GRB 071003: Broadband Follow-up Observations of a Very Bright Gamma-Ray Burst in a Galactic Halo
D. A. Perley, W. Li, R. Chornock, J. X. Prochaska, N. R. Butler, P. Chandra, L. K. Pollack, J. S. Bloom, A. V. Filippenko, H. Swan, F. Yuan, C. Akerlof, M. W. Auger, S. B. Cenko, H.-W. Chen, C. D. Fassnacht, D. Fox, D. Frail, E. M. Johansson, D. Le Mignant, T. McKay, M. Modjaz, W. Rujopakarn, R. Russell, M. A. Skinner, G. H. Smith, I. Smith, M. A. van Dam, S. Yost
aa r X i v : . [ a s t r o - ph ] S e p To appear in ApJ 2008 November 10
Preprint typeset using L A TEX style emulateapj v. 03/07/07
GRB 071003: BROADBAND FOLLOW-UP OBSERVATIONS OF A VERY BRIGHT GAMMA-RAY BURST IN AGALACTIC HALO
D. A. Perley , W. Li , R. Chornock , J. X. Prochaska , N. R. Butler , P. Chandra , L. K. Pollack ,J. S. Bloom , A. V. Filippenko , H. Swan , F. Yuan C. Akerlof , M. W. Auger , S. B. Cenko , H.-W. Chen ,C. D. Fassnacht , D. Fox , D. Frail , E. M. Johansson , D. Le Mignant , T. McKay , M. Modjaz ,W. Rujopakarn , R. Russell , M. A. Skinner , G. H. Smith , I. Smith , M. A. van Dam , and S. Yost To appear in ApJ 2008 November 10
ABSTRACTThe optical afterglow of long-duration GRB 071003 is among the brightest yet to be detected fromany GRB, with R ≈
12 mag in KAIT observations starting 42 s after the GRB trigger, includingfiltered detections during prompt emission. However, our high S/N ratio afterglow spectrum displaysonly extremely weak absorption lines at what we argue is the host redshift of z = 1 . II absorption systems observed at lower redshifts. Together withKeck adaptive optics observations which fail to reveal a host galaxy coincident with the burst position,our observations suggest a halo progenitor and offer a cautionary tale about the use of Mg II for GRBredshift determination. We present early through late-time observations spanning the electromagneticspectrum, constrain the connection between the prompt emission and early variations in the light curve(we observe no correlation), and discuss possible origins for an unusual, marked rebrightening thatoccurs a few hours after the burst: likely either a late-time refreshed shock or a wide-angle secondaryjet. Analysis of the late-time afterglow is most consistent with a wind environment, suggesting amassive star progenitor. Together with GRB 070125, this may indicate that a small but significantportion of star formation in the early universe occurred far outside what we consider a normal galacticdisk. Subject headings: gamma rays: bursts — gamma-ray bursts: individual: 071003 INTRODUCTION
Concurrent observations of long-wavelength afterglowand ongoing gamma-ray burst (GRB) activity should, inprinciple, yield important constraints on the nature ofthe physical processes of the emission (eg., Kobayashi2000). However, as a GRB typically lasts less than100 s, it is challenging for large ground-based opti-cal/infrared follow-up facilities to react to a GRB alert
Electronic address: (dperley,wli)@astro.berkeley.edu Department of Astronomy, University of California, Berkeley,CA 94720-3411. Department of Astronomy and Astrophysics, UCO/Lick Ob-servatory; University of California, 1156 High Street, Santa Cruz,CA 95064. Department of Astronomy, University of Virginia, P.O. Box400325, Charlottesville, VA 22904. Jansky Fellow, National Radio Astronomy Observatory. Sloan Research Fellow. University of Michigan, Randall Laboratory of Physics, 450Church St., Ann Arbor, MI, 48109-1040. Department of Physics, University of California, 1 Shields Av-enue, Davis, CA 95616. Division of Physics, Mathematics, and Astronomy, MS 105-24,California Institute of Technology, Pasadena, CA 91125. Department of Astronomy and Astrophysics, University ofChicago, 5640 S. Ellis Ave, Chicago, IL 60637. National Radio Astronomy Observatory, P.O. Box O, Socorro,NM 87801. W. M. Keck Observatory, 65-1120 Mamalahoa Highway, Ka-muela, HI 96743. Steward Observatory, Tucson, AZ 85721. The Aerospace Corporation, Mail Stop M2-266, PO Box92957, Los Angeles, CA 90009-29957. The Boeing Company, 535 Lipoa Parkway, Suite 200, Kihei,HI 96753. Department of Physics and Astronomy, Rice University, 6100South Main, MS108, Houston, TX 77251-1892. College of St. Benedict, St. Joseph, MN 56374. quickly and take data during the prompt phase. Multi-color observations, which provide vital information onthe emission mechanism, are even more difficult to ob-tain during the prompt phase because of the addedoverhead associated with changing filters. Nevertheless,due to the coordinated efforts of recent space missions(
HETE-II , Ricker et al. 2003;
Swift , Gehrels et al. 2004)to detect GRBs and various ground-based optical follow-up programs, observations during the prompt phase ofGRBs are no longer uncommon — the optical after-glows (OAs) of several dozen GRBs have been observed(e.g., Akerlof et al. 1999; Vestrand et al. 2006; Yost et al.2007) during gamma-ray emission, and multi-color opti-cal data have been obtained in a handful of cases (e.g.,Blake et al. 2005; Nysewander et al. 2007).Observations of GRBs in the past several yearshave also revealed a rich demography in OA behav-ior. Some OAs have monotonic power-law decays (e.g.,Li et al. 2003b; Laursen & Stanek 2003), while othershave plateau (e.g., Rykoff et al. 2006) and rebrighten-ing (e.g., Wo´zniak et al. 2006) phases. Even amongGRBs with relatively simple behavior, however, short-timescale features not predicted in the basic shock mod-els often appear in sufficiently well-sampled data. Var-ious modifications to the standard picture have beenproposed to explain such observations, including thepresence of a jet with single (e.g., Sari et al. 1999)or multiple (e.g., Berger et al. 2003) components, re-freshed shocks (Zhang et al. 2006), central engine activ-ity (Kocevski et al. 2007; Chincarini et al. 2007), gravi-tational microlensing (Garnavich et al. 2000), and den-sity irregularity in the GRB environment (Holland et al.2003). Observationally, constraints on the change inthe afterglow color and the spectral energy distribution(SED) play an important role in limiting the viability ofmodels for a particular GRB.The question of the nature of the GRB itself is in-timately tied to the question of its environment andorigins. At intermediate to late times, spectroscopy ofthe afterglow (e.g., Prochaska et al. 2007; D’Elia et al.2007) and deep imaging of the host environment (e.g.,Bloom et al. 2002; Fruchter et al. 2006) can help estab-lish the nature of the GRB’s progenitor and environment,connecting what we learn about the burst itself to thelarger question of its origins and place in the early uni-verse.In this paper, we report on our photometric and spec-troscopic observations of GRB 071003 with various tele-scopes from the prompt phase to late times. In § § § §
5. We assume H ◦ = 71 km s − Mpc − , Ω M = 0 .
3, andΩ Λ = 0 . OBSERVATIONS
BAT/XRT Observations
On 2007 October 3, 07:40:55 UT (defined as t = 0 inthis paper; UT dates are used throughout), a bright GRBtriggered the Burst Alert Telescope (BAT) onboard the Swift satellite (trigger 292934 Schady et al. 2007). Thefirst GCN notice was distributed within 16 s. Unfortu-nately,
Swift was still returning to normal observationsafter its 2007 August gyro failure, but it did slew to theposition after 22 ks and began observations using theX-Ray Telescope (XRT).We downloaded the
Swift
BAT and XRT data fromthe
Swift
Archive and quicklook data site. The XRTand BAT spectra were fitted using ISIS .The XRT data were processed with version 0.11.4of the xrtpipeline reduction script from the HEA-soft 6.3.1 software release. We employ the latest(2007 December 4) XRT calibration files. Our reduc-tion of XRT data from cleaned event lists output by xrtpipeline to science-ready light curves and spectrais described in detail by Butler & Kocevski (2007a). Weuse the latest calibration files from the 2007 September24 BAT database release. We establish the energy scaleand mask weighting for the BAT event mode data by run-ning the bateconvert and batmaskwtevt tasks. Spec-tra and light curves are extracted with the batbinevt task, and response matrices are produced by running batdrmgen . To produce the BAT spectra, we apply thesystematic error corrections to the low-energy BAT spec-tral data as suggested by the BAT Digest Web site ,and fit the data in the 15–150 keV band. The spectral ftp://legacy.gsfc.nasa.gov/swift/data \protect http://swift.gsfc.nasa.gov/cgi-bin/sdc/ql \protect http://space.mit.edu/CXC/ISIS \protect http://heasarc.gsfc.nasa.gov/docs/software/lheasoft/ \protect http://swift.gsfc.nasa.gov/docs/swift/analysis/bat digest.html trigger (sec)0.00.20.40.6 R a t e ( − k e V ) • • • • • • O p ti ca l f l ux ( µ J y ) VRclearI
Fig. 1.—
Light curve from the
Swift
BAT of GRB 071003, withoptical photometry from KAIT and P60, and the optical light curvemodel discussed in Section 4.3, overplotted. The GRB is dominatedby a complicated, spiky emission episode in the first 30 s, but apulse is also observed much later, at 150 s. Optical data points(all from KAIT, except one R -band measurement from the P60),by contrast, show a power-law decay at early times followed by aslow-rising “bump.” Here the V and I filtered observations havebeen offset to match the R and unfiltered points based on therelative colors at 2000 s. normalizations are corrected for satellite slews using the batupdatephakw task.The burst exhibits one dominant emission episode ofduration dt ≈
30 s, followed by a minor pulse ∼
150 slater of duration ∼
20 s. The total duration is T =148 ± placing it clearly into the long GRB class.The primary pulse is resolved into multiple pulses. Thegamma-ray light curve is shown in Figure 1, overplottedwith early-time photometry from KAIT and P60 (dis-cussed in § § t = − . t = 169 s is acceptably fitted ( χ /ν = 47 . /
55, where ν is the number of degrees of freedom) by a power-lawmodel, with photon index α = − . ± . S γ = (1 . ± . × − erg cm − (15–350 keV).The main emission episode ( t = − . t = 22 . α = − . ± . S γ = (1 . ± . × − ergcm − , χ /ν = 56 . / t = 131–169 s) is softer ( α = − . ± . S γ = 1 . +0 . − . × − ergcm − , χ /ν = 41 . / t ≈ × s is well fitted by a power-lawtime decay t − . ± . . The time-integrated spectrumis well fitted ( χ /ν = 48 . /
54) by an absorbed power-law model [photon index Γ = 2 . ± .
12, unabsorbed F X = (5 . ± . × − erg cm − s − ]. The equiva-lent H column density, N H = (2 . ± . × cm − ,is marginally consistent with the expected Galactic col-umn density in the source direction, N H = 1 . × cm − (Dickey & Lockman 1990). Examining the X-rayhardness ratio (e.g., Butler & Kocevski 2007b), there isno evidence for spectral evolution during the XRT obser-vation. All uncertainties quoted in this paper are 1 σ , except wherespecified otherwise. KAIT Observations
The Katzman Automatic Imaging Telescope (KAIT)is a 0.76-m robotic telescope at Lick Observatory thatis dedicated to searching for and observing supernovaeand monitoring other variable or ephemeral celestialphenomena. It is equipped with a Finger Lakes In-strument (FLI) ProLine PL77 back-illuminated CCDcamera having a resolution of 0 . ′′ − and a to-tal field of view (FOV) of ∼ . ′ × . ′
8. More in-formation on KAIT can be found in Li et al. (2000),Filippenko et al. (2001), and Filippenko (2005), whilethe KAIT GRB alert system is described in detail byLi et al. (2003b). Notable KAIT observations of GRBsinclude GRB 021211 (Li et al. 2003a), GRB 051111(Butler et al. 2006), GRB 060210 (Li 2006, Li et al. 2008in preparation), and GRB 080319B (Bloom et al. 2008).Several improvements have been implemented for theKAIT GRB alert system since the description given byLi et al. (2003b). An FLI PL77 camera has replaced theApogee AP7 camera, offering a much faster readout time(1.2 s for FLI vs. 11.0 s for Apogee). A new feature hasbeen incorporated into the software so the system caneasily terminate an ongoing exposure in preparation forthe GRB response sequence. Most importantly, a real-time image-processing pipeline has been developed tocompare the KAIT images to archival Digital Sky Survey(DSS) images to identify new objects. Astrometry solu-tions are derived for the KAIT images by matching thedetected objects to the USNO B1 catalog (Monet et al.2003), providing coordinates to any new objects to a pre-cision of ∼ ′′ .2. Point-spread-function (PSF) fitting pho-tometry is also performed on new objects, and calibratedto the red magnitudes of the stars in the USNO B1 cat-alog. The image-processing results are displayed in realtime on a website. For GRB 071003, the KAIT GRB alert program re-ceived the GCN socket notice at t = 16 s. The systemimmediately terminated the ongoing supernova searchprogram and began to slew the telescope to the GRB po-sition. After slewing from close to meridian to an hourangle of 4.2 hr, a sequence of 5 × t = 42 s. KAIT then switched to a sequencethat alternated with 20 s V , I , and unfiltered images.Finally, the sequence converted to 20 s I and unfilteredimages. Because of the physical west hour angle limitof 4.7 hr, KAIT only finished part of this pre-arrangedsequence. In total, 56 images were obtained in the V , I ,and unfiltered passbands from t = 42 to 1628 s, with fullwidth at half-maximum intensity (FWHM) of ∼ ′′ .Visual inspection of the image-processing results re-vealed a true new object, first reported by our group(Li 2007), measured at 12.8 mag at a position of α =20 h m s . δ = +10 ◦ ′ ′′ . σ astrometric uncertainty 0 ′′ .3). Our can-didate OA was subsequently confirmed by observationsfrom the automated Palomar 60-inch (1.5 m) telescope(P60; Cenko & Fox 2007). Further preliminary analysisof OA early-time behavior from the KAIT observationswas reported by Li et al. (2007). Figure 2 shows a se-quence of the KAIT images for the OA of GRB 071003.An 80 ′′ × ′′ section is shown for the first and fifth unfil- \protect http://hercules.berkeley.edu/grbdata/grbfinder.gif . tered 5 s image and a 20 s unfiltered image that startedat t = 431 s. As seen in Figure 2 and reported by severalgroups (Li 2007; Cenko & Fox 2007; Misra et al. 2007;Li et al. 2007), a bright ( R ≈
11 mag) foreground staris located 6 ′′ .5 west of the OA of GRB 071003. As dis-cussed in §
3, the presence of this bright star complicatesthe photometry for the OA, and various methods havebeen used to minimize its contamination.
P60 Observations
The Palomar 60-inch telescope (P60; Cenko et al.2006) automatically responded to the
Swift trigger forGRB 071003, beginning a pre-programmed sequence ofobservations at 07:43:51 UT (176 s after the trigger).Observations were taken in the Kron R , Sloan i ′ and z ′ ,and Gunn g filters at large airmass ( > . V -band filter in relatively short (30 s) exposures. A se-quence of 30 images was obtained. AEOS Observations
The 3.6-m US Air Force Advanced Electro-OpticalSystem (AEOS) telescope, located at the Maui SpaceSurveillance System on Haleakala , observed theOA of GRB 071003 with the AEOS Burst Camera(ABC, Flewelling-Swan et al. 2006). ABC has a back-illuminated 2048 × ′′ .189 pixel − and a FOV of ∼ . ′ × . ′
5. Becausethere is no direct internet access to AEOS, after
Swift detected the GRB, a FAX alert was automatically sentto the AEOS control room, to initiate a series of Target-of-Opportunity (ToO) observations.The AEOS observations of GRB 071003 are all unfil-tered 10 s exposures. The first batch of images started at ∼ t ≈
83 minutes, all with very good imagequality (FWHM ≈ ′′ .9). The second batch of imagesstarted at t ≈
205 minutes, and 56 images were observeduntil t ≈
222 minutes. Due to the large airmass for theseobservations and the degraded seeing conditions, how-ever, the images have rather poor quality. We have triedvarious methods to measure the brightness of the OA inthese images but failed. Accordingly, only the first batchof 238 images is analyzed in this study. Preliminary anal-ysis of the AEOS observations is reported by Swan et al.(2007).
Keck I/Gemini-S observations
In response to the detection of the OA of GRB 071003,we organized a campaign to obtain spectroscopy andlate-time photometry with the 10-m Keck I and the 8-mGemini-S telescopes. At t ≈ . Based on data from the Maui Space Surveillance System,which is operated by Detachment 15 of the U.S. Air Force ResearchLaboratory’s Directed Energy Directorate. t = 42 − 47 s t = 67 − 72 s t = 431 − 451 s NE Fig. 2.—
Sequence of KAIT images for the OA of GRB 071003. An 80 ′′ × ′′ section is shown for the first and fifth unfiltered 5 s imagesand for a 20 s unfiltered image that started at t = 431 s. The OA is the central object in the circles. It is well detected in the early imagesand rapidly fades. The image quality is poor owing to the very high airmass of the object. I, but the data are of poor signal-to-noise ratio (S/N)and no obvious lines were detected (Perley et al. 2007b).Just before the HIRES spectroscopy started, we also ob-tained guider images for the OA, providing importantphotometric coverage during a gap in the photometryobtained elsewhere (see § ′′ .37 pixel − with a FOV of 53 ′′ .5 × ′′ .3.On 2007 October 4, we observed the GRB 071003 OAwith the Low Resolution Imaging Spectrometer (LRIS;Oke et al. (1995)) on Keck I. Anticipating significant fad-ing of the OA, a series of deep 300 s images was takenwith the g and R filters under excellent seeing conditions(FWHM ≈ ′′ .5). Inspection of the images reveals thatthe OA was still bright and saturated in most of the im-ages. Consequently, only a single image in each of the g and R bands, where the OA is not saturated, is analyzedin this study. LRIS uses a beamsplitter to separate thelight between two arms, red and blue. Both the blue andred cameras have a usable FOV of ∼ . ′ × . ′
8. The redcamera used a back-illuminated Tek 2048 × ′′ .215 pixel − , while the blue cam-era has a mosaic of two 2048 × ′′ .135 pixel − .Encouraged by the brightness of the OA, we also per-formed LRIS spectroscopy of the OA. A preliminaryanalysis of the spectrum is reported by Perley et al.(2007a), and a more detailed analysis is presented in § u , g , V , and R filters. Thepresence of the very bright star presents a significantchallenge to extracting useful data on the OA, as itsdiffraction spikes change positions and intensity accord-ing to the time and seeing conditions of the observations.Unfortunately, observations on 2007 October 8 were ad-versely affected by diffraction spikes and poor seeing, andwere not usable. The data taken on 2007 October 15 areseriously affected by clouds, and do not provide an inter-esting limit to the brightness of the OA, so they are notused in this study.We also triggered our TOO program (GS-2007B-Q-2; PI H.-W. Chen) for GRBs with the Gemini-S tele-scope and obtained g -, r -, i -, and z -band images with theGMOS camera on 2007 Oct. 5 and 6. The GMOS camerais equipped with three back-illuminated EEV 2048 × × ′′ .146 pixel − anda FOV of ∼ ′ .5 × ′ .5. Unfortunately, the 2007 October 5 images are badly affected by bleeding from the verybright star and are not used in this study.As part of the efforts to follow the evolution of theOA of GRB 071003, we also performed adaptive op-tics (AO) observations with Keck I on 2007 October 19(Pollack et al. 2007). The details of the AO observationscan be found in § Radio Observations
GRB 071003 was observed with the Very Large Array(VLA) on various occasions. We made the observa-tion in the B configuration array. We used VLA source1950+081 as phase calibrator for 4.86 GHz (C) band ob-servations and 2001+104 for 8.46 GHz (X) band obser-vations. The data were analyzed using standard data re-duction routines of the Astronomical Image ProcessingSystem (AIPS). The first observation took place on 2007October 5 in the X band with flux density of 393 ± µ Jy. Since then we made six observations in the X bandand three observations in the C band (Table 6). DATA REDUCTION
The bright star in the neighborhood of the OA of GRB071003 makes it a challenge to measure reliable photom-etry from the data described in §
2. In this section wedescribe the methods used to minimize its contamina-tion.
Photometric Calibrations
For photometric calibrations, the field of GRB 071003was observed in B , V , R , and I on two photometricnights (2007 October 7 and 8) at Lick Observatory, us-ing both KAIT and the Lick Nickel 1-m telescope. Abouta dozen Landolt standard-star fields (Landolt 1992) wereobserved at different airmasses throughout each photo-metric night. Photometric solutions to the Landolt stan-dard stars yield a scatter of ∼ BV RI images with different depth on both nights.The photometric solutions are used to calibrate a set oflocal standard stars in the GRB 071003 field. Becausethe GRB 071003 field is quite crowded, the number of cal-ibrated local standard stars is large (Table 1). A finderfor a subset of 23 relatively bright local standard stars The NRAO is a facility of the National Science Foundation,operated under cooperative agreement by Associated Universities,Inc.
Fig. 3.—
Finder chart for a subset of local standard stars inthe field of GRB 071003. The field of view is 6 ′ .6 × ′ .6. North isup and east is to the left. The displayed image is the unfilteredtemplate taken with KAIT on 2007 October 9. is in Figure 3. As seen in Table 1, the local standardstars in the field of GRB 071003 are well calibrated, withstandard deviation of the mean (SDOM) of ∼ BV RI bands. We refer to this calibration asthe “Lick calibration” throughout the rest of the paper.Several Landolt standard-star fields were also observedwith LRIS at Keck I: in the u , g , and R bands on 2007October 9, and in the V band on 2007 October 11. As thenumber of the observed standard-star fields is small, itis not possible to derive a complete photometric solutionfor either night. Since the GRB field was observed atsimilar airmasses with some of the standard-star fields,we can treat the LRIS filters as standard and derive themagnitudes for the local standard stars via differentialphotometry. Unfortunately, this procedure suggests thatthe 2007 Oct. 9 night was not photometric, as differentstandard-star observations yield somewhat different zeropoints. The 2007 October 11 night was photometric, butonly the V -band standard stars were observed.We elected to use the Lick calibration as the founda-tion for all the photometric calibrations, except in thecase of the u band. The Lick-calibrated magnitudes arein BV RI , and can be reliably converted to the g , r , and i bands using color transformation equations (Jester et al.2005). The conversion to the z band (Rodgers et al.2006) is somewhat problematic, and as a result we adopta relatively large uncertainty for the converted magni-tudes. For the u band, only two standard-star fieldswere observed with LRIS on 2007 Oct. 9, and they givea difference of 0.30 mag in the zero points. We chose tocalibrate the GRB 071003 field with the standard-starfield that is closer in time of GRB observation, but weadded an uncertainty of 0.30 mag to all the calibratedmagnitudes. We note that the true error for the u -bandcalibration may be higher than 0.30 mag due to the non-photometric conditions on 2007 October 9. KAIT Data Reduction (a) (b)(c) (d)
Fig. 4.—
Illustration of using image subtraction to remove thecontamination of the bright nearby star to the OA of GRB 071003.The KAIT image subtraction code is demonstrated here. ( a ) An80 ′′ × ′′ section of the original 20 s unfiltered KAIT image of theOA taken at t = 431 s; ( b ) the same section after image subtractionof the central 50 ′′ × ′′ using an unfiltered template image after theOA has faded; ( c ) an 80 ′′ × ′′ section of the combined unfilteredAEOS image at t = 5002 . d ) the same section after imagesubtraction of the central 30 ′′ × ′′ using a hand-made templateimage. See text for more details. The KAIT data were automatically processed with biasand dark current subtraction and flat-fielding. The PSFof the OA is seriously affected by the bright star whichis less than 10 pixels away in the KAIT images. Con-sequently, normal PSF-fitting photometry cannot fit thepeak and background of the OA simultaneously to pro-duce a reliable measurement.We use image subtraction to remove the contamina-tion of the bright star. To generate template images forsubtraction, KAIT imaged the GRB 071003 field in theunfiltered mode and in the V and I filters for the nextseveral nights after the burst. To make sure the brightstar is not saturated, short (5 s) exposures were used,and 50–100 images for each filter were acquired to en-sure high S/N in the combined images. As discussed in §
4, the GRB OA was still reasonably bright in the sec-ond night after the burst, so we used the images obtainedat 4–6 days after the burst as the template for the fieldwithout significant OA contribution. Our image subtrac-tion code is based on the ISIS package (Alard & Lupton1998) as modified by B. Schmidt for the High- z Super-nova Search Team (Schmidt et al. 1998). An illustrationof the image subtraction is presented in the top panelsof Figure 4.The Lick calibration was used to transform the KAITinstrumental magnitudes to the standard Johnson V andCousins I passbands, with proper color terms measuredfrom the photometric nights. We also find that the com-bination of the KAIT optics and the quantum efficiencyof the FLI CCD camera makes the KAIT unfiltered ob-servations mostly mimic the R band. During the twophotometric nights, unfiltered observations of the Lan-dolt standard-star fields were also performed. Analysisof these images indicates that the KAIT unfiltered mag-nitudes can be effectively transformed to the R band, t−t trigger (sec)18.016.014.012.0 m a gn it ud e ( R )
10 20 50 100 200 500 1000t−t trigger (sec) 10 F ν ( µ J y ) VRcleariIz P KA I T A E O S BAT
Fig. 5.—
Early-time light curve of the optical afterglow ofGRB 071003 using KAIT photometry, supplemented by observa-tions from P60 and AEOS. The gamma-ray light curve from theBAT is overplotted in gray (scaled arbitrarily). A clearly additive“bump” at 100–500 s is apparent. Photometric follow-up observa-tions continued after 2000 s with P60 and AEOS, as well as withGemini and Keck in subsequent nights; the complete 16-day opticallight curve is presented in Figure 10. with a relatively large color term and an rms of ∼ § AEOS Data Reduction
The ABC images were processed using dark subtrac-tion only. Because of highly variable stray light and vi-gnetting, we did not apply a flat field to these images.We used SExtractor (Bertin & Arnouts 1996) to find allthe sources in the images, from which we were able todetermine the astrometry.We employed the NN2 flux difference method(Barris et al. 2005; hereafter the NN2 method) for con-structing the AEOS light curve. The NN2 method alsouses image subtraction to measure the fluxes for a vari-able source, but it does not designate one particular im-age as the template. Instead, given N total observations,the NN2 method solves for the vector of fluxes from theindividual images using the antisymmetric matrix of fluxdifferences from the N ( N − / R -band data during theoverlap period and assume that the unfiltered AEOSdata have no color term to the R band. The finalAEOS photometry is listed in Table 3. The reported er-ror bars are only those output by the NN2 method, anddo not include a possible large systematic error due tocalibration. If the throughput of the AEOS telescope inthe unfiltered mode is not drastically different from thatof KAIT, we estimate the systematic error to be ∼ t <
20 minutes),and ∼ t > Keck I/Gemini-S Data Reduction
Due to the large aperture of the Keck I and Gemini-S telescopes, the bright star close to the GRB 071003OA produces numerous diffraction spikes, as well as twolarge blooming spikes along the readout direction. Be-cause the orientation, width, and intensity of the spikeschange with the seeing conditions, the exposure duration,and the time of the observations, it is difficult to cleanlyremove them using the template image subtraction orthe NN2 method. However, due to the high resolution ofthese images, the spikes are well sampled and show dis-tinct axial symmetry. We developed a saturation spikesubtraction method, in which we divide the image of thebright star in half, flip the right side, and subtract itfrom the left side. Due to the symmetry in the spikes,this subtraction process leaves a reasonably clean regionaround the GRB OA. PSF-fitting photometry was thenperformed on the GRB OA in the spike-subtracted im-ages, and on a series of local standard stars. The Lickcalibration is used to calibrate the Keck I and Gemini-Sinstrumental magnitudes to the standard system.The final Keck I and Gemini-S photometry is reportedin Table 4. The error bars of the magnitudes are the un-certainties from the PSF-fitting photometry and thosein the calibration process added in quadrature. One spe-cial data point is the Keck I HIRES guider image at t = 9523 . R -bandimages. The measured R -band magnitudes from these We attempted to quantify the color term of the unfilteredAEOS data to the standard R system using the local standardstars in the field of GRB 071003, but found no apparent correlationbetween the scatter of the (unfiltered − R ) differences versus thecolors of the stars. three methods show a scatter of ∼ P60 Data Reduction
The P60 data reduction is presented in this section be-cause it employs several methods (illustrated in Figure 4)discussed earlier in the paper. We obtained template im-ages for the field after the OA of GRB 071003 has faded.However, the saturation spikes of the bright star closeto the GRB ruined the template images in the R and i ′ bands, so we were only able to run image subtractionfor the data in the g and z ′ bands. We also employedthe saturation spike subtraction methods as described in § ′ . × ′ . α = 20 h m s . δ =+10 ◦ ′ ′′ . ′′ . ′′ . Keck AO Data Reduction
On 2007 October 19 (starting at UT 05:14) we observedthe GRB 071003 OA with the NIRC2 (Van Dam et al.2004) narrow-field camera (0 ′′ .01 pixel − ) on Keck II us-ing natural guide star adaptive optics (NGS AO). Whilethe extremely bright nearby star greatly complicated theoptical analysis, it was ideal to be used as the natu-ral guide star during NGS AO imaging. We took 15science exposures, each of 60 s and 2 coadds, resultingin a total integration time of 30 minutes. The imageswere reduced using standard techniques, including darksubtracting, flat fielding, and filtering for deviant pix-els. Each frame was dewarped using the recommendedmethod for NIRC2, and the resulting images were regis-tered to a common origin and combined.The GRB OA is well detected 2 weeks after the burst,as shown in the final combined image in Figure 6. Tomeasure the brightness of the OA, we created a model ofthe PSF using short-exposure, unsaturated images of anearby Two Micron All Sky Survey (2MASS) star ( K s =12 . ± .
024 mag, d = 7 . ′′ ), taken immediately priorto the science exposures. We then subtracted this modelPSF from the OA. With the same 2MASS star as thephotometric calibrator, we measure the OA to have K ′ GalaxyAfterglowPhotometric calibratorNatural Guide StarNE
Fig. 6.—
NGS AO image of the GRB 071003 field taken withKeck II on 2007 October 19, 16 days after the burst. The FOVis approximately 12 ′′ × ′′ . The afterglow is well detected with K ′ = 21 . ± .
03 mag. No host-galaxy emission is detected. = 21.65 ± A K ′ ≈ .
05 mag is negligible along this sightline and has notbeen applied.)
Keck LRIS Spectroscopy Reduction
We obtained low-resolution optical spectroscopy of theoptical afterglow of GRB 071003 on 2007 October 4.335using the LRIS on the Keck I telescope. A pair of 600 sdithered exposures was taken under clear conditions atairmass 1.2 with 0.6 ′′ seeing. We used both the blueand red arms of LRIS, with the light split by the D680dichroic. The 300/5000 grism on the blue side gave aspectral resolution of 8.4 ˚A over the range 3300–6500 ˚A.We used the 600/10000 grating to achieve 4.1 ˚A resolu-tion over the range 6500–8630 ˚A. The spectrophotomet-ric standard star Feige 110 (Stone 1977) was observed thefollowing night in the same setup. Intermittent cloudswere present the night of the standard-star observation,so the absolute flux scale is unreliable.The long, 1.0 ′′ -wide slit was oriented at a position an-gle of 10 ◦ for the afterglow observations, which was notthe parallactic angle (Filippenko 1982). However, theCassegrain Atmospheric Dispersion Compensator mod-ule (Phillips et al. 2006) was mounted, so the derivedspectral shape should be reliable. The exception is inthe spectral range of 6000–6500 ˚A, where second-orderblue light contamination is prominent in the spectrumof the standard star. An attempt was made to cor-rect for the contamination, but the spectral slope inthis section is more uncertain than in the rest of thespectrum. We also fitted an extinction-corrected powerlaw to the flux-calibrated spectrum (excluding line andsecond-order contaminated regions) in an attempt to es-timate the spectral slope, but the estimated slope of f ν ∝ ν − . differs significantly from the spectral slopeestimated from multi-band late-time photometry ( § f λ ( − e r g s / c m / s / Å ) Mg II 2796Mg II 2803 z = 1.604
Fe II 2586Fe II 2599
Fig. 7.—
Spectrum of the GRB 071003 afterglow covering thefull observed spectral range. The spectrum has been flux-calibratedand corrected for Galactic reddening of E ( B − V ) = 0 .
148 mag.The inset shows an expanded view of the region surrounding theFe and Mg absorption system at the burst redshift. A power-lawcontinuum was fitted to the regions of the spectrum shown in green,chosen to avoid strong absorption lines and the wavelength rangecontaminated by second-order blue light. The thick solid blue lineshows the resultant fit ( f λ ∝ λ − . , or f ν ∝ ν − . ), but it differsin slope from our more reliable fit to the broadband photometry;thus, it is used only to normalize the spectrum. spectral index.The largely featureless spectrum (Figure 7) has a S/N > − down to ∼ z GRB < (3500 / − .
88. Numerousmetal-line absorption lines (but no emission lines) arevisible in the spectrum. We have fitted the equivalentwidths of all & σ features in the normalized spectrumusing a Gaussian profile and report the rest-frame valuesin Table 7.We previously presented (Perley et al. 2007a) analy-sis of this spectrum, identifying Mg II absorption sys-tems at z = 0 .
372 and z = 1 . z = 0 . z = 1 .
100 system originated from thehost galaxy (Figure 8). Surprisingly, however, a morethorough investigation revealed a fourth, weak absorp-tion system at a higher redshift of z = 1 .
604 (Figure 9).Contrary to our expectation, the gas at this redshift hasthe weakest Mg II absorption of the four systems.This is remarkable: absorption lines associated withGRB environments are generally very strong with rest-frame equivalent widths exceeding several angstroms(Savaglio et al. 2003; Prochaska et al. 2008a). Figure 8also indicates, however, the presence of fine-structureFe II transitions at this redshift. With the excep-tion of active galactic nucleus environments, these tran-sitions have only been identified in gas surroundingthe GRB phenomenon (Prochaska et al. 2006). Thesetransitions are excited by the GRB afterglow itselfthrough indirect ultraviolet pumping (Prochaska et al.2006; Vreeswijk et al. 2007) of gas in the interstellarmedium (ISM) of the host galaxy. Altogether, the coin- N o r m a li ze d F l ux Galactic:CaII H+Kz=0.372: FeII 2586, 2600; MgII 2796, 2803, MgI 2852 z
GRB =1.604:CIV 1548, 1550 N o r m a li ze d F l ux z=1.100: FeII 2586, 2600; MgII 2796, 2803, MgI 2852 z GRB =1.604:FeII 2344, 2374, 2382FeII* 2396, 2405, 2411, 2414z=0.372: CaII 3934, 3969z=0.937: MgII 2796, 2803; MgI 2852
Fig. 8.—
Portions of the normalized Keck LRIS spectrum of theGRB 071003 afterglow. We mark the positions of several metalabsorption-line features from four distinct extragalactic systemsincluding a series of Fe II and Fe II * transitions associated withthe host galaxy of GRB 071003 ( z GRB = 1 . II doublet marked as Galactic may be due to the very brightGalactic star offset by 6.5 ′′ from GRB 071003 as opposed to theGalactic ISM. cidence of (1) the absence of any higher-redshift absorp-tion systems in our spectrum, (2) the positive detectionof fine-structure Fe II transitions, and (3) the absence ofintergalactic medium absorption at λ > z = 1 .
604 as the redshift of GRB 071003.It might seem unusual to have detected fine-structureFe II transitions in such a late-time spectrum ( t ≈ . z = z GRB . We re-port the positive detections of C IV λ II λλ II λ II λ II λ II ,for example, is fully an order of magnitude below thegeneral population (Cenko et al. 2008), with the sole ex-ception of GRB 070125, and the equivalent width for theC IV gas ( W = 0 . ± .
06 ˚A), represents the lowestmeasurement to date (Prochaska et al. 2008a). RESULTS AND MODELING
Light Curve: General Observations
The multi-color photometric evolution of the GRB071003 OA is shown in Figure 10, fitted by our preferredmodel (described later). Visual inspection of the lightcurves reveals what appear to be three distinct compo-nents: an overall power-law decline that has already setin by the very first measurement at 42 s, a small “bump”feature at ∼ CIV 1548
CIV 1550
AlII 1670 −500 0 5000.80.91.0
FeII 2382
FeII 2586
FeII 2600
MgII 2796 −500 0 500
MgII 2803 N o r m a li ze d F l ux Relative Velocity (km s −1 ) Fig. 9.—
Velocity plot of strong, resonance-line transitions forgas associated with GRB 071003 ( z GRB = 1 . IV absorptionis the weakest yet reported for a GRB afterglow (Prochaska et al.2008a). t − α ) for theclear-band data both before this period and after it, thepower-law indices ( α = 1 .
47 and α = 1 .
49, respectively)are fully consistent with each other and with the overalldecay index over both periods ( α = 1 . t ≈ . R -band limit is reportedat t ≈ > U -banddetection at t ≈ . Optical to Gamma-Ray and X-Ray Comparison
The BAT and XRT light curves we derive forGRB 071003 are also shown in Figure 10. Unfortunately, because
Swift was still in the process of returning to nor-mal operations after its gyro failure (Gehrels 2007), au-tomatic slewing to GRB 071003 was disabled at the timewhen the GRB was detected. As a result, there wereno prompt XRT observations for GRB 071003, leaving along gap in the gamma-ray/X-ray light curve at t = 200–20000 s. In particular, there are no X-ray observationsuntil approximately the peak of the rebrightening in theoptical band. Nevertheless, direct comparison of the dataavailable reveals three relevant facts.First, there is no obvious optical prompt counterpartto the last spike of the gamma-ray light curve. However,this spike is nearly contemporaneous with the much moreslowly rising optical bump feature; we return to this pos-sible connection in our later modeling ( § α X = 1 . ± .
04. Inaddition, the late-time OA behavior (after t ≈ × s) isconsistent with a single power-law decay with an index of α O = 1 . ± .
07, fully consistent with this value. As wenote later, an extrapolation of the X-ray spectral index isalso consistent with the optical observations, suggestingthat at late times there is no need for an additional X-raycontribution (such as inverse Compton) or large amountsof host-galaxy extinction.Finally, while the gamma rays are scaled arbitrarily inFigure 10, we note that if we extrapolate the gamma-ray spectrum into the X-rays to compare the BAT andXRT light curves, the evolution between the end of theprompt emission and the start of the XRT observationsis nearly consistent with a simple extension of the late-time XRT power law back to earlier times, without aneed for a rebrightening or break. However,
Swift hasshown previously (Nousek et al. 2006) that early-time X-ray light curves can conceal a wide variety of complexfeatures, so we will not speculate further as to whetheror not this was actually the case.
Detailed Optical Modeling
The procedure used to model the optical light curveis generally the same as that employed by Perley et al.(2008), but further generalized. For our fit model, weemploy an unbroken power-law decay (component 0) plustwo Beuermann et al. (1999) functions (broken power-law pairs, components 1 and 2), but allow for differentvalues of the functional parameters for each filter andcomponent. The functional form is F ν = F ,ν ( t − dt ) − α + F ,ν (0 . t − dt t p ) − s α ,b + 0 . t − dt t p ) − s α ,a ) − s + F ,ν (0 . t − dt t p ) − s α ,b + 0 . t − dt t p ) − s α ,a ) − s , (1)where for component 0, α is the power-law decay in-dex, and dt is an adjustment to the Swift /BAT triggertime. For cmponent 1, α ,b and α ,a are the power-lawdecay indices for the rising and declining components, re-spectively, dt is an adjustment to the Swift /BAT triggertime, t p is the time of the peak flux, and s is the sharp-ness parameter. Component 2 has a similar function ascomponent 1.0 m a gn it ud e ( R ) F ν ( µ J y , R ) o r F ν , X ( n J y ) ugVRcleariIzK’ P KA I T A E O S K ec k G e m i n i − S XRTBAT α = 1.466 ± α = −1.710 ± α = 5.174 ± α = −1.124 ± α = 1.724 ± ∆β = 0.756 ± ∆β = 1.112 ± ∆β = 0.801 ± χ = 111.1/77t−t trigger (sec)−0.6−0.4−0.2−0.00.20.40.6 r e s i du a l ( m a g ) t−t trigger (sec) 0.61.0 f l ux r a ti o Fig. 10.—
Multi-color, early through late-time light curves of the OA of GRB 071003. The magnitudes are offset according to their early-time colors, showing the color evolution between early and late times. Overplotted colored curves indicate the best-fit three-component,color-evolution model described in the text; the dashed lines represent the individual components that compose this model (a uniformpower-law decay, a chromatic early-time bump, and a monochromatic late-time rebrightening). The X-ray and gamma-ray afterglows arealso overplotted for comparison. The gamma-ray light curve is scaled arbitrarily; if scaled based on the likely gamma-to-X-ray spectralindex it would fall on or near the extrapolation of the X-ray light curve back to early times.
Fitting this function with no constraints generates un-realistic results because of non-uniform sampling in dif-ferent filters. However, we can make the following phys-ically motivated assumptions to tie specific parametersand produce more physically meaningful results.1. We assume that the temporal decay index at anygiven time is independent of the filter, as is implicitin the notation ( α does not depend on ν ). Thismeans that the color of a component cannot changeexcept while the light curve of that component isbreaking.2. Most importantly, we assume that differences be-tween the spectra of the various model componentscan be described by changes in the power-law indexof the intrinsic spectrum, modified by an arbitrary,but fixed, extinction law. Mathematically, this con-straint is expressed as F i,ν = ν ∆ β ij F j,ν . Physi- cally, this assumption requires that external effectssuch as extinction, which might cause the spectrumof any component to deviate from a power law,affect all components equally and are not time-dependent. The extinction law itself (as well asthe absolute underlying index of any specific com-ponent) is fully general and can be fitted accordingto various models later.3. In addition, we assume that the rising segments ofeach component are also power laws, but not neces-sarily the same power laws as the falling segment,to allow for chromatic breaks. This imposes thefollowing condition: ( t p ,xt p ,y ) = ( ν y ν x ) ∆ β ba / ∆ α ba . Here b and a refer to “before” and “after” the break of aspecific component (0, 1, or 2), where the compo-nent index is omitted for clarity, and x and y referto two different filters.1Fitting is performed, under these assumptions, usingthe IDL package mpfit .The assumptions involved in these constraints are, ofcourse, oversimplifications for the full array of modelsthat might be considered. In particular, this model al-lows only one break per component, but with an evolvingsynchrotron light curve plus a jet we may expect as manyas three. However, it has the advantage of being simpleand generates a single physically motivated parameterquantifying color change over each component.We perform a variety of fits under varying combina-tions of assumptions. Some of the possibilities we con-sidered include the following:1. Forcing the bump (component 1) to have the samecolor as the uniform decay (component 0), or al-lowing it to be a different color overall.2. Forcing the bump itself to be achromatic over itsevolution, or allowing it to contain a chromaticbreak.3. Forcing the late rebrightening (component 2) tohave the same color as the uniform decay, or al-lowing it to have a different color.4. Fixing dt for the early steep decay to be zero (theBAT trigger time), or allowing it to be free to vary.5. Fixing dt for the bump component to be zero, tobe equal to the beginning of the prompt-emissionpulse that is nearly contemporaneous with it, orallowing it to be free to vary.6. Fixing dt for the late rebrightening to be zero, orallowing it to be free to vary.The results under various combinations of these as-sumptions are presented in Tables 8 and 9. We discussthe implications of these results in the remainder of thepaper. Color Change
Detection of a GRB afterglow in filtered observationsduring prompt emission, as was the case here, is rare.The situation is even more intriguing since our multi-color prompt OA observations show an apparent bumpfeature (component 1) that is nearly contemporaneouswith a rebrightening pulse in the gamma-ray light curve.Therefore, it is of great interest to attempt to measurethe color of Component 1. By the same token, we havegood spectral coverage of the afterglow both during theprimary normal decay and during the fading of the dra-matic late rebrightening, and any color difference mayshed light on the origin of these features.We tested for color differences in three places: betweencomponent 0 (rapid decay) and component 1 (bump),between component 0 and component 2 (rebrightening),and over the break of component 1 itself (since the ris-ing spectral index may differ from the falling spectralindex). In all cases we find evidence for color variation,although in each case only at the ∼ σ level. The fadingcomponent of the bump is redder than the fading com-ponent of the uniform decay by ∆ β = 0.75 ± \protect http://cow.physics.wisc.edu/ ∼ craigm/idl/idl.html . bump feature is chromatic with a shift from the rising tofalling component of ∆ β = 1.11 ± β = 0.84 ± § Energy Injection Times
It is often unclear what time is most appropriate touse as t when fitting a power law to a GRB afterglow.Thanks to the extremely early-time clear-band data, it ispossible to fit t and constrain this within a few secondsin the case of GRB 071003. This fit, notably, gives a t of exactly the trigger time ( dt = − . ± .
01 s).The gamma-ray light curve (Figure 1) fluence is stronglydominated by the initial pulse, which rises sharply andpeaks within a few seconds, so this is not necessarilysurprising.Some authors (Blake et al. 2005; Vestrand et al. 2005,2006; Yost et al. 2007) have presented evidence of anoptical component rising coincident with the promptemission, although significantly longer lasting. Wecan analyze whether the bump component observed inGRB 071003 may be such a feature by determiningwhether or not it can be fitted with a pulse that risesabruptly, contemporaneous with the prompt emission.While our power-law model is somewhat simplified andthe sampling of the rise is extremely poor, we find thatit generally does not: the best-fit t is intermediate be-tween the trigger time and the time of the prompt emis-sion spike ( ∼
125 s) at dt = 60 ±
20 s. This is a model-independent result, although it rests mostly on one data-point: the initial V -band measurement, representing anintegration from 97 to 117 s after the BAT trigger ( ∼
18 sbefore the rise of the prompt emission spike), lies 0.14mag above a simple power-law extrapolation from re-gions of the data excluding the bump, compared to aphotometric error of only 0.03 mag. While it is possibleto envision scenarios where a relatively slow optical risemight follow a gamma-ray pulse (any broadband featurewith hard-to-soft evolution, or perhaps a late internalshock that later collides with and energizes the exter-nal shock), no model to our knowledge can explain whyan optical flare would precede a gamma-ray pulse, so wetake this as evidence that the two features are physicallyunconnected.While our sampling around the rise and peak of thelate-time rebrightening is poor (and dominated by thedifficult-to-calibrate AEOS and HIRES guider images),we can also attempt to fit the t for the rebrighteningcomponent. This is significantly different from t = 0,with a best-fit initial time of dt = 1245 ±
311 s. (Thisis well short of its peak time of approximately 20 ks, sothe effect on the light curve is minor.) No prompt-likefluctuations or other features are observed in the light2 t−t trigger (sec)1 • • • • • t−t trigger (sec)60100200400 F ν ( µ J y ) V L A α = 0.334 + 0.096 _ χ = 6.1/6 Fig. 11.—
VLA radio light curve fitted to an unbrokenpower law. The uncertainties in the measurements have beenincreased compared to their statistical values to take into ac-count the effect of interstellar scintillation. Some contempo-raneous late-time optical points (scaled arbitrarily) are shownfor comparison. curve in this region.
Radio Modeling
GRB 071003 is rare among
Swift bursts for having abright radio afterglow. We were able to successfully de-tect the afterglow at two frequencies and several epochsspanning ∼ ∼ χ /ν = 15 . /
6. A single, monochro-matic break improves the fit dramatically ( χ /ν =2 . / χ /ν = 1 . / χ /ν ≈
1. Properties of the temporal fitsare given in Table 10.The uncertainty due to scintillation is in any eventtoo large to allow any firm conclusions about the lightcurve. However, since only refractive scintillation is ex-pected to be significant, the refractive timescale is muchlonger than the several-hour timescale of individual ob-servations, and the C-band observations were in all casestaken immediately after the X-band observations, we doconsider the measurement of the radio spectral index( β R = − . ± .
42) to be trustworthy regardless of anyscintillation uncertainty.
Spectral Energy Distribution and ExtragalacticExtinction
If our modeling assumptions are accurate (or nearlyso), we can use our model to calculate the SED at anytime using a combination of all the data available, ratherthan restricting the measurement to a small subset of thephotometry and filters, even if the data were acquired atvery different times in the evolution of the GRB and thecolor is not constant.We calculate the SED at two epochs. First, we calcu-late the SED at t = 2 .
67 days after the burst, the timeof our four-color Gemini-South observations. In calculat-ing this SED, we perform a slightly modified light-curvefit: we do not perform any filter transformations (e.g., toconvert r to R ), but we fix all non-SED parameters tothat derived from the light-curve analysis. In addition,we add in quadrature a calibration uncertainty equal to5% in all filters, with a few exceptions. For z , we usea 15% uncertainty. For u , we use a 30% uncertainty,for reasons described earlier. Finally, for K ′ , we use alarge extra uncertainty of 50% due to the possibility ofa temporal break sometime between our last optical ob-servations and the AO observations. (However, if such abreak is absent, then the K ′ observation is much moreprecise than is given on the plots.) Unfiltered observa-tions are not used. We also calculate an early-time SEDduring the “normal” power-law decay at t = 1000 s, us-ing a fit excluding late-time measurements and measure-ments during the (possibly chromatic) bump. Additionof uncertainties is as for the late-time SED.The resulting SEDs are plotted in Figures 12 and 13.After removing the effects of Galactic extinction (but notyet considering non-Galactic extinction), both SEDs area reasonable fit to a power law, providing a general con-firmation of our assumptions as well as indicating thatthe host or intervening galaxies do not impose a greatdeal of frequency-dependent extinction. In support ofour analysis from the light-curve modeling, the spectralindices appear to differ from early to late times: β s = 0.62 ± β . d = 1.25 ± Swift was unable to slewrapidly. However, this GRB was observed nearly simul-taneously in X-rays, optical, and radio during the de-clining phase of the late rebrightening. Therefore, itis possible to calculate a coeval late-time spectrum atall wavelengths simultaneously. The values at 2.67 days3 λ eff (Å)17.817.617.417.217.016.8 M a gn it ud e ( A B ) λ eff (Å) 20004000 λ eff,rest (Å) 300500 F ν ( µ J y ) β = 0.29 ± 0.49A V = 0.21 (fixed)R V = 2.74 (fixed) χ /dof = 0.71 / 4 β = 0.29A V = 0 t = 1000 sCorrected for Galactic A V = 0.45Dust redshift z = 0.372 gVRiIz Fig. 12.—
Optical SED of the GRB 071003 OA at 1000 s afterthe burst, fitted using the extinction constraints derived using thelate-time SED. The intrinsic (pre-extinction) model spectrum isalso shown. λ eff (Å)22.021.521.020.520.0 M a gn it ud e ( A B ) λ eff (Å) 2000400060008000 λ eff,rest (Å) 102040 F ν ( µ J y ) β = 0.93 ± 0.03A V = 0.21 ± 0.08R V = 2.74 (fixed) χ /dof = 5.95 / 7 β = 0.93A V = 0 t = 2.67 dCorrected for Galactic A V = 0.45Dust redshift z = 0.372 ugVrRiIzK’ Fig. 13.—
Same as Figure 12 but for t = 2 .
67 d after theburst. The data (plus an X-ray normalization, not shown) havebeen fitted with an SMC-like extinction law, with the best-fit curveoverplotted. The intrinsic (pre-extinction) model spectrum is alsoshown. (the same as the first optical-only SED, above, which isalso contemporaneous with XRT observations and withinabout half a day of the first VLA observation) are givenin Table 11 and plotted in Figure 14.Even without considering host-galaxy extinction, theoptical and X-ray observations are nearly consistent witha common spectral index: β O = 1.25 ± β X = 1.14 ± β OX = 0.90 ± α O = 1.72 ± α X = 1.68 ± II is not an exact tracer of the presence ofdust, the extremely weak line absorption at the likelyhost-galaxy redshift of z = 1 .
604 suggests that the dustcolumn at that redshift is nearly negligible. Among theremaining absorbers, the Mg II system at z = 0 .
372 is byfar the strongest (by a factor of ∼ z = 1 . z > F = 0.036 ± µ Jy at2.67 days. This value has already been corrected forphotoelectric absorption ( § A V .Four different extinction models were tested. In ad-dition to a control fit with no extinction, we fit forMilky Way-like, Small Magellanic Cloud (SMC)-like,and Large Magellanic Cloud (LMC)-like extinction usingthe parameterization of the Fitzpatrick & Massa (1990)(“FM”) model, and a model for extinction in starburstgalaxies parameterized by Calzetti et al. (2000). In allcases the standard average value of the ratio of total-to-selective extinction R V in the reference galaxy in ques-tion was used. (Fits with varying R V were attempted,but lacking infrared or ultraviolet measurements we wereunable to constrain this parameter.) We performed sep-arate fits assuming dust at z = 0 . f -test: 96% confidence) for a small amount ( A V = 0.1–0.3 mag, depending on the model) of extinctionalong the light of sight. We cannot strongly constrain itsnature; all four extinction laws, at each of the three pos-sible redshifts, give reasonable fits to the observations.The intrinsic (pre-extinction) spectral slope β is stronglyconstrained to be 0.94 ± β = 1.14 ± § Photometric Limits on a Host Galaxy andIntervening Absorbers
Neither our LRIS imaging nor our late-time NGS AOimaging show any evidence of extension or host-galaxyemission consistent with the afterglow position. Wesearched for emission from a host coincident with the OAposition by smoothing and binning the PSF-subtractedAO image. No host emission was detected to a conser-vative upper limit of K ′ ≈
23 Vega mag.In our first-night LRIS image (when the seeing was bestand contamination from the bright nearby star relativelyminimized), a faint, extended source is visible slightlysouthwest of the OA. The same source is also visible inthe AO image, clearly resolved into a faint galaxy with K ′ ≈
19 mag at an offset of 2.07 ′′ southwest of the OA.4 −5 −4 −3 −2 −1 E (eV)10 −3 −2 −1 F ν ( µ J y ) ν (Hz) t = 2.67 dayXRT β X =1.14 ± 0.12IR/Optical β opt = 1.13 ± 0.06VLA β rad = −1.15 ± 0.42 β = . ± . Fig. 14.—
Broadband SED at t = 2 .
67 days from radiothrough X-ray observations. The shaded region shows an un-broken extrapolation of the X-ray fit (90% confidence region),which is consistent with the optical measurements. The op-tical points are corrected for Galactic but not extragalacticextinction; a best-fit model for the effects of host-galaxy andintervening-galaxy extinction is shown (thin cyan line). Thelocations of the cooling break and peak frequency shown arearbitrarily chosen; the actual frequencies are not constrainedby the available data except that both are located betweenthe radio and optical bands.
We know from the spectral analysis that there are atleast four systems that intersect the sightline betweenthe z = 1 .
604 GRB and Earth, including the host itself.Of these, the strongest candidate for association with theobserved galaxy is clearly the z = 0 .
372 system, whichboth is closest and exhibits the strongest absorption sig-nature. (Unfortunately, we have no spectra of the galaxyto confirm this.) This source appears to be a small ir-regular galaxy, which at this redshift would be offset by ∼
10 kpc (a reasonable distance to explain the observedabsorption) and approximately 0.5 kpc in half-light ra-dius.No other extended sources are detected within 3 ′′ ofthe afterglow, so our upper limit rules out detection ofboth a host galaxy and any absorbing systems within thisdistance. The corresponding limit on a galaxy luminosityis only mild, compared to the known GRB host distribu-tion. At the presumptive GRB redshift of z = 1 .
6, anyhost galaxy is limited to a K -band absolute magnitudeof M ( K ′ ) = − . K -band luminositieson the order of 0.1 L ∗ , and are bluer and fainter thantypical SCUBA galaxies (Le Floc’h et al. 2003). Spectroscopic Constraints on the Host Galaxy andIntervening Absorbers
The very weak absorption at the host redshift in ourspectrum suggests a lower than average H I column den-sity along the sightline and/or a metal-poor gas. Be-cause of our low spectral resolution, however, the ab-sorption is unresolved and the line profiles may be satu-rated (Prochaska 2006). We may conservatively report alower limit to the column densities by assuming the weaklimit. In this manner, we estimate N Mg + > . cm − based on the equivalent width of Mg II λ N HI > cm − .This is a conservative estimate because the gas metallic-ity is presumably subsolar. Nevertheless, it is unlikelythat the gas has an H I column density matching thevalues typical of most GRBs.In addition to the gas associated with GRB 071003,the afterglow spectrum reveals three foreground Mg II absorbers. Two of these have moderate rest-frame equiv-alent widths ( W ≈ . z = 0 . W =2 . II absorption at z < . z ≈ . W > ℓ ( z ) = 0 .
13 at z = 0 .
5, and the incidence of absorbers with W > W > z < . a posteriori statistics, this analysis re-minds one of the apparent enhancement of strong Mg II absorbers along GRB sightlines (Prochter et al. 2006b).Given its low redshift, this system will be an excellentcase to perform follow-up observations and examine theproperties of the galaxies hosting such systems (Pollacket al. 2008, submitted) The bright nearby star, however,poses a formidable obstacle for non-AO ground-based ob-servations. Energetics
The measured gamma-ray fluence of 5.32 ( − × − erg cm − (Konus, 20 keV–4 MeV:Golenetskii et al. 2007) can be converted to an isotropic-equivalent total energy release in the host frame: E iso =3.4 ( − × erg — well in the upper range of Swift events.No clear jet break is observed over the course of ourobservations, in either the optical bands or the X-ray, outto at least 6 × s. There is a possible monochromaticbreak in the radio bands at around 8 days (7 × s), butit appears likely to be a scintillation artifact (see § η = 0 .
5) and circumburst density ( n = 3 . − ) (the end result is nearly insensitive to these pa-rameters), we have θ jet = 6 . ◦ ( t jet d ) / ( n − ) / ( 1 + z − / ( E iso /η erg ) − / . (2)However, as we discuss later, the late-time afterglow be-havior in this case favors a wind model. Thus, followingLi & Chevalier (2003) we have θ jet = 5 . ◦ ( t jet d ) / ( A ∗ ) / ( 1 + z − / ( E iso /η erg ) − / (3)The upper limit on t jet of 7 days gives a limit on theopening of at least 3.1 ( A ∗ / . / deg. (As discussed5later in § A ∗ ≈ . E γ & × ( A ∗ / . / erg.It is also possible that the jet break is hidden by thecomplicated evolution of the burst, including the re-brightening, which would imply more modest energeticsfor this burst. However, as the late-time slope is stillrelatively shallow ( α = 1 .
72; generally we expect α ≥ DISCUSSION
Initial Power-Law Decline
We first turn our attention to the rapidly decliningpower law. The temporal behavior of this feature is quitesimple, with a decay constant α = 1.466 ± β = 0.62 ± β = 0.29 ± α and β (as in, for example, Price et al.2002), all environment models (ISM, wind, and jet) areconsistent with the constraints derived from the data,largely because the early-time constraint on β is poor.(We discuss the forward vs. reverse-shock models for thisemission again in § The Bump: Internal Shock Origin Without aPrompt-Emission Connection
The bump feature is of considerable interest, since it isnearly simultaneous with a prompt-emission pulse. How-ever, as discussed earlier, the temporal analysis seems todisfavor the interpretation as a prompt reverberation:the bump seems to be already rising even before theprompt spike.Another possible explanation for the origin of thisfeature is a large density variation in the surroundingmedium (a large clump or other discrete physical fea-ture in the path of the expanding shock). The observedpulse width ∆ t/t ≈ The Late Rebrightening
The rebrightening phase of this burst is quite dramatic.While our observations do not sample the peak of theemission, a fit with a reasonable assumption of the sharp-ness parameter suggests that the flux increased by ap-proximately 1 mag, and the amount of integrated opticalflux released during the rebrightening is comparable toor more than that emitted by the early afterglow. A risein optical flux of more than a magnitude at intermediatetimes (well after the end of prompt emission, but beforeany supernova component) has to our knowledge beenseen in only a handful of previous cases: GRBs 970508(Castro-Tirado et al. 1998), 041219A (Blake et al. 2005),060729 (Grupe et al. 2007), 070420 (Jel´ınek et al. 2007),and 070311 (Guidorzi et al. 2007).The rebrightening is also notable because it appearsto differ subtly from the early decay, even though theevolution of both curves is generally quite simple. Thedecay index and spectral index both steepen, by ∆ α =0.25 ± β = 0.80 ± ν < ν c before cooling, and ν > ν c after cooling, consistent withthe changes observed), or because of a shift in the elec-tron index p by approximately ∆ p = +0.4.We consider several physical origins for the rebrighten-ing feature: the appearance of the forward shock whenthe burst ejecta first decelerate against the ISM, thelate-time peak of a pre-existing forward shock due toevolution of the critical frequencies, impact of the for-ward shock through a density variation, and rebrighten-ing caused by a refreshed shock. Appearance of forward shock — When the GRB ejectafirst begin to sweep up an amount of matter from theISM comparable to the energy in the ejecta, they beginto decelerate, and reverse and forward shocks are propa-gated back into the ejecta and forward into the ISM, re-spectively; depending on the Lorentz factor, both shockscan then rise very quickly. We consider this scenarioextremely unlikely to be relevant, since by necessity theforward and reverse shocks must rise simultaneously, andthere is no explanation for the bright early-time compo-nent in the burst — save for a prompt model connectedwith internal shocks, but as we have already shown, there6is no evidence linking the early optical behavior with thehigh-energy emission.
Spectral peak of existing forward shock — A more rea-sonable model postulates that the reverse and forwardshocks both formed extremely early, but because theyevolve differently (the reverse shock, whose synchrotronparameters are boosted down by factors of γ , beginsto fade immediately, while the forward shock will riseat lower frequencies), the reverse shock fades rapidly,while the forward shock can rise and peak when the syn-chrotron frequency ν m passes through the optical band.This model has, for example, been invoked to explainearly-time bumps in the light curves of GRB 021004 (e.g.,Kobayashi & Zhang 2003), GRB 050525A (Shao & Dai2005), and GRB 080319B (Bloom et al. 2008), whichlevel off significantly (but do not rebrighten) at around10 s. However, this model is problematic here: althoughwe have only sparse observations of the rebrightening,the observed rising temporal index of α = − ± F ∝ t (2 − s ) / (4 − s ) (= t / for a constant-density ISM and constant for awind). Therefore, the synchrotron peak of the forwardshock alone cannot explain this feature. Density variation — A third possibility, not invokingthe transition between reverse and forward shocks, mightbe a dramatic density variation: for example, the impactof the shock wave into a previously ejected circumstellarshell, or emergence of the shock from a low-density cav-ity into a dense external medium. Density fluctuationshave been successfully invoked to explain low-level varia-tions in several previous studies (e.g., Lazzati et al. 2002)and the timescale of the rebrightening (∆ t/t ≈
1) is con-sistent with a density-fluctuation origin (Nakar & Piran2003). However, in this case we would expect neither achange in the spectral index (as is probably observed)nor such a slow decline after the peak, with a tempo-ral index that differs significantly but only slightly fromthe value of the initial decay. Furthermore, detailed nu-merical studies by numerous authors (Huang et al. 2006;Nakar & Piran 2003; Nakar & Granot 2007) have failedto reproduce anything but the smallest rebrightening sig-natures in previous GRBs using density variations.
Multi-component jet — The complicated light curve ofGRB 030329 has been interpreted (Berger et al. 2003) asthe result of two separate forward shocks, arising fromtwo different jet components: a narrow, highly relativis-tic jet whose emission peaks extremely early, plus a wide,more mildly relativistic jet that dominates the late-timeand radio evolution. Could this model conceivably ex-plain the observations of GRB 071003? While a completeanalysis is beyond the scope of this paper, we note thatthe observations do seem consistent: the similarities oflate-time decay of both rapid and late-time componentsare naturally explained, the timescale of our rebright-ening is similar to that observed in GRB 030329, and(notably) the most significant criticism of the two-jet in-terpretation of GRB 030329 (that the rebrightening rosetoo rapidly and peaked too sharply — Huang et al. 2006)does not apply here: the rebrightening in this case ismuch smoother than that observed for GRB 030329.
Refreshed shock — Finally, we consider the possibilitythat this feature is due to a discrete energy reinjection energizing the forward shock, such as via a slow-movingshell that catches up to the forward shock at late timesafter it decelerates. This seems consistent with all obser-vations, although largely by virtue of not making strongpredictions; by invoking a customized pattern of energyreinjection at the right times, a very broad space of lightcurve behavior can be modeled (Huang et al. 2006). Wedo note that a large, sudden rebrightening of this na-ture may also produce a (second) reverse shock, whichwould be observable in radio and decline rapidly withtime. The radio flux does in fact decay somewhat (incontrast to the expectation from a forward-shock model,where the radio flux is constant or rising), and the mea-sured α = 0 . ± .
10 is not far from the predicted decayconstant for a reverse shock of α ≈ / ν < ν m frequency regime (Kobayashi 2000). However, the radiodecay could conceivably be due to other effects (e.g., latejet break), and without an independent measurement ofthe synchrotron peak frequency ν m and late-time Lorentzfactor Γ we are unable to further constrain the presenceor absence of such a feature with the limited observationsavailable.We therefore find that only the multi-component jetand refreshed shock models are consistent with all avail-able data. Unfortunately, we do not have sufficient ob-servations during the rising phase of the rebrighteningto distinguish the two models; in particular, we can setno constraints on the color evolution and lack a detailedlight curve of the rise to peak of the rebrightening. Wedo note that the X-ray observations are already decayingwell before the (probable) optical peak by an extrapola-tion of our observations (Figure 10), which may suggesthard-to-soft evolution in this feature as well. However,as noted earlier, the X-ray decay extrapolates back tothe BAT light curve without explicit need for a rebright-ening, so without earlier X-ray measurements this asso-ciation is speculative. Environmental Constraints
In the simplest models, the late-time light curve of anyGRB is fixed by a number of basic parameters: micro-physical parameters ǫ B (the fraction of energy in mag-netic fields), ǫ e (the fraction of energy in electrons), and p (the electron energy index); macroscopic parameters E K (the blastwave energy) and θ j (the jet opening an-gle); and a parameter quantifying the density of the sur-rounding medium, n (for a uniform density) or A ∗ (foran r − density profile). Our broadband observations(spanning from radio to X-rays) should, in principle, al-low us to firmly constrain most of these parameters forGRB 071003 — or, more accurately, to its late rebright-ening phase, as this component is dominant at late times.The indices α and β are both well constrained at latetimes in the optical through X-ray bands, thanks to thewide range of temporal and spatial sampling: α O + X =1.71 ± β OX = 0.93 ± ρ ∝ r − ) in which p ≈ .
9, and amodel in which the jet break has already occurred with p ≈ . p = 2). Notably, ISM mod-els are a poor fit: the late-time decay rate is too fastfor the shallow spectral index. The radio observationsappear to support this conclusion: the rising light curvepredicted by the ISM model is clearly ruled out, and7while the slow radio decay ( α R = 0 . ± .
1) is inconsis-tent in detail with the wind prediction of constant evo-lution as well, it is conceivable that variations from anexact s = − t ≈ β ≈ − . β = − ν a and a spectrum of β = − . ∼
90% confidence) tells us that, if the spec-trum is really synchrotron, the absorption break is likelyto be very close to these frequencies, although exact con-straints are difficult with only two frequencies since thebreak is likely to be quite soft. The radio evolution ap-pears nearly achromatic, which would argue against thisinterpretation, but considering the relatively narrow timeand frequency window of the observations and unknownbreak sharpness, we feel that this is not a major concern.Because the ISM model is notably discrepant with themeasured values of α and β , we unfortunately cannot usethe afterglow as a probe of the ambient density. If thewind model, which is more consistent with the observa-tions in this case, is correct, we can calculate the param-eter A ∗ using (for example) equation 2 in Chevalier & Li(1999): F ν m = 20 mJy( d L z ) / ( ǫ B . − / E / A ∗ t / . (4)While we have no direct measurement of F ν m , it is con-strained by the radio and optical observations (see Figure14) to be ∼ ∼ A ∗ = 0 . ǫ B / . / , an interestingly low valueregardless of the value of ǫ B . While ǫ B is not stronglyconstrained, the absence of a cooling break between theX-ray and optical bands during the first 5 days (thecooling frequency ν c increases in a wind model) requires ǫ B & . ν > ν c late it must have been early as well under stan-dard synchrotron evolution. However, because the re-brightening appears to be either a separate phenomenonor a large energy impulse that could conceivably have“reset” the synchrotron parameters [including ν c ] to newvalues, this may not be a major concern.) No jet breakis observed in the light curve, but it is possible that a jetsignature was concealed by the rebrightening. This casewould certainly rule out the wide-angle jet interpretationof the secondary peak and would significantly reduce theenergetics. Spectral Implications on the Environment and HostGalaxy
The late-time spectroscopy and imaging tell acoherent story: unlike the vast majority of GRBs (Wainwright et al. 2007; Prochaska et al. 2008a),GRB 071003 did not occur in a gas-rich galaxy. Theenvironment is more consistent with a progenitor locatedin an outer galactic halo, or in an extremely small (evencompared to “normal” long-duration GRB hosts) andgas-poor galaxy. While the possibility of line saturationprevents us from setting definitive upper limits, thecolumn density through any host is consistent withbeing 3 orders of magnitude below typical GRB-derivedvalues, and the contrast to the overall GRB population- which is dominated by subluminous galaxies to beginwith (e.g., Fruchter et al. 2006; Fynbo et al. 2008), isdramatic.While it is well established that long-duration GRBsgenerally originate from massive stars, we should be care-ful to ensure that our prior experience does not blind usto the existence of rarer subclasses of events. We notethat one other GRB on record, GRB 070125, had verysimilar properties: extremely low Mg II absorption andno coincident host (Cenko et al. 2008), as well as a verybright afterglow and extreme energetics ( E γ = 3 × erg; Chandra et al. 2008), and even a (mild) late-timerebrightening (Updike et al. 2008). Both are also amongthe few Swift bursts detected at radio wavelengths.However, GRB 070125 and GRB 071003 show evidencefrom their broadband light curves of origins typical ofordinary long GRBs. In the case of GRB 070125, a con-stant but very high circumstellar density suggested thatit occurred in what was locally a dense environment, notan empty galactic halo, despite the near absence of alarge-scale gas signature in the spectrum. In our case,for GRB 071003, we find evidence of a wind-like strati-fied environment, a characteristic of a massive star. To-gether, these events appear to suggest an origin for these“halo” bursts similar to those of all other GRBs.If GRB 071003 did occur in a star-forming region,then there are two possibilities consistent with the ex-tremely small metal absorption in the spectrum. First,the burst may simply have formed in an extremely sublu-minous galaxy — necessarily, the number or distributionof such objects at very high redshift is not observation-ally constrained, but most simulations predict an abun-dance of small, highly sub-Galactic halos in the universethat could very well harbor limited star formation. Al-ternatively, GRB 071003 may have occurred in a tidallystripped tail from another, larger galaxy. In this case,further follow-up observations should reveal a disturbed,star-forming host in the close vicinity of the burst.Either scenario seems plausible to explain the con-straints derived on the burst environment. In either case,if GRBs are shown to be reasonable tracers of star for-mation at high redshift, then future large-sample GRBspectroscopy missions may be able to place importantconstraints on the star-formation history of the universenot possible by any other means. While the samplesize of such low-column-density GRBs is now small (two Since our measurement is based on magnesium, we are directlymeasuring the metal column, not the gas column. An alternatepossibility, therefore, is that the host is “normal” but extraordi-narily metal-poor, less than 10 − of the average solar abudance.However, we consider a highly subluminous host a more likely pos-sibility. Both effects may be in play: low-luminosity galaxies, andthose with low equivalent widths, tend to be relatively metal-poor(Prochaska et al. 2008b). II equivalent widths turns out to be bimodal.On a related note, the existence of GRB 071003 andGRB 070125 may have important implications regardingthe escape fraction of ionizing photons and the reioniza-tion history of the universe. Although the relatively lowredshift of these systems keeps the Lyman- α and Lyman-break absorption features out of our spectral range andprevents us from measuring the H I column density di-rectly (Chen et al. 2007), these GRBs provide evidencethat massive stars can form well outside of gas-densehosts, where there is little to shield the intergalacticmedium from their ionizing UV radiation. If the frac-tion of these events is more than a few percent at z > x H (e.g., McQuinn et al. 2007) without the interferenceof saturated line profiles originating from the host galaxy. CONCLUSIONS
Although the temporal evolution of the optical after-glow of GRB 071003 is complicated, our early throughlate-time photometric follow-up data clearly resolve theoptical light curve into separate components. Observa-tions from KAIT during the prompt phase of the GRBrevealed a slowly rising, slowly falling bump or flare com-ponent, superimposed on a simple fading power law thathas no observable correlation with the prompt emission,suggesting that while early internal-shock flares can beobserved in the optical, they are not necessarily the sameas those producing the high-energy signatures. Our late-time observations revealed one of the most dramatic laterebrightenings ever recorded in a GRB light curve, andsuggest that this feature is not due to a reverse-forwardshock transition or density variation, requiring either an-gular jet structure or very discrete late-time re-energizingof the optical afterglow. This may have important im-plications for the interpretation of other, less dramaticbumps and rebrightenings at similar timescales that ap-pear to be common features in GRB afterglows.The spectroscopic study of GRB 071003 offers a cau-tionary tale about the standard use of Mg II to infer aredshift: while it is common practice to use the highest-redshift Mg II system observed (especially in the caseswhen the absorption is quite strong) under the assump-tion that the GRB host system should show significantmetal absorption, here we have a clear case where thisassumption is fundamentally flawed. Were the S/N ofthe spectrum worse, or the host-galaxy absorption even weaker by a factor of only 2–3, it is likely that we wouldhave missed the higher-redshift system entirely and pro-ceeded with the assumption that this burst was at a red-shift of 1.100 instead of 1.604. In light of this fact, previ-ous and future GRB redshift claims based solely on iden-tification of Mg II absorption should be regarded withincreased skepticism.The intervening absorption systems are neverthelessalso remarkable. With three completely independentMg II systems along the line of sight, GRB 071003 isamong the most dramatic examples yet of the bizarreoverabundance of these systems in GRB afterglows rela-tive to those of quasars. Further study of this sightline,especially using AO systems, may help shed light on thismysterious result.KAIT and its ongoing operation were made possibleby donations from Sun Microsystems, Inc., the Hewlett-Packard Company, AutoScope Corporation, Lick Obser-vatory, the National Science Foundation, the Universityof California, the Sylvia & Jim Katzman Foundation, andthe TABASGO Foundation. J.S.B.’s group is supportedin part by the Hellman Faculty Fund, Las Cumbres Ob-servatory Global Telescope Network, and NASA/ Swift
Guest Investigator grant NNG05GF55G. A.V.F.’s groupis supported by NSF grant AST–0607485 and theTABASGO Foundation, as well as by NASA/
Swift
GuestInvestigator grants NNG05GF35G and NNG06GI86G.N.R.B. is partially supported by a SciDAC grant fromthe Department of Energy. J.X.P. is partially supportedby NASA/
Swift
Guest Investigator grant NNG05GF55Gand NSF CAREER grant AST–0548180. H. Swan hasbeen supported by NSF grant AST-0335588 and bythe Michigan Space Grant Consortium. F. Yuan hasbeen supported under NASA/Swift Guest InvestigatorGrant NNX-07AF02G. We acknowledge helpful discus-sions with E. Ramirez-Ruiz and thank D. Whalen andA. Heger for their calculations of the photon flux frommassive stars. We offer particular thanks to D. A. Kannfor useful discussions and feedback.This research is based in part on observations obtainedat the Gemini Observatory, which is operated by the As-sociation of Universities for Research in Astronomy, Inc.,under a cooperative agreement with the NSF on behalfof the Gemini partnership. Some of the data presentedherein were obtained at the W. M. Keck Observatory,which is operated as a scientific partnership among theCalifornia Institute of Technology, the University of Cal-ifornia, and NASA; the Observatory was made possibleby the generous financial support of the W. M. KeckFoundation. We wish to extend special thanks to thoseof Hawaiian ancestry on whose sacred mountain we areprivileged to be guests. We are grateful to the staffs atthe Gemini, Keck, and Lick Observatories for their assis-tance.
REFERENCESAkerlof, C., et al. 1999, Nature, 398, 400Alard, C., & Lupton, R. H. 1998, ApJ, 503, 325Barris, B. J., Tonry, J. L., Novicki, M. C., & Wood-Vasey, W. M.2005, AJ, 130, 2272Berger, E., et al. 2003, Nature, 426, 154 Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393Beuermann, K., et al. 1999, A&A, 352, L26Blake, C. H., et al. 2005, Nature, 435, 181Bloom, J. S., Kulkarni, S. R., & Djorgovski, S. G. 2002, AJ, 123,1111 Bloom, J. S., et al. 2008, ApJin press. (ArXiv e-prints 0803.3215)Butler, N. R. & Kocevski, D. 2007a, ApJ, 663, 407—. 2007b, ApJ, 668, 400Butler, N. R., et al. 2006, ApJ, 652, 1390Calzetti, D., Armus, L., Bohlin, R. C., Kinney, A. L., Koornneef,J., & Storchi-Bergmann, T. 2000, ApJ, 533, 682Castro-Tirado, A. J., et al. 1998, Science, 279, 1011Cenko, S. B., & Fox, D. B. 2007, GCN Circular 6839Cenko, S. B., et al. 2006, PASP, 118, 1396Cenko, S. B., et al. 2008, ApJ, 677, 441Chandra, P., et al. 2008, ArXiv e-prints, 0802.2748Chen, H.-W., Prochaska, J. X., & Gnedin, N. Y. 2007, ApJ, 667,L125Chevalier, R. A. & Li, Z.-Y. 1999, ApJ, 520, L29Chincarini, G., et al. 2007, ArXiv Astrophysics e-prints, 0702371D’Elia, V., et al. 2007, A&A, 467, 629D’Elia, V., et al. 2008, ArXiv e-prints, 0804.2141Dessauges-Zavadsky, M., Chen, H.-W., Prochaska, J. X., Bloom,J. S., & Barth, A. J. 2006, ApJ, 648, L89Dickey, J. M. & Lockman, F. J. 1990, ARA&A, 28, 215Fenimore, E. E., in ’t Zand, J. J. M., Norris, J. P., Bonnell, J. T.,& Nemiroff, R. J. 1995, ApJ, 448, L101Filippenko, A. V. 1982, PASP, 94, 715Filippenko, A. V. 2005, in Astronomical Society of the PacificConference Series, Vol. 332, The Fate of the Most Massive Stars,ed. R. Humphreys & K. Stanek, 33Filippenko, A. V., Li, W. D., Treffers, R. R., & Modjaz, M. 2001, inAstronomical Society of the Pacific Conference Series, Vol. 246,IAU Colloq. 183: Small Telescope Astronomy on Global Scales,ed. B. Paczy´nski, W.-P. Chen, & C. Lemme, 121Fitzpatrick, E. L. & Massa, D. 1990, ApJS, 72, 163Flewelling-Swan, H., et al., & Moore, K. 2006, in Bulletin ofthe American Astronomical Society, Vol. 38, Bulletin of theAmerican Astronomical Society, 134Fruchter, A. S., et al., & Woosley, S. E. 2006, Nature, 441, 463Fugazza, D., Fiore, F., D’Elia, V., D’Avanzo, P., Piranomonte, S.,Antonelli, L. A., Chincarini, G., Covino, S., Tagliaferri, G., Valle,M. D., & Soto, A. F. 2007, GCN Circular 6851Fynbo, J. P. U., et al. 2008, ArXiv e-prints, 0801.3273Garnavich, P. M., Loeb, A., & Stanek, K. Z. 2000, ApJ, 544, L11Gehrels, N. 2007, GCN Circular 6760Gehrels, N., et al. 2004, ApJ, 611, 1005Golenetskii, S., Aptekar, R., Mazets, E., Pal’shin, V., Frederiks,D., & Cline, T. 2007, GCN Circular 6849Granot, J., & Kumar, P. 2006, MNRAS, 366, L13Grupe, D., et al. 2007, ApJ, 662, 443Guidorzi, C., et al. 2007, A&A, 474, 793Holland, S. T., et al. 2003, AJ, 125, 2291Huang, Y. F., Cheng, K. S., & Gao, T. T. 2006, ApJ, 637, 873Jel´ınek, M., Tristram, P., Postigo, A. d. U., Gorosabel, J., Hale, F.,Gilmore, A., Kilmartin, P., & Castro-Tirado, A. J. 2007, GCNCircular 6338Jester, S., et al. 2005, AJ, 130, 873Kobayashi, S. 2000, ApJ, 545, 807Kobayashi, S., & Zhang, B. 2003, ApJ, 582, L75Kocevski, D., Butler, N., & Bloom, J. S. 2007, ApJ, 667, 1024Laursen, L. T. & Stanek, K. Z. 2003, ApJ, 597, L107Lazzati, D., Rossi, E., Covino, S., Ghisellini, G., & Malesani, D.2002, A&A, 396, L5Le Floc’h, E., et al. 2003, A&A, 400, 499Li, W. 2006, GCN Circular 4727—. 2007, GCN Circular 6838Li, W., Chornock, R., Filippenko, A. V., Perley, D. A., & Bloom,J. S. 2007, GCN Circular 6844Li, W., Filippenko, A. V., Chornock, R., & Jha, S. 2003a, ApJ,586, L9—. 2003b, PASP, 115, 844Li, W. D., et al. 2000, in American Institute of Physics ConferenceSeries, Vol. 522, American Institute of Physics Conference Series,ed. S. S. Holt & W. W. Zhang, 103–106Li, Z.-Y. & Chevalier, R. A. 2003, in Lecture Notes in Physics,Berlin Springer Verlag, Vol. 598, Supernovae and Gamma-RayBursters, ed. K. Weiler, 419–444McQuinn, M., Lidz, A., Zaldarriaga, M., Hernquist, L., & Dutta,S. 2007, ArXiv e-prints, 0710.1018Misra, K., Pandey, S. B., Roy, R., & Castro-Tirado, A. J. 2007,GCN Circular 6840 Monet, D. G., et al., Schoening, W., Hartley, M., Read, M. A., &Tritton, S. B. 2003, AJ, 125, 984Nakar, E., & Granot, J. 2007, MNRAS, 380, 1744Nakar, E., & Piran, T. 2003, ApJ, 598, 400Nestor, D. B., Turnshek, D. A., & Rao, S. M. 2005, ApJ, 628, 637Nousek, J. A., et al. 2006, ApJ, 642, 389Nysewander, M., Reichart, D. E., Crain, J. A., Foster, A., Haislip,J., Ivarsen, K., Lacluyze, A., & Trotter, A. 2007, ArXiv e-prints,0708.3444Oke, J. B., Cohen, J. G., Carr, M., Cromer, J., Dingizian, A.,Harris, F. H., Labrecque, S., Lucinio, R., Schaal, W., Epps, H.,& Miller, J. 1995, PASP, 107, 375Perley, D. A., et al. 2008, ApJ, 672, 449Perley, D. A., Chornock, R., Bloom, J. S., Fassnacht, C., & Auger,M. W. 2007a, GCN Circular 6850Perley, D. A., Prochaska, J. X., Bloom, J. S., & Smith, G. 2007b,GCN Circular 6843Phillips, A. C., Miller, J., Cowley, D., & Wallace, V. 2006,in Presented at the Society of Photo-Optical InstrumentationEngineers (SPIE) Conference, Vol. 6269, Ground-based andAirborne Instrumentation for Astronomy. Edited by McLean,Ian S.; Iye, Masanori. Proceedings of the SPIE, Volume 6269,pp. 62691O (2006).Pollack, L. K., Dam, M. A. v., Mignant, D. L., Johansson, E. M.,Perley, D. A., & Bloom, J. S. 2007, GCN Circular 7010Price, P. A., et al. 2002, ApJ, 572, L51Prochaska, J. X. 2006, ApJ, 650, 272Prochaska, J. X., Chen, H.-W., & Bloom, J. S. 2006, ApJ, 648, 95Prochaska, J. X., Chen, H.-W., Dessauges-Zavadsky, M., & Bloom,J. S. 2007, ApJ, 666, 267Prochaska, J. X., Chen, H.-W., Wolfe, A. M., Dessauges-Zavadsky,M., & Bloom, J. S. 2008a, ApJ, 672, 59—. 2008b, ApJ, 672, 59Prochter, G. E., Prochaska, J. X., & Burles, S. M. 2006a, ApJ, 639,766Prochter, G. E., et al. 2006b, ApJ, 648, L93Ricker, G. R., et al. 2003, in American Institute of PhysicsConference Series, Vol. 662, Gamma-Ray Burst and AfterglowAstronomy 2001: A Workshop Celebrating the First Year of theHETE Mission, ed. G. R. Ricker & R. K. Vanderspek, 3–16Rodgers, C. T., Canterna, R., Smith, J. A., Pierce, M. J., & Tucker,D. L. 2006, AJ, 132, 989Rykoff, E. S., et al. 2006, ApJ, 638, L5Sari, R., Piran, T., & Halpern, J. P. 1999, ApJ, 519, L17Savaglio, S., Fall, S. M., & Fiore, F. 2003, ApJ, 585, 638Schady, P., Evans, P. A., Gehrels, N., Guidorzi, C., Holland, S. T.,Kennea, J. A., Markwardt, C. B., O’Brien, P. T., Page, K. L.,Palmer, D. M., Sato, G., & Starling, R. L. C. 2007, GCN Circular6837Schmidt, B. P., et al. 1998, ApJ, 507, 46Shao, L., & Dai, Z. G. 2005, ApJ, 633, 1027Shih, C. Y., Lee, Y. H., Urata, Y., & Huang, K. Y. 2007, GCNCircular 6846Stone, R. P. S. 1977, ApJ, 218, 767Swan, H., Smith, I., Rujopakarn, W., Yuan, F., Yost, S., Akerlof,C., Skinner, M., & Russell, R. 2007, GCN Circular 6841Thoene, C. C., Fynbo, J. P. U., & Jakobsson, P. 2006, GCNCircular 5747Updike, A. C., et al. 2008, ArXiv e-prints, 0805.1094Van Dam, M. A., Le Mignant, D., & Macintosh, B. A. 2004,Applied Optics, 43, 5458Vestrand, W. T., et al. 2005, Nature, 435, 178Vestrand, W. T., et al., & Fenimore, E. 2006, Nature, 442, 172Vreeswijk, P. M., et al. 2007, A&A, 468, 83Wainwright, C., Berger, E., & Penprase, B. E. 2007, ApJ, 657, 367Walker, M. A. 1998, MNRAS, 294, 307—. 2001, MNRAS, 321, 176Wo´zniak, P. R., Vestrand, W. T., Wren, J. A., White, R. R., Evans,S. M., & Casperson, D. 2006, ApJ, 642, L99Yost, S. A., Harrison, F. A., Sari, R., & Frail, D. A. 2003, ApJ,597, 459Yost, S. A., et al. 2007, ApJ, 657, 925Zhang, B., et al. 2006, ApJ, 642, 354 TABLE 1Photometry of comparison stars in the field of GRB 071003 ID α J δ J B N B V N V R N R I N I Note . — Uncertainties (standard deviation of the mean) are indicated in parentheses. This table hasbeen truncated; additional standard stars are available in the online material. a In degrees.
TABLE 2KAIT photometry of GRB 071003 t starta Exp. time Mag Error Filter(s) (s)42.0 5 12.791 0.019 R b R R R R V I R V I R V I R V I R V I R I R I R c ×
20 16.830 0.113 I c ×
20 17.362 0.121 R c ×
20 17.314 0.148 I c ×
20 17.711 0.147 R c ×
20 18.103 0.154 R c ×
20 17.473 0.135 I a The start time of the exposure, in seconds afterthe BAT trigger. b The R-band photometry is derived from unfil-tered observations. c The time (s) at the middle point of severalcombined images. TABLE 3AEOS R -bandphotometry fromunfilteredobservations a t midb R σ R (s)568.6 16.708 0.016681.3 17.046 0.018794.0 17.337 0.020906.8 17.573 0.0201019.6 17.766 0.0201132.3 17.940 0.0201245.1 18.101 0.0201357.9 18.229 0.0231470.7 18.339 0.0251583.5 18.454 0.0261696.3 18.545 0.0281809.1 18.640 0.0341922.0 18.724 0.0332034.8 18.814 0.0402147.6 18.865 0.0372260.3 18.924 0.0422373.1 18.941 0.0482485.9 18.941 0.0642598.7 19.042 0.0442711.4 19.087 0.0492824.2 19.104 0.0502937.0 19.109 0.0543049.7 19.142 0.0543162.5 19.150 0.0523275.3 19.160 0.0563388.2 19.169 0.0553501.0 19.158 0.0503613.8 19.139 0.0563726.5 19.185 0.0553839.3 19.206 0.0564062.4 19.188 0.0564175.3 19.196 0.0554288.1 19.194 0.0574401.0 19.208 0.0514513.8 19.167 0.0544626.6 19.150 0.0534739.4 19.163 0.0654852.2 19.160 0.0575002.6 19.145 0.050 a The original data setwas grouped and com-bined into 39 images. b The time at the middlepoint of several combinedimages.
TABLE 4Keck/Gemini-S photometry of GRB 071003 t mida Exp. time Mag Error Filter Telescope(s) (s)9523.7 2 ×
20 18.59 b C R Keck I76891.8 300 20.32 0.07 g Keck I77044.0 300 19.43 0.06 R Keck I231174.7 450 22.33 0.20 g Gemini-S231802.3 450 21.57 0.32 i Gemini-S232430.8 450 21.97 0.22 r Gemini-S233056.5 450 21.35 0.40 z Gemini-S515855.0 1485 22.61 0.30 R Keck I516250.5 975 23.56 0.30 g Keck I517510.4 720 23.56 0.45 u Keck I604978.0 660 23.06 0.50 R Keck I605144.4 780 24.05 0.40 g Keck I682211.1 330 23.56 0.40 V Keck I682211.9 660 24.42 0.50 u Keck I682940.0 840 23.40 0.60 R Keck I1373589. 1800 21.58 0.03 K ′ Keck II a The time (s) at the middle point of the observations. b Measured from unfiltered images from the Keck I HIRESguider. TABLE 5P60 photometry of GRB 071003 t starta Exp. time Mag Error Filter(s) (s)176.0 60 14.57 0.06 R i ′ z ′ R i ′ z ′ R i ′ z ′ z ′ g z ′ g g a The start time of the exposure, in secondsafter the BAT trigger.
TABLE 6Radio observations of GRB 071003
UT Date t mid Frequency Flux density ErrorObservation hr GHz µ Jy µ Jy2007 Oct. 05, 1.85 42.168 8.46 393 552007 Oct. 07, 3.38 91.698 8.46 430 502007 Oct. 07, 3.92 92.238 4.86 220 542007 Oct. 12, 1.03 209.248 8.46 431 512007 Oct. 14, 14.84 271.158 8.46 332 672007 Oct. 24, 23.58 519.898 8.46 260 422007 Oct. 25, 0.04 520.358 4.86 119 462007 Nov. 05, 0.01 785.328 8.46 109 452007 Nov. 07, 0.18 833.336 4.86 93 52
TABLE 7Absorption Lines in the Afterglow Spectrum ofGRB 071003 λ z
Transition W a σ ( W ) b (˚A) (˚A) (˚A)3549.69 0.37223 FeII 2586 < . .
33 0.593837.72 0.37223 Mg II .
48 0.203847.65 0.37223 Mg II .
14 0.193915.45 0.37223 MgI 2852 1 .
02 0.174032.63 1.60435 CIV 1548 0 .
22 0.064039.88 1.60435 CIV 1550 < . < . .
20 0.055276.54 1.60435 ZnII 2026 < . .
61 0.075417.99 0.93740 Mg II .
61 0.055432.79 c .
46 0.055447.85 0.37223 CaII 3969 0 .
46 0.075872.31 1.10019 Mg II .
80 0.055888.27 1.10019 Mg II .
68 0.066105.90 1.60435 FeII 2344 < . .
26 0.046240.46 1.60435 FeII* 2396a 0 .
25 0.046265.95 1.60435 FeII* 2405 < . .
18 0.036284.57 0 .
72 0.126734.47 0 .
97 0.156737.28 1.60435 FeII 2586 0 .
16 0.046772.60 1.60435 FeII 2600 0 .
27 0.057301.58 1.60435 Mg II .
17 0.057430.06 1.60435 MgI 2852 < . .
92 0.138436.10 d .
86 0.268534.91 d .
72 0.178599.02 d .
34 0.17 a Equivalent widths are rest-frame values and assumethe redshift given in Column 2. b Limits are 2 σ statistical values. c Blended with Mg II λ z = 0 . d These features may be residuals from sky subtrac-tion. TABLE 8Optical Light-Curve Fits: Color Change
Model Description ∆ β − ∆ β ( b − a ) ∆ β − β χ /ν Fully monochromatic 0 0 0 0.72 ± ± ± ± ± ± ± ± ± ± ± ± Note . — Summary of relevant parameters and χ for models allowing or disallowing color transi-tions and chromatic breaks between the various components. Values without uncertainties are fixed.Component 0 is the fast-decay component, Component 1 is the bump, and Component 2 is the laterebrightening. The absolute late-time spectral index β is not a model parameter, but is fit externallyafter completion of the fit. TABLE 9Optical Light-Curve Fits: t Model Description dt dt dt χ /ν (s) (s) (s)Reference 0 0 0 113.713 / 78Decay − . ± .
01 0 0 113.713 / 77Bump 0 60.5 ± ±
311 111.149 / 77
Note . — Summary of relevant parameters and χ for models using a t different from the trigger time. In all cases, the favored color-change model(chromatic bump and rebrightening) was used. Values without uncertainties arefixed. Component 0 is the fast-decay component, Component 1 is the bump,and Component 2 is the late rebrightening. TABLE 10Radio Modeling of GRB 071003
Parameter Value (broken power law) Value (unbroken) Value (unbroken w/scintillation a ) α b − ± ± ± α a ± t break ± β − ± − ± − ± χ /ν Note . — Best-fit parameters of a fit to the radio afterglow of GRB 071003 using aBeuermann et al. (1999) broken power-law model versus an unbroken power-law model. The im-provement for the broken power-law fit is significant given the flux uncertainties, but due to inter-stellar scintillation may be coincidental. If a small amount of interstellar scintillation uncertaintyis added in quadrature, an unbroken power-law fit is reasonable. a In this model, we added a 15% error to all X-band points and a 22% error to all C-band points.
TABLE 11Model fluxes at t = 2 . days Band/Filter E Flux UncertaintyeV µ Jy µ JyX-ray 1000 0.036 0.006 u g V r R i I z K ’ 0.584 33.59 16.8X 3.5e-5 414.6 91.8C 2.0e-5 256.1 73.9 Note . — Fluxes of the afterglow interpo-lated to t = 2 .
67 d after the BAT trigger us-ing all available X-ray, optical, and radio data.Galactic extinction ( E ( B − V ) = 0 .
148 mag)is not accounted for; however, the X-ray fluxis corrected for photoelectric absorption. TABLE 12Extinction models for optical/X-ray fits of GRB071003 model A V R V β χ /ν none 0 - 0.913 ± z = 0 . ± ± ± ± ± ± ± ± z = 1 . ± ± ± ± ± ± ± ± z = 1 . ± ± ± ± ± ± ± ± Note . — Results of various fits to the contemporaneousoptical and X-ray fluxes for extinction due to either the hostgalaxy or the intervening absorbers at z = 0 .
372 and z = 1 . z = 0 ..