2026-2-25

Occlusion-aware pattern measurement for see-through BSE images of HV-SEM via contour- and pixel-based approaches

• Event: SPIE Advanced Lithography + Patterning 2026
• Shinji Murakami, Yoshihiro Midoh, Masahiro Oya, Yosuke Okamoto, Shinichi Nakazawa, Kotaro Maruyama, Yuichiro Yamazaki, Noriyuki Miura
  
  Proc. SPIE 13981, Metrology, Inspection, and Process Control XL (25 Feb 2026)

ABSTRACT

As semiconductor devices become increasingly complex, direct measurement of overlay and edge placement error (EPE) on actual device patterns is required. We developed a high-voltage SEM-based contour extraction framework for occluded patterns in see-through BSE images and compared a contour-based method (CVNet) with a pixel-based segmentation method (UNet+FCN). Using synthetic SEM datasets with realistic variations in noise, contrast, and occlusion, both methods successfully extracted hidden patterns. The pixel-based approach achieved higher region extraction accuracy, while the contour-based approach provided superior boundary precision, indicating that method selection should be optimized according to measurement objectives.

Key words: High voltage, EBI, SEM, D2DB, CD, EPE, Overlay, CVNet