Pair Natural Orbital Restricted Open-Shell Configuration Interaction (PNO-ROCIS) Approach for Calculating X-ray Absorption Spectra of Large Chemical Systems.

Abstract

In this work, the efficiency of first-principles calculations of X-ray absorption spectra of large chemical systems is drastically improved. The approach is based on the previously developed restricted open-shell configuration interaction singles (ROCIS) method and its parametrized version, based on a density functional theory (DFT) ground-state determinant ROCIS/DFT. The combination of the ROCIS or DFT/ROCIS methods with the well-known machinery of the pair natural orbitals (PNOs) leads to the new PNO-ROCIS and PNO-ROCIS/DFT variants. The PNO-ROCIS method can deliver calculated metal K-, L-, and M-edge XAS spectra orders of magnitude faster than ROCIS while maintaining an accuracy with calculated spectral parameters better than 1% relative to the original ROCIS method (referred to as canonical ROCIS). The method is of a black box character, as it does not require any user adjustments, while it scales quadratically with the system size. It is shown that for large systems, the size of the virtual molecular orbital (MO) space is reduced by more than 90% with respect to the canonical ROCIS method. This allows one to compute the X-ray absorption spectra of a variety of large "real-life" chemical systems featuring hundreds of atoms using a first-principles wave-function-based approach. Examples chosen from the fields of bioinorganic and solid-state chemistry include the Co K-edge XAS spectrum of aquacobalamin [H2OCbl]+, the Fe L-edge XAS spectrum of deoxymyoglobin (DMb), the Ti L-edge XAS spectrum of rutile TiO2, and the Fe M-edge spectrum of α-Fe2O3 hematite. In the largest calculations presented here, molecules with more than 700 atoms and cluster models with more than 50 metal centers were employed. In all the studied cases, very good to excellent agreement with experiment is obtained. It will be shown that the PNO-ROCIS method provides an unprecedented performance of wave-function-based methods in the field of computational X-ray spectroscopy.