1: \documentclass{emulateapj}

2: % \documentclass[12pt,manuscript]{aastex}

3: % \documentclass[12pt,preprint]{aastex}

4:  

5: \def\ssp{\baselineskip=11pt plus 1pt minus 1pt}

6: \def\dsp{\baselineskip=22pt plus 1pt minus 1pt}

7: \def\rf{\hfill\break}

8: \def\sm{M_{\odot}}

9: \def\ppp{\par \smallskip \noindent \hangindent .5in \hangafter 1}

10: \def\e{\eta}

11: \def\pt{\ \ \ \ .}

12: \def\com{\ \ \ \ ,}

13: \def\rt{\it T}

14: \def\ir{\it R}

15: \def\g{\gamma}

16: \def\G{\bar \gamma}

17: \def\s{\sigma}

18: \def\a{\alpha}

19: \def\b{\beta}

20: \def\ap{\approx}

21: \def\p{\partial}

22: \def\t{\theta}

23: \def\T{\tau}

24: \def\D{\Delta}

25: \def\d{\delta}

26: \def\r{\rho}

27: \def\st{\sigma _T}

28: \def\bl{\bar \lambda}

29: \def\nep{\nu ^{'}}

30: \def\pt{\ \ \ \ .}

31: \def\com{\ \ \ \ ,}

32: \def\cl{\centerline}

33: \def\sles{\lower2pt\hbox{$\buildrel {\scriptstyle <}

34:    \over {\scriptstyle\sim}$}}

35: \def\sgreat{\lower2pt\hbox{$\buildrel {\scriptstyle >}

36:    \over {\scriptstyle\sim}$}}

37: \def\sg{r^2}

38: \def \ar{$\Longrightarrow$}

39: \def \bc{\begin{center}}

40: \def \ec{\end{center}}

41: 

42: \shorttitle{Curvature Effect of a Non-powerlaw Spectrum } \shortauthors{Zhang et

43: al. }

44: \begin{document}

45: 

46: \title{Curvature Effect of a Non-Power-Law Spectrum and Spectral Evolution of

47: GRB X-Ray Tails}

48: 

49: \author{Bin-Bin Zhang\altaffilmark{1}, Bing Zhang \altaffilmark{1},

50:   En-Wei Liang \altaffilmark{2}, Xiang-Yu Wang \altaffilmark{3} 

51: } \altaffiltext{1}{Department of Physics and Astronomy, University of Nevada,

52: Las Vegas,

53: NV 89154, USA; zbb,bzhang@physics.unlv.edu}

54: \altaffiltext{2}{Department of Physics, Guangxi University, Nanning 530004,

55: China}

56: \altaffiltext{3}{Department of Astronomy, Nanjing University, Nanjing 210093,

57: China}

58: \begin{abstract}

59: The apparent spectral evolution observed in the steep decay

60: phase of many GRB early afterglows raises a great concern of the

61: high-latitude ``curvature effect'' interpretation of this phase. 

62: However, previous curvature effect models only invoked a simple power 

63: law spectrum upon the cessation of the prompt internal emission. We 

64: investigate a model that invokes the ``curvature effect'' of a more

65: general non-power-law spectrum and test this model with

66: the Swift/XRT data of some GRBs. We show that one can

67: reproduce both the observed lightcurve and the apparent spectral 

68: evolution of several GRBs using a model invoking a power-law spectrum

69: with an exponential cut off. GRB 050814 is presented as an example.

70: \end{abstract}

71: 

72: 

73: 

74: \keywords{gamma-rays: bursts}

75: 

76: \section{INTRODUCTION\label{sec:intro}}

77: Most of the early X-Ray afterglows detected by Swift (Gehrels

78: et al. 2004)

79: show a steep decay phase around 100$\sim$1000

80: seconds after the burst trigger (Tagliaferri et al. 2005). 

81: The main characteristics of this steep decay phase include the

82: following. (1) It

83: connects smoothly to the prompt $\gamma$-ray light curve extrapolated 

84: to the X-ray band, suggesting that it is the ``tail'' of the prompt 

85: emission (Barthelmy et al. 2005, O'Brien et 2006,

86: Liang et al 2006).  (2) The decay slope is typically $3 \sim 5$ when

87: choosing the GRB trigger time as the zero time point $t_0$

88: (Tagliaferri et al. 2005; Nousek et al. 2006; Zhang et al 2006). 

89: (3) The time-averaged spectral index of the steep decay phase is

90: much different from that of the later shallow decay phase, indicating

91: that it is a distinct new component that is unrelated to the

92: conventional afterglow components (Zhang et al 2006; Liang et

93: al. 2007). (4) Strong spectral evolution exists in about one third

94: of the bursts that have a steep decay phase (Zhang et al. 2007, 

95: hereafter ZLZ07; Butler \& Kocevski 2007; 

96: Starling et al. 2008). All these features suggest that

97: the steep decay phase holds the key to understand the connection

98: between the prompt emission (internal) phase and the traditional 

99: afterglow (external) phase. Any proposed model (see M\'esz\'aros

100: 2006; Zhang 2007 for reviews) should be able to 

101: explain these features.

102: 

103: The so called ``curvature effect'', which accounts for the delayed 

104: photon emission from high latitudes with respect to the line of 

105: sight upon the abrupt cessation of emission in the prompt emission

106: region (Fenimore et al. 1996; Kumar \& Panaitescu 2000; 

107: Dermer 2004; Dyks et al. 2005; Qin 2008a), 

108: has been suggested to play an important role in shaping the sharp 

109: flux decline in GRB tails (Zhang et al. 2006; Liang et al. 

110: 2006; Wu et al. 2006; Yamazaki et al. 2006). In the simplest model, 

111: it is assumed that 

112: the instantaneous spectrum at

113: the end of the prompt emission is a simple power law 

114: with a spectral index $\beta$. The predicted temporal decay index

115: of the emission is (with the convention $F_{\nu} \propto t^{-\alpha} 

116: \nu^{-\beta}$)

117: \begin{equation}

118:  \alpha=2+\beta~,

119: \label{curvature-1}

120: \end{equation}

121: if the time origin to define the $\log-\log$ light curve, $t_0$, 

122: is taken as the beginning of the last emission episode

123: before the cessation of emission. Adopting a time-averaged

124: $\beta$ in the tails, Liang et al. (2006) 

125: found that Eq.(\ref{curvature-1})

126: is generally valid. 

127: The strong spectral evolution identified 

128: in a group of GRB tails (ZLZ07)

129: apparently violates Eq.(\ref{curvature-1}), which is valid only

130: for a constant $\beta$. ZLZ07 then investigated a curvature effect

131: model by assuming a structured jet with varying $\beta$ at different

132: latitudes and that the line of sight is near the jet 

133: axis\footnote{Notice

134:  that this structured jet model is different from the traditional

135:  one that invokes an angle-dependent energy/Lorentz factor, but

136:  not the spectral index (Zhang \& M\'esz\'aros 2002; Rossi et al. 2002).}.

137: One would then expect that Eq.(\ref{curvature-1}) is roughly

138: satisfied, with both $\alpha$ and $\beta$ being time-dependent.

139: ZLZ07 found that this model does not fit the data well.

140: 

141: These facts do not rule out the curvature effect interpretation of

142: GRB tails, however. This is because the instantaneous spectrum upon

143: the cessation of prompt emission may not be a simple power law.

144: If the spectrum has a curvature, as the 

145: emission from progressively higher latitudes reach the observer,

146: the XRT band is sampling different segments of the intrinsic

147: curved spectrum (Fig.1). This would introduce an apparent spectral

148: evolution in the decaying tail. 

149: The main goal of this paper is to test this more general

150: curvature effect model using the available Swift XRT data.

151: 

152: 

153: 

154: 

155: 

156: \section{Curvature Effect of a Non-powerlaw Spectrum }

157: 

158: We consider a general non-power-law spectrum in the form of 

159: \begin{equation}

160:  F_\nu(\nu)=F_{\nu,c}G(\nu)~,

161: \end{equation}

162: where $G(\nu)$ is the function form of the spectrum

163: with a characteristic frequency $\nu_c$ so that $G(\nu_c)$=1, 

164: and $F_{\nu,c}=F_\nu (\nu_c)$ is the normalization of the 

165: spectrum at $\nu=\nu_c$. 

166: 

167: The curvature effect states that given a same spectrum

168: at different latitudes with respect to the line of sight,

169: one has 

170: $F_{\nu,c}\propto {\cal D}^2$ and $\nu_c \propto {\cal D}$,

171: where ${\cal D}$ is the Doppler factor. If the high-latitude

172: angle $\theta \gg \Gamma$, the Dopper factor ${\cal D}

173: \propto t^{-1}$, so that $F_{\nu,c} \propto t^{-2}$,

174: $\nu_c \propto t^{-1}$ (Kumar \& Panaitescu 2000).

175: Considering the $t_0$ effect (Zhang et al. 2006; Liang

176: et al. 2006), this can be written as

177: \begin{equation}\label{fe}

178: F_{\nu,c}(t) = F_{\nu,c,p}  

179: \left(\frac{t-t_0}{t_p-t_0}\right)^{-2}

180: \end{equation}

181: and 

182: \begin{equation}\label{ep}

183:  \nu_c(t) = \nu_{c,p} 

184: \left(\frac{t-t_0}{t_p-t_0}\right)^{-1}

185: \end{equation}

186: for $t \gg t_p$, 

187: where $t_0$ refers to the time origin of the last pulse in the 

188: prompt emission and $t_p$ is the epoch when the curvature-effect 

189: decay starts (or the ``peak'' time of the lightcurve), 

190: $F_{\nu,c,p}=F_{\nu,c}(t_p)$ and $\nu_{c,p}=\nu_c(t_p)$.

191: Notice that in the case of $G(\nu)=(\nu/\nu_c)^{-\beta}$

192: (a pure power law spectrum), one derives $F_\nu \propto

193: (t-t_0)^{-\beta-2}$. This is the relation Eq.(\ref{curvature-1}).

194: 

195: We consider several physically motivated non-powerlaw spectra with 

196: a characteristic frequency $\nu_c$, including the cut-off power law 

197: spectrum and the

198: Band-function (Band et al. 1993). To explore the compatibility with 

199: the data, we also investigate different forms of the cutoffs with 

200: varying sharpness. In all cases, the $F_{\nu_p}(t)$ and

201: $\nu_p(t)$ follow Eqs.(\ref{fe}) and

202: (\ref{ep}). When $\nu_c(t)$ drops across an observational narrow 

203: energy band, e.g. the Swift/XRT band, it introduces an apparent

204: spectral softening with time, which, if fitted by a power law,

205: shows an increase of photon index with time.

206: In the meantime, the flux within the observing band

207: drops down rapidly, leading to an apparent steep decay phase 

208: in the lightcurve (Fig.1). 

209: 

210: \begin{table}

211:   \caption[]{Best-fitting parameters and their 1-sigma errors for 

212: the cutoff power curvature effect model for GRB050814.  }

213: \begin{tiny}

214:   \begin{center}\begin{tabular}{lllllll}

215:   \hline\noalign{\smallskip}

216: $N_{0,p}$ & $E_{c,p}({\rm keV})$ & $\Gamma$ & $t_0({\rm s})$ & nH$_{host}$&

217: k & $\chi^2$/dof\\

218:   \hline\noalign{\smallskip}

219: $0.67(0.12)$ &$10.2(1.3)$ &$1.56(0.25)$ &$103.5(3.4)$ &0.002(0.04)&1 (fixed)&$10.7/9$\\

220:   \hline\noalign{\smallskip}

221: 

222:   \end{tabular}\end{center}

223: \end{tiny}

224: \end{table}

225: 

226: 

227: \begin{figure}\label{cartoon}

228:  \includegraphics[angle=0,scale=0.7]{f1.eps}

229: 

230: \caption{A schematic picture showing that shifting a set of

231: non-power-law spectra in time can equivalently give an apparent spectral 

232: evolution in a fixed band. The dashed lines represent a set of 

233: exponential-like spectra, whose $F_{\nu_p}(t)$ and $\nu_p(t)$ 

234: drop down with time according to Eqs. (\ref{fe}) and (\ref{ep}). 

235: The two vertical solid lines bracket the observed energy band. 

236: The thick solid lines denote the effective power law fits

237: to the time-dependent spectra at each time step.} 

238: %\\ \end{tiny}

239: \end{figure}

240: 

241: 

242: \begin{figure}\label{050814fitting}

243:  \includegraphics[angle=0,scale=.45]{f2.ps}

244: 

245: \caption{The lightcurve (upper panel) and spectral evolution (lower

246: panel) of the X-ray tail of GRB 050814 with the best-fit theoretical

247: model (black curve in upper panel and green curve in lower panel).

248: The blue and red data points are the window

249: timing and photon counting data, respectively. The inset shows 

250: time-dependent theoretical spectra with the XRT band (0.3-10 keV)

251: bracketed by two vertical lines.  The integers denote the 

252: time segments for the time-resolved spectral analysis.}

253: 

254: \end{figure}

255: 

256: 

257: \begin{figure}

258: \begin{center}

259:   \includegraphics[angle=270,scale=0.13]{f3a.ps}

260:   \includegraphics[angle=270,scale=0.13]{f3c.ps} \\

261:   \includegraphics[angle=270,scale=0.13]{f3d.ps} 

262:   \includegraphics[angle=270,scale=0.13]{f3f.ps}

263: \caption{\textit{Upper panel}: Examples of simulated time-dependent spectra 

264: of GRB050814 with the best-fit parameters. The time intervals are 

265: 1,6 repectively as denoted in Fig. 2. In each panel, the data histogram

266: displays the simulated spectrum, and the solid line displays the

267: best-fit ($\chi^2/dof = 39.0/61,  25.2/25$) power

268: law model ($wabs*zwasb*powerlaw$ in XSPEC) that is used to derive the

269: time-dependent photon index $\Gamma$. \textit{Lower pabel}: The

270: corresponding observed spectrum in the three time intervals and

271: their power law fits ($\chi^2/dof = 47.1/46, 22.0/19$).}

272: \end{center}

273: 

274: \end{figure}

275: 

276: 

277: \begin{figure}

278: \begin{center}

279:  \includegraphics[angle=270,scale=0.25]{f4.ps}

280: \end{center}

281: 

282: \caption{The simulated cut-off power law spectrum at $t_p=144$ s 

283: based on the best fit model confronted by the BAT data in the time 

284: interval $(141.5-146.5)$ s. The reduced $\chi^2$ of the fitting 

285: is 1.2 with dof$=197$.}

286: \end{figure}

287: 

288: 

289: \section{Data Reduction and Simulation Method\label{sec:sim}}

290: 

291: We consider a time-dependent cutoff power law photon spectrum taking 

292: the form of  

293: \begin{equation}\label{eqmain}

294: N(E,t)=N_{0}(t) \left(\frac{E}{1~{\rm keV}}\right)^{-\Gamma} 

295: \exp\left[-\left(\frac{E}{E_c(t)}\right)^k \right]

296: \end{equation}

297: where $\Gamma=\beta+1$ is the power law photon index, and

298: $k$ is a parameter to define the sharpness of the high energy 

299: cutoff in the spectrum, $E_c(t)$ is the time-dependent 

300: characteristic photon energy, and $N_0(t)$ is a 

301: time-dependent photon flux (in units of

302: ${\rm photons\cdot keV^{-1}cm^{-2}s^{-1}}$) at 1 keV (Arnaud 1996). 

303: The choice of this function was encouraged by the fact that

304: the spectral evolution of some GRB tails can be fitted by such

305: an empirical model (Campana et al. 2006; ZLZ07; Yonetoku et al. 2008).

306: According to Eqs.(\ref{fe}) and (\ref{ep}), and 

307: noticing the conversion between the photon flux and the emission 

308: flux density, i.e. $F_\nu \propto EN(E)$,  we get

309: \begin{equation}\label{Ec}

310: E_c(t) = E_{c,p} \left(\frac{t-t_0}{t_p-t_0}\right)^{-1}

311: \end{equation}

312: and $N(E_c,t) = N_{c,p} \left[(t-t_0)/(t_p-t_0)\right]^{-1}$,

313: where $N_{c,p}=N(E_c,t_p)$, and $E_{c,p}=E_c(t_p)$. This gives

314: \begin{equation}\label{Nc_02}

315: N_{0}(t) = N_{0,p} \left(\frac{t-t_0}{t_p-t_0}\right)^{-(1+\Gamma)}~.

316: \end{equation}

317: Notice that $t_p$ is the beginning of the steep decay, which is

318: a parameter that can be directly constrained by the data. For

319: a complete lightcurve, we read $t_p$ off from

320: the lightcurve. In the case of an observational gap, usually 

321: $t_p$ can be reasonably fixed to the end of the prompt emission. 

322: We therefore do not include this parameter into the fits, and

323: derive the other five parameters, namely, $N_{0,p}$, $E_{c,p}$, 

324: $\Gamma$, $t_0$, and $k$ from the data. At any time $t$, the

325: model spectrum can be determined once these parameters are 

326: given. One can then confront the model with the real GRB data. 

327: 

328: The procedure includes the following steps.

329: (1) For a given burst, we extract its Swift/XRT light curve and $n$

330: slices of time-dependent spectra using the standard HEASoft/Swift

331: Package. The details of the data reduction method were described 

332: in ZLZ07. (2) Given a trial set of parameters in the theoretical 

333: spectra\footnote{Notice that $k$ is fixed 

334: to a certain value for a particular model, and is varied when 

335: different models are explored.} \{$N_{0,p}$, $E_{c,p}$, $\Gamma$, 

336: $t_0$\}, using Eqs.(\ref{eqmain}-\ref{Ec}) we model $n$

337: time-dependent {\em theoretical spectra} that correspond to 

338: the time bins that are used to derive the time-dependent

339: observed spectra.

340: (3) Based on the theoretical spectra of each time slice, we 

341: simulate the corresponding {\em model spectra} by taking account 

342: of the observational effects, including the Swift/XRT response 

343: matrix, the absorption column densities ($N_H$) of both the Milky 

344: Way (extracted from the observations from step 1) and the host galaxy 

345: of the burst (a free parameter),

346: the redshift (if applicable), and a Poisson noise background. 

347: Notice that $n_{H,host}$ is another parameter introduced in

348: the model spectra (besides the other parameters introduced

349: in the theoretical spectra). All these faked spectra can

350: be obtained using HEASoft (Version 6.4) and Xspec (Version 12.4) 

351: (4) We fit the faked model spectra with a simple power law model, 

352: i.e. $wabs*wabs*powerlaw$ (or $wabs*zwabs*powerlaw$ if the redsift 

353: is available) in XSPEC and get the simulated fluxes and spectral

354: indices of the $n$ slices. Here the column densities of both the 

355: Milk Way and the host galaxy are fixed to the obsered values as 

356: in Step 1. (5) We compare the simulated fluxes and spectral indices with the 

357: observed ones and access the goodness of the fits using $\chi^2$

358: statistics. (6) We refine the trial set of parameters based on the comparison 

359: and repeat steps (2)-(5) when necessary. We test whether we can 

360: reach a set of best-fitting parameters that can reproduce 

361: both the light curve and the apparent spectral evolution as 

362: observed.

363: 

364: 

365: \section{An example: GRB050814}

366: 

367: We apply the method to GRB050814, a typical burst with well-observed

368: X-ray tail with strong spectral evolution. As seen in

369: Fig.2, the tail has a steep decay index of $\sim 3.2$, and a strong 

370: spectral evolution is apparent 

371: at\footnote{The PC mode spectra become harder at

372: the end of this tail, which might be due to the contamination of

373: the harder shallow decay component. For simplification, we focus on 

374: the WT mode data only.} 

375: $t<600$ s. These features are common in most of the GRB X-ray tails. 

376: We first fix $k=1$ in Eq. \ref{eqmain}, which corresponds to the 

377: simplest cutoff powerlaw model. The initial

378: trial parameters we choose are ($\Gamma$, $N_{0,p}$, $t_0$,

379: $E_{p,0}$, $n_{H,host}$) $=(1.2, 0.4, 72.0, 30.0, 0.05)$.

380: The peak time $t_p$ is fixed to 143.6 s, which corresponds to the

381: end of the prompt emission.

382: Some IDL scripts are developed to follow the procedure described 

383: in \S3 to automatically search for the best-fit parameters  

384: to match both the observed light curve and the time-dependent 

385: spectral index. The final best-fitting parameters are shown in Table

386: 1. The corresponding simulated light curve (black curve) and 

387: spectral indices (green curve) are shown in Fig.2. Figure 2 suggests 

388: that the sharp decay and the spectral evolution in the tail of GRB 050814

389: can be indeed explained by the curvature effect with a cutoff power

390: law spectrum. In Fig.3 we present the comparison between the simualted

391: and observed spectra in the time steps 1 and 6 (as examples),

392: which show reasonable consistency.

393: 

394: Our model predicts that the prompt emission spectrum at $t_p \sim 144$

395: s should be a cut-off power law with the parameters in Table 1. In 

396: order to confirm this, we subtract the BAT-band spectrum in the time 

397: interval $(141.5 - 146.5)$ s, and compare with the data with the model

398: prediction. As shown in Fig.4, the BAT data is roughly consistent with

399: the model prediction, suggesting the validity of the model.

400: 

401: Some physical parameters can be constrained according to our model.

402: The time interval from $t_p$ to the beginning of the steep decay phase

403: $t_{tail,0}$ may be related to the angular spreading time scale

404: $\tau_{ang}=(t_{tail,0}-t_p)/(1+z)$. Noticing $z \sim 5.3$ for

405: GRB050814 (Jakobsson et al. 2005), we can estimate the Lorentz factor 

406: of the fireball as $\Gamma=({R}/2c\tau_{ang})^{1/2}

407:  \simeq 69 R_{15}^{1/2}$, where $R_{15}=R/(10^{15}~{\rm cm})$ is the

408: normalized emission radius. Since we

409: know the spectral peak energy $E_p$ at $t_p$, we can also

410: estimate the corresponding electrons' Lorentz factor for synchrotron

411: emission by

412: $\gamma_{e,p}=\left[E_{p}/({\hbar \Gamma \frac{eB}{mc}})\right]^{1/2}

413: \sim 2.4\times 10^3 R_{15}^{-1/4}B_3^{-1/2}$. 

414: From the rest frame duration of the X-ray tail we are analyzing 

415: $\tau_{tail} = (t_{tail,e} - t_{tail,0})/(1+z) \sim (378-165)/6.3 

416: = 33.8$ s, one can constrain the 

417: minimum jet opening angle as $\theta_j > (2c \tau_{tail}/R)^{1/2}

418: =2.6^{\rm o} \times R_{15}^{-1/2}$. 

419: These values are generally consistent with those derived from

420: various other methods.

421: 

422: We find that the abruptness parameter $k$ cannot be very 

423: different from unity. A Band-function spectrum introduces a 

424: less significant spectral evolution and it

425: cannot reproduce the data (cf. Qin 2008b). 

426: 

427: 

428: \section{Discussions and conclusions}

429: 

430: 

431: We have successfully modeled the lightcurve and spectral evolution

432: of the X-ray tail of GRB050814 using the curvature effect model of

433: a cutoff power law spectrum with

434: an exponential cutoff ($k=1$).

435: It has been

436: discussed in the literature (e.g. Fan \& Wei 2005; Barniol-Duran

437: \& Kumar 2008) that the GRB central engine may not die abruptly,

438: and that the observed X-ray tails may reflect the dying history of

439: the central engine. If this is indeed the case, the strong spectral 

440: evolution in the X-ray tails would demand a time-dependent 

441: particle acceleration mechanism that gives a progressively soft

442: particle spectrum. Such a behavior has not been predicted by 

443: particle acceleration theories. Our results suggest that at

444: least for some tails, the spectral evolution is simply a consequence

445: of the curvature effect: the observer views emission from the

446: progressively higher latitudes from the line of sight, so that

447: the XRT band is sampling the different segments of a curved spectrum.

448: This is a simpler interpretation.

449: 

450: The phenomenology of the X-ray tails are different from case to case

451: (ZLZ07). We have applied our model to some other clean X-Ray tails,

452: such as GRB050724, GRB080523, and find that they can be also interpreted

453: by this model. Some other tails have superposed 

454: X-ray flares, making a robust test of the model difficult. A systematic

455: survey of all the data sample is needed to address what fraction

456: of the bursts can be interpreted in this way or they demand

457: other physically distinct models (e.g. Barniol-Duran \& Kumar 2008; 

458: Dado et al. 2008). This is beyond the scope of this Letter.

459:   

460: \acknowledgments

461: We thank Pawan Kumar for stimulative discussions and comments. This

462: work is

463: supported by NASA NNG05GB67G, NNX07AJ64G, NNX08AN24G and NNX8AE57A 

464: (BBZ\&BZ), by the National Natural Science Foundation of China

465: under Grant 10463001 (EWL) and 10221001 (XYW), and by the National 

466: ``973'' Program of China under Grant 2009CB824800 (EWL\& XYW). 

467: BBZ \& BZ also acknowledges the President's Infrastructure Award 

468: from UNLV.

469: 

470: \begin{thebibliography}{}

471: \bibitem[]{419}Arnaud, K.A., 1996, Astronomical Data Analysis Software and Systems

472: V, eds. Jacoby G. and Barnes J., p17, ASP Conf. Series volume 101.

473: \bibitem[]{421} Barniol Duran, R \& P. Kumar, 2008, arXiv:0806.1226

474: \bibitem[]{422} Band, D et al. 1993, ApJ, 413, 281

475: \bibitem[]{423} Barthelmy, S et al. 2005, Nature, 438, 994

476: \bibitem[]{424} Butler, N. R. \& Kocevki, D. 2007, ApJ, 668, 400

477: \bibitem[]{425} Campana, S. et al. 2006, Nature, 442, 1008

478: \bibitem[]{426} Dado, S., Dar, A., De Rujula, A. 2008, ApJ, 681, 1408

479: \bibitem[]{427} Dermer, C. D.  2004, ApJ, 614, 284

480: \bibitem[]{428} Dyks, J., Zhang, B. \& Fan, Y. Z. 2005 , astro-ph/0511699

481: \bibitem[]{429} Fan, Y. Z. \& Wei, D. M. 2005, MNRAS, 364, L42

482: \bibitem[]{430} Fenimore, E. E., Madras, C. D., \& Nayakshin, S. 1996, ApJ, 473,

483: 998

484: \bibitem[]{432} Gehrels, N., et al. 2004, ApJ, 611, 1005

485: \bibitem[]{452} Jakobsson, P. et al. 2005, A\&A, 447, 897

486: \bibitem[]{433} Kumar, P. \& Panaitescu, A. 2000, ApJ, 541, L51

487: \bibitem[]{434} Liang, E. W., Zhang, B., O'Brien, P. T., Willingale, R., et al.

488: 2006, ApJ, 646, 351

489: \bibitem[Liang et al.(2007)]{2007ApJ...670..565L} Liang, E.-W., Zhang,

490:      B.-B., Zhang, B., 2007. ApJ 670, 565.

491: \bibitem[]{438} M\'esz\'aros, P. 2006, Rep. Prog. Phys. 69, 2259

492: \bibitem[]{439} Nousek, J. A., Kouveliotou, C., Grupe, D., Page, K. L., et al.

493: 2006, ApJ, 642, 389

494: \bibitem[]{441} O'Brien, P. T., Willingale, R., Osborne, J., Goad, M. R.,

495:   et al. 2006, ApJ, 647,1213

496: \bibitem[]{443} Qin, Y.-P. 2008a, ApJ, in press (arXiv:0804.2175)

497: \bibitem[]{444} Qin, Y.-P. 2008b, ApJ, submitted (arXiv:0806.3339)

498: \bibitem[]{445} Rossi, E., Lazzati, D., \& Rees, M. J. 2002, MNRAS, 332, 945

499: \bibitem[]{446} Starling, R. L. C. et al. 2008, MNRAS, 384, 504

500: \bibitem[]{447} Tagliaferri, G., Goad, M., Chincarini, G., Moretti, A., et al.

501: 2005, Nature, 436, 985

502: \bibitem[]{449} Wu, X. F. et al. 2006, 36th COSPAR Sci. Ass. \#731

503:   (astro-ph/0512555)

504: \bibitem[]{471}Yamazaki et al. 2006, MNRAS, 369, 311

505: \bibitem[]{451} Yonetoku, D. et al. 2008, PASJ, 60, S352

506: \bibitem[]{452} Zhang, B. 2007, ChJAA, 7, 1

507: \bibitem[]{453} Zhang, B., Fan, Y. Z, Dyks, J., Kobayashi, S., et al. 2006, ApJ,

508: 642, 354

509: \bibitem[]{455} Zhang, B. \& M\'esz\'aros, P.  2002, ApJ, 571, 876

510: \bibitem[Zhang et al.(2007a)]{2007ApJ...666.1002Z} Zhang, B.-B., Liang, E.-W.,

511: Zhang, B., 2007, ApJ 666, 1002 (ZLZ07)

512: \end{thebibliography}

513: 

514: 

515: \end{document}