Emerging iron based porous metallopolymeric material with cross-linked networks for the separation of ultra-trace arsenic from aqueous environment and simulation with artificial neural network

Abstract

Numerous health problems caused by the aquatic environment's extreme arsenic poisoning influence millions of people, and in the case of prolong indigestion of arsenic containing drinking water, can be potentially fatal. Herein, a new porous polymeric network knitted with integrated iron moiety i.e., poly(ferric tri methacrylate) (pFeM) is fabricated via the suspension polymerization technique. The relative reactants i.e., monomer, crosslinker, and porogen were varied to get a highly efficient pFeM adsorbent for arsenic. The prepared pFeM revealed a significant affinity for arsenic owing to a cohesive iron moiety in the polymeric chain. The prepared adsorbent was characterized by instrumental techniques such as, SEM, XRD, FTIR, BET, and XPS. The pFeM has exhibited high adsorption capacities (q_{e max exp}) of 41.39 mg g⁻¹ for As(V) and 37.35 mg g⁻¹ for As(III), which is close to the adsorption capacities (q_{e max theo}) of 45.10 mg g⁻¹ for As(V) and 40.88 mg g⁻¹ for As(III) achieved by Langmuir adsorption model. The high adsorption capacities of pFeM might be owing to its unique architecture of integrated iron with porous texture surface area (SA: 197 m² g⁻¹) and porosity (Pore size: 0.93–2.36 µm). The pFeM exhibited high arsenic adsorption capacity (>94 % for both forms of arsenic) in the neutral water pH range of 6.0 to 8.0. The phosphate (PO₄³⁻), and bicarbonate (HCO₃⁻) ions negatively affect the adsorption of arsenic, as demonstrated by the effect of interfering ions in water. The pseudo-second-order kinetic model is best fitted with a high correlation coefficient (R²) of 0.9918 and 0.9789 for As(V) and As(III), respectively. The artificial neural network (ANN) was trained and validated using a three-layer back propagation network with the ideal structure of 4–10–1. The values of correlation coefficients (R > 0.99) reveal the high accuracy of the ANN model.