Determining Ion Activity Coefficients in Ion-Exchange Membranes with Machine Learning and Molecular Dynamics Simulations

The activity coefficients of ions in polymeric ion-exchangemembranes(IEMs) dictate the equilibrium partitioning coefficient of the ionsbetween the membrane and the liquid. It also affects ion transportprocesses, such as conductivity, in ion-exchange membranes. Accuratelypredicting the ion activity coefficient without experimental datahas been elusive as most models are empirical or semi-empirical. Thiswork employs an embedding process that maps microscopic and macroscopicproperties for modeling of ion activity coefficients in IEMs withmolecular dynamics (MD) and machine learning (ML). This strategy iseffective for accurately predicting activity coefficients in variousIEM materials, including random copolymer and block copolymer systems.ML algorithms are increasingly being used for the analysis of complexsystems when limited knowledge is available. The framework uses smallexperimental activity coefficient datasets in conjunction with polymerstructure information and molecular attributes describing the solvationof ions and polymers to predict the ion activity coefficient in IEMs.Two different ML models were developed to estimate the molecular attributesand the ion activity coefficient. The best ML model accurately predictsthe solvation descriptors and ion activity coefficient with an averagemean absolute error of <7 and 10%, respectively. Adopting the saidapproach allows for the estimation of ion activity coefficients inIEMs without the need for new time-consuming MD simulation runs andexperiments.