Faraday Discuss. Chem. SOC., 1989, 87, 79-90 Computational-chemical Assessments of well characterised Uniform Catalysts Anthony K. Cheetham,*T Julian D. Gale,? Andreas K. Nowak, Brian K. Peterson,? Stephen D. Pickett and John M. Thomas Davy Faraday Research Laboratory, Royal Institution of Great Britain, 21, Albemarle Street, London W1 X 4BS The use of computer-simulation procedures to model the behaviour of zeolites and other well characterised, uniform catalysts is discussed. We describe the prediction of the location of solute molecules and the estimation of heats of adsorption by molecular mechanics (MM) procedures. I n addi- tion, the use of Monte Carlo (MC) and molecular dynamics (MD) techniques to study sorbates at high loadings is considered, together with the calculation of diffusion coefficients by MD.The assessment of the stability of known and hypothetical zeolites by lattice-simulation procedures is described and we explore the extension of the method to pillared clays. The possibility of modelling a catalytic reaction mechanism by quantum-mechanical pro- cedures is also examined. 1. Introduction There is a very large class of heterogeneous catalysts in which all, or almost all, of the atoms in the bulk of the solid participate directly, or are implicated indirectly, in the key catalytic processes of the overall reaction. Well known examples are zeolites, silico-aluminophosphates (SAPOs), as well as certain clays and pillared clays. In all these cases, the active sites are uniformly distributed in the bulk of the solid. Because of the microporosity of these structures, however, the active sites are only accessible to gaseous reactant species of the requisite size and shape. Uniform heterogeneous catalysts, primarily because their structures may be elucidated by the traditional tech- niques applicable to the study of bulk solids,'-5 are therefore very well characterised in comparison with heterogeneous catalysts consisting of supported or multiphasic solids6 In addition to being able to pinpoint the nature of the active sites in uniform heterogeneous catalysts, a wealth of experimental methods is also available for the determination of other key features of their catalytic performance.Thus the siting of reactants, products and poisons in actual or model zeolitic catalysts, as well as their binding energies and mobilities, emerge from the application of a wide variety of spectroscopic, diffraction-based and thermodynamic methods.Neutron and X-ray diffraction studies on powdered uniform catalysts are especially revealing, as are high- resolution, multinuclear solid-state n.m.r. methods, high-resolution electron microscopy, and 'inelastic' spectroscopies, be they Raman- or neutron-based. The class of uniform heterogeneous catalysts is so large that experimental techniques alone are unlikely to be able to cope with the almost endless number of variants (brought about, for example, by changes (subtle or marked) in Si/Al ratios or distributions, in framework compositions and structures, in the character and distribution of exchange- able cations etc.) of individual catalysts that are theoretically, or in practice, available for use. We turn under these circumstances to computational assessments, both as a t Also at Chemical Crystallography Laboratory, University of Oxford, 9 Parks Road, Oxford OX1 3PD.7980 Compu ta t ional-chemica 1 Assessment of Catalysts means of rationalising information already gained and of guiding our quest to retrieve that which is yet to be won. With the appropriate combination of simulations and quantum-mechanical calculations, insights into the mechanisms of heterogeneous cataly- sis that are elusive by direct, experimental methods may also be obtained. In this paper we focus on the following topics, all of which are amenable, to a greater or lesser degree, to computational assessments: (i) the preferred sites at which molecules (which might be reactants or products) are sorbed within a zeolite or clay catalyst, (ii) the energetics of sorption processes and the dynamics of diffusion, (iii) the stability and structure of existing and hypothetical (new) catalysts and (iv) catalytic reaction mechan- isms.Some recent reviews of related may be useful to the reader. 2. Location of Sorbed Molecules Perhaps the simplest objective that we can address is the prediction of the siting of an adsorbed molecule in a zeolite cavity at low temperatures. Under these circumstances, entropy effects can be ignored and we need only locate the global minimum for the interaction energy between the host and the guest. Several approximations are made: (i) the zeolite is assumed to be rigid and unperturbed by the presence of the solute molecule, (ii) the hydrocarbon is assumed to be rigid and present in very low concentra- tions, (iii) the interaction is described by a simple atom-atom potential, using, for example, the parameterisations (for different Si/Al ratios of the zeolite) of Kiselev' and (iv) the parameterisations are assumed to be transferable from one zeolite to another.We have used a potential of the form: +(tot)=C ( rA x-<+- c:qj) . ij Values of A and B were determined semiempirically' by fitting experimental data for the heat of adsorption of methane in zeolite-Y (one of the faujasite family) and the charges for the electrostatic term were obtained from molecular-orbital calculations on both the hydrocarbons and fragments of the zeolites.Whereas, at ambient temperatures, the molecules are likely to be distributed over several sites, their observed positions at low temperatures should provide a useful test of the computer simulations since they should correspond to the positions of +(min). There have been virtually no single-crystal diffraction studies of zeolite-guest complexes, but several recent studies by high-resolution powder neutron diffraction have been reported. These include the location of Xe in zeolite p,", CO in zeolite-A," benzene in zeolite-Y'* and pyridine in zeolite-L.13 In the last instance, the molecule occupies a single site in the main channel, at 4 K, coordinated to a potassium ion through the nitrogen of the pyridine ring.The molecule is thus able to form an acid-base complex with the cation whilst benefiting from a non-bonding interaction with the cavity wall. Evaluation of the global minimum predicts the location and orientation of the molecule within 0.2 A of the observed position, thus lending credence to the validity of both the experimental result and the simulation. The principal discrepancy is to be found in the N-K distance, which is estimated to be slightly longer than the observed value. This may stem from the neglect of any covalent contribution to the acid-base interaction. 3. The Energetics of Sorption Processes and the Dynamics of Diffusion 3.1. Calculation of Heats of Adsorption The evaluation of the heat of adsorption at, say, room temperature is a more complex problem since it requires us to evaluate eqn ( 1 ) for all orientations of the guest moleculeA.K . Cheetham et al. 81 Table 1. Internal energies of adsorption of hydrocarbons in zeolite-Y (kJrn01-I)’~ calcd exptl CH4 -13.3 -15.2 C2H6 -21.5 -23.3 C3H8 -30.1 -32.3 C4H 10 -35.2 -37.4 and at all positions in the cage. The following integrations, which describe the Boltzmann distribution of the molecule over the available energy levels, can then be performed: I , = \ exp [-$(tot)/RT] du (2) l2 = $(tot) exp [-$(tot)/RT] du. (3 1 V I, The internal energy of adsorption is then given by AU(ads) = $(tot) = ]?/Ii. (4) This treatment accounts for entropy effects that distribute the molecule over an increasing number of higher enthalpy sites as the temperature is raised.For example, though the minimum energy, 4(min), of methane in zeolite-Y is -23.0 kJ mol-’, the molecule would only be expected to occupy the position with this energy at very low temperatures. At room temperature, AU(ads) is only -13.3 kJ mol-I. In table 1 we compare the calculated and experimental adsorption energies for a series of hydrocarbons in zeolite-Y. As the molecules become larger, the energies increase. We emphasise that these values correspond to the energies of isolated molecules and are only valid at infinite dilution when intermolecular interactions can be ignored. Below, we discuss the extension of this treatment to cavities with multiple occupancy. Many of the interesting properties of zeolites are dependent upon cooperative effects between sorbate molecules.The internal energy of adsorption, for instance, will have components due both to the adsorbate/zeolite interactions and to the adsorbate/adsor- bate interactions. In such a case, eqn (2-4) must be evaluated over all configuration space and one must turn to more sophisticated approaches to determine the properties of interest. The techniques of Monte Carlo and molecular dynamics, as used in liquid- state theory, are also applicable here. MD also offers the opportunity to study dynamic processes such as diffusion, a topic that will be addressed in section 3.2. The first use of Monte Carlo simulations in the study of adsorption in zeolites appears to be that by Stroud et al? They studied methane in zeolite 5A and calculated thermodynamic properties such as the isosteric heat of adsorption and the heat capacity.They also calculated the adsorption isotherm, though by a tedious method using coupling parameters and simulations for systems with potentials other than the one of interest. Kretschmer and FiedlerI6 also performed some early Monte Carlo work, simulating alkanes in zeolites. However, their method was restricted to one molecule per cavity and therefore corresponds to the ideal-gas limit. They were particularly interested in the configurations of the sorbate molecules within the zeolite cavities. More recently, Yashonath et al.” found good agreement between MC simulations and experimental results for the heat of adsorption of methane at zero coverage in sodium zeolite-Y, and Smit and den OudenI8 have done zero-coverage MC studies of methane in the zeolites faujasite, mordenite and ZSM-5.Both studies used a potential of the type given in eqn ( 1 ) . A very interesting feature of the work of Smit and den82 Computational-chemical Assessment of Catalysts Ouden was that they varied the Si/AI ratio in mordenite and found a sharp change in the heat of adsorption at Si/Al =r 6.7. They were able to rationalise this result as the blocking of high-energy adsorption sites by sodium cations. This prediction has yet to be verified by experiment. Woods and Rowlinson’’ have recently performed a grand canonical Monte Carlo (GCMC) simulation (in which the chemical potential is held fixed, rather than the number of particles) for xenon and methane adsorbed in zeolites X and Y.They used crystallographically determined coordinates for the zeolites and compared adsorption isotherms and heats of adsorption with experimental data. These workers used fluid-fluid potential parameters available in the literature and fitted the fluid-zeolite potential parameters to zero-coverage data. Although they fitted the potentials at one temperature and used them at a higher temperature, they were able to reproduce all of the qualitative features of the heat of adsorption and adsorption isotherm throughout the range of zeolite cavity occupancies. The potential they used was a particularly simple one, including only the Lennard-Jones terms and a constant “background” correction term. Even with this simple potential, including no explicit polarization or electrostatic terms, many of the features of the experiments were reproduced semi-quantitatively. Soto and Myers?” had previously used GCMC to study hard-sphere and Lennard- Jones fluids in zeolite-13X.They were the first to use this method for zeolite adsorption and were able to demonstrate its usefulness. They did not achieve a level of agreement with experiment comparable with that obtained by Woods and Rowlinson, but instead chose to use simple potentials to study the qualitative effects of including various interactions. Among their conclusions was that the hard-sphere model (used in many earlier theories of adsorption) worked well for the heat of adsorption (because the energy is mostly determined by the fluid-zeolite contributions), but not for the adsorption isotherm (because the fluid-fluid interactions determine the chemical potential).3.2. Dynamics of Sorbate Molecules The MC methods mentioned above are very useful for probing the configuration space of a molecular system, but to obtain information about its time dependence MD must be used. M D has a long history in the study of bulk liquids and gases,” and the methods developed there should easily transfer to the study of adsorption in zeolites. In its simplest form it involves the solution of Newton’s equations of motion for N particles in a specified volume and with a specified total energy. It is very similar to MC except that the system evolves new configurations naturally in time. Quantities such as the diffusivity, which depend on the state of the system at more than one time, can also be calculated.The diffusivity is usually found from the Einstein diffusion equation, ( 5 ) where Ax is the displacement of a particle from its initial position and the brackets denote averages over numbers of particles or over separate experiments. I t will be especially enlightening to compare the values of D calculated from simulations with resu 1 t s from pu 1 s ed - fi e 1 d - gr a d i en t n . in. r . e x pe ri men t s . The only MD study to appear in the literature on diffusion in zeolites is that of Yashonath et al.” They studied methane in Na-Y and investigated the effect of tem- perature on the mobility of the sorbate. A complicated RMKZ3 potential was used to model the methane/methane interactions, and Lennard-Jones plus electrostatic terms to model the methane/zeolite interactions.Only one loading was studied ( 6 molecules per cage), but they studied several temperatures from 50 to 300 K. Yashonath et al. calculated cage- and site-residence times for the methane molecules at the various temperatures. As expected, the methane becomes much more mobile at higher temperatures, with a large drop in the residence times being obtained in the ( S.u2) = 6 DtA. K . Cheetham et al. 83 range 50-150 K. At all of the temperatures studied, the molecules remained close to the walls of the cages and hence the mode of transport is surface diffusion. The trajectory at 300 K was analysed with eqn (5) and D = 2.0 x lop8 m2 s-’ was found, which compares well with an experimental (n.m.r.) value of 1.5 x lo-’ m’ s-l.One feature of the Yashonath et al.’* work reminds us that simulation methods have their limits. At the lowest temperature studied (50 K), the cage residence time was of the same magnitude as the length of a simulation run (ca. 25 ps). When this is the case, the simulation run is not long enough to give a reliable estimate of the diffusivity or the residence time. This will be true whenever the time scales that govern the phenomena of interest are longer than the amount of time than can be afforded for a simulation. This is further emphasized by the fact that they found different results at 50 K, depending on whether the sample was heated from the minimum energy (OK) configuration or cooled from a higher-temperature configuration.If there are metastable states in the vicinity and the lifetime of these states is of the same order as the simulation time, the system can be trapped in them. This problem is particularly acute at low temperatures when mobilities are low. Some of the flexibility offered by the various MC methods can also be obtained by M D. Constant temperature, rather than constant energy, simulations can be performed via a variety of methods.” Also, the chemical potential can be determined (at least for systems with not too high a density) via the potential distribution theory of W i d ~ m . ’ ~ These conveniences, along with its dynamic nature, endow MD with the properties of a very useful tool. Future developments in this area will involve the extension of the treatment to flexible sorbates and non-rigid hosts, thus leading to the possibility of modelling shape-selectivity in microporous catalysts.4. The Stability and Structure of Catalysts 4.1. New Zeolitic Materials Computer-simulation techniques have not only been applied to the study of adsor- bate/adsorbent systems, but also to the investigation of adsorbent structures themselves. Catlow and co-workers have developed lattice simulation techniques that yield informa- tion on the stability of crystal Such techniques are essential in the search for new structures which could be of interest in catalysis, because they allow a comparison of relative stabilities of known and hypothetical structures. Several different approaches have been reported for predicting new structures, but we focus here on the method developed by Akporiaye and Price26 which was used in the investigation of possible zeolitic structures consisting of a mordenite/mazzite intergrowth.*’ The basic assumption is that zeolite structures can be represented as combinations of component layers.The layers are obtained from projections of observed zeolite structures, such as that of zeolite ECR-1A2’ shown in fig. 1, by allowing different topologies of the layer. A combination of these layers in three dimensions using mirror or translational symmetry operations results in a number of hypothetical structures. Fig.2 shows ten of the component layers with the ECR-1A projection, but different topologies. On the basis of the coordination sequence of the tetrahedral atoms we found layer C to result in the most appropriate structure for further investigation [shown in fig.3( a ) ] . The structure, named DF (Davy Faraday), can be described as an intergrowth of two separate structures, ( a ) strings of cancrinite cages showing 6-rings [fig. 3 ( 6 ) ] and ( 6 ) strings of connected 4-, 5-, and 6-rings [fig. 3( c ) ] . Lattice-simulation techniques were then applied to compare the stabilities of DF and ECR-1A at 0 K. The simulations combine two-body short- and long-range potentials and three-body terms in the evaluation of lattice energies and optimum atomic coordin- ates. This approach was applied successfully to various silicate systems. 29~30 Our results84 Compu ta tional-chem ica 1 Assessment of Catalysts MWd I Ma22 Mord I I I I 1 I I I Fig.1. Structure of ECR-1A projected along the a axis. Alternating component strips of mordenite and mazzite are indicated; large spheres denote oxygen atoms, small spheres are T atoms (Si and Al). I I Fig. 2. Ten possible structures with the ECR-1A projection, generated by the methodology of Akporiaye and Price.26 Arrowheads indicate T atoms lying above the plane of atoms with unmarked bonds.A. K . Cheetham et al. 85 Fig. 3. ( a ) Perspective view of the framework of DF, ( b ) the cancrinite cage and ( c ) the layer of 4-, 5- and 6-rings which is found in DF. predict a lower lattice energy for DF than for ECR-lA, the more stable variant of ECR-1, by ca. 0.16 eV per Si02 , thus suggesting the existence of the crystal structure on energetic grounds. Similar predictions of new silicate structures can be made from many different component layers based on more than 50 known zeolite structures.Their use in catalysis, ion-exchange and molecular sieving can also be predicted roughly by studying three- dimensional physical or computer-generated models, or by computer simulations of adsorption and diffusion. A residual difficulty, however, is that it is still not possible to design the synthesis of a new zeolite catalyst from basic principles. 4.2. Pillared Clays Following the successful applications of computer modelling to zeolite structures and their sorbate complexes, we have begun to explore the extension of such simulations to other aluminosilicate catalysts. The pillared clays represent an interesting and chal- lenging group of materials that in many respects are closely related to the zeolites.A typical clay is composed of negatively charged, magnesio-aluminosilicate layers, bound together by hydrated cationic species. Typical examples of naturally occuring clays are vermiculite: [( Mg2.36Fe0.48A10. 16)0ct(S~2.72~~ 1 .28)tet010(OH 121 -0'64[Mg0.32(H20) n 1 +Od4 and montmorillonite, with a somewhat lower layer charge: [ ( Mg0.33A1 1.67)oct( si4)teto 121 -0.33"a0.33(H20) n Several members of the clay family (notably montmorillonite, beidelite, hectorite, and vermiculite) exhibit powerful catalytic properties. Synthetic variants of these naturally occuring solids, such as fluorotetrasilicic mica, labelled FTSM, are also good 'clay' catalysts, and are readily converted to their pillared (more open, accessible) states by the insertion of Keggin ions (such as [Al,,O,(OH),,( H20)12]7+) into their interlamellar regions.So far as the uniform (acidic) catalytic properties of clays are concerned, the active centres are either the quasi-free protons generated by hydrolysis of such cations as A13+, introduced by cation exchange, or the weak, Brprnsted-acidic centres that are attached to the interlamellar macrocations of the Keggin type. A wide variety of organic reactions can be catalysed by modified or synthetic In several of the reactions86 Computational-chemical Assessment of Catalysts ? Fig. 4. The experimental location of the anilinium ion in the interlamellar region of vermiculite, showing hydrogen bonding inferred by the short N - 0 distances.catalysed by sheet silicates, it has been that layer charge, and the density of the charge, are of key importance. From the modelling point of view, several new challenges are presented by pillared clays. First, the pillars are charged and it is therefore essential that our description of the electrostatic interaction between pillar and layer is reliable. Secondly, unlike the zeolites, we can no longer assume that the host is unperturbed by the presence of the guest species; for example, we would certainly expect the interlamellar spacing to be altered. Thirdly, hydrogen bonding is likely to play a key role in the host-sorbate interactions. On the other hand, computer modelling has a great deal to offer in this area because there is frequently a dearth of reliable experimental data.For example, powder X-ray studies have been reported for a large number of organic intercalate^,^^ but only the 001 reflections have been obtained in most cases and, consequently, only one-dimensional projections of the structures are known. Additional structural informa- tion relies upon indirect methods such as m.a.s.n.m.r., infrared and Mossbauer spectros- copies. In the exceptional case of anilinium-pillared vermiculite, Slade and Stone36 have succeeded in obtaining single crystals from which a full, three-dimensional crystal structure, excluding hydrogen atoms, has been obtained by X-ray diffraction. Although later work3’ indicates the presence of a supercell, the salient features may be seen from the reduced unit cell.Anilinium ions are sandwiched between opposing silicate 6-rings, with the amine group hydrogen bonding to a triangle of oxygen atoms (fig.4). Two anilinium sites are observed, according to whether the molecule points up or down. We38 have simulated this system at the idealised 1 : 1 composition, using the programme THBREL, developed by Catlow and co-w~rkers’~ (see above). Two- and three-bodyA. K . Cheetham et al. 87 Table 2. Results of computer simulations of three structural models for the anilinium-vermiculite complex ~ model lattice energy/eV AE/kJ mol-' a / A b / A c/A PI" exptl" - - 5.330 9.268 14.892 97.02 1 -1184.621 0.0 5.412 9.373 14.97 1 96.86 2 -1184.181 42.4 5.414 9.385 14.923 97.40 - 1 184.360 25.2 5.414 9.384 18.358 1 1 1.24 3 The experimental data of Slade and Stone are shown for comparison. A E represents the difference in energy between the preferred model and the alternatives shown in fig.5. " Ref. (36). Fig. 5. Three possible configurations of the anilinium ion in vermiculite at the high-density limit, ( a ) model 1, ( b ) model 2, (c) model 3. potentials for the aluminosilicate framework are already well known,'9 while those derived from crystal structures of organic molecules4" have been used to complete the force field. The geometry of the anilinium ion was initially held rigid at that determined for anilinium hydrochloride, but the structure of the layers was allowed to relax.41 There has been much debate about the orientation of organic molecules in the interlamellar region of clays. For high-charge-density clays, such as vermiculite, the do,, spacing would be equally well satisfied by two molecules arranged perpendicular to the sheets with either an antiparallel (model 1 ) or parallel (model 2) orientation, or by the molecules lying flat and being stacked one upon the other (model 3, fig.5). The lattice energies and unit cell parameters for the energy-minimised structures correspond- ing to the three starting configurations are given in table 2, together with the experimental unit cell. The two models which contain the cation perpendicular to the layer yield unit cells that agree reasonably well with the experimental results, but the structure with the antiparallel arrangement of anilinium ions gives a markedly lower energy. In the minimisation starting from the third orientation, the anilinium ions partially reorient towards the most favourable packing arrangement before becoming trapped in a local minimum.In nearly all respects, our simulated structure (fig. 6) is in remarkable accord with the experimental findings of Slade and Stone. A further well documented property of clay intercalates is that as the layer charge is decreased, there is an increasing preference for aromatic cations or molecules to orientate themselves parallel to the layers. We have made a preliminary investigation of this phenomenon by reducing the layer charge to half the value used in the previous simulations and correspondingly lowering the number of anilinium ions per unit cell to one. Initial results indicate that this is reproduced by our model.88 Computational-chemical Assessment of Catalysts Fig.6. The calculated minimum-energy packing arrangement, viewed perpendicular to the silicate layers. Our preliminary work in this area has confirmed that computer simulation has the potential to be a powerful method in the structural characterisation of layered silicates and we are now planning to extend the work to a wide variety of pillared clays and other low-dimensional solids. 5. Catalytic Reaction Mechanisms By its very nature, it is difficult to study by direct experiment the critical act of catalytic conversion. It is one thing to derive, by spectroscopic or diffraction procedures, the nature of the bound reactant, but it is quite another to track, on the femtosecond timescale, the rupture and formation of bonds.(One notes, in passing, that only for simple photo-excited, gas-phase reactions has it proved possible, very recently,42 to do just this). In principle, however, a combination of ab initio quantum-mechanical calcula- tions and, say, MD simulations can cope with the rearrangements that are involved in a catalytic reaction within a zeolite, and the quantum-dynamics approach of Car and Parrinel10~~ appears to offer a strategy for future calculations of this type. In the following paragraph, we review the progress that has been achieved to date in this area. Vetrivel et aZ.44*45 have examined the preferred site of binding of methanol in a model ZSM-5 catalyst containing framework Al at the so-called T2 site (for which there is n.m.r. evidence46).After predicting the placement of the physisorbed molecule in the vicinity of the Bronsted-acid site by the energy-minimisation procedure, modified configurations were then explored by using ab initio SCF calculations on the resulting cluster (the GAMESS code developed by Guest and Kendrick4’ at the Cray X-MP at the Rutherford Appleton Laboratory was used for this task). The final configuration arrived at by Vetrivel et al.44 shows one of the methyl hydrogens of the methanol essentially dissociated and re-bound to a framework oxygen of the catalyst. For this dissociation, there appears to be essentially no activation energy. In essence, the mechanism arrived at in this quantum-mechanical fashion signifies that the methanol is activated at the Brgnsted-acid site to yield a CH20H species, leaving the site free for access by other reactant species.It is still not clear4’ precisely which intermediates are critically impli- cated in the catalytic conversion of methanol to gasoline (and especially in carbon-carbonA. K . Cheetham et al. 89 bond formation) so that this computationally derived mechanism cannot yet be tested. Experiments using 13C-enriched methanol, with cross-polarisation m.a.s.n.m.r., recently initiated by Anderson et ~ l . , ~ ~ may help to clarify the situation. 6. Conclusions The foregoing examples serve to illustrate that computer simulations have, in a relatively short time, contributed a great deal towards our understanding of well characterised catalysts, but it is important to stress the future opportunities that are now presenting themselves.We note that it has recently become feasible, on a routine basis, to obtain full, three-dimensional structures from new materials which, although eminently suitable for catalysis, cannot be obtained in single-crystal form. This has been achieved by powder diffraction methods, using, for example, a6 initio techniques, which have recently been applied both to zeolites’” and other inorganic materials. ”,” An alternative strategy was adopted for the zeolite catalyst, ZSM-23.5’ Its structure was solved by augmenting, computationally, the information derived from powder X-ray diffraction (which gave unit-cell dimensions and an indication of the space group), electron diffraction (which confirmed the space group), and from adsorption measurements (which yielded framework density).In effect, this procedure, which could be generalised, arrives at the atomic structure of a catalyst that is so microcrystalline that it could not be solved by a6 initio powder methods. It is significant, then, that armed with the appropriate structural details, computational methods are now capable of assessing the suitability of such new materials for catalysis. The thermodynamic and dynamical behaviour of reactant and product molecules can be interrogated, shape-selectivity can be examined, and we are on the verge of being able to treat the catalytic reactions themselves. A great deal remains to be done and problems have to be solved, especially in relation to force fields; for example, two recent simulations of the siting of p-xylene in silicalite yielded different results, apparently because different parameterisations of the (same) interatomic potential were u~ed.”~” Nevertheless, the successes to date speak for themselves, and the ever-increasing power of the computer augurs well for future progress in this important area.References 1. J . M. Thomas, Proc. 8th Intl. Congre. Catal., 1984, 1, 33 2 A. K. Cheetham, A. K. Nowak and P. W. Betteridge, Proc. Indian Acad. Sci. (Chem. S c i ) , 1986,96,411. 3 R. M. Barrer, in Zeolites and Clay Minerals (Academic Press, London, 1978). 4 W. J. Mortier and R. A. Schoonheydt, Progr. Solid State Chem., 1985, 16, 1. 5 I . E. Maxwell, Adv. Catal., 1982, 31, 1. 6 J . M. Thomas, Angew. Chem., Int. Ed., 1988, 27, 1673.7 S. Ramdas, J. M. Thomas, P. W. Betteridge, A. K. Cheetham and E. K. Davies, Angew. Chem., Int. Ed., 1984, 23, 671; also S. Ramdas and J . M. Thomas, Chem. Br., 1985, 21, 49. 8 A. K. Cheetham and B. K. Peterson, i n Computer-aided Molecular Design, ed. W. G. Richards (1989); also S. Ramdas, J. Computer-aided Mol. Design, 1988, 2, 137. 9 A. V. Kiselev, A. G. Bezus, A. A. Lopatkin and Pham Quang Du, J. Chem. Soc., Faraday Trans. 2, 1978, 74, 367. 10 P. A. Wright, J. M. Thomas, S. Ramdas and A. K. Cheetham, J. Chem. Soc., Chem. Commun., 1984, 1338. 11 J . M. Adams and D. A. Haselden, J. Solid Slate Chem., 1984, 55, 209. 12 A. N. Fitch, H. Jobic and A. Renouprez, J. Chem. Soc., Chem. Commun., 1985, 284. 13 P. A. Wright, J. M . Thomas, A. K. Cheetham and A.Nowak, Nature (London), 1985, 318, 61 1. 14 A. K. Nowak and A. K. Cheetham, in New Developments in Zeolite Science and Technology, Proc. 7th Int. Zeolite Conf., ed. M. Murakami, A. lijima and J. W. Ward ( Kodansha-Elsevier, Tokyo-Amsterdam 1986), p. 475. 15 H . J . F. Stroud, E. Richard, P. Limcharoen and N. G. Parsonage, J. Chem. Soc., Faraday Trans. 1, 1976, 72, 942. 16 R. G. Kretschmer and K. Fiedler, 2. Phys. Chem., 1977, 258, 1045. 17 S. Yashonath, J. M . Thomas, A. K. Nowak and A. K. Cheetham, Nature (London), 1988, 331, 601.90 Computational-chemical Assessment of Catalysts 18 B. Smit and C . J. J. den Ouden, J. Phys. Chem., 1988, submitted. 19 G . B. Woods and J. S. Rowlinson, J. Chem. Soc., Faraday Trans. 2, 1988, submitted. 20 J. L. Soto and A.L. Myers, Mol. Phys., 1981, 42, 971. 21 M. P. Allen and D. J. Tildesley, in Computer Sirnulation qf Liquids (Clarendon Press, Oxford, 1987). 22 S. Yashonath, P. Demontis and M. L. Klein, Chem. Phys. Lett., 1988, submitted. 23 A. R. Penner, N. Meinender and G. C . Tabisz, Mol. Phys., 1985, 54, 479. 24 B. Widom, J. Ph.vs. Chem., 1982, 86, 869. 25 R. A. Jackson and C. R. A. Catlow, Molecular Simulation, 1988, 1 , 207. 26 D. E. Akporiaye and G. D. Price, Zeolites, in press. 27 D. E. Akporiaye, S. D. Pickett, A. K. Nowak, J. M. Thomas and A. K. Cheetham, Catal. Lett., 1988, 28 M. E. Leonowicz and D. E. W. Vaughan, Nature (London), 1987, 329, 819. 29 S. C . Parker, C . R. A. Catlow and A. N. Cormack, Acta Crystallogr., Sect. B, 1984, 40, 200. 30 C. R. A. Catlow, J .M. Thomas, S. C. Parker and D. A. Jefferson, Nature (London), 1982, 295, 658. 31 J. M. Thomas, in Intercalation Chemistr),, ed. M. S. Whittingham and A. J. Jacobson (Academic Press, 32 J. P. Rupert, W. T. Grandquist, and T. J. Pinnavaia, in Chemistrj* of Clays and Clay Minerals, ed. 33 A. Wiess, Angen: Chem., Int. Ed., 1981, 20, 859. 34 J. H. Purnell, P. A. Diddams and J. M. Thomas, submitted. 35 R. Greene-Kelly, Trans. Faraday. Soc., 1955, 51, 412. 36 P. G. Slade and P. A. Stone, Claj*s Clay Mineral., 1985, 33, 200. 37 P. G. Slade, C. Dean, P. K. Schultz and P. G. Self, C1aj.s Cia), Mineral., 1987, 35, 177. 38 J. D. Gale, A. K. Cheetham, J. M. Thomas, R. A. Jackson and C. R. A. Catlow, to be published. 39 C. R. A. Catlow and A. N. Cormack, Int. Rev. Phys. Chem., 1987, 6 , 227. 40 T. Oie, G. M. Maggiora, R. E. Christoffersen and D. J. Duchamp, Int. J. Quantum Chem. Quantum B i d . Symp., 1981, 8, 1. 41 C. J. Brown, Acta Crystallogr., 1949, 2, 228. 42 M. J. Rosker, T. S. Rose and A. H. Zewail, C’hem. Phys. Lett., 1988, 146, 175. 43 R. C a r and M. Parrinello, Phys. Rev. Lett., 1985, 55, 2471. 44 R. Vetrivel, C. R. A. Catlow and E. A. Colbourn, Proc,. R. Soc. London, Ser. A , 1988, 417, 81. 45 R. Vetrivel, C . R. A. Catlow and E. A. Colbourn, J. Phys. Chem., submitted. 46 E. G. Derouane and R. A. Hubard, Chem. Phys. Lett., 1986, 132, 315. 47 M. F. Guest and J. Kendrick, CAMESS User Manual, Introductory Guide (University of Manchester Computer Centre, 1986). 48 S. D. Hellring, K. D. Schmith and C. D. Chang, J. Chem. Soc., Chem. Commun., 1987, 1320; see also J. K. A. Clarke, R. Darcey, B. F. Hegarty, E. O’Donogue, V. Amir-Ebrahimi and J. J . Rooney, J. Chem. Soc., Chem. Commun., 1986, 425. 1, 133. New York, 19821, p. 5 5 . A. C. D. Newman (Mineralogical Soc. Monograph no. 6, J. Wiley, Chichester, 1987), p. 317. 49 M. W. Anderson et al., in preparation. 50 P. A. Wright, J. M. Thomas, G. F. Millward, S. Ramdas and S. A. I . Barri, J. Chem. Soc., Chem. Commun., 1985, 11 17. 51 A. K. Cheetham, W. I . F. David, M. M. Eddy, R. J. B. Jakeman, M W. Johnson and C. C. Torardi, Nature (London), 1986, 320, 46. 52 J. P. Attfield, A. W. Sleight and A. K. Cheetham, Nature (London), 1986, 322, 620. 53 L. B. McCusker, J. Appl. Cr?*staIlogr., 1988, 21, 305. 54 S. D. Pickett, A. K. Nowak, A. K. Cheetham and J . M. Thomas, Molecular Simulation, 1989, 2, 353. 55 P. T. Reischman, K. D. Schmitt and D. H. Olson, J. Phys. Chem., 1988, 92, 5165. Paper 8/05061 I; Received 22nd December, 1988