Small closed-form CI expansions for electronic g-tensor calculations

In recent years, major progress has been made toward devising accurate and practical computational methods for predicting electronic g-tensors of molecules, thus raising hopes of extending the capability to solid-state defects and solvated radicals. The best agreement with experiment has been obtained at the multireference configuration interaction level via explicit sum-over-states (SOS) expansions. This method is computationally very intensive, however, and the explicit expansion may prove unworkably convoluted for larger, more complex species. SOS perturbation expansions of g-tensors are dominated by states that magnetically couple with the ground state. For this reason, one may optimize g-tensor expansions by neglecting configurations not coupled magnetically with the ground state. This criterion yields a configuration space that is specially tailored to magnetic response, yet is small enough to allow for a closed-form configuration interaction (CFCI) treatment of the entire resulting excited-state manifold. This CFCI method has been applied herein to predict g-tensors for H2O+, MgF, NO2, CO+, and H2CO+. CFCI results compare very favorably with other high-level theoretical methods and are clearly superior to Restricted Open Shell Hartree Fock (ROHF) treatment, for which new numbers are reported herein. Expansions limited to one-open-shell determinants yield excellent results for H2O+ and MgF and decent values for CO+ and NO2. Additional three-open-shell configurations improve results for both CO+ and NO2, and do not significantly undermine the good agreement in H2O+ and MgF. H2CO+, however, remains problematic within this formalism, probably due either to the shortcomings of the uncorrelated around state or in the approximate treatment of the excited manifold.