EDM 1.0: Electron direct methods

Machine learning in electron microscopy for advanced nanocharacterization: current developments, available tools and future outlook

In the last few years, electron microscopy has experienced a new methodological paradigm aimed to fix the bottlenecks and overcome the challenges of its analytical workflow. Machine learning and artificial intelligence are answering this call providing powerful resources towards automation, exploration, and development. In this review, we evaluate the state-of-the-art of machine learning applied to electron microscopy (and obliquely, to materials and nano-sciences). We start from the traditional imaging techniques to reach the newest higher-dimensionality ones, also covering the recent advances in spectroscopy and tomography. Additionally, the present review provides a practical guide for microscopists, and in general for material scientists, but not necessarily advanced machine learning practitioners, to straightforwardly apply the offered set of tools to their own research. To conclude, we explore the state-of-the-art of other disciplines with a broader experience in applying artificial intelligence methods to their research (e.g., high-energy physics, astronomy, Earth sciences, and even robotics, videogames, or marketing and finances), in order to narrow down the incoming future of electron microscopy, its challenges and outlook.

Objective crystallographic symmetry classifications of a noisy crystal pattern with strong Fedorov-type pseudosymmetries and its optimal image-quality enhancement

Statistically sound crystallographic symmetry classifications are obtained with information-theory-based methods in the presence of approximately Gaussian distributed noise. A set of three synthetic patterns with strong Fedorov-type pseudosymmetries and varying amounts of noise serve as examples. Contrary to traditional crystallographic symmetry classifications with an image processing program such as CRISP, the classification process does not need to be supervised by a human being and is free of any subjectively set thresholds in the geometric model selection process. This enables crystallographic symmetry classification of digital images that are more or less periodic in two dimensions (2D), also known as crystal patterns, as recorded with sufficient structural resolution from a wide range of crystalline samples with different types of scanning probe and transmission electron microscopes. Correct symmetry classifications enable the optimal crystallographic processing of such images. That processing consists of the averaging over all asymmetric units in all unit cells in the selected image area and significantly enhances both the signal-to-noise ratio and the structural resolution of a microscopic study of a crystal. For sufficiently complex crystal patterns, the information-theoretic symmetry classification methods are more accurate than both visual classifications by human experts and the recommendations of one of the popular crystallographic image processing programs of electron crystallography.

When does atomic resolution plan view imaging of surfaces work?

Surface structures that are different from the corresponding bulk, reconstructions, are exceedingly difficult to characterize with most experimental methods. Scanning tunneling microscopy, the workhorse for imaging complex surface structures of metals and semiconductors, is not as effective for oxides and other insulating materials. This paper details the use of transmission electron microscopy plan view imaging in conjunction with image processing for solving complex surface structures. We address the issue of extracting the surface structure from a weak signal with a large bulk contribution. This method requires the sample to be thin enough for kinematical assumptions to be valid. The analysis was performed on two sets of data, c(6×2) on the (100) surface and (3×3) on the (111) surface of SrTiO₃, and was unsuccessful in the latter due to the thickness of the sample and a lack of inversion symmetry. The limits and the functionality of this method are discussed.

Transition from Reconstruction toward Thin Film on the (110) Surface of Strontium Titanate

The surfaces of metal oxides often are reconstructed with a geometry and composition that is considerably different from a simple termination of the bulk. Such structures can also be viewed as ultrathin films, epitaxed on a substrate. Here, the reconstructions of the SrTiO3 (110) surface are studied combining scanning tunneling microscopy (STM), transmission electron diffraction, and X-ray absorption spectroscopy (XAS), and analyzed with density functional theory calculations. Whereas SrTiO3 (110) invariably terminates with an overlayer of titania, with increasing density its structure switches from n × 1 to 2 × n. At the same time the coordination of the Ti atoms changes from a network of corner-sharing tetrahedra to a double layer of edge-shared octahedra with bridging units of octahedrally coordinated strontium. This transition from the n × 1 to 2 × n reconstructions is a transition from a pseudomorphically stabilized tetrahedral network toward an octahedral titania thin film with stress-relief from octahedral strontia units at the surface.

Computation in electron microscopy

Some uses of the computer and computation in high-resolution transmission electron microscopy are reviewed. The theory of image calculation using Bloch wave and multislice methods with and without aberration correction is reviewed and some applications are discussed. The inverse problem of reconstructing the specimen structure from an experimentally measured electron microscope image is discussed. Some future directions of software development are given.

Determination of the incommensurate modulated structure of Bi(2)Sr(1.6)La(0.4)CuO(6+δ) by aberration-corrected transmission electron microscopy

The incommensurate modulated structure (IMS) of Bi2Sr1.6La0.4CuO6+δ (BSLCO) has been studied by aberration-corrected transmission electron microscopy in combination with a high-dimensional (HD) space description. Two images are deconvoluted in the negative Cs imaging (NCSI) and positive Cs imaging (PCSI) modes. Similar results for the IMS have been obtained from two corresponding projected potential maps (PPMs), and the size of the dots representing atoms in the NCSI PPM is found to be smaller than that in the PCSI PPM. Considering that the object size is one of the factors that influence the precision of the structural determination, modulation functions for all unoverlapped atoms in BSLCO were determined on the basis of the NCSI PPM in combination with the HD space description.

Surface determination through atomically resolved secondary-electron imaging

Unique determination of the atomic structure of technologically relevant surfaces is often limited by both a need for homogeneous crystals and ambiguity of registration between the surface and bulk. Atomically resolved secondary-electron imaging is extremely sensitive to this registration and is compatible with faceted nanomaterials, but has not been previously utilized for surface structure determination. Here we report a detailed experimental atomic-resolution secondary-electron microscopy analysis of the c(6 × 2) reconstruction on strontium titanate (001) coupled with careful simulation of secondary-electron images, density functional theory calculations and surface monolayer-sensitive aberration-corrected plan-view high-resolution transmission electron microscopy. Our work reveals several unexpected findings, including an amended registry of the surface on the bulk and strontium atoms with unusual seven-fold coordination within a typically high surface coverage of square pyramidal TiO5 units. Dielectric screening is found to play a critical role in attenuating secondary-electron generation processes from valence orbitals.

Structure refinement from precession electron diffraction data

Electron diffraction is a unique tool for analysing the crystal structures of very small crystals. In particular, precession electron diffraction has been shown to be a useful method for ab initio structure solution. In this work it is demonstrated that precession electron diffraction data can also be successfully used for structure refinement, if the dynamical theory of diffraction is used for the calculation of diffracted intensities. The method is demonstrated on data from three materials - silicon, orthopyroxene (Mg,Fe)(2)Si(2)O(6) and gallium-indium tin oxide (Ga,In)(4)Sn(2)O(10). In particular, it is shown that atomic occupancies of mixed crystallographic sites can be refined to an accuracy approaching X-ray or neutron diffraction methods. In comparison with conventional electron diffraction data, the refinement against precession diffraction data yields significantly lower figures of merit, higher accuracy of refined parameters, much broader radii of convergence, especially for the thickness and orientation of the sample, and significantly reduced correlations between the structure parameters. The full dynamical refinement is compared with refinement using kinematical and two-beam approximations, and is shown to be superior to the latter two.

A software tool for automatic analysis of selected area diffraction patterns within Digital Micrograph™

A software package "SADP Tools" is developed as a complementary diffraction pattern analysis tool. The core program, called AutoSADP, is designed to facilitate automated measurements of d-spacing and interplaner angles from TEM selected area diffraction patterns (SADPs) of single crystals. The software uses iterative cross correlations to locate the forward scattered beam position and to find the coordinates of the diffraction spots. The newly developed algorithm is suitable for fully automated analysis and it works well with asymmetric diffraction patterns, off-zone axis patterns, patterns with streaks, and noisy patterns such as Fast Fourier transforms of high-resolution images. The AutoSADP tool runs as a macro for the Digital Micrograph program and can determine d-spacing values and interplanar angles based on the pixel ratio with an accuracy of better than about 2%.

Vacant-site octahedral tilings on SrTiO(3) (001), the (sqrt[13]×sqrt[13])R33.7° surface, and related structures

The structure of the SrTiO(3) (001) (sqrt[13]×sqrt[13])R33.7° surface reconstruction has been determined using transmission electron diffraction combined with direct methods and density functional theory. It has a TiO(2)-rich surface with a 2D tiling of edge or corner-sharing TiO(5)□ octahedra. Additionally, different arrangements of these octahedral units at the surface, dictated by local bond-valence sums, form 2D networks that can account for many ordered surface reconstructions as well as disordered glasslike structures consistent with the multitude of structures observed experimentally, and potentially other materials and interfaces.

Structural insights into P-glycoprotein (ABCB1) by small angle X-ray scattering and electron crystallography

P-glycoprotein (ABCB1) is an ATP-binding cassette protein that is associated with the acquisition of multi-drug resistance in cancer and the failure of chemotherapy in humans. Structural insights into this protein are described using a combination of small angle X-ray scattering data and cryo-electron crystallography data. We have compared the structures with bacterial homologues, and discuss the development of homology models for P-glycoprotein based on the bacterial Sav1866 structure.

A quantitative analysis of the cone-angle dependence in precession electron diffraction

Precession electron diffraction (PED) is a technique which is gaining increasing interest due to its ease of use and reduction of the dynamical scattering problem in electron diffraction. To further investigate the usefulness of this technique, we have performed a systematic study of the effect of precession angle on the mineral andalusite where the semiangle was varied from 6.5 to 32 mrad in five discrete steps. The purpose of this study was to determine the optimal conditions for the amelioration of kinematically forbidden reflections, and the measurement of valence charge density. We show that the intensities of kinematically forbidden reflections decay exponentially as the precession semiangle (varphi) is increased. We have also determined that charge density effects are best observed at moderately low angles (6.5-13 mrad) even though PED patterns become more kinematical in nature as the precession angle is increased further.

Rapid structure determination of a metal oxide from pseudo-kinematical electron diffraction data

The electron precession diffraction technique is employed to provide quasi-kinematical data for determination of atom positions in the (Ga,In)2SnO5m-phase. Precession data are compared with conventional diffraction data captured under identical conditions and show a distinct superiority because they exhibit kinematical characteristics in the structure-defining reflections. Precessed data are not usable within a kinematical interpretation in all cases, and a simple basis is presented for omission of errant reflections to improve adherence to kinematical behavior. A second approach is demonstrated where intensities are used with direct methods instead of amplitudes, enhancing the contrast between strong and weak beams. The unrefined atom positions recovered a priori via direct methods are consistent between the two approaches and fall on average within 4 picometers of positions in the previously refined structure.