Developments in scanning transmission electron microscopy (STEM) have opened up new possibilities for time-resolved imaging at the atomic scale. Recent examples include a study of the diffusion of dopant atoms in semiconductors  and, using environmental STEM, in situ studies of catalytic reactions . Rapid imaging of single atom dynamics brings with it a new set of challenges. High frame rates and long total acquisition times mean novel methods are needed for handling and processing “big data” sets. Further, the need for short exposure times leads to severe problems with noise, but by exploiting the spatial and temporal correlations between frames, it is possible to considerably improve the signal-to-noise ratio using a method known as singular value thresholding . Crucially, by employing robust procedures to automatically estimate the noise and motion characteristics, it is possible to optimize the process with little user input (Figure 1a,b). The identity and positions of individual atoms in the denoised data can then be determined using a newly-developed intensity-based classification algorithm. Building on the theme of automation, the classifier can be trained using simulated STEM images to robustly process long image sequences, where manual identification would be prohibitive.
As an example, we have applied these methods to investigate the diffusive behaviour of copper atoms on the (110) surface of silicon. The noise removal and atom identification steps are used along with particle tracking software  to extract a set of atomic trajectories from a series of annular dark-field STEM images (Figure 1c-e). The form of these trajectories can be related to the underlying silicon substrate, as in Figure 1f, which suggests the existence of preferred pinning sites for copper atoms. The interaction between adatoms and the substrate can be explored with unprecedented spatio-temporal resolution using rapid imaging, and interpreted by modelling with density functional theory (DFT) calculations (Figure 1g). This highlights the potential for combining time-resolved STEM with theory, forming a powerful approach to investigating and understanding the dynamic behaviour of materials at the atomic scale.
 Ishikawa R, et al. (2014). Phys. Rev. Lett. 113, 155501.
 Gai P, et al. (2014). Chemical Physics Letters. 592, 355-359.
 Furnival T, Leary R, Midgley PA. (2016). Manuscript submitted.
 Chenouard N, et al., (2013). IEEE Trans. Pattern Anal. Mach. Intell., 35, 2736-2750.
The research leading to these results has received funding from the European Research Council under the European Union’s Seventh Framework Programme (FP7/2007-2013)/ERC grant agreement 291522-3DIMAGE.
To cite this abstract:Tom Furnival, Eric Schmidt, Rowan Leary, Daniel Knez, Ferdinand Hofer, Paul D Bristowe, Paul A Midgley; STEM imaging of atom dynamics: novel methods for accurate particle tracking. The 16th European Microscopy Congress, Lyon, France. https://emc-proceedings.com/abstract/stem-imaging-of-atom-dynamics-novel-methods-for-accurate-particle-tracking/. Accessed: October 31, 2020
EMC Abstracts - https://emc-proceedings.com/abstract/stem-imaging-of-atom-dynamics-novel-methods-for-accurate-particle-tracking/