Owing to narrow energy band of textit{Swift}/BAT, several urgent issues are required to pay more attentions but unsolved so far. We systematically study the properties of a refined sample of 283 textit{Swift}/BAT gamma-ray bursts with well-measured spectral peak energy ($E_{text p}$) at a high confidence level larger than 3$sigma$. It is interestingly found that duration ($T_{90}$) distribution of textit{Swift} bursts still exhibits an evident bimodality with a more reliable boundary of $T_{90}simeq$1.06 s instead of 2 s for previously contaminated samples including bursts without well-peaked spectra, which is very close to $sim$1.27 s and $sim$0.8 s suggested by some authors for Fermi/GBM and textit{Swift}/BAT catalogs, respectively. The textit{Swift}/BAT short and long bursts have comparable mean $E_{text p}$ values of $87^{+112}_{-49}$ and $85^{+101}_{-46}$ keV in each, similar to what found for both types of BATSE bursts, which manifests the traditional short-hard/long-soft scheme may not be tenable for the certain energy window of a detector. In statistics, we also investigate the consistency of distinct methods for the $E_{text p}$ estimates and find that Bayesian approach and BAND function can always give consistent evaluations. In contrast, the frequently-used cut-off power-law model matches two other methods for lower $E_{text p}$ and will overestimate the $E_{text p}$ more than 70% as $E_{text p}>$100 keV. Peak energies of X-ray flashes, X-ray rich bursts and classical gamma-ray bursts could have an evolutionary consequence from thermal-dominated to non-thermal-dominated radiation mechanisms. Finally, we find that the $E_{text p}$ and the observed fluence ($S_{gamma}$) in the observer frame are correlated as $E_psimeq [S_{gamma}/(10^{-5} erg cm^{-2})]^{0.28}times 117.5^{+44.7}_{-32.4}$ keV proposed to be an useful indicator of GRB peak energies.
Download