Parts | Days | Selection | Search | Updates | Downloads | Help
MM: Fachverband Metall- und Materialphysik
MM 64: Liquid and Amorphous Materials IV
MM 64.4: Talk
Thursday, March 21, 2024, 17:30–17:45, C 243
Towards in-depth atomistic understanding of polymer-derived silicon oxycarbides using machine-learning potentials — Niklas Leimeroth, •Jochen Rohrer, and Karsten Albe — Institute of Materials Science, Technical University Darmstadt, Otto-Berndt-Strasse 3, 64287, Darmstadt, Germany
Polymer-derived amorphous silicon oxycarbides (SiOC) show promising properties for advanced applications in a variety of fields such as high-temperature coatings, biomedicine and batteries. This outstanding versatility is due to their highly tunable composition and microstructure. Simultaneously, this tunability poses a challenge for a thorough knowledge and understanding of structure-property relations in this system. In this work, we present a machine-learning potential (MLP) for SiOCs based on the atomic cluster expansion (ACE) and trained to a diverse set of actively-learned density functional theory (DFT) data. We demonstrate the capability of the MLP to model glass-phase and microstructure formation from commonly used polymer-precursor fragments, contrast these microstructures with experimental findings and show how atomistic simulations can be used to understand complex structure-property relations on the example of Young’s Moduli in relation to phase volumes and different types of bonding in the system.
Keywords: Machine-learning potentials; Atomistic simulations; Ceramics; Glasses