SKM 2023 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
O: Fachverband Oberflächenphysik
O 85: Electronic Structure of Surfaces II
O 85.11: Vortrag
Donnerstag, 30. März 2023, 17:30–17:45, REC C 213
Ultra-fast machine learning potentials for hydrogen under pressure — •Thomas Bischoff1, Bastian Jäckl1, and Matthias Rupp1,2 — 1Department of Computer and Information Science, University of Konstanz, Germany — 2Materials Research and Technology Department, Luxembourg Institute of Science and Technology (LIST), Luxembourg
Hydrogen exhibits remarkably complex behavior at high pressures. Its rich phase space with multiple solid and liquid polymorphs is the subject of controversial scientific debate [1,2].
We apply ultra-fast potentials (UFPs) to model hydrogen under pressure. UFPs are recent robust interpretable machine-learning potentials that enable accurate simulations of large atomistic systems over long time scales [3].
We examine the accuracy of UFPs for training data from density functional theory and quantum Monte Carlo calculations.
We also demonstrate the physical interpretability of UFPs for reference configurations composed of atomic and molecular hydrogen.
With the obtained machine-learning potential, we investigate solid-liquid and liquid-liquid phase transitions for an extensive part of the phase diagram of dense hydrogen.
[1] (a) B. Cheng et al., Nature 585: 217, 2020; (b) V.V. Karasiev et al., Nature 600: E12, 2021; (c) B. Cheng et al., Nature 600: E15, 2021.
[2] A. Tirelli et al., Physical Review B 106(4): L041105, 2022.
[3] S. R. Xie et al., arXiv 2110.00624, 2021.