Graphene-Based Membranes for Water Desalination and Gas Separation: A Review of Advances in Molecular Dynamics and Machine Learning Approaches

The search for more efficient routes to clean water and low-carbon gas separations has renewed attention toward graphene-based membranes, particularly as traditional polymeric systems approach their intrinsic performance limits. Graphene and related two-dimensional derivatives provide an unusual combination of atomic-scale thickness, mechanical robustness, and chemically adaptable pore environments, making them promising candidates for applications that require both rapid transport and strict molecular discrimination. Over the past decade, molecular dynamics (MD) simulations have been instrumental in resolving how water and gas molecules interact with graphene pores and layered structures at the atomic level. In parallel, machine-learning (ML) techniques have begun to influence membrane research by assisting in property prediction, guiding design choices, and enabling the exploration of large structural spaces that are otherwise inaccessible through simulations alone. In this review, we draw together recent developments where MD and ML inform one another, with a focus on desalination and gas separation performance, pore-size engineering, chemical functionalization, multilayer configurations, and the influence of operating conditions. Particular attention is given to how ML models can complement MD by identifying structure–property trends and navigating the typical permeability–selectivity constraints faced by membrane materials. The discussion also outlines the present advantages and limitations of MD–ML integration, as well as the key challenges that must be overcome before computational discoveries translate reliably into scalable membrane technologies.