Dresden 2017 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
MM: Fachverband Metall- und Materialphysik
MM 65: Topical session: Data driven materials design - uncertainty approaches
MM 65.2: Talk
Thursday, March 23, 2017, 16:00–16:15, BAR 205
Numerical Quality Control for DFT–based Materials Databases — C. Carbogno1, K. S. Thygesen2, •B. Bieniek1, C. Draxl4, L. Ghiringhelli1, A. Gulans4, O. T. Hofmann3, K. W. Jacobsen2, S. Lubeck4, J. J. Mortensen2, M. Strange2, E. Wruss3, and M. Scheffler1 — 1FHI Berlin, Germany — 2DTU, Lyngby, Denmark — 3TU Graz, Austria — 4HU Berlin, Germany
Electronic-structure theory has become an invaluable tool in materials science. Still, the precision of different approaches has only recently been scrutinized thoroughly (for the PBE functional) using extremely accurate numerical settings [1]. Little is known, however, about code- and method-specific deviances and error bars that arise under numerical settings commonly used in actual calculations. We shed light on this issue by systematically investigating the deviances in total and relative energies as function of typical settings for basis sets, k-grids, etc. for 71 elemental [1] and 81 binary solids in four different electronic–structure codes. On the basis of the observed trends, we propose analytical models for the estimation of error bars. They are also cross-validated using data obtained from the NoMaD Repository [2]. Eventually, we discuss the extensibility of our approach towards more complex materials properties, e.g., band gaps, and its general applicability in the context of computational materials databases.
This work received funding from The Novel Materials Discovery (NOMAD) Laboratory, a European Center of Excellence.
[1] K. Lejaeghere et al., Science 351 (2016).
[2] https://nomad-repository.eu