Process Condition Prediction for Target Workpiece End States
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional semiconductor manufacturing processes are hindered by the need for indirect control setting values, which slow down device development and are not directly targeted to achieve desired shape processing and surface treatment.
Innovation Solution
An information processing device that predicts the end state of a workpiece by using model data elements to determine the optimal process conditions based on initial state data and target end state data, allowing for direct determination of control setting values for semiconductor manufacturing processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional control setting values are used to control semiconductor manufacturing processes, then the process can be executed with standard control components, but the development speed is slowed down and the control is not directly targeted to achieve desired shape processing and surface treatment
Solution Approach 1:
The patent changes the control parameters from indirect control setting values to direct control parameters that directly influence the workpiece shape and surface treatment. By using machine learning models to predict optimal control parameter settings based on desired outcomes, the system directly targets shape processing and surface treatment goals, thereby accelerating device development while maintaining manageable system complexity through automated parameter optimization.
2Manufacturing precision
If indirect control setting values are used for process control, then the control system is simpler to implement, but the accuracy of achieving desired shape processing and surface treatment is reduced
Solution Approach 1:
The patent introduces machine learning models as intermediary components that translate desired shape processing and surface treatment goals into optimal control parameter settings. These models act as mediators between the control system and the manufacturing process, enabling high precision by directly predicting the relationship between control parameters and outcomes, while the complexity is managed through the use of trained predictive models rather than complex real-time control algorithms.
3Loss of time
If conventional model-based optimization is used with multiple model parameters, then the association between etched feature profile and input parameters can be established, but the process is still indirect and slows down development
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models with extensive process data to establish the relationships between control parameters and workpiece outcomes. This preliminary training phase captures the complex associations between parameters and results, allowing the system to quickly predict optimal settings for new scenarios without time-consuming iterative optimization, thereby reducing development time while maintaining ease of operation through simple query-based parameter retrieval.
Data Source
AI summary
An information processing device includes: a recording means storing model data that reproduces a change of a workpiece, the model data being constructed, as an effect of process treatment, from differential data between initial state data and end state data of the workpiece, the end state data representing state of the workpiece to which the process treatment is applied under a predetermined process condition; an input receiving means for receiving an input of the initial state data and target end state data of the workpiece; a predicting means for predicting the end state data from the received initial state data, by using the model data and a combination of multiple model data in the recording means; and a determining means for determining a process condition of process treatment to be applied to the workpiece, based on a proximity between the predicted end state data and the target end state data.