Automated Semiconductor Inspection Recipe Generation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for generating inspection recipes for semiconductor devices are time-consuming and lack effective feedback, requiring multiple iterations and manual management, leading to inefficiencies and inconsistencies across different wafers.

Innovation Solution

A method and apparatus for automated inspection recipe generation that allows offline tuning and refinement using a 'tune map' or 'reference map' database, enabling the selection and classification of dies of interest, and automatic analysis of inspection results to minimize overkill and underkill, without the need for continuous re-inspection of entire wafers.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual recipe tuning is performed by reviewing individual die images and modifying parameters iteratively, then inspection accuracy can be improved, but the time required for recipe generation increases significantly

Engineering Contradiction:
Improveinspection accuracyVSAvoidrecipe generation time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system implements automated feedback by analyzing inspection results and automatically adjusting recipe parameters based on detected defects and performance metrics, eliminating the need for manual parameter tuning while maintaining high inspection accuracy

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The inspection system performs self-tuning by automatically generating and optimizing inspection recipes based on reference wafer data and defect patterns, allowing the system to service itself without operator intervention for recipe creation and optimization

Inventive Principle:
Principle #25Self-service

2Reliability

If the complete wafer is re-inspected multiple times during recipe tuning, then inspection results can be validated, but productivity decreases due to repeated full-wafer inspections

Engineering Contradiction:
Improveinspection result validationVSAvoidwafer inspection throughput
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The system segments the validation process by selecting and inspecting only representative regions or areas of interest on the wafer rather than re-inspecting the entire wafer, maintaining validation reliability while significantly reducing inspection time and increasing productivity

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system performs partial inspection by focusing on critical areas or representative samples during recipe tuning, which is sufficient for validation purposes without requiring complete wafer re-inspection, thereby improving productivity

Inventive Principle:
Principle #16Partial or excessive action

3Measurement precision

If recipe parameters are manually modified and saved as new versions, then inspection performance can be optimized, but device complexity increases due to multiple recipe versions and manual management

Engineering Contradiction:
Improveinspection performanceVSAvoidrecipe management complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system automatically manages recipe versions and parameter optimizations through self-tuning algorithms, eliminating the need for manual recipe version creation and management, thereby reducing operational complexity while maintaining inspection performance

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system optimizes inspection performance by dynamically adjusting recipe parameters based on automated analysis of inspection results and defect patterns, achieving performance improvement without the complexity of manual parameter modification and version management

Inventive Principle:
Principle #35Parameter changes

4Measurement precision

If inspection results are reviewed on individual die level, then detection precision can be maintained, but loss of time increases due to manual review of each die

Engineering Contradiction:
Improvedefect detection precisionVSAvoidresult review time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system implements automated feedback mechanisms that analyze inspection results across multiple dies simultaneously, providing comprehensive defect detection and classification without requiring manual review of individual dies, thus maintaining precision while reducing time loss

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The system replaces manual mechanical review of individual dies with automated computational analysis that processes multiple die images and defect data simultaneously, maintaining detection precision while dramatically reducing the time required for result review

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP2828882B1Method, computer system and apparatus for recipe generation for automated inspection semiconductor devices
Publication Date: 2019.09.18 KLA CORP
  • EP2828882B1 patent drawingFigure 1
  • EP2828882B1 patent drawingFigure 2
  • EP2828882B1 patent drawingFigure 3

AI summary

A method, a computer system and an apparatus are disclosed for inspection recipe generation for the automated inspection of semiconductor devices. In order to generate the inspection recipe a reference data set is used. Automatic inspection is carried out with an initial recipe on images of dies of the reference data set (reference wafermap). The detected inspection results from the automatic inspection are classified and the classified inspection results are compared with an expert classification of defects in dies. Overkill and underkill numbers are automatically generated. According to the overkill and underkill numbers the inspection recipe parameters are modified. Automatic inspection is repeated if the detection and/or the classification are below a predefined threshold.