2014MNRAS.439.3439P

We have analysed in a systematic way about nine years of INTEGRAL data (17-100 keV) focusing on supergiant fast X-ray transients (SFXTs) and three classical high-mass X-ray binaries (HMXBs). Our approach has been twofold: image-based analysis, sampled over a ∼ ks time frame to investigate the long-term properties of the sources and light-curve-based analysis, sampled over a 100 s time frame to seize the fast variability of each source during its ∼ ks activity. We find that while the prototypical SFXTs (IGR J17544-2619, XTE J1739-302 and SAX J1818.6-1703) are among the sources with the lowest ∼ ks-based duty cycle (<1 per cent activity over nine years of data), when studied at the 100 s level, they are the ones with the highest detection percentage, meaning that, when active, they tend to have many bright short-term flares with respect to the other SFXTs. To investigate in a coherent and self-consistent way all the available results within a physical scenario, we have extracted cumulative luminosity distributions for all the sources of the sample. The characterization of such distributions in hard X-rays, presented for the first time in this work for the SFXTs, shows that a power-law model is a plausible representation for SFXTs, while it can only reproduce the very high luminosity tail of the classical HMXBs, and even then, with a significantly steeper power-law slope with respect to SFXTs. The physical implications of these results within the frame of accretion in wind-fed systems are discussed.