Molecular Heterogeneous Catalysis, Wiley (2006), 352729662X

46 Chapter 2

then subsequently undergoes a β-hydride elimination from the acetoxyethyl intermediate to form vinyl acetate. This mechanism has shown experimentally to be more favorable than the route through vinyl.

In addition to the topological surface-ensemble e ect, there is increasing awareness of the importance of the activation of molecules by surface defect sites such as kinks or steps compared with their reactivity over terrace sites. For example, studies on single crystal surfaces of Ni demonstrated that for the steam reforming reaction CH4 + H2O → CO + 3H2[36], CH4 activation is rate limiting. Model single crystal surface experiments have shown that the D2/CH4 exchange reaction tends to proceed at step sites where the activation barrier which is 82 kJ/mol, is significantly lower than the barrier of 101 kJ/mol which is found at the terrace sites. In addition, the C atoms that form prefer to bind at the surface edges where their recombination leads to deactivating graphene-carbon formation. Alloying Ni with Au tends selectively to poison the coordinatively unsaturated sites which occur at both kink and step sites. Indeed, the nickel catalyst is promoted by alloying it with Au, which results in a significant increase of the activation energy for methane activation (CH4 activation now only occurs at the terraces of Ni). In addition, there is an increase in the stability of the catalyst since coke formation is now suppressed.

The di erence in surface energies of the metal components that form the alloy helps to determine the level of enrichment with the low surface energy component at coordinatively unsaturated surface atom sites. The enrichment depends exponentially on the degree of coordinative unsaturation of the surface atoms. For the statistical mechanical ideal solution model of an alloy, one can derive the surface composition, xs, to be

where xs and xb are respectively surface and bulk concentration, m is the coordinative unsaturation of a surface atom and 11 and 22 are the bond energies between two atoms in the pure components[37] . The energy expression used to deduce Eq. (2.23) is derived from the elementary broken-bond model, that assumes the energy to be linear in the number of neighbor atoms.

Relation (2.23) implies that in alloys such as Ni with Cu, there is substantial enrichment at the surface where the low surface energy component, or a complex or a phase on a metal oxide, segregates to the surface. This enrichment preferably occurs at the more coordinatively unsaturated step edge and kink sites.

The degree of surface segregation can change significantly in the presence of a reactive environment. Adsorbates that bond strongly to the surface can readily lead to surface reconstruction as well as changes in the segregation behavior. This is the result of di erences between the bond strengths between the adsorbate and metal 1 and the adsorbate and metal 2. A more accurate analysis of the surface composition would require following the dynamics of surface segregation, and surface reactivity all at the same time.

The di erence in the dependence of the two reactions shown in Fig. 2.13 (of dehydrogenation and hydrogenolysis) on the degree of alloying may also be due to the di erences in the reactivity of steps and kinks versus terrace sites. The C–C bond cleavage reaction will occur preferentially at step edges (see Chapter 3, page 139). The very rapid decrease in hexane hydrogenolysis rate with Cu alloying is then explained by preferential poisoning of the step edges, which are only present in small amounts. This point of view is supported by experiments by Blakely and Somorjai[38] that show on Pt single crystal surfaces with varying step density structural independence for the hydrogenation reaction but structure

dependence for the hydrogenolysis reaction. As we will describe in the next section, as the particles become less than some critical size less than 1 nm, the chemistry may be controlled by unique electronic and structural properties of the particle that are dictated by metal-support interactions.

The highly reactive step edges also provide adsorption sites which have strong binding energies. Therefore, these sites are most likely to be readily poisoned during reaction. Experiments using radiochemical labels[39] have provided significant evidence for the formation of carbonaceous overlayers on transition-metal catalysts. It was found, for example, that the small Pt particles used in bifunctional zeolite catalysts to establish hexane-hexene equilibrium are only selective in this reaction after being deactivated with a carbonaceous residue. This carbonaceous residue is thought to poison step edge or kink type sites that are selective towards the hydrogenolysis reaction. The terrace atom sites, however, appear to be responsible for carrying out dehydrogenation/hydrogenation reactions.

The two points of view that we have outlined here so far are the Langmuirian uniform surface view[1] versus the Taylorian[2] specific reactive site view. Clearly, the demonstration of the unique properties of step and kink sites versus terrace sites supports the Taylorian view that catalysis occurs by uniquely active sites, that are sometimes only present in very small numbers.

2.3.4 Cluster Size E ects and Metal-Support Interaction

2.3.4.1 Metal-Support E ects and Promotion; Relation to Catalyst Synthesis

We continue with our investigation of the features of the extrinsic environment which influence the intrinsic kinetics of the active site. It is well established that the particle size and shape as well as the support on which they sit can significantly influence their performance for structure sensitive reactions. In order to provide a perspective of the physical chemistry of supported clusters, we first provide a brief overview here of the basic catalyst preparation methods. These methods ultimately dictate the surface composition of the support, and thus control the size, shape, morphology and adhesion of the metal particles that form on the support. Understanding the chemistry that occurs at the metalsupport interface is therefore important for improving the preparation of heterogeneous catalysts. The aim of catalyst synthesis is usually to produce a catalyst with a high dispersion of catalytically active surface components that therefore have a small particle size. In catalyst synthesis, the catalytically active precursor complexes are first dissolved in an aqueous solution and then contacted with the catalyst support. After impregnation, the catalyst is dried and subsequently activated. In the aqueous phase, the catalyst support is typically covered with terminal hydroxyl groups of varying basicity and acidity. Their relative concentration is determined by the colloid chemistry of the system. The pH of the solution with respect to the isoelectric point of the oxide plays a key role (see, for example, Farauto and Bartholomew[40] and van Santen et al.[41]). The choice of the catalyst precursor complex relates to whether basic hydroxyl or acidic groups cover the surface of the support. For example, as will be explained in Chapter 5, the commonly used alumina supports are dominated by basic hydroxyl species whereas silica supports contain only weakly acidic hydroxyl surface species. An ion exchange with surface hydroxyls creates surface complexes with single metal cations or small metal clusters, and hence provides a good strategy to synthesize catalysts that are highly dispersed. The catalyst precursor used depends on the nature of the surface hydroxyls that dominate on the support surface. For example, to prepare a highly dispersed Pt on alumina catalyst a

48 Chapter 2

negatively charged Pt complex is typically used such as PtCl4 2−. This complex can ion exchange with the basic hydroxyls of the alumina support. The initial complex on the catalyst support is considered to be a surface complex such as (AlOH)(PtCl4).

A very di erent system is used, however, in preparing a catalyst supported by silica, which contains only weakly acidic silanol groups. In preparing a Pt on silica catalyst, Pt(NH3 )4 2+ is typically used to carry out the ion-exchange reaction with silanol protons.

In subsequent catalyst activation steps, a complex set of reactions can take place, and depend on catalyst loading as well as the chemistry. For surface complexes that are not easily reduced, such as for Co2+ or Co3+ reacted with alumina or silica, the support-metal particle interface may exist as a Co-aluminate or -silicate layer. Under reducing conditions, this interface is covered with reduced Co. In addition, subsequent catalyst preparation steps carried out on a reactive support such as TiO2 can increase the interface with a reduced metal particle by partially covering the reduced metal particle with the oxide or by an increased wetting of the particle surface which will increase the interfacial area[42a,b].

The chemical reactivity of the catalyst support may make important contributions to the catalytic chemistry of the material. We noted earlier that the catalyst support contains acidic and basic hydroxyls. The chemical nature of these hydroxyls will be described in detail in Chapter 5. Whereas the number of basic hydroxyls dominates in alumina, the few highly acidic hydroxyl groups also present on the alumina surface can also dramatically a ect catalytic reactions. An example is the selective oxidation of ethylene catalyzed by silver supported by alumina. The epoxide, which is produced by the catalytic reaction of oxygen and ethylene over Ag, can be isomerized to acetaldehyde via the acidic protons present on the surface of the alumina support. The acetaldehyde can then be rapidly oxidized over Ag to CO2 and H2O. This total combustion reaction system is an example of bifunctional catalysis. This example provides an opportunity to describe the role of promoting compounds added in small amounts to a catalyst to enhance its selectivity or activity by altering the properties of the catalyst support. To suppress the total combustion reaction of ethylene, alkali metal ions such as Cs+ or K+ are typically added to the catalyst support. The alkali metal ions can exchange with the acidic support protons, thus suppressing the isomerization reaction of epoxide to acetaldehyde. This decreases the total combustion and improves the overall catalytic selectivity.

The chemistry at the interface of a transition metal and a reactive metal oxide, such as TiO2, can be quite di erent to that carried out over large metal particles alone. Reducible oxide supports such as TiO2 or V2O5 can help to promote the chemistry on the metal and behave quite di erently to than oxides such as alumina. The dissociation of CO, for example, is usually considerably enhanced at the interface of the metal and a reactive oxide, where it can dissociate leaving a carbon atom attached to the metal and the oxygen at a cation site such as Ti or V on the metal oxide support.

Consecutive reactions with hydrogen or CO lead to the removal of this oxygen as H2O or CO2, respectively. In reactions such as Fischer–Tropsch , where CO dissociation is rate-limiting, the addition of such promoters helps to enhance the activity and even the selectivity for chain growth reactions. The increase in selectivity is the result of increasing the concentration of reactive carbon atoms on the transition metal.[43a,b].

Another example of the influence of the metal/metal oxide interface on the chemistry of the metal refers to the promotional e ects that take place in the presence of alkali metal oxides on transition-metal catalyzed reactions. Alkali and alkaline earth metal oxides are known to promote the catalytic activation of di erent molecules such as CO (in Fischer– Tropsch) and N2 (in ammonia synthesis) over supported metal particles. Coadsorbed alkali

metal generates an electrostatic field that favors electron donation from the transition metal to the adsorbate. These e ects can be understood by generally lowering of the work function in the presence of the alkali and alkaline metal oxide promoters[43c,d,e].

This lowers the activation energy of the bond dissociation reaction. For more details on the chemistry of the promoter e ects we refer to Thomas and Thomas[3], Somorjai[44], and Diehl and McGrath [45].

To maximize the rate of a reaction, one needs the maximum exposure of metalor catalytically active atoms to the reactants. Hence there is a great desire to stabilize small particles on catalyst supports. In the next two subsections on transition metals we will provide a detailed description of changes in the chemical reactivity of transition metals when the particle size decreases. This provides a short background to aid in understanding the e ects of particle size on catalysis. In the next subsection we discuss cluster size dependence e ects and in the subsections that follow we will summarize the specific e ort on supported Au clusters.

2.3.4.2 Cluster Size Dependence

One can distinguish at least three di erent characteristic regions for transition metal particles[46] and their catalytic activity:

(a)Specific molecular structures with sizes that are less than 40 atoms. These are comprised predominantly of surface atoms. There are very few, if any, bulk atoms.

(b)The nanoparticle region falls in between that for the molecular structures and that for bulk particles. The relative energies of the particles are dominated by di erences in the surface energies. Structures di erent from the bulk may become stabilized. Surface energies are of the same order of magnitude as the di erences in the bulk energies.

(c)Crystallites of bulk lattice structures, with faceted morphologies.

We will discuss in detail cases (a) and (b). The shapes of crystallites in catagory (c) are controlled by bulk-metal energies and therefore do not require separate treatment.

(a) Molecular clusters

To introduce the subject, we examine a range of di erent Rhx clusters and their corresponding energies. Figure 2.17 depicts 21 highly symmetric cluster shapes, that small metal atoms can assume. DFT calculated formation energies for each structure, normalized per Rh atom, are also given in Fig. 2.17.

While the bulk formation energy of metallic Rh is –555 kJ/at, the most stable Rh13 cluster results in a corresponding value of only –299 kJ/at. The di erences may reflect the much lower average coordination number of the cluster atoms compared with the bulk. One also notes that a few selected clusters such as Rh3, D3h ; Rh4, Dh4; Rh4, D3h ; Rh9, Oh; Rh13, Ih have similar formation energies, whereas the atoms in these clusters have very di erent coordination numbers. The planar configurations appear to be the preferred clusters at least for the smaller sized clusters.

Quantum chemical bonding details determine the relative stability and structure of these clusters. Because of their lower stability, the small metal clusters can be expected to be generally highly active. The reactivities usually show a maximum at a particle size between three and seven metal atoms[48,49]. Three parameters appear to be important in controlling the reactivity of these clusters: the coordinative unsaturation of the surface atoms, the availability of enough cluster atoms to bind with an adsorbing atom or

50 Chapter 2

Figure 2.17. The topologies of 21 small spin-optimized Rhx rhodium clusters and their corresponding formation energies per atom in kJ/at[47].

molecule, and the ionization potential or electron a nity of the cluster. A low reaction barrier for dissociative adsorption of a gas-phase molecule requires electron donation from

the metal cluster into an antibonding orbital of the dissociating bond.

For the alkali metals[50] and noble metals (i.e. Cu, Ag and Au), the di erences in the relative stability and electronic structure of these clusters as a function of cluster size N can be understood by using a spherical free electron model. The clusters contain welldefined orbital structures analogous to those found for atoms and molecules rather than the band structure found for bulk metal systems. Electron-shell closure and maximum stability are predicted for N = 8, 20, 34, 40, 58, and 92 atom clusters.

An example of an experimentally measured and simulated sequence of measurements for Au clusters[51] is shown in Fig. 2.18.

It appears that on practical catalysts which contain reactive hydroxyl groups or coadsorbed water, small metal particles are highly reactive towards oxygen and, hence, are di cult to reduce. Temperatures for reduction of small metal oxides may di er from

Figure 2.18. Electron a nities of Au1−79. Predicted shell closings at 8, 20, 34, and 58 are observed[51]. The gradual increase in electron a nity, and also the decrease in ionization potential can be understood electrostatically. When a conductive sphere increases in size, the charge can be better accommodated because it can distribute over a larger surface. The low electron a nities are found at the shell closure values, because the empty orbitals have a high energy with respect to the occupied orbitals. The differences in energy between the oddand even-number clusters reflect di erences in energy between the corresponding HOMO (Highest Occupied Molecular Orbital) and LUMO (Lowest Unoccupied Molecular Orbital).

temperatures for reduction of the corresponding metals by several hundred of Kelvin. Also during reaction, the highly reactive small particles may form cluster compounds with unique reactivity determined by cluster-complex chemistry.

(b) The Intermediate Nanoparticle Region

This regime is of most interest to practical catalysis, since most of the supported metal particles used industrially fall in this size range. The nanoparticles that fall within this size range tend to form three basic types of structure: cuboctahedron (Fig. 2.19a); decahedron (Fig. 2.19b) and icosahedron FCC metals tend to take on the cuboctahedron cluster as it usually works out to be a minimum in energy following closely the structure of atoms in the infinite bulk structure. This structure is similar to the D3h Rh13 cluster in Fig. 2.19. The icosahedron structure can be recognized as Ih Rh13 and the D5d decahedron as Rh13, respectively. Crossover between the di erent structures occurs as a function of particle size. Such crossovers are due to an increase in the relative importance of the surface energies when the size of a particle decreases. The simple expression for the energy of a droplet-like cluster with one type of surface can be used to help understand the relative stability of di erent sizes. With symmetrical potentials between the atoms it can be written in the form (Fig. 2.19c).

NU = A + BN −1/3 + CN −2/3

where N refers to the number of atoms in the cluster, and U is the potential energy of

52 Chapter 2

Figure 2.19. Structures of 309 atom clusters: (a) fcc cuboctahedron; (b) decahedron; and (c) icosahedron.

the cluster. In this polynomial function A describes the volume (bulk) contribution, B the surface contribution and C the edge contribution to the energy.

In the structures shown in Fig.2.19, the bulk energies, A, are di erent. Hence, when crossover from one particle structure to another occurs, the bulk energy A also changes. Table 2.1 summarizes the calculated cluster sizes at which crossover takes place for various di erent metals [52]. The results were calculated for di erent metals using molecular mechanics simulations.

Table 2.1. Crossover numbers between the di erent morphologies for the structures in Fig. 2.19[52]

The equilibrium shape of a particle is determined by the Wul rule[53] , according to which the ratio of the surface areas are inverse proportional to their surface energy.

For a metal such as Co, there is a phase transition from the low-temperature bulk hcp structure to the high-temperature fcc structure. In moving to small Co particles, the fcc structure now becomes stabilized because of the low surface energies of the fcc faces as compared with those of the hcp particles. The crossover between hcp to fcc occurs when the particles become smaller than 175 nm.

An extensive number of studies have examined the equilibrium shapes of Au clusters in the range of 50–5000 atoms.[46]. A broad range of di erent structural forms have been analyzed. This includes truncated octahedra in addition to the structures given in Fig. 2.20. This figure summarizes the computed predictions.

The equilibrium shape of an fcc particle is the truncated octahedron with “magic” numbers of N = 38, 201, 585,· · ·. In Fig. 2.20 these sizes are recognizable as the points on a drawn line. Over a wide range of sizes the decahedral clusters with varying ratios of their surface edges are found to be most stable.

Figure 2.20. Energies of structurally optimized Aun (N ≤ 520) clusters plotted as (E − εB N )/N 2/3 vs N (on an N 1/3 scale), where εB = 3.93 eV is the cohesive energy of an atom in bulk Au. Various

structural motifs are denoted as Oh(- - - -), Ih (... ...), TO (– –), t–TO ( ), To+ (+), t–TO+ (|−−|), i-Dh (•), and m-Dh ( ), with the denoting m-Dh clusters in the enhanced stability region. The 75, 101, and 146 atom m-Dh clusters corresponding to particular stable structural sequences are denoted by .

The (m, n, p) indices of m-Dh are shown in the inset for a (5,5,2) cluster. Adapted from C.L. Cleveland et al.[46].

2.3.4.3 Gold Catalysis; an Example of Coordination, Particle Size and Support E ects

The importance of metal particle size and metal support e ects on catalytic reactivity is probably best illustrated by examining the current flurry of work in the literature on the catalysis of supported Au nanoparticles. Pioneering work by Haruta et al.[54] led to the first discovery of the unique catalytic performance of nanometer-sized Au particles on various di erent supports including α-Fe2O3, Co3O4 and NiO. It is now well established that finely dispersed Au nanoparticles are highly active for a range of di erent catalytic reactions including CO oxidation, H2 oxidation, water gas shift, hydrogenation of unsaturated hydrocarbons and alkene epoxidation[55−58]. The nature of the support in

these systems plays a very important role since bulk Au is quite noble and hence nonreactive. Since the first discovery by Haruta et al.[54] in 1989, the research e orts on Au

catalysis have increased exponentially.

Despite the substantial experimental and theoretical e orts, the nature of the active site and the features which control its reactivity are still intensely debated in the literature. There are three predominant explanations for the low-temperature activity of supported Au. The unique activity is attributed to: (1) Changes in the specific particle’s shape, atomic structure or size which ultimately controls the relative ratios of edge, corner and terrace sites[55,59−62]. The coordinatively unsaturated edge and corner sites are defect

Principles of Molecular Heterogeneous Catalysis 55

than the structure given in Fig. 2.21c. The latter corresponds to the equilibrium shape sketched in Fig. 2.21a.

A decrease in particle size increases the relative ratio of surface atoms and atoms which have lower coordination numbers. Therefore, the reactivity of these particles tends to increase. Interestingly, often the activity per surface atom also changes. For supported catalysts, the turnover number (TON) typically goes through a maximum[77]. Smaller particles lead to increased coordinatively unsaturated sites. The adsorption energies at these sites are typically the highest. Dissociative addition reactions, therefore preferentially occur on such sites also for reactions that are positioned to the left of the Sabatier maximum. This is consistent with an increase in TON. Second, the active sites are sometimes positioned at the interface of metal particle and catalyst support. With a decrease in particle size, this interface between the metal and support increases, which should also increase the overall activity.

Decreasing the particle size to very small atomic ensembles , however, can sometimes lead to reduced activity. This is related to two factors. Metal atoms in the cluster which have very low coordination numbers form very strong bonds with the adsorbate to compensate for the smaller number of metal–metal bonds. This increased binding energy between the metal and the adsorbate leads to higher reduction temperatures. This shifts the reactivity patterns. There is therefore a shift in the position along the Sabatier curve. Larger Au particles tend to lie closer to the left of the Sabatier maximum whereas these smaller particles tend to lie to the right of the maximum whereby the products or other intermediates begin to inhibit the surface reactivity. In addition to the enhanced metal-adsorbate bonding, the bonding between the metal and the oxide support becomes stronger, and as such, may begin to deactivate the metal clusters. For example, metal atoms bound to an oxide support tend to transfer electrons to the support and become oxidized. Bogicevic and Jennison [78] have shown that the nature of the metal-oxygen bond for single metal atom and small metal clusters at very low coverages on the oxide support is primarily ionic regardless of the metal chosen. For larger particles there is a trade-o between metal–metal and metal-oxygen bonding.

Results of DFT calculations predict that Pd, as well as Pt, on the ideal MgO(100) surface will tend to form clusters rather then isolated ions[79]. Metals that lie closer to the

right in the periodic table, such as Cu, form much weaker metal–metal bonds. Copper, silver and gold tend to prefer isolated ions which tend wet the surface. For a detailed review on metal-support interactions we refer to the review by Campbell[80].

Figure 2.22 illustrates the particle size e ects for Au particles of nanometer size supported by TiO2. The catalytic properties of Au are altered in a unique fashion. Au particles between 2 and 4 nm supported on titania show unique activity for the low temperature oxidation of CO, whereas large particles are non-reactive.

As the particles become smaller, the fraction of metal atoms in the cluster that reside at the surface increases. This increases the ratio of corner and edge atoms over terrace atoms. For a few di erent surfaces the adsorption energies of O and CO are computed and plotted as a function of coordination number of the metal surface-atoms. The lower the degree of coordination, the higher is the degree of coordinative unsaturation representative of edge atoms. The computed adsorption energies, given as a function of the Au surface-atom coordination number, are shown in Fig. 2.23[59]. The increased degree of coordinative unsaturation will act to increase the rate up to the point where the adsorbate binding is too strong (Sabatier maximum). The rate will then decrease owing to the inability to remove CO from the surface. This is analogous to conventional CO oxidation catalysts