CO Disproportionation on a Nanosized Iron Cluster - American

Jul 1, 2009 - Giorgio Lanzani,† Albert G. Nasibulin,‡ Kari Laasonen,*,† and Esko I. Kauppinen*,‡. Department of Chemistry, UniVersity of Oulu,...
0 downloads 0 Views 3MB Size
12939

2009, 113, 12939–12942 Published on Web 07/01/2009

CO Disproportionation on a Nanosized Iron Cluster Giorgio Lanzani,† Albert G. Nasibulin,‡ Kari Laasonen,*,† and Esko I. Kauppinen*,‡ Department of Chemistry, UniVersity of Oulu, P.O. Box 3000, FIN-90014 Finland, and Department of Applied Physics and Center for New Materials, Helsinki UniVersity of Technology, P.O. Box 5100, FIN-02150 Espoo, Finland ReceiVed: May 5, 2009; ReVised Manuscript ReceiVed: June 16, 2009

First-principles electronic structure calculations, fully incorporating the effects of spin polarization and noncollinear magnetic moments, have been used to investigate CO disproportionation on an isolated Fe cluster. After CO dissociation, which occurs on a vertex between the facets, O atoms remain on the surface while C atoms move into the cluster as the initial step toward carbide formation. The lowest CO dissociation barrier found (0.77 eV) is lower than that on most of the studied Fe surfaces. Several possible paths for CO2 formation were identified. The lowest reaction barrier was 1.08 eV. The carbon nanotubes (CNTs) are of great interest since they exhibit unique and useful chemical and physical properties related to toughness, electrical/thermal conductivity, and magnetism.1 The chemical vapor deposition (CVD) methods are widely used as a CNTs synthesis method since they open a way to highly controlled and continuous CNT production.2-5 In these processes, metal nanoparticles are produced in a mixed flow of carbon precursors and other gases (e.g., hydrogen), and the growth process is driven by cleaving the carbon atoms from the precursors, and these atoms will form carbon structures on the nanoparticles’ surface. All properties, like diameter and chirality, of the nanotube are determined by the metal particle. In addition to the CNT synthesis, the metal nanoclusters with a size of less than 10 nm have attracted a great deal of attention due to their applications in magnetism,6 electronics,7 and catalysts.8 The metal nanoparticles are widely used in several real-world catalytic applications, including the car exhaust catalysts, where reactions happen on 3-8 nm size Pt group metal particles. Often, the good catalytic activity can be related to catalytic sites, like atomic size steps, on the cluster. Due to the high curvature of the clusters, the special site density and distribution is much higher than that on almost flat surfaces. Furthermore, the real nanoclusters have several unique active sites like facets and vertexes between the facets, which can have catalytic properties that differ drastically from the ones of almost flat surfaces. The research related to the active sites is mainly limited to atomic steps, and the nanosized clusters have received much less attention.9 Experimentally, many investigators have studied metal nanostructures using a wide range of surface science techniques, but these studies were done with rather a arbitrary size of clusters because it is very difficult to prepare fixed size clusters.9 If we want to understand the active sites on clusters, we have to know which cluster we are studying. Even then, each cluster has several different active sites, and it is * To whom correspondence should be addressed. E-mail: Kari.Laasonen@ oulu.fi (K.L); [email protected] (E.I.K.). † University of Oulu. ‡ Helsinki University of Technology.

10.1021/jp904200e CCC: $40.75

very difficult to know which of them is the most active one. For this reason, the computational approach, where precise sites can be studied, is very attractive. The present study is addressing the CO disproportionation CO(g) + CO(g) h CO2(g) + C(s) on an iron nanocluster during the CVD method for the synthesis of single-walled carbon nanotubes (SWCNTs).2,10,11 The chemistry on the surface of SWCNT catalyst transition-metal nanoparticles is largely unknown, but it is believed that the adsorbed CO first dissociates, CO(s) f C(s) + O(s), and the O will react with an undissociated CO(s) to form surface CO2(s) (Figure 1).12,13 In order to understand the role of the cluster in this reaction, we have used first-principle calculations to study the steps of this reaction on a 55 atom nanocluster. There are only a few ab initio studies of nanosized clusters,14,15 but these works do not address any chemical reactions. The computational work is combined with experimental investigations of the same reaction on larger nanoclusters. A gas-phase process of SWCNTs formation, based on thermal decomposition of ferrocene in the presence of carbon monoxide (CO), was investigated in ambient pressure laminar flow reactors in the temperature range of 600-1300 °C.10 In situ sampling carried out at 1000 °C showed that the SWCNT’s growth occurred from individual metal particles in the heating section of the furnace in the temperature range of 891-928 °C, in which the growth rate was estimated to exceed 2 µm/s.16 Kinetic investigations of the CO disproportionation reaction were performed in a horizontal quartz tube at a heating rate of 5 °C/ min from room temperature. A silica substrate with deposited 15 nm sized iron particles was placed inside of the tube. Investigations show appreciable reaction rates in the temperature interval from 470 to 820 °C, with a maximum rate at about 625 °C. The region of the CO2 concentration increase from about 325 to about 600 °C is the kinetic region where the rate of the CO disproportionation reaction can be measured. By plotting the kinetic region in the coordinates of ln XCO2 versus 1/T, one gets an Arrhenius dependence, XCO2 ) k0 exp(-Ea/RT), where XCO2 is the carbon dioxide mole fraction, k0 ) 7.16091, and Ea  2009 American Chemical Society

12940

J. Phys. Chem. C, Vol. 113, No. 30, 2009

Letters

Figure 1. Reaction mechanism proposed for the CO disproportionation (Fe, gray; C, green; and O, red).

) 63031 J/mol ) 0.65 eV is the energy barrier for the CO disproportionation reaction. For electronic structure calculations, we have used density functional theory (DFT) calculations with the PW91’s generalized gradient approximation (GGA-PW91) for the exchange and correlation functional.17,18 The calculations has been performed with the Vienna ab initio simulation package (VASP19) code with the projector augmented wave (PAW) method.20,21 The adequacy of the functional form, pseudopotentials, and plane waves basis set has been tested against the calculation of the cluster geometry. Iron has a large magnetic moment, and in the cluster geometry, the spin orientation is subtle. Therefore, we have fully incorporated the effects of spin polarization and noncollinear magnetic moments in our calculations.22 To our knowledge, this is the first time these effects are included when molecules or reactivity on nanoclusters is considered. Tests on the magnetic and structural properties of iron clusters up to Fe55 have been performed using ECutoff up to 270 eV and a cubic super cell up to 30 Å. Good results were obtained with a supercell size of 21 Å, ECutoff ) 250 eV, and these settings are also used for the rest of the calculations. This cluster is the best candidate for our study because it is the largest one that we have been able to simulate obtaining a good representation of the magnetic and geometric properties.22 The cluster has an icosahedral geometry and has a formation energy of -3.87 eV/ atom. The largest Fe-Fe distance was 9.68 Å, the dipole moment was (-0.05; -0.03; 0.04) el Å, and the magnetic moment was (2.33; 0.48; 1.06) µB/atom. On the basis of CO adsorption and C/O coadsorption data, CO dissociation paths from the most stable CO adsorption geometries (on F and D sites, Figure 2) to the best C/O coadsorption structures were mapped out with the CI-NEB method.23 According to this algorithm, a set of 3N-dimensional images of the adsorbate system is generated between the end point configurations. A harmonic interaction between adjacent images is added to ensure continuity of the path, thus mimicking an elastic band. An optimization of the band, involving the minimization of the forces acting on the images, brings the band to the minimum-energy path. Several minimum-energy paths (MEPs) were calculated, and the obtained barriers vary between 0.77 and 3.27 eV. Here, we focus to the two lowest barrier paths, COF f C/OCO-2 (barrier 0.99 eV) and COF f C/OCO-6 (barrier 0.77 eV), also shown in Figure 3. The path COF f C/OCO-6 is thermodynamically the most favored one of the studied paths. This reaction is exothermic by 0.69 eV. In the transition state, the C-O bond is elongated to 1.25 Å, with the O atom close to a bridge site (Figure 3.a). Although the path COF f C/OCO-2 has the largest ∆E, the COF f C/OCO-6 (Figure 3b) path is kinetically the most favored one. This reaction is 0.16 eV exothermic. The CO first moves to the edge of the face, reorients itself, and then goes through a transition state where CO tilts on the vertex of the cluster. If the predicted MEPs for CO dissociation are compared to the calculations on other surfaces, it is possible to notice that the paths are similar, but the presence of the vertex lowers the barrier of the dissociation reaction.

These results agree rather well with the previous computational study; the effect of lowering of the barrier due to the presence of a surface step has been previously observed. Using the PW91 functional, Borthwick et al. found the maximum energy along the dissociation pathway of CO on Fe(211) at a bond length of 1.83 Å with a energy of 0.78 eV.24 It is remarkable that the use of the same functional in similar reaction environments provides analogous results. Sorescu et al. calculate a barrier of 1.06 eV per molecule for dissociation of CO from a tilted state on Fe(100),25 while Bromfield et al. calculate barriers of 1.11 and 1.18 eV at 0.25 and 0.5 ML coverages on the same surface.26 The activation barrier of 0.77 eV obtained during this calculation is significantly lower than most of the reported barriers for different surfaces, suggesting that the Fe55 is much more reactive toward the CO dissociation than most of the stepped and flat Fe surfaces. It seems that the ability of the metal atoms to activate reactants changes substantially as the coordination of them is reduced on both the atomic steps and the vertexes. The obtained results show that the catalytically most active sites are the vertexes where the molecules also have a relatively low coordination number. In order to study the formation of CO2 from CO and its fragment adsorbed on the cluster, the CO/O/C and CO2/C coadsorption geometries have been studied. Taking CO_C_O_1 as the starting point, possible paths to CO2_C_1 can be easily conceived. The corresponding structures of all transition states and intermediates are depicted in Figure 4a together with the energy profile. The obtained formation energy is -0.37 eV with a barrier of 1.33 eV. This path presents an elementary rearrangement of atoms and molecules on the surface. A more complex reaction path was found by taking CO_C_O_2 as the starting point and CO2_C_2 as the final point of the path. This path has a formation energy of +0.88 eV and barrier of 1.08 eV. We can also compare the calculated results with our experimental results reported in the initial part of this letter. The DFT results above show that CO disproportionation on a nanosized iron cluster is a multistep process in which the ratelimiting step is the formation of CO2 from O and CO. We studied different reaction channels, and the lowest-energy channel showed a barrier of 1.08 eV. This is consistent with the kinetic investigations that suggest a barrier of 0.65 eV for the overall process. The agreement is reasonable considering the difference in the size of the clusters and the unknown coverage in the experiments. The computed nanoparticle is very small and regular and, thus, it has only a few types of active sites. In this sense, it is probably rather unreactive as a nanocluster. The larger clusters relevant in the experiments will have much more active sites, and some of them can have lower activation barriers. Also, our calculations are at 0 K; therefore, the Fe atom diffusion is not taken into account, whereas the experiments are at around 900 K where the Fe atoms are very mobile. The iron mobility can also cause new active sites that are very difficult to model with the static calculations. Even

Letters

J. Phys. Chem. C, Vol. 113, No. 30, 2009 12941

Figure 2. High-symmetry adsorption sites on a triangular face of the icosahedral Fe55; A, on the plane hcp site; B, on the plane bridge site; C, on the edge bridge site; D, on the vertex top site; E, on the plane hcp site; and F, on the edge top site.

Figure 4. Calculated minimum-energy paths for CO2 formation on Fe55 starting from CO and coadsorbed atomic C/O. Several intermediate structures along the path are shown (Fe, blue; C, green; and O, red).

Figure 3. Calculated minimum-energy paths for CO dissociation on Fe55. Several intermediate structures along the path are shown (Fe, blue; C, green; and O, red).

the difference between the experimental and simulated cluster is rather large, and we wanted to emphasize that the calculations have given new information of the reactivity of nanoclusters

and that the comparisons to real-world systems is more relevant at lower temperatures. In summary, we have presented an ab initio study of the CO adsorption, dissociation, and CO2 formation on a small iron nanoparticle. To our knowledge, this is the first computational study of a chemical reaction on a real nanosized iron particle. Even more importantly, we have studied active sites that have not been studied before, the facets and vertexes. Overall, we see the small clusters as new and important model systems for nanocatalyst studies. They exhabit new active sites that complement the stepped surface calculations. When the computational capacity will permit, larger clusters can also be studied. Presently, detailed chemical reactions can be studied at clusters with a size of 50-100 atoms. Another novel aspect of our study is the use of the spin polarization with noncollinear magnetic moments. In our calculations, the Fe55 cluster has high catalytic activity. We have observed one important effect, the increased reactivity of low-coordinated molecules on the vertexes. Such sites are particularly abundant on the small nanosized particles, and they very likely explain the catalytic activity of them.9 We have been able to show that the reaction barriers for both the CO dissociation and CO2 formation are relatively low, and an exothermic path for surface bound elemental carbon has been found. The carbon atom formation is the initial step for the CNT formation. Under the CVD conditions, the 1.5-2.5 nm size clusters are the CNT catalysts. These clusters probably have a

12942

J. Phys. Chem. C, Vol. 113, No. 30, 2009

similar geometry to what we have studied here, and we believe that also in this case, the vertexes are the catalytically active sites. In further studies, we will focus on carbon-carbon and nitrogen-carbon reactions on the nanoparticle since we are interested in the CNT formation. We need to include other chemicals, like H2, in our study. Overall, the CNT formation chemistry on the iron particle is very challenging. It is not very difficult to model the full CNT growth process with ab initio methods, but we can model elementary reactions between small C fragments (at least up to approximately C12) with C, O, and H. Such studies will give a lot of details of the chemistry on iron nanoparticles. Acknowledgment. The authors would like to express their sincere thanks to Dr. A. S. Foster and Dr. A. V. Krasheninnikov for useful discussions. The proposed project has mainly used the existing computational facilities available at CSC (Center for Scientific Computing), Espoo. This work has been supported, in part, by the European Commission under the 6 Framework Programme (STREP project BNC Tubes, Contract Number NMP4-CT-2006-03350) and the Academy of Finland (Project Number 128445). Supporting Information Available: Additional methods and results. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Saito, R.; Dresselhaus, G.; Dresselhaus, M. S. In Physical properties of carbon nanotubes; Imperial College Press: London, 1998. (2) Nikolaev, P.; Bronikowski, M. J.; Bradley, R. K.; Rohmund, F.; Colbert, D. T.; Smith, K. A.; Smalley, R. E. Chem. Phys. Lett. 1999, 313, 91. (3) Harutyunyan, A.; Pradhan, B.; Kim, U.; Chen, G.; Eklund, P. Nano Lett. 2002, 2, 525. (4) Hata, K.; Futaba, D. N.; Minuzo, K.; Namai, T.; Yumura, M.; Ijima, S. Science 2004, 306, 1362.

Letters (5) Moisala, A.; Nasibulin, A. G.; Brown, D. P.; Jiang, H.; Khriachtchev, L.; Kauppinen, E. I. Chem. Eng. Sci. 2006, 61, 4393. (6) Ziolo, R. F.; Giannelis, E. P.; Weinstein, B. A.; O’Horo, M. P.; Ganguly, B. N.; Mehrotra, V.; Russell, M. W.; Huffman, D. R. Science 1992, 257, 219. (7) Simon, U.; Scho¨n, G.; Schmid, G. Angew. Chem., Int. Ed. Engl. 1993, 32 (2), 250. (8) Yu, W.; Liu, H.; Tao, Q. Chem. Commun. 1996, 15, 1773. (9) Campbell, C. T. Science 2004, 306, 234. (10) Nasibulin, A. G.; Queipo, P.; Shandakov, S. D.; Brown, D. P.; Jiang, H.; Pikhitsa, P. V.; Tolochko, O. V.; Kauppinen, E. I. J. Nanosci. Nanotechnol. 2006, 6 (5), 1233. (11) Nasibulin, A. G.; Pikhitsa, P. V.; Jiang, H.; Brown, D. P.; Krasheninnikov, A. V.; Anisimov, A. S.; Queipo, P.; Moisala, A.; Gonzalez, D.; Lientschnig, G.; Hassanien, A.; Shandakov, S. D.; Lolli, G.; Resasco, D. E.; Choi, M.; Toma´nek, D.; Kauppinen, E. I. Nat. Nanotechnol. 2007, 2 (3), 156. (12) Kandalam, A. K.; Chatterjee, B.; Khanna, S. N.; Rao, B. K.; Jena, P.; Reddy, B. V. Surf. Sci. 2007, 601, 4873. (13) Li, P.; Miser, D. E.; Rabiei; Yadav, R. T.; Hajaligo, M. R. Appl. Catal., B 2003, 43, 151. (14) Raty, J. Y.; Gygi, F.; Galli, G. Phys. ReV. Lett. 2005, 95, 096103. (15) Feng, D.; Bolton, K.; Rose´n, A. Comput. Mater. Sci. 2006, 35, 243. (16) Nasibulin, A. G.; Moisala, A.; Brown, D. P.; Jiang, H.; Kauppinen, E. I. Chem. Phys. Lett. 2005, 402, 227. (17) Perdew, J. P.; Chevary, J. A.; Vosko, S. H.; Jackson, K. A.; Pederson, M. R.; Singh, D. J.; Fiolhais, C. Phys. ReV. B 1992, 46, 6671. (18) Perdew, J. P.; Chevary, J. A.; Vosko, S. H.; Jackson, K. A.; Pederson, M. R.; Singh, D. J.; Fiolhais, C. Phys. ReV. B 1993, 48, 4978. (19) Kresse, G.; Hafner, J. J. Phys.: Condens. Matter 1994, 6, 8245. (20) Blochl, P. E. Phys. ReV. B 1994, 50 (N24), 17953. (21) Kresse, G.; Joubert, J. Phys. ReV. B 1999, 59, 1758. (22) Kohler, C.; Seifert, G.; Frauenheim, T. Comput. Mater. Sci. 2006, 35, 297. (23) Henkelman, G.; Uberuaga, B. P.; Jonsson, H. J. Chem. Phys. 2000, 113, 9901. (24) Borthwick, D.; Fiorin, V.; Jenkins, S. J.; King, D. A. Surf. Sci. 2008, 602 (13), 2325. (25) Sorescu, D. C.; Thompson, D. L.; Hurley, M. M.; Chabalowski, C. F. Phys. ReV. B 2002, 66, 035416. (26) Bromfield, T. C.; Curulla Ferre, D.; Niemantsverdriet, J. W. ChemPhysChem 2005, 6 (2), 254.

JP904200E