Assessing exchange-correlation functional performance for structure and property predictions of oxyfluoride compounds from first principles

Motivated by the resurgence of electronic and optical property design in ordered fluoride and oxyfluoride compounds, we present a density functional theory (DFT) study of 19 materials with structures, ranging from simple to complex, and variable oxygen-to-fluorine ratios. We focus on understanding the accuracy of the exchange-correlation potentials (Vxc) to DFT for the prediction of structural, electronic, and lattice dynamical properties at four different levels of theory, i.e., the local density approximation (LDA), generalized gradient approximation (GGA), metaGGA, and hybrid functional level which includes fractions of exact exchange. We investigate in detail the metaGGA functionals MS2 [Sun et al., Phys. Rev. Lett. 111, 106401 (2013)] and SCAN [Sun et al., Phys. Rev. Lett. 115, 036402 (2015)], and show that although the metaGGAs show improvements over the LDA and GGA functionals in describing the electronic structure and phonon frequencies, the static structural properties of fluorides and oxyfluorides are often more accurately predicted by the GGA-level Perdew-Burke-Ernzerhof functional for solids, PBEsol. Results from LDA calculations are unsatisfactory for any compound, regardless of oxygen concentration. The PBEsol and Heyd-Scuseria-Ernzerhof (HSE06) functionals give good performance in all-oxide or all-fluoride compounds. For the oxyfluorides, PBEsol is consistently more accurate for structural properties across all oxygen concentrations; however, we stress the need for detailed property assessment with various functionals for oxyfluorides with variable composition. The "best" functional is anticipated to be more strongly dependent on the property of interest. Our study provides useful insights in selecting an Vxc that achieves optimal performance compromises, enabling more accurate predictive design of functional fluoride-based materials with density functional theory.