Wafer Image Segmentation for Predictive Process Non-Compliance Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing semiconductor manufacturing processes lack high-throughput and sensitive methods for detecting non-compliance based on previously unknown patterns, relying on slow and reactive techniques that fail to provide in-line feedback, leading to random defects and inefficient maintenance.
Innovation Solution
A method and device for generating a model that uses segmentation masks and metrics to analyze wafer images across multiple modalities, enhancing signal-to-noise ratio and enabling predictive process control by identifying statistically significant combinations of segment masks, metrics, and wafer images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional low wafer sampling inspection is used, then detection sensitivity is improved, but throughput is reduced
Solution Approach 1:
The inspection system segments the wafer into multiple regions of interest (ROIs) and applies different inspection strategies to each region. This allows high-throughput inspection for most areas while maintaining sensitive inspection for critical regions, resolving the contradiction between detection sensitivity and throughput.
Solution Approach 2:
The system performs partial inspection by focusing computational resources and inspection depth only on regions where defects are most likely to occur or where process non-compliance is most critical. This partial action approach maintains high throughput while achieving sufficient detection sensitivity for quality control.
2Device complexity
If process monitoring is performed only after defects are created, then device complexity is reduced, but reliability is worsened
Solution Approach 1:
The system performs preliminary analysis of wafer images to identify patterns and trends that indicate potential process non-compliance before actual defects are created. By detecting early signs of process drift through image pattern recognition, the system maintains reliability without requiring complex real-time monitoring infrastructure.
Solution Approach 2:
The system establishes feedback loops where inspection results are continuously fed back to adjust inspection parameters and alert operators to emerging process issues. This feedback mechanism enables reliable process control by detecting and responding to non-compliance conditions as they develop, rather than waiting for definitive defects.
3Ease of operation
If inspection focuses on pre-defined zones with known patterns, then ease of operation is improved, but adaptability is reduced
Solution Approach 1:
The inspection system dynamically adapts its inspection parameters, regions of interest, and analysis methods based on the specific wafer type, process step, and detected patterns. This dynamic adjustment maintains ease of operation through automated parameter selection while achieving high adaptability to different defect types and process conditions.
Solution Approach 2:
The system changes inspection parameters such as imaging conditions, analysis algorithms, and evaluation criteria based on the specific process module and wafer characteristics. This parameter adaptation enables the system to maintain simple operation through automated configuration while achieving versatility across different inspection scenarios and defect patterns.
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
Methods and devices are disclosed for automated detection of a status of wafer fabrication process based on images. The methods advantageously use segment masks to enhance the signal-to-noise ratio of the images. Metrics are then calculated for the segment mask variations in order to determine one or more combinations of segment masks and metrics that are predictive of a process non-compliance. A model can be generated as a result of the process. In another embodiment, a method uses a model to monitor a process for compliance.