Enabling emergency response to arsenic contamination: simultaneous and rapid identification of arsenic speciation by a machine learning-driven fluorescent sensor array

Abstract

The rapid identification of arsenic speciation is critical for assessing its toxicity and guiding emergency response during water contamination events, yet it remains a significant challenge for current analytical methods. Herein, a novel machine learning-driven fluorescent sensor array was designed for the differentiation of four arsenic species, including arsenite (AsIII), arsenate (AsV), monomethylarsonic acid (MMAV), and dimethylarsinic acid (DMAV). Two Fe-based luminescent metal-organic frameworks (NH2-MIL-88(Fe) and OH-MIL-88(Fe)) were synthesized by functionalizing MIL-88 (Fe) with 2-amino-terephthalic acid and 2-hydroxy-terephthalic acid, respectively, both of which presented promising fluorescence behavior. Remarkably, varying arsenic species differentially regulated the fluorescence intensity of NH2-MIL-88(Fe) and OH-MIL-88(Fe), which was further analyzed by pattern recognition methods to develop a fluorescence sensor array for the rapid, simultaneous identification of four arsenic species and their mixtures. Furthermore, a machine learning algorithm was employed to integrate with the fluorescent sensor array to establish a stepwise prediction model to precisely identify and predict four arsenic species, which was successfully applied to actual water samples. Thus, our findings presented a robust, rapid, and intelligent platform for arsenic speciation, offering a powerful tool for water quality assessment and emergency response.