Semiconductor Metrology Library Optimization via Maximum Likelihood
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing measurement library optimization methods in semiconductor metrology are inefficient, leading to inaccuracies and significant time delays due to the lack of real fabrication process knowledge and the need for recalculating pre-computed libraries, which increases computational effort and delays production schedules.
Innovation Solution
A maximum likelihood based approach is employed to optimize library control parameters without regenerating the measurement library, by iteratively evolving parameter values to minimize differences between reference and estimated measurements, using trusted reference data from instruments like TEMs and AFMs, and applying these optimized parameters to improve measurement accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the measurement library is regenerated to improve measurement accuracy, then measurement precision is improved, but loss of time and computational effort increases significantly
Solution Approach 1:
The patent applies preliminary action by pre-computing a comprehensive measurement library that covers a wide range of parameter values and conditions in advance. This pre-computed library is then optimized using maximum likelihood estimation with reference data to adjust control parameters, avoiding the need to regenerate the entire library while improving accuracy. The preliminary computation of the full library enables rapid optimization iterations.
Solution Approach 2:
The patent employs parameter changes by using maximum likelihood estimation to optimally adjust library control parameters based on reference measurement data. Instead of regenerating the measurement library with different geometric models, the approach changes the control parameters (such as weighting factors, signal selection, or measurement conditions) to maximize the likelihood of obtaining accurate measurements, thereby improving precision without time-consuming library regeneration.
2Measurement precision
If the measurement library is regenerated to improve measurement accuracy, then measurement precision is improved, but productivity decreases due to delays in production schedules
Solution Approach 1:
The measurement library is pre-computed to include a comprehensive set of parameter variations before production begins. During production, only lightweight optimization of control parameters using reference data is performed, enabling rapid adaptation without interrupting production schedules. This preliminary preparation eliminates the need for time-consuming library regeneration during manufacturing.
Solution Approach 2:
The patent enables rapid parameter optimization using maximum likelihood estimation with reference data, allowing the measurement system to adapt to actual production conditions without regenerating the measurement library. This parameter-based optimization approach maintains high measurement accuracy while keeping production delays minimal, as parameter adjustment is computationally efficient compared to full library regeneration.
3Measurement precision
If complex optical tools and multiple measurement technologies are used to characterize nanoscale structures, then measurement capability is improved, but device complexity increases
Solution Approach 1:
The patent creates a universal measurement library that can be used across multiple measurement technologies and configurations. The pre-computed library covers various optical tools, wavelengths, and measurement conditions, allowing a single optimized library to serve multiple measurement systems. This multi-functionality reduces the need for separate libraries for each tool, simplifying the overall measurement system while maintaining comprehensive characterization capability.
Solution Approach 2:
The patent uses parameter-based optimization where control parameters (such as measurement conditions, signal weights, or analysis settings) are adjusted to optimize performance across different measurement technologies. This approach allows the same measurement library to be adapted to various optical tools and configurations through parameter changes rather than requiring separate specialized libraries for each system, thereby reducing device complexity.
4Ease of operation
If the measurement library is optimized using traditional methods without maximum likelihood estimation, then ease of operation is maintained, but measurement precision deteriorates due to modeling inaccuracies
Solution Approach 1:
The patent implements feedback by using reference measurement data to optimize library control parameters through maximum likelihood estimation. The reference data provides feedback on the accuracy of the measurement library, and this feedback is used to iteratively adjust control parameters to maximize measurement precision. This feedback mechanism maintains ease of operation while significantly improving accuracy compared to traditional optimization methods.
Solution Approach 2:
The patent applies parameter changes through maximum likelihood estimation to optimally adjust library control parameters based on reference data. This statistical approach automatically determines the optimal parameter values that maximize the likelihood of accurate measurements, improving precision without requiring complex manual optimization procedures. The method maintains ease of operation by using automated statistical optimization rather than manual trial-and-error approaches.
Data Source
AI summary
Methods and systems for optimizing a set of measurement library control parameters for a particular metrology application are presented herein. Measurement signals are collected from one or more metrology targets by a target measurement system. Values of user selected parameters of interest are resolved by fitting a pre-computed measurement library function to the measurement signals for a given set of library control parameters. Values of one or more library control parameters are optimized such that differences between the values of the parameters of interest estimated by the library based measurement and reference values associated with trusted measurements of the parameters of interest are minimized. The optimization of the library control parameter values is performed without recalculating the pre-computed measurement library. Subsequent library based measurements are performed by the target measurement system using the optimized set of measurement library control parameters with improved measurement performance.