2024-11-12

CVNet を用いた頑健なパターン輪郭抽出法の検討

• Event: 第44回ナノテスティングシンポジウム(NANOTS2024)
• 村上慎治, 大家政洋, 岡本陽介, 中澤伸一, 丸山浩太郎, 山崎裕一郎, 御堂義博, 三浦典之
  
  第44回ナノテスティングシンポジウム(NANOTS2024), pp.18–23, Nov. 2024.

ABSTRACT

We evaluated the applicability of CVNet, a contour extraction method that combines the classical Active Contour Model with deep learning, for pattern metrology using high-voltage SEM images. The study focused on lower-layer patterns affected by process variations and challenging imaging conditions, including low signal-to-noise ratio, low contrast, and multilayer pattern overlap. Using synthetic SEM model images, we assessed the robustness of the proposed method and its suitability for overlay measurement. The results indicate that CVNet effectively tolerates these disturbances and enables accurate contour extraction for overlay metrology. In addition, a novel loss function improved contour-tracking stability and simplified hyperparameter optimization.

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