Cross-layer common-unique analysis for semiconductor wafer defect filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor wafer inspection methods are inefficient in distinguishing defects of interest (DOI) from nuisance defects, leading to high nuisance rates and labor-intensive processes, especially as design rules shrink, making it difficult to focus process control on relevant defects.
Innovation Solution
A system and method utilizing a light source, detector, and processor to perform common-unique analysis and filter defects based on locations and attributes across multiple layers of the semiconductor wafer, employing machine learning and image processing techniques to identify and separate DOI from nuisance defects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If aggressive defect detection thresholds are used to detect more defects, then the likelihood of detecting defects of interest increases, but the nuisance rate increases substantially
Solution Approach 1:
The defect classification process is segmented into multiple independent analysis layers: first layer defect attributes are analyzed separately from second layer defect attributes, allowing independent optimization of detection thresholds for each layer while filtering common nuisance defects across layers
Solution Approach 2:
A processor acts as an intermediary between defect detection and final classification, receiving defect locations from multiple layers and applying common-unique analysis to filter nuisance defects before reporting final defect of interest locations
2Measurement precision
If manual classification of defects is performed to verify defects, then classification accuracy improves, but the process becomes tedious and time-consuming
Solution Approach 1:
The system performs self-service classification by automatically analyzing defect attributes from multiple layers and using common-unique analysis to classify defects as nuisance or defect of interest without requiring manual review, thereby maintaining accuracy while eliminating time loss
Solution Approach 2:
The manual mechanical classification process is replaced with an automated electronic system that uses processor-based analysis of defect attributes and common-unique algorithms to perform classification automatically
3Productivity
If automatic detection and classification schemes are used to save time, then processing speed increases, but the system is prone to nuisance defects and requires large computation power
Solution Approach 1:
The system adds a new dimension to automatic detection by analyzing defects across multiple layers simultaneously rather than single-layer analysis, using vertical stacking of defect attribute data to improve classification accuracy while maintaining speed
Solution Approach 2:
The system performs preliminary analysis of defect attributes from multiple layers before final classification, pre-processing the data to identify common nuisance patterns that can be filtered out before the final defect of interest determination
4Measurement precision
If inspection processes are made more sensitive to detect smaller defects, then detection capability improves, but the population of nuisance defects detected increases dramatically
Solution Approach 1:
The system extracts and separates nuisance defects from the total defect population by identifying common defects across multiple layers, removing them from the final defect of interest list while preserving genuinely relevant defects
Solution Approach 2:
The system applies different analysis approaches to different defect populations: aggressive detection thresholds are applied locally to each layer to ensure no defect of interest is missed, while global common-unique analysis across layers filters out the nuisance portion
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 reduces nuisance rates and improves sensitivity to DOI, allowing for more effective defect detection and process control by distinguishing between relevant and non-relevant defects, thereby enhancing yield and reducing the complexity of inspection processes.
Implementation Method 1
a light source that generates a beam of light; a stage configured to hold a semiconductor wafer in a path of the beam of light; a detector that receives the beam of light reflected from the semiconductor wafer
Data Source
AI summary
Common events between layers on a semiconductor wafer are filtered. Common events should contain the majority of defects of interest. Only nuisance events that are common between layers on the semiconductor wafer remain, which reduces the nuisance rate. Defects that are common across layers can be filtered based on, for example, defect coordinates, a difference image, or defect attributes.