AI-enabled design of extraordinary daytime radiative cooling materials

Abstract

Here we developed an artificial intelligence (AI)–based deep generative model, combined with a one-dimensional convolutional neural network (1D-CNN), for the inverse design of extraordinary passive daytime radiative cooling (PDRC) materials in a probabilistic manner. This AI-enabled strategy delivered a comprehensive solution for the one-to-many mapping problem of inverse design by predicting the optical properties—specifically, the refractive index (n) and extinction coefficient (k)—of hypothetical new materials. We then used the Kramers–Kronig relations and Lorentz–Drude model to validate the predicted results, and discovered a new record-breaking PDRC material that provided a decrease of approximately 79 K relative to ambient temperature and of approximately 12 K relative to that provided by the conventional ideal selective emitter under conditions of perfect insulation and a perfect electric conductor substrate. This AI-extrapolated approach toward extraordinary PDRC materials provides new guidelines for designing PDRC materials and connects the gap between ideal selective emitters and real materials.