Defect Sampling for Electron Beam Review
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for electron beam review in semiconductor manufacturing face limitations due to the lack of accurate defect classification from optical inspection, leading to low capture rates and prolonged time in generating ground truth defect classifications, as they rely on limited information from optical inspection and review, making targeted DOI sampling difficult.
Innovation Solution
A computer-implemented method that combines defect attributes from optical inspection and optical review to separate defects into bins, enabling targeted sampling for electron beam review, thereby improving capture rates and reducing the time to achieve ground truth defect classifications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If optical inspection and review are used to classify defects before electron beam review, then the time required for electron beam review is reduced, but the accuracy and purity of defect classification deteriorates
Solution Approach 1:
The defect review process is segmented into multiple classification stages: optical inspection provides initial defect detection and basic classification, optical review performs secondary classification to filter and prioritize defects, and electron beam review focuses only on the most critical defects. This segmentation allows each tool to operate at its optimal capability level, reducing overall review time while maintaining classification accuracy through the cumulative effect of multiple filtering stages.
Solution Approach 2:
Optical inspection and optical review perform preliminary defect classification and filtering before electron beam review. By pre-processing the defect population with optical tools that can quickly identify and categorize defects, the system prepares the defect list in advance, allowing electron beam review to focus only on high-priority candidates. This preliminary action significantly reduces the time required for electron beam review without sacrificing final classification accuracy.
2Productivity
If optical inspection classifiers are used to generate defect classification results, then the process is faster, but the purity and accuracy of classification results deteriorates, particularly for defects near classification thresholds
Solution Approach 1:
The system merges the results from multiple classification tools (optical inspection classifier and optical review classifier) to produce a final defect classification. By combining the strengths of different classification approaches and cross-validating their results, the system achieves both high processing speed and high classification purity. Defects that are consistently classified by multiple tools are identified with higher confidence, while borderline cases are flagged for additional review.
Solution Approach 2:
The system implements feedback loops where electron beam review results are fed back to validate and refine the optical classification algorithms. Classification accuracy is continuously improved by using ground truth data from electron beam review to retrain and adjust optical inspection and review classifiers. This feedback mechanism ensures that as processing speed increases, classification purity also improves over time through iterative optimization.
3Ease of operation
If limited information from optical inspection is used for electron beam review sampling, then the system is simpler to operate, but the capture rate of actual defects deteriorates
Solution Approach 1:
Optical inspection and optical review perform preliminary defect identification and characterization before electron beam review. These tools gather and analyze defect information in advance, creating a prioritized defect list with relevant attributes and confidence scores. This preliminary action enriches the information available for sampling decisions, enabling the system to maintain high defect capture rates while keeping the operation simple through automated prioritization algorithms.
Solution Approach 2:
Optical review acts as an intermediary between optical inspection and electron beam review, bridging the information gap. It performs intermediate analysis to extract additional defect characteristics and validate optical inspection results, generating enhanced defect metadata that improves sampling accuracy. This intermediary layer ensures that electron beam review receives sufficient information for effective sampling without requiring direct complex integration between optical inspection and electron beam systems.
Data Source
AI summary
Various embodiments for generating a defect sample for electron beam review are provided. One method includes combining, on a defect-by-defect basis, one or more first attributes for defects determined by optical inspection of a wafer on which the defects were detected with one or more second attributes for the defects determined by optical review of the wafer thereby generating combined attributes for the defects. The method also includes separating the defects into bins based on the combined attributes for the defects. The bins correspond to different defect classifications. In addition, the method includes sampling one or more of the defects for the electron beam review based on the bins into which the defects have been separated thereby generating a defect review sample for the electron beam review.