Interfacial Mechanisms of Reactive Dye Adsorption on a Waste-Cotton Derived PVA/TiO2/Cellulose Biofilm: A Molecular Dynamics and Physics-Informed Predictive Framework

Résumé fourni par la source

Textile dye effluents remain a major source of water pollution, and adsorptive bio-based composite films are attractive for their low cost and environmental footprint. This study identifies the interfacial mechanism governing reactive dye removal by a poly(vinyl alcohol) (PVA)/TiO2/microcrystalline cellulose (MCC) biofilm, in which the MCC was extracted from waste cotton fabric (WCF) and citric acid served as the crosslinker, and shows that this mechanism can be captured in a predictive model. All-atom molecular dynamics simulations of the PVA–MCC–citric acid–TiO2dye system resolve an electrostatically dominated oxide interface together with an extensive PVA hydrogen-bond and hydrophobic-contact network that also explains the composite’s elevated glass-transition temperature; combined with the post-adsorption S=O infrared response and the TiO2 point of zero charge, this identifies attraction between the protonated surface and the dye’s sulfonate groups as the leading interaction. The resulting chemisorption-dominated picture is corroborated by two further approaches: classical kinetic and equilibrium analysis, best described by the pseudo-second-order (R2 = 0.998) and Freundlich (R2 = 0.978) models; and a physics-informed neural network constrained by the pseudo-second-order rate law, which achieved the strongest generalization among six benchmarked machine-learning models (leave-one-out R2 = 0.989, MAE = 0.021 mg/g). A four-parameter unified surface coupling the Freundlich isotherm and pseudo-second-order kinetics through a mass balance reproduces both datasets (R2 = 0.962 and 0.992) across dosages differing by a factor of 2.4, enabling prediction of performance under untested operating conditions. FTIR, XRD, particle-size analysis, FE-SEM/EDX, mechanical testing, DSC and swelling studies confirm composite formation and ester crosslinking, and the biofilm removed up to 49% of the reactive dye Avitera Light Red SE at 50 ppm within 170 minutes. Together these analyses offer a transferable framework for coupling molecular simulation with physics-informed machine learning in the study of bio-based sorbents.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.