Parameter Estimation Network for Metrology in Charged Particle Microscopy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current metrology techniques for charged particle microscopy require heavy user interaction due to size, shape, noise, and contrast variations, leading to failed or incorrect measurements, and lack robustness in automating analysis and measurement processes, especially for small structures that change in shape and consistency.
Innovation Solution
A parameter estimation network (PEN) using a recurrent convolutional neural network aligns a feature model to reference images, optimizing model parameters to provide precise metrology data, reducing the need for manual intervention and improving measurement accuracy by recursively refining model parameters based on image analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual feature identification and measurement is performed by skilled operators, then measurement accuracy can be maintained, but productivity is reduced and human error persists
Solution Approach 1:
The system enables self-service by training the neural network on manually annotated examples, allowing it to automatically identify features and perform measurements without continuous human intervention. The manual annotations serve as training data, after which the system operates autonomously.
Solution Approach 2:
The patent replaces the mechanical system of manual operator interaction with an automated neural network-based system. The neural network learns from training data and automatically performs feature identification and measurement, substituting human operators in the measurement process.
2Productivity
If automated measurement recipes are implemented, then productivity increases, but robustness decreases due to inability to adapt to changes
Solution Approach 1:
The system transitions from static measurement recipes to a dynamic neural network model that can adapt to changes in imaging conditions and manufacturing processes. The neural network learns from training data and can generalize to new conditions without requiring manual recipe re-tuning.
Solution Approach 2:
The neural network automatically adjusts its internal parameters and decision boundaries based on the input image characteristics, allowing it to adapt to variations in imaging conditions and manufacturing processes without requiring external reconfiguration.
3Measurement precision
If measurement recipes are manually re-tuned for each change, then measurement accuracy is maintained, but loss of time increases
Solution Approach 1:
The system performs preliminary action by training the neural network on comprehensive training data that covers various imaging conditions and manufacturing variations. This pre-training enables the system to handle future changes without requiring re-tuning, as the network has already learned to recognize features under diverse conditions.
4Reliability
If operators screen false positives to ensure accuracy, then measurement reliability improves, but productivity decreases
Solution Approach 1:
The neural network performs self-service by automatically identifying and measuring features without requiring operator screening. The system handles the entire measurement process autonomously, from feature detection to measurement extraction, eliminating the need for manual false positive screening.
Data Source
AI summary
Methods and apparatuses are disclosed herein for parameter estimation for metrology. An example method at least includes optimizing, using a parameter estimation network, a parameter set to fit a feature in an image based on one or more models of the feature, the parameter set defining the one or more models, and providing metrology data of the feature in the image based on the optimized parameter set.