How accurately can we calculate molecular CH and CC re distances by DFT methods? Dependence on basis sets and functionals, estimations of experimentally inaccessible re distances and distance-dependent scaling factors for approximations of triple-zeta quality

Abstract

Optimized bond lengths have been calculated by DFT methods using six functionals and 10 basis sets. None of the predicted values agreed completely with the experimental r e distances and they are in general too large. Available accurate experimental r e distances for CH and CC bonds of simple organic molecules have been linearly correlated with corresponding DFT optimized molecular structures. These regression equations can be used for prediction of accurate experimental r e equilibrium distances of CH and CC bonds of medium-sized to large molecules. The derived regression equations also enable the introduction of distance dependent scaling factors which allow very accurate predictions of higher order basis set geometries from lower order calculations. Analytical equations for the determination of these scaling factors have been derived. Thus, large savings of CPU time (dependent on the size of the system) are possible. Applications to both small hydrocarbons with up to six carbon atoms and larger molecules such as alanin dipeptide and vitamin A are presented. Likewise, more critical test cases such as singlet and triplet methylencarbene and strained cyclic hydrocarbons are considered.