From Materials to Machine Learning: A Photonics-Based Design of Thermophotovoltaic Emitters

Thermophotovoltaics (TPV) generate electricity through the conversion of radiation from an optical emitter, whose emissive spectra can be shaped to optimize efficiency. Among proposed emitter designs, thin-film multilayers offer a practical balance of spectral control, thermal stability, and manufacturability. In this Perspective, we first analyze the theoretical limits of spectral shaping on TPV efficiency and power, second we examine real materials as bulk emitters. Third, we emphasize how multilayer coatings with high refractive index contrast and aperiodic thicknesses enhance TPV performance by mitigating spectral losses. Fourth, we highlight machine learning as a scalable tool for navigating the multilayer parameter space through a representative comparison against a traditional algorithm. Finally, we discuss practical considerations for implementing emitters and further potential of machine learning for TPV. Together, these insights outline a materials- and photonics-driven pathway for next-generation TPV systems, where selective substrates, robust coatings, and data-driven optimizations push device efficiencies toward their fundamental limits.