Theoretical Study of CO and NO Chemisorption on RhCu (111) Surfaces

Parc Cientı´fic de Barcelona, C/Martı´ i Franque`s 1, 08028 Barcelona, Spain, and ... NO reduction by CO in the three way converter in car ... aut...
0 downloads 0 Views 202KB Size
Download PDF
4654

J. Phys. Chem. B 2005, 109, 4654-4661

Theoretical Study of CO and NO Chemisorption on RhCu(111) Surfaces Silvia Gonza´ lez,†,‡ Carmen Sousa,† and Francesc Illas*,† Departament de Quı´mica Fı´sica i Centre Especial de Recerca en Quı´mica Teo` rica, UniVersitat de Barcelona i Parc Cientı´fic de Barcelona, C/Martı´ i Franque` s 1, 08028 Barcelona, Spain, and Departamento de Quı´mica, UniVersidad Auto´ noma Metropolitana-Iztapalapa, P.O. Box 55-534, C.P. 09340 Me´ xico D.F., Me´ xico ReceiVed: October 8, 2004; In Final Form: January 10, 2005

A systematic density functional theory study using periodic models is presented concerning the chemisorption of CO and NO on various sites of RhCu(111) surfaces. The properties of the adsorbed molecules on various mono- and bimetallic sites of these alloy surfaces have been obtained and compared to those corresponding to the pure Rh(111) and Cu(111) surfaces. It is shown that that the interaction of small probe molecules such as CO or NO on RhCu alloys is essentially dominated by the atomic nature of the surface active site with little influence of the rest of the metallic system. Moreover, it is suggested that it is possible to control the adsorption site of these molecules by appropriate choice of the surface composition.

I. Introduction NO reduction by CO in the three way converter in car exhausts is one of the most important catalytic reactions in the automobile industry. It eliminates two polluting gases and produces two inoffensive ones, CO2 and N2,1,2 which are already components of clean air. The interest in this reaction has been growing by the need to reduce atmosphere pollutants all over around the world. Various metals have been used to catalyze this reaction; the best performance is achieved by Rh or Pt.1-3 However, these precious metals are both very expensive, and indeed, both are easily poisoned. Therefore, it is desirable to modify their chemistry, for instance by using a second metal. The idea is to obtain a bimetallic catalyst with better performance.1,4-6 This is to reduce fabrication cost and to decrease poisoning. To this end, several bimetallic catalysts such as PtRh,7 PtSn,1,4 PdMn,8 PdCu,9 and RhCu5 have been studied either in experiments or in computational models. The bimetallic PtRh system is the essential component of the catalyst under current use since it exhibits the best chemical activity toward NO reduction and hydrocarbon and CO oxidation.1,7 Still, this catalyst involves two precious metals, thus resulting in a too high production cost. Hence, there is an important motivation to pursue research on other bimetallic catalysts. In particular, RhCu is of interest because it has been suggested that its activity toward NO reduction by CO is even higher than that of Rh5 although the reasons behind this behavior are essentially unknown. Moreover, many researches have reported peculiar catalytic properties of this alloy. For example, a small amount of Cu on Rh catalysts increases the H2 uptake until a maximum adsorption is reached. However, a larger Cu concentration decreases the amount of adsorbed H2.10-15 Recently, a few experimental and theoretical studies have proposed explanations for the extraordinary catalytic properties of this alloy, namely change of adsorption site with respect to pure Rh provoked by the presence of Cu atoms,11,16 charge transfer between the two metals13,17 and changes in the electronic density of the alloy as compared to pure Rh.18 * Corresponding author. E-mail [email protected]. † Universitat de Barcelona i Parc Cientı´fic de Barcelona. ‡ Universidad Auto ´ noma Metropolitana-Iztapalapa.

The adsorption of probe molecules such as CO or NO on various metallic and bimetallic surfacessCu, Rh and RhCu among themshas been extensively studied with the aim to infer the surface reactivity and has also been studied because it constitutes a previous step in some catalytic reactions.19-21 In particular, the interaction of CO with Rh surfaces has been studied in detail because this metal is an excellent oxidation catalyst22-25 and is also active for NO dissociation.2,3 This special chemical activity of Rh surfaces has triggered a considerable number of research papers concerning CO and NO adsorption on Rh surfaces, either from experimental26-28 or theoretical18,29-33 sides. Similarly, Cu is another metal catalytically active in many reactions1,34-36 including NO reduction by CO oxidation.37 The interaction of CO with various bimetallic alloys, and in particular with RhCu, has been studied by Rodriguez et al.17,38 These authors found a linear correlation between changes in the CO chemisorption energy and shifts in surface core-level binding energy. The presence of the Cu surface atoms in the Rh substrate results in a CO adsorption energy which is larger than that corresponding to monometallic Cu. Hence, Rh activates Cu adatoms by a strong perturbation of the electronic density due to the formation of the heterometallic bond. For Cu monolayers on electron-rich metals such as Rh, X-ray photoelectron spectroscopy measurements suggest that the electron density of Cu atoms increases; therefore, one may infer that the π-back-donation capacity with respect to a surface atoms in Cu(100) is enhanced and therefore one would expect that the CO-Cu interaction becomes more favorable.17 The existence of the linear correlation between metal core level shifts and CO adsorption energy has also been predicted from ab initio cluster model calculations on PdCu alloys.39 However, a decomposition of the interaction energy based on the constrained space orbital variation method40-42 shows that this is not simply due to an enhancement of the π-back-donation from the metal to CO, the Pauli repulsion playing also a key role. Another important conclusion reached by Rodriguez and Goodman38 is that the CO vibrational frequency is not representative of the composition of bimetallic surfaces. Finally, it is worth pointing out that kinetic studies of CO oxidation on Cu supported on Rh(100) at various reaction conditions permit

10.1021/jp0454016 CCC: $30.25 © 2005 American Chemical Society Published on Web 02/17/2005

CO and NO Chemisorption on RhCu(111) Surfaces one to conclude that the presence of Cu on Rh(100) enhances the CO oxidation catalytic activity when compared to that of Rh(100).43 This improvement in the catalytic activity is attributed to an increase in the surface oxygen coverage on Cu/ Rh with respect to the Rh(100) surface. Despite the potential activity of RhCu described among others by Rodriguez and Goodman17,38 and by Ferna´ndez-Garcı´a et al.,14 the number of experimental studies involving this system is rather limited.10-15,17 The number of theoretical studies is even smaller.16,18 The present paper reports a systematic theoretical study of the interaction of CO and NO on Rh(111), on Cu(111) and on two RhCu(111) surfaces mimicking different alloy compositions. To this end, density functional theory (DFT) periodic calculations with large unit cells have been carried out. Changes in the interaction energy, adsorption geometry and CO or NO vibrational frequencies with respect to the composition of surface are analyzed. The present DFT calculations establish that, for the considered composition range, changes in the analyzed adsorption properties with respect to the composition of surface are rather small, indicating that the site-molecule interaction dominates over the metallic environment. II. Computational Details The DFT calculations were performed using the VASP (Vienna ab initio simulation package) computational code44,45 implemented in a series of parallel computers. The XCrySDen program46,47 (version 0.9.3) has been used to visualize and represent the structures of interest. Exchange and correlation energies were estimated using the Perdew-Wang48 (PW91) version of the generalized gradient approximation (GGA). The projected augmented wave (PAW) method49,50 was used to reproduce the atomic cores effects in the electronic density of valence electrons. The PAW is essentially an all electron frozen core method combining the accuracy of all electron methods, such as the full potential linearized plane wave method, and the computational simplicity of the pseudopotential approach. This is especially the case in the realization of Kresse and Joubert.51 The Kohn-Sham one-electron wave functions were expanded in a basis of plane waves with kinetic energy below 415 eV. The residual minimization method direct inversion in the iterative subspace (RMM-DIIS) was employed as electronic minimization algorithm. The quasi-Newton algorithm was used to relax the atomic positions in the geometry optimization. To determine how the partial occupancies are set for each Bloch function, the order 2 Methfessel-Paxton scheme was used, and calculations were carried out using a kBT ) 0.2 eV smearing of the electron density in the corresponding one electron wave functions, but the total energy was extrapolated to kBT ) 0 eV upon convergence of the self-consistent field procedure. Harmonic frequencies were obtained from the forces of each adsorbate in the three spatial coordinates using a finite difference scheme. The spin-restricted formalism has been used although NO shows spin polarization; however, this comparative study does not pretend to obtain accurate absolute values but trends in adsorption properties with respect to the surface composition. In any case, several tests indicate that taking spin polarization into account results in a very minuscule effect, ∼0.0018 kcal mol-1 in the total energy. Therefore, for the present purposes it is possible to consider that spin polarization effects in the energy of the systems are negligible. III. Surface Models Rh, Cu, and RhCu surfaces were modeled by the periodic supercell approach. This model includes a two-dimensional

J. Phys. Chem. B, Vol. 109, No. 10, 2005 4655 metallic slab with a vacuum region and the slab plus vacuum are repeated in the third direction for computational convenience. The Cu and Rh lattice parameters were optimized with respect to the energy of each bulk metal. The optimized values, 3.63 and 3.85 Å, are sufficiently close to the experimental ones, 3.62 and 3.80 Å, respectively.52 The slabs used to model mono and bimetallic surfaces have five atomic layers, although it was found that, for most purposes, four layers provide an accurate enough representation of the Rh surface. The two uppermost layers of the slab were allowed to relax completely and the remaining three were fixed at bulk distances to provide an adequate environment to the surface atomic layers. The vacuum region between the periodically repeated slabs was set to 10 Å which is larger than the vertical distance corresponding to four atomic layers. The Brillouin zone of the surface unit cell was described with a 5 × 5 × 1 Monkhorst-Pack mesh. The parameters described above have been found to provide converged values for the adsorption of CO on the Rh(111) surface and have also been used to represent Cu(111) and the different RhCu surface models. The present settings are similar to those used in previous studies.31,33,53,54 In the present periodic models of the (111) surface of mono and bimetallic systems a (4 × 4) supercell has been chosen, each atomic layer contains 16 metal atoms and, hence, the unit cell used in the calculations involves a total of 80 metal atoms. This large supercell guarantees that in the RhCu models there are always metal surface atoms with the coordination corresponding to the bulk metal. A (2 × 2) supercell is still too limited to permit this representation. The bimetallic compositions were determined following the alloy phase diagram.55 This indicates a very low miscibility and hence Rh- or Cu-rich compositions. To model a Rh-rich bimetallic system (hereafter denoted as Cu/Rh) a surface Rh atom was substituted by a Cu atom. A similar strategy was used to model the Cu-rich (Rh/ Cu) alloy. The lattice parameter of the corresponding alloys was not reoptimized nor interpolated from Rh and Cu bulk distances. This is justified since the bimetallic models do only differ from the corresponding monometallic systems by a single surface atom. The relaxation in the interlayer distances in the surface models, with respect to the bulk distance is denoted by % ∆12, this is the change in the distance between the first and the second layer of the slab with respect to the unrelaxed distance and given as a percent value. Similarly, % ∆23 is the percentage change in the distance between the second and third layer. For both metals, the effect of surface relaxation is very small; the two uppermost layers of Rh surface suffer a compression of 1.90% (∼0.05 Å) with respect to the Rh bulk spacing in good agreement with an experimental value of 1.4%. The effect on ∆23 is even smaller, 0.4% or 0.01 Å. In the case of Cu, surface relaxation is almost negligible, the first layer is slightly compressed % ∆12 ) 0.04 or ∆12 ∼ 0.001 Å. In the monometallic surfaces, the substitution of one metal atom by another one of a different element, for instance one Cu atom replacing a Rh one in the Rh(111) surface causes a buckling. On the average, the extent of this buckling is similar to the relaxation of the bare Rh(111) and, hence, needs to be taken into account. Moreover, in Rh/Cu, the buckling is larger than the Cu surface relaxation. Still, the resulting absolute effect is very small. In both cases, the metal impurity tends to move outward the surface, 1.1% for Cu and even less for Rh (Table 1). This is in agreement with findings on other Cu containing bimetallic alloys, the Cu atom impurity always tends to segregate more than the other component.56 Here, % ∆Rh or % ∆Cu would

4656 J. Phys. Chem. B, Vol. 109, No. 10, 2005

Gonza´lez et al.

TABLE 1: Summary of Calculated Interlayer Relaxation for the Mono- and Bimetallic Systemsa % ∆12 % ∆23 % ∆Cu

Rh

Cu/Rh

-1.9 (-1.420) -0.4

-2.1 -0.4 +1.1

% ∆12 % ∆23 % ∆Rh

Cu

Rh/Cu

-0.04 (-0.720) 0.0

-0.4 -0.5 +0.2

a % ∆12, % ∆23 are the average relative relaxation of the outermost atomic layers; % ∆Rh and % ∆Cu are a measure of the surface rumpling. Available experimental values are given in parentheses.

Figure 1. Top view of possible adsorption sites for CO and NO on monometallic and bimetallic surface models. (a) Possible sites of adsorption on the Rh(111) or Cu(111) surface. (b) Possible adsorption sites for the Cu/Rh model (Rh atoms are light gray and Cu atoms are dark gray). A similar definition has been used for the Rh/Cu alloy.

correspond to the % ∆12 parameter of the monometallic surfaces but here the total effect is measured from the position of the impurity atom. Hence, % ∆Rh indicates the change of the vertical distance between the Rh atom and the center of the metal layer immediately beneath it. A similar definition for % ∆Cu is used in the Cu/Rh model. The interaction of NO and CO with the different surface models has been studied by placing the molecule just on one side of the slab model. With the supercell size describe above this results in a CO or NO surface coverage of 0.0625. An upright geometry with the N or C atoms pointing the metallic atoms of the uppermost layer of the slab has been considered. For the adsorption of CO, atop and bridge sites were considered because it is generally accepted that both sites are almost isoenergetic at high coverave although the PW91 functional tends to incorrectly favor the three-hollow sites.57,58 In fact, CO adsorption on the atop site is the experimentally observed site for both Rh and Cu surfaces and, hence, the most stable one, at least at low coverage20,27,59 whereas the present DFT calculations do also predict the hollow site as slightly more stable as previously found for CO on Pt(111).58 On the contrary, NO prefers multicoordinated sites,20,59 in the case of the pure Rh or Cu(111) surfaces, the possible sites are 3-fold (fcc and hcp) and bridge hollows. Nevertheless, the atop site is included to complete these study. Bimetallic surfaces exhibit an increased number of possible active sites depending on the distance of the impurity atom to the actual site or on whether the site itself involves the impurity atom. Therefore, in the bimetallic models, in addition to the sites considered for the monometallic surfaces, one needs to take into account near or far (as in monometallic surfaces) to Cu on Cu/Rh or to Rh on Rh/Cu and bimetallic 2and 3-fold hollows. Figure 1 shows all the possible active sites considered on the monometallic (Figure 1a) and bimetallic surfaces (Figure 1b). IV. CO Adsorption on Monometallic Sites of the Rh, Cu, and RhCu Surface Models Adsorption energy (Eads), distance from the molecule to the metal surface, molecular structure and the corresponding CO

stretching frequency are the properties chosen to analyze the effect of the surface composition. More in detail, we focus on the active site composition and its environment. To this end, the results of CO adsorption on the different metallic and bimetallic surfaces will be discussed separately. First, we consider adsorption on the pure Rh and Cu surfaces and compare with the abundant experimental literature. Next, CO chemisorption in the different monometallic sites of the bimetallic systems is considered. Table 2 summarizes the results obtained for the properties of CO adsorption when this interaction involves only rhodium atoms either in the Rh surface or in the Rh/Cu and Cu/Rh bimetallic surfaces. Top and bridge sites have been considered whereas adsorption of CO on the threecoordinated sites has not been taken into account for two reasons. On one hand there is no experimental evidence of such coordination either on Rh(111) or Cu(111) at low coverage and, on the other hand, the present DFT computational approach does not properly describe this interaction.57,58 Different situations have been considered for the interaction of CO on top a Rh atom the depending on whether this atom is (a) surrounded only by other Rh atoms, (b) surrounded only by Cu atoms, or (c) surrounded by Rh atoms but with one single Cu atom either near or far from the adsorption site (see Figure 1). These four situations correspond to the four leftmost columns of Table 2. Similarly, for the adsorption on bridge the CO molecule interacts always with two Rh atoms but these can be (d) surrounded only by Rh atoms or (e) surrounded by Rh atoms but with one Cu atom either near of far from one of the two Rh atoms constituting the bridge site (three rightmost columns of Table 2). Table 3 condenses the results obtained for the properties of CO adsorption when this interaction involves Cu and again either on the Cu surface or on the Rh/Cu and Cu/Rh bimetallic models. The same adsorption possibilities have been considered but exchanging the role of the Rh and Cu atoms. Let us first consider the geometry of the CO adsorbed molecule on a Rh on top site (Table 2). The average C-O distance in the different surface models is 1.16 Å with very small deviations (of ∼0.002 Å) on going from Rh to either Rh/ Cu or Cu/Rh. This is very close to the experimental value for CO on Rh(111) which is 1.15 ( 0.07 Å28 and to the values predicted by DFT cluster model calculations.18 Hence, the presence of Cu atoms near the Rh active site does not introduce any significant effect on the geometry of adsorbed CO. A similar situation is encountered when the interaction involves a Cu atom (Table 3). In this case the average CO distance is 1.15 Å again with small variations induced by the composition change. The calculated value is again in good agreement with the experimental value for CO on Cu (100) surface which is 1.13 Å.60 Hence, the CO distance cannot be a probe of the surface composition. In addition, there is a significant effect on the surface geometry, the metal atom interacting directly with CO exhibits an additional buckling when compared to the clean surfaces. A measure of the relaxation effects induced by the adsorbed molecule is given by the % ∆Rh and % ∆Cu values reported in Tables 2 and 3, respectively. For the interaction on top of Rh and Cu, the metal-C distances (dM-C) are both 1.85 Å, very close to the experimental values for the monometallic (111) surfaces; 1.84 ( 0.07 Å28 and 1.90 Å60 for Rh and Cu, respectively. From the structural study one can readily conclude that geometric properties are similar for all compositions, in agreement with previous cluster model studies.18 Interestingly enough, Eads depends essentially on the type of the adsorbing metal atom. In fact, the adsorption energy of CO on the different Rh sites differs by less than 2 kcal mol-1 depending on the

CO and NO Chemisorption on RhCu(111) Surfaces

J. Phys. Chem. B, Vol. 109, No. 10, 2005 4657

TABLE 2: Calculated Properties for CO Adsorbed on Top and on Bridge Sites of Rh, Cu/Rh and Rh/Cu Surfacesa top

bridge Cu/Rh

Cu/Rh

site

Rh

Rh/Cu

Cu near

Cu far

Rh

Cu near

Cu far

% ∆12 % ∆23 % ∆Rh dCO (Å) dRh-C (Å) Eads (kcal mol-1) ν (cm-1)

-2.0 -0.2 +5.3 1.16 (1.1528) 1.84 (1.8428) -45.6 (-38.061) 2028 (207027)

-0.3 -0.7 +2.8 1.16 1.85 -45.1 2027

-2.2 -0.2 +4.2 1.16 1.85 -44.6 2020

-2.2 -0.3 +4.9 1.16 1.84 -43.7 2019

-2.0 -0.2 +4.1 1.18 (1.1520) 2.03 -46.0 1824 (186127)

-2.1 -0.2 +2.3 1.19 2.02 -46.9 1818

-2.0 -0.4 +2.1 1.18 2.02 -42.9 1820

a The CO molecule always interacts directly with Rh atoms although those may have a different environment (see section 4). % ∆12, % ∆23 are the average relative relaxation of the outermost atomic layers and % ∆Rh is a measure of the surface rumpling. dCO and dRh-C, are the internuclear CO distance and the distance from the C atom to the surface nearest metal atom; ν is the adsorbed CO stretching frequency and Eads is the interaction energy. Experimental values are shown between parentheses.

TABLE 3: Calculated Properties for CO Adsorbed on Top and on Bridge Sites of Cu, Cu/Rh, and Rh/Cu Surfacesa top

bridge Rh/Cu

Rh/Cu

site

Cu

Cu/Rh

Rh near

Rh far

Cu

Rh near

Rh far

% ∆12 % ∆23 % ∆Cu dCO (Å) dCu-C (Å) Eads (kcal mol-1) ν (cm-1)

-0.8 -0.3 +2.0 1.15 (1.1360) 1.85 (1.960) -17.1 (-11.762) 2027 (207864)

-2.1 -0.2 +4.2 1.15 1.85 -16.8 2024

-0.6 -0.6 +2.2 1.15 1.85 -14.8 2028

-0.6 -0.6 +1.8 1.15 1.85 -16.0 2029

-0.7 -0.8 +0.8 1.18 1.98 -18.9 (-16.636) 1857 (182075)

-0.6 -0.4 +6.1 1.17 2.00 -16.7 1884

-0.2 -0.4 +6.5 1.17 2.00 -17.3 1884

a The CO molecule always interacts directly with Cu atoms although those may have a different environment (see section 4). % ∆12, % ∆23 are the average relative relaxation of the outermost atomic layers and % ∆Rh is a measure of the surface rumpling. dCO and dCu-C, are the internuclear CO distance and the distance from the C atom to the surface nearest metal atom; ν is the CO stretching frequency and Eads is the interaction energy. Experimental values are shown between parentheses.

composition surface; this is smaller than the difference with respect to the experimental value for the Rh(111) surface; -45 kcal mol-1 vs an experimental value of -38 kcal mol-1.61 The CO adsorption energy on Cu is -16.9 kcal mol-1 which deviates in a similar way from the experimental value of -11.7 kcal mol-1.62 Also in this case, the composition does not largely affect Eads. Although the values for the clean surfaces are overestimated by 7 kcal mol-1, the difference between the two metals is well reproduced. For the interaction of CO with the Rh bridge sites the situation is very similar, the surface composition does not affect the CO internuclear distance although this is somewhat larger than on the top site. There is a small variation on the Rh-C distance with composition but with a very tiny effect on Eads. For the interaction of CO with the Cu bridge sites the same trend is predicted although in this case the presence of Rh atoms on the Cu surface induces a decrease of the adsorption energy of CO. The very small differences in either geometry or Eads induced by the changes in the surface composition are also found in the CO internal vibrational frequency (Tables 2 and 3). Therefore, it is possible to conclude that as far as CO chemisorption on monometallic sites of the RhCu alloy is concerned, the presence of atoms of the minority component either near or far from these active sites does not significantly affect the adsorption properties. Let us end this discussion by pointing out that CO adsorption induces some metal separation of the surface atoms directly interacting with the adsorbate. This surface distortion suggests that the COmetal interaction is strong enough to separate the adsorbing metal atom from other surface metal atoms. This phenomenon is larger in the Rh and Rh-rich surfaces, almost 5 and 4% respectively, than in Cu and Cu-rich ones, which are 2 and 6%; this fact indicates a stronger interaction between Rh and CO

than between Cu and CO, which is consistent with the adsorption energies reported in Tables 2 and 3. The calculated vibrational frequencies for CO on Rh(111), either on top or bridge, are in good agreement with experimental value obtained at CO coverage of 0.25 and 0.75, respectively. In the case of the linear adsorption the deviation is of around 2% but it is reduced to almost zero if one considers the experimental value obtained in the limit of zero coverage which is in the 1995-2015 cm-1 range.27,63 For CO on Cu(111), the calculated vibrational frequency of CO adsorbed on both sitess 2027 and 1857 cm-1 for on top and bridge sitessshows again a 2% deviation with respect to experiment.64 For the on top interaction a rather large value (2100 cm-1) has been predicted from cluster model calculations within the B3LYP functional.18 Interestingly enough, this value is very close to the vibrational frequency of adsorbed CO measured on Cu supported particles which is of ∼2120 cm-1.65,66 This difference between extended surfaces and metal particles does not exist in the case of Rh.67,68 This is in agreement with evidence showing that some properties of chemisorbed CO on extended surfaces and small supported particles are very similar27,61,63,67 and hence the cluster model description may be adequate. In the case of CO on Cu the proper description of the CO vibrational frequency would require very large cluster models. The negligible influence of the surface composition in the calculated values of vibrational frequency of adsorbed CO is in agreement with the results discussed above for the geometry and adsorption energy as well as with experimental results.11,38 V. CO Adsorption on Bimetallic Sites of the RhCu Surface Models This section describes the results obtained for the CO adsorption properties when this molecule interacts on hetero-

4658 J. Phys. Chem. B, Vol. 109, No. 10, 2005

Gonza´lez et al.

TABLE 4: Calculated Properties for CO Adsorbed on Heterometallic Bridge Sites of Cu/Rh and Rh/Cu Surfacesa % ∆12 % ∆23 % ∆RhCu dCO (Å) dRh-C (Å) dCu-C (Å) angle (deg) Eads (kcal mol-1) ν (cm-1)

Cu/Rh

Rh/Cu

-2.1 -0.4 +1.6 1.17 1.87 2.48 27 -40.6 1942 (199913)

-1.4 -0.6 +1.4 1.17 2.07 2.59 23 -42.2 1948

a % ∆12, % ∆23 are the average relative relaxation of the outermost atomic layers and % ∆RhCu is a measure of the surface rumpling. dCO, dCu-C and dRh-C are the internuclear CO distance and the distance from the C atom to the surface nearest Rh and Cu metal atoms; ν is the CO stretching frequency and Eads is the interaction energy. Experimental values are shown in parentheses.

Figure 2. Adsorption geometry of CO on RhCu bridge site in Cu/Rh surface (Rh atoms are light gray, Cu atoms are dark gray, the C atom is black, and the O atom is white).

metallic sites in bimetallic surfaces. The only heterometallic site considered is a bridge one, either in Rh/Cu or Cu/Rh. These two sites differ only in the surface composition; in the Rh/Cu the Rh atom is surrounded by Cu atoms and the opposite holds for the Cu/Rh. Interaction of CO with these sites has some interesting differential features with respect to the monometallic bridge sites described in the previous section (Table 4). First of all, the CO molecule appears always tilted with a deviation of ∼30° with respect to the perpendicular (Figure 2). Moreover, the optimized geometry shows two well different metal-C distances as expected from the different atomic sizes. However, the Rh-C distance in Cu/Rh is significantly smaller than that corresponding to Rh homometallic bridge sites. The opposite is found for the Cu-C distance; it appears to be considerably larger than that corresponding to a clean Cu(111) surface. Another interesting aspect is provided by the adsorption energy. From interpolation arguments one may predict that Eads of CO on Rh/Cu will be similar to that of CO on Cu(111) and, conversely, that the Eads of CO on Cu/Rh will be similar to that of CO on Rh(111).69 The data reported in Table 4 show that this is not the case. The interpolation arguments seem to hold for CO on Cu/Rh but cannot be sustained for CO on Rh/Cu because the calculated Eads values for CO on Rh/Cu and CO on Cu/Rh are extremely close to each other and only slightly lower that the value corresponding to Rh(111) monometallic on top or bridge sites. This seems to suggest that the interaction is dominantly local and driven by the two atoms bonded to CO and independent of the environment of these two atoms. The interpretation above is supported by the analysis of the CO internuclear distance and of the CO vibrational frequency. The corresponding values for CO on Rh/Cu or on Cu/Rh in Table 4 are almost the same. The vibrational frequency is of

particular interest because it is significantly different from the values reported in Tables 2 and 3 and therefore would permit the identification of CO on this heterometallic sites in RhCu alloys. Indeed, this is precisely the case since the calculated values are in very good agreement with experimental results corresponding to RhCu bimetallic particles.13 The IR experiments show a number of features which are assigned to the different sites described in the present work. The close similarity between the Eads values for CO on Rh or on RhCu heterometallic sites justifies the competition of CO for the different sites and their occupancy. VI. NO Adsorption on Monometallic Sites of the Rh, Cu, and RhCu Surface Models The interaction of NO with the monometallic sites of the different surface models has been studied using the strategy followed for CO on the same substrates. However, in the case of NO one needs to consider several adsorption sites: on top, bridge, fcc, and hcp three-hollow sites. To discuss the large amount of results in a systematic way, Table 5 reports results for the interaction of NO on the four sites mentioned above when the atoms interacting directly with NO are all Rh. In a similar way, Table 6 reports the corresponding results for the interaction of NO with pure Cu sites. The interaction with heterometallic sites is discussed in the next section. First, we comment on the results corresponding to the interaction of NO with the pure Rh(111) and Cu(111) surfaces since comparison to experiment is possible. The present DFT calculations suggest that the interaction of NO with Rh(111) is stronger on the three-hollow sites in agreement with experiment2,3,70,71 and previous theoretical values using a very similar approach.72 Moreover, the energy difference between both fcc and hcp sites is negligible. The structural parameters (dNO and dRh-N) as well as the NO stretching frequency are in agreement with experimental data obtained using various surface science techniques70,71 (see Table 5) thus validating the present theoretical approach. The geometric properties of NO adsorbed on the 3-fold hollows are very similar and close to those corresponding to the bridge site with differences of only 0.01 and 0.06 Å in the dNO and dRh-N, respectively. The stronger interaction in the 3-fold hollow sites leads to a larger molecular relaxation as expected from back-donation arguments. Similarly to adsorption on Rh sites, NO adsorption on pure Cu sites is preferred on hollow sites (Table 6) although the trend followed by the Eads on the different adsorption sites is similar to that corresponding to NO on Rh sites. The difference in adsorption energy between both metals, ∼ 30 kcal mol-1 indicates the preference of NO molecule to bind Rh atoms. This will be discussed later when considering heterometallic sites. Now we turn our attention to the adsorption properties of NO interacting with either pure Rh or pure Cu sites but on the bimetallic surface models. Inspection of Tables 5 and 6, shows that, at least for the composition ranges studied in the present work, the surface composition does not influence the adsorption properties of adsorbed NO. Thus, the adsorption properties are fully determined by the nature of atoms directly interacting with the NO molecule with very small variation induced by the surface composition; again within the considered surface range. This is in full agreement with the results in section IV corresponding to CO on the various monometallic sites of the bimetallic surfaces and constitutes one of the main conclusions of the present work. Finally, notice that, at variance of the behavior exhibited by adsorbed CO, the vibrational frequency of adsorbed NO is not

CO and NO Chemisorption on RhCu(111) Surfaces

J. Phys. Chem. B, Vol. 109, No. 10, 2005 4659

TABLE 5: Calculated Properties for NO Adsorbed on the Top, Bridge, fcc and hcp Sites of Rh, Cu/Rh, and Rh/Cu Surfacesa top site

Rh

% ∆12 % ∆23 % ∆Rh dNO (Å) dRh-N (Å) Eads (kcal mol-1) ν (cm-1)

-2.2 -0.1 +10.1 1.17 1.78 -46.5 1845 (184070)

bridge

Cu/Rh Rh/Cu Cu near Cu far -1.0 -0.5 +10.2 1.17 1.79 -47.6 1840

-2.3 -0.5 +6.4 1.17 1.78 -48.5 1862

-2.2 -0.5 +6.3 1.17 1.78 -47.1 1854

fcc Cu/Rh Cu near Cu far

Rh

Cu/Rh Cu far

Rh

-1.8 -0.1 +3.6 1.20 (1.1570) 1.96 -56.1 1620 (163071)

-2.1 -0.5 +2.7 1.20 1.96 -55.7 1622

-2.0 -0.6 +2.3 1.22 (1.1720) 2.04 (2.1770) -60.1 1546 (151570)

-2.1 -0.5 +2.6 1.22 2.04 -61.3 1517

-2.1 -0.5 +2.3 1.22 2.05 -59.7 1518

hcp Rh

Cu/Rh

-1.9 -0.4 +1.5 1.22 2.04 (2.1770) -61.3 1523 (151570)

-2.2 -0.4 +2.0 1.22 2.06 -60.7 1504

a The NO molecule always interacts directly with Rh atoms although those may have a different environment (see section 4). % ∆12, % ∆23 are the average relative relaxation of the outermost atomic layers and % ∆Rh is a measure of the surface rumpling. dNO and dRh-N, are the internuclear NO distance and the distance from the N atom to the surface nearest metal atom; ν is the NO stretching frequency and Eads is the interaction energy. Experimental values are shown in parentheses.

TABLE 6: Calculated Properties for NO Adsorbed on the Top, Bridge, fcc and hcp Sites of Cu, Cu/Rh, and Rh/Cu Surfacesa top

bridge Rh/Cu

site

Cu

% ∆12 % ∆23 % ∆Cu dNO (Å) dCu-N (Å) Eads (kcal mol-1) ν (cm-1)

-1.0 -0.3 +6.3 1.18 1.81 -18.9 1721 (>165076)

Cu/Rh Rh near Rh far -2.1 -0.4 +2.4 1.17 1.92 -17.1 1730

-0.8 -0.5 +6.1 1.18 1.82 -14.7 1742

-0.9 -0.5 +5.6 1.18 1.82 -16.9 1741

fcc Rh/Cu

Cu -1.0 -0.4 +3.3 1.21 1.94 -28.7 1538 (150076)

Rh near Rh far -0.8 -0.6 +2.6 1.21 2.01 -25.4 1540

-0.8 -0.6 +2.5 1.21 1.94 -25.5 1545

Rh/Cu Cu -1.1 -0.5 +2.5 1.22 2.00 -31.4 1472 (140076)

Rh near Rh far -0.8 -0.7 +2.8 1.22 2.01 -26.4 1479

-1.0 -0.6 +1.8 1.22 2.00 -28.1 1482

hcp Cu -1.2 -0.6 +1.8 1.22 2.01 -30.7 1475 (140076)

Rh/Cu Rh far -1.0 -0.7 +1.9 1.22 2.00 -26.7 1486

a The NO molecule always interacts directly with Cu atoms although those may have a different environment (see section 4). % ∆12, % ∆23 are the average relative relaxation of the outermost atomic layers and % ∆Cu is a measure of the surface rumpling. dNO and dCu-N, are the internuclear NO distance and the distance from the N atom to the surface nearest metal atom; ν is the NO stretching frequency and Eads is the interaction energy. Experimental values are shown in parentheses.

only in agreement with experimental available data but does also permit one to discern among the different active sites on the monometallic surfaces (as is well-known from experiment) and also to distinguish the atomic composition of the different sites in both monometallic and heterometallic surfaces. VII. NO Adsorption on Bimetallic Sites of the RhCu Surface Models The interaction of NO with various sites involving simultaneously Rh and Cu atoms has also been considered. The bridge site, composed by one Rh and one Cu already discussed in section V for CO adsorption, has also been also used to study NO adsorption. Since NO tends to prefer 3-fold hollow sites and since there is no appreciable difference between the fcc and hcp sites, the discussion will be limited to the latter. This hcp site is made either of one Cu and two Rh or one Rh and two Cu atoms depending on whether Cu/Rh or Rh/Cu compositions are considered. Nevertheless, one must be aware that geometry optimization can considerably distort the initial geometry. This is precisely the case for NO on the bimetallic hcp site of Cu/Rh. The NO interacts preferentially with the two Rh atoms and therefore the final geometry is very close to that corresponding to NO on the bridge site of the Rh(111) surface; see Figure 4 and Table 7. This similarity in the final adsorption site is also found on the structural parameters which again are close to those of NO on the bridge site of Rh(111). This includes the adsorption energy, which is similar to that corresponding to the interaction of NO on the bridge site of Rh(111), thus pointing out the local character of the interaction. An important conclusion from these results is that by monitoring the surface composition it would be possible to have all NO molecules on bridge sites instead of occupying the more stable 3-fold sites.

Figure 3. Adsorption geometry of CO on RhCu bridge site in Rh/Cu surface (Rh atoms are light gray, Cu atoms are dark gray, the C atom is black, and the O atom is white).

Figure 4. Adsorption geometry of NO on RhCuRh hcp hollow in Cu/ Rh surface (Rh atoms are light gray, Cu atoms are dark gray, the N atom is black, and the O atom is white). The two distances Rh-N and the angles O-N-Rh are equivalent.

This will be an ensemble effect since the presence of Cu in Cu/Rh simply kills many of the Rh 3-fold sites. The results for NO adsorption on the heterometallic bridge sites show that the Eads values are intermediate values between

4660 J. Phys. Chem. B, Vol. 109, No. 10, 2005

Gonza´lez et al.

TABLE 7: Calculated Properties for NO Adsorbed on Heterometallic Bridge and hcp Sites of Cu/Rh and Rh/Cu Surfacesa bridge

hcp

site

Cu/Rh

Rh/Cu

Cu/Rh

Rh/Cu

% ∆12 % ∆23 % ∆RhCu dNO (Å) dRh-N (Å) dCu-N (Å) q (deg) Eads (kcal mol-1) ν (cm-1)

-2.0 -0.4 +1.8 1.20 1.91 2.01 ∼0 -43.7 1601

-0.8 -0.7 +4.4 1.20 1.90 2.02 ∼0 -41.2 1616

-1.9 -0.5 +1.8 1.21 1.97 2.50 20 -54.9 1577

-0.9 -0.7 +2.9 1.21 1.94 2.06 ∼0 -40.5 1528

a

% ∆12, % ∆23 are the average relative relaxation of the outermost atomic layers and % ∆RhCu is a measure of the surface rumpling. dNO, dCu-N and dRh-N are the internuclear NO distance and the distance from the N atom to the surface nearest Rh or Cu metal atoms; ν is the NO stretching frequency and Eads is the interaction energy. θ is the tilting angle as defined in Figure 4.

those calculated for pure Cu- or Rh-bridge sites. Hence, for both heterometallic bridge sites, the calculated Eads is almost the same spite of the fact that one corresponds to Rh/Cu and the other to Cu/Rh. Once again, this is a clear expression of the local character of the interaction. However, the ν values are close to those of the pure Rh bridge. This seems to indicate that the interaction with the Rh atom is dominant. This is supported by the analysis of the metal to N distance which is always shorter for Rh. Notice, however, in all these cases the Eads value is significantly smaller than the one corresponding to the pure Rh sites. This implies that none of these sites will be occupied at low NO coverage, all molecules sticking on the surface will prefer to adsorb on monometallic Rh 3-fold sites. The situation is perhaps somewhat different for the interaction of NO at the heterometallic hcp site of the Cu/Rh surface which, as commented above, ends up being effectively an NO interaction with the pure Rh bridge site. VIII. Concluding Remarks In this work a systematic computational study concerning the interaction of CO and NO with various sites of RhCu(111) surfaces has been presented. The properties of the adsorbed molecules have also been compared to those corresponding to the pure Rh(111) and Cu(111) surfaces. One important finding of the present DFT calculations is that, for the composition range chosen in this work, changes in the various adsorption propertiess adsorbate geometry, distances to the nearest metal atoms, adsorbate stretching frequency, and adsorption energyswith respect to the composition of surface are rather small. This seems to suggest that interaction of the molecule with the surface is dominated by the atomic site composition with little influence of the remaining metallic environment. The analysis of the adsorption energies for the heterometallic bridge sites in Rh/Cu and Cu/Rh compositions shows that these results cannot be predicted from the interpolation arguments suggested by Jacobsen et al. for the Co-Mo bimetallic system.69 The interpolation arguments seem to hold for CO on Cu/Rh but cannot be sustained for CO on Rh/Cu because the predicted Eads values for these very distinct alloy compositions are extremely close to each other. This adds further support to the conclusion above suggesting that the interaction is dominantly local and driven essentially by the two atoms bind to CO and independent of the environment of these two atoms. Another important aspect is the close similarity between the Eads values

for CO on Rh or on RhCu heterometallic sites. This suggests a competition of CO for the different sites and their occupancy. From the previous discussion it follows that, in principle, CO will always try to bind monometallic on-top or bridge Rh sites but their number is very different in Rh/Cu and Cu/Rh alloys. In the latter, they are very numerous, but in the former, they are very scarce. Therefore, it would be possible to monitor the number of Rh-Cu sites in a RhCu alloy and thus control the concentration of adsorbed CO. A similar conclusion holds for the interaction of NO on these bimetallic surfaces. However, there is an important difference because in the monometallic surfaces: NO prefers the 3-fold hollow sites whereas on the heterometallic sites the final adsorption site is reminiscent of the bridge site. Therefore, using bimetallic surfaces of the appropriate composition would permit to have NO on bridge sites. This will be a ligand effect, since the presence of Cu in Cu/Rh simply kills many of the Rh 3-fold sites. On these CuRh bridge sites adsorbed NO perhaps exhibits a different reactivity. Of course, this would require working with Rh-rich alloys. Finally, one may attempt to relate the present results to the catalytic activity of the alloy as compared to the single metal surfaces. However, this is a difficult task because the catalytic performance of a given RhCu alloy will depend on the particular chemical reaction of interest, although it is also often found there is a close correlation between adsorption and activation energies for surface reactions.73 The adsorption of CO and NO may be viewed as the first step for some chemical reactions involving these molecules such as the NO reduction and CO oxidation. The present work suggests that it is possible to prepare the catalyst in such a way that the surface concentration of CO or NO can be monitored. This is of particular importance for the CO + O reaction where it is know that reaction rate is usually determined by CO desorption. However, the very little influence of the adsorption site surroundings on the adsorption energies seem to indicate that in the present case it is not possible to extract information regarding the catalytic activity of the different RhCu alloys from adsorption data alone. To address the problem of the different chemical reactivity of the RhCu alloys one would need to investigate the energy barriers for a given chemical process and to determine the role of the surface composition in the reaction energy profile. This work is currently being carried out in our laboratory for the CO + NO reaction on these bimetallic surfaces.74 To summarize, DFT calculation carried out on a series of RhCu alloys show that the interaction of small probe molecules such as CO or NO is essentially dominated by the atomic nature of the surface active site with little influence of the rest of the metallic systems. Moreover, it is suggested that it is possible to control the adsorption site of these molecules by appropriate choice of the surface composition. Acknowledgment. S.G. thanks the CONACyT (Me´xico) and the UniVersitat de Barcelona for supporting her predoctoral research. Financial support has been provided by the Spanish Ministry of Science and Technology (Project BQU2002-04029CO2-01) and, in part, by Generalitat de Catalunya (Projects 2001SGR0043 and Distincio´ per a la Promocio´ de la Recerca UniVersita` ria de la Generalitat de Catalunya granted to F.I.). The authors wish to express their gratitude to CESCA, CEPBA, and CIRI, supercomputer centers for generous computer time allocation and assistance. References and Notes (1) Ertl, G.; Kno¨zinger, H.; Weitkamp, J. Handbook of Heterogeneous Catalysis; Wiley-VCH: Munich, Germany, 1997; Vols. 3-5.

CO and NO Chemisorption on RhCu(111) Surfaces (2) Garin, F. Appl. Catal., A 2001, 222, 183. (3) Comelli, G.; Dhanak, V. R.; Kiskinova, M.; Price, K. C.; Rosei, R. Surf. Sci. Rep. 1998, 32, 165. (4) Rodriguez, J. A. Surf. Sci. Rep. 1996, 24, 223. (5) Petrov, L.; Soria, J.; Dimitrov, L.; Catalun˜a, R.; Spasov, L.; Dimitrov, P. Appl. Catal., B 1996, 8, 9. (6) Granger, P.; Lecomte, J. J.; Leclercq, L.; Leclercq, G. Appl. Catal., A 2001, 208, 369. (7) Ponec, V. Appl. Catal., A 2001, 222, 31. (8) Loffreda, D.; Delbecq, F.; Simon, D.; Sautet, P. J. Chem. Phys. 2001, 115, 8101. (9) Deubage, Y.; Abon, M.; Bertolini, J. C.; Massardier, J.; Rochefort, A. Appl. Surf. Sci. 1995, 90, 15. (10) Chou, S. C.; Yeh, C. T.; Chang, T. H. J. Phys. Chem. B 1997, 101, 5828. (11) Anderson, J. A.; Rochester, C. H.; Wang, Z. J. J. Mol. Catal. A 1999, 139, 285. (12) Coq, B.; Dutartre, R.; Figueras, F.; Rouco, A. J. Phys. Chem. 1989, 93, 4904. (13) Guerrero-Ruiz, A.; Bachiller-Baeza, B.; Ferreira-Aparicio, P.; Rodrı´guez-Ramos, I. J. Catal. 1997, 171, 374. (14) Ferna´ndez-Garcı´a, M.; Martı´nez-Arias, A.; Rodrı´guez-Ramos, I.; Ferreira-Aparicio, P.; Guerrero-Ruiz, A. Langmuir 1999, 15, 5295. (15) Reyes, P.; Pecchi, G.; Fierro, J. L. G. Langmuir 2001, 17, 522. (16) Gonza´lez, S.; Sousa, C.; Ferna´ndez-Garcı´a, M.; Bertin, V.; Illas, F. J. Phys. Chem. B 2002, 106, 7839. (17) Rodrı´guez, J. A.; Campbell, R. A.; Goodman, D. W. J. Phys. Chem. 1990, 94, 6936. (18) Gonza´lez, S.; Sousa, C.; Illas, F. Surf. Sci. 2003, 531, 39. (19) Morrison, S. R. The chemical physics of surfaces; Plenum Press: New York, 1977. (20) Somorjai, G. A. Introduction to Surface Chemistry and Catalysis, John Wiley and Sons: New York, 1994. (21) Kiselev, V. F.; Krylov, O. V. Adsorption and catalysis on transition metals and their oxides; Gomer, R., Ed.; Springer Series in Surface Science 9; Springer-Verlag: Berlin, 1988. (22) Hopstaken, M. J. P.; Niemantsverdriet, J. W. J. Chem. Phys. 2000, 113, 5457. (23) Campbell, C. T.; Shi, S. K.; White, J. M. J. Phys. Chem. 1979, 83, 2255. (24) Colonell, J. I.; Gibson, K. D.; Sibener, S. J. J. Chem. Phys. 1995, 103, 6677. (25) Peden, C. H. F.; Goodman, D. W.; Blair, D. S.; Berlowitz, P. J.; Fisher, G. B.; Oh, S. H. J. Phys. Chem. 1988, 92, 1563. (26) Smedh, M.; Beutler, A.; Borg, M.; Nyholm, R.; Andersen, J. N. Surf. Sci. 2001, 491, 115. (27) Linke, R.; Curulla, D.; Hopstaken, M. J. P.; Niemantsverdriet, J. W. J. Chem. Phys. 2001, 115, 8209. (28) Gierer, M.; Barbieri, A.; Van Hove, M. A.; Somorjai, G. A. Surf. Sci. 1997, 391, 176. (29) Eichler, A.; Hafner, J. J. Chem. Phys. 1998, 109, 5585. (30) Zhang, C. J.; Hu, P.; Lee, M. H. Surf. Sci. 1999, 432, 305. (31) Eichler, A. Surf. Sci. 2002, 498, 314. (32) Loffreda, D.; Simon, D.; Sautet, P. J. Chem. Phys. 1998, 108, 6447. (33) Birgersson, M.; Almbladh, C. O.; Borg, M.; Andersen, J. N. Phys. ReV. B 2003, 67, 045402. (34) Milushev, A.; Hadjiivanov, K. Phys. Chem. Chem. Phys. 2001, 3, 5337. (35) Borovkov, V. Y.; Jiang, M.; Fu, Y. J. Phys. Chem. B 1999, 103, 5010. (36) Greeley, J.; Mavrikakis, M. J. Catal. 2002, 208, 291. (37) Shelef, M. Chem. ReV. 1995, 95, 209. (38) Rodriguez, J. A.; Goodman, D. W. Science 1992, 257, 897.

J. Phys. Chem. B, Vol. 109, No. 10, 2005 4661 (39) Illas, F.; Lopez, N.; Ricart, J. M.; Clotet, A.; Conesa, J. C.; Ferna´ndez-Garcı´a, M. J. Phys. Chem. B 1998, 102, 8017. (40) Bagus, P. S.; Hermann, K.; Bauschlicher, C. W., Jr. J. Chem. Phys. 1984, 81, 1966. (41) Bagus, P. S.; Hermann, K. Phys. ReV. B 1986, 33, 2987. (42) Bagus, P. S.; Illas, F. J. Chem. Phys. 1992, 96, 8962. (43) Szanyi, J.; Goodman, D. W. J. Catal. 1994, 145, 508. (44) Kresse, G.; Furthmu¨ller, J. Phys. ReV. B 1996, 54, 11169. (45) Kresse, G.; Hafner, J. Phys. ReV. B 1993, 47, 558. (46) Kokalj, A.; Causa, M. X-window CRYstalline Structures and DENsities package. 1999. (47) Kokalj, A. J. Mol. Graphics Modell. 1999, 17, 176. (48) Perdew, J. P.; Wang, Y. Phys. ReV. B 1992, 45, 13244. (49) Blo¨chl, P. E. Phys. ReV. B 1994, 50, 17953. (50) Blo¨chl, P. E.; Margl, P.; Schwarz, K. Ab initio molecular dynamics with the projector augmented wave method. In Chemical application of Density-Functional Theory; American Chemical Society: Washington, DC,, 1996. (51) Kresse, G.; Joubert, D. Phys. ReV. B 1998, 59, 1758. (52) See for instance: http://www.webelements.com. (53) Lopez, N.; Norskov, J. K. Surf. Sci. 2001, 477, 59. (54) Loffreda, D.; Delbecq, F.; Simon, D.; Sautet, P. J. Phys. Chem. B 2001, 105, 3027. (55) Massalsiki, T. B.; Okamoto, H.; Subramanian, P. R.; Kacprezak, L. Binary Alloy Phase Diagrams, 2nd ed.; ASM: Materials Parks: OH, 1992. (56) Ruban, A. V.; Skriver, H. L.; Norskov, J. K. Phys. ReV. B 1999, 59, 15990. (57) Hammer, B.; Hansen, L. B.; Norskov, J. K. Phys. ReV. B 1999, 59, 7413. (58) Feibelman, P. J.; Hammer, B.; Norskov, J. K.; Warner, F.; Scheffler, M.; Stumpf, R.; Waywe, R.; Dumesic, J. J. Phys. Chem. B 2001, 105, 4018. (59) Rider, K. B.; Hwang, K. S.; Salmeron, M.; Somorjai, G. A. J. Am. Chem. Soc. 2002, 124, 5588. (60) Andersson, S.; Pendry, J. B. J. Phys. C 1980, 13, 3547. (61) Wei, D. H.; Skelton, D. C.; Kevan, S. D. Surf. Sci. 1997, 381, 49. (62) Kirstein, W.; Kru¨ger, B.; Thieme, F. Surf. Sci. 1986, 176, 505. (63) de Jong, A. M.; Niemantsverdriet, J. W. J. Chem. Phys. 1994, 101, 10126. (64) Eve, J. K.; McCash, E. M. Chem. Phys. Lett. 1999, 313, 575. (65) Hadjiivanov, K.; Venkov, T.; Kno¨zinger, H. Catal. Lett. 2001, 75, 55. (66) Coloma, F.; Bachiller-Baeza, B.; Rochester, C. H.; Anderson, J. A. Phys. Chem. Chem. Phys. 2001, 3, 4817. (67) Suzuki, A.; Inada, Y.; Yamaguchi, A.; Chilhara, T.; Yuasa, M.; Nomura, M.; Iwasawa, Y. Angew. Chem., Int. 2003, 42, 4795. (68) Krishnamurthy, R.; Chiang, S. S. C. Thermochim. Acta 1995, 262, 215. (69) Jacobsen, C. J. H.; Dahl, S.; Clausen, B. S.; Bahn, S.; Logadottir, A.; Nørskov, J. K. J. Am. Chem. Soc. 2001, 123, 8404. (70) Kim, Y. J.; Thevuthasan, S.; Herman, G. S.; Peden, C. H. F.; Chambers, S. A.; Belton, D. N.; Permana, H. Surf. Sci. 1996, 359, 269. (71) Root, T. W.; Fisher, G. B.; Schmidt, L. D. J. Chem. Phys. 1986, 85, 4679. (72) Loffreda, D.; Simon, D.; Sautet, P. Chem. Phys. Lett. 1998, 291, 15. (73) Mavrikakis, M.; Stoltze, P.; Norskov, J. K. Catal. Lett. 2000, 64, 101. (74) Gonza´lez, S.; Sousa, C.; Illas, F. Work in progress. (75) Raval, R.; Parker, S.; Penble, M.; Hollins, P.; Pritchard, J.; Chesters, M. Surf. Sci. 1988, 203, 353. (76) Dumas, P.; Suhren, M.; Chabal, Y. J.; Hirschmugl, C. J.; Williams, G. P. Surf. Sci. 1997, 371, 200.