SKM 2023 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

MM: Fachverband Metall- und Materialphysik

MM 29: Data Driven Materials Science: Big Data and Work Flows – Electronic Structure

MM 29.5: Vortrag

Mittwoch, 29. März 2023, 12:45–13:00, SCH A 251

A machine-learned interatomic potential for silica and mixed silica-silicon systems — •Linus C. Erhard¹, Jochen Rohrer¹, Karsten Albe¹, and Volker L. Deringer² — ¹Institute of Materials Science, TU Darmstadt, Darmstadt, Germany — ²Department of Chemistry, Inorganic Chemistry Laboratory, University of Oxford, Oxford, United Kingdom

The interface between silica and silicon has enormous relevance in various applications, including semiconductors and novel battery materials. However, atomistic modeling of this interface is a challenge due to the different charge states of silicon and the limitations of classical interatomic potentials. To overcome these limitations, we introduce a machine-learning-based interatomic potential based on the non-linear atomic cluster expansion (ACE) for various Si-O phases. This model is based on the previously developed database for silica [1], which was substantially extended by active learning. The new model shows improved performance for high-pressure silica and is also able to describe silica surfaces. Moreover, the use of the ACE formalism enables us to reach more than 100 times longer time or larger length scales compared to the Gaussian approximation potential (GAP). Finally, the potential is able to describe off-stoichiometric mixtures of Si and SiO₂. This capability is used to investigate the nanostructure of silicon monoxide.

[1] Erhard et al. A machine-learned interatomic potential for silica and its relation to empirical models. npj Comput Mater 8, 90 (2022)