1.1. Basic property of carbon nanotube
Carbon nanotubes (CNTs) are cylinders of graphene which is a single planar sheet of sp2-bonded carbon atoms. From their unique structures, they exhibit several novel properties, such as high strength, high performance in electrical and thermal conductors, etc. Therefore, since the discovery by Iijima in 1991 (Iijima, 1991), CNTs have been one of the hottest topics in both physics and material science. Mainly, their structures are specified with four parameters; the number of walls, diameter, length and chirality. Dependent of the number of walls, CNTs are categorized as single-walled nantoubes (SWNTs) and multi-walled nanotubes (MWNTs) and SWNT structure can be specified uniquely with a chiral vector
The band gap of semiconducting SWNTs is inversely proportional to the diameter. Therefore, it is important to control and obtain SWNTs with unique chirality at will, for realization of CNT devices.
1.2. Carbon nanotube growth by surface decomposition of SiC
There have been three main methods to produce CNTs; arc discharge, laser ablation and chemical vapour deposition (CVD). Arc discharge and laser ablation utilize the evaporation of a graphite target to create gas-phase carbon fragments which recombine to form CNTs. Although both produce CNTs with high crystallinity due to the high growth temperature, it is difficult to control the structure of CNTs. Recently, CVD has been widely used for CNT growth because it is suitable for organizing CNTs over a large surface. CVD utilize metal catalyst particles in the gas phase or on surfaces, decomposing a carbon containing feedstock gas, such as acetylene or ethanol. These procedures can be applied to control the structure of CNTs through the size and shape of catalyst. So far, the diameters of CNTs could be controlled to some extent through the control of catalyst particle size (Jeong et al., 2007). However, there still remains some distribution in the CNT diameter, and also the control of chirality has never been achieved.
As distinct from these conventional growth techniques, surface decomposition of SiC has been reported as a novel CNT growth technique during recent years. It was first discovered by M. Kusunoki et al. in 1997, who found that CNTs grew on SiC(-1-1-1) surface after heating a 3C-SiC single crystal at 1700ºC using YAG laser during a transmission electron microscope (TEM) observation (Kusunoki et al., 1997). They observed that CNTs were mostly oriented along the  direction. Later, they reported that CNTs could be easily produced on 6H-SiC(000-1) surface only after heating in a vacuum electric furnace (Kusunoki et al., 2000). The schematic picture of the growth process of CNTs by surface decomposition of SiC is shown in Fig. 2. As heating temperature rises, Si atoms are desorbed and CNTs grew into the SiC substrate the axes of which are kept perpendicular to the substrate surface. This growth method has an exclusive characteristics because it needs no catalyst, and self-aligned CNTs were grown with desorption of Si atoms.
Also, they pointed out that CNT growth was strongly dependent on the polar faces of SiC. Figures 3(a) and (b) shows schematic cross-sectional and top view of 6H-SiC, respectively, and SiC (000-1) and (0001) faces are shown at the top layer and bottom in Fig. 3(a). On the (000-1) face (carbon–face), there are dangling bonds from carbon atoms which are perpendicular to the substrate surface, while dangling bonds from Si atoms are seen on the (0001) face (Si-face). Kusunoki et al. pointed out that CNT growth was observed only on carbon-faces of SiC single crystals, such as (000-1) face of 6H-SiC and (-1-1-1) face of 3C-SiC, while graphite sheets were formed on the surface after heating Si-faces (Kusunoki et al., 2000).
So far, through high-resolution TEM (HRTEM) images, they revealed several characteristics of CNTs grown by this method: (a) most CNTs are two to five layered 2-5 nm in diameter, (b) The CNTs are grown perpendicular to the SiC surface, (c) The density of CNTs is above 1012 cm-2, which is higher than those grown by other methods, (d) The CNTs are atomically bonded with the SiC crystal at the interface and (e) most of the CNTs have unique chirality, that is, zigzag-type (Fig. 3(c)) (Kusunoki et al., 2002). Although there are still several problems to be resolved for device fabrication, such as purity of the CNT film and homogeneity in the number of walls and diameter, CNTs grown by surface decomposition of SiC have several advantages to realize CNT devices. Among these useful properties, the unique chirality of grown CNTs is the most important and mysterious, because the growth of CNTs with unique chirality has never been attained by other growth techniques. Therefore, surface decomposition method might play an important role in fabrication of CNT devices. Also, elucidation of the mechanism of selectively chirality alignment will be invaluable and useful to control and align chirality of CNTs grown with other methods.
1.3. Carbon nanocap
The formation process in the initial stage of CNT growth by surface decomposition of SiC has been reported by several groups. TEM studies by Kusunoki’s group showed that arced sheets of carbon with an outer diameter of 5 nm and height of 1-2 nm appeared on the SiC(000-1) face after heating at 1250ºC for 30 min (Kusunoki et al., 2000). After heating at 1300ºC, two-, to three-layered carbon nanocaps 3-5 nm in diameter and 3-5 nm in height were generated in a dense formation on the surface. After heating at higher temperature, CNTs grew below these carbon nanocaps with desorption of Si atoms indicating that these carbon nanocaps correspond to the top of CNTs. This means that the structure of CNTs is closely related to the carbon nanocaps, and elucidation of the formation mechanism of the carbon nanocaps is a key point to clarify unique chirality of CNTs grown by this method.
Except for pioneering and essential investigation by Kusunoki’s group, few studies have been reported about the formation mechanism of carbon nanocaps. Watanabe et al. carried out TEM and scanning tunnelling microscope (STM) observation for CNT growth on 3C-SiC(-1-1-1) and proposed a model for carbon nanocap formation. The schematic figure of their model is shown in Fig. 4 (Watanabe et al., 2001). In their model, at first, amorphous carbon was formed on the SiC surface with the evaporation of Si atoms. Then, they were crystallized and graphene flakes were formed and aligned parallel to the SiC surface with the temperature rise. Finally, lifting of part of the graphene, carbon nanocaps were formed. This model is significant as first attempt to explain this unique phenomenon. However, intuitively, their model seems to have some discrepancy against the observed experimental results; in their model, the diameter of a CNT would be enlarged, however, TEM observation showed that the diameter was maintained during the growth. Also, it seems to be difficult to explain the reason why enormously high-density CNTs are grown. To clarify the real growth process, analysis at the atomic level is essential, including the alignment and direction of C-C bonding during the crystallization
In this study, we investigated the details of formation process of carbon nanocaps using various experimental techniques; STM, X-ray photoemission (XPS) and near edge X-ray absorption fine structure (NEXAFS) spectroscopies. Based on these experimental results, we attempted to propose a formation mechanism of carbon nanocaps.
2. Formation mechanism of carbon nanocap
2.1. Formation process of carbon nanocap
To clarify the details of formation process of carbon nanocaps, it is essential to clarify the surface structure and composition. However, few studies have been reported with combing XPS measurement and STM observation for formation process of carbon nanocaps. Here, we carried out both XPS measurements and STM observations for carbon nanocaps formed on 6H-SiC(000-1) (Maruyama et al., 2006, 2007). Figure 5(a) and (b) show XPS spectra of 6H-SiC(000-1) surface before and after heating at 1100ºC in UHV, respectively, where the spectrum of Fig. 5(a) is that from the sample after etching with 10% HF for 15 min. To grow carbon nanocaps, the temperature was slowly increased, keeping the heating rate below 2ºC/min in the temperature region above 400ºC. Both spectra were measured at room temperature with a laboratory photon source (
Surface structure change with the heating temperature is shown in Fig. 6. These STM images were observed at room temperature after heating at each temperature in UHV. After heating at 800ºC, step-like structures appeared on some portion of the surface (a). The area of each terrace was small and several tens of nanometers. After heating at 1100ºC, densely packed grains, around 1-2 nm in size, were observed over the entire surface (b). These nanoparticles formed clusters at 1150ºC, keeping the sizes of individual particles unchanged (c). The cluster size ranged from around 3 to 5 nm. Taking into account the result of XPS measurements, the nanoparticles were elemental carbon which were accumulated after desorption of Si. After heating at 1200ºC, the SiC surface was covered by a large number of grains ranging in size from around 3 to 5 nm, as shown in STM images of Fig. 6(d). The blowup in Fig. 6(d) shows that their surfaces were covered with cobweb-like patterns, indicating that the carbon nanocaps were formed at this temperature.
Figure 7 shows another STM images for carbon nanocaps formed on SiC(000-1) which were annealed at 1250ºC in a vacuum electric furnace (Bang et al., 2006). Carbon nanocaps formed on the entire surface were clearly shown in (a). Figure 7(b) shows an atomic-resolution STM images of the blowup of a cap structure produced on a 6H-SiC(000-1) substrate. They showed convex structures, 3-5 nm in diameter and 1-2 nm in height, exhibiting cobweb-like patterns, as shown in Fig. 6(e).
Figure 8(a) shows one portion of the surface of a carbon nanocap formed on 6H-SiC(000-1) after annealing in a vacuum electric furnace and (b) and (c) show magnified feature of STM images of (a). Figure (b) and (c) are the same except for the markers shown by open circles and a square. (Bang et al., 2006). Figure 8(b) clearly reveals the cobweb-like pattern, which corresponds to carbon networks, on the surface of carbon nanocaps. In Fig. 8(c), one large circle, one large square, and nine small circles can be seen. There are six small open circles around the right large open circle, which confirms that this pattern is a carbon network composed of hexagons. In contrast, the left large square is surrounded by only five small circles, which indicates that the portion of network represented by the left large open square is a pentagon on the nanocap surface. It is well known that a convex carbon network contains several pentagons, as observed on the surface of fullerenes. Given the convex shape of the nanocap, it is reasonable to observe a pentagon on the surface. These STM observations support the assumption that the carbon nanocap had been already crystallized when it was formed on the SiC surface.
Crystallization during the formation of carbon nanocaps was directly confirmed by
As seen in Fig. 9, for both incident angles, and irrespective of the temperature, sharp C-C π* resonance peaks were observed at approximately 285 eV in all spectra. In addition, at 1250C, broad σ* resonances were observed in the region from 290 to 315 eV, accompanied by a sharp σ* bound exciton at ~291.5 eV ((a) and (b)). These two peaks are characteristic of carbon materials composed of well-ordered graphene sheets. Therefore, the spectra in Figs. 9(a) and (b) are expected to be associated with carbon nanocaps. At 1190C, slight differences were observed between the spectra for the two incident angles. At
2.2. Mechanism of chirality alignment to zigzag-type
One of the most significant and important characteristics of CNT films grown by surface decomposition of SiC is their unique chirality. As stated above, Kusunoki et al. pointed out that the CNTs mostly exhibit zigzag-type chirality (Kusunoki et al., 2002). Taking into account that the carbon nanocaps were the tips of CNTs and that CNTs grow from below the carbon nanocaps, this suggests that they have chemical bonds perpendicular to the SiC(000-1) surfaces at its periphery, connected with zigzag-type CNTs (Fig. 3(c)). On the other hand, considering that aligned CNTs were apt to form on the carbon-face of SiC, Irle et al. pointed out that dangling bonds of carbon-face play a key role to form cap structures, based on quantum chemical molecular dynamics simulations (Irle et al., 2006).
Our XPS and STM results show that amorphous carbon nanoparticles are accumulated on SiC surface before the carbon nanocap formation (Maruyama et al., 2006). These nanoparticles assemble and coalesce as the temperature rises and finally are crystallized to carbon nanocaps. From the NEXAFS results, majority of C-C bonds are directed toward perpendicular to the surface at the beginning of crystallization. Therefore, utilizing dangling bonds on SiC surface, crystallization occurred, which serves to form carbon nanocaps. Considering that the final structure is a dome-like, the bond alignment perpendicular to the surface should occur circumferentially.
On the basis of the various experimental results, we propose a model for carbon nanocap formation by surface decomposition of SiC. Schematic images are shown in Fig. 10. (1) At first, carbon nanoparticles are accumulated as Si atoms are desorbed from SiC surface. (~1100ºC) (2) Upon heating, carbon particles start migration on the SiC surface as a result of the thermal energy, forming clusters of several tens of carbon nanoparticles to reduce the surface energy further (~1150ºC) (Fig. 10(b)). Then, the redistribution of carbon atoms begins within each nanoparticle cluster, and the nanoparticles begin to coalesce (Fig. 10(c)). At around 1200ºC, coalescence proceeds within the clusters, and crystallization occurs, forming CNT nanocaps. In the crystallization, the size of clusters is presumably almost the same as that of the formed nanocaps. Thus, we believe that the size of clusters determines the nanocap size. (5) Above 1200ºC, CNT growth into SiC proceeds by the sublimation of Si atoms from the SiC(000-1) surface. The NEXAFS spectra which suggest that the C-C bonds are directed nearly parallel to the surface at the beginning of crystallization, indicates that the direction of C-C bonds were changed during the crystallization. (Some difference might be seen in the measurement temperature between STM and NEXAFS because of the detection limitation of a pyrometer.) Taking into account the final structure of carbon nanocaps, we speculate that crystallization occurs at the periphery of amorphous carbon dome, and formation of C-C bonding utilize the dangling bonds from carbon atoms on the SiC(000-1) surface (Fig. 10). From the standpoint of crystal growth, this means that nucleation starts to occur at the edge in the carbon nanocaps.
Recently, graphene growth from SiC(000-1) carbon-face by surface decomposition has been reported and a lot of researchers have been focusing on it because of its superiority in electronic property (Jernigan et al., 2009, Tedesco et al., 2009). From the point of view of growth technique, the method of graphene growth on SiC(000-1) is remarkably similar to that of CNT growth by surface decomposition, but one significant difference is the heating rate. In the growth of graphene, in general, the temperature of SiC is quickly increased to the growth temperature, typically, about several minutes to ~1400ºC (Gamara et al., 2008). On the other hand, in the CNT growth by surface decomposition, the heating rate is typically 1-15ºC/min (Kusunoki et al. 1999, 2002) and much slower than that in the graphene growth. This low heating rate might be necessary to enhance the migration of carbon atoms, assembling and leading to formation of carbon nanocaps.
3. Effect of ambient oxygen gas
In early studies, Kusunoki et al. proposed a chemical reaction process for CNT growth by surface decomposition of SiC as follows (Kusunoki et al. 1999):
In their CNT growth, heating of SiC was carried out using a vacuum electric furnace whose pressure was about 10-2 Pa. In this circumstance, residual gases, such as O2, H2O, CO etc., should exist in the furnace during the growth. They considered that these residual gases reacted with Si atoms on SiC surface, leading to the desorption.
Fig. 11. shows C 1s XPS spectra for 6H-SiC at 1270ºC under various oxygen pressures. The main peak at around 284.3 eV and the shoulder peak at around 282.7 eV are derived from carbon sp2 and SiC components, respectively, and these spectra are normalized with the main peaks. It is clearly shown that the ratio of the peak intensity of the SiC component to that of the carbon sp2 component decreased as the oxygen partial pressure increased, confirming that oxygen enhances the decomposition of the SiC surface and promoting the carbon nanocap formation. Thus, surface reaction represented with eq. (2) seems to be reasonable. In the case of CNT growth in a vacuum electric furnace, the main reaction may be replaced by the reactions as follows, since residual gases are mainly composed of CO and H2O.
As for the reaction of oxygen with SiC surface, thermal oxidation of SiC surface has been an interesting theme from a fundamental surface science perspective, since it is chemically much more complex than that of Si. So far, three distinct regions have been observed in the phase diagram describing the SiC+O2 interaction; “passive oxidation”, “active oxidation” and “surface segregation of carbon” (Song & Smith, 2002). In the passive oxidation, which occurs under high oxygen pressure and low temperature, SiO2 layer grows on the surface, and, in the active oxidation, which occurs under low oxygen pressure and high temperature, SiC surface is etched by the reaction with oxygen. In addition, under the lower oxygen pressure and the higher temperature, surface segregation of carbon has been observed. The CNT growth by surface decomposition of SiC should occur under the third mode where segregation of carbon atoms proceeds on the surface. However, it has not been well investigated between the reaction with oxygen gas and the CNT growth in surface decomposition of SiC. To clarify these, it is necessary to minutely control oxygen pressure and investigate the relation among the pressure, temperature and surface decomposition of SiC during the CNT growth.
Fig. 12 shows a phase diagram for the interaction of O2 with the SiC surface, where formation of carbon nanocaps and SiO2 are separately shown (Maruyama et al., 2007). All data were obtained by heating the 6H-SiC(000-1) in a vacuum electronic furnace and in a UHV chamber where the oxygen gas was introduced by controlling the pressure, and the heating rates were kept below 1ºC/min which were low enough to form carbon nanocaps. In the figure, the surface covered with carbon nanocaps, elemental carbon (not formed nanocaps), SiO2, and tiny amount of SiO2 (weak SiO2 peak was observed in XPS spectra) are shown as open circles, open squares, solid squares and solid triangles, respectively. In addition, the boundary between passive oxidation (P. O.) and active oxidation (A. O.) and that between active oxidation (A. O.) and “surface segregation of carbon (S. C) ” which were obtained by extrapolating the experimental results for thermal oxidation of 6H-SiC(000-1) reported by Song and Smith (Song & Smith 2002) to wider ranges in both oxygen pressure and temperature. In their data, the critical oxygen pressure at which the active-transition occurred was estimated to be,
Previous studies reported that graphitization of amorphous carbon occurs just above 1000ºC (Botti, 2005; Chadderton & Chen, 1999). In the case of CNT growth by surface decomposition of SiC, both accumulation of sufficient carbon atoms and carbon migration are important to form carbon nanocap. This might lead to necessity of the higher temperature. On the other hand, SiO2 layers were formed on the SiC surface in the passive oxidation region, except for the UHV region, where slight elementary carbon was seen because of insufficient oxygen gas. Although, in our experimental result, data were not obtained in the just active oxidation region, shown between the dashed line and solid line, it was clearly shown that carbon nanocaps were formed above 1200ºC in the “surface segregation of carbon” region.
4. Effect of ambient hydrogen gas
Effect of hydrogen gas on carbon nanocap and CNT formation is another interesting theme, since reduction of residual oxides on SiC surfaces has been reported after annealing SiC in ambient hydrogen gases (Seyller, 2004) and this may affect the formation of carbon nanocaps. Figure 13 shows survey XPS spectra for SiC(000-1) surfaces after annealing at 1250ºC for 30 min in a hydrogen atmosphere at various pressures (Ueda et al., 2010). The O 1s peak was not observed for the samples annealed at 10-3 and 10-2 Pa, whereas the O 1s peak was clearly observed for that annealed at 10-4 Pa. Figures 14(a) and (b) shows Si 2p and C 1s spectra for 6H-SiC(000-1) after annealing at 1250ºC in a hydrogen atmosphere of 10-2 Pa.
For comparison, Si 2p spectrum of the sample after annealing at 1250ºC in a UHV is also shown in (c). In the Si 2p spectrum for the sample after annealing in a hydrogen atmosphere, only SiC component was observed and the component of SixCyOz related oxides were negligible, while there still remain some oxides after annealing in a UHV. These results confirm removal of residual SixCyOz oxides by annealing a hydrogen atmosphere of sufficient high pressure, as reported by Seyller. Figure 14(b) shows a C 1s XPS spectrum of the same sample. Except for a peak derived from adsorbed CO molecules, the spectrum consists of only a SiC-component at 282.7 eV and carbon sp2 hybridization bonding at 284.3 eV, and no SixCyOz related oxides components were observed, consistent with the Si 2p spectrum in (a).
STM images of SiC(000-1) substrates after annealing at 1200ºC for 30 min in the UHV and a hydrogen atmosphere of 10-2 Pa are shown in Fig. 15 (Ueda et al., 2010). The images reveal that convexities corresponding to carbon nanocaps were presented over the entire surface, independent of the annealing conditions. On the sample surface annealed in a UHV, the resulting carbon nanocaps were mainly in the diameter range of 1-3 nm, but nanocaps of more than 4 nm were observed. There were also some areas where no carbon nanocaps were formed. On the other hand, carbon nanocaps whose diameters were mainly distributed between 3 and 4 nm were formed on the SiC surface annealed in a hydrogen atmosphere, as shown in Fig. 15(b). The change in the diameter distribution is summarized in Fig. 16, which shows histograms of carbon nanocap diameter after annealing in a hydrogen atmosphere of 10-2 Pa. They indicates that the distribution of carbon nanocap diameter after annealing in the hydrogen atmosphere was narrower than that found in the UHV, demonstrating that the homogeneity of carbon nanocaps can be improved by reducing the native oxide from the SiC(000-1) surface.
XPS result suggests that native SixCyOz oxide on 6H-SiC still remain even after carbon nanocap formation without pre-annealing in H2 ambient (Fig. 5(b)). Therefore, carbon nanocap formation should be disturbed among the areas covered with the oxides, because those areas should be suppressed sublimation of Si atoms. Such a localized suppression will induce an inhomogeneous distribution of an amorphous carbon layer, which, in turn, will have a deleterious effect on the size uniformity of the resulting carbon nanocaps. On the other hand, if there is no native oxide on the surface, simultaneous desorption of Si atoms would occur and homogeneous distribution of amorphous carbon layer would be formed. This process would produce carbon nanocaps with a uniform size.
CNT growth by surface decomposition of SiC is not only an interesting method as peculiar crystal growth but also has various characteristics useful for device applications. Development of this growth technique should directly lead to realization of CNT devices in future, replaced by conventional semiconductor devices. There still remain several problems to be solved for fabrication of CNT devices. To understand its growth mechanism is essential to control structural parameters of CNTs, which is absolutely necessary for realization of CNT devices. To clarify the formation process of carbon nanocaps, which are precursors of CNTs, determining the structures of CNTs, is the first step to reach this objective.
The authors are grateful to Prof. Amemiya for his help during NEXAFS measurements and also thank Dr. Ito, Mr. Kawamura, Ms. Fujita and Mr. Tanioku for assisting with our study on STM observation. A part of this work was supported by the Japan Society for the Promotion of Science (JSPS), a Grant-in-Aid for Scientific Research (C) 17560015 and 21510119, and by the Ministry of Education, Culture, Sports, Science and Technology (MEXT) through the 21st century COE program “Nano Factory”. We are grateful to Professor T. Yokoyama and Dr. T. Nakagawa of the Institute for Molecular Science (IMS), Okazaki, for providing the XPS facility through the Nanotechnology Support Project in Central Japan, financially supported by the Nanotechnology Network. The NEXAFS measurements were performed under the approval of Photon Factory Program Advisory Committee (Proposal No. 2009G530). We would also like to to thank the Academic Research Institute of Meijo University for their financial support.