Design and optimization of phosphorus- and titanium-doped graphitic carbon nitride photocatalysts for the degradation of Beibrich Scarlet dye under visible light irradiation using machine learning
Kamilia Madi, Derradji Chebli, Oumnia Rayane Benkouachi, Reguia Boudra, Rachida Bouallouche, Hichem Tahraoui, Mohammed Kebir, A. Gil, Jie Zhang, Abdeltif Amrane
Abstract
Most traditional photocatalysts show poor performance under visible light. Moreover, optimizing experimental conditions is often repetitive and time-consuming. This study presents the synthesis, characterization, and photocatalytic evaluation of a novel composite P-g-C₃N₄@Ti-g-C₃N₄ designed for efficient degradation of Beibrich Scarlet (BS) dye under visible light irradiation. The materials were prepared by a simple polymerization process followed by physical mixing. Structural and morphological features were analyzed using Fourier-transform infrared spectroscopy (FTIR), X-ray diffraction (XRD), scanning electron microscopy coupled with energy dispersive X-ray spectroscopy (SEM-EDX), transmission electron microscopy (TEM), ultraviolet–visible diffuse reflectance spectroscopy (UV-DRS), and photoluminescence spectroscopy (PL), confirming successful phosphorus and titanium doping, improved crystallinity, reduced optical bandgap (2.56 eV), and enhanced charge separation. The composite demonstrated superior photocatalytic performance, achieving complete degradation of BS (30 mg/L) within 100 min under optimal conditions (pH 6.1, 1 g/L catalyst dose), following pseudo-first-order kinetics (k = 0.0312 min −1 ). Radical trapping experiments indicated hydroxyl radicals (•OH) as the dominant reactive species, followed by photogenerated holes (h + ) and superoxide radicals (O₂• − ). A machine learning model based on a decision tree with least squares boosting (DT_LSBOOST) accurately predicted the degradation efficiency with excellent correlation ( R = 0.9999) and minimal error (root mean square error, RMSE <0.005). Optimization using the Improved Grey Wolf Optimizer (IGWO) enabled rapid identification of ideal operational parameters. This work highlights a powerful and intelligent approach combining material engineering, visible-light-driven photocatalysis, and artificial intelligence (AI)-based prediction and optimization for the development of sustainable wastewater treatment technologies.