Design and analysis of a slotted square 1 × 2 MIMO terahertz patch antenna with machine-learning-based gain prediction and equivalent rlc validation

This paper presents the design and analysis of a compact graphene-based terahertz (THz) microstrip patch antenna in a 1 × 2 MIMO configuration for potential 6G wireless communication applications. The proposed antenna consists of slotted square graphene radiating patches printed on a polyimide substrate with a partial copper ground plane. The circular and rectangular slot arrangement, together with optimized substrate and feed dimensions, is used to achieve multiband operation, improved impedance matching, and reduced mutual coupling between the two antenna elements. The antenna exhibits three resonant modes at 4.63 THz, 5.3425 THz, and 6.15 THz, with bandwidths of approximately 0.532 THz, 0.465 THz, and 0.50 THz, respectively. It achieves a maximum gain of 14.6 dB, radiation efficiency of approximately 92%, and mutual coupling of (-33.03) dB. In addition, a low envelope correlation coefficient (ECC) of approximately 0.0001 and a diversity gain (DG) of approximately 10 dB demonstrate strong MIMO diversity performance. An equivalent RLC circuit model is developed to analyze the antenna’s multiresonant behavior, impedance matching, and coupling characteristics. Machine-learning regression models are also employed to predict antenna gain from simulation-derived design parameters. Among the evaluated models, XGBoost achieved the best performance, with an MAE of 3.79%, MSE of 0.74%, RMSE of 8.60%, and (R^2) of 91.77%. The proposed antenna is a promising candidate for future high-speed THz communication and sensing applications.