Design and optimization of a reconfigurable multi-material THz MIMO antenna

Reconfigurable THz MIMO antennas are attracting increasing interest for next-generation high-frequency wireless and sensing systems, where compactness, high isolation, wide bandwidth, and low correlation are simultaneously required. In this paper, a surrogate-assisted optimization framework is proposed for a reconfigurable multi-material THz MIMO antenna integrating graphene, vanadium dioxide (VO 2 ), and black phosphorus (BP). The electromagnetic response is first modeled using an improved WCIP full-wave formulation, analytically validated for ultrathin conductive sheets and shown to provide a computational speedup of about 71 compared with conventional full-wave simulation. To enable efficient exploration of the material design space, a compact multi-output deep neural network is trained on nearly 1000 WCIP-generated samples and achieves high predictive accuracy, with average metrics of RMSE = 0.0767, MAE = 0.0621, and R 2 = 0.9995 . The optimized antenna operates at 1.5 THz and achieves 0.52 THz bandwidth, − 70.8 dB isolation, 14.25 dB gain, 91.21% efficiency, and an envelope correlation coefficient (ECC) of 0.001. These features make the proposed antenna a promising candidate for compact adaptive THz front-ends, high-data-rate short-range wireless links, and reconfigurable sensing platforms. These results demonstrate that the proposed framework provides an efficient and physically consistent route for the design of compact high-performance reconfigurable THz MIMO antennas.