Optical Wafer Overlay Estimation Using Machine Learning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Scanning electron microscopes provide high resolution for overlay metrology but have low throughput, while optical inspection lacks the accuracy needed for precise overlay measurement due to feature sizes below the diffraction limit, limiting the ability to resolve structures and extract overlay information accurately.

Innovation Solution

Employing machine learning to analyze optical images of semiconductor structures, using trained models to estimate overlay offsets between process layers, enhancing the accuracy of optical inspection for overlay metrology.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If scanning electron microscopy is used for overlay metrology, then measurement precision is improved, but productivity deteriorates

Engineering Contradiction:

Improveoverlay measurement accuracyVSAvoidthroughput

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent replaces scanning electron microscopy (a mechanical/electronic imaging system) with optical inspection systems combined with machine learning algorithms. This substitution allows the system to achieve SEM-level overlay measurement accuracy using optical methods, thereby dramatically improving throughput while maintaining measurement precision.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If optical inspection is used for overlay metrology, then productivity is improved, but measurement precision deteriorates

Engineering Contradiction:

ImprovethroughputVSAvoidoverlay measurement accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent transforms optical inspection from a qualitative imaging tool to a quantitative measurement tool by changing the analysis parameters. Machine learning models analyze optical images to extract overlay information that was previously invisible, effectively changing the measurement parameters from direct structural visualization to statistical pattern recognition, thereby achieving high precision with optical methods.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent introduces machine learning algorithms as an intermediary between optical images and overlay measurements. These algorithms act as a mediator that extracts meaningful overlay information from optical images that classical algorithms cannot resolve, bridging the gap between optical inspection capabilities and overlay measurement requirements.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Ease of operation

If classical algorithms are used to extract overlay information from optical images, then ease of operation is maintained, but measurement precision deteriorates

Engineering Contradiction:

Improvealgorithm simplicityVSAvoidoverlay measurement accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent transitions from static, fixed algorithms to dynamic, adaptive machine learning models. These models can adapt to different semiconductor structures, process conditions, and image qualities, automatically optimizing their analysis approach. This dynamic capability enables high measurement precision while maintaining ease of operation, as the system self-adjusts without requiring manual reconfiguration.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS12535744B2Overlay estimation based on optical inspection and machine learning

Publication Date: 2026.01.27 KLA CORP

AI summary

One or more optical images of a portion of a semiconductor wafer are obtained. The one or more optical images show a first structure in a first process layer and a second structure in a second process layer. The one or more optical images are provided to a machine-learning model trained to estimate an overlay offset between the first structure and the second structure. An estimated overlay offset between the first structure and the second structure is obtained from the machine-learning model.