Cosmic ray antiproton/electron discrimination capability of the CAPRICE silicon-tungsten calorimeter using neural networks

A data analysis based on an artificial neural network classifier is proposed to identify cosmic ray antiprotons detected with the CAPRICE silicon-tungsten imaging calorimeter against electron background in the energy range 1.2–4.0 GeV. A set of new physical variables, describing the events inside the calorimeter on the base of their different patterns, are introduced in order to discriminate between hadronic and electromagnetic showers. The ability of the artificial neural network classifier to perform a careful multidimensional analysis gives the possibility to identify antiprotons with an electron rejection 408±85 (stat) at 95.0±0.2 (stat)% of signal detection efficiency. The high accuracy achieved by this method improves substantially the efficiency in the evaluation of the cosmic ray antiproton spectrum.