14860
J. Phys. Chem. B 2006, 110, 14860-14869
Catalytic Oxidation Activity of Pt3O4 Surfaces and Thin Films Nicola Seriani,*,†,⊥ Wolfgang Pompe,† and Lucio Colombi Ciacchi‡,§ Institut fu¨r Werkstoffwissenschaft, Technische UniVersita¨t Dresden, Hallwachsstrasse 3, 01069 Dresden, Germany, Theory of Condensed Matter Group, CaVendish Laboratory, UniVersity of Cambridge, J J Thomson AVenue, Cambridge CB3 0HE, United Kingdom, and Fraunhofer Institut fu¨r Werkstoffmechanik, Wo¨hlerstrasse 11, 79108 Freiburg, Germany ReceiVed: May 28, 2006
The catalytic oxidation activity of platinum particles in automobile catalysts is thought to originate from the presence of highly reactive superficial oxide phases which form under oxygen-rich reaction conditions. Here we study the thermodynamic stability of platinum oxide surfaces and thin films and their reactivities toward oxidation of carbon compounds by means of first-principles atomistic thermodynamics calculations and molecular dynamics simulations based on density functional theory. On the Pt(111) surface the most stable superficial oxide phase is found to be a thin layer of R-PtO2, which appears not to be reactive toward either methane dissociation or carbon monoxide oxidation. A PtO-like structure is most stable on the Pt(100) surface at oxygen coverages of one monolayer, while the formation of a coherent and stress-free Pt3O4 film is favored at higher coverages. Bulk Pt3O4 is found to be thermodynamically stable in a region around 900 K at atmospheric pressure. The computed net driving force for the dissociation of methane on the Pt3O4(100) surface is much larger than that on all other metallic and oxide surfaces investigated. Moreover, the enthalpy barrier for the adsorption of CO molecules on oxygen atoms of this surface is as low as 0.34 eV, and desorption of CO2 is observed to occur without any appreciable energy barrier in molecular dynamics simulations. These results, combined, indicate a high catalytic oxidation activity of Pt3O4 phases that can be relevant in the contexts of Pt-based automobile catalysts and gas sensors.
1. Introduction The majority of platinum metal demand, namely more than hundred tons per year, originates from its application as a catalyst in the automotive industry.1 In automobile catalysts, platinum is the primary element used to oxidize uncombusted hydrocarbons and carbon monoxide present in the exhausts, while rhodium is employed to reduce nitrogen oxide compounds.2 Although the properties of platinum as an oxidation catalyst have been investigated already by Langmuir3 almost a century ago, the precise nature of the catalytically active sites is still the subject of intense investigation efforts.4-8 An oxide layer is thought to form on the surface of the Pt clusters of a car catalyst under oxygen-rich conditions,9 and the total oxidation of carbon compounds most probably follows a Mars-van Krevelen reaction mechanism.10 Namely, the oxygen atoms required to form the reaction products are extracted by the reactants from the oxide lattice, which is subsequently restored upon reaction with molecular oxygen. This hypothesis was put forward a decade ago by Mallens et al.11 after a series of carefully conducted experiments of methane oxidation on platinum gauzes at about 1000 K. In more recent years Hendriksen and Frenken4 investigated the temporal evolution of a Pt(110) surface in situ during the catalytic combustion of CO in a high-pressure flow reactor coupled to a scanning * Address correspondence to this author. E-mail: nicola.seriani@ univie.ac.at. ⊥ Present address: Institut fu ¨ r Materialphysik, Universita¨t Wien, Sensengasse 8, A-1090 Wien, Austria. † Technische Universita ¨ t Dresden. ‡ University of Cambridge. § Fraunhofer Institut fu ¨ r Werkstoffmechanik.
tunneling microscope. These authors found that a sudden increase in the catalytic activity of the surface is associated with a dramatic change in its morphology, which was attributed to the formation of a superficial oxide layer. Theoretical investigations into the atomic structure of this layer and its catalytic activities have only recently been carried out.5,7 However, very little is known about the structural details or the catalytic oxidation activity of thin oxide layers grown on (100) and (111) Pt surfaces, which are the main facets of small Pt metal crystallites such as those in an automobile catalyst.12-15 In this work we investigate the structures of platinum oxide layers grown on (100) and (111) Pt surfaces, and probe their activity toward the dissociative adsorption of methane and the oxidation of CO via a combination of first-principles atomistic thermodynamics and first-principles molecular dynamics (FPMD) techniques based on the density functional theory (DFT). In general, the catalytic activity of thin oxide layers on transition metals is the result of a delicate dynamical equilibrium between reactants and products, and the availability of suitable reactive sites. This leads to reaction channels which are active only in a narrow range of temperatures and oxygen pressures, at equilibrium conditions which may be reached on a time scale of seconds, and may be dominated by kinetic effects.16 An automobile catalyst is a very complex system whose working state is difficult to define,17 whose features vary with different crystallographic orientation of the metal crystallites, and whose activity may be heavily influenced by the ceramic support.17 Furthermore, at very high temperatures decomposition of the oxide layer and formation of volatile oxide species cannot be excluded.18 This makes the experimental investigation of the catalytic mechanisms active on the surface of Pt particles
10.1021/jp063281r CCC: $33.50 © 2006 American Chemical Society Published on Web 07/06/2006
Catalytic Oxidation Activity of Platinum Particles difficult to perform at the atomic scale. Only recently, Hammer and co-workers19 and Frenken and co-workers7 have shown that surface science techniques, coupled with state-of-the-art atomistic modeling techniques, can be used to provide insight into the formation of oxide thin films. Atomistic modeling is routinely used nowadays to investigate the relative thermodynamic stabilities of different oxide layers on metallic surfaces at the pressure and temperature conditions typical of catalytic applications.20 In particular, the recent developments of so-called first-principles atomistic thermodynamics techniques have led to the characterization of many surface oxide structures on a variety of transition metal surfaces.16 Notably, thermodynamically stable ultrathin oxide layers can be noncommensurate with respect to the underlying substrate, nonstoichiometric, and can present structures which reflect the short-range ordering of the bulk phases in typical coordination motifs around the metal ions, but do not have the same long-range order.21 We attempt here to identify thermodynamically stable superficial platinum oxide phases which present particularly high reactivity toward methane decomposition and CO oxidation. The starting point of our study is the commonly formed platinum oxide structures. These comprise PtO, Pt3O4, R-PtO2, and β-PtO2.22,23 As reported by Muller and Roy,22 R-PtO2 is the thermodynamically stable bulk phase at low temperatures and low oxygen pressures. At higher temperatures, the formation of Pt3O4 becomes more favorable, while the PtO phase has been observed not to be stable in any interval of temperature and pressure. The β-PtO2 phase becomes stable only at oxygen partial pressures of several hundred atmospheres. Similar results were reported by Punnoose and co-workers, who examined the oxidation state of supported Pt clusters in air at increasing temperatures.24 These authors found the occurrence of two-phase transitions near 910 and 1070 K, and related them to the transitions R-PtO2 f Pt3O4 f Pt. Finally, nanoscopic platinum clusters were observed to fully oxidize after 2 h of treatment at 770 K at nearly atmospheric air pressures, and a strong dependence of the oxidation state on cluster size has been established.25 The structure and composition of platinum oxide surface layers have been the subject of recent experimental and theoretical investigations. Tang and co-workers26 have produced a phase diagram of oxygen adsorbed on Pt(111) up to half monolayer coverage by using first-principles methods. Li et al.5 investigated the initial oxide formation on Pt(110) by using a combination of experimental and first-principles techniques, and suggested the phase formed to be PtO2. Under slightly different reaction conditions, the same surface was observed to be covered either by an incommensurate overlayer of R-PtO2 or by a carbonate-stabilized commensurate oxide film that is only stable in the simultaneous presence of O2 and CO. Li and Hammer6 proposed that R-PtO2 submonolayers could form on Pt(111) and showed with DFT calculations that three-phase boundaries involving these structures could lower the barrier for CO oxidation in comparison to a metallic surface. In a recent work,19 the formation of one-dimensional PtO2 at Pt(111)-type steps between Pt(111) terraces on the Pt(332) surface has been detected and related to increased reactivity of the surface toward CO oxidation. Moreover, the adsorption of CO on β-PtO2 has been recently investigated by Gong et al.8 It is interesting to note that the formation of superficial oxide phases has been observed to enhance the reactivity of the Pt(110) surface toward CO oxidation, while the same reaction has been observed to be impeded after initial formation of a surface oxide on Pt(100).27 In general, the structural details and reactivity of oxide films
J. Phys. Chem. B, Vol. 110, No. 30, 2006 14861 formed on Pt(100), a surface that is known to be normally present in platinum clusters,28 have not been investigated to date. In the following, after a brief description of the computational technique employed (Section 2), we will calculate the relative stabilities of bulk oxides (Section 3.1), their surfaces (Section 3.2), and thin oxide films on the Pt(111) and Pt(100) surfaces (Section 3.3). Moreover, we will compute the enthalpic driving force for methane dissociation on a variety of superficial oxide phases (Section 3.4) and study the oxidation of CO molecules to CO2 (Section 3.5). A discussion of the results and a summary of the main findings will conclude the paper. 2. Computational Methods Electronic and structural optimizations as well as FPMD simulations29 are performed by using the Car-Parrinello (CP) method30 and Density Functional Theory (DFT) within the PW91 Generalized Gradient Approximation (GGA),31 as implemented in the LAUTREC code.32 We use periodic boundary conditions and a plane-wave basis set with a kinetic energy cutoff of 50 Ry (increased up to 70 Ry to test the convergence of energy differences in selected systems), and separable normconserving pseudopotentials which were intensively tested and used in previous works.33-35 Total energies have been calculated within a spin-paired approach. Test calculations have been carried out including spin-polarization to compute the dissociative adsorption energy of methane on the Pt3O4 surface. We found that the maximum difference between the values obtained in spin-paired vs spin-polarized calculations is 50 meV, which does not affect our conclusions. All systems have been fully relaxed. Special k-point sets for integration in the first Brillouin zone are produced with the method of Monkhorst and Pack36 and tested for convergence in all cases. The Pt(111) and Pt(100) surfaces have been modeled with a slab of five atomic layers separated by a corresponding vacuum layer in the z direction, and periodically repeated in the x and y directions. Using a (2 × 2) surface cell with a 3 × 3 × 1 k-point mesh we obtain a surface energy of 103 meV/Å2 for Pt(111), which is consistent with other DFT-GGA results (e.g., a value of 97 meV/ Å2 has been reported in ref 37). The considered oxide surface slabs were four to five layers thick, separated by a corresponding vacuum layer in the z direction, and periodically repeated in the x and y directions. k-point meshes with density in reciprocal space corresponding to those of the metallic surfaces have been employed. Phonon calculations are performed within the Density Functional Perturbation Theory38 as implemented in the ABINIT package.39,40 When needed, the GGA+U method41,42 has been employed as implemented in the Quantum-ESPRESSO package,43 using ultrasoft pseudopotentials with a cutoff energy of 40 Ry. 2.1. Ab Initio Atomistic Thermodynamics. Gibbs’ free energies of formation are calculated by combining DFTcalculated free energies with data for enthalpy and entropy of the oxygen gas taken from thermodynamic tables,44 following the scheme originally proposed by Scheffler and Reuter.45 In particular, the Gibbs’ free energy of formation of a bulk PtOx phase is calculated as the difference:
x gas f bulk ∆GPtO (T,p) ) gPtO (T,p) - gbulk Pt (T,p) - gO2 (T,p) (1) x x 2 where gOgas2 is the free energy of a dioxygen molecule in the gas phase and gbulk can be written as gbulk ) flatt + fvibr + pV. In this expression flatt is the calculated total energy per Pt atom of the ideal crystal (in the case of Pt oxides, it is the total energy
14862 J. Phys. Chem. B, Vol. 110, No. 30, 2006
Figure 1. Crystal structure of platinum oxides: (a) PtO; (b) Pt3O4; (c) R-PtO2; and (d) β-PtO2.
per PtOx formula unit), the vibrational free energy fvibr can be obtained analytically within harmonic approximations from the calculated phonon dispersion curves,46 and the pV term can be neglected in solid phases. Calculation of the phonon contribution fvibr to the free energy is computationally quite demanding. To estimate the importance of this term, we have calculated the full phonon spectra for R-PtO2 and Pt, and found that the equilibrium point between the two phases at a partial oxygen pressure of 0.4 atm is shifted from 936 to 966 K when the fvibr term is taken into account. Since such a shift is within the error bar associated with the other approximations inherent in the DFT implementation used,16 the reported phase diagrams have been calculated without taking into account phonon contributions, following ref 45. The surface free energy γ relative to a surface of a PtOx phase for a slab model containing NPt platinum atoms and NO oxygen atoms is calculated as follows:45
γPtOx(T,p) )
1 slab bulk [G (T,p,NPt,NO) - NPtgPtO (T,p) + x 2A (xNPt - NO)µO(T,p)] (2)
bulk where Gslab is the calculated free energy of the slab used, gPtO x is the free energy of the bulk oxide, µO is the chemical potential of oxygen, and A is the area of the slab in the xy plane. In the case of thin oxide films on a metallic surface, we have approxbulk imated gPtO with the free energy of the bulk metal (x ∼ 0). x
3. Results 3.1. Bulk Platinum Oxides. A comprehensive listing of commonly available platinum oxide phases, along with their temperature and pressure stabilities in the 400-900 °C temperature and 20-3500 atm pressure ranges, has been reported in ref 22. These comprise PtO, Pt3O4, R-PtO2, and β-PtO2.22,23 We start our study with total energy calculations and structural relaxation of all these bulk structures (Figure 1). The optimized lattice parameters, bulk moduli, and enthalpies of formation are reported in Table 1. r-PtO2 crystallizes in the hexagonal CdI2 structure and presents a laminar structure consisting of stacked O-Pt-O trilayers which are covalently bonded in plane, while the interactions between trilayers are of van der Waals type.6,22 While a good agreement between the experimental and computed edge of the hexagonal cell could be achieved (3.10 vs 3.14 Å), the GGA approximation heavily underestimates the interlayer forces,48 similarly as in the case of graphite.49 The
Seriani et al. energy versus layer separation curve presents a very shallow minimum (37 meV per Pt atom) at a c/a ratio of 1.85 (the reported experimental value is about 1.4 [ref 22]). As a comparison, at the LDA level we obtain a c/a value of 1.29 and a larger interlayer binding energy (49 meV per Pt atom). The existence of weak interlayer bonding is consistent with other theoretical investigations6 and a number of experimental observations, such as poor crystallization in the c-axis direction.22 β-PtO2 crystallizes in an orthorhombic CaCl2 structure, which is closely related to the rutile structure, differing from the latter only through a rotation of the oxygen pairs around the central Pt atom. This is quantified by the rotation angle or alternatively by the reduced coordinates x and y of the oxygen atoms. We calculate x ) 0.26 and y ) 0.35, which corresponds to a rotation angle of 8.5° (to be compared with the experimental value of 7.7°,50 and with the value of 9.4° computed within DFT in ref 51). Contrary to the experimental evidence, but in agreement with previous DFT calculations,5 R-PtO2 is less stable than β-PtO2 by about 10 meV/Pt atom. This may either be due to DFT approximations such as the exchange-correlation functional and the pseudopotential employed or, most probably, to the underestimation of the interlayer binding energy at the GGA level. For Pt3O4 there are two proposed crystal structures: body centered cubic (BCC)52 or simple cubic (SC).22 In the case of the BCC structure, we compute an optimized lattice constant of 5.49 Å, which is much smaller than the value of 6.23 Å originally proposed to explain the X-ray data. Moreover, the computed formation enthalpy is 2.45 eV higher than the minimum energy corresponding to the SC structure, for which the agreement between experimentally proposed and calculated lattice parameter is much better (5.59 Å vs 5.65 Å). The electronic structure of SC Pt3O4 at the DFT-GGA level presents no band gap, i.e., the oxide appears to have metallic character. To check the consistency of this result, we have performed calculations at the GGA+U level, increasing the U parameter up to 15 eV. However, no opening of a band gap in the electronic structure has been observed, even though the occupied d-bands are shifted to considerably lower energy. The structure of PtO (tetragonal, PtS structure) was described for the first time by Moore and Pauling in ref 23. Different from the case of palladium, where PdO is the most stable oxide stoichiometry,53 the formation enthalpy of PtO is found to be higher than all other oxide phases (Table 1). The calculated band structure displays metallic character, even though PtO is believed to have a gap larger than 0.7 eV.54 Notably, in this case the absence of band gap could be corrected at the GGA+U level. Band gap values of 0.38 and 0.83 eV are obtained by using values of 9 and 15 eV for the U parameter.55 The variation of free energy of formation of the different oxide phases with temperature and oxygen partial pressure now can be computed as described in the Methods section (eq 1). A graph of free energy versus temperature relative to an oxygen pressure of 1 atm is reported in Figure 2. Consistently with the available experimental literature,22,24 at low temperatures up to about 870 K the thermodynamically stable phase is PtO2. Metallic platinum is thermodynamically stable at temperatures higher than about 970 K. In agreement with the finding of Muller,22 there is a temperature range of thermodynamic stability of an oxide phase with intermediate stoichiometry, namely Pt3O4. In our calculations, this temperature range is about 100 K wide. We note, however, that the real stability interval can be considerably affected by the approximations inherent in the
Catalytic Oxidation Activity of Platinum Particles
J. Phys. Chem. B, Vol. 110, No. 30, 2006 14863
TABLE 1: Calculated Structural Properties, Bulk Moduli, and Enthalpies of Formation (in eV) of Platinum Oxide Phases with Experimental Data in Parentheses structure a (Å) b/a c/a Pt-O (Å) Pt-Pt (Å) B (GPa) ∆Hf(0,0)f ∆Hf(0,0)g a
PtOa
Pt3O4b
R-PtO2b
β-PtO2c
PdS, tetragonal 3.10 (3.08) 1.745 (1.735) 2.06 (2.04) 3.10 (3.08) 368 -0.76 -0.79
simple cubic 5.65 (5.59) 2.00 (1.97) 2.83 (2.79) 262 -1.42 -1.38
CdI2, hexagonal 3.14 (3.10) 1.85 (1.38-1.42) 2.05 3.14 (3.10) -2.05 -1.94
CaCl2, orthorombic 4.49 (4.48) 1.05 (1.01) 0.70 (0.70) 2.03 (1.99) 3.15 (3.14) 246 (265)d -2.06 -1.93
Reference 47. b Reference 22. c Reference 50. d Calculated with DFT, ref 51. e Norm-conserving pseudopotentials. f Ultrasoft pseudopotentials.
Figure 3. Ball-and-stick models of platinum oxide surfaces: (a) Pt3O4(100) (O-terminated); (b) Pt3O4(100) (Pt-terminated); (c) Pt3O4(110) (Pt-terminated); (d) Pt3O4(110) (O-terminated); (e) PtO(100) (O terminated); and (f) PtO(101).
TABLE 2: Calculated Surface Free Energies of Pt, PtO (O-terminated), Pt3O4 (O-terminated and Pt-terminated), and r-PtO2 (See Figure 3)
Figure 2. (a) Calculated Gibb’s free energy of formation of R-PtO2 (red line), Pt3O4 (green line), and PtO (blue line) with respect to Pt (black line) at 1 atm of oxygen pressure, as a function of temperature. (b) Phase diagram showing the temperature and pressure ranges of stabilities of bulk platinum oxides.
computational method,16 so the calculated transition temperatures can be considered as qualitative values only. Within these approximations, our calculated data appear to be in reasonable agreement with the results of Punnoose et al.,24 where phase transitions attributed to the PtO2 f Pt3O4 and Pt3O4 f Pt transition have been reported to take place at 910 and 1070 K for fully oxidized supported clusters subjected to thermal annealing. 3.2. Surfaces. Since we are interested in the catalytic activity of oxidized platinum, it is important to determine the thermodynamic stability of exposed surfaces. We therefore calculate the surface free energy (see Methods section, eq 2) of a number of low-index surfaces of the metallic and oxide phases using symmetric slab models (i.e., models presenting equal composition of the top and bottom surfaces of the slab). All results are reported in Table 2, while the structure of selected surfaces is shown in Figure 3. Common features among the different oxide
surface
γ (meV/Å2)
surface
γ (meV/Å2)
Pt(111) Pt(100) PtO(100)-O PtO(101) R-PtO2(0001)
103 131 -31 56 2
Pt3O4(100)-O Pt3O4(100)-Pt Pt3O4(110)-O Pt3O4(110)-Pt
27 128 77 187
surfaces are their lower surface free energy compared with bare metallic surfaces, and their preference for oxygen-rich terminations at low temperatures. In the case of metallic platinum we consider here the two lowest energy surfaces, namely (111) and (100), which correspond to the common termination of metallic nanoparticles.56 In the case of R-PtO2, we consider only the (0001) surface to be relevant, given the layered, graphite-like structure of such a phase. In the case of Pt3O4 we have considered the (100) and (110) surfaces and different terminations (oxygen-rich and oxygen-poor). The (100) oxygen-rich termination displays square-planar arrangements of oxygen atoms, half of which are occupied by a platinum atom placed in the center of the square (Figure 3a). Therefore, each oxygen atom of this surface is only bound to two Pt atoms, while the oxygen atoms of the bulk are bound to three Pt atoms. Despite the presence of uncoordinated atoms, this surface presents a very low surface energy, as shown in Table 2. As far as PtO is concerned, even if this is a metastable phase, we cannot exclude a priori the formation of thin PtO films as a result of platinum oxidation processes. We have thus taken into account the O-terminated PtO(100) and PtO(101) surfaces. Interestingly, the former system is characterized by a negative surface energy (Table 2), which is consistent with the thermo-
14864 J. Phys. Chem. B, Vol. 110, No. 30, 2006
Seriani et al.
Figure 5. Models of oxide layers on Pt surfaces: (a) R-PtO2 on (2 × 2) Pt(111)sthe line delimitates the elementary surface cell, the squares indicate the positions of the Pt atoms of the topmost metallic layer, the circles indicate the positions of the Pt atoms in the oxide monolayer; (b) R-PtO2 on (1 × 1) Pt(111)ssymbols as in part a; (c) PtO-like monolayer on Pt(100); (d) PtO-like monolayer on Pt(100) with further adsorbed oxygen; (e) Pt3O4-like monolayer on Pt(100); (f) x5 × x5R27° oxide monolayer on Pt(100); (g) a Pt3O4 film corresponding to 3 ML on Pt(100); and (h) a PtO film corresponding to 3 ML on Pt(100).
Figure 4. Calculated surface free energies of oxide systems on Pt surfaces at 1 atm of oxygen pressure as a function of temperature. The vertical black lines delimitate the stability regions of bulk oxides. (a) On Pt(111): R-PtO2 ML on (2 × 2) Pt(111) (dark green); coherent (1 × 1) R-PtO2 ML (orange); 1 ML O (red); 0.75 ML O (blue); 0.50 ML O (light green); and 0.25 ML O (black, growing). The horizontal black line represents the surface energy of clean Pt(111). (b) On Pt(100): PtO-like ML (red); PtO-like ML with adsorbed oxygen (brown); 1 ML O (light green); 0.5 ML O (blue); x5 × x5R27° (orange); and Pt3O4like ML (dark green). The horizontal black line represents the surface energy of clean Pt(100). See also Figure 5.
dynamic instability of the PtO phase toward higher oxidation states. The structure of PtO(100) resembles the structure of the oxygen-rich Pt3O4(100) surface (see Figure 3), with the important difference that all square-planar motifs formed by oxygen atoms coordinate a Pt atom in their center. Therefore, all oxygen atoms of the surface are 3-fold coordinated, which will have important implications in the different chemical activity of the two surfaces, as will be shown later in this paper. 3.3. Thin Oxide Films. While bulk oxide phases are thermodynamically stable, it is common knowledge that platinum does not undergo spontaneous oxidation at low temperatures. However, formation of thin films of superficial oxide, which protect the bulk metal from further oxidation, is in principle a feasible process. We therefore investigate a variety of oxide structures on the Pt(111) and Pt(001) surfaces up to a coverage of 1 ML of oxygen, paying particular attention to possible coherent interfaces, and considering structures proposed in the existing literature for the oxidation of palladium20 and platinum6 surfaces. The surface systems are modeled with a slab comprising 5 Pt layers, with equal oxide composition on both the top and the bottom sides of the slab,58 and the free energy is calculated as described in the Methods section (eq 2). The results are described below, and diagrams of the thermodynamic stability of the considered phases are reported in Figure 4.
A schematic representation of relevant thin film structures is shown in Figure 5. Pt(111). We start our investigation with the adsorption of oxygen atoms in fcc hollow sites of the (111) surface. By using a (2 × 2) surface unit cell, the oxygen coverage can be varied from 0.25 to 1.0 ML. Consistently with experimental measurements,59 the binding energy per atom of oxygen chemisorbed on the surface decreases significantly with increasing oxygen coverage. Namely, the calculated values of adsorption energy are 1.62, 1.38, 1.01, and 0.59 eV/O atom for coverages of 0.25, 0.50, 0.75, and 1.0 ML, respectively. Interestingly, increasing the oxygen coverage from 0.75 to 1.0 ML is associated with a loss in free energy. Indeed, at 1 atm of partial oxygen pressure a complete monolayer of adsorbed oxygen is never the thermodynamically favorable state in any temperature interval (Figure 4a), in agreement with the result of a recent experimental study.59 Thus, 0.75 ML is the energetically most favorable coverage up to about 500 K, where the interval of stability of 0.5 ML begins. This is in reasonable agreement with the lowest temperature peak of thermal desorption experiments performed in ref 59, which is centered at 560 K. Increasing the temperature is expected to lead to further oxygen desorption (0.25 ML is the stable coverage at T higher than about 850 K), until the bare metallic surface becomes the thermodynamically stable phase at about 1400 K. We can thus expect that coverages larger than 0.75 ML can only be obtained by overcoming the enthalpy barrier toward insertion of oxygen atoms under the surface, with formation of a thin oxide layer. For instance, the formation of R-PtO2 thin films has been suggested to take place on Pt(111) (see ref 6 and references therein) and XPS spectra peaks very close to those of PtO2 have been measured for the Pt 4f5/2 states on a Pt(111) surface with an oxygen coverage of 2.6 ML.59 Such a structure can be the outcome of the oxidation of the first layer of the metallic surface, when the resulting oxide layer relaxes over the edge of surface steps. Indeed, an R-PtO2 layer can be obtained from a (111) atomic layer of metallic platinum upon adsorption of one oxygen atom above and one oxygen atom below the plane for each Pt atom. However, the Pt-Pt distance in R-PtO2 is 3.14 Å, much larger than the Pt-Pt distance in the
Catalytic Oxidation Activity of Platinum Particles Pt(111) plane (2.82 Å). Hence, the formation of a (1 × 1) commensurate Pt(111)/R-PtO2(0001) interface is associated with a very large compressive strain of the oxide layer, which results in thermodynamic instability of this structure (Figure 4a, orange curve). A nearly unstrained R-PtO2 layer can be accommodated on a (2 × 2) Pt(111) surface cell after rotation of the oxide layer of 30° with respect to the platinum surface, i.e., placing the edge of the hexagonal cell of R-PtO2 along the long diagonal of the rhomboedral cell of the (111) surface. In this configuration, the strain in the oxide layer is only about 3.5%. As can be seen in Figure 4 (dark green line), this oxide layer is the most stable phase up to about 852 K, which nicely corresponds with the interval of stability of bulk PtO2 and is compatible with the 750 K peak of thermal desorption spectra experimentally measured under ultrahigh vacuum for an oxygen coverage of 2.9 ML on Pt(111).59 Pt(100). Oxygen adsorption on the Pt(100) surface is investigated by using a symmetric slab with five Pt layers. On the unreconstructed surface, oxygen atoms adsorb preferentially on bridge positions.60 We have studied the relative stabilities of 0.5 and 1.0 ML of adsorbed oxygen at increasing temperature, and find that the two free energy curves cross at a temperature of about 772 K. Moreover, a PtO-like monolayer, which can be obtained by “lifting and shifting” every second row of Pt atoms to obtain the square-planar coordination of Pt atoms between four oxygen atoms such as in the PtO(100) surface described in the previous section, is found to be more stable than a layer of adsorbed oxygen (Figure 4b, red curve). A transition to the unreconstructed surface with 0.5 ML of chemisorbed oxygen is at around 950 K. We note that the difference between lattice parameters of Pt and PtO (see Table 1) implies that this layer is under considerable compressive stress. Namely, the layer is compressed by 9% in the direction of the Pt rows and strained by 4% in the other direction with respect to the bulk PtO(100) surface. This might lead to a driving force for formation of a Pt3O4-like monolayer upon lifting each second Pt atom of the square-planar motifs into a bridge position between oxygen atoms (Figure 5e). Indeed, the lattice parameter of the cubic cell of Pt3O4 (5.65 Å) is almost exactly twice the Pt-Pt distance in the primitive cubic cell of platinum (2.82 Å). This means that a coherent interface can form between the Pt(100) and the Pt3O4(100) surfaces. However, at a 1 ML coverage, due to the evident undercoordination of Pt atoms, this structure is highly unfavored with respect to the PtO-like layer. We have also considered the formation of a thin x5 × x5R27° surface oxide (Figure 5f), which has been found to be a very stable structure on Pd(100).20 This structure can be seen as a single adsorbed PtO(101) layer (cf. Figure 3) terminated by oxygen. Contrary to the case of Pd, however, in the case of Pt this oxide film is not thermodynamically stable in any range of temperatures (Figure 4b, orange curve). It is important to note that none of the thin oxide films considered is stable outside the stability range of bulk oxides. This indicates that the formation of oxides at low temperature is hindered by kinetic effects. On the other hand, at increasing temperature, when diffusion processes become more and more important, the formation of thicker oxide layers becomes more and more favorable, due to the large thermodynamic driving force associated with oxide formation. In particular on the “open” (100) surface, atomic rearrangements could easily lead to the formation of an oxide layer thicker than 1 ML. Assuming formation of stoichiometric oxides and of a coherent interface, candidate phases are PtO and Pt3O4. Therefore, we have compared the stability of thicker layers of PtO and Pt3O4
J. Phys. Chem. B, Vol. 110, No. 30, 2006 14865 corresponding to an oxygen coverage of 3 ML. At this coverage and at the oxygen pressure of 1 atm we have found that the Pt3O4-like structure is more stable than the PtO-like structure up to about 1250 K. Hence, while a PtO stoichiometry can remain stable up to a coverage of about 1 ML, as the oxide film becomes thicker, the formation of a thin Pt3O4 layer, which can be accommodated on Pt(100) coherently and stress-free, is strongly favored. 3.4. Dissociative Adsorption of Methane. In the previous sections we have addressed the problem of calculating the relative thermodynamic stabilities of platinum oxide bulk, surface, and thin film phases. We would like now to correlate the existence of such oxide phases with their catalytic oxidation reactivity. To address this issue, we compute in this section the thermodynamic driving force for the dissociative adsorption of CH4 on relevant surfaces, to obtain adsorbed CH3 and H species. A comprehensive analysis of methane dissociation would ideally involve the calculation of the barrier for the C-H bond activation. This, on the other hand, depends on vibrational and translational degrees of freedom of the CH4 molecules in a highly nontrivial manner,61,62 and is well-known to involve structural defects such as, e.g., steps and adatoms.63 Therefore, we consider here the thermodynamic driving force for the dissociative adsorption to be a qualitative indication of the reactivity of the surfaces toward methane dissociation. This corresponds to an implicit assumption that the probability of finding a defect on which the activation of methane takes place be equal on all considered structures, and that the height of the enthalpy barrier for the dissociation event be roughly proportional to the net enthalpy gain. Indeed, this has been demonstrated to be true in the case of methane adsorption on a number of metallic surfaces.64,65 The enthalpic gain for dissociative adsorption of methane on a slab model can be calculated as follows:
∆E ) ECH3 + EH - 2Esur - Emeth
(3)
where ECH3 and EH are the calculated energies of separate surface slab models with the CH3 and H species adsorbed on one side, Esur is the energy of the clean surface model, and Emeth is the energy of the isolated methane molecule. A (2 × 2) surface cell was used for all modeled systems. All values of ∆E obtained for selected surfaces are reported in Table 3. The dissociative adsorption reaction is found to be more exothermic on Pt(100) than on Pt(111), consistent with the higher surface energy, and thus higher reactivity, of the (100) surface. On the (100) surface, CH3 adsorbs preferentially on a bridge position, where the carbon atom binds to a Pt atom and one of the hydrogen atoms is attracted toward a second Pt atom, while the corresponding C-H bond is stretched by 0.08 Å with respect to the initial length of 1.10 Å in CH4. When the (100) surface begins to oxidize, i.e., after the formation of a thin PtOlike layer of oxide corresponding to a coverage of 1 ML of oxygen, the dissociative reaction becomes endothermic (Table 3). This strongly indicates that the formation of an ultrathin oxide layer on the (100) surface results in a less reactive system, which makes thermodynamically unfavorable the dissociative methane adsorption. However, after further oxidation and formation of a thin Pt3O4 film at a coverage of 3 ML of adsorbed oxygen, the exposed Pt3O4 surface suddenly becomes very reactive toward methane dissociation. The calculated driving force for dissociative adsorption is 1.87 eV, much higher than that on all other systems considered, despite the very low surface free energy of this surface (Table 2).66
14866 J. Phys. Chem. B, Vol. 110, No. 30, 2006
Seriani et al.
TABLE 3: Calculated Enthalpy Differences (∆E) of the Dissociative Adsorption of Methane and Adsorption Sites of the CH3 and H Species on Various Pt and Pt Oxides Surfaces system
CH3
H
∆E (eV)
Pt(111) Pt(111) Pt(100) Pt(100)
fcc top top bridge
top top top top
0.59 0.03 -0.01 -0.07
PtO-ML PtO-ML
on Pt on O
on Pt on Pt
0.74 0.89
Pt3O4(100) Pt3O4(100) Pt3O4-ML
on Pt on O on O
on Pt on Pt on O
1.56 0.18 -1.47
R-PtO2(0001)
on O
on O
2.03
The strongly exothermic character of the dissociation reaction on this surface does not depend on the particular thickness of the formed oxide. Indeed, we calculate a very similar driving force for methane dissociation of 1.84 eV on the surface of bulk Pt3O4, and of 1.47 eV on the Pt3O4-like monolayer on the Pt(100) surface (see Figure 5e). The latter structure, despite being metastable with respect to the PtO-like monolayer, presents an interesting feature, namely under-coordinated Pt atoms bridging O atoms. These sites can be considered as native defects of the PtO-like ML structure, and are the natural Pt termination of the Pt3O4(100) structure (see Figure 3b). The bridging Pt atoms may represent active sites for the initial methane activation. In fact, in a FPMD simulation where a CH4 molecule approaches one of these sites with an initial kinetic energy of 1 eV, spontaneous methane dissociation is observed to take place (Figure 6). We note that, in a number of analogous simulations performed at the same impinging energy of 1 eV, dissociation has not been observed on regular metallic or oxide surfaces, but only on adatom systems (e.g., on a Pt adatom on a Pt(111) surface). Interestingly, the most stable sites for adsorption of CH3 and H are oxygen surface atoms. Thus, methane dissociation on Pt3O4 surface or thin films may take place with direct formation of a C-O bond. However, after the adsorption we do not observe spontaneous break of additional C-H bonds, or desorption of reaction intermediates such as, e.g., formaldehyde. This implies that the further oxidation of the adsorbed group is also associated with a nonnegligible enthalpic barrier. As far as the R-PtO2(0001) surface is concerned, we found that dissociative adsorption is largely endothermic (by 2.03 eV), indicating that methane dissociation is strongly unfavored on this oxide phase. The peculiar stability of the single (0001) layers is the result of complete saturation of all Pt and O atoms composing the layer, so that adsorption becomes unfavorable. These striking differences in the chemistry of the oxide phases preferentially formed on the (100) and (111) surfaces may suggest a dependence of the catalytic activity toward hydro-
Figure 6. Snapshots from a FPMD simulation in which a CH4 molecule with an initial kinetic energy of 1 eV dissociates over a Pt atom of the Pt-terminated Pt3O4(100) surface: (a) initial configuration; (b) after 53 fs of simulated time; and (c) after 65 fs of simulated time.
CH3
H
∆E (eV)
Pt(111) Pt(111) Pt(100) Pt(100)
fcc top top bridge
fcc fcc bridge bridge
0.44 -0.13 -0.38 -0.44
PtO-ML PtO-ML
on Pt on O
on O on O
0.02 0.18
Pt3O4(100) Pt3O4(100) Pt3O4-3ML
on Pt on O on O
on O on O on O
-0.46 -1.84 -1.87
system
carbon oxidation on the crystallographic direction as well as on the reaction temperature. However, we have to note that the presence of defects, not considered in this work, is expected to have a profound effect on the reactivities of the surfaces, and will need further investigation. 3.5. Oxidation of CO. To get a more complete picture of the catalytic oxidation properties of platinum oxide phases we consider in this section the process of CO adsorption on the metallic (111) surface, on the R-PtO2(0001) surface, and on the Pt3O4(100) surface. Adsorption on Pt(111) is associated with an energy gain of about 1.9 eV. In contrast with the experimental evidence, but in agreement with previous DFT calculations,67 preferential adsorption takes place on fcc hollow sites. The reason for this wrong prediction lies in the too low π-levels of the CO molecule as calculated in pure DFT approaches, which leads to errors in the adsorption energy.68,67 However, the differences among the different adsorption sites on the Pt(111) surface are within 0.1 eV, i.e., much smaller than the absolute value of binding energy.69 We can therefore assume here that a comparison among adsorption energy values between different phases is meaningful when the energy differences are larger than a few tenths of 1 eV. However, no simple adsorption of CO molecules is observed on the R-PtO2(0001) or the Pt3O4(100) surfaces. Several structural minimization attempts of the CO molecule near the R-PtO2(0001) surface led to spontaneous desorption of CO. This result, combined with the endothermic methane dissociation reaction described in the previous section, confirms that this oxide structure, in the absence of defects, is rather inert as far as oxidation reactions of combustion exhausts are concerned. In contrast, Pt3O4 is once again revealed to be extremely active toward catalytic oxidation. Namely, several attempts of structural minimization of CO molecules adsorbed on the Pt3O4(100) surface resulted in spontaneous desorption of CO2 molecules. In this reaction, CO binds strongly to one of the four undercoordinated oxygen atoms of the surface, while the covalent bonds between this O atom and the underlying Pt atoms break. As a result, a linear CO2 molecule is expelled in the gas phase (Figure 7). Since in our simulations the motion of the atomic nuclei was damped with the intention of reaching a minimum of the potential energy surface, we can conclude that the ratelimiting step for CO oxidation on this surface is the approach of a CO molecule from the gas phase and the formation of an initial bond with the surface. To compute the enthalpy barrier associated with this process, a series of total energy calculations have been performed decreasing the distance between the C atom of the CO molecule and an O atom of the surface. At every step the geometry of the whole system is fully minimized keeping fixed the CCO-O surf distance. The results are reported in Figure 7, and show that the maximum enthalpy barrier for
Catalytic Oxidation Activity of Platinum Particles
Figure 7. Calculated enthalpy barrier for the adsorption and oxidation of a CO molecule on the Pt3O4(100) surface. Enthalpy values E are calculated at fixed distances d between the C atom of CO and a surface O atom. The initial configuration (a), the transition state (b), and the final configuration (c) of the simulated system are shown in the balland-stick models, in which only the topmost atomic layers are represented. The zero enthaply value corresponds to d ) 5 Å. The total driving force for the reaction is 1.49 eV, and the maximum enthalpy is 0.34 eV.
CO binding to the surface and formation of CO2 is as low as 0.34 eV. This is in particular a much lower value than the typical barriers of more than about 0.7 eV which characterize the Langmuir-Hinshelwood reaction of CO oxidation on clean Pt surfaces.70,71 4. Discussion The catalytically active platinum particles of automobile catalysts are known to undergo partial oxidation under oxygenrich conditions (see, e.g., ref 9 and references therein). Catalytic oxidation of hydrocarbons and carbon monoxide is therefore thought to take place according to a Mars-van Krevelen mechanism, where the reactants extract oxygen atoms from a superficial oxide layer.11 Indeed, the formation of superficial oxide has been associated with a sudden increase of the catalytic oxidation activity of the Pt(110) surface.4 To get an insight into possible oxide phases present on platinum particles which contribute to thir oxidation activity, we have calculated the relative thermodynamic stabilities of relevant oxide structures formed on the (100) and (111) Pt surfaces, and addressed the driving force for the dissociative adsorption of methane and the reaction with CO. 4.1. Formation of Platinum Oxide Phases. Metallic platinum is a noble metal since its oxidation at low temperature is kinetically hindered. However, from a thermodynamic point of view, formation of platinum oxide is a favorable process, and indeed platinum particles of typical sizes of a few nanometers can be fully oxidized upon thermal annealing in air at about 770 K.25 By means of ab initio thermodynamics methods we found that R-PtO2 is the thermodynamically stable phase at atmospheric pressure and at temperatures lower than ∼870 K, above which formation of Pt3O4 becomes favorable. Metallic platinum is stable at temperatures higher than 970 K. These results are consistent with the observation that R-PtO2 starts decomposing in air above 870 K,18 and with measured phase transitions of platinum oxide particles upon increasing temperature, at 910 and 1070 K.24 Because of the peculiar layered structure of the bulk, ultrathin layers of R-PtO2 are extremely stable, and could in principle form on a variety of crystallographic orientations, thus hindering the diffusion of oxygen toward the bulk and the further oxidation of the metal. In particular, we found that a single R-PtO2 layer
J. Phys. Chem. B, Vol. 110, No. 30, 2006 14867 is the thermodynamically stable oxide structure on the Pt(111) surface at oxygen coverages larger than about 0.75 ML. On Pt(100), however, at oxygen coverage of 1 ML, very limited atomic rearrangements are required to obtain a PtO-like structure starting from an adsorbed ML of O atoms: namely every second Pt row must be lifted and shifted to obtain this oxide monolayer. At the monolayer level, this is the most stable structure among those investigated, despite the considerable lattice mismatch between PtO and Pt, which leads to substantial compressive stress in the oxide layer. It is to be noted that the ultrathin oxide layers investigated are not thermodynamically favored with respect to the formation of bulk oxides (see Figure 2). Therefore, at increasing temperature, when the hindrance for the diffusion of oxygen atoms is expected to decrease, oxide growth is expected to take place. In particular, on the (100) surface the growth of Pt3O4 can take place without the building up of any stress at the metal/oxide interface, and is thermodynamically favored. The oxidation of Pt(001) may thus proceed via adsorption of oxygen, followed by the initial formation of an ultrathin PtO-like layer, followed eventually by the growth of a thicker Pt3O4 layer. The fact that Pt3O4 grows coherently on Pt(001) suggests that the kinetic hindrance to its formation might be reduced, as opposed to the formation of PtO2 on Pt(111), which requires the diffusion of O atoms under the compact (111) planes. Oxide films formed on metallic Pt surfaces after treatment with oxygen plasma have been found to be ∼2.5 nm thick.72 This means that small Pt particles could indeed, under similar circumstances, be easily oxidized entirely to bulk oxide. In general, the formation of oxide starting from small particles is expected to be much easier than that on extended surfaces, because of favorable chemical and mechanical effects due to the presence of steps and edges, and to the release of stresses possibly present at the metal/oxide interface.73 The presence of Pt3O4 on the Pt(001) surface is consistent with the formation of Pt3O4 starting from clusters larger than about 2-3 nm,25 which tend to grow along the principal directions of the cubic primitive cell. On the other hand, smaller clusters are preferentially of tetrahedral shapes, exposing mainly the (111) plane, so that the formation of R-PtO2 is expected to be dominant. In fact, oxidation of clusters of 1 nm size has been observed to lead preferentially to PtO2.25 4.2. Catalytic Activity of Oxide Surfaces. We addressed the catalytic activity of oxide phases formed on platinum surfaces considering the driving force for the dissociative adsorption of methane and oxidation of CO molecules. As mentioned in Section 3.4, calculation of the barrier for the activation of a C-H bond goes beyond the scope of this work. We therefore consider here the driving force for methane dissociation as an indication of the reactivity of a given surface toward the initial oxidation reaction. We have found that the dissociation of methane is slightly exothermic on clean (111) and (100) metallic surfaces. When either an R-PtO2 layer on Pt(111) or a PtO-like ultrathin layer on Pt(100) start forming after reaction with oxygen molecules, then the driving force for methane oxidation is found to decrease. Namely, we found that methane dissociation on both surfaces is endothermic, and in particular the defect-free R-PtO2(0001) surface is expected to be completely inactive due to the high stability of the PtO2 layers. As a consequence, the formation of ultrathin oxide films at oxygen coverages of about 1 ML is detrimental to dissociative adsorption of methane on both Pt(111) and Pt(100). Notably, R-PtO2(0001) is also found to be inactive toward adsorption and oxidation of CO. Possible
14868 J. Phys. Chem. B, Vol. 110, No. 30, 2006 catalytic activity of this phase must therefore be confined to defects (e.g., oxygen vacancies) and interface regions.6 These results are consistent with a number of previous experimental findings. Namely, Ge´lin et al.9 reported that a dispersed PtO2 phase is less reactive toward methane oxidation. Yazawa and co-workers74 found a correlation between low activity of platinum for propane oxidation and the presence of PtO2. On the other hand, PtO2 is known to act as a potent catalyst for hydrosilylation and oxidation of ethanol in the liquid phase.75,76 It is assumed that the oxidation of the substrates causes a reduction of the oxide to metallic platinum according to mechanisms which are not identified to date, and could possibly involve the presence of oxygen vacancies.76 At increasing thickness of the oxide layer, we expect facile formation of Pt3O4 on the Pt(100) surface since a coherent, stress-free film can form during oxide growth. This is expected to be associated with a sudden increase in the reactivity toward methane oxidation, as is suggested by the very large energy gain (∼1.9 eV) upon dissociation on the Pt3O4(100) surface. The Pt3O4(100) surface is also found to promote oxidation of CO molecules to CO2 according to a Mars-van Krevelen mechanism. Our results show that a barrier as low as 0.34 eV needs to be overcome for a CO molecule to adsorb on an oxygen atom of this surface from the gas phase. In FPMD simulations, after the initial formation of a bond between the C atom of CO and an O atom of the surface, the bonds between this atom and the underlying Pt atoms are observed to break and a CO2 molecule is observed to spontaneously desorb from the surface. This leaves oxygen vacancies in the oxide, which are expected to react with molecular oxygen to restore the thermodynamically stable oxygen-terminated surface, and thus its full catalytic activity. 5. Summary We have studied the formation of bulk platinum oxide phases and thin oxide films on the Pt(111) and Pt(100) surfaces by means of ab initio thermodynamics techniques. At atmospheric oxygen pressures, R-PtO2 is found to be the thermodynamically stable phase at low temperatures, while Pt3O4 is stable between 870 and 974 K. Metallic Pt is stable at higher temperatures. On metallic surfaces, at oxygen coverages of about 1 ML, R-PtO2 is the stable structure on the (111) surface, while a PtO-like structure is expected to form on the (100) surface. Both structures are found not to be active toward either dissociative adsorption on methane or oxidation of CO. However, at higher oxygen coverages a thin film of Pt3O4 can grow on Pt(100) with formation of a coherent and stress-free metal/oxide interface, and this phase has been found to present surprisingly high catalytic oxidation activity. Namely, the driving force for the dissociative adsorption of methane is about 1.9 eV, much higher than on any other system considered. Moreover, CO molecules are found to adsorb on O atoms of the Pt3O4(100) surface after overcoming a small enthalpy barrier of 0.34 eV, after which they spontaneously oxidize and desorb in the form of CO2. These results indicate that the formation of Pt3O4 could contribute significantly to the peculiar total catalytic oxidation activity of Pt nanoparticles in gas sensors or automobile catalysts. Acknowledgment. We would like to thank M. Bobeth, A. Ullrich, and A. De Vita for useful discussions, and M. Stengel for his precious work of code development. Computing resources were provided by the Centre for High Performance Computing of the Dresden University of Technology, Germany,
Seriani et al. by the Cambridge-Cranfield High Performance Computing Facility, UK, and the HPCx computing facilities through the UKCP Consortium, UK. N.S. is grateful to the Theory of Condensed Matter Group at the Cavendish Laboratory, Cambridge for its hospitality in August 2003 and October 2004. L.C.C. acknowledges support by the EPSRC, UK (GR/S61263/ 01), and by the Alexander von Humboldt Foundation. This work has been financially supported by the German Federal Ministry of Education and Research (BMBF) within the project BIOKADS. Supporting Information Available: Complete ref 39. This material is available free of charge via the Internet at http:// pubs.acs.org. References and Notes (1) Platinum 2005, Interim Review; Johnson Matthey, 2005. (2) Diesel Cars & Platinum Demands. In Platinum 2002; Johnson Matthey, 2002. (3) Langmuir, I. J. Am. Chem. Soc. 1918, 40, 1361. (4) Hendriksen, B.; Frenken, J. Phys. ReV. Lett. 2002, 89, 046101. (5) Li W. X.; O ¨ sterlund L.; Vestergaard E. K.; Vang R. T.; Matthiesen J.; Pedersen T. M.; Lægsgaard E.; Hammer B.; Besenbacher F. Phys. ReV. Lett. 2004, 93, 146104. (6) Li, W. X.; Hammer, B. Chem. Phys. Lett. 2005, 409, 1. (7) Ackermann, M. D.; Pedersen, T. M.; Hendriksen, B. L. M.; Robach, O.; Bobaru, S. C.; Popa, I.; Quiros, C.; Kim, H.; Hammer, B.; Ferrer, S.; Frenken, J. W. M. Phys. ReV. Lett. 2005, 95, 255505. (8) Gong X.-Q.; Raval R.; Hu P. Phys. ReV. Lett. 2004, 93, 106104. (9) Ge´lin, P.; Primet, M. Appl. Catal., B 2002, 39, 1. (10) Mars, P.; van Krevelen, D. W. Chem. Eng. Sci. 1954, 3, 41. (11) Mallens E. P. J.; Hoebink J. H. B. J.; Marin G. B. Catal. Lett. 1995, 33, 291. (12) Arenz, M.; Mayrhofer, K. J. J.; Stamenkovic, V.; Blizanac, B. B.; Tomoyuki, T.; Ross, P. N.; Markovic, N. M. J. Am. Chem. Soc. 2005, 127, 6819. (13) Birgersson, H.; Eriksson, L.; Boutonnet, M.; Ja¨rås, S. G. Appl. Catal.,B 2004, 54, 193. (14) Reinhard, D.; Hall, B. D.; Berthoud, P.; Valkealahti, S.; Monot, R. Phys. ReV. Lett. 1997, 79, 1459. (15) Graoui, H.; Giorgio, S.; Henry, C. R. Philos. Mag. B 2001, 81, 1649. (16) Reuter, K.; Scheffler, M. Phys. ReV. Lett. 2003, 90, 46103. (17) Hodnett, B. K. Heterogeneous catalytic oxidation; John Wiley & Sons Ltd.: Chichester, UK, 2000; p 227. (18) Rothschild, W. G.; Yao, H. C.; Plummer, H. K., Jr. Langmuir 1986, 2, 588. (19) Wang, J. G.; Li, W. X.; Borg, M.; Gustafson, J.; Mikkelsen, A.; Pedersen, T. M.; Lundgren, E.; Weissenrieder, J.; Klikovits, J.; Schmid, M.; Hammer, B.; Andersen, J. N. Phys. ReV. Lett. 2005, 95, 256102. (20) Lundgren, E.; Gustafson, J.; Mikkelsen, A.; Andersen, J. N.; Stierle, A.; Dosch, H.; Todorova, M.; Rogal, J.; Reuter, K.; Scheffler, M. Phys. ReV. Lett. 2004, 92, 46101. (21) Lundgren, E.; Kresse, G.; Klein, C.; Borg, M.; Andersen, J. N.; De Santis, M.; Gauthier, Y.; Konvicka, C.; Schmid, M.; Varga, P. Phys. ReV. Lett. 2002, 88, 246103. (22) Muller O.; Roy R. J. Less-Common Metals 1968, 16, 129. (23) Moore, W. J., Jr.; Pauling, L. J. Am. Chem. Soc. 1941, 63, 1392. (24) Punnoose, A.; Seehra, M. S.; Wender, I. Fuel Process. Technol. 2001, 74, 33. (25) Wang C.-B.; Lin H.-K.; Hsu S.-N.; Huang T.-H.; Chiu H.-C. J. Mol. Catal. A: Chem. 2002, 188, 201. (26) Tang H.; Van der Ven A.; Trout B. L. Phys. ReV. B 2004, 70, 45420. (27) Dicke, J.; Rotermund, H. H.; Lauterbach, J. Surf. Sci. 2000, 454456, 352. (28) Petroski J. M.; Wang Z. L.; Green T. C.; El-Sayed M. A. J. Phys. Chem. B 1998, 102, 3316. (29) Payne M. C.; Teter M. P.; Allan D. C.; Arias T. A.; Joannopoulos J. D. ReV. Mod. Phys. 1992, 64, 1045. (30) Car, R.; Parrinello M. Phys. ReV. Lett. 1985, 55, 2471. (31) Perdew, J. P.; Wang, Y. Phys. ReV. B 1976, 13, 4274. (32) De Vita, A.; Canning, A.; Car, R. EPFL Supercompt. J. 1994, 6, 22. (33) Ciacchi, L. C.; Pompe, W.; De Vita A. J. Am. Chem. Soc. 2001, 123, 7371. (34) Ciacchi, L. C.; Pompe, W.; De Vita A. J. Phys. Chem. B 2003, 107, 1755. (35) Mertig, M.; Ciacchi, L. C.; Seidel, R.; Pompe, W.; De Vita, A. Nano Lett. 2002, 2, 841.
Catalytic Oxidation Activity of Platinum Particles (36) Monkhorst H. J.; Pack J. D. Phys. ReV. B 1976, 13, 5188. (37) Boisvert, G.; Lewis, L. J.; Scheffler, M. Phys. ReV. B 1998, 57, 1881. (38) Giannozzi P.; de Gironcoli S.; Pavone P.; Baroni S. Phys. ReV. B 1991, 43, 7231. (39) The ABINIT code is a common project of the Universite´ Catholique de Louvain, Corning Incorporated, and other contributors (URL http:// www.abinit.org); Gonze, X. et al. Comput. Mater. Sci. 2002, 25, 478. (40) Gonze, X. Phys. ReV. B 1996, 55, 10337. (41) Anisimov, V. I.; Zaanen, J.; Andersen, O. K. Phys. ReV. B 1991, 44, 943. (42) Cococcioni, M.; de Gironcoli, S. Phys. ReV. B 2005, 71, 35105. (43) Baroni, S.; Dal Corso, A.; de Gironcoli, S.; Giannozzi, P.; Cavazzoni, C.; Ballabio, G.; Scandolo, S.; Chiarotti, G.; Focher, P.; Pasquarello, A.; Laasonen, K.; Trave, A.; Car, R.; Marzari, N.; Kokalj, A. http://www.pwscf.org/. (44) Otto, J.; Thomas, W. In Landolt-Bo¨rnstein, 6th ed.; Hausen H., Ed; Springer-Verlag: Berlin, Germany, 1967; Vol. IV/4a, pp 230-231. (45) Reuter, K.; Scheffler, M. Phys. ReV. B 2001, 65, 35406. (46) Lee, C.; Gonze, X. Phys. ReV. B 1995, 51, 8610. (47) McBride, J. R.; Graham, G. W.; Peters, C. R.; Weber, W. H. J. Appl. Phys. 1991, 69, 1596. (48) Hasegawa, M.; Nishidate, K. Phys. ReV. B 2004, 70, 205431. (49) Baskin, Y.; Meyer, L. Phys. ReV. 1955, 100, 544. (50) Weber, W. H.; Graham, G. W.; McBride, J. R. Phys. ReV. B 1990, 42, 10969. (51) Wu, R.; Weber, W. H. J. Phys.: Condens. Matter 2000, 12, 6725. (52) Galloni E. E.; Roffo A. E. J. Chem. Phys. 1941, 9, 875. (53) Pies, W.; Weiss, A. In Landolt-Bo¨rnstein, New Series; Hellwege, K.-H., Hellwege, A. M., Eds; Springer-Verlag: Berlin, Germany, 1975; Vol. NS III/7b1, p 601. (54) Uddin, J.; Peralta, J. E.; Scuseria, G. E. Phys. ReV. B 2005, 71, 155112 and references therein. (55) The differences between formation energies of all considered oxide phases are found to change only slightly at the GGA+U level with respect to the values at the simple GGA level. In particular, inclusion of the U term does not affect the relative stabilities of the considered phases. Therefore, in the following of the paper all reported numbers refer to the standard GGA approach, unless explicitly stated. (56) Unreconstructed Pt(100) has been considered, since its quasihexagonal reconstruction is lifted already at very low CO pressures.57
J. Phys. Chem. B, Vol. 110, No. 30, 2006 14869 (57) Van Beurden, P.; Bunnik, B. S.; Kramer, G. J.; Borg, A. Phys. ReV. Lett. 2003, 90, 66106. (58) In the case of a thicker Pt3O4 layer on the Pt(100) surface, only one of the two slab surfaces has been covered with the oxide phase, to reduce the computational effort. In this case, the presence of a free surface of the slab has been taken into account via the calculated free surface energy of the clean Pt(100) surface. (59) Weaver, J. F.; Chen, J.-J.; Gerrard, A. L. Surf. Sci. 2005, 592, 83. (60) Deskins, N. A.; Lauterbach, J.; Thomson, K. J. Chem. Phys. 2005, 122, 184709. (61) Walker, A. V.; King, D. A. Phys. ReV. Lett. 1999, 82, 5156. (62) Beck, R. D.; Maroni, P.; Papageorgopoulos, D. C.; Dang, T. T.; Schmid, M. P.; Rizzo, T. R. Science 2003, 302, 98. (63) Gee, A. T.; Hayden, B. E.; Mormiche, C.; Kleyn, A. W.; Reidmu¨ller, B. J. Chem. Phys. 2003, 118, 3334. (64) Liu, Z.-P.; Hu, P. J. Chem. Phys. 2001, 114, 8244. (65) Michaelides, A.; Liu, Z.-P.; Zhang, C. J.; Alavi, A.; King, D. A.; Hu, P. J. Am. Chem. Soc. 2003, 125, 3704. (66) To check the consistency of this result, we have calculated the adsorption energy of CH4 on the Pt3O4(100) also at the GGA+U level. Increasing the U parameter from 0 to 4 eV has the effect of increasing the driving force for dissociation by 0.44 eV. (67) Feibelman, P. J.; Hammer, B.; Nørskov, J. K.; Wagner, F.; Scheffler, M.; Stumpf, R.; Watwe, R.; Dumesic, J. J. Phys. Chem. B 2001, 105, 4018. (68) Gil, A.; Clotet, A.; Ricart, J. M.; Kresse, G.; Garcia-Hernandez, M.; Ro¨sch, N.; Sautet, P. Surf. Sci. 2003, 530, 71. (69) Olsen, R. A.; Philipsen, P. H. T.; Baerends, E. J. J. Chem. Phys. 2003, 119, 4522. (70) Alavi, A.; Hu, P.; Deutsch, T.; Silvestrelli, P. L.; Hutter, J. Phys. ReV. Lett. 1998, 80, 3650. (71) Eichler, A.; Hafner, J. Phys. ReV. B 1999, 59, 5960. (72) Li, Z.; Beck, P.; Ohlberg, D. A. A.; Stewart, D. R.; Williams, R. S. Surf. Sci. 2003, 529, 410. (73) Celino, M.; Cleri, F.; D’Agostino, G.; Rosato, V. Phys. ReV. Lett. 1996, 77, 2495. (74) Yazawa, Y.; Yoshida, H.; Hattori, T. Appl. Catal., A 2002, 237, 139. (75) Sabourault, N.; Mignani, G.; Wagner, A.; Mioskowski, C. Org. Lett. 2002, 4, 2117. (76) Zhensheng, J.; Chanjuan, X.; Qingmei, Z.; Feng, Y.; Jiazheng, Z.; Jinzhen, X. J. Mol. Catal. A: Chem. 2003, 191, 61.
