Regensburg 2022 – wissenschaftliches Programm

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

DY: Fachverband Dynamik und Statistische Physik

DY 44: Poster Session: Statistical Physics and Critical Phenomena

DY 44.20: Poster

Donnerstag, 8. September 2022, 15:00–18:00, P2

Detection of defects in soft quasicrystals with neural networks — •Ali Doener and Michael Schmiedeberg — Theoretische Physik I, Erlangen, DE

The aim of this work is to construct and employ a neural network for the detection of topological defects in dodecagonal quasicystalline patterns. Even though quasircrystals are aperiodic, they exhibit a longe-range order. Furthermore, in principle any discrete rotational symmetry can occur.

In this work, dodecagonal quasicrystalline patterns in two-dimensions with a built-in dislocation are generated and employed as input images of the neural network. The network then should figure out not only the position but also the type of the Burgers vector of the defect.

Our trained neural network is able to recognize the type of the Burgers vector very good. The position of the dislocation is recognized up to a mean deviation from the real position that is much smaller that the small length scale in the quasicrystals. In future, we want to train the network with patterns that contain multiple dislocations as well as phasonic excitations.