Berlin 2018 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
MM: Fachverband Metall- und Materialphysik
MM 66: Topical Session (Symposium MM): Big Data in Materials Science - Managing and exploiting the raw material of the 21st century
MM 66.4: Vortrag
Donnerstag, 15. März 2018, 18:30–18:45, H 0107
Numerical-Error Estimates for DFT Calculations and Materials Databases — C. Carbogno1, K.S. Thygesen2, B. Bieniek1, C. Draxl1,3, L. Ghiringhelli1, A. Gulans3, O.T. Hofmann4, K.W. Jacobsen2, •S. Lubeck3, J.J. Mortensen2, M. Strange2, E. Wruss4, and M. Scheffler1 — 1FHI Berlin, Germany — 2DTU, Lyngby, Denmark — 3HU Berlin, Germany — 4TU Graz, Austria
Density-functional theory (DFT) has become an invaluable tool in materials science. Whereas the precision of different approaches has been scrutinized for the PBE functional using extremely accurate numerical settings [1], little is yet known about code- and method-specific errors that arise under more commonly used numerical settings. Recently, this has become a severe issue, since it prevents repurposing publicly available DFT data created using different settings and/or codes. To overcome this, we study the convergence of different properties (geometries, total and relative energies) in four conceptually-different DFT codes (exciting, FHI-aims, GPAW, VASP) for typical settings. Specifically, we discuss relative and absolute errors as a function of the numerical settings, e.g., basis sets and k-grids, for 71 elemental solids [1]. Using this data, we propose analytical models that allow for reliable error estimates for any compound, as we explicitly demonstrate for binary and ternary solids. We discuss the extensibility of our approach towards more complex materials properties and its applicability in computational materials databases.
[1] K. Lejaeghere et al., Science 351, aad3000 (2016).