SKM 2023 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
TT: Fachverband Tiefe Temperaturen
TT 24: Quantum Dots: Transport (joint session HL/TT)
TT 24.2: Talk
Tuesday, March 28, 2023, 09:45–10:00, POT 151
Predicting charge density maps in 2D nanostructures with machine learning techniques — •Amanda Teodora Preda1,2,3, Calin Andrei Pantis-Simut1,2,3, Nicolae Filipoiu2,3, Lucian Ion2, Andrei Manolescu4, and George Alexandru Nemnes1,2,3 — 1Research Institute of the University of Bucharest (ICUB), Sos. Panduri 90, Bucharest, Romania — 2University of Bucharest, Faculty of Physics, 077125 Magurele-Ilfov, Romania — 3Horia Hulubei National Institute for Physics and Nuclear Engineering, 077126 Magurele-Ilfov, Romania — 4Department of Engineering, Reykjavik University, Menntavegur 1, IS-102 Reykjavik, Iceland
Machine learning (ML) models have the potential to significantly improve and assist the design process of nanodevices that require precise control of the quantum states.
For 2D nanoelectronic structures, charge and spin densities are relevant observables and are also suited for ML techniques which employ image processing. The model systems that we considered are two dimensional quantum dots with multiple electrons and random confinement potentials. With convolutional neural networks, we built a ML model to predict whether a configuration displays singlet-triplet transitions in the ground state. For image translation problems, we used models based on conditional generative adversarial networks in order to predict the charge density distribution for arbitrary interacting systems taking as input either the non-interacting cases or just the shape of the confining potential.