This contribution was initially motivated by Axel Becke’s ingenious work on treating nondynamic correlation in density functional theory (DFT) [1]. Becke’03, ‘05 and ’13 showed that the multireference character in electron correlation could be defined and treated with the single-determinant DFT, contrary to common beliefs. We modified the Becke’13 method by modeling the adiabatic connection for the correlation kinetic energy, resulting in a functional with far fewer parameters. The new functional, abbreviated as KP16/B13 [2], is not only capable of recovering the majority of strong correlations for extreme cases such as the dissociation of a covalent bond, but also competitive to contemporary heavily parameterized methods in standard benchmarking. The method was further extended to correct the charge delocalization error of DFT in some common cases.

Still, KP16/B13 was unable to handle general ensemble-representable densities. Indeed, some basic concepts related to that type of density, such as the locality, physical necessity of fractional charge, and v-representability, are still being debated. In this presentation, we argue that any physically reasonable density is ensemble v-representable within a finite accuracy [3,4]. We start from a hypothetical physical principle that defines molecules in a limited volume. The principle shows that all molecules are entangled in the number of electrons and leads to the possibility of fractional numbers of electrons. The universal v-representability admits the definition of an exchange-correlation functional for non-interacting ensemble densities. It is essentially equivalent to exact exchange for a single-determinant v-representable density and recovers most of the exchange-correlation effect for ensemble densities.

[1] A.D. Becke, J. Chem. Phys, 2013, 138, 074109.
[2] J. Kong and E. Proynov, J. Chem. Theory Comput., 2016, 12, 133.
[3] J. Kong, The Journal of Chemical Physics, 2024, 161, 224111.
[4] J. Kong, arXiv preprint arXiv:2405.00203