Basis Sets for Molecular Interactions. 1. Construction and Tests on (HF)2 and (H2O)2

TLDR: In this paper, simple modifications of basis sets of only moderate size (e.g., 6−31G) can accomplish the same end at much reduced computational expense, including reoptimization of the orbital exponents within the framework of the relevant molecules, plus addition of a single diffuse shell of sp orbitals on nonhydrogen centers.

Abstract: Basis set superposition error (BSSE) remains one of the major difficulties besetting current ab initio calculations of molecular interactions. Despite the widespread notion that lowering of the BSSE to negligible magnitude requires extremely large basis sets, we show that simple modifications of basis sets of only moderate size (e.g., 6‐31G**) can accomplish the same end at much reduced computational expense. These modifications include reoptimization of the orbital exponents within the framework of the relevant molecules, plus addition of a single diffuse shell of sp orbitals on nonhydrogen centers. Subsequent addition of a second set of d‐functions further lowers the SCF BSSE, bringing it below 0.1 kcal/mol for both (HF)2 and (H2O)2. It is notable that addition of the latter d‐functions without prior reoptimization of the valence orbitals produces the opposite effect of an increase in the BSSE. Although the MP2 BSSE is also substantially decreased by the above modifications, it appears difficult to reduce this quantity below about 0.4 kcal/mol.