Smart ellipsometry: artificial intelligence for thin film characterization

Spectroscopic ellipsometry (SE) has been developed as almost the standard tool for characterizing thin films. Nevertheless, deducting unknown sample properties can only be achieved by solving an inverse problem, and thus ellipsometry data analysis has been recognized as a major challenge. We have developed a Smart Ellipsometry Network (SENet) for automated determination of material optical constants. SENet adopts a physics-informed learning-based inference and is capable of self-tuning its network weights to correct the prediction deviations. During the experiments on a variety of thin films covering metals, dielectrics, and semiconductors in an ultraviolet-visible-near-infrared broad spectral range, SENet exhibited remarkable generalization in various scenarios without the need to retrain or reconstruct the network and the intervention of human-experts. The developed method opens a new avenue for automated, accurate and fast determination of material optical constants in a single click manner.