Universal Correlation Model for Substrate Film Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In semiconductor fabrication, accurately calculating film thickness across multiple substrate treatment systems is challenging due to differences in image states, requiring cumbersome generation of correlation data for each system, especially when multiple systems perform common treatments.
Innovation Solution
A substrate inspection method that creates a correlation model based on pixel values from a captured image of a model creating substrate, allowing for estimation of film feature amounts in other systems without needing individual correlation models for each, and calculates an offset amount to correct feature amount calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If individual correlation models are created for each substrate treatment system, then measurement precision is improved, but device complexity and time consumption increase
Solution Approach 1:
The patent creates a single universal correlation model that can be applied across multiple substrate treatment systems. Instead of generating separate correlation models for each system, the invention uses one standardized model that works universally, thereby reducing the complexity of model management while maintaining measurement precision through the offset amount correction mechanism.
Solution Approach 2:
The patent introduces an offset amount as an intermediary parameter that mediates between the universal correlation model and the specific measurement conditions of different substrate treatment systems. This offset amount corrects the estimated feature amount to align with actual measurements, enabling a single model to accurately serve multiple systems without requiring system-specific customization.
2Measurement precision
If individual correlation models are created for each substrate treatment system, then measurement accuracy is improved, but loss of time increases
Solution Approach 1:
The patent establishes a universal correlation model that serves multiple substrate treatment systems simultaneously. This eliminates the need to create separate models for each system, dramatically reducing the time required for model creation and setup while maintaining accurate feature amount calculations through the offset correction approach.
Solution Approach 2:
The patent performs preliminary calibration to determine the offset amount before actual measurements are taken. This preliminary action establishes the correction parameter in advance, so that subsequent measurements using the universal correlation model can proceed quickly without requiring time-consuming system-specific model creation for each measurement task.
3Ease of operation
If a universal correlation model is used across multiple substrate treatment systems, then ease of operation is improved, but measurement precision may deteriorate
Solution Approach 1:
The patent introduces an offset amount as a corrective intermediary that compensates for variations between different substrate treatment systems. This offset amount is determined through preliminary calibration and is applied to the estimated feature amount from the universal correlation model, thereby maintaining measurement precision despite the simplicity of using a single model across multiple systems.
Solution Approach 2:
The patent adjusts the correlation model by applying an offset amount parameter that is specific to each substrate treatment system or measurement condition. This parameter change allows the universal model to adapt to different systems, maintaining precision while preserving the ease of operation associated with using a single standardized model.
Data Source
AI summary
A method of inspecting a substrate, includes: creating a model indicating a relation between a pixel value in a captured image of the substrate and a feature amount of a film on the substrate, based on a measured feature amount of a film on a creating substrate and a captured image generated by imaging the creating substrate by an apparatus in a first system; imaging an object substrate by an apparatus in a second system to generate a captured image, and calculating an estimated feature amount of a film on the object substrate, based on the captured image and the model; calculating a statistical value of the estimated feature amounts of the object substrates; and calculating an offset amount for the estimated feature amount from a measured feature amount of a film formed by performing a same treatment on an offset substrate in the second system and the statistical value.