Spatial Filter Algorithm for Noisy Semiconductor Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect detection methods in semiconductor wafer inspection are inaccurate for detecting two-dimensional defects in noisy inspection data, as they rely solely on signal strength at individual points, leading to missed detections and misclassification due to background noise.
Innovation Solution
A computer-implemented method applying a spatial filter algorithm to inspection data across a substrate area to identify regions with higher probability of being a selected type of non-point defect, generating a two-dimensional map, and searching for spatial characteristics matching the defect, thereby enhancing signal-to-noise ratio and accurately detecting faint two-dimensional defects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If simple one-dimensional scans with signal thresholding are used for defect detection, then the detection method is simple and fast, but the detection accuracy deteriorates in noisy inspection data causing faint defects to be lost
Solution Approach 1:
The patent transitions from one-dimensional scan data to two-dimensional spatial maps, allowing defects to be visualized and detected in their true spatial context. This dimensional transformation enables the preservation of spatial relationships between data points, making it possible to distinguish faint defects from noise by analyzing their spatial patterns rather than relying solely on signal thresholding at individual points.
Solution Approach 2:
The patent combines multiple raw inspection data points corresponding to substantially the same locations on the substrate to generate integrated inspection data. This merging process consolidates signal information across multiple measurements, improving the signal-to-noise ratio and enabling more reliable defect detection in noisy environments while maintaining detection efficiency.
2Measurement precision
If point defect detection methods are used, then individual defects can be identified, but two-dimensional defects are treated as disconnected points leading to misclassification
Solution Approach 1:
The patent creates two-dimensional spatial maps that preserve the continuous spatial relationships of defects across the substrate surface. This allows extended two-dimensional defects to maintain their spatial coherence and be recognized as unified defect structures rather than disconnected point defects, enabling accurate defect type classification.
Solution Approach 2:
The patent introduces spatial filtering as an intermediary processing step between raw data acquisition and defect classification. This spatial filter algorithm acts as a mediator that enhances the spatial continuity information in the inspection data, allowing the system to recognize the true two-dimensional nature of extended defects while filtering out noise and disconnected artifacts.
3Measurement precision
If multiple raw inspection data are combined to generate inspection data, then the signal-to-noise ratio improves, but the processing complexity increases
Solution Approach 1:
The patent performs preliminary combination of multiple raw inspection data points before the main defect detection and classification processes. By pre-integrating the signal information from multiple measurements at substantially the same locations, the system improves the signal-to-noise ratio in advance, reducing the need for complex noise filtering and enhancement operations in subsequent processing stages.
Solution Approach 2:
The patent introduces a spatial filter algorithm as an intermediary that efficiently processes the combined inspection data to generate two-dimensional spatial maps. This spatial filter serves as a computationally efficient mediator that enhances spatial patterns while suppressing noise, achieving improved signal-to-noise ratio without requiring excessively complex processing operations.
Data Source
AI summary
Methods and systems for detection of selected defects in relatively noisy inspection data are provided. One method includes applying a spatial filter algorithm to inspection data acquired across an area on a substrate to determine a first portion of the inspection data that has a higher probability of being a selected type of defect than a second portion of the inspection data. The selected type of defect includes a non-point defect. The inspection data is generated by combining two or more raw inspection data corresponding to substantially the same locations on the substrate. The method also includes generating a two-dimensional map illustrating the first portion of the inspection data. The method further includes searching the two-dimensional map for an event that has spatial characteristics that approximately match spatial characteristics of the selected type of defect and determining if the event corresponds to a defect having the selected type.