GRB 081029: A GAMMA-RAY BURST WITH A MULTI-COMPONENT AFTERGLOW

The BAT data analysis was performed using the Swift HEASOFT 6.5.1 software package. The burst pipeline script, batgrbproduct, was used to process the BAT event data. We used the position of the optical afterglow as the source's input position during the process.

Figure 1 shows the BAT energy-resolved light curves of GRB 081029 with 10 s binning. The light curve shows an extremely weak and smooth profile with a T₉₀ duration of 280 ± 50 s (1σ, statistical). The 1 s peak flux in the 15–150 keV band measured in the 1 s time window starting from 20.6 s after the BAT trigger time is (2.8 ± 1.3) × 10⁻⁸ erg cm⁻² s⁻¹. The energy fluence in the 15–150 keV band is (2.0 ± 0.2) × 10⁻⁶ erg cm⁻². The time-average spectrum is well fitted by a simple power law with a photon index of 1.5 ± 0.2.

Zoom In Zoom Out Reset image size

Because of the weak, smooth light curve of the prompt emission, GRB 081029 satisfies the BAT possible high-z criteria (Ukwatta et al. 2008). The BAT possible high-z criteria are basically selecting those bursts with weak, smooth light curves and hard spectra. Figure 2 shows the distributions of GRB 081029, the BAT known-z bursts which satisfy the Ukwatta et al. (2008) BAT possible high-z criteria, and the BAT long GRBs in the peak flux and the fluence plane. The BAT parameters are from the BAT1 catalog (Sakamoto et al. 2008b). As seen in the figure, GRB 081029 has a lower peak flux and fluence than the higher redshift bursts such as GRB 050904, GRB 060510B, and GRB 050814. We also note that some very faint GRBs such as GRB 071122 and GRB 080604 occurred at low redshifts. Therefore, the weakness and the smoothness of the GRB 081029 hard X-ray light curve in the prompt emission might be more related to the central engine of the burst rather than the cosmological redshift effect. The Ukwatta et al. (2008) test gives a reasonable indication that a burst may be at high redshifts but the false negatives—such as GRB 080913A, GRB 090423, GRB 090429B, and GRB 090429B (Greiner et al. 2009; Tanvir et al. 2009; Zhang et al. 2009; Cucchiara et al. 2011), which were at high redshift but did not satisfy the criteria—mean that the test should be used with extreme caution.

Zoom In Zoom Out Reset image size

XRT began to observe GRB 081029 2448 s after the BAT trigger. The UVOT-enhanced X-ray position is R.A., Decl. = 23:07:05.51, −68:09:21.9 (J2000.0) with an uncertainty of 15 (radius, 90% confidence). The observational data were processed by the Swift Data Center at NASA/GSFC and further calibrated with xrtpipeline. For details of how the light curve was produced, see Evans et al. (2007). All the XRT data for GRB 081029 were collected in Photon Counting mode.

The X-ray light curve can be modeled by a broken power law (f(t)∝t^−α). The best-fitting model has decay indices of α_{X, 1} = 0.56 ± 0.03 and α_{X, 2} = 2.56 ± 0.09 with a break time of t_{X, b} = 18 230 ± 346 s yielding a goodness-of-fit of χ²/dof = 93.947/77 = 1.22. The X-ray light curve with this fit is shown in Figure 3. Alternately, if we fit a smoothly varying broken power law (Beuermann et al. 1999) we find α_{X, 1} = 0.45 ± 0.11, α_{X, 2} = 2.65 ± 0.23, and a smoothness parameter of n = 2.3 ± 1.5 with χ²/dof = 91.260/76 = 1.20. The initial X-ray light curve shows some evidence for flaring between approximately 2500 and 5000 s after the BAT trigger. The sawtooth behavior of the X-ray emission during this period is consistent with flares with Δt/t ≲ 1. It is possible that the X-ray photons that we see at this time are due to flaring on top of a power-law decay. The lack of X-ray data before 2000 s could be causing us to miss the rise of the flare and thus give the impression that the X-ray photons seen between 2000 s and 5000 s are due solely to the plateau phase of the X-ray light curve.

Zoom In Zoom Out Reset image size

The X-ray spectrum can be fitted by an absorbed power law with a photon index Γ_X = 1.98 ± 0.08, corresponding to β_X = 0.98 ± 0.08. The assumed Galactic column density value in the direction of the burst is N_H = 2.8 × 10²⁰ cm⁻² (Kalberla et al. 2005) and the fitted intrinsic column density in the host galaxy is N_H = 4.9^+3.9_−2.7 × 10²¹ cm⁻². Assuming an SMC-like relation between the neutral hydrogen column density and the V-band extinction of N_H = (15.4 × 10²¹)A_V (Equation (4) and Table 2 of Pei 1992) this corresponds to A_V = 0.3^+0.3_−0.2 mag in the rest frame of the host galaxy. However, the observed gas-to-dust ratio for GRB host galaxies varies by about a factor of 10 (Schady et al. 2007), so the X-ray data alone can only constrain the rest-frame V-band extinction to be A_V ≲ 2 mag. The observed 0.3–10 keV flux is 3.1 × 10⁻¹² erg cm⁻² s⁻¹, which corresponds to an unabsorbed value of ∼3.5 × 10⁻¹² erg cm⁻² s⁻¹. This was computed using the time-average spectrum between 2.7 × 10³ s and 6.2 × 10⁴ s after the BAT trigger.

The Swift/UVOT began settled observations 2708 s after the BAT trigger (Sakamoto et al. 2008a). An optical afterglow was detected in the initial white exposure with a magnitude of 20.47^−0.29_+0.39. The afterglow increased in luminosity until approximately 9000 s and then faded. The UVOT position of the afterglow is R.A., Decl. = 23:07:05.34, −68:09:20.0 with an estimated internal uncertainty of 014 and an estimated systematic uncertainty relative to the International Celestial Reference System (Fey et al. 2004) of 042 (Breeveld et al. 2010). These uncertainties are the 90% confidence intervals. This corresponds to Galactic coordinates of ℓ^II, b^II = 3165827, −461091. The field of GRB 081029 is shown in Figure 4. The afterglow is well isolated from other sources in the field, so there is no contamination from neighboring sources when doing aperture photometry.

Zoom In Zoom Out Reset image size

We obtained the UVOT data from the Swift Data Archive.¹⁶ These data have had bad pixels identified, mod-8 noise corrected, and have been transformed into FK5 coordinates. We used the standard UVOT data analysis software distributed with HEASOFT 6.10 along with version 20110131 of the UVOT calibration data. Photometry was done using uvotsource with circular aperture of radius 25 and a nearby circular background region with a radius of 10''. The background region was selected to have similar background properties to those at the location of the afterglow, and to be free of contaminating sources. The UVOT photometry is presented in Table 1. The photometry was calibrated to the UVOT photometric system described in Poole et al. (2008) and Breeveld et al. (2011). We have followed the Poole et al. (2008) convention and used lowercase letters to identify the UVOT bandpasses. Figure 5 shows the UVOT light curves for filters where a detection was found.

Zoom In Zoom Out Reset image size

2.4.1. REM Data

2.4.2. ROTSE Data

2.4.3. ANDICAM Data

2.4.4. UVES Spectrum

The GRB 081029 optical afterglow was observed with the high-resolution UV-visual echelle spectrograph (UVES; Dekker et al. 2000), mounted on the VLT-UT2 telescope, in the framework of the ESO program 082.A-0755. Observations began on 2008 October 29 at 02:06:37 UT (∼23 min after the Swift/BAT trigger), when the magnitude of the afterglow was R ∼ 18. Two UVES exposures of 5 and 10 minutes were obtained using both the blue and the red arms. The slit width was set to 1'' (corresponding to a resolution of R = 40,000) and the read-out mode was rebinned to 2 × 2 pixels. The spectral range of our observation is ∼3300 Å to ∼9500 Å.

The data reduction was performed using the UVES pipeline (version 2.9.7; Ballester et al. 2000). Due to the faintness of the target, the decaying magnitude during the observations, and the exposure times, the signal-to-noise ratio was not high enough to study line variability. The signal-to-noise ratio of the combined spectrum is ∼3–4, allowing the identification of the main spectral features, but not a reliable estimation of their column densities. We are able to put a 2σ upper limit on the equivalent width of the intervening Mg ii(λ2796) absorption line of 0.6 Å. A portion of the UVES spectrum with a collection of absorption features is shown in Figure 6. The redshift path analyzed is 1.2 ⩽ z ⩽ 2.7.

Zoom In Zoom Out Reset image size

The line-of-sight Galactic extinction in the direction of GRB 081029 is E_{B − V} = 0.03 ± 0.01 mag (Schlegel et al. 1998). Using the extinction law given in Roming et al. (2009) yields extinctions in the UVOT filters of A_v = 0.09, A_b = 0.12, A_u = 0.15, A_uvw1 = 0.20, A_uvm2 = 0.28, and A_uvw2 = 0.25, and A_white = 0.13 mag. For the ground-based data, we adopted the Schlegel et al. (1998) extinction values of A_R = 0.08, $A_{I_C} = 0.06$, A_J = 0.03, A_H = 0.02, and A_K = 0.01 mag in this direction.

SEDs for GRB 081029 were produced at three epochs. The first SED was constructed for T + 4000 s using data between 3000 s and 5000 s after the BAT trigger. This epoch had X-ray data, GROND data, and UVOT b- and v-band data and corresponds to the period when the optical flux was rising. The second SED was constructed for T + 12000 s using data between 9000 s and 14000 s. This epoch had X-ray and the ground-based data and corresponds to the period when the optical and near-infrared fluxes were near their peak. The final SED was computed for T + 20000 s using data between 15000 s and 25000 s after the BAT trigger. This epoch had X-ray and ground-based data and corresponds to the decay after the peak optical/near-infrared flux. Data were interpolated to a common time within each epoch using the observed light curves for each filter during the appropriate epoch.

We used uvot2pha v1.3 to convert UVOT image data to spectral files compatible with the spectral fitting package XSpec. Version 104 of the UVOT response matrix calibration was adopted for the responsivity curves. For the ground-based data, spectral files were produced for each filter using the appropriate responsivity curves and setting the magnitude to those determined from the light-curve interpolations. R and I responsivity curves were taken from Bessell (1990), and the J-, H-, and K-band responsivity curves were taken from Cohen et al. (1992a, 1992b) and Bessell et al. (1998). The GROND filter response functions²² were used for the GROND data.

XRT spectra were extracted within xselect (v2.4) over the 0.3–10 keV energy range. Source counts were extracted from a circular region centered on the source with a 50'' radius, and the background count rate was measured from a circular, source-free area in the field of view with a 150'' radius. The spectral files were grouped to ⩾20 counts per energy channel. Effective area files corresponding to the spectral files were created using the xrtmkarf tool (v0.5.6), where the exposure map was taken into account in order to correct for hot columns. Response matrices from version 10 of the XRT calibration files were used. The spectrum was normalized to correspond to the 0.3–10 keV flux of the X-ray afterglow at the epoch of the SED. The normalization was determined from the best-fit power-law decay model to the afterglow light curve, in the same way as was done for the UVOT and ground-based data.

The SEDs were fitted using XSpec (v12.4.0), first using a single power-law and then using a broken power-law spectral model. In both the power-law and broken power-law models two independent dust and gas components were included to correspond to the Galactic and the host galaxy photoelectric absorption and dust extinction, where the Galactic components were frozen to the column density and reddening values taken from Kalberla et al. (2005) and Schlegel et al. (1998), respectively. The dependence of the dust extinction on wavelength in the GRB host galaxy was modeled on the Milky Way, the Large Magellanic Clouds (LMC) and the Small Magellanic Clouds (SMC) empirical extinction laws using the XSpec model zdust, which is based on the extinction coefficients and extinction laws from Pei (1992). The total-to-selective extinction, R_V = A_V/E_{B − V} was taken to be R_V = 3.08, 2.93, and 3.16 for the Galactic, SMC and LMC extinction laws, respectively (Pei 1992). The equivalent neutral hydrogen column density in the host galaxy was determined from the soft X-ray absorption, where solar abundances were assumed.

To model the Lyman-series absorption in the 912–1215 Å rest-frame wavelength range, we used the prescription provided in Madau (1995) to estimate the effective optical depth from the Lyman series as a function of wavelength and redshift, which was coded into a local model for XSpec. As well as estimating the hydrogen absorption caused by intervening systems, Madau (1995) also determined the error on this due to statistical fluctuations in the number of absorption clouds along the line of sight. This error was added in quadrature to the photometric uncertainty of any optical data at rest-frame wavelengths blueward of Lyα.

We found that the best-fitting models were consistent with there being no measurable dust in the host galaxy along the line of sight to the burst. Since many GRB host galaxies exhibit an SMC extinction law (Stratta et al. 2004; Kann et al. 2006; Starling et al. 2007; Schady et al. 2010) we adopted this for the fits to GRB 081029's SED. However, since the amount of fitted extinction is negligible (A_V < 0.02 mag at the 3σ level, which is barely consistent with the upper limit derived in Section 2.2), the details of the extinction law do not significantly affect our results. There is no evidence for a spectral break between the X-ray and optical bands, so we simultaneously fit a simple power-law spectrum to all three epochs. Our best-fitting models for each epoch are given in Table 5 and shown in Figure 7. The extinction and H i column density were assumed to be the same at every epoch. In order to test our use of the Madau (1995) method of handling absorption from the intergalactic medium we removed the ultraviolet photometry with wavelengths less than 1215 Å and refitted the data. This resulted in no significant change to the fit presented in Table 5, so we conclude that the intergalactic medium does not significantly affect the SED of GRB 081029.

Zoom In Zoom Out Reset image size

The simple power-law model assumes that the optical and X-ray photons are produced by the same mechanism. We find that the spectrum becomes steeper by Δβ = 0.08 ± 0.02 between ∼4000 s and 12000 s. This steepening occurs at about the same time that the light curve rebrightens indicating that there is a physical change in the mechanism that produces the light during the rebrightening.

Our SED is in agreement with the results of Nardini et al. (2011), which is to be expected because most of our optical and near-infrared data were taken from their paper.

The flux-density light curves for the X-ray, optical, and infrared afterglows are presented in Figure 5. Between about 100 and 2000 s after the BAT trigger the REM optical data decay slowly with a decay index of α_{opt, 1} = 0.52 ± 0.02. The ROTSE R-band data are consistent with a smooth decay with the same decay index. We do not see the achromatic break at 940 s reported by Nardini et al. (2011) in the GROND data. However, the GROND data have considerably less photometric scatter than either the REM or ROTSE data, so this feature appears to be washed out in our data. There is, however, considerable variation in the R-band and infrared luminosity before ≈500 s. It is not clear if this variability is real or an artifact of the photometry. We note that the earliest GROND infrared data show evidence for a deviation from the smooth decay seen in the GROND optical data taken at the same time. This deviation is comparable to the variability seen in the early (t ≲ 500 s) ROTSE and REM data.

Between about 3000 and 5000 s the X-ray light curve has a decay index of α_X = 0.56 ± 0.03. The optical and infrared light curves at this time, however, rise rapidly, as is clearly seen in the GROND data (Nardini et al. 2011). The rise index is α ≈ −8 during this period. Our data are sparse during this period, but they are consistent with the GROND data. After approximately 10000 s our data show the same decay as the GROND, although our data have a decay index of α_{opt, 2} = 1.89 ± 0.25 while GROND finds α = 2.5. We attribute this difference to the sparsity of our data. The late-time X-ray decay (α_{X, 2} = 2.56 ± 0.09) is consistent with the late-time GROND optical decay.

The rapid rise seen in the GROND data (Nardini et al. 2011) (α ≈ −8) is inconsistent with the expected rise index of α = −0.5 before the cooling break. Further, the rebrightening at about 3000 s does not show any color evolution, so it is not possible to interpret the rise that is seen in the optical and infrared as the synchrotron frequency passing though the optical on its journey toward lower energies. Therefore, we conclude that the rise is not due to a synchrotron peak.

The break seen at 940 s in the early GROND light curves has a magnitude of Δα = 0.77 ± 0.08 and is achromatic. The post-break decay index is too small for the change in decay index to be due to a jet break, and the achromatic nature of the break argues against it being due to the passage of the cooling break through the optical bands. The break can be explained by energy injection turning off at 940 s. If we assume a constant density interstellar medium, and that the cooling break is above the optical bands, then having α = 0.38 ± 0.05 during energy injection and α = 1.12 ± 0.6 after energy injection stops implies an energy injection index of q = 0.5 ± 0.1 where L(t)∝t^−q and an electron distribution index of p = 2.5 ± 0.1 where N(E)∝E^−p (De Pasquale et al. 2009). Using these values we predict the spectral decay index after the early-time break to be β = 0.75 ± 0.05. However, in Section 3.1 we find that the spectral decay index between the optical and X-ray bands at 4000 s is β_OX = 0.90 ± 0.01, which is only barely consistent (a 3σ difference) with the expected value in the energy injection scenario. We computed the expected spectral index for a wind-stratified circumburst medium but were unable to find values of p and q that produced a spectral decay index that was consistent with the observed value regardless of the location of the cooling break. The only scenario that gives a spectral index that is roughly consistent with the observed β_OX = 0.90 is a constant density environment with the cooling break above the X-ray band between 900 s and 3000 s.

The energy injection scenario can explain the early-time behavior of the optical and infrared light curves, but it cannot explain the rebrightening seen at 3000 s. We need some other mechanism to do this.

We are able to reproduce most of the observed X-ray, optical, and infrared light curves as well as the observed SED if we assume a two-component jet model for the afterglow of GRB 081029. The afterglow is characterized by a rebrightening in the optical and infrared bands with a simultaneous flattening in the X-ray band. This implies that a new mechanism was contributing to the flux in the optical regime starting at about 3000 s. In the two-component jet model, the early afterglow emission (t < 2500 s) was produced by the narrow, fast component while the late rebrightening was attributed to the emergence of the radiation powered by the wider, slower component.

In our fit the deceleration time for the wide component occurs earlier than 3000s, and we find that the synchrotron frequency of the wide jet component of the afterglow is between the X-ray and optical bands (i.e., ν_opt < ν_{m, w} < ν_X) for 3000 < t < 9000 s. The passage of the wide jet's synchrotron break through the optical band cannot reproduce the rapid rise that is seen in the optical and infrared photometry at about 3000 s, suggesting that there is another process at work that contributes to the sudden increase in the flux. Further, Nardini et al. (2011) find that optical flux rises as t⁻⁸ during this time.

The physical parameters of the two components are summarized in Table 6. The half-opening angle of the jet is denoted by θ_j, Γ₀ is the Lorentz factor, E_{K, iso} is the isotropic equivalent kinetic energy in the jet, p is the electron index, _e and _B are the fractions of the energy in electrons and magnetic fields, respectively, n is the density of the circumburst medium, and z is the redshift. Details of the model and the numerical code used are given in Jin et al. (2007).

Table 6. Model Fits for a Two-component Jet

Parameter	Narrow Jet	Wide Jet
θj (rad)	0.015	0.025
Γ0	500	100
EK, iso (erg)	4.0 × 1054	3.0 × 1054
p	2.05	2.20
e	0.05	0.10
B	0.0001	0.0002
n (cm−3)	10	10
z	3.8479	3.8479

Download table as:  ASCIITypeset image

We find that the narrow, inner jet has a half-opening angle of θ_{j, n} = 0.015 rad (09) and an initial Lorentz factor of Γ_{0, n} = 500. This component gives rise to the X-ray flux and the pre-jump optical and infrared flux. The wider, outer jet has θ_{j, w} = 0.025 rad (14) and an initial Lorentz factor of Γ_{0, w} = 100. This component dominates the afterglow after about 3000 s. The total electromagnetic energy in the afterglow is approximately equally divided between the two jets.

Our two-component model predicts that the optical spectrum of the wide jet emission (which dominates at late time) has β_opt = 0.6 and the X-ray spectrum has β_X = 1.1, with the break being located at ∼10¹⁵ Hz (approximately the u band), at ∼9000 s. This is not consistent with the fits to the observed SEDs presented in Section 3.1. We tried varying the amount of extinction when fitting our models and found that the observed SEDs can be made to weakly agree with the model if there is extinction in the host galaxy along the line of sight to the burst. This is consistent with the constraints on the extinction (A_V ≲ 2 mag in the host galaxy) from the X-ray data alone (Section 2.2) but inconsistent with the stronger constraint on the host extinction (A_V < 0.03 mag) found by combining the X-ray, optical, and infrared data (Section 3.1).

As shown in Figure 8 the two-component jet model can reproduce the X-ray and some of the optical/infrared data reasonably well. However, this model fails to reproduce the rapid rise seen in the UVOT white data. Further, GROND observations show that this rapid rise occurs in all filters between the GROND g' and K_s bands and has a power-law index of α ∼ −8 (Nardini et al. 2011). This is somewhat steeper than can be accommodated with the two-component jet model. Finally, the two-component model predicts either a spectral break at about 10¹⁵ Hz or extinction along the line of sight in the host galaxy. Both of these predictions are inconsistent with the data. The two-component jet model can explain the observed light curves of the afterglow, but not the spectral behavior. Further, the two-component jet model has trouble handling the rapid rise in the flux seen at ≈3000 s.

Zoom In Zoom Out Reset image size

Vlasis et al. (2011) have proposed that collisions between ejecta shells can produce flares in the optical light curve. In their scenario two shells are ejected by the central engine. The first shell has a lower Lorentz factor than the second shell, so the second shell will eventually catch up with the first. The first shell sweeps up a uniform interstellar medium and decelerates. The second shell has a higher Lorentz factor and overtakes the first shell. The collision between the two shells produces an optical flare with properties that depend on the Lorentz factor of the second shell and the isotropic energy (E_iso). Vlasis et al. (2011) find that the collision between two shells of material with different Lorentz factors can produce an optical flare with Δt/t ∼ 1. This is approximately consistent with the GROND data for the rebrightening episode. The simulations of Vlasis et al. (2011) suggest that the magnitude of the flare, relative to the underlying synchrotron light curve, depends on the Lorentz factor and the isotropic energy. They find typical values for the increase in the flux (f ) resulting from the collision of Δf/f ∼ 2–5 for typical GRB values of Γ and E_iso. This is consistent with what is seen during the rebrightening of GRB 081029.

Figures 4 and 5 of Vlasis et al. (2011) show predicted optical light curves produced by colliding ejecta shells for four sets of Lorentz factors and isotropic energies. All four cases result in light curves that exhibit flares that have shapes and intensities that are similar to the rebrightening seen in GRB 081029. The simulations assume Γ = 23 for the first shell, which results in the onset of a flare at ∼20000 s in the rest frame. Our data suggest that GRB 081029's flare started at ∼3000 s in the observer's frame (∼600 s in GRB 081029's rest frame). However, the time of the collision will depend on the time that the second (faster) shell was ejected relative to the first (slower) shell, and on the distance of the first shell from the central engine (and the first shell's Lorentz factor). Detailed simulations will be needed to test this scenario and determine the physical properties of the ejecta.

In general, we find that by themselves neither a one-component jet nor continuous energy injection from the central engine can explain the observed light curves and SED of the X-ray, optical, and infrared afterglows of GRB 081029. A two-component jet model, similar to what is seen in some other GRB afterglows, does provide a reasonable fit to the light curves, but the Lorentz factor of the fast, narrow jet is less than expected given that the peak time of the light curve is earlier than 89 s in the observer's frame. Further, the two-component jet model is not able to reproduce the observed SEDs during the unusual optical activity.

We find that the rise in the optical light curves of the afterglow of GRB 081029 can be broadly explained by the collision of a fast-moving ejecta shell with a slower shell that has been decelerated by sweeping up a uniform interstellar medium. This scenario does not, however, address the shallow decay phase of the afterglow. The early shallow decay requires that the emission is due to a different emission component from the late-time emission. We favor the multi-component jet explanation because it does not require energy injection. The discrepancy between the observed SED (with β = 1 and A_V < 0.05 mag) and the SED that is predicted by the two-jet model (with β = 0.6 and A_V ∼ 0 mag) can be explained by the spectrum at ∼12000 s being dominated by emission from the flare caused by the collision between the two shells. Evidence for this is that at 4000 s, during the onset of the flare, the SED had β ≈ 1.0, which suggests a transition between the intrinsic SED of the narrow jet and the SED of the emission from the collision. If this is the case there is no need to invoke extinction to explain the difference between the observed and predicted spectral decay indices. This leads to a picture where GRB 081029 had a two-component jet and a collision between two ejecta shells at about 3000 s. At this time the afterglow is making a transition from being dominated by the narrow jet to being dominated by the wide jet, so it is not possible to tell if the collision between the two ejecta shells occurred in the narrow or wide jet.

The nature of GRB afterglows has been a matter of much debate over the past decade. There is a general agreement that they are the result of a combination of a forward shock due to a relativistic jet moving into the circumstellar medium surrounding the burst and a reverse shock that propagates back into the jet. However, the details of how these shocks affect their environment, the role of magnetic fields, and the structure of the jets are the subject of much research. Several GRBs have had afterglows that are difficult or impossible to explain using a single, uniform jet. A multi-component jet structure has been postulated to explain unusual behavior in the light curves of some GRBs. An example of a multi-component jet is given by Berger et al. (2003), who invoked a two-component jet to explain radio observations of the long–soft burst GRB 030329. Oates et al. (2007) found that a two-component jet explained GRB 050802's afterglow, and Holland et al. (2007) found that a two-component jet could explain the lack of a jet break in the light curves of XRF 050416A. Racusin et al. (2008) found that the afterglow emission from the "naked-eye burst" GRB 080319B is best explained using a two-component jet. A multi-component jet can also explain the afterglow of the short–hard burst GRB 051221A (Jin et al. 2007). However, the physical parameters of multi-component jets vary considerably from one GRB to another, so there does not appear to be a universal jet structure.

Multi-component or structured jets are predicted by simulations of the relativistic outflow from GRBs. Kumar & Granot (2003) found that the bulk Lorenz factor decreases as one moves away from the axis of the jet resulting in a jet with a fast inner core surrounded by a slower outer envelope. Simulations of outflows from accretion disks about collapsed massive stars show that multi-component jets can form with the outer jet carrying far more energy than the inner jet (Vlahakis et al. 2003). In this scenario the inner and outer jet will have different energies and different bulk Lorentz factors. The interaction of each jet with the circumburst medium will produce separate afterglow emission components, which can result in complex light curves (Peng et al. 2005).

Several mechanisms have been proposed to create late-time activity in the central engine of a GRB that would result in multiple ejecta events. King et al. (2005) suggest that the fragmentation of a rapidly rotating stellar core could result in multiple accretion events onto the newly formed compact object. Perna et al. (2006) pointed out that the fragmentation of an accretion disk that is undergoing viscous evolution can result in an accretion disk that results in highly variable accretion onto the central compact object. Each accretion event would restart the central engine resulting in new shells being ejected. Each accretion event would be independent, so the initial Lorentz factors of the ejecta could vary considerably leading to collisions between shells from different events.

There is evidence for colliding shells in the optical light curves of a few GRBs. The afterglow of GRB 081029 has an optical light curve that is similar to that of GRB 060206 (Stanek et al. 2007; Woźniak et al. 2006), and GRB 970508 (Sokolov et al. 1998) exhibited a late-time flare similar to what is expected from colliding shells. After 10 years such flares have only been observed in a handful of GRB afterglows, which suggests that discrete late-time accretion events may be fairly uncommon in GRBs.

A combination of the colliding shell scenario and a multi-component jet can reproduce the broad features of the light curves, but detailed modeling will be needed to determine the physical parameters governing this afterglow. Several GRB afterglows have shown evidence for multi-component jets, and there is evidence that some GRBs undergo multiple accretion events that result in late-time impulsive energy injection into the afterglow. GRB afterglows appear to be complex phenomena that require detailed modeling to be fully understood.

We detect both hydrogen absorbing features (Lyα and Lyβ) and several metallic transitions in the spectrum of the optical afterglow. The latter belong both to neutral elements (O i(λ1039) and (λ1302)), low ionization species (C ii(λ1334), S ii(λ1250), (λ1253), and (λ1259), Si ii(λ1264), (λ1304), and (λ1526), Fe ii(λ1608), Al ii(λ1670), Al iii(1854)) and high ionization species (N v(λ1238) and (λ1242), Si iv(λ1393) and (λ1402), C iv(λ1548) and (λ1550), O vi(λ1031) and (λ1037)). In addition to these features, lines from several fine structure levels of C ii(λ1334), O i(λ1304) and (λ1306), Si ii(λ1264), (λ1309), and (λ1533), and Fe ii(λ1618), (λ1621), (λ1631), and (λ1636) are detected. These features are excited by the ultraviolet flux from the GRB afterglow. Estimates of the typical distance from a GRB to absorption systems suggest distances of ∼0.1–1 kpc (Vreeswijk et al. 2007; D'Elia et al. 2009) implying that all these absorption features are due to the GRB 081029 host galaxy. The common redshift of these features is z = 3.848, which we take at the redshift of the host. We find no evidence for intervening metal absorption lines in our combined spectrum.