Ultra-Wideband Multi-Resonance High-Performance Graphene-Enhanced Multi-Square Slotted Two-Port MIMO Antennas With Machine Learning-Driven Optimization for 6G Terahertz Communication

A compact MIMO antenna operating in the THz region is designed and analyzed for potential application in future 6G systems in this paper. The graphene/polyimide-based antenna addresses critical problems, such as miniaturization, mitigation of substrate losses, wideband operation, and energy efficiency, that are essential for high-frequency transceivers. A full optimization of the MIMO architecture follows the sequence from the single-antenna-element design and is supported by detailed simulations and parametric studies. The antenna exhibits resonances at 3.1615 THz, 3.97 THz, 4.806 THz, and 5.834 THz, with a wideband of 3.5218 THz. As a result, it has a gain of 11.23 dB, an efficiency of 86.87%, and the best isolation of -34.174 dB, which reduces mutual coupling among elements. The antenna's resonant frequencies and input impedance are corroborated using an RLC circuit model. In addition, ML-based optimization is applied to enhance the design process and achieve optimal performance by accurately predicting key parameters such as resonance, gain, isolation, and RLC characteristics. Among the tested machine learning methods, XGB Regression achieves the highest accuracy, with an R-squared value of 93. 29% with the smallest mean absolute error of 4.26%. This demonstrates the ability of machine learning to accelerate antenna design and improve performance. The results reveal the suitability of the designed antenna for high-speed, high-capacity wireless communication, sensing, and bio-applications, making it a good candidate for next-generation 6G systems.