Simultaneous Entropy, Enthalpy and Free Energy Prediction using a Physics-Informed Neural Network and Multi-task Learning
Daniel J. Fowles, Rose G. McHardy, Abdullah Ahmad, and David S. Palmer
Department of Pure and Applied Chemistry, University of Strathclyde, Thomas Graham Building, 295 Cathedral Street, Glasgow G1 1XL, Scotland, UK
The One-Dimensional Reference Interaction Site Model (1D-RISM) is a statistical mechanics-based method for modelling molecular solutions that is computationally inexpensive but is too inaccurate for routine use in its common form. By replacing an end-point free energy functional in 1D-RISM with a deep learning model trained using multi-task learning, we show that predictions approaching chemical accuracy can be obtained for entropy, enthalpy and free energy of solvation. The generalisability of the model is demonstrated by accurate predictions for neutral and ionised solutes in aqueous and non-aqueous solvents at a wide-range of temperatures. The method has been implemented in our in-house software package (pyRISM), which is made freely available to the community.
