Abstract text (incl. figure legends and references)

Even though the advent of the aberration-corrected microscopes has enabled the possibility to perform STEM up to and below 50 pm resolution it is still not trivial to achieve atomic resolved images for each material. This is not because of the optics of the electron microscope but due to the electron beam damage which occurs when the highly energetic incoming electron interacts with the specimen. The common method to reduce the induced damage is by reducing the total dose, by making short exposure times, or reducing the electron beam current. Both result in lower signal-to-noise ratio images limiting the accuracy and precision of the quantified parameters extracted from such data. Experimental evidence has pointed out that not only the dose and dose rate influence the damage which occurs in a specimen during a STEM acquisition but that even the scan sequence and dwell time can modify the induced damage on a specimen [1,2]. One clear experimental evidence is found in the work of Velazco et al. [1] where a substantial decrease in damage is observed on a Linde Type A zeolite when decreasing the dwell time or changing the scan pattern while keeping the total dose and dose rate equal.

In the present work, the aim is to qualitatively model the damage observed using a minimal amount of physical parameters (diffusion constant and threshold). A diffusion-based model is chosen since the damage was experimentally found to depend on the scan sequence indicating a non-local component where the induced damage depends on the time at which the neighboring pixels are visited. By only taking the diffusion of a quantity to describe the damage there is no difference in the total induced damage when changing the scan sequence. Therefore, a threshold is introduced below which no damage is induced. Note that no assumption is made on which physical parameter is diffusing and why. Furthermore, the experimental observations observed in Velazco et al. [1] are used to estimate the two parameters of the model and verify the applicability of the model to describe the experiments. By developing such model, it is possible to predict how the scan sequence and dwell time can influence the total induced damage. In the future these predictions could be taken as guidelines to optimize experiments for different samples.

References:

[1] Velazco A. et al., (2022), Ultramicroscopy, 232, 113398.

[2] Jones L. et al. (2018), Microscopy, Issue 67, Pages i98-i113

[3] Jannis, D. et al. (2022), Ultramicroscopy, 240, 113568

[4] Acknowledgment: AB, AV, DJ and JV acknowledge financial support from the Flemish region through FWO grants G093417N and G042920N innovation program under grant agreement No 101017720, FET-Proactive EBEAM