We have recently developed a machine learning potential with the accuracy of DFT but a cost similar to a force field. With this potential, which has analytical forces, we can optimize geometries and perfom molecular dynamics, from small gas phase molecules to reactions in condensed phases. We will show that using an Active Learning methodology, we can see reactions happen in real time, from carbon structures (tubs and balls), to the products of a primordial soup run.

Molecular dynamics with machine learning potentials. From gas phase to solution chemistry, at low cost and high accuracy

255th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.