Online Thickness Determination with Position Averaged Convergent Beam Electron Diffraction using Convolutional Neural Networks

Accurate thickness determinations of specimens are important for many STEM-applications. A convenient method for crystalline samples is the position averaged convergent-beam electron diffraction (PACBED) method [1]. The thickness is determined by finding the best match of the recorded PACBED pattern with a series of simulated PACBED, calculated at different crystal thickness. This is a time-consuming process and is not practical during a microscope session. Xu and LeBeau showed the automatization of such PACBED analysis by convolutional neural networks (CNNs) [2]. Although, this enables fast analysis, simulating a large training dataset of PACBEDs and training the CNNs has high computational costs and these CNNs are only valid for the simulated system. Since many scientists are working with the same materials, the scientific community would benefit of a shared database of trained CNNs to predict the thickness of the specimen without having to simulate PACBEDs and train CNNs by themselves.
We build prototype of a server-based PACBED thickness determination by CNNs, which is integrated in GMS to enable an easy and fast PACBED analysis during a microscope session. 

This project recieved fundings from the European Union's Horizon 2020 research and innovation programme under grant agreement No 823717 – ESTEEM3 and the Austrian Science Fund (FWF) I4309.

[1] J. M. LeBeau, S. D. Findlay, L. J. Allen, and S. Stemmer, Ultramicroscopy, vol. 110, no. 2 (2010)
[2] W. Xu and J. M. LeBeau, Ultramicroscopy, vol. 188 (2018)