Johannes Haust (Marburg / DE), Akshaiy Jerath (Marburg / DE), Oliver Maßmeyer (Marburg / DE), Jürgen Belz (Marburg / DE), Andreas Beyer (Marburg / DE), Kerstin Volz (Marburg / DE)
Abstract text (incl. figure legends and references)
Introduction:
Phase retrieval in scanning transmission electron microscopy (STEM) by different ptychography algorithms gained more and more attention. It allows sub-angström resolution1 in 2D materials, is suitable for light elements and dose efficient2. Furthermore, by combining ptychography with the multislice approach its application can also be extended to thicker samples3.
Objectives:
The extended ptychographical iterative engine (ePIE) was already applied successfully for imaging 2D materials1. In this work we aim at utilizing an improved algorithm, the regularized PIE (rPIE)4,5, for the phase retrieval of 2D structures as well as thicker samples e.g. given for battery materials.
Materials and Methods:
rPIE solves the phase problem by taking the recorded diffraction patterns one by one and by iteratively revising the probe and the object4,5. It has an improved update function, which allows more control over the update rate and gives an enhanced robustness and stability in comparison to ePIE4,5.
In this work we benchmark the rPIE approach by simulated data in a first step and later on with experimental data. For STEM simulations our in-house developed multislice algorithm STEMsalabim6 is used. The experimental images are acquired with an aberration-corrected JEOL JEM-2200FS STEM equipped with a fast pixelated direct electron detector (DED) by PNDetectors.
Results:
In a first benchmark test rPIE is already applied to a simulated dataset of tungsten disulfide (WS2). The calculated phase as well as the amplitude image, as shown in figure 1, exhibit a clearly enhanced resolution in comparison to the annular dark field (ADF) image. With a subsequent simulation study, the optimal instrumental parameters like scan density, cutoff angle, and angular resolution will be determined, which will allow later on efficient acquisition of experimental data suitable for phase reconstruction. Also, the probability of extending rPIE with the multislice approach will be investigated.
Conclusion:
rPIE could be successfully applied on a simulated WS2 dataset. This opens the path for the first experimental datasets and as well as the extension of the rPIE algorithm.
| Figure 1. a) ADF image of a simulated WS2 dataset with 80 kV acceleration voltage and 21 mrad semi-convergence angle. b) and c) Object amplitude and phase obtained with rPIE after 10 iterations. |
References:
[1] Jiang, Y. et al. Nature 559, 343–349 (2018).
[2] Song, W. et al. Joule 6, 1049–1065 (2022).
[3] Chen, Z. et al. Science 372, 826–831 (2021).
[4] Maiden, A. et al Optica 4, 736 (2017).
[5] Springer Handbook of Microscopy, 819-904 (2019).
[6] Oelerich, J. O. et al. Ultramicroscopy 177, 91–96 (2017).