Automatic Vortex Cluster Detection in Bose-Einstein Condensates Through Machine Learner Image Processing

Project Level: Honours/Masters

Bose-Einstein condensates (BECs) are an example of a superfluid system, described by a macroscopic wavefunction. The bulk of the superfluid is irrotational (zero circulation) and only poles in the wavefunction can support quantised amounts of circulation. These poles are referred to as quantised vortices and are of interest when studying quantum turbulence in BEC superfluids.

One difficult task when analysing these vortices is defining clustering in an objective manner. Cluster sorting can be considered a subset of the tasks appropriate for machine learning algorithms to solve. In addition, the experimental results available to us are images.

Therefore, image processing is also required which is another task suitable for machine learners. The aim of the project is to develop a complete machine learning algorithm to detect and graphically map quantised vortex clusters from single-shot images taken in the experiment. To accompany this, the algorithm will need to be error-tested using simulations of the system. Some research into the most appropriate algorithm will also be required.