Neural Network Chamber Matching for Semiconductor Tools
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current chamber matching techniques in semiconductor manufacturing are recipe-dependent, slow to scale, expensive to develop and maintain, and unable to handle multi-chamber matching effectively, leading to inefficiencies and potential device damage due to inaccurate monitoring and fault detection.
Innovation Solution
A method utilizing a neural network trained on time-series data from reference and test processing chambers to calculate errors and declare mismatches based on a threshold, enabling more accurate and efficient chamber matching without the need for destructive metrology or extensive substrate processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional chamber matching techniques are used, then chamber matching can be performed, but the process is slow to scale and expensive to develop and maintain
Solution Approach 1:
The patent replaces traditional mechanical and manual chamber matching methods with an optical-based sensor system and automated image processing algorithm. The system captures chamber images using sensors, processes them through algorithms to identify chamber features and characteristics, and automatically generates matching results, eliminating the need for manual measurement and comparison methods.
Solution Approach 2:
The patent transforms chamber matching from a manual measurement process to an automated image analysis process by changing the measurement parameters from physical dimensions to visual features captured by sensors. The system analyzes image parameters such as chamber geometry, component positions, and visual characteristics to determine chamber matching, enabling faster processing and scaling.
2Reliability
If comprehensive plasma monitoring is implemented, then process control is improved, but system complexity increases
Solution Approach 1:
The patent implements a multi-functional monitoring system that uses a single sensor platform to capture multiple types of chamber information simultaneously. The image sensors can detect chamber geometry, component positions, plasma characteristics, and other visual features, eliminating the need for multiple specialized sensors and reducing overall system complexity while improving reliability.
Solution Approach 2:
The patent introduces an intermediate image processing layer that translates complex plasma and chamber data into simplified visual representations. The processing algorithms act as intermediaries between the raw sensor data and the control system, making the data more manageable and reducing the complexity of the overall monitoring system while maintaining high reliability.
3Measurement precision
If manual chamber matching procedures are used, then detailed analysis is possible, but the process is too slow for production environments
Solution Approach 1:
The patent replaces manual measurement and analysis methods with automated image capture and processing systems. The sensors rapidly capture chamber images, and algorithms automatically analyze the images to extract precise chamber characteristics, maintaining measurement precision while increasing throughput by eliminating manual operations.
Solution Approach 2:
The patent enables continuous chamber monitoring and matching by implementing an automated system that can process multiple chambers in sequence without interruption. The image-based system allows for rapid successive measurements and comparisons, maintaining continuous productive action rather than intermittent manual analysis, thereby increasing throughput while preserving accuracy.
Data Source
AI summary
A server trains a neural network by feeding a first set of input time-series data of one or more sensors of a first processing chamber that is within specification to the neural network to produce a corresponding first set of output time-series data. The server calculates a first error. The server feeds a second set of input time-series data from corresponding one or more sensors associated with a second processing chamber under test to the trained neural network to produce a corresponding second set of output time-series data. The server calculates a second error. Responsive to the difference between a second error between the second set of input time-series data and the corresponding second set of output time-series data and a first error between the first set of input time-series data and the corresponding first set of output time-series data being equal to or exceeding a threshold amount, the server declares that the second processing chamber under test mismatches the first processing chamber that is within specifications.