Berlin 2024 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
MM: Fachverband Metall- und Materialphysik
MM 53: Data Driven Material Science: Big Data and Workflows VI
MM 53.5: Vortrag
Donnerstag, 21. März 2024, 11:15–11:30, C 243
Complete Basis Set Limit Extrapolation in Density Functional Theory Calculations using Statistical Learning — •Daniel Speckhard1,2, Claudia Draxl1,2, and Matthias Scheffler2,3 — 1Humboldt-Universität zu Berlin, Physics Department and IRIS Adlershof, Berlin, Germany — 2Max-Planck-Institut für Festkörperforschung, Stuttgart, Germany — 3The NOMAD Laboratory at the FHI of the Max-Planck-Gesellschaft and IRIS-Adlershof of the Humboldt-Universität zu Berlin
The numerical precision of density-functional-theory (DFT) calculations depends on a variety of computational parameters, one of the most critical being the basis-set size. The ultimate precision is reached with an infinitely large basis set, i.e., in the complete basis-set (CBS) limit. Our aim is to find a machine-learning model that extrapolates finite basis-size calculations to the CBS limit. Quantile random forests and symbolic regression, applying the SISSO approach, are used to estimate total energies, lattice parameters, and band gaps as a function of the basis-set size. The random-forest model outperforms previous approaches in the literature for both codes, while SISSO outperforms the random-forest model for the exciting code. Our approach also provides prediction intervals, which quantify the models’ uncertainty. We evaluate our work on datasets consisting of 63 binary solids and 4000 binary semiconductors, respectively.
[1] C. Carbogno et al., npj Comput. Mater. 8, 69 (2022).
Keywords: Density functional theory; Machine learning; Delta learning