Use of distributed Gaussian basis sets for calculating energy levels of weakly bound complexes

A basis set method for calculating the energy levels of weakly bound complexes is investigated. The basis is constructed from distributed Gaussian functions in both the bending and stretching coordinates. This produces a set of functions which are localized in the full internal space and allows the construction of nondirect product basis sets which model the known characteristics of the wave function. Low lying states which are restricted to a small portion of space may be described by functions placed just in the well region. Highly excited states usually occupy large regions of space but may still be efficiently described by modeling the regions of greatly differing wave function curvature. Application to the van der Waals complex Ar--CO2 shows that such basis sets give a reduction of about a third in basis set size when compared with a more conventional basis of Legendre functions for the monomer rotation and distributed Gaussians for the stretching motion. Such savings should increase for larger, more anisotropic systems