Monte Carlo Configuration Generation: A Parallel Approach for CI Calculations

J. C. Greer ( ), NMRC, University College, Lee Maltings, Prospect Row, Cork, Ireland.

A discussion of the Monte Carlo configuration interaction and its performance on parallel computers is given. The method relies on a random generation of configuration state functions, and an iterative procedure for accepting/rejecting configurations.

The Monte Carlo generation of configurations makes the method well-suited for parallel computing. The calculations include configurations without bias for excitation level, thereby allowing accurate treatment of dissociation energies and electronic excitations. Results are presented for small molecular systems and performance of the algorithm on a Beowulf cluster is evaluated.