Machine-Learning-Optimized Refractory Metasurfaces for Thermophotovoltaic Energy Conversion

非技术性：开发新的环保能源是社会面临的重大工程挑战之一。热光伏器件将废热转化为可用的电能，并吸引了极大的研究兴趣，以满足日益增长的电力需求。热光伏（TPV）系统可以利用许多能源，包括太阳能和来自化石燃料和工业过程的废热。TPV系统可以使重量轻，多功能和紧凑的发电机，是无声的，低维护和能源效率。实现高效TPV系统需要提高材料，光子学和设计的基础知识。一个关键的挑战是优化多功能TPV组件及其组成材料，以便在环境和极端温度下稳定运行。该项目将使用人工智能（机器学习）将光学材料的知识与先进的优化相结合，以实现高效的TPV系统。该项目将创建一个全新的机器学习辅助优化框架，用于实现先进的TPV组件。该项目将利用可定制光学材料的扩展知识和数据库，并将机器学习算法与光子设计相结合。技术：近年来，通过使用光子超颖表面进行有效的能量转换（包括热光致发光），对材料的光学和光谱特性进行工程设计的研究兴趣很大。该计划将先进的光子拓扑优化与基于深度学习的逆向设计方法和全面的材料数据库相结合，以解锁非正统的光学设计，从而实现TPV应用的高效组件。这项工作将扩大设计参数空间，并结合机器学习方法，以实现拓扑优化的速度和效率的显着提高，以及建立一个大型的文件材料数据库。通过非传统的光学设计，该计划旨在通过增强辐射传热过程来开发高效的TPV能量转换方法。所提出的TPV器件可以通过将发射器的发射率与商业光伏电池（例如硅、锑化镓、砷化铟镓）的带隙相匹配来实现无与伦比的能量转换效率，该能量转换效率可能超过50%。这种方法可以提升纳米光子设计到以前不可用的制度，并可以应用到光子系统超越TPV。这个奖项反映了NSF的法定使命，并已被认为是值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估的支持。

Nontechnical:Developing new, environmentally friendly energy sources is one of the grand engineering challenges faced by society. Thermophotovoltaic devices convert waste heat into usable electricity and have attracted a great research interest to satify the increasing need for electrical power. Thermophotovoltaic (TPV) systems can take advantage of many energy sources, including solar energy and waste heat from fossil fuels and industrial processes. TPV systems could enable low-weight, versatile and compact electricity generators that are noiseless, low-maintenance and energy-efficient. Realizing high-efficiency TPV systems requires advancing fundamental knowledge of materials, photonics, and design. A key challenge is to optimize multi-functional TPV components and their constituent materials for stable operation under environment and extreme temperatures. This project will use artificial intelligenc (machine learning) to merge the knowledge of optical materials with advanced optimization to achieve highly efficient TPV systems. The project will create a fundamentally new, machine-learning-assisted optimization framework for the realization of advanced TPV components. This project will leverage the extended knowledge and database of tailorable optical materials and integrate machine-learning algorithms with photonic designs.Technical:In the recent years, there has been significant research interest in engineering the optical and spectral properties of materials through the use of photonic metasurfaces for efficient energy conversion, including thermophotovoltaics. The proposed program merges advanced photonic topology optimization with deep-learning-based inverse design methods and a comprehensive material database to unlock unorthodox optical designs for the realization of highly-efficient components for TPV applications. This effort will expand the design parameter space and incorporate machine-learning approaches to achieve the dramatic improvement of the speed and efficiency of topology optimization, as well as to build a large documented materials database. Through unconventional optical design, the program aims to develop highly efficient TPV energy conversion approaches by enhancing radiative heat transfer process. The proposed TPV device could enable unparalleled energy conversion efficiency potentially exceeding 50% by matching the emissivity of the emitter to the bandgap of commercial photovoltaic cells such as silicon, gallium antimonide, indium gallium arsenide. This approach could elevate nanophotonic designs into previously unavailable regimes and can be applied to photonic systems beyond TPV.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.