placed in a capsule and vigorously vibrated under HSVM conditions for 30 min in a glovebox filled with nitrogen. Separation of the reaction mixture by flash chromatography on silica gel, eluted with hexane/toluene and then with toluene/o-dichlorobenzene (ODCB), gives 70% of recovered C60 and 18% of C120 [925, 926]. It was found that this reaction can take place efficiently also by the use of potassium salts such as K2 CO3

Figure 39. Schematic 3D structure for a C120 (C60 dimer) molecule.

C60 -Based Materials

452 and CH3 CO2 K which contain nucleophilic anions, metals such as Li, Na, K, Mg, Al, and Zn (while Ni and Cu are not effective), and organic bases such as 4-(dimethylamino)and 4-aminopyridine [925, 927]. Single crystals of C120 can grown in an ODCB solution since it is hardly soluble in most common organic solvents but has a reasonable solubility in ODCB [925, 926]. The quasihydrostatic compression method for the dimerization of C60 is achieved under a high-pressure, hightemperature treatment [928, 930, 931]. It was shown that the dimerization occurs even at room temperature in the entire pressure range above ∼1.0 GPa. However, the formation of a dimer phase as a stable modification is at least at temperatures above 400 K. No condition was found that may yield pure C60 dimers so far. A typical condition with the pressure, temperature, and treatment time of 1.5 GPa, 423 K, and 1000 s yields about 80% C60 dimers [930]. Squeezing of a C60 compound was reported to be a high yield selective synthesis method of C60 dimers [929]. Squeezing the organic molecular compound crystal (ET)2 C60 [ET = bis(ethylenedithio)tetrathiafulvalene] at 5 GPa and 200  C followed by removing unreacted ET molecules by sonication in CH2 Cl2 produces C60 dimers, with a yield of about 80%. Most of the product is soluble in o-dichlorobenzene, while there remains a small amount of insoluble material, which may be one-dimensional oligomers or ladder polymers.

5.2.2. Structure The isolated C120 forms a dark brown powder. The 13 C NMR spectrum exhibited 15 signals (including one overlapped signal) in the sp2 region and one signal at 76.22 ppm in the sp3 region, which are fully consistent with the assigned structure with D2h symmetry [925]. It was also indicated that C120 is indeed the essential subunit of these polymers by comparing the data with the 13 C NMR spectra of C60 polymers. The single crystal structure was determined by X-ray crystallography; it was found that there is no obvious difference in the length of the four bonds in the 92 + 2: cycloaddition four-atom ring. The intracage bond length was determined to be 1.575(7) Å, and intercage bond length was found to be 1.581(7) Å [925]. In single crystals grown in an ODCB solution, the C120 molecules are arrayed in highly order layers, which is different from the face-centered cubic structure of C60 [925].

5.2.4. Properties Far-infrared vibrational property investigation of the C60 dimer showed that all the modes are nondegenerate. Hg (1)derived modes are allowed in the dimer and appear as a weak quadruplet near 250 cm−1 [933]. The number of T3u (1)derived modes is three, in agreement with group-theory predictions for the D2h symmetry of the dimer. In room-temperature solution, a systematic study of the photophysical properties and nonlinear absorptive optical limiting responses of the C60 dimer was given [934]. It was found that the absorption, fluorescence (spectrum, quantum yield, lifetime), and photoinduced electron-transfer properties of the C60 dimer are somewhat different from those of C60 but qualitatively similar to those of other C60 derivatives, indicating that the dimer may be considered as a pair of 1,2-functionalized C60 derivatives. The triplet–triplet absorption of the C60 dimer is noticeably weaker than those of C60 and other C60 derivatives, corresponding to lower optical limiting responses of the C60 dimer at 532 nm. It was also concluded that the C60 dimer is photochemically stable since there is no meaningful change in both absorption and fluorescence spectra of the dimer solution after 10 h of continuous photoirradiation. The thermal behavior of the C60 dimer was investigated in the range 80–220  C at a rate of 1  C min−1 using differential scanning calorimetry [925]. An endothermic peak from 150 to 175  C and centered at 162  C was found in the heating process, while no such peak was observed in the cooling scan. It was also found that the C60 dimer could dissociate quantitatively into C60 by heating its ODCB solution at 175  C for 15 min. C120 should have equal chemical reactivity to C60 considering their close reduction potentials.

5.3. Endohedral C60 Endohedral fullerenes are novel forms of fullerene-based materials in which the fullerenes are encapsulated by atom(s) (even some small molecules) in their inner hollow space (see Fig. 40). These new fullerene derivatives are usually described using the symbol “@” conventionally and conveniently with the encaged atom(s) listed to the left of the “@” symbol. For example, a C60 -encaged metal species (M) is written as M@C60 [935]. In recent years endohedral fullerenes have attracted a lot of interest due to their fascinating properties and novel electronic structures.

5.2.3. Electronic Structure The lowered symmetry of the C120 as a consequence of dimerization molecule will result in an increased splitting of energy levels and a narrowing of the optical bandgap [920]. The presence of sp3 intermolecular bonds and the caused distortion of the C120 molecule also change its electronic structure. The electronic DOS confirms universal expectations about the electronic and vibrational energy levels. Due to the dimerization of C60 , the LUMO t1u states split into a quasidoublet and singlet state but the HOMO hu state mixes strongly with the lower gg and hg levels leading to a generally broader DOS [932].

Figure 40. Schematic 3D structure for an endohedral fullerene molecule.

C60 -Based Materials

453

5.3.1. M@C60 , M = Metal

5.3.3. X@C60 , X = Noble Gases

The first evidence for endohedral metallofullerenes based on C60 (M@C60
was proposed from a magic number due to LaC60 in a mass spectrum prepared by laser vaporization of a LaCl2 -impregnated graphite rod [936]. Further evidence showing that the La C+ 60 ions did not react with H2 , O2 , No, and NH3 confirmed that the active metal atoms are indeed encapsulated inside the C60 cage [937]. Original soot containing M@C60 (M = Y, Ba, La, Ce, Pr, Nd, Gd, Er, Eu, and Dy) is usually produced by arcdischarge-heating of an Mx Oy /graphite (Mx Oy = Y2 O3 , BaO, La2 O3 , CeO2 , Pr6 O11 , Nd2 O3 , Gd2 O3 , Er, and Eu2 O3
composite rod under a 50–100 Torr He atmosphere [938– 941]. But the extraction of M@C60 from the soot is difficult because of almost all known M@C60 so far showing high reactivity and they are insoluble and unstable in normal fullerene solvents such as toluene and carbon disulfide (CS2
. Ca@C60 has been extracted by pyridine or aniline [942], M@C60 (M = Y, Ba, La, Ce, Pr, Nd, Gd, and Sr) can be extracted by aniline [943], and Li@C60 prepared by an ion implantation technique can be extracted by CS2 [942, 943]. Pure samples of Eu@C60 , Er@C60 , and Dy@C60 have obtained by combining sublimation and HPLC with aniline as an eluent [939–941]. The X-ray-absorption near edge structure spectrum showed that the valence of Eu atom in Eu@C60 was +2 [939], while the Raman spectrum indicated that the valence of Dy atom in Dy@C60 is +3 [941] with 3 electrons transfering from the Dy atom to the C60 cage. It has also been indicated that the Dy in Dy@C60 is located at an off-center position 1.25–1.30 Å from the center of the C60 cage, and the UV-visible–IR spectrum suggested that the HOMO–LUMO gaps of Dy@C60 , Er@C60 , and Eu@C60 molecules are small [939–941].

Noble gas atoms can also be trapped in C60 cages. In fact, the C60 molecule cage is large enough to enclose all of the noble gases including helium, neon (Ne), argon (Ar), krypton (Kr), and xenon (Xe) [952, 953]. The noble-gas endohedral fullerenes are usually prepared by treating the fullerene with the gas under high temperature and pressure [952, 953]. Treatment at ∼650  C and ∼3000 atm yields a C60 /X@C60 (X = He, Ne, Ar, Kr, Xe) mixture with the ratio of encapsulated to empty fullerenes only on the order of 10−3 . Recently, remarkable enrichment of Ar@C60 (with a purity of 40%) [954], Xe@C60 (with a purity of 31%) [955], and Kr@C60 samples [956] with a purity as high as 90% has been realized by HPLC using a semipreparative PYE[2-(1-pyrenyl)ethylsilyl] column (10 × 250 mm) with toluene as the eluent [957]. The fact that noble gases with smaller masses do not interact significantly [952, 953] (the interaction is suggested to be van der Waals force mainly) with the inner parts of -orbitals of the C60 cage implies that the chemical and physical properties of these compounds will not differ much from those of empty C60 molecules. But 129 Xe NMR spectroscopy study shows that in the case of Xe, the 5p electrons of the Xe are much closer to and interact much more strongly with the -electrons of the C60 [955].

5.3.2. R@C60 , R = N, P Endohedrals of group-V elements N@C60 and P@C60 can be prepared by an ion implantation method [944, 945]. The C60 fullerene is continuously evaporated onto a substrate and simultaneously bombarded with low-energy (40 eV) nitrogen or phosphorous ions. The resulting film is subsequently scratched from the substrate and dissolved in toluene and filtered. The soluble part contains only intact fullerenes, with the ratio of encapsulated to empty fullerenes typically on the order of 10−4 . HPLC techniques are used to obtain pure endohedral fullerenes. These two systems are soluble in organic solvents and stable at room temperature while As@C60 is calculated to be unstable [946]. Experimentally, ESR studies [944, 945, 947–950] revealed that the encapsulated nitrogen and phosphorus are located at the center of the cage, the N or P atom is in quartet electronic spin ground state S = 3/24 S3/2
, and the highly reactive 4 S spin states of N and P do not interact with the C60 cage. The fact of well-isolated electron spin of group-V atoms encapsulated in fullerenes and the easy positioning of endohedral C60 offer a very flexible possible implementation of quantuminformation bits for a promising spin quantum computer [951].

5.4. C60 Mn The exohedral metal complexes of C60 , usually written as C60 Mn , where M is the metal species, can be categorized as the third type of C60 complex compared to the hetero- and endohedral fullerenes (see Fig. 41). The study of gas phase C60 Mn compounds has been stimulated by the discovery of the superconductivity of alkali doped fullerites. C60 Mn complexes are suitable model systems for studying the electronic properties of nanostructures and understanding the fundamental interactions between two nanosystems, particularly metal–covalent interfaces.

5.4.1. Alkali Metals and Alkaline Earth Metals In a pioneering experimental work, Wang and co-workers first studied the photoelectron spectra of C60 K− n clusters, n = 0–3, and they found evidence for charge transfer from the alkali atoms to the C60 molecules [958]. The electron affinities of the species were found to have a linear relationship

Figure 41. Schematic 3D structures for two types of exohedral metal– fullerene clusters (C60 Mn
. (a) A “wet” structure where metal atoms spread over the fullerene cage. Here a C60 M12 molecule with one metal atom on each 12 pentagon facet. (b) A “dry” structure where metal atoms form an adsorbed droplet on C60 .

C60 -Based Materials

454 with the number of K atoms. This indicated that C60 Kn are quite ionic because each K atom donates its outer 4s electron to the t1u LUMO of C60 . A dipole-supported state is found in KC− 60 , yielding an inter-K-C60 stretching vibration of 140 cm−1 . Zimmermann et al. have given a systematic study on alkali metal–fullerene clusters by using photoionization time-offlight mass spectrometry [959]. The C60 Mn clusters were produced by coevaporation of fullerenes and metal in a gas aggregation cell. It was suggested that the stability of the alkali metal–fullerene clusters seems to be determined primarily by the electronic rather than the geometric configuration. The C60 M6 clusters are found to be particularly stable formations for any alkali metal. Coating a fullerene with more than seven alkali metal atoms led to an even–odd alternation in the mass spectra, which signals the beginning of metal–metal bonding [959]. But C60 Li12 was found to be an exception to the electronically determined cluster stability. It retains the iscosahedral symmetry of the C60 molecule because each Li atom was found to be stable when above one of the pentagonal faces. C60 Li12 is very stable both in the singly and the doubly ionized state. The origin of the enhanced stability might be interpreted to be of geometric origin [959, 960]. Alkaline earth metal covered fullerenes C60 Mn (n = 0–500; M = Ca, Sr, Ba) were also studied in these works [959, 961]. It was suggested that these clusters are formed by the completion of metal layers around a central C60 molecule; the first layer is complete when one metal atom is situated above each of the 32 facets of the C60 cage. Recently, the electronic and geometric properties of − C60 Na− n and C60 Aun cluster anions were investigated by TOF-MS and PES [962]. The C60 –Na and C60 –Au clusters were produced in a double-rod laser vaporization source. Both the adiabatic electron affinity and the vertical detachment energy, extracted form the PES spectra, exhibit even– odd alteration with the successive addition of Au atoms on C60 . So odd-n species have high relative stability. C60 Aun takes a “dry” structure where Au atoms aggregate as a cluster (see Fig. 41a). In contrast, C60 Nan adopts the direct binding of the Na atoms into stable trimers, each sodium atom lying above a pentagonal ring; thus C60 Nan takes a “wet” structure where the Na atoms spread over the fullerene cage (see Fig. 41b). For C60 Nan clusters, strong periodical oscillations are observed with a high cluster stability at smaller sizes (n = 3, 6, 9, and 12). When n > 12, the sodium atoms begin to exhibit a metallic layer behavior considering the much smoother adiabatic electron affinity (AEA) and vertical detachment energy (VDE) variations [963]. Additional PES experiments, performed with a 5.83 eV photodetachment energy, have shown that for n = 1 to 3 the charge transfer from the valence orbital of the Na atoms to the fullerene LUMO is approximately complete, whereas at n = 4 this charge transfer is only partial. This result not only points out the specificity of Na compared with the larger alkali elements but also suggests a full rearrangement of the cluster geometric configuration at the intermediate sizes. In contrast to the previous work, another individual study based on the measurement of the polarizability and dipole of isolated C60 Nan (n = 1–34) molecules in a static electric field showed the existence of a permanent electric dipole for every size with increasing n from 1 to 34; these results

cannot be explained by a metal shell around the C60 with a regular distribution but are in agreement with a sodium cluster bound to the C60 molecule (i.e., a sodium droplet on C60 ) [964]. For clusters with more than one sodium atom (n > 2), it was suggested that there is a strong decrease in ionization potential as the sodium cluster size increases, and a full electron transfer from sodium atoms to the C60 molecule is expected. The electric dipole moment of isolated single-alkali-atomC60 molecules (C60 Li, C60 Na, C60 K, C60 Rb, and C60 Cs) has been measured by molecular beam deflection experiments [965]. It was shown that the dipole increases from 12.4 D for C60 Li to 21.5 D for C60 Cs. These results indicated a strong electron transfer from the alkali atom to the C60 cage, which is partial for Na and almost complete for K and Cs. It was also found that the C60 K molecules have a giant polarizability at room temperature, equal to 2506 ± 250 Å3 , but no permanent dipole by using the molecular beam deflection technique [966]. The addition of a potassium atom enhances by more than a factor of 20 the polarizability of a pure C60 molecule. This giant polarizability is caused by the free skating of the K atom on the C60 surface. C60 K behaves like a paraelectric system with an electric dipole of 17 7 ± 0 9 D.

5.4.2. Transition Metals In the case of transition metals, the compound clusters C60 Mx (x = 0  150; M = Ti, Zr, V, Y, Ta, Nb, Fe, Co, Ni, Rh) have been produced using modified laser vaporization of a transition metal target in a low-pressure inert gas condensation cell [917, 967, 968]. It was found that the metals form complete layers around the central fullerene molecule (i.e., forming transition metal coated fullerenes). Mass spectrometric studies on these free metal–fullerene clusters showed that upon heating with a intensive laser pulse, these metal–fullerene clusters transform into metal carbides (M = V, Ti) and bulklike metallo-carbohedrene clusters (M = Nb, Ta where x < 2) [967]. Photofragmentation mass spectra of metal–fullerene clusters C60 Mx (M = Fe, Co, Ni, Rh; x = 0  30) reveal the existence of a reaction channel which yields clusters having the composition of metal in-cage doped heterofullerenes C59−2n M (n = 0  10) [917]. Additional tandem TOF experiments on mass selected C59 M indicated that the initial fragmentation step of this new kind of substitutionally doped fullerene is the loss of a neutral MC molecule [917]. Recently, evidence for the existence of C60 Mx (M = Sm, Pt) has also been observed in photofragmentation study by excimer laser ablation-TOF mass spectrometry [969].

5.5. C60 @Carbon Nanotubes C60 can be encapsulated into carbon nanotubes in the form of self-assembled one-dimensional chains and form a new type of self-assembled hybrid structure called “buckypeapods” (see Fig. 42). The novel structure may give potential applications in data storage [970] and promising high-temperature superconductors [971]. Pulsed laser vaporization of graphite with certain metallic catalysts can produce both carbon nanotubes and C60 molecules [972]. To obtain nanotube production, most of

C60 -Based Materials

Figure 42. Schematic 3D view for C60 @carbon nanotubes.

the C60 is removed by purification and annealing. Nevertheless, HRTEM observation showed that C60 encapsulated in single-wall cabon nanotubes (SWCNTs), usually written as C60 @SWCNTs, can also be obtained via the same method. The encapsulated fullerenes can coalesce into longer capsules under the action of the electron beam [973] or under a high temperature treatment [974, 975]. The encapsulated C60 molecules are positioned in a way that a preferred van der Waals separation (0.3 nm) was maintained between the cages and the nanotubes. The presence or absence of encapsulated cages was strongly correlated with the tubes’ diameters, mostly observed to be 1.3–1.4 nm [976]. In a row of encapsulated C60 molecules concentric with a tube axis, the center-to-center distance between two C60 molecules was observed to be 1.0 nm [973], whereas for paired C60 molecules, this distance was 0.9 nm [976]. In another study using electron diffraction, it was found that the intermolecular spacing of adjacent fullerenes in SWNTs is 0.97 nm, smaller than that of the three-dimensional bulk C60 crystal but larger than that of polymerized C60 crystal [977]. Yields of almost 100% C60 @SWCNTs have been obtained by heating SWCNTs in dry air for 50 hours at 400  C [978]. Another individual work which yielded ∼90% C60 @SWCNTs in abundance by a vapor phase method was confirmed by HRTEM observation indicating a simple lattice of 1.0 nm of the encapsulated one-dimensional C60 molecules [979]. Both experimental results based on HRTEM observation [974] and theoretical results using molecular dynamical calculation [980] suggested an optimum temperature range from 350 to 450  C and that the filling mechanism is most likely through the defect on the surface of SWCNTs but not through the ends of them. EELS experiments study on C60 @SWCNTs showed that the electronic and optical properties of the encapsulated C60 molecules are very similar to fcc C60 with only small changes suggesting a weak van der Waals interaction between the SWCNTs and the C60 molecules [981]. Refined measurements of electrical resistivity, thermopower, and thermal conductivity of highly C60 -filled SWCNTs and unfilled controls from 1.5 to 300 K have been performed recently [982]. It was shown that additional

455 conductive paths form via the filling of interior C60 chains, which can effectively bridge defects on the tube walls, reduce the electrical resistivity, and also reduce the effects of charge carrier localization at low temperature [982]. It can also be indicated from the measurements that the filling of C60 also increases the photon scattering, enhances the thermal conductivity in some sense, and effectively prevents other molecules (such as oxygen molecules) from doping into SWCNTs [982]. Another direct-current electric resistance measurement carried out by the four-probe method using the matlike films of C60 @SWCNTs showed somewhat different results in that, for the temperature range 5 < T < 300 K, the resistances in logarithmic scale exhibit linear temperature dependence when the abscissa is represented by T −1/4 , suggesting an electron transport associated with three-dimensional VRH [978]. Hongo et al. measured the transport properties and their temperature dependences of C60 @SWCNTs bundles on a membrane structure consisting of a 50–100 nm thick silicon nitride on a Si wafer and Pt/Ti(30/30 nm) gap electrodes [983]. Their results showing high resistances of the peapod bundles, with two kinds of temperature dependence characters: one exhibits linear temperature dependence with respect to T −1/4 ; the resistance increases as temperature decreases, which agrees with the VRH aforementioned mechanism. The other shows resistance saturation at low temperature, which suggested the existence of weak localization or disorder [984]. A variety of fullerenes were found to exist inside SWCNTs, including C60 molecules and 0.7 nm diameter fullerene capsules of various lengths [976]. Many of the hybrid structures spontaneously jump during observation. Under the present experimental conditions, chains containing 2–5 C60 melecules are most often seen to jump and each jump is typically 1–10 nm. This dynamic behavior could be driven by beam–specimen interactions.

5.6. (C60 )n In a (C60
n cluster, C60 molecules interact with each other by weak van der Waals forces. Because the interaction has a very short range, investigations of C60 clusters are focused on structure behavior. Experimental studies of fullerene clusters indicate they probably have icosahedral structure, similar to inert gas clusters [985]. Systematic optimization of the lowest energy structures of the (C60
n (n ≤ 57) clusters has been carried out with a genetic algorithm and database conjugate-gradient method [986]. These structures show transitions at n = 17 from icosahedral to decahedral and at n = 38 from decahedral to fcc. The second energy difference curve indicates that n = 13, 38, 55 (C60
n clusters are especially stable. However, taking account of the thermal effect, the icosahedral structure instead of decahedral and close-packed ones is the most stable structure when the cluster is cooled from a high-temperature state by a molecular dynamics simulation.

7

9

24

19

35

46

86

134

187

259

450

616

C80

C82

C84

C86

C88

C90

C92

C94

C96

C98

C100

C102

a

10

1

0 0 1

0

0

0 0

Chiral isomers [41]

D5d , D2 , C2v , C2v′ , D2 (Ih
, D3 , D5h C2 , Cs , C2 , Cs , C2 , Cs , C3v , C3v , C2v D2 ,D2d (two major isomers)

D3 , C2v , C2v′ , D3h , D3h

D6h D3h D2 , Td

D5h

Ih

Ih D6h , D2d [42]

Point groupa

9.542 (max.)(C1 ), 6.918 (min.) (C1
[47]

8.70 [43] 7 79 × 8 45 × 8 16 (D2
, 7 90 × 8 30 × 8 50 (D2d
[45, 46]

8.20 [43]

8 75 × 7 50 × 6 54 [44]

7.76 [43]

7.10 [23]

∼5 [42]

Diameter or length of the 3 principal axes (Å)

Some higher fullerenes have too many isomers, structures, point groups therefore all are not listed here.

5

C78

C104 C106

1 1 2

C72 C74 C76

1

1

1812 [19]

C60

0 0

IPR isomers [40]

C70

1 15

C20 C36

Fullernes

Non-IPR isomers

Appendix A. Inserting Properties on horizontal axis and Fullerenes on Vertical axis.

21 (C2v
 22 (C2v′
8 (D3h
 8 (D3h′
13 (D3 ) [41]

19 (D2
[41]

4 [41]

5 [41]

1 [41]

1 3 (D6h


Distinct carbon atoms

3.42 [50] 3.39 [50]

3.38 [50]

3.32 [50]

3.26 [50]

3.28 [50]

3.21 [50]

3.20 [50]

3.27 [50]

3.20 [50]

3.14 (first) [50, 51] 2.3 (second) [51] 3.23 [50]

3.14 [50]

3.17 [50]

3.10 [50]

3.09 [50] 3.28 [50] 2.89 [50]

2.666 [36] 2.73 [50]

2.65 [50]

2.25 [48] 2.5 [49]

Electron affinity (eV)

[38] [52] [52] [52] [38] [52] [38] [52]

6.96 [52] 6.92 [52]

6.89 [52]

6.95 [52]

6.95 [52]

6.92 [52]

6.96 [52]

7.03 [52]

7.09 [52]

7.09 [52]

7.16 [52]

7.15 [38] 7.17 [52]

7.25 [52]

7.30 [52]

7.48 7.36 7.31 7.13 7.10 7.34 7.05 7.26

7.95 [38] 7.57 [52]

Ionization energy (eV)

776.2 735.4 745.2 804.6 784.0 763.7 826.7 757.5 813.0 839.3 820.1 878.3 822.0 836.7 794.8 793.9 806.1 852.2

694.3 697.8 697.8 702.3 702.6 727.2 724.6 733.6 720.1 750.8 716.4 768.1

(C2v
(C2
[55] (C2
(D2
[55] (C2v
(D5h
[55] (C2v
(D2
[55] (C2v
(Cs
[55] (C2v
(D3h
[55] (Cs
(Cs
[55] (C1
(C1
[55] (D3
(C2v
[55]

(C2v′
(D3
(D3h′
(D3h
[41] (C2v
(D3h
[55] (D5d
(D3
[55] (C2
(C3v
[55] (D2d
(D2
[55]

704.8 [55] 694.6 [55] 699.1 [55]

657.7 [55]

723 [53] 840.8 [54] 618.1 [55]

Heat of formation (kcal/mol)

456 C60 -Based Materials

>10−7 [42]

70 (cal.) [250]

0.9 [261]

Color of solution/solid

Resistivity [( cm)−1 ] Transition temperature

Bulk modulus (GPa)

Bandgap (eV)

Static magnetic susceptibility (cgs ppm per mole of carbon)

Dielectric constant

Heat of sublimation (kJ mol−1
Sublimation standard enthalpy (kJ mol−1


1.05 (cal.) [250]

yellow–brown (pyridine)/black [42]

Mass density (g/cm3


Unit-cell parameter

C36

hcp [42] fcc [250] (polymerized) 6.68 Å [42]

Structure

Property

C60

200 ± 6 (at 298 K) [265] ∼4.6 [269] −7 85 ± 0 71 [271] −8 82 ± 0 11 [259]

∼4 [262] −4 33 ± 0 33 [271] −4 38 ± 0 10 [259]

43 ± 5 [264]

1.8 [262]

280 K, 340 K [253, 258] 25 [252]

190 ± 7 (at 764 K), 202 ± 18 (at 803 K) [266] 206 ± 4 [267] 4–5 [262] −7 16 ± 0 10 [259]

1.3 [262]

170–135 K, glass [259]

′ 0.7 (C2v ) 0.85 (C2v
[263] 47 ± 18 [264]

1.64 (estim.) [254] bright yellow–green C2v -chestnut brown color both in solution D3 -golden-yellow [41] and crystalline state [41]

5–6 [270]

210 ± 6 (at 950 K) [268]

59 ± 6 [264]

1.2 [262]

181 K, 235 K [260] 20 GPa [252]

greenish yellow (mix of two major isomers of ′ D2d and D2
[256] ′ D2d -green D2 -yellow brown [257]

(hcp) 15.92 Å [255], 15.935 Å [250] (fcc) 11.34 [43]

(hcp) 11.12 Å (fcc) 15.6 Å [43]

(hcp) 11.00 Å [43] (fcc) 15.3 Å [254]

(hcp) 10.70 Å [43] (fcc) 15.01 Å [253]

port-wine red/ reddish brown [14]

fcc/hcp [43, 251, 252]

C84

fcc/hcp [43]

C78

fcc/hcp [43]

C76

fcc/hcp

C70

38 [120] 39 ± 5 [264] 181 ± 2 (at 298 K) [39]

6.8 [113] ∼12 [114] 18 [252] 1.7 [262]

∼10−8 260 K, 90 K [116]

magenta/black brown [14]

(hcp) 10.02 Å (fcc) 14.17 Å [236] 1.72 [109]

fcc

Appendix B. Properties of solid fullerenes from C36 to C84 .

C60 -Based Materials

457

C60 -Based Materials

458

EQUATIONS T = h/2e2
R−1

(1)

n n
=1− 0
2 720

(2)

"t = "t0 expEAt /kB T


(3)

 = 0 exp−E' /kT


(5)

2

P T
= P0 exp−5H /kT


(4)

0 = 00 exp−E' /kT0
2

2

(6)

−5

 T
∼ e kT 9N EF
: ' 9ln;ph
/:

(9)

#
= #1 
+ i#2 


(10)

N 
= 9n
+ ik
:

#
= 9N 
:2 = 9n
+ ik
:2 C
= #1 
#0 A/d 3

(15) (16)

= C44

(17)

= C11 − C12 + C44
/3 C;L2

(18)

= C11

K = C11 + 2C12
/3 G = 9C11 − C12
/2C44 :

(19) 2/5

Hc1 T
/Hc1 0
= 1 − T /Tc
Jc = AM+ − M−
/R

Jc ∼ Jc 0
91 − T /Tc
2 :n 1/4

C = 0 079exp1/T


(12) (14)

k = Cvs B/3 C;T2

(11) (13)

D
= #2 
/#1 


C;T2

(7) (8)

#
= #1 
+ i#2 


CP T
= C1 T + C3 T

4

− 1:

(20) 2

(21) (22) (23) (24)

GLOSSARY Buckminsterfullerene Or buckyball, C60 , hollow cage carbon molecule consisting 60 carbon atoms named for R. Buckminster Fuller because of the resemblance of its molecular structure to his geodesic domes. Buckminsterfullerene (C60 ) was originally detected in soot in 1985; isolation was first reported in 1990. C120 Also called C60 dimer. A dumb-bell-shaped all-carbon molecule consisting of two C60 molecules connected by two single C–C bonds via cycloaddition. Endohedral C60 Novel C60 -based complex in which the fullerene molecule (C60 ) is encapsulated by atom(s) or some small molecules in the inner hollow space of its carbon cage. These kinds of molecules are usually written as M@C60 ; M is the endohedral species. Exohedral C60 Novel C60 -based complex in which the fullerene molecule (C60 ) is adsorbed by exohedral atoms (usually metal atoms, including alkali metals, alkaline earth

metals, and transition metals), forming gas-phase stable compounds. Fullerene superconductivity Some kinds of metal doped C60 compounds have superconductivity. The M3 C60 (M = alkali metal) system is the most widely studied of them. The highest Tc found in M3 C60 is about 30 K (Rb3 C60 ). It is widely believed that M3 C60 are s-wave BCS-like superconductors, driven by the coupling to the intramolecular Hg phonons and probably with some strong-coupling effects. Fullerene Any of a class of carbon molecules in which the carbon atoms are arranged into 12 pentagonal faces and 2 or more hexagonal faces to form a hollow sphere, cylinder, or similar figure. The smallest possible fullerene molecule may have as few as 32 atoms of carbon, and fullerene-like molecules as few as 20 carbon atoms. The most common and stable fullerene is buckminsterfullerene (C60 ). Fullerite Solid state form of fullerene; usually means solid state C60 . It usually forms black powders or black–brownish crystals. Heterofullerene Fullerenes in which one or more carbon atoms in the carbon cage structure is substituted by a noncarbon atom (heteroatom), also called “dopeyballs” or heterohedral fullerene, including borafullerenes (C59 B ), azafullerenes (C59 N C58 N2  C69 N ), and heterofullerenes with heteroatoms other than N, B. Metal doped C60 compound Or metal doped fullerite. A kind of material formed by intercalating metal atom into the interstitial sites of C60 host lattice. The metal usually used includes alkali metal, alkaline earth metal, and some lanthanide metals. Some kinds of these materials have superconductivity. Metal/C60 composite A kind of phase-separated solid of M/C60 . These metals (such as Al, Cu, Ag, etc.) usually have high cohesive energy and cannot be intercalated C60 lattice. The most common composites are multilayer films and the nanodispersion films. Polymerized C60 Solid C60 can be polymerized by irradiation of intense visible or ultraviolet light, or under high-temperature, high-pressure treatment. During polymerization, some of the internal C–C bonds of C60 molecules are broken and connected via intermolecular cycloaddition.

REFERENCES 1. H. W. Kroto, J. R. Heath, S. C. O’Brien, R. F. Curl, and R. E. Smalley, Nature 318, 162 (1985). 2. W. Krätschmer, L. D. Lamb, K. Fostiropoulos, and D. R. Huffman, Nature 347, 354 (1990). 3. C. Joachim, J. K. Gimzewski, and A. Aviram, Nature 408, 541 (2000). 4. J. B. Howard, J. T. McKinnon, Y. Makarovsky, A. L. Lafleur, and M. E. Johnson, Nature 352, 139 (1991). 5. D. S. Bethune, C. H. Kiang, and M. S. Devries, Nature 363, 605 (1993). 6. M. Bachman, J. Griesheimer, and K. H. Homann, Chem. Phys. Lett. 223, 506 (1994). 7. W. J. Grieco, A. L. Lafleur, K. C. Swallow, H. Richter, K. Taghizadeh, and J. B. Howard, in “27th Symposium (International) on Combustion,” p. 1669. The Combustion Institute, Pittsburgh, PA, 1998.

C60 -Based Materials 8. R. F. Curl, Carbon 30, 1149 (1992). 9. R. Taylor, G. J. Langley, H. W. Kroto, and D. R. M. Walton, Nature 366, 728 (1993). 10. M. C. Zumwalt and D. R. Huffman, in Proc. Eletrochem. Soc. 94–24, 360 (1994). 11. C. Crowley, R. Taylor, H. W. Kroto, D. R. M. Walton, P.-C. Cheng, and L. T. Scott, Synth. Met. 77, 17 (1996). 12. J. Osterodt, A. Zett, and F. Vögtle, Tetrahedron 52, 4949 (1996). 13. N. R. Conley and J. J. Lagowski, Carbon 40, 949 (2002). 14. R. Taylor, J. P. Hare, A. K. Abdul-Sada, and H. W. Kroto, Chem. Commun. 1423 (1990). 15. W. A. Scrivens, P. V. Bedworth, and J. M. Tour, J. Am. Chem. Soc. 114, 7917 (1992). 16. W. A. Scrivens and J. M. Tour, J. Org. Chem. 57, 6922 (1992). 17. A. D. Darwish, H. W. Kroto, R. Taylor, and D. R. M. Walton, Chem. Commun. 15 (1994). 18. L. T. Scott, M. M. Boorum, B. J. McMahon, S. Hagen, J. Mack, J. Blank, H. Wegner, and A. Meijere, Science 295, 1500 (2002). 19. D. E. Manolopolous, Chem. Phys. Lett. 192, 330 (1992). 20. J. G. Hou, J. Yang, H. Wang, Q. Li, C. Zeng, L. Yuan, B. Wang, D. M. Chen, and Q. Zhu, Nature 409, 304 (2001). 21. H. W. Kroto, Nature 329, 529 (1987). 22. S. T. G. Chmalz, W. A. Seitz, D. J. Klein, and G. E. Hite, J. Am. Chem. Soc. 110, 1113 (1988). 23. R. D. Johnson, D. S. Bethune, and C. S. Yannoni, Acc. Chem. Res. 25, 169 (1992). 24. A. Hirsch, in “Fullerenes and Related Structures, Topics in Current Chemistry” (A. Hirsch, Ed.), Vol. 199, p. 13. Springer-Verlag, Berlin, 1999. 25. R. C. Haddon, J. Am. Chem. Soc. 108, 2837 (1986). 26. R. C. Haddon, J. Am. Chem. Soc. 109, 1676 (1987). 27. R. C. Haddon, Acc. Chem. Res. 21, 243 (1988). 28. R. C. Haddon, J. Am. Chem. Soc. 112, 3385 (1990). 29. R. C. Haddon, Acc. Chem. Res. 25, 127 (1992). 30. M. Schulman, R. L. Disch, M. A. Miller, and R. C. Peck, Chem. Phys. Lett. 141, 45 (1987). 31. M. Häser, J. Almhöf, and G. E. Scuseria, Chem. Phys. Lett. 181, 497 (1991). 32. C. S. Yannoni, P. P. Bernier, D. S. Bethune, G. Meijer, and J. R. Salem, J. Am. Chem. Soc. 113, 3190 (1991). 33. W. I. F. David, R. M. Ibberson, J. C. Matthewman, K. Prassides, T. J. S. Dennis, J. P. Hare, H. W. Kroto, R. Taylor, and D. R. M. Walton, Nature 353, 147 (1991). 34. S. Liu, Y. J. Lu, M. M. Kappes, and J. A. Ibers, Nature 254, 408 (1991). 35. S. Saito and A. Oshiyama, Phys. Rev. B 44, 11532 (1991). 36. L. S. Wang, J. Conceicao, C. Changming, and R. E. Smalley, Chem. Phys. Lett. 182, 5 (1991). 37. A. Tokmakoff, D. R. Haynes, and S. M. George, Chem. Phys. Lett. 186, 450 (1991). 38. H. Steger, J. Holzapfel, A. Hielscher, W. Kamke, and I. V. Hertel, Chem. Phys. Lett. 234, 455 (1995). 39. V. Piacente, G. Gigli, P. Scardala, A. Giustini, and D. Ferro, J. Phys. Chem. 99, 14052 (1995). 40. P. W. Fowler and D. E. Manolopoulos, “An Atlas of Fullerenes,” p. 254. Clarendon Press, Oxford, 1995. 41. F. Diederich and R. L. Whetten, Acc. Chem. Res. 25, 119 (1992). 42. C. Piskoti, J. Yarger, and A. Zettl, Nature 393, 771 (1998). 43. D. L. Dorset and J. R. Fryer, J. Phys. Chem. B 105, 2356 (2001). 44. S. Saito, S. Sawada, N. Hamada, and A. Oshiyama, Mater. Sci. Eng. B 19, 105 (1993). 45. X. Q. Wang, C. Z. Wang, B. L. Zhang, and K. M. Ho, Chem. Phys. Lett. 207, 349 (1993). 46. S. Saito, S. Sawada, and N. Hamada, Phys. Rev. B 45, 13845 (1992). 47. S. Okada and S. Saito, Chem. Phys. Lett. 247, 69 (1995).

459 48. H. Prinzbach, A. Weller, P. Landenberger, F. Wahl, J. Worth, L. Scot, M. Gelmont, D. Olevano, and B. V. Issendorff, Nature 407, 60 (2000). 49. A. Ito, T. Monobe, T. Yoshii, and T. Tanaka, Chem. Phys. Lett. 328, 32 (2000). 50. O. V. Boltalina, E. V. Dashkova, and L. N. Sidorov, Chem. Phys. Lett. 256, 253 (1996). 51. O. V. Boltalina, D. B. Ponomarev, and L. N. Sidorov, Mass Spectrom. Rev. 16, 333 (1997). 52. O. V. Boltalina, I. N. Ioffe, L. N. Sidorov, G. Seifert, and K. Vietze, J. Am. Chem. Soc. 122, 9745 (2000). 53. A. Ito, T. Monobe, T. Yoshii, and K. Tanaka, Chem. Phys. Lett. 330, 281 (2000). 54. E. G. Gal’pern, A. R. Sabirov, I. V. Stankevich, A. L. Chistyakov, and L. A. Chernozatonskii, JETP Lett. 73, 556 (2001). 55. J. Cioslowski, N. Rao, and D. Moncrieff, J. Am. Chem. Soc. 122, 34 (2000). 56. C. Joachim, J. K. Gimzewski, R. R. Schlittler, and C. Chavy, Phys. Rev. Lett. 74, 2102 (1995). 57. C. Chavy, C. Joachim, and A. Altibelli, Chem. Phys. Lett. 214, 569 (1993). 58. N. L. Allinger, R. A. Kok, and M. R. Iman, J. Comp. Chem. 9, 591 (1988). 59. R. Landauer, Z. Phys. B 64, 217 (1987). 60. C. Joachim and J. K. Gimzewski, Chem. Phys. Lett. 265, 353 (1997). 61. D. Porath, Y. Levi, M. Tarabiah, and O. Millo, Phys. Rev. B 56, 9829 (1997). 62. C. Zeng, H. Wang, B. Wang, J. Yang, and J. G. Hou, Appl. Phys. Lett. 77, 3595 (2000). 63. H. Park, J. Park, A. K. L. Lim, E. H. Anderson, A. P. Alivisatos, and P. L. McEuen, Nature 407, 57 (2000). 64. J. I. Pascual, J. Gomez-Herrero, D. Sanchez-Portal, and H.-P. Rust, J. Chem. Phys. 117, 9531 (2002). 65. G. Binnig, N. Garcia, and H. Rohrer, Phys. Rev. B 32, 1336 (1985). 66. B. N. J. Persson and A. Baratoff, Phys. Rev. Lett. 59, 339 (1987). 67. B. C. Stipe, M. A. Rezaei, and W. Ho, Science 280, 1732 (1998). 68. J. R. Hahn, H. J. Lee, and W. Ho, Phys. Rev. Lett. 85, 1914 (2000). 69. J. I. Pascual, J. J. Jackiw, Z. Song, P. S. Weiss, H. Conrad, and H.-P. Rust, Phys. Rev. Lett. 86, 1050 (2001). 70. J. Gaudioso, L. J. Lauhon, and W. Ho, Phys. Rev. Lett. 85, 1918 (2000). 71. L. J. Lauhon and W. Ho, J. Phys. Chem. A 104, 2463 (2000). 72. N. Lorente, M. Persson, L. J. Lauhon, and W. Ho, Phys. Rev. Lett. 86, 2593 (2001). 73. J. I. Pascual, J. Gomez-Herrero, C. Rogero, A. M. Baro, D. Sanchez-Portal, E. Artacho, P. Ordejon, and J. M. Soler, Chem. Phys. Lett. 321, 78 (2001). 74. H. Ajie, M. M. Alvarez, S. J. Anz, R. D. Beck, F. Diederich, K. Fostiropoulos, D. R. Huffman, W. Krätschmer, Y. Rubin, K. E. Schriver, D. Sensharma, and R. L. Whetten, J. Phys. Chem. 94, 8630 (1990). 75. R. D. Johnson, G. Meijer, J. R. Salem, and D. S. Bethune, J. Am. Chem. Soc. 113, 3619 (1991). 76. R. D. Johnson, D. S. Bethune, and C. S. Yannoni, Acc. Chem. Res. 25, 169 (1992). 77. J. W. Arbogast, A. P. Darmanyan, C. S. Foote, Y. Rubin, F. N. Diederich, M. M. Alvarez, and R. L. Whetten, J. Phys. Chem. 95, 11 (1991). 78. Y.-P. Sun, P. Wang, and N. B. Hamilton, J. Am. Chem. Soc. 115, 6378 (1993). 79. B. Ma and Y.-P. Sun, J. Chem. Soc., Perkin Trans. 2, 2157 (1996). 80. C. S. Foote, in “Physics and Chemistry of the Fullerenes” (K. Prassides, Ed.), p. 79. Kluwer Academic, Dordrecht, 1994. 81. S. K. Leach, in “Physics and Chemistry of the Fullerenes” (K. Prassides, Ed.), p. 117. Kluwer Academic, Dordrecht, 1994. 82. A. T. Werner, H. J. Byrne, and S. Roth, Fullerene Sci. Technol. 4, 757 (1996).

460 83. D. K. Palit and J. P. Mittal, Fullerene Sci. Technol. 3, 643 (1996). 84. J. S. Ahn, K. Suzuki, Y. Iwasa, and T. Mitani, J. Lumin. 72, 464 (1997). 85. D. Wang, J. Zuo, Q. Zhang, Y. Luo, Y. Ruan, and Z. Wang, J. Appl. Phys. 81, 1413 (1997). 86. M. Ichida, M. Sakai, T. Yajima, and A. Nakamura, J. Lumin. 72, 499 (1997). 87. S. M. Argentine, K. T. Kotz, T. Rudalevige, D. Zaziski, A. H. Francis, R. Zand, and J. A. Schleuter, Res. Chem. Intermed. 23, 601 (1997). 88. A. Fedorov, M. N. Berberan-Snatos, J.-P. Lefevre, and B. Valeur, Chem. Phys. Lett. 267, 467 (1997). 89. Y.-P. Sun, in “Molecular and Supramolecular Photochemistry” (V. Ramamurthy and K. S. Shanze, Eds.), Vol. 1, p. 325. Dekker, New York, 1997. 90. C. S. Foote, in “Topics in Current Chemistry: Electron Transfer I” (J. Mattay, Ed.), p. 347, Springer-Verlag, Berlin, 1994. 91. S. Nath, H. Pal, D. K. Palit, A. V. Sapre, and J. P. Mittal, J. Phys. Chem. B 102, 10158 (1998). 92. F. Negri and G. Orlandi, J. Chem. Phys. 108, 9675 (1998). 93. A. Sassara, G. Zerza, F. Negri, and G. Orlandi, J. Chem. Phys. 107, 8731 (1997). 94. A. Sassara, G. Zerza, and M. Chergui, J. Phys. Chem. A 102, 3072 (1998). 95. J. S. Ahn, K. Suzuki, Y. Iwasa, N. Otsuka, and T. Mitani, J. Lumin. 76, 201 (1998). 96. M. Terazima, N. Hirota, H. Shinohara, and Y. Saito, J. Phys. Chem. 95, 9080 (1991). 97. L. Biczok, H. Linschitz, and R. I. Walter, Chem. Phys. Lett. 195, 339 (1992). 98. R. V. Bensasson, T. Hill, C. Lambert, E. J. Land, S. Leach, and T. G. Truscott, Chem. Phys. Lett. 201, 326 (1993). 99. C. Luo, M. Fujitsuka, A. Watanabe, O. Ito, L. Gan, Y. Huang, and C.-H. Huang, J. Chem. Soc. Faraday Trans. 94, 527 (1998). 100. M. R. Wasielewski, M. P. O’Neil, K. R. Lykke, M. J. Pelin, and D. M. Gruen, J. Am. Chem. Soc. 113, 2774 (1991). 101. T. W. Ebbesen, K. Tanigaki, and S. Kuroshima, Chem. Phys. Lett. 181, 501 (1991). 102. Y. Kajii, T. Nakagawa, S. Suzuki, Y. Achiba, K. Obi, and K. Shibuya, Chem. Phys. Lett. 181, 100 (1991). 103. G. L. Closs, P. Gautam, D. Zhang, P. J. Krusic, S. A. Hill, and E. Wassserman, J. Phys. Chem. 96, 5228 (1992). 104. D. K. Palit, A. V. Sapre, J. P. Mittal, and C. N. Rao, Chem. Phys. Lett. 195, 1 (1992). 105. N. M. Dimitrijevic and P. V. Kamat, J. Phys. Chem. 96, 4811 (1992). 106. R. B. Weisman, in “Optical and Electronic Properties of Fullerenes and Fullerene-Based Materials” (J. Shinar, Z. V. Vardeny, and Z. H. Kafafi, Eds.), p. 83. Dekker, New York, 2000. 107. L. Tutt and A. Kost, Nature 356, 225 (1992). 108. D. Mclean, R. Sutherland, M. Brant, D. Brandelik, P. Fleitz, and T. Pottenger, Opt. Lett. 18, 858 (1993). 109. L. Smilowitz, D. McBranch, V. Klimov, J. Robinson, A. Koskelo, M. Grigorova, B. Mattes, H. Wang, and F. Widl, Opt. Lett. 21, 922 (1996). 110. J. E. Riggs and Y.-P. Sun, J. Chem. Phys. 112, 4221 (2000). 111. Y.-P. Sun, G. E. Lawson, J. E. Riggs, B. Ma, N. Wang, and D. K. Moton, J. Phys. Chem. A 102, 5520 (1998). 112. L. Tutt and T. Boggess, Progr. Quant. Electron. 17, 299 (1993). 113. J. Perry, K. Mansour, S. Marder, K. Perry, J. D. Alvarez, and I. Choong, Opt. Lett. 19, 625 (1994). 114. J. Shirk, R. Pong, F. Bartoli, and A. Snow, Appl. Phys. Lett. 63, 1880 (1993). 115. T. Xia, D. J. Hagan, A. Dogariu, A. A. Said, and E. W. V. Stryland, Appl. Opt. 36, 4110 (1997). 116. J. W. Perry, in “Nolinear Optics of Organic Molecules and Polymers” (H. S. Nalwa and S. Miyata, Eds.), p. 813. CRC Press, New York, 1997.

C60 -Based Materials 117. M. Prato and M. Maggini, Acc. Chem. Res. 31, 519 (1998). 118. N. Martin, L. Sanchez, B. Illescas, and I. Perez, Chem. Rev. 98, 2527 (1998). 119. Y.-P. Sun and J. E. Riggs, Int. Rev. Phys. Chem. 18, 43 (1999). 120. M. Lee, O.-K. Song, J.-C. Seo, and D. Kim, Chem. Phys. Lett. 196, 325 (1992). 121. J. E. Riggs and Y.-P. Sun, J. Phys. Chem. A 103, 485 (1999). 122. S. Mishra, H. Rawat, M. P. Joshi, and S. Mehendale, J. Phys. B 27, L157 (1994). 123. B. Justus, Z. Kafafi, and A. Huston, Opt. Lett. 18, 19, 1603 (1993). 124. D. Gust, T. A. Moore, and A. L. Moore, Res. Chem. Intermed. 23, 621 (1997). 125. H. Imahori and Y. Sakata, Adv. Mater. 9, 537 (1997). 126. H. Imahori, K. Hagiwara, T. Akiyama, M. Aoki, S. Taniguchi, T. Okada, M. Shirakawa, and Y. Sakata, Chem. Phys. Lett. 263, 545 (1996). 127. Q. Xie, E. Perez-Cordero, and L. Echegoyen, J. Am. Chem. Soc. 114, 3978 (1992). 128. G. E. Lawson, A. Kitaygorodskiy, B. Ma, C. E. Bunker, and Y.-P. Sun, Chem. Commun. 2225 (1995). 129. K. F. Liou and C. H. Cheng, Chem. Commun. 15 (1996). 130. A. Hirsch, “The Chemistry of the Fullerenes,” Thieme, Stuttgart, 1994. 131. F. Diederich and C. Thilgen, Science 271, 31 (1996). 132. M. Prato, J. Mater. Chem. 7, 1097 (1997). 133. M. Prato and M. Maggini, Acc. Chem. Res. 31, 519 (1998). 134. E. K. Miller, K. Lee, K. Hasharoni, J. C. Hummelen, F. Wudl, and A. J. Heeger, J. Chem. Phys. 108, 1390 (1998). 135. K. Yoshino, K. Tada, A. Fujii, K. Hosoda, S. Kawabe, H. Kajii, M. Hirohata, R. Hidayat, H. Araki, A. A. Zakhidov, R. Sugimoto, M. Iyoda, M. Ishikawa, and T. Masuda, Fullerene Sci. Technol. 5, 1359 (1997). 136. K. Y. T. Akashi, S. Morita, M. Yoshida, M. Hamaguchi, K. Tada, A. Fujii, T. Kawai, S. Uto, M. Ozaki, M. Onoda, and A. A. Zakhidov, Synth. Met. 70, 1317 (1995). 137. R. E. Haufler, L.-S. Wang, L. P. F. Chibante, C.-M. Jin, and S. K. Kim, Chem. Phys. Lett. 179, 449 (1991). 138. D. Kim, M. Lee, Y. D. Suh, and S. K. Kim, J. Am. Chem. Soc. 114, 4429 (1992). 139. R. S. Ruoff, R. Malhotra, and D. L. Huestis, Nature 361, 140 (1993). 140. N. Sivaraman, R. Dhamodaran, I. Kaliappan, T. G. Srinivasan, P. R. V. Rao, and C. K. Mathews, J. Org. Chem. 57, 6077 (1992). 141. N. Sivaraman, R. Dhamodaran, I. Kaliappan, T. G. Srinivasan, P. R. V. Rao, and C. K. Mathews, Fullerene Sci. Technol. 2, 233 (1994). 142. N. Sivaraman, R. Dhamodaran, I. Kaliappan, T. G. Srinivasan, P. R. V. Rao, and C. K. Mathews, in “Recent Advances in the Chemistry and Physics of Fullerenes and Related Materials” (K. M. Kadish and R. S. Ruoff, Eds.), 1994, Rep. 1211. 143. R. S. Ruoff, D. S. Tse, R. Malhotra, and D. C. Lorents, J. Phys. Chem. 97, 3379 (1993). 144. W. A. Scrivens and J. M. Tour, J. Chem. Soc. Chem. Commun. 1207 (1993). 145. W. A. Scrivens, A. M. Cassell, K. E. Kinsey, and J. M. Tour, in “Recent Advances in the Chemistry and Physics of Fullerenes and Related Materials” (K. M. Kadish and R. S. Ruoff, Eds.), 1994. 146. M. T. Beck, G. Mandi, and S. Keki, Fullerene Sci. Technol. 2, 1510 (1995). 147. M. T. Beck and G. Mandi, Fullerene Sci. Technol. 3, 32 (1996). 148. M. T. Beck and G. Mandi, Fullerene Sci. Technol. 5, 291 (1997). 149. M. T. Beck, G. Mandi, and S. Keki, in “Fullerenes” (R. S. Ruoff and K. S. Kadish, Eds.), Vol. 2. The Electrochem. Soc., Pennington, NJ, 1995. 150. M. T. Beck and G. Mandi, in “Fullerenes” (R. S. Ruoff and K. S. Kadish, Eds.), Vol. 3. The Electrochem. Soc., Pennington, NJ, 1996.

C60 -Based Materials 151. M. T. Beck and G. Mandi, Fullerene Sci. Technol. 5, 291 (1997). 152. D. Heyman, Fullerene Sci. Technol. 4, 509 (1996). 153. X. H. Zhou, J. B. Liu, Z. X. Jin, Z. N. Gu, Y. Q. Wu, and Y. L. Sun, Fullerene Sci. Technol. 5, 285 (1997). 154. Y. Marcus, J. Phys. Chem. B 101, 8617 (1997). 155. Y. Marcus, A. L. Smith, M. V. Korobov, A. L. Mirakyan, N. Avramenko, and E. B. Stukalin, J. Phys. Chem. B 105, 2499 (2001). 156. N. Sivaraman, T. G. Srinivasan, P. R. V. Rao, and R. Natarajan, J. Chem. Inf. Comput. Sci. 41, 1067 (2001). 157. I. Z. Kiss, G. Mandi, and M. T. Beck, J. Phys. Chem. A 104, 8081 (2000). 158. C. Reichardt, “Solvents and Solvent Effects in Organic Chemistry,” 2nd ed., p. 534. VCH, Weinheim, 1998. 159. V. N. Bezmelnitsyn, A. V. Eletskii, and E. V. Stepanov, in “Progress in Fullerene Research” (H. Kuzmany et al., Eds.), p. 45. World Scientific, Singapore, 1994. 160. V. N. Bezmelnitsyn, A. V. Eletskii, and E. V. Stepanov, J. Phys. Chem. 98, 6665 (1994). 161. G. V. Andrievsky, M. V. Kosevich, O. M. Vovk, V. S. Shelkovsky, and L. A. Vashenko, J. Chem. Soc., Chem. Commun. 1281 (1995). 162. G. V. Andrievsky, M. V. Kosevich, O. M. Vovk, V. S. Shelkovsky, and L. A. Vashchenko, in “Chemistry and Physics of Fullerenes and Related Materials” (R. S. Ruo and K. M. Kadish, Eds.), Electrochem. Soc. Proc. Series, PV 95-10, p. 1591. Electrochem. Soc., Pennington, NJ, 1995. 163. G. V. Andrievsky, V. K. Klochkov, E. L. Karyakina, and N. O. Mchedlov-Petrossyan, Chem. Phys. Lett. 300, 392 (1999). 164. G. V. Andrievsky, V. K. Klochkov, A. Bordyuh, and G. I. Dovbeshko, Chem. Phys. Lett. 364, 8 (2002). 165. X. Wei, M. Wu, L. Qi, and Z. Xu, J. Chem. Soc., Perkin Trans. 2, 1389 (1997). 166. A. D. Roslyakov, G. V. Andrievsky, A. Yu. Petrenko, and L. T. Malaya, Zh. Akad. Med. Nauk. Ukrainy 5, 338 (1999) [in Russian]. 167. N. O. Mchedlov-Petrossyan, V. K. Klochkov, , and G. V. Andrievsky, J. Chem. Soc., Faraday Trans. 93, 4343 (1997). 168. B. V. Derjaguin, “The Theory of Stability of Colloids and Thin Films.” Nauka, Moscow, 1986 [in Russian]. 169. T. Steiner, Biophys. Chem. 95, 195 (2002). 170. T. R. Ohno, Y. Chen, S. E. Harvey, G. H. Kroll, J. H. Weaver, R. E. Haufler, and R. E. Smalley, Phys. Rev. B 44, 13747 (1991). 171. S. Modesti, S. Cerasari, and P. Rudolf, Phys. Rev. Lett. 71, 2469 (1993). 172. J. K. Gimzewski, S. Modesti, and R. R. Schlittler, Phys. Rev. Lett. 72, 1036 (1994). 173. G. K. Wertheim and D. N. E. Buchanan, Phys. Rev. B 50, 11070 (1994). 174. B. W. Hoogenboom, R. Hesper, L. H. Tjeng, and G. A. Sawatzky, Phys. Rev. B 57, 11939 (1998). 175. L. H. Tjeng, R. Hesper, A. C. L. Heessels, A. Heeres, H. T. Jonkman, and G. A. Sawazky, Solid State Commun. 103, 31 (1997). 176. R. Hesper, L. H. Tjeng, and G. A. Sawazky, Europhys. Lett. 40, 177 (1997). 177. S. J. Chase, W. S. Basca, M. G. Mitch, L. J. Pilione, and J. S. Lannin, Phys. Rev. B 46, 7873 (1992). 178. P. W. Murray, M. Ø. Pedersen, E. Laegsgaard, I. Stensgaard, and F. Besenbacher, Phys. Rev. B 55, 9360 (1997). 179. K.-D. Tsuei, J.-Y. Yuh, C.-T. Tzeng, R.-Y. Chu, S.-C. Chung, and K.-L. Tsang, Phys. Rev. B 56, 15412 (1997). 180. M. R. C. Hunt, S. Modesti, P. Rudolf, and R. E. Palmer, Phys. Rev. B 51, 10039 (1995). 181. T. Kobayashi, C. Tindall, O. Takaoka, Y. Hasegawa, and T. Sakurai, J. Korean Phys. Soc. 31, S5 (1997). 182. C. T. Chen, L. H. Tjeng, P. Rudolf, G. Meigs, J. E. Rowe, J. Chen, J. P. McCauley, Jr., A. B. Smith III, A. R. McGhie, W. J. Romanow, and E. W. Plummer, Nature 352, 603 (1991). 183. C.-T. Tzeng, W.-S. Lo, J.-Y. Yuh, R.-Y. Chu, and K.-D. Tsuei, Phys. Rev. B 61, 2263 (2000).

461 184. A. J. Maxwell, P. A. Brühwiler, S. Andersson, D. Arvanitis, B. Hernnäs, O. Karis, D. C. Mancini, N. Mårtensson, S. M. Gray, M. K.-J. Johansson, and L. S. O. Johansson, Phys. Rev. B 52, R5546 (1995). 185. A. J. Maxwell, P. A. Bruhwiler, D. Arvantis, J. Hasselstrom, M. K.-J. Johansson, and N. Martensson, Phys. Rev. B 57, 7312 (1998). 186. C. Cepek, A. Goldoni, and S. Modesti, Phys. Rev. B 53, 7466 (1996). 187. E. I. Altman and R. J. Colton, Surf. Sci. 279, 49 (1992). 188. E. I. Altman and R. J. Colton, Surf. Sci. 295, 13 (1993). 189. T. Hashizume and T. Sakurai, J. Vac. Sci. Technol. B 12, 1992 (1994). 190. T. David, J. K. Gimzewski, D. Purdie, B. Reihl, and R. R. Schlittler, Phys. Rev. B 50, 5810 (1994). 191. E. I. Altman and R. J. Colton, Phys. Rev. B 48, 18244 (1993). 192. K. Motai, T. Hashizume, H. Shinohara, Y. Saito, H. W. Pickering, Y. Nishina, and T. Sakurai, Jpn. J. Appl. Phys. 32, L450 (1993). 193. Y. Kuk, D. K. Kim, Y. D. Suh, K. H. Park, H. P. Noh, S. J. Oh, and S. K. Kim, Phys. Rev. Lett. 70, 1948 (1993). 194. M. Pedio, R. Felici, X. Torrelles, P. Rudolf, M. Capozi, J. Rius, and S. Ferrer, Phys. Rev. Lett. 85, 1040 (2000). 195. T. Hashizume, K. Motai, X. D. Wang, H. Shinohara, Y. Maruyama, K. Ohno, Y. Nishina, H. W. Piclering, Y. Kuk, and T. Sakurai, Phys. Rev. Lett. 71, 2959 (1993). 196. R. Gaisch, R. Berndt, J. K. Gimzewski, B. Reihl, R. R. Schlittler, W. D. Schneider, and M. Tschudy, J. Vac. Sci. Technol. B 12, 2153 (1994). 197. R. Fasel, P. Aebi, R. G. Agostino, D. Naumovic, J. Osterwalder, A. Santaniello, and L. Schlapbach, Phys. Rev. Lett. 76, 4733 (1996). 198. R. Fasel, R. G. Agostino, P. Aebi, and L. Schlapbach, Phys. Rev. B 60, 4517 (1999). 199. C. Cepek, R. Fasel, M. Sancrotti, T. Greber, and J. Osterwalder, Phys. Rev. B 63, 125406 (2001). 200. J. Weckesser, C. Cepek, R. Fasel, J. V. Barth, F. Baumberger, T. Greber, and K. Kern, J. Chem. Phys. 115, 9001 (2001). 201. J. K. Gimzewski, S. Modesti, T. David, and R. R. Schlittler, J. Vac. Sci. Technol. B 12, 1942 (1994). 202. S. Modesti, R. R. Schlittler, and J. K. Gimzewski, Surf. Sci. 331– 333, 1129 (1995). 203. V. Saltas and C. A. Papageorgopoulos, Surf. Sci. 488, 23 (2001). 204. M. Pedio, K. Hevesi, N. Zema, M. Capozi, P. Perfetti, R. Gouttebaron, J.-J. Pireaux, R. Caudano, and P. Rudolf, Surf. Sci. 437, 249 (1999). 205. X. D. Wang, T. Hashizume, H. Shinohara, Y. Saito, Y. Nishina, and T. Sakurai, Japan. J. Appl. Phys. 31, L983 (1992). 206. D. Chen and D. Sarid, Phys. Rev. B 49, 7612 (1994). 207. S. Suto, K. Sakamoto, T. Wakita, C. W. Hu, and A. Kasuya, Phys. Rev. B 56, 7439 (1997). 208. P. Moriarty, M. D. Upward, A. W. Dunn, Y.-R. Ma, P. H. Beton, and D. Teehan, Phys. Rev. B 57, 362 (1998). 209. K. Sakamoto, M. Harada, D. Kondo, A. Kimura, A. Kakizaki, and S. Suto, Phys. Rev. B 58, 13951 (1998). 210. C. Cepek, P. Schiavuta, M. Sancrotti, and M. Pedio, Phys. Rev. B 60, 2068 (1999). 211. Y. Z. Li, M. Chander, J. C. Partin, J. H. Weaver, L. P. F. Chibante, and R. E. Smalley, Phys. Rev. B 45, 13837 (1992). 212. J. G. Hou, J. Yang, H. Wang, Q. Li, C. Zeng, H. Lin, B. Wang, D. M. Chen, and Q. Zhu, Phys. Rev. Lett. 83, 3001 (1999). 213. H. Wang, C. Zeng, Q. Li, B. Wang, J. Yang, J. G. Hou, and Q. Zhu, Surf. Sci. 442, L1024 (1999). 214. D. Chen, J. Chen, and D. Sarid, Phys. Rev. B 50, 10905 (1994). 215. H. Xu, D. M. Chen, and W. N. Creager, Phys. Rev. Lett. 70, 1850 (1993). 216. Y. Fujikawa, K. Saiki, and A. Koma, Phys. Rev. B 56, 12124 (1997). 217. D. Chen and D. Sarid, Surf. Sci. 329, 206 (1995). 218. D. Klyachko and D. M. Chen, Phys. Rev. Lett. 75, 3693 (1995).

462 219. S. Suto, K. Sakamoto, D. Kondo, T. Wakita, A. Kimura, A. Kakizaki, C.-W. Hu, and A. Kasuya, Surf. Sci. 438, 242 (1999). 220. D. Chen, M. J. Gallagher, and D. Sarid, J. Vac. Sci. Technol. B 12, 1947 (1994). 221. X. Yao, T. G. Ruskell, R. K. Workman, D. Sarid, and D. Chen, Surf. Sci. 366, L743 (1996). 222. X. Yao, R. K. Workman, C. A. Peterson, D. Chen, and D. Sarid, Appl. Phys. A 66, S107 (1998). 223. A. V. Hamza, M. Balooch, and M. Moalem, Surf. Sci. 317, L1129 (1994). 224. D. Chen, R. Workman, and D. Sarid, Surf. Sci. 344, 23 (1995). 225. D. Klyachko and D. Chen, J. Vac. Sci. Technol. B 14, 974 (1996). 226. D. V. Klyachko, J.-M. Lopez-Castillo, J.-P. Jay-Gerin, and D. M. Chen, Phys. Rev. B 60, 9026 (1998). 227. R. D. Aburano, H. Hong, K.-S. Chung, M. C. Nelson, P. Zschack, H. Chen, and T.-C. Chiang, Phys. Rev. B 57, 6636 (1999). 228. H. Xu, D. M. Chen, and W. N. Creager, Phys. Rev. B 50, 8454 (1994). 229. A. Goldoni, C. Cepek, M. De Seta, J. Avila, M. C. Asensio, and M. Sancrotti, Phys. Rev. B 61, 10411 (2000). 230. Y. Z. Li, J. C. Patrin, M. Chander, J. H. Weaver, L. P. F. Chibante, and R. E. Smalley, Science 252, 547 (1991). 231. Y. Z. Li, J. C. Patrin, M. Chander, J. H. Weaver, L. P. F. Chibante, and R. E. Smalley, Science 253, 429 (1991). 232. Q. Xue, T. Ogino, Y. Hasegawa, H. Shinohara, and T. Sakurai, Phys. Rev. B 53, 1985 (1996). 233. T. Sakurai, Qikun Xue, T. Hashizume, and Y. Hasegawa, J. Vac. Sci. Technol. B 15, 1628 (1997). 234. A. R. Kortan, N. Kopylov, S. H. Glarum, E. M. Gyorgy, A. P. Ramirez, R. M. Fleming, F. A. Thiel, and R. C. Haddon, Nature 355, 529 (1992). 235. P. A. Heiney, J. E. Fischer, A. R. McGhie, W. J. Romanow, A. M. Denenstein, J. P. McCauley, Jr., A. B. Smith, and D. E. Cox, Phys. Rev. Lett. 66, 2911 (1991). 236. P. W. Stephens, L. Mihaly, P. L. Lee, R. L. Whetten, S.-M. Huang, R. B. Kaner, F. Diederich, and K. Holczer, Nature 351, 632 (1991). 237. L. Forró and L. Mihaly, Rep. Progr. Phys. 64, 649 (2001). 238. P. A. Heiney, G. B. M. Vaughan, J. E. Fischer, N. Coustel, D. E. Cox, J. R. D. Copley, D. A. Neumann, W. A. Kamitakahara, K. M. Creegan, D. M. Cox, J. P. McCauley, Jr., and A. B. Smith III, Phys. Rev. B 45, 4544 (1992). 239. J. E. Fischer, P. A. Heiney, A. R. McGhie, W. J. Romanow, A. M. Denenstein, J. P. McCauley, Jr., and A. B. Smith III, Science 252, 1288 (1991). 240. A. Lundin and B. Sundqvist, Phys. Rev. B 53, 8329 (1996). 241. J. E. Schirber, G. H. Kwei, D. Jorgensen, R. L. Hitterman, and B. Morosin, Phys. Rev. B 51, 12014 (1995). 242. X. D. Shi, A. R. Kortan, J. M. Williams, A. M. Kini, B. M. Saval, and P. M. Chaikin, Phys. Rev. Lett. 68, 827 (1992). 243. W. I. F. David, R. M. Ibberson, T. J. S. Dennis, J. P. Hare, and K. Prassides, Europhys. Lett. 18, 219 (1992). 244. G. A. Samara, J. E. Schirber, B. Morosin, L. V. Hansen, D. Loy, and A. P. Sylwester, Phys. Rev. Lett. 67, 3136 (1991). 245. G. A. Samara, L. V. Hansen, R. A. Assink, B. Morosin, J. E. Schirber, and D. Loy, Phys. Rev. B 47, 4756 (1993). 246. C. Pan, M. P. Sampson, Y. Chai, R. H. Hauge, and J. L. Margrave, J. Phys. Chem. 95, 2944 (1991). 247. J. Abrefah, D. R. Olander, M. Balooch, and W. J. Siekhaus, Appl. Phys. Lett. 60, 1313 (1992). 248. H. Kataura, N. Irie, N. Kobayashi, Y. Achiba, K. Kikuchi, T. Hanyu, and S. Yamaguchi, Jpn. J. Appl. Phys. 32, L1667 (1993). 249. M. B. Jost, N. Troullier, D. M. Poirier, J. L. Martins, J. H. Weaver, L. P. F. Chibante, and R. E. Smalley, Phys. Rev. B 44, 1966 (1991). 250. V. Rosato, M. Celino, G. Bvenedek, and S. Gaito, Phys. Rev. B 60, 16928 (1999). 251. S. Margadonna, C. M. Brown, T. J. S. Dennis, A. Lappas, P. Pattison, K. Prassides, and H. Shinohara, Chem. Mater. 10, 1742 (1998).

C60 -Based Materials 252. I. Margiolaki, S. Margadonna, K. Prassides, S. Assimopoulos, K. P. Meletov, G. A. Kourouklis, T. J. S. Dennis, and H. Shinohara, Physica B 318, 372 (2002). 253. G. B. M. Vaughan, P. A. Heiney, J. E. Fischer, D. E. Luzzi, D. A. Ricketts-Foot, A. R. McGhie, Y. W. Hui, A. L. Smith, D. E. Cox, W. J. Romanow, B. H. Allen, N. Coustel, J. P. McCauley, and A. B. Smith III, Science 254, 1350 (1991). 254. Y. Saito, N. Fujimoto, K. Kikychi, and Y. Achiba, Phys. Rev. B 49, 14794 (1994). 255. S. Margadonna, C. M. Brown, T. J. S. Dennis, A. Lappas, P. Pattison, K. Prassides, and H. Shinohara, Chem. Mater. 10, 1742 (1998). 256. F. Diederich, R. Ettl, Y. Rubin, R. L. Whetten, R. Beck, M. Alvarez, S. Anz, D. Sensharma, F. Wudl, K. C. Khemani, and A. Koch, Science 252, 548 (1992). 257. T. John, S. Dennis, T. Kai, T. Tomiyama, and H. Shinohara, Chem. Commun. 619 (1998). 258. Y. Maniwa, A. Ohi, K. Mizoguchi, K. Kume, K. Kikuchi, K. Saito, I. Ikemoto, S. Suzuki, and Y. Achiba, J. Phys. Soc. Jpn. 62, 1131 (1993). 259. Y. Maniwa, K. Kume, K. Kikuchi, K. Saito, I. Ikemoto, S. Suzuki, and Y. Achiba, Phys. Rev. B 53, 14196 (1996). 260. R. Almairac, D. Tranqui, J. P. Lauriat, J. Lapasset, and J. Moret, Solid State Commun. 106, 437 (1998). 261. P. G. Collins, J. C. Grossman, M. Cote, M. Ishigami, C. Piskoti, S. G. Louie, M. L. Cohen, and A. Zettl, Phys. Rev. Lett. 82, 165 (1998). 262. R. Kuzuo, M. Terauchi, and M. Tanaka, Phys. Rev. B 51, 11018 (1995). 263. M. Knupfer, O. Knauff, M. S. Golden, J. Fink, M. Burk, D. Fuchs, S. Schuppler, R. H. Michel, and M. M. Kappes, Chem. Phys. Lett. 258, 513 (1996). 264. M. Moalem, M. Balooch, A. V. Hamza, and R. S. Ruoff, J. Phys. Chem. 99, 16736 (1995). 265. V. Piacente, G. Gigli, P. Scardala, A. Giustini, and G. Bardi, J. Phys. Chem. 100, 9815 (1996). 266. O. V. Boltalina, V. Yu. Markov, A. Ya. Borschevskii, L. N. Sidorov, V. N. Bezmelnitsin, A. V. Eletskii, and R. Taylor, Rapid Commun. Mass Spectrom. 12, 1028 (1998). 267. B. Brunetti, G. Gigli, E. Giglio, V. Piacente, and P. Scardala, J. Phys. Chem. B 101, 10715 (1997). 268. V. Piacente, C. Palchetti, G. Gigli, and P. Scardala, J. Phys. Chem. A 101, 24, 4303 (1997). 269. H. Kataura, H. Y. Endo, T. Hanyu, S. Yamaguchi, Y. Achiba, and K. Kikuchi, J. Phys. Chem. Solid 58, 1913 (1997). 270. J. F. Armbruster, M. Roth, H. A. Romberg, M. Sing, M. Schmidt, P. Schweiss, P. Adelmann, M. S. Golden, J. Fink, R. H. Michel, J. Rockenberger, F. Hennrich, and M. M. Kappes, Phys. Rev. B 50, 4933 (1994). 271. R. C. Haddon, L. F. Schneemeyer, J. V. Waszczak, S. H. Glarum, R. Tycko, G. Dabbagh, A. R. Kortan, A. J. Muller, A. M. Mujsce, M. J. Rosseinsky, S. M. Zahurak, A. V. Makhija, F. A. Thiel, K. Raghavachari, E. Cockayne, and V. Elser, Nature 350, 46 (1991). 272. R. M. Fleming, A. P. Ramirez, M. J. Rosseinsky, D. W. Murphy, R. C. Haddon, S. M. Zahurak, and A. V. Makhija, Nature 352, 787 (1991). 273. P. A. Heiney, J. E. Fischer, A. R. McGhie, W. J. Romanow, A. M. Denestein, J. P. McCauley, Jr., A. B. Smith III, and D. E. Cox, Phys. Rev. Lett. 67, 1468 (1991). 274. R. Tycko, G. Dabbagh, R. M. Flemming, R. C. Haddon, A. V. Makhija, and S. M. Zahurak, Phys. Rev. Lett. 67, 1886 (1991). 275. C. S. Yannoni, R. D. Johnson, G. Meijer, D. S. Bethune, and J. R. Salem, J. Phys. Chem. 95, 9 (1991). 276. R. Tycko, R. C. Haddon, G. Dabbagh, S. H. Glarum, D. C. Douglass, and A. M. Mujsce, J. Phys. Chem. 95, 518 (1991). 277. D. A. Neumann, J. R. D. Copley, R. L. Cappelleti, W. A. Kamitakahara, R. M. Lindstrom, K. M. Creegan, D. M. Cox, W. J.

C60 -Based Materials

278.

279.

280. 281.

282. 283.

284. 285. 286.

287. 288. 289. 290. 291. 292. 293. 294. 295. 296. 297. 298. 299. 300. 301. 302. 303. 304. 305. 306. 307.

Romanow, N. Coustel, J. P. McCauley, Jr., N. C. Maliszewskyj, J. E. Fischer, and A. B. Smith III, Phys. Rev. Lett. 67, 3808 (1991). R. F. Kiefl, J. W. Schneider, A. MacFarlane, K. Chow, T. L. Duty, T. L. Wstle, B. Hitti, R. L. Lichti, E. J. Ansaldo, C. Schwab, P. W. Percival, G. Wei, S. Wlodek, K. Kojima, W. J. Romanow, J. P. McCaulay, Jr., N. Coustel, J. E. Fischer, and A. B. Smith III, Phys. Rev. Lett. 68, 1347 (1992). R. F. Kiefl, J. W. Schneider, A. MacFarlane, K. Chow, T. L. Duty, T. L. Wstle, B. Hitti, R. L. Lichti, E. J. Ansaldo, C. Schwab, P. W. Percival, G. Wei, S. Wlodek, K. Kojima, W. J. Romanow, J. P. McCaulay, Jr., N. Coustel, J. E. Fischer, and A. B. Smith III, Phys. Rev. Lett. 68, 2708 (1992). R. D. Johnson, C. S. Yannoni, H. C. Dorn, J. R. Salem, and D. S. Bethume, Science 255, 1235 (1992). P. C. Chow, X. Jiang, G. Reiter, P. Wochner, S. C. Moss, J. D. Axe, J. C. Hanson, R. K. McMullan, R. L. Meng, and C. W. Chu, Phys. Rev. Lett. 69, 2943 (1992). L. Pintschovius, S. L. Chaplot, G. Roth, and G. Heger, Phys. Rev. Lett. 75, 2843 (1995). R. J. Papoular, G. Roth, G. Heger, M. Haluska, and H. Kuzmany, in “Electronic Properties of Fullerenes” (H. Kuzmany, J. Fink, M. Mehring, and S. Roth, Eds.), Springer Series in Solid State Sciences, Vol. 117, p. 189. Springer, Berlin/Heidelberg, 1993. W. I. F. David, R. M. Ibberson, and T. Matsuo, Proc. Roy. Soc. London Ser. A 442, 129 (1993). L. Pintschovius, Rep. Progr. Phys. 57, 473 (1996). J. D. Axe, S. C. Moss, and D. A. Neumann, in “Solid State Physics” (H. Ehrenreich and F. Spaepen, Eds.), Vol. 48, p. 149. Academic, New York, 1994. J. Yu, R. K. Kalia, and P. Vashishta, Appl. Phys. Lett. 63, 3152 (1993). R. A. Jishi, R. M. Mirie, and M. S. Dresselhaus, Phys. Rev. B 45, 13685 (1992). J. L. Feldman, J. Q. Broughton, L. L. Boyer, D. E. Reich, and M. D. Kluge, Phys. Rev. B 46, 12731 (1992). D. E. Weeks and W. G. Harter, Chem. Phys. Lett. 144, 366 (1988). E. Brendsdal, B. N. Cyvin, J. Brunvoll, and S. J. Cyvin, Spectrosc. Lett. 21, 313 (1988). S. J. Cyvin, E. Brendsdal, B. N. Cyvin, and J. Brunvoll, Chem. Phys. Lett. 143, 377 (1988). R. S. Ruoff and A. L. Ruoff, Appl. Phys. Lett. 59, 1553 (1991). Z. C. Wu, D. A. Jelski, and T. F. George, Chem. Phys. Lett. 137, 291 (1987). R. E. Stanton and M. D. Newton, J. Phys. Chem. 92, 2141 (1988). Z. Slanina, J. M. Rudzinski, M. Togasi, and E. Osawa, J. Mol. Struct. 202, 169 (1989). F. Negri, G. Orlandi, and F. Zerbetto, Chem. Phys. Lett. 144, 31 (1988). G. B. Adams, J. B. Page, O. F. Sankey, K. Sinha, J. Menendez, and D. R. Huffman, Phys. Rev. B 44, 4052 (1991). R. Jones, C. D. Latham, M. I. Heggie, V. J. B. Torres, S. Öberg, and S. K. Estreicher, Philos. Mag. Lett. 65, 291 (1992). J. Kohanoff, W. Andreoni, and M. Parrinello, Phys. Rev. B 46, 4371 (1992). Z. H. Dong, P. Zhou, J. M. Holden, P. C. Eklund, M. S. Dresselhaus, and G. Dresselhaus, Phys. Rev. B 48, 2862 (1993). M. Martin, J. Fabian, J. Godard, P. Bernier, J. M. Lambert, and L. Mihaly, Phys. Rev. B 51, 2844 (1995). K. A. Wang, A. M. Rao, P. C. Eklund, M. S. Dresselhaus, and G. Dresselhaus, Phys. Rev. B 48, 11375 (1993). V. Schettino, M. Pagliai, and G. Cardini, J. Phys. Chem. A 106, 1815 (2002). V. Schettino, M. Pagliai, L. Ciabini, and G. Cardini, J. Phys. Chem. A 105, 11192 (2001). C. H. Choi, M. Kertesz, and L. Mihaly, J. Phys. Chem. A 104, 102 (2000). P. Giannozzi and S. Baroni, J. Chem. Phys. 100, 8537 (1994).

463 308. G. B. Adams, J. B. Page, O. F. Sankey, and M. O’ Keeffe, Phys. Rev. B 50, 17471 (1994). 309. J. Menéndez and J. B. Page, Light scattering in solids VIII, in “Topics in Applied Physics” (M. Cardona and G. Güntherodt, Eds.), Vol. 76, p. 27. Springer-Verlag, Berlin/Heidelberg, 2000. 310. M. C. Martin, X. Q. Du, J. Kwon, and L. Mihaly, Phys. Rev. B 50, 173 (1994). 311. V. Schettino, P. R. Salvi, R. Bini, and G. Cardini, J. Chem. Phys. 101, 11079 (1994). 312. P. Bowmar, W. Hayes, M. Kurmoo, P. A. Pattenden, M. A. Green, P. Day, and K. Kikuchi, J. Phys.: Condens. Matter 6, 3161 (1994). 313. L. Akselrod, H. J. Byrne, S. Donovan, and S. Roth, Chem. Phys. 192, 307 (1995). 314. W. Brockner and F. Menzel, J. Mol. Struct. 378, 147 (1996). 315. A. Graja, A. Lapinski, and S. Król, J. Mol. Struct. 404, 147 (1997). 316. M. S. Dresselhaus, G. Dresselhaus, and P. C. Eklund, “Science of Fullerenes and Carbon Nanotubes.” Academic, San Diego, 1995. 317. M. S. Dresselhaus, G. Dresselhaus, P. C. Eklund, and R. Saito, in “Optical and Electronic Properties of Fullerenes and FullereneBased Materials” (J. Shinar, Z. V. Vardeny, and Z. H. Kafafi, Eds.), Vol. 8, p. 217. Dekker, New York, 1999. 318. V. A. Davydov, L. S. Kashevarova, A. V. Rakhmanina, V. M. Senyavin, R. Ceolin, H. Szwarc, H. Allouchi, and V. Agafonov, Phys. Rev. B 61, 11936 (2000). 319. K. A. Wang, Y. Wang, P. Zhou, J. M. Holden, S. L. Ren, G. T. Hager, H. F. Ni, P. C. Eklund, G. Dresselhaus, and M. S. Dresselhaus, Phys. Rev. B 45, 1955 (1992). 320. D. S. Bethune, G. Meijer, W. C. Tang, and H. J. Rosen, Chem. Phys. Lett. 174, 219 (1990). 321. P. Zhou, K. A. Wang, Y. Wang, P. C. Eklund, M. S. Dresselhaus, G. Dresselhaus, and R. A. Jishi, Phys. Rev. B 46, 2595 (1992). 322. P. Zhou, A. M. Rao, K. A. Wang, J. D. Robertson, C. Eloi, M. S. Meier, S. L. Ren, X. X. Bi, P. C. Eklund, and M. S. Dresselhaus, Appl. Phys. Lett. 60, 2871 (1992). 323. S. J. Duclos, R. C. Haddon, S. H. Glarum, A. F. Hebard, and K. B. Lyons, Science 254, 1625 (1991). 324. H. Kuzmany, R. Winkler, and T. Pichler, J. Phys.: Condens. Matter 7, 6601 (1995). 325. J. M. Hawkins, A. Mayer, S. Loren, and R. Nunlist, J. Am. Chem. Soc. 113, 9394 (1991). 326. S. P. Love, D. McBranch, M. I. Salkola, N. V. Coppa, J. M. Robinson, B. I. Swanson, and A. R. Bishop, Chem. Phys. Lett. 225, 170 (1994). 327. A. Rosenberg and C. Kendziora, Phys. Rev. B 51, 9321 (1995). 328. M. C. Martin, J. Fabian, J. Godard, P. Bernier, J. M. Lambert, and L. Mihaly, Phys. Rev. B 51, 2844 (1995). 329. P. J. Horoyski, M. L. W. Thewalt, and T. R. Anthony, Phys. Rev. Lett. 74, 194 (1995). 330. S. Suha, J. Menendez, J. B. Page, G. B. Adams, G. S. Spencer, J. P. Lehman, P. Giannozzi, and S. Baroni, Phys. Rev. Lett. 72, 3359 (1994). 331. J. Menendez, J. B. Page, and S. Guha, Philos. Mag. B 70, 651 (1994). 332. S. Guha, J. Menendez, J. B. Page, and G. M. Adams, Phys. Rev. B 56, 15431 (1997). 333. J. R. Soto, A. Calles, and J. L. Moran-Lopez, J. Chem. Phys. 113, 1055 (2000). 334. R. Heid, L. Pintschovius, and J. M. Godard, Phys. Rev. B 56, 5925 (1997). 335. R. L. Cappeletti, J. R. D. Copley, W. A. Kamitakahara, F. Li, J. S. Lannin, and D. Ramage, Phys. Rev. Lett. 66, 3261 (1991). 336. K. Prassides, T. J. S. Dennis, J. P. Hare, J. Tomkinson, H. W. Kroto, R. Taylor, and D. R. M. Walton, Chem. Phys. Lett. 187, 455 (1991). 337. C. Coulombeau, H. Jobic, P. Bernier, C. Fabre, D. Schultz, and A. Rassat, J. Phys. Chem. 96, 22 (1992). 338. C. Coulombeau, H. Jobic, C. J. Carlile, S. M. Bennington, C. Fabre, and A. Rassat, Fullerene Sci. Technol. 2, 247 (1994).

464 339. F. Gompf, B. Renker, W. Schober, P. Adelmann, and R. Heid, J. Supercond. 7, 643 (1994). 340. J. R. D. Copley, D. A. Neumann, and W. A. Kamitakahara, Can. J. Phys. 73, 763 (1995). 341. G. Gensterblum, L.-M. Yu, J. J. Pireaux, P. A. Thiry, R. Caudano, P. Lambin, A. A. Lucas, W. Kratschmer, and J. E. Fischer, J. Phys. Chem. Solids 53, 1427 (1992). 342. A. A. Lucas, J. Phys. Chem. Solids 53, 1415 (1992). 343. G. Gensterblum, L. M. Yu, J. J. Pireaux, P. A. Thiry, R. Caudano, J. M. Themlin, S. Bouzidi, F. Coletti, and J. M. Debever, Appl. Phys. A 56, 175 (1993). 344. G. Gensterblum, J. J. Pireaux, P. A. Thiry, R. Caudano, J. P. Vigneron, P. Lambin, A. A. Lucas, and W. Kratschmer, Phys. Rev. Lett. 67, 2171 (1991). 345. A. Lucas, G. Gensterblum, J. J. Pireaux, P. A. Thiry, R. Taylor, and P. Lambin, Phys. Rev. B 45, 13692 (1991). 346. M. K. Nissen, S. M. Wilson, and M. L. W. Thewalt, Phys. Rev. Lett. 69, 2423 (1992). 347. A. Sassara, G. Zerza, and M. Chergui, J. Phys. B 29, 4997 (1996). 348. A. Sassara, G. Zerza, and M. Chergui, Chem. Phys. Lett. 261, 213 (1996). 349. W. I. F. David, R. M. Ibberson, T. J. S. Dennis, J. P. Hare, and K. Prassides, Europhys. Lett. 18, 735 (1992). 350. T. Atake, T. Tanaka, R. Kawaji, K. Kikuchi, K. Saito, S. Suzuki, I. Ikemoto, and Y. Achiba, Physica C 185–189, 427 (1991). 351. W. P. Beyermann, M. F. Hundley, J. D. Thompson, F. N. Diederich, and G. Gruner, Phys. Rev. Lett. 68, 2046 (1992). 352. R. C. Yu, N. Tea, M. B. Salamon, D. Lorens, and R. Malhotra, Phys. Rev. Lett. 68, 2050 (1992). 353. H.-B. Bürgi, E. Blanc, D. Schwarzenbach, S. Liu, Y. Lu, M. M. Kappes, and J. A. Ibers, Angew. Chem. Int. Ed. Engl. 31, 640 (1992). 354. P. Heiney, J. Phys. Chem. Solids 53, 1333 (1992). 355. J. D. Axe, S. C. Moss, and D. A. Neumann, Solid State Phys. 48, 149 (1994). 356. O. Blaschko, G. Krexner, Ch. Maier, and R. Karawatzki, Phys. Rev. B 56, 2288 (1997). 357. F. Gugenberger, R. Heid, C. Meingast, P. Adelmann, M. Braun, H. Wuhl, M. Haluska, and H. Kuzmany, Phys. Rev. Lett. 69, 3774 (1992). 358. W. Schranz, A. Fuith, P. Dolinar, H. Warhanek, M. Haluska, and H. Kuzmany, Phys. Rev. Lett. 71, 1561 (1993). 359. G. B. Alers, B. Golding, A. R. Kortan, R. C. Haddon, and F. A. Theil, Science 257, 511 (1992). 360. D. A. Neumann, J. R. D. Copley, W. A. Kamitakahara, J. J. Rush, R. L. Paul, and R. M. Lindstrom, J. Phys. Chem. Solids 54, 1699 (1993). 361. T. Matsuo, H. Suga, W. I. F. David, R. M. Ibberson, P. Bernier, A. Zahba, C. Fabre, A. Rassat, and A. Dworkin, Solid State Commun. 83, 711 (1992). 362. R. Sachidanandam and A. B. Harris, Phys. Rev. Lett. 66, 1467 (1991). 363. A. Dworkin, H. Szwarc, S. Leach, J. P. Hare, T. J. Dennis, H. W. Kroto, R. Taylor, and D. R. M. Walton, C. R. Acad. Sci., Paris Ser. II 312, 979 (1991). 364. R. D. Johnson, G. Meijer, and D. S. J. Bethune, J. Am. Chem. Soc. 112, 8983 (1990). 365. R. Blinc, J. Seliger, Dolinsek, and D. Arcon, Eur. Phys. Lett. 23, 355 (1993). 366. A. Dworkin, C. Fabre, D. Schutz, G. Kriza, R. Ceolin, H. Szwarc, P. Bernier, D. Jerome, S. Leach, A. Rassat, J. P. Hare, T. J. Dennis, H. W. Kroto, R. Taylor, and D. R. M. Walton, C. R. Acad. Sci., Paris Ser. II 313, 1017 (1991). 367. E. Grivei, B. Nysten, M. Cassart, A. Demain, and J. P. Issi, Solid State Commun. 85, 73 (1993). 368. G. Pitsi, J. Caerels, and J. Thoen, Phys. Rev. B 55, 915 (1997).

C60 -Based Materials 369. L. Pintschovius, B. Renker, F. Gompf, R. Heid, S. L. Chaplot, M. Haluška, and H. Kizmany, Phys. Res. Lett. 69, 2662 (1992). 370. D. A. Neumann, J. R. D. Copley, W. A. Kamitakahara, J. J. Rush, R. L. Cappelletti, N. Coustel, J. E. Fischer, J. P. McCauley, Jr., A. B. Smith III, K. M. Creegan, and D. M. Cox, J. Chem. Phys. 96, 8631 (1992). 371. G. Van Tendeloo and S. Amelinckx, in “Electron Microscopy of Fullerenes and Related Materials, Characterization of Nanophase Materials” (Z. L. Wang, Ed.), Ch. 12. Wiley–VCH, Weinheim, 2000. 372. S. Hoen, N. G. Chopra, X. D. Xiang, R. Mostovoy, J. Hou, W. A. Vareka, and A. Zettl, Phys. Rev. B 46, 12737 (1992). 373. P. H. M. van Loosdrecht, P. J. M. van Bentum, and G. Meijer, Phys. Rev. Lett. 68, 1176 (1992). 374. P. H. M. van Loosdrecht, P. J. M. van Bentum, M. A. Verheijen, and J. Meijer, Chem. Phys. Lett. 198, 587 (1992). 375. M. Matus and H. Kuzmany, Appl. Phys. A 56, 241 (1993). 376. J. De Bruyn, A. Dworkin, H. Szwarc, J. Godard, R. Céolin, C. Fabre, and A. Rassart, Europhys. Lett. 24, 551 (1993). 377. W. Luo, H. Wang, R. S. Ruoff, J. Cioslowski, and S. Phelps, Phys. Rev. Lett. 73, 186 (1994). 378. P. Byszewski, F. L. Diduszko, E. Kowalska, J. Fink-Finowicki, and A. Witowski, Appl. Phys. Lett. 61, 2981 (1992). 379. E. Grivei, B. Nysten, M. Cassart, J. P. Issi, L. Langer, B. Nysten, J. P. Michenaud, C. Fabre, and A. Rassat, Phys. Rev. B 48, 8514 (1993). 380. N. H. Tea, R. C. Yu, M. B. Salamon, D. C. Lorents, R. Malhotra, and R. S. Ruoff, Appl. Phys. A 56, 219 (1993). 381. G. Van Tendeloo, S. Amelinckx, M. A. Verheijen, P. H. M. van Loosdrecht, and G. Meijer, Phys. Rev. Lett. 69, 1065 (1992). 382. P. J. Benning, F. Stepniak, and J. H. Weaver, Phys. Rev. B 48, 9086 (1993). 383. C. Cepek and S. Modesti, Phys. Rev. B 54, 2890 (1996). 384. A. Glebov, V. Senz, J. P. Toennies, and G. Gensterblum, J. Appl. Phys. 82, 2329 (1997). 385. D. Passerone and E. Tosatti, Surf. Rev. Lett. 4, 859 (1997). 386. Z. Y. Li, Surf. Sci. 441, 366 (1999). 387. H. Wang, C. Zeng, B. Wang, J. G. Hou, Q. Li, and J. Yang, Phys. Rev. B 63, 85417 (2001). 388. C. Laforge, D. Passerone, A. B. Harris, P. Lambin, and E. Tosatti, Phys. Rev. Lett. 87, 85503 (2001). 389. D. M. Poirier, D. W. Owens, and J. H. Weaver, Phys. Rev. B 51, 1830 (1995). 390. R. Glang, in “Handbook of Thin Film Technology” (L. I. Maissel and R. Glang, Eds.), Ch. 1. McGraw–Hill, New York, 1970. 391. A. F. Hebard, R. C. Haddon, R. M. Fleming, and A. R. Kortan, Appl. Phys. Lett. 59, 2109 (1991). 392. R. M. Fleming, A. R. Kortan, B. Hessen, T. SieGrist, F. A. Thiel, R. C. Haddon, R. Tycho, G. Dabbaagh, M. L. Kaplan, and A. M. Mujscc, Phys. Rev. B 44, 888 (1991). 393. B. Morosin, P. P. Newcomer, R. J. Baughman, E. L. Venturini, D. Loy, and J. E. Schirber, Physica C 184, 21 (1991). 394. N. A. Fortune, K. Murata, F. Iga, Y. Nishihara, K. Kikuchi, S. Suzuki, I. Ikemoto, and Y. Achiba, Physica C 185–189, 425 (1991). 395. H. Ogata, T. Inable, H. Hoshi, Y. Maruyama, Y. Achiba, S. Suzuki, K. Kikuchi, and I. Ikemoto, Jpn. J. Appl. Phys. 31, L166 (1992). 396. D. L. Dorset, J. Phys. Chem. 99, 16748 (1995). 397. R. A. Assink, J. E. Schirber, D. A. Loy, B. Morosin, and G. A. Carlson, J. Mater. Res. 7, 2136 (1992). 398. K. Kikychi, S. Suzuki, K. Saito, H. Shiromaru, I. Ikeromoto, and Y. Achiba, Physica C 185–189, 415 (1991). 399. S. Pekker, G. Faigel, K. Fodorcsorba, L. Granasy, E. Jakab, and M. Tegze, Solid State Commun. 83, 423 (1992). 400. F. Michaud, M. Barrio, S. Toscani, D. O. Lopez, J. Ll. Tamatit, V. Agafonov, H. Szwarc, and R. Ceolin, Phys. Rev. B 57, 10351 (1998).

C60 -Based Materials 401. S. Toscani, H. Allouchi, J. Ll. Tamarit, D. O. López, M. Barrio, V. Agafonov, A. Rassat, H. Szwarc, and R. Céolin, Chem. Phys. Lett. 330, 491 (2000). 402. R. L. Meng, D. Ramirez, X. Jiang, P. C. Chow, C. Diaz, K. Matsuishi, S. C. Moss, P. H. Hor, and C. W. Chu, Appl. Phys. Lett. 59, 3402 (1991). 403. J. Z. Liu, J. W. Dykes, M. D. Lan, P. Klavins, R. N. Shelton, and M. M. Olmstead, Appl. Phys. Lett. 62, 531 (1992). 404. J.-B. Shi, W.-Y. Chang, S.-R. Su, and M.-W. Lee, Jpn. J. Appl. Phys. 35, L45 (1996). 405. X. D. Xiang, J. G. Hou, G. Briceno, W. A. Vareka, R. Mostovoy, A. Zettl, V. H. Crespi, and M. L. Cohen, Science 256, 1190 (1992). 406. J. Li, S. Komiya, T. Tamura, C. Nagasaki, J. Kihara, K. Kishio, and K. Kitazawa, Physica C 195, 205 (1992). 407. M. Haluska, H. Kuzmany, M. Vybornov, P. Rogl, and P. Fejdi, Appl. Phys. A 56, 161 (1993). 408. M. Tan, B. Xu, H. Li, Z. Qi, and Y. Xu, J. Crystal Growth 182, 375 (1992). 409. K.-C. Chiu, J.-S. Wang, C.-Y. Lin, T.-Y. Lin, and C.-S. Ro, Mater. Res. Bull. 30, 883 (1995). 410. K. Matsumoto, E. Schonherr, and M. Wojnowski, J. Crystal Growth 135, 154 (1994). 411. K. Murakami, K. Matsumoto, H. Nagatomo, E. Schönherr, and M. Wojnowski, Ultramicroscopy 73, 191 (1998). 412. H. Li, Y. Xu, J. Zhang, P. He, H. Li, T. Wu, and S. Bao, Progr. Natural Sci. 11, 427 (2001). 413. A. Al-Mohamad and A. W. Allaf, Synth. Metals 104, 39 (1999). 414. R. C. Haddon, A. S. Perel, R. C. Morris, T. T. M. Palstra, A. F. Hebard, and R. M. Fleming, Appl. Phys. Lett. 67, 121 (1995). 415. C. P. Jarret, K. Pichler, R. Newbould, and R. H. Friend, Synth. Metals 77, 35 (1996). 416. K. Horiuchi, K. Nakada, S. Uchino, S. Hashii, A. Hashimoto, N. Aoki, and Y. Ochiai, Appl. Phys. Lett. 81, 1911 (2002). 417. W. Krakow, N. M. Rivera, R. A. Roy, R. S. Ruoff, and J. J. Cuomo, Appl. Phys. A 56, 185 (1993). 418. Z. Dai, H. Naramoto, K. Narumi, S. Yamamoto, and A. Miyashita, Thin Solid Films 360, 28 (2000). 419. E. A. Katz, D. Faiman, S. Shtutina, and A. Isakina, Thin Solid Film 368, 49 (2000). 420. W. M. Tong, D. A. A. Ohlberg, H. K. You, R. S. Williams, S. J. Anz, M. M. Alvarez, R. L. Whetten, Y. Rubin, and F. N. Diederich, J. Phys. Chem. 95, 4709 (1991). 421. Z.-M. Ren, Z.-F. Ying, X.-X. Xiong, M.-Q. He, Y.-F. Li, F.-M. Li, and Y.-C. Du, J. Phys. D 27, 1499 (1994). 422. S. Isoda, H. Kawakubo, S. Nishikawa, and O. Wada, Nucl. Instrum. Methods B 112, 94 (1996). 423. H. Gao, Z. Xue, and Sh. Pang, J. Phys. D 29, 1870 (1996). 424. X. Shi, X. J. Fan, H. X. Guo, and Q. Fu, Solid State Commun. 99, 445 (1996). 425. X. Zou, S. Zhu, J. Xie, and J. Feng, J. Crystal Growth 200, 441 (1999). 426. J. G. Hou, J. Zeng, Y. Li, and Z. Q. Wu, Thin Solid Films 320, 179 (1998). 427. D. Schmicker, S. Schmidt, J. G. Skofronick, J. P. Toennies, and R. Vollmer, Phys. Rev. B 44, 10995 (1991). 428. J. E. Fischer, E. Werva, and P. A. Heincy, Appl. Phys. A 56, 193 (1993). 429. T. Thundat, R. J. Warmack, D. Ding, and F. L. N. Compton, Appl. Phys. Lett. 63, 891 (1993). 430. A. Fatash, Appl. Phys. Lett. 64, 1877 (1994). 431. S. Henke, K. M. Thürer, and S. Geier, Appl. Phys. A 60, 383 (1995). 432. G. Ma, Y. Yang, and G. Chen, Mater. Lett. 31, 297 (1997). 433. D. Stifter and H. Sitter, Appl. Phys. Lett. 66, 679 (1995). 434. D. Stifter and H. Sitter, J. Cryst. Growth 156, 79 (1995). 435. D. Stifter and H. Sitter, Thin Solid Films 280, 83 (1996). 436. J. K. N. Linder, S. Henke, B. Rauschenbach, and B. Stritzker, Thin Solid Films 279, 106 (1996).

465 437. K. Tanigaki, S. Kuroshima, J. Fujita, and T. W. Ebbesen, Appl. Phys. Lett. 63, 2351 (1993). 438. K. Tanigaki, S. Kuroshima, and T. W. Ebbesen, Thin Solid Films 257, 154 (1995). 439. H. Gaber, Appl. Phys. Lett. 65, 378 (1994). 440. J. Fujita, S. Kuroshima, T. Satoh, J. S. Tsai, T. W. Ebbesen, and K. Tanigaki, Appl. Phys. Lett. 63, 1008 (1993). 441. D. J. Kenny and R. E. Palmer, Surf. Sci. 447, 126 (2000). 442. M. Sakurai, H. Tada, K. Saiki, and A. Koma, Jpn. J. Appl. Phys. 30, L1892 (1991). 443. M. Sakurai, H. Tada, K. Saiki, A. Koma, H. Funasaka, and Y. Kishimoto, Chem. Phys. Lett. 208, 425 (1993). 444. G. Chen and G. Ma, Thin Solid Films 323, 309 (1998). 445. D. Bernaerts, G. Van Tendeloo, S. Amelinckx, K. Hevesi, G. Gensterblum, L. M. Yu, J.-J. Pireaux, F. Grey, and J. Bohr, J. Appl. Phys. 80, 3310 (1996). 446. U. D. Schwarz, W. Allers, G. Gensterblum, J.-J. Pireaux, and R. Wiesendanger, Phys. Rev. B 52, 5967 (1995). 447. J.-M. Themlin, S. Bouzidi, F. Coletti, J.-M. Debever, G. Gensterblum, L.-M. Yu, J.-J. Pireaux, and P. A. Thiry, Phys. Rev. B 46, 15 602 (1992). 448. G. Gensterblum, L. M. Yu, J.-J. Pireaux, P. A. Thiry, R. Caudano, Ph. Lambin, A. A. Lucas, W. Kräschmer, and J. E. Fischer, J. Phys. Chem. Solids 53, 1427 (1992). 449. J. Dura, P. M. Pippenger, N. J. Halas, X. Z. Xiong, P. C. Chow, and S. C. Moss, Appl. Phys. Lett. 63, 3443 (1993). 450. T. Ichihashi, K. Tanigaki, T. W. Ebbesen, S. Kuroshima, and S. Iijima, Chem. Phys. Lett. 190, 179 (1992). 451. Z. Dai, H. Naramoto, K. Narumi, S. Yamamoto, and A. Miyashita, Appl. Phys. Lett. 74, 1686 (1999). 452. Y. Kim, L. Jiang, T. Iyoda, K. Hashimoto, and A. Fujishima, Appl. Phys. Lett. 71, 3489 (1997). 453. Y. Kim, L. Jiang, T. Iyoda, K. Hashimoto, and A. Fujishima, Appl. Surf. Sci. 130–132, 602 (1998). 454. H. Yanagi and T. Sasaki, Appl. Phys. Lett. 65, 1222 (1994). 455. H. Yanagi, S. Doumi, T. Sasaki, and H. Tada, J. Appl. Phys. 80, 4990 (1996). 456. E. J. Snyder, M. S. Anderson, W. M. Tong, R. S. Williams, S. J. Anz, M. M. Alvarez, Y. Rubin, F. N. Diederich, and R. L. Whetten, Science 253, 171 (1991). 457. S. Fölsch, T. Maruno, A. Yamashita, and T. Hayashi, Appl. Phys. Lett. 62, 2643 (1993). 458. A. Fartash, Appl. Phys. Lett. 67, 3901 (1995). 459. A. Fartash, Phys. Rev. B 52, 7883 (1995). 460. G. Costantini, S. Rusponi, E. Giudice, C. Boragno, and U. Valbusa, Carbon 37, 727 (1999). 461. T. Hashizume, K. Motai, X. D. Wang, H. Shinohara, Y. Saito, Y. Maruyama, K. Ohno, Y. Kawazoe, Y. Nishina, H. W. Pickering, Y. Kuk, and T. Sakurai, Phys. Rev. Lett. 71, 2959 (1993). 462. A. Fartash, J. Appl. Phys. 79, 742 (1996). 463. J. Zeng, Y. Wang, Y. Li, and J. G. Hou, Physica C 282–287, 739 (1997). 464. W. Xu and J. G. Hou, J. Appl. Phys. 86, 4660 (1999). 465. J. G. Hou, W. Xu, H. Wang, and Y. Li, J. Appl. Phys. 84, 2906 (1998). 466. S. Saito and A. Oshiyama, Phys. Rev. Lett. 66, 2637 (1991). 467. N. Troullier and J. L. Martins, Phys. Rev. B 46, 1754 (1992). 468. M. Knupfer, Surf. Sci. Rep. 42, 1 (2001). 469. E. L. Shirley and S. G. Louie, Phys. Rev. Lett. 71, 133 (1993). 470. J. H. Weaver, J. Phys. Chem. Solids 53, 1433 (1992). 471. R. W. Lof, M. A. van Veenendaal, B. Koopmans, H. T. Jonkman, and G. A. Sawatzky, Phys. Rev. Lett. 68, 3924 (1992). 472. M. Skibowski and L. Kipp, J. Electron Spectrosc. Relat. Phenom. 68, 77 (1994). 473. T. L. Makarova, Semiconductors 35, 243 (2001). 474. B. Mishori, E. A. Katz, D. Faiman, and Y. Shapira, Solid State Commun. 102, 489 (1997).

466 475. E. Sohmen, J. Fink, and W. Krätchmer, Z. Phys. B 86, 87 (1992). 476. T. Rabenau, A. Simon, R. K. Kremer, and E. Sohmen, Z. Phys. B 90, 69 (1993). 477. T. Ohgami, Y. Shimada, H. Kubota, H. Tanaka, S. Matsuzaki, and M. Nagata, Physica B 239, 32 (1997). 478. P. Mondal, P. Lunkenheimer, and A. Loidl, Z. Phys. B 99, 527 (1996). 479. M. Hosoya, K. Ichimura, Z. H. Wang, G. Dresselhaus, M. S. Dresselhaus, and P. C. Eklund, Phys. Rev. B 49, 4981 (1994). 480. J. Hora, P. Pánek, K. Navrátil, B. Handlírová, J. Humlícek, H. Sitter, and D. Stifer, Phys. Rev. B 54, 5106 (1996). 481. A. Skumanich, Chem. Phys. Lett. 182, 486 (1991). 482. R. Schwedhelm, L. Kipp, A. Dallmeyer, and M. Skibowski, Phys. Rev. B 58, 13176 (1998). 483. B. Mishori, E. A. Katz, D. Faiman, A. Belu-Marian, and Y. Shapira, Fullerene Sci. Technol. 6, 113 (1998). 484. E. A. Katz, D. Faiman, B. Mishori, Y. Shapira, A. I. Shames, S. Shtutina, and S. Goren, J. Appl. Phys. 84, 3333 (1998). 485. M. Kaiser, W. K. Maser, H. J. Byme, J. Beichenbach, J. Anders, A. Mittelbach, and S. Roth, in “Electronic Properties of Fullerenes” (H. Kuzmany, J. Fink, M. Mehring, and S. Roth, Eds.), Vol. 117, p. 418. Springer-Verlag, Berlin/Heidelberg, 1993. 486. S. Krummacher, M. Biermann, M. Neeb, A. Liebsch, and W. Eberhardt, Phys. Rev. B 48, 8424 (1993). 487. E. L. Shirley, L. X. Benedict, and S. G. Louie, Phys. Rev. B 54, 10970 (1996). 488. M. Knupfer and J. Fink, Phys. Rev. B 60, 10731 (1999). 489. K. Kaneto, K. Yamanaka, K. Rikitake, T. Akiyama, and W. Tahashima, Jpn. J. Appl. Phys. 35, 1802 (1996). 490. G. Priebe, B. Pietzak, and R. Könenkamp, Appl. Phys. Lett. 71, 2160 (1997). 491. C. Wen, J. Li, K. Kitazawa, T. Aida, I. Honma, H. Komiyama, and K. Yamada, Appl. Phys. Lett. 61, 2162 (1992). 492. H. Habuchi, S. Nitta, D. Han, and S. Nonomura, J. Appl. Phys. 87, 8580 (2000). 493. A. Hamed, Y. Y. Sun, Y. K. Tao, R. L. Meng, and P. H. Hor, Phys. Rev. B 47, 10873 (1993). 494. A. Zahab and L. Firlej, Solid State Commun. 87, 893 (1993). 495. He Peimo, Xu Yabo, Z. Xuanjia, and Li Wenzhou, Solid State Commun. 89, 373 (1994). 496. J. Mort, R. Ziolo, M. Machonkin, D. R. Huffman, and M. I. Ferguson, Chem. Phys. Lett. 186, 284 (1991). 497. K. Hoshimono, S. Fujimori, and S. Fujita, Jpn. J. Appl. Phys. 2 32, L1070 (1993). 498. T. Takahashi, S. Suzuki, T. Morikawa, H. Katayama-Yoshida, S. Hasegawa, H. Inokuchi, K. Seki, K. Kikuchi, S. Suzuki, K. Ikemoto, and Y. Achiba, Phys. Rev. Lett. 68, 1232 (1992). 499. S. Fujimori, K. Hoshimono, and S. Fujita, Solid State Commun. 89, 437 (1994). 500. S. Gonda, M. Kawasaki, T. Arakane, and H. Koinuma, Mater. Res. Soc. Symp. Proc. 349, 25 (1994). 501. K. Rikitake, T. Akiyama, W. Takashima, and K. Kaneto, Synth. Met. 86, 2357 (1997). 502. J. Paloheimo and H. Isotalo, Synth. Met. 55, 3185 (1993). 503. J. C. Wang and Y. F. Chen, Appl. Phys. Lett. 73, 948 (1998). 504. Y. Saito, H. Shinohara, M. Kato, H. Nagashima, M. Ohkohchi, and Y. Ando, Chem. Phys. Lett. 189, 236 (1992). 505. S. Matsuura and T. Ishiguro, Fullerene Sci. Technol. 3, 437 (1995). 506. T. Arai, Y. Murakami, H. Suematsu, K. Kikuchi, Y. Achiba, and I. Ikemoto, Solid State Commun. 84, 827 (1992). 507. M. Kaiser, W. K. Maser, H. J. Byrne, A. Mittelbach, and S. Roth, Solid State Commun. 87, 281 (1993). 508. F. Yan, Y. N. Wang, Y. N. Huang, M. Gu, Q. M. Zhang, and H. M. Shen, Phys. Lett. A 201, 443 (1995). 509. I. G. Austin and N. F. Mott, Adv. Phys. 18, 41 (1969). 510. N. I. Nemchuk, T. L. Makarova, O. I. Konkov, Y. F. Biriulin, and A. Y. Vul, Mol. Cryst. Liq. Cryst. C 7, 183 (1996).

C60 -Based Materials 511. T. Rabenau, S. Roth, and R. K. Kremer, Acta Phys. Pol. A 87, 881 (1995). 512. R. K. Kremer, T. Rabenau, W. K. Maser, M. Kaiser, A. Simon, M. Haluska, and H. Kuzmany, Appl. Phys. A 56, 211 (1993). 513. M. K. Kelly, P. Etchegoin, D. Fuchs, W. Krätschmer, and K. Fostiropoulos, Phys. Rev. B 46, 4963 (1992). 514. T. L. Makarova, Mol. Cryst. Liq. Cryst. C 7, 199 (1996). 515. P. Milani, M. Manfredini, G. Guizzetti, F. Marabelli, and M. Patrini, Solid State Commun. 90, 639 (1994). 516. S. L. Ren, Y. Wang, A. M. Rao, E. McRae, G. T. Hager, K. A. Wang, W. T. Lee, H. F. Ni, J. Selegue, and P. C. Eklund, Appl. Phys. Lett. 59, 2678 (1991). 517. Y. Wang, J. M. Holden, A. M. Rao, W. T. Lee, X. X. Bi, S. L. Ren, G. W. Lehman, G. T. Hager, and P. C. Eklund, Phys. Rev. B 45, 14396 (1992). 518. S. Leach, M. Vervloet, A. Despres, A. Despres, E. Breheret, J. P. Hare, T. J. Dennis, H. W. Kroto, R. Taylor, and D. R. M. Walton, Chem. Phys. 160, 451 (1992). 519. T. Gotoh, S. Nonomura, S. Hirata, and S. Nitta, Appl. Surf. Sci. 114, 278 (1997). 520. M. S. Dresselhaus, G. Dresselhaus, A. M. Rao, and P. C. Eklund, Synth. Met. 78, 313 (1996). 521. T. N. Thomas, J. F. Ryan, R. A. Taylor, D. Mihaiovic, and R. Zamboni, Int. J. Mod. Phys. B 6, 3931 (1992). 522. U. D. Venkateswaran, D. Sanzi, A. M. Rao, P. C. Eklund, L. Marques, J.-L. Hodeau, and M. Núñez-Regueiro, Phys. Rev. B 57, R3193 (1998). 523. S. Mochizuki, M. Sasaki, and R. Ruppin, J. Phys.: Condens. Matter 10, 2347 (1998). 524. M. Braga, S. Larsson, A. Rosen, and A. Volosov, Astron. Astrophys. 245, 232 (1991). 525. G. F. Bertsch, A. Bulgac, D. Tománek, and Y. Wang, Phys. Rev. Lett. 67, 2690 (1991). 526. K. Harigaya and S. Abe, Phys. Rev. B 49, 16746 (1994). 527. T. Pichler, M. Matus, J. Kürti, and H. Kuzmany, Solid State Commun. 81, 859 (1992). 528. K. Yabana and G. F. Bertsch, Chem. Phys. Lett. 197, 32 (1992). 529. M. I. Salkola, Phys. Rev. B 49, 4407 (1994). 530. M. S. Dresselhaus, G. Dresselhaus, and P. C. Eklund, J. Mater. Res. 8, 2054 (1993). 531. J. Mort, K. Okumura, M. Machonkin, R. Ziolo, D. R. Huffman, and M. I. Ferguson, Chem. Phys. Lett. 186, 281 (1991). 532. M. Kaiser, J. Reichenbach, H. J. Byrne, J. Anders, W. Maser, S. Roth, A. Zahab, and P. Bernier, Solid State Commun. 81, 261 (1992). 533. A. Fartash, Phys. Rev. B 54, 17215 (1996). 534. A. M. Rao, P. Zhou, K.-A. Wang, G. T. Hager, J. M. Holden, Y. Wang, W.-T. Lee, X.-X. Bi, P. C. Eklund, D. S. Cornett, M. A. Duncan, and I. J. Amster, Science 259, 955 (1993). 535. B. Pevzner, A. F. Hebard, and M. S. Dresselhaus, Phys. Rev. B 55, 16439 (1997). 536. S. D. Senturia, J. N. F. Sheppard, H. L. Lee, and D. R. Day, J. Adhesion 15, 69 (1982). 537. N. F. Sheppard, D. R. Day, H. L. Lee, and S. D. Senturia, Sensors Actuators 2, 263 (1982). 538. F. Yan and Y. N. Wang, Appl. Phys. Lett. 72, 3446 (1998). 539. J. S. Su, Y. F. Chen, and K. C. Chiu, Appl. Phys. Lett. 75, 1607 (1999). 540. V. Elser and R. C. Haddon, Nature 325, 792 (1987). 541. V. Elser and R. C. Haddon, Phys. Rev. A 36, 4579 (1990). 542. A. Pasquarello, M. Schlüter, and R. C. Haddon, Science 257 (1992). 543. B. T. Kelly, “Physics of Graphite.” Applied Science, London, 1981. 544. R. S. Ruoff, D. Beach, J. Cuomo, T. Mcguire, R. L. Whetten, and F. Diederich, J. Phys. Chem. 95, 3457 (1991). 545. R. C. Haddon, L. F. Schneemeyer, J. V. Waszczak, S. H. Glarum, R. Tycko, G. Dabbagh, A. R. Kortan, A. J. Muller, A. M. Mujsce,

C60 -Based Materials

546. 547. 548. 549. 550. 551. 552.

553.

554. 555.

556. 557. 558. 559. 560. 561. 562. 563.

564. 565. 566. 567. 568. 569. 570. 571. 572. 573. 574.

575.

576.

M. J. Rosseinsky, S. M. Zahurak, A. V. Makhija, F. A. Thiel, K. Raghavachari, E. Cockayne, and V. Elser, Nature (London) 350, 46 (1991). P. W. Fowler, P. Lasseretti, and R. Zanasi, Chem. Phys. Lett. 165, 79 (1990). R. C. Haddon and V. Elser, Chem. Phys. Lett. 169, 362 (1990). N. Ganguli and K. S. Krishnan, Proc. Roy. Soc. London Ser. A 177, 168 (1941). X. K. Wang, R. P. H. Chang, A. Patashinski, and J. B. Ketterson, J. Mater. Res. 9, 1578 (1994). J. Heremans, C. H. Olk, and D. T. Morelli, Phys. Rev. B 49, 15122 (1994). Y. G. Dorfman, “Diamagnetism and the Chemical Bond.” Elsevier, New York, 1966. P. M. Allemand, K. C. Khemani, A. Koch, F. Wudl, K. Holczer, S. Donovan, G. Grüner, and J. D. Thompson, Science 253, 301 (1991). V. E. Antonov, I. O. Bashkin, S. S. Khasanov, A. P. Moravsky, Yu. G. Morozov, Yu. M. Shulga, Yu. A. Ossipyan, and E. G. Ponyatovsky, J. Alloys Compd. 330, 365 (2002). Y. M. Shul’ga, A. S. Lobach, I. N. Ivleva, Y. G. Morozov, and V. N. Spector, Mol. Cryst. Liq. Cryst. 10, 201 (1998). T. L. Makarova, B. Sundqvist, R. Höhne, P. Esquinazi, Y. Kopelevich, P. Scharff, V. A. Davydov, L. S. Kasherova, and A. V. Rakhmanina, Nature 413, 716 (2001). J. R. Olson, K. A. Topp, and R. O. Pohl, Science 259, 1145 (1993). S. P. Tewari, P. Silotia, and K. Bera, Solid State Commun. 107, 129 (1998). W. P. Beyermann, M. F. Hundley, and J. D. Thompson, Phys. Rev. Lett. 68, 2046 (1992). P. W. Anderson, B. I. Halperin, and C. M. Varma, Philos. Mag. 25, 1 (1972). K. Allen and F. Hellman, Phys. Rev. B 60, 11765 (1999). J. E. Fischer, A. R. McGhie, J. K. Estrada, M. Haluška, H. Kuzmany, and H.-U. ter Meer, Phys. Rev. B 53, 11418 (1996). W. DeSorbo and W. W. Tyler, J. Chem. Phys. 21, 1660 (1953). G. B. M. Vaughan, P. A. Heiney, D. E. Cox, J. E. Fischer, A. R. McGhie, A. L. Smith, R. M. Strongin, M. A. Cichy, and A. B. Smith III, Chem. Phys. 178, 599 (1993). A. T. Pugachev, N. P. Churakova, N. I. Gorbenko, Kh. Saadli, and E. S. Syrkin, J. Exp. Theor. Phys. 87, 1014 (1998). Ya. M. Soifer and N. P. Kobelev, Mol. Mater. 7, 267 (1996). F. Yan, M. Gu, and Y. N. Wang, J. Phys. IV 6, C8 (1996). N. P. Kobelev, R. K. Nikolaev, N. S. Sidorov, and Ya. M. Soˇıfer, Phys. Solid State 44, 429 (2002). N. P. Kobelev, R. K. Nikolaev, Ya. M. Soˇıfer, and S. S. Khasanov, Phys. Solid State 40, 154 (1998). N. P. Kobelev, Ya, M. Soˇıfer, R. K. Nikolaev, and V. M. Levin, Phys. Status Solidi B 214, 303 (1999). E. Burgos, E. Halac, and H. Bonadeo, Phys. Rev. B 49, 15544 (1994). N. P. Kobelev, R. K. Nikolaev, Ya. M. Soˇıfer, and S. S. Khasanov, Chem. Phys. Lett. 276, 263 (1997). N. P. Kobelev, A. P. Moravskˇı, Ya, M. Soˇıfer, I. O. Bashkin, and O. G. Rybchenko, Phys. Solid State 36, 1491 (1994). H. Coufal, K. Meyer, R. K. Grygier, M. de Vries, D. Jenrich, and P. Hess, Appl. Phys. A 59, 83 (1994). A. F. Hebard, M. J. Rosseinsky, R. C. Haddon, D. W. Murphy, S. H. Glarum, T. T. M. Palstra, A. P. Ramirez, and A. R. Kortan, Nature 350, 600 (1991). R. C. Haddon, A. F. Hebard, M. J. Rosseinsky, D. W. Murphy, S. J. Duclos, K. B. Lyons, B. Miller, J. M. Rosamilia, R. M. Fleming, A. R. Kortan, S. H. Glarum, A. V. Makhija, A. J. Muller, R. H. Eick, S. M. Zahurak, R. Tycko, G. Dabbagh, and F. A. Thiel, Nature 350, 320 (1991). L. Holczer, O. Klein, S. M. Huang, R. B. Kaner, K. J. Fu, R. L. Whetten, and F. Diederich, Science 252, 1154 (1991).

467 577. J. P. McCauley, Jr., Q. Zhu, N. Coustel, O. Zhou, G. Vaughan, S. H. J. Idziak, J. E. Fischer, S. W. Tozer, D. M. Froski, N. Bykovetz, C. L. Lin, A. R. McGhie, B. H. Allen, W. J. Romanow, A. M. Denenstein, and A. B. Smith III, J. Am. Chem. Soc. 113, 8537 (1991). 578. X. D. Xiang, J. G. Hou, G. Briceno, W. A. Vareka, R. Mostovoy, A. Zettl, V. H. Crespi, and M. L. Cohen, Science 256, 1190 (1992). 579. X. D. Xiang, J. G. Hou, V. H. Crespi, A. Zettl, and M. L. Cohen, Nature 361, 54 (1993). 580. F. Benseda, B. Xiang, and L. Kevan, J. Phys. Chem. 96, 6118 (1992). 581. M. Tokumoto, Y. Tanaka, N. Kinoshita, and T. Kinoshita, J. Phys. Chem. Solids 54, 1667 (1993). 582. T. Yildirim, L. Barbedette, J. E. Fischer, G. M. Bendele, P. W. Stephens, C. Lin, C. Goze, F. Rachdi, J. Robert, P. Petit, and T. T. M. Palstra, Phys. Rev. B 54, 981 (1996). 583. T. Inabe, H. Ogata, Y. Maruyama, Y. Achiba, S. Suzuki, K. Kikuchi, and I. Ikemoto, Phys. Rev. Lett. 69, 3797 (1992). 584. Q. Zhu, O. Zhou, N. Coustel, G. B. M. Vaughan, J. P. McCauley, Jr., W. J. Romanow, J. E. Fischer, and A. B. Smith III, Science 254, 545 (1991). 585. P. W. Stephens, L. Mihaly, J. B. Wiley, S. M. Huang, R. B. Kaner, F. Diederich, R. L. Whetten, and K. Holczer, Phys. Rev. B 45, 543 (1992). 586. P. Zhou, K. A. Wang, Y. Wang, and P. C. Eklund, Phys. Rev. B 46, 2595 (1992). 587. M. K. Kelly and C. Thomsen, Phys. Rev. B 50, 18572 (1994). 588. J. Winter and H. Kuzmany, Phys. Rev. B 54, 655 (1996). 589. X. H. Chen, T. Takenobu, T. Muro, H. Fudo, and Y. Iwasa, Phys. Rev. B 60, 12462 (1999). 590. I. D. Hands, J. L. Dunn, and C. A. Bates, Phys. Rev. B 63, 245414 (2001). 591. K. J. Fu, W. L. Karney, O. L. Chapman, S. M. Huang, R. B. Kaner, F. Diederich, K. Holczer, and R. L. Whetten, Phys. Rev. B 46, 1937 (1992). 592. A. F. Hebard, Phys. Today 45, 26 (1992). 593. C. H. Pennington and V. A. Stenger, Rev. Mod. Phys. 68, 855 (1996). 594. S. Satpathy, V. P. Antropov, O. K. Andersen, O. Jepsen, O. Gunnarsson, and A. I. Lichtenstein, Phys. Rev. B 46, 1773 (1992). 595. M. P. Gelfand and J. P. Lu, Phys. Rev. Lett. 68, 1050 (1992). 596. E. J. Mele and S. C. Erwin, Phys. Rev. B 50, 2150 (1994). 597. S. Satpathy, V. P. Antropov, O. K. Anderson, O. Jepsen, O. Gunnarsson, and A. I. Lichtenstein, Phys. Rev. B 46, 1773 (1992). 598. S. C. Erwin, “Buckminsterfullerenes” (W. E. Billups, and M. A. Ciufoini, Eds.), pp. 217–255. VCH, New York. 599. S. E. Erwin and W. E. Pickett, Science 254, 842 (1992). 600. M. Z. Huang, Y. N. Xu, and W. Y. Ching, Phys. Rev. B 46, 6572 (1992). 601. D. L. Novikov, V. A. Gubanov, and A. J. Freeman, Physica C 191, 399 (1992). 602. J. L. Martins and N. Troullier, Phys. Rev. B 46, 1776 (1992). 603. A. Oshiyama, S. Saito, N. Hamada, and Y. Miyamoto, J. Phys. Chem. Solids 53, 1457 (1992). 604. W. Andreoni, P. Giannozzi, and M. Parrinello, Phys. Rev. B 51, 2087 (1995). 605. K. P. Bohnen, R. Heid, K. M. Ho, and C. T. Chan, Phys. Rev. B 51, 5805 (1995). 606. O. Gunnarsson, S. C. Erwin, E. Koch, and R. M. Martin, Phys. Rev. B 57, 2159 (1998). 607. P. J. Benning, D. M. Poirier, T. R. Ohno, Y. Chen, M. B. Jost, F. Stepniak, G. H. Kroll, J. H. Weaver, J. Fure, and R. E. Smalley, Phys. Rev. B 45, 6899 (1992). 608. M. Merkel, M. Knupfer, M. S. Golden, and J. Fink, Phys. Rev. B 47, 11740 (1993). 609. M. Knupfer, M. Merkel, M. S. Golden, and J. Fink, Phys. Rev. B 47, 13944 (1993).

468 610. P. J. Benning, F. Stepniak, D. M. Poirier, J. L. Martins, J. H. Weaver, L. P. F. Chibante, and R. E. Smalley, Phys. Rev. B 47, 13843 (1993). 611. M. De Seta and F. Evangelisti, Phys. Rev. Lett. 71, 2477 (1993). 612. O. Gunnarsson, H. Handschuh, P. S. Bechthold, B. Kessler, G. Gantefor, and W. Eberhardt, Phys. Rev. Lett. 74, 1875 (1995). 613. M. S. Golder, M. Knupfer, J. Fink, J. F. Armbruster, T. R. Cummins, H. A. Romberg, M. Roth, M. Sing, M. Schmidt, and E. Sohmen, J. Phys.: Condens. Matter 7, 8219 (1995). 614. O. Gunnarsson, A. I. Liechtenstein, V. Eyert, M. Knupfer, J. Fink, and J. F. Armbruster, Phys. Rev. B 53, 3455 (1996). 615. O. Gunnarsson, V. Eyert, M. Knupfer, J. Fink, and J. F. Armbruster, J. Phys.: Condens. Matter 8, 2557 (1996). 616. A. I. Liechtenstein, O. Gunnarsson, M. Knupfer, J. Fink, and J. F. Armbruster, J. Phys.: Condens. Matter 8, 4001 (1996). 617. H. Li, Y. Xu, S. Bao, H. Li, P. He, J. Zhang, J. Wang, H. Qian, F. Liu, and Y. Kuirisi, Phys. Rev. B 61, 13256 (2000). 618. R. Hesper, L. H. Tjeng, A. Heeres, and G. A. Sawatzky, Phys. Rev. Lett. 85, 1970 (2000). 619. J. H. Weaver and D. M. Poirier, in “Solid State Physics 48” (H. Ehrenreich and F. Saepen, Eds.), p. 225. Academic, New York, 1994. 620. F. Stepniak, P. J. Benning, D. M. Poirier, and J. H. Weaver, Phys. Rev. B 48, 1899 (1993). 621. R. C. Haddon, A. S. Perel, R. C. Morris, S. H. Chang, A. T. Fiory, A. F. Hebard, T. T. M. Palstra, and G. P. Kochanski, Chem. Phys. Lett. 218, 100 (1994). 622. M. Knupfer, J. F. Armbruster, H. A. Romberg, and J. Fink, Synth. Met. 70, 1321 (1995). 623. E. Sohmen, J. Fink, and W. Kratschmer, Europhys. Lett. 17, 51 (1992). 624. T. Yildirim, L. Barbette, J. E. Fischer, C. L. Lin, J. Robert, P. Petit, and T. T. M. Palstra, Phys. Rev. Lett. 77, 167 (1996). 625. V. P. Antropov, O. Gunnarsson, and A. I. Liechtenstei, Phys. Rev. B 46, 7651 (1993). 626. T. T. M. Palstra, R. C. Haddon, A. F. Hebard, and J. Zaanen, Phys. Rev. Lett. 68, 1054 (1991). 627. O. Klein, G. Gruner, S. M. Huang, J. B. Wiley, and R. B. Kaner, Phys. Rev. B 46, 11247 (1992). 628. A. F. Hebard, T. T. M. Palstra, R. C. Haddon, and R. M. Fleming, Phys. Rev. B 48, 9945 (1993). 629. A. Zettl, L. Lu, X. D. Xiang, J. G. Hou, W. A. Vareka, and M. S. Fuhrer, J. Supercond. 7, 639 (1994). 630. W. A. Vareka and A. Zettl, Phys. Rev. Lett. 72, 4121 (1994). 631. T. T. M. Palstra, A. F. Hebard, R. C. Haddon, and P. B. Littlewood, Phys. Rev. B 50, 3462 (1994). 632. J. G. Hou, L. Lu, V. H. Crespi, X. D. Xiang, A. Zettl, and M. L. Cohen, Solid State Commun. 93, 973 (1995). 633. L. Degiorgi, G. Gruïner, P. Wachter, S. M. Huang, J. Wiley, R. L. Whetten, R. B. Kaner, K. Holczer, and F. Diederich, Phys. Rev. B 46, 11250 (1992). 634. L. Degiorgi, G. Briceno, M. S. Fuhrer, A. Zettl, and P. Wachter, Nature 359, 541 (1994). 635. L. Degiorgi, E. J. Nicol, O. Klein, G. Gruner, P. Wachter, S. M. Huang, J. B. Wiley, and R. B. Kaner, Phys. Rev. B 49, 7012 (1994). 636. O. Gunnarsson, Rev. Mod. Phys. 69, 575 (1997). 637. G. Sparn, J. D. Thompson, S. M. Huang, R. B. Kaner, F. Diederich, R. L. Wetten, G. Grüner, and K. Holczer, Science 252, 1829 (1991). 638. K. Tanigaki, T. W. Ebbesen, S. Saito, J. Mizuki, J. S. Tsai, Y. Kubo, and S. Kuroshima, Nature 352, 222 (1991). 639. O. Zhou, G. B. M. Vaughan, Q. Zhu, J. E. Fischer, P. A. Heiney, N. Coustel, J. P. McCauley, Jr., and A. B. Smith III, Science 255, 833 (1992). 640. R. J. Rosseinsky, A. P. Ramirez, S. H. Glarum, D. W. Murphy, R. C. Haddon, A. F. Hebard, T. T. M. Palstra, A. R. Kortan, S. M. Zahurak, and A. V. Makhija, Phys. Rev. Lett. 66, 2830 (1991).

C60 -Based Materials 641. J. E. Schirber, D. L. Overmeyer, H. H. Wang, J. M. Williams, K. D. Carlson, A. M. Kini, U. Welp, and W. K. Kwok, Physica C 178, 137 (1991). 642. G. Sparn, J. D. Thompson, R. L. Whetten, S. M. Huang, R. B. Kaner, F. Diederich, G. Gruner, and K. Holczer, Phys. Rev. Lett. 68, 1228 (1992). 643. T. W. Ebbesen, J. S. Tsai, K. Tanigaki, J. Tabuchi, Y. Shimakawa, Y. Kubo, I. Hirosawa, and J. Mizuki, Nature 355, 620 (1992). 644. A. P. Ramirez, A. R. Kortan, M. J. Rosseinsky, S. J. Duclos, A. M. Mujsce, R. C. Haddon, D. W. Murphy, A. V. Makhija, S. M. Zahurak, and K. B. Lyons, Phys. Rev. Lett. 68, 1058 (1992). 645. C. C. Chen and C. M. Lieber, J. Am. Chem. Soc. 114, 3141 (1992). 646. C. C. Chen and C. M. Lieber, Science 259, 655 (1993). 647. Z. Zhang, C. C. Chen, S. P. Kelty, H. Dai, and C. M. Lieber, Nature 353, 333 (1991). 648. V. A. Stenger, C. H. Pennington, D. R. Buffinger, and R. P. Ziebarth, Phys. Rev. Lett. 74, 1649 (1995). 649. R. F. Kiefl, W. A. MacFarlane, K. H. Chow, S. Dunsiger, T. L. Duty, T. M. S. Johnston, J. W. Schneider, J. Sonier, L. Brard, R. M. Strongin, J. E. Fischer, and A. B. Smith III, Phys. Rev. Lett. 70, 3987 (1993). 650. D. Koller, M. C. Martin, and L. Mihaly, Phys. Rev. Lett. 77, 4082 (1996). 651. C. Gu, B. W. Veal, R. Liu, A. P. Paulikas, P. Kostic, H. Ding, K. Gofron, J. C. Campuzano, J. A. Schlueter, H. H. Wang, U. Geiser, and J. M. Williams, Phys. Rev. B 50, 16566 (1994). 652. Z. Zhang, C. C. Chen, and C. M. Lieber, Science 254, 1619 (1991). 653. P. Jess, U. Hubler, S. Behler, V. Thommen-Geiser, H. P. Lang, and H. J. Güntherodt, Synth. Met. 77, 201 (1996). 654. R. Tycko, G. Dabbagh, M. J. Rosseinsky, D. W. Murphy, A. P. Ramirez, and R. M. Fleming, Phys. Rev. Lett. 68, 1912 (1992). 655. J. G. Hou, V. H. Crespi, X. D. Xiang, W. A. Vareka, G. Briceno, A. Zettl, and M. L. Cohen, Solid State Commun. 86, 643 (1993). 656. J. G. Hou, X. D. Xiang, V. H. Crespi, M. L. Cohen, and A. Zettl, Physica C 228, 175 (1994). 657. V. Buntar and H. W. Weber, Supercond. Sci. Technol. 9, 599 (1996). 658. R. Tycko, G. Dabbagh, M. J. Rosseinsky, D. W. Murphy, R. M. Fleming, A. P. Ramirez, J. C. Tully, Science 253, 884 (1991). 659. K. Holczer, O. Klein, G. Grüner, J. D. Thompson, F. Diederich, and R. L. Whetten, Phys. Rev. Lett. 67, 271 (1991). 660. C. E. Johnson, H. W. Jiang, K. Holtczer, R. B. Kaner, R. L. Whetten, and F. Diederich, Phys. Rev. B 46, 5880 (1992). 661. G. S. Boebinger, T. T. M. Palstra, A. Passner, M. J. Rosseinsky, and D. W. Murphy, Phys. Rev. B 46, 5676 (1992). 662. S. Foner, E. J. McNiff, D. Heiman, S. M. Huang, and R. B. Kaner, Phys. Rev. B 46, 14936 (1992). 663. J. G. Hou, X. D. Xiang, M. L. Cohen, and A. Zettl, Physica C 232, 22 (1994). 664. N. R. Werthamer, E. Helfand, and P. C. Hohenberg, Phys. Rev. 147, 295 (1966). 665. V. Buntar, F. M. Sauerzopf, and H. W. Weber, Phys. Rev. B 56, 14128 (1997). 666. S. Chu and M. E. McHenry, Phys. Rev. B 55, 11722 (1997). 667. C. Politis, A. I. Sokolov, and V. Buntar, Mod. Phys. Lett. B 6, 351 (1992). 668. V. Buntar, F. M. Sauerzopf, and H. W. Weber, Phys. Rev. B 54, R9651 (1996). 669. V. Buntar, U. Eckern, and C. Politis, Mod. Phys. Lett. B 6, 1037 (1992). 670. S. H. Irons, J. Z. Liu, P. Klavins, and R. N. Shelton, Phys. Rev. B 52, 15517 (1995). 671. M. Dressel, L. Degiorgi, O. Klein, and G. Grüner, J. Phys. Chem. Solids 54, 1411 (1993). 672. Y. J. Uemura, A. Keren, L. P. Le, G. M. Luke, B. J. Sternlieb, W. D. Wu, J. H. Brewer, R. L. Whetten, S. M. Huang, S. Lin, R. B. Kaner, F. Diederich, S. Donovan, G. Grüner, and K. Holczer, Nature 352, 605 (1991).

C60 -Based Materials 673. Y. J. Uemura, L. P. Le, and G. M. Luke, Synth. Met. 56, 2845 (1993). 674. L. Degiorgi, P. Wachter, G. Grüner, S. M. Huang, J. Wiley, and R. B. Kaner, Phys. Rev. Lett. 69, 2987 (1992). 675. C. P. Bean, Phys. Rev. Lett. 8, 250 (1962). 676. W. A. Fietz and W. W. Webb, Phys. Rev. 178, 657 (1969). 677. C. Politis, V. Buntar, W. Krauss, and A. Gurevich, Europhys. Lett. 17, 175 (1992). 678. M. Baenitz, M. Heinze, E. Straube, H. Werner, R. Schlögl, V. Thommen, H. J. Güntherodt, and K. Lüders, Physica C 228, 181 (1994). 679. J. D. Thompson, G. Sparn, K. Holczer, O. Klein, G. Grüner, R. B. Kaner, F. Diederich, and R. L. Whetten, in “Physical and Material Properties of High Temperature Superconductors” (S. K. Malic, and S. S. Shah, Eds.), p. 139. Nova, Commack, NJ, 1994. 680. C. Politis, V. Buntar, and V. P. Seminozhenko, Int. J. Mod. Phys. B 7, 2163 (1993). 681. M. W. Lee, M. F. Tai, S. C. Luo, and J. B. Shi, Physica C 245, 6 (1995). 682. M. F. Tai, G. F. Chang, and M. W. Lee, Phys. Rev. B 52, 1176 (1995). 683. T. T. M. Palstra, O. Zhou, Y. Iwasa, P. E. Sulewski, R. M. Fleming, and B. R. Zegarski, Solid State Commun. 93, 327 (1995). 684. M. J. Rosseinsky, D. W. Murphy, R. M. Fleming, R. Tycko, A. P. Ramirez, T. Siegrist, G. Dabbagh, and S. E. Barrett, Nature 356, 416 (1992). 685. K. Tanigaki, I. Hirosawa, T. W. Ebbesen, J. Mizuki, Y. Shimakawa, Y. Kubo, J. S. Tsai, and S. Kuroshima, Nature 356, 419 (1992). 686. K. Tanigaki and K. Prassides, J. Mater. Chem. 5, 1515 (1995). 687. I. Hirosawa, K. Tanigaki, J. Mizuki, T. W. Ebbesen, Y. Shimakawa, Y. Kubo, and S. Kuroshima, Solid State Commun. 82, 979 (1992). 688. J. Winter and H. Kuzmany, Solid State Commun. 84, 935 (1992). 689. P. W. Stephens, Nature 356, 383 (1992). 690. Q. Zhu, O. zhou, J. E. Fischer, A. R. McGhie, W. J. Romanow, R. M. Strongin, M. A. Cichy, and A. B. Smith, III, Phys. Rev. B 47, 13948 (1993). 691. D. M. Poirier, T. R. Ohno, G. H. Kroll, P. J. Benning, F. Stepniak, J. H. Weaver, L. P. F. Chibante, and R. E. Smalley, Phys. Rev. B 47, 9870 (1993). 692. D. M. Poirier and J. H. Weaver, Phys. Rev. B 47, 10959 (1993). 693. P. W. Stephens, G. Bortel, G. Faigel, M. Tegze, A. Janossy, S. Pekker, G. Oszlanyi, and L. Forro, Nature 370, 636 (1994). 694. O. Chauvet, G. Oszlanyi, L. Forro, P. W. Stephens, M. Tegze, G. Faigel, and A. Janossy, Phys. Rev. Lett. 72, 2721 (1994). 695. S. Pekker, L. Forro, L. Mihaly, and A. Janossy, Solid State Commun. 90, 349 (1996). 696. S. Pekker, A. Janossy, L. Mihaly, O. Chauvet, M. Carrard, and L. Forro, Science 265, 1077 (1994). 697. G. Faigel, G. Bortel, M. Tegze, L. Granasy, S. Pekker, G. Oszlanyi, O. Chauvet, G. Baumgartner, L. Forro, P. W. Stephens, L. Mihaly, and A. Janossy, Phys. Rev. B 52, 3199 (1995). 698. T. Kalber, G. Zimmer, and M. Mehring, Z. Phys. B 97, 1 (1995). 699. H. Alloul, V. Brouet, E. Lafontaine, L. Malier, and L. Forro, Phys. Rev. Lett. 76, 2922 (1996). 700. B. Renker, H. Schober, F. Gompf, R. Heid, and E. Ressouche, Phys. Rev. B 53, R14701 (1996). 701. L. Cristofolini, C. M. Brown, A. J. Dianoux, M. Kosaka, K. Prassides, K. Tanigaki, and K. Vavekis, Chem. Commun. 2465 (1996). 702. G. B. Adams, J. B. Page, O. F. Sankey, and M. O’Keeffe, Phys. Rev. B 50, 17471 (1994). 703. S. C. Erwin, G. V. Krishna, and E. J. Mele, Phys. Rev. B 51, 7345 (1995). 704. T. Ogitsu, T. M. Briere, K. Kusakabe, S. Tsuneyuki, and Y. Kateyama, Phys. Rev. B 58, 13925 (1998). 705. Q. Zhu, D. E. Cox, and J. E. Fischer, Phys. Rev. B 51, 3966 (1995). 706. G. Oszlanyi, G. Bortel, G. Faigel, L. Granasy, G. Bendele, P. W. Stephens, and L. Forro, Phys. Rev. B 54, 11849 (1996).

469 707. M. Kosaka, K. Tanigaki, T. Tanaka, T. Take, A. Lappas, and K. Prassides, Phys. Rev. B 51, 12018 (1995). 708. R. M. Fleming, M. J. Rossiensky, A. P. Ramirez, D. W. Murphy, J. C. Tully, R. C. Haddon, T. Siegrist, R. Tycko, S. H. Glarum, P. Marsh, G. Dabbagh, S. M. Zahurak, A. V. Makhija, and C. Hampton, Nature 352, 701 (1991); Erratum; Nature 353, 868 (1991). 709. O. Zhou, J. E. Fisher, N. Coustel, S. Kycia, Q. Zhu, A. R. McGhie, W. J. Romanow, J. P. McCauley, Jr., A. B. Smith III, and D. E. Cox, Nature 351, 462 (1991). 710. P. Dahlke, P. F. Henry, and M. J. Rosseinskyi, J. Mater. Chem. 8, 1571 (1998). 711. A. K. Christine, G. M. Bendele, and P. W. Stephens, Phys. Rev. B 55, R3366 (1997). 712. R. F. Kiefl, T. L. Duty, J. W. Schneider, A. MacFarlane, K. Chow, J. W. Elzey, P. Mendels, G. D. Morris, J. H. Brewer, E. J. Ansaldo, C. Niedermayer, D. R. Noakes, C. E. Stronach, B. Hitti, and J. E. Fischer, Phys. Rev. Lett. 69, 2005 (1992). 713. S. C. Erwin and C. Bruder, Physica B 199–200, 600 (1994). 714. H. Suematsu, Y. Murakami, T. Arai, K. Kikuchi, Y. Achiba, and I. Ikemoto, Mater. Sci. Eng. B 19, 141 (1993). 715. C. Goze, F. Rachdi, and M. Mehring, Phys. Rev. B 54, 5164 (1996). 716. M. Knupfer and J. Fink, Phys. Rev. Lett. 79, 2714 (1997). 717. G. A. Sawatzky, “Physics and Chemistry of Fullerenes and Derivatives ” (H. Kuzmany, J. Fink, M. Mehring, and S. Roth, Eds.), 1995. World Scientific, Singapore, p. 373. 718. J. P. Lu, Phys. Rev. B 49, 5687 (1994). 719. Y. Iwasa, T. Kaneyasu, M. Nagata, and N. Mizutani, Synth. Met. 70, 1361 (1995). 720. R. Kerkoud, P. Auban-Senzier, D. Jerome, S. Brazovskii, N. Kirova, I. Lukyanchuk, F. Rachdi, and C. Goze, Synth. Met. 77, 205 (1996). 721. R. Kerkoud, P. Auban-Senzier, D. Jerome, S. Brazovskii, I. Kukyanchuk, N. Kirova, F. Rachdi, and C. Goze, J. Phys. Chem. Solids 57, 143 (1996). 722. D. W. Murphy, M. J. Rosseinsky, R. M. Fleming, R. Tycko, A. P. Ramirez, R. C. Haddon, T. Siegrist, G. Dabbag, J. C. Tully, and R. E. Walstedt, J. Phys. Chem. Solids 53, 1321 (1992). 723. L. Hajji, F. Rachdi, C. Goze, M. Mehring, and J. E. Fischer, Solid State Commun. 100, 493 (1996). 724. O. Zhou and D. E. Cox, J. Phys. Chem. Solids 53, 1373 (1992). 725. W. Andreoni, F. Gygi, and M. Parrinello, Phys. Rev. Lett. 68, 823 (1992). 726. T. Yildirim, J. E. Fischer, A. B. Harris, P. W. Stephens, D. Liu, L. Brard, R. M. Strongin, and A. B. Smith III, Phys. Rev. Lett. 71, 1383 (1993). 727. K. Tanigaki, I. Hirosawa, T. W. Ebbesen, J. I. Mizuki, and J. S. Tsai, J. Phys. Chem. Solids 54, 1645 (1993). 728. T. Yildirim, O. Zhou, J. E. Fischer, R. A. Strongin, M. A. Cichy, A. B. Smith III, C. L. Lin, and R. Jelinek, Nature 369, 568 (1992). 729. G. Oszlányi, G. Baumgartner, G. Faigel, and L. Forró, Phys. Rev. Lett. 78, 4438 (1997). 730. G. Oszlányi, G. Baumgartner, G. Faigel, L. Granasy, and L. Forró, Phys. Rev. B 58, 5 (1998). 731. K. Tanigaki, I. Hirosawa, T. W. Ebbesen, J. Mizuki, and S. Kuroshima, Chem. Phys. Lett. 203, 33 (1993). 732. K. Prassides, K. Vavekis, K. Kordatos, K. Tanigaki, G. M. Bendele, and P. W. Stephens, J. Am. Chem. Soc. 119, 834 (1997). 733. G. M. Bendele, P. W. Stephens, K. Prassides, K. Vavekis, K. Kordatos, and K. Tanigaki, Phys. Rev. Lett. 80, 736 (1998). 734. J. M. Fox, P. F. Henry, and J. M. Rosseinsky, Chem. Commun. 2299 (1996). 735. T. Yildirim, J. E. Fischer, R. Dinnebier, P. W. Stephens, and C. L. Lin, Solid State Commun. 93, 269 (1995). 736. Y. Chabre, D. Djurado, M. Armand, W. J. Romanow, N. Coustel, J. P. McCauley, Jr., J. E. Fischer, and A. B. Smith III, J. Am. Chem. Soc. 114, 764 (1992).

470 737. E. Ozdas, A. R. Kortan, N. Kopylov, A. P. Ramirez, T. Siegrist, and P. H. Citrin, in “Fullerenes: Recent Advances in the Physics and Chemistry of Fullerenes and Related Materials” (K. M. Kadish and R. S. Ruoff, Eds.), The Electrochemical Society Proceedings, Vol. 3. pp. 1176–1185. Electrochem Soc., Pennington, NJ, 1996. 738. L. Cristofolini, M. Riccò, and R. D. Renzi, Phys. Rev. B 59, 8343 (1999). 739. K. Tanigaki, T. W. Ebbesen, J.-S. Tsai, I. Hirosawa, and J. Mizuki, Europhys. Lett. 23, 57 (1993). 740. I. Hirosawa, K. Prassides, J. Mizuki, K. Tanigaki, M. Gevaert, A. Lappas, and J. K. Cockcroft, Science 264, 1294 (1994). 741. M. Kosaka, K. Tanigaki, K. Prasside, S. Margadonna, A. Lappas, C. M. Brown, and A. N. Fitch, Phys. Rev. B 59, R6628 (1999). 742. M. J. Rosseinsky, D. W. Murphy, R. M. Fleming, and O. Zhou, Nature 364, 425 (1993). 743. O. Zhou, T. T. M. Palstra, Y. Iwasa, R. M. Fleming, A. F. Hebard, P. E. Sulewski, D. W. Murephy, and B. R. Zegarski, Phys. Rev. B 52, 483 (1995). 744. H. Shimoda, Y. Iwasa, Y. Miyamoto, Y. Maniwa, and T. Mitani, Phys. Rev. B 54, R15653 (1996). 745. O. Zhou, R. M. Fleming, D. M. Murphy, M. J. Rosseinsky, A. P. Ramirez, R. B. van Dover, and R. C. Haddon, Nature 362, 433 (1993). 746. A. R. Kortan, N. Kopylov, S. Glarum, E. M. Gyorgy, A. P. Ramirez, R. M. Fleming, O. Zhou, F. A. Thiel, P. L. Trevor, and R. C. Haddon, Nature 360, 566 (1992). 747. M. Baenitz, M. Heinze, K. Luders, H. Werner, R. Schlogl, M. Weiden, G. Sparn, and F. Steglich, Solid State Commun. 96, 539 (1995). 748. Th. Schedel-Niedrig, M. C. Böhm, H. Werner, J. Schulte, and R. Schlögl, Phys. Rev. B 55, 13542 (1997). 749. B. Gogia, K. Kordatos, H. Suematsu, K. Tanigaki, and K. Prassides, Phys. Rev. B 58, 1077 (1998). 750. Th. Schedel-Niedrig, M. C. Böhm, H. Werner, J. Schulte, and R. Schlögl, Solid State Commun. 98, 463 (1996). 751. D. M. Deaven, P. E. Lammert, and D. S. Rokhsar, Phys. Rev. B 52, 16377 (1995). 752. S. Saito and A. Oshiyama, Phys. Rev. Lett. 71, 121 (1993). 753. S. Saito and A. Oshiyama, Solid State Commun. 83, 107 (1992). 754. S. Saito and A. Oshiyama, J. Phys. Chem. Solids 54, 1759 (1993). 755. Y. Chen, D. M. Porier, M. B. Jost, C. Gu, T. R. Ohno, J. L. Martins, J. H. Weaver, L. P. F. Chibante, and R. E. Smalley, Phys. Rev. B 46, 7961 (1992). 756. X. H. chen, X. J. Zhou, and S. Roth, Phys. Rev. B 54, 3971 (1996). 757. Y. Z. Li, J. C. Patrin, M. Chander, J. H. Weaver, L. P. F. Chibante, and R. E. Smalley, Phys. Rev. B 46, 12914 (1992). 758. J. E. Schirber, W. R. Bayless, A. R. Kortan, and N. Kopylov, Physica C 213, 190 (1993). 759. R. C. Haddon, G. P. Kochanski, A. F. Hebard, A. T. Fiory, and R. C. Morris, Science 258, 1636 (1992). 760. R. C. Haddon, G. P. Kochanski, A. F. Hebard, A. T. Fiory, R. C. Morris, and A. S. Perel, Chem. Phys. Lett. 203, 433 (1993). 761. H. Romberg, M. Roth, and J. Fink, Phys. Rev. B 49, 1427 (1994). 762. A. R. Kortan, N. Kopylov, E. Ozdas, A. P. Ramirez, R. M. Fleming, and R. C. Haddon, Chem. Phys. Lett. 223, 501 (1994). 763. A. R. Kortan, N. Kopylov, R. M. Fleming, O. Zhou, F. A. Thiel, and R. C. Haddon, Phys. Rev. B 47, 13070 (1993). 764. G. Sparn, A. Link, F. Steglich, M. Baenitz, K. Luders, H. Werner, and R. Schlogl, J. Low. Temp. Phys. 105, 1703 (1996). 765. Y. Iwasa, M. Kawaguchi, H. Iwasaki, T. Mitani, N. Wada, and T. Hasegawa, Phys. Rev. B 57, 13395 (1998). 766. Y. Iwasa, H. Hayashi, T. Furudate, and T. Mitani, Phys. Rev. B 54, 14960 (1996). 767. Y. Iwasa, H. Hayashi, T. Furudate, M. Kawaguchi, and T. Mitani, Synth. Met. 86, 2309 (1997). 768. T. Yildirim, L. Barbedette, J. E. Fischer, G. M. Bendele, P. W. Stephens, C. L. Lin, C. Goze, F. Rachdi, J. Robert, P. Petit, and T. T. M. Palstra, Phys. Rev. B 54, 11981 (1996).

C60 -Based Materials 769. Y. Errammach, A. Rezzouk, and F. Rachdi, Synth. Met. 129, 147 (2002). 770. H. Tou, Y. Manjwa, S. Taga, Y. Iwasa, T. Mitani, K. Tanigaki, B. Gogia, and H. Suematsu, Synth. Met. 102, 1463 (1999). 771. E. Özdas, A. R. Kortan, N. Kopylov, A. P. Ramirez, T. Siegrist, K. M. Rabe, H. E. Bair, S. Schuppler, and P. H. Citrin, Nature 375, 126 (1995). 772. P. H. Citrin, E. Ozdas, S. Schuppler, A. R. Kortan, and K. B. Lyons, Phys. Rev. B 56, 5213 (1997). 773. K. M. Rabe and P. H. Citrin, Phys. Rev. B 58, R551 (1998). 774. X. H. Chen and G. Roth, Phys. Rev. B 52, 15534 (1995). 775. X. H. Chen, S. Y. Li, G. G. Qian, K. Q. Ruan, and L. Z. Cao, Phys. Rev. B 57, 10770 (1998). 776. D. Claves, Y. Ksari, G. Chouteau, A. Collomb, and Ph. Touzain, Solid State Commun. 99, 359 (1996). 777. Y. Ksari-Habiles, D. Claves, G. Chouteau, P. Touzain, C. Jeandey, J. L. Oddou, and A. Stepanov, J. Phys. Chem. Solids 58, 1771 (1997). 778. D. Claves, Y. Ksari-Habiles, G. Chouteau, and P. Touzain, Solid State Commun. 106, 431 (1998). 779. A. S. Ginwalla, A. L. Balch, S. M. Kauzlarich, S. H. Irons, P. Klavins, and R. N. Shelton, Chem. Mater. 9, 278 (1997). 780. Q. Zhu, D. E. Cox, J. E. Fischer, K. Kniaz, A. R. McGhie, and O. Zhou, Nature 366, 712 (1992). 781. M. Kobayashi, Y. Akahama, H. Kawamura, H. Shinohara, H. Sato, and Y. Saito, Solid State Commun. 81, 93 (1992). 782. Y. Akahama, M. Kobayashi, H. kawamura, H. Shinohara, H. Sato, and Y. Saito, Mater. Sci. Eng. 19, 100 (1993). 783. H. Werner, M. Wesemann, and R. Schlögl, Europhys. Lett. 20, 107 (1992). 784. N. G. Park, S. W. Cho, S. J. Kim, and J. H. Choy, Chem. Mater. 8, 324 (1996). 785. Y. Maniwa, T. Shibata, K. Mizoguchi, K. Kume, K. Kikuchi, I. Ikemoto, S. Suzuki, and Y. Achiba, J. Phys. Soc. Jpn. 61, 2212 (1992). 786. R. E. Haufler, J. J. Conceicao, L. P. F. Chibante, Y. Chai, N. E. Byrne, S. Flanagan, M. M. Haley, S. C. O’Brien, C. Pan, Z. Xiao, W. E. Billups, M. A. Ciufolini, R. H. Hauge, J. L. Margrave, L. J. Wilson, R. F. Curl, and R. E. Smalley, J. Phys. Chem. 94, 8634 (1990). 787. H. Selig, C. Lifshitz, T. Peres, J. E. Fischer, A. R. McGhie, W. J. Romanow, J. P. McCauley, Jr., and A. B. Smith III, J. Am. Chem. Soc. 113, 5475 (1997). 788. J. H. Holloway, E. G. Hope, R. Taylor, G. J. Langley, A. G. Avent, T. J. Dennis, J. P. Hare, H. W. Kroto, and D. R. M. Walton, Chem. Commun. 966 (1991). 789. F. Cataldo, Carbon 32, 437 (1994). 790. P. R. Birkett, P. B. Hitchcock, H. W. Kroto, R. Taylor, and D. R. M. Walton, Nature 357, 479 (1992). 791. F. N. Tebbe, R. L. Halow, D. B. Chase, D. L. Thorn, G. C. Campbell, Jr., J. C. Calabrese, N. Herron, R. J. Young, Jr., and E. Wasserman, Science 256, 822 (1992). 792. M. Gu, T. B. Tang, C. Hu, and D. Feng, Phys. Rev. B 58, 659 (1998). 793. J. E. Schirber, R. A. Assink, G. A. Samra, B. Morosin, and D. Loy, Phys. Rev. B 51, 15552 (1995). 794. S. A. FitzGerald, T. Yildirim, L. J. Santodonato, D. A. Neumann, J. R. D. Copley, J. J. Rush, and F. Trouw, Phys. Rev. B 60, 6439 (1999). 795. S. A. FitzGerald, S. Forth, and M. Rinkoski, Phys. Rev. B 65, 140302 (2002). 796. B. Morosin, R. A. Assink, R. G. Dunn, T. M. Massis, J. E. Schirber, and G. H. Kwei, Phys. Rev. B 56, 13611 (1997). 797. W. R. Datars, T. R. Chien, R. K. Nkum, and P. K. Ummat, Phys. Rev. B 50, 4937 (1994). 798. W. R. Datars, J. D. Palidwar, and P. K. Ummat, J. Phys. Chem. Solids 57, 977 (1996). 799. R. Francis, P. K. Ummat, and W. R. Datarsi, J. Phys. Condens. Matter 9, 7223 (1997).

C60 -Based Materials 800. I. W. Locke, A. D. Darwish, H. W. Kroto, K. Prassides, R. Taylor, and D. R. M. Walton, Chem. Phys. Lett. 225, 186 (1994). 801. R. E. Douthwaite, M. L. H. Green, S. J. Heyes, M. J. Rosseinsky, and J. F. C. Turner, Chem. Commun. 1367 (1994). 802. G. Roth and P. Adelmann, Appl. Phys. A 56, 169 (1993). 803. J. D. Crane, P. B. hitchock, H. W. Kroto, R. Taylor, and D. R. M. Walton, Chem. Commun. 1764 (1992). 804. A. L. Balch, J. W. Lee, B. C. Noll, and M. M. Olmstead, J. Chem. Soc., Chem. Commun. 56 (1993). 805. M. F. Meidine, P. B. Hitchcock, H. W. Kroto, R. Taylor, and D. R. M. Walton, Chem. Commun. 1534 (1992). 806. P. W. Stephens, D. Cox, J. W. Lauher, L. Mihaly, J. B. Wiley, P. M. Allemand, A. Hirsch, K. Holczer, Q. Li, J. D. Thompson, and F. Wudl, Nature 355, 311 (1992). 807. P. Zhou, A. M. Rao, K. A. Wang, J. D. Robertson, C. Eloi, M. S. Meier, S. L. Ren, X. X. Bi, P. C. Eklund, and M. S. Dresselhaus, Appl. Phys. Lett. 60, 2871 (1992). 808. P. Zhou, Z.-H. Dong, A. M. Rao, and P. C. Eklund, Chem. Phys. Lett. 211, 337 (1993). 809. F. Cataldo, Polym. Int. 48, 143 (1999). 810. D. S. Cornett, I. J. Amster, M. A. Duncan, A. M. Rao, and P. C. Eklund, J. Phys. Chem. 97, 5036 (1993). 811. B. Burger, J. Winter, and H. Kuzmany, Synth. Met. 86, 2329 (1997). 812. A. Hassanien, J. Gasperic, J. Demsar, I. Muševic, and D. Mihailovic, Appl. Phys. Lett. 70, 417 (1997). 813. B. Sundqvist, Adv. Phys. 48, 1 (1999). 814. Y. Iwasa, T. Arima, R. M. Fleming, T. Siegrist, O. Zhou, R. C. Haddon, L. J. Rothberg, K. B. Lyons, H. L. Carter, Jr., A. F. Hebard, R. Tycko, G. Dabbagh, J. J. Krajewski, G. A. Thomas, and T. Yagi, Science 264, 1570 (1994). 815. G. Oszlanyi and L. Forro, Solid State Commun. 93, 265 (1995). 816. M. Núñez-Regueiro, L. Marques, J.-L. Hodeau, O. Béthoux, and M. Perroux, Phys. Rev. Lett. 74, 278 (1995). 817. B. Sundqvist, O. Andersson, U. Edlund, A. Fransson, A. Inaba, P. Jacobsson, D. Johnels, P. Launois, C. Meingast, R. Moret, T. Moritz, P.-A. Persson, A. Soldatov, and T. Wågberg, in “Fullerenes: Recent Advances in the Chemistry and Physics of Fullerenes and Related Materials” (K. M. Kadish and R. S. Ruoff, Eds.), ECS Proceedings, Vol. 3, p. 1014. The Electrochemical Society, Pennington, NJ, 1996. 818. R. Moret, P. Launois, P.-A. Persson, and B. Sundqvist, Europhys. Lett. 40, 55 (1997). 819. V. A. Davydov, L. S. Kashevarova, A. V. Rakhmanina, V. Agafonov, H. Allouchi, R. Ceolin, A. V. Dzyabchenko, V. M. Senyavin, and H. Szwarc, Phys. Rev. B 58, 14786 (1998). 820. V. A. Davydov, V. Agafonov, H. Allouchi, R. Ceolin, A. V. Dzyabchenko, and H. Szwarc, Synth. Met. 103, 2415 (1999). 821. V. A. Davydov, L. S. Kashevarova, A. V. Rakhmanina, V. Agafonov, H. Allouchi, R. Ceolin, A. V. Dzyabchenko, V. M. Senyavin, H. Szwarc, T. Tanaka, and K. Komatsu, J. Phys. Chem. B 103, 1800 (1999). 822. R. Moret, P. Launois, T. Wågberg, and B. Sundqvist, Eur. Phys. J. B 15, 253 (2000). 823. X. Chen and S. Yamanaka, Chem. Phys. Lett. 360, 501 (2002). 824. T. L. Makarova, T. Wågberg, B. Sundqvist, Xiao-Mei Zhu, E. B. Nyeanchi, M. E. Gaevski, E. Olsson, V. Agafonov, V. A. Davydov, A. V. Rakhamanina, and L. S. Kashevarova, Mol. Mater. 13, 151 (2000). 825. X. Chen, S. Yamanaka, K. Sako, Y. Inoue, and M. Yasukawa, Chem. Phys. Lett. 356, 291 (2002). 826. L. A. Chernozatonskii, N. R. Serebryanaya, and B. N. Mavrin, Chem. Phys. Lett. 316, 199 (2000). 827. M. Ata, N. Takahashi, and K. Nojima, J. Phys. Chem. 98, 9960 (1994). 828. N. Takahashi, H. Dock, N. Matsuzawa, and M. Ata, J. Appl. Phys. 74, 5790 (1993).

471 829. J. Kastner, H. Kuzmany, and L. Palmetshofer, Appl. Phys. Lett. 65, 543 (1994). 830. Y. Nakamura, Y. Mera, and K. Maeda, Appl. Phys. Lett. 77, 2834 (2000). 831. Y. B. Zhao, D. M. Poirier, R. J. Pechman, and J. H. Weaver, Appl. Phys. Lett. 64, 577 (1994). 832. A. Matsutani, F. Koyama, and K. Iga, Jpn. J. Appl. Phys. 37, 4211 (1998). 833. L. Marques, J.-L. Hodeau, and M. Núñez-Regueiro, Mol. Mater. 8, 49 (1996). 834. V. A. Davydov, L. S. Kashevarova, A. V. Rakhmanina, V. M. Senyavin, R. Céolin, H. Szwarc, H. Allouchi, and V. Agafonov, Phys. Rev. B 61, 11936 (2000). 835. V. A. Davydov, V. Agafonov, A. V. Dzyabchenko, R. Céolin, and H. Szwarc, J. Solid State Chem. 141, 164 (1998). 836. S. Okada, S. Saito, and A. Oshiyama, Phys. Rev. Lett. 83, 1986 (1999). 837. V. V. Brazhkin and A. G. Lyapin, Phys. Rev. Lett. 85, 5671 (2000). 838. S. Okada, S. Saito, and A. Oshiyama, Phys. Rev. Lett. 85, 5672 (2000). 839. L. Marques, M. Mezouar, J.-L. Hodeau, M. Núñez-Regueiro, N. R. Serebryanaya, V. A. Ivdenko, V. D. Blank, and G. A. Dubitsky, Science 283, 1720 (1999). 840. N. R. Serebryanaya and L. A. Chernozatonskii, Solid State Commun. 114, 537 (2000). 841. V. Agafonov, V. A. Davydov, L. S. Kashevarova, A. V. Rakhmanina, A. Kahn-Harari, P. Dubois, R. Céolin, and H. Szwarc, Chem. Phys. Lett. 267, 193 (1997). 842. V. A. Davydov, L. S. Kashevarova, A. V. Rakhmanina, A. V. Dzyabchenko, V. N. Agafonov, P. Dubois, R. Céolin, and H. Szwarc, JETP Lett. 66, 120 (1997). 843. K. Tanaka, Yu. Matsuura, Y. Oshima, T. Yamabe, Y. Asai, and M. Tokumoto, Solid State Commun. 93, 163 (1995). 844. C. H. Xu and G. E. Scuseria, Phys. Rev. Lett. 74, 274 (1995). 845. S. Okada and S. Saito, Phys. Rev. B 55, 4039 (1997). 846. S. Okada and S. Saito, Phys. Rev. B 59, 1930 (1999). 847. V. V. Belavin, L. G. Bulusheva, A. V. Okotrub, D. Tománek, J. Phys. Chem. Solids 61, 1901 (2000). 848. Y. Iwasa, K. Tanoue, T. Mitani, and T. Yagi, Phys. Rev. B 58, 16 374 (1998). 849. M. Shiraishi, M. Ramm, and M. Ata, Appl. Phys. A 74, 613 (2002). 850. F. Bommeli, L. Degiorgi, P. Wachter, Ö. Legeza, A. Jánossy G. Oszlanyi, O. Chauvet, and L. Forro, Phys. Rev. B 51, 14794 (1995). 851. T. L. Makarova, P. Scharff, B. Sundqvist, B. narymbetov, H. Kobayashi, M. Tokumoto, V. A. Davydov, A. V. Rakhmanina, and L. S. Kashevarova, Synth. Met. 121, 1099 (2001). 852. T. L. Makarova, B. Sundqvist, P. Scharff, M. E. Gaevski, E. Olsson, V. A. Davydov, A. V. Rakhmanina, and L. S. Kashevarova, Carbon 39, 2203 (2001). 853. A. V. Okotrub, V. V. Belavin, L. G. Bulusheva, V. A. Davydov, T. L. Makarova, and D. Tománek, J. Chem. Phys. 115, 5637 (2001). 854. S. G. Buga, V. D. Blank, G. A. Dubitsky, L. Edman, X.-M. Zhu, E. B. Nyeanchi, and B. Sundqvist, J. Phys. Chem. Solids 61, 1009 (2000). 855. Y. Wang, J. M. Holden, A. M. Rao, P. C. Eklund, U. D. Venkateswaran, D. Eastwood, R. L. Lidberg, G. Dresselhaus, and M. S. Dresselhaus, Phys. Rev. B 51, 4547 (1995). 856. C. Xu, G. Chen, E. Xie, and J. Gong, Appl. Phys. Lett. 70, 2641 (1997). 857. G. Lyapin, V. V. Brazhkin, S. G. Lyapin, S. V. Popova, T. D. Varfolomeeva, R. A. Voloshin, A. A. Pronin, N. E. Sluchanko, A. G. Gavrilyuk, and I. A. Trojan, Phys. Status Solidi B 211, 401 (1999). 858. R. A. Wood, M. H. Lewis, M. R. Lees, S. M. Bennington, M. G. Cain, and N. Kitamura, J. Phys.: Condens. Matter 14 L385 (2002). 859. V. M. Levin, V. D. Blank, V. M. Prokhorov, Ja. M. Soifer, and N. P. Kobelev, J. Phys. Chem. Solids 61, 1017 (2000).

472 860. V. D. Blank, S. G. Buga, N. R. Serebryanaya G. A. Sulyanov, M. Yu. Popov, V. N. Denisov, A. N. Ivlev, and B. N. Mavrin, Phys. Lett. A 220, 149 (1996). 861. V. Blank, M. Popov, G. Pivovarov, N. Lvova, K. Goglinsky, and V. Reshetov, Diamond Relat. Mater. 7, 427 (1998). 862. G. K. Wertheir and D. N. E. Buchanam, Solid State Commun. 88, 97 (1993). 863. A. F. Hebard, C. B. Eom, Y. Iwasa, K. B. Lyons, G. A. Thomas, D. H. Rapkine, R. M. Fleming, R. C. Haddon, J. M. Phillips, J. H. Marshall, and R. H. Eick, Phys. Rev. B 50, 17740 (1994). 864. B. Wang, H. Wang, Y. Li, B. Xie, and J. G. Hou, Physica C 282, 735 (1997). 865. L. Xiang, H. Wang, W. N. Wang, Y. J. Tang, H. W. Zhao, W. S. Zhan, and J. G. Hou, J. Phys.: Condens. Matter 13, 3987 (2001). 866. J. G. Hou, Y. Wang, W. Xu, Li Yang, and Y. H. Zhang, Appl. Phys. Lett. 70, 3100 (1997). 867. J. G. Hou, Y. Wang, W. Xu, L. Yang, and Y. H. Zhang, Mater. Lett. 34, 36 (1998). 868. J. G. Hou, Y. Li, Y. Wang, W. Xu, J. Zuo, and Y. H. Zhang, Phys. Status Solidi A 163, 403 (1997). 869. J. G. Hou, X. Li, and H. Wang, Adv. Mater. 11, 1124 (2000). 870. R. Popescu, D. Macovei, A. Devenyi, R. Manaila, P. B. Barna, A. Kovacs, and J. L. Lábár, Eur. Phys. J. B 13, 737 (2000). 871. B. Wang, H. Q. Wang, Y. Q. Li, S. Y. Zhang, and J. G. Hou, Mater. Res. Bull. 35, 511 (2000). 872. Y. Li, Y. Wang, W. Xu, J. G. Hou, and Y. H. Zhang, Physica C 282, 737 (1997). 873. X. D. Zhang, W. B. Zhao, K. Wu, J. L. Zhang, C. Y. Li, D. L. Yin, Z. N. Gu, X. H. Zhou, and Z. X. Jin, Chem. Phys. Lett. 228, 100 (1994). 874. J. G. Hou, X. Li, H. Wang, and B. Wang, J. Phys. Chem. Solids 61, 995 (2000). 875. X. Li, Y. J. Tang, H. W. Zhao, and W. S. Zhan, Appl. Phys. Lett. 77, 984 (2000). 876. X. Li, Y. J. Tang, H. Wang, H. W. Zhao, W. S. Zhan, and J. G. Hou, Phys. Status Solidi A 186, 57 (2001). 877. H. Q. Wang, J. G. Hou, O. Takeuchi, Y. Fujisuku, and A. Kawazu, Phys. Rev. B 61, 2199 (2000). 878. Z. Zhao, H. Q. Wang, B. Wang, J. G. Hou, G. L. Liu, and X. F. Jin, Phys. Rev. B 65, 235413 (2002). 879. J. A. Block, K. Parvin, J. L. Alpers, T. Sezen, and R. LaDuca, IEEE Trans. Magn. 34, 982 (1998). 880. B. Zhao, J. Y. Chow, and X. Yan, J. Appl. Phys. 79, 6022 (1996). 881. L. A. Zheng, B. M. Lairson, E. V. Barrera, and R. D. Shull, Appl. Phys. Lett. 77, 3242 (2000). 882. T. C. Arnolduseen, in “Magnetic Recording Technology” (C. D. Mee and E. D. Daniel, Eds.), pp. 4–36. McGraw–Hill, New York, 1995. 883. K. E. Johnson, J. B. Mahlke, K. J. Schultz, and A. C. Wall, IEEE Trans. Magn. 29, 215 (1993). 884. T. Guo, C. Jin, and R. E. Smalley, J. Phys. Chem. 95, 4948 (1991). 885. Z. C. Ying, R. L. Hettich, R. N. Compton, and R. E. Haufler, J. Phys. B 29, 4935 (1996). 886. Y. F. Jia, Q. Zhu, and K. M. Thomas, Chem. Phys. Lett. 364, 171 (2002). 887. N. L. Clipston, T. Brown, Y. Y. Vasil’ev, M. P. Barrow, R. Herzschuh, U. Reuther, A. Hirsch, and T. Drewello, J. Phys. Chem. A 104, 9171 (2000). 888. J. C. Hummelen, B. Knight, J. Pavlovich, R. Gonzˆalez, and F. Wudl, Science 269, 1554 (1995). 889. T. Grösser, M. Prato, V. Lucchini, A. Hirsch, and F. Wudl, Angew. Chem. Int. Ed. 34, 1341 (1995). 890. B. Nuber and A. Hirsch, Chem. Commun. 1421 (1996). 891. W. Andreoni, A. Curioni, K. Holczer, K. Prassides, M. KeshavarzK, J. C. Hummelen, and F. Wudl, J. Am. Chem. Soc. 118, 11335 (1996).

C60 -Based Materials 892. C. Bellavia-Lund, R. González, J. C. Hummelen, R. G. Hicks, A. Sastre, and F. Wudl, J. Am. Chem. Soc. 119, 2946 (1997). 893. A. Gruss, K.-P. Dinse, A. Hirsch, B. Nuber, and U. Reuther, J. Am. Chem. Soc. 119, 8728 (1997). 894. P.-M. Allemand, G. Srdanov, A. Koch, K. Khemani, F. Wudl, Y. Rubin, M. Diederich, M. M. Alvarez, S. J. Anz, and R. L. Whetten, J. Am. Chem. Soc. 113, 2780 (1991). 895. J. Dong, J. Jiang, Z. D. Wang, and D. Y. Xing, Phys. Rev. B 51, 1977 (1995). 896. W. Andreoni, F. Gygi, and M. Parrinello, Chem. Phys. Lett. 190, 159 (1992). 897. T. Pichler, M. Knupfer, R. Friedlein, S. Haffner, B. Umlauf, M. Golden, O. Knauff, H.-D. Bauer, J. Fink, M. Keshavarz-K, C. Bellavia-Lund, A. Sastre, J. C. Hummelen, and F. Wudl, Synth. Met. 86, 2313 (1997). 898. S. Haffner, T. Pichler, M. Knupfer, B. Umlauf, R. Friedlein, M. Golden, J. Fink, M. Keshavarz-K, C. Bellavia-Lund, A. Sastre, J. C. Hummelen, and F. Wudl, Eur. Phys. J. B 1, 11 (1998). 899. C. M. Brown, E. Beer, C. Bellavia, L. Cristofolini, R. Gonzalez, M. Hanfland, D. Häusermann, M. Keshavarz-K, K. Kordatos, K. Prassides, and F. Wudl, J. Am. Chem. Soc. 118, 8715 (1996). 900. C. M. Brown, L. Cristofolini, K. Kordatos, K. Prassides, C. Bellavia-Lund, R. Gonzalez, M. Keshavarz-K., F. Wudl, A. K. Cheetham, J. P. Zhang, W. Andreoni, A. Curioni, A. N. Fitch, and P. Pattison, Chem. Mater. 8, 2548 (1996). 901. H.-J. Muhr, R. Nesper, B. Schnyder, and R. Kötz, Chem. Phys. Lett. 249, 399 (1996). 902. T. Nakamura, K. Ishikawa, K. Yamamoto, T. Ohana, S. Fujiwara, and Y. Koga, Phys. Chem. Chem. Phys. 1, 2631 (1999). 903. U. Reuther and A. Hirsch, Chem. Commun. 1401 (1998). 904. Y. Tobe, H. Nakanishi, M. Sonoda, T. Wakabayashi, and Y. Achiba, Chem. Commun. 1625 (1999). 905. L. Hultman, S. Stafström, Z. Czigány, J. Neidhardt, N. Hellgren, I. F. Brunell, K. Suenaga, and C. Colliex, Phys. Rev. Lett. 87, 225503 (2001). 906. T. Nakamura, K. Ishikawa, A. Goto, M. Ishihara, T. Ohana, and Y. Koga, Diamond Relat. Mater. 10, 1228 (2001). 907. D. A. Jelski, J. R. Bowser, R. James, X. Xia, G. Xinfu, J. Gao, and T. F. George, J. Cluster. Sci. 4, 173 (1993). 908. T. Kimura, T. Sugai, and H. Shinohara, Chem. Phys. Lett. 256, 269 (1996). 909. M. Pellarin, C. Ray, J. Lermé, J. L. Vialle, M. Broyer, X. Blase, P. Kéghélian, P. Mélinon, and A. Perez, Eur. Phys. J. D 9, 49 (1999). 910. M. Pellarin, C. Ray, P. Mélinon, J. Lermé, J. L. Vialle, P. Kéghélian, A. Perez, and M. Broyer, Chem. Phys. Lett. 277, 96 (1997). 911. C. Ray, M. Pellarin, J. L. Lermé, J. L. Vialle, M. Broyer, X. Blase, P. Kéghélian, P. Mélinon, and A. Perez, Phys. Rev. Lett. 80, 5365 (1998). 912. I. M. L. Billas, F. Tast, W. Branz, N. Malinowski, M. Heinebrodt, T. P. Martin, M. Boero, C. Massobrio, and M. Parrinello, Eur. Phys. J. D 9, 337 (1999). 913. I. M. L. Billas, C. Massobrio, M. Boero, M. Parrinello, W. Branz, F. Tast, N. Malinowski, M. Heinebrodt, and T. P. Martin, J. Chem. Phys. 111, 6787 (1999). 914. J. F. Christian, Z. Wan, and S. L. Anderson, Chem. Phys. Lett. 199, 373 (1992). 915. J. J. Stry and J. F. Garvey, Chem. Phys. Lett. 243, 199 (1995). 916. C. Moschel and M. Jansen, Z. Anorg. Allg. Chem. 625, 175 (1999). 917. W. Branz, I. M. L. Billas, N. Malinowski, F. Tast, M. Heinebrodt, and T. P. Martin, J. Chem. Phys. 109, 3425 (1998). 918. L. D. Strout, R. L. Murry, C. H. Xu, W. C. Eckhoff, G. F. Odom, and G. E. Scuseria, Chem. Phys. Lett. 214, 576 (1993). 919. N. Matsuzawa, M. Ata, D. A. Dixon, and G. Fitzgerald, J. Phys. Chem. 98, 2555 (1994).

C60 -Based Materials 920. M. Menon, K. R. Subbaswamy, and M. Sawtarie, Phys. Rev. B 49, 13966 (1994). 921. D. Porezag, M. R. Pederson, T. Frauenheim, and T. K. Köhler, Phys. Rev. B 52, 14963 (1995). 922. J. Kürti and K. Németh, Chem. Phys. Lett. 256, 119 (1996). 923. G. E. Scuseria, Chem. Phys. Lett. 257, 583 (1996). 924. S. Osawa, M. Sakai, and E. Osawa, J. Phys. Chem. A 101, 1378 (1997). 925. G.-W. Wang, K. Komatsu, Y. Murata, and M. Shiro, Nature 387, 583 (1997). 926. K. Komatsu, G.-W. Wang, Y. Murata, T. Tanaka, and K. Fujiwara, J. Org. Chem. 63, 9358 (1998). 927. S. Lebedkin, A. Gromov, S. Giesa, R. Gleiter, B. Renker, H. Rietschel, and W. Krätschmer, Chem. Phys. Lett. 285, 210 (1998). 928. V. A. Davydov, L. S. Kashevarova, A. V. Rakhmanina, V. M. Senyavin, V. Agafonov, R. Ceolin, and H. Szwarc, JETP Lett. 68, 928 (1998). 929. Y. Iwasa, K. Tanoue, T. Mitani, A. Izuoka, T. Sugawara, and Yagi, Chem. Commun. 1411 (1998). 930. V. A. Davydov, L. S. Kashevarova, A. V. Rakhmanina, V. M. Senyavin, R. Céolin, H. Szwarc, H. Allouchi, and V. Agafonov, Phys. Rev. B 61, 11936 (2000). 931. A. V. Dzyabchenko, V. Agafonov, and V. A. Davydov, J. Phys. Chem. A 103, 2812 (1999). 932. M. R. Pederson and A. A. Quong, Phys. Rev. Lett. 74, 2319 (1995). 933. V. C. Long, J. L. Musfeldt, K. Kamarás, G. B. Adams, J. B. Page, Y. Iwasa, and W. E. Mayo, Phys. Rev. B 61, 13191 (2000). 934. B. Ma, J. E. Riggs, and Y.-P. Sun, J. Phys. Chem. B 102, 5999 (1998). 935. Y. Chai, T. Guo, C. Jin, R. E. Haufler, L. P. F. Chibante, J. Fure, L. Wang, J. M. Alford, and R. E. Smalley, J. Phys. Chem. 95, 7564 (1991). 936. J. Heath, S. C. O’Brien, Q. Zhang, Y. Liu, R. F. Curl, H. W. Kroto, F. K. Tittel, and R. E. Smalley, J. Am. Chem. Soc. 107, 7779 (1985). 937. F. D. Weiss, J. L. Elkind, S. C. O’Brien, R. F. Curl, and R. E. Smalley, J. Am. Chem. Soc. 110, 4464 (1988). 938. Y. Kubozono, H. Maeda, Y. Takabayashi, K. Hiraoka, T. Nakai, S. Kashino, S. Emura, S. Ukita, and T. Sogabe, J. Am. Chem. Soc. 118, 6998 (1996). 939. T. Inoue, Y. Kubozono, S. Kashino, Y. Takabayashi, K. Fujitaka, M. Hida, M. Inoue, T. Kanbara, S. Emura, and T. Uruga, Chem. Phys. Lett. 316, 381 (2000). 940. T. Ogawa, T. Sugai, and H. Shinohara, J. Am. Chem. Soc. 122, 3538 (2000). 941. T. Kanbara, Y. Kubozono, Y. Takabayashi, S. Fujiki, S. Iida, Y. Haruyama, S. Kashino, S. Emura, and T. Akasaka, Phys. Rev. B 64, 113403 (2001). 942. A. Gromov, W. Krätschmer, N. Krawez, R. Tellgmann, and E. E. B. Campbell, Chem. Commun. 2003 (1997). 943. Ch. Kusch, N. Krawez, R. Tellgmann, B. Winter, and E. E. B. Campbell, Appl. Phys. A 66, 293 (1998). 944. T. Almeida Murphy, T. Pawlik, A. Weidinger, M. Höhne, R. Alcala, and J.-M. Spaeth, Phys. Rev. Lett. 77, 1075 (1996). 945. J. A. Larsson, J. C. Greer, W. Harneit, and A. Weidinger, J. Phys. Chem. 116, 7849 (2002). 946. M. Waiblinger, K. Lips, W. Harneit, and A. Weidinger, Phys. Rev. B 64, 159901 (2001). 947. B. Pietzak, M. Waiblinger, T. Almeida Murphy, A. Weidinger, M. Höhne, E. Dietel, and A. Hirsch, Chem. Phys. Lett. 279, 259 (1997). 948. C. Knapp, N. Weiden, H. Käss, K.-P. Dinse, B. Pietzak, M. Waiblinger, and A. Weidinger, Mol. Phys. 95, 999 (1998). 949. A. Weidinger, M. Waiblinger, B. Pietzak, and T. A. Murphy, Appl. Phys. A 66, 287 (1998).

473 950. E. Dietel, A. Hirsch, B. Pietzak, M. Waiblinger, K. Lips, A. Weidinger, A. Gruss, and P. Dinse, J. Am. Chem. Soc. 121, 2432 (1999). 951. W. Harneit, Phys. Rev. A 65, 32322 (2002). 952. M. Saunders, H. A. Jiménez-Vázquez, R. J. Cross, and R. J. Poreda, Science 259, 1428 (1993). 953. M. Saunders, R. J. Cross, H. A. Jiménez-Vázquez, R. Shimshi, and A. Khong, Science 271, 1693 (1996). 954. B. A. Dicamillo, R. L. Hettich, G. Guiochon, R. N. Compton, M. Saunders, H. A. Jiménez-Vázquez, A. Khong, and R. J. Cross, J. Phys. Chem. 100, 9197 (1996). 955. M. S. Syamala, R. J. Cross, and M. Saunders, J. Am. Chem. Soc. 124, 6216 (2002). 956. K. Yamamoto, M. Saunders, A. Khong, R. J. Cross, Jr., M. Grayson, M. L. Gross, A. F. Benedetto, and R. B. Weisman, J. Am. Chem. Soc. 121, 1591 (1999). 957. K. Kimata, K. Hosoya, T. Areki, and N. Tanaka, J. Org. Chem. 58, 282 (1993). 958. L.-S. Wang, O. Cheshnovsky, R. E. Smalley, J. P. Carpenter, and S. J. Hwu, J. Chem. Phys. 96, 4028 (1992). 959. U. Zimmermann, N. Malinowski, A. Burkhardt, and T. P. Martin, Carbon 33, 995 (1995). 960. U. Zimmermann, A. Burkhardt, N. Malinowski, U. Näher, and T. P. Martin, J. Chem. Phys. 101, 2244 (1994). 961. U. Zimmermann, N. Malinowski, U. Näher, S. Frank, and T. P. Martin, Phys. Rev. Lett. 72, 3542 (1994). 962. B. Palpant, Y. Negishi, M. Sanekata, K. Miyajima, S. Nagao, K. Judai, D. M. Rayner, B. Simard, P. A. Hackett, A. Nakajima, and K. Kaya, J. Chem. Phys. 114, 8459 (2001). 963. B. Palpant, A. Otake, F. Hayakawa, Y. Negishi, G. H. Lee, A. Nakajima, and K. Kaya, Phys. Rev. B 60, 4509 (1999). 964. Ph. Dugourd, R. Antoine, D. Rayane, I. Compagnon, and M. Broyer, J. Chem. Phys. 114, 1970 (2001). 965. R. Antoine, D. Rayane, E. Benichou, Ph. Dugourd, and M. Broyer, Eur. Phys. J. D 12, 147 (2000). 966. D. Rayane, R. Antoine, Ph. Dugourd, E. Benichou, A. R. Allouche, M. Aubert-Frécon, and M. Broyer, Phys. Rev. Lett. 84, 1962 (2000). 967. F. Tast, N. Malinowski, S. Frank, M. Heinebrodt, I. M. L. Billas, and T. P. Martin, Phys. Rev. Lett. 77, 3529 (1996). 968. F. Tast, N. Malinowski, S. Frank, M. Heinebrodt, I. M. L. Billas, and T. P. Martin, Z. Phys. D 40, 351 (1997). 969. Q. Kong, Y. Shen, L. Zhao, J. Zhuang, S. Qian, Y. Li, Y. Lin, and R. Cai, J. Chem. Phys. 116, 128 (2002). 970. Y.-K. Kwon, D. Tománek, and S. Iijima, Phys. Rev. Lett. 82, 1470 (1999). 971. R. F. Service, Science 292, 45 (2001). 972. A. G. Rinzler, J. Liu, H. Dai, P. Nikolaev, C. B. Huffman, F. J. Rodríguez-Macías, P. J. Boul, A. H. Lu, D. Heymann, D. T. Colbert, R. S. Lee, J. E. Fischer, A. M. Rao, P. C. Eklund, and R. E. Smalley, Appl. Phys. A 67, 29 (1998). 973. B. W. Smith, M. Monthioux, and D. E. Luzzi, Nature 396, 323 (1998). 974. B. W. Smith and D. E. Luzzi, Chem. Phys. Lett. 321, 169 (2000). 975. S. Bandow, M. Takizawa, K. Hirahara, M. Yudasaka, and S. Iijima, Chem. Phys. Lett. 337, 48 (2001). 976. B. W. Smith, M. Monthioux, and D. E. Luzzi, Chem. Phys. Lett. 315, 31 (1999). 977. K. Hirahara, S. Bandow, K. Suenaga, H. Kato, T. Okazaki, H. Shinohara, and S. Iijima, Phys. Rev. B 64, 115420 (2001). 978. K. Hirahara, K. Suenaga, S. Bandow, H. Kato, T. Okazaki, H. Shinohara, and S. Iijima, Phys. Rev. Lett. 85, 5384 (2000). 979. B. W. Smith, R. M. Russo, S. B. Chikkannanavar, and D. E. Luzzi, J. Appl. Phys. 91, 9333 (2002). 980. S. Berber, Y.-K. Kwon, and D. Tománek, Phys. Rev. Lett. 88, 85502 (2002).

474 981. K. Hirahara, S. Bandow, K. Suenaga, H. Kato, T. Okazaki, H. Shinohara, and S. Iijima, Phys. Rev. B 65, 45419 (2001). 982. J. Varo, M. C. Llaguno, B. C. Satishkumar, D. E. Luzzi, and J. E. Fischer, Appl. Phys. Lett. 80, 1450 (2002). 983. H. Hongo, F. Nihey, M. Yudasaka, T. Ichihashi, and S. Iijima, Physica B 323, 244 (2002).

C60 -Based Materials 984. L. Langer, V. Bayot, E. Grivei, J.-P. Issi, J. P. Heremans, C. H. Olk, L. Stockman, C. Van Haesendonck, and Y. Bruynseraede, Phys. Rev. Lett. 76, 479 (1996). 985. T. P. Martin, U. Näher, H. Schaber, and U. Zimmermann, Phys. Rev. Lett. 70, 3079 (1993). 986. W. Zhang, L. Liu, J. Zhuang, and Y. Li, Phys. Rev. B 62, 8276 (2000).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Calixarenes Chebrolu P. Rao, Mishtu Dey Indian Institute of Technology, Mumbai, India

CONTENTS 1. Introduction 2. Calixarene Conformations 3. Organic Functionalization 4. Metallocalixarenes 5. Physical Properties of Calixarenes 6. Spectral Properties of Calixarenes 7. Structures by X-Ray Crystallography 8. Applications 9. Conclusions Glossary References

1. INTRODUCTION 1.1. Genesis of Calixarenes About 150 years ago, Adolph von Baeyer demonstrated the formation of a hard, resinous, and noncrystalline product from the reaction of aqueous formaldehyde with phenol that remained essentially uncharacterized for almost 70 years. However, three decades later, Leo Baekeland smartly designed a process for this product, which he marketed under the name “bakelite,” with tremendous commercial success [1]. In 1942 Zinke and Ziegler researched the products of the reaction of phenol and its substitutes with formaldehyde. Phenol itself reacts with formaldehyde in acid or base at the ortho- and para-positions to form a highly crosslinked three-dimensional polymer in which each phenolic residue is attached to three other neighbors as shown in Figure 1. On the other hand, para-alkylphenol can react only at its two ortho-positions and thereby forms a linear onedimensional polymer (Fig. 2) by reducing the cross-linking possibilities, in addition to the formation of several macrocylic oligomers. However, the same reaction in the presence of base at high temperatures yielded high melting and high ISBN: 1-58883-057-8/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

molecular weight substance to which Zinke assigned a cyclic tetrameric structure [2], where the product was proven to be a mixture later on as shown in Figure 3. This problem was not overcome until late 1970s when Gutsche became intrigued by the emerging area of enzyme mimics where he explored the Zinke cyclic tetramers as cavity-containing substances appropriate for enzyme mimic building blocks. He later showed that a careful control of reaction conditions could provide good yields of pure compounds of various ring sizes, calix[n]arenes.

1.2. Scope of the Chapter The main aim of this chapter is to bring into focus the newer developments of calixarenes and its derivatives and the many applications exhibited by these systems. In order to fulfill such a focus, the authors take the approach of showing the developments that took place since the realization of simple calixarenes and their putative applications in biology and materials leading to the nanoscience of such supramolecular systems. Since the work carried out as well as the understanding of calix[6 & 8]arenes is rather limited in the literature, we confine our discussions mostly to calix[4]arenes.

1.3. Nomenclature The name calixarene is of Greek origin (Greek calix, a chalice or cup; arene referring to the aromatic ring), having recognized the cyclic tetramer in the conformation in which all of its four aryl groups are oriented in the same direction of a chalice or cuplike shape [3–7] similar to that of a Greek crater vase (Fig. 4). The substitution at the paraposition preceeds calix[n]arene in order to accommodate the number of aryl groups present by giving a numerical value to [n]. For example, the cyclic tetramer derived from ptert-butylphenol is written as p-tert-butylcalix[4]arene. For a more systematic application of the calixarene nomenclature, the basic name calix[n]arene is retained, and identities of all substituents are indicated and their positions are specified by a number (e.g., calixarene is derived from p-tert-butylphenol and formaldehyde is named as 5,11,17,23-tetra-tert-butyl25,26,27,28-tetrahydroxycalix[4]arene as shown in Fig. 5).

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 1: Pages (475–497)

Calixarenes

476 OH

OH

OH 2

OH

2 1

OH

HO

41

35

OH 1

1 2 37 OH 38 OH HO 42

2 OH OHHO 2 2 OH 2

29 2 5

26

OH

HO

5

8

35

HO 41

39 OH

1

14

HO

51 OH

49

2 5

HO 56 8

52 OH 11 29

40

23

1

50 OH

32

OH 8

20

HO

38

32 1

47

44

HO OH 53

26

17

HO

55 11

54 14

20 23

17

OH Figure 5. Structure and numbering scheme for calix[n]arenes, where n = 4 6, and 8. Figure 1. Schematic view of a highly cross-linked three-dimensional polymer, bakelite, obtained from the condensation of phenol with formaldehyde.

OH

OH

OH

As this nomenclature is very inconvenient, in this chapter we have also used the appropriate short form as used by others in the literature.

OH

2. CALIXARENE CONFORMATIONS R

R

R

R

Figure 2. Schematic view of one-dimensional polymer formed from the linear condensation of para-substituted phenol with formaldehyde. R = any alkyl group.

OH -+ CH2O HO

OH OH HO OH

m

Figure 3. Cyclic condensation of p-tert-butyl phenol with formaldehyde resulting in the formation of highly ordered structures referred to as calixarenes. Ring size can be controlled based on the reaction conditions, leading to p-tert-butyl calix[4, 6, 8]arenes m = 1 3 5.

Calixarenes generally show conformational mobility in solution and are frozen into a particular conformation only upon crystallization. X-ray crystallographic structures have been established for simple and functionally modified calixarenes. Simple calixarenes have many conformational isomers because of two possible rotational modes of phenol units, the para-substituent and the oxygen through the annulus rotation (Fig. 6), and hence shows fluxional behavior [8]. Functionalized calixarenes, at either the lower or upper rim, would have further influence on the conformational mobility and in some derivatives such mobility is frozen. Calix[4]arene shows four conformers and these are designated by Gutsche and co-workers as the “cone” (C2v , C4v ), “partial-cone” (Cs ), “1,2-alternate” (C2h ), and “1,3alternate” (D2d ) forms [9] as shown in Figure 7. All the four conformations are accessible by rotation about the bridging methylene groups. In the cone conformation all four phenolic OH groups are involved in circular hydrogen bonding. On the other hand, calix[6]arene can assume 8 up–down conformational isomers, designated as cone, partial-cone, 1,2-alternate, 1,3-alternate, 1,4-alternate, 1,2,3-alternate, 1,2,4-alternate, and 1,3,5-alternate. Because of the increased flexibility of this system, additional conformations are possible in which one or more aryl rings assumes a position approximately in the average plane of the molecule designated as an out alignment and named a winged conformation by Gutsche. In p-tert-butylcalix[8]arene, the molecule

R'

Figure 4. Structure of calix[4]arene resembling Greek crater vase. Adapted from [3], V. Bohmer and J. Vicens, “Calixarenes: A Versatile Class of Macrocyclic Compounds.” Kluwer Academic, Dordrecht, 1991. © 1991, Kluwer Academic Publishers.

O R

Para-substituent through the annulus rotation

n

Oxygen through the annulus rotation

Figure 6. Annulus rotation in calixarenes. R = any substituent including tert-butyl group.

Calixarenes

477

Figure 7. Conformations exhibited by calix[4]arene: (a), (b), (c), and (d) represent cone, partial-cone, 1,2-alternate, and 1,3-alternate conformations respectively. R1 = any group and R2 = any group including the tert-butyl group.

exists in a pleated loop form. For calix[8]arene, there are 16 up–down forms, as well as many others in which one or more aryl groups assume the out alignment [10].

3. ORGANIC FUNCTIONALIZATION Since the pathway for conformational inversion in the calix[4]arene involves rotation of the aryl groups in a direction that brings the OH groups through the annulus of the macrocyclic ring, the most obvious way to curtail this motion is by derivatization. Thus, functionalization of calixarenes at the lower rim (phenolic OH groups) or at the upper rim (para-position of the aromatic moiety) often produces derivatives with reduced conformational mobility (Fig. 8). It is indeed possible to selectively introduce organic functionalities both on the lower rim as well as on the upper rim of calix[4]arenes. Due to chemical modifications, calixarenes and their derivatives can act as useful receptors or baskets for a range of guests including neutral, cationic, and anionic species. In these modifications the calix[4]arene acts as a platform to which an array of functional groups is attached resulting in the cone conformation and hence is capable of acting as the primary binding site.

3.1. Lower Rim Functionalization The lower rims of the calixarenes are already functionalized with hydroxyl groups which provide excellent sites for the introduction of other functional groups, by either direct acylation or alkylation (Williamson method) by bringing in groups such as –CH2 C(O)R (R = alkyl [11, 12], or aryl [13], or N(alkyl)2 [14, 15], or –NHCH(R′ )COOH (gly, ala, asp, glu) [16] or –NH-CH2 Py [16], –CH2 CH2 OR [17, 18], –CH2 P(O)Ph2 [19, 20], CH2 –pyridyl [21, 22], CH2 CN [12, 23], –CH2 C CH) [23]. Further, selective alkylation in terms of the degree as well as the site specificity can be achieved by appropriately manipulating the OH OH OH

HO OH OH OH OH

R

R

R

R

Lower rim

experimental conditions [24, 25]. In the literature, strategies are reported for di-functionalization of distal (1,3) and proximal (1,2) positions in p-R-calix[4]arenes [26]. We have recently obtained 1,3-di-derivatives built through amide linkages originating from amino acids, heterocyclic amines, and saccharide amines and further derivatized so that 2,4-OH groups are methylated [27, 28]. The conformational freezing can also be achieved by bridging between transannular oxygens at the lower rim resulting in crown etherlike structures [29, 30].

3.2. Upper Rim Functionalization The p-tert-butyl group attached to the phenyl ring is removed by de-alkylation using AlCl3 -phenol as catalyst [31, 32]. A variety of methods are available to selectively or totally introduce functional groups into the p-position of the phenyl ring, which includes electrophilic substitution, such as halogenation (Cl, Br, and I) [33, 34], nitration [35], sulphonation [36, 37], acylation [31], and formylation [38–40]; para-Claisen rearrangement [41]; chloromethylation [42] and subsequent replacement of chloride, resulting in the formation of a variety of potential ligands; and Mannich reaction with dialkylamines followed by conversion of nitrogen into quaternary structure and then treatment with nucleophiles [43].

4. METALLOCALIXARENES Calixarenes and their functionalized derivatives provide the basic structural framework for developing coordination chemistry, and when properly tuned with appropriate functional groups to provide biomimetic model systems. Both simple (parent) and functionalized calixarenes possessing O, S, N donor groups either at the upper rim or at the lower rim have been used as ligands to complex the transition metal ions.

4.1. Metal Ion Complexes of Simple Calixarenes Simple calixarenes can bind to metal ions via deprotonated phenolate moieties at the lower rim. Neutral complexes of p-tert-butyl-calix[4]arene (metal to ligand ratio) using Ti(IV) (2:2), Fe(III) (2:2), Co(III) (3:2), Nb(V) (1:1 with Nb O), Ta(V) (1:1 with Ta O), Mo(VI) (1:1 with M NR), and W(VI) (1:1 with W O or W(Cl)2 ) were reported in the literature as shown in Figure 9a–f [44–49]. Harrowfield et al. [50–54] have reported a series of lanthanide metal ion complexes of europium, terbium, and cerium with p-tert-buytlcalix[4]arene. The two europium ions were coordinated to the calixarene either as a 2:2 metal/calix[4]arene complex (Fig. 9g) or as a 2:1 metal/calix[8]arene complex.

4.2. Calixarenes Possessing Pendant Binding Sites

Upper rim R

R

R

R

Figure 8. “Lower rim” and “upper rim” in calix[4]arene where R = tertbutyl.

Introduction of appropriate functional groups on either the upper or lower rim leads to the derivatives of a versatile nature and would be of great use in a variety of applications. Calixarenes possessing different donor groups (N, P, S)

Calixarenes

478 t But Bu

But

t But Bu

But

O

O

Ti O

O

O

O Me3SiO

NH3

O

NH3

Fe

Fe

O

O

O

OSiMe3

Co

SiMe3

O

Co

Co

1

O

O

H2N

But But

But

But

(a)

But

(b)

O

R

But

O

O

N

N

R

R

R

OH OH O

N

N

(c)

O

OSO Mo

R

R

R

t But Bu

But

But O

O

O

O

O

O

W

But

t But Bu

But

O

O

Cl

(e)

Cl

(f)

O

O

OH OH O

But

But

O O

But

O dmf

O

O

O

N

But

O Eu

H N

N H

O

Eu

O

But

O

dmf

dmf

But

But

t But Bu

W

O

(d)

(b)

But

O

N

N

N

(a) R

But

t But Bu

O

thf

O

But But

But

But

SiMe3

O But But

But

NH2 NH2

H2N

O

O

O

3

O

O

O

But

But

O

Ti O

t But Bu

But

But

N

But

dmf

N But

including pendant acids have been synthesized [16]. Some of these molecules and their coordination chemistry aspects are summarized in this section.

4.3. Coordination Chemistry Aspects of Calixarenes with Pendant Nitrogen Donors Nitrogen containing donor groups have been introduced at either the lower or upper rim of the calix[4]arene as shown in Figure 10 [55, 56]. Interaction of ligand (a) with Ni(II), Cu(II), Pd(II), Co(II), and Fe(II) metal salts yields no complexes attributable to the presence of four mobile ethylamine arms that prevented the formation of rigid and/or closed cavities. However, when ligand (b) is reacted with Ni(ClO4 )2 ·6H2 O and NaN3 , it yields a binuclear complex (d) and when ligand (c) is reacted with Ru(bipy)2 complex, it results in the formation of a (bipy)2 Ru(bipy-calixarene) complex.

A limited number of thio-derivatives of calixarenes, viz., tetra mercaptocalix[4]arene and dihydroxy-di mercaptocalix[4]arene, were reported in the literature [57–61]. The former exists in 1,3-alternate conformation both in the free

N

N

N

Ni

N

N

N

Ni N

N

N

N

N

(d)

Figure 10. (a)–(c) Calix[4]arene derivatives with pendant nitrogen donors; (d) dinuclear nickel complex of (b); R = p-O2 SC6 H4 Br.

state as well as in its complexed state. The complex with Hg(acetate)2 yielded 2:1 with tetra-mercaptocalix[4]arene and 1:1 with dihydroxy-dimercaptocalix[4]arene, and in both cases the calixarene exists in 1,3-alternate conformation as shown in Figure 11.

But

But

But

But Hg

SH SH

S SH

SH

S S

S Hg But But

But But

But

But Hg But

4.4. Coordination Chemistry Aspects of Calixarenes with Pendant Sulphur Donors

N

N

(c)

(g)

Figure 9. (a)–(f) Transition and (g) lanthanide metal ion complexes of p-tert-butyl calix[4]arene, where dmf = N ,N -dimethylformamide; thf = tetrahydrofuran; R = t-Bu, H; S = CH3 CN.

OH OH O N

But

But

But

S

S OH

OH

But But

OH SH SH OH Figure 11. Two complexes.

thiocalix[4]arene

derivatives

and

their

mercury

Calixarenes

479

5. PHYSICAL PROPERTIES OF CALIXARENES A few physical properties, such as melting point, solubility, and pKa’s, are discussed in this section both for the organic derivatives as well as for their inorganic ones.

5.1. Melting Temperatures of Calixarenes Melting point is a characteristic property of many calixarenes. The parent calixarenes containing the free hydroxyl groups melt at high temperatures and increase progressively going from calix[4]arene to calix[6]arene to calix[8]arene and the melting points are generally in the range 340– 450  C. It has been noticed that even the nature of the substituent present in the p-position affects the melting point (Table 1, compounds 1, 4, 6). It is understood from the literature that [62] calix[6]arenes produced from p-nalkyl-phenols exhibit melting points as low as 110  C when the substituent is p-n-octadecyl. The melting point is also known to be affected by the position of the substituents on

the “upper rim” of the calix[4]arene (Table 1, compounds 7a–7c). The effect of substituents on the melting points has been observed in the case of isomeric compounds that differ only in the arrangement of identical groups around the “upper rim” of the calixarene. Thus Bohmer’s compounds [63] 7a, 7b, and 7c in Table 1 have melting points 185–190, 270, and 368  C respectively. Hence derivatization of the calixarenes greatly affects the melting point. Lower rim calixarene derivatives of esters and ethers generally melt at temperatures lower than the parent compound (Table 1). For example, 8 and 9 melt at 226–228 and 230–231  C, respectively (Table 1). However, there are exceptions where the trimethylsilyl ether of p-tert-butylcalix[4]arene at the lower rim melts at 411–412  C and the tetraacetate melts at 383–386  C. Among the calix[4]arene derivatives 12–17 (Table 1), the esters melt at lower temperatures than their carboxylic acid counterparts indicating a decrease in the hydrogen-bond interactions. Generally the decrease in the melting point upon derivatization at the lower rim is consistent with a decrease in hydrogen bonding interactions and a similar trend observed with the derivatives at the upper rim is consistent with an increase in the hydrophobic nature.

Table 1. Melting points and solubilities of calixarene derivatives. Codes 1 2 3 4 5 6

Compounds p-tert-butylcalix[4]arene p-tert-butylcalix[6]arene p-tert-butylcalix[8]arene p-phenylcalix[4]arene p-phenylcalix[8]arene p-n-octadecylcalix[4]arene

7

Melting point ( C)

Solubility

342–344 380–381 411–412 407–409 >450 110

less soluble in CHCl3

more soluble in CHCl3

R2

OH R1

OH

R3

HO OH

R4

R1

8 9 10 11 12 13 14 15 16 17 18 19 20

R2

R3

R4

a

Ph

t-Bu

CO2Et

Me

b

Ph

t-Bu

Me

CO2Et

c

Me

t-Bu

CO2Et

Ph

tetramethyl ether of p-tert-butylcalix[4]arene tetrabenzyll ether of p-tert-butylcalix[4]arene trimethylsilyl ether of p-tert-butylcalix[4]arene tetraacetate of p-tert-butylcalix[4]arene gly-ester of p-tert-butylcalix[4]arene gly-dev of p-tert-butylcalix[4]arene ala-ester of p-tert-butylcalix[4]arene ala-dev of p-tert-butylcalix[4]arene glu-ester of p-tert-butylcalix[4]arene glu-dev of p-tert-butylcalix[4]arene p-acetylcalix[4]arene calix[4]arene tetraacetate tetracarboxymethyl ether of p-tert-butylcalix[4]arene

185–190, 270 368 226–228 230–231 411–412 383–386 182 210 176–178 196–198 168–170 172

CHCl3 DMSO CHCl3 DMSO CHCl3 DMSO Hot bezene Insoluble in benzene Water

Calixarenes

480 5.2. Solubility Simple calixarenes are insoluble in water and aqueous base and are only slightly soluble in organic solvents; however, they are sufficiently soluble in chloroform, pyridine, or carbon disulfide [64]. It is this feature that makes some of the calixarenes difficult to isolate, purify, and characterize. Alkyl groups in the para-position of a calix[4]arene affect the solubility in organic solvent and the solubility increases as the chain length increases (Table 1, compounds 1 and 6) [65]. Lower rim derivatives of esters and ethers of calix[4]arenes show increased solubility and the difference in solubility may be used to separate the components from a mixture. For example, No et al. [66] separated the product, p-acetylcalix[4]arene, from its reactant, calix[4]arene tetraacetate, which is soluble in hot benzene. Water soluble calixarenes were developed due to their ability to interact with ions and molecules and also due to their catalytic properties. A first water soluble calixarene, tetracarboxymethyl ether of p-tert-butylcalix[4]arene, was synthesized by Ungaro et al. [67] whose solubility was found to be in the range 5 × 10−3 to 5 × 10−4 M depending upon the accompanying cation. p-Carboxy-calix[4–8]arenes were found to be soluble in 10−3 M aqueous base as reported by Gutsche et al. [33, 42, 43]. However, p-sulphonato calix[4– 8]arenes were much more soluble than the counter-carboxyderivatives and the former is soluble up to an extent of 0.1 M as reported by Shinkai et al. [36, 37]. However, the upper rim aminocalixarenes are moderately soluble in dilute aqueous acid [42, 43, 55].

5.3. pKa Due to their low solubility it was difficult to obtain the dissociation constants for the OH groups of the p-tertbutylcalixarenes. The pK1 values were found to be 6.0 and 4.3 for the mono-nitrocalixarenes (Figure 12a and b) in 1:1 water:methanol by measuring their ultraviolet absorption as a function of pH as reported by Bohmer et al. [68]. Thus these nitro-calixarenes are somewhat more acidic than p-nitrophenol. The pK difference of 1.7 units between the

two compounds that differ only in a single p-alkyl substituent is attributed to the conformational effects arising from the relative sizes of the methyl and tert-butyl groups.

6. SPECTRAL PROPERTIES OF CALIXARENES Spectral properties, such as vibrational, electronic, nuclear magnetic resonance (NMR), and mass, of both organic and inorganic derivatives of calixarenes are discussed in this section by taking several examples. Luminescence spectral studies are presented in Sections 8.5 and 8.6.

6.1. Vibrational Spectroscopy One of the most distinct features in the infrared spectra of the calixarenes is the observation of very low frequency for OH , viz., 3150 cm−1 in the case of calix[4]arene and 3300 cm−1 in case of calix[5]arene, which is attributed to the strength of intramolecular hydrogen bond interaction involving the –OH groups at the lower rim in these calixarenes [69, 70]. In the fingerprint region between 1500 and 900 cm−1 , the spectra look almost similar in all the calixarenes. However, in the 900–500 cm−1 region, the patterns vary to some extent depending upon the derivative. As shown in Scheme 1, the parent, p-tert-butylcalix [4]arene 1, shows a band at 3186 cm−1 for OH . Esterification of 1 resulted in 2 and in the spectrum of 2 the OH and C O appear at 3425 and 1757 cm−1 indicating the disruption of the circular hydrogen bond to some extent upon derivatization. The base hydrolysis of 2 resulted in the corresponding diacid 3, where OH and C O appear at 3424 and 1747 cm−1 respectively. In case of 3, Phe-OH overlaps with the carboxylic OH and thereby exhibits a broad band as compared to its precursor, 2. Upon treating with SOCl2 , the acid is converted to acid-chloride, 4, where both peaks have shifted to

O

(i)

O

O

(ii)

R R

R

But

R=

R R

1

R

R

R

R

OH

HO

Me

OH

HO

N H

t-Bu Me

O

O

O

(v)

O

(iii)

N H

O

O

O

NH

(vi)

O

R

OH

R

R

R2

4

NO2 Figure 12. (i) Mono-nitro calix[4]arenes: R1 = Me, 1a; R1 = t-Bu, 1b. (ii) Tetraesters and ethers of p-tert-butylcalix[4]arene: R2 = OCOCH3 , 2a; R2 = OCO2 Et, 2b; R2 = OCH2 CO2 Et, 2c; R2 = OCH2 CO2 But , 2d; R2 = OCH2 CONEt2 , 2e; R2 = OCH2 COCH3 , 2f.

R

R

6

R

R

R R

R

4

O

OH OH O

R R

CH2

R

NH O

O OHOHO

O OHOHO

N

O

O OHOHO

(iv)

N

Cl

Cl

' R' NH

R''

R'' NH

R

OR'

R'O

O

O

OH

R

R

3

2

R''

O

O OHOHO

O OHOHO

R1

OH

HO

OEt

EtO

OHOHOH OH

R

R

R

7

5

Scheme 1. Stepwise syntheses of p-tert-butylcalix[4]arene derivatives. (i) BrCH2 CO2 Et, K2 CO3 , acetone, reflux; (ii) 15% aq. NaOH, EtOH, reflux; (iii) SOCl2 , benzene, reflux; (iv) amino acid ester.HCl, Et3 N, THF, RT; (v) LiOH, THF/H2 O; (vi) 2-(2-amino methylpyridine), Et3 N, RT. R′ = CH3 , R′′ = Gly, 5a; R′ = CH3  R′′ = Ala, 5b; R′ = CH2 Ph, R′′ = Asp, 5c; R′ = CH2 CH3  R′′ = Glu, 5d; R′ = Gly, 6a; R′ = Ala, 6b; R′′ = Asp, 6c; R′′ = Glu, 6d.

Calixarenes

481

higher frequencies (Fig. 13). The reaction of acid-chloride, 4, with the amino acid esters [71] yields the corresponding amides (5a–5d) that exhibit characteristic ester and amide bands in the ranges 1741–1754, 1679–1685 (amide I), and 1531–1539 cm−1 (amide II, NH ). Thus the data are indicative of the formation of the pendants with peptide linkages. Conversion of these by base hydrolysis to the corresponding carboxylic acids (6a–6d) resulted in a frequency shift of amide bands to lower values shown in Figure 14.

6.2. Ultraviolet Spectroscopy The linear and cyclic oligomers show two absorption bands at 280 and 288 nm in the ultraviolet spectra. The ratio of the absorptions at these two wavelengths is a function of the size of the ring. Conforth and co-workers [2] carried out the lower rim alkylation of calixarenes by monitoring the disappearance of the absorption at 300 nm (characteristic of the free phenol) and the growth of absorption at 270–280 nm arising from the ether.

Figure 14. FTIR spectra corresponding to the structures shown under each spectrum. (A), (B), (C), and (D) respectively correspond to the compounds 5c, 6c, 5d, and 6d which were given in Scheme 1.

6.3. Nuclear Magnetic Resonance Spectroscopy

Figure 13. Fourier transform infrared (FTIR) spectra corresponding to the structures showed under each spectrum. (A), (B), (C), (D), and (E) respectively correspond to the compounds 1, 2, 3, 4, and 7 which were given in Scheme 1.

In the 1 H NMR spectrum of p-tert-butylcalix[4]arene at room temperature, resonances from the aromatic protons, the tert-butyl protons, and the hydroxyl protons are observed as singlets, and those from the CH2 protons are observed as a pair of doublets. However, these doublets collapse to a singlet when the temperature is raised from 20 to 60  C in chloroform [72]. On the basis of nuclear Overhauser effect experiments, Ungaro et al. [67] assigned the higher field doublet to the equatorial protons (closer to the aromatic rings) and the lower field doublet to the axial protons (closer to the hydroxyl groups). Gutsche et al. [73, 74] have shown that calix[8]arenes behave in a similar manner but calix[6]arenes show only a singlet resonance even at room temperature and split at lower temperature. Calix[4]arene and calix[8]arene can be differentiated [75] by the fact that in pyridine the former retains the pair of doublets at 0  C whereas the latter shows only a singlet at temperatures as low as −90  C. Rao et al. [16] reported that the 1 H NMR spectra of 1 (Scheme 1) shows a single tert-butyl signal at 1.26 and an AB quartet for bridged methylene group at 3.50 and 4.30 ppm and a singlet for aromatic protons at 7.05 ppm. The phenolic proton appeared as a singlet at 10.30 ppm that disappears upon D2 O exchange. The 1 H NMR spectra of

Calixarenes

482 1,3-di-derivatives (2, 3, and 4) exhibited doubling of peaks for tert-butyl groups (0.93–1.05 and 1.26–1.27 ppm) and aromatic protons (6.78–6.93 and 7.06–7.08 ppm), one each corresponding to the substituted part and the unsubstituted one. Spectra of these also exhibited an AB quartet for the bridged methylene group in the ranges 3.32–3.41 and 4.13– 4.45 ppm. A singlet for the methylene group arising from O–CH2 –CO was observed in the range 4.66–5.03 ppm. The spectrum of 2 shows signals corresponding to the ethyl ester group which is absent in 3 and 4. Thus the data suggest that the substitution in the di-derivatives is of 1,3-alternate type and the calix[4]arene exists in the cone conformation. The splitting pattern observed for the bridged –CH2 group in 1 H NMR spectra provides a signature for the type of conformation present. In case of the 1,3-di-derivatives built using peptide linkages, 5 and 6, the protons and/or carbon (in 13 C NMR) as well as that of calixarene resonances were assigned by correlation spectroscopy (COSY) and hetero multiple quantum coherence experiments. The 1 H NMR data corresponding to these derivatives are given in Table 2. In the 1 H NMR spectra, the amide N–H proton signals appeared in the range 9.27 to 9.63 (CDCl3 ) and 8.97 to 9.20 ppm (DMSOd6 ) for 5a–5d and 8.72 to 9.14 ppm for 6a–6d (DMSO-d6 ). These chemical shifts were comparable with those observed in case of the tetra-substituted amino acid ester derivatives [76]. The positions of amide proton signals were confirmed through the cross-coupling peaks of C H in COSY experiments. While the amide –NH protons were not exchanged upon addition of D2 O in the case of 5, these were found to be exchanged in the case of 6 due to the changes in the confirmation of the pendant groups. From these results it can be concluded that there exists strong intramolecular hydrogen bonding in 5 in CDCl3 . However, 5 when measured in DMSO-d6 showed upfield shifts of amide proton signals indicating the disruption of intramolecular hydrogen bond interactions present in the molecule. However, the phenolic –OH proton signals disappear upon D2 O exchange. The aromatic protons (Ar–H) in 5a (gly- derivative) appeared as two singlets whereas 5b–5d (ala-, asp-, and glu- derivatives) showed two doublets due to the presence of pendant chiral amino acid units. For the same reason, the axial and equatorial calixarene bridged CH2 protons showed two pairs of doublets in 5b–5d. This chiral effect is also seen in the signals of OCH2 CO protons as two doublets. In case of 6a–6d, a similar splitting pattern was observed. All the spectra were

indicative of the presence of cone conformations of these 1,3-di-derivatives.

6.4. Mass Spectrometry Rao and co-wokers [16] have reported mass spectral data for the calixarene 1,3-diderivatives of amino acid and their esters formed through peptide linkage. The molecular weights were confirmed by the presence of molecular ion peaks in the mass spectra obtained by the FAB mass method or the EI method.

7. STRUCTURES BY X-RAY CRYSTALLOGRAPHY Single crystal X-ray diffraction provides a powerful means of determining the molecular and crystal structures. The lower rim derivatives of the calix[4]arenes can exist in any one or more of the possible four conformations as given in Section 2. Undoubtedly, X-ray crystallography is the most confirmative way to ascertain the conformations exhibited by the calixarenes and their derivatives. In several cases the solid state structural studies were further augmented by NMR conformational analysis in solution.

7.1. Crystal Structures of Calixarenes and Their Derivatives The earliest example was reported by Rizzoli et al. [77], wherein the tetraacetate of p-tert-butylcalix[4]arene was shown to exist in partial-cone conformation as shown in Figure 15. Ester and ether derivatives, such as 1b–1f (Fig. 12), exist in cone conformation [14, 29, 78, 79] for which X-ray structures have been obtained. The synthesis of 25, 27-dimethoxy-26, 28-dimethylester-p-tertbutylcalix[4]arene was reported and its structure was established wherein it is observed that the calixarene asssumes a partial-cone conformation [80]. The corresponding potassium salt of this ester exists in 1,3-alternate conformation whereas in the sodium salt the calixarene is in a distorted cone or flattened partial-cone conformation (Fig. 16). The crystal structure of a tetra(benzyl)amide of calix[4]arene exhibits a distorted cone conformation [81]. Syntheses of lower rim substituted aryl ethers of p-tert-butylcalix[4]arene were reported [82] where the crystal structure of one of the 1,2-disubstituted compounds exists in a partial-cone conformation. One of the aryl ether pendants is sandwiched

Table 2. 1 H NMR data of 1,3-diderivatives of p-tert-butylcalix[4]arene with aminoacids and their corresponding esters. Compounda

NH

OH

Ar-H

C H

OCH2 CO

ArCH2 Ar

5a 5b 5c 5d 6a 6b 6c 6d

9.27 9.42 9.63 9.37 8.89 9.09 8.88 8.87

7.83 7.85 7.92 7.78 8.41 8.30 8.20 8.17

7.25, 6.92 7.08, 6.93

4.14 4.74–4.69 5.03 4.65 4.03 4.41–4.36 4.76 4.40–4.33

4.63 4.53, 4.69–4.74 4.77, 4.37 4.85, 4.31 4.53 4.77, 4.31 4.61 4.75, 4.40–4.33

4.23, 3.43 4.31, 3.40, 4.18, 3.46 4.29, 4.14, 3.42, 3.24 4.24, 4.13, 3.40, 3.26 4.23, 3.47 4.41–4.36, 4.15, 3.54, 3.45 4.26, 3.43 4.40–4.33, 4.21, 3.51, 3.44

a

6.99, 6.83 7.17 7.22–7.17 7.14 7.20–7.12

Refer to Scheme 1 for compound names.

C H

C H

1.47 2.97 2.29

3.97

1.40 2.81–2.78 2.08–1.99

2.36–2.32

C(CH3 3 1.27, 1.28, 1.24, 1.19, 1.20, 1.20, 1.20, 1.20,

1.03 0.97 1.04 0.96 1.13 1.12 1.17 1.11

Calixarenes

483

Figure 15. Lower rim tetraacetate derivative of p-tert-butylcalix[4]arene. Open circles denote carbons and filled circles denote oxygens. Adapted from [77], C. Rizzoli et al., J. Mol. Struct. 82, 133 (1982). © 1982, Elsevier Science.

between two neighboring O-unsubstituted phenyl rings within the calixarene cavity. Synthesis and structure determinations of thiacalix[4]arenes bearing four keto-groups and four amido-groups on the lower rim have been reported. All three derivatives have been found to adopt 1,3-alternate conformation [83]. A calix tube is formed when two thiabridged calix[4]arenes were connected at their lower rim phenolic oxygens through four ethylene linkers [84]. This is shown to adopt a flexible C2v flattened cone arrangement as shown in Figure 17. Zinic et al. [85] reported the synthesis of chiral calix[4]arene derivatives with four O-(N -acetylphenylgly-OMe) or O-(N -acetyl-Leu-OMe) strands. X-ray structure analysis of the glycine derivative revealed distorted cone conformation with C2 symmetry and the structure is organized into infinite chains by intra- and intermolecular hydrogen bonds. The solid and solution structures of the corresponding sodium complex were identical with cone conformation having C4 symmetry. A number of lower rim O-alkylated derivatives of mono-, di-, tri-, and tetra-substituted picolyl ones with or without p-tert-butyl groups were synthesized and their conformations

Figure 16. Crystal structures of (a) 25,27-dimethoxy-26,28-dimethylester-p-tert-butylcalix[4]arene (partial-cone conformation); (b) its potassium ion complex (1,3-alternate conformation), and (c) its sodium ion complex (flattened partial-cone conformation). Open circles denote carbons, filled circles denote oxygens, and circles with a line denote nitrogens. Sodium and potassium ion centers are labeled in the figure. Adapted from [80], I. Oueslati et al., Tetrahedron Lett. 41, 8439 (2001). © 2001, Elsevier Science.

Figure 17. Crystal structure of thiacalix[4]arene tube. Big and small open circles denote carbons and hydrogens respectively; filled circles denote oxygens and circles with lines denote sulphurs. Adapted from [84], S. E. Matthews et al., New J. Chem. 25, 1355 (2001). © 2001, Royal Society of Chemistry.

were established [21]. It has been found that the type of base used during the synthesis [20 equivalents of the picolylchloride (HCl), NaH (70–80% cone), K2 CO3 (36% of partial cone, 55% of 1,3-alternate, and 9% of cone), or Cs2 CO3 (54% of partial cone, 18% of 1,3-alternate, and 9% of cone)] influences the conformation of the tetra-substituted picolyl derivative of calix[4]arene obtained. Such influence of the base used in the reaction on the conformation of the tetraalkylated calix[4]arenes was also reported by Reinhoudt and co-workers [86]. This group also reported that the tetraethyl ether of p-tert-butylcalix[4]arene exists in a fixed 1,2-alternate conformation [87] for which the X-ray structure was determined as shown in Figure 18. The reaction of p-tert-butyl calix[4]arene with a flexible and reactive sebacoyl chloride yielded the singly intrabridged calix[4]arene derivative [88]. The crystal structure of this compound showed a distorted cone conformation (Fig. 19). Selective functionalization of calix[4]arenes at the

Figure 18. Molecular structures of (a) tetraethyl ether of p-tertbutylcalix[4]arene in a fixed 1,2-alternate conformation; (b) 1,4dimethyl-2,3-diethyl ether of p-tert-butylcalix[4]arene in partial-cone conformation. Open circles denote carbons and filled circles denote oxygens. Adapted from [87], L. C. Groenen et al., J. Am. Chem. Soc. 113, 2385 (1991). © 1991, American Chemical Society.

484

Figure 19. Singly intrabridged structure (using sebacoyl chloride) connecting 1,3-phenolic moieties in p-tert-butylcalix[4]arene in a distorted cone conformation. Open circles denote carbons and filled circles denote oxygens. Adapted from [88], J.-D. Van Loon et al., J. Org. Chem. 55, 5176 (1990). © 1990, American Chemical Society.

upper rim led to the formation of 26,28-dimethoxy-11,23dinitrocalix[4]arene, the crystal structure of which (shown in Fig. 20) reveals that the macrocycle is in a flattened cone conformation [89]. A series of bridged calix[4]arenes with two opposite phenolic units connected by an aliphatic chain were synthesized [90]. The length of the chain was varied from 5 and 16 carbon atoms. 1 H NMR spectra support that the cone conformation is fixed in these compounds. X-ray analysis also revealed a distortion of the ideal cone conformation with shorter connecting chains, which is attributed to the weaker intramolecular hydrogen bonds.

7.2. Structures of the Metal Ion Complexes of Calixarenes and Its Derivatives The treatment of calixarenes and its derivatives with inorganic salts and/or compounds results in the formation of the complexes, whose three-dimensional structures can be established by X-ray crystallography. Power and co-workers established the first crystal structure of a calix[4]arene– transition metal ion complex by the reaction of p-tert-

Figure 20. Structure of 26,28-dimethoxy (lower rim)-11,23-dinitro (upper rim) calix[4]arene in a flattened cone conformation. Open circles denote carbons, filled circles denote oxygens, and circles with a line denote nitrogens. Adapted from [89], J.-D. Van Loon et al., Tetrahedron Lett. 30, 2681 (1989). © 1989, Elsevier Science.

Calixarenes

butylcalix[4]arene with Ti(NMe2 )4 [44]. The structure shows two titanium ions sandwiched between two calixarene units and the calixarene exists in a partial-cone conformation with the titanium centers in tetrahedral coordination geometry. Similarly, iron and cobalt complexes of calix[4]arene were synthesized and their structures were determined by X-ray crystallography. The iron assumes a distorted trigonal pyramidal geometry and that of cobalt is irregular (Fig. 9). Andreeti and co-workers [91] treated p-tert-butylcalix[6] arene with Ti(OPr)4 and the resultant product exhibited a structure consisting of two calixarene units in a cone conformation coordinated with two titanium centers as a Ti–O–Ti unit, with distorted trigonal bipyramidal geometry around each titanium center. Harrowfield et al. [92] have synthesized an europium complex of p-tert-butylcalix[8]arene whose crystal structure was found to contain two europium atoms in a pinched conformer of the cyclic octamer. Ripmeester and co-workers have reported [93] that xenon is encapsulated in the p-tertbutylcalix[4]arene cavity and behaves both as a guest and as a cavity probe. 129 Xe NMR and X-ray diffraction data show that the Xe occupies two cavity sites (Fig. 21). Oxomolybdenum (VI) binds four oxygen atoms from a calix[4]arene with cone conformation, and in the lattice this unit is co-crystallized with a free calix[4]arene by trapping a nitrobenzene between these two units [47]. Reaction of mercury(II)acetate with p-tert-butyl-1,3-dihydroxy2,4-disulfanylcalix[4]arene yielded a mononuclear complex in which the Hg(II) ion is coordinated linearly to two arenethiolate groups [61] and the calixarene adopts a 1,3-alternate conformation (Fig. 22). Calix[4]arene–W(VI)– (OAr)2 complex is exhibited in the lattice self-assembly of these, resulting in a columnar structure as reported by Floriani’s group [94]. Reaction of a 10-fold excess of NaH and LiH with p-tert-butylcalix[4]arene in THF results in the formation of the corresponding salts of mono-oxyanion and di-oxyanion calixarenes respectively [95], where the additional metal coordination sites are filled by MeOH and H2 O to result in tetra-coordination around Li+ and penta-coordination around Na+ as shown in Figure 23a and b. Reaction of the hexamethyl ether of p-tert-butylcalix[6]arene with TiCl4 in toluene [96] resulted in crystals of [(calix[6]arene)– (Cl3 TiOTiCl2 )2 ] (Fig. 24).

Figure 21. Clathration of Xe (labeled in the figure) into the p-tertbutylcalix[4]arene cavity as a guest. Open circles denote carbons and filled circles denote oxygens. Adapted from [93], E. B. Brouwer et al., J. Chem. Soc., Chem. Commun. 939 (1997). © 1997, Royal Society of Chemistry.

Calixarenes

Figure 22. (a) Cone conformation of p-tert-butyl-1,3-dihydroxy-2, 4-disulfanylcalix[4]-arene (top view) and (b) its Hg complex (linear) in 1,3-alternate conformation (lateral view). Open circles denote carbons, circles with lines represent sulphurs; filled circles denote oxygens in (a); circles with dots represent oxygens in (b); and filled circles in (b) denote Hg. Adapted from [61], X. Delagiue et al., J. Chem. Soc., Chem. Commun. 609 (1995). © 1995, Royal Society of Chemistry.

Reactions of aluminium alkyls with a number of calixarene methyl ethers have been established [97, 98]. The reaction of tetramethyl ether of p-tert-butylcalix[4]arene with Me3 Al in benzene–toluene solution resulted in crystals, the structure of which shows that the calixarene is in 1,2alternate conformation with two molecules of Me3 Al per calixarene. On the other hand the reaction with MeAlCl2 or EtAlCl2 resulted in a complex having the same composition but the calix[4]arene takes up an 1,3-alternate rather than the 1,2- conformation observed in the previous case. The 1,3-bis(trimethylsilyl)ether of p-tert-butylcalix[4]arene has been synthesized and used as a dianionic ligand for complexing Ge and Sn [99]. The crystal structure of the Ge complex exhibits exo/endo isomerism (Fig. 25a and b). Transition metals can be used for shaping and electronically enriching the calix[4]arene cavity so that the -basic cavities can be adapted to alkali–metal cation complexation in the hydrophobic zone [100]. Thus the structure of the calixarene–alkali metal cation was determined by X-ray diffraction (XRD) wherein the calixarene assumes

Figure 23. Structures of (a) [Na(MeOH)(H2 O)2 ][p-tert-butylcalix[4] arene]·MeOH and (b) [Li(MeOH)2 (-H2 O)][p-tert-butylcalix[4]arene]. MeOH. Open circles denote carbons, filled circles denote oxygens, and circles with crosses denote metal ions. Adapted from [95], F. Hamada et al., Supramol. Chem. 2, 19 (1993). © 1993, Taylor & Francis, Ltd.

485

Figure 24. Structure of calix[6]arene[TiCl2 (-O)TiCl3 ]2 . Adapted from [96], S. G. Bott et al., J. Chem. Soc., Chem. Commun. 610 (1986). © 1986, Royal Society of Chemistry.

a flattened cone conformation. Complexes of Fe(III) and Er(III) of a calix[4]arene diamide were reported [101] to form 1:1 complexes where the calix[4]arene is found to be in cone conformation. Two extended-array metallocalixarenes derived from the ring opening of imido-molybdenum precursors have been synthesized and their structures were determined [102]. It is observed that the metallocalixarenes are organized in a “cup-to-cup” arrangement that gives rise to a calixarene “socket” and this potentially hosts guest molecules. A calix[4]arene derivative, 1-carboxylic acid-3-diethyl amide, exhibited efficient extraction properties toward rareearth ions and the crystal structure of a neutral 1:1 uranyl complex of dinuclear nature was determined [103]. Syntheses and X-ray crystal structures of neutral dimeric europium (Fig. 26), samarium, and monomeric lutetium complexes and the selective extraction properties of the 1,3-acid-diethyl amide substituted calix[4]arene ligand have been determined [104]. While the monomeric lutetium complex is seven-coordinated (Fig. 27), both the dimeric complexes exhibit eight-coordination around each metal ion. The structures show that the substituted calix[4]arene exists in cone conformation.

Figure 25. Molecular structures of [p-tert-butylcalix(trimethylsilyl)2 ]Ge: (a) exo-form and (b) endo-form. Open circles denote carbons, filled circles denote oxygens, and circles with lines denote Ge. The big open circles represent Si. Adapted from [99], T. Hascall et al., J. Chem. Soc., Chem. Commun. 101 (1998). © 1998, Royal Society of Chemistry.

486

Figure 26. Crystal structure of [Eu(1,3-diacid-diethylamide substituted calix[4]arene)(EtOH)]2 . Open circles denote carbons, filled circles denote oxygens, and circles with a line denote nitrogens. Adapted from [104], P. D. Beer et al., J. Chem. Soc., Chem. Commun. 1117 (1996). © 1996, Royal Society of Chemistry.

Complexes of La(III), Eu(III), Gd(III), Tb(III), and Lu(III) with the lower rim functionalized tetrakis[2(diethylcarbamoyl-methoxy)ethoxy]calix[4]arene have been synthesized and the crystal structure of the Lu complex was established where Lu(III) is 9-coordinated in which the calixarene is in the cone conformation [105]. Calixarene derivatives of phosphine oxide moieties [calix(OCH2 CH2 POPh2 )n , where n = 4, 6, 8] attached to the lower rim have been reported to exist in distorted cone conformation [106]. These act as receptors for the extraction of Eu(III), Th(IV), Pu(IV), and Am(IV) in nuclear waste treatment. The reaction of 5,11,17,23-tetra-tert-butyl-25,27-dihydroxy-26,28-bis[{N -(2-diphenyl phosphino)phenyl} carboxyamidemethoxy]calix[4]arene with [Cu(CH3 CN)4 ]ClO4 and Pt(COD)Cl2 formed complexes whose structures were determined by X-ray analysis. The copper complex has two seven-membered metallocycles, the platinum complex

Figure 27. Crystal structure of [Lu(1,3-diacid-diethylamide substituted calix[4]arene)(H2 O)] monomer. Open circles denote carbons, filled circles denote oxygens, and circles with a line denote nitrogens. Adapted from [104], P. D. Beer et al., J. Chem. Soc., Chem. Commun. 1117 (1996). © 1996, Royal Society of Chemistry.

Calixarenes

has one five-membered metallocycle, and both complexes assume the cone conformation [107]. X-ray studies reveal that the dimeric supramolecular assemblies of water-soluble calix[4]arenes are formed by a combination of hydrophobic effects and either hydrogen bonding or secondary bonding [108]. It is observed that for copper complex, the dimeric association is based on a secondary bonding interaction between the copper(II) ion and a sulfonato oxygen atom combined with hydrophobic interactions. The coordination form and hence the cavity of the calix[4]arene ligands can be controlled by other ligands in transition-metal calix[4]arene complexes, if strong directing co-ligands such as oxo groups are used. Radius [109] described the synthesis and structure of Ti(IV) and Mo(VI) complexes of 1,3-dimethoxy-calix[4]arene where TiCl2 and MoO2 moieties arising from TiCl4 and MoO2 Cl2 control the coordination form and hence the cavity. While the ligand adopts an elliptically distorted cone conformation in TiCl2 complex as shown in Figure 28, the same adopts a partialcone conformation in the molybdenum complex (Fig. 29). Transition-metal complexes of 5,11,17,23-tetra-tert-butyl25,26,27,28-tetrakis(diethyl-carbamoylmethoxy)calix[4]arene in the cone conformation have been synthesized and the crystal structures of Fe(II), Ni(II), Cu(II), Zn(II), and Pb(II) complexes have been established by Ogden and co-workers [110]. The structures of the complexes of Fe(II), Zn(II), and Pb(II) are almost similar with the metal ions coordinated to all eight oxygen atoms of the ligand as can be seen in Figure 30. The copper complex is somewhat different with four bonds slightly shorter than the remaining four. On the other hand, the nickel structure is completely different where the ligand has undergone rearrangement to accommodate the metal cation in a distorted octahedral environment (Fig. 31). The same group also reported Fe(III) complexes of 5,11,17,23-tetra-tert-butyl-25-hydroxy, 26,27,28 tris(diethylcarbamoylmethoxy)-calix[4]-arene whose structure was determined by XRD [111]. The structure shows that Fe(III) is bound to all seven oxygen atoms of calix[4]arene with Fe–Ophenolic distance being the shortest (Fig. 32).

Figure 28. Molecular structure of [p-tert-butylcalix[4]arene(OMe)2 O2 TiCl2 ] in elliptical distorted cone conformation. Open circles denote carbons, filled circles denote oxygens, and circles with lines denote chlorines. Adapted from [109], U. Radius, Inorg. Chem. 40, 6637 (2001). © 2001, American Chemical Society.

Calixarenes

487

Figure 29. Molecular structure of [p-tert-butylcalix[4]arene(OMe)2 O2 MoO2 ] in partial conelike mode. Open circles denote carbons and filled circles denote oxygens. Adapted from [109], U. Radius, Inorg. Chem. 40, 6637 (2001). © 2001, American Chemical Society.

8. APPLICATIONS Parent calixarenes are insoluble in water, have generally low solubility in most organic solvents, and are characterized by their high melting points. However, through appropriate derivatization these exhibit good solubility in either medium [112]. Aqueous solutions (0.1 M) can be prepared with p-sulphonato-calix[4]arene. The acylated calixarenes possess antioxidant and heat stabilizing properties that are useful for the plastics. Thus simple calixarenes, their functionalized

Figure 31. Structure of the nickel(II) complex of p-tert-butylcalix[4]arene tetramide with included acetonitrile molecule. Open circles denote carbons, filled circles denote oxygens, and circles with a line denote nitrogens. Adapted from [110], P. D. Beer et al., J. Chem. Soc., Dalton Trans. 1273 (1995). © 1995, Royal Society of Chemistry.

derivatives, and the metal ion complexes of all these offer a number of applications having impacts in chemistry, biology, and materials. Details of some such applications are presented in this section.

Figure 30. Structure of lead(II) complex of p-tert-butylcalix[4]arene tetraamide together with the included MeCN. Open circles denote carbons, filled circles denote oxygens, and circles with a line denote nitrogens. Adapted from [110], P. D. Beer et al., J. Chem. Soc., Dalton Trans. 1273 (1995). © 1995, Royal Society of Chemistry.

Figure 32. Molecular structure of [Fe(p-tert-butylcalix[4]arenetriamideH)](FeCl4 )2 ]. Open circles denote carbons, filled circles denote oxygens, and circles with a line denote nitrogens. Adapted from [111], M. I. Ogden et al., J. Chem. Soc., Dalton Trans. 3073 (2001). © 2001, Royal Society of Chemistry.

Calixarenes

488 8.1. Supramolecular Formation Undoubtedly the most useful property of calixarenes and its derivatives is their ability to function as molecular baskets and bind neutral and/or ionic guests resulting in supramolecular arrays. Rapid and extensive developments have taken place through exploration of the receptor property of the simple calix[4]arene and its functionally modified derivatives, thereby contributing to the area of analytical devices. Inclusion complexes of neutral guest molecules, such as toluene, benzene, xylene, anisole, and pyridine, in the chalicelike molecular cavity of calix[4]arene have been reported [113]. Attractive CH3 · · ·  interactions may contribute to the binding. Based on the X-ray diffraction studies, Atwood [114] and co-workers demonstrated a water inclusion complex of an upper rim sulphonated calix[4]arene (Fig. 33). The water molecule occupies a distorted conical cavity with its two hydrogen atoms directed toward the two nearest juxtaposed benzene rings. Structures of this type may provide important clues about how water molecules interact with aromatic moieties in biological systems. A 1:1 clatharate complex of calix[4]arene–tetracarbonate and acetonitrile is reported to have C4v symmetry. In this clatharate, the acetonitrile resides in the hydrophobic cavity with nitrogen atom protruding from the lower rim.

8.2. Calixarenes as Receptors Calixarenes can function as ion receptors by selective transport of alkali cations from one aqueous phase to another through a chloroform membrane containing the p-tert-butyltetramer, -hexamer, or -octamer [115]. When the source phase was neutral, practically no transport was observed, but when the source phase was made basic to allow ionization of the hydroxyl groups ion transport takes place, the octamer showing the fastest rate with cation selectivity order Cs+ > Rb+ ≫ K+ > Na+ > Li+ . The presumption that neutral alkali phenoxides contain the cation in or near the cavity has received support from a recent X-ray analysis of neutral cesium salt of the tetramer [116]. Here the cation is found in the center of the cavity. The four phenolic C–O bonds are completely equivalent, indicating a rapidly exchanging network of hydrogen bonds around the four oxygen atoms. It is reported that the calix[4]arene was selective for sodium as compared to the other group I and II metal ions [117]. The studies also showed that there was a size selective effect manifested by a peak extraction of

cesium by the ester derivative of calix[6]arene. Thus, among the alkali cations, calix[4]arene derivatives have the highest affinity for Na+ . Some advantages of calixarene as receptors are evident from the complexation of UO2+ 2 species because the uranyl ion requires hexagonal coordination unlike many competing heavy metal ions that have an octahedral coordination sphere. Shinkai and co-workers [118] have found stabilwith sulphonic and carboxylic acid ity constants for UO2+ 2 derivatives of calix[6]arene, up to 1019 mol−1 and selectivity factors of 1011 –1019 when compared to the divalent cations, viz., Mg2+ , Zn2+ , Ni2+ . Though most of the complexation studies with calixarenes deal with alkali and alkali earth cations, the derivatives containing soft donor atoms like N and S act as effective receptors for a range of transitionmetal ions. Therefore, changing the nature of the substituents/pendants at the lower rim can produce dramatic changes in the selectivity. For instance, if the ester moiety at the lower rim of p-tert-butyl calix[4]arene is substituted by soft donors, like thio and thioamide groups, the selectivity is shifted from Na+ to Ag+ . Similarly, phosphine derivatives have been synthesized to study their complexation with lanthanide and actinide ions for the selective extraction of radioactive elements from nuclear waste.

8.3. Calixarenes as Ion-Selective Electrodes Calixarenes containing cation complexing groups at the lower rim possess the molecular requirements for the type of ionophores used in ion-selective electrodes. This is so, because calixarenes possess rigid cavities that are predisposed to selectively complex with cations and the high lipophilicity associated with the t-butyl groups prevents the organic derivative and its cation-complexed product from leaching nonpolar membrane into aqueous solution. Ion selective electrodes for K+ and Cs+ have been successfully developed using dioxacalix[4]arene [119] and calix[6]arene [120] respectively. Calix[4]arene derivatives containing S and N donor groups have been developed as ion selective electrodes as well as used to make chemically modified electrodes suitable to determine Hg2+ , Pb2+ , Ag+ , and Cu2+ [121]. Ion selective field effect transistors containing a calixspherand as the ionophore enable the quantitative determination of K+ in the presence of even a large excess of Na+ [122].

8.4. Calixarenes as Biosensors A tetra-ester derivative of calix[4]arene, 1 (Fig. 34), has been used to make bench and miniaturized Na+ selective electrodes [123] and for Na+ detection in flow injection analysis [124]. These devices have been used to determine Na+ in human blood plasma [125]. This sensor has also been used along with several other sensors in an array giving accurate assays of ions in blood [126] and mineral water [127].

8.5. Calixarenes as Optical Sensors Figure 33. Inclusion complex of a calix[4]arene sulphonate with water as guest. Adapted from [51], J. M. Harrowfield et al., J. Chem. Soc., Chem. Commun. 1159 (1991). © 1991, Royal Society of Chemistry.

The signal in potentiometric sensors is generated by selective exchange of cations into the membrane phase, which is mediated by a calixarene ligand. However, for optical

Calixarenes

489 R1

R3

CH2

CH2 n

OR2

OR4

m

O

O

n

O

O O 1 m=0, n=4, R1=R3= tBu, R2 =R4 =

OMe

CH2

OH O

O

OH

O 2 m=0, n=4, R1=R3= tBu, R2 =R4 =

CH2

OC2H5

3 m=0, n=4, R1=R3=tBu, R2=R4= O CH2

N

N

O

O

OH

O N

N

NO2

Figure 34. Schematic representation of the structure of calix[4]arene derivatives used as optical sensors (suitable for absorption studies) in their “cone” conformation.

transduction, an optically responsive mechanism must be coupled with the ion–ligand complexation process. This can be achieved by co-immobilizing a lipophilic, acido-chromic dye in the membrane with the ligand. The complexation of the metal ion by the ligand results in the expulsion of a proton from the dye to the sample phase, which maintains overall charge neutrality [128]. This approach has been used to design a series of ion-selective optical sensors (optodes), based on the same type of calixarene ligands (Fig. 34) that is used for ion-selective electrodes. The dyes, such as acridine derivatives, can produce optodes with longer lifetimes [129]. It is also possible to attach acidochromic groups directly on the upper or lower rim of the calixarene [130]. A calix[4]arene derivative [131] containing the nitrophenol-azaphenol chromophore, 3 (Fig. 34), in chloroform changes color upon addition of Li+ in the presence of tri-n-dodecylamine (TDDA). On complexation of Li+ , the labile proton of the dye is transferred to TDDA, and the color changes from a pale yellow to deep brownish red by shifting the absorption maximum from ∼380 to ∼530 nm. A mixture of tetraester of calix[4]arene, 2 (Fig. 34), Girard’s reagent P derivative of butyraldehyde, and chromoionophore ETH5294 has been used to detect low molecular weight aldehydes. 1,1′ -Binaphthyl (at the lower rim) as well as the bis(indolphenol) (at the upper rim)derived calix[4]arene crown ether derivative (Fig. 35) was developed as a sensor to differentiate the entiomers of host amines and amino acids through color change. While one of the enantiomers respond through change in visual coloration, the other does not when these are interacted with this calix[4]arene derivative [132].

Figure 35. Schematic structure of 1,1′ -binaphthyl (on lower rim) as well as bis(indophenol) (on upper rim)-derived calix[4]arene as molecular sensor for chiral recognition (n = 0 1).

Some of these chromogenic calixarene derivatives reported are novel and have important analytical applications, such as light-switching properties, where the derivative possesses a fluorescent pyrene group in the 1-position and a p-nitrobenzyl quencher in the 3-position as shown in Figure 36 [133]. Fluorescence is absent in the cation-free form of the calixarene, but the presence of a cation guest in the cavity forces the lower rim into a more open conformation, with quencher and fluorophore spatially separated, thus allowing fluorescence to occur. Thus, these receptors are of use in fluorescence-based ion sensors. Also, calixarenes functionalized on their lower rim to contain luminescent ruthenium(II) trisbipyridal complex [134] are found to act as potential amine sensors.

H

H H

H

O O O O YY O O O

O

O 2N Y= CH2COOC2H5 Figure 36. Schematic representation of the structure of calix[4]arene derivative possessing pyrene group on the lower rim as sensor that is suitable for fluorescence studies.

Calixarenes

490 8.6. Electrochemiluminescence of Calixarenes Upper rim carboxylic acid derivatives of calixarenes shown in Figure 37, when coated on a porous silicon electrode surface, exhibited interesting luminescence properties [135, 136] in 1.0 M H2 SO4 electrolyte solution through anodic oxidation of the calixarene derivative. It was found that both the duration as well as efficiency of the light emission are affected depending upon the nature of the calixarene derivative used. Electrochemiluminescence can be generated for longer periods of time for these coated porous Si junctions relative to the untreated porous Si ranging from up to 3 h for calix[4]arene-coated porous Si to about 30 min for the calix[n]arene (n = 5, 7, 8) derivatives. These observations are attributed to packing differences between different calixarene carboxylic acid derivatives as thin films and subsequent bias-induced Si nanoparticle oxidation in the porous matrix. The studies also focused on the effect of electrochemiluminescence on the size of the calixarene as well as on the chain length of the carboxylic tether to the rim.

8.7. Calixarenes as Catalysts Shinkai and co-workers [137, 138] have studied kinetics with a series of sulfonato-calixarene derivatives shown in Figure 38 as catalysts for the reaction of addition of water to 1-benzyl-1, 4-dihydronicotinamide to form 1-benzyl-6hydroxy-1,4,5,6-tetrahydronicotinamide. The results showed that the calixarenes having a protic lower rim (a and b) are better catalysts than those lacking this feature (c and d). The same reaction was carried out using p-(carboxyethyl)calixarenes [43] (Fig. 37, with n = 6), and it was found that p-sulphonato calix[6]arene is about four times more effective as a catalyst than p-(carboxyethyl)calix[6]arene. This is attributed to the greater concentration of negative charge at the upper rim of the calixarene, the six sulfonato groups being held more rigidly in place than the more flexible carboxyethyl moieties.

CONH2

CONH2 N

AH / H2O

HO

CH2

SO3Na

CH2

a b c d

R= H R=CH2CO2H R= Me R= n-hexyl

OR 6

Figure 38. Addition of water to 1-benzyl-1,4-dihydronicotinamide catalyzed by sulphonato calix[6]arenes in the presence of acid (AH).

reported to exhibit trans-acylase activity in the methanolysis of p-nitro phenyl acetate [147]. Calix[4]arene-based binuclear complexes (Fig. 39) have been shown to efficiently catalyze phosphate diester cleavage with a high degree of cooperation between the two ZnII or the two CuII centers [139–143]. It has been observed that the binuclear zinc complex 3-Zn2 having calixarene moiety exhibits a very high activity in the trans-esterification of the RNA model substrate HPNP (2-hydroxypropyl p-nitrophenyl phosphate). The catalytic activity of mononuclear calix[4]arene 2-Zn and the reference complex 1-Zn, lacking the flexible calix[4]arene backbone, are low by a factor of 50 and 300 respectively [144, 145]. The binuclear ZnII complex of the bis(aminomethyl)pyridine ligand 3 shows a higher catalytic rate than its CuII analog 3-Cu2 . This is in

N Zn

2+

Zn

N

2+

Zn

2+

N

N

N

8.8. Biomimetic Catalysis

N

N

N N

Although calix[4]arene-based molecular receptors have been reported [139, 140], calix[4]arene-based enzyme models have hardly been developed [141–146]. A Ba(II) complex of crown ether attached to the lower rim of calix[4]arene was

N

CH2

N

N N Zn

2+

OR OROR OR

OR OROR OR 1-Zn 2-Zn

CO2H

N

N N

N Zn N

Zn

2+

3-Zn2

Cu

N 2+

Zn

2+

N

N

N

N N

HO

2+

Cu N

N

n

OR OROR OR 4-Zn3

Figure 37. Schematic representation of the structure of calixarene carboxylic acid derivatives (n = 5–8) at the upper rim used in electrochemiluminescence.

N

N

N

N OH

CH2 OH

2+

OR OROR OR 5-Cu2

Figure 39. Calix[4]arene based model systems for metallophosphodiesterases [R = –CH2 CH2 OCH2 CH3 ].

Calixarenes

491

contrast to the dinucleating bisimidazolyl ligand 5, which forms a highly active CuII complex, 5-Cu2 , and a weakly active ZnII complex, 5-Zn2 . Comparison of dinuclear CuII calix[4]arene 5-Cu2 with binuclear ZnII shows that the high catalytic activity of 5-Cu2 in HPNP trans-esterification is mainly a result of a high turnover rate (kcat ) combined with a moderate substrate-catalyst binding constant (Kass ). However, the observed low activity of 3-Zn2 is mainly due to strong substrate-catalyst affinity along with a relatively moderate turnover rate. This agrees well with the reported higher phosphate affintiy of ZnII . The differences observed in the reactivity between the ZnII complexes (3-Zn2 and 5-Zn2 ) and the CuII complexes (3-Cu2 and 5-Cu2 ) are reflective of their preferential coordination geometries [142, 143]. It was therefore concluded that the tether length and the geometry of the binding arms of the ligands determine the catalytic activity and the mechanism of catalysis. Proposed mechanisms for HPNP cleavage by calix[4]arene complexes of 3-Zn2 and 5-Cu2 are shown in Figure 40. Hence these calix[4]arene based binuclear complexes can be used as models for binuclear metallo-phosphodiesterases [141– 145]. Extension to systems with a third metal ion binding site afforded a trinuclear ZnII complex 4-Zn3 (Fig. 39) that mimics trinuclear ZnII phosphodiesterases [142, 146] due to the cooperative action of three ZnII centers resulting in an enhanced reactivity.

8.9. Calixarenes as Liquid Crystals Tungsten complexes with 8 and 12 dodecyloxy side chains on lower rim were found to exhibit liquid crystalline phase [148]. Swager et al. reported a new type of columnar liquid crystal with a rigid bowlic core based on tungsten–oxo calix[4]arene complexes which is stable over a very wide temperature range. These complexes display mesophases of unusually high stability and novel host–guest effects, which suggests that head-to-tail organization occurs in the mesomorphic state.

NO2

N Zn

O

H

O

O

O HO

O 2+

O

NO2

P

P O

H

:B

O

O -

B:

N N

N N HO

OR OROR OR 3-Zn2

Organic molecules with  electron systems and unsymmetrical charge distributions are useful in nonlinear optics, for frequency-doubling of laser light and electro-optic switching [149, 150]. Reinhoudt and co-workers [151] reported that the tetranitrocalix[4]arene, shown in Figure 41, contains four nonconjugated D--A dipoles within a single molecule and it can be present in four idealized conformations with different relative orientations of the intramolecular nonlinear optic-phores. The high  value of the tetranitrocalix[4]arene enables a high degree of orientation upon poling in a polymeric matrix [152]. It is reported that calix[4]arene based chromophore, calix[4]stilbazole, forms densely packed, highly ordered monolayers on silica surfaces [153]. The second order nonlinear optical properties of calix[4]arene for its different conformations have been investigated theoretically [154]. For further details on this subject, one may refer a recent textbook [155], Nonlinear Optics of Organic Molecules and Polymers.

8.11. Nanocavities The design and synthesis of nanoscale molecular containers such as cavitands [156–158], hemicarcerands [159], and capsules [160–162] result in the selective binding, separation, and sensing of smaller molecules and ions, molecular transport and delivery, stabilization of reactive intermediates, and catalysis through encapsulation [163]. They also mimic the hydrophobic pockets of enzymes [164]. Synthetic calixarene derivatives having cavities of 15–20 Å contribute to molecular recognition thus giving new directions into nanotechnology, microfabrication, and molecular biology [165, 166]. The nanocavities can be used for the encapsulation of drugs and their active transport/delivery through cell membranes. In calix[4]arenes, the aromatic rings were fixed in the perfect cone conformation through the covalent bridging of proximal phenol oxygens with diethylene glycolic chains [167]. Thus the calix-cavitands formed are more rigid and show better affinity toward guest molecules. The crystal structures of calixarenes and the smaller cavitands derived from them showed that the guest molecule is positioned above the plane of the four carbon atoms of the cyclic polyaromatic skeleton [165, 166, 168–170]. The cavity’s depth is generally less than 4 Å and this space is just sufficient for one methylsized fragment to be included. Several approaches have been developed to enlarge the cavity and achieve better guest entrapment. For example, triethylene glycol footed deep cavitands are functionalized with four m-amidinium groups

Cu

Cu

N N

2+

2+

N

2+

CH3 Zn

O

-

8.10. Calixarenes in Nonlinear Optics

N

N

N

N

nPrO

N

N OH

nPrO

OnPr OnPr

OR OROR OR 5-Cu2

Figure 40. Proposed mechanism for 2-hydroxypropyl para-nitrophenyl phosphate cleavage by calix[4]arene based model complexes, 3-Zn2 (left) and 5-Cu2 (right) [R = –CH2 –CH2 –O–CH2 –CH3 ].

O 2N

NO2 NO2

NO2

Figure 41. para-Nitro-calix[4]arene where phenolic oxygens are alkylated with n-propyl group.

Calixarenes

492 at the upper rim in order to effectively bind mono- and dinucleotides in buffered aqueous and methanol solutions [171]. In addition to charged interactions and hydrogen bonding, the cavity also contributes to the overall binding process. A proposed adenosine monophosphate complex of this cavitand is shown in Figure 42. 1 H NMR and molecular modeling studies suggest that the adenine part is located deep inside the host cavity [172]. The cavitand shown in Figure 43 is a much deeper and open ended container molecule with a nanoscale cavity ∼14 Å deep and 12 Å wide reported to date [173]. According to molecular modelling, such a cavity is able to accommodate up to three benzene or chloroform sized molecules. However, its complexing ability toward small organic guests in nonpolar solutions appeared to be poor. The electron-rich interior resulted in considerable interactions with electron deficient fullerene C60 in toluene as shown in Figure 43. The open-ended nature of the cavitand is responsible for its host–guest properties. An alternative strategy to make deep cavitands is to employ self-assembly through intermolecular hydrogen bonding. Multicomponent cavitands have been constructed by MacGillivray in the solid state by co-crystallization of resorcinarene (R = Me) and 4,4′ -bipyridine from ethanol in the presence of an aromatic or polycyclic guest. This resulted in an one-dimensional wavelike polymer structure (guest = ferrocene, 1,1′ -diacetyl ferrocene, p-chlorotoluene, adamantanon, and [2.2]paracyclophane), where the guest molecules interact with the aromatic walls through multiple C–H· · ·  contacts as shown in Figure 44 [174–176].

8.12. Clamshell-Shaped Molecular Containers Geometrically, clamshell-shaped hosts impose more constrictive binding than cavitands and their inner cavities are still exposed to the bulk exterior. Unlike self-assembling

Figure 42. Proposed complex of cavitand with adenosine monophosphate as molecular recognition unit [R = –(CH2 )3 (OCH2 CH2 )3 OCH3 ]. Adapted from [171], K. T. Chapman and W. C. Still, J. Am. Chem. Soc. 111, 3075 (1989). © 1989, American Chemical Society.

Figure 43. Deep cavitand with nanoscale cavity having electron rich interior provides considerable interaction for filling with fullerene, C60 . Adapted from [173], F. C. Tucci et al., J. Org. Chem. 64, 4555 (1999). © 1999, American Chemical Society.

capsules, their cavities are constructed through covalent bonding. Reinhoudt et al. reported the synthesis of large clamshell-shaped hydrophobic surfaces by linking two calix[4]arenes with one resorcinarene or one calix[4]arene with two resorcinarenes as shown in Figure 45 [177]. Four strong intramolecular hydrogen bonds between the bridging acetamide CH2 C(O)–NH and the oxygens of the cavitand acetal bridges were found to significantly rigidify the structure of these extended surfaces. The resultant hosts (Fig. 45) bind corticosteroids, carbohydrates, nucleosides, and alkaloids in CDCl3 solutions.

8.13. Self-Assembling Nanocapsules Self-assembly represents an alternative way to construct nanocavities and their host–guest complexes [160, 161]. With appropriate curvature and carefully designed positioning of interacting sites, calixarene-based self-assembling systems generate capsules. Capsules have been organized through hydrogen bonding and metal–ligand interactions. An example of self-assembling cavities is the calix[4]arene tetraurea dimer shown in Figure 46 [178, 179]. This has an internal volume of only 50–250 Å 3 and can entrap a single guest molecule of moderate dimensions [180, 181].

Figure 44. Self-assembling cavitand formed between recorcinarene and 4,4′ -bipyridine through intermolecular hydrogen bonds. The cavities are filled with guest molecules, such as ferrocene, p-chlorotoluene, adamantanon, and [2.2]paracyclophane. Adapted from [174], L. R. MacGillivray and J. L. Atwood, J. Am. Chem. Soc. 119, 6931 (1997). © 1997, American Chemical Society.

Calixarenes

493 8.14. Spheres and Tubular Structures

Figure 45. Clamshell-shaped molecular containers formed by linking calix[4]arene and resorcinarene (either in 1:2 or in 2:1 ratio) through hydrogen bonds, possessing hydrophobic interiors to bind carticosteroids, carbohydrates, nucleosides, or alkaloids. Adapted from [177], I. Higler et al., J. Org. Chem. 61, 5920 (1996). © 1996, American Chemical Society.

The p-sulfonatocalix[4]arene has been reported to assemble in an up–up fashion to result in curved structures by Atwood and co-workers [182], using a reaction system consisting of Na5 [p-sulfonatocalix[4]arene, pyridine N -oxide, and La(NO3 )3 ·6H2 O in 2:2:1 ratio. The hydrophobic regions of the calixarenes are assembled in an up–up radial symmetric fashion along the surface of a sphere, where they constitute an organic shell around an aqueous polar core. This spherical assembly measures ∼28 Å across and has a volume of ∼11,000 Å 3 . The interior of the sphere consists of two Na(H2 O)6 ions surrounded by an additional 24 water molecules. The cavities of the calixarenes are situated just below the polar surface of the sphere and constitute a series of hydrophobic pockets. Twelve pyridine N -oxide molecules penetrate the polar surface of the sphere and are bound within the hydrophobic pockets via -stacking interactions. Their oxygen atoms extend outward and coordinate to La3+ ions above the surface of the sphere. The hydrophilic interstitial regions between the spheres contain, in addition to La3+ ions, an array of water molecules and hydrated Na+ ions. The core of each sphere has a diameter of ∼15 Å and a volume of ∼1700 Å 3 . Adjacent spheres are tethered to one another by coordination of sulfonate oxygen atoms to interstitial lanthanum ions. Thus these La(III) ion centers are both first- and second-sphere coordinated to the calixarene molecules. When the three components are taken in a molar ratio of 2:8:1, crystals obtained revealed a tubular assembly of ∼28 Å in diameter. The calixarene molecules are arranged along the surface of a cylinder that is analogous to the spherical assembly in that there is a polar core, a hydrophobic midregion constituting the tube, and a polar outer shell. However, the organic shell is not composed of purely calixarene molecules but contains pyridine N -oxide molecules intercalated between the aromatic rings of adjacent macrocycles. The calixarene and the pyridine N -oxide molecules form a chiral helical assembly along the length of the tube. The helix consists of a single strand of alternating calixarene and pyridine N -oxide molecules and there are 4.5 such units per turn. Both the hydrophilic core and the hydrophilic interstices between the tubules contain complex arrays of sodium, lanthanum, and water molecules.

8.15. Synthetic Nanotubes

Figure 46. Self-assembling calixarene capsule formed from calix[4]arene tetraurea dimer through intermolecular hydrogen bonds. R = Alk, R1 = Alk, Ar, Ts. Adapted from [178], K. D. Shimizu and J. Rebek, Jr., Proc. Natl. Acad. Sci. USA 92, 12403 (1995). © 1995, National Academy of Sciences.

Calix[4]arenes in 1,3-alternate conformations have a -basic benzene hole or interlayers through which metal ions can easily pass with the help of cation– interaction [183]. This plays a crucial role in metal ion transport across ion channels, metal ion inclusion in fullerenes, and intercalation of metal cations into graphite layers [184]. Nanotubes of 1,3alternate calix[4]arenes having a well-defined inner diameter for the metal tunneling have been reported (Fig. 47) [185].

Calixarenes

494

O

O

CH2

CH2

Pr

Pr

O O O

O

CH2 CH2 O

O

X

X

O

O

CH2 CH2 Pr

Pr O

O

O O Pr Pr CH2

to suit special applications of these based on their intrinsic properties. Properties such as metal ion binding and the capacity to hold and/or recognize molecules or species of calix[4]arenes have been guided and thoroughly refined by the rational design followed by the development. The metal ion complexation chemistry of calix[4]arene derivatives is still going on at a very slow pace. While the biomimetic chemistry of calix[4]arene derivatives is in its early phase, their ability to host organic and/or inorganic species in their neutral, cationic, or anionic form seems to have grown very rapidly during the past few years. Characterization of both the organic and the corresponding inorganic derivatives of calixarenes by different spectral and crystallographic methods has become an important feature in the current literature on calixarenes. Very recent studies have shown the putative utility of carefully designed calix[4]arene derivatives to act as catalysts, hosts, receptors, sensors, cavitands, and nanomaterials as reported in this review chapter. Based on the rate at which the studies are going on with the calixarenes in the literature, it is logical to draw a conclusion that the smartly designed derivatives of calix[4]arene would be occupying a pivotal role in molecular recognition, analytical devices, and materials science and technology that leads to the enrichment of the knowledge of nanoscience in order to eventually get into nanotechnology of calixarenes in the very near future.

CH2

O

O

X

X

O

O CH2

CH2

Pr

Pr

O O O

O Pr

Pr

CH2 CH2 O

O

X=

Figure 47. Synthetic nanotubes of 1,3-alternate-calix[4]arene having a -basic hole for metal tunneling.

9. CONCLUSIONS It is evident from the literature that the field of calixarenes has been going through rapid phase changes during the past two decades ever since these were synthesized in pure and in large quantities. During the initial period, the literature was primarily focused on developing methods for organic derivatization, particularly of calix[4]arenes, at both its lower and upper rims. As a result of this, the literature of calix[4]arenes on one hand became very rich with a spectrum of organic derivatives; on the other hand, such an outcome led to the rational design of calix[4]arene derivatives

GLOSSARY Arene cavity The cavity formed by the enclosure of the aromatic rings in a calixarene. Calix[n]arenes A group of phenolic macrocyclic molecules resulting from the condensation of para-substituted phenol and formaldehyde (n refers to the number of phenolic units). Crystal structure Three-dimensional arrangement of atoms derived from single crystal X-ray diffraction that provides information about the structure, conformation, and geometry of the molecule. Functionalization Chemical modification on the lower or upper rim leading to the formation of derivatives containing appropriate functional groups. Lower rim The rim containing the phenolic –OH groups in a para-substituted calixarene. Metalloenzymes Those enzymes that exhibit their function as a result of the presence of a specific metal ion. Receptor A molecule that possesses the ability to bind to neutral or ionic species through recognition. Upper rim The rim containing the groups present at the para-position in a calixarene.

ACKNOWLEDGMENTS C. P. R. expresses his appreciation to his former student, Dr. P. Venkateswara Rao, who has ventured into this exciting field of calixarene chemistry and acknowledges the help rendered by the groups of Professor E. Kolehmainen and Professor K. Rissanen, Finland. We thank Amjad Ali for help

Calixarenes

during the preparation of this review chapter. C. P. R. gratefully acknowledges the financial support extended by the Department of Science and Technology, BRNS of Department of Atomic Energy, and Council for Scientific and Industrial Research (CSIR), New Delhi. M. D. acknowledges CSIR for the award of a senior research fellowship.

REFERENCES 1. C. D. Gutsche, Calixarenes, in “Monographs in Supramolecular Chemistry” (J. F. Stoddart, Ed.). Royal Society of Chemistry, London, 1989. 2. J. W. Cornforth, E. D. Morgan, K. T. Potts, and R. J. W. Rees, Tetrahedron 29, 1659 (1973). 3. V. Bohmer and J. Vicens, “Calixarenes: A Versatile Class of Macrocylic Compounds.” Kluwer Academic, Dordrecht, 1991. 4. C. D. Gutsche and R. Muthukrishnan, J. Org. Chem. 43, 4905 (1978). 5. C. D. Gutsche and M. Iqbal, Org. Syn. 68, 234 (1989). 6. C. D. Gutsche, B. Dhawan, M. Leonis, and D. Stewart, Org. Syn. 68, 238 (1989). 7. J. H. Munch and C. D. Gutsche, Org. Syn. 68, 243 (1989). 8. A. Ikeda and S. Shinkai, Chem. Rev. 97, 1713 (1997). 9. V. Bocchi, D. Foina, R. Ungaro, and G. D. Andreetti, Tetrahedron 38, 373 (1982). 10. C. D. Gutsche, A. E. Gutsche, and A. I. Karaulov, J. Incl. Phenom. 3, 447 (1985). 11. G. Ferguson, B. Kaitner, M. A. McKervey, and E. M. Seward, J. Chem. Soc., Chem. Commun. 584 (1987). 12. E. M. Collins, M. A. McKervey, and S. J. Harris, J. Chem. Soc., Perkin Trans. 1 372 (1989). 13. F. Arnaud-Neu, E. M. Collins, M. Deasy, G. Ferguson, S. J. Harris, B. Kaitner, A. J. Lough, M. A. McKervey, E. Marques, B. L. Ruhl, M.-J. Schwing-Weill, and E. M. Seward, J. Am. Chem. Soc. 111, 8681 (1989). 14. G. Calestani, F. Ugozzoli, A. Arduini, E. Ghidini, and R. Ungaro, J. Chem. Soc., Chem. Commun. 344 (1987). 15. A. Arduini, E. Ghidini, A. Pochini, R. Ungaro, G. D. Andreetti, G. Calestani, and F. Ugozzoli, J. Incl. Phenom. 6, 119 (1988). 16. P. V. Rao, C. P. Rao, E. Kolehmainen, E. K. Wegelius, and K. Rissanen, Chem. Lett. 1176 (2001). 17. S.-K. Chang and I. Cho, J. Chem. Soc., Perkin Trans. 1 211 (1986). 18. K. Araki, A. Yangi, and S. Shinkai, Tetrahedron 49, 6763 (1993). 19. C. Loeber, D. Matt, A. De Cian, and J. Fischer, J. Organomet. Chem. 475, 297 (1994). 20. C. Dieleman, C. Loeber, D. Matt, A. De Cian, and J. Fischer, J. Chem. Soc., Dalton Trans. 3097 (1995). 21. S. Pappalardo, L. Giunta, M. Foti, G. Ferguson, J. F. Gallagher, and B. Kaitner, J. Org. Chem. 57, 2611 (1992). 22. S. Pappalardo, G. Ferguson, P. Neri, and C. Rocco, J. Org. Chem. 60, 4576 (1995). 23. E. M. Collins, M. A. McKervey, E. Madigan, M. B. Moran, M. Owens, G. Ferguson, and S. J. Harris, J. Chem. Soc., Perkin Trans. 1 3137 (1991). 24. H. Yamamoto, T. Skaki, and S. Shinkai, Chem. Lett. 469 (1994). 25. L. C. Groenen, B. H. M. Ruel, A. Casnati, W. Verboom, A. Pochini, R. Ungaro, and D. N. Reinhoudt, Tetrahedron 47, 8379 (1991). 26. J.-D. Van Loon, W. Verboom, and D. N. Reinhoudt, Organic Preparations Procedures Int. 24, 437 (1992). 27. M. Dey, C. P. Rao, E. Kolehmainen, and P. Guionneau, to be published. 28. A. Ali, C. P. Rao, and E. Kolehmainen, to be published. 29. G. Alfieri, E. Dradi, A. Pochini, R. Ungaro, and G. D. Andreeti, J. Chem. Soc., Chem. Commun. 1075 (1983).

495 30. R. Ungaro, A. Pochini, and G. D. Andreeti, J. Inclusion Phenom. 2, 199 (1984). 31. C. D. Gutsche and I.-G. Lin, Tetrahedron 42, 1633 (1986). 32. C. D. Gutsche and J. A. Levine, J. Am. Chem. Soc. 104, 2652 (1982). 33. C. D. Gutsche and P. F. Pagoria, J. Org. Chem. 50, 5795 (1985). 34. P. Timmerman, W. Verboom, D. N. Reinhoudt, A. Arduini, S. Grandi, A. R. Sicuri, A. Pochini, and R. Ungaro, Synthesis 2, 185 (1994). 35. Z. Asfari and J. Vicens, J. Incl. Phenom. 19, 85 (1994). 36. S. Shinkai, S. Mori, T. Tsubaki, T. Sone, and O. Manabe, Tetrahedron Lett. 25, 5315 (1984). 37. S. Shinkai, K. Araki, T. Tsubaki, T. Arimura, and O. Manabe, J. Chem. Soc., Perkin Trans. 1 2297 (1987). 38. A. Arduini, G. Manfredi, A. Pochini, A. R. Sicuri, and R. Ungaro, J. Chem. Soc., Chem. Commun. 936 (1991). 39. A. Arduini, S. Fanni, G. Manfredi, A. Pochini, R. Ungaro, A. Sicuri, and F. Ugozzoli, J. Org. Chem. 60, 1448 (1995). 40. A. Arduini, M. Fabbi, M. Mantovani, L. Mirone, A. Pochini, A. Secchi, and R. Ungaro, J. Org. Chem. 60, 1454 (1995). 41. C. D. Gutsche, J. A. Levine, and P. K. Sujeeth, J. Org. Chem. 50, 5802 (1985). 42. M. Almi, A. Arduini, A. Casnati, A. Pochini, and R. Ungaro, Tetrahedron 45, 2177 (1989). 43. C. D. Gutsche and I. Alam, Tetrahedron 44, 4689 (1988). 44. M. M. Olmstead, G. Sigel, H. Hope, X. Xu, and P. P. Power, J. Am. Chem. Soc. 107, 8087 (1985). 45. F. Corazza, C. Floriani, A. Chiesi-Villa, and C. Guastini, J. Chem. Soc., Chem. Commun. 1083 (1990). 46. J. A. Acho, L. H. Doerrer, and S. J. Lippard, Inorg. Chem. 34, 2542 (1995). 47. F. Corazza, C. Floriani, A. Chiesi-Villa, and C. Guastini, J. Chem. Soc., Chem. Commun. 640 (1990). 48. V. C. Gibson, C. Redshaw, W. Clegg, and M. J. R. Elsegood, J. Chem. Soc., Chem. Commun. 2371 (1995). 49. F. Corazza, C. Floriani, A. Chiesi-Villa, and C. Rizzoli, Inorg. Chem. 30, 4465 (1991). 50. J. M. Harrowfield, M. I. Ogden, and A. H. White, Aust. J. Chem. 44, 1249 (1991). 51. J. M. Harrowfield, M. I. Ogden, W. R. Richmond, and A. H. White, J. Chem. Soc., Chem. Commun. 1159 (1991). 52. J. M. Harrowfield, M. Mocerino, B. W. Skelton, C. R. Whitaker, and A. H. White, Aust. J. Chem. 47, 1185 (1994). 53. B. M. Furphy, J. M. Harrowfield, M. I. Ogden, B. W. Skelton, A. H. White, and F. R. Wilner, J. Chem. Soc., Dalton Trans. 2217 (1989). 54. J. M. Harrowfield, M. I. Ogden, A. H. White, and F. R. Wilner, J. Chem. Soc., Dalton Trans. 2153 (1991). 55. C. D. Gutsche and K. C. Nam, J. Am. Chem. Soc. 110, 6153 (1988). 56. P. D. Beer, M. G. B. Drew, P. B. Leeson, K. Lyssenko, and M. I. Ogden, J. Chem. Soc., Chem. Commun. 929 (1995). 57. C. G. Gibbs and C. D. Gutsche, J. Am. Chem. Soc. 115, 5338 (1993). 58. H. Kwart and E. R. Evans, J. Org. Chem. 31, 410 (1966). 59. M. S. Newman and H. A. Karnes, J. Org. Chem. 31, 3980 (1966). 60. X. Delagiue, J. M. Harrowfield, M. W. Hosseini, A. De Cain, J. Fischer, and N. Kyritsakas, J. Chem. Soc., Chem. Commun. 1579 (1994). 61. X. Delagiue, M. W. Hosseini, N. Kyritsakas, A. De Cain, and J. Fischer, J. Chem. Soc., Chem. Commun. 609 (1995). 62. Z. Asfari and J. Vicens, Tetrahedron Lett. 29, 2659 (1988). 63. V. Bohmer, F. Marschollek, and L. Zetta, J. Org. Chem. 52, 3200 (1987). 64. S.-I. Chen, Ph.D. Thesis, Washington University, St. Louis, 1985, pp. 46–47. 65. J. H. Munch, Makromol. Chem. 178, 69 (1985). 66. K. No, Y. Noh, and Y. Kim, Bull. Korean Chem. Soc. 7, 442 (1986).

496 67. A. Arduini, A. Pochini, S. Reverberi, and R. Ungaro, J. Chem. Soc., Chem. Commun. 981 (1984). 68. V. Bohmer, E. Schade, and W. Vogt, Makromol. Chem., Rapid Commun. 5, 221 (1984). 69. W. Saenger, C. Betzel, B. Hingerty, and G. M. Brown, Angew. Chem. Int. Ed. Engl. 22, 883 (1983). 70. S. W. Keller, G. M. Schuster, and F. L. Tobiason, Polym. Mater. Sci. Eng. 57, 906 (1987). 71. P. V. Rao, Ph.D. Thesis, Indian Institute of Technology Bombay, India, 2000, Chaps. 5 and 6. 72. H. Kammerer, G. Happel, and F. Caesar, Makromol. Chem. 162, 179 (1972). 73. C. D. Gutsche, B. Dhawan, K. H. No, and R. Muthukrishnan, J. Am. Chem. Soc. 103, 3782 (1981). 74. C. D. Gutsche, B. Dhawan, K. H. No, and R. Muthukrishnan, J. Am. Chem. Soc. 106, 1891 (1984). 75. C. D. Gutsche and L. J. Bauer, Tetrahedron Lett. 22, 4763 (1981). 76. F. Sansone, S. Barboso, A. Casnati, M. Fabbi, A. Pochini, F. Ugozzoli, and R. Ungaro, Eur. J. Org. Chem. 897 (1998). 77. C. Rizzoli, G. D. Andreetti, R. Ungaro, and A. Pochini, J. Mol. Struct. 82, 133 (1982). 78. M. A. McKervey, E. M. Seward, G. Ferguson, and B. L. Ruhl, J. Org. Chem. 51, 3581 (1986). 79. M. A. McKervey, E. M. Seward, G. Ferguson, B. Ruhl, and S. J. Harris, J. Chem. Soc., Chem. Commun. 388 (1985). 80. I. Oueslati, R. Abidi, H. Amri, P. Thuery, M. Nierlich, Z. Asfari, J. Harrowfield, and J. Vicens, Tetrahedron Lett. 41, 8439 (2000). 81. A. Hamdi, R. Abidi, M. T. Ayadi, P. Thuery, M. Nierlich, Z. Asfari, and J. Vicens, Tetrahedron Lett. 42, 3595 (2001). 82. S. Chowdhury and P. E. Georghiou, J. Org. Chem. 66, 6257 (2001). 83. R. Lamartine, C. Bavoux, F. Vocanson, A. Martin, G. Senlis, and M. Perrin, Tetrahedron Lett. 42, 1021 (2001). 84. S. E. Matthews, V. Felix, M. G. B. Drew, and P. D. Beer, New J. Chem. 25, 1355 (2001). 85. L. Frkanec, A. Visnjevac, B. Kojic-Prodic, and M. Zinic, Chem. Eur. J. 6, 442 (2000). 86. L. C. Groenen, B. H. M. Ruel, A. Casnati, P. Timmerman, W. Verboom, S. Harkema, A. Pochini, R. Ungaro, and D. N. Reinhoudt, Tetrahedron Lett. 32, 2675 (1991). 87. L. C. Groenen, J.-D. Van Loon, W. Verboom, S. Harkema, A. Casnati, R. Ungaro, A. Pochini, F. Ugozzoli, and D. N. Reinhoudt, J. Am. Chem. Soc. 113, 2385 (1991). 88. J.-D. Van Loon, D. Kraft, M. J. K. Ankone, W. Verboom, S. Harkema, W. Vogt, V. Bohmer, and D. N. Reinhoudt, J. Org. Chem. 55, 5176 (1990). 89. J.-D. Van Loon, A. Arduini, W. Verboom, R. Ungaro, G. J. Van Hummel, S. Harkema, and D. N. Reinhoudt, Tetrahedron Lett. 30, 2681 (1989). 90. H. Goldmann, W. Vogt, E. Paulus, and V. Bohmer, J. Am. Chem. Soc. 110, 6811 (1988). 91. G. D. Andreeti, G. Calestani, F. Ugozzoli, A. Arduini, E. Ghidini, A. Pochini, and R. Ungaro, J. Inclusion Phenom. 5, 123 (1987). 92. B. M. Furphy, J. M. Harrowfield, D. L. Kepert, B. W. Skelton, A. H. White, and F. R. Wilner, Inorg. Chem. 26, 4231 (1987). 93. E. B. Brouwer, G. D. Enright, and J. A. Ripmeester, J. Chem. Soc., Chem. Commun. 939 (1997). 94. A. Zanotti-Gerosa, E. Solari, L. Giannini, C. Floriani, A. ChiesiVilla, and C. Rizzoli, J. Chem. Soc., Chem. Commun. 119 (1996). 95. F. Hamada, K. D. Robinson, G. W. Orr, and J. L. Atwood, Supramol. Chem. 2, 19 (1993). 96. S. G. Bott, A. W. Coleman, and J. L. Atwood, J. Chem. Soc., Chem. Commun. 610 (1986). 97. S. W. Bott, A. W. Coleman, and J. L. Atwood, J. Inclusion Phenom. 5, 747 (1987). 98. S. W. Bott, A. W. Coleman, and J. L. Atwood, J. Am. Chem. Soc. 108, 1709 (1986).

Calixarenes 99. T. Hascall, A. L. Rheingold, I. Guzei, and G. Parkin, J. Chem. Soc., Chem. Commun. 101 (1998). 100. A. Zanotti-Gerosa, E. Solari, L. Giannini, C. Floriani, A. ChiesiVilla, and C. Rizzoli, J. Chem. Soc., Chem. Commun. 183 (1997). 101. P. D. Beer, M. G. B. Drew, P. B. Leeson, and M. I. Ogden, Inorg. Chim. Acta 246, 133 (1996). 102. V. C. Gibson, C. Redshaw, W. Clegg, and M. R. J. Elsegood, J. Chem. Soc., Chem. Commun. 1969 (1998). 103. P. D. Beer, M. G. B. Drew, D. Hesek, M. Kan, G. Nicholson, P. Schmitt, P. D. Sheen, and G. Williams, J. Chem. Soc., Dalton Trans. 2783 (1998). 104. P. D. Beer, M. G. B. Drew, A. Grieve, M. Kan, P. B. Leeson, G. Nicholson, M. I. Ogden, and G. Williams, J. Chem. Soc., Chem. Commun. 1117 (1996). 105. F. De M. Ramirez, L. Charbonniere, G. Miller, R. Scopelliti, and J.-C. G. Bunzil, J. Chem. Soc., Dalton Trans. 3205 (2001). 106. J. F. Malone, D. B. Marrs, M. A. McKervey, P. O’Hagan, N. Thompson, A. Walker, F. Arnaud-Neu, O. Mauprivez, M.-J. Schwing-Weill, J.-F. Dozol, H. Rouquette, and N. Simon, J. Chem. Soc., Chem. Commun. 2151 (1995). 107. X.-X. Han, L.-H. Weng, X.-B. Leng, and Z.-Z. Zhang, Polyhedron 20, 1881 (2001). 108. J. L. Atwood, G. W. Orr, F. Hamada, S. G. Bott, and K. D. Robinson, Supramol. Chem. 1, 15 (1992). 109. U. Radius, Inorg. Chem. 40, 6637 (2001). 110. P. D. Beer, M. G. B. Drew, P. B. Leeson, and M. I. Ogden, J. Chem. Soc., Dalton Trans. 1273 (1995). 111. M. I. Ogden, B. W. Skelton, and A. H. White, J. Chem. Soc., Dalton Trans. 3073 (2001). 112. J. L. Atwood and S. G. Bott, “Calixarenes, a Versatile Class of Macrocyclic Compounds” (J. Vicens and V. Bohmer, Eds.), p. 199. Kluwer, Dordrecht, 1991. 113. G. D. Andreeti and F. Ugozzoli, “Calixarenes, a Versatile Class of Macrocyclic Compounds” (J. Vicens and V. Bohmer, Eds.), p. 87. Kluwer, Dordrecht, 1991. 114. J. L. Atwood, F. Hamad, K. D. Robinson, G. W. Orr, and R. L. Vincent, Nature (London) 349, 683 (1991). 115. S. R. Izatt, R. T. Hawkins, J. J. Christensen, and R. M. Izatt, J. Am. Chem. Soc. 107, 63 (1985). 116. J. M. Harrowfield, M. I. Ogden, W. R. Richmond, and A. H. White, J. Chem. Soc., Chem. Commun. 1159 (1991). 117. M. J. Schwing-Weill and M. A. McKervey, “Calixarenes, a Versatile Class of Macrocyclic Compounds” (J. Vicens and V. Bohmer, Eds.), p. 149. Kluwer, Dordrecht, 1991. 118. S. Shinkai, in “Calixarenes, a Versatile Class of Macrocyclic Compounds” (J. Vicens and V. Bohmer, Eds.), p. 173. Kluwer, Dordrecht, 1991. 119. A. Cadogan, D. Diamond, S. Cremin, A. M. McKervey, and S. J. Harris, Anal. Proc. 28, 13 (1991). 120. A. Cadogan, D. Diamond, M. R. Smyth, G. Svehla, M. A. McKervey, E. M. Seward, and S. J. Harris, Analyst 115, 1207 (1990). 121. F. Cadogan, P. Kane, M. A. McKervey, and D. Diamond, Anal. Chem. 71, 5544 (1999). 122. J. A. Brunink, J. R. Haak, J. G. Bomer, D. N. Reinhoudt, M. A. McKervey, and S. J. Harris, Anal. Chim. Acta 254, 75 (1991). 123. D. Diamond, G. Svehla, E. Seward, and M. A. McKervey, Anal. Chim. Acta 204, 223 (1988). 124. M. Telting-Diaz, D. Diamond, and M. R. Smyth, Anal. Chim. Acta 251, 149 (1991). 125. M. Telting-Diaz, F. Regan, D. Diamond, and M. R. Smyth, J. Pharm. Biomed. Anal. 8, 695 (1990). 126. R. J. Forster, F. Regan, and D. Diamond, Anal. Chem. 63, 876 (1991). 127. R. J. Forster and D. Diamond, Anal. Chem. 64, 1721 (1992). 128. D. Diamond and K. Nolan, Anal. Chem. 73, 22A (2001). 129. S. O’Neill, S. Conway, J. Twellmeyer, O. Egan, K. Nolan, and D. Diamond, Anal. Chim. Acta 398, 1 (1999).

Calixarenes 130. W. H. Chan and X. J. Wu, Analyst 123, 2851 (1998). 131. M. McCarrick, B. Wu, S. J. Harris, D. Diamond, G. Barrett, and M. A. McKervey, J. Chem. Soc., Chem. Commun. 1287 (1992). 132. Y. Kubo, S. Maeda, S. Tokita, and M. Kubo, Nature 382, 522 (1996). 133. I. Aoki, T. Sakaki, S. Tsutsui, and S. Shinkai, Tetrahedron Lett. 33, 89 (1992). 134. R. Grigg, J. M. Holmes, S. K. Jones, and W. D. J. A. Norbert, J. Chem. Soc., Chem. Commun. 185 (1994). 135. L. Zhang and J. L. Coffer, J. Phys. Chem. B 101, 6874 (1997). 136. L. Zhang, J. L. Coffer, J. Wang, C. D. Gutsche, J.-J. Chen, and O. Chyan, J. Am. Chem. Soc. 118, 12840 (1996). 137. S. Shinkai, H. Koreishi, S. Mori, T. Sone, and O. Manabe, Chem. Lett. 1033 (1985). 138. S. Shinkai, S. Mori, H. Koreishi, T. Tsubaki, and O. Manabe, J. Am. Chem. Soc. 108, 2409 (1986). 139. V. Bohmer, Angew Chem. Int. Ed. Engl. 34, 713 (1995). 140. J. Scheerder, J. P. M. Van Duynhoven, J. F. J. Engbersen, and D. N. Reinhoudt, Angew. Chem. Int. Ed. 35, 1090 (1996). 141. P. Molenveld, S. Kapsabelis, J. F. J. Engbersen, and D. N. Reinhoudt, J. Am. Chem. Soc. 119, 2948 (1997). 142. P. Molenveld, W. M. G. Stikvoort, H. Kooijman, A. L. Spek, J. F. J. Engbersen, and D. N. Reinhoudt, J. Org. Chem. 64, 3896 (1999). 143. P. Molenveld, J. F. J. Engbersen, H. Kooijman, A. L. Spek, and D. N. Reinhoudt, J. Am. Chem. Soc. 120, 6726 (1998). 144. P. Molenveld, J. F. J. Engbersen, and D. N. Reinhoudt, J. Org. Chem. 64, 6337 (1999). 145. P. Molenveld, J. F. J. Engbersen, and D. N. Reinhoudt, Eur. J. Org. Chem. 3269 (1999). 146. P. Molenveld, J. F. J. Engbersen, and D. N. Reinhoudt, Angew. Chem. Int. Ed. 38, 3189 (1999). 147. R. Cacciapaglia, A. Casnati, L. Mandolini, and R. Ungaro, J. Am. Chem. Soc. 114, 10956 (1992). 148. B. Xu and T. M. Swager, J. Am. Chem. Soc. 115, 1159 (1993). 149. D. J. Williams, Angew. Chem. Int. Ed. Engl. 23, 6 (1984). 150. “Nonlinear Optical Properties of Organic Molecules and Crystals” (D. S. Chemla and J. Zyss, Eds.), Vols. 1 and 2. Academic Press, Orlando, 1987. 151. E. Kelderman, L. Derhaeg, G. J. T. Heesink, W. Verboom, J. F. J. Engbersen, N. F. Van Hulst, A. Persoons, and D. N. Reinhoudt, Angew. Chem. Int. Ed. Engl. 31, 1075 (1992). 152. S. Nijhuis, G. L. J. A. Rikken, E. E. Havinga, W. Ten Hoeve, H. Wynberg, and E. W. Meijer, J. Chem. Soc., Chem. Commun. 1093 (1990). 153. D. Q. Li, X. Yang, and D. McBranch, Synthetic Metals 86, 1849 (1997). 154. E. Brouyere and J. L. Bredas, Synthetic Metals 71, 1699 (1995). 155. “Nonlinear Optics of Organic Molecules and Polymers” (H. S. Nalwa and S. Miyata, Eds.), p. 885. CRC Press, Boca Raton, FL, 1997. 156. J. R. Moran, S. Karbach, and D. J. Cram, J. Am. Chem. Soc. 104, 5826 (1982).

497 157. R. C. Helgeson, M. Lauer, and D. J. Cram, J. Chem. Soc., Chem. Commun. 101 (1983). 158. Cavitands: D. M. Rudkevich and J. Rebek, Jr., Eur. J. Org. Chem. 1991 (1999). 159. A. Jasat and J. C. Sherman, Chem. Rev. 99, 931 (1999). 160. M. M. Conn and J. Rebek, Jr., Chem. Rev. 97, 1647 (1997). 161. D. Philp and J. F. Stoddart, Angew. Chem. Int. Ed. Engl. 35, 1155 (1996). 162. Self-assembling capsules: (a) J. de Mendoza, Chem.-Eur. J. 4, 1373 (1998); (b) J. Rebek, Jr., Acc. Chem. Res. 32, 278 (1999). 163. Reactions inside cavities: R. Warmuth, Eur. J. Org. Chem. 423 (2001). 164. C. D. Gutsche, Acc. Chem. Res. 16, 161 (1983). 165. D. J. Cram and J. M. Cram, “Container Molecules and their Guests.” Royal Society of Chemistry, Cambridge, UK, 1997. 166. C. D. Gutsche, “Calixarenes Revisited.” Royal Society of Chemistry, Cambridge, UK, 1998. 167. A. Arduini. W. M. McGregor, D. Paganuzzi, A. Pochini, A. Sechhi, F. Ugozelli, and R. Ungaro, J. Chem. Soc., Perkin Trans. 2 839 (1996). 168. D. J. Cram, S. Karbach, H.-E. Kim, C. B. Knobler, E. F. Maverick, J. L. Ericson, and R. C. Helgeson, J. Am. Chem. Soc. 110, 2229 (1988). 169. D. J. Cram, K. D. Stewart, I. Goldberg, and K. N. Trueblood, J. Am. Chem. Soc. 107, 2574 (1985). 170. J. A. Tucker, C. B. Knobler, K. N. Trueblood, and D. J. Cram, J. Am. Chem. Soc. 111, 3688 (1989). 171. K. T. Chapman and W. C. Still, J. Am. Chem. Soc. 111, 3075 (1989). 172. L. Sebo, F. Diedrich, and V. Gramlich, Helv. Chim. Acta 83, 93 (2000). 173. F. C. Tucci, D. M. Rudkevich, and J. Rebek, Jr., J. Org. Chem. 64, 4555 (1999). 174. L. R. MacGillivray and J. L. Atwood, J. Am. Chem. Soc. 119, 6931 (1997). 175. L. R. MacGillivray and J. L. Atwood, J. Chem. Soc., Chem. Commun. 181 (1999). 176. L. R. MacGillivray, H. A. Spinney, J. L. Reid, and J. A. Ripmeester, J. Chem. Soc., Chem. Commun. 517 (2000). 177. I. Higler, P. Timmerman, W. Verboom, and D. N Reinhoudt, J. Org. Chem. 61, 5920 (1996). 178. K. D. Shimizu and J. Rebek, Jr., Proc. Natl. Acad. Sci. USA 92, 12403 (1995). 179. O. Mogek, V. Bohmer, and W. Vogt, Tetrahedron 52, 8489 (1996). 180. S. Mecozzi and J. Rebek, Jr., Chem. Eur. J. 4, 1016 (1998). 181. R. Meissner, X. Garcias, S. Mecozzi, and J. Rebek, Jr., J. Am. Chem. Soc. 119, 77 (1997). 182. G. W. Orr, L. J. Barbour, and J. L. Atwood, Science 285, 1049 (1999). 183. A. Ikeda and S. Shinkai, J. Am. Chem. Soc. 116, 3102 (1994). 184. R. A. Kumpf and D. A. Dougherty, Science 261, 1708 (1993). 185. A. Ikeda and S. Shinkai, J. Chem. Soc., Chem. Commun. 2375 (1994).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Cantilever-Based Sensors D. Then, C. Ziegler University of Kaiserslautern, Kaiserslautern, Germany

CONTENTS 1. Introduction 2. Sensors and Biosensors 3. Transduction Principles 4. Sensor Applications in the Gas Phase 5. Sensor Applications in the Liquid Phase 6. Summary and Future Trends in Microcantilever Sensors Glossary References

1. INTRODUCTION The invention of scanning force microscopy (SFM) by Binnig et al. in 1986 [1] was a first bridge between microscopy and micromechanics. SFM measures the interaction forces between two surfaces from the micrometer level down to the atomic scale. Intermolecular forces are not the only forces that can be detected by this technique. A wide variety of different signal domains such as magnetic [2], electrical [3], electrochemical [4], elastic [5], thermal [6], biological [7], and chemical [8] signals can be detected with high sensitivity and (moderate) speed through transduction into a mechanical motion of a cantilever beam. SFM measures the small forces acting on a sharp tip at the apex of a thin beam clamped at one side, the cantilever. The force on the tip bends the cantilever, which acts as a force transducer. The support of the cantilever is mounted into a holder and the motion of the beam is measured via the tunneling effect, with an interferometer, or via the optical lever deflection technique, the latter being the most commonly used method in commercially available SFMs due to its high resolution and its low costs. Newly developed smart force sensors use micromachined devices with integrated piezoresistive, integrated piezoelectric, and integrated capacitance detection of the cantilever deflection. It is possible to produce inexpensive force sensors in a reproducible way by taking advantage of the batch silicon micromachining techniques developed for integrated circuit (IC) process technology. Miniaturization and mass

ISBN: 1-58883-057-8/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

production are the main reasons for the success of SFM in research laboratories and industrial applications. These lead to cantilevers with low spring constants for a high sensitivity to the applied forces or high resonance frequencies for faster response times and insensitivity to external disturbances. The commercial cantilevers are typically made of silicon, silicon nitride, or silicon oxide and are available in a great variety of different shapes, dimensions, and force sensitivities [9–14]. Recent developments combine the newest IC and complementary metal oxide semiconductor (CMOS) technologies to get intelligent cantilevers [15–23], extremely small cantilevers [24–31], or large arrays [32–39]. In the last few years a second evolution in the world of nanoscience was seen, and cantilevers were applied in scenarios other then detecting surfaces. Because of their above-mentioned benefits of high sensitivity, small dimensions, short response times, and low prize they were used to detect changes in temperature, surface stress, and mass in picogram amounts of analyte without touching the surface any more. Cantilevers were used in sensor systems to get a completely new type of miniaturized transducer based on fundamental principles of physics such as the bimetallic effect, surface stress, or the harmonic oscillator. With cantilever sensors it is possible to push “the powers of 10” into a region in which otherwise only time-consuming and expensive analytical tools can be applied or which is completely inaccessible by means of macroscopic techniques. In this chapter the fundamental physics of microcantilever sensors will be described as well as recent methods, experiments, and prospective directions of the newest sensing systems. The discussion is limited to chemical and/or biological sensors as well as radiation sensors. Physical sensing devices such as accelerometers [40, 41], gyroscopes [42], or flow sensors [43, 44] with beam architecture will be excluded, but micromachined cantilevers show interesting abilities to enhance the power of these devices, too.

2. SENSORS AND BIOSENSORS (Bio-)chemical sensors are miniaturized devices that convert a chemical state into an electrical signal. The chemical state is determined by different concentrations, partial pressures, or activities of particles such as atoms, molecules,

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 1: Pages (499–516)

500 ions, or biologically relevant compounds to be detected in the gas, liquid, or solid phase [45]. A sensor can be divided into three main components. The first is the recognition site with a sensitive or selective layer that recognizes the signal of interest. In some cases there is also a preprocessing element (e.g., filter, temperature, catalyst, or preconcentration) that can add further selectivity or sensitivity. The second part of a sensor system is the signal transducer. It converts the chemical or biological information into the physical signal (e.g., mass change into frequency change of a quartz crystal microbalance [46]) which then can be detected, filtered, amplified, and displayed by a readout system as the third part (Fig. 1). There are a huge variety of transducer principles, but in the present context only acoustic (often called mass-sensitive) transducers are of importance. The advantage of using acoustic sensors compared with classical analytical techniques or most optical sensors is the possibility of working label-free. An optimal sensor system should be highly selective, stable, and sensitive to the analyte (“sss”). It should be able to work continuously and time-resolved to allow for on-line or in-line detection of changes in concentrations. This is only possible if the sensor response is completely reversible. By the combination of several different sensors into arrays with pattern recognition and multicomponent analysis of the complex data pattern the challenge to analyze complex mixtures could be met successfully. Such arrays are often called “electronic nose” [47] or “electronic tongue” [48] although there are only a few similarities to the functions of a human nose and tongue. In an ideal electronic nose or tongue, each single sensor should respond to only one analyte with a signal proportional to the analyte concentration, ensuring complete reversibility. The vector space spanned by such sensor signals would consist of orthogonal basis vectors, which would make the required data processing easy and the sensor response unique. In reality, sensor signals are never unique and cross-sensitivities occur. The scientist or user has to select the best sensors with respect to the application, that is, sensitive layers that cover the range of possible interactions. A biosensor is often defined as a compact analytical device incorporating a biological or biologically derived sensing element either integrated within or intimately associated with a physicochemical transducer [49]. Its recognition layer therefore consists of an enzyme, antibody, nucleic acid, microorganism, cell, or tissue. However, biological molecules can also be detected unspecifically on a nonbiological sensor surface, which will also be covered in this review.

Cantilever-Based Sensors

The detection elements may be divided into three groups: In catalytic biosensors enzymes catalyze a reaction. In affinity sensors a specific recognition reaction between a receptor and a ligand, that is, a key–lock interaction, is detected. This includes antibody–antigen, nucleic acid–nucleic acid, enzyme–enzyme inhibitor, and other receptor systems. The third class includes all other principles, for example, membrane transport through specific pores or whole-cell sensors, in which one or more of the above-mentioned interactions can occur. In this chapter only those systems in which mass changes take place are relevant. Therefore, mainly affinity sensors will be stressed in the following. One problem in the definition of an affinity biosensor is the antagonism of specificity and reversibility. A rule of thumb is that with increasing strength of the chemical interaction between sensitive layer and target analyte, sensitivity and selectivity are enhanced whereas the reversibility of the sensor decreases. Reversibility can be achieved either by selecting an appropriate chemical interaction (e.g., solubility interactions) or by refreshing cycles of the system (e.g., purging with reference/neutralizing gas/solution, adsorbing/desorbing steps) [50]. This reversibility is in strong contrast to the highly specific binding process of, for example, antibodies and antigens with high-affinity constants. Therefore, most affinity biosensors are not sensors according to the defintion given above but can be better described as dosimeters because the biomolecules pile up on the surface with exposure time.

3. TRANSDUCTION PRINCIPLES In general there are three different methods to transduce the recognition event into a micromechanical motion. First, the frequency change due to additional mass loading or a change in the force constant can be measured; that is, the cantilever is used as a microbalance. Second, the bending of a bimetallic cantilever can be used as temperature sensor. In addition, cantilevers can work as stress sensors by measuring the bending due to changes in the surface stress at one side of the cantilever. The transduction principles are shown in Figure 2.

3.1. Resonant Behavior of Cantilever Beams A basic understanding of the mechanical behavior of cantilever beams is crucial to develop cantilever sensors with optimized sensitivity and geometry. (However, those readers only interested in the general subject may skip this section.) The spring constant, resonance frequency, and corresponding quality factor will be derived in this section using the one-dimensional Euler–Bernoulli differential equation of an initially flat, thin, homogenous cantilever beam [51]: 4 2
 wxt + A+  wxt + wxt = qxt (1) EI 4 x t 2 t

Figure 1. Schematic drawing of a sensor system. The information about the chemical nature and concentration of the analyte will be transduced into an electrical signal that can be displayed and stored.


and denote the apparent Young’s modulus [see Eq. (2)] E and specific mass density of the beam material, A = hb, the cross-sectional area, and I is the moment of inertia. is the damping coefficient per unit length per unit velocity,

501

Cantilever-Based Sensors

The variable of position can be written as
d4 W x EI − 20 W x = 0

A + dx4

(6)

The solution of a homogenous fourth-order differential equation is the sum of four linear independent solutions,          x + c2 sin x + c3 sinh x W x = c1 cos L L L    x (7) + c4 cosh L

Figure 2. Schematic drawings of the possible cantilever transduction principles: (a) bimetallic temperature and heat sensor; (b) surface stress sensor; (c) microbalance due to additional mass loading.

and is an additional mass per unit length. qx t represents the driving load in N/m. The additional mass can either exist due to adsorbing molecules on the surface or describe the mass of molecules of the surrounding media that are accelerated by the vibration of the beam. This leads to dependence of the resonance frequency, respectively, the resonance behavior, in general from the surrounding media. Effects of rotary inertia and shear deformations are not included in Eq. (1) nor are nonlinear terms due to a large vibration amplitude. Under the assumption of a beam structure with b ≫ h, the apparent Young’s modulus is given by
= E

 b 0

h/2 −h/2

 4

A20  =
L EI

(9)

The first two boundary conditions are due to the fact that one end of the beam is clamped. The third excludes any bending moment at the free end of the cantilever and the fourth excludes any torsion [53]. By assuming the boundary conditions, a characteristic equation can be calculated: cos  cosh  = −1

(10)

(2) 0 = 1"875

z2 dz dy =

1 bh3 12

(3)

By assuming a harmonic time dependence a variable separation with wx t = W x T t can be performed. For the case without an additional driving load this leads to [52] 4

with L as the length of the cantilever and the nondimensional parameter, , defined by

with discrete values a for  for the first four modes [54]

E 1 − 2

where  is the Poisson’s ratio and E is the Young’s modulus of the beam material. The moment of inertia for a rectangular beam can be calculated as I=

with the constants c1 , c2 , c3 , and c4 assuming the boundary conditions d2 W x  dW x  =0 =0 W 0 = 0   dx x=0 dx2 x=L d3 W x  =0 (8)  dx3 x=L

2

 T t  W x T t


EI 2 x4 t =− − t ≡ 20 = const

A+ W x

A+ T t T t

(4)

2 = 7"86

3 = 11"00

(11)

By inserting these parameters into Eq. (9) the resonance frequency of a clamped beam with rectangular cross-section can be calculated as  
E EI 2a h 2a = 2 a 0 = 2#fa 0 = √ 2 L

A + 12L2 1 −  + with

a = 0 1 2 " " ""

(12)

With the introduction of the quality factor Q=

 A +

(13)

the resonance frequency of a cantilever beam can then be expressed as

The time variable can be written as dT t

d2 T t + 20 T t = 0 + dt 2

A + dt

1 = 4"69

(5)

The solution is a function of time which is equal to that of the model of the one-dimensional harmonic oscillator with T t = Be−t sin0 t +  .

  1 −1/2 a = a 0 · 1 + 4Q2

(14)

a 0 is the resonance frequency in the undamped case = 0. From Eq. (13) one can expect the quality factor to increase linearly with the resonance frequency of the cantilever beam.

502

Cantilever-Based Sensors

Looking at Eq. (12) one can use the model of the onedimensional harmonic oscillator k 1 (15) f = 2# m · n with n = 0"24 as a geometrical parameter for the fundamental mode of a rectangular cantilever beam and k as the force constant:  3 h b
(16) k=E L 4

An additional mass uniformly distributed over the surface can be correlated with a frequency change by   1 1 k (17) − )m = − 2 4# n f02 fm2 where fm is the frequency of the loaded beam  1 f k fm = = 0 2# m∗ + n)m 1 + )m

(18)

m

A Taylor series taking into account only the first order and fm = f0 + )f results in   1 )m 1 )m )f fm ≈ f0 1 − ≈− (19) ⇒ 2 m f0 2 m By using the material parameters the sensitivity per unit area can be written as √

3 1 −  2 )m 2 ≈ −4# n · L · · 2)f (20) A E The minimum detectable mass is given if one takes the minimum measurable frequency change. By considering that all material parameters are constant, the only variable parameter is the length of the cantilever. To achieve a better mass resolution and therefore a higher mass sensitivity the length of the cantilever has to be reduced. However, the length reduction is usually accompanied by a reduction of area also. The second possibility for getting a higher frequency change for a given mass change is to operate the system at higher modes. Besides the frequency characteristics, the force constant is another important parameter. According to Eq. (17) the force constant has to be known if a quantitative calculation of the mass change due to the frequency change is performed. In the literature the mechanical behavior and the determination of spring constants is well described for various methods [55–61]. In Eq. (16) the force constant can be calculated from the material parameters. In reality the material parameters are unknown variables because of the production process of the cantilevers. The length and the width of the cantilever can be controlled quite well during the etching process, but the thickness of the cantilever is the critical parameter. In addition, the Young’s modulus of the bulk material can change due to reflective coatings on one side of the cantilever. This could be calculated, too, but there are more elegant possibilities to calculate the force constant.

One possibility is to measure the static deflection of the cantilever because of the influence of defined forces [62, 63] or pressing with a reference spring [64, 65]. Dynamic methods will use the mean square amplitude of the thermal noise [53, 66] or the frequency shift due to adding a known mass at the apex [67] of the cantilever. For the thermal noise method especially, an easy and fast method for determination of the spring constant of soft cantilevers, the sensitivity of the optical detection system has to be known. The most common method for the determination of this lever sensitivity is that the cantilever is brought into contact with a hard surface and then moved a known distance. The slope of the resulting cantilever deflection versus distance yields the sensitivity, assuming that the piezo scanner driving the cantilever is calibrated. A method without touching the surface by using the drag forces of a moving cantilever is presented in [68]. The third important parameter to characterize the movement of a cantilever is the quality factor Q. The quality factor of a resonator is defined as Q = 2#

W0 )W

(21)

where W0 is the stored vibrational energy and )W is the total energy lost per cycle [compare Eq. (13) for Q described by material parameters]. The frequency stability of a resonator depends on different parameters. One is the noise acting on the resonator as well as the ability to store the energy that is applied to it. The minimum detectable frequency shift * of a cantilever with resonance frequency 0 under the influence of a thermal noise kB T can be estimated to be [69]  0 kB T B * = (22) kQA2 where B, k, kB , and T denote the bandwidth, spring constant of the cantilever, Boltzmann constant, and absolute temperature respectively. A2 is the mean square vibrational amplitude. From Eq. (22) it becomes obvious that an improved frequency resolution of the resonator is associated with an improved quality factor. The dissipation mechanisms that contribute to )W can be distinguished as internal damping due to the physical structure of the resonator and external damping caused by viscous and turbulent flow of the surrounding media. The inverse quality factor can then be written as [70]  1 1 1 1 = = + Q Q Q Q i int ext i

(23)

Qint represents all dissipation effects due to the internal structure of the cantilever, whereas Qext includes all external effects (like viscous media) that act on the cantilever. The gas damping in microstructures can be accessed by a finite-element simulation and is described in [71]. Another approach uses the model of a moving string of beads [72] instead of a beam-shaped geometry. The resulting drag force is the sum of the drag forces of the individual spheres. By assuming that the motion of the cantilever is in a viscous medium (e.g., liquids) the Reynolds numbers are not small

503

Cantilever-Based Sensors

and the drag force becomes not only proportional to the speed but also to the acceleration. The damping can then be expressed as   R

= 6#,R 1 + (24) *L The additional or induced mass can be expressed as [73]     2 9 *L = (25) #R3 L 1 + 3 2 R whereas R is the radius of the sphere, , is the viscosity of the medium, and L is the density. *L = 2,/ L  can be derived from the Navier–Stokes equation and equals the thickness of an additional layer of molecules sticking at the sphere that have to be moved during the oscillation. The mass of the cantilever can therefore be described by a virtual mass m′ = m∗ + mI (mI describes the additional mass of the accelerated molecules) [74]. The relative frequency change due to damping can be approximated by   ) #R3 L 9 *L =− 1+ (26) 0 3 Lb 2 R The change in resonance characteristics leads to a change in the quality factor [75] a bh2 E/12 Q= (27) 6#,RL1 + R/*L where and E denote the mass density and the Young’s modulus of the beam material and *L is again the width of the boundary layer perpendicular to the direction of motion around the sphere. This result indicates an increasing quality factor with decreasing length or increasing resonance frequency. For more information on this complex theme, see, for example, [74, 76–78]. It can be seen from Eq. (22) and the above derivations that the quality factor is an important parameter in the resonance characteristics of a cantilever. An increased quality factor indicates a decreasing minimum detectable frequency change and therefore a better resolution (and sensitivity) of the sensor system. To increase the quality factor the method of active feedback can be used (Fig. 3). This was first introduced in [79] and is normally used for increasing speed and resolution in tapping-mode atomic force microscopy [80, 81]. The active feedback electronic consists of a variable phase shifter . and a gain amplifier G. The equation of motion of a damped feedback system can be written as ¨ +  zt ˙ + 2 zt = F0 eit + F2 zt

(28)

where k is the force constant,  is the damping constant, and F2 is a second periodic force generated by the feedback circuit. The original signal of the oscillator zt = Aeit is amplified with G and phase shifted with #2 and fed back into the original system: F2 = G · Ae

it+#/2

= Ge

i#/2

zt

(29)

Figure 3. Schematic drawing of an active feedback circuit with a variable phase shifter and gain amplifier (black and dotted lines) first presented by [79] and a self-oscillating system (only black lines).

The first-time derivative dz = ieit = zei#/2 ⇒ F2 = dt



 G dz  dt

(30)

leads to an effective reduced damping constant by inserting Eq. (30) into Eq. (28): eff =  −



G 



(10)

The reduced damping leads to a reduced full width at half-peak maximum and therefore to a smaller minimum detectable frequency shift. Besides the actively driven system the electronic feedback circuit can be designed in a way that the system is selfoscillating and no further actuation has to be done. The electronic circuit has to be designed in a way that only the resonance frequencies of the cantilever are amplified and fed back and that the environmental noise is high enough to start the oscillation; that is, it has to contain the resonance frequency. This principle is well known from quartz crystals driven by an oscillation circuit to build a microbalance [82, 83]. Examples of cantilever sensors driven by a self-oscillating system will be discussed later.

3.2. Bending Behavior of Cantilever Beams A uniform surface stress acting on an isotropic material increases (compressive stress) or decreases (tensile stress) the surface area. If this stress is not compensated for at the opposite side of a thin plate or beam, the whole structure will bend (Fig. 4). Between the areas of compressive stress and tensile stress there is a neutral plane that is not deformed. Because of the bending, a force F is acting at the distance x and in the neutral plane results in a bending moment M = F · x. Therefore the radius of curvature R is given by M d2 z 1 = = 2
R dx EI

(31)

504

Cantilever-Based Sensors

dependence. With the assumption of a constant Poisson’s ratio the relative change of the resonance frequency with temperature can be written as   1 1 L 1 b 1 E 1 h 1 f = −3 + + 3 f0 T 2 h0 T L0 T b0 T E0 T

(35)

The subscript zero indicates the variable to be at the starting temperature T0 . The isotropic thermal expansion coefficient is given by 5= Figure 4. Side view of a thin beam. The beam bends around a neutral plane N with a curvature R due to a force F . The origin of the bending can be also be a change in the surface stress (see text).


is the apparent Young’s modulus and I is the where E moment of inertia, which is given by Eq. (3) for rectangular beams. The change in the surface stress at one side of the beam will cause a static bending and the bending moment can be calculated [84] by )3bh M= 2

61 −  1 )3 = R Eh2

(33)

6=

3L2 1 −  )3 Eh2

(34)

Changes in surface stress can be the result of an adsorption process or electrostatic interactions between charged molecules on the surface as well as changes in the surface hydrophobicity and conformational changes of the adsorbed molecules. In addition to surface stress-induced bending, the volume expansion of bimaterial cantilevers can result in a static bending. A bimaterial cantilever undergoes bending due to gas adsorption if the volume expansion coefficients of the two materials are different [87].

3.3. Thermal Effects on Cantilevers By assuming a beam with a constant cross-section consisting of only one material, a change in the surrounding temperature will change the physical dimensions of the beam due to thermal expansion and the elastic properties of the beam material. In considering Eq. (12) the length L, width b,
show a temperature height h, and the Young’s modulus E

1 E E0 T

(37)

If these equations are inserted into Eq. (35), the temperature-dependent relative frequency change will be 1 f 1 = 5 + 6 f0 T 2

(38)

These considerations do not take into account different beam materials in a sandwich structure cantilever. A composite beam of two materials with two different thermal expansion coefficients undergoes a static bending if the temperature changes. For uniform heating the bending z is proportional to the adsorbed heating power and is expressed by the differential equation [88]

By taking into account the boundary conditions of a cantilever (R ≫ L [86], Eq. (33) can be solved and the displacement of the cantilever s can be written as s=

(36)

and the temperature coefficient of the Young’s modulus by

(32)

)3 = 31 − 32 is the differential surface stress with 31 and 32 as surface stress at the upper and lower sides of the cantilever, respectively. Inserting this into Eq. (31) yields Stoney’s formula [85]

1 h 1 L 1 b = = h0 T L0 T b0 T

d2 z h +h = 651 − 52 1 2 2 T − T0 L2 2 dx h2 K

(39)

with K =4+6



h1 h2



+4



h1 h2

2

+

    E1 h1 3 E2 h2 + E2 h2 E1 h1

(40)

By assuming  as the thermal conductivities of layers 1 and 2 and P as the total power on the sensor, the bending is derived as z=

5 t +t L3 51 − 52 1 2 2 P 4 t2 K 1 T1 + 2 T2 b

(41)

Devices can be calibrated in-situ by following the bending of the sensor for a known power transduced into heat at the sensor [89]. Because of the above explanations, on the one side the cantilever can be used as a highly sensitive temperature sensor, but on the other side every temperature change will cause trouble if another parameter like the frequency change is the measured quantity.

505

Cantilever-Based Sensors

3.4. Cantilever Deflection Detection Techniques The most common method to measure the deflection of a cantilever is the optical lever technique [90]. This technique works with a focused laser diode beam at the free end of the cantilever, which is reflected into a split photodiode or a position-sensitive detector. When the cantilever bends, the reflected light moves on the photodetector surface. This movement is proportional to the cantilever deflection. Another optical detection method is based on the interference effect [91–93]. The interference occurs between a reference laser beam and the reflection from the cantilever. This can be realized with a cleaved end of an optical fiber brought close to the cantilever [94]. One part of the light is reflected at the interface fiber/surrounding media and the other part is reflected at the cantilever back into the fiber. These two beams interfere inside the fiber and the interference signal can be measured with a photodiode. Interferometry is very sensitive and provides a direct and absolute measurement of the displacement, but it is limited by the wavelength of the laser light and therefore works only for small deflections. If the distance between two parallel plates is changed, the capacitance between these plates is changing, too. This is used in capacitive sensors to measure the distance between the plates. If one of the plates is a cantilever, small changes in distance can be measured [95–97]. This technique is highly sensitive and provides absolute displacements, but it is unable to measure large displacements and does not work in electrolyte solutions because of the Faradaic currents between the capacitor plates [98]. It is therefore only of limited use in biosensor applications. Alternative approaches use interdigital cantilevers as an optical diffraction grating. The reflected laser light forms a diffraction pattern, where the intensity is proportional to the cantilever deflection [99, 100]. This can be used for atomic force microscopy and infrared detection [101, 102] and is proposed for chemical sensing [103]. An alternative method uses piezoresistive cantilevers. If a piezoresistive material is set under stress, it changes its electrical conductivity. Implemented into cantilevers this effect can be used to monitor stress and therefore the bending of the cantilever. Such stress sensors can be integrated on a cantilever structure with a Wheatstone bridge measuring the resistivity [104–107]. A lot of developments are under way in the field of piezoresisitive cantilevers to gain better resolution and error-free measurements. The advantage compared to standard optical techniques is that neither additional optical components nor laser alignment is needed. Also the readout electronics can be integrated on the same chip using CMOS fabrication [16, 17], and they are unaffected by optical artifacts arising from distributions of the surrounding medium. The disadvantage of most piezoresistive cantilevers is that they only can be used in nonliquid environments because of Faradaic currents that will resut in a short circuit. Recent improvements allow the fabrication of thin, passivated resistors on cantilevers that can be used in electrolyte solutions [108]. Drift compensation is possible due to the development of a symmetrical configuration where the output signal is the difference between a sensing and a reference cantilever [23]. Piezoresistive cantilevers can also vary

their surface temperature by increasing the electrical current flow through the resistor layer. This is proposed to be implemented as a tool for breaking ligand–receptor bindings for the regeneration process of the sensing layer in biosensor applications [98].

3.5. Cantilever Actuation Methods For cantilever sensors based on the measurement of the frequency change it is necessary to drive the cantilever. Without any external actuation the cantilever is only driven by the thermal noise kB T . At room temperature the thermal energy is very low, and long measurement times are the consequence. A better way to drive the cantilever is to apply an external force. In the most commercially available scanning probe microscopes this is done by an external driving piezo mounted near the cantilever. The excitation takes part through mechanical stimulation by shaking the cantilever as well as the complete measurement setup. In a liquid environment the disadvantage of this method is that a lot of standing waves are generated through the acoustic stimulation of the medium. This can be avoided by direct driving of the cantilever using a piezoresistive layer similar to the piezoresisitive readout [22, 106, 109–112]. Lorentz forces can be used for the actuation of a cantilever either by evaporating a magnetic layer on the cantilever and applying an external magnetic field [113–117] or by a conductive pathway on the cantilever with an alternating current and the application of an constant external field [118–121]. Instead of magnetic actuation it is also possible to drive the cantilever with electrostatic forces [118, 122–125] or by an additional heater at the bottom of the cantilever [16, 18, 126].

4. SENSOR APPLICATIONS IN THE GAS PHASE As is described above two basic operation principles are possible for micromachined beams: measurement of the frequency change or measurement of the static bending. In this section an overview from the first experiments to recent developments in the field of sensor applications in the gas phase is given. First, experiments with the deflection detection will be introduced. The bending of a cantilever is very sensitive to changes in the environment of the cantilever and, more interesting, is very sensitive to changes at the surface of the cantilever. Even the smallest changes of the material properties due to adsorbing molecules will cause a static bending. One problem is that the changes have to occur only on one side of the cantilever and that the processes are very complicated and often not completely understood. Often a bending that occurs cannot be explained in full detail and therefore it is nearly impossible to make quantitative measurements. The other problem of the bending measurements compared to frequency measurements is that they are very prone to disturbances in the environment such as temperature changes (if one does not want to measure them) or vibrations. Measuring the frequency changes seems to be better because one can easily quantify mass changes and the operation is a lot more stable. But there are other problems to

506 deal with: First of all the measurement setup is more complicated because of the active driving of the cantilever and one needs a complicated electronic device for the frequency feedback. In addition, the frequency change can arise from temperature changes and/or changes in the force constant. This shows that each method has advantages and disadvantages and one has to decide upon the nature of the task that one has chosen. It is best to measure both the bending and the frequency change at one time. In this case the surface effects can be separated from mass effects. Most modern sensor systems measure the bending and the frequency change simultaneously.

4.1. Deflection Detection The first experiments with microcantilevers beams published in 1994 dealt with temperature measurements and surface stress-induced deflections. Tiny heat fluxes were detected with bimetallic cantilevers by measuring the static deflection of the apex of a cantilever due to temperature changes. In [127] the observation of a chemical reaction with a sensitivity of 1 pJ and a temperature limit of ∼ 10−5 K was published, whereas in [128] thermal and ambient temperature-induced deflections were described. Then calorimeters and photothermal spectroscopy were presented with bending cantilevers as sensors. The use as a calorimeter for investigating surface catalytic reactions was achieved by coating the sensor with a thin Pt layer, which catalyzes the exothermic conversation of H2 and O2 to form water [88]. The thermal analysis of picoliter volumes was shown with the enthalpy change detection during the phase transition of n-alkanes [129–131]. The measurement of heat transfer during a chemical reaction was also described in [132]. By coating the sensor with a specific photon-absorbing material, it was possible to monitor heat effects as a function of wavelength [133]. With a multilayer architecture cantilevers could be used for photothermal spectroscopy as well [134]. In [135] a radiation dosimeter was presented with ultraviolet cross-linking polymers coated at one side of a cantilever as optical detectors. The exposure to radiation causes a bending in the cantilever. Sensing infrared radiation with high sensitivity and resolution was already shown in [136–139]. In [140] cantilevers were introduced as vibration sensors to control the power between two optical fibers. The capability of micromachined cantilevers as stress sensors was shown with the adsorption of molecules to a freshly cleaved gold surface. The unspecific adsorption of ambient molecules like water or hydrocarbons [128] could be seen as well as the forming of self-assembled monolayers of thiols on the gold surfaces [141–144]. The alkanthiols formed a monolayer producing a compressive stress, which can be time-resolved, monitored, and fitted with a Langmuir kinetic. Cantilevers can be used to monitor thin film growth processes [84, 145] too. Good overviews of early cantilever sensor achievements are given in [89] and [146]. By measuring the electric field emanating from a charged-particle collecting sphere, cantilevers can be used as charged-particle flux detectors [147]. Bending microcantilevers can be used as chemical sensors by coating one side with a sensitive layer. This layer will amplify the bending effect compared with an uncoated

Cantilever-Based Sensors

cantilever. This was shown for the water adsorption of an uncoated and a gelatine-coated cantilever [148]. For (specific) chemical sensing a sensitive or selective layer is necessary to get any kind of signal, because without coating there will not be any interaction between the cantilever and the analyte. Polymer-coated cantilevers can be used to monitor concentration changes of different volatile organic compounds [149, 150]. It is easy to apply the polymers to the sensor surface with spray coating, a polymer solution with an airbrush pistol, or the use of a patch-clamp pipette and a micromanipulator. The polymer sticks due to physisorption at the surface and no further modification is necessary. By using array architectures with sensitively coated cantilevers and uncoated reference cantilevers, it is possible to distinguish between a real signal and an error signal and to improve the signal-to-noise ratio. Cantilever arrays with differently coated cantilevers have been realized by a number of different groups. In [151–156] a system with a sequential optical readout was described for an array of eight cantilevers. Through coating with different polymers a great variety of organic compounds could be analyzed using multivariate data analysis techniques. The qualitative and quantitative analysis of mixtures of different organic solvents, perfume oils, flavors, and whiskies was performed. A multiple input chemical sensing unit with coated cantilevers, which are produced by including microelectromechanical system (MEMS) processes, was presented in [157, 158]. In [159] a complete microcantilever system with coated cantilevers was described. In [103] chemical sensing using optical diffraction from an array of microcantilevers was proposed. The micromachined cantilevers contained embedded deformable diffraction gratings and are functionalized with polymer coatings. The latest inventions use array techniques combined with the high resolution of bending cantilevers for radiation detection arrays such as quantum well photon detectors [160] or electromagnetic and nuclear radiation [161]. A new field of investigation and invention for microcantilevers is array infrared detectors for building uncooled infrared devices [37, 102, 162–165] or thermal nanoprobes [33, 166]. Another approach in chemical sensing combines the adsorption-induced stress and the photoinduced stress of target molecules in one device. Microcantilevers that have adsorbed molecules will undergo photoinduced bending that depends on the number of adsorbed molecules on the surface. Also, when microcantilevers undergo photoinduced bending, molecules will adsorb on their surface differently. Depending on the wavelength and cantilever material used, the bending direction can be altered. By combining this with a chemically sensitive coating, a tunable selectivity is achieved [167–170]. A last approach of a microcantilever sensing device brings an array of piezoresistive cantilevers in contact with a sensitive layer. The swelling of the sensitive layer due to analyte exposure is measured directly by measuring the bending of the cantilever [171].

4.2. Frequency Detection Mass-sensitive devices are well established in the wide field of sensors. One of the most popular representatives is the quartz crystal microbalance (QCM) device. They are well

Cantilever-Based Sensors

described in the literature, work in gaseous and liquid environments, and are relatively cheap and easy to handle. They are driven by an oscillation circuit and are therefore very insensitive against external disturbances, providing very good mass sensitivity [172–174]. The reason for new masssensitive devices based on microcantilevers is the low mass of cantilevers compared with that of quartz crystal systems. If one calculates the mass sensitivity of a commercially available QCM with 10 MHz compared with that of a cantilever made of single crystal silicon with a fundamental frequency of 100 kHz, one gets a 1000 times higher mass resolution for the cantilever [175]. The mechanical properties together with the low price and fast response times make vibrating beams ideal candidates for new sensor systems. The first use of a coated beam structure as a chemical sensor was reported in [176], and it was first shown in practice with scanning probe microscopy cantilevers in [128], probing the resonance frequency shift due to water adsorption on the cantilever surface. The first CMOS processed beams for sensor purpose were presented in [177]. With these early devices a mass resolution ranging from nanograms to picograms was predicted. The sensitivity of the humidity sensing device could be enhanced by coating with gelatine or phosphorous acid [148, 178]. An early experiment for testing cantilevers as mass-sensing devices was the adsorption of mercury vapors to gold-coated cantilevers [146, 148, 179]. The first interactions between adsorption, force constant change, and frequency change were described in [84]. If the evaporated mercury layer is too thick, the change in the elastic properties of the beam material will completely compensate for the mass effects due to adsorption. A way to enhance the sensitivity of microdevices to mass loading is the use of materials with a high surface-to-mass ratio as a coating material. This was shown for humidity sensing with a few zeolite single crystals attached to a cantilever sensor [180–182]. Good overviews of the first experiments with cantilever sensors are given in [89] and [146]. A micromechanical sensor for charged-particle flux detection is demonstrated in [147] with 5 particles. Frequency and damping rate variation due to electrostatic force gradients were measured. In [30] cantilevers are used as a low-cost detector in molecular beam experiments to measure the ratio of neutral to ionized clusters and the molecular beam profile with an array architecture of cantilevers. In all these experiments no additional electronic circuit for active feedback was integrated. They were all driven either by the thermal energy or with an additional piezoelectric crystal as is used in scanning probe microscopes for the tapping mode. A typical application of cantilever sensors is the detection of volatile organic compounds with an array of differently coated cantilevers. The selective coating consists of polymers that have shown their versatility in QCM systems [183]. The effect is based on a solubility interaction between an adsorbed analyte and the polymer. The closer the chemical properties of the analyte and the polymer, the more analyte is solved in the polymer matrix. This effect increases the density and the polymer gets heavier and starts to swell. With these effects and the chemical selectivity of different polymers it is possible to build a sensor array for analyzing mixtures of organic solvents.

507 A chemical sensor based on a micromechanical cantilever array consisting of eight differently coated cantilevers with a sequential readout system using a time multiplexing scheme was introduced in [154, 184]. In this setup the cantilever motion is detected via beam deflection with an array of vertical cavity surface-emitting lasers and a position-sensitive detector. The time-multiplexing scheme allows only one cantilever at a time to be monitored. In the dynamic mode the cantilever is driven by a piezoelectric actuator and the frequency change is controlled with a phase locked loop (PLL) controller. This system enables the simultaneous detection of the frequency change and the static bending. A different approach uses an array of piezoresistive cantilevers with an integrated microheater for calorimetry and mass detection [33]. For microbalance purposes the frequency change during the outgasing process of photoresist due to ultraviolet exposure was monitored. In [185–188] the parallel frequency readout of an array of polymer-coated mass-sensitive transducers for sensor applications is described. The coated cantilevers were exposed to vaporized organic solvents. Parallel detection of the thermal noise of the cantilevers was possible with an optical beam deflection technique. By adding a cylindric lens in front of the laser diode a line focus is formed, which can be used to illuminate a whole array of cantilevers. The light is reflected from the beams and collected in a split photodiode. The Fourier spectrum of the detector signal represents the response of the cantilever array to thermal excitation and reveals the resonance frequencies of each individual cantilever. The advantage is the simultaneous detection of several cantilevers and the improvement of the sensitivity by measuring higher eigenmodes. However, this method is very time consuming (the Fourier analysis requires long recording times), and it is impossible to monitor the bending of the cantilever at the same time. The data were evaluated with principal component regression (partial least squares, one method of the multivariate data analysis) for quantitative and qualitative analysis. Therefore, the prediction of unknown concentrations in a binary mixture was possible. A new way to enhance sensitivity is the use of bridge structures, which are both-ends-supported beams. In [189] the fabrication of a micromechanical mass-sensitive resonator using silicon-on-insulator substrates was presented. Bridge structures have a 6.3 times higher fundamental frequency compared with one-end-supported beam structures and therefore have a higher mass sensitivity. The resonators are driven electrostatically and read out optically by means of the beam deflection method. In all these experiments no additional electronic circuits for active feedback have been integrated. They were all driven either by the thermal energy or with an additional driving system (piezoelectric crystal or electrostatical) as is used in scanning probe microscopes for the tapping mode. As described in Section 3.1 the sensitivity can be enhanced by enhancing the quality factor. The quality factor is related to damping and an active feedback loop reduces damping. This was realized in the latest cantilever sensor systems where one can find stable self-oscillating cantilevers with extremely enhanced sensitivity.

508

was described in [175]. The cantilevers were either actuated electrostatically or magnetically and read out optically with the beam deflection technique. For the electrostatic actuation they were integrated in a capacitor-like circuit in which either the cantilever or the counter electrode was supplied with alternating current voltage and the other electrode with direct current voltage or vice versa. For magnetic actuation the cantilevers were coated with a magnetic material like iron or cobalt and premagnetized normal to the cantilever axis. An array was built up of two times three cantilevers at each substrate with different polymer coatings. The measurement chamber was integrated in a gas-mixing station to provide different gases and gas mixtures to the measurement system. The readout can only be done sequentially so far. The cantilever system was integrated in a closed-loop feedback system in which the signal from the cantilever is filtered, amplified, and fed back. The closed loop setup provides stable self-oscillation at the resonance frequency of the cantilever with a short-term stability of 0.005 Hz for the electrostatic and 0.01 Hz for the magnetic actuation [196]. The quality factor can be enhanced for a cantilever with 100 kHz from ∼200 with vibrating in air driven by the thermal energy up to 400,000 driven actively. This enhancement provides a 1000 times higher frequency stability compared with the systems actuated with thermal energy (Fig. 6). With this measurement electronics it is possible to measure frequency changes and the static bending of the cantilever simultaneously. For gas-sensing purposes different polymers were chosen so that with multivariate data analysis it was possible to analyze quantitatively and qualitatively a ternary mixture of three different organic solvents. A qualitative principal component analysis of the mixture after normalizing the data set is shown in Figure 7. In a principal component analysis (PCA), a high dimensional vector space (one dimension from each cantilever sensor if only one variable like maximum frequency change is evaluated) can be projected to two (or at least less than the original) dimensions, which nonetheless contain nearly the whole information. In the so-called scores plot of a PCA, data points which reflect similar properties cluster in a certain region. The more similar the properties, the closer two points are together.

0 ,0 2 0

0 ,0 0 0 0 0 3 5

0 ,0 0 0 0 0 3 0 0 ,0 1 5

0 ,0 0 0 0 0 2 5 0 ,0 1 0 0 ,0 0 0 0 0 2 0 0 ,0 0 5 0 ,0 0 0 0 0 1 5

0 ,0 0 0

0 ,0 0 0 0 0 1 0

74000

74500

75000

75500

V o lta g e w ith fe e d -b a c k

V o lta g e w ith o u t fe e d -b a c k

In [190] a feedback driven system was developed. In this approach an additional feedback circuit is added and PLL device actively drives the frequency. Due to the magnetic actuation this system can be used in a liquid environment as well (see Section 5.3). A promising approach for microcantilever applications is the combination of the sensitivity of micromachined beams with CMOS technology. By including the CMOS process in cantilever fabrication, it is possible to integrate the electronics on the substrate and to get extremely compact devices. A CMOS resonant beam gas-sensing system with on-chip self-excitation was presented in [38, 50, 121, 191–194]. In these devices the cantilever is thermally excited with two heating resistors in the cantilever base that are heated periodically. The temperature increase on the cantilever generates a bending moment due to the difference in thermal expansion coefficients. The cantilever motion is detected with a piezoresistive Wheatstone bridge at the base of the cantilever. The active feedback circuit is integrated on-chip. The output signal of the Wheatstone bridge is filtered and amplified, and the phase shift is controlled by a Schmitt trigger that can be adjusted externally. The phase is shifted to ensure positive feedback. The result is an integrated oscillator operating at the cantilever resonance frequency with short-term stability of better than 0.1 Hz [38, 195]. The cantilever has an on-chip thermal actuation and a closed feedback loop preamplifier circuitry (Fig. 5). For sensor test purposes the cantilevers were coated with different polymers for the identification of organic vapors. With multivariate data analysis it is possible to analyze quantitatively and qualitatively a binary mixture of two different organic solvents. Advantages of this system are the high sensitivity, its compact design, the array architecture, and the integrated electronics. The disadvantages are the limitations with a gaseous environment (the CMOS circuits have not been stable in liquids so far) and the fact that the simultaneous detection of static bending is not possible. Another approach for the use of scanning probe microscopy cantilevers as transducers in gas sensor systems

Cantilever-Based Sensors

76000

F re q u e n c y [H z ]

Figure 5. Micrograph of a 150 9m long resonant cantilever with an on-chip feedback circuit, piezoresistive Wheatstone bridge, and thermal actuation. Reprinted with permission from [38], D. Lange et al., Anal. Chem. 74, 3084 (2002). © 2002, American Chemical Society.

Figure 6. Frequency spectrum of a cantilever driven by thermal noise (upper curve, triangles) and of the same cantilever driven with a closed feedback loop and magnetic actuation (lower curve, squares), measured with the spectrum analyzer function of a hp4195A [175].

509

Cantilever-Based Sensors

5. SENSOR APPLICATIONS IN THE LIQUID PHASE In biosensing applications there is a need for fast, easy to use, cheap, and highly sensitive methods for the recognition of biomolecules. A high grade of parallelization is a must because of the demand of the pharmaceutical industry for high throughput screening. All these points can be fulfilled by micromachined cantilever sensors, and therefore they are the ideal candidates for biosensing applications.

5.1. Surface Modification Figure 7. PCA scores plot for the calibration of an array of six polymercoated cantilevers showing the first two principal components for qualitative and quantitative analysis of a ternary mixture of organics solvents (before calculation of the PCA the data were normalized).

Which property leads to the clustering depends on the particular system. Here, data points cluster if interactions with the sensors and hence their chemical natures are similar. As can be seen in Figure 7 the data points rearrange in clusters. The larger the distance between two clusters the better the sensor array is able to separate the two analytes. After the system was calibrated with a training set of different ternary mixtures, a linear regression model with the partial least squares (PLS) method was applied and a prediction of unknown concentrations could be done. In Figure 8 the predicted values are plotted against the true concentration values of an organic solvent. The concentration is given in parts per million of the carrier gas (synthetic air). Obviously, an acceptable prediction is achieved. Advantages of this system are the high short-term stability, the easy setup, and the possible use in liquid environments. The disadvantage is that so far a parallel readout is not possible.

Figure 8. Predicted concentrations vs. true concentrations for 1-butanol. The model used for this prediction was a linear regression model (PLS2). The ternary mixture was evaluated for only one unknown concentration after training the model with the data from Figure 7.

As defined in Section 2 a liquid sensor or, more specifically, a biosensor has as its main part a transducer combined with a recognition unit. The recognition element is in most cases a single-strand DNA, a protein, or an antibody. These molecules usually will not stick at the surface themselves. Their long-term stability on the surface is in most cases not given. Therefore, a chemical surface functionalization has to be applied to deposit a receptor layer on a cantilever surface. The ideal case of a surface functionalization is a closed, uniform monolayer of the receptor that is covalently anchored to the surface with enough space to interact with the specific ligand. The receptor activity should be maintained over time and resist several regeneration cycles. To anchor a molecule covalently to a surface the sensor has to be modified and functionalized. Most cantilevers are made of silicon, silicon nitride, or silicon dioxide. Such a surface can be activated and functionalized with silanes [197]. The silanes form a dense polymer covalently linked to the surface with free functional side groups. These groups can react with other side groups of the target molecule. Another popular method for surface functionalization is the evaporation of thin layers of noble metals like gold. Alkane chain molecules with thiol side groups form ordered self-assembled monolayers (SAM) on gold substrates [198]. These uniform, densely packed, and covalently linked monolayers can be synthesized with different chain lengths and end groups. The end groups are used to link other molecules to the functionalized surface. With the chain length one can vary the spacer length between the surface and the active site in the recognition layer and therefore guarantee the best possible activity. Because of their variability SAMs are one of the most popular methods used for biosensor applications. From the application point of view, one has to decide which surface modification technique is required. Among others, the following parameters are involved in the choice of the chemical surface modification protocol: type of substrate (e.g., plastic, glass, silicon, gold, or ceramics), the functional unit that has to be attached (e.g., polymers, antibodies, or other proteins or lipids), the type of sample (e.g., DNA, DNA–protein mixtures, pure proteins, or cells), and the type of liquid transport (e.g., pressure driven or electroosmotic). Additional introductions to surface coating techniques and reviews can be found in [199, 200]. If one wants to use the deflection detection method for cantilevers, only one side has to be functionalized. Using thiol monolayers this can be done by evaporating gold only on one side of the cantilever or incubating only one side

510 of a double-sided gold-coated cantilever in the thiol solution. Incubating only one side always causes stress that has to be taken into account [143]. In [201] a procedure is described for coating each side of the cantilever with a different thiol. Nevertheless, surface functionalization is as easy as it sounds, but if deflection detection is used the experimenter has to ensure that the specific interaction occurs only at one side of the cantilever and, even more important, unspecific interaction is blocked. One easy way to block surfaces against unspecific adsorptions is treatment with bovine serum albumin (BSA), a protein with the ability to adsorb on nearly every surface and to be very resistant to a wide variety of chemical and physical conditions such as changes in salt concentration, weak acids and alkaline solutions, and temperature changes. Besides the property of blocking surfaces, the unspecific protein adsorption can be used for surface modification too. By using functionalized BSA a surface can easily be modified biologically with a close, uniform, and compact layer. One example is the modification with biotinylated BSA to get a biotin surface coverage [202]. Another way to modify surfaces with adsorbing proteins is protein A. Protein A has nearly the same adsorbing characteristics as BSA, and it is able to bind the Fc fragment of most immunoglobulin G antibodies [203], which comprise the largest class of antibodies. With this method an antibodycoated surface can be prepared. Besides blocking the surface against unspecific adsorption with BSA or other proteins, the surface can be covered with a monolayer of nonreactive macromolecules or polymers. Polyethyleneglycols have proven their suitability for protecting surfaces against the unspecific absorption of proteins [204, 205]. Organic layers can be created on solid substrates without the need for reactive surface sites with Langmuir–Blodgett (LB) film transfer. The LB deposition transfers ordered layers of amphiphilic molecules from the water/air to the solid/air interface and allows the precise control of multilayer formation [206]. Sol–gels can be used for surface functionalization too [207]. They allow the creation of layers of porous materials with controlled pore size [98]. An increase in the active area can act as a catalyst with the pores serving as a mechanical filter to improve the specificity of the interaction [208].

5.2. Deflection Detection With the deflection detection method in liquids it is possible to use cantilevers as stress sensors by performing electrocapillary curves with gold- or platinum-coated cantilevers as working electrodes [209]. The change in voltage causes a static bending and enables the study of the thermodynamics of the electrode–electrolyte interface. Similar electrocapillary curves were reported in [89, 210]. In [211] the bending of a cantilever due to the electrochemical deposition of polyaniline to one side of the cantilever is reported. By cyclic oxidation and reduction of the polymer the cantilever is reversibly deflected and restored to its original position. Changes in surface energy can also be measured with bending microcantilevers. In [212] a cantilever deflects due to changes in the pH value or salt concentration of the surrounding liquid. This effect could also be shown for thiol-modified cantilevers [201]. Differences in the

Cantilever-Based Sensors

pH dependence of a mercaptohexadecanoic acid-coated cantilever and a hexadecanethiol-covered reference cantilever were reported in [213]. Additional experiments have been presented with metal-modified and aminosilanemodified surfaces. Only the change in the pH value influences the bending; negligible bending was found for variations in salt concentrations [214]. Trace amounts of Hg2+ can be detected by using a gold-coated microcantilever [215]. The bending occurs due to accumulation of the Hg2+ on the gold surface. Coating the surface with a SAM of a long-chain thiol compound could improve the sensitivity. The detection of cesium and calcium ions was possible with the modification of the surface with an ion-selective SAM. A method for in-situ monitoring of ions in the range 10−7 – 10−11 M was developed in [216, 217]. In addition, the use as a temperature sensor in liquid environment was shown in [89]. A polymer-coated cantilever can be used for the detection of different amounts of ethanol in water. It was proposed in [150] that the bending of the cantilever due to a change in ethanol concentration is caused by a swelling of the polymer matrix. In addition, the adsorption-induced variation in the spring constant of microcantilevers due to Na+ ions was proposed in [218]. All of these experiments based on the measurement of the bending of the cantilever and a wide range of signal domains such as pH value, salt concentration, temperature, and electrochemical activity can and will be transduced by a static deflection of the cantilever. During the measurement of a biological affinity reaction all these parameters have to be kept constant or the result is ambiguous. This problem and the difficulties in the quantification of the results are always present during the experiments. It is possible to build a chemical sensor based on polymercoated microcantilevers in a liquid environment. In [219] thiolated 6-cyclodextrins form a SAM on the cantilever surface, which enables an unspecific interaction with analyte molecules such as a naphthalene derivative. Similar results could be obtained for other cyclodextrin layers on smooth and nanostructured gold surfaces. Since the introduction of microcantilevers many biological experiments using static deflection have been presented. In [220–222] a so-called force-amplified biological sensor was used to take advantage of the high sensitivity of scanning probe microscope cantilevers to detect the presence of small magnetic particles bound to a cantilever by a sandwich immunoassay technique. The magnetic particles were functionalized with a specific biological species whose counterpart is on the cantilever surface. To detect the binding event the deflection of the cantilever is measured while the magnetic particles amplify the deflection through magnetization of a strong permanent magnet and a modulated additional magnetic field. With this technique concentrations down to 10−18 M were proposed to be detectable. An important field of investigation is the unspecific adsorption of proteins and cells. In [223] the interrogation of living cells grown directly on the cantilever surface to external chemical stimuli was proposed. The unspecific adsorption of proteins such as BSA on one side of the cantilever causes stress and therefore a static bending of the cantilever [153, 212, 224]. The differentiation between the adsorption of a low-density lipoprotein and its oxidized form on heparin

511

Cantilever-Based Sensors

was monitored in [225] as well as the surface stress induced by slow conformational changes of proteins in [226]. Besides the field of unspecific adsorption, the field of specific binding events is very promising. In [146] the variation in an antibody-coated microcantilever deflection as a function of time due to interactions with antigens was reported. The adsorption kinetics and mechanical properties of thiolmodified DNA oligonucleotides on gold were investigated with microcantilever sensors [227, 228]. The sensor used had an integrated piezoresistive readout and is integrated in a microliquid handling system. The specific detection of two forms of prostate-specific antigen over a wide range of concentrations in a background of different proteins was reported in [229]. The antigen–antibody interaction of biotin and streptavidin or the binding of a herbicide to its monoclonal antibody with a bending cantilever as transducer is described in [201, 230]. This enables the label-free detection of an antigen–antibody interaction with a miniaturized micromechanical system. With cantilever sensors one can also enter the field of genomics. A first example of an array of cantilevers detecting DNA hybridization was presented in [231]. The hybridization of complementary oligonucleotides showed that a single mismatch between two 12-mer oligonucleotides is clearly detectable. The principle of this hybridization reaction is shown in Figure 9. These promising results were obtained in

Figure 9. Schematic illustration of a hybridization experiment. Each cantilever is functionalized on one side with a different 12-mer oligonucleotide base (red or blue) differing in only one base. (a) The differential deflection signal is set to zero. (b) After injection of the first complementary oligonucleotide (green), hybridization occurs on the cantilever providing the matching sequence (red) and increasing the differential signal )x. (c) Injection of the second complementary oligonucleotide (yellow) causes the cantilever functionalized with the second oligonucleotide (blue) to bend. Reprinted with permission from [231], J. Fritz et al., Science 288, 316 (2000). © 2000, American Association for the Advancement of Science.

[232] with a 10-mer oligonucleotide. It is possible to measure with cantilever sensors single nucleotide polymorphisms fast and easily.

5.3. Frequency Detection All of the previously mentioned examples of cantilever liquid sensors are based on the bending of the micromachined beam. The most important problem that has to be solved if the frequency change in liquids should be the measured value is the damping in liquids. As described in Section 3.1, in liquids high damping occurs with a decreasing resonance frequency and, even more important, a decreasing quality factor. Generally the quality factor decreases by 2 orders of magnitude if the surrounding medium is changed from air to water. Because of these effects it is nearly impossible to measure frequency changes below 10–20 Hz. The damping is caused by changes in the density and the viscosity in the surrounding medium and can be explained by an increase of the effective mass due to an added mass of the liquid being dragged with the cantilever. The resulting frequency change can be used for quantitative measurements of the density/viscosity of different liquids [74, 233]. In [146] the hybridization of an 8-mer DNA oligonucleotide is proposed to be measured with the frequency change of a resonating cantilever, but no data have been published. With a cantilever operated in the resonating mode in [234] the adsorption of bacteria was investigated. The number of bacteria adsorbed onto an antibody-coated surface could be weighed and counted. The measurements were not only measurements in the liquid phase, they were also performed on a dry cantilever in air to get a higher frequency resolution. In [235] cells were been immobilized, but measurement of the frequency change was not reported. The main problem in all these measurements is the high damping, not the biological preparation. A solution to this problem is to take advantage of the closed feedback loop described earlier. If it is possible to design a closed feedback loop that will work in liquids in an appropriate manner, the damping could be reduced and the quality factor enhanced. This was done in [190] to measure ethanol in water with a polymer-coated cantilever and the antigen–antibody interaction STAR71-BRAC30. In this measurement setup the cantilever is driven with a PLL-controlled oscillator connected to a feedback loop with a variable phase shifter and gain amplifier. The results obtained are a first step in the direction of measuring biological interactions in the liquid phase with resonating microcantilevers. Another promising result is described in [175]. In this setup a self-oscillating system is designed that needs no further actuation. This easy design enables the stable oscillation of cantilevers in the liquid environment. The versatile use of this system was proven with the quantitative measurement of the viscosity and density changes in different ethanol/water mixtures. The short-term frequency stability obtained is approximately 0.3 Hz and therefore an enhancement of the quality factor of 2 orders of magnitude was possible.

512

Figure 10. Optical microscope images of a sensor chip which consists of 10 piezoresistive cantilevers in a channel and 4 inlets/outlets. The piezoresistor, which serves as a readout mechanism to detect the bending of the cantilever, is fully encapsulated and can be used in liquids. Reprinted with permission from [227], J. Thaysen et al. “International Conference of Microelectromechanical Systems,” pp. 401–404, 2001. © 2001, IEEE.

6. SUMMARY AND FUTURE TRENDS IN MICROCANTILEVER SENSORS In this chapter it was shown that most of the fundamental problems in cantilever sensor technology are solved in the laboratory state. It is possible to fabricate small cantilevers with low force constants that are very sensitive to bending moments. On the other hand, stiff cantilevers can be produced with high frequencies that are very sensitive to mass changes or changes in the force constant. There are a large variety of different methods for measuring the movement of the cantilever, starting with optical techniques and ending with CMOS-manufactured piezoresistive cantilevers. A few different methods for coating the surface with different polymers for gas sensing purposes and with biological layers for biosensing purposes were introduced. First approaches to work with arrays of cantilevers for reference measurements and pattern recognition to improve signal quality were presented. In addition, some methods based on closed feedback loop operations for the enhancement of the quality factor in gaseous and liquid environment were shown. Most of these measurements are promising approaches to a completely new class of miniaturized sensor systems. However, under real conditions sensors have to be stable long-term with low power consumption in an extremely small housing for low cost. Ease of use and reproducibility are additional factors for an industrial sensor system. In biomedicine and biotechnology only small amounts of reagents are available and high sensitivity and high throughput screening are requirements. Furthermore, the preparation of high amounts of reproducible biofunctionalized surface modifications in an easy-to-use liquid handling system is necessary for biosensing applications. At present there are several approaches to solving these problems. In Figure 5 CMOS-processed cantilevers combined with subsequent CMOS-compatible micromachining

Cantilever-Based Sensors

were presented with the complete electronic circuitry for cantilever driving and signal conditioning [38, 195]. With this industrial technique it is easy to mass produce large and compact arrays of integrated cantilevers for the detection of gaseous analytes. Besides an integrated readout and driving system it is necessary to combine sensing systems with the requirements of biotechnology. One approach is an array of 10 cantilevers with a piezoresistive readout for deflection detection integrated into a microfluidic handling system onto a silicon chip as shown in Figure 10. These cantilevers are completely encapsulated and can be operated stably in liquid environments. Another development focuses on larger arrays. At the moment an array of 2 to 10 cantilevers is state of the art, but this is not enough for the requirements of the industry for extremely high parallelization. Large sensor arrays with 1024 cantilevers combined with MEMS technology are used in [34, 236, 237] as ultra-high density storage devices. Besides the piezoresistive readout of these large arrays, new optical readout techniques are being explored [37, 103, 165, 238] for parallel optical detection. One next step will be to go to smaller cantilevers with higher frequencies and therefore better mass resolution in the dynamic mode of operation. These approaches are presented in [24–31] for an optical detection system. Nanometer-sized cantilevers with frequencies in the megahertz range and integrated CMOS circuitry are the newest steps in the future direction of high resolution mass sensitivity [239, 240].

GLOSSARY Biosensor Often defined as a compact analytical device incorporating a biological or biologically derived sensing element either integrated within or intimately associated with a physicochemical transducer. Complementary metal oxide semiconductor (CMOS) Special semiconductor technology, standard process for most integrated circuits. Integrated circuit (IC) Semiconductor device (e.g., amplifier, oscillator, or microprocessor), in which all resistors, capacitors, and transistors which perform the desired function of the circuit are processed together. Multivariate data analysis Methods for dealing with large tables of data by displaying the data in a few meaningful plots that contain the most important information. Methods for displaying are, for example, principal component analysis for an overview over groups and trends, the classification or discriminant analysis for analyzing the groups, and the multiple regression analysis or partial least squares projection to build quantitative models for prediction of unknown analytes. Scanning force microscopy (SFM) A method to measure the interaction forces between two surfaces for getting a three-dimensional picture of the surface. Sensor A miniaturized device that converts a chemical state into an electrical signal. It mainly consists of three parts: the recognition site, the transducer, and the readout system.

Cantilever-Based Sensors

REFERENCES 1. G. Binnig, C. F. Quate, and C. Gerber, Phys. Rev. Lett. 56, 930 (1986). 2. H. Hosoi, M. Kimura, K. Hayakawa, K. Sueoka, and K. Mukasa, Appl. Phys. A 72, S23 (2001). 3. W. F. Heinz and J. H. Hoh, Biophys. J. 76, 528 (1999). 4. S. M. Smith and J. L. Gilbert, in “Compatability of Biomedical Implants, Proceedings of the Electrochemical Society,” 1994, pp. 229–240. 5. A. L. Weisenhorn, M. Khorsandi, S. Kasas, V. Gotzos, and H.-J. Butt, Nanotechnology 4, 106 (1993). 6. M. Maywald, R. J. Pylkki, and L. J. Balk, Scanning Microsc. 8, 181 (1994). 7. R. Ros, F. Schwesinger, C. Padeste, A. Plückthun, D. Anselmetti, H.-J. Güntherodt, and L. Tiefenauer, Proc. SPIE-Int. Soc. Opt. Eng. 3607, 84 (1999). 8. G. Haehner, A. Marti, and N. D. Spencer, Tribol. Lett. 3, 359 (1997). 9. Olympus Optical Co., Ltd., http://www.olympus.co.jp/probe/index. html (2002). 10. Nanosensors GmbH & Co. KG, http://www.nanosensors.com/ products.htm (2002). 11. NT-MDT Co., http://www.ntmdt.ru/ (2002). 12. MikroMasch, http://www.spmtips.com/ (2002). 13. NanoDevices Inc., http://www.nanodevices.com (2002). 14. TM Microscopes, http://www.spmprobes.com/ (2002). 15. M. Ono, D. Lange, O. Brand, C. Hagleitner, and H. Baltes, Ultramicroscopy 91, 9 (2002). 16. D. Lange, T. Akiyama, C. Hagleitner, A. Tonin, H. R. Hidber, P. Niedermann, U. Staufer, N. F. De Rooij, O. Brand, and H. Baltes, “Technical Digest of the 12th, IEEE International Conference on Micro Electro Mechanical Systems,” 1999, pp. 447–452. 17. C. Hagleitner, D. Lange, T. Akiyama, A. Tonin, R. Vogt, and H. Baltes, Proc. SPIE-Int. Soc. Opt. Eng. 3673, 240 (1999). 18. W. Franks, D. Lange, S. Lee, A. Hierlemann, N. Spencer, and H. Baltes, Ultramicroscopy 91, 21 (2002). 19. H. Takahashi, K. Ando, and Y. Shirakawabe, Ultramicroscopy 91, 63 (2002). 20. N. Satoh, K. Kobayashi, S. Watanabe, T. Fujii, T. Horiuchi, H. Yamada, and K. Matsushige, Appl. Surf. Sci. 188, 425 (2002). 21. I. W. Rangelow, P. Grabiec, T. Gotszalk, and K. Edinger, Surf. Interface Anal. 33, 59 (2002). 22. Y. Miyahara, M. Deschler, T. Fujii, S. Watanabe, and H. Bleuler, Appl. Surf. Sci. 188, 450 (2002). 23. J. Thaysen, A. Boisen, O. Hansen, and S. Bouwstra, Sens Actuators 83, 47 (2000). 24. T. E. Schaffer, J. Appl. Phys. 91, 4739 (2002). 25. A. Chand, M. B. Viani, T. E. Schaffer, and P. K. Hansma, J. Microelectromech. Syst. 9, 112 (2000). 26. T. E. Schäffer, M. Viani, D. A. Walters, B. Drake, E. K. Runge, J. P. Cleveland, M. A. Wendman, and P. K. Hansma, Proc. SPIEInt. Soc. Opt. Eng. 3009, 48 (1997). 27. M. B. Viani, T. E. Schäffer, A. Chand, M. Rief, H. E. Gaub, and P. K. Hansma, J. Appl. Phys. 86, 2258 (1999). 28. M. B. Viani, T. E. Schäffer, G. T. Paloczi, L. I. Pietrasanta, B. L. Smith, J. B. Thompson, M. Richter, M. Rief, H. E. Gaub, K. W. Plaxco, A. N. Cleland, H. G. Hansma, and P. K. Hansma, Rev. Sci. Instrum. 70, 4300 (1999). 29. D. A. Walters, M. Viani, G. T. Paloczi, T. E. Schäffer, J. P. Cleveland, M. A. Wendman, G. Gurley, V. Ellings, and P. K. Hansma, Proc. SPIE-Int. Soc. Opt. Eng. 3009, 43 (1998). 30. T. Bachels and R. Schäfer, Rev. Sci. Instrum. 69, 3794 (1998). 31. M. Hoummady and H. Fujita, Nanotechnology 10, 29 (1999). 32. E. M. Chow, G. G. Yaralioglu, C. F. Quate, and T. W. Kenny, Appl. Phys. Lett. 80, 664 (2002). 33. N. Abedinov, P. Grabiec, T. Gotszalk, T. Ivanov, J. Voigt, and I. W. Rangelow, J. Vac. Sci. Technol. A 19, 2884 (2001).

513 34. P. Vettiger, M. Despont, U. Drechsler, U. Dürig, W. Häberle, M. I. Lutwyche, H. E. Rothuizen, R. Sturz, R. Widmaer, and G. K. Binnig, IBM J. Res. Dev. 44, 323 (2000). 35. U. Durig, G. Cross, M. Despont, U. Drechsler, W. Haberle, M. I. Lutwyche, H. Rothuizen, R. Stutz, R. Widmer, P. Vettiger, G. K. Binnig, W. P. King, and K. E. Goodson, Tribol. Lett. 9, 25 (2000). 36. P. Vettiger, J. Brugger, M. Despont, U. Drechsler, U. Durig, W. Haberle, M. Lutwyche, H. Rothuizen, R. Stutz, R. Widmer, and G. Binnig, Microelectron. Eng. 46, 11 (1999). 37. Y. Zhao, M. Mao, R. Horowitz, A. Majumdar, J. Varesi, P. Norton, and J. Kitching, J. Microelectromech. Syst. 11, 136 (2002). 38. D. Lange, C. Hagleitner, A. Hierlemann, O. Brand, and H. Baltes, Anal. Chem. 74, 3084 (2002). 39. R. McKendry, J. Zhang, Y. Arntz, T. Strunz, M. Hegner, H. P. Lang, M. K. Baller, U. Certa, E. Meyer, H.-J. Guntherodt, and C. Gerber, Proc Nat. Acad. Sci. USA 99, 9783 (2002). 40. K. Nilsson and E. Nilsson, U.S. Patent, 6252335, 2001. 41. D. L. Polla, R. S. Muller, and R. M. White, IEEE Electron Device Lett. EDL-7, 254 (1986). 42. X. Li, M. Bao, H. Yang, S. Shen, and D. Lu, Sens. Actuators, A A72, 217 (1999). 43. Y. Su, A. G. R. Evans, A. Brunnschweiler, and G. Ensell, “ESSDERC ’96, Proceedings of the 26th European Solid State Device Research Conference,” 1996, pp. 717–720. 44. T. Matsuura, M. Taguchi, K. Kawata, and K. Tsutsumi, Sens. Actuators, A A60, 197 (1997). 45. W. Göpel and K. Schierbaum, in “Sensors: A Comprehensive Survey” (W. Göpel, J. Hesse, and J. N. Zemel, Eds.), Vol 2, pp. 1–27. VCH, Weinheim, 1991. 46. M. Kaspar, H. Stadler, T. Weiss, and C. Ziegler, Fresenius J. Anal. Chem. 366, 602 (2000). 47. T. C. Pearce, J. W. Gardener, and W. Göpel, in “Sensors Update” (W. Göpel, J. Hesse, and H. Baltes, Eds.), Vol. 3, pp. 61–130. Wiley-VCH, Weinheim, 1998. 48. Cardiff University, Electronic ‘tongue’ for environmental monitoring, http://www.cardiff.ac.uk/news/releases/0204/020404.html (2002). 49. A. P. F. Turner, I. Karube, and G. S. Wilson, “Biosensors: Fundamentals and Application” p. 770, Oxford University Press, Oxford, 1987. 50. H. Baltes, A. Koll, and D. Lange, “Proceedings of the IEEE, ISIE ’97,” 1997, pp. SS152–SS157. 51. W. Weaver, S. P. Timoschenko, and D. H. Young, “Vibration Problems in Engineering, Wiley, New York, 1990. 52. O. Brand and H. Baltes, in “Sensors Update” (H. Baltes, W. Göpel, and J. Hesse, Eds.), Vol. 4, pp. 3–51. VCH, Weinheim, 1998. 53. H.-J. Butt and M. Jaschke, Nanotechnology 6, 1 (1995). 54. R. D. Blevin, “Formulas for Natural Frequencies and Mode Shapes,” Van Nostrand Reinhold, New York, 1979. 55. J. M. Neumeister and W. A. Ducker, Rev. Sci. Instrum. 65, 2527 (1994). 56. J. E. Sader, Rev. Sci. Instrum. 66, 4583 (1995). 57. J. E. Sader, I. Larson, P. Mulvaney, and L. R. White, Rev. Sci. Instrum. 66, 3789 (1995). 58. J. E. Sader, J. W. M. Chon, and P. Mulvaney, Rev. Sci. Instrum. 70, 3967 (1999). 59. J. E. Sader and L. White, J. Appl. Phys. 74, 1 (1994). 60. R. Lévy and M. Maaloum, Nanotechnology 13, 33 (2002). 61. J. D. Holbery, V. L. Eden, M. Sarikaya, and R. M. Fisher, Rev. Sci. Instrum. 71, 3769 (2000). 62. H.-J. Butt, P. Siedle, K. Seifert, K. Fendler, T. Seeger, A. Bamberg, A. L. Weisenhorn, K. Goldie, and A. Engel, J. Microsc. 169, 75 (1993). 63. T. J. Senden and W. A. Ducker, Langmuir 10, 1003 (1994). 64. E.-L. Florin, V. T. Moy, and H. E. Gaub, Science (Washington, DC) 264, 415 (1994).

514 65. Y. Q. Li, N. J. Tao, J. Pan, A. A. Garcia, and S. M. Lindsay, Langmuir 9, 637 (1993). 66. J. L. Hutter and J. Bechhoefer, Rev. Sci. Instrum. 64, 1868 (1993). 67. J. P. Cleveland, S. Manne, D. Bocek, and P. K. Hansma, Rev. Sci. Instrum. 64, 403 (1993). 68. Asylum Research, http://www.asylumresearch.com/QuantMolec ForcesAbstract.pdf (2002). 69. T. R. Albrecht, P. Grütter, D. Horne, and D. Rugar, J. Appl. Phys. 69, 668 (1991). 70. D. Lange, “Cantilever-Based Microsystems for Gas Sensing and Atomic Force Microscopy,” Dissertation, Swiss Federal Institute of Technology, Zurich, 2000. 71. J. Mehner, S. Kurth, D. Billep, C. Kaufmann, K. Kehr, and W. Dötzel, “Proceedings of IEEE Micro Electro Mechanical Systems (MEMS),” 1998, pp. 240–245. 72. K. Kokobun, M. Hirata, M. Ono, H. Murakami, and Y. Toda, J. Vac. Sci. Technol. A5, 2450 (1987). 73. L. D. Landau and E. M. Lifshitz, “Fluid Mechanics,” New York, Pergamon Press, 1959. 74. G. Y. Chen, R. J. Warmack, T. Thundat, D. P. Allison, and A. Huang, Rev. Sci. Instrum. 65, 2532 (1994). 75. F. R. Blom, S. Bouwstra, M. Elwenspoek, and J. H. J. Fluitman, J Vac Sci. Technol. B 10, 19 (1992). 76. J. E. Sader, J. Appl. Phys. 84, 64 (1998). 77. W. Y. Shih, X. Li, H. Gu, W.-H. Shih, and I. A. Aksay, J. Appl. Phys. 89, 1497 (2001). 78. S. Rast, C. Wattinger, U. Gysin, and E. Meyer, Rev. Sci. Instrum. 71, 2772 (2000). 79. B. Anczykowski, J. P. Cleveland, D. Krüger, V. Ellings, and H. Fuchs, Appl. Phys. A 66, S885 (1998). 80. T. Sulchek, R. Hsieh, J. D. Adams, G. G. Yaralioglu, S. C. Minne, C. F. Quate, J. P. Cleveland, A. Atalar, and D. M. Adderton, Appl. Phys. Lett. 76, 14731 (2000). 81. J. Tamayo, A. D. L. Humphris, and M. J. Miles, Appl. Phys. Lett. 77, 582 (2000). 82. G. Sauerbrey, Z. Phys. 155, 206 (1959). 83. F. Eichelbaum, R. Borngräber, R. Lucklum, P. Hauptmann, and S. Rösler, Technisches Messen 65, 434 (1998). 84. G. Y. Chen, T. Thundat, E. A. Wachter, and R. J. Warmack, J. Appl. Phys. 77, 3618 (1995). 85. G. G. Stoney. “Proceedings of the Royal Society of London A82,” 1909, p. 172. 86. T. Honda, K. I. Arai, and M. Yamaguchi, J. Appl. Phys. 76, 6994 (1994). 87. Z. Hu, T. Thundat, and R. J. Warmack, J. Appl. Phys. 90, 427 (2001). 88. J. R. Barnes, R. J. Stephenson, C. N. Woodburn, S. J. O´Shea, M. E. Welland, T. Rayment, J. K. Gimzewski, and C. Geber, Rev. Sci. Instrum. 65, 3793 (1994). 89. R. Berger, C. Gerber, H. P. Lang, and J. K. Gimzewski, Microelectron. Eng. 35, 373 (1997). 90. G. Meyer and N. M. Amer, Appl. Phys. Lett. 53, 1045 (1988). 91. R. Erlandsson, G. M. McClelland, C. M. Mate, and S. Chiang, J. Vac. Sci. Technol., A 6, 266 (1988). 92. C. A. J. Putman, B. G. D. Grooth, N. F. V. Hulst, and J. Greve, J. Appl. Phys. 72, 6 (1992). 93. Y. Martin, C. C. Williams, and H. K. Wickramasinghe, J. Appl. Phys. 61, 4723 (1987). 94. D. Rugar, H. J. Mamin, and P. Guethner, Appl. Phys. Lett. 55, 2588 (1989). 95. N. Blanc, J. Brugger, N. F. d. Rooij, and U. Dürig, J. Vac. Sci. Technol., B 14, 901 (1996). 96. Y. Shiba, T. Ono, K. Minami, and M. Esashi, Trans. Ins. Electr. Eng. Jpn. 118-E, 647 (1998). 97. J. Brugger, N. Blanc, P. Renaud, and N. F. d. Rooij, Sens. Actuators, A 43, 339 (1994). 98. R. Raiteri, M. Grattarola, and R. Berger, Mater. Today 22 (2002).

Cantilever-Based Sensors 99. G. G. Yaralioglu, A. Atalar, S. R. Manalis, and C. F. Quate, J. Appl. Phys. 83, 7405 (1998). 100. S. R. Manalis, S. C. Minne, and A. Atalar, Appl. Phys. Lett. 69, 3944 (1996). 101. T. Perazzo, M. Mao, O. Kwon, A. Majumdar, J. B. Varesi, and P. Norton, Appl. Phys. Lett. 74, 3567 (1999). 102. S. R. Manalis, S. C. Minne, C. F. Quate, G. G. Yaralioglu, and A. Atalar, Appl. Phys. Lett. 70, 3311 (1997). 103. T. Thundat, E. Finot, Z. Hu, R. H. Ritchie, G. Wu, and A. Majumdar, Appl. Phys. Lett. 77, 4061 (2000). 104. M. Tortonese, R. C. Barrett, and C. F. Quate, Appl. Phys. Lett. 62, 834 (1993). 105. R. Linnemann, T. Gotszalk, L. Hadjiiski, and I. W. Rangelow, Thin Solid Films 264, 159 (1995). 106. S. C. Minne, S. R. Manalis, and C. F. Quate, Appl. Phys. Lett. 67, 3918 (1995). 107. B. W. Chui, T. D. Stone, T. W. Kenny, H. J. Mamin, B. D. Terris, and D. Rugar, Appl. Phys. Lett. 69, 2767 (1996). 108. O. Hansen and A. Boisen, Nanotechnology 10, 51 (1999). 109. S. S. Lee and R. M. White, Sens. Actuators, A 52, 41 (1996). 110. S. R. Manalis, S. C. Minne, A. Atalar, and C. F. Quate, Rev. Sci. Instrum. 67, 3294 (1996). 111. P.-F. Indermühle, G. Schürmann, G.-A. Racine, and N. F. d. Rooij, J. Micromech. Microeng. 7, 218 (1997). 112. C. Lee, T. Itoh, T. Ohashi, R. Maeda, and T. Suga, J. Vac. Sci. Technol., B 15, 1559 (1997). 113. W. Han, S. M. Lindsay, and T. Jing, Appl. Phys. Lett. 69, 4111 (1996). 114. M. A. Lantz, S. J. O’Shea, and M. E. Welland, Appl. Phys. Lett. 65, 409 (1994). 115. E.-L. Florin, M. Radmacher, B. Fleck, and H. E. Gaub, Rev. Sci. Instrum. 65, 639 (1994). 116. A. Schemmel and H. E. Gaub, Rev. Sci. Instrum. 70, 1313 (1999). 117. I. Revenko and R. Proksch, J. Appl. Phys. 87, 526 (2000). 118. E. Sarraute and I. Dufour, “Proceedings of the 1999 International Conference on Modeling and Simulation of Microsystems,” 1999, pp. 266–269. 119. M. Dufour, M. T. Delaye, F. Michel, J. S. Danel, B. Diem, and G. Delapierre, Sens. Actuators, A 34, 201 (1992). 120. A. Buguin, O. D. Roure, and P. Silberzan, Appl. Phys. Lett. 78, 2982 (2001). 121. D. Lange and O. Brand, European Patent 1197726, 2002. 122. S. Tsuchitani and R. Kaneko, Trans. Inst. Electr. Eng. Jpn. 118-E, 6 (1998). 123. D. F. Cunningham, L. Jenkins, and M. A. H. Khalid, Sens. Actuators, A 63, 125 (1997). 124. P. Attia, M. Boutry, A. Bosseboeuf, and P. Hesto, Microelectronics J. 29, 641 (1998). 125. G. Abadal, Z. J. Davis, B. Helbo, X. Borrise, R. Ruiz, A. Boisen, F. Campabadal, J. Esteve, E. Figueras, F. Perez-Murano, and N. Barniol, Nanotechnology 12, 100 (2001). 126. T. Akiyama, U. Staufer, N. F. de Rooij, D. Lange, C. Hagleitner, O. Brand, H. Baltes, A. Tonin, and H. R. Hidber, J. Vac. Sci. Technol., B 18, 2669 (2000). 127. J. K. Gimzewski, C. Gerber, E. Meyer, and R. R. Schlittler, Chem. Physi. Lett. 217, 589 (1994). 128. T. Thundat, R. J. Warmack, G. Y. Chen, and D. P. Allison, Appl. Phys. Lett. 64, 2894 (1994). 129. R. Berger, C. Gerber, J. K. Gimzewski, E. Meyer, and H.-J. Güntherodt, Appl. Phys. Lett. 69, 40 (1996). 130. R. Berger, C. Gerber, and J. K. Gimzewski, “Symposium on Microscale Thermal Phenomena in Electronic Systems,” 1996. 131. R. Berger, C. Gerber, and J. K. Gimzewski, “2nd International Symposium on Micro Total Analysis Systems,” 1996. 132. T. G. Thundat and M. J. Doktycz, World Patent 9947911, 1999. 133. J. R. Barnes, R. J. Stephenson, M. E. Welland, C. Gerber, and J. K. Gimzewski, Nature (London) 372, 79 (1994).

Cantilever-Based Sensors 134. G. Li, L. W. Burggraf, and W. P. Baker, Appl. Phys. Lett. 76, 1122 (2000). 135. T. Thundat, S. L. Sharp, W. G. Fisher, R. J. Warmack, and E. A. Wachter, Appl. Phys. Lett. 66, 1563 (1995). 136. E. A. Wachter, T. Thundat, P. I. Oden, R. J. Warmack, P. G. Datskos, and S. L. Sharp, Rev. Sci. Instrum. 67, 3434 (1996). 137. J. Varesi, J. Lai, T. Perazzo, Z. Shi, and A. Majumdar, Appl. Phys. Lett. 71, 306 (1997). 138. P. I. Oden, P. G. Datskos, T. Thundat, and R. J. Warmack, Appl. Phys. Lett. 69, 3277 (1996). 139. P. G. Datskos, P. I. Oden, T. Thundat, E. A. Wachter, R. J. Warmack, and S. R. Hunter, Appl. Phys. Lett. 69, 2986 (1996). 140. E. Peiner, D. Scholz, A. Schlachetzki, and P. Hauptmann, Sens. Actuators, A A65, 23 (1998). 141. R. Berger, E. Delamarche, H. P. Lang, C. Gerber, J. K. Gimzewski, E. Meyer, and H.-J. Güntherodt, Appl. Phys. A 66, S55 (1998). 142. R. Berger, H. P. Lang, E. Delamarche, C. Gerber, J. K. Gimzewski, C. Andreoli, J. Brugger, M. Despont, and P. Vettiger, in “8th International Conference on Sensors and Their Applications,” 1997. 143. R. Berger, E. Delamarche, H. P. Lang, C. Gerber, and J. K. Gimzewski, Science (Washington, DC) 276, 2021 (1997). 144. P. G. Datskos and I. Sauers, Sens. Actuators, B B61, 75 (1999). 145. D. Sander, A. Enders, and J. Kirschner, Rev. Sci. Instrum. 66, 4734 (1995). 146. T. Thundat, P. I. Oden, and R. J. Warmack, Microscale Thermophys. Eng. 1, 185 (1997). 147. A. C. Stephan, T. Gaulden, A. D. Brown, M. Smith, L. F. Miller, and T. Thundat, Rev. Sci. Instrum. 73, 36 (2002). 148. E. A. Wachter and T. Thundat, Rev. Sci. Instrum. 66, 3662 (1995). 149. H. Jensenius, J. Thaysen, A. A. Rasmussen, L. H. Veje, O. Hansen, and A. Boisen, Appl. Phys. Lett. 76, 2615 (2000). 150. A. Boisen, J. Thaysen, H. Jensenius, and O. Hansen, Ultramicroscopy 82, 11 (2000). 151. R. Berger, H. P. Lang, J. P. Ramseyer, F. Battiston, J. H. Fabian, L. Scandella, C. Andreoli, J. Brugger, M. Despont, P. Vettiger, E. Meyer, H.-J. Güntherodt, C. Gerber, and J. K. Gimzewski, Research Report IBM Research Division RZ 2986, 1997, pp. 1–9. 152. H. P. Lang, R. Berger, F. Battiston, J.-P. Ramseyer, E. Meyer, C. Andreoli, J. Brugger, P. Vettinger, M. Despont, T. Mezzacasa, L. Sandella, H.-J. Gütherodt, C. Gerber, and J. K. Gimzewski, Appl. Phys. A 66, S61 (1998). 153. M. K. Baller, H. P. Lang, J. Fritz, C. Gerber, J. K. Gimzewski, U. Drechsler, H. Rothuizen, M. Despont, P. Vettiger, F. M. Battiston, J. P. Ramseyer, P. Fornaro, E. Meyer, and H.-J. Güntherodt, Ultramicroscopy 82, 1 (2000). 154. F. M. Battiston, J.-P. Ramseyer, H. P. Lang, M. K. Baller, C. Gerber, J. K. Gimzewski, E. Meyer, and H.-J. Güntherodt, Sens. Actuators, B B77, 122 (2001). 155. H. P. Lang, M. K. Baller, R. Berger, C. Gerber, J. K. Gimzewski, F. M. Battiston, P. Fornaro, J. P. Ramseyer, E. Meyer, and H.-J. Güntherodt, Research Report IBM Research Division, 1998, pp. 1–7. 156. H. P. Lang, R. Berger, C. Andreoli, J. Brugger, M. Despont, P. Vettinger, C. Gerber, J. K. Gimzewski, J. P. Ramseyer, E. Meyer, and H.-J. Güntherodt, Appl. Phys. Lett. 72, 383 (1998). 157. C. L. Britton, R. L. Jones, P. I. Oden, Z. Hu, R. J. Warmack, S. F. Smith, W. L. Bryan, and J. M. Rochelle, Ultramicroscopy 82, 17 (2000). 158. C. L. Britton, Jr., R. J. Warmack, W. L. Bryan, R. L. Jones, P. I. Oden, and T. Thundat, U.S. Patent 6167748, 2001. 159. T. G. Thundat and E. A. Wachter, U.S. Patent 5719324, 1998. 160. P. G. Datskos, S. Rajic, L. R. Senesac, and I. Datskou, Ultramicroscopy 86, 191 (2001). 161. T. G. Thundat, R. J. Warmack, and E. A. Wachter, World Patent 9726556, 1997.

515 162. Y. Zhao, J. Yamaguchi, J. Choi, S. Morales, R. Horowitz, A. Majumdar, P. Norton, J. Kitching, H. Li, and M. Athavale, in 369-1, “Proceedings of the ASME Heat Transfer Division— 2001,” Vol. 1, 2001, pp. 367–370. 163. M. Mao, T. Perazzo, O. Kwon, Y. Zhao, A. Majumdar, J. Varesi, and P. Norton, Microelectromech. Syst. 1, 309 (1999). 164. M. Mao, T. Perazzo, O. Kwon, A. Majumdar, J. Varesi, and P. Norton, “Technical Digest of the 12th IEEE International Conference on Micro Electro Mechanical Systems,” 1999, pp. 100–105. 165. T. Perazzo, M. Mao, O. Kwon, A. Majumdar, J. B. Varesi, and P. Norton, Appl. Phys. Lett. 74, 3567 (1999). 166. I. W. Rangelow, T. Gotszalk, N. Abedinov, P. Grabiec, and K. Edinger, Microelectron. Eng. 57–58, 737 (2001). 167. P. G. Datskos, S. Rajic, M. J. Sepaniak, N. Lavrik, C. A. Tipple, L. R. Senesac, and I. Datskou, J. Vac. Sci. Technol., B 19, 1173 (2001). 168. P. G. Datskos, World Patent 0058688, 2000. 169. P. G. Datskos, M. J. Sepaniak, C. A. Tipple, and N. Lavrik, Sens. Actuators, B 76, 393 (2001). 170. T. G. Thundat, U.S. Patent 6212939, 2001. 171. T. L. Porter, M. P. Eastman, D. L. Pace, and M. Bradley, Sens. Actuators, A 88, 47 (2001). 172. J. Rickert, G. L. Hayward, B. A. Ravi, M. Thompson, and W. Göpel, “Sensors Update,” VCH, Weinheim, 1998. 173. K. Bodenhöfer, A. Hierlemann, G. Noetzel, U. Weimar, and W. Göpel, “Proceedings of Transducers ’95,” 1995, pp. 728–731. 174. M. Haug, K. D. Schierbaum, G. Gauglitz, and W. Göpel, Sens. Actuators, B 11, 383 (1993). 175. A. Vidic, D. Then, and C. Ziegler, Ultramicroscopy, accepted, 2002. 176. J. M. O’Connor and J. C. Patton, European Patent 72744, 1983. 177. S. Prescesky, M. Parameswaran, A. Rawicz, R. F. B. Turner, and U. Reichl, Can. J. Phys. 70, 1178 (1992). 178. T. Thundat, G. Y. Chen, R. J. Warmack, D. P. Allison, and E. A. Wachter, Anal. Chem. 67, 519 (1995). 179. T. Thundat, E. A. Wachter, S. L. Sharp, and R. J. Warmack, Appl. Phys. Lett. 66, 1695 (1995). 180. L. Scandella, G. Binder, T. Mezzacasa, J. Gobracht, R. Berger, H. P. Lang, C. Gerber, J. K. Gimzewski, J. H. Koegler, and J. C. Jansen, Research Report IBM Research Division, 1997, pp. 1–8. 181. J.-H. Fabian, J. Gobrecht, L. Scandella, R. Berger, H. P. Lang, C. Gerber, J. K. Gimzewski, and E. Meyer, Research Report IBM Research Division, 1998, pp. 1–4. 182. L. Scandella, G. Binder, T. Mezzacasa, J. Gobrecht, R. Berger, H. P. Lang, C. Gerber, J. K. Gimzewski, J. H. Koegler, and J. C. Jansen, Microporous Mesoporous Mater. 21, 403 (1998). 183. K. Bodenhöfer, Medien Verlag Köhler, Tübingen, 1997. 184. H. P. Lang, F. M. Battiston, M. K. Baller, R. Berger, J.-P. Ramseyer, P. Fornaro, E. Meyer, H.-J. Güntherodt, C. Andreoli, J. Brugger, M. Despont, P. Vettiger, J. H. Fabian, T. Mezzacasa, L. Scandella, C. Gerber, and J. K. Gimzewski, Research Report, p. 4. 185. M. Maute, S. Raible, F. E. Prins, D. P. Kern, H. Ulmer, U. Weimar, and W. Göpel, Sens. Actuators, B 58, 505 (1999). 186. M. Maute, S. Raible, F. E. Prins, D. P. Kern, U. Weimar, and W. Göpel, Microelectron. Eng. 46, 439 (1999). 187. B. H. Kim, M. Maute, F. E. Prins, D. P. Kern, M. Croitoru, S. Raible, U. Weimar, and W. Göpel, Microelectron. Eng. 53, 229 (2000). 188. B. H. Kim, F. E. Prins, D. P. Kern, S. Raible, and U. Weimar, Sens. Actuators, B 3944, 1 (2001). 189. B. H. Kim, D. P. Kern, S. Raible, and U. Weimar, Microelectron. Eng. 61–62, 947 (2002). 190. J. Tamayo, A. D. L. Humphris, A. M. Malloy, and M. J. Miles, Ultramicroscopy 86, 167 (2001). 191. D. Lange, C. Hagleitner, O. Brand, and H. Baltes, “IEEE Micro Electro Mechanical Systems 14th International Conference,” 2001, pp. 547–552.

516 192. C. Hagleitner, A. Hierlemann, D. Lange, A. Kummer, N. Kerness, O. Brand, and H. Baltes, Nature (London) 414, 293 (2001). 193. A. Hierlemann, D. Lange, C. Hagleitner, N. Kerness, A. Koll, O. Brand, and H. Baltes, Sens. Actuators, B 70, 2 (2000). 194. A. Roncaglia, L. Colalongo, D. Lange, and M. Rudan, Sens. Actuators, B 69, 320 (2000). 195. C. Hagleitner, A. Hierlemann, D. Lange, A. Kummer, N. Kerness, O. Brane, and H. Baltes, Nature (London) 414, 293 (2001). 196. D. Then, Miniaturisierte massensensitive Sensoren und deren Anwendung in der Gas- und Flüssigkeitsanalytik, Dissertation 2002. 197. J. Piehler, A. Brecht, K. E. Geckeler, and G. Gauglitz, Bios. Bioelectron. 11, 579 (1996). 198. M. Mrksich and G. Whitesides, Biosens. Bioelectron. 14, 247 (1996). 199. B. Ratner, Biosens. Bioelectron. 10, 797 (1995). 200. N. Alcantar, E. Aydil, and J. Israelachvili, J. Biomedi. Mater. Res. 51, 343 (2000). 201. R. Raiteri, M. Gratturola, H.-J. Butt, and P. Skládal, Sens. Actuators, B 79, 115 (2001). 202. V. T. Moy, E.-L. Florin, and H. E. Gaub, Science (Washington, DC) 266, 257 (1994). 203. R. Krapf, Dissertation, Universität Tübingen, 2001. 204. K. Feldman, G. Hähner, N. D. Spencer, P. Harder, and M. Grunze, J. Am. Chem. Soc. 121, 10134 (1999). 205. K. L. Prime and G. M. Whitesides, J. Am. Chem. Soc. 115, 10714 (1993). 206. M. C. Pretty, Langmuir-Blodgett Films. Cambridge University Press, Cambridge, UK, 1996. 207. B. C. Fagan, C. A. Tipple, Z. Xue, M. J. Sepaniak, and P. G. Datskos, Talanta 53, 599 (2000). 208. C. J. Brinker and G. W. Scherer, Sol-Gel Science: The Physics and Chemistry of Sol-Gel Processing. Academic Press, San Diego, 1990. 209. R. Raiteri and H.-J. Butt, J. Phys. Chem. 99, 15728 (1995). 210. S. J. O´Shea, M. E. Welland, T. A. Brunt, A. R. Ramadan, and T. Rayment, J. Vac. Sci. Technol., B 14, 1383 (1996). 211. M. Lahav, C. Durkan, R. Gabai, E. Katz, I. Willner, and M. E. Welland, Angew. Chem. 113, 4219 (2001). 212. H.-J. Butt, J. Colloid Interface Sci. 180, 251 (1996). 213. J. Fritz, M. K. Baller, H. P. Lang, T. Strunz, E. Meyer, H.-J. Güntherodt, E. Delamarche, C. Gerber, and J. K. Gimzewski, Langmuir 16, 9694 (2000). 214. H.-F. Ji, K. M. Hansen, Z. Hu, and T. Thundat, Sens. Actuators, B 72, 233 (2001). 215. X. Xu, T. G. Thundat, G. M. Brown, and H.-F. Ji, Anal. Chem. 74, 3611 (2002). 216. H.-F. Ji, E. Finot, R. Dabestani, T. Thundat, G. M. Brown, and P. F. Britt, Chem. Commun. 457 (2000). 217. H.-F. Ji and T. Thundat, Biosens. Bioelectron. 17, 337 (2002). 218. S. Cherian and T. Thundat, Appl. Phys. Lett. 80, 2219 (2002).

Cantilever-Based Sensors 219. C. A. Tipple, N. V. Lavrik, M. Culha, J. Headrick, P. Datskos, and M. J. Sepaniak, Anal. Chem. 74, 3118 (2002). 220. D. R. Baselt, G. U. Lee, and R. J. Colton, J. Vac. Sci. Technol., B14, 789 (1996). 221. D. R. Baselt, G. U. Lee, K. M. Hansen, L. A. Chrisey, and R. J. Colton, Proc. IEEE 85, 672 (1997). 222. D. R. Baselt, G. U. Lee, M. Natesan, S. W. Metzger, P. E. Sheenhan, and R. J. Colton, Biosens. Bioelectron. 13, 731 (1998). 223. M. D. Antonik, N. P. D’Costa, and J. H. Hoh, IEEE Eng. Med. Biol. 16, 66 (1997). 224. J. Fritz, M. K. Baller, H. P. Lang, T. Strunz, E. Meyer, H.-J. Güntherodt, E. Delamarche, C. Gerber, and J. K. Gimzewski, Langmuir 16, 9694 (2000). 225. A. M. Moulin, S. J. O’Shea, and M. E. Welland, Ultramicroscopy 82, 23 (2000). 226. A. M. Moulin, S. J. O’Shea, R. A. Badley, P. Doyle, and M. E. Welland, Langmuir 15, 8776 (1999). 227. J. Thaysen, R. Marie, and A. Boisen, “International Conference of Microelectromechanical Systems,” 2001, pp. 401–404. 228. R. Marie, H. Jensenius, J. Thaysen, C. B. Christensen, and A. Boisen, Ultramicroscopy 91, 29 (2002). 229. G. Wu, R. H. Datar, K. M. Hansen, T. Thundat, R. J. Cote, and A. Majumdar, Nat. Biotechnol. 19, 856 (2001). 230. R. Raiteri, G. Nelles, H.-J. Butt, W. Knoll, and P. Skládal, Sens. Actuators, B 61, 213 (1999). 231. J. Fritz, M. K. Baller, H. P. Lang, H. Rothuizen, P. Vettiger, E. Meyer, H.-J. Güntherodt, C. Gerber, and J. K. Gimzewski, Science (Washington, DC) 288, 316 (2000). 232. K. M. Hansen, H.-F. Ji, G. Wu, R. Datar, R. Cote, A. Majumdar, and T. Thundat, Anal. Chem. 73, 1567 (2001). 233. S. Weigert, M. Dreier, and M. Hegner, Appl. Phys. Lett. 69, 2834 (1996). 234. B. Ilic, D. Czaplewski, H. G. Craighead, P. Neuzil, C. Campagnolo, and C. Batt, Appl. Phys. Lett. 77, 450 (2000). 235. S. Prescesky, M. Parameswaran, A. Rawicz, R. F. B. Turner, and U. Reichl, Can. J. Phys. 70, 1178 (1992). 236. M. Lutwyche, C. Andreoli, G. Binnig, J. Brugger, U. Drechsler, W. Haberle, H. Rohrer, H. Rothuizen, P. Vettiger, G. Yaralioglu, and C. Quate, Sens. Actuators, A 73, 89 (1999). 237. G. Binnig, M. Despont, U. Drechsler, W. Haberle, M. Lutwyche, P. Vettiger, H. J. Mamin, B. W. Chui, and T. W. Kenny, Appl. Phys. Lett. 74, 1329 (1999). 238. J. P. Roger, A. C. Boccara, M.-C. Potier, M. Guirardel, and C. Bergaud, Proc. SPIE 4434, 138 (2001). 239. G. Abadal, Z. J. Davis, B. Helbo, X. Borrise, R. Ruiz, A. Boisen, F. Campabadal, J. Esteve, E. Figureras, F. Perez-Murano, and N. Barniol, Nanotechnology 12, 1 (2001). 240. Z. J. Davis, G. Abadal, O. Kuhn, O. Hansen, F. Grey, and A. Boisen, J. Vac. Sci. Technol., B 17, 612 (2000).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Carbon Nanomaterials Maheshwar Sharon Indian Institute of Technology, Mumbai, India

CONTENTS 1. Introduction 2. Classification of Carbon 3. Diamond-Type Carbon with Pure sp3 Configured Carbon Atoms 4. Graphite-Type Carbon with Pure sp2 Carbon Configured Atoms 5. Intermediate Forms of Carbon Containing Mixture of sp2 and sp3 Configured Carbon Atoms 6. Need for Precursors not Derived from Fossil Fuels 7. Summary Glossary References

1. INTRODUCTION The element carbon occupies a special place in nature. The whole living world is built of carbon, which makes a huge number of (organic) compounds. This is possible because of the electronic structure of the carbon atom, resulting in different chemical bonds. Though carbon belongs to the same group as silicon, thanks to the carbon bonding degeneracy, unlike silicon, carbon can exist in various allotropic forms, starting from sp2 bonded graphite, across a whole variety of carbon materials including glassy carbon, amorphous sp2 bonded carbon, polymeric materials, fullerenes, nanotubes, amorphous diamond-like carbon with majority of sp3 bounded atoms, to diamond with strong sp3 tetrahedral bonds. Carbon has been accepted by the power source industry because of its unique properties of good electronic conductivity, good intercalation capability, chemical inertness, low cost, wide availability, ease of fabrication, and availability in many different forms. One of the earliest uses of carbon in a commercial battery was in 1841, when Robert Bunsen proposed replacing the expensive platinum current collector of a Grove cell with a more cost-effective carbon one. ISBN: 1-58883-057-8/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

Nowadays the unique properties of carbon make it a preferred constituent of virtually all power source technologies. Better performance of carbon in lithium battery is achieved using Chevron acetylene black, which has a surface area of 40 m2 g−1 , instead of graphite (5 m2 g−1 ). High-surface-area carbons, such as Ketjenblack (1000 m2 g−1 ) or Black pearl 2000 (2000 m2 g−1 ), could further improve its performance. Various forms of carbon have been considered as hydrogen storage media, such as carbon nanotubes or carbon nanofibers [1]. If the high storage capacity of nanofibers can be realized in a practical system, it could revolutionize hydrogen storage for fuel cell application. Carbon in supercapacitors is also being developed as a rechargeable power source for electric vehicles and consumer electronics. Carbon is one of the most common electrode materials, either in powdered form or as woven cloths with surface areas in the range of 1000 to 2000 m2 g−1 and pore size from 5 nm. Intense research is being conducted worldwide into development of carbon nanomaterials for lithium battery technology, hydrogen storage, and fuel cells. Carbon has also been gaining importance for its ability to induce different properties and to form composite materials. The term composite materials is often taken to mean continuous fiber-reinforced polymer-matrix composites. This family includes glass fiber-reinforced polyester resins, used for boats and large tanks, and carbon fiber-reinforced epoxy resins, used on many aircraft. They offer a very good combination of high stiffness and low density. Another form of commercial carbon is “whiskers,” which are fibrous single crystal and which have been produced commercially since 1962. Relatively inexpensive silicon carbide whiskers can be formed by the pyrolysis of rice husk. The husk of rice is rich in silica. During pyrolysis the cellulose dehydrates to carbon; this reduces the silica to silica monoxide. Silicon carbide whiskers are apparently formed by the interaction of silicon monoxide vapor with carbon [2]. Prior to 1985, it was known that two allotropic forms of carbon exist: diamond and graphite. However, after the discovery of the fullerene family by Kroto et al. [3], and their microscopic preparation by Kratchmer et al. [4], it has become well known that carbon has three allotropic forms, namely diamond, graphite, and the fullerene family. But is it really so? In addition to these forms, we have many other

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 1: Pages (517–546)

Carbon Nanomaterials

518

2. CLASSIFICATION OF CARBON Conventionally carbon has been classified in two forms: diamond and graphite. Has this type of classification given us any help in searching for newer allotropes of carbon? After the discovery of the fullerenes family, a third allotrope of carbon, “fullerenes,” was adopted. Hence there is a need to reconsider the classification procedure for carbon to evolve a classification which can help to discover newer carbon materials. A classification of carbon based on the nature of carbon atoms present in the carbonaceous material perhaps may be more appropriate and may open a Pandora’s box for developing many new varieties of carbon than what we have been able to even think of. For example, diamond has atoms with 100% sp3 type of carbon, while graphite has 100% sp2 carbon atoms. The former is an insulator due to its very large bandgap (5.5 eV) while the latter is a conductor (bandgap being in vicinity of 0.25 eV or low). These two examples immediately give us a clue that there is a possibility of millions of permutations and combinations of sp3 and sp2 , giving an innumerable variety of carbon with varied bandgaps in the range of 0.25 to 5.5 eV. Diamond-like carbon (DLC) and glassy carbon are two examples of such combinations. Neither of them can be classified as diamond or graphite. Therefore, a classification of carbon on the basis of ratio of sp3 and sp2 carbon atoms present in the material may be more appropriate than the conventional classification. We can thus classify carbon again in three forms: (i) diamond (100% sp3 ), (ii) intermediate carbon (with different ratios of sp3 /sp2 carbon atoms), and (iii) graphite (100% sp2 ). All forms of carbons will fall in one of these three categories. This classification also gives us an opportunity

to understand the reasons for carbon showing its novelty. Carbon nanobeads, carbon nanotubes, carbon fibers, active carbon, charcoal, and fullerenes all fall in the intermediate group of carbon because they contain different ratios of sp3 /sp2 . Perhaps it may be a good idea to devote some time to discuss structural aspects, applications, and properties of carbon belonging to these three new groups of carbon.

3. DIAMOND-TYPE CARBON WITH PURE sp3 CONFIGURED CARBON ATOMS 3.1. Structural Aspect Material containing 100% sp3 bonded carbon is classified as “diamond.” Carbon in the form of diamond belongs to the cubic system and is the hardest material known. It has the same space lattice as zinc blend (Fig. 1a) except that alternate tetrahedral sites and all octahedral sites are filled with carbon atoms instead of sulfur and zinc, respectively. The essential point of this structure is that every carbon atom is surrounded by four other carbon atoms situated at the corners of a regular tetrahedron. Thus, all carbon in diamond has the sp3 configuration. The distance between the centers of two adjoining atoms is 1.54 Å which, strangely, corresponds very closely to the distance between two carbon atoms attached to each other by a single covalent bond in aliphatic organic compounds. This agreement, together with the fact that each carbon atom in diamond has four others situated around it at the corners of a rectangular tetrahedron, suggests that every atom is joined to four others by covalent linkages. It will be noticed from Figure 1b that the carbon atoms form a series of hexagonal “puckered” rings similar to those in cyclohexane, where the carbon atoms (a)

(b)

(c)

= Zn =S

(d)

Basal plane surface

A Prismatic surfaces

forms of carbon such as active carbon, black carbon, carbon beads, nanobeads, nanotubes, fibers, etc. Which class of carbon should these materials therefore be considered as? For example, has the classification of carbon as diamond and graphite led us to think of the existence of the fullerene family? Certainly not. Moreover, does the classical classification give us any impetus to think of the possibility of developing any new forms of carbon? The answer is No. Thus, there is a need to evolve a new classification of carbon which on one hand, can give us an insight for developing new forms of carbon, and on the other, should be able to classify all the existing forms of carbon. For this purpose, perhaps it would be advisable to examine some of the properties of carbon, which are distinctly different from other forms of carbon. We shall discuss this aspect here in more detail. However, it would be a difficult task to cover all aspects of carbon in this small review, since each aspect of carbon can form a book in itself. Hence, in this review, an inquiry is made into the suitability of the conventional classification of various allotropes of carbon and whether classification gives an insight into developing newer forms of carbon. This is followed by a brief description of conventional types of synthesis, properties, and applications of carbon, which have been discovered lately. In order to keep the list of references to a minimum number, references appearing within a given paper are, as far as possible, not included in the list of references.

B Layer plane spacing A

Outline of unit cell

6 0.24 nm

c – = 0.3354 nm 2

2 14 0. nm

Figure 1. (a) Zinc sulfide (zinc blend) lattice and (b) space lattice of diamond forming hexagonal “puckered” ring like cyclohexane. (c) Carbon atom surrounded by four other carbon atoms forming channels in diamond structure.

Carbon Nanomaterials

are joined by single linkages. This figure, however, clearly shows the possibility of the existence of many interlocked hexagonal nanochannels, and the diameter of the channel would have a dimension similar to the hexagonal cyclohexane ring (Fig. 1c). Hence diamond can also exist in hexagonal form (lonsdaleite) with a wurtzite crystal structure where the C C bond length is 1.52 Å. The crystal density of both types of diamond is 3.52 g cm−3 .

519 (a)

3.2. Synthesis of Diamond and Diamond Films Until the 1950s, nature was the only source of diamond, because of the extreme conditions required for its formation. Since at room temperature and atmospheric pressure the trigonal bond is more stable, a very high temperature (greater than 2500 K) and high pressure (around 100,000 atm) are required to induce the formation of tetrahedral bond. In the 1950s, scientists simulated these conditions in the laboratory to produce synthetic diamond. Two decades later, in the 1970s, scientists showed that diamond could be produced at lower temperature (1000 K) and subatmospheric pressure by thermal dissociation of hydrocarbon gases, for example, methane, ethane, acetylene, etc. This process involves dissociation of hydrocarbon, typically methane-hydrogen mixture, with the addition of small amount of oxygen. Techniques such as hot filament, flame (acetylene torch), or microwave excitation are used for the dissociation process. The growth process consists of a competition between the tetrahedral and trigonal bonds where the former are promoted and the latter suppressed by their etching generated by atomic hydrogen and oxygen or internal rearrangement of carbon atoms [5]. It was believed earlier that the substrate’s temperature in excess of 1000 K was necessary to get a diamond film, but now it has been shown that diamond film can be grown at 700 K or less also. Sharon and co-workers [6] have been able to prepare diamond by chemical vapor deposition of soot of camphor (Fig. 2). The standard free energy changes for the process Cgas → graphite and Cgas → diamond are −671.26 and −668.36 kJ mol−1 at 298 K, respectively. This comparison suggests that the thermodynamic driving forces for forming graphite and diamond from vapor phase are similar. Hence large numbers of chemical vapor deposition (CVD) diamond deposition techniques have been developed. The following are the essential requirements for deposition of diamond film by CVD process: (i) A carbon source gas that is rich in hydrogen but dilute (typically ∼1.0 vol%) in the carbon-containing gas. The gas mixture may contain oxygen in form of CO. (ii) A process to activate the gas to produce free radicals, excited molecular species, or plasma. (iii) A temperature-controlled substrate (typically at 500– 900  C). CVD processes can be classified as thermal methods (e.g., hot filament methods) and plasma methods (direct current, radio frequency, and microwave). Film deposition rates range from less than 0.1 m h−1 to ∼1 mm h−1 depending upon the method used, although over a small area,

0.00P

SE

20 . 0 k V

1976

(b)

5.000P

SE

20 . 0 k V

–4924

Figure 2. (a) Diamond thin film deposited over quartz glass from the soot of camphor. (b) Magnified SEM view of the diamond film. Reprinted with permission from [6], K. Mukhopadhyay et al., Mater. Chem. Phys. 49, 252 (1997). © 1997, Elsevier Science.

high deposition rates of 100 mm h−1 or more have been reported. Thinner films with good adhesion to substrates (e.g., silicon wafer) have been achieved, but in thicker films spalling or delamination occurs because of built-in-stress [7]. CVD diamond films can be deposited on a wide range of substrates (metals, semiconductors, insulators, single crystals and polycrystalline solids, glassy and amorphous solids). Most of the CVD diamond films reported to date have been grown on single-crystal Si wafers, mainly due to availability, low cost, and favorable properties of Si wafers. However, one could use any other substrate for deposition of diamond provided it meets the following requirements [8]: 1. Substrate must have melting point higher than the temperature required for diamond growth. This precludes the use of plastic, aluminum, and glasses, GaAs, etc. 2. Thermal expansion of substrate material must be comparable to that of diamond. Otherwise, diamond film will experience compressive stresses from the shrinking substrate, leading to bowing of the sample and/or cracking, flaking, or even delamination of the entire film. 3. In order to form adherent films, it is a requirement that substrate material be capable of forming a carbide layer to a certain extent. The carbide layer can be pictured as the “glue” which promotes the growth of diamond and aids the adhesion. Therefore, material which does not have many tendencies to form carbide (e.g., Cu, Sn, Pb, Ag, Au, and Ge, sapphire, alumina) forms only few

Carbon Nanomaterials

520 layers of carbide. This layer of carbide helps the growth of the subsequent layer of diamond as well as helps to form a strong adhesion with the substrate. But those materials which have greater ability to form carbide (e.g., Pt, Pd, Rh, Ni, Ti, and Fe) promote the growth of a thicker layer of carbide over the substrate. These materials, in addition to helping the growth of diamond, also modify the property of deposited diamond. There are some materials which form carbide easily (e.g., metals like Ti, Zr, Hf, V, Nb, Ta, Cr, Mo, W, Co, Ni, Fe, Y, Al, etc., and nonmetals like B), and when these materials are used as substrate, most of the layers of the diamond are in form of carbide; thus the property of diamond is completely changed. It is believed that mechanical polishing with diamond grit ranging from 10 nm to 10 m aids nucleation by creating appropriately shaped scratches in the surface, which act as growth templates, or by embedding nanometer-sized fragments of diamond into the surface, which then act as seed crystals, or a combination of both [9]. Diamond thin films have been grown from a diluted mixture of a hydrocarbon gas (e.g., methane) in hydrogen. For example, a gas mixture consisting of methanol and hydrogen introduced at pressure of 10 Torr gives a deposition of diamond on the substrate maintained at 800  C. Atomic hydrogen prevents the surface reconstruction from a saturated sp3 diamond to sp2 graphite microstructure. Hydrogen also helps in abstracting hydrogen from hydrocarbon surface sites and gas-phase hydrocarbon species to form reactive radicals. Sharon and Chaterjee [10] have been able to get diamond film by pyrolysis of turpentine oil with iron catalyst at 650  C under argon gas (Fig. 3). The purity and the property of the film are strongly controlled by the graphitic carbon content in diamond film. Doping by phosphorus gives n-type diamond. This was achieved by carrying vapor of phosphorus/methanol mixture into the reactor from a solution of P2 O5 in methanol [11]. Microwave plasma-assisted CVD from a mixture of H2 S in 1% CH4 /H2 on a Si substrate [12] has also given n-type diamond. Diamond has also been doped with nitrogen by allowing vapor of urea into the reactive chamber where n-type diamond is being formed [13]. For this purpose a saturated solution of urea and methanol in acetone is vaporized and

20 . 0 0 P

20 . 0 k V

1081

Figure 3. SEM micrograph of diamond film obtained at 650  C from pyrolysis of turpentine oil in presence of ferrocene catalysts.

carried into the hot-filament chemical vapor deposition unit along with acetone and hydrogen. Nitrogen could be doped to ∼3.5 × 1020 N atoms per cubic centimeter into the diamond. Such a high concentration of N is usually difficult to incorporate in diamond when using nitrogen or ammonia as the dopants. Boron doping of diamond has been achieved by dissolving B2 O3 in methanol [14]. This process gives volatile trimethylborate (CH3 O)3 B. Ferreira et al. [15] prepared boron doped diamond by hot-filament technique. CH4 flow (0.5 s cm−3 ) was maintained along with flow of hydrogen through a solution of B2 O3 in methanol. Boron could be carried by hydrogen from this mixture. Furnace temperature was maintained at 800  C.

3.3. Properties of Diamond Thin Films Although the standard enthalpies of diamond and graphite only differ by 2.9 kJ mol−1 , a large activation barrier separates the two phases, preventing interconversion between them at room temperature and pressure. Ironically, this large energy barrier, which makes diamond so rare, is also responsible for its existence, since diamond, once formed, cannot spontaneously be converted to the more stable graphite phase. Thus diamond is said to be metastable, that is, kinetically stable but thermodynamically unstable. Diamond possesses several technologically important properties, including extreme hardness, high electrical resistance, chemical inertness, high thermal conductivity, high electron and hole mobility, and optical transparency (Table 1). As per the solid-state concept of vacuum scale, electrons present in or above the vacuum scale are expected to be free electrons. Since the conduction band of diamond lies above the vacuum scale, this means that if electrons of diamond are excited from valence band to conduction band, they will behave like free electrons. This condition is different from a state of any other material whose conduction band lies below the vacuum scale. With such material, although electrons excited from its valence band to conduction band are free, they still belong to the same material, whereas with diamond, they become nonbound free electrons. This also suggests that if diamond were heavily doped such that its Fermi level shifted to about 0.1 eV below the conduction band, diamond would become a free electron emitter. Unfortunately, it has not been possible to dope diamond with donor atoms to this extent so far. It is interesting to observe that naturally occurring diamond, though in pure form (intrinsic diamond), is p-type. The reasons for intrinsic diamond being the p-form are not very well understood, though many explanations have been given. The author of this chapter would like to propose the following arguments to explain this behavior. Conversion of material into either n- or p-type depends upon the nature of dopant added into the materials. If dopant has a greater number of electrons than the host atom, the material becomes n-type, and if the dopant has a smaller number of valence electrons than the host, the material becomes p-type. This behavior is quite reasonably understood. Why diamond in pure form should exist in p-form is not very well understood. Perhaps, this strange behavior of diamond can be explained by considering the electronic configuration of carbon atoms bonded into the structure.

Carbon Nanomaterials

521

Table 1. Some outstanding properties of diamond. Property Atomic mass Crystalline structure Symmetry Chemical bond Lattice parameter Density Melting temperature Highest bulk modulus Hardness Linear expansion coefficient at 300 K Lowest compressibility Highest room temperature thermal conductivity Thermal expansion coefficient at room temperature Highest sound propagation velocity Room temperature resistivity (undoped) Bandgap Breakdown voltage Biological compatibility Mobility (electron) at 300 K [27] Mobility (hole) at 300 K Raman shift Work function Dangling bond Boron doping Phosphorus doping [28] Nitrogen doping

Electron affinity of the bare surface of diamond [29] EA for hydrogenated C (111):H surface [30]

Value 12.01 fcc Fd3m covalent sp3 3.56 3.51 g cm−3 4000  C 12 × 1012 N m−2 6000–10000 kg mm−2 118 × 10−6 deg−1 183 × 10−13 m2 N−1 20 × 103 W m−1 K−1 10 × 10−6 K 17.5 km s−1 83 × 10−13 m2 N−1 1.5 eV (indirect) 10 × 107 V cm−1 yes 1000–2200 cm2 V−1 s−1 1400–1600 cm2 V−1 s−1 1332 cm−1 4.7 eV (close to graphite) ∼1.4 eV above valence band acceptor level at 0.38 eV above Ev donor level 0.46 eV below Ec creates two levels, one at 1.7 eV below Ec and the other one doubly occupied just above Ev due to quasi-N lone pair +0.37 eV −1.27 eV. Thus Ev lies between 5.9 eV and 4.2 eV below vacuum level

The combination of one s orbital with two p orbitals leads to the formation of three sp2 hybridized orbitals. Each of these orbitals has one large and one small lobe. All these hybridized orbitals lie in the same plane (the plane containing the two p orbitals from which they were formed) and point toward the corners of an equilateral triangle. Each sp2 hybridized carbon atom has one electron in a p valence orbital that is perpendicular to the plane of the  bonds. These p orbitals can overlap to form  bonds. This type of bond places electron density above and below the line joining the bonded atoms (i.e., above or below the plane in which  bond lies). In contrast to this, the mixing of the s and all three p orbitals yields four sp3 hybridized orbitals. Again, each of these orbitals has a large and a small lobe. The large lobes point at the corners of the tetrahedron. Unlike sp2 hybridized carbon atoms, due to absence of  bonds, the charge density in sp3 carbon atoms can be

assumed to be zero; that is, the plane perpendicular to the plane containing  bonds will have no charge (Fig. 4). Therefore, in contrast to diamond, graphite should have some charge density in space, perpendicular to the  bonds of diamond. In other words, removal of this charge from  bond by addition of hole should lead to a structure of diamond. If we assume the existence of some kind of equilibrium between the graphite and diamond (based on the charge density), then the following equilibrium between sp2 and sp3 carbon atom can be conceived: sp3 -C-C-sp3 + h+ missing of electron charge in plane ⊥ to  bond (Fig. 4b)


 

sp2 -C-C-sp2 + e− addition of electron charge due  bonds ⊥ to  bond (Fig. 4a)

If this concept is accepted, then diamond should intrinsically exist as p-type material and graphite should exist as n-type material. This is what is normally observed with these materials. The conversion of graphite to diamond can therefore be viewed as substitution of sp3 configured into sp2 configured carbon atoms. This explanation also suggests that while pure diamond and pure graphite should exhibit p- and n-type character, respectively, materials containing carbonbonded atoms of some suitable ratio of sp3 and sp2 should exhibit a true intrinsic character. Sharon et al. [16] observed that pyrolyzing camphor under inert atmosphere and sintering it under inert atmosphere for different time, changes the ratio of sp2 /sp3 carbon. They also observed that if this ratio is 33%, the resulting carbon behaves like true intrinsic material. If this ratio is greater than 33%, pyrolyzed carbon shows n-type character, and if it is less than this value, pyrolyzed carbon shows p-type character. This change in behavior is observed only due to change in the ratio of sp2 and sp3 and not due to any external doping (Fig. 5). This result has suggested that in carbonaceous materials, the meaning of n- and p-type needs to be addressed differently and should not perhaps be treated as we normally do with inorganic materials. Thus, naturally occurring p-type diamond need not be due to presence of dopant. It may be due to presence of larger concentration of sp3 carbon atom. In other words, p-character is due to the absence of electron in  bond. Perhaps this may be the reason why it is difficult to either make or naturally obtain n-type diamond. However, substitution with impurity atoms like nitrogen of some of the carbon atoms during the diamond film growth makes the film n-type. Nitrogen atoms in the diamond make bonds to both carbon and hydrogen. Therefore, the total bonding configuration changes by nitrogen addition. It forms a distorted substitutional site with one long bond forming a

A

B

Figure 4. (a) sp2 carbon with  bond (shown in black square) exchanges with (b) sp3 carbon, creating empty  bond (shown in empty square).

Carbon Nanomaterials

522 50 n

45 1000°C

sp2-carbon (%)

n n

40

n

35 p

p

30 900°C

p

p

p

p 950°C

25

p 20 1

2

3

4

5

6

7

Pyrolyzing time (h)

Figure 5. Variation in the sp2 carbon content in the camphor pyrolyzed carbon films as a function of pyrolyzing time. Reprinted with permission from [16], M. Sharon et al., Mater. Chem. Phys. 56, 284 (1998). © 1998, Elsevier Science.

deep singly occupied donor level, 1.7 eV below the conduction band Ec , and also a doubly occupied state just above Ev due to quasi-N lone pair of this C N bond [17]. The decrease of hydrogen bonded to carbon atoms can cause the increase of sp2 bonding fraction, hence the decrease of the optical bandgap [18]. Doping diamond [19] with boron produces a shallow substitutional acceptor with a level at 0.38 eV above Ev . Phosphorus also acts like a substitutional donor, with a level about 0.46 eV below Ec . Replacement of one CH group by one phosphorus atom causes diamond to become n-type. Typical Raman spectra for high-quality diamond and boron-doped diamond thin films feature one intense band at 1332 cm−1 with a line width of 5 to 8 cm−1 . The significantly broadened and shifted diamond line for the noncrystalline films results from decreasing grain size to the nanometer scale. To a first approximation, the linewidth is a measure of the phonon lifetime. The more defects (e.g., grain boundaries) and impurities (e.g., nitrogen) there are, the shorter the phonon lifetime and the broader the linewidth. Weak scattering intensity in the region of 1520 cm−1 is observed due to nondiamond sp2 bonded carbon peak resulting from the -bonded carbon atoms in the grain boundaries [20, 21] impurities in the film. Graphite, however, gives rise to a broader peak at 1580 cm−1 . The ratio of the intensities of these two peaks indicates how much of each phase is present. It is also dependent on the wavelength of excitation (normally excitation is done with 532 nm). It can be seen that the ratio of the diamond to the nondiamond band intensities, I1339 /I1560 , is largest for the film deposited without any added N2 and decreases for the film deposited with the gas. However, the ratio is independent of the N2 level [22]. Raman spectra of boron-doped diamond films were studied by Ferreira et al. [15]. It is interesting to note that the intensity of the peak at 1332 cm−1 decreased as the concentration of boron was increased. Moreover, a new

peak started to appear at around 1200 cm−1 , whose intensity also increased with increase in doping concentration. This suggests that one may be able to establish a relationship between boron concentration and the intensity of either the 1332 cm−1 or the 1200 cm−1 peak. Some of the CVD grown diamond film exhibits photoconductive signal if it contains absorption bands between 1.3 eV and 1.5 eV. This photoconductive response in the visible region can impair the performance of diamond detectors made for ultraviolet application [23]. Photoconductivity studies, therefore, are a useful technique to characterize diamond. It is interesting to note that Hall mobility in boron doped diamond decreases as the concentration of dopant is increased [24]. While at hole concentration of 5 × 1014 cm−3 , Hall mobility is in the vicinity of 1500 cm2 /Vs, for 5 × 1015 cm−3 hole concentration it falls down to 250 cm2 /Vs. Three low-index morphologies are usually obtained during growth of diamond film: (100), (110), and (111), the latter being the most abundant. It is observed that while the morphology of the (111) and (110) surface is rough, (100) crystallites grow as platelets or tiles with partial overlap between neighboring ones. Most electrochemical studies are carried out with surfaces preponderant with (111) crystallites [25]. Robertson and co-workers [26] have suggested that chemical termination of (111) diamond surfaces shifts the electron affinity from +0.35 eV for clean surface to −1.3 eV for H terminated ones and about +2 eV for oxygen terminated ones. The shift occurs because the terminating groups are slightly polar and introduce a surface dipole layer. In the case of C H groups, the H atoms are positive and the dipole reduces the electron affinity, while oxygen has a negative change and the C O bond dipole makes the affinity more positive. As these effects are localized on the chemical bond, similar affinity changes may be expected on ta-C surfaces. In Table 1, various properties of diamond are enumerated.

3.4. Electrochemical Properties of Diamond Film In electrochemical work, there is a need to use an electrode which is inert and does not need larger additional potential (known as overpotential) than the equilibrium potential for carrying out electrochemical reactions. Moreover, one also needs a large potential difference between oxidation and reduction potential of water so that the electrode can be used safely for electrochemical reaction within the so-called window of the potential (i.e., within the range of potential in which water does not get electrolyzed) in aqueous solution. For this purpose, platinum or mercury has been used by electrochemists. Conducting diamond, being inert to alkali and acid, has been used as an electrode. Since doping of diamond has been possible, lately, boron-doped diamond films are getting more popularity in electrochemical studies than carbon electrodes (glassy carbon). Diamond thin-film electrode shows a very small background current and low double-layer capacitance. With 0.1M KCl solution between −1000 to +1000 mV, current is observed in the range of 50 mA/cm2 and is highly stable. Sometimes a small oxidation peak is observed on the forward sweep at ∼1200 mV just prior to the onset of chlorine evolution [31]. Boron-doped

Carbon Nanomaterials

diamond thin film shows a wide working potential window for solvent-electrolyte electrolysis in aqueous media. This means that diamond has large overpotentials for both hydrogen evolution and oxygen evolution. In 0.1M KCl as well as in 0.5M H2 SO4 , the working potential window is 3.5 V (i.e., from −1.25 to 2.3 V vs standard hydrogen electrode) as compared to 2.5 V for glassy carbon. These ranges greatly exceed the theoretical range of stability of water under ambient conditions (1.23 V). Diamond film is reported to show an increase in background current by less than 5% even after constant use for 15 hr, whereas in glassy carbon it increases to a value greater than 30.0%. Another application of diamond film is anodic stripping voltammetry of heavy-metal ions in aqueous media. It has been possible to electrodeposit Pt, Hg, and Pb on boron-doped diamond thin films [32]. Ferro et al. [33] have observed chlorine evolution at highly boron-doped diamond electrodes. They reported the chlorine evolution below 1 V versus SCE with NaCl solution. Since this electrode is stable in perchlorate solution, one can think of making a fuel cell operating with chlorine/hydrogen gas.

3.4.1. Electron Field Emission Effect The field emission display (FED) is a new type of flat panel display which overcomes many limitations of liquid crystal display, because of its superior brightness, viewing angle, color rendition, response time, and operating temperature range. The activity in field emission display, sometimes called cold emission devices, is concentrated now mainly in vacuum microelectronic devices based on broad area multiple arrays, for example, flat panel displays, microwave devices, vacuum electron sources for electron guns for electron microscopes [34], etc. For a perfectly flat metal surface with a typical work function of 5 eV, the electric field necessary to get measurable emission current is in the range of 2.5 V nm−1 , or in more appropriate units, 2500 V m−1 . To generate such high electric fields, one has to use the effect of field enhancement at tip-like structures. For such purpose, silicon or metal microtips are generally used. Nevertheless, these materials suffer emission degradation due to sputter erosion and chemical contamination and therefore require a high vacuum environment for operation. On the other hand, carbon and especially diamond is found to emit at a very low electric field. The chemical inertness of carbon field emitters especially is one of the most important advantages over silicon or metal micro-tips. Field emission from semiconductors in general can originate from the valence band or the conduction band, from an accumulation layer at the surface [35]. The great interest in electron emission from diamond was kindled by the work of Himpsel et al. [36], who reported high quantum yield for photon-electron emission from the (111) hydrogen terminated surface of diamond. Later it was determined [37] that the (111) hydrogen terminated diamond surface exhibits a negative electron affinity. The electron affinity of a semiconductor describes the energy barrier between the vacuum level outside the surface and the conduction band minimum inside the sample. Given this definition, the electron affinity of these prepared surfaces of diamond would exhibit a negative value, because the energy of electrons at the conduction

523 band minimum for diamond would be above the vacuum level, and this condition is termed negative electron affinity. The presence of a negative electron affinity means that electrons in the conduction band can be freely emitted into vacuum. Field emission from high-quality diamond is actually quite difficult, largely because of its high resistivity. Field emission from polycrystalline diamond is easier and is found to vary inversely with the grain size. Development of conducting diamond thin films thus has led to new areas of cold electron emission from diamond. Recent results from diamond cold cathode demonstrate high emission current, good current stability, uniform emission, from an array of emitters, and emissivity even under poor vacuum conditions. Two approaches using ungated field emission cathodes with diamond coating have been proposed: (i) flat-film display, where patterned diamond film is used as addressable cathodes (lines) [38], and (ii) arrays of silicon tips with diamond coating, where addressing occurs by switching of p-n junctions formed in the Si substrate [39]. The turn-on fields of diamond films have been in the range of 1–30 V mm−1 ; that is, the electric field needed to inject electrons from diamond film is in the range of 1–30 V mm−1 . A schematic of a test circuit of the field emission [40] is shown in Figure 6. Sometimes, a series resistance of 5 M is also connected to the sample to protect the equipment. The emitting area could be of the size 15 mm × 5 mm separated by a distance in the order of a few hundred micrometers. The system is kept in vacuum (∼10−7 torr or sometimes even lower than this). When potential is applied across anode and cathode, current will appear after certain threshold voltage is applied. To show the nature of CV characteristics, a typical current density (A cm−2 ) versus applied electric field obtained with kerosene-pyrolyzed carbon film deposited over alumina plate at 1000  C is shown in Figure 7 [40]. It will be noticed that there is a threshold voltage of 2.20 V m−1 before which no emission occurs. It is also necessary to confirm that this emission is not due to electric discharge of gas present in the system. In order to confirm, normally polarity of electrode is altered and current density (A cm−2 ) versus applied electric field (V m−1 ) is measured. If current characteristic as shown in Figure 7 were not obtained, this would suggest the absence of any current due

µA

Platinum + –

Mica Carbon film on Alumina plate Vacuum

Figure 6. Schematic diagram of the field emission device. Reprinted with permission from [40], M. Kumar et al., Diamond Related Mater. 10, 883 (2001). © 2001, Elsevier Science.

Carbon Nanomaterials

524

found that phosphorus doping improves the emission in diamond. The process of electron emission can be summarized as follows:

30

Current Density (µA/cm2)

25

1. Tunneling from the substrate (e.g., metal or semiconductor) into the conduction band (CB) of diamond. This is designated as electron injection process into the CB. 2. Transportation of the electrons to the diamond surface. 3. Escape of the electrons from the diamond surface to a vacuum.

20

15

10

5

0

2

3

4

5

6

El. Field (V/µm)

Figure 7. Current density (A cm−2 ) vs applied electric field (V m−1 ) obtained from kerosene-pyrolyzed thin film of carbon deposited over alumina plate utilizing configuration as shown in Figure 6. Reprinted with permission from [40], M. Kumar et al., Diamond Related Mater. 10, 883 (2001). © 2001, Elsevier Science.

to electric discharge process. It has been reported that some fluctuation of field emission occurs. It is believed that the fluctuation of field emission current is caused by migration of absorbed gas molecules over the emitter surface, and ion bombardment to the emitter surface [41–43]. So the fluctuation can be increased as the impinging rates of the gas molecules and ions are increased. In order to reduce this effect, high vacuum is necessary. However, before a reliable emitter structure can be developed, the electron transport properties of the material must be well understood, especially regarding limitations on the current density and energy spread of the transmitted beam. Furthermore, an understanding of the emission process is necessary to accurately interpret emission [44]. There is ambiguity in explaining the actual phenomena which are responsible for the electron emission from diamond or materials like diamond. A simplified mechanism is discussed here so that this process can be qualitatively understood. A basic process for injection of electrons into diamond is the tunneling of electrons from the cathode substrate through a potential barrier at the interface (i.e., a forbidden energy region) into the conduction band of diamond film [45]. It is expected that the cathode or metal over which diamond is deposited forms a metal/Schottky type of contact. An energy barrier is formed at the interface of the metal/diamond contact. When a negative bias is applied to the metal electrode (cathode), electrons are injected into the diamond film from the metal. The number of injected electrons depends on the barrier height and thickness. Therefore, the electron emission characteristics are improved by forming a low barrier height and by enhancing the electric field at the barrier of the diamond. In the case of the ideal Schottky junction, choosing the metal work function can change the barrier height. In order to enhance the electric field at the metal/diamond contact, on the other hand, it is desirable to use diamond films doped with shallow donor impurities because of the formation of a space charge layer due to ionized impurities. Incidentally, Geis et al. [46] have

The electrode of the behind diamond film, such as Si, Mo, etc. (i.e., substrate), forms a Schottky barrier. The electrons are injected from the electrode to the conduction band of the behind diamond film through the Schottky barrier by applying the electric field (step 1). The electrons at the diamond surface, on the other hand, can easily be emitted to a vacuum even by applying low electric field, because the diamond surface has negative electron affinity (step 3). However, it is difficult to transport electrons from the behind diamond film to the surface (step 2) because there is no effective means for n-type doping in the case of diamond. The emission current density strongly depends on the electron-transportation mechanism in the diamond film. In addition to these, there is a possibility that the following process may also work to facilitate the electron injection [47]. 1. Tunneling from the valence band (VB) of the semiconductor substrate into the CB of diamond. This is designated as hole injection from the diamond into the semiconductor substrate. 2. Zener tunneling, which is transition across the energy gap from the VB to the CB of a semiconductor or insulator under the internal applied field. This process is negligible for a wide-bandgap material such as diamond, except for extremely high fields (∼104 V m−1 ). 3. Transition to the CB from the local impurity or imperfection level. This is field ionization process. Druz et al. [48] have reported that reversibility of emission current versus applied potential (I-E cycles) depends upon the magnitude of emission current. It is also interesting to study whether electron field emission can be sustained by diamond films for a long period. In this context, they also have studied the time stability of the emission current of the DLC films. They observed the I-E curves to be reproducible after many measuring cycles, or after many hours without field applied; emission current was found to be constant at different fixed fields during at least one hour [49]. Sugino et al. [45] reported that emission property of polycrystalline diamond film depends upon the work function of the metal (e.g., Cu/diamond, Al/diamond, and Si/diamond) used for making the back contact with diamond. The reduction in the threshold voltage for these metals as compared to silicon is believed due to formation of a low energy barrier to electrons at the metal/diamond contact. On the contrary, Robertson and co-workers [50] have suggested that threshold field, instead of depending on the work function of substrate, depends upon the extent of nitrogen doping.

Carbon Nanomaterials

The field at which emission begins is also found to depend on sp3 content [45]. Sugino et al. [45] and Robertson and co-workers [50] have reported that variation in threshold energy depends upon the nitrogen content in the diamond film: as the concentration increases, the threshold field decreases to a minimum value at 0.4% N in the film, and thereafter it increases again with increase in nitrogen content in the film. It is concluded that the reduction in emission at higher N content suggests that bulk sp2 bonding in diamond does not favor emission. 3.4.1.1. Fowler–Nordheim Theory Electron field emission for metal has been known for many years, and among the various theories put forward to explain this process, Fowler–Nordheim theory is used most for explaining the process in diamond or diamond-like carbon. It would be difficult to deal here in detail with this theory. However, Forbes [51] has given an extremely good review on the Fowler– Nordheim theory and its limitations. The theory of cold field emission of electrons from metal into vacuum is often called Fowler–Nordheim FN theory, after the first author to treat this effect as wave-mechanical tunneling through a triangular barrier. The standard physical assumptions of FN theory are that: (i) the metal has a free-electron band structure; (ii) electrons are in thermodynamic equilibrium and obey Fermi–Dirac statistics; (iii) the surface is smooth and flat; (iv) local work function is uniform across the emitting surface and is independent of external field; (v) there is a uniform electric field above the emitting surface. Considering these assumptions, the magnitude of emission current J  may be written as J = Aa−1 E 2 exp−b3/2 /E ≡ Aa−1 E 2 exp−b3/2 /E

525

4. GRAPHITE-TYPE CARBON WITH PURE sp2 CARBON CONFIGURED ATOMS 4.1. Structure of Graphite Graphitic carbon is carbonaceous material with a layered structure but with a number of structural defects. Its density is 2.36 g cm−3 . It consists of layers in which each carbon atom is linked to three other carbon atoms by strong interplanar, covalent bonds which are stronger than those in diamond, with a C C bond length of 1.42 Å (Fig. 8). However, a weak carbon-carbon bond exists between widely separated layers (interlayer distance of 3.354 Å). These widely separated layers account for the ease with which the layers glide over each other and cause the familiar softness of graphite. Graphite, thus, has a structure in which trigonal planar carbon atoms are present in horizontal planes and each plane is separated by a distance of 3.354 Å. XRD of graphite is shown in Figure 9 [52]. For comparison, XRD of graphite, graphene sheet, hard-shell carbon, and multiwall carbon nanotubes are also shown. The most intense peak corresponds to (002). From a crystallographic point of view, the term “graphite” is applicable only to carbon materials having a layered lattice structure with a perfect stacking order of graphene layers, either the more prevalent ABABAB    (hexagonal graphite) or the less common ABCABCABC    (rhombohedral graphite). Due to the small energy required for the transformation of an AB- into an ABC-stacking (and vice versa), perfectly stacked graphite crystals are practically not available. Therefore, the term “graphite” is commonly used regardless of the stacking order. In some carbons, the aggregates are large and relatively free of defects, for example, in highly oriented pyrolytic graphite (HOPG). For both forms of graphite, the in-plane C C distance is 142 pm. This length is in between the bond length of Csp3 Csp3 (i.e., 153 pm) and Csp2 Csp2 (132 pm). There is a large difference between in-plane C C distance, 142 pm, and the interlayer distance, 335.4 pm, in graphite (Fig. 8) that results from the different types of chemical bonding. Within planes

Or Ln J /E 2 = Ln Aa−1  − b3/2 /E where a ≡ e3 /8h = 1541434 × 10−6 A eV V−2 (in SI unit), b ≡  4/3 2me 1/2 /e
= 6830888 × 109 eV−3/2 V m−1 (in SI Unit),  = the local work function of the emitting surface, E = external electric field (taken as a positive quantity), and A = area of emission. Thus a plot of Ln J /E 2 versus 1/E would give a linear plot with a slope equal to b3/2 . Since value of constant “b” is known, local work function of the emitting surface () can be calculated from this analysis. It has been also observed that when current-voltage characteristics of the electron emission are plotted in the Fowler–Nordheim coordinates of log 1/E 2  versus 1/E, broken lines are observed, and near the break, plots have a more or less extended “shelf” segment where emission current is only weakly dependent on voltage. This behavior of getting broken line is in contrast to expected continuous line.

6.706Å

b 3.364Å

a 1.421Å

Figure 8. Schematic drawing of the crystal structure of hexagonal graphite showing AB-layer stacking sequence and unit cell.

Carbon Nanomaterials

526

(11) (006)

hard shell

(002)

Intensity (arbitrary units)

(004)

(002)

(004)

graphene sheets

(002)

(004)

purified MWNTs

(110) (112) (006)

(004)

(100) (101)

graphite

10

20

30

40

50

60

70

80

90

2 Theta

Figure 9. XRD of graphene sheet, hard shell, purified MWNT, and graphite. Reprinted with permission from [52], X. Zhao et al., J. Cryst. Growth 198/199, 934 (1999). © 1999, Elsevier Science.

the C C bonds are trigonal sp2 hybrid  bonds with delocalized  bonds. The large interlayer spacing suggests that the contribution to interlayer bonding from -bond overlap is negligible. When the turbostatically disordered structure becomes more distinct and dominates among the crystallites, the carbonaceous material is no longer graphitic and can be considered as a nongraphitic carbon. Nongraphitic carbonaceous materials consist of carbon atoms with most of them arranged in a planar hexagonal network but without the crystallographic order in the c-direction of the graphite structure. The ability for graphitization is higher (i) if the precursor material comprises highly condensed aromatic hydrocarbons, and (ii) if neighboring graphene layers or graphitic crystallites are suitably oriented to each other. Unlike diamond, in which all the orbitals of the four valent electrons of carbon atom (one s and three p’s) are hybridized (sp3 hybridization) to make four equivalent, strong strictly directed bonds, in graphite only three of the four orbitals are combined (sp2 hybridization), making three equivalent, strong, coplanar bonds with an angle of 60 between three  bonds. The remaining fourth electron is quasi-free, that is, free to move within its layer, and it makes very weak bonds ( bonds) directed perpendicular to the

hexagon layers, thus bonding these layers to each other. The atoms within the layer are strongly bound and therefore the layers are compact and difficult to break. On the other hand, neighboring layers are hardly bound to each other, so that they can be easily moved with respect to each other, or even separated. Since it is very difficult to break the layers (i.e., by stretching them), graphite’s strength is very large parallel to the layers, but since the layers can be easily separated from each other, strength is nearly zero across them. Along the layers, electrons can move easily and therefore graphite is a good conductor of electricity parallel to the layers, but an insulator across them. Traditional graphite is made of a very large number of tiny crystals randomly oriented; that is, they point to different directions. Therefore, if one crystal is a good conductor in one direction, its neighbor may be an insulator in that direction, so that the net result is no bulk anisotropy at all. On the other hand, it is clear that if we could make all crystals to point in the same direction, we would have an anisotropic material. This is what we have in a pyrolytic carbon. Pyrolytic carbon is a monolithic polycrystalline graphitizable carbon material obtained by pyrolysis of a hydrocarbon vapor and deposition of carbon on a solid substrate heated to high temperature. The structure of pyrolyzed carbon is characterized by a strong preferred orientation of the constituent crystallites with their basal planes mostly parallel to the substrate surface. Properties of such carbon depend upon preparation conditions and are mostly determined by the deposition temperature. Graphitization of carbon by metal is one of the processes to prepare graphite at lower temperature. It is observed that temperature at which graphitization occurs depends upon the initial state of the carbon. The lowest temperature of graphitization by Ni occurs for diamond or diamond-like carbon at 700  C and 500  C, respectively [53]. There has been considerable interest in the doping of graphite with elements such as boron, nitrogen, and silicon in order to modify the electronic, mechanical, and oxidative properties of this material. In recent years, a number of carbon-boron, carbon-nitrogen, and carbon-boron-nitrogen alloy structures have been proposed from the results of theoretical calculations, notably BC3 , C3 N4 , and BC2 N [54–57]. In terms of boron doping within the graphitic structure, it has been shown that the thermodynamic solubility of B in the C lattice [58] is around 2.35 at.%. Introduction of boron in graphite-like lattice induces the presence of disordered phases. These domains occur when boron content level is higher than a few percent. The exact nature of the boron sites in the higher-boron-content material still remains unclear. The perfectly conjugated polyne (( C C )n ), called “carbyne,” is known as one of the carbon allotropes. Though many attempts have been made to synthesize it, synthesis is still under investigation [59, 60]. Carbyne has been considered to be a transition phase between graphite and diamond and a carbynoid material was experimentally observed to form diamond [61]. On the other hand, polyne was suggested to be a precursor for fullerene [62].

Carbon Nanomaterials

4.2. Application of Graphite Graphite has been used for many industrial applications. It has been used as an electrode in electrochemical works, an inert crucible for high-temperature application, a solid lubricant, a precursor for synthesis of fullerenes, in carbon nanotubes, etc. It would be impossible to discuss applications of graphite in this chapter because of space limitations; also, they are well known and documented nicely at many places. However, recently graphite has found an application in an area related to environmental problems, for example, leakage of oil during transportation; its large absorbent capacity and easy manipulation are expected to avoid serious pollution problem. So far, some porous polymers, such as poly(propylene) and poly(ethylene terephthalate), have been used for the absorption of spilled oil, of which maximum absorption capacity is ∼10–20 g of heavy oil per 1 g of polymer [63]. Moreover, most of the polymers absorb water, as well as heavy oil. They do not have special selectivity for heavy oils. On the other hand, exfoliated graphite was found to absorb selectively a large amount of heavy oil. It is reported that about 80 g of heavy oil floating on water can be removed from 1 g of exfoliated graphite [64].

5. INTERMEDIATE FORMS OF CARBON CONTAINING MIXTURE OF sp2 AND sp3 CONFIGURED CARBON ATOMS In previous discussions we dealt with carbon with 100% either sp2 carbon (graphite) or sp3 (diamond). Now we shall discuss carbon material possessing both forms of carbons. It will be noticed that the variation in the percentage composition of sp3 and sp2 carbon yields different forms of carbon. Therefore, a new method of classification has been suggested here, where the structure and properties of carbon materials are mainly governed by the ratio of these two forms of carbon. Under a suitable condition of preparation, it is possible to create different forms of carbon having different ratio of sp3 and sp2 carbon atoms. Diamond-like carbon and glassy carbon are two examples of such configurations, because they have different structures from graphite and possess both types of carbons (i.e., carbon with sp3 and sp2 configuration). We shall now try to discuss this class of carbon materials in brief.

527 properties such as bandgap, while the sp3 regions control mechanical properties such as rigidity and hardness. Because of this, diamond-like structure shows almost all properties of diamond and the conductance properties of graphite. DLC films, depending upon the ratio of sp2 and sp3 carbon atoms, can also be stable, hard, and chemically inert like diamond. Depending upon the ratio of sp2 and sp3 nature of carbon, various types of DLC films can be obtained [65]. Furthermore, there are two types of semiconducting amorphous carbon: (a) with hydrogen (a-C:H); and (b) pure carbon without hydrogen (a-C). In a-C:H, commonly obtained when a hydrogen gas is used as the plasma source, the rigidity of the amorphous network, and hence the “diamond-like” content, is influenced by the fraction of polymeric C H bonds in different configurations. In a-C, it is the relative amounts of sp2 and sp3 C C bonds in the network which determines the macroscopic electronic, optical, and mechanical properties of the materials. Thus, DLC covers a wide range of materials, from the a-C:H containing sp3 content in the range of 20 to 50%, to the tetrahedral amorphous carbon (ta-C) containing essentially no hydrogen and sp3 contents up to 70–85%. A systematic classification of amorphous carbon films has been given by Biswas et al. [66]. According to them, classification of amorphous carbon films can be shown graphically in form of a triangle. The corners at the base of the triangle correspond to graphite (100% sp2  and diamond (100% sp3 carbon). The apex represents 100% H, but the upper limit for formation of solid films is defined by the tie line between the compositions of polyethene, (CH2 n− , and polyethyne, (CH)n− . The majority of a-C films contain mainly sp2 carbon, but the sp3 carbon content can be varied over the range 5–55%; the hardness of the films increases with sp3 content. The H content of a-C:H films can be varied over a wide range and the hardness of a-C:H film is inversely related to the hydrogen content. Films with very high sp3 content (∼80–90%) and correspondingly high hardness have been called tetrahedrally bonded amorphous carbon films (ta-C films). Hard a-C:H films are called “diamond-like carbon” (DLC). Thus, in general, hardness increases with sp3 carbon content, as the proportion of DLC increases. Conversely, the films become softer as the sp2 carbon content and/or hydrogen content increases [67, 68].

5.1.2. Synthesis of DLC Films 5.1. Diamond-Like Carbon 5.1.1. Structural Aspects of DLC Film Elemental carbon has six electrons, with configuration 1s2 2s2 2p2 . Thus when carbon atoms combine, there is flexibility in terms of allocation (hybridization) of the electrons in the 2s and 2p orbitals, giving rise to graphite (sp2 ), diamond (sp3 ), or a third allotrope of carbon, diamond-like carbon containing a mixture of sp2 and sp3 carbon atoms. When the diamond structure is incorporated with both tetrahedral (sp3 ) and trigonal planar (sp2 ) carbon-bonded atoms, the resulting film takes the form of diamond-like structure. In other words, in DLC, the clusters of three coordinated sp2 carbon atoms are embedded in a sp3 bonded matrix. The sp2 regions of the DLC film are found to control electronic

There is a good review on the DLC films by Zhang et al. [69]. Generally, two basic ways have been used to synthesize a DLC film: one is through the vapor phase via hydrocarbon gases, and the other is by sputtering of a solid carbon source. The latter process usually gives nonhydrogenated or a-C films. In this process, classified as plasma vapor deposition (PVD), plasma of argon is allowed to hit the carbon target to sputter carbon, which in turn is deposited at the substrate to give a DLC film. In other processes, like chemical vapor deposition (CVD), hydrocarbon gas or mixture of hydrogen and a hydrocarbon such as methane (CH4 ), n-butane (C4 H10 ), acetylene (C2 H2 ), camphor, etc., are used as the carbon source. These gases are activated by directcurrent glow discharge, rf discharge, microwave discharge, ion beam, pyrolysis, etc. Sharon and co-workers [6] have

528 been able to get DLC film by pyrolysis of various naturally occurring precursor camphors.

5.1.3. Properties of DLC Films 5.1.3.1. Electrical Properties In DLC, the clusters of three coordinated sp2 carbon atoms are embedded in a sp3 bonded matrix. Consequently, the bandgap of DLC film can be varied from 1 to nearly 4 eV according to the fraction of sp3 bonding. Heat treatment of amorphous carbon films leads to the evolution of hydrogen and causes a structural change. Gradual release of hydrogen at the temperature range from 400  C to 600  C converts tetrahedral sp3 into trigonal sp2 structure. If this happens, then the bandgap of DLC film will decrease as ratio of sp2 /sp3 increases. Sharon et al. [16] have shown that pyrolysis of camphor can give DLC films with various bandgaps depending upon the concentration of sp2 present in the film. In addition to this, Sharon et al. [70] also studied the electrical conductivity of DLC film prepared by pyrolysis of camphor vapor and observed that if films are thermally heated below 600  C, values of electrical conductivity are reproducible. This suggested that up to this temperature, there seems to be not much change in the ratio of sp3 and sp2 carbon and activation energy for conduction remains 1–1.3 eV. However, when the film is heated above this temperature, electrical value takes an altogether different reproducible path and the activation energy for conduction becomes much lower than 0.1 eV. This suggested that concentration of sp2 carbon must have increased to make it highly conducting. These observations suggest that while ratio of sp2 /sp3 controls the bandgap, pyrolyzing below 650  C or sintering at temperature below 600  C does not alter the ratio of sp2 /sp3 carbon present in the DLC film. This property of DLC is useful if we wish to make semiconducting carbon of a specific bandgap and to know the temperature up to which the ratio of sp2 /sp3 carbon present in the film does not alter. It has been suggested that in a-C:H and a-C, it is the  bonding and  ∗ antibonding electronic states resulting from p-orbital bonding at sp2 sites which determine the optical bandgap [71, 72]. The electronic properties of a-C are also determined by the  and  ∗ states as they in effect form the valence band and the conduction band edge states, respectively. In ta-C with very low sp2 content, the  and  ∗ electronic states are highly localized within the much-wider-gap (∼5 eV) - ∗ states arising mainly from the sp3 hybridized diamond-like bonds. Does this mean that any carbon material with sp3 configuration will always result in bandgap of ∼5 eV? This cannot be true. Earlier we discussed the reasons why natural diamond is p-type (Fig. 4), which has 100% sp3 carbon, and why graphite with 100% sp2 carbon is an electronic conductor. When carbon system contains both sp3 and sp2 carbon, a complex interaction of  bonding and  ∗ antibonding with  bonding and  ∗ antibonding should be considered in the formation of bandgap of a DLC film rather than only interaction of  bonding and  ∗ antibonding, because presence of  bonding coming out of sp3 carbon in DLC film neither can be neglected nor can be assumed to be ∼5 eV. It is therefore necessary to take a fresh look into the exact mechanism of the formation of bandgap in carbon films containing both sp2 and sp3 carbon. This is more

Carbon Nanomaterials

important because bandgap of DLC film can be varied from 1 to nearly 4 eV according to the fraction of sp3 bonding. In an amorphous network, however, the orientation of the  orbitals is randomized, so that two  bonds which are next to each other will not interact electronically if their orientations are orthogonal to each other. Therefore, in amorphous carbon even when there are a significant number of - ∗ states, conduction is through a hopping related process and consequently electronic properties such as mobility will remain very low compared to other amorphous semiconductors such as a-S:H [73]. Robertson and co-workers [74] have observed that ta-C changes its bandgap as per the doping level of nitrogen and so does the conductivity of the film. At nitrogen content above about 2%, the bandgap begins to close and bonding reverts towards sp2 . Like diamond, pure DLC films and those containing Cu exhibit p-type conduction, and those containing Ti and Si have n-type conduction [75]. However, they also reported that Cu concentration does not exceed 3.0 at.%. Incorporation of Cu into DLC films decreases the resistivity significantly. Sharon et al. [76] have studied electrochemical properties of DLC film deposited on an alumina plate by pyrolysis of kerosene at 1000  C in argon atmosphere. The electrochemical potential window of this film in contact with H2 SO4 solution was investigated in potential range 2.5 V to −1.5 V vs SCE. The same experiment was repeated with platinum and glassy electrodes using a Pt electrode as a counterelectrode. The kerosene DLC film has the widest potential window (2.91 V) as compared to Pt (2.02 V) and glassy carbon (2.79 V). This result has suggested that kerosene-pyrolyzed carbon electrode can be used as an oxygen electrode more efficiently. They have suggested that DLC film prepared from kerosene is as good as platinum electrode. 5.1.3.2. Spectroscopic Properties Since DLC film is a mixture of sp3 and sp2 carbon atoms, there is no crystallinity in the film. It is quite amorphous in nature and hence XRD cannot be taken as a confirmation for DLC film. Raman spectroscopy has been found very handy to characterize DLC film or to differentiate it from either graphite or diamond. If films contain amorphous sp2 -bonded (graphitic) carbon and sp3 -bonded (diamond-like) carbon (DLC), one gets broad ill-defined bands with high-frequency cut-off near 1332 cm−1 (in the range of 1350–1372 cm−1 , known as D-band or Raman active E2g2 mode of graphite) for the sp3 bonded carbon, and around 1580 cm−1 (usually in the range of 1519–1580 cm−1 , known as G-band, also referred to as the A1g mode frequency) for the sp2 -bonded carbon [77, 78]. Schwan et al. [79] have suggested that the G peak is not necessarily composed only of the E2g2 mode of graphite, but also of sp2 C C stretch vibrations because single benzene rings exhibit a Raman peak at 1588 cm−1 . This feature is valid for all aromatic rings and condensed benzene rings. Therefore, it is suggested that possible contribution to the G peak may arise from the C C sp2 stretch vibrations of olefinic or conjugated carbon chains at about 1620 cm−1 . Thus, the existence of a G peak in the Raman spectrum does not necessarily prove that the a-C film consists only of graphite or fused benzene rings. The G peak might as well have its origin in olefinic chains in the carbon films. The existence of

Carbon Nanomaterials

1/La ∝ I1355 /I1575 = ID /IG It has been suggested that when G linewidth increases beyond 50 cm−1 , the resulting decrease in the ID /IG ratio would be indicative of a cluster size smaller than 10 Å. Raman spectra of graphite rod, graphene, sheet and multiwall nanotube are shown in Figure 10 to illustrate these behaviors.

Graphene Sheets

Outer Shell 1837

1579

1350 1338 1200

Diamond

1576

1348 1344

1567

1581

1332 cm-1

Intensity (a.u)

a peak centered at about 1350 cm−1 proves the existence of aromatic or benzene clusters incorporated in the amorphous (hydrogenated) carbon films. G-band is normally observed to be very broad. In principle, there may be four significant reasons for linebroadening feature in a-C:H films, namely, the cluster size, the cluster distribution, the influence of the stress in the films, and broadening due to chemical bonding. For graphite, Sakata et al. [80] found that the half-width of line1580 cm−1 peaks increases linearly with increasing stress. In hydrogenated DLC (a-C:H) films or unhydrogenated DLC (a-C) films, D-band is broad as compared to G-band. Nevertheless, intensity of D-band is greater with a-C type films as compared to a-C:H films [81]. The presence of the D-band and G-band peaks, therefore, can also indicate the co-existence of sp2 bonds with sp3 bonds. The shifting of the G-band resonance peak towards 1580 cm−1 infers the decrease in the sp3 to sp2 bond ratio. This is supported by the observations made by Cheng et al. [82]. They have shown that as the temperature of the substrate (gold coated silicon) over which DLC films was grown increases, the intensity of D-band (an indication of increase of sp3 constituent of the film) increases. It is reported that in a DLC film prepared below 400  C, no Raman signal is observed unless it is annealed above 400  C [83]. It is observed that with increase of the annealing temperature, the intensity ratio of ID /IG increased slightly. Shifting of G-line position from 1580 to about 1606 cm−1 and D-line position from 1380 to about 1345 cm−1 , and change in the intensity, are the most sensitive indications that the film is changing from one with bond-angle disorder to one containing three fold-cordinated crystallites, that is, the film is undergoing a graphitization process. In principle, these reasons may also be true for the linewidth of D-band. It is possible to get some useful information about the cluster size in amorphous carbon from ratio of intensity of D-peak and G-peak (i.e., ID /IG ) and the G linewidth [79]. In pure graphite, D peak is expected to be absent [84]. But if graphite is assumed to be disturbed by presence of sp3 bonded carbon, the D peak will develop and rise in magnitude in the Raman spectrum. When material is composed entirely of sp3 bonded carbon (and/or olefinic chains connected by sp3 bonded carbon), the D peak should not be detected, because there exist no benzene clusters that give rise to disorder regions, which are responsible for the D peak. In other words, in a completely sp3 network the ID /IG ratio will be zero. Schwan [79] thus suggested that the ID /IG ratio would go through a maximum if it is plotted versus G linewidth. Moreover, Tuinstra and Koenig [84] found that the Raman intensity of ID /IG is inversely proportional to the graphite crystallite size (La ) as determined by XRD:

529

MWNT

Graphite Rod 1400

1600

1800

2000

-1

Raman Shift (cm )

Figure 10. Raman spectra in the frequency range 1200–2000 cm−1 for graphene sheets, outer shell, purified MWNTs, and graphite rod. Reprinted with permission from [85], Y. Ando et al., J. Inorg. Mater. 1, 77 (1999). © 1999, Elsevier Science.

Ager [86] has suggested that hydrogen incorporation into DLC films saturates double bonds, thus forming smaller sp2 domains and helping sp3 to cross-link; this increases both the optical gap to higher energy and the hardness of the film. Veerasamy et al. [87] have shown two prominent IR absorption peaks at 2920 cm−1 and 2840 cm−1 . They have assigned these peaks to stretching mode of C H bond in sp3 configuration in C H3 and C H sites, respectively. Optical absorption coefficient [88] of a typical DLC film is reported to be 104 to 105 cm−1 . 13 C NMR spectrum of a-C:H film shows two peaks, one at around 30–40 ppm due to sp3 bonding configuration and the other at about 150 ppm due to sp2 configuration [89]. DLC can thus be amorphous carbon (a-C) or hydrogenated amorphous carbon (a-C:H) with a bandgap in the range of 1 to 4 eV according to the fraction of sp3 bonding. Sharon et al. [16] have shown the dependence of bandgap with ratio of sp2 /sp3 of DLC film (Fig. 11) prepared from pyrolysis of camphor. 5.1.3.3. Photovoltaic Effect Because DLC is known to possess a bandgap in the range of 1.0 to 4.1 eV, there have been some efforts to make a photovoltaic cell from semiconducting carbon like DLC. One of the constituents of most photovoltaic cells developed is inorganic semiconducting material such as Si, GaAs, etc. This has raised a question whether photocurrent obtained from such cell is due to inorganic semiconductor (e.g., due to Si or GaAs) or due to carbon as well. It is also questioned whether in these cells, carbon is acting as a metal, thus forming a metal-Schottky-type cell. In addition, none of these photovoltaic cells could be more economical than the silicon photovoltaic cell. Nevertheless, following are a few examples of

Carbon Nanomaterials

530

thin films has also been studied exhaustibly, principally by Pleskov and his co-workers [97–101] and by Fujishima and his co-workers [102]. Ago et al. [103] have made a photovoltaic cell of configuration “Al(transparent)/PPV(poly(pphenylene vinylene)/MWNT/ITO (glass substrate).” They studied its current-voltage characteristics as well as its quantum efficiency (1.8%). A heterojunction carbon solar cell of configuration “Al/n-C60 /p-Si/Al” has been reported by Katz et al. [104]. Under natural sunlight conditions the shortcircuit current density was found to be 42 A cm−2 and the open-circuit voltage was 322 mV. A significant rectification property is obtained in the I-V characteristic of the junction. In spite of this work, there is a need to make efforts to develop a homojunction carbon photovoltaic cell which is cheaper than the silicon photovoltaic cell, giving efficiency at least 10% so that it can be commercially viable.

2.0

Bandgap (eV)

1.6

1.2

0.8

0.4

0.0 20

25

30

35

40

45

50

sp2-carbon (%)

Figure 11. Variation of the bandgap of camphor pyrolyzed carbon as a function of the sp2 carbon contents in the grown films. Reprinted with permission from [16], M. Sharon et al., Mater. Chem. Phys. 56, 284 (1998). © 1998, Elsevier Science.

the attempts made in developing a photovoltaic cell utilizing carbon as one of the constituents of the cell. A photovoltaic cell of configuration “C/n-Si” fabricated from the carbonaceous amorphous thin film prepared from the pyrolysis of 2,5-dimethyl-p-benzoquinone in argon atmosphere at 500  C has been studied exhaustibly by Hong-an et al. [90]. They have also given the band diagram of the junction. Since the optical bandgap of carbon was observed to be 0.25 eV and the nature of carbon (n- or p-type) is not known, this photovoltaic cell may be considered as metal-Schottky-type cell, where carbon behaves like a metal. Recently, Romero et al. [91] have fabricated a carbon nanotube/BCHA-PPV. This shows a rectifying behavior as well as photocurrent. A semiconducting polymer, commercially available, soluble di-cholestanol derivative of poly(p-phenylenevinylene), poly(2,5-bis(cholestanoxy)-1,4-phenylenevinylene) (BCHAPPV), was used as one component of the cell. The surface of a carbon nanotube was covered with a solution of BCHA-PPV in xylene under an overpressure of 2 atmospheres of argon in order to allow the polymer to penetrate the space between the tubes. A final layer of the polymer was subsequently spin cast onto the carbon nanotube. A second metallic contact of aluminum or gold thermally evaporated onto the polymer completed the device. Sharon et al. have developed a photoelectrochemical cell with DLC, carbon prepared from camphor [92]. They also developed a homojunction [93–94] carbon solar cell from semiconducting carbon obtained by pyrolysis of camphor of configuration p-C (100 nm)/n-C(100 nm)/n-Si(380 nm) giving an efficiency of 1.2%. Soga et al. [95] have fabricated homo- and heterojunction carbon solar cells of configuration “n-C/p-Si,” “p-C/n-Si,” and “n-C/p-C/p-Si” using amorphous carbon (a-C and a-C:H). The efficiency of these cells is about 2.1%. A p-n junction has been formed by bonding n-GaAs with p-diamond epitaxial layer [96]. A photoelectrochemical behavior of semiconducting boron-doped diamond

5.1.3.4. Electron Field Emission Effect Details of theory and techniques to study electron field emission have been discussed in the diamond section. Here we shall discuss this behavior in relation to DLC films only. Field emission from DLC is generally found more easily than from diamond. It is observed that the substrate over which diamond-like film is deposited has an essential impact upon the electron emission characteristics of the film. Either the electron properties of the interface itself or the growth morphology and structure of the film influence the electron field emission properties. When a diamond-like film is deposited on a flat substrate through a mask, the emission sites are most abundant at the edges of the film. It is likely that the most favorable conditions for emission will always be attained somewhere at the film edge [105]. Xu et al. [106] have proposed that, in the DLC films consisting of a mixture of sp3 and sp2 bonds, the sp3 bonds in the clusters are the electron emission sites, whereas the sp2 bonds in the clusters play the role of a transport path for supplying electrons. Cheng et al. [107] have studied the electron emission behavior of DLC film deposited over Cr/Si tip by laser ablation process (600 mJ laser energy). They reported that as the temperature of substrate increases, Raman shift at around 1550 cm−1 increases, which in turn increases, the current density of 340–436 A cm−2 . This suggests that as the sp2 content in film increases, emission current also increases. This observation also implies that emission current may increase with increase in sp2 concentration up to some optimum concentration of sp2 ; otherwise graphite should have given the maximum emission current. There has been considerable discussion about whether field emission from diamond and DLC is intrinsic or due to surface damage caused during its formation. There is a need to examine this behavior in detail. Sharon et al. [16] have suggested that carbon with 33.3% sp2 behaves like an intrinsic material, and while higher concentration makes it n-type, lower concentration of sp2 makes it p-type carbon. It would be interesting to study the electron field emission with carbon possessing different concentration of sp2 carbon to establish any possible relationship with intrinsic character of carbon. Threshold emission field for DLC is 20–40 V m−1 which can be lowered to about 10 V m−1 for optimized films, and to 5 V m−1 for nitrogen-doped films [108, 109]. On the contrary, Kuo et al. [110] have observed that hydrogenated

Carbon Nanomaterials

DLC film grown in presence of NH3 needs higher threshold voltage than film grown in presence of nitrogen for the electron field emission. They suggested that nitrogen doping is more effective with NH3 than with nitrogen. It appears from the discussions that exact reasons for field emission with DLC films need more rigorous studies to arrive at some definite conclusion.

5.2. Fullerenes 5.2.1. Structural Aspect The discovery of C60 , one of the family members of fullerene, was first reported in 1985 when Kroto et al. [3] measured a carbon cluster containing 60 atoms arising from their laser vaporization apparatus. The C60 molecules contain 12 pentagons and 20 hexagons. In the C60 molecules each carbon atom is bonded to three others by two longer bonds (length ∼145 pm) and one shorter bond (bond length ∼140 pm). This stable pure carbon molecule was found to have geodesic properties of a truncated icosahedral hollow cage and was named Buckminsterfullerene (C60 ). An excellent and informative review on the growth of carbon nanostructure and fullerenes is given by Popov and co-workers [111]. Fullerene family are the most ordered structures among the known carbon clusters. They are said to be ordered structures because other forms of carbon are believed to be amorphous (e.g., carbon black, charcoal, active carbon, etc.). Crystals of C60 formed by vacuum sublimation have a face-centered cubic (fcc), crystal structure at room temperature a0 = 1417 pm. Those grown from solution have a variety of crystal structures depending upon the solvent used, for example, fcc, hexagonal close-packed (hcp), or orthorhombic structures. Since there are many good reviews, in-depth discussions on this material will not be made.

5.2.2. Properties The discovery of Buckminsterfullerene has created a torrent of activity spanning the scientific disciplines. It has been speculated that the fullerenes should be excellent lubricants due to their low intermolecular shear strength and high volume compressibility. The optical and electrical properties and the electronic structure of C60 are still a matter of dispute. For example, the bandgap of C60 has been reported in the range of 1.5–2.15 eV by various authors [112]. Interestingly, Faiman et al. [113] have shown that the optical bandgap of C60 thin film can be shifted at will by controlling the deposition conditions leading to effective optical bandgap in the range of 1.5–2.5 eV. Kuzmany et al. [114] have shown a heterojunction photovoltaic solar cell (Al/ C60 /p-Si/Al) giving a open-circuit voltage (under the sunlight illumination) of 322 mV. Licht et al. [115] fabricated a photoelectrochemical solar cell with C60 with various redox electrolytes (aqueous and nonaqueous) and have shown that such type of cell may be feasible with this material provided this material is cheaply available. Some of the properties of crystalline forms of carbon are shown in Table 2. More information is becoming available about the biological properties of C60 and its derivatives. For example, C60 inhibits the replication of simian immunodeficiency virus

531 Table 2. Properties of diamond, graphite, and C60 . Property

Diamond [116] (cubic)

Bond length (pm) Density (g cm−3 ) Bulk modulus (Gpa) Young’s modulus (Gpa) Melting point (K) Thermal conductivity (W m−1 K−1 )

154 352 442 1054 4500 15000

Graphite [117] (hexagonal) 142335 226 286 1020363 4450 2800

C60 [118] fcc 146144 172 68 16 1180 04

(SIV) in vitro and the activity of Moloney murine leukemia virus (M-MulV) reverse transcriptase [119]. A polyfullerenol has been found to suppress the levels of the microsomal enzymes in vivo and decrease the activities of P450dependent monooxygenase and mitochondrial oxidative phosphorylation in vitro [120]. Water-soluble polythylene modified C60 accumulated at the site of tumors during testing, and after light irradiation the volume of the tumor was significantly reduced [121].

5.2.3. Synthesis Much evidence suggests that fullerenes are a by-product of soot formation; on the other hand, there are observations pointing out that soot produced in fullerene processes is somewhat special compared to other soots. Small carbon clusters (less than 12 atoms) have been observed in sooting flames, in acetylene pyrolysis, and in carbon plasma produced by different methods. It is suggested that fullerenes and soot have the same precursors [122]. C60 has been produced from carbon materials other than graphite, such as coals, polymers, benzene soot, glassy carbons, kerosene, coorongite camphor [123, 124], etc. Unique fullerene-related materials to be synthesized include metallofullerenes and nanotubes and polyhedra. Many metals like cesium, potassium, barium, etc. have been successfully encapsulated into a fullerene cage. Laser vaporization [125] of graphite disk impregnated with lanthanum oxide gives a soot containing La2 C82 .

5.3. Carbon Nanomaterials 5.3.1. Structural Aspects There are many mechanisms proposed to elucidate the formation of carbon nanotubes. One of the most accepted mechanisms is formation of carbon nanotube over a molten catalyst. A nucleation of metal and hydrocarbons takes place on the particle surface, it diffuses, and dissociates at the contact angle of the particle with the reactor wall. A carbon shell is produced which develops by lateral growth into a carbon nanotube. A good description of nanotubes and various aspects related to carbon nanotubes are nicely dealt with in an article published by Kroto and co-workers [126]. Carbon nanotubes consist of graphene sheets wrapped to a cylinder. In other words, nanotubes consist of concentric hexagon-rich cylinders, made up of sp2 hybridized carbons, as in graphite, and terminated by end-caps arising from the presence of

Carbon Nanomaterials

532 12 pentagons (six per end). It is possible to construct a cylinder by rolling up a hexagonal graphite sheet in different ways. There exist three forms of nanotubes (NTs): achiral NTs of the “armchair type,” achiral “zigzag” NTs, and chiral or helical NTs (each pair of the hexagon sides is arranged at an angle other than 0 and 90 relative to the NT axis). First two of these are “nonhelical” in the sense that graphite lattices at the top and the bottom of the tube are parallel. In the armchair structure, two C C bonds on opposite sides of each hexagon are perpendicular to the tube axis, whereas in the zigzag arrangement, these bonds are parallel to the tube axis. The structures of these NTs are usually described by means of two indices, n and m, which are related unambiguously to the NT diameter (d) and the chiral angle (!, characterizing the deviation from the “zigzag” configuration and ranging from 0 to 30 ): √ d = a 3 n2 + m2 + mn/ √ ! = arctan − 3m/ 2n + m where a is the interatomic distance in the planar network. (n m) nanotubes can be characterized by the chiral vector Ch = na1 + ma2 on the honeycomb lattice of a flat graphene sheet, where a1 and a2 denote the lattice vectors of graphite. The cylinders are formed by rolling up graphene sheet so that sites defined by lattice vector R fall on top of sites given by the lattice vector R + Ch . The chiral angle ! is given with respect to the (a1 + a2 ) direction. Thus achiral NT of the “armchair” type are characterized by indices

n m, those of the “zigzag” type have indices (n, 0) and chiral ones are described by (n, m). Because of the sixfold symmetry of the honeycomb lattice, several different integer pairs (n, m) describe equivalent tubes. Dresselhaus and co-workers [127] observed that the intershell spacing (d002 ) between the two graphene sheets of the nanotubes ranges from 0.34 to 0.39 nm and d002 increases as the tube diameter decreases. The empirical equation for the best fit to the data is found to be d002 = 0344 + 01 e−D/2 (nm) for

D≥0

where D is the inner tube diameter. The intershell spacing decreases exponentially and approaches 0.344 nm for tube of diameter greater than ∼10 nm. Apart from these classifications, two kinds of carbon nanotubes are now available: multiwalled nanotubes (MWNTs) and single-walled nanotubes (SWNTs). The diameter of MWNTs is typically in a range from 10 to 50 nm, and length is more than 10 m. On the other hand, SWNTs are much thinner, being only 1.0 to 1.4 nm in diameter, while the length is on the order of 100 m. The number of walls in MWNTs is theoretically unlimited but normally it does not exceed a few dozen. The distances between the neighboring shells are close to the interlayer spacing in graphite (0.34 nm); thus, the smallest diameter of carbon NT is ∼0.7 nm. The diameter of the second and subsequent coaxial atomic layers is stipulated by the diameter for the innermost layer. In this connection, the structure of NT resembles that of onion fullerenes; if the inner shell is C60 , the second shell is C240 , the third one is C540 , etc. In most cases, the tips of NT are covered by hemispherical or conical caps, which

contain not only hexagons but also pentagons, in which the configuration of the carbon atoms is less stable. These caps are somewhat more chemically reactive than the lateral surfaces. Raman-scattering spectra [128] of single-walled nanotubes measured with incident laser wavelength from 488 to 783 nm shows two peaks, one at 1590.9 cm−1 and the other at 1567.5 cm−1 , with broad peak at 1549.2 cm−1 as compared to graphite peak at 1580 cm−1 . Their measurements suggest that as the diameter of the nanotubes increases, the resolution of Raman-scattering peaks, especially of 1567.5 cm−1 and 1549.2 cm−1 , diminishes and finally becomes one broad peak around 1580 cm−1 . When the diameter becomes about 5 nm, it is exhibited by multiwall nanotubes, polyhedral graphitic nanoparticles, or even graphitic layers. The XRD patterns of graphene sheets, hard shell, purified MWNTs, and graphite rod are shown in Figure 9. In the XRD pattern of MWNTs, the (002) and (004) peaks shift to a lower angle as compared to those from raw graphite, indicating the wide d002 spacing of MWNTs. One of the useful aspects of carbon nanometer tubes is their size-dependent physical and chemical properties. Theoretically, nanotubes are expected to be either an insulator or metal depending upon their diameter [129]. There have been some efforts to establish experimentally as well as theoretically a relationship between the diameter of the semiconducting carbon nanotubes and the bandgap. Smalley and others [130, 131] have reported a relationship between the bandgap and the diameter of the CNT which has been found to be in good agreement with the theoretical estimation [132]. Amaratunga et al. [133] have measured the activation energy (∼0.6 eV for high-temperature range) for the electrical conductivity as well as the Hall mobilities (275 × 10−5 cm2 V−1 s−1 ) and optical bandgap (in the range of 1.5–2.0 eV) of carbon nanotube prepared under different conditions. It is interesting to note that Hall mobility is very low compared to diamond (1000–2200 cm2 V−1 s−1 ). Moreover, there seems to be a large difference between activation energy for electrical conduction and optical bandgap, suggesting that CNTs may possess many defect states promoting electrical conduction.

5.3.2. Preparation of Nanotubes 5.3.2.1. Carbon Arc Process Iijima and co-workers were the first to discover carbon nanotubes in the deposit on the cathode of carbon arc [134, 135]. Since then, there have been various approaches adopted to prepare carbon nanotubes. Rakov [136] has given a very exhaustive review on methods for preparation of carbon nanotubes. Nowadays, carbon nanotubes are produced by diverse techniques such as arc-discharge, pyrolysis of hydrocarbons over catalysts, laser evaporation of graphite, condensed phase electrolysis, for example, electrolysis of metal salts using graphite electrode, etc. All these techniques would be difficult to discuss in detail here. Hence, only a few of them are highlighted here. The arc-discharge method is similar to that used for the synthesis of fullerenes. An inert gas atmosphere (preferably He) is flowed through a reaction vessel [137] at a controlled pressure. Two graphite rods constitute the electrodes,

Carbon Nanomaterials

between which a potential difference is applied. As the rods are brought closer, a discharge occurs resulting in formation of plasma. A deposit, which may contain carbon nanotubes under certain conditions, forms on the large negative electrode (cathode) while the smaller positive electrode (anode) is consumed. When a metal catalyst is to be used along with graphite, a hole is drilled in the carbon anode and it is filled with a mixture of metal and graphite powder. In this case, most nanotubes are found in the soot deposited on the arcchamber wall. It has been found that vaporization of an anode containing Co, Co-Ni, Co-Y, Co-Fe, Ni, Ni-Y, Ni-Lu, and Ni-Fe gives deposits looking like a lace collar or a soft belt being formed around the cathode deposit. They contain SWNTs mixed with amorphous carbon and metal particles. Some catalysts (Cu, Cu with Ni, Pt, Y, or Fe; Ni, Ni with Y, Lu, or Fe) cause the formation of a “web.” The addition of sulfur to Co catalysts also results in a greater amount of the web-like products. Higher yields of SWNTs with diameter ranging from 0.55 to 6 nm have been obtained with the positive graphite electrode mixed with different metals (Ni, Fe, Co, La) [138]. Since sulfur itself does not catalyze the formation of nanotube, its role seems to reduce to the stabilization of dangling bonds. The addition of Bi and Pb also causes increase in the NT diameter. 5.3.2.2. Catalytic-Assisted Pyrolysis (Chemical Vapor Deposition ) Pyrolysis of hydrocarbon (chemical vapor deposition) has also been extensively used for this purpose. Catalytic pyrolysis of hydrocarbons had been used to prepare carbon fibers even before the discovery of NT and fullerenes [139]. In this method, chemical precursor along with inert carrier gas is allowed to pass through a quartz tube maintained at some specific temperature. Organic precursor decomposes and the products are deposited on alumina plate. This method has given large quantity of nanotubes. Under this method one could use one furnace [140] and allow the vapor of precursor to enter the furnace for the pyrolysis. Alternatively, two furnaces could be used [141], one for pyrolysis and the other for vaporization of the precursor. Metal nanoparticles employed in the CVD synthesis of carbon nanotubes have been prepared by varieties of methods. Catalysts have also been used for this purpose. Catalyst powder of nanoparticles size could be kept in the furnace over which vapor of hydrocarbon is catalytically decomposed to form carbon nanotubes. Catalysts could also be mixed with precursor and vaporized to enter the furnace along with the vapor of precursor. Alternatively, catalysts could be in form of thin film deposited over some ceramic materials. One method is to use nanoporous or mesoporous substrates, such as zeolites [142], porous Si [143], and anodized Al2 O3 [144], in which metal nanoparticles are confined through evaporation or impregnation. Another method is to etch a metal substrate or a metalcoated substrate by laser ablation [145] or plasma treatment [146]. The physical etching generates metallic nanoparticles, which play a role as catalysts for nanotubes growth. Many parameters, including the temperature and duration of the treatment, the gas composition and flow rate, and catalyst nature and size affect the nature of carbon species in the resulting materials [147]. For each type of precursor, one has to establish the best conditions to get the desired type

533 of carbon nanotubes. One of the best and easiest methods is to follow Taguchi Analysis [148]. It has been observed that nickel-catalyzed decomposition of CH4 (99.99%) and H2 (in 9:1) helps the growth of carbon nanotubes at 600  C better than either 500  C or 700  C. In general, admixtures of Li, Cu, Ag, Zn, Cd, B, Al, In, Y, La, lanthanides, Si, Ge, Sn, Ti, Hf, Pb, Bi, S, Se, Cr, W, Mn, Ru, Pd, Pt, mixtures of two metals or of a metal with a nonmetal, and several carbides and oxides have been tested as catalyst for such purpose [149–153]. Huang et al. [154] have synthesized large-area highly oriented carbon nanotubes at temperature below 666  C by plasma-enhanced hot-filament chemical vapor deposition. Acetylene gas is used to provide carbon and ammonia gas is used for dilution and catalyst. Pyrolytic methods have been used to generate helical nanotubes [155], hemitoroidal nanotube caps [156, 157], and bent nanotubes [158]. Pyrolysis of metallocene such as ferrocene, cobaltocene, and nickelocene yields carbon nanotubes and metal-filled onion-like structure [159, 160]. Decomposition of Fe and Co phthalocyanines is also accompanied by the formation of nanotube, whereas Cu phthalocyanine gives neither nanotube nor graphitized particles. Ferrocene has been used as the floating catalyst in the pyrolysis of thiophene giving relatively thick “ropes” and ribbons colored silvery black. The longest ropes were 3–4 cm long and have a diameter of 0.1 mm. Lee et al. [161] observed that nickel deposited over SiO2 substrate dipped in dilute solution of HF for 100–200 s, or etched by NH3 , results in Ni particles acting as the nucleation seeds for the growth of carbon nanotubes. They could get vertically well-aligned CNTs of length 50 m. Cao et al. [162] have been able to get bundles of aligned MWNTs by pyrolysis of xylene containing ferrocene. In a horizontal furnace maintained at 800  C, mixture of hydrogen (150 sccm) and argon (900 sccm) has allowed to carry drop by drop a solution of ferrocene in xylene (0.12 g/ml) at an interval of 2 min. This arrangement gave bundles of MWNTs of average length 60 m. Cheng et al. [163] have developed a high yield, low cost, and continuous nanotube growth process. Benzene or methane is transported to the furnace maintained at 1100–1200  C with carrier gas hydrogen along with ferrocene as a catalyst. To enhance the growth of SWNTs a sulfur-containing additive (thiophene) was employed. Helically coiled nanotubes [164] have been prepared by catalytic decomposition of benzene on Co/SiO2 catalyst at 873 K. They have a 10 nm coil diameter, 50 nm coil pitch, and around 1 m length. Carbon nanotubes have been synthesized by chemical vapor deposition at 700  C and 800  C using Niphthalocynanine as starting material. The tubes grew perpendicularly on the quartz surface. This precursor was preferred because Ni-phthalocyanine (NiC32 N8 H16 ) can supply Ni, C, N, and H at the atomic ratio of 1:32:8:16 to preparation system [165]. Zhao et al. [166] have developed a simple technique to grow aligned MWNTs. Al sheet is electropolished in phosphoric/glycerol solution at 13–15 V for 10 min at 70–80  C until a mirror finished is obtained. This Al sheet is anodized in 0.3 M oxalic acid solution for about 1 hr under 40 V at 25  C. This sheet is immersed in 5% H3 PO4 for some time

534 to remove partial alumina template, and then it is anodized again under the same condition. This gives a regular pore structure. Over this sheet cobalt metal was deposited in the pores by an ac current source of 50 Hz. Alumina membrane is removed from substrate by immersing in saturated HgCl2 solution. By this process, a deposition of Co in the pores was achieved. This system was used for depositing carbon nanotubes by keeping it in a furnace at 650  C and passing acetylene vapor. In this way, carbon nanotubes were deposited in the pores of the membrane. Advantage of this method is that carbon nanotubes are deposited as aligned nanotubes over metallic aluminum plate. Recently Maruyama et al. [167] have demonstrated a simple catalytic vapor deposition technique to synthesize high-purity SWNTs at low temperature (550  C) by using methanol as the carbon source. It is suggested that the etching effect of decomposed OH radical attacking carbon atoms with a dangling bond, impurities such as amorphous carbon, MWNT, metal particles, and carbon nanoparticles are completely suppressed even at relatively low reaction temperature. Alcohol was sucked into the furnace rather than pushing through a carrier gas. Sharon and his group [168] have been trying to synthesize carbon nanotubes by pyrolysis of natural precursors. Their group has been able to synthesize various forms of carbon nanotubes from natural precursor like turpentine and camphor (Fig. 12). 5.3.2.3. Laser Technique Laser technique has been successfully used for the growth of carbon nanotubes. Singlewalled carbon nanotube yield depends upon the power of laser beam (532 nm) used to ablate carbon target, irradiation time, the composition of the target, the pressure and flow rate of inert gas, and the target temperature. Kroto and coworkers [169, 170] pyrolyzed 2-amino-4,6-dichloro-s-triazine over cobalt thin films (deposited on an inverted silica substrate) and etched using laser technique. A bundle of aligned nanotubes of uniform length of around 100 mm and diameter of 30–50 nm grown perpendicularly over the substrate were observed. Guo et al. [171] have proposed an original method for the synthesis of SWNTs in which mixture of carbon and transition-metal are vaporized by a laser impinging on a metal-graphite composite target. They pointed out that in contrast to the arc method, direct vaporization allows far greater control over growth condition permitting continuous operation, and produces nanotubes in higher yield and better quality. Rinzler et al. [172] have been able to develop laser ablation technique to produce SWNTs yielding in the range of 60–90 vol.% of the carbonaceous soot material and at a production rate of about 1 g day−1 . There is a good review on the production of carbon nanotubes by Wolfgang et al. [173]. Some standard conditions for synthesis of nanotubes (SWNT and MWNT) using laser are given in Table 3. A self-explanatory schematic diagram of the laser evaporation technique is shown in Figures 13 and 14. 5.3.2.4. Electrochemical Method Electrochemical method has lately been tried for the growth of carbon nanotubes. Matveev et al. [174] have been able to get MWNTs from acetylene at room temperature for the first time. Gaseous acetylene was purified by passing it through iron chloride solution. Gas was then dried by passing it through

Carbon Nanomaterials

a layer of freshly melted KOH flakes. Purified acetylene was dissolved in liquid ammonia to get 15–20 mol% acetylene solution. This solution was electrolyzed at 233 K using n-Si (100) electrodes (separated by 4 cm) at 150 V for 5–10 h. Carbon curled nanotubes were found on the electrodes and exhibited a high aspect ratio (length/diameter) >1000. Liquid ammonia helps to dissolve hydrocarbon as well as to initiate chain radical reactions in the solution of hydrocarbon to facilitate growth of carbon nanotubes. Windle and co-workers [175] as well as Kroto and co-workers [176] have synthesized carbon nanotubes and nanoparticles electrolytically from normal graphite in molten alkali chlorides such as LiCl, NaCl, and KCl. Carbon nanotubes as well as nanoparticles and onion-like structures have been generated. The nature of products depends upon several factors including the electrolysis voltage and current, depth of electrode immersion in the electrolyte, the length of time the current is maintained, and the electrolyte. It is reported that carbon consumption for nanotube formation occurs at the graphite cathode where alkali metals form during the electrolysis. However, the major hurdle is the nonconducting nature of most organic precursors and insolubility of the organic materials in polar solvents or in solvents with high dielectric constant. Hence, though this technique is very suitable, these limitations have to be overcome before it can be used for large-scale synthesis of carbon nanostructures. 5.3.2.5. Effect of Particle Size of Catalyst It has been observed that the particle size of the catalyst plays a role in the growth of MWNTs as well as SWNTs. For multiwalled nanotubes, the particle size needs to be larger than the nanotube diameter since catalytic process requires carbon species to diffuse through catalyst like Ni in order to grow multiwall nanotubes. For the growth of single-walled carbon nanotubes, the particle, which is attached to one end of the nanotubes, has to have the same size as the nanotube diameter where the carbon shell on the particles grows and transforms to a carbon nanotube. These relationships agree with the experiential evidences [177–179]. 5.3.2.6. Opening and Insertion of Metal into Carbon Nanotubes Carbon nanotubes prepared by any of these methods would be normally closed from one end of the tube. Hence, efforts have also been made to open one of the ends so that some chemistry could be done with these carbon nanotubes. One simple method is by a chemical route for opening multiwalled carbon nanotubes. This is achieved by boiling suspensions of nanotubes in aqueous nitric acid for several hours (∼24 h) at 150  C [180]. Several efforts have been made to encapsulate metallic particles like iron into carbon nanoparticles [181]. After opening the closed end of carbon nanotubes, filling in-situ with metal oxides such as NiO [182] or even metal like palladium has been achieved [183]. Ajayan and co-workers have been able to coat carbon nanotube with a single layer of metal oxide (V2 O5 ) [184]. A novel electrolytic technique has been discovered by Kroto and co-workers [185] which yields fully filled metallic Sn nanowires. They also succeeded in coating carbon nanotubes with Sn nanowires by electrolyzing graphite electrode in molten mixture of LiCl and SnCl2 under argon atmosphere.

Carbon Nanomaterials

535

500 nm

250 nm

1 µm 8..80K

200 nm 38.0K

1759

(A)

(B)

(C)

1762

(D)

200 nm

500 nm

1 µm 11.5K

(E)

200 nm 50.0K

1690

(F)

(G)

0708

(H)

200 nm

100 nm 88.0K

0895

10 µa

(I)

(J)

(K)

(L)

Figure 12. Various types of carbon nanomaterials synthesized from turpentine and camphor. Bamboo-shaped branched CNT (A, B, C, F); Y-junction CNT (D, E); coiled double-walled CNT (G, I, J); enlarged portion of (G) to show double-walled CNT (H); formation of branching in CNT (K); Yofork type CNT (L) [168].

In this fashion many types of elements such as Pb, Bi, Cs, S, Se, etc. have been introduced into the carbon nanotubes easily. It has also been possible to dope potassium into single-walled nanotube. Due to donation of electron by potassium, its doping causes generation of electron carriers in the carbon nanotube. The Hall mobility for the electron is found to be 20–60 cm2 V−1 s−1 , a value which is similar to hole effective mobility in nanotubes [186].

5.3.3. Purification of Carbon Nanotubes When one measures the physical properties of MWNTs or SWNTs and plans their practical application, impurities like carbon nanoparticles are not desirable. There is a good review on the purification of SWNTs by Chiang et al. [187]. Many attempts have been made for purification of MWNTs [188]. Carbon single-walled nanotubes or multiwalled carbon

Carbon Nanomaterials

536 Table 3. Some favorable standard laser ablation conditions for synthesis of nanotubes. Source conditions

Source Single pulsed Nd:YAG Double pulsed Nd:YAG Cw-CO2

" 1064 nm

Solar light

Sunlight

Target

Temperature

Gas

Results

" 532 nm

Graphite (G)

Furnace 1200  C

Argon

MWNTs

" 1064 nm

G (98)/Ni (0.6)Co (0.6) at.% G (95)/Ni (4)/ Y (1) at.% G (96)/Ni (2) Co (2) at.%

Furnace 1200  C

Argon

No furnace

Argon

No furnace

Argon

SWNTs 60–90 vol.%, 1 g/day SWNTs 80 vol.%, 130 mg/h MWNTs and few SWNTs

Source: Reprinted with permission from [173], K. Wolfgang et al., Carbon 40, 1685 (2002). © 2002, Elsevier Science.

nanotubes are typically accompanied by other carbonaceous materials when synthesized. A number of purification methods have been developed. They can be categorized into four major methods: (i) acid oxidation, (ii) gas oxidation, (iii) filtration, and (iv) chromatography. In the acid reflux procedure raw nanotubes are refluxed in nitric acid to oxidize the metals and impurity carbon. Sharon and co-workers [189] have purified carbon nanotube by refluxing it with concentrated HNO3 for 24 h. TEM images of CNT films are shown in Figure 15. While Figure 15a shows the CNT before purification, Figure 15b shows CNT after purification. Globular particles at the end of the CNT (Fig. 15a) are the Co catalyst, which are absent in the purified material (Fig. 15b). Unfortunately, acid-treated nanotubes are thought to have carboxylic acid groups at the tube ends, possibly, at defects on the sidewalls. Hence, acid-purified tubes must be checked before its application. Gas phase oxidation is used for the purification of MWNT [190]. This process suffers from the disadvantage of destroying SWNT. Combination of acid wash and gas oxidation process has also been applied for purification. Oxidation of acid-treated products at 550  C for 30 min has given a good result to purify SWNT. TGA studies indicate that these purified tubes can withstand temperature up to 600  C in air. Microfiltration and chromatography methods have also been found suitable, but are not so popular. Purification of multiwalled carbon nanotubes containing necessary carbon particles has been also achieved by heating the sample with infrared radiation in air [191, 192]. Unfortunately, more than 95% of the nanotubes are also destroyed in this process and the remaining nanotubes lose their protective end-caps. Recently, it has been suggested to

treat carbon nanomaterials (impure) with HCl and then with HNO3 .

5.3.4. Properties of Carbon Nanotubes Before discussing various applications of carbon nanotubes, perhaps it would be appropriate to spend some time discussing the general properties of carbon nanotubes. Measurements of Young’s modulus show that single nanotubes are stiffer than commercial carbon fibers. MWNTs can pass a very high current density of the order of 108 A cm−2 without adverse effect [193]. Properties of nanotubes depend on atomic arrangement (how the sheets of graphite are rolled), the diameter and length of the tubes, and the morphology or nanostructure. Work functions of MWNTs and graphite have been evaluated from the secondary electron in UPS HeI spectra by Ago et al. [194]. These values are given in Table 4. X-ray profile of purified and annealed SWNT has been given by Smalley and co-workers [196], showing welldefined reflections at 2! corresponding to 6.3, 10.5, 16.5, and 22 . An essential problem in understanding conduction in nanotubes is a crossover from metallic to nonmetallic behavior as the temperature is lowered [197]. Recent electrical transport measurements on bulk nanotubes, individual Quartz-Tube

Plume

Target Furnace Nd:YAG Laser

Argon

Nanotube Deposit

Co2 Laser

Plume target Water cooled Cu Collector 1200 °C

Argon

Quartz-Tubes Inner: ∅ 2.5 cm Outer: ∅ 5 cm

Figure 13. Sketch of the up-scales Nd:YAG laser evaporation experimental setup. Reprinted with permission from [173], K. Wolfgang et al., Carbon 40, 1685 (2002). © 2002, Elsevier Science.

Figure 14. Drawing of cw-CO2 laser evaporation system. Reprinted with permission from [173], K. Wolfgang et al., Carbon 40, 1685 (2002). © 2002, Elsevier Science.

Carbon Nanomaterials

537 value to approximately 20 V K−1 (at a temperature of 350 K) in air to negative values of approximately −12V K−1 in vacuum, and so ascribed the positive thermopowers observed to the adsorption of oxygen. However, the negative thermopower magnitude seen in vacuum is still larger than typical metallic values instead of being close to zero. Grigorian et al. [199] found a negative thermopower for SWNT mats doped with Cs, again reaching a rather large magnitude of −15 V K−1 near 150 K before decreasing in magnitude at higher temperature. Carbon nanostructured materials are also considered for some electrochemical applications such as membrane support of catalyst for the electrocatalytic reduction of oxygen and the oxidation of methanol [200, 201], storage of lithium [202, 203], and supercapacitors [204, 205]. These properties are discussed later in detail.

(a) 200 nm

5.3.5. Applications of Carbon Nanotubes

(b) 100 nm

Figure 15. TEM image of the CNT (a) before purification and (b) after purification. Reprinted with permission from [189], A. K. Chatterjee et al. Electrochim. Acta. 48, 3439 (2003). © 2003, Elsevier Science.

multilayered tubes, and single-walled tubules have revealed that their conducting properties depend markedly on the degree of graphitization, helicity, and diameter. Carbon nanotubes are found to behave like metal, insulator, or semiconductor, depending upon the method of production— which controls the degree of graphitization, the helicity, and the diameter. Collins et al. [198] have recently found that the thermopower of SWNT thin films changes from a large positive Table 4. Work function of MWNTs.

Sample Purified MWNTs Air-oxidized MWNTs Plasma-oxidized MWNTs Acid-oxidized MWNTs Highly oriented pyrolytic graphite (HOPG) Single-walled nanotube (SWNTs) C60

Work function (eV) Ago [194] Shiraishi [195] 43 44 48 51 44

495 — — — 480

— —

505 65

5.3.5.1. Biochemical Applications SWNTs can be cut into smaller sections using sonication in a mixture of concentrated sulfuric acid and nitric acid. This process makes the tube opened. Once opened, tubes can be filled with a variety of materials, for example, enzymes, protein [206], and DNA [207]. The biomolecules appear to be encapsulated within an environment which offers some protection, and subsequent analysis of the catalytic activity of the immobilized enzymes showed that a significant amount of the enzyme retains its activity. Opening of tube tips has enabled the resultant OH and COOH groups to be functionalized using the standard chemical reaction conditions. For example, SWNTs have been functionalized at the opened tips by reaction with thionyl chloride and octadecylamine: the derivatized tubes are more soluble in organic solvents than their parent compounds and can subsequently be reacted further along the waists of the molecules [208]. This type of research is expected to help in drug delivery to infected human organs. A suitable drug can be inserted into the hollow space of the nanotubes and can be injected into body (with functionalized soluble carbon nanotubes). Drug will start operating slowly at the site, which is infected. 5.3.5.2. Hydrogen Storage by Carbon Nanotubes Good reviews on hydrogen storage by adsorption in carbon nanotubes are given by Darkrim et al. [209] and Simonyan et al. [209, 210]. There are conflicting reports about the magnitude of hydrogen adsorbed by carbon nanotubes. However, soon we shall arrive at some results, which are reproducible. Dillon et al. [211] have reported that crystalline SWNTs have hydrogen capacity of 5–10 wt% at pressure less than 1 bar near room temperature. It has also been observed that some nanotubes can electrochemically store relatively large amount of (110 mA h g−1  hydrogen, corresponding to a hydrogen storage capacity of 0.39 wt%. The reaction is reversible, indicating that the nanotubes can be used to produce electrodes for rechargeable batteries. Considering these properties, Rajalakshmi et al. [212] have made electrodes of purified and open single-walled carbon nanotubes and found them to behave like metal hydride electrodes in Ni-MH batteries, showing high electrochemical reversible charging capacity up to 800 mA h g−1 corresponding to hydrogen storage capacity of 2.9% compared

Carbon Nanomaterials

538 to known AB5 , AB2 metal hydride electrodes. The charging and discharging potential was around −0.8 and −0.6 V versus SCE, respectively. In a separate attempt, Qin et al. [213] reported that MWNTs are promising candidates for hydrogen storage. The electrochemical capacity of MWNTsNi electrode can reach over 200 mA h g−1 at a current density of 200 mA g−1 . MWNT-Ni electrode showed a high rate of discharge capacity and long charge-discharge cycle life. The electrochemical reaction of MWNTs-Ni was controlled by the step of the hydrogen adsorption and the hydrogen oxidation, that is, Ni + H2 O + e− → NiHad + OH−

(this being the rate-determining step)

NiHad + MWNT → MWNTHad + Ni If this could be improved, then low-weight, high-power density Ni-carbon hydrogen battery could become a reality. Zuttel et al. [214] have tried to study the hydrogen storage capacity of carbon samples at room temperature by means of electrochemical galvanostatic measurement in a 6M KOH electrolyte. During the charge process, the water in the electrolyte dissociates at the working electrode (sample, negative electrode) into the adsorbed atomic hydrogen and OH− ions remaining in the electrolyte. The adsorbed atomic hydrogen may intercalate in the electrode or recombine at the surface to molecular hydrogen and diffuse into the electrode or form gas bubbles at the surface of the electrode. During discharge process, the hydrogen in the electrode recombines with OH− ions in the electrolyte to form water. This reaction is accompanied with a charge transfer and therefore, the amount of hydrogen desorbed from the electrode can be measured by measuring the electric charge, which is equal to the product of current and time in galvanostatic setup. In Zuttel’s experiment, there is no evidence that discharged hydrogen has been completely adsorbed in the carbon. Hence, the experiment has to be designed such that one can be sure beyond any doubt discharge current is proportional to quantity of hydrogen gas adsorbed in the carbon matrix. This hydrogen storage is very attractive from the viewpoint of fuel cell technology, which requires a very high density of hydrogen molecules for clean combustion. The temperature programmed desorption (TPD) profiles of closed- and open-tip SWNTs with diameters of around 1.2 nm are compared with the results of activated carbon, which has micropores with diameter of around 3.0 nm. The peak at 133 K is seen to originate due to hydrogen molecules adsorbed inside the contamination amorphous carbon. This hydrogen is not useful because it cannot be used at room temperature. A peak at 288 K found only in the open-tip CNTs suggests that the hydrogen molecules are incorporated inside the hollow spaces of CNTs. The high-temperature desorption (288 K) of a large number of hydrogen molecules is a fascinating property for this application. The heat of adsorption for this site (19.6 kJ mol−1 ) is larger than those for graphite (∼4 kJ mol−1 ) and activated carbons (12 kJ mol−1 for low coverage and 4 kJ mol−1 for high coverage). The well-aligned SWNTs with the diameter of 2.0 nm were predicted to be good candidates for use in vehicles, much better than metal hydrides [215]. It is reported that 4 wt% or higher of hydrogen can be stored in SWNT.

5.3.5.3. Lithium Secondary Battery One of the biggest advantages of lithium battery is that one can get operating voltage in the range of 4.7–5.0 V. The theoretical capacity of lithium metal is considerably high 3862 mA h g−1 . Since lithium reacts vigorously with water, it is difficult to get this power in aqueous solution. Alternative method is to use nonaqueous solution and some device such that lithium could be transported into the solution. Herold [216] reported the first synthesis of Li-intercalated graphite. Later Besenhard [217, 218] observed electrochemical intercalation of alkali metal in graphite that is immersed in a nonaqueous electrolyte. Cyclic voltammetry of graphite in dimethylsulfoxide (DMSO), 1,2-dimethoxyethane (DME), or PC indicated reversible electrochemical reduction at potential more positive than that for Li deposition. This was attributed to an intercalation reaction involving lithium. General formula for intercalated lithium can be Lix Cy ; for x = 1, y is equal to 6 and for x = 05, y is equal to 12. For LiC6 , the electrochemical capacity is equivalent to 372 mA h g−1 C, and 186 mA h g−1 for LiC12 . In some special cases, it has been possible to get 1600 mA h g−1 capacity but soon after first cycle of operation, the capacity falls down to 460 mA h g−1 . Introducing defect by ball-milling is found to improve the capacity to 1000 mA h g−1 . This observation attracted scientists to develop lithium battery using graphite as an electrode. The overall charge (Li intercalation in carbon, de-intercalation of metal oxide) reaction in a Li-ion battery that contains graphitic carbon and a lithiated metal oxide (MO2 , M = Co, Ni, or Mn) is LiMO2 + 6C → Li1−x MO2 + Lix C6 The electrochemical synthesis of Li-intercalated graphite in a nonaqueous electrolyte containing a lithium salt can be represented by xLi+ + 6C + xe− → Lix C6 During charging, Li from the metal oxide is transferred to the carbon to form the negative electrode. Correspondingly, discharge (de-intercalation of Li from carbon) of Li-ion cell involves the reverse reaction. Lithiated carbon, in which the Li species are intercalated between the layer planes (i.e., in graphite) or associated with other sites (i.e., in disordered carbons), are produced by both chemical and electrochemical process. This transport of lithium back and forth during the charge and discharge process has led to the use of terminology such as “rocking chair,” “shuttlecock,” “SWING,” etc., to describe Li-ion cell. After the discovery of carbon nanotube, this idea has been extended to carbon nanotubes. The electrochemical intercalation of lithium in MWNT was achieved by placing lithium metal foil, a polypropylene porous separator soaked with the electrolyte solution, and a cathode prepared by mixing nanotubes with 5% of black carbon and 5% of polytetrafluoroethylene (PTFE) as a binder. Electrolytes are normally prepared with LiPF6 (1M) in carbonate mixture (EC:PC:3DMC) [219]. Shinoda et al. [220], however, has suggested that lithium intercalation is possible with open SWNT. Sharon and co-workers [221–223] have studied the application of carbon nanotubes and carbon nanobeads prepared

Carbon Nanomaterials

from pyrolyzed camphor and observed that capacity of battery is as good as that obtained with graphitic carbon. Slurry of camphoric carbon nanotubes was made with acetylene black and ethylene propylene diene monomer as a binder in a proportion of 86:10:4. This slurry was spread evenly onto a finally polished copper sheet. Lithium metal was used as counterelectrode. A Celgard 2500 microporous polypropylene, wetted with dimethyl carbonate, served as a separator. 1M LiPF6 solution, prepared in 1:1 mixture of dimethyl carbonate and ethylene carbonate, was used as the electrolyte. Typically, the potential of the carbon is >1 V before Li intercalation takes place. In the case of highly graphitized carbon, when current is applied to intercalate Li, the potential initially drops rapidly to near 0.8 V (vs Li/Li+ ) where electrolyte decomposition and the formation of a surface film occur. When these reactions are taking place, the potential remains close to a constant value. Following electrolyte decomposition, the potential declines and the majority of Li intercalation occurs at 100,000 cycles), simple principle and mode of construction, short charging time, safety, and high power density. The energy density of supercapacitors, however, is smaller than that of secondary batteries. The most common applications of double layer capacitor are as button cell, memory backup for electronic equipment. They are also used as auxiliary power sources in small appliances like laptop computers, alarms, VCR, telephones, etc. They are well suited as a backup source because of their energy storage density, low cost, and maintenancefree long life operation. They have been used as a memory backup device because of their high cycle efficiency and the long cycle life. Moreover, recently EDLC is expected as the subpower source for hybrid electric vehicle (HEV) because EDLC can provide high power density (from 0.5 to 20 kW kg−1 ) during few seconds or more, which leads to specific energy densities ranging from 0.5 to 10 Wh kg−1 . The electrochemical capacitor can use either nonaqueous electrolyte system or aqueous electrolyte system. Nonaqueous system can work at high working voltage, but internal resistance is high and needs hermetic seal to avoid moisture absorption. Absorption of moisture spoils the performance of the capacitor. On the other hand, aqueous solvent system has high ion mobility and high dielectric constant, making the internal resistance very small. However, this works at low voltage (around 1.3 V). There are three different types of EDLC supercapacitors: (i) carbon/carbon, (ii) metal oxide, and (iii) electronically conducting polymers. In carbon supercapacitors, the electric charge is stored between a high surface area of carbon electrode/electrolyte interface. In these systems, the use of organic electrolytes allows the increase of the working voltage as compared to aqueous electrolyte. The electrode is composed of a current collector (e.g., nickel foam) in contact with an activated carbon. The current collector has to be electrochemically inactive in the potential window in which the system works. The active material is an activated carbon. Under a powder form, activated carbon (approximately 95%) is mixed with a binder (about 5% mixture of binder like carboxymethylcellulose and PTFE) and sometimes with an electronic conductor in order to reach good mechanical

Carbon Nanomaterials

540 and electronic properties to the electrode. After drying, the active material is laminated on each side of the current collector, to make about 600-m-thick electrode. A basic typical design of the electrochemical capacitor of configuration CNTH2 SO4 H2 SO4 CNT sandwiched between two pieces of carbon paper assembled between Perspex plates is shown in Figure 16. Pellet of CNT (B) is charged with external potential and when needed the same charge can be discharged externally to get the power. In order to increase the capacitance, carbon nanotube needs to be activated. Activated carbon used in supercapacitors must possess the following characteristics: (i) a high specific surface area to ensure high specific capacitance value, (ii) a low resistivity, and (iii) a microstructure well adapted in order to allow good electrolyte accessibility into inner surface of the electrode. The practical electrode material for EDLC is porous carbon such as activated carbons. The high capacitance (100–200 F g−1 ) is due to high specific area (1000 m2 g−1 ) produced by many micropores (2 nm > pore width). In general, it is believed that there is a proportional relationship between the specific surface area and the electric double layer capacitance of the activated carbon. The electric double layer capacitance can be calculated by the amount of electricity passed during the charge or discharge process by using the following equation [234]: c = it/w&V where c = electric double layer capacitance (F g−1 ), i = current (A), t = charge or discharge time (s), w = weight of activated carbon in the electrode (g), and &V = potential change during charge or discharge process. The surface area of activated carbons results from a complicated porous structure which involves pores of different size: macropores (>500 Å wide), mesopores (20–500 Å), and micropores (= 2 (33) e
one can easily see from relations (26) and (31) that 2 Av =
U0 Cv 3

Figure 26. Schematic illustration of the situation at a surface under field emission conditions and the resulting energy distribution of the field emitted electrons (calculated for F = 2700 V/mm and
= 5 3 eV).

⇒


=

3 Av 2U0 Cv

(34)

where Av is the slope of the Fowler–Nordheim plot (28) with respect to the applied voltage U0 and Cv is the exponential factor of the low energy part of the field emission energy distribution measured at the applied voltage. By determining Av and Cv the work function
at the emission site can be determined without the need to know the field enhancement factor . Knowing the work function the local field present at the emission site can be determined from  2me
0 5 (35) F =2 e
Cv

Carbon Nanostructures for Cold Electron Sources

565

In these considerations and derivations a triangular emission barrier as described by relation (7) is assumed. A significant improvement of the emission model results from the influence of the image charge potential on the derivation of the Fowler–Nordheim law and the total energy distribution. The Fowler–Nordheim relation (16) becomes 
 4 2me
1 5 1 e3 2 exp − F vy (36) j= 422

t 2 y 3
eF with A
F  y= =




e3 F 40




(37)

where ty and vy are the so-called elliptical functions which are tabulated [103]. The parameter y is the ratio if the field dependent work function lowering A
F , due to the influence of the image charge potential (Schottky effect) and the work function
of the emitter. The total energy distribution transforms into  2me eF 1 P E = f E! EF ! T  222
2
0 5 ty  
4 2me
1 5 vy × exp − 3
eF 
 0 5 2 2me
tyE − EF  (38) × exp
eF Figure 27 displays the values of the elliptical functions ty and vy. In the field and work function range for carbon field emitters the parameter y varies between 0.3 and 0.4. In this regime the following approximations for ty and vy can be made: ty ≈ 1 02

vy ≈ 0 98–1 1y

(39) 2

(40)

Figure 27. Plot of the elliptical functions ty and vy. The markers represent the tabulated values [103] and the solid lines are approximations to ty and vy.

Substituting (39) and (40) into (37) the Fowler–Nordheim relation becomes
2  e 2me e3 F2 1 exp j= 1 04 422

3
0
0 5  
4 2me
1 5 (41) × exp −0 98 3
eF From relations (41) and (38) one can see that the estimation of the work function
taking into account the image charge potential, using the slope of the Fowler–Nordheim plot Av and the exponential factor Cv , will lead to
=

3 Av 1 02 3 Av = 1 04 0 98 2 Cv U0 2U0 Cv

(42)

This is about 4% larger than without taking the image charge potential into account (34).

2.2.3. Example: Determination of the Work Function
of MWNT Experimental Setup The energy resolved field emission measurements were carried out in an OMICRON surface analysis system (base pressure p < 10−10 mbar) equipped with a VSW EA 125 HR electron energy analyzer. The measurements were carried out in the constant analyzer energy mode with pass energy of 5 eV which corresponds to an energy resolution of 40 meV between 100 and 1000 eV kinetic energy. The high voltage was set to the sample by a Keithly 237 instrument with a ripple voltage less than 5 mV. The field emission spectra were measured between 1 pA and 1 nA total emission current, with count rates of 100 to 106 cps. Figure 28 shows the cut through the sample holder used for the measurements. Details of the experiment can be found in [104]. Results Figure 29 displays a series of FES measured from a sample of MWNTs recorded at room temperature at different applied voltages and therefore different electric fields. The applied voltage U0 was subtracted from the kinetic energy scale of all spectra giving the energy relative to the

Figure 28. Schematic cross-sectional view of the sample holder used for the FES measurements.

Carbon Nanostructures for Cold Electron Sources

566

Figure 30 shows the field emission spectrum from the series in Figure 29 recorded at an applied voltage of U0 = 680 V. The dashed line is a Gaussian curve fitted to the high energy side of the peak, whereas the solid line is a fit according to relation (32). One can see the asymmetric shape of the field emission distribution, where the low energy side of the peak shows the exponential decrease with decreasing energy as expected. Using the definition (33) for Cv the relation for the total energy distribution P E (31) can be simplified to

0 5 P E = f E! EF ! T BF !
 exp Cv E − EF  (43) F Using this relation the low energy side of the FES can be fitted and we can obtain a value of Cv
0 5 F −1 = Cv
0 5 U0−1 −1 = 9 09 (eV)−1 With this value, the value for FNslope , and relation (34) the work function
at the emission site can be determined: Figure 29. Series of FES of a MWNT field emitter recorded at different applied voltages. The inset shoes the FN plot derived from the FE spectra. Av is the slope of the FN plot. Reprinted with permission from [46], O. Gröning, Ph.D. Thesis No. 1258, University of Fribourg, Switzerland, 1999.

Fermi energy EF . The solid lines are fits to the data using relation (32) assuming T = 0 K and taking the spectrometer resolution into account. By integrating the spectra over the energy we obtain a value which is proportional to the emission current density. Therefore we obtain a current– voltage characteristic that can be displayed in a FN plot (inset Fig. 29). The FES in Figure 29 show a very narrow, asymmetric shaped energy distribution (FWHM = 375 meV) of the emitted electrons located at the Fermi energy EF . As discussed in Section 2.1 such an energy distribution is characteristic for Fowler–Nordheim tunneling (Fig. 25a). The straight line in the FN plot (inset Fig. 29) over five orders of magnitude in the emission current proves the Fowler– Nordheim emission for the MWNT field emitter. As the slope of the FN plot with respect to the field F at the emission site depends on the work function alone by
1 5 , the work function
of the emitter can be determined. Yet we have to realize that in a field emission experiment current–voltage characteristics are measured. In the case of two parallel plates acting as anode and cathode separated by a distance d, the applied field is given by F0 = U0 /d. The local field F at the emission site is then F = F0 · . As discussed  is generally not known and the FN plot is plotted with coordinates lnj/F0  vs F0−1 . Hence the slope of the FN plot is proportional to
1 5 −1 . The FN plot depicted in the inset of Figure 29 is given with respect to the applied voltage and the slope FNslope = −Av
1 5 −1 (28) equals 20,152 V. The local field F at the emission site can be expressed by F = U0 · , where  is the field enhancement factor, this time not with respect to the applied field F0 but to the applied voltage U0 . Therefore  has in this case the dimension [m−1 ].

2 A Av
1 5 −1 = v
U0 =
U0 Cv 3 Cv
0 5 U0−1 −1 ⇒


=

Av
1 5 −1 3 2U0 Cv
0 5 U0−1 −1


=

20156 (V) −1 3 = 4 89 eV · 2 · 680 (V) 9 09 (eV)

As can be seen the unknown field enhancement factor  is cancelled out in this calculation. The obtained work function value
= 4 89 ± 0 3 eV is a bit higher than for highly oriented pyrolythic graphite with a work function of 4 55 ± 0 1 eV but is in very good agreement with the work function of sputtered graphite (highly disordered) of 4 9 ± 0 2 eV.

Figure 30. FES spectrum recorded at an applied voltage U0 = 680 V from the FES series depicted in Figure 29. Reprinted with permission from [46], O. Gröning, Ph.D. Thesis No. 1258, University of Fribourg, Switzerland, 1999.

Carbon Nanostructures for Cold Electron Sources

This example shows that the work function
at the emission site can be determined accurately by measuring the I–V characteristic and the energy distribution of the emitted electrons. Knowing the emitter work function
, the local electric field F present at the emission site can be determined from definition (33) using Cv
0 5 F −1 = 9 09 [eV]−1 = 5 68 × 1019 [J−1 ) ⇒

F = Cv




5 68 × 1019 [J−1 )  2me 8 93 × 10−10 [J0 5 ) · =2 e
5 68 × 1019 [J−1 ) = 2 5 × 109 [Vm−1 )

This value for the electric field at the emission site corresponds very well to the electric field of 2700 V/m predicted by the Fowler–Nordheim relation (16).

3. CARBON BASED ELECTRON EMITTERS Since the first observation and investigations of the low field electron emission properties of natural and chemical vapor deposited diamond in the early 1990s [20, 21], the interest in this field has steadily increased until the present day. This interest is not of a scientific nature only (over 150 patents [105] were issued on the subject of diamond field emitters alone) although the physical nature of the low field electron emission from diamond is still under dispute. Our works on carbon thin films show clearly that one has to distinguish strictly between the applied electric field (defined as the applied voltage devised by the anode– cathode distance) and the local field present at the emission site. From the technological point of view the applied field is the relevant parameter to characterize the field emission, as it determines what voltage has to be applied for a given anode–cathode distance. Yet, with regard to the physical nature of the emission, only the local field present at the emission site is relevant. In a field emission experiment in general only the applied electric field can be determined directly and it is very difficult to estimate the local fields on a real cathode surface. The combined measurement of the current–field characteristic and the field emitted electron energy distribution of an emitter allows one to determine independently the local field F and the work function
of the emitter. In all cases we have investigated (CVD diamond, DLC films, and CNTs) we found that the local fields present at the emission site are always in the order of 2500 V m−1 and higher, although the applied field ranges between 1 and 50 V m−1 . The emission process is in all cases ordinary Fowler–Nordheim tunneling with emitter work functions around 5 eV.

3.1. CVD Diamond Thin Film Emitter On CVD diamond films exhibiting a good crystalline quality, with a low defect density and large (>1 mm) crystallites (Fig. 16a), we found in most cases that the field emission showed very distinct, nonreversible activation characteristics

567 [106–109]. Figure 31 shows a typical current–field characteristic of the activation of boron doped good crystalline CVD diamond film. In the first voltage sweep from zero upward, the measured current remained below 1 pA up to a critical field EC . At this field, a sudden and persistent increase of the current occurs. The value of EC can vary significantly with the film under investigation because the occurrence of vacuum discharge depends on many parameters such as residual gas pressure, surface roughness, surface contamination, and anode–cathode geometry [110]. In our setup, typical values for EC ranged between 100 and 400 V m−1 . When the voltage is lowered, a reversible field emission characteristic can be observed at a turn-on field Eon . The values for Eon can range between 2 and 50 V m−1 . Typically the arc formed a small crater in the diamond film. By means of atomic force microscopy and Raman spectroscopy a graphitization of the diamond film around the crater can be observed. This results in an increased local conductivity and an increased roughness in the submicrometer range in this zone. Field emission spectroscopy showed that after the activation process the field emitted electrons originated from the states at the Fermi energy EF indicating the metallic nature of the emission site. The local field F at the emission sites is in the order of 3000 V m−1 and the work function
of the emitter is around 5 eV. CVD diamond films exhibiting bad crystalline quality, nanocrystalline films (Fig. 16b), do not require an activation step. Usually these films are characterized by a Raman spectrum showing a very weak or even absent 1332 cm−1 diamond line but strong features of sp2 bonded carbon. These films exhibit turn-on fields below 5 V m−1 [111, 112]. The current–field characteristics of these films are Fowler– Nordheim-like in the low current regime. In the high current regime deviations from the Fowler–Nordheim law can be observed depending on the resistivity of the deposited

Figure 31. Typical current–field characteristic of the activation of a weakly emitting CVD diamond film. At a critical field EC the discharge activation takes place and leads to a drastically and persistently increased current (1–2). If the field is subsequently lowered (4–5), a reversible field emission characteristic around Eon can be observed (5– 4–3). (The constant current observed for higher fields is due to the current limitation of the voltage source.) Reprinted with permission from [107], O. Gröning et al., J. Vac. Sci. Technol. B 17, 1064 (1999). © 1999, American Institute of Physics.

568

Carbon Nanostructures for Cold Electron Sources

film. From FES an average work function
of 5.7 eV for nanocrystalline diamond can be determined. The local field F present at emission sites is around 2500 V m−1 , whereas the applied fields range between 2 and 10 V m−1 . This implies that strong field enhancement effects are the basis for the low field emission property of this material. Internal field emission and injection of electrons into the conduction band of the diamond and subsequent emission due to negative electron affinity could be ruled out as a possible emission model. This is due to the absence of internal electric fields in the diamond film under field emission conditions. The strong field enhancement effect is most probably due to the surface morphology of the films with a large peak to peak roughness of up to several micrometers and sharp local features on the nanometer scale.

3.2. Diamondlike Carbon Thin Film Emitter Diamondlike carbon films show field emission properties very similar to those of CVD diamond films. On smooth, structureless (with regard to the surface morphology) DLC films, we only could observe discharge activated field emission [113]. This activation can be observed on polymer films as well as on carbon films with high sp3 content. This shows that in the activation step the film acts mainly as a carbon source whereas the chemical nature of the carbon film we start is not of major importance. Nonactivated low field electron emission from DLC films with applied field range of 10–30 V m−1 can be observed when graphitic clusters are present in or on the film. Compared to the smooth parts of the DLC films these clusters show an enhanced conductivity. As in the case of the nanocrystalline CVD diamond the current–field characteristic of the emission is Fowler–Nordheim-like in the low current regime and the FES exhibit the typical shape of Fowler–Nordheim tunneling. Local fields F of 2500– 3000 V m−1 and emitter work functions
around 5 eV have been determined. The investigation of the graphitic clusters on the DLC films by high resolution scanning electron microscopy revealed nanotubelike structures, with diameters down to 7 nm and lengths up to several hundred nanometers (Fig. 32). Laser ablation and filtered cathodic arc deposition seem to be the methods of choice to deposit good emitting DLC films. Both methods are used in the production of nanotubes as well.

3.3. Carbon Nanotube Field Emitter 3.3.1. Work Function of Carbon Nanotubes The investigations on the field emission properties of CVD diamond and DLC emitters show clearly that the electron emission of carbon based materials is ordinary Fowler– Nordheim tunneling and field enhancement structures on the surface are required for low field electron emission. Nanotubes (CNTs) have the ideal shape to act as field enhancing structures and they can be produced relatively simply in large quantities. We have investigated the field emission properties of MWNTs and SWNTs. On thin films we measured emission site densities of 10,000 cm−2 at applied fields below

Figure 32. SEM pictures of graphitic structures present on good emitting DLC films.

5 V m−1 [104, 114, 115]. Simultaneous field and photoemission electron spectroscopy (XPS) allowed us to compare the work function at the emission site with the average work function of the CNT films. In the case of the MWNTs we determined an emitter work function
= 4 9 ± 0 3 eV by FES. By XPS we measured an average function of 4 8 ± 0 1 eV. In the case of SWNTs we determined a lower work function of 3 7 ± 0 3 eV by FES. The reduced work function for SWNTs can be attributed to an electrostatic effect, due to the small radius of curvature of the SWNTs at their ends. One of the constituents to the work function is the work an electron has to perform against the image force. In the case of nanostructures, it is important to notice that the image potential for a plane and a small sphere (radius R

(45)

where R is the radius of the sphere. It can be easily seen that for r/R = 1 + d (d ≪ 1) sphere coincides with plane

Carbon Nanostructures for Cold Electron Sources

569

with x = R · d. The difference in work function between the planar and the spherical case equals [117]
sphere =
plane −

e2 5 8R−1 40

(46)

This effect was demonstrated by the observation that the ionization potential of small clusters depends on their size. An increase of the ionization potential with decreasing size was observed [118, 119]. According to (46) in SWNTs with diameter around 1.4 nm the work function is reduced by 1.3 eV. Compared to MWNTs with work function values around 5 eV SWNTs with 1.4 nm diameter would have work function values of 3.7 eV which corresponds well with the values we determined by FES.

3.3.2. Environmental Stability The term environmental stability is used to describe the interaction of electron emitters with the gaseous environment in which they are operated. Such interactions typically lead to emission current fluctuations on a time scale of seconds to minutes. If the current fluctuations are large they may negatively affect the device performance. In addition the device long-term stability is dependent on the emitter interactions with its environment, wherefore investigations on environmental stability are important. Such investigations have been done in the past on etched metal tips using field emission microscopy. The field emission microscope (FEM) invented by Müller [120] in 1937 has for a long time been the state of the art to investigate adsorption and diffusion phenomena at surfaces with almost atomic resolution. Figure 33 shows FEM patterns commonly observed from SWNTs and MWNTs. These emission patterns are not stable in time but show switching from one configuration to another. The switching frequency increases with increasing currents and is in the order of 1 Hz. Associated with this instability of the emission spot are fluctuations in the emission current as indicated by the different intensities of the emission patterns in Figure 33b. Each emission pattern has

Figure 33. (a) The lobed FEM pattern of a MWNT at 300 K. (b) Some examples of a large variety of such patterns commonly observed from SWNTs and MWNTs. Reprinted with permission from [121], L. O. Nilsson, Ph.D. Thesis No. 1337, University of Fribourg, Switzerland, 2001.

a well defined emission current for a given applied voltage. Therefore the emission current fluctuates between different current levels associated to the different emission patterns. The difference between the highest and lowest current level can be up to a factor of 10. Such kind of random telegraphic noise was observed by different groups [46, 121, 122]. It is recognized that the FEM patterns on the screen reflect the electronic structure, the anisotropy of the work function, local variation of the microscopic field, and electron transmission probability of a CNT cap. This means that atomically sized areas on the CNT cap, where the work function is lower or the local electric field is higher than the surrounding regions, are reproduced with higher intensity on the phosphor screen. Therefore the different emission patterns of CNTs can be understood in terms of switching between differents states with relatively small energy differences in levels close to the Fermi energy. In order to illustrate how these states are formed and what they may look like one can take the -states of a benzene molecule (C6 H6  as an example (Fig. 34). In analogy to the C6 H6 molecule the -states of CNT cap can be thought as linear combinations of px orbitals which protrude out perpendicular to the carbon lattice on the tube apex. The limited variety in shape, style, and energy of such -states can only partly explain the large number of emission patterns observed by the FEM. It therefore seems plausible to think that the energy levels of the CNT at the apex are modified through adsorbed molecules coming from the gas phase during field emission. Sometimes the sticking is very firm and may result in a resonant tunneling state which can lead to increased tunneling probability and anisotropy in the field emission current [123]. We observed that the field emission fluctuations are not sensitive to H2 and H2 O, the two the main components of residual gases in vacuum systems, up to 10−4 mbar gas pressure. Increased partial pressure of O2 causes degradation of the field emission due to

Figure 34. The px orbitals (a) of benzene C6 H6 can form six  hybrid orbitals of delocalized electrons (b), of which three are in binding stabilized state and three are in an antibinding state. Reprinted with permission from [121], L. O. Nilsson, Ph.D. Thesis No. 1337, University of Fribourg, Switzerland, 2001.

570

Carbon Nanostructures for Cold Electron Sources

reactive ion etching. The suggestion of Rinzler et al. [122] that the emission fluctuations are due to individual desorption processes induced by ion bombardment seems very convincing.

4. MICROSCOPIC CHARACTERIZATION OF THIN FILM EMITTERS The development of field emission electron sources has been hampered for half a century by the fact that tiplike field enhancing structures are needed to create locally high fields for field electron emission to take place. Usually the higher the field enhancement, the smaller the effective emission area of the tips gets, as the radius of curvature gets increasingly small. Although the field emission current density can easily exceed 1000 A cm−2 , the total emission current per single tip remains small as the emitting surface, being the apex of the field enhancing tip, is small. The total current per single tip in field emission rarely surpasses 0.1 mA. In contrast, thermionic emitters can deliver electron emission currents of up to 400 A cm−2 . This situation resulted in the fact that field emitters are only used where high brightness rather than high currents are requested, for example, in the case of high resolution electron microscopy (SEM, TEM). The key to reaching planar field emitters of high current density approaching 1 A cm−2 resides in the integration of a large number of field enhancing tips on a surface.

4.1. Field Enhancement Distribution Function Whereas a single field electron emitter can be said to be completely characterized in terms of its work function
at the emission site, the field enhancement factor , and the resistance R, the description of the overall emission behavior from a thin film emitter with a large number of emission sites requires a statistical approach [121]. The number of degrees of freedom with regard to geometrical alignment, orientation, length, diameter, and interemitter distances within the thin film emitter ensemble is much larger than for a single emitter. It can also be expected that
and R exhibit some variation. Electrostatic screening has to be taken into account for thin film emitters, since the presence of a large number of emitters may affect the local electric field F at the emission sites. The emission properties can therefore vary considerably from one position to another within the thin film emitter. An example of a thin film emitter with a large scatter in the spatial emission current Ix! y is shown in Figure 35a, where parts of the surface show no emission at all. The spatial scatter in the electron emission Ix! y is highly undesirable, because the device performance with regard to brightness homogeneity is largely reduced. In order to understand and improve the electron emission homogeneity of thin film emitters with a large number emitter sites a quantitative description of the Ix! y map in terms of field enhancement factor , work function
, emission area A, and resistance R is needed. Thin film electron emitters are often characterized in terms of a “threshold” field Fthr required to obtain a given

Figure 35. (a) Variation of the emission current Ix! y under a constant applied voltage U0 = 320 V (tip–cathode distance d ∼ 7 m). (b) Corresponding local variation of the field enhancement factor x! y. Reprinted with permission from [121], L. O. Nilsson, Ph.D. Thesis No. 1337, University of Fribourg, Switzerland, 2001.

field emission current density in a typical large anode diode type field emission experiment. Such characterization can be misleading since a cathode with a spatial extension in the mm2 range may contain a small number (sometimes only one) of strong emitters and thus can have a low threshold field Fthr which obviously will not guarantee a high emission current density, because the emitter ensemble shows a certain distribution of the field enhancement factor  and therefore in the emission current density. In this picture the field emission properties of a thin film must be represented by a function f , where the number of emitters dN on a surface area A with field enhancement factors in the interval [!  + d] is given by dN  = A · f  · d

(47)

The field enhancement distribution function f  is then defined as f  =

1 dN A d

[emitter/cm2 ]

(48)

According to the definition, f  is proportional to the probability of the field enhancing structures in the interval [!  + d] per unit area. Consequently one may view f  as a probability distribution. Assuming the same relation for the emission current as a function of the local field F at the emission site for all emitters of the thin film, the field emission properties are completely determined by the field emission distribution function f . In terms of the Fowler–Nordheim relation (16) the assumption of a general relation for the current–field dependence of the electron emission means that all emitters have the same emitting area and the same work function
. The assumption that the work function
is constant for all emitters can partially be justified by the fact that the emitters consist of the same material. Actually, we have found in FES that the work function at the emission site of carbon based materials is always in the range of 4.7–5.1 eV. The argument that the emission area should be constant can for CNTs is partially justified by the fact that all the emitters are of the same dimensions. In order to describe the emission current as a function of the local field F the simplified Fowler–Nordheim relation

Carbon Nanostructures for Cold Electron Sources

as proposed by Brodie and Spindt [3], where the elliptic functions are approximated, can be used  

1 5 10 4 1 5 × 10−6 7
exp −6 44 × 10 · F 2 exp I =A·

0 5 F (49) I is the emission current in amperes per single emitter, A is the emission area of the emitter in cm2 ,
is the work function in eV, and F is the local electric field [V/cm] present at the emission site. In a field emission device, the external electric field F0 is generated by a voltage U0 applied between the cathode with the emitting structures and an anode. In configurations where anode and cathode are represented by two parallel plates separated by a vacuum gap d the applied field is F0 = U0 /d. The local field at the emission site is then defined as F = F0 · . Relation (49) becomes 
1 5 × 10−6 10 4 2 2 I =A· · F0 ·  exp

0 5
 1 5
× exp −6 44 × 107 (50) F0 ·  This is now the relation for the emission current I of a single emitter for a given applied field F0 which depends only on the field enhancement factor . The field emission behavior of an ensemble of field emitters is therefore completely determined by the field enhancement distribution function f . Important parameters such as the threshold field Fthr and the emission site density can be immediately derived from the field enhancement distribution function f . Using relation (50) for the emission current the threshold field Fthr can be determined from the following relation  f  · I! Fthr  d Ithr = A 0


  1 5 × 10−6 10 4 2 · exp d · f  · Fthr · 2

0 5 0 

1 5 (51) × exp −6 44 · 107 Fthr · 

=A·

571 current larger than some threshold current (e.g., 10 nA). Therefore the ESD as function of the applied field F0 can be calculated from f  by ESDF0  =





f  d

(cm−2 )

(52)

min =Fmin /F0

Fmin is the local field at the emission site required to produce a minimum detectable emission current Imin . In an experimental setup using a phosphorous screen Imin corresponds to the minimum emission site current, required to observe the emission spot by eye. The emission current density is calculated in a similar manner jF0  =





0

f  · j! F0  d

(A cm−2 )

(53)

Unfortunately in many publications the performance of planar field emission cathodes is still characterized by I–V measurements alone. Figure 36 displays the field emission spots on a phosphorous screen of two carbon thin film emitters at the same applied field and emitting roughly the same emission current density of about 0.1 mA cm−2 , so that from the I–V characteristic both thin film emitters would yield a comparable behavior and one would judge them to have the same field emission performance. It is, however, obvious that the ESD is considerably higher for sample B. Whereas on sample B the emission current originates from a relative high density of emission sites the emission current of sample A is delivered from only a few but quite strong emitters. Therefore the emission performance of sample B should be judged considerably higher than for sample A because emitter degradation will more easily occur on sample A at higher fields. In conclusion, any standards of characterization of thin film electron emitters must include the field enhancement distribution function f  and knowledge of the spatial emitting properties x! y. With these two parameters the collective emitting behavior of thin film electron emitters is completely characterized.

Here A denotes the surface of the cathode under investigation. It is worth noting that the threshold field Fthr will depend on the cathode area which is measured, so that it is only sensible to define the threshold field Fthr with regard to a given emission current density jthr . In other words, the measured threshold field Fthr decreases with increasing surface area measured. The characterization of the emission properties of a thin film emitter just by stating the threshold field Fthr is therefore very questionable.

4.2. Emission Site Density The emission site density (ESD), which gives the number of active emitters per unit area at a given applied field or voltage, is one of the key parameters in characterizing the field emission performance or quality of a planar field emission cathode. An active emitter is an emitter giving an emission

Figure 36. Optical micrographs of field emission spots collected on a phosphorous screen of two different diamond thin film emitters at the same applied electric field of 4.5 V m−1

Carbon Nanostructures for Cold Electron Sources

572 4.3. Characterization by Scanning Anode Field Emission Microscopy The scanning anode field emission microscope (SAFEM) allows the investigation of the field emission properties of thin film emitters with lateral resolution in the submicrometer range [124]. The SAFEM is based on a micrometer-sized tip which serves as an anode. By recording the emission as a function of the spatial position of the tip above the thin film emitter, field emission maps with submicrometer resolution can be obtained. Such maps can be extended up to 5000 × 5000 m2 . The SAFEM can operate in two different modes. In the first mode the emission current is recorded for a constant applied voltage U0 . This mode is called the constant voltage mode (CVM). The Ix! y map obtained in the CVM is the analog to the electron emission on a phosphor screen, albeit with higher resolution. In the second mode of operation the emission current is maintained on a constant level for every tip position along the field emission scan and the voltage applied to the tip must therefore be adjusted according to the individual emitter efficiency. Since the spatially recorded emission voltage V x! y is determined for a constant emission current this mode is called the constant current mode (CCM). As shown in the previous section, the knowledge of the field enhancement distribution function f  gives an almost complete characterization of field emission behavior of thin film emitters. Therefore the experimental characterization of thin film emitters should include the determination of f . The field enhancement distribution function f  can be obtained from the spatial distribution of field enhancement factors , the so-called field enhancement map x! y. In theory x! y can be calculated from the topography of the surface hx! y, for example, by solving the Poission equation at the surface. However, in general the Poission equation cannot be solved because it is extremely difficult to measure hx! y with nanometer resolution. Therefore another method is used to derive x! y. Through relation (50), the emission current I is directly related to field enhancement factor ; therefore the local emission current Ix! y on thin film emitter is given by

go into the resistor limited emission regime cannot be used as the basis to calculate x! y. In order to derive x! y and avoid current saturation effects, not to say emitter destruction, the scanned electron emission measurement has to performed in the CCM. In the CCM the voltage applied to the anode varies as a function of the anode position over the thin film emitter, in order to maintain the same emission current for every point. If the current is reasonably low (∼50 nA), one will obtain a spatially resolved field emission image of the extraction voltage V x! y, without current saturation effects or emitter destruction. This is both experimentally and conceptually beautiful since V x! y is actually an inverted image of the field enhancement landscape x! y according to x! y =

d·F V x! y

(55)

where d is the tip–sample distance and F is the local field at the emission site required for a fixed emission current I predicted by the Fowler–Nordheim relation (54) (see Fig. 37).

4.4. Carbon Nanotube Thin Film Emitters As discussed, the key for reaching planar field emitters of high current density approaching 1 A cm−2 resides in the integration of a large number of field enhancing tips on a surface. Depending on the required emission current density, the density of the emitting tips has to be in the range of 106 –108 cm−2 . This limits their size to micrometer or even submicrometer dimensions. CNTs have proven to be ideally suited to serve as field enhancing structures for field emission applications due to their exceptionally high aspect ration with lengths in the micrometer range and diameters ultimately down to one nanometer. Furthermore, they can be readily produced in large numbers by relatively simple and cost effective techniques. One has to acknowledge that when it comes down to creating free electrons in vacuum the method most widely employed to produce these electrons is thermionic emission. The advantage of thermionic electron


1 5 × 10−6 10 4 · F02 · x! y2 exp

0 5

1 5 × exp −6 44 × 107 (54) F0 · x! y

Ix! y = A ·

Hence, the measurement of Ix! y could theoretically be used to calculate x! y. Scanning anode field emission measurements operating in the CVM can be principally used to obtain spatially resolved emission current Ix! y. Since the anode is located very close to the emitting surface, the surface topography and the local field enhancement cause the applied electric field to fluctuate several times from one site to another. Such variation in the applied electric field causes emission current fluctuations of many orders of magnitude into the regime of resistor limited emission and emitter degradation [125]. Since expression (54) is describes pure Fowler–Nordheim emission, field emission measurements (CVM) which may

Figure 37. SAFEM measurement from a CNT thin film in the CCM. Extraction voltage map V x! y for 10 nA emission recorded with an tip–cathode distance of 7 m.

Carbon Nanostructures for Cold Electron Sources

sources is their simple operation. However, thermionic electron sources have some major drawbacks. They are power controlled, via the temperature, and therefore direct modulation of the emission is rather slow. Further, due to operation at high temperatures (∼2000 K) the realization of micrometer sized emitters is practically impossible due to insufficient heat dissipation. Micro- and nanosized field electron emitters hold the possibility of miniaturizing vacuum electronic devices. In this context CNT thin film emitters offer an attractive possibility to produce the key component of an electron source, namely the field emitting tip, in large numbers using inexpensive processes. In CNT thin films as they are produced by CVD over supported catalysts the growth yields to nanotubes with rather ill defined diameter, length, orientation, and spatial distribution. All these properties, however, influence the field enhancement and therefore the field emission properties. The field enhancement of a metallic needle shaped structure perpendicular in a homogeneous electric field is in first approximation h/r. From this it is clear that the length, diameter, and orientation of the CNT with respect to the electric field will immediately influence the field enhancement. In addition, the tube density will influence the field enhancement due to electrostatic shielding [126]. As discussed in the previous section, the ensemble of field emitters represented by a CNT thin film is therefore difficult to be characterized a single parameter such as threshold electric field Fthr to obtain a given field emission current density. Because the CNT emitter ensemble shows a certain distribution in the field enhancement the emission properties have to be described statistically by the field enhancement function f  according definition (48). It was shown that under the assumption that all emitters have the same work function
, the local emission current Ix! y on thin film emitter can be written as 
10 4 1 5 × 10−6 · F02 · x! y2 exp

0 5

1 5 × exp −6 44 × 107 F0 · x! y

Ix! y = A ·

In order to account for experimental observed emission currents and field emission spectra of MWNTs, A = 10−9 cm2 is a reasonable value for the emission area of a MWNT. This value might seem too large for a MWNT with a typical diameter of 30 nm. Yet one has to consider that though the pre-exponential factor Av [see (27)] has the dimension of a surface, it not only takes account of the emitting area but also depends on the band structure of the emitter. Assuming a mean value of 4.9 eV for the work function and using the emission area A mentioned previously the emission current I for a MWNT is given by

573 relation (51) Ithr = A



=A





0  0

f  · I! Fthr  d d · f  · 3 36


 6 99 × 108 2 × 10−14 · Fthr · 2 exp − Fthr · 

(57)

A calculated x! y landscape compiled from a extraction voltage map V x! y is presented in Figure 38a. By counting the field enhancement values AN in the interval [!  + d] one can get a histogram with the distribution of the field enhancement factors , as shown in Figure 38b. It is worth noting that the amplitude of the histogram is, however, dependent on the choice of the bin with d. The normalized distribution dN /d equals the scanned area A times a functional dependence f . At first glance the shape of the normalized distribution dN /d ∝ f  appears to be asymmetric with a maximum located at the  value of 50. This apparent asymmetric behavior is, however, due to a measurement artifact.  values below 50 are “screened” due to neighboring field emitters with large  values. The probability to detect field emitters is thus a monotonic increasing function of  according to 
 A E ∝ 1 − A0 f  d → 1 when  →  (58) 

The integral can be understood as the area AE occupied by emitters with field enhancement factors larger than . The effective  distribution
 = AE f , as calculated from the true  distribution f , can be fitted to the experimental values as shown in Figure 38b. It can be shown that f  can be fitted to experimental data over five orders of magnitudes using scanning and integral (screen) field emission measuring techniques according to f  = k1 exp−k2 

(59)

where k1 is the measured macroscopic emission site density and k2 is connected with the structural order/disorder of the field enhancing structures. By changing k1 , k2 the field emission properties of any surface can thus be tuned. Engineering technological relevant field emission cathodes is


 6 99 × 108 Ix! y = 3 36 × 10−14 · F02 · x! y2 exp − F0 · x! y (56) and the threshold field Fthr for a MWNT thin film emitter can be determined from the following relation according to

Figure 38. (a) Field enhancement x! y landscape calculated from an extraction voltage map V x! y which is an analog of the one presented in Figure 37. (b) Histogram of the field enhancement factors  in (a).

574 thus the story of how to change/manipulate the fundamental relation (59). The question as to what a technologically relevant and useful ESD for a CNT thin film emitter will be has to be answered in view of the application the emitter should be used in and of course in view of the competing technologies. Comparison with silicon or metal microtip arrays indicates that an ESD of the order of 106 up to 108 cm−2 at applied electric fields below 50 V m−1 should be aimed at. The characterization of EDSs above 104 cm−2 using a phosphorous screen setup is challenging due to the overlap of the emission spots. As in the phosphorous screen the anode–cathode distance gives the resolution of the SAFEM. Figure 39 displays three field enhancement maps of the same region of a randomly oriented MWNT thin film at different tip–sample distances evidencing the increasing resolution with decreasing tip–sample distance. As can be seen in Figure 39c a resolution is given in first approximation as twice the tip–sample distance. This means that in order to resolve 107 emitters per cm2 the tip–sample distance should be of the order of 2–3 m. Such a tip–sample distance is of the same order as the typical height of the CNTs and therefore is about the limit of what can be achieved. Figure 40 displays a high resolution field enhancement map derived from a SAFEM voltage map V x! y recorded at a tip–sample distance of 2–3 m on a randomly oriented CNT thin film. The black diamond dots indicate the position of emission sites with field enhancement factors larger than 130. On the investigated area of 2 4 × 10−5 cm−2 164 such emitters could be detected resulting in an ESD of 6 8 × 106 cm−2 . The field enhancement was determined under the assumption that the local field present at the CNT apex is 3800 V m−1 in order to emit 11 nA, which is the current at which the SAFEM V x! y map was recorded. A field enhancement factor larger than 130 therefore means that the applied field is lower than 30 V m−1 . This measure-

Figure 39. SAFEM field enhancement maps recorded at 11 nA emission current of the same region of a randomly oriented CNT thin film emitter with different tip–sample distances (d = 33! 13! 7 m).

Carbon Nanostructures for Cold Electron Sources

Figure 40. High resolution SAFEM field enhancement map x! y of a randomly oriented CNT thin film recorded at 11 nA emission current and a tip–sample distance of ∼3 m.

ment demonstrates that relatively crude, as-deposited CNT thin films exhibit already the potential to achieve technologically relevant ESDs at moderate applied fields. However, it is important to point out that the high ESD indicated of a few million per cm−2 is to a certain extent academic as it cannot be exploited on large cathode surfaces. The reason for this is the large spread in the field enhancement factor  the individual emitters exhibit. For the randomly oriented CNT thin films we discuss here the field enhancement distribution f  follows an exponential law as described in (57). This kind of field enhancement results in the situation that on a large cathode area (e.g., 1 cm−2  there are a few very strong emitters of very high field enhancement (∼1500) resulting in a threshold field of 1–2 V m−1 . However the ESD is very low, only 1–10 cm−2 . Upon an increase in the applied field the ESD increases rapidly, but at the same time the emission current from the first few emitters increases exponentially. With further increase of the applied field the point is reached where the emission current of the first strong emitters is so large that they get destroyed. From this point the ESD hardly increases anymore as a situation is reached where the appearance of a new emitter is balanced by the disappearance of old strong emitters upon increase of the applied field. For as-deposited MWNT thin film emitters this point is reached typically at about 10 V m−1 with an ESD of ∼10,000 cm−2 and an emission current density of ∼10 mA cm−2 over the whole thin film emitter. It is clear that under such circumstances the cathode suffers irreversible degradation and that the operation of a cathode in such conditions is highly critical from the point of view of stability and lifetime. The values for the field, the ESD, and the emission current density given are not of a general nature as they depend on numerous parameters. However, it is important to notice that the emitter degradation at low applied fields is the limiting factor for the performance of such cathodes. In order to enhance the field emission performance of such cathodes a detailed understanding of the degradation of the degradation mechanism and subsequently the improvement of the degradation behavior is required. Figure 41 shows the SAFEM field enhancement map x! y of a randomly oriented MWNT thin film emitter, with regions which have suffered high emission current degradation previously. The region from x = 0 m to x = 55 m has been strained with 1.5 A and the region from

Carbon Nanostructures for Cold Electron Sources

575

Figure 41. SAFEM field enhancement map x! y of a randomly oriented MWNT thin film with two high emission current strained regions on the left hand side. Region I (0 < x < 55 m) has been strained previously with 1.5 A and region II (55 < x < 110 m) with 2 A.

x = 55 m to x = 110 m has been strained with 2 A emission current by the SAFEM. The region x > 110 m has not been current strained. After current straining a field enhancement map x! y was recorded at 11 nA and a tip–sample distance of 4–5 m. One can observe that in the current strained regions the emitters with a high field enhancement factor  have been removed, whereas they are still present in the unstrained region. The values at which the current degradation occurs can vary from 500 nA up to a few A, which is more than 100 times lower than for a single CNT welded on a metal tip. I–V measurements on two point contacted CNTs have shown that straight and probably defect-free CNTs, repeatably can conduct currents up to 3 mA ballistically without heat dissipation [127, 128]. The emitters which are then left over with a lower field enhancement factor  are typically very stable and robust, which means that they can sustain emission currents well above 1 A even up to 100 A for short periods of time. A closer inspection of the regions strained by 1.5 and 2 A shows that the average field enhancement is higher in the region strained with the lower current. This behavior indicates that the high current degradation is linked to the enhancement, where the critical current for degradation decreases with increasing field enhancement and therefore increasing aspect ratio. Figure 42 illustrates the degradation of a single emitter as opposed to the collective current degradation behavior displayed in Figure 41. Figure 42a and b shows field enhancement maps x! y (@ 11 nA emission current) before and after the central emitter has been subjected to high current degradation. The I–V characteristic displayed in Figure 42c shows that the irreversible degradation occurs abruptly at an applied voltage of 290 V, which corresponds to an applied electric field of 14.5 V m−1 , and at a current level of 2.2 A. The degradation manifests itself by a current drop of almost three orders of magnitude. As can be evaluated from the field enhancement maps the value of the field enhancement at this position has decreased from 630 to 250. It has to be pointed out that most probably the

Figure 42. Single emitter high current degradation measured on a MWNT cathode. The field enhancement maps (a) and (b) show the reduction in the field enhancement factor for the emitter indicated by the arrow (a) before and (b) after the degradation event. In the maps white corresponds to high and black to low field enhancement. Diagram (c) displays the I–V characteristic of the degradation where the crosses indicate the experimental data, the black solid line the theoretical Fowler–Nordheim behavior, and the gray solid line the resistor limited I–V behavior. The schematic drawings (d) and (e) represent emitter degradation due to large power dissipation due to either high intratube resistance or high contact resistance.

emission after degradation does not originate from the same CNT as before but from a tube nearby, which was concealed by the original emitter. The black curve in Figure 42c indicates the theoretical Fowler–Nordheim emission behavior. One can observe that in the low current regime the measured I–V characteristic is well described by the Fowler–Nordheim law. However, above 100 nA emission current a pronounced deviation from the Fowler–Nordheim behavior can be observed. This current saturation continues to a voltage of about 290 V and an emission current of 2.2 A, where a sudden and irreversible decrease of the emission current to a value of 7 nA occurs. The emission behavior in the non-Fowler–Nordheim regime can be modeled using a resistor limitation approach, where it is assumed that there is a voltage drop between the emission site (being the CNT apex) and the electron reservoir (being the electrical contact of the CNT on the substrate). This results in the situation that the full applied voltage does not generate the extraction field F for the electron emission, but this voltage is reduced by the product of the emission current and the resistance in the current path. Assuming an ohmic resistor the high voltage behavior in Figure 42c can

576 be modeled reasonably well with the following relation:
 C2 I = C1 · V − AV 2 exp − V − AV
 C2 2 (60) = C1 · V − R · I exp − V −R·I C1 and C2 are positive constants in Eq. (60) which is equivalent to the Fowler–Nordheim relation (16) taking into account the proportionality between the voltage and the field. Relation (60) is of course recursive and has to be evaluated numerically. From the emission current and the voltage drop one can estimate the amount of power dissipated in the resistor. In principle the voltage drop could be determined from the model; it has to be stressed, however, that in order to do so the influence of the voltage drop on the local field at the apex of the tube should be known. This in turn requires information on the exact location of the voltage drop in the current path and the amount of electrostatic shielding the emitter experiences. Both are generally unknown. The draws in Figure 42d and e illustrate the possible origins of the voltage drop due to either a high internal resistance in the CNT (e.g., due to defects) or a high contact resistance. It can be estimated that the voltage drop in the saturation regime has to be of the order of 10 V at 2 A emission current which gives a power of 20 W dissipated. Taking into account the very small volume of the contact Vc  dtube 3 ≈ 5 × 10−17 cm−3 one estimates a power density in the range of 1012 W/cm3 dissipated in the contact. This can easily explain the emission degradation and the observed local substrate melting due to high emission currents [125]. From this consideration it becomes obvious that the operating the emitter in the current saturation regime is always risky as the current saturation is a sign of power dissipation and therefore of possible degradation. From the I–V curve in Figure 42c it becomes apparent that the voltage window where the emitter can be operated usefully with an emission current between 10 nA and 2 A is rather small. This finding can be related generally to CNT emitters by realizing that the window of operation for an emitter corresponds to about twice the threshold field. Together with the large spread in the field enhancement distribution function f , this leads to the limitation in the emission performance of randomly oriented CNT thin film emitters. The threshold field of macroscopic CNT thin film emitters of 1 cm2 area is generally around 2 V m−1 . SAFEM investigations, however, show that an interesting ESD of the order of 106 cm−2 is reached only for applied fields higher than 25 V m−1 . According to the previous consideration emitter degradation has to be expected for applied fields around 5 V m−1 . It is therefore not possible to reach that required for the high ESD without very severe emitter degradation.

Carbon Nanostructures for Cold Electron Sources

electrostatic shielding between the emitting structures. To quantify the negative influence of electrostatic shielding on the emission properties we have carried out computer simulations, calculating the field emission current density for CNT arrays of tube different densities. Figure 43 illustrates the effect of the electrostatic shielding in CNT films with different tube densities. A quantitative analysis of the computer simulation shows that optimum field emission properties for a CNT film can be expected, if the mean distance between the tubes is about twice their height. The simulations show further that the field enhancement factor  of the tubes in a film shows a wide distribution, as it is influenced by the orientation of the tube with respect to the applied electric field and by the degree of electrostatic shielding a tube receives from its neighboring tubes. In real CNT films the differences of length and diameter of the individual tubes will further contribute to the distribution of the field enhancement factor  among the CNTs. This distribution gives rise to inhomogeneous field emission properties of CNT thin film emitters. The simulations shown demonstrate clearly why a homogenous emitting CNT film must be formed by identically ordered CNTs. The ideal case of such a CNT film is displayed in Figure 44, where all tubes have the same length, diameter, orientation, and distance from each other. As shown in Section 1.5.3, it is possible to grow films with well-oriented CNTs. The problem of all these films for field emission is that the spacing between the individual tubes is not large enough to prevent electrostatic shielding. It is a big challenge for the nanotechnology to develop processes required for the deposition of CNT films, optimal for field emission. In the case of a flat panel display a CNT based field emission technology will ultimately need to fulfill the following requirements to be competitive with the current metal tip technology: • growth of individual CNTs at well-defined positions, tolerance ∼100 nm • growth of individual CNTs with well-defined orientation • growth of individual CNTs with a well-defined aspect ration ( = h/r), tolerance d/ ∼ 5%

4.5. Optimal Structure for Carbon Nanotube Based Electron Field Emitters As shown in the previous sections the field emission properties of a thin film emitter can be fully characterized by the field enhancement distribution function f , which depends on the geometry of the emitting structure and the

Figure 43. Two-dimensional simulation of the electrostatic potential in films with random distributed and oriented CNTs and different tube densities. (All tubes have the same length and diameter.)

Carbon Nanostructures for Cold Electron Sources

577

GLOSSARY

Figure 44. Two-dimensional simulation of the electrostatic potential in an ordered CNT array.

• massively parallel processing (108 tubes per cm2  • large area deposition (∼m2  • low cost processes

5. CONCLUSIONS Significant progress in the understanding of the electron emission has been achieved since the first reports on the low threshold field electron emission from carbon thin film emitters in the early 1990s. It seems certain that the low threshold field electron emission from carbon thin film electron emitters can be accurately described by Fowler–Nordheim tunneling. Fowler–Nordheim tunneling requires high electric fields in the order of 3000 V/m generated by field enhancing structures. Contrary to a single emitter, the emission properties of a thin film emitter require a statistical treatment, since the field enhancing structures exhibit a distribution with regard to the filed enhancement . Therefore the electron emission properties of thin film emitter with an ensemble of field enhancement structures have be characterized in terms of the spatial field enhancement, the field enhancement map x! y. The variation in  for different field enhancement structures leads to a  distribution, the field enhancement distribution function f , which can be seen as a probability function of the field enhancing structures. The field enhancement distribution function f  can be said to give an almost complete characterization of the field emission properties, contrary to the threshold field Fth , of thin film emitters. The field enhancement distribution function f  has been shown to be exponential for randomly oriented CNT thin films. In order to obtain a field emission device with uniform brightness and high emission site density from thin films, stringent requirements must be placed on the permitted spatial variation of the emitting sites and therefore the variation of the field enhancement . Homogeneous electron emission from an ensemble of field emitting structures, defined as dI/I < 50, requires a relative variation of the field enhancement d/ smaller than 4%. Therefore it is desirable to optimize the density of sites close to the highest field enhancement of the ensemble to the limit of electrostatic screening effect. The example of CNT based field electron emission technology demonstrates clearly the difficult path from nanoscience to nanotechnology. The development of devices based on nanostructures needs, in addition to the process technology, characterization methods and knowledge which cover all aspects, from the fundamental physical phenomena of the individual nanostructure up to the integral behavior of the large number of nanostructures forming the macroscopic device.

Carbon nanotube (CNT) A carbon allotrope consisting of rolled-up graphene sheets. Emission site density (ESD) The number of emitters per unit area, which delivers for a given applied electric field an emission current larger than a given threshold current. Field emission spectroscopy (FES) Measurement of the energy distribution of the field emitted electrons. Scanning anode field emission microscope (SAFEM) A scanning probe microscope working with a micrometer-sized tip scanned at constant distance over the cathode surface and with applied fields high enough for field electron emission.

ACKNOWLEDGMENTS The authors acknowledge the immense contributions of Dr. O. M. Küttel, Dr. E. Maillard-Schaller, and Dr. R. Clergereaux to this work. The results of some of the work reported in this chapter were partly supported by the Swiss TOP Nano 21 Initiative on Nanotechnology and Motorola Inc. Res., Tempe, AZ (USA).

REFERENCES 1. P. R. Schwoebl and I. Brodie, J. Vac. Sci. Technol. B 13, 1391 (1995). 2. Hewlett Packard, Sci. Ame. May 2000, p. 50. 3. I. Brodie and C. Spindt, Adv. Electron. Electron Phys. 83, 6 (1992). 4. A. Einstein, Ann. Phys. 17, 132 (1905). 5. O. W. Richardson, Philos. Trans. Roy. Soc. London 201, 497 (1903). 6. L. W. Nordheim, Proc. Roy. Soc. London Ser. A 121, 626 (1928). 7. R. W. Wood, Phys. Rev. Ser. I 5, 1 (1897). 8. R. H. Fowler and L. W. Nordheim, Proc. Roy. Soc. London Ser. A 119, 173 (1928). 9. The WKB method was published in 1926 independently by three authors: G. Wenzel, Z. Phys. 38, 518 (1926); H. A. Kramers, Z. Phys. 39, 828 (1926); L. Brillouin, C. R. Acad. Sci. Paris 183, 24 (1926). 10. H. F. Ivey, Phys. Rev. 76, 567 (1949). 11. F. Rohrbach, CERN Report 71-5/TC-L, 1971. 12. I. Langmuir, J. Am. Chem. Soc. 54, 2798 (1932). 13. G. Gärtner, P. Geittner, H. Lydtin, and A. Ritz, Appl. Surf. Sci. 111, 11 (1997). 14. P. Ruffieux, O. Gröning, M. Bielmann, P. Mauron, L. Schlapbach, and P. Gröning, Phys. Rev. B, in press. 15. J. van der Weide, Z. Zhang, P. K. Baumann, M. G. Wensell, J. Bernholc, and R. J. Nemanich, Phys. Rev. B 50, 5803 (1994). 16. L. Diederich, O. M. Küttel, P. Aebi, and L. Schlapbach, Surf. Sci. 418, 219 (1998). 17. S. P. Bozeman, P. K. Baumann, B. L. Ward, M. J. Cuomo, R. J. Nemanich, and D. L. Dreifuss, Diamond Relat. Mater. 5, 802 (1996). 18. J. Robertson, Mater. Res. Soc. Symp. Proc. 509, 83 (1998). 19. J. E. Jaskie, U.S. Patent 5, 619, 092. 20. J. Robertson, Mater. Res. Soc. Symp. Proc. 498, 197 (1998). 21. J. Robertson, J. Vac. Sci. Technol. B 17, 659 (1999). 22. L. S. Pan and D. R. Kania, “Diamond: Electronic Properties and Applications.” Kluwer Academic, Boston, 1995. 23. G. Davis, “Properties and Growth of Diamond.” Inspec, London, 1994. 24. S. R. P. Silva, J. Robertson, W. I. Mine, and G. A. J. Amaratunga, “Amorphous Carbon: State of the Art.” World Scientific, Singapore, 1997.

578 25. P. R. Wallace, Phys. Rev. 71, 622 (1947). 26. A. Santoni, L. J. Terminello, F. J. Himpsel, and T. Takahashi, Appl. Phys. A 52, 299 (1991). 27. E. Herbig, Astrophys. J. 196, 129 (1975). 28. T. G. Dietz, M. A. Duncan, D. E. Powers, and R. E. Smalley, J. Chem. Phys. 74, 6511 (1981). 29. S. Iijima, Nature 354, 56 (1991). 30. M. S. Dresselhaus, G. Dresselhaus, and P. C. Eklund, “Science of Fullerenes and Carbon Nanotubes.” Academic Press, San Diego, 1996. 31. B. V. Deryagin and D. V. Fedoseev, Science 233, 102 (1975). 32. B. Dischler and C. Wild, “Low-Pressure Synthetic Diamond.” Springer-Verlag, Berlin, 1998. 33. P. K. Bachmann, in “Properties and Growth of Diamond” (G. Davis, Ed.), EMIS Datareviews Series, p. 349. Short Run Press Ltd., Exeter, 1994. 34. P. K. Bachmann, D. Leers, and H. Lydt, Diamond Relat. Mater. 1, 1 (1991). 35. L. S. Plano, Growth of CVD diamond for electronic applications, in “Electronic Properties and Applications” (L. S. Pan and D. R. Kania, Eds.). Kluwer, Boston, 1995. 36. J. E. Butler and R. L. Woodlin, Philos. Trans. Roy. Soc. London Ser. A 342, 209 (1993). 37. S. Skotov, B. Weimer, and M. Franklach, J. Phys. Chem. 99, 5616 (1995). 38. C. J. Chu, R. H. Hauge, J. L. Margrave, and M. P. D’Evelyn, Appl. Phys. Lett. 61, 1393 (1992). 39. S. J. Harris, Appl. Phys. Lett. 56, 2298 (1990). 40. M. Frenklach and K. E. Spear, J. Mater. Res. 3, 133 (1988). 41. H. Lui and D. S. Dandy, Diamond Relat. Mater. 4, 535 (1995). 42. M. Ece, B. Oral, J. Patscheider, and K.-H. Ernst, Diamond Relat. Mater. 4, 720 (1995). 43. B. R. Stoner, G. H. M. Ma, S. D. Wolter, and J. T. Glass, Phys. Rev. B 45, 11067 (1992). 44. C. Wild, P. Koidl, W. Müller-Sebert, H. Walcher, R. Kohl, N. Herres, R. Locher, and R. Brenn, Diamond Relat. Mater. 2, 158 (1993). 45. E. Maillard-Schaller, O. M. Küttel, P. Gröning, O. Gröning, R. G. Agostino, P. Aebi, L. Schlapbach, P. Wurzinger, and P. Pongartz, Phys. Rev. B 55, 15895 (1997). 46. O. Gröning, Field Emission Properties of Carbon Thin Films and Carbon Nanotubes, Ph.D. Thesis No. 1258, University of Fribourg, Switzerland, 1999. 47. W. Zhu, G. P. Kochanski, S. Jin, and L. Seibles, J. Appl. Phys. 78, 2707 (1995). 48. S. Aisenberg and R. Chabot, J. Appl. Phys. 42, 2953 (1971). 49. J. Robertson, Progr. Solid State Chem. 21, 199 (1991). 50. D. R. McKenzie, Rep. Progr. Phys. 59, 1611 (1996). 51. J. Robertson, in “Amorphous Carbon: State of the Art.” World Scientific, Singapore, 1998. 52. P. K. Bachmann, Ullman’s Encyclopaedia Ind. Chem. A 26, 720 (1996). 53. H. Tsai and D. B. Bogy, J. Vac. Sci. Technol. A 5, 3287 (1987). 54. R. Hauert, J. Patscheider, L. Knoblauch, and M. Diserens, Adv. Mater. 11, 175 (1999). 55. S. R. P. Silva, J. D. Carey, R. U. A. Kahn, E. G. Gerstner, and J. V. Anguita, in “Handbook of Thin Film Materials” (H. S. Nalwa, Ed.), Vol. 4, p. 403. Academic Press, San Diego, 2002. 56. J. Robertson and E. P. O’Reilly, Phys. Rev. B 35, 2946 (1987). 57. N. F. Mott and E. A. Davis, in “Electronic Processes in NonCrystalline Materials,” Ch. 1. Oxford Univ. Press, Oxford, 1971. 58. R. A. Street, “Hydrogenated Amorphous Silicon.” Cambridge Univ. Press, Cambridge, UK, 1991. 59. F. Darkrim and D. Levesque, J. Chem. Phys. 109, 4981 (1998). 60. T. W. Ebbesen, Annu. Rev. Mater. Sci. 24, 235 (1194). 61. T. Guo, P. Nikolaev, A. Thess, D. T. Colbert, and R. E. Smalley, J. Phys. Chem. 55, 10694 (1995).

Carbon Nanostructures for Cold Electron Sources 62. M. Endo, K. Takeuchi, S. Igarashi, K. Kobori, M. Shiraishi, and H. W. Kroto, J. Phys. Chem. Solids 54, 1841 (1993). 63. S. Iijima, T. W. Ebbesen, and P. M. Ajayan, Nature 358, 229 (1992). 64. “Graphite Fibres and Filaments” (M. S. Dresselhaus, G. Dresselhaus, I. L. Spain, and H. A. Goldberg, Eds.). Springer-Verlag, New York, 1998. 65. R. T. K. Baker and P. S. Harris, Chem. Phys. Carbon 14, 83 (1978). 66. S. H. Tsai, C. W. Chao, C. L. Lee, and H. C. Shih, Appl. Phys. Lett. 74, 3462 (1999). 67. J. Li, C. Papadopoulus, J. M. Xu, and M. Moskivits, Appl. Phys. Lett. 75, 367 (1999). 68. C. Bower, Appl. Phys. Lett. 77, 830 (2000). 69. J. Kong, H. T. Soh, A. M. Cassell, C. F. Quate, and H. Dai, Nature 385, 878 (1998). 70. S. Fan, M. G. Chapline, N. R. Franklin, T. W. Tombler, A. M. Cassell, and H. Dai, Science 283, 512 (1999). 71. H. Kind, J.-M. Bonard, Ch. Emmenegger, L.-O. Nilsson, K. Hernadi, E. Maillard Schaller, L. Schlapbach, L. Forro, and K. Kern, Adv. Mater. 11, 1285 (1999). 72. B. Q. Wei, R. Vajtai, Y. Jung, J. Ward, R. Zhang, G. Ramanath, and P. M. Ajayan, Nature 416, 495 (2002). 73. Z. J. Zhang, B. Q. Weis, G. Ramanath, and P. M. Ajayan, Appl. Phys. Lett. 77, 3764 (2000). 74. Y. Tu, Z. P. Huang, D. Z. Wang, J. G. Wen, and Z. F. Ren, Appl. Phys. Lett. 80, 4018 (2002). 75. L. Delzeit, I. MacAninch, B. Cruden, D. Hash, B. Chen, J. Han, and M. Meyyappan, J. Appl. Phys. 91, 6027 (2002). 76. M. Tanemura, K. Iwata, K. Takahashi, Y. Fujimoto, and F. Okuyama, J. Appl. Phys. 90, 1529 (2001). 77. Y. S. Woo, D. Y. Jeon, I. T. Han, N. S. Lee, J. E. Jung, and J. M. Kim, Diamond Relat. Mater. 11, 59 (2002). 78. Ch. Bower, W. Zhu, S. Jin, and O. Zhou, Appl. Phys. Lett. 77, 830 (2000). 79. K. B. K. Teo, M. Chowalla, G. A. J. Amaratunga, W. I. Milne, D. G. Hasko, G. Pirio, P. Legagneux, F. Wyczisk, and D. Pribat, Appl. Phys. Lett. 79, 79 (2001). 80. B. C. Djubua and N. N. Chubun, IEEE Trans. Electron Devices 38, 2314 (1991). 81. C. Wang, A. Garcia, D. C. Ingram, M. Lake, and M. E. Kordesch, Electron. Lett. 27, 1459 (1991). 82. M. W. Geis, N. M. Elfremov, J. D. Woodhouse, M. D. McAleese, M. Marchywka, D. G. Socker, and J. F. Hochedez, IEEE Electron Device Lett. 12, 456 (1991). 83. L. N. Dobretsov and M. V. Gomoyunova, “Emission Electronics.” (Izdatel’stvo Nauka, Moscow, 1966); Israel Programme for Scientific Translations, Jerusalem, 1971. 84. See www.pixtech.com 85. K. Okano, K. Hoshina, and M. Iida, Appl. Phys. Lett. 64, 2742 (1994). 86. Yu. V. Gulyaev, L. A. Chernozatonskii, Z. Ya. Kosakovskaya, N. I. Sinitzyn, G. V. Torgashov, and Yu. F. Zakharchenko, in “Proc. 7th Int. Vac. Microelectronics Conf.,” Grenoble, France, 1994, p. 322. 87. A. Y. Tcherepanov, A. G. Chakhovskoi, and V. V. Sharov, in “Proc. 7th Int. Vac. Microelectronics Conf.,” Grenoble, France, 1994, p. 76. 88. D. Hong and M. Aslam, J. Vac. Sci. Technol. B 12, 427 (1995). 89. V. V. Zirnov, E. I. Givargizov, and P. S. Plekhanov, J. Vac. Sci. Technol. B 13, 418 (1995). 90. G. A. J. Amaratunga and S. R. P. Silva, Appl. Phys. Lett. 68, 2529 (1996). 91. K. Okano, S. Koizumi, S. Ravi, P. Silva, and G. A. J. Amaratunga, Nature 381, 140 (1996). 92. Z.-H. Huang, P. H. Cutler, N. M. Miskovsky, and T. E. Sullivan, in “Proc. 7th Int. Vac. Microelectronics Conf.,” Grenoble, France, 1994, p. 92. 93. N. S. Xu, Y. Tzeng, and R. V. Latham, J. Phys. D 27, 1988 (1994). 94. R. G. Forbes, Solid-State Electron. 45, 779 (2001).

Carbon Nanostructures for Cold Electron Sources 95. M. W. Geis, J. C. Twichell, and T. M. Lyszczarz, J. Vac. Sci. Technol. B 14, 2060 (1996). 96. F. J. Himpsel, J. A. Knapp, J. A. Van Vechte, and D. E. Eastman, Phys. Rev. B 20, 624 (1979). 97. J. van der Weide, Z. Zang, P. K. Baumann, M. G. Wensell, J. Bernholc, and R. J. Nemanich, Phys. Rev. B 50, 5803 (1994). 98. C. Bradis and B. B. Pate, Appl. Phys. Lett. 69, 4123 (1996). 99. R. G. Farrer, Solid State Comm. 7, 685 (1969). 100. E. Aluker, St. Chernov, and V. Garilov, Phys. Status Solidi K 25, 172 (1992). 101. G. Gärtner, P. Geittner, H. Lydtin, and A. Ritz, Appl. Surf. Sci. 111, 11 (1997). 102. R. Stratton, Phys. Rev. A 135, 794 (1964). 103. R. E. Burges, H. Kroemer, and J. M. Houston, Phys. Rev. 90, 515 (1953). 104. O. Gröning, O. M. Küttel, Ch. Emmenegger, P. Gröning, and L. Schlapbach, J. Vac. Sci. Technol. B 18, 665 (2000). 105. IBM Patent Server, www.patents.ibm.com. 106. O. Gröning, O. M. Küttel, E. Schaller, P. Gröning, and L. Schlapbach, Appl. Phys. Lett. 68, 476 (1996). 107. O. Gröning, O. M. Küttel, P. Gröning, and L. Schlapbach, J. Vac. Sci. Technol. B 17, 1064 (1999). 108. O. Gröning, O. M. Küttel, P. Gröning, and L. Schlapbach, Appl. Phys. Lett. 71, 2253 (1997). 109. O. M. Küttel, O. Gröning, Ch. Emmenegger, L. O. Nilsson, E. Maillard, L. Diederich, and L. Schlapbach, Carbon 37, 745 (1999). 110. R. Latham, “High Voltage Vacuum Insulation.” Academic Press, San Diego, 1995. 111. O. Gröning, O. M. Küttel, P. Gröning, and L. Schlapbach, J. Vac. Sci. Technol. B 17, 1970 (1999). 112. O. Gröning, L. O. Nilsson, P. Gröning, and L. Schlapbach, Solid State Electron. 45, 929 (2001). 113. O. Gröning, O. M. Küttel, P. Gröning, and L. Schlapbach, Appl. Surf. Sci. 111, 135 (1997).

579 114. O. M. Küttel, O. Gröning, Ch. Emmenegger, and L. Schlapbach, Appl. Phys. Lett. 73, 2113 (1998). 115. K. A. Dean, O. Gröning, O. M. Küttel, and L. Schlapbach, Appl. Phys. Lett. 75, 2773 (1999). 116. L. D. Landau and E. M. Lifshitz, “Electrodynamics of Continuous Media,” Pergamon, New York, 1960. 117. D. M. Wood, Phys. Rev. Lett. 46, 749 (1981). 118. G. Markov, A. Nitzan, and L. E. Brus, J. Chem. Phys. 88, 5076 (1988). 119. V. Bonacic-Koutecky, P. Fantucci, and J. Koutecky, Chem. Rev. 91, 1035 (1991). 120. E. W. Müller, Z. Phys. 106, 541 (1937). 121. L. O. Nilsson, Microscopic Characterisation of Electron Field Emission from Carbon Nanotubes and Carbon Thin Film Emitters, Ph.D. Thesis No. 1337, University of Fribourg, Switzerland, 2001. 122. A. G. Rinzler, J. H. Hafner, P. Nikolaev, L. Lou, S. G. Kim, D. Tomanek, P. Nordlander, D. T. Colbert, and R. E. Smalley, Science 269, 1550 (1995). 123. K. Dean and B. R. Chalamala, Appl. Phys. 85, 3822 (1999). 124. L. O. Nilsson, O. Gröning, O. Küttel, P. Gröning, and L. Schlapbach, J. Vac. Sci. Technol. B 20, 326 (2002). 125. L. O. Nilsson, O. Gröning, P. Gröning, and L. Schlapbach, Appl. Phys. Lett. 79, 1036 (2001). 126. L. O. Nilsson, O. Gröning, Ch. Emmenegger, O. M. Küttel, E. Schaller, and L. Schalpbach, Appl. Phys. Lett. 76, 2071 (2000). 127. S. Frank, P. Poncharal, Z. L. Wang, and W. A. de Heer, Science 280, 1744 (1998). 128. A. Zettel and J. Cumings, in “Proc. XIV Inter. Winterschool on Electr. Prop. of Novel Materials,” Kirchberg, Tirol, Austria, 2000, p. 526. 129. P. Ruffieux, Ph.D. Thesis No. 1387, University of Fribourg, Switzerland, 1999. 130. E. Maillard-Schaller, Ph.D. Thesis No. 1145, University of Fribourg, Switzerland, 1996.

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Carbon Nanotube Growth by Chemical Vapor Deposition M. Meyyappan NASA Ames Research Center, Moffett Field, California, USA

CONTENTS 1. Introduction 2. Growth Apparatus 3. Catalyst Preparation 4. Growth Results 5. Growth Mechanisms 6. Applications for CVD/PECVD Grown Nanotubes Glossary References

1. INTRODUCTION The interest in carbon nanotubes (CNTs) and the level of research activities across the world have been extraordinary in the last decade. CNTs were first discovered by Sumio Iijima of the NEC Corporation in 1991 [1] in the soot of an arc-discharge apparatus. These are elongated fullerenes with diameters as small as 0.7 nm and lengths of up to several microns. Single-walled carbon nanotubes (SWNTs) exhibit unique electronic properties in that they can be metallic or semiconducting, depending on their helicity. This allows the formation of semiconductor–semiconductor and semiconductor–metal junctions useful in device fabrication. SWNTs also possess extraordinary mechanical properties. The Young’s modulus of individual SWNTs has been estimated to be around 1 TPa, and the yield strength can be as large as 120 GPa [2]. The interesting combination of electronic and mechanical properties of CNTs has led to wide-ranging investigations of their potential in future electronics and computing, field-emitter devices, sensors, electrodes, high-strength composites, and storage of hydrogen, lithium, and other metals. For a detailed discussion on the properties and applications, the reader is referred to several volumes edited by Saito et al. [2], Meyyappan and Srivastava [3], and other review articles [4]. This chapter focuses on CNT growth and large-scale production by chemical vapor deposition (CVD) and related techniques such as plasma-enhanced

ISBN: 1-58883-057-8/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

CVD (PECVD). The early processes used for CNT production were laser ablation and an arc-discharge approach. Both are able to produce single- and multiwalled nanotubes, and continue to be popular among researchers. Details on these processes can be found in [4–6]. CVD is amenable for nanotube growth on patterned surfaces, suitable for the fabrication of electronic devices, sensors, field emitters, and other applications where controlled growth over masked areas is needed for further processing. A variety of different CNT structures is possible by CVD and related techniques. A SWNT is a rolled-up tubular shell of graphene sheet which is composed of benzenetype hexagonal rings of carbon atoms. A multiwalled carbon nanotube (MWNT) is a rolled-up stack of graphene sheets into concentric cylinders. The walls of each layer of the MWNT or the graphite basal planes are parallel to the central axis (
= 0). In contrast, a stacked-cone arrangement (also known as a Chevron structure, ice-cream-cone structure, or piled-cone structure) is also seen, where the angle between the graphite basal planes and the tube axis is nonzero [7]. Nolan et al. [7] suggest that hydrogen satisfies the valences at cone edges in such structures. (An MWNT has no graphite edges, and therefore there is no need for valence-satisfying species such as hydrogen.) Since the stacked-cone structures exhibit only small
values, and are not solid cylinders but are mostly hollow, they can be called multiwalled carbon nanofibers (MWNFs) [8]. Note that the terminologies graphitic carbon fibers (GCFs) and vaporgrown carbon fibers (VGCFs) have long been used to denote solid cylinders. The CNT growth literature has grown rapidly in recent years. A reasonably complete (but not exhaustive) list of references in areas discussed in this chapter is given below: • • • •

early works on carbon fibers/filaments [9–15] thermal CVD of SWNTs [16–27, 59] thermal CVD of MWNTs [7, 28–55] CNT growth on atomic force microscope (AFM) cantilevers [56–59] • electric-field assistance in thermal CVD [60–62] • floating catalyst CVD of SWNTs [63–69] and MWNTs [70, 71]

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 1: Pages (581–589)

582

Carbon Nanotube Growth by CVD

• PECVD of MWNTs and MWNFs [8, 59, 72–92] — — — —

dc plasma [83–85, 87–91] plasma-assisted hot filament [72, 73, 75] microwave plasma [74, 76–82] RF capacitive and inductive [8, 59, 86, 92]

• theoretical analysis of CNT growth [93–95] • computational model and simulation of CNT growth [96–98]. This chapter is organized as follows: Section 2 presents an overview of the growth apparatus used in thermal CVD and PECVD; catalyst preparation techniques are outlined in Section 3; growth results from thermal CVD of SWNTs and MWNTs, the effect of the electric field on alignment, and plasma-grown nanotubes are presented in Section 4. Growth mechanism is the subject of Section 5, and a brief overview of applications for the thermal CVD- and PECVD-grown material is presented in Section 6.

2. GROWTH APPARATUS As markets for CNTs are not developed, there is no commercial equipment industry now, unlike the case with the semiconductor equipment industry. So, the equipment is mostly homemade, low-throughput batch reactors. When the market changes to large-scale product-oriented (devices, sensors, nanoelectromechanical systems, etc.) from the current academic-research-dominated scenerio, the worldwide equipment vendors (currently catering to the semiconductor and MEMs market) will undoubtedly move in to meet the demands since the growth hardware is very much similar, as will be seen below.

2.1. Thermal CVD The thermal CVD apparatus for CNT growth is very simple [18, 24, 59]. It consists of a quartz tube (diameter of 1–2 in) inserted into a tubular furnace capable of maintaining ±1  C over a 25 cm zone. Thus, it is a hot-wall system, and primarily atmospheric pressure CVD. Cold-wall reactors, common in the semiconductor industry, where the substrate holder is directly heated from below (resistance, inductance, or other heaters), have not been reported in the literature. Since the growth is catalyst promoted at temperatures of 500– 1000  C and does not depend on precursor dissociation at these temperatures, either a hot- or a cold-wall system could be designed to be effective for CNT growth. The substrate, typically about or smaller than 1 in now, is placed inside the quartz tube. In thermal CVD, either CO or some hydrocarbon such as methane, ethane, ethylene, acetylene, or other higher hydrocarbons is used without any dilution. The feedstock is metered through a mass flow controller. A typical growth run would involve purging the reactor first with argon or some other inert gas until the reactor reaches the desired growth temperature. Then the gas flow is switched to the feedstock for the specified growth period. At the end, the gas flow is switched back to the inert gas while the reactor cools down to 300  C or lower before exposing the nanotubes to air. Exposure to air at elevated temperatures can cause damage to the CNTs. Typical growth rates range from a few nm/min to 2– 5 m/min. CNT growth is largely empirical at present, and

diagnostics and modeling studies are rare. For example, the effects of reactor length and diameter, flow rate, and so on, on growth characteristics are completely unknown. One study using a mass spectrometer [99] and another involving a computational fluid dynamics modeling [98] confirm that the precursor dissociation (in the case of methane at 900  C) indeed is very small, and the growth proceeds due to the catalytic activity of methane on particle surfaces. For growth on substrates, the catalyst mixture needs to be applied to the substrate (see Section 3) prior to loading it inside the reactor. This is the so-called supported catalyst approach. In contrast, CVD can also be used to grow large quantities of nanotubes using a “floating-catalyst” approach [63–71]. In this case, a nozzle system [64, 65] may be used to inject the vaporized catalyst precursor into the flowing CO or hydrocarbon. A second furnace may be required to heat up the catalyst precursor system to its dissociation temperature [63, 70]. The floating-catalyst approach is amenable to scale up for large-scale production to meet the “commodity market” demands of nanotubes in a variety of structural applications.

2.2. PECVD The plasma enhancement in CVD first emerged in microelectronics because certain processes cannot tolerate the high wafer temperatures of the thermal CVD. For example, charring of photoresists on patterned wafers could be a problem. The plasma CVD allowed an alternative at substantially lower wafer temperatures (room temperature to 100  C), and hence, has become a key step in integratedcircuit manufacturing. The low-temperature operation is possible since the precursor dissociation (necessary for the deposition of all common semiconductor, metallic, and insulating films) is enabled by the high-energy electrons in an otherwise cold plasma. In contrast to this familiar scenerio, CNT growth does not need precursor dissociation, and as such, the well-known efficiency of a plasma to tear apart the precursor gas for reactive radical generation should not be a factor. What is known to date about the catalytic activity of the transition metal particles for CNT growth tells us that growth does not occur below 550  C. Hence, the cold wafer scenario is out of the question, at least for now, until new knowledge emerges. It is not clear, then, what the true role of the plasma in CNT growth is, despite a large number of articles in this field [8, 59, 72–92]. However, there is enough empirical evidence that the cold plasma enables more vertically aligned CNTs than thermal CVD [59, 75, 81, 85, 92]. Whereas any marginal alignment seen in thermal CVD samples is due to a crowding effect (nanotubes supporting each other by van der Waals attraction), individual, free-standing, and vertically oriented CNTs are possible with PECVD, as will be seen in Section 4. As in the case of earlier technology waves such as microelectronics materials and diamond deposition, researchers have attempted dc [83–85, 87–91], RF [86], hot-filament aided with dc [72, 73, 75], microwave [74, 76–82], electron cyclotron resonance, and inductively coupled plasma reactors [8, 59, 92]. An overview of plasma fundamentals and plasma equipment can be found in [100, 101]. A dc plasma reactor consists of a pair of electrodes in a grounded chamber, with one electrode grounded and the second connected

583

Carbon Nanotube Growth by CVD

to a power supply. The negative dc bias applied to the cathode leads to a breakdown of the feedgas. The wafer with the catalyst layer may be placed either on the anode or cathode for CNT deposition. The electrode holding the wafer may need an independent heating source to raise the wafer temperature to the desired growth temperature. Instead of a resistive heater underneath the electrode, a tungsten wire suspended in the plasma stream may serve as the heating source. This is the so-called hot-filament system with plasma assistance. These two systems have successfully yielded MWNTs and MWNFs, as evidenced from [72, 73, 75, 83–85, 87–91]. However, the bias on the wafer in these systems appears to be very high (>300 V). It is well known that ions gain substantial energy in the sheath from high electric fields, and the high-energy ion bombardment of the wafer often leads to damage, which is well known from silicon and III–V semiconductor literature. It is for this reason and the poor efficiency of the dc systems that the semiconductor industry abandoned these sources a few decades ago. Since the plasma can dissociate the hydrocarbon, creating many reactive radicals, pure hydrocarbon feedstock in plasma reactors may lead to substantial amorphous carbon deposition. Therefore, it is desirable to dilute the hydrocarbon with argon [8], hydrogen [8, 59], or ammonia 75 81. The pressure in the reactor typically ranges from 1–20 torr, with a hydrocarbon fraction of up to 20%. Atmospheric pressure operation of plasma systems is uncommon due to power coupling problems. Low-pressure operation in the tens of millitorrs would result in very slow growth rates. So, PECVD reactors are typically operated at 1–20 torr pressure levels for CNT growth. At these pressure levels, inductive coupling to hydrocarbon/H2 systems appears to be difficult; the coupling appears to have a large capacitive component [8, 92]. The capacitive component decreases with an increasing fraction of argon. Microwave sources are very popular at these pressures and power levels of up to 2 KW, and have been widely used for diamond deposition. Following this success, CNT literature also consists of several successful demonstrations of MWNT growth using microwave sources [79–82]. In addition to the source, matching network, and other power-coupling components, the equipment consists of mass flow controllers and one or more vacuum pumps. The growth chamber itself is grounded. All plasma reactors are cold-wall systems, with the substrate directly heated using some form of heat source from below the substrate holder. First, the wafer is loaded in the reactor, and the system is pumped down to 10−5 torr or below to minimize impurities and water vapor. The substrate holder is heated to the desired temperature. Then the feedstock is admitted, and the flow rate and chamber pressure can be set to desired levels independent of each other with the aid of a throttle valve. Next, the power from the power source is coupled to the plasma. At the end of the run, the heater, power source, and gas flow are turned off, and the system is purged with argon flow. The wafer is removed after the reactor cools down below 300  C.

3. CATALYST PREPARATION Although there have been some studies reporting CNT growth without catalysts in arc discharge processes, it is widely acknowledged that transition metal catalysts are needed for SWNT, MWNT, and MWNF growth by CVD. It is also believed that the catalyst on the substrate must be in the form of particles instead of smooth, continuous films. There have been several studies indeed correlating the catalyst particle size and the diameter of the resulting nanotubes [26, 41, 53, 55, 79, 86, 88, 102]. The metals used to date as catalysts include Fe, Ni, Co, and Mo. It is possible to apply these onto the substrate from solutions containing them or they can be directly deposited using some physical techniques. These two approaches are different in terms of needed resources, time and cost, and the nature of the resulting products. A brief overview is provided on the two routes to supported-catalyst preparation, and also the floating-catalyst approach for large-scale production.

3.1. Solution-Based Catalyst Preparation The literature contains numerous recipes for preparing catalysts from solutions, and one such recipe is given below [44]. First, 0.5 g (0.09 mmol) of Pluronic P-123 triblock copolymer is dissolved in 15 cm3 of a 21 mixture of ethanol and methanol. Next, SiCl4 (0.85 cm3 , 7.5 mmol) is slowly added using a syringe into the triblock copolymer/alcohol solution, and stirred for 30 min at room temperature. Stock solutions of AlCl3  6H2 O CoCl2  6H2 O, and FeNO3 3  6H2 O are prepared at the concentration of the structure directing agent (SDA) and inorganic salts. The catalyst solutions are filtered through 0.45 m polytetrafluoroethylene membranes before applying onto the substrate. The substrate with the catalyst formulation is loaded into a furnace, and heated at 700  C for 4 h in air to render the catalyst active by the decomposition of the inorganic salts and removal of the SDA. Admission of hydrocarbon feedstock into the reactor at this point would initiate nanotube growth. It is noted that a mixture of transition-metal-containing compounds along with structure-directing agents is used in the above recipe. It is difficult to find optimum concentrations of each constituent in a trial-and-error approach as the number of trials is large. Cassell and coauthors pioneered a combinatorial optimization process for catalyst discovery for the growth of SWNTs [25] and MWNTs [44]. This rapid throughput approach, coupled with characterization techniques, allows the development of catalyst libraries with a minimal number of growth experiments. In general, even with the right formulation known a priori, solution-based approaches are time consuming. A typical preparation includes such steps as dissolution, stirring, precipitation, refluxing, separation, cooling, gel formation, reduction, drying/annealing/calcinations, and so on. The overall process is cumbersome and time consuming; some recipes even call for overnight annealing. Another problem is the difficulty in confining the catalyst within small patterns.

584 3.2. Physical Techniques for Catalyst Preparation Physical techniques, such as electron gun evaporation [83– 85, 102], thermal evaporation [88], pulsed laser deposition [46], ion beam sputtering [8, 24, 55, 59, 92], and magnetron sputtering [75, 78, 79, 86, 91], have been successfully used in catalyst preparation. These techniques are quick, easy, and amenable to produce small patterns, in contrast to the solution-based approaches discussed in the previous section. Typically, a thin catalyst film ( EB (pores) > EB (surface) [61, 62]. Williams et al. calculated the binding energy and specific surface area contributions () of these sites to be (1) channels: EB = 0119 eV,  = 45 m2 /g, (2) grooves: EB = 0089 eV,  = 22 m2 /g, (3) pores: EB = 0062 eV,  = 783 m2 /g, and (4) surfaces: EB = 0049 eV,  = 483 m2 /g [61, 62]. Clearly, the channels and grooves have much larger binding energy but less available surface area. The outer surface has only weak adsorption consistent with the

Figure 6. Schematic structure of a SWCNT bundle showing different sites for gas adsorption. Reprinted with permission from [65], C. K. W. Adu et al., Chem. Phys. Lett. 337, 31 (2001). © 2001, Elsevier Science.

study of a single SWCNT surface [56]. These studies provide a good fundamental understanding of the gas adsorption on SWCNTs. For individual MWCNTs and MWCNT films, much less work has been done due to the complexity in tube structure. However, both the external and internal surfaces are expected to contribute to the gas adsorption. The defects, in particular, could play a more important role in changing the electrical properties of MWCNTs upon gas adsorption.

3.1.2. Gas Sensors Based on Single Carbon Nanotube FETs The Si-based field-effect transistors are sensitive to the electron transfer induced by gas adsorption and therefore could be used as gas sensors for certain applications [27]. SWCNTs are known to be semiconducting if their chirality m n satisfies m − n = 3 × integer. It was demonstrated that a semiconducting SWCNT (S-SWCNT) connected to two metal electrodes exhibits p-type transistor characteristics with the conductance tunable over several orders of magnitude by changing the gate voltage applied on the Si substrate [13]. Such a phenomenon was attributed to the adsorption of O2 as electron acceptor in ambient environment. The theoretical work mentioned above predicts about 0.1 electron transfer per O2 molecule [56]. With the typical diameter (∼2 nm) and length (∼3
m) of the S-SWCNT in the FET device (see Fig. 3), the adsorption/desorption of even a single O2 molecule can produce about a 1 × 1016 e/mm3 charge density in the SWCNT if simply calculated from the volume. This is about the level of the charge density in a moderately doped Si device. In reality, the quantum wire nature of SWCNT makes the conductance of the tube even more sensitive to any local charge along the one-dimensional

596

Molecular interactions were found not only to affect the conductance of a single S-SWCNT but also to change that of bulk samples consisting of SWCNT bundles/ropes. Br and I intercalations were reported to enhance the conductance of bulk film consisting of SWCNT ropes [62, 63]. Kong et al. demonstrated that the resistance of a SWNT mat decreased to about a half of the original value after exposure to a 200 ppm NO2 flow for ∼10 min while it increased to about 1.5 times of the original value in a 1% NH3 flow [13]. Apparently, both metallic and semiconducting SWCNTs in the bulk sample are affected by the gas adsorption. The effect of different gas molecules such as NH3 and NO2 on the bulk sample is very consistent with that of the single S-SWCNT transistors. Collins et al. also found that the electrical resistance R measured with a thin film of SWCNT bundles is particularly sensitive to O2 in the environment [23]. An experiment measuring the electrical resistance switched by 10 to 15% as the chamber was alternately flooded with air or evacuated to ∼10−6 Torr. Measurable changes in R were found at oxygen partial pressure as low as 10−10 Torr. However, other major gas constituents of air such as N2 , CO2 , H2 O, etc. and inert gas molecules such as Ar and He have little effect on the electrical properties of the SWCNT film. Further transport experiments indicate that once SWCNTs have been exposed to oxygen, it is not possible to fully deoxygenate them at room temperature even under highvacuum conditions. This irreversibility is similar to that of

(a)

25

R (Kohm)

3.1.3. Electrical Resistance Based Gas Sensors

NO2 adsorption on a S-SWCN transistor gas sensor, which limits the application of CNTs in gas sensors. Apparently, electron acceptors such as O2 and NO2 have dramatic effects on the electrical properties of SWCNTs. Many previously reported electronic properties of CNTs are dominated by preadsorbed O2 from ambient environment. The slow recovery time or irreversibility is likely due to the strong adsorption or capillary effects, particularly when using open-ended SWCNT bundles. MWCNTs with much larger tube diameter, on the other hand, can change from microporous (such as SWCNT bundles) to mesoporous or even macroporous materials. MWCNTs could potentially have much less interaction with the gas molecules. The sensitivity of electrical resistance R to the gas environment may drop but the reversibility could be improved. Ng et al. reported a study of gas sensing using a 1.5 mm × 1.5 mm × 25
m freestanding MWCNT membrane grown on a polymer template by thermal CVD (as shown in Fig. 4) [25]. Due to the decomposition and shrinkage of the polymer material, the MWCNT–catalyst composite membrane peeled off from the substrate while still maintaining the desired pattern from the substrate. Thus formed membrane contains loosely entangled MWCNTs mixed with catalyst and carbon particles. As shown in Figure 7, simple electrical resistance measurements following cycling the chamber pressure between ambient atmosphere and high vacuum

20 15 10 5 0 10 20 30 40 50 60 70 80 90 100 110 12

Time (min) 0.8

(b)

0.7 0.6

∆R/R

wire. Therefore, the sensitivity of single molecule adsorption/desorption could, in principle, be achieved with such miniaturized S-SWCNT FETs. Kong et al. [13] demonstrated that the conductance of a single S-SWCNT sample decreased by about 100-fold after exposure to NH3 (0.1 to 1% in Ar or air) for about 10 minutes. On the other hand, exposure to NO2 molecules [2 to 200 parts per million (ppm) in Ar] increased the conductance of the S-SWCNT sample by about 1000 times when it was initially depleted by a back-gate voltage Vg of +4 V. The sensitivity, that is, the ratio between the resistance after and before gas exposure (Rafter /Rbefore ), was found to be significantly better than other room-temperature operated gas sensors based on conventional materials. Since the SWCNT is a hole-doped semiconductor, the adsorption of electron donating NH3 molecules causes the hole depletion and reduced conductance while NO2 exposure results in enriched hole carriers and enhanced conductance. The chemical nature of the gas molecules is indicated in the I–Vg curve which is characteristically shifted by −4 V after exposure to NH3 and shifted by +4 V after exposure to NO2 . It is possible to fabricate arrays of such miniaturized sensing devices with ultrahigh sensitivity as well as chemical characteristics. A significant drawback is the slow recovery time (∼12 hr) after exposure to NO2 , which is likely due to the slow desorption of strong adsorbates. The influence of O2 in the ambient environment to the initial sample is also worth further investigation. Much more systematic study is still needed to develop this system for practical applications.

Carbon Nanotube Sensors

0.5 0.4 0.3 0.2 0.1 0.0 0

5

10

15

20

25

30

Time (min) Figure 7. Variations in the electrical resistance (R) of the MWCNT membrane with the surrounding medium, (a) spontaneous changes in R while cycling the environment between vacuum and ambient atmosphere, and (b) the ratio of the electrical resistance drop from the value at the base pressure vs different oxygen dosing pressure: •, 64 × 10−7 Torr;
31 × 10−4 Torr;  90 × 10−4 Torr; 14 × 10−3 Torr; ♦ 21 × 10−3 Torr, and  27 × 10−3 Torr. Reprinted with permission from [25], H. Ng et al., J. Nanosci. Nanotech. 1, 375 (2001). © 2001, American Scientific Publishers.

597

Carbon Nanotube Sensors

show a very reversible response. R drops by a factor of three (from about 24 to about 7 kOhm) when venting the vacuum with air. Similar to Collins et al. [23], Ar or N2 did not show any effect. The recovery time is about 2 to 3 minutes which is likely limited by the pumping and venting speed. Stepwise increase of the oxygen base pressure indicates that R is most sensitive to O2 in the range from 10−6 to 10−3 Torr. About 80% of the changes in R occurred at oxygen base pressure less than 10−3 Torr. Over about 3 × 10−3 Torr, the change in R is saturated likely because all the surface sites have already been occupied. It is worth noting that most studies [13, 23, 25] were carried out at room temperature, at which oxygen is difficult to remove once they adsorbed onto the CNTs. As a result, the measured electrical resistance R was dominated by the preadsorbed O2 molecules and little change was observed when exposing the sample to inert gases such as N2 and Ar. However, Sumanasekera et al. found that the fourprobe electrical resistance RT of a mat of SWNT bundles showed easily measurable change even upon exposure to 1 atm of N2 or He after being degassed in high vacuum at elevated temperature (∼500 K) [24]. The change in R is small (∼2–3% increase from vacuum to 1 atm N2 and ∼10% increase from vacuum to 1 atm He). The recovery time is slow (>2 hr) but quite reversible. The increase in R upon exposure to inert gas was attributed to the gas collisions with the nanotube wall.

3.1.4. Thermoelectric “Nanonoses” The thermoelectric power (TEP) S, defined as S = V /T

(1)

where V and T are voltage difference and temperature difference between two spots in the sample, is another parameter strongly correlated with the electronic properties of the materials. Collins et al. studied the change in TEP upon cycling the environment between vacuum and pure atmospheric pressure oxygen [23]. The response was found to be in good consistency with the change of R. The oxygen exposed SWCNT film has a positive TEP of about +20
V/K. The value quickly decreases and reaches about −10
V/K a few minutes after oxygen was removed from the chamber. This phenomenon was attributed to the change in the carrier (hole) density upon the adsorption/desorption of O2 as electron acceptors. Eklund et al. did a series of careful work at elevated temperature (∼500 K) at which the adsorption/desorption of both strong adsorbates and inert gas molecules is completely reversible [24, 65, 66]. They attributed the strongly negative TEP after degassing to the change of metallic tubes to n-type semiconductors. As we mentioned earlier, at 500 K, easily measurable electrical resistance change was obtained even upon exposure to inert gases. Both TEP and the SWNT bundle resistivity  can be measured at different wt% of gas adsorption. Eklund et al. [65] attributed the total bundle resistivity to two things: (1) the scattering mechanisms preexisting in the bundle before gas adsorption (0 ), for example, from phonons and wall defects, and (2) the extra impurity scattering due to the adsorbed gas (a ) with  = 0 + a

(2)

Approximately, the TEP and resistivity have the Nordheim– Gorter (NG) relation as S = S0 + a /0 Sa − S0

(3)

where S0 and Sa are the contributions to the thermoelectric power from the host resistivity 0 and additional impurity resistivity a from the adsorbed gas, respectively. If the gas molecules are physisorbed on the tube walls by van der Waals interaction, EF will be unaffected and Sa − S0 is a constant, resulting in linear NG curves as shown in Figure 8 for N2 , He, and H2 . On the other hand, strong chemisorption has a much more profound effect on the host band structure and EF . Thus, (Sa − S0 ) must depend on the gas coverage, resulting in nonlinear NG plots as shown in Figure 8 for O2 and NH3 . It is noteworthy that the NG plots for even inert gases such as N2 , He, and H2 are significantly distinct from each other. Potentially, the NG plots at elevated temperature can differentiate different molecules and find applications in thermoelectric “nanonose” as proposed by Eklund et al. [65].

3.1.5. Remote Gas Sensors The above-mentioned gas sensors based on single CNT FET, electrical resistance, and TEP all rely on direct electrical connections to the sensing elements exposed in the sampling environments. For some applications such as detection of corrosive gases and samples in sealed environments, remote or non-direct-contact sensors are required. Recently, Chopra et al. reported a carbon-nanotube-based resonantcircuit sensor for ammonia [47]. Similar to the electrical resistance and TEP sensors, tens of micrometer thick CNT film was used as the adsorbent material due to the large uptake capability defined by its huge surface area. Both

Figure 8. NG plots (S vs ) showing the effect of gas adsorption on the electrical properties of the SWCNT mat. Reprinted with permission from [65], C. K. W. Adu et al., Chem. Phys. Lett. 337, 31 (2001). © 2001, Elsevier Science.

598 SWCNTs and MWCNTs were used. The CNT films were simply deposited on a disk shaped Cu resonator on a Duroid board. A radio frequency transmitter sends a microwave signal to the interrogation region of the Cu resonator, producing a strong signal at its resonant frequency (∼4 GHz). The resonant frequency was found shift downward by a few MHz after exposure to NH3 gas over ∼100 ppm. The response and recovery time were found to be about 10 min, a reasonable short time. The resonant frequency shift was attributed to the change in the effective dielectric constant of the CNT film upon exposure to NH3 . In principle, other strong adsorbates may affect the downshift of the resonant frequency differently and the chemical nature of the gas may be derived. Even though the sensitivity is low, the remote sensor may find applications in certain corrosive or toxic environments as well as those that require nondestructive testing.

3.2. Liquid-Phase Chemical Sensors From the structural point of view, the main body of CNTs, particularly MWNTs, can be considered as graphitic layers rolled into cylinders. The sidewall of CNTs is expected to present similar physical and electronic properties to the graphite basal plane. The studies discussed in previous sections on the gas adsorption on the sidewall of CNTs is, in many aspects, consistent with previous studies on graphite surfaces [67–69] except for the difference in the surface area. Along the tube axis, it is expected to present high conductivity similar to graphite in-plane properties. For SWCNTs, the large curvature and small tube dimension are likely to cause interesting quantum effects to deviate from graphite properties. The electronic properties of SWCNTs are also easily affected by adsorbates since almost every carbon atom is exposed at the surface. MWCNTs, on the other hand, have essentially only the outmost and innermost shells exposed to other molecules and thus are less sensitive to the perturbation from adsorbates. Their larger size also makes them more similar to graphite cylinders. In liquid solutions, the sidewall of MWCNTs is expected to present an anomalously small effective capacitance and a slow electron transfer rate similar to graphite basal plane electrodes. However, the open end is a very active electrode similar to the graphite edge plane. Electron transfer along the tube axis is expected to be fast similar to that within a graphite plane. These properties made CNTs, particularly MWCNTs, very interesting electrode materials. Electrons could be caught at the active end and transported along the tube to the other end without much interference from the solution. For this purpose, the interest is in the nanostructure of CNTs which improves the performance over conventional materials rather than the influence of their electronic properties by other molecules. In this section, we summarize some of the studies using CNTs as electrodes, which may potentially lead to dramatic improvements in the performance of current chemical and biological sensors by incorporating unique properties of CNTs. Nugent et al. demonstrated an ideal Nerstain behavior of the K4 Fe(CN)6 reaction using an open end WCNT bundle produced by electric arc-discharge [70]. The fast electron transfer rate from the tube ends along the tube axis agrees well with the cylindrical graphite model. Ideal reversible

Carbon Nanotube Sensors

cyclic voltammetries were also obtained in measuring the oxidation of neurotransmitter dopamine [48] and redox proteins such as cytochrome c and a “blue” copper protein azurin [71]. Efficient electrocatalytic reduction of dissolved O2 was also demonstrated [72]. However, all these studies use compact CNT paste or big bundles in which CNTs mixed with each other as well as organic binders in an uncontrolled way. Even though the specific capacitance of the graphitic sidewall is relatively small, they still contribute a huge capacitive background due to the enormous surface area [49, 73, 74]. For chemical sensor applications, the exposed surface area needs to be minimized. It was demonstrated that a loosely packed MWCNT membrane is able to detect neurotransmitter catechols at concentrations down to a few
Ms [25]. A series of CNTs with different densities was carefully studied by Li et al. to explore these issues [50]. Much work is desired to develop well-defined CNT electrodes from both the fundamental and application points of view. Campbell et al. demonstrated that single MWCNTs ∼80–200 nm in diameter and 15–50
m in length show reversible characteristic steady-state radial diffusion behavior consistent with fast electron transfer [75]. The sidewall of the CNT can be insulated with polymer and only the very end of the tube is exposed. This can find unique applications in the analysis of redox biospecies such as neurotransmitters and proteins in a live single cell [76]. Another effort in the development of well-defined CNT electrodes was demonstrated by NASA scientists using vertically aligned MWCNTs [50]. Li et al. used vertically aligned CNTs with an average diameter of ∼50–100 nm and nearestneighbor distance of ∼200–500 nm on metal coated Si substrates. SiO2 was deposited by TEOS CVD to insulate both CNTs and subtrate metal layers. A mechanical polishing process followed to planarize the sample to remove excess materials which results in an array of nanoelectrodes with only the open ends of MWCNTs exposed. This method provides a unique bottom-up scheme to fabricate nanoelectrode ensembles (NEEs) similar to those demonstrated by Menon and Martin [77] by filling metals into the onedimensional pores in the filtration membranes. The sensitivity is expected to be improved by orders of magnitude for measuring small redox species using such NEEs compared to conventional electrodes. Ordered nanoelectrode arrays with precisely controlled size and density could be achieved by employing appropriate lithographic techniques. The nanoelectrode array approach is very promising since it takes full advantage of CNTs via miniaturization to achieve ultimate sensitivity and may be incorporated into microfluidic devices, microelectromechanical systems (MEMS) chromatography instruments, and biomedical/ environmental monitoring devices.

3.3. Biosensors As we described in the discussion of S-SWCNT FET gas sensors [13], even small electron transfer between S-SWCNT and adsorbates can cause a significant change in the conductance of the device. Biomolecules such as DNA, RNA, and proteins are all heavily charged molecules in most conditions. The adsorption of such molecules onto the CNT

599

Carbon Nanotube Sensors

surface is expected to change its electronic properties similar to gas molecules. Therefore, the FET devices could be employed as biosensors. Particularly, recognition of specific biomolecules could be realized through the specific interaction with preimmobilized molecules. For this purpose, the functionalization of biomolecules such as DNA and proteins was intensively pursued in the past few years. Dai et al. have reported a series of work for sidewall functionalization of single SWCNTs for protein immobilization [78, 79]. The SWNT is intended to be used as the sensing element based on its electronic properties. As a result, the graphitic sp2 sidewall structure has to be preserved. Noncovalent bonding of some precursor molecules on the CNT sidewall was demonstrated for this purpose, either by the -stacking of the conjugate pyrenyl group of 1-pyrenebutanoic, succinimidyl ester, or coating with a surfactant (Triton) mixed with amine-terminated poly(ethylene glycol). Both methods efficiently modified the CNT surface and incorporated functional groups so that further covalent bonding of protein molecules such as biotin can be followed. The functionalized CNTs were then demonstrated with specific recognition of ferritin and streptavidin. These methods can be extended to the recognition of other biomolecules based on specific interactions of antibody–antigen and complimentary DNA strands. However, the response in electrical conductance has not been reported so far. The very end of the CNT, particularly the open end, is chemically much more active due to the dangling sp2 bonds similar to the graphite edge plane. The chemistry to functionalize these sites is more mature. Wong et al. demonstrated that the open end of the CNT is rich in COOH group and thus can be used for selective covalent bonding of primary amine molecules by carbodiimide chemistry [18, 80]. NASA scientists further demonstrated that DNA oligoprobes can be covalently functionalized, by the same chemistry, to an array of vertically aligned MWNTs embedded in SiO2 matrix with only the open ends exposed [81]. Similar to other DNA microarrays, the covalently bonded DNA probes can sustain more stringent control such as washing at elevated temperatures so that only specifically binded target DNA molecules can be hybridized with the probes. A NASA–NCI joint work has demonstrated an electrochemical DNA sensor based on an array of DNA functionalized nanotube electrodes [51, 81, 82]. The nano- bioelectronics platform is expected to provide unsurpassed electrochemical sensitivity, a high degree of miniaturization, great ease, and low cost operation.

4. DISCUSSIONS AND FUTURE DEVELOPMENTS In summary, CNTs are attractive materials for various sensor applications. Their small dimension and large surfaceto-volume ratio make them very sensitive to adsorbates from the environment. The electronic properties such as resistance/conductance, thermoelectric power, and dielectric constant can directly indicate the adsorption of target species. For applications which do not involve direct measurements of the inherent electrical properties of CNTs, the finite size and unique graphitic properties of CNTs can

still dramatically improve the performance, as demonstrated in electrochemical applications. It has been reported that CNTs can be used in sensing gases, liquid-phase chemicals, and biomolecules. The work summarized here only covers the CNT which is only one of the doors leading us to the nanotechnology world. Other nanomaterials such as semiconducting or metallic nanowires [83], nanoparticles [84, 85], quantum dots [86–88], and dendrimers [89] have all shown great potential in sensor applications. These nanomaterials allow bridging of the gap between microdevices and biomolecules to achieve an integrated micro-nano-bio heterogeneous system. We should expect great progress in this field in the near future. A few papers reported the most recent progress on gas adsorption on CNTs during the period in which manuscript was written. A Ricca et al. [90] studied the physisorption and chemisorption of O2 on the inner and outer walls of a 9 0 CNT using ONIOM method and the MP2 correlation treatment which includes long-range dispersion interactions. In contrast to previous theoretical work using density functional theory [55, 56], only physisorption of triplet O2 occurs both inside and outside the tube with very weak binding energy and little charge transfer. Chemisorption likely occurs with singlet O2 which has a very high barrier and is not popular under normal conditions. The CNT FETs were also proposed to operate as unconventional “Schottky barrier transistors”, in which transistor action occurs primarily by varying the contact resistance rather than channel conductance [91]. Therefore, the mechanism of the CNT FET sensor for gas adsorption needs further investigation.

GLOSSARY Carbon nanotube Seamless graphitic cylindrical fibers made of carbon atoms. It could be single walled if there is only one graphitic sheet in the cylindrical structure or multiwalled if there is more than one sheets rolled into the tube. Chemical mechanical polishing (CMP) A process typically used in the semiconductor industries to planarize the surfaces of a water to achieve optically flat topography. A slurry which consists of small spherical particulates and additives is used during the polishing process. The compositions of the slurry could be tailored to selectively polish the material of interest. Chemical vapor deposition (CVD) A process which involves deposition of a thin film using a selected choice of chemical precursors at elevated temperature. Deoxyribonucleic acid (DNA) A macromolecule consisting of one or two strands of linked deoxyribonucleotides. DNA hybridization Formation of a duplex structure by two complementary single strands of DNA. Polydimethylsiloxane (PDMS) A network of polymer chains with Si–O backbones and methyl (CH3) side group. Single-walled carbon nanotube ropes/bundles Multiple single-walled carbon nanotubes aggregated into hexagonally packed ropes/bundles side by side due to the strong van der Waals interactions between them.

600 Soft lithography A lithographic technique which involves non-optical components to achieve patterning of features with dimensions ranging from the millimeter to nanometer length scale.

ACKNOWLEDGMENTS We acknowledge helpful discussions with Jie Han, Jing Kong, Jing Li, and M. Meyyappan during preparation of this manuscript.

REFERENCES 1. M. S. Dresselhaus, G. Dresselhaus, and P. C. Eklund, “Science of Fullerenes and Carbon Nanotubes.” Academic Press, New York, 1996. 2. T. W. Ebbessen, “Carbon Nanotubes: Preparation and Properties.” CRC Press, Boca Raton, FL, 1997. 3. R. Saito, M. S. Dresselhaus, and G. Dresselhaus, “Physical Properties of Carbon Nanotubes.” World Scientific, New York, 1998. 4. D. Tománek and R. Enbody, “Science and Application of Nanotubes.” Kluwer Academic, New York, 2000. 5. S. Iijima, Nature 354, 56 (1991). 6. S. J. Tans, A. R. M. Verschueren, and C. Dekker, Nature 393, 49 (1998). 7. M. S. Fuhrer, J. Nygard, L. Shih, M. Forero, Y.-G. Yoon, M. S. C. Mazzoni, H. J. Choi, J. Ihm, S. G. Louie, A. Zettl, and P. L. McEuen, Science 288, 494 (2000). 8. C. W. Zhou, J. Kong, E. Yenilmez, and H. Dai, Science 290, 1552 (2000). 9. T. Rueckes, K. Kim, E. Joselevich, G. Y. Tseng, C. L. Cheung, and C. M. Lieber, Science 289, 94 (2000). 10. V. Derycke, R. Martel, J. Appenzeller, and Ph. Avouris, Nano Letters 1, 453 (2001). 11. A. Bachtold, P. Hadley, T. Nakanishi, and C. Dekker, Science 294, 1317 (2001). 12. X. L. Liu, C. Lee, C. W. Zhou, and J. Han, Appl. Phys. Lett. 79, 3329 (2001). 13. J. Kong, N. R. Franklin, C. W. Zhou, M. G. Chapline, S. Peng, K. Cho, and H. Dai, Science 287, 622 (2000). 14. B. Vigolo, A. Penicaud, C. Coulon, C. Sauder, R. Pailler, C. Journet, P. Bernier, and P. Poulin, Science 290, 1331 (2000). 15. W. A. de Heer, A. Chatelain, and D. Ugarte, Science 270, 1179 (1995). 16. A. G. Rinzler, J. H. Hafner, P. Nikolaev, L. Lou, S. G. Kim, D. Tomanek, P. Nordlander, D. T. Colbert, and R. E. Smalley, Science 269, 1550 (1995). 17. H. Dai, J. H. Hafner, A. G. Rinzler, D. T. Colbert, and R. E. Smalley, Nature 384, 147 (1996). 18. S. Wong, E. Joselevich, A. Woolley, C. Cheung, and C. Lieber, Nature 394, 52 (1998). 19. J. Li, A. Cassell, and H. Dai, Surf. Interface Anal. 28, 8 (1999). 20. C. V. Nguyen, K. J. Chao, R. M. D. Stevens, L. Delzeit, A. Cassell, J. Han, and M. Meyyappan, Nanotechnology 12, 363 (2001). 21. C. F. Liu, Y. Y. Fan, M. Liu, H. T. Cong, H. M. Chen, and M. S. Dresselhaus, Science 286, 1127 (1999). 22. G. Che, B. B. Lakshmi, E. R. Fisher, and C. R. Martin, Nature 393, 346 (1998). 23. P. G. Collins, K. Bradley, M. Ishigami, and A. Zettl, Science 287, 1801 (2000). 24. G. U. Sumanasekera, C. K. W. Adu, S. Fang, and P. C. Eklund, Phys. Rev. Lett. 85, 1096 (2000). 25. H. T. Ng, A. Fang, J. Li, and S. F. Y. Li, J. Nanosci. Nanotech. 1, 375 (2001). 26. G. Heiland, Sensors Actuators 2, 343 (1982).

Carbon Nanotube Sensors 27. A. Mandelis and C. Christofides, “Physics, Chemistry, and Technology of Solid State Gas Sensor Devices.” Wiley, New York, 1993. 28. M. S. Wrighton, Science 231, 32 (1986). 29. M. C. Longergan, E. J. Severin, B. J. Doleman, S. A. Beaber, R. H. Grubbs, and N. S. Lewis, Chem. Mater. 8, 2298 (1996). 30. S. Iijima and T. Ichihashi, Nature 363, 603 (1993). 31. D. S. Bethune, C. H. Kiang, M. S. de Vries, G. Gorman, R. Savoy, J. Vazquez, and R. Beyers, Nature 363, 605 (1993). 32. A. Thess, R. Lee, P. Nikolaev, H. Dai, P. Petit, J. Robert, C. Xu, Y. H. Lee, S. G. Kim, A. G. Rinzler, D. T. Colbert, G. E. Scuseria, D. Tománek, J. E. Fischer, and R. E. Smalley, Science 273, 483 (1996). 33. H. Kajiura, S. Tsutsui, H. Huang, M. Miyakoshi, Y. Hirano, A. Yamada, and M. Ata, Chem. Phys. Lett. 346, 356 (2001). 34. C. Journet, W. Maser, P. Bernier, A. Loiseau, M. Delachapelle, S. Lefrant, P. Deniard, R. Lee, and J. Fischer, Nature 388, 756 (1997). 35. J. Kong, A. M. Cassell, and H. Dai, Chem. Phys. Lett. 292, 567 (1998). 36. M. Su, B. Zheng, and J. Liu, J. Chem. Phys. Lett. 322, 321 (2000). 37. J.-F. Colomer, C. Stephan, S. Lefrant, G. Tendeloo, I. Willems, Z. Kónya, A. Fonseca, C. Laurent, and J. Nagy, J. Chem. Phys. Lett. 317, 83 (2000). 38. E. Flahaut, A. Govingaraj, A. Peigney, C. Laurent, and C. N. Rao, Chem. Phys. Lett. 300, 236 (1999). 39. J. Hafner, M. Bronikowski, B. Azamian, P. Nikolaev, D. Colbert, and R. E. Smalley, Chem. Phys. Lett. 296, 195 (1998). 40. M. J. Bronikowski, P. A. Willis, D. T. Colbert, K. A. Smith, and R. E. Smalley, J. Vac. Sci. Technol. A 19, 1800 (2001). 41. S. Bandow, A. M. Rao, K. A. Williams, A. Thess, R. E. Smalley, and P. C. Eklund, J. Phys. Chem. B 101, 8839 (1997). 42. K. Tohji, H. Takahashi, Y. Shinoda, N. Shimizu, B. Jeyadevan, I. Matsuoka, J. Phys. Chem. B 101, 1974 (1997). 43. A. G. Rinzler, J. Liu, H. Dai, P. Nikolaev, C. B. Huffman, F. J. Rodríguez-Macías, P. J. Boul, A. H. Lu, D. Heymann, D. T. Colbert, R. S. Lee, J. E. Fischer, A. M. Rao, P. C. Eklund, and R. E. Smalley, Appl. Phys. A 67, 29 (1998). 44. Z. Shi, Y. Lian, F. Liao, X. Zhou, Z. Gu, Y. Zhang, and S. Iijima, Solid State Comm. 112, 35 (1999). 45. M. Cinke, J. Li, B. Chen, A. Cassell, L. Delzeit, J. Han, and M. Meyyappan, Chem. Phys. Lett. 365, 69 (2002). 46. H. T. Soh, C. F. Quate, A. F. Morpurgo, C. M. Marcus, J. Kong, and H. Dai, Appl. Phys. Lett. 75, 627 (1999). 47. S. Chopra, A. Pham, J. Gaillard, A. Parker, A. M. Rao, Appl. Phys. Lett. 80, 4632 (2002). 48. J. M. Nugent, K. S. V. Santhanam, A. Rubio, and P. M. Ajayan, Nano Letters 1, 87 (2001). 49. P. J. Britto, K. S. V. Santhanam, and P. M. Ajayan, Bioelectrochem. Bioenergetics 41, 121 (1996). 50. J. Li, A. Cassell, L. Delzeit, J. Han, and M. Meyyappan, J. Phys. Chem. B 106, 9299 (2002). 51. J. Li, R. Stevens, L. Delzeit, H. T. Ng, A. M. Cassell, J. Han, and M. Meyyappan, Appl. Phys. Lett. 81, 910 (2002). 52. A. C. Dillon, K. M. Jones, T. A. Bekkedahl, C. H. Kiang, D. S. Bethune, and M. J. Heben, Nature 386, 377 (1997). 53. Y. A. Ye, C. C. Ahn, C. Witham, B. Fultz, J. Liu, A. G. Rinzler, D. Colbert, K. A. Smith, and R. E. Smalley, Appl. Phys. Lett. 74, 2307 (1999). 54. S. M. Lee and Y. H. Lee, Appl. Phys. Lett. 76, 2877 (2000). 55. S.-H. Jhi, S. G. Louie, and M. L. Cohen, Phys. Rev. Lett. 85, 1710 (2000). 56. J. Zhao, A. Buldum, J. Han, and J. P. Lu, Nanotechnology 13, 195 (2002). 57. J. Liu, A. G. Rinzler, H. Dai, J. H. Hafner, R. K. Bradley, P. J. Boul, A. Lu, T. Inverson, K. Shelimov, C. B. Huffman, F. RodriguezMacias, Y.-S. Shon, T. R. Lee, D. T. Colbert, and R. E. Smalley, Science 280, 1253 (1998).

Carbon Nanotube Sensors 58. A. Kuznetsova, D. B. Mawhinney, V. Naumenko, J. T. Yates Jr., J. Liu, and R. E. Smalley, Chem. Phys. Lett. 321, 292 (2000). 59. E. B. Mackie, R. A. Wolfson, L. M. Arnold, K. Lafdi, and A. D. Migone, Langmuir 13, 7197 (1997). 60. K. G. Ayappa, Langmuir 14, 880 (1998). 61. K. A. Williams and P. C. Eklund, Chem. Phys. Lett. 320, 352 (2000). 62. G. Stan, V. H. Crespi, M. W. Cole, and M. Boninsegni, J. Low Temp. Phys. 113, 447 (1998). 63. R. S. Lee, H. J. Kim, J. E. Fischer, A. Thess, and R. E. Smalley, Nature 388, 255 (1997). 64. L. Grigorian, K. A. Williams, S. Fang, G. U. Sumanasekera, A. L. Loper, E. C. Dickey, S. J. Pennycook, and P. C. Eklund, Phys. Rev. Lett. 80, 5560 (1998). 65. C. K. W. Adu, G. U. Sumanasekera, B. K. Pradhan, H. E. Romero, and P. C. Eklund, Chem. Phys. Lett. 337, 31 (2001). 66. H. E. Romero, G. U. Sumanasekera, G. D. Mahan, and P. C. Eklund, Phys. Rev. B 65, 205410 (2002). 67. K. Lenghaus, and D. H. Solomon, in “Encyclopedia of Surface and Colloidal Science” (A. Hubard, Ed.), p. 609. Dekker, Santa Barbara, 2002. 68. L. S. Singer, in “Proceedings of the 5th Carbon Conference,” Vol. 2, p. 37. Pergamon, New York, 1961. 69. S. M. Lee, Y. H. Lee,Y. G. Hwang, J. R. Hahn, and H. Kang, Phys. Rev. Lett. 82, 217 (1999). 70. J. M. Nugent, K. S. V. Santhanam, A. Rubio, and P. M. Ajayan, Nano Letters 1, 87 (2001). 71. J. J. Davis, R. J. Coles, and H. A. O. Hill, J. Electroanal. Chem. 440, 279 (1997). 72. P. J. Britto, K. S. V. Santhanam, A. Rubio, J. A. Alonso, and P. M. Ajayan, Adv. Mater. 11, 154 (1999). 73. C.-Y. Liu, A. J. Bard, F. Wudl, I. Weitz, and J. R. Heath, Electrochem. Solid-State Lett. 2, 577 (1999).

601 74. J. N. Barisci, G. G. Wallace, and R. H. Baughman, J. Electrochem. Soc. 147, 4580 (2000). 75. J. K. Campbell, L. Sun, and R. M. Crooks, J. Am. Chem. Soc. 121, 3779 (2000). 76. T. Vo-Dinh, J.-P. Alarie, B. M. Cullum, and G. D. Griffin, Nature Biotechnol. 18, 764 (2000). 77. V. P. Menon and C. R. Martin, Anal. Chem. 67, 1920 (1995). 78. R. J. Chen, Y. Zhang, D. Wang, and H. Dai, J. Am. Chem. Soc. 123, 3838 (2001). 79. M. Shim, N. W. S. Kam, R. J. Chen, Y. Li, and H. Dai, Nano Letters 2, 285 (2002). 80. S. S. Wong, A. T. Woolley, E. Joselevich, C. L. Cheung, and C. M. Lieber, J. Am. Chem. Soc. 120, 8557 (1998). 81. C. V. Nguyen, L. Delzeit, A. M. Cassell, J. Li, J. Han, and M. Meyyappan, Nano Letters 2, 1079 (2002). 82. J. Li, H. Chen, A. M. Cassell, W. Fan, J. E. Koehne, C. V. Nguyen, H. Ng, R. Stevens, J. Han, and M. Meyyappan, in preparation. 83. Y. Cui, Q. Wei, H. Park, and C. M. Lieber, Science 293, 1289 (2001). 84. R. Elghanian, J. J. Storhoff, R. C. Mucic, R. L. Letsinger, and C. A. Mirkin, Science 277, 1078 (1997). 85. S.-J. Park, T. A. Taton, and C. A. Mirkin, Science 295, 1503 (2002). 86. M. Bruchez Jr., M. Moronne, P. Jin, S. Wiess, and A. P. Alivisatos, Science 281, 2031 (1998). 87. W. C. W. Chan and S. Nie, Science 281, 2016 (1998). 88. M. Han, X. Gao, J. Z. Su, and S. Nie, Nature Biotechnol. 19, 631 (2001). 89. S. Uppuluri, D. R. Swanson, L. T. Piehler, J. Li, G. L. Hagnauer, and D. A. Tomalia, Adv. Mater. 12, 796 (2000). 90. A. Ricca, and J. A. Drocco, Chem. Phys. Lett. 362, 217 (2002). 91. S. Heinze, J. Tersoff, R. Martel, V. Derycke, J. Appenzeller, and Ph. Avouris, Phys. Rev. Lett. 89 (10), 106801 (2002).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Carbon Nanotubes: Synthesis by Arc Discharge Technique Yoshinori Ando Meijo University, Tenpaku-ku, Nagoya, Japan

CONTENTS 1. Introduction 2. Preparation of Multiwalled Carbon Nanotubes (MWNTs) by Carbon Arc 3. Preparation of Single-walled Carbon Nanotubes (SWNTs) by Carbon Arc 4. Properties of SWNTs and MWNTs 5. Applications of MWNTs and SWNTs Glossary References

1. INTRODUCTION Carbon nanotubes were serendipitously discovered by Iijima [1] in 1991 in a specimen of the present author [2]. Arc discharge between two graphite rods was carried out to make fullerenes [3], and the nanotubes were observed in the residual of the cathode deposit. The method was similar to Krätschmer’s ac contact arc [4], with the difference that the two graphite rods were 1–2 mm apart and a dc arc was used. It was done by replacing the silicon block of one electrode by a graphite rod in the apparatus used for making ultrafine powder of SiC [5]. On the other hand, Iijima had been interested in observing various kinds of carbon [6, 7] by using high-resolution transmission electron microscopy (HRTEM), since before the discovery of fullerenes [8]. The observation of a fullerene byproduct, the cathode deposit of the dc arc discharge, was also one of them. As a result, he found one-dimensional fibers having quite a new morphology of carbon [1]. They are made of rolled graphene sheets classified into two categories depending on the number of concentric shells: single-walled carbon nanotubes (SWNTs) or multiwalled carbon nanotubes (MWNTs). After the discovery of MWNTs in the cathode deposit of the carbon arc [1], two years later SWNTs were found [9, 10] in sootlike deposits formed in the carbon arc chamber. It should be noticed that both MWNTs and SWNTs were first prepared by dc arc discharge of graphite, though there is a difference in the catalyst requirement [9, 10]. Later on, ISBN: 1-58883-057-8/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

in order to prepare MWNTs, the chemical vapor deposition (CVD) method for large-scale synthesis [11] and thermal decomposition of SiC for aligned MWNT synthesis [12] were applied. However, highly crystallized MWNTs could be produced only by the arc method, because of the high power density under the growth condition. A high yield of SWNTs could be produced by laser ablation [13], whereas the arc discharge [14] and CVD [15] methods were found suitable for their large-scale production. This review focuses only on the arc-discharge method for producing MWNTs and SWNTs. In the case of the arc method, apart from dc/ac, arc current, gas pressure, and so on, the nature of the ambient gas also plays an important role [16, 17]. Especially, the crystallinity of the MWNTs is highly elaborated when evaporating in pure hydrogen gas [18]. The characteristic features of H2 arc MWNTs are high crystallinity and a very thin innermost tube. The physical properties and Raman spectra of H2 arc MWNTs are extremely intrinsic [19–23], and H2 arc MWNTs really resemble quantum fibers. Electrical and mechanical applications of these H2 arc MWNTs are expected.

2. PREPARATION OF MULTIWALLED CARBON NANOTUBES BY CARBON ARC 2.1. The Method To prepare MWNTs, dc arc discharge between two pure graphite rods has been carried out [2, 16, 24, 25] using a vacuum chamber, as shown in Figure 1. Of course, the alignment of the two graphite rods can be vertical [16, 24, 25] or horizontal [17, 26]. After the chamber is evacuated to a vacuum higher than 10−5 torr by using an oil diffusion pump, an inert gas (or appropriate active gas) is introduced into the chamber. Then a dc arc voltage of approximately 25 V is applied between the two graphite rods which are 1–2 mm apart. Within a few minutes, the anode is consumed and carbon soot is formed in the chamber. When an inert gas is used as the ambient gas, fullerenes are included in the soot [3].

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 1: Pages (603–610)

604

Carbon Nanotubes: Synthesis by Arc Discharge Technique

Figure 2. HRTEM micrographs of MWNTs: (a) five-wall, diameter 6.7 nm; (b) two-wall, diameter 5.5 nm; (c) seven-wall, diameter 6.5 nm and the smallest hollow diameter 2.2 nm. Reprinted with permission from [1], S. Iijima, Nature 354, 56 (1991). © 1991, Macmillan Magazines Ltd.

Figure 1. Schematic diagram of arc apparatus producing MWNTs. Reprinted with permission from [25], X. Zhao et al., Bull. Res. Inst. Meijo Univ. 1, 7 (1996). © 1996, Research Institute of Meijo University.

Another part of the evaporated carbon is deposited on the top of the cathode. In the cathode deposit [2, 24] and on the top of the cathode deposit [27], MWNTs and coexisting carbon nanoparticles are observed by using scanning electron microscopy (SEM). HRTEM observation of the cathode deposit dispersed on a TEM microgrid has been carried out. There is a concentric Russian doll structure of rolled graphene sheets with some helical arrangement [1], as shown in Figure 2. The distance between each layer is the same as the {002} lattice spacing of graphite, namely, 0.34 nm. A schematic atomic model of a four-wall MWNT is shown in Figure 3, which has a helical arrangement and the edge is zigzag type [28]. When the cathode deposit is observed by using SEM, the MWNTs are seen as fibers, as shown in Figure 4. Usually, not only fibrous MWNTs but also a number of coexisting carbon nanoparticles are observed [2, 27], as shown in Figure 4. The evaporation of graphite rods has been carried out in He ambient gas, and fibrous MWNTs are observed in the cathode deposit and on the top surface of the cathode deposit as well [27].

Ar, and CH4 within a pressure range of 20–200 torr. At an intermediate pressure of 100 torr, the quantity of prepared MWNTs is found as follows: Ar < He < CH4 [24]. It is known that fullerenes such as C60 cannot be produced in an ambient gas containing hydrogen atoms. On the contrary, thin and long MWNTs with few coexisting nanoparticles can be prepared in CH4 gas more than in an inert gas; as an example, see the SEM micrograph [16] shown in Figure 5. From the mass spectra of the chamber gas before and after evaporation [17], it was confirmed that methane undergoes a thermal decomposition as follows: 2CH4 = C2 H2 + 3H2

2.2. Effect of Ambient Gas The first specimen of MWNTs was prepared by dc arc discharge evaporation in the ambient gas of He [3], although Iijima erroneously wrote Ar in his first report [1]. Since then, not only inert gas such as He or Ar but also active CH4 gas has been tried as the ambient gas of arc discharge [2, 16, 17, 24]. As a result, MWNTs are formed in all gases of He,

Figure 3. Atomic model of four-walled carbon nanotube.

Carbon Nanotubes: Synthesis by Arc Discharge Technique

Figure 4. SEM micrograph of MWNTs prepared at dc arc current of 70 A in He 200 torr.

In this reaction, the calculated mole ratio of the gas before and after evaporation was 1 : 2. This led to the necessity to carry out the arc evaporation of the graphite rod in pure C2 H2 or pure H2 ambient gas. In the case of pure C2 H2 gas, a similar result as in CH4 gas is obtained. On the contrary, arc discharge in pure H2 gas shows quite remarkable improvement [18]. A typical optical photograph of the top of the cathode [18, 29] after H2 arc evaporation is shown in Figure 6, which is obtained after 3 min evaporation by a dc arc current of 50 A in H2 gas at 60 torr. The top surface is composed of three regions: a central black region A, a surrounding silver-gray region B, and a very thin outer region C. When the outer region C is observed using SEM, “carbon roses” are seen [18, 29, 30], as shown in Figure 7. Each petal of a carbon rose is made of very thin graphene sheets [30], that is, nanopetals. In fact, they are composed of several lay-

Figure 5. SEM micrograph of thin, long MWNTs on the top surface. The original specimen was prepared at CH4 20 torr and dc arc current of 30 A. Reprinted with permission from [16], X. Zhao et al., Jpn. J. Appl. Phys. 35, 4451 (1996). © 1996, Institute of Pure and Applied Physics, Japan.

605

Figure 6. Top-view photograph of the deposited cathode after H2 arc discharge. Reprinted with permission from [29], X. Zhao et al., J. Cryst. Growth 198/199, 934 (1999). © 1999, Elsevier Science.

ers of parallel graphene sheets, the thinnest one consisting of only two graphene sheets. The region B is a hard, thick deposit which is observed to be a graphitic layered structure. MWNTs are observed in the central thick black deposit A, as shown in Figure 8a. There are fewer nanoparticles coexisting with the pristine MWNTs in H2 (Fig. 8a) than in He or Ar gas. Therefore, purification is very easily achieved by infrared radiation in air at 500  C for 30 min [20, 31], as shown in Figure 8b. Such purified MWNTs look like a sponge (Fig. 9) [31], which is easy to manipulate with tweezers for various measurements. The size of the sponge is 10 mm2 area × 0.1 mm thickness, and the mass is about 1 mg.

2.3. Structure of H2 Arc MWNTs The structural features of MWNTs produced by H2 arc discharge can be elucidated by HRTEM [18, 21]. As shown in Figure 10a, as-grown MWNTs are highly crystallized with perfect coaxial tube layers [21]. The diameter of the innermost tube A is about 1.2 nm and the outer diameter of

Figure 7. SEM micrograph of “carbon rose.” Reprinted with permission from [18], X. Zhao et al., Carbon 35, 775 (1997). © 1997, Elsevier Science.

606

Carbon Nanotubes: Synthesis by Arc Discharge Technique

Figure 8. SEM micrographs of MWNTs produced by hydrogen arc discharge: (a) as grown, (b) after purification. Reprinted with permission from [20], Y. Ando et al., Trans. Mater. Res. Soc. Jpn. 25, 817 (2000). © 2000, Materials Research Society of Japan.

Figure 10. HRTEM micrographs of MWNT prepared by hydrogen arc discharge: (a) side view, (b) edge-on view. Reprinted with permission from [21], Y. Ando et al., Carbon 39, 569 (2001). © 2001, Elsevier Science.

the MWNT is 13 nm. There are 16 graphene sheets on both sides with a regular spacing of 0.34 nm. The arrow B points out amorphous carbon seen in pristine MWNTs. A crosssectional view of an MWNT (but not of the same one as in Fig. 10a) is shown in Figure 10b. Concentric circles are seen, but with half-fringe discrepancy at each arrowed position. Based on HRTEM observation, the innermost diameter distribution can be obtained [22]. The mean value of the innermost tube is 1.0 nm, whereas that for He arc MWNTs is usually more than 3 nm. The innermost tube observed in H2 arc MWNTs is 0.4 nm (Fig. 11) [32]. The innermost tube is supposed to possess an armchair (3, 3) structure.

SWNTs were found in the specimen produced by arc discharge using a graphite rod including a metal catalyst (Fe), not in the cathode deposit but in the carbon soot. Incorporation of the metal catalyst is essential to produce SWNTs. In Iijima’s case, a mixture of methane (10 torr) and argon (40 torr) was used as the ambient gas [9]. Bethune et al. prepared SWNTs under slightly different conditions using a Co catalyst in helium gas [10]. However, in both cases, the yield of SWNTs was very small.

3. PREPARATION OF SINGLE-WALLED CARBON NANOTUBES BY CARBON ARC

3.2. Mass-Scale Production High-yield SWNTs were first obtained by laser ablation [13], in which unwanted nanoparticles were negligible as compared to SWNTs. However, the overall quantity of SWNTs

3.1. Lab-Scale Production Two years after the discovery of MWNTs, SWNTs were found by Iijima and Ichihashi [9] and by Bethune et al. [10], independently. An HRTEM micrograph of the singleshell tube with diameter 1.37 nm [9] is shown in Figure 12.

Figure 9. SEM micrograph of purified MWNTs sponge. Reprinted with permission from [31], Y. Ando et al., Jpn. J. Appl. Phys. 37, L61 (1998). © 1998, Institute of Pure and Applied Physics, Japan.

Figure 11. HRTEM micrograph of a 4-Å tubule. Reprinted with permission from [32], L.-C. Qin et al., Nature 408, 50 (2000). © 2000, Macmillan Magazines Ltd.

Carbon Nanotubes: Synthesis by Arc Discharge Technique

607

Figure 12. HRTEM micrograph of SWNT. Reprinted with permission from [9], S. Iijima and T. Ichihashi, Nature 363, 603 (1993). © 1993, Macmillan Magazines Ltd.

obtained was still very small. Later on, Journet et al. succeeded in obtaining a large quantity of SWNTs by the arc discharge method [14]. They used a bimetallic catalyst; the best results were obtained with 1 at% Y and 4.2 at% Ni. An SEM micrograph of SWNTs as high density of entangled carbon filaments is shown in Figure 13, which is formed as a collar around the cathode deposit and is not the cathode deposit itself. Such SWNTs are also obtained in other places, that is, rubbery soot condensed on the chamber walls, weblike structures between the cathode and the reactor walls, and a cylindrical deposit at the cathode end.

3.3. Arc Plasma Jet Method Unlike the case of MWNTs, the cathode deposit of arc discharge using a metal-doped anode does not include SWNTs [14]. Hence, an attempt was made to reduce the cathode deposit and increase the quantity of soot generated that contains SWNTs [33]. A pure graphite cathode and metal-doped anode were set at an angle of 30 , and discharge was carried out with the arc flame occurring along the inclined cathode. This new method to produce SWNTs is known as the arc plasma jet (APJ) method [33]. The evaporation rate of the anode and the production rates of the carbon soot and cathode deposit measured in SWNT production experiments under He gas at 500 torr [33] are shown in Figure 14a and b. In the case of APJ, more than 80% of the evaporated anode yields cotton-like soot containing SWNTs and less than 20% appears as cathode deposit. On the contrary, the quantity of deposit in the conventional arc method is larger than the soot at a dc

Figure 13. SEM micrograph of large-scale production of SWNTs by arc discharge. Reprinted with permission from [14], C. Journet et al., Nature 388, 756 (1997). © 1997, Macmillan Magazines Ltd.

Figure 14. Evaporation rate of the anode, production rate of carbon soot, and production rate of deposit measured in SWNT production experiments under He gas of 500 torr: (a) APJ method, (b) conventional arc method. Reprinted with permission from [33], Y. Ando et al., Chem. Phys. Lett. 323, 580 (2000). © 2000, Elsevier Science.

arc current higher than 70 A. Figure 15 shows a characteristic SEM micrograph of cotton-like carbon soot produced by the APJ method [33]. A large number of entangled SWNT bundles and nanoparticles of Ni are seen. The catalyst metal particles are usually covered by thick amorphous carbon, which is hard to remove. Nevertheless efforts to purify asgrown SWNTs are in progress.

4. PROPERTIES OF SWNTs AND MWNTs 4.1. Structures The helical structures of MWNTs and SWNTs are elucidated from their electron diffraction patterns observed by TEM [1, 9]. The atomic structure of individual SWNTs can be determined by scanning tunneling microscopy (STM) [34, 35]. All three types of possible structures predicted theoretically—chiral, zigzag, and armchair—were confirmed experimentally by STM images [34–36]. X-ray diffraction (XRD) measurement of H2 arc MWNTs is performed using a synchrotron radiation source [37]. The

Figure 15. SEM micrograph of carbon soot containing SWNTs which were prepared by the APJ method. Reprinted with permission from [33], Y. Ando et al., Chem. Phys. Lett. 323, 580 (2000). © 2000, Elsevier Science.

608

Carbon Nanotubes: Synthesis by Arc Discharge Technique

powder XRD patterns (wavelength 0.9942 Å) are collected using a powder X-ray diffractometer equipped with an imaging plate. H2 arc MWNT samples are sealed in a quartz capillary and the temperature is controlled in a range from room temperature (RT) to 900 K. XRD profiles are taken as a function of temperature. The XRD profiles of the (002) peak at RT and 873 K are analyzed by Gaussian line shape as shown in Figure 16. Thus, each peak is decomposed into three kinds of peaks: A, B, and C. The first two peaks, A and B, are due to defective MWNTs, while peak C is most probably due to a perfect Russian-doll-type structure.

4.2. Electric Properties Not only the helical structure of SWNTs but also the electronic density of states can be measured by STM [34, 35]. It has been confirmed that SWNTs are metallic or semiconducting depending on the wrapping angle and the tube diameter. The electrical transport measurements of individual SWNTs [38] also coincide with the theoretical predictions [39]. On the other hand, the electric resistance of individual MWNTs has been measured by a micro-manipulator system [40]. Individual H2 arc MWNTs can conduct high electric current [19], 10 × 107 A/cm2 . The temperature dependence of their electric resistance in curve a (Fig. 17) shows that the resistance decreases with increasing temperature. This

Figure 17. Temperature dependence of electric resistance of MWNT. Curve a is obtained from a single MWNT and curve b is obtained from a bundle of MWNTs. The arrows show the direction of the temperature change. Curve a is semiconducting and curve b is metallic. Reprinted with permission from [19], Y. Ando et al., Diamond Relat. Mater. 9, 847 (2000). © 2000, Elsevier Science.

temperature dependence corresponds to the semiconducting character of MWNTs. In curve b, the resistance increases with increasing temperature, namely, metallic character. Field emission measurements of the spongy bulk of H2 arc MWNTs have also been carried out [19]. A practically desirable field emission current density of 1 mA/cm2 is achieved only at an anode voltage of 580 V. The value of the applied voltage is reduced to a field (applied voltage/gap distance) of 1.2 V/m. A light source device, which emits primary color of super-high luminance and fast switching, has been developed by using the spongy bulk of the H2 arc MWNTs [41]. High current density is obtained and enough electron emission is maintained for a long time. A super-high green luminance of 10 × 106 cd/m2 is achieved by the light source device.

4.3. Raman Spectra

Figure 16. X-ray diffraction profile of H2 arc MWNTs. Reprinted with permission from [37], Y. Maniwa et al., Phys. Rev. B 64, 073105 (2001). © 2001, American Institute of Physics.

The Raman spectra of SWNTs show their characteristic features [42, 43]. In the low-frequency region, radial breathing modes (RBMs) are observed at frequency values inversely proportional to the diameter of the SWNTs [42, 43]. The observation of RBM peaks is necessary to confirm the presence of SWNTs. In the high-frequency region, the splitting of G-band modes and its splitting due to the cylindrical structure of SWNTs are observed. In the past decade, extensive Raman experiments have been performed on MWNTs synthesized by arc discharge evaporation in helium gas and by CVD. However, the reported Raman spectra closely resemble that of graphite, and no RBMs have yet been found [44–46] except by Jantoljak et al. [47]. On the other hand, many Raman-active modes [22, 48] are observed for H2 arc MWNTs in the lowfrequency region. The resonance effect of each mode [49] is similar to that of the RBM in an SWNT. It is known that RBMs are expected to have an A1g symmetry, and the wave number of the RBM is proportional to the inverse of the diameter of the SWNT [42, 43]. Figure 18 shows the micro-Raman spectra [22] in the low-frequency region (100–600 cm−1 ) for three specimens of

Carbon Nanotubes: Synthesis by Arc Discharge Technique

609 correspond to the innermost diameters of H2 arc MWNTs: 1.09, 0.58 and 1.23 nm, respectively. SERS spectra A, B, and C in Figure 19b show multiple splitting of the G-band modes in H2 arc MWNTs. These peaks are analyzed by Lorentz fit [23] and decomposed into a broad graphite-like line at approximately 1580(20) cm−1 and sharp lines with FWHM of about 4 cm−1 . The latter originate from the thin innermost tubes of H2 arc MWNTs [23].

4.4. Nanocomposite Structure

Figure 18. Raman spectra of MWNT mat in low-frequency region. Reprinted with permission from [22], X. Zhao et al., Chem. Phys. Lett. 361, 169 (2002). © 2002, Elsevier Science.

H2 arc MWNT mats, (A)–(C). For comparison, the Raman spectra of SWNTs produced by the APJ method [33] and highly oriented pyrolytic graphite (HOPG) are also displayed in Figure 18. In contrast to HOPG, many sharp Raman-active peaks are observed in the Raman spectra of the MWNT mats. The highest frequency peak is observed at 570(3) cm−1 in MWNTs (A), where the full width at halfmaximum (FWHM) is given in parentheses. By observing the polarized Raman spectra from single straight bundles of MWNTs, each Raman-active peak in Figure 18 is assigned an RBM [22]. In the case of MWNTs, the RBM comes from the thin innermost diameter. From the RBM frequency the diameter of the innermost tube can be determined. Thus, the highest frequency peak 570 cm−1 is due to the smallest innermost tube of 0.4 nm [32]. The characteristic features of the Raman spectra are also observed in the high-frequency region for individual H2 arc MWNTs dispersed onto a silver substrate [23]. Surfaceenhanced Raman scattering (SERS) spectra are taken from a single H2 arc MWNT. Three examples of Raman spectra are shown in Figure 19a and b for the low- (100–600 cm−1 ) and high- (1450–1650 cm−1 ) frequency regions, respectively. In Figure 19a, single RBM peaks of different frequencies with sharp FWHM can be seen for each H2 arc MWNT, A, B, and C. The RBM frequencies of 216, 403, and 192 cm−1

Figure 19. Surface-enhanced Raman spectra from individual MWNT: (a) low-frequency region, (b) high-frequency region. Reprinted with permission from [23], X. Zhao et al., Appl. Phys. Lett. 81, 2550 (2002). © 2002, American Institute of Physics.

By removing the cap of the MWNT or SWNT, other materials such as some metals or fullerenes can be inserted into the tube [50–52]. Nanotubes act as molds for the fabrication of metallic wires, some of which are less than 2 nm in diameter [50]. From these viewpoints, new composite materials have been made. As an interesting phenomenon, heat treatment of SWNTs filled with C60 [52] leads to the formation of double-wall carbon nanotubes (DWNTs) [53]. This is a new method for making DWNTs.

5. APPLICATIONS OF MWNTs AND SWNTs Potential applications of carbon nanotubes are predicted for electrical, electronic, mechanical, and chemical applications, such as field emission [54], flat-panel display source, quantum wires [38], X-ray source [54], contact needle of atomic force microscopy, hydrogen storage, gas sensors, capillarity-induced filling of metal [50], and fullerene peapods [52]. Highly crystallized MWNTs and SWNTs produced by arc discharge should have more potential for electronic and mechanical applications. Application-based research of SWNTs and MWNTs is continuously increasing worldwide.

GLOSSARY Arc discharge Evaporation method of graphite electrode using arc discharge. Usually carried out in inert gas. Arc plasma jet (APJ) Method to prepare SWNTs by a special kind of arc discharge. Carbon nanoparticles Nanometer-sized particles made of carbon, which usually has onion-type layered structure. Carbon nanotube Tube composed of carbon atoms only, having a diameter on the order of nanometers. Each tube is made of a rolled graphene sheet. Carbon roses Petal-like carbon made of graphene sheets, which resemble a rose. Chemical vapor deposition (CVD) An alternative method to make fibers or carbon nanotubes. Field emission Electron emission from metal just by applying electric field without heating the metal. Fullerenes Represented by C60 , which is a macromolecule having a cage structure consisting of hexagonal and pentagonal carbon rings. G band Typical Raman peak of graphite observed at 1582 cm−1 , which has E2 symmetry.

610 H2 arc MWNTs MWNTs prepared by dc arc discharge in H2 gas. Multiwalled carbon nanotubes (MWNTs) Consisting of many rolled graphene sheets. Radial breathing mode (RBM) Observed in low-frequency region of Raman spectra of thin tube. The frequency is proportional to the inverse of the tube diameter. Raman spectra Spectra of Raman shift measured by laser acquisition. Single-walled carbon nanotubes (SWNTs) Consisting of only one rolled graphene sheet. X-ray diffraction (XRD) Pattern to clarify crystal structures.

REFERENCES 1. S. Iijima, Nature 354, 56 (1991). 2. Y. Ando and S. Iijima, Jpn. J. Appl. Phys. 32, L107 (1993). 3. Y. Saito, M. Inagaki, H. Shinohara, H. Nagashima, M. Ohkohchi, and Y. Ando, Chem. Phys. Lett. 200, 643 (1992). 4. W. Krätschmer, L. D. Lamb, K. Fostiropoulos, and D. R. Huffman, Nature 347, 354 (1990). 5. Y. Ando and M. Ohkohchi, J. Cryst. Growth 60, 147 (1982). 6. S. Iijima, J. Cryst. Growth 50, 675 (1980). 7. S. Iijima, J. Phys. Chem. 91, 3466 (1987). 8. H. W. Kroto, J. R. Heath, S. C. O’Brien, R. F. Curl, and R. E. Smalley, Nature 318, 162 (1985). 9. S. Iijima and T. Ichihashi, Nature 363, 603 (1993). 10. D. S. Bethune, C. H. Kiang, M. S. de Vries, G. Gorman, R. Savoy, J. Vazquez, and R. Beyers, Nature 363, 605 (1993). 11. M. José-Yacamán, M. Miki-Yoshida, L. Rendón, and J. G. Santiesteban, Appl. Phys. Lett. 62, 202 (1993). 12. M. Kusunoki, T. Suzuki, K. Kaneko, and M. Ito, Philos. Mag. Lett. 79, 153 (1999). 13. A. Thess, R. Lee, P. Nikolaev, H. Dai, P. Petit, J. Robert, C. Xu, Y. H. Lee, S. G. Kim, A. G. Rinzler, D. T. Colbert, G. E. Scuseria, D. Tománek, J. E. Fischer, and R. E. Smalley, Science 273, 483 (1996). 14. C. Journet, W. K. Maser, P. Bernier, A. Loiseau, M. Lamy de la Chapelle, S. Lefrant, P. Deniard, R. Lee, and J. E. Fischer, Nature 388, 756 (1997). 15. H. M. Cheng, F. Li, G. Su, H. Y. Pan, L. L. He, X. Sun, and M. S. Dresselhaus, Appl. Phys. Lett. 72, 3282 (1998). 16. X. Zhao, M. Wang, M. Ohkohchi, and Y. Ando, Jpn. J. Appl. Phys. 35, 4451 (1996). 17. M. Wang, X. Zhao, M. Ohkohchi, and Y. Ando, Fullerene Sci. Technol. 4, 1027 (1996). 18. X. Zhao, M. Ohkohchi, M. Wang, S. Iijima, T. Ichihashi, and Y. Ando, Carbon 35, 775 (1997). 19. Y. Ando, X. Zhao, H. Kataura, Y. Achiba, K. Kaneto, M. Tsuruta, S. Uemura, and S. Iijima, Diamond Relat. Mater. 9, 847 (2000). 20. Y. Ando, X. Zhao, H. Kataura, Y. Achiba, K. Kaneto, S. Uemura, and S. Iijima, Trans. Mater. Res. Soc. Jpn. 25, 817 (2000). 21. Y. Ando, X. Zhao, and H. Shimoyama, Carbon 39, 569 (2001). 22. X. Zhao, Y. Ando, L.-C. Qin, H. Kataura, Y. Maniwa, and R. Saito, Chem. Phys. Lett. 361, 169 (2002). 23. X. Zhao, Y. Ando, L.-C. Qin, H. Kataura, Y. Maniwa, and R. Saito, Appl. Phys. Lett. 81, 2550 (2002).

Carbon Nanotubes: Synthesis by Arc Discharge Technique 24. Y. Ando, Fullerene Sci. Technol. 2, 173 (1994). 25. X. Zhao, M. Wang, M. Ohkohchi, and Y. Ando, Bull. Res. Inst. Meijo Univ. 1, 7 (1996). 26. T. W. Ebbesen and P. M. Ajayan, Nature 358, 220 (1992). 27. Y. Ando, Jpn. J. Appl. Phys. 32, L1342 (1993). 28. R. Saito, G. Dresselhaus, and M. S. Dresselhaus, “Physical Properties of Carbon Nanotubes.” Imperial College Press, 1998. 29. X. Zhao, M. Ohkohchi, and Y. Ando, J. Cryst. Growth 198/199, 934 (1999). 30. Y. Ando, X. Zhao, and M. Ohkohchi, Carbon 35, 153 (1997). 31. Y. Ando, X. Zhao, and M. Ohkohchi, Jpn. J. Appl. Phys. 37, L61 (1998). 32. L.-C. Qin, X. Zhao, K. Hirahara, Y. Miyamoto, Y. Ando, and S. Iijima, Nature 408, 50 (2000). 33. Y. Ando, X. Zhao, K. Hirahara, K. Suenaga, S. Bandow, and S. Iijima, Chem. Phys. Lett. 323, 580 (2000). 34. J. W. G. Wildöer, L. C. Venema, A. G. Rinzler, R. E. Smalley, and C. Dekker, Nature 391, 59 (1998). 35. T. W. Odom, J.-L. Huang, P. Kim, and C. M. Lieber, Nature 391, 62 (1998). 36. S. G. Lemay, J. W. Janssen, M. van den Hout, M. Mooij, M. J. Bronikowski, P. A. Willis, R. E. Smalley, L. P. Kouwenhoven, and C. Dekker, Nature 412, 617 (2001). 37. Y. Maniwa, R. Fujiwara, H. Kira, H. Tou, E. Nishibori, M. Takata, M. Sakata, A. Fujiwara, X. Zhao, S. Iijima, and Y. Ando, Phys. Rev. B 64, 073105 (2001). 38. S. J. Tans, M. H. Devoret, H. Dai, A. Thess, R. E. Smalley, L. J. Geerligs, and C. Dekker, Nature 386, 474 (1997). 39. N. Hamada, A. Sawada, and A. Oshiyama, Phys. Rev. Lett. 68, 1579 (1992). 40. Y. Ando, X. Zhao, H. Shimoyama, G. Sakai, and K. Kaneto, Inorg. Mater. 1, 77 (1999). 41. J. Yotani, S. Uemura, T. Nagasako, H. Kurachi, H. Yamada, T. Esaki, Y. Saito, Y. Ando, X. Zhao, and M. Yumura, J. Vac. Soc. Jpn. 44, 956 (2001) (in Japanese). 42. A. M. Rao, E. Richter, S. Bandow, B. Chase, P. C. Eklund, K. A. Williams, S. Fang, K. R. Subbaswamy, M. Menon, A. Thess, R. E. Smalley, G. Dresselhaus, and M. S. Dresselhaus, Science 275, 187 (1997). 43. S. Bandow, S. Asaka, Y. Saito, A. M. Rao, L. Grigorian, E. Richter, and P. C. Eklund, Phys. Rev. Lett. 80, 3779 (1998). 44. P. C. Eklund, J. M. Holden, and R. A. Jishi, Carbon 33, 959 (1995). 45. W. Li, H. Zhang, C. Wang, Y. Zhang, L. Xu, K. Zhu, and S. Xie, Appl. Phys. Lett. 70, 2684 (1997). 46. A. M. Rao, A. Jorio, M. A. Pimenta, M. S. S. Dantas, R. Saito, G. Dresselhaus, and M. S. Dresselhaus, Phys. Rev. Lett. 84, 1820 (2000). 47. H. Jantoljak, J.-P. Salvetat, L. Forró, and C. Thomsen, Appl. Phys. A 67, 113 (1998). 48. X. Zhao and Y. Ando, Jpn. J. Appl. Phys. 37, 4846 (1998). 49. H. Kataura, Y. Achiba, X. Zhao, and Y. Ando, Mater. Res. Soc. Symp. Proc. 593, 113 (2000). 50. P. M. Ajayan and S. Iijima, Nature 361, 333 (1993). 51. S. C. Tsang, Y. K. Chen, P. J. F. Harris, and M. L. H. Green, Nature 372, 160 (1994). 52. B. W. Smith, M. Monthioux, and D. E. Luzzi, Nature 396, 323 (1998). 53. S. Bandow, M. Takizawa, K. Hirahara, M. Yudasaka, and S. Iijima, Chem. Phys. Lett. 337, 48 (2001). 54. H. Sugie, M. Tanemura, V. Filip, K. Iwata, K. Takahashi, and F. Okuyama, Appl. Phys. Lett. 78, 2578 (2001).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Carbon Nanotube-Based Field Emitters Seong Chu Lim, Hee Jin Jeong, Kay Hyeok An, Dong Jae Bae, Young Hee Lee Sungkyunkwan University, Suwon, Republic of Korea

Young Min Shin Samsung Advanced Institute of Technology, Suwon, Republic of Korea

Young Chul Choi Samsung SDI, Gongse-ri, Kiheung-eup, Youngin, Republic of Korea

CONTENTS 1. Introduction 2. Advantages of Carbon Nanotubes 3. Field Emission Properties of Carbon Nanotubes 4. Applications 5. Conclusions Glossary References

1. INTRODUCTION The source of the most commonly used cathode ray tube (CRT) is an electron, which is thermally ejected from a hot cathode surface. To facilitate the electron ejection, a material with low work function such as a barium oxide compound has been used [1]. The ejected electrons are accelerated through the electric field for rastering and exciting the phosphor screen to display images. The electron emission and acceleration process lead the CRT to consume a lot of power, to operate at high voltage, and to be very bulky. Even though there are some advantages to conventional CRTs such as wide viewing angle, fast response time, unlimited colors, and low price, these CRTs are losing the competition in the display market these days. One major reason is that, in nature, the large dimension and high operational voltage do not allow the CRT to be portable. Portability, however, ISBN: 1-58883-057-8/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

is one of the essential features that a display must satisfy in the information age to permit accessibility to the intellectual resources at any time and any place. Therefore, the demand for a flat-panel display, which is slim, light, and operating at low voltage and at low power consumption, has become even stronger. The annual revenue of the fast-growing global market of flat-panel displays (FPDs) is supposed to reach 80 billion dollars in 2010 [2]. Such tremendous growth of the FPD market has been empowered by the wide distribution of the laptop computer and technical developments in the area of FPDs. In the FPD market, there are several types of displays: liquid crystal display (LCD), plasma display panel (PDP), organic electroluminescence display (OELD), and field emission display (FED). At present, the LCD, dominating about 80% of the market, is very light and slim and operates at low voltage, but contains problems such as limited colors, narrow viewing angle, narrow operation voltage, and high power consumption. LCDs should overcome these problems in order to maintain their position in the market. At the same time, other types of FPDs have to show technical superiority to survive in the market. For instance, PDPs can be readily scaled up to over 60 in., but consume heavy power for plasma generation and still do not show the wide color range that CRTs can display. The OELD, which is flexible and can be fit into almost any shape, operates successfully only for a few hundred hours and degrades fast and furthermore shows low thermal stability. Among these displays, the FED has captured only a small portion of the market. The CRT and FED have many internal structural features in common, except for the electron

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 1: Pages (611–624)

Carbon Nanotube-Based Field Emitters

612 ejection gun. In FEDs, the electrons are ejected by the strong local electric field from many sharp tips. The electric field can be switched on and off swiftly, contributing to the fast response time. FEDs can also provide unlimited colors, a wide viewing angle, and a wide operating temperature, but with low power consumption and low applied voltage. In addition to the above advantages, there are millions of tips in a pixel. Therefore, if one of the emitters stops functioning, another emitter contributes to the field emission. All theses properties of the FED make it an almost perfect display. However, many technical issues have yet to be resolved to meet the requirements for commercialization. The field emission phenomenon was not clearly understood until the adoption of quantum mechanics, because field emission is different from thermionic emission, where sufficient kinetic energy should be given to the electrons in order to surmount the potential barrier at the interface between the vacuum and the surface of the emitter [3–5]. Field emission is a tunneling phenomenon, which cannot be described by classical mechanics. In the presence of a strong electric field, electrons can tunnel through the potential barrier since the barrier becomes dramatically narrow under such a strong field, as shown in Figure 1. Therefore, the field emission is expected to vary with the strength of the applied field and the work function of the emitter materials. Some efforts neglecting this tunneling effect were made in the 20th century, but turned out to be unsuccessful for characterizing the relationship among the above parameters [6], until Fowler and Nordheim formulated the relationship based on quantum mechanics. In their formula, the emission currents can be described as follows: J = AV 2 exp−B3/2 /V 

(1)

where  is the work function of the tip, A and B are constants, and  is a field enhancement factor depending on the geometrical structure [7]. Metal tips were used to study the field emission in the early days. The metal tips are prepared by electrochemical etching [8, 9] and lead to an intensification of the electric field. Among various metals, refractory metals, such as tungsten, and molybdenum are utilized for field emitters

because of their high melting temperature and mechanical robustness to high electrical stress [10, 11]. An emitter coated with a thin film of low work function, such as Cs, can render lower operation voltage, but the chemically reactive Cs becomes easily oxidized, decreasing the current density [12]. However, the high work function and low field enhancement factor of metal emitters of about 10 require an ultrahigh vacuum environment and appreciable emission voltage in the range of 3–7 × 109 V/m, which would eventually result in the ionization of the residual gases, physical sputtering of the emitter tip, and fluctuation of the emission currents. For these reasons, coating the emitter tips with wide bandgap materials (WBGMs) exhibiting negative electron affinity (NEA) such as diamond, SiC, c-BN, LiF, and CaF have been attempted [13–16]. Even though the high mechanical hardness and chemical inertness of the WBGM are very beneficial for field emitters, concurrently a wide bandgap becomes a prime obstacle that hinders large electron emission and lowers electrical conductivity unless the coating thickness is controlled with high precision. Another attempt to circumvent the high turn-on voltage of the metal emitters is miniaturization of the device by microfabrication techniques from the semiconductor industry. The pioneering achievement was first made by Spindt in 1968 [17]. He fabricated a gate-structured cathode consisting of a cathode and a gate electrode. For the cathode, thermal evaporation of molybdenum was conducted through the microcavity, which had been preformed on a layered structure of gate/insulator/cathode. The gap between the gate and the conical-shaped emitter is only about 0.5 m and the cathode is a matrix-addressed Mo emitter array. The proximity of the gate electrode and the high population of emitters, 2 5 × 107 emitters/cm2 , contribute to the large available emission current of 1000 A/cm2 at a bias voltage of 100 V [18]. Despite the fact that such prominent advancements enable us to operate FEDs at low voltage with large emission current, the operation voltage is still too high to be met by the current integrated circuit (IC) chip driving circuit and furthermore thin-film deposition for making the emitter array is not an effective approach for lowering the manufacturing cost and scale-up of the display.

2. ADVANTAGES OF CARBON NANOTUBES Vacuum level

(a)

f

f Eff

Fermi level

(c) (b) Figure 1. Schematic diagram of potential barrier (a) on the emitter surface, (b) under an applied electric field, and (c) with the effect of image potential.  and eff are the work functions on the emitter surface without and with the combination of the applied field and image charge, respectively.

As briefly mentioned in the previous section, field emitters should be robust in order to sustain a high electric field and large emission current. Therefore, the emitter material should have a high melting point, high thermal and electrical conductivity, and high physical stiffness. In addition, low work function is one of the important factors in making a field emission device operable at low voltage. High mechanical stiffness would help emitters to minimize the geometric deformation from ion bombardments; chemical inertness must be a significant advantage making emitters immune to the residual gases. However, it is not easy to find a material satisfying all these requirements. Nonetheless, carbon nanotubes (CNTs) are a potential material for field emitters in many aspects. The CNT first reported by Iijima et al. is a seamless tubular structure formed by rolling

Carbon Nanotube-Based Field Emitters

up and matching a graphene sheet [19]. The cylindrical CNT is composed of single layers or concentric multilayers of the graphene sheet. The geometric structure of the CNTs makes them very effective field emitters when they are placed inside the electric field due to the large aspect ratio coming from a small diameter ranging from one to several nanometers for single-walled CNTs (SWCNTs) and a few tens of nanometers for multiwalled CNTs (MWCNTs) and a length of several micrometers. CNTs, sp2 -bonded carbon networks, are supposed to have a high Young’s modulus with low mass density. However, the technical difficulties involved in the manipulation of the nanotubes mainly come from the small diameter and the tendency of the CNTs to form bundles, which makes direct measurement of the mechanical properties a challenging subject. As a consequence, only a few measurements on the Young’s modulus have been reported. The studies have been carried out under or inside a microscope. One approach is to correlate it with the amplitude of a standalone oscillating CNT whose bottom is fixed at the substrate [20, 21]. The detailed characteristics of the CNT motions observed inside a transmission electron microscope (TEM) are analytically studied after printing the micrographs. Measurements made on many SWCNTs reveal that the Young’s modulus is 1.25 TPa, which is greater than the modulus along the basal plane of the graphite. Another approach is to use atomic force microscopy (AFM) to bend an SWCNT bridging a pore of an ultrafiltration alumina membrane using an AFM tip [22, 23]. But, regardless of the experimental methods, the Young’s modulus of the SWCNT is about 1 TPa. Also, in the case of the multiwalled CNT with a diameter larger than 2 nm, the Young’s modulus is on the same order as that of the graphite. This suggests that a large diameter does not induce stiffening [22]. CNTs, quasi-one-dimensional structures with the unique electronic structure of a graphene sheet [24], exhibit very high electric and thermal conductivity [25]. The electronic properties are highly responsible for the geometric structures [26]. Although graphite is a semimetal, the CNT can be a metal or a semiconductor, depending on the diameter and chirality of the tube, with various energy bandgaps usually below 1 eV [26–29]. The influence of these versatile electronic properties on the geometrical structures originates from the unique mirror-image band structure of the graphene sheet and the quantization of wave vectors along the circumferential direction. The periodic boundary condition along the circumference allows only a certain set of k states of the planar graphite sheet to be available, depending on the tube diameter and chiral angle. In general, the chiral angle of each single-walled CNT can be designated by an integer pair (n m), where n and m are indices of two basis vectors of the graphene sheet. Any choice of (n m) classifies CNTs into two categories, depending on the metallicity. All (n n) tubes have a band degeneracy, which is induced by crossing  and  ∗ states at k = 2/3/a0 . Therefore, these tubes should be metallic. The (n m) tubes satisfying n − m = 3i, where i is a nonzero integer, should also be metallic. All others are large-gap semiconductors. For semiconducting tubes, the bandgap decreases inversely proportional to the diameter. The above band-folding scheme is valid for tubes with a large diameter. However, for the

613 (n 0) tube with a small diameter, this scheme is not adequate because the strain energy dramatically increases as the degree of curvature becomes larger. This curvature effect or strain opens up the small bandgap, whereas the tube is practically metallic at room temperature due to the small bandgap. The curvature effects become more prominent, at which point the tube is under mechanical stress because of an enhanced mixing of the  and  states [30]. To construct the closure of the graphene sheet, the incorporation of topological defects, pentagons, is essential. It results that, many different structures can exist at the cap. The variety in cap structures stems from the variable number of pentagons involved and their relative positions with respect to the hexagons [31]. Due to this property, it is possible that there exist several different cap structures at the same tube diameter and the consequence is remarkable differences in the electronic structures at the cap. The electronic structures in correlation with the chirality and the diameter of the tube have been examined by scanning tunneling spectroscopy (STS) [32–34]. Raman spectroscopy has proved to be an easily accessible, reliable, and nondestructive method to probe the chirality and diameter of the individual tube [35]. However, these approaches are not applicable for the investigation of the cap geometry. The structure of the cap has been investigated using field emission microscopy (FEM) [36–38]. A pentagon ring of carbon is clearly seen. In addition, since the pentagon is a defect, the electronic structure at the cap should be different from that on the wall. As mentioned previously, the diversity in structure results in various local electronic structures at the cap [39]. Understanding the local density of states (LDOS) originating from these defects has huge significance for discovering the field emission properties since, in some topological configurations, the localized states can emerge very near the Fermi level at which the field emission take places [40]. At pentagonal sites, the field emission is observed to be stronger and gas adsorption is more active than at other hexagonal areas [36]. Dean et al. have proposed that the nonmetallic behavior observed at the high voltage region might be related to the defect states [41]. Unfortunately, complete comprehension of the emission behavior correlated with the cap geometry is not well established. This will require systematic studies in the near future. When the carbon nanotube was first discovered in 1991, it was a soot synthesized by arc discharge [19]. Usually, this growth method together with laser ablation is known to produce highly crystalline single-walled and multiwalled nanotubes [42]. The generated CNTs are entangled in bundles, with amorphous carbon, graphitic particles, and other forms of carbons as by-products [43, 44]. The entanglement of the CNTs with amorphous carbon layer and catalyst particles along the tube leads to the laborious purification process consisting of sonication, air oxidation, and acid treatments [44]. After the purification process, only a small amount of CNT with high quality is obtained. However, in the two growth methods, since the synthesis takes places over a very local area at extremely high temperature, the scale-up of the growth system is practically impossible. Also, without the removal of amorphous carbon, metal particles, and graphitic particles, the application of pristine

614 CNTs as field emitters causes the emission properties to be uncontrollable. However, if we understand the growth mechanism, where catalytic decomposition of hydrocarbon is involved, the control of diameter, length, directional growth, and scale-up of growth system, which are impossible in arc discharge and laser ablation, would be achieved in a chemical vapor deposition (CVD) system. In CVD, all the growth parameters from arc discharge and laser ablation are precisely controlled and slowed down over a prolonged growth period. First, in CVD, the heating of catalyst metals is carried out after the CVD chamber is filled with carrier gases and hydrocarbon sources up to a certain pressure. Usually, catalyst metals such as Fe, Ni, and Co are deposited on a Si substrate or porous support such as alumina and zeolite to hold the nanoparticles inside the channels [45–47]. During the temperature elevation, the control of diffusion and agglomeration of metal particles due to thermal energy is very critical to monitor the tube diameter and density [48]. The growth temperature varies with the gas species and catalyst metals, but ranges between 600 and 1000  C. As carbon sources, C2 H2 , CH4 , CO, ethylene, and even alcohol are used [48–51]. Recently, the growth of CNTs aligned in the perpendicular direction to the substrates has successfully been accomplished by many groups [52–55]. CVD tubes grown for 20–30 min are, on average, about 10–50 m long and 50 nm wide. The huge progress in CNT growth using the CVD method has led to a reduction in the purification steps and resulted in vertically aligned tubes. Still, the growth of CNTs below 550  C over a large area in an economic way requires further investigation for their practical application.

3. FIELD EMISSION PROPERTIES OF CARBON NANOTUBES Before de Heer and his group studied the field emission using highly oriented CNTs [56], the growth of the CNTs using the CVD method had been demonstrated [57, 58]. Although CVD growth promises several advantages for applications as field emitters compared to arc discharge and laser ablation synthesis, the realization of the field emitter was postponed due to poor alignment of the CVD-grown tubes. However, later, more adaptable CNTs with better orientation were grown on mesoporous media by the CVD method by Li et al. [59]. The above pioneering research has ignited worldwide research on CNTs as field emitters. MWCNTs can be grown in a preferential direction using catalyst particles deposited on Si substrates using a magnetron sputter. However, the complexity of the growth parameters such as temperature, flow rate, catalyst metal, gas species, pressure, and grain size of catalyst metal results in wide diversities in tube structure, diameter, density, length, and crystal quality. Characterizations of electron field emission from CNTs, which are placed inside a strong electric field, have been carried out in many different ways. We know from (1) that the emission currents are a function of work function and field enhancement factor. However, the work function, in particular, at the emitting site, is not the same as that at the bulk and is very sensitive to local electronic structures. Many groups have demonstrated from FEM studies

Carbon Nanotube-Based Field Emitters

that field emission occurred through the cap, and the cap structures observed on the phosphor screen are different from tube to tube. This implies that emission currents might change dramatically depending on the cap geometry, but systematic studies have not been carried out due to the various geometries. Some theoretical calculations on the effects of cap structures have been conducted only on certain types of tubes [39, 60]. Kim et al. proposed that the localized states of the pentagon defect contribute to the field emission only at strong electric field and the degree of charge accumulation at the cap shows stronger dependence on the tip sharpness rather than the local atomic geometry [61]. From the aspect of electronic structures, CNTs are different from previous metal emitters. Nevertheless, the Fowler– Nordheim (F–N) characteristic curve obtained from CNTs resembles that of metal emitters. Due to the advantage of the tip geometry, the CNT has a lower turn-on voltage, between 1 and 2 V/m. The SWCNT has a lower turnon voltage than the MWCNT. The evaluated field enhancement factor from the F–N plot based on a work function of 5.0 eV is about 1000–3000 and 2500–10,000 for MWCNTs and SWCNTs, respectively [62]. The higher field enhancement factor for SWCNTs is attributed to the smaller diameter. Bonard et al. have also tested the emission stability of both tubes for long-term operation. The emission currents from the SWCNTs degrade about 10 times faster than those from the MWCNTs. This degradation could be explained possibly by the structural weakness of the SWCNT edge to the ion bombardment of residual gases [62]. The higher stability of the MWCNTs could also be explained by the tube edge interactions, where the tube edges are stabilized by lip–lip interactions via spot-welded adatoms [63]. The directional dependence of the emission properties was also studied. The field emission has been examined as a function of the angle of CNTs with respect to the substrate [64]. Contrary to the general belief that the low turn-on voltage of CNTs is attributed to the high aspect ratio, where the aspect ratio is presumed to be even greater with the CNTs aligned in parallel with the electric field, the CNTs parallel to the substrate eject electrons at a lower turn-on voltage. It has been proposed that numerous defect sites along the tube wall contribute to easier electron emission. Another possible explanation is the catalyst existing at the end of the tube, which degrades the field emission currents. With regard to the fabrication of flat-panel displays based on the CNTs, depositing the emitters with an optimum number of CNTs per unit area is crucial, since the poor emission properties observed from the CNT films grown by CVD are speculated to originate from high tube density due to the shielding effect of the electric fields. Several approaches have been attempted to regulate the CNT emitter density on the substrates. One approach is to regulate the grain size and density of the catalytic particles during thin-film deposition using typically RF (radio frequency) magnetron sputter, which enables one to control the grain size and density by the plasma power, substrate temperature, and deposition time [65]. Another way to deposit catalyst particles is to print using a stamp. This is called micro-contact printing (CP), which uses a patterned elastomeric stamp to transfer a self-assembling material (SAM)–forming ink. The stamp is

Carbon Nanotube-Based Field Emitters

made by casting a polydimethylsiloxane (PDMS) elastomer on a master with a designed pattern prepared by MEMS (microelelctromechanic system) techniques. A catalystcontaining solution is used for the ink. [66–68]. The transfer of catalyst particles is made by compressing a stamp against the substrate after dipping the stamp into the ink or placing a droplet on the stamp. The primary advantages of CP are its simplicity and scalability. One can readily change the concentration of the catalyst ink and generate reproducibly the same pattern over a large area. Bonard et al. have examined the correlation of concentration of catalyst ink to CNT growth. In their study, up to a certain concentration, the alignment of CVD-grown CNT becomes better along the vertical direction and then CNTs do not grow above the optimum concentration [69]. Another way to control the emitter density is to use a template, which arranges the metal particles along geometric channels. The template is composed of arrays of pores, whose diameter and pore–pore distance are well defined through the anodizing process [70]. During the anodization of high-purity aluminum, we can change the pore diameter, pore–pore distance, and pore depth by varying the voltage, temperature, reaction time, and proper choice of chemicals. Using an anodized aluminum oxide (AAO) template, monodispersed pores with a high regularity over a large area have been obtained by several groups [71–74]. One of the benefits of using the AAO template is that we can precisely control the ratio of height to diameter, which is an essential factor for regulating field emission and minimizing screening effects, as shown in Figure 2a and b. The detailed significance of screening effects due to the neighboring tubes is well documented in the literature [75]. The most efficient ratio of tube height to tube–tube distance is known to be

Figure 2. SEM images of CVD-grown CNTs using AAO template. The average tube diameter and tube–tube distance are (a) 38 and 104 nm and (b) 19 and 65 nm, respectively. The variation of field enhancement factors in the function of tube height. The intertube distance is (a) 104 nm and (b) 65 nm. Reprinted with permission from [75], J. S. Suh et al., Appl. Phys. Lett. 80, 2392 (2002). © 2002, American Institute of Physics.

615 about unity. The field enhancement factor at this configuration ranges from 2500 to 3500, as shown in Figure 2c and d. All the approaches mentioned above manage the asgrown tube density by arranging catalyst particles before the tube growth. The following screen-printing method is used to handle the ready-grown nanotubes. In the screen-printing approach, the as-grown or purified SWCNTs are dispersed in isopropyl alcohol and mixed with an organic binder such as nitrocellulose and ethlycellulose. The mixture with homogeneously dispersed CNTs is pasted on the prepatterned cathode electrodes [76, 77]. However, the number of emitters per unit area is also very sensitive to the pasting techniques. Another parameter as significant as the tube density in achieving emission uniformity is the film roughness. Initially, the emission pattern of the CVD-grown tube is very poor. The reason is that tubes protruding out from their neighborhood dominate the emission currents. Improvement in the uniformity of the emission pattern requires removal of those protruded tubes. The removal process is known as high-voltage annealing, since the tube is gradually burned away by resistive heating at large emission currents. While the duration of the annealing process is strongly dependent on the applied voltage and emission currents, it usually lasts longer than 24 h. The field emission characteristics of CNTs at room temperature exhibit current suppression at a high-field region, that is, a decrease in the slope of the F–N curve, which is different from those of the typical metal tips [18, 78]. The turn-on voltage of the CNTs is as low as a few volts per micrometer, lower than that of the typical metal tips by a factor of 100 [79]. Several mechanisms such as space charge, screening effects, metal particles at the cap, and gas adsorbates have been proposed in order to explain the current suppression at high field. Lim et al. have checked various possible causes and have concluded from a slope change during only rise sweep, not fall sweep, in their I–V curves that the release of gas adsorbates at large emission currents are the origin of current suppression [80]. Several groups have observed thermal motions of the gas adsorbates at the cap using FEM [36, 37, 81, 82]. The cause of thermal fluctuation in FEM images is local heating for large emission currents. At the stage that gas adsorbates initially reside on the surface of cap, the emission currents are stable and the FEM image is motionless. As the increasing emission currents start to heat up the cap resistively, local heating, the emission currents fluctuate as do the FEM images. At even higher bias voltage, a complete desorption of gas adsorbates on the cap is observed with a significant current decrease. At desorption, the temperature is reported to be around 1000 K [41, 82]. The cyclic repetition of voltage sweep generates consistent I–V characteristics and FEM micrographs. As a consequence, local resonant states present on the surface of the cap temporarily disappear and reappear in parallel with an increase and decrease in the emission currents. Field emission enhanced through the resonant states becomes suppressed in the absence of adsorbates. For the purpose of clarifying the effects of each gas species, Dean et al. intentionally exposed CNTs to different gases and studied the variation of the emission properties.

Carbon Nanotube-Based Field Emitters

616 Saito et al. have investigated the variation of the emission properties upon adsorption and desorption of gas adsorbates [36]. They observed the migration of gas adsorbates on the cap of the CNT and found that the pentagon site is favorable for electron emission because of not only geometrical protrusion but also the presence of strongly localized states. It is also found that the gas adsorbates preferred to reside on the pentagon sites. To remove the gas adsorbates, heating to about 1300 K is required. This temperature is somewhat higher than the desorption temperature witnessed by Dean et al. After heating, the stepwise decrease in emission currents was observed as the gas adsorbates came off the cap, as shown in Figure 3. Among the species introduced into a vacuum system, water vapor is the only species enhancing emission currents. Water molecules are removed from the surface of the CNT at about 900 K and degrade the emission. Lim et al. studied the effects of different gases on the field emission properties as a function of exposure times. In their experiment, oxygen gas reduces the emission currents significantly through the oxidative etching process. The erosive etching of oxygen gases destroys the sp2 network and depletes the 2p– electron states, which contribute to the field emission currents [83, 84]. The destruction of tube structure and reduction in electronic states near the Fermi level result in high turnon voltage and low emission currents. After oxidation, the contents of the sp3 -bonded structures increase. The reverse transition from sp3 to sp2 is partially achieved as the MWCNTs were annealed above 750  C. Theoretical calculations on the electronic structures in the presence of gas adsorbates and an adsorption mechanism have provided a clearer understanding of the effects of field emission upon gas species [85–87]. The influence of water molecules on field emission has been studied. Water molecules, whether they maintain the molecular structure or break into ions on the cap, always increase the emission currents. Maiti and his colleagues have assessed the influence

of water molecules that exist on the cap. In their study, they chose a closed tube and water molecules to be physisorbed on the cap. Polar water molecules construct complexes stable up to several hundred degrees, with a corresponding decrease in the ionization potential and stability of the highest occupied molecular orbital (HOMO). The resulting electronic structure becomes more favorable to field emission. The preservation of the molecular structure of water neglects the possibility that water molecules can break into H and OH ions. The residence of water fragments such as H and OH ions on the open edge of armchair and zigzag tubes is theoretically well simulated by Hwang and Lee [88]. The resulting electronic states with the Fermi level shifted toward the conduction band by 0.14–0.17 eV enhance the field emission for both armchair and zigzag tubes, where the effect is more pronounced in the zigzag tube edge because of the introduction of additional states near the Fermi level. Kim et al. have explored the adsorption of hydrogen, nitrogen, and oxygen molecules at the open edge of CNTs. The shift of the Fermi level upon each gas is presented in Table 1. In their study, oxygen gases with stronger electronegativity suppress the emission currents by creating a dipole layer toward the inner tube at the edge of the tip. In the case of hydrogen gases, the weaker electronegativity in the hydrogen atom than in the carbon atom is supposed to promote the emission currents. However, experimental results show that the hydrogen gases do not affect the field emission appreciably. Such a result may be attributed to the absence of resonant states, even though adsorption of hydrogen gases shifts the Fermi level toward the conduction band. Nitrogen gases are supposed to be less effective than other gases, since the binding state to the edge is relatively weak and the dipole strength is weak due to a negligible charge transfer [85]. As CNTs are exposed to oxygen gases, the situation becomes more complicated due to the progressive change in the cap geometry. The etching process must be more active around the cap area than the wall due to the pentagonal defect sites. Park et al. have pointed out that at the stage of molecular adsorption emission currents are enhanced by the electronic structures and local field intensification at the adsorption sites. In the case of atomic adsorption, the latter effect is more dominating, whereas molecular oxygen generates more currents from the new energy states. However, for prolonged exposure, the cap structure becomes open and the edge can be terminated with atomic oxygen, resulting in a decline in the emission currents [87]. Strengthening of the electric field due to sharp geometry occurs at the cap. The amplified field narrows the potential barrier so that electrons tunnel more easily into a vacuum. Therefore, the field emission electron strongly reflects the Table 1. Change in the Fermi level due to gas adsorption at various sites.

Figure 3. FEM images showing progressive desorption of gas adsorbates with the change of emission currents. The first pattern was taken 60 s after the beginning of heating. The following patterns were taken (b) 90 s, (c) 120 s, and (d) 130 s after heating, respectively. (e) Heating was turned off. Reprinted with permission from [36], K. Hata et al., Surf. Sci. 490, 296 (2001). © 2001, Elsevier Science.

Armchair, top O2 H2 N2

−0 31 0 08 Note:

®

Armchair, seat −0 50

®

indicates the resonant states. Units are in eV.

Zigzag, top −0 12® 0 15 0 04

Carbon Nanotube-Based Field Emitters

local density of states (DOS) at the emitting site. As mentioned earlier, the integration of pentagon rings into the hexagon network creates significant modification in the local DOS. In addition, it is shown by scanning tunneling spectroscopy (STS) that the number of pentagon rings and their relative positions to the hexagon generates a wide variation in electronic structures [39]. The theoretical evaluation of a (5 5) metallic nanotube, whose cap is a hemisphere of C60 , reveals the existence of localized states about 0.8 eV above the Fermi level. Therefore, this means that the local DOSs at the cap strongly influence the field emission behavior. The energy distribution of electrons emitted from the MWCNT was measured by Fransen et al. [37]. A distinguishable difference between the metal tip and the MWCNT tip is the shift in the peak position upon extraction voltage. This opens up two possibilities. One is that the nanotube is semiconducting and the other is that the insulation layer between the W tip and the nanotube induces a voltage drop. It is not clear yet which mechanism dominates. The reason that we have studied various emitter materials and examined the environmental stabilities is to securely preserve the tip from degradation. The degradation process either breaks out abruptly or progressively develops over a long period. The degradation of the tips can be divided into two types: chemical contamination and physical destruction of the tip geometry. Since current suppression, triggered by the chemical adsorption of gas molecules, was addressed previously, here, the cause of the geometrical damage due to large emission currents and ion bombardments will be discussed. Usually, in metal emitters, vacuum breakdown results from the abrupt increase in pressure between the electrodes, which is responsible for the introduction of materials from the cathode and anode. When the tip is involved in the vacuum breakdown, the sudden increase in pressure between the two electrodes is attributed to large emission currents, which melt and evaporate the tips. This catastrophic vacuum breakdown forms a conducting path between the electrodes, leaving the devices malfunctional. Another deformation of the tip is triggered by the bombardment of ionized residual gases. The gases existing between the gap become ionized by the emitted electrons and the ionized gas molecules are accelerated toward the tip. The sputtering process continues progressively during the operation of the FED and can end up with a vacuum breakdown in the long run. Although CNTs show exceptional conductivity and physical stiffness, they develop slow variations in their structures unless the driving voltage is less than the sputtering threshold. The circumstance in which CNTs are exposed to energetic ion particles has been well simulated inside a field ion microscope (FIM) by Hata et al., who found that CNTs are damaged at an electric field above 10 V/nm [89]. The study also reports that the field evaporation of MWCNTs and SWCNTs produces carbon clusters. The major product constituting the cluster is C20 , which is the smallest cage structure composed of hexagons and pentagons. For the field-induced unraveling process, it was shown by Lee et al. from theoretical simulation that the zigzag tube is more favorable for the unraveling of carbon atoms and the field strength is similar to the result from the previous experiment [90]. Dean et al. also witnessed a slow disintegration of carbons from the tip [91]. In their field

617 emission microscopy (FEM) studies, an individual SWCNT is able to withstand emission currents up to 1–2 A, corresponding to 109 A/cm2 . Above this current level, CNTs begin to show local heating evidenced by blurred ring images spinning continuously, as shown in Figure 4. The rotating ring on an FEM image is evidence that the CNT is being eroded away from thermal evaporation at about 1600 K. To enhance environmental stability and durability to ion bombardment, chemical contamination, and local heating, various materials have been used as coatings at the apex of the metal tip [92–94]. However, in the case of CNTs, only a couple of studies have been carried out. The criteria for the choice of coating material include chemical inertness, mechanical hardness, and thermal conductivity. Diamondlike carbons (DLCs), which consist of sp2 and sp3 carbon, are good electric and thermal conductors and are physically hard as well due to the tetrahedral sp3 bond. It is reported that deposition of 20-nm-thick DLC reduces the turn-on voltage from 2.6 to 1.5 V/m [95]. However, such advantageous effects disappear as the coating thickness increases to 50 nm. Another material studied is MgO deposited on CVD-grown vertically aligned MWCNTs. The role of the MgO layer is to lower the turn-on voltage by resonantly enhancing tunneling inside the MgO layer [96]. Doping is another way to enhance the field emission properties by altering the electronic structure. The most commonly used doping materials are boron and nitrogen. Several studies report remarkable changes in the structure and growth behavior. However, the correlation of those changes with field emission effects has not been reported. Only one group has reported that Cs-doped SWCNT film shows a remarkable decrease in work function and turn-on voltage, which is similar to the metal tip [97]. Although the encouraging reduction in work function of the CNT has been achieved, Cs deposition is not applicable to the usual vacuum environment of FED due to degradation [98].

Figure 4. (a) I–V behavior of SWCNT beyond F–N region. (b) Evolution of the FEM image with an increase in emission currents. The image at 1050 V rotates with a frequency over 5 Hz. Reprinted with permission from [81], K. A. Dean and B. R. Chalamala, Appl. Phys. Lett. 76, 375 (2000). © 2000, American Institute of Physics.

618

4. APPLICATIONS Several applications such as FED, electron multiplier, and X-ray gun have been demonstrated. Among these applications, the FED will have the largest impact on our society by revolutionizing the display. The fast-growing FPD market and the superior properties of FED have led companies such as Motorola and Candescent to invest in research and development of microtip FEDs. In spite of well-established fabrication technologies, the high production expense, tip degradation, and difficulty in scale-up make Mo tip-based FED uncompetitive with other types of FPDs in the market. The array of the microtip, called the Spindt cathode, was made by evaporating an emitter material through the gate holes. As the size of the cathode becomes larger, the nonuniformity and production cost go up rapidly. To overcome the problems arising from scalability and fabrication cost, several concepts for low-cost and scalable cathodes were proposed by different industries such as Canon, Matsushita, Hitachi, Samsung, and ISE Electronics. Among these enterprises, Samsung and ISE Electronics are using CNTs as field emitters. The first FED taking advantage of CNTs, demonstrated by Wang et al., has a different form of cathode array [99]. For this cathode construction, a composite of nanotube and epoxy is squeezed into microchannels on a glass substrate since selectively CVD-grown CNT emitters are not accessible to a large area and sodalime glass substrate. The growth temperature inside the CVD chamber is between 650 and 1000  C. The emitter density of the as-prepared cathode in the FED is about two tubes/m2 . A panel, 1 × 1 cm2 in diode structure, successfully operates inside a vacuum system at an applied voltage of 300 V. A fully sealed FED, 4.5 in. in diagonal, with a high brightness of 1800 cd/cm2 has been demonstrated by Choi et al. [76]. To achieve an operable pressure, a nonevaporable getter was used. In the display, purified SWCNT was mixed with an organic binder. This composite was squeezed into a premade substrate through the metal mesh. In this method, the uniform dispersion and vertical alignment of the CNTs must be satisfied in order to meet the practical performance requirements. To meet the requirements, after pasting the composite, a pretreatment such as rubbing and electric field conditioning was carried out to align the embedded CNTs in the vertical direction, followed by burning of the organic binder [100]. This fully sealed diode-type display (200-m glass spacer) turns on at less than 1 V/m and emits 1.5 mA at 3 V/m. The matrix-addressed CNT-FED proves that the cathode array produced by mechanical printing of the mixture of the CNT and epoxy/organic binder on the electrode-patterned sodalime glass is scalable, enabling one to enlarge the panel without limitation in size and with low production cost. The above two prototype FEDs are configured in a diode structure, consisting of a cathode and anode. Under this configuration, the ejection of electrons from the emitter surface is achieved at high voltage since the vacuum gap is about 200 m. The diode structure is not able to tune the colors and brightness once the electrons leave the surface of the emitter. Imposing an additional gate electrode at a typical distance of about 1 m above the emitter makes for

Carbon Nanotube-Based Field Emitters

low-voltage operation and a full range of colors. A few gate structures and fabrication processes have been suggested [99, 101–103]. The common aim of all these different structures is to efficiently tune the emission currents, simplify the fabrication process, and collimate the electron beam to the corresponding phosphor cells with a finer color tuning. The most common triode structure is a normal gate. The gate electrode is placed between the emitter and the anode. Samsung developed a fully sealed normal-gated FED with an active area of 5 in. [101]. The CNT emitters inside the microcavity are electrically isolated from the gate electrode and sit on a resistive amorphous Si layer. The gate hole and the emitter diameter are 30 m and 20 m, respectively. The applied voltage over the 1.5-m insulator gap between the gate and the cathode turns on the display panel at 65–70 V. The emitted electrons are accelerated through a 1.1-mm vacuum gap at 1.0–1.5 kV. Researchers at Samsung observed that at a higher anode voltage the anode current increased, since the transmittance of the ejected electrons became higher. The device renders 256 gray scales. Since Samsung prepared the emitter array using the printing method, the whole process can be carried out below 400  C. The organic binder degrades the vacuum inside the FED, even after combustion of the organic binder. Therefore, depositing CNT emitters in the microcavity using the thermal/MPE-CVD method might be effective to avoid the current fluctuation due to outgassing [104, 105]. Current fluctuation due to outgassing is observed to be larger than 7% in a fully sealed FED [76]. In the CVD approach, the critical issue is how to control the tube number and vertical alignment inside the microcavity. Precise control of these parameters requires monitoring the grain shape, the size of the Ni particles, and the morphology of the TiN sublayer. Pribat et al. implanted a single CNT in each cavity with a diameter of 200 nm by controlling the size of the catalyst particles. The normal-gated FED fabricated by Samsung has displayed promising picture quality. However, one of the problems observed with the normal-gated FED is crosstalk. The circular shape of the cathode in the microcavity enhances the electric field at the edge of the cathode, leading to a strong widening of the electron trajectory [106]. Therefore, the emission pattern is brighter at the edge rather than at the center due to crosstalk with neighboring pixels. Another disadvantage of the normal-gate structure arises from the cathode fabrication. Since the gate electrode is formed at the final fabrication step, the contamination of the CNT with gate materials is unavoidable. In the under-gate structure, the gate electrode is located below or at the same height as the cathode so that the trajectory of electrons is tilted with respect to the normal direction. The schematic view of an under-gate structure is shown in Figure 5a. The CNT film is pasted at the edge of the cathode electrode. This geometry, which concentrates the electric field at the edge, contributes to a narrowing of the electron trajectory, resulting in a uniform emission pattern and better color separation. A benefit of the under-gate configuration in terms of fabrication is that the manufacturing process becomes greatly simplified. A prototype monochromatic FED, 15 in. in diagonal, designed with the under-gate structure is shown in Figure 5b.

Carbon Nanotube-Based Field Emitters

Figure 5. (a) Schematic view of an under-gate structure. (b) Monochromatic 15-in. FED fabricated with an under-gate structure.

ISE Electronics also demonstrated a CNT-based FED, which is shown in Figure 6a. This FED has many different features compared to the FED made by Samsung. In the display from ISE Electronics, the emitters are prepared using the thermal CVD method. The benefit of the CVD-grown tubes is good uniformity in the emission pattern, resulting from the smooth surface and round edges of the CNT electrodes [107]. However, the problems found during operation of the display are thermal distortion and vibration of the gate electrode. Deformation of the gate electrode is not a result of the growth temperature. The gate electrode is made of metal and is formed like a long metal wire [108]. The gate electrode shows local thermal expansion by emission currents. The inhomogeneous emission pattern contributed by the thermal distortion was circumvented by introducing a gate–insulator composite between the ribs of the anode

Figure 6. (a) Schematic of a triode structure with gate–insulator composite having through holes. (b) CNT-FED panel in 40 in. diagonal. Reprinted with permission from [110], S. Uemura et al., “SID’02 Digest,” 2002, p. 1132. © 2002, Society of Information Display.

619 and cathode electrodes. The composite is compressed by an atmospheric pressure during the evacuation process, which enables the gate–insulator composite to compensate for the thermal distortion. In addition, the combination of metal and insulator is also advantageous in preventing thermal vibration due to the insulator. All these new features in the display enhance the emission pattern tremendously by maintaining the exact distance between the gate and the cathode electrodes over the whole area. ISE Electronics demonstrated a 40-in. CNT-FED panel with the benefits of all the above features, as shown in Figure 6b. The pixel size of the panel is 2 54 mm × 7 62 mm [3 (RGB) ×2 54 mm] [109, 110]. The application of a CNT as an electron multiplier does not seem to be attractive because of the metallic characteristics of the sp2 bond. Therefore, the secondary electron emission (SEE) is not expected to be high. Usually, a large amount of SEE has been observed on the surface of an insulator such as MgO, SiO2 , and diamond. The common property among these materials is that they are insulators with sparse free electrons. Therefore, once the valence electrons are released by energetic incident electrons, the valence electrons do not lose energy by an inelastic collision with conduction electrons. Consequently, those electrons arriving at the surface can escape into the vacuum as long as their kinetic energy is larger than the work function. From all these aspects, the electronic structures of the CNT are not good for the SEE. Nevertheless, what makes the nanotube attractive as an electron multiplier is its geometric structure. The possible application of CNTs to electron multipliers was demonstrated for the first time by Yi et al. [111]. In their experiment, catalytically CVD-grown MWCNTs on a silica glass were coated with a MgO layer, which is one of the SEE materials. As they bombarded the surface of the MgO, a huge electron emission was detected with this geometry. The SEE yield in this case is measured to be 22,000 and is very sensitive to the backbias, a voltage applied to the MWCNTs [112]. This value is promisingly large compared to the MgO by itself, whose SEE yield is about 800. Kim et al. proposed that Townsend avalanche effects due to field amplification at the tip of the MWCNT is responsible for the generation of the huge number of secondary electrons in the MgO film. The large field enhancement factor of CNTs, however, highly exaggerates the electric field at the tip and triggers the secondary electron emission. The field enhancement factor of the MWCNTs is about 1000–2000. The thickness of the MgO layer is also one of the critical parameters determining the SEE yield. The amplification phenomenon is most prominent in a MgO layer with a nominal thickness of about 500–2000 Å. Luminescent lighting elements equipped with carbon nanotube emitters have been successfully exhibited [113, 114]. The lighting elements shown so far are, in principle, working on field emission. In particular, the flashlight designed by Saito et al. is similar to a gated-structured FED. The technical advancements accomplished by CNTs, compared to conventional thermionic filaments, include superior environmental and emission stability together with low power consumption. This device can operate in a high vacuum range

620 and sustain stable emission currents over 4000 h. In addition, the intensity of luminance is twice that of conventional devices [115, 116]. Another application of the lighting element was accomplished by Bonard and his colleagues [114]. A luminescent tube was constructed in a cylindrical shape to take advantage of the stronger electric field than diode structure at the same gap distance. Inside the tube, tungsten wire, which was coated with catalyst particles, was used for the growth of MWCNTs and placed at the center of the tube for electron emission. The I–V characteristics of the tube show a good agreement with the Fowler–Nordheim equation. The brightness obtained from a tube 50 mm long and 21 mm wide is reported to be 10,000 cd/cm2 . Most X-ray tubes currently on the market produce electrons by resistively heating a tungsten filament above 2000  C. Some of the problems naturally inherent in thermionic electron emission have been mentioned before. In addition, the thermionic cathode intrinsically has low response time and the chemical reaction with residual gases shortens the lifetime, which leads to failure of the cathode as well. To liberate the current X-ray tube from the abovementioned problems, the hot cathode has been replaced by a CNT-based cold cathode [117]. In this setup, tungsten wire, on which MWCNTs were grown after deposition of cobalt as a catalyst, was used as the electron source. The MWCNTcovered W cathode contributes to a better X-ray image compared to the thermionic cathode due to the narrower energy distribution of the emitted electrons. It also shows a high operating vacuum pressure of 10−7 torr. However, low emission currents and short lifetime at this pressure prove that the tube is still far from the marketable stage. Yue et al. resolved the problems of coating SWCNTs on a flat metal disk by electrophoresis [118]. In this case, the solid attachment of the CNT to a metal disk is very crucial, particularly under a strong electric field. To enhance the adhesion between the tube and the metal disk, an iron interlayer was evaporated onto the substrate. The as-prepared cathode easily yields a current density of 30 mA/cm2 with negligible fluctuation during operation for 18 h. The above results show that CNT-based X-ray tubes seem to overcome the technical difficulties in present X-ray tubes such as lifetime, environmental stability, and portability.

5. CONCLUSIONS The availability of CNT-based vacuum microelectronics on the market is dependent on technical advancements and a profound understanding of the fundamental field emission mechanism. All the studies carried out on individual or multiple tubes have shown that the field emission properties are well explained by the Fowler–Nordheim formula. Nevertheless, there exist many scientific issues that are not clearly answered. For instance, the effects of defect states existing at the cap, at which field emission actively takes places, have not been studied well. In addition to this defect effect, comprehension of the difference between an open and a closed tube under intensive electric field is also important. The deposition of CNTs over a large area and at a temperature below 500  C is another issue to be solved. The cause of

Carbon Nanotube-Based Field Emitters

field penetration observed from FEM studies is still ambiguous whether it is due to the existence of a semiconducting tube or an insulating layer between the CNT emitter and the metal support. The effective work function at the cap of the clean CNT is one of the intrinsic parameters to be clarified. Several groups have measured the work function of the CNT and reached the conclusion that the work function is slightly less than that of graphite. However, field emission preferentially takes place at the pentagon defects, in which strong localized states are observed near the Fermi level. Therefore, the work function at the emitting site is expected to be different from the previously known values. Comprehension of the work function at the emitting sites can elucidate the real field enhancement factor appearing in the F–N plots, since these two parameters are strongly correlated. There are unique properties originating from low dimensionality. The ballistic movement and Luttinger liquid behavior of conduction electrons have been observed. The details of these properties are well described in the literature [119, 120]. How these low-dimensional properties influence the field emission currents are to be explored in the future. Regardless of the lack of understanding of the intrinsic field emission properties of CNTs, various realizations of CNT-based applications have shown that the CNT is very efficient for field emitters. However, these applications are just prototypes. To meet commercializable performance, the existing technical obstacles should be resolved. One of the key parameters that significantly influences the performance of the FED is phosphor. Even though phosphor has a long history of use, the FED has demanded a new phosphor suitable for low voltage operation. In the cathode ray tube, the electrons are sufficiently accelerated to excite the phosphor screen, which is protected by a thin metal film. The phosphor used in CRTs can generate a full range of colors with acceptable efficiency. In addition, conventional high-voltage phosphors are well characterized, commercially acceptable, and widely used. The most widely used phosphors for color display are P22 red, green, and blue triplets. Among these, green has the most significant effect on brightness. The use of high-voltage phosphor for FED turned out to be unsuitable due to the degradation in lifetime and luminescence efficiency. Large emission currents can be supplied to compensate for the low brightness, but this leads to shorter lifetime due to significant chemical degradation. In addition, the larger emission currents will end up increasing the manufacturing cost for high-power circuits. Numerous materials have been tried and their luminescence properties have been examined. Following are most the promising luminescent materials for low-voltage full-color FED applications: for red, SrTiO3 : Pr, Y2 O3 : Eu, Y2 O2 S : Eu; for blue, ZnGa2 O4 , Y2 SiO5 : Ce, ZnS : Ag, Cl; Y3 (Al, Ga)5 O12 : Tb, for green, Zn(Ga, Al)2 O4 : Mn, Y2 SiO5 : Tb, ZnS : Cu, Al [121]. Among the above phosphors, only the ZnO : Zn phosphor shows a lifetime longer than 10,000 h and stable emission and is comparable in performance to monochromatic FEDs [122]. To improve the current phosphor, various modifications have been made and the modifications can be classified into three categories: formation of conducting films, improvement

Carbon Nanotube-Based Field Emitters

of chemical and radiation stability, and increase in the luminance efficiency [123]. Modification of the luminescent materials has consisted of the formation of the corresponding salt and solution with subsequent annealing in air. The reduction process results in an insulating layer such as SiO2 , MgO, TiO2 , and SnO2 . These layers are resistant to electron bombardment, are chemically stable, and are known to increase the luminescence efficiency and conductivity as well. With all the above properties, new low-voltage phosphor is supposed to have a fast decay time, thermal stability during the panel manufacturing process, no evaporable substances contaminating the emitters, and acceptable grain size for precise coating [124]. The cathode arrays can be prepared using thermal evaporation through a gate hole. This approach, however, is not easy for scale-up and the production cost may be high. Screen printing is a new approach to overcome the scalability problem with an easy manufacturing process. Regarding the size of the cathode, this approach has no limitations. Printing the mixture of CNT and organic binder requires a firing step to burn out the organic binder at about 350  C. After removing the binder, the solid attachment of the CNT to the electrode is very important. Therefore, an appropriate choice of electrode material can reduce the firing temperature [125]. The width of the pasted CNTs is about 120 m on the 390-m-wide cathode electrodes [106]. The large space between the cathode electrodes fabricated by the screenprinting method provides only 240 × 576 lines in a 9-in. FED. The smallest pixel size that one can obtain from the current screen-printing method is about 70 m. Therefore, the finer pitch pattern could define pictures more precisely. A new patterning process permitting a higher resolution FED with finer pitches such as the liftoff method and photolithography has been shown [77, 125]. Using the photolithography approach, the pixel was downsized to 30 m. The solid insulation between the cathode and the anode should be obtained for field emission. To insulate between two electrodes, a thin insulating layer, called a spacer, is used. The choice of spacer materials brings many important issues into consideration. For instance, the spacer must be thin, mechanically strong, vacuum compatible, exhibiting low leakage current, and withstanding a strong electric field [126]. Since the spacer is supporting the voltage difference between the cathode and the anode and being exposed to emitted electrons, the resistance of the spacer should be high enough to prevent leakage current and, simultaneously, low enough to dissipate accumulated charges. Electron bombardment can charge the spacer negatively or positively, depending on the energy of the emitted electrons. Induced charges on the surface of the spacer will significantly change the trajectory of the electrons, resulting in image distortion. Choi et al. have simulated the trajectory of electrons in the presence of the charged spacer. Their results show that the image of the dot arrays near the charged spacer appears less luminescent and the dot size looks smaller than one near the uncharged spacer [127]. In certain cases generation of the secondary electron emission along the spacer surface can lead to disastrous failure of the device. This voltage breakdown along the

621 spacer, called surface flashover, is initiated in the triplejunction region, where spacer, vacuum, and electrode join [128]. Although the detailed mechanism of how the electron ejection at the triple-junction area develops into surface flashover is not clearly understood, the most widely accepted mechanism is secondary electron emission avalanche (SEEA), which is initiated at the triple junction. Since the surface flashover is ignited by electrons emitted from the triple junction, the introduction of a void at the junction area lowers the electric field. In addition to the void, the insertion of metal strips into the spacer can be advantageous in two ways. First, the metal strips reduce the flashover since it is strongly correlated with charging of the spacer. Second, the image distortion due to charge accumulation on the spacer could be minimized by the conducting strips. However, besides the above two parameters, other parameters also have significant effects on the surface flashover. For instance, the length of the spacer, the spacer surface roughness, the sharpness of the electrodes at the edge, the surface condition of the spacer, and the spacer material itself can all trigger a catastrophic vacuum breakdown. Physical sputtering of the CNT emitters shortens the lifetime of the FED rapidly. The prototype CNT-FEDs introduced in the literature degrade significantly in a few hours. While an effort to extend the lifetime has been made using double-wall nanotubes (DWNTs), the FED cannot survive long enough to meet the commercial requirement, about 20,000 h. In this review, we have discussed many important aspects of carbon nanotube–based field emitters that are still under intensive research and development. Because of their short history, some of the phenomena observed during field emission are not well identified. Nevertheless, it has been proven that CNTs are a great applicant for field emitters and tremendous progress has been made. Also, various types of applications have been exhibited. Bringing the prototype devices to the point of acceptable performance and marketable manufacturing cost requires overcoming many of the challenging problems described previously.

GLOSSARY Adsorbate A substance that adheres on the surface of the material under investigation. Anodization Method of plating metal for corrosion resistance, electrical insulation, thermal control, abrasion resistance, sealing, improving paint adhesion, and decorative finishing. Anodizing consists of electrically depositing an oxide film from an aqueous solution onto the surface of a metal, often aluminum, which serves as the anode in an electrolytic cell. Plate properties such as porosity, abrasion resistance, color, and flexibility depend on the type, concentration, and temperature of the electrolyte, the strength of the electrical current and the processing time, and the type of metal being plated. Electronegativity In chemistry, the tendency of an atom to attract an electron pair shared with another atom in a chemical bond.

622 Electrophoresis The electrostatic attractive movement of charged particles in a colloidal solution that is under the influence of an electric field. Fermi level A measure of the energy of the least tightly bound electrons within a solid. The value of the Fermi level at absolute zero is called the Fermi energy and is a constant for each solid. The Fermi level changes as the temperature increases and as electrons are added to or withdrawn from the solid. According to quantum mechanics, each energy level can accommodate only two electrons. The Fermi level is the energy level having the probability that it is exactly half filled with electrons. Levels of lower energy than the Fermi level tend to be entirely filled with electrons, whereas energy levels higher than the Fermi level are unoccupied. Field enhancement Amplification of the applied electric field at the edge of the emitter. Getter A material with a high sticking coefficient. Therefore, residual gases physically colliding with the surface of getter materials are permanently removed from a vacuum system. After the saturation of the getter surface, the getter is reactivated and adsorbed gas molecules sink into the bulk of the getter. Th, V, Ti, and Zr are used for pumping a vacuum system. Microcavity A cavity is about 1 m deep and a few micrometers wide in Spindt structure. This cavity is fabricated by MEMS techniques to form metal emitters through the thermal evaporator. Shielding effect Weakening of the applied electric field at the tip of the emitters due to the high density of the emitters. Thermionic emission Emission of electrons that has high enough energy to overcome the surface potential from thermal excitation. Work function Required energy for electrons to overcome the surface potential barrier.

ACKNOWLEDGMENTS The authors extend their appreciation to those who allowed their prestigious work to be presented in this review. We also thank Hyun Jin Kim and Im Bok Lee for their valuable support. This work was supported by the Ministry of Science and Technology through the National Research Laboratory program and the KOSEF through the Center for Nanotubes and Nanostructured Composites at Sungkyunkwan University.

REFERENCES 1. S. H. Magnus, D. N. Hill, and W. L. Ohlinger, Appl. Surf. Sci. 111, 42 (1997). 2. H. D. Park, “Field Emission Workshop ’02,” 2002, p. 143. 3. E. L. Murphy and R. H. Good, Jr., Phys. Rev. 102, 1461 (1962). 4. P. A. Redhead, J. Vac. Sci. Technol., A 16, 1394 (1998). 5. E. Guth and C. J. Mullin, Phys. Rev. 61, 339 (1942). 6. W. Schottky, Phys. Z. 15, 872 (1914). 7. R. H. Fowler and L. W. Nordheim, Proc. R. Soc. London 119, 173 (1928). 8. W. P. Dyke, J. K. Trolan, E. E. Martin, and J. P. Barbour, Phys. Rev. 91, 1043 (1953).

Carbon Nanotube-Based Field Emitters 9. J. P. Barbour, W. W. Dolan, J. K. Trolan, E. E. Martin, and W. P. Dyke, Phys. Rev. 92, 45 (1953). 10. E. W. Muller and T. T. Tsong, “Field Ion Microscopy.” Elsevier, New York, 1969. 11. C. A. Spindt, I. Brodie, L. Humphrey, and E. R. Westerberg, J. Appl. Phys. 47, 5248 (1976). 12. J. M. Macaulay, I. Brodie, C. A. Spindt, and C. E. Holland, Appl. Phys. Lett. 61, 997 (1992). 13. V. V. Zhirnov, G. J. Wojak, W. B. Choi, J. J. Cuomo, and J. J. Hren, J. Vac. Sci. Technol., A 15, 1733 (1997). 14. M. C. Benjamin, C. Wang, R. F. Davis, and R. J. Nemanich, Appl. Phys. Lett. 64, 3288 (1994). 15. D. A. Lapiano-Smith, E. A. Eklund, F. J. Himpsel, and L. J. Terminello, Appl. Phys. Lett. 59, 2174 (1991). 16. S. C. Lim, R. E. Stallcup II, I. A. Akwani, and J. M. Perez, Appl. Phys. Lett. 75, 1179 (1999). 17. C. A. Spindt, J. Appl. Phys. 39, 3504 (1968). 18. C. A. Spindt, Surf. Sci. 266, 145 (1992). 19. S. Iijima, Nature (London) 354, 56 (1991). 20. M. M. J. Treacy, T. W. Ebbesen, and J. M. Gibson, Nature 381, 678 (1996). 21. A. Krishnan, E. Dujardin, T. W. Ebbesen, P. N. Yianilos, and M. M. J. Treacy, Phys. Rev. B 58, 14013 (1998). 22. J.-P. Salvetat, A. J. Kulik, J.-M. Bonard, G. Andrew, D. Briggs, T. Stöckli, K. Méténier, S. Bonnamy, F. Béguin, N. A. Burnham, and L. Forró, Adv. Mater. 11, 161 (1999). 23. E. W. Wong, P. E. Sheehan, and C. M. Lieber, Science 277, 1971 (1997). 24. P. Chen, X. Wu, X. Sun, J. Lin, W. Ji, and K. L. Tan, Phys. Rev. Lett. 82, 2548 (1999). 25. S. Berber, Y.-K. Kwon, and D. Tománek, Phys. Rev. Lett. 84, 4613 (2000). 26. J. W. Mintmire, B. I. Dunlap, and C. T. White, Phys. Rev. Lett. 68, 631 (1992). 27. R. Saito, M. Fujita, G. Dresselhaus, and M. S. Dresselhaus, Phys. Rev. B 46, 1804 (1992). 28. C. T. White, D. H. Robertson, and J. W. Mintmire, Phys. Rev. B 47, 5485 (1993). 29. R. Saito, G. Dresselhaus, and M. S. Dresselhaus, J. Appl. Phys. 73, 494 (1993). 30. A. Rochefort, D. R. Salahub, and Ph. Avouris, Chem. Phys. Lett. 297, 45 (1998). 31. R. Tamura and M. Tsukada, Phys. Rev. B 49, 7697 (1994). 32. W. Rivera, J. M. Perez, R. S. Ruoff, D. C. Lorents, R. Malhotra, S. Lim, Y. G. Rho, E. G. Jacobs, and R. F. Pinizzotto, J. Vac. Sci. Technol., B 13, 327 (1995). 33. K. Sattler, Carbon 33, 915 (1995). 34. C. H. Olk and J. P. Heremans, J. Mater. Res. 9, 259 (1994). 35. A. M. Rao, E. Richter, S. Bandow, B. Chase, P. C. Eklund, K. A. Williams, S. Fang, K. R. Subbaswamy, M. Menon, A. Thess, R. E. Smalley, G. Dresselhaus, and M. S. Dresselhaus, Science 275, 187 (1997). 36. K. Hata, A. Takakura, and Y. Saito, Surf. Sci. 490, 296 (2001). 37. M. J. Fransen, Th. L. van Rooy, and P. Kruit, Appl. Surf. Sci. 146, 312 (1999). 38. Y. Saito and S. Uemura, Carbon 38, 169 (2000). 39. D. L. Carroll, P. Redlich, P. M. Ajayan, J. C. Charlier, X. Blase, A. De Vita, and R. Car, Phys. Rev. Lett. 78, 2811 (1997). 40. R. Tamura and M. Tsukada, Phys. Rev. B 52, 6015 (1995). 41. K. A. Dean, P. von Allmen, and B. R. Chalamalab, J. Vac. Sci. Technol., B 17, 1959 (1999). 42. T. Guo, P. Nikolaev, A. Thess, D. T. Colbert, and R. E. Smalley, Chem. Phys. Lett. 243, 49 (1995). 43. Y. S. Park, K. S. Kim, H. J. Jeong, W. S. Kim, J. M. Moon, K. H. An, D. J. Bae, Y. S. Lee, G.-S. Park, and Y. H. Lee, Synth. Met. 126, 245 (2002).

Carbon Nanotube-Based Field Emitters 44. J.-M. Moon, K. H. An, Y. H. Lee, Y. S. Park, D. J. Bae, and G.-S. Park, J. Phys. Chem. B 105, 5677 (2001). 45. H. J. Jeong, S. Y. Jeong, Y. M. Shin, J. H. Han, S. C. Lim, S. J. Eum, C. W. Yang, N. G. Kim, C. Y. Park, and Y. H. Lee, Chem. Phys. Lett. 361, 189 (2002). 46. G. Che, B. B. Lakshmi, C. R. Martin, E. R. Fisher, and R. S. Rouff, Chem. Mater. 10, 260 (1998). 47. J. Kong, H. T. Soh, A. M. Cassell, C. F. Quate, and H. Dai, Nature 395, 878 (1998). 48. H. J. Jeong, Y. M. Shin, S. Y. Jeong, Y. C. Choi, Y. S. Park, S. C. Lim, G.-S. Park, I.-T. Han, J. M. Kim, and Y. H. Lee, Chemical Vapor Deposition 8, 11 (2001). 49. O. Groning, O. M. Kuttel, Ch. Emmenegger, P. Groning, and L. Schlapbach, J. Vac. Sci. Technol., B 18, 665 (2000). 50. S. Maruyama, R. Kojima, Y. Miyauchi, S. Chiashi, and M. Kohno, Chem. Phys. Lett. 360, 229 (2002). 51. Y. Zhang, A. Chang, J. Cao, Q. Wang, W. Kim, Y. Li, N. Morris, E. Yenilmez, J. Kong, and H. Daia, Appl. Phys. Lett. 79, 3155 (2001). 52. Y. C. Choi, D. J. Bae, Y. H. Lee, B. S. Lee, I. T. Han, W. B. Choi, N. S. Lee, and J. M. Kim, Synth. Met. 108, 159 (2000). 53. C. J. Lee, D. W. Kim, T. J. Lee, Y. C. Choi, Y. S. Park, W. S. Kim, Y. H. Lee, W. B. Choi, N. S. Lee, J. M. Kim, Y. G. Choi, and S. C. Yu, Appl. Phys. Lett. 75, 1721 (1999). 54. Y. C. Choi, D. J. Bae, Y. H. Lee, B. S. Lee, G.-S. Park, W. B. Choi, N. S. Lee, and J. M. Kim, J. Vac. Sci. Technol., A 18, 1864 (2000). 55. A. Cao, X. Zhang, C. Xu, J. Liang, and D. Wu, Appl. Phys. Lett. 79, 1252 (2001). 56. D. Ugarte, A. Châtelain, and W. A. de Heer, Science 268, 845 (1995). 57. M. Endo, K. Takeuchi, S. Igarashi, K. Kobori, M. Shiraishi, and H. W. Kroto, J. Phys. Chem. Solids 54, 1841 (1993). 58. V. Ivanov, J. B. Nagy, Ph. Lambin, A. Lucas, X. B. Zhang, X. F. Zhang, D. Bernaerts, G. Van Tendeloo, S. Amelinckx, and J. Van Landuyt, Chem. Phys. Lett. 223, 329 (1994). 59. W. Z. Li, S. S. Xie, L. X. Qian, B. H. Chang, B. S. Zou, W. Y. Zhou, R. A. Zhao, and G. Wang, Science 274, 1701 (1996). 60. S. Suzuki, Y. Watanabe, T. Kiyokura, K. G. Nath, T. Ogino, S. Heun, W. Zhu, C. Bower, and O. Zhou, Phys. Rev. B 63, 245418 (2001). 61. C. Kim, B. Kim, S. M. Lee, C. Jo, and Y. H. Lee, Phys. Rev. B 65, 165418 (2001). 62. J.-M. Bonard, J.-P. Salvetat, T. Stöckli, W. A. de Heer, L. Forró, and A. Châtelain, Appl. Phys. Lett. 73, 918 (1991). 63. Y. K. Kwon, Y. H. Lee, S. G. Kim, Ph. Jund, R. E. Smalley, and D. Tomanek, Phys. Rev. Lett. 79, 2065 (1997). 64. A. L. C. Triques, C. M. B. Cordeiro, V. Balestrieri, B. Lesche, W. Margulis, and I. C. S. Carvalho, Appl. Phys. Lett. 76, 2469 (2000). 65. Y. C. Choi, Y. M. Shin, S. C. Lim, D. J. Bae, Y. H. Lee, and B. S. Lee, J. Appl. Phys. 88, 4898 (2000). 66. G. Gu, G. Philipp, X. Wu, M. Burghard, A. M. Bittner, and S. Roth, Adv. Funct. Mater. 11, 295 (2001). 67. R. Kurt, J.-M. Bonard, and A. Karimi, Thin Solid Films 398, 193 (2001). 68. A. M. Cassell, N. R. Franklin, T. W. Tombler, E. M. Chan, J. Han, and H. Dai, J. Am. Chem. Soc. 121, 7975 (1999). 69. J.-M. Bonard, H. Kind, T. Stöckli, and L.-O. Nilsson, Solid-State Electron. 45, 893 (2001). 70. D. N. Davydov, P. A. Sattari, D. AlMawlawi, A. Osika, T. L. Haslett, and M. Moskovits, J. Appl. Phys. 86, 3983 (1999). 71. J. Li, C. Papadopoulos, and J. M. Xu, Appl. Phys. Lett. 75, 367 (1999). 72. J. S. Suh and J. S. Lee, Appl. Phys. Lett. 75, 2047 (1999). 73. S. H. Jeong, H. Y. Hwang, K. H. Lee, and Y. Jeong, Appl. Phys. Lett. 78, 2052 (2001).

623 74. Z. H. Yuan, H. Huang, H. Y. Dang, J. E. Cao, B. H. Hu, and S. S. Fan, Appl. Phys. Lett. 78, 3127 (2001). 75. J. S. Suh, K. S. Jeong, and J. S. Lee, Appl. Phys. Lett. 80, 2392 (2002). 76. W. B. Choi, D. S. Chung, J. H. Kang, H. Y. Kim, Y. W. Jin, I. T. Han, Y. H. Lee, J. E. Jung, N. S. Lee, G. S. Park, and J. M. Kim, Appl. Phys. Lett. 75, 3129 (1999). 77. Y. R. Cho, J. H. Lee, C. S. Hwang, Y. H. Song, H. S. Uhm, D. H. Kim, S. D. Ahn, C. H. Chung, B. C. Kim, and K. I. Cho, Jpn. J. Appl. Phys., Part 1 41, 1532 (2002). 78. C. A. Spindt, J. Appl. Phys. 33, 2917 (1962). 79. V. T. Binh and Ch. Adessi, Phys. Rev. Lett. 85, 864 (2000). 80. S. C. Lim, H. J. Jeong, Y. M. Shin, K. H. An, D. J. Bae, Y. H. Lee, N. S. Lee, and J. M. Kim, Adv. Mater. 13, 1536 (2001). 81. K. A. Dean and B. R. Chalamala, Appl. Phys. Lett. 76, 375 (2000). 82. L. Nilsson, O. Groning, P. Groning, O. Kuttel, and L. Schlapbach, Thin Solid Films 383, 78 (2001). 83. H. Ago, T. Kugler, F. Cacialli, W. R. Salaneck, A. H. Windle, and R. H. Friend, J. Phys. Chem. B 103, 8116 (1999). 84. S. C. Lim, C. Jo, H. J. Jeong, Y. M. Shin, Y. H. Lee, I. A. Samayoa, and J. Choi, Jpn. J. Appl. Phys., Part 1 41, 5635 (2002). 85. C. Kim, Y. S. Choi, S. M. Lee, J. T. Park, B. Kim, and Y. H. Lee, J. Am. Chem. Soc. 124, 9906 (2002). 86. A. Maiti, J. Andzelm, N. Tanpipat, and Ph. von Allmen, Phys. Rev. Lett. 87, 155502 (2001). 87. N. Park, S. Han, and J. Ihm, Phys. Rev. B 64, 125401 (2001). 88. Y. G. Hwang and Y. H. Lee, unpublished. 89. K. Hata, M. Ariff, K. Tohji, and Y. Saito, Chem. Phys. Lett. 308, 343 (1999). 90. Y. H. Lee, G. K. Seong, and D. Tománek, Chem. Phys. Lett. 265, 667 (1997). 91. K. A. Dean, T. P. Burgin, and B. R. Chalamala, Appl. Phys. Lett. 79, 1873 (2001). 92. V. V. Zhirnov, A. N. Alimova, and J. J. Hren, Appl. Surf. Sci. 191, 20 (2002). 93. J. H. Jung, B. K. Ja, Y. H. Lee, J. Jang, and M. H. Oh, IEEE Electron Device Lett. 18, 197 (1997). 94. C. Kimura, T. Yamamoto, T. Hori, and T. Sugino, Appl. Phys. Lett. 79, 4533 (2001). 95. S. Dimitrijevic, J. C. Withers, V. P. Mammana, O. R. Monteiro, J. W. Ager III, and I. G. Brown, Appl. Phys. Lett. 75, 2680 (1999). 96. S. Yu, W. Yi, T. Jeong, W. S. Kim, J. Lee, H. Heo, C. S. Lee, J.-B. Yoo, Y. H. Lee, and J. M. Kim, Physica B 323, 177 (2002). 97. A. Wadhawan, R. E. Stallcup II, and J. M. Perez, Appl. Phys. Lett. 78, 108 (2001). 98. J. M. Macaulay, I. Brodie, C. A. Spindt, and C. E. Holland, Appl. Phys. Lett. 61, 997 (1994). 99. Q. H. Wang, A. A. Setlur, J. M. Lauerhaas, J. Y. Dai, E. W. Seelig, and R. P. H. Chang, Appl. Phys. Lett. 72, 2912 (1998). 100. J. M. Kim, W. B. Choi, N. S. Lee, and J. E. Jung, Diamond Relat. Mater. 9, 1184 (2000). 101. D.-S. Chung, S. H. Park, H. W. Lee, J. H. Choi, S. N. Cha, J. W. Kim, J. E. Jang, K. W. Min, S. H. Cho, M. J. Yoon, J. S. Lee, C. K. Lee, J. H. Yoo, J.-M. Kim, J. E. Jung, Y. W. Jin, Y. J. Park, and J. B. You, Appl. Phys. Lett. 80, 4045 (2002). 102. C. G. Lee, S. H. Jo, E. J. Chi, J. S. Lee, S. J. Lee, S. H. Cho, J. W. Kim, B. G. Lee, J. C. Cha, H. S. Han, S. H. Ahn, K. W. Jung, K. S. Ryu, S. Y. Park, S. H. Jin, Y. S. Choi, J. W. Nam, H. Y. Kim, Y. S. Han, H. J. Lee, S. J. Lee, J. H. You, and J. M. Kim, “IDMC ’02,” 2002. 103. Q. H. Wang, M. Yan, and R. P. H. Chang, Appl. Phys. Lett. 78, 1294 (2001). 104. D. Pribat, G. Pirio, P. Legagneux, K. B. K. Teo, M. Chhowalla, G. A. J. Amaratunga, W. I. Milne, and D. G. Hasko, “IDMC ’02,” 2002. 105. Y.-H. Lee, Y.-T. Jang, D.-H. Kim, J.-H. Ahn, and B.-K. Ju, Adv. Mater. 13, 479 (2001).

624 106. N. S. Lee, D. S. Chung, I. T. Han, J. H. Kang, Y. S. Choi, H. Y. Kim, S. H. Park, Y. W. Jin, W. K. Yi, M. J. Yun, J. E. Jung, C. J. Lee, J. H. You, S. H. Jo, C. G. Lee, and J. M. Kim, Diamond Relat. Mater. 10, 265 (2001). 107. S. Uemura, “IDRC ’00,” 2000, p. 398. 108. J. Yotanoi and S. Uemura, “IDRC ’01,” 2001, p. 1209. 109. S. Uemura, J. Yotani, T. Nagasako, H. Kurachi, H. Yamada, T. Ezaki, T. Maesoba, Y. Saito, and M. Yumura, “IDMC ’02,” 2002. 110. S. Uemura, J. Yotani, T. Nagasako, H. Kurachi, H. Yamada, T. Ezaki, T. Maesoba, and T. Nakao, “SID ’02 Digest,” 2002, p. 1132. 111. W. Yi, S. Yu, W. Lee, I. T. Han, T. Jeong, Y. Woo, J. Lee, S. Jin, W. Choi, J. Heo, D. Jeon, and J. M. Kim, J. Appl. Phys. 89, 4091 (2001). 112. W. S. Kim, W. Yi, S. Yu, J. Heo, T. Jeong, J. H. Lee, C. S. Lee, J. M. Kim, H. J. Jeong, Y. M. Shin, and Y. H. Lee, Appl. Phys. Lett. 81, 1098 (2002). 113. Y. Saito, S. Uemura, and K. Hamaguchi, Jpn. J. Appl. Phys. 37, L346 (1998). 114. J.-M. Bonard, T. Stöckli, O. Noury, and A. Châtelain, Appl. Phys. Lett. 78, 2775 (2001). 115. S. Uemura, Y. Seko, H. Kamogawa, M. Morikawa, and T. Shimojo, ITE Tech. Rep. 17, 31 (1999). 116. M. Morikawa, Y. Seko, H. Kamogawa, M. Morikawa, S. Uemura, and T. Shimojo, “Japan Display ’92,” 1992, p. 385.

Carbon Nanotube-Based Field Emitters 117. H. Sugie, M. Tanemura, V. Filip, K. Iwata, K. Takahashi, and F. Okuyama, Appl. Phys. Lett. 78, 2578 (2001). 118. G. Z. Yue, Q. Qiu, B. Gao, Y. Cheng, J. Zhang, H. Shimoda, S. Chang, J. P. Lu, and O. Zhou, Appl. Phys. Lett. 81, 355 (2002). 119. S. Frank, Ph. Poncharal, Z. L. Wang, and W. A. de Heer, Science 280, 1744 (1998). 120. Z. Yao, H. W. Ch. Postma, L. Balents, and C. Dekker, Nature 402, 273 (1999). 121. S. Itoh, H. Toki, F. Kataoka, Y. Sato, K. Tamura, and Y. Kagawa, IEICE Trans. Electron. E82-C, 1808 (1999). 122. S. Itoh, T. Watanabe, T. Yamaura, and K. Yano, “Technical Digest of Asia Display ’95,” 1995, p. 617. 123. S. A. Bukesov, “Proceedings of SPIE Advanced Display Technologies: Basic Studies of Problems in Information Display (FLOWERS 2000),” 2000, p. 43. 124. S. Itoh, Y. Yonezawa, H. Toki, F. Kataoka, K. Tamura, and Y. Sato, “Extended Abstracts of the Third International Conference on Science and Technology of Display Phosphor,” 1997, p. 275. 125. Y.-R. Cho, J. H. Lee, Y.-H. Song, S.-Y. Kang, M.-Y. Jung, C.-S. Hwang, and K. I. Cho, J. Vac. Sci. Technol. 19, 1012 (2001). 126. X. Ma and T. S. Sudarshan, IEEE Trans. Dielectric Electric Insul. 7, 277 (2000). 127. Y. S. Choi, S. N. Cha, S. Y. Jung, J. W. Kim, J. E. Jung, and J. M. Kim, IEEE Trans. Electron Devices 47, 1673 (2000). 128. H. C. Miller, IEEE Trans. Electric Insul. 28, 512 (1993).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Carbon Nanotube-Based Supercapacitors Young Hee Lee, Kay Hyeok An, Ji Young Lee, Seong Chu Lim Sungkyunkwan University, Suwon, Korea

CONTENTS 1. Introduction 2. Principles of Operation 3. Differences Between a Supercapacitor and a Battery 4. Electrode Materials for Supercapacitors 5. Carbon Nanotube-Based Supercapacitors 6. Conclusions Glossary References

1. INTRODUCTION Recently, there have been considerable attempts to use carbon nanotubes (CNTs) for the electrodes of electrochemical energy storage systems, such as Li-ion secondary batteries [1–6], electrochemical hydrogen storage [7–9], fuel cells [10, 11], and supercapacitors [12–21]. Carbon nanotubes are attractive materials for the electrodes of electrochemical energy storage devices due to their superb characteristics of chemical stability, low mass density, low resistivity, and large surface area. Recent developments in massive synthesis of carbon nanotubes [22–24] have accelerated new applications of these materials to the area of electrical energy storage systems. In particular, applications of CNTs as the electrode materials for supercapacitors have ignited significant worldwide investigations on their microscopic and macroscopic porous structures and electrochemical behaviors [12–14, 20–21, 25– 30]. The CNT electrodes exhibit a unique pore structure and high usage efficiency of specific surface areas [12, 15, 17, 19– 21]. The CNT electrodes for the electric double-layer type of supercapacitor have excellent absorption characteristics due to the accessible mesopores formed by the entangled individual CNTs [12, 13, 19–21]. Supercapacitors [31–36] (also called electrochemical capacitors, electric double-layer capacitors, or ultracapacitors) have several advantages compared to the secondary ISBN: 1-58883-057-8/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

battery, for instance, long cycle life (>100000 cycles), simple principle and mode of construction, short charging time, safety, and high power density. Among them, the power density of supercapacitors is the most remarkable property. The power density of supercapacitors is larger than that of the secondary battery by about 10 times [31, 35, 36], although the energy density of supercapacitors is smaller than that of the secondary battery, which is a serious drawback to be applied for practical devices. Supercapacitors have been used as small-scale energy storage devices in electronic stationary, such as memory backup devices and solar batteries with semipermanent charge–discharge cycle life. Now the capability of supercapacitors with high power density extends its application to various other novel devices of load leveling, hybrid capacitor–battery system, cold-starting assistance, and catalytic converter preheating [31]. To apply the supercapacitors to various practical devices—even to electric vehicles—the development of supercapacitors with both high power density and high energy density is prerequisite. The principle of electric energy storage by an electric double layer in a charged capacitor has been known since 1745. The idea of utilizing this principle to store the electrochemical energy for practical purposes of battery cells had been first proposed in the patent granted to Becker in 1957 [37]. The patent proposed electrical energy storage by means of the charges held at the interfacial double layer of a porous carbon material in aqueous electrolytes. After Becker, the SOHIO Corporation in Cleveland commercialized the electric double-layer capacitor (EDLC) with high surface area of carbon materials in a nonaqueous solvent (organic solvent) containing a dissolved tetraalkylammonium salt electrolyte [38]. Such organic electrolyte systems provided higher operating voltages (2.3–3.0 V) owing to the higher decomposition voltages of organic electrolytes than those of aqueous ones. Therefore, they can accommodate higher charge densities and provide larger specific energy storage increase by the square of the operating voltage attainable on charge. A different principle of the capacitor using metal oxide electrodes, originating from the redox reaction as in a battery, was developed between 1975 and 1981 by the Conway group [39–42]. Although these systems have some problems

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 1: Pages (625–634)

Carbon Nanotube-Based Supercapacitors

626 for commercial applications in that the electrode materials are expensive and have a limiting nominal voltage of 1 V, useful military applications have been developed [31]. After the discovery of conducting polymers in the 1980s, the possibility of employing electroactive conducting polymers, such as polypyrrole or polyaniline, for electrochemical pseudocapacitor materials was suggested by the Gottesfeld group [43–45]. These conducting polymer systems offer opportunities for storing electrochemical energy through the redox capacitance, the so called pseudocapacitance. The large capacitance compared to conventional capacitors (on the order of several or more farads per gram) that can be developed with metal oxides and conducting polymer–pseudocapacitor systems and also with carbon double-layer capacitors led to the terms “supercapacitor” and “ultracapacitors” for these two types of high-specificcapacitance devices [31, 35]. Recently, it has been suggested that the more general term “electrochemical capacitors” be used to refer to these systems [31, 35]. The operating principles of electric double-layer capacitors and electrochemical pseudocapacitors, the differences between these two types of capacitors, the different types of electrode materials, and the differences between a supercapacitor and a battery are discussed in Sections 2–4.

2. PRINCIPLES OF OPERATION In a conventional capacitor (condenser), the charge accumulation is achieved electrostatically by positive and negative charges residing on two interfaces separated by a vacuum or a molecular dielectric (a film of mica, a space of air, or an oxide film). Supercapacitors store the electric energy in an electrochemical double layer formed at the interfaces between the polarizable electrodes and the electrolyte solution. Positive and negative ionic charges within the electrolyte accumulate at the surface of the polarizable electrodes and compensate for the electronic charges at the electrode surface, as shown in Figure 1. This charge distribution layer is called the electric double layer (or electrochemical double layer). Figure 1 presents the principle of an electrochemical capacitor. The thickness of the double layer depends on the concentration of the electrolyte and on

the size of the ionic clusters and is typically on the order of 5–10 Å for concentrated electrolytes [46]. The capacitance, C, accumulated in the electric double layer formed at the interface between the polarizable electrodes and the electrolyte solution is defined by C = /4

+ + - + + + - + + + - + -

+ + + + + + + +

charge

+ + -

discharge

+ -

-

+ + + + + + + +

-

+ + + + + +

-

Electrode II

Electrolyte

electrolyte

electrolyte

(1)

dS

where  is the dielectric constant of the electrolyte,  is the distance from the electrode interface to the center of the ion, and S is the surface area of the electrode interface. The corresponding electric field in the electrochemical double layer is very high and assumed to be up to 106 V/cm easily [31, 35]. Compared to conventional capacitors where a total capacitance is typically on the order of picofarads and microfarads, the capacitance and the energy density stored in the supercapacitor by the electrochemical double layer is much higher. To achieve a higher capacitance, the surface area of the electrode is additionally enlarged by using porous electrodes, where an extremely large internal surface area is expected. There are several techniques for determining the specific capacitance, such as a unit-cell test (two-electrode system), a half-cell test (three-electrode system), and an impedance test. The unit-cell and half-cell tests are mainly used to determine the specific capacitance of the supercapacitor. The specific capacitances reported in the literature are not consistent, mainly due to the experimental methods used to determine them. For the sake of consistency, it is worth specifying the electrochemical technique for intercalculating the specific capacitance between the two-electrode and three-electrode systems. Figure 2a shows the double layer of electrodes used in the two-electrode system (2E), which represents a real double-layer supercapacitor device and its equivalent circuit. Figure 2b shows the double layer of electrodes used in the three-electrode system (3E), which is used in the lab cell with a reference electrode and its equivalent circuit. Assuming that the weight of each individual electrode is m, then C1 = C2 = C. The capacitance measured for Electrode I

load




-

+ + + + + +

Potentiostat

-

+ + + + + +

Reference electrode Counter electrode

Working electrode

y0+ y1 y0 C1

y0- y1

(a)

C2

(a)

C

(b)

(b)

Figure 1. Principle of an electric double-layer capacitor.

Figure 2. Electric double layer and its equivalent circuit in (a) twoelectrode system and (b) three-electrode system.

Carbon Nanotube-Based Supercapacitors

627

the two-electrode system is C2E = 21 C. The specific capacitance turns out be Cspec-2E = C2E /2m = 41 C/m . However, for the three-electrode system, the double-layer capacitance measured is C3E = C and the specific capacitance is Cspec-3E = C3E /m = C/m . Thus, the relationship between the specific capacitance measured with the two-electrode and three-electrode techniques is Cspec-3E = 4Cspec-2E . In the double layer at plane electrodes, charge densities of about 16–50 F/cm2 are commonly realized [31]. Taking an average value of 30 F/cm2 , the capacitance of a single polarizable electrode with a typical surface area of 1000 m2 /g for porous materials leads to a specific capacitance of 300 F/g. At 1 V in an aqueous electrolyte, the maximum storage energy, E, is E = CVi2 /2 = 300 × 12 /2 = 150 W-s/g, 150 kJ/kg, or 42 W-h/kg, theoretically. This value is considerably lower than that obtained for available batteries but much higher than that for conventional capacitors. It should be mentioned that the above value depends on the double-layer capacitance, the specific surface area of the respective electrode material, the wetting behavior of the pores, and the nominal cell voltage. The maximum power density of a supercapacitor is given by Pmax = Vi2 /4R Vi = initial voltage, R = equivalent series resistance (ESR)] [31]. Therefore, the key factors determining the power of supercapacitors are the resistivity of the electrode itself, the resistivity of the electrolyte within the porous layer of electrode, and the contact resistance between the electrode and the current collector, as shown in Figure 3. In evaluating the performance of supercapacitors, the characterization of their energy density and power

current collector

electrode material

separator

electrolyte

current collector

A1 A2 A3 An

Rsn

+ +

Cn + +

+

resistance attributed to collect charges in pores

+

contact resistance

RCn

RLn

+

capacitor element

density is one of the most important factors of rating electrochemical power devices. In addition, from practical and fundamental points of view, there is a question of how the energy density and power density are related for various types of electrochemical power sources, including fuel cells and rechargeable batteries.

3. DIFFERENCES BETWEEN A SUPERCAPACITOR AND A BATTERY The most important difference between a supercapacitor and a battery is the principle of electrochemical energy storage. Electrochemical energy can be stored in two fundamentally different ways. In a battery, the potentially available chemical energy storage requires Faradic oxidation and reduction of electrochemically active reagents to release charges that can perform electric work when they flow between two electrodes having different potentials; that is, the charge storage is achieved by an electron transfer that produces a redox reaction in the electroactive materials according to Faraday’s law [31, 47, 48]. With an electric double-layer capacitor (EDLC), the charge storage process is non-Faradic; that is, ideally, no electron transfer takes place across the electrode interface and the storage of electric charge and energy is electrostatic. Actual electron charges are accumulated on the electrode surfaces with lateral repulsion and involvement of redox chemical changes. Table 1 summarizes the perceived advantages and disadvantages of EDLC energy storage. Because the charging and discharging of such EDLCs involve no chemical phase and composition changes, such capacitors have a high degree of cyclability on the order of 106 times and a high specific power density, although the specific energy density is rather small [31]. However, in some cases of the supercapacitor based on pseudocapacitance (redox type of supercapacitor), the essential process is Faradic; that is, the charge storage is achieved by an electron transfer that produces a redox reaction (Faradic reaction) in the electroactive materials according to Faraday’s law. The supercapacitors based on pseudocapacitance have higher specific capacitance than the EDLCs, due to the redox reaction as in a battery, although the redox reaction gives rise to

resistance attributed to Ion migration

Table 1. Advantages and disadvantages of supercapacitor energy storage.

(a) RL Rc R1

C1 C2

R2

Rn

Cn

(b) Figure 3. Equivalent circuit of an electrochemical capacitor.

Advantages Long cycle life, >100000 cycles Excellent power density, >106 W/kg Simple principle and mode of construction Combines state of charge indication Can be combined with secondary battery for hybrid applications (electric vehicles) Disadvantages Limited energy density Poor volume energy density Low working voltage Requires stacking for high potential operation (electric vehicles)

Carbon Nanotube-Based Supercapacitors

628 high internal resistance in supercapacitors, resulting in a decrease in specific power density. The typical electrodes of the supercapacitors based on pseudocapacitance are metal oxides (i.e., RuO2 , IrO2 , Co3 O4 and conducting polymers (i.e., polypyrrole, polyaniline, polythiophene) [39–45]. A supercapacitor requires two equivalent electrodes, one of which is charged negatively with respect to the other, the charge storage and separation being electrostatic. At each electrode, the charge storage and separations are established across the electrode interface. Usually, the electrodes of supercapacitors have high surface area and porous matrices. However, batteries have bipolar electrode configurations for higher voltage series combinations [31]. For a battery, the maximum Gibbs energy is the product of charge Q and the difference of potential, E, between the Nernstian reversible potentials of the two electrodes, that is, G = Q · E. In the capacitor case, for a given charge Q, G is 21 QV . For a given electrode potential difference, E = V , it is then evident that the energy stored by a twoelectrode cell accommodating a given Faradic charge Q at voltage E = V , is twice that stored in a capacitor charged with the same Q at the same voltage. In the process of charging a pure electric double-layer capacitor, every additional element of charge has to do electrical work (Gibbs energy) against the charge density already accumulated on the electrodes, progressively increasing the interelectrode potential difference [31, 47–50]. In a battery cell being charged, a thermodynamic potential (ideally) exists independent of the extent of charge Q added, as long as two components (reduced and oxidized forms) of the electroactive material remain coexisting. Thus, the potential difference (electromotive force) of the battery cell is ideally constant throughout the discharge or recharge half-cycles, so that G is Q · E rather than Q · 21 E (or 1 V ) [31, 47–50]. This difference can be illustrated by the 2 discharge curves shown schematically in Figure 4, where the voltage in the capacitor declines linearly with the extent of charge, while that for an ideal battery remains constant as long as two phases remain in equilibrium. The decline in the supercapacitor voltage arises formally since C = Q/V or V = Q/C; therefore, dV /dQ = 1/C [31, 47–50]. The ideal battery cell voltages on discharge and recharge, as a function of state of charge, are shown as parallel lines of zero slope in Figure 4. In the slope of the discharge and recharge charge

Potential

Battery discharge

charge Supercapacitor

discharge

Time Figure 4. Difference in discharge and recharge relationships for a supercapacitor and a battery.

Table 2. Overal comparison of supercapacitor and battery characteristics. Item

Supercapacitor

Slope of charge and discharge curve Intrinsic stage of charge indication Energy density Power density Cyclability and cycle life Origin of internal IR

Declining slope

Constant slope

Good

Bad

Activation polarization Lifetime Cell stacking by bipolar system

Battery

Poor Good Good Poor Excellent Bad High area matrix Active electrode + electrolyte materials + electrolyte Little or none Significant (Faradic resistance) Long Poor Possible Impossible

lines for the supercapacitor in Figure 4, there is a significant IR drop, depending on the discharging and recharging rates. An overall comparison of electrochemical capacitor and battery characteristics is given in Table 2.

4. ELECTRODE MATERIALS FOR SUPERCAPACITORS 4.1. Metal Oxides The concept and use of metal oxide as an electrode material in electrochemical capacitors can be traced back to the paper by Trasatti and Buzzanca [41] on ruthenium dioxide (RuO2 as a new interesting electrode material. Some other oxides, such as IrO2 , Co3 O4 , MoO3 , WO3 , and TiO2 , as electrode materials in electrochemical capacitors have been discovered [51–54]. The cyclic voltammogram of the metal oxide electrodes has an almost rectangular shape and exhibits good capacitor behavior [31, 35, 55, 56]. However, the shape of the cyclic voltammogram is not a consequence of pure double-layer charging, but a consequence of the redox reactions occurring in the metallic oxide, giving rise to the redox pseudocapacitance [31, 35, 55, 56]. A very high specific capacitance of up to 750 F/g was reported for RuO2 prepared at relatively low temperatures [57]. Conducting metal oxides such as RuO2 or IrO2 were the favored electrode materials in early electrochemical capacitors used for space or military applications [31, 35]. The high specific capacitance in combination with the low resistance resulted in very high specific powers. An energy density of 8.3 W-h/kg and a power density of 30 kW/kg was achieved in a prototype 25-V electrochemical capacitor built only with RuO2 · xH2 O material and electrolyte [31]. These capacitors, however, turned out to be too expensive. A rough calculation of the capacitor cost showed that 90% of the cost resides in the electrode material [31, 35]. In addition, these capacitor materials are only suitable for aqueous electrolytes, thus limiting the nominal cell voltage to 1 V [31, 35]. Several studies have attempted to take advantage of the material properties of such metal oxides at a reduced cost.

Carbon Nanotube-Based Supercapacitors

The dilution of the costly noble metal by the formation of perovskites was investigated by Guther et al. [58]. Other forms of metal compounds such as nitrides were investigated by Liu et al. [59]. However, these materials are not yet commercially available in the electrochemical capacitor market.

4.2. Conducting Polymers The discovery of conducting polymers has given rise to a rapidly developing field of electrochemical polymer science. conducting polymers, such as polyacetylene, polyaniline, polythiophene, and polypyrrole, have been suggested by several authors for electrochemical capacitors [43–45, 49]. The conducting polymers have fairly high electronic conductivities, typically of magnitudes of 1–100 S/cm [31, 35, 49]. The electrochemical processes of conducting polymers are electrochemical redox reactions associated with sequential Lewis acid or Lewis base production steps so that the polymer molecules are converted to multiply charged structures through electrochemical Lewis-type reactions involving electron withdrawal or electron donation [31, 35, 49]. Therefore, the pseudocapacitance by Faradic redox processes in conducting polymer-based electrochemical capacitors is dominant, although about 2–5% of double-layer capacitance is included in the total specific capacitance [31, 35]. Such polymer electrode materials are much cheaper than RuO2 or IrO2 and can generate comparably large specific capacitance. However, the polymer electrode materials do not have the long-term stability and cycle life during cycling, which may be a fatal problem in applications. Swelling and shrinking of electroactive conducting polymers is well known and may lead to degradation during cycling [31, 35, 49]. Therefore, these electroactive conducting polymers are also far from being commercially used in electrochemical capacitors.

4.3. Carbons Carbon materials for electrochemical energy devices, such as secondary batteries [60–62], fuel cells [63–65], and supercapacitors [66–69], have been extensively studied. However, each type of electrochemical energy device requires different physical properties and morphology. For supercapacitors, the carbon material for the EDLC type must have (i) high specific surface area, (ii) good intra- and interparticle conductivity in porous matrices, (iii) good electrolyte accessibility to intrapore surface area, and (iv) the available electrode production technologies [31, 35, 36]. Carbons for supercapacitors are available with a specific surface area of up to 2500 m2 /g as powders, woven cloths, felts, or fibers [31, 35]. The surface conditioning of these carbon materials for supercapacitor fabrication is of substantial importance for achieving the best performance, such as good specific surface area, conductivity, and minimum selfdischarge rates. Carbons with high specific surface area have many oxygen functional groups, such as ketone, phenolic, carbonyl, carboxylic, hydroquinoid, and lactone groups, introduced during the activation procedure for enlarging the surface area [31, 35]. These oxygen functional groups on activated carbons or activated carbon fibers give rise to one kind of electrochemical reactivity, oxidation or reduction. Oxidation or

629 reduction of the redox functional groups shows pseudocapacitance, which amounts to about 5–10% of the total realizable capacitance [31]. However, the various surface functionalities in activated carbons are one of the factors that increase the internal resistance (equivalent series resistance; ESR) due to the redox reaction [31]. Activated carbons are much cheaper than metal oxides and conducting polymers and they have much larger specific surface area than the others. Activated carbon-based supercapacitors have been commercialized for small memory backup devices. However, activated carbons show lower conductivity than metal oxides and conducting polymers, resulting in a large ESR, which gives smaller power density. In addition, the observed specific capacitances of the carbon-based supercapacitors are about one-fourth the theoretical capacitance in spite of their high specific surface area (2000–3000 m2 /g), which is attributed to the existence of micropores [31, 35]. This is a weak point of active carbons as electrode materials in supercapacitors with high energy density and power density.

5. CARBON NANOTUBE-BASED SUPERCAPACITORS During the last decades, the application of activated carbons as the electrode materials in supercapacitors has been intensively investigated because of their high specific surface area and relatively low cost [31, 35, 36, 66–69]. Since the specific capacitance of a supercapacitor is proportional to the specific surface area, the effective surface area of the electrode materials is important [31, 35, 36]. Theoretically, the higher the specific surface area of an activated carbon, the higher the specific capacitance should be. Unfortunately, the theoretical capacitance of the activated carbons is not in good agreement with the observed value, because a significant part of the surface area remains in the micropores (C O, enhance the capacitance of the CNT electrode. However, the IR drop at the initial stage of discharging is large due to the Faradic reactions of the functional groups, giving rise to a large ESR. The discharge profile is also different from the typical profile of electric double-layer capacitors. The discharging slope of V /t is not linear due to the Faradic reaction by the functional groups. Frackowiak et al. [15] investigated the electrochemical characteristics of supercapacitors built from multiwalled CNT electrodes with a specific surface area of 430 m2 /g in 1 M KOH aqueous solution and the correlation of the microtexture and elemental composition of materials. They argued that the presence of mesopores due to the central

Carbon Nanotube-Based Supercapacitors

canal and/or entanglement of CNTs is the origin of an easy accessibility of the ions to the electrode/electrolyte interface for charging the electrical double layer. They detected pure electrostatic attraction of ions as well as quick pseudoFaradic reactions upon varying surface functionality, which was induced during acidic oxidation. The values of the specific capacitance varied from 4 to 135 F/g, depending on the type of nanotubes or/and their posttreatments (acidic oxidation). They demonstrated that even with a moderate specific surface area below 470 m2 /g, due to their accessible mesopores, the multiwalled CNT electrodes represent attractive materials for supercapacitors as compared to the best activated carbons. Zhang et al. [17] studied a supercapacitor using multiwalled CNT electrodes in an organic electrolyte system. The multiwalled CNT electrodes showed a specific surface area of approximately 100 m2 /g and a measured specific capacitance up to 18.2 F/g (16.6 F/cm3 with 1 M LiClO4 in a mixture of ethylene carbonate and propylene carbonate (1  1 in volume ratio) as an organic electrolyte solution. The measured specific capacitance is relatively lower than that of other groups using aqueous electrolytes, due to the low specific surface area of the multiwalled CNT electrodes and the organic electrolyte solution system. However, the energy density of the supercapacitor can reach up to 20 W-h/kg with a 10-mA discharge current density, due to the high operation voltage in the organic electrolyte solution. The relative volume of mesopores and macropores of the used electrodes exceeds about 92%, and the micropores are nearly negligible. For organic electrolytes, because of their larger molecular structure, only the mesopores and macropores of their electrode pores are accessible and are much larger than those of activated carbons. Raymundo-Piñero et al. [70] also studied supercapacitors using multiwalled CNTs with a high surface area prepared by KOH activation. Micropores of pure multiwalled CNTs have been highly developed using chemical KOH activation. Depending on the nanotubular material, the burn-off ranged from 20 to 45% after the activation process. The surface area was at least doubled with maximum values of 1050 m2 /g for a treatment condition of KOH  CNT in a ratio of 4  1. The activated multiwalled CNTs still possess a nanotubular morphology with many defects on the outer walls, giving a significant increase in the micropore volume, while keeping a noticeable mesoporosity. The high efficiency of KOH activation in the development of the surface area is attributed to the redox reactions between carbon nanotubes and KOH followed by potassium intercalation and separation of the graphitic layers. This process is very efficient for the formation of pores, especially for the multiwalled CNTs with wellgraphitized walls, while preserving nanotubular morphology. The creation of defects on the nanotube walls developed microporosity, which is responsible for excellent capacitance behavior. Such activated multiwalled CNTs have been used for the electrode materials of supercapacitors in alkaline, acidic, and aprotic medium. Enhanced values of capacitance are always observed after activation: In some cases, the capacitance increased almost 7 times from 15 F/g (for nonactivated nanotubes) to 90 F/g (after chemical activation) in alkaline medium (6 M KOH). In an organic electrolyte, the capacitance of activated multiwalled CNTs reached 65 F/g

Carbon Nanotube-Based Supercapacitors

631

with a boxlike shape characteristic of cyclic voltammetric behavior. Such a significant modification of surface microporosity for essentially mesoporous nanotubular materials is of great interest for many electrochemical applications. Frackowiak et al. [19] also studied supercapacitors comparatively for multiwalled and single-walled CNT electrodes. They obtained a maximum specific capacitance of 40 F/g using single-walled CNTs. This value is twice as small as that for multiwalled CNTs, which have a maximum specific capacitance of 80 F/g. They argued that the low specific capacitance of the single-walled CNT electrodes is attributed to the perfect bundle structure of the single-walled CNTs. An et al. [20, 21, 27] reported on the key factors determining the performance of supercapacitors using single-walled CNT electrodes, which were synthesized by direct-current (dc) arc discharge. The single-walled CNTs consisted of bundles with diameters of 10–20 nm. The sample purity was roughly estimated to be about 30% [71]. A mixture of the single-walled CNTs and poly-vinylidene chloride or polyvinyl alcohol used as a binder was molded and then carbonized at 500–1000  C for 30 min in an argon gas ambient. The thickness of the electrode is about 150 m. The measured apparent mass density of the electrode is 0.75 g/cm3 . They obtained a maximum specific capacitance of 180 F/g and a measured power density of 20 kW/kg at an energy density of 7 W-h/kg in a solution of 7.5 N KOH. They demonstrated that heat treatment at high temperature is necessary to increase the capacitance and reduce the CNT electrode resistance. The increased capacitance is due to the enhancement of the specific surface area and the abundant pore distribution at lower pore sizes of 30–50 Å estimated from the BET(N2 measurements. They also found that most of the BET surface area of the electrode contributes to the theoretically estimated specific capacitance, unlike activated carbons. They also claimed that minimization of the contact resistance is independent of the specific capacitance but directly related to maximization of the power density. The performance of several types of supercapacitors using various carbon nanotubes reported in recent years is summarized in Table 3. Table 3 shows that the types of CNTs, the CNT morphologies, and the electrode preparation conditions are critically important in determining the specific

capacitance, the specific power density, and the specific energy density of supercapacitors. Although many of them are still to be determined in order to obtain the best performance for real applications, there is plenty of room to improve the performance of supercapacitors using CNT electrodes.

5.2. Carbon Nanotube–Composite Electrodes Even if the CNT electrodes exhibit a unique pore structure and high usage efficiency of specific surface areas and some processes for disintegrating the bundle structure of the CNTs may enhance the utilization efficiency of their surface area, there is a limitation to widening the effective surface area of the CNTs. The specific capacitance of supercapacitors using CNT electrodes is still smaller than that of supercapacitors using activated carbons and activated carbon fibers (∼250 F/g), which are commonly used as electrode materials for supercapacitors. Therefore, it is necessary to improve the SWNTs for applications that require a high capacitance. The small specific energy density of supercapacitors is a serious drawback to their application in practical devices, and, consequently, the main interest in supercapacitors is their ability to deliver high specific capacitance and energy density. In the energy storage by supercapacitors, there are two fundamentally different approaches, as discussed in the previous section. A hybrid of an electric double-layer system and a pseudo-Faradic system could be a candidate for a supercapacitor with high specific capacitance and energy density. Currently, there is considerable interest in studying a new type of nanocomposite using CNTs and conducting polymers to improve the conductivity, electronic transport, and electromagnetic properties of organic conducting polymers [72–79]. These nanocomposites are regarded as one of the fascinating candidates for applications in nanoelectronic element and electro-optical devices. Recently, Jurewicz et al. [18] demonstrated that the composite electrode based on multiwalled carbon nanotubes (MWNTs) coated with polypyrrole (Ppy) can be used in supercapacitors. They realized that a homogeneous layer of Ppy could be deposited on the nanotubular materials by

Table 3. Summary of the performance of many types supercapacitors using various carbon nanotubes reported in recent years.

Type of CNT

Treatment of CNT

Surface area of CNT (m2 /g)

Type of binder

Type of electrolyte 38wt% H2 SO4 38 wt% H2 SO4 1 M KOH 1M LiClO4/ EC + PC 6 M KOH 6 M KOH 7.5 M KOH

MWNT

Nitric acid

430

—

MWNT

120

MWNT MWNT

Sulfuric/nitric acid carbonization Nitric acid Nitric acid

Phenolic resin PVDF PTFE

SWNT MWNT SWNT

— KOH activation Carbonization

475 ∼100 — 1050 360

PVDF PVDF PVdC

Maximum capacitance (F/g)

Power density (kW/kg)

Energy density (W-h/kg)

Ref.

102

>8

—

[12]

27

—

—

[14]

135 18

— —

— 20

[15] [17]

40 95 180

— — >20

— — 7

[19] [70] [20]

Note: MWNT, multiwalled carbon nanotube; SWNT, single-walled carbon nanotube; PVDF, polyvinylidene fluoride; PTFE, polytetrafluoroethylene; PVdC, polyvinylidene chloride; EC, ethylene carbonate; PC, propylene carbonate.

Carbon Nanotube-Based Supercapacitors

632 electrochemical polymerization of pyrrole. The composite electrodes in all cases enhanced the specific capacitance. They obtained a maximum value of 163 F/g for MWNTs prepared at 600  C and modified by a Ppy layer of 5 nm, whereas it is only 50 F/g for the pristine nanotubes. It has been proposed that the nanocomposite electrode favors the formation of a three-dimensional electrical double layer, allowing a more effective contribution to the pseudo-Faradic properties of Ppy. Frackowiak et al. [19] also reported on composite electrodes based on MWNTs deposited with Ppy for supercapacitors, although they prepared the nanocomposite electrodes of MWNTs and Ppy by chemical polymerization of pyrrole using an oxidant in an acidic solution. The value of the specific capacitance obtained from MWNT electrodes modified by Ppy has reached 172 F/g, about twice that given either by pristine MWNTs (ca. 80 F/g) or by pure Ppy (ca. 90 F/g). They also observed pseudocapacitance and speculated the origin of pseudocapacitance as the surface functionalities, metallic particles, or conducting polymers. Lee and co-workers [80, 81] introduced SWNT–Ppy nanocomposite electrodes to improve the specific capacitance of the supercapacitor by combining the electric double layer and the redox reaction. The individual nanotubes and nanoparticles are uniformly coated with Ppy by in-situ chemical polymerization of pyrrole. In comparison to the pure Ppy and the SWNTs, the SWNT–Ppy nanocomposite electrodes exhibit a specific capacitance that is 5–10 times higher. They show that this large specific capacitance of the SWNT–Ppy nanocomposite is due to the uniformly coated Ppy on the SWNTs, increasing the active sites on Ppy chains. Lee and co-workers emphasized that the pseudocapacitance by the redox reaction and the capacitance by the electric double layer are simultaneously enhanced by the enlarged surface area of Ppy. They also studied the effects of the conducting agent added in the nanocomposite electrodes on the specific capacitance and the internal resistance of supercapacitors. They obtained a maximum specific capacitance of 265 F/g from the SWNT–Ppy nanocomposite electrodes containing 15 wt% of the conducting agent. They argued that the addition of a conducting agent into the SWNT–Ppy nanocomposite electrodes gives rise to the increase in the specific capacitance by reducing the internal resistance of the supercapacitor.

6. CONCLUSIONS Some of the important challenges in this world are the conversion, storage, and distribution of energy. These are clearly connected with the development of several key technologies such as transportation communications, and electronics. The environmental problems and economical aspects related to the development and use of electrochemical energy storage devices are of significance. Air pollution and abusive use of the earth’s natural resources are of major concern. If only 10% of vehicles are electrically powered, petroleum consumption will be cut by 700 million barrels per year, the reduction in expenses would be 14 billion at $20 per barrel, and air pollution would be reduced by 1011 kg of CO2 [82]. Therefore, new applications and development of supercapacitors are directly related to technologies of an

electric vehicle (EV) or a hybrid electric vehicle (HEV). The supercapacitor in an EV or an HEV will serve as a shortterm energy storage device with high power capability and will allow us to store the energy obtained from regenerative braking. This energy will be reused in the next acceleration and boost. It also allow us to reduce the size of the primary power sources [batteries (EV), internal combustion engine (HEV), fuel cell] and keep them running at an optimized operation point [31, 35]. High-power supercapacitors for an EV or an HEV will require a high working voltage of 100– 300 V with low resistance and high energy density by the series and parallel connections of elemental capacitors. Note also that homogeneous performance in each supercapacitor unit is essential in such a system [31, 35]. The remaining problems in the technology of supercapacitors are connected with the engineering of high-rate production lines for the fabrication of very well matched units for stacking in multicell and high-voltage devices and the production of reliably sealed, low-ESR, bipolar, stacked devices. The lowered ESR for high-power supercapacitors strongly depends on the manufacturing technology. However, the energy density of the supercapacitors is mainly governed by the properties of the electrode materials. Therefore, the development of novel electrode materials will accelerate the supercapacitor having an energy density as high as a secondary battery. Another future application is the microsupercapacitor for the power source of micro (or nano) electromechanical systems (MEMS or NEMS). In recent years, MEMS (or NEMS) technologies have attracted attention worldwide for many applications, including medical devices, communications equipment, sensors, and actuators. There are many technical problems with the development of these types of microdevices. One of the most important challenges is to develop the optimal micropower source to operate these devices. In many cases, the MEMS (or NEMS) requires extremely low current and power. This may be possible by using microsupercapacitors as the power sources for these devices.

GLOSSARY Battery A device that stores electrical energy using electrochemical cells. Chemical reactions occur spontaneously at the electrodes when they are connected through an external circuit, producing an electrical current. Capacitor An electrical device that stores electricity or electrical energy. It has three essential parts: two electrical conductors, which are usually metal plates, separated and insulated by the third part, called the dielectric. The plates are charged with equal amounts of positive and negative electrical charges, respectively. Carbon nanotubes Graphitic layers rolled into the form of a hollow cylinder whose diameter is an order of nanometer scale. Cell A device that converts chemical energy into electrical energy or vice versa when a chemical reaction occurs in the cell. Typically, it consists of two metal electrodes immersed into an electrolyte with electrode reactions at the electrode– solution surfaces.

Carbon Nanotube-Based Supercapacitors

Electric double layer Development of a net charge at the particle surface affects the distribution of ions in the surrounding interfacial region, resulting in an increased concentration of counterions (ions of opposite charge to that of the particle) close to the surface. Electrodes The two electronically conducting parts of an electrochemical cell. Electrolyte A chemical compound (salt, acid, or base) that dissociates into electrically charged ions when dissolved in a solvent. The resulting electrolyte solution is an ionic conductor of electricity. Equivalent series resistance The single resistor that draws the same current as the combination of constituent resistors. Faradic redox process An oxidation–reduction reaction that occurs in an electrochemical cell. The essential feature is that the simultaneous oxidation–reduction reactions are spatially separated. Polarizable electrode An electrode that is easily polarizable. That is, the potential of the electrode will change significantly from its equilibrium potential with the application of even a small current density. Pseudocapacitance The capacitance due to the Faradic redox process in a supercapacitor. Separator A thin structural material (usually a sheet) used to separate the electrodes of a devided electrochemical cell into two or more compartments. Supercapacitor A high-energy version of a conventional electrolytic capacitor holding hundreds of times more energy per unit volume or mass than a normal capacitor.

REFERENCES 1. F. Leroux, K. Metenier, S. Gautier, E. Frackowiak, S. Bonnamy, and F. Beguin, J. Power Sources 81–82, 317 (1999). 2. G. T. Wu, C. S. Wang, X. B. Zhang, H. S. Yang, Z. F. Qi, P. M. He, and W. Z. Li, J. Electrochem. Soc. 146, 1696 (1999). 3. E. Frackowiak, S. Gautier, H. Gaucher, S. Bonnamy, and F. Beguin, Carbon 37, 61 (1999). 4. E. Jouguelet, C. Mathis, and P. Petit, Chem. Phys. Lett. 318, 561 (2000). 5. B. Gao, A. Kleinhammes, X. P. Tang, C. Bower, L. Fleming, Y. Wu, and O. Zhou, Chem. Phys. Lett. 307, 153 (1999). 6. B. Gao, C. Bower, J. D. Lorentzen, L. Fleming, A. Kleinhammes, X. P. Tang, L. E. McNeil, Y. Wu, and O. Zhou, Chem. Phys. Lett. 327, 69 (2000). 7. C. Nutzenadel, A. Zuttel, D. Chartouni, and L. Schlapbach, Electrochem. Solid-State Lett. 2, 30 (1999). 8. S. M. Lee, K. S. Park, Y. C. Choi, Y. S. Park, J. M. Bok, D. J. Bae, K. S. Nahm, Y. G. Choi, S. C. Yu, N. G. Kim, T. Frauenheim, and Y. H. Lee, Synth. Met. 113, 209 (2000). 9. A. K. M. Fazle Kibria, Y. H. Mo, K. S. Park, K. S. Nahm, and M. H. Yun, Int. J. Hydrogen Energy 26, 823 (2001). 10. B. Rohland, K. Eberle, J. Scholta, and J. Garche, Electrochim. Acta 43, 3841 (1998). 11. N. Muradov, Catal. Comun. 2, 89 (2001). 12. C. Niu, E. K. Sichel, R. Hoch, D. Moy, and H. Tennent, Appl. Phys. Lett. 70, (1997) 1480. 13. L. Diederich, E. Barborini, P. Piseri, A. Podesta, P. Milani, A. Schneuwly, and R. Gallay, Appl. Phys. Lett. 75, 2662 (1999). 14. R. Z. Ma, J. Liang, B. Q. Wei, B. Zhang, C. L. Xu, and D. H. Wu, J. Power Sources 84, 126 (1999).

633 15. E. Frackowiak, K. Metenier, V. Bertagna, and F. Beguin, Appl. Phys. Lett. 77, 2421 (2000). 16. E. Frackowiak and F. Beguin, Carbon 39, 937 (2001). 17. B. Zhang, J. Liang, C. L. Xu, B. Q. Wei, D. B. Ruan, and D. H. Wu, Mater. Lett. 51, 539 (2001). 18. K. Jurewicz, S. Delpeux, V. Bertagna, F. Beguin, and E. Frackowiak, Chem. Phys. Lett. 347, 36 (2001). 19. E. Frackowiak, K. Jurewicz, S. Delpeux, and F. Beguin, J. Power Sources 97–98, 822 (2001). 20. K. H. An, W. S. Kim, Y. S. Park, Y. C. Choi, S. M. Lee, D. C. Chung, D. J. Bae, S. C. Lim, and Y. H. Lee, Adv. Mater. 13, 497 (2001). 21. K. H. An, W. S. Kim, Y. S. Park, J.-M. Moon, D. J. Bae, S. C. Lim, Y. S. Lee, and Y. H. Lee, Adv. Funct. Mater. 11, 387 (2001). 22. C. Journet, W. K. Maser, P. Bernier, A. Loiseau, M. L. de la Chapelle, S. Lefrant, P. Deniard, R. Lee, and J. E. Fischer, Nature London 388, 756 (1997). 23. A. Thess, R. Lee, P. Nikolaev, H. Dai, P. Petit, J. Robert, C. Xu, Y. H. Lee, S. G. Kim, A. G. Rinzler, D. T. Colbert, G. E. Scuseria, D. Tomanek, J. E. Fisher, and R. E. Smalley, Science 273, 483 (1996). 24. H. M. Cheng, F. Li, G. Su, H. Y. Pan, L. L. He, X. Sun, and M. S. Dresselhaus, Appl. Phys. Lett. 72, 3282 (1998). 25. J. N. Barisci, G. G. Wallace, and R. H. Baughman, J. Electrochem. Soc. 147, 4580 (2000). 26. C.-Y. Liu, A. J. Bard, F. Wudl, I. Weitz, and J. R. Heath, Electrochem. Solid-State Lett. 2, 577 (1999). 27. K. H. An, K. K. Jeon, W. S. Kim, Y. S. Park, D. J. Bae, S. C. Lim, and Y. H. Lee, J. Korean Phys. Soc. 39, S511 (2001). 28. A. Peigney, Ch. Laurent, E. Flahaut, R. R. Bacsa, and A. Rousset, Carbon 39, 507 (2001). 29. R. R. Bacsa, Ch. Laurent, A. Peigney, W. S. Bacsa, Th. Vaugien, and A. Rousset, Chem. Phys. Lett. 323, 566 (2000). 30. M. Eswaramoorthy, R. Sen, and C. N. R. Rao, Chem. Phys. Lett. 304, 207 (1999). 31. B. E. Conway, “Electrochemical Supercapacitors: Scientific Fundamentals and Technological Applications.” Kluwer Academic/Plenum, New York, 1999. 32. S. T. Mayer, R. W. Pekala, and J. L. Kaschmitter, J. Electrochem. Soc. 140, 446 (1993). 33. A. Yoshida, S. Nonaka, I. Aoki, and A. Nishino, J. Power Sources 60, 213 (1996). 34. Y. Kibi, T. Saito, M. Kurata, J. Tabuchi, and A. Ochi, J. Power Sources 60, 225 (1996). 35. R. Kotz and M. Carlen, Electrochim. Acta 45, 2483 (2000). 36. M. Endo, T. Takeda, Y. J. Kim, K. Koshiba, and K. Ishii, Carbon Sci. 1, 117 (2001). 37. H. E. Becker, U.S. Patent 2,800,616 (to General Electric), 1957. 38. D. I. Boos, U.S. Patent 3,536,963 (to Standard Oil, SOHIO), 1970. 39. B. E. Conway and H. A. Kozlowska, Acc. Chem. Res. 14, 49 (1981). 40. S. Hadzi-Jordanov, H. A. Kozlowska, M. Vukovic, and B. E. Conway, J. Electrochem. Soc. 60, 1471 (1978). 41. S. Trasatti and G. Buzzanca, J. Electroanal. 29, 1 (1971). 42. R. Galizzioli, F. Tantardini, and S. Trasatti, J. Appl. Electrochem. 4, 57 (1974). 43. S. Gottesfeld, A. Redondo, and S. W. Feldberg, J. Electrochem. Soc. 134, 271 (1987). 44. A. Rudge, I. Raistrick, S. Gottesfeld, and J. P. Ferraris, Electrochem. Soc. 39, 273 (1994). 45. A. Rudge, J. Davey, I. Raistrick, and S. Gottesfeld, J. Power Sources 47, 89 (1994). 46. B. E. Conway, J. O’M. Bockris, and I. A. Ammar, Trans. Faraday Soc. 47, 756 (1951). 47. G. Hambitzer, K. Pinkwart, C. Ripp, and C. Schiller, in “Handbook of Barrey Materials” (M. Wakihara and O. Yamamotu, Eds.), Chap. 1. Wiley–VCH, Weinheim, 1999. 48. G. W. Vinal, “Storage Battery.” Wiley, New York, 1955.

634 49. A. Burke, J. Power Sources 91, 37 (2000). 50. B. E. Conway, V. Birss, and J. Wojtowicz, J. Power Sources 66, 1 (1997). 51. J. B. Goodenough, in “Progress in Solid-State Chemistry” (N. Reiss, Ed.), Vol. 5. Pergamon, New York, 1971. 52. W. D. Ryden, A. W. Lawson, and C. C. Sartain, Phys. Lett. A 26, 209 (1968). 53. H. Schafer, G. Schneidereit, and W. Gerhardt, Z. Anorg. Allg. Chem. 319, 372 (1963). 54. S. Trasatti and G. Lodi, in “Conductive Metal Oxides” (S. Trasatti, Ed.), Vol. A, p. 338. Elsevier, Amsterdam, 1980. 55. S. Ardizzone, G. Fregonara, and S. Trasatti, Electrochim. Acta 35, 263 (1990). 56. R. Kotz and S. Stucki, Electrochim. Acta 31, 1311 (1986). 57. J. P. Zheng, P. J. Cygan, and T. R. Jow, J. Electrochem. Soc. 142, 2699 (1995). 58. T. J. Guther, R. Oesten, J. Garche, F. M. Delnick, D. Ingersoll, X. Andrieu, and K. Naoi, “Electrochemical Capacitors II, Proceedings of the Electrochemical Society,” 1997, Vol. 96-25, p. 16. 59. T.-C. Liu, W. G. Pell, B. E. Conway, and S. L. Roberson, J. Electrochem. Soc. 145, 1882 (1998). 60. Y. Sato, K. Tanuma, T. Takayama, K. Kobayakawa, T. Kawai, and A. Yokoyama, J. Power Sources 97–98, 165 (2001). 61. W. Lu and D. D. L. Chung, Carbon 39, 493 (2001). 62. Y. Kida, K. Yanagida, A. Funahashi, T. Nohma, and I. Yonezu, J. Power Sources 94, 74 (2001). 63. S. Lj. Gojkovic and T. R. Vidakovic, Electrochim. Acta 47, 633 (2001). 64. A. L. Sauvet and J. Fouletier, J. Power Sources 101, 259 (2001). 65. X. Wang, I.-M. Hsing, and P. L. Yue, J. Power Sources 96, 282 (2001). 66. J. Gamby, P. L. Taberna, P. Simon, J. F. Fauvarque, and M. Chesneau, J. Power Sources 101, 109 (2001).

Carbon Nanotube-Based Supercapacitors 67. D. Qu and H. Shi, J. Power Sources 74, 99 (1998). 68. Y. Kibi, T. Saito, M. Kurata, J. Tabuchi, and A. Ochi, J. Power Sources 60, 219 (1996). 69. H. Shi, Electrochim. Acta 41, 1633 (1996). 70. E. Raymundo-Piñero, D. Carzorla-Amorós, A. Linares-Solano, S. Delpeuk, E. Frackowiak, K. Szostak, and F. Béguin, Carbon 40, 1597 (2002). 71. J.-M. Moon, K. H. An, Y. H. Lee, Y. S. Park, D. J. Bae, and G.-S. Park, J. Phys. Chem. B 105, 5677 (2001). 72. A. B. Kaiser, G. U. Flanagan, D. M. Stewart, and D. Beaglehole, Synth. Met. 117, 67 (2001). 73. B. H. Chang, Z. Q. Liu, L. F. Sun, D. S. Tang, W. Y. Zhou, G. Wang, L. X. Qian, S. S. Xie, J. H. Fen, and M. X. Wan, J. Low Temp. Phys. 119, 41 (2000). 74. S. Lefrant, I. Baltog, M. Lamy de la Chapelle, M. Baibarac, G. Louarn, C. Journet, and P. Bernier, Synth. Met. 100, 13 (1999). 75. K. Yoshino, H. Kajii, H. Araki, T. Sonoda, H. Take, and S. Lee, Fullerene Sci. Technol. 7, 695 (1999). 76. H. S. Woo, R. Czerw, S. Webster, D. L. Carroll, J. W. Park, and J. H. Lee, Synth. Met. 116, 369 (2001). 77. G. Z. Chen, M. S. P. Shaffer, D. Coleby, G. Dixon, W. Zhou, D. J. Fray, and A. H. Windle, Adv. Mater. 12, 522 (2000). 78. J. Fan, M. Wan, D. Zhu, B. Chang, Z. Pan, and S. Xie, Synth. Met. 102, 1266 (1999). 79. J. Fan, M. Wan, D. Zhu, B. Chang, Z. Pan, and S. Xie, J. Appl. Polym. Sci. 74, 2605 (1999). 80. K. H. An, K. K. Jeon, J. K. Heo, S. C. Lim, D. J. Bae, and Y. H. Lee, J. Electrochem. Soc. 149, A1058 (2002). 81. Y. H. Lee, K. H. An, S. C. Lim, W. S. Kim, H. J. Jeong, W. Lee, and J. M. Kim, New Diamond Frontier Carbon Technol. 14, 209 (2002). 82. R. M. Dell and D. A. J. Rand, J. Power Sources 100, 2 (2001).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Carbon Nanotubes in Composite Materials Ch. Laurent, A. Peigney Université Paul-Sabatier, Toulouse, France

CONTENTS 1. 2. 3. 4. 5.

Introduction Carbon Nanotube–Metal–Matrix Composites Carbon Nanotube–Ceramic–Matrix Composites Carbon Nanotube–Polymer–Matrix Composites Conclusions Glossary References

1. INTRODUCTION 1.1. Carbon Nanotube Discovery and Description Although hollow carbon fibers have been prepared for several decades, it is the report by Iijima [1] in 1991 on obtaining of carbon tubes with a diameter in the nanometer range, the so-called carbon nanotubes (CNTs), and on their relations to the recently discovered fullerenes that initiated a worldwide research effort devoted to improving their synthesis, determining their structure, and calculating and measuring their physical properties. Rolling a graphene sheet onto itself so that the open edges match exactly produces a hollow cylinder which is termed a single-walled CNT (SWNT). Two such concentric cylinders form a double-walled CNT (DWNT). A CNT with more walls is named a multiwalled CNT (MWNT). To close the tips, half fullerenes or more complex structures that include pentagons are required. However, the cylindrical part of the CNT, which contains only hexagons, is considered infinite in most studies. There are several ways to roll the graphene sheet, which can be described by a helical vector (or chiral vector, or roll-up vector) represented by a pair of integers (n m) [2, 3] as described in Figure 1 [4]. The limiting cases are referred to as zizzag (n 0) and armchair (n n) SWNTs, with a helical (or chiral) angle equal to 0 and 30
, respectively. Other SWNTs have a helical (often termed chiral) structure. As mentioned in a later section, the helicity has significant implications on the electronic properties of CNTs. For MWNTs, the measured interlayer distance (about

ISBN: 1-58883-057-8/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

0.34 nm) is close to that measured between graphene sheets in graphite and no particular correlation appears between the helicity of concentric layers. SWNTs have a diameter in the range 0.4–6 nm and a length reaching several tens or even hundreds of micrometers, resulting in a very high aspect ratio: indeed, the length/diameter ratio could be in the range 1000–100,000, although the upper limit is difficult to evaluate because the length is difficult or impossible to measure. Another notable characteristic of CNTs is the high specific surface area (1315 m2 /g considering the outer surface of infinite SWNTs) which is mostly dependent on the number of walls and, in a lesser way, on the diameter [5]. The hollow structure of the CNTs is of course also interesting because it offers a nanometric cavity which can be filled with various materials. More on the fundamental relations governing the geometry of CNTs can be found in the early review by Dresselhaus et al. [6] which also highlights the implications of symmetry on the vibrational and electronic structure of onedimensional carbon. The following sections will very briefly describe the synthesis and the general properties of the CNTs.

1.2. Synthesis The first report on the preparation of filamentous carbon by the decomposition of carbonaceous gases dates back to the 19th century [7]. Rodriguez [8] reviewed the considerable amount of work done on the topic since then. Although very small hollow nanofilaments were prepared by catalytic chemical vapor deposition (CCVD) in the 1970s [9], Iijima’s report [1] and the subsequent development of the field triggered an enormous amount of research aiming at the synthesis of CNTs. We will only briefly describe here the formation mechanisms of the CNTs, the main synthesis methods (arc discharge, laser vaporization, and CCVD), and the characteristics of the so-prepared CNTs. Rakov [10] published a review, with 282 references, on the topic. It is important to note something which was not fully appreciated at the beginning of the development of the “nanotube” field, that is, that each method produces different CNTs samples, that differences subsist depending on the experimental conditions for

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 1: Pages (635–653)

636

Carbon Nanotubes in Composite Materials

a1 Armchair tubule [4,4]

a2

Ch

Zigzag tubule [7,0] ma2

C na1 θ

Chiral tubule [4,3] two-dimensional graphene sheet

Figure 1. The helical (chiral) vector OC or Ch = na1 + ma2 is defined on the hexagonal lattice of carbon atoms by unit vectors a1 and a2 and the helical (chiral) angle  with respect to the zigzag axis. Reprinted with permission from [4], K. T. Lau and D. Hui, Composites B 33, 263 (2002). © 2002, Elsevier Science.

a given method, and that even different batches from a single source may not be exactly identical. However, the field is evolving extremely quickly. The formation mechanisms of the SWNTs (and DWNTs) and that of the MWNTs are basically different (see [11] for an early review). Arc-discharge MWNTs are prepared without the use of a metal catalyst, so mechanisms involving the incorporation of carbon atoms (C1, C2, C3) and thickening by lateral growth of graphene planes on existing surfaces were proposed [12–16]. Typically, the outer diameter of arcdischarge MWNTs ranges between 2 and 20 nm and the inner diameter ranges between 1 and 3 nm [3, 15, 17, 18]. The length is of the order of one to several micrometers. MWNTs form regularly organized bundles, which may contain tens or hundreds of individual tubes. The bundles in turn are combined into fibers with a diameter of about 50 m, which themselves form threads with a diameter of about 1 mm. It was noted that the arc-discharge MWNTs contain many defects [19–21]. When using CCVD, relatively large metal particles (about 10–30 nm) can produce MWNTs with a matching diameter by mechanisms established long ago for hollow filaments [9, 22, 23], resembling the vapor– liquid–solid (VLS) mechanism [24]. For SWNTs, two cases are observed: one particle produces either one CNT or many CNTs, depending on its size and other factors. Accordingly, two formation mechanisms, also bearing a certain relation with VLS, were proposed. The “one particle–many CNTs” mechanism [25] is dominant when using arc discharge or laser vaporization. A large number of SWNTs bundles originate from the surface of a metal particle several tens of nm in diameter [25–27]. All observed SWNTs are narrow (0.8–2 nm). A narrow catalyst particle distribution centered on 15 nm appears to give the highest SWNT yields [28]. The diameter distribution of arc-discharge SWNTs is very narrow, generally in the range 0.7–1.6 nm [14, 29, 30]. However, various carbon species

are found in the deposits including SWNTs, MWNTs, amorphous carbon, spherical metallic nanoparticles, empty or filled graphitic nanoparticles, a few graphitic sheets, as well as round spherical metallic particles. The SWNTs are mostly organized in bundles consisting of a few to a few tens of individual tubes forming a triangular lattice and more or less covered with some carbon soot. Lengths reach up to several micrometers. It is claimed that laser vaporization [31] allows far greater control over growth conditions and produces SWNTs in higher yield (70–90%) and of better quality. The SWNTs are about 1–1.6 nm in diameter, with a very narrow distribution, and are arranged in large bundles that may be 100 m long. No amorphous coating is observed. The diameter distribution can be slightly changed and SWNTs close to 6 nm in diameter were reported [32, 33]. In the “one particle–one CNT” case, observed when using CCVD, the diameter of the SWNTs is in the range 0.5–5 nm and is directly related to that of the catalyst particle. Hafner et al. [34] calculated a critical diameter of 3 nm above which the formation of concentric carbon layers around the particle is favored, in good agreement with experimental results [34, 35]. The so-called “yarmulke” mechanism [36] allows for a second cap (and thus a second CNT) to form underneath the first one. This is also observed experimentally [34]. CCVD specimens may contain several species in addition to Iijima-type CNTs, including amorphous carbon, carbon particles, and carbidic carbon particles, but tremendous progress on the selectivity has been made in recent years. Several groups have reported progresses in the yield, control of the diameter, and the number of walls [34–49]. The CCVD SWNTs are quite flexible and in general contain more defects than those produced by arc discharge and laser vaporization. Note that CCVD methods are becoming increasingly attractive due to their great potential for the production of large quantities of CNTs at a low cost. Length is in the range between ca. 100 nm and several micrometers (or tens of micrometers) and the surface is free of the amorphous carbon coating typically observed in CCVD MWNTs, which generally are even longer. Purification procedures are required to eliminate the residual metal catalyst and the nontubular carbon species. Several techniques, usually involving thermal and acidic treatments, have been proposed. The sensitivity of SWNTs to chemical processing was investigated by transmission electron microscopy (TEM) [50]. The chemistry of CNTs, which encompasses the aspects of purification and functionalization, is a very important field for the development of future applications. Rakov [51] published a review with 573 references on the topic. Chemical functionalization was also reviewed by Sinnott [52]. Several methods for attachment of oxygen-containing groups to the CNT surfaces have been developed [53, 54]. Carboxylic, carbonyl, and hydroxyl groups were found covering the CNT walls [55]. The concentration of acid groups on the surfaces of HNO3 treated CNTs was higher than that found for graphite treated under the same conditions [56]. Sonication of CNTs prior to the functionalization in acids increases the concentration of acid groups. Oxidation of CNTs with a H2 SO4 –HNO3 mixture also leads to a higher concentration of functional groups on the surface [57]. Long-term acid treatment of CNTs also induces modifications. Functional groups can be removed

637

Carbon Nanotubes in Composite Materials

from the surface of CNTs by heating. It should be noted that SWNTs obtained by different methods exhibit different behavior in the course of acid treatment and subsequent oxidation.

1.3. General Properties This section will very briefly highlight some general properties of CNTs which are treated in detail in several books [58–60]. A complete coverage of the electrical, thermal, and mechanical characteristics of the different kinds of CNTs (SWNTs, DWNTs, MWNTs, and bundles) is indeed far beyond the scope of this chapter. The electrical characteristics of CNTs have been investigated by theoretical and experimental works [2, 3, 61– 65] which have brought to light some remarkable features related to the unique one-dimensional nature of the CNTs. A review on the electronic properties of CNTs is published in this Encyclopedia [66]. The most important characteristic to note for use of the CNTs in composites is that either metallic or semiconducting behavior may be observed depending on the helicity. The electrical conductivity of a metallic CNT could reach 10,000 S/cm and that of a semiconducting CNT is in the range 0.1–100 S/cm. The thermal conductivity of CNTs was addressed both theoretically [67–69] and experimentally [70–72]. Calculations showed that the thermal conductivity exhibits a peaking behavior as a function of temperature and that it strongly depends on CNT diameter but does not or weakly depends on helicity. The extrapolation of measurements on compressed mats of CNTs bundles gave values of thermal conductivity ranging from 1750 to 5850 Wm−1 K−1 , to be compared with the highest known thermal conductivity, that of the in-plane value of graphite (3000 Wm−1 K−1 ). Reviews on the mechanical properties of CNTs were published by several authors [73–75]. A low density of defects confers excellent mechanical properties to CNTs. Furthermore, the intrinsic strength of the carbon–carbon sp2 bond is expected to give CNTs the highest strength and modulus among all existing whiskers. Theoretical [76–78] and experimental studies have shown that CNTs have excellent elastic properties. Treacy et al. [79] first investigated the elastic modulus (Young’s modulus) of MWNTs using thermal vibration experiments performed in a transmission electron microscope. The average value, obtained over 11 isolated MWNTs, was 1.8 TPa. Atomic force microscopy (AFM) was used to perform a direct measurement of the stiffness and strength of individual MWNTs [80]. An elastic modulus of 1.26 TPa was obtained and the average bending strength measured was 14.2 ± 8 GPa. Salvetat et al. [81] measured the properties of SWNT bundles with the AFM. The axial and shear moduli decrease significantly upon an incsrease of the bundle diameter, suggesting slipping of the SWNTs within the bundle. A yield strength of 45 ± 7 GPa was calculated for SWNT bundles from experimental strain measurements on the AFM, with an elastic modulus taken at 1.25 TPa [82]. The Poisson ratio is defined by the variation of the radius of the SWNT resulting from longitudinal deformations along the axis. Ab initio calculations indicate that it retains graphitic value (0.17) except for a possible slight reduction for small radii and shows a helicity dependence [74].

The deformation behavior of CNTs under stress is also remarkable. Several studies [83–86] have shown that both thin- and thick-walled CNTs exhibit compressive strengths one order of magnitude higher than any other known fiber. Moreover, the flexibility of CNTs is remarkable [87, 88] and the bending may be fully reversible up to a critical angle value as large as 110
for SWNTs [89]. However, this leads to the formation of kinks and highly strained regions. The flexibility arises from the ability of the sp2 C–C bonds to reversibly change hybridization when deformed out of plane. Tensile-load experiments on MWNTs ropes found a tensile strength and modulus of 3.6 and 450 GPa, respectively [90]. Other tension tests performed on MWNT [91] and SWNT ropes [92] reported tensile strengths in the range of 11– 63 GPa, in reasonable agreement with the theoretical value of about 100 GPa and no apparent dependence on the outer wall diameter. Thostensson et al. [73] noted that 63 GPa is a value 50 to 60 times greater than that of high-quality steels. In the case of MWNTs, the failure of the outermost tube occurred followed by pullout of the inner CNTs. The effectiveness of load transfer to the inner walls or within a bundle and its influence on stiffness is still under debate and was notably studied on DWNTs [67, 93, 94]. A study of the radial compressibility of CNTs by AFM [95] showed that at low pressure, the modulus is similar to that of graphite along the c-axis (ca. 10 GPa), while at higher pressure, repulsion forces from the more deformed sides increase the elastic modulus. The collapse of CNTs into nanoribbons has been observed [96], which could account for the competition between the van der Waals attraction and elastic energy. Regarding the use of CNTs in composites, it is important to note [74] that strengthening but also toughening are major concerns in structural elements involving a brittle matrix. For example, fibers longer than the critical length and a strong interfacial bond are required for fiber–polymer composites with a high fracture strength, whereas a weaker interface will promote a high toughness through fiber-matrix debonding and pullout. With so many unique or exceptional characteristics, the potential applications of CNTs are plentiful, notably as a component of composite materials. Applications in car body panels [97], car and truck bumpers, bullet-proof jackets, the construction of earthquake-proof buildings, support of a human Martian mission, electromagnetic shielding, waveguides, etc., have already been envisioned [51]. In the following sections, the works reported on composite materials containing CNTs will be addressed, successively reviewing materials with metal, ceramic, and polymer matrix. Earlier reviews on CNT composites were published by Thostenson et al. [73] and Lau and Hui [4]. Note that materials consisting of “simply” filled or coated CNTs will not be considered here. A review on CNT filling was published by Monthioux [98].

2. CARBON NANOTUBE– METAL–MATRIX COMPOSITES A few papers report the fabrication of CNT–metal composites by powder metallurgy techniques and some results on the stability of CNTs toward the solid metal matrix at high

638 temperatures or on the influence of CNTs on the electrical and mechanical properties of the materials. Xu et al. [99] studied the electrical properties of MWNT– Al composites. The MWNTs (CCVD, diameter 30 nm) were ground with an aluminum powder and the resulting mixtures were hot-pressed at 793 K. In the obtained material, MWNTs are located mainly at Al grain boundaries and some aluminum carbides were formed, which was attributed to a reaction between amorphous carbon and Al. The electrical resistivity at room temperature increases slightly (3.4– 5.5  cm) with increasing MWNT volume fraction. From room temperature down to 80 K all the composites show the typical metallic decrease of the electrical resistivity. At about 80 K their resistivity abruptly drops by more than 90%. Kuzumaki et al. [100] also investigated MWNT–Al composites using unpurified arc-discharge MWNTs. The composites (5 or 10 vol% MWNTs) were produced by hot-pressing at 873 K and hot-extrusion at 773 K. The MWNTs were found undamaged and, on the contrary to conventional carbon fiber/systems, no aluminum carbides were detected at the interface with the matrix, even after a heat treament at 983 K. The tensile strength and elongation of the composites are only slightly affected by annealing at 873 K in contrast to those of aluminum. The same group [101] also studied MWNT–Ti composites using similar samples of MWNTs (diameter 5–50 nm), containing graphitic particles and amorphous carbon. The powders were mixed and hot-pressed at 1208 K. The relative density of a pure Ti specimen prepared as a 13 reference was 98.6% and that of the composite was 95.8%. The formation of TiC was observed but TEM observations showed that the MWNTs themselves did not react with the titanium matrix. Probably, TiC comes from the reaction between amorphous carbon and the matrix. The Young’s modulus of the composite is about 1.7 times that of pure Ti. This result may be caused by both TiC formation and the addition of MWNTs. The composite has about 5.5 times the Vickers’ hardness of pure Ti. This is associated with the suppression of coarsening of the Ti grains, TiC formation, and addition of MWNTs with an extremely high Young’s modulus. It is proposed that the MWNTs play an important role in blocking migration of dislocations. Dong et al. [102] prepared CCVD MWNT–Cu composites by similar techniques. The hardness of the composites increases while the coefficient of friction and wear loss both decrease with the increase in the MWNT fraction (0– 25 vol%). Rolling tests reveal that the composites can reach a deformation of 50–60% and have a good isotropy of the mechanical properties. The structure and superconductivity of CCVD MWNT– MgB2 prepared by mixing and heat treatment (950
C, 2 h, Ar) were reported by Wei et al. [103]. Carbon atoms coming from the MWNTs do substitute for boron in the MgB2 lattice. Depending on the MWNT content and on the extent of substitution, the composites are either semiconductive or metallic. Thus, these first works have shown the stability of MWNTs in some solid metal environments during the hotpressing or hot-extrusion processes. But the homogeneity of the dispersion of CNTs has to be improved and the use of

Carbon Nanotubes in Composite Materials

SWNTs has to be tested. The study of more optimized materials will probably confirm the beneficial influence of CNTs on some mechanical properties of metal–matrix composites. Li et al. [104] prepared a MWNT–metallic glass Fe82 P18 composite by rapid solidification of a pressed and melted mixture of MWNTs (CCVD, 3 wt%), Fe–P, and Fe powders. The MWNTs, some of which are filled with an iron alloy, are well dispersed in the so-obtained composite ribbons (40 m thick). Thus, it seems that techniques involving molten metals can also be considered to fabricate CNT– metal composites. Coatings with CNTs are also studied with the aim to obtain excellent tribological properties. Chen et al. [105] reported the electrodeposition of CCVD MWNTs and nickel on a carbon steel substrate. The MWNT content in the deposit was increased upon the increase to a maximum value of a number of factors including the concentration of the MWNTs in the solution, the current density, and the agitation rate. Ball-milling the MWNTs to decrease their length was found to be beneficial to increase the maximum concentration in the solution. Observation of fracture surfaces shows that the MWNTs appear well dispersed in the nickel matrix, with one end embedded into it and the other end protruding from the surface. CCVD MWNT–Ni–P composite coatings were prepared by electroless deposition [106–108]. These coatings exhibit a high wear resistance and a low friction coefficient compared to SiC–Ni–P and graphite–Ni–P coatings.

3. CARBON NANOTUBE– CERAMIC–MATRIX COMPOSITES Studies on CNT–ceramic composites using oxide, carbide, and carbon matrices have been reported and usually involve mixing raw or purified CNT samples with the matrix or a corresponding precursor before a densification step. An alternative route involves composite powders with in-situ grown CNTs which are very homogeneously dispersed. Ma et al. [109] prepared MWNT–SiC composites (10 wt% MWNTs) by hot-pressing mixtures of CCVD MWNTs (diameter 30–40 nm) and nano-SiC powders. A densification of 94.7% was achieved of by hot-pressing at 2000
C and using 1 wt% B4 C as a sintering aid. It is claimed that the presence of the MWNTs provides an increase of about 10% of both the bending strength and fracture toughness, notably by crack deflection and MWNT debonding. The application of SWNTs for thermal management in ceramics was reported [110]. Purified laser SWNTs were mixed into a slurry containing partially stabilized zirconia. Samples (1% SWNTs) with thickness over 200 m were prepared by tape casting. Thermal treatments were performed for binder burnout and sintering. Although the SWNTs do not survive the final treatment, they remain intact long enough to serve as templates to create a SWNT-based microstructure. However, the thermal conductivity of these composites is higher than that of the pure matrix, which is opposed to what was aimed for. This could be a consequence of the presence of residual metal catalyst particles. In contrast, the dispersion of 1 wt% vapor-grown carbon fibers (VGCF), which withstand the sintering, provokes a decrease of about 30% of the thermal conductivity.

Carbon Nanotubes in Composite Materials

MWNTs were included in a TiO2 matrix by a sol–gel method [111]. CCVD MWNTs (diameter 15 nm) covered by a thin layer of amorphous carbon were dispersed by various methods into a TiO2 -precursor sol. Films were prepared by dip coating and were annealed in air at different temperatures. The MWNTs are preserved for treatments at or below 400
C and TEM observation reveals that they appear to act as preferential sites for the crystallization of rutile-TiO2 nanocrystals. The porosity of the films is about 20%. The effect of CNT addition on the tribological behavior of carbon/carbon composites was reported [112]. Slurries of MWNTs (CCVD, 12–15 walls) were infiltrated into the composites (0–20 wt% CNTs) which were then carbonized in N2 . The surface of the final composite is covered with a CNT/carbon layer. The friction coefficient increases and the wear loss decreases upon the increase in CNT loading, so that the wear resistance is doubled with 20 wt% CNTs compared to the 0 wt% material. Andrews et al. [113] reported the dispersion of purified laser-ablation SWNTs in isotropic petroleum pitch matrices to form composite carbon fibers with a diameter of about 18 m. SWNTs loadings of 8 and 10 wt% were found to be too high to form a fiber. The tensile strength, elastic modulus, and electrical conductivity of a composite fiber (5 wt% SWNTs) are increased by about 90%, 150%, and 340% respectively, as compared to the corresponding values for unmodified isotropic pitch fibers. Hwang and Hwang [114] prepared SiO2 glass rods of micrometer size by using surfactant–MWNT (arc-discharge) co-micelles as templates. The rods were mixed with SiO2 and dense composites (6 wt%) MWNTs were prepared. It is claimed that the Vickers hardness of the MWNTs/SiO2 rods–SiO2 materials is twice that of SiO2 rods–SiO2 composites but it is unclear whether this is due to the MWNTs themselves or to the change in rod morphology they provide. Arc-discharge SWNT–SiO2 glass composites were prepared using a sol–gel technique [115]. The purified and shortened (300–500 nm) SWNTs were functionalized with dendron methyl 3,5-di(methyltriglycoloxy)benzylic alcohol prior to ultrasonic blending in the sol phase and gelification. Heating the glasses to 600
C produced materials with 80% densification and a good optical transparency, revealing the homogenous dispersion of the SWNTs (0.1–1.0 wt%). Nonlinear optical transmission was observed. The synthesis of MWNT/sol–gel composites [116] was also performed by adding various silane precursor sols (methyl-, ethyl-, and propyltrimethoxysilane) to CCVD MWNTs and mechanically mixing to ensure a uniform composition. The mixture was packed and allowed to dry for 24–48 h depending on the amount of sol. The final composites contained 34, 44, or 61 wt% of MWNTs. However, the composition of the matrix at this stage is unclear. The authors note that negatively charged silanol groups may be present. SEM observations reveal that the composites are quite porous. The development of electrochemical devices was investigated. It was found that varying the sol and the amount of MWNTs allowed one to obtain a double-layer capacitance on a wide range (0.2–100 mF/cm2 ). A high capacitance is required for energy-storage applications while a low capacitance is required for electrochemical sensors. Fast electron-transfer kinetics are observed and it is proposed that the so-called

639 selectivity of the composite (followed by the shift of the oxidation peak of ascorbic acid on cyclic voltammograms) could be controlled. The present authors have reported a novel catalytic route for the in-situ formation of SWNTs and thin MWNTs in a composite powder by reduction in H2 –CH4 of an aluminabased solid solution [42]. From the beginning, one of the aims was to further use such composite powders to produce dense composites. Incidentally, to the best of our knowledge, this was the first time that the formation of SWNTs by CCVD of a hydrocarbon was reported. With the aim to favor the formation of tubular carbon (notably SWNTs and DWNTs) with respect to other carbon species (fibers, hollow or filled particles, disordered carbon), several parameters related to the material were studied: allotropic form of the solid solution [117], iron content [118], specific surface area [119]. Parameters related to the reduction treatment such as the composition of the H2 –CH4 atmosphere [49, 120] and the temperature [35] were also studied in detail. The method was expanded to MgAl2 O4 -based solid solutions in order to be able to use divalent cations and therefore compare the catalytic efficiency of Fe, Co, and Ni [121, 122] and their binary alloys [123]. Solid solutions in the form of foams rather than powders were prepared and the formation of SWNTs from such foams was reported for the first time [124]. The method was further expanded to MgO-based solid solutions [43, 44, 125–127]. The CNTs can be extracted without damage from the CNT–metal–MgO composite powders by a mild treatment (soaking in dilute HCl at room temperature). They thus can be used for dispersion in other matrices such as epoxy as mentioned in a following section. The specific surface area of the so-obtained CNTs specimens is very high (over 800 m2 /g) [44], in good agreement [5] with the distributions of the number of walls and diameter. The CNTs being grown in-situ in the powders are thus very homogeneously dispersed between the metal–oxide grains (Fig. 2) in a way that seems very difficult to achieve by methods involving mechanical mixing. The CNTs gather in extensively branched bundles, forming a weblike network between the matrix agglomerates. The bundles are always smaller than 100 nm in diameter, they appear to be very flexible, and some have been traced for 100 m. Several studies have dealt with the preparation of dense CNT–Fe–Al2 O3 , CNT–Co–MgO, and CNT–FeCo–MgAl2 O4 composites by hot-pressing [128–131] and hot-extrusion [132] and on the determination of some of their mechanical and electrical characteristics. The precise CNT content in the dense materials is difficult to determine, so the carbon content in the starting composite powders (2–10 wt%) and a unique parameter derived from specific surface area measurements [5, 42, 128] representing the quantity of CNTs were used to compare the different specimens. The early studies [128–130] showed that in comparison to similar carbon-free nanocomposites, the relative densities of the CNT–Fe–Al2 O3 composites are lower, the matrix grains are smaller, and the fracture strength and fracture toughness are generally markedly lower. Nevertheless, scanning electron microscopy (SEM) observations of composite fractures indicated that the CNT bundles locally act to dissipate some of the fracture energy. The too-low volume fraction of CNTs

640

Carbon Nanotubes in Composite Materials

(a)

1 µm

(b)

1 µm

Figure 2. Scanning electron micrographs of CNT–Fe–Al2 O3 composite powders with micronic (a) and submicronic (b) matrix grains. Long bundles of CNTs are observed.

and the presence of others species, notably large diameter (>15 nm) filamentous carbon, could explain the lack of mechanical reinforcement at the macroscopic scale. As mentioned, works on the powder synthesis allowed one to increase both the quantity of CNTs and the quality of carbon (much less nanofibers) in the CNT–Fe–Al2 O3 powders and to prepare novel composite powders, CNT– Fe/Co–MgAl2 O4 and CNT–Co–MgO. Hot-pressed composites were investigated [131]. For the sake of comparison, ceramics (MgAl2 O4 and MgO) and metal–oxide composites

(i.e., without CNTs) were prepared in the same conditions. With the Al2 O3 matrix, the increase of the quantity of CNTs in the powder leads to a refinement of the microstructure of the hot-pressed specimen. With both the Al2 O3 and the MgAl2 O4 matrix, a fraction of the CNTs seems to be destroyed during hot-pressing at 1500
C (in a vacuum). When using the MgO matrix, most CNTs are destroyed during a hot-pressing at 1600
C, but the CNTs are not damaged if the treatment is limited to 1200
C. It seems that the quantity of CNTs retained in the massive composite is more dependent on the treatment temperature than the nature of the oxide matrix. CNT damage produces disordered graphene layers which gather at matrix grain junctions. Probably due to a too-low relative density (87–93%), the fracture strength (measured by three-point bending) and the fracture toughness (measure by the single-etched notched beam technique) of the CNT-containing composites (Tables 1 and 2) are generally lower than those of the carbon-free metal–oxide composites and only marginally higher than those of the ceramics. SEM observations (Fig. 3) showed that some CNTs are trapped inside the matrix grains or at grain boundaries and seem to be wetted by the matrix in the case of alumina. Most of these CNTs are cut near the fracture surface after some pull-out (Fig. 4) and could contribute to a mechanical reinforcement. However, this is not demonstrated at a macroscopic scale. It was shown that CNTs confer an electrical conductivity to ceramic–matrix composites. Indeed, whereas the corresponding ceramics and metal–oxide nanocomposites are insulators, the CNT–metal–oxide composites are electrical conductors (electrical conductivity in the range 0.2– 4.0 S cm−1 ) due to the percolation of the CNTs. The values of the electrical conductivity are fairly well correlated to the relative quantity of CNTs, the specimens becoming insulators when the CNTs are destroyed. Recently [133], it was found that the percolation threshold is below 1 vol% of CNT and that controlling the CNT quantity in the starting powders allows one to control the electrical conductivity in the range 0.01–10 S cm−1 . Peigney et al. [132] reported that extrusion at high temperatures (in a vacuum) allows one to align CNTs in ceramic–matrix bulk nanocomposites, as opposed to fibers or thin films. The superplastic forming is made easier by the CNTs, which inhibit the matrix grain growth and also

Table 1. Relative density and microstructural characteristics of the hot-pressed ceramics and nanocomposites. Ceramic or composite

Label

d %

Gm m

dmetal m

dZrO2 m

CNT–Fe–Al2 O3 CNT–Fe–Al2 O∗3 MgO Co–MgO CNT–Co–MgO MgAl2 O∗4 Fe/Co–MgAl2 O∗4 CNT–Fe/Co–MgAl2 O∗4

CMA1 CMA2 B MB CMB D MD CMD

887 875 901 966 929 997 982 906

03 03 5 7 3 13 08 05

≤05 ≤05 — ≤2 ≤15 — ≤05 ≤05

— ≤1 — — — ≤1 ≤1 ≤1

Note: d: relative density calculated by assuming that all carbon has the density of graphite, with dgraphite = 225 g/cm3 ; Gm : average grain size of the oxide; dmetal : diameter of the larger metal particles; dZrO2 : diameter of the larger ZrO2 particles in specimens prepared from powders attrition-milled before reduction (∗). Source: Reprinted with permission from [131], E. Flahaut et al., Acta Mater. 48, 3803 (2000). © 2000, Elsevier Science.

641

Carbon Nanotubes in Composite Materials Table 2. Electrical conductivity (), fracture strength f ), fracture toughness (KIc ), and fracture mode of the hot-pressed ceramics and nanocomposites. Ceramic or composite

Label or reference

 (S cm−1 )

f (MPa)

KIc (MPa m1/2 )

Al2 O3 Fe–Al2 O3 CNT–Fe–Al2 O3 CNT–Fe–Al2 Oa3 CNT–Fe–Al2 O3 CNT–Fe–Al2 Oa3 MgO Co–MgO CNT–Co–MgO MgAl2 Oa4 Fe/Co–MgAl2 Oa4 CNT–Fe/Co–MgAl2 Oa4

[34] [34] [17] [18] CMA1 CMA2 B MB CMB D MD CMD

— — — — 0.4–0.8 2.8–4.0 n.m. n.m. 0.2 n.m. n.m. 1.5–1.8

335 630 540 295 400 296 202 283 254 308 212 221

44 72 36 27 50 31 — — — — 294 171

Fracture mode intergranular mixed mixed intergranular mixed intergranular intergranular mixed intergranular transgranular mixed mixed

Note: n.m.: not measurable; mixed: the fracture presents both the inter- and transgranular characters. Source: Reprinted with permission from [131], E. Flahaut et al., Acta Mater. 48, 3803 (2000). © 2000, Elsevier Science. a Specimens prepared from powders attrition-milled before reduction.

act as a lubricating agent. The CNTs withstand the extreme shear stresses occurring during the extrusion up to 1500
C. In addition to electron microscopy revealing the alignment, the composites show an anisotropy of the electrical conductivity [about 30 times larger in the longitudinal direction (20 S cm−1 ) than in the transverse direction], which could be adjusted by controlling the amount of CNTs (carbon content in the range 0.6–8 wt%). Siegel et al. [134] mixed arc-discharge MWNTs with a nanometric alumina powder in ethanol dispersion using an ultrasonic probe. The mixture was dried in two steps, first in an ultrasonic bath and then in an oven. Hot-pressing was performed in argon at 1300
C. A single composite was prepared (10 vol% MWNTs). It is claimed that the fracture toughness (derived from Vickers microhardness

(a)

(b)

200 nm

100 nm

(c)

100 nm

Figure 3. Scanning electron micrographs of the fractures of CNT–Fe– Al2 O3 composites. Reprinted with permission from [130], A. Peigney et al., Ceram. Int. 26, 677 (2000). © 2000, Elsevier Science.

measurements) is increased by 24% (to 4.2 MPa m1/2 ). However, basic materials characteristics such as the density and grain size are not given, making it difficult to consider this result as evidence for the beneficial role of the MWNTs themselves. Zhan et al. [135] prepared SWNT–alumina composites using small-diameter SWNT bundles obtained by the HIPCo technique and a nanometric alumina powder (50 nm). The two components were mixed by ball-milling, which produced a reasonably homogeneous dispersion without damaging the SWNTs. Fully dense composites were obtained by the spark plasma sintering (SPS) technique, in which the specimen is annealed at lower temperatures (1150
C) and during much lower times (3 min) than using other sintering processes. SPS does not damage the SWNTs. The bundles can be seen to be located mainly at the boundaries of the alumina grains and have good contact with the alumina matrix. The authors report a very large gain in fracture toughness (derived from Vickers indentation) in alumina composites containing 10 vol% SWNT (from 3.7 MPa m1/2 for pure Al2 O3 to 9.7 MPa m1/2 ). Moreover, one of these SWNT– Al2 O3 composites which was not fully densified (about 14% of residual porosity) presented a quite remarkable fracture toughness—about twice that of the fully densified but unreinforced alumina matrix. Sun et al. [136] reported the development of a colloidal process to coat CNTs with alumina prior to another mixing step with a concentrated alumina suspension, so that the final material contains only 0.1 wt% CNTs. SPS was used to fully densify (at 1300
C) the composite. The Vickers hardness is slightly higher than that of pure alumina (17.6 versus 16.9 GPa) and the Vickers indentation-derived toughness is increased from 3.7 to 4.9 MPa m1/2 upon the addition of 0.1 wt% CNTs. It is proposed that coating the CNTs before sintering results in an improved bonding with the matrix. Crack bridging and CNT pullout could also lead to the increase in fracture toughness. These studies [135, 136] have thus demonstrated that CNT–ceramic composites warrant further investigation with respect to the mechanical properties [137]. For example, the preparation of larger specimens to test the fracture strength would be of interest [137].

642

Carbon Nanotubes in Composite Materials

(a)

100 nm

(b)

100 nm

MWNT–lanthanum cobaltate composites were tested for a possible application as an oxygen electrode in rechargeable zinc/air batteries [138]. La1−x Mx CoO3 (M = Ca Sr) particles were coated with a citrate precursor containing more cobaltate with a slight excess of cobalt oxide. The mixture was treated in acetylene (700
C) to produce MWNTs (diameter 20–30 nm, length about 20 m) grown on the oxide surface. An electrode was manufactured and promising results were obtained. The CNTs formed directly at the oxide surface could improve the electrical junction between the oxide and the carbon-containing substrate. MWNTs (arc discharge, 3 wt%) were embedded into Bi2 Sr2 CaCu2 O8+ (a Bi-2212 superconductor) in order to improve pinning properties [139]. The insulator– supraconductor transition is slightly lower in the composite than in the Bi-2212 material (87 vs 95 K). The increase in current densities shows that the MWNTs function like columnar defects produced by heavy-ion irradiation.

4. CARBON NANOTUBE– POLYMER–MATRIX COMPOSITES

h (c)

(d)

h

100 nm

100 nm

(e)

100 nm

(f)

100 nm

Figure 4. FEG-SEM images of the fracture of the CNT–Fe–Al2 O3 composite showing some aspects of the CNT–matrix interactions. Reprinted with permission from [131], E. Flahaut et al., Acta Mater. 48, 3803 (2000) © 2000, Elsevier Science.

CNT–polymer composites are by far the most studied. In the following sections, we will successively review composites with epoxy, polymethylmethacrylate, and other matrices. Lordi and Yao [140] noted that three key issues affect the performance of a fiber–polymer composite: the strength of the fiber phase, the toughness of the fiber phase, and good interfacial bonding which is crucial for load transfer to occur and fiber orientation. These authors [140] performed calculations related to the interfacial adhesion in CNT–polymer composites. It is shown that interfacial binding energies are only slightly correlated with adhesion in actual composites, as determined by comparison with experimental observations of fiber pullout by various authors. It is also reported that hydroxy side-groups, and to a lesser extent phenyl sidegroups, on the polymer seem advantageous for strong interfaces. The key factor in forming a strong bond the CNTs and the matrix lies in the polymer’s ability to form largediameter helices around individual CNTs. The strength of the interface may result from molecular-level entanglement of the two phases and forced long-range ordering of the polymer. In addition, it is proposed that weak frictional interactions between layers of MWNTs and between SWNTs in bundles indicate that isolated SWNTs are desired for dispersion in a matrix. Schadler et al. [141] noted that a high interfacial shear stress between the fiber and the matrix will transfer the applied load to the fiber over a short distance, whereas a low interfacial shear stress will require a long distance. The first main mechanism of load transfer is micromechanical interlocking, which could be difficult with CNTs due to their atomically smooth surface. Other mechanisms include chemical bonding and van der Waals bonding between the CNTs and the matrix.

4.1. Epoxy–Matrix Composites The first work on a CNT–polymer composite was reported by Ajayan et al. in 1994 [142]. Purified arc-discharge MWNTs were dispersed in a liquid-epoxy base resin by mechanical stirring. After evacuation to remove trapped air

643

Carbon Nanotubes in Composite Materials

bubbles, pastes of the material were hardened by heating above 6
C for 2 hours. Thin slices were cut and observed by TEM. It was demonstrated that, during cutting, the majority of MWNTs were straightened in the strain direction rather than broken. These results suggested, first, that MWNTs have excellent mechanical properties and second that anisotropy can be induced by a flow in such a material. Only several years later, in 1998, were other works published on CNT–polymer composites. Schadler et al. [141] reported the dispersion of 5 wt% MWNTs in an epoxy resin by an ultrasonic treatment. Although the MWNTs were well separated, they remain poorly distributed. Mechanical tests and associated Raman spectroscopy revealed that the load transfers to the MWNTs were much higher in compression than in tension. It was inferred that, under loading, only the outer layers of the MWNTs are stressed in tension because all the inner tubes are sliding within the outer, whereas all the layers are stressed in compression. Further investigations on the load transfer in such materials were reported. Both laser-ablation SWNTs and arcdischarge MWNTs were used. It was found [85] that the compressive strength of thin- and thick-walled CNTs is more than two orders of magnitude higher than the compressive strength of any known material, which is close to 0.05 GPa. Compressive stresses were induced in composites films, 200 m thick, from the shrinkage of the matrix due to polymerization and by further quenching from room temperature to low temperatures (223–81 K) [143, 144]. The frequency of some Raman bands being stress-sensitive, it has been demonstrated that this spectroscopy can be used to study the strain distribution in fiber composites and to calculate the Young’s modulus [145]. The obtained values were close to those previously reported for SWNTs and MWNTs, about 5 and 1.8 TPa, respectively [79, 146]. The fracture of a SWNT–epoxy composite, observed in real time by TEM, revealed a good polymer–CNT wetting and significant CNT–polymer adhesion [147]. Raman spectroscopy investigations of SWNT–epoxy and MWNT–epoxy composites [148] allowed one to determine the effective modulus of the CNTs (about 1 TPa and 0.3 MPa for the SWNTs and MWNTs, respectively). Stress transfer between the different phases was demonstrated. The formation of damage doublets in adjacent MWNTs was reported in microtomed thin slices of arc-discharge MWNT–epoxy composite films [149] (Fig. 5). It was concluded that despite the enormous difference in scale, fundamental concepts pertaining to continuum mechanics of fiber composites such as “stress concentration factor” and “effective matrix radius” could be valid at the nanometric level, at least to some degree. Interactions between CNTs should also be taken into account. Mechanical tests conducted on 5 wt% SWNT–epoxy composites [150] showed that SWNT bundles were pulled out of the matrix during the deformation of the material. In some areas, the SWNTs appeared broken and relaxed but in other regions, they were fully stretched and bridging a crack (Fig. 6). On the basis of Raman spectroscopy results and SEM observations, the authors inferred that the load transfer is limited because the SWNTs are slipping, under tension

(a)

(b)

60 nm

48 nm

Figure 5. (a) TEM image of aligned MWNTs close to the film surface, revealing CNT fragmentation and the formation of a doublet of adjacent breaks (bottom arrow). A doublet consisting of a CNT end and a nanotube break is also shown (top arrow). (b) TEM picture of two aligned MWNTs in close contact, revealing the formation of two doublets of adjacent breaks (arrows). Reprinted with permission from [149], O. Lourie and H. D. Wagner, Comp. Sci. Tech. 59, 975 (1999) © 1999, Elsevier Science.

strain, within the bundles. It was also reported [151] that the combination of standard mechanical tests and Raman scattering tests allows the determination of residual strains due to matrix shrinkage, the elastic properties of embedded SWNTs/bundles, their dispersion, and load transfer effectiveness. The influence of the interfacial interaction between CNTs and the epoxy matrix was evidenced by Gong et al. [152]. Dynamic mechanical measurements and SEM observations were performed on 1 wt% CNT–epoxy composites, prepared either with or without a nonionic surfactant addition (polyoxyethylene 8 lauryl). For the former composite, the glass transition temperature was increased from 63 to 88
C and the elastic modulus was also increased by more than 30% in comparison with the matrix. In contrast, the latter composite, prepared without surfactant addition, presents only moderated improvements of these properties. SEM observations confirmed that the surfactant addition improved the dispersion of the CNTs, thus allowing the dramatic effects of CNTs on the properties of the material.

644

Carbon Nanotubes in Composite Materials

(a)

Conductivity (Sm-1)

10-1

a

10-3

b 10-5

c 10-7

(b)

10-9 0.01

0.1

1

Carbon loading (vol%)

Figure 7. Comparative log–log plot of the conductivity of nanocomposites containing: (a) CCVD CNTs, (b) carbon black with copper-chloride, and (c) carbon black only, as a function of the filler volume content. Lines are included as visual aids. Reprinted with permission from [154], J. Sandler et al., Polymer 40, 5967 (1999). © 1999, Elsevier Science.

Figure 6. (a) SEM image of SWNT bundles stretched across cracks observed in a CNT–carbon composite pellet for certain critical crack widths. (b) Curved bundles of SWNTs that are no longer under load, as observed in a SEM image. The scale bars in the images correspond to 1 m. Reprinted with permission from [150], P. M. Ajayan et al., Adv. Mater. 12, 750 (2000). © 2000, Wiley–VCH.

Allaoui et al. [153] claimed to have quadrupled the elastic modulus and the yield strength (measured through tensile test) of an epoxy matrix by the incorporation of 4 wt% of both very large CCVD MWNTs and VGCFs (15–400 nm in diameter). Conductivity measurements revealed that the insulator–conductor transition occurred for MWNT contents in the range 0.5–1 wt%. CCVD MWNTs (inner diameter 5 nm, outer diameter 10 nm, length a few micrometers) were dispersed in an epoxy matrix (0.0225–0.15 wt% MWNTs) [154]. The exposure to ultrasound early in the process and the subsequent intense stirring of the resin dramatically improved the dispersion of the MWNTs in the matrix. The percolation threshold was found to be between 0.0225 and 0.04 wt% MWNTs. An electrical conductivity of about 10−2 S/m, sufficient for antistatic applications, was achieved with 0.04 wt% MWNTs, that is, using a loading about one-tenth of that needed using carbon black (Fig. 7). It is noted that at these low filler fractions, neither the processing behavior of the matrix nor the surface finish of the samples is adversely affected and that the mechanical properties of the epoxy matrix should not be compromised. Using CCVD CNTs (80% SWNTs or DWNTs) prepared by the solid solution route as mentioned in a previous section, Barrau et al. [155] prepared CNT–epoxy composites,

with CNT fractions in the range 0.04–2.5 wt%. An analysis of the dc and ac electrical conductivity in the temperature range 20–110
C and frequency range 10−2 Hz–1 MHz was reported. There is an increase in dc conductivity of 10 orders of magnitude between 0.2 and 0.6 wt% CNTs, the highest conductivity being 10−4 S/cm for 2.5 wt%. A percolation threshold was found at 0.3 wt% CNTs independent of the temperature. The critical exponent  of the scaling law of the percolation theory  ∝ f − fc  increases with the temperature (171 ± 020 at 100
C) and tends toward the value (1.94) predicted by the tridimensional percolation theory of randomly distributed objects. Indications that tunneling conduction may be the main conduction mechanism are presented. The frequency dependence of the conductivity obeys the universal dynamic response. The behavior of the frequency exponent with the temperature is consistent with the correlated barrier hopping mechanism. CCVD SWNTs were used to augment the thermal transport properties of industrial epoxy [156]. Samples loaded with 1 wt% unpurified SWNTs (containing 15–25 wt% residual Fe catalyst) showed an increase in thermal conductivity of 70% and 125% at 40 K and at room temperature, respectively. Electrical conductivity data showed a percolation threshold in the range 0.1–0.2 wt% SWNTs. The Vickers hardness rose monotonically with SWNT loading up to a factor of 3.5 at 2 wt%. It is noted that these results suggest that the thermal and mechanical properties of SWNT–epoxy composites are improved without the need to chemically functionalize the SWNTs. Thostenson et al. [157] reported the direct growth of MWNTs on carbon fibers by CCVD. Embedding the coated fibers in an epoxy matrix produces a multiscale composite. Single-fiber composites were prepared. It is reported that the MWNT coating improves the interfacial load transfer, possibly by local stiffening of the polymer matrix near the fiber/matrix interface.

645

Carbon Nanotubes in Composite Materials

4.2. Polymethylmethacrylate–Matrix Composites Unpurified and purified laser-ablation SWNTs were dispersed and aligned in polymethylmethacrylate (PMMA) using a combination of solvent mixing and melt mixing, a method expected to be applicable to most thermoplastic polymers [158]. Composite films (1–8 wt% SWNTs) showed a higher electrical conductivity along the flow direction than perpendicular to it. The conductivity increases with SWNTs loading. In addition, composite fibers with well aligned SWNTs were prepared by melt spinning. The highest draw ratio was achieved for pure PMMA and specimens with 1% SWNTs. However, the fiber diameter is not uniform along the axis and specimens with the higher SWNT contents tend to fracture during the melt-spinning process, revealing a much increased viscosity of the melt. The weight fraction of SWNTs necessary for percolation increases as the SWNTs are aligned. Unpurified arc-discharge MWNTs (diameter 30 nm) meltblended with PMMA were compressed (8–9 MPa, 210
C, 5 min) to produce transparent films [159]. Different contents (4–30 wt%) of MWNTs were used and it is claimed that they are well dispersed in the matrix with no apparent damage. Thermogravimetric analysis (TGA) shows that, compared to pure PMMA, the thermal degradation occurs at a slightly higher temperature when 26 wt% MWNTs are added. The dynamic mechanical behavior was investigated, showing that the storage modulus (E ′ ) is increased upon the incorporation of MWNTs, particularly at high temperatures (120–140
C) (Fig. 8). It was, however, thought necessary to improve the wetting between the MWNTs and the PMMA, so the MWNTs were mixed and therefore coated with poly(vinylidene fluoride) prior to melt blending with PMMA [160]. The presence of a small amount of poly(vinylidene fluoride) indeed results in an increased storage modulus, but this effect diminishes at higher temperatures. Jia et al. [161] mixed purified CCVD MWNTs with a free radical initiator (2,2′ -azobisisobutyrontrile) and the

storage modulus (MPa)

1000

100

pure PMMA 4% MWNT 9% MWNT

10

11% MWNT 17% MWNT 26% MWNT 1 40

80

120

160

temperature (degree)

Figure 8. Storage moduli of PMMA and various composites. Reprinted with permission from [159], Z. Jin et al., Chem. Phys. Lett. 337, 43 (2001). © 2001, Elsevier Science.

monomer MMA to obtain MWNT–PMMA composites by an in-situ process. Two kinds of MWNTs were used: as-purified specimens and ball-milled purified samples. The ball-milling treatment produces much shorter and better separated MWNTs. It is claimed that the MWNTs take part in the polymerization reaction, obstructing the growth of PMMA. C–C bonds are formed between the MWNTs and PMMA results in a strong interface. The tensile strength, toughness, hardness, and deflection temperature of the ballmilled MWNT–PMMA composites are improved compared to pure PMMA up to 7 wt% MWNTs. A higher loading is reported to generate residual stresses in the matrix which causes the composite to become very brittle. In contrast the as-purified MWNT–PMMA composites show weaker mechanical properties because of a poor dispersion of the MWNTs. MWNT–PMMA (arc-discharge MWNTs, diameter 10– 15 nm, length 2–3 m) and SWNT–PMMA composites (laser-ablation SWNTs, diameter 1.4–2 nm, length 200–400 m, in bundles) were prepared using a dry powder mixing method [162]. The final step of the preparation was to extrude the materials in order to align the CNTs. The efficiency of the method is demonstrated by SEM, TEM, and Knoop hardness analyses. In particular, the Knoop hardness is at a maximum when the long diagonal of the indenter is perpendicular to the orientation of the CNTs. The tensile modulus was almost insensitive to the presence of the CNTs while the impact strength was significantly increased with only 0.1 wt% of SWNTs. It is proposed that this apparent increase in toughness is due to a weak interfacial adhesion, to the high flexibility of the SWNTs, and to the pullout and sliding of individual SWNTs within a bundle. The characterization by Raman spectroscopy of SWNT– PMMA composite films (2.5–20 wt% SWNTs) prepared by spin casting was reported [163, 164]. The SWNTs, prepared by different methods (arc discharge, laser ablation, and solar energy), are organized in bundles. In the case of arc discharge, composites were prepared with SWNT samples from different parts of the so-called collaret. Comparing the Raman spectra allowed one to get a diameter distribution of the SWNTs and to confirm the presence of armchair SWNTs. Is was shown that the diameter distribution slightly changes from one collaret sample to the other, whereas spectra from composites with laser ablation or solar energy SWNTs are rather similar. Furthermore, for low SWNT loadings, the amorphous carbon formed during the synthesis of the SWNTs was found to be dispersed in the PMMA and the bundles appeared to be undone. It is expected that such composite films could be used in multilayer diodes. The transport properties of SWNT–PMMA composite films (10 m thick) were studied in great detail [165, 166]. The SWNTs (diameter 1.3–1.5 nm) are produced by arc discharge with a typical purity of about 70–90%. Most SWNTs form bundles with a typical size of 7–12 nm. The electrical conductivity strongly depends on the SWNTs content. It increases by nine orders of magnitude from 0.1 to 8 wt% (Fig. 9). The room-temperature conductivity is well described with the standard percolation theory. Transport is due to localized carriers originating from the metallic SWNTs. The critical behavior  ∝ f − fc  is observed

646

Carbon Nanotubes in Composite Materials

4.3. Polymer Composites with Other Matrices

100

10-2

0

10

-1

10

10-3

-2

10

-4

σ (Ω.cm)-1

Conductivity (Ω.cm)-1

10

-1

10-5 10

-6

10

mc = 0.33% β = 2.1

-3

10

-4

10

-5

10

-6

10

10-7

0.1

1

10

Reduced fraction (m-mc)/mc

10-8 0

1

2

3

4

5

6

7

8

9

Mass fraction (%)

Figure 9. Dependence of the room-temperature (RT) composite conductivity with the SWNT content. Inset: RT power law dependence of  versus the reduced mass fraction of SWNTs. Reprinted with permission from [165], J. M. Benoit et al., Synth. Met. 121, 1215 (2001). © 2001, Elsevier Science.

with a percolation threshold fc = 033 ± 003% and a critical exponent  = 21 ± 01. The low value of fc arises from the very high aspect ratio of the SWNTs, estimated at about 850. The critical exponent  is in good agreement with the conventional 1.94 exponent found for random tridimensional connectivity. This quasi-ideal percolation behavior confirms the good dispersion of the SWNTs in the matrix. The temperature (4–300 K) dependence of the resistivity of some composites was studied, revealing a behavior usually ascribed to hopping in the presence of Coulomb interactions opening a soft Coulomb gap. This suggests that a distribution of intertube/interbundle barriers is superposed to the topological percolation network. It is proposed that the Coulomb gap is due to the Coulomb charging energy involved in the transport process from one bundle (or one SWNT) to the neighboring one. It is further suggested that using composites close to the percolation threshold may be a way to access intrinsic SWNT properties at a macroscopical level and that this approach may be developed for the thermal conductivity as well. Grimes et al. [167] investigated SWNT–PMMA composite films (thickness about 500 m) with much higher SWNT loading. Indeed 0–23 wt% of arc-discharge SWNTs (diameter 10 nm, length 1 m) were dispersed. The electrical conductivity varied over nine orders of magnitude and the percolation threshold was found to be close to 3 wt%. Addition of the SWNTs has a dramatic effect on the measured permittivity spectra (for example at 500 MHz, the real permittivity increases by a factor of 35 and the complex permittivity increases by over three orders of magnitude upon increasing the SWNT loading from 0 to 23 wt%). Both sets of data (electrical conductivity and complex permittivity) fall on a reasonably smooth curve indicating a good dispersion of the SWNTs in the matrix. It is, however, noted that no percolation threshold is observed in the 500 MHz data. Comparison of the permittivity measurements (500– 5500 MHz) with theoretical predictions using an effective medium theory were found to be in reasonable agreement.

Shaffer and Windle [168] prepared CCVD MWNT–polyvinyl alcohol (PVA) composite films with a wide range of MWNT loadings (5–60 wt%). The dry film thickness varied from 53 to 44 m, with increased loading. SEM images revealed an even dispersion of the MWNTs and pullout lengths of about 100 nm were observed on fracture surfaces. TGA showed that although there is a retardation of the onset of PVA decomposition, resulting possibly from the absorption of the decomposition-generated free radicals by activated carbon surfaces, the subsequent progress of the degradation of PVA is not affected. Below the glass transition of the matrix, the tensile elastic modulus increases due to the presence of the MWNTs. The stiffness of the MWNTs was evaluated at about 150 MPa, well below values reported for isolated tubes. This low experimental value could reflect defects in the MWNTs or the difficulty of shear stress transfer between the walls. Above the glass transition, a large proportion of the stiffness observed below is retained, denoting a much more significant effect of the MWNTs. The onset of the glass transition was not affected by the MWNTs but the high-temperature side was broadened, because polymer molecules near the MWNT surface would have a reduced mobility. The percolation threshold was found to be in the range 5–10 wt%. It is noted that the electrical conductivity is rather lower compared to that of MWNT–epoxy composites containing very low MWNTs loadings [154], possibly because the adsorbed layer of polymer on each MWNT reduces the quantity and quality of electrical contacts between the MWNTs. Vigolo et al. [169] proposed an original method for the fabrication of macroscopic composite fibers and ribbons incorporating flow-aligned SWNTs. The elastic modulus was found to be equal to 15 GPa, lower than that of individual CNTs or conventional carbon fibers, but nevertheless giving hope for future improvements. It was shown that such fibers can be netted and even knotted. Starting with arc-discharge MWNTs randomly dispersed in a thermoplastic polymer (polyhydroxyaminoether) prepared by solution casting, Jin et al. [170] prepared composites with aligned MWNTs by mechanical stretching above the glass transition temperature of the polymer. X-ray diffraction (XRD) analysis showed that the fraction of aligned MWNTs increased with increasing stretching ratio. Decreasing the MWNT content also favored alignment, due to a higher mobility. Interestingly, no alignment was achieved with SWNTs using similar stretching ratios. A further study [171] revealed that buckling was observed in bent MWNTs of large curvature, some being completely collapsed. Buckling is thought to be caused by the shrinkage of the polymer during cooling down to room temperature in the final stage of preparation. In some cases, heating by electron irradiation in the TEM restored the cylindrical shape, showing that buckling is an elastic deformation, at least at moderate strain. The onset buckling strain was estimated at about 5%, in good agreement with results from AFM lateral force measurements [80]. The observation of fractured surfaces revealed that failure occurred from MWNT pullout and fracture of the matrix but that the load transfer was not

Carbon Nanotubes in Composite Materials

sufficient to fracture the MWNTs. The fracture strain was found to be greater than 18%. Wagner et al. [172] prepared arc-discharge MWNT– polyurethane acrylate films 200 m thick. Tensile test were performed and subsequently thin slices (70 nm) cut in a direction parallel to the film surface were observed by TEM. A progressive fragmentation process was detected which was thought to originate in the large stresses induced in the MWNTs during the tensile tests. The MWNT-matrix stress transfer efficiency was estimated to be at least one order of magnitude larger than in fiber-based composites. This could arise from a strong reactivity of MWNTs toward double bonds in the polymeric chains upon ultraviolet (UV) curing of the specimen. The macrofragmentation and microfragmentation phenomena were studied in detail in a further work [173]. Wood et al. [174] prepared 0.1 wt% SWNT–polyurethane acrylate films 150 m thick using a flow orientation technique to align the SWNTs. Micro-Raman spectra for specimens cut both parallel and perpendicular to the flow direction were significantly different, as a function of mechanical strains, allowing one to probe the alignment. It is proposed that the adhesion between the SWNTs and the polymer exceeds the shear yield strength of the matrix. A further work [175] showed that using polarized resonance Raman spectroscopy allows one to use unoriented SWNTs as strain sensors. The Raman spectral shifts of the SWNTs were also used to study the glass transition and secondary transitions in the polyurethane acrylate and polycarbonate polymers [176]. The rheological behavior of MWNT–polycarbonate composites was reported by Pötschke et al. [177]. CCVD MWNTs (diameter 10–15 nm, length 1–10 m) were used. One-kilogram mixtures of polycarbonate with the MWNTs were extruded at 240
C using a twin-screw extruder. Composites with MWNTs loadings of 0.5, 1, 2, and 5 wt% (corresponding to about 0.34, 0.68, 1.37, and 3.40 vol%, respectively) were formed into bars by compression molding. SEM observations showed that the MWNTs are randomly dispersed in the matrix and that a layer of polycarbonate covers the MWNTs, indicating some wetting and phase adherence. MWNT–polyamide-6 prepared for the sake of comparison did not exhibit the same MWNT-covering features. Electrical resistivity measurements revealed that the percolation threshold is between 1 and 2 wt%. The rheological behavior was investigated by oscillatory rheometry at 260
C. The viscosity of composites with less than 2 wt% MWNTs had similar frequency dependency than pure polycarbonate and reached a Newtonian plateau at low frequencies. Above 2 wt%, the viscosity decreases to a greater extent with frequency and a non-Newtonian behavior is obtained to much lower frequencies. The coincidence between the electrical and rheological thresholds indicates that the rheology of the material is sensitive to the interconnectivity of the MWNTs. The increase in viscosity with MWNT content is much higher than that reported for nanofibers and carbon black, because of the higher aspect ratio of the MWNTs. The viscosity increase appears along an increase of the elastic melt properties, the increase of the storage modulus being much more important than that of the loss modulus.

647 MWNT–polycarbonate and MWNT–polybutylene terephthalate (CCVD MWNTs, 8–15 walls, loading 3.5 and 5.0 wt%) were reported [178] to have improved mechanical and electrical properties compared to the corresponding pure polymer. It is noted that the MWNTs must be untangled for optimized performance, which can be done by high shear in a compounding operation. The synthesis of MWNT–polyaniline composites by in-situ polymerization was reported [179]. Arc-discharge MWNTs (10, 20, 30, and 50 wt% with respect to aniline monomer) were used. A composite of 30 wt% MWNTs and ex situ polymerized polyaniline was also prepared for comparison. In this process, polyaniline exists in its primary doped form called emeraldine salt. X-ray diffration analysis showed that no additional order had been introduced. Comparison of the different Raman spectra revealed that a site-selective interaction between the quinoid ring of the doped polymer and the MWNTs occurs as a consequence of the in-situ polymerization. This interaction may influence charge-transfer processes. The transport properties of MWNTs, polyaniline and the different composites were studied between 300 and 1.25 K. At the higher temperatures, the electrical conductivity is dominated by the polyaniline. At low temperatures, the MWNT network which shows a very weak temperature dependence becomes more conductive than the matrix. Interestingly, it is shown that the conductivity of both the polyaniline and the MWNTs has increased during the in-situ polymerization, indicating improved electrical contacts between the MWNTs and the polyaniline grains. Thus the composite material is more conductive than the starting components. Studies using arc-discharge SWNTs [180] revealed that the dispersion in polyaniline is poor for all SWNT contents (2, 5, 10, 20, 30, 50 wt% with respect to aniline monomer), probably because the SWNTs contain residual catalyst particles and amorphous carbon, as well as form entangled bundles. Qian et al. [181] reported the preparation of MWNT– polystyrene films about 400 m thick by a solutionevaporation method. One wt% of MWNTs 30 nm in diameter and either 15 or 50 m long were used. Tensile tests revealed a 25% increase in break stress and a 36– 42% increase in elastic modulus. A higher elastic modulus being achieved with longer MWNTs reveals that the external tensile loads are successfully transmitted to the MWNTs through the interface with polystyrene. It is claimed that the MWNTs are superior to VGCF and high-modulus carbon fibers due to their very large length-to-diameter ratio. In-situ TEM observations showed that cracks tend to nucleate at low MWNTs density areas then propagate along weak interfaces or relatively low MWNT density areas. The MWNTs align perpendicular to the crack direction and bridge the crack faces. They break and/or pull out from the matrix when the crack opening displacement exceeds 800 nm. A theoretical study on the interfacial characteristics of SWNT– and DWNT–polystyrene composites using molecular mechanics simulations and elasticity calculations was reported by Liao and Li [182]. It is found that in the absence of atomic bonding between the CNTs and the matrix, the nonbond interactions (electrostatic and van der Waals interactions), the deformation induced by these forces, and the stress/deformation arising from the thermal expansion

648 coefficient mismatch all contribute to the interfacial stress transfer ability, which is the critical parameter controlling material performance. The interfacial shear stress derived from a CNT pullout simulation is about 160 MPa, significantly higher than for most carbon fiber reinforced polymer composites. Fisher et al. [183] showed by a combined finite element and micromechanical approach that curved CNTs significantly reduce the effective reinforcement when compared to straight CNTs. The model that is developed suggests that the so-called CNT waviness limits the modulus enhancement of CNT–polymer composites. Watts et al. studied the electrical behavior of polystyrene– matrix composites loaded with boron-doped MWNTs [184] and Fe-filled MWNTs [185]. In the first report [184], arcdischarge MWNTs (diameter 5–40 nm, length 4–10 m) and B-MWNTs (1–5% boron, diameter 5–40 nm, length over 20 m) were mixed with a solution of polystyrene in toluene (2.5, 5, 7.5, 10, 12.5 wt% carbon-to-polymer). The solutions were sonicated and evacuated to remove toluene, resulting in the formation of composite films ca. 300 m thick. The electrical conducting path is established via bundling or tube–tube crossing. Polymer-coated CNTs were also observed. Impedance measurements were carried out under compression or bending. The composites exhibited relatively low resistance, being essentially ohmic conductors at frequencies 1–104 Hz. The B-MWNTs showed a lower resistivity because boron doping results in an increase in charge density on MWNTs. In the second work [185], Fe-filled MWNTs are prepared by CCVD. Encapsulated Fe nanowires over 1 m long were frequently found inside the MWNTs. Current–voltage measurements showed that the film resistivity decreased with increasing MWNT content. Impedance measurements (1 MHz–0.1 Hz) revealed that the composites exhibit inductive, capacitive, and resistive phases. The extent of inductive phase increases with the increase of Fe content, while the extent of the capacitive phase increases with the decrease of MWNT content. It is proposed that electromagnetic induction arises from the encapsulation of long Fe crystals in helical MWNTs that form a solenoid-like structure. Frankland et al. [186] reported simulations on load transfer in (10, 10) SWNT–polyethylene composites. It is predicted that in the case of nonbonded SWNT–matrix interactions, SWNTs longer than 10–100 m are needed for significant load-bearing ability. In contrast, SWNTs crosslinked to the matrix should increase the shear yield by one or two orders of magnitude, the critical SWNT length thus being reduced to 1 m. A further study [187] showed that the introduction of a relatively low density (20 nm) nanotabular structures or structures containing defects/amorphous carbon have also been classified as nanofibers. In this work, we prefer to use the term nanotube to describe carbon filaments with tubular graphene walls parallel to the axis and use the term nanofiber for carbon filaments with graphene layers at other angles. This is because special physical properties arise from the “nanotube” structure which distinguish it from the “nanofiber” structure, which itself has other advantageous properties, as will be described later. Carbon nanofibers and nanotubes have been synthesized since the 1960s, but why has one particular form (i.e. the nanotube) received so much attention recently? In 1991, Iijima reported that highly graphitized carbon nanotubes, formed from the arc discharge of graphite electrodes, contained several coaxial tubes and a hollow core [9]. This important discovery led to the realization that with graphene tubes parallel to the filament axis, these highly crystallized tubular carbon structures would inherit several

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 1: Pages (665–686)

Catalytic Synthesis of Carbon Nanotubes and Nanofibers

666

Figure 1. The three structural forms of carbon nanofibers, classified by the angle of the graphene layers/platelets with respect to the filament axis. Adapted with permission from [5], N. M. Rodriguez et al., Langmuir 11, 3862 (1995). © 1995, American Chemical Society.

important properties of “intraplane” graphite. In particular, a nanotube exhibits high electrical conductivity, thermal conductivity, and mechanical strength along its axis. As there are very few open edges and dangling bonds in the structure, nanotubes are also very inert and species tend to be physically adsorbed onto graphene walls rather than chemically react with them. Note that carbon nanotubes are completely covalently bonded which implies that as electrical conductors, they would not suffer from electromigration or atomic diffusion like metals. These properties make carbon nanotubes a technologically important material for various electronic and mechanical applications as will be listed later. The stacked and herringbone nanofibers tend to be investigated solely for energy storage applications such as electrodes for lithium batteries or fuel cells as small

ions/molecules can enter via open edges and intercalate between the graphene layers. Already, large amounts of nanofibers can be purchased commercially from companies such as Hyperion Catalysis International, Applied Sciences Incorporated, Catalytic Materials, Showa Denko, and Nanomirae [10]. Nanotubes are also available commercially from companies such as Carbon Nanotechnologies Inc., Iljin Nanotech, and NanoLab among many others listed in [11]. Nanotubes are further classified into two types, namely multiwall and single wall [9, 12, 13]. The multiwall carbon nanotube contains several concentric, coaxial graphene cylinders with interlayer spacings of ∼0.34 nm [14]. Note that this spacing is larger than single crystal graphite (0.335 nm). Recent studies have shown that the intershell spacing can actually range from 0.34 to 0.39 nm, where the intershell spacing decreases with increasing carbon nanotube diameter, and this effect is more pronounced in small diameter nanotubes (3000 W mK [29]

∼60 GPa [25]

Electrical properties Typical resistivity of single and multiwall nanotube Typical maximum current density Quantized conductance, theoretical/measured

Thermal conductivity of single wall nanotube Thermal conductivity of multiwall nanotube

Electronic properties 10

−6

 m [22, 30]

107 –109 A cm2 [31, 32] (6.5 k)−1 /(12.9 k)−1 [22, 33]

Single wall nanotube bandgap— whose n–m is divisible by 3 whose n–m is nondivisible by 3 Multiwall nanotube bandgap

0 eV [22] (metallic) 0.4–0.7 eV [20, 21] (semiconducting) ∼0 eV [22] (nonsemiconducting)

Catalytic Synthesis of Carbon Nanotubes and Nanofibers

nonsemiconducting. Table 1 lists several reported properties of carbon nanotubes. Carbon nanotubes and nanofibers are being investigated for a wide range of applications today. Reviews of carbon nanotube applications are presented in [24, 35–42], and only an overview is presented in this chapter. Let us first consider the nanotube as a high aspect ratio, electrically conductive wire with diameter in the nanometer range. These structures are highly desirable as field emission tips for applications such as field emission displays [43–48], X-ray tubes [49], electron sources for microscopy and lithography [50], gas discharge tubes [51], and vacuum microwave amplifiers. The use of nanotubes as a field emission electron source has recently been commercialized in a portable X-ray source by Oxford Instruments [52]. Nanofibers have also been investigated as electron sources [53, 54]. The high aspect ratio and small diameter of the nanotube is also desirable for scanning probe tips [55–57]. In fact, nanotube-based scanning probe tips (“Probemax”) are commercially available today from nPoint (also known as Piezomax) [58]. Single wall carbon nanotubes, which can be electronically semiconducting, are also being investigated as transistors or logic elements [59–62]. Containing only one carbon tube, the electronic properties of the carbon nanotubes are highly sensitive to adsorbed molecules/species [63, 64]. Although this implies that in logic circuits, the nanotubes must be suitably encapsulated, the high sensitivity of the nanotubes can be advantageously utilized in chemical or biological sensors to detect poisonous or dangerous gases in the environment. In addition, the coherent nature of electron transport in wellcrystallized nanotubes would find these structures applicable in spin–electronic devices [65]. Carbon nanotubes could also be used as electromechanical sensors as their electrical characteristics respond to mechanical deformation of their structure [66]. Another interesting application for these structures is as electrodes in electrochemical supercapacitors [35, 67–71]. When nanotubes/fibers are produced enmasse (i.e., woollike or forestlike), they have large surface areas which could lead to higher charge storage capabilities than standard capacitors and batteries [35, 72]. The cycle characteristics of lead acid and lithium ion batteries can also be improved when carbon nanofibers are used as fillers in the battery electrodes [73]. The high electrical conductivity and relative inertness of nanotubes make them potential candidates as electrodes in electrochemical reactions too [74, 75]. The large surface area of nanotubes, both inside and outside, can be usefully employed to support reactant particles in catalytic conversion reactions [76, 77]. It was also proposed that hydrogen could also be stored among and inside nanotubes/nanofibers for fuel cell applications [7, 76–81], although recent results show that the amount of hydrogen stored is not as high as originally anticipated [82]. Nanotubes can also mechanically deflect under electric stimulation (e.g., due to charge induced on the nanotubes) and this opens up applications such as cantilevers or actuators [35, 83–85]. The use of nanotubes and nanofibers as filters or membranes for molecular transport has been recently proposed [86]. The exceptional mechanical properties and low weight of nanotubes and nanofibers make them potential filling materials in polymer composites. Nanotubes and nanofibers

667 can improve the strength and stiffness of a polymer, as well as add multifunctionality (such as electrical conductivity) to polymer based composite systems [87–95]. Carbon nanotubes should be ideal reinforcing fibers for composites due to their high aspect ratio and high in-axis strength [96]. Furthermore, carbon nanotubes, unlike macroscopic carbon fibers, are short enough to flow through conventional polymer processing equipment so that complicated shapes or small parts could be molded from their composites [96, 97]. As fillers for composites, single wall carbon nanotubes are preferred to multiwall nanotubes because the inner layers of the multiwall nanotubes contribute little under structural loading and thus would reduce the stiffness for a given volume fraction of tubes [26, 96]. It is evident that for the various different applications, nanotubes or nanofibers of different morphologies are required. For instance, scanning probe applications require a single high aspect ratio nanotube whereas polymer strengthening requires “masses” of nanotubes/nanofibers. Field emission applications ideally require vertically aligned nanotubes which are spaced about twice their height apart [98–100], whereas horizontally aligned nanotubes are more suited for electrical transport or electronic (transistor/spin) applications. The great flexibility of catalytic chemical vapor deposition (CVD) is that this technique can be adapted to producing nanotubes and nanofibers for virtually all these applications. In the case of nanotubes, chemical vapor deposition is very different from the other two common methods used for nanotube production, namely arc discharge [101, 102] and laser ablation [30]. Arc discharge and laser ablation can be classified as high temperature (>3000 K) and short time reactions (s–ms), whereas catalytic chemical vapor deposition is a medium temperature (700–1400 K) and long time reaction (typically minutes to hours). Although carbon filament/nanofiber growth by catalytic chemical vapor deposition was established in the 1960s–1980s [3, 4, 103–105], much of the fundamental work on the properties of nanotubes in the early 1990s was performed on nanotubes produced by arc discharge and laser ablation because of their superior straightness and crystallinity due to the high temperature deposition. The main technological drawbacks with arc discharge and laser ablation were that the nanotubes had to be produced separately (i.e., not directly on substrates), purified [106, 107], and then manipulated onto substrates before use. At that time, most CVD-grown nanotubes were “spaghetti-like” and largely defective, but the potential of the technique to satisfy technological requirements was recognized. From 1998 onward, substantial and rapid progress was made in the development of CVD to establish it as a highly controlled technology for the production of carbon nanotubes and nanofibers: today, it is possible to fabricate high quality single wall carbon nanotubes [108, 109] or multiwall carbon nanotubes [110], horizontally [111, 112] or vertically aligned [113–115], as an individual nanotube [116–118] or “en masse” [110, 119], with controlled diameter [120, 121] and length [122, 123], structurally as a tube or stacked layered nanofiber form [5, 124], directly onto substrates or in bulk as a raw material [4, 125]. A major advantage of CVD is that the nanotubes/nanofibers can be used directly without

668 further purification unless the catalyst particle is required to be removed, methods for which will be discussed later. The rest of this chapter discusses the growth mechanism of carbon nanotubes and nanofibers, the various methods of catalyst preparation, and variations in the chemical vapor deposition technique. This provides the technologist with a repertoire of techniques from which he/she can choose the most suitable one for his/her specific application.

2. GROWTH MECHANISM OF CARBON NANOTUBES AND NANOFIBERS 2.1. General Mechanisms In general, carbon nanotube and nanofiber growth by the catalytic CVD method require catalyst nanoparticles (usually Fe, Co, or Ni), a carbon feedstock (e.g., hydrocarbon or CO), and heat. The diameter of the filament produced is often closely related to the physical dimension of the metal catalyst. The peculiar ability of these transition metals to form graphitic carbon is thought to be related to a combination of factors that include their catalytic activity for the decomposition of volatile carbon compounds, the formation of metastable carbides, and the diffusion of carbon through the metal particles [126]. We present some of the growth models that have been proposed both for nanotubes and nanofibers which are widely accepted by the research community. The most commonly accepted mechanism was postulated by Baker et al. in the early 1970s, who explained the growth of carbon filaments by catalytic decomposition of the carbon feedstock and bulk diffusion of carbon [1]. According to this mechanism (Fig. 2a), the hydrocarbon gas decomposes on the front-exposed surfaces of the metal particle to release hydrogen and carbon, which dissolve in the particle. The dissolved carbon diffuses through the particle and is precipitated at the trailing end to form the body of the carbon filament. Due to the exothermic decomposition of hydrocarbons, it is believed that a temperature gradient exists across the catalyst particle. Since the solubility of carbon in a metal is temperature dependent, precipitation of excess carbon will

Figure 2. The growth of carbon nanotubes and nanofibers involves the catalytic decomposition of a carbon feedstock (hydrocarbon or CO), carbon diffusion, and its precipitation as a filament. In (a), the carbon diffuses through the bulk of the metal catalyst “M” as proposed by the Baker model [3]. In (b), the carbon diffuses over the surface of the metal catalyst and forms a tubular structure from the circumference of the catalyst as proposed by the Oberlin model [134]. In (c), angled graphene layers are precipitated from a faceted catalyst particle to form a nanofiber as proposed by Rodriguez and Terrones [5, 147].

Catalytic Synthesis of Carbon Nanotubes and Nanofibers

occur at the colder zone behind the particle, thus allowing the solid filament to grow with the same diameter as the width of the catalyst particle. Such a process will continue until the leading tip of the catalyst particle is “poisoned” or deactivated. A common cause of catalyst poisoning is the formation of carbon around it, thus preventing the gas from reaching the catalyst particle. Support for this bulk diffusion model comes from experiments on the kinetics of growth of carbon filaments from acetylene (C2 H2  catalyzed by Ni particles, which yielded an activation energy of (eV) 140 kJ mole−1 [2]. This value is similar to the activation energy for bulk diffusion of carbon through solid Ni (i.e., 133 kJ mole−1  [127]. Similarly, the enthalpies for the growth of filaments with -Fe, -Fe, Ni, Co, Fe–Ni, and Cu catalyst were found to be similar to the enthalpy of diffusion. Thus, the rate limiting step in the growth is believed to be the diffusion of carbon through the catalyst. In general, the filament length depends on the duration of the catalytic process, where longer durations result in longer filaments [128, 129]. This general bulk diffusion mechanism accounts for the formation of both nanofibers and nanotubes. However, two irregularities of this growth mechanism should be noted. First, not all hydrocarbon dehydrogenation reactions are exothermic (e.g., methane), and yet growth has been observed from these hydrocarbons. Moreover, it is unlikely that there is a temperature gradient across such a small metal particle. This is because the metal particle has a high thermal conductivity and thus a small temperature gradient implies that a massive heat flow is occurring through the particle, which is physically intangible. The exothermic decomposition of the hydrocarbon probably raises the temperature of the entire filament, and the growth of the carbon filament is also probably driven by a concentration gradient of carbon across the particle. One common question asked is whether the catalyst is a liquid or solid during nanotube/nanofiber growth. If we assume Fe as the metal catalyst, most of the growth experiments are typically well below the melting temperature of iron (1534  C) and also below the iron–carbon eutectic temperature (1147  C). The formation of graphite platelets from certain crystallographic faces of the catalyst particle (see Section 2.2) suggests that the catalyst is in a solid form. The agreement between the enthalpy for the growth of filaments and the enthalpy of diffusion for the bulk catalyst metals as discussed earlier also suggests that the catalyst is in the solid phase. Although growth is performed below the eutectic temperature and metal melting point (N.B. usually these temperatures are quoted at 1 atm and will change with operating pressure), one should note that the catalyst metal nanoparticles will behave completely differently than their bulk metal form because these small particles will have exceptionally high surface energy, area, and mobility. For example, Hou et al. reported that annealing iron encapsulated carbon particles in argon between 1000 to 1100  C completely removes the iron, indicating that these small iron particles were highly mobile at these temperatures [130]. Because of the high mobility and reactivity of the metal atoms, the catalyst nanoparticles are often in the shape of metallic clusters or have been observed to undergo certain surface reconstruction. If temperatures above the metal– carbon eutectic are used, the growth would be similar to the

Catalytic Synthesis of Carbon Nanotubes and Nanofibers

general vapor–liquid–solid mechanism proposed by Wagner [131, 132] where diffusion through a liquid-phase particle is responsible for the synthesis of filaments. Instead of bulk diffusion, another common growth model is a catalytic process involving the surface diffusion of carbon around the metal particle [133–135] as shown in Figure 2b. The carbon atoms diffuse over the catalyst surface to form a tubular structure which emanates from the circumference of the catalyst. Note that the tubular structure is favored for carbon filaments with nanometric diameters. A single graphene layer of finite size has many dangling bonds which correspond to high energy states, and for such a small structure, there would be an enormous percentage of dangling bonds if a stacked planar graphite was formed [17, 136]. By the formation of closed tubular carbon shells, the total energy of the system is reduced [136]. An alternative model based on the minimization of surface energy of nanoparticles was suggested by Dai et al. [137]. Nanoparticles contain a very high percentage of surface atoms; as a result a large amount of surface energy exists. Excess carbon can help solve this problem by assembling a graphene cap on the particle surface, called a “yarmulke,” with its edges strongly chemisorbed to the metal. Since the basal plane of graphite has an extremely low surface energy (10–20 times smaller than most metals), the total surface energy diminishes. A crucial feature of the yarmulke mechanism was its avoidance at all stages of growth of any open graphene edges, which would expose energetically costly dangling bonds. It also provides an automatic solution to forming caps and resulting structures are tubes which have no seams. Carbon can add to the cylindrical section of a growing layer. Once the smallest yarmulke has formed, insertion of new carbon between the tube edge and the catalytic particle is the best solution, as long as complete overcoating of the particle (i.e., encapsulation) is avoided which would deactivate it. The actual composition of the active catalyst particle is a widely debated issue, and further research could be performed in this area. High carbon content carbides were determined to be a prerequisite for carbon fiber growth [138–140]. However, there are many conflicting reports concerning the actual composition of the catalyst particle; for example, in the case of an iron catalyst, a hexagonal form with composition Fe2 2 C or Fe2 C was postulated, rather than Fe3 C [134]. There are also arguments as to whether a carbide particle is indeed the active catalyst [103]. These are based on the findings that the loss of catalytic activity of iron was accompanied by a gradual conversion of the catalyst to a stable carbide. The catalyst could be reactivated by treatment with hydrogen, reducing the carbide back to iron.

2.2. Nanofiber Growth Let us now focus our discussion on catalysts used for nanofiber growth and why nanofibers form. The ability to control and tailor the structure of nanofibers (stacked or herringbone) has been demonstrated by Rodriquez et al. [5, 124]. The general concept used here is the creation of a faceted catalyst particle [5, 124, 141] so that carbon feedstock decomposition occurs at certain faces whereas carbon precipitation (in the form of graphite layers) occurs at other

669 faces as shown in Figure 2c. The graphitic platelets are precipitated parallel to the surface of the faceted catalyst particle, and hence the angle between the planes and the fiber axis is determined by the shape of the catalyst particle, as proposed by Boellard et al. [138]. Under certain conditions of gas composition, temperature, and catalyst composition, the catalyst particles undergo surface reconstruction to form unique geometrical shapes which drive the formation of nanofibers [5, 124, 142, 143]. For example, the herringbone structure was found to grow from Fe–Cu (7:3) particles in a C2 H4 –H2 (4:1) gas mixture at 600  C, whereas the stacked structure formed from Fe-based catalyst in a CO–H2 (4:1) gas mixture at 600  C [5]. The formation of herringbone structures is favored when the catalyst particle is an alloy [124, 144–146], although Pd has also been used alone under certain growth conditions to yield similar structures [147]. Nolan et al. [148] have suggested that hydrogen plays a significant role in the formation of nanofibers. This is because the presence of hydrogen in abundance can terminate the large number of dangling bonds at the edges of the stacked graphite platelets, whereas without hydrogen termination, the more stable form of the carbon filament would be closed tubular graphene shells where there are no dangling bonds. In plasma enhanced CVD, the carbon filaments formed are often nanofibers rather than nanotubes. This is thought to be due to the large amount of atomic hydrogen formed in the gas phase due to plasma decomposition of the hydrocarbon gas or the use of hydrogen as a dilution gas. Delzeit et al. showed that by controlling the relative amount of hydrogen in the gas phase via altering the plasma parameters, one could change the structure from nanotubes to herringbone nanofibers, with high hydrogen content favoring the latter [149].

2.3. Multiwall and Single Wall Nanotube Growth Let us now concentrate on the formation of nanotubes. Without an abundance of dangling bond terminating species (e.g., H) in the gas phase, carbon nanotubes will tend to form when the diameter of the filaments is ∼50 nm or less. This is because a single graphene layer of finite size has many dangling bonds, and these dangling bonds correspond to high energy states [17, 136]. The total energy of a small number of carbon atoms is reduced by eliminating these dangling bonds, even at the expense of increasing the strain energy, thereby promoting the formation of the closed tubular structure [17, 136, 150]. The catalytic activity of the metal catalyst in the formation of nanotubes has also been studied in considerable detail. Besides the commonly used Fe, Co, and Ni catalysts, other metals (such as Mo, Cu) or metal mixtures (Fe–Ni, Fe–Mo, Fe–Co, Co–Ni, and Co–Mo) have been used for nanotube synthesis [109, 113, 122, 128, 137, 151–162]. Using thermal CVD, Co and Fe catalysts generally tend to form hollow and well-graphitized nanotubes, whereas Ni and Cu produced structures which were not as well graphitized [158, 163]. One should note that different metal catalysts would have their optimum catalytic activity at different temperatures [128, 129, 163]. The yield and crystallinity of nanotubes can be improved by the use of metal

670 catalyst mixtures such as Co–Fe or Co–Ni [158, 164]. Furthermore, it has also been reported that the addition of Mo to Fe or Co [109, 122, 151–157] increases the yield of single wall nanotubes compared to when a single metal catalyst is used [161]. A Co to Mo ratio of 1:2 is reported to be optimal for a high synthesis yield of single wall carbon nanotubes. Figure 3 shows various forms of nanotubes. Here, the nanotubes are examined using transmission electron microscopy and hence we are looking at a cross section of the nanotubes. The solid black lines in the micrographs represent the hexagonal sheets of carbon atoms which make up the walls of the nanotubes. Under ideal growth conditions, the nanotubes produced should be straight (e.g., Fig. 3a) and contain graphene walls parallel to the tube axis without

Figure 3. (a) TEM of straight nanotubes grown from the floating catalyst method using ferrocene–toluene at 760  C. (b) Helical nanotubes grown from CVD of C2 H2 /Ar at 700  C with Ni catalyst. (c) Straight nanotube with bamboo compartments grown by plasma enhanced CVD of C2 H2 /NH3 at 700  C with Fe catalyst.

Catalytic Synthesis of Carbon Nanotubes and Nanofibers

any defects. Dai proposed that a high growth temperature is required to anneal out defects so that well-crystallized and straight nanotubes could be obtained [165]. The occurrence of defects (e.g., pentagons or heptagons) would cause the nanotube to bend during growth. When carbon nanotubes are formed by the electric arc or laser ablation, temperatures of ∼3000 K are obtained and this possibly explains why mostly straight and well-crystallized nanotubes are obtained from these processes. In lower temperature CVD processes (∼700–1400 K), “curly” and “coiled” nanotubes (Fig. 3b) are common variations to the perfectly linear nanotube. The growth of various shapes of nanotubes, especially wavy and helical tubes, was investigated by Amelinckx et al. [166]. The concept of a spatial velocity was introduced to describe the extrusion of carbon from the catalyst particle to form the nanotube. Essentially, when the extruded carbon material was uniform, straight nanotubes are obtained, whereas nonuniform extrusion caused the nanotube to deform elastically into complicated patterns such as the helix shape. Nanotubes containing bamboo compartments are also commonly observed, as shown in Figure 3c. A growth model for bamboo-shaped carbon nanotubes was proposed by Lee and Park [167]. Their transmission electron microscopy (TEM) evidence showed that the bamboo-shaped compartment layers were due to the surface geometry of the catalyst particle and the precipitation of carbon sheets from the bulk of the catalyst particle. Li et al. found that when using a higher deposition pressure of carbon feedstock, the nanotubes became bamboo in structure [168]. They argued that at high pressures, the carbon concentration was sufficiently high to cause bulk diffusion of carbon through the catalyst, forming the bamboo compartments behind the catalyst particle. In the literature, bamboo structures are sometimes called “nanofibers.” However, note from Figure 3c that the bamboo structure actually contains graphene walls parallel to the filament axis, which suggest that these structures would inherit the physical properties of the “nanotube.” It is possible to obtain the growth of straight nanotubes by close-packed growth, use of porous templates, electric field directed growth, or plasma-induced alignment as will be discussed later. Under what conditions is the growth single wall carbon nanotubes preferred? The size of the catalyst is probably the most important parameter for the nucleation of single wall carbon nanotubes. Conclusive evidence on the dependence of catalyst size on the formation of single wall carbon nanotubes has been reported in [159, 160, 169]. Li et al. (Duke University) prepared catalyst nanoparticles of uniform diameters (between 3 to 14 nm) by thermal decomposition of metal carbonyl complexes using a mixture of long-chain carboxylic acid and long-chain amine as protective agents [169]. Their results indicate that the upper limit for single wall nanotube growth occurred at catalyst sizes between 4 and 8 nm. Above 8.5 nm, no more single wall structures were observed [169]. Li et al. (Stanford University) also grew single wall carbon nanotubes from discrete catalytic nanoparticles of various sizes [159]. Discrete nanoparticles were prepared by placing a controllable number of metal atoms into the cores of apoferritin. Their TEM studies indicated that the nanotube diameters were closely correlated to the size of the catalytic nanoparticles.

Catalytic Synthesis of Carbon Nanotubes and Nanofibers

Furthermore, the nanotubes grew by a base-growth mechanism with the nanoparticles seen to be anchored to the support material. Smaller nanoparticles (103 spots per cm2  and experimental flexibility with different mixtures of catalyst solutions. For example, this method was used to study the effectiveness of different metal salts to determine the optimal catalyst composition for the growth of multiwall and single wall carbon nanotubes [194, 195]. It is also possible to deposit catalyst nanoparticles onto a substrate by electrochemical deposition with a metal salt solution. Tu et al. used a solution of 0.01 M of Ni(SO4  and 0.01 M of H3 BO3 to electroplate Ni nanoparticles onto a Si substrate which was metallized with Cr [196]. By changing both the current density and time during electrochemical deposition, it is possible to control the density of the particles and hence the density of the carbon nanotubes on the substrate. The catalyst support for nanotubes/nanofibers could also be in the form of a powder/nanoparticles (typically alumina, silica, or graphite). In the wet catalyst method, these powders are impregnated with the catalyst [163, 173]. For example, graphite can be impregnated with a 40 vol% ethanol/60 vol% water of iron (III) oxalate to form a 2.5 wt% Fe/graphite sample as described in [173]. This sample was next dried in nitrogen at 250  C and reduced in hydrogen to form metallic iron which was then used to catalyze the growth of multiwall nanotubes using acetylene at 700  C. The impregnation method has also been used to prepare catalyst for single wall nanotubes as well [161, 197]. The most promising catalyst that has been reported by such a method is a Co–Mo metal supported on silica as described previously [155, 157]. The impregnation technique is often used for the bulk production of nanotubes and nanofibers. At the end of the reaction, the support can be removed by dissolution in strong acid or alkali to yield the carbon structures. Another wet catalyst preparation route is co-precipitation which involves the reaction of metal salt solutions with

673 ammonium bicarbonate to form metal carbonates. The metal carbonates can be reduced to metal oxides by calcination [5] and further reduced to the metal catalyst during growth using hydrogen. Wet catalysts are especially useful for coating nonplanar geometries such as wires or tips [198–201]. These surfaces can either be dipped in the catalyst solution or the solution can be spin-coated onto the substrate.

3.3. Thin Film Metals Another common technique of depositing the metal catalyst is by physical vapor deposition. A very thin film of Fe, Co, or Ni is carefully deposited on a substrate using sputtering or evaporation. The film thickness is usually in the few nanometer range and is monitored during deposition using a quartz oscillator-type film thickness monitor for accuracy. When this thin film is heated up to a high temperature (such as the growth temperature), the thin film breaks up and coalesces to form nanoclusters due to increased surface mobility and strong cohesive forces of the metal atoms [117, 183]. These nanoclusters then catalyze the growth of the carbon nanotubes or nanofibers. In general, the size of the nanoclusters formed can be controlled by the thickness of the catalyst film [182, 183, 202], by the temperature [203, 204], or by the annealing time. Thicker films, higher temperatures, and lengthy annealing times lead to the formation of larger metal clusters due to increased surface migration of the metal atoms. Although these parameters may be used to control the average size of the nanoclusters (i.e., diameter of resultant nanofibers/nanotubes), one should note that the formation of nanoclusters from the metal film is a random process and thus there will still be a distribution in the diameters of the structures [183, 202]. Multilayer metal films have also been used to catalyze nanotube growth. A noncatalyst metallic underlayer can be used to control the surface properties of the catalyst or the deposition yield as discussed earlier. Single wall nanotubes have been grown using a three layer metal film containing 0.2 nm Mo on 1 mm Fe on 10 nm Al on a silicon substrate [171]. These authors showed that the metal films formed Fe/Mo catalyst particles of ∼2 nm diameter which seeded the single wall nanotubes. The advantage of using catalyst thin films is that they can easily and accurately be patterned by using masking or etching techniques based around photolithography or electron beam lithography. In fact, individual freestanding nanotubes and nanofibers have been deposited using this process [116–118]. Figure 5 shows some examples where a 7 nm thin Ni film was patterned with lines and dots, which were then used to nucleate nanotubes. To obtain single freestanding structures, the Ni film must be patterned into dots of ∼300–350 nm or less [117, 205]. When the Ni film dots are heated up to the growth temperature (700  C), the film forms a single catalyst cluster of equal volume and catalyzes the growth of the nanotube. The vertically standing nanotubes of Figure 5 were deposited by plasma enhanced chemical vapor deposition which will be discussed later. Note that when the metal film thickness exceeds a few tens of nanometers, nanoclusters are no longer formed and the film forms islands a few micrometers in size. These large

674

Catalytic Synthesis of Carbon Nanotubes and Nanofibers

techniques generate the small catalyst particles needed to grow carbon filaments. An alternative technique is to oxidize the metal surface through heating in oxygen or by rusting [192]. During growth, the metal catalyst nanoparticles are formed from the rusty, porous oxide surface by decomposing it with a reducing gas (e.g., H2 .

3.5. Colloids

Figure 5. A 7 nm film of Ni catalyst, prepared by sputtering, was patterned using lithography into lines (a,b) and dots (c). The substrate used was Si, and hence a TiN diffusion barrier (also prepared by sputtering) was used to support the catalyst as Ni reacts with Si at the growth temperature. The growth was performed by plasma enhanced CVD of C2 H2 :NH3 at 700  C. The nanotubes formed are straight, vertically aligned, and typically bamboo in structure (like the TEM shown in Fig. 3c). Using plasma enhanced CVD, it is also possible to attain slightly conical shaped structures by altering the synthesis conditions which is described in [268].

micrometer-size islands/particles usually do not catalyze filament growth but instead absorb carbon into their bulk. Nanotube/nanofiber growth is usually found around the grain boundaries of metal islands. This is because smaller (submicrometer) catalyst particles are often present around the grain boundaries. Baker et al. also observed that carbon filament growth occurred only from the edges of macroscopic Fe foil [192]. In this case, the catalyst material probably easily detached at the metal foil edges to form small catalyst particles which nucleated the carbon filaments. Note that large (micrometer-sized) catalyst grains or thick continuous catalyst films do not nucleate nanotubes. The uniform growth of nanotubes inside large metal grains or catalyst metal substrates is only possible if nanoparticles are present on the surface of the metal as described in the next section.

3.4. Thick Metal Catalyst Films or Metal Catalyst Substrates Nanotubes or nanofibers can be grown with high yield on thick catalyst metal films or catalyst metal substrates when surface treatment techniques are first used to roughen the substrate surfaces. Mechanical roughening (e.g., by using sandpaper) or electrochemical etching may be used to generate a coarse surface. Plasma etching or ion bombardment has also been used to increase the roughness of the metal surface and to generate submicrometer metal islands/particles [113, 206–208]. These surface roughening

Colloidal metal (or oxide) particles have also been used to catalyze the growth of nanotubes. Colloids are usually synthesized (or bought) in liquid suspensions where the colloids are separated by adsorbed charged species or organic molecules. The advantage of using colloids is that these can be highly homogenous in size and can be synthesized in diameters down to 2 nm. Thus, the use of colloidal catalysts allows the growth of nanotubes with well defined diameters, in contrast to the other techniques mentioned above which tend to produce nanotubes with a significant variation in diameter (except for the growth of single nanotubes from a small catalyst patterned dot where the catalyst is essentially fixed in size/volume). Cheung et al. describes the preparation of monodisperse Fe clusters with different diameters and uses these to catalyze the growth of single wall nanotubes supported on an oxidized Si substrate [160]. Li et al. have also synthesized colloids of diameters varying from 3 to 14 nm to catalyze nanotube growth [169]. Colloidal metal suspensions can in general be applied to the substrate using similar techniques as for the wet catalyst. Additionally, colloidal solutions in which the particles are separated by charge can be easily applied/adhered to substrates which have been functionalized with oppositely charged surface layers [209].

3.6. Sol–Gel Technique The sol–gel technique has also been used to prepare catalyst for both multiwall and single wall carbon nanotube synthesis [153 158]. Sol–gels impregnated with metal catalysts have very high surface area, high porosity, and ultralow density— these characteristics lead to a high yield of nanotubes during growth. For example, a Fe–Mo catalyst was prepared by the sol–gel technique based on supercritical CO2 drying and then used for single wall carbon nanotube synthesis, as described in [153]. This catalyst was reported to be capable of deposition yield of over 200% compared with the original weight of the catalyst for a 1 hour deposition. The catalyst was active for 6.5 hours of growth, yielding 600% weight gain in total. The high yields were the result of the aerogel having a high surface area, high porosity, and good metal–support interaction. As yet, no other research group has been able to exceed the yields of single wall carbon nanotubes on a supported substrate by sol–gel [153].

3.7. Unsupported/Floating Catalyst Method The floating catalyst method is commonly used for the bulk/mass production of nanotubes/nanofibers by CVD. The main advantage of using this technique is that purification is not required to recover nanotubes from the substrate. The simplest method is to inject catalyst nanoparticles (e.g., in the form of a colloidal/particle suspension or organometallic precursors with a carbon feedstock) directly into the CVD

Catalytic Synthesis of Carbon Nanotubes and Nanofibers

chamber. In this case, a vertical CVD chamber is usually used so that the nanotubes/nanofibers grow as the catalyst particles fall from top to the bottom of the chamber. This technique has been used to prepare vapor grown carbon fibers for over 20 years [4 105 210–212]. Organometallic compounds are often used as precursors for the catalyst. Examples of organometallic compounds that have been commonly used are metallocenes, iron pentacarbonyl, and iron (II) phthalocyanine [4 172 211–219]. These precursors are usually sublimed and catalyst nanoparticles are formed in-situ when the compound is decomposed/reduced by heat or hydrogen. A double stage furnace is typically needed because of the different temperatures needed for organometallic sublimation and nanotube growth. In general, the sublimation of metallocenes offers little control over the structural parameters of the nanotubes such as length and diameter, although it has been shown that by varying the relative concentration of the metallocene to carbon in the gas phase the average diameter of the structures may be changed [219 220]. An improvement over the double stage furnace is to use a syringe pump and atomizer to continuously feed a metallocene–liquid carbon feedstock solution into a single stage furnace where nanotube growth occurs [219 221–223]. Aligned, high yield and pure multiwall carbon nanotubes can be obtained with conversion rates of 25% of the carbon input using this method. Very often, the floating catalyst technique leads to highly dense/close-packed nanotube deposition where essentially only upward (i.e., “aligned”) growth of the nanotubes is possible. Interestingly, the floating catalyst method can also be used to selectively grow nanotubes on substrates [224 225]. It was observed that multiwall carbon nanotubes grown by CVD of ferrocene and xylene at 800  C only occurred on silica (SiO2  surfaces and not on Si surfaces. Thus, by using lithographic means to pattern SiO2 on a Si substrate, selective growth of nanotubes was obtained [225]. The multiwall nanotubes were aligned and grew perpendicularly from the SiO2 surfaces. In this case, it was suggested that it was the good catalyst–support interaction between SiO2 and Fe that led to the growth of nanotubes. If metal layers of nickel (Ni) were patterned onto Si, the nanotubes were seen to lift these metal patterns during growth [226]. Such aligned nanotubes have also been grown on gold and MgO substrates and palladium seeds [227–229]. Typically, metallocene assisted chemical vapor deposition of hydrocarbons (e.g., benzene, xylene) produces multiwall carbon nanotubes at lower temperatures (∼700  C) whereas a combination of multiwall and single wall nanotubes are produced at much higher (>900  C) temperatures. For example, pyrolysis of iron pentacarbonyl with benzene at 900  C leads to single wall nanotube formation [220]. Nickelocene and cobaltocene were reported to be more favorable for single wall nanotube synthesis than ferrocene, although no differences in single wall nanotube yields were observed when binary mixtures of metallocenes were used, except that the nanotubes appeared to be “cleaner” when mixtures were used [230]. The addition of trace amounts of thiophene (sulfur containing compound) to liquid hydrocarbons has also been reported to promote the growth of single wall carbon

675 nanotubes [170 231 232], although higher concentrations of thiophene were reported to revert the growth back to multiwall in structure (>5 wt%) [170 231]. Recently, extremely long ropes (several centimeters in length) of high purity single wall nanotubes have been synthesized with the vertical floating catalyst method using a ferrocene, n-hexane, and thiophene mixture with hydrogen as the carrier gas [232–234]. These macroscopic ropes will definitely enable the use of single wall nanotubes in mechanical applications. “HiPCO,” developed at Rice University, is also a process involving a high pressure gas phase catalytic process for single wall nanotube growth [172 235]. The catalyst is formed in-situ by the thermal decomposition of iron pentacarbonyl in a heated flow of CO, and growth is performed at pressures ranging from 1 to 10 atmospheres (atm) and temperatures ranging from 800 and 1200  C. The optimum condition for maximum yield was at 1200  C and 10 atm. The rate at which the reactant gases were heated also had substantial effects on the amount and quality of nanotubes produced. The addition of small amounts of methane (0.7% by volume) produced clean nanotubes and increased the yield as well. Although milligram quantities were obtained, such a process is continuous and is being currently scaled up to produce larger quantities of single wall nanotubes (marketed as HiPCO™ single wall carbon nanotubes by Carbon Nanotechnologies Inc.).

4. CHEMICAL VAPOR DEPOSITION CONFIGURATIONS AND CONSIDERATIONS 4.1. Horizontal Furnace The horizontal furnace is the most popular configuration for the production of carbon nanofibers and nanotubes [4 124 137 173]. In its simplest form, it is a heated quartz tube in which the substrates/catalyst are placed. The reactant gases are flowed over the substrates/catalyst which sit in a removable ceramic boat/holder in the center of the quartz tube (see Fig. 6a). The horizontal furnace is advantageous because there is no (or small) temperature gradient within the heated zone. In most cases, the length of the nanotubes/nanofibers can be simply controlled by the length of the deposition time. When samples are first put into the chamber, the quartz tube is first flushed with a “carrier” gas. The most popular carrier gases are argon, hydrogen, and nitrogen [4 115 158 161 221 232]. Argon is mostly used as it easily displaces air and therefore easily forms an inert atmosphere in the chamber. The furnace is then heated up to the growth temperature in the inert atmosphere. Hydrogen is often added to the gas flow to reduce the catalyst particles (e.g., oxides) during heating. Even if the chamber is evacuated by a pump, it is important to maintain a forward flow of inert/reducing gases during heating as it is possible that nanotubes may undesirably grow from the catalytic cracking of pump oil (e.g., from back streaming). When the growth temperature is reached, the carbon feedstock is introduced. As discussed in the growth mechanism section, the choice of carbon feedstock and other additives is based on whether nanofibers, multiwall nanotubes,

676

Catalytic Synthesis of Carbon Nanotubes and Nanofibers

Figure 7. The free enthalpy of formation of some carbon compounds, calculated from [236, 237].

Figure 6. Types of chambers used for catalytic CVD of nanofibers and nanotubes. The most commonly used is the horizontal furnace (a). For mass production, the vertical furnace (b) has been employed. (c) The fluidized bed reactor and (d) a basic plasma enhanced CVD system based around a vacuum chamber.

or single wall nanotubes are desired. Reactions are usually conducted at temperatures below 1000  C to reduce the formation of undesirable carbon deposits such as amorphous carbon [151]. The amorphous carbon is deposited from the thermal decomposition (pyrolysis) of the carbon feedstock gas, whereas the carbon nanotubes/nanofibers are grown from the catalytic decomposition of the carbon feedstock gas. In most cases, “clean” (i.e., amorphous carbonfree) growth of highly crystallized structures is desired and hence the highest deposition temperature without significant self-decomposition of the carbon feedstock is preferred. In order to determine the highest growth temperature possible using a particular carbon feedstock, it is necessary to consider the thermodynamic stability of the compound depending on temperature. The driving force for the pyrolytic reactions involving gaseous components can be derived from plots of the free enthalpy of formation versus temperature (generated using [236]) as shown in Figure 7. The term pyrolytic is defined as converting the carbon feedstock to solid carbon as the main product and to different

volatile compounds as by-products [237]. The stability of the compound increases with the free enthalpy, whereas the driving force for breaking the compound into its elements decreases. For example, the forward reaction producing elementary carbon from carbon monoxide is favored at temperatures below 700  C. In general, there is a compromise between obtaining high purity carbon nanotubes/fibers and high crystallinity. The growth temperature affects the crystallinity of the structure produced, but too high a temperature leads to the formation of pyrolytic amorphous carbon. Furthermore, the diffusion of carbon through the catalyst is a thermally activated process, and hence, in general, higher temperatures lead to higher growth rates which may be desirable for massproduction processes. One strategy to reduce the formation of amorphous carbon is to decrease the contact time between the carbon feedstock and the substrates. This is achieved by using very high gas flow rates (in cases where only carbon feedstock is used in the flow) or high dilution [129 163 175 238]. Thus, the carbon feedstock is often diluted by the carrier gases. Normally there is an optimum ratio of carbon feedstock to carrier gas; for example, using acetylene and nitrogen, the optimized combination was 9% by volume of the total gas flow. If the acetylene combination is increased, amorphous carbon begins to form due to the self-pyrolysis carbon feedstock and the increased exposure of the sample to the carbon feedstock [129]. A small amount of hydrogen in the gas flow is useful in keeping the catalyst particle active by reducing it. It is well known that the growth of nanofibers is enhanced in a hydrogen atmosphere [4 148 189 239]. Hydrogen also reduces the formation of undesirable carbon deposits from the pyrolysis of carbon feedstock [159]. This is because hydrogen rehydrogenates the reactive carbon species in the gas phase. Franklin et al. found that for clean single wall carbon nanotube growth at 900  C, the optimum flow of hydrogen was between 100 and 150 ml/min in a predominantly CH4 flow (CH4 = 1500 ml/min) [176]. The flow of hydrogen had to be increased to 200 ml/min to maintain

Catalytic Synthesis of Carbon Nanotubes and Nanofibers

“clean” growth when a temperature of 950  C was used. In the absence of hydrogen flow, the CH4 was found to pyrolyze and form amorphous carbon deposits all over the substrate. Alternatively, the floating catalyst method can also be used in the horizontal furnace configuration for carbon nanofiber or nanotube growth [4 212]. As discussed before, the catalyst can be sublimed in a preceding furnace stage and flowed into the main furnace for growth or injected directly into the growth furnace using a syringe pump or atomizer spray [4 213–219 221–223 240]. The growth of nanotubes would then occur all over the walls of the quartz tube and also on any substrates placed inside the furnace. This technique is very useful for bulk production since the material can simply be removed from the walls of the chamber or from the substrates after growth. Note that selective area growth is also possible with the floating catalyst technique by using different substrate materials; for example, SiO2 supports nanotube growth whereas Si does not [224 225]. Finally, note that a furnace is not necessarily required to generate the heat needed for nanotube growth. It is possible to grow nanotubes on wires through direct heating with electrical current [241]. Furthermore, plasma or radio-frequency (rf)-induced heating of the catalyst particle can also generate sufficient temperature locally for carbon nanofiber/nanotube growth at lower process temperatures [242].

677 The fluidization process involves the supported catalysts to remain much longer in the furnace than in the vertical floating technique. The fluidization method is relatively new for the bulk production of carbon nanotubes [244–246], and thus far, production rates up to tens of kilograms of multiwall carbon nanotubes a day have been achieved [244].

4.3. Aligned and Directed Nanotube Growth In some applications, the deposition of well aligned carbon nanotubes or nanofibers on substrates is desired. In this area, chemical vapor deposition is uniquely superior to the other nanotube production techniques (i.e., electric arc discharge and laser ablation) in that nanotube/nanofiber alignment on substrates is readily achievable during growth. Whenever very dense and closely packed nanotubes are deposited (see Fig. 8a), they are forced to grow in an upward ensemble (i.e., perpendicularly) from the substrate—this is sometimes referred to as “self-oriented” or “self-assembled” growth. It is generally believed that the nanotube ensemble is held together by van der Waals interaction and that the nanotubes are so closely packed that the only possible

4.2. Vertical Furnace The vertical furnace configuration, as shown in Figure 6b, is usually employed for the continuous production of carbon fibers, nanofibers, and nanotubes [4 125 211 232 243]. The catalyst and carbon source is injected at the top of the furnace and the resultant filaments grow during flight and are collected at the bottom of the chamber. The vertical furnace can be run continuously for the mass production of carbon nanotubes and nanofibers. Ultrafine metal catalyst particles are either introduced into the reactor directly or formed insitu using precursors such as metallocenes as discussed previously. Note that the residence time of the catalyst particle in the vertical furnace is relatively short compared to the horizontal furnace. As discussed earlier, the growth of multiwall or single wall nanotubes is dependent on the temperature and gases used, with higher temperatures and the addition of a sulfur containing compound (e.g., thiophene) favoring single wall nanotube production [170 231 232]. The main advantage with the vertical furnace configuration is the continuous nature of nanofiber/nanotube production and high purity product eliminating the need for purification or removal from the substrate. The vertical furnace technique has been commercialized for the production of multiwall nanotubes and nanofibers, in quantities of tons per year. Most of this material is used in the electrodes of lithium-ion batteries [73] and as fillers in conductive polymers [35]. The fluidized bed reactor (Fig. 6c) is a variation of the vertical furnace. Fluidization is defined as the transformation of solid particles into a fluidlike state through suspension in a gas or liquid. As seen in Figure 6c, supported catalysts are usually placed in the center of the furnace and an upward flow of carbon feedstock gases is used.

Figure 8. Methods of aligning nanotubes during growth. In (a), densely packed nanotubes grow approximately vertically aligned, as shown in this peeled section of a nanotube “film.” This was deposited using the floating catalyst technique employing a ferrocene–toluene solution injected into a heated furnace (750  C). In (b), perfectly straight, same diameter, and same length nanotubes are grown in the pores of a nanochannel alumina template. In (c), a lateral electric field of 0.5 V/m is used to guide these single wall nanotubes horizontally between electrodes during growth. In (d), straight, vertically aligned carbon nanotubes were grown using plasma enhanced CVD. It is believed that the electric field in the plasma sheath which forms over the substrate aligns the structures during growth. (b) Reprinted with permission from [250], J. Li et al., Appl. Phys. Lett. 75, 367 (1999). © 1999, American Institute of Physics. (c) Reprinted with permission from [111], Y. G. Zhang et al., Appl. Phys. Lett. 79, 3155 (2001). © 2001, American Institute of Physics.

678 growth direction is upward. Thus, the key factor in achieving this type of dense, aligned growth is the preparation of dense and active catalyst particles on the substrate surface. One of the most spectacular examples of this was reported by a Chinese research group [114 119]. The authors prepared iron oxide nanoparticles in the pores of mesoporous silica, which was then reduced to iron particles by H2 /N2 flow at 550  C, followed by reaction with C2 H2 at 700  C for nanotube growth. Dense arrays of nanotubes grew perpendicularly outward from the mesoporous silica, and bundles of nanotubes up to 2 mm in length were synthesized [119]. Fan et al. also synthesized “towers” of densely packed nanotubes by using a 5 nm thin film of evaporated iron on electrochemically etched porous silicon [110]. The authors explained that during growth, the walls of the nanotubes interact with their neighbors via van der Waals to form a rigid bundle during growth. In fact, close examination of these nanotube “towers” revealed that there were no nanotubes which “branched out” from the main tower. As the catalyst was rooted at the base of the structure (i.e., base growth), the porous substrate played a key role in the growth as it allowed the deposition gas to continuously feed the catalyst. The strong interaction between the substrate and the catalyst also meant that the catalyst was well rooted and did not sinter to form larger catalyst particles at the growth temperature. Dense and aligned nanotube growth on substrates has also been observed with evaporated iron on oxidized Si [167], from laser-ablated Co catalyst on silica substrates [115], and from substrates coated by nanotubes from the floating catalyst method as discussed earlier [218 220 221 225]. In general, the closepacked nanotubes are not perfectly straight and exhibit some degree of waviness as seen in Figure 8a. Another common means of achieving aligned growth is through the use of templates, the most popular of which are vertical nanopores created by the electrochemical processing (anodization) of aluminum. Porous alumina membranes typically contain vertical nanopores which are a few to hundreds of nanometers in diameter and lengths which can range from a few micrometers to hundreds of micrometers. The electrochemical parameters can be varied to control the pore diameter, length, and density—a review of the template preparation and growth of nanomaterials in templates is given in [247]. The intent is to grow nanotubes within the alumina pores so that the diameter, length, density, and alignment of the structures reflect that of the original template. The use of porous alumina in the synthesis of carbon nanotubes was first reported in 1995 [248]. In general, two types of template grown structures are possible, namely catalyzed and pyrolytic (no catalyst). The latter usually requires higher temperatures in order to decompose the carbon feedstock gas. Che et al. prepared carbon nanotubes with diameters ∼20 nm using an alumina template in ethylene/pyrene with a Ni catalyst at 545  C or without catalyst at 900  C [249]. After growth, the alumina template can be removed by dipping in HF or NaOH solution to reveal an array of well-ordered carbon tubules standing perpendicular on the substrate. Figure 8b shows what is possible with this technique—a well-aligned, hexagonally packed nanotube ensemble with highly homogenous diameter and lengths. This was prepared by Li et al. using an alumina

Catalytic Synthesis of Carbon Nanotubes and Nanofibers

template together with Co/Ni catalysis of acetylene at 650  C [250]. The smallest nanotubes, 4 Å (0.4 nm) in diameter, have been fabricated using zeolite templates [251]. Another strategy of obtaining vertical nanotube/nanofiber growth is through the use of electric fields. Avigal et al. used a vertical electric field during growth to achieve vertical alignment of nanotubes [252]. A similar effect arises during plasma enhanced chemical vapor deposition (see Fig. 8d) which is discussed in the next section. The above techniques yield the growth of vertically aligned structures—but what about horizontally aligned ones? Horizontally aligned nanotubes are necessary for the mass production of nanotube electronic and spintronic devices. The group led by Dai, at Stanford University, is a pioneer in this area. In their early work, straight, single wall nanotubes were grown from patterned catalyst but in random directions [151]. In some cases, the nanotubes grew in the correct direction (e.g., between electrical contacts) which allowed the fabrication of nanotube electronic devices [63 253 254]. Clearly there was a need to control the growth direction in order to achieve a higher fabrication yield of nanotube devices. In [177], they fabricated silicon pillars and contact printed the catalyst for single wall nanotubes on top of the pillars. A suspended network of nanotubes was observed from the pillars after growth. Fascinatingly, the direction of the suspended nanotubes followed the pattern of the pillars—for example, when the pillars were lined up in rows, nanotubes would be found to be suspended from the pillar tops resembling a power line. Using this technique, 100–150 m long single wall nanotubes were grown. If four pillars were arranged in a square, a square arrangement of suspended nanotubes would join the tops of the pillars. The authors reasoned that the nanotubes growing toward a pillar would adhere to the pillar and become suspended, whereas nanotubes growing in other directions do not meet a pillar and would fall down toward the substrate. Thus, the arrangement of pillars essentially defined the growth direction of the nanotubes. Lateral electric fields can also be used to guide nanotubes during growth. Zhang et al. prepared electrode and catalyst “fingers” on a quartz substrate which were biased during chemical vapor deposition in order to create a lateral electric field [111 112]. The authors found that electric fields of 0.13–0.5 V/m were needed to guide and align the single wall nanotubes during growth, as shown in Figure 8c. The mechanism of alignment was due to the electric dipole polarization of the nanotubes which made them align in the applied electric field. Lee et al. also presented a technique for the directed growth of lateral nanotubes. These authors made a sandwich structure comprised of SiO2 –Ni–Nb on silicon [255]. Using microfabrication, only one face of the sandwich was left exposed to the gases and it was from this face that the Ni catalyzed the outward/lateral growth of multiwall nanotubes in C2 H2 /N2 at 650  C.

4.4. Plasma Enhanced Chemical Vapor Deposition Plasma enhanced chemical vapor deposition (PECVD) is a relatively new technique of producing vertically aligned carbon nanotubes and nanofibers, and it is considerably

Catalytic Synthesis of Carbon Nanotubes and Nanofibers

different from the horizontal and vertical furnace techniques. A plasma is an excited/ionized gas, and the processing plasmas, usually known as “cold” plasmas, are generated using dc, rf, or microwave excitation. The simplest plasma system to implement is the dc plasma which is shown in Figure 6d. The dc glow discharge plasma is generated by grounding the anode and applying a negative dc bias of a few hundred volts to the cathode (note that the substrate for nanotube/nanofiber growth is on the cathode in this case such that the plasma would form a sheath over it with a large voltage drop). Plasma systems are commonly used in semiconductor processing for etching or low temperature deposition of thin films. A review presenting the different configurations of plasma systems is presented in [256]. Today, carbon nanotube/nanofiber growth has been demonstrated with hot filament assisted PECVD [113 206–208], microwave PECVD [257–262], dc glow discharge PECVD [117 182], inductively coupled plasma PECVD [74 149], and rf PECVD [242 263]. Plasma depositions are very stable—this leads to highly controllable and reproducible growth conditions. PECVD is usually used to produce vertically aligned carbon nanotubes/nanofibers or grow nanofibers at low temperatures. Chen et al. used a combination of hot filament and dc PECVD (via negative bias on the substrate) to grow aligned carbon nanofibers using a single crystal Ni100 surface with CH4 and N2 gases [206]. This process was later refined using 3% C2 H2 in N2 and a polycrystalline Ni substrate to grow nanofibers of 60–70 nm average diameter and few micrometers in length [207]. The structures produced by Chen et al. were classed as nanofibers because their TEM showed that the graphene planes were slightly tilted/herringbone, although the nanofiber was hollow in the center. These films showed excellent field emission characteristics with low turnon electric fields and high emission currents. A major application for carbon nanotubes is the electron source in field emission displays (a flat panel display technology). The major drawback, however, is that nanotube growth typically requires temperatures of 700  C or higher which exceeds the strain point of the best “display” glass (by Corning) of 666  C. Hence, there was great excitement in 1998 when Ren et al. demonstrated that vertically aligned carbon nanotubes could indeed be deposited below that temperature on glass, using hot filament assisted dc-PECVD of C2 H2 and NH3 [113]. The aligned nanotubes produced were uniform and very straight and stood “individually” as shown in Figure 8d, in contrast to the aligned nanotubes produced by dense growth in which nanotubes were bundled together and wavy. The authors used a thin film Ni catalyst which was sputtered onto the glass substrates and also showed that plasma bombardment could be used to break up the thin film into islands. Furthermore, this work also demonstrated that the initial film thickness of the film could be used to control the diameter of the resultant nanotubes. Huang et al. later used this process to grow nanotubes uniformly on a polycrystalline Ni substrate which was pre-etched to create Ni islands [208]. Ren et al. [116] later used electron beam lithography to pattern submicrometer Ni dots directly on a silicon substrate and achieved the growth of single, freestanding vertical nanotubes. These nanotubes, however, were not very uniform in terms of height and yield which

679 could be due to the absence of a diffusion barrier between the Ni catalyst and Si substrate. Huang et al. [264] next studied thin film Co, Fe, and Ni catalysts on a Ti substrate using the hot filament dc-PECVD process and found that the nanotubes produced with Ni were structurally the best in terms of graphitization, straightness, lack of amorphous carbon overcoating, and structural defects such as openings in the walls. The nanotubes nucleated from Ni also had the fastest growth rate (in terms of length). The authors noted that the diameters of the Ni catalyzed nanotubes were larger than the nanotubes catalyzed by Co and Fe, indicating that Ni had the weakest interaction with the Ti substrate and hence formed the largest catalytic clusters. Huang et al. studied the growth of individual nanotubes from patterned Ni dots and forests of nanotubes from a larger area Ni film [265]. They found that the Ni catalyst particle attained a particular orientation after growth, namely with the 220 orientated in the direction of the plasma. The individual nanotubes were also better crystallized with tubular walls, compared with the forests of nanotubes which had herringbone-like structures (i.e., nanofibers). They suggested that plasma focusing and heating of the catalyst particle in the case of the individual nanotube were responsible for its different structure. Bower et al. devised an elegant set of experiments to investigate the alignment mechanism of the nanotubes [257]. A microwave PECVD system was used to grow carbon nanotubes catalyzed from a Co thin film on a Si substrate. Carbon nanotubes were deposited using PECVD and then the plasma was stopped for conventional thermal CVD to continue. The resultant nanotubes were thus straight for the plasma-grown section and curly for the thermally grown section. Additionally, when nonplanar or angled substrates were introduced into the plasma, the nanotubes still grew perpendicularly from the substrate surfaces because the microwave plasma formed a sheath around these objects. The authors concluded that the alignment was indeed a plasma induced effect and that it was probably the electric field in the plasma sheath formed around objects which guided the growth of the nanotubes perpendicularly from the objects. The electric field in the plasma sheath was estimated to be 0.1 V/m, which is of the same order of magnitude as the electric field used to align single wall carbon nanotubes laterally during growth discussed previously. Chen et al. [266], who used a hot filament dc-PECVD system, showed that by placing a substrate at an angle to the biased cathode electrode, nanotubes also grew at an angle which followed the direction of the electric field in the plasma sheath. Tsai et al., who used a microwave plasma to synthesize aligned nanotubes, proposed a model based on anisotropic etching in order to explain the vertical alignment of the nanotubes [267]. They suggested that nanotubes which grew in random orientations were unprotected by their metal catalyst particle and were hence anisotropically etched away in the plasma [267]. Merkulov et al. developed a simple dc-PECVD system to grow vertically aligned carbon nanofibers [117]. The system used a resistively heated cathode which was biased at −550 V to generate a dc glow discharge. The gases used were C2 H2 and NH3 , and Ni, deposited on top of a Ti

680 barrier layer on silicon substrates, was used as the catalyst. They investigated the lithographic conditions necessary to nucleate single, freestanding nanofibers as well as single-file lines of nanofibers. The maximum catalyst dimensions were determined to be 350 nm for a catalyst dot and 200 nm for the width of a catalyst line (i.e., for a single file line of nanofibers). Above these dimensions, the catalyst was observed to break into multiple nanoclusters at the growth temperature which nucleated more than one nanofiber. Merkulov et al. named their filaments nanofibers because they were highly disordered and bamboolike. In further studies, these authors found that the thickness of the catalyst layer controlled the average diameter of the nanofibers grown by PECVD [202], the ratio of C2 H2 in the gas flow caused the nanofibers to attain a conical shape due to amorphous carbon buildup on walls of the nanofiber as it grew upward [268], and the direction of the nanofiber growth was determined by the local electric field in the plasma sheath [269]. The authors studied the field emission properties of their nanofibers and also fabricated gated microelectronic field emitters and electrochemical electrodes based on single nanofibers [54 75 270–272]. Chhowalla et al. performed a parametric study of the dc-PECVD growth of carbon nanotubes using C2 H2 and NH3 gases with Ni catalyst [182]. The structures were termed nanotubes because the filaments had well crystallized graphene walls parallel to the filament axis (see Fig. 3c for a typical TEM), with bamboo compartments along their axis. They studied the effect of the initial catalyst thickness (with thicker films leading to larger diameter nanotubes which were shorter and of lower areal density), the effect of the C2 H2 ratio in the gas flow (with higher ratios leading to conical structures), the effect of pressure and deposition time (with higher pressures and longer times leading to longer nanotubes), and the effect of the dc bias on the substrate (with higher biases leading to straighter, aligned nanotubes). By examining the plasma characteristics, it was determined that 0.15 V/m was the minimum electric field in the plasma sheath necessary for vertical alignment of the nanotubes. Clearly, to obtain a high degree of vertical alignment, one should thus maximize the electric field in the plasma sheath by increasing the substrate bias or by increasing the gas pressure (which increases ionization, leading to a higher field in the sheath). By investigating the effect of the growth temperature, it was found that the activation energy for PECVD growth was less (i.e., 0.76 eV) than that for thermal CVD (i.e., 1.21 eV), the latter being close to the bulk diffusion of carbon through Ni [204]. This indicates that the plasma plays a role in lowering the activation barrier necessary for nanotube growth. Teo et al. used the dc PECVD process to demonstrate the high yield and uniform growth of patterned areas of nanotubes and individual nanotubes [118 273], examples of which are shown in Figure 5. In patterned growth especially, the C2 H2 to NH3 ratio was found to be important in achieving amorphous carbon free growth. During growth, the C2 H2 is continuously being decomposed by the plasma (at a rate much faster than thermal pyrolysis) to form amorphous carbon on the substrate surface, and the role of NH3 in the plasma is to etch away this unwanted amorphous carbon. The N and H species in the NH3 plasma react with the amorphous carbon to form

Catalytic Synthesis of Carbon Nanotubes and Nanofibers

volatile C–N and C–H species. Hence, there exists an optimum condition, which was determined to be a flow ratio of 40:200 sccm of C2 H2 :NH3 at 700  C, where the production and etching of amorphous carbon is balanced, thus yielding substrates which are free of amorphous carbon. Teo et al. also found that individual vertical nanotubes, nucleated from Ni clusters of controlled size/volume, could be grown with a high degree of uniformity in terms of tip radius and height. The standard deviations in the tip diameter and height were found to be 4.1 and 6.3% of the average respectively [205]. This high degree of structural uniformity leads good uniformity in terms of field emission characteristics from adjacent nanotubes in an array [205 274]. Delzeit et al. performed a parametric study of vertically aligned carbon nanotube and nanofiber growth in an inductively coupled plasma [149]. The authors used thin film Fe catalyst which was deposited on an aluminum layer on Si substrates, and the nanofibers/nanotubes were grown using CH4 , H2 , and Ar gases. By varying the process conditions, the authors found a clear transition between two structural forms of carbon filaments grown (i.e., the nanotube with parallel graphene walls and nanofiber with walls which were inclined to the filament axis). Until then, it was known that some plasma processes produced nanotubes (which were mostly bamboo), but other plasma processes produced nanofibers (with angled graphene layers but mostly hollow inside). The authors conclusively found that it was the relative amount of hydrogen species in the gas phase which determined whether nanotube or nanofiber growth was favored. Nanofiber growth was favored under process conditions which contained a large amount of hydrogen because the hydrogen could terminate the large number of dangling bonds protruding from angled graphene layers. The lack of hydrogen (e.g., achieved by Ar dilution) in the gas phase produced nanotubes which have closed, tubular graphene shells with few dangling bonds. One of the most significant results recently is the demonstration of room temperature growth of nanofibers using rfPECVD [242]. The authors used a CH4 and H2 plasma with Ni powder catalyst to grow nonaligned nanofibers at room temperature and at 100 and 250  C. The authors claimed that instead of thermal energy, the energy of the plasma and induced rf heating of the catalyst particle allowed the formation of the carbon nanofibers at ambient temperature.

4.5. Growth of Branched Structures The growth of branched filaments allows direct “wiring” of nanotube/nanofiber structures. Y-branched nanotubes would also allow switching and rectification behavior over a network of wires, similar to neural networks in biological systems. The growth of Y-branched nanotubes and nanofibers has been reported by several groups using the catalytic CVD [275–280]. Using the catalytic methods, the Y-junctions are generally formed at high temperatures (1000–1100  C) [277 279] or by using mixed metal or Cu catalysts [275 276 280–282]. For example, Chambers et al. found that branching is promoted when Co catalyst is alloyed with 2% Cu [275 276] while Li et al. [280] observed branching when the catalyst is doped with Si or Ca. Another method for fabricating Y-junctions is to use a nanochannel

Catalytic Synthesis of Carbon Nanotubes and Nanofibers

Figure 9. Branched structures as a result of a rapid drop in temperature from 700 to 550  C during growth. The drop in temperature caused the Ni catalyst here (bright dot at the tip of the structure) to break up. After the temperature perturbation, the temperature was brought back to 700  C for normal growth.

alumina template that has been prepared such that the pores are Y-shaped [283 284]. The electrical measurements on these Y-junction nanotubes showed nonlinear conductance and reproducible rectification [284]. It is also possible to grow branched structures without catalyst additives, templates, or high temperatures. In fact, all that is required is a simple perturbation during deposition in order to promote the formation of smaller catalyst particles (to catalyze branches) from a single, larger catalyst (which forms the main “stem”). In Figure 9, the branched nanofiber structures were formed by growth using dc-PECVD of C2 H2 :NH3 with Ni catalyst at 700  C for 3 minutes, then rapidly decreasing the growth temperature from 700 to 550  C and then normal deposition at 700  C again for 3 minutes. As the solubility of carbon in the catalyst decreases at lower temperatures, the rapid drop in temperature causes the catalyst to become oversaturated and to split up to form more surface area to expel the carbon. After the catalyst has been broken up by the rapid drop in temperature, further growth at the normal temperature (700  C) is used to extend the length of the nanofiber branches.

5. SUMMARY Catalytic chemical vapor deposition is an extremely versatile technique for the production of carbon nanofibers and nanotubes. This chapter has shown that by controlling the catalyst and synthesis conditions, one can control many aspects of the growth, such as the structure (nanofiber vs nanotube), diameter, length, and alignment. The catalytic CVD technique can be adapted for mass production purposes or for the controlled growth of nanotubes/fibers at particular sites on a substrate for various applications.

GLOSSARY Carbon nanofiber The carbon nanofiber is a generic term used to describe filaments/whiskers of carbon which have diameters less than 500 nm. The term fiber usually implies that the structures have a high aspect ratio, and hence the lengths of nanofibers are usually in the range of few micrometers or more. Although the carbon nanotube (see next) can be classed as a form of nanofiber, the carbon nanotube is typically used to describe structures which are comprised of tubular graphene walls parallel to the fiber axis. Hence,

681 recently, the term carbon nanofiber is used to imply filaments which is are comprised of graphene layers which are stacked at an angle to the fiber axis (such as the herringbone, cup-stacked, or stacked type filaments). Also, carbon nanofibers can be used to describe filaments of disordered/amorphous carbon. Carbon nanotube The carbon nanotube is comprised of graphene sheets (containing hexagonally arranged, sp2 bonded carbon atoms) which have been rolled up to form a seamless tubular structure. The simplest form is known as the single wall carbon nanotube in which the structure consists of a single graphene shell/tube. Multiwall carbon nanotubes consist of multiple, concentric tubular graphene shells.

ACKNOWLEDGMENTS The authors thank M. Castignolles for the TEM image in Figure 3c, D. G. Hasko for use of electron beam lithography equipment to prepare Figure 5, and R. Lacerda for sample preparation. We also thank G. A. J. Amaratunga for invaluable discussions and help. K.B.K.T. acknowledges the support of Christ’s College Cambridge. C.S. acknowledges the financial support of SIRIM Berhad (Malaysia), the Cambridge Commonwealth Trust, Selwyn College, and the Lundgren Fund.

REFERENCES 1. R. T. K. Baker, M. A. Barber, P. S. Harris, F. S. Feates, and R. J. Waite, J. Catal. 26, 51 (1972). 2. R. T. K. Baker, P. S. Harris, R. B. Thomas, and R. J. Waite, J. Catal. 30, 86 (1973). 3. R. T. K. Baker, Carbon 27, 315 (1989). 4. M. Endo, Chemtech 18, 568 (1988). 5. N. M. Rodriguez, A. Chambers, and R. T. K. Baker, Langmuir 11, 3862 (1995). 6. J. Gavillet, A. Loiseau, C. Journet, F. Willaime, F. Ducastelle, and J. C. Charlier, Phys. Rev. Lett. 8727, 275504 (2001). 7. C. A. Bessel, K. Laubernds, N. M. Rodriguez, and R. T. K. Baker, J. Phys. Chem. B 105, 1115 (2001). 8. M. Endo, Y. A. Kim, T. Hayashi, Y. Fukai, K. Oshida, M. Terrones, T. Yanagisawa, S. Higaki, and M. S. Dresselhaus, Appl. Phys. Lett. 80, 1267 (2002). 9. S. Iijima, Nature 354, 56 (1991). 10. http://www.fibrils.com, http://www.apsci.com/home.html, http:// www.nanomirae.com/eng/main.htm, http://www.sdk.co.jp. 11. http://www.personal.rdg.ac.uk/∼scsharip/tubes.htm. 12. D. S. Bethune, C. H. Kiang, M. S. Devries, G. Gorman, R. Savoy, J. Vazquez, and R. Beyers, Nature 363, 605 (1993). 13. S. Iijima and T. Ichihashi, Nature 363, 603 (1993). 14. Y. Saito, T. Yoshikawa, S. Bandow, M. Tomita, and T. Hayashi, Phys. Rev. B 48, 1907 (1993). 15. M. Endo, K. Takeuchi, T. Hiraoka, T. Furuta, T. Kasai, X. Sun, C. H. Kiang, and M. S. Dresselhaus, J. Phys. Chem. Solids 58, 1707 (1997). 16. C. H. Kiang, M. Endo, P. M. Ajayan, G. Dresselhaus, and M. S. Dresselhaus, Phys. Rev. Lett. 81, 1869 (1998). 17. M. S. Dresselhaus and M. Endo, in “Carbon Nanotubes Synthesis, Structure, Properties and Applications” (M. S. Dresselhaus, G. Dresselhaus, and P. Avouris, Eds.), p. 11. Springer-Verlag, New York, 2001.

682 18. A. Bachtold, C. Strunk, J. P. Salvetat, J. M. Bonard, L. Forro, T. Nussbaumer, and C. Schonenberger, Nature 397, 673 (1999). 19. M. Endo, S. Iijima, and M. S. Dresselhaus, “Carbon Nanotubes.” Elsevier Science, Oxford, 1996. 20. T. W. Odom, J. L. Huang, P. Kim, and C. M. Lieber, Nature 391, 62 (1998). 21. J. W. G. Wildoer, L. C. Venema, A. G. Rinzler, R. E. Smalley, and C. Dekker, Nature 391, 59 (1998). 22. L. Forro and C. Schnonenberger, in “Carbon Nanotubes Synthesis, Structure, Properties and Applications” (M. S. Dresselhaus, G. Dresselhaus, and P. Avouris, Eds.), p. 329. Springer-Verlag, New York, 2001. 23. R. Saito, G. Dresselhaus, and M. S. Dresselhaus, “Physical Properties of Carbon Nanotubes.” Imperial College Press, London, 1998. 24. P. G. Collins and P. Avouris, Sci. Am. 283, 62 (2000). 25. E. W. Wong, P. E. Sheehan, and C. M. Lieber, Science 277, 1971 (1997). 26. M. F. Yu, O. Lourie, M. J. Dyer, K. Moloni, T. F. Kelly, and R. S. Ruoff, Science 287, 637 (2000). 27. M. F. Yu, B. S. Files, S. Arepalli, and R. S. Ruoff, Phys. Rev. Lett. 84, 5552 (2000). 28. J. Hone, M. Whitney, C. Piskoti, and A. Zettl, Phys. Rev. B 59, R2514 (1999). 29. P. Kim, L. Shi, A. Majumdar, and P. L. McEuen, Phys. Rev. Lett. 8721, 215502 (2001). 30. A. Thess, R. Lee, P. Nikolaev, H. J. Dai, P. Petit, J. Robert, C. H. Xu, Y. H. Lee, S. G. Kim, A. G. Rinzler, D. T. Colbert, G. E. Scuseria, D. Tomanek, J. E. Fischer, and R. E. Smalley, Science 273, 483 (1996). 31. C. Schonenberger, A. Bachtold, C. Strunk, J. P. Salvetat, and L. Forro, Appl. Phys. A 69, 283 (1999). 32. W. A. de Heer and R. Martel, Phys. World 13, 49 (2000). 33. S. Frank, P. Poncharal, Z. L. Wang, and W. A. de Heer, Science 280, 1744 (1998). 34. http://www.pa.msu.edu/cmp/csc/ntproperties/, compiled by T. A. Adams II. 35. R. H. Baughman, A. A. Zakhidov, and W. A. de Heer, Science 297, 787 (2002). 36. P. J. F. Harris, “Carbon Nanotubes and Related Structures. New Materials for the Twenty-first Century.” Cambridge Univ. Press, Cambridge, UK, 1999. 37. T. W. Ebbesen, Phys. Today 49, 26 (1996). 38. H. J. Dai, Surf. Sci. 500, 218 (2002). 39. P. M. Ajayan, Chem. Rev. 99, 1787 (1999). 40. P. M. Ajayan and O. Z. Zhou, Carbon Nanotubes Topics Appl. Phys. 80, 391 (2001). 41. C. N. R. Rao, B. C. Satishkumar, A. Govindaraj, and M. Nath, Chem. Phys. Chem. 2, 78 (2001). 42. L. M. Dai and A. W. H. Mau, Adv. Mater. 13, 899 (2001). 43. W. B. Choi, D. S. Chung, J. H. Kang, H. Y. Kim, Y. W. Jin, I. T. Han, Y. H. Lee, J. E. Jung, N. S. Lee, G. S. Park, and J. M. Kim, Appl. Phys. Lett. 75, 3129 (1999). 44. A. G. Rinzler, J. H. Hafner, P. Nikolaev, L. Lou, S. G. Kim, D. Tomanek, P. Nordlander, D. T. Colbert, and R. E. Smalley, Science 269, 1550 (1995). 45. I. T. Han, H. J. Kim, Y. J. Park, N. Lee, J. E. Jang, J. W. Kim, J. E. Jung, and J. M. Kim, Appl. Phys. Lett. 81, 2070 (2002). 46. Y. H. Lee, Y. T. Jang, D. H. Kim, J. H. Ahn, and B. K. Ju, Adv. Mater. 13, 479 (2001). 47. D. S. Chung, S. H. Park, H. W. Lee, J. H. Choi, S. N. Cha, J. W. Kim, J. E. Jang, K. W. Min, S. H. Cho, M. J. Yoon, J. S. Lee, C. K. Lee, J. H. Yoo, J. M. Kim, J. E. Jung, Y. W. Jin, Y. J. Park, and J. B. You, Appl. Phys. Lett. 80, 4045 (2002). 48. W. B. Choi, Y. W. Jin, H. Y. Kim, S. J. Lee, M. J. Yun, J. H. Kang, Y. S. Choi, N. S. Park, N. S. Lee, and J. M. Kim, Appl. Phys. Lett. 78, 1547 (2001).

Catalytic Synthesis of Carbon Nanotubes and Nanofibers 49. G. Z. Yue, Q. Qiu, B. Gao, Y. Cheng, J. Zhang, H. Shimoda, S. Chang, J. P. Lu, and O. Zhou, Appl. Phys. Lett. 81, 355 (2002). 50. N. deJonge, Y. Lanny, K. Schoots, and T. H. Ooesterkamp, Nature 420, 393 (2002). 51. R. Rosen, W. Simendinger, C. Debbault, H. Shimoda, L. Fleming, B. Stoner, and O. Zhou, Appl. Phys. Lett. 76, 1668 (2000). 52. http://www.oxfordxtg.com/products/coldcath.htm, Oxford Instruments plc., 2002. 53. C. J. Lee, T. J. Lee, and J. Park, Chem. Phys. Lett. 340, 413 (2001). 54. M. A. Guillorn, A. V. Melechko, V. I. Merkulov, E. D. Ellis, C. L. Britton, M. L. Simpson, D. H. Lowndes, and L. R. Baylor, Appl. Phys. Lett. 79, 3506 (2001). 55. H. J. Dai, J. H. Hafner, A. G. Rinzler, D. T. Colbert, and R. E. Smalley, Nature 384, 147 (1996). 56. H. Nishijima, S. Kamo, S. Akita, Y. Nakayama, K. I. Hohmura, S. H. Yoshimura, and K. Takeyasu, Appl. Phys. Lett. 74, 4061 (1999). 57. C. V. Nguyen, K. J. Chao, R. M. D. Stevens, L. Delzeit, A. Cassell, J. Han, and M. Meyyappan, Nanotechnology 12, 363 (2001). 58. http://www.piezomax.com, nPoint, 2002. 59. S. J. Tans, A. R. M. Verschueren, and C. Dekker, Nature 393, 49 (1998). 60. R. Martel, T. Schmidt, H. R. Shea, T. Hertel, and P. Avouris, Appl. Phys. Lett. 73, 2447 (1998). 61. V. Derycke, R. Martel, J. Appenzeller, and P. Avouris, Nano Lett. 1, 453 (2001). 62. A. Bachtold, P. Hadley, T. Nakanishi, and C. Dekker, Science 294, 1317 (2001). 63. J. Kong, N. R. Franklin, C. W. Zhou, M. G. Chapline, S. Peng, K. J. Cho, and H. J. Dai, Science 287, 622 (2000). 64. P. G. Collins, K. Bradley, M. Ishigami, and A. Zettl, Science 287, 1801 (2000). 65. K. Tsukagoshi, B. W. Alphenaar, and H. Ago, Nature 401, 572 (1999). 66. T. W. Tombler, C. W. Zhou, L. Alexseyev, J. Kong, H. J. Dai, L. Lei, C. S. Jayanthi, M. J. Tang, and S. Y. Wu, Nature 405, 769 (2000). 67. K. H. An, W. S. Kim, Y. S. Park, Y. C. Choi, S. M. Lee, D. C. Chung, D. J. Bae, S. C. Lim, and Y. H. Lee, Adv. Mater. 13, 497 (2001). 68. E. Frackowiak, K. Metenier, V. Bertagna, and F. Beguin, Appl. Phys. Lett. 77, 2421 (2000). 69. C. M. Niu, E. K. Sichel, R. Hoch, D. Moy, and H. Tennent, Appl. Phys. Lett. 70, 1480 (1997). 70. M. Hughes, M. S. P. Shaffer, A. C. Renouf, C. Singh, G. Z. Chen, J. Fray, and A. H. Windle, Adv. Mater. 14, 382 (2002). 71. M. Hughes, G. Z. Chen, M. S. P. Shaffer, D. J. Fray, and A. H. Windle, Chem. Mater. 14, 1610 (2002). 72. E. Frackowiak and F. Beguin, Carbon 40, 1775 (2002). 73. M. Endo, Y. A. Kim, T. Hayashi, K. Nishimura, T. Matusita, K. Miyashita, and M. S. Dresselhaus, Carbon 39, 1287 (2001). 74. J. Li, R. Stevens, L. Delzeit, H. T. Ng, A. Cassell, J. Han, and M. Meyyappan, Appl. Phys. Lett. 81, 910 (2002). 75. M. A. Guillorn, T. E. McKnight, A. Melechko, V. I. Merkulov, P. F. Britt, D. W. Austin, D. H. Lowndes, and M. L. Simpson, J. Appl. Phys. 91, 3824 (2002). 76. G. L. Che, B. B. Lakshmi, E. R. Fisher, and C. R. Martin, Nature 393, 346 (1998). 77. E. S. Steigerwalt, G. A. Deluga, D. E. Cliffel, and C. M. Lukehart, J. Phys. Chem. B 105, 8097 (2001). 78. A. Chambers, C. Park, R. T. K. Baker, and N. M. Rodriguez, J. Phys. Chem. B 102, 4253 (1998). 79. C. Park, P. E. Anderson, A. Chambers, C. D. Tan, R. Hidalgo, and N. M. Rodriguez, J. Phys. Chem. B 103, 10572 (1999). 80. A. C. Dillon, K. M. Jones, T. A. Bekkedahl, C. H. Kiang, D. S. Bethune, and M. J. Heben, Nature 386, 377 (1997).

Catalytic Synthesis of Carbon Nanotubes and Nanofibers 81. B. Gao, A. Kleinhammes, X. P. Tang, C. Bower, L. Fleming, Y. Wu, and O. Zhou, Chem. Phys. Lett. 307, 153 (1999). 82. M. Hirscher, M. Becher, M. Haluska, U. Dettlaff-Weglikowska, A. Quintel, G. S. Duesberg, Y. M. Choi, P. Downes, M. Hulman, S. Roth, I. Stepanek, and P. Bernier, Appl. Phys. A 72, 129 (2001). 83. P. Kim and C. M. Lieber, Science 286, 2148 (1999). 84. R. H. Baughman, C. X. Cui, A. A. Zakhidov, Z. Iqbal, J. N. Barisci, G. M. Spinks, G. G. Wallace, A. Mazzoldi, D. De Rossi, A. G. Rinzler, O. Jaschinski, S. Roth, and M. Kertesz, Science 284, 1340 (1999). 85. J. Cumings and A. Zettl, Science 289, 602 (2000). 86. L. Zhang, A. V. Melechko, V. I. Merkulov, M. A. Guillorn, M. L. Simpson, D. H. Lowndes, and M. J. Doktycz, Appl. Phys. Lett. 81, 135 (2002). 87. J. Sandler, P. Werner, M. S. P. Shaffer, V. Demchuk, V. Altstadt, and A. H. Windle, Compos. A 33, 1033 (2002). 88. W. B. Downs and R. T. K. Baker, J. Mater. Res. 10, 625 (1995). 89. D. D. L. Chung, Carbon 39, 1119 (2001). 90. D. D. L. Chung, Carbon 39, 279 (2001). 91. J. Sandler, M. S. P. Shaffer, T. Prasse, W. Bauhofer, K. Schulte, and A. H. Windle, Polymer 40, 5967 (1999). 92. E. Kymakis and G. A. J. Amaratunga, Appl. Phys. Lett. 80, 112 (2002). 93. B. Vigolo, A. Penicaud, C. Coulon, C. Sauder, R. Pailler, C. Journet, P. Bernier, and P. Poulin, Science 290, 1331 (2000). 94. R. Andrews, D. Jacques, A. M. Rao, T. Rantell, F. Derbyshire, Y. Chen, J. Chen, and R. C. Haddon, Appl. Phys. Lett. 75, 1329 (1999). 95. P. A. O. Muisener, L. Clayton, J. D’Angelo, J. P. Harmon, A. K. Sikder, A. Kumar, A. M. Cassell, and M. Meyyappan, J. Mater. Res. 17, 2507 (2002). 96. P. Calvert, Nature 399, 210 (1999). 97. L. S. Schadler, S. C. Giannaris, and P. M. Ajayan, Appl. Phys. Lett. 73, 3842 (1998). 98. J. M. Bonard, N. Weiss, H. Kind, T. Stockli, L. Forro, K. Kern, and A. Chatelain, Adv. Mater. 13, 184 (2001). 99. L. Nilsson, O. Groening, C. Emmenegger, O. Kuettel, E. Schaller, L. Schlapbach, H. Kind, J. M. Bonard, and K. Kern, Appl. Phys. Lett. 76, 2071 (2000). 100. O. Groning, O. M. Kuttel, C. Emmenegger, P. Groning, and L. Schlapbach, J. Vac. Sci. Technol. B 18, 665 (2000). 101. T. W. Ebbesen and P. M. Ajayan, Nature 358, 220 (1992). 102. C. Journet, W. K. Maser, P. Bernier, A. Loiseau, M. L. delaChapelle, S. Lefrant, P. Deniard, R. Lee, and J. E. Fischer, Nature 388, 756 (1997). 103. “Chemistry and Physics of Carbon” (J. Philip, L. Walker, and P. A. Thrower, Eds.), Vol. 14. Dekker, New York/Basel, 1978. 104. R. T. K. Baker and P. S. Harris, in “Chemistry and Physics of Carbon” (J. P. L. Walker and P. A. Thrower, Eds.), Vol. 14, p. 83. Dekker, New York/Basel, 1978. 105. M. S. Dresselhaus, G. Dresselhaus, K. Sugihara, I. L. Spain, and H. A. Goldberg, “Graphite Fibers and Filaments,” Vol. 5. Springer-Verlag, Berlin, 1988. 106. G. S. Duesberg, J. Muster, H. J. Byrne, S. Roth, and M. Burghard, Appl. Phys. A 69, 269 (1999). 107. A. G. Rinzler, J. Liu, H. Dai, P. Nikolaev, C. B. Huffman, F. J. Rodriguez-Macias, P. J. Boul, A. H. Lu, D. Heymann, D. T. Colbert, R. S. Lee, J. E. Fischer, A. M. Rao, P. C. Eklund, and R. E. Smalley, Appl. Phys. A 67, 29 (1998). 108. J. Kong, A. M. Cassell, and H. J. Dai, Chem. Phys. Lett. 292, 567 (1998). 109. J. H. Hafner, M. J. Bronikowski, B. R. Azamian, P. Nikolaev, A. G. Rinzler, D. T. Colbert, K. A. Smith, and R. E. Smalley, Chem. Phys. Lett. 296, 195 (1998). 110. S. S. Fan, M. G. Chapline, N. R. Franklin, T. W. Tombler, A. M. Cassell, and H. J. Dai, Science 283, 512 (1999).

683 111. Y. G. Zhang, A. L. Chang, J. Cao, Q. Wang, W. Kim, Y. M. Li, N. Morris, E. Yenilmez, J. Kong, and H. J. Dai, Appl. Phys. Lett. 79, 3155 (2001). 112. A. Ural, Y. Li, and H. Dai, Appl. Phys. Lett. 81, 3464 (2002). 113. Z. F. Ren, Z. P. Huang, J. W. Xu, J. H. Wang, P. Bush, M. P. Siegal, and P. N. Provencio, Science 282, 1105 (1998). 114. W. Z. Li, S. S. Xie, L. X. Qian, B. H. Chang, B. S. Zou, W. Y. Zhou, R. A. Zhao, and G. Wang, Science 274, 1701 (1996). 115. M. Terrones, N. Grobert, J. Olivares, J. P. Zhang, H. Terrones, K. Kordatos, W. K. Hsu, J. P. Hare, P. D. Townsend, K. Prassides, A. K. Cheetham, H. W. Kroto, and D. R. M. Walton, Nature 388, 52 (1997). 116. Z. F. Ren, Z. P. Huang, D. Z. Wang, J. G. Wen, J. W. Xu, J. H. Wang, L. E. Calvet, J. Chen, J. F. Klemic, and M. A. Reed, Appl. Phys. Lett. 75, 1086 (1999). 117. V. I. Merkulov, D. H. Lowndes, Y. Y. Wei, G. Eres, and E. Voelkl, Appl. Phys. Lett. 76, 3555 (2000). 118. K. B. K. Teo, M. Chhowalla, G. A. J. Amaratunga, W. I. Milne, D. G. Hasko, G. Pirio, P. Legagneux, F. Wyczisk, and D. Pribat, Appl. Phys. Lett. 79, 1534 (2001). 119. Z. W. Pan, S. S. Xie, B. H. Chang, C. Y. Wang, L. Lu, W. Liu, M. Y. Zhou, and W. Z. Li, Nature 394, 631 (1998). 120. L. An, J. M. Owens, L. E. McNeil, and J. Liu, J. Am. Chem. Soc. 124, 13688 (2002). 121. C. H. Choi, W. Kim, D. Wang, and H. Dai, J. Phys. Chem. B, in press. 122. H. J. Dai, J. Kong, C. W. Zhou, N. Franklin, T. Tombler, A. Cassell, S. S. Fan, and M. Chapline, J. Phys. Chem. B 103, 11246 (1999). 123. Y. C. Choi, Y. M. Shin, Y. H. Lee, B. S. Lee, G. S. Park, W. B. Choi, N. S. Lee, and J. M. Kim, Appl. Phys. Lett. 76, 2367 (2000). 124. N. M. Rodriguez, J. Mater. Res. 8, 3233 (1993). 125. H. G. Tennent, J. J. Barber, and R. Hoch, Hyperion Catalysis, Cambridge, MA, 1996. 126. S. B. Sinnott, R. Andrews, D. Qian, A. M. Rao, Z. Mao, E. C. Dickey, and F. Derbyshire, Chem. Phys. Lett. 315, 25 (1999). 127. G. G. Tibbetts, Appl. Phys. Lett. 42, 666 (1983). 128. A. Fonseca, K. Hernadi, P. Piedigrosso, J. F. Colomer, K. Mukhopadhyay, R. Doome, S. Lazarescu, L. P. Biro, P. Lambin, P. A. Thiry, D. Bernaerts, and J. B. Nagy, Appl. Phys. A 67, 11 (1998). 129. P. Piedigrosso, Z. Konya, J. F. Colomer, A. Fonseca, G. Van Tendeloo, and J. B. Nagy, Phys. Chem. Chem. Phys. 2, 163 (2000). 130. H. Q. Hou, A. K. Schaper, F. Weller, and A. Greiner, Chem. Mater. 14, 3990 (2002). 131. R. S. Wagner and W. C. Ellis, Appl. Phys. Lett. 4, 89 (1964). 132. R. S. Wagner, in “Whiskers Technology” (A. P. Levitt, Ed.), p. 47. Wiley, New York, 1970. 133. T. Baird, J. R. Fryer, and B. Grant, Carbon 12, 591 (1974). 134. A. Oberlin, M. Endo, and T. Koyama, J. Crystal Growth 32, 335 (1976). 135. A. Oberlin, M. Endo, and T. Koyama, Carbon 14, 133 (1976). 136. G. G. Tibbetts, J. Crystal Growth 66, 632 (1984). 137. H. J. Dai, A. G. Rinzler, P. Nikolaev, A. Thess, D. T. Colbert, and R. E. Smalley, Chem. Phys. Lett. 260, 471 (1996). 138. E. Boellaard, P. K. Debokx, A. Kock, and J. W. Geus, J. Catal. 96, 481 (1985). 139. P. K. Debokx, A. Kock, E. Boellaard, W. Klop, and J. W. Geus, J. Catal. 96, 454 (1985). 140. A. Kock, P. K. Debokx, E. Boellaard, W. Klop, and J. W. Geus, J. Catal. 96, 468 (1985). 141. E. Tracz, R. Scholz, and T. Borowiecki, Appl. Catal. 66, 133 (1990). 142. N. Krishnankutty, N. M. Rodriguez, and R. T. K. Baker, J. Catal. 158, 217 (1996). 143. A. Chambers and R. T. K. Baker, J. Phys. Chem. B 101, 1621 (1997). 144. M. S. Kim, N. M. Rodriguez, and R. T. K. Baker, J. Catal. 134, 253 (1992).

684 145. C. Park, E. S. Engel, A. Crowe, T. R. Gilbert, and N. M. Rodriguez, Langmuir 16, 8050 (2000). 146. C. Singh, T. Quested, C. B. Boothroyd, P. Thomas, I. A. Kinloch, A. I. Abou-Kandil, and A. H. Windle, J. Phys. Chem. B 106, 10915 (2002). 147. H. Terrones, T. Hayashi, M. Munoz-Navia, M. Terrones, Y. A. Kim, N. Grobert, R. Kamalakaran, J. Dorantes-Davila, R. Escudero, M. S. Dresselhaus, and M. Endo, Chem. Phys. Lett. 343, 241 (2001). 148. P. E. Nolan, D. C. Lynch, and A. H. Cutler, J. Phys. Chem. B 102, 4165 (1998). 149. L. Delzeit, I. McAninch, B. A. Cruden, D. Hash, B. Chen, J. Han, and M. Meyyappan, J. Appl. Phys. 91, 6027 (2002). 150. H. W. Kroto, J. R. Heath, S. C. Obrien, R. F. Curl, and R. E. Smalley, Nature 318, 162 (1985). 151. J. Kong, H. T. Soh, A. M. Cassell, C. F. Quate, and H. J. Dai, Nature 395, 878 (1998). 152. A. M. Cassell, J. A. Raymakers, J. Kong, and H. J. Dai, J. Phys. Chem. B 103, 6484 (1999). 153. M. Su, B. Zheng, and J. Liu, Chem. Phys. Lett. 322, 321 (2000). 154. B. Kitiyanan, W. E. Alvarez, J. H. Harwell, and D. E. Resasco, Chem. Phys. Lett. 317, 497 (2000). 155. W. E. Alvarez, B. Kitiyanan, A. Borgna, and D. E. Resasco, Carbon 39, 547 (2001). 156. J. E. Herrera, L. Balzano, A. Borgna, W. E. Alvarez, and D. E. Resasco, J. Catal. 204, 129 (2001). 157. W. E. Alvarez, F. Pompeo, J. E. Herrera, L. Balzano, and D. E. Resasco, Chem. Mater. 14, 1853 (2002). 158. A. Kukovecz, Z. Konya, N. Nagaraju, I. Willems, A. Tamasi, A. Fonseca, J. B. Nagy, and I. Kiricsi, Phys. Chem. Chem. Phys. 2, 3071 (2000). 159. Y. M. Li, W. Kim, Y. G. Zhang, M. Rolandi, D. W. Wang, and H. J. Dai, J. Phys. Chem. B 105, 11424 (2001). 160. C. L. Cheung, A. Kurtz, H. Park, and C. M. Lieber, J. Phys. Chem. B 106, 2429 (2002). 161. J. F. Colomer, G. Bister, I. Willems, Z. Konya, A. Fonseca, G. Van Tendeloo, and J. B. Nagy, Chem. Commun. 1343 (1999). 162. J. Geng, C. Singh, D. Shephard, M. Shaffer, B. F. G. Johnson, and A. H. Windle, Chem. Commun. 2666 (2002). 163. V. Ivanov, J. B. Nagy, P. Lambin, A. Lucas, X. B. Zhang, X. F. Zhang, D. Bernaerts, G. Vantendeloo, S. Amelinckx, and J. Vanlanduyt, Chem. Phys. Lett. 223, 329 (1994). 164. L. F. Sun, J. M. Mao, Z. W. Pan, B. H. Chang, W. Y. Zhou, G. Wang, L. X. Qian, and S. S. Xie, Appl. Phys. Lett. 74, 644 (1999). 165. H. J. Dai, Phys. World 13, 43 (2000). 166. S. Amelinckx, X. B. Zhang, D. Bernaerts, X. F. Zhang, V. Ivanov, and J. B. Nagy, Science 265, 635 (1994). 167. C. J. Lee and J. Park, Appl. Phys. Lett. 77, 3397 (2000). 168. W. Z. Li, J. G. Wen, Y. Tu, and Z. F. Ren, Appl. Phys. A 73, 259 (2001). 169. Y. Li, J. Liu, Y. Q. Wang, and Z. L. Wang, Chem. Mater. 13, 1008 (2001). 170. H. Ago, S. Ohshima, K. Uchida, and M. Yumura, J. Phys. Chem. B 105, 10453 (2001). 171. L. Delzeit, B. Chen, A. Cassell, R. Stevens, C. Nguyen, and M. Meyyappan, Chem. Phys. Lett. 348, 368 (2001). 172. P. Nikolaev, M. J. Bronikowski, R. K. Bradley, F. Rohmund, D. T. Colbert, K. A. Smith, and R. E. Smalley, Chem. Phys. Lett. 313, 91 (1999). 173. M. Joseyacaman, M. Mikiyoshida, L. Rendon, and J. G. Santiesteban, Appl. Phys. Lett. 62, 657 (1993). 174. M. Endo, K. Takeuchi, S. Igarashi, K. Kobori, M. Shiraishi, and H. W. Kroto, J. Phys. Chem. Solids 54, 1841 (1993). 175. A. M. Benito, Y. Maniette, E. Munoz, and M. T. Martinez, Carbon 36, 681 (1998). 176. N. R. Franklin, Y. M. Li, R. J. Chen, A. Javey, and H. J. Dai, Appl. Phys. Lett. 79, 4571 (2001).

Catalytic Synthesis of Carbon Nanotubes and Nanofibers 177. N. R. Franklin and H. J. Dai, Adv. Mater. 12, 890 (2000). 178. V. I. Merkulov, A. V. Melechko, M. A. Guillorn, D. H. Lowndes, and M. L. Simpson, Chem. Phys. Lett. 350, 381 (2001). 179. D. S. Y. Hsu, Appl. Phys. Lett. 80, 2988 (2002). 180. C. Bower, W. Zhu, D. Shalom, D. Lopez, L. H. Chen, P. L. Gammel, and S. Jin, Appl. Phys. Lett. 80, 3820 (2002). 181. G. Pirio, P. Legagneux, D. Pribat, K. B. K. Teo, M. Chhowalla, G. A. J. Amaratunga, and W. I. Milne, Nanotechnology 13, 1 (2002). 182. M. Chhowalla, K. B. K. Teo, C. Ducati, N. L. Rupesinghe, G. A. J. Amaratunga, A. C. Ferrari, D. Roy, J. Robertson, and W. I. Milne, J. Appl. Phys. 90, 5308 (2001). 183. C. Bower, O. Zhou, W. Zhu, D. J. Werder, and S. H. Jin, Appl. Phys. Lett. 77, 2767 (2000). 184. J. I. Sohn, S. Lee, Y. H. Song, S. Y. Choi, K. I. Cho, and K. S. Nam, Appl. Phys. Lett. 78, 901 (2001). 185. S. M. Sze, “VLSI Technology,” 2nd ed., p. 309. McGraw-Hill, New York, 1988. 186. K. B. K. Teo, M. Chhowalla, G. A. J. Amaratunga, W. I. Milne, G. Pirio, P. Legagneux, F. Wyczisk, D. Pribat, and D. G. Hasko, Appl. Phys. Lett. 80, 2011 (2002). 187. A. M. Rao, D. Jacques, R. C. Haddon, W. Zhu, C. Bower, and S. Jin, Appl. Phys. Lett. 76, 3813 (2000). 188. T. de los Arcos, F. Vonau, M. G. Garnier, V. Thommen, H. G. Boyen, P. Oelhafen, M. Duggelin, D. Mathis, and R. Guggenheim, Appl. Phys. Lett. 80, 2383 (2002). 189. M. S. Dresselhaus, G. Dresselhaus, K. Sugihara, I. L. Spain, and H. A. Goldberg, “Graphite Fibers and Filaments” Vol. 5, SpringerVerlag, Berlin, 1988. 190. C. Emmenegger, P. Mauron, A. Zuttel, C. Nutzenadel, A. Schneuwly, R. Gallay, and L. Schlapbach, Appl. Surf. Sci. 162, 452 (2000). 191. H. Kind, J. M. Bonard, C. Emmenegger, L. O. Nilsson, K. Hernadi, E. Maillard-Schaller, L. Schlapbach, L. Forro, and K. Kern, Adv. Mater. 11, 1285 (1999). 192. R. T. K. Baker, J. R. Alonzo, J. A. Dumesic, and D. J. C. Yates, J. Catal. 77, 74 (1982). 193. S. K. Hwang, K. D. Lee, and K. H. Lee, Jpn. J. Appl. Phys. 2 Lett. 40, L580 (2001). 194. A. M. Cassell, S. Verma, L. Delzeit, M. Meyyappan, and J. Han, Langmuir 17, 260 (2001). 195. B. Chen, G. Parker, J. Han, M. Meyyappan, and A. M. Cassell, Chem. Mater. 14, 1891 (2002). 196. Y. Tu, Z. P. Huang, D. Z. Wang, J. G. Wen, and Z. F. Ren, Appl. Phys. Lett. 80, 4018 (2002). 197. J. F. Colomer, C. Stephan, S. Lefrant, G. Van Tendeloo, I. Willems, Z. Konya, A. Fonseca, C. Laurent, and J. B. Nagy, Chem. Phys. Lett. 317, 83 (2000). 198. J. M. Bonard, T. Stockli, O. Noury, and A. Chatelain, Appl. Phys. Lett. 78, 2775 (2001). 199. K. Matsumoto, S. Kinosita, Y. Gotoh, T. Uchiyama, S. Manalis, and C. Quate, Appl. Phys. Lett. 78, 539 (2001). 200. J. H. Hafner, C. L. Cheung, and C. M. Lieber, Nature 398, 761 (1999). 201. E. B. Cooper, S. R. Manalis, H. Fang, H. Dai, K. Matsumoto, S. C. Minne, T. Hunt, and C. F. Quate, Appl. Phys. Lett. 75, 3566 (1999). 202. Y. Y. Wei, G. Eres, V. I. Merkulov, and D. H. Lowndes, Appl. Phys. Lett. 78, 1394 (2001). 203. M. P. Siegal, D. L. Overmyer, and P. P. Provencio, Appl. Phys. Lett. 80, 2171 (2002). 204. C. Ducati, I. Alexandrou, M. Chhowalla, G. A. J. Amaratunga, and J. Robertson, J. Appl. Phys. 92, 3299 (2002). 205. K. B. K. Teo, S. B. Lee, M. Chhowalla, D. G. Hasko, H. Ahmed, G. A. J. Amaratunga, W. I. Milne, V. Semet, V. T. Binh, O. Groening, M. Castignolles, A. Loiseau, P. Legagneux, G. Pirio, and D. Pribat, Nanotechnology, 14, 204 (2003).

Catalytic Synthesis of Carbon Nanotubes and Nanofibers 206. Y. Chen, Z. L. Wang, J. S. Yin, D. J. Johnson, and R. H. Prince, Chem. Phys. Lett. 272, 178 (1997). 207. Y. Chen, S. Patel, Y. G. Ye, S. T. Shaw, and L. P. Luo, Appl. Phys. Lett. 73, 2119 (1998). 208. Z. P. Huang, J. W. Wu, Z. F. Ren, J. H. Wang, M. P. Siegal, and P. N. Provencio, Appl. Phys. Lett. 73, 3845 (1998). 209. T. Sato, D. G. Hasko, and H. Ahmed, J. Vac. Sci. Technol. B 15, 45 (1997). 210. T. Koyama, Carbon 10, 757 (1972). 211. T. Kato, K. Kusakabe, and S. Morooka, J. Mater. Sci. Lett. 11, 674 (1992). 212. L. J. Ci, Y. H. Li, B. Q. Wei, J. Liang, C. L. Xu, and D. H. Wu, Carbon 38, 1933 (2000). 213. M. Endo, K. Takeuchi, K. Kobori, K. Takahashi, H. W. Kroto, and A. Sarkar, Carbon 33, 873 (1995). 214. B. C. Satishkumar, A. Govindaraj, and C. N. R. Rao, Chem. Phys. Lett. 307, 158 (1999). 215. R. Sen, A. Govindaraj, and C. N. R. Rao, Chem. Phys. Lett. 267, 276 (1997). 216. S. M. Huang, L. M. Dai, and A. W. H. Mau, J. Phys. Chem. B 103, 4223 (1999). 217. Y. Y. Yang, S. M. Huang, H. Z. He, A. W. H. Mau, and L. M. Dai, J. Am. Chem. Soc. 121, 10832 (1999). 218. D. C. Li, L. M. Dai, S. M. Huang, A. W. H. Mau, and Z. L. Wang, Chem. Phys. Lett. 316, 349 (2000). 219. C. Singh, M. S. P. Shaffer, and A. H. Windle, Carbon 41, 363 (2003). 220. R. Sen, A. Govindaraj, and C. N. R. Rao, Chem. Mater. 9, 2078 (1997). 221. R. Andrews, D. Jacques, A. M. Rao, F. Derbyshire, D. Qian, X. Fan, E. C. Dickey, and J. Chen, Chem. Phys. Lett. 303, 467 (1999). 222. R. Kamalakaran, M. Terrones, T. Seeger, P. Kohler-Redlich, M. Ruhle, Y. A. Kim, T. Hayashi, and M. Endo, Appl. Phys. Lett. 77, 3385 (2000). 223. M. Mayne, N. Grobert, M. Terrones, R. Kamalakaran, M. Ruhle, H. W. Kroto, and D. R. M. Walton, Chem. Phys. Lett. 338, 101 (2001). 224. Z. J. Zhang, B. Q. Wei, G. Ramanath, and P. M. Ajayan, Appl. Phys. Lett. 77, 3764 (2000). 225. B. Q. Wei, R. Vajtai, Y. Jung, J. Ward, R. Zhang, G. Ramanath, and P. M. Ajayan, Nature 416, 495 (2002). 226. B. Q. Wei, Z. J. Zhang, G. Ramanath, and P. M. Ajayan, Appl. Phys. Lett. 77, 2985 (2000). 227. A. Y. Cao, L. J. Ci, D. J. Li, B. Q. Wei, C. L. Xu, J. Liang, and D. H. Wu, Chem. Phys. Lett. 335, 150 (2001). 228. R. Vajtai, K. Kordas, B. Q. Wei, J. Bekesi, S. Leppavuori, T. F. George, and P. M. Ajayan, Mater. Sci. Eng. C 19, 271 (2002). 229. B. Q. Wei, R. Vajtai, Z. J. Zhang, G. Ramanath, and P. M. Ajayan, J. Nanosci. Nanotechnol. 1, 35 (2001). 230. C. N. R. Rao, A. Govindaraj, R. Sen, and B. C. Satishkumar, Mater. Res. Innov. 2, 128 (1998). 231. H. M. Cheng, F. Li, G. Su, H. Y. Pan, L. L. He, X. Sun, and M. S. Dresselhaus, Appl. Phys. Lett. 72, 3282 (1998). 232. H. W. Zhu, C. L. Xu, D. H. Wu, B. Q. Wei, R. Vajtai, and P. M. Ajayan, Science 296, 884 (2002). 233. H. W. Zhu, B. Jiang, C. L. Xu, and D. H. Wu, Chem. Commun. 1858 (2002). 234. B. Q. Wei, R. Vajtai, Y. Y. Choi, P. M. Ajayan, H. W. Zhu, C. L. Xu, and D. H. Wu, Nano Lett. 2, 1105 (2002). 235. M. J. Bronikowski, P. A. Willis, D. T. Colbert, K. A. Smith, and R. E. Smalley, J. Vac. Sci. Technol. A 19, 1800 (2001). 236. HSC Chemistry Version 4.1, Outokumpu Research Oy, 2002. 237. “Chemistry and Physics of Carbon” (J. Philip, L. Walker, and P. A. Thrower, Eds.), Vol. 7. Dekker, New York, 1971. 238. V. Ivanov, A. Fonseca, J. B. Nagy, A. Lucas, P. Lambin, D. Bernaerts, and X. B. Zhang, Carbon 33, 1727 (1995).

685 239. J.-B. Donnet and R. C. Bansal, “Carbon Fibers,” 2nd ed., Vol. 10. Dekker, New York, 1990. 240. C. Singh, M. S. P. Shaffer, I. Kinloch, and A. H. Windle, Physica B 323, 339 (2002). 241. J. M. Bonard, M. Croci, F. Conus, T. Stockli, and A. Chatelain, Appl. Phys. Lett. 81, 2836 (2002). 242. B. O. Boskovic, V. Stolojan, R. U. A. Khan, S. Haq, and S. R. P. Silva, Nature Mater. 1, 165 (2002). 243. S. Ohshima, H. Ago, H. Inoue, and M. Yumura, New Diam. Front. Carbon Technol. 11, 437 (2001). 244. Y. Wang, F. Wei, G. Luo, H. Yu, and G. Gu, Chem. Phys. Lett. 364, 568 (2002). 245. Y. Wang, F. Wei, G. Gu, and H. Yu, Physica B 322, 327 (2002). 246. A. Weidenkaff, S. G. Ebbinghaus, P. Mauron, A. Reller, Y. Zhang, and A. Zuttel, Mater. Sci. Eng. C 1ss9, 119 (2002). 247. A. Huczko, Appl. Phys. A 70, 365 (2000). 248. T. Kyotani, L. F. Tsai, and A. Tomita, Chem. Mater. 7, 1427 (1995). 249. G. Che, B. B. Lakshmi, C. R. Martin, E. R. Fisher, and R. S. Ruoff, Chem. Mater. 10, 260 (1998). 250. J. Li, C. Papadopoulos, J. M. Xu, and M. Moskovits, Appl. Phys. Lett. 75, 367 (1999). 251. N. Wang, Z. K. Tang, G. D. Li, and J. S. Chen, Nature 408, 50 (2000). 252. Y. Avigal and R. Kalish, Appl. Phys. Lett. 78, 2291 (2001). 253. H. T. Soh, C. F. Quate, A. F. Morpurgo, C. M. Marcus, J. Kong, and H. J. Dai, Appl. Phys. Lett. 75, 627 (1999). 254. J. Kong, C. Zhou, A. Morpurgo, H. T. Soh, C. F. Quate, C. Marcus, and H. Dai, Appl. Phys. A 69, 305 (1999). 255. Y. H. Lee, Y. T. Jang, C. H. Choi, D. H. Kim, C. W. Lee, J. E. Lee, Y. S. Han, S. S. Yoon, J. K. Shin, S. T. Kim, E. K. Kim, and B. K. Ju, Adv. Mater. 13, 1371 (2001). 256. H. Conrads and M. Schmidt, Plasma Sources Sci. Technol. 9, 441 (2000). 257. C. Bower, W. Zhu, S. H. Jin, and O. Zhou, Appl. Phys. Lett. 77, 830 (2000). 258. L. C. Qin, D. Zhou, A. R. Krauss, and D. M. Gruen, Appl. Phys. Lett. 72, 3437 (1998). 259. N. Wang and B. D. Yao, Appl. Phys. Lett. 78, 4028 (2001). 260. M. Okai, T. Muneyoshi, T. Yaguchi, and S. Sasaki, Appl. Phys. Lett. 77, 3468 (2000). 261. Y. C. Choi, Y. M. Shin, S. C. Lim, D. J. Bae, Y. H. Lee, B. S. Lee, and D. C. Chung, J. Appl. Phys. 88, 4898 (2000). 262. H. Murakami, M. Hirakawa, C. Tanaka, and H. Yamakawa, Appl. Phys. Lett. 76, 1776 (2000). 263. K.-Y. Lee, K. Fujimoto, S. Ohkura, S. Honda, M. Katayama, T. Hirao, and K. Oura, Mater. Res. Soc. Symp. Proc. 675, W3.1 (2001). 264. Z. P. Huang, D. Z. Wang, J. G. Wen, M. Sennett, H. Gibson, and Z. F. Ren, Appl. Phys. A 74, 387 (2002). 265. J. G. Wen, Z. P. Huang, D. Z. Wang, J. H. Chen, S. X. Yang, Z. F. Ren, J. H. Wang, L. E. Calvet, J. Chen, J. F. Klemic, and M. A. Reed, J. Mater. Res. 16, 3246 (2001). 266. Y. Chen, D. T. Shaw, and L. P. Guo, Appl. Phys. Lett. 76, 2469 (2000). 267. S. H. Tsai, C. W. Chao, C. L. Lee, and H. C. Shih, Appl. Phys. Lett. 74, 3462 (1999). 268. V. I. Merkulov, M. A. Guillorn, D. H. Lowndes, M. L. Simpson, and E. Voelkl, Appl. Phys. Lett. 79, 1178 (2001). 269. V. I. Merkulov, A. V. Melechko, M. A. Guillorn, M. L. Simpson, D. H. Lowndes, J. H. Whealton, and R. J. Raridon, Appl. Phys. Lett. 80, 4816 (2002). 270. L. R. Baylor, V. I. Merkulov, E. D. Ellis, M. A. Guillorn, D. H. Lowndes, A. V. Melechko, M. L. Simpson, and J. H. Whealton, J. Appl. Phys. 91, 4602 (2002). 271. M. A. Guillorn, M. D. Hale, V. I. Merkulov, M. L. Simpson, G. Y. Eres, H. Cui, A. A. Puretzky, and D. B. Geohegan, Appl. Phys. Lett. 81, 2860 (2002).

686 272. V. I. Merkulov, D. H. Lowndes, and L. R. Baylor, J. Appl. Phys. 89, 1933 (2001). 273. K. B. K. Teo, M. Chhowalla, G. A. J. Amaratunga, W. I. Milne, G. Pirio, P. Legagneux, F. Wyczisk, J. Olivier, and D. Pribat, J. Vac. Sci. Technol. B 20, 116 (2002). 274. V. Semet, V. T. Binh, P. Vincent, D. Guillot, K. B. K. Teo, M. Chhowalla, G. A. J. Amaratunga, W. I. Milne, P. Legagneux, and D. Pribat, Appl. Phys. Lett. 81, 343 (2002). 275. A. Chambers, N. M. Rodriguez, and R. T. K. Baker, J. Phys. Chem. 99, 10581 (1995). 276. A. Chambers, N. M. Rodriguez, and R. T. K. Baker, J. Mater. Res. 11, 430 (1996). 277. B. C. Satishkumar, P. J. Thomas, A. Govindaraj, and C. N. R. Rao, Appl. Phys. Lett. 77, 2530 (2000).

Catalytic Synthesis of Carbon Nanotubes and Nanofibers 278. F. L. Deepak, A. Govindaraj, and C. N. R. Rao, Chem. Phys. Lett. 345, 5 (2001). 279. J. M. Ting and C. C. Chang, Appl. Phys. Lett. 80, 324 (2002). 280. W. Z. Li, J. G. Wen, and Z. F. Ren, Appl. Phys. Lett. 79, 1879 (2001). 281. B. Gan, J. Ahn, Q. Zhang, S. F. Yoon, Rusli, Q. F. Huang, H. Yang, M. B. Yu, and W. Z. Li, Diam. Relat. Mater. 9, 897 (2000). 282. B. Gan, J. Ahn, Q. Zhang, Rusli, S. F. Yoon, J. Yu, Q. F. Huang, K. Chew, V. A. Ligatchev, X. B. Zhang, and W. Z. Li, Chem. Phys. Lett. 333, 23 (2001). 283. J. Li, C. Papadopoulos, and J. Xu, Nature 402, 253 (1999). 284. C. Papadopoulos, A. Rakitin, J. Li, A. S. Vedeneev, and J. M. Xu, Phys. Rev. Lett. 85, 3476 (2000).

Encyclopedia of Nanoscience and Nanotechnology

www.aspbs.com/enn

Ceramic Nanoparticle Synthesis Xiangdong Feng Ferro Corporation, Ohio, USA

Michael Z. Hu Oak Ridge National Laboratory, Oak Ridge, Tennessee, USA

CONTENTS 1. Introduction 2. Methods for Categorizing Nanoparticle Synthesis 3. Descriptions of Major Synthesis Methods by Media 4. Summary Glossary References

1. INTRODUCTION Nanotechnology has been viewed as the “little big science” [1], which is considered the impetus for the next industrial revolution. The ability to work at the molecular level, atom by atom, to create large structures with fundamentally new properties and functions is the core of nanoscale science, engineering, and technology [2]. One of the building blocks of nanotechnology is nanostructured materials such as nanoparticles [3]. Nanoparticles may be defined according to their physical dimensions as particles with a size range of 1 to 100 nm [3, 4]. A typical nanoparticle most often consists of several small “primary particles” [5]. The size of nanoparticles usually refers to the “secondary” particle size. Primary particles agglomerate into these “secondary” particles [5]. The primary particles have a very high surface-to-volume ratio. The proportion of atoms on the surface and at the grain boundaries of a nanoparticle is significant. Nanoparticles have about 98, 40, and 10% of their atoms on the surfaces or at the grain boundaries if their primary particles are 1, 5, and 20 nm, respectively. The high proportion of surface and grain boundary atoms is responsible for the different properties (electronic, optical, electrical, magnetic, chemical, and mechanical) of nanoparticles with respect to the ISBN: 1-58883-057-8/$35.00 Copyright © 2004 by American Scientific Publishers All rights of reproduction in any form reserved.

bulk materials [6, 7]. This represents new opportunities in science, such as the change of properties and laws as the dimensions of the nanomaterials become comparable to, or smaller than, relevant length scales, such as the meanfree path or the coherence length. New material properties result when the surface and the grain boundary atoms represent a large proportion of the total number of atoms, or when processes become dominated by quantum effects, or— in general—by the transition from atomic and molecular behavior to condensed matter behavior [8]. The size range for nanoparticles may be more accurately defined as the size range from 1 nm to the largest nanometer dimension, where the ratio of the number of atoms on surfaces and at grain boundaries to those at the center is such that new physical, chemical, and biological properties occur, compared to bulk materials [9]. The upper limit of the nanoparticle size may be different from 100 nm, depending upon the sizes of the atoms and molecules in question. Nanoparticles have large surface areas [7] and high reactivity due to the unsaturated bonds on their pristine surfaces. Nanoparticles tend to react among themselves to form necking, and hence agglomerate into large secondary particles. This imposes extreme challenges for industry to produce large quantities of nanoparticles with identical properties inexpensively and reliably. Selecting the proper manufacturing methods for a specific application requires a thorough understanding of the fundamental aspects of nanoparticle formation. There are numerous reported synthetic methods for the production of nanoparticles [10]. The challenge is to control the nanoparticle size, size distribution, morphology, crystallinity, shape, and properties, to assemble the nanoparticles for a given purpose, and to make them from a variety of materials. The synthetic methods reported can be classified according to synthesis strategies, the branches of science involved in the process, sources of energy input, and the media in which nanoparticles are formed.

Encyclopedia of Nanoscience and Nanotechnology Edited by H. S. Nalwa Volume 1: Pages (687–726)

Ceramic Nanoparticle Synthesis

688

2. METHODS FOR CATEGORIZING NANOPARTICLE SYNTHESIS 2.1. By Synthesis Strategy Strategies are important for the synthesis of nanoparticles. There are two basic strategies used in the syntheses of nanoparticles: bottom up and top down [9, 11].

2.1.1. Bottom-Up Strategy This strategy is more in line with the basic strategy of nanoscience and technology. Here, nanoparticles are built up atom by atom, or molecule by molecule, at the will of their creators. Most nanoparticle syntheses belong to this category. A simple example of this approach is the flame synthesis of TiO2 from gaseous TiCl4 and oxygen according to TiCl4 + O2 → TiO2 + 2Cl2
It has been shown that nucleation and surface growth determine the primary particle size of titania [12–14]. Other studies [15, 16] show that coagulation of the primary particles of titania could also dominate particle growth during flame synthesis of TiO2 . The molecular TiO2 formed in the flame can grow into particles through nucleation and surface oxidation of TiCl4 , aggregation, or the combination of both mechanisms. Under certain processing conditions, either one of the mechanisms could dominate the overall particle growth.

2.1.2. Top-Down Strategy This strategy is most applied in the traditional particlemaking process. Here, nanoparticles are synthesized by breaking down bulk materials gradually into smaller sizes until they are nanosized. The end results could also be achieved by leaching out one type of molecule in a bulky homogeneous molecular solid mixture. Mechanical breakdown, such as high-energy ball milling, is a simple case of the top-down strategy. –Al2 O3 powders have been ball milled into nanocrystalline alumina powders with different surface areas of up to 100 m2 /g, depending on the milling atmosphere of air or inert gases [17]. Highenergy ball milling is useful for deriving nanocrystalline materials of high-temperature phases without going to extreme heat treatment, which would promote significant grain growth and surface area loss. For example, nanocrystalline –Al2 O3 with a high surface area was obtained through ball milling [18]. Another example of the top-down strategy is illustrated by the synthesis of nanosize yttrium-stabilized zirconia by leaching bulky yttrium-doped BaZrO3 or bulky yttrium-doped Na2 ZrO3 [19]. In this approach, yttrium-doped BaZrO3 or yttrium-doped Na2 ZrO3 is first synthesized from BaCO3 , ZrO2 , and Y2 O3 , and from Na2 CO3 , ZrO2 , and Y2 O3 by a conventional solid-state reaction method. The synthesized yttrium-doped BaZrO3 or yttrium-doped Na2 ZrO3 is then boiled to leach away the unwanted species, BaO or Na2 O, either in a diluted HNO3 solution for BaZrO3 or in water for Na2 ZrO3 . The residues of the skeleton form fine, nanosize yttrium-stabilized zirconia of 3–15 nm.

2.2. By Nature of the Process Different nanoparticle-making techniques may involve different scientific processes, such as physical, chemical or biological, or some combination of these processes.

2.2.1. Physical Methods A physical process involves only changes in physical state, such as size, shape, and phase of the matter. Mechanical size reduction is a physical process. Condensing gaseous metal vapor into nanoparticles is another example. The formation of nanoparticles in the gas phase is accomplished by a rapid quenching of supersaturated metal vapors with roomtemperature or cold inert gas [20].

2.2.2. Chemical Methods Most nanoparticle syntheses involve chemical changes. For instance, a sol–gel process converts metal alkoxides into oxide particles [21]. A hydrothermal process converts titanium isopropoxide and barium hydroxide into BaTiO3 [22]. Other chemical methods will be described in more detail in following discussions.

2.2.3. Biological Methods Nanostructured materials are synthesized in nature by a process known as biomineralization (the in vivo formation of inorganic crystals and/or amorphous particles in biological systems [23, 24]). In nearly all cases, the growth of the inorganic phases is controlled by biologically produced membranes or templates, which are usually composed of proteins and/or polysaccharides [25]. In some cases, the phase of the inorganic element is dictated by the organic template, by molecular complementarities, or by epitaxial matching of the organic and inorganic surfaces. In some cases, the biological membrane may act to limit the size and morphology of the resulting crystal by providing a defined space and chemical environment within which precipitation can occur. Fendler [26, 27] has shown that a mimetic approach in nanoparticle synthesis addresses the essential requirement of biomineralization. It is to have compartments capable of providing structural, spatial, and chemical controls for nanoparticle generation and stabilization. The mimetic approach using aqueous micelles, reverse micelles, microemulsions, vesicles, polymerized vesicles, monolayers, self-organized multilayers, and bilayer lipid membranes has successfully generated nanocrystalline Ag2 O [28], and less than 10 nm monodispersed Pt, Pd, Rh, and Ir particles [29, 30]. Pierre et al. have also demonstrated the synthesis of nanoscale (1500  C (adiabatic). These aerosol synthesis methods, aerosol pyrolysis and flame–aerosol pyrolysis, have produced BaTiO3 particles less than 100 nm. The BaTiO3 particles in Figure 2 also show broad particle size distributions. In general, these aerosol syntheses can produce spherical particles with clean surfaces, desired Ba/Ti ratio control, and good crystallinity via a continuous process. There are many reported studies in which nanoparticles of a few nanometers are produced [52, 74, 85–87]. In many cases, these are academic studies involving the collection of a few particles on a TEM grid. These nanoparticles are extremely reactive, and agglomerate quickly into large particles in an industrial collection system. The industrial process for producing large quantities of particles of nanosizes may require such a large dilution of the particle stream with the carrying gas as to make this process inefficient. The scale up to the production of tonnage quantity for the flame–aerosol process is much easier than for that of aerosol pyrolysis. In general, it is extremely challenging to make unagglomerated particles with a narrow size distribution.

3.2. Liquid-Phase Synthesis Liquid-phase syntheses are the most common and diverse methods for nanoparticle synthesis. The critical particle formation step occurs in the liquid phase; however, under supercritical conditions, the particle formation phase is at both the liquid and gas phase since there is no distinction between gas and liquid under those conditions. The most common solvent is water, but more and more organic solvents are also used in an effort to reduce the agglomeration of nanoparticles. Common to all of the liquid-phase syntheses is that they begin with solution preparation. A true solution means homogeneity at the atomic or molecular scale.

Ceramic Nanoparticle Synthesis

The central goal for the diverse solution synthetic methods is to preserve as much of this homogeneity as possible in the particle formation process. Liquid-phase syntheses are generally wet-chemical solution syntheses, sometimes called soft solution-processing routes. There is an increasing role that solution synthesis plays in the preparation of ceramic powders, particularly nanometer-sized particles with controlled size and morphology [88]. Wet-chemical processes have been widely recognized as an efficient approach to prepare nanocrystalline fine particles at low temperatures. In relation to other methods such as vapor-phase synthesis and physical methods (such as laser ablation and electron beam evaporation), major advantages of the liquid-phase synthesis include: (1) the process can be scaled up easily due to its bulk-processing nature; (2) the process is more controllable because chemical reactor technologies developed for chemical synthesis may be adapted and applied to the production of nanoparticles; (3) some processes require only mild conditions such as low temperature and atmospheric pressure; (4) the chemical composition of nanoparticles can be tailored in a liquid phase; (5) in the liquid phase, surface-controlling agents can be applied during and after nanoparticle formation to control the size and prevent unnecessary agglomeration; and (6) monodispersed nanoparticles can be made in the liquid phase via a homogeneous nucleation, growth, and aggregation mechanism. Uniform particle size is extremely important to nanoparticle selfassembly-based systems. Wet-chemical syntheses are particularly suited to the synthesis of nanoparticles (solid, core–shell, or hollow) with tightly controlled parameters, such as: (1) the size and shape, (2) the monodispersity, (3) the chemical composition and purity, (4) the bulk substructure, (5) the crystallinity (polycrystalline, single crystalline, or amorphous), (6) the surface functional group (thus, the interfacial free energy and surface-charge density), (7) the shell-layer thickness in a core–shell nanoparticle, or even (8) the size and shell thickness of a hollow sphere particle. For ceramic processing, the above characteristics of nanoparticles will affect the processing behavior, as well as the properties of the final, casted, sintered ceramic materials. The use of “nanoparticle(s)” as a popular technical term in the literature of wet-chemical synthesis of truly nanosized (i.e.,