Dresden 2026 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
O: Fachverband Oberflächenphysik
O 97: Solid-liquid interfaces: Reactions and electrochemistry III
O 97.2: Talk
Friday, March 13, 2026, 10:00–10:15, TRE/PHYS
Mechanism of Fe(II) Chemisorption on Hematite(001) Revealed by Machine Learning Molecular Dynamics — •Kit Joll1, Philipp Schienbein1,2, Kevin Rosso3, and Jochen Blumberger1 — 1Department of Physics and Astronomy and Thomas Young Centre, University College London, London WC1E 6BT, United Kingdom; — 2Ruhr-Universität Bochum, 44780 Bochum, Germany; Research Center Chemical Sciences and Sustainability, Research Alliance Ruhr, 44780 Bochum, Germany — 3Pacific Northwest National Laboratory, Richland, Washington 99354, United States;
Understanding ion adsorption at mineral-water interfaces is key to explaining geochemical processes such as redox cycling, dissolution, and crystal growth. However, probing these reactions experimentally at atomic resolution remains challenging. Here, we employ reactive neural network potential molecular dynamics (NNP-MD) with umbrella sampling and reactive flux transition state theory to simulate the chemisorption of aqueous Fe2+ on hematite(001) at density-functional theory (DFT) accuracy and nanosecond time scales. We identify a dissociative mechanism for Fe2+ water-ligand exchange and multiple stable surface complexes, nonadsorbed, physisorbed, monodentate, and tridentate chemisorbed species, along the adsorption pathway. The overall process is exergonic, in contrast to previous classical force-field predictions. These results demonstrate how machine-learning potentials bridge the accuracy of ab-initio methods with the sampling efficiency of classical molecular dynamics, providing new atomistic insight into reactive processes at oxide-water interfaces.
Keywords: Adsorption; Molecular dynamics; Machine learning potentials; Kinetics; Enhanced Sampling