Chemical State and Morphology of Co-Mo Catalysts on Quartz and Silicon Substrates for Growth of Single-Walled Carbon Nanotubes

Transcription

1 Chemical State and Morphology of Co-Mo Catalysts on Quartz and Silicon Substrates for Growth of Single-Walled Carbon Nanotubes Minghui Hu, + Yoichi Murakami, ++ Masaru Ogura, + Shigeo Maruyama, ++ Tatsuya Okubo,+ Department of Chemical System Engineering, and Department of Mechanical Engineering, The University of Tokyo, Hongo, Bunkyo-ku, Tokyo , Japan Received ABSTRACT Bimetallic Co-Mo catalysts with a stoichiometry of and a nominal thickness of a few monolayers (~1 nm) were prepared on quartz and silicon substrates from metal acetate solutions, by using a procedure of dip coating, calcination and reduction. X-ray photoelectron spectroscopy and high-resolution transmission electron microscopy studies on the chemical state and morphology of the Co-Mo catalysts during their preparation and activation provided critical evidence about the mechanism of their high selectivity and activity exhibited in the growth of high-quality single-walled carbon nanotubes (SWNTs) during an alcohol catalytic chemical vapor deposition process. The results showed that catalyst particles with diameters of 1 2 nm and a number density of ~ /m 2 correlate were well dispersed on the substrates, and that such nano-sized particles account for the selective growth of high-quality SWNTs with diameters of nm. Based on our results, Co molybdate underlayers between metallic particles and substrates surfaces stabilize nano-sized metallic cobalt particles against agglomerating into larger particles. This study reveals that the stable existence of well-dispersed nano-sized catalyst particles is required for the highly selective growth of high-quality SWNTs in chemical vapor deposition processes. Corresponding Author. Fax: okubo@chemsys.t.u-tokyo.ac.jp + Department of Chemical System Engineering ++ Department of Mechanical Engineering Page 1 of 30

2 Introduction Single-walled carbon nanotubes 1 (SWNTs) exhibit superior chemical reactivity, electrical conductivity, optical activity, and mechanical strength. 2 Catalytic chemical vapor deposition (CCVD) can produce large-scale SWNTs at low cost, 3,4 compared with arc discharge 5 and laser ablation 6 methods. Compared with monometallic catalysts, bimetallic Co-Mo catalysts formed with various methods show a high selectivity 7-10 and a seemingly preferential chirality distribution 11 for the growth of SWNTs. Especially, Co-Mo catalysts fabricated using a dip-coating process from metal acetate solutions followed by an alcohol CCVD process 12,13 can produce high-quality SWNTs directly on oxide substrates without supporting materials. 14,15 This will significantly promote the development of silicon-based SWNT device systems. To control the growth and structure of SWNTs, it is important to understand why Co-Mo catalysts exhibit characteristics of high selectivity and activity. At temperatures as low as ~800 C, monometallic Mo catalysts are inactive for the growth of SWNTs, whereas bimetallic Co-Mo catalysts show an enhanced selectivity and activity compared with monometallic Co catalysts. 7,10,14-16 Moreover, Co and Mo are not uniformly distributed in each particle of the bimetallic catalysts. 17 Reports suggest that at an optimum Co/Mo ratio, Co molybdate upperlayers stabilize Co species during reduction, and Mo carbides release active carbon to catalytic metallic Co particles from which SWNTs selectively grow during the CVD reaction Although Co-Mo catalysts have also been studied for decades due to their high activity in hydrodesulfurization, 21,22 the correlation between the chemical state and the morphology of Co-Mo catalysts during their preparation and activation has never been verified. The effect of Mo in bimetallic catalysts on the growth of SWNTS still remains unclear. In this study, to clarify this effect, we prepared Co-Mo catalysts on quartz substrates and silicon substrates from metal acetate solutions by using a procedure of dip coating, calcination and reduction. We characterized the chemical state of metal species using X-ray photoelectron spectroscopy (XPS), and characterized catalyst morphology at the nanoscale by using high-resolution transmission electron microscopy (HRTEM). A model was then developed to explain why Co-Mo catalysts prepared from metal acetates by using the dip-coating method have high selectivity and activity for the growth of high-quality SWNTs in alcohol catalytic CVD processes. Experimental Section Catalyst preparation and SWNT growth. Co-Mo catalysts were prepared on fused quartz substrates and silicon substrates with naturally oxidized layers by using a procedure of dip coating, calcination and reduction. 14 First, Co acetate (CH 3 COOH) 2 Co-4H 2 O and Mo acetate (CH 3 COOH) 2 Mo (Co:Mo = 0.01%:0.01% in metal weight = 2:1 in atomic ratio) were dissolved in ethanol under sonication, and then transferred onto the substrates by using the dip-coating process at a withdrawal Page 2 of 30

3 speed of 4 cm/min. Assuming that the liquid film had a thickness of ~50 µm, Co-Mo catalysts with a nominal thickness of ~1 nm, i.e., a few monolayers (MLs), were formed on the substrate surfaces, estimated from the concentrations of metal acetates and densities of bulk metals. After drying in air at room temperature (r.t.), the substrates were calcined in an open furnace at 400 C for 5 min. These substrates are referred to here as calcined catalysts/substrates. Then, these calcined Co-Mo catalysts were activated with a hydrogen reduction process prior to the CVD reaction. The freshly calcined substrates were immediately transferred from the furnace into a tubular CVD reactor that was then evacuated by a rotatory pomp. They were heated from r.t. to 800 C in 30 min in an Ar/H 2 (3% H 2 ) stream of 300 sccm at 300 Torr. These substrates are referred here to as reduced catalysts/substrates. Finally, SWNTs were grown on these reduced substrates by the alcohol CVD process. 12,13 Immediately after reduction, alcohol vapor was supplied as a carbon source at 800 C at 10 Torr for 1 h. After that, the substrates were cooled to r.t. in an Ar/H 2 stream. These products are referred to here as as-grown SWNTs. Characterization of SWNTs and catalysts. The calcined catalysts, reduced catalysts, and as-grown SWNTs, were characterized using spectroscopy and microscopy methods. The selectivity and quality of the as-grown SWNTs were evaluated using micro-raman scattering and scanning electron microscopy (SEM). The Raman scattering measurements were done using a Chromex 501is spectrometer and an Andor DV401-FI CCD system equipped with a Seki Technotron STR250 optical system. The morphology of the as-grown SWNTs was observed using a Hitachi S-900 field-emission SEM (FE-SEM). Chemical state of the calcined and reduced Co-Mo catalysts was characterized by using ex-situ XPS and the morphology by using TEM. XPS measurements were done using a PHI 1600 X-ray photoelectron spectrometer equipped with an Mg Kα ( ev) source operating at 300 W and an electron take-off angle of 45. The dual-anode X-ray source was located at 54.7 relative to the analyzer axis. Binding energies were referenced to the C 1s peak at ev to compensate for the charging effect. Because Co 2p binding energies are broad, only Co 2p 3/2 components were decomposed. The binding energy curves were fitted using a mixed Gaussian/Lorentzian curve after subtracting the spectral background using the Shirley algorithm, 23 in which spin-orbital-splitting intensities were fixed at the theoretical ratios. TEM observations were done using a JEOL JEM-2010F operating at 200 kv with a point resolution of 0.19 nm. For TEM specimens, a 10-nm SiO 2 layer was grown onto the substrates by sputter deposition to avoid structural variations occurring during TEM specimen preparation and observations. These specimens were then prepared using conventional mechanical grinding, polishing, and dimpling, followed by Ar ion milling at an acceleration voltage of 4 kev and an incidence angle of 6. To examine the influence of oxidation and contamination on the results during ex-situ XPS and TEM analyses, as soon as the calcined and reduced substrates coated with Co-Mo catalysts were Page 3 of 30

4 removed from the furnace or the CVD reactor, the quartz substrates were sealed in a plastic box filled with Ar gas, whereas the silicon substrates were exposed to air without any protection. Results Selectivity and quality of as-grown SWNTs. Figure 1 shows the Raman spectra of as-grown SWNTs on the quartz substrates measured with 488-nm excitation. The strong peak at 1590 cm -1 (G-band) arises from an in-plane oscillation of carbon atoms in the sp 2 graphene sheet. The weak peak at 1350 cm -1 (D-band) reflects the degree of defects or dangling bonds contained in the sp 2 arrangement of graphene sheet. The high G/D ratio of about 30 indicates the highly selective growth of high-quality SWNTs, and the absence of amorphous impurities has been confirmed by previous TEM observations. 14,15 The inset frame in Figure 1 shows the Raman spectrum of radial breathing mode (RBM) of these as-grown SWNTs. According to the relationship d = 248 / υ, where d (nm) is the diameter and υ (cm -1 ) is the Raman shift of a SWNT, 24,25 the estimated diameters of these SWNTs were nm. Figure 2 shows the typical morphology of as-grown SWNTs on the quartz substrates after 1 h CVD at 800 C. These SWNTs self-assembled into randomly oriented bundles with diameters of 7 15 nm and lengths of nm. No metal particles with diameters larger than 5 nm were observed at the substrate edge. Such metal particles often appear for catalytic metals prepared with other methods. 16,26-28 The selectivity and quality of as-grown SWNTs on the silicon substrates were also investigated using Raman spectroscopy and FE-SEM. 14,15 It seems that except for that the quartz substrates showed a higher reproducibility for the growth of SWNTs than the silicon substrates, no distinct difference was found between the two substrates. Moreover, we previously investigated catalyst performances for different compositions of catalysts for the growth of SWNTs, and found that the performance had the following order: Co-Mo acetates > Co acetates >> Co (nitrides)-mo (acetates) >> Mo acetates. 15 Based on these results, we concluded that the bimetallic Co-Mo catalysts on the quartz substrates and silicon substrates prepared from metal acetate solutions using the dip-coating method show a high activity in the alcohol CCVD growth of SWNTs with high selectivity and high quality. Chemical state of calcined and reduced Co-Mo catalysts. Figure 3 shows binding energies of C 1s, Si 2p, and O 1s levels for the calcined and reduced catalysts on the quartz substrates. The binding energies of Si 2p at ev (Figure 3b), and O 1s at ev (Figure 3c) agree well with quartz references (Si: ev, O: ev). 29,30 This good agreement reveals that the shift in binding energy due to the charging effect was well compensated using the binding energy of C 1s as the reference. For both the calcined and reduced catalysts, the spectra of C 1s (Figure 3a) and Si 2p (Figure 3b) indicate the absence of acetate residues ( ev), 31,32 metal silicides ( ev), and metal carbides ( ev) Moreover, seen from the spectra of O 1s Page 4 of 30

5 (Figure 3c), metal oxide species existed in both the calcined and reduced samples, indicated by the weak peaks at ev. 39 Figure 4 shows binding energies of Mo 3d 5/2 and 3d 3/2 levels for the calcined and reduced catalysts on the quartz substrates. Both catalysts showed a pair of spin-orbit peaks at and ev. For the reduced catalysts, a pair of new spin-orbit peaks appeared at and ev. The Mo 3d 5/2 component at ev is attributed to Mo 6+ species in either MoO 3 ( ev) 18,38,40-43 or non-stoichiometric Co molybdates, CoMoO x ( ev for stoichiometric Co molybdates where x = 4). 18,41,42 Without other evidence, it is difficult to further identify the chemical state of this component due to the small difference in binding energy between MoO 3 and CoMoO x. On the other hand, the Mo 3d 5/2 component at ev with a distance of 3.4 ev to Mo 6+ 3d 5/2 peak is attributed to Mo 4+ species in MoO 2 because of the good agreement with the binding energy of Mo 4+ in MoO 2 ( ev) and its distance ( ev) to Mo 6+ 3d 5/2 in MoO 3. 18,38,40,41,43 Based on these results, we infer that that Mo oxide species in Co-Mo catalysts were, partially at least, reduced into metallic Mo under our experimental conditions because bulk MoO 3 can be reduced into MoO 2 by H 2 at 400 C, and further into metallic Mo above 500 C. However, metallic Mo was not detected by our XPS analyses, possibly because highly reactive metallic Mo was oxidized as soon as the samples were removed from the CVD reactor. Figure 5 shows binding energies of Co 2p 3/2 and 2p 1/2 levels for the calcined and reduced Co-Mo catalysts on the quartz substrates. As indicated by the arrows and dotted lines, the spectrum for the reduced catalysts differed from that for the calcined ones in two aspects. One is the appearance of a new component at ev, which is attributed to metallic Co ( ev). 18,41,44,45 The other is the decreased distance by as large as 1.6 ev between the 2p 3/2 and 2p 1/2 spin-orbit peaks. These results indicate that besides the formation of metallic Co, other changes in chemical state of Co oxide species occurred during reduction. To clarify what happens to the Co oxide species in bimetallic catalysts during reduction, we decomposed and fitted the Co 2p 3/2 binding energies. As shown in Figure 6, the calcined and reduced catalysts had two characteristics in common. One is two 2p 3/2 peaks at ev and ev, and the other is two distinct shake-up satellite peaks at ev and ev. Only from binding energies of Co 2p 3/2, it is difficult to distinguish the component at ev from Co 2+ in CoO ( ev), 18,44-48 Co 2+/3+ in Co 3 O 4 ( ev) 18,42,45-48 and Co 3+ in Co 2 O 3 ( ev) In fact, it is well known that Co 2+ ions (tetrahedral sites) are in high spin state (s=3/2), whereas Co 3+ ions (octahedral sites) are diamagnetic (s=0). As a consequence, Co 2+ in CoO has an intense shake-up satellite peak at the higher binding energy side, whereas Co 2+/3+ in Co 3 O 4 and Co 3+ in Co 2 O 3 do not. 42,45,47,49,50,52-54 Due to the intense shake-up satellite peaks at the position ev higher than the 2p 3/2 peak at ev, the Co oxide species at ev should at least contain Co 2+ in CoO. However, the existence of a small amount of Co 3 O 4 and Co 2 O 3 can not be completely excluded because their binding energy peaks might be hidden within that of CoO and thus are difficult to separate by curve fitting. On the other hand, the component at Page 5 of 30

6 ev is attributed to Co 2+ incorporated in MoO 3 matrix, i.e., CoMoO x, which was supported by two facts. One is that the distance to its shake-up satellite peak was ev, consistent with that for Co ,45,47,52-54 The other is that the distance to the 2p 3/2 peak of Co 2+ in CoO was ev, similar to the distance of ev for 2p 3/2 peak of Co 2+ in CoMoO 4. 41,42,44,55 Therefore, the component at ev is attributed to Co 2+ in CoMoO x. The relatively large deviation of its binding energy from that of CoMoO 4 might result from the polarization effect in a highly oxidized matrix, 44,55,56 or from the specific structure due to the non-stoichiometry. The change in peak intensities indicates that more Co 2+ in CoMoO x was formed at the expense of Co oxides. Based on the above results, we concluded that during calcination, Co acetates were decomposed into Co oxide species in the form of CoO and CoMoO x, and during reduction, CoO was converted to CoMoO x and metallic Co. The chemical state of the calcined and reduced catalysts on the silicon substrates was also investigated. Table 1 shows the results of binding energies, peak intensities, and full width of half maximums for the Co-Mo catalysts on the quartz and silicon substrates. The Co-Mo catalysts on the silicon substrates exhibited similar characteristics as those on the quartz substrates, except for two differences seen in the reduced catalysts. One is that the Mo 3d 3/2 binding energy of the new species at ev on the silicon substrates appeared at a higher binding energy side than those at ev on the quartz substrates. The component at ev is attributed to Mo oxide species (Mo 5+ ) located between MoO 3 and MoO 2. The other is that metallic Co was not detected on the silicon substrates, but was detected on the quartz substrates. These two differences can be understood from the difference in the exposure of the samples to air before XPS analyses, i.e., the reduced catalysts on the silicon substrates without any protection were oxidized more than those on the quartz substrates sealed in the Ar environment after they were removed from the CVD reactor. Table 2 shows surface atomic ratios of the calcined and reduced quartz and silicon substrates coated with Co-Mo catalysts calculated from peak intensities in the XPS spectra and atomic sensitivity factors. The metals only account for less than 10% of the tatal composition, and is as low as that of carbon impurities. This low quantity of metals supports the approximate evaluation (see Experimental section) that the substrates might be covered with only a few monolayers (~1 nm) of the bimetallic catalysts. The atomic ratios of Co to Mo were relatively constant at after calcination and after reduction, approximately consistent with the initial value of 2 in metal acetate solutions. However, the atomic ratios of Co and Mo to total elements decreased by 63 78% and 68 64% after reduction, respectively, compared with those before reduction. The decrease in metal atomic ratios is attributed to the decrease in metal coverage, arising from either de-wetting or volume shrinkage. 45 In-situ experiments clearly indicated that if the re-oxidation of once reduced metal particles restores its original atomic composition, the volume shrinkage during reduction from Co oxides to metallic Co is responsible for the decreased coverage; if it is not restored, the weakened wettability during reduction is responsible. 45 Our XPS analyses have showed that on the quartz substrates, Co species in the reduced catalysts existed as CoO and metallic Co, whereas on the Page 6 of 30

7 silicon substrates, they only existed as CoO, due to the difference in exposure to air. Thus, the quartz and silicon substrates can be reasonably considered as partially reduced and re-oxidized samples, respectively. Because both substrates showed similar decreases in atomic percentages of Co (63% on quartz and 78% on silicon) and Mo (68% on quartz and 64% on silicon), the reduction might lead to the de-wetting of catalyst particles on the substrates. Morphology of calcined and reduced Co-Mo catalysts. Here, only Co-Mo catalysts on the silicon substrates were reported because the above XPS results and previous studies on the growth of SWNTs 14,15 showed the similarity between the Co-Mo catalysts on the quartz substrates and those on the silicon substrates in chemical state of catalysts and the growth of SWNTs. Figure 7 shows plan-view TEM and HRTEM images of the calcined and reduced Co-Mo catalysts on the silicon substrates. For both samples, well-dispersed nano-sized particles can be seen on the substrate surfaces, indicating that the dip-coating process using metal acetates is more suitable to fabricate catalyst particles with high dispersion and controlled sizes, compared with other methods For the calcined samples (Figure 7a), most catalyst particles were about 1 3 nm in diameter, but some areas had larger particles with diameters of 4 6 nm, as indicated by arrows. On the other hand, for the reduced samples (Figure 7d), these larger particles were absent. Comparing Figures 7b and 7e, the catalyst particles after reduction seem to be dispersed more uniformly with smaller sizes than those after calcination. These differences between the calcined and reduced catalysts support the conclusion from XPS analyses that the de-wetting of catalyst particles occurred on the substrates during reduction, which results in the decrease in particle size and metal coverage. These well-dispersed catalyst particles had diameters of 1 2 nm and a number density of ~ /m 2. Because these nano-sized particles did not agglomerate into larger particles during their de-wetting, SWNTs with diameters of nm were grown in the subsequent alcohol CVD reaction. 14,15 HRTEM analyses revealed that most of the nano-sized catalyst particles were amorphous, and some of them were crystalline. These crystalline particles (Figures 8c and 8f) had lattice distances of Å, which are approximately consistent with CoO (200) lattice constant (2.13 Å), but significantly differ from those of Co (111) (2.05 Å), Co 3 O 4 (311) (2.44 Å), MoO 2 (110, 111 ) (3.41 Å), MoO 3 (021) (3.26 Å), and CoMoO 4 (002, 220) (3.36 Å) lattices. Because the Co/Mo atomic ratio was about , i.e., Co was in excess, we further concluded that most of those nano-sized particles, whether crystalline or amorphous, should mainly be composed of CoO. This is consistent with the XPS result that Co oxides were detected on both the calcined and reduced substrates because once reduced metallic Co was re-oxidized in air. Discussion The evolution in the chemical state and morphology of Co-Mo catalysts on the quartz and silicon substrates can be summarized as follows based on our results. The Co and Mo catalysts with a nominal Page 7 of 30

8 thickness of a few MLs (~1 nm) are coated onto substrates with a stoichiometry of by using the dip-coating method from dilute metal acetate solutions. After calcination in air at 400 C, metal acetates are decomposed into metal oxides, CoO and MoO 3, and CoMoO x. These species form nano-sized particles that are well dispersed on the substrates. After reduction in H 2 up to 800 C, metallic Co and Mo, and CoMoO x are formed at the expense of CoO and MoO 3. No distinct agglomeration occurs although the reduction weakens the wettability of catalyst particles. These particles consisting of metallic Co are uniformly dispersed on substrates and have diameters of about 1 2 nm and a number density of ~ /m 2. Based on these results, we developed the model shown in Figure 8 to describe the chemical composition and morphology of the Co-Mo catalysts during calcination and reduction, and to clarify their high activity and selectivity during the growth of SWNTs in alcohol CVD processes. As shown in Figure 8a, after calcination in air at 400 C, Co and Mo acetates are decomposed to metal oxide species. If well-dispersed Mo and Co coexist on SiO 2 surfaces, Mo preferentially promotes the formation of metal oxides at catalyst/sio 2 interfaces, because Mo has a stronger affinity to oxygen than Co. 57 Then, part of the excess Co easily diffuses into MoO 3, 55 and forms CoMoO x underlayers/boundaries, whereas the residue Co exists as CoO particles either dispersed on CoMoO x underlayers or attached to CoMoO x boundaries. The decomposition of metal acetates and the formation of CoMoO x underlayers/boundaries significantly contribute to the size and dispersion of CoO particles. Some studies indicate that compared with ionic metal nitrides, calcined metal acetates form well-dispersed nano-sized metal oxide particles on SiO 2 surfaces, because metal acetates have strong interactions with SiO 2 substrates possibly resulting from the coordination of carboxylic groups. 58,59 In addition, for Co-Mo catalysts, the formation of CoMoO x underlayers/boundaries further restricts the agglomeration of nano-sized metal oxide particles into large ones during calcination. Therefore, compared with Co acetates and Co (nitrides)-mo (acetates), Co-Mo acetates show the best performance for growth of SWNTs. After reduction in hydrogen up to 800 C, as shown in Figure 8b, CoO and MoO 3 are reduced into metallic Co and Mo, respectively, whereas CoMoO x underlayers/boundaries remain unchanged, because Co molybdates have a higher stability against reduction compared with monometallic Co and Mo oxides, possibly due to the polarization effect. 60,61 Although metallic Co particles can vigorously migrate on SiO 2 surfaces, 45 they are easily trapped on CoMoO x underlayers/boundaries, owing to the strong interactions between MoO 3 and metallic Co. 55 Because Co is in excess, most CoMoO x underlayers are mainly covered with metallic Co, but some might be covered with Co and a little Mo. Such Co and Mo within a single particle might be distributed separately, because bulk Co-Mo alloys can only be formed at temperatures higher than 1335 C. 62 Although metallic Co particles might have a relatively weakened wettability compared with CoO particles, the interactions between CoMoOx and Co are strong enough to stabilize nano-sized metallic Co particles, limit their mobility, and prevent them from agglomerating into large particles. Without strong interfacial interactions, such nano-sized metal particles could not be well dispersed at a number density as high as ~10 17 /m Generally, the Page 8 of 30

9 growth of SWNTs requires catalyst particles with diameters less than 2 3 nm These well-dispersed Co particles with diameters of 1 2 nm directly confirmed in HRTEM images (figures 7d, e, and f) correlate well with the selective growth of high-quality SWNTs with diameters of nm. 14,15 If the interactions between Co and Mo oxide species are interrupted with the addition of sodium salts, 42 the catalyst selectivity for the growth of SWNTs disappears due to the agglomeration of Co particles. 20 Because we did not investigate the catalyst structures during or after the growth of SWNTs, we can only speculate as follows on their reactive and catalytic mechanism during the CVD reaction, based on published studies. The ethanol vapor is decomposed on nano-sized metal particles, leading to the formation of surface- or bulk-phase metal carbide species. 18,38,70,71 The growth of SWNTs starts with the formation of graphic sheets on cobalt particles, because Mo itself does not show any activity in carbon nanotube growth at this temperature. 7,15,19 An appropriate amount of oxygen radicals produced from the decomposition of ethanol on catalytic metal particles removes amorphous carbon impurities with dangling bonds in SWNTs grown on metal particle surfaces. Such impurities are obstacles in creating high-purity SWNTs. This etching effect makes the alcohol CCVD process exceptionally suitable for synthesizing high-quality SWNTs, 12,13 compared with other carbon sources, such as carbon monoxide and methane. This study revealed that the stable existence of well-dispersed nano-sized catalytic metal particles is indispensable for the highly selective growth of high-quality SWNTs in CVD processes. The combination of the alcohol CCVD processes and the Co-Mo catalysts prepared from metal acetate solutions by using the dip-coating method, satisfies the requirement of in-situ purification of SWNTs and size control of catalyst particles, thereby being able to produce SWNTs with high selectivity and high quality. However, some issues on the working state of catalysts, such as do molybdenum carbide species act as a main site for precursor decomposition and a carbon sink to release active carbon to cobalt particles, need further investigation. Conclusions Bimetallic Co-Mo catalysts prepared from metal acetate solutions with a dip-coating process showed high selectivity and activity in fabricating high-quality SWNTs with diameters of nm directly grown on quartz and silicon substrates. XPS and HRTEM studies on the chemical state and morphology of calcined and reduced Co-Mo catalysts provided critical evidence for the high selectivity and high activity of these catalysts. The results showed that calcination in air at 400 C decomposes metal acetates to metal oxide species. The excess CoO exists as nano-sized wetting particles either dispersed on or attached to CoMoO x underlayers/boundaries. After treatment in hydrogen up to 800 C, CoO is reduced to metallic Co which then forms well-dispersed Co particles with diameters of about 1 2 nm and a Page 9 of 30

10 number density of ~ /m 2. CoMoO x underlayers at catalyst/sio 2 interfaces stabilize nano-sized metallic Co particles against agglomerating into larger particles by limiting their mobility. This study suggests that the stable existence of well-dispersed nano-sized catalytic metal particles is required for the highly selective growth of high-quality SWNTs in CVD processes. The combination of first preparing Co-Mo catalysts on quartz and silicon substrates prepared from metal acetate solutions by using the dip-coating method, and then using an alcohol CCVD process is suitable for synthesizing SWNTs with high selectivity and high quality. Acknowledgments. We gratefully acknowledge Prof. Y. Shimogaki (The University of Tokyo) for his support in the XPS analyses, Mr. H. Tsunakawa (The University of Tokyo) for his assistance in TEM observations, and Mr. T. Sugawara (The University of Tokyo) for his technical advice about FE-SEM observations. Part of this work was financially supported by from JSPS and #13GS0019 from MEXT. Page 10 of 30

11 FIGURE CAPTIONS Figure 1. Raman spectra of SWNTs synthesized on quartz substrates measured with 488-nm excitation. Figure 2. Cross-sectional SEM image of as-grown SWNTs on quartz substrates, observed at a fracture edge with a tilted angle. Figure 3. XPS spectra of (a) C 1s, (b) Si 2p, and (c) O 1s levels for calcined and reduced Co-Mo catalysts on quartz substrates. Figure 4. XPS spectra of Mo 3d5/2 and 3d3/2 for calcined and reduced Co-Mo catalysts on quartz substrates. Figure 5. XPS spectra of Co 2p3/2 and 2p1/2 for calcined and reduced Co-Mo catalysts on quartz substrates. Figure 6. XPS spectra of Co 2p3/2 levels for calcined and reduced Co-Mo catalysts on quartz substrates. Figure 7. Plan-view TEM images of (a) calcined and (d) reduced Co-Mo catalysts on silicon substrates, and HRTEM images of (b) (c) calcined and (e) (f) reduced Co-Mo catalysts on silicon substrates. Figure 8. Chemical state and morphology of Co-Mo catalysts on quartz and silicon substrates during (a) calcination and (b) reduction. Page 11 of 30

12 Table 1. Binding energies (BE), peak intensities (PI), and full width of half maximums (FWHMs) of all the elements for calcined and reduced Co-Mo catalysts on quartz and silicon substrates. Quartz substrates Silicon substrates Element Calcined Reduced Calcined Reduced BE / PI / FWHM BE / PI / FWHM BE / PI / FWHM BE / PI / FWHM (ev / cps / ev) (ev / cps / ev) (ev / cps / ev) (ev / cps / ev) C 1s / 3323 / / 2024 / / 1095 / / 1451 / 1.46 O Si Mo 1s 2p 3d 5/2 3d 3/ / / / / / / / / / 2771 / / 1098 / / 1200 / / 437 / / / / / / 3207 / / 1599 / / 1822 / / 3557 / / 8972 / / 5373 / / 4348 / / 1997 / / 1603 / / 1001 / / 5985 / / 3584 / / 2900 / / 1332 / / 1069 / / 668 / 2.45 Co 2p 3/2 shake-up satellite / / / 5119 / / 6132 / / 1662 / / / / / / 5997 / / 2803 / / 1805 / / / / 8561 / / 3566 / / 1577 / / 4747 / / 1413 / / 360 / / 1019 / 2.77 Page 12 of 30

13 Table 2. Surface composition of quartz and silicon substrates coated with Co-Mo catalysts after calcination and after reduction. Co-Mo catalysts Treatment Atomic ratio (%) Co Mo Si O C On quartz On silicon Calcination Reduction Calcination Reduction Page 13 of 30

14 References (1) Iijima, S.; Ichihashi, T. Nature 1993, 363, 603. (2) Saito, R.; Dresselhaus, G.; Dresselhaus, M. S. Physical Properties of Carbon Nanotubes; Imperial College Press: London, (3) Kong, J.; Cassell, A. M.; Dai, H. J. Chem. Phys. Lett. 1998, 292, 567. (4) Su, M.; Zheng, B.; Liu, J. Chem. Phys. Lett. 2000, 322, 321. (5) Journet, C.; Maser, W. K.; Bernler, P.; Loiseau, A.; de la Chapelle, M. L.; Lefrant, S.; Denlard, P.; Lee, R.; E., F. J. Nature 1997, 388, 756. (6) Thess, A.; Lee, R.; Nikolaev, P.; Dai, H. J.; Petit, P.; Robert. J; Xu, C. H.; Lee, Y. H.; Kim, S. G.; Rinzler, A. G.; Colbert, D. T.; Scuseria, G. E.; Tomanek, D.; Fischer, J. E.; Smalley, R. E. Science 1996, 273, 483. (7) Kitiyanan, B.; Alvarez, W. E.; Harwell, J. H.; Resasco, D. E. Chem. Phys. Lett. 2000, 317, 497. (8) Ago, H.; Ohshima, S.; Uchida, K.; Yumura, M. J. Phys. Chem. B. 2001, 105, (9) Yoon, Y. J.; Bae, J. C.; Baik, H. K.; Cho, S. J.; Lee, S.; Song, K. M.; Myung, N. S. Physica B 2002, 323, 318. (10) Alvarez, W. E.; Pompeo, F.; Herrera, J. E.; Balzano, L.; Resasco, D. E. Chem. Mater. 2002, 14, (11) Bachilo, S. M.; Balzano, L.; Herrera, J. E.; Pompeo, F.; Resasco, D. E.; Weisman, R. B. J. Am. Chem. Soc. 2003, 125, (12) Maruyama, S.; Kojima, R.; Miyauchi, Y.; Chiashi, S.; Kohno, M. Chem. Phys. Lett. 2002, 360, 229. (13) Murakami, Y.; Miyauchi, Y.; Chiashi, S.; Maruyama, S. Chem. Phys. Lett. 2003, 374, 53. (14) Murakami, Y.; Miyauchi, Y.; Chiashi, S.; Maruyama, S. Chem. Phys. Lett. 2003, 377, 79. (15) Murakami, Y.; Chiashi, S.; Miyauchi, Y.; Maruyama, S. Jpn. J. Appl. Phys., Part , Submitted. (16) Yoon, Y. J.; Bae, J. C.; Baik, H. K.; Cho, S. J.; Lee, S.; Song, K. M.; Myung, N. S. Chem. Phys. Lett. 2002, 366, 109. (17) Liao, X. Z.; Serquis, A.; Jia, Q. X.; Peterson, D. E.; Zhu, Y. T. Appl. Phys. Lett. 2003, 82, (18) Herrera, J. E.; Balzano, L.; Borgna, A.; Alvarez, W. E.; Resasco, D. E. J. Catal. 2001, 204, 129. (19) Alvarez, W. E.; Kitiyanan, B.; Borgna, A.; Resasco, D. E. Carbon 2001, 39, 547. (20) Herrera, J. E.; Resasco, D. E. J. Catal. 2003, in press. (21) Topsoe, H.; Clausen, B. S.; Candia, R.; Wivel, C.; Morup, S. J. Catal. 1981, 68, 433. (22) Okamoto, Y.; Kubota, T. Catal. Today 2003, 86, 31. (23) Shirley, D. A. Phys. Rev. B 1972, 135, (24) Saito, R.; Dresselhaus, G.; Dresselhaus, M. S. Phys. Rev. B 2000, 61, (25) Jorio, A.; Saito, R.; Hafner, J. H.; Lieber, C. M.; Hunter, M.; McClure, T.; Dresselhaus, G.; Dresselhaus, M. S. Phys. Rev. Lett. 2001, 86, (26) Delzeit, L.; Chen, B.; Cassell, A.; Stevens, R.; Nguyen, C.; Meyyappan, M. Chem. Phys. Lett. 2001, 348, 368. (27) Nerushev, O. A.; Morjan, R. E.; Ostrovskii, D. I.; Sveningsson, M.; Jonsson, M.; Rohmund, F.; Campbell, E. E. B. Physica B 2002, 323, 51. (28) Zhang, R. Y.; Amlani, I.; Baker, J.; Tresek, J.; Tsui, R. K. Nano Lett. 2003, 3, 731. (29) Wagner, C. D.; Passoja, D. E.; Kinisky, H. F.; Six, H. A.; Jansen, W. T.; Taylor, J. A. J. Vac. Sci. Technol. 1982, 21, 933. Page 14 of 30

15 (30) Duval, Y.; Mielczarski, J. A.; Pokrovsky, O. S.; Mielczarski, E.; Ehrhardt, J. J. J. Phys. Chem. B. 2002, 106, (31) Gelius, U.; Heden, P. F.; Hedman, J.; Lindberg, B. J.; Manne, R.; Nordberg, R.; Nording, C.; Siegbahn, K. Phys. Scr. 1970, 2, 70. (32) Hagstrom, S. B. M.; Heden, P. O.; Lofgren, H. Solid State Commun. 1970, 8, (33) Shabanova, I. N.; Trapeznikov, V. A. J. Electron Spectrosc. Relat. Phenom. 1975, 6, 297. (34) Brainard, W. A.; Wheeler, D. R. J. Vac. Sci. Technol. 1978, 15, (35) Herranen, M.; Bauer, A. D.; Carlsson, J.-O.; Bunshah, R. F. Surf. Coat. Technol. 1997, 96, 245. (36) Ramqvist, L.; Hamrin, K.; Johannson, G.; Fahlman, A.; Nording, C. J. J. Phys. Chem. Solids 1969, 30, (37) Wang, H.; Wong, S. P.; Cheung, W. Y.; Ke, N.; Lau, W. F.; Chiah, M. F.; Zhang, X. X. Materials Science and Engineering C: 2001, 16, 147. (38) Solymosi, F.; Cserenyi, J.; Szoke, A.; Bansagi, T.; Oszko, A. J. Catal. 1997, 165, 150. (39) Nefedov, V. I.; Gati, D.; Dzhurinskii, B. F.; Sergushin, N. P.; Salyn, Y. V. Zh. Neorg. Khim. 1975, 20, (40) Chen, G.; Ford, T. E.; Clayton, C. R. J. Colloid Interface Sci. 1998, 204, 237. (41) Probst, M.; Vob, M.; Denecke, R.; Viscido, L.; Heras, J. M.; Borgmann, D.; Steinruck, H.-P. J. Electron Spectrosc. Relat. Phenom. 2001, , 539. (42) Parola, V. L.; Deganello, G.; Tewell, C. R.; Venezia, A. M. Appl. Catal. A: 2002, 235, 171. (43) Quincy, B. R.; Houalla, M.; Proctor, A.; Hercules, D. M. J. Chem. Phys. 1990, (44) Kleyna, R.; Mex, H.; Vos, M.; Borgmann, D.; Viscido, L.; Heras, J. M. Surf. Sci. 1999, , 723. (45) Riva, R.; Miessner, H.; Vitali, R.; Piero, G. D. Appl. Catal. A: 2000, 196, 111. (46) Wang, W.-J.; Chen, Y.-W. Appl. Catal. A: 1991, 77, 223. (47) Jimenez, V. M.; Espinos, J. P.; Gonzalez-Elipe, A. R. Surf. Interface Anal. 1998, 26, 62. (48) Brik, Y.; Kacimi, M.; Ziyad, M.; Bozon-Verdurazy, F. J. Catal. 2001, 202, 118. (49) Casella, I. G.; Guascito, M. R. Journal of Electroanalytical Chemistry 1999, 476, 54. (50) Christoskova, S. G.; Stoyanova, M.; Georgieva, M.; Mehandjiev, D. Materials Chemistry and Physics 1999, 60, 39. (51) Fulton, W. S.; Smith, G. C.; Titchener, K. J. Applied Surface Science 2003, in press. (52) Venezia, A. M.; Parola, V. L.; Deganello, G.; Cauzzi, D.; Leonardi, G.; Predieri, G. Appl. Catal. A: 2002, 229, 261. (53) Barreca, D.; Massignan, C. Chem. Mater. 2001, 13, 588. (54) Hassel, M.; Freund, H.-J. Surf. Sci. 1995, 325, 163. (55) Viscido, L.; Vob, M.; Borgmann, D.; Heras, J. M. J. Mol. Catal. A 2001, 167, 199. (56) Zazhigalov, V. A.; Haber, J.; Stoch, J.; Pyatnitzkaya, A.; Komashko, G. A.; Belousov, V. M. Applied Catalysis A: General 1993, 96, 135. (57) Hu, M.; Noda, S.; Komiyama, H. Surf. Sci. 2002, 513, 530. (58) Matsuzaki, T.; Takeuchi, K.; Hanaoka, T.; Arakawa, H.; Sugi, Y. Catal. Today 1996, 28, 251. (59) Sun, S.; Tsubaki, N.; Fujimoto, K. Appl. Catal. A: 2000, 202, 121. (60) Deboer, M.; Koch, E. P. F. M.; Blaauw, R. J.; Stobbe, E. R.; Hoffmann, A. N. J. M.; Boot, L. A.; Vandillen, A. J.; Geus, J. W. Solid State Ionics 1993, 63-65, 736. (61) Halawy, A. A.; Mohamed, M. A.; Bond, G. C. J. Chem. Technol. Biotechnol. 1993, 58, 237. Page 15 of 30

16 (62) Massalski, T. B.; Okamoto, H.; Subramanian, P. R.; Kacprozak, L. Binary Alloy Phase Diagrams, 2nd ed.; American Society for Metals: Materials Park, OH, (63) Hu, M.; Noda, S.; Komiyama, H. J. Appl. Phys. 2003, 93, (64) Hu, M.; Noda, S.; Okubo, T.; Yamaguchi, Y.; Komiyama, H. J. Appl. Phys. 2003, 94-95, (65) Hu, M.; Noda, S.; Tsuji, Y.; Okubo, T.; Yamaguchi, Y.; Komiyama, H. Journal of Vaccum Science and Technology A 2002, 20, 589. (66) Dai, H.; Rinzler, A. G.; Nikolaev, P.; Thess, A.; Colbert, D. T.; Smalley, R. E. Chem. Phys. Lett. 1996, 260, 471. (67) Sinnott, S. B.; Andrews, R.; Qian, D.; Rao, A. M.; Mao, Z.; Dickey, E. C.; F., D. Chem. Phys. Lett. 1999, 315, 25. (68) Cheung, C. L.; Kurtz, A.; Park, H.; Lieber, C. M. J. Phys. Chem. B. 2002, 106, (69) Shibuta, Y.; Maruyama, S. Chem. Phys. Lett. 2003, in press. (70) Pinheiro, J. P.; Schouler, M. C.; Dooryhee, E. Solid State Commun. 2002, 123, 161. (71) Laplaze, D.; Alvarez, L.; Guillard, T.; Badie, J. M.; Flamant, G. Carbon 2002, 40, Page 16 of 30

17 Figure 1. Raman spectra of SWNTs synthesized on quartz substrates measured with 488-nm excitation. Figure 2. Cross-sectional SEM image of as-grown SWNTs on quartz substrates, observed at a fracture edge with a tilted angle.

18 Figure 3. XPS spectra of (a) C 1s, (b) Si 2p, and (c) O 1s levels for calcined and reduced Co-Mo catalysts on quartz substrates.

19 Figure 4. XPS spectra of Mo 3d 5/2 and 3d 3/2 for calcined and reduced Co-Mo catalysts on quartz substrates. Figure 5. XPS spectra of Co 2p 3/2 and 2p 1/2 for calcined and reduced Co-Mo catalysts on quartz substrates.

20 Figure 6. XPS spectra of Co 2p 3/2 levels for calcined and reduced Co-Mo catalysts on quartz substrates.

21 Figure 7. Plan-view TEM images of (a) calcined and (d) reduced Co-Mo catalysts on silicon substrates, and HRTEM images of (b) (c) calcined and (e) (f) reduced Co-Mo catalysts on silicon substrates. Figure 8. Chemical state and morphology of Co-Mo catalysts on quartz and silicon substrates during (a) calcination and (b) reduction.