Dresden 2026 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
MM: Fachverband Metall- und Materialphysik
MM 25: Interface Controlled Properties, Nanomaterials, and Microstructure Design I
MM 25.2: Talk
Wednesday, March 11, 2026, 16:00–16:15, SCH/A215
Accelerating grain boundary segregation studies in ferritic steels with machine learned interatomic potentials — •Han Lin Mai1, Tilmann Hickel2, and Jörg Neugebauer1 — 1Max Planck Institute for Sustainable Materials GmbH, Dusseldorf, Germany — 2BAM Federal Institute for Materials Research and Testing, Berlin, Germany
The segregation of solute and impurities to grain boundaries (GBs) can critically alter the mechanical properties of steels. To predict segregation phenomena, ab-initio methods such as density functional theory (DFT) are frequently used, but their computational expense limits the size of the model GBs viable for computation. This limitation has rendered access to segregation statistics out of reach and therefore has hindered our understanding of segregation phenomena. Here, we present a machine learned interatomic potential (MLIP) for multiple Fe-X binary alloys that predict segregation energies approaching ab-initio accuracy with automated dataset generation techniques. We use these potentials to conduct high-throughput segregation studies to generate segregation energy spectra and compare these to those commonly found in GBs accessible to DFT studies. Commonly purported relationships between GB quantities such as excess volume, site volume and GB energies and the strength of segregation binding of solutes and impurities are re-examined and revised.
Keywords: Grain boundary; Segregation; Cohesion; Machine learned interatomic potentials; Density functional theory