Abstract
Interpreting electronic spectra of uranium-containing compounds is an important component of fundamental chemistry as well as in the assessment of waste streams in the nuclear fuel cycle. Here we employ multiconfigurational calculations with CASSCF or DMRGSCF methods on exemplar uranium molecules [U(VI)O(2)Cl(4)](2-), [U(V)(TREN(TIPS))(N)](-), and [U(IV)Cl(5)(THF)](-), featuring an array of geometries and oxidation states, to determine their effectiveness in predicting electronic spectra, compared to literature calculations and experimental data. For [U(VI)O(2)Cl(4)](2-), DMRGSCF alone shows poor agreement with experiment, which can be improved by adding corrections for dynamic correlation with MC-PDFT to give results of similar quality to TD-DFT. However, for [U(V)(TREN(TIPS))(N)](-) the addition of dynamical correlation via MC-PDFT or CASPT2 made no improvements over CASSCF, suggesting that perhaps other factors such as solvation effects could be more important in this case. Finally, for [U(IV)Cl(5)(THF)](-), dynamical correlation included via MS-CASPT2 on top of CASSCF calculations is crucial to obtaining a quantitatively correct spectrum. Here, MC-PDFT fails to even qualitatively describe the spectrum, highlighting the shortcomings of single-state methods in cases of near-degeneracy.