Machine learning-assisted terahertz plasmonic biosensor with multi-material architecture for rapid bacterial water quality monitoring

Rapid detection of bacterial contamination in aquatic systems requires field-deployable sensing platforms with reproducible analytical performance. This study presents a multilayer plasmonic biosensor proposed for the simulation-based detection of waterborne bacteria. The sensor operates via localized electromagnetic field enhancement and resonance frequency shifts induced by bacterial analyte adsorption. Finite-element simulations in COMSOL Multiphysics were performed across graphene chemical potentials of 0.1–0.9 eV and incident angles of 0–80°. The sensor achieved an average refractive index sensitivity of 811 GHz/RIU, computed from linear regression of resonance frequency against refractive index across the full detection range (R² > 0.70, n = 1.33–1.3921 RIU). A secondary local maximum sensitivity of 976 GHz/RIU was additionally observed near the upper end of the detection range and is reported for completeness. The minimum detection limit is 0.071 RIU. Under analyte detection conditions within the refractive index range n = 1.33–1.3921 RIU, with resonance frequencies spanning 0.167–0.175 THz and a constant full width at half maximum of 0.054 THz, quality factors ranged from 3.093 to 3.241. Polynomial regression models predicted resonance shifts with coefficients of determination up to 0.94, enabling real-time calibration and performance monitoring. The proposed fabrication protocol employs chemical vapour deposition for graphene synthesis, electron-beam lithography for pattern definition, and thin-film deposition for multilayer stack formation using exclusively graphene, copper, aluminium, BaTiO₃, and SiO₂, conforming to standard microfabrication procedures adaptable for environmental monitoring and water quality assessment. All results are derived exclusively from electromagnetic simulations and numerical analysis. No experimental fabrication or measurements were taken.