Inverse design and performance study of ultra-wideband radiative cooling metasurfaces based on hierarchical micro-nano structures

Passive radiative cooling technology represents a promising sustainable approach to addressing the energy consumption crisis associated with temperature control and supporting the achievement of the “dual carbon” goals, while metamaterials with precise spectral control capabilities are the key means of realizing high-performance passive radiative cooling. Traditional trial-and-error methods suffer from low optimization efficiency and are prone to being trapped in local optima in high-dimensional parameter spaces. Moreover, machine learning approaches often exhibit limited generalization capability and struggle with multi-objective collaborative optimization. To address these challenges, this study proposes a reverse design framework based on a conditional generative adversarial network (cGAN) with a forward-simulation surrogate-model constraint. Using this framework, the automated global optimization of a ZrO 2 /PDMS/Ag hierarchical radiative cooling metasurface is achieved. The designed metasurface adopts a three-tier composite architecture, achieving a weighted-average reflectance of 96.51% in the 0.25-2.5 μm solar wideband and a weighted-average emissivity of 97.55% in the 8-13 μm atmospheric-transparency window. Under AM1.5 standard spectral noon peak irradiance conditions, its cooling power reaches 114.33 W/m 2 , enabling stable sub-ambient cooling at an ambient temperature of 40 °C, while demonstrating excellent robustness across a wide range of wind speeds and under complex meteorological conditions. Furthermore, this metasurface exhibits outstanding potential for year-round energy savings and carbon reduction in various global climate zones. These characteristics indicate that the proposed inverse design framework and its optimized metasurface structure hold promising application prospects in fields such as building energy conservation and electronic device heat dissipation.