Today the digital computer is a major tool of research in chemistry and the chemical sciences. However, although computers have been employed in chemical research since their very inception, it is only in the past ten or fifteen years that computational chemistry has emerged as a field of research in its own right. The computer has become an increasingly valuable source of chemical information, one which can complement and sometimes replace more traditional laboratory experiments. The computational approach to chemical problems can not only provide a route to information which is not available from laboratory experiments but can also afford additional insight into the problem being studied, and. as it is often more efficient than the alternative, the computational approach can be justified in terms of economics.
The applications of computers in chemistry are manifold. A broad overview of both the methods of computational chemistry and their applications in both the industrial research laboratory and the academic research environment is given in the book Chemistry by Computer (Plenum Press, 1986). Applications of the techniques of computational chemistry transcend the traditional divisions of chemistry - physical, inorganic and organic - and include many neighbouring areas in physics, biochemistry, and biology. Numerous applications have been reported in fields as diverse as solid-state physics and pesticide research, catalysis and pharmaceuticals, nuclear physics and forestry, interstellar chemistry and molecular biology, surface physics and molecular electronics. The range of applications continues to increase as research workers in chemistry and allied fields identify problems to which the methods of computational chemistry can be applied.
The techniques employed by the computational chemist depend on the size of the system being investigated, the property or range of properties which are of interest, and the accuracy to which these properties must be determined. The methods of computational chemistry range from quantum mechanical studies of the electronic structure of small molecules to the determination of bulk properties by means of Monte Carlo of molecular dynamics simulations, from the study of protein structures using the methods of molecular mechanics to the investigation of simple molecular collisions, from expert systems for the design of synthetic routes in organic chemistry to the use of computer graphics techniques to investigate interactions between biological molecules.
The computers employed in chemical calculations vary enormously, from the small microcomputers used for data analysis to the large state-of-the-art machines which are frequently necessary for contemporary ab initio calculations of molecular electronic structure. Increasingly large mainframe computers are departing from the traditional von Neumann architecture with its emphasis on serial computation and a similar change is already underway in smaller machines. With the advent of vector processing and parallel processing computers, the need to match an algorithm closely to the target machine has been recognised. Whereas different implementations of a given algorithm on traditional serial computers may lead to programs which differ in speed by a factor of about two, factors of twenty were not uncommon with the first vector processors, the CRAY 1 and CYBER 205. Larger factors can be expected as software adapted to the new generation of multitasking vector processor, the CRAY X-MP and Y-MP and ETA 10 series, becomes available.
A wide range of Chemical Applications Software is available for CRAY supercomputers. The following list is taken from the Directory of Applications Software for Cray Supercomputers and gives an idea of the variety of work being carried out internationally. the range of packages available and the problems which they can be used to study.
Programs marked * are fully supported and distributed by a third party, other programs are not fully supported. Full details are given in the Directory of Applications Software for Cray Supercomputers. Further notes in ARClight will expand on individual applications.
