Fluorescent graphene quantum dots-enhanced machine learning for the accurate detection and quantification of Hg2+ and Fe3+ in real water samples

Abstract

Selective, accurate, fast detection with minimal usage of instrumentation has become paramount nowadays in the areas of environmental monitoring. Herein, 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 in real water samples. The probe is obtained via the electrosynthesis of CVD graphene in the presence of urea, followed by functionalization with 1-nitroso-2-naphthol (NN). The functionalization with NN moieties dramatically improves the selectivity and sensitivity of the probe toward Hg²⁺ and Fe³⁺, as demonstrated by LODs of 0.001 and 0.003 mg L⁻¹, respectively. Simulations performed by time-dependent density functional theory (TD-DFT) reveals that the NN molecules within the GQDs are responsible for 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. Our work demonstrates how chemical modification of GQDs, guided by an efficient ML model, markedly enhances sensitivity and selectivity in detecting harmful ions while critically reducing experiments and instrument handling.