AUTOMATED SELECTION OF OPTIMAL GAUSSIAN FITS TO ARBITRARY FUNCTIONS IN ELECTRONIC-STRUCTURE THEORY

We present a method of fitting arbitrary functions to linear combinati ons of Gaussians. In particular, we discuss an adaptation of Prony's m ethod, or separation of exponentials, which allows us to automatically select appropriate exponents for these Gaussians. We then apply this technique to the selection of dealiasing sets for pseudospectral elect ron correlation methods. We show that it can successfully choose funct ions that generally improve the accuracy of pseudospectral correlation energies while reducing the size of the dealiasing set chosen. (C) 19 98 John Wiley & Sons, Inc.