Length Measurement System for Semiconductor Edge Roughness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for measuring micropattern sizes and edge roughness in semiconductor manufacturing face challenges with noise interference, leading to inaccurate measurements and potential pattern deformation, especially when high-frequency components are lost or signal profiles are distorted due to averaging or smoothing processes.
Innovation Solution
The development of five calculation methods and a system that optimizes smoothing parameters to obtain true size and roughness measurements by minimizing noise influence, allowing for precise determination of edge points and reducing measurement variations, thereby improving measurement accuracy and throughput.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If averaging or smoothing is applied to reduce noise influence, then measurement precision is improved, but high-frequency components are lost and pattern deformation occurs
Solution Approach 1:
The patent applies parameter changes by introducing a smoothing parameter Sm that can be adjusted to control the degree of smoothing. By varying Sm, the system optimizes the balance between noise reduction and preservation of high-frequency components, allowing accurate measurement while minimizing information loss through controlled parameter adjustment rather than fixed averaging
Solution Approach 2:
The patent implements feedback by calculating the dependency of measured values on the smoothing parameter Sm and using this information to determine the optimal Sm value. The system evaluates how measurement results change with different Sm values and selects the parameter that provides the best measurement accuracy while preserving essential pattern information, creating a closed-loop optimization process
2Measurement precision
If integration time is increased to improve signal S/N ratio, then measurement precision is improved, but inspection time increases and pattern deformation occurs
Solution Approach 1:
The patent replaces the mechanical approach of increasing integration time with a computational approach using image processing algorithms. Instead of physically integrating signals for longer durations, the system uses mathematical operations (smoothing with variable Sm) to enhance signal quality, thereby achieving improved measurement precision without the time penalty and pattern deformation associated with extended physical integration
3Measurement precision
If averaging is applied in direction parallel to pattern edges, then noise is reduced, but short-period roughness components are lost
Solution Approach 1:
The patent applies local quality by making the smoothing treatment location-dependent through the use of variable smoothing parameter Sm. Different regions of the pattern can have different smoothing intensities based on their specific requirements, allowing the system to reduce noise in areas where it is most problematic while preserving short-period roughness components in areas where they are critical for measurement, thus achieving localized optimization rather than uniform processing
Data Source
AI summary
Disclosed herewith is a length measurement system, which obtains a value closer to its true one when figuring out the size and edge roughness of a pattern from a noise-included pattern image. Among plural band-like regions representing a portion around an edge in an image respectively, the system calculates the dependency of the edge point position on the image processing parameter at each of a narrow width band-like portion and a wide width band-like portion to calculate an image processing condition that calculates each measured value closer to its true value or estimates the true value itself.