Machine Learning and Quantum-Inspired Optimization of Microwave-Synthesized High-Entropy Alloy Magnetic Nanoparticles for Photocatalytic Diazinon Degradation

This study reports the synthesis of high-entropy alloy-functionalized magnetic nanoparticles (HEA@MNP) photocatalysts and investigates their application for the photocatalytic degradation of diazinon pesticide under visible LED irradiation. The core-shell nanoarchitecture of Fe₃O₄ magnetic core and Co-Ni-Cu-Zn-Mn HEA-like shell was prepared using a co-precipitation method combined with a microwave-assisted functionalization technique. FTIR, SEM-EDX, PXRD, and TG did characterization of HEA@MNP. The HEA@MNP photocatalysts exhibited excellent photocatalytic activity with a high degradation efficiency of 97.4% for diazinon, good adsorption capacity of 169.5 mg/g, and excellent reusability retaining 93% retained efficiency after seven cycles without noticeable emission of heavy metal ions. For process modeling and optimization, a full-fledged machine learning framework with four algorithms (LSTM-XGBoost, Gaussian Process, Polynomial Ensemble, and Ultimate Ensemble) was established. Among the evaluated models, the Ultimate Ensemble achieved the highest test R² of 0.936; however, overfitting analysis (ΔRMSE% = 49.5%) revealed moderate model instability. The Polynomial Ensemble demonstrated the most robust generalization performance (ΔRMSE% = 7.7%, Generalization Score = 0.969), suggesting it may be more reliable for practical applications despite slightly lower test metrics. For the multi-objective optimization, a new quantum-inspired genetic algorithm (QGA) was designed, and it has proven to be very effective in the solution processes with computational efficiency compared to the traditional methods (31 iterations are needed to converge instead of 300). Application of the QGA optimization approach enabled a 21.6% reduction in catalyst loading while sustaining a degradation efficiency of 86.48%, thereby enhancing the overall cost-effectiveness of the process by approximately 28.5%, Herein, photocatalytic scavenging experiments were carried out and showed that holes generated during the irradiation process are responsible for the predominant oxidation, and hydroxyl/superoxide radicals are the secondary oxidants. This combination of advanced materials, machine learning, and quantum-inspired optimization represents an elegant platform for sustainable water treatment technologies.