Assessment of metageneralized gradient approximation and screened Coulomb hybrid density functionals on bulk actinide oxides

Abstract

Recent advances in the field of density functional theory afford increasingly accurate and efficient studies on a wide range of materials, but validation of any new computational method requires comprehensive benchmarking of its performance on various classes of systems. In the present work, we assess two newly developed density functionals on bulk uranium and plutonium oxides, for which structural, magnetic, and one-electron properties are calculated. The new functionals are the metageneralized gradient approximation (meta-GGA) of Tao, Perdew, Staroverov, and Scuseria (TPSS) and the screened Coulomb hybrid functional of Heyd, Scuseria, and Ernzerhof (HSE). Their predictions are compared to those of the local spin density approximation (LSDA), the Perdew-Burke-Ernzerhof (PBE) realization of the GGA, and a hybrid implementation of the latter (PBE0). The nonhybrid density functionals LSDA, PBE, and TPSS generally fail to provide a satisfactory qualitative description of the electronic and magnetic structure of actinide oxides. TPSS improves upon the LSDA in the prediction of bulk parameters, but only to a level of accuracy comparable to that of PBE. It however outperforms both of them for one-electron properties, as it predicts a nonzero band gap for antiferromagnetic plutonium oxides. HSE is computationally more efficient than its parent functional PBE0, while being at least as accurate for structural, one-electron and magnetic properties. The predictions of HSE agree well with experiment where known, making it suitable for calculations on transitional and $f$-element compounds.