Control and Optimization of Absorbing Behavior in Graphene-Based Multiple Narrowband Metamaterial Absorber by Machine Learning

Abstract

Graphene, with its tunable properties and high optical response characteristics, has a wide range of applications in the development of metamaterial absorbers: it is frequently employed as a substitute for the periodic metal structure of conventional metamaterial devices, or utilized as an intermediate layer to composite with the devices. However, precise control over absorption performance within specific narrowbands and the optimization of multi-peak absorption intensity remain key challenges. In this study, a machine learning approach is integrated into the device design process to develop a multilayer heterogeneous composite metamaterial absorber based on graphene, effectively addressing these issues. By combining a patterned metal layer, a graphene thin film layer, and a metal enhancement layer with two dielectric layers and a metallic ground plane, multi-narrowband resonance is successfully achieved. Firstly, machine learning supervision is utilized to effectively control the device’s multi-narrowband absorption behaviors within the 0.5-1.2 THz frequency range, leading to the realization of double-peak, triple-peak, and quadruple-peak multi-narrowband absorption behaviors, respectively. Then, leveraging the predictive capabilities of the machine learning model, the structural parameters of the device are tuned to achieve comprehensive optimization of its multi-narrowband absorption performance. As a result, the absorption performance across multiple frequency ranges exceeds 90%. This method avoids the traditional trial-and-error optimization and provides a scalable design framework for customized multi-narrow band high-performance terahertz absorber.

Keywords: Behavior regulation, comprehensive optimization, electric field loss, machine learning (ML), metamaterial absorber, multiple narrowband absorption