Molecular Heterogeneous Catalysis, Wiley (2006), 352729662X

126 Chapter 3

3.6.3 Association Reactions; Carbon–Carbon Bond Formation

The activation of CO is one of the critical elementary steps that controls Fischer–Tropsch synthesis in the production of higher hydrocarbons from synthesis gas. It is well established that the Fischer–Tropsch reaction proceeds by activating CO to form surface carbon and oxygen[26]. The surface carbon subsequently hydrogenates to form various CHx intermediates which can react further with hydrogen, couple with other CHx fragments and ultimately desorb as di erent hydrocarbon products.

In addition to the CO dissociation paths discussed in the previous subsection, the presence of coadsorbed hydrogen o ers a third potential reaction path for the activation of the CO bond. CO bond activation, which proceeds through interaction with the metal surface, becomes more di cult for Group VIII metals at the bottom-right corner of the periodic table. In contrast, the weaker M–CO bonds tend to help promote the hydrogenation of CO to form formaldehyde or methanol. This reaction proceeds through the formation of surface formyl (CHO) intermediates.

The calculated barrier for the reaction of CO and hydrogen to form the surface formyl intermediate over Ru is 143 kJ/mol (see Fig. 3.39). The subsequent CO bond activation of the the formyl intermediate on Ru is then only 30 kJ/mol. Hence, on the Ru terrace this scheme of first adding hydrogen before CO dissociation will be the preferred path to cleave the CO bond compared with the direct CO dissociation path. On the step, the barrier to formyl formation is similar to that on the terrace, hence higher than the barrier to direct CO dissociation. On surface steps the latter dissociation step path will be the preferred path for the formation of C1 adsorbed species.

Figure 3.39. The reaction energy diagram for the C–O bond-cleavage reaction via intermediate formyl formation[23].

The trends for the insertion reaction are opposite of those for dissociation since association is the microscopic reverse of dissociation. The weaker the adsorption of CO and H, the easier their recombination should be. The calculated activation barrier for the association of surface CO and hydrogen to form the formyl intermediate on Pd(111) is 70 kJ/mol[27], which is much lower than the earlier value reported over Ru. This is directly in line with expectation, since CO and H bind more strongly to Ru than they do to Pd. The formation of the adsorbed formyl intermediate is lowered now to 40 kJ/mol endothermic. In the

The Reactivity of Transition-Metal Surfaces 127

presence of hydrogen, the path to consecutive hydrogenation to methanol now competes with the CO cleavage reaction. Because of the weak metal adsorbate bonds the reactions steps towards methanol formation are also exothermic. In contrast, the CO dissociation reaction over Pd(111) has a fairly high activation barrier and is highly endothermic[27] .

The comparison of the CO activation and CO hydrogenation barriers over Pd strongly indicates that CO hydrogenation to methanol is much more prevalent than CO activation subsequently to form CH4. This discussion also illustrates why on a transition-metal surface (in the absence of any steric constraints) there can be a preference for a particular reaction channel. On Pd(111) the overall reaction barrier for methanol formation is significantly lower than that for the CO dissociation reaction.

3.7 Organometallic Chemistry of the Hydroformulation Reaction

Homogeneous catalytic reactions can typically provide considerable insight into the mechanisms that govern analogous heterogeneous catalyzed pathways[28]. Organometallic complexes are structurally well defined and can be characterized by a number of in-situ spectroscopic techniques. In the previous section, we discussed possible reaction paths and mechanisms for the hydrogenation of CO to methanol. This reaction is analogous to the homogeneous hydroformulation reaction. Hydroformulation involves the reaction of an alkene with hydrogen and CO over a homogeneous organometallic catalyst and is used in the production of aldehydes and alcohols. This reaction is thought to proceed by the hydrogenation of ethylene to form an ethyl fragment followed by the insertion of the ethyl intermediate into a metal–CO bond. We will discuss here the elementary reaction steps involved in the hydroformulation reaction as proposed for the cobalt carbonyl complex. Cationic Pd or Rh complexes with phosphine ligands are currently also widely used for this reaction. The structure of the HCo(CO)4 complex that reacts with ethylene is given in Fig. 3.40. In order to react with ethylene, one of the CO ligands has to desorb so that alkyl formation can occur from the reaction of π-bonded ethylene with the H atom bound to CO. In complex II, which is shown in Fig. 3.41, Co has a formal charge of +1 and a d-electron occupation of d8. The insertion reaction can be considered to occur on the (OC)3Co fragment we discussed before (Section 3.4), with three empty ligand positions. The three dangling bonds can form one symmetric and two antisymmetric combinations. In complex II, one of the antisymmetric dxz symmetric orbitals is unoccupied. Complex II can subsequently react to form complex III.

Figure 3.40. The structure of HCo(CO)4 (complex I).

In the first step, a CO molecule inserts into the M–C bond of adsorbed ethyl. Subsequently, H2 adsorbs on the vacant ligand position. In the final step, H2 dissociates heterolytically and subsequently reacts with the CH3CH2CO fragment to form the corresponding aldehyde along with regeneration of the catalytically active carbonyl complex.

128 Chapter 3

Figure 3.41. Formation of a Co alkyl–carbonyl complex.

Figure 3.42. Insertion of CO and aldehyde formation.

The barrier necessary to activate CO for insertion stems from the Pauli repulsion between the doubly occupied σ-ethyl orbital and the doubly occupied 5σ-CO orbital of the inserting CO molecule[29]. When the ethyl carbon atom and the CO carbon atom approach each other, a bonding and antibonding orbital combination comprised of the two σ-type orbitals is formed. Both are doubly occupied, thus resulting in Pauli repulsion. This Pauli repulsion is reduced by the donative interaction of this doubly occupied antibonding σ orbital combination with the antisymmetric unoccupied Co dxz atomic orbital. This is a very general process and has been investigated for several di erent metal systems[29] . This is schematically illustrated in Fig. 3.43 for the insertion of CO into the metal–methyl bond of a Pd2+ phosphine complex. The complex in Fig. 3.43 is planar. Donation of electrons from the antibonding 5σ–CH3 orbital into the empty low energy dx2−y2 metal orbital lowers the activation barrier for the insertion reaction.

3.8 Activation of CH4 , NH3 and H2O

The dissociation of CO, NO, and other diatomic molecules predominantly occurs through the back-donation of electrons from the metal surface into the low-energy π antibonding orbitals on the adsorbate, thus leading to a significant weakening of the adsorbate– adsorbate bond. In the transition state the intramolecular bond is typically nearly broken. The dissociating complex essentially exists as two fragments coordinated to the surface that tend to share bonding with a common metal atom. This metal atom sharing typically leads to repulsive interactions between the product fragments. The activation barrier with respect to the dissociated fragments is primarily determined by the degree of weakening of the bonds between the dissociating fragments and the metal surface. The barrier height for recombination is only weakly dependent upon the variation of the adatom metal surface interaction energy.

The Reactivity of Transition-Metal Surfaces 129

Figure 3.43. (a) Molecular orbital scheme for the Pd(CO)(CH3 )(PH3)2 complex. (b) The Pd dx2−y2 orbital of the Pd(CO)(CH3 )(PH3)2 complex. c) Release of the repulsive interaction between CH3 and CO by back-donation into the empty dx2−y2 orbital.

The activation of C–H or N–H or alkane C–C σ bonds is much more di cult than for π bonds since the unoccupied antibonding orbitals of the former are much higher in energy. The bond that is to be broken has to be significantly stretched from its initial state as one proceeds along the minimum energy reaction path, but still maintain a significant bond order. In the transition state the CH bond is extensively stretched. The corresponding transition states are therefore best characterized as late transition states (see Chapter 2 for the definition).

The activation of the bond that is broken occurs via a similar mechanism to that for the oxidative addition over an organometallic substrate. Dissociation leads to the formation of negatively charged adsorbate-fragment orbitals with formally oxidized metal surface atoms. Dissociation typically occurs over the top of a surface atom. The critical point in the activation of the C–H or N–H bond occurs when it is stretched su ciently such that the empty antibonding bond orbital lowers close enough to the Fermi level to allow for back-donation and electron transfer from metal into the antibonding state of the absorbate. This is illustrated in Fig. 3.44 for the dissociation of H2 over di erent metal surfaces (see also ref. [4]). In the oxidative addition reaction, the reactant-molecular

130 Chapter 3

Figure 3.44. The DOS projected onto σg and σu for H2 in the dissociation transition state on Cu(111), Ni(111), and Au(111), and Pt(111) surfaces, adapted from Hammer and Nørskov[4].

bond distance has to be increased so that that empty antibonding molecular orbitals are low enough to become occupied by electron back-donation. In the reverse reaction, the reductive elimination, in order to form a chemical bond, the antibonding intramolecular orbital has to be pushed above the Fermi level so as to reduce the repulsive intramolecular interaction by electron donation to the metal, similarly as illustrated in Fig. 3.44. The activation of methane occurs as depicted in Fig. 3.45.

The activation of the CH bond can occur by stretching it over the top of an Ni atom on the surface or through a valley between Ni atoms. Methane activation occurs with slight preference for the path that proceeds through the valley between Ni atoms. In contrast, CH3 activation occurs by stretching the CH bond over the top of an Ni atom. With the exception of Cu, Pd, Pt and possibly Ir, the CHx fragments that are generated prefer adsorption in three-fold coordination sites on many of the close-packed surfaces. On the excepted metals, the interaction between adsorbate and the highly occupied, spatially extended d-valence electrons forces the CH3 fragment to the atop position (CH3 has one empty sp3 lone-pair orbital that binds to the metal) and the CH2 fragment to a twofold position (CH2 has two empty orbitals), as predicted according to bond hybridization arguments[3] . The di erence in the coordination of CH3 to Ni on which CH3 prefers threefold coordination and Pt where CH3 binds atop derives from the small spatial extension of the d-atomic orbitals of Ni compared with Pt. The repulsive surface–adsorbate d–lectron interactions on Ni are significantly reduced.

Because of the particular geometry of the Ru(1120) surface, the adatoms now prefer two-fold coordination surface sites. The coordination number of the metal atoms on the Ru(1000) surface is nine, on the more open Ru(1120) surface the coordination number is seven. The lower coordinated Ru(1120) surface should therefore be more reactive. The activation energy for CH4 dissociation is found to be lowered by over 30 kJ/mol moving from the Ru(0001) to the Ru(1120) surface, see Fig. 3.46 a and b. In this figure (moving from right to left) the activation energies are compared for subsequent elementary C-H bond activation steps in the decomposition of CH4 to adsorbed carbon. These surfaces were studied at two di erent surface coverages (25% versus 10%). At higher surface occupations the repulsive interactions between adsorbates decrease bond energies and, hence

The Reactivity of Transition-Metal Surfaces 131

Figure 3.45. DFT-computed reaction paths for the dissociation of CH4 and subsequent reaction intermediates on the Ni(111) surface[30].

Figure 3.46a. Reaction energy diagram for CH4 decomposition over the Ru(0001) surface at di erent coverages of intermediates[31a]. Reaction proceeds from right to the left. Top curves (2 x 2) unit cell, low curves (3 x 3) unit cell.

increase activation barriers. Of the dissociated fragments, the relative energies are compared immediately after dissociation and when they are at an infinite distance. Whereas on the Ru(0001) surface the CH adsorbed fragment is most stable, CH2 and CH have comparable energies on the more open surface. CH prefers high coordination sites on the (0001) surface, but is limited on the (1120) surface to two-fold coordination sites. Trends in reactivity for di erent metals are illustrated in Table 3.5.

132 Chapter 3

Figure 3.46b. DFT-calculated reaction diagram of CH4 decomposition on the Ru (1120) surface; (2 x 2) unit cell[31b].

Activation barriers for row 4 metals appear to be the highest, those for the row 5 metals are lower and those for row 6 metals tend to be lowest. Surprising is the low value found for Pd, which may be an artifact of the calculation.

Table 3.5. DFT-computed activation energies in kJ/mol for CH4 decomposition to CH3 computed in a (2 x 2) unit cell[21,32−34,41]

There is an important di erence between the trends found here for methane activation and those reported earlier for CO dissociation. The CO dissociation energy trend is determined by the Oads adsorption energy. Note that whereas CO dissociation on Pt(111) is more di cult than on the Ni(111) surface, it is the reverse for CH4 activation.

Also the activation of C–C bonds and the formation of C–C bonds for partially hydrogenated intermediates do not have to behave similar to that for C–H activation. This follows, for example, from inspection of Fig. 3.51 which will be discussed more extensively later. In this figure, the C–C coupling reactions on Co and Ru are compared. While the barrier for the C–H cleavage of CH4 is higher on Co than on Ru, the barriers for C–C bond formation and C–C bond dissociation are lower on Co than on Ru. On Ru the stronger

The Reactivity of Transition-Metal Surfaces 133

M–C interaction in the product as well as reactant state results in weaker interaction between reacting hydrocarbon fragments, so that barriers for the reaction in both directions increase. This in line with the observed lower rate of hydrocarbon hydrogenolysis observed for Pt as compared with Ni. On Ni the rate of C–H activation is lowered more than that for the CHx–CHy bond cleavage reaction. Because of the stronger metal–carbon bond of Pt than Ni, the rate of methane formation by recombination of adsorbed hydrogen with adsorbed CH3 will be lower on the former.

Now let us compare the dissociation of water with methane. On Ni, the activation energy for H2O dissociation is lower than that of CH4 because of the substantially stronger metal–OH interaction on Ni, Eads(OH)= 326 kJ/mol, as compared with the metal–CH3 interaction. In Fig. 3.47 the reaction energy diagrams are shown for decomposition of H2O to adsorbed oxygen and hydrogen. The dissociation of water over Ni(111) proceeds via a path through the valley of three Ni atoms. The dissociative adsorption of water also strongly depends on coordinative unsaturation of the surface metal atoms. Bengaard et al.[20] predict that the activation energy for H2O dissociation is decreased by 50 kJ/mol on comparing the open (211) surface with the close-packed less reactive Ni(111) surface.

Returning to our discussion of methane activation, as has been pointed out by Liu and Hu[32] and by Abbott and Harrison[35], the experimentally measured activation energies for CH4 di er by more than 30 kJ/mol for the same surfaces. Whereas this may be partially due to an overestimate of the activation energies due to the inaccuracy of density functional theory itself, a more likely explanation is crystal imperfection on the experimental single crystal surfaces. For Rh and Pd the large reductions in activation energies for CH4 and CO dissociation on surface kinks and steps are compared with those on terraces in Table 3.6.

Figure 3.47a. Reaction energy diagrams of H2 O dissociation on the Ni(111) surface; energies with respect to adsorbed hydrogen.

134 Chapter 3

Figure 3.47b. Reaction energy diagram of H2 O dissociation on the Ru(0001) surface, energies with respect to H2 gas phase.

Table 3.6. The calculated dissociation barriers[33] (Eadis) for CH4 (g) → CH3 +H and CO → C+O and the barriers (Eaas) for their reverse reactions on di erent Rh and Pd surfacesa

aThe least coordination number (CN) of the metal atoms involved in the TSs on flat surfaces, steps and kinks are also listed for comparison. The unit of the barriers is in eV.

One notes the large decreases in activation energies for the activation of CH4 on the more coordinatively unsaturated edge and kink atoms. The reaction path proceeds over the top of a single metal atom. Interestingly, for the recombination reaction of Hads and CH3ads, this di erence in activation energies largely disappears. For CO dissociation, on the other hand, the activation energy for recombination also shows a large reduction.

The behavior of CH4 activation follows that which is expected for late transitions as was discussed in Chapter 2. This only applies, however, for systems that have similar reaction pathways. This is not the case, however, to CO. There is a significantly large decrease in the activation energies for CO, where the C and O atom that form in the transition state do not share the same metal atoms, whereas they do on a terrace. Since the reaction paths are very similar on the step and kink sites, the di erences between the activation energies in the forward and backward directions are relatively small.

The Reactivity of Transition-Metal Surfaces 135

Figure 3.48. The activation of NH3 by Pt(111). Reaction energy paths and structures of reaction

intermediates and their corresponding transition states. Reaction energy diagram for the transformation of NH3 to Nad . − − − NH3 ; − · −· NH3 +Oads ; · · · NH3 +OHads [36].

Two classes of reactions can be distinguished: structure sensitive and structure insensitive. The cleavage of the C–H bond that occurs for the oxidative addition is structure sensitive. The reverse reaction for the reductive elimination, however, is not (α = 1!). This is due to the reaction path, that involves C–H activation via a transition state that involves a metal atom that shares bonding to the molecule CH3 and H fragments. More coordinatively unsaturated metal atoms bind the reaction fragments more strongly and this acts to help lower the barrier for C–H activation. In contrast, CO dissociation, as well as CO formation, are both structure-sensitive reactions. The interaction with a large ensemble of atoms with the right topology (a geometric e ect) lowers the energy of the transition state. Low activation barriers are found for the forward and reverse reactions when the fragments do not share metal surface atoms in the transition state. This concept of metal-atom sharing can also be used to help understand the promotion of adsorbate bond activation by coadsorption. Ammonia decomposition, for example, can be activated in the presence of surface oxygen. Figure 3.48 illustrates the reaction energy paths for NH3 activation as found on a Pt(111) surface[36] . On clean Pt, the dissociation of adsorbed NH3 is thermodynamically unfavorable. Experiments indicate that at low temperature NH3 does not decompose on the close-packed Pt surfaces. The presence of oxygen, however, can help to promote this reaction. The activation of NH3 over the clean Pt surface is compared here with activation of NH3 in the presence of coadsorbed atomic oxygen and OH. Nitrogen is the main product at low temperature. The competitive product is NO. Reaction energies and reaction paths for the recombination of N and O adatoms are shown in Fig. 3.49. The recombination of nitrogen adatoms to form N2 is quite similar.