A State-Averaged Formulation for Variational Multiconfigurational Pair-Density Functional Theory

The accurate and efficient prediction of excited states in systems exhibiting both static and dynamic correlation remains a significant challenge in computational chemistry. To handle such challenges, we present a new state-averaged formulation for variational multiconfiguration pair-density functional theory. We assess the performance of this method on excitation energies for a diverse set of open-shell single radicals and closed-shell organic molecules. We compare the translated pure, global hybrid, and range-separated PDFT functionals to regular spin-TD-DFT and high-level wave function methods. For small open-shell radicals, which are characterized by strong correlation, we find that MC-PDFT with pure and─more significantly─global hybrid GGA translated functionals remarkably improve results of standard Kohn-Sham DFT (B3LYP, CAM-B3LYP) and EOM-CCSD. When range separation is included, mean absolute errors against MRCISD+Q are as low as 0.06 eV. Regarding closed-shell medium-sized organic molecules, our analysis reveals a critical dependency on the electronic nature of the transition. While translated pure GGAs struggle with π → π* transitions ─often severely underestimating energies─the introduction of range separation effectively corrects even large deviations. The range-separated sr-ctBLYP functional provides a balanced description of both n → π* and π → π* states and is shown to be competitive with high-level wave function methods such as CC3 and NEVPT4, offering a robust and computationally efficient strategy for modeling complex chemical systems.