Design and Optimization of a Graphene-Enhanced Terahertz Metasurfaces Surface Plasmon Resonance Biosensor for High-Sensitivity Peptide Detection with Machine Learning Optimization Based on XGBoost Regressor

This study introduces a novel terahertz (THz) biosensor based on metasurfaces, designed for high-sensitivity peptide detection. The sensor features a multi-layered structure comprising gold, silver, and graphene elements deposited on a silicon dioxide substrate. Finite element method simulations were employed to evaluate the sensor's performance across two frequency bands: 0.32–0.44 THz and 0.11–0.15 THz. Key design parameters—including graphene chemical potential, resonator dimensions, and incident angle—were optimized to maximize sensor performance. The optimized device demonstrated exceptional sensitivity of 1500 GHzRIU⁻¹, with a figure of merit of 60 RIU⁻¹ and a detection limit of 0.016 RIU. The sensor successfully detected subtle changes in peptide concentrations, exhibiting frequency shifts of up to 60 GHz. Furthermore, an XGBoost regression model was implemented to predict sensor behaviour, achieving perfect accuracy (R² = 100%) across multiple parameter spaces. This high-performance biosensor shows significant promise for biochemical analysis applications, particularly in peptide detection and characterization, offering potential advances in drug discovery and disease diagnosis.