Neural-network based molecular dynamics simulation of the Fe$_3$O$_4$(001)-water interface Salvatore Romano 1 , Pablo Montero de Hijes 1 , Alexander Gorfer 1,2 , Christoph Dellago 1 1 Faculty of Physics, University of Vienna, Austria, 2 Department of Lithospheric Research, University of Vienna, Austria The behavior of water at the interface with magnetite can be modeled atomistically using first principles 1 . However, such calculations are very demanding. Here, we consider the subsurface cation vacancy model, as it is supported by both experiments 2 and density functional theory energy minimizations 1 . In this work, we run ab initio machine learning simulations of the magnetite(001)/water interface. To do so, we develop a neural-network potential 3 , allowing us to run molecular dynamics simulations of large systems over long simulation times. We study kinetic processes like adsorption and desorption. Moreover, we test the behavior of the subsurface cation vacancy model under full water coverage and explore other possible reconstructions. References 1. Meier, M., Hulva, J., Jakub, Z., Pavelec, J., Setvin, M., Bliem, R., ... & Parkinson, G. S. (2018). Water agglomerates on Fe3O4 (001).Proceedings of the National Academy of Sciences,115(25), E5642-E5650. 2. Bliem, R., McDermott, E., Ferstl, P., Setvin, M., Gamba, O., Pavelec, J., & Parkinson, G. S. (2014). Subsurface cation vacancy stabilization of the magnetite (001) surface.Science,346(6214), 1215-1218. 3. Singraber, A., Behler, J.& Dellago, C. (2019). Library-based LAMMPS implementation of high-dimensional neural network potentials.Journal of chemical theory and computation,15(3), 1827-1840.
P27
© The Author(s), 2023
Made with FlippingBook Learn more on our blog