Deep learning-assisted inverse design of nanoparticle-embedded radiative coolers
Min Ju Kim, June Tae Kim, Mi Jin Hong, Sang Wook Park, Gil Ju Lee
Abstract
Radiative cooling is an energy-efficient technology without consuming power. Depending on their use, radiative coolers (RCs) can be designed to be either solar-transparent or solar-opaque, which requires complex spectral characteristics. Our research introduces a novel deep learning-based inverse design methodology for creating thin-film type RCs. Our deep learning algorithm determines the optimal optical constants, material volume ratios, and particle size distributions for oxide/nitride nanoparticle-embedded polyethylene films. It achieves the desired optical properties for both types of RCs through Mie Scattering and effective medium theory. We also assess the optical and thermal performance of each RCs.