Process Model Accuracy via 2-D Pattern Kernel Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional process models in semiconductor manufacturing struggle to accurately differentiate between 1-D and 2-D patterns, leading to inaccuracies in modeling semiconductor manufacturing processes, particularly in photolithography, which affects the efficacy of optical proximity correction and resolution enhancement techniques.
Innovation Solution
A system that generates an improved process model by using a 2-D-pattern detecting kernel, which differentiates between 1-D and 2-D regions in a layout, by categorizing empirical data and fitting kernels to both 1-D and 2-D patterns, enhancing the model's accuracy for both types of patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional process models are used without 2-D-pattern detecting kernels, then the model structure is simpler and easier to implement, but the manufacturing precision deteriorates due to inability to accurately differentiate between 1-D and 2-D patterns
Solution Approach 1:
The patent segments the pattern recognition function by introducing separate 1-D pattern detecting kernels and 2-D pattern detecting kernels. This segmentation allows the model to handle different pattern types with specialized kernels, improving manufacturing precision by accurately differentiating between 1-D and 2-D patterns while maintaining manageable model complexity through modular kernel design
Solution Approach 2:
The patent transitions from conventional single-dimension pattern modeling to multi-dimensional pattern modeling by incorporating 2-D pattern detecting kernels. This dimensionality change enables the model to capture two-dimensional pattern characteristics that were previously inaccessible, significantly improving process model accuracy for complex semiconductor patterns
2Measurement precision
If a single unified process model is used for both 1-D and 2-D patterns, then the model structure is simpler, but the measurement precision deteriorates due to inability to capture pattern-specific characteristics
Solution Approach 1:
The patent applies local quality by assigning different kernel characteristics to different pattern types. The 1-D pattern detecting kernels are optimized for linear pattern characteristics, while the 2-D pattern detecting kernels are optimized for two-dimensional pattern characteristics. This localized optimization improves measurement precision for each pattern type without requiring a completely different model structure
Solution Approach 2:
The patent segments the empirical data into 1-D empirical data and 2-D empirical data, and correspondingly segments the modeling process into separate kernel fitting stages. This segmentation allows precise measurement of pattern-specific characteristics while maintaining a unified overall model framework, balancing measurement precision with model structure manageability
Data Source
AI summary
One embodiment provides a system for determining an improved process model that models one or more semiconductor manufacturing processes. During operation, the system can receive a first process model. Next, the system can receive a 2-D-pattern detecting kernel which can detect 2-D patterns. The system can then receive a second set of empirical data which is associated with 2-D patterns in a test layout. Next, the system can determine an improved process model using the first process model, the 2-D-pattern detecting kernel, the test layout, and the second set of empirical data.