Multi-Scale Autoencoders for Semiconductor Feature Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing metrology solutions struggle to accurately characterize and inspect crystalline material features in semiconductor workpieces at multiple length scales, from workpiece-sized features down to micron-scale or sub-micron features, which are challenging due to their specific spatial distributions and the evolution of new features during crystal growth and processing.

Innovation Solution

A method utilizing multi-scale autoencoders, specifically deep convolutional multiscale variational autoencoders (MS-VAEs), processes workpiece images to generate residual images at various scales, enabling efficient detection and characterization of features by encoding and decoding semiconductor workpiece images, including downsampling, upsampling, and subtracting smoothed images to highlight higher-resolution details.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If high-resolution image processing is used to detect features at multiple length scales, then measurement precision is improved, but computational resources and processing time increase significantly

Engineering Contradiction:

Improvefeature detection accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the image processing task by applying different resolution levels to different spatial scales. Low-resolution images process workpiece-sized features, while high-resolution images process micron-scale features. This segmentation allows efficient processing by matching computational resources to the actual detection needs at each scale, reducing overall processing time while maintaining measurement precision.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces a resolution dimension to the processing system, creating a multi-resolution framework. By transforming the single-resolution processing into multi-resolution processing, the system can efficiently handle features at different scales simultaneously, reducing the computational burden of processing entire high-resolution images for all feature types.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Measurement precision

If high-resolution image processing is used to detect features at multiple length scales, then measurement precision is improved, but computational resources increase significantly

Engineering Contradiction:

Improvefeature detection accuracyVSAvoidcomputational resources

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent segments the computational workload by resolution level, assigning low-resolution processing to capture workpiece-sized features and high-resolution processing only to capture micron-scale features. This segmentation reduces the total computational resources required compared to processing full high-resolution images for all feature detection tasks.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies high-resolution processing only partially, specifically only where and when micron-scale features need to be detected, rather than applying it excessively to the entire image processing task. This partial application of high-resolution processing significantly reduces computational resource consumption while maintaining detection accuracy for critical features.

Inventive Principle:

Principle #16Partial or excessive action

3Ease of manufacture

If existing metrology solutions are used to characterize crystalline material features, then manufacturing process is simplified, but measurement precision and feature detection capability are insufficient

Engineering Contradiction:

Improveprocess simplicityVSAvoidfeature characterization accuracy

Core Design Contradiction:

Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent creates a universal processing framework that handles multiple feature types and scales through a single integrated system. The multi-resolution processing approach can detect workpiece-sized features, micron-scale features, and sub-micron features using the same fundamental methodology, improving measurement precision across all feature types while maintaining process simplicity through automation.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent replaces traditional mechanical/optical metrology systems with an automated image processing system using machine learning models. This substitution improves measurement precision by enabling automated detection and characterization of features at multiple scales, while the systematic approach maintains ease of manufacture through programmable processing pipelines.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS20250363621A1Multi-Scale Autoencoders for Semiconductor Workpiece Understanding

Publication Date: 2025.11.27 WOLFSPEED INC

AI summary

An example method includes obtaining a workpiece image of a semiconductor workpiece. The example method includes providing the workpiece image as input to a machine-learned encoding model. The example method includes obtaining an output from the machine-learned encoding model, the output includes an encoding corresponding to the semiconductor workpiece. The example method includes determining one or more characteristics of the semiconductor workpiece based at least in part on the encoding or modifying a semiconductor manufacturing process based at least in part on the encoding.