A multilayer 2D material plasmonic metasurface terahertz biosensor with machine learning optimization for high-sensitivity tuberculosis detection.

Tuberculosis (TB) remains a leading global infectious disease, with over 10.6 million new cases annually and persistent diagnostic gaps in low-resource settings. Existing methods such as sputum smear microscopy and nucleic acid amplification tests suffer from low sensitivity, long turnaround times, and infrastructure dependency. To address these shortcomings, this work proposes a novel five-layer Au/Graphene/MoS₂/MXene/Phosphorene/WS₂ plasmonic metasurface terahertz (THz) biosensor for label-free, high-sensitivity TB biomarker detection. The hierarchical architecture integrates a MXene-functionalized elliptical central resonator for selective TB biomarker capture (LAM and MPT64 antigen), concentric Au/phosphorene ring resonators for polarization-dependent field enhancement, peripheral WS₂ and MoS₂ resonators for multiband absorption and chemical stability, and a gate-tunable graphene layer enabling dynamic post-fabrication tunability. Finite-element electromagnetic simulations and the transfer matrix method were employed to characterize the full spectral and near-field response. Bayesian Ridge regression was applied to optimize geometric parameters, resonator dimensions, and incidence angles, achieving predictive R2 values exceeding 0.9996. The optimized sensor delivers a maximum sensitivity of 1000 GHz/RIU, a figure of merit of 30.303 RIU-1, an electric field enhancement of 1.1 × 108 V/m at resonance (0.996 THz), and off-resonance transmittance consistently above 98% across graphene chemical potentials from 0.1 to 0.9 eV and incidence angles from 0° to 80°. These results establish the proposed sensor as a competitive platform for rapid, accurate, point-of-care TB diagnostics.