Berlin 2024 – wissenschaftliches Programm

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

MM: Fachverband Metall- und Materialphysik

MM 17: Poster Ib

MM 17.19: Poster

Montag, 18. März 2024, 18:30–20:30, Poster F

Accurate ab initio vacancy properties in concentrated Mo-Ta alloys from machine-learning potentials — •Xiang Xu^1,3, Xi Zhang¹, Sergiy Divinski², and Blazej Grabowski¹ — ¹Institute for Materials Science, University of Stuttgart, Germany — ²Institute for Materials Physics, University of Münster, Germany — ³Institute for Materials Testing, Materials Science and Strength of Materials, University of Stuttgart, Germany

We utilize a bespoke machine-learning interatomic potential, i.e., moment tensor potential (MTP) to predict thermodynamic properties of vacancy formation. The highly optimized MTP is trained with snapshots from ab initio molecular dynamics simulations within the active learning framework. For the vacancy formation energy, we utilize the special quasi-random structure approach in conjunction with a statistical analysis, from which temperature-dependent formation Gibbs energies as well as averaged atomic environments can be extracted. We show that the temperature-dependent vacancy formation Gibbs energy due to "configurational excitations" has a negative entropy contribution while thermal vibrations provide a positive entropy. The local chemical environment effect and general trends are also analyzed.

Keywords: vacancy property; machine learning potential; Mo-Ta alloy; ab initio molecular dynamics