About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Computational Thermodynamics and Kinetics
|
| Presentation Title |
A Magnetism-Aware Interatomic Machine-Learning Potential: Application to Surface Phase Diagrams for Magnetite |
| Author(s) |
Mira Todorova, Baptiste Bienvenu, Matous Mrovec, Ralf Drautz, Joerg Neugebauer |
| On-Site Speaker (Planned) |
Mira Todorova |
| Abstract Scope |
Iron oxides and their surfaces are important in applications such as catalysis, spintronics, and iron ore reduction. Deviations from stoichiometry, which have been observed in both the bulk and the surface of these oxides, result in complex surface structures and reconstructions. Our recently developed Fe-O machine-learning interatomic potential [npj Comput. Mater. 11 (2025) 81], based on the Atomic Cluster Expansion (ACE), allows us to efficiently explore the configurational space of possible surface structures. Performing hybrid molecular dynamics – Monte Carlo (MD-MC) simulations in a grand canonical ensemble under varying oxygen and iron chemical potentials, we scan possible surface terminations of the low-index facets of Magnetite (Fe3O4) and uncovered some previously unreported surface structures. Validating all identified low energy structures by density-functional theory (DFT) calculations, we construct fully ab initio surface phase diagrams and show that the computed Wulff shapes match experimental observed ones for Hematite nanoparticles. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, Machine Learning, Modeling and Simulation |