Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

A: Fachverband Atomphysik

A 25: Poster IV

A 25.1: Poster

Mittwoch, 13. März 2024, 17:00–19:00, Tent A

Analysing Single Particle Trajectories Of Ultracold Atoms With Artificial Intelligence — •Marco Mohler, Silvia Hiebel, Dennis Wagner, Sabrina Burgardt, Julian Feß, Marius Kloft und Artur Widera — University of Kaiserslautern-Landau, Kaiserslautern, Germany

Artificial Intelligence can be a helpful tool in analysing large datasets. In the presented work, we analyze the diffusion of Cs atoms, which are trapped in a far-detuned optical dipole trap and driven by an optical molasses. As the atoms absorb and reemit photons from the molasses laser beams they receive small momentum kicks in a random direction. This fluctuating force together with Doppler damping due to the laser beams detuning results in diffusive behaviour and is similar to Brownian motion. A small imbalance in the power of the counterpropagating molasses beams results in a small drift away from the stronger beam. This is to be avoided as it is a disturbance to experiments. Here, we present a neural network trained to learn the underlying force field behind the diffusive cesium trajectory which originates from the details of the laser setup. Applying the network to experimental data might reduce everyday readjustment time by telling which parameters to adjust to negate the drift from a reduced number of recorded trajectories. Initial training is carried out on simulated data because producing this data requires less ressources. Therefore trajectories are calculated for different laser imbalances and presented to the neural network so that it learns how the imbalance affects the atoms` movement. Currently the simulation is being tested before the neural network is set up.

Keywords: artificial intelligence; diffusion; trajectories; imbalance; neural network