Patch-Based Defect Detection in Semiconductor Wafer Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect detection techniques in semiconductor wafer inspection are inadequate for high precision and uniformity, particularly in identifying defects in submicron features, as they rely on low-resolution initial inspections and require multiple phases with potential alignment errors.
Innovation Solution
A computerized method and system for inspecting articles, which processes inspection images by defining source patches in reference images, determining similarity levels with reference patches, and rating inspected pixels based on these similarities to detect defects with improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional two-phase inspection procedures are used, then defect detection is performed, but the process complexity and inspection time increase
Solution Approach 1:
The inspection process is segmented into distinct phases: a first phase for rapid low-resolution defect mapping, and a second phase for thorough high-resolution analysis of suspected locations only. This segmentation allows the system to maintain high defect detection accuracy while reducing overall process complexity by avoiding full high-resolution inspection of the entire wafer surface.
Solution Approach 2:
The first phase performs preliminary defect mapping at low resolution to identify suspected defect locations before committing resources to detailed high-resolution inspection. This preliminary action filters out large areas of the wafer that are likely defect-free, reducing the complexity and time of the subsequent detailed inspection phase.
2Measurement precision
If high-resolution inspection is applied to the entire wafer surface, then defect detection precision improves, but inspection time and processing complexity increase significantly
Solution Approach 1:
The wafer inspection is divided into two resolution phases applied to different spatial regions. The first phase covers the entire wafer at low resolution to identify defect candidates, while the second phase applies high resolution only to small suspected regions. This segmentation maintains high measurement precision for defects while dramatically reducing total inspection time compared to uniform high-resolution inspection.
Solution Approach 2:
Instead of applying full high-resolution inspection to the entire wafer (excessive action), the system applies high resolution only partially to suspected defect regions. This partial action achieves the necessary measurement precision for defect detection while avoiding the time penalty of inspecting the entire wafer surface at high resolution.
3Reliability
If multiple detectors are used for two-phase inspection, then defect detection capability improves, but device complexity and cost increase
Solution Approach 1:
The inspection system uses a single multi-functional detector that can operate in two distinct modes: a first mode for rapid low-resolution defect mapping and a second mode for detailed high-resolution analysis. This universal detector performs both inspection phases, eliminating the need for separate detector systems and reducing overall device complexity while maintaining improved defect detection capability.
Solution Approach 2:
The detector operates dynamically, switching between two operational phases with different resolution and speed characteristics. The system adapts its detection parameters based on the inspection phase and region being examined, allowing a single detector to replace what would traditionally require multiple specialized detectors, thereby reducing device complexity.
Data Source
AI summary
A system capable of inspecting an article for defects, the system including: a patch comparator, configured to determine with respect to each of a plurality of reference patches in a reference image a similarity level, based on a predefined patch-similarity criterion and on a source patch defined in the reference image; an evaluation module, configured to rate each inspected pixel out of multiple inspected pixels of the inspection image with a representative score which is based on the similarity level of a reference patch associated with a reference pixel corresponding to the inspected pixel; a selection module, configured to select multiple selected inspected pixels based on the representative scores of the multiple inspected pixels; and a defect detection module, configured to determine a presence of a defect in the candidate pixel based on an inspected value of the candidate pixel and inspected values of the selected inspected pixels.