Machine Learning for Deep Trench Bottom Width Measurements using Scatterometry : AM: Advanced Metrology

We present a machine learning enhanced metrology method to measure the bottom width of deep trenches (about 42 µm in depth) using optical scatterometry. For this study, 2D line trenches as well as circular 3D trenches with varying trench side-wall angles were investigated. A combination of reference sets from SEM cross-sections, inline CD-SEM as well as depth measurements obtained from reflectance fringes are used to train the machine learning model. The results are cross verified against SEM cross-sections.