from total energies in the random phase approximation*

While the limitations (semi-)local density functionals for the electronic structure are widely acknowledged, total energies are usually considered to be rather accurate. However, in transition metal compounds, standard density or even hybrid functionals often predict the wrong ground state structure. Total energy calculations in the random phase approximation (RPA) greatly improve the phase stability prediction, e.g., rock-salt vs wurtzite in MnO, but quantitative predictions are still sensitive on input wavefunction for the RPA energy [1]. Here, we calculate the phase stability for MnSe and MnTe in the rocksalt, nickeline, and wurtzite structures, for the "negative-pressure" phase in MnSeTe alloys [2]. To account for the wavefunction dependence, we perform a variational minimization of the RPA energy with respect to the onsite potentials U and V [1]. In VO2, the failure of standard DFT to produce a band gap can be remedied by simple DFT+U calculation, at the expense of the incorrect prediction that the undistorted antiferromagnetic phase is lower in energy than the experimentally known nonmagnetic monoclinic phase. We further discuss results based on the SCAN functional.
[1] H. Peng and S. Lany, PRB 87, 174113 (2013)
[2] S. Siol et al, Sci. Adv. 4, eaaq1442 (2018)