The static and dynamic adsorptive performance of a nitrogen and sulfur functionalized 3D chitosan sponge for mercury and its machine learning evaluation

Abstract

The use of chitosan-based sponge materials for Hg(II) removal has gained attention recently due to their effectiveness. However, the complex preparation, limited performance, and poor acid resistance remained major drawbacks. Herein, a nitrogen‑sulfur functionalized macroporous chitosan sponge was successfully synthesized via two mild amidation reactions and exhibited abundant interconnected mesopores. These features endowed the functionalized chitosan-based sponge with high adsorption capacity (1227.15 mg g⁻¹), fast reaction rate (8.27 × 10⁻³ g mg⁻¹·min⁻¹), broad pH adaptability (1–7), and high selectivity, even in the artificial chlor-alkali wastewater. Furthermore, the impressive saturation capacity of 1329.24 mg g⁻¹ was achieved in various heights and injection rates in the fixed-bed column test, and the good removal efficiency (>85 %) was maintained after six dynamic regeneration cycles. The excellent performance was primarily attributed to the chemisorption of CS groups. Among the three machine learning models, the ANFIS algorithm owned the best results of the smallest RMSE (0.00315) and highest R² (0.9752) for predicting dynamic adsorptive behaviors. Overall, this research provided a reference for preparing a promising mesoporous sponge as an alternative recyclable and efficient candidate for industrial wastewater treatment and offered a machine learning model to predict the dynamic adsorptive performance.