Speaker
Description
Removing constraints of 4D STEM with a framework for event-driven acquisition and processing
Arno Annys1,2*, Hoelen L. Lalandec Robert1,2, Saleh Gholam1,2, Joke Hadermann1,2 and Jo Verbeeck1,2
1Electron Microscopy for Materials Science (EMAT), University of Antwerp, Groenenborgerlaan 171, 2020 Antwerp, Belgium
2NANOlight Center of Excellence, University of Antwerp, Groenenborgerlaan 171, 2020 Antwerp, Belgium
* Corresponding author: arno.annys@uantwerpen.be
Electron microscopy is a powerful and versatile tool for characterizing a wide range of samples in the fields of materials science and life sciences. In scanning transmission electron microscopy (STEM), a nanometer to picometer sized electron probe is raster scanned over a specimen, typically with microsecond dwell times. The most commonly used detection technique for transmitted electrons in STEM involves a single-pixel annular dark field detector, producing image contrast that scales with mass and thickness. However, this approach results in the loss of significant amounts of information, both from undetected electrons and from the limited information content per detected electron.
Recent advancements in detector technology, particularly the development of fast hybrid pixel array direct electron detectors, have revolutionized STEM by enabling the recording of the full scattering distribution at each probe position during a scan. This typically results in a 2D diffraction pattern for each point in the 2D scan grid, so that the technique is referred to as 4D STEM by the community. 4D STEM enables a wide range of information retrieval methods for various applications, including e.g. dose-efficient high-resolution imaging with ptychography and strain, orientation, and phase mapping in nano-beam electron diffraction.
The maximum frame rate of frame-based HPAD in STEM is generally still a factor 10-100 times slower than the megahertz regime usually desired. More importantly, because the number of electrons used in a single probe position is often much smaller than the number of detector pixels required, frame-based measuring becomes increasingly inefficient in terms of both data and computation as the detector speeds approach the desired rates. These limitations on speed and data size, among others, put stringent practical constraints on 4D STEM, making it still considered an off-line expert technique.
The sparse and continuous operation mode of event-driven HPADs, such as those based on Timepix3 and Timepix4, has been shown to mitigate some of the practical challenges in 4D STEM [1]. However, fully leveraging the benefits of event-driven detection requires redesigning the entire acquisition and processing pipeline, optimizing for sparse operation. To address this, we introduce a framework designed to bridge the gap between event-driven detection and the 4D STEM community, tackling key challenges in synchronization and processing efficiency. By redesigning current workflows to operate on individual electron events, we eliminate the need for expensive dense 4D arrays. We introduce a new algorithm that enables analytical electron ptychography to be performed directly on the raw event stream, achieving reconstruction rates that exceed maximum acquisition speeds on a standard consumer computer. This paves the way towards routinely driving experiments on the live feedback from 4D STEM measurements.
We compare the scaling laws of data size and computational requirements for event-driven and frame-based operation across various scenarios, demonstrating the benefits of event-driven workflows across a broad range of techniques. Eliminating detector speed limitations facilitates the investigation of beam-sensitive materials at extremely low doses. The advantages of high detector resolution without added data size are shown in nanobeam electron diffraction. Additionally, increasing scan resolution without penalty allows for large fields of view with sufficient sampling, which is especially important for techniques like ptychography. Finally, we illustrate how dividing a measurement into multiple scans without additional cost provides direct insights into sample dynamics, such as beam damage effects or responses to external stimuli.
References:
1. D. Jannis et al. Ultramicroscopy. 233, 113423 (2022)
2. This work received funding from the Horizon Europe framework program for research and innovation under grant agreement n. 101094299 (IMPRESS), the European Union’s Horizon 2020 research and innovation program under grant agreement n. 101017720 (FET-Proactive EBEAM) and an SBO FWO national project under grant agreement n. S000121N (AutomatED)
| Workshop topics | Applications |
|---|
