$D^{*+}$ identification using a neural network method in $\nu_{\mu}$ charged current interactions in the NOMAD experiment

NOMAD is a neutrino experiment designed to search for nu /sub tau / appearance in the CERN-SPS wide band nu /sub mu / beam. The nu /sub mu / charged current interactions in the experiment target could also be used to study charm quark production. A search was made among nu /sub mu / charged current events to isolate the signal: D*/sup +/ to D/sup 0/+ pi /sup +/ to K/sup -/+ pi /sup +/. A neural network method, with backpropagation algorithm, has been used for this purpose and has been compared to a likelihood ratio method. The results showed a large improvement in signal identification with the neural network method, for the same background rejection level. A 90% purity signal sample have been extracted by the likelihood ratio method (26 events) and by the neural network method (47 events), out of which 26 events have been selected by both methods. (4 refs).