Exploring intermolecular interactions through Coupled Cluster, Density Functional, and Multireference Symmetry-Adapted Perturbation Theories

Abstract

Intermolecular interactions are computed at successively lower levels of theory to establish the relative accuracy of each level. A set of 21 small molecules was first computed using the CCSDT(Q) level of theory to establish the post-CCSD(T) uncertainty of approximately 3\%. Methane and CO$_2$ bound to a series of polycyclic aromatic hydrocarbons (PAHs) are then computed using a variety of tools to approximate the CCSD(T)/CBS interaction energy, the MP2/CBS+$\Delta$CCSD(T)-F12avg/aDZ method demonstrated a mean error of just 2\% from benchmark results. The accuracy of a set of dispersion including DFT methods is explored for methane and CO$_2$ bound to curved coronene systems. While these DFT methods exhibited mean errors of 5-15\% at the van der Waals minima their error at shorter ranges rose dramatically. In order to mitigate these short-range errors, the damping parameters of the DFT-D3 method were refitted to a large database of 8,299 intermolecular interactions. It was found that through refitting the average error of the DFT-D3 methods was reduced by 10-50\%, the greatest reduction in error came from the largest DFT-D3 outliers. The resulting refitted DFT-D3 method is more accurate and the error is less variable with respect to the choice of underlying DFT functional and damping form. In addition, symmetry-adapted perturbation theory (SAPT) will be extended to multiconfigurational self-consistent field (MCSCF) wavefunctions. To this end, optimization techniques for MCSCF wavefunctions are detailed and density-fitting is introduced into these equations to reduce their overall cost.