Anisotropic carbon nanotube papers fabricated from multiwalled carbon nanotube webs

Anisotropic carbon nanotube papers fabricated from multiwalled carbon nanotube webs

CARBON 4 9 ( 2 0 1 1 ) 2 4 3 7 –2 4 4 3 available at journal homepage: Anisotropic carbon nan...

949KB Sizes 0 Downloads 9 Views


4 9 ( 2 0 1 1 ) 2 4 3 7 –2 4 4 3

available at

journal homepage:

Anisotropic carbon nanotube papers fabricated from multiwalled carbon nanotube webs Yoku Inoue a,*, Yusuke Suzuki a, Yoshitaka Minami a, Junichi Muramatsu a, Yoshinobu Shimamura b, Katsunori Suzuki c, Adrian Ghemes d, Morihiro Okada d, Shingo Sakakibara d, Hidenori Mimura d, Kimiyoshi Naito e a

Department of Electrical and Electronic Engineering, Shizuoka University, 3-5-1 Johoku, Hamamatsu 432-8561, Japan Department of Mechanical Engineering, Shizuoka University, 3-5-1 Johoku, Hamamatsu 432-8561, Japan c Graduate School of Science and Technology, Shizuoka University, 3-5-1 Johoku, Hamamatsu 432-8011, Japan d Research Institute of Electronics, Shizuoka University, 3-5-1 Johoku, Hamamatsu 432-8011, Japan e National Institute for Materials Science, Hybrid Materials Center, Composite Materials Group, 1-2-1 Sengen, Tsukuba 305-0047, Japan b



Article history:

We fabricated large-scale anisotropic carbon nanotube (CNT) paper sheets by stacking

Received 27 October 2010

long-lasting multiwalled CNT (MWCNT) webs without using binder materials. The

Accepted 6 February 2011

MWCNTs are highly aligned in the webs and they retain their alignment in the fabricated

Available online 26 February 2011

paper. Although MWCNTs are just connected by van der Waals force, tensile strength is as strong as 75.6 MPa. In addition, resistivity and thermal conductivity is as good as 2.5 · 103 X cm and 70 W/m K, respectively. The present high anisotropy ratios of 7.3 in resistivity and of 8.1 in thermal conductivity are due to the high alignment of the ultra-long MWCNTs which have lengths of millimeters. High-speed web drawing with a draw speed of over 10 m/s enables very rapid fabrication. The material properties of CNT structures can be measured by conventional methods for macroscopic samples rather than methods designed for nanomaterials. CNT web technology will enable CNTs to be used in new applications.  2011 Elsevier Ltd. All rights reserved.



Carbon nanotubes (CNTs) are an attractive functional nanomaterial because of their diverse features, especially their good electrical, mechanical and thermal properties. Individual CNTs have an electrical resistivity of 104 to 105 X cm [1], a tensile strength of 150 GPa [2] and a thermal conductivity of 3500 W/m K [3]. However, it is difficult to take full advantage of these excellent properties when CNTs are formed into large-scale structures, such as nanotube buckypaper [4]. For example, the total electrical conductivity of large-scale CNT structures in which CNTs make contact with each other by

their surfaces or via binding materials is mainly determined by the contact resistance at CNTCNT connections [4,5]. Moreover, the spaces between adjacent CNTs (i.e., voids) increase the size of structures made from CNTs. In addition, for multiwalled CNT (MWCNT) structures, since little charge penetrates to the interior walls, the outermost walls predominantly function as charge transfer channels. However, the electrical conductivity of a material is calculated by dividing the conductance of an object by its dimensions (i.e., length and cross-sectional area). Consequently, the conductivity of a CNT structure is significantly lower than that of an individual CNT. Nevertheless, it is worthwhile to attempt to fabricate

* Corresponding author: Fax: +81 53 478 1356. E-mail address: [email protected] (Y. Inoue). 0008-6223/$ - see front matter  2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.carbon.2011.02.010



4 9 ( 2 0 1 1 ) 2 4 3 7 –2 4 4 3

large-scale CNT structures that exploit the excellent properties of CNTs. Since most properties of CNTs are related to their cylindrical structures, it is preferable to fabricate highly aligned CNT structures. Successive fabrication of MWCNT yarns [6–11] and sheets [12–14] has recently been reported. MWCNTs are horizontally drawn from a vertically aligned MWCNT array by taking one MWCNT after another, which is called dry spinning. This process transforms a three-dimensional array into a horizontally aligned web, which is a two-dimensional network of CNTs. The spinning manner resembles spinning a silkworm cocoon into a thread. It can be used to produce large-scale structures from MWCNTs. Although it is similar to conventional short-fiber spinning processes, it differs in that it is not necessary to twist the MWCNTs together. The large surface areas of MWCNTs give rise to strong van der Waals forces between them. We recently developed a rapid method for producing MWCNT arrays [15]. Millimeter-scale MWCNT arrays were grown by chloride-assisted chemical vapor deposition (CVD). The arrays had higher drawabilities than previously reported CNT arrays, as demonstrated by the fact that the MWCNT webs could be drawn from 2.0 mm-high MWCNT arrays. Previously, the height of drawable arrays has been limited up to 1 mm [7,9,11,14]. By inventing a new growth method, we could achieve very long drawable arrays. Because of their high stickiness, MWCNT webs could be easily assembled into large-scale structures without the need for twisting. Wellaligned MWCNT paper was fabricated by stacking and shrinking the MWCNT webs. This study investigates the material properties of the MWCNT paper. The measured electrical, mechanical and thermal properties exhibited high anisotropies; these anisotropies are due to the high aspect ratios (over 50,000) of the MWCNTs and the high degree of alignment of the MWCNTs in the paper. The web-stacked MWCNT paper is a novel material that has multiple properties that are anisotropic (e.g., electrical, mechanical, and thermal properties). Due to its large size, all the measurements were performed by conventional measurement methods designed for macrosized samples, rather than by measurement methods especially developed for nanomaterials. The technique used to produce MWCNT webs is promising for realizing large-scale CNT applications.



MWCNT arrays were synthesized using a conventional thermal CVD system. Iron chloride, FeCl2, was used as the precursor for a Fe catalyst. Since catalyst nanoparticles are formed by a vapor-phase reaction between FeCl2 and acetylene at the beginning of the growth process, the substrate does not need to be predeposited with a metal. An unprocessed substrate of quartz or oxidized Si was placed in the growth chamber. Details of the growth method are described in Ref. [15]. The advantages of this method are its rapid growth rate of over 0.1 mm/min and a reaction efficiency of over 50% with acetylene. In addition, every array can be drawn for as long as MWCNTs remain on the substrate. As-grown arrays can be drawn without post-processing.


Results and discussion

MWCNTs with diameters of 30–50 nm are highly aligned perpendicular to the substrate and very highly bundled over the entire array. Since all the MWCNTs are strongly attached to each other, the array can be easily peeled from the substrate and it is self-supporting; it resembles a thick black mat. It is essential that bundled MWCNTs spread across the entire substrate for high drawability. A high drawability is achieved due to highly aligned MWCNTs that are straight from their roots to their tips (Fig. 1a) (even MWCNTs with lengths of the order of millimeters). The areal density is estimated by measuring the average distance between MWCNTs from side-view scanning electron microscopy (SEM) images of the arrays. Since the average distance of MWCNTs was ranging from 200 nm to 300 nm, the areal density was estimated to be 1–3 · 109/ cm2. The density can be changed by changing the growth conditions, such as growth pressure and gas flow rate, and the high areal density is one of important issues for highly drawable arrays. Because the MWCNTs have high drawabilities, no specialized equipment is required to form a web (Fig. 1b). They can be easily drawn by just pinching and pulling out an edge of the array using tweezers. SEM observations were performed to observe the transformation from a three-dimensional array to a two-dimensional web. Fig. 1c shows horizontal drawing of a web from a 2.0 mm-high MWCNT array that had been grown with a growth time of 16 min. A drawn MWCNT or MWCNT bundle pulls another one from the array; this process repeats until all the MWCNTs have been removed from the substrate. Individual or bundled MWCNTs are aligned in the drawing direction in the MWCNT web (Fig. 1d). The MWCNTs are automatically aligned in the drawing direction. This transformation into a web appears to be a characteristic of CNTs. Previously reported methods for aligning a large number of CNTs require a lot of effort [16]. Forming webs of MWCNTs is a good method to obtain an aligned CNT structures. We believe this technique for forming MWCNT webs, which we term CNT web technology, will become a key CNT technology. To further investigate the array structure, polarized Raman spectroscopy measurements were performed in backscattering geometry with incident light (laser light with a wavelength of 532 nm) normal to the side surface of the array. The incident light was polarized parallel and perpendicular to the MWCNT alignment (see the inset of Fig. 2). For both polarizations, the ratio of the intensity of the graphite-like G band at 1580 cm1 to that of the disorder-induced D band at 1350 cm1 (IG/ID) is as high as 3.0. This high value indicates the high crystal quality of our MWCNTs and the low amount of amorphous carbon. The high quality of the MWCNTs is attributed to the high growth temperature (830 C), which improves the crystalline quality of the graphene layers, and to the short growth time (5–16 min), which suppresses deposition of amorphous carbon. On the other hand, the G-band intensity ratio for the two polarizations, R = Iparallel/Iperpendicular, provides information regarding the degree of MWCNT alignment in the array, because Raman scattering is more intense when the polarization of the incident light is parallel to the axis of a MWCNT [16]. The obtained R is 4.4, which strongly


4 9 ( 20 1 1 ) 2 4 3 7–24 4 3


Fig. 1 – (a) Roots of a MWCNT array. (b) MWCNT web initiation using tweezers. (c) Transformation from a vertical array to a horizontal web. (d) MWCNT web.

Fig. 2 – Polarized Raman spectra of a MWCNT array. Inset shows the measurement configuration. For the G band at 1580 cm1, the parallel polarization intensity (red) is 4.4 times higher than the perpendicular polarization intensity (blue). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

indicates a high degree of alignment [17]. This alignment is due to the close proximity of growth sites. To form aligned buckypapers, long-lasting MWCNT webs were wound on a drum (Fig. 3a). Because of the high drawability and the high stickiness of the webs, large MWCNT paper sheets were easily formed in a short time with a draw speed

of over 10 m/s (Supplementary material movie 1). To densify the as-stacked MWCNT paper, ethanol was sprayed on the sheets and then evaporated. The MWCNTs in the sheet were bonded to each other only by van der Waals forces; no binder material was needed. The thickness of the paper sheets could be controlled by varying the amount of webs rolled on the drum. Fig. 3b shows a photograph of an A4 sheet of MWCNT paper, which was produced by cutting open a roll of paper. Due to the high-drawability of the MWCNT array, an A4 sheet of paper can be fabricated in several minutes (see Supplementary material). The MWCNT paper produced in this study was 1.8 ± 0.1 lm thick and had a density of 0.84 ± 0.08 g/cm3. As a result of the high alignment of the MWCNTs in the webs, the MWCNTs are highly aligned in the paper (Fig. 3c). Therefore, the electrical, mechanical and thermal properties were significantly anisotropic. In the present study, all measurements of the MWCNT paper sheets were performed using conventional methods for macroscopic samples. The MWCNT paper sheets are not nanomaterials, but macro-sized objects that can be handled just as easily as normal materials. Therefore, we consider that MWCNT paper sheets should be regarded as conventional sheets and should not be treated by techniques used for determining the physical parameters of CNTs (e.g., normalizing quantities by the density or subtracting the empty spaces in the material). Consequently, the parameters of the whole sample (i.e., the sample length, cross-sectional area and weight) were used when determining the material properties, which is conventional for macroscopic samples.



4 9 ( 2 0 1 1 ) 2 4 3 7 –2 4 4 3

Fig. 4 – Current–voltage characteristics measured parallel and perpendicular to MWCNT alignment.

Table 1 – Sheet resistance and resistivity measured parallel and perpendicular to MWCNT alignment. Direction

Sheet resistance (X/sq) Resistivity (X cm)

Parallel 13.8 Perpendicular 100.1

2.5 · 103 1.8 · 102

resistance than the perpendicular direction. The high anisotropy is attributed to the MWCNTs having lengths of the order of millimeters and being highly aligned in the paper sheet. Fig. 5 shows the stress–strain characteristics, which were measured using a tensile tester (Shimadzu, EZ-L) for MWCNT paper samples that were 1 cm · 1 cm in size and were mounted in a paper mount. The tensile strength, which was obtained by dividing the maximum load by the sample cross-sectional area, was 75.6 MPa parallel to the MWCNT alignment. Although the MWCNTs were bonded by van der Fig. 3 – (a) Fabricating an aligned MWCNT paper by winding MWCNT webs on a drum. (b) A4 sheet of aligned MWCNT buckypaper and (c) surface image of the MWCNT paper.

The current–voltage characteristics were measured on a 1 cm · 1 cm area of a paper sheet parallel and perpendicular to the MWCNT alignment. Pd plates were used as ohmic electrodes. Linear current–voltage characteristics were observed (Fig. 4; see Table 1 for a summary). The sheet resistance in the parallel direction (13.8 X/sq) is lower than that in the perpendicular direction (100.1 X/sq) and the anisotropy ratio is 7.3. The resistivity of a 1.8 lm-thick paper sheet in the parallel direction is thus 2.5 · 103 X cm. The MWCNT paper is lightweight and highly electrically conductive with a high anisotropy. This macroscopic resistance of the MWCNT paper is mainly determined by the contact resistance between adjacent MWCNTs. Consequently, the parallel direction, which has fewer connecting points along current paths, has a lower

Fig. 5 – Stress–strain characteristics of the MWCNT paper measured parallel and perpendicular to MWCNT alignment.


4 9 ( 20 1 1 ) 2 4 3 7–24 4 3


Table 2 – Thermal diffusivity and thermal conductivity measured parallel and perpendicular to MWCNT alignment. Direction

Thermal diffusivity (m2/s)

Thermal conductivity (W/m K)

Parallel Perpendicular

1.22 · 104 1.50 · 105

69.6 8.6

Waals forces alone, the sheet strength is higher than that of disordered buckypaper [18], and is comparable to that of pure aluminum or nickel [19]. Unlike typical stress–strain curves for spun MWCNT fibers, which show that fracture occurs suddenly [6–11], the MWCNT paper fractures gradually due to sliding of MWCNTs, as indicated by the smooth peak structure in Fig. 5. In contrast, the tensile strength in the perpendicular direction is very low (see the inset in Fig. 5). The high strength in the parallel direction is attributed to the high alignment of MWCNTs that have lengths of the order of millimeters and that are connected by very large surface areas, which increases the amount of van der Waals bonds that form between adjacent MWCNTs. Table 2 lists the highly anisotropic thermal properties. So far, for the aligned CNT sheet, electrical property has been measured [14]. However, thermal conductivity has not been well investigated. The thermal diffusivities of the MWCNT paper sheets, a, were measured using a scanning laser heating thermal diffusivity meter (Ulvac-Riko, Inc., Laser PIT). The scanning laser heating provides a highly uniform and high intensity energy source. An analysis method using both the amplitude decay and the phase shift was employed to eliminate the effects of heat loss. MWCNT paper sheets (5 mm · 25 mm in size) were measured at room temperature in a vacuum of less than 0.01 Pa. The experimental details of the measurement system are given in Ref. [20]. The thermal conductivities of the paper sheets, K, were calculated using K = aqC, where q is the density and C is the specific heat of the MWCNT paper. The specific heat of a MWCNT was assumed to be 0.713 J/gÆK, which is almost identical to that of graphite at room temperature [21]. The thermal conductivity of the paper in the parallel direction was estimated to be 69.6 W/mÆK, which is much higher than that of reported buckypaper [22]. However it is still lower than the thermal conductivity of single CNTs [3]. Since MWCNT paper consists of a huge number of MWCNTs, it is reasonable to consider that the large-scale thermal conductivity is governed by the heat resistance at MWCNTMWCNT interfaces [23]. The anisotropy ratio for the thermal conductivity is as high as 8.1, which is due to the high alignment of ultra-long MWCNTs. When MWCNTs are used as large-scale industrial materials, it is important to consider that irrespective of whether they are used for transferring electric current, mechanical stress or heat, the outermost walls of the MWCNTs form the main channel. Since electrons are injected generally from the outermost walls and the electrical conductivity in the graphene plane is much higher than that between the graphene planes, electron transport predominately occurs through the outermost wall [24]. In addition, when an electron transfers between two MWCNTs, it passes through the interface

Fig. 6 – (a) Schematic of the effective cross-sectional area of an aligned MWCNT structure. Calculation of the effective cross-sectional area ratio in the total cross-sectional area can be reduced to that of a hexagon, which contains a MWCNT. The green area is the effective cross section in the red hexagon. The spacing between the walls in a MWCNT and the distance between adjacent MWCNTs are assumed to be equal to the spacing of graphite (0 0 2) planes (0.34 nm). (b) Calculated effective cross-sectional area as a function of MWCNT diameter. The inset tabulates the results.

between the two surfaces. Consequently, macroscopic electrical conduction is governed by the outermost walls. For heat transport, in which phonons are the dominant carriers [25], similar reasoning can be used to predict that phonons are mainly transported through the outer wall [23]. On the other hand, for load transfer in MWCNT structures such as buckypapers and spun fibers [6–11], the external load is balanced by shear forces at MWCNTMWCNT interfaces. Since the total tensile strength of a MWCNT structure is much less than that of an individual MWCNT, fracture occurs due to sliding



4 9 ( 2 0 1 1 ) 2 4 3 7 –2 4 4 3

of MWCNTs. Therefore, only the outermost walls are responsible for the shear force. Consequently, in addition to the hollow areas and interspaces between MWCNTs, the cross-sectional area of the inner walls does not participate in transfer of charge, load and heat. Nevertheless, when MWCNT structures are used as industrial materials, their total cross-sectional area should be taken into account, as is done with conventional macroscopic materials. Fig. 6b shows the calculated relationship between the effective cross-sectional area and the MWCNT diameter. Assuming that the aligned MWCNTs are hexagonally close packed, the effective cross section is simply given by the cross-sectional area of the outermost wall that is contained in a hexagon (see Fig. 6a). The ratio of the effective cross-sectional area to the total cross-sectional area, Xeffective, is expressed by: 2pdD Xeffective ¼ pffiffiffi 3ðd þ DÞ2


where D is the diameter of the MWCNT. In this calculation, the spacing between the walls of a MWCNT is assumed to be identical to the distance between adjacent MWCNTs; this distance is denoted by d and set to be 0.34 nm, which is the spacing between graphite (0 0 2) planes. The effective crosssectional area decreases drastically with increasing diameter. The calculation reveals that even for a CNT structure fabricated from 1 nm-diameter single-walled nanotubes, 30% of the total cross-sectional area does not participate in transfer of charge, load and heat. In the present study, the MWCNTs were 30 nm in diameter; thus, at most only 4% of the crosssectional area is effective. Based on the density of the MWCNT paper sheets being lower than expected, it is obvious that they are not ideally packed. The actual effective crosssectional area is speculated to be considerably less than 4%. To enhance the properties of CNT structures, it is critical to densely pack thin CNTs so as to reduce ineffective spaces.



In this work, we have grown high-drawability MWCNT arrays by chloride-assisted CVD. The drawability was so high that webs could be drawn at high speeds from a 2.0 mm-high MWCNT array. By stacking and shrinking the webs, aligned MWCNT paper sheets could be easily fabricated without using a binding material. The MWCNT paper produced is lightweight and flexible, while being mechanically strong and having a high electrical conductivity. In the present study, although condensed packing was not achieved, high anisotropies were obtained in the electrical and thermal conductivities and in tensile strength. The present high anisotropies, arising from the high alignment of MWCNTs with lengths of the order of millimeters, are useful for developing novel applications.

Acknowledgement This work was supported in part by the Japan Science and Technology Agency, the Japan Chemical Innovation Institute and the Murata Science Foundation.

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.carbon.2011.02.010.


[1] Ebbesen TW, Lezec HJ, Hiura H, Bennett JW, Ghaemi HF, Thio T. Electrical conductivity of individual carbon nanotubes. Nature 1996;382:54–6. [2] Demczyk BG, Wang YM, Cumings J, Hetman M, Han W, Zettl A, et al. Direct mechanical measurement of the tensile strength and elastic modulus of multiwalled carbon nanotubes. Mater Sci Eng A 2002;334:173–8. [3] Pop E, Mann D, Wang Q, Goodson K, Dai H. Thermal Conductance of an individual single-wall carbon nanotube above room temperature. Nano Lett 2006;6:96–100. [4] Rinzler AG, Liu J, Dai H, Nikolaev P, Huffman CB, RodriguezMacias FJ, et al. Large-scale purification of single-wall carbon nanotubes: process, product, and characterization. Appl Phys 1998;A67:29–37. [5] Fischer JE, Zhou W, Vavro J, Llaguno MC, Guthy C, Haggenmueller R, et al. Magnetically aligned single wall carbon nanotube films: Preferred orientation and anisotropic transport properties. J Appl Phys 2003;93:2157–63. [6] Jiang K, Li Q, Fan S. Spinning continuous carbon nanotube yarns. Nature 2002;419:801. [7] Zhang M, Atkinson KR and Baughman RH. Multifunctional carbon nanotube yarns by downsizing an ancient technology. Science 2004;306:1358-1361 (2004). [8] Aliev AE, Guthy C, Zhang M, Fang S, Zakhidov AA, Fischer JE, et al. Thermal transport in MWCNT sheets and yarns. Carbon 2007;45:2880–8. [9] Tran CD, Humphries W, Smith SM, Huynh C, Lucas S. Improving the tensile strength of carbon nanotube spun yarns using a modified spinning process. Carbon 2009;47:2662–70. [10] Nakayama Y. Synthesis, nanoprocessing, and yarn application of carbon nanotubes. Jpn J Appl Phys 2008;47:8149–56. [11] Zhang X, Li Q, Tu Y, Li Y, Coulter JY, Zheng L, et al. Strong carbon-nanotube fibers spun from long carbon-nanotube arrays. Small 2007;3:244–8. [12] Kuznetzov AA, Lee SB, Zhang M, Baughman RH, Zakhidov AA. Electron field emission from transparent multiwalled carbon nanotube sheets for inverted field emission displays. Carbon 2010;48:41–6. [13] Lepro X, Lima MD, Baughman RH. Spinnable carbon nanotube forests grown on thin, flexible metallic substrates. Carbon 2010;48:3621–7. [14] Liu K, Sun Y, Chen L, Feng C, Feng X, Jiang K, et al. Controlled growth of super-aligned carbon nanotube arrays for spinning continuous unidirectional sheets with tunable physical properties. Nano Lett 2008;8:700–5. [15] Inoue Y, Kakihata K, Hirono Y, Horie T, Ishida A, Mimura H. One-step grown aligned bulk carbon nanotubes by chloride mediated chemical vapor deposition. Appl Phys Lett 2008;92:213113. [16] Fischer JE, Zhou W, Vavro J, Llaguno MC, Guthy C, Haggenmueller R, et al. Magnetically aligned single wall carbon nanotube films: Preferred orientation and anisotropic transport properties. J App Phys 2003;93:2157–63. [17] Zhang Y, Zou G, Doorn SK, Htoon H, Stan L, Hawley ME, et al. Tailoring the morphology of carbon nanotube arrays: from spinnable forests to undulating foams. Nano 2009;3:2157–62.


4 9 ( 20 1 1 ) 2 4 3 7–24 4 3

[18] Park JG, Smithyman J, Lin CY, Cooke A, Kismarahardja AW, Li S, et al. Effects of surfactants and alignment on the physical properties of singlewalled carbon nanotube buckypaper. J Appl Phys 2009;106:104310. [19] Boyer HE. T. L. Gall, Metals Handbook, Desk edition. Ohio: American Society for Metals; 1985. [20] Naito K, Yang JM, Xu Y, Kagawa Y. Enhancing the thermal conductivity of polyacrylonitrile- and pitch-based carbon fibers by grafting carbon nanotubes on them. Carbon 2010;48:1849–57. [21] Hone J, Llaguno MC, Biercuk MJ, Johnson AT. B. Batlogg B, Z. Benes Z, et al. Thermal properties of carbon

[22] [23]




nanotubes and nanotube-based materials. Appl Phys 2002;A74:339–43. Yue Y, Huang X, Wang X. Thermal transport in multiwall carbon nanotube buckypapers. Phys Lett A 2010;374:4144–51. Kim P, Shi L, Majumdar A. McEuen PL. Thermal transport measurements of individual multiwalled nanotubes. Phys Rev Lett 2001;87:215502. Bourlon B, Miko C, Forro L, Glattli DC. Bachtold A, Determination of the intershell conductance in multiwalled carbon nanotubes. Phys Rev Lett 2004;93:176806. Yang DJ, Zhang Q, Chen G, Yoon SF, Ahn J, Wang SG, et al. Phys Rev B 2002;66:165440.