Uncertainty-Guided Data Selection for Patterning Model Calibration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing prediction models in semiconductor manufacturing lack certainty and consistency in their predictions, leading to uncertainties in mask layout generation and potential functional issues in integrated circuits due to repeated model usage.
Innovation Solution
A method to reduce prediction uncertainty by determining a prediction uncertainty parameter based on prediction data, selecting a subset of patterning process data, and recalibrating the prediction model using calibration data and the selected subset, iteratively until convergence is achieved.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If prediction models are used repeatedly in semiconductor manufacturing, then productivity is improved, but prediction uncertainty and inconsistency increase
Solution Approach 1:
The patent implements a feedback mechanism where prediction uncertainty parameters are calculated from prediction data, and this information is fed back to select subsets of patterning process data for recalibrating the prediction model. This iterative feedback loop continuously reduces prediction uncertainty while maintaining high productivity through automated model recalibration.
Solution Approach 2:
The patent changes the parameters of the prediction model by recalibrating it with selected subsets of patterning process data. The model parameters are updated based on prediction uncertainty parameters, allowing the model to adapt and reduce uncertainty in its predictions while maintaining efficient operation.
2Measurement precision
If prediction models are recalibrated frequently to reduce uncertainty, then prediction accuracy is improved, but computational time and process complexity increase
Solution Approach 1:
The patent applies partial action by selecting only a subset of patterning process data for recalibration rather than using all available data. This is determined by the prediction uncertainty parameter, which identifies the most critical data points needed for accurate prediction, thereby reducing recalibration time while maintaining prediction accuracy.
Solution Approach 2:
The patent performs preliminary action by calculating prediction uncertainty parameters before selecting data for recalibration. This preliminary analysis identifies which data points are most important for reducing uncertainty, allowing the recalibration process to be more efficient and targeted, reducing overall computational time.
3Measurement precision
If all patterning process data is used for model calibration, then measurement precision is improved, but device complexity and data processing requirements increase
Solution Approach 1:
The patent extracts only the essential data needed for calibration by using prediction uncertainty parameters to select subsets of patterning process data. This extraction approach removes unnecessary data processing complexity while maintaining the precision needed for accurate model calibration, focusing computational resources only on critical data points.
Data Source
AI summary
Systems and methods for reducing prediction uncertainty in a prediction model associated with a patterning process are described. These may be used in calibrating a process model associated with the patterning process, for example. Reducing the uncertainty in the prediction model may include determining a prediction uncertainty parameter based on prediction data. The prediction data may be determined using the prediction model. The prediction model may have been calibrated with calibration data. The prediction uncertainty parameter may be associated with variation in the prediction data. Reducing the uncertainty in the prediction model may include selecting a subset of process data based on the prediction uncertainty parameter; and recalibrating the prediction model using the calibration data and the selected subset of the process data.