Hydrogen Activation on Silver: A Computational ... - ACS Publications

Jan 15, 2008 - Hydrogen Activation on Silver: A Computational Study on Surface and Subsurface Oxygen. Species. Amjad B. Mohammad,† Ilya V. Yudanov,â...
0 downloads 0 Views 291KB Size
1628

J. Phys. Chem. C 2008, 112, 1628-1635

Hydrogen Activation on Silver: A Computational Study on Surface and Subsurface Oxygen Species Amjad B. Mohammad,† Ilya V. Yudanov,†,‡ Kok Hwa Lim,†,§ Konstantin M. Neyman,| and Notker Ro1 sch*,† Department Chemie, Theoretische Chemie, Technische UniVersita¨t Mu¨nchen, 85747 Garching, Germany, BoreskoV Institute of Catalysis, Russian Academy of Sciences, 630090 NoVosibirsk, Russia, School of Chemical and Biomedical Engineering, Nanyang Technological UniVersity, Singapore 639798, and Institucio´ Catalana de Recerca i Estudis AVanc¸ ats (ICREA), 08010 Barcelona and Departament de Quı´mica Fı´sica & Institut de Quı´mica Teo` rica i Computacional, UniVersitat de Barcelona, 08028 Barcelona, Spain ReceiVed: August 14, 2007; In Final Form: NoVember 2, 2007

Clean silver is known to be inert toward H2 dissociation. Nevertheless, silver catalysts recently have been found to exhibit a noteworthy selectivity in the hydrogenation of unsaturated aldehydes to unsaturated alcohols. Experimental studies indicate that pretreatment in oxygen atmosphere activates the catalyst. To examine the role of oxygen in activation of hydrogenation catalysts, we carried out a density functional study on periodic slab models of H2 dissociation at various oxygen species on silver surfaces, including subsurface oxygen. According to our calculations all oxygen forms under scrutiny promote dissociation of molecular hydrogen. With hydrogenation reactions in mind, we discuss a mechanism according to which an oxygen species, before it desorbs as a water molecule, produces one or two active hydrogen atoms on a metal terrace.

1. Introduction Silver, which is widely used to catalyze oxidation reactions,1 has recently been found to exhibit a relatively high selectivity in the hydrogenation of low R,β-unsaturated aldehydes to unsaturated alcohols.2 For instance, acrolein hydrogenation on supported silver catalysts yields up to 42% of allyl alcohol,3-5 significantly more than on a platinum catalyst (5%).6 The regioselectivity of acrolein hydrogenation strongly depends in a complicated way on the preparation and the structure of the catalyst,3,5 which renders clarification of the hydrogenation mechanism difficult. One of the key issues to be unraveled is the activation of hydrogen. Indeed, H2 dissociation on clean silver surfaces is quite unlikely according to a variety of experimental surface science7-11 and theoretical12-16 studies of different Ag systems. In a recent density functional (DF) study,16 we addressed the dissociation of H2 molecules on the experimentally wellcharacterized reconstructed added-row structure p(2 × 1)O/Ag(110). On this model of an oxygen-pre-covered surface, we calculated H2 dissociation at the added-row -Ag-O-Ag-Oto be thermodynamically feasible and kinetically much more favorable than on clean Ag(110). On the latter surface, H2 dissociation was determined to be endothermic, by ∼40 kJ mol-1, in contrast to the added-row model, where H2 dissociation was determined to be strongly exothermic, by about -200 kJ mol-1. Also, the activation barrier of H2 dissociation decreases from ∼120 kJ mol-1 on clean Ag(110) to ∼70 kJ mol-1 on the oxygenated surface.16 Beneficial effects of oxygen pretreatment of Ag/SiO2 catalysts in the acrolein hydrogenation have been recently demonstrated * Corresponding author. E-mail: [email protected]. † Technische Universita ¨ t Mu¨nchen. ‡ Russian Academy of Sciences. § Nanyang Technological University. | ICREA/Universitat de Barcelona.

by Claus et al.17 The overall hydrogenation activity increased two- to threefold after oxidative pretreatment of the catalyst as well as in the presence of a small amount of oxygen. Oxygen pretreatment also increased the selectivity toward the desired product allyl alcohol, to more than 50%.17 Regenerative oxygen pretreatment of a deactivated Ag/SiO2 catalyst not only restored the initial conversion level, but also significantly increased the hydrogenation activity and improved the selectivity.17 Clearly, the added-row reconstructed surface p(2 × 1)O/Ag(110) is a model far too idealized if one aims at representing the surface of a hydrogenation catalyst. Therefore, quite a few questions remained after our previous study of hydrogen activation on oxygenated silver surfaces.16 The added-row reconstruction takes place on Ag(110) at relatively low oxygen coverage;18 the ordered phase with the lowest oxygen coverage known, 0.125 ML, is p(8 × 1).19 However, such a phase is unlikely to exist under hydrogenation conditions. A more general problem is related to the fact that atomic hydrogen is strongly bound to added-row oxygen and therefore is unlikely to participate in hydrogenation reactions. In the present work, we considered the activation of molecular hydrogen on silver surfaces with oxygen species, which do not involve extensive surface reconstruction and can represent even the limiting case of isolated adsorbed oxygen atoms. We studied oxygen species both at the surface and in subsurface positions. We also extended our study of hydrogen activation on the p(2 × 1)O/Ag(110) substrate by considering alternative pathways, including removal of oxygen via formation of water. Furthermore, we addressed another important aspect of the hydrogenation mechanism: transportation of hydrogen, activated on oxygen species, to clean silver regions where hydrogenation of organic reactants possibly occurs. In view of the complexity of the “working” hydrogenation catalysts based on silver,2-5 the models under scrutiny cannot be exhaustive. Thus, the goal of the present work is not to

10.1021/jp0765190 CCC: $40.75 © 2008 American Chemical Society Published on Web 01/15/2008

Hydrogen Activation on Silver

Figure 1. Adsorption sites for oxygen on a Ag(110) surface: short bridge (SB), long bridge (LB), top (T), threefold hollow (3f), fourfold hollow (4f). A rectangular frame traces the projection of a 2 × 2 unit cell onto the surface plane. Only atoms of the two top layers of the five-layer slab are shown; spheres in dark shading represent the second layer.

establish univocally, which sites of such catalysts are responsible under real catalytic conditions for the supply of atomic hydrogen reactants. Rather, this paper aims at examining the propensity for H2 dissociation of potential active sites on silver surfaces, formed because of the presence of oxygen species. 2. Models and Computational Details The calculations were performed with the plane-wave based Vienna ab initio simulation package (VASP)20-22 using the PW91 exchange-correlation functional.23 The interaction between atomic cores and electrons was described by the projector augmented wave (PAW) method.24,25 For integration over the Brillouin zone, we combined Monkhorst-Pack grids26 of (5 × 5 × 1) k points for all unit cells with the first-order MethfesselPaxton smearing technique (smearing value 0.15 eV).27 The final energies were extrapolated to zero smearing. In all cases, we chose an energy cutoff of 400 eV. To model the Ag(110) surface, we employed unit cells with four Ag atoms per layer,28 enabling hydrogen coverage on the surface of 1/4 ML or higher. Figure 1 illustrates the O adsorption sites studied. For the stepped Ag(221) [4(111) × (111)] surface, we used unit cells of 12 Ag atoms per layer. We modeled subsurface oxygen and H adsorption at such sites on the Ag(111) surface, using (2 × 2) with 4 and (3 × 3) surface unit cells with 9 Ag atoms per layer. The latter model was also invoked when studying oxygen diffusion from surface to subsurface sites; subsurface O atoms were located in interstitial octahedral subsurface (oss) sites between the top and the second layer of metal atoms (referred to as Ooss atoms). We did not consider here interstitial tetrahedral subsurface sites as they leave too little room for oxygen; for Ag29 and some other metals,30,31 their occupation is known to be less favorable than that of oss sites. A vacuum spacing of about 1 nm was adopted to separate the periodically repeated slabs. The adsorbed moieties were placed on one side of a five-layer slab for Ag(110) and Ag(221); the Ag(111) surface was represented by a four-layer slab. For these clean surfaces and in the corresponding adsorption studies, we relaxed the atomic positions of the “top” two Ag layers and, if present, of all adsorbates. The remaining three (or two) layers at the “bottom” of a metallic slab were kept fixed at the optimized geometry of the bulk material (Ag-Ag ) 293 pm). Previously, we had justified this type of model for studies of surface reactions.16,31,32 The added-row structure -Ag-O- arises by formal deposition of two Ag and two O atoms per surface unit cell of Ag(110); as in our preceding study,16 we chose p(2 × 1)O/Ag(110)

J. Phys. Chem. C, Vol. 112, No. 5, 2008 1629 models. We modeled the Ag(110) substrate as five-layer slab, fixed at the geometry optimized for the bare Ag(110) slab by relaxing the top two layers. In adsorption and reaction studies, we allowed the positions of the H adsorbates and those of the -Ag-O- added row to relax. The induced dipole moment is taken into account by applying a dipole correction.33 The energy of adsorption, Ead ) Etot(ad_sub) - [Etot(ad) + Etot(sub)], was calculated by subtracting the sum of the total energy Etot(ad) of the adsorbed species in the gas phase and the total energy Etot(sub) of the clean substrate from the total energy Etot(ad_sub) of the slab covered by the adsorbate in the optimized geometry. With this definition, a negative value of Ead implies favorable adsorption, associated with an energy release. Adsorption complexes at metal surface were calculated in a spin-restricted fashion. To obtain reference energies for the gas phase, we placed molecules in a cubic unit with edges of 1 nm34 and, where necessary (atomic H, O), accounted for the open-shell ground states by invoking a spin-polarized approach. We located transition states (TS) with the nudged elastic band method (NEB)35 by identifying first approximate TS configurations for models with a reduced slab thickness. Afterward, we extended the calculations to accurate models as described above. We checked all TS structures with a normal-mode analysis that comprised all degrees of freedom of the adsorption complexes involved in the structure variation, to ensure the existence of a single imaginary frequency. 3. Oxygen Adsorption on Silver Surfaces and Diffusion to Subsurface Sites Oxygen species interacting with silver surfaces attracted a great deal of attention for quite some time, mainly from the point of view of epoxidation studies.36 It is well-established now that various oxygen species can be formed on the surface depending on the experimental conditions.1,37,38 Moreover, oxygen atoms have been reported to diffuse subsurface39-42 and are supposed to reside there even in reductive ambient conditions.17 It is difficult to establish correspondence between catalytically active oxygen species and structures known from surface science experiments, but recently it has been attempted.37 A nucleophilic form of oxygen was correlated with linear structures, similar to added rows on reconstructed p(2 × 1)O/ Ag(110), whereas electrophilic oxygen centers were assigned to O species at threefold hollow sites on Ag(111).37 In our previous computational study,16 we provided detailed structural parameters of the reconstructed p(2 × 1)O/Ag(110) surface. As basis for studying H2 activation on oxygenated Ag surfaces, we first briefly describe various pertinent models of silver surfaces with isolated oxygen. We also discuss calculated activation barriers for the diffusion of atomic oxygen from surface to subsurface sites. 3.1. O/Ag(110). According to calculations for the nonreconstructed Ag(110) surface, atomic O can form three stable adsorption complexes which are quasi-degenerate (Table 1, Figure 1): at long bridges (LB) in [001] direction with an adsorption energy of -351 kJ mol-1, at fourfold hollow (4f) sites in the troughs with an adsorption energy of -353 kJ mol-1, and at threefold hollow sites (3f), located at the sides of the troughs, with an adsorption energy of -359 kJ mol-1. The two quasi-degenerate sites along the troughs, LB and 4f, are separated by a small, almost symmetric barrier, 7 kJ mol-1 from the 4f side, whereas the barrier from the LB site toward the most stable site 3f is hardly 1 kJ mol-1. Vibrational frequency analyses confirmed all three sites as local minima, but revealed

1630 J. Phys. Chem. C, Vol. 112, No. 5, 2008

Mohammad et al.

TABLE 1: Calculated Interatomic Distances O-Ag, Heights h(O) of the Atom O above the Surface (110) Plane (Both in Picometers), and Adsorption Energies Ead (Kilojoules per Mole) of Atomic Oxygen on the Unreconstructed Ag(110) Surface for a Coverage θO ) 1/4 sitea

O-Ag

h(O)

Ead

LBb 3fb 4f SBc Tc

221, 232 215, 222 222 204 193

32 60 54 134 193

-351 -359 -353 -309 -207

a See Figure 1 for the definition of the adsorption sites. b Distances to nearest-neighbor Ag atoms of the first and the second layer, respectively. c Not a local minimum (see text).

adsorption complexes at short-bridge (SB) sites, with an adsorption energy of -309 kJ mol-1, to be saddle points of the potential energy surface. Adsorption complexes with oxygen on top of silver centers (T), with Ead ) -207 kJ mol-1, are characterized by two imaginary vibrational frequencies and thus represent local maxima of the potential energy with respect to lateral motion. In the 3f adsorption complex, the O-Ag distances to the first and second layer Ag centers are 215 and 222 pm, respectively; the corresponding values for the LB complex are 221 and 232 pm, and the distances in 4f complex amount to 222 pm. The oxygen centers are located at heights of 60 pm (3f), 54 pm (4f), and 32 pm (LB) above the top layer plane of the surface (Table 1). Preceding computational studies43,44 reported preferred adsorption on Ag(110) at LB and 4f sites, but neither specified the topology of the potential energy surface nor considered the 3f site. In quantitative agreement with our results, the binding energy of atomic oxygen on the fourfold hollow site had previously been calculated at about -340 kJ mol-1, with oxygen located at 57 pm above the top crystal plane.45 The adsorption energy of atomic O on Ag(110), derived from experimental data,46 is about -416 kJ mol-1. This estimate is for an oxygen coverage of θO ) 0.15 ML.47 In this coverage range, formation of ordered added-row phases (7 × 1) (θO ) 0.143) and (6 × 1) (θO ) 0.166) was observed.46 Early surface extended X-ray absorption fine-structure (SEXAFS) studies48,49 assumed oxygen atoms to be adsorbed on the Ag(110) at longbridge sites at 2 pm above the top crystal plane. However, with the help of scanning tunneling microscopy (STM) results, ordered (n × 1) phases later on were identified as reconstructed, with added rows running along the [001] direction.50,51 Thus, the above experimental adsorption energy estimate46 probably corresponds to oxygen in a more stable added-row structure rather than to adsorption on the unreconstructed Ag(110) surface. Adsorption complexes LB, 4f, and 3f of oxygen at such low coverage that added-row moieties are not yet formed exhibit very similar features (Table 1). Thus, any of them appears to be similarly adequate as representative model of H2 activation at the unreconstructed Ag(110) surface. (2 × 2) surface unit cells as used here have been justified elsewhere52 for studying adsorption of small molecules on metal surfaces. 3.2. Subsurface Oxygen Species. As an alternative to H2 activation via oxygen species on silver surfaces, one may invoke adsorption complexes near subsurface oxygen atoms. Small atoms are known to diffuse in appreciable quantities into lattices of transition metals, affecting bulk properties and causing various macroscopic phenomena, such as corrosion and embrittlement.53,54 When these impurities reside in interstices near the surface, they may directly affect surface properties. Subsurface oxygen centers can be viewed as the initial stage of corrosion and oxide formation.54 At low oxygen exposure,55

TABLE 2: Calculated Binding Energies Ead(fcc) of Atomic O at Threefold Hollow fcc Sites and the Corresponding Interaction Energies E(oss) at Octahedral Subsurface (oss) Sites of Ag(110) (Figure 1), Ag(111) and Ag(221) (Figure 2) Slabs, as well as Energies E(fccfoss) for fcc f oss Oxygen Transfer, and the Corresponding Activation Barriers Eaa surface site Ag(110) Ag(111) Ag(221)

3f fcc fcc 1 fcc 2 fcc 3

Ead(fcc)

E(oss)

E(fccfoss)

-359 -358 -363 -348 -342

-278 -292 -292 -299 -284

81 66 70 49 57

Ea(fccfoss)b 81 (∼0) 80 (14) 70 (∼0) 51 (2) 76 (19)

a All energies in kilojoules per mole. b Values in parentheses are activation barriers for reverse diffusion, from subsurface sites to the surface.

detected oxygen species were assumed to occupy an “underlayer” site;56 when surface coverage increased to 0.05-0.08 ML, a well-ordered p(4 × 4)-O oxide-like phase started to nucleate.56 Occupation of subsurface octahedral and tetrahedral sites as well as substitutional oxygen adsorption were discussed as possible models of subsurface oxygen species; however, no definite assignment of detected species was made.56 On the other hand, in line with results of the present study (see below), surface adsorption of oxygen on silver at low coverage was calculated to be notably more favorable.29 Whereas the adsorption energy of oxygen at the Ag(111) surface decreases considerably with increasing coverage, the binding energy of subsurface oxygen depends only weakly on the concentration.29 In addition, Ooss species were favored over those in tetrahedral subsurface sites over the whole surface coverage range studied, from 0.11 to 1 ML.29 We now consider in more detail various models of low content oxygen in subsurface sites of different Ag surfaces (see Table 2). 3.2.1. On Ag(110). Oxygen underneath an LB site on Ag(110) (Figure 1), was reported to be ∼20 kJ mol-1 more strongly bound than at the LB site; also, a calculated activation barrier of 86 kJ mol-1 for oxygen diffusion from the LB to the subLB position was reported, which corresponds to a barrier of ∼66 kJ mol-1 for the reverse movement.43 We attempted to reproduce these findings but failed;57 according to our calculations, oxygen leaves the sub-LB site for the more stable LB site without activation barrier. We confirmed the complex of oxygen in the interstitial oss site underneath the SB site of Ag(110) (Figure 1) as local minimum of the potential energy surface. However, this complex is 81 kJ mol-1 higher in energy than oxygen adsorbed at the neighboring surface 3f site (Table 2); the reverse process does not encounter a notable activation barrier, but a vibrational analysis identified the subsurface complex as a local minimum. Subsurface incorporation of oxygen induces significant structural stress. Nearest-neighbor distances between Ag centers close to Ooss are elongated to 340-350 pm, from 293 pm. In these models, we relaxed three layers of the Ag(110) slab (one more than normally, see section 2) to determine more precisely the optimized positions of Ag centers close to oss sites. 3.2.2. On Ag(111). On the Ag(111) surface, we obtained a similar picture for oxygen diffusion from the fcc threefold hollow sites to oss sites. At an oxygen coverage of 0.11 ML, adsorption at the surface threefold hollow fcc site is 66 kJ mol-1 more favorable than in the corresponding oss site, E(oss) ) -292 kJ mol-1 (Table 2). The calculated activation barrier for oxygen diffusion fcc f oss is 80 kJ mol-1; therefore, the reverse process is also activated, by 14 kJ mol-1. Our results agree very well with those of similar computational study for the same

Hydrogen Activation on Silver

Figure 2. Threefold hollow fcc sites 1, 2, 3 on a (111) terrace of the Ag(221) surface which were considered as starting points for oxygen diffusion to subsurface oss sites. The frame traces the unit cell in the top layser as used in the calculations. Only the two top layers of the five-layer slab are shown.

oxygen coverage,29 where five-layer slab models had been used, in contrast to four-layer models of the present work. That earlier study also addressed the effect of oxygen surface coverage on the stability of subsurface species on Ag(111); for coverages up to 0.5 ML, surface positions were calculated to be favored over subsurface sites.29 3.2.3. On Ag(221). We also inspected oxygen diffusion to subsurface sites of the surface (221), which exhibits (111) terraces terminated by steps. We considered three types of fcc and the corresponding oss sites (Figure 2), distinguished by their location on the terrace with respect to a step. The unit cell shown in Figure 2, with single oxygen species, corresponds to an oxygen coverage of 1/12. In view of the more open structure of Ag(221) compared with the ideal Ag(111) surface, one may expect more flexibility for incorporating oxygen species in the subsurface region. On first glance, on Ag(221), the energy profiles for oxygen subsurface diffusion fcc f oss are similar to that on Ag(111) (Table 2). This holds in particular for fcc sites 1 near the top of a step, except that the corresponding oss site, a very shallow local minimum, exhibits essentially no barrier for Ooss atoms to be released to surface. The fcc sites 2 and 3 further “inward” on the terrace are notably less bound, up to 20 kJ mol-1, while the associate oss sites show much smaller variations in stability. This notably reduces the associated energy change upon oxygen release to the surface, which has to overcome small activation barriers, up to 19 kJ mol-1 (site 3, Figure 2). As in the case of Ag(110) and Ag(111), occupation of oss positions of Ag(221) by oxygen induces significant structural stress: nearest-neighbor distances among Ag atoms surrounding the oss site reach 330-350 pm. Thus, all Ag surface systems considered in this section show subsurface oxygen species to be unstable (in the absence of adsorbed oxygen). Recall that according to DF and ab initio thermodynamics modeling on oxygen at Ag(111), bulk-dissolved oxygen also was calculated to be energetically disfavored at low oxygen concentrations compared with surface oxygen.29 Yet, oxygen species in subsurface interstitial positions of silver, even if only as metastable complexes, cannot be excluded under (hydrogenation) reaction conditions. 4. H2 Activation at Oxygen Species Previously, we found that dissociation of the H2 molecule can take place on the added rows of the reconstructed p(2 × 1)O/Ag(110) surface.16 However, the resulting two hydroxyl groups are unlikely to act as hydrogen source for reactions with organic molecules on silver catalysts, because of the strong binding of hydrogen in the hydroxyl groups formed on added rows; the corresponding overall reaction energy is -224 kJ mol-1 (Table 3).16 Here, we start by addressing H2 dissociation on single oxygen centers on Ag(110): one atom of a dissociating

J. Phys. Chem. C, Vol. 112, No. 5, 2008 1631 H2 molecule is to interact with an oxygen center, forming a hydroxyl group, and the other H atom is released to a clean area of silver surface, so that subsequently it is able to participate in a hydrogenation process. In initial states (IS) of H2 activation considered in the following, the reactant H2 was placed at 300 pm above the corresponding surface. These IS structures with adsorption energies of a few kilojoules per mole correspond to weak physisorption, similar to the situation on clean Ag surfaces.16 4.1. O/Ag(110). To model H2 dissociation on the unreconstructed O/Ag(110) surface, we chose atomic O centers at LB sites (Figure 1) from a series of almost degenerate adsoprtion sites (section 3.1) because these centers exhibit a local environment that may qualify them as initial state for the formation of an added row. In the TS (Figure 3a), one H atom interacts with an oxygen center (O-H distance 134 pm) and the other H atom is located above a silver ridge, at Ag-H ) 207 pm (Table 3). The barrier for H2 dissociation, 47 kJ mol-1, is much lower than that calculated for clean Ag(110), ∼120 kJ mol-1, and even notably lower than that for the added-row structure p(2 × 1)O/ Ag(110), ∼70 kJ mol-1.16 The H-H distance in the TS, 93 pm (Table 3), is significantly shorter than calculated for the TSs on p(2 × 1)O/Ag(110) (122 pm) and clean Ag(110) (136 pm), indicating that H2 dissociation O/Ag(110) proceeds via an early transition state. The reaction was calculated exothermic, by -99 kJ mol-1. After breaking the H-H bond, the hydrogen center above silver ridge moves to a LB site above a trough (Figure 3a). The potential energy surface of H atom on clean Ag(110) is rather flat with adsorption energy differences between LB, SB, and 3f sites of at most 10 kJ mol-1.16 Therefore, H atom can easily diffuse over the clean Ag(110) terrace to approach a target organic reactant or intermediate to be hydrogenated. We also considered hydroxyl groups, formed on Ag(110) via interaction of adsorbed oxygen with H2 molecules, as possible sites for activating a second H2 molecule. Such a reaction produces H2O and one more H atom, released to a silver terrace and available as subsequent hydrogenation agent (Figure 3b). The position of the OH group differs significantly from that of the original O center. While oxygen adsorbed at an LB position above the trough is almost in the plane of top silver atoms (Figure 3a, left panel), the O center of an OH group in the same position is 110 pm above that plane (Figure 3a, right panel). Moreover, a hydroxyl group at a SB position is slightly more stable, ∼5 kJ mol-1, than a hydroxyl group at an LB site (adsorption on top of an Ag center is 73 kJ mol-1 weaker); therefore, we calculated H2 dissociation for OH at an SB site (Figure 3b). Interestingly, the calculated activation barrier for H2 dissociation is extremely low, 10 kJ mol-1 only. The geometric characteristics of the TS complex are similar to those calculated for the previous case of H2 dissociation at an adsorbed O center (Table 3). H2 activation at an OH group is also exothermic, by -48 kJ mol-1. The resulting H species binds at a SB position of the ridge neighboring the ridge where the reacting OH was located (Figure 3b). As on adsorbed O, the produced H atom can easily diffuse from its SB site to a clean Ag(110) terrace. Similar to other noble metal surfaces,58 for example, Ag(111),59 the H2O molecule formed is weakly bound on Ag(110), about -20 kJ mol-1 at an top site of the ridges and thus can easily desorb. In summary, the mechanism considered here involves an isolated O center adsorbed on Ag(110) surface, which is able to activate two H2 molecules, thereby producing two H atoms adsorbed at oxygen-free Ag(110) terraces and a H2O molecule. The overall reaction O(ads) + 2H2(gas) f H2O(gas) + 2H-

1632 J. Phys. Chem. C, Vol. 112, No. 5, 2008

Mohammad et al.

TABLE 3: Calculated Reaction Energies Er and Corresponding Barriers Ea (kJ mol-1) for H2 Dissociation on Silver Surfaces with Pre-Adsorbed Oxygen where Selected Interatomic Distances (pm) in the Transition States of H2 Dissociation Are Also Shown substrate

final statea

Er

Ea

H-H

O-H

O/Ag(110) OH/Ag(110) p(2 × 1)O/Ag(110) step 1b,c step 2b,c first alternative, step 2c second alternative, step 2c p(2 × 1)O/Ag(110) step 1c step 2c alternative step 2c p(2 × 1)OH/Ag(110) Ag(111)/Osub, θoss ) 0.25 ML Ag(111)/Osub, θoss ) 0.5 ML

OH/ + H/ H2O/ + H/ OH/ar + H/ar 2 OH/ar OH/ar + H/(LB) OH/ar + H/(SB) H2O/ar H2O desorption OH/ar + H/(SB) H2O/ar + H/ 2 H/ 2 H/

-99 -48 32 -256 -118 -120 -150 20 77 -50 15; -13d -80d

47 10 71 12 11 11 137

93 89 122

134 147 109

207 226 189

123

103

268

89 142 109d

144

192 172; 208

156 22 98 71d

Ag-H

a Adsorption complexes are indicated by / (or /ar if formed on the added row of a reconstructed surface). b Reference 16. c Reaction energies Er for each step, i.e., the initial state of the first step comprises H2 while the initial state of the second step is the final state of the first step. d Values from ref 76. In that work, different parameters were used in the models: energy cutoff 340 eV vs our 400 eV. k-grid: 18 special Chadi-Cohen k-points vs our 5 × 5 × 1 Monkhorst-Pack grid.

Figure 4. Alternative directions for transfer of a hydrogen atom from an Ag center; situation after the assumed dissociation of an H2 molecule at an Ag-O pair of the added row on a p(2 × 1)O/Ag(110) surface. See Figure 3 for the gray scale coding.

Figure 3. Dissociation of H2 on different models of oxygenated silver surfaces: (a) O adsorbed at a LB position of nonreconstructed Ag(110); (b) OH group adsorbed at a SB position of Ag(110); (c) addedrow reconstructed p(2 × 1)O/Ag(110) surface; (d) added-row reconstructed p(2 × 1)OH/Ag(110) surface; (e) Ag(111) surface with subsurface oxygen, coverage θoss ) 1/4. Only two top layers (as well as the added row for the reconstructed surfaces) of multilayer slabs are shown. Structures corresponding to initial state (IS), transition state (TS), and final state (FS) are sketched. O centers are represented as black spheres, Ag centers are medium gray spheres, and H centers are light gray spheres.

(ads) is exothermic by -126 kJ mol-1 and both H2 dissociation steps exhibit small activation barriers of 47 and 10 kJ mol-1, respectively. 4.2. p(2 × 1)O/Ag(110). On the added-row reconstructed p(2 × 1)O/Ag(110) surface, we previously had modeled H2 activation, keeping the H-H axis parallel to the -O-Ag- chain.16 The interaction with H2 was found to occur in two steps. First, the H-H bond is broken at a Ag-O pair with an activation

barrier Ea ∼ 70 kJ mol-1 (Table 3), furnishing an OH group and a H atom bound at a Ag center of the added row; this step is endothermic, by ∼30 kJ mol-1. Subsequently, the H atom migrates to another O center of the added row with an activation barrier of only 12 kJ mol-1, forming a second OH group.16 However, as mentioned above, OH groups are unlikely to be highly active in the hydrogenation of organic substances as hydrogen is very strongly bound, by -324 kJ mol-1 (where a H atom in the gas phase is taken as reference).16 For comparison, recall that the strongest bonds of H atoms at 3f, SB, and LB positions of a clean Ag(110) surface are about -200 kJ mol-1.16 Indeed, reversible dehydrogenation of allyl alcohol to acrolein was reported for an oxygen pre-dosed Ag(110) surface;60,61 thus, pre-adsorbed oxygen on a silver surface is able to act as dehydrogenating agent. The large difference in binding energies also makes transfer of H atoms from OH groups at the added row to metal centers of the Ag(110) substrate thermodynamically unfavorable by about 120 kJ mol-1. In the present study, we considered also migration of H atoms, formed as intermediates after the first step of H2 dissociation, from added-row Ag centers16 to the Ag(110) substrate (Figure 4), as an alternative to migration to nearest-neighbor oxygen centers (see above). The activation barriers for these alternative directions are almost the same, 12 kJ mol-1 in [11h0] direction to the O center16 and 11 kJ mol-1 in [100] direction to the neighboring Ag(110) ridge. However, diffusion of H from the added-row Ag site to O is 138 kJ mol-1 more exothermic than movement to the neighboring LB site [p(2 × 1)O/Ag(110), Table 3], whereas diffusion to the SB position at the ridge of the Ag(110) substrate is almost iso-energetic (cf. first and second alternatives of step 2; Table 3 and Figure 4).

Hydrogen Activation on Silver In analogy to the approach taken here for unreconstructed O/Ag(110) (section 4.1), as an alternative to the formation of two OH groups in the added row, we also considered a mechanism where dissociative adsorption of H2 leads to the formation of only one OH group in the added row, releasing the second H atom to an oxygen-free area of the Ag surface. To this end, we oriented the axis of the dissociating H2 molecule perpendicular to the added row, along the [11h0] direction (Figure 3c). However, we found that this H2 dissociation pathway with activation energy of 137 kJ mol-1 leads directly to formation of water molecule; this process is exothermic by -150 kJ mol-1 (Table 3). The water molecule formed is weakly bound on the added row, with an adsorption energy about -20 kJ mol-1, and can easily desorb, similar to the unreconstructed Ag(110) surface (subsection 4.1). Dissociation of H2O to produce an OH group sitting within the added row and a hydrogen atom on the metallic part of the surface is endothermic by 77 kJ mol-1 and exhibits a high activation barrier of 156 kJ mol-1. Note that water formation is about 70 kJ mol-1 less favorable than formation of two OH groups in the added row. Thus, the mechanism releasing one hydrogen atom per dissociated H2 molecule to an oxygen-free part of the catalyst surface is kinetically unfavorable on the reconstructed surface. The most likely result when molecular hydrogen from the gas phase interacts with an addedrow structure is complete hydrogenation of the added row. Above, we discussed the interaction of H2 with an isolated OH group on Ag(110). In the same spirit, we also studied dissociation of H2 via interaction with an OH group, previously formed on an added row. In fact, in experiment, a OH(1 × 2)/ Ag(110) structure has been observed to form by H2O dissociation on p(4 × 1)O/Ag(110).62 An oxygen-covered Ag(110) surface is extremely reactive to water, even below room temperature.62 However, the exact structure of the ordered hydroxyl layer OH(1 × 2) formed in this way [or OH(1×n/2) in the more general case of an initial p(n × 1)O/Ag(110) phase] remains under discussion. Although -OH-Ag-OH- chains, preserving the added-row motif of an oxygen-covered surface, were suggested as a possibility,63 very recently it has been suggested that the added rows dissolve, forming a hydroxyl layer that comprises just OH rows on Ag(110) substrate.62 Thus, in our model studies, we took two types of surface hydroxyl groups into account: (i) OH on the unreconstructed Ag(110) and (ii) hydroxylated added rows. In section 4.1, we had already considered H2 dissociation on OH groups at unreconstructed Ag(110); in the following, we will address a hydroxylated surface with preserved added-row superstructure. For this purpose, we selected the added-row structure p(2 × 1)OH/Ag(110) as model (Figure 3d), which formally can be obtained by full hydrogenation of a p(2 × 1)O/Ag(110) structure. As in the case of isolated OH, we determined a very low activation barrier, 22 kJ mol-1, for the dissociation of H2 via interaction with OH groups of the added row; this reaction was calculated to be exothermic, by -50 kJ mol-1 (Table 3). Thus, a hydrogen atom can be released to a silver terrace via dissociation of H2 on a previously formed OH group of an added row. Finally, we consider recombination of H atoms from the Ag(110) substrate with OH groups of the added row, producing water molecules. The activation barrier of this process was calculated at 85 kJ mol-1, and the reaction energy was calculated at -66 kJ mol-1. [Similar values, 76 kJ mol-1 and -81 kJ mol-1, respectively, were calculated on a Ag(111) surface relative to isolated OH and H(fcc) adsorbates.59] The present result for silver differs qualitatively from what we had calculated

J. Phys. Chem. C, Vol. 112, No. 5, 2008 1633 for Cu(111) where we had determined that reaction as almost thermo-neutral, -6 kJ mol-1.64 Obviously, hydrogenation of OH groups may compete with hydrogenation of organic molecules. Hydrogen activation discussed in this subsection assumes a two-step mechanism: (i) pairwise hydrogenation of oxygen centers of an added-row -Ag-O- of a reconstructed p(n × 1)O/Ag(110) surface and (ii) dissociation of H2 via interaction with an OH group to produce H2O that desorbs and a H atom on an Ag(110) substrate (terrace) that can react further with an organic reactant. Overall, this mechanism of hydrogen production is less efficient than the mechanisms discussed in subsection 4.1 for isolated oxygen species because on p(n × 1)O/Ag(110) only one of three H atoms becomes available for the hydrogenation of an organic molecule. For oxygen-covered Ag(111) surfaces, it seems obvious to postulate a chemistry of H2 activation very similar to that of on added-row model of Ag(110). On Ag(111), O-induced surface reconstruction leads to the ordered phase p(4 × 4)O/Ag(111), extensively studied both experimentally65-67 and theoretically.29,68-70 Very recently the atomistic structure of the p(4 × 4)O/Ag(111) phase has been revisited.70-72 While originally this phase was interpreted as a Ag1.83O structure, formed by hexagonal -Ag-O- rings and related to an epitaxial overlayer of the oxide Ag2O, recent STM studies supplemented by DF modeling suggested the formation of silver terraces bridged by oxygen centers.71,72 In any case, these different interpretations of the p(4 × 4) phase share the local structural motif Ag-OAg where an oxygen center bridges two Ag centers. This motif resembles the models considered in the present study for the added-row p(2 × 1)O/Ag(110) surface. In fact, the oxygen coverage of the latter surface, 1/2, is quite close to the value 3/8 of the oxide-covered surface p(4 × 4)O/Ag(111). Photoelectron spectroscopy suggests an additional argument that supports the generalization of results obtained for the addedrow modified Ag(110) surface to a partly oxygen-covered Ag(111) surface. Indeed, O 1s ionization energies of the p(4 × 4)O/Ag(111) structure and the p(2 × 1)O/Ag(110) structure differ only ∼0.1 eV,65,73 in line with DF estimates.72 4.3. Ooss/Ag(111). As model of H2 activation near subsurface oxygen, we chose the Ag(111) surface with Ooss atoms underneath the top layer of Ag atoms, but one should keep in mind that more complicated structures may occur under real catalytic conditions.74,75 However, detailed structural information from experiment on such species is lacking. Our goal was to estimate the effect of subsurface oxygen (at low concentrations) on the surface chemistry of hydrogen. We started by considering the adsorption of atomic H near an Ooss/Ag(111) complex (Figure 5). The threefold hollow fcc and hcp sites on (111) surfaces of fcc metals usually exhibit very similar propensity for adsorption of H; for instance, in our DF study, H adsorption at fcc sites of Pd(111) was just 6 kJ mol-1 stronger than at hcp sites, while on Cu(111), both adsorption complexes were calculated degenerate.52 For Ag(111), our previous calculations also resulted in identical adsorption energies, -198 kJ mol-1, for atomic H at fcc and hcp sites at a H coverage of 1/4 ML.16 Experimentally, the concentration of subsurface O atoms can be assumed to be very low.5,17 We represented this situation by the model Ooss/Ag(111) with the coverage θoss ) 1/4 ML and a p(2 × 2) unit cell (Figure 5). We distinguished two fcc (F1, F2) and two hcp (H1, H2) threefold hollow sites by their proximity to Ooss. Not surprisingly (Figure 5), the strongest effect of Ooss on the adsorption of H was calculated for F1 sites, located directly above the occupied oss site. Adsorption

1634 J. Phys. Chem. C, Vol. 112, No. 5, 2008

Mohammad et al. hydrogen adsorption, the effect of subsurface oxygen centers is indirect, via induction of an upward shift of the d manifold of the Ag atoms in the surface.76 5. Conclusions

Figure 5. Three-fold hollow hcp (H1, H2) and fcc (F1, F2) adsorption sites of atomic hydrogen on a Ag(111) surface, modified by oxygen centers occupying subsurface oss sites at a coverage of 1/4 ML. The solid frame indicates a 2 × 2 surface unit cell. Calculated adsorption energies of single atoms H are also shown (in kilojoules per mole). Two adjacent F2 sites represent the final state for the hydrogen activation process studied (see text). The subsurface oxygen center is represented as black sphere.

complexes of H at F1 are strongly destabilized, by about 75 kJ mol-1 compared with H adsorbed on the ideal Ag(111) surface without subsurface impurities, where Ead ) -198 kJ mol-1.16 At adsorption site H1, in nearest-neighbor position of Ooss, the calculated destabilization of H with respect to the ideal Ag(111) surface is very small, only 5 kJ mol-1. More importantly, at the somewhat more distant sites F2 and H2, adsorption of hydrogen becomes more stable than on the clean Ag(111) surface, by 12-13 kJ mol-1. [We calculated qualitatively similar H adsorption energies at the lower content of subsurface O, θoss ) 1/9 ML (not shown).] Note that, on Ag(111), with all octahedral interstitial positions beneath the top Ag layer occupied by O atoms (θoss ) 1 ML), the adsorption interaction of H atoms strengthens by as much as 88 kJ mol-1, compared with the clean Ag(111) surface.76 In the model with coverage θoss ) 1/4 ML and a p(2 × 2) unit cell, we chose as final state (FS) of the H2 activation process the most stable co-adsorption configuration of two H atoms at two neighboring fcc sites F2, θH ) 1/2 ML (Figure 3e). H2 dissociation on this surface was calculated endothermic by 15 kJ mol-1 (Table 3); that is, it is ∼15 kJ mol-1 more favorable than on the clean Ag(110) surface.16 The corresponding activation energy for H2 dissociation is 98 kJ mol-1; the activated complex with H-H ) 142 pm represents a late TS (Table 3). The calculated activation energy of H2 on Ooss/Ag(111) is ∼30 kJ mol-1 higher than on the p(2 × 1)O/Ag(110) surface,16 but ∼25 kJ mol-1 lower than the lowest value on a clean Ag(110) surface. This latter finding agrees with other DF results according to which subsurface O reduces the barrier of H2 activation compared with clean silver surfaces.76 The calculated activation barrier for H2 was shown to decrease from 107 kJ mol-1 for the clean Ag(111) surface [or from 126 kJ mol-1 on Ag(110)16] to 71 kJ mol-1 for subsurface oxygen at a coverage 1/2 ML.76 Our activation energy Ea ) 98 kJ mol-1 for θoss ) 1/4 ML is 27 kJ mol-1 higher than that preceding result, because a higher coverage of subsurface O in the latter case resulted in a more stable final state. Note that the barrier of 71 kJ mol-1 by Xu et al.76 is very close to the H2 activation energies of 6976 kJ mol-1 determined in our study for added-row modified p(n × 1)O/Ag(110) surfaces.16 These similar values of the activation barrier seem accidental in view of the fact that the corresponding mechanisms of H2 activation on silver systems with oxygen at the surface and in the subsurface region are different. Whereas surface oxygen centers of added-row systems act directly as active centers for

With the goal to rationalize possible sources of atomic hydrogen for the hydrogenation of organic molecules on silver catalysts, we studied computationally the activation (dissociation) of H2 on different model oxygen species adsorbed on various Ag surfaces as well as trapped in the subsurface region: O/Ag(110), OH/Ag(110), p(2 × 1)O/Ag(110), p(2 × 1)OH/Ag(110), and Ooss/Ag(111). In line with our previous result on H2 dissociation on a -Ag-O- added-row chain of the reconstructed p(2 × 1)O/Ag(110) surface,16 we found that other oxygen species, present on or underneath a silver surface, as modeled by the systems just mentioned, are also able to promote dissociation of molecular hydrogen. For isolated oxygen centers adsorbed on Ag(110), we suggested a mechanism that involves successive activation of two H2 molecules. Interaction with a first H2 molecule produces a hydroxyl group and a H atom on the metal surface; the OH group is able to react with another H2 molecule to produce water (which desorbs) and one more H atom on the metal surface. Surprisingly, activation of H2 on adsorbed hydroxyl groups exhibits the lowest activation barrier among the species considered. As a single surface oxygen center can produce two atomic H species (that can further participate in hydrogenation) and afterward desorbs as H2O molecule, the catalytic protocol in this model should include an oxygen source. A similar mechanism was considered for the reconstructed p(2 × 1)O/ Ag(110) surface with the difference that a single O center gives rise to only one H atom on a metal terrace, ready to participate in hydrogenation of other molecules. Subsurface oxygen species Ooss, located in interstitial octahedral positions, decrease the barrier for H2 activation compared with clean Ag surfaces. However, they seem less efficient for H2 activation than surface oxygen as the corresponding activation barrier was calculated higher. Moreover, in the absence of adsorbed oxygen, subsurface oxygen species are unstable and predicted to leave the subsurface sites with a low (or without any) activation barrier. Nevertheless, subsurface oxygen species can be regarded as sites promoting the formation of atomic hydrogen required for hydrogenation reactions on silver catalysts. Various sources of oxygen come to mind.17 Oxygen may be supplied by the oxide support of the silver clusters, from the “bulk” region of large silver particles, or even from decay products of the compound to be hydrogenated (e.g., in the case of unsaturated aldehydes). Clearly, further studies of this topic are highly desirable, in particular, efforts to characterize the structure of the catalyst in more detail. Acknowledgment. We thank M. Bron and P. Claus for stimulating discussions. This work was supported by Deutsche Forschungsgemeinschaft and Fonds der Chemischen Industrie (Germany). I.Y. thanks for support by the Russian Foundation for Basic Research (Grant 06-03-33020). K.M.N. is grateful for financial support from the Spanish Ministry of Education and Science (Grants CTQ2005-08459-CO2-01 and UNBA05-33001) and the Generalitat de Catalunya (Grants 2005SGR-00697 and 2005PEIR 0051/69). References and Notes (1) Van Santen, R. A.; Kuipers, H. P. C. E. AdV. Catal. 1987, 35, 265. (2) Claus, P. Topics Catal. 1998, 5, 52.

Hydrogen Activation on Silver (3) Bron, M.; Kondratenko, E.; Trunschke, A.; Claus, P. Z. Phys. Chem. 2004, 218, 1. (4) Bron, M.; Teschner, D.; Knop-Gericke, A.; Steinhauer, B.; Scheybal, A.; Ha¨vecker, M.; Wang, D.; Fo¨disch, R.; Ho¨nicke, D.; Wootsch, A.; Schlo¨gl, R.; Claus, P. J. Catal. 2005, 234, 37. (5) Bron, M.; Teschner, D.; Knop-Gericke, A.; Jentoft, F. C.; Kro¨hnert, J.; Hohmeyer, J.; Volckmar, C.; Steinhauer, B.; Schlo¨gl, R.; Claus, P. Phys. Chem. Chem. Phys. 2007, 9, 3559. (6) Mohr, C.; Hofmeister, H.; Lucas, M.; Claus, P. Chem. Eng. Technol. 2000, 23, 324. (7) Avouris, Ph.; Schmeisser, D.; Demuth, J. E. Phys. ReV. Lett. 1982, 48, 199. (8) Sprunger, P. T.; Plummer, E. W. Phys. ReV. B 1993, 48, 14436. (9) Christmann, K. Surf. Sci. Rep. 1988, 9, 1. (10) Zhou, X.; White, J. M. Surf. Sci. 1989, 218, 201. (11) Lee, G.; Plummer, E. W. Phys. ReV. B 1995, 51, 7250. (12) Mijoule, C.; Russier, V. Surf. Sci. 1991, 254, 329. (13) Eichler, A.; Kresse, G.; Hafner, J. Surf. Sci. 1998, 397, 116. (14) Greeley, J.; Mavrikakis, M. J. Phys. Chem. B 2005, 109, 3460. (15) Montoya, A.; Schlunke, A.; Haynes, B. S. J. Phys. Chem. B 2006, 110, 17145. (16) Mohammad, A. B.; Lim, K. H.; Yudanov, I. V.; Neyman, K. M.; Ro¨sch, N. Phys. Chem. Chem. Phys. 2007, 9, 1247. (17) Bron, M.; Teschner, D.; Wild, U.; Steinhauer, B.; Knop-Gericke, A.; Wang, D.; Volckmar, C.; Wootsch, A.; Schlo¨gl, R.; Claus, P., in preparation. (18) We take the oxygen coverage throughout as the ratio of O atoms to Ag atoms in the surface layer. (19) Canepa, M.; Salvietti, M.; Traverso, M.; Mattera, L. Surf. Sci. 1995, 331, 183. (20) Kresse, G.; Furthmu¨ller, J. Phys. ReV. B 1996, 54, 11169. (21) Kresse, G.; Hafner, J. Phys. ReV. B 1993, 47, 558. (22) Kresse, G.; Furthmu¨ller, J. Comput. Mater. Sci. 1996, 6, 15. (23) Perdew, J. P.; Wang, Y. Phys. ReV. B 1992, 45, 13244. (24) Blo¨chl, P. E. Phys. ReV. B 1994, 50, 17953. (25) Kresse, G.; Joubert, D. Phys. ReV. B 1999, 59, 1758. (26) Monkhorst, H. J.; Pack, J. D. Phys. ReV. B 1976, 13, 5188. (27) Methfessel, M.; Paxton, A. T. Phys. ReV. B 1989, 40, 3616. (28) When studying hydrogen activation on O/Ag(110) surface, we used a larger unit cell with six Ag atoms per layer to minimize interactions between O atoms in the neighboring unit cells. (29) Li, W.-X.; Stampfl, C.; Scheffler, M. Phys. ReV. B 2003, 67, 045408. (30) Yudanov, I. V.; Neyman, K. M.; Ro¨sch, N. Phys. Chem. Chem. Phys. 2004, 6, 116. (31) Lim, K. H.; Neyman, K. M.; Ro¨sch, N. Chem. Phys. Lett. 2006, 432, 184. (32) Chen, Z.-X.; Lim, K. H.; Neyman, K. M.; Ro¨sch, N. J. Phys. Chem. B 2005, 109, 4568. (33) Makov, G.; Payne, M. C. Phys. ReV. B 1995, 51, 4014. (34) Lim, K. H.; Zakharieva, O.; Shor, A. M.; Ro¨sch, N. Chem. Phys. Lett. 2007, 444, 280. (35) Mills, G.; Jo´nsson, H.; Schenter, G. K. Surf. Sci. 1995, 324, 305. (36) Grant, R. B.; Lambert, R. M. J. Catal. 1985, 92, 364. (37) Bukhtiyarov, V. I.; Ha¨vecker, M.; Kaichev, V. V.; Knop-Gericke, A. R.; Mayer, W.; Schlo¨gl, R. Phys. ReV. B 2003, 67, 235422. (38) Bao, X.; Muhler, M.; Schedel-Niedrig, Th.; Schlo¨gl, R. Phys. ReV. B 1996, 54, 2249. (39) Outlaw, R. A.; Wu, D.; Hoflund, G. B.; Davidson, M. R. J. Vac. Sci. Technol., A 1992, 10, 1497. (40) Rehren, C.; Muhler, M.; Bao, X.; Schlo¨gl, R.; Ertl, G. Z. Phys. Chem. 1991, 174, 11. (41) Savio, L.; Gerbi, A.; Vattuone, L.; Baraldi, A.; Comelli, G.; Rocca, M. J. Phys. Chem. B 2006, 110, 942. (42) Savio, L.; Vattuone, L.; Rocca, M. Appl. Phys. A 2007, 87, 399.

J. Phys. Chem. C, Vol. 112, No. 5, 2008 1635 (43) Sun, Q.; Wang, Y.; Fan, K.; Deng, J. Surf. Sci. 2000, 459, 213. (44) Pazzi, V. I.; Herman, P.; Philipsen, T.; Baerends, E. J.; Tantardini, G. F. Surf. Sci. 1999, 443, 1. (45) Olsson, F. E.; Lorente, N.; Persson, M. Surf. Sci. 2003, 522, L27. (46) Engelhardt, H. A.; Menzel, D. Surf. Sci. 1976, 57, 591. (47) In original publication (ref 46), the oxygen coverage θO is given at 0.3 ML, as the coverage of 1 ML was taken to correspond to saturated (2 × 1)O/Ag(110), i.e., θO ) 0.5 ML according to our definition; see ref 18. (48) Puschmann, A.; Haase, J. Surf. Sci. 1984, 144, 559. (49) Becker, L.; Aminipirooz, S.; Schmalz, A.; Hillert, B.; Pedio, M.; Haase, J. Phys. ReV. B 1991, 44, 13655. (50) Hashizume, T.; Taniguchi, M.; Motai, K.; Lu, H.; Tanaka, K.; Sakurai, T. Jpn. J. Appl. Phys., Part 2 1991, 30, L1529. (51) Taniguchi, M.; Tanaka, K.; Hashizume, T.; Sakurai, T. Surf. Sci. 1992, 262, L123. (52) Chen, Z.-X.; Neyman, K. M.; Lim, K. H.; Ro¨sch, N. Langmuir 2004, 20, 8068. (53) Wert, Ch. A. In Hydrogen in Metals II; Alefeld, G., Vo¨lkl, J., Eds.; Springer-Verlag: Berlin, 1978; Vol. 29, p 305. (54) Over, H.; Seitsonen, A. P. Science 2002, 297, 2003. (55) In accordance with our general definition, ref 18, we measure the density θoss of occupied oss sites with respect to the number of Ag atoms in the top layer of the (111) surface. Therefore, θoss ) 1.0 ML refers to a situation where all oss sites below the top (111) layer are occupied. (56) Carlisle, C. I.; Fujimoto, T.; Sim, W. S.; King, D. A. Surf. Sci. 2000, 470, 15. (57) The binding energies of oxygen, reported in ref 43 for all sites (LB, 4f, SB, top) are about 250 kJ mol-1 larger than those obtained in the present study and other works. (58) Michaelides, A.; Ranea, V. A.; de Andres, P. L.; King, D. A. Phys. ReV. Lett. 2003, 90, 216102. (59) Montoya, A.; Haynes, B. S. J. Phys. Chem. C 2007, 111, 1333. (60) Solomon, J. L.; Madix, R. J. J. Phys. Chem. 1987, 91, 6241. (61) Solomon, J. L.; Madix, R. J.; Sto¨hr, J. J. Chem. Phys. 1988, 89, 5316. (62) Guillemot, L.; Bobrov, K. Surf. Sci. 2007, 601, 871. (63) Canepa, M.; Cantini, P.; Mattera, L.; Narducci, E.; Salvietti, M.; Terreni, S. Surf. Sci. 1995, 322, 271. (64) Lim, K. H.; Moskaleva, L. V.; Ro¨sch, N. Chem. Phys. Chem. 2006, 7, 1802. (65) Campbell, C. T. Surf. Sci. 1985, 157, 43. (66) Rovida, G.; Pratesi, F.; Maglietta, M.; Ferroni, E. Surf. Sci. 1974, 43, 230. (67) Carlisle, C. I.; King, D. A.; Bocquet, M.-L.; Cerda, J.; Sautet, P. Phys. ReV. Lett. 2000, 84, 3899. (68) Li, W.-X.; Stampfl, C.; Scheffler, M. Phys. ReV. Lett. 2003, 90, 256102. (69) Michaelides, A.; Bocquet, M.-L.; Sautet, P.; Alavi, A.; King, D. A. Chem. Phys. Lett. 2003, 367, 344. (70) Michaelides, A.; Reuter, K.; Scheffler, M. J. Vac. Sci. Technol., A 2005, 23, 1487. (71) Schnadt, J.; Michaelides, A.; Knudsen, J.; Vang, R. T.; Reuter, K.; Lægsgaard, M.; Scheffler, M.; Besenbacher, F. Phys. ReV. Lett. 2006, 96, 146101. (72) Schmid, M.; Reicho, A.; Stierle, A.; Costina, I.; Klikovits, J.; Kostelnik, P.; Dubay, O.; Kresse, G.; Gustafson, J.; Lungren, E.; Andersen, J. N.; Dosch, H.; Varga, P. Phys. ReV. Lett. 2006, 96, 146102. (73) Campbell, C. T.; Paffett, M. T. Surf. Sci. 1984, 143, 517. (74) Nagy, A. J.; Mestl, G.; Herein, D.; Weinberg, G.; Kitzelmann, E.; Schlo¨gl, R. J. Catal. 1999, 182, 417. (75) Nagy, A. J.; Mestl, G.; Schlo¨gl, R. J. Catal. 1999, 188, 58. (76) Xu, Y.; Greeley, J.; Mavrikakis, M. J. Am. Chem. Soc. 2005, 127, 12823.