Processing Chamber Pressure Control With Real-Time Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing semiconductor processing systems face challenges in accurately controlling chamber pressure during recipe step transitions, leading to pressure oscillations, increased settling time, and variability across multiple chambers due to non-linear dynamic responses and unmodeled system errors, which affect processing throughput and plasma stability.
Innovation Solution
Implementing real-time learning and optimization of dynamic pressure responses by measuring and analyzing pressure data during recipe step transitions, comparing it to a pressure-based model, and adjusting valve operations to reduce errors and optimize pressure control, thereby reducing pressure oscillations and settling time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If real-time learning and optimization is implemented by measuring and analyzing pressure data during recipe step transitions, then pressure control precision is improved, but device complexity increases
Solution Approach 1:
The system measures actual pressure responses during recipe step transitions, compares them to model predictions, and uses the error information to calculate and apply adjustments to valve operations in subsequent runs. This closed-loop feedback mechanism continuously improves pressure control precision by learning from past performance and adapting control parameters accordingly.
Solution Approach 2:
The system performs self-optimization by automatically measuring its own pressure responses, analyzing the data, calculating optimal adjustments, and applying corrections without external intervention. The real-time learning capability enables the system to autonomously improve its pressure control performance across multiple recipe runs.
2Stability of the object's composition
If valve operations are adjusted to reduce pressure oscillations, then pressure stability is improved, but response time increases
Solution Approach 1:
The system calculates optimal valve adjustments in advance based on learned error patterns from previous recipe runs. By pre-determining the correction amount before the next transition occurs, the system can apply the adjustment immediately when needed, achieving both stability and fast response without the trade-off.
Solution Approach 2:
The system dynamically adapts valve operation adjustments based on the specific conditions of each recipe step transition. Rather than using fixed control parameters, the system modifies valve adjustments in real-time based on actual pressure responses and learned error patterns, enabling optimal balance between stability and response speed for each transition scenario.
Data Source
AI summary
System and methods of improving dynamic pressure response during recipe step transitions. An exemplary method may include changing at least one of a plurality of recipe parameters in accordance with a processing recipe while running the processing recipe on a semiconductor substrate in a processing chamber. The method may further include measuring a pressure response in the processing chamber responsive to the changing of the at least one of the plurality of recipe parameters, and determining a response error based on the pressure response and a model pressure response calculated based on the processing recipe. The method may further include, in response to determining that the response error may be greater than a threshold value, calculating an adjustment to an operation of a valve downstream of the processing chamber when changing the at least one of the plurality of recipe parameters in accordance with the processing recipe in subsequent runs.