Abstract
Abstract
Scanning Electron Microscopy (SEM) leverages electron wavelengths for nanoscale imaging. Achieving optimal imaging requires precise adjustment of parameters like focus, stigmator, and aperture alignment, and therefore a process traditionally relies on skilled personnel and time-consuming adjustments. Existing auto-focus (AF) and auto-stigmation (AS) methods face challenges due to the interdependent nature of these parameters and sample diversity. This paper introduces a novel beam kernel estimation method, designed to independently optimize SEM parameters, irrespective of sample variations. Our approach disentangles the mutual influences among parameters, enabling concurrent optimization of focus, stigmator x, y, and aperture-align x, y. This method demonstrates robust performance, yielding average errors of 1.00µm for focus, 0.30% for stigmators, and 2.28% for aperture alignment, significantly outperforming the sharpness-based approach with its average errors of 6.42µm for focus and 2.32% for stigmators, and lacking in aperture-align capabilities. The key innovation of our approach lies in its ability to address the complex interplay of SEM parameters through a blind deconvolution model, facilitating rapid and automated optimization. This advancement not only enhances the precision and efficiency of SEM operations but also broadens its applicability across various scientific and industrial fields.
Publisher
Research Square Platform LLC
Reference60 articles.
1. Electron microscopy in semiconductor inspection;Nakamae K;Meas. Sci. Technol.,2021
2. Automatic recognition of defect areas on a semiconductor wafer using multiple scanning electron microscope images;Nakagaki R;Meas. Sci. Technol.,2009
3. Failure localization with active and passive voltage contrast in FIB and SEM;Rosenkranz R;J Mater Sci: Mater Electron,2011
4. Newell, T., Tillotson, B., Pearl, H. & Miller, A. Detection of electrical defects with SEMVision in semiconductor production mode manufacturing. in 2016 27th Annual SEMI Advanced Semiconductor Manufacturing Conference (ASMC) 151–156 (2016). doi:10.1109/ASMC.2016.7491149.
5. Sun, W., Goto, Y., Yamamoto, T. & Hitomi, K. Evaluation of deep learning model for 3D profiling of HAR features using high-voltage CD-SEM. in Metrology, Inspection, and Process Control for Semiconductor Manufacturing XXXV vol. 11611 160–166 (SPIE, 2021).