Gamma-ray bursts in the Swift era

Gamma-ray burst (GRB) research has undergone a revolution in the last two years. The launch of Swift, with its rapid slewing capability, has greatly increased the number and quality of GRB localizations and x-ray and optical afterglow lightcurves. Over 160 GRBs have been detected, and nearly all have been followed up with the on-board narrow field telescopes. Advances in our understanding of short GRBs have been spectacular. The detection of x-ray afterglows has led to accurate localizations from ground based observatories, which have given host identifications and redshifts. Theoretical models for short GRB progenitors have, for the first time, been placed on a sound foundation. The hosts for the short GRBs differ in a fundamental way from the long GRB hosts: short GRBs tend to occur in non-star forming galaxies or regions, whereas long GRBs are strongly concentrated within star forming regions. Observations are consistent with a binary neutron star merger model, but other models involving old stellar populations are also viable. Swift has greatly increased the redshift range of GRB detection. The highest redshift GRBs, at z∼5–6, are approaching the era of reionization. Ground-based deep optical spectroscopy of high redshift bursts is giving metallicity measurements and other information on the source environment to much greater distance than other techniques. The localization of GRB 060218 to a nearby galaxy, and association with SN 2006aj, added a valuable member to the class of GRBs with a detected supernova. The prospects for future progress are excellent given the >10 year orbital lifetime of the Swift satellite.

Swift spends 56% of its time observing GRBs and their afterglows, with observations continuing for weeks and even months in some cases. The mission policy is to give highest priority to GRB science. The remaining time is shared between non-GRB planned targets, Target of Opportunity (TOO) observations of non-GRB transients, and calibration sources. ToOs are open to community proposal, with the decision to observe them made by the Swiff Principal Investigator based on scientific merit and observational constraints. To date, more than 150 T o 0 targets have been observed. Afterglow from 8 GRBs from other observatories has been detected by XRT.

Swift GRB Observations
As of 31 August 2006, BAT has detected 168 GRBs (annual average rate since December 2004 of -100 per year and since August 2005 of -1 10 per year). Approximately 90% of the BAT-detected GRBs have repointings within 5 minutes (the remaining 10% have spacecraft constraints that prevent rapid slewing). Of those, virtually all bursts observed promptly have detected X-ray afterglow, the only exceptions being 3 short GRBs (050906,050925,051 1054). The fraction of rapid-pointing GRBs that have UVOT detection is -30%. Combined with ground-based optical observations, about 50% of Swift GRB has optical afterglow detection.
There are 57 Swift GRBs with redshifts as listed in Table 1. This total from the first 1.7 years of Swift operations is more than the number found from all previous observations since 1997. The distribution in redshift is given in Figure 1. It is seen that Swift is detecting GRBs at higher redshift than previous missions due to its higher sensitivity and rapid afterglow observations. The average redshift for the Swift GRBs is <z> = 2.3 compared to <z> = 1.2 for previous observations. Jakobsson et al. 1121 find that the Swift redshift distribution is consistent with models where the GRB rate is proportional to the star formation rate in the universe.    Another way of considering the distances of GRBs is to plot the distribution of their look-back time. This is done for the Swij? bursts with redshift detelmiilations in Figure 2. The era of Swift GRBs is seen to have peaked at >10 Gyr in the past. The duration distribution of Stllij? detected GRBs is shown in Figure 3. Swift's shortburst fraction is -10% which is smaller than BATSE's -25% because Swift has a lower energy range than BATSE and short GRBs have hard spectra. Still, the detection rate of short bursts is 10 per year and high enough for considerable progress as discussed in the following section. Figure 4 shows the duration distribution in the source frame for those bursts with redshift determinations. The typical duration in the source frame is a factor of -3 less than that in the observer frame as one would expect from the (l+z) time dilation and average redshift of -2.3. Long GRBs have true physical durations of typically 10-20 s and not 30-60 s that we observe.

Short GRBs
At the time of Swift's launch, the greatest mystery of GRB astronomy was the nature of short-duration, hard-spectrum bursts. Although more than 50 long GRBs had afterglow detections, no afterglow had been found for any short burst. In May 2005 (GRB 050509B), Swifi provided the first short GRB X-ray afterglow localization 1201. This burst plus the HETE-2 GRB 050709 and Swift GRB 050724 led to a breakthrough in our understanding 120-24,13,25] of short bursts. BAT has now detected -13 short GRBs, most of which with XRT detections, and about half of which with host identifications or redshifts (an additional two have been detected by HETE-2).
In stark contrast to long bursts, the evidence to date on short bursts is that they typically originate from regions with low star formation rate. GRB 050509B and 050724 were from elliptical galaxies with low current star formation rates while GRB 050709 was from a region of a star forming galaxy with no nebulosity or evidence of recent star formation activity in that location. This is illustrated in Figure 5 where the images of these 3 short bursts are contrasted to 3 typical HST images of long bursts showing them coincident with regions of star formation 11261. Taken together, these results support the interpretation that short bursts are associated with an old stellar population, and may arise from mergers of compact binaries [i.e., double neutron star or neutron starblack hole (NS-BH) binaries].

Long GRBs
Short GRBs cD elliptical A list of short GRBs detected to date since GRB 050509B is given in Table 2. The list includes all bursts that researchers have discussed in the context of short events. Some, such as GRB 05091 1,060505 and 060614 are uncertain as to their long or short classification. From the 5 definite short events with firm redshifts, the concentration is seen to be near 2=0.2, but with some events as far away as 2=2, or possibly higher. It has 0.5) and farther away (z > 1). With the caveat that statistics are poor and the population appears diverse, the redshifts for short bursts are smaller on average by a factor of -4 than those of long bursts (<z,,,,> =0.5, <z,,,,,> = 2.3), and their isotropic energies are smaller by a factor of -100. Table 2. List of short GRBs with accurate localizations of sensitive searches for afterglow.
Measurements or constraining limits on beaming from light cuive break searches have been hard to come by with the typically weak afterglow of short GRBs. Figure 6 shows the best data available comparing the inferred beaming angle distributions for long and short GRBs. Based on the limited statistics available, and bearing in mind the large uncertainties involved in determining reliable breaks for the short GRB light curves, it appears short GRBs have larger beaming angles on average than for long GRBs.  Opening Angle (degrees) Swift observations also reveal new and puzzling features. Long (-100 s) "tails" with softer spectra than the first episode is seen to follow the prompt emission for about 25% of short bursts [33,34]. Also, X-ray flares on late timescales in the afterglow [35J are not easily explained by the standard coalescence model. Perhaps these flares result from a complex energy extraction process from the nascent black hole, or self-gravitational clumping instabilities at large radii in the fall-back disk 1361, or other possibilities 1371. GRB 060614 is a particularly interesting case that may or may not be a short burst with a exceptioilally bright tail as discussed in section 7.2.
Swij? localization of a short GRB increases the sensitivity of gravitational wave interferometers to detect gravitational waves from that GRB by a large factor due to the much narrower search window that can be used 1381. Detection of gravitational waves from a Swift GRB would be an enormous discovery with great scientific payoff for merger physics, progenitor types, and NS equations of state. Short GRBs are also "cosmic sirens" that can provide constraints on the properties of dark energy, if they are detected by gravitational wave detectors [39]. Even if this requires Advanced LIGO in 2012, it is feasible for Swift to be operating at that time.
We already know from the 27 December 2004 extremely luminous giant flare from SGR 1806-20 that such events could be detected to -60 Mpc and would look identical to short GRBs [40]. With Swift, we can determine whether some short GRBs are magnetar flares or if the SGR 1806-20 giant flare was an extremely rare event. A recent study [26] that searched for nearby galaxies (z<0.025) within the error boxes of six well-localized, pre-S~vift short GRBs failed to find any plausible hosts as would be expected from magnetar progenitors, and concludes that magnetar hype~flares constitute 4 5 % of all short GRBs.

Slzort GRB F~mcre Progress
Swift will provide a statistically significant sample of short GRBs as it continues to operate, with prompt emission and afterglow obsenrations for dozens of short bursts over 5 years. The key topics that will be addressed are: 1) Origin of short GRBs. Secure galaxy localizations for short GRBs now total less than 6, and hint at an older population than for long GRBs. The basic scenario of short GRBs as NS-NS mergers is supported, but many other models are also viable [41]. Increased statistics of the hosts are badly needed. The few bright, well observed bursts that Swift will provide over the coming years will lead to the most progress.
2) Sub-classes. Two of the short GRBs, 050813 and 060121, have potential host galaxies at cosmological redshifts z>l. The existence of a new class of short GRB lying at much greater distance may reveal a new class of more energetic phenomenon [28J. At the other extreme, the magnetar giant flare event of 27 December 2004, with its short duration, hard spectrum, and total energy -0.01 that of a typical short GRB, also indicates the possibility of at least one additional sub-class existing at lower luminosities. Again, more statistics are needed.
3) Prompt emission tails. The observation of soft emission lasting 10's of seconds after the prompt hard episode is a discovery that will have profound implications for models. A sample size twice as big as the current one is need to firmly establish the observational characteristics of this feature.

Afterglow Physics
Swift was specifically designed to investigate GRB afterglows by filling the temporal gap combined power of the BAT and XRT has revealed that in long GRBs the prompt X-ray emission smoothly transitions into the decaying afterglow (Figure 7 & 8). Often, a steepto-shallow transition (phases I -I1 in Figure 7) is found suggesting that prompt emission and the afterglow are distinct emission components. This also seems to be the case for short bursts [20,22]. The early steep-decay phase seen in the majority of GRBs is a real surprise. The current best explanation is that we are seeing high-latitude emission due to termination of central engine activity [43,42,44]. This phase is usually followed by an equally unexpected shallow decay phase with that begins within the first hour. The shallow phase can last for up to a day, and, although faint, is energetically very significant. It is likely due to the forward shock being constantly refreshed [42, 45,46] by either late central engine activity or less relativistic material emitted during the prompt phase. Granot et al. 1473 show how the two-component jet model [48] in which a narrow, initially highly relativistic conical jet (producing the prompt emission) embedded within a mildly relativistic coaxial cone that decelerates markedly as it plows into the CSM, can account for the early-time, flat decay (following the initial steep decay) in the XRT light curves.
Most Swift-localized GRBs are optically faint at early times 1151, in contrast to pre-Swift expectations. In some GRBs, the afterglow decays more gradually after the prompt emission. These tend to be the GRBs that are detected early with the UVOT. Here, the afterglow emission may be dominated by the external shock, as expected prior to Swift (phase I11 in Figure 7). Swift has found erratic flaring behavior (phase V in Figure 7), lasting long after the prompt phase, in some cases for several hours after the burst. The most extreme examples are flares with integrated power similar to or exceeding the initial burst 1351. The rapid rise and decay, multiple flares in the same burst, and cases of fluence comparable with the prompt emission suggest that these flares are due to continuing activity of the central engine.
There is a lack of evidence for jet breaks (breaks in temporal decay slope, phase III-IV transition in Figure 7) in the Swift X-ray afterglow 149,501. Although possible jet breaks have been measured in some bursts, the number of bursts in which such breaks are seen is small and they do not satisfy the empirical relations previously found from optical observations [29,51]. We have detected one textbook version of an achromatic jet break in both X-ray and optical (GRB 050525A, Figure 9). Whether these results invalidate the jet picture inferred from earlier optical observations remains to be seen.

aft erg lo^^ Physics F~iture Progress
Results obtained with Swijl so far have led to significant progress in understanding GRB outflows, but most issues are far from settled. In the next few years Swift will address the following topics: 1) Afterglow origin. Long-duration monitoring of additional bursts will address whether the radiative efficiency in the prompt phase is much higher than in the afterglow, providing clues as to whether the prompt emission requires a Poyntingdominated ejecta and whether the afterglow efficiency or shock microphysics varies in time. The late evolution of the light curve will also allow searches for unambiguous achromatic jet breaks to constrain jet width and intrinsic luminosity P I .
2) Rare Bright Optical GRBs. Detection of more bright optical bursts will test whether prompt optical emission is correlated with a high isotropic luminosity.
Based on experience from years 1 and 2, Swift will detect -2 of these bursts per year.
Comparison of bright optical-flash GRBs with a large sample of early UVOT detections and severe upper limits, combined with detailed modeling of forward shock and reverse shock emission, directly addresses whether GRB fireballs are baryonic or magnetic in origin (e.g., 153-561).
3) High Redshift Fireballs. A large sample of high redshift bursts will determine whether their fireball physics is similar to that of nearby bursts, or whether it evolves as a function of redshift. 4) X-ray Flares. A large sample of bursts with X-ray flares will constrain how flares evolve during an individual burst and how they correlate with other GRB properties. Such correlations test if the flares are powered by the central engine. This will also test disk models with fragmentation or MHD-dominated accretion as the explanation of flaring behavior. 5) Central Engine. Monitoring the temporal and spectral evolution of large numbers of GRBs during the shallow decay phase will constrain the possible late ejection of and/or the range in initial Lorentz factor of the entrained material in the relativistic jet. These data can be compared to detailed numerical simulations of the various GRB progenitors to study the behavior of the central engine.

High Redshift GRBs and Cosmology
GRBs, as the most brilliant explosions we know of, offer the potential to probe the early Universe into the epoch of reionization. They can trace the star formation, re-ionization, and metallicity histories of the Universe [57-601. GRBs are 100 -1000 times brighter at early times than are high redshift QSOs (the near infrared afterglow of GRB 050904 was J = 17.6 at 3.5 hours). Also, they are expected to occur out to z > 10, whereas QSOs drop off beyond z = 3. Another benefit is that GRB afterglows produce no "proximity effects" on intergalactic distances scales, and have simple power-law spectra and no emission lines. Thus GRBs are "clean" probes of the intergalactic medium (IGM). Figure 1 and Table 1 show that 6 of the 8 highest redshift GRBs ever seen were discovered by Swift, including bursts at redshifts ~~5 . 3 , and 6.3 [61-631. Of the GRBs with measured redshift, we find that 4 out of 50 or -8% of Swift GRBs lie at z > 5, consistent with model predictions [60,12]. These same models predict that Swift can detect GRBs to redshifts of z>8. A great deal of effort is currently being invested in order to rapidly recognize such bursts and obtain redshifts with large ground-base IR spectrographs.
The time evolution of gamma-ray and X-ray fluxes of 4 high-z GRBs is shown in Figure 10. All of these bursts are exceptionally luminous and long-lasting, and their evolution can be very complex.  Swift's rapid localizations have provided new opportunities for spectroscopy of highredshifi GRB afterglows. Observed at low resolution, the host galaxy appears as a damped Ly-a (DLA) system along with a rich array of metallic lines which can be used to infer metal abundances. At high resolution, the host absorption lines split into an array of fine-structure transitions, which allows the inference of gas densities and even of diffuse radiative conditions in the host galaxy [64,65]. Figure 11 is an example of an optical spectrum for a high redshift (~~4 . 3 ) GRB [651. Countless lines are evident in the spectrum included a damped Ly-a feature corresponding to a neutral hydrogen column density of 1 022 ~r n -~. The lines imply a density of 100 cm-3 in the source region. Absorption lines observed in infrared spectroscopic observations of GRB 050904 gave a metallicity measurement of 5% solar 1631, the first metallicity determination at such high redshift demonstrating that the observed evolution in the mass-and luminosity-metallicity relationships from z = 0 to 2 continues to z>6 [66]. . BAT  z=4.275 as well as two foreground absorbers 165 1.

High Redslzift GRB Fut~lre Progress
We consider here the important cosmological topics that Swift observations of GRBs and their afterglows can address over the next few years.
1) Reionization. Observation of one Swifi GRB at 2 7 would provide more information about reionization than all of the SDSS quasars combined. It is impressive that these studies will probe the IGM less than 1 Gyr after the Big Bang. In no other way can such unique observations be made.
2) Star Formation Rate. The connection between long GRBs and SNe opens the possibility of using the redshifts of long GRBs to infer the cosmological star formation history, with relatively minor (or in any case unique) selection effects [57, 60,67,68]. Preliminary estimates of the star formation rate derived from Swift bursts 1691 shows a flat or (at the highest redshifts) slowly-declining star formation rate, consistent with color-selected galaxy observations r70J.
3) The First Generation of Stars. Whether massive Population 111 stars can produce GRBs is not yet known [71,72]. If such stars, perhaps stripped of their outer envelopes by a binary companion, do produce GRBs, Swift may detect them for two reasons: first, because GRBs are so bright; and second, because metal enrichment of the IGM is expected to be heterogeneous. Regions of low metallicity are consequently expected to survive for a substantial period of timepossibly to z = 10, or even z = 6. Detection of a GRB from the collapse of a massive Pop I11 star would provide a demonstration of the existence of such stars.

Observations of GRB 060218 1 SN 2006aj
On 18 February 2006 Swifl detected the remarkable burst GRB 060218 that provided considerable new information on the connection between SNe and GRBs. It lasted longer than and was softer than any previous burst, and was associated with SN 2006aj at only z=0.033. The BAT trigger enabled XRT and UVOT observations during the prompt phase of the GRB and initiated multiwavelength observations of the supernova starting at the time of the initial core collapse.
The spectral peak in prompt emission at -5 keV places GRB 060218 in the X-ray flash category of GRBs [73]. Combined BAT-XRT-UVOT observations provided the first direct observation of shock-breakout in a SN [73]. This is inferred from the evolution of a soft thermal component in the X-ray and UV spectra, and early-time luminosity variations. Concerning the supernova, SN 2006aj was dimmer by a factor -2 than the previous SNe associated with GRBs, but still -2-3 times brighter than normal SN Ic not associated with GRBs 174,751. GRB 06021 8 was an underluminous burst, as were 2 of the other 3 previous cases. Because of the low luminosity, these events are only detected when nearby and are therefore rare occurrences. However, they are actually -10 time more common in the universe than normal GRBs 1761.

Tlze Pecwliar Case of GRB 060614
GRB 060614 was a low-redshift, long-duration burst with no detection of a coincident supernova to deep limits. It was a bright burst (fluence in 15-150 keV band of 2.2~10." erg ~m '~) and well studied in the X-ray and optical. With a Tw duration of 102 s, it seemly falls squarely in the long burst category. A host galaxy was found [77-791 at z=0.125 and deep searches made for a coincident supernova. All other well-observed nearby GRBs have had supernovae detected, but this one did not to limits >I00 times fainter than previous detections [77-791.
We have found that GRB 060614 shares some characteristics with short bursts [80]. The BAT light curve shows a first short, hard-spectrum episode of emission (lasting 5 s) followed by an extended and somewhat softer episode (lasting -100 s). The total energy content of the second episode is five times that of the first [fluence of (1.69+0.02)~10~~ curve appearance (short hard episode followed by long soft emission) is similar in many respects to that of several recent Swift and HETE-2 short-duration bursts (GRB 050709, 050724,05091 1,051227) and a subclass of BATSE short bursts r811. There are differences in that the short episode of this burst is longer than the previous examples and the soft episode is relatively brighter.
Another similarity with short bursts comes from a lag analysis of GRB 060614 1801. Figure 12 shows the peak luminosity (L,,,,) in Swift GRBs as a function of their spectral lag (t,,,) between the 50 -100 keV and 15-25 keV bands. It is possible for the first time to include short bursts in such a plot with the redshift determinations for several short events from the past 2 years. For long bursts there is an anti-correlation between t,,, and L,,,,, whereas short bursts have small t,,, and small L,,,, and occupy a separate area of parameter space. The lag for GRB 060614 for the first 5 s is 3 t 6 ms which falls in the same region of the lag-luminosity plot as short bursts. Figure 12. Spectral lag as a function of peak luminosity showing GRB 060614 in the region of short GRBs. The lags and peak luminosities are corrected to the source frame of the GRB. The data points labeled as long bursts are from Swift, with the exception of GRB 030528 which is a very long-lagged HETE-2 burst. The blue data points for short bursts are from S~d~ift. In green are the 4 nearby long GRBs with associated SNe. The three of the four (980425,031203 and 060218) fall below the long-burst correlation, while the only SN-associated GRB with normal luminosity (030329) falls near the long-burst line. From ref. [80].

Log [t,,d(l+~)O-~~]
It is difficult to determine unambiguously which category of burst the well-observed GRB060614 falls into. It is a long event by the traditional definition, but it lacks an associated SN as had been seen in all other nearby long GRBs. It shares some similarities with Swift short bursts, but has important differences such the brightness of the extended soft episode. If it is due to a collapsar, it is the first indication that some massive star collapses either fail as supernovae or highly underproduce '6Ni. If it is due to a merger, then the bright long-lived soft episode is hard to explain for a clean NS-NS merger where little accretion is expected at late time but might fit in a NS-BH scenario. I11 any case, this peculiar burst is challenging our classifications of GRBs.
We note that GRB 060505 appears to also be another nearby long GRB with no coincident SN [82]. It was an unitrigger Swift burst found in ground processing, and so does not have much data from the on-board instruments aside from a BAT light curve and XRT position. The duration was T,, = 4.0 s. Ground-based studies of the optical afterglow gave an association with a galaxy at z=0.089 and no coincident supernova to deep limits.

GRB-SN Corzrzection Put~ire Progress
Although the average redshift of Swift bursts is large, there are still a good number of events detected at small enough distance for sensitive supernova searches. Table 1 shows that 3 events have z<0.1 giving a nearby-burst detection rate of more than one per year. It is probable that Swift will detect 2 or more GRBs with well-observed coincident supernovae (or deep limits) over the next 5 years. The Swift supernova-GRB data set will then be about as large as all previous detections. In addition, the rapid response of the satellite will give coverage to the full supernova light curve from core collapse through the fading of the '"Co decay. Key topics to address in the coming years are: 1) Population of underluminous GRBs. Although rarely detected, the nearby weak bursts with coincident SN greatly outnumber normal GRBs. A uniform search for such events with Swift over many years will give a much better determination of the population size.
2) GRB -SN relationship. A key open question is whether all long GRBs have coincident SNe associated with them. Observations over several years with deep optical searches for SNe will answer this question. There is already a hint from GRB 060614 and 060505 that some long bursts have no associated SN or very faint onesor perhaps we do not yet know how to distinguish mergers from collapsars.
3) GRB jet physics. Supernova GRBs observed at low redshift provide unique observations of the emergence ofjets from the stellar envelop. The Swift data are particularly valuable because they start at the time of the collapse and give multiwavelength coverage of the jet emergence. It is anticipated that Swift will make such observations about once every 2 years.

Conclusions
Our understanding of GRBs has advanced greatly in the past 2 years. Swift is providing rapid and accurate localizations, which lead to intensive observing campaigns by Swifl and ground-based observatories starting -1 minute after the GRB trigger. Unifonn multiwavelength afterglow light curves are available for the first time for a large number of bursts. The data have led to a break-through in our understanding of short GRBs, have extended our knowledge of the high redshift universe, have elucidated the physics taking place in the highly relativistic GRB fireball outflows and have added significantly to the study of the connection between GRBs and SNe. The Swifl mission has an orbital lifetime of >10 years and no expendable resources on board, and so is likely to greatly expand on these results with detailed observations of >I000 bursts.