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

MM: Fachverband Metall- und Materialphysik

MM 4: Data Driven Material Science: Big Data and Workflows I

MM 4.6: Vortrag

Montag, 18. März 2024, 11:45–12:00, C 243

Transferable interatomic potential of water with the atomic cluster expansion — •Eslam Ibrahim, Yury Lysogorskiy, and Ralf Drautz — ICAMS, Ruhr Universität Bochum, 44780 Bochum, Germany

We present a transferable parameterization of water using the Atomic Cluster Expansion (ACE). Our approach efficiently samples liquid water by employing static calculations of various ice phases. The active learning feature of ACE-based D-optimality algorithm is utilized to select relevant water configurations, circumventing computational challenges associated with ab-initio molecular dynamics (AIMD) simulations. Our results demonstrate that ACE descriptors enable a potential fitted solely on ice structures to provide a very good description of liquid water. The developed potential shows remarkable agreement with first-principles references, accurately capturing structural and dynamic properties of liquid water. This includes pair correlation functions, covalent bonding profiles, hydrogen bonding profiles, diffusion coefficient, and thermodynamic properties like the melting point of water. This work introduces an efficient sampling technique for machine learning potentials in water simulations, along with a transferable interatomic potential that rivals the accuracy of ab-initio references. This advancement enhances our understanding of water’s behavior at the atomic level and opens new avenues for studying complex aqueous systems.

Keywords: machine learning potentials; local atomic environments; ab initio molecular dynamics; water and water interfaces