Spectroscopic characterization of three fungicides using terahertz spectroscopy and DFT theory, combined with convolutional neural networks for quantitative analysis of tebuconazole.

Fungicides are essential agrochemicals for the prevention and control of plant diseases. Counterfeit products, often lacking enough active ingredients, can compromise disease management and pose risks to agricultural safety. Precise quantification of the chemical structure and concentration of active components enables reliable authentication of fungicide formulations, ensuring their efficacy in crop protection and supporting the quality and safety of agricultural production. This study presents a comprehensive analysis of three fungicides-tebuconazole (TEB), thiophanate-methyl (TPM), and fludioxonil (FDL)-using a combined approach integrating terahertz time-domain spectroscopy (THz-TDS), density functional theory (DFT), and convolutional neural networks (CNN). Characteristic absorption features in the 0.5-2.5 THz range were first identified, followed by data preprocessing through wavelet transform denoising and asymmetric least squares baseline correction. Vibrational mode assignments corresponding to the experimental absorption peaks were performed based on DFT calculations and GaussView analysis. A CNN model was then developed using the spectral data to accurately predict pesticide concentrations and evaluate the corresponding limits of detection. This work reports, for the first time, the terahertz fingerprint spectra of these three fungicides and offers a novel strategy for the quantitative analysis of active ingredients in pesticide formulations.