Source-linked AI summary
Resolution enhancement in scanning electron microscopy using deep learning
Kevin de Haan, Zachary S. Ballard, Yair Rivenson, Yichen Wu, Aydogan Ozcan
TL;DR
SEM can reveal nanoscale features, but electron-beam exposure can charge and damage samples. The paper trains a neural network on co-registered low- and high-resolution SEM images; its outputs recover spatial details and frequency distributions consistent with high-resolution images, while the demonstrated twofold resolution increase corresponds to fourfold faster acquisition.
Problem
SEM provides nanoscale characterization, but its focused electron beam can cause charge buildup and damage, particularly in soft or dielectric materials.
Source: p. 2Method
A neural network trained on co-registered low- and high-resolution SEM image pairs blindly super-resolves a low-resolution image.
Source: pp. 2–3Results
The network recovered spatial details matching corresponding high-resolution SEM images, and its output frequency distribution was consistent with the high-resolution images.
Source: pp. 4, 5, 7Takeaways & Limitations
The demonstrated twofold resolution increase allows fourfold faster image acquisition with fewer electrons interacting with the sample.
Source: p. 7
Abstract
from arXiv · showhide
We report resolution enhancement in scanning electron microscopy (SEM) images using a generative adversarial network. We demonstrate the veracity of this deep learning-based super-resolution technique by inferring unresolved features in low-resolution SEM images and comparing them with the accurately co-registered high-resolution SEM images of the same samples. Through spatial frequency analysis, we also report that our method generates images with frequency spectra matching higher resolution SEM images of the same fields-of-view. By using this technique, higher resolution SEM images can be taken faster, while also reducing both electron charging and damage to the samples.