astro-ph0512602/ms.tex
1: 
2: \documentclass[12pt,preprint]{aastex}
3: \usepackage{epsfig}
4: \usepackage{amssymb,amscd}
5: \usepackage{graphicx}% Include figure files
6: \usepackage{dcolumn}% Align table columns on decimal point
7: \usepackage{bm}% bold math
8: \usepackage{natbib}
9: 
10: \bibliographystyle{plainnat}
11: \begin{document}
12: \title{Quiescent times in Gamma-Ray-Bursts:\\ hints of a dormant inner engine}
13: 
14: \author{Alessandro Drago\altaffilmark{1,2} \& Giuseppe Pagliara\altaffilmark{3,2}}
15: 
16: 
17: 
18: 
19: 
20: 
21: \altaffiltext{1}{Dipartimento di Fisica, Universit{\`a} di Ferrara, 44100 Ferrara, Italy}
22: \altaffiltext{2}{INFN, Sezione di Ferrara, 44100 Ferrara, Italy}
23: \altaffiltext{3}{Dipartimento di Fisica, Politecnico di Torino, 10129 Torino, Italy}
24: 
25: 
26: 
27: \begin{abstract}
28: 
29: We perform a statistical analysis of the temporal
30: structure of long Gamma-Ray-Bursts (GRBs). 
31: First we consider a sample of bursts in which a long
32: quiescent time is present. Comparing
33: the pre-quiescence with the post-quiescence emission we show
34: that they display similar
35: temporal structures, hardness ratios and emitted powers, but,
36: on the average, the post-quiescence emission is roughly
37: twice as long as the pre-quiescence emission.
38: We then consider a sample of long and bright GRBs.
39: We show that the duration of each emission period
40: is compatible with the duration of an active 
41: period computed in various inner engine models.
42: At the contrary, if the inner engine is
43: assumed to be always active, i.e. also during the quiescent times, 
44: in several cases the total duration
45: of the burst largely exceeds the theoretical durations.
46: Our analysis therefore does not support the interpretation of long
47: quiescent times in terms of stochastic modulation of a continuous
48: wind. Instead the quiescent times can be interpreted as
49: dormancy periods of the inner engine. Before and after a dormancy period
50: the inner engine produces similar emissions. 
51: 
52: 
53: \end{abstract}
54: 
55: \keywords{gamma rays: bursts -- dense matter  }
56: 
57: \maketitle
58: 
59: \section{Introduction}
60: The time structure of GRBs is usually complex and
61: it often displays several short pulses separated by time intervals
62: lasting from fractions of second to several ten of seconds. The
63: analysis of the light curves can provide hints on the activity of the
64: inner engine although the relation between the observed signal and the
65: Physics of the inner engine is not yet completely understood.
66: 
67: A previous statistical analysis \citep{NP} has shown that there are
68: three time-scales in the GRB light curves: the shortest one is the
69: variability scale determining the pulses' durations and the intervals
70: between pulses; the largest one describes the total duration of the
71: bursts and, finally, an intermediate time scale is associated with long
72: periods within the bursts having no activity, the so called {\it
73: quiescent times}
74: \footnote{QTs are not the same as the
75: gaps between the precursors and the main pulses.  
76: In the case of the precursors, the gaps are between a softer
77: and weaker component (the precursor) and a harder and much stronger
78: one (the main burst). Here we observe gaps between pulses with the
79: same characteristics.}. The origin of these periods of quiescence is still unclear.
80: 
81: 
82: Here we show, through a statistical analysis, that if a quiescent time
83: longer than a few ten of seconds is present in the light curve then
84: the pre-quiescence and the post-quiescence emissions (PreQE, PostQE) have similar
85: variability scales but, on the average, the PostQE
86: is longer and only marginally softer than the PreQE. 
87: The similarities between the first and the second emission periods
88: strongly suggest that both emissions are produced by the same mechanism.
89: Moreover we will show that the average durations of PreQE and of PostQE, separately,
90: are compatible with the theoretical durations predicted by various inner enngine models. 
91: 
92: 
93: \section{Data analysis}
94: 
95: 
96: We have performed a statistical analysis of the time intervals $\Delta
97: t$ between adjacent peaks using the peak finding
98: algorithm of \citet{Li-Fe} and borrowing from \citet{NP1}
99: the definition of active periods separated by Quiescent Times (QTs).
100: In Fig.~\ref{grb} we show an example of burst where the previous
101: quantities are illustrated. We have applied this analysis to all the
102: light curves of the BATSE catalogue.
103: 
104: In a first investigation we have merged all the bursts of the
105: catalogue into one sample from which we compute the cumulative
106: probability $c(\Delta t)$ of finding time intervals $\Delta t$ which are not
107: QTs i.e.  we compute the distribution of the time intervals within each active period.  
108: In Fig.~\ref{rit}a, we show that $c(\Delta t)$ is well described by a
109: log-normal distribution. In Fig.~\ref{rit}b, the histogram of QTs is
110: displayed together with a log-normal distribution. As already observed
111: by previous authors \citep{NP}, there is an evident deviation of the
112: data points respect to the log-normal distribution for time intervals
113: longer than a few seconds, indicating an excess of long $\Delta t$.
114: In Fig.~\ref{rit}c we show a power law fit of the tail of the QTs
115: distribution which displays a very good agreement with the data, as
116: already observed by \citet{quilligan}.  The
117: physical interpretation of this distribution will be discussed later.
118: Finally, in Fig.~\ref{rit}d we show a correlation function, indicating
119: the probability of finding at least 2 QTs longer than $\Delta T$ in a
120: same GRB. As shown in the figure this probability rapidly decreases
121: and it essentially vanishes for $\Delta T > 40$ s.
122: 
123: 
124: A technical remark. In our analysis we have
125: considered two possible values for $\sigma$, the parameter used to
126: discriminate signal from noise.  A standard choice is
127: $\sigma=\sqrt{background}$.  Using that definition we can reproduce
128: the results of \citet{NP} on the cumulative distribution of $\Delta t$.  
129: Instead, the results presented in our figures have
130: been obtained using another
131: definition: $\sigma=\sqrt{\mathrm{Max}}$, where Max is the maximum number of
132: counts in the light curve. The reason for this choice is 
133: that we
134: want to study the main events of a GRB light curve (the ones
135: having a large luminosity) and not the faint
136: micro-structures (as for instance precursors, 
137: which we would exclude
138: from our sample).  It is important to remark that, using this larger
139: value of $\sigma$ we can recover the main results of 
140: Nakar and Piran, but the
141: distinction between a) time intervals which are not QTs and b) QTs 
142: is now even more clear on a
143: statistical basis: the firsts are perfectly interpreted by a
144: lognormal distribution, while long QTs are very well fitted by a
145: power-law, as discussed before.
146: 
147: 
148: 
149: 
150: 
151: 
152: 
153: 
154: 
155: We can now define a subsample of the BATSE catalogue composed of all
156: the bursts having a QT longer than 
157: $40$ s \footnote{The subsample contains the following 36 GRBs of BATSE catalogue:
158: 142, 222, 869, 1328, 1989, 2138, 2148, 2156, 2211, 2213, 2922, 3336, 3351, 3488, 3634,
159: 3776, 5421, 5478, 5486, 5585, 6295, 6335, 6454, 6472, 6629, 6745, 6892, 7170, 7185, 7301,
160: 7503, 7549, 7769, 8001, 8063, 8087.}.  In Fig.~\ref{rit}a we also
161: show the distribution of $\Delta t$ within the subsample.  The
162: distributions of the full BATSE catalogue and of the subsample are
163: essentially equal.  This indicates that the
164: subsample is not composed of bursts having an anomalously
165: large redshift because instead all time scales within the subsample
166: would be homogeneously dilated.
167: 
168: In our analysis we will now concentrate on the subsample.  From the
169: result of Fig.~\ref{rit}d, the bursts of the subsample contain only one
170: long QT and it is therefore possible to divide each burst into a
171: PreQE and a PostQE of which
172: we will compare the temporal and spectral structure.
173: 
174: 
175: In Fig.~\ref{twoemission} we display the cumulative distributions
176: $c_1(\Delta t)$ and $c_2(\Delta t)$ within each of the two emission
177: periods. In panel a we display only $\Delta t$ which are not QTs 
178: (same as in Fig.~\ref{rit}a),
179: while in panel b QTs are included.
180: In both cases the two distributions are very similar. The $\chi^2$-test
181: provides a significance of $28\%$ for panel a and of $34\%$ for 
182: panel b that the two data sets are drawn
183: from the same distribution function \footnote{The $\chi^2$-test has been performed
184: on the set of $\Delta t$ after a binning and not on the cumulative distribution.
185: We have checked that using binnings in the range 2--8 s the result of
186: the test 
187: remains stable.}.  
188: Let us remind that within the
189: internal-external-shocks model \citep{piran2,meszaros}, external shocks produce
190: emissions lacking the short time scale variability produced by
191: internal shocks \citep{saripiran}. The result of Fig.~\ref{twoemission}
192: rules out a scenario in which PostQE is dominated by external shocks
193: and PreQE by internal shocks.  This in turn excludes the possibility
194: of associating the QTs with the time needed to the jet to reach and
195: interact with the interstellar medium. Clearly enough, the statistical
196: analysis we are presenting does not rule out the existence of specific
197: GRBs in which the second episode is indeed associated with external
198: shocks. For instance GRB960530 and GRB980125 are examples of bursts in
199: which PostQE has a smoother morphology and a
200: softer spectral evolution than PreQE \citep{hakkila}.
201: 
202: We perform now a statistical analysis of the durations $D1$ and $D2$
203: of the two emission periods.  As shown in Fig.~\ref{durations}a, the
204: two data sets are well fitted by two log-normal distributions (the
205: Kolmogorov-Smirnov test provides a significance of $\sim 90\%$ ).  The
206: two distributions have different mean values ($D1_{ave}\sim 21s$,
207: $D2_{ave}\sim 41s$) 
208: and almost
209: identical standard deviations ($\sigma_1=36s$, $\sigma_2=33s $). We
210: address now the following question: is the longer duration of PostQE a
211: manifestation of a progressive increase in the active periods'
212: durations during the burst?  To answer this question we have
213: repeated the previous analysis by dividing PreQE and PostQE each in
214: two parts, using the longest QT
215: within each emission as a divider.
216: The distributions of the duration of all parts are
217: shown in Fig.~\ref{durations}b. The durations of the two parts within
218: each emission period share the same distribution (the $\chi^2$-test provides
219: significances larger than $50\%$ in both cases) but, in agreement
220: with the previous findings, the average durations of the two parts of
221: PostQE are longer than the two parts of PreQE. Therefore, the longer
222: duration of PostQE cannot be attributed to a continuous modification
223: of the emission but is a specific feature of the second part of the
224: GRB.
225: 
226: To estimate the emitted energy during PreQE and PostQE we have
227: analyzed the hardness ratios, defined as the ratios between the
228: photon counts in two BATSE channels (the second and the third in our
229: case).  The average hardness of PostQE turns out to be only marginally
230: smaller ($\sim 20\%$) than the average hardness of PreQE.  
231: Since, as shown above, the average durations of PreQE and of PostQE
232: are in the relation $D2_{ave}\sim 2 \, D1_{ave}$, the total energy emitted during PostQE 
233: is also about a factor of 2 larger than the one emitted during PreQE.
234: This is also evident from the result of Fig.~\ref{power}, where the distributions of 
235: the powers emitted during PreQE and PostQE are displayed, showing that 
236: the average powers of the two emissions are essentially the same. 
237:  
238: It is interesting to compare our result on the durations with the correlation between
239: duration of QT and of PostQE found by \citet{ramireza}. By computing
240: the average duration of QT and of PostQE in our subsample (which is 
241: not the same analyzed by those authors)
242: we obtain the following result: PostQE $\sim$ 41 s and 
243: QT $\sim$ 80 s.  
244: Therefore, in our sample and using the Nakar and Piran
245: algorithm we cannot confirm the correlation found by \citet{ramireza}.
246: Indeed, if that correlation was present
247: the average durations of QT and of
248: PostQE should be comparable, since the correlation law suggested by
249: those authors is a straight line with a slope of order one.
250: Actually, in our sample the statistical correlation 
251: between the durations of QT and of PostQE is very small, $r \sim $ 0.13.
252: 
253: \section{Discussion}
254: 
255: \subsection{Dormant engine scenario {\it vs.} wind modulation model}
256: 
257: 
258: 
259: As observed by \citet{ramirez},
260: within the internal shocks model it is possible to explain the QTs
261: either as a turn-off of the IE or as a modulation of a
262: continuous relativistic wind emitted by the IE
263: (Wind Modulation Model WMM). Both
264: hypothesis are consistent with the result of
265: Fig.~\ref{twoemission}.
266: 
267: The main difference between the WMM and the dormant engine scenario
268: is that in the WMM the inner engine has to provide a constant
269: power during the whole duration of the burst. 
270: In our subsample, we have several bursts whose total duration
271: (including the QT) approaches 300 s. 
272: These durations have to be corrected taking into account the average
273: redshift of the BATSE catalogue, $z_{ave} \sim 2 $ \citep{piran}, but
274: even after this renormalization, durations of a hundred seconds or more are not
275: too rare. 
276: This time scale has to be compared with the typical duration of the
277: emission period of the inner engine, as estimated in various models.
278: For instance, in 
279: all numerical investigations of the collapsar model \citep{woosley,woosley1}
280: the IE remains switched-on
281: during some 20s.  
282: There is at the moment no indication that a 
283: "steady state disk", characterized by a mass infall rate 
284: large enough to power the GRB, can last hundreds of seconds instead
285: of tens of seconds. Also in the quark deconfinement model
286: \citep{cheng,datta,ouyed,apj,haensel}
287: the inner engine remains active during periods of the order of
288: a few ten seconds corresponding to the cooling time of the
289: compact stellar object.
290: 
291: To better clarify our argument concerning both the duration of an
292: emission period and the energy injected by the inner engine, let us
293: first discuss an example of a bright and long burst, GRB \#7301
294: of the BATSE catalogue (see Fig.~\ref{b2}). In
295: that burst a large QT is present lasting $\sim$ 100 s followed by a
296: PostQE of
297: $\sim$ 70 s and with a PreQE of $\sim$ 40 s. To compare these
298: time-scales to the typical durations provided by the models for the inner engine
299: we have first to divide all the observed durations by a factor of 3, due to
300: the average redshift of the BATSE catalogue.  While the resulting
301: durations of PreQE and of PostQE are then in agreement with the typical
302: time duration of an emission period in the collapsar model or in the
303: quark deconfinement model ($\sim$ 20 s), the total duration of the burst
304: ($\sim$ 73 s, including the QT lasting $\sim$ 35 s), is roughly a factor of three larger
305: than the duration in the theoretical models. It is therefore
306: difficult to explain the long duration of the total burst within the
307: current inner engine models if the WMM is adopted.
308: 
309: Obviously, the previous remark would be meaningless if the
310: example we are providing is not typical of the time structure of
311: long GRBs. Instead, by checking all the GRBs of the 
312: "current catalogue" of BATSE, having T$_{90} >$ 100 s and peak photon flux
313: in the 256 ms channel $>$ 5 photons$^{-1}$cm$^{-2}$, we can come to
314: the following conclusion: all the lightcurves 
315: of this subsample can be interpreted within the dormant engine model
316: by rescaling
317: the total duration by a factor 1/(z+1) = 1/3 and by splitting the
318: burst in two emission episodes, separated by a 
319: dormancy period corresponding to a chosen QT.
320: In Figs.~\ref{b1},\ref{b2} we show all the GRBs of this sample
321: and we indicate the longest QT. 
322: Clearly enough, there are bursts
323: in which it is not unambiguous to decide which QT has to be interpreted as the
324: dormancy period, because more than one long QT is present. 
325: In the Figure we indicate the only two cases 
326: (out of 15 GRBs analyzed) in which
327: we suggest a dormancy period which is different from the
328: longest QT. In both those cases the suggested
329: dormancy period corresponds to the second longest QT present in the
330: burst. This rather small ambiguity can easily be explained by
331: noticing that a small fraction of long QTs can be generated by 
332: stochastic fluctuations (described by the tail of the 
333: log-normal distribution) and not by a dormancy
334: period. 
335: As discussed above,
336: the association of a QT with a dormancy period is
337: totally unambiguous for bursts in which a QT longer than
338: $\sim$ 40 s is present.
339: The last light curve presented in Fig.~\ref{b2} (GRB
340: $\#$ 6454) corresponds to a faint burst which does not
341: belong to the subsample discussed in this section
342: (GRBs of high luminosity).  It is
343: interesting to notice that in this burst PreQE and PostQE are so
344: long that also after dividing their durations by a redshift factor of
345: three, 
346: the durations are not compatible with the existing IE
347: models. Anyway, the faintness of this burst
348: is probably due to the extreme distance of the source and 
349: a larger red-shift correction should then be applied.
350: Similar considerations
351: are also valid for other superlong bursts presented in \citet{superlong}
352: (GRB $\#$ 6454 is one of the burst discussed in that paper). 
353: \footnote{
354: Using the algorithm of \citet{NP1} some 
355: faint components of the lightcurves are not considered as
356: parts of the signal because their counts number is smaller than 4-5
357: $\sigma$ and therefore they cannot be considered as active periods. 
358: This reflects on the estimated duration of the bursts.
359: In particular there are bursts whose $T_{90}$ durations exceed 100 s,
360: while their durations based on the active periods are shorter
361: and therefore these bursts are excluded from our sample.
362: This selection criteria clearly depends on the value of $\sigma$.
363: We have performed our analysis using for $\sigma$ both choices
364: described in Sec.~2. 
365: In Figs. 6-7 we show the results obtained using 
366: $\sigma=\sqrt{background}$, which is the most conservative choice
367: (in this case the excluded GRBs are $\#$ 1157, $\#$ 1886, $\#$ 3458, $\#$ 3930, $\#$ 7527). 
368: Using instead the larger value for $\sigma$, further 
369: components of the lightcurves would be excluded, and our 
370: conclusions would be even stronger.
371: }
372:  
373: 
374: 
375: Another problem with the WMM is due to the
376: prevision, within that model, that the emitted power during PostQE is
377: larger than during PreQE. This prevision is based again on
378: the existence of a continuous emission of shells also during the QT
379: \citep{ramirez}. The analysis displayed in Fig.~\ref{power} does
380: not support this prevision.
381: 
382: It is also worth recalling that, on the average, PostQE lasts roughly twice PreQE.
383: While it is probably possible to fix the parameters of
384: the WMM so to satisfy that constraint, no reason is
385: provided within the WMM model to explain that feature
386: in terms of the activity of the inner engine or
387: of the dynamics of the jet formation. 
388: At the contrary, models for a dormant inner engine can
389: associate the durations of the emission periods with
390: the durations of physical phenomena taking place within the
391: IE. This point will be discussed in the next Section.
392: 
393: We conclude that long QTs most probably correspond to periods of
394: inactivity of the IE.  On the other hand, it is possible
395: that the WMM is responsible for short QTs
396: occurring within the two emission periods, as suggested by
397: Fig.~\ref{durations}b where it is shown that the durations of the two parts of a single
398: emission period have the same distribution.
399: Finally, while our subsample is composed of GRBs having QTs longer
400: than 40 s, it is surely possible that in many cases the IE switches-off
401: for a much shorter time. Unfortunately, in those cases it is
402: not trivial to distinguish between QTs generated by the switch-off and
403: QTs generated by the WMM, as it also results from the discussion of 
404: Figs.~\ref{b1},\ref{b2}.
405: 
406: 
407: 
408: \subsection{Models for a dormant inner engine}
409: 
410: 
411: Let us now discuss how to generate dormancy periods using
412: various models of the IE.  Within the most
413: popular model, the collapsar model \citep{woosley}, there are two
414: possible scenarios: a temporary interruption of the jet produced by
415: Kelvin-Helmhotz instabilities \citep{woosley2} or the fragmentation of
416: the collapsing stellar core before its merging with the black hole \citep{osborne}. 
417: In both scenarios it can be possible to produce
418: long QTs.
419: For instance, in the scenario proposed by \citet{osborne}
420: the durations of the emission periods are related to the
421: durations of the accretion disks generated by each fragment.
422: Our result on the durations of PreQE and of PostQE can be explained
423: if, e.g.,
424: the average mass of the second fragment is larger than that of the 
425: first fragment. A discussion of the mass distribution of the fragments can be found 
426: in \citet{perna} \footnote{Disk fragmentation models could explain the 
427: recent X-ray-flare observations by Swift 
428: \citep{xray} which suggests the
429: possibility of a re-brightening of the IE.  In those cases,
430: however, the detected signal was softer than the GRB prompt emission
431: (the X-ray flares have not been detected by the BAT instrument on
432: Swift, but by the XRT). The re-brightening phenomena which we call
433: PostQEs belong instead to the prompt emission and they are only
434: marginally softer then the PreQEs. X-ray flares belong therefore to another class of phenomena.}.
435: 
436: Another model for the IE is based on the conversion of a
437: metastable hadronic star into a star containing quark
438: matter \citep{cheng,datta,ouyed,apj,haensel}. In the last years the 
439: possibility of forming
440: a diquark condensate at the center of a compact star has been widely
441: discussed in the literature \citep{raj}. The formation of a color
442: superconducting quark core can increase the energy released by a
443: significant amount \citep{noiprd}. It has also been shown that the
444: conversion from normal to gapped quark matter goes through a first
445: order transition \citep{shovk}. It is therefore tempting to associate
446: PreQE with the transition from hadronic to normal quark matter and
447: PostQE with the formation of the superconducting phase \citep{noi}.  In
448: this scenario the two dimensional scales regulating the durations of
449: PreQE and PostQE are the energies released in the two transitions.
450: Finally, let us remark that the power-law fitting the long QTs
451: distribution can originate from a superposition of exponential
452: distributions with different decay times \citep{flare}. In the quark
453: deconfinement model, the different decay times are associated with
454: slightly different masses of the metastable compact star.  
455: If the interpretation based on the quark deconfinement model
456: is correct then the GRBs data analysis provides very stringent bounds on the physics 
457: of high density matter.
458: 
459: 
460: A possible signature of the models in which the IE goes dormant
461: would be the detection of external shock emissions at the end of both 
462: PreQE and PostQE, indicating that the two emissions are physically
463: disconnected.  
464: 
465: 
466: \section{Conclusions}
467: 
468: In this paper we have studied the problem of quiescent times in long GRBs
469: borrowing the technique developed by Nakar and Piran.
470: The main results obtained in our analysis are the following:
471: 
472: \noindent
473: --- the cumulative distribution of time intervals between peaks 
474: within a same active period is 
475: well fitted by a lognormal distribution, while
476: long quiescent times are well fitted by a powerlaw distribution.
477: This strengthens \citet{NP} conclusion about the different origin
478: of long quiescent times respect to shorter time intervals;
479: 
480: \noindent
481: --- the two components of the emission, preceding 
482: and following the quiescent time are statistically similar from the viewpoint of
483: their temporal microstructure, their hardness ratio and emitted power.
484: These results suggest a unique mechanism at the origin of both
485: the pre-quiescence and the post-quiescence emission.
486: Interestingly, the second emission lasts on the average twice the first,
487: what provides an important constraint to the inner engine models;
488: 
489: \noindent
490: --- the durations of the activity periods of the inner engine, computed within the existing 
491: theoretical models do not exceed few ten seconds. These theoretical estimates
492: compare favorably with the durations of the pre-quiescent and of the 
493: post-quiescent emissions, separately. On the other hand, 
494: in several bursts the total duration
495: including the quiescent time, is roughly three times longer. 
496: This is an indication in favor of a dormant inner engine 
497: respect to wind modulation models.
498: 
499: 
500: 
501: \section{Acknowledgments}
502: 
503: 
504: It is a pleasure to thank Filippo Frontera, Enrico Montanari and Rosalba Perna for many useful
505: suggestions and for help in the data analysis.
506: 
507: 
508: \begin{thebibliography}{1}
509: 
510: 
511: \bibitem[Berezhiani et~al.(2003)]{apj}
512: Berezhiani, Z., Bombaci, I., Drago, A., Frontera, F. \& Lavagno, A., \newblock 2003,
513: Astrophys.J., 586, 1250
514: 
515: \bibitem[Bombaci \& Datta(2000)]{datta}
516: Bombaci~I. \& Datta~B., \newblock 2000,
517: Astrophys.J., 530, L69
518: 
519: \bibitem[Burrows et~al.(2005)]{xray}
520: Burrows, D.N. et~al. 
521: \newblock 2005, Science 309, 1833
522: 
523: \bibitem[Cheng \& Dai(1996)]{cheng}
524: Cheng, K.S.\& Dai~Z.G., \newblock 1996,
525: Phys.Rev.Lett., 77, 1210
526: 
527: \bibitem[Drago et~al.(2004)]{noiprd}
528: Drago, A., Lavagno, A. \& Pagliara, G., \newblock 2004,
529: Phys.Rev., D69, 057505
530: 
531: \bibitem[Drago et~al.(2005)]{noi}
532: Drago, A., Lavagno, A. \& Pagliara, G., \newblock 2005,
533: astro-ph/0510018, Proceedings QM2005, August 2005 Budapest.
534: 
535: \bibitem[Hakkila \& Giblin(2004)]{hakkila}
536: Hakkila, J. \& Giblin, T.W., \newblock 2004,
537: Astrophys.J., 610, 361
538: 
539: \bibitem[King et~al.(2005)]{osborne}
540: King, A., O'Brien, P.T., Goad, M.R., Osborne, J., Olsson, E. \& Page, K., \newblock 2005,
541: Astrophys.J., 630, L113
542: 
543: \bibitem[Li \& Fenimore(1996)]{Li-Fe}
544: Li, H. \& Fenimore, E., \newblock 1996,
545: Astrophys.J., 469, L115
546: 
547: \bibitem[MacFadyen \& Woosley (1999)]{woosley}
548: MacFadyen, A.L. \& Woosley, S.E., \newblock 1999,
549: Astrophys.J., 524, 262
550: 
551: \bibitem[Nakar \& Piran(2002{\natexlab{a}}))]{NP1}
552: Nakar, E. \& Piran, T., \newblock 2002{\natexlab{a}},
553: Mon. Not. Roy. Astron. Soc., 330, 920
554: 
555: \bibitem[Nakar \& Piran(2002{\natexlab{b}}))]{NP}
556: Nakar, E. \& Piran, T., \newblock 2002{\natexlab{b}},
557: Mon. Not. Roy. Astron. Soc., 331, 40
558: 
559: \bibitem[Ouyed \& Sannino(2002)]{ouyed}
560: Ouyed, R. \& Sannino, F., \newblock 2002,
561: Astron. Astrophys., 387, 725
562: 
563: \bibitem[Paczynski \& Haensel(2005)]{haensel}
564: Paczynski, B. \& Haensel, P., \newblock 2005,
565: Mon. Not. Roy. Astron. Soc. Lett., 362, L4 
566: 
567: \bibitem[Perna et~al.(2006)]{perna}
568: Perna, R., Armitage, P.J. \& Zhang, B., \newblock 2006,
569: Astrophys.J., 636, L29
570: 
571: \bibitem[Piran(1999)]{piran}
572: Piran, T., \newblock 1999, Phys.Rept., 314, 575
573: 
574: \bibitem[Piran(2004)]{piran2}
575: Piran, T., \newblock 2004, Rev.Mod.Phys., 76, 1143
576: 
577: \bibitem[Quilligan et~al.(2002)]{quilligan}
578: Quilligan, F., McBreen, B., Hanlon~L., McBreen~S., Hurley~K.J. \& Watson~D., \newblock 2002,
579: Astron. \& Astrophys., 385, 377
580: 
581: \bibitem[Rajagopal \& Wilczek(2000)]{raj}
582: Rajagopal, K. \& Wilczek, F., \newblock 2000,
583: hep-ph/0011333.
584: 
585: \bibitem[Ramirez-Ruiz et~al.(2001a)]{ramireza}
586: Ramirez-Ruiz, E., Merloni, A., \newblock 2001,
587: Mon. Not. Roy. Astron. Soc., 320, L25
588: 
589: \bibitem[Ramirez-Ruiz et~al.(2001b)]{ramirez}
590: Ramirez-Ruiz, E., Merloni, A. \& Rees, M.J., \newblock 2001,
591: Mon. Not. Roy. Astron. Soc., 324, 1147
592: 
593: \bibitem[Ruster et~al.(2005)]{shovk}
594: Ruster, S.B. , Werth, V., Buballa, M., Shovkovy, I.A. \& Rischke, D.H., \newblock 2005,
595: Phys. Rev. D73, 034025
596: 
597: \bibitem[Sari \& Piran(1997)]{saripiran}
598: Sari, R. \& Piran, T \newblock 1997, Astrophys.J., 485, 270
599: 
600: \bibitem[Tikhomirova \&  Stern(2005)]{superlong}
601: Tikhomirova , Y. \& Stern, B \newblock 2005, Astron. Lett. 31, 291
602: 
603: \bibitem[Wheatland(2000)]{flare}
604: Wheatland, M.S., \newblock 2000,
605: Astrophys. J., 536, L109
606: 
607: \bibitem[Woosley et~al.(2003)]{woosley2}
608: Woosley, S.E., Zhang, W. \& Heger~A., \newblock 2003,
609: AIP Conference Proceedings, 662, 185
610: 
611: \bibitem[Zhang \& Meszaros(2004)]{meszaros}
612: Zhang, B. \& Meszaros, P., \newblock 2004, Int.J.Mod.Phys., A19, 2385
613: 
614: \bibitem[Zhang et~al.(2003)]{woosley1}
615: Zhang, W., Woosley, S.E. \& MacFadyen, A.I.., \newblock 2003,
616: Astrophys. J., 586, 356
617: 
618: 
619: 
620: 
621: 
622: \end{thebibliography}
623: 
624: \begin{figure}
625: \begin{center}
626: \includegraphics[scale=1]{f1.epsi}
627: \end{center}
628: \caption{{\bf  Light curve of a typical GRB} Time profile of BATSE
629: burst $\# 5486$ in the energy range $55 Kev <E< 320 Kev $. For each
630: light curve, the background is first determined using a linear fit and then
631: subtracted from the data. The signal is extracted taking as
632: initial and final bins the ones with counts above $2\sigma$.  
633: The active periods are defined as periods in which 
634: bins with counts exceeding 4$\sigma$ are present.
635: Active periods begin (dark-gray dots) and end (light-gray dots) when the signal 
636: drops below 2$\sigma$.
637: Within each active period we search for peaks using the peak
638: finding algorithm of \citet{Li-Fe}: indicating with $C_p$ the counts of a
639: candidate peak and with $C_1$ and $C_2$ the counts in the bins to the
640: left and to the right of the candidate, a ``true'' peak must satisfy the
641: relations $C_p -C_{1,2} \geq N_{var} \sqrt{C_p}$, where $N_{var} = 5$ in our
642: analysis. The true peaks are indicated by the black points.
643: \label{grb}}
644: \end{figure}
645: 
646: 
647: 
648: 
649: 
650: 
651: 
652: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
653: 
654:  \begin{figure}
655:     \begin{centering}
656: \hbox{\hskip-0.cm \epsfig{file=f2a.epsi,height=4.6cm}\hskip 1.cm 
657: \epsfig{file=f2b.epsi,height=4.6cm}}
658: \vskip 1.cm    
659: \hbox{\hskip 0.3cm
660: \epsfig{file=f2c.epsi,height=4.6cm}
661: \hskip 1.6cm \epsfig{file=f2d.epsi,height=4.6cm}}
662:     \caption{{\bf  Analysis of time intervals between peaks} 
663: {\bf a} The cumulative distribution of time intervals $\Delta t$ which are not QTs
664: (black triangles), is compared with its best fit log-normal distribution (solid
665: green line). The data come from the full BATSE sample after
666: eliminating bursts displaying data gaps. The boxes correspond to
667: the cumulative distribution of $\Delta t$ taken from the subsample of
668: bursts containing a QT longer than $40$s (see text). {\bf b} Histogram
669: of the QTs and its log-normal fit (dashed line).
670: {\bf c} Histogram of QTs and power-low fit of its tail (dashed line).
671: The fit is based only on QTs longer than $40$s.
672: {\bf d} Frequency of bursts containing at least two QTs longer than $\Delta T$.
673: \label{rit} }
674:    \end{centering}
675:    \end{figure} 
676: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
677: 
678: 
679: \begin{figure}
680: \begin{center}
681: \includegraphics[scale=1]{f3.epsi}
682: \end{center}
683: \caption{{\bf Analysis of time intervals between peaks within the two emission periods} 
684: {\bf a} The cumulative distribution of time intervals $\Delta t$ which are not QTs
685: are shown for the two emission episodes, PreQE and PostQE.
686: {\bf b} The cumulative distributions of $\Delta t$ are shown for the two emission episodes including QTs. 
687: \label{twoemission} }
688: \end{figure}
689: 
690: 
691: 
692: \begin{figure}
693: \begin{center}
694: \includegraphics[scale=1]{f4.epsi}
695: \end{center}
696: \caption{{\bf Analysis of the durations $D$ of the two emission periods}
697: {\bf a} Cumulative distributions of durations of PreQE (filled triangles) and of PostQE
698: (filled boxes) and their best-fit log-normal distributions (dotted lines).
699: {\bf b} Cumulative distributions of durations of the first and second part 
700: of PreQE (empty triangles and boxes) and of PostQE (filled triangles and boxes).
701: \label{durations} }
702: \end{figure}
703: 
704: 
705: 
706: \begin{figure}
707: \begin{center}
708: \includegraphics[scale=1]{f5.epsi}
709: \end{center}
710: \caption{{\bf Analysis of the powers $P$ of the two emission periods}
711: Cumulative distributions of powers of PreQE (filled triangles) and of PostQE (filled boxes)
712: \label{power} }
713: \end{figure}
714: 
715: 
716: 
717: \begin{figure}
718: \begin{center}
719: \includegraphics[scale=0.9]{f6.epsi}
720: \end{center}
721: \caption{{\bf Long and bright bursts  }
722: GRBs of the "current catalogue" of BATSE, having T$_{90} >$ 100 s and peak photon flux
723: in the 256 ms channel $>$ 5 photons$^{-1}$cm$^{-2}$. The dashed lines indicate the fit of
724: the background and the 2$\sigma$ and 5$\sigma$ levels used to define the active periods.
725: Dark-gray dots and light-gray dots mark the initial and the
726: end point of each active period, respectively.
727: Black arrows indicate the longest QT.
728: Gray arrows indicate the suggested dormancy period when
729: it does not coincide with the longest QT (two cases only).   
730: \label{b1} }
731: \end{figure}
732: 
733: 
734: \begin{figure}
735: \begin{center}
736: \includegraphics[scale=0.9]{f7.epsi}
737: \end{center}
738: \caption{{\bf Long and bright bursts}
739: Same as in Fig.~\ref{b1}. 
740: Here the suggested dormancy period
741: always coincide with the longest QT. We also display
742: a faint burst $\#$ 6454 (it does not satisfy the
743: limit on the peak photon flux) whose very long
744: duration could be due to a large red-shift.
745: \label{b2} }
746: \end{figure}
747: 
748: 
749: 
750: \end{document}
751: 
752: 
753: 
754: 
755: 
756: 
757: 
758: 
759: 
760: