Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
HL: Fachverband Halbleiterphysik
HL 23: Quantum Dots and Wires: Transport (joint session HL/TT)
HL 23.3: Vortrag
Dienstag, 18. März 2025, 11:45–12:00, H13
Fast Machine-Learning assisted characterisation of current quantisation — •Wang Ngai Wong1, Yannic Rath1, Nikolaos Schoinas1, Shota Norimoto1, Masaya Kataoka1, Alessandro Rossi1,2, and Ivan Runnger1,3 — 1National Physical Laboratory, Teddington, TW11 0LW, UK — 2Department of Physics, SUPA, University of Strathclyde, Glasgow G4 0NG, UK — 3Department of Computer Science, Royal Holloway, University of London, Egham, TW20 0EX, UK
Characterisation of single-electron pumps (SEPs) has long been bottlenecked by the process of fine-tuning measurement parameters to study their novel properties. This limits potential experimental parameters to those that can remain static throughout the fine-tuning process. We demonstrate a novel method assisted by machine learning which has led to an eightfold speedup in the measurement process (see Appl. Phys. Lett. 125, 124001 (2024)), and in so doing opens the door to further characterisation experiments which are impossible using conventional methods. Our method is based around an active learning cycle to navigate the information landscape of the gate voltage parameter space, while also significantly reducing the number of measurement points required. This is paired with a post-processing approach which allows us to accurately predict and characterise the small operational regimes significantly more efficiently than conventional sweeps across the parameter space. We exploit the framework to characterise the behaviour of multiplexed GaAs multi-pump devices across a range of magnetic fields.
Keywords: Machine-Learning; Single-Electron-Pumps; Quantum-Metrology; Quantum-Dots