NEUP CINR-21-24312: Accelerating the development of reliable and robust machine learning-based interatomic potentials for the prediction of molten salt structure and properties

Machine learning-based interatomic potentials (MLIPs) used in molecular dynamics (MD) can accurately and efficiently predict molten salt properties. However many machine learning-based methods require large training sets, and can fail unpredictably. This project will overcome these challenges by developing a method for efficiently sampling diverse configurations from MD to train reliable and robust neural network potentials, and develop new models for predicting errors in MLIPs.

Additional Info

Field	Value
Awarded Institution	University of Massachusetts, Lowell
PI	Stephen Lam