Statistical Evidence for Three Classes of Gamma-ray Bursts

Two different multivariate clustering techniques, the K-means partitioning method and the Dirichlet process of mixture modeling, have been applied to the BATSE Gamma-ray burst (GRB) catalog, to obtain the optimum number of coherent groups. In the standard paradigm, GRB are classified in only two groups, the long and short bursts. However, for both the clustering techniques, the optimal number of classes was found to be three, a result which is consistent with previous statistical analysis. In this classification, the long bursts are further divided into two groups which are primarily differentiated by their total fluence and duration and hence are named low and high fluence GRB. Analysis of GRB with known red-shifts and spectral parameters suggests that low fluence GRB have nearly constant isotropic energy output of 1052 ergs while for the high fluence ones, the energy output ranges from 1052 to 1054 ergs. It is speculated that the three kinds of GRB reflect three different origins which may be mergers of neutron star systems, white dwarf with neutron stars and collapse of massive stars. Subject headings: Gamma Rays: Bursts Methods: Data Analysis Methods: Statistical