Localized Feature Quantification in Surface Metrology Tools
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methodologies for wafer topography metrology struggle with accurately capturing and quantifying localized features, leading to measurement noise and reduced accuracy, especially at the wafer edge regions, due to signal attenuation and artifacts from traditional filtering methods.
Innovation Solution
A method involving high-order surface fitting and the use of two-dimensional Localized Feature Metrics (LFM) windows to suppress edge and corner artifacts, combined with adaptive filtering to stabilize extreme data sample measurements, improves the accuracy and robustness of localized feature quantification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional Double Gaussian (DG) filtering schemes are used for nanotopography characterization, then full wafer characterization is optimized, but localized feature signal is attenuated and measurement precision deteriorates
Solution Approach 1:
The patent applies different filtering strategies to different regions of the wafer surface. Instead of using a uniform Double Gaussian filter across the entire wafer, the invention implements localized filtering approaches that adapt to regional characteristics, particularly preserving high-frequency components in regions containing localized features while maintaining noise suppression in other areas.
Solution Approach 2:
The patent modifies the filtering parameters and approaches based on the specific characteristics of the measurement region. By changing the filtering parameters dynamically rather than using fixed DG filter settings, the system optimizes both signal preservation and noise suppression for localized feature measurement.
2Measurement precision
If high frequency features are measured in localized feature metric with extreme statistics, then localized features are detected, but measurement noise increases and repeatability deteriorates
Solution Approach 1:
The patent implements feedback mechanisms that use measurement results to adjust subsequent measurement and filtering parameters. By analyzing the characteristics of detected features and measurement quality, the system dynamically adjusts filtering and processing parameters to maintain both detection accuracy and repeatability across varying measurement conditions.
Solution Approach 2:
The patent transitions from static filtering and processing approaches to dynamic methods that adapt to the specific characteristics of each measurement. The system dynamically adjusts parameters based on real-time analysis of the data, allowing optimal performance for different feature types and measurement conditions while maintaining consistency.
3Ease of operation
If rectangular region construction is used for localized feature quantification, then feature detection is simplified, but artifact errors are introduced at region edges and corners
Solution Approach 1:
The patent addresses the symmetry issue of rectangular regions by applying asymmetric correction factors or weighting schemes to edge and corner regions. Rather than treating all regions equally, the invention introduces asymmetric processing that compensates for the inherent geometric artifacts at boundaries, improving accuracy while maintaining the simplicity of rectangular region definition.
Data Source
AI summary
A method for enabling more accurate measurements of localized features on wafers is disclosed. The method includes: a) performing high order surface fitting to more effectively remove the low frequency shape components and also to reduce possible signal attenuations commonly observed from SEMI standard high pass, such as Gaussian and Double Gaussian filtering; b) constructing and applying a proper two dimensional LFM window to the residual image from the surface fitting processing stage to effectively reduce the residual artifacts at the region boundaries; c) calculating the metrics of the region using the artifact-reduced image to obtain more accurate and reliable measurements; and d) using site-based metrics obtained from front and back surface data to quantify the features of interest. Additional steps may also include: filtering data from measurements of localized features on wafers and adjusting the filtering behavior according to the statistics of extreme data samples.