Regensburg 2025 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

MM: Fachverband Metall- und Materialphysik

MM 17: Development of Calculation Methods

MM 17.3: Talk

Wednesday, March 19, 2025, 10:45–11:00, H22

Development of on-the-fly kinetic Monte Carlo framework with neural network potentials for surface chemistry — •Tomoko Yokaichiya, Tatsushi Ikeda, Koki Muraoka, and Akira Nakayama — The University of Tokyo, Tokyo, Japan

It is established that adsorbate-adsorbate lateral interactions in heterogeneous catalysis significantly influence adsorption and activation energies. To further our understanding of the degree to which such lateral interactions affect catalytic properties, we develop an "on-the-fly" adaptive kinetic Monte Carlo (kMC) simulation scheme. The scheme proceeds by energetically evaluating each configuration, including its local adsorbate-rich environment, using a neural-network potential then stores them to a database for efficient reuse in later kMC iterations. We apply this scheme to the industry-relevant interactions of H adsorption and diffusion on Pd and Pt surfaces as well as CO oxidation on Pt surfaces. With this scheme, we are able to elicit the extent lateral adsorbate-adsorbate interactions influence surface reactions and diffusion.

Keywords: Kinetic Monte Carlo simulation; Neural network potential; Lateral interaction; Hydrogen diffusion; CO oxidation