Abstract:

Austin-Petersson-Frisch (APF) is a new hybrid density functional method that combines B3PW91 and PBE0. APF-D provides an additional empirical dispersion correction method based on a spherical atom model (SAM), which is different from the Grimme's empirical dispersion correction method. APF-D accurately describes the binding energy and the potential energy surfaces of complexes of noble gas atoms and small hydrocarbon dimers. However, APF-D is not accepted as a standard method to study intermolecular interactions because the results often show a large deviation from the normal range when using the APF-D method to calculate the binding energy of hydrogen bonded complexes, C-H…π and π…π interactions. Our research identified that such a deviation arises from some long-range dispersion that has been double counted by the APF function and the SAM dispersion correction. Therefore, we propose an improved APF-D method, termed APF-D^*. By taking advantage of ζ, which is independent of SAM dispersion, we were able to solve effectively the problem of excessive dispersion compensation in APF-D. By comparing the results from S66 and L7 benchmark sets, we find that APF-D^* greatly improved the precision of calculations over the traditional APF-D method. The overall accuracy of APF-D^* was found to be comparable to or better than current leading DFT methods, such as B3LYP-D3 and ωB97X-D. However, both B3LYP-D3 and ωB97X-D have a much larger computational cost than APF-D^*. We believe that APF-D^* is a better method to calculate of the intermolecular energy of large molecules.

Key words: DFT-D, APF functional, SAM dispersion correction