Ultra-sensitive terahertz biochemical detector via gold–graphene synergistic metastructures enhanced by machine learning optimization

The research introduces gold and graphene in a metasurface engineered to identify substances with subtle refractive index variations. The sensor features a multi-layered resonator configuration comprising a central circular resonator and concentric square rings fabricated on a SiO₂ substrate. Through systematic optimization using COMSOL Multiphysics and machine learning technique, the final design achieves exceptional sensitivity of 800 GHz/RIU within the refractive index range of 1.29–1.38 RIU. The sensor demonstrates remarkable performance metrics, including 8.000 RIU^–1, 0.157, and 13.140 as FOM (figure of merit), detection limit and quality factor. A stacking ensemble regressor approach reduced computational requirements by 88 % while maintaining 96–100 % prediction accuracy. Combining high sensitivity with reliable operation, this design excels in detecting low refractive indices with precision in the THz spectrum. Optimization in machine learning enhances sensitivity by fine-tuning parameters, reducing errors, improving accuracy, refining models, and boosting overall performance effectively.