Emissivity prediction framework for freestanding multilayer nanomembranes

Freestanding multilayer nanomembranes serve as essential components in emerging applications, including absorber membranes for infrared (IR) bolometers, extreme ultraviolet pellicles, and X-ray/electron transmission windows. Owing to their extremely low heat capacity and limited conductive/convective cooling, their thermal reliability strongly depends on precise emissivity control. However, the optical constants of ultrathin metallic films vary nonlinearly with thickness, which complicates the design of multilayer radiative structures. In this study, we constructed and experimentally validated a framework for predicting the emissivity of multilayer nanomembranes based on the thickness-dependent optical constants of their metallic sublayers. To develop this framework, we fabricated freestanding Ru and Mo nanomembranes and extracted optical constants from mid-IR measurements. We then developed a physics-regularized neural network to expand the extracted constants to Å-scale thickness resolution while preserving physical consistency. Using the expanded optical constants, we simulated Ru/SiN multilayer stacks and experimentally confirmed that emissivity can be significantly modulated by sublayer configuration even at a same nominal total thickness. These results show that resolving the nonlinear optical behavior of ultrathin metallic films enables multilayer architecture to serve as an effective design lever for thermal radiation control in multilayer nanomembranes.