Downloaded from http://rsta.royalsocietypublishing.org/ on September 30, 2016 Phil. Trans. R. Soc. A (2007) 365, 1281–1291 doi:10.1098/rsta.2006.1967 Published online 9 February 2007 Swift-BAT results on the prompt emission of short bursts B Y S COTT D. B ARTHELMY * NASA/Goddard Space Flight Center, Greenbelt, MD 20771, USA This is a brief review of short hard bursts (SHBs) from previous missions and from SwiftBAT; in particular, a review of the developing class of gamma-ray bursts which are similar to SHBs in that they have the short hard initial spike (0.1 to a few seconds), but that they also have a long extended phase of soft emission (50–200 s). Further, we suggest that a class of events discovered by Horvath in the T90 versus hardness ratio plane is this SHB with extended emission. Keywords: gamma-ray burst; SHB; extended emission; long GRB 1. Introduction This will be an abbreviated version of the talk given at the Royal Society discussion meeting. This proceedings paper will concentrate only on the observations made on short hard bursts (SHBs) and, in particular, those SHBs with the so-called extended emission (SHBwEE). There will be a brief review of the T90 versus hardness ratio (HR) and spectral lag work, mostly to provide a groundwork for the Burst Alert Telescope (BAT) SHBwEE results. There will also be a brief section on the status of the BAT instrument and its performance. It will not cover (i) the redshift distribution results, (ii) the BAT burst catalogue results that allow a pseudo-redshift determination or (iii) the transition from the prompt gamma-ray emission phase into the X-ray emission phase. 2. Burst duration distributions The work of Norris (1984) with KONUS-13/14 and ISEE-3 using the durations of bursts gave indications that there were two distinct types or populations of bursts. The work by Kouveliotou et al. (1993) using the much larger BATSE sample of bursts plus spectral information confirmed this. Figure 1 shows the distribution of T90 durations for bursts detected by the CGRO-BATSE instrument. The distribution clearly shows two peaks: one peaking at approximately 0.5 s and the other at approximately 30 s. These are the so-called SHBs and the long bursts (sometimes also called the long soft bursts; LSBs). The hard versus soft spectral aspects will be discussed later. The minimum between these two peaks is at approximately 2 s. It is this minimum that has led to a *scott@milkyway.gsfc.nasa.gov One contribution of 35 to a Discussion Meeting Issue ‘Gamma-ray bursts’. 1281 This journal is q 2007 The Royal Society Downloaded from http://rsta.royalsocietypublishing.org/ on September 30, 2016 1282 S. D. Barthelmy 80 number of bursts 60 40 20 short 0 10–3 10–2 10–1 1 duration (s) long 10 102 103 Figure 1. A histogram of the T90 durations for 586 bursts detected by CGRO-BATSE. The distribution clearly shows the two populations of bursts: short bursts and long bursts. The minimum between these two peaks is approximately 2 s. general working approximation within the burst community that SHBs have durations less than 2 s and that LSBs have durations longer than 2 s. It should be noted that these two distributions have a significant overlap. It is important to note that the details of this duration distribution are dependent on the detecting instrument. In particular, they are dependent on the energy window of the instrument and the trigger criteria used. As an example, the BeppoSAX Wide Field Camera did not detect any short bursts during the 6 years of operations. The KONUS-13/14 distribution is similar (Norris 1984). Figure 2 shows the T90 distribution for the Swift-BAT instrument. The ratio of SHB to LSB is much lower in BAT than for BATSE. This ratio difference is generally due to the softer energy range of the BAT trigger (15–150 keV) compared with the BATSE one (50–300 keV). The spectral differences between the short and the long bursts are seen in figure 3 (ignore the ellipses for the moment; they will be covered in §4c). Here, the HR is plotted against the T90 duration. While the separation of the two populations in the duration (T90) dimension is clearly seen, the separation of the two populations in the spectral (HR) dimension is not so strong. There is a much greater overlap in the HR for the two populations. 3. Short hard bursts As already mentioned, there are two classical populations of gamma-ray bursts (GRBs): SHB and LSB. Swift-BAT has confidentally detected 15 (and maybe as many as 18) SHBs. They are listed in table 1 along with some related parameters (redshift, lag, host galaxy and comments). For a complete discussion of this table and results, please see Gehrels (2007). Phil. Trans. R. Soc. A (2007) Downloaded from http://rsta.royalsocietypublishing.org/ on September 30, 2016 1283 Swift-BAT results and short bursts 25 number of bursts 20 15 10 5 0 10–3 0.01 0.1 1 T90 (s) 10 100 1000 Figure 2. The T90 distribution of bursts detected by Swift-BAT. The SHB peak is only weakly present in the BAT-detected bursts. 2 1.5 HR 1 –0.5 –0 –0.5 –1 –2 –1.5 –1 –0.5 0 0.5 T90 1 1.5 2 2.5 3 Figure 3. BATSE T90–HR scatterplot with the three populations fitted with two-dimensional Gaussians (shown with ellipses). The original SHB population (upper left), the LSB population (lower right) and the third population fitted by Horvath (middle ellipse). Note that the SHB and LSB have no correlation of T90 with HR, but that the ‘third’ population has T90 anti-correlated with HR. Adapted from Horvath et al. (2006). Phil. Trans. R. Soc. A (2007) Downloaded from http://rsta.royalsocietypublishing.org/ on September 30, 2016 1284 S. D. Barthelmy Table 1. Short bursts detected by Swift-BAT. Question marks indicate uncertainty on that particular entry. name redshift afterglow host 050509B 050709b 050724 050813 050906 060614 050925d 051105A 051210 051221A 051227 060121e 060313 060502B 060505 060614 060801 061006 0.225 0.161 0.258 0.7 (?) 1.8 — 0.165? — — — 0.547 — 1.5 (?) 4.5 — — 0.089 0.125 1.131 — X X, O X, O, R X — late slew — — X X, O, R X X, O X, O X X, O X, O X X, O? Eiso/ (1050 erg) elliptical 0.01 SF galaxy 0.6 elliptical 3 — galaxyc galaxy? — 0.9? clusterc in galaxy plane — — 0.098 — cluster?c SF galaxy 0.2 — — galaxy — — cluster?c galaxy? — galaxy w0.5 dwarf galaxy 3.7 galaxy? w13 no bright galaxy — duration (s)a lag (ms) origin? 0.042 0.07 0.276 0.012 0.162 16 0.056 0.028 0.852 1.35 0.9 1.97 0.736 0.048 4.0 103(5) 0.5 130 4.3G3.2 0.0G2.0 K4.2G8.2 K9.7G14.0 — K20G20 — 6.3G5.3 K5.3G24.0 0.8G0.5 2G10? 2G29 0.3G0.7 K0.2G2.8 — 3G6 K8G8 5G3 NS–NS NS–NS NS–NS — — is it short? new SGR? — — — — — — — is it short? — — — a T90 in 30–350 (400) keV. bHETE GRB. cGalaxy in cluster. dSoft spectrum. eNot related to massive stars, consistent with NS–NS merger models. (a ) Redshift distribution Figure 4 shows the distribution of redshifts of bursts detected by BAT. The short bursts have a clearly lower distribution than the long bursts. This is an instrumental effect based on the lower fluence of soft photons available for the SHB and thus the reduced sensitivity of the BAT triggers. On a related topic to SHB and distance distributions, soft gamma-ray repeaters (SGRs) superflares are also short in duration. On 27 December 2004, BAT detected a superflare from SGR1806-20. Even though these SGR superflare events tend to be softer than SHBs (as are regular SGR outbursts), many authors have noted that some of the SHBs in the upper left cluster in figure 3 (at least the lower portion thereof) could be due to SGR superflares. This is an improbable situation for the BAT instrument because the 041227 superflare event, which is already 100 times more energetic than any of the other two superflares, could not be seen by BAT beyond 40 Mpc (Palmer et al. 2005). All of the detected Swift-BAT SHBs are well beyond the Virgo cluster of galaxies. (b ) Fast temporal variations in short hard burst Unlike many SHBs which have a single spike of emission in their lightcurves, figure 5 shows Swift-BAT GRBs 051221a and 060313, which have many separate, statistically significant peaks in their lightcurves. Many of these spikes have rise times less than 1 ms and a few have fall times less than 1 ms. Phil. Trans. R. Soc. A (2007) Downloaded from http://rsta.royalsocietypublishing.org/ on September 30, 2016 1285 Swift-BAT results and short bursts 10 Noccur long GRBs 5 diamonds: short GRBs 0 0.01 0.1 1 10 redshift Figure 4. Distribution of redshifts of BAT bursts with measured redshifts. The long bursts are shown in the red histogram. The SHBs are shown as the blue diamonds. This was adapted from Norris & Bonnell (2006). (a) (b) counts /1–ms bin 200 100 150 80 60 100 40 50 0 20 0.10 0.20 0.30 0.5 s 0.40 0.50 0 0.2 0.4 0.6 0.8 1.0 1.0 s Figure 5. Lightcurve for Swift-BAT (a) GRB 051221A and (b) GRB 060613 in the 15–350 keV band. Note the many peaks and the very fine temporal structure (some of the peaks have rise times less than 1 ms). These are mask-weighted lightcurves which means background has been subtracted. The error bars are approximately 5 counts sK1 for both lightcurves. (c ) Spectral lag Norris & Bonnell (2006) have done extensive work on the spectral lag properties of GRBs, in particular, the spectral lag as a tool for separating short from long bursts. Figure 6 shows the lag values for 260 BATSE bursts that have ‘zero lag’ (i.e. anything with a lag in the G20 ms range). Figure 6a shows that there is no correlation of lag with peak flux (i.e. neither correlation with an intrinsic brightness of the SHB nor with distance). Figure 6b collapses figure 6a in the flux dimension to produce a histogram of lags for SHB with the x -axis now in sigma units. The distribution is tightly confined around zero with only a small tail on the positive side (demarcated by the blue arrow). These extra bursts with small positive lag values are due to long bursts with extremely small lag values. Norris & Bonnell (2006) estimate that only one or two out of the 20 events seen in this tail are true SHB, yielding that approximately 18 are not true SHBs. The level of contamination is 7%. Phil. Trans. R. Soc. A (2007) Downloaded from http://rsta.royalsocietypublishing.org/ on September 30, 2016 1286 S. D. Barthelmy (b) 30 25 20 Noccur Fp (ph cm–2 s–1, 50–300 kev) (a) 50 10 15 10 5 2 –0.050 –0.025 0 0.025 0.050 lag (s): 25–50 keV, 100–300 keV 0 –10 –5 0 lag in sigmas 5 10 Figure 6. Spectral lag distribution for short GRBs (adapted from Norris & Bonnell (2006)). (a) The flux versus lag. The grey diamonds are 260 BATSE bursts with ‘zero lag’. The black and open diamonds are eight bursts from figure 9 (TTE and PREBCDISCSC data, respectively). (b) collapses (a) to make a histogram of the lag values with the x -axis now in sigma units. The arrow shows the long tail of positive lags which is a contamination of long bursts (7%). 4. Short hard bursts with extended emission There is a growing body of evidence that there is a class of events that is at least observationally different, if not intrinsically, from the SHB and LSB classical populations of GRBs. This putative new class starts out like a SHB (the initial short pulse), but then it has ongoing emission—well beyond the 2 s, or even few seconds, durations that have previously been defined as the duration limit for SHBs. This ongoing (extended) emission is softer than the initial emission of the initial spike. Using this observational-based definition, I will call these events SHBs with extended emission (SHBwEE). (a ) Swift-BAT SHBwEE Figures 7 and 8 show the lightcurves for GRBs 050724 and 060614, respectively. Swift-BAT has detected two other SHBwEEs: GRBs 051227 and 061006, plus 10 traditional SHBs (i.e. no extended emission above the BAT background level). This 4 out of 14 ratio is much higher than the 8 out of 260 ratio that Norris & Bonnell (2006) found in BATSE. This difference is probably due to the somewhat softer energy window that BAT has over BATSE, and that BAT has an image-trigger capability which BATSE did not (note that GRB 061006 was not detected by BAT using a rate trigger on the initial spike, but rather using an image trigger on the extended emission phase because the initial spike occurred during a spacecraft slew manoeuvre when all triggers are disabled). (b ) BATSE, KONUS and HETE SHBwEEs Figure 9 shows the lightcurves of eight SHBwEEs that Norris & Bonnell (2006) identified from the BATSE catalogue. All eight have zero lag (figure 6) Phil. Trans. R. Soc. A (2007) Downloaded from http://rsta.royalsocietypublishing.org/ on September 30, 2016 1287 Swift-BAT results and short bursts 15–25 keV (counts/s/det) 0.02 0.01 0 –50 0 50 100 time (s) 150 200 250 Figure 7. Lightcurve of Swift-BAT GRB 050724, a SHBwEE. The initial spike at TC0 s is suppressed in amplitude compared with the extended emission phase owing to the time binning. The extended emission starts to rise at approximately TC20 s, peaks around TC80 s and returns to instrumental background around TC200 s. 15–150 keV (counts/s/det) 2.5 2.0 1.5 1.0 0.5 0 0 100 50 time since BAT trigger (s) 150 Figure 8. Lightcurve of Swift-BAT GRB 060614, a SHBwEE. The extended emission starts to rise at approximately TC10 s, peaks around TC45 s and returns to instrumental background around TC150 s. The spectral lag for the spike phase is 9G6 and 3G6 ms for the extended emission phase. The apparent periodic structure in the extended emission is not statistically significant (Gehrels et al. 2006). and are therefore, in the Norris & Bonnell (2006) definition, true short bursts. These eight bursts were previously classified as long bursts by the BATSE team because they were using only T90 and HR; no spectral lag information was used. It is only recently that the spectral lag factor is being incorporated into the Phil. Trans. R. Soc. A (2007) Downloaded from http://rsta.royalsocietypublishing.org/ on September 30, 2016 1288 S. D. Barthelmy 105 104 103 1025 10 104 103 102 GRB 9210226 trigg 1997 GRB 9907126 trigg 7547 104 GRB 961225 trigg 5725 counts (s –1) 103 102 104 GRB 910709 trigg 503 103 102 104 GRB 920525 trigg 1625 103 102 104 GRB 931222 trigg 2703 103 102 GRB 000107 trigg 7934 103 102 GRB 951007 trigg 3853 103 102 0 10 20 30 40 50 time (s) 60 70 80 90 100 Figure 9. BATSE SHBwEE. Out of the 260 short bursts found by using the ‘zero-lag’ test on the BATSE catalogue of burst, these are eight that have extended emission and still have a zero-lag value. Adapted from Norris & Bonnell (2006). classification method. Given that there is a 7% chance of contamination (§3c), it is improbable that any of these eight are long bursts. Note that all eight have the same basic lightcurve envelope: they start with an initial intense and short spike (0.5–2 s), then a hiatus of emission, followed by the onset of the soft emission for 50–100 s. HETE GRB 050709 (figure 10) shows the identical temporal envelope. Figure 11 shows three KONUS bursts that are clearly in the same SHBwEE temporal and spectral envelope. Additionally, KONUS has six more bursts with nearly the same signature (Mazets et al. 2002)—the only difference is that there is no clear hiatus of emission immediately after the initial spike. Given that four instruments have seen these SHBwEEs and at least one BAT SHBwEE would have been classified as a traditional SHB (i.e. the extended emission would not have been detectable in BATSE; Barthelmy et al. 2005b), it seems probable that this is a new (sub)class of bursts. Phil. Trans. R. Soc. A (2007) Downloaded from http://rsta.royalsocietypublishing.org/ on September 30, 2016 1289 Swift-BAT results and short bursts WXM counts (s) (2–25 keV) 1400 1200 1000 800 600 400 200 0 0 100 time (s) 200 300 Figure 10. HETE SHBwEE. This is the lightcurve of GRB 050709 that was detected by HETE. This burst has the same profile of a SHB (the initial spike of intense spectrally hard emission, followed by a hiatus of emission, then increasing spectrally soft emission, peaking and then decreasing back to instrumental background). Note that in this plot, there is variable-width time binning to preserve the narrowness of the initial spike. (b) 3500 (c) G1+G2+G3 3500 3000 4500 2500 4000 2000 3500 1500 –20 0 20 40 60 –100 –50 G1+G2+G3 8500 coumts/2.048 s coumts/1.024 s G1+G2+G3 coumts/4.086 s (a) 8000 7500 7000 0 50 100 150 –100 –50 T–T0, s T–T0, s 0 50 100 150 T–T0, s Figure 11. KONUS SHBwEE. This shows three of the best examples of SHBwEE from the KONUS catalogue (Mazets et al. 2002). These three bursts also show the profile of initial spike, followed by a hiatus of emission, then increasing soft emission, peaking and then decreasing emission. (c ) The Horvath third class of bursts Horvath has performed fits in the HR–T90 plane with two-dimensional Gaussians (the three ellipses in figure 3) and finds a third population of burst to be 11% of the total. While the traditional SHB and LSB populations have no significant correlation between T90 and HR, his third population clearly shows an anticorrelation. The longer the burst (i.e. the larger the T90 value), the softer the HR. While Horvath states that he does not know of a reason for this anticorrelation, it seems reasonable that this third population is SHBwEE. The extended emission part of the burst is softer than the initial spike phase of the Phil. Trans. R. Soc. A (2007) Downloaded from http://rsta.royalsocietypublishing.org/ on September 30, 2016 1290 S. D. Barthelmy Table 2. BAT basic characteristics. item value energy range energy resolution location accuracy field of view detectors 12–300 keV 5 keV (at 60 keV) 1–3 arcmin (1s, depends on rising edge of burst) 2 sr 5200 cm2 of CZT, 32 768 elements GRB. For a given short burst, as the soft extended emission phase extends longer and longer, the total HR will be decreased. This is the same as the anticorrelation behaviour seen in the Horvath third population. (d ) Conclusions on short hard bursts We have shown that there is a sizable and distinct population of bursts whose lightcurves and spectral evolutions are different than the traditional short and long bursts. There are two differences. Firstly, they have an initial spectral hard and very short (less than a couple of seconds) emission phase which is identical to the SHBs. And secondly, they have an extended emission phase (lasting as much as 100–200 s) which is spectrally softer and is very close to the duration and spectral properties of the LSBs. Owing to this combination of properties of both the short and the long bursts, these SHBwEEs have been misidentified (owing to a lack of sufficient low-energy response) as to a type in previous missions (i.e. they were most commonly classified as LSB). This plus the results of Horvath (§4c) suggest that there is a third class or population of bursts (which we call SHBwEE). However, we cannot rule out the possibility that (i) there are still only two populations, (ii) all SHBs have extended emission, and (iii) the ratio of initial, hard emission to extended, soft emission has a large dynamic range such that most SHBs have such a low ratio that the extended emission is not detected by the past and the current missions. 5. Status of the Swift-BAT instrument The BAT on Swift is fully described in Barthelmy et al. (2005a) and the top-level instrumental parameters are shown in table 2. BAT has and continues to operate flawlessly since it was turned on and commissioned on 17 December 2004. Between then and 16 September 2006, there have been a total of 12.7 days (2% downtime) when BAT was not in a fully operational and triggerable state. This downtime was due to two computer crashes (both early in the mission and the cause has been fixed), two flight software upload sessions, a time when one of the two redundant loop heatpipes temporarily shutdown, and finally due to a handful of times when the spacecraft went into safehold state. In addition to these system downtimes, the BAT instrument is not triggerable to GRB (or hard X-ray transients) while in the SAA or while the spacecraft is in slew mode (each approximately 15% of the total time). There have been 173 GRBs detected, localized and distributed to the world Phil. Trans. R. Soc. A (2007) Downloaded from http://rsta.royalsocietypublishing.org/ on September 30, 2016 Swift-BAT results and short bursts 1291 community which yields an annual rate of 100G8 GRBs yrK1. Of these, approximately 10 per year are short GRBs. There have been slight increases due to two refinements in the triggering system. The first involved a change in the threshold levels in the short rate trigger criteria (4–32 ms duration triggers) and other trigger control parameters. The second involved a bug in the BAT flight software that allowed the long image triggers (5–43 min) to be enabled. Since launch, approximately 16% of the 32 768 CZT detectors have become permanently or semi-permanently noisy and had to be disabled. I wish to acknowledge and thank Jay Norris for many useful conversations that contributed significantly to this paper. I also wish to acknowledge and thank John K. Cannizzo and John D. Myers for their help in the preparation of this manuscript. References Barthelmy, S. D. et al. 2005a The burst alert telescope on the swift MIDEX mission. Space Sci. Rev. 120, 143–164. (doi:10.1007/s11214-005-5096-3) Barthelmy, S. D. et al. 2005b An origin for short GRBs unassociated with current start formation. Nature 438, 994–996. (doi:10.1038/nature04392) Gehrels, N. 2007 Swift observations of gamma-ray bursts. Phil. Trans. R. Soc. A 365, 1119–1128. (doi:10.1098/rsta.2006.1975) Gehrels, N. et al. 2006 A new gamma-ray burst classification scheme from GRB 060614. Nature 444, 1044–1046. (doi:10.1038/nature05376) Horvath, I. et al. 2006 A new definition of the intermediate group of gamma-ray bursts. Astron. Astrophys. 447, 23–30. (doi:10.1051/0004-6361:20041129) Kouveliotou, C. et al. 1993 Identification of two classes of gamma-ray bursts. Astrophys. J. 413, L101. (doi:10.1086/186969) Mazets, E. P. et al. 2002 Konus catalog of short GRBs (astro-ph/0209219). Norris, J. P. 1984 Frequency of fast, narrow gamma-ray bursts. Nature 308, 434–435. (doi:10.1038/ 308434a0) Norris, J. P. & Bonnell, J. T. 2006 Short gamma-ray bursts with extended emission. Astrophys. J. 643, 266–275. (doi:10.1086/502796) Palmer, D. M. et al. 2005 A giant gamma-ray flare from the magnetar SGR 1806–20. Nature 434, 1107–1109. (doi:10.1038/nature03525) Phil. Trans. R. Soc. A (2007)