Semiconductor Run-to-Run Recipe Control Using Metrology Forecasts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing semiconductor manufacturing processes face challenges in achieving precise control over process outputs due to the complexity of sensor data and the inability of current machine learning techniques to reliably invert models for real-time run-to-run adjustments, leading to increased scrap and rework rates.
Innovation Solution
A combination of a process model, metrology forecast model, and actor model is employed to leverage complex sensor data for real-time recipe optimization, using a metrology forecast model to predict outcomes and an actor model to adjust inputs reliably, while incorporating cost functions and process models for invertible predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional machine learning models are used for process control, then model simplicity is maintained, but the ability to reliably invert models for real-time adjustments is lost
Solution Approach 1:
The control system is divided into separate functional modules: a process model for predicting outputs, a metrology forecast model for predicting measurements, and an actor model for determining input adjustments. This segmentation allows each component to be optimized for its specific function while maintaining overall system reliability.
Solution Approach 2:
The actor model serves as an intermediary component that bridges the gap between predicted metrology values and actual process input adjustments. It reliably translates predictions into actionable control decisions without requiring direct inversion of the complex process model.
2Manufacturing precision
If real-time run-to-run control is implemented, then manufacturing precision is improved, but the complexity of data processing and model coordination increases
Solution Approach 1:
The system performs preliminary predictions using the process model and metrology forecast model before actual measurements are available. This allows control decisions to be made in real-time based on predicted outcomes rather than waiting for metrology results, improving precision while managing complexity through advance computation.
Solution Approach 2:
The actor model continuously receives feedback from both actual metrology measurements and predicted values, adjusting process inputs in real-time to maintain manufacturing precision. This closed-loop feedback mechanism enables precise control while distributing computational complexity across multiple specialized models.
3Manufacturing precision
If complex sensor data is leveraged for control, then manufacturing precision improves, but the complexity of data processing increases
Solution Approach 1:
The actor model serves multiple functions simultaneously: it processes complex sensor data, coordinates predictions from multiple models, determines optimal input adjustments, and manages real-time control decisions. This multi-functionality consolidates data processing complexity into a single coordinated component while improving manufacturing precision.
Data Source
AI summary
In some embodiments, a computer-implemented method of controlling a semiconductor manufacturing process is provided. A computing system generates predicted metrology values for a current run and a next run by providing metrology forecast inputs to a metrology forecast model. The computing system generates an updated recipe for executing at least one semiconductor manufacturing process step using the predicted metrology values for the current run and the next run.