Abstract
Proper description of noncovalent interactions requires, among other things, the use of diffuse atomic orbital (AO) basis sets. However, the presence of diffuse functions, especially in extended molecular systems, can lead to linear dependent AO basis sets. This in turn results, for example, in molecular orbital optimization problems or, when dependencies are removed in unpredictable and possibly geometry-dependent accuracy fluctuations. In this work, an alternative approach is proposed which suffers no linear dependence problems and delivers comparably accurate noncovalent interaction energies. An algorithm is proposed and implemented to construct a grid of off-center s-type Gaussian functions surrounding the molecule; substituting the presence of atom-centered diffuse basis functions. While the number of basis functions in the grid is comparable to the number of diffuse basis functions in aug-cc-pVXZ (for each cardinality number "X") basis sets for small molecular systems, the ratio becomes more favorable with increasing system size. The grid is constructed in a way that it is unique for a molecule (monomer) and, thus, independent of noncovalent complex/cluster geometry. The grid parameters, such as the density of grid points and s-function exponents, are obtained via optimization toward the S22 data set on the MP2 level. The quality, transferability, and versatility of the grid is tested on the S66 data set as well as on several cuts through the potential energy surface for noncovalent complexes, such as methyl-guanine···methyl-cytosine conversion from stacked to hydrogen-bonded structure.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have