Accuracy of molecular dynamics simulations depends crucially on the interatomic potential used to generate forces. The gold standard would be first-principles quantum mechanics (QM) calculations, but these become prohibitively expensive at large simulation scales. Machine learning (ML) based potentials aim for faithful emulation QM drastically reduced computational cost. accuracy and robustness...

Marc J. Cawkwell, Anders M. N. Niklasson, T-1 Molecular dynamics (MD) simulations are used heavily in materials science, chemistry, and biology to study the evolution of structures, defects, and non-equilibrium phenomena at the atomic scale. In an MD simulation atoms move over a number of finite time steps according to the force acting on them. These forces are computed from the interatomic pot...