Advanced learning strategies for machine learned interatomic potentials

In this project the application of machine learned interatomic potentials (MLIPs) as surrogate models for molecular dynamics (MD) simulations with ab-initio accuracy for room temperature ionic liquids is investigated. This will be accomplished by generating training data using density functional theory followed by fitting MLIPs and the deployment of such in MD simulations. This project also involves the development of software packages for the design and management of the MLIP workflow. These tools involve workflow, data and experiment management in general (ZnTrack), specialized in the MLIP development (MLSuite) and post processing the MD simulations (MDSuite).