Electrostatic and Capillary Force-Enhanced Dynamic SERS Combined with Machine Learning for Rapid Quantification of Structurally Similar Triazole Pesticides in Tobacco

Abstract

Rapid and accurate detection of pesticide residues is crucial for safeguarding both environmental and human health. Nonetheless, the quantitative detection of structurally similar pesticides, which are frequently employed concurrently in agricultural practices to manage a range of pests and diseases, presents significant challenges. In this study, we introduced a methodology that integrates dynamic surface-enhanced Raman spectroscopy (D-SERS) with machine learning to quantify the binary mixtures of triadimenol and triadimefon. Initially, halogen-modified nanoparticles (Ag–Cl NPs) were synthesized to serve as SERS substrates, and the D-SERS method identified an optimal hot spot duration of approximately 60 s through the combined effects of electrostatic and capillary forces. The SERS spectra acquired at this hot spot exhibited high stability (relative deviation < 9.67%), enabling sensitive detection, with limits of detection (LODs) of 74 ppb for triadimenol and 68 ppb for triadimefon. Using these spectra, a PCA-SVM model was developed and trained to accurately quantify both pesticides, individually and in mixtures, in tobacco samples. This integrated D-SERS and machine learning platform provides a rapid, sensitive, and reliable analytical method for the simultaneous quantification of structurally similar multicomponent pesticides, offering a promising solution for addressing pesticide residue issues in agricultural products.