Interpretable machine learning and greenness assessment for sustainable tetracycline adsorption and phytotoxicity...

This study presents, the conversion of almond peel waste into activated carbon (APAC) and its use for removing tetracycline (TC) antibiotic from water. The prepared APAC possesses a rough, highly porous, and mesoporous carbon structure with a high surface area, as revealed by FESEM and BET analysis. FTIR, XRD, and XPS results showed the presence of oxygenated and phosphorus-containing surface functionalities and a predominantly amorphous carbon framework, which favored effective TC adsorption. Adsorption experiments indicated that TC removal was best described by the Elovich kinetic model (R² = 0.9999), while equilibrium data fit the Freundlich isotherm model (R² = 0.9961), suggesting adsorption on a heterogeneous surface. The maximum adsorption capacity obtained from the Langmuir model was 87.91 mg/g. Thermodynamic results confirmed that the process was spontaneous and exothermic, with a ΔH° value of -14.76 kJ/mol. After 3 reuse cycles, APAC retained approximately 76.65% of its initial removal efficiency and showed reliable TC removal in real water samples. Phytotoxicity tests using Brassica juncea demonstrated that APAC treatment significantly reduced the toxic effects of TC on plant growth. Machine learning models developed from 180 experimental data points showed excellent prediction accuracy, with CatBoost giving the best performance (Test R² = 0.9853). SHAP assessment of the optimized CatBoost model identified TC concentration and time as the dominant predictors. Greenness and applicability assessment gave AGREE, ComplexMoGAPI, and BAGI scores of 0.53, 73, and 67.5, respectively, indicating moderate overall sustainability and practical applicability with scope for improvement. Overall, APAC shows promise as an effective biomass derived adsorbent for TC removal from wastewater.