Lithography Model Bounds for Printability Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current mask printability verification in optical lithography lacks accuracy in predicting errors, especially when process parameters are outside the bounds of calibration data, leading to potential costly mistakes in semiconductor manufacturing due to reliance on unverified simulation models.
Innovation Solution
The method involves creating tight clusters of calibration data to define predictable regions, using convex hulls or ellipsoids to identify points within or outside these clusters, and reporting errors when process parameters exceed predetermined boundaries, thereby enhancing the accuracy of printability verification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If empirical data from a single set of nominal exposure conditions is used to calibrate the lithographic process model, then calibration simplicity is maintained, but prediction accuracy for varied process parameters deteriorates
Solution Approach 1:
The patent applies preliminary action by performing model calibration across multiple sets of varied exposure conditions before actual printability verification. This pre-calibration with diverse parameters ensures the model is prepared to accurately predict outcomes for a wide range of process variations, rather than calibrating simply with a single nominal condition and then encountering accuracy issues during verification.
2Measurement precision
If the lithographic process model is calibrated with varied exposure conditions, then prediction accuracy for different process parameters is improved, but calibration complexity and data requirements increase
Solution Approach 1:
The patent utilizes parameter changes by systematically varying exposure conditions (such as focus, dose, and other process parameters) during model calibration. By intentionally changing these parameters across multiple calibration sets, the model learns to accurately predict outcomes for a broad range of conditions. This approach balances the increased calibration complexity with the significant gain in prediction accuracy, as the model becomes robust to parameter variations.
3Productivity
If printability verification is performed without checking model predictability bounds, then verification speed is maintained, but error detection accuracy deteriorates
Solution Approach 1:
The patent introduces an intermediary mechanism by implementing predictability bound checks as an intermediate step during printability verification. Before declaring a verification result, the system checks whether the process parameters fall within the calibrated model's predictability bounds. This intermediary check ensures that only reliable predictions are accepted, significantly improving error detection accuracy while adding minimal overhead to the verification process.
4Adaptability or versatility
If process parameters fall outside calibration data bounds, then the model cannot reliably predict errors, but conventional verification still proceeds without warning
Solution Approach 1:
The patent implements feedback by continuously monitoring whether process parameters during verification fall within the calibrated model's predictability bounds. When parameters exceed these bounds, the system provides feedback by flagging the prediction as unreliable or erroneous. This feedback mechanism ensures that the verification process is aware of its own limitations and can appropriately handle cases where the model cannot reliably predict outcomes, thereby maintaining both adaptability and reliability.
Data Source
AI summary
Embodiments of the present invention provide a method of performing printability verification of a mask layout. The method includes creating one or more tight clusters; computing a set of process parameters associated with a point on said mask; comparing said set of process parameters to said one or more tight clusters; and reporting an error when at least one of said process parameters is away from said one or more tight clusters.