Prediction of energy, force and response properties of metal nanoclusters using machine learning techniques

Abstract

Calculating potential energy using density functional theory (DFT) is computationally very expensive. The complexity escalates significantly when determining the first-order derivative of energy with respect to cartesian coordinates and the second-order derivatives of the energy with respect to the external electric field, which is essential for predicting forces and polarizability in metallic nanoclusters. Since force calculations are essential for running molecular simulations, such as molecular dynamics (MD) simulations, the use of DFT-based methods becomes impractical for studying larger metallic nanoclusters. Empirical and semi-empirical methods are widely used for modeling atomic systems. However, they often lack the accuracy needed for complex systems, failing to capture intricate electronic interactions.