Failure Pattern Prediction Using Optical Feature Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity of semiconductor patterning processes makes it difficult to accurately predict whether patterns will form correctly on a substrate, leading to increased difficulty in predicting patterning failures.
Innovation Solution
A method involving obtaining optical parameters, performing dimensional reduction, and generating a failure pattern prediction model through supervised learning to classify patterns as failures or successes in a virtual three-dimensional space.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional prediction methods are used for pattern formation, then the process is simpler, but the prediction accuracy is insufficient due to increasing process complexity
Solution Approach 1:
The patent segments the complex optical parameters into multiple key dimensions (e.g., intensity, slope, curvature) and processes each dimension separately through dimensional reduction. This allows the system to handle complex patterning processes by breaking them down into manageable components that can be individually analyzed and combined for accurate prediction.
Solution Approach 2:
The patent performs dimensional reduction on multi-dimensional optical parameters to extract the most significant features. By transforming high-dimensional optical parameter data into a reduced feature space, the system maintains prediction accuracy while simplifying the computational complexity of analyzing complex patterning processes.
2Measurement precision
If more optical parameters are analyzed to improve prediction accuracy, then the prediction becomes more accurate, but the computational complexity and data processing requirements increase
Solution Approach 1:
The patent extracts only the most critical and informative optical parameters from the complete set of available parameters. By identifying and extracting key features that have the greatest impact on pattern formation prediction, the system achieves high accuracy without the computational burden of processing all possible optical parameters.
Solution Approach 2:
The patent reduces the dimensionality of optical parameter data by transforming multi-dimensional parameter sets into a smaller number of representative features. This dimensional reduction simplifies data processing and analysis while preserving the essential information needed for accurate pattern formation prediction.
3Reliability
If comprehensive optical parameter analysis is performed to identify potential defects, then the yield improves, but the processing time and computational resources increase
Solution Approach 1:
The patent performs prediction analysis on optical parameters before actual mask fabrication and semiconductor manufacturing. By identifying potential pattern formation failures in advance through computational modeling and dimensional reduction, the system allows for design corrections to be made before committing to expensive and time-consuming manufacturing processes, thereby improving yield without extending development time.
Solution Approach 2:
The patent creates virtual models and simulations of the patterning process using optical parameter data. By working with computational copies and predictions rather than physical prototypes, the system can comprehensively analyze multiple scenarios and optimize designs rapidly without the time and resource constraints of physical experimentation.
Data Source
AI summary
A failure pattern prediction method includes obtaining first optical parameters of first patterns, extracting first data by performing a first dimensional reduction method on the first optical parameters and generating a failure pattern prediction model by performing supervised learning on the first patterns based on the first data.