Semiconductor Metrology Using Electron Microscopy Image Analysis

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current semiconductor metrology methods using electron-microscope images are noisy and inaccurate for measuring complex structures, fail to determine non-geometric parameters like lithographic focus and dose, and do not effectively utilize information outside measured features.

Innovation Solution

A computer-based system that trains models to predict semiconductor-fabrication parameters from electron-microscope images, using neural networks and regression analysis to improve accuracy and exploit image information beyond measured features.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional explicit measurement methods are used to measure dimensions in electron-microscope images, then measurement process is simple, but measurement precision deteriorates due to noise and inaccuracy

Engineering Contradiction:

Improvedimension measurement accuracyVSAvoidmeasurement system complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces traditional mechanical/geometric measurement methods with machine learning-based image analysis. Neural networks and regression models analyze electron-microscope images to predict fabrication parameters, substituting direct dimensional measurement with pattern recognition and statistical analysis, thereby improving precision while accepting increased computational complexity

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent transforms the measurement approach by changing from direct geometric parameter extraction to indirect prediction through trained models. The system uses electron-microscope images as input and predicts fabrication parameters through learned relationships, changing the measurement paradigm from direct to indirect observation

Inventive Principle:

Principle #35Parameter changes

2Loss of information

If traditional measurement methods focus only on measured features, then measurement process is straightforward, but loss of information increases due to ignoring image context

Engineering Contradiction:

Improveimage information utilizationVSAvoidparameter determination difficulty

Core Design Contradiction:

Loss of informationVSDifficulty of detecting and measuring

Solution Approach 1:

The patent makes the measurement system universal by enabling it to determine multiple types of parameters (geometric and non-geometric) from the same electron-microscope images. The trained models can predict various fabrication parameters simultaneously, making the system multi-functional rather than specialized for single measurements

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent introduces trained machine learning models as intermediaries between the raw electron-microscope images and the fabrication parameters. These models act as mediators that extract and interpret information from images, translating visual data into quantitative parameter predictions without direct geometric measurement

Inventive Principle:

Principle #24Intermediary (Mediator)

3Adaptability or versatility

If explicit dimensional measurements are performed, then geometric parameters can be obtained, but non-geometric parameters (lithographic focus, dose) cannot be determined

Engineering Contradiction:

Improveparameter type coverageVSAvoidparameter determination accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent replaces geometric measurement mechanisms with statistical prediction mechanisms. Instead of measuring physical dimensions to infer parameters, the system uses trained models that predict both geometric and non-geometric parameters directly from image patterns, expanding parameter coverage beyond traditional measurement capabilities

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent changes the nature of measurable parameters from exclusively geometric to including non-geometric fabrication parameters. By training models on diverse parameter types, the system transforms the measurement output to include lithographic focus, dose, and other process parameters that cannot be obtained through dimensional measurement alone

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS10580673B2Semiconductor metrology and defect classification using electron microscopy

Publication Date: 2020.03.03 KLA CORP

AI summary

In some embodiments, a first plurality of electron-microscope images for respective instances of a semiconductor structure is obtained from a first source. The electron-microscope images of the first plurality show different values of one or more semiconductor-fabrication parameters. A model is trained that specifies a relationship between the first plurality of electron-microscope images and the values of the one or more semiconductor-fabrication parameters. A second plurality of electron-microscope images for respective instances of the semiconductor structure on one or more semiconductor wafers is collected. The one or more semiconductor wafers are distinct from the first source. Values of the one or more semiconductor-fabrication parameters for the second plurality of electron-microscope images are predicted using the model.