Data-Driven Design of Nanopore Graphene for Water Desalination

Development of energy-efficient and low-cost desalination techniques is of pivotal importance, and reverse osmosis (RO) is regarded as one of the most promising solutions to tackle the world water crisis and has been widely deployed for large-scale and distributed water desalination. Graphene with nanopores was considered as a promising desalination membrane due to its unique properties. However, the intrinsic complexity of the desalination process, together with the various tunable properties of the membranes/nanopores themselves, makes accurate prediction of the performance or designing of new materials challenging. Machine learning (ML) techniques are superior in analyzing physical processes from multiple aspects, which could facilitate the rational design of high-performance desalination membranes. In this work, it was discovered that salt rejection mainly depends on the pore shape, pore area, and applied pressure and that water permeation mainly depends on the pore area and applied pressure from the ML study. The physical–chemical analysis based on the ion density and water density along the nanopore offers us a deep understanding of the effect of the pore shape on salt rejection and water permeation. In light of the results of ML and the analysis of physicochemical properties, we design the graphene pore with a particular pore shape, which could achieve high water permeation with high salt rejection. ML combined with high-throughput computation methods could help us design the material with excellent performance for desalination.