Machine learning-based design optimization of aperiodic multilayer coatings for enhanced solar reflection

Abstract

Multilayered coatings are promising and successful for applications in semiconductors, optical mirrors, and energy harvesting technologies. Amongst these, optical mirrors are essential for passive radiative cooling. Building upon the multilayer radiative cooling systems observed in snails and drawing from previous research, this study showcases the efficacy of machine learning algorithms in optimizing and gaining insights into multilayer structures. Due to the constraint of low sky window emissivity in biologically found calcite shells, focusing on solar reflectance becomes crucial to maximize the biological phenomenon found in snails. The manual search of the periodic multilayer design space for calcite with air gaps points to the maximum solar reflectance of ∼89% at 170 nm layer thickness for 20 μm coating. To unlock the full potential of these multilayers, we then employ machine learning-based evolutionary optimization method - a genetic algorithm. The optimized aperiodic coating shows a significant enhancement of solar reflectance to ∼99.8% for a 20 μm coating. Interestingly, the same average layer thickness of 170 nm provides maximum solar reflectance in 20 μm periodic and aperiodic calcite multilayer. Investigation of the spectral reflectance shows that layer thickness is crucial in tuning the solar reflectance. For small coatings, wavelengths with higher solar intensity are prioritized. Increasing the coating thickness allows inclusion of thicker layers to reflect longer wavelengths, leading to increasing trend of average calcite layer thickness. Further work exploring radiative cooling materials shows that calcite and barium sulfate reflect sunlight significantly better than silicon dioxide due to their refractive index contrast. Our findings and insights using bio-inspired design can provide superior solar reflectance utilizing thin coatings with modern manufacturing technology.