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

MM: Fachverband Metall- und Materialphysik

MM 31: Data Driven Materials Science: Big Data and Work Flows – Machine Learning

MM 31.8: Talk

Wednesday, March 29, 2023, 17:45–18:00, SCH A 251

Data-driven magneto-elastic interatomic potentials for discovering novel phases of transition metal alloys — •Mani Lokamani¹, Kushal Ramakrishna⁴, Julian Tranchida³, Svetoslav Nikolov², Hossein Tahmasbi⁴, Michael Wood², and Attila Cangi⁴ — ¹HZDR Dresden, Germany — ²SNL New Mexico, USA — ³CEA Cadarache, France — ⁴CASUS Görlitz, Germany

Structural prediction methods are used for identifying stable and metastable structures in a broad spectrum of materials. The presence of the electron spin degree of freedom in magnetic materials increases the complexity of finding such structures, constraining the analysis to the thermodynamically most relevant structures in a narrow range of temperatures and pressures. We achieve a search over much wider temperature and pressure conditions by utilizing machine-learning interatomic potentials based on the spectral neighbor analysis method within the coupled spin-molecular dynamics framework implemented in LAMMPS. This data-driven methodology enables predicting the properties of magnetic materials on much larger spatial, spin, and temporal domains and is parametrized by first-principles data. Leveraging this methodology, we predict the formation of metastable crystalline structures in transition metal alloys (FeNi, FeMn, FeCr, FeCo, FeGd) at high temperature-pressure conditions and assess their magnetic properties. This enables studying long-range spin structures in novel phases of transition metal alloys and complements the quest for permanent magnets for renewable energy applications that do not depend on rare-earth elements.