Die DPG-Frühjahrstagung in Dresden musste abgesagt werden! Lesen Sie mehr ...
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
MM: Fachverband Metall- und Materialphysik
MM 30: Poster Session II
MM 30.36: Poster
Dienstag, 17. März 2020, 18:15–20:00, P4
Application of Machine Learning Interatomic Potentials to Carbon Nanostructures — •Tom Rothe1, Erik Lorenz1,2, Gustav Johansson3, Fabian Teichert1, Daniel Hedman3, Andreas Larsson3, and Jörg Schuster1,2 — 1Chemnitz University of Technology, Chemnitz, Germany — 2Fraunhofer Institute for Electronic Nano Systems (ENAS), Chemnitz, Germany — 3Luleå University of Technology, Luleå, Sweden
Machine Learning Interatomic Potentials (ML-IAPs) are a new class of non-empirical IAPs for atomistic simulations that are created using Machine Learning methods. Promising near quantum mechanical accuracy while being orders of magnitudes faster than first principle methods, they are the new "hot topic" in material simulation research.
This work investigates the state of the research in the field of ML-IAPs for simulation of carbon nanostructures (CNS). Publicly available ML-IAPs are used for simulation of defect induced deformation of carbon nanotubes. Comparing the results with previously published density-functional tight-binding results and our own empirical IAP geometry optimizations show that ML-IAPs can already be used for simulations of CNS. They are indeed faster than and nearly as accurate as first-principle methods.
We also present results for a new Neural Network-based ML-IAP trained on graphene, haeckelite, carbon nanotube, and fullerene structures, with and without defects.