The Rates of type I X-ray Bursts from Transients Observed with RXTE: Evidence for Black Hole Event Horizons

We measure the rates of type I X-ray bursts from a likely complete sample of 37 non-pulsing Galactic X-ray transients observed with the RXTE ASM during 1996-2004. Pointed RXTE observations of these sources are used to identify bursts via timing and spectral criteria applied to PCA data. These X-ray transients have been classified as dynamical black hole binaries (9), black hole candidates (10), neutron-star systems (14), and type unknown (4), and 99% of the PCA archive is distributed among the first three categories. Our strategy is to determine (i) whether the burst rates or upper limits, measured as a function of the bolometric luminosity, are consistent with the predictions of a burst model for accreting neutron stars and (ii) whether the rates measured in black hole systems are consistent with those expected for heavy compact objects with a solid surface, or if they in fact require true black holes with event horizons. The burst models we use are augmented versions of the models developed by Narayan & Heyl (2002; 2003). The selection of X-ray transients provides burst rate measurements over a wide range in source luminosity, and this target sample is one of the differences between the present study and the investigation of burst rates in the context of event horizons by Tournear et al. (2003). The neutron-star group consists of known bursters, although not all of them exhibited X-ray bursts during the RXTE observations that we analyzed. Burst searches yield 137 candidates from 3.7 Ms of PCA light curves, and 135 of these are confirmed as type I X-ray bursts via successful modeling of the burst spectrum with a blackbody function. The measured burst function is found to be consistent with the model predictions for neutron stars, when considerations of the distance uncertainties are factored in. Searches of 3.8 Ms of PCA light curves of black-hole binaries and 2.7 Ms for black-hole candidates yield 20 burst candidates. However, none of these events passes the spectral test for blackbody emission from a thermonuclear explosion. Isolating the blackhole binaries, for which there are reasonable estimates of mass and distance, we find that the upper limits for X-ray burst rates deviate significantly from the predictions of the burst model for heavy compact objects with a solid surface. The estimated probability that the measurements are consistent with this burst model are below 6×10. The addition of exposure times for the black hole candidates further decreases the agreement with the solid-surface model to a probability level of 9 × 10. We conclude that the burst model performs well for neutron stars, and that the lack of type I X-ray bursts for black holes and candidates rules out the presence of surfaces in these objects and provides indirect evidence that black holes do have an event horizon. Subject headings: neutron star physics — black hole physics — general relativity — X-rays: stars