Imaging error compensation method for through-focus scanning optical microscopy images based on deep learning.

Area of Science:

• Optics
• Metrology
• Machine Learning

Background:

• Through-focus scanning optical microscopy (TSOM) is a nanoscale metrology technique combining microscopy and simulations.
• TSOM image quality can be compromised by mechanical vibrations and optical noise, impacting measurement accuracy.

Purpose of the Study:

• To develop and validate a deep learning-based method for compensating imaging errors in TSOM images.
• To improve the accuracy of nanoscale linewidth estimation using optimized TSOM images.

Main Methods:

• A deep learning approach utilizing the U-Net architecture was employed for image error compensation.
• The U-Net was trained using a supervised learning strategy, with simulated TSOM images serving as the ground truth.
• The network minimizes the difference between experimental and simulated images by iteratively updating weights and biases.

Main Results:

• The proposed U-Net based method effectively compensates for imaging errors in TSOM images.
• Optimized TSOM images derived from the compensation method show significantly enhanced accuracy for nanoscale linewidth estimation.

Conclusions:

• Deep learning with U-Net offers a robust solution for improving TSOM image quality and measurement accuracy.
• The developed compensation method demonstrates significant potential for advancing nanoscale metrology applications.