NEUP CINR-21-24582: Machine-Learning-Accelerated Molecular Dynamics Approaches for Molten Salts

We will develop new machine learning potential (MLP) approaches and new MLPs to enable rapid prediction of molten salt (FLiBe and Nal-MgCl2 with impurities) properties with near ab initio quantum mechanical accuracy. Uncertainty quantification with active learning and on-the-fly fitting will greatly accelerate MLP training. This work will support dramatically increased simulation speeds and associated data generation and understanding for molten salts.

Additional Info

Field	Value
Awarded Institution	University of Wisconsin-Madison
PI	Dane Morgan