A decade of atom-counting in STEM: From the first results toward reliable 3D atomic models from a single projection

A simple circularity-based approach for nanoparticle size histograms beyond the spherical approximation

Conventional Transmission Electron Microscopy (TEM) is widely used for routine characterization of the size and shape of an assembly of (nano)particles. While the most basic approach only uses the projected area of each particle to infer its size (the "circular equivalent diameter" corresponding to the so-called "spherical approximation"), other shape descriptors can be determined and used for more elaborate analyses. In this article we present a generic model of particles, considered to be made of a few individual grains, and show how the equivalent size (i.e. a particle volume information) can be reliably deduced using only two basic parameters: the projected area and the perimeter of a particle. We compare this simple model to the spherical and ellipsoidal approximations and discuss its benefits. Then, partial coalescence of grains in a particle is also considered and we show how a simple analytical approximation, based on the circularity parameter of each particle, can improve the experimental determination of a particle size histogram. The analysis of experimental observations on nanoparticles assemblies obtained by mass-selected cluster deposition is presented, to illustrate the efficiency of the proposed approach for the determination of particle size just from conventional TEM images. We show how the presence of multimers offers an excellent opportunity to validate our improved and simple procedure. In addition, since the circularity plays a central role in this approach, attention is attracted on the perimeter determination in a pixelated image.

Atomic-Scale Imaging of Polymers and Precision Molecular Weight Analysis

Polymer design requires fine control over syntheses and a thorough understanding of their macromolecular structure. Herein, near-atomic level imaging of polymers is achieved, enabling the precise determination of one of the most important macromolecular characteristics: molecular weight. By judiciously designing and synthesizing different linear metal(loid)-rich homopolymers, subnanoscale polymer imaging is achieved through annular dark field-scanning transmission electron microscopy (ADF-STEM), owing to the incorporation of high Z atoms in the side chain of the monomeric units. The molecular weight of these polymers can be precisely determined by detecting and counting their metal(loid) atoms upon ADF-STEM imaging, at sample concentrations as low as 10 μg·mL-1. Notably, a commonly used C, H, and O-containing polymer (i.e., poly(methyl acrylate)) that was thus far inaccessible at the atomic scale is derivatized to allow for subnano-level imaging, thus expanding the scope of our approach toward the atomic-level visualization of commodity polymers.

Precise Size Determination of Supported Catalyst Nanoparticles via Generative AI and Scanning Transmission Electron Microscopy

Transmission electron microscopy (TEM) plays a crucial role in heterogeneous catalysis for assessing the size distribution of supported metal nanoparticles. Typically, nanoparticle size is quantified by measuring the diameter under the assumption of spherical geometry, a simplification that limits the precision needed for advancing synthesis-structure-performance relationships. Currently, there is a lack of techniques that can reliably extract more meaningful information from atomically resolved TEM images, like nuclearity or geometry. Here, cycle-consistent generative adversarial networks (CycleGANs) are explored to bridge experimental and simulated images, directly linking experimental observations with information from their underlying atomic structure. Using the versatile Pt/CeO2 (Pt particles centered ≈2 nm) catalyst synthesized by impregnation, large datasets of experimental scanning transmission electron micrographs and physical image simulations are created to train a CycleGAN. A subsequent size-estimation network is developed to determine the nuclearity of imaged nanoparticles, providing plausible estimates for ≈70% of experimentally observed particles. This automatic approach enables precise size determination of supported nanoparticle-based catalysts overcoming crystal orientation limitations of conventional techniques, promising high accuracy with sufficient training data. Tools like this are envisioned to be of great use in designing and characterizing catalytic materials with improved atomic precision.

Obtaining 3D Atomic Reconstructions from Electron Microscopy Images Using a Bayesian Genetic Algorithm: Possibilities, Insights, and Limitations

The Bayesian genetic algorithm (BGA) is a powerful tool to reconstruct the 3D structure of mono-atomic single-crystalline metallic nanoparticles imaged using annular dark field scanning transmission electron microscopy. The number of atoms in a projected atomic column in the image is used as input to obtain an accurate and atomically precise reconstruction of the nanoparticle, taking prior knowledge and the finite precision of atom counting into account. However, as the number of parameters required to describe a nanoparticle with atomic detail rises quickly with the size of the studied particle, the computational costs of the BGA rise to prohibitively expensive levels. In this study, we investigate these computational costs and propose methods and control parameters for efficient application of the algorithm to nanoparticles of at least up to 10 nm in size.

Localization and segmentation of atomic columns in supported nanoparticles for fast scanning transmission electron microscopy

To accurately capture the dynamic behavior of small nanoparticles in scanning transmission electron microscopy, high-quality data and advanced data processing is needed. The fast scan rate required to observe structural dynamics inherently leads to very noisy data where machine learning tools are essential for unbiased analysis. In this study, we develop a workflow based on two U-Net architectures to automatically localize and classify atomic columns at particle-support interfaces. The model is trained on non-physical image simulations, achieves sub-pixel localization precision, high classification accuracy, and generalizes well to experimental data. We test our model on both in situ and ex situ experimental time series recorded at 5 frames per second of small Pt nanoparticles supported on CeO2(111). The processed movies show sub-second dynamics of the nanoparticles and reveal site-specific movement patterns of individual atomic columns.

Seeing the Supracolloidal Assemblies in 3D: Unraveling High-Resolution Structures Using Electron Tomography

Transmission electron microscopy (TEM) imaging has revolutionized modern materials science, nanotechnology, and structural biology. Its ability to provide information about materials' structure, composition, and properties at atomic-level resolution has enabled groundbreaking discoveries and the development of innovative materials with precision and accuracy. Electron tomography, single particle reconstruction, and microcrystal electron diffraction techniques have paved the way for the three-dimensional (3D) reconstruction of biological samples, synthetic materials, and hybrid nanostructures at near atomic-level resolution. TEM tomography using a series of two-dimensional (2D) projections has been used extensively in biological science, but in recent years it has become an important method in synthetic nanomaterials and soft matter research. TEM tomography offers unprecedented morphological details of 3D objects, internal structures, packing patterns, growth mechanisms, and self-assembly pathways of self-assembled colloidal systems. It complements other analytical tools, including small-angle X-ray scattering, and provides valuable data for computational simulations for predictive design and reverse engineering of nanomaterials with the desired structure and properties. In this perspective, I will discuss the importance of TEM tomography in the structural understanding and engineering of self-assembled nanostructures with specific emphasis on colloidal capsids, composite cages, biohybrid superlattices with complex geometries, polymer assemblies, and self-assembled protein-based superstructures.

The role of high-resolution transmission electron microscopy and aberration corrected scanning transmission electron microscopy in unraveling the structure-property relationships of Pt-based fuel cells electrocatalysts

Platinum-based fuel cell electrocatalysts are structured on a nano level in order to extend their active surface area and maximize the utilization of precious and scarce platinum. Their performance is dictated by the atomic arrangement of their surface layers atoms via structure-property relationships. Transmission electron microscopy (TEM) and scanning transmission electron microscopy (STEM) are the preferred methods for characterizing these catalysts, due to their capacity to achieve local atomic-level resolutions. Size, morphology, strain and local composition are just some of the properties of Pt-based nanostructures that can be obtained by (S)TEM. Furthermore, advanced methods of (S)TEM are able to provide insights into the quasi-in situ, in situ or even operando stability of these nanostructures. In this review, we present state-of-the-art applications of (S)TEM in the investigation and interpretation of structure-activity and structure-stability relationships.

Atom counting from a combination of two ADF STEM images

To understand the structure-property relationship of nanostructures, reliably quantifying parameters, such as the number of atoms along the projection direction, is important. Advanced statistical methodologies have made it possible to count the number of atoms for monotype crystalline nanoparticles from a single ADF STEM image. Recent developments enable one to simultaneously acquire multiple ADF STEM images. Here, we present an extended statistics-based method for atom counting from a combination of multiple statistically independent ADF STEM images reconstructed from non-overlapping annular detector collection regions which improves the accuracy and allows one to retrieve precise atom-counts, especially for images acquired with low electron doses and multiple element structures.