F39: Density Functional Theory in Chemical Physics: Multiconfigurational Pair-DFT, Strong Correlation and Other Beyond-DFT Techniques

Tue. March 5, 8:48 a.m. – 9:00 a.m. CST

103E

Multiconfigurational wave function methods have several advantages over density functional theory: they give a clear picture of the physical state of the molecule with good quantum numbers and matrix elements. However, there is often overwhelming choice in choosing the configurational degrees of freedom to vary in order to obtain a good wave function with these methods (i.e, the problem of "active space selection"). Thus, multiple wave function results can be generated with no hint as to which result is better. In this work, we show that density functional theory can help solve this problem. By calculating the energy of multiconfigurational wave functions using density functionals instead of the molecular Hamiltonian (i.e., using multiconfigurational pair-density functional theory, MC-PDFT), one can select between the wave function results variationally. We apply this strategy to a large database of 400 vertical excitations by generating four possible multiconfigurational results and choosing the result lowest in the MC-PDFT energy. Using this approach, we find that we can produce the results of previous multiconfigurational benchmarks with a mean absolute error of 0.17 eV. Thus, this discrete variational selection (DVS) approach to generating multiconfigurational wave functions appears promising for the high-throughput application of multiconfigurational theories, and shines light on the interpretation of MC-PDFT in the context of density functional theory.