Substrate Processing Controller Machine Learning Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Substrate processing systems face challenges in maintaining uniformity of processing results across different substrates due to deviations in measurement values and cumulative operating time issues, leading to inconsistent film thickness and processing quality.
Innovation Solution
A substrate processing system equipped with a controller and devices that perform machine learning on measurement data to generate update parameters for adjusting processing settings, ensuring consistent film formation across substrates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If a processing program is adjusted based on a computed relationship between film thickness and processing program, then processing quality uniformity is improved, but measurement precision deteriorates due to deviations in measurement values
Solution Approach 1:
The patent implements a feedback mechanism where measurement values from substrates are continuously fed back to the learning part in the controller. The learning part updates the relationship between processing programs and film thickness based on this feedback, enabling dynamic adjustment of processing parameters to compensate for measurement deviations and maintain uniform processing quality.
Solution Approach 2:
The system dynamically changes processing parameters by generating update parameters that adjust the processing program based on learned relationships. The learning part computes updated relationships between processing programs and film thickness, then applies these changes to optimize processing conditions and compensate for measurement inaccuracies.
2Device complexity
If processing parameters are fixed based on initial computations, then device complexity is reduced, but processing uniformity deteriorates due to cumulative operating time issues
Solution Approach 1:
The patent transforms the static processing parameters into dynamic ones by introducing a learning part that continuously updates the relationship between processing programs and film thickness. This dynamic adjustment mechanism allows the system to adapt to cumulative operating time effects and maintain processing uniformity without requiring complex manual recalibration.
Solution Approach 2:
The learning part enables the system to self-adjust and self-optimize by automatically learning from measurement data and generating update parameters. This self-service capability allows the processing apparatus to maintain high uniformity over time without external intervention, balancing simplicity with adaptive performance.
Data Source
AI summary
There is provided a technique of manufacturing a semiconductor device, including: by a processing performing part, processing a substrate based on setting parameter corresponding to process recipe stored in a controller; by a first transceiver, transmitting measurement value of the processing performing part to the controller; by the controller, causing a learning part to perform machine learning process on the measurement value received from the first transceiver as learning data; by the controller, after the act of causing the learning part to perform the machine learning process, generating update data for updating the setting parameter; by the controller, causing an arithmetic part to generate update parameter for updating the setting parameter based on the update data; by the controller, causing a second transceiver to transmit the update parameter to the first transceiver; and by the updating part, updating the setting parameter based on the update parameter received from the controller.