Bi–TiO2 nanomaterials in treating toxic water dye pollutants with machine learning models

Abstract

In the current investigation, bismuth-loaded TiO₂ nanoparticles were effectively synthesized via a single-step ultrasonication technique. The phase identification of the Bi–TiO₂ nanomaterials was carried out using powder X-ray diffraction technique, other physical and chemical characterization methods. Furthermore, variations in bismuth concentrations (0.3 M, 0.5 M, and 0.7 M) led to the observation of a notable decrease in the band gap (3.20, 2.95, 2.91, and 2.80 eV). The efficiency of these samples was evaluated by photocatalytic degradation of Congo red azo dye, a toxic pollutant in aqueous environment. 0.7 M Bi-loaded sample exhibited about 95 % Congo red dye degradation efficacy in 480 min. Additionally, the photodegradation experiments were carried out using a variety of scavengers (Ethylene diamine tetra acetic acid (EDTA), Isopropyl alcohol (IPA), and benzoquinone) to illustrate that the degradation is mainly by OH^● radicals. Machine learning (ML) models were used to predict the photocatalytic degradation efficiency of Bi–TiO₂ nanomaterials in treating Congo red dye. Using Random Forest (RF), Gradient Boosting Decision Trees (GBDT), and XGBoost models, we optimized several experimental conditions such as catalyst concentration, dye concentration, and scavenger presence to maximize degradation efficiency.