An Intelligent Reconfigurable Graphene–VO 2 Hybrid Terahertz Bandpass Filter Architecture for Differential Biomolecular Sensing and Machine Learning–Driven Analyte Characterization

A compact hybrid graphene–VO 2 terahertz (THz) bandpass filter integrating reconfigurable operation, differential biosensing, and machine learning (ML)–based prediction is proposed and numerically investigated. The device consists of a symmetric dual‐stepped‐impedance resonator (SIR) fabricated on a quartz/silicon substrate, where graphene provides electrical tunability and a VO 2 phase‐change layer enables dynamic modulation. Three independent tuning approaches are demonstrated: geometric control through SIR dimensions, graphene chemical potential variation (0.0–1.0 eV), and VO 2 conductivity switching (200 − 2 × 10 5 S/m). These mechanisms enable broadband frequency reconfiguration from 0.82 to 2.47 THz, achieving a tuning ratio of 3.01:1. For sensing applications, the differential resonator configuration provides self‐referenced operation, delivering a maximum refractive index sensitivity of 412 GHz/RIU and a figure of merit of 8.24 RIU −1 . Hemoglobin detection is demonstrated over clinically relevant concentrations (0–160 g/L) with a detection limit of 4.2 g/L. Furthermore, six ML models trained on 15,000 physics‐informed synthetic samples are evaluated for analyte prediction. Gradient boosting provides superior performance with an R 2 value of 0.9971, RMSE of 0.0061 RIU, and MAE of 0.0044 RIU. Random forest classification achieves 98.7% accuracy in identifying VO 2 phase states. The proposed graphene–VO 2 platform offers a versatile solution for biomedical sensing, environmental monitoring, and adaptive THz communication systems.