Benchmarking PBE+D3 and SCAN+rVV10 methods using potential energy surfaces generated with MP2+ΔCCSD(T) calculation

We develop a benchmark system for van der Waals interactions obtained with MP2+ΔCCSD(T) method at complete basis set limit. With this benchmark, we examine the widely used PBE+D3 method and recently developed SCAN+rVV10 method for density functional theory calculations. Our benchmark is based on two molecules: glycine (or Gly, an amino acid) and uracil (or U, an RNA base). We consider six dimer configurations of the two monomers and their potential energy surfaces as a function of relative distance and rotation angle. The Gly-Gly, Gly-U, and U-U pairs represent London dispersion, hydrogen bonding, and π–π stacking interactions, respectively. Our results show that both PBE+D3 and SCAN+rVV10 methods can yield accuracy better than 1 kcal/mol, except for the cases when the distance between the two monomers is significantly smaller than the equilibrium distance. In such a case, neither of these methods can yield uniformly accurate results for all the configurations. In addition, it is found that the SCAN and SCAN+rVV10 methods can reproduce some subtle features in a rotational potential energy curve, while the PBE, PBE+D3, and the local density approximation fail. (paper)