Molecular Heterogeneous Catalysis, Wiley (2006), 352729662X

36 Chapter 2

Figure 2.10. The active site and its local environment for di erent catalytic systems. (A) The single monoatomic zeolitic cation site (Fe in ferrierite); (B) the single site enzyme and its local cavity which provide multiple points of additional contact (methanol oxidase) the ensemble of metal atoms, (C) site for Fe-catalyzed ammonia synthesis.

contribute equally to the catalytic cycle. The second follows the view put forth by Taylor indicating that the sites are not equivalent, and that only a selective few actually control the rate. In either case, what controls the activity and selectivity are the nature and the strength of the chemical bond that forms in the adsorbate surface complex. The detailed atomic structure or topology of the surface can also be important for controlling reactivity. Some reactions are quite sensitive to the topology whereas others are not. These are better known as structure-sensitive and structure-insensitive reactions.

The local chemical interactions predominantly control the activity of the reaction. Environmental e ects around the active site are essential for stereochemical selectivity. Examples include the influence of the cavity on controlling reaction selectivity in zeolites and enzymes. The e ects of ligands in homogeneous catalysis belong to this same category of sterochemical control. In addition, the external environment can also considerably influence the activity. The environment in which the active site sits can be rather complex, o ering a variety of features that could alter the intrinsic kinetics. Catalytic reactions carried out over supported metal particles in the presence of an aqueous medium, for example, can demonstrate changes in the rate and selectivity based on the changes in the transition metal used, the metal particle size and morphology, the exposed surface facets, defect sites on the metal, the support that is used, the interaction between the metal and the support, the composition and relative position of a second metal, the solution phase and its properties near the active metal substrate. Many of these features are captured in Figure 2.11.

In order to elucidate which of the structural features control catalytic performance, it will be critical to establish not only the local bond-making and bond-breaking events that drive the catalytic reaction but also the influence of the active site environment. We examine some of the general ideas on the influence of the environment in the next few sections of this chapter.

We start by distinguishing between two major e ects. The first involves the influence of the medium n which the reaction is carried out, while the second refers to the influence of the solid-state matrix in which the active site is embedded. We first summarize cavity, ligand and solvent e ects which comprise the catalytic medium. This is then followed by a brief discussion on the di erent matrix e ects.

Zeolites form their own category of heterogeneous systems in that they have well-defined ordered micropores with dimensions that are similar to those of substrate molecules.

For reactions in zeolites and enzymes, the cavity created by the inorganic framework of

Figure 2.11. The complexity of the catalytic center in its reaction environment is illustrated here with a schematic of the hydgrogenation of glucose over carbon-supported Pd particles that are alloyed with Ru in an aqueous medium. The interaction between the metal and the support, the surface composition of Pd and Ru and possible nucleation of Ru clusters, the aqueous medium and its wetting of the metal and the support, the presence of other metal and support surface intermediates such as hydroxyl groups can all act to influence the catalytic reaction.

the zeolite and the peptide backbone of the enzyme can o er specific stereochemical control of the reaction. For reacting molecules adsorbed in these cavities, the intermolecular bonding can be broken down into interactions between the framework and the adsorbate that lead to activation of bonds in the substrate molecules and interactions of the substrate with the framework through physical forces such van der Waals and electrostatic interactions and hydrogen bonds that a ect the positioning of the substrate molecules in the cavity. For larger molecules, the latter forces control the matching of size and shape of the cavity with the size and shape of the reacting molecules. Enzymes, as was discussed above, have many point contacts between the protein backbone that comprises the reaction cavity and the substrate molecule. More detailed discussions on the influence of the cavity on both activity and selectivity in zeolites and enzyme catalysis will be presented in Chapters 4 and 7 respectively.

On two-dimensional surfaces the stereochemical control can be moderated by the addition of enantiomeric molecules that play a similar role to the enzyme cavity in providing steric control induced by the weak physical interactions in three dimensions between the enantiomeric coadsosorbed molecule and the reacting substrate molecules. These interactions are analogous to those that occur between the ligands and the substrate molecules that control the selectivity in organometallic homogeneous catalysis. Moderation of heterogeneous catalytic surfaces and ligand e ects in homogeneous catalysis are discussed in Sections 2.3.6 and 2.4.

Solvent e ects play a role similar that of the cavity in influencing the catalytic activity by providing a medium with a di erent dielectric constant that can stabilize or destabilize polar transition states. The di erence here, however, is the lack of stereochemical control. Protonic solvents open up new reaction channels involving proton transfer-mediated

38 Chapter 2

reactions. In polar solvents, the electrochemical activation of molecules can also become possible thus, allowing the surface chemical potential as an additional variable in tuning the activity. The influence of solvent and electrochemical modulation on the catalytic activity is described in more detail in Chapter 7.

In order to understand the influence of the solid-state matrix on activity, some of the relevant structural aspects of the di erent heterogeneous catalytic systems will first be reviewed. The catalytically active material for many heterogeneous systems is usually dispersed over an inert, high surface area support in order maximize the surface area and stabilize the particle size of the active material on the support. The chemical interactions between the support and the active particle often cannot be ignored as they can influence catalytic activity. This interaction with the support occurs for a number of heterogeneous systems including metallic, bimetallic, metal oxide and metal sulfide particles.

The size and shape of the active particles as well as their activity can be greatly a ected by the interaction between the active particle and the support. In addition, the structure of the support including its shape, size and porosity can have a significant influence on the overall catalytic behavior, predominantly due to extrinsic factors that result from mass and heat transfer phenomena. These extrinsic support e ects are not explicitly covered in this book. The interested reader is referred to the books by Thomas and Thomas[3] and Farrauto and Bartholomew[21].

For heterogeneous systems, the solid-state matrix can influence the activity by altering both electronic and structural features about the active site. We can distinguish two types of solid-state matrix e ects. The first involves the embedding of the active site within the catalytically active particle and the indirect changes that arise from the interaction of the active particle and the support. Examples of the direct e ects include the overall size, shape and morphology of the metal particle and the composition and the specific atomic arrangements in alloy particles.

The indirect e ects related to the particle-support interactions include the wetting of the particle on the support, which influence both the shape and size of the particles that form, and the stress or strain that the support imparts on the electronic structure of the particle.

A second e ect involves the chemical nature of the interface between the particle and the support, which can influence the reactivity of the active particle. For instance, in metal particles, the metal atoms at the interface are often not completely reduced. This can lead to unique activity at the interface and can lead to a perturbation in the chemical reactivity of active centers removed from the interface. As a third e ect, the support can impart unique properties to the particle due to charge transfer between the active particle and the support and electronic perturbations due to structural defects in the support which would influence its reactivity.

We conclude this introductory with a short outline of the sections that follow. In Section 2.3.2, we o er a brief introduction to pressure and material gap problems that arise when model catalytic systems and conditions are used to emulate working catalytic systems. In Section 2.3.3, we describe the local aspects of the catalytically reactive sites in the section titled “Ensemble e ects and defect sites”. This is followed by four sections on environmental influences on the reactivity entitled “Cluster size and metal supports”, “Cooperativity”, “Surface moderation by coadsorption of organic molecules” and “Stereochemistry of homogeneous catalysts”. The chapter is concluded with a section on surface kinetics, dealing with surface reconstruction and transient reaction intermediates.

40 Chapter 2

determine their catalytic kinetics. Various general explanations may be possible that help to explain this comparison of the rates between the two di erent surfaces. One explanation might be that the dominant exposed Ni surface on the supported catalyst may just so happen to be the same as that for the Ni(100) surface. Alternatively, during reaction there might be a surface reconstruction that acts to expose the same surfaces under reaction conditions on both the model single crystal and the supported particles. Alternatively, the reaction may by catalyzed by surface defect sites that are similar on both the model and actual catalytic surfaces. As we have learned in Section 2.2, CO dissociation preferentially at a step defect site. This is considered the rate-limiting step for the methanation reaction and may play an important role. We will elaborate more on this issue in the next section.

In turning to the pressure-gap problem, we note that this same reaction carried out under UHV would show a marked decrease in the rate of reaction as measured by the catalytic turnover number. The turnover number is defined as the rate of a reaction normalized per reactive surface site.

The di erence in the rates is due to the di erence in the pressure when the reaction under atmospheric conditions is compared with vacuum conditions (10−6 bar). In general, the desorption energy of adsorbed CO from a transition metal surface tends to be close to the activation energy for its dissociation (see Section 3.3 for details). Since the activation entropy for a surface reaction is low (tight transition state), but the activation entropy

for desorption is high (∆S‡surf/∆Sdes‡ ≈ 10−1), the rate of CO desorption will be orders of magnitude faster than that of dissociation under UHV conditions. Only at significantly

higher pressures, such as performed under atmospheric conditions, will the steady-state surface concentration of CO remain high enough at the temperatures necessary to dissociate CO. Under atmospheric conditions, CO dissociates with an overall rate fast enough to compete with desorption.

High reaction pressures are needed for many other systems as well in order to convert the surface into a uniquely reactive state such as has been found for ethylene epoxidation. The epoxidation reaction of ethylene catalyzed by silver shows a distinct pressure gap. Higher oxygen pressures are needed in order to convert the silver surface into a silver-

oxide overlayer where weakly adsorbed oxygen atoms are formed, that selectively epoxidize ethylene[23] .

Higher surface coverages will not only influence the rate but can also change the selectivity for a reaction. An interesting example is the low-temperature oxidation of ammonia over Pt (see Sections 3.3 and 6.1). Single crystal studies of ammonia oxidation demonstrate the presence of only two products, N2 and NO, when a Pt single crystal is exposed to molecular beams of ammonia and oxygen. Ammonia oxidation studies carried out under atmospheric conditions, however, reveal the formation of N2O. N2O can be formed from NO at step edges, which may not be present on the ideal single crystal surfaces studied, or requires weakly adsorbed intermediates not formed under UHV conditions but only at high coverage. Unique intermediates such as NO3− may also form when the surface is slightly oxidized owing to its exposure to higher oxygen pressure.

An example of such high-pressure e ects are the studies of ethylene hydrogenation. Hydrogenation of ethylene[24] by a Pt(111) surface under atmospheric conditions occurs by a weakly adsorbed ethylene species (π-bonded), that predominantly forms only under reaction conditions, when the surface is nearly completely covered with strongly adsorbed (σ-bonded) ethylene. Similar results were also found computationally over Pd(111) with the exception that both π and weakly held di-σ intermediates were found to be active.

42 Chapter 2

increasing Cu concentration. The dehydrogenation of cyclohexane requires only a small ensemble of reactive atoms, whereas the hydrogenolysis requires a large ensemble of atoms. To explain the latter, the intermediates sketched in Fig. 2.14 have been proposed. The bond cleavage between atoms 1 and 2 along with the subsequent hydrogenation of the intermediates that form result in the formation of hydrocarbon fragments adsorbed on the transition-metal surface atoms.

Figure 2.14. Multi-point adsorption of heptane on a metal surface. Surface ensemble requirement for the hydrogenolysis reaction.

The ensemble e ect is typically inferred when one realizes that a molecule that dissociates requires at least two di erent surface sites to accommodate the molecular fragments. For instance, the rate constant for CO dissociation can in an elementary fashion be written as:

where kdiss is the elementary rate constant for the dissociation of CO. This expression predicts a strong dependence of the rate on CO coverage. This is the result of the ensemblesize requirement for the dissociating the CO molecule. Experimental confirmation of this coverage dependence can be deduced from measurements of the rate of methane formation from pre-adsorbed CO, as for instance illustrated in Fig. 2.15. As was discussed earlier, the rate of methanation is typically controlled by the dissociation of CO. We can, therefore, use the rate of methane formation to help probe CO activation.

Figure 2.15. The rate of methane formation reflects the rate of CO dissociation as a function of CO coverage on the surface of a rhodium catalyst[29].

Using the Langmuir adsorption expression for CO, Eq. (2.21) can be rewritten as:

1 + Keqads p(CO)

One recognizes here Sabatier behavior in the volcano-type dependence of the reaction rate as a function of the adsorption equilibrium constant for CO. One can also note the positive order of the rate constant in CO pressure at the left of the Sabatier maximum and the negative order in CO pressure to the right of the Sabatier maximum.

The structure insensitivity of hydrogenation or dehydrogenation reactions may be the result of competing influences. One has to distinguish changes in adsorbate metal-surface atom bond energies by electronic changes on the metal atoms, due to their altered environment (Scheme 2.2a) from changes in the adatom bond energies due to changes in the specific coordination of the adatom (Scheme 2.2b).

Scheme 2.2 (a) The interaction of A with M changes with a change in the electronic structure of M due to neighboring atoms M . (b) The interaction A with the surface changes because one of the coordinating atoms M is replaced by M .

Experimental[30] and theoretical[12,31] results are available that indicate a general ligandtype surface chemical e ect occurs upon alloying. The ligand e ect refers to change in the environment about the adsorption site without changes to the geometric structure of the local adsorption site. While the structure and composition of the adsorption site remain the same (Scheme 2.2a), the local environment of the M atoms within the adsorption complex have di erent electronic properties owing to their external environment that arises from its interaction with M . This subsequently alters the A–M interaction.

While the ligand e ect can change the nature of the adsorbate-metal surface bonding and reactivity, it is an indirect electronic e ect and is therefore less predominant than more direct changes in the A–M adsorption complex. For instance, in Scheme 2.2b, when M is substituted by atom M and the M–M bond energy is larger than the M–M bond energy, the adatom surface-metal cluster interaction energy will decrease much more strongly than if M were only at a neighboring site. This is in agreement with qualitative deduction based on the principle of Bond Order Conservation (BOC)[32] . According to BOC principles, more extensively discussed in Chapter 3, the interaction energy of an adsorbate with a surface atom depends on the number of bonds and their bond strength with neighboring atoms of the surface atom. The interaction energy increases with a decrease in the bond order of the interacting surface-metal atoms with their embedding environment. The ensemble e ect (Scheme 2.2b) ascribes changes in adsorption energy to direct di erences in atoms within the adsorption complex.

Neurock and co-workers[33] used ab initio calculations to determine the influence of both Ag and Au on Pd for the hydrogenation reaction of ethylene and acetylene.

44 Chapter 2

The electronic ligand e ect due to alloying Pd with Au or Ag is significantly smaller than the geometric (ensemble) e ect but does play a role. The activation barriers are reduced by about 10 kJ/mol over the alloy owing to weakening of the metal-hydrogen and metal-carbon bonds as the result of the electronic e ect[33b] . The binding energies for both ethylene and hydrogen , however, are reduced more substantially if Au (or Ag) is actually part of the adsorption complex. This is known as the geometric or ensemble e ect (Scheme 2.2b). The weaker metal-hydrogen and metal-hydrocarbon bonds enhance the intrinsic hydrogenation activity. This promotional e ect, however, is o set by the fact the hydrogen-Au interactions are so weak that they limit the amount of hydrogen on the surface. The rate of hydrogenation is typically first order or lower in hydrogen. Therefore, decreases in the hydrogen surface coverage will inherently decrease the intrinsic rate. Upon alloying, the order of the reaction rate in hydrogen tends to increase.

The ab initio results were used to develop an ab initio kinetic Monte Carlo scheme to follow the rates of ethylene hydrogenation and the influence of Au[33a,c,d]. The de-

tails of the kinetic Monte Carlo method and its application to ethylene are presented in more detail at the end of Chapter 3. The simulations explicitly followed the adsorption, surface reaction, desorption and di usion steps, surface site specificity, and lateral interactions between surface adsorbates throughout the simulation. The simulations examined the steady-state catalytic hydrogenation of ethylene at temperatures and pressures of interest. The surface was then alloyed with di erent compositions and di erent atomic configurations of Au to examine its e ects on the molecular transformations and on the overall rate and selectivity to specific pathways. A snapshot from one of the simulation runs which captures the atomic level detail of the simulation is shown on the cover of this book. The hydrogen atoms in the simulation predominantly sit at the three-fold Pd sites. This significantly lowers the surface coverage of hydrogen, which should act to lower the rate since the reaction is first order in hydrogen. The intrinsic activation barrier for hydrogenation over the alloy, however, is also lowered owing to the presence of Au. This increases the likelihood for each reaction. Two of these e ects (lower hydrogen surface coverage and weaker adsorbate bonding) should o -set one another, maintaining more of a constant rate. Indeed, the simulated turnover for ethylene hydrogenation showed little change upon changing the relative amount of Au or its specific location. These simulation results are consistent with experiments carried out by Davis and Boudart[34] that showed that alloying Pd with Au had little e ect on the measured turnover frequency.

Whereas there is little change in catalytic activity, the selectivity over the alloy is significantly improved. The hydrogenation of ethylene can produce significant amounts of ethylidyne or other carbonaceous intermediates on pure Pd surfaces especially at higher temperatures where competing dehydrogenation and carbon-carbon bond breaking steps become more prevalent. These paths are typically much more structure-sensitive since they require larger surface ensembles. Ab initio based simulations showed that the addition of group IB metal such as Au or Ag can act to minimize the number of these larger ensembles and thus reduce the unselective decomposition paths.[33a,33c,33d]

There are a range of di erent catalytic reactions where explicit changes in the ensembles upon alloying can change the activity and the selectivity. An understanding of how the surface-ensemble size, morphology and specific atomic arrangement influence activity and selectivity could ultimately be used in the “design” of specific arrangements to optimize catalytic performance. Vinyl acetate synthesis, which involves the acetoxylation of ethylene in the presence of oxygen carried out over Pd and PdAu alloys, is significantly influenced by alloy composition. Kinetic experiments in which ethylene, acetic acid and

Figure 2.16. (a) Snapshot from the steady-state simulation of VAM synthesis over pure Pd(111).

(b) Simulation of surface coverage for vinyl acetate synthesis in the presence of PdAu alloys[33].

oxygen were pulsed over supported Pd and PdAu alloys to elucidate the temporal synthesis of vinyl acetate indicate that both the activity and the selectivity of this reaction were improved by alloying Pd with Au[35] . Ab initio calculations were coupled with kinetic Monte Carlo simulations to examine how changes in the surface composition, ensemble size, shape and specific structural arrangements of Pd and Au for model substrates influence the simulated activity and selectivity[33]. The simulation of vinyl acetate synthesis over Pd(111) results in a very low production of vinyl acetate. The surface is essentially poisoned by the acetate and oxygen intermediates that form. At the steady-state, ethylene has a di cult time finding free Pd sites available on which to adsorb. Higher pressures of ethylene are required in order to adsorb ethylene to any appreciable degree. This is consistent with experimental results.

Alloying Au into the surface opens up surface sites for ethylene to adsorb, since ethylene is only 20 kJ/mol more weakly bound to Au than to Pd. Acetate and oxygen, however, are much more strongly bound to the surface and will therefore tend to bind selectively only to the Pd sites. Au decreases the surface coverage of acetate, and even more importantly, provides sites where ethylene can adsorb exclusively. Well-dispersed Au will therefore create ensembles where ethylene is surrounded by acetate intermediates, thus creating more active local environments. Figure 2.16, for example, shows the steady-state surface population (coverage) over a well-dispersed PdAu (111) alloy surface.

Acetate and oxygen adsorb exclusively on the bridge and three-fold Pd sites, respectively. Ethylene, however, adsorbs on Au sites where it reacts with neighboring Pd to form vinyl intermediates. These vinyl groups react with neighboring acetate to form vinyl acetate. The simulated activity and selectivity in this system was improved by a factor of 2 and 3%, respectively. The size and shape of the Pd and Au ensembles were subsequently tailored to enhance the self assembly of ethylene and acetate under reaction conditions and optimize the simulated activity and selectivity. The Au island sizes in Fig. 2.16, for example, were found computationally to improve the activity by an order of magnitude and the selectivity by 7%. Whether these surfaces can be made and whether or not they are stable is an important issue. Note that qualitatively very similar results were found for simulations which follow the mechanism where ethylene first couples with acetate and