Regensburg 2019 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
TT: Fachverband Tiefe Temperaturen
TT 27: Focus Session: Designer Quantum Systems I (joint session O/TT)
TT 27.6: Vortrag
Dienstag, 2. April 2019, 12:15–12:30, H15
Machine learning the 3D shape of non-planar molecules from AFM images — Prokop Hapala1, Fedor Utirev1, Niko Oinonen1, Ondřej Krejčí1, Filippo Federici Canova1, Benjamin Alldritt1, Juho Kannala2, Peter Liljeroth1, and •Adam Foster1 — 1Department of Applied Physics, Aalto University — 2Department of Computer Science, Aalto University
In recent decade Atomic Force Microscopy with tip functionalized by carbon monoxide (CO) provided unique tool to experimentally image sub-molecular details of individual organic molecules [1]. Yet up to now most experiments are limited to flat aromatic molecules, due to difficulties with interpretation of highly distorted images originating from non-planer molecules due to mechanical relaxation of tip or sample. These problems can be partially overcome using a simple mechanical model [2] which can reproduce those distortions, therefore simulate AFM images for given molecular structure. Testing many possible candidate structures is, however, laborious. Instead we attempt to develop automatic tool to conduct inverse task - i.e. to recover molecular structure from given set of AFM images. Preliminary results suggests that convolutional neural network (CNN) [3] trained on simulated AFM images can learn this inverse mapping rather easily. Yet application of the method on real experimental data, and identification of atomic species remains to be a challenge. [1] Gross, L., et al., Science, 325(5944), 1110-1114 (2009). [2] Lecun, Y., et al., Proceedings of the IEEE, 86(11), 2278-2324 (1998). [3] Hapala, et al. PRB, 90(8), 085421 (2014).