Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

MM: Fachverband Metall- und Materialphysik

MM 25: Data Driven Materials Science: Computational Frameworks / Chemical Complexity

MM 25.5: Topical Talk

Mittwoch, 7. September 2022, 17:15–17:45, H46

Understanding Dislocation Flow and Avalanches in High Entropy Alloys by Machine Learning-based Data Mining of In-Situ TEM Experiments — •Stefan Sandfeld — FZJ/IAS-9, 52068 Aachen, Dennewartsstr. 25-27

This talk will give an overview over recent developments in the field of material informatics and materials data science, in particular over current, state-of-the-art machine learning and data-mining techniques in the context of TEM experiments.

As a main example, the goal is to understand some of the many open questions concerning the underlying structure-property relations in High Entropy Alloys (HEAs). Although in-situ Transmission Electron Microscopy (TEM) allows high-resolution studies of the structure and dynamics of moving dislocations and -- in a way -- makes the local obstacle/energy "landscape" visible through the geometry of dislocations; a 3D analysis and high-throughput data-mining of the resulting data is still not possible.

We introduce a novel data-mining approach that is based on spatio-temporal coarse graining of TEM dislocation data, making ensemble averaging of a large number of snapshots in time possible. Using dislocations as "probes" we investigate the effect of pinning points on the dislocation glide behavior of CoCrFeMnNi alloy during in-situ TEM straining. Additionally, we use our Deep Learning-based dislocation extraction and 3D reconstruction to analyze the strain avalanche statistics of in-situ TEM recordings and discuss the dependency of the power law exponent based on 3D dislocation dynamics simulations.