SKM 2023 – scientific programme

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

MM: Fachverband Metall- und Materialphysik

MM 32: Topical Session: Defect Phases I

MM 32.2: Talk

Wednesday, March 29, 2023, 16:15–16:30, SCH A 216

Fully Automated Calculation of Defect Phase Diagrams — •Marvin Poul, Erik Bitzek, and Joerg Neugebauer — Max-Planck-Institut fuer Eisenforschung, Duesseldorf, Deutschland

Understanding the thermodynamics of segregation at crystal defects is an important part of successful materials engineering.[1] We present an efficient method that constructs finite temperature Defect Phase Diagrams (DPDs) for binary alloys from fully relaxed molecular calculations using machine learning interatomic potentials (MLIP) without user intervention and implemented it as a pyiron[2] workflow. A major challenge that we had to address is the combinatorially growing number of different segregation configurations at any extended defect. The proposed method is able to efficiently tackle hundreds of thousands to millions of configurations and is based on a fast proxy model. This model is based on the ACE descriptors and avoids having to evaluate a full MLIP. This proxy model together with the MAXVOL active learning algorithm allows to pre-screen which configurations to calculate with the underlying MLIP. We apply the workflow on the example of Al and Ca segregation to Mg grain boundaries.

[1]: Korte-Kerzel, S. et al. (2022) Defect phases: thermodynamics and impact on material properties, International Materials Reviews, 67:1, 89-117

[2]: Janssen, J, et al. "pyiron: An integrated development environment for computational materials science." Computational Materials Science 163 (2019): 24-36.