AI-Optimized Functional Nanocomposite for Soil and Water Remediation with Integrated Crop and Microbial Safety Analysis

Uranium along with other pollutants often found together in water and soil creates serious risks for both ecosystems and human health. But traditional remediation methods are not very selective and they tend to produce secondary waste . Here, we employ a machine learning-informed approach to design a GO-biochar-PEI (GBP) functional nanocomposite for multi-pollutant removal. Random Forest model trained on 150 engineered formulations identifies the optimal composite, which integrates GO for high surface area and reactive sites, biochar for structural porosity, and PEI for amine functionality. SEM analysis reveals micron-scale aggregates (2-14 µm) composed of intimately mixed nano-components, while BET analysis confirms a high specific surface area (300-500 m²/g). This ternary nanocomposite also catalytically reduces U(VI) to less mobile U(IV) via electron transfer from GO and PEI amine groups, establishing its triple-catalyst functionality. The optimized GBP exhibits a uranium adsorption capacity of 543.5 mg/g (Langmuir isotherm, R² ≥ 0.991) and follows pseudo-second-order kinetics (R² = 0.999), confirming monolayer chemisorption. Thermodynamic analysis reveals endothermic (ΔH° = +60.57 kJ/mol) and spontaneous (ΔG° < 0) adsorption. Under environmentally relevant conditions (pH 6, 60 °C), GBP achieves 120 mg/g uptake and outperforms GO-PEI and biochar-PEI, with 87.5% removal in groundwater, 84.0% in river water, and 81.2% in soil. Leaching tests show minimal uranium release (3.2%), indicating low environmental risk. Additionally, GBP significantly reduces uranium uptake (30–34%) and translocation (TF < 1) in brassica and sunflower, while enhancing plant physiological and antioxidant defenses. This AI-guided framework links nanoscale properties to remediation performance and ecological safety, providing a scalable strategy for sustainable environmental remediation of radionuclides, heavy metals, and emerging contaminants across complex matrices.