Evidence Boundaries and Scale-Up for Tapered Graphene Desalination: A Mechanism-to-Decision Synthesis

Abstract

A tapered graphene channel may perform convincingly in simulation or a controlled test while remaining poorly characterized as a manufacturable desalination system. This review maps the boundaries between molecular transport evidence, pore fabrication, membrane assembly, module hydraulics, feed variability, maintenance, and water-quality decisions. It examines physics-informed AI, surrogate modelling, and domain adaptation as tools for scale-up, but requires each transfer to retain its assumptions, uncertainty, and comparator. Out-of-distribution detection identifies geometry, chemistry, or operating regimes beyond the training evidence; causal and sensitivity analysis test whether the model relies on defensible mechanism rather than incidental correlation. A digital provenance thread connects simulations and specimens to models, validation results, and release decisions. No new performance figure is inferred from the cited studies. The article's contribution is a mechanism-to-decision scale-up framework that uses AI to integrate evidence while making clear where laboratory success ends, what uncertainties remain, and who is accountable for authorizing the next level of use.