2022ApJ...940....5D

We select the largest sample of Swift gamma-ray bursts (GRBs) so far to reexamine the classification in terms of time duration, hardness ratio, and physical collapse model. To analyze the sample selection effect, we divide the observed Swift GRB sample into four subsamples according to signal-to-noise level, spectral quality, and extended emission. First, we find that only the sample of Swift GRBs with well-measured peak energy can be evidently divided into two types at a boundary of ∼1 s, and other data sets are well described by three Gaussian functions. Using Swift GRBs with known redshift, a Kolmogorov-Smirnov test shows the intrinsic duration distributions of five data sets are equally distributed. Second, we ascertain in the plane of hardness ratio versus duration that the hardness ratio of short GRBs is significantly higher than those of middle classes and long GRBs, while the latter two components are the same in statistics, implying the so-called middle class to be artificial. Third, we apply a collapse model to discriminate the boundaries between collapse and noncollapse Swift bursts. It is interesting to find that a significant fraction, ≥30%, of Swift short GRBs could have originated from the collapsing progenitors, while all long GRBs are produced from the collapsars only. Finally, we point out that short GRBs with extended emission are the main contributors to the noncollapsar population with longer duration.