Presentation Information
[8a-N102-6]Automated Analysis in Spectroscopic Ellipsometry Using Large Language Model
〇Hitoshi Nakatani1 (1.HORIBA, Ltd.)
Keywords:
ellipsometry,large language model
Spectroscopic ellipsometry enables highly accurate evaluation of the optical properties of thin films, but its analysis requires specialized expertise. Although automated analysis methods have been reported, they do not use optical models and therefore cannot explicitly handle material-specific optical responses. In this study, we propose a method that uses a large language model to construct an optical model from measurement conditions and material information, and estimates its parameters by combining machine learning with fitting. Applied to five thin films on Si substrates, the proposed method achieved a relative difference of less than 0.5% compared with analysis results obtained by an expert.