NANO LETTERS
Individual Fe-Co Alloy Nanoparticles on Carbon Nanotubes: Structural and Catalytic Properties
2008 Vol. 8, No. 9 2738-2743
Jian Zhang, Jens-Oliver Mu¨ller, Weiqing Zheng, Di Wang, Dangsheng Su,* and Robert Schlo¨gl Department of Inorganic Chemistry, Fritz-Haber-Institute of the Max Planck Society, Faradayweg 4-6, D-14195 Berlin, Germany Received April 28, 2008; Revised Manuscript Received July 10, 2008
ABSTRACT We report the synthesis, characterization, and catalytic performance of Fe-Co alloy nanoparticles inside the tubular channel of carbon nanotubes. The homogeneous distributions of Fe and Co in the isolated nanoparticles were evidenced confidentially by bulk and surface structural and compositional characterizations, that is, scanning electron microscopy, high-resolution transmission electron microscope in combination with elemental mapping by energy dispersive X-ray spectroscopy and electron energy-loss spectroscopy, powder X-ray diffraction, and X-ray photoelectron spectroscopy. We also demonstrate for the first time an unusual synergism in alloy catalysis. The alloy nanoparticles with widely varying Co/Fe ratio are kept as active as Co for the H2 production from NH3 decomposition. The stability of Co was significantly improved by alloying with Fe. We expect our experimental method to be a general approach to elucidate the synergism phenomenon in alloy catalysis.
Carbon nanotube (CNT)-supported Fe and/or Co nanoparticles have attracted intensive interests in several important interdisciplinary research fields. First, CNTs have a hollow tubular channel and extremely high mechanistic strength of carbon walls to encapsulate magnetic nanowires or nanorods up to micrometers in length. Their potential applications include drug delivery, magnetic data storage, toners and inks for xerography, magnetic resonance imaging, and so forth.1-3 The graphitic wall as a protective shell is expected to prevent the sintering of nanoparticles during the shaping process and to avoid the dipolar relaxation between two neighboring magnetic centers, which are major drawbacks of traditional spinel nanoparticles.3 Furthermore, as a novel carbon material, CNTs have been employed as supports for active metals in heterogeneous catalysis. Compared with the traditional oxides supports, CNTs have high thermal and electronic conductivities, good resistance to acidic/basic chemicals at high temperature, controllable porosity, and tunable surface properties.4 We focus our attention on the catalytic applications of the CNT-supported Fe-Co bimetallic nanoparticles. Alloys are proved to have superior catalytic properties with respect to the reactivity as compared with the pure metals.5 Mixtures of Fe and Co are of particular interest in a number of reactions, including the Fischer-Tropsch reaction,6 carbon nanotube growth,7 water gas shift reaction,8 NH3 synthesis,9 * Corresponding author. E-mail:
[email protected]. Phone: + 49 30 8413 5406. Fax: + 49 30 8413 4401. 10.1021/nl8011984 CCC: $40.75 Published on Web 08/02/2008
2008 American Chemical Society
and so forth. The probe reaction chosen in the present study is catalytic decomposition of NH3, which is a feasible route to provide COx-free hydrogen for fuel cell applications.10,11 However, it is still challenging to explore a highly active and economical decomposition catalyst to meet the strict requirements of the “green” energy system based on hydrogen. Synergic improvement on activity is expected for NH3 synthesis by alloying two active metals (Fe, Co, Ni, Mo, etc.).12 Because of the similarity in limiting factors of ratedetermining steps, that is, dissociative adsorption of N2 molecule and combinative desorption of surface N* atoms for synthesis and decomposition, respectively, bimetallic or alloy catalysts have been thought as the main-stream trend to optimal solution of the decomposition catalysts. However, previous experimental attempts have been restricted in using the metal oxides as the support.13 During thermal pretreatment or reaction at high temperature, the oxide support can partially transform the metallic alloy into the less reducible ternary oxide with spinel structure (e.g., Ni-Al, Co-Al, Fe-Al, and Co-Si). The observed variance in activity probably arises from the change in reducibility and thus active sites over the surface. Furthermore, in sufficient researches, the enhanced performance is often ascribed to the special electronic and/or structural effects of alloying process. However, it is not convincing on the basis of such experimental data to conclude this since there is always a lack of direct evaluations of the alloying status in individual
Scheme 1. Scheme of Synthesis Route to CNTs-Supported FeCo Alloy Nanoparticles
nanoparticles. An attempt is thus necessary to clarify the nature of synergism in alloy nanoparticles with a highly alloyed surface in a real catalyst. Herein, we investigated the NH3 decomposition reaction over Fe-Co bimetallic nanoparticles and employed CNTs to support metallic phases and to provide an efficient electron transfer during the reaction. Either nitrides or carbides that could result from less well-defined carbon surfaces were absent from all exposed active surfaces. It is for the first time possible to identify individual particles with the alloying state for this reaction. Substituting Co by the much cheaper Fe was found to keep the high activity of Co and, most importantly, improve the stability to a great extent. This is obviously different from the classical researches using oxides as supports, in which the improved activities having a volcano curve are usually reported.13,14 CNTs encapsulated Fe-Co alloy nanoparticles were produced on the basis of the capillary phenomenon in the channel of CNTs, as depicted in Scheme 1. Commercial CNTs with a big inner diameter of 20∼50 nm were used as support. First, a mixed solution of Fe and Co nitrates was dropped on the functionalized CNTs powder under continuous stirring, during which metal precursors are distributed on both outer and inner surfaces. Then, less amounts of solvent were dropped over the sample in the same way, and most of the residual metal ions adsorbed on the external walls were pulled inside of CNTs channel by the capillary force under the assistance of ultrasonic treatment. Finally, the obtained sample was dried at room temperature overnight in a fume hood until the moisture disappeared. After calcination and H2 reduction, most of the metal nanoparticles were deposited on the inner walls of the CNTs. Tuning the Nano Lett., Vol. 8, No. 9, 2008
size of the nanotubes could achieve an improved spatial confinement. As the inner diameter of CNTs decreased to some extent, metal particles that were tightly contacted with the concave inner walls would tend to aggregate into onedimensional nanowires or nanorods, providing a practical strategy to synthesize a coaxial carbon-FeCo nanocomposite with an unusual magnetic behavior. As compared with the reported techniques,2 our route is simple, reproducible, and free of template or expensive organic precursors. Elemental maps obtained by the energy filtered transmission electron microscopy (EFTEM) of fresh Fe-Co samples were presented in Figure 1a-c. As can be clearly deduced from the mapping, every metal particle contained both Fe and Co atoms with homogeneous distributions in most particles, suggesting a high alloying extent throughout the nanoparticles. The used samples after the NH3 decomposition reaction also have such a high-extent mixing feature. As shown in Figure S1, Co-K and Fe-K profiles as a function of distance along the line scans revealed the homogeneous distributions of Co and Fe in each individual nanoparticle. Figure 1d showed the Fe-L23 ionization edge at 708 eV and the Co-L23 ionization edge at 779 eV. With reduction of the Co/Fe ratio from 2.0 to 0.2, the Co signal decreased while the Fe signal increased. A significant signal from oxygen (O-K ionization edge at 532 eV) was also recorded by electron energy-loss spectroscopy (EELS), which mainly arose from the slight oxidation due to air exposure. Powder X-ray diffraction (XRD) data of the fresh Fe-Co samples were shown in Figure 2, in which no signal corresponding to single metallic or metal oxide phases were observed. All characteristic peaks could be ascribed to graphite and metallic Fe-Co alloy phases. As the Co/Fe ratio decreased, the peak of alloy gradually shifted from 44.7 to 45.1°. Deconvolution of each peak revealed the coexistence of several typical alloy phases with different Fe/Co ratio, for example, Fe3Co7, FeCo, and Fe7Co3, regardless of the nominal composition, which is essentially related with the intrinsic properties of Fe and Co metals to form continuous solid-solution series. The metallic-oxide core-shell structure could be clearly identified by the high-resolution TEM image and oxygen map of representative alloy particles in the used samples (Figure 3a). The lattice distances of core and shell were 0.20 and 0.25 nm, respectively, which could be attributed to the characteristic of the Fe-Co alloy (110) and CoFe2O4 (311) planes. This evidence agreed well with the measurements done with EFTEM and EELS. As shown in the insert, the thickness of oxide shell was close to 10 atomic layers, being thus too thin to be detected by the XRD technique. The existence of such oxide shell were always observed and independent of the location of particles, for example, inside or outside of CNTs. Statistics of 150 particles for each sample revealed a similar average particle size of 13 ( 2 nm, which is far away from the critical zone of particle size effect, that is, ∼2 nm, for NH3 decomposition over metals.15 Therefore, the slight difference in the size of the present “big” nanoparticles will not affect the specific activity to a large extent. After reaction, no change in the alloying extent 2739
Figure 1. Characterization results of CNTs-supported Fe-Co catalysts. EELS elemental mappings of fresh samples: (a) 5% Co2Fe/CNTs, (b) 5% CoFe/CNTs, and (c) 5% CoFe5/CNTs; (d) EELS spectra of fresh samples.
Figure 2. XRD spectra of fresh CNTs-supported Fe-Co catalysts.
between Fe and Co could be found (Figure S2). Different from the pure Fe catalyst, in which the iron nitride phase was formed as active sites during reaction,11 no nitrogen signal could be identified in the Fe-Co samples, indicating the active phase is the metallic alloy. Catalytic reaction happens over the gas-solid interfaces, and thus surface analysis is required to deeply understand how alloy nanoparticles work. The chemical state and relative abundance of Co and/or Fe were determined by an X-ray photon spectroscopy (XPS). As shown in Figure 3b, the signal corresponding to Co can be resolved into two components, that is, the Co3+ ions in an octahedral environ2740
ment (EB ) 780.0-780.4 eV) and the Co2+ ions in tetrahedral sites (EB ) 782.3-782.6 eV). The XPS spectra of Fe 2p3/2 displayed two components, that is, Fe3+ (Fe2O3, 710.6-710.8 eV) and Fe2+ (FeO, 708.8 eV).16 With increasing Co/Fe ratio, the part of Co shifted from lower oxidation state to higher oxidation state (ICo3+/ICo2+ increased), while that of Fe shifted from higher oxidation state to lower oxidation state (IFe3+/ IFe2+ decreased). This indicates an electron transfer between Co and Fe and the near-distance interaction between two metal atoms, indicating Fe-Co alloy phase might form on the surface during reaction. The stability of the Fe-Co alloy phase over the surface can be related with their negative segregation energy and positive mixing energy.17 Quantitative analysis of XPS data showed the accordance in Fe/Co ratios between surface and bulk of each sample, which also agreed with the good homogeneity in each individual nanoparticle by the microscopic technique. Note that the derived concentration of surface metals from XPS spectra ranged at 1.3 ( 0.5 wt %, which was less than the bulk concentration, that is, 5 wt %. Signals were contributed from particles (1) at the mouse of tubes, (2) inside CNTs which were parallel to the beam line, and (3) outside CNTs with the amount around 0.2 wt % (see the statistics on the weight ratio of inside particles below). Catalytic NH3 decomposition was conducted according to the experimental procedure in Supporting Information. For each reaction run, the outlet gas mixture only comprised N2, H2, and unreacted NH3, without CH4 that might form from the methanation of CNTs (C + 2H2 f CH4).18 Figure 4a Nano Lett., Vol. 8, No. 9, 2008
Figure 3. Characterization results of CNT-supported Fe-Co catalysts. (a) HRTEM image of the used 5% CoFe/CNTs and oxygen maps and corresponding images of used 5% Co2Fe/CNTs; (b) XPS spectra of used catalysts.
compared the steady-state H2 productivity and the activation energies of all samples. At 550 °C, the conversion over pure Co was 2.7 times that of pure Fe. However, the high activity of Co was kept over Fe-Co alloys with the Co/Fe ratio widely ranging from 0.2 to 2.0. Substituting Co by Fe resulted in an increase in the overall activation energy of NH3 from 79 to 105∼109 kJ mol-1, which is still lower than that over pure Fe, that is, 147 kJ mol-1. The highest barrier of Fe can be related with the high coverage of nitrogen over the reconstructed surface due to the very strong binding of nitrogen species on Fe.19 Since the recombinative desorption of surface nitrogen atoms acts as the rate-limiting step,20 one would expect the lowest activity of pure Fe for NH3 decomposition. Co has one more d band electron than Fe, and thus the formation of a surface Fe-Co bond will transfer the electron from electron-rich Co to Fe. The modified electronic properties of Fe are expected to assist the adsorbed nitrogen atoms to desorb from surface, thus leading to a decrease in the activation barrier. As compared with such a pronounced electron donating from Co to Fe in the near distance, the electronic charge from a domain of one alkali Nano Lett., Vol. 8, No. 9, 2008
metal oxide (e.g., K2O) to one Fe particle seems to be far inefficient, which can well explain the absence of the promotion effect of K on a fused Fe catalyst in previous studies.21 Figure 4b showed the relative activity of each catalyst with the same weight loading of metal along with reaction time. Significant deactivation was clearly seen over Co and Fe monometallic catalysts. Especially for Fe, only 45% of the initial activity remained after reaction for 1000 min. Interestingly, all tested bimetallic catalysts were much more stable than monometallic ones, and the stability increased with increasing fraction of Fe in alloys. Considering the absence of metal carbide or nitride phase, one can attribute the loss in activity mainly to the sintering of nanoparticles. It is therefore concluded that the significantly enhanced stability originated from the suppressing agglomeration and coalescence via atomic and/or crystallite migration by alloy formation. The tendencies on activity and stability agreed well with the morphological analysis. To clarify the confinement effect of CNTs channel on catalysis, we synthesized a CoFe5-out-CNT (73 wt % outside) 2741
Figure 4. Performance of CNT-supported FeCo catalysts for NH3 decomposition. (a) Steady-state activity at 550 °C and activation energy (450∼550 °C, 36000 mL of NH3 g-1 h-1). (b) Catalytic performance (600 °C, 36000 mL of NH3 g-1 h-1).
sample to compare with the CoFe5-in-CNT (96 wt % inside) sample as described above. Location of nanoparticles on CNTs was confirmed by tilting the sample in TEM, as shown in Figure S3. Morphological analyses of fresh CoFe5-outCNT and CoFe5-in-CNT samples did not reveal any observable difference, as evidenced by the similar size distribution in Figure 5a and the prevailing characteristic of CoFe2O4 (311) plane (not shown). CoFe5-out-CNTs approached the
steady state after 400 min and finally displayed a NH3 conversion of 24%, which was less than that over CoFe5in-CNTs, that is, 48% (Figure 5b). Significant sintering of outside particles was observed on CoFe5-out-CNTs, and the mean size increased from 14.8 to 29.6 nm, resulting in a sharp decrease in ratio of exposed metal atoms according to its reciprocal proportion to the mean particle size.22 On the contrary, the mean size in CoFe5-in-CNTs remained almost unchanged, that is, 13.7∼14.7 nm. The calculated metal dispersion of CoFe5-in-CNTs was 6.8%, which was twice that of CoFe5-out-CNTs. Normalization of the conversion to metal dispersion reported that both catalysts displayed a nearly same turnover rate, that is, 5.4 ( 0.1 molecule H2 site-1 s-1. The activation energy, that is, 116 ( 5 kJ/mol, was also not influenced by the location of alloy particles. Therefore, we can expect that the higher conversion over inside particles may originate from their superior thermal stability. We then compared the sintering behaviors of particles inside and outside of CNTs by heating the samples in a TEM chamber. Figure 6a showed the overview images of nanoparticles outside of CNTs at room temperature. Location of the selected particles was confirmed by tilting the CoFe5out-CNT sample from -20° to 25°. There was no observable change as the temperature was increased to 550 °C (Figures 6b-d). However, when the temperature approached 600 °C, outside particles immediately aggregated into bigger ones, which can well explain the inferior performance of CoFe5out-CNTs from the beginning of the reaction. TEM images of the used sample (Figure S4) revealed a typical agglomerate characteristic of some outside particles, evidencing the aggregation of several smaller particles at the reaction temperature. Particles inside CNTs were resistant to sinter even at 650 °C and thus displayed a stable catalytic performance. Most recently, the superior confinement effect of CNTs has been reported for hydrogenation reactions, for example, conversion of CO and H2 to ethanol,23 hydrogenation of cinnamaldehyde,24 and the Fischer-Tropsch reaction.25 For instance, Pan et al. found that Rh-Mn particles inside CNTs
Figure 5. Characterization and activity evaluation of CoFe5-in-CNTs and CoFe5-out-CNTs: (a) particle size distributions and (b) comparison of catalytic performance. Reaction conditions: 50 mg catalyst, space velocity 36 000 mL of NH3 g-1 h-1, 600 °C. 2742
Nano Lett., Vol. 8, No. 9, 2008
Supporting Information Available: Description of experimental methods, line scan EDX spectra, elemental maps of the used catalysts, tilting TEM images, and HRTEM images of the used CoFe5-out-CNTs. This material is available free of charge via Internet at http://pubs.acs.org. References
Figure 6. Characterization of the thermal stability of CoFe5/CNTs 54% out. (a) TEM images during tilting the sample holder from -20° to 25°; (b-f) aggregation of nanoparticles outside of CNT channels while heating sample to 650 °C.
displayed 16 times higher yield to ethanol than those outside CNTs.23 The yield of C5 hydrocarbons over confined Fe catalysts was found to be twice that over the outside Fe.25 However, we did not observe a superior turnover rate over the CNT-confined catalyst. This could be due to the specific features of the applied reaction in the present study. NH3 decomposition approaches via a series of dehydrogenation into H2 and N2. The turnover rate is controlled by the recombinative desorption of surface N* species, which was essentially determined by electronic properties of the active metal. On the contrary, hydrogenation reaction involves a series of chain growths into big molecules on the active sites. Spatial restriction of the reaction inside nanoscale channels might prolong the contact time on the local domain and thus favor the growth into the wanted long-chain hydrocarbons.23 In summary, we report the combinative characterizations and the unusual synergism and confinement in catalysis of Fe-Co alloy nanoparticles located within CNTs. The highextent alloying of Fe and Co was confirmed by elemental analysis, XRD, and XPS, by which the structure-activity correlation can be substantial. In NH3 decomposition as a test reaction, high activity can be achieved on Fe by alloying with a little amount of Co. Another important feature of alloyed nanoparticles is the superior stability. Confinement inside CNTs was found to improve the thermal stability rather than the turnover rate of the nanoparticles. Our work will shed new light on the development of NH3 decomposition technology as a practical route to produce COx-free H2. Acknowledgment. We gratefully acknowledge financial support by the ENERCHEM project of Max Planck Society, and European Associated Laboratory for Catalysis and Surface Science (ELCASS), technical assistance from Massimiliano Comotti (MPI fu¨r Kohlenforschung, Mu¨lheim), Raoul Blume and Edith Kitzelmann, and helpful discussion with Dr. Jean-Philippe Tessonnier at the Fritz-Haber Institute.
Nano Lett., Vol. 8, No. 9, 2008
(1) (a) Bianco, A.; Prato, M. AdV. Mater. 2003, 15, 1765–1768. (b) Majetich, S. A.; Artman, J. O.; McHenry, M. E.; Nuhfer, N. T.; Staley, S. W. Phys. ReV. B 1993, 48, 16845–16848. (2) (a) Korneva, G.; Ye, H.; Gogotsi, Y.; Halverson, D.; Friedman, G.; Bradley, J.-C.; Kornev, K. G. Nano Lett. 2005, 5, 879–884. (b) Elias, A. L.; Rodriguez-Manzo, J. A.; McCartney, M. R.; Golberg, D.; Zamudio, A.; Baltazar, S. E.; Lopez-Urias, F.; Munoz-Sandoval, E.; Gu, L.; Tang, C. C.; Smith, D. J.; Bando, Y.; Terrones, H.; Terrones, M. Nano Lett. 2005, 5, 467–472. (3) Pham-Huu, C.; Keller, N.; Estourne`s, C.; Ehret, G.; Ledoux, M. J. Chem. Commun. 2002, 1882–1883. (4) Serp, P.; Corrias, M.; Kalck, P. Appl. Catal. A: Gen. 2003, 253, 337– 358. (5) Rodriguez, J. A. Surf. Sci. Rep. 1996, 24, 223–287. (6) (a) van Steen, E.; Prinsloo, F. F. Catal. Today 2002, 71, 327–334. (b) Guczi, L.; Stefler, G.; Geszti, O.; Koppa´ny, Z.; Ko´nya, Z.; Molna´r, E´.; Urba´n, M.; Kiricsi, I. J. Catal. 2006, 244, 24–32. (7) Corrias, M.; Kihn, Y.; Kalck, P.; Serp, P. Carbon 2005, 43, 2820– 2823. (8) Pereira, A. L. C.; dos Santos, N. A.; Ferreira, M. L. O.; Albornoz, A.; Rangel, M. C. Stud. Surf. Sci. Catal. 2007, 167, 225–230. (9) Raro´g-Pilecka, W.; Jedynak-Koczuk, A.; Petryk, J.; Mis´kiewicz, E.; Jodzis, S.; Kaszkur, Z.; Kowalczyk, Z. Appl. Catal. A: Gen. 2006, 300, 181–185. (10) (a) Schlo¨gl, R. Angew. Chem., Int. Ed. 2003, 42, 2004–2008. (b) Christensen, C. H.; Johannessen, T.; Sørensen, R. Z.; Nørskov, J. K. Catal. Today 2006, 111, 140–144. (11) Zhang, J.; Comotti, M.; Schu¨th, F.; Schlo¨gl, R.; Su, D. S. Chem. Commun. 2007, 1916–1918. (12) (a) Jacobsen, C. J. H.; Dahl, S.; Clausen, B. S.; Bahn, S.; Logadottir, A.; Norskov, J. K. J. Am. Chem. Soc. 2001, 123, 8404–8405. (b) Boisen, A.; Dahl, S.; Nørskov, J. K.; Christensen, C. H. J. Catal. 2005, 230, 309–312. (13) (a) Guczi, L. Catal. Lett. 1990, 7, 205–212. (b) Takanabe, K.; Nagaok, K.; Nariai, K.; Aika, K. J. Catal. 2005, 232, 268–275. (c) Mirzaei, A. A.; Habibpour, R.; Faizi, M.; Kashi, E. Appl. Catal. A: Gen. 2006, 301, 272–283. (14) Kojima, R.; Aika, K. Appl. Catal. A 2001, 215, 149–160. (15) (a) Hardeveld, V.; Montfoort, V. Surf. Sci. 1966, 4, 396–430. (b) Jacobsen, C. J. H.; Dahl, S.; Hansen, P. L.; To¨rnquvist, E.; Jensen, L.; Topsøe, H.; Prip, D. V.; Møenshaug, P. B.; Chorkendorff, I. J. Mol. Catal. A: Chem. 2000, 163, 19–26. (c) Zhang, J.; Xu, H. Y.; Li, W. Z. Appl. Catal. A: Gen. 2004, 296, 257–267. ´ lvarez-Galva´n, M. C.; Campos-Martin, (16) (a) OrsquoShea, V. A. P.; A J. M.; Mene´ndez, N. N.; Tornero, J. D.; Fierro, J. L. G. Eur. J. Inorg. Chem. 2006, 24, 5057–5068. (b) Zhang, J.; Wang, H.; Dalai, A. K. J. Catal. 2007, 249, 298–308. (17) Larsen, J. H.; Chorkendorff, I. Surf. Sci. Rep. 1999, 35, 163–222. (18) Yin, S. F.; Xu, B. Q.; Wang, S. J.; Ng, C. F.; Au, C. T. Catal. Lett. 2004, 96, 113–116. (19) Mortensen, J. J.; Ganduglia-Pirovano, M. V.; Hansen, L. B.; Hammer, B.; Stoltze, P.; Nørskov, J. K. Surf. Sci. 1999, 422, 8–16. (20) Ertl, G.; Ru¨stig, J. Surf. Sci. 1982, 119, L314–L318. (21) Jedynak, A.; Kowalczyk, Z.; Szmigiel, D.; Raro´g, W.; Zielin´ski, J. Appl. Catal. A: Gen. 2002, 237, 223–225. (22) Bartholomew, C. H.; Pannell, R. B. J. Catal. 1980, 65, 390–401. (23) Pan, X. L.; Fan, Z. L.; Chen, W.; Ding, Y. J.; Luo, H. Y.; Bao, X. H. Nat. Mater. 2007, 6, 507–511. (24) Ma, H. X.; Wang, L. C.; Chen, L. Y.; Dong, C.; Yu, W. C.; Huang, T.; Qian, Y. T. Catal. Commun. 2007, 8, 452–456. (25) Chen, W.; Fan, Z. L.; Chen Pan, X. L.; Bao, X. H. J. Am. Chem. Soc., available on line.
NL8011984
2743
