اشترك بالحزمة الذهبية واحصل على وصول غير محدود شمرا أكاديميا
تسجيل مستخدم جديدA deep learning approach to identify local structures in atomic-resolution transmission electron microscopy images
78
0
0.0
(
0
)
اسأل ChatGPT حول البحث
ﻻ يوجد ملخص باللغة العربية
Recording atomic-resolution transmission electron microscopy (TEM) images is becoming increasingly routine. A new bottleneck is then analyzing this information, which often involves time-consuming manual structural identification. We have developed a deep learning-based algorithm for recognition of the local structure in TEM images, which is stable to microscope parameters and noise. The neural network is trained entirely from simulation but is capable of making reliable predictions on experimental images. We apply the method to single sheets of defected graphene, and to metallic nanoparticles on an oxide support.
قيم البحث
اقرأ أيضاً
Chiral indices determine important properties of carbon nanotubes (CNTs). Unfortunately, their determination from high-resolution transmission electron microscopy (HRTEM) images, the most accurate method for assigning chirality, is a tedious task. We
develop a Convolutional Neural Network that automatizes this process. A large and realistic training data set of CNT images is obtained by means of atomistic computer simulations coupled with the multi-slice approach for image generation. In most cases, results of the automated assignment are in excellent agreement with manual classification, and the origin of failures is identified. The current approach, which combines HRTEM imaging and deep learning algorithms allows the analysis of a statistically significant number of HRTEM images of carbon nanotubes, paving the way for robust estimates of experimental chiral distributions.
Single atoms can be considered as basic objects for electron microscopy to test the microscope performance and basic concepts for modeling of image contrast. In this work high-resolution transmission electron microscopy was applied to image single pl
atinum atoms in an aberration-corrected transmission electron microscope. The atoms are deposited on a self-assembled monolayer substrate which induces only negligible contrast. Single-atom contrast simulations were performed on the basis of Weickenmeier-Kohl and Doyle-Turner scattering factors. Experimental and simulated intensities are in full agreement on an absolute scale.
Quantitative differential phase contrast imaging of materials in atomic-resolution scanning transmission electron microscopy using segmented detectors is limited by various factors, including coherent and incoherent aberrations, detector positioning
and uniformity, and scan-distortion. By comparing experimental case studies of monolayer and few-layer graphene with image simulations, we explore which parameters require the most precise characterisation for reliable and quantitative interpretation of the reconstructed phases. Coherent and incoherent lens aberrations are found to have the most significant impact. For images over a large field of view, the impact of noise and non-periodic boundary conditions are appreciable, but in this case study have less of an impact than artefacts introduced by beam deflections coupling to beam scanning (imperfect tilt-shift purity).
For quantitative electron microscopy high precision position information is necessary so that besides an adequate resolution and sufficiently strong contrast of atoms, small width of peaks which represent atoms in structural images is needed. Size of
peak is determined by point spread (PS) of instruments as well as that of atoms when point resolution reach the subangstrom scale and thus PS of instruments is comparable with that of atoms. In this article, relationship between PS with atomic numbers, sample thickness, and spherical aberration coefficients will be studied in both negative Cs imaging (NCSI) and positive Cs imaging (PCSI) modes by means of dynamical image simulation. Through comparing the peak width with different thickness and different values of spherical aberration, NCSI mode is found to be superior to PCSI considering smaller peak width in the structural image.
Progress in functional materials discovery has been accelerated by advances in high throughput materials synthesis and by the development of high-throughput computation. However, a complementary robust and high throughput structural characterization
framework is still lacking. New methods and tools in the field of machine learning suggest that a highly automated high-throughput structural characterization framework based on atomic-level imaging can establish the crucial statistical link between structure and macroscopic properties. Here we develop a machine learning framework towards this goal. Our framework captures local structural features in images with Zernike polynomials, which is demonstrably noise-robust, flexible, and accurate. These features are then classified into readily interpretable structural motifs with a hierarchical active learning scheme powered by a novel unsupervised two-stage relaxed clustering scheme. We have successfully demonstrated the accuracy and efficiency of the proposed methodology by mapping a full spectrum of structural defects, including point defects, line defects, and planar defects in scanning transmission electron microscopy (STEM) images of various 2D materials, with greatly improved separability over existing methods. Our techniques can be easily and flexibly applied to other types of microscopy data with complex features, providing a solid foundation for automatic, multiscale feature analysis with high veracity.
سجل دخول لتتمكن من نشر تعليقات
التعليقات
جاري جلب التعليقات
سجل دخول لتتمكن من متابعة معايير البحث التي قمت باختيارها
