11272
J. Phys. Chem. B 2006, 110, 11272-11276
CO Adsorption on Ag(100) and Ag/MgO(100) Changyong Qin, Laura S. Sremaniak, and Jerry L. Whitten* Department of Chemistry, North Carolina State UniVersity, Raleigh, North Carolina 27695 ReceiVed: December 26, 2005; In Final Form: April 7, 2006
Theoretical studies of CO adsorption on a two-layer Ag(100) film and on a two-layer Ag film on a MgO(100) support are reported. Ab initio calculations are carried at the configuration interaction level of theory using embedding methods to treat the metal-adsorbate region and the extended ionic solid. The metal overlayer is considered in two different structures: where Ag-Ag distances are equal to the value in the bulk solid, and for a slightly expanded lattice in which the Ag-Ag distances are equal to the O-O distance on the MgO(100) surface. The calculated adsorption energy of Ag(100) on MgO(100) is 0.58 eV per Ag interfacial atom; the Ag-O distance is 2.28 Å. A small transfer of electrons from MgO to Ag occurs on deposition of the silver overlayer. CO adsorption at an atop Ag site is found to be the most stable for adsorption on the two-layer Ag film and also for adsorption on Ag deposited on the oxide; CO adsorption energies range from 0.12 to 0.19 eV. The CO adsorption energy is reduced for the Ag/MgO system compared to adsorption on the unsupported metal film thereby providing evidence for a direct electronic effect of the oxide support at the metal overlayer surface. Expansion of the Ag-Ag distance in the two-layer system also reduces the adsorption energy.
1. Introduction Nanoparticles or nanofilms deposited on oxide surfaces are of growing importance in heterogeneous catalysis. Modification of the catalytic activity of a metal by the oxide support could have major implications in future catalyst design. For example, although bulk gold is inert, when deposited as thin, small particles on TiO2, its activity in the catalytic oxidation of CO and hydrocarbons is greatly increased.1-3 Studies by Valden et al.1 show that the activity is most pronounced for particle diameters of 3.0-3.5 nm and for thickness of two Au layers, corresponding to a cluster size of about 300 atoms. Two possibilities exist: a perimeter effect in which both the metal and oxide support participate in the reaction or a direct electronic effect induced by the support through the thin metal layer. Accurate theoretical calculations can lead to a better understanding of the role an ionic solid plays in modifying the surface properties of thin metal films. The Ag/MgO(100) model system has received considerable attention due to its appealing property of having only a 3% lattice mismatch between the metal and the oxide (the lattice constant is 4.09 Å for fcc Ag and 4.21 Å for MgO). Threedimensional island and two-dimensional layer-by-layer growth modes of Ag on MgO(100) have been reported experimentally. Both growth modes provide a basis for theoretical studies of silver nanoparticles and nanofilms deposited on MgO(100).4-8 It is found that silver atoms adsorb above the surface O2- anions and the Ag-O distance has been determined experimentally to be 2.5 Å.6-8 Calculations give Ag-O distances between 2.4 and 2.7 Å, showing good agreement with experiments.9-18 However, calculated adsorption energies, as reported in Table 1., show large variations, depending on the methods used and the silver film thickness. The interaction between Ag and MgO is reported to involve only minor charge transfer and orbital mixing. * Corresponding author. Telephone: (919)515-7960. Fax: (919)5152959. E-mail:
[email protected].
TABLE 1: Calculated Adsorption Energies (eV per silver atom) and Ag-O Distances (Å) of Ag(100) Films on MgO(100) method
dAg-O
year
ref
HF CCa HF FLAPW-LDA DFT-LDA DFT-GGA DFT-GGA DFT-PWGGA
Monolayer 0.20 2.64 0.26 2.56 0.30 2.69 0.33 2.61 0.16 2.75 0.15 2.77 0.14 2.60
1996 2001 1993 1999 1999 2001 2004
14 11 9 15 15 12 16
CI (present) DFT-GGA DFT-GGA DFT-PWGGA
Two Layers 0.58 2.28 0.33 2.50 0.67 0.24 2.48
2005 2003 2001 2004
13 12 16
HF DFT-GGA DFT-PWGGA
Three Layers 0.46 2.43 0.73 0.27 2.47
2001 2001 2004
10 12 16
a
Eads
Include correlation correction.
Adsorption of CO on a silver surface is only slightly exothermic and the initial heat of adsorption has been determined experimentally to be only 0.28 eV for CO on Ag(111).19 DFTPW91 calculations found the atop site to be the most stable with a CO adsorption energy of 0.16 eV.20 Although there are numerous studies on separate Ag/Mg(100) and CO/Ag(100) systems, only CO adsorption on metal adatoms and metal monolayers have been reported for the CO/Ag/MgO(100) system.17,18 This system which is simpler than Au on TiO2 could provide insight into the origin of electronic effects accompanying noble metal deposition on oxides. An experimental study by Kim et al. on Ag adsorption on Mg(100) surfaces and on defect surfaces created by ion sputtering is also relevant to the present work.21 In the present paper, we report ab initio configuration
10.1021/jp0575207 CCC: $33.50 © 2006 American Chemical Society Published on Web 05/24/2006
CO Adsorption
J. Phys. Chem. B, Vol. 110, No. 23, 2006 11273
Figure 1. Embedded Ag9/Mg9O9 cluster model for Ag/MgO(100). The large system shows the Ag9/Mg9O9 central cluster plus one-electron Ag atoms and the ionic core nuclei. The inset shows the labeling of nuclei in the Ag9/Mg9O9 central cluster.
interaction (CI) calculations for the CO/Ag/MgO(100) system and compare the results with CO adsorption on Ag(100). 2. Theory and Method The CO/Ag/MgO(100) system is described in the present work using an embedded cluster method that allows an accurate computation of adsorbate-surface interactions. There are two distinct components of the embedding: (1) the localization transformation that defines the electronic subspace treated by configuration interaction, (2) the use of effective core potentials for ions in the oxide that are outside the region of the oxide that is treated accurately.22-25 A brief summary of the method is given below. Calculations are performed by first solving the total electronic Hamiltonian of the adsorbate/surface system using selfconsistent field (SCF) methods. The occupied and unoccupied SCF orbitals are then localized separately with respect to the surface region of the metal and oxide atoms. Next, the localized orbitals with a large degree of localization and corresponding electrons are put into the CI active space. Finally, the CI wave function, Ψ, is generated from the dominant SCF configuration, Ψ0, plus other important configurations by multiple excitations,
Ψ ) Ψ0 +
∑k CkΨk
(1)
All generated configurations, Ψk, are retained if an interaction threshold
|〈Ψk|H|Ψ0〉|2/(Ek - E0) > 1 × 10-6 au
(2)
is satisfied. Contributions of excluded configurations are estimated using second-order perturbation theory. A second aspect of embedding concerns the treatment of the oxide where an all electron Mg9O9 cluster is embedded in a lattice of +2/-2 ions. The ions are represented by core electron densities, FMg and FO, plus Phillips-Kleinmann projectors for Mg 1s-2p and O 1s-2p electrons. This construction ensures that cores are correctly more repulsive than would otherwise be the case if only +2/-2 point charges were used. See ref 25 for details of this type of effective core potential. The model for the Ag/MgO(100) substrate containing a Ag9/ Mg9O9 central cluster is shown in Figure 1. The Mg9O9 portion is embedded in an array of ionic (Mg2+/O2-) core potentials (ICPs) and the Ag9 cluster is extended to a two-layer Ag(100) film by adding 32 one-electron silver atoms (12 for the surface layer, 20 for the second layer). The basis sets for Mg and O are similar to those for Mg2+ and O2- but the orbitals used were obtained by variational energy minimization of the molecule
MgO embedded in a lattice of +2/-2 charges. Silver atoms in Ag9/Mg9O9 are treated with an effective 1s-3d core potential plus valence 4s, 4p, 4d, 5s, and 5p orbitals, while those Ag atoms outside the central cluster are described with a 1s-4d core and only 5s orbitals. Core potential functions are placed on all silver atoms as described in refs 26 and 27. For the carbon and oxygen of CO, the same s,p,d-type Gaussian basis functions as described in refs 28-30 are adopted; d-polarization functions have an exponent of 0.8. Calculations are carried out for two Ag lattices, one with the Ag-Ag distance the same as in bulk Ag and the second for an Ag-Ag distance equal to the O-O distance in MgO(100). To study the effect of the MgO(100) support on CO adsorption on the silver overlayer, the CI active space must cover two interactions: one between the MgO(100) and the silver film; and the other between the CO molecule and the silver film. In the current study, electrons and orbitals determined by localization with respect to the five MgO(100) surface oxygen atoms, nine central silver atoms and the CO molecule are put into the CI active space. Basis set superposition errors, which are found to be quite small, are evaluated for each system. 3. Results and Discussion In the following sections, the calculated results (geometrical parameters, energetics, and electronic properties) are reported. 3.1. Two-layer Silver Film on MgO(100). Previous experimental and computational studies have shown that silver adsorption on MgO(100) prefers the oxygen anion site, and only this site has been considered in the present study. The adsorption energy curve calculated for a two-layer silver film on MgO(100) is shown in Figure 2. For deposition of Ag(100) films on MgO(100), adsorption energies reported in Table 1 range from 0.15 eV for the monolayer to 0.73 eV for a three-layer system for different methods. In the present study, the adsorption energy is calculated to be 0.57 eV per interfacial Ag atom for a two-layer Ag(100)/ MgO(100) system at the SCF level, and 0.58 eV at the CI level. Our Ag adsorption energy is 0.25 eV larger than a DFT-GGA value reported by Purton et al.15 using Perdew-Wang GGA functional, but only 0.09 eV smaller than that by Cho et al.12 using Perdew-Burke-Ernzerhof GGA functional. The calculated equilibrium Ag-O distance is 2.28 Å at both SCF and CI levels, approximately 0.1 Å shorter than the experimental value of 2.43 Å; other theoretical values are in the range (2.5-2.7 Å). The calculated net electron transfer from surface oxygen anions to silver has been reported to be less than 0.1 e at the SCF level.10,11 The present calculations show by Mulliken population analysis that 0.34 e is transferred from MgO(100)
11274 J. Phys. Chem. B, Vol. 110, No. 23, 2006
Qin et al.
Figure 2. Potential energy curve of two-layer Ag(100) adsorption on MgO(100).
TABLE 2: Adsorption Energies, Equilibrium Distances, and Mulliken Charges for CO Adsorption on the Cluster Model of Ag(100) and Ag/MgO(100) surface bulka expandedb Ag/MgO(100)c gas phase a
site
Eads eV
dC-Surf Å
C charge e
O charge e
CO charge e
CO π population
atop bridge hollow atop bridge hollow atop bridge hollow CO
0.19 0.13 0.14 0.17 0.12 0.13 0.12 0.05 0.10
3.44 4.50 4.23 3.50 3.76 4.29 3.29 4.48 3.42
+0.349 +0.306 +0.309 +0.333 +0.315 +0.319 +0.302 +0.283 +0.300 +0.218
-0.226 -0.241 -0.244 -0.228 -0.234 -0.244 -0.219 -0.237 -0.222 -0.218
+0.123 +0.065 +0.065 +0.105 +0.081 +0.075 +0.083 +0.046 +0.078 +0.000
1.923 1.951 1.945 1.925 1.950 1.948 1.931 1.951 1.943 1.952
Bulk, Ag-Ag distance is 2.89 Å. b Expanded, Ag-Ag distance is 2.98 Å. c Ag sits on oxygen anion with a 2.3 Å Ag-O distance.
to the two-layer silver film per oxygen anion. However, the Mulliken value can be misleading because of the partitioning of the overlap charge when diffuse Ag 5s and 5p orbitals are involved. A more reliable analysis based on core level shifts is presented in the next section. 3.2. CO Adsorption on Ag(100). Table 2 gives calculated adsorption energies, equilibrium distances, and Mulliken charges for the adsorption of a single CO molecule on the cluster model of Ag(100). Calculations are performed for two Ag overlayers: (1) one where the Ag-Ag distance is the experimental bulk lattice value, 2.89 Å and (2) one for a slightly expanded AgAg distance equal to the O-O distance in MgO(100), 2.98 Å. For Ag/Mg(100), the expanded two-layer Ag(100) is lattice matched with MgO(100) with Ag above the oxygen anion at an Ag-O distance of 2.3 Å. In all calculations, the C-O bond distance is optimized to be 1.138 Å at the CI level, which is very close to the experimental gas-phase value, 1.132 Å.31 Thus, the change of the CO bond length upon adsorption to Ag(100) is found to be negligible. On Ag(100) with no oxide present, the calculated adsorption energy at the 1-fold atop site is 0.19 eV at the CI level, which is 0.06 and 0.05 eV larger than the values at the 2-fold bridge and 4-fold hollow sites, respectively (see Table 2). A previously calculated value for CO on Ag(111) at the atop site is in the range ∼0.14-0.16 eV obtained using DFT-GGA-PW91 methods.20 However, an adsorption energy of -0.18 eV is obtained using a revised Perdew-Burke-Erzernhof (RPBE) functional, indicative of repulsive interaction of CO on Ag(111).20 An
experimental value obtained using thermal desorption spectroscopy (TDS) was reported to be 0.28 eV on Ag(111) at the 1-fold atop site.19 The present result is in fairly good agreement with that from DFT-GGA-PW91 methods, but is slightly lower than the experimental value on Ag(111). The net electron transfer from CO to Ag(100) is found to be 0.123 e, mainly from the C lone pair orbital to Ag. Adsorption energies of CO on Ag(100) with the slightly expanded Ag-Ag distance are calculated to be 0.17, 0.12, and 0.13 eV at the 1-fold atop, 2-fold bridge, and 4-fold hollow sites, respectively. Thus, the interaction of CO with Ag(100) is slightly weakened by the 3% lattice expansion, but the site preference is not changed. Correspondingly, the net electron transfer from CO to Ag(100) is reduced to 0.105 e. Figure 3 shows that energy levels of the Ag 4d and 5s orbitals increase on lattice expansion, which apparently inhibits slightly the electron transfer from the CO lone pair orbital to Ag, resulting in a slightly smaller CO adsorption energy. We now consider CO adsorption on the two-layer Ag lattice deposited on the oxide, Ag/MgO(100). Comparison with the previous CO/Ag(100) results allows us to probe the electronic effects associated with the oxide support. The calculated CO adsorption energies are 0.12, 0.05, and 0.10 eV at the 1-fold atop, 2-fold bridge, and 4-fold hollow sites, respectively. Thus, it is found that the CO interaction with Ag/Mg(100) is weaker than that with the two-layer unsupported Ag(100) for the same slightly expanded Ag-Ag distance. On Ag(100), CO adsorption at the 2-fold bridge and 4-fold hollow sites showed essentially
CO Adsorption
J. Phys. Chem. B, Vol. 110, No. 23, 2006 11275
Figure 3. Energies of occupied orbitals for a two-layer silver film and for the film on a MgO support. In the Ag/MgO column, levels primarily identifiable as O 2p are shifted to the right. Part a shows an upward shift of Ag levels for the expanded lattice compared to the bulk lattice. Deposition on the oxide leads to a further upward shift of Ag levels and a downward shift of Mg and O eigenvalues. Part b depicts general trends in core and CO level shifts when the CO/Ag system is deposited on the MgO support. All values are for CO adsorbed at an atop Ag site.
no site preference. However, CO adsorption at the 4-fold hollow site is increased on Ag/MgO(100) and becomes favored over the 2-fold bridge site. The 4-fold site is still less favorable than the atop site. Compared to CO/Ag(100), the adsorption energy of CO on Ag/MgO(100) decreases 30%, illustrating a direct electronic effect of the oxide support that extends to the metalvacuum surface. Since CO adsorption energies on silver are small for both supported and unsupported systems, one should not expect to find a dramatic electronic structure feature that completely changes the character of CO bonding to the surface. There are however several consistent trends that correlate with the observed slight decrease in the adsorption energy for the supported system. These trends are identified and discussed below for CO adsorbed atop Ag, the most favorable adsorption site. (1) Electron Transfer from MgO to Ag. Examination of core orbital eigenvalues shows a decrease in energy (stabilization) of both Mg and O 1s orbital eigenvalues when the Ag slab is deposited on the surface. Correspondingly there is an upward shift of the Ag 4s and 4p eigenvalues. We can use these orbital eigenvalue shifts to probe electron transfer. The argument is strictly electrostatic: an upward shift in orbital eigenvalue occurs when electrons are transferred to a spatial region in the vicinity of the orbital. The eigenvalue changes which are on
the order of -0.8 eV for Mg, -1.6 eV for O (of MgO) and +0.3 eV for Ag are indicative of a transfer of electrons from the oxide to the metal. The presence of the oxide thus causes electron transfer to and a charge redistribution in the metal overlayer. The fact that all of the Ag core eigenvalues are affected shows that the effect extends to the silver-vacuum surface. Consistent with these core level shifts, the Mg 2s and 2p also shift downward and the Ag 4s and 4p shift upward. Work function measurements on submonolayer Ag on MgO(100) are consistent with the calculated downward energy shift (stabilization) of MgO levels on deposition of Ag overlayers.21 (2) CO Level Shifts. Compared to CO adsorbed on the twolayer unsupported Ag slab, the C1s, O1s, COσ, and COπ levels undergo an 0.25 eV upward shift (destabilization) for CO adsorbed on MgO-supported Ag. This shift is due to a slight decrease in electron transfer from CO to silver on the MgOsupported surfaces. However, the Cσ lone pair that normally is viewed as donating to the metal surface is strongly mixed with Ag d and surface s,p orbitals, primarily because of its near degeneracy with metal orbitals. (3) Orbital Effects. The usual analysis of CO bonding to transition metals focuses on the Cσ lone pair and COπ orbitals (the Blyholder model). For Ag, however, the d-levels are deeper and less important in back-bonding to the π system. In the present case, when CO adsorbs on MgO-supported Ag, there is
11276 J. Phys. Chem. B, Vol. 110, No. 23, 2006 only a trivially small change in population of the CO σ and π orbital levels, and there is no significant shift in charge between C and O. The most pronounced orbital effect, when CO adsorbs on Ag/MgO, lies in the region between Ag and CO where there is a shift of electrons from the Ag s-type orbitals to Ag p-type orbitals presumably to accommodate the Cσ lone pair. (4) Electron Correlation. We find no evidence from the configuration interaction calculations that electron correlation plays an important role in differentiating between the supported and unsupported systems. This is not surprising since the CO molecule undergoes essentially no overall change in electron density nor are there internal shifts in charge. In summary, except for the electron transfer discussed above and corresponding level shifts shown in Figure 3, we find no other gross feature that differentiates the bonding of CO to Ag vs CO to Ag/MgO. Other effects of the surface, noted above, are more subtle and affect slightly the relative stability of the three sites investigated. As pointed out in the core eigenvalue analysis, it is clear that the effects of the MgO support extend to the metal-vacuum surface layer. For other adsorbates or reactions on the surface, the surface effect may be of greater consequence, but for CO there is only a slight decrease in adsorption energy on the supported system. 4. Conclusions Theoretical studies of CO adsorption on a two-layer Ag(100) film and on a two-layer Ag film on a MgO(100) support are reported. Ab initio calculations are carried at the configuration interaction level of theory using embedding methods to treat the metal-adsorbate region and the extended ionic solid. The metal overlayer is considered in two different structures: where Ag-Ag distances are equal to the value in the bulk solid, and for a slightly expanded lattice in which the Ag-Ag distances are equal to the O-O distance on the MgO(100) surface. The calculated adsorption energy of Ag(100) on MgO(100) is 0.58 eV per Ag interfacial atom; the Ag-O distance is 2.28 Å. A small transfer of electrons from MgO to Ag occurs on deposition of the silver overlayer. CO adsorption at an atop Ag site is found to be the most stable for adsorption on the two-layer Ag film and also for adsorption on Ag deposited. Acknowledgment. The authors gratefully acknowledge the U.S. Department of Energy for financial support of this research work.
Qin et al. References and Notes (1) Valden, M.; Lai, X.; Goodman, D. W. Science 1998, 281, 1647. (2) Haruta, M. Catal. Today 1997, 36, 153. (3) Bamwenda, G. R.; Tsubota, S.; Nakamura, T.; Haruta, M Catal. Lett. 1997, 44, 83. (4) Larsen, J. H.; Ranney, J. T.; Starr, D. E.; Musgrove, J. E.; Campbell, C. T. Phys. ReV. B 2001, 63, 195410. (5) Didier, F.; Jupille, J. Surf. Sci. 1993, 307-309, 587. (6) Robach, O.; Renaud, G.; Barbier, A. Phys. ReV. B 1999, 60, 5858. (7) Flank, A. M.; Delaunay, R.; Lagarde, P.; Pompa, M. Phys. ReV. B 1996, 53, 1737. (8) Gue´nard, P.; Renaud, G.; Villette, B. Physica B 1996, 221, 205. (9) Li, C.; Wu, R.; Freeman, A. J.; Fu, C. L. Phys. ReV. B 1993, 48, 8317. (10) Zhukovskii, Yu. F.; Kotomin, E. A.; Jacobs, P. W. M.; Stoneham, A. M.; Harding, J. H. Surf. Sci. 1999, 441, 373. (11) Zhukovskii, Yu. F.; Kotomin, Fuks, D.; Dorfman, S.; Gordon, A. Surf. Sci. 2001, 482-485, 66. (12) Cho, J.; Kim, K. S.; Chan, C. T.; Zhang, Z. Phys. ReV. B 2001, 63, 113408. (13) Herschend, B.; Hermansson, K.; Alfredsson, M.; Zhukovskii, Y. F.; Kotomin, E. A.; Jacobs, P. W. M. J. Phys. Chem. B 2003, 107, 11893. (14) Heifets, E.; Kotomin, A.; Orlando, R. J. Phys.: Condens. Matter. 1996, 8, 6577. (15) Purton, A.; Bird, D. M.; Parker, S. C.; Bullett, D. W. J. Chem. Phys. 1999, 110, 8090. (16) Ferrari, A. M. Chem. Phys. Lett. 2004, 400, 541. (17) Gronbeck, H.; Broqvist, P. J. J. Phys. Chem. B 2003, 107, 12239. (18) Giordano, L.; Del Vitto, A.; Pacchioni, G.; Ferrari, A. M. Surf. Sci. 2003, 540, 63. (19) McElhiney, G.; Papp, H.; Pritchard, J. Surf. Sci. 1976, 54, 617. (20) Gajdosˇ, M.; Eichler, A.; Hafner, J. J. Phys.: Condes. Matter 2004, 16, 1141. (21) (a) Kim, Y. D.; Stultz, J.; Wei, T, Goodman, D. W. J. Phys. Chem. B 2002, 106, 6827. (b) Wendt, S.; Kim, Y. D.; Goodman, D. W. Prog. Surf. Sci. 2003, 74 (1-8), 141-159. (22) Whitten, J. L. J. Chem. Phys. 1993, 177, 387. (23) Cremaschi, P.; Whitten, J. L. Theor. Chim. Acta 1987, 72, 485. (24) Madhavan, P.; Whitten, J. L. J. Chem. Phys. 1982, 77, 2673. (25) Whitten, J. L.; Yang, H. Surf. Sci. Rep. 1996, 24, 55. (26) Okada, S.; Matsuoka, O. J. Chem. Phys. 1989, 91, 4183. (27) Whitten, J. L.; Yang, H. Int. J. Quantum Chem.: Quantum Chem. Symp. 1995, 29, 41. Whitten, J. L.; Yang, H. In Transition State Modeling for Catalysis; American Chemical Society: Washington, DC, 1998. (28) Yang, H.; Whitten, J. L. J. Chem. Phys. 1988, 89, 5329. (29) Qin, C.; Whitten, J. L. J. Phys. Chem. B 2005, 109, 8852. (30) Whitten, J. L. J. Chem. Phys. 1966, 44, 359. (31) Huber, K. P.; Herzberg, G. Molecular Spectra and Molecular Structure: Constants of Diatomic Molecules; van Nostrand-Reinhold: New York, 1979; Vol. 4.
