Preview

PDF, English Download (5MB) | Terms of use

Abstract

This thesis reports on a setup for a quasi two-dimensional Bose-Einstein condensate of 39K, which is an isotope well suited for interaction tuning due to a broad magnetic Feshbach resonance, and presents the application of the setup to a specific type of physical computing.

The first part gives an overview of the experimental components to prepare a quasi two-dimensional condensate with a configurable shape. Particular focus is put on the control of the magnetic field for the adjustment of atomic interactions and the configurable potential, which is realized with a digital micromirror device. The imaging setup is presented in detail and a strategy for absorption imaging at high magnetic field is elaborated. This strategy is necessary to properly exploit the magnetic Feshbach resonance. It relies on a scheme for an approximately closed four level optical cycle.

The second part introduces an approach for the implementation of a shallow artificial neural network with a physical system. Subsequently, a specific implementation that utilizes a quasi one-dimensional Bose-Einstein condensate is presented. Regression and interpolation of a non-inear function are performed successfully as a proof-of-concept, and the results are compared for different experimental parameters.