Accurate prediction of core-level spectra of radicals at density functional theory cost via square gradient minimization and recoupling of mixed configurations
Hait, Diptarka, Haugen, Eric A, Yang, Zheyue, Oosterbaan, Katherine J, Leone, Stephen R, Head-Gordon, Martin
Abstract
State-specific orbital optimized approaches are more accurate at predicting\ncore-level spectra than traditional linear-response protocols, but their\nutility had been restricted on account of the risk of `variational collapse'\ndown to the ground state. We employ the recently developed square gradient\nminimization (SGM, J. Chem. Theory Comput. 16, 1699-1710, 2020) algorithm to\nreliably avoid variational collapse and study the effectiveness of orbital\noptimized density functional theory (DFT) at predicting second period element\n1s core-level spectra of open-shell systems. Several density functionals\n(including SCAN, B3LYP and $\\omega$B97X-D3) are found to predict excitation\nenergies from the core to singly occupied levels to high accuracy ($\\le 0.3$ eV\nRMS error), against available experimental data. Higher excited states are\nhowever more challenging by virtue of being intrinsically multiconfigurational.\nWe thus present a CI inspired route to self-consistently recouple single\ndeterminant mixed configurations obtained from DFT, in order to obtain\napproximate doublet states. This recoupling scheme is used to predict the C\nK-edge spectra of the allyl radical, the O K-edge spectra of CO$^+$ and the N\nK-edge of NO$_2$ to high accuracy relative to experiment, indicating\nsubstantial promise in using this approach for computation of core-level\nspectra for doublet species (vs more traditional time dependent DFT, EOM-CCSD\nor using unrecoupled mixed configurations). We also present general guidelines\nfor computing core-excited states from orbital optimized DFT.