TSOM Image Acquisition Method and Semiconductor Device Inspection Method

TSOM Semiconductor Inspection Method Combining Virtual Focus Image Generation and Deep Learning

This technology generates Through-Focus Scanning Optical Microscopy (TSOM) images for semiconductor inspection using a minimal number of actual captured images. By combining a few real focus images with FMM-based optical system analysis, it generates virtual images at various focus positions to create a complete TSOM image, which is then analyzed for defects using a deep learning-based inspection tool.

While TSOM allows for the measurement of sub-10nm 3D patterns using standard optical microscopes, it traditionally requires significant time and effort due to the need to repeatedly adjust the sample or lens position to capture images at different focal depths. This also makes it difficult to build the large-scale TSOM databases required for deep learning applications.

This technology calibrates optical transformation equations by comparing actual images with virtual images at the same focus, then generates precise TSOM images using densely arranged virtual images near the focal point. By using these as training data for deep learning, both inspection speed and accuracy are improved. It can be applied to semiconductor wafer pattern metrology and process defect inspection, enabling electron microscope-level detection using low-cost optical equipment.

‍

‍

Key Features:

• Acquiring actual images of the inspection target at different focal positions and their respective defocus distances using optical means
• Analyzing the optical system via FMM and obtaining virtual images at other focal positions using a transformation equation derived from comparing them with virtual images at the same focus as the actual images
• Obtaining a Through-Focus Scanning Optical Microscopy (TSOM) image of the inspection target using the actual and virtual images
• Determining an inspection tool through deep learning using TSOM images of semiconductor devices with known categories, and classifying the categories of unknown images

‍

This invention was developed with support from the Ministry of Trade, Industry and Energy for multi-focus optical/metrology inspection technology.

‍