Semantic Segmentation for Semiconductor Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect inspection and metrology processes in semiconductor manufacturing rely on reference images, which can introduce noise and make it difficult to detect defects, especially in cases like repeater defects where a suitable reference image is hard to obtain.
Innovation Solution
A system and method utilizing a semantic segmentation model to assign labels to each pixel in a specimen image without requiring a reference image, allowing for defect detection and information determination directly from the specimen image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If reference images are used in defect inspection and metrology processes, then measurement and detection can be performed, but noise is introduced and sensitivity is reduced
Solution Approach 1:
The patent extracts and removes the reference image component from the inspection system, transitioning from a two-image comparison approach (test image vs. reference image) to a single-image analysis approach. This extraction eliminates the noise source inherent in reference images while preserving the essential defect detection functionality through semantic segmentation of the test image alone.
Solution Approach 2:
The patent introduces semantic segmentation as an intermediary computational process that bridges the gap between raw test image data and defect identification. This intermediary layer processes the test image to extract meaningful features and patterns without requiring reference image subtraction, thereby maintaining measurement precision while eliminating noise introduction.
2Reliability
If reference images are used for defect detection, then defects can be identified, but it becomes difficult to detect repeater defects where suitable reference images are hard to obtain
Solution Approach 1:
The patent enables the inspection system to perform defect detection using only the test image itself, without external reference images. The semantic segmentation model processes the test image to identify defects autonomously, making the system self-sufficient and particularly effective for repeater defects where reference images cannot be obtained or are unsuitable.
3Productivity
If traditional defect inspection algorithms are used, then processing can be performed, but the system requires reference images that introduce complexity and limitations
Solution Approach 1:
The patent removes the reference image requirement from the inspection algorithm, simplifying the system architecture from a two-component input system (test image + reference image) to a single-component input system (test image only). This extraction reduces complexity while maintaining productivity through efficient semantic segmentation processing.
Data Source
AI summary
Methods and systems for determining information for a specimen are provided. One system includes a computer subsystem and one or more components executed by the computer subsystem. The one or more components include a semantic segmentation model configured for assigning labels to each of multiple pixels in an image responsive to what is represented in each of the multiple pixels. The image is an image of a specimen generated by an imaging subsystem. The computer subsystem is configured for determining information for the specimen from the assigned labels and without a reference image for the specimen.