Co-TPA@g-C₃N₄ nanocomposites for visible light-induced photocatalytic degradation: Synthesis and optimization using RSM, ANN and ANFIS

Heteropolyacids (HPAs) are known for their exceptional electron storage and charge-separation capabilities in photocatalysis. While traditionally used in homogeneous systems due to their polar solubility, immobilizing HPAs on high-surface-area semiconductors enhances their utility in heterogeneous photocatalysis by improving light absorption, charge transfer and recyclability. In this study, graphitic carbon nitride (g-C3N4) was modified with varying concentrations (10 %, 20 % and 30 %) of cobalt-exchanged tungstophosphoric acid (Co-TPA) to synthesize Co-TPA@g-C3N4 nanocomposites. They were characterized by FT-IR, XRD, SEM-EDS, BET, UV-Vis and PL analyses, and tested for photocatalytic degradation of methylene blue (MB) under visible light. Among the synthesized composites, the 10% Co-TPA@g-C3N4 nanocomposite had the smallest particle size (142.3 nm) and the highest surface area (61.57 m(2)/g) with SEM images confirming a smoother, cohesive layered structure, indicating better dispersion of Co-TPA. Its band gap (2.45 eV) enabled better visible light absorption, while low PL intensity indicated improved charge separation. This sample also achieved the highest degradation efficiency (92.36 %) under optimized conditions (50 mg catalyst, 10 ppm dye and 120 min). Kinetic analysis revealed that pseudo-first-order behavior with a rate constant of 0.014 min(-1), and energy efficiency evaluation confirmed the lowest EEO value (540.6 kWh/m(3)order). Furthermore, this nanocomposite retained over 87.57 % of its activity after four reuse cycles, affirming its stability and reusability. Among the predictive modeling techniques evaluated, the Adaptive Neuro-Fuzzy Inference System (ANFIS) demonstrated superior performance over Artificial Neural Networks (ANN) and Response Surface Methodology (RSM), achieving an R-2 of 0.998 and an MAE of 0.296. These results highlight ANFIS's enhanced capability to model complex, nonlinear relationships between process parameters and photocatalytic performance.