Large scale scientific computation via minicomputer

UWROtWCTION Theoretical chemists have traditionally been among the heaviest users of electronic computers. The earliest (Parr & Crawford, 1952; Boys et al., 1956). and still the most extensive (Schaefer, 1972), use of computers by theoretical chemists was in the calculation of electronic wave functions for atoms and molecules. A somewhat more recently developed (Wall et ol., 1958), but equally important, area of research is the study of the dynamics of chemical reactions using classical trajectories (Bunker, 1971), semi-classical methods (Miller, 1971), or even fully quantum mechanical approaches (Light, 1971). The relation between the two fields is clear; while the potential energy surface is frequently the goal of the electronic wave function computations, the same surface is the most important prerequisite to the study of the reaction dynamics. The above discussion is particularly pertinent to the present time, since it is only during the past year that the first computations of the type we refer to (Bender et al., 1972; George & Miller, 1972; Doll et al., 1973) have been carried out. It can be argued that the above areas are part of a new branch of chemistry, computational chemistry. In any event, the emerging importance of computation in theoretical chemistry will demand the availability of large amounts of relatively inexpensive machine time during the next decade. It has customarily been assumed that such large scale scientific computations require the use of the largest and fastest computers available. In the present paper we raise some questions concerning the economic feasibility of using machines such as the CDC 8600 for theoretical chemistry. We propose an experiment to determine the feasibility of an alternative route for large scale scientific computation-the use of minicomputers.