DFT‐based solution to the gap problem of antiferromagnetic transition metal oxides and parent compounds of high‐Tc superconductors

Currently used approximations to the exchange‐correlation potential in density functional theory (DFT) are known to fail in describing the properties of certain compounds, of which we discuss here only two examples: CoO and stoichiometric La2CuO4. Both materials are insulating and antiferromagnetic. A DFT calculation on CoO yields antiferromagnetic order, but Co‐associated magnetic moments that are by ∼1 μB smaller than the experimental value, and one obtains the electronic structure of a metal. The latter applies also to La2CuO4, and, in contrast to the experiment, the calculation does not even yield nonzero moments associated with the Cu atoms. We exploit the fact that approximate exchange‐correlation potentials lead necessarily to spin‐dependent densities that differ from the exact ones. We therefore derive modified Kohn–Sham (KS) equations in which the effective potentials depend on the exact spin densities rather than on the standard KS densities. If the latter are modified by adding small (spin‐up, spin‐down) portions that individually integrate to zero within the lattice unit cell and do not change the total charge density, the inconsistencies with the experiment can be removed.