2002A&A...382..947K

We report on ten type I X-ray bursts originating from GX 17+2 in data obtained with the RXTE/PCA in 1996-2000. Three bursts were short in duration (∼10s), whereas the others lasted for ∼6-25min. All bursts showed spectral softening during their decay. There is no evidence for high-frequency (>100Hz) oscillations at any phase of the bursts. We see no correlations of the burst properties with respect to the persistent X-ray spectral properties, suggesting that in GX 17+2 the properties of the bursts do not correlate with inferred mass accretion rate. The presence of short bursts in GX 17+2 (and similar bright X-ray sources) is not accounted for in the current X-ray bursts theories at the high mass accretion rates encountered in this source. We obtain satisfactory results if we model the burst emission with a black body, after subtraction of the persistent pre-burst emission. The two-component spectral model does not fit the total burst emission whenever there is a black-body component present in the persistent emission. We conclude that in those cases the black-body contribution from the persistent emission is also present during the burst. This implies that, contrary to previous suggestions, the burst emission does not arise from the same site as the persistent black-body emission. The black-body component of the persistent emission is consistent with being produced in an expanded boundary layer, as indicated by recent theoretical work. Five of the long bursts showed evidence of radius expansion of the neutron star photosphere (independent of the spectral analysis method used), presumably due to the burst luminosity reaching the Eddington value. When the burst luminosity is close to the Eddington value, slight deviations from pure black-body radiation are seen at energies below ≃10keV. Similar deviations have been seen during (long) X-ray bursts from other sources; they can not be explained by spectral hardening models. The total persistent flux just before and after the radius expansion bursts is inferred to be up to a factor of 2 higher than the net peak flux of the burst. If both the burst and persistent emission are radiated isotropically, this would imply that the persistent emission is up to a factor of 2 higher than the Eddington luminosity. This is unlikely and we suggest that the persistent luminosity is close to the Eddington luminosity and that the burst emission is (highly) anisotropic (ξ∼2). Assuming that the net burst peak fluxes equal the Eddington limit, applying standard burst parameters (1.4M_☉ neutron star, cosmic composition, electron scattering opacity appropriate for high temperatures), and taking into account gravitational redshift and spectral hardening, we derive a distance to GX 17+2 of ∼8kpc, with an uncertainty of up to ∼30%.