Berlin 2024 – scientific programme

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

MM: Fachverband Metall- und Materialphysik

MM 45: Mechanical Properties and Alloy Design: e.g. Light-Weight, High-Temperature, Multicomponent Materials IV

MM 45.2: Talk

Wednesday, March 20, 2024, 16:00–16:15, C 230

Machine-learning structural stability of complex intermetallic phases — •Mariano Forti, Ralf Drautz, and Thomas Hammerschmidt — ICAMS, Ruhr-Universität Bochum. Universitätsstr. 150 44801 Bochum

The understanding of the precipitation of topologically close packed (TCP) phases in single-crystal superalloys is of central importance for the design of these materials for high-temperature applications. However, the structural complexity of these intermetallic compounds and the chemical complexity of the superalloys with typically up to ten elements hampers the exhaustive sampling of chemical space by density-functional theory (DFT) calculations. For example, the computation of the convex hull of the R phase with 11 inequivalent lattice sites would require N¹¹ DFT calculations in an N-component system. We overcome this computational limitation by combining machine learning (ML) techniques with descriptors of the local atomic environment of the TCP phases. In particlar, we use descriptors derived from bond order potentials (BOP) and atomic cluster expansions (ACE) that retain structural and electronic information. The resulting ML models predict the relative stability of complex TCP phases with very good precision in binary and ternary systems even for small training-data sets of only few hundred data points. We explore strategies for knowledge based feature selection that make it possible to handle the exponentially growing number of features in multicomponent systems, and to obtain a prediction for the convex hull of the R phase in Cr-Co-W system.

Keywords: TCP phases; Machine learning; formation enthalpies