From theory to practice: DFT-guided Raman spectral analysis and machine learning for high-accuracy pesticide identification

Pesticide residues pose threats to the environment and food safety, and accurate and efficient detection technologies are key to addressing the challenges of pesticide residues. Theoretical calculations of Raman spectroscopy can guide molecular structure analysis, signal analysis, and optimization of detection methods, significantly improving detection accuracy. This study aims to verify the reliability of Raman spectroscopy in identifying pesticides and their isomers, and to provide theoretical and technical support for pesticide residue detection. Based on density functional theory (DFT), we calculated the Raman spectra of 166 pesticides, focused on analyzing the Raman peaks and vibrational modes of 22 heterocyclic pesticides, and explored the effects of functional group isomers and chain isomers on the spectra; additionally, we used PCA and t-SNE machine learning algorithms for the identification of these 22 pesticides. The results clarified the spectral characteristics of different pesticides and the regularity of isomers' influence on spectra, and the machine learning algorithms achieved accurate identification of the 22 pesticides. Raman spectroscopy combined with theoretical calculations and machine learning exhibits high reliability and application potential in the identification of pesticides and their isomers, providing strong technical support for environmental and food safety supervision.