J. Phys. Chem. C 2009, 113, 18523–18526
18523
Analysis of Fe Catalyst during Carbon Nanotube Synthesis by Mo¨ssbauer Spectroscopy Hisayoshi Oshima,*,† Tomohiro Shimazu,† Milan Siry,† and Ko Mibu‡ Research Laboratories, DENSO Corporation, 500-1 Minamiyama, Komenoki, Nisshin, Aichi 470-0111, Japan, and Graduate School of Engineering, Nagoya Institute of Technology, Showa-ku, Nagoya, Aichi 466-8555, Japan ReceiVed: June 3, 2009; ReVised Manuscript ReceiVed: September 11, 2009
The behavior of Fe catalyst, from its deposition on a SiO2/Si substrate with or without an Al buffered layer to the end of carbon nanotube (CNT) synthesis under hydrogen- and water vapor-added conditions, was investigated by Mo¨ssbauer spectroscopy and scanning electron microscopy. The analyses revealed that the chemical states of the catalyst, not only the morphology, change drastically in the presence of the Al buffered layer. In the case of Fe/SiO2/Si without Al, a quarter of the deposited R-Fe changed to Fe2SiO4 at the heat-up stage, and then not only R-Fe but also Fe2SiO4 appeared to act as CNT synthesis catalysts. In the case of Fe/Al/SiO2/Si, the amount of R-Fe once decreased down to 8 at. % at the heat-up stage and then increased up to 76 at. % during CNT synthesis. The phenomenon in the latter case can be explained by the two following steps: (i) formation of nonmagnetic and magnetic Fe-Al alloys and (ii) segregation of Fe on the surface (where it changes to Fe3C) and at the Al2O3/SiO2 interface (where it stays as R-Fe) accompanied by selective water vapor oxidation of Al in the Fe-Al alloys. The segregated Fe on the surface is thought to act as a CNT synthesis catalyst. Introduction Use of Al or alumina for a buffered layer of metallic catalyst combined with addition of water vapor in chemical vapor deposition (CVD) ambient conditions is an excellent method to obtain vertically aligned long carbon nanotubes (CNTs).1,2 The buffered layer and the water vapor allegedly act as an inhibiter of catalyst agglomeration and an enhancer of CNT synthesis rate2 and as a catalyst activity enhancer and preserver,3,4 respectively. Although a variety of CNTs have been extensively synthesized by this method,5-7 there are few reports investigating interactions between Al and catalysts. Bartsch et al.8 described that, in the case of Fe-Mo/Al/SiO2/Si systems, the Fe-Mo interdiffused into the oxidized Al (AlOx) layer and segregated at the interface between the AlOx layer and the SiO2 layer. When Co was used as a catalyst, segregation of Co at the interface was also observed.5 However, it was not clarified what kind of mechanisms take place for the segregation. The behaviors of catalyst, that is, its chemical states, are frequently investigated by X-ray photoelectron spectroscopy (XPS).9-11 Although XPS is a suitable tool for chemical analysis of catalysts, it is difficult to detect the signal from catalysts when carbon is deposited on them. So applicability of XPS is limited only to the analysis in the stages from catalyst deposition to very early stage of the CNT synthesis. Furthermore, for Fe catalysts, it is difficult to distinguish Fe and iron carbides because of the small difference in the chemical shift of Fe 2p peaks when Fe is bonded with carbon. Compared with XPS, Mo¨ssbauer spectroscopy using Fe K-conversion electrons with 7.3 keV has a potential to measure 100 nm in depth12 and is suitable for an analysis of Fe catalysts.13,14 In this paper, we apply Mo¨ssbauer spectroscopy for an analysis of Fe catalyst behavior, from the catalyst deposition on a SiO2/Si substrate with or without an Al buffered * To whom correspondence should be addressed: e-mail: hoosima@ rlab.denso.co.jp. † DENSO Corporation. ‡ Nagoya Institute of Technology.
layer to the end of the CNT synthesis. The effect of Al from the viewpoint of an interaction between Al and the catalyst is discussed, and a model of Fe catalyst behavior on the Al buffered layer is proposed. Experimental Section Two types of substrates for CNT synthesis were used for this study. One was Fe/SiO2/Si, and the other was Fe/Al/SiO2/Si. A 6-in. silicon wafer was thermally oxidized to form SiO2 with a thickness of 100 nm, and then the wafer was cut down to 14 × 14 × 0.6 mm3 in size. Al and 57Fe (95% enriched) were deposited on the cut wafer by use of an evaporator. The film thicknesses of Al and 57Fe were 15 and 5 nm, respectively. CNT was synthesized on the substrates as follows. A substrate was inserted into a 30 mm i.d. quartz tube. The tube was then heated up to the synthesis temperature (1113 K) under atmospheric pressure of H2 (150 sccm) gas flow, and water vapor was supplied by a water bubbling unit with Ar (50 sccm) gas flow. After 20 min, the temperature reached the synthesis temperature and C2H4 (30 sccm) gas was added for the start of CNT synthesis. The synthesis was terminated by stopping the C2H4 gas flow. The synthesis time was varied from 3 to 180 min. For reference, heat-treated substrates were also prepared in the CNT synthesis tube at 1113 K under atmospheric pressure of flowing Ar (200 sccm) gas for 5 min. 57 Fe Mo¨ssbauer spectra were measured by means of conversion electron Mo¨ssbauer spectroscopy (CEMS) using a gas-flow counter with He + 1% (CH3)3CH at room temperature. R-Fe was used for velocity calibration and also as the standard of isomer shift. A multichannel analyzer with 512 channels was used for data collection, and the typical off-resonance count rate of each channel was 1-10 counts/s. To estimate the composition ratio of Fe sites and species, all spectra were deconvoluted by use of singlet, doublet, and sextet patterns with Lorentzian-shaped curves. The fitted hyperfine parameters are summarized later in Table 3. Surface morphologies were
10.1021/jp905195b CCC: $40.75 2009 American Chemical Society Published on Web 10/07/2009
18524
J. Phys. Chem. C, Vol. 113, No. 43, 2009
Oshima et al. TABLE 1: Evolution of Percentage of Fe-Containing Compounds on Fe/SiO2/Si Substrate from Mo¨ssbauer Spectra after annealing CVD for 3 min CVD for 60 min
R-Fe
Fe3C
Fe2SiO4
unknown phase
γ-Fe
62 37 19
0 49 69
22 3 2
11 8 6
5 3 4
CNT synthesis temperature.17 Therefore, we expect from the Mo¨ssbauer spectroscopy results that Fe deposited on Al/SiO2/ Si reacts with Al and form 59 at. % nonmagnetic and 12 at. % magnetic Fe-Al alloys during the heat-up stage, and the morphology change is caused by melting, aggregation, and reaction of Al. A small amount of Fe3C also exists. Although only Ar gas was flowed during the heat-up stage, the Fe3C may be formed by a reaction of Fe with residual hydrocarbon gases in the reactor tube.18 The Fe of Fe3C would be supplied from the Fe not reacted with Al. Indeed, some amounts of R-Fe remained after the heat-up stage as shown in Figure 1b. In general, Al oxidizes to form AlOx, termed native oxides, in air at room temperature. The native oxides prevent the reaction of Fe with Al partially in our experiments. The composition ratios of Fe species are summarized in Tables 1 and 2. The influence of CNT synthesis on Mo¨ssbauer spectra and surface morphologies were investigated further. Figure 2 shows (a) the cross-section SEM image and (c) the Mo¨ssbauer spectrum of Fe/SiO2/Si without Al after 3 min of CNT synthesis. Figure 2 panels b and d are the corresponding images after 60 min of CNT synthesis. The Mo¨ssbauer peaks in Figure 2c,d mainly consist of R-Fe and Fe3C. As the results of the composition evolution of the catalyst at each stage (Table 1), we conclude that (i) Fe is scarcely oxidized under the employed CNT synthesis condition even though water vapor has been introduced, and (ii) R-Fe and Fe2SiO4 observed in Figure 1c are carbonized rapidly at the early stage of CNT synthesis. According to the vapor-liquid-solid (VLS) model19,20 accepted as the CNT synthesis mechanism, a formation of catalyst with appropriate size is required before the CNT synthesis and then the catalyst reacts with carbon source gas to form CNT in the early stage of CNT synthesis. It is reported that, during the CNT synthesis, the catalyst shows a metal carbide phase.21 Therefore, Fe3C particles with the appropriate size formed from R-Fe and/ or Fe2SiO4 are candidates for CNT synthesis catalysts. Analysis of the unknown nonmagnetic phase in Figure 1c needs further investigation with consideration of the superparamagnetic effect of ultrafine magnetic particles.22 CNT syntheses were performed with Fe/Al/SiO2/Si as well. Figure 3 shows (a) a tilted cross-section SEM image, (b) a surface SEM image, and (c, d) Mo¨ssbauer spectra after the CNT synthesis. CNT synthesis times were 3 min for Figure 3a,c and 180 min for Figure 3b,d. Before the SEM images were obtained, the CNTs were removed from the substrates. Deconvolution of the Mo¨ssbauer spectra was achieved with nine chemical sites of six chemical species: R-Fe, Fe3C, two nonmagnetic sites of Fe-Al alloys, three magnetic sites of Fe-Al alloys, one nonmagnetic site of Fe-O-Al compound, and γ-Fe. These
Figure 1. Surface SEM images and Mo¨ssbauer spectra of (a, c) Fe/ SiO2/Si substrate and (b, d) Fe/Al/SiO2/Si substrate, annealed at 1113 K for 5 min. Red areas (a, b) indicate the high-contrast areas of SEM images obtained in BSE mode. In the Mo¨ssbauer spectra, solid lines indicate spectra fitted by use of Mo¨ssbauer parameters listed in Table 3. Peak positions for individual sites with relative ratio more than 5% are also shown. An unknown nonmagnetic phase and nonmagnetic components are indicated as UNP and NM, respectively. Contributions from magnetic Fe-Al alloys, same as those in Figure 3c, also exist in panel d.
observed by scanning electron microscopy (SEM; Hitachi S-5500). To confirm particles containing heavy Fe atoms, SEM images obtained in the backscattered electron (BSE) observation mode were compared with the images in the same area obtained by the secondary electron (SE) observation mode. The chemical stateofAlwasalsomeasuredbyXPS(ULVAC-PHIESCA5400MC). Results and Discussion Typical SEM images and Mo¨ssbauer spectra of heat-treated Fe/SiO2/Si and Fe/Al/SiO2/Si without CNT synthesis are shown in Figure 1. In the case of the Fe/SiO2/Si, Fe aggregated and formed droplets that have a diameter of several tens of nanometers (Figure 1a). The Mo¨ssbauer spectrum in Figure 1c shows that these droplets mainly consist of R-Fe. Deconvolution of the spectrum revealed that the spectrum consisted of 62 at. % R-Fe and three additional components: γ-Fe (5 at. %), Fe2SiO4 (22 at. %), and an unknown phase (11 at. %) that is attributed to a superparamagnetic phase in ref 15. In contrast, the morphology of Fe/Al/SiO2/Si was quite different (Figure 1b). There are many particles and grains, but those would not consist of Fe judging from the weak contrast of the BSE image, and it is difficult to find R-Fe droplets like those in Figure 1a. This result is correlated with the Mo¨ssbauer spectrum (Figure 1d). The peaks from R-Fe were very weak and peaks from other phases appeared. One singlet and two doublets are related to nonmagnetic Fe alloys. There are also three magnetic Fe components (not indicated by peak position bars in Figure 1d but referred to later in Figure 3c) just inside the peaks of R-Fe needed for the deconvolution.16 Fe and Al can form nonmagnetic Fe-Al alloys such as FeAl3 and magnetic Fe-Al alloys at the
TABLE 2: Evolution of Percentage of Fe-Containing Compounds on Fe/Al/SiO2/Si Substrate from Mo¨ssbauer Spectra Fe-Al alloys after annealing CVD for 3 min CVD for 180 min
Fe-O-Al compound
R-Fe
Fe3C
nonmagnetic
magnetic
nonmagnetic
γ-Fe
8 16 76
18 36 11
59 4 3
12 32 2
2 11 1
1 1 6
Mo¨ssbauer Analysis of Fe Catalyst for CNT Synthesis
J. Phys. Chem. C, Vol. 113, No. 43, 2009 18525 TABLE 3: Mo¨ssbauer Parameters for Deconvolution of Spectraa components R-Fe Fe3C Fe2SiO4 unknown nonmagnetic phase γ-Fe nonmagnetic Fe-Al alloy nonmagnetic Fe-Al alloy magnetic Fe-Al alloy magnetic Fe-Al alloy magnetic Fe-Al alloy nonmagnetic Fe-OAl compound
Figure 2. Cross-section SEM images and Mo¨ssbauer spectra of Fe/ SiO2/Si substrates after CNT growth at 1113 K for (a, c) 3 min and (b, d) 60 min. In the Mo¨ssbauer spectra, solid lines indicate spectra calculated from Mo¨ssbauer parameters listed in Table 3. Peak positions for individual sites with relative ratio more than 5% are also shown.
Figure 3. (a) Tilted cross-section SEM image and (c) Mo¨ssbauer spectrum of Fe/Al/SiO2/Si after CNT growth at 1113 K for 3 min; (b) surface SEM image and (d) Mo¨ssbauer spectrum of Fe/Al/SiO2/Si after CNT growth for 180 min. Red areas (a, b) indicate the high-contrast areas of SEM images obtained in BSE mode. In the Mo¨ssbauer spectra, the solid lines indicate spectra calculated from Mo¨ssbauer parameters listed in Table 3. Peak positions for individual sites with relative ratio more than 5% are also shown. M and NM in panel c indicate magnetic and nonmagnetic, respectively.
Mo¨ssbauer parameters are summarized in Table 3. Here the Al layer was oxidized by water vapor during CNT synthesis, which was also confirmed by XPS. Table 2 shows evolution of the percentage of Fe-containing compounds from annealing to CVD. Surprisingly, the amount of R-Fe, which decreased to 8 at. % during the heat-up process, increased up to 76 at. %, and the amount of Fe-Al alloys and Fe-O-Al compound decreased to less than a few atom percent after 180 min of CVD. The behavior of R-Fe is completely opposite to the results of the Fe/SiO2/Si case. Furthermore, it seems a wonder that R-Fe is stable in the hydrocarbon gas atmosphere at the CVD temperature, because Fe can easily react with carbon at high temperatures. To explain the behavior of the catalyst on Fe/Al/SiO2/Si consistently, we propose a formation model of CNT catalyst. A set of schematic drawings explaining the chemical state of Fe and surface morphologies is shown in Figure 4. First, Al and Fe are deposited on SiO2/Si (Figure 4a). Then the substrate
δ (mm/s) 2εQ (mm/s) ∆EQ (mm/s) Bhf (T) 0.00 0.20 1.15 0.28
0.00 0.01 2.75 1.01
-0.04 0.21 0.24 0.08 0.10 0.15 0.26
33.0 20.5
0.00 0.00 1.36 0.00 0.00 0.00
31.7 30.0 25.1 0.55
a δ, isomer shifts; 2εQ, quadrupole shifts; ∆EQ, quadrupole splitting; Bhf, hyperfine field.
Figure 4. Fe catalyst formation model. (a) Deposition of Fe on Al/ SiO2/Si; (b) alloying and aggregation of Al during heat-up to the CNT growth temperature; (c) segregation of Fe on the surface and at the Al2O3/SiO2 interface by selective oxidation of pure Al and Al in the Fe-Al alloys in hydrogen- and water vapor-added gases for CNT growth; (d) segregated and nonreacting Fe on the surface acting as CNT growth catalyst.
is heated up to a CNT synthesis temperature of 1113 K under Ar atmosphere. At this stage, the Al melts, aggregates, and reacts with Fe to form Fe-Al alloys and a rugged surface (Figure 4b). With the film thickness of deposited Fe and Al and the composition ratio of Fe-Al alloys considered, the melted layer of Fe-Al is thought to consist of pure Al and Fe-Al alloys. Before heating, the Al is partially oxidized by the air and forms native aluminum oxides as mentioned above. The native oxides layer prevents the formation of Fe-Al alloys partially, and Fe remains on them. Then the residual Fe changes to Fe3C at the CVD stage. When hydrogen and water vapor are added to the gas flow, pure Al and Al in the Fe-Al alloys are selectively
18526
J. Phys. Chem. C, Vol. 113, No. 43, 2009
oxidized. This oxidation process would lead to extrusion of Fe from the alloys and formation of pure R-Fe particles on the surface and at the buried interface (Figure 4c). Indeed, these two places with Fe atoms were observed by SEM as in Figure 3a. The segregation phenomenon of catalyst to the buried interface was previously observed.5,8 These buried Fe atoms can be protected from carbonization. This stage corresponds to Figure 1b,d. Fe could be oxidized as well, but FeOx is reduced by the hydrogen on the surface, judging from the fact that there are no significant iron oxide peaks in all the Mo¨ssbauer spectra. Finally, Fe particles on the surface could act as CNT synthesis catalysts, and Fe particles buried at the interface are inactive because of the passivation by the Al2O3 layer (Figure 4d). This stage corresponds to Figure 3b,d. Additionally when the diameter of the residual Fe on the native oxides layer is as small as the diameter of CNT, the residual Fe could also act as a CNT synthesis catalyst. Our model suggests that reduction of the amount of Fe segregated at the buried interface by selection of appropriate Al layer thickness is important to increase the catalyst activity ratio. Conclusions We investigated the behavior of Fe catalyst on SiO2/Si substrate with or without an Al buffered layer from Fe deposition to CNT synthesis, using Mo¨ssbauer spectroscopy and SEM observation. The results revealed that in the case of Fe/SiO2/ Si, Fe2SiO4 was formed just before CNT synthesis and both R-Fe and Fe2SiO4 would be partially carbonized and act as CNT synthesis catalysts. In contrast, in the case of Fe/Al/SiO2/Si, (i) the majority of Fe once reacts with Al to form Fe-Al alloys at the heat-up stage and then (ii) segregates to the surface and the Al2O3/SiO2 interface through selective oxidation of Al in the Fe-Al alloys by water vapor. The segregated Fe on the surface and/or residual Fe on the native Al2O3 layer are thought to act as CNT synthesis catalysts. Acknowledgment. We thank K. Okeyui of Research Laboratories, DENSO Corporation, for substrate preparation and CNT synthesis. Measurement of Mo¨ssbauer spectra was supported by the Nanotechnology Network Project of MEXT, Japan.
Oshima et al. References and Notes (1) Hata, K.; Futaba, D. N.; Mizuno, K.; Nanami, T.; Yumura, M.; Iijima, S. Science 2004, 306, 1362. (2) Noda, S.; Hasegawa, K.; Sugime, H.; Kakehi, K.; Zhang, Z.; Maruyama, S.; Yamaguchi, Y. Jpn. J. Appl. Phys. (Exp. Lett.) 2007, 46, L399. (3) Futaba, D. N.; Hata, K.; Yamada, T.; Mizuno, K.; Yumura, M.; Iijima, S. Phys. ReV. Lett. 2005, 95, 056104. (4) Yoshihara, N.; Ago, H.; Tsuji, M. J. Phys. Chem. C 2007, 111, 11577. (5) Ohno, H.; Takagi, D.; Yamada, K.; Chiashi, S.; Tokura, A.; Homma, Y. Jpn. J. Appl. Phys. 2008, 47, 1956. (6) Yun, Y.; Shanov, V.; Tu, Y.; Schulz, M. J.; Yarmolenko, S.; Neralla, S.; Sankar, J.; Subramaniam, S. Nano Lett. 2006, 6, 689. (7) Li, Q.; Zhang, X.; DePaula, R. F.; Zheng, L.; Zhao, Y.; Stan, L.; Holesinger, T. G.; Arendt, P. N.; Peterson, D. E.; Zhu, Y. T. AdV. Mater. 2006, 18, 3160. (8) Bartsch, K.; Arnold, B.; Kaltofen, R.; Taschner, C.; Thomas, J.; Leonhardt, A. Carbon 2007, 45, 543. (9) Liang, Q.; Gao, L. Z.; Li, Q.; Tang, S. H.; Liu, B. C.; Yu, Z. L. Carbon 2001, 39, 897. (10) Arcos, T.; Garnier, M.; Seo, J.; Oelhafen, P.; Thommen, V.; Mathys, D. J. Phys. Chem. B 2004, 108, 7728. (11) Okita, A.; Ozeki, A.; Suda, Y.; Nakaura, J.; Oda, A.; Bhattacharyya, K.; Sugawara, H.; Sakai, Y. Jpn. J. Appl. Phys. 2006, 45, 8323. (12) Tanuma, S.; Powell, C. J.; Penn, D. R. J. Appl. Phys. 2008, 103, 063707. (13) Coquay, P.; Vandenberghe, R. E.; De Grave, E.; Fonseca, A.; Piedigrosso, P.; Nagy, B. J. Appl. Phys. 2002, 92, 1286. (14) De Resende, V. G.; De Grave, E.; Peigney, A.; Laurent, C. J. Phys. Chem. C 2008, 112, 5756. (15) Pe’rez-Cabero, M.; Taboada, J. B.; Guerrero-Ruiz, A.; Overweg, A. R.; Rodriguez-Ramos, I. Phys. Chem. Chem. Phys. 2006, 8, 1230. (16) Stearns, B. M. J. Appl. Phys. 1964, 35, 1095. (17) Greenwood, N. N., Gibb, T. C. Mo¨ssbauer Spectroscopy; Chapman and Hall: New York, 1971. (18) Liu, K.; Liu, P.; Jiang, K.; Fan, S. Carbon 2007, 45, 2379. (19) Kukovitsky, E. F.; L’vov, S. G.; Sainov, N. A. Chem. Phys. Lett. 2000, 317, 65. (20) Chhowalla, M.; Teo, K. B. K.; Ducati, C.; Rupesinghe, N. L.; Amaratunga, G. A. J.; Ferrari, A. C.; Roy, D.; Robertson, J.; Milne, W. I. J. Appl. Phys. 2001, 90, 5308. (21) Peigney, A.; Coquay, P.; Flahaut, E.; Vandenberghe, R. E.; De Grave, E.; Laurent, C. J. Phys. Chem. B 2001, 105, 9699. (22) Gangopadhyay, S.; Hadjipanayis, C. G.; Dale, B.; Sorensen, M. C.; Klabunde, J. K.; Papaefthymiou, V.; Kostikas, A. Phys. ReV. B 1992, 45, 9778.
JP905195B