Effects of Hydrogen on the Reactivity of O2 toward Gold Nanoparticles

Nov 30, 2007 - Laura Barrio,† Ping Liu,‡ Jose A. Rodriguez,*,‡ Jose M. Campos-Martin,† and. Jose L. G. Fierro†. Instituto de Cata´lisis y Petroleoquı´...
0 downloads 0 Views 587KB Size
J. Phys. Chem. C 2007, 111, 19001-19008

19001

Effects of Hydrogen on the Reactivity of O2 toward Gold Nanoparticles and Surfaces Laura Barrio,† Ping Liu,‡ Jose A. Rodriguez,*,‡ Jose M. Campos-Martin,† and Jose L. G. Fierro† Instituto de Cata´ lisis y Petroleoquı´mica, CSIC, Marie Curie No. 2, Cantoblanco, 28049 Madrid, Spain, and Chemistry Department, BrookhaVen National Laboratory, Upton, New York, 11973 ReceiVed: May 9, 2007; In Final Form: September 21, 2007

Density functional theory (DFT) was used to study the coadsorption of O2 and H2 or H on Au(111) and Au(100) and on free pyramidal clusters (Au25 and Au29) observed on titania- and ceria-based catalysts. For the adsorption of isolated O2, no significant interaction was obtained between the flat metal surfaces and the oxygen molecule. Gold clusters (Au14, Au25, and Au29) showed a higher reactivity, but this reactivity depended strongly on the type of uncoordinated sites exposed, ensemble effects, and the fluxionality of the metal nanoparticles. On the Au14 and Au29 clusters, a superoxo species and a peroxo moiety were formed, respectively. In contrast, no interaction between O2 and Au25 was seen because of the high coordination number of the exposed metal sites. In general, H2 dissociates much easier on the gold clusters than O2, but again Au25 is less reactive than Au29 or Au14. The pyramidal Au25 and Au29 clusters illustrate the importance of geometry in the chemistry of gold nanoparticles: A smaller particle size does not necessarily imply a bigger chemical reactivity. With the presence of predissociated hydrogen on a gold system, the adsorption energy of oxygen is much higher in absolute value than for the O2/Au systems, accompanied by the formation of a hydroperoxo species. The origin of the synergistic effect observed for the coadsorption of hydrogen and O2 is the formation of O-H bonds that enhance the OTAu interactions, reducing (0.15-0.30 Å) the O-Au bond lengths. The reaction H(a) + OOH(a) f H2O2(a) was found to be highly exothermic on the Au clusters. An interesting case is when the oxygen orientation allows for simultaneous interaction with two H adatoms: The formation of hydrogen peroxide is readily achieved with an energy release of more than 30 kcal/mol. This reaction is only hindered by the need of a concerted approach of O2 to the H adatoms.

Introduction Nanosized gold particles have attracted the attention of researchers because of their high activity in a wide range of reactions.1 Supported gold nanoparticles catalyze hydrogenation reactions,2 low-temperature CO oxidation,3,4 propylene epoxidation,3,5 and hydrogen peroxide production6,7 among others. The activity of gold nanoparticles depends on many factors. Size,8 shape,3 and support9,10 are considered the main parameters to explain the properties of these clusters. Because of the high activity in oxidation reactions, the reactivity of gold clusters with oxygen has been extensively analyzed. However, no clear picture emerges from the work done about the mechanism for inducing the reaction of O2. It is well-established that gold surfaces are inactive toward oxygen adsorption unless activated by deposition of atomic O11 or unless sufficient energy is provided to break the O-O double bond.12 Supported gold nanoparticles have been shown to activate molecular oxygen.13-15 For Au/TiO2 catalysts, it has been suggested that the activation step occurs on the support, which provides active oxygen to the gold cluster.16 Density functional theory (DFT) has been used to investigate the interaction of oxygen with free gold particles. Luo et al.17 performed a study for Au24 clusters and concluded that the oxygen adsorption was dependent on particle shape and coordination number of the * To whom correspondence should be addressed. E-mail: rodrigez@ bnl.gov. † Instituto de Cata ´ lisis y Petroleoquı´mica. ‡ Brookhaven National Laboratory.

interacting gold atoms. Oxygen was found to adsorb preferentially by interaction with more than one gold atom. The chemisorption of oxygen on a Au32 cluster was studied by Wang and Gong,18 and they found that oxygen dissociation was more favorable than molecular adsorption for an icosahedral structure. Barton and Podkolzin found very weak interactions between O2 and Au55 but substantial bonding with Au13 and Au5.19 Theoretical work on supported gold clusters suggests that the interface between the metal particle and oxide support is the most probable active site.9,20 Thus, it seems that there are mechanisms which depend mainly on intrinsic gold particle activity and that there are other mechanisms which are supportinduced or even support-assisted.21 The coadsorption of O2 and H2 on gold or gold/oxide catalysts is a very interesting topic. Experimentally, an enhancement of activity has been observed for oxidation reactions, such as preferential CO oxidation22 or direct epoxidation of propylene, in the presence of H2.3,5 How can the presence of hydrogen, a reducing agent, promote oxidation activity? The formation of a peroxo moiety is the most accepted hypothesis to explain the mechanism. This peroxo species has been experimentally detected by Sivadinarayana et al.23 who carried out an inelastic neutron scattering study of gold nanoparticles supported on titania. Hydrogen and oxygen are the reactants for the direct synthesis of hydrogen peroxide (H2 + O2 f HOOH), and gold catalysts are known to be active in this reaction.6 The energetics of the elementary reactions involved in the synthesis of hydrogen peroxide over noble metal catalysts (Pd, Pt, Ag, and Au) estimated with the bond order conservation-Morse potential

10.1021/jp073552d CCC: $37.00 © 2007 American Chemical Society Published on Web 11/30/2007

19002 J. Phys. Chem. C, Vol. 111, No. 51, 2007

Barrio et al.

Figure 1. Cluster models investigated in this study: (a) Au14; (b) Au29; (c) Au25; (d) Au4; (e) Au5.

(BOC-MP) approach showed that the hydrogen peroxide could be formed more selectively on gold catalysts.24 Such empirical calculations agree with experimental results of direct hydrogen peroxide synthesis on supported gold nanoparticles.25-27 It was found that the rate of H2O2 formation decreased with an increase in the mean diameter of the Au nanoparticles.26 This tendency was also corroborated by preparing a Au/SiO2 catalyst in which the mean diameter of Au particles was about 30 nm and which produced only H2O but in no case hydrogen peroxide.26 This catalytic behavior was also studied with theoretical calculations, which revealed that small gold clusters can act as viable sites for the direct synthesis of hydrogen peroxide.28-30 A thorough kinetic and DFT study of gold catalysts for water formation from H2 and O2 was reported by Barton and Podkolzin19 who found a strong dependence of particle size toward hydrogen peroxide formation, both from experimental and theoretical results. A 0.7 nm size gold particle was found to be the most reactive one, smaller particles were not able to stabilize the hydroperoxyl intermediate, and the icosahedral Au55 particle was inert toward oxygen adsorption. Wells et al.28 performed DFT calculations on hydrogen peroxide formation on very small Au3 clusters and found that the addition of H2 to an O2-Au3 cluster led to the formation of hydroperoxo-like species with an energy release of 18 kcal/mol. On Au4+, Au5, or Au5- clusters, the formation of hydroperoxo groups was also obtained, but for these systems, an energy barrier had to be overcome.30 To obtain realistic results for the activity of gold clusters, it is necessary to model systems with the right size and shape. Also, the fluxionality of the nanoparticles has been found to be a determining factor,31,32 thus, structural constraints should be avoided. The aim of this work is to analyze the reactivity of medium-size (0.7-1.2 nm) gold nanoparticles toward oxygen adsorption and how the presence of hydrogen can affect interactions between gold and oxygen. Small clusters (Au4 and Au5) have also been considered for comparison purposes. We are particularly interested in the reactivity of the Au25 and Au29 clusters shown in Figure 1. These clusters have a pyramidal structure formed by the interconnection of (111) and (100) faces of the bulk metal.10,32 Their shape mimics the experimentally observed structures of supported Au nanoparticles. Such pyramidal gold clusters have been observed by TEM for Au/TiO2 catalysts33 and by scanning tunneling microscopy for Au/CeO2 systems.34 They expose sites that have different coordination numbers and are spatially arranged in a different way. Our previous studies examining the dissociation of H2 and the water-gas shift reaction (CO + H2O f H2 + CO2) on Au29 show that this cluster has quite interesting chemical properties because of ensemble effects and its fluxional properties.10,32,35

Theoretical Methods In the present study, DFT was employed to study the interaction of O2 and O2/H2 with gold nanoparticles (see Figure 1) and with periodic bulk surfaces like Au(111) and Au(100). All the calculations were performed with the DMol3 program36,37 that allows the treatment of molecules, nanoparticles, and extended surfaces with the same level of approximation. In the past, this code has been successfully used to study the interaction of gold nanoparticles with different types of adsorbates, including O2 and H2.10,19,35,38 Gold atoms were expressed by effective core potentials to account for relativistic effects.39,40 Nineteen electrons of the gold atoms’ valence shell were explicitly considered in the calculations. The DF calculations were carried out using a numerical basis set that included the 5s, 5p, 5d, 6s, 6p, and 6d orbitals of gold and that contained polarization and diffuse functions for O and H. The cutoff value for the numerical basis set employed was 9.45 bohr (∼5 Å). The DF calculations were performed employing the generalized gradient approximation with the PW91 functional41 and allowing spin polarization (i.e., testing different spin multiplicities) as used in previous works.9,10 In the literature, there is no agreement on which is the best functional to study the bonding of molecules to gold systems. The PW91 functional has been used in many recent works.10,15,31,32,35,42 Usually, the PW91 functional overestimates the bonding energy of molecules to transition-metal surfaces or clusters,43 but for the adsorption of O2 on a series of metal clusters (Au3, Au5, Au8, Au14, Au25, Au29, Au55), this functional predicts adsorption energies (-6 to -32 kcal/mol) that are not as big as those measured with calorimetry for O2 on Au nanoparticles supported on silica (-35 to -50 kcal/mol).19 Modified GGA functionals like the PBE, RPBE, or B3LYP give bonding interactions between O2 and gold that are even weaker than those predicted by the PW91 functional43c and, consequently, give a bigger error when compared to experimental values.19 Our studies for the adsorption of O2/O on extended Au surfaces using the PW91 functional give qualitative trends that agree very well with experimental results.11,12 In this article, our interest is not the absolute values of the calculated bonding energies but is mainly in the trends observed when changing the adsorption sites for the bonding of O2 or when coadsorbing O2 and H2. To simulate the bulk surfaces, we used the supercell approach with a slab thickness of three layers and a vacuum of 11 Å.10,32,35 Reciprocal-space integration over the Brillouin zone was approximated employing a Monkhorst-Pack grid of 4 × 4 × 1 k-points. The top layer of the Au slabs was optimized during the DFT calculations. The clusters in Figure 1 expose adsorption

Effects of Hydrogen on the Reactivity of O2

J. Phys. Chem. C, Vol. 111, No. 51, 2007 19003 partition methods have well-known limitations, but nevertheless they provide a helpful and familiar tool for qualitative analysis.45 The Mulliken analysis is known to be very sensitive toward the basis set employed. The Hirshfeld method is less dependent on the used basis set, but it has a tendency to underestimate the atomic charges.36,37,45 The net spin population on the oxygen atoms was found to be zero when adsorbed on Au14, Au25, Au29, or extended surfaces. Results and Discussion

Figure 2. Unit cells for O2 adsorption on (a) Au(111) and (b) Au(100) surfaces.

sites that are not present on Au(111) or Au(100). Pyramidal clusters like Au29 and Au25 have been observed by TEM for Au/TiO2 catalysts33 and by scanning tunneling microscopy for Au/CeO2 systems.34 A frequency calculation on the geometries of Au14, Au25, and Au29 showed no negative values and revealed that the structures in Figure 1 are true minima. The Au25 and Au29 structures represent the shape of supported gold nanoparticles.33,34 In a catalyst, only their first (i.e., bottom) layer will be in contact with an oxide support. Therefore, the oxide is expected to have only a small effect on the second and third layers of the clusters, where the most reactive gold atoms are. The structure of the Au14 cluster (shown in Figure 1) allows testing the reactivity of a particular configuration of Au atoms. Furthermore, it is close in energy to other structures reported in the literature.19,38 For comparison purposes, the reactivity of Au4 and Au5 clusters has also been analyzed. A Au4 cluster represents the top layer of the Au29 cluster and has a planar square shape. It has a size of 0.4 nm and has been used as a model in many studies investigating the chemical properties of supported Au nanoparticles.28,42c The Au5 cluster has a 0.5 nm size, and its planar shape has already been considered in the bibliography.19 The geometry of the Au4, Au5, Au14, Au25, and Au29 clusters was fully relaxed to allow structural changes during the bonding of O2 and H2. This is quite important since the fluxionality of the gold clusters determines their reactivity.32,44 In this work, adsorption energies are defined as

∆E ) E(adsorbate + gold substrate) - E(free adsorbate) E(gold substrate) A negative ∆E denotes an exothermic process and bonding to a particular gold substrate. Charge and spin population on each atom were calculated by the Mulliken and Hirshfeld population analyses.36,37,45 Charge

A. Adsorption of O2. The adsorption of molecular oxygen on bulk metal surfaces was included to establish a reference for the adsorption on molecular clusters. The most stable surfaces of gold, Au(100) and Au(111), were analyzed. A representative cell of three layers of gold atoms was considered to model the metal surfaces (see Figure 2), relaxing only the top layer of atoms.32 The coordination number of gold atoms in any of these surfaces is at least eight, which makes them very stable and poorly active. The interaction of O2 with Au(111) or Au(100) was very weak, and no chemisorption was observed (Table 1), in agreement with that found in previous DFT calculations.19,46 The rigidity of the bulk metal does not allow for a strong stabilization of the adsorbate. These results are in agreement with several experimental data for the reaction of gold surfaces with oxygen,11,12 which indicate that bulk surfaces have a very low reactivity unless some activation treatment is performed. Next, we examine the bonding of O2 to the gold clusters in Figure 1. First of all, the simplest unit cell for a face-centered cubic (fcc) structure formed by 14 gold atoms has been studied. The particle size is 0.7 nm. On this Au14 cluster, highcoordinated and inert gold atoms are present, but also more active atoms at the corner positions (which possess only three neighbor atoms) exist. Here, O2 exhibits a weak interaction (3.5 kcal/mol). The oxygen molecule binds to two gold atoms and forms a superoxo structure (Figure 3a). An important effect is the distortion of the cluster upon adsorption, indicating that geometry relaxation of the gold atoms to hold the oxygen molecule is necessary for obtaining an adsorbed state. This is in agreement with the work done by Barton and Podkolzin19 who found also an exothermic adsorption of molecular oxygen on a Au13 cluster. Okumura et al.29 could only obtain a metastable adsorption of oxygen on their icosahedral Au13 cluster probably because they fixed the geometry of the gold atoms. The reactivity of oxygen on Au29 is the most interesting case. The Au29 cluster has a pyramidal structure that mimics experimentally observed structures for Au nanoparticles on TiO233 and CeO2.34 On Au29, the oxygen is adsorbed on the top and interacts with four gold atoms (Figure 3b). This finding is very similar to that obtained in our previous study with H2 adsorption on this cluster where we proved that the top position is the most reactive one because it allows the cooperation of four active/low-coordinated gold atoms.32 The O-O bond distance is elongated up to 1.5 Å, which corresponds to a peroxo

TABLE 1: Selected Structural and Thermochemical Data for the Adsorption of O2 on Gold Surfaces and Clusters gold system

∆Eads (kcal/mol)

adsorbed species

O-Obonddist. (Å)

O-Aubonddist. (Å)

Au(111) Au(100) Au4 Au5 Au14 Au29 Au25

-1.26 -0.97 -13.67 -20.86 -3.56 -11.00 -2.60

not adsorbed not adsorbed superoxo superoxo superoxo peroxo not adsorbed

1.231 1.225 1.325 1.341 1.353 1.502 1.229

3.109 3.534 2.167 2.140 2.443 2.240 3.031

19004 J. Phys. Chem. C, Vol. 111, No. 51, 2007

Barrio et al.

Figure 3. O2 adsorption on gold clusters: (a) Au14; (b) Au29; (c) Au25; (d) Au4; (e) Au5. On Au25, there is no bonding of the oxygen molecule (see text).

Figure 5. Calculated energy changes for the O2 dissociation on Au29.

Figure 4. Electron density plots mapped by electrostatic potential for (a) Au29 and (b) O2‚Au29 systems.

moiety. This is known to be a very active species in many oxidation reactions.47 The corresponding electron density plots in Figure 4 show a large accumulation of charge on the adsorbed molecule and a redistribution of charge for the metal atoms in the cluster. A Mulliken population analysis gives a charge of -0.30e on the oxygen atoms, whereas for a Hirshfeld analysis the corresponding charge is -0.20e. Electrons flow from Au(6s,6p) states into the 2π* orbitals of the adsorbate. Thus, the oxygen is highly activated when bonded to the Au29 cluster.

The binding energy for the adsorption is significant (11 kcal/ mol, Table 1). Because of the large elongation of the O-O distance upon adsorption, the dissociation of O2 may appear as a probable pathway for the reaction of the molecule on the Au29 cluster. Figure 5 shows calculated changes in energy as a function of the O-O separation. To obtain these values, the O-O bond distance was elongated in steps of 0.1 Å while fully relaxing the gold atoms in the cluster. Also, a full transition state optimization of the whole structure was performed to better estimate the energy barrier. A frequency analysis revealed that the only imaginary vibration corresponded to the O-O stretching, corroborating the true nature of the transition state. As it is illustrated in Figure 5, the dissociation of O2 is not thermodynamically favored. To break the O-O bond, a small energy barrier (∼9 kcal/mol) needs to be overcome, but more importantly, the dissociated state is equally stable to the peroxo one, so the dissociation process has no energy reward. In such clusters, the most probable chemisorbed species after interaction with O2 would be the peroxo one. The size and shape of the Au29 are more favorable toward O2 adsorption than that of the Au14 cluster. For Au29, a distorted structure is also obtained upon O2 adsorption, indicating the importance of the fluxionality for particle reactivity. The gold cluster is distorted in a way that the preferential exposed face is Au(111) over Au(100), see Figure 3b. This is the most stable face of gold, so the oxygen adsorption helps to stabilize the nanoparticle. Luo et al.17 also found an exothermic adsorption

Effects of Hydrogen on the Reactivity of O2

J. Phys. Chem. C, Vol. 111, No. 51, 2007 19005

TABLE 2: Selected Structural and Thermochemical Data for the Coadsorption of O2 and H2 on Gold Clusters and Surfaces structure

∆Eads (kcal/mol)

adsorbed species

O-Obonddist. (Å)

Au-Obonddist. (Å)

O2‚H2‚Au14a H2O2‚Au14b O2‚H2‚Au29a O2‚H2‚Au29 symma H2O2‚Au29b O2‚H2‚Au25a H2O2‚Au25b O2‚H2‚Au4a O2‚H2‚Au5a O2‚H‚Au(111)c O2‚H‚Au(100)c O2‚H‚Au14c O2‚H‚Au29c O2‚H‚Au25c O2‚H‚Au4c O2‚H‚Au5c

-28.77 -20.28 -20.08 -35.12 -13.79 -11.55 -24.33 -30.76 -25.51 -11.48 -16.28 -13.23 -21.46 -14.56 -18.82 -13.17

hydroperoxo H2O2 formation hydroperoxo H2O2 formation H2O2 formation hydroperoxo H2O2 formation hydroperoxo superoxo hydroperoxo hydroperoxo hydroperoxo hydroperoxo hydroperoxo hydroperoxo hydroperoxo

1.468 1.473 1.485 1.466 1.469 1.454 1.466 1.432 1.361 1.448 1.439 1.502 1.448 1.436 1.4053 1.480

2.126 2.777 2.116 2.657 3.167 2.115 2.720 2.048 2.148 2.265 2.293 2.100 2.139 2.137 2.103 2.201

a Stabilization energies have been estimated by ∆E ) E(O2‚H2‚Au) - E(O2) - E(H2‚Au) to account more explicitily for the energetics of O2 adsorption. b OOH‚H‚Aun systems have been considered as reference, so this data accounts for the energetics of hydrogen peroxide formation. ∆E ) E(H2O2‚Au) - E(O2‚H2‚Au). c Structures containing atomic hydrogen have been considered as reference for the estimation of stabilization energies.

of O2 on a Au24 cluster. They estimated an adsorption energy of 10.15 kcal/mol. They also found that the O-O bond was elongated but only to form a superoxo structure (O-Obond distance ) 1.37 Å). The difference in the kind of oxygen adsorbed species may be because the oxygen was able to interact only with two atoms on their Au24 cluster instead of four atoms as in our calculations for O2/Au29. Interestingly, on top of the Au29 cluster, O2 binds in a different way than on Au4 or Au5 clusters, where a side-on configuration is preferred (see Figure 3d and 3e). For the O2/Au5 system, we find a structure and a bonding energy (21 kcal/mol) that are close to those found by Barton and Podkolzin.19 Finally, the adsorption of O2 on the Au25 cluster was investigated. This cluster is formed by removing the top layer of the Au29 cluster and exposing a large number of gold atoms. Surprisingly, on this cluster, no adsorption of molecular oxygen was found (Figure 3c). Different symmetry and geometry constraints as well as different attack positions were attempted, but in no case did the oxygen molecule make a substantial bond with the gold particle. The coordination number of the top layer of gold atoms on the Au25 cluster is much higher on average than that for the Au29 cluster. Thus, this result indicates that the high coordination of exposed gold atoms inhibits the activity of the cluster. The Au24 cluster considered in the literature17 was active toward oxygen chemisorption because the exposed gold atoms possessed a low coordination number. While small clusters like Au4 and Au5 exhibit a significant reactivity toward O2, the trends found for the bonding of the molecule to Au14, Au25, and Au29 indicate that not only particle size governs the oxygen adsorption process; also, cluster shape and morphology have a significant influence on chemical reactivity. Besides, the superoxo species obtained on the small gold clusters are not as big and powerful oxidant agents as the peroxo-one obtained on the Au29 cluster. The results for the adsorption of O2 on the pyramidal Au25 and Au29 clusters highlight the importance of geometry in the chemistry of gold nanoparticles. A cluster that has a lower number of Au atoms and only two layers (Au25) is not necessarily more active than a cluster that has a bigger number of atoms and three layers (Au29). In the TEM or STM images seen for Au/TiO2 and Au/ CeO2,33,34 a smaller particle size does not necessarily imply a bigger chemical reactivity. Several factors are responsible for the high activity of Au29. Not only the fluxionality of the particle, which is also present for Au14 and Au25 clusters, facilitates the interaction with adsorbates but also an ensemble effect involving

the simultaneous cooperation of four active gold atoms makes unique the chemical activity of the pyramidal Au29 cluster.32 This ensemble effect stresses the need of modeling and obtaining the right particle shape, which allows the simultaneous interaction of many gold atoms with substrates. B. Coadsorption of Hydrogen and O2. Bulk surfaces of gold are inactive toward H2.32,48 The reactivity of gold clusters toward hydrogen was investigated in our previous work32 in which spontaneous dissociation of the molecule on Au14 and Au29 systems was found. Thus, if such gold nanoparticles are exposed to a flow of pure H2 or a mixture of H2 and O2, it would be feasible to obtain dissociation of the hydrogen molecule on these systems. As it was stated in the Introduction, experimentally, it is well-known that the presence of H2 promotes oxidation reactions on gold, but the causes behind this important phenomenon are not known, and there is much debate.3,5 Therefore, a study of the effect of adsorbed H2 or H on the reactivity of O2 seems very interesting from academic and practical viewpoints. In principle, O2 gas could react directly with the preadsorbed H atoms (Eley-Rideal mechanism) or could adsorb first and then react (Langmuir-Hinshelwood mechanism). The second possibility will require a substantial interaction between O2 and the gold system which may not always occur (see Table 1). Barton and Podkolzin have investigated the O2(ads) + H(ads) f OOH(ads) reaction on Au(111) and Au(211) and on gold clusters Au55, Au13, and Au5.19 They found this Langmuir-Hinshelwood process to be endothermic on the small Au5 and Au13 clusters and to be exothermic on the Au55 cluster and on the Au(111) and Au(211) surfaces. It appears that the reaction requires a relatively large ensemble of sites (>10 atoms). What can we expect for Au14, Au25, and Au29? Can Au29 catalyze the formation of OOH or HOOH exposing only four gold atoms with low coordination? In a systematic way, we analyzed the effects of hydrogen on the reactivity of the clusters shown in Figure 1 toward oxygen

O2(gas) + 2H/Au f {OOH + H}/Au

(1)

{OOH + H}/Au f {HOOH}/Au

(2)

Thermochemical and structural data are summarized in Table 2. On this table, the reference for the estimation of adsorption energies is the H2‚Au systems in order to account more explicitly for the effects of O2 adsorption. If free H2, O2, and Au systems

19006 J. Phys. Chem. C, Vol. 111, No. 51, 2007

Figure 6. Final state for the reaction of O2 with H2 predissociated on Au14. Initially, the O2 molecule was close to only one of the H adatoms.

are taken as reference, the stabilization energy would include also the H-Au interactions. The corresponding final states for eq 1 are represented in Figures 6-9. To obtain these structures, the O2 molecule was set within bonding distance of each Au cluster near one or two of the hydrogen adatoms, and the geometry of the whole system was optimized. The produced OOH was reactive toward H adatoms, and eq 2 was exothermic on the Au14, Au25, and Au29 clusters. The presence of dissociated hydrogen atoms on the Au14 cluster (H2‚Au14 in our notation) enhances the reactivity of this system toward O2. Thus, the adsorption of O2 on H2‚Au14 leads to the formation of a hydroperoxo complex (Figure 6). The stabilization energy is more than 8 times higher than single O2 adsorption. The coadsorption of both molecules has a high synergistic effect on the thermodynamics in part as a result of HsO bonding interactions. Looking at the O-Au bond distances in O2‚Au14 and H2‚O2‚Au14 (compare Tables 1 and 2), it is clear that the oxygen atoms become closer (∼0.3 Å) to the metal cluster when the coadsorption with H occurs. So, the presence of hydrogen also makes the bond between the gold cluster and the oxygen species stronger. The oxygen molecule was able to interact with one single H adatom, but what would happen if the second H is forced to interact with the hydroperoxo moiety? The formation of H2O2, eq 2, is achieved in an exothermic process (∆E ) -20.2 kcal/mol). The strong interaction of the oxygen with hydrogen makes weaker the bonding of the H2O2 adsorbate to the gold cluster, as can be deduced by the high Au-O distance obtained in Table 2. On the H2‚Au29 cluster, several possibilities for the bonding of O2 were tested. First, the oxygen molecule was adsorbed near one H atom on the most stable structure obtained for H2‚Au29.30 The reaction between molecular oxygen and this system results in the formation of a very stable hydroperoxo moiety (Figure 7). The O-O bond distance is elongated up to 1.485 Å. The stabilization energy for the process is 20.1 kcal/mol, which is almost twice the sum of the separate adsorption energies for O2 and H2. Again, a synergistic effect is obtained when coadsorbing O2 in the presence of hydrogen. Also, a shortening (∼0.13 Å) of the oxygen-gold bond distance is obtained, suggesting once more a stronger Au-O bond than in the absence of hydrogen. The electron density plot for this system (not shown) gave a large accumulation of charge on the O atoms. The calculated Mulliken charges were -0.38e for the O bonded to the H and -0.32e for the other one. The Hirshfeld charge analysis gave a charge of -0.22e for the free oxygen atom and -0.14e for the hydroperoxo one. The formation of a new O-H bond and H2O2, eq 2, led to an exothermic process releasing 13 kcal/mol.

Barrio et al.

Figure 7. Final state for the reaction of O2 with H2 predissociated on Au29. Initially, the O2 molecule was close to only one of the H adatoms.

Another possibility has been taken into account for the reaction of oxygen on a H2‚Au29 cluster. It involves the symmetric attack of O2 to a preadsorbed H2 molecule on Au29. Although this is a much less probable starting condition,32 this concerted pathway leads to the barrierless formation of hydrogen peroxide in a very exothermic process as two O-H bonds are formed. We changed the O2-H2 separation manually, in a systematic way, and were not able to find a local minimum or reaction barrier. In general, if the orientation for the O2 attack allows for the interaction with two hydrogen atoms, the formation of H2O2 should occur spontaneously. As it can be seen in Figure 8, the gold cluster undergoes some distortion during this reaction. After the formation of hydrogen peroxide, this molecule moves away from the gold cluster. Such finding is in agreement with experimental data: Supported gold nanoparticles are well-known catalysts for direct hydrogen peroxide synthesis from their elements.6,7,25-27 The limitation for the complete formation of H2O2 on the surface of gold nanoparticles could be restricted to the right approach and collision conditions for the H2 and O2 to meet. The performed calculations show that there is no energy barrier to overcome either for the H-H bond breaking or for the O-O bond elongation when the process shown in Figure 8 takes place on a Au29 cluster. Next, we will consider the case of a Au25 system. As we stated above, no significant interaction was obtained for the adsorption of O2 on this structure (Figure 3c and Table 1). Nevertheless, molecular hydrogen was able to dissociate on Au25, although a very weak stabilization energy was obtained (-0.45 kcal/mol). Once the hydrogen was dissociated, a highly distorted structure was found. When the adsorption of oxygen was tried on this structure, a hydroperoxo species was formed with a stabilization energy of 15 kcal/mol. Thus, the presence of dissociated H2 makes possible the adsorption of O2. Once more, the formation of a bond between O and H atoms is favored thermochemically and allows the generation of an oxidant species on the gold particle. When the formation of hydrogen peroxide according to eq 2 is considered on the surface of the Au25 cluster, an exothermic process is found (∆E ) -24.3 kcal/ mol), as seen on Au14 and Au29. Interestingly, for the Au5 cluster, the oxygen adsorption in the presence of dissociated H2 leads to the formation of a hydroperoxo structure with an energy release of 25 kcal/mol. This value is close to the 20 kcal/mol release for the single oxygen adsorption. Although a new bond has been made, the O-O bond distance is not as large as for a peroxo-like bond. Table 2 shows that the O-O bond is around 1.36 Å, which is better assigned to a superoxo-like bond than to the peroxo-like bond distance (1.5 Å). On Au5, the formation of peroxo-like

Effects of Hydrogen on the Reactivity of O2

J. Phys. Chem. C, Vol. 111, No. 51, 2007 19007

Figure 8. Initial and final state for the reaction of O2 with H2 predissociated on Au29. There was a symmetric attack of O2 on the two preadsorbed H atoms.

Figure 9. Final state for the reaction of O2 with H2 predissociated on Au25. Initially, the O2 molecule was close to one of the H adatoms.

Au(100) or Au(111) was found (see Table 1 and refs 11, 12, 19, and 46). The exothermicity that we calculate for the O2(gas) + H(ads) f OOH(ads) reaction on the extended gold surfaces is consistent with the exothermicity found by Barton and Podkolzin for the O2(ads) + H(ads) f OOH(ads) reaction on Au(111) and Au(211).19 An Eley-Rideal mechanism cannot be discarded because the O2 molecule interacts too weakly with the periodic Au substrates (Table 1). Thus, O2 either bonds directly to the adsorbed H or adsorbs on the surface with the help of HTO2 bonding interactions. Summing up, the obtained results show that the reactivity of medium-size gold clusters and surfaces toward oxygen is highly enhanced by the presence of preadsorbed hydrogen. The interaction of gold with oxygen is stronger (i.e., shorter Au-O bonds by 0.15-0.3 Å) and leads to easier formation of very active oxidant species (OOH). This result can explain why H2 accelerates oxidation reactions,3,5,22 since it allows the easy and fast formation of hydroperoxo moieties, a very powerful oxidant species. The Au14, Au25, and Au29 clusters have the ensemble of sites and fluxionality necessary to perform the O2(gas) + H(ads) f OOH(ads) reaction. Conclusions

Figure 10. Final state for the reaction of O2 with preadsorbed H on gold surfaces: (a) Au(111) and (b) Au(100).

structrure is inhibited and, thus, the adsorption energy is not much larger than that obtained for single O2 adsorption (i.e., no synergistic effect on the oxygen adsorption by the presence of hydrogen). Thus, the nature of eq 1 is affected by the size and shape of the gold clusters. Although extended Au surfaces are inert toward molecular hydrogen adsorption, it is possible to find a stable structure when only atomic H is adsorbed (the H could be generated by dissociation of H2 on a hot filament or by spillover from another surface). If a H atom is present on Au(111) or Au(100), bonding of O2 can occur. On H/Au(111) and H/Au(100), there is oxygen adsorption to form a hydroperoxo-like species (see Figure 10) with energy releases of 11.5 and 16 kcal/mol, respectively. In the absence of H, no significant interaction between O2 and

DFT was used to study the adsorption of O2 on Au(111) and Au(100), and no significant interaction was obtained between the flat metal surfaces and the oxygen molecule. The adsorption of O2 on Au14, Au25, and Au29 nanoparticles showed a complex behavior. On the Au14 cluster, a superoxo species was obtained upon adsorption whereas for Au29 a peroxo moiety was formed. The shape and size of the Au29 cluster are more favorable toward oxygen activity. Not only is the adsorption energy bigger for this system (11.0 vs 3.6 kcal for Au14), but also a more active oxidant species is formed. In contrast, the study on the Au25 cluster led to no interaction between oxygen and the gold atoms because of the high coordination number of the exposed metal sites. Au14 and the pyramidal Au25 and Au29 clusters illustrate the importance of geometry in the chemistry of gold nanoparticles: A smaller particle size does not necessarily imply a bigger chemical reactivity. When O2 adsorbs on a gold nanoparticle in which H2 has been previously dissociated, a hydroperoxo species is obtained. The ∆E of the process is very exothermic, and the general trend is that the adsorption energy is much bigger in absolute value than that for the sum of the H2/Au and O2/Au systems. On the Au25 cluster and on the extended Au(111) and Au(100) surfaces, the presence of hydrogen allows for the adsorption of oxygen, which was not possible for the bare gold systems. The origin

19008 J. Phys. Chem. C, Vol. 111, No. 51, 2007 of the synergistic effect observed for the coadsorption of hydrogen and O2 is the formation of O-H bonds that enhance the OTAu interactions. The reaction H(a) + OOH(a) f H2O2(a) was found to be highly exothermic on the Au clusters. An interesting case is when the oxygen orientation allows for simultaneous interaction with two H atoms; the spontaneous formation of hydrogen peroxide is achieved with an energy release of more than 30 kcal/mol. Our results indicate that in this case there is no energy barrier for hydrogen peroxide formation on gold nanoclusters, and only the need for a concerted approach of O2 hinders this reaction. Acknowledgment. CICyT (Spain) and Comunidad de Madrid/CSIC are acknowledged for the financial support in the project ENE2004-07345-C03-01/ALT and 200680M24, respectively. CTI of CSIC is also acknowledged for CPU facilities. The work done at Brookhaven National Laboratory was supported by the U.S. Department of Energy, Division of Chemical Sciences (DE-AC02-98CH10886). References and Notes (1) Bond, G. C.; Louis, C.; Thompson, D. T. Catalysis by gold; Imperial College Press: London, U.K., 2006. (2) Pawelec, B.; Cano-Serrano, E.; Campos-Martin, J. M.; Navarro, R. M.; Thomas, S.; Fierro, J. L. G. Appl. Catal. A:, Gen. 2004, 275, 127. (3) Haruta, M. CATTECH 2002, 6, 102. (4) Zanella, R.; Giorgio, S.; Shin, C.; Henry, C.; Louis, C. J. Catal. 2004, 222, 357. (5) Nijhuis, T. A.; Visser, T.; Weckhuysen, B. M. Angew. Chem., Int. Ed. 2005, 44, 1115. (6) Campos-Martin, J. M.; Blanco-Brieva, G.; Fierro, J. L. G. Angew. Chem., Int. Ed. 2006, 45, 6962. (7) Landon, P.; Collier, P. J.; Carley, A. F.; Chadwick, D.; Papworth, A. J.; Burrows, A.; Kiely, C. J.; Hutchings, G. J. Phys. Chem. Chem. Phys. 2003, 5, 1917. (8) Chen M. S.; Goodman D. W. Science 2004, 306, 252. (9) Molina L. M.; Rasmussen M. D.; Hammer B. J. Chem. Phys. 2004, 120, 7673. (10) Rodriguez, J. A.; Liu, P.; Hrbek, J.; Evans, J.; Pe´rez, M. Angew. Chem., Int. Ed. 2007, 46, 1329. (11) Deng, X.; Min, B. K.; Guloy, A.; Friend, C. J. Am. Chem. Soc. 2005, 127, 9267. (b) Stiehl, J. D.; Kim, T. S.; McClure, M.; Mullins, C. B. J. Phys. Chem. B 2005, 109, 6316. (12) Weiher, N.; Beesley, A. M.; Tsapatsaris, N.; Delannoy, L.; Louis, C.; van Bokhoven, J. A.; Schroeder, S. L. M. J. Am. Chem. Soc. 2007, 129, 2240. (13) Miller, J. T.; Kropf, A. J.; Zhac, Y.; Regalbuto, J. R.; Delannoy, L.; Louis, C.; Bus, E.; van Bokhoven, J. A. J. Catal. 2006, 240, 222. (14) Bokhoven, J. A.; Louis, C.; Miller, J. T.; Tromp, M.; Safonova, O. V.; Grlatzel, P. Angew. Chem., Int. Ed. 2006, 45, 4651. (15) Zhang, C.; Yoon, B.; Landman, U. J. Am. Chem. Soc. 2007, 129, 2228. (16) Wang, J. G.; Hammer, B. Top. Catal. 2007, 44, 49. (17) Luo, C.; Fa, W.; Dong, J. J. Chem. Phys. 2006, 125, 084707.

Barrio et al. (18) Wang, Y.; Gong, X. G. J. Chem. Phys. 2006, 125, 124703. (19) Barton, D. G.; Podkolzin, S. J. Phys. Chem. B 2005, 109, 2262. (20) Rasmussen, M. D.; Molina, L.; Hammer, B. J. Chem. Phys. 2005, 123, 161104. (21) Remediakis, I. N.; Lopez, N.; Nørskov, J. K. Appl. Catal., A: Gen. 2005, 291, 13. (22) Schubert, M. M.; Plzak, V.; Garche, J.; Behm, R. J. Catal. Lett. 2001, 76, 143. (23) Sivadinarayana, C.; Choudhary, T. V.; Daemen, L. L.; Eckert, J.; Goodman, D. W. J. Am. Chem. Soc. 2004, 126, 38. (24) Olivera, P. E.; Patrito, M.; Sellers, H. Surf. Sci. 2004, 313, 25. (25) Landon, P.; Collier, P. J.; Papworth, A. J.; Kiely, C. J.; Hutchings, G. J. Chem. Commun. 2002, 2058. (26) Okumura, M.; Kitagawa, Y.; Yamagcuhi, K.; Akita, T.; Tsubota, S.; Haruta, M. Chem. Lett. 2003, 32, 822. (27) Edwards, J. K.; Solsona, B. E.; Jandon, P.; Carley, A. F.; Herzing, A.; Kiely, C. J.; Hutchings, G. J. J. Catal. 2005, 236, 39. (28) Wells, D. H.; Delgass, W. N.; Thomson, K. T. J. Catal. 2004, 225, 69. (29) Okumura, M.; Kitagawa, Y.; Haruta, M.; Yamaguchi, K. Appl. Catal., A: Gen. 2005, 291, 37. (30) Joshi, A. M.; Delgass, W. N.; Thomson, K. T. J. Phys. Chem. B 2005, 109, 22392. (31) Ha¨kkinen, H.; Abbet, S.; Sanchez, A.; Heiz, U.; Landman, U. Angew. Chem., Int. Ed. 2003, 42, 1297. (32) Barrio, L.; Liu, P.; Rodriguez, J. A.; Campos-Martin, J. M.; Fierro, J. L. G. J. Chem. Phys. 2006, 125, 164715. (33) Okazaki, K.; Ichikawa, S.; Maeda, Y.; Haruta, M.; Kohyama, M. Appl. Catal., A 2005, 291, 45. (34) Nakamura K., Tokyo Institute of Technology, Yokohama, Japan; private communication. (35) Liu, P.; Rodriguez, J. A. J. Chem. Phys. 2007, 126, 164705. (36) Delley, B. J. Chem. Phys. 1990, 92, 508. (37) Delley, B. J. Chem. Phys. 2000, 113, 7756. (38) Rodriguez, J. A.; Perez, M.; Jirsak, T.; Evans, J.; Hrbek, J.; Gonzalez, L. Chem. Phys. Lett. 2003, 378, 526. (39) Andrae, D.; Huasserman, U.; Dolg, M.; Stoll, H.; Preuss, H. Theor. Chim. Acta 1990, 77, 123. (40) Pyykko¨, P. Angew. Chem., Int. Ed. 2004, 43, 4412. (41) Perdew, J. P.; Chevary, J. A.; Vosko, S. H.; Jackson, K. A.; Pederson, M. R.; Singh, D. J.; Fionlhais, C. Phys. ReV. B 1992, 46, 6671. (42) (a) Herna´ndez, N. C.; Sanz, J. F.; Rodriguez, J. A. J. Am. Chem. Soc. 2006, 128, 15600. (b) Corma, A.; Boronat, M.; Gonzalez, S.; Illas, F. Chem. Commun., 2007, 3371. (c) Rodriguez, J. A.; Vin˜es, F.; Illas, F.; Liu, P.; Takahashi, Y.; Nakamura, K. J. Chem. Phys., in press. (43) (a) Hammer, B.; Jacobsen, K. W.; Nørskov, J. K. Phys. ReV. Lett. 1992, 69, 1971. (b) Paier, J.; Marsman, M.; Kresse, G. J. Chem. Phys. 2007, 127, 024103. (c) Ding, X.; Li, Z.; Hou, J. G.; Zhu, Q. J. Chem. Phys. 2004, 120, 9594. (44) Landman, U.; Yoon, B.; Zhang, C.; Heiz, U.; Arenz, M. Top. Catal. 2007, 44, 145. (45) (a) Szabo, A.; Ostlund, N. S. Modern Quantum Chemistry; McGraw-Hill: New York, 1989. (b) Wiberg, K. B.; Rablen, P. R. J. Comput. Chem. 1993, 14, 1504. (46) Mavrikakis, M.; Stoltze, P.; Nørskov, J. K. Catal. Lett. 2000, 64, 101. (47) Sheldon, R. A.; Wallau, M.; Arends, I.; Schuchardt, U. Acc. Chem. Res. 1998, 31, 485. (48) Hammer, B.; Nørskov, J. K. Nature 1995, 376, 238.