A case study of density functional theory and domain-based local pair natural orbital coupled cluster for vibrational effects on EPR hyperfine coupling constants: vibrational perturbation theory versus ab initio molecular dynamics

Local approximations of high-level ab initio methods make superior accuracy in the computation of molecular properties accessible by drastically decreasing computational times. As a consequence, these methods become applicable not only for large systems but also in schemes for which large numbers of calculations are necessary. In this work, we apply a recently developed open-shell implementation of the domain-based pair natural orbital coupled cluster singles doubles (DLPNO-CCSD) approach for the computation of vibrational corrections to the isotropic values of electron paramagnetic resonance (EPR) hyperfine coupling constants. We assess density functional theory (DFT) and DLPNO-CCSD approaches using two common but very different schemes: (1) vibrational perturbation theory based on equilibrium geometries, and (2) explicit canonical ensemble averages using configuration snapshots sampled from revPBE0-D3(0) ab initio molecular dynamics simulations. Both approaches are found to yield very similar results for the spin probe 2,2,3,4,5,5-hexamethylperhydroimidazol-1-oxyl (HMI) and are both feasible for systems of around 30 atoms. However, the numerical stability required for higher derivatives can become a limitation for local correlation methods in the case of vibrational perturbation theory.