Substrate Process Modeling for Fast Condition Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies for optimizing substrate processing conditions rely heavily on expert knowledge or machine learning, which can be time-consuming, require a large number of experiments, and lack robustness, making it difficult to optimize process conditions efficiently and accurately across different environments and devices.
Innovation Solution
An information processing apparatus that acquires a mathematical model corresponding to a process result from a substrate processing apparatus, fits parameters of the model based on measurement data, and predicts process results, allowing for efficient optimization of process conditions with high accuracy, even without expert intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If expert knowledge or machine learning is used to optimize substrate processing conditions, then optimization accuracy can be improved, but the time required and number of experiments increase significantly
Solution Approach 1:
The patent changes the fundamental parameter of the optimization approach by using a physics-based mathematical model instead of data-driven machine learning. The model uses physical equations to directly calculate process conditions, eliminating the need for extensive experimental data collection and training, thus achieving fast and accurate optimization simultaneously
Solution Approach 2:
The patent replaces the mechanical trial-and-error experimentation process with a computational mathematical model. Instead of physically conducting numerous experiments to optimize conditions, the system uses mathematical calculations based on physical principles to determine optimal parameters directly
2Reliability
If machine learning models are trained to predict process results, then prediction capability is enhanced, but the number of experiments required increases
Solution Approach 1:
The patent changes the approach from statistical learning to physics-based calculation. The mathematical model uses fundamental physical equations to predict process results directly, requiring minimal experimental data for calibration while maintaining high prediction reliability across different operating conditions
Solution Approach 2:
The mathematical model is designed to be universal and applicable across different substrate processing conditions and equipment configurations. Once calibrated with minimal data, the same model can predict results for various process parameters without requiring separate training datasets for each condition
3Ease of operation
If conventional optimization methods are used, then process conditions can be optimized, but adaptability to different environments and devices is poor
Solution Approach 1:
The patent uses a mathematical model with adjustable parameters that can be calibrated for different environments and devices. The model structure remains the same, but parameters can be fitted to match specific equipment characteristics, providing both ease of operation and adaptability
Solution Approach 2:
The patent performs preliminary calibration of the mathematical model using a small set of measurement data from the specific equipment. This preliminary action of fitting model parameters to actual device characteristics enables the system to adapt to different environments while maintaining efficient operation
Data Source
AI summary
An information processing apparatus includes a model acquisition unit configured to acquire a mathematical model corresponding to a process result obtained by a substrate processing apparatus; a data acquisition unit configured to acquire measurement data indicating the process result measured using the substrate processing apparatus; a fitting unit configured to fit a parameter of the mathematical model based on the measurement data; and a prediction unit configured to predict the process result based on the mathematical model with the parameter fitted by the fitting unit.