EllipsoNet—Deep-Learning-Enabled Optical Ellipsometry for Complex Thin Films

Analysis of optical spectroscopy data often requires intensive model fitting. Reflectometry and ellipsometry are commonly used methods to measure the optical dielectric functions or the complex refractive indices of optical thin films such as 2D materials. However, the available fitting models are extremely computational intense, require calibrations from human-expert and very specific to the optical structures of the samples. The substrate structures need to be simple (a single, thick, and transparent substrate is ideal) and perfectly defined. In addition, ellipsometry also requires expensive setup. In this study, we developed a deep learning method based on an encoder-decoder convolutional neural network that is capable of extracting refractive indices of thin films including 2D materials on arbitrary complex multilayer substrates named EllipsoNet. EllipsoNet is trained with numerically generated data. Without any prior knowledge of stack materials or human intervention, EllipsoNet can predict the complex refractive indices of thin films from experimentally obtained optical reflectance with high accuracies. Kramers-Kronig relations are spontaneously learned by the model without purposely teaching. This approach enables the in-situ optical characterization of functional materials and components in actual complex optoelectronic devices, a task previously not feasible with traditional reflectometry or ellipsometry methods.