Semiconductor Chamber Spatial Modeling for Variability Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Semiconductor wafer processing complexity increases with shrinking device sizes, requiring extensive manual tuning of multiple knobs, and lacks precise in-line metrology, leading to suboptimal process solutions and resource wastage.
Innovation Solution
A spatial model using machine-learning to predict performance, combining virtual and on-board metrology data, enabling digital design of experiments to optimize processes without physical wafers, and adaptive control models for chamber-to-chamber and wafer-to-wafer variability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional manufacturing processes are used without advanced modeling, then device performance variability is reduced, but process development time and cost increase significantly
Solution Approach 1:
A spatial model is built during the process development phase to predict chamber performance characteristics before actual manufacturing. This preliminary modeling allows optimization of process parameters and identification of potential variability issues ahead of time, reducing both development time and device performance variability during production
Solution Approach 2:
The system incorporates periodic updates of the process control model using machine learning techniques that learn from actual manufacturing data. This feedback mechanism continuously refines the spatial model predictions, improving manufacturing precision while maintaining efficient process development through adaptive control
2Loss of time
If spatial modeling is implemented for process optimization, then process development time is reduced, but system complexity increases
Solution Approach 1:
A spatial model acts as an intermediary between process parameters and chamber performance predictions. This computational model simplifies the complex relationships by providing a predictive framework that guides process optimization without requiring direct experimentation with all parameter combinations, thus reducing development time while managing system complexity
Solution Approach 2:
The system optimizes process development by changing and adjusting model parameters iteratively during the development phase. Machine learning techniques automatically tune spatial model parameters based on observed data, reducing the need for manual system configuration and lowering operational complexity while accelerating process development
3Manufacturing precision
If periodic model updates are performed using machine learning, then device performance variability is controlled, but computational resources and processing time increase
Solution Approach 1:
The process control model is updated periodically at predetermined intervals (e.g., wafer-to-wafer, lot-to-lot, chamber-to-chamber) rather than continuously. This periodic updating approach maintains device performance variability control through regular model refinement while significantly reducing computational resource consumption compared to continuous updates
Solution Approach 2:
The system applies machine learning updates at selective intervals rather than for every single process run. This partial updating strategy provides sufficient model accuracy to control device performance variability while avoiding the excessive computational burden of updating after every measurement, optimizing the balance between precision and resource usage
Data Source
Figure 1A
Figure 1B
Figure 2~3
AI summary
A spatial model is built to predict performance of a processing chamber. The spatial model is used to converge faster to a desired process during the process development phase. A system for controlling device performance variability during manufacturing includes a process platform, on-board metrology (OBM) tools, and a machine-learning based process control model. The system receives SEM metrology data, and updates the process control model periodically (e.g., wafer-to-wafer, lot-to-lot, chamber-to-chamber etc.) using machine learning techniques. Periodic update of the process control model may account for chamber-to-chamber variability.