Regensburg 2022 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
MM: Fachverband Metall- und Materialphysik
MM 34: Data Driven Materials Science: Interatomic Potentials / Reduced Dimensions
MM 34.6: Vortrag
Donnerstag, 8. September 2022, 17:15–17:30, H45
Machine Learning of ab-initio grain boundary Segregation Energies — •Christoph Dösinger1, Daniel Scheiber2, Oleg Peil2, Vsevolod Razumovskiy2, Alexander Reichmann1, and Lorenz Romaner1 — 1Montanuniversität Leoben, Department of Materials Science, Leoben, Austria — 2Materials Center Leoben Forschung GmbH, Leoben, Austria
Grain-boundary (GB) segregation is an important phenomenon in alloys, where the resulting GB excess can strongly influence their properties, for example induce intergranular fracture or lead to phase transformations. A fundamental quantity that uniquely describes the propensity of a solute towards GB segregation is the segregation energy. It determines the tendency of a solute atom to enrich or deplete at the GB. This quantity can be directly calculated from first principles. However, such calculations are computationally expensive and can become computationally unfeasible as the complexity of the GB crystal structure increases. The aim of this work is to reduce the computational cost of GB segregation energies by applying machine learning methods trained at series of representative DFT calculations and expanding them to more complex GB structures. The atomic structure, together with the segregation energies are used to train a model, which then is employed to predict the segregation energy for arbitrary segregation sites and GB types. In our work we apply this method to tungsten alloys. The results show, that this approach indeed gives reliable results for the segregation energies and can be used to get a complete description of segregation profiles.