Ultra-broadband tunable terahertz metasurface absorber with multi-mode regulation based on artificial neural network

The multi-window tunable absorption of terahertz (THz) waves faces challenges such as a small tunable bandwidth, discontinuous tuning windows, and a complex design process. This study addresses these issues by employing inverse design based on artificial neural networks (ANN) to achieve an ultra- broadband tunable THz metasurface absorber (UTTMA). Leveraging the electrically tunable property of graphene and the phase transition property of vanadium dioxide (VO2), the UTTMA achieves multi-domain dynamic operation across a large frequency band. The design allows for arbitrary switching between multiple windows by transforming between graphene surface-localized plasmon resonance (LSPR), fundamental order surface plasmon polaritons (SPP) resonance, and graphene surface plasmon resonance (SPR), and second order SPP resonance. The tunable windows cover the range between 1.90 and 10.00 THz, encompassing four-fifths of the entire THz band, surpassing previously reported absorbers. The accuracy of the ANN inverse design is verified, and a systematic analysis of the optical performance of the UTTMA is conducted. These findings offer a viable and effective method for expanding the tunable bandwidth and simplifying the design process for THz metasurfaces.