The Hubbard U correction for iron-bearing minerals: A discussion based on (Mg,Fe)SiO3 perovskite

Using iron-bearing magnesium silicate perovskite as an example, we show that the Hubbard U correction is not always necessary for calculating the structural and elastic properties of iron-bearing minerals. Instead, it is the choice of DFT functionals (LDA or GGA) that may have greater impact on the quality of the predictive calculation. For iron-bearing minerals, calculations adopting LDA (LDA + U) are generally in better agreement with experimental data than those adopting GGA (GGA + U) after including zero point motion and thermal vibrational effects, as demonstrated by the room-temperature compression curve of (Mg,Fe)SiO3 perovskite. A criterion indicating the necessity of the Hubbard U correction is also discussed. As long as the standard DFT functional produces insulating ground state with correct orbital occupancy, the Hubbard U correction does not affect the computed structural and elastic properties.