Repeater Defect Capture on Semiconductor Wafer
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect detection methods in semiconductor manufacturing, such as multi-die adaptive threshold (MDAT) and standard reference die (SRD), face challenges with high nuisance rates and low repeater capture rates due to their chronological inspection approach and inability to utilize repeater defect information for improved detectability, leading to inefficient defect analysis and long inspection times.
Innovation Solution
A method involving repeater analysis on a semiconductor wafer using a processor to identify repeater defects by obtaining signed difference images, calculating mean normalized values, and applying a repeater threshold to filter out non-repeater defects, which can be performed using a single or double detection algorithm, thereby enhancing defect detection efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If chronological inspection approach is used (MDAT, SRD), then inspection can be performed systematically, but nuisance rate increases and repeater capture rate decreases
Solution Approach 1:
Instead of inspecting swath after swath chronologically without revisiting, the invention inverts the approach by continuously returning to previously inspected regions to apply newly learned information. This allows the system to progressively refine defect detection by re-evaluating old data with updated knowledge, thereby improving repeater capture rate while maintaining systematic inspection
Solution Approach 2:
The invention implements feedback by using information learned from newly inspected swaths to re-analyze previously inspected swaths. This feedback loop allows the system to continuously improve its defect detection capability, reducing nuisance rates by eliminating false positives identified through iterative learning while maintaining high repeater capture rates
2Measurement precision
If hot scan is performed with aggressive detection settings, then representative sample of defects is collected, but billions of defect candidates are generated increasing analysis time
Solution Approach 1:
The invention extracts and isolates repeater defect patterns from the billions of defect candidates generated by hot scan. By identifying and separating repetitive defect patterns that originate from reticle defects, the system focuses analysis only on these extracted patterns, dramatically reducing the time required for defect analysis while maintaining high detection sensitivity
Solution Approach 2:
The invention changes the parameter of defect candidate evaluation by applying repeater analysis specifically to defect patterns showing periodicity. Instead of analyzing all billions of defect candidates equally, the system transforms the approach to identify and analyze only those defects that exhibit repeater characteristics, thereby reducing analysis time while maintaining comprehensive defect detection
Data Source
AI summary
Repeater analysis at a first threshold identifies repeater defects. The repeater defects are located at a coordinate that is the same on each reticle. Images on every reticle of the semiconductor wafer at the coordinate are received, and a plurality of signed difference images are obtained. A repeater threshold for signed difference images is calculated, as is consistency of the polarity. The threshold is applied to the images and a number of defects per each repeater that remain are determined. A secondary repeater threshold can be applied for nuisance filtering.