Green synthesis of Ag/MgO nanocomposites from red pepper seed waste for efficient sunlight-driven diclofenac...

The occurrence of pharmaceutical residues in aquatic environments has become a major environmental concern due to their persistence, low biodegradability, and incomplete removal by conventional wastewater treatment processes. Among these contaminants, diclofenac (DCF) is frequently detected in surface and municipal wastewaters, highlighting the need for efficient and sustainable remediation technologies. In this study, Ag/MgO NCs were green synthesized using Capsicum annuum L. (red pepper) seed extract as a natural reducing and capping agent. Physicochemical characterization by XRD, FTIR, SEM-EDX, UV-Vis DRS, and pH_PZC analyses confirmed the successful formation of the Ag/MgO NCs, with crystallite sizes ranging from 12.0 to 9.0 nm and an extended visible-light absorption edge shifting to an apparent threshold of 1.83 eV due to localized surface plasmon resonance (LSPR) and interfacial states, thereby significantly enhancing solar-light utilization. The photocatalytic performance under natural sunlight was systematically evaluated by investigating the effects of irradiation time, catalyst dosage, Ag loading, solution pH, initial DCF concentration, electrolyte type and concentration, and irradiation source. The 10% Ag/MgO photocatalyst exhibited the highest activity, achieving 80% degradation of 10 mg L^-1 DCF within 180 min under optimum operating conditions. Radical trapping experiments identified hydroxyl radicals (˙OH) as the dominant reactive species, while metallic Ag acted as an efficient electron sink, suppressing charge-carrier recombination and enhancing photocatalytic efficiency. To accurately predict the degradation performance, a Decision Tree coupled with Least-Squares Boosting (DT-LSBoost) model was developed and its hyperparameters were optimized using the Electric Eel Foraging Optimization (EEFO) algorithm. The optimized model demonstrated excellent predictive capability (R = 0.9999; RMSE = 1.2 × 10^-3), while residual analysis confirmed its robustness and generalization ability. A MATLAB-based graphical user interface was further developed to enable rapid prediction of photocatalytic performance under different operating conditions. The integration of green-synthesized Ag/MgO NCs with an EEFO-optimized DT-LSBoost model provides a novel and efficient framework for predicting solar-driven diclofenac degradation. This integrated strategy highlights the potential of combining green nanotechnology with explainable machine learning to support the design, optimization, and practical implementation of sustainable photocatalytic water-treatment systems.