HAR Cross-Section Segmentation for Accurate 3D Pillar Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 2D imaging techniques fail to accurately capture the true path and shape properties of high aspect ratio (HAR) structures in integrated semiconductor devices, such as pillars or channels, due to their unpredictable shapes and the potential for imprecise measurements and curtaining effects, which can lead to errors and defects that affect device performance.
Innovation Solution
A two-step machine learning-based approach is employed to segment and analyze cross-section images of HAR structures, utilizing a first machine learning logic for binary segmentation and a second logic for multi-level annotation, followed by parameter determination and statistical analysis to reconstruct the 3D shape and trajectory of these structures with high precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If high aspect ratio structures are formed in semiconductor devices, then device functionality and performance are improved, but the structures become difficult to manufacture with conventional lithography techniques
Solution Approach 1:
The patent divides the formation of high aspect ratio structures into multiple sequential lithography steps, where each step creates a portion of the final structure. This segmentation allows conventional lithography tools to manufacture structures that would otherwise require advanced lithography techniques, thereby improving ease of manufacture while maintaining device functionality.
Solution Approach 2:
The patent performs preliminary patterning actions to create mandrels or support structures before forming the final high aspect ratio features. These preliminary structures serve as templates or guides that enable subsequent lithography steps to achieve the desired high aspect ratio geometries using conventional equipment.
2Ease of manufacture
If conventional lithography techniques are used, then manufacturing simplicity is maintained, but the ability to form high aspect ratio structures is insufficient
Solution Approach 1:
The patent employs dynamic process control where lithography parameters such as exposure dose, focus, and resist development conditions are adjusted between steps based on the evolving structure geometry. This dynamic adaptation enables conventional lithography tools to achieve high aspect ratio formation capability while maintaining manufacturing simplicity.
Solution Approach 2:
The patent changes critical lithography parameters including numerical aperture, wavelength, and process conditions across multiple steps to optimize structure formation at each stage. These parameter changes enable the system to achieve high manufacturing precision for high aspect ratio structures while using conventional lithography equipment.
Data Source
Figure 1a
Figure 1b
Figure 2
AI summary
The present invention relates to identifying ring structures in pillars of high aspect ratio (HAR) structures. For segmentation of rings, a machine learning-logic is used. A two-step training method for the machine learning logic is described.