Item

Abstract

In this study, a detailed calibration of the performance of modern ab initio wave function methods in the domain of X-ray absorption spectroscopy (XAS) is presented. It has been known for some time that for a given level of approximation, for example, using time-dependent density functional theory (TD-DFT) in conjunction with a given basis set, there are systematic deviations of the calculated transition energies from their experimental values that depend on the functional, the basis set, and the chosen treatment of scalar relativistic effects. This necessitates a linear correlation for a given element/functional/basis set combination to be established before chemical applications can be performed. This is a laborious undertaking since it involves sourcing trustworthy experimental data, lengthy geometry optimizations, and detailed comparisons between theory and experiment. In this work, reference values for the element-specific shifts of all the first-row transition metal atoms and the main group elements C, N, O, F, Si, P, S, and Cl have been computed by using a protocol that is based on the complete active space configuration interaction in conjunction with second-order N-electron valence state perturbation theory (CASCI/NEVPT2). It is shown that by extrapolating the results to the basis set limit of the method and taking into account scalar relativistic effects via the second-order Douglas–Kroll–Hess (DKH2) corrections, the predicted element shifts are on average less than 0.75 eV across all the absorption edges and a very good agreement between theory and experiment in all the studied XAS cases is observed. The transferability of these errors to molecular systems is thoroughly investigated. The constructed CASCI/NEVPT2 database of element shifts is used to evaluate the performance and to automatically calibrate prior to comparison with the experiment two commonly used methods in X-ray spectroscopy, namely, DFT/Restricted open shell configuration interaction singles (DFT/ROCIS) and TD-DFT methods.