Optimizing Quantum Gas Production by an Evolutionary Algorithm

TL;DR

This paper demonstrates how an evolutionary algorithm can optimize various steps in producing a Bose-Einstein condensate, significantly reducing production time and revealing parameter dependencies in complex quantum gas experiments.

Contribution

It introduces the application of differential evolution, an EA, to optimize each step of BEC production, providing a new approach for experimental enhancement and parameter analysis.

Findings

Improved loading rates and phase-space density of BEC

Reduced BEC production time significantly

Revealed parameter correlations for further optimization

Abstract

We report on the application of an evolutionary algorithm (EA) to enhance performance of an ultra-cold quantum gas experiment. The production of a $^{87}$Rubidium Bose-Einstein condensate (BEC) can be divided into fundamental cooling steps, specifically magneto optical trapping of cold atoms, loading of atoms to a far detuned crossed dipole trap and finally the process of evaporative cooling. The EA is applied separately for each of these steps with a particular definition for the feedback the so-called fitness. We discuss the principles of an EA and implement an enhancement called differential evolution. Analyzing the reasons for the EA to improve \eg, the atomic loading rates and increase the BEC phase-space density, yields an optimal parameter set for the BEC production and enables us to reduce the BEC production time significantly. Furthermore, we focus on how additional information about the experiment and optimization possibilities can be extracted and how the correlations revealed allow for further improvement. Our results illustrate that EAs are powerful optimization tools for complex experiments and exemplify that the application yields useful information on the dependence of these experiments on the optimized parameters.