Fluorescent Graphene Quantum Dots-Enhanced Machine Learning for Accurate Detection and Quantification of Hg2+ and Fe3+ in Real Water Samples

Abstract

Selective, accurate, fast, and with minimal usage of instrumentation has become paramount nowadays in areas of environmental monitoring. Here, we chemically modified fluorescent graphene quantum dots (GQDs) and trained a Machine Learning (ML) algorithm for the selective quantification of Hg2+ and Fe3+ ions present within real water samples. The probe is obtained by an electrosynthesis of CVD graphene in the presence of urea, followed by the functionalization with 1-nitroso-2-naphthol (NN). The functionalization with NN moieties dramatically improve selectivity and sensitivity toward Hg2+ and Fe3+, as demonstrated by LODs as low as 0.001 mg L-1 and 0.003 mg L-1; respectively. Time-dependent density-functional theory (TD-DFT) reveals that the NN molecules within the GQDs are responsible of the florescence emission of the probe. The emission spectra profiles exhibited distinct characteristics between Hg2+ and Fe3+ enabling the ML model to precisely quantify and differentiate between both analytes present in natural and drinking waters. The ML results were further validated by measurements via cold vapor-atomic fluorescence spectroscopy and UV-vis spectroscopy. The efficiency of the ML model eliminates the necessity of extensive training values and validations, making it a reliable tool for precisely quantifying the presence of Fe and Hg ions in real water samples.