Blank Substrate Inspection Using AI for Exposure Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing inspection methods for photolithography processes in semiconductor manufacturing are limited by the influence of pattern shape, making it difficult to assess optical performance of exposure apparatuses independently of pattern formation.
Innovation Solution
An information processing apparatus uses a learning model to analyze captured images of developed substrates without patterns, utilizing machine learning to determine the developed state based on learned relationships between exposure amounts and substrate states.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pattern shape comparison is used to identify defects, then pattern shape defects can be detected, but the inspection is influenced by pattern shape making it difficult to assess optical performance independently
Solution Approach 1:
The patent extracts the inspection object from the patterned substrate context and inspects a blank substrate instead. By removing the pattern shape comparison step and inspecting a substrate without patterns, the inspection isolates optical performance assessment from pattern shape influences, allowing pure optical characteristic evaluation while maintaining defect detection capability through developed state analysis
2Measurement precision
If substrates with and without background patterns are exposed and compared, then exposure amount can be determined, but the inspection process becomes complex and time-consuming
Solution Approach 1:
The patent extracts only the essential inspection element (substrate developed state) while removing the complex comparison process between patterned and non-patterned substrates. By inspecting only the developed state of a blank substrate using machine learning, the system determines exposure amount without requiring multiple substrate types or complex comparison procedures
Solution Approach 2:
The patent replaces the mechanical/physical comparison method (visual or manual comparison of pattern shapes) with a machine learning-based automated inspection system. The learning model processes captured images of developed substrates to determine exposure amounts, substituting complex manual comparison procedures with automated intelligent processing
3Measurement precision
If pattern shape comparison is performed to identify defects, then defect location can be identified, but variability in determination results increases due to pattern shape influences
Solution Approach 1:
The patent extracts the inspection focus from pattern shape analysis to developed state analysis of blank substrates. By removing pattern shape comparison entirely and inspecting only the developed state characteristics, the system eliminates variability introduced by pattern shape differences, achieving more consistent and reliable determination results
Solution Approach 2:
The patent changes the inspection parameter from pattern shape characteristics to developed state characteristics. By evaluating the developed state (resist film removal characteristics) rather than pattern shape, the inspection achieves parameter-independent results that are not affected by pattern shape variations, improving reliability
Data Source
AI summary
An information processing apparatus configured to inspect a developed state of a substrate includes an inspection unit configured to input a captured image of a developed second substrate to a learning model to acquire inspection data that include information indicating a second developed state of the second substrate. The learning model is obtained by learning using learning data that include a captured image of a developed first substrate and information indicating a first developed state of the first substrate.