Regensburg 2022 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

MM: Fachverband Metall- und Materialphysik

MM 25: Data Driven Materials Science: Computational Frameworks / Chemical Complexity

MM 25.3: Talk

Wednesday, September 7, 2022, 16:30–16:45, H46

Atomic cluster expansion: a universal machine learning potential for magnesium — •Eslam Ibrahim, Yury Lysogorskiy, Matous Mrovec, and Ralf Drautz — ICAMS, Ruhr Universität Bochum, 44780 Bochum, Germany

In this work, we developed a general-purpose parametrization of the atomic cluster expansion (ACE) for Mg. The model shows an outstanding transferability over a broad range of atomic environments and is able to capture physical properties of bulk as well as defective Mg phases in excellent agreement with reference first-principles calculations. We demonstrate the computational efficiency and the predictive power of ACE by calculating the phase diagram covering temperatures up to 3000 K and pressures up to 80 GPa using state-the-art thermodynamic integration techniques implemented in the CALPHY software package. The ACE predictions are compared with those of common interatomic potentials, such as the embedded atom method or the angular-dependent potential, as well as a recently developed neural network potential. The comparison reveals that ACE is the only model that is able to predict both qualitatively and quantitatively correctly the phase diagram in close experiment with experimental observations.