Expanding global population, industrialization, and climate change have resulted in water shortages, increasing the need for innovative water desalination technologies. The anisotropic physicochemical characteristics and superior properties (e.g., high surface-to-volume ratio and high-water permeability due to atomic thinness) of 2D materials make them attractive for seawater desalination. Graphene, molybdenum disulfide (MoS2), and hexagonal boron nitride (hBN) are emerging as potential 2D materials for water desalination. Experimental studies on using these 2D nanosheets for nanofiltration, reverse osmosis (RO), forward osmosis (FO), and gas separations have been carried out. To support experimental development, classical molecular dynamics (MD) simulations have been used to obtain insights into the design of these membranes. In particular, MD gives valuable information on the rates of transport of molecules and ions through nanopores, helping understand water permeation and salt rejection. As the properties of water in nanoconfinement are different from bulk water, it is important to understand water-nanoporous 2D material interactions at the atomic-level.

Machine-learning (ML) potentials have been explored to describe the interactions of water-2D material systems with the accuracy of ab-initio methods and the efficiency of classical MD[7]. ML potentials are better than their empirical counterparts because of the former’s near-first-principles accuracy.  Such ML potentials have been derived for hBN- and graphene-water systems using neural networks and have been applied to water permeation through nanocapillaries. However, so far, no study has used ML potentials to understand water and ion transport through nanoporous defective 2D materials (which is more realistic).  Given the importance of accurately modeling water- realistic 2D systems, in this work, ML potentials will be developed for defective hBN-water system using density functional theory calculations. After that, these potentials will be used to obtain insights on water permeation through the nanoporous hBN by calculating various properties like friction and diffusion coefficients. Overall, this proposal will lead to the development of highly accurate ML potentials for simulating water- realistic hBN system for applications in membrane separations.