Parts | Days | Selection | Search | Updates | Downloads | Help
MM: Fachverband Metall- und Materialphysik
MM 25: Data Driven Materials Science: Computational Frameworks / Chemical Complexity
MM 25.8: Talk
Wednesday, September 7, 2022, 18:15–18:30, H46
Databases for Machine Learning of Grain Boundary Segregation — •Alexander Reichmann1, Christoph Dösinger1, Daniel Scheiber2, Oleg Peil2, Vsevolod Razumovskiy2, and Lorenz Romaner1 — 1Department of Materials Science, Leoben, Austria — 2Materials Center Leoben Forschung GmbH, Leoben, Austria
The chemical and structural state of grain boundaries (GBs) is of great importance for the design and performance of many technologically relevant materials. On the basis of atomistic simulations, the relevant quantities of GB, in particular the segregation energy has been calculated for many materials. On the experimental side, the concentration of solute elements at the GBs can be measured with a variety of techniques including in particular Auger spectroscopy or Atom Probe Tomography. When comparing calculated segregation energies with segregation energy gained from experimental excess data, good agreement is not always observed. In this talk we will present our current and planned activities regarding creation of segregation databases and application of data driven models. One of these is the Bayesian inference framework, which we used in combination with Markov chain Monte Carlo simulations for uncertainty quantification and model evaluation. These activities shall lead to a better understanding of the deviation between DFT-calculated and experimentally determined GB excess.