Recipe Generation Device for Semiconductor Inspection Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge lies in automatically generating a processing recipe for semiconductor manufacturing using inspection images, which are affected by various imaging conditions, making it difficult to evaluate and achieve the desired processed shape without expert intervention.
Innovation Solution
A processing recipe generation device converts inspection images to reduce the impact of imaging conditions, using a structure prediction unit to generate a predicted image and iteratively correct the recipe to match a target image, eliminating the need for manual dimension measurement and expert input.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If a processing recipe is automatically generated using the inspection image itself, then expert intervention is eliminated and automation is improved, but the inspection image varies due to imaging conditions making automatic generation difficult
Solution Approach 1:
The patent introduces a conversion process as an intermediary between the raw inspection image and the recipe generation. The inspection image conversion unit converts the inspection image to a standardized format, mediating the variability caused by different imaging conditions. This conversion acts as a buffer that transforms diverse input images into a consistent representation suitable for automatic recipe generation, thereby enabling automation while maintaining reliability.
Solution Approach 2:
The patent changes the parameters of the inspection image through conversion processes. The conversion condition includes adjusting imaging parameters such as acceleration voltage and degree of charge to standardize the image representation. By transforming the image parameters to a common reference frame, the system can reliably generate recipes automatically despite variations in original imaging conditions.
2Manufacturing precision
If dimension measurement by expert is used to adjust processing recipe, then recipe accuracy is improved, but manual operation and time consumption increase
Solution Approach 1:
The system enables self-service by allowing the inspection image and conversion result to directly inform recipe generation without requiring expert measurement. The structure prediction unit and recipe generation unit work together to automatically determine processing parameters based on the converted image, making the system self-sufficient and eliminating the need for manual expert intervention in dimension measurement and recipe adjustment.
Solution Approach 2:
The patent replaces the mechanical process of manual dimension measurement with an automated image processing and analysis system. Instead of experts physically measuring dimensions, the system uses conversion units and structure prediction algorithms to extract relevant information from the inspection image, substituting human mechanical measurement with automated computational analysis.
3Extent of automation
If inspection image is used directly for recipe generation, then automation is improved, but the varying imaging conditions reduce measurement precision
Solution Approach 1:
The patent applies preliminary action by performing image conversion before recipe generation. The inspection image conversion unit pre-processes the inspection image to eliminate the effects of varying imaging conditions. This preliminary conversion ensures that when the structure prediction unit and recipe generation unit process the image, they work with standardized data, thereby maintaining measurement precision while enabling automation.
Data Source
AI summary
An object of the present invention is to provide a technique that allows automatically generating a processing recipe from an inspection image even when the inspection image varies by being affected by an imaging condition of a processing device. A processing recipe generation device according to the present invention generates a converted image in which components relying on the imaging condition of the inspection image are reduced and generates a processing recipe using a target image generated using a conversion condition same as that of the converted image (see FIG. 2).