Semiconductor Defect Analysis via Frequency Domain Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing defect distribution analysis methods struggle to identify defect causes in semiconductor wafer manufacturing when defect patterns are weak or have low density, as they often classify such cases as global and require user-defined events, failing to provide efficient identification of failure causes.
Innovation Solution
A method that classifies defect distributions into specific categories such as repeated, clustered, arc-shaped, radial, line-type, ring, and blob-type defects using techniques like Voronoi diagrams, Hough transforms, and geometric template patterns, enabling automatic identification of defect causes without requiring user-defined events.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional defect distribution analysis methods are used, then defect causes can be identified when remarkable defect distribution patterns appear, but weak defect distribution patterns (low defect density or small difference between inside and outside of pattern) cannot be properly classified and analyzed
Solution Approach 1:
The patent transforms the defect distribution analysis from direct pattern recognition to frequency domain analysis by applying Fast Fourier Transform (FFT). This parameter transformation enables the detection of weak periodic defect patterns that are not visible in the spatial domain, thereby improving both measurement precision and adaptability to weak defect patterns.
Solution Approach 2:
The patent introduces an intermediary processing step (FFT transformation) between defect data acquisition and pattern recognition. This intermediary converts spatial defect distribution into frequency domain representation, allowing weak periodic patterns to be extracted and analyzed effectively, thus resolving the contradiction between detection accuracy and handling weak patterns.
2Adaptability or versatility
If user-defined events are required for defect classification, then flexibility in categorization is achieved, but analysis speed decreases and user intervention is needed
Solution Approach 1:
The patent implements self-service by enabling the system to automatically classify defects into standardized categories (repeated, clustered, arc-shaped, radial, line-type, ring, blob-type, random) based on FFT analysis results without requiring user-defined events. This automation maintains classification flexibility through predefined categories while eliminating user intervention, thereby improving analysis speed and productivity.
Solution Approach 2:
The patent segments the defect analysis process into distinct automated classification categories based on frequency domain characteristics. By dividing defect patterns into specific types (repeated, clustered, arc-shaped, etc.), the system achieves both flexibility in categorization and high-speed automated processing without user intervention.
3Productivity
If automated defect classification is implemented, then analysis speed increases, but the ability to handle various weak defect patterns may be limited
Solution Approach 1:
The patent introduces FFT transformation as an intermediary that enables automated classification to effectively handle weak periodic patterns. This frequency domain transformation allows the automated system to detect periodicities and weak patterns that would be invisible in spatial domain analysis, thus maintaining both high speed and high precision.
Solution Approach 2:
By transforming from spatial domain to frequency domain parameters, the patent enables automated classification to detect weak periodic defect patterns. This parameter change allows the system to maintain high analysis speed while improving detection capability for weak patterns through frequency spectrum analysis.
Data Source
AI summary
In a process for manufacturing a semiconductor wafer, defect distribution state analysis is performed so as to facilitate identification of the defect cause including a device cause and a process cause by classifying the defect distribution state according to the defect position coordinates detected by the inspection device, into one of the distribution characteristic categories: repeated defects, clustered defects, arc-shaped regional defects, radial regional defects, line type regional defects, ring and blob type regional defects, and random defects.