Chemometric modeling of the adsorption mechanism of Cu(II) in aqueous solution onto functionalized materials: Integrating artificial neural networks and porous structure characterization

Abstract

Traditional methods for evaluating adsorption mechanisms rely on material characterization and its linear relationship with adsorption capacity. However, this approach has limitations, as it assumes a linear correlation, and when this fails, it is often speculated that multiple mechanisms are involved without detailing their contributions. This study overcomes these challenges by using artificial intelligence to analyze the adsorption of Cu(II) onto alternative adsorbents. An Artificial Neural Network (ANN) combined with 3D porous texture simulations, based on mercury intrusion porosimetry, established non-linear correlations among 13 textural and chemical characteristics and adsorption capacity.

The material with the highest adsorption capacity (107 mg g⁻¹) featured an accessible porous texture rich in –COOH groups. The ANN quantified the contributions of two governing mechanisms: diffusion through the porous texture (67.07 %) and interaction with –COOH sites (32.93 %). Chemometric analysis revealed that the greatest weight in the ANN model was attributed to the average pore diameter (17.11 %), which was consistent with the characterization of the saturated material by SEM-EDX, showing that adsorption occurs primarily in the exposed cavities of the material.

The adsorption mechanism proposed by the ANN study explains the atypical points observed in the different materials, showing that the adsorption process is governed by a combination of two mechanisms: one associated with the porous texture and the other with surface chemistry. The findings provide a deeper understanding of the key variables influencing adsorption and offer guidance for optimizing material synthesis.