Parts | Days | Selection | Search | Updates | Downloads | Help
Q: Fachverband Quantenoptik und Photonik
Q 55: Poster VI
Q 55.2: Poster
Thursday, March 14, 2024, 17:00–19:00, Tent B
Machine Learning techniques in Quantum Gas Transport Experiments — •Gabriel Müller1, Victor J. Martínez-Lahuerta1, Philipp-Immanuel Schneider2,3, Ivan Sekulic2,3, and Naceur Gaaloul1 — 1Leibniz University Hannover, Germany — 2JCMwave GmbH, Berlin, Germany — 3Zuse Institute Berlin, Germany
Precision atom interferometry (AI) requires an accurate quantum state engineering of the atomic ensembles at the input port. With Bose-Einstein Condensates (BECs), quick and robust transports have been experimentally realised using shortcut to adiabaticity (STA) protocols [N. Gaaloul et al., Nature communications 13(1), 7889 (2022)]. These STA protocols, however, as well as alternative approaches featuring Optimal Control Theory (OCT) [S. Amri et al. Scientific Reports 9(1), 5346 (2019)], are either limited by approximations to avoid expensive computations or by a limited number of control parameters.
To address these limitations, we propose a novel approach that utilises Bayesian optimisation with Gaussian processes as machine learning surrogates. We evaluate its level of control in comparison to STA and OCT methods and later extend the application to reduce the amount of approximations and open up more degrees of freedom.
Once these methods are verified, one could consider dual-species transport and improve its robustness by taking into account experimental imperfections on ground and in microgravity.
Acknowledgements: Funded by the German Space Agency (DLR) with funds under Grant No. 50WM2253A/B (AI-quadrat).
Keywords: Bose Einstein Condensate; Quantum engineering; STA transport; Machine learning; Atom Interferometry