Defect Classification Method Using Randomized Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for distinguishing between systematic and random defects in semiconductor manufacturing require user intervention and are prone to inaccuracies, often misidentifying random defects as systematic due to higher defect counts and reliance on user-defined thresholds.
Innovation Solution
A computer-implemented method that compares the number of actual defects to randomly generated defects, using Pareto charts and statistical analysis to determine if defects are systematically or randomly distributed without user intervention, by normalizing defect counts and generating defects with similar distributions across the wafer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If user-defined thresholds are used to identify systematic defects, then the method is simple to implement, but it requires user intervention and judgment leading to inaccuracies
Solution Approach 1:
The system performs self-service by automatically generating random defect groups and comparing them with actual defect groups to determine systematic defects without requiring user intervention. The computer system autonomously executes the entire classification process using predefined algorithms and statistical comparisons.
Solution Approach 2:
The patent replaces the mechanical/manual process of user-defined threshold setting with an automated computer-based system that uses statistical algorithms. The mechanical action of manual threshold selection is substituted by computational methods that automatically generate and compare defect distributions.
2Ease of manufacture
If defect counts are used to identify systematic defects, then the method is straightforward, but random defects with higher counts are misidentified as systematic
Solution Approach 1:
The patent segments defects into different groups based on their spatial distribution patterns and compares each group's characteristics. By dividing the defect population into systematically distributed groups and randomly distributed groups, the method can accurately distinguish between the two types without relying solely on total defect counts.
Solution Approach 2:
The system creates a copy of random defect groups with similar characteristics to actual defect groups. By generating synthetic random defect datasets that mirror the actual defect distribution patterns, the system can compare and distinguish real systematic defects from random variations without bias toward higher-count defects.
3Productivity
If all defects are corrected to eliminate yield loss, then manufacturing yield improves, but the cost and complexity of corrections increases significantly
Solution Approach 1:
The patent extracts and isolates only the systematic defects from the total defect population through automated classification. By separating systematic defects (which require correction) from random defects (which do not require correction), the system enables targeted correction efforts that improve yield without the unnecessary complexity of correcting all defects.
Data Source
AI summary
Various computer-implemented methods for determining if actual defects are potentially systematic defects or potentially random defects are provided. One computer-implemented method for determining if actual defects are potentially systematic defects or potentially random defects includes comparing a number of actual defects in a group to a number of randomly generated defects in a group. The actual defects are detected on a wafer. A portion of a design on the wafer proximate a location of each of the actual defects in the group and each of the randomly generated defects in the group is substantially the same. The method also includes determining if the actual defects in the group are potentially systematic defects or potentially random defects based on results of the comparing step.