2007MNRAS.380L..45S

We compute the probability of detecting long gamma-ray bursts (GRBs) at z ≥ 5 with Swift, assuming that GRBs form preferentially in low-metallicity environments. The model fits both the observed Burst and Transient Source Experiment (BATSE) and Swift GRB differential peak flux distributions well and is consistent with the number of z ≥ 2.5 detections in the 2-yr Swift data. We find that the probability of observing a burst at z ≥ 5 becomes larger than 10 per cent for photon fluxes P < 1 ph/s/cm2, consistent with the number of confirmed detections. The corresponding fraction of z ≥ 5 bursts in the Swift catalogue is ∼10-30 per cent depending on the adopted metallicity threshold for GRB formation. We propose to use the computed probability as a tool to identify high-redshift GRBs. By jointly considering promptly available information provided by Swift and model results, we can select reliable z ≥ 5 candidates in a few hours from the BAT detection. We test the procedure against last year Swift data: only three bursts match all our requirements, two being confirmed at z ≥ 5. Another three possible candidates are picked up by slightly relaxing the adopted criteria. No low-z interloper is found among the six candidates.