Regensburg 2019 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
DY: Fachverband Dynamik und Statistische Physik
DY 55: Poster: Noneq. Stat. Phys., Stat. Bio. Phys., Brownian
DY 55.10: Poster
Thursday, April 4, 2019, 15:00–18:00, Poster B2
Unravelling the energetics of stochastic surface growth with artificial neural networks — •Thomas Martynec1, Stefan Kowarik2, and Sabine H.L. Klapp1 — 1Institut fuer Theoretische Physik, Technische Universitaet Berlin, Hardenbergstr. 36, 10623 Berlin, Germany — 2Institut fuer Physik, Humboldt-Universitaet zu Berlin, Newtonstrasse 15, 12489 Berlin, Germany
Stochastic surface growth by means of molecular beam epitaxy (MBE) is one of the most widely used techniques to fabricate thin film devices for various technological applications. It involves a competition between the adsorption of particles and various diffusion processes. Initially, particles like atoms, colloids or organic molecules are adsorbed on an ideally flat and defect free discrete substrate at rate F. This is followed by thermally activated Arrhenius-type diffusion processes to neighboring lattice sites. Once clusters are formed, nucleation on top of these clusters sets in. Depending on the strength Ees of the Ehrlich-Schwoebel barrier that reduces the interlayer transport rate of particles, surface growth proceeds either in a smooth or rough fashion. We demonstrate that an artificial neural network can precisely determine the value of the interlayer transport barrier from images of the growing islands with a 'wedding cake' morphology.
[1] C. M. Bishop, Rev. Sci. Instr. 65 1803 (1994)
[2] T.Martynec and S.H.L. Klapp Phys. Rev. E 98, 042801 (2018)