Abstract text (incl. figure legends and references)

Advancements in the semiconductor industry allow for ever smaller miniaturization. The simulation of scanning transmission electron microscopy (STEM) images can assist in understanding measurements necessary for the characterization of those nanoscale structures.

In multislice simulations of STEM images the total potential of the specimen commonly gets constructed by a linear superposition of atomic potentials, while the atomic potentials themselves are related to scattering amplitudes in the first Born approximation. This approach is called independent atom model, since bonding effects are neglected [1]. On the other hand, the muffin-tin approximation is a common method to account for the differences in the electron distribution in solids [2], in which the atomic potentials are assumed to be radially symmetric inside a sphere around the nucleus and constant outside of it.

In this work we compare the results of differently derived potentials to gain a better understanding of the impact of those potentials on multislice simulations. We compare independent atom model potentials using a common parametrization [1] with atomic potentials derived from charge densities limited by muffin-tin radii, which are based on the ELSEPA software package [2]. We further compare those results with the full density functional theory (DFT) calculations of Susi et al. [3].

The potentials show a significant deviation between free atoms and those limited by the muffin-tin spheres as well as the DFT results. Comparison with experimental data will reveal the appropriateness of models.

[1] E. J. Kirkland, "Advanced computing in electron microscopy" (Springer, 2020).
[2] F. Salvat et al., Computer Physics Communications 165, 157-190 (2005).
[3] T. Susi et al., Ultramicroscopy 197, 16-22 (2019).