Multiconfiguration pair-density functional theory (MC-PDFT) and linearized pair-density functional theory (L-PDFT) are efficient methods for calculating ground and excited-state electronic energies of molecular systems. Here we present MC25, which is… Click to show full abstract
Multiconfiguration pair-density functional theory (MC-PDFT) and linearized pair-density functional theory (L-PDFT) are efficient methods for calculating ground and excited-state electronic energies of molecular systems. Here we present MC25, which is a new hybrid meta on-top-density functional developed for multiconfiguration pair-density functional theory (MC-PDFT) and linearized pair-density functional theory (L-PDFT). The new functional is an improvement of our previously developed MC23 functional; it is obtained by adding a more diverse set of databases with electronic excitation energies to the training set. MC25 has improved accuracy for the excitation energies with a mean unsigned error, averaged over same-spin and spin-change excitation energies, of 0.14 eV. Combining this with its approximately as-good-as-MC23 performance for ground-state energies means it has the best overall accuracy on the combination of both ground-state energies and excitation energies of any available on-top functional. It has more accuracy on both excitation energies and ground-state energies than 18 other methods that we tested on the same data. In addition to developing MC25, we also developed a local functional using the same set of databases and optimization algorithm. Although the local functional, which is named MC25L, is not as accurate as MC25, it is still more accurate than the three other local on-top functionals that are tested in this work (tPBE, tM06-L, and tτ-HCTH). We recommend using MC25 for both ground-state energies and excitation energies for both strongly- and weakly correlated systems, and MC25L may be useful in contexts where an on-top functional is applied with reference density matrices but without a CASSCF energy.
               
Click one of the above tabs to view related content.