Real‐Time Reconstruction of Arbitrary Slices for Quantitative and In Situ 3D Characterization of Nanoparticles

Abstract

strongly connected to their 3D structure. [1][2][3][4][5][6][7][8][9][10] Heterostructures, in which several compounds are combined within a single nano-object, provide even more flexibility to tune their final properties. For example, bimetallic nanoparticles can display superior properties compared to their monometallic counterparts. [11][12][13] To understand the connection between structure/ composition and properties, nanoparticles are often investigated by transmission electron microscopy (TEM). Although TEM has become an indispensable tool for studying nanomaterials, it remains difficult to perform a 3D characterization. Indeed, conventional TEM provides 2D projection images of 3D objects, therefore missing a wealth of information. Electron tomography was developed to overcome this issue. [14][15][16][17] In 2003, Midgley et al. combined high angle annular dark field scanning TEM (HAADF-STEM) with tomography, [18,19] which has since been successfully applied to investigate a broad variety of nanostructures. [20][21][22][23][24] During a typical electron tomography experiment, a series of 2D projection images are collected along various tilt angles, to cover an angular range that is as large as possible. After alignment of the tilt series, they serve as the input to a mathematical algorithm that reconstructs the 3D structure of the object. Although the acquisition of a tilt series can be automated, it can take (many) hours to obtain all images, depending on the complexity of the experiment. In addition, both the alignment and the reconstruction of the acquired projection images are carried out through offline post-processing procedures, performed at a dedicated workstation. These steps are computationally demanding, leading to a total data processing time of at least 1 h. To dramatically accelerate the acquisition of tilt series, socalled "fast tomography" was recently introduced in both TEM and HAADF-STEM modes. [25][26][27] The methodology is based on continuously tilting the holder and simultaneously acquiring projection images, ideally while focusing and tracking the particle at the same time.Fast HAADF-STEM tomography enables a new range of experiments, during which the dynamic behavior of nanoparticles can be probed in 3D. For example, recently this technique was combined with in situ heating to investigate the thermal stability of Au and Au/Pd nanoparticles. [27,28] These experiments are at the state of the art with respect to acquisition time, and we were able to record a full HAADF-STEM tilt series within about 5 min. However, since the alignment A detailed 3D investigation of nanoparticles at a local scale is of great importance to connect their structure and composition to their properties. Electron tomography has therefore become an important tool for the 3D characterization of nanomaterials. 3D investigations typically comprise multiple steps, including acquisition, reconstruction, and analysis/quantification. Usually, the latter two steps are performed offline, at a dedicated workstation. This...

Ultra-high heating rate effects on the sintering of ceramic nanoparticles: an in situ TEM study

“…Due to the limited projected view of TEM imaging, there could be overlapping of particles and removal of porosity in the direction of the electron beam which could not be observed. This makes the quantification of morphological changes challenging and will require either 3D tomography technique to be implemented [41], or analysis through annular dark field image and EELS mass-thickness maps [27].…”

Fast Electron Tomography for Nanomaterials

“…Such on-the-fly reconstruction has just been realized allowing for real-time 3D feedback and onthe-fly quantification, paving the path for high-throughput and routine NP 3D characterization. 71 The combined effort in fast tomography acquisition and advanced reconstruction algorithms will most definitely lead to new scientific breakthroughs in material science. The main areas that benefit from that development will be beam-sensitive materials, such as perovskites and zeolites, as well as in situ and operando microscopy.…”

Noise2Filter: fast, self-supervised learning and real-time reconstruction for 3D Computed Tomography

“…These slices are interactive and can be manipulated by the technician of the tomographic experiment. This technique for real-time visualization has been successfully applied in practice to acquisitions in micro-CT systems [5], synchrotron tomography [3], and electron tomography [20].…”

“…In the presence of noise, determining the correct center of rotation for a dataset can be difficult and is often performed after acquiring the measured projection data. Using the tools developed in [20], the center of rotation can be adapted interactively in real-time. In Figure 9 we show Noise2Filter and FBP reconstructions with shifted centers of rotation at the first time step.…”

“…The information gained from this quasi-3D visualization can be used to directly steer the tomographic experiment, for instance, by adjusting an external parameter -such as temperature -in response to changes in the interior of the object. In addition, the object can be re-positioned, or other acquisition parameters can be adjusted to facilitate the best possible reconstruction [20].…”