EUV Lithography Defect Calibration Using Pattern Shape Repair
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting and addressing stochastic defects in semiconductor manufacturing are impractical and costly due to the need for extensive experimental data and geometric grouping, which are not scalable for complex design layouts, leading to inefficiencies and potential under/over prediction of defect-free process windows.
Innovation Solution
A method and system that utilize a stochastic defect prediction model to identify high-probability defect locations, group them by pattern shapes, determine expected defect counts, and selectively repair these shapes, reducing the need for extensive experimental data and enabling efficient calibration and repair.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If extensive experimental data and geometric grouping are used to calibrate defect prediction models, then prediction accuracy is improved, but data requirements and computational complexity increase significantly
Solution Approach 1:
The patent transforms the calibration approach by changing from using extensive experimental data to using a small set of manufactured samples with known defect states. The model is calibrated using synthetic data generated from these manufactured samples, fundamentally altering the data input requirements from physical measurements to computational representations.
Solution Approach 2:
The patent creates synthetic copies of defect patterns through computational modeling rather than relying on physical experimental data. By generating virtual representations of defect states from manufactured samples, the system replicates the need for extensive data without actually collecting it through experimentation.
2Device complexity
If geometric grouping of defect locations is performed to categorize pattern shapes, then defect analysis is simplified, but scalability to complex design layouts is reduced
Solution Approach 1:
The patent segments the defect analysis by creating separate, modular components: a defect detection module, a pattern shape identification module, and a calibration module. This segmentation allows each component to handle specific aspects of the problem independently, making the overall system more scalable to complex layouts while maintaining manageable complexity.
Solution Approach 2:
The patent develops a universal calibration approach that can be applied across different design layouts and defect types. The synthetic data generation and model calibration processes are designed to be layout-independent, allowing the same calibration methodology to serve multiple functions across various semiconductor device designs.
3Reliability
If traditional calibration methods are used for stochastic defects, then prediction reliability is improved, but time to resolve defects increases
Solution Approach 1:
The patent performs preliminary calibration using manufactured samples with known defect states before actual defect resolution is needed. By pre-calibrating the model with synthetic data from controlled manufacturing, the system prepares reliable prediction capabilities in advance, eliminating the need for time-consuming calibration during defect resolution processes.
4Quantity of substance
If stochastic defect models are calibrated with limited data, then data efficiency is improved, but prediction accuracy may be compromised
Solution Approach 1:
The patent replaces the mechanical process of collecting and processing extensive experimental data with a computational approach using synthetic data generation. This substitution allows the system to achieve accurate predictions without the physical data collection process, maintaining prediction accuracy while dramatically reducing data requirements.
Data Source
AI summary
Based on an initial probability of occurrence of a stochastic defect over a layout of a workpiece, a subset of locations on the workpiece are selected where the initial probability is above a threshold. The subset of locations are grouped by pattern shapes. An expected defect count is determined for each of the pattern shapes. A subset of the pattern shapes is then selected for repair.