Dynamic Binning for Wafer Defect Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect sampling methods struggle to achieve maximum defect type diversity, especially as the dimensionality of the feature space increases, leading to a decrease in effectiveness in achieving multiple diversification objectives simultaneously.
Innovation Solution
A computer-implemented method that acquires inspection results for a wafer, identifies diverse attribute values, generates bins based on these values, separates defects into bins, and selects defects within each bin based on diversity in signal attributes, allowing for dynamic binning and intuitive configuration to achieve multiple diversification objectives effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If traditional defect sampling methods are used, then the sampling process is simple, but the defect type diversity is insufficient especially in high-dimensional feature spaces
Solution Approach 1:
The patent segments the defect sampling process into multiple stages: initial diverse defect selection, feature space analysis, dynamic binning creation, and iterative refinement. This segmentation allows the system to achieve comprehensive defect type diversity by systematically exploring different regions of the feature space rather than using a single-step sampling approach.
Solution Approach 2:
The patent implements dynamic binning where the bins are not fixed but adaptively created and adjusted based on the defect population characteristics. The binning structure evolves iteratively as defects are selected and removed, allowing the sampling process to dynamically respond to the underlying defect distribution and maintain effectiveness in high-dimensional feature spaces.
2Measurement precision
If more defect attributes are considered to achieve better diversification, then the defect type identification improves, but the computational complexity increases
Solution Approach 1:
The patent extracts and focuses on the most discriminative feature combinations that are most useful for defect type identification. Rather than processing all possible feature combinations, the system identifies and extracts the key features that provide the greatest diversification benefit, reducing computational complexity while maintaining identification accuracy.
Solution Approach 2:
The patent applies partial action by selecting a subset of defects and features that are sufficient to achieve the diversification objective. The iterative process stops when adequate defect type coverage is achieved, avoiding the excessive computation that would result from exhaustively processing all defects and features.
3Productivity
If static binning is used, then the process is computationally efficient, but it cannot adapt to high-dimensional feature spaces effectively
Solution Approach 1:
The patent transforms the static binning approach into a dynamic one where bins are created and adjusted based on the actual defect population and feature space characteristics. This dynamic adaptation allows the system to maintain computational efficiency while effectively handling high-dimensional feature spaces, as the binning structure evolves to match the data distribution.
Solution Approach 2:
The dynamic binning process is self-adjusting, automatically adapting to the defect population characteristics without requiring external intervention. The system serves itself by iteratively refining the binning based on the defects it selects, making the process both efficient and adaptable to high-dimensional feature spaces.
Data Source
AI summary
Methods and systems for generating a defect sample for a wafer are provided. One method includes separating defects detected on a wafer into bins having diversity in values of a first set of one or more first attributes of the defects. The method also includes selecting, independently from one or more of the bins, defects within the bins based on diversity in a second set of one or more second attributes of the defects. The selected defects are then used to create a defect sample for the wafer. In this manner, defects having diverse values of multiple attributes can be easily selected.