Regensburg 2025 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
O: Fachverband Oberflächenphysik
O 35: Poster Solid-Liquid Interfaces: Structure
O 35.2: Poster
Tuesday, March 18, 2025, 13:30–15:30, P3
High-Dimensional Neural Network Potentials for Molecular Dynamics Simulations of Mineral-Water Interfaces — •Maite Böhm1, 2, Bernadeta Prus1, 2, and Jörg Behler1, 2 — 1Theoretische Chemie II, Ruhr-Universität Bochum, Germany — 2Research Center Chemical Sciences and Sustainability, Research Alliance Ruhr, Germany
In recent years, High-Dimensional Neural Network Potentials (HDNNP), a frequently used type of machine learning potential, have become a popular tool for simulations of complex systems such as mineral-water interfaces. Here, we present a HDNNP trained to density functional theory energies and forces for tricalcium aluminate (Ca3Al2O6, C3A)-water interfaces, which are of high interest for concrete chemistry. After validation, the obtained HDNNP is applied in large-scale molecular dynamics simulations to unravel the interactions of water with this material by computing a series of structural and dynamical properties.
Keywords: Machine Learning Potentials; Neural Networks; interfaces; water; Molecular Dynamics