Session: 11-58-01: Nanoscale Thermal Transport

Paper Number: 119845

119845 - Machine Learning-Based Design Optimization of Aperiodic Multilayer Coatings for Enhanced Solar Reflection 

Sphincterochila Boissieri is a species of snail found in hot and dry climatic conditions in Israel and Egypt. They are known for their radiative cooling ability using calcite (CaCO₃) shells. Using solar reflectance and radiative cooling capability, snails' multilayered structures help them stay cool in extreme heat. A fundamental understanding of these bioinspired multilayers is critical, and there is a scope for further optimization to boost solar reflectance in our reflector technologies.

Our work explores a multilayer design using calcite as a reflective material with air gaps bioinspired from the snail shell structure. We assume 1-D morphology for these multilayers enabling the use of transfer matrix method (TMM) to study the photonic interaction with the nanostructure. We set an initial benchmark using a manual search for periodic photonic crystals varying the layer thickness for a fixed coating thickness. The periodic multilayer shows a maximum solar reflectance of 89% at 170 nm layer thickness for a 20 μm coating. We optimize the multilayer structure using a machine learning-based genetic algorithm (GA). The evolutionary GA performs selection, crossover, and mutation, starting with a randomly generated population to find the maximum reflectance in a vast design space. We obtain a 99.8% solar reflectance for the same total thickness with a non-intuitive aperiodic structure obtained from the GA optimization. Interestingly, the optimized structure's average calcite layer thickness is 170 nm, the same as the periodic multilayer. Investigation of spectral reflectance shows that a layer thickness distribution is crucial in tuning solar reflectance. We optimize coatings of various thicknesses in 5 μm to 40 μm range. For small coatings, wavelengths with higher solar intensity are prioritized and localized effectively, while the longer wavelengths show oscillating electric fields indicating transmission. Increasing the coating thickness allows the inclusion of thicker layers that can reflect longer wavelengths, leading to an increasing trend of average calcite layer thickness with increasing coating size.

Additionally, we employ our GA on other well-known radiative cooling materials like barium sulfate (BaSO₄) and silicon dioxide (SiO₂) multilayers to compare them with calcite multilayers. Our findings demonstrate the slight superiority of barium sulfate to calcite, while silicon dioxide gives the worst performance. All the resulting GA optimizations lead to approximately 50% material-to-air gap ratio as the optimized volume fraction. This comparison and analysis of the non-intuitive optimized configurations shows a dependence on design factors like layer thickness, coating thickness, volume fraction, and material properties like refractive indices.

Our work helps us understand the underlying mechanisms thriving in the ecological environment and artificially developed multilayer systems. It will help exploit the multilayer design for enhanced reflectance in bio-inspired design and manufacturing in the future.

Presenting Author: Krutarth Khot Purdue University

Presenting Author Biography: Krutarth is a third year PhD student in Mechanical Engineering at Purdue University (West Lafayette). His research is focused on nanoscale thermal and radiative heat transfer across interfaces with the use of machine learning and other computational techniques.

Authors:

Krutarth Khot Purdue University
Prabudhya Roy Chowdhury Purdue University
Xiulin Ruan Purdue University