Applications of maximum likelihood estimations for analyzing photon counts in few atom experiments

Abstract: We present a method for determining the atom number distribution of few atoms in a tight optical tweezer from their fluorescence distributions. In the tight tweezer regime, the detection light causes rapid atom loss due to light-assisted collisions. This in turn leads to non-Poissonian and overlapping fluorescence distributions for different initial atom numbers, and commonly used threshold techniques fail. We use maximum likelihood estimation algorithms to fit model distributions that account for the atom loss. This gives accurate atom number distributions for relatively few experimental runs (about 600 is sufficient) to sample a photon number distribution. We show that the method can be extended to situations when the photon number distributions for known initial atom numbers cannot be modeled, at the cost of requiring a higher number of experimental runs.