The existence of band gaps in Mott insulators such as perovskite oxides with partially filled 3d shells has been traditionally explained in terms of strong, dynamic inter-electronic repulsion codified by the on-site repulsion energy U in the Hubbard Hamiltonian. The success of the "DFT+U approach" where an empirical on-site potential term U is added to the exchange-and correlation Density Functional Theory (DFT) raised questions on whether U in DFT+U represents interelectronic correlation in the same way as it does in the Hubbard Hamiltonian, and if empiricism in selecting U can be avoided. Here we illustrate that ab-initio DFT without any U is able to predict gapping trends and structural symmetry breaking (octahedra rotations, Jahn-Teller modes, bond disproportionation) for all ABO3 3d perovskites from titanates to nickelates in both spin-ordered and spin disordered paramagnetic phases. We describe the paramagnetic phases as a supercell where individual sites can have different local environments thereby allowing DFT to develop finite moments on different sites as long as the total cell has zero moment. We use a recently developed exchange and correlation functional ("SCAN") that is sanctioned by the usual single-determinant, mean-field DFT paradigm with static correlations, but has a more precise rendering of self-interaction cancelation. Our results suggest that strong dynamic electronic correlations are not playing a universal role in gapping of 3d ABO3 Mott insulators, and opens the way for future applications of DFT for studying a plethora of complexity effects that depend on the existence of gaps, such as doping, defects, and band alignment in ABO3 oxides.
Read full abstract