Regensburg 2022 – wissenschaftliches Programm

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

O: Fachverband Oberflächenphysik

O 62: Surface Reactions and Heterogeneous Catalysis 1

O 62.5: Vortrag

Donnerstag, 8. September 2022, 11:30–11:45, H4

Machine-Learning Driven Global Optimization of Surface Adsorbate Geometries — •Hyunwook Jung, Sina Stocker, Karsten Reuter, and Johannes T. Margraf — Fritz-Haber-Institut der MPG, Berlin, Germany

The adsorption energy is an essential descriptor for predicting catalytic activity in theoretical models of heterogeneous catalysis. Although established scaling relations facilitate the prediction of adsorption energies for small adsorbates like OH, they are not applicable to larger adsorbates that are frequently encountered in syngas chemistry. Such systems often feature complex potential energy surfaces due to their flexibility and the possibility of multidentate binding to the surface. Consequently, computing adsorption energies for such adsorbates implies a complex global optimization to find the ground state geometry. This is prohibitively expensive at the density functional theory (DFT) level for routine applications. To tackle this issue, we present a global optimization protocol for adsorbate geometries which trains a surrogate Gaussian Approximation Potential on-the-fly. The approach is applicable to generic surface models (i.e. without defining surface sites) and minimizes both user intervention and the number of DFT calculations by iteratively updating the training set with configurations explored by the algorithm. We demonstrate this approach for diverse adsorbates on the Rh (111) and (211) surfaces.