Regensburg 2022 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

MM: Fachverband Metall- und Materialphysik

MM 22: Data Driven Materials Science: Experimental Data Treatment and Machine Learning

MM 22.6: Vortrag

Mittwoch, 7. September 2022, 12:15–12:30, H46

Motif Extraction from Crystalline Images in Real Space — •Amel Shamseldeen Ali Alhassan and Benjamin Berkels — AICES Graduate School, RWTH Aachen University, Germany

Using modern transmission and scanning transmission electron microscopes ([S]TEM), atomic resolution images are readily available. In particular, the amount of data produced is so large that automatic analysis tools are needed.

During the last decade, automatic data analysis methods concerning different aspects of crystal analysis have been developed, for example, unsupervised primitive unit cell extraction and automated crystal distortion and defects detection. However, an automatic, dedicated motif extraction method is still called for by experimentalists. While previous works on motif extraction did good work in, for example, finding the plane symmetries and restoring smeared out features or finding positions in atomic columns, they were either not automated enough, not applicable to atomic scale images, or required special calibration.

Here, we propose and demonstrate a novel method for automatic direct motif extraction from crystalline images based on variational methods. Given an atomic resolution crystalline image, our method employs unit cell extraction to find the atomic structure then solves a minimization problem involving the unit cell projection operator to find the motif. The method was tested on various synthetic and experimental data sets. The results are a representation of the motif in form of an image, primitive unit cell vectors and a denoised reconstruction of the input image.