Abstract

Machine learning models systematically interpolate interatomic potentials from electronic structure calculations, thereby exploiting the smoothness of the Born-Oppenheimer potential energy surface. We can therefore use such models to achieve the accuracy of DFT at a computational cost that is orders of magnitude smaller. I am exploring the application of machine learning potentials to simulate molecular systems, where interactions between molecules, which must be accounted for separately, play a key role in determining the properties of the material. I will present the progress of my work specifically to create a potential for liquid hydrocarbons with applications to fuels and lubricants research.