Dresden 2026 – wissenschaftliches Programm

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

MM: Fachverband Metall- und Materialphysik

MM 39: Hydrogen in Materials II

MM 39.5: Vortrag

Freitag, 13. März 2026, 11:15–11:30, SCH/A215

Machine Learning Potential Approach to Complex Hydrogen-Surface Interactions: GRACE Applied to the H/Sn/Ru System — •Manuel Enns and Daniel F. Urban — Fraunhofer IWM, Wöhlerstraße 11, 79108 Freiburg, Germany

Understanding hydrogen interactions with contaminated metal surfaces remains a challenging problem in computational materials science, requiring methods that balance accuracy with computational efficiency to capture complex multi-component dynamics. We demonstrate the application of a GRACE machine learning interatomic potential to study hydrogen behavior on tin-contaminated ruthenium surfaces - a system that exemplifies the challenges of modeling reactive adsorbates on multi-element surfaces. Data from density functional theory calculations on hydrogen penetration mechanisms through ruthenium surfaces with varying tin coverages was used to fine-tune the GRACE potential, enabling computationally efficient molecular dynamics (MD) simulations. The resulting MD simulations successfully capture key dynamic processes including hydrogen diffusion and jump mechanisms, tin growth kinetics at a prior hydrogen coverage, and low-energy hydrogen deposition. This case study demonstrates GRACE's capability to handle complex multi-component systems involving reactive species, providing atomic-level insights into how surface contamination affects hydrogen penetration pathways and adsorbate interactions.

Keywords: Machine Learning Potential; Surfaces; Hydrogen; Surface Contamination; DFT