DUV-Vis-IR Optical Metrology Enhancements with Machine Learning

Machine learning algorithms can supplement traditional data analysis methods and enable new capabilities for optical metrology systems. This includes measurements of semiconductor processes that are difficult or not possible with the standard approach. By integrating standard models with machine learning, we leverage the strengths of both methodologies. This paper will provide three application examples, using UV-Vis-IR spectroscopic reflectance data, where machine learning adds significant value: 1) Lead (Pb) classification within a glass frit bonding layer, 2) Phosphorus content measurement within a TEOS film, and 3) CD measurements for a Deep Trench Isolation (DTI) structure.