astro-ph0406027/ms.tex
1: 
2: \documentstyle[psfig]{mn2e}
3: 
4: \def\ie{i.e$.$~} \def\eg{e.g$.$~} \def\etal{et al$.$~} \def\cm3{\rm cm^{-3}}
5: \def\Msunyear{M_\odot\,{\rm yr^{-1}}} \def\kms{\rm km\,s^{-1}}
6: \def\simg{\mathrel{%
7:       \rlap{\raise 0.511ex \hbox{$>$}}{\lower 0.511ex \hbox{$\sim$}}}}
8: \def\siml{\mathrel{%
9:       \rlap{\raise 0.511ex \hbox{$<$}}{\lower 0.511ex \hbox{$\sim$}}}}
10: \def\epsi{\varepsilon_i} \def\epsB{\varepsilon_B} \def\epse{\epsilon_e}
11: \def\Mesz{M\'esz\'aros~}
12: 
13: 
14: \begin{document}
15: 
16: \title[GRB Afterglows 990123 and 021211]
17:        {Analysis of Two Scenarios for the Early Optical Emission of the GRB Afterglows 990123 and 021211}
18: 
19: \author[A. Panaitescu \& P. Kumar]{A. Panaitescu and P. Kumar \\
20:        Department of Astronomy, University of Texas, Austin, TX 78712}
21: 
22: \maketitle
23: 
24: \begin{abstract}
25: 
26:  The optical light-curves of GRB afterglows 990123 and 021211 exhibit a steep decay at 100--600
27: seconds after the burst, the decay becoming slower after about 10 minutes. We investigate two 
28: scenarios for the fast decaying early optical emission of these GRB afterglows. In the 
29: {\sl reverse-forward shock} scenario, this emission arises in the reverse shock crossing the GRB 
30: ejecta, the mitigation of the light-curve decay occurring when the forward shock emission overtakes 
31: that from the reverse shock. Both a homogeneous and wind-like circumburst medium are considered. 
32: In the {\sl wind-bubble} scenario, the steeply decaying, early optical emission arises from the 
33: forward shock interacting with a $r^{-2}$ bubble, with a negligible contribution from the reverse 
34: shock, the slower decay starting when the blast wave reaches the bubble termination shock and 
35: enters a homogeneous region of the circumburst medium. 
36: 
37:  We determine the shock microphysical parameters, ejecta kinetic energy, and circumburst density 
38: which accommodate the radio and optical measurements of the GRB afterglows 990123 and 021211. 
39: We find that, for a homogeneous medium, the radio and optical emissions of the afterglow 990123 can 
40: be accommodated by the reverse-forward shock scenario if the microphysical parameters behind the 
41: two shocks differ substantially. A wind-like circumburst medium also allows the reverse-forward 
42: shocks scenario to account for the radio and optical properties of the afterglows 990123 and 021211, 
43: but the required wind densities are at least 10 times smaller than those of Galactic Wolf-Rayet 
44: stars. The wind-bubble scenario requires a variation of the microphysical parameters when the 
45: afterglow fireball reaches the wind termination shock, which seems a contrived feature.
46: 
47: \end{abstract}
48: 
49: 
50: \section{Introduction}
51: 
52: 
53:  There are currently two GRB afterglows for which a fast falling-off optical emission was detected
54: at early times, only $\sim 100$ seconds after the burst. The general consensus is that this emission 
55: arises from the GRB ejecta which is energized by the reverse shock ({\bf RS}) crossing the ejecta and 
56: caused by the interaction of the ejecta with the circumburst medium ({\bf CBM}). This interaction also 
57: drives a forward shock ({\bf FS}) energizing the swept-up CBM, to which the later afterglow emission 
58: is attributed (the "reverse-forward shock" scenario).
59:   
60:  The RS emission was first calculated by \Mesz \& Rees (1997), who considered the cases of a 
61: frozen-in and turbulent magnetic field in the ejecta, and showed that, in either case, a bright 
62: optical emission ($m_V \sim 9$) is obtained at the end of the burst. 
63: \Mesz \& Rees (1999) extended their previous calculations of the RS emission to a radiative evolution
64: of the fireball Lorentz factor and pointed out the importance of spectral information in constraining 
65: the RS dynamics and the magnetic field origin from the observed $t^{-2}$ power-law decay of the very 
66: early optical light-curve of the afterglow 990123 (Akerlof \etal 1999). They also pointed out the 
67: possibility that optical flashes arise in the same internal shocks which generate the burst emission.
68: 
69:   Sari \& Piran (1999) have shown that, if the peak frequency of the RS emission is assumed to be
70: in the optical at the time when the optical emission of the afterglow 990123 peaks (50 seconds 
71: after the burst), then the expected softening of the RS emission and self-absorption effects can 
72: account for the radio flare reported by Kulkarni \etal (1999a). Kobayashi \& Sari (2000) confirm 
73: the RS interpretation of this radio flare through numerical calculations of the RS dynamics.
74: 
75:  Chevalier \& Li (2000) have presented calculations of the RS synchrotron emission until it crosses 
76: the GRB ejecta, for the case of a wind-like CBM. For their choice of a high magnetic field parameter, 
77: the RS cooling frequency falls well below the optical domain, which leads to a RS optical emission 
78: much dimmer than that observed for the afterglow 990123 at its peak (40 seconds after the burst).
79: Furthermore, such a low cooling frequency implies that the early afterglow optical emission should
80: cease when the RS has crossed the ejecta shell, \ie at the peak time of the RS emission. Since this
81: is in contradiction with the observations of the afterglow 990123, Chevalier \& Li (2000) have 
82: concluded that a wind-like CBM cannot explain the early optical emission of the afterglow 990123.
83: 
84:  Constraints on the fireball initial Lorentz factor have been obtained by Soderberg \& Ramirez-Ruiz
85: (2003) for several afterglows by comparing the observed radio emission at $\sim 1$ day with the 
86: model RS emission, under the assumption that the RS magnetic field and typical electron energy 
87: parameters (which we shall call {\bf microphysical parameters}) are those determined for the FS from 
88: fits to the broadband emission of those afterglows (Panaitescu \& Kumar 2001). Zhang, Kobayashi \& 
89: \Mesz (2003) have noted that the ratios of the RS and FS peak fluxes, and peak and cooling frequencies 
90: depend only on the fireball initial Lorentz factor and the ratio of the magnetic fields, to develop 
91: a method of constraining these two quantities, and have shown that the optical emission of the afterglow 
92: 990123 requires a magnetized outflow. 
93: 
94:  In this work we use the general properties (flux, epochs during which power-law decays are observed,
95: decay slopes -- see Table 1) of the radio and optical emissions of the afterglows 990123 and 021211 
96: to constrain the ejecta (isotropic-equivalent) kinetic energy, CBM density, and the microphysical 
97: parameters for the reverse-forward shock scenario (\S\ref{RF}), for either a homogeneous or wind-like 
98: profile of the CBM. In contrast with other works, we take into account all constraints arising from 
99: the radio and optical measurements of the GRB afterglows 990123 and 021211 and we do not assume certain 
100: values for any of the model parameters.
101:  We also investigate a "wind-bubble" scenario (\S\ref{BB}), where all the radio and optical emission 
102: arises in the FS, with a negligible contribution from the RS (which is verified numerically), the 
103: mitigation of the optical decays observed in the afterglows 990123 and 021211 being due to the FS 
104: crossing the bubble termination shock, i.e. transiting from a wind-like CBM structure to a homogeneous 
105: region of shocked wind. 
106:  For both scenarios, we consider either adiabatic or radiative dynamics, the resulting microphysical 
107: parameters being checked for consistency with the assumed dynamical regime.
108: 
109: 
110: \section{Physical Parameters at the Ejecta Shock-Crossing Radius}
111: \label{dynamics}
112: 
113:  We begin by calculating the spectral properties (break frequencies and peak flux) of the RS emission 
114: at the radius $R_+$ where the RS finishes crossing the ejecta shell and the injection of fresh electrons 
115: by the RS ceases. Because most of the ejecta kinetic energy has been transferred to the forward shock
116: at $R_+$, the radius $R_+$ marks the onset of a steeper power-law decrease of the bulk Lorentz factor 
117: $\Gamma$ with radius. After $R_+$, the spectral properties of the RS emission can be calculated from the 
118: adiabatic evolution of the electrons and magnetic field. The spectral properties of the FS emission can 
119: also be calculated from those at $R_+$ or directly from the dynamics of the fireball after $R_+$ 
120: (\ie without passing through the parameters at $R_+$) if the shock dynamics is adiabatic. 
121: 
122:  Each shock compresses the fluid ahead of it by a factor $4 \Gamma' + 3$, where $\Gamma'$ is the Lorentz 
123: factor of the shocked fluid as measured in the frame of the yet unshocked gas, and heats it to a energy 
124: per particle equal to $\Gamma'-1$. Therefore, the pressure equality at the contact discontinuity which 
125: separates the shocked ejecta and CBM, implies that
126: \begin{equation}
127:  (4 \Gamma' + 3) (\Gamma' - 1) n_{ej} = (4 \Gamma + 3)  (\Gamma - 1) n 
128: \label{G1}
129: \end{equation}
130: where $\Gamma$ is the Lorentz factor of the shocked fluid in the laboratory frame, and $n_{ej}$, $n$ 
131: are the proton number densities of the unshocked ejecta and of the CBM, respectively, each measured 
132: in the corresponding comoving frame. From addition of velocities in special relativity,
133: \begin{equation}
134:  \Gamma' = \Gamma_0 \Gamma (1 - \beta_0 \beta) \simeq 
135:           \frac{1}{2} \left( \frac{\Gamma}{\Gamma_0} + \frac{\Gamma_0}{\Gamma} \right) 
136: \label{G2}
137: \end{equation}
138: where $\Gamma_0$ is the initial Lorentz factor of the ejecta, $\beta$ denotes velocities and 
139: $\Gamma_0 \gg 1$, $\Gamma \gg 1$ were used in the approximation. 
140: Substituting equation (\ref{G2}) in (\ref{G1}), one obtains a fourth-degree equation for $\Gamma$,
141: which can be cast in the form
142: \begin{equation}
143:  \frac{(\Gamma_0 - \Gamma)^2}{\Gamma \Gamma_0} \left[  \frac{(\Gamma_0 + \Gamma)^2}{\Gamma \Gamma_0} 
144:                - \frac{1}{2} \right] n_{ej} = 4 \Gamma^2 n \;.
145: \label{G3}
146: \end{equation} 
147: Because $\Gamma \leq \Gamma_0$, the first term in the square bracket is at least eight times larger
148: than the last term. Ignoring the $1/2$ term, the solution of equation (\ref{G3}) is 
149: \begin{equation}
150:  \Gamma = \frac{\Gamma_0}{\left[ 1 + 2 \Gamma_0 \left( \frac{\displaystyle n}
151:                    {\displaystyle n_{ej}} \right)^{1/2} \right]^{1/2}} \;.
152: \label{G4}
153: \end{equation}
154: The limiting cases for the Lorentz factor are
155: \begin{equation}
156:  \Gamma = \left\{ \begin{array}{ll} 
157:                   \Gamma_0 & \frac{\displaystyle n_{ej}}{\displaystyle n} \gg 4 \Gamma_0^2 \\
158:                   \left( \frac{\displaystyle \Gamma_0}{\displaystyle 2} \right)^{1/2} 
159:                       \left( \frac{\displaystyle n_{ej}}{\displaystyle n} \right)^{1/4} & 
160:                       \frac{\displaystyle n_{ej}}{\displaystyle n} \ll 4 \Gamma_0^2 
161:                \end{array} \right.  \;.
162: \label{G5}
163: \end{equation}
164: Therefore, in the limit of very dense ejecta, the Lorentz factor of the shocked fluid is the same as
165: that of the unshocked ejecta, while for more tenuous ejecta, $\Gamma$ depends on the ratio of the
166: comoving densities. 
167:  Note that the ratio $n_{ej}$ and $n_{cmb}$ changes with the fireball radius $R$:
168: \begin{equation}
169:   n_{ej} = \frac{E}{4 \pi m_p c^2 \Gamma_0 (\Gamma_0 \Delta) R^2} \;,\; n = A R^{-s} 
170: \end{equation}
171: where $E$ is the fireball ejecta energy (or, if the outflow is collimated, its isotropic equivalent),
172: $m_p$ is the proton's mass, $\Delta \ll R$ is the ejecta geometrical thickness measured in the laboratory 
173: frame (thus $\Gamma_0 \Delta$ is the comoving frame thickness), and we restricted our calculations to two 
174: simple radial structures of the CBM, either homogeneous ($s=0$) of particle density $n$ or a $R^{-2}$ 
175: stratification ($s=2$) corresponding to a wind expelled by the GRB progenitor. In the latter case, 
176: $A = 3 \times 10^{35} A_* \; {\rm cm^{-1}}$, where $A_*$ is the mass-loss rate to wind speed ratio, 
177: normalized to $10^{-5} \Msunyear/ 10^3 \kms$. By denoting 
178: \begin{equation}
179:   X = \frac{E}{4 \pi A m_p c^2}
180: \end{equation}
181: the ratio of the comoving densities is
182: \begin{equation}
183:  \frac{n_{ej}}{n} = \frac{X}{\Delta \Gamma_0^2 R^{2-s}} \;.
184: \label{n}
185: \end{equation}
186: The ejecta-shell thickness, $\Delta$, is the largest of its initial thickness $c \tau$, where $\tau$
187: is the laboratory frame duration of the GRB ejecta release by their source, and the expansion due to 
188: a spread in the radial outflow velocity of the ejecta particles. This velocity spread can be either a 
189: relic of the initial, super-Eddington radiation pressure in the fireball, or to an imperfect collimation
190: in the radial direction of the ejecta particles at the end of the fireball acceleration. The former
191: leads to a comoving-frame expansion of the shell at the sound speed, the latter is expected to produce
192: a spread of order $1/\Gamma_0$ in the ejecta particles direction of motion (\Mesz, Laguna \& Rees 1993). 
193: In either case, the resulting contribution to the shell thickness evolution is $R/\Gamma_0^2$, therefore
194: \begin{equation}
195:  \Delta = \left\{ \begin{array}{ll} 
196:              c \tau & R < 2 \Gamma_0^2 c \tau  \\ 
197:             \frac{\displaystyle R}{\displaystyle 2 \Gamma_0^2} & R >  2 \Gamma_0^2 c \tau 
198:           \end{array} \right.  \;.
199: \label{D}
200: \end{equation}
201: As shown by Kumar \& Panaitescu (2003), the difference between the laboratory frame speeds of the unshocked 
202: ejecta and that of the RS is
203: \begin{equation}
204:  \beta_0 - \beta_{RS} = \frac{1.4}{\Gamma_0^2} \left( \frac{\Gamma_0^2 n}{n_{ej}} \right)^{1/2} 
205: \label{bdiff}
206: \end{equation}
207: for a wide range of the ratio $\Gamma_0^2 n/n_{ej}$. From here, one can calculate the radius $R_+$ at 
208: which the RS finishes crossing the ejecta shell:
209: \begin{equation}
210:   \Delta (R_+) = \int_0^{R_+} (\beta_0 - \beta_{RS})\; dR \;. 
211: \label{Rcross}
212: \end{equation}  
213: Once $R_+$ is known, equations (\ref{G4}), (\ref{n}) and (\ref{D}) give the Lorentz factor $\Gamma$
214: at the shock-crossing radius $R_+$ and all the properties of the RS and FS emissions at $R_+$,
215: which can be then extrapolated at $R > R_+$. We proceed by considering separately ejecta shells 
216: for which $2 \Gamma_0^2 c \tau < R_+$ and $R_+ < 2 \Gamma_0^2 c \tau$. In the former case, the 
217: ejecta shell undergoes a significant spreading while the RS propagates into it, while in the latter 
218: case the spreading is negligible. The usual terminology (Sari \& Piran 1995) is that of "thin ejecta" 
219: for the former and "thick ejecta" for the latter.
220: 
221: 
222: \subsection{Thin Ejecta Shell -- $\Delta = R/(2\Gamma_0^2)$}
223: \label{thin}
224: 
225:  The substitution of equations (\ref{n}) and (\ref{bdiff}) in (\ref{Rcross}) leads to
226: \begin{equation}
227:  (s=0):\; R_+ = \left( \frac{\displaystyle 1.6 X}{\displaystyle \Gamma_0^2} \right)^{1/3} \! \! \! \! = 
228:              0.94 \times 10^{17} \left( \frac{\displaystyle E_{53}}
229:              {\displaystyle n_0 \Gamma_{0,2}^2} \right)^{1/3} {\rm cm} 
230: \label{R1}
231: \end{equation}
232: \begin{equation}
233:  (s=2):\; R_+ = \frac{\displaystyle 0.57 X}{\displaystyle \Gamma_0^2} \! =
234:                   1.0  \times 10^{15} \frac{\displaystyle E_{53}}
235:                          {\displaystyle A_* \Gamma_{0,2}^2}\; {\rm cm}
236: \label{R2}
237: \end{equation}
238: where the usual notation $Q_n = Q/10^n$ has been used.
239: The defining condition for this case, $R_+ > 2 \Gamma_0^2 c \tau$, becomes
240: \begin{equation}
241:  \Gamma_0 < \left[ \frac{0.20 X}{(c \tau)^3} \right]^{1/8} = 
242:               670\; \left( \frac{E_{53}}{n_0 \tau_0^3} \right)^{1/8} \;\; {\rm for} \;\; s=0 
243: \label{Gmin0}
244: \end{equation}
245: \begin{equation}
246:  \Gamma_0 < \left( \frac{0.29 X}{c \tau} \right)^{1/4} =
247:               120\; \left( \frac{E_{53}}{A_* \tau_0} \right)^{1/4}  \;\; {\rm for} \;\; s=2  \;.
248: \label{Gmin2}
249: \end{equation}
250: From equation (\ref{n}), the density ratio at $R_+$ is
251: \begin{equation}
252:   \frac{n_{ej}}{n} (R_+)= \frac{2X}{R_+^{3-s}} = \left\{ \begin{array}{ll} 
253:                         1.3\; \Gamma_0^2 & s=0 \\ 3.5\; \Gamma_0^2 & s=2
254:                               \end{array} \right. \;.
255: \label{nratio}
256: \end{equation}
257: Comparing with equation (\ref{G4}), this shows that at $R_+$ the Lorentz factor of shocked gas is 
258: in an intermediate regime:
259: \begin{equation}
260:  \Gamma_+ \stackrel{s=0}{=} 0.60\, \Gamma_0 \;\;,\;\; \Gamma_+ \stackrel{s=2}{=} 0.70\, \Gamma_0  \;.
261: \label{G6}
262: \end{equation}  
263: 
264:  Besides $\Gamma(R_+)$, two other quantities are of interest at the ejecta shock-crossing radius:
265: the energy of the swept-up CBM, $E_{cbm} (R_+)$, and the corresponding observer frame time, $t_+$. 
266: The total energy of the shocked CBM is $\Gamma^2$ larger than its rest-mass energy, thus 
267: \begin{equation}
268:  E_{cbm} (R) = m_p c^2 \int\limits_0^R 4 \pi r^2  A r^{-s} \Gamma^2(r) \; dr \;.
269: \label{Ecbm}
270: \end{equation}
271: The arrival time $t$ corresponding to the contact discontinuity and the fluid moving toward 
272: the observer is given by 
273: \begin{equation}
274:  t_{CD} (R) = (1+z) \int_0^R \frac{dr}{2c\Gamma^2} \;, 
275: \label{dt}
276: \end{equation}
277: where $z$ is the burst redshift.
278: Substituting $\Gamma$ from equation (\ref{G4}) in equations (\ref{Ecbm}) and (\ref{dt}), one
279: obtains 
280: \begin{equation}
281:  (s=0):  E_{cbm} (R_+) = 0.25\,E ,\; 
282:            t_+ = 350\; (1+z) \left( \frac{E_{53}}{n_0 \Gamma_{0,2}^8} \right)^{1/3}  \! \! \! \!{\rm s} 
283: \label{Et0}
284: \end{equation}
285: \begin{equation}
286:  (s=2):  E_{cbm} (R_+) = 0.34\,E ,\; 
287:            t_+ = 2.9\; (1+z) \frac{E_{53}}{A_* \Gamma_{0,2}^4}\, {\rm s} \;.
288: \label{Et1}
289: \end{equation}
290: Thus less than half of the initial ejecta energy has been dissipated by the FS by the time when the 
291: RS crosses the ejecta shell. This means that the shock crossing radius $R_+$ is slightly smaller than 
292: the usual deceleration radius $R_d$ defined by $E_{cbm}(R_d) = 0.5 E$. 
293: 
294: 
295: \subsection{Thick Ejecta Shell -- $\Delta = c \tau$}
296: \label{thick}
297: 
298:  Once again, using equations  (\ref{n}) and (\ref{bdiff}) in (\ref{Rcross}), one obtains that
299: \begin{equation}
300:  R_+ \stackrel{s=0}{=} (2.0 X c \tau)^{1/4} = 
301:                        0.24 \times 10^{17} \left( \frac{\displaystyle E_{53} \tau_0} 
302:                                         {\displaystyle n_0} \right)^{1/4}\; {\rm cm} 
303: \label{R3}
304: \end{equation}
305: \begin{equation}
306:  R_+ \stackrel{s=2}{=} (0.51 X c \tau)^{1/2} = 
307:                        0.52 \times 10^{15} \left( \frac{\displaystyle E_{53} \tau_0} 
308:                                         {\displaystyle A_*} \right)^{1/2}\; {\rm cm}  \;.
309: \label{R4}
310: \end{equation}
311: The requirement that $R_+ < 2 \Gamma_0^2 c \tau$ leads to the reversed inequalities given in equations
312: (\ref{Gmin0}) and (\ref{Gmin2}). 
313: The density ratio (equation \ref{n}) at $R_+$ is
314: \begin{equation}
315:   \frac{n_{ej}}{n} (R_+) = \frac{X}{\Gamma_0^2 c \tau R_+^{2-s}} \;.
316: \label{nr}
317: \end{equation}
318: Because in the thick shell case the ejecta density is lower than for a thin shell, the values given
319: in the $rhs$ of equation (\ref{nratio}) are upper limits for the density ratio. Substituting
320: equation (\ref{nr}) in the second regime given in equation (\ref{G5}) leads to
321: \begin{equation}
322:  (s=0) : \Gamma_+ =  \left[ \frac{X}{32(c \tau)^3} \right]^{1/8} =
323:                              530\; \left( \frac{E_{53}}{n_0 \tau_0^3} \right)^{1/8} 
324: \label{G7}
325: \end{equation}
326: \begin{equation}
327:  (s=2) : \Gamma_+ =  \left( \frac{X}{4c \tau} \right)^{1/4} =
328:                              110\; \left( \frac{E_{53}}{A_* \tau_0} \right)^{1/4} \;.
329: \label{G8}
330: \end{equation}
331: 
332:  Note from equations (\ref{G7}) and (\ref{G8}) that, for a thick ejecta shell, the Lorentz factor
333: $\Gamma$ of the shocked ejecta is independent of that of the unshocked ejecta, $\Gamma_0$. Thus, for
334: a sufficiently tenuous ejecta, the contrast between $\Gamma_0$ and $\Gamma$ can be sufficiently large
335: that the RS is relativistic (equation \ref{G2}) in the frame of the incoming ejecta. In contrast, in
336: the thin ejecta case, equations (\ref{G2}) and (\ref{G6}) lead to $\Gamma' - 1 = 0.13$ for $s=0$ and
337: $\Gamma' - 1 = 0.064$ for $s=2$, \ie the RS propagating in a thin ejecta shell is trans-relativistic.
338: 
339:  The energy dissipated by the FS at $R=R_+$ and the ejecta shell shock-crossing time are
340: \begin{equation}
341:   E_{cbm} (R_+) = 0.36\, E \;, \quad t_+ = 0.71\, (1+z)\, \tau 
342: \label{Et2}
343: \end{equation}
344: for either type of medium.
345: Note that, as for a thin ejecta shell, the shock-crossing radius $R_+$ is close to the usual 
346: deceleration radius. 
347: Furthermore, for a thick ejecta shell, the observer frame shock-crossing time $t_+$ is fairly close 
348: to the laboratory frame duration of the ejecta release. Given that $t_+$ is roughly the timescale
349: for dissipating the ejecta kinetic energy and that, most likely, the shock-accelerated electrons cool 
350: faster than the dynamical timescale, we expect that the duration of an external shock GRB is close 
351: to $t_+$. The simple temporal structure of the GRBs 990123 and 021211 may suggest that they originate 
352: in an external shock, nevertheless it is entirely possible that both bursts were produced in internal 
353: shocks occurring in an outflow with a fluctuating ejection Lorentz factor (Rees \& \Mesz 1994). In this 
354: case, the internal 
355: shocks take place at the radius $R_{is} \sim \Gamma_{min}^2 c \tau$, where $\Gamma_{min}$ is the Lorentz 
356: factor of the slower shells; the shocked fluid moves at $\Gamma \sim (\Gamma_{min} \Gamma_{max})^{1/2}$, 
357: where $\Gamma_{max}$ is the Lorentz factor of the faster shells; therefore the observed burst duration 
358: is $t_\gamma = R_{is}/(c \Gamma^2) \sim (\Gamma_{min}/\Gamma_{max}) \tau$. Since a high dissipation 
359: efficiency requires a large contrast between the Lorentz factors of various ejecta shells ($\Gamma_{min} 
360: \ll \Gamma_{max}$), the burst duration sets only a lower limit on the duration of the ejecta release,
361: $\tau$. We shall use this constraint when choosing the shock-crossing time $t_+$ for the GRBs 990123 
362: and 021211 for the case of a thick ejecta shell.
363: 
364: 
365: \section{Synchrotron Emission}
366: \label{radiation}
367: 
368:  The synchrotron emission from either shock at any observing frequency is determined by the peak 
369: flux $F_p$, the three break frequencies -- absorption $\nu_a$, injection $\nu_i$, and cooling $\nu_c$ 
370: -- and the slope of the afterglow spectrum between its break frequencies.
371: 
372:  The peak of the $F_\nu$ synchrotron spectrum, which is at the frequency $\nu_p = \min\{\nu_i,\nu_c\}$,
373: is given by
374: \begin{equation}
375:  F_p = \frac{1+z}{4\pi d_L^2(z)} N_e \Gamma P'_p  
376: \label{Fp}
377: \end{equation}
378: where $z$ is the burst redshift, $d_L$ the luminosity distance (in a $H_0 = 70 \kms/Mpc$, $\Omega_M 
379: = 0.3$, $\Omega_\Lambda = 0.7$ universe), $N_e$ is the number of radiating electrons, and the factor
380: $\Gamma$ accounts for the average relativistic boost of the comoving frame synchrotron power $P'_p$
381: per electron at the peak frequency $\nu_p' = \nu_p/\Gamma$. The Doppler boost $\Gamma$ appears at only 
382: the first power in equation (\ref{Fp}) because the observer receives emission for an area subtending 
383: an angle of $\Gamma^{-1}$ radians, \ie from a fraction $\Gamma^{-2}$ of the total number of electrons
384: $N_e$. Since we will be interested in the early time afterglow emission, the possible collimation of 
385: the outflow is not an issue here. After the shock crossing radius $R_+$, the number of electrons energized 
386: by each shock is
387: \begin{equation}
388:  N_e^{(RS)} = \frac{E}{m_p c^2 \Gamma_0} \;, \quad  N_e^{(FS)} = \frac{4 \pi}{3-s} A R^{3-s} \;.
389: \label{Ne}
390: \end{equation}
391: The comoving power per electron of equation (\ref{Fp}) is
392: \begin{equation}
393:  P'_p = \sqrt{3} \phi_p \frac{e^3}{m_e c^2} B
394: \end{equation}
395: where $\phi_p$ is the order-unity coefficient calculated by Wijers \& Galama (1999), $e$ is the
396: electron charge, and $B$ is the magnetic field strength. The magnetic field is parameterized by 
397: the fraction $\epsB$ of the post-shock energy density stored in it. Taking into account that
398: the FS compresses the CBM by a factor $4 \Gamma $ and heats it is to an energy per proton of
399: $\Gamma m_p c^2$, the magnetic field is
400: \begin{equation}
401:  B = \left( 32 \pi\, \epsB\, A R^{-s}\, m_p c^2\, \Gamma^2 \right)^{1/2} \;.
402: \label{B1}
403: \end{equation}
404: 
405:  The break frequencies are calculated from the corresponding electron Lorentz factors $\gamma_{a,i,c}$
406: \begin{equation}
407:  \nu_{a,i,c} = \frac{3 x_p}{4\pi} \frac{e}{m_e c} \gamma_{a,i,c}^2 B \Gamma
408: \label{nuaic}
409: \end{equation}
410: where $x_p$ is another order-unity factor calculated by Wijers \& Galama (1999) and the last factor
411: in the $rhs$ is for the average relativistic boost of the fireball emission by its relativistic
412: expansion. The typical electron Lorentz factor after shock-acceleration is parameterized as
413: \begin{equation}
414:  \gamma_i = \epsi \frac{m_p}{m_e} (\Gamma' - 1)
415: \label{gi}
416: \end{equation}
417: where $\Gamma' = \Gamma$ for the FS, while $\Gamma' (R_+)$ for the RS is that given in equation (\ref{G2}) 
418: for $\Gamma = \Gamma_+$.
419: We assume that each shock injects in the downstream fluid electrons with a power-law energy
420: distribution
421: \begin{equation}
422:   \frac{dN}{d\gamma} (\gamma > \gamma_i) \propto \gamma^{-p} \;. 
423: \label{dNdg}
424: \end{equation}
425: The acceleration of new electrons by the RS ceases at $R_+$, when all the ejecta have been swept-up, 
426: but continues at the FS. The cooling electron Lorentz factor is that for which the radiative losses 
427: timescale is equal to the dynamical timescale:
428: \begin{equation}
429:  \gamma_c = 6 \pi \frac{m_e c^2}{\sigma_T} \frac{\Gamma}{(Y + 1) R B^2}
430: \label{gc}
431: \end{equation}
432: where $\sigma_T$ is the Thomson cross-section for electron scattering and $Y$ is the Compton parameter, 
433: \ie the ratio of inverse Compton to synchrotron power. The Compton parameter is calculated from the 
434: electron distribution, as described by Panaitescu \& Kumar (2000). The random Lorentz factor of the 
435: electrons radiating at the self-absorption frequency $\nu_a$ is given by 
436: \begin{equation}
437:  \gamma_a = \left\{ \begin{array}{ll} 
438:             \gamma_p \tau_p^{3/10} & \tau_p < 1 \\ \gamma_p \tau_p^{1/(p+4)} & \tau_p > 1
439:              \end{array} \right.
440:            \;,\;\;\; \tau_p = \frac{5 e}{B \gamma_p^5} \frac{N_e}{4\pi R^2}
441: \label{ga}
442: \end{equation}
443: where $\gamma_p = \min\{\gamma_i,\gamma_c\}$ and $N_e$ is given in equation (\ref{Ne}). 
444: 
445:  Equations (\ref{Fp}) -- (\ref{ga}) provide the characteristics of the RS and FS synchrotron emissions 
446: at $R > R _+$, with the values of $B$, $\gamma_i$ and $\gamma_c$ for the RS calculated at $R = R_+$
447: from equations (\ref{B1}), (\ref{gi}) and (\ref{gc}), respectively (using equations \ref{R1},
448: \ref{R2}, \ref{G6}, \ref{R3}, \ref{R4}, \ref{G7}, \ref{G8}) and then extrapolated at $R > R_+$
449: as described in \S\ref{rs}. Once $\Gamma (R > R_+)$ is known, the fireball radius can be related with 
450: the observer-frame arrival-time of the photons emitted at that radius and the evolution of the
451: RS and FS spectral characteristics can be calculated. 
452: 
453:  The slopes of the piece-wise synchrotron spectrum from the broken power-law electron distribution
454: with energy resulting from shock-acceleration and radiative cooling are described in detail by Sari,
455: Piran \& Narayan (1998). Note that, for $R > R_+$, the RS electron distribution has a sharp cut-off 
456: at $\gamma_c$, as the injection of electrons stops when the RS has crossed the ejecta shell, which 
457: leads to an abrupt switch-off of the RS emission at a given frequency when $\nu_c$ drops below that
458: frequency.
459: 
460: 
461: \subsection{Forward Shock}
462: \label{fs}
463: 
464:  If the dynamics of the FS is adiabatic, the Lorentz factor $\Gamma_F$ of the FS follows immediately 
465: from energy conservation $\Gamma_F^2 M_{cbm} c^2 = E$, where $M_{cbm}$ is the mass of the swept-up 
466: CBM. From here one obtains the Blandford-McKee solution (Blandford \& McKee 1976):
467: \begin{equation}
468:  \Gamma_F (R > R_+) = \Gamma_+ \left( \frac{R}{R_+} \right)^{-(3-s)/2} \;.
469: \label{GFadb}
470: \end{equation}
471: Equation (\ref{dt}) also gives the photon arrival time for the emission arising from a patch on the FS
472: moving at an angle $\Gamma_F^{-1}$ relative to the fireball center -- observer axis, from where most
473: of the emission arises. Its integration with $\Gamma$ from equation (\ref{GFadb}) leads to 
474: \begin{equation}
475:  t_F (R)=\frac{1+z}{4-s} \frac{R_+}{c\Gamma_+^2} \left[\left(\frac{R}{R_+}\right)^{4-s}+1-\frac{s}{2}\right] 
476: \label{time}
477: \end{equation} 
478: where the weak deceleration at $R < R_+$ has been ignored in the calculation of the last term in the $rhs$. 
479: For $t \gg t_+$, equations (\ref{GFadb}) and (\ref{time}) lead to 
480: \begin{equation}
481:  \Gamma_F (t) \propto t^{-(3-s)/(8-2s)} \;,\; R (t) \propto t^{1/(4-s)} \;.
482: \end{equation}
483: Substituting in equations (\ref{Ne}), (\ref{B1}), (\ref{gi}), (\ref{gc}) and (\ref{ga}), one obtains
484: the following scalings
485: \begin{equation}
486:  F_p \propto t^{-s/(8-2s)} \;,\; \nu_i \propto t^{-3/2} 
487: \end{equation}
488: \begin{equation}
489:  \nu_c \propto t^{-(4-3s)/(8-2s)} \;,\; \nu_a \propto t^{-3s/(20-5s)} 
490: \end{equation}
491: where $\nu_a < \nu_i < \nu_c$ was assumed for the last equation. These scalings are used to calculate
492: the characteristics of the FS synchrotron emission at any $t > t_+$ from those at $t = t_+$.
493: 
494:  If the swept-up CBM radiates half of its internal energy faster than the dynamical timescale, the 
495: dynamics of the afterglow is described by $\Gamma_F M = const$, which leads to $\Gamma_F \propto 
496: R^{-(3-s)}$. Therefore the two extreme regimes of the fireball dynamics are
497: \begin{equation}
498:   \Gamma_F^{(adb)} \propto t^{-(3-s)/(8-2s)} \;,\; \Gamma_F^{(rad)} \propto t^{-(3-s)/(7-2s)}
499: \label{GFrad}
500: \end{equation}
501: where the former is the adiabatic case and the latter is a highly radiative regime.
502: To estimate the effect of high radiative losses, we calculate the dependence of the observed flux on 
503: the fireball Lorentz factor, $F_\nu \propto \Gamma_F^x$ (with $x$ frequency-dependent), and adjust 
504: the fluxes obtained in the adiabatic case by a factor 
505: \begin{equation}
506:  \frac{F_\nu^{(rad)}}{F_\nu^{(adb)}} (t) = \left( \frac{\Gamma_F^{(rad)}}{\Gamma_F^{(adb)}} \right)^x = 
507:               \left( \frac{t}{t_+} \right)^{-(3-s)x/[(8-2s)(7-2s)]} \;.
508: \label{rad}
509: \end{equation}
510: The radiative correction factors at $t \sim 0.1$ day are close to those resulting from the expressions 
511: of the afterglow flux in the adiabatic case (\eg equations B1-B9 and C1-C9 in Panaitescu \& Kumar 2000)
512: if the fireball energy is decreased by a factor around 10. This means that a highly radiative regime 
513: corresponds to a fractional energy loss of about 90\% within the first day after the burst. Note that 
514: for such high radiative losses to occur in an afterglow, FS electrons should acquire a substantial 
515: fraction of the energy dissipated by the shock and should radiate it quicker that the dynamical timescale, 
516: which requires a sufficiently high magnetic field, \ie a sufficiently large parameter $\epsB$ and dense 
517: CBM.
518: 
519: 
520: \subsection{Ejecta Shell}
521: \label{rs}
522: 
523:  We calculate the dynamics of the shocked ejecta by assuming adiabatic dynamics and that the ejecta is
524: in equilibrium pressure with the energized CBM, whose radial profile (Lorentz factor, density, pressure) 
525: is described by the Blandford-McKee solution. For adiabatic dynamics, the pressure in the ejecta shell
526: evolves as 
527: \begin{equation}
528:  p_R \propto (R^2 \Delta')^{-a}
529: \label{pR}
530: \end{equation}
531: where $\Delta'$ is the comoving thickness of the ejecta and $a$ is the adiabatic index. From the 
532: Blandford-McKee solution, the pressure equilibrium at the contact discontinuity implies that
533: \begin{equation}
534:  p_R = p_F (\chi_R) = p_F(\chi=1) \chi_R^{-(17-4s)/(12-3s)} 
535: \end{equation}
536: where $p_F (\chi_R)$ is the pressure of the shocked CBM at the coordinate $\chi_R$ of the RS, and 
537: \begin{equation}
538:  p_F(\chi=1) \propto \Gamma_F^2 n(R) \propto R^{-3}
539: \label{pF}
540: \end{equation}
541: is the pressure immediately behind the FS, the last relation resulting from the dynamics of the FS
542: (equation \ref{GFadb}). Equations (\ref{pR})--(\ref{pF}) imply that the Blandford-McKee coordinate 
543: for the RS satisfies
544: \begin{equation}
545:  \chi_R \propto (R^{2a-3} \Delta'^a)^{3(4-s)/(17-4s)} \;.
546: \label{chiR}
547: \end{equation}  
548: Equations (\ref{GFadb}) and (\ref{chiR}) and the Blandford-McKee solution for the post-FS Lorentz factor
549: \begin{equation}
550:  \Gamma (\chi) \propto \Gamma_F \chi^{-1/2}
551: \end{equation}  
552: lead to that the Lorentz factor at the location of the contact discontinuity, \ie the RS Lorentz
553: factor, evolves as
554: \begin{equation}
555:  \Gamma_R \propto R^{-(3-s)/2} (R^{2a-3} \Delta'^a)^{-1.5(4-s)/(17-4s)} \;.
556: \label{GR}
557: \end{equation}  
558: 
559:  The adiabatic index $a$ (equation \ref{GR}) of the ejecta is initially $4/3$ if the RS is relativistic, 
560: which corresponds to the thick ejecta case discussed in \S\ref{thick}, and lower for a 
561: mildly relativistic shock (thin ejecta -- \S\ref{thin}), decreasing to $5/3$ due to the 
562: ejecta cooling. Between these limiting cases, the exponents of $R$ and $\Delta'$ given in equation 
563: (\ref{GR}) change by 0.24 and 0.14, respectively. Such variations do not lead to significant changes 
564: in the solutions presented in \S\ref{RF}. For ease of further calculations, we will use 
565: $a=1.5$ in equation (\ref{GR}). 
566: 
567:  The only uncertainty left in the evolution of $\Gamma_R$ is that of the ejecta shell comoving thickness
568: $\Delta'$. This uncertainty also affects the adiabatic cooling of the electrons and the evolution of
569: the magnetic field in the ejecta. The evolution of the electron Lorentz factors $\gamma_i$ and $\gamma_c$ 
570: at $R > R_+$ is 
571: \begin{equation}
572:  \gamma_{i,c} \propto V'^{-(a_e-1)} \propto (R^2 \Delta')^{-1/3} 
573: \label{gic}
574: \end{equation}
575: where $V'$ is the comoving frame ejecta volume and an adiabatic index $a_e = 4/3$ for the relativistic
576: electrons has been used for the last term. Because the ejecta emission switches off when the decreasing
577: cooling frequency $\nu_c^{(RS)}$ falls below the observing frequency $\nu$, we will search for afterglow 
578: parameters for which $\nu_c^{(RS)}(t_{max}) > \nu$, where $t_{max}$ is the latest time when the RS
579: emission was (or thought to have been) observed. Therefore the electrons radiating at frequency $\nu$
580: cool mostly adiabatically and the ejecta radiative cooling after $R_+$ can be ignored. 
581: For the magnetic field, the flux-freezing condition yields 
582: \begin{equation}
583:   B_\perp \propto (R\Delta')^{-1} \;,\; B_\parallel \propto R^{-2}
584: \label{B2}
585: \end{equation}
586: where $B_\perp$($B_\parallel$) is the magnetic field perpendicular (parallel) to the radial direction 
587: of the fireball motion.
588:  
589:  To assess the effect of the uncertainty of the behaviour of $\Delta'$ on the ejecta synchrotron emission,
590: we consider two extreme cases: $\Delta' = const$, as could result from the compression ejecta against
591: the decelerating contact discontinuity, and $\Delta' = R/\Gamma_R$, corresponding to a comoving-frame
592: expansion of the ejecta shell at a speed comparable to the speed of light. For the former case, relating 
593: the ejecta radius with the observer time through $R \propto \Gamma_R^2 t$, leads to $R \propto t^{1/4}$ 
594: for $s=0$ and $R \propto t^{1/2}$ for $s=2$ (just as for the FS). Substituting in equation (\ref{B2}) 
595: shows that $B_\perp \propto R^{-1}$ decays slower than $B_\parallel$. Then, equations (\ref{Fp}), 
596: (\ref{nuaic}), (\ref{ga}), and (\ref{gic}) yield
597: \begin{equation}
598:  (s=0) \;\; F_p \propto t^{-0.63} \;,\; \nu_{i,c} \propto t^{-0.96} \;,\; \nu_a \propto t^{-0.61}
599: \end{equation}
600: \begin{equation}
601:  (s=2) \;\; F_p \propto t^{-0.75} \;,\; \nu_{i,c} \propto t^{-1.42} \;,\; \nu_a \propto t^{-0.72} \;.
602: \end{equation}
603: For an ejecta shell spreading law $\Delta' = R/\Gamma_R$, one obtains that $R \propto t^{1/9}$ for $s=0$ 
604: and $R \propto t^{1/5}$ for $s=2$. Equation (\ref{B2}) shows that $B_\parallel \propto R^{-2}$ decays 
605: slower than $B_\perp$, leading to
606: \begin{equation}
607:  (s=0) \;\; F_p \propto t^{-0.67} \;,\; \nu_{i,c} \propto t^{-1.19} \;,\; \nu_a \propto t^{-0.41}
608: \label{scale0}
609: \end{equation}
610: \begin{equation}
611:  (s=2) \;\; F_p \propto t^{-0.80} \;,\; \nu_{i,c} \propto t^{-1.47} \;,\; \nu_a \propto t^{-0.47} \;.
612: \label{scale2}
613: \end{equation}
614: From the above scalings, it can be seen that the temporal index of the break frequencies changes by 
615: $\sim 0.2$ ($0.05-0.25$) for $s=0$ ($s=2$), while that of the peak flux by 0.05 for either type of 
616: medium. The solutions presented in \S\ref{RF} vary little between the above assumed behaviours 
617: of $\Delta'$.
618: 
619:  Equations (\ref{scale0}) and (\ref{scale2}) are used to calculate the characteristics of the RS synchrotron 
620: emission at $t > t_+$ from those at $t_+$.  The effect of radiative losses on the RS emission is estimated 
621: in a similar way as for the FS (equation \ref{rad}), by adjusting the fluxes obtained in the adiabatic case 
622: by a factor which accounts for the faster deceleration of the FS due to the radiative losses. For the highly 
623: radiative regime, the evolution of the RS Lorentz factor $\Gamma_R$ is calculated as in the adiabatic case 
624: (equations \ref{pR} and \ref{GR}) but using the scaling of the FS Lorentz factor $\Gamma_F$ with radius
625: corresponding to the radiative dynamics case (equation \ref{GFrad})\footnotemark. The absorption of the RS 
626: radio emission in the FS is also taken into account.
627: \footnotetext{Because the calculation of $\Gamma_R$ makes use of the Blandford-McKee solution for adiabatic 
628:   dynamics, our calculation of $\Gamma_R$ for radiative dynamics is only a crude approximation} 
629: 
630: 
631: \section{Reverse-Forward Shock Scenario}
632: 
633:  The formalism presented in \S\ref{dynamics} and \S\ref{radiation} allows the calculation of the 
634: RS and FS emission at a given observer time and observing frequency. These emissions depend on the dynamics 
635: of the FS and ejecta shell, \ie on the fireball kinetic energy $E$ and the particle density of 
636: the CBM ($n$ for a homogeneous medium or $A_*$ for a wind surrounding a massive star). The ejecta-shell 
637: shock-crossing time $t_+$ also depends on the fireball initial Lorentz factor $\Gamma_0$, for thin ejecta,
638: or on the duration $\tau$ of the fireball ejection, for thick ejecta. The initial Lorentz factor also 
639: determines the number of electrons in the ejecta and, therefore, the RS emission. Finally, the RS and FS 
640: emissions depend on the two microphysical parameters $\epsi$ and $\epsB$ which quantify the typical 
641: electron energy and the magnetic field. Thus, the RS emission is determined by five parameters in the 
642: thin ejecta case, or six in the opposite case, while the FS emission depends on four parameters. Note 
643: that $E$ and $n$ (or $A_*$) determine the emission of both shocks. 
644: 
645:  In this section we determine in the framework of the reverse-forward shock scenario the values of the 
646: above parameters allowed by the radio and optical emissions of the GRB afterglows 990123 and 021211, 
647: the only two afterglows for which an optical emission has been detected at early times, $\sim 100$ seconds 
648: after the burst. 
649: 
650:  Table 1 lists the properties of the burst, optical, and radio emissions of the two afterglows. 
651: For the afterglow 021211, the optical emission 
652: is decaying since the first measurement, at $t_1 = 130$ seconds after the burst (Li \etal 2003). For the 
653: afterglow 990123, the emission begins to decay at $\sim 45$ seconds (Akerlof \etal 1999), after which the 
654: burst exhibits some variability. This raises the possibility of some energy injection in the RS after 45 
655: seconds. For this reason, we choose $t_1 = 73$ seconds as the beginning of the afterglow decay, as after 
656: this epoch the burst exhibits a weaker, decaying emission. 
657: 
658:  In both cases, the early optical emission fall-offs steeper than at later times. For the afterglow 
659: 021211, the transition between these two regimes has been observed: it occurs at $t_* = 550 - 750$ 
660: seconds (Li \etal 2003). For GRB 990123, the transition is inferred to occur at $t_* = 400 - 700$ seconds 
661: (Li \etal 2003), \ie around or after the last early optical measurement at $t_2 = 610$ seconds (Akerlof 
662: \etal 1999) but prior to the next available measurement at $t_3 \sim 4.0$ hours after the burst (Kulkarni 
663: \etal 1999b).
664: 
665: \begin{table*}
666:  \begin{minipage}{180mm}
667:  \caption{Properties of optical and radio emissions used to constrain the afterglow parameters} 
668: \begin{tabular}{cccccccccccccccccccccc} \hline 
669:   GRB & z & $t_\gamma$(s) & $t_1$(s) & $F_1$(mJy) & $\alpha_{12}$ & $t_2$(s) & $t_3$ & $\alpha_{34}$ &
670:                $t_4$(h) & $F_4$($\mu$Jy) & $\beta$ & $t_{5,6,7}$(d) & $F_{5,6,7}$($\mu$Jy)     \\
671:         & (1) & (2) & (3) & (4) & (5) & (6) & (7) & (8) & (9) & (10) & (11) & (12) & (13)      \\ \hline
672:  990123&1.6 & 70-100 & 73 & 400 & 1.80 & 610 & 4.2h & 1.10 & 8.3 & 67 & 0.68-0.82 & 0.2,1.2,4.2 & 130,320,68\\
673:  021211&1.0 & 2-8 & 130 & 4.1 & 1.56 & 650 & 650s & 0.94 & 2.5 & 25 & 0.55-0.98 & 0.1,0.9,3.9 & 84,44,91  \\
674:                \hline 
675: \end{tabular}
676:  (1): burst redshift,
677:  (2): range of burst duration in various $X$-ray bands,
678:  (3): selected epoch for early optical measurement,
679:  (4): $R$-band flux at $t_1$ (Akerlof \etal 1999, Li \etal 2003),
680:  (5): $t_1 - t_2$ temporal index (Li \etal 2003) of the early optical light-curve, $F_o \propto t^{-\alpha_{12}}$,
681:  (6): last available measurement or end of the steeply falling-off optical afterglow,
682:  (7): beginning of slower decaying optical emission,
683:  (8): $t_3 - t_4$ optical light-curve index (Kulkarni \etal 1999a; Li \etal 2003), $F_o \propto t^{-\alpha_{34}}$,
684:  (9): selected epoch after $t_3$,
685:  (10): $R$-band flux at $t_4$ (Fox \etal 2003),
686:  (11): slope of optical continuum measured after $t_3$ (Holland \etal 2000; Pandey \etal 2003),
687:           $F_\nu \propto \nu^{-\beta}$,
688:  (12): selected epochs of radio measurements,
689:  (13): $2\sigma$ upper limits on radio fluxes (Kulkarni \etal 1999b; Fox \etal 2003).
690: \end{minipage}
691: \end{table*}
692: 
693: 
694: \subsection{Constraints on the FS and RS Emissions}
695: \label{RF}
696: 
697:  The slower decaying optical emission lasting for days is naturally attributed to the FS energizing the CBM. 
698: Its $t^{-1}$ decay implies that the FS injection frequency is below the optical domain at the first epoch,
699: $t_3$, when the slower decay begins:
700: \begin{equation}
701:  \nu_i^{(FS)} (t_3) < 5 \times 10^{14} \; {\rm Hz} \quad ({\rm constraint \; 1}) \;.
702: \end{equation}
703: A second constraint is set by the optical flux normalization; for this we choose an epoch $t_4 > t_3$
704: when the model flux, $F_o^{(FS)}$, is required to be within a factor 3 of the observed flux, $F_o^{(obs)}$,
705: making allowance for some uncertainty in our calculations\footnotemark:
706: \begin{equation}
707:  \frac{1}{3}\, F_o^{(obs)} (t_4) < F_o^{(FS)} (t_4) < 3\, F_o^{(obs)} (t_4) \quad ({\rm constraint \; 2}) \;.
708: \end{equation}
709: \footnotetext{This uncertainty factor determines the width of the region of allowed $\epsi-\epsB$ solutions
710:  in the lower left corner -- upper right corner direction in the figures, and does not affect significantly 
711:  the conclusions that will be drawn.}
712: 
713:  With the exception of the flare seen at $t = 1.2$ days in the radio afterglow of GRB 990123, there are 
714: no other detections in the radio down to 0.1 mJy or less. We use the $2\sigma$ upper limits on the radio 
715: flux at three epochs ($t_{5,6,7}$) spanning the interval 0.1--4 days, to constraint the radio FS 
716: emission, $F_r^{(FS)}$:
717: \begin{equation}
718:  F_r^{(FS)} (t_{5,6,7}) < F_r^{(2\sigma)} (t_{5,6,7})  \quad ({\rm constraint \; 3}) \;.
719: \end{equation}
720: 
721:  That the early afterglow exhibits a steep fall-off requires that the RS has crossed the ejecta. 
722: The burst duration, $t_\gamma$, sets a lower bound on the ejecta crossing-time $t_+$, as discussed in 
723: \S\ref{thick}, thus $t_+$ is constrained by
724: \begin{equation}
725:  t_\gamma <  t_+ < t_1  \quad ({\rm constraint \; 4}) \;.
726: \end{equation}
727: Furthermore, the steep $t^{-1.6}$ decay of the RS emission at $t_1$ requires that the RS injection frequency 
728: is below the optical domain at that time:
729: \begin{equation}
730:  \nu_i^{(RS)} (t_1) < 5 \times 10^{14} \; {\rm Hz} \quad ({\rm constraint \; 5}) \;.
731: \end{equation}
732: Matching the observed flux to within a factor of 3,
733: \begin{equation}
734:  \frac{1}{3}\, F_o^{(obs)} (t_1) < F_o^{(RS)} (t_1) < 3\, F_o^{(obs)} (t_1) \quad ({\rm constraint \; 6}) \;.
735: \end{equation}
736: is another requirement set on the calculated RS emission. The detection of RS emission until epoch $t_2$, 
737: when the early observations end (GRB 990123) or when the transition to the FS emission is observed (GRB 
738: 021211), implies that the RS cooling frequency remains above the optical domain until at least epoch $t_2$
739: \begin{equation}
740:  \nu_c^{(RS)} (t_2) > 5 \times 10^{14} \; {\rm Hz} \quad ({\rm constraint \; 7}) 
741: \end{equation}
742: otherwise, the RS optical emission would exhibit a sharp drop when $\nu_c^{(RS)}$ falls below optical.
743: Finally, the radio upper limits are imposed on the RS emission as well:
744: \begin{equation}
745:  F_r^{(RS)} (t_{5,6,7}) < F_r^{(2\sigma)} (t_{5,6,7})  \quad ({\rm constraint \; 8}) \;.
746: \end{equation}
747: 
748:  We search for afterglow parameters $(\Gamma_0,E;n/A_*;\epsi,\epsB)$ that lead to FS and RS emissions
749: satisfying the constraints 1--3 and 5--8, respectively. For GRB 990123, constraint 4 requires that 
750: $t_+ \sim 70$ s, as for this burst $t_\gamma \sim t_1$. For GRB 021211, the same constraint allows 
751: that $4\,{\rm s} < t_+ < 130\,{\rm s}$. In the thin ejecta case, $t_+$ determines the ejecta initial
752: Lorentz factor, $\Gamma_0$ (equations \ref{Et0} and \ref{Et1}). For thick ejecta, $t_+$ determines the 
753: duration of the ejecta release, $\tau$, (equation \ref{Et2}).  Note that in the case of a thick ejecta 
754: shell, the reversed inequalities given in equations (\ref{Gmin0}) and (\ref{Gmin2}) provide only a lower 
755: limit on $\Gamma_0$. Nevertheless, $\Gamma_0$ remains a relevant parameter because its sets the number 
756: of electrons in the ejecta. 
757: 
758:  The index $p$ of the electron energy distribution (equation \ref{dNdg}) is determined from the exponent 
759: $\alpha$ of the optical power-law decay, $F_\nu \propto t^{-\alpha}$, for each shock. The available 
760: measurements of the slope $\beta$ of the optical continuum,  $F_\nu \propto \nu^{-\beta}$, at $t \sim 1$ 
761: day, constrain the index $p$ through that the intrinsic afterglow spectral slope $\beta_o$ cannot be
762: larger than observed, as intrinsic spectra harder than observed ($\beta_o < \beta$) can be attributed
763: to dust reddening in the host galaxy. 
764: 
765: 
766: \subsection{Results}
767: 
768:  The search for afterglow parameters is done by choosing various combinations of parameters $(E,n)$ 
769: (for $s=0$) or $(E,A_*)$ (for $s=2$) and by identifying the regions in the $(\epsi, \epsB)$ parameter 
770: space which satisfy the above constraints. Various values of the initial Lorentz factor $\Gamma_0$ 
771: satisfying constraint 4 are tried, to maximize the allowed $(\epsi, \epsB)$ parameter range. For thin 
772: ejecta, the $(E,n/A_*;\epsi,\epsB)$ parameter space for the RS is significantly smaller than for thick 
773: ejecta. For brevity, we present here only solutions for the latter case.
774: 
775:  Figures \ref{jans0}--\ref{decs2} show the RS and FS solutions
776: $(\epsi,\epsB)$ for various combinations $(E,n)$ or $(E,A_*)$ for which both RS and FS solutions exist. 
777: For parameters $E$, $n$ (or $A_*$) different by a factor 10 than those shown, the emission from one of 
778: the shocks, or from both, fails to satisfy the above constraints. For all these three figures, the 
779: adiabatic index in the shocked ejecta (\S\ref{rs}) was set to $a=1.5$ and the comoving thickness 
780: of the ejecta was assumed to evolve as $\Delta' = R/\Gamma_R$, where $\Gamma_R$ is the bulk Lorentz 
781: factor of the shocked ejecta at $R > R_+$. Taking $a = 4/3$, as for a relativistic RS, or assuming
782: a constant $\Delta'$ lead to a modest change in the solutions shown and leave unaltered the conclusions
783: below. To assess the effect of radiative losses, the fluxes obtained in the adiabatic case were decreased
784: as expected from the faster deceleration of $\Gamma$ in the case of radiative dynamics (equation \ref{rad}).
785: 
786:  For the reverse-forward shock scenario we reach the following conclusions about the RS and FS 
787: parameters: 
788: 
789: \begin{figure*}
790: \begin{minipage}{18cm}
791: \centerline{\psfig{figure=fig1.eps,width=12cm}}
792: \caption{{\bf Reverse-forward shock scenario}, {\sl GRB 990123, homogeneous circumburst medium (CBM)}: 
793:   Reverse and forward shock microphysical parameters $(\epsi,\epsB)$ for the minimal electron energy 
794:   and magnetic field energy, satisfying the constraints described in \S\ref{RF}. 
795:   An uncertainty of a factor 3 is assumed in the calculated optical fluxes, however no uncertainty 
796:   factor is assigned to the analytical radio fluxes, as they are compared with $2\,\sigma$ observational 
797:   upper limits.  "RS" and "FSb" denote reverse and forward shock solutions, respectively, the FS solutions 
798:   corresponding to a cooling frequency above (blueward of) the optical domain.
799:   The solutions shown are for thick ejecta (\S\ref{thick}), with an observer-frame RS ejecta crossing-time 
800:   $t_+ = 70$ s, and for highly radiative dynamics (\S\ref{fs}). There are no FS solutions for adiabatic 
801:   dynamics. Each panel specifies the isotropic-equivalent of the blast-wave kinetic energy, $E$ (in ergs),  
802:   and CBM particle density, $n$ (in $\cm3$). For values of $(E,n)$ differing by a factor 10 or more than 
803:   those shown here, the emission of at least one of the shocks becomes incompatible with the observations. 
804:   For comparison, the isotropic equivalent of the gamma-ray output of this burst is $E_\gamma = 3 \times 
805:   10^{54}$ ergs (Kulkarni \etal 1999b).  }
806: \label{jans0}
807: \end{minipage}
808: \end{figure*}
809: 
810: \begin{figure*}
811: \begin{minipage}{18cm}
812: \centerline{\psfig{figure=fig2.eps,width=16.5cm}}
813: \caption{{\bf Reverse-forward shock scenario}, {\sl GRB 990123, wind-like CBM}:
814:   Solutions for radiative dynamics and a CBM shaped by the wind blown by a massive star. 
815:   The parameter $A_*$ parameterizes the wind particle density $n(r) = \dot{M}/(4\pi m_p r^2 v_w) = 
816:   3 \times 10^{35} A_* r_{cm}^{-2}$, where $\dot{M}$ is the mass-loss rate and $v_w$ is the wind 
817:   velocity, with $A_*=1$ corresponding to $\dot{M}/v_w = 10^{-5}\,\Msunyear / 10^3\, \kms$. 
818:   Higher values of the $A_*$ than shown here, such as those expected for Wolf-Rayet stars 
819:   ($A_* \simeq 1$), do not allow RS solutions, for either adiabatic or radiative dynamics. 
820:   Overlapping FS and RS solutions are shown with larger symbols. There are no RS solutions 
821:   for adiabatic dynamics.  }
822: \label{jans2}
823: \end{minipage}
824: \end{figure*}
825: 
826: (1) {\sl GRB 990123, homogeneous CBM (figure \ref{jans0}):} 
827:   RS and FS microphysical parameters differ, with $\epsB^{(RS)} \simg 100\, \epsB^{(FS)}$ and 
828:   $\epsi^{(RS)} < 0.1 \, \epsi^{(FS)}$.
829:   A RS magnetic field larger than that behind the FS points to an ejecta which was initially magnetized.
830:   FS solutions correspond to a cooling frequency, $\nu_c^{(FS)}$, above the optical domain. 
831:   The power-law decay of the FS optical light-curve, $\alpha_{34}$, sets the electron index: 
832:   $p = (4\alpha_{34} + 3)/3 = 2.47 \pm 0.07$. This implies that the slope of the intrinsic 
833:   optical continuum is $\beta_o = (p-1)/2 = 0.74 \pm 0.04$, consistent with the value reported 
834:   by Holland \etal (2000), at $t=1-3$ days after the burst. FS solutions are obtained only 
835:   for high radiative losses. For the FS parameters shown in figure \ref{jans0}, the ratio of 
836:   the (inverse-Compton) cooling to injection electron Lorentz factors is $\gamma_c/\gamma_i 
837:   \sim 10\, (t/0.1\,d)^{2/3}$, therefore the energy given to electrons which cool radiatively 
838:   ($\gamma > \gamma_c$) is a fraction $f(t) = (\gamma_c/\gamma_i)^{2-p} = 0.34\, (t/0.1\,d)^{-1/3}$ 
839:   of the total electron energy. If the injected electron distribution extends to arbitrarily 
840:   high energies, the fraction of the post-shock energy in electrons is $\epse = [(p-1)/(p-2)] 
841:   \epsi = 3\, \epsi \in (0.3, 1)$. Then, at $t = 0.1$ day, the radiative losses over one dynamical 
842:   timescale are a fraction $\xi_{rad} (0.1\, d)= f \epse \simeq \epsi \in (0.1,0.3)$ of the FS 
843:   energy. If the total electron energy does not exceed equipartition ($\epse \leq 0.5$), then
844:   $\xi_{rad} (0.1\, d)\siml 0.15$. Therefore, for the FS solutions shown in figure \ref{jans0}, 
845:   the fireball dynamics is between the adiabatic and highly radiative regimes. 
846: 
847:  The ejecta kinetic energy $E \simg 10^{55}$ ergs (if spherical) and the ambient medium density 
848: $n \simg 0.1\, \cm3$ for which the RS emission accommodates the early optical emission of the afterglow 
849: 990123 are 10 and 50 times larger than their upper limits found by us (Panaitescu \& Kumar 2001) from 
850: multiwavelength afterglow modelling. The difference is caused by the inclusion in the current 
851: calculations of the early ($t < t_*$) optical emission, which cannot be accommodated by the RS if 
852: the fireball energy and CBM density were those determined from modelling the afterglow emission at 
853: $t > 4$ hours, as the constraints 6 and 8 in \S\ref{RF} cannot be satisfied simultaneously. 
854: Conversely, for the ($E,n$) values for which the RS emission accounts for the early optical emission, 
855: the FS solutions identified in this work satisfy the general constraints imposed by observations 
856: (\S\ref{RF}) but provide a $\chi^2$-wise poorer fit to the afterglow data after 4 hours than 
857: the best fit obtained numerically with more accurate calculations of radiative losses and the
858: integration of received emission over the photon equal arrival time surface. 
859: 
860: \begin{figure*}
861: \begin{minipage}{18cm}
862: \centerline{\psfig{figure=fig3.eps,width=16.5cm}}
863: \caption{{\bf Reverse-forward shock scenario}, {\sl GRB 021211, wind-like CBM}:
864:   Solutions for radiative dynamics and $t_+ = 100$ s. Adiabatic dynamics also allow solutions with 
865:   the same microphysical parameters behind the shocks. There are no RS solutions for denser winds 
866:   (larger $A_*$), or fireball energies $E \siml 3\times 10^{52}$ ergs, for either adiabatic or 
867:   radiative dynamics. The isotropic-equivalent $\gamma$-ray output of this burst is $E_\gamma = 
868:   6 \times 10^{51}$ ergs (Vreeswijk \etal 2002).  }
869: \label{decs2}
870: \end{minipage}
871: \end{figure*}
872: 
873: (2) {\sl GRB 990123, $r^{-2}$ CBM (figure \ref{jans2}):} 
874:   there are RS and FS solutions with the same microphysical parameters. RS solutions exist only 
875:   for radiative dynamics. Denser wind environments are found to allow RS solutions if the kinetic 
876:   energy $E$ is increased, but even for $E = 10^{56}$ ergs, which is 30 times larger than the GRB 
877:   output, the allowed wind density is well below that of a Wolf-Rayet (WR) star ($A_* \simeq 1$). 
878:   FS solutions correspond to $\nu_c^{(FS)}$ above the optical domain, therefore the electron 
879:   index is $p = (4\alpha_{34} + 1)/3 = 1.83 \pm 0.08$, leading to $\beta_o = (p-1)/2 = 0.42 
880:   \pm 0.04$, \ie significantly harder than any reported measurement. The RS and FS solutions
881:   shown in figure \ref{jans2} with $\epsi \sim 10^{-2}$ are consistent with high radiative
882:   losses if electrons reach equipartition ($\epse = 0.5$), because $p < 2$ allows most of the 
883:   electron energy to be at $\gamma > \gamma_c$. However, for the solutions with $\epsi \sim 0.1$, 
884:   equipartition energies require a cut-off in the accelerated electron distribution at 
885:   $\gamma_{cut} < \gamma_c$. The synchrotron characteristic frequency for the $\gamma_{cut}$
886:   electrons is above the optical domain (thus the cut-off does not affect the optical afterglow), 
887:   but $\gamma_{cut} < \gamma_c$ implies that all injected electrons are cooling adiabatically, 
888:   \ie the assumption of high radiative losses becomes invalid for $\epsi \simg 0.1$. 
889: 
890: (3) {\sl GRB 021211, homogeneous CBM :} 
891:    in our previous study (Kumar \& Panaitescu 2003) we have shown that, in the thin ejecta shell 
892:    case, the RS magnetic field parameter must be $10^3-10^4$ times larger than for the FS. For a 
893:    thick ejecta shell, corresponding to a RS shell-crossing time longer than the burst duration 
894:    (2--8 s) but shorter the time of the first optical measurement (130 s), there are RS and FS 
895:    solutions with the same microphysical parameters if the dynamics is radiative.
896: 
897: (4) {\sl GRB 021211, $r^{-2}$ CBM (figure \ref{decs2}):} 
898:   RS and FS may have the same microphysical parameters, for either adiabatic or radiative dynamics. 
899:   There are also solutions with the same parameter $\epsi$ and a RS magnetic field parameter larger 
900:   than for the FS, indicative of a frozen-in magnetic field. FS solutions correspond to $\nu_c^{(FS)}$ 
901:   above the optical domain, leading to $p = 1.59 \pm 0.08$ and $\beta_o = 0.29 \pm 0.04$, which is 
902:   $2\sigma$ below the hardest slope reported (Pandey \etal 2003, at $t=20$ hours). Because $p < 2$, 
903:   the dynamics may be radiative for $\epsi < 0.1$ and $\epse = 0.5$, while a significantly lower 
904:   $\epse$ would ensure an adiabatic dynamical regime. Note that a high $E = 10^{54}$ ergs, more than 
905:   100 times larger than the burst output, is required by a wind density corresponding to $A_* = 0.3$, 
906:   \ie slightly below that of a WR star. From the constraints on the FS parameters, Chevalier, Li 
907:   \& Fransson (2004) have also concluded that a weak wind with $A_* \sim 0.01$ is required for the 
908:   afterglow 021211, however, in their calculations, the FS cooling frequency was placed below the 
909:   optical. If we impose the same constraint here, we do not find any solutions for the RS microphysical 
910:   parameters.
911: 
912: 
913: 
914: \section{Wind-Bubble Scenario}
915: \label{BB}
916: 
917:  The similarity of the decay indices of the optical light-curves of the afterglows 990123 and 021211 
918: before and after $t_* \sim 600$ may suggest that a single mechanism produces the entire optical 
919: afterglow emission. Time-varying microphysical parameters, including the slope of the power-law 
920: electron energy distribution, could cause a change in the optical light-curve decay index, however 
921: such an explanation is in contradiction with the consistency seen in many afterglows between the 
922: optical spectral slope and light-curve decay index at times of order 1 day.
923: 
924:  If the FS is the only mechanism producing the detected afterglow emission and if microphysical 
925: parameters are constant, then the non-monotonic behaviour of the optical light-curves of the afterglows 
926: 990123 and 021211 at $t_* \sim 600$ seconds must be tied with the fireball dynamics. The fireball 
927: dynamics is determined by the ejecta initial energy and CBM density. A substantial energy injection 
928: can mitigate the afterglow dimming rate, however the energy deposition would have to last for the 
929: entire duration of the slower power-law decay, \ie until at least a few days, and could lead to a 
930: too bright radio emission from the RS. 
931: 
932:  Besides energy injection, a sudden variation in the radial profile of the CBM could also alter 
933: the afterglow behaviour. Such a variation is suggested by the association of GRBs with the death 
934: of massive stars, which drive powerful winds, and that the modelling of multiwavelength afterglow
935: measurements starting a few hours after the burst leads to better fits for a homogeneous medium
936: than a wind. This discrepancy may be resolved if the afterglow does not arise in a freely
937: expanding wind, but in the environment resulting from the interaction of the wind with the
938: circumstellar gas (Wijers 2001) or with the winds blown by other stars (Scalo \& Wheeler 2001).
939: It is then possible that the GRB ejecta run into a CBM whose density profile at
940: smaller radii is the $r^{-2}$ expected for a uniform, free wind, and closer to uniformity at 
941: larger distances. Then, if the FS cooling frequency is above the optical domain, the optical 
942: afterglow light-curve index should decrease by $\delta \alpha = 0.5$ when the wind termination 
943: shock is reached. The resulting index decrease is slightly smaller than observed, but it is possible 
944: that deviations from uniformity of the environment outside the unperturbed wind account for the 
945: difference.
946: 
947:  One important issue for this wind-bubble scenario is under what conditions the wind termination 
948: shock is located at the radius $R_*$ of the afterglow at the time $t_*$ when the light-curve transits 
949: from a steeper to a slower decay. From equations (\ref{R2}) and (\ref{time}), one obtains
950: \begin{equation}
951:  R_* \sim 1.4\times 10^{16} \; \left( \frac{E_{53}}{A_*} \frac{t_*}{600\;{\rm s}} \frac{2}{1+z} 
952:                    \right)^{1/2} \; {\rm cm} \;.
953: \label{Rtrans}
954: \end{equation} 
955: $R_*$ is higher by a factor of a few for the wind parameters $A_* \sim 0.1$ which we find for 
956: this scenario, and by an extra factor  $\simg 10$ for the high ejecta kinetic energy obtained 
957: for the afterglow 990123. Thus, we shall find that $R_* \sim 0.3$ pc for the afterglow 990123 
958: and $R_* \siml 0.02$ pc for the afterglow 021211. Because the afterglow radius increases as 
959: $R \propto t^{1/4}$ at $R > R_*$ and because the slower decay of the afterglows 990123 and 021211 
960: is seen from $t = t_*$ untill $t \sim$ few days, the uniform part of the CBM must extend up to
961: at least $5\, R_*$.
962: 
963:  Castor, McCray \& Weaver (1975) have derived the major physical properties of a bubble resulting 
964: from the interaction of stellar winds with the interstellar gas, taking into account the cooling 
965: and the diffusion of the interstellar gas into the shocked wind. The radius of the wind termination 
966: shock $R_t$ can be estimated from the equality of the wind ram pressure and that inside the bubble,
967: leading to $R_t = 4\, \dot{M}_{-5}^{0.3} v_{w,3}^{0.1} n_0^{-0.3} t_5^{0.4}$ pc, where $\dot{M}_{-5}$ 
968: is the mass-loss rate in $10^{-5}\, \Msunyear$, $v_{w,3}$ is the wind velocity in km/s, $n_0$ is
969: the interstellar gas density in $\cm3$, and and $t_5$ is the duration of the wind measured in 
970: $10^5$ years. From here, the ratio of the contact discontinuity radius $R_{cd}$ to that of the wind 
971: termination shock is $R_{cd}/R_t = 2.3\, \dot{M}_{-5}^{-0.1} v_{w,3}^{0.3} n_0^{0.1} t_5^{0.2}$. 
972: Thus for a GRB occurring in a dense cloud ($n > 10^5\, \cm3$), the termination shock radius 
973: could be at location required by the CBM scenario (equation \ref{Rtrans}) and the shocked wind 
974: shell could be sufficiently thick. 
975: 
976:  However, WR winds do not interact with the interstellar medium but with the wind expelled during 
977: the red supergiant (RSG) phase, which collides with the main-sequence phase wind decelerated 
978: by the interaction with the interstellar medium.
979: The numerical hydrodynamical calculations of Ramirez-Ruiz \etal (2001) take into account the 
980: wind history and show that $R_t \sim 0.02$ pc for $n = 1 \,\cm3$ and $t = 10^6$ years. Such a 
981: termination shock radius is suitable for the wind-bubble scenario and the afterglow 021211,
982: however the wind-bubble size (0.3 pc) shown by  Ramirez-Ruiz \etal (2001) is surprisingly small,
983: being 100 times less than that expected from the analytical results of Castor \etal (1975). 
984: Chevalier \etal (2004) have considered the possibility that a high interstellar pressure may 
985: stall the bubble expansion. For the external pressure expected in a intense starburst region, 
986: their numerical simulations lead to a wind shock termination radius $R_t = 0.4$ pc and a contact
987: discontinuity located at $R_{cd}  = 4\, R_t$, which is about right for the wind-bubble 
988: scenario and the afterglow 990123.
989: 
990:  Alternatively, the uniformity of the CBM at $R > R_*$ required by the wind-bubble scenario 
991: might arise from a sudden increase in the wind speed, leading to a inner shock propagating into 
992: the incoming wind. The self-similar solutions derived by Chevalier \& Imamura (1983) for colliding 
993: winds show that a thick, uniform density shell forms behind the inner shock if the termination 
994: shock moves at less than 1\% of the unshocked wind speed, which requires the fluctuation in the 
995: wind to consist of a decrease in the mass-loss rate by a factor 100 and an increase of the wind 
996: speed increases by a factor 100 or larger. Such a dramatic change in the wind properties exceeds 
997: that expected at the transition from a luminous blue variable (LBV) wind ($\dot{M}\simg 10^{-3}\, 
998: \Msunyear$, $v_w = 200\, \kms$; Garcia-Segura, Mac Low, \& Langer 1996) or a RSG wind ($\dot{M} = 
999: 10^{-4}\, \Msunyear$, $v_w \siml 100\, \kms$; Garcia-Segura, Langer \& Mac Low 1996) to a Wolf-Rayet 
1000: wind ($\dot{M} = 10^{-5}\, \Msunyear$, $v_w \simg 1000\, \kms$). The inner shock speed $v_{sh} = 
1001: v_w^{(WR)}/100 = 10\, \kms$ and the location of the termination shock required by the afterglows 
1002: 990123 and 021211 imply a WR lifetime $R_*/v_{sh} = 3\times 10^4$ and $2 \times 10^3$ years, 
1003: respectively, \ie much shorter than predicted by evolutionary models for such stars.
1004: 
1005: \begin{figure*}
1006: \begin{minipage}{18cm}
1007: \centerline{\psfig{figure=fig4.eps,width=12cm}}
1008: \caption{{\bf Wind-bubble scenario}, {\sl GRB 990123}:
1009:   Forward shock microphysical parameters shown separately for the wind-like, inner region of the CBM
1010:   (marked "$n \propto r^{-2}$"), corresponding to the early, fast decaying, optical afterglow ($t < 650$s), 
1011:   and the homogeneous, outer part of the CBM (denoted by "$n \propto r^0$"), where the $t > 4$ hours, 
1012:   slower falling-off emission arises. These solutions were calculated assuming radiative dynamics, 
1013:   which is consistent with the resulting values of the microphysical parameters for $n \propto r^0$. 
1014:   Each panel indicates the fireball kinetic energy $E$, the wind parameter $A_*$, and the resulting
1015:   density $n$ (equation \ref{nout}) of the uniform outer medium. There are no solutions for the 
1016:   the $n \propto r^0$ region for denser winds or lower fireball energies. The lack of overlap between 
1017:   the wind and uniform medium solutions indicate that this scenario requires the parameters for magnetic
1018:   field and minimal electron energy behind the FS to vary when the wind-bubble termination shock is 
1019:   encountered. }
1020: \label{janwb}
1021: \end{minipage}
1022: \end{figure*}
1023: 
1024: \begin{figure*}
1025: \begin{minipage}{18cm}
1026: \centerline{\psfig{figure=fig5.eps,width=12cm}}
1027: \caption{{\bf Wind-bubble scenario}, {\sl GRB 021211}:
1028:  Same as in figure \ref{janwb} but for the afterglow 021211. The allowed regions for the microphysical
1029:  parameters for the wind and uniform portions of the CBM are shown. In the left panel, the assumption 
1030:  of adiabatic dynamics is justified by the resulting total electron energy $\epse = (p-1)\epsi/(p-2) = 
1031:  5 \epsi < 0.1$ for the required electron distribution index $p = 2.25$. In the right panel, for the 
1032:  homogeneous CBM region, neither dynamical regime is consistent with the microphysical parameters shown, 
1033:  thus the correct $n \propto r^0$ solutions should lie somewhere between the two extreme regimes,
1034:  most likely not overlapping with the wind ($n \propto r^{-2}$) solutions.} 
1035: \label{decwb}
1036: \end{minipage}
1037: \end{figure*}
1038: 
1039:  Having found some support for the wind-bubble scenario in the stalled WR wind-bubble model
1040: of Chevalier \etal (2004), and less so in the interaction between the RSG/LBV and WR winds, we
1041: proceed with testing it against the radio and optical observations of the afterglow 990123 and 
1042: 021211. The wind-bubble scenario must satisfy the constraints given in \S\ref{RF},
1043: the first three pertaining to the emission from the fireball interacting with the homogeneous 
1044: portion of the CBM, while the last three referring to the FS propagating in the $r^{-2}$ bubble
1045: (instead of the "RS", as indicated in those equations). In addition, the constraint
1046: \begin{equation}
1047:  \nu_c^{(FS)} (t_1, t_4) > 5 \times 10^{14} \; {\rm Hz}
1048: \label{nuc}
1049: \end{equation}
1050: must be imposed, to explain the decrease of the afterglow dimming rate at $t_*$, when the uniform 
1051: medium is encountered. We note that, for $s=2$, $\nu_c$ increases in time, while for $s=0$ it decreases. 
1052: Thus, if condition (\ref{nuc}) is satisfied, then $\nu_c^{(FS)}$ is above the optical domain for any 
1053: $t \in (t_1,t_4)$.  We also note that this scenario has only four parameters, two ($E$ and $A_*$) for 
1054: the FS dynamics and two ($\epsi$ and $\epsB$) for the emission. The density of the uniform region of 
1055: the CBM is determined by the compression by a factor 4 of the wind density at the location of the 
1056: termination shock. From equation (\ref{Rtrans}), one obtains that this density is
1057: \begin{equation}
1058:   n_{s=0} = \frac{10^{36}\,A_*}{R_{*,{\rm cm}}^2} = 6 \times 10^3\,
1059:           \frac{A_*^2}{E_{53}} \left( \frac{t_*}{600\;{\rm s}} \frac{2}{1+z} \right)^{-1} \; \cm3 \;.
1060: \label{nout}
1061: \end{equation}
1062: The density jump across the termination shock should lead to a brief brightening of the afterglow
1063: (Wijers 2001).
1064: Such a behaviour may have been missed in the afterglow 990123, where there is a gap in the optical 
1065: observations from 10 minutes to 4 hours. It is not seen in the optical measurements of the afterglow
1066: 021211 at the "break" time of 10 minutes determined by Li \etal (2003), instead an optical emission 
1067: brighter by 0.5 magnitudes than the double power-law fit used there is seen at 2 hours after the 
1068: burst. The lack of a brightening at the right time in the afterglow 021211 may be problematic for 
1069: the wind-bubble scenario considered here.
1070: 
1071:  Figures \ref{janwb} and \ref{decwb} display separately the FS microphysical parameters which 
1072: accommodate the observations before and after $t_*$, for a few combinations of fireball energy 
1073: $E$ and wind parameter $A_*$. Smaller values of the former parameter or larger values for the latter 
1074: do not allow the FS to accommodate the $t > t_*$ radio and optical emission of the afterglows
1075: 990123 and 021211. Note that, because the RS emission has at least one free parameter ($\tau$ and
1076: $\Gamma_0$), requiring that it does not exceed the measured optical fluxes or the radio upper limits, 
1077: does not constrain the FS parameters. 
1078: 
1079:  As shown in figure \ref{janwb}, the parameters $\epsi$ and $\epsB$ satisfying the observational 
1080: constraints cannot be constant across $R_*$ for the afterglow 990123, with the electron energy 
1081: parameter increasing by a factor $\simg 100$ and the magnetic field parameter decreasing by a 
1082: factor $\simg 100$ when the fireball crosses the wind termination shock. A similar conclusion is 
1083: reached for the afterglow 021211 (figure \ref{decwb}). 
1084: 
1085: 
1086: \section{Conclusions}
1087: 
1088:  We have investigated two scenarios that can account for the behavior of the early optical emission
1089: of GRB afterglows 990123 and 021211, whose light-curves fall-off as $t^{-1.7 \pm 0.1}$ at $t < t_* 
1090: \simeq 600$ seconds, while the decay at $t > t_*$ follows $t^{-1.0 \pm 0.1}$. The first scenario is 
1091: the widely used reverse-forward shock scenario, where the fast decay of the early optical emission 
1092: is attributed to the GRB ejecta energized by the reverse shock (RS), and the slower decaying phase 
1093: is associated with circumburst medium (CBM) swept-up by the forward shock (FS).
1094: 
1095:  Figure \ref{jans0} shows that, for a homogeneous medium, the reverse-forward shock scenario can 
1096: accommodate the radio and optical measurements of the afterglow 990123 if the ejecta magnetic field 
1097: is $\simg 10$ times larger than in the shocked CBM, which implies an ejecta frozen-in magnetic field 
1098: (\Mesz \& Rees 1997, Zhang \etal 2003), and if the RS parameter for the typical electron energy is 
1099: $\simg 10$ times smaller than for the FS. This differences between the microphysical parameters behind 
1100: the two shocks are too large to be explained away by the inaccuracies in the calculations of the 
1101: afterglow emission presented in \S\ref{radiation}. 
1102: 
1103:  For a wind-like CBM (figure \ref{jans2}), the reverse-forward shock scenario can explain the major 
1104: properties of the radio and optical emissions of the afterglow 990123 with the same microphysical 
1105: parameters behind both shocks (figure \ref{jans2}), but it requires a wind density corresponding 
1106: to a mass-loss rate to speed ratio less than $10^{-6}\, \Msunyear/ 10^3\, \kms$ (\ie $A_* < 0.1$). 
1107: We obtain a similarly tenuous wind also from modelling the broadband data of the afterglow 990123 
1108: at $t > 0.2$ days, but it should be noted that the best fit with a wind-like CBM provides a poorer 
1109: fit to the data than a uniform medium. 
1110: 
1111:  Within the framework of the reverse-forward shock scenario, the same microphysical parameters are 
1112: obtained for the afterglow 021211, for a thick ejecta shell and either a uniform or a wind-like CBM 
1113: (the latter case is shown in figure \ref{decs2}). For a uniform CBM, we obtain $n \simg 30\, \cm3$, 
1114: larger than the $n \siml 1\, \cm3$ inferred by us (Kumar \& Panaitescu 2003) for a thin ejecta shell. 
1115: For a wind-like CBM, the afterglow 021211 requires a wind density corresponding to a mass-loss rate 
1116: to speed ratio below $10^{-6}\, \Msunyear/ 10^3\, \kms$, a result similar to that obtained for the 
1117: afterglow 990123.
1118: 
1119:  The second scenario considered in this work (\S\ref{BB}), that of wind-bubble having a inner $r^{-2}$ 
1120: wind-like region surrounded by a zone of uniform density, is motivated by that the decrease in the dimming 
1121: rate of the optical afterglows 990123 and 021211 seen at $t_* \sim 650$ seconds matches fairly well 
1122: the expectations for such a density profile. The required CBM structure finds support in the scenario 
1123: of WR wind-bubbles stalled by a high interstellar pressure discussed by Chevalier \etal (2004). 
1124: For this scenario to explain the general properties of the radio and optical emissions of the 
1125: afterglows 990123 and 021211, the magnetic field and electron energy parameters would have to 
1126: decrease and increase, respectively, by a factor of about 100 at $t=t_*$, when the wind termination 
1127: shock is reached (figures \ref{janwb} and \ref{decwb}), a contrived feature without a physical
1128: foundation. We also find that the wind-bubble scenario requires winds which are as tenuous as those 
1129: for the reverse-forward shock scenario. 
1130: 
1131:  Thus the reverse-forward shock scenario provides a more natural explanation than the wind-bubble
1132: scenario for the steep early decay of the optical emission of the afterglows 990123 and 021211.
1133: Given that the GRB ejecta can be initially magnetized and that the RS is less relativistic than 
1134: the FS, the microphysical parameters might differ behind the two shocks. If their equality is 
1135: required for a simpler scenario, with fewer assumptions, then a wind-like CBM is favoured by the
1136: reverse-forward shock scenario, though a problem still exists: the low wind density inferred in 
1137: each case ($A_* < 0.1$), which is similar to that derived by Chevalier \etal (2004) for the 
1138: afterglows 020405 and 021211 and by Price \etal (2002) for the afterglow 011211. 
1139:  In the sample of 64 Galactic WR stars analyzed by Nugis \& Lamers (2000) there is only one star 
1140: with $A_* < 0.1$, the majority of the other stars having a mass-loss rate $\dot{M} \in (0.5 - 7) 
1141: \times 10^{-5} \Msunyear$, a wind velocity $v_w \in (1000-3000)\, \kms$, and $A_* \in (0.5,3)$. 
1142: The dependence of the mass-loss rate on stellar mass and metallicity inferred by Nugis \& Lamers 
1143: (2000) lead to $\dot{M} \sim 10^{-6} (M/M_\odot)^{1.1} Y^{2.2}\, \Msunyear$ for WN stars and  
1144: $\dot{M} \sim 10^{-5} (M/M_\odot)^{1.1} Y^2 Z\, \Msunyear$ for WCs, which may suggest that 
1145: the tenuous winds required by the reverse-forward shock scenario for the afterglows 990123 and 
1146: 021211 arise WR stars which are less massive and less metal rich than Galactic WRs (Wijers 2001, 
1147: Chevalier \etal 2004). If such stars do not exist, then either the microphysical parameters
1148: must be different behind the RS crossing the GRB ejecta and the FS sweeping-up the CBM or the fast 
1149: declining early optical emission of the afterglows 990123 and 021211 is not arising in the RS. 
1150: One possibility is that the early optical afterglow emission is produced in internal shocks 
1151: occurring in an unsteady wind (\Mesz \& Rees 1999), a scenario which was not investigated in this 
1152: work.
1153: 
1154: \clearpage
1155: 
1156: \begin{thebibliography}{99} 
1157: \bibitem{} Akerlof C. \etal, 1999, Nature, 398, 400 
1158: \bibitem{} Blandford, R. \& McKee, C. 1976, Phys. of Fluids, vol. 19, no. 8, 1130
1159: \bibitem{} Castro J., McCray R., Weaver R., 1976, ApJ, 200, L107
1160: \bibitem{} Chevalier L. \& Imamura J., 1983, ApJ, 270, 554
1161: \bibitem{} Chevalier L. \& Li Z.-Y., 2000, ApJ, 536, 195
1162: \bibitem{} Chevalier L., Li Z.-Y. \& Fransson C., 2004, ApJ, submitted (astro-ph/0311326) 
1163: \bibitem{} Fox D. \etal, 2003, ApJ, 586, L5
1164: \bibitem{} Garcia-Segura G., Mac Low M.-M. \& Langer N., 1996, A\&A, 305, 229
1165: \bibitem{} Garcia-Segura G., Langer N. \& Mac Low M.-M., 1996, A\&A, 316, 133
1166: \bibitem{} Holland S., Bjornsson G., Hjorth J., Thomsen B., 2000, A\&A, 364, 467
1167: \bibitem{} Kobayashi S. \& Sari R., 2000, ApJ, 542, 819
1168: \bibitem{} Kulkarni S. \etal, 1999a, ApJ, 522, L97
1169: \bibitem{} Kulkarni S. \etal, 1999b, Nature, 389, 398
1170: \bibitem{} Kumar P. \& Panaitescu A., 2003, MNRAS, 346, 905
1171: \bibitem{} Li W., Filippenko A., Chornock R., Jha S., 2003, ApJ, 586, L9
1172: \bibitem{} \Mesz P., Laguna P. \& Rees M.J., 1993, ApJ, 415, 181 
1173: \bibitem{} \Mesz P. \& Rees M., 1997, ApJ, 476, 232
1174: \bibitem{} \Mesz P. \& Rees M., 1999, MNRAS, 306, L39
1175: \bibitem{} Nugis T. \& Lamers H., 2000, A\&A, 360, 227
1176: \bibitem{} Pandey S. \etal, 2003, A\&A, 408, L21
1177: \bibitem{} Panaitescu A. \& Kumar P., 2000, ApJ, 543, 66
1178: \bibitem{} Panaitescu A. \& Kumar P., 2001, ApJ, 560, L49
1179: \bibitem{} Price P. \etal, 2002, ApJ, 572, L51
1180: \bibitem{} Ramirez-Ruiz E., Dray L., Madau P. \& Tout C. 2001, MNRAS, 327, 829
1181: \bibitem{} Rees M.J. \& \Mesz P., 1994, ApJ, 430, L93
1182: \bibitem{} Sari R. \& Piran T., 1995, ApJ, 455, L143
1183: \bibitem{} Sari R., Piran T. \& Narayan R., 1998, ApJ, 497, L17
1184: \bibitem{} Sari R., \& Piran T., 1999, ApJ, 517, L109
1185: \bibitem{} Scalo J. \& Wheeler J., 2001, ApJ, 562, 664
1186: \bibitem{} Vreeswijk P., Fruchter A., Hjorth J., Kouveliotou C., 2002, GCN Circ. 1785
1187: \bibitem{} Wijers R. \& Galama T., 1999, ApJ, 523, 177
1188: \bibitem{} Wijers R. 2001, "Gamma-Ray Bursts in the Afterglow Era", eds. E. Costa, F. Frontera,
1189:            J. Jorth (Berlin:Springer-Verlag), 306
1190: \bibitem{} Zhang B., Kobayashi S., \& \Mesz P., 2003, ApJ, 595, 950
1191: \end{thebibliography}
1192: 
1193: 
1194: 
1195: 
1196: 
1197: \end{document}
1198: