Semiconductor Mask Inspection via Die-to-Die and Database Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The downsizing of semiconductor devices leads to a lowered degree of pattern forming faithfulness in mask manufacturing, resulting in false defects during inspection, which increases workload and limits the reliability of automated classification, causing a significant number of harmless defects to be misidentified as true defects, overwhelming defect inspection apparatuses.
Innovation Solution
A mask forming method that includes generating design data with identical repetition patterns, creating inspection control information for positional and sensitivity data, and using a combination of Die-to-Database and Die-to-Die comparison methods for defect inspection, where repetition patterns are inspected with higher sensitivity and non-repetition patterns with reduced sensitivity to minimize false defect detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high sensitivity defect detection is applied to mask patterns with fine graphics formed by OPC process, then defect detection sensitivity is improved, but false defects increase due to lowered pattern forming faithfulness
Solution Approach 1:
The mask pattern is segmented into multiple types: repetition patterns (first pattern type) and non-repetition patterns (second pattern type). Different inspection methods are applied to different segments - Die-to-Die comparison for repetition patterns and Die-to-Database comparison for non-repetition patterns. This segmentation allows optimization of inspection sensitivity for each pattern type separately, reducing false defects from repetition patterns while maintaining detection capability for non-repetition patterns.
Solution Approach 2:
Different inspection sensitivities and methods are applied to different regions/pattern types on the mask. Repetition patterns are inspected with Die-to-Die comparison at controlled sensitivity to reduce false defects, while non-repetition patterns are inspected with Die-to-Database comparison at higher sensitivity. This local differentiation of inspection quality resolves the contradiction between overall sensitivity and false defect rate.
2Productivity
If automated classification of defects is used to handle false defects, then productivity is improved, but reliability of classification is insufficient
Solution Approach 1:
The inspection method performs preliminary action by preventing false defect generation at the source through selective application of Die-to-Die comparison for repetition patterns. By controlling the inspection approach before defects are detected and classified, the system reduces the burden on automated classification systems and eliminates the need for complex post-detection classification work.
3Measurement precision
If Die-to-Database comparison method is used for mask defect inspection, then measurement precision is improved, but device complexity increases due to storage requirements
Solution Approach 1:
The inspection system is segmented to apply Die-to-Database comparison only to non-repetition patterns (second pattern type) rather than the entire mask. Repetition patterns are handled by Die-to-Die comparison which has lower storage requirements. This segmentation reduces the overall storage capacity requirements while maintaining high detection accuracy for patterns that require it.
Data Source
AI summary
A mask forming method includes preparing design data of mask including pattern regions having identical repetition patterns respectively, generating mask pattern data of mask based on the design data, generating inspection control information for controlling inspection of defect on mask based on the mask pattern data, the information including positional information of the pattern regions and inspection sensitivity information of the repetition pattern, providing the inspection control information to mask pattern data, forming mask pattern of mask based on the mask pattern data, and inspecting the mask pattern based on the mask pattern data comprising inspecting portion in the mask pattern different from the pattern regions by Die-to-Database comparison method, the inspecting the portion including selecting portion corresponding to repetition pattern from the mask pattern based on the positional information, and inspecting the selected portion by Die-to-Die comparison method at an inspection sensitivity corresponding to inspection sensitivity information.