The 27th International Conference on Atomic Physics

Dynamics following an interaction quench in the BEC-BCS crossover and machine-learning the phase diagram

18 Jul 2022, 17:00

1h 30m

Hart House (Hart House)

Poster presentation Degenerate gases, many-body physics, and quantum simulation Poster session

Description

With a specially designed coil [Rev. Sci. Instrum. 92, 093202 (2021)], we quench the interaction strength of an ultracold cloud of ${}^6$Li atoms instantaneously, i.e. faster than the Fermi time. In the short time dynamics following the quench we observe first indications of an oscillation in the condensate fraction reminiscent of the Higgs-mode.

The observation was facilitated by a deep neural network trained to determine the amount of superfluidity in the cloud from simple time-of-flight images. This quantity is hidden for conventional fitting techniques, especially under real experimental conditions.
We record a high-resolution phase diagram of the BEC-BCS crossover with the neural network and show that such networks are also able to detect the existance of a phase transition without any additional training input.