Parts | Days | Selection | Search | Updates | Downloads | Help
O: Fachverband Oberflächenphysik
O 83: Frontiers of Electronic Structure Theory: Focus on Artificial Intelligence Applied to Real Materials 4
O 83.3: Talk
Friday, September 9, 2022, 11:00–11:15, S054
Machine learning TCP phases with domain knowledge of the interatomic bond — •Mariano Forti, Alesya Burakovskaya, Ralf Drautz, and Thomas Hammerschmidt — ICAMS, Ruhr-Universität Bochum, Universitätsstr. 150, 44801 Bochum, Germany.
The understanding of the precipitation of topological 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 N=5-10 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. We present descriptors that are derived from bond order potential (BOP) theory which retain domain knowledge of the interatomic interaction from tight-binding Hamiltonians. We demonstrate that these descriptors enable us to predict the structural stability of TCP phases with simple regression algorithms. We apply this methodology to several systems with experimental evidence of R phase formation.