A review: the development of terahertz meta-devices using artificial intelligence technology in advanced optoelectronic applications

Abstract

Artificial intelligence (AI) techniques have been increasingly integrated into the design and optimization of metamaterial devices (meta-devices). This approach reduces the dependence on conventional methods that require extensive expertise in electromagnetic theory and time-consuming numerical simulations, thereby accelerating the design process. Despite advances in this field, the development of terahertz (THz) meta-devices remains underexplored. This study critically examines conventional and AI-assisted design methodologies for THz metamaterial devices or THz meta-devices, focusing particularly on the current research progress of tunable and multifunctional meta-devices. Here, we review how AI techniques, including machine learning, evolutionary algorithms, and deep learning, can be utilized to optimize THz meta-devices by establishing data-driven correlations between structural configurations and electromagnetic responses. Furthermore, we address future developments and challenges in the meta-device field, to inform future research in this area. We also provide a perspective on emerging AI-assisted optical computing hardware that could accelerate THz device design and real-time signal processing.