Rapid and flexible segmentation of electron microscopy data using few-shot machine learning

AbstractAutomatic segmentation of key microstructural features in atomic-scale electron microscope images is critical to improved understanding of structure–property relationships in many important materials and chemical systems. However, the present paradigm involves time-intensive manual analysis that is inherently biased, error-prone, and unable to accommodate the large volumes of data produced by modern instrumentation. While more automated approaches have been proposed, many are not robust to a high variety of data, and do not generalize well to diverse microstructural features and material systems. Here, we present a flexible, semi-supervised few-shot machine learning approach for segmentation of scanning transmission electron microscopy images of three oxide material systems: (1) epitaxial heterostructures of SrTiO3/Ge, (2) La0.8Sr0.2FeO3 thin films, and (3) MoO3 nanoparticles. We demonstrate that the few-shot learning method is more robust against noise, more reconfigurable, and requires less data than conventional image analysis methods. This approach can enable rapid image classification and microstructural feature mapping needed for emerging high-throughput characterization and autonomous microscope platforms.

Download Full-text

Rapid and Flexible Semantic Segmentation of Electron Microscopy Data Using Few-Shot Machine Learning

10.21203/rs.3.rs-346102/v1 ◽

2021 ◽

Author(s):

Sarah Akers ◽

Elizabeth Kautz ◽

Andrea Trevino-Gavito ◽

Matthew Olszta ◽

Bethany Matthews ◽

...

Keyword(s):

Electron Microscopy ◽

Machine Learning ◽

Semantic Segmentation ◽

Atomic Scale ◽

Oxide Material ◽

Structure Property ◽

Chemical Systems ◽

Material Systems ◽

Scanning Transmission ◽

Microscopy Data

Abstract Semantic segmentation of key microstructural features in atomic-scale electron microscope images is critical to improved understanding of structure-property relationships in many important materials and chemical systems. However, the present paradigm involves time-intensive manual analysis that is inherently biased, error-prone, and unable to accommodate the large volumes of data produced by modern instrumentation. While more automated approaches have been proposed, many are not robust to a high variety of data, and do not generalize well to diverse microstructural features and material systems. Here, we present a flexible, semi-supervised few-shot machine learning approach for semantic segmentation of scanning transmission electron microscopy images of three oxide material systems: (1) epitaxial heterostructures of SrTiO3 / Ge, (2) La0.8Sr0.2FeO3 thin films, and (3) MoO3 nanoparticles. We demonstrate that the few-shot learning method is more robust against noise, more reconfigurable, and requires less data than conventional image analysis methods. This approach can enable rapid image classification and microstructural feature mapping needed for emerging high-throughput and autonomous microscope platforms.

Download Full-text

A Review of Grain Boundary and Heterointerface Characterization in Polycrystalline Oxides by (Scanning) Transmission Electron Microscopy

Crystals ◽

10.3390/cryst11080878 ◽

2021 ◽

Vol 11 (8) ◽

pp. 878

Author(s):

Hasti Vahidi ◽

Komal Syed ◽

Huiming Guo ◽

Xin Wang ◽

Jenna Laurice Wardini ◽

...

Keyword(s):

Electron Microscopy ◽

Transmission Electron Microscopy ◽

Experimental Studies ◽

Atomic Scale ◽

Polycrystalline Materials ◽

Structure Property ◽

Ceramic Oxides ◽

Structure Property Relationships ◽

Transmission Electron ◽

Scanning Transmission

Interfaces such as grain boundaries (GBs) and heterointerfaces (HIs) are known to play a crucial role in structure-property relationships of polycrystalline materials. While several methods have been used to characterize such interfaces, advanced transmission electron microscopy (TEM) and scanning TEM (STEM) techniques have proven to be uniquely powerful tools, enabling quantification of atomic structure, electronic structure, chemistry, order/disorder, and point defect distributions below the atomic scale. This review focuses on recent progress in characterization of polycrystalline oxide interfaces using S/TEM techniques including imaging, analytical spectroscopies such as energy dispersive X-ray spectroscopy (EDXS) and electron energy-loss spectroscopy (EELS) and scanning diffraction methods such as precession electron nano diffraction (PEND) and 4D-STEM. First, a brief introduction to interfaces, GBs, HIs, and relevant techniques is given. Then, experimental studies which directly correlate GB/HI S/TEM characterization with measured properties of polycrystalline oxides are presented to both strengthen our understanding of these interfaces, and to demonstrate the instrumental capabilities available in the S/TEM. Finally, existing challenges and future development opportunities are discussed. In summary, this article is prepared as a guide for scientists and engineers interested in learning about, and/or using advanced S/TEM techniques to characterize interfaces in polycrystalline materials, particularly ceramic oxides.

Download Full-text

Probing vacancies and structural distortions at individual defects in ceramics using EELS

Proceedings, annual meeting, Electron Microscopy Society of America ◽

10.1017/s0424820100165859 ◽

1996 ◽

Vol 54 ◽

pp. 678-679

Author(s):

D. J. Wallis ◽

N. D. Browning

Keyword(s):

Structural Information ◽

Core Loss ◽

Nearest Neighbors ◽

Ceramic Materials ◽

Scanning Transmission Electron Microscope ◽

Real Space ◽

Atomic Scale ◽

Structure Property ◽

Scanning Transmission ◽

Key Issues

In electron energy loss spectroscopy (EELS), the near-edge region of a core-loss edge contains information on high-order atomic correlations. These correlations give details of the 3-D atomic structure which can be elucidated using multiple-scattering (MS) theory. MS calculations use real space clusters making them ideal for use in low-symmetry systems such as defects and interfaces. When coupled with the atomic spatial resolution capabilities of the scanning transmission electron microscope (STEM), there therefore exists the ability to obtain 3-D structural information from individual atomic scale structures. For ceramic materials where the structure-property relationships are dominated by defects and interfaces, this methodology can provide unique information on key issues such as like-ion repulsion and the presence of vacancies, impurities and structural distortion.An example of the use of MS-theory is shown in fig 1, where an experimental oxygen K-edge from SrTiO3 is compared to full MS-calculations for successive shells (a shell consists of neighboring atoms, so that 1 shell includes only nearest neighbors, 2 shells includes first and second-nearest neighbors, and so on).

Download Full-text

Electron Microscopy Studies of The Chemistry And Microstructure of BaF2 / YBa2Cu3O7-x Thin Films

Proceedings, annual meeting, Electron Microscopy Society of America ◽

10.1017/s0424820100171717 ◽

1994 ◽

Vol 52 ◽

pp. 796-797

Author(s):

J. L. Lee ◽

C. A. Weiss ◽

R. A. Buhrman ◽

J. Silcox

Keyword(s):

Thin Film ◽

Electron Microscopy ◽

Thin Films ◽

Scanning Transmission Electron Microscope ◽

Atomic Scale ◽

Island Growth ◽

Multilayer Systems ◽

Transmission Electron ◽

Scanning Transmission ◽

Mgo Substrate

BaF2 thin films are being investigated as candidates for use in YBa2Cu3O7-x (YBCO) / BaF2 thin film multilayer systems, given the favorable dielectric properties of BaF2. In this study, the microstructural and chemical compatibility of BaF2 thin films with YBCO thin films is examined using transmission electron microscopy and microanalysis. The specimen was prepared by using laser ablation to first deposit an approximately 2500 Å thick (0 0 1) YBCO thin film onto a (0 0 1) MgO substrate. An approximately 7500 Å thick (0 0 1) BaF2 thin film was subsequendy thermally evaporated onto the YBCO film.Images from a VG HB501A UHV scanning transmission electron microscope (STEM) operating at 100 kV show that the thickness of the BaF2 film is rather uniform, with the BaF2/YBCO interface being quite flat. Relatively few intrinsic defects, such as hillocks and depressions, were evident in the BaF2 film. Moreover, the hillocks and depressions appear to be faceted along {111} planes, suggesting that the surface is smooth and well-ordered on an atomic scale and that an island growth mechanism is involved in the evolution of the BaF2 film.

Download Full-text

Aberration Measurement and Correction in Scanning Transmission Electron Microscopy using Machine Learning

Microscopy and Microanalysis ◽

10.1017/s1431927621003226 ◽

2021 ◽

Vol 27 (S1) ◽

pp. 814-816

Author(s):

Ryusuke Sagawa ◽

Fuminori Uematsu ◽

Keito Aibara ◽

Tomohiro Nakamichi ◽

Shigeyuki Morishita

Keyword(s):

Electron Microscopy ◽

Machine Learning ◽

Transmission Electron Microscopy ◽

Scanning Transmission Electron Microscopy ◽

Scanning Transmission Electron ◽

Transmission Electron ◽

Scanning Transmission

Download Full-text

Atomic-scale structural and compositional analyses of Ruddlesden-Popper planar faults in AO-excess SrTiO3 (A = Sr2+, Ca2+, Ba2+) ceramics

Journal of Materials Research ◽

10.1557/jmr.2009.0321 ◽

2009 ◽

Vol 24 (8) ◽

pp. 2596-2604 ◽

Cited By ~ 8

Author(s):

Sašo Šturm ◽

Makoto Shiojiri ◽

Miran Čeh

Keyword(s):

Electron Microscopy ◽

Transmission Electron Microscopy ◽

Dark Field ◽

Atomic Scale ◽

Transmission Electron ◽

Initial Stage ◽

Final Microstructure ◽

Scanning Transmission ◽

The Matrix ◽

Annular Dark Field

The microstructure in AO-excess SrTiO3 (A = Sr2+, Ca2+, Ba2+) ceramics is strongly affected by the formation of Ruddlesden-Popper fault–rich (RP fault) lamellae, which are coherently intergrown with the matrix of the perovskite grains. We studied the structure and chemistry of RP faults by applying quantitative high-resolution transmission electron microscopy and high-angle annular dark-field scanning transmission electron microscopy analyses. We showed that the Sr2+ and Ca2+ dopant ions form RP faults during the initial stage of sintering. The final microstructure showed preferentially grown RP fault lamellae embedded in the central part of the anisotropic perovskite grains. In contrast, the dopant Ba2+ ions preferably substituted for Sr2+ in the SrTiO3 matrix by forming a BaxSr1−xTiO3 solid solution. The surplus of Sr2+ ions was compensated structurally in the later stages of sintering by the formation of SrO-rich RP faults. The resulting microstructure showed RP fault lamellae located at the surface of equiaxed BaxSr1-xTiO3 perovskite grains.

Download Full-text