Specimen Image Anomaly Detection via Statistical Moment Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor manufacturing processes face challenges in detecting subtle defects and anomalies in images of specimens due to limited data utilization and reliance on traditional defect thresholds, which can lead to undetected critical defects, especially in advanced designs with tighter geometry and smaller process windows.
Innovation Solution
A system and method that uses an imaging subsystem to generate images of specimens and determine characteristics without applying a defect detection algorithm, allowing computer subsystems to identify anomalies based on these characteristics, enabling more sensitive detection of defects and process variations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional defect detection algorithms with fixed thresholds are used, then the inspection process is simple and fast, but subtle defects and anomalies cannot be detected
Solution Approach 1:
The patent transforms the inspection approach by changing from fixed threshold parameters to dynamic statistical parameters. Instead of using predetermined defect thresholds, the system calculates statistical parameters (mean, standard deviation, skewness, kurtosis) from the image data itself, allowing the detection criteria to adapt to the actual process variation and detect subtle anomalies that deviate from the statistical norm
Solution Approach 2:
The patent adds a new dimension to defect detection by introducing statistical moment analysis (first moment for mean, second moment for standard deviation, third moment for skewness, fourth moment for kurtosis) beyond traditional binary defect/non-defect classification. This multi-dimensional statistical approach enables detection of subtle process variations and anomalies that single-threshold methods miss
2Reliability
If defect thresholds are set based on inspection operating parameters, then the inspection process is straightforward, but critical defects with subtle manifestations are hidden
Solution Approach 1:
The system replaces fixed inspection thresholds with dynamic statistical parameters calculated from the actual image data. By computing mean, standard deviation, skewness, and kurtosis for each image or region, the system adapts to process variations and identifies anomalies based on statistical deviation rather than arbitrary thresholds, preventing loss of subtle defect information
Solution Approach 2:
The system establishes a feedback mechanism where statistical parameters are continuously calculated from inspected images and used to identify anomalies. The anomaly identification process feeds back into process control, allowing the system to learn from actual process variations and improve detection reliability over time by adjusting to the specific characteristics of each process run
3Loss of information
If image data is processed using conventional defect inspection methods, then processing is fast, but the full information content of the images is not utilized
Solution Approach 1:
The system extracts meaningful information from images by calculating statistical parameters (mean, standard deviation, skewness, kurtosis) that capture the essential characteristics of the image data. This extraction approach distills the full image information into compact statistical representations that retain the ability to detect anomalies while reducing processing complexity compared to analyzing every pixel
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 enhances the detection of subtle defects and anomalies, improving yield and device characterization by fully utilizing image data and identifying issues early in the semiconductor manufacturing process.
Implementation Method 1
an imaging subsystem configured for generating images of a specimen by directing energy to and detecting energy from the specimen
Data Source
AI summary
Methods and systems for detecting anomalies in images of a specimen are provided. One system includes one or more computer subsystems configured for acquiring images generated of a specimen by an imaging subsystem. The computer subsystem(s) are also configured for determining one or more characteristics of the acquired images. In addition, the computer subsystem(s) are configured for identifying anomalies in the images based on the one or more determined characteristics without applying a defect detection algorithm to the images or the one or more characteristics of the images.