Terahertz Nondestructive Measurement of Multilayer Films via Polarization Modeling Driven Deep Learning

In modern film application technology, according to the different characteristics of each layer of film, the structure of multilayer films is often used to improve the overall comprehensive performance of the films, but the loss of multilayer films in the long-term use process will affect the performance of the multilayer films, thus affecting the system involved in the films. Therefore, it is of great significance to measure the thickness and optical constants of each layer of multilayer films. The existing thin film measurement technology usually uses multi-wavelength wide spectrum measurement methods to measure the optical constants or thickness of the thin film, but the existing technology cannot measure the thickness and optical constants of each layer of the multilayer films at the same time. Here, utilizing the deep penetration and nondestructive nature of terahertz radiation, we present an integrated methodology combining photophysical modeling with deep neural network architecture. Through systematic experimental validation, our single-wavelength polarization measurement framework enables real-time, high precision determination of each layer thickness and optical constants within multilayer films structure.