Context Offset Estimation in Semiconductor Run-to-Run Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current run-to-run control systems in semiconductor fabrication require numerous pilot runs to estimate offsets, which are expensive and time-consuming, especially in highly-mixed fabrication environments where many different products and contexts are involved.
Innovation Solution
A state-space model is developed to estimate individual context offsets for each thread, allowing these offsets to be calculated without additional pilot runs for subsequent threads with similar contexts, thereby reducing the number of required pilot runs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a pilot run is performed for each thread to estimate offsets, then the accuracy of offset estimation is improved, but the time consumption and cost increase significantly
Solution Approach 1:
The patent segments the offset estimation problem by decomposing thread-specific offsets into context-specific offset components. Instead of estimating a complete offset for each entire thread through separate pilot runs, the system divides the offset into manageable context-level segments (e.g., tool context, product context, layer context) that can be estimated independently and then combined. This segmentation allows reuse of context offset estimates across multiple threads sharing the same contexts, dramatically reducing the number of required pilot runs while maintaining estimation accuracy.
2Adaptability or versatility
If the number of products and contexts increases in highly-mixed fabrication, then the versatility of the fabrication facility is improved, but the number of required pilot runs increases exponentially
Solution Approach 1:
The patent implements universality by creating context offset estimates that serve multiple threads simultaneously. A context offset estimate derived from one thread can be universally applied to other threads that share the same context (e.g., the same tool, product, or layer), making the offset estimation process multi-functional. This universal approach allows a single context offset estimate to benefit multiple threads, reducing the total number of pilot runs required as the facility handles more diverse products and contexts.
3Measurement precision
If separate pilot runs are performed for each unique combination of product, tool, and layer, then the precision of thread-specific offset estimation is improved, but the complexity of the control system increases
Solution Approach 1:
The patent merges the offset estimation process by combining context-specific offset components to form thread-specific offsets. Instead of treating each thread's offset as a completely separate entity requiring independent estimation, the system merges reusable context offset estimates (from shared tool, product, layer contexts) to construct thread-specific offsets. This merging approach maintains thread-specific precision while avoiding the complexity of completely separate estimation processes for each thread.
Data Source
AI summary
A method and system for estimating context offsets for run-to-run control in a semiconductor fabrication facility is described. In one embodiment, contexts associated with a process are identified. The process has one or more threads, and each thread involves one or more contexts. A set of input-output equations describing the process is defined. Each input-output equation corresponds to a thread and includes a thread offset expressed as a summation of individual context offsets. A state-space model is created that describes an evolution of the process using the set of input-output equations. The state-space model allows to estimate individual context offsets.