Neural Network Overlay Error Correction for Asymmetric Targets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing overlay metrology in semiconductor manufacturing struggles to accurately account for target structure asymmetry, leading to reduced measurement accuracy and increased uncertainty.
Innovation Solution
A neural network is trained to generate correction values for overlay measurements by inputting distance-to-origin measurements for asymmetry at multiple wavelengths, allowing for improved accuracy in overlay measurement by accounting for asymmetric perturbations in the target structure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional overlay measurement methods are used, then the measurement process is simple, but the measurement accuracy is reduced due to unaccounted target structure asymmetry
Solution Approach 1:
A machine learning model serves as an intermediary component that processes measured parameters and generates correction values. The model takes as input parameters such as distance-to-origin measurements and target structure characteristics, processes them through trained neural networks, and outputs corrected overlay measurements that account for asymmetry effects, thereby improving accuracy without requiring complex hardware modifications
Solution Approach 2:
The system changes the parameters used in overlay measurement by introducing correction values derived from machine learning models. Instead of using raw measured values directly, the system adjusts the measurement parameters by adding correction terms that compensate for target structure asymmetry, effectively transforming the measurement approach to achieve higher precision
2Reliability
If target structure asymmetry is not accounted for, then the measurement process remains simple, but measurement uncertainty increases
Solution Approach 1:
The machine learning model incorporates feedback mechanisms by continuously learning from measurement data and refining its correction predictions. The model processes input parameters, generates correction values based on trained patterns, and outputs improved measurements, creating a feedback loop that reduces uncertainty by accounting for asymmetry effects through data-driven insights
3Manufacturing precision
If simple overlay measurement is used, then the process is efficient, but alignment precision is reduced due to asymmetric perturbations
Solution Approach 1:
The machine learning model performs preliminary action by pre-calculating and providing correction values before the final measurement is completed. The model processes input parameters and generates correction terms in advance, allowing the measurement system to efficiently apply corrections without significantly delaying the overall measurement process, thus maintaining productivity while improving alignment precision
Data Source
AI summary
A correction to an error of overlay measurement which accounts for target structure asymmetry using a neural network is described. According to embodiments, an overlay measurement accuracy can be improved by accounting for multiple and/or asymmetric perturbations in the target structure. A trained neural network is described which generates a correction value for overlay measurement based on a measure of asymmetry. Based on an as-measured overlay measurement, which may not account for target structure asymmetry, and the correction value, a true overlay measurement is determined-which can exhibit improved accuracy and reduced uncertainty versus uncorrected values.