We show that three-body loss of trapped atoms leads to sub-Poissonian atom-number fluctuations. We prepare hundreds of dense ultracold ensembles in an array of magnetic microtraps which undergo rapid three-body decay. The shot-to-shot fluctuations of the number of atoms per trap are sub-Poissonian, for ensembles comprising 50-300 atoms. The measured relative variance or Fano factor F=0.53+/-0.22 agrees very well with the prediction by an analytic theory (F=3/5) and numerical calculations. These results will facilitate studies of quantum information science with mesoscopic ensembles.

We report on the storage and manipulation of hundreds of mesoscopic ensembles of ultracold 87Rb atoms in a vast two-dimensional array of magnetic microtraps, defined lithographically in a permanently magnetized film. The atom numbers typically range from tens to hundreds of atoms per site. The traps are optically resolved using absorption imaging and individually addressed using a focused probe laser. We shift the entire array, without heating, along the surface by rotating an external bias field. We evaporatively cool the atoms to the critical temperature for quantum degeneracy. At the lowest temperatures, density dependent loss allows small and well defined numbers of atoms to be prepared in each microtrap. This microtrap array is a promising novel platform for quantum information processing, where hyperfine ground states act as qubit states, and Rydberg excitation may orchestrate interaction between neighbouring sites.

We demonstrate improved detection of small trapped atomic ensembles through advanced postprocessing and optimal analysis of absorption images. A fringe-removal algorithm reduces imaging noise to the fundamental photon-shot-noise level and proves beneficial even in the absence of fringes. A maximum-likelihood estimator is then derived for optimal atom-number estimation in well-localized ensembles and is applied to real experimental data to measure the population differences and intrinsic atom shot noise between spatially separated ensembles each comprising between 10 and 2000 atoms. The combined techniques improve our signal-to-noise ratio by a factor of 3, to a minimum resolvable population difference of 17 atoms, close to our ultimate detection limit.