Molecular Heterogeneous Catalysis, Wiley (2006), 352729662X

66 Chapter 2

Figure 2.29. The enantioselective hydrogenation by homogeneous Rh–BINAP catalyst.

reactions are catalyzed by the same reaction center. Chiral phosphine ligands attached to cationic Rh catalysts are e cient enantiomeric hydrogenation catalysts. The non-bonding interactions between the reactant alkenes adsorbed to the metal center and bulky ligands force the reactant molecules to adsorb in a strongly preferred orientation. Figure 2.29 illustrates this for enantioselective hydrogenation to produce the L-Dopa product molecule. Mechanistic studies by Landis and Halpern[103] helped to elucidate the physical organic chemistry of this reaction. They performed kinetic studies on the hydrogenation of methyl(Z)-acetamidocinnamate by the Rh(DiPAMP) catalyst (Fig. 2.30) and proposed two specific pathways, namely the hydride and the alkene pathways. According to the hydride pathway H2 initially dissociates over the Rh center to form a alkene Rh hydride. This is subsequently followed by the reaction between an alkene and dihydride. The alternative pathway is termed the alkene pathway. Hydrogenation now occurs after alkene binding. This actually appears to be the preferred pathway (see Fig. 2.31).

Interestingly, the major product comes from the very rapid hydrogenation of the less stable diastereomer with H2. This feature that the least-stable intermediate is actually the most reactive has been called anti-lock and key behavior. The enantioselectivity of the reaction is a ected by the competition between the rate of hydrogenation and the interconversion of the two diastereomers.

An increase in the hydrogen pressure suppresses the enantiomeric yield. Figure 2.32 [105], shows the computed free energy diagrams for this hydrogenation reaction. The calcu-

lations were performed using hybrid methods, involving DFT calculations embedded into a Molecular Mechanics force field to describe properly the interaction between reactants

68 Chapter 2

The results can be best explained as being representative of a loose transition state where the barrier height is dominated by the need to weaken the interaction between catalyst and substrate. The unfavorable adsorption state appears to be the most reactive. It is also the adsorption mode in which the entropy is a maximum.

2.4 Surface Kinematics

2.4.1 Surface Reconstruction

The adsorption of adatoms, and also other strongly bound surface intermediates, weakens the metal–metal bonds in the surface layer and between surface and subsurface. The bond strength between the adatom (A) and the surface-metal atom (M) is sensitive to the coordinative unsaturation of the surface metal atoms. The combination of these two e ects can result in a rearrangement of the surface metal atoms to increase the surface energy when the surface is covered by an overlayer of adsorbed atoms.

Figure 2.33. Overlayer energies of O adsorbed on Cu(110)[129].

Figure 2.33, for example, illustrates how the removal of a row of Cu atoms on the Cu(110) surface releases the stress in the strained overlayer of adsorbed atomic oxygen, thus enhancing surface reconstruction.

The details of the surface reconstruction depend on the surface as well as the adsorbate overlayer composition. Figure 2.34 illustrates this for the adsorption of CO, H2 and O2 on the Pt(110) surface. Each di erent adsorbate overlayer generates a di erent surface metal atom topology. Surface reconstruction has several important consequences. When such a reconstruction occurs, the density of the surface atoms changes. Hence, there are local regions on the surface where there is accumulation of surface atoms and other regions where there is a depletion. This creates defects such as edge and kink sites. As we mentioned in the previous section, the edge or kink sites are often the sites where dissociative adsorption occurs. Reconstruction of surfaces can significantly a ect the reactivity of catalytically active surfaces because of this generation of highly reactive defect sites.

Figure 2.34. The reconstruction of the Pt(110) surface upon exposure to H2, O2 and CO. (a) nested missing-row reconstructions; (b) (1:1) micro incets; (c) unreconstructed (1 x 1) terraces separated by multiple height steps. Adapted from B.J. McIntyre et al.[106].

A predictive theory of heterogeneous catalysis, therefore, should include the prediction of the restructuring phenomena that occur when a catalyst is brought into its reactive phase. This is especially important when surface roughening occurs with the creation of reactive steps and kinks. Surface reconstruction is driven by the surface’s desire to minimize its surface energy. Sometimes a surface which is free of adsorbates will reconstruct its bulk terminated surface, as for instance the hexagonal reconstruction of the Pt(100) surface. Surface reconstruction of clean surfaces occurs predominantly on transition metals with spatially extended d-valence atomic orbitals and with high electron occupations

such as Pt and Au.

According to van Beurden and Kramer[107], the clean surfaces of these metals reconstruct because of the low values for vacancy formation as compared with the respective

cohesion energies. The surface atom density changes upon reconstruction, therefore the heat of reconstruction, ∆Hr, is given by[108]

where ∆N is the di erence in the number of atoms in the reconstructed layer and N is the total number of atoms in the reconstructed layer, En is the total energy before reconstruction and Er is the total energy of the reconstructed system. The heat of reconstruction is seen to be explicitly dependent upon the cohesive energy. The large spatial extent of the d-valence atomic orbitals, characteristic of the group 5d transition metals, and the nearly complete electron occupation generate a strong repulsive contribution to the metal–metal bonds to be overcome by the attractive contribution of the free–electron type s- and p- valence electrons. Vacancy formation becomes easier because it reduces these repulsive interactions. A similar reduction of the repulsive interactions between the highly occupied d-valence atomic orbitals assists the stabilization of the longer bond distances found in the reconstructed surfaces.

Covering the surface with an adsorption layer changes the surface energy. For Langmuir

70 Chapter 2

where γ is the surface energy of the surface layer covered with adsorbate, γ0 is the surface energy of the free surface, a is the surface area of the adsorbate, θ is the coverage, p is the partial pressure of adsorbate and K is the adsorption equilibrium constant. Surface reconstruction occurs when the increased surface energy of the surface metal atoms with low coordination is compensated for by the increased adsorption energies. This is illustrated schematically in Fig. 2.35, where the changes of the two surface energies are plotted as a function of partial pressure of the adsorbate. Surface (2) is the less stable surface, with the higher adsorption energies. If the two curves cross, as indicated in the figure, there is a critical pressure pc beyond which there is a driving force for reconstruction.

The existence of a critical pressure pc, beyond which surface reconstruction occurs, defines a critical coverage θc for each surface that is related by pc through the adsorption isotherm.

Figure 2.35. Schematic illustration of the dependence of the surface energy on partial pressure. Surface

(2) has surface atoms of lower coordination than surface (1), and hence is most reactive. pc is the partial pressure beyond which there is a driving force for reconstruction of surface (1) to surface (2).

Theory can be used to compute surface energies as well as the adsorption energies and entropies. Hence a theoretical prediction of Fig. 2.35 can be made for any two surfaces. Surfaces covered with strongly bonded atoms such as C are often found to have a critical coverage beyond which there is a driving force for reconstruction.

A theoretical analysis of the surface reconstruction dynamics is possible using first principle calculations. For example, Molecular Dynamics was recently used to study the reconstruction of the Pt(111)hex surface phase by CO as a function of time[107,111]. These simulations employed embedded atom potentials that were developed based on quantummechanical calculations of the surface.

The local surface concentration of molecules and adatoms can change over the course of a catalytic reaction. This is intrinsic to the catalytic reaction cycle. Reconstruction can

deactivate the catalyst surface due to presence of strongly bound adatoms or the formation of a non-reactive surface. In addition, reconstruction can drive particular elementary steps that lead to autocatalytic behavior. For example, surface reactions in which a product molecule is formed and subsequently desorbs, often require a vacancy to activate a reactant molecule and usually create more vacancies than were initially present (see Chapter 8). Such reactions are autocatalytic in the number of vacancy sites as is shown in the following equation

R + −→ P + 2

where R and P refer to the gas-phase reactant and product, respectively, and and R refer to free and reactant-covered sites.

Under particular conditions, the combination of deactivation that results from surface reconstruction and the activation that occurs via autocatalytic steps in the surface overlayer ultimately lead to transient collective self organizing pattern formation such as moving spirals, pulsars or other patterns[112] , known as Turing structures[113] (see Section 8.3). These structures are named after the mathematician Turing, who discovered such self organized structures when di usion of the participating components is very di erent. In the Turing system, one chemical component is autocatalytic. The product itself enhances the rate of its own production. It also catalyzes the production of another chemical component. The second chemical, however, inhibits production of the first. Self organizing phenomena in catalysis will be extensively discussed in Chapter 8.

In the heterogeneous catalytic system, the reacting molecules, and the metal-surface atoms, can be quite mobile. This leads to locally ordered structures when synchronization or self organization phenomena are present however, disorder in the surface layer prevails when these phenomena are absent. For example, it has been proposed that the catalytic oxidation over mixed metal oxides, discussed in Section 2.3.5, actually occurs in the disordered overlayer that forms at the surface under reaction conditions.

The epoxidation of ethylene which is catalyzed by Ag and promoted by chlorine compounds, for example, is thought to occur in a surface overlayer that has features similar to a melt of Ag ions. The silver-oxychloride reactive surface layer requires Ag3+ ions (as in the electrochemical system, see Scheme 2.1) to enhance the overall selectivity. Reduced Ag clusters, however, are required to activate molecular oxygen. Dynamic events between these two states are necessary to close the catalytic cycle. Chlorine in combination with Cs is added to promote the Ag catalyst. Eutectic melting points of this phase are close to the reaction temperature[114] .

An interesting phenomenon that nicely illustrates the consequence of the dynamic surface events is shown in Fig. 2.36, which shows the structure of an Ir-surface, after exposure at 1000 ◦C to an oxidizing mixture of methane giving CO, CO2 and H2O.

Within half an hour the initial surface, which was flat, is transformed into a mountainous landscape with altitude di erences on the order of 1 µm. Kramer[115] has estimated that

in processes that result in these transformations each surface atom can jump at a rate of 106/sec, whereas the elementary reaction rates that occur under these conditions occur at a rate of 104–105/sec. The most adequate picture of this reactive surface was suggested to be that of a dynamically changing surface that reacts with a quasi-static ensemble of adsorbed molecules or molecular fragments. The surface etching process is the result of the balance between momentous stable surface reconstruction and destabilization of

72 Chapter 2

Figure 2.36. An interference–microscopic photograph of a iridium surface after one hour of catalyzing the CPO process at 1000 ◦C. The di erent color densities indicate di erences in height of about 1 µm[115] .

the surface due to reaction events. In the case of the Turing instability, this leads to self organization e ects that are stationary during reaction. In the above example, no balance between the two is reached. As a consequence, the reconstruction processes persist in an irreversible manner until finally a stationary situation is reached.

A second well-known example where strong restructuring of the surface occurs is the high-temperature ( 1200 K) Ostwald process in which ammonia is oxidized to NO over a Pt/Rh gauze. Under industrial conditions, so-called cauliflower structures develop. These changes occur during the first hours of operation and are assumed to play an important role in improving catalyst performance[40] . Restructuring occurs in two stages with initial formation of parallel facets followed by growth of microcrystals[116].

In the previously discussed Fischer–Tropsch reaction catalyzed by Co, adsorbed reaction intermediates were able to lead to reconstruction, thus resulting in the formation of a roughened surface which contains edge, kink and hollow sites[117,118].

As we will discuss in Chapter 3, two factors control the unique reactivity of the step versus the terrace. The first deals with the fact that the step provides the ability to separate the reaction fragments so as they do not share bonds to the same metal atoms in the transition state. This will favor both bond-making and bond-breaking reactions. CO dissociation and carbon–carbon coupling are two examples of bond-breaking and bondmaking reaction steps in FT catalysis that are significantly enhanced at the step edges as a result of lowering metal atom sharing. These elementary surface reaction steps are sensitive to the surface topology in both directions.

The second feature of the step is the under-coordinative saturation of the metal atoms at the top edge of the step. For reactions that only proceed over the edge atom, these sites tend to enhance bond-breaking steps but may slow bond-making steps owing to the enhanced binding of the fragments. For reactions that proceed through a transition state whereby the reaction fragments share a metal atom, the rate of dissociation is enhanced at the step due to the decrease in metal atom sharing and the enhanced binding of the product fragments. The rate of bond-making will depend on the degree of metal-atom sharing in the transition state since the enhancement due to the decrease in metal atom sharing now competes with the impedance due to stronger metal-adsorbate bonds of the reactant at the step edge. When the transition state is late, only the forward bond cleavage reaction step will be sensitive to the surface topology. The backward reaction will not depend on the surface topology. Methanation reactions, for example, are typically suppressed at the step edge whereas carbon–carbon coupling is enhanced.

The above arguments illustrate the importance of edge and kink sites in catalysis. As a consequence, reconstruction phenomena that change also the edge and kink site distribution can have a large e ect not only on the overall rate of a catalytic reaction but also on its selectivity. The latter occurs when competing elementary reaction steps have di erent sensitivities.

Surface reconstruction is inherent to surface oxidation and sulfidation chemistry. In involves essentially surface corrosion and surface compound formation phenomena. The state of a surface can change from a metallic state to that of a solid oxide, sulfide, carbide or nitride depending upon the reaction environment. The surface of the epoxidation catalyst, discussed earlier, in the absence of Cl or Cs, for example, has a composition similar to AgO in the oxidizing reaction environment of the epoxidation system. The oxidation of CO over Ru can readily lead to the formation of surface RuO2 (see Chapter 5). In desulfurization reactions the transition-metal surface is converted to a sulfide form. The reactivity of the surface in these systems begins to look chemically more similar to that of coordination complexes. This we will illustrate in Chapter 5 for the CoS/MoS2 system.

2.4.2 Transient Reaction Intermediates in Oxidation Catalysis

Earlier in the section on the pressure gap, we saw that adspecies with unique reactivity may appear when the surface state changes as a function of coverage. Weak and highly reactive species that are not stable at low coverage may develop. The π-bonded ethylene intermediate which forms at higher surface coverages present under actual reaction conditions on di erent metal surfaces is one such intermediate which has been found to be important in the hydrogenation of ethylene. Notwithstanding the short residence time of the weakly bonded reactive intermediate, the rate of reaction is finite owing to the increased concentration at high pressure. The residence time is equal to kdes −1. The rate depends on the surface coverage of the reactive intermediate, θr , which is a strong function of the surface state.

Short-lived, highly reactive intermediates can form on a surface upon molecule dissociation. Carley et al.[119], for example, demonstrated that a uniquely reactive atomic oxygen (O−) is formed in the oxidation of ammonia by oxygen over Cu, Zn, and Mg. This unique O− develops the instant that O2 dissociates. The oxygen atom that forms does not fully equilibrate with the surface to form O2−. Upon dissociation, molecular oxygen moves through a high energy state to overcome the reaction barrier. Immediately after reaction, the atoms that result are not energetically equilibrated, and move over a metal atom position before they adsorb and equilibrate. These “hot” O−-type atoms have

been shown to have unique reactivity. Their short lifetime relates to the time required to equilibrate[13b,120] the high energy-atoms to the surface. This will be typically on the order of a few picoseconds. At low pressure this time is independent of the pressure of the reaction gas. It will become shorter at higher pressures where gas-phase molecules collide with the hot atoms. In oxidation catalysis the reactivity of such hot atoms has to be distinguished from the reactivity of short-lived molecular oxygen species such as O2 −, that are highly reactive but have a short residence time not because they desorb, but because they readily dissociate. The surface concentration of such reactant molecules has to be high enough and the activation barrier low enough in order for these species to compete with dissociation. Owing to their short residence times, it is usually di cult to isolate such intermediates spectroscopically or identify their role in the mechanism. The decomposition of NH3 by oxygen is an interesting example that has been studied in detail

74 Chapter 2

over di erent surfaces by Au and Roberts[121].

NH3ads + (O2−)a −→ H2O + Hads− + NO

Theoretical calculations have been performed to demonstrate their importance of the O2− intermediate for this reaction over Cu[122].

On oxide catalysts O2− species have often been proposed to be the intermediates for total combustion. However, it is not always easy to distinguish between the role of short-

lived O2− intermediates or the presence of uniquely reactive oxygen atoms. Unraveling the selective epoxidation mechanism by silver is a good example[123] of the di culty in

establishing the reactive oxygen form.

In the earlier mentioned ethylene epoxidation reaction catalyzed by Ag, initial spectroscopic isotope exchange and chemisorption data indicated that adsorbed O2 species

were responsible for epoxide formation, and that the oxygen adatoms were responsible for activating C–H bonds, hence leading to total combustion[124]. These conclusions have been disputed on the basis of kinetic experiments by Force and Bell[125] that lead to the

interpretation that the reactivity of adsorbed oxygen atoms to Ag strongly depends on the state of the oxygen overlayer. Low concentrations of adsorbed oxygen create electronegative oxygen adspecies that help to activate the C–H bond; oxygen adsorbed at high

oxygen coverage is much more electrophilic and therefore prefers to insert into the ethylene π bond. The main evidence[123] now supports the Force and Bell point of view. The

intermediate leading to oxidation is proposed to be the oxymetallocycle[126] as sketched in Scheme 2.4.

Scheme 2.4 The oxymetallocycle was proposed by Barteau[128] to be the active intermediate for ethylene epoxidation. Surface experiments by Bocquet et al.[127] indicate that electron-deficient Ag atoms coordinate with adsorbed ethylene with moderate energy. This is the precursor state to the oxymetallocycle complex. Oxygen activation can also occur directly on the oxide overlayer.

An interesting example where the O2− intermediate forms and appears to play an important role in the catalytic reactivity is the high-temperature oxidation of methane over La2O3. The reactive O2− intermediate here is not generated directly from adsorption of O2 on the surface, but indirectly by a unique dissociation of molecular oxygen over the lattice oxygen atoms of La2O3. Lanthanum oxide is non-reducible oxide which has a

high a nity for oxygen which helps to make this path possible. Calculations by Palmer et al.[128] have demonstrated that the activation of oxygen given in Scheme 2.5 occurs with

low endothermicity of 50 kJ/mol and an activation barrier of 132 kJ/mol.

Owing to the endothermicity of the reaction, the equilibrium concentration of the resulting O2− will be low and their residence time short. In Scheme 2.5, the reaction of O2 over the La2O3 surface converts the surface O2− anion which is attached to La to a reactive surface superoxide O2− intermediate. These anions are of radical type character and can subsequently activate CH4 to produce a CH3 radical and OOH. This reaction

Scheme 2.5 The reaction of O2 over La2O3 to create reactive O2− surface peroxide intermediates.

was found to 116 kJ/mol endothermic. Subsequent water formation from the recombination of surface OH radicals is an exothermic process. The reactivity of the oxygen atoms of the stoichiometric La2O3 surface was found to be very low, as can be deduced from the endothermicity of the reaction with CH4 (370 kJ/mol). In contrast, the reactivity of surface-defect centers is high, but such centers require high energy to be generated.

2.5 Summary: Concepts in Catalysis

In this chapter we introduced the basic physical chemistry that governs catalytic reactivity. The catalytic reaction is a cycle comprised of elementary steps including adsorption, surface reaction, desorption, and di usion. For optimum catalytic performance, the activation of the reactant and the evolution of the product must be in direct balance. This is the heart of the Sabatier principle. Practical biological, as well as chemical, catalytic systems are often much more complex since one of the key intermediates can actually be a catalytic reagent which is generated within the reaction system. The overall catalytic system can then be thought of as nested catalytic reaction cycles. Bifunctional or multifunctional catalysts realize this by combining several catalytic reaction centers into one catalyst. Optimal catalytic performance then requires that the rates of reaction at di erent reaction centers be carefully tuned.

To predict catalyst performance, one needs to predict the rates of the elementary reaction steps at the catalyst surface. This must ultimately be integrated into a kinetic simulation which treats the interactions between the many di erent adsorbates present on the catalyst surface. In this chapter, we presented rate expressions derived from transition state reaction rate theory as a bridge to connect ab initio quantum mechanical information to reaction rate predictions. In Chapter 3, we present a more extensive treatment of kinetic simulations including many-body interactions and their influence on the catalytic performance.

We use the constructs of transition state theory in order to define the Brønsted–Evans– Polanyi (BEP) relationship, which relates the equilibrium thermodynamics (reaction enthalpy or free energy) with non-equilibrium thermodynamic features, namely the activation energy and activation entropy. A small value of the proportionality parameter in the BEP relationship, α, is identified with an early transition state, whereas values of α that are close to 1 relate to a late transition state. Microscopic reversibility ensures that if the forward reaction is an early transition state then the backward reaction must be a late transition state and vice versa.

The rates of reaction that proceed through early transition states are rather insensitive to changes in the reaction enthalpy, and hence variations in the catalyst, provided that there is no change in the reaction path. On the other hand, reactions with late transition states depend strongly on the reaction energy and hence are quite sensitive to variations in