1: \documentclass[12pt,preprint]{aastex}
2:
3:
4: %\renewcommand{\baselinestretch}{1}
5: \shorttitle{GRBs: Restarting the Engine}
6: \shortauthors{Andrew King et al.}
7:
8: \begin{document}
9: \def\grb{{GRB050502b}}
10: \def\lsun{{\rm L_{\odot}}}
11: \def\msun{{\rm M_{\odot}}}
12: \def\rsun{{\rm R_{\odot}}}
13: \def\go{
14: \mathrel{\raise.3ex\hbox{$>$}\mkern-14mu\lower0.6ex\hbox{$\sim$}}
15: }
16: \def\lo{
17: \mathrel{\raise.3ex\hbox{$<$}\mkern-14mu\lower0.6ex\hbox{$\sim$}}
18: }
19: \def\simeq{
20: \mathrel{\raise.3ex\hbox{$\sim$}\mkern-14mu\lower0.4ex\hbox{$-$}}
21: }
22:
23:
24: %\input epsf.sty
25: %\input psbox.tex
26:
27: \title{Gamma-ray bursts: Restarting the Engine}
28: \author{Andrew King, Paul T. O'Brien, Michael R. Goad, Julian Osborne,\\
29: Emma Olsson, Kim Page}
30: \affil{Department of Physics \& Astronomy,
31: University of Leicester, Leicester LE1 7RH, UK}
32:
33: %\maketitle
34: \begin{abstract}
35: Recent gamma--ray burst observations have revealed late--time, highly
36: energetic events which deviate from the simplest expectations of the
37: standard fireball picture. Instead they may indicate that the central
38: engine is active or restarted at late times. We suggest that
39: fragmentation and subsequent accretion during the collapse of a
40: rapidly rotating stellar core offers a natural mechanism for this.
41: \end{abstract}
42: \keywords{Gamma Rays: bursts --- black hole physics ---
43: accretion disks}
44:
45: \section{Introduction}
46:
47: It is now widely believed that gamma-ray bursts (GRBs) lasting longer than
48: about 1~s are produced by a class of supernovae, called hypernovae
49: (Paczy\'nski, 1998; MacFadyen \& Woosley, 1999). Collapse of a massive,
50: rapidly-rotating stellar core is assumed to lead to the formation of a black
51: hole, while the remaining core material has enough angular momentum to form a
52: massive accreting neutron torus around it. The infalling torus radiates its
53: gravitational energy as neutrinos or converts it directly to a beam by MHD
54: processes. This evacuates the rotation axis of the core, allowing both an
55: observable burst of gamma rays and the expulsion of a jet of matter at high
56: Lorentz factors. The interaction of this jet with the surroundings produces
57: the X-ray afterglow, which has proved so useful in locating and studying
58: gamma-ray bursts.
59:
60: In all cases observed until now, once the initial gamma-ray emission
61: has faded away the energy output is dominated by the afterglow which
62: has far lower total energy than the GRB. However, the recent \grb \
63: is quite different (Burrows et al. 2005). An X-ray flare occurred
64: starting some 400 s after the beginning of the GRB, and released at
65: least as much energy as the original burst. There have been several
66: efforts to explain \grb \ by adding elements to the standard picture
67: of long GRBs such as late internal shocks, or synchrotron
68: self--Compton emission (Burrows et al. 2005; Kobayashi et
69: al. 2005). However there is considerable difficulty in understanding
70: how \grb \ can put so much energy into its surroundings at such late
71: times, and what physical parameter specifies this very different
72: behaviour.
73:
74: Here we consider an idea which offers a way out of these
75: difficulties. In effect it offers a physical model for the late
76: internal shock suggested by Burrows et al. (2005).
77:
78: \section{Observations}
79:
80: GRB050502b was detected by the Burst Alert Telescope (BAT) on the {\it
81: Swift} satellite (Gehrels et al. 2004) at 09:25:40 UT on 2005 May 2
82: (Falcone et al. 2005). The burst lasted 17.5 s (T90) with a fluence in
83: the 15--350 keV band of $(8\pm 1) \times 10^{-7}$ erg cm$^{-2}$
84: (Cummings et al. 2005). The gamma-ray spectrum is well fitted by a
85: single power law with a photon index, $\Gamma = 1.6\pm 0.1$. Falcone
86: et al. reported that initially the X-ray count rate seen by {\it
87: Swift} was too low to obtain an on-board centroid position with the
88: {\it Swift} X-ray Telescope (XRT) in the 0.3--10 keV band, but
89: subsequent ground-processing revealed an initially fading X-ray
90: source. This X-ray source brightened dramatically starting some 400 s
91: after the initial burst, rising to a peak which lasted for several
92: hundred seconds before fading away (Burrows et al. 2005).
93:
94: Both GRB050406 (Cummings et al. 2005) and GRB050607 (Krimm et al.
95: 2005) show similar behaviour to GRB050502b, in that they have a
96: re-brightening of their X-ray emission a few hundred seconds after the
97: initial burst. In both cases, however, the X-ray flares are much
98: weaker both relative to their respective bursts and to the bright
99: flare seen in GRB050502b. All three of these bursts are relatively low
100: in GRB fluence, particularly GRB050406, but are not unusually
101: faint. Here we report an analysis of all three GRBs, but concentrate
102: on GRB050502b. As we do not know the redshifts of these bursts, all
103: physical parameters are quoted in the observer's frame of reference.
104:
105: The spectrum of the X-ray flare in GRB050502b can be well fitted by a
106: power law with $\Gamma=2.2\pm 0.03$ but with a time-dependent
107: absorbing column. Similar evidence for spectral variability is seen
108: during the flares for the other GRBs, but cannot be well constrained.
109: In \grb \ the column is highest at the start of the flare, (${\rm N_H}
110: \approx 1\times 10^{21}$ cm$^{-2}$ above Galactic)
111: and then decreases strongly suggesting ionization (Burrows et al.
112: 2005). This material was not ionized by the GRB so presumably
113: is off the line of sight illuminated by the initial jet.
114: The absorbing column was derived assuming zero redshift so the
115: intrinsic column will be higher. To derive the X-ray fluence we
116: performed time-dependent spectral fits for GRB050502b. For the other
117: two bursts we assumed an intrinsic absorbing column of ${\rm N_H} = 1\times
118: 10^{21}$ cm$^{-2}$ with $\Gamma = 2.4$ and 2.3 for GRB050406 and
119: GRB050607 respectively.
120:
121: In Table 1 we quote fluences in the original detector bandpass and
122: that derived by extrapolating the power law spectra over the 0.3--350
123: keV bandpass (all bandpasses quoted are in the observer's frame).
124: Extending the energy range to lower energies (e.g. 0.1 keV), would
125: further enhance the relative strength of the X-ray flare in GRB050502b
126: (e.g. it would be 30\% brighter if a low energy cutoff of 0.1 keV were
127: adopted).
128:
129: Our spectral analysis of the X-ray flare in GRB050502b shows that the
130: total fluence of the flare is comparable to or higher than that of the
131: initial GRB (Table 1). This is also illustrated in Figure 1 where we
132: show the flux light curve derived for the XRT bandpass (0.3--10 keV).
133: For the other GRBs the fluence of the X-ray flare is $\le 10$\% of the initial
134: burst.
135:
136:
137: \begin{deluxetable}{lccccl}
138: %\tablewidth{0pc}
139: \tablecaption{GRB fluences for the initial GRB and the X-ray flare}
140: \tablecolumns{6}
141: \tablehead{
142: \colhead{GRB} &
143: \colhead{Burst} & \colhead{Flare} &
144: \colhead{Burst} & \colhead{Flare} & \colhead{References} \\
145: \colhead{} & \colhead{($10^{-7}$ erg cm$^{-2}$)} &
146: \colhead{($10^{-7}$ erg cm$^{-2}$)} &
147: \colhead{($10^{-7}$ erg cm$^{-2}$)} &
148: \colhead{($10^{-7}$ erg cm$^{-2}$)} & \colhead{}\\
149: \colhead{} & \colhead{15--350 keV} & \colhead{0.3--10keV} &
150: \colhead{0.3--350 keV} & \colhead{0.3--350keV} & \colhead{}
151: }
152: \startdata
153: %\hline
154: 050502b & 8.0 & 8.8 & 11 & 14 & 1, 2\\
155: 050406 & 9.0 & 0.9 & 57 & 1.1 & 1, 3 \\
156: 050607 & 8.9 & 1.1 & 17 & 1.5 & 1, 4\\
157: \enddata
158: \tablerefs{References --- (1) This paper; (2) Cummings et al. 2005;
159: (3) Krimm et al. 2005; (4) Retter et al. 2005}
160: \end{deluxetable}
161:
162:
163: \begin{figure}
164: \centerline{\includegraphics[scale=0.6,angle=270]{f1.eps}}
165: %\epsscale{0.2}
166: %\plotone{f1.eps}
167: \caption{
168: The unabsorbed 0.3-10 keV flux light curve for the early phase of
169: GRB050502b. The horizontal bars represent the time intervals over
170: which the fluxes were calculated. The first point is the initial GRB
171: extrapolated to the 0.3-10 keV band.}
172: \end{figure}
173:
174: \section{Energetics}
175:
176: We now consider how the central engine in a GRB might restart.
177: Our line of argument is straightforward. The original GRB in 050502b
178: signalled the formation of a compact object (neutron star or black
179: hole) and the accretion of a stellar mass on to it on a very short
180: timescale, presumably from a neutron torus. The simplest explanation
181: for the comparable energy of the X-ray flare is that a second, similar
182: mass accreted after a delay $t_d$ of up to 400~s (allowing for
183: possible light-travel effects). Clearly this large mass cannot have
184: already been in the first torus or it would have been accreted along
185: with it. Thus a second stellar-mass torus must form and accrete after
186: $t_d$.
187:
188: In 2D models of fallback the extra mass is already in a torus, by
189: construction, and the delay between the GRB and X--ray flare in \grb \
190: would have to reflect some viscous angular momentum transport
191: timescale before the torus became dynamically unstable and rapidly
192: accreted. In contrast if we do not assume high symmetry, an obvious
193: possibility is that this mass is a second compact `star' (i.e. a
194: self-gravitating neutron lump), formed because the collapsing core had
195: enough angular momentum to fragment. Gravitational radiation then
196: drags the lump in towards the first compact object where tides smash
197: it up into a torus which can be accreted, releasing an energy
198: comparable to the original GRB. This effectively amounts to
199: restarting, or re-activating, the central engine at late times.
200:
201: Re-starting the GRB central engine also accounts for the
202: extraordinarily rapid rate of increase and decrease in the X-ray flare
203: which goes as greater than $t^7$ if we use the time ($t$) of the
204: initial burst as the origin $t=0$. Such rates are hard to accommodate
205: in the standard fireball model for GRB afterglows and instead strongly
206: support a re-activated central engine. The BAT on {\it Swift} did not
207: detect gamma-rays from the X-ray flare in GRB050502b, possibly because
208: the flare spectrum was soft and the energy was spread over many
209: seconds.
210:
211: %Alternatively the optical depth
212: %$\tau_{\gamma - \gamma}$ for gamma-ray annihilation may have been too
213: %large, or the Lorentz factor of the second outflow lower.
214: %
215: %All the energy must end up in an explosive outflow and be
216: %radiated when this hits some of the debris caused by the first event.
217: %This must be more massive than usual as stellar matter is added to the
218: %swept-up interstellar medium involved in the first afterglow. As a
219: %result the outflow is stopped fairly quickly, and radiates its very
220: %large luminosity in a much brighter afterglow.
221:
222:
223: \section{Core collapse and fragmentation}
224:
225: The two-stage collapse described above is almost exactly that
226: envisaged by Davies et al. (2002) to explain why hypernovae are rare
227: among supernovae, and any possible delays between supernovae and
228: gamma-ray bursts. Davies et al. (2002) argued by comparison with
229: theoretical studies of star formation (cf. Bonnell \& Pringle, 1995),
230: that the collapse of a rapidly rotating stellar core was likely to
231: lead to fragmentation and the initial formation of more than one
232: compact object of nuclear density. They considered a case where the
233: collapses producing these objects make a supernova rather than an
234: initial gamma-ray burst, i.e. where none of them is surrounded by an
235: accreting torus. Depending on the way fragmentation occurs, cases
236: where the initial collapse produces a GRB are clearly possible.
237:
238: The compact objects subsequently coalesce under the effect of
239: gravitational radiation. Provided that at least one of them is a
240: neutron star rather than a black hole, Davies et al. show that there
241: must be a stage where a neutron torus accretes on to a central object,
242: thus releasing an energy comparable to the initial burst. Eq. (7) of
243: that paper shows that gravitational radiation emission drags an
244: orbiting fragment in, to the point where it disrupts and forms a
245: neutron torus, on a timescale
246: \begin{equation}
247: \tau_{\rm GR} = 0.18\times {(a_0/1000~{\rm km})^4\over
248: m_1m_2(m_1+m_2)}~{\rm hours.}
249: \label{gr}
250: \end{equation}
251: Here $a_0$ is the initial circular orbit radius and the central black
252: hole and orbiting fragment have masses $m_1\msun, m_2\msun$
253: respectively. This
254: gives a gravitational radiation delay timescale $\tau_{\rm
255: GR} = 0.18$~hr = 640 s if the merging masses are $\sim 1\msun$ and
256: their initial separation is $a_0= 1000$~km, which in turn requires a
257: specific angular momentum $j = 10^{17}$ cm$^2$ s$^{-1}$ --- exactly
258: what hypernova models require. This will be true for any required GR
259: delay (which is also affected by the redshift) since $j \propto
260: a^{1/2} \propto \tau_{\rm GR}^{1/8}$. One can arrange the
261: gravitational radiation delay to be shorter than the observed delay if
262: the X--ray event corresponds to a second faster outflow overtaking the
263: initial one, although a substantial overtaking delay is rather
264: contrived.
265:
266: The initial orbit of the fragment may be somewhat misaligned from the
267: spin axis of the central black hole. Tidal dissipation in the orbiting
268: fragment and viscous dissipation in the neutron torus cause rapid
269: alignment through the Lense--Thirring effect (cf. King et al.,
270: 2005). The second burst (X--ray flare) thus has a jet axis close to
271: the first one. This may explain why the flare is spectrally softer and
272: lasts longer than the original gamma--ray burst. In the internal shock
273: picture, the peak spectral energy scales as $E_p \propto
274: \Gamma^{-2}t_v^{-1}$ (e.g. Table 2 of Zhang \& Meszaros, 2002), where
275: $\Gamma$ is the bulk Lorentz factor and $t_v$ the variability
276: timescale of the outflow. The cleaner environment for the second jet
277: may reduce its baryon loading and thus increase $\Gamma$. It is also
278: possible that tidal effects make the second accretion event smoother
279: than the first (increasing $t_v$), although a full hydrodynamical
280: calculation is needed to check this.
281:
282: \section{Smaller X-ray flares}
283:
284: Several other GRBs have been observed to have significant X-ray flares
285: superimposed on their fading afterglows (see Table 1 and Piro et al.
286: 2005), although none are nearly as bright or quite as late as seen in
287: \grb. \grb \ itself has another, smaller X-ray event $\approx 10^5$ s
288: after the burst (Burrows et al. 2005) with a similar fluence to that
289: of the event in GRB050607 (some 10\% of the initial GRB). These
290: smaller flares may be evidence of a similar process involving the
291: accretion of smaller neutron clumps by the central object: for a
292: sufficiently complex fragmentation process there may even be several
293: events. A lower limit to the relative luminosity of such smaller
294: events to the main GRB arises from the result (Davies et al., 2002)
295: that the minimum mass of a neutron-rich clump is $\sim 0.2\msun$
296: (smaller masses make nuclei in their centres). The energies of a
297: merger event to the original GRB must be in the ratio $r =
298: \eta_2M_2/\eta_1M_1$, where $M_1, M_2$ are the masses of the first
299: compact object and the subsequent merging object, and $\eta_1, \eta_2$
300: are the efficiencies of the collapse and merger events respectively.
301: For comparable efficiencies a merger must have $r\ga 0.2\msun/M_1 \ga
302: 0.02$ (for $M_1 \la 10\msun$). The observed X-ray/GRB ratio depends on
303: the spectral index of the emitted flux.
304:
305:
306: \section{Discussion}
307:
308: The observed fluence of the late X-ray flare in \grb \ is comparable
309: to that of the main GRB and is spectrally softer. Smaller, but
310: possibly related, flares have also been seen in other bursts. Whether
311: the unusually bright X-ray flare event in GRB050502b is unique or
312: simply and extreme example of a pattern of behaviour common to GRBs is
313: unclear. For example the time of the late X-ray flare in GRB050502b is
314: within the known range in GRB durations (e.g. Paciesas et al. 1999)
315: which extends to at least a thousand seconds. Many bursts, but not
316: all, show spectral evolution such that they become softer later
317: (e.g. Norris et al. 1986). It has also been shown that if a GRB engine
318: is quiescent for a long time the subsequent emission outburst could be
319: unusually large (Ramirez-Ruiz and Merloni 2001). The fragmentation
320: picture we have suggested here may be able to explain these
321: phenomena. Continuing accretion at a slower rate from a fragment may
322: be the origin of the long, faint X--ray afterglows seen in many {\it
323: Swift} GRBs, which probably require the central engine to remain on
324: for a long time (cf Zhang \& Meszaros 2001). We will investigate this
325: possibility in a future paper.
326:
327: We have suggested that X-ray flares may result from the fragmentation
328: of the collapsing stellar core, and the subsequent merger of a
329: significant fragment with the most massive compact object formed in
330: the collapse. This is a departure from the current hypernova picture,
331: in which only one object is assumed to form directly in the
332: collapse. However the rapid rotation apparently required to make a
333: hypernova is known to lead to fragmentation in other situations such
334: as star formation. A simple test of our idea will come from combining
335: gamma-ray burst observations with gravitational-wave detections by
336: LIGO. Long GRBs should be significant sources of gravitational
337: radiation, with a chirp signal characteristic of a merging binary
338: system. If this proves successful it would also give a clean
339: determination of the masses involved, providing major insight into the
340: process of core collapse in rapidly rotating stars.
341:
342:
343:
344: \paragraph*{Acknowledgments}
345:
346: We thank Evert Rol for useful discussions and the referee for a very
347: helpful report. ARK acknowledges a Royal Society -- Wolfson Research
348: Merit Award. Emma Olsson is supported by a European Union Research
349: and Training Network grant. The {\it Swift} Project and theoretical
350: astrophysics research at the University of Leicester are supported by
351: PPARC.
352:
353:
354: \begin{thebibliography}{}
355: \item{}
356: Bonnell, I.A., Pringle, J.E., 1995, MNRAS, 285, L12
357: \item{}
358: Burrows, D.N. et al., 2005, Science, submitted
359: \item{}
360: Cumming, J. et al., 2005, GCN Circular 3339
361: \item{}
362: Davies, M.B., King, A.R., Rosswog, S., Wynn, G.A., 2002, ApJL, 579, L63
363: \item{}
364: Falcone, A. et al., 2005, GCN Circular 3330
365: \item{}
366: Gehrels, N. et al., 2004, ApJ, 611, 1005
367: \item{}
368: King, A.R., Lubow, S.L., Ogilvie, G.G., Pringle, J.E., 2005, MNRAS in
369: press (astro--ph/0507098)
370: \item{}
371: Krimm, H. et al., 2005, GCN Circular 3183
372: \item{}
373: Kobayashi, S., Zhang, B., Meszaros, P., Burrows, D.N., 2005, ApJ,
374: submitted
375: \item{}
376: MacFadyen, A.L., Woosley, S.E., 1999, ApJ, 524, 262
377: \item{}
378: Norris, J.P. et al., 1986, ApJ, 301, 213
379: \item{}
380: Paciesas, W.S. et al., 1999, ApJS, 122, 465
381: \item{}
382: Paczynski B., 1998, ApJ, 494, L45
383: \item{}
384: Piro, L. et al., 2005, ApJ, 623, 314
385: \item{}
386: Ramirez-Ruiz, E., Merloni, A., 2001, MNRAS, 320, L25
387: \item{}
388: Retter, A. et al. 2005, GCN Circular 3525
389: \item{}
390: Zhang, B., Meszaros, P., 2001, ApJ, 552, L35
391: \item{}
392: Zhang, B., Meszaros, P., 2002, ApJ, 581, 1236
393: \end{thebibliography}
394: \end{document}
395:
396:
397:
398: