Inverse design of colored daytime radiative coolers using deep neural networks

Passive daytime radiative cooling is an eco-friendly and cost-efficient cooling strategy that is achieved by selectively reflecting the solar irradiance and emitting heat to cold background of the universe through the atmospheric window (AW) at infrared wavelengths. The daytime radiative coolers traditionally exhibit a grey or white color due to the requirement of high solar irradiance reflectance. Here, we present a colored daytime radiative cooler (CDRC) that has high reflectance in the NIR, high emissivity in the AW, and the capability to generate subtractive primary colors based on Fabry-Pe<acute accent>rot interference using metal-insulator-metal (MIM) structures. The structural parameters of the MIM multilayers are inversely designed using tandem neural networks to achieve cooling powers of 11.2-38.2 W/m2 with on-demand color generation. The proposed CDRCs have potential to be used for cooling thermal sensitive electronic and optoelectronic devices and aesthetic applications.