Regensburg 2022 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
O: Fachverband Oberflächenphysik
O 27: Electron-Driven Processes
O 27.5: Talk
Tuesday, September 6, 2022, 11:30–11:45, H6
Distribution of Charge and Lattice Defects via Machine Learning. — Viktor Birschitzky1, Michael Prezzi1, Marco Corrias1, Lorenzo Papa1, Igor Sokolovic2, Alexander Gorfer1, Martin Setvin2,3, Michael Schmid2, Ulrike Diebold2, Cesare Franchini1,4, and •Michele Reticcioli1 — 1University of Vienna (Austria) — 2Institute of Applied Physics, TU Wien (Austria) — 3Charles University, Prague (Czech Republic) — 4University of Bologna (Italy)
Lattice defects and localized charge on oxide surfaces impact the properties of the material to a different degree depending on their spatial distribution. However, the high number of possible defect configurations poses practical challenges to first-principles studies. Here, we propose a machine-learning-accelerated approach to explore in the framework of density functional theory the spatial configurations of charge and lattice point defects. We apply this approach to analyze the distribution of surface oxygen vacancies on rutile TiO2(110). The attractive interaction with small polarons (electrons localized on the Ti atoms) are revealed to weaken the repulsion between oxygen vacancies, favoring particular arrangements of the vacancies. The resulting distribution can be compared with the patterns identified by computer vision algorithms on scanning-probe microscopy images.