Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
O: Fachverband Oberflächenphysik
O 85: New Methods: Theory
O 85.2: Vortrag
Donnerstag, 20. März 2025, 10:45–11:00, H25
Efficient implementation of charge-equilibration schemes for fourth-generation machine learning potentials — •Moritz R. Schäfer1,2, Moritz Gubler3, Stefan Goedecker3, and Jörg Behler1,2 — 1Theoretische Chemie II, Ruhr-Universität Bochum, Germany — 2Research Center Chemical Sciences and Sustainability, Research Alliance Ruhr, Germany — 3Department of Physics, University of Basel, Switzerland
Fourth-generation high-dimensional neural network potentials (4G-HDNNPs) are a modern technique to compute energies and forces with close to ab initio accuracy for conducting extensive molecular dynamics simulations of complex systems. They are based on global information and include long-range charge transfer, electrostatics and atomic energies to describe the interactions in a system. A central component of 4G-HDNNPs is a charge equilibration (Qeq) step, which due to its non-local nature dominates the computational costs. Here, we discuss efficient implementation strategies, and show their performance on selected benchmark systems.
Keywords: Machine Learning Potentials; HDNNPs; Simulation; MD