Refinery-Waste-Derived Carbon Quantum Dots for Selective Metal Ion Fluorescence Sensing: Experimental, DFT, and Deep Learning Insights

In this study, we present a sustainable approach for synthesizing carbon quantum dots (CQDs) from petroleum refinery waste, transforming industrial residues into high-performance optical nanomaterials. Three different CQDs samples (QVR1, QVR2, and QVR3) were synthesized and extensively characterized using transmission electron microscopy (TEM), X-ray photoelectron spectroscopy (XPS), X-ray diffraction (XRD), and Fourier transform infrared spectroscopy (FTIR). Optical properties, assessed via UV–vis absorption and photoluminescence (PL), revealed strong and selective fluorescence enhancement in response to 30 environmentally relevant metal cations and anions. Specifically, QVR1, QVR2, and QVR3 exhibited selective and rapid interactions with Na+ (half-life = 11.33 s), Cd2+ (t1/2 = 7.38 s), and K+ (t1/2 = 17.32 s), respectively. Simplified Molecular Input Line Entry System (SMILES) was extracted using density functional theory (DFT) by RDKit and Open Babel, and combined with experimental PL data and binding energy values in a deep learning framework. The Graph Convolutional Network (GCN) model demonstrated superior predictive performance over ChemBERT, achieving an overall accuracy of 83.3% and a 42.9% recall for minority classes, compared to ChemBERT's 80.6% accuracy and 0% recall for minority classes. This work highlights the synergistic use of experimental, theoretical, and AI-driven cheminformatics approaches for understanding nanoscale sensing behaviors and establishes refinery-derived CQDs as promising low-cost materials for selective and sensitive optical sensing.