Intelligent Defect Classification and Sampling for Semiconductor Yield
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional defect inspection methods in semiconductor manufacturing struggle to effectively identify and classify systematic defects, particularly in design layout patterns, leading to yield reduction and potential fatal errors due to the inability to recognize weak defect patterns early in the manufacturing process.
Innovation Solution
A computer-implemented intelligent defect classification and sampling system that uses a software and hardware framework to screen and match weak defect patterns through defect inspection data analysis, employing techniques like process window qualification (PWQ) and focus exposure matrix (FEM) to select critical defect samples and verify systematic defects, utilizing statistical process control (SPC) probability indices and critical area analysis (CAA) to prioritize defect sampling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional defect inspection methods are used, then all defects are detected, but weak systematic defects cannot be effectively identified and classified
Solution Approach 1:
The patent segments defects into different categories (systematic defects vs. random defects) and further divides systematic defects into priority levels (first priority, second priority, etc.). This segmentation allows the inspection system to focus on identifying and classifying weak systematic defects separately from other defect types, improving recognition accuracy for these critical defects.
Solution Approach 2:
The patent changes the inspection parameters by introducing priority-based classification criteria and systematic defect identification algorithms. By adjusting the inspection focus from generic defect detection to targeted systematic defect identification with priority levels, the system overcomes the difficulty of detecting weak systematic defects that conventional methods miss.
2Reliability
If comprehensive defect inspection is performed, then all defect types are detected, but manufacturing efficiency decreases due to lack of prioritization
Solution Approach 1:
The patent performs preliminary classification of defects into priority levels before detailed analysis and corrective actions are taken. By pre-identifying first priority systematic defects and second priority defects, the system enables manufacturers to focus resources on the most critical issues first, protecting yield while improving manufacturing efficiency through prioritized response.
Solution Approach 2:
The patent implements a feedback mechanism where defect inspection results are classified by priority and fed back to the manufacturing process. This allows systematic defects to be identified and addressed in priority order, with first priority defects receiving immediate attention and second priority defects addressed subsequently, balancing yield protection with manufacturing efficiency.
3Loss of information
If detailed analysis of all defects is conducted, then complete defect characterization is achieved, but time consumption increases significantly
Solution Approach 1:
The patent extracts and prioritizes critical defect information by classifying defects into priority levels. Instead of conducting detailed analysis on all defects equally, the system extracts and focuses detailed characterization on first priority systematic defects and second priority defects, while treating other defects with less intensive analysis, thus reducing overall inspection time while maintaining information completeness for critical defects.
Data Source
AI summary
Disclosure is related to a method and a system for intelligent defect classification and sampling, and a computer-readable storage device. The computer-implemented method acquires in-line defect inspection file, and retrieves the defect patterns over a device under test, e.g. a wafer from a fab. The system incorporates a defect pattern recognition engine to recognize the defect signature patterns from the defect patterns. A sampling scheme is performed to acquire weak defect patterns. A critical area analysis based on failure probability of weak patterns is incorporated to performing the sampling. The defect layout pattern groups probably causing the open or short failure can be obtained. The defect signature patterns through sampling are then displayed using a browsing system. Through a user interface, the user can perform functions, such as filtering, selection and merging, onto the defect patterns.