Molecular Heterogeneous Catalysis, Wiley (2006), 352729662X

136 Chapter 3

The coadsorption of atomic oxygen enhances the adsorption energy of ammonia by 18 kJ/mol. This enhancement of the adsorption energy occurs as long as adsorbed NH3 and atomic oxygen do not share a bond with the same metal atom. According to the Bond Order conservation principle, the weakened metal–metal bonds next to adsorbed oxygen enhance the reactivity of the surface metal atom to which NH3 adsorbs. The reaction of ammonia with coadsorbed oxygen reduces the activation barrier by over 68 kJ/mol. In the transition state, the oxygen atom moves to a two-fold position and the hydrogen–oxygen bond has already been partially formed.

The barrier heights for the subsequent elementary reaction steps with adsorbed O increase. This is due to the fact that Oad and NH2ad have to share metal surface atoms in the transition state. This results in strong repulsive interactions. The overall result is that the rate of the initial dissociative NH3 adsorption with coadsorbed O is increased, but the reactivity of adsorbed NH2 and NH are reduced.

On Pt, OH adsorbs on an atop site. Therefore, its interactions with NH3, NH2 and NH with OH in the transition state do not share binding to the same metal atom. Reactions with adsorbed OH to form H2O lower the activation energies for all three of these cases.

The e ect of surface steps on the activation of the NH bond over Pt was found to be negligible. Nevertheless, the experimental rate of ammonia activation will be higher near the step than on the terrace. This is due to an increase in the surface concentration of ammonia at the step over the concentration on the terrace rather than the intrinsic activation barriers. The higher coverage of ammonia at the step is the result of a high adsorption energy of ammonia at the step. The heat of adsorption of NH3 is 20 kJ/mol higher on a terrace than on the step. Hence the apparent activation energy for the dissociative adsorption, at low coverage, which is equal to ET ST + Eads, is decreased by the same amount. The overall e ect will be an increased rate of dissociative adsorption.

The apparent activation energy for the dissociative adsorption of NH3 with preadsorbed O is significantly lower. It then follows that the e ect of coadsorbed oxygen is much larger than that of the activation by steps.

The low-temperature oxidation reaction of NH3 is extensively discussed in Chapter 6. The experimental evidence indicates that ammonia will only dissociatively adsorb when coadsorbed oxygen is present on the Pt surface. N2 is the initial product that forms in the presence of oxygen in a flow system operating under mild conditions. N2O is formed as a co-product as the catalyst begins to deactivate.

The recombination of nitrogen adatoms occurs at at step edges, with a low barrier (Eact = 70 kJ/mol) and, hence, at low temperature. The formation of N2 at low temperature, however, can also occur via the reaction of NO with NH3. This reaction requires the formation of surface NH2 to proceed (see Chapter 6). Because of its high adsorption energy (Eads = 187 kJ/mol), NO once formed will not desorb, especially at low temperatures. NO, however, can be indirectly detected if it reacts to form N2O. N2O readily desorbs and is a co-product that indeed is observed.

The activation energies and reaction paths for the recombination of N and O to give NO on a terrace and a stepped Pt(111) surface are shown in Fig. 3.49. The activation energy for N + O recombination (Eact = 223 kJ/mol)is slightly lower than that for the N + N recombination reaction (Eact = 234 kJ/mol). The decrease of the activation energy for the formation of NO on the terrace step (Eact = 65 kJ/mol is so large that it occurs readily at low temperatures. However, the desorption energy of NO from a terrace is 187 kJ/mol and from the hollow position on the step is 133 kJ/mol, hence NO will not desorb at low temperatures. The low reactivity of the Pt(111) surface, without steps, is clearly

The Reactivity of Transition-Metal Surfaces 137

Figure 3.49. The recombination reaction path for nitrogen and oxygen on a stepped Pt(111) surface. Reactions at terraces and steps are compared.

Figure 3.50. Coadsorbed oxygen atoms are not always promoters[45].

due to the high barriers for the recombination reaction of Nads + Oads. The presence of steps, however, provide low activation barrier paths for these recombination reactions. As discussed in the previous subsection, NO has a unique reactivity on the Pt(100) surface. Qe and Neurock[24a] showed that NO dissociation is strongly enhanced, but the N and O adsorption energies are less increased. A low barrier ( 7 kJ/mol) for Nads and Oads recombination is found owing to the unique transition-state configuration over the fourfold hollow site made up of squarely arranged Pt atoms. NO recombination therefore also proceeds over the hollow site of squarely arranged Pt atoms. The N and O atoms do not share binding to the same metal atom in the transition state.

The reaction of NH3 with adsorbed O to form hydroxyl intermediates on Pt decreases the reaction energy for the dissociative adsorption of NH3. In contrast, such a reaction on Rh acts to increase the reaction energy. This relates to the much stronger Rh–Oads bond (–110 kcal/mol) as compared with the Pt–Oads (–66 kcal/mol) bond. For dissociative adsorption of methanol, this is illustrated in Fig. 3.50. The stronger oxygen adatom bond to Rh than to Pt may have a beneficial selectivity e ect in reactions where hydrogen

138 Chapter 3

formation competes with water formation. Hickman and co-workers[39] studied methane oxidation on Rh and Pt monoliths. The main products from this reaction were CO, CO2, H2 and H2O. On Rh they found for reaction conditions which enhanced the selectivity to CO production over that of CO2 a substantially higher selectivity to H2 compared with H2O. The activation energy for H2O formation from 2Hads and Oads over Rh equals 120 kJ/mol. On Pt the corresponding value is only 60 kJ/mol. The bond energy of hydrogen with Rh or Pt is quite similar and, hence, the activation energy for H2 recombination also will be similar. The improved H2 selectivity on Rh is due to the suppressed rate of H2O formation on this surface caused by the stronger metal–oxygen bond energy.

A final remark has to be made on the nature of the transition states. The transition states for the first bond cleavage reaction of CH4, NH3 and H2O to produce Hads and adsorbed CH3, NH2 or OH, respectively, have slightly higher entropies than the bond cleavage of the second or successive X–H bonds. In the transition state for the NH cleavage of NH3, the entropy remains relatively high because the HNH part of the molecule remains nearly freely rotating. The lower transition-state entropies for NH2 and NH cleav-

age support the view that the bond cleavage reactions proceed through tight transition states[40] . The higher transition-state entropy for the cleavage of the first X–H bond im-

plies a slightly looser transition state and, hence, a smaller dependence on the degree of coordinative unsaturation of the surface atoms. On the other hand, surface reactions tend to proceed through activation barriers of low activation entropy and, hence, bond cleavage reactions show a strong dependence on overall reaction enthalpy. This will be

reflected in BEP transition state with α values close to 1. For reactions involving CHx and OHads surface species this has been confirmed by Michaelides et al.[41].

3.9 Carbon–Carbon Bond Cleavage and Formation Reactions, a Comparison with CO Oxidation

Watwe et al.[42] describe the C–C bond cleavage over Pt(111) and Pt(211) surfaces. Despite the di erences between the two surfaces, the structures for the adsorbates on the two surfaces are similar. A common feature of the transition states is the significant extension of the C–C bond compared with that in the ground state. In the transition state, the C–C bond was found to be at least 25% longer. The transition states are late. As follows from Table 3.7, the barrier for breaking the C–C bond increases with increasing bond order, and is lower for those bonds which generate fragments that are more stabilized by the metal surface.

There is a substantial lowering of the barrier when the C–C bond is activated over a step site. In the gas phase, the activation of σ-C–C bonds is thermodynamically preferred over that of alkane C–H bonds. On the surface, however, C–H activation is found to be easier than C–C bond activation. This relates to some extent to the di erences between the M–CH3 and M–H bond energies. The C–H bond is more readily activated on the surface since there is initially less steric hinderance than that for HxC–CHy activation. In the adsorbed state, the hydrogen atoms of an alkyl intermediate directly interact with the surface through weak van der Waals interactions. The carbon atoms, on the other hand, are not in direct contact.

Carbon–carbon bond formation reactions are of critical importance to the Fischer– Tropsch synthesis of linear hydrocarbons over transition metals such as Co, Fe or Ru. Fischer–Tropsch synthesis involves the activation of CO and H2 over the metal to form adsorbed carbon, oxygen and hydrogen. The carbon atoms that form hydrogenate to form

The Reactivity of Transition-Metal Surfaces 139

di erent CHx intermediates that can subsequently couple to form longer hydrocarbon chain intermediates and products. We present here theoretical results obtained out over well-defined transition-metal surfaces for some of the critical steps in the mechanism. Zheng et al.[43] demonstrated that on terraces of the transition metals the CHx–CHy recombination reaction requires low values of x and y otherwise repulsive interactions between the hydrogen atoms on the CH2 and CH3 groups will prevent recombination. This conclusion is in line with the high barrier for the ethyl C–C cleavage reaction found on the Pt(111) surface reported in Table 3.7. The barrier is due to the large repulsive interaction that the hydrogen atoms experience with the surface.

Table 3.7. Transition-state energies for C–C bond cleavage on two surfaces of Pt[42]

Figure 3.51. Schematic potential energy surfaces (relative energies) on Co and Ru for C–C coupling of CH with CH2 [45].

A similar argument holds for the recombination of CH3 and CO on the terrace of a transition metal. The strong repulsive interactions between the CH and the metal surface in the transition state with the CO surface species has to be overcome. Reaction paths that cross over a step edge or occur directly at a step edge are expected to experience far less repulsion interactions between these adsorbates. Calculated barriers for CHx–CHy recombination on stepped and unstepped Ru surfaces are given in Table 3.8. Ethylene formation is a preferred reaction on the steps. Also C + CH and C + CH2 recombination are preferred recombination steps.

The low barrier for the recombination of CH and CH3 surface species on terrace sites is quite striking. Work by Ge et al.[45] allows a comparison of the CH + CH2 recombination

140 Chapter 3

reaction on the less reactive Co(0001) compared with that on the Ru(0001) surface. The recombination reaction occurs with a substantially lower barrier on Co than Ru (see Fig. 3.51). This is due to the weaker M–CHx bonds on Co. The proposition that the C1 adsorbed intermediates, namely CHads, are the most abundant reaction intermediates (MARI) in Fischer–Tropsch synthesis is consistent with the fact that the the CHads species is calculated to be more stable on the surface than the Cads species. Chain growth would then require the following reaction steps:

The calculated transition state of 77 kJ/mol for step (2) implies that in this cycle of steps, that hydrogen addition may be the rate-limiting step. While the results from theoretical predictions suggest that we can speculate as to the MARI and the rate-limiting step, we can not be sure unless we carry out full simulations of the elementary processes occurring simultaneously to establish the actual kinetic outcome.

Table 3.8. Calculated reaction barriers (Ea) for C–C coupling and some hydrogenation addition reactions on the Ru surface (eV)[44]

As long as there are enough CH species available on the surface, the chain growth reaction may be faster than the chain termination reaction. On steps and surfaces where CH2 species are stabilized [e.g. the Ru(1121) surface] ethylene formation will compete strongly with the chain growth reaction. Alkene formation can also proceed by the β-CH cleavage of adsorbed alkyl species. For Pd the activation barrier for this reaction is 70 kJ/mol, which competes with the activation energy for hydrogenation of ethyl. On Ru the activation barrier of the β-CH reaction is even lower. Therefore, alkene formation is preferred over alkane formation in the chain termination step.

142 Chapter 3

Figure 3.53. Fraction of adsorbed C1 species formed via CH4 decomposition at 400 ◦C[48] , cooled to room temperature and subsequently converted to ethane and propane upon hydrogenation at room temperature[47].

The third alternative termination path is by direct insertion of CO into the metal– alkyl bond. This insertion reaction requires the presence of coadsorbed CO and occurs on surfaces where CO bond cleavage competes with C–C bond formation as shown in Section 3.7. CO insertion or hydroformulation requires cationic metal centers and is therefore an unfavorable reaction at a reduced metal surface.

There is ample experimental evidence[26] that suggests that the Fischer–Tropsch reaction proceeds via the formation of C1 species, as proposed above. Experimental evidence indicates that the C1 species are favored as surface CHx intermediates[46].

In order to dissociate CO, the barrier for CO activation has to be low. As illustrated in Table 3.4, the rate of dissociative CO adsorption appears to be fastest over Ru, Co and Fe, the preferred transition metals for the Fischer–Tropsch reaction. The barrier for CO dissociation, however, is very high compared with that of the chain growth or termination reactions, especially on the close-packed terraces. While the presence of steps helps to lower the CO activation barrier, it is still considered to be rate limiting.

The requirement for a low activation energy for CO dissociation is opposite to that for e cient carbon–carbon bond formation, which favors weak metal–carbon bonds. This is nicely illustrated in Fig. 3.52a and b where we compare the C–C bond formation and C–O bond formation energies on Ru(111) and Ni(111) surfaces, respectively. Table 3.4 summarizes the transition-state energies for CO dissociation, and indicates that Ni is less reactive than Rh. The M–C bonds are weaker on Ni than on Ru as reflected by the higher endothermicity for carbon deposition on Ni. Whereas C–C bond formation on Ru has a substantial barrier, such a barrier is absent on Ni. In the presence of hydrogen, the C–C bond formation reaction, which is the precursor to graphite formation, is suppressed. Once CH species are present on the surface, the graphitization reaction becomes much more di cult and typically does not occur. In the presence of hydrogen, the CHx-CHy bond formation reaction begins to compete with the methanation reaction. In the methanation reaction, the metal–carbon bonds become completely broken. The C–C bond formation reaction, however, only depends weakly on variation in the M–C bond energies. Hence methanation will be more strongly dependent on the M–C bond energy than the chain growth reaction is. This prediction is confirmed by experiments[47] in which CHads species were generated by dissociative adsorption of CH4 on the surface. In the presence of hydrogen, C–C coupling occurs and the alkanes desorb. The competitive reaction here is CH4 formation. The selectivity towards higher hydrocarbons is shown in Fig. 3.53. The stronger the metal–carbon bond, the larger is the selectivity towards higher hydrocarbons. There is no reaction on metals such as Cu, or Ag where H2 cannot dissociate. There is also no reaction when the M–C bond is too strong, as is the case for Fe.

The Reactivity of Transition-Metal Surfaces 143

3.10 Lateral Interactions

3.10.1 Introduction

Up to this point, we have focused on modeling the intrinsic reactivity of individual molecules on a surface which can be used for comparison with experiments carried out at low surface coverages under UHV conditions. Reactions that are carried out under more industrially relevant conditions, however, are typically run at pressures which are many orders of magnitude greater. This is known as the pressure-gap problem in surface science. These higher pressures can lead to significantly higher surface coverages, which can subsequently change the surface composition as well as the catalytic performance. At higher surface coverages, the intermolecular interactions between adsorbed intermediates become important in dictating the bond strength of the adsorbate to the surface and also

its reactivity. The interactions between adsorbed species are known as lateral interactions and can be defined in terms of through-space or through-surface interactions. The through-

space interactions are typically due to local steric or electrostatic interactions between two or more species. These interactions are based solely on the position of the adsorbate with respect to other molecules or intermediates on the surface. They exist even in the absence of the metal surface. Through-surface interactions are the result of electronic interactions between two or more adsorbates that are mediated through changes in the electronic structure of the surface metal atoms. We discuss the chemical bonding aspects of lateral interactions in more detail in Section 3.10.2.

The interactions between coadsorbed molecules or atoms can be either attractive or repulsive depending upon the local positioning of the adsorbates with respect to one another. If two or more adsorbates form bonds with the same metal atom, the interactions are typically repulsive, as would be predicted by Bond Order Conservation (BOC) principles. The interactions between adsorbates that are distant from one another by a single metal bond are typically attractive, which again would follow BOC principles. At higher surface coverages, the interactions between adsorbates are typically, but not always, repulsive. Repulsive interactions weaken the adsorption energy and thus lead to a decrease in the desorption energy with coverage. The metal-mediated electronic changes can lead to overlayer ordering and, in some cases, even drive surface reconstruction, as was discussed in Chapter 2. The substantial weakening of the adatom or adsorbed molecule interaction occurs when two or more adsorbates bind to the same surface metal atom. The reduction in the binding energy with coverage implies that such adsorption states are likely populated only at low temperature and high pressure and not likely accessible under UHV conditions.

The adsorbates which are more weakly bound to the surface are more likely to interact with other surface species through bond-making processes. An example of this situation will be discussed in Section 3.10.3 where we examine the ethylene hydrogenation mechanism as a function of surface coverage. We specifically analyze the elementary reaction steps for both π- and σ-bonded ethylene intermediates.

Lateral interactions influence the reactants, products, intermediates and even transition states for a reaction. Reactant molecules likely adsorb in di erent local environments and are therefore exposed to di erent lateral interactions depending upon the relative number, type and position of neighboring adsorbates. Stochastic kinetic methods provide the best hope of capturing these molecular di erences. Traditional deterministic modeling of catalytic systems average over the surface coverage and thus provide only a mean field description. Individual surface sites, as well as intermolecular interactions, however, can be

The Reactivity of Transition-Metal Surfaces 145

that might arise from first-principle calculations. Currently the only practical way to include lateral interactions would be to develop a simpler coarse-grained model that can be used to calculate the interactions as they arise within the simulation.

Various models have been proposed in the literature to model adsorbate–adsorbate interaction[49]. At the simplest level, the interactions can be described by a single parameter ωAA which treats the repulsion (ω > 0) and attractive (ω < 0) interactions between two species labeled A:

This approach has been adapted for systems whereby the lateral interactions are lumped into a single parameter. More advanced treatments add a second parameter (ωAA ) in order to describe next-nearest neighbor interactions. Q0 here refers to the binding energy of the isolated A molecule. The simplest molecular level treatment would view these as pairwise additive interactions. The change in the binding energies and activation barriers can easily be computed by simply adding (or subtracting) the e ect of all pairwise interactions that result from direct nearest (ω) and next-nearest neighbors (ω ), This is shown for the adsorption of A in the following expression:

A much more sophisticated treatment for modeling all interaction pairs as well as trimer interactions was developed by Kreuzer and co-workers[50] . They subsequently extended the approach to examine oxygen on Ru using first principle DFT calculations[51]. They calculated a large number of di erent possible configurations for oxygen on Ru, and then regressed the coe cients for both pair and ternary interactions in the model to these constants. This is an elegant study whereby a theory was used to establish a coarsegrained interaction model. The results follow the experimental TPD curves for oxygen on Ru(0001) very well. The number ab initio calculations necessary to establish a ternary model for catalytic systems with multiple di erent adsorbates would be prohibitive.

A second approach which may be attractive for more complex surface systems involves the application of the Bond Order Conservation model that was developed by Shustorovich and co-workers[52−65]. The BOC model treats the interaction between the adsorbate and the surface atom through the use of a Morse potential. The total heat of adsorption is then described by summing all interactions. The BOC model is based on the concept that the bonding potential for every atom in the system is conserved. The heat of adsorption for an atomic species A is described by the following expression:

where A and n refer to the adsorbate A and the number of metal atoms, respectively. Shustorovich and Sellers [61,65] developed a systematic set of equations which can be

used to estimate activation barriers for adsorption, surface reaction and desorption processes based upon the strength of atomic interactions Although this approach may not yield quantitative predictions for all systems, it has been very e ective in estimating the

barriers and adsorption energies for a number of systems.

Hansen and Neurock[66] showed how such an approach can be coupled with firstprinciple DFT calculations in order to model quantitatively the interactions of O/Ru(100).