Computationally tractable approaches to ultra-high temperature anharmonicity

Advances in high-performance computing and first principles methodologies herald greater opportunities than ever to make high-temperature predictions at the ab initio level of theory. This has led to recent predictions up to the melting point, for metals such as aluminium,[1] and for ceramics such as ZrC.[2] Despite major advances in beyond-quasiharmonic approaches, for real materials, full anharmonic treatments are still prohibitively expensive computationally.

In this work we present two first principles approximations to account for anharmonicity at high temperature: a local-uncoupled potential model, and a thermodynamic integration based methodology. The accuracy of anharmonic free energies and derivatives thereof are bench-marked for ZrC against predictions from the two-stage up-sampled thermodynamic integration of Langevin dynamics (TU-TILD) method.[2] The anharmonic approximations we present provide meaningful improvements in computational efficiency, and incur relatively small losses in accuracy. Efficiency gains are sufficiently large to compute anharmonic vacancy formation energies using moderate computational resources.

[1] B. Grabowski et al., Physical Review B, 79, 134106, 2009
[2] A.I. Duff et al., Physical Review B, 91, 214311, 2015