Regensburg 2019 – scientific programme

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

MM: Fachverband Metall- und Materialphysik

MM 16: Methods in Computational Materials Modelling (methodological aspects, numerics)

MM 16.6: Talk

Tuesday, April 2, 2019, 11:45–12:00, H45

Machine learning and the thermodynamics of grain boundary segregation — •Liam Huber, Raheleh Hadian, Blazej Grabowski, and Joerg Neugebauer — Max-Planck-Institut für Eisenforschung GmbH, Düsseldorf, Germany

Solute-grain boundary (GB) interaction critically impacts the evolution and stabilization of grain structure and thus has a strong impact on the final material properties. At the atomic scale, structural disorder at the GB can create a wide variety of local environments for segregating atoms, and thus a wide distribution of segregation energies. Using classical molecular statics, we perform high-throughput calculations of six solutes to 38 different boundaries in Al obtaining 1.4 million segregation energies. With this rich dataset, we demonstrate that the traditional Langmuir-McLean model, which approximates solute-GB interaction with a single effective energy, is insufficient. By applying machine learning techniques, we provide a new and computationally highly efficient path to obtain the full energy distribution of the solvents. Extending this approach we have also calculated a corresponding set of surface segregation energies. Using a similar distribution-based approach provides us with a direct route to assess the role of solutes on GB embrittlement.