Benchmarking electronic structure calculations on the bare UO ₂²⁺ ion: How different are single and multireference electron correlation methods?

In a recent investigation by some of us on the spectrum of the uranyl (UO₂^2-) ion [Réal, F.; Vallet, V.; Marian, C.; Wahlgren, U. J. Chem. Phys. 2007, 126, 214302], a sizable difference between CASPT2 and linear response coupled cluster (LRCC) was observed both with and without the perturbative inclusion of spin-orbit coupling. This poses a serious question as to which of the two would be more reliable for investigating molecules containing actinides. In this paper we address this question by comparing CASPT2 and LRCC to a method known to accurately describe the spectra of actinide-containing molecules: the fourcomponent intermediate Hamiltonian Fock-space coupled cluster (IHFSCC) method, where electron correlation and spin-orbit coupling are treated on an equal footing. Our results indicate that for UO₂²⁺ there is little difference between treatments of spin-orbit coupling, making electron correlation the main cause of discrepancies. We have found IHFSCC and LRCC to be the most alike in the overall description of excited states, even though individual LRCC energies are blue-shifted in comparison to IHFSCC due to the difference in the parametrization of the excited states' wave functions. CASPT2, on the other hand, shows good agreement with IHFSCC for individual frequencies but significantly less so for the spectrum as a whole, due to the difference in the degree of correlation recovered in both cases.