Density Functional Study of the Adsorption of Propene on Silver Clusters, Agqm (m=1 5; q=0, +1)

Chemistry Publications Chemistry 11-2004 Density Functional Study of the Adsorption of Propene on Silver Clusters, Agqm (m=1 5; q=0, +1) Steeve Chrétien University of California - Santa Barbara Mark S. Gordon Iowa State University, mgordon@iastate.edu Horia Metiu University of California - Santa Barbara Follow this and additional works at: http://lib.dr.iastate.edu/chem_pubs Part of the Chemistry Commons The complete bibliographic information for this item can be found at http://lib.dr.iastate.edu/ chem_pubs/438. For information on how to cite this item, please visit http://lib.dr.iastate.edu/ howtocite.html. This Article is brought to you for free and open access by the Chemistry at Iowa State University Digital Repository. It has been accepted for inclusion in Chemistry Publications by an authorized administrator of Iowa State University Digital Repository. For more information, please contact digirep@iastate.edu.

Density Functional Study of the Adsorption of Propene on Silver Clusters, Agqm (m=1 5; q=0, +1) Abstract Density functional theory has been used to investigate the binding of propene to small Ag clusters in the gas phase. The binding mechanism based on frontier orbital theory, which we used previously to describe the binding between propene and the Au clusters, works for the pure Ag clusters as well. Among other things, it explains the trends of the desorption energy of propene as a function of the Ag cluster size. We show that one can predict the binding site of propene by examining the shape of the lowest unoccupied molecular orbitals (LUMOs) of the bare clusters and correlate the strength of the bond to the orbital energies of the LUMOs of the bare cluster. Keywords Silver, Density functional theory, Adsorption, Binding sites, Desorption Disciplines Chemistry Comments The following article appeared in Journal of Chemical Physics 121 (2004): 9925, and may be found at doi:10.1063/1.1809600. Rights Copyright 2004 American Institute of Physics. This article may be downloaded for personal use only. Any other use requires prior permission of the author and the American Institute of Physics. This article is available at Iowa State University Digital Repository: http://lib.dr.iastate.edu/chem_pubs/438

Density functional study of the adsorption of propene on silver clusters, Ag m q (m=1 5; q=0, +1) Steeve Chrétien, Mark S. Gordon, and Horia Metiu Citation: The Journal of Chemical Physics 121, 9925 (2004); doi: 10.1063/1.1809600 View online: http://dx.doi.org/10.1063/1.1809600 View Table of Contents: http://scitation.aip.org/content/aip/journal/jcp/121/20?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Density functional study of the interaction between small Au clusters, Au n ( n = 1 7 ) and the rutile Ti O 2 surface. I. Adsorption on the stoichiometric surface J. Chem. Phys. 127, 084704 (2007); 10.1063/1.2770462 Characterization of methoxy adsorption on some transition metals: A first principles density functional theory study J. Chem. Phys. 122, 044707 (2005); 10.1063/1.1839552 Density functional study of the adsorption of propene on mixed gold-silver clusters, Au n Ag m : Propensity rules for binding J. Chem. Phys. 121, 9931 (2004); 10.1063/1.1809601 Binding of propene on small gold clusters and on Au(111): Simple rules for binding sites and relative binding energies J. Chem. Phys. 121, 3756 (2004); 10.1063/1.1769366 Kohn Sham density-functional study of the adsorption of acetylene and vinylidene on iron clusters, Fe n / Fe n + (n=1 4) J. Chem. Phys. 119, 12279 (2003); 10.1063/1.1626625

JOURNAL OF CHEMICAL PHYSICS VOLUME 121, NUMBER 20 22 NOVEMBER 2004 Density functional study of the adsorption of propene on silver clusters, Ag q m mä1 5; qä0, 1 Steeve Chrétien Department of Chemistry and Biochemistry, University of California, Santa Barbara, California 93106 Mark S. Gordon Department of Chemistry, Iowa State University, Ames, Iowa 50011 Horia Metiu a) Department of Chemistry and Biochemistry, University of California, Santa Barbara, California 93106 Received 28 July 2004; accepted 2 September 2004 Density functional theory has been used to investigate the binding of propene to small Ag clusters in the gas phase. The binding mechanism based on frontier orbital theory, which we used previously to describe the binding between propene and the Au clusters, works for the pure Ag clusters as well. Among other things, it explains the trends of the desorption energy of propene as a function of the Ag cluster size. We show that one can predict the binding site of propene by examining the shape of the lowest unoccupied molecular orbitals LUMOs of the bare clusters and correlate the strength of the bond to the orbital energies of the LUMOs of the bare cluster. 2004 American Institute of Physics. DOI: 10.1063/1.1809600 I. INTRODUCTION Recent experiments have shown that small Au and Ag clusters 1 16 supported on TiO 2 catalyze the epoxidation of propene with very high selectivity. The present paper is part of a series that uses density functional theory DFT to try to gain some understanding of the mechanisms of these reactions. We are particularly interested in the possibility of using clusters consisting of a few atoms as catalysts. The catalytic properties of such clusters often change quite substantially, upon alloying, upon contact with the support, or upon changing the number of atoms in the cluster. 17 In principle, this makes them more amenable to tuning their properties to optimize the catalytic process. In a previous paper, 18 we have proposed four simple rules, inspired by the frontier orbital theory, 19,20 which can be used to predict the binding site of propene to Au clusters, and to estimate the strength of the bond. There are a lot of papers in the literature where the frontier orbital theory was succefully applied. For example, Wells, Delgass, and Thomson used it to predict the binding site of O 2 to small Au clusters using the highest occupied molecular orbital of the naked cluster. 21 Our goal is not to provide a survey of the literature, but rather to provide very simple guidance rules that can be used to make useful predictions related to the Au catalyst. To explain these rules we use the following nomenclature: we call SOMO, an orbital that is occupied by one electron singly occupied MO and reserve the names LUMO1, LUMO2,... for the lowest lying empty orbitals. Among these, LUMO1 has the lowest energy, the energy of LUMO2 is the next lowest, etc. Rule 1. Propene binds preferentially to the site on the a Electronic mail: metiu@chem.ucsb.edu naked cluster where the LUMO1 of the naked cluster protrudes most in the vacuum i.e., has the highest overlap with the orbital of propene. Rule 2. If a naked Au n cluster of a given shape has several LUMOs that have low energy, they can all contribute to propene binding. Various LUMOs account for various binding sites of propene, hence various isomers of the Au n (C 3 H 6 ) q complex q is the total charge of the complex. The strongest propene Au n bond is the one whose binding site is controlled by the overlap with LUMO1, the next strongest is the one corresponding to sites controlled by the overlap with LUMO2, etc. Rule 3. The lower the energy of LUMO1 of the naked cluster, the higher the desorption energy of propene to the site where that LUMO1 protrudes in space. In other words, the variation of the desorption energy of the most stable isomer of Au n (C 3 H 6 ) q with n, tracks the dependence of the energy of LUMO1 on n. Rule 4. If LUMO1 has protruding lobes at several nonequivalent sites, the propene will make the strongest bond at the site having the lowest coordination. We have also noted that since propene tends to donate electron density to a Au cluster, the bond strength correlates with the ability of the cluster to accept electrons. 18 If these observations can be extended to pure Ag and to mixed Au Ag clusters, we expect the following trends: 1 For a given cluster size, propene should bind most strongly to a positive ion, less strongly to a neutral one and even less to a negative one. 2 Propene should bind more strongly to a Au cluster than to the corresponding Ag one because, for a given cluster size, the electron affinity of the Ag cluster is smaller than that of the Au cluster. 22,23 3 We anticipate that the desorption energies of propene from mixed Au Ag clusters are intermediate between 0021-9606/2004/121(20)/9925/6/$22.00 9925 2004 American Institute of Physics

9926 J. Chem. Phys., Vol. 121, No. 20, 22 November 2004 Chrétien, Gordon, and Metiu the ones of the pure Ag and pure Au clusters because, according to our calculations, the electronic affinities of the mixed clusters are in between the ones of the pure Ag and Au clusters. 4 We expect propene to bind more strongly to a Ag atom in a mixed cluster than on the same atom in a the pure Ag cluster of the same size and shape; since Au is more electronegative than Ag, an electron transfer from Ag to Au atoms should occur in the mixed clusters. 24,25 This creates an electron deficiency on the Ag atoms in the mixed clusters and activates these atoms towards the adsorption of propene. In the present work we use DFT to calculate the desorption energy of propene to neutral and positively charged Ag clusters and to test whether the rules proposed for propene adsorption on Au clusters work for Ag. The adsorption of propene on the neutral mixed Au Ag clusters is discussed in an accompanying paper. It is possible in principle to derive these rules from quantum mechanics, but such an approach would be complex and not necessarily more enlightening than these empirical observations based on the results of DFT calculations. The rules are useful in systematizing and explaining trends in desorption energy. They also allow us to cut down substantially the number of calculations, since we can use the properties of the naked clusters the shape and the energy of their LUMOs to guess reliably which propene-cluster isomers are likely to have the largest desorption energy. r-pbe functional correlates with an upward shift of 0.1 0.2 ev of the LUMO energy of the bare Ag clusters. The desorption energy of propene in the Ag C 3 H 6 )] complex has been measured recently by Manard, Kemper, and Bowers 39 to be 1.65 0.04 ev; the r-pbe calculation gives 1.72 ev, PW91 gives 2.02 ev, and B3LYP gives 1.52 ev. 39 This latter value include a zero-point energy correction of 0.04 ev while the other two do not contain such a correction. The geometries of the bare clusters and the propenecluster complexes have been optimized by using a conjugated gradient algorithm 40 starting from many different structures, without symmetry constraints. We used the methodology proposed in Ref. 18 to reduce the number of initial structures to be optimized for the adsorption of propene on the Ag clusters. As we showed previously, 18 we need to consider only the adsorption of propene on the clusters at the positions where a low-lying LUMO protrudes into the vacuum. Some tests have been performed to ensure that this applies to Ag clusters as well. We have not considered the dissociative adsorption of propene. All of the starting structures were optimized in a 15 Å cubic supercell. Optimized structures were not characterized by vibrational analysis because this procedure is not available in the version of VASP we used. Consequently, we cannot exclude the possibility that some of the low-lying isomers presented in this paper correspond to transition states. The number of unpaired electrons was kept fix during the geometry optimization. We have considered only the singlet state for the molecules containing II. COMPUTATIONAL DETAILS an even number of electrons and the doublet state for A detailed presentation of the method of computation is odd numbers of electrons. Our calculations on the adsorption available in Ref. 18. Briefly, density functional calculations of propene on small Au clusters showed that the equilibrium have been performed with the VASP program version structures with higher number of unpaired electrons are less 4.4.5. 26 29 The potential energy surfaces for all the systems stable by at least 2.0 ev. We have made some tests to ensure discussed in this paper were initially explored using the combination of the exchange and correlation functionals of that this is true as well for the adsorption of propene on Ag clusters. Perdew and Wang 1991 PW91. 30 33 Ultrasoft pseudopotentials with one hydrogen, four carbon, and eleven electrons silver and gold optimized for the PW91 functional were III. RESULTS AND DISCUSSION A. The properties of the naked Ag q m clusters used. 34 Relativistic effects were partially taken into account We do not report our results for the naked neutral and through the use of relativistic scalar pseudopotentials for all charged Ag clusters since they are close to the ones presented atoms. The Brillouin zone was sample at only one k point. in Refs. 41 and 42. The energy cutoff for the plane-wave expansion was 180.7 and 349.4 ev for the bare clusters and the complexes, respectively. The convergence criterion was 10 4 B. The geometries of the Ag for the selfconsistent electronic minimization and for the change of the m C 3 H 6 q mä1 5; qä0, 1 complexes total free energy between two consecutive ionic steps. All the low-energy structures obtained with the PW91 The lowest-energy structures of the Ag m (C 3 H 6 ) and Ag m (C 3 H 6 ) (m 1 5) complexes are shown in Figs. 1 functional were reoptimized again using the revised Perdew and 2, respectively. The shapes of the Ag q m clusters in the Burke Ernzerhof r-pbe functional 35 and the projector complexes Ag m (C 3 H 6 ) q (m 1 5; q 0, 1 are close to augmented-wave pseudopotentials 36 optimized for the PBE those of the bare Ag q m clusters; the adsorption of propene functional. 37 It has been demonstrated, recently, that this slightly changes the geometry of the cluster but not its shape. functional gives better desorption energies 35 than the PW91 functional. The latter seems to systematically overestimate the value of the desorption energies of propene and CO from Au clusters. 18,38 The desorption energies of propene adsorbed There are many similarities between the geometries of the complexes propene makes with the Ag and the Au clusters: 1 The lowest energy structures of the M n (C 3 H 6 ) q complexes are similar, whether M is Ag or Au see Figs. 1 on Ag clusters, Ag q m (m 1 5; q 0, 1, computed with and 2 of the present paper and Figs. 1 and 5 of Ref. 18 ; 2 r-pbe are smaller, by about 0.27 ev, than those calculated with PW91. This decrease of the desorption energies with the when bonded to a single Ag atom located at the periphery of a Ag cluster, the propene molecule rotates almost freely

J. Chem. Phys., Vol. 121, No. 20, 22 November 2004 Adsorption of propene on metal clusters 9927 FIG. 1. Lowest energy isomers of the Ag m (C 3 H 6 ) complexes (m 1 5) calculated with the r-pbe functional. E is the excess energy of an isomer as compared to the energy of the most stable isomer. D e is the desorption energy of the propene from the cluster see text for the definition of the desorption energy. The orbital energies of the LUMO involved in the binding of propene is indicated also. All energies are in ev. FIG. 2. Lowest energy isomers of the Ag m (C 3 H 6 ) complexes (m 1 5) calculated with the r-pbe functional. E is the excess energy of an isomer as compared to the energy of the most stable isomer. D e is the desorption energy of the propene from the cluster see text for the definition of the desorption energy. The orbital energies of the LUMO involved in the binding of propene is indicated also. All energies are in ev. ference between this conformation and the one in structure around that Ag atom, just as it does for Au clusters; 3 the 2I without propene is less than 0.1 ev with the PW91 structure of the propene molecule in complexes with Au or functional. Consequently, we have used the energy of the Ag is very close to its gas phase structure; 4 propene does later structure to calculate the desorption energy of propene not bind to a flat facet of a Ag or Au cluster. from Ag 5 in structures 2J, 2K, 2L, and 2M see Fig. 2. The desorption energies of propene from Ag m (C 3 H 6 ) q (m 1 5; q 0, 1, are given in Figs. 1 and 2 and a comparison between propene desorption energies to Au and Ag C. The desorption energy of propene from the Ag m C 3 H 6 q mä1 5; qä0, 1 complexes clusters neutral and positively charged is made in Table I. The desorption energy of propene from a cluster is calculated with the formula electrons to them, we expect the bond strength to correlate Since propene binds to a Au or a Ag cluster by donating with the tendency of the cluster to accept electrons. Since the D e E Ag q m E C 3 H 6 E Ag m C 3 H 6 q. 1 electron affinities of the Ag clusters are less than those of the Au clusters of the same size and similar shape, we expect Here E Ag q m is the energy of the naked Ag q m cluster obtained by energy minimization starting with the geometry of the Ag cluster having the same shape as in the TABLE I. Comparison of the desorption energies D e in ev of propene Ag m (C 3 H 6 ) q from the Ag and the Au clusters, computed with the r-pbe functional. complex. For example, if the Ag atoms in Ag 3 (C 3 H 6 ) q form a triangle, then E Ag q 3 is the lowest energy triangular Ag q Neutral Cation 3 cluster. No zero-point energy correction is made since the difference between the zero-point en- n Au n Ag n Au n n Ag ergies of the products and that of the reactants, is small 1 0.03 0.38 1.72 3.13 compared to D e. For example, this correction is 0.04 ev for 2 0.24 1.01 1.14 2.18 the Ag C 3 H 6 )] complex. 39 3 0.47 1.27 0.92 1.77 All attempts to optimize an Ag 4 0.52 1.19 0.83 1.53 5 cluster with a trapezoidal shape failed with the r-pbe functional. The energy dif- 5 0.18 0.69 0.74 1.70

9928 J. Chem. Phys., Vol. 121, No. 20, 22 November 2004 Chrétien, Gordon, and Metiu FIG. 3. Desorption energies of propene left column and absolute value of the energy of the LUMO of the naked Ag clusters right column as a function of cluster size. The upper and lower rows correspond, respectively, to the cationic and neutral cluster. All energies are in ev and were calculated with the r-pbe functional. FIG. 4. LUMO orbital density plots i 2 for the two most stable Ag 4 clusters. The orbital energies are in ev, and the plots show an equal density surface of 0.025 e/å 3. propene to bind more strongly to gold. Moreover, propene should bind more strongly to a positive cluster than to the neutral one for a given cluster size. Indeed, this is what the DFT calculations show see Table I. However, it turns out that the electron affinity of the cluster is not an adequate predictor of propene bond strength. The electron affinity of a naked cluster oscillates with the number of electrons in the cluster: it is large for clusters with an odd number of electrons, and small when the number of electrons is even. 22,23 The desorption energy of electron acceptors i.e., oxygen oscillates with the electron affinity of the cluster, 43 48 but that of electron donors propene, 18 CO, 38 propene oxide, and acetone 49 does not. Therefore, we cannot use the electron affinity to predict how the desorption energy varies with the number of atoms in the cluster. There is however a good correlation between the desorption energy of propene from the lowest energy isomer of M n (C 3 H 6 ) q ) and the energy of the LUMO of the naked M q n cluster: the plot of the desorption energy of propene in stant value. We call these, orbital density surfaces or LUMO plot. As an example we show, in Fig. 4, the orbital density of the LUMOs of two Ag 4 clusters the ones having the lowest energies. In our previous study of propene binding to Au clusters 18 we have shown that the most stable bond is formed at the sites where the orbital density of the LUMO sticks out in the vacuum, to ensure a good overlap with the MO of propene. The same is true for all Ag m and Ag m clusters examined here. For example, propene will bind to the cluster shown in the upper panel of Fig. 4 in position 1 or 3 which are equivalent, since the LUMO protrudes in space at those positions. Propene binds to the Ag 4 cluster, shown in the lower pannel, to the position 1 or 3, for a similar reason. In our previous work with Au clusters 18 we have also found that the desorption energy of propene correlates with M n (C 3 H 6 ) q versus n is very similar to the plot of the the energy of the LUMO of the naked cluster. This rule can LUMO energy of M q n versus n see Fig. 3. In making the above corelation we have ignored, for the clusters with odd number of electrons, the SOMO. Had we assumed that propene binding engages the SOMO and that the bond energy correlates with the SOMO energy, we would have predicted that the desorption energy oscillates with the number of electrons in the bare cluster. This would be in disagreement with the DFT results. Our disregard of the SOMO can be justified by the following simple physical be used to predict to which Au q n isomer propene will make the strongest bond. For example, the LUMO energies of the Ag 3 clusters having a triangular and a bent configuration are 3.2 and 2.6 ev, respectively. As we can see from Fig. 1, propene binds more strongly to the isomer with the triangular shape structure 1C than the bent one structure 1D by 0.28 ev, in agreement with the rule. The rules predict other features that have been found by DFT calculations. Propene does not bind to the flat faces of argument. 38 There is some electronic repulsion between the the Ag q m clusters because the LUMO density allong these electrons of propene and the partially occupied SOMO of the Au and Ag clusters. This repulsion is absent when the accepting orbital is empty LUMO. The similarity between propene binding to the Ag and to the Au clusters is due to the similarity between the LUMOs of the naked clusters. To describe qualitatively the properties of an orbital i (x,y,z) we use plots of the surface on which the orbital wave function square, i (x,y,z) 2, takes a con- facets is very small. We have found that it cost very little energy to rotate propene around an axis going through the atom to which the propene is bound. This happens because the plot of the LUMO density is spherically symmetric and rotating the molecule does not change the overlap of its orbital with the cluster s LUMO. Finally, propene does not adsorb perpendicularly to a Ag Ag bond on the edge of the silver clusters, because of the lack of good overlap between

J. Chem. Phys., Vol. 121, No. 20, 22 November 2004 Adsorption of propene on metal clusters 9929 the frontier orbitals the of propene and the LUMO of the cluster of the two fragments. D. Exceptions to the rules The rule works for all Ag clusters studied here with relatively few exceptions. One example of failure is provided by the isomers 1G and 1H in Fig. 1, in which propene binds to a trapezoidal Ag 5 cluster. According to our rules, the binding of propene at the long base of the trapeze 1H in Fig. 1 involves LUMO1, while the binding of propene at the shorter base of the trapeze 1G in Fig. 1 involves the LUMO2 of the naked trapezoidal Ag 5 cluster. Based on the LUMOs energy, the rules suggest that the most stable isomer is obtained when propene binds to the corner of the long base of the trapeze 1H in Fig. 1. The DFT calculations say that this is not the case: the bond to the top edge of the trapeze 1G of Fig. 1 is stronger by 0.03 ev. However, the orbital energies of the two LUMOs 2.6 and 2.4 ev and the desorption energies to the two sites 0.15 and 0.18 ev are very close to each other. These energy differences are rather small and it would be very surprising if our empirical rules had the high accuracy needed to discriminate between these two cases. IV. SUMMARY AND CONCLUSIONS In this paper we have considered the adsorption of propene on Ag clusters, Ag q m (m 1 5; q 0, 1. We have demonstrated that the binding mechanism describing the interaction of propene with the Au clusters 18 can be applied to the adsorption of propene on Ag clusters. We can predict, by examining at the lowest LUMOs of the naked clusters, the site to which propene binds more strongly. Moreover, the lower energy of the LUMO involved in the binding, the stronger the propene molecule bond. This qualitative binding mechanism can be particularly useful to study the adsorption of propene and similar ligands such as CO and ethylene on larger clusters. The possible number of stable isomers for the bare clusters and also the number of adsorption site for the ligand increase with the cluster size. Based on the binding mechanism, one needs only to consider the adsorption of propene and similar ligands at the positions indicated by a low-energy LUMO. In this way, the computational effort involved in such study can be decreased dramatically. The four simple rules work because the energy separation between the LUMO s of the small Au and Ag clusters is large. Unfortunately, the energy separation between the LUMO s decreases as the cluster grow and eventually reaches zero for the bulk surface. In this case, our rules predict that any LUMO near the Fermi level which is localized and protrudes into the vacuum could be involved in the binding of propene. In that situation, we have proposed 18 to look at the local density of states above the Fermi level. 20 Propene binding will take place where such density is localized and protude in the vacuum. Initiative on Nanotechnology DURINT grant is gratefully acknowledged. We are grateful to Greg Mills for helpful discussions. 1 T. Hayashi, K. Tanaka, and M. Haruta, J. Catal. 178, 566 1998. 2 E. E. Stangland, K. B. Stavens, R. P. Andres, and W. N. Delgass, J. Catal. 191, 332 2000. 3 T. A. Nijhuis, B. J. Huizinga, M. Makkee, and J. A. Moulijn, Ind. Eng. Chem. Res. 38, 884 1999. 4 M. Haruta and M. Daté, Appl. Catal., A 222, 427 2001. 5 A. L. de Oliveira, A. Wolf, and F. Schüth, Catal. Lett. 73, 157 2001. 6 M. Haruta, B. S. Uphade, S. Tsubota, and A. Miyamoto, Res. Chem. Intermed. 24, 329 1998. 7 Y. A. Kalvachev, T. Hayashi, S. Tsubota, and M. Haruta, J. Catal. 186, 228 1999. 8 B. S. Uphade, M. Okumura, S. Tsubota, and M. Haruta, Appl. Catal., A 190, 43 2000. 9 C. Qi, T. Akita, M. Okumura, and M. Haruta, Appl. Catal., A 218, 81 2001. 10 B. S. Uphade, Y. Yamada, T. Akita, T. Nakamura, and M. Haruta, Appl. Catal., A 215, 137 2001. 11 B. S. Uphade, T. Akita, T. Nakamura, and M. Haruta, J. Catal. 209, 331 2002. 12 M. Haruta, Cattech 6, 102 2002. 13 A. K. Sinha, S. Seelan, T. Akita, S. Tsubota, and M. Haruta, Appl. Catal., A 240, 243 2003. 14 G. Mul, A. Zwijnenburg, B. van der Linden, M. Makkee, and J. A. Moulijn, J. Catal. 201, 128 2001. 15 A. Zwijnenburg, M. Saleh, M. Makkee, and J. A. Moulijn, Catal. Today 72, 59 2002. 16 R. Wang, X. Guo, X. Wang, and J. Hao, Catal. Lett. 90, 57 2003. 17 A. Sanchez, S. Abbet, U. Heiz, W.-D. Schneider, H. Häkkinen, R. N. Barnett, and U. Landman, J. Phys. Chem. A 103, 9573 1999. 18 S. Chrétien, M. S. Gordon, and H. Metiu, J. Chem. Phys. 121, 3756 2004. 19 K. Fukui, Science 218, 747 1982. 20 R. Hoffmann, Rev. Mod. Phys. 60, 601 1988. 21 D. H. Wells, Jr., W. N. Delgass, and K. T. Thomson, J. Chem. Phys. 117, 10597 2002. 22 K. J. Taylor, C. L. Pettiette-Hall, O. Cheshnovsky, and R. E. Smalley, J. Chem. Phys. 96, 3319 1992. 23 J. Ho, K. M. Ervin, and W. C. Lineberger, J. Chem. Phys. 93, 6987 1990. 24 V. Bonačić-Koutecký, J. Burda, R. Mitrić, M. Ge, G. Zampella, and P. Fantucci, J. Chem. Phys. 117, 3120 2002. 25 H. M. Lee, M. Ge, B. R. Sahu, P. Tarakeshwar, and K. S. Kim, J. Phys. Chem. B 107, 9994 2003. 26 G. Kresse and J. Hafner, Phys. Rev. B 47, 558 1993. 27 G. Kresse and J. Hafner, Phys. Rev. B 49, 14251 1994. 28 G. Kresse and J. Furthmüller, Comput. Mater. Sci. 6, 15 1996. 29 G. Kresse and J. Furthmüller, Phys. Rev. B 54, 11169 1996. 30 J. P. Perdew, J. A. Chevary, S. H. Vosko, K. A. Jackson, M. R. Pederson, D. J. Singh, and C. Fiolhais, Phys. Rev. B 46, 6671 1992. 31 J. P. Perdew, J. A. Chevary, S. H. Vosko, K. A. Jackson, M. R. Pederson, D. J. Singh, and C. Fiolhais, Phys. Rev. B 48, 4978 E 1993. 32 J. P. Perdew, K. Burke, and Y. Wang, Phys. Rev. B 54, 16533 1996. 33 J. P. Perdew, K. Burke, and Y. Wang, Phys. Rev. B 57, 14999 E 1998. 34 A. Rappe, K. M. Rabe, E. Kaxiras, and J. D. Joannopoulos, Phys. Rev. B 41, 1227 1990. 35 B. Hammer, L. B. Hansen, and J. K. Nørskov, Phys. Rev. B 59, 7413 1999. 36 G. Kresse and D. Joubert, Phys. Rev. B 59, 1758 1999. 37 J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77, 3865 1996. 38 X. Wu, L. Senapati, S. K. Nayak, A. Selloni, and M. Hajaligol, J. Chem. Phys. 117, 4010 2002. 39 M. J. Manard, P. R. Kemper, and M. T. Bowers private communication. 40 W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical Recives in Fortran: The Art of Scientific Computing Cambridge ACKNOWLEDGMENTS University Press, Cambridge, 1992. 41 P. Weis, T. Bierweiler, S. Gilb, and M. M. Kappes, Chem. Phys. Lett. 355, Financial support from the Air Force Office of Scientific 355 2002. Research AFOSR through a Defense University Research 42 R. Fournier, J. Chem. Phys. 115, 2165 2001.

9930 J. Chem. Phys., Vol. 121, No. 20, 22 November 2004 Chrétien, Gordon, and Metiu 43 B. E. Salisbury, W. T. Wallace, and R. L. Whetten, Chem. Phys. 262, 131 2000. 44 W. T. Wallace and R. L. Whetten, J. Am. Chem. Soc. 124, 7499 2002. 45 G. Mills, M. S. Gordon, and H. Metiu, Chem. Phys. Lett. 359, 493 2002. 46 G. Mills, M. S. Gordon, and H. Metiu, J. Chem. Phys. 118, 4198 2003. 47 S. A. Varganov, R. M. Olson, M. S. Gordon, and H. Metiu, J. Chem. Phys. 119, 2531 2003. 48 B. Yoon, H. Häkkinen, and U. Landman, J. Phys. Chem. A 107, 4066 2003. 49 S. Chrétien and H. Metiu, J. Chem. Phys. to be published.