Multi-Step Image Subtraction for Semiconductor Defect Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As semiconductor feature sizes shrink, defect inspection systems face challenges with weak measurable signals and lower signal-to-noise ratios (SNRs) due to noise from previous process steps, limiting accuracy and reliability in defect detection.
Innovation Solution
The system generates a multi-step difference image by weighted subtraction of images from current and previous process steps, using a controller to identify defects by isolating noise from previous layers, thereby improving SNR and detection sensitivity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If feature sizes are shrunk to continue scaling, then device performance is improved, but signal-to-noise ratio deteriorates due to weaker defect signals
Solution Approach 1:
The patent extracts and removes noise components from previous process steps by performing subtraction of images from previous steps from current step images. This isolates the defect signals from the overwhelming background noise, enabling detection of defects even as feature sizes shrink and signals weaken.
Solution Approach 2:
The patent performs preliminary noise characterization by capturing and processing images from previous process steps before the actual defect inspection. By pre-identifying and subtracting noise patterns from earlier steps, the system prepares a cleaned reference for subsequent defect detection, improving sensitivity to weak signals.
2Reliability
If noise from previous process steps is included in inspection, then inspection coverage is maintained, but noise level increases reducing detection accuracy
Solution Approach 1:
The patent selectively extracts and removes noise components from previous process steps by performing image subtraction. This maintains the beneficial coverage of inspecting multiple steps while eliminating the harmful noise that reduces detection accuracy, allowing defects to be clearly identified against a cleaned background.
Solution Approach 2:
The patent converts the harmful noise from previous process steps into a useful reference by capturing, analyzing, and subtracting these noise patterns. The previously harmful noise becomes a subtracted reference that enhances defect detection, transforming the problem of noise into a solution for improved signal clarity.
3Measurement precision
If multiple images from current and previous steps are processed, then noise reduction is improved, but processing complexity increases
Solution Approach 1:
The patent segments the image processing into distinct, manageable steps: capturing images from multiple process steps, characterizing noise from previous steps, subtracting noise images from current step images, and identifying defects in the difference images. This segmentation makes the complex multi-step processing more manageable and implementable.
Data Source
AI summary
An inspection system may generate first-step images of multiple sample regions after a first process step and generate second-step images of the sample regions after a second process step, where the second process step modifies the sample in at least one of the sample regions. The system may further identify one of the sample regions as a test region and at least some of the remaining sample regions as comparison regions, where the second-step image of the test region is a test image and the second-step images of the comparison regions are comparison images. The system may further generate a multi-step difference image by subtracting a combination of at least one of the second-step comparison images and at least two of the first-step images from the test image. The system may further identify defects in the test region associated with the second process step based on the multi-step difference image.