Platinum Metals Rev., 1978, 22, (2), 47
Cluster Compounds of the Platinum Metals
Considerable Potential as Industrial Catalysts
In recent years metal cluster compounds with large numbers of metal atoms in the molecule have been prepared and characterised. The intermediate role of such clusters between monometal co-ordination compounds and metal surfaces is apparent, and is of great importance in studies of surface reactions for use in catalytic and other industrial processes. Investigation of the modification of the behaviour of carbon monoxide and simple organic ligands on co-ordination to such “metal fragments” is consequently one of the most rapidly expanding areas of chemical research.
In recent years a remarkable growth has occurred in the knowledge of cluster compounds which contain transition metal ions. The number of known compounds of this type has increased several fold and the information about simple substances such as the binary carbonyls of ruthenium and osmium, M3(CO)12 (M=Ru or Os), has expanded greatly. The reason for this growth is to be found largely in the awakened interest in metal-metal bonds initiated by Professors Lewis and Cotton and the late Sir Ronald Nyholm, and also the development of instrumental, spectroscopic methods of structure characterisation. Chief among these are the availability of high resolution mass spectrometers, more rapid methods of X-ray structure analysis, and the advent of nuclear magnetic resonance (NMR) instruments capable of examining nuclei other than 1H, 19F and 31P. An additional factor is the potential usefulness in industry of some of the compounds, in particular carbonyl clusters of rhodium which have proved to be effective catalysts for the production of ethylene glycol from carbon monoxide and hydrogen. Still a third stimulus is the expectation that the chemistry of small transition metal clusters should prove to be valuable in providing information about the reactions of small molecules, for example carbon monoxide, hydrogen, oxygen, olefins and acetylenes, with the surfaces of macroscopic crystals used in heterogeneous catalysis.
In this review we shall emphasise this latter point. For the most part we shall illustrate our general arguments by examples of our own work on the carbonyls of ruthenium and osmium. However, it must be appreciated that these arguments will also apply to clusters formed by elements throughout the transition-metal block. We will examine the chemistry of polynuclear clusters according to Scheme 1 and concentrate mainly on clusters containing carbonyl (CO) ligands.
Consider collections of metals Mm. These may be conveniently separated into the following three overlapping classes. Clusters, where m is between 3 and 50; microcrystallites, where the upper range of m reaches several hundreds; and macroscopic crystals, where m extends towards infinity.
At one time there was the general view that in terms of structure there was a steady progression from a single metal ion, where m=1, to the macroscopic crystal with m=infinity. The arrangement of the metal ions was essentially the same irrespective of whether there were few, say 13, or many metal ions present. Microcrystallites and clusters, in the absence of ligands, were merely fragments of the bulk metal structure. This idea has been questioned particularly by Burton who has argued—and his arguments are apparently well supported by experimental evidence—that microcrystallites do not necessarily have close-packed arrangements of metal atoms, that is corresponding to hexagonal close packing or cubic close packing, but may in some cases be based on pentagonal symmetries. Clearly then there could be a major discontinuity in the build-up of macroscopic crystals which, of course, possess close-packed arrangements of metal atoms.
Geometry apart, recent calculations have led to the conclusion that even for relatively small clusters, where m=8 or 12, there is a remarkable similarity of electronic structure for these and those of highly crystalline metals. In essence band theory may also apply to small aggregates of metals. However, caution must be exercised here particularly in view of the proposed geometrical changes, and much further work in this area is essential.
Irrespective of these two aspects it is clear that clusters, microcrystallites, and macroscopic crystals have in common the faces, for example, M3 (triangular), M4 (square planar) or M4 (butterfly), which they present on the outside surface. In Figure 1 we illustrate these faces as they are found in known transition metal cluster compounds.
It is conceivable therefore that by studying the mode of behaviour of small molecules with such cluster faces, information leading to a better understanding of surface reactivity may be obtained about:
The bonding modes adopted by these molecules with triangular (111), square planar (110) and similar metal ion arrangements, and their consequent chemistries.
The energies of interconversion of one bonding mode to another. For example, it is known that in metal carbonyl clusters the CO ligand may bond in at least four different ways, shown below.
Other bonding modes will almost certainly be discovered. These modes parallel those considered for the interaction of CO with metal surfaces and clearly some idea of the energy necessary to convert one mode to another is of importance.
Whether or not the metal cluster geometry is itself sensitive to the reacting species. For many years there has been the implicit assumption that the geometry of the cluster species is insensitive to reactions on its surface. This may not be true, however, even for some reactions with microcrystallites and macroscopic crystals.
The chemistry, catalytic and otherwise, of the metal cluster.
At this stage it is important to realise that in our discussions of cluster reactivity we have two extremes to consider. First, we can have a “naked” cluster. That is to say a cluster which has no ligands associated with it and sits in an inert matrix. This is in reality a very difficult situation to reach. Secondly we can have the more conventional cluster species MmLn, in which the metallic cluster is surrounded by donor ligands L. Clusters of this type can themselves be separated out into:
Low-valent clusters. In these the metal ions are assumed to be in the zerovalent, or less, oxidation state and are commonly found for L=CO, PR3 and so on.
High-valent clusters. In this class L is a ligand such as Cl− or O2−. The so-called “naked clusters” supported in O2− containing matrices will almost certainly be of this type.
Clusters of type (i) are in general terms the more easily studied. Certainly these are potential homogeneous catalysts and are clearly important in reactions involving, for example, the activation of CO (Fischer-Tropsch). For these clusters, which, for m=3 to 13, form cage like polyhedra (tetrahedra, octahedra, etc.), the effect of the peripheral ligands L is very important. They can exert an electronic influence. Consider the tetrahedral carbonyl Ir4(CO)12. It has been shown that on reaction with triphenyl-phosphine, substitution of up to three CO groups by Ph3P can occur. The remarkable feature of this reaction, however, is that the rates of substitution increase dramatically in the order 3rd>2nd>1st. Substitution has been shown to occur on separate metal ions and substitution on the first iridium atom labilises a CO group on the iridium atom adjacent to it. They can also exhibit steric constraints. Most simple carbonyl containing species undergo reaction with the iodide ion to produce a metal iodide species, for example:
The hexanuclear carbonyl, Os6(CO)18, is different. This carbonyl reacts with I− (and other anions) to produce the dianion [Os6(CO)18]2− in quantitative amounts and iodine is liberated. Here we believe that the eighteen carbonyl groups effectively shield the Os6 nucleus from attack. Iodide attack therefore occurs at the CO ligand leading to an electron transfer reaction through an incipient CO-bridge. This serves to illustrate a common problem found for carbonyl clusters namely that in many reactions the rate determining step is CO-dissociation, to expose the Mm cluster, and this is often sluggish.
Much structural information is now available for carbonyl compounds and a wide variation in metal polyhedra have been found. Some are shown in Scheme 2.
It has been known for some time that in solution at least, some carbonyl clusters undergo a structural change. Notable in this respect is Fe3(CO)12 which is known to possess two carbonyl bridges in the solid (Scheme 3) but there is little evidence to support their presence in solution. This has been interpreted in terms of fluxional behaviour. In solution, where the CO ligands are removed from the constraints of the lattice, migration of the CO ligands about the Fe3 triangle occurs generating species with and without CO-bridges. The energy of activation for this process has been estimated to be >5 kcal/mol and the non-bridged species are considered to dominate in solution.
For some carbonyls it has been found possible to deduce ground-state geometries in solution from 13C NMR studies of isotopically 13CO enriched species. For these compounds, which may also exhibit fluxional behaviour, activation energies for CO migration must be in excess of 5 kcal/mol. We will consider the following three examples, each of which illustrates a different aspect of CO-fluxionality.
Rh4(CO)12. The 13C NMR spectra for this species (70 per cent 13CO enriched) over a range of temperatures are shown in Figure 2. At low temperatures the spectrum is totally compatible with the observed solid state structure exhibiting three doublets in the intensity ratio 3 : 3 : 3 and one triplet of intensity 3. The appearance of the triplet clearly identifies the CO group as a bridge spanning two rhodium nuclei (103Rh 100 per cent, ). As the temperature is raised, all signals collapse uniformly until at 40°C only a well resolved quintet is observed indicating that all CO groups are seeing four equivalent rhodium nuclei, in other words the twelve CO ligands are rapidly migrating around the Rh4 cluster. An activation energy of ∼12 kcal/mol has been calculated for this process.
Co3Rh(CO)12. In the solid the complex has a structure related to that of Rh4(CO)12. The 13C NMR spectrum at low temperature is consistent with this geometry, however, in contrast to Rh4(CO)12, as the temperature is raised two types of fluxional behaviour become apparent. At the intermediate temperature of −30°C all but two CO groups, namely those bound terminally to the rhodium atom (Figure 3), undergo interchange. This has the effect of equilibrating the two CO groups. At high temperatures all CO groups become equivalent and are clearly migrating over the whole Co3Rh cluster. The intermediate behaviour is of interest because it shows that CO-migration preferentially follows a course which enables bridges to be in association with the rhodium atoms at all times.
Os6(CO)18. The remarkable feature of this molecule is the presence of three different types of osmium atoms, Figure 4. Osmium(1) is six-co-ordinate being adjacent to three osmium atoms and three CO groups. Osmium(2) is seven-co-ordinate and osmium (3) eight. This is clearly apparent from the 13C NMR. At low temperatures the spectrum is as expected for this geometry (Figure 4) with three sets of signals [A+A′], [B+B′] and [C+C′]. Each set consists of two signals in the ratio of 2 : 1 corresponding to the two CO environments about each osmium atom. As the temperature is raised, these latter signals collapse giving rise eventually to three single signals. In this case equilibration occurs about each independent osmium atom and there is no evidence for CO transfer from one osmium to another in the temperature range examined. The effect of co-ordination number and electronic differences from one osmium to another is very dramatic.
These three examples serve to show that ground state geometries may be accessible for carbonyl clusters in solution, and that a variety of fluxional processes are available.
The triangular clusters of ruthenium and osmium have been the most widely studied. This is largely due to their availability. Consider the reaction established for M3(CO)12, Scheme 4, with H2, C2H4 and C2H2.
On reaction of M3(CO)12 ethylene undergoes C-H bond cleavage to produce M3-hydrido species containing either a C : CH2 or a CH : CH unit. In an alternative reaction sequence Os3(CO)12 reacts first with H2 to produce H2Os3(CO)10 which will then react with C2H2 to give first HOs3(CO)10(CH: CH2) which on heating generates one of the products of the reaction of C2H4 with Os3(CO)12.
Hydrogenation of H2Os3(CO)9(C : CH2) occurs not only on the C : CH2 fragment but also on the Os3 triangle to produce a species containing a CMe fragment bound symmetrically to the metal triangle. In contrast, reduction of H2Os3(CO)9 (CH : CH) liberates olefin and forms the tetranuclear cluster H4Os4(CO)12. The bonding modes of the various C2 fragments are shown in Figure 5.
These observations are important not only for the intrinsic interest in cluster chemistry but also because they provide some insight into the bonding modes that olefins might adopt on contact with metal, or related, surfaces. They certainly provide evidence of bonding modes not previously contemplated by the chemist interested in heterogeneous catalysis. It is also interesting to note that on reaction with Os3(CO)12, ethylene undergoes C-H bond cleavage whereas if H2 is present hydrogenation to C2H6 occurs.
Similar behaviour has been noted for a wide range of substituted olefins and acetylenes, and also a variety of cyclic systems.
A tetranuclear cluster might intuitively be expected to behave similarly since it merely consists of four triangular faces. Alternatively, it could be considered as an M3 triangle in association with an additional metal:
Clearly this additional metal ion will exert some influence on the metal triangle. Some reactions of H4Ru4(CO)12 and H4Os4(CO)12, two readily available tetrahedral clusters, are shown in Scheme 5.
There are two observations of note. First H4Os4(CO)12 undergoes reactions which differ from those of H4Ru4(CO)12. With C2H4 the osmium cluster retains its tetrahedral geometry whereas the ruthenium compound undergoes cluster fragmentation to yield among other products, H3Ru3(CO)9CMe. With cyclic olefins the ruthenium cluster again undergoes structural rearrangement to give clusters based on the butterfly arrangement of four metal atoms. This is simply derived from the tetrahedron by a bond-break mechanism:
The osmium compound again retains its tetrahedral configuration and in general terms gives products similar to those derived from linear olefins. Interestingly the same product is obtained irrespective of whether the reacting substance is an olefin or acetylene. In contrast to the M3 system, on heating the vinylic derivations HM4(CO)11(CH : CH2), C-H bond cleavage to give the 1,2 disubstituted olefin derivative only and none of the 1,1 disubstituted complex. Finally, and again in contrast to the behaviour of the M3 system, H3Os4(CO)11(CH : CH2) undergoes acetylene insertion into the Os-C bond to generate linear polymers. Clearly the effect of the additional M is to modify the reactivity of the M3 triangle.
Geometrical Changes in Clusters
There are available a number of theories which permit the rationalisation of cluster geometrics. None are totally successful but the Effective Atomic Number (E.A.N.) rule and the approach put forward by Wade have proved to be especially useful. The E.A.N. rule works very well for clusters with five metals or less, while Wade Theory may be more generally applied. Wade Theory very satisfactorily accounts for the bicapped tetrahedral geometry adopted by Os6(CO)18 but even more successful was the prediction that on reduction to [Os6(CO)18]2− the geometry should change to regular octahedron. In essence Wade Theory says that cluster shape is a function of the number of electron pairs available for cluster bonding irrespective of the source of these electrons, that is charge or ligand.
Recently we were able to produce a series of new binary carbonyls of osmium by the pyrolysis of Os3(CO)12 in a sealed tube. The structures of all these carbonyls, which are based on Os5, Os6, Os7 and Os8 units, have been established. These in turn undergo reaction with base (OH−) to produce anionic species, for example [Os5(CO)15]2−, which react with H+ to produce hydrido-clusters. These reactions are summarised in Scheme 6.
Consider Os5(CO)16, which has been shown by single crystal X-ray analysis to possess a trigonal bipyramidal arrangement of five osmium atoms. Treatment with base produces [Os5(CO)15]2−, which also has a trigonal bipyramidal geometry in keeping with Wade Theory, and reacts with acid (H+) to form first HOs5(CO)15− and then [H2Os3(CO)15]. The structure of the latter species, established by X-ray analysis, is an edge-bridged tetrahedron. Since the three species Os5(CO)16, [Os5(CO)15]2− and H2Os5(CO)15 have the same number of electrons, similar geometries might have been expected. The observed change clearly emphasises the effect of the incoming ligands, in this case H2. Similar behaviour is found for the Os6 series. Thus [Os6(CO)18]2− has a regular octahedral geometry whereas H2Os6(CO)18 has a monocapped square base pyramidal arrangement. Such changes cannot possibly be restricted to hydrido-species and other ligands will play a similar role. The implications of these results are important. They lead us to suggest that in the course of chemical reaction it is unrealistic to expect the cluster geometry to remain intact. It could be argued that similar changes might occur in surface reactions and there is growing evidence to support this.
In this review we have attempted to point out the various points of interest in cluster chemistry. Two points of significance have emerged from our more recent studies. First the bonding modes adopted by small molecules on contact with clusters are far more numerous than previously supposed from a simple comparison with monometal systems. Secondly, there is growing evidence that clusters may undergo facile structure exchange on contact with substrate.