The thermal stability of polyethylene terephthalate (PET) microplastics (MPs) is crucial in diverse applications, necessitating a comprehensive understanding of their behavior under high temperatures. This study explores the impact of chemical modifications and aging on the thermal degradation of different PET MPs, including pristine (Pr-PET), acid-modified (Mod-PET), and oxidatively aged (Ag-PET), using kinetics, thermodynamics, and machine learning algorithms using non-isoconversional thermogravimetric analysis (TGA) data. The calculated activation energy (E) and pre-exponential factor (A) for Coats-Redfern model showed Pr-PET with highest E value of 21.09 and 15.24 kJ/mol for zero-order and first order kinetic model, indicating its increased energy requirement for chemical reactions, while Mod-PET and Ag-PET demonstrated improved thermal resistance with lower E values. Thermodynamic data from kinetic parameters showed positive values of Gibbs free energy (ΔG°), enthalpy (ΔH°) and negative values of entropy (ΔS°) (−0.17. kJ/mol·K to −0.20 kJ/mol·K) indicated non-spontaneous degradation process, necessitated energy input, and reduced disorder with decreased reactivity. For accurate predictions of thermal behavior, various machine learning algorithms, including Support Vector Machine (SVM), Decision Tree Regressor (DTM), Random Forest Regressor (RFM), and Artificial Neural Network (ANN), were utilized. The RFM consistently outperformed other models, exhibiting the lowest error metrics and providing reliable predictions of weight loss and derivative thermogravimetric (DTG) for different PET MPs with a predicted-R2 of 0.999. This study sheds light on the thermal stability of PET MPs, considering chemical modifications and aging. The integration of these methods offers a powerful approach to understanding and predicting PET MPs under thermal conditions.
Read full abstract