Abstract text (incl. figure legends and references)

The introduction pixelated detectors for scanning transmission electron microscopy (STEM) having frame rates of several thousand fps opened the exciting new field of 4DSTEM, since they allow to acquire a full diffraction pattern at each scan point. This facilitates to measure built-in electric fields using momentum-resolved STEM (MRSTEM) by evaluating the introduced shift of the diffraction pattern e.g. by tracking the direct beam or calculating the center-of-mass (COM) of the pattern"s intensity. In this way, fields with extensions in the atomic range or even longer-ranges up to micrometers, which are present in actual devices, like solar cells or batteries, can be addressed. As an example, recently, MRSTEM was used to determine the electric fields in GaAs-based pn-junctions quantitatively [1].

Real-life devices, however, generally do not consist of one material only but are composed of different ones. Accordingly, key properties, like the crystal structure, crystal orientation, the mean inner potential (MIP) or the strain state can change across the interfaces involved, complicating the determination of the electric fields.

In this contribution, we investigate the different contributions to the COM signal at interfaces with the help of systematic multi-slice simulations utilizing the STEMsalabim code [2]. These simulations are probably the only way to disentangle the different contributions of COM, since the input parameters like MIP and internal fields, e.g. due to built-in charges, can be varied independently. Moreover, optimum experimental parameters can be derived before the actual experiment. To this end, we utilize various well-defined model interfaces, i.e. GaP/Si, GaAs/GaInP, GaAs/AlAs, in which a notable difference in MIP is present and/or an intentional doping was applied to create an additional electric field . To verify our theoretical findings, complementary 4DSTEM datasets were acquired in an aberration corrected JEOL JEM 2200FS operating at 200 kV, equipped with a pnCCD detector and a NanoMEGAS Astar precession electron diffraction (PED) system.

It turns out that the choice of experimental parameters is indeed crucial to measure meaningful signals, e.g. low convergence angles of a few mrad proved to be beneficial. In addition, we find that dynamic diffraction leads to thickness-dependent oscillations in the COM signal, which do not reflect the internal fields anymore. However, PED can be applied to minimize the impact of dynamic diffraction at the cost of a reduction of spatial resolution. Under optimum conditions, MIP and fields can be detected via MRSTEM. This is shown in Fig. 1 with the example of a simulated GaAs/AlAs interface. The fields and potentials derived from the simulated COM data are in very good agreement to the input data.

In this contribution, we elucidate the different contributions to the COM signal at interfaces.

Fig1.: Simulation study of the GaAs/AlAs interface: The Input electric fields (dashed black lines) are shown alongside the fields derived from the simulated COM data (solid red lines), considering the difference in mean inner potential (MIP) only (a), an internal electric field only (b) and both MIP and field (c). The corresponding calculated potential profiles are shown in (d) – (f).

[1] A. Beyer et al., Nano Lett. 2021, 21 (5), 2018–2025.

[2] J. O. Oelerich et al., Ultramicroscopy 2017, 177.