Artificial Intelligence-Enhanced Metamaterial Bragg Multilayers for Radiative Cooling

Abstract

A full numerical study combining artificial intelligence (AI) methods and electromagnetic simulation software on a multilayered structure for radiative cooling (RC) is investigated. The original structure is made of SiO₂/Si nanometer-thick layers that make a Bragg mirror for wavelengths in the solar irradiance window (0.3–4 μm). The structures are then optimized in terms of the calculated net cooling power and characterized via the reflected and absorbed incident light as a function of their structural parameters. This investigation provides with optimal designs of beyond-Bragg, all-dielectric, ultra-broadband mirrors that provide net cooling powers in the order of ≈100 W m⁻², similar to the best-performing structures in literature. Furthermore, it explains AI's success in producing these structures and enables the analysis of resonant conditions in metal-free multilayers with unconventional layer thickness distributions, offering innovative tools for designing highly efficient structures in RC.