Deep Learning-Based Multicapturer SERS Platform on Plasmonic Nanocube Metasurfaces for Multiplex Detection of Organophosphorus Pesticides in Environmental Water.

Abstract

In situ rapid detection of contaminants in environmental water is crucial for protecting the ecological environment and human health; however, it is always hindered by the complexity of sample matrices, trace content, and unknown species. Herein, we demonstrate a deep learning-based multicapturer surface-enhanced Raman scattering (SERS) platform on plasmonic nanocube metasurfaces for multiplex determination of organophosphorus pesticides (OPPs) residues. Poly(vinylpyrrolidone), 4-mercaptobenzoic acid, and l-cysteine are assembled on Ag nanocubes (AgNCs) and act as capturers to chemically define OPPs. Meanwhile, the OPPs-captured AgNCs efficiently close the interparticle distance and generate plasmonic metasurfaces, guaranteeing ultrasensitive and reproducible SERS analysis. Furthermore, by strategically combining all capturer-OPP SERS spectra, comprehensive "combined-SERS spectra" are reconstructed to enhance spectral variations of each OPP. Based on the combined-SERS spectra, a deep learning model is trained to predict OPPs, which significantly improve the qualitative and quantitative analysis accuracy. We successfully identified multiple OPPs in farmland, river, and fishpond water using this strategy. The whole detection procedure requires only 30 min, including sampling, SERS measurements, and deep learning analyses. This combination of a multicapturer SERS platform with the deep learning algorithm creates a rapid and reliable analytical strategy for multiplex detection of target molecules, providing a potential paradigm shift for environment-related research.