Dresden 2020 – wissenschaftliches Programm
Die DPG-Frühjahrstagung in Dresden musste abgesagt werden! Lesen Sie mehr ...
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
CPP: Fachverband Chemische Physik und Polymerphysik
CPP 104: Topical Session: Data Driven Materials Science - Machine Learning Applications (joint session MM/CPP)
CPP 104.2: Vortrag
Donnerstag, 19. März 2020, 17:45–18:00, BAR 205
Investigation of short-range order in multicomponent alloys with the use of machine-learning interatomic potentials — •Tatiana Kostiuchenko1, Alexander Shapeev1, Fritz Körmann2,3, and Andrey Ruban4,5 — 1Skolkovo Institute of Science and Technology, Moscow, Russia — 2Delft University of Technology, Delft, The Netherlands — 3Max-Planck-Institut für Eisenforschung GmbH, Düsseldorf, Germany — 4KTH Royal Institute of Technology, Stockholm, Sweden — 5Materials Center Leoben Forschung GmbH, Leoben, Austria
Multicomponent alloys, such as high-entropy alloys or multi-principal element alloys, are promising structural materials. However, the vast range of chemical compositions and long annealing times offer an opportunity for their study with ab initio methods. The drawback of ab initio methods is their huge computational costs. We address this drawback by developing efficient data-driven interaction models with accuracy close to those of ab initio methods, namely the low-rank interatomic potential (LRP) [Shapeev A., 2017]. They are used in a Canonical Monte Carlo algorithm as an “on-lattice” model that can take into account local lattice distortions. In this work, we investigate the equiatomic VCoNi system. It represents a medium-entropy alloy with a distorted fcc lattice which leads to an outstanding strength-ductility relationship as reported in [Sohn S. et al., 2019]. We simulate this system by including, implicitly, magnetism into LRP by fitting on spin-polarized DFT calculation. The LRP has the error of about 1-3 meV/atom and is used to study the formation of the short-range order.