3D-NAND CDSEM Metrology for Staircase Line Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional CD SEM metrology algorithms struggle with accurately measuring 3D-NAND staircase structures due to assumptions about constant feature numbers and the presence of contacts, leading to incorrect line detection and process control issues.
Innovation Solution
The method involves creating a contacts mask, assigning random grey level values to preserve image statistics, and using a modified LEA algorithm to measure individual steps in 3D-NAND structures, treating each step as a separate feature and considering variable line numbers within each step.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional CD SEM metrology algorithms are used to measure 3D-NAND staircase structures, then the measurement process is simple, but the measurement precision deteriorates due to incorrect line detection caused by assumptions about constant feature numbers and presence of contacts
Solution Approach 1:
The patent applies segmentation by creating a contacts mask that separates contact regions from step regions in the SEM image. This allows the measurement algorithm to process different feature types independently, improving line detection accuracy for steps while accounting for the variable presence of contacts. The segmentation enables the algorithm to handle 3D-NAND staircase structures with varying numbers of steps without being confused by contact features.
Solution Approach 2:
The patent changes the parameter assumptions of the measurement algorithm. Instead of assuming a constant number of features, the algorithm now accepts variable numbers of steps and uses parameter passing to convey information about feature types. The grey level values are transformed to preserve image statistics while enabling differentiation between contacts and steps, allowing accurate measurement despite structural variations.
2Reliability
If the measurement algorithm assumes constant feature numbers and presence of contacts, then the algorithm is simpler to implement, but the reliability of measurement deteriorates for 3D-NAND structures with variable steps
Solution Approach 1:
The patent implements dynamics by making the measurement algorithm adaptive to variable numbers of steps. The algorithm dynamically adjusts to the actual number of steps present in each 3D-NAND structure rather than assuming a fixed number. This is achieved through parameter passing that conveys variable feature information and processing that handles optional contact features, ensuring reliable measurements across different device configurations.
Solution Approach 2:
The patent introduces an intermediary contacts mask that mediates between the raw SEM image and the measurement algorithm. This mask serves as an intermediate representation that separates contact and step features, allowing the algorithm to process step measurements reliably without being affected by variable contact presence. The intermediary structure enables the algorithm to handle structural variations while maintaining measurement consistency.
3Measurement precision
If contacts are included in the image processing, then the complete structure is analyzed, but the line edge detection accuracy deteriorates due to interference from contact features
Solution Approach 1:
The patent applies segmentation by creating a contacts mask that divides the image into contact regions and step regions. This segmentation allows the line edge detection algorithm to focus on step features while excluding contact features that would interfere with detection. The complete structure is still analyzed through separate processing of the contacts mask, maintaining versatility while improving detection accuracy.
Solution Approach 2:
The patent extracts contact features from the main image processing flow by creating a separate contacts mask. This extraction removes the interfering contact features from the line edge detection process while preserving them for separate analysis. The step features can then be measured with high accuracy without contact interference, while the complete structure including contacts is still fully characterized through the mask.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables precise measurement of 3D-NAND staircase dimensions, improving process control by accurately detecting line edges and distances between steps, even in the presence of contacts, thereby enhancing the reliability of semiconductor fabrication.
Implementation Method 1
The image may be generated by scanning the semiconductor structure e.g. with a charged particle beam and signals emanating from the semiconductor structure may be collected
Data Source
AI summary
A method for process control of a semiconductor structure fabricated by a series of fabrication steps, the method comprising obtaining an image of the semiconductor structure indicative of at least two individual fabrication steps; wherein the image is generated by scanning the semiconductor structure with a charged particle beam and collecting signals emanating from the semiconductor structure; and processing, by a hardware processor, the image to determining a parameter of the semiconductor structure, wherein processing includes measuring step/s from among the fabrication steps as an individual feature.