Now Available: A New Parameterization of the DFT/CIS Method with Applications to X-ray Spectroscopy

Our latest Q-Chem webinar (Webinar 77), which was presented by Aniket Mandal on October 24, is now available! You can view the archived recording of the webinar here.

Time-dependent density functional theory (TD-DFT) within a restricted excitation space is used as an efficient and reasonably accurate means to compute core-level spectra, using only a small subset of the occupied orbitals in what is known as the core/valence separation approximation. TD-DFT generates spectra with accurate relative peak positions, however, TD-DFT using standard exchange-correlation functionals affords core-to-valence excitation energies significantly smaller than experimental values, due in part to TDDFT’s failure to describe Rydberg and charge-transfer excited states. This leads to TD-DFT spectra needing significant scalar shifts to align with experimental spectra.

To avoid these problems, while retaining the favorable scaling of TD-DFT, an empirically-corrected combination of the configuration interaction with single substitutions (CIS) method based on Kohn-Sham orbitals has been implemented, which is known as “DFT/CIS”. This semiempirical approach is well-suited for simulating x-ray spectra, as it contains additional exact exchange over what is already present in the Kohn-Sham functional to model charge-transfer excitations yet retains the parent functionals low-cost description of dynamical election correlation. Static correlation in the excited state, and “many-electron” effects on oscillator strengths, are described by the CI wave function. The use of a fixed set of optimized parameters for a specific functional makes this a black-box method, and we have tested new variants based on range-separated hybrid functionals. The newly-parameterized DFT/CIS approach is applied to simulate x-ray absorption and emission spectra for a variety of systems, primarily at elemental K-edges. Results compare favorably to the best-available ab initio benchmarks.