SKM 2023 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

HL: Fachverband Halbleiterphysik

HL 14: 2D Materials III (joint session HL/CPP)

HL 14.8: Talk

Tuesday, March 28, 2023, 11:30–11:45, POT 81

Fully automated platform for 2D material flake detection using real-time machine learning techniques — •Jan-Lucas Uslu, Taoufiq Ouaj, Bernd Beschoten, Lutz Waldecker, and Christoph Stampfer — JARA-FIT and 2nd Institute of Physics A, RWTH Aachen University, Aachen, Germany

As of today, most of fundamental experimental 2D material research is based on mechanically exfoliated flakes, finding suitable flakes for the fabrication of van der Waal heterostructures is time-consuming and time-critical part requiring expert knowledge and manpower.

In order to mitigate this problem, we demonstrate a simple and robust real time-capable algorithm based on Gaussian mixture models, a machine learning technique, to allow for a fast automated search of exfoliated flakes of different 2D materials in a single run with an automated microscope setup to analyze batches of exfoliated material.

The algorithm solves the task of automatically detecting various flakes on Si++/SiO2 wafer dices, allows to index the location and segmentation of each flake and provides metrices such as size, thickness and shape.

The algorithm is evaluated on more than 500.000 images of different 2D materials including graphene and multilayer graphene, hexagonal boron nitride, transition metal dichalcogenides and 2D magnets.