Virtual Substrate Condition Sensing in Manufacturing Chambers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for characterizing and predicting the effects of processing conditions in manufacturing chambers, especially near substrates, are impractical due to the inability to perform direct measurements at the substrate location, leading to inefficiencies and inconsistencies in substrate production.
Innovation Solution
The use of a physics-informed compressed sensing machine learning model to virtually measure properties proximate to a substrate in a manufacturing chamber, by training the model with measurements from sensors placed elsewhere in the chamber and using a reduced order model to infer property values at the substrate location.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensors are placed at the substrate location for direct measurement, then measurement precision is improved, but device complexity and ease of operation deteriorate due to the inability to access substrate location
Solution Approach 1:
The patent uses an intermediary machine learning model that acts as a mediator between the accessible sensor measurements and the inaccessible substrate conditions. The model learns the mapping relationship from measurements at accessible locations to properties at the substrate location, enabling indirect measurement without physical access to the substrate.
Solution Approach 2:
The patent creates a virtual copy of the substrate conditions through machine learning modeling. Instead of directly measuring the substrate, the system learns to predict substrate properties by creating a computational replica that mimics the substrate's response to chamber conditions based on measurements from other locations.
2Ease of operation
If sensors are placed elsewhere in the chamber for easier measurement, then ease of operation is improved, but measurement precision deteriorates due to distance from substrate
Solution Approach 1:
The system implements feedback by continuously learning the relationship between measurements at accessible locations and substrate conditions. The machine learning model is trained on historical data and continuously refined, using the feedback from known substrate properties to improve its predictions of substrate conditions based on remote sensor measurements.
Solution Approach 2:
The patent performs preliminary action by training the machine learning model in advance using historical measurement data and known substrate conditions. This pre-training phase establishes the mapping relationships before actual substrate processing begins, enabling the system to make accurate predictions without real-time access to substrate location.
3Reliability
If direct measurement at substrate location is performed, then reliability is improved, but device complexity increases due to access requirements
Solution Approach 1:
The machine learning model serves multiple functions: it predicts substrate temperature, analyzes chamber conditions, detects anomalies, and provides process optimization guidance. This multi-functional approach replaces what would otherwise require multiple specialized measurement systems, reducing overall device complexity while maintaining reliability.
Solution Approach 2:
The patent replaces the mechanical approach of physically placing sensors at the substrate location with a computational system. Instead of mechanical sensor placement and wiring to inaccessible locations, the system uses algorithms and data processing to achieve the same measurement objectives, significantly reducing device complexity.
Data Source
AI summary
A method includes measuring a subset of property values within a manufacturing chamber during a process performed on a substrate within the manufacturing chamber. The method further includes determining property values in the manufacturing chamber at locations removed from the locations the measurements are taken. The method further includes performing a corrective action based on the determined properties.