Band theory and Mott insulators: Hubbard U instead of Stoner I.

TLDR: A form for the exchange-correlation potential in local-density band theory, appropriate for Mott insulators, and finds that all late-3d-transition-metal monoxides, as well as the parent compounds of the high-${\mathit{T}$ compounds, are large-gap magnetic insulators of the charge-transfer type.

Abstract: We propose a form for the exchange-correlation potential in local-density band theory, appropriate for Mott insulators The idea is to use the ``constrained-local-density-approximation'' Hubbard parameter U as the quantity relating the single-particle potentials to the magnetic- (and orbital-) order parameters Our energy functional is that of the local-density approximation plus the mean-field approximation to the remaining part of the U term We argue that such a method should make sense, if one accepts the Hubbard model and the success of constrained-local-density-approximation parameter calculations Using this ab initio scheme, we find that all late-3d-transition-metal monoxides, as well as the parent compounds of the high-${\mathit{T}}_{\mathit{c}}$ compounds, are large-gap magnetic insulators of the charge-transfer type Further, the method predicts that ${\mathrm{LiNiO}}_{2}$ is a low-spin ferromagnet and NiS a local-moment p-type metal The present version of the scheme fails for the early-3d-transition-metal monoxides and for the late 3d transition metals