Page 243 - Read Online

Page 8 of 17 Cabral et al. Microstructures 2023;3:2023040 https://dx.doi.org/10.20517/microstructures.2023.39

characterization and analysis. Although in-depth details of these techniques are beyond the scope of this
perspective, comprehensive reviews are available that outline ongoing and future research directions of in-
situ S/TEM [63-65] .

Analysis tools for electron microscopy data
With recent advancements in S/TEM instrumentation and technology, ease of use has increased
dramatically. Many modern aberration-corrected microscopes are now equipped with software allowing for
automatic tuning, greatly speeding up the process and ease of use for users. The coupled with improved
microscope stability allows users to align the microscope and tune the corrector within a matter of minutes.
Furthermore, many electron microscopes in shared user facilities are housed in carefully designed rooms
that mitigate instabilities resulting from noise, temperature variations, mechanical vibrations, acoustical
sources, and magnetic sources . With all these advancements, high-quality data collection can be
[66]
performed much more efficiently, allowing for more time to perform post processing. In the following
sections, we will detail various software tools that allow the correction of distortion in electron microscopy
data and simplified post processing.

Correcting drift and distortion in STEM images
Despite significant improvements in the spatial resolution and stability of electron microscopes, sample drift
and scanning distortion remain challenges for achieving accuracy and precision in STEM data. These issues
arise from both intrinsic scanning distortion caused by the movement of the electron probe across the
[67]
sample and external sources of drift . As a result, captured atomic resolution images may exhibit a
combination of expansion, compression, and shearing, which hampers the accurate measurement of
structural features such as lattice distances, atom column intensities, and polarization. Even the most stable
microscopes can experience small quantities of drift over long acquisition times. To address these
distortions, various techniques have been developed for correction in electron microscopy and other high
spatial resolution techniques, such as scanning probe microscopy (SPM). One approach involves utilizing
prior knowledge of the atomic features to correct distortions . However, this information may not be
[68]
available when studying novel materials. Another method involves using a known standard sample (e.g., Si
or SrTiO ) to measure drift and distortion, which can then be applied to the samples of interest . In the
[69]
3
case of beam-sensitive samples, a non-rigid image registration method can be employed using a series of
images acquired at a low electron dose. This approach is particularly useful for noisy data, as it allows
alignment of similar features from multiple images to a single image with a sufficient signal-to-noise
ratio [70-72] . These techniques offer effective ways to address sample drift and scanning distortion, enabling
more accurate characterization of materials at the atomic scale.

One of the most frequently used techniques to correct sample drift and distortion is capturing a series of
images and recombining them during post-processing. Many electron microscope user interfaces have tools
built in for this approach, such as drift-corrected frame integration (DCFI) on Thermo Fisher Scientific
instruments , or they can be installed as plugins, such as those available for Digital Micrograph. However,
[73]
although these tools are easy to use, they may not correct for non-linear drift and scan distortions. In such
cases, other post-processing tools, including revolving STEM (RevSTEM), can be employed . RevSTEM
[74]
captures a series of images with a short time/pixel dwell time and a 90° scan rotation between subsequent
frames. By rotating the scanning system between successive frames, sample drift and distortion can be
[74]
quantified and corrected through post-processing . Scanning distortion can also be corrected by
quantifying it on a standard sample (such as Si or SrTiO ) and mathematically applying it to the other
3
images using an affine transformation. Another procedure for correcting non-linear drift distortion involves
using orthogonal image pairs (90° rotation between images) to align the images by fitting contrast variations
in the slow scan direction on a line-by-line basis . Although only two images are necessary for this process,
[67]