Classical and machine-learning-based molecular dynamics
Title: Classical and machine-learning-based molecular dynamics
DNr: LiU-gpu-2023-2
Project Type: LiU Compute
Principal Investigator: Davide Sangiovanni <>
Affiliation: Linköpings universitet
Duration: 2023-04-12 – 2024-05-01
Classification: 10304


In this project, GPUs will be used for running molecular dynamics simulations (MD) based on the MEAM potential as well as machine-learning interatomic potentials (MLIP). We will train and test two MLIP implementations : the Moment Tensor Potentials (MTP developed by Alexander Shapeev) and the Atomic Cluster Expansion (ACE developed by Ralf Drautz). The ACE method has been implemented to run in parallel on GPUs. All MD simulations will be carried out using LAMMPS compiled for GPU's and LAMMPS interfaced with MTP and ACE. The simulations will focus on modeling materials subject to load at various temperatures as well as the properties of extended crystallographic defects.