Live User-Guided Low Dose Scanning Transmission Electron Microscopy Imaging

Electron beam-sensitive materials remain a permanent challenge in scanning/transmission electron microscopy (S/TEM), requiring a drastic reduction in electron dose to minimize sample damage while maximizing the information extracted from each transmitted electron. A variety of advanced techniques have been developed to both mitigate electron beam damage and reduce the required dose for imaging. These include digital electron counting to eliminate detector afterglow and Gaussian read-out noise, multi-frame acquisition schemes, as well as subsampling methods with image reconstruction [1-3]. However, all these approaches rely on a sufficiently good initial seed of data from the sample to enable accurate alignment and/or extrapolation. In most cases, data acquisition still requires a final manual alignment step, and identifying the optimal imaging conditions under extreme low-dose conditions remains particularly challenging. In this study, we present a visual solution to assist users in the final imaging adjustments under conditions where conventional imaging modes fail to provide instant feedback. The proposed method involves borrowing from techniques already established in the scanning electron microscopy field. Here, we use DigitalMicrograph scripting to implement a continuous live rolling average, enhancing the signal-to-noise ratio (SNR) in real-time to the operator. Prior to data acquisition, memory is pre-allocated to store multiple frames, enabling, for example, to collect two minutes of live scanning. Then, the data seen by the user is continuously inserted into the pre-allocated stack during all navigation, search, focus and fine-stigmation operations. Simultaneously, a weighted average of the stored images running backwards into the recent past is presented to the user. This approach improves the SNR while maintaining an acceptable response speed. The image displayed can be further filtered (e.g. bandpass filter) or represented via its Fourier transform to assist the operator and obtain real-time feedback to refine key alignment settings manually (primarily focus and astigmatism). We applied this methodology to a RbFe(MoO4)2 (RFMO) material exhibiting unconventional magnetoelectricity through its coupling with ferroaxiality. Characterizing RFMO at the atomic scale is particularly challenging due to its vulnerability to electron beam exposure. After reducing the electron dose to a level where nearly no visible features were discernible, we employed the live feedback process to optimize imaging conditions before acquiring a series of 512 x 512 STEM High-Angle Annular Dark Field (HAADF) images with a dwell time of 0.9 µs. Fig. 1a shows an example of a single frame, where the crystalline lattice is barely visible. The complete dataset, consisting of 70 images, was then aligned using SmartAlign [2], and the rigid-registered result is presented in Fig. 1b. Further improvement in the signal-to-noise ratio was achieved using the template matching module from the SmartAlign plug-in, as demonstrated in Fig 1c. The resulting atomic structure closely matches the expected crystalline arrangement simulated using the py4DSTEM package [4], as shown in Fig 1d. Using the live feedback routine, the RFMO crystalline structure along the [100]h zone axis is revealed, and results will be further compared with structural models [5]. Low dose STEM HAADF of the RFMO sample along the [100]h direction. a) One single frame of the dataset. b) Sum of the 70 frames after rigid-registration c) Motif reconstruction from b) using the template matching module from SmartAlign. d) STEM HAADF image simulation of the RFMO crystal structure along the [100]h .c) and d) have enlarged insets in the lower left corners.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.