Nature of Ag Islands and Nanoparticles on the CeO2(111) Surface

ARTICLE pubs.acs.org/JPCC

Nature of Ag Islands and Nanoparticles on the CeO2(111) Surface Paola Luches S3, Istituto Nanoscienze
CNR, Via G. Campi 213/a, 41125 Modena, Italy

Federico Pagliuca and Sergio Valeri Dipartimento di Fisica - Universit
a di Modena e Reggio Emilia and S3, Istituto Nanoscienze
CNR, Via G. Campi 213/a, 41125 Modena, Italy

Francesc Illas Departament de Química Física & Institut de Química Te
orica i Computacional (IQTCUB), Universitat de Barcelona, C/Martí i Franqu
es 1, 08028 Barcelona, Spain

Gloria Preda and Gianfranco Pacchioni* Dipartimento di Scienza dei Materiali, Universit
a di Milano-Bicocca, Via R. Cozzi, 53
20125, Milano, Italy ABSTRACT: Ag nanoparticles have been deposited on stoichiometric and reduced thin CeO2 films grown on Pt(111). The nucleation and growth of the Ag nanoparticles has been characterized by STM and XPS (X-ray photoemission spectroscopy) measurements complemented with DFT calculations on Ag atoms, clusters, and extended layers deposited on slab models of the CeO2(111) surface. The XPS spectra clearly show a reduction of the ceria support by Ag deposition (formation of Ce3+ ions). This is accompanied by a positive shift of the Ag 3d core levels, in which final state effects related to the finite size of the Ag deposits come into play. The DFT calculations support the view of a direct electron transfer from the Ag clusters and nanoparticles to the ceria support. Other possible origins of the reduction of the ceria substrate, like the occurrence of oxygen reverse spillover on the Ag nanoparticles, are ruled out based on energy considerations.

1. INTRODUCTION CeO2, often in combination with other oxides, with supported metals, or with dopants of various nature, is one of the most important industrial and environmental catalysts.1,2 It is considered a prototype of reducible oxides, due to its capability to lose and adsorb oxygen, a property which is at the basis of its usage in many catalytic reactions. It's no surprise that it is attracting increasing interest also from a fundamental point of view. In fact, the nature of reduced CeO2 and of oxygen vacancies in the bulk or on the surface of this oxide have been the subject of several studies.3
10 An aspect which has been investigated in detail is the role of transition metal particles on the activity of CeO2 in oxidation processes, water
gas shift, and soot combustion in diesel antiparticulate filters.11,12 It is generally assumed that the oxidation processes proceed via a Mars
Van Krevelen mechanism where an adsorbed species is oxidized by lattice oxygens with formation of oxygen vacancies and consequent reduction of CeO2 (having only formally Ce4+ cations) to CeO2‑x (exhibiting formally both types of cations Ce4+ and Ce3+). The role of supported transition metals is probably that to facilitate the dissociation of the gasphase oxygen molecules, thus lowering the energy demand for a crucial step in the whole catalytic cycle. r 2011 American Chemical Society

Of particular interest for this study is the role of silver and the nature of the CeO2
Ag interface. The addition of Ag to nanocrystalline CeO2 leads to an efficient catalyst for CO and hydrocarbon oxidation.13 Machida et al.14 have reported that loading the ceria catalyst with Ag metal leads to an improved catalytic activity and a lower operating temperature in soot combustion. In their experiments, CeO2 has been mixed with various amounts of Ag, from a few percent to about 20% in weight. The results show a reduction of the combustion temperature from about 390 °C to about 340 °C for 10% Ag content. This change in activity has been tentatively explained with the fact that metallic silver can promote the formation of superoxo species.15,16 In fact, the electron paramagnetic resonance (EPR) signal associated to O2
increases slightly when CeO2 is loaded with metallic silver.14 In another study, Aneggi et al.17 have shown that Ag/CeO2 samples (prepared by impregnation of CeO2 with an aqueous solution of AgNO3 followed by calcination) contain Ag crystallites separated from the ceria particles by an Ag2O phase, Received: October 25, 2011 Revised: November 29, 2011 Published: December 06, 2011 1122

dx.doi.org/10.1021/jp210241c | J. Phys. Chem. C 2012, 116, 1122–1132

The Journal of Physical Chemistry C raising some questions about the characteristics of the active catalytic phase, Ag, Ag2O, or their interfaces with ceria. The possible role of silver
ceria boundaries in solid oxide fuel cells has been emphasized in a theoretical study by Wang et al.18 Other authors have reported that ceria is able to maintain silver in a positive oxidation state.19 In a recent study, Shimuzu et al.20 reported a novel method of preparation of Ag/CeO2 catalysts based on calcination of Ag metal and CeO2 which leads to well dispersed Ag nanoparticles of about 8 nm diameter and in an active catalyst for soot oxidation. From a more fundamental side, Ag adsorption on CeO2 welldefined surfaces has been studied with sophisticated surface science techniques.21
23 The conclusions of these papers are quite relevant for the present discussion and will be analyzed in detail below. We will concentrate on the CeO2(111) surface, used for the majority of the studies dealing with cerium oxidebased model systems. We have performed a series of experiments on the deposition of Ag nanoparticles on thin CeO2 films grown on Pt(111). Various amounts of Ag have been deposited on ceria films with different degrees of reduction. The nucleation process has been followed by STM (scanning tunneling microscopy), whereas the electronic structure of the Ag aggregates has been studied by XPS (X-ray photoemission spectroscopy) as a function of temperature. These studies have been complemented by first principles density functional theory (DFT) calculations on stoichiometric and reduced CeO2 surface models. One problem connected to the description of reduced ceria with standard DFT is that this method fails in describing the localized nature of the Ce 4f states.24,25 This can be overcome by using hybrid functionals26 or the so-called DFT+U approach27 where the standard DFT energy functional is augmented by an on-site Hubbard-like interaction. Here we have followed this second approach.

2. EXPERIMENTAL AND COMPUTATIONAL METHODS 2.1. Experiments. The samples used for this study were grown in an ultrahigh vacuum apparatus equipped with an ion gun (Omicron ISE10) for surface cleaning, evaporators, a gas line, a conventional X-ray source, a hemispherical electron analyzer (Omicron EA 125), a low-energy electron diffraction (LEED) apparatus (Omicron SPECTALEED), and a room temperature scanning tunneling microscope (Omicron UHV AFM/STM). The substrate used was a Pt(111) single crystal, cleaned by repeated cycles of sputtering (1 keV, 1 μA) and annealing (1040 K). After the cleaning procedure, the substrate showed flat terraces hundreds of nanometers wide, the concentration of impurities on the sample surface was below the XPS detection limit, and the LEED showed a sharp (1  1) pattern. Cerium oxide ultrathin films were grown by reactive evaporation of Ce, using an e-beam evaporator with a rate of approximately 0.2 Å/min, measured by a quartz microbalance. For this study films of 3 and 6 ML thickness were used. During the evaporation molecular oxygen was supplied through a nozzle in a background O2 pressure of 1  10
7 Torr. The Pt substrate was kept at room temperature during the evaporation. After the growth the cerium oxide films were annealed at 1040 K in 1  10
7 Torr O2 pressure for 15 min, to obtain epitaxial films with large flat terraces and to minimize the Ce3+ concentration.28 Ag was evaporated from a Knudsen cell on the cerium oxide films at room temperature. The Ag evaporation rate used for the samples described in this work was 0.3 Å/min. In the following,

ARTICLE

the amount of deposited ceria is expressed in terms of ceria monolayers (ML), defined as the thickness of an O
Ce
O trilayer in the (111)-oriented fluorite structure, corresponding to 3.12 Å. The nominal thickness of deposited Ag is expressed in Å. The XPS measurements have been performed using Al Kα photons and detecting photoelectrons at an emission angle of 65° from sample normal, to increase the surface sensitivity. The analysis of the Ce 3d photoemission spectra has been performed following the procedure introduced by Skala and co-workers.29,30 The values reported for the Ce3+ concentration after Ag deposition are obtained with the assumption that all Ce3+ sites formed are located in the surface layer. This assumption does not hold for the cerium oxide samples in which a relatively high concentration of Ce3+ sites is induced by high temperature annealing. For these samples the Ce3+ concentration refers to the total volume probed by XPS. The Ce3+ concentration values within the XPS probing depth are affected by an absolute error of (2%, due to some intrinsic asymmetries of the Ce4+ related peaks,29 which give rise to non negligible uncertainties in the fitting procedure. When the Ce3+ concentration is assumed to be within the topmost layer, the absolute error becomes 3%. The Ag 3d photoemission lines have been fitted by a Voigt-shaped spin
orbit split doublet, while the Pt 4f lines were fitted by a Doniach-Sunjich shaped doublet. We carefully checked that X-ray exposure did not modify the shapes and positions of the XPS lines at the doses used for the measurements. The binding energy (BE) of the XPS spectra were aligned to have the same Pt 4f BE in all samples, to avoid extrinsic shifts due to sample charging effects. The STM images were acquired at RT in constant current mode using electrochemically etched tungsten tips and a positive sample-to-tip bias. The removal of some Ag nanoparticles in the typical measuring conditions (2 V, 0.1
0.2 nA) was often observed. Care has been taken in trying to minimize the effect by adjusting the measurement parameters. The STM images were processed using WSXM31 and other commercial software. 2.2. DFT Calculations. Spin polarized DFT calculations have been carried out to study the interaction between Ag atoms, clusters, and extended one or two monolayers (ML) and the CeO2(111) surface. The calculations have been carried out with the Vienna ab initio simulation package (VASP)32,33 using a plane wave basis set and the projector augmented wave (PAW) method34 to describe the interaction between the valence electrons and the atomic cores. The valence electron density is defined by the twelve electrons of each Ce atom (5s25p6 6s25d14f1), the six electrons of each O atom (2s22p4), and the eleven electrons of Ag (5s14d10). The plane-wave expansion includes all plane waves with kinetic energy smaller than a cutoff of 415 eV. The LDA+U or GGA+U approaches35
37 were chosen to account for exchange and correlation. The LDA and GGA parts of these functionals are those of Vosko et al.38 (VWN) and Perdew
Wang39,40 (PW91), respectively. In the LDA+U (and GGA+U) method, part of the self-interaction energy is corrected by explicit inclusion of a Hubbard like Ueff term for the 4f electrons penalizing partial occupancy of this atomic level. In the present work, we used the formalism of Dudarev et al.41 Numerical integration in the reciprocal space was carried out using a 2  2  1 Monkhorst-Pack special k-points grid.42 A Methfessel-Paxton smearing width σ = 0.2 eV was applied to help to converge the electronic density43 although only meaningful energy values properly obtained with σ = 0 eV are considered when necessary. 1123

dx.doi.org/10.1021/jp210241c |J. Phys. Chem. C 2012, 116, 1122–1132

The Journal of Physical Chemistry C

Figure 1. STM images of (a) a 3 ML CeO2 film on Pt(111) (I = 1.5 nA, V = 3 V); (b) a detail of panel a evidencing O
Ce
O steps (A), linear defects (B), bidimensional defects (C), and holes down to the Pt substrate (D) (I = 1.5 nA, V = 3 V); (c) 0.07 Å Ag on a 3 ML CeO2 film (I = 0.1 nA, V = 2.5 V); (d) detail of panel c revealing the tendency for Ag nanoparticles to decorate the steps between ceria terraces, inset: detail showing hexagonal nanoparticles (I = 0.1 nA, V = 2.5 V); (e and f) 0.2 Å Ag on a 3 ML CeO2 film (I = 1.5 nA (e) and I = 0.2 nA (f), V = 2 V).

Geometry optimization has been performed within the LDA+U exchange-correlation potential with Ueff = 5 eV scheme because it provides a lattice parameter which is close to experiment;27 magnetic moments and Bader charges44
46 have been obtained at the same level of theory whereas energies were calculated at the GGA+U level with Ueff = 3 or 4 eV (in general the GGA approach is better for energy considerations, not necessarily for other properties). This choice allows one to properly describe localization in bulk Ce2O3 as well as in ceria nanoparticles containing both Ce3+ and Ce4+ atoms.8 To model the O-terminated CeO2(111) surface 2  2 (Ce8O16), 3  3 (Ce18O36), or 4  4 (Ce32O64) supercells with 6 or 9 atomic layers (2 or 3 ML, respectively) have been used keeping the three bottom layers fixed and relaxing the uppermost ones. The slab model was cut from the bulk cubic (Fm3m) CaF2like structure using the optimized LDA+U lattice parameter a0 of 5.40 Å, in excellent agreement with experimental a0 ≈ 5.41 Å. The supercell includes a vacuum width of about 12 Å which is large enough to avoid interaction between the slabs obtained after replication in the three space dimensions.

3. EXPERIMENTAL RESULTS The substrates used for this study were cerium oxide ultrathin films grown on Pt(111), annealed at 1040 K in O2 after the growth. For nominal 3 ML CeO2 thickness approximately 90% of the Pt surface is covered by atomically flat cerium oxide terraces several tens of nanometers wide (Figure 1 a), separated by single O
Ce
O steps (A in Figure 1 b). If the CeO2 nominal

ARTICLE

Figure 2. (a) Ag 3d XPS spectra of samples with different amount of Ag and for a 50 Å Ag film deposited on a 3 ML CeO2 film. The spectra have been normalized in area to the bulk Ag value. (b) Ag 3d5/2 binding energy as a function of Ag nominal thickness. The Ag 3d5/2 binding energy of a 50 Å Ag film is also shown as a solid line.

thickness is increased to 6 ML the Pt(111) surface coverage increases to almost 100%. Linear defects, several tens of nanometers long, ascribed to domain boundaries, are also present on the terraces (B in Figure 1b). Furthermore, a number of defects with a few nanometers lateral size are present (C in Figure 1b), together with atomic scale defects (not shown).28 Most of the darker holes with lateral size of a few tens of nanometers (D in Figure 1b) extend down to the Pt surface, while only some of them expose the ceria layers underneath. The ceria films show a sharp (1.4  1.4) LEED pattern, which indicates an epitaxial growth on Pt(111).28 The fitting of the Ce 3d XPS spectra of the bare films indicates a Ce3+ concentration below 1
2% within the XPS probing depth.28 The deposition of 0.07 Å of Ag at RT leads to the formation of Ag nanoparticles on the ceria surface, as shown by the STM image in Figure 1c,d. The smaller scale image (Figure 1d) reveals the tendency for Ag nanoparticles to decorate ceria
ceria terrace step edges, while the steps between CeO2 and Pt, the linear defects and the large scale defects on the ceria terraces are not decorated. Some nanoparticles are found also on the terraces, possibly nucleating at some smaller scale surface defect. The Ag nanoparticle shape appears mainly hexagonal (inset of Figure 1d), suggesting a (111) orientation of the Ag nanoparticles surface, confirmed by the LEED patterns measured at higher Ag coverage (not shown). The Ag particle lateral size and height range from approximately 1 to 3 nm and from 0.2 to 1 nm, respectively. The particle density is 1.2  1012 np/cm2. By increasing the amount of deposited Ag to 0.2 Å the nanoparticles increase mainly in lateral size and height (Figure 1e,f), while their density is only slightly increased to 1.6  1012 np/cm2. The electronic properties of the Ag/CeO2 system were investigated by XPS. Figure 2a shows the Ag 3d XPS lines for samples with different amount of Ag and for a 50 Å thick Ag film, 1124

dx.doi.org/10.1021/jp210241c |J. Phys. Chem. C 2012, 116, 1122–1132

The Journal of Physical Chemistry C

Figure 3. (a) XPS spectra of a 3 ML CeO2 film before (black curve) and after (red curve) the deposition of 0.2 Å Ag. The arrows indicate the regions in which Ce3+-related peaks appear. (b) Ce3+ concentration in the topmost layer as a function of Ag nominal thickness, evaluated from the fitting of Ce 3d XPS spectra.

the latter used as a reference for bulk Ag. The signals have been normalized in area to the bulk Ag value, to allow for a comparison of the shape and position of the peaks in the different samples. Figure 2b reports the BE of the Ag 3d5/2 peak as a function of Ag nominal coverage for Ag deposited on a 3 ML and on a 6 ML CeO2 film. At the lowest Ag coverages the Ag 3d5/2 BE is shifted to higher values compared to the bulk by up to 0.5 eV. The Ag 3d5/2 BE decreases with increasing Ag deposited amount reaching almost the bulk value above 4 Å. The observed shift toward high binding energy values for low amounts of deposited Ag can be ascribed to the reduced coordination of Ag atoms in the nanoparticles compared to the bulk. A similar shift was frequently observed on metal nanoparticles of few nanometers size.47,23,48,49 A charge transfer from Ag nanoparticles to the ceria substrate can also induce a shift in the binding energy value. At variance with most metals, in which the binding energy is higher in the positive ionic state than in the metallic state due to initial state effects, in bulk Ag oxides the Ag 3d binding energy is lower than in the metallic state due a large extra-atomic relaxation energy contribution, which dominates over the positive shift in the initial state.50 In our system we certainly have dimensionality related effects, which hinder the possibility of measuring eventual shifts due to charge transfer. In order to have additional information on possible charge transfer effects we investigated the evolution of Ce 3d XPS spectra. Figure 3a shows the spectrum of a 3 ML ceria film measured before and after the growth of 0.2 Å of Ag. The change of the shape of the Ce 3d spectrum after Ag growth is consistent with the appearance of the most intense of Ce3+-related doublet with the main peak at 885 eV and an 18.5 eV spin
orbit splitting, and it indicates that ceria is reduced by Ag growth. The fitting of the spectra gives a concentration of Ce3+ of 5% in the topmost layer after the deposition of 0.2 Å of Ag. The observed ceria reduction can be due to an electron transfer from Ag or by other effects, such as O-reverse spillover or oxygen loss due to oxidation of carbonaceous species present in the background

ARTICLE

Figure 4. STM images of 0.2 Å Ag deposited on (a) a stoichiometric film (cCe3+ = 1%; I = 1.5 nA, V = 2 V); (b) a film reduced in UHV for 1 h (cCe3+ = 21%; I = 0.2.5 nA, V = 2 V); (c) a film reduced in UHV for 2 h (cCe3+ = 31%; I = 0.05 nA, V = 2 V); (d) density of Ag nanoparticles measured on the STM images as a function of Ce3+ concentration of the substrate measured within the XPS probing depth and linear fit of the data.

pressure. To have further information on this issue we performed a fitting of the Ce 3d spectra after the deposition of different amounts of Ag. As shown in Figure 3b the Ce3+ concentration in the topmost layer increases as a function of the amount of deposited Ag up to 14% when all of the ceria surface is covered by Ag. This behavior is qualitatively consistent with the hypothesis of a progressive charge transfer with increasing the fraction of CeO2 surface in contact with Ag nanoparticles. The growth of Ag on cerium oxide films with different Ce3+ concentration was also investigated. We have shown that UHV annealing at 1040 K of CeO2 films allows one to obtain surfaces with different degree of surface reduction.28 Figure 4a
c shows the topography of 0.2 Å Ag deposited on a stoichiometric cerium oxide film and on two reduced cerium oxide films, obtained after 1 and 2 h UHV annealing, with 21% and 31% Ce3+ concentration within the XPS probing depth, respectively. A linear increase of Ag np density as a function of Ce3+ concentration can be observed (Figure 4d), at least in the investigated Ce3+ concentration range, indicating that probably surface oxygen vacancies formed on the annealed surfaces hinder the mobility of Ag atoms and act as nucleation centers for Ag nanoparticles. As a consequence of the increase in nanoparticles density, their height appears decreased from 1.1 to 0.9 nm on the reduced ceria surfaces. The possibility to resolve a significant decrease of lateral size in the Ag nanoparticles on the reduced ceria substrates is partially hindered by the finite size of the STM tip. A similar increase in nanoparticle density on reduced ceria surfaces was observed also by Zhou and co-workers in a recent study of Pt nanoparticles on ceria,47 whereas Rh grows with a very similar morphology on oxidized and reduced ceria surfaces.51 1125

dx.doi.org/10.1021/jp210241c |J. Phys. Chem. C 2012, 116, 1122–1132

The Journal of Physical Chemistry C

Figure 5. (a) Ag 3d XPS spectra of 0.2 Å Ag samples deposited on ceria films with different Ce3+ concentration (measured within the XPS probing depth) and for a 50 Å Ag film. The 50 Å Ag film spectrum has been normalized to the average area of the other spectra for comparison. (b) Ag 3d5/2 binding energy as a function of the density of Ag nanoparticles. The Ag 3d5/2 binding energy of a 50 Å Ag film is also shown as a solid line.

Figure 5a shows the Ag 3d XPS lines for 0.2 Å Ag deposited on the three substrates with a different degree of reduction and the Ag 3d line of the thick Ag film for comparison. In this case only the thick film signal was normalized to the average area of the other samples. The Ag 3d area of the 0.2 Å Ag samples on the three ceria films with different surface reduction are approximately the same, thus confirming the hypothesis that the observed increase in nanoparticles density is accompanied by a consequent size reduction; that is, the Ag amount is the same on the three samples. The Ag 3d peaks are progressively shifted to higher BE compared to the bulk as the Ag np density increases and the particle size decreases (Figure 5b). This evidence confirms that there is a significant contribution to the Ag 3d BE shift induced by dimensionality effects. In the already discussed case of the fully oxidized substrate the Ce3+ concentration in the topmost layer increases from 1% to 5% after 0.2 Å Ag growth. The 1 h UHV annealed sample shows the same Ce3+ concentration before and after Ag growth (21%). The 2 h UHV annealed sample instead shows an apparently slightly lower Ce3+ concentration (29%) after the growth than before it (31%). In the latter case the small decrease of Ce 3+ relative weight in the Ce 3d spectrum after the growth may also be induced by the preferential attenuation of photoemission from Ce3+ sites due to the nucleation of Ag clusters on surface O vacancies. The thermal stability of the Ag/CeO2 samples was also studied by STM and XPS. Figure 6 shows some selected images for the 0.07 b and 0.2 Å Ag samples as grown and after UHV annealing for 15 min at increasing values of temperature. No apparent change was measured in the STM images with annealing up to 470 K (images not shown). After annealing to higher temperatures the clusters progressively coalesce, increasing in size and decreasing in density. In particular, in the 0.07 Å Ag/CeO2 sample after annealing at 570 K the cluster density is significantly

ARTICLE

Figure 6. STM images of (a
c) 0.07 Å Ag on a 3 ML CeO2 film as grown and after annealing at 570 and 620 K [I = 0.1 nA, V = 2.5 V (a); I = 0.2 nA, V = 2 V (b); I = 0.4 nA, V = 2 V (c)]; (d
f) 0.2 Å Ag on a 3 ML CeO2 film as grown and after annealing at 570 and 620 K [I = 1.5 nA, V = 2 V (d); I = 0.2 nA, V = 3 V (e); I = 0.4 nA, V = 2 V (f)].

reduced (Figure 6b) and at 620 K all of the Ag nanoparticles disappear from the cerium oxide surface (Figure 6c). The 0.2 Å Ag/CeO2 sample (Figure 6, panels d and e) shows the same qualitative behavior with annealing, but the cluster density is systematically higher than for the thinner sample. The thermal stability of 0.2 Å Ag on the reduced ceria films was also tested, without finding significantly different results compared to the case of the oxidized films, i.e., in spite of the initially larger density of Ag particles on the reduced ceria films, the density of nanoparticles on the oxidized and reduced samples decrease to the same values at the same temperatures. XPS spectra were measured on the 0.2 and 0.07 Å Ag samples on the oxidized ceria film and on the 0.2 Å Ag on the ceria film with 21% Ce3+ concentration before and after the thermal treatments. Figure 7 a reports the Ag 3d5/2 BE as a function of postgrowth annealing temperature for the three samples. After annealing at temperatures between 570 and 670 K the BE decreases to the bulk value. This decrease is consistent with the increase in dimensionality of the clusters in the considered temperature range. Above a specific temperature, different for the three samples and corresponding to temperatures for which all of the Ag nanoparticles are removed from the ceria surface, a negative Ag 3d5/2 BE shift is observed with respect to the bulk value. The observed effect is ascribed to the nanoparticles which remain in the holes of the ceria film in contact with the Pt substrate, as confirmed by the BE position of a sample of 0.2 Å of Ag deposited on the Pt(111) surface and annealed to 670 K (dotted gray line in Figure 7a). Figure 7b reports the values of Ce3+ concentration within the XPS probing depth as a function of postgrowth annealing temperature. The observed increase within the annealing temperature range is of the same amount in the 3 samples, regardless of the initial Ce3+ concentration, suggesting a thermal effect independent of the presence of Ag nanoparticles. The same increase in Ce3+ concentration is observed also for a bare CeO2 sample of the same thickness. 1126

dx.doi.org/10.1021/jp210241c |J. Phys. Chem. C 2012, 116, 1122–1132

The Journal of Physical Chemistry C

ARTICLE

a reduced Ce3+ ion Ag1 þ CeO2 f Ag1 þ =CeO2
ð1Ce3þ Þ

Figure 7. (a) Ag 3d5/2 binding energy as a function of annealing temperature for a 0.07 Å Ag (light blue dots) and 0.2 Å Ag (blue dots) deposited on a stoichiometric CeO2 film and for 0.2 Å Ag (dark blue dots) deposited on a reduced ceria film (the solid lines are guidelines to the eye). (b) Ce3+ concentration measured within the XPS probing depth as a function of annealing temperature for 0.07 Å Ag (light blue squares) and 0.2 Å Ag (blue squares) deposited on a stoichiometric CeO2 film (left scale) and for 0.2 Å Ag (dark blue squares) deposited on a reduced ceria film (right scale).

4. THEORETICAL RESULTS 4.1. Ag Atoms on CeO2(111) and CeO2‑x(111). Deposition of Ag atoms on the CeO2(111) surface can result in Ag species adsorbed on regular sites of the surface but, under special conditions, could also result in the substitution of Ce lattice ions. This second process has been considered by Metiu and co-workers who demonstrated by DFT calculations that the presence of Ag ions replacing Ce ions in the surface of CeO2(111) causes a considerable weakening of the Ce
O bond and a much lower formation energy of oxygen vacancies.52 According to this model, if the Ag ions segregate toward the surface they facilitate oxidation mechanisms based on the Mars
Van Krevelen mechanism since the activation barrier for oxygen transfer is considerably reduced. Incorporation of Ag atoms in the ceria lattice, however, requires very different preparation conditions than used in this work. Here Ag vapor is deposited from a Knudsen cell at room temperature on the CeO2(111) thin films and it is likely that isolated atoms arrive on the surface. The interaction of an Ag atom with the regular and defective surfaces of CeO2(111) has already been considered53,54 and is discussed here only for comparative purposes. Two stable adsorption sites have been identified: in one case (preferred) Ag is above an O atom of the third layer and is bound by 0.7 eV to three O atoms of the first atomic layer; in the other case Ag is directly on-top of an O atom of the first atomic layer (Ead = 0.49 eV). Notice that the absolute value of the Ag adsorption energy should be taken with care since it can change significantly with the computational setup adopted.53 In all sites Ag behaves as an electron donor with respect to CeO2 with formation of an Ag+ cation and

ð1Þ

This is clearly shown (1) by the value of the Bader charge on Ag, +0.54 e, (2) by the spin density maps which indicate that the 5s electron of Ag is no longer present while an unpaired electron is localized on the 4f states of one Ce ion, and (3) by the DOS plots which show the appearance of a singly occupied 4f state in the band gap of the material. Thus, Ag atoms act as alkali metals and result in the formation of reduced Ce3+ ions. Quite different is the interaction of Ag atoms with O vacancies. The removal of one oxygen atom corresponds to a reduction of the system and in fact two electrons are found in localized 4f states on two surface Ce ions which change their configuration from Ce4+(4f0) to Ce3+(4f1).3
9 Recent studies have shown that there is a distribution of Ce sites where the localization of the electrons associated to the oxygen vacancy can take place so that multiple structures exist with formally the same electronic configuration but with similar total energies.55,56 An Ag atom binds rather strongly, by 1.4 eV, to an O vacancy at the CeO2‑x(111) surface. The atom is directly on-top of an O atom of the third layer and in direct contact with three Ce atoms around the vacancy. Since two Ce3+ 4f1 ions are associated to the vacancy and Ag has a 5s1 valence configuration, there is a maximum of three unpaired electrons. In the final structure, however, only one unpaired electron remains. This can be interpreted as due to an electron transfer from one Ce3+ ion to Ag1 which formally becomes Ag
(5s2) Ag1 þ CeO2
x ð2Ce3þ Þ f Ag1
=CeO2
x þ ð1Ce3þ Þ

ð2Þ

The process can be seen also as due to the coupling of the Ce 4f1 with the Ag 5s1 electrons with formation of a covalent Ce
Ag bond. The Bader charges on Ag (
0.48 e) and the DOS curves which show the double occupancy of the Ag 5s level point more toward the first interpretation. 4.2. Ag Clusters on CeO2(111) and CeO2‑x(111). Preliminary calculations on an Ag5 cluster adsorbed on the CeO2(111) surface have shown the tendency of Ag clusters to be oxidized and of ceria to be reduced.54 Gas-phase Ag5 has a doublet ground state and when deposited to the CeO2(111) surface is bound by 1.96 eV (computed with respect to a bypiramidal trigonal gasphase isomer). The structure of the Ag5 cluster is just one of the several possible isomers. As for a single Ag atom, also for Ag5 there is a net charge transfer toward the oxide. In the ground state (LDA+U and GGA+U = 4), two electrons are transferred from Ag5 to the oxide support, see Figure 8 (the Bader charge on Ag5 is +0.84, the magnetic moments on two Ce ions 0.87 and 0.76). In a second solution, 0.23 eV higher in energy, only one electron is transferred to a surface Ce ion which changes oxidation state from Ce4+ to Ce3+. This second solution (one transferred electron) is nearly degenerate with the other one (two transferred electrons) at the GGA+U (U = 3 eV) level, indicating that while there is a clear tendency to transfer electrons from silver to ceria, the exact amount of charge transferred is method dependent. In general, the Ag atoms of the cluster at the boundary with ceria are slightly oxidized while the apical Ag atom (2nd layer) keeps a metallic character. The oxidation of Ag5 and the concomitant reduction of ceria could be biased by the presence of an unpaired electron on the cluster and by an easy ionization of this system. Furthermore, very small clusters have discrete energy levels whose energy 1127

dx.doi.org/10.1021/jp210241c |J. Phys. Chem. C 2012, 116, 1122–1132

The Journal of Physical Chemistry C

ARTICLE

Figure 8. Ag5 cluster deposited on the stoichiometric CeO2(111) surface. (a) Structure and spin distribution; (b) DOS curves. Red, O; dark gray, Ce; pale gray, Ag.

Figure 9. Ag10 cluster deposited on the stoichiometric CeO2(111) surface. (a) Structure and spin distribution; (b) DOS curves. Red, O; dark gray, Ce; pale gray, Ag.

depends on several factors like coordination number, shape of the cluster, etc. In other words, the occurrence of a charge transfer could be an artifact related to the special electronic structure of Ag5. For this reason we have considered a larger Ag10 cluster with closed shell ground state. The cluster contains 7 Ag atoms in the first layer and 3 Ag atoms in the second layer, Figure 9. It is only one of the several possible isomers of supported Ag10 but it is representative of a twolayer nanoparticle. The Ag10 cluster has been adsorbed with the more extended face on CeO2(111) and the geometry has been fully reoptimized. The spin density analysis shows that after Ag10 deposition there are three unpaired electrons localized on the 4f states of three Ce ions which are thus reduced to Ce3+. Some spin density is also found on the top layer of the Ag cluster, Figure 9. The formation of Ce3+ ions is confirmed by the DOS curves which show that the corresponding states are below the Fermi level. So, not only the Ag10 cluster is oxidized by interaction with the ceria surface, but this involves more than a single electron (LDA+U and GGA+U = 4 results). These results have been obtained with a 2 ML thick CeO2 support. We further checked that the charge transfer occurs also when the ceria support is modeled by a thicker 3 ML slab but the LDA+U results confirm the occurrence of a multiple charge transfer. This is in line with recent theoretical studies on Au adsorption on CeO2(111) which showed no dependence of the results on the thickness of the ceria support.57 Also in this case, however, the extent and nature of the charge transfer changes when the GGA+U (U = 3 eV) method is adopted: for Ag10 on CeO2(2 ML) the charge transfer is still

present, but the spin on the cerium oxide is delocalized over several Ce ions. In the following we discuss only LDA+U for geometry and properties and GGA+U (U = 4 eV) results, keeping in mind that for smaller values of the U parameter the results may be partly different and probably physically less meaningful. In summary, Ag atoms or Ag nanoclusters deposited on a stoichiometric CeO2 surface are directly oxidized with consequent reduction of the Ce ions of the support. However, the extent of the charge transfer and spin localization is a rather subtle question which depends on the level of accuracy of the theoretical treatment. The next step has been to consider the same Ag5 and Ag10 species adsorbed on an oxygen vacancy (interaction of reduced ceria with silver species). When Ag5 is adsorbed in the vicinity of an O vacancy on the CeO2‑x(111) surface the bonding is slightly larger than on the regular surface (Ead = 2.21 eV versus 1.96 eV) and we found a total of three Ce3+ ions, Figure 10; although two of these ions are related to the presence of the vacancy, the third one is due to an electron transfer from the Ag cluster. This is consistent with a positive Bader charge on the cluster (+0.4 e). Thus, even on a reduced surface there is a tendency of Ag5 to be oxidized. This is further confirmed by the results obtained on Ag10 adsorbed on CeO2‑x(111), Figure 10. Here we have a total of four Ce3+ ions, two of which are due to an electron transfer from the Ag cluster to the reduced ceria surface. A positive charge of +0.94e is present on the Ag cluster. Also in this case there is a moderate increase in adsorption energy going from Ag10 adsorbed on the regular surface, Ead = 3.66 eV, to the same cluster 1128

dx.doi.org/10.1021/jp210241c |J. Phys. Chem. C 2012, 116, 1122–1132

The Journal of Physical Chemistry C

ARTICLE

Figure 10. (a) Ag5 and (b) Ag10 clusters deposited on the reduced CeO2‑x(111) surface. Red, O; dark gray, Ce; pale gray, Ag.

adsorbed on reduced CeO2, Ead = 4.39 eV. This suggests that the mobility of Ag clusters should be smaller on the reduced ceria surface. In summary, on both stoichiometric and reduced ceria surfaces the Ag clusters act as electron donors and become oxidized at the expense of the ceria support. This process is spontaneous, and can involve more than one electron per cluster. 4.3. Extended Ag Deposits (Monolayer and Bilayer). So far we have considered the interaction between a finite system, Ag atoms or clusters, and the CeO2(111) surface and we have observed that ceria is always reduced due to an electron transfer from the supported species. This effect could be related to the nanodimensionality of the adsorbate and to the presence of orbitals at high energy which favor the electron transfer to the Ce 4f states. Large Ag nanoparticles, with diameter of a few nanometers, could behave quite differently due to the development of a metallic character (the work function in bulk Ag is 4.2 eV). For this reason we have considered a monolayer and a bilayer Ag film which has been adapted to the CeO2(111) support. We have also considered two different models of the CeO2(111) surface, CeO2(2 ML) and CeO2(3 ML). The difference is that in the CeO2(2 ML) system all Ce atoms are surface atoms (either top or bottom atomic layers), while in CeO2(3 ML) the central layer is more “bulk-like”. To simulate the Ag monolayer a 4  4 Ag(111) cell (16 Ag atoms) has been adapted to a 3  3 Ce18O36 supercell (CeO2(2 ML)); the two cells have nearly the same dimensions and the strain is about 1% (the Ag layer has been contracted to adapt to the ceria lattice constant). When we considered 1 ML Ag on CeO2(2 ML) we found clear evidence of a charge transfer, Figure 11. At both the LDA+U and GGA+U (U = 4 eV) levels we found that one electron per unit cell is transferred from the Ag deposit to the oxide with formation of a single Ce3+ ion. The Ag film has a non-negligible rumpling and the average distance between the top layer of Ce ions and the Ag film is of 2.44 Å with the shortest Ag
O distance being 2.19 Å; this is only slightly longer than the Ag
O distances in Ag2O (2.05 Å)58 or AgO (2.05
2.18 Å).59 There is a partial spin polarization of the Ag layer (μAg = 0.36) and an average charge on the Ag layer of +0.50e. Also for a 2 ML Ag film grown on CeO2 (2 ML) the tendency to reduce the oxide is confirmed: at both LDA+U and GGA+U (U = 4 eV) levels we have clear evidence of four Ce4+ ions which are reduced to Ce3+. The occurrence of a charge transfer is further confirmed by the Bader charges on the Ag layer at contact with CeO2 which is +1.48e while the upper Ag layer is in zero oxidation state.

We further checked the results by using a 3 ML CeO2 support. For a single Ag layer the solution where no charge transfer occurs is preferred at all levels of treatment (LDA+ U, GGA+U either with U = 3 or 4 eV). The charge transfer solution is about 0.3 eV higher in energy, but the spin on the ceria support is delocalized and not easy to quantify. When we consider 2 ML of Ag on 3 ML of CeO2 we have indications of the occurrence of a multiple electron transfer. These results seem to indicate that, at variance with the Ag cluster models, for an extended Ag monolayer the tendency to reduce the ceria support is less pronounced and depends on the thickness of the ceria support used. As we mentioned before, the main difference between 2 and 3 ML of ceria is that in the former case there is a large number of undercoordinated Ce ions. This suggests that the electron transfer from the Ag deposits is easier in the presence of low-coordinated sites, a situation which is common on small ceria nanoparticles or islands. 4.4. Oxygen Spillover on Silver. The DFT calculations indicate the reduction of the ceria support by deposition of Ag atoms and clusters. While this fits with a number of experimental observations (see also the discussion in section 5, below), the reduction of ceria by deposition of metal clusters could also be due to a reverse spillover of O atoms from the ceria surface to the metal cluster or nanoparticle, with formation of a reduced understoichiometric CeO2‑x support and of an oxygen covered Ag adsorbate. The process considered is thus Agn =CeO2 f Agn O=CeO2
x ð2Ce3þ Þ

ð3Þ

We first considered the small Ag5 cluster. If the O atom is included inside the Ag5 cage, in an interstitial position, the system is unstable and the O atom spontaneously moves toward the surface refilling the vacancy. On the contrary, if the O atom is bound to the top of the Ag5 cluster we found a local minimum, Figure 12a, which is about 2.9 eV less stable than the reactants, i.e., Ag5/CeO2(111). The same process has been considered using the Ag10 supported cluster. In this case we did not consider an interstitial O atom since the results for Ag5 clearly show the preference for this species to diffuse to the vacancy and restore the original stoichiometry. Thus, the O atom has been moved from the Ag/CeO2 interface to the top layer of the cluster where it is bound to Ag in a three-hollow position, Figure 12b. In this position the O atom is strongly bound to the supported Ag cluster, De = 3.01 eV, but this is far from compensating the cost of creating a vacancy and the entire system is 2.53 eV less stable than a Ag10 cluster adsorbed on a stoichiometric CeO2(111) surface. 1129

dx.doi.org/10.1021/jp210241c |J. Phys. Chem. C 2012, 116, 1122–1132

The Journal of Physical Chemistry C

ARTICLE

Figure 11. Structure and DOS of Ag films on CeO2(111). (Top) Ag(1 ML)/CeO2(111)(2 ML); (bottom) Ag(2 ML)/CeO2(111)(2 ML). Red, O; dark gray, Ce; pale gray, Ag.

Figure 12. Oxygen reverse spillover from the CeO2(111) regular surface to Ag5 and Ag10 clusters. (a) Structure of Ag5O/CeO2‑x. (b) Structure of Ag10O/CeO2‑x. In both cases the Ag cluster is adsorbed above the O vacancy site generated by moving the O atom to the cluster surface. Red, O; dark gray, Ce; pale gray, Ag.

These results indicate that the cost of moving one O atom from the ceria surface to the adsorbed Ag nanoparticle is much too high to consider the reverse spillover of oxygen as a plausible mechanism for ceria reduction.

5. DISCUSSION AND CONCLUSIONS The analysis of the XPS spectra of the deposited Ag nanoparticles on the thin ceria films grown on Pt(111) provide clear evidence for the reduction of the oxide by Ag deposition. This can be deduced from the line shape of the Ce 3d states which

show the characteristic change associated to the presence of Ce3+ ions for the deposition of Ag on stoichiometric CeO2(111) (Figure 3). This effect is accompanied by a positive shift of the Ag 3d core level binding energy by decreasing the size of the deposited nanoparticles: at a nominal thickness of 0.2 Å the Ag 3d5/2 binding energy is 368.5 eV, 0.3 eV higher than for metallic silver. As the nominal thickness further decreases, the Ag 3d5/2 core level binding energy becomes 368.7 eV, with a shift of about 0.5 eV with respect to metallic silver. These results are similar to those recently reported by Kong et al.23 on Ag nanoparticles 1130

dx.doi.org/10.1021/jp210241c |J. Phys. Chem. C 2012, 116, 1122–1132

The Journal of Physical Chemistry C deposited on CeO2(111)/Ru(0001) films. However, while there is no doubt about the fact that the ceria support is reduced upon silver deposition, the origin of this phenomenon is still under debate. In fact, core level binding energies of supported metal particles are affected by final state effects arising from final state screening, which strongly depend on the size of the nanoparticles. In this respect, the simple analysis of the core level shifts of the metal nanoparticles can be misleading.60 An attempt to disentangle the initial and final state contributions to the Ag 3d core level shifts has been done in ref 23 based on a procedure originally suggested by Wagner.61,62 According to this method, the shift of the Auger peak and of the core level binding energy approximately equals twice the final state contributions to the core level shifts. Using this delicate decomposition technique, it was concluded that the initial state contribution to the Ag 3d core level shifts is negligible, that the observed positive shift is entirely due to final state screening, and that the Ag nanoparticles on CeO2(111) are metallic (zero oxidation state). This conclusion contrasts with the results from DFT calculations described above which show a clear tendency of the Ag nanoparticles to become oxidized by simple contact with the ceria surface. Of course, in our experiments as well as in those of Kong et al.23 there is also clear evidence that the ceria support is reduced by Ag deposition. According to the DFT calculations this is the natural consequence of the charge transfer from Ag, while a different origin has to be invoked if one assumes that the Ag particles remain in zero oxidation state. In order to rationalize the apparent reduction of ceria, in ref 23 it has been suggested that spillover of lattice oxygen occurs at the Ag/CeO2 boundary, and that oxygen diffuses on the surface of the Ag nanoparticles, thus leading to a CeO2‑x/AgnOm system. This mechanism has been explicitly investigated in our DFT calculations, see reaction 3, and the results clearly show that oxygen reverse spillover from the CeO2(111) surface to Ag is an energetically very unfavorable process. This is in line with the low affinity of Ag for O and with thermodynamic data. In fact, ΔH°(Ag2O),
31.1 KJ/mol, is more than 30 times smaller than that of CeO2 (ΔH°(CeO2) =
1088.7 KJ/mol). This is also consistent with the relatively high cost to create an O vacancy on the CeO2(111) surface, 3.54 eV if computed with respect to 1/2 O2, 6.62 eV with respect to atomic O. This energy is hardly compensated by the formation of the bond between adsorbed O and an Ag particle. Experimentally, the initial heat of adsorption of oxygen on Ag single crystal surfaces is of about 170 KJ/mol,63,64 corresponding to an adsorption energy of atomic oxygen of about 330 KJ/mol, a value well reproduced by DFT GGA calculations (about 310 KJ/mol).65 Thus, the heat of adsorption of O2 on Ag surfaces is 3
4 times smaller than the oxidation enthalpy of ceria which oscillates between 500 KJ/mol at low extents of reduction to 760 KJ/mol for highly reduced samples.66 All these data consistently point toward an energetically unfavorable transfer of oxygen from the ceria support to Ag particles. A recent joint experimental-theoretical study on the oxygen reverse spillover on Pt clusters supported on ceria nanoparticles has shown that the process is greatly facilitated on small ceria crystallites because of the presence of low-coordinated O atoms and of a substantial lowering of the formation energy of an O vacancy at these sites.67 However, the experiments of ref 23 as well as those reported here are dealing with large CeO2(111) islands where the role of the border should be much less important than on ceria nanocrystals.

ARTICLE

Finally, we notice that the observed increase of the concentration of Ce3+ ions with the amount of Ag deposited (Figure 3b) is a further support for the hypothesis of ceria reduction by electron transfer from Ag. In fact, we observed by STM that an increase in the amount of deposited Ag results in particles with larger size (Figure 1c
f), with almost the same density, at least in the submonolayer range, in agreement also with the results obtained by Farmer et al.22 The increase in Ce3+ concentration with the amount of Ag deposited is consistent with the observed increase of the ceria surface coverage. The O reverse spillover process, instead, is not expected to be more pronounced as the Ag particle size increases. Furthermore, the observed increase of Ce3+ concentration with annealing the Ag/CeO2 system is comparable with the one of the bare CeO2 surface, at variance with systems in which O reverse spillover was found to be dominant.67

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ ACKNOWLEDGMENT Gloria Preda thanks the Consortium CORIMAV between Pirelli and the University of Milano Bicocca for a PhD fellowship and HPC Europa for supporting her stay at the University of Barcelona. Stimulating discussions with Luca Castellani and Livia Giordano are gratefully acknowledged. F.I. acknowledges financial support provided by Spanish MICINN (Grant FIS200802238) and by Generalitat de Catalunya (Grants 2009SGR1041 and XRQTC). F.I. also acknowledges additional support through 2009 ICREA Academia award for excellence in research. The work has been supported by the Italian MIUR through the FIRB Project RBAP115AYN “Oxides at the nanoscale: multifunctionality and applications”, a CINECA award under the ISCRA initiative for the availability of high performance computing resources and by Regione Lombardia and CILEA Consortium through a LISA Initiative (Laboratory for Interdisciplinary Advanced Simulation) 2011 grant [link: http://lisa.cilea.it]. ’ REFERENCES (1) Kaspar, J.; Fornasiero, P.; Graziani, M. Catal. Today 1999, 50, 285. (2) Trovarelli, A. Catal. Rev. 1996, 38, 439. (3) Skorodumova, N. V.; Simak, S. I.; Lundqvist, B. I.; Abrikosov, I. A.; Johansson, B. Phys. Rev. Lett. 2002, 16, 166601. (4) Nolan, M.; Grigoleit, S.; Sayle, D. C.; Parker, S. C.; Watson, G. W. Surf. Sci. 2005, 576, 217. (5) Hay, P. J.; Martin, R. L.; Uddin, J.; Scuseria, G. E. J. Chem. Phys. 2006, 125, 034712. (6) Mei, D.; Deskins, A. N.; Dupuis, M.; Ge, Q. J. Phys. Chem. C 2007, 111, 10514. (7) Branda, M. M.; Loschen, C.; Neyman, K. M.; Illas, F. J. Phys. Chem. C 2008, 112, 17643. (8) Loschen, C.; Migani, A.; Bromley, S. T.; Illas, F.; Neyman, K. M. Phys. Chem. Chem. Phys. 2008, 10, 5730. (9) Watkins, M. B.; Foster, A. S.; Shluger, A. L. J. Phys. Chem. C 2009, 111, 15337. (10) Ganduglia-Pirovano, M. V.; Hofmann, A.; Sauer, J. Surf. Sci. Rep. 2007, 62, 219. (11) Avgouropoulos, G.; Manzoli, M.; Boccuzzi, F.; Tabakova, T.; Papavasiliou, J.; Ioannides, T.; Idakiev, V. J. Catal. 2008, 256, 237. (12) Dyakonov, A. J.; Little, C. A. Appl. Catal. B: Environ. 2006, 67, 52. 1131

dx.doi.org/10.1021/jp210241c |J. Phys. Chem. C 2012, 116, 1122–1132

The Journal of Physical Chemistry C (13) Bera, P.; Patil, K. C.; Hegde, M. S. Phys. Chem. Chem. Phys. 2000 2, 3715. (14) Machida, M.; Murata, Y.; Kishikawa, K.; Zhang, D.; Ikeue, K. Chem. Mater. 2008, 20, 4489. (15) Outka, D. A.; St€ohr, J.; Jark, W.; Stevens, P.; Solomon, J.; Madix, E. J. Phys. Rev. B 1987, 35, 4119. (16) Gravil, P. A.; Bird, D. M.; White, J. A. Phys. Rev. Lett. 1996, 77, 3933. (17) Aneggi, E.; Llorca, J.; de Leitenburg, C.; Dolcetti, G.; Trovarelli, A. Appl. Catal. B: Environ. 2009, 91, 489. (18) Wang, J. H.; Liu, M. L.; Lin, M. C. Sold State Ionics 2006, 177, 939. (19) Murriel, L. L.; Carlin, R. T. J. Catal. 1996, 159, 479. (20) Shimizu, K.; Kawachi, H.; Komai, S.; Yoshida, K.; Sasaki, Y.; Satsuma, A. Catal. Today 2011, 175, 93. (21) Farmer, J. A.; Campbell, C. T. Science 2010, 329, 933. (22) Farmer, J. A.; Baricuatro, J. H.; Campbell, C. T. J. Phys. Chem. C 2010, 114, 17166. (23) Kong, D.; Wang, G.; Pan, Y.; Hu, S.; Hou, J.; Pan, H.; Campbell, C. T.; Zhu, J. J. Phys. Chem. C 2011, 115, 6715. (24) Esch, F.; Fabris, S.; Zhou, L.; Montini, T.; Africh, C.; Fornasiero, P.; Comelli, G.; Rosei, R. Science 2005, 309, 752. (25) Fabris, S.; Vicario, G.; Balducci, G.; de Gironcoli, S.; Baroni, S. J. Phys. Chem. B 2005, 109, 22860. (26) Da Silva, J. L. F.; Ganduglia-Pirovano, M. V.; Sauer, J. Phys. Rev. B 2007, 75, 045121. (27) Loschen, C.; Carrasco, J.; Neyman, K. M.; Illas, F. Phys. Rev. B 2007, 75, 035115. (28) Luches, P.; Pagliuca, F.; Valeri, S. J. Phys. Chem C 2011, 115, 10718. (29) Skala, T.; Sutara, F.; Prince, K. C.; Matolín, V. J. Electron Spectrosc. Relat. Phenom. 2009, 169, 20. (30) Skala, T.; Sutara, F.; Skoda, M.; Prince, K. C.; Matolín, V. J. Phys: Condens. Matter. 2009, 21, 055005. (31) Horcas, I.; Fernandez, R.; Gomez-Rodríguez, J. M.; Colchero, J.; Gomez-Herrero, J.; Baro, A. M. Rev. Sci. Instrum. 2007, 78, 013705. (32) Kresse, G.; Hafner, J. Phys. Rev. B 1993, 47, 558. (33) Kresse, G.; Furthm€uller, J. Phys. Rev. B 1996, 54, 11169. (34) Bl€ochl, P. E. Phys. Rev. B 1994, 50, 17953. (35) Anisimov, V. I.; Aryasetiawan, F.; Lichtenstein, A. I. J. Phys.: Condens. Matter 1997, 9, 767. (36) Anisimov, V. I.; Solovyev, I. V.; Korotin, M. A.; Czyzyk, M. T.; Sawatzky, G. A. Phys. Rev. B 1993, 48, 16929. (37) Solovyev, I. V.; Dederichs, P. H.; Anisimov, V. I. Phys. Rev. B 1994, 50, 16861. (38) Vosko, S. H.; Wilk, L.; Nusair, M. Can. J. Phys. 1980, 58, 1200. (39) Perdew, J. P.; Chevary, J. A.; Vosko, S. H.; Jackson, K. A.; Pederson, M. R.; Singh, D. J.; Fiolhais, C. Phys. Rev. B 1992, 46, 6671. (40) Perdew, J. P.; Chevary, J. A.; Vosko, S. H.; Jackson, K. A.; Pederson, M. R.; Singh, D. J.; Fiolhais, C. Phys. Rev. B 1993, 48, 4978. (41) Dudarev, S. L.; Botton, G. A.; Savrasov, S. Y.; Humphreys, C. J.; Sutton, A. P. Phys. Rev. B 1998, 57, 1505. (42) Monkhorst, H. J.; Pack, J. D. Phys. Rev. B 1976, 13, 5188. (43) Methfessel, M.; Paxton, A. T. Phys. Rev. B 1989, 40, 3616. (44) Bader, R. F. W. Atoms in Molecules; Oxford University Press: Oxford, 1990. (45) Henkelman, G.; Arnaldsson, A.; Jonsson, H. Comput. Mater. Sci. 2006, 36, 254. (46) Tang, W.; Sanville, E.; Henkelman, G. J. Phys.: Condens. Matter 2009, 21, 084204. (47) Campbell, C. T. Surf. Sci. Rep. 1997, 27, 1. (48) Zhou, Y.; Perket, J. M.; Zhou, J. J. Phys. Chem. C 2010, 114, 11853. (49) Torelli, P.; Giordano, L.; Benedetti, S.; Luches, P.; Annese, E.; Valeri, S.; Pacchioni, G. J. Phys. Chem. C 2009, 113, 19957. (50) Bao, X.; Wild, U.; Muhler, M.; Pettinger, B.; Schl€ogl, R.; Ertl, G. Surf. Sci. 1999, 425, 224. (51) Zhou, J.; Baddorf, A. P.; Mullins, D. R.; Overbury, S. J. Phys. Chem. C 2008, 112, 9336.

ARTICLE

(52) Shapovalov, V.; Metiu, H. J. Catal 2007, 245, 205. (53) Branda, M. M.; Hernandez, N. C.; Sanz, J. F.; Illas, F. J. Phys. Chem. C 2010, 114, 1934. (54) Preda, G.; Pacchioni, G. Catal. Today 2011, 177, 31. (55) Ganduglia-Pirovano, M. V.; Da Silva, J. L. F.; Sauer, J. Phys. Rev. Lett. 2009, 102, 026101. (56) Li, H.-Y.; Wang, H.-F.; Gong, X.-Q.; Guo, Y.-L.; Lu, G.; Hu, P. Phys. Rev. B 2009, 79, 193401. (57) Branda, M. M.; Castellani, N. J.; Grau-Crespo, R.; de Leeuw, N. H.; Hernandez, N. C.; Sanz, J. F.; Neyman, K. M.; Illas, F. J. Chem. Phys. 2009, 131, 094702. (58) Tjeng, L. H.; Meinders, M. B. J.; Vanelp, J.; Ghijsen, J.; Sawatzky, G. A.; Johnson, R. L. Phys. Rev. B 1990, 41, 3190. (59) Scatturin, V.; Bellon, P. L. J. Electrochem. Soc. 1961, 108, 819. (60) Bagus, P. S.; Illas, F.; Pacchioni, G.; Parmigiani, F. J. Electron Spectrosc. Relat. Phenom. 1999, 100, 215. (61) Wagner, C. D. Anal. Chem. 1972, 44, 967. (62) Wagner, C. D.; Gale, L. H.; Raymond, R. H. Anal. Chem. 1979, 51, 466. (63) Enegelardt, H. A.; Menzel, D. Surf. Sci. 1979, 57, 591. (64) Campbell, C. T. Surf. Sci. 1986, 173, L641. (65) Li, W. X.; Stampfl, C.; Scheffler, M. Phys. Rev. B 2002, 65, 075407. (66) Zhou, G.; Shah, P. R.; Montini, T.; Fornasiero, P.; Gorte, R. J. Surf. Sci. 2007, 601, 2512. (67) Vayssilov, G. N.; Lykhach, Y.; Migani, A.; Staudt, T.; Petrova, G. P.; Tsud, N.; Skala, T.; Bruix, A.; Illas, F.; Prince, K. C.; Matolin, V.; Neyman, K. M.; Libuda, J. Nat. Mater. 2011, 10, 310.

1132

dx.doi.org/10.1021/jp210241c |J. Phys. Chem. C 2012, 116, 1122–1132