Bowtie-patterned MIM SERS platform assisted by machine learning for detection of pesticide residues in food matrices

The increasing use of pesticides and their mixtures poses a serious risk to human health and the environment. This increases the demand for simple, cost-effective, and reliable methods for detecting these residues. In this study, a highly sensitive in-house SERS platform based on a metal-insulator-metal (MIM) nanoarray structure was employed to acquire Raman fingerprint spectra of Pyrimethanil (PYM), Imidacloprid (IMI), and Chlormequat chloride (CCC) in pepper juice, yielding spectra with high signal-to-noise ratios. The detection limit for PYM in pepper juice (0.16 mg/kg) was well below both the EFSA (2 mg/kg) and EPA (2 mg/kg) limits. Among the tested pesticides, PYM shows the lowest detection limit, indicating a more efficient signal enhancement for the π-metal interaction. This strong affinity results in significantly enhanced Raman scattering activity. Furthermore, the unsupervised machine learning analysis techniques (e.g., PCA and HCA) used showed a concentration-dependent separation in spiked samples. The same approach also enabled detection and discrimination in real food samples obtained from different regions. These results demonstrate the potential of the developed platform for rapid, on-site monitoring of pesticide residues in complex food matrices.