Graphene and graphene oxide-enabled PVDF Janus membranes: Design, fabrication, and first principles-machine-learning insights for water desalination

Membrane distillation is often constrained by low permeate flux and long-term instability caused by membrane wetting, fouling, and scaling. Herein, we report Janus PVDF membranes with asymmetric functionalization by nonsolvent-induced phase separation (NIPS), incorporating graphene (G) on the feed-facing surface and graphene oxide (GO) on the permeate-facing surface to induce a controlled wettability gradient. The hydrophobic G layer enhances liquid entry pressure (LEP) and wetting resistance, while the oxygen rich GO layer imparts hydrophilicity and promotes vapor condensation on the permeate side. This functional asymmetry results in a mechanically stable Janus architecture with pronounced wettability contrast, achieving a 194-220% increase in permeate flux compared to a commercial PVDF membrane while maintaining 99.97-99.99% NaCl rejection. The optimized membrane demonstrates strong resistance to wetting and fouling when treating hypersaline and chemically complex feedwaters containing up to 70,000 ppm NaCl, surfactants, divalent salts, and humic substances. To elucidate the underlying mechanisms, Hansen solubility parameters, DLVO theory, and density functional theory (DFT) were employed to analyze polymer-filler interactions and interfacial stability. The results reveal that van der Waals interactions and hydrogen bonding govern the compatibility and stability of the PVDF-G/GO system, consistent with observed wettability and LEP trends. In addition, machine learning models correlate membrane structure and operating conditions with performance, with an artificial neural network achieving an R² of 0.984. These findings provide mechanistic insight into Janus membrane behavior and highlight the potential of asymmetric PVDF-based membranes for advanced desalination applications.