Adaptive Tile Statistics for Large Defect and Misalignment Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect detection methods in semiconductor manufacturing face challenges with fixed threshold max (Tmax) settings that fail to detect large defects and image misalignment, leading to undetected defects and sensitivity loss due to inappropriate parameter settings and silent misalignment.
Innovation Solution
A system and method that dynamically adjusts threshold max (Tmax) and detects image misalignment by analyzing statistical data from tiles in difference image frames, identifying outlier tiles, and applying adaptive defect detection parameters to enhance defect detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a fixed Tmax value is used for defect detection, then the inspection process is simple and fast, but large defects and populations of small defects cannot be detected accurately
Solution Approach 1:
The patent implements dynamic Tmax adjustment by dividing the wafer into multiple tiles and calculating separate Tmax values for each tile based on local signal cloud characteristics. This allows the inspection system to adapt to varying defect densities and sizes across different regions, improving detection accuracy without requiring a complete wafer scan for parameter setting.
Solution Approach 2:
The patent segments the wafer into multiple tiles and processes each tile independently with its own Tmax value. This segmentation allows localized defect detection parameters to be optimized for each region, enabling accurate detection of both large defects and populations of small defects while maintaining computational efficiency.
2Measurement precision
If Tmax is set low to detect large defects, then large defects can be detected, but the inspection time increases and production yield decreases
Solution Approach 1:
By dividing the wafer into tiles and processing them independently with localized Tmax values, the system can use lower Tmax values only where needed (in tiles with large defects) rather than across the entire wafer, maintaining high detection sensitivity while preserving overall inspection throughput.
Solution Approach 2:
The patent performs preliminary tile-based statistical analysis to identify regions with large defects before applying detailed inspection. This allows the system to focus computational resources on problematic areas, reducing overall inspection time while maintaining sensitivity for large defect detection.
3Reliability
If image alignment search range is limited, then the inspection process is fast, but misaligned frames cannot be detected
Solution Approach 1:
The patent segments the alignment verification process by examining individual tiles for misalignment characteristics. This allows rapid identification of misaligned frames through tile-level statistical analysis without requiring exhaustive full-frame alignment searches, maintaining fast inspection speed while improving misalignment detection reliability.
4Measurement precision
If full wafer scanning is performed to set optimal parameters, then accurate defect detection is achieved, but inspection time increases significantly
Solution Approach 1:
The patent performs parameter optimization at the tile level rather than requiring full wafer scanning. By analyzing statistical properties of individual tiles, the system can determine appropriate Tmax values locally without time-consuming comprehensive wafer examination, achieving accurate parameter optimization efficiently.
Solution Approach 2:
The patent performs preliminary statistical analysis on tile data to establish defect detection parameters before the main inspection process. This preliminary parameter determination based on tile-level statistics enables accurate defect detection without requiring time-consuming full wafer scans for parameter optimization.
Data Source
AI summary
Methods and systems for detecting defects on a specimen are provided. One method includes generating one or more statistics for each of multiple tiles in difference image frames generated for a job of the images. The method also includes identifying outlier tiles in the multiple tiles based on the statistic(s) generated for each of the multiple tiles. The method further includes determining one or more defect detection parameters of a defect detection method for each of the difference image frames based on the identified outlier tiles and detecting defects on the specimen with the determined defect detection parameter(s). The embodiments described herein are particularly suitable for identifying large defects, grossly defective image frames, and image misalignment in difference image generation that can be missed by currently used inspection methods and systems.