Green synthesis of Ag/MgO nanocomposites from red pepper seed waste for efficient sunlight-driven diclofenac degradation: experimental investigation and EEFO-optimized DT_LSBoost predictive modeling

Ag/MgO nanocomposites were successfully green-synthesized using red pepper seed extract as reducing and capping agent, with XRD, FTIR, SEM-EDX, UV-Vis DRS, and pH_PZC confirming formation; Ag incorporation decreased crystallite size from 12.0 to 9.0 nm and lowered the apparent visible-light absorption threshold to 1.83 eV. It uses an agricultural byproduct (pepper seeds from harissa production) as the synthesis medium and extends the optical response from pristine MgO's wide 5.65 eV band gap into the visible region, providing a materials basis for solar-driven applications. Multiple complementary characterization techniques (XRD, EDX, SEM including backscattered mode, FTIR, UV-Vis DRS, pH-drift method) corroborate one another and yield quantitative structural parameters such as crystallite size, lattice constant, and cell volume.

Under natural sunlight, 10% Ag/MgO achieved about 80% degradation of 10 mg/L diclofenac within 180 min, outperforming pristine MgO and 2%–8% Ag-loaded samples, and surpassing artificial UV (26 W) and halogen (150 W) sources. It systematically scanned Ag loading, pH, catalyst dosage, initial DCF concentration, KCl/NaCl ionic strength, and irradiation source, identifying 0.5 g/L dosage, natural pH (5.48), and 10 ppm initial concentration as optimal, and provides a comparison table against previously reported catalysts. Based on controlled batch experiments with multi-parameter single-factor investigation, including pseudo-first-order kinetics fitting (k = 0.0084 min⁻¹, R² = 0.961) and TOC mineralization data (53.54% at 180 min), which show a gap between degradation and mineralization.

Radical trapping experiments indicate hydroxyl radicals (˙OH) and superoxide radicals (O₂˙⁻) are the dominant reactive species, with metallic Ag acting as an electron sink that suppresses electron–hole recombination and enhances photocatalytic efficiency. It attributes the enhanced activity of Ag/MgO to synergistic mid-gap states and localized surface plasmon resonance (LSPR) introduced by Ag, and presents stepwise reaction equations describing ROS generation and DCF oxidation pathways. Reactive species contributions are inferred from comparative experiments with four scavengers (T-BuOH, benzoquinone, ammonium oxalate, AgNO₃); this is inferential mechanistic evidence without direct quantitative radical measurements such as transient spectroscopy.

An EEFO-optimized DT_LSBoost model was built, achieving R = 0.9999 and RMSE around 1.2 × 10⁻³ for predicting normalized residual concentration C/C₀ across training, testing, validation, and combined datasets, with a MATLAB graphical interface for rapid prediction. It combines a population-based metaheuristic (EEFO) with ensemble learning (DT_LSBoost) for photocatalytic degradation prediction and packages the model as a decision-support tool requiring no direct code interaction, extending the usability of data-driven methods in photocatalytic process optimization. Reports residual diagnostics over 264 samples (247 residuals within −8 × 10⁻³ to 6 × 10⁻³), a residual histogram, and residual distributions across three subsets, supporting internal consistency of model stability and generalization.

This work addresses a laboratory-scale borosilicate beaker system with 10 mg/L diclofenac aqueous solution, 0.5 g/L catalyst dosage, and clear-sky summer sunlight in Constantine, Algeria (UV index 8–9, average irradiance 850 ± 50 W/m²); its conclusions apply to Ag/MgO photocatalytic degradation prediction and parameter screening within this operating domain, and serve as a starting point for subsequent pilot-scale scale-up, real wastewater characterization, and economic and environmental impact assessment.

Readers may continue to watch: degradation intermediates and transformation pathways remain unidentified, and the gap between mineralization (53.54%) and degradation (80%) raises questions about intermediate fate; the influence of competing organics and ions in real wastewater matrices; the mechanism of activity loss and long-term stability after five reuse cycles; and the model's generalization beyond the trained operating domain. In addition, although this is a full-text parse, some equations and tables appear as images, so specific numerical details should be checked against the original figures and tables.