A Density Functional with Spherical Atom Dispersion Terms

Abstract

A new hybrid density functional, APF, is introduced, which avoids the spurious long-range attractive or repulsive interactions that are found in most density functional theory (DFT) models. It therefore provides a sound baseline for the addition of an empirical dispersion correction term, which is developed from a spherical atom model (SAM). The APF-D empirical dispersion model contains nine adjustable parameters that were selected based on a very small training set (15 noble gas dimers and 4 small hydrocarbon dimers), along with two computed atomic properties (ionization potential and effective atomic polarizability) for each element. APF-D accurately describes a large portion of the potential energy surfaces of complexes of noble gas atoms with various diatomic molecules involving a wide range of elements and of dimers of small hydrocarbons, and it reproduces the relative conformational energies of organic molecules. The accuracy for these weak interactions is comparable to that of CCSD(T)/aug-cc-pVTZ calculations. The accuracy in predicting the geometry of hydrogen bond complexes is competitive with other models involving DFT and empirical dispersion.