Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
MI: Fachverband Mikrosonden
MI 8: Scanning Probe Microscopy (SPM)
MI 8.1: Vortrag
Donnerstag, 23. März 2017, 10:00–10:15, MER 02
Machine Learning in Spectroscopic Scanning Probe Microscopy: a Magnetoelectric Composite Case Study — •Harsh Trivedi1, Vladimir V. Shvartsman1, Doru C. Lupascu1, and Robert C. Pullar2 — 1Institute for Materials Science and Center for Nanointegration Duisburg-Essen (CENIDE), University of Duisburg-Essen, 45141, Essen, Germany — 2CICECO, University of Aveiro, 3810-193, Aveiro, Portugal
Recent years have seen a tremendous growth in the spectroscopic Scanning Probe Microscopy (SPM) modes, where spectra against various principle variables like voltage (Piezoresponse/Current), distance (force), frequency (amplitude) etc. are collected over a 2-dimensional grid. In certain cases however, the variable in concern is time dependent, and a point by point variation is not ideal. In such cases a convenient alternative is the sequential SPM, where a sequence of SPM images is generated as a function of the principle variable. However, such a method suffers from the absence of a concrete correlation between the consecutive images, and also the nature of each individual spectra differs with the spatial location. In the absence of a universal physical model that fits the data, a proper application of unsupervised machine-learning to such problems could lead to fast extraction of qualitative components of the spectra, which can conveniently be used to device the necessary physical models. Presented in this work is a case study of application of simple unsupervised machine-learning to variable magnetic field Piezoresponse Force Microscopy (PFM) of BaTiO3 based magnetoelectric composites.