NH3 Oxidation on Oxygen-Precovered Au(111): A Density Functional

The competitive path, that is, NO formation, can be obtained from adsorbed N and O atoms on the surface; the final product is an adsorbed NO molecule ...
0 downloads 0 Views 548KB Size
J. Phys. Chem. C 2008, 112, 247-252

247

NH3 Oxidation on Oxygen-Precovered Au(111): A Density Functional Theory Study on Selectivity Nu´ ria Lo´ pez,* Mo´ nica Garcı´a-Mota, and Jaime Go´ mez-Dı´az Institute of Chemical Research of Catalonia, ICIQ, AVgda. Paı¨sos Catalans 16, 43007 Tarragona, Spain ReceiVed: September 7, 2007; In Final Form: October 10, 2007

We have performed density functional theory simulations applied to slabs to study the dehydrogenation of ammonia on the surface of clean and atomic oxygen or hydroxyl precovered Au(111). Ammonia does not dissociate unimolecularly on the surface and needs a proton scavenger to drive decomposition. O atoms or hydroxyl groups on the surface are basic enough to attract the first H atom from the molecule and are also needed in the resting dehydrogenation steps. Recombination steps to form N2 and NO are hindered by barriers lower than those on Pt or Rh(111) due to the low interaction energies of molecular moieties with the Au(111) surface. The barrier for N2 formation is in any case lower than that of NO in agreement with the high selectivity observed toward N2. Finally, coverage effects are discussed for the competing reactions.

I. Introduction Gold has proven to be a good catalyst provided that nanoparticles are employed.1,2 Gold catalysts are active in selective processes including oxidations and hydrogenations,3-5 and they have been widely studied for the low-temperature CO oxidation.6-8 In contrast, the (111) surface of gold is known to be completely noble.9 Because of the important structure selectivity found for Au,10 standard ultrahigh vacuum experiments on single crystals are unable to retrieve the exceptional catalytic behavior of gold nanoparticles. However, experimentalists have been trying to bridge this material gap to be able to perform activity experiments on Au surfaces but mimicking the results obtained with dispersed nanoparticles. For instance, to test possible new oxidation reactions on the surface of gold nanoparticles, gold crystals in ultrahigh vacuum experiments can be fed directly with O atoms.11-13 Although this approach avoids O2 dissociation, which is the main hindering step for oxidation reactions on Au(111) surface, that is, O2 dissociation, it can be seen as a suitable way to test new catalytic capabilities of gold-based systems in a controlled manner. Very recently, Mullins et al.12 have reported the oxidation of ammonia on atomic O preadsorbed on Au(111) and the effect of oxygen coverage on the products by this technique. Ammonia is a toxic component in waste streams that can be removed by oxidation. Depending on the reaction conditions, the process leads to the formation of water, hydrogen, nitrogen, NO, and even N2O.14 To improve ammonia abatement, the key reaction is the improvement of the N2 route, thus selectively avoiding the conversion of nitrogen into its oxides. Ammonia dehydrogenation15 and oxidation16-19 have been extensively studied. In particular for the oxidation, Pt and Ir are the most active catalysts, but they produce significant amounts of nitrogen oxides. In contrast, for oxygen precovered Au(111) at low coverages, ammonia can be selectively decomposed to form water and N2 (90%) without any traces of nitrogen oxides. However, when the coverage of oxygen doubles that of ammonia, only 53% of ammonia reacts, and the amount of NO * To whom correspondence should be addressed. E-mail: nlopez@ iciq.es.

is 30% of that of N2. In the temperature-programmed desorption, the recombinative N2 produces a peak at 460-470 K while that of NO is a broad peak at 480 K. The aim of our work is to investigate, through a theoretical approach, ammonia oxidation on the O-precovered Au(111) surface to obtain the fundamental reasons for the selectivity toward N2 observed. II. Computational Approach Density functional theory (DFT) applied to slabs has been used to determine the energy profiles for ammonia oxidation on Au(111). The calculations have been performed with the VASP code.20 The energies have been calculated through the RPBE functional.21 The inner electrons were replaced by PAW pseudopotentials22,23 while the monoelectronic functions corresponding to the valence electrons were expanded in plane waves with a cutoff energy of 400 eV. With this approach, the unit cell parameter is 4.208 Å to be compared to the experimental 4.0782 Å.24 For the adsorption and reaction processes the slabs contain 4 layers and a vacuum of 12 Å. The two outermost surface layers and the adsorbates have been allowed to relax. For these surfaces, a p(2 × 2) unit cell was used for the adsorption and reaction studies. In fact, the p(2 × 2) superstructure is well suited to represent the coverage employed in the experiments in ref 12: about 0.25 ML of NH3 and the same coverage for atomic O. The k-point sampling was done up to 5 × 5 × 1 for the surfaces.25 Tests on p(3 × 3) systems have been performed for the N-N and N-O recombinations so as to preserve a coverage comparable to that in the experiments, that is, 0.22-0.25 ML for atomic adsorbates. The barriers for the elementary steps have been estimated by the climbing-image version of the Nudged Elastic Band (NEB) method.26 The optimized transition state structures have been characterized by vibrational analysis. III. Results The reaction scheme for the dehydrogenation of ammonia on the clean, oxygen and hydroxyl covered surface can be summarized as follows.

10.1021/jp077205f CCC: $40.75 © 2008 American Chemical Society Published on Web 12/08/2007

248 J. Phys. Chem. C, Vol. 112, No. 1, 2008

Lo´pez et al.

TABLE 1: Reaction Energies, ∆E in eV, for the Different Reaction Stepsa system

Au

Au-O

Au-OH

NH3 f NH2 + H NH2 f NH + H NH f N + H N+N f N2 N+O f NO

1.52 1.66 1.16 -5.43 -1.45

0.21 0.35 -0.14

-0.07 0.08 -0.42

a Au stands for the clean Au(111) surface, Au-O represents the oxygen-covered surface, and Au-OH represents the hydroxyl. Positive values indicate endothermic processes while negative stand for exothemic processes. Results correspond to the calculated isolated molecular moieties in p(2 × 2)-Au(111) (0.25 ML coverage).

TABLE 2: Energy Barriers for Each Elementary Step, Ea in eV, and Imaginary Frequencies at the Transition State in cm-1a system

Au

Au-O

Au-OH

Ea NH3 f NH2 + H ωNH3 f NH2 + H Ea NH2 f NH + H ωNH2 f NH + H Ea NH f N + H ωNH f N + H

2.44 1231 2.38 394 2.08 664

0.72 279 0.37 145 0.58 1238

0.51 303 0.72 1196 0.29 1187

a

The coverages of NHx, oxygen, and hydroxyl groups are 0.25 ML.

Activation steps: NH3 T NH2 + H NH3 + O T NH2 + OH NH3 + OH T NH2 +H2O Dehydrogenation steps: NH2 T NH + H NH2 + O T NH + OH NH2 + OH T NH + H2O NH T N + H NH + O T N + OH NH + OH T N + H2O Termination Steps: N + N T N2 N + O T NO The activation, dehydrogenation, and recombination reaction sets are discussed in the following sections. The thermodynamics for all the represented steps are reported in Table 1 while the kinetic parameters are reported in Tables 2 (dehydrogenation steps) and III (recombination). A. Activation. The initial, transition state, and final structures for the first dehydrogenation of ammonia on the clean, Oprecovered and OH-precovered surface are shown in Figure 1, and the relevant energies are shown in Tables 1 and 2. In all cases, the oxygen, hydroxyl, and NHx species coverage is 0.25 ML. For the clean Au(111) surface, the adsorption of ammonia to the surface is almost thermoneutral (exothermic by 0.05 eV). The N to Au distance is 2.402 Å and the H-N-Au angle is close to 107°. This is similar to the structures previously found for NH3 on Pt(111) and Rh(111).27-31 The first dehydrogenation leads an H atom sitting on an fcc site, the NH2 sits on a bridge site, and the process is highly endothermic (>1.8 eV). Consequently, the calculated energy barrier for H elimination is large, 2.44 eV. At the transition state, the H-N distance is increased to 1.803 Å and reaches 3.356 Å at the final state. Therefore, Au(111) shows a very low tendency to promote NH3 activation and dehydrogenation on this surface, as seen in the experiments.12 As comparison, the barrier for the direct dissociation of ammonia on Pt obtained with a similar setup is lower than

Figure 1. Initial, transition, and final state for the first dehydrogenation on Au(111), top clean surface, center O-precovered surface, and bottom hydroxyl precovered surface. Yellow spheres represent Au atoms, blue represents N, light blue represents H, and red represents O. The coverage of ammonia, oxygen, and hydroxyl groups is 0.25 ML.

1 eV.15,28 The difference in the energy barriers represents more than 10 orders of magnitude in the dehydrogenation rate on Au when compared to Pt at 500 K. The adsorption of O2 molecules to the Au(111) surface has been extensively reported and therefore will not be discussed here. The main result is that the barrier for O2 dissociation is high for this surface and consequently the process is unlikely,32,33 and any oxidation reaction will not take place without previous activation of O2. In the experiments,12 the oxygen dissociation problem is avoided by directly adsorbing O atoms. Such a process is exothermic by -3.12 eV/atom (with respect to O 3P) and marginally endothermic when O gas phase is employed 2 as a reference, 0.07 eV. The adsorbed O atoms are placed at the fcc sites. The presence of preadsorbed O atoms on the Au(111) surface increases the binding energy of ammonia when compared to the clean surface, BE ) -0.38 eV. The reason behind the stabilization is the formation of donor-acceptor pairs in the NH3-O substructure. In the coadsorbed state, the N-Au distance is shortened to 2.393 Å and the H-N-Au angle increased to 109°. Moreover, there is a H-bridge between one of the hydrogen atoms of ammonia and the O on the surface, and the shortest H-O distance is 2.752 Å. Similarly, OH groups on the surface (adsorbed on top sites) stabilize an ammonia molecule in the neighboring site with the binding energy of ammonia being -0.44 eV. For this structure, a tight network of H bonds and donor-acceptor couples is formed and is responsible for the large binding energy found. In the coadsorbed state, the N-Au distance is 2.312 Å, the H-N-Au angle is 109°, and the shortest H-O distance 2.170 Å (see bottom left of Figure 1). In the coadsorbed structures, both the donating (NH3) and the abstracting groups (O, OH) are present and can lead to first hydrogen stripping. When O is the stripping agent, the process is endothermic by 0.38 eV, and the resulting products are NH2 on a bridge site and a hydroxyl group on top an Au atom on the surface. This elementary step is hindered by a barrier of 0.72 eV. At the transition state, the H-N distance is 2.659 Å while the new OH bond is almost completed, the distance being 0.978 Å. The process is more endothermic and with a larger barrier than for Pt(111) (∆E ) 0.2 and Ea ) 0.4 eV);28 this is because the more basic character of O on the Au(111) surface34

NH3 Oxidation on Oxygen-Precovered Au(111)

Figure 2. Initial, transition, and final states for the second dehydrogenation on Au(111), top clean surface, center O-precovered surface, and bottom hydroxyl precovered surface. The coverage of NH2, oxygen, and hydroxyl groups is 0.25 ML. Same color code as in Figure 1.

cannot compensate for the very low tendency of Au to adsorb the NH2 fragment compared to Pt(111). When hydroxyl groups are adsorbed on the surface, the first hydrogen abstraction is endothermic by 0.30 eV with the final products being an NH2 at the bridge site and a desorbed water molecule. The barrier for OH-assisted dehydrogenation is 0.51 eV. At the transition state structure, the H-N distance is 2.380 Å, shorter than that when O is the abstracting agent, while the H-O distance is 0.981 Å, slightly larger than for O. The more compact nature of the donor-acceptor system in the hydroxyl case at the transition state and the final compression relief due to the formation of nascent water molecules are behind the better energetics found in this case. In summary, the barriers for ammonia activation when O or OH are present are less than a third of those corresponding to the direct reaction on the clean surface, which shows that without the assistance of a proton scavenger ammonia dehydrogenation is impossible under Au surface in normal conditions. At high O coverage (0.50 ML), the adsorption of 0.25 ML of ammonia is not favorable anymore, the binding energy of ammonia to the oxygen precovered surface is endothermic by 0.16 eV. This explains why only a fraction of ammonia reacts in the experiments in ref 12 when the O coverage is increased. The rest of the ammonia is desorbed from the surface before dehydrogenation reaction starts. B. Dehydrogenation Steps. The second hydrogen abstraction from ammonia can take place unimolecularly or by any of the fragments on the surface (oxygen and hydroxyl groups). The relevant structures for dehydrogenation steps on clean, O and OH precovered surfaces are shown in Figure 2. The second dehydrogenation step of NH2 on Au(111) is endothermic by 1.79 eV on the clean surface. The barrier for the direct reaction is high (2.38 eV), and therefore will not take place under the conditions described. At the transition state, the relevant N-H distance is 2.294 Å and the final products are atomic H and NH on neighborings fcc sites. For Pt and Rh(111), early reports indicate reaction barriers of 1.10 and 1.07 eV, respectively.29 In contrast, when O atoms are present the reaction is less endothermic (0.21 eV), and the barrier for H stripping from preadsorbed O atoms is 0.37 eV. At the initial state, the O atom is close to a hydrogen atom from the NH2 moiety, the distance being 2.239 Å. At the transition state, the O atom is activated

J. Phys. Chem. C, Vol. 112, No. 1, 2008 249

Figure 3. Initial, transition, and final state for the third dehydrogenation on Au(111), top clean surface, center O-precovered surface, and bottom hydroxyl precovered surface. The coverages of NH, oxygen, and hydroxyl groups is 0.25 ML. Same color code as in Figure 1.

and the O-H distance is reduced to 0.978 Å while the N-H bond is enlarged to 2.497 Å. At the final state, NH is sitting on a hollow site while the hydroxyl group is on top Au. Hydroxyl groups on the surface can also produce hydrogen stripping. Starting from NH2, dehydrogenation and water formation shows a barrier of 0.72 eV with the reaction being endothermic by 0.39 eV. At the starting coadsorbed structure, OH + NH2, the shortest distance between a H from the NH2 moiety and the OH group is 1.721 Å. At the transition state, this distance is reduced to 1.257 Å and the H-N enlarged to 1.285 Å. In the final state, the water molecule is desorbed from the surface while the remaining NH is placed on a fcc hollow site. The complete dehydrogenation can take place again by a unimolecular or a bimolecular reaction; see Figure 3. NH direct decomposition is prevented by the low affinity of gold surfaces for the hydrogen atoms, and therefore the reaction is endothermic by about 1.7 eV and the barrier amounts 2.08 eV. In comparison, on Pt and Rh(111) the barriers for this process are only 1.18 and 1.15 eV, respectively.28,29 As for the structures, in the initial state the N-H distance is 1.023 Å, which is enlarged to 1.932 Å at the transition state. In the final state, both atoms occupy fcc centers. The corresponding reaction starting from O atoms and the NH fragment on fcc neighboring sites on the surface is exothermic by 0.31 eV, and the barrier amounts to 0.58 eV. The H-O distance in the initial configuration is 3.098 Å, which is reduced to 1.347 Å at the transition state while the N-H distance increases from 1.024 to 1.210 Å. In the final state, N remains in an fcc site while the hydroxyl group is placed on top Au. The reaction of NH (fcc) with the OH (on top) fragments on the surface is exothermic by 0.47 eV, while the barrier is 0.29, which is the smallest of all the calculated barriers. The relevant geometric parameters for such process are the H-O distances: in the initial state configuration the distance is 3.419 Å, which gets reduced to 1.294 Å at the transition state and to 0.976 Å at the final state. Simultaneously, the N-H distance starts at 1.023 and is enlarged to 1.256 and 2.752 Å at the transition and final states. In the final state, the water molecule is desorbed, and the N atom remains at the fcc site.

250 J. Phys. Chem. C, Vol. 112, No. 1, 2008

Lo´pez et al.

Figure 4. N-N recombination energy plots at two different coverages: p(2 × 2) and p(3 × 3). Relevant energies are shown in eV.

TABLE 3: Diffusion Energies, ∆E, and Recombination Barriers, Ea, Both in eV for N2 and NO at Two Coveragesa energy

system

supercell

X)N

X)O

∆E ∆E Ea Ea Ea

N + X far f N + X close N + X far f N + X close N + X close f NX N + X close f NX N + X far f NX

p(2 × 2) p(3 × 3) p(2 × 2) p(3 × 3) p(2 × 2)

0.52 0.04 0.68 (464) 0.95 1.20

0.62 0.04 0.86 (411) 1.03 1.48

a Imaginary frequencies for the recombination steps are shown in parenthesis in cm-1.

C. Recombination. After the nitrogen atoms have been formed on the surface, there are two ways they can be eliminated: by either N-N or N-O recombinations. The relevant structures for the finalization steps of the N2 reaction are shown in Figure 4 and are equivalent to the NO reaction. The corresponding energies are displayed in Tables 1 and 3. The N-N recombination is exothermic by about 6 eV and the resulting N2 molecule readily leaves the surface. The barrier for the recombination has been calculated in the p(2 × 2) unit cell, corresponding in this case to a total N coverage of 0.5 ML. From the energy difference between the p(2 × 2) supercells containing one and two N atoms, the intrinsic energy cost of bringing the two atoms together can be estimated. In this case, the energy for the formation of the initial configuration is 0.52 eV. However, this estimate contains a term coming from the change in the overall nitrogen coverage (from 0.25 to 0.50 ML). The strong repulsion found is because each N shares two metal atoms with other two N in the initial configuration,35 and this situation is presented twice for each p(2 × 2) unit cell. However, such an approach has been employed extensively28 to obtain lateral interactions. The N-N recombination barrier from the high coverage 0.50 ML initial configuration is only 0.68 eV. To test the effect of the change of coverage in the N2 recombination, we have employed a larger p(3 × 3) supercell. Under these conditions the N coverage is kept close to the 0.25 ML used in the experiments. The effect of diffusion and reaction processes has been analyzed by performing two chained NEB calculations. In the initial configuration of the first process, the two N atoms are far apart by 2 fcc sites in the p(3 × 3) unit cell; see Figure 4. There is a diffusion process leading to two N atoms in neighboring fcc sites that involves two barriers of 0.24 and of 0.64 eV (calculated at the NEB level). It is important to notice that while the first diffusion barrier follows the results obtained for the clean surface,36 the second is affected by the

presence of the second N atom. This final configuration with the 2 N atoms sitting in neighboring fcc sites is less stable than the long distance N-N configuration by 0.04 eV. This points out that in the low-coverage range (i.e., below 0.25 ML) lateral interactions are small.37 From the N-N coadsorbed state, the recombination barrier is 0.95 eV. This barrier is smaller than that reported for other metals.38 Finally, as for the geometries at the transition state the N-N distance is 2.080 Å for the p(2 × 2) unit cell, and the N atoms are sitting almost at two contiguous bridging sites. For the lower coverage, the structure is qualitatively similar (both atoms at neighboring bridging sites) but the N-N distance is further reduced to 1.921 Å. Moreover, at high coverage the Au in contact with both N atoms is lifted from the surface plane by about 0.65 Å. A smaller extraction from the surface plane, 0.58 Å, is found when the larger supercell is employed. Similar effects have been found for CH4 dissociation on Ir(111).39 The competitive path, that is, NO formation, can be obtained from adsorbed N and O atoms on the surface; the final product is an adsorbed NO molecule on top an Au atom on the surface,40,41 which is very weakly adsorbed, -0.11 eV, and slightly tilted. The energy needed to bring the N and O reactants together in the small cell p(2 × 2) from two separated p(2 × 2) calculations is 0.62 eV. The recombination barrier from the coadsorbed state is 0.86 eV, which is larger than the barrier observed for the N2 formation. When inspecting the p(3 × 3) results and starting from a configuration where the N and O atoms are far away in fcc centers, the barriers for diffusion are 0.40 and 0.70 eV if the diffusing entity is O and slightly smaller for N. The final configuration with N and O sitting on neighboring fcc sites is 0.05 eV more unstable than the far apart configuration. From the neighboring sites, the barrier for NO formation is 1.03 eV. As for the geometries, at high coverage the N-O distance at the transition state is 2.044 Å while at low coverage it is 2.024 Å, whereas in the final state the distance is 1.175 Å. Therefore, N-N or N-O repulsion is strongly dependent on the surface coverage. The range of lateral interactions starts at 0.05 eV in diluted systems (0.25 ML), but when neighbors are present at high coverages a 0.30 eV contribution per occupied nearest neighbor structure can be estimated with respect to lowcoverage adsorption energies. This agrees well with the values obtained for NO on Rh(100) and illustrates how DFT results can help in the determination of lateral interactions.37,42 Moreover, because NO is weakly adsorbed on the Au(111) surface, it is unlikely to react with other coadsorbed N or O on the surface to form other nitrogen oxides. IV. Discussion Bradley, Hopkinson, and King devised a model (BHK model) for the selectivity of ammonia oxidation on Pt(100)17 indicating that the reaction system depends directly on the relative coverage of ammonia and oxygen on the surface. Because ammonia decomposes, the model indicates that only N and O coverages on the surfaces are responsible for the observed selectivity toward N2 and/or NO. In this section, a similar model will be derived to describe the experimental results in ref.12. As seen in the calculations, ammonia hardly adsorbs on the inert Au(111) surface, and adsorption is assisted by the presence of O atoms on the surface to which NH3 forms hydrogen bonds. Oxygen atoms (or hydroxyl groups) on the surface of Au(111) are needed in order to strip H atoms. Then, the issue of selectivity is related to the competition between the recombinative steps, N + O and N + N. For Au(111), NO is weakly

NH3 Oxidation on Oxygen-Precovered Au(111)

J. Phys. Chem. C, Vol. 112, No. 1, 2008 251

Figure 5. Selectivity toward N2 in Au(111) at 450 K expressed as a function of the relative O/N coverage. The black solid line represents the model with the barriers for p(3 × 3), the red dashed line is that of p(2 × 2), and the green dotted line is that of p(2 × 2) including the initial repulsion. The data from ref 12 (blue squares) is also included for comparison.

adsorbed on the surface in contrast to other more active metals (i.e., Pt), and therefore the selectivity toward N2, s, can be written as

s)

r N2 rN2 + rNO

)

kN2 kN2 + kNO(θO/θN)

(1)

where rN2 and rNO are the rates for N2 and NO formation, respectively, and kN2 and kNO are the corresponding constants. On the other hand, the kinetic constants can be written as Arrhenius equations in the form: kXY ) A exp(-EaXY/RT). If we assume that (i) the prefactors in the rate constants are similar, (ii) the reaction paths and structures are very similar, and (iii) the mass of the reactants are similar, the simplest expression for the selectivity can be written as

s)

exp(-EaN2/RT) exp(-EaN2/RT) + exp(-EaNO/RT)(θO/θN)

(2)

Therefore, we can describe the selectivity as a function of the relative oxygen to nitrogen coverage. The values for the barriers, Ea, are taken from our calculated data, and the temperature employed is 450 K following the experimental conditions described in reference.12 Figure 5 shows that exclusive formation of N2 is observed at low O coverages. This would be the desired regime for the selective ammonia oxidation toward N2 in agreement with the experiments.12 Moreover, we have employed the results for p(3 × 3) and p(2 × 2) to obtain the effect of coverage on the selectivity. In fact, at high coverages N2 is more likely to be formed than in the diluted case. The selectivity toward N2 is even larger when the barriers employed are those of p(2 × 2) but with respect to the isolated atoms calculated on p(2 × 2) structures. In addition, the black line representing the selectivity at low coverages is in good agreement with the experimental results represented by the two squares in Figure 5. These points have been extracted from the data in ref 12 and correspond to (1) 0.25 ML O and 0.18 ML ammonia and (2) 0.64 ML O and 0.25 ML ammonia. In the first case, 89% of the ammonia reacts giving 100% selectivity toward N2, whereas in the second set of experiments only 53% of ammonia is oxidized. Because we

Figure 6. Effect of the temperature on the selectivity toward N2 in Au(111) expressed as a function of the relative O/N coverage. The data from ref 12 blue squares, is also included for comparison.

know from the DFT calculations that O and OH groups are needed for the complete ammonia dehydrogenation and that dehydrogenation steps take place at lower temperatures than recombination, a detailed mass balance can be applied to both experiments. In the first case, 0.18 ML of ammonia would stoichiometrically need 0.27 ML for the complete dehydrogenation (and concomitant atomic N adsorption). This means that the surface is no longer covered by O and thus the selectivity would be 100% with an oxygen coverage of zero. The total yield would be smaller because a part of the ammonia would not find enough O atoms or hydroxyl groups to completely dehydrogenate. This explains that the yield observed is lower than 100%. In the second case, the nominal O coverage is 0.64 ML and ammonia conversion 53%. Because we know that the unreacted ammonia is desorbed from the surface (as our calculations for high O-coverage indicate), the coverage of N atoms on the surface is 0.13 ML, and ammonia oxidation consumes 0.20 ML of the oxygen on the surface. Therefore, before recombination the N coverage is 0.13 while that of O is 0.44 ML. Under these conditions ca. 77% selectivity toward N2 was obtained in the experiments. In summary, DFT calculations do provide a clear picture, even up to almost quantitative values, for the selectivity toward N2 observed. We have performed the same kind of selectivity analysis at different temperatures; see Figure 6. In the plot, the effect of the temperature is as expected: the higher the temperature, the lower the selectivity toward N2. Thus, higher selectivities could be achieved at lower temperature operation. V. Conclusions We have studied the dehydrogenation and oxidation of ammonia on oxygen precovered Au(111) by means of DFT. O atoms or hydroxyl groups are the stripping agents needed for NH3 dehydrogenation on Au(111). The reason being is that the high basic character of O atoms adsorbed on this surface. Whenever OH groups are formed on the surface, they are eliminated; therefore these fragments are short-life species. All the ammonia decomposition steps on Au(111) have larger barriers than those on Pt.28 On the contrary, recombination processes are much less energetically demanding. The reason for such behavior is the inertness of gold, that does not favor the formation of bonds between molecule moieties and the Au(111) surface.43 In addition, all oxygen or hydroxyl-assisted dehydrogenation steps have lower barriers than the recombination ones under relevant conditions. N2 recombination is

252 J. Phys. Chem. C, Vol. 112, No. 1, 2008 hindered by a barrier of 0.95 eV whereas that of NO is 1.03 eV. Thus, NO is never the preferred thermodynamic or kinetic product. Finally, NO adsorbs only weakly on the metal surface, and readsorption or recombination with other N, O atoms on the surface are unlikely events. Moreover, the effect of coverage on N2 or NO recombination barriers is large; N2 formation at high N coverage or NO formation under high N or O coverage are hindered by smaller barriers (the effect is on the order of 0.3 eV). In that case, the barrier difference is even more favorable toward N2 formation. Finally, ammonia and oxygen mixtures on Au(111) when oxygen is below the stoichiometric value for complete dehydrogenation only produce N2, which is in agreement with experiments. When higher O coverages are employed, the selectivity observed can be traced back to the barriers of the competing reactions (i.e., NO and N2 formation) but considering that a stoichiometric amount of oxygen is needed for the dehydrogenation of ammonia with the corresponding water formation. Our results point out to the possibility of employing density functional theory-based calculations for the study of complex reaction schemes and complex properties like selectivity. Furthermore, the results warn about the incorrect use of small unit cells and different coverage to evaluate lateral interactions that may result in too large repulsions. Acknowledgment. We thank the ICIQ foundation, MEC (CTQ2006-00464BQU, CSD2006-003), and GenCat for financial support and BSC for computational resources. Professor Pe´rez-Ramı´rez is acknowledged for fruitful discussions. References and Notes (1) Haruta, M. Catal. Today 1997, 36, 153. (2) Hashmi, A. S. K.; Hutchings, G. J. Angew. Chem., Int. Ed. 2006, 45, 7896. (3) Remediakis, I. N.; Lopez, N.; Nørskov, J. K. Appl. Catal., A 2005, 291, 13. (4) Segura, Y.; Lopez N.; Pe´rez-Ramı´rez, J. J. Catal. 2007, 247, 383. (5) Corma, A.; Serna, P. Science 2006, 313, 332. (6) Chen, M. S.; Goodman, D. W. Science 2004, 306, 252. (7) Remediakis, I. N.; Lopez, N.; Nørskov, J. K. Angew. Chem., Int. Ed. 2005, 44, 1824. (8) Matthey, D.; Wang, J. G.; Wendt, S.; Matthiesen, J.; Schaub, R.; Laegsgaard, E.; Hammer, B.; Besenbacher, F. Science 2007, 315, 1692. (9) Hammer, B.; Nørskov, J. K. Nature 1995, 376, 238. (10) Bahn, S. R.; Lopez, N.; Nørskov, J. K.; Jacobsen, K. W. Phys. ReV. B 2002, 66, 081405. (11) Deng, X. Y.; Friend, C. M. J. Am. Chem. Soc. 2005, 127, 17178. (12) Gong, J. L.; Ojifinni, R. A.; Kim, T. S.; White J. M.; Mullins, C. B. J. Am. Chem. Soc. 2006, 128, 9012.

Lo´pez et al. (13) Kim, T. S.; Gong, J.; Ojifinni, R. A.; White, J. M.; Mullins, C. B. J. Am. Chem. Soc. 2006, 128, 6282. (14) Pe´rez-Ramı´rez, J.; Kondratenko, E. V.; Kondratenko, V. A.; Baerns, M. J. Catal. 2004, 227, 90. (15) Abild-Pedersen, F.; Greeley, J.; Studt, F.; Rossmeisl, J.; Munter, T. R.; Moses, P. G.; Skulason, E.; Bligaard, T.; Nørskov, J. K. Phys. ReV. Lett. 2007, 99, 016105. (16) Weststrate, C. J.; Bakker, J. W.; Rienks, E. D. L.; Martı´nez, J. R.; Vinod, C. P.; Lizzit, S.; Petaccia, L.; Baraldi, A.; Nieuwenhuys, B. E. J. Catal. 2005, 235, 92. (17) Bradley, J. M.; Hopkinson, A.; King, D. A. J. Phys. Chem. 1995, 99, 17032. (18) Kim, M.; Pratt, S. J.; King, D. A. J. Am. Chem. Soc. 2000, 122, 2409. (19) Lescop, B.; Galtayries, A.; Fanjoux, G. J. Phys. Chem. B 2004, 108, 13711. (20) (a) Kresse, G.; Hafner, J. Phys. ReV. B 1993, 47, 558. (b) Kresse, G.; Hafner, J. Phys. ReV. B 1994, 49, 14251. (c) Kresse, G.; Furthmu¨ller, J. Comput. Mat. Sci. 1996, 6, 15. (d) Kresse, G.; Furthmu¨ller, J. Phys. ReV. B 1996, 54, 11169. (21) Hammer, B.; Hansen, L. B.; Nørskov, J. K. Phys. ReV. B 1999, 59, 7413. (22) Blo¨chl, P. E. Phys. ReV. B 1994, 50, 17953. (23) Kresse, G.; Joubert, D. Phys. ReV. B 1999, 59, 1758. (24) Web of Elements, www.webelements.com, (accessed 2007). (25) Monkhorst, H. J.; Pack, J. D. Phys. ReV. B 1976, 13, 5188. (26) Henkelman, G.; Uberuaga, B. P.; Jonsson, H. J. Chem. Phys. 2000, 113, 9901. (27) Garcı´a-Herna´ndez, M.; Lopez, N.; Moreira, I.; de, P. R.; Illas, F. Surf. Sci. 1999, 430, 18. (28) Offermans, W. K.; Jansen, A. P. J.; van Santen, R. A. Surf. Sci. 2006, 600, 1714. (29) Popa, C.; Offermans, W. K.; van Santen, R. A.; Jansen, A. P. J. Phys. ReV. B 2006, 74, 155428. (30) Novell-Leruth, G.; Valca´rcel, A.; Clotet, A.; Ricart, J. M.; Pe´rezRamı´rez, J. J. Phys. Chem. B 2005, 109, 18061. (31) Novell-Leruth, G.; Valca´rcel, A.; Perez-Ramirez, J.; Ricart, J. M. J. Phys. Chem. C 2007, 111, 860. (32) Mavrikakis, M.; Stoltze, P.; Nørskov, J. K. Catal. Lett. 2000, 64, 101. (33) Lopez, N.; Janssens, T. V. W.; Clausen, B. S.; Xu, Y.; Mavrikakis, M.; Bligaard, T.; Nørskov, J. K. J. Catal. 2004, 223, 232. (34) Torres, D.; Lopez, N.; Illas, F.; Lambert, R. M. Angew. Chem., Int. Ed. 2007, 46, 2055. (35) Hammer, B. Phys. ReV. Lett. 1999, 82, 3681. (36) Nilekar, A. U.; Greeley, J.; Mavrikakis, M. Angew. Chem., Int. Ed. 2006, 45, 7046. (37) Hermse, C. G. M.; van Bavel, A. P.; Niewenhuys, B. E.; Lukkien, J. J.; van Santen, R. A.; Jansen, A. P. J. Langmuir 2005, 21, 8302. (38) Go´mez-Dı´az, J.; Pe´rez-Ramı´rez, J.; Lopez, N. in preparation. (39) Henkelman, G.; Jonsson, H. Phys. ReV. Lett. 2001, 86, 664. (40) Gajdos, M.; Hafner, J.; Eichler, A. J. Phys.: Condens. Matter 2006, 18, 13. (41) Gajdos, M.; Hafner, J.; Eichler, A. J. Phys.: Condens. Matter 2006, 18, 41. (42) Hermse, C. G. M.; Frechard, F.; van Bavel, A. P.; Lukkien, J. J.; Niemantsverdriet, J. W.; van Santen, R. A.; Jansen, A. P. J. J. Chem. Phys. 2003, 118, 7081. (43) Lopez, N.; Nørskov, J. K. J. Am. Chem. Soc. 2002, 124, 11262.