1: %% Beginning of file 'sample.tex'
2: %% Modified 2005 December 5
3: %% This is a sample manuscript marked up using the
4: %% AASTeX v5.x LaTeX 2e macros.
5: %% The first piece of markup in an AASTeX v5.x document
6: %% is the \documentclass command. LaTeX will ignore
7: %% any data that comes before this command.
8: %% The command below calls the preprint style
9: %% which will produce a one-column, single-spaced document.
10: %% Examples of commands for other substyles follow. Use
11: %% whichever is most appropriate for your purposes.
12: %% manuscript produces a one-column, double-spaced document:
13: %%\documentclass[manuscript]{aastex}
14: %% preprint2 produces a double-column, single-spaced document:
15: %% \documentclass[preprint2]{aastex}
16: %% Sometimes a paper's abstract is too long to fit on the
17: %% title page in preprint2 mode. When that is the case,
18: %% use the longabstract style option.
19: %% \documentclass[preprint2,longabstract]{aastex}
20: %% If you want to create your own macros, you can do so
21: %% using \newcommand. Your macros should appear before
22: %% the \begin{document} command.
23: %% If you are submitting to a journal that translates manuscripts
24: %% into SGML, you need to follow certain guidelines when preparing
25: %% your macros. See the AASTeX v5.x Author Guide
26: %% for information.
27: %% You can insert a short comment on the title page using the command below.
28: %\slugcomment{Not to appear in Nonlearned J., 45.}
29: %% If you wish, you may supply running head information, although
30: %% this information may be modified by the editorial offices.
31: %% The left head contains a list of authors,
32: %% usually a maximum of three (otherwise use et al.). The right
33: %% head is a modified title of up to roughly 44 characters.
34: %% Running heads will not print in the manuscript style.
35: %% This is the end of the preamble. Indicate the beginning of the
36: %% paper itself with \begin{document}.
37:
38:
39: \documentclass[12pt,preprint]{aastex}
40: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
41: %\usepackage{subfigure}
42:
43: %TCIDATA{OutputFilter=Latex.dll}
44: %TCIDATA{Version=5.50.0.2890}
45: %TCIDATA{<META NAME="SaveForMode" CONTENT="1">}
46: %TCIDATA{BibliographyScheme=Manual}
47: %TCIDATA{LastRevised=Monday, June 16, 2008 01:21:09}
48: %TCIDATA{<META NAME="GraphicsSave" CONTENT="32">}
49:
50: \newcommand{\vdag}{(v)^\dagger}
51: \newcommand{\myemail}{skywalker@galaxy.far.far.away}
52: \shorttitle{The softening phenomenon} \shortauthors{Qin}
53:
54: \begin{document}
55:
56: \title{The softening phenomenon due to the curvature effect: in the case of extremely short intrinsic
57: emission}
58: \author{Y.-P. Qin\altaffilmark{1,2}}
59:
60: \altaffiltext{1}{Center for Astrophysics, Guangzhou University,
61: Guangzhou 510006, P. R. China; ypqin@gzhu.edu.cn}
62:
63: \altaffiltext{2}{Physics Department, Guangxi University, Nanning
64: 530004, P. R. China}
65:
66: \begin{abstract}
67: Both the light curve and spectral evolution of the radiation from a
68: relativistic fireball with extremely short duration are studied, in
69: order to examine the curvature effect for different forms of the
70: radiation spectrum. Assuming a $\delta$ function emission we get
71: formulas that get rid of the impacts from the intrinsic emission
72: duration, applicable to any forms of spectrum. It shows that the
73: same form of spectrum could be observed at different times, with the
74: peak energy of the spectrum shifting from higher energy bands to
75: lower bands following $E_{peak}\propto t^{-1}$. When the emission is
76: early enough the $t^{2}f_{\nu }(t)$ form as a function of time will
77: possess exactly the same form that the intrinsic spectrum as a
78: function of frequency has. Assuming $f_\nu \propto \nu^{-\beta}
79: t^{-\alpha}$ one finds $\alpha=2+\beta$ which holds for any
80: intrinsic spectral forms. This relation will be broken down and
81: $\alpha > 2+\beta $ or $\alpha \gg 2+\beta $ will hold at much later
82: time when the angle between the moving direction of the emission
83: area and the line of sight is large. An intrinsic spectrum in the
84: form of the Band function is employed to display the light curve and
85: spectral evolution. Caused by the shifting of the Band function
86: spectrum, a temporal steep decay phase and a spectral softening
87: appear simultaneously. The softening phenomenon will appear at
88: different frequencies. It occurs earlier for higher frequencies and
89: later for lower frequencies. The terminating softening time
90: $t_{s,max}$ depends on the observation frequency, following
91: $t_{s,max}\propto \nu^{-1}$. This model predicts that the softening
92: duration would be linearly correlated with $t_{s,max}$; the observed
93: $\beta_{min}$ and $\beta_{max}$ are determined by the low and high
94: energy indexes of the Band function; both $\beta_{min}$ and
95: $\beta_{max}$ are independent of the observation frequency.
96: \end{abstract}
97:
98: \keywords{gamma-rays: bursts --- gamma-rays: theory --- relativity}
99:
100: %% LaTeX will automatically break titles if they run longer than
101: %% one line. However, you may use \\ to force a line break if
102: %% you desire.
103:
104: %% Use \author, \affil, and the \and command to format
105: %% author and affiliation information.
106: %% Note that \email has replaced the old \authoremail command
107: %% from AASTeX v4.0. You can use \email to mark an email address
108: %% anywhere in the paper, not just in the front matter.
109: %% As in the title, use \\ to force line breaks.
110:
111: %% Notice that each of these authors has alternate affiliations, which
112: %% are identified by the \altaffilmark after each name. Specify alternate
113: %% affiliation information with \altaffiltext, with one command per each
114: %% affiliation.
115:
116:
117: %% Mark off your abstract in the ``abstract'' environment. In the manuscript
118: %% style, abstract will output a Received/Accepted line after the
119: %% title and affiliation information. No date will appear since the author
120: %% does not have this information. The dates will be filled in by the
121: %% editorial office after submission.
122:
123: %% Keywords should appear after the \end{abstract} command. The uncommented
124: %% example has been keyed in ApJ style. See the instructions to authors
125: %% for the journal to which you are submitting your paper to determine
126: %% what keyword punctuation is appropriate.
127:
128: %% From the front matter, we move on to the body of the paper.
129: %% In the first two sections, notice the use of the natbib \citep
130: %% and \citet commands to identify citations. The citations are
131: %% tied to the reference list via symbolic KEYs. The KEY corresponds
132: %% to the KEY in the \bibitem in the reference list below. We have
133: %% chosen the first three characters of the first author's name plus
134: %% the last two numeral of the year of publication as our KEY for
135: %% each reference.
136:
137: %% Authors who wish to have the most important objects in their paper
138: %% linked in the electronic edition to a data center may do so by tagging
139: %% their objects with \objectname{} or \object{}. Each macro takes the
140: %% object name as its required argument. The optional, square-bracket
141: %% argument should be used in cases where the data center identification
142: %% differs from what is to be printed in the paper. The text appearing
143: %% in curly braces is what will appear in print in the published paper.
144: %% If the object name is recognized by the data centers, it will be linked
145: %% in the electronic edition to the object data available at the data centers
146: %%
147: %% Note that for sources with brackets in their names, e.g. [WEG2004] 14h-090,
148: %% the brackets must be escaped with backslashes when used in the first
149: %% square-bracket argument, for instance, \object[\[WEG2004\] 14h-090]{90}).
150: %% Otherwise, LaTeX will issue an error.
151:
152: \section{Introduction}
153:
154: The successful launch of the Swift satellite (Gehrels et al. 2004)
155: has made great advance for the observations of the X-ray afterglows
156: of gamma-ray bursts (GRBs) in the past few years (for a recent
157: review, see Zhang 2007). Among the many new findings obtained by the
158: Swift instruments, some emerge as a puzzling. One is the spectral
159: evolution which was detected in the tails of some bursts (Campana et
160: al. 2006; Ghisellini et al. 2006; Gehrels et al. 2006; Mangano et
161: al. 2007; Zhang et al. 2007a). The phenomenon has not been predicted
162: by the curvature effect nor by the standard external shock afterglow
163: model. Soon after that, Zhang et al. (2007b) performed a systematic
164: analysis on this issue and found that 33 of the 44 bursts with
165: bright steep decay tails show an obvious spectral evolution --- an
166: observed softening. This suggests that, the detected spectral
167: evolution is not a rare phenomenon, but instead, it is quite common
168: (at least detectable in the majority of GRBs that have bright steep
169: decay tails).
170:
171: Several attempts have been made to interpret this softening
172: phenomenon. It was proposed that a central engine producing a soft
173: but decaying afterglow emission might be responsible for some of
174: these bursts (Campana et al. 2006; Fan et al. 2006; Zhang et al.
175: 2007b). Some bursts with strong softening might be accounted for by
176: a cooling of the internal-shocked region (Zhang et al. 2007b). A
177: possible thermal component has also been tried. However, Yonetoku et
178: al. (2008) showed that introducing a thermal component is not
179: sufficient to explain all of the spectral softening, and thus
180: additional spectral evolution is required. In investigating the
181: origins for the spectral evolution of GRB 070616, Starling et al.
182: (2008) ruled out the possibility that a superposition of two
183: power-laws causes the evolution. In stead, they considered a
184: possibility of an additional component dominant during the late
185: prompt emission. They proposed that a combination of the spectral
186: evolution and the curvature effect may cause the observed steep
187: decay phase of the light curve. Another scenario is based on the
188: cannonball model of GRBs, which was suggested to be responsible for
189: both the temporal behavior and the spectral softening of the bursts
190: observed by Swift (Dado et al. 2008).
191:
192: As shown in Zhang et al. (2007b), the spectral softening is
193: accompanied by a very steep decay phase. This phase is seen directly
194: following the prompt emission and is naturally (and generally)
195: regarded as the tail of the prompt emission (Tagliaferri et al.
196: 2005; Barthelmy et al. 2005; Liang et al. 2006). The tail was
197: suspected to arise from the emission of the high latitude of the
198: fireball surface, which is often referred to as the curvature effect
199: (Kumar \& Panaitescu 2000; Dermer 2004; Dyks et al. 2005; Liang et
200: al. 2006; Panaitescu et al. 2006; Zhang et al. 2006, 2007b; Butler
201: \& Kocevski 2007; Qin 2008; Starling et al. 2008). A full
202: consideration of the curvature effect includes the delay of time and
203: the shifting of the intrinsic spectrum as well as other relevant
204: factors of an expanding fireball (for detailed explanation and
205: analysis, see Qin 2002; Qin et al. 2004, 2006; Qin 2008). The effect
206: has been intensively studied in the prompt gamma-ray phase, such as
207: the profile of the light curve of pulses, the spectral lags, the
208: power-law relation between the pulse width and energy, the evolution
209: of the hardness ratio and the evolution of the peak energy (Fenimore
210: et al. 1996; Sari \& Piran 1997; Qin 2002; Ryde \& Petrosian 2002;
211: Kocevski et al. 2003; Qin \& Lu 2005; Shen et al. 2005; Lu et al.
212: 2006, 2007; Peng et al. 2006; Qin et al. 2004, 2005, 2006; Jia
213: 2008).
214:
215: In a recent investigation, Butler \& Kocevski (2007) concluded that the
216: early emission in $>90\%$ of early afterglows has a characteristic $\nu
217: f_{\nu}$ spectral energy $E_{peak}$, which likely evolves from the $\gamma$%
218: -rays through the soft X-ray band on timescales of $10^2 - 10^4$ s
219: after the GRB. Many careful analyses revealed that there do exist
220: some bursts with their peak energy $E_{peak}$ decreasing from a
221: higher band to a much lower band. These bursts include: GRB 060124,
222: from 108 keV to 1.3 keV (Butler \& Kocevski 2007); GRB 060614, from
223: 8.6 keV to 1.1 keV (Butler \& Kocevski 2007; Mangano et al. 2007);
224: GRB 060904A, from 163 keV to 2.28 keV (Yonetoku et al. 2008); GRB
225: 061121, from 270 keV to 0.95 keV (Butler \& Kocevski 2007); GRB
226: 070616, from 135 keV to $\sim$4 keV (Starling et al. 2008). For GRB
227: 070616, the spectral softening evolution was observed even in the
228: prompt emission phase: its duration is $T_{90}=402.4s$, while the
229: softening starts from 285s and extends to 1200s (Starling et al.
230: 2008). Among them, GRB 060614, GRB 060904A and GRB 061121 are
231: members of the Zhang et al. (2007b)'s sample. Although the curvature
232: effect was rejected to interpret the softening by some authors,
233: Starling et al. (2008) insisted that the observed shifting of the
234: peak energy is in agreement with what expected by the curvature
235: effect: the peak energy of the Band function spectrum passes through
236: the $\gamma$-ray band at a relatively early time, while it passes
237: through the X-ray band at a later time due to the high latitude
238: emission. They proposed that both the curvature effect and a strong
239: spectral evolution cause the steep decline in flux. Based on the
240: explicit illustration of the evolution of the whole spectral form in
241: Butler \& Kocevski (2007), we suspect that the curvature effect
242: alone might be responsible for both the spectral softening and the
243: accompanied steep decay light curve.
244:
245: To reveal the pure curvature effect and get rid of the possible
246: impacts from the emission duration, we concern in this paper only
247: extremely short intrinsic emission. Focusing on this emission has
248: two advantages. The first is that the formulas become very simple,
249: and the second is that many key characteristics of the effect can be
250: plainly illustrated. The paper is organized as follows. In Section
251: 2, we present basic formulas of the full curvature effect,
252: applicable to any temporal and spectral forms of emission. We
253: discuss light curves and spectral evolution arising from an
254: extremely short emission in Section 3. In Section 4, we assume an
255: intrinsic spectrum in the form of the Band function and illustrate
256: the corresponding light curve and spectral evolution in detail.
257: Parameters that affect the results are discussed in Section 5. In
258: Section 6, we apply the model to the XRT band and also to a Swift
259: burst. Discussion and conclusions are presented in Section 7.
260:
261: \section{Equations of flux densities influenced by the curvature effect}
262:
263: In the following, we study the emission from an expanding fireball
264: shell. Suppose the emission occurs over the fireball area confined
265: by $\theta _{\min }\leq \theta \leq \theta _{\max }$, where $\theta
266: $ is the angle between the normal of the area concerned with respect
267: to the line of sight (which is also the angle of the moving
268: direction of the emitting region with respect to the direction to
269: the observer), and within proper time interval $t_{0,\min }\leq
270: t_{0}\leq t_{0,\max }$. Let us consider the following situation: the
271: Lorentz factor of the emitting shell is constant, the energy range
272: of the emission is unlimited, and the intrinsic radiation intensity
273: is independent of direction. Basic formulas of the flux density
274: which is expected by a distant observer measured at laboratory time
275: $t_{ob}$ and other relevant quantities for this simple situation are
276: presented in Qin (2008; see equations 1-5 there).
277:
278: To meet and/or approximate the conventional definition of
279: observation time, we assign (see also Qin 2008)
280: \begin{equation}
281: t\equiv t_{ob}-t_{c}+R_{c}/v-D/c,
282: \end{equation}%
283: where $t_{c}$ as a time constant is defined in the observer frame,
284: $D$ is the distance of the fireball to the observer, $v$ is the
285: speed of the shell, $R_{c}$ is the radius of the shell measured at
286: $t_{c}$ by local observers who are stationary in the explosion area.
287: Equation (1) is a definition of observation time. With this
288: definition of time, the basic formulas can be written as
289: \begin{equation}
290: f_{\nu }(t)=\frac{2\pi c^{2}}{D^{2}(\Gamma v/c)^{2}t^{2}}\int_{\widetilde{t%
291: }_{0,\min }}^{\widetilde{t}_{0,\max }}I_{0,\nu }(t_{0},\nu
292: _{0})[R_{c}/c+(t_{0}-t_{0,c})\Gamma v/c]^{2}[(t_{0}-t_{0,c})\Gamma
293: +R_{c}/v-t]dt_{0},
294: \end{equation}%
295: with%
296: \begin{equation}
297: \widetilde{t}_{0,\min }=\max \{t_{0,\min
298: },\frac{t-R_{c}/v+(R_{c}/c)\cos \theta _{\max }}{[1-(v/c)\cos \theta
299: _{\max }]\Gamma }+t_{0,c}\},
300: \end{equation}%
301: \begin{equation}
302: \widetilde{t}_{0,\max }=\min \{t_{0,\max
303: },\frac{t-R_{c}/v+(R_{c}/c)\cos \theta _{\min }}{[1-(v/c)\cos \theta
304: _{\min }]\Gamma }+t_{0,c}\},
305: \end{equation}%
306: \begin{equation}
307: \nu _{0}=\frac{t}{R_{c}/v+(t_{0}-t_{0,c})\Gamma }\Gamma \nu ,
308: \end{equation}%
309: and%
310: \begin{equation}
311: %
312: \begin{array}{l}
313: \lbrack 1-(v/c)\cos \theta _{\min }][(t_{0,\min }-t_{0,c})\Gamma
314: +R_{c}/v]\leq t \\
315: \leq \lbrack 1-(v/c)\cos \theta _{\max }][(t_{0,\max }-t_{0,c})\Gamma
316: +R_{c}/v]%
317: \end{array}%
318: ,
319: \end{equation}%
320: where $t_{0,c}$ denotes the moment of $t_{c}$, measured by a
321: co-moving observer; $I_{0,\nu }(t_{0},\nu _{0})$ is the intrinsic
322: radiation intensity; $\Gamma =1/\sqrt{1-v^{2}/c^{2}}$. Equations
323: (2)-(6) are more general than equations (8)-(12) in Qin (2008). The
324: former can be applied to any forms of the intrinsic spectrum, while
325: the later are applicable only in the case of a single power-law
326: spectrum. According to equation (1), referring also to equation (8)
327: in Qin et al. (2004), $t=0$ corresponds to the moment of the
328: emission that occurs at the spot of the explosion (say, $ R_{c}=0$).
329: Or precisely, $t=0$ is the moment when photons emitted from $
330: R_{c}=0$ reach the observer. As explained in Qin (2008), observation
331: time $t$ approximates the time defined by the trigger time (e.g.,
332: $t=t_{ob}-t_{ob,trig}$). As long as the Lorentz factor is large
333: enough and the trigger event is early enough, the offset between the
334: two definitions of observation time would be very small (Qin 2008).
335:
336: \section{Light curves and the spectral evolution of the fireball arising
337: from an extremely short intrinsic emission}
338:
339: The simplest emission is a $\delta$ function emission which can
340: always simplify the equations concerned. Perhaps the most important
341: reason for considering a $\delta$ function emission is that effects
342: arising from the duration of real intrinsic emission will be omitted
343: and therefore those merely coming from the expanding motion of the
344: fireball surface will be clearly seen (Qin 2008). In practical
345: situation, when an emission is extremely short, one could regard it
346: as a $\delta$ function emission. In order to reveal the main
347: properties of the curvature effect in the case of
348: X-ray afterglow, we consider only this kind of radiation and hence assume a $%
349: \delta$ function emission through out this paper.
350:
351: Let the concerned intrinsic emission be
352: \begin{equation}
353: I_{0,\nu }(t_{0},\nu _{0})=I_{0,0}\delta
354: (\frac{t_{0}}{t_{0,0}}-\frac{t_{0,c}}{t_{0,0}})g_{\nu }(\nu _{0}),
355: \end{equation}%
356: where $t_{0,0}>0$ is any assigned time constant (e.g.,
357: $t_{0,0}=1s$), $I_{0,0}$ is a constant in units of $erg\cdot cm^{-2}
358: s^{-1} Hz^{-1}$, and $g_{\nu }(\nu _{0})$ is the intrinsic spectrum
359: of the emission in a dimensionless form. We consider the emission
360: from the whole fireball surface and take $\theta _{\min }=0$ and
361: $\theta _{\max }=\pi /2$. In this situation, one gets from (6) that
362: (see also Qin 2008)
363: \begin{equation}
364: (1-v/c)R_{c}/v\leq t\leq R_{c}/v.
365: \end{equation}%
366: Within this observation time interval, equation (2) becomes
367: \begin{equation}
368: f_{\nu }(t)=\frac{2\pi I_{0,0}t_{0,0}R_{c}^2}{D^{2}(\Gamma
369: v/c)^{2}}\frac{g_{\nu }[\nu _{0}(t_0 = t_{0,c})](R_{c}/v-t)}{t^{2}}.
370: \end{equation}%
371: According to equation (5), $\nu _{0}$ is related to $t$ and $\nu$ by
372: \begin{equation}
373: \nu _{0}=\frac{\Gamma v}{R_{c}}t\nu .
374: \end{equation}
375: Inserting equation (10) into equation (9) comes to a plain result:
376: \begin{equation}
377: f_{\nu }(t)=\frac{2\pi I_{0,0}t_{0,0}R_{c}^2}{D^{2}(\Gamma v/c)^{2}}%
378: (R_{c}/v-t)g_{\nu }(\frac{\Gamma v}{R_{c}}t\nu )t^{-2}.
379: \end{equation}
380:
381: A straightforward consequence of equation (11) comes from the
382: situation when $g_{\nu }(\nu_0)\propto \nu_0^{-\beta}$, that gives
383: rise to $f_{\nu }(t) \propto (R_{c}/v-t)t^{-2-\beta}\nu^{-\beta}$
384: (see Qin 2008). When the event occurs early enough, the $R_{c}/v-t$
385: term will become constant and then we come to the well-known form of
386: flux density $f_{\nu }(t) \propto t^{-2-\beta}\nu^{-\beta}$
387: (Fenimore et al. 1996; Kumar \& Panaitescu 2000; Qin 2008).
388:
389: Several conclusions are reached from equation (11). a) For a certain
390: observation time satisfying equation (8), the spectrum observed is
391: merely a shifted intrinsic one. The shifting factor is $\Gamma
392: vt/R_{c}$. The peak energy will decline following
393: $E_{peak}=E_{peak,0}R_{c}/ \Gamma vt$ (namely, $E_{peak}\propto
394: t^{-1}$). b) For a certain observation frequency, the light curve
395: observed depends entirely on the emission spectrum. In the case when
396: $R_{c}/v\gg t$, the form $t^{2}f_{\nu }(t)$, $\propto g_{\nu
397: }(\frac{\Gamma v\nu }{R_{c}}t)$, as a function of time takes the
398: same form that the intrinsic spectrum as a function of frequency
399: has. Or, from $t^{2}f_{\nu }(t)|_{\nu=const}$, when replacing $t$
400: with variable $\nu$ and multiplying it with a constant to alter its
401: dimension one will get exactly the intrinsic spectral form (this
402: might be useful in checking the curvature effect in further
403: investigations). This is plain in the pure power-law emission where
404: $f_{\nu }(t) \propto t^{-2-\beta}\nu^{-\beta}$ in the case
405: $R_{c}/v\gg t$. But the conclusion holds for any intrinsic spectral
406: forms, which is unaware previously. As explained in Qin (2008), the
407: term $R_{c}/v-t$ reflects the projected factor of the infinitesimal
408: fireball surface area in the angle concerned to the distant
409: observer, known as $\cos \theta $ (in fact, $R_{c}/v-t$ is $\cos
410: \theta $ multiplying a constant; see Appendix A). When the emission
411: area is close to the line of sight region, condition $R_{c}/v\gg t$
412: (i.e., $\cos \theta \sim 1$) can easily be satisfied, while when the
413: angle between the moving direction of the emission region and the
414: line of sight is large enough the term $R_{c}/v-t$ will be
415: important. At a much later time, the emission area would be close to
416: that of $\theta =\pi /2$, and the term $R_{c}/v-t$ and then the flux
417: will approach to zero (see Appendix A and Appendix B). c) Under the
418: condition of $R_{c}/v\gg t$, which will hold at an earlier time, the
419: temporal power law index and the spectral power law index will be
420: well related. For a given time $t$ and a given frequency $\nu $, we
421: assign the flux as $f_{\nu }(t)=f_{0}t^{-\alpha }$ and assign the
422: intrinsic spectrum as $g_{\nu }(\frac{\Gamma v}{R_{c}}t\nu
423: )=g_{0}\cdot (\frac{\Gamma v }{R_{c}}t\nu )^{-\beta }$. These can
424: easily be satisfied when one carefully chooses $f_{0}$, $\alpha $,
425: $g_{0}$ and $\beta $ in the vicinity of $t$ (e.g., by performing a
426: fit). Inserting $g_{\nu }(\frac{\Gamma v}{R_{c}}t\nu )=g_{0}\cdot
427: (\frac{\Gamma v }{R_{c}}t\nu )^{-\beta }$ into $f_{\nu
428: }(t)=f_{0}t^{-\alpha }$ under the $R_{c}/v\gg t$ condition one will
429: come to the well-known $\alpha =2+\beta $ relation from equation
430: (11). Note that this relation will hold for any intrinsic spectral
431: forms as long as the condition $R_{c}/v\gg t$ is satisfied. When the
432: angle $\theta$ is large enough (say, when $R_c/v-t \ll R_{c}/c$; see
433: Appendix A) the $\alpha =2+\beta $ relation will be broken down
434: since the influence of the $R_c/v-t$ term on the differential of the
435: light curve (which is associated with $\alpha$) will no more be
436: ignored.
437:
438: \section{In the case of the Band function spectrum}
439:
440: According to the above analysis, for an extremely short emission
441: burst, its observed spectrum and its intrinsic spectrum take the
442: same form. It is known that the observed spectra of most GRBs
443: possess the Band function form (Band et al. 1993). Since emission
444: from some of such bursts might be extremely short, it is likely that
445: the intrinsic emission of some bursts takes the Band function form.
446: In this section, we consider an intrinsic Band function emission and
447: assume (Band et al. 1993)
448: \begin{equation}
449: g_{\nu }(\nu _{0})=\{%
450: \begin{array}{c}
451: (\frac{\nu _{0}}{\nu _{0,p}})^{1+\alpha _{B,0}}\exp [-(2+\alpha _{B,0})\frac{%
452: \nu _{0}}{\nu _{0,p}}]\qquad \qquad \qquad \qquad (\frac{\nu _{0}}{\nu _{0,p}%
453: }<\frac{\alpha _{B,0}-\beta _{B,0}}{2+\alpha _{B,0}}) \\
454: (\frac{\alpha _{B,0}-\beta _{B,0}}{2+\alpha _{B,0}})^{\alpha
455: _{B,0}-\beta _{B,0}}\exp (\beta _{B,0}-\alpha _{B,0})(\frac{\nu
456: _{0}}{\nu _{0,p}})^{1+\beta
457: _{B,0}}\qquad (\frac{\alpha _{B,0}-\beta _{B,0}}{2+\alpha _{B,0}}\leq \frac{%
458: \nu _{0}}{\nu _{0,p}})%
459: \end{array}%
460: ,
461: \end{equation}%
462: where $\nu _{0,p}$, $\alpha _{B,0}$, and $\beta _{B,0}$ are
463: constants.
464:
465: With this spectral form, we can produce the light curve and the
466: spectral evolution using equation (11). In the following analysis we
467: take $\Gamma =100 $, $R_{c}=10^{15}cm$, $1+\alpha _{B,0}=-0.5$,
468: $1+\beta _{B,0}=-2.5$, and $h\nu _{0,p}=E_{0,p}=1keV$ as a primary
469: set of parameters. In the consequent analysis below they will be
470: replaced one by one to reveal their influences on the light curve
471: and the spectral evolution. We assign $F_0 \equiv 2\pi
472: I_{0,0}R_{c}^2/D^{2}(\Gamma v/c)^{2}$ and $t_{0,0}=1s$. The flux
473: density will be calculated in units of $F_0$ through out this paper.
474:
475: \subsection{Expected at the 1 keV observation frequency}
476:
477: First, let us explore the spectral evolution at $E=1keV$ by assuming
478: a power
479: law of flux within a limited (or narrow) band including $E=1keV$:%
480: \begin{equation}
481: f_{\nu }(t)=I(t)\nu ^{-\beta }.
482: \end{equation}
483:
484: %\clearpage
485: \begin{figure}[tbp]
486: \begin{center}
487: \includegraphics[width=5in,angle=0]{f1.eps}
488: \end{center}
489: \caption{The light curve (the upper panel) and spectral evolution
490: (the lower panel) of a $\delta$ function emission from an expanding
491: fireball, expected at 1 keV observation frequency. Observation time
492: $t$ is defined by equation (1) and in units of $s$. The flux density
493: is calculated with equation (11) and in units of $F_0$ [$F_0 \equiv
494: 2\pi I_{0,0}R_{c}^2/D^{2}(\Gamma v/c)^{2}$], while the intrinsic
495: spectrum is determined by equation (12). The adopted parameters
496: include: $\Gamma =100 $, $R_{c}=10^{15}cm$, $1+\alpha _{B,0}=-0.5$,
497: $1+\beta _{B,0}=-2.5$, and $h\nu _{0,p}=1keV$. Equation (13) is used
498: to defined and evaluate the spectral index $\beta$.} \label{Fig.
499: 1_01}
500: \end{figure}
501: %\clearpage
502:
503: The resulting light curve and spectral evolution are displayed in
504: Fig. 1. A temporal steep decay phase and a spectral softening are
505: observed within the range of $1-10^{3}$ s, which occur
506: simultaneously. Determined or influenced by the intrinsic spectral
507: form, the light curve decays in a milder manner at an earlier time
508: (influenced by the lower energy index of the Band function) and then
509: turns to be steeper at a much later time (influenced by the higher
510: energy index of the Band function). Connecting these two segments is
511: a breaking feature which appears when the peak energy of the
512: spectrum passes through the observation band (here, the 1 keV
513: observation frequency). This feature is also viewable in Fig. 1 of
514: Kumar \& Panaitescu (2000) (at $\sim$ 400 s), and it is interpreted
515: as due to the passing through the observation band as well. We
516: observe that, at about 1400 s, the softening stops and the spectral
517: index becomes constant. This occurs after the peak energy has
518: sufficiently passing through the observation band. At about 30000 s,
519: the light curve ends with a cutoff tail which is determined by the
520: term $R_{c}/v-t$. As mentioned above, the term $R_{c}/v-t$ comes
521: from the projected factor of the infinitesimal fireball surface area
522: in the angle concerned (say, $\theta $) to the distant observer,
523: known as $\cos \theta $. As the angle between the moving direction
524: of the dominant emission area and the line of sight becomes larger,
525: the $\cos \theta $ term becomes smaller. When $\theta $ approaches
526: to $\pi /2$ (the edge of the half fireball surface that faces the
527: observer), $\cos \theta $ approaches to zero and then the tail comes
528: into being (it is expectable that any emission from a fireball must
529: be limited due to its limited size).
530:
531: The observed XRT light curves are ranging from $60 s$ to $1\times
532: 10^7 s$ and from $5\times 10^{-15} erg\cdot cm^{-2} s^{-1}$ to
533: $8\times 10^{-8} erg\cdot cm^{-2} s^{-1}$. From the data of Fig. 1
534: we find that, at $60s$ $f_{\nu}\sim 20 F_0$, at $1000s$ $f_{\nu}\sim
535: 4\times 10^{-3} F_0$ (where the breaking feature appears), and at
536: $30000s$ $f_{\nu}\sim 1\times 10^{-10} F_0$ (where the light curve
537: cutoff tail, or the broken down feature, emerges). For the adopted
538: parameter set, the flux at $1000s$ is smaller than that at $60s$
539: about 4 orders of magnitude, and then the breaking feature is
540: reasonably expectable. The broken down feature cannot be observable
541: since the flux associated with it is smaller than that at $60s$
542: about 11 orders of magnitude. However, this does not mean that this
543: feature will never be observable since the magnitude of the flux
544: associated with it depends strongly on the fireball radius, the
545: intrinsic peak energy, and the energy indexes of the Band function
546: spectrum. For example, for a much smaller high energy index [say,
547: when $-(1+\beta _{B,0})$ is much smaller] the problem will be
548: significantly eased, and for a smaller fireball radius the interval
549: between the observable start time of the XRT light curve and the
550: broken down feature will be much shorter and the difference between
551: their flux magnitudes will be much smaller (see Fig. 6 and also the
552: discussion below).
553:
554: %\clearpage
555: \begin{figure}[tbp]
556: \begin{center}
557: \includegraphics[width=5in,angle=0]{f2.eps}
558: \end{center}
559: \caption{The $\nu f_{\nu}$ (in units of $F_0 keV$) spectra of the
560: emission considered in Fig. 1, at different observation times. The
561: four dotted lines represent the 0.1, 1, 10, and 100 keV frequencies
562: respectively.} \label{Fig. 1_02}
563: \end{figure}
564: %\clearpage
565:
566: Development of the whole spectrum over the same period concerned is
567: displayed in Fig. 2, where the $\nu f_{\nu }$ curves at 5, 10, 50,
568: 100, 500, 1000, 5000, and 10000 s, spanning from 0.01 to 1000 keV,
569: are presented. This figure plainly illustrates that, due to the
570: contribution of the high latitude emission (where angle $\theta$
571: becomes larger and larger), it is indeed that the shifting of the
572: Band function spectrum causes the softening observed in Fig. 1. When
573: the peak energy has passed through the adopted bandpass (say, the 1
574: keV observation frequency), the higher energy power law portion in
575: the intrinsic Band function spectrum gradually dominates the
576: emission. After $\sim$ 1000 s, the expected flux density (at
577: $h\nu=1keV$) is mainly contributed by this emission (i.e., the $\nu
578: _{0}^{1+\beta _{B,0}}$ portion emission), and then the
579: $t^{-2-\beta}\nu^{-\beta}$ curve comes into being (here
580: $-\beta=1+\beta _{B,0}$). In this period, the spectral index $\beta$
581: is of course constant (see Fig. 1). As mentioned above, the
582: $t^{-2-\beta}\nu^{-\beta}$ curve is deduced by assuming an intrinsic
583: emission with a power law spectrum, emitted from an expanding
584: fireball (Fenimore et al. 1996; Kumar \& Panaitescu 2000; Qin 2008),
585: which is a consequence of equation (11).
586:
587: %\clearpage
588: \begin{figure}[tbp]
589: \begin{center}
590: \includegraphics[width=5in,angle=0]{f3.eps}
591: \end{center}
592: \caption{Evolution of the peak energy $E_{peak}$ (the lower panel)
593: of the observed spectrum of the emission considered in Fig. 1, where
594: the light curve (the upper panel) is the same in Fig. 1.}
595: \label{Fig. 1_03}
596: \end{figure}
597: %\clearpage
598:
599: The evolution of the peak energy is shown in Fig. 3, where the
600: $E_{peak}\propto t^{-1}$ law is visible.
601:
602: %\clearpage
603: \begin{figure}[tbp]
604: \begin{center}
605: \includegraphics[width=5in,angle=0]{f4.eps}
606: \end{center}
607: \caption{Comparison of the temporal index $\alpha$ (the solid line
608: in the lower panel) and spectral index $\beta$ (presented in
609: $2+\beta$) (the dash line in the lower panel) of the emission
610: considered in Fig. 1, where the light curve (the upper panel) is the
611: same in Fig. 1.} \label{Fig. 1_04}
612: \end{figure}
613: %\clearpage
614:
615: The $\alpha =2+\beta $ relation discussed in last section can be
616: directly checked by plotting and comparing the $\alpha$ vs. $t$
617: curve and the $2+\beta$ vs. $t$ curve. This is shown in Fig. 4. When
618: the observation time is not too late (say, $< 2000$ s), the $\alpha
619: =2+\beta $ relation firmly stands. As expected, at later times the
620: $\alpha =2+\beta $ relation is broken down as the $\cos \theta $
621: term (the $R_{c}/v-t$ term) becomes important. This happens at
622: $\sim$ 2000 s. After that, the $\cos \theta $ term dominates the
623: flux, and then the temporal index $\alpha$ rapidly increases.
624:
625: \subsection{Expected at other observation frequencies}
626:
627: Next, let us consider the light curve and spectral evolution
628: expected at other observation frequencies. One finds from Fig. 2
629: that the peak energy passes through 0.1 keV frequency at a much
630: later time while it passes through 100 keV at a very early time.
631: This suggests that for the adopted set of parameters, the softening
632: would appear at lower and very higher energy bands as well, but the
633: corresponding times would be very different. Displayed in Fig. 5 are
634: the light curves and the spectral evolution expected at frequencies
635: 0.1, 1, 10 and 100 keV respectively. The spectral index $\beta$
636: reaches its maximum at $\sim$ $1.3\times 10^4$, $1.3\times 10^3$,
637: $1.3\times 10^2$ and $13$ s, for frequencies 0.1, 1, 10 and 100 keV,
638: respectively.
639:
640: %\clearpage
641: \begin{figure}[tbp]
642: \begin{center}
643: \includegraphics[width=5in,angle=0]{f5.eps}
644: \end{center}
645: \caption{Comparison of the light curve (the upper panel; the units
646: are the same as in Fig. 1) and the spectral evolution (the lower
647: panel) of the same emission considered in Fig. 1, expected at
648: different observation frequencies. In the lower panel, solid lines
649: from the left to the right correspond to observation frequencies
650: 100, 10, 1, and 0.1 keV respectively, whilst in the upper panel,
651: solid lines from the bottom to the top are associated with 100, 10,
652: 1, and 0.1 keV respectively.} \label{Fig. 1}
653: \end{figure}
654: %\clearpage
655:
656: This analysis leads to the following conclusions: the softening due
657: to the curvature effect will appear at different frequencies; it
658: occurs earlier for higher frequencies and later for lower
659: frequencies; the terminating softening time (defined as the time
660: when the spectral index $\beta$ reaches its maximum), $t_{s,max}$,
661: depends strictly on the observation frequency, and it follows the
662: $t_{s,max}\propto \nu^{-1}$ law, where $\nu$ is the observation
663: frequency. Note that the $t_{s,max}\propto \nu^{-1}$ law can also be
664: deduced from equation (10) and/or equation (11). This prediction
665: holds as long as the intrinsic emission is short enough.
666:
667: Indeed, the spectral softening evolution was observed in the prompt
668: emission of GRB 070616, starting from 285s and extending to 1200s
669: (Starling et al. 2008). This favors our new finding that the
670: softening is also expectable in higher energy bands.
671:
672: \section{Parameters that affect the results}
673:
674: The light curve and spectral evolution discussed in last section are
675: produced by adopting a certain set of parameters. Here we
676: investigate how these parameters affect the results.
677:
678: \subsection{Lorentz factor, fireball radius and peak energy of the intrinsic
679: spectrum}
680:
681: The softening time scale exhibited in Fig. 1 must be affected by the
682: Lorentz factor, fireball radius and peak energy of the intrinsic
683: spectrum. We repeat the above analysis by replacing $\Gamma =100$,
684: $R_{c}=10^{15}cm$ and $h\nu _{0,p}=E_{0,p}=1keV$ with $\Gamma =10$,
685: $R_{c}=10^{14}cm$ and $h\nu _{0,p}=E_{0,p}=0.1keV$ one by one. The
686: results are displayed in Fig. 6.
687:
688: %\clearpage
689: \begin{figure}[tbp]
690: \begin{center}
691: \includegraphics[width=5in,angle=0]{f6.eps}
692: \end{center}
693: \caption{The light curve (the upper panel) and spectral evolution
694: (the lower panel) of the emission considered in Fig. 1 (in the same
695: units), produced by replacing the parameters adopted there. In the
696: lower panel, solid lines from the left to the right correspond to
697: the cases of replacing $R_{c}=10^{15}cm$ with $R_{c}=10^{14}cm$,
698: without replacement, and replacing $\Gamma =100 $ with $\Gamma =10
699: $, respectively, and the dash line corresponds to the case of
700: replacing $h\nu _{0,p}=1keV$ with $h\nu _{0,p}=0.1keV$. In the upper
701: panel, solid lines from the bottom to the top are associated with
702: the cases of replacing $R_{c}=10^{15}cm$ with $R_{c}=10^{14}cm$,
703: without replacement, and replacing $\Gamma =100 $ with $\Gamma =10
704: $, respectively, and the dash line corresponds to the case of
705: replacing $h\nu _{0,p}=1keV$ with $h\nu _{0,p}=0.1keV$.} \label{Fig.
706: 1_05}
707: \end{figure}
708: %\clearpage
709:
710: We find that a smaller Lorentz factor extend both the light curve
711: (marked by its breaking feature which corresponds to the moment when
712: the peak energy passes though the observation frequency) and the
713: spectral softening to a larger time scale. This must be resulted
714: from the less contraction of time. The cutoff tail of the steep
715: decay phase remains in the same time position. This is not surprise
716: since the tail is associated with the fireball radius, entirely
717: independent of the Lorentz factor (see Qin 2008).
718:
719: As expected, in the case of $R_{c}=10^{14}cm$, the softening and the
720: cutoff tail (as well as the breaking feature) shift to earlier time.
721: The softening appears as early as $<3$ s and ends as early as
722: $\sim100$ s, and the cutoff tail appears at $\sim 3000$ s. At $60s$
723: we find $f_{\nu}\sim 0.3 F_0$, and at $3000s$ (where the light curve
724: cutoff tail appears) $f_{\nu}\sim 1\times 10^{-9} F_0$. For this
725: adopted parameter set, the flux at $3000s$ is smaller than that at
726: $60s$ about 8 orders of magnitude, which is only 1 order of
727: magnitude smaller than the usual observation magnitude range (7
728: orders). The undetectable problem raised above is now largely eased.
729:
730: In the case of $h\nu _{0,p}=E_{0,p}=0.1keV$, the spectral softening
731: curve is almost overlapped with that of $ R_{c}=10^{14}cm$. Compared
732: with the case of $h\nu _{0,p}=E_{0,p}=1keV$, the breaking feature
733: shifts to earlier time, whilst the cutoff tail remains in the same
734: time position.
735:
736: We notice from Fig. 6 together with Fig. 5 that the softening
737: process, associated with various parameters, spans a time scale
738: comparable to the terminating softening time $t_{s,max}$. In terms
739: of mathematics, this is due to the fact that, relative to the moment
740: $t=0$, the start time of the softening is much smaller (more than
741: one order of magnitude smaller) than $t_{s,max}$. We come to this
742: conclusion: the time interval of the softening is in the same order
743: of magnitude of the terminating softening time, and the two
744: quantities much be linearly correlated. We suspect that it is the
745: geometric property of the fireball surface that gives rise to this
746: relation.
747:
748: \subsection{High and low energy spectral indexes}
749:
750: High and low energy spectral indexes of the Band function must have
751: influences on the spectral softening. We replace $1+\alpha
752: _{B,0}=-0.5$ with $1+\alpha _{B,0}=0$ and $1+\alpha _{B,0}=-0.8$,
753: and replace $1+\beta _{B,0}=-2.5$ with $1+\beta _{B,0}=-2$ and
754: $1+\beta _{B,0}=-5$ respectively to produce the light curve and the
755: spectral evolutionary curve. In doing so, other parameters remain
756: unchanged.
757:
758: %\clearpage
759: \begin{figure}[tbp]
760: \begin{center}
761: \includegraphics[width=5in,angle=0]{f7.eps}
762: \end{center}
763: \caption{The light curve (the upper panel) and spectral evolution
764: (the lower panel) of the emission considered in Fig. 1 (in the same
765: units), produced by replacing the indexes adopted there. The dash,
766: dot, dash dot, and dash dot dot lines correspond to the cases of
767: replacing $1+\alpha _{B,0}=-0.5$ with $1+\alpha _{B,0}=0$, replacing
768: $1+\alpha _{B,0}=-0.5$ with $1+\alpha _{B,0}=-0.8$, replacing
769: $1+\beta _{B,0}=-2.5$ with $1+\beta _{B,0}=-2$, and replacing
770: $1+\beta _{B,0}=-2.5$ with $1+\beta _{B,0}=-5$, respectively. The
771: two solid lines in Fig. 1 are presented in gray solid lines.}
772: \label{Fig. 1_06}
773: \end{figure}
774: %\clearpage
775:
776: The results are shown in Fig. 7. As expected, both indexes have
777: influences on the softening curve: the low energy index $\alpha
778: _{B,0}$ puts a lower limit to the observed spectral index $\beta$,
779: making $\beta_{min} = - (1+\alpha _{B,0})$; and the high energy
780: index $\beta _{B,0}$ confines the upper limit of $\beta$, following
781: $\beta_{max}=-(1+\beta _{B,0})$. The ratio between the fluxes at
782: $60s$ and at $3000s$ is also influenced by the indexes, making the
783: undetectable problem to be eased or worse.
784:
785: Combining Figs. 5 and 7 we come to the following conclusions: the
786: observed $\beta_{min}$ and $\beta_{max}$ are determined by the low
787: and high energy indexes of the observed Band function spectrum (note
788: that in the case of $\delta$ function emission the observed spectrum
789: and the intrinsic spectrum share the same form); the observed
790: $\beta_{min}$ and $\beta_{max}$ for different observation
791: frequencies would be unchanged as long as the whole softening
792: process appears within the whole available observation time at the
793: concerned frequencies.
794:
795: \section{Application}
796:
797: \subsection{The light curve and spectral evolution expected in the XRT band}
798:
799: Our theoretical analysis carried above does not directly correspond
800: to real observational data. In fact, instead of being defined at a
801: particular observation frequency, the XRT light curve is measured
802: within an energy band which is $0.3 - 10$ keV. Therefore, it is
803: necessary to investigate the light curve as well as the spectral
804: evolution over this energy range. The available flux of the XRT
805: light curve is always that has been integrated over this band. We
806: use $f_{\nu,int}$ to denote this flux which is determined by
807: \begin{equation}
808: f_{\nu,int}(t)=\int_{0.3keV}^{10keV}f_{\nu }(t)d\nu.
809: \end{equation}
810: The flux is now in units of $erg\cdot cm^{-2} s^{-1}$.
811:
812: There is a difficulty in evaluating the spectral index over a band.
813: If the band is large enough one might not be able to consider it
814: acting still as a power-law. Although we can impose a power-law on
815: the spectrum over this band, the power-law index is still hard to be
816: defined and hence hard to be determined. In terms of observation, we
817: can collect all data points within this band and then figure out the
818: index by performing a power-law fit. This method is hard to be
819: adopted in theoretical investigation since one can create countless
820: data points. We therefor turn to consider a simpler but well defined
821: approach. First, we assume a power-law over the whole XRT band and
822: then calculate the index by considering only the fluxes at the lower
823: (0.3 keV) and upper (10 keV) limits of the band. Second, a power-law
824: is assumed over a smaller band ($0.6 - 5$ keV) and then the index is
825: calculated by employing only the fluxes at the new lower (0.6 keV)
826: and upper (5 keV) limits. Presented in Fig. 8 are the spectral
827: evolution so evaluated and the light curve of (14), where the Band
828: function is also adopted and parameters other than the observation
829: frequency are the same as those adopted in Fig. 1.
830:
831: %\clearpage
832: \begin{figure}[tbp]
833: \begin{center}
834: \includegraphics[width=5in,angle=0]{f8.eps}
835: \end{center}
836: \caption{The light curve over the XRT band (the dot line in the
837: upper panel; in unites of $F_0 keV$) and the corresponding spectral
838: evolution (the lower panel) of the emission considered in Fig. 1.
839: The dot line in the lower panel represents the spectral index
840: evaluated by considering a power-law spanning from 0.3 keV to 10
841: keV; the dash line stand for that calculated by assuming a power-law
842: spanning from 0.6 keV to 5 keV. For the sake of comparison, both
843: solid lines in Fig. 1 (in the same units as adopted there) are also
844: presented.} \label{Fig. 1_01}
845: \end{figure}
846: %\clearpage
847:
848: The upper panel of Fig. 8 shows that the XRT light curve is very
849: similar to the 1 keV light curve. The power-law index $\alpha$ of
850: the former seems slightly different from that of the latter. This
851: will lead to a deviation from the $\alpha =2+\beta $ relation. From
852: the lower panel we find that the lower and upper limits of the
853: corresponding spectral index ($\beta_{min}$ and $\beta_{max}$) are
854: the same as that measured at 1 keV, but the spectral evolutionary
855: curve deviates significantly from that measured at 1 keV. The
856: deviation is so large that the $\alpha =2+\beta $ relation would be
857: violently violated. This might help us to understand why this
858: relation is not commonly detected. The lower panel also shows that
859: the narrower the power-law range assumed, the closer the spectral
860: evolutionary curve to that measured at 1 keV. This suggests that,
861: the narrower band to concern, the more chance of detecting the
862: $\alpha =2+\beta $ relation.
863:
864: \subsection{GRB 060614}
865:
866: Although the condition that the emission is extremely short might be
867: rare, there might be some bursts their early X-ray emission can
868: roughly be accounted for by equation (11). Once a burst is selected,
869: there are two ways of testing. One is to directly fit the light
870: curve data and the spectral data with equation (11). The other is to
871: check if their temporal and spectral indexes obey the $\alpha
872: =2+\beta $ relation. We adopt the second method since the result
873: does not depend on fitting parameters.
874:
875: Among the bursts (up to March 28, 2008) analyzed by the UNLV GRB
876: Group (see http://grb.physics.unlv.edu), GRB 060614 might be one of
877: such bursts that can be accounted for by the $\delta$ function
878: emission curvature effect model. There is an obvious softening in
879: the steep decay phase for this burst. The light curve in this phase
880: is relatively smooth, suggesting that, besides the main decay
881: emission, it is unlikely that other components obviously influence
882: the light curve. In this way, the temporal index can be well
883: evaluated.
884:
885: %\clearpage
886: \begin{deluxetable}{cccc}
887: \tablewidth{0pt} \tablecaption{Temporal indexes of GRB 060614.}
888: \tablehead{ \colhead{section} & \colhead{$t_1$(s)} &
889: \colhead{$t_2$(s)} & \colhead{$\alpha$} } \startdata
890:
891: 1 & 103.3 & 126.8 & 2.73 $\pm$ 0.23 \\
892: 2 & 126.8 & 168.8 & 2.25 $\pm$ 0.13 \\
893: 3 & 169.3 & 309.8 & 3.841 $\pm$ 0.064 \\
894: 4 & 315.0 & 408.4 & 3.78 $\pm$ 0.25 \\
895: 5 & 408.4 & 468.8 & 5.89 $\pm$ 0.48 \\
896: \enddata
897: %\tablenotetext{a}{in units of $s$.}
898: \end{deluxetable}
899: %\clearpage
900:
901: According to the quality of the light curve data of this burst, we
902: divide them into five sections in this phase, requiring that, for
903: each section, $log f_{\nu}$ and $log t$ are linearly correlated. We
904: then estimate the temporal index in each section by fitting the
905: corresponding data with a power law function. Time intervals of
906: these sections as well as the fitting results (the estimated
907: temporal index $\alpha$) are listed in Table 1, where $t_1$ and
908: $t_2$ are the lower and upper limits of the observation time of the
909: corresponding sections respectively.
910:
911: %\clearpage
912: \begin{figure}[tbp]
913: \begin{center}
914: \includegraphics[width=5in,angle=0]{f9.eps}
915: \end{center}
916: \caption{Comparison of the temporal index in $\alpha-2$ (the filled
917: circles in the lower panel) and spectral index $\beta$ (the open
918: circles in the lower panel) of GRB 060614 in its steep decay phase.
919: The data of the flux (the upper panel; in units of $erg \cdot
920: cm^{-2} s^{-1}$) and $\beta$ are taken from the UNLV GRB Group
921: website.} \label{Fig. 1_07}
922: \end{figure}
923: %\clearpage
924:
925: The temporal index obtained from the light curve by the fit and the
926: spectral index measured in the X-ray band by the UNLV GRB Group are
927: presented in Fig. 9. We find that the $\alpha -2$ vs. $t$ curve is
928: roughly in agreement with the $\beta$ vs. $t$ curve in the concerned
929: steep decay phase, suggesting that the curvature effect might
930: probably be the main cause of the steep decay curve.
931:
932: As analyzed in last subsection, the $\alpha =2+\beta $ relation will
933: not be strictly obeyed if one considers the light curve and the
934: spectral index over a band (here, the XRT band) instead of at a
935: fixed frequency. A deviation between the $\alpha -2$ vs. $t$ curve
936: and the $\beta$ vs. $t$ curve is hence expectable. However, the
937: temporal index measured in the last time section in this phase is so
938: large that it is likely to have other causes. Although the effect of
939: the light curve over a band and the effect of the duration of the
940: intrinsic emission have not been considered, we still suspect that
941: the large temporal index might be a consequence of larger absorption
942: for higher latitude photons.
943:
944: \section{Discussion and conclusions}
945:
946: We investigate the influence of the curvature effect on both the
947: light curve and the spectrum of late emission of GRBs, attempting to
948: explain the observed softening phenomenon in early X-ray afterglows
949: of Swift. As the first step of investigation, we explore only the
950: case of extremely short intrinsic emission, for which we assume and
951: apply a $\delta$ function emission. Although how an emission is
952: extremely short is currently unclear (this deserves a detailed
953: investigation in the near future) and the condition that the
954: emission is extremely short might be rare, the investigation is
955: necessary since by considering such emission the possible impacts
956: from the emission duration can be ignored and then the pure
957: curvature effect can plainly be illustrated.
958:
959: Formulas presented in Qin (2008) are employed to study this issue.
960: Unlike what investigated in Qin (2008) and other relevant
961: theoretical analyses (e.g., Fenimore et al. 1996; Sari et al. 1998;
962: Kumar \& Panaitescu 2000), we do not limit our study on an intrinsic
963: power law spectrum. Instead, we consider more general spectral form
964: of emission. Assuming a $\delta$ function emission we get formulas
965: that get rid of the impacts from the intrinsic emission duration,
966: which are applicable to any forms of spectrum. According to these
967: formulas, one would observe the same form of spectrum at different
968: times, with the peak energy of the spectrum shifting from higher
969: energy bands to lower bands. This was detected recently in the early
970: X-ray afterglows of some GRBs (see Butler \& Kocevski 2007). The
971: peak energy is expected to decline following $E_{peak}\propto
972: t^{-1}$. In the case when the emission is early enough so that the
973: emitting area on the fireball surface is not far from the line of
974: sight (say, $\theta \ll \pi/2$ or $t \ll R_{c}/v$), the temporal
975: power law index and the spectral power law index will be well
976: related by $\alpha =2+\beta $. As a consequence, the form
977: $t^{2}f_{\nu }(t)$ will possess exactly the intrinsic spectral form
978: (say, when replacing $t$ with $\nu$ one will get from $t^{2}f_{\nu
979: }(t)$ as a function of $\nu$ that take exactly the same form of the
980: intrinsic spectrum). The $\alpha =2+\beta $ relation will be broken
981: down and $\alpha
982: > 2+\beta $ or $\alpha \gg 2+\beta $ will hold (see Fig. 4) at much later time when
983: the angle between the moving direction of the emission area and the
984: line of sight is large.
985:
986: As revealed in Butler \& Kocevski (2007) and suggested by Starling
987: et al. (2008), we focus our attention to the emission with a Band
988: function spectrum. This spectrum has a power law behavior in both
989: lower and higher energy bands, where the two power laws are smoothly
990: connected (Band et al. 1993). Using this spectral form, we plot the
991: light curve and the spectral evolution with our formulas, expected
992: at 1 keV and other frequencies. The analysis shows that there do
993: exist a temporal steep decay phase and a spectral softening which
994: occur simultaneously. As Fig.2 reveals, both the steep decay light
995: curve and the spectral softening are caused by the shifting of the
996: Band spectrum. As mentioned above, Starling et al. (2008) suggested
997: that the steep decline in the flux of GRB 070616 may be caused by a
998: combination of the strong spectral evolution and the curvature
999: effect. Based on the above argument, we insist that both the
1000: spectral evolution and the steep decline observed in GRB 070616 and
1001: other Swift bursts are likely to be caused merely by the curvature
1002: effect. In addition, we find that, just as what is illustrated in
1003: Qin (2008), the $cos\theta$ term, $R_{c}/v-t$, plays a role in
1004: producing the light curve, which ``attaches'' a cutoff tail to the
1005: latter (see Fig. 1). The spectral softening terminates when the
1006: emission is dominated by that from the high energy portion of the
1007: Band function spectrum. Thus, there exists a maximum of the spectral
1008: index, $\beta_{max}$. After the $\beta_{max}$ appears, it lasts to
1009: the end of the emission, including the phase of the cutoff tail. Our
1010: analysis shows that the softening due to the curvature effect will
1011: appear at different frequencies; it occurs earlier for higher
1012: frequencies and later for lower frequencies; a characteristic of the
1013: softening, the terminating softening time $t_{s,max}$ (when the
1014: $\beta_{max}$ appears), depends strictly on the observation
1015: frequency, which follows $t_{s,max}\propto \nu^{-1}$. Although this
1016: is concluded based on the assumption of extremely short emission, we
1017: tend to believe that its main characters hold in most cases since
1018: the light curve from any finite emission is contributed by countless
1019: $\delta$ function emission. The combination of these countless
1020: $\delta$ function emission would change the values of some
1021: quantities such as $t_{s,max}$, but would not change the trend.
1022: Starling et al. (2008) showed, the spectral evolution of GRB 070616
1023: starts earlier at $ \gamma$-ray energies (while the X-ray flux is
1024: still at an approximately constant level) and begins much later at
1025: X-ray energies around the onset of the steep X-ray decay. In terms
1026: of the curvature effect, this is due to the following fact: the peak
1027: energy of the Band function spectrum passes through the $\gamma$-ray
1028: band at a relatively early time, while it passes through the X-ray
1029: band at a later time.
1030:
1031: Parameters that might have impacts on the light curve and the
1032: spectral evolution are also studied. Whilst a smaller Lorentz factor
1033: shifts the spectral softening to larger time scales, smaller values
1034: of the fireball radius and the rest frame peak energy make the
1035: occurrence of the softening earlier. The analysis shows that the
1036: terminating softening time appears much later than the start time of
1037: the softening. The former is always larger than the latter by about
1038: one order of magnitude. It is therefore predicted that the duration
1039: of the softening and the terminating softening time would be
1040: linearly correlated. It also shows that the low energy index $\alpha
1041: _{B,0}$ puts a lower limit to the observed spectral index $\beta$
1042: and the high energy index $\beta _{B,0}$ confines the upper limit of
1043: $\beta$. The following conclusions are reached: the observed
1044: $\beta_{min}$ and $\beta_{max}$ are determined by the low and high
1045: energy indexes of the observed Band function spectrum; $\beta_{min}$
1046: and $\beta_{max}$ for different observation frequencies would remain
1047: unchanged as long as the whole softening process appears within the
1048: whole available observation time at the concerned frequencies.
1049:
1050: As application, we study the light curve and the spectral evolution
1051: over the XRT band. That is, the light curve is that has been
1052: integrated over the $0.3-10$ keV band and the spectral index is
1053: estimated by assuming a power-law over this band. The analysis shows
1054: that the light curve slightly deviate from that measured at 1 keV,
1055: whilst the spectral evolutionary curve significantly betrays that
1056: measured at 1 keV. This suggests that the $\alpha =2+\beta $
1057: relation will be violently violated if one measures the light curve
1058: and estimates the spectral index over a wide band.
1059:
1060: Another application is to check the $\alpha =2+\beta $ relation by
1061: employing the light curve and spectrum data of GRB 060614. The
1062: temporal index $\alpha$ in the steep decay phase of this burst is
1063: evaluated. We compare $\alpha$ and the spectral index $\beta$ in the
1064: same softening phase (it is also the steep decay phase of the light
1065: curve). It shows that the $\alpha-2$ vs. $t$ curve and the $\beta$
1066: vs. $t$ curve are roughly in agreement, suggesting that the $\alpha
1067: =2+\beta $ relation is roughly satisfied in this phase and the
1068: softening of this burst might possibly be due to the curvature
1069: effect.
1070:
1071: What we have investigated are based on the assumption of extremely
1072: short emission which ignores the possible impacts from the emission
1073: duration. While the contribution of the emission duration might have
1074: less affect on the spectrum, it might obviously affect the temporal
1075: profile and hence the temporal index. Therefore the $\alpha =2+\beta
1076: $ relation might not hold in this case. Combining this with the
1077: problem arising from the estimation of the spectral index in the
1078: softening phase over an energy band rather than at a fixed
1079: frequency, it might be able to explain why the $\alpha=2+\beta$
1080: relation is not common in Swift bursts (see Liang et al. 2006 and
1081: also Zhang 2007 for a detailed discussion). Yonetoku et al. (2008)
1082: showed that the two characteristic break energies they considered
1083: have a time dependence of $ \propto t^{-3}-t^{-4}$. This is not in
1084: agreement with the prediction of the $E_{peak}\propto t^{-1}$ law.
1085: Perhaps an intrinsic softening might be responsible for this
1086: difference when the emission duration is taken into account. Another
1087: possibility is that the $E_{peak}$ shifting of this kind of burst is
1088: due to structure jets, where in high latitude, $E_{peak}$ would be
1089: much smaller than it is in uniform jets. In addition to test the
1090: $\alpha =2+\beta $ relation and the $E_{peak}\propto t^{-1}$ law, we
1091: suggest to detect the softening at different frequencies as well.
1092: The observed relations might not strictly follow what the $\delta$
1093: function emission model predicts, but as argued above, the trend of
1094: the relevant effects would be maintained (which deserves a detailed
1095: investigation in the near future).
1096:
1097: %\clearpage
1098: \begin{figure}[tbp]
1099: \begin{center}
1100: \includegraphics[width=5in,angle=0]{f10.eps}
1101: \end{center}
1102: \caption{Comparison of the functions of the $t^{3}f_{\nu }(t)$ vs.
1103: $t$ curve (the upper panel) and the $\nu f_{\nu}$ vs. $\nu$ curve
1104: (the spectrum; the lower panel) of the emission considered in Fig.
1105: 1. The three solid lines in the lower panel are the same lines of
1106: 50, 100, and 500 s respectively in Fig. 1.} \label{Fig. 1_08}
1107: \end{figure}
1108: %\clearpage
1109:
1110: In addition to these more conventional tests, we propose to try a
1111: totally new test which is to check if the $t^{3}f_{\nu }(t)$ vs. $t$
1112: curve is in agreement with the $\nu f_{\nu}$ vs. $\nu$ curve when
1113: replacing $t$ with $\nu$ in the $t^{3}f_{\nu }(t)$ and $t$ forms and
1114: multiplying a constant to match the corresponding dimensions (see
1115: Section 3). Illustrated in Fig. 10 is an example of the comparison
1116: (where we compare only the relevant functions and thus do not
1117: replace variables or multiply constants to change the dimensions).
1118: An advantage of doing so is that we can guess spectrum form merely
1119: from the light curve data. For example, when we find that the
1120: $t^{3}f_{\nu }(t)$ form being a perfect power-law function of time,
1121: then the spectrum is guessed to be a pure power-law (if the spectral
1122: data are found not to obey a power-law, then we will have reasons to
1123: doubt if this burst is not affected by the curvature effect). Or,
1124: when we find that the $t^{3}f_{\nu }(t)$ form is a Band function of
1125: time, then we will have reasons to guess that the spectrum takes a
1126: Band function form. Another usage of plotting the $t^{3}f_{\nu }(t)$
1127: vs. $t$ curve is to find out the time when the peak energy passes
1128: through the observation band (se also Kumar \& Panaitescu 2000),
1129: which can be directly checked by observation and hence becomes a
1130: test to the curvature effect as well. According to equation (11),
1131: the moment when the peak of $t^{3}f_{\nu }(t)$ appears is the time
1132: when the peak energy passes through the observation frequency $\nu$,
1133: or, it is the time when $E_{peak}=h\nu$. In plotting the
1134: $t^{3}f_{\nu }(t)$ vs. $t$ curve (see the upper panel of Fig. 10),
1135: one can also check the influence of factors other than that of the
1136: pure curvature effect by observing and measuring the deviation of a
1137: real light curve from that of equation (11). An extra usage of this
1138: plot might be a direct comparison of light curves of different
1139: bursts, probably enabling us to divide them according to their
1140: temporal properties (we strongly suggest a detailed investigation on
1141: this issue in the near future).
1142:
1143: %\clearpage
1144: \begin{figure}[tbp]
1145: \begin{center}
1146: \includegraphics[width=5in,angle=0]{f11.eps}
1147: \end{center}
1148: \caption{The $t^{3}f_{\nu }(t)$ (in units of $erg \cdot cm^{-2}
1149: s^{2}$) vs. $t$ curve of GRB 060614 in its steep decay phase. The
1150: data are taken from the UNLV GRB Group website.} \label{Fig. 1_09}
1151: \end{figure}
1152: %\clearpage
1153:
1154: Displayed in Fig. 11 is the $t^{3}f_{\nu }(t)$ vs. $t$ curve of GRB
1155: 060614. It shows a function of time bearing the Band function form
1156: (when replacing $\nu$ with $t$) attached with a cutoff tail. If we
1157: believe that the softening of this burst is due to the curvature
1158: effect in the case of extremely short emission, according to Fig. 11
1159: it would be expectable that the peak energy of this burst passes
1160: through the corresponding observation energy range (the XRT band) at
1161: $\sim $175 s (or, at $\sim $175 s, the peak energy of the observed
1162: spectrum is just within the energy range of the light curve).
1163: Indeed, as presented in Table 4 of Mangano et al. (2007), within the
1164: time interval $128-190$ s, the peak energy (when fitting the
1165: spectrum with a Band model) is about $2-4$ keV which is well within
1166: the $0.3-10$ keV band. A similar result for this burst is also
1167: visible in Fig. 1 of Butler \& Kocevski (2007).
1168:
1169: Based on their fits to the composite light curves, Sakamoto et al.
1170: (2007) confirmed the existence of an exponential decay component
1171: which smoothly connects the BAT prompt data to the XRT steep decay
1172: for several GRBs. Yonetoku et al. (2008) also showed that the
1173: spectrum of GRB 060904A contains a cutoff tail in its higher energy
1174: range, which can be represented by an exponential function.
1175: According to the above analysis, the spectral form obviously affects
1176: the light curve if the curvature effect is at work and the intrinsic
1177: emission is short enough. An intrinsic spectrum with an exponential
1178: tail might probably lead to an exponential decay light curve. We
1179: notice that the projected factor (the $\cos \theta $ term) also
1180: produces a very steep decay phase when the angle between the moving
1181: direction of the dominant emission area and the line of sight is
1182: large enough. However, it is unlikely that many of them (if not only
1183: few of them) are due to the $cos\theta$ term, since this term always
1184: appears at a very late time (see Fig. 6 and also Qin 2008). The
1185: following factors might also be the cause of this tail: one is the
1186: large absorption for higher latitude emission and the other is the
1187: limited open angle of jets. Both will lead to a steeper tail. While
1188: the former might probably give rise to a smooth decay curve, the
1189: latter might probably lead to a sharp feature.
1190:
1191: In addition to the shifting of the peak energy, Starling et al.
1192: (2008) also observed softening of the low energy power law slope.
1193: They measured a softening of the low energy spectral slope from
1194: $\alpha \sim 0.1 - 1.3$. This implies that the intrinsic spectrum
1195: might evolved itself. This cannot be taken into account in the
1196: current investigation since a $\delta$ function emission has no
1197: evolution. We suggest to explore in a later investigation the impact
1198: from the emission duration and the effect from the intrinsic
1199: spectral evolution.
1200:
1201: \acknowledgments
1202:
1203: Special thanks are given to the anonymous referee for his or her
1204: comments and suggestions which have improved the paper greatly. This
1205: work was supported by the National Natural Science Foundation of
1206: China (No. 10573005 and No. 10747001) and by the Guangzhou Education
1207: Bureau and Guangzhou Science and Technology Bureau.
1208:
1209: \appendix
1210:
1211: \section{Projected factor as a function of time}
1212:
1213: Emission from a distant area is proportional to the projected factor
1214: of the area relative to the observer, which is known as $cos\theta$,
1215: where $\theta$ is the angle between the normal of the area and the
1216: line of sight. For a face-on area (its normal is parallel to the
1217: line sight), $\theta=0$ and then $cos\theta=1$; and for an edge-on
1218: area (its normal is perpendicular to the line sight), $\theta=\pi/2$
1219: and then $cos\theta=0$. The projected factor varies from
1220: $cos\theta=1$ to $cos\theta=0$ for the half fireball surface facing
1221: the observer. Due to time delation, emission from the area of
1222: $\theta=0$ reaches the observer earlier and that of $\theta=\pi/2$
1223: reaches the observer later, and the former emission is not reduced
1224: by the projected factor whilst the latter will be reduced to zero.
1225: It has been pointed out that the term $R_{c}/v-t$ in equation (11)
1226: comes from nothing but the projected factor $cos\theta$ (Qin 2008).
1227: Here we analyze this issue in detail.
1228:
1229: Taking into account the increase of the radius of an expanding
1230: fireball (equation A.3 in Qin 2002) and the time delay associated
1231: with different latitudes in the fireball surface (equation A.6 in
1232: Qin 2002), one obtains the following relation between the
1233: observation time $t_{ob}$ and the angle relative to the line of
1234: sight (equation A.7 in Qin 2002)
1235: \begin{equation}
1236: t_{emit}=\frac{t_{ob}-D/c+[R_{c}/c-(v/c)t_{c}]cos\theta}{1-
1237: (v/c)cos\theta },
1238: \end{equation}
1239: which could be written as
1240: \begin{equation}
1241: cos\theta=\frac{t_{emit}+D/c-t_{ob}}{R_{c}/c+(v/c)(t_{emit}-t_{c})}.
1242: \end{equation}
1243: Replacing $t_{ob}$ with $t$ (see equation 1) and considering the
1244: time contraction $t_{emit}-t_c=\Gamma (t_0-t_{0,c})$ (equation A.1
1245: in Qin 2002), we get
1246: \begin{equation}
1247: cos\theta=\frac{(t_0-t_{0,c})\Gamma +R_c/v-t}{[(t_0-t_{0,c})v\Gamma
1248: +R_{c}]/c}.
1249: \end{equation}
1250: As revealed in equation (A.4) in Qin (2002), $(t_0-t_{0,c})v\Gamma
1251: +R_{c}$ is the fireball radius measured at $t_0$. Equation (A3)
1252: suggests that, for any particular intrinsic emission time $t_0$ (at
1253: which the fireball radius is fixed), $cos\theta$ varies with
1254: observation time $t$. This is due to the time delay, as mentioned
1255: above (see also equation A.6 in Qin 2002). According to equation
1256: (A3), the term $(t_0-t_{0,c})\Gamma +R_c/v-t$ in equation (2) is the
1257: projected factor $cos\theta$ multiplying the fireball radius divided
1258: by the speed of light. Note that the term $(t_0-t_{0,c})\Gamma
1259: +R_c/v-t$ in equation (2) directly comes from the term $cos\theta$
1260: in equation (A.15) in Qin (2002) as a consequence of the projected
1261: effect (see equation A.11 in Qin 2002).
1262:
1263: For the $\delta$ function emission considered in this paper, the
1264: intrinsic emission is assigned to occur at $t_0=t_{0,c}$ (see
1265: equation 7), which leads to
1266: \begin{equation}
1267: cos\theta=\frac{R_c/v-t}{R_{c}/c}.
1268: \end{equation}
1269: This explains why we interpret the term $R_c/v-t$ in equation (11)
1270: as that reflecting the projected factor of the emission area in the
1271: fireball surface.
1272:
1273: Note that observation time $t$ is confined by equation (8) which
1274: arises from the limit of the fireball surface in the case of
1275: $\delta$ function emission (see Qin 2002; Qin et al. 2004; Qin
1276: 2008). According to equation (A4), $cos\theta=1$ when
1277: $t=(1-v/c)R_c/v$ which is the observation time when photons from the
1278: face-on area ($\theta=0$) of the fireball surface arrive the
1279: observer. When taking $t=R_c/v$ we get $cos\theta=0$. Note that
1280: $t=R_c/v$ is the observation time when photons from the edge-on area
1281: ($\theta=\pi/2$) reach the observer.
1282:
1283: In the case of $R_c=10^{15}cm$ and $\Gamma=100$, $R_c/v \sim 3\times
1284: 10^4 s$ which is much larger than the steep decay segment of most
1285: GRBs detected by Swift. Even for $R_c=10^{14}cm$ we get $R_c/v \sim
1286: 3\times 10^3 s$ which is still larger than the steep decay segment
1287: of many Swift GRBs. For the observation time satisfying $t\ll
1288: R_c/v$, the projected factor approaches a unit and then can be
1289: ignored. Only when the observation time is comparative to $R_c/v$,
1290: the projected factor will play an important role (see the broken
1291: down feature in Fig. 1). The broken down feature also exists in the
1292: case of the pure power-law spectrum (Qin 2008). When this feature is
1293: observed, the fireball radius will be well determined, no matter the
1294: observed spectrum is a pure power-law or a Band function form.
1295:
1296:
1297: \section{Projected factor in flux}
1298:
1299: Here we show how the projected factor plays a role in the flux per
1300: unit frequency from a relativistic source.
1301:
1302: As shown in Qin (2002), the amount of energy emitted from
1303: differential area $ds_{emit}$ radiating towards the observer is
1304: \begin{equation}
1305: dE_{emit}=\frac{I_{\nu }cos\theta ds_{emit}d\nu dt \Delta
1306: s_{ob}}{D^2},
1307: \end{equation}
1308: where $I_{\nu }$ is the intensity of radiation measured by local
1309: observers near the emitting source, $cos\theta$ is the projected
1310: factor of $ds_{emit}$ towards the distant observer, $d\nu$ and $dt$
1311: are the observed frequency and time intervals respectively, $\Delta
1312: s_{ob}$ is the area of the detector, and $D$ is the distance between
1313: the observer and the emitter. The flux density measured from this
1314: amount of energy is
1315: \begin{equation}
1316: d f_{\nu}=\frac{I_{\nu }cos\theta ds_{emit}}{D^2}.
1317: \end{equation}
1318:
1319: Observer frame intensity $I_{\nu }$ is related with the intrinsic
1320: intensity $I_{0,\nu }$ by
1321: \begin{equation}
1322: I_{\nu }=(\frac{\nu}{\nu_0})^3I_{0,\nu },
1323: \end{equation}
1324: where $\nu_0$ is the intrinsic frequency. Applying this relation and
1325: the Doppler effect we get
1326: \begin{equation}
1327: d f_{\nu}=\frac{I_{0,\nu }cos\theta ds_{emit}}{D^2\Gamma^3
1328: [1-(v/c)cos\theta]^3}.
1329: \end{equation}
1330: If the area is on the tip of the fireball surface, $\theta=0$. That
1331: gives rise to
1332: \begin{equation}
1333: d f_{\nu,0}=\frac{I_{0,\nu } ds_{emit}}{D^2\Gamma^3 [1-(v/c)]^3}.
1334: \end{equation}
1335: The relation between $d f_{\nu}$ and $d f_{\nu,0}$ is
1336: \begin{equation}
1337: \frac{d f_{\nu}}{d
1338: f_{\nu,0}}=[\frac{1-(v/c)}{1-(v/c)cos\theta}]^3cos\theta.
1339: \end{equation}
1340:
1341: For a large Lorentz factor, we get
1342: \begin{equation}
1343: \frac{d f_{\nu}}{d
1344: f_{\nu,0}}=\frac{cos\theta}{8\Gamma^6[1-(1-1/2\Gamma^2)cos\theta]^3}.
1345: \end{equation}
1346: Here are several particular results: a) when $\theta \ll 1/\Gamma$,
1347: $d f_{\nu}/d f_{\nu,0}=1$; b) when $1/\Gamma \ll \theta \ll 1$, $d
1348: f_{\nu}/d f_{\nu,0}=1/(\theta\Gamma)^{6}$, which was previously
1349: pointed out by Kumar \& Panaitescu (2000); c) when $\theta=\pi/3$,
1350: $d f_{\nu}/d f_{\nu,0}=1/2\Gamma^{6}$; d) when $\theta=\pi/2$, $d
1351: f_{\nu}/d f_{\nu,0}=0$.
1352:
1353:
1354: \begin{thebibliography}{99}
1355: \bibitem[]{1034} Band, D. et al. 1993, ApJ, 413, 281
1356:
1357: \bibitem[]{1036} Barthelmy, S. D. et al. 2005, ApJ, 635, L133
1358:
1359: \bibitem[]{1038} Butler, N. R. \& Kocevski, D. 2007, ApJ, 663, 407
1360:
1361: \bibitem[]{1040} Campana, S. et al. 2006, Nature, 442, 1008
1362:
1363: \bibitem[]{1042} Dado, S., Dar, A., Rujula, A. D. 2008, astro-ph/0709.4307
1364:
1365: \bibitem[]{1044} Dermer, C. D. 2004, ApJ, 614, 284
1366:
1367: \bibitem[]{1046} Dyks, J. et al. 2005, preprint, astro-ph/0511699
1368:
1369: \bibitem[]{1048} Fan, Y. Z. et al. 2006, JCAP, 9, 13
1370:
1371: \bibitem[]{1050} Fenimore, E. E. et al. 1996, ApJ, 473, 998
1372:
1373: \bibitem[]{1052} Ghisellini, G. et al. 2006, MNRAS, 372, 1699
1374:
1375: \bibitem[]{1054} Gehrels, N. et al. 2004, ApJ, 611, 1005
1376:
1377: \bibitem[]{1056} Gehrels, N. et al. 2006, Nature, 444, 1044
1378:
1379: \bibitem[]{1058} Jia, L.-W. 2008, ChJAA, 8, 451
1380:
1381: \bibitem[]{1060} Kocevski, D., Ryde, F., Liang, E. 2003, ApJ, 596, 389
1382:
1383: \bibitem[]{1062} Kumar, P., \& Panaitescu, A. 2000, ApJ, 541, L51
1384:
1385: \bibitem[]{1064} Liang, E. W. et al. 2006, ApJ, 646, 351
1386:
1387: \bibitem[]{1066} Lu, R.-J., Qin, Y.-P., Zhang, Z.-B., Yi, T.-F. 2006, MNRAS, 367,
1388: 275
1389:
1390: \bibitem[]{1069} Lu, R.-J., Peng, Z.-Y., Dong, W. 2007, ApJ, 663, 1110
1391:
1392: \bibitem[]{1071} Mangano, V. et al. 2007, A\&A, 470, 105
1393:
1394: \bibitem[]{1073} Panaitescu, A. et al. 2006, MNRAS, 366, 1357
1395:
1396: \bibitem[]{1075} Peng, Z.-Y., Qin, Y.-P., Zhang, B.-B., Lu, R.-J., Jia, L.-W.,
1397: Zhang, Z.-B. 2006, MNRAS, 368, 1351
1398:
1399: \bibitem[]{1078} Qin, Y.-P. 2002, A\&A, 396, 705
1400:
1401: \bibitem[]{1080} Qin, Y.-P. et al. 2004, ApJ, 617, 439
1402:
1403: \bibitem[]{1082} Qin, Y.-P. \& Lu, R.-J. 2005, MNRAS, 362, 1085
1404:
1405: \bibitem[]{1084} Qin, Y.-P. et al. 2005, ApJ, 632, 1008
1406:
1407: \bibitem[]{1086} Qin, Y.-P. et al. 2006, Phys. Rev. D 74, 063005
1408:
1409: \bibitem[]{1088} Qin, Y.-P. 2008, ApJ, 683, 900
1410:
1411: \bibitem[]{1090} Ryde, F., \& Petrosian, V. 2002, ApJ, 578, 290
1412:
1413: \bibitem[]{1092} Sakamoto, T. et al. et al. 2007, ApJ, 669, 1115
1414:
1415: \bibitem[]{1094} Sari, R., \& Piran, T. 1997, ApJ, 485, 270
1416:
1417: \bibitem[]{1096} Sari, R., Piran, T., \& Narayan, R. 1998, ApJ, 497, L17
1418:
1419: \bibitem[]{1098} Shen, R.-F. et al. 2005, MNRAS, 362, 59
1420:
1421: \bibitem[]{1100} Starling, R. L. C. et al. 2008, MNRAS, 384, 504
1422:
1423: \bibitem[]{1102} Tagliaferri, G. et al. 2005, Nature, 436, 985
1424:
1425: \bibitem[]{1104} Yonetoku, D. et al. 2008, PASJ, 60, S351
1426:
1427: \bibitem[]{1106} Zhang, B. 2007, ChJAA, 7, 1
1428:
1429: \bibitem[]{1108} Zhang, B. et al. 2007a, ApJ, 655, L25
1430:
1431: \bibitem[]{1110} Zhang, B.-B. et al. 2007b, ApJ, 666, 1002
1432:
1433: \bibitem[]{1112} Zhang, B., Fan, Y. Z., Dyks, J. et al. 2006, ApJ, 642, 354
1434:
1435: %\item[]Yamazaki, R., Toma, K., Ioka, K., \& Nakamura, T. 2006, MNRAS, 369,
1436: %311
1437: \end{thebibliography}
1438:
1439: \end{document}
1440: