Universal Framework for Multiconfigurational DFT

Strong correlation remains a significant challenge for DFT with no satisfying solutions found yet within the standard Kohn-Sham framework. Instead, for decades, a number of different approaches have been suggested to combine the accuracy of multiconfigurational methods with the efficiency of DFT. In this article, we demonstrate that many of these methods are or would be significantly improved by being reformulated as variants of multiconfigurational pair-density functional theory (MC-PDFT). This work presents the first implementation of these methods within the recently proposed variational formulation of MC-PDFT. It also provides for the first time a systematic comparison of their accuracy across representative examples of strongly correlated systems. By analyzing their accuracy and formal properties, we provide design guidelines to inform the development of future functionals.