Autoencoder Abnormality Detection for Optical Film Thickness Measurement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional optical film-thickness measurement systems in semiconductor manufacturing face challenges in accurately detecting abnormalities in reference intensity data, leading to potential large errors in film thickness measurements due to unpredictable defects or setup errors, requiring labor-intensive verification processes.
Innovation Solution
An abnormality detection method using a trained autoencoder model based on machine learning to analyze preset spectrum data, calculating differences between input and output data to determine if the data contains abnormalities, thereby automatically identifying issues before polishing processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional methods are used to detect abnormalities in reference intensity data by checking actual film-thickness measurement results during device startup or consumable replacement, then measurement accuracy can be maintained, but the process requires long time and labor
Solution Approach 1:
The patent applies preliminary action by detecting abnormalities in reference intensity data before they affect film thickness measurements. The system automatically checks for abnormalities in preset spectrum data (including reference intensity data) prior to actual measurement operations, preventing potential measurement errors from occurring in the first place. This is achieved through automated comparison of current reference data against stored normal data patterns.
Solution Approach 2:
The system performs self-service by automatically detecting and identifying abnormalities in its own reference intensity data without requiring external intervention. The automated detection system compares current reference data against historical normal data patterns and independently determines whether abnormalities exist, eliminating the need for manual verification by operators.
2Measurement precision
If reference intensity data is measured in advance to remove optical system noise, then measurement accuracy is improved, but unpredictable defects or setup errors can cause abnormal values in the reference data
Solution Approach 1:
The patent implements feedback by continuously monitoring and comparing current reference intensity data against stored normal data patterns. The system provides feedback when abnormalities are detected in the reference data, allowing operators to be alerted to potential issues before they affect measurements. This feedback mechanism enables the system to self-diagnose problems with reference data quality.
Solution Approach 2:
The system performs preliminary detection of abnormalities in reference intensity data before actual film thickness measurements are performed. By checking for abnormal values in advance and alerting operators, the system prevents potentially erroneous measurements from being produced, thereby maintaining both accuracy and reliability.
3Productivity
If automated abnormality detection using machine learning is implemented, then detection speed and accuracy are improved, but device complexity increases
Solution Approach 1:
The patent replaces manual mechanical verification processes with automated computer-based detection using machine learning algorithms. Instead of requiring operators to manually analyze measurement results for abnormalities, the system uses computational models to automatically identify patterns and detect anomalies in reference intensity data, significantly improving detection speed while reducing human labor.
Solution Approach 2:
The system creates a digital copy or replica of normal reference data patterns through machine learning training. The trained model stores characteristic patterns of normal reference intensity data and can automatically compare current data against these learned patterns to detect abnormalities. This copying approach enables rapid, accurate detection without requiring complex real-time analysis hardware.
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 enables rapid and accurate detection of abnormality in reference intensity data, reducing measurement errors and streamlining the verification process, allowing for more efficient and reliable film thickness measurements in semiconductor manufacturing.
Implementation Method 1
an optical film-thickness measuring device configured to direct light, emitted by a light source, to a surface of a wafer and analyze a spectrum of reflected light from the wafer
Implementation Method 2
analyze a spectrum of reflected light from the wafer to detect a film thickness
Implementation Method 3
inputting the preset spectrum data to an autoencoder which is a trained model constructed by machine learning using training data including a plurality of normal preset spectra data; calculating a difference between output data output from the autoencoder and the preset spectrum data
Data Source
AI summary
An abnormality detection method that can automatically detect abnormality in preset spectrum data, such as reference intensity data (base intensity data) used for optical measurement of a film thickness is disclosed. The abnormality detection method includes: creating the preset spectrum data before polishing of the workpiece; inputting the preset spectrum data to an autoencoder which is a trained model constructed by machine learning using training data including a plurality of normal preset spectra data; calculating a difference between output data output from the autoencoder and the preset spectrum data; and determining that there is an abnormality in the preset spectrum data when the difference is larger than a threshold value.