Machine learning approach for ion imprinted (IIP) and non-imprinted (NIP) polymer discrimination based on pyrolysis kinetic data

Abstract

The improper disposal of effluents containing toxic elements, such as nickel and mercury, can result in the gradual deterioration of surface water quality, impacting aquatic ecosystems and human health. Ionically imprinted polymers (IIPs) are promising candidates for remediating toxic areas due to their high removal rates and selectivity. In this study, multifunctional IIPs were developed using Ni2+ and Hg2+ as template ions and dithizone, methacrylic acid, and ethylene glycol dimethacrylate as complexing agent, monomer, and cross-linking agent, respectively, to remove Ni (II) and Hg (II) from effluents. The thermal decomposition behavior and degradation mechanism of ion imprinted polymers were investigated to address the issue of limited lifespan. The kinetic study revealed that the samples showed significant thermal stability, and the IIP-Duo sample exhibited higher activation energy compared to the non-imprinted polymer (NIP) and the mono-imprinted polymer (IIP), indicating that double impression alters the structure of the polymer and consequently its thermal degradation. Three machine learning models were applied to classify the samples according to imprinting, successfully predicting 95% or more of the classes correctly without overfitting or overmodelling. The master plot method showed that imprinting did not affect the thermal decomposition pathway, with the IIP, IIP-Duo, and NIP samples presenting the same mechanism geometrical contraction model (R3) at a conversion rate between 0.4 and 0.9. These findings suggest that multifunctional IIPs are promising materials for remediating areas impacted by toxic elements.