Developing Machine Learning Potential to Unravel Quantum Effect on Ionic Hydration and Transport in Nanoscale Confinement

Developing Machine Learning Potential to Unravel Quantum Effect on Ionic Hydration and Transport in Nanoscale Confinement using Large-scale Simulations This project will investigate the quantum effect on the thermodynamics of ionic selectivity of nanopores. The high selectivity of ions is a long-standing challenge for separations. Nanopores have been considered a solution for this challenge because the nanoscale confinement amplifies the difference between ions. The quantum effect shall play an essential role in determining the non-covalent interactions between ions, water molecules, and nanopores within such a narrow space. However, the classical force fields used for membrane investigation cannot describe the quantum effects precisely. The quantum mechanical calculations can describe the quantum effects well, but it is incredibly time-consuming to use quantum mechanical calculations to investigate the thermodynamics of ionic solvation and selectivity in nanopores. We propose to investigate the quantum effect on ionic solvation and selectivity in nanopores using molecular simulations and machine learning. The goal is to develop a more accurate computational model for investigating and designing nanoporous materials for ion separations. This goal will be achieved based on the development of novel force fields based on a neural network instead of pre-setting functions. The research activities can be divided into two categories. (1) to develop machine learning force fields that can involve quantum effect. The force field will combine long-range electrostatic potential and short-range neural network potential. The short- range neural network potential will involve the quantum effect and (2) to investigate the structure, dynamics, and thermodynamics of ion and water molecules in nanopores using the developed force field. This work will use Li+, Mg2+, Ca2+ as the model ions and carbon nanotubes and graphene slits as the model pores. We hypothesized that a neural network potential trained on a proper size of configuration could well balance the ability of the force field to describe the impact of quantum effect and long-range non-bonded interactions on the structure and dynamics of ions and molecules in nanopores. Based on this hypothesis, we proposed three research tasks: (1) To probe the ability of classical-machine learning hybrid force field to describe thermodynamics and structure of ion solvation in bulk aqueous solution, (2) To investigate the quantum effect on structure and dynamics of water molecules in carbon nanotubes and graphene slits, and (3) To investigate the effect of nanoscale confinement on the free energy of ions to enter and solvate in nanopores using classical-machine learning hybrid force field. We will assess the accuracy of our research by comparing the results with the literature reports of thermodynamics and structures of ions in aqueous solution and nanoscale confinement. The outcome will include (a) a classical- machine learning hybrid force field for ions and water molecules in nanoscale pores and (b) a quantitative or semi-quantitative description of quantum effects on the solvation-transport-free energy relationship for ions confined in nanopores. The primary intellectual merit is to develop tools and harness knowledge. Evidence have shown that the quantum effect plays a vital role in determining the behavior of ions and polarizable molecules confined in a 1-2 nm space. The success of this project will provide new computational tools for investigating the quantum effect in nanoscale confinement and harness knowledge of the quantum effect on thermodynamics and kinetics of ionic selectivity of nanopores. The tools and knowledge will enable a precise design of nanoporous materials for ionic selectivity and other substrates that are hard to be separated.