SEM Roughness Measurement Using Inverse Linescan Noise Subtraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for measuring roughness in pattern structures, particularly in noisy SEM images, suffer from bias due to measurement noise, leading to inaccurate and unreliable results that affect the quality of manufacturing processes and devices.
Innovation Solution
A system and method that utilize an inverse linescan model to detect edges in SEM images without filtering, generating a biased power spectral density dataset, evaluating noise models, and subtracting noise to obtain unbiased roughness measures, thereby filtering out artifacts and revealing true feature roughness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If noise filtering is applied to SEM images for roughness measurement, then measurement reliability is improved, but measurement precision deteriorates due to loss of true feature information
Solution Approach 1:
The measurement process is segmented into separate stages: first acquiring the raw signal containing both feature information and noise, then separately characterizing the noise through PSD analysis, and finally subtracting the noise component mathematically rather than through signal filtering. This segmentation allows preservation of true feature information while removing noise.
Solution Approach 2:
Power spectral density (PSD) analysis serves as an intermediary tool that characterizes the noise properties without directly modifying the raw signal. The PSD provides a frequency-domain representation that enables identification and quantification of noise components, which then can be subtracted from the original measurement to obtain the true roughness.
2Ease of operation
If conventional roughness measurement methods are used in noisy images, then ease of operation is maintained, but measurement precision deteriorates due to noise bias
Solution Approach 1:
The method incorporates feedback through iterative PSD analysis and noise subtraction. The process uses the measured signal to characterize noise, then applies corrections based on that characterization, and can iterate to refine the roughness measurement. This feedback loop automatically compensates for noise without requiring manual intervention.
Solution Approach 2:
The patent replaces mechanical or manual noise filtering operations with mathematical signal processing in the frequency domain. Instead of applying physical filters that modify the signal directly, the system uses PSD-based noise modeling and subtraction, which achieves noise removal while preserving the true feature spectrum.
3Reliability
If multiple measurements are averaged to reduce noise, then measurement reliability is improved, but loss of time increases
Solution Approach 1:
The method changes the approach from temporal averaging (taking multiple measurements over time) to frequency-domain analysis (examining the spectral characteristics of a single measurement). By transforming the problem from the time domain to the frequency domain through PSD analysis, the system achieves noise reduction without requiring multiple sequential measurements.
Data Source
AI summary
In one embodiment, a method includes receiving measured linescan information describing a pattern structure of a feature, applying the received measured linescan information to an inverse linescan model that relates measured linescan information to feature geometry information, and identifying, based at least in part on the applying the received measured linescan model to the inverse linescan model, feature geometry information that describes a feature that would produce a linescan corresponding to the received measured linescan information. The method also includes determining, at least in part using the inverse linescan model, feature edge positions of the identified feature, analyzing the feature edge positions to determine errors in the manufacture of the pattern structure, and controlling a lithography tool based on the analysis of the feature edge positions.