SEM Depth Estimation Using Simulated Images and Neural Networks

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing scanning electron microscopes (SEMs) provide only two-dimensional planar images, which are inadequate for analyzing the structure of semiconductor samples, and obtaining depth information is challenging due to the lack of sufficient ground truth data for training artificial neural network models.

Innovation Solution

A method involving a simulator and an artificial neural network (ANN) model is used to generate and refine depth maps, utilizing data augmentation techniques to improve the accuracy of depth estimation, even in scenarios where ground truth data is scarce.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If SEM is used for depth information estimation, then analysis speed is improved, but measurement precision deteriorates

Engineering Contradiction:

Improveanalysis speedVSAvoiddepth information accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent creates simulated TEM images from SEM depth maps as virtual copies, allowing the neural network to learn from synthesized training data that combines the speed of SEM with the depth accuracy of TEM, resolving the contradiction between analysis speed and measurement precision

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent transforms the analysis approach by changing from direct TEM imaging to using SEM images enhanced with simulated depth information, altering the fundamental parameters of how depth data is obtained while maintaining both speed and accuracy

Inventive Principle:

Principle #35Parameter changes

2Measurement precision

If TEM is used for depth information estimation, then measurement precision is improved, but device complexity and cost increase

Engineering Contradiction:

Improvedepth information accuracyVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent generates simulated TEM images from SEM depth maps, creating virtual copies that provide TEM-quality training data without requiring actual TEM imaging, thereby reducing device complexity while maintaining measurement precision

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent introduces a simulator as an intermediary that translates SEM depth maps into simulated TEM images, serving as a bridge between the two imaging modalities and eliminating the need for complex TEM systems

Inventive Principle:

Principle #24Intermediary (Mediator)

3Measurement precision

If TEM is used for depth information estimation, then measurement precision is improved, but loss of time and resource consumption increase

Engineering Contradiction:

Improvedepth information accuracyVSAvoiddata acquisition time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary simulation of TEM images from SEM depth maps to create training data in advance, allowing the neural network to be pre-trained with high-quality depth information without requiring time-consuming actual TEM imaging during analysis

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12614296B2Method and apparatus with depth information estimation

Publication Date: 2026.04.28 SAMSUNG ELECTRONICS CO LTD

AI summary

A method of estimating depth information includes generating a first simulated image using a simulator provided with a first depth map, training an artificial neural network model based on the first depth map and the first simulated image, generating a second depth map by inputting an actual image into the trained artificial neural network model, and generating a second simulated image using the simulator provided with the second depth map.