Molecular Heterogeneous Catalysis, Wiley (2006), 352729662X

56 Chapter 2

Figure 2.22. E ects of particle size on the activity of titania-supported Au for the oxidation of CO[63] .

Figure 2.23. The correlation between the binding energies, for CO molecules and O atoms, with respect

to the coordination number of Au atoms in a series of environments. Binding energies, reported in eV, here are referred to gas-phase CO and O2 [59].

that demonstrate suppression of the rate by CO inhibition at low temperatures.

The adsorbate bond energy increases with increase in the degree of coordinatively unsaturated metal atoms. This is due to the decrease in the localization energy of electrons on the Au surface atoms for structures with fewer neighboring atoms.

As we will discuss in more detail in Chapter 3, the delocalization of electrons is proportional to the square root of the number of coordinating atoms[37]. One would therefore expect adsorbate binding energies to increase with decreasing particle size, owing to the increased number of coordinatively unsaturated surface atoms. The reactivity of these particles with respect to cluster size will then depend the position of the adsorbate bond energy with respect to the Sabatier curve maximum.

2.3.4.5 Quantum Size E ects

As was briefly mentioned above, the increased reactivity of small Au nanoparticles may be the result of the unique electronic characteristics of small Au clusters. There is a clear shift from the bulk metal properties which readily allow electron transfer between the valence and conduction bands of the metal due to small energy di erences between these states. In reducing the size of a metal particle to the nanometer size scale, we lose the band structure

of the metal and form more of a molecular-like structure with discrete energy di erences between molecular orbitals. This results in the formation of a more appreciable energy gap between the highest occupied and lowest unoccupied states and thus the transition from metallic to insulator electronic properties. This is know as a quantum size e ect. The influence of quantum size e ects on the unique low-temperature activity of Au has been speculated by Valden et al.[63], who demonstrated unique reactivity for 3.5 nm Au particles supported on model TiO2 substrates. The size was consistent with the transition from metal to insulator. Theoretical studies by Mills et al.[60] also suggest quantum size e ects. A more extensive description of these e ects can be found in Section 2.3.4.2.

2.3.4.6 Support E ects

The support clearly a ects the rate of some Au-catalyzed reactions. The support can play various roles. First, the support can change the nature of the metal particle adhesion to the surface, and thus change the metal particle size that forms, as was discussed above. Second, the support can act to strain the metal–metal bonds, which would significantly change the electronic properties of the metal atoms near the interface and thus their catalytic properties. Third, there can be electron transfer between the metal and the support, which would change the electronic properties of the metal. Neutral and positively and negatively charged Au clusters have all been proposed to be catalytically active for specific reactions in the literature. Lastly, the interface between the metal and the support can act to create unique bifunctional sites which demonstrate enhanced reactivity. We discuss the last three e ects below. The e ect of particle size on the catalytic performance was discussed in detail in the previous section.

Mavrikakis et al.[64] have nicely shown that the strain induced on the metal–metal bonds by the misalignment of the metal lattice to the registry of the oxide support leads to a shift in the center of the d-band. This change in the electronic structure alters the adsorbate bond strength at these sites, which ultimately dictates the reactivity of the metal. While these e ects may die o for large particles on the support, they can clearly play a role for smaller nanoparticles that are in direct contact with the support.

2.3.4.7 Elucidating Mechanisms and the Nature of Active Sites

Much of the current work on the mechanisms responsible for the unique Au activity have focused on understanding CO oxidation. CO oxidation is generally agreed to proceed by

the adsorption of both CO and oxygen, the activation of oxygen, and the subsequent formation and desorption of CO2.[55,62,66,77,81] It is still debated as to whether atomic or molecular oxygen is the reactive oxygen species. A fair amount of the current evidence appears to point to a bifunctional-type site where CO is adsorbed to the Au at the Au/oxide interface whereas molecular oxygen is activated on a nearby site on the oxide.[55,62,66,77,81] This, however, is still debated. The mechanism, however, is strongly dependent upon the charge state of the metal. We will, therefore, discuss the current thoughts on the mechanisms in the subsections that follow and more specifically analyze the di erent charged states of the metal.

2.3.4.8 Electron Transfer E ects

The unique properties of small Au particles responsible for the low-temperature catalytic activity have not been given a definitive explanation[55]. Neutral and negatively and positively charged gold particles have been identified on di erent metal oxide supports and speculated in catalyzing di erent reactions. The formation of neutral and positively and

58 Chapter 2

negatively charged gold clusters is clearly influenced by the nature of the support and its ability to transfer or accept charge. The reaction conditions can also act to influence charge transfer and modify the oxidation state of the metal. The sensitivity of the reaction conditions, the support used and the reaction examined suggests that active sites for di erent reactions may actually be di erent. That is, the active site for CO oxidation may be di erent to that for water gas shifts and alkene hydrogenation. We discuss experimental evidence along with theoretical results for each of the three di erent Au charged states as potential reaction sites.

2.3.4.9 Neutral Au Clusters

While much of the work in the literature has speculated on the presence of neutral Au clusters, there has been very little in-situ experimental evidence to support these ideas. Calla and Davis[65] claerly showed the presence and the activity of neutral Au clusters under reaction conditions. They followed the CO oxidation reaction over Au supported on Al2O3 using in-situ X-ray absorption spectroscopy as well as with transient isotopic labeling studies. Their results indicate that Au3+ is reduced during calcination and that the active species throughout the reaction appears to be metallic Au[65].

The results from many of the theoretical studies on idealized models of TiO2 and MgO surfaces indicate that Au metal clusters remain nearly neutral. Some of the earliest theoretical studies suggested that the increased activity of Au was the result of the increased presence of coordinatively unsaturated sites such as step edges and corner sites for small Au clusters. CO, as well as oxygen, showed significantly enhanced adsorption energies at these sites, as depicted in Fig. 2.23. The increased binding energies at these sites were thought to enhance the rate of elementary steps involved in CO oxidation[59].

In subsequent studies, Liu et al.[66] calculated the activation barriers for CO oxidation

over Au(211) slabs supported on TiO2. The barriers for the oxidation of CO by atomic (O*) and molecular (O2*) oxygen were calculated to be 25 and 60 kJ/mol, respectively[66] .

The barrier required to dissociate O2 to form atomic oxygen, however, was found to be over 100 kJ/mol. The reaction path involving molecular O2, therefore, appears to be the favored route. On small metal particles, the increased adsorption energies at step edges increases the surface coverage of CO. O2, on the other hand, adsorbs at the positively charged Ti sites and results in a charge transfer from the supported Au into the antibonding 2π orbital of O2, thus activating O2 for reaction[66,67].

Theoretical studies of Au-supported clusters on MgO by Molina and Hammer[68] support this same view that the adsorption of molecular oxygen is stabilized at the interface between Au and the MgO support. Molecular oxygen is found to be the reactive oxygen intermediate. The active interface is shown in Fig. 2.24 which identifies di erent proposed CO binding configurations and reaction intermediates for CO oxidation on various shaped Au[68] clusters supported on the model MgO(100) surface.

More recent results by Remediakis et al.[62] identify two possible mechanisms for CO oxidation. The first suggests that the reaction takes place between CO on the metal O2

which is bound to the titanium support at the Au interface. This is similar to the studies by both Liu et al.[66,67] and Molina et al.[68,82] This mechanism strongly depends on the na-

ture of the support and its ability to stabilize O2 at the interface. The second mechanism, however, is one which is nearly independent of the support. This mechanism proceeds solely over Au. Low coordination sites are found to be important for activating CO and O2. The presence of these two routes agrees with the current experimental evidence which shows strong support dependence for some systems with little to no dependence on the

Figure 2.24. Proposed CO–O2 binding configurations on various shaped Au[68] clusters on Mg(100). CO adsorption at the interface between the support and cluster is disadvantageous owing to steric e ects as the CO is repulsed by the support. Oxygen is bound to both the cluster and the support, ultimately forming a CO–O–O complex at the cluster’s edge with the most favorable arrangement shown in (b). When the complex rearranges as depicted in (d), the barrier to reaction is lowered significantly and CO2 is readily formed. The second oxygen remaining on the gold particle is even more easily reacted with CO.

(e) shows an alternative arrangement of the reaction interface. However, in this case the adsorbates are not bound to low-coordinated gold as in (b) or (d) so the reaction is less favorable[68].

support for other reactions.

Gold catalysts dispersed on TiO2 also demonstrate unique catalytic reactivity for the epoxidation of propylene with hydrogen and oxygen. Hydrogen peroxide is proposed to be the active intermediate. It is also thought to be formed at the interface of Au and TiO2

in a mechanism very similar to that proposed for the reactive O2/CO complex involved in CO oxidation.[69].

2.3.4.10 Negatively Charged Au Clusters

Studies on the soft landing of Au clusters on an MgO support taken together with ab initio calculations suggest that anionic Au is the active surface species necessary for oxdizing CO. [73−75] The authors suggest the presence of F-center defects, formed as the result

of oxygen vacancies enable charge transfer from the support to the metal. Hakkinen et al.[73] et al., for example, concluded that the unique reactivity of an Au8 particle attached to the MgO(100) surface is the result of its adsorption to an MgO defect site. Due to the presence of an oxygen vacancy (F center) the cluster is charged via (partial) electron transfer from the oxidic support. Whereas low-temperature reaction channels show chemistry consistent with the Molina and Hammer results, Hakkinen et al. find that at high temperatures defect-rich MgO(100) shows strongly enhanced reactivity. CO2 is formed initially with CO adsorbed on the top facet of the Au8 cluster and the peroxo O2 molecule bonded to the periphery of the interfacial layer of the cluster. As we have learned in the section on cluster size e ects, Au8 is a non-reactive species due to electron shell closure. On the other hand, Au8− is a highly reactive intermediate because it will have a very low ionization potential. This is consistent with the need for electron donation

60 Chapter 2

to Au8 in order to produce its high activity.

More recent ab initio calculations carried out by Pacchioni et al.[83], however, indicate that the formation of the anionic Au particles may be due to electron transfer via tunneling from the underlying Mo substrate (used to grow the films) to the Au particles through the very thin MgO films used experimentally. This would suggest that defect sites may not be necessary, and that the application of these elegant surface results may not translate to the unique catalytic particles of supported particles. Recent DFT calculations by Molina and Hammer[84] show that the presence of F centers does not appear to lead to appreciable charge transfer to the metal particle. Their results do, however, show that artificially charging the tetrahedral Au20 cluster on the MgO support results in a significant improvement in the CO oxidation kinetics.

2.3.4.11 Positively Charged Au Clusters

Cationic gold species have also been speculated as well as detected in-situ for various different reactions. The unique reactivity of cationic Au clusters supported on ceria has been suggested to be important in catalyzing the water gas shift (WGS) reaction. Fu et al.[70] deposited small Au clusters on an La-doped reducible CeO2 substrate and demonstrated unique, highly reactive Au particles for WGS. The surface was subsequently washed with a basic NaCN solution in order to leach out metallic Au. Despite the removal of metallic Au, the WGS activity over these leached systems remained the same, suggesting that the active catalytic sites were comprised of cationic gold.

The reactivity of cationic gold particles has also been speculated for Au supported on Mg(OH)2 and Fe2O3 as well as for Au atoms in zeolites. Gusman and Gates[71] used X-ray

absorption spectroscopy to identify single Au3+ atoms as the active form of Au present in the zeolite and responsible for ethylene hydrogenation. As we will see in Chapter 3, metal cations with partially empty d-shells are active catalysts for insertion reactions; such a reaction is, for instance, the formation of adsorbed ethyl from ethylene and adsorbed hydrogen. The reaction between surface CO and surface hydroxyl intermediates

has also been suggested to occur over cationic Au thus leading to the formation of formate species[56] .

In a more recent study, Guzman et al.[81] used in-situ time-resolved XANES and Raman spectroscopy to probe the nature of the active oxidation state of the metal along with the active oxygen form for CO oxidation over Au nanoclusters supported on CeO2−x. Their results indicate that there is a Ce(4+)/Ce(3+) redox couple. In-situ Raman spectroscopy shows evidence for molecular oxygen bound as both an η1 superoxide as well as a peroxide surface species form at defect sites in the oxide. The η1 superoxide appears to be the reactive form. Au helps to promote O2 on the oxide support. XANES data suggest that cationic Au is responsible for carrying out the chemistry.

Others have speculated that cationic Au along with metallic Au must be present[72] . Hydroxylated Au+ species can form at the periphery of the Au/oxide interface and may be the active species.

We have seen that the chemistry of the catalytic system strongly depends on the Au particle size, the support used and the reaction studied. The relation between catalytic reactivity and the charge state of the reactive Au center may depend strongly on the size of the Au particle. A second important e ect is the interaction between the metal particle and the support. Here again, the metal particle size as well as the reducibility of the oxide can critically impact the catalytic behavior. Theory has o ered valuable insights into each of these proposed e ects by providing elementary reaction energetics along with detailed

features of the reaction mechanism.

2.3.5 Cooperativity

In heterogeneous catalysis, there are many examples where addition of a second component can change the overall catalytic reactivity in the system by changing its solid-state chemistry. An example of this includes the addition of Co2+ to the MoS2 and NiS2 systems discussed in Chapter 5. The mixed metal sulfides o er significantly increased activity due to changes in the chemical reactivity of the sulfide surface. We introduce here the solid-state chemistry of oxide catalysts (see also reference 3). A more detailed discussion on mixed metal oxides is presented in Chapter 5.

Figure 2.25a. Proposed propane ammoxidation mechanism[85] over Mo–V–Nb–Te–Ox catalysts.

The inorganic chemistry of a multi-component heterogeneous catalyst is often very complex, as it is quite di cult to obtain structural information at the molecular level to help establish the fundamental processes. As an example, we discuss the chemistry of the complex mixed metal oxide catalyst Mo7.5V1.5NbTeO29 shown in Fig. 2.25, which is known to catalyze the ammoxidation of propane to acrylonitrile. The active centers in this system are multifunctional metal oxide assemblies that are spatially isolated from one another owing to their unique crystal structures.

The Nb5+ centers are thought to stabilize the primary active Mo and V centers by structurally isolating them. This site isolation appears to be a prerequisite for selective oxidation. The presence of excess adsorbed oxygen would otherwise tend to lead to total combustion. The activation of propane is thought to proceed via the abstraction of the

methylene hydrogen by a V5+ center, which subsequently changes its oxidation state to V4+.

5+V=O ←→ 4+V−O•

62 Chapter 2

Figure 2.25b and c. (b) Catalytically active center of Mo7.5 V1.5 NbTe29 in [001] projection and schematic depiction of the active site. (c)2 x 2 unit cell structure model of Mo7.5 V1.5 NbTeO29 in [001] projection showing four isolated catalytically active sites[85].

Para n activation is thought to be free radical in character. H-atom abstraction by O

radical centers is a facile process. The adsorbed propyl radical can then lose a methyl-H to oxygen at an adjacent Te4+(6+) center, thus forming propylene. The catalyst is consid-

ered to be bifunctional in character whereby the ammoxidation of propylene subsequently occurs at the Mo6+ centers. The reduced Mo and V centers are regenerated by lattice oxygen originating at a reoxidation site that is physically separated from the active site. Dioxygen dissociates at the redox site and can then be incorporated as lattice oxygen. Hence the ammoxidation sites of the reaction are regenerated by oxygen atoms, formed at other sites on the surface via the dissociative adsorption of O2. The oxide medium

helps to facilitates the transport of this oxygen to the catalytically reactive and selective centers. This is known as the Mars–van Krevelen mechanism[86]. The well-defined

crystallographic structure of these catalysts, Fig 2.25(c), and in particular in the organic chemistry of their synthesis, allows them to be viewed as self-assembled multifunctional catalysts in which covalently bonded intermediates are exposed to topologically optimum atomic configurations for catalysis. The catalyst can, in some sense, be considered as a solid-phase analogue of the Wacker catalyst in its ability to simulate the catalytic cycle and regenerate the active redox site by the remote activation of oxygen.

In multifunctional catalysts, reactions occur in consecutive steps where each step can be catalyzed by a di erent active site. Typically many reactions are at equilibrium and one or two of the steps are rate-limiting. The selectivity then is controlled by the relative composition of the di erent reaction sites which catalyze specific steps. For example, oxidation catalysts are typically optimum when the V=O sites that catalyze the alkane to alkene reaction are present in a high enough concentration that an alkane/alkene equilibrium is reached in the particular oxidation system of interest. The subsequent catalytic oxidation steps which are involved in the functionalization of the alkene then become rate limiting.

2.3.6 Surface Moderation by Coadsorption of Organic Molecules

The coadsorption of surface moderating organic molecules can be used to induce significant steric control of the reaction product selectivity if the size and shape of coadsorbed molecules will form three-dimensional structures a the site of reaction on the 2D surface (such as shrubs or trees emerging from the bottom of a forest). These groups induce preference for reactant molecules to adsorb or form intermediates due to their optimal fit. Coadsorption of enantiomeric molecules have been used to “design” heterogeneous catalytic surfaces that show enantiomeric selectivity[87] . The best known is the e ect of the addition of the cinchonidine complex to Pt. The cinchona-modified Pt/Al2O3 catalysts have been designed with enantioselectivity higher than 90% for the hydrogenation of α-keto-esters. The higher selectivity observed here is thought to be executed by analogues of the large organometallic clusters immobilized on a transition metal surface. On model catalysts comprised of enantiomeric tartaric acid adsorbed on the Cu(110) surfaces, the surface overlayer is found to be enantiomerically active and creates chiral channels that expose bare metal atoms[88]. Vayner et al.[89] proposed a mechanism for the reduction of pyruvate that proceeds through an intermediate covalently bonded to the chinonidine (Scheme 2.3).

Scheme 2.3

This bond is subsequently broken through a hydrogen addition step. The stereochemical arrangement and interaction between the catalyst ligands and the reactant molecule can result in important selectivity preferences. The impact on selectivity is especially true for homogeneous catalysts whereby catatalyst design for homogeneous catalysts often involves screening a variety of di erent ligands. These same selectivity influences are also seen in enzymes. Both homogeneous, and enzyme catalysts typically involve single metal atom centers which impart stereochemistry by the topological positioning of the ligands and the reactant molecules about the active center.

Heterogeneous supported transition-metal catalysts, on the other hand, lack individual molecular centers and instead are comprised of surfaces with di erent metal ensembles. Surfaces that contain step edges can begin to exploit the stereochemistry of the step to induce topologically controlled stereoselectivity. For instance, enantioselective catalysis has been realized electrochemically by the use of oriented transition-metal catalysts with

logue of organometallic coordination complexes. Indeed, a wealth of knowledge on HDS has come out of well-defined studies using metal sulfide clusters and complexes [94−100].

Similarly, the surfaces of heterogeneous solid-state metal oxide catalysts can also be analogues of corresponding coordination complexes. These catalysts operate as sulfide or oxide phases since these are the stable phases under reaction conditions.

We have now seen that specific steric factors can control catalytic selectivity by guiding specific reactants to specific products. Most heterogeneous catalysts, with the exception of zeolites, can be expected to be intrinsically less selective, since they typically do not contain a three-dimensional architecture that can help to guide the formation of specific products. Therefore, they are optimum for those applications where thermodynamics prescribes the formation of one particular product over the others. The careful choice of transition metals and promoters, however, can appreciably alter chemical reactivity and bias specific reactions, thus altering the relative rates of competing elementary reaction steps.

2.3.7 Stereochemistry of Homogeneous Catalysts. Anti-Lock and Key Concept

The selectivity of organometallic complexes used in homogeneous catalysis can be significantly improved by changing the metal center, its oxidation state, or by manipulating the structure as well as the electronic properties of the ligands about the active metal center. The bulkiness of these ligands can even be tuned to help develop more highly selective enantiomeric catalysts. The Noyori hydrogenation catalyst provides a good example. The transition-metal cation in this catalyst is coordinated to the enantiomeric BINAP catalyst (Fig. 2.28).

Figure 2.28. The BINAP ligand[101].

Since the BINAP ligand is optically active, the catalytic reaction is enantiomerically selective. The rigidity of this phosphine ligand derives from the connectedness of the phosphine groups.

Stereochemical control is typically due to the interaction between reactants and the bulky ligands. Stereochemical control by enzymes can be comparable to the highly selective organometallic catalysts. Stereochemical control of a catalytic reaction is another example of the use of molecular recognition to control the relative adsorption strength of reaction intermediates. One of the greatest challenges in chemocatalysis is to design catalytic systems that combine di erent catalytic functions in a controlled fashion, so as to integrate di erent reactions into a single catalytic system.

A successful homogeneous system involves the catalytic hydroaminomethylation of internal alkenes to produce the linear products over an organometallic Rh complex which contains the rigid bulky diphosphorus ligand[102]. It helps to catalyze the combined sequence of bond isomerization, CO insertion and amination at the Rh center with stereochemical control. The unique feature of this catalyst is that all three of these di erent