Failure Rate Process Windows for Sparse Logic Pattern Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to accurately determine a process window for features in random logic structures during device manufacturing, particularly due to sparse occurrences and high variability, leading to inefficiencies in defect-based process control.
Innovation Solution
A method is developed to group features based on metrics like process window and critical dimension (CD) mean, generating a base failure rate model for each group, which is then used to create feature-specific models to identify critical features and optimize the patterning process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If features are analyzed individually in random logic structures, then feature-specific process optimization is achieved, but measurement data scarcity and high variability prevent accurate failure rate modeling
Solution Approach 1:
The patent combines multiple features into groups based on similarity metrics (critical dimension, process window, geometry) to aggregate measurement data. This merging enables statistically significant failure rate modeling by pooling data from multiple features while maintaining feature-specific optimization through group-based modeling.
Solution Approach 2:
The patent creates representative feature groups that copy the essential characteristics of individual features. By modeling failure rates at the group level and applying these models to individual features within the group, the system overcomes data scarcity while preserving feature-specific accuracy.
2Reliability
If process window is determined using traditional methods, then general process control is maintained, but sparse feature occurrences in random logic structures prevent accurate defect-based process window determination
Solution Approach 1:
The patent merges measurement data from multiple sparse features into grouped datasets, enabling statistically robust process window determination. By combining data across features with similar characteristics, the system achieves sufficient data quantity for accurate defect-based process window modeling.
Solution Approach 2:
The patent performs preliminary grouping of features based on similarity metrics before conducting failure rate analysis. This preliminary organization ensures that measurement data is aggregated in a way that maximizes statistical power and enables accurate process window determination even with limited data per individual feature.
3Measurement precision
If feature grouping is performed based on multiple metrics, then modeling accuracy is improved, but computational complexity and model generation time increase
Solution Approach 1:
The patent segments the feature grouping process into hierarchical stages: initial grouping by geometry, then refinement by critical dimension, and final adjustment by process window. This segmentation reduces computational complexity by breaking down the multi-metric clustering problem into manageable steps while maintaining modeling accuracy.
Solution Approach 2:
The patent applies a practical threshold approach where grouping is performed on the most critical metrics (geometry and critical dimension) with optional refinement on less critical metrics. This partial application of multiple metrics achieves sufficient modeling accuracy without the full computational burden of exhaustive multi-metric clustering.
Data Source
AI summary
A method for determining a process window of a patterning process based on a failure rate. The method includes obtaining a plurality of features printed on a substrate, grouping, based on a metric, the features into a plurality of groups, and generating, based on measurement data associated with a group of features, a base failure rate model for the group of features, wherein the base failure rate model identifies the process window related to the failure rate of the group of features. The method can further include generating, using the base failure rate model, a feature-specific failure rate model for a specific feature, wherein the feature-specific failure rate model identifies a feature-specific process window such that an estimated failure rate of the specific feature is below a specified threshold.