Crystal-clear simulation

Our ability to model and explore complex three dimensional structures, from molecules to galaxies, has been revolutionised by new computer technology.

This allows us to construct, display and manipulate models at a previously impossible level of detail.

This article will emphasise the role of the computer in constructing models at the atomic level of processes and structures.

The role of computers in contemporary science is profound. Their constantly expanding processing power and memory allows us to translate fundamental knowledge into models of reality.

The computer explores, develops and displays a model specified by the scientist, a model that obeys the basic laws of science. Crucially, the model then confronts reality; and is refined and improved. A deeper and more predictive understanding of nature is developed.

This procedure is illustrated by the case of "atomistic" modelling, constructing models for the complex reality of matter at the atomic level. We understand the fundamentals which control the behaviour of atoms and molecules. The distribution and energy levels of electrons within molecules and solids can be calculated using the celebrated "Schrodinger" equation originally formulated in the 1920s and which can be solved for molecules and solids of increasing complexity using computational methods. For highly complex assemblies of atoms a simpler approach is possible. We can feed into the computer data which expresses how the energies of an assembly of atoms varies with the arrangement of the atoms. These interatomic potentials can be derived from both theoretical and experimental sources. Armed with this information the computer can predict the structure and dynamics of matter at the atomic level.

There have been exciting developments in the modelling of crystalline solids where the arrangement of the atoms is regular and repeating; one ambition in the modelling of solids is to be able to predict the structure at the atomic level of a crystal from knowledge only of its constituents. This field is progressing very rapidly. Tim Bush at the Royal Institution and Clive Freeman at BIOSYM have developed computer generated models for complex oxides such as lanthanum nickelate whose structure is related to the high temperature superconductor materials, the discovery of which eight years ago caused such a sensation in solid state science. Their model was developed by making no assumptions about the arrangement of the atoms in this crystal, and is a faithful representation of the real structure. In collaboration with Peter Battle in the Inorganic Chemistry Laboratory, Oxford, Bush has solved the structure of a complex oxide (lithium ruthenate), whose structure eluded experimental definition.

Models of glassy materials have been evolved rapidly, for example by Behnam Vessal and co-workers on silicate systems. They used a computer to simulate the melting of a special silicate, rubidium silicate, and then simulated a rapid quench -- the real way we make glasses. The resulting beautiful model for the disordered structure shows fascinating features -- the rubidium ions group together into channels -- a process which had been proposed previously by Neville Greaves and now emerges naturally from the computer models.

In atomistic modelling of biological structures there has been equally impressive progress. The first example is a recent simulation, by Tim Forrester, Bill Smith and Julian Clarke on the CRAY T3D at Edinburgh, of a cell membrane. It is vital to be able to move ions across the cell membrane.

In addition, simulation work by Julia Goodfellow at Birkbeck College on DNA shows how it flexes and bends -- a vital process by which literally metres of the molecules are packed into microns in chromosomes.

This type of detailed understanding of biochemical processes at the atomic level is of enormous value in the design and optimisation of pharmaceuticals which mimic the docking process. Of course, what determines the behaviour of molecules and materials in many circumstances is not what happens in their interior, but what happens on their surface. Chemical reactions often take place on surfaces -- and surfaces can promote chemical reactions, as in heterogeneous catalysis. Friction, the overcoming of which consumes huge amounts of energy, takes place between surfaces sliding over each other. The growth of crystals takes place at surfaces; and the surface must be modified to prevent it.

Both crystal growth and catalysis depend on molecules or atoms docking on to surfaces. Simulation of these crucial processes has been carried out by Andrew Rohl at the Royal Institution. He has done detailed simulations on barium sulphate, which is highly insoluble and so precipitates, for example, in oil pipes; a major problem in the oil industry. But this precipitation can be prevented by adding inhibitors. Recent work at the RI in collaboration with ICI has shown how these inhibitors work -- how they "dock in" on the surface. Once they have docked at the surface, they block further crystal growth. The challenge is to design new materials -- new inhibitors -- which block more effectively; with computational techniques this is a real possibility.

To promote the desired chemical reaction -- to make it selective and specific for particular molecules -- the surface must be engineered at the atomic level. One example is the Ziegler Natta catalysts which convert the gas ethylene into the polymer, polyethylene. Here, methylated titanium chloride molecules are deposited onto a magnesium chloride surface. Computer simulations at the RI by J. S. Lin have shown how the Ti/Cl molecules bind to the surface, creating sites for coordination of the ethylene molecules.

One of the most ambitious examples of computational surface science is the work of Gillan, Payne and co-workers who looked at how the chlorine molecule dissociates on the surface of silicon. By using parallel computers they followed the process in enormous detail and watched how the electrons redistributed.

Computers can also simulate the behaviour of crystals which surface are all zeolites and other microporous solids.

Because of the porous structures, molecules can diffuse into zeolites and they are sorbed and diffuse at different rates, so these solids are widely used by industry as separators. They also catalyse chemical reactions of the sorbed molecules; the nature of these reactions are controlled by the structure of the surrounding crystal -- the phenomenon of shape selective catalysis.

The really exciting problems in zeolite science all relate to the behaviour of molecules inside these porous architectures. Computers can show how molecules diffuse inside the pores of these materials -- a crucial process in both gas separation and catalysis.

Zeolites are synthesised from gels to which are added "template" molecules -- organic bases which direct the synthesis towards specific architectures. Recent work at the RI in collaboration with BIOSYM Technologies in the United States has shown how computers can pick the best template for a particular zeolite, by docking different templates into a given structure and finding out which has the lower energy which fits the best. This offers the opportunity to design molecules to provide specific new structures for use in catalysis and gas separation.

Docking of molecules is central to key biological processes. Computational work played an important role in work on the enzyme methane monoxygenase, which can convert methane into methanol. Being gaseous it is difficult to transport; but if converted to methanol which is a liquid, transport is more straightforward.

These examples show that computers, like the Cray T3D at Edinburgh are essential in contemporary physical and biological sciences. Their role will enormously enhance one of the older of scientific activities -- the building of models to represent reality.

Richard Catlow is the Wolfson professor of natural philosophy at The Royal Institution of Great Britain.