Graphene Metasurface Based Biosensor for COVID-19 Detection in the Terahertz Regime with Machine Learning Optimization using K-Nearest Neighbours Regression

This investigation presents a novel terahertz-based biosensing platform for SARS-CoV-2 detection, utilizing a hybrid architecture comprising nanostructured TiO2, black phosphorus, and graphene-based metasurfaces. The sensor architecture was systematically optimized through parametric analysis, exhibiting robust performance characteristics within the 0.1-0.3 THz. Comprehensive characterization was conducted by modulating critical parameters, including the graphene chemical potential, electromagnetic wave incident angle, and resonator geometrical configurations, to demonstrate the sensor's electromagnetic response characteristics. The optimized configuration achieved a sensitivity coefficient of 600 GHzRIU-1, with a corresponding figure of merit of 18.75 RIU-1 and a minimum detection threshold of 0.064 RIU. Electromagnetic field distribution analyses were performed to elucidate the underlying sensing mechanisms. The investigation incorporated machine learning methodology through implementation of a K-Nearest Neighbours regression algorithm for predicting absorption coefficients across diverse structural and operational parameters, consistently demonstrating coefficient of determination (R2) values of 1.00 across multiple validation datasets. This synergistic integration of computational electromagnetics and machine learning frameworks for designing the proposed sensor demonstrates significant potential for rapid, high-sensitivity SARS-CoV-2 detection utilizing terahertz spectroscopy.