Thermally Switchable Absorption in VO₂–Graphene Terahertz Metamaterials Enabled by Machine Learning Optimization

Abstract VO₂, a prototypical phase-change material, exhibits a reversible insulator-to-metal transition near 68°C, accompanied by several orders of magnitude change in electrical conductivity while preserving structural integrity. Graphene, renowned for its tunable electronic properties and superior optical response, has emerged as a promising alternative to conventional periodic metal structures in metamaterials, or as an interfacial layer in composite devices. In this study, we integrate graphene and VO₂ into a multilayer heterostructured metamaterial absorber and incorporate machine learning techniques to optimize its geometric parameters, to achieve switchable high-performance absorption behavior. The designed absorber consists of a patterned metallic top layer, a graphene sheet, a VO₂-based functional layer, two dielectric layers, and a metallic aluminum ground plane. Leveraging the thermally induced phase transition of VO₂, the device enables dynamic switching of different absorption modes without changing its geometric shape and parameters. Specifically, in the metallic state of VO₂ (> 68°C), the absorber demonstrates broadband absorption performance with an average absorption exceeding 90% across the 1.14–1.305 THz range. In contrast, when VO₂ is in its insulating state (< 68°C), the device exhibits triple-band narrowband absorption with three sharp resonance peaks, achieving maximum absorptivities of 72%, 71%, and 99.7%, respectively. This work introduces a thermally switchable metamaterial absorber with fixed geometry, integrating VO₂ and graphene to achieve thermally switchable absorption behaviors, offering a practical solution for multifunctional terahertz applications.