Artificial Intelligence-Integrated Overcoupled Resonator for Multifunctional Pesticide Spectral Classification and...

Pesticide residue detection plays a critical role in ensuring food safety, protecting human health, promoting environmental governance, and supporting sustainable agricultural practices. However, the growing diversity of pesticides, coupled with complex and overlapping spectral signatures and low residue concentrations, significantly limits the efficiency and applicability of conventional detection methods. Here, we present an overcoupled (OC) resonator platform integrated with artificial intelligence (AI) for multifunctional pesticide analysis. The OC resonator exhibits an ultrabroadband spectral response spanning 1650-750 cm^-1, representing up to a 1685-fold bandwidth enhancement compared with conventional narrowband resonators. Owing to this broadband characteristic, the OC resonator eliminates the need for resonance tuning when detecting different pesticide molecules. In addition, the OC resonator features high sensitivity and inherent immunity to Fano asymmetry, enabling identification and trace-level detection of multiple pesticide species. Experimental results demonstrate that the platform achieves a minimum limit of detection as low as 12.5 ng·μL^-1 for pesticide molecules. To resolve the complexity and overlap in molecular spectral features, we incorporate AI algorithms for spectral classification, concentration prediction, and signal reconstruction, achieving 100% classification accuracy across complex mixtures. Furthermore, we validate the platform's real-world applicability by detecting pesticide residues on apple peels and in lake water, demonstrating excellent selectivity and strong interference suppression in complex backgrounds. This study not only expands the scope of OC resonator-based pesticide detection but also establishes a versatile framework for manipulating light-matter interactions, designing advanced plant sensors, and enabling ultratrace molecular diagnostics.