Semiconductor Specimen Defect Examination via ML Focus Calibration

Solution Overview

Problem

Current semiconductor examination processes face challenges in maintaining image focus over time and varying patterns due to physical factors like charging effects and position variations, leading to defocusing issues that affect defect detection and measurement accuracy.

Innovation Solution

A computerized system using a machine learning model to estimate image focus and perform runtime focus calibration, adjusting the focus plane to achieve optimal focus scores for each image, ensuring that images are acquired at optimal focus planes for defect examination.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If focus calibration is performed frequently to maintain image quality, then measurement precision is improved, but productivity deteriorates due to time loss from repeated calibrations

Engineering Contradiction:

Improveimage focus qualityVSAvoidexamination throughput

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The system performs focus calibration in advance by acquiring a calibration image and determining the optimal focus plane before actual defect examination begins. This preliminary calibration action ensures that subsequent examination images are captured at the correct focus, eliminating the need for frequent interruptions during production examination.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system uses feedback from calibration image analysis to automatically adjust the focus plane setting. By analyzing the calibration image to determine focus quality metrics and automatically adjusting the focus plane based on this feedback, the system maintains optimal focus without requiring manual intervention or frequent recalibration, thus preserving productivity.

Inventive Principle:

Principle #23Feedback

2Productivity

If focus calibration is performed less frequently to maintain productivity, then productivity is improved, but measurement precision deteriorates due to defocusing issues

Engineering Contradiction:

Improveexamination throughputVSAvoidimage focus quality

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The system performs focus calibration in advance by acquiring a calibration image and determining the optimal focus plane before actual defect examination begins. This preliminary calibration action ensures that subsequent examination images are captured at the correct focus, eliminating the need for frequent interruptions during production examination.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system uses feedback from calibration image analysis to automatically adjust the focus plane setting. By analyzing the calibration image to determine focus quality metrics and automatically adjusting the focus plane based on this feedback, the system maintains optimal focus without requiring manual intervention or frequent recalibration, thus preserving productivity.

Inventive Principle:

Principle #23Feedback

3Measurement precision

If manual focus adjustment is used to achieve optimal focus, then measurement precision is improved, but device complexity increases due to additional hardware requirements

Engineering Contradiction:

Improveimage focus qualityVSAvoidhardware configuration

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs automatic focus calibration by acquiring calibration images, analyzing focus quality metrics from these images, and self-adjusting the focus plane setting without requiring manual intervention. This self-service capability eliminates the need for additional manual focus adjustment hardware while maintaining optimal focus quality through automated image-based feedback.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS11961221B2Defect examination on a semiconductor specimen

Publication Date: 2024.04.16 APPL MATERIALS ISRAEL LTD

AI summary

There is provided a system and method of runtime defect examination of a semiconductor specimen, comprising obtaining a first image representative of at least part of the semiconductor specimen, the first image acquired by an examination tool configured with a first focus plane; estimating whether the first image is in focus using a machine learning (ML) model, wherein the ML model is previously trained for classifying images into focused images and defocused images; upon an estimation that the first image is out of focus, performing focus calibration on the examination tool to select a second focus plane associated with an optimal focus score; and obtaining a second image acquired by the examination tool configured with the second focus plane, and estimating whether the second image is in focus using the ML model. The second image, upon being estimated as being in focus, is usable for defect examination on the specimen.