Machine-Learning-Assisted Design of a Robust Biomimetic Radiative Cooling Metamaterial - Abstract

Aug 5, 2026

Upload by Hazel

Recently, biomimetic photonic structural materials have advanced radiative cooling performance, yet biomimetic metamaterials often focus on biological radiative cooling mechanisms and insufficiently address sensitive parameter variations. Traditional numerical design methods are costly and labor-intensive and may produce high-performance microstructures only within narrow preparation intervals. A deep-learning-based data-driven approach is introduced to link biomimetic metamaterial structures to spectral responses. Using Batocera lineolata-inspired design and a truncated cone array on SiO2, robust optimal parameters spanning H/Dbottom from 0.8 to 2.4 are achieved, yielding average emissivity of 0.985, and experiments show an approximate 8.3 °C temperature drop at midday. This enables faster design and optimization for standardized passive radiative cooling.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages28

Reading Time1 h 11 min

AccessFree to download

Introduction

Biomimetic photonic micro/nanostructures and passive radiative cooling

Prior biological inspirations and emissivity enhancement mechanisms

Limitations of conventional metamaterial design and numerical optimization

Need for parameter–spectrum correlation and robust preparation guidance

ISTITUTO NAZIONALE DI RICERCA METROLOGICA Repository Istituzionale Machine-Learning-Assisted Design of a Robust Biomimetic Radiative Cooling Metamaterial This is the author's submitted version of the contribution published as: Original Machine-Learning-Assisted Design of a Robust Biomimetic Radiative Cooling Metamaterial / Ding, Zm; Li, X; Ji, Qx; Zhang, Yc; Li, Hl; Zhang, Hl; Pattelli, L; Li, Y; Xu, Hb; Zhao, Jp. - In: ACS MATERIALS LETTERS. -ISSN 2639-4979. - (2024) . [10 . 1021/acsmaterialslett.4c00337] Availability: This version is available at: 11696/81019 since: 2024-08-02T07:44:04Z Publisher: AMER CHEMICAL SOC Published DOI:10.1021/acsmaterialslett.4c00337 Terms of use: This article is made available under terms and conditions as specified in the corresponding bibliographic description in the repository Publisher copyright American Chemical Society (ACS) Copyright © American Chemical Society after peer review and after technical editing by the publisher. To access the final edited and published work see the DOI above. (Article begins on next page) 29 August 2024 Machine-Learning-Assisted Design of a Robust Biomimetic Radiative Cooling Metamaterial Zhenmin Ding, Xin Li, Qingxiang Ji *, Yunce Zhang, Honglin Li, Hulin Zhang, Lorenzo Pattelli, Yao Li, Hongbo Xu *, Jiupeng Zhao * Abstract: Recently, biomimetic photonic structural materials have significantly advanced the radiative cooling performance of designed materials. However, the biomimetic metamaterials have largely concentrated on the biological radiative cooling mechanism, with little effort to investigate sensitive parameter variations. Additionally, traditional numerical methods are not only costly and laborious, but also challenging to uncover optimal solutions, often yielding high-performance microstructures with a narrow range of preparation intervals. To address this, machine learning (ML) was introduced into the design of a biomimetic photonic structure based on the Batocera LineolataHope, and silicon dioxide (SiO2) was chosen as the substrate material. By employing a deep learning model, we can understand the correlation between biomimetic metamaterial and their corresponding spectral responses. We managed to attain a large range (H/Dbottom between 0.8 and 2.4) of optimal performance parameters for a truncated cone array, which resulted in an average emissivity of 0.985. Subsequently, the fabricated truncated cone array samples showed a temperature drop of about 8.3 °C at midday in experiments. By taking the above data-driven approach, the process of designing and optimizing robust biomimetic radiative cooling metamaterials can be expedited, making it an ideal option for standardized passive radiative cooling applications. Keywords: Biomimetic photonic materials, Machine learning, Photonic design, Infrared emissivity, Radiative cooler 1. Introduction Metamaterials created through photonic micro/nanostructures have been extensively studied in the fields of thermal manipulation, i.e., thermal camouflage, radiative cooling, signaling, and so forth. 1-3 Of particular interest is the imitation of microstructures based on natural creatures, as this allows for the direct analysis of the microstructure's functionality through the external behavior of plants and animals.4-9 The application of biomimetic photonic structures in the realm of infrared radiation is a promising trend, which can greatly enhance the radiative cooling performance of designed materials.10-11 This form of passive cooling, different from active cooling systems such as air conditioning, requires no extra energy input and produces no pollution, showing a significant contribution to energy conservation and environmental protection. 12-13 Researchers have developed materials and systems that are proficient in radiative cooling by examining the heat radiation characteristics of living organisms. Shi et al.14 first discovered that the Sahara si

FAQs

Why do current biomimetic radiative cooling metamaterial designs face limitations?

They largely follow biological radiative cooling mechanisms while neglecting how sensitive structural parameters vary. This weakens the connection between sample spectra and the parameters needed for practical preparation guidance.

What role does machine learning play in the proposed design workflow?

A deep learning model is used to learn the correlation between biomimetic metamaterial structures and their spectral responses. This supports efficient identification of optimal structures while improving robustness across preparation intervals.

What performance and experimental cooling results are reported for the optimized truncated cone array?

The optimized parameters (H/Dbottom between 0.8 and 2.4) yield an average emissivity of 0.985. Fabricated samples show an about 8.3 °C temperature drop at midday in experiments.

Aug 5, 2026

Upload by Hazel