Wafer Defect Detection Using Base Pattern Fingerprint Data

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing defect detection techniques for semiconductor structures on wafers, such as D2DB defect detection, are error-prone, time-consuming, and require significant manual efforts, and are not process or imaging modality stable, often failing to accurately compare microscopic images with CAD layouts due to missing fabrication and imaging modality features.

Innovation Solution

A method using fingerprint data for base pattern classes associated with semiconductor structures, which allows for accurate defect detection without the need for mapping between microscopic images and CAD layouts, by obtaining and comparing fingerprint data from a database with microscopic images, and populating a database with fingerprint data for each base pattern class to facilitate defect detection.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a direct comparison between microscopic images and CAD layout is performed, then the defect detection process is simple, but the accuracy is poor due to missing fabrication and imaging modality features

Engineering Contradiction:
Improvedefect detection accuracyVSAvoidcomparison process complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary approach by using base pattern classes as mediators between the CAD layout and microscopic images. Instead of directly comparing the detailed microscopic images with the simplified CAD layout, the system first extracts base pattern classes from the CAD layout and compares these abstracted patterns with corresponding regions in the microscopic images. This intermediary layer captures essential structural information while filtering out irrelevant fabrication-specific details, thereby improving detection accuracy without requiring complex direct comparison algorithms.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent segments the semiconductor structure into base pattern classes, which are fundamental repeating units or motifs. By dividing the complex semiconductor structure into these basic building blocks, the system can compare only the essential pattern information rather than processing the entire detailed microscopic image against the entire CAD layout. This segmentation reduces the comparison complexity while maintaining the ability to detect defects in the segmented pattern elements.

Inventive Principle:
Principle #1Segmentation

2Measurement precision

If synthetic microscopic image generation based on CAD layout is performed, then the comparison can account for fabrication features, but the process becomes difficult and error-prone due to lack of detailed process knowledge

Engineering Contradiction:
Improvedefect detection accuracyVSAvoidsimulation process ease
Core Design Contradiction:
Measurement precisionVSEase of manufacture

Solution Approach 1:

The patent extracts only the essential base pattern classes from the CAD layout, separating the critical structural information from the detailed fabrication process parameters. By taking out just the base pattern information rather than attempting to generate a complete synthetic image, the system avoids the complexity of simulating every fabrication step while still capturing the essential structural features needed for accurate defect detection.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Instead of generating a synthetic microscopic image from the CAD layout by simulating fabrication processes (the conventional approach), the patent inverts the approach by extracting base pattern classes from the CAD layout and using these abstracted patterns as the reference for comparison. This inversion eliminates the need for complex synthetic image generation while maintaining the ability to account for fabrication features through the base pattern class extraction process.

Inventive Principle:
Principle #13The other way round (Inversion)

3Adaptability or versatility

If manual mapping between microscopic images and CAD layout is performed, then the comparison can be adapted to process variations, but the process becomes time-consuming and requires significant manual effort

Engineering Contradiction:
Improveprocess adaptation capabilityVSAvoidmanual processing time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent performs preliminary action by pre-extracting base pattern classes from the CAD layout before the actual defect detection process. These base pattern classes are prepared in advance and stored as reference data. During the defect detection phase, the system simply needs to compare the base pattern classes with corresponding regions in the microscopic images, eliminating the need for time-consuming manual mapping during the actual inspection process.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent creates a simplified copy or representation of the CAD layout in the form of base pattern classes. These base pattern class copies capture the essential structural information without the complexity of the full CAD layout or the variability of manual mapping. The system then compares these standardized base pattern class copies with the microscopic images, providing a consistent and efficient comparison process that adapts to process variations without requiring manual intervention.

Inventive Principle:
Principle #26Copying

Data Source

PatentUS20230260105A1Defect detection for semiconductor structures on a wafer
Publication Date: 2023.08.17 CARL ZEISS SMT GMBH
  • US20230260105A1 patent drawing
  • US20230260105A1 patent drawing
  • US20230260105A1 patent drawing

AI summary

A method of a defect detection of a plurality of semiconductor structures arranged on a wafer includes obtaining a microscopic image of the wafer. The microscopic image depicts the plurality of semiconductor structures. The method also includes obtaining, from a database, fingerprint data for each base pattern class of a set of base pattern classes associated with respective one or more semiconductor structures of the plurality of semiconductor structures. The method further includes performing the defect detection based on the fingerprint data and the microscopic image.