Integrated RSM-ANN Modelling and Mechanistic Evaluation of Arsenate Adsorption onto Click-Functionalized Magnetic NanoSorbent (M-TACA)

Abstract

A click-functionalized magnetic nano-adsorbent (M-TACA) incorporating N-methyl-D-glucamine (NMDG) ligands was systematically evaluated for arsenate [As(V)] removal using a newly generated multivariate experimental dataset. The adsorption behaviour was modelled using an integrated response surface methodology (RSM) and artificial neural network (ANN) framework to assess the combined effects of initial As(V) concentration, solution pH, contact time, and adsorbent dose. Both modelling approaches demonstrated excellent predictive performance, with coefficients of determination exceeding 0.99 (R-2 > 0.99). Under the RSM-derived optimal conditions (pH 8.0, initial As(V) concentration of 200 mg L-1, contact time of 150 min, and adsorbent dose of 1.5 g L-1), adsorption capacities of 97.3 mg g(-1) (experimental) and 99.8 mg g(-1) (ANN-predicted) were obtained. Mechanistic interpretation based on pH-dependent zeta potential measurements and aqueous arsenate speciation indicated that electrostatic attraction governs As(V) uptake below the point of zero charge (pH(p)zc approximate to 7.7), whereas surface complexation and hydrogen-bonding interactions become increasingly relevant under near-neutral conditions. The presence of NMDG moieties introduces multiple hydroxyl and amine functional groups, enhancing arsenate affinity across a broad pH range and supporting the formation of inner-sphere surface interactions. In comparison with other Fe3O4-based sorbents, M-TACA exhibits a higher adsorption capacity together with a wider operational pH tolerance. This study presents the first multivariate, AI-assisted optimization of a click-functionalized magnetic sorbent for As(V) removal and demonstrates that the hybrid RSM-ANN framework provides improved predictive capability and mechanistic insight for sustainable water treatment applications.