Predicting the properties of salt water using neural network potentials and continuum solvent theory

Electrolyte solutions play a fundamental role in a vast range of important industrial and biological applications. Yet their thermodynamic and kinetic properties still can not be predicted from first principles. There are three central challenges that need to be overcome to achieve this. Firstly, the dynamic nature of these solutions requires long time scale simulations. Secondly, the long-range Coulomb interactions require large spatial scales. Thirdly, the short-range quantum mechanical (QM) interactions require an expensive level of QM theory. Here, we demonstrate a methodology to address these challenges. Data from a short \emph{ab initio} molecular dynamics (AIMD) simulation of aqueous sodium chloride is used to train an equivariant graph neural network interatomic potential (NNP) that can reliably reproduce the short-range QM forces and energies at a moderate computational cost. This NNP is combined with a continuum solvent description of the long-range electrostatic interactions to enable stable long time and large spatial scale simulations. From these simulations, ion-water and ion-ion radial distribution functions (RDFs), as well as ionic diffusivities, can be determined. The ion-ion RDFs are then used in a continuum solvent approach to calculate the osmotic and activity coefficients. Good experimental agreement is demonstrated up to the solubility limit of sodium chloride in water. This result implies that classical electrostatic theory can describe electrolyte solution over a remarkably wide concentration range as long as it is combined with an accurate description of the short-range interactions. This approach should be applicable to determine the thermodynamic and kinetic properties of many important electrolyte solutions for which experimental data is insufficient.