Performance Enhancement of Graphene-based Linear to Circular Polarization Converter for Terahertz Frequency Using a Novel Parameter Prediction Methodology

This paper presents a graphene-based linear-to-circular polarization converter (LTCPC) for the terahertz frequency regime. Besides these, it suggests a technique for effective and accurate metasurface designing using a novel approach, which we call the machine learning parameter prediction method (MLPPM). It provides an efficacious solution for the limitations of the existing parametric variation method and over-reliance on human precision and perception in metasurface designing. The periodic unit cell is designed using a gold-backed silicon dioxide (SiO2) containing an elliptical-shaped graphene ring with splits. The proposed design converts linearly polarized incident waves into circularly polarized reflected waves from 2.44 to 3.70 THz with a wide bandwidth of 1.26 THz and about 41% of fractional bandwidth. The unit cell dimensions are optimized using the proposed MLPPM, and the operational bandwidth from 2.30 to 4.0 THz, i.e., 54% fractional bandwidth (FBW), is reported. That is, using this method, 31.7% of the overall performance of the unit cell is improved. The empirical model for the technique is also proposed. The tunability of the graphene monolayer is achieved by varying the chemical potential and relaxation time. The incident angle invariability of the proposed unit cell is tested, and up to 50-degree incident angle invariability is reported.