Metrology Method Using Trained Model for Overlay Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing overlay inference methods in lithographic processes lack flexibility with respect to process variations, suffer from accuracy and robustness issues, and inefficiently use signal dimensions, leading to suboptimal overlay estimation.
Innovation Solution
A method that uses a trained model to relate asymmetry signals to the parameter of interest, incorporating proxies for nuisance components, allowing for improved overlay inference by optimizing training parameters and efficiently utilizing multiple information channels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional overlay inference methods are used, then the measurement process is simple, but the accuracy and robustness of overlay estimation deteriorates due to lack of flexibility with process variations
Solution Approach 1:
The patent transforms the overlay inference problem by changing parameters - specifically introducing proxy parameters for nuisance components (process variations) and using a trained model that relates asymmetry signals to overlay while accounting for these proxies. This allows the system to maintain simplicity in data collection while achieving high precision through sophisticated parameter transformation and model-based inference.
Solution Approach 2:
The patent introduces intermediary proxy components that mediate between the raw asymmetry signals and the overlay parameter. These proxies represent nuisance components (focus, alignment, process variations) and allow the trained model to separate their effects from the true overlay signal, thereby improving measurement precision without requiring direct measurement of all influencing factors.
2Reliability
If existing metrology methods are used, then the measurement process is straightforward, but robustness to process variations deteriorates
Solution Approach 1:
The patent implements feedback through the trained model that continuously learns from and adapts to process variations. The model uses proxy components to capture the effects of process variations and adjusts overlay estimates accordingly, providing robustness without requiring manual intervention or complex operational changes. The feedback loop is embedded in the training process and model inference.
Solution Approach 2:
The patent creates copies of the measurement process by using trained models that replicate the relationship between asymmetry signals and overlay under various process conditions. The model captures the essential relationships through training data, allowing it to predict overlay accurately across different process variations without requiring repeated manual measurements or complex operational procedures.
3Measurement precision
If signal dimensions are used inefficiently, then the measurement approach is simple, but the overlay estimation becomes suboptimal
Solution Approach 1:
The patent applies dimensionality change by introducing additional dimensions to the measurement space through proxy components that represent nuisance factors. The trained model operates in this expanded dimensionality, relating multiple asymmetry signals (including those modulated by proxies) to overlay. This allows efficient utilization of signal dimensions by capturing correlations between different signal components and process variations, thereby improving precision without information loss.
Solution Approach 2:
The patent segments the measurement problem into separate components: asymmetry signals containing overlay information, proxy components representing nuisance factors, and the trained model that integrates them. This segmentation allows each component to be processed and utilized optimally - the asymmetry signals provide overlay data, the proxies capture process variation effects, and the model synthesizes them into accurate overlay estimates, preventing information loss through comprehensive use of all signal dimensions.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The method enhances the accuracy and robustness of overlay inference by effectively separating overlay signals from nuisance components, improving flexibility with respect to process variations, and optimizing the use of signal dimensions.
Implementation Method 1
These devices direct a beam of radiation onto a target and measure one or more properties of the scattered radiation—e.g., intensity at a single angle of reflection as a function of wavelength
Implementation Method 2
Diffraction-based overlay using dark-field detection of the diffraction orders enables overlay measurements on smaller targets
Implementation Method 3
obtaining a trained model having been trained or configured to relate said at least one asymmetry signal to the parameter of interest, the trained model comprising at least one proxy for at least one nuisance component
Data Source
AI summary
Disclosed is a method for determining a parameter of interest relating to at least one target on a substrate. The method comprises obtaining metrology data comprising at least one asymmetry signal, said at least one asymmetry signal comprising a difference or imbalance in a measurement parameter from the target; obtaining a trained model having been trained or configured to relate said at least one asymmetry signal to the parameter of interest, the trained model comprising at least one proxy for at least one nuisance component of the at least one asymmetry signal; and inferring said parameter of interest for said at least one target from said at least one asymmetry signal using the trained model.