Next-generation terahertz gas sensor based on a machine learning-optimized hybrid plasmonic metasurface

This study presents a gas sensor employing a multi-resonator architecture composed of gold, MXene, copper, and graphene layers deposited on a silicon dioxide substrate. The device incorporates concentric circular and square ring resonator geometries designed to operate across the 0.1–1 THz frequency range. Finite element analysis using COMSOL Multiphysics demonstrates significant tunability through modulation of graphene chemical potential (0.1–0.9 eV) and incident angle variation (0°–80°). The sensor exhibits a refractive index sensitivity of 300 GHzRIU −1 , a quality factor of 13.143, and a figure of merit of 4.762 RIU −1 . Integration of machine learning algorithms, specifically polynomial regression models, facilitates enhanced performance prediction with coefficient of determination (R 2 ) values approaching unity. The proposed sensor architecture is compatible with standard photolithographic fabrication processes, demonstrates resilience to environmental perturbations, and presents a viable solution for portable gas detection systems. Compared to previously reported designs, this multi-material configuration offers enhanced practicality and detection precision for industrial sensing applications.