Progress in understanding the nature of short gamma-ray bursts (GRBs) has been rapid since the discovery of the first afterglows in mid-2005. The emerging picture appears to be of short GRBs, which originate at moderate redshift (a few tenths) and appear in galaxies of all ages. This discovery has been used to argue for their origin in compact binary mergers. However, this population does not describe all short bursts. Here, I will present results of observations of several short GRBs, which challenge the conclusions drawn from the early observations. The observations show that some short GRBs originate in the very low redshift Universe (below 100 Mpc), while some may also lie at redshifts comparable with the long GRBs (i.e. z>2). I will discuss the properties of these bursts and the implications they have for the progenitors of short GRBs.
Short gamma-ray bursts (SGRBs) are a distinct subset of gamma-ray bursts (GRBs) with durations of less than 2 s and typically hard gamma-ray spectra. Their origin was largely unconstrained by observation until mid-2005 when the first X-ray, optical and radio afterglows were discovered (Berger et al. 2005; Gehrels et al. 2005; Hjorth et al. 2005). The first few afterglows all pointed to galaxies at moderate redshifts (z∼0.2), and populations of all ages, crucially including those with no ongoing star formation. This immediately suggested a differing progenitor system from the long-duration GRBs (LGRBs), whose origin in the core collapse of massive stars (e.g. Hjorth et al. 2003) naturally links them to young, star-forming galaxies (e.g. Fruchter et al. 2006). The favourite for this system was the merging of two compact objects (e.g. neutron star–neutron star, neutron star–black hole). It is thought that such mergers can create the necessary conditions for GRB formation (Rosswog et al. 2003; Davies et al. 2005) and also explain their origin in galaxies with old populations.
However, these early observations only uncovered one aspect of the SGRB phenomena, and it is becoming increasingly apparent that the short-burst population is significantly more diverse than was initially appreciated. It is these newer observations that challenge conclusions drawn from the first bursts, which are discussed in more detail here.
2. The short gamma-ray burst population
At the time of writing (September 2006), approximately six short bursts have optical afterglow identifications, with four of these yielding moderately secure host galaxy redshifts1. The mean redshift of these bursts is z≈0.3, with a range of 0.16<z<0.54. Although a small population, this is markedly different from the redshift distribution of LGRBs detected by Swift, which have a mean of z=2.8 (Jakobsson et al. 2006b). Further, the host galaxies selected by these bursts include those where significant star formation has not occurred for several gigayears, in contrast to the young populations seen in LGRB hosts. Thus, it is clear that SGRBs are the result of a different progenitor system to the LGRBs, with binary mergers being the most widely discussed, although other progenitor systems may also explain the observations.
3. Short gamma-ray bursts in the local Universe
On 27th December 2004, the Galactic Soft Gamma Repeater (SGR) 1806-20 underwent a hypergiant flare, emitting approximately 1046 ergs in a 0.2 s duration spike (e.g. Hurley et al. 2005). Such a bright flare would have been visible as an SGRB, had it occurred in an external galaxy out to a distance of approximately 60 Mpc, and raises the obvious possibility that some fraction of short bursts could be due to similar flares occurring in external galaxies.
(a) Searches for SGR flares from BATSE bursts
The giant flare from SGR 1806-20 led many authors to search for similar events in the BATSE catalogue by searching for those events that exhibited similar spectral properties (Lazzati et al. 2005), or for an excess of events in the direction of Virgo (Popov & Stern 2005). An alternative approach is to correlate the locations of BATSE bursts with large-scale structure. This approach is potentially more sensitive than others, since it does not look only at one region of the sky, nor does it necessarily require the brightest bursts (i.e. those that have sufficient signal to noise for constraining spectral fitting). Such an approach was taken by Tanvir et al. (2005), who found that approximately 10% of the SGRBs originate from those with 25 Mpc. The energies of these bursts would typically be of order 1046 ergs and would fit naturally in the range of magnetar flares. Interestingly, this analysis shows a markedly stronger correlation when restricted to early-type galaxies (although as the PSCz catalogue is IRAS selected, this predominantly spirals with relatively few elliptical galaxies). This may be simply that earlier-type galaxies tend to be more strongly clustered, thereby demonstrating that SGRBs trace large-scale structure; however, it may also indicate that even local SGRBs prefer older populations. If these bursts are due to SGR flares, then it would require their production in old stellar systems, perhaps by accretion-induced collapse following the merger of two white dwarfs (Levan et al. 2006a).
(b) Possible local examples: gamma-ray bursts 050906, 051103 and 060502B
Although BATSE has produced by far the most extensive sample of short-duration bursts, its large error boxes bar the identification of any individual host galaxies. However, a number of bursts have been identified by other missions (e.g. IPN, HETE, Swift), with error boxes only a few square arcminutes in area. Given the conclusions reached above, it is therefore reasonable to ask whether any of these bursts could plausibly have originated via SGR giant flares in a moderately nearby galaxy.
The most probable candidate for such a flare is SGRB 051103, detected by the IPN, whose error box overlapped significantly with the outer edge of both M81 and M82, and may well represent the nearest short burst to the Milky Way (Frederiks et al. in press; Ofek et al. 2006). Other possible bursts for which this is true are SGRB 050906 and 060502B, which may have been associated with IC328 (Levan et al. in preparation) and UGC 11292 (Bloom et al. 2007), respectively. These galaxies, lying approximately 100 Mpc distant, would require SGR flares of somewhat greater luminosity than those seen so far in the galaxy. However, creating such flares remains plausible on theoretical grounds, while the luminosity function for giant flares is largely unconstrained due to the paucity of observed events.
(c) The difficulties in probabilistic associations
A crucial problem in the studies of SGRBs to date has been the frequently poor localizations for the bursts. Even moderately accurate X-ray positions (i.e. those that locate the bursts to within a few arcseconds) are not necessarily sufficient to unambiguously locate the host galaxy. For example, in the case of GRB 050509B, the X-ray error box contains a large number of galaxies, any of which could in theory have contained the GRB (even though the probability is that it originated in the brightest elliptical galaxy). The danger of these associations has been demonstrated by the recent GRB 060912A. This burst was located approximately 10′′ from the core of a bright elliptical galaxy at z=0.09 suggesting an association; indeed, with an X-ray-only localization, this would have probably been the conclusion. However, the optical afterglow allowed deep observations to pinpoint a more distant host galaxy, and showed that the burst actually originated at z=0.93 (Jakobsson et al. 2006a; Levan et al. 2006b), demonstrating that, at least in this case, there really was a chance coincidence with a more nearby galaxy.
4. Short gamma-ray bursts at high redshift
The early observations of short bursts essentially confined them to galaxies at z<1, and led several authors to suggest that this represented that long delay times were necessary for the formation of SGRBs. However, other possibilities should also be discussed. For example, the SGRBs observed at lower redshifts had typical energies of approximately 1048 ergs; it is therefore plausible that at higher redshift, these bursts remain invisible, and thus only the brightest (and rarest) SGRBs can be seen. Indeed, it appears that a population of higher redshift and, subsequently, much more luminous bursts are starting to be found.
(a) Gamma-ray burst 060121
GRB 060121 was detected by HETE-2 on 21st January 2006, with an afterglow being discovered in observations taken approximately 2 h post-burst (Malesani et al. 2006; Levan et al. 2006c). The afterglow was faint in the optical and very red (de Ugarte Postigo et al. 2006). Unlike previous SGRBs, there was no obvious host galaxy underlying the GRB position. Indeed, deep Hubble Space Telescope (HST) observations were required to uncover a faint, red host galaxy with R∼27 (Levan et al. 2006d). The properties of both afterglow and host galaxy can most readily be explained if the burst occurred at markedly higher redshift than the first GRBs. The two probable scenarios for the redshift of the burst are that (i) it occurred at z∼1.5, but with significant host galaxy extinction or (ii) it occurred at z∼4.5. Both of these require at least an order of magnitude more energy than bursts observed at z∼0.3.
(b) Gamma-ray burst 060313
GRB 060313 exhibits one of the best-studied X-ray/optical afterglows for a short GRB (Roming et al. 2006), and being detected by Konus-WIND in addition to Swift enabled its prompt properties to be well studied. In particular, its location in the hardness duration diagram (figures 1 and 2) lies closest to those of the SGRBs detected by BATSE. Deep observations with the VLT reveal only a very faint possible host galaxy, with R∼26. This host galaxy is much fainter than those at moderate redshift, and therefore is likely to be at significantly higher z.
5. Implications for progenitor models
The current rate of magnetar giant flares in the galaxy is weakly constrained, although estimates of around one per century have been made. However, it is worth noting that even if the rate of 1 per 1000 years would result in a local rate of approximately 20 yr−1 in a volume of 100 Mpc radius, making it difficult to avoid any contribution from SGR flares to the SGRB rate. Giant flares from magnetars are thought to be produced by the global reconfiguration of the magnetic field. Over the course of the flare, a few percent of the total field is lost and this provides a limit on the plausible energy of the flare. Such considerations imply that giant flares could, at most, be approximately 10 times brighter than the flare from SGR 1806-20, and so the energies of the more distant bursts (e.g. greater than 1048 ergs) would seem to be beyond the range of energies plausible, although it should be noted that any beaming in the flare would act to reduce the total energy output and may provide a viable explanation.
At higher redshifts, it seems probable that another progenitor system is needed. For the systems observed at moderate redshift, this could well be double degenerate mergers, since their long delay times and potentially large natal kicks can naturally explain the observed host galaxy properties. At very high redshifts (e.g. z>4), the delay times are necessarily smaller, since the age of the Universe at z∼4 is only 1.5 Gyr. However, this is not necessarily a problem for degenerate mergers, since it is predicted that a moderate fraction of them will merge within 106 years of their formation. Potentially, more troubling is the energy budget required at the high redshift. Lower redshift bursts have isotropic energy releases in the region of 1048 ergs, while at higher–z, 1051 ergs are required, thereby requiring a broad luminosity function. At the present time, the expected luminosity functions of degenerate mergers are largely unknown, and so strong constraints on the progenitors cannot be made via this route. Nonetheless, it appears that the luminosity function (and delay time distribution) for short bursts may be significantly wider than has previously been considered (e.g. Nakar et al. 2006).
The mystery of SGRBs is only beginning to be unravelled some 30 years since their first detection. As with much progress in GRB astronomy, the first few observations pointed to a relatively simple picture—SGRBs in all ages of stellar populations and at moderate redshifts—however, as the sample has grown, it has become clear that this represents only one aspect of a population, which is, if anything, even more diverse than the long GRBs that have been the subject of the most study over the past decade. Future progress into the nature of these short bursts will require the building of a significantly larger sample, and crucially on securing redshift identifications (from afterglows where possible) even for the faintest and, possibly, most distant SGRBs.
One contribution of 35 to a Discussion Meeting Issue ‘Gamma-ray bursts’.
↵Although it should be noted that at the current time there are no absorption redshifts for SGRBs.
- © 2007 The Royal Society