Modelling van der Waals interactions in molecular systems

It is shown that modest computational techniques corrected for vdW interactions do not provide sufficient accuracy for modelling physisorption interaction energies. Dispersion corrected Hartree-Fock (HF) and Density functional theory (DFT) are compared to highly accurate coupled cluster results to assess the accuracy of these less computationally intensive methods. Grimme¿s D3(BJ) correction is applied to the B3LYP, TPSS, PBE, and PBE0 functionals, while the Tang and Toennies (TT) correction is applied to HF. As to establish a foundational knowledge of how well these methods perform, the separation dependent interaction energies are modelled first for the inert gas dimers Ne2, Ar2, Kr2, Ne-Ar, and Ar-Kr as well as the simple molecular systems (H2)2, (CH4)2, Benzene2, H2-Benzene, and CH4-Benzene. Interestingly, PBE and PBE0 performed better without the correction. However, of all methods, B3LYP+D3(BJ) provided the most consistent accuracy across all systems when compared to coupled cluster results. B3LYP+D3(BJ) is then compared to a coupled cluster approximation ALMO+rCCD(T) for H2-Graphene and NO2-Graphene interactions in which the respective molecules physisorb to the graphene surface.

In the H2-Graphene interaction, B3LYP+D3(BJ) gave a reasonable fit to the ALMO+rCCD(T) interaction potential, but overestimated the binding energy by 17%. For the NO2-Graphene interaction, there is sufficient error in the fit of B3LYP+D3(BJ) to ALMO+rCCD(T) that prevents any conclusion that it could accurately model the interaction energy.