Basis set convergence and performance of density functional theory including exact exchange contributions for geometries and harmonic frequencies

Abstract

The performance of the Becke three-parameter Lee-Yang-Parr (B3LYP) method for geometries and harmonic frequencies has been compared with other density functional methods and accurate coupled cluster calculations, and its basis set convergence investigated. In a basis of [3s2p1d] quality, B3LYP geometries are more accurate than CCSD(T) due to an error compensation. Using simple additivity corrections, B3LYP/[4s3p2d1f] calculations allow the prediction of geometries to within 0.002 Angstrom, on average. Except for certain special cases where frequencies are especially sensitive to the basis set, B3LYP/[4s3p2d1f] frequencies do not represent a clear improvement over B3LYP/[3s2p1d], while the latter are of nearly the same quality as CCSD(T)/[3s2p1d]. Applications to ethylene, benzene, furan and pyrrole are presented. For the latter three molecules, our best structures and harmonic frequencies are believed to be the most accurate computed values available.