Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.mso.anu.edu.au/~kcf/pubs_top20/2.pdf Äàòà èçìåíåíèÿ: Sat Oct 15 11:29:49 2011 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 00:53:26 2012 Êîäèðîâêà: Windows-1251

A PHOTOMETRIC REDSHIFT OF z 9.4 FOR GRB 090429B
A. Cucchiara1,2,3 , A. J. Levan4 , D. B. Fox1 , N. R. Tanvir5 , T. N. Ukwatta6,7 , E. Berger8 , T. Kruhler9,10 , A. Kup cu Yoldaå11,12 , X. F. Wu1,13 , K. Toma1 , J. Greiner9 , F. Olivares E.9 , Å ÅÅ s A. Rowlinson5 , L. Amati14 , T. Sakamoto7 , K. Roth15 , A. Stephens15 , A. Fritz15 , J.P.U. Fynb o16 , J. Hjorth16 , D. Malesani16 , P. Jakobsson17 , K. Wiersema5 , P. T. O'Brien5 , A. M. Soderb erg8 , R. J. Foley8 , A. S. Fruchter18 , J. Rhoads19 , R. E. Rutledge20 , B. P. Schmidt21 , M. A. Dopita21 , P. Podsiadlowski22 , R. Willingale5 , C. Wolf22 , S. R. Kulkarni23 , AND P. D'Avanzo24 acucchiara@lbl.gov

arXiv:1105.4915v3 [astro-ph.CO] 31 May 2011


-2- ABSTRACT Gamma-ray bursts (GRBs) serve as p owerful prob es of the early Universe, with their luminous afterglows revealing the locations and physical prop erties of star forming galaxies at the highest redshifts, and p otentially locating first generation (Population I I I) stars. Since GRB afterglows have intrinsically very simple sp ectra, they allow robust redshifts from low signal to noise sp ectroscopy, or photometry. Here we present a photometric redshift of z 9.4 for the Swift detected GRB 090429B based on
Department of Astronomy & Astrophysics, 525 Davey Lab oratory, Pennsylvania State University, University Park, PA 16802, USA
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 1

Lawrence Berkeley National Lab oratory, M.S. 50-F, 1 Cyclotron Road, Berkeley, CA 94720, USA Department of Astronomy, 601 Campb ell Hall, University of California, Berkeley, CA 94720-3411, USA Department of Physics, University of Warwick, Coventry, CV4 7AL, UK Department of Physics and Astronomy, University of Leicester, University Road, Leicester, LE1 7RH, UK Department of Physics, The George Washington University, Washington, D.C. 20052, USA NASA Goddard Space Flight Center, Greenb elt, MD 20771, USA Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, USA Max-Planck-Institut fur extraterrestrische Physik, Giessenbachstr. 1, 85740 Garching, Germany Å Universe Cluster, Technische UniversitÅt Munchen, Boltzmannstraße 2, D-85748, Garching, Germany a Å Europ ean Southern Observatory, Karl-Schwarzschild-Str. 2, 85748 Garching, Germany Institute of Astronomy, University of Cambridge, Madingley Road, CB3 0HA, Cambridge, UK Purple Mountain Observatory, Chinese Academy of Sciences, Nanjing 210008, China INAF - IASF Bologna, via P. Gob etti 101, 40129 Bologna, Italy Gemini Observatory, 670 North A'ohoku Place, Hilo, HI 96720, USA

Dark Cosmology Centre, Niels Bohr Institute, Cop enhagen University, Juliane Maries Vej 30, 2100 Cop enhagen Ø, Denmark Centre for Astrophysics and Cosmology, Science Institute, University of Iceland, Dunhagi 5, IS-107 Reykjav?k, i Iceland
18 19 20 21 17

Space Telescop e Science Institute, 3700 San Martin Drive, Baltimore, MD21218, USA School of Earth & Space Exploration, Arizona State University, Box 871404, Temp e, AZ 85287-1404, USA Physics Department, McGill University, 3600 rue University, Montreal, QC H3A 2T8, Canada

Research School of Astronomy & Astrophysics, The Australian National University, Cotter Road, Weston Creek ACT 2611, Australia
22 23 24

Department of Physics, Oxford University, Keble Road, Oxford, OX1 3RH, UK Department of Astronomy, California Institute of Technology, MC 249-17, Pasadena, CA 91125, USA INAF-Osservatorio Astronomico di Brera, via Bianchi 46, 23807 Merate, Italy


-3- deep observations with Gemini-North, the Very Large Telescop e, and the GRB Optical and Near-infrared Detector. Assuming an Small Magellanic Cloud dust law (which has b een found in a ma jority of GRB sight-lines), the 90% likelihood range for the redshift is 9.06 < z < 9.52, although there is a low-probability tail to somewhat lower redshifts. Adopting Milky Way or Large Magellanic Cloud dust laws leads to very similar conclusions, while a Maiolino law does allow somewhat lower redshift solutions, but in all cases the most likely redshift is found to b e z > 7. The non-detection of the host galaxy to deep limits (Y (AB) 28, which would corresp ond roughly to 0.001L at z = 1) in our late time optical and infrared observations with the Hubble Space Telescope, strongly supp orts the extreme redshift origin of GRB 090429B, since we would exp ect to have detected any low-z galaxy, even if it were highly dusty. Finally, the energetics of GRB 090429B are comparable to those of other GRBs, and suggest that its progenitor is not greatly different to those of lower redshift bursts. Subject headings: early Universe - galaxies: high-redshifts - gamma-rays bursts: individual (GRB 090429B) - techniques: photometric

1.

Intro duction

The burst detections and rapid afterglow identifications of the Swift satellite (Gehrels et al. 2009), combined with intensive ground-based follow-up efforts, have confirmed some gamma-ray bursts (GRBs) as among the most distant ob jects known in the universe (Tanvir et al. 2009; Salvaterra et al. 2009), illuminating the conditions of star formation at the earliest ep ochs. As burst detections push toward progressively higher redshifts, the mere existence of GRBs at these times will provide imp ortant constraints on models of gravitational collapse, galaxy formation, and the early generations of stars. At the same time, high-quality sp ectroscopy of the burst afterglows can b e exp ected to reveal element abundances (e.g. Starling et al. 2005; Kawai et al. 2006; Berger et al. 2005), host galaxy kinematics, and p otentially, the H i fraction of the intergalactic medium (IGM), as the process of cosmic reionization unfolds (e.g. Barkana & Loeb 2004; Totani et al. 2006; Tanvir & Jakobsson 2007; McQuinn et al. 2008). GRBs offer some advantages over other techniques for the selection and study of distant galaxies. Most notably, they have unprecedented luminosity, b oth of the prompt emission, and afterglow (e.g. Racusin et al. 2008; Bloom et al. 2009), enabling them to provide detailed diagnostics of their environments, and pinp ointing their host galaxies however faint. However, this utility comes at a price - GRB afterglows achieve such brightness only fleetingly, and so the time available to obtain redshifts and other information for a burst is often very short (normally < 24 hr). In order to realize the ambitions of finding bursts at extreme redshift, and efficiently exploiting high-redshift GRBs as prob es of this early cosmic ep och, it is necessary to devote increasing effort to the rapid identification of GRB near-infrared (NIR) afterglows. In addition to workhorse NIR instrumen-


-4- tation at large observatories, a growing numb er of dedicated facilities and instruments have b een commissioned, with a primary aim of rapidly locating distant GRBs (e.g. PAIRITEL (Bloom et al. 2006); Gamma-Ray Burst Optical and Near-Infrared Detector (GROND) (Greiner et al. 2008)). Follow-up sp ectroscopy of these candidates has proved several to b e at very high redshift (e.g. Kawai et al. 2006; Greiner et al. 2009), culminating in GRB 090423 at z 8.2 (Tanvir et al. 2009; Salvaterra et al. 2009). However, in some cases rapid sp ectroscopy is not p ossible, and we must fall back on photometric redshift measurements (e.g. Jakobsson et al. 2006; Haislip et al. 2006). Here again, GRBs offer some advantages over galaxies for the application of such techniques. First, there is little intrinsic variation in the sp ectral shap e of an afterglow - it can b e modeled simply as a p ower-law plus host galaxy dust extinction Ly- absorption. This is in contrast to the diverse sp ectra of galaxies, which can have contributions from young/old p opulations (or a mixture), dust in complex configurations , exhibit intrinsic curvature, Balmer breaks etc, none of which are a concern for GRB afterglows. Second, the identity of a GRB afterglow is unambiguous from its fading, and thus there is no chance of mistaking a GRB afterglow with e.g. a Galactic L or T dwarf, which can also confuse high-z galaxy searches. It has b een shown that GRB photometric redshifts are generally robust for these reasons (Kruhler et al. 2011). Indeed, while the fundamental accuracy is limited by the bandwidths Å and bands used, GRBs are much less sub ject to the "catastrophic" failure of photometric redshift determination, that can impact individual galaxy measurements. In this pap er we discuss the discovery and multi-wavelength follow-up of GRB 090429B. The afterglow was not visible in deep early optical imaging, but was found in deep IR observations starting 2.5 hr after the burst. While sp ectroscopic observations were curtailed by p oor weather conditions, our photometry does allow us to construct a sp ectral energy distribution (SED) for the burst, and to infer a photometric redshift of z 9.4 , making GRB 090429B one of the most distant ob jects known to date. The pap er is structured as follows: in Section 2 we present our full dataset on GRB 090429B and the uncertainties of our photometric measurements; in Section 3 we derive our photometric redshift, supplemented with deep host observations. Finally, in Section 4 we summarize our conclusions, highlighting the imp ortance rapid-resp onse NIR imaging and sp ectroscopic capability on large telescop es for the study of the early universe using GRBs. Throughout this pap er we assume CDM cosmology with H0 = 72 km s-1 Mp c-1 , M = 0.27, = 0.73, and use a standard nomenclature to describ e the variation of the afterglow flux density as F t- - .


-5- 2. Observations and Analysis 2.1. Swift Observations

The Burst Alert Telescop e (BAT; Barthelmy et al. 2005) ab oard the Swift satellite triggered on GRB 090429B at T0 = 05:30:03 UT. The 15-350 keV light curve is comp osed of three distinct p eaks with a total duration T90 = 5.5 s, and the time-integrated sp ectrum can b e fitted by a single p ower law with an exp onential cut off. The derived total fluence in the 15-150 keV band is 3.1 × 10-7 erg cm-2 , with Epeak = 49 keV. This p eak energy is among the few detected by Swift within the BAT bandpass. After 106 s, the narrow-field instruments b egan their standard burst-resp onse observation sequence. The X-ray Telescop e (XRT; Burrows et al. 2005) identified an uncataloged fading source . at RA(J2000)=14h 02m 40s 10, Dec(J2000)=+32 10 14. 6; no optical/UV counterpart was seen in the UV-Optical Telescop e (UVOT; Roming et al. 2005) data. The X-ray data has b een characterized using standard routines in Heasoft, Xspec, and QDP, with the light curve fitting process as describ ed in Evans et al. (2009). For some analyses, we have used the automatic data products produced by the UK Swift Science Data Centre (Evans et al. 2007, 2009). Our presentation of parameters derived from the Swift data follows the convention of quoting errors at the 90% level. The time-averaged 0.3-10 keV X-ray sp ectrum from 97 to 29893 s after the burst is b est fit by a 0 p ower-law with photon sp ectral index X = 2.01+0.16 and with a total absorption column density - .24 of NH = 10.1+4.6 × 1020 cm-2 , mildly (2.7 ) in excess of the Galactic absorption of 1.2 × 1020 cm-2 ; -5.3 we discuss the p ossible significance of this finding in Section 3.5. The X-ray light curve, given in Table 1 and illustrated in Figure 1, is adequately fit by a combination of brightening and fading temp oral p ower laws: initially, the X-ray flux rises with 0. temp oral index X1 = -0.96+0.43 , referenced to the burst time; following the p eak time TX = - 52 + 0 589-146 s, the light curve then breaks to a p ower-law decay with X2 = 1.20+0.08 . 80 - .07

2.2. Basic reduction steps ware1 . Photometric analys custom scripts. Errors in equal to that of the source

Optical and Near-IR Observations

for all optical and NIR photometry were p erformed using IRAF softis used b oth IRAF and the Starlink GAIA software, as well as our own the sky subtraction step are estimated from multiple ap ertures of size ap erture, placed around the field of the GRB.

Optical images were calibrated using field stars from the Sloan Digital Sky Survey (SDSS) Data-Release 7 (DR7) catalog (Abaza jian et al. 2009) in the and NIR images were provisionally
IRAF is distributed by the National Optical Astronomy Observatory, which is op erated by the Association of Universities for Research in Astronomy, Inc., under coop erative agreement with the National Science Foundation
1


-6- calibrated directly to the Two Micron All Sky Survey (2MASS) catalog, but subsequently refined as describ ed b elow. Detections and limits on the brightness of any associated source are presented in Table 2.

2.2.1.

ESO2.2m/GROND Observations

The GROND (Greiner et al. 2008) observed the field of GRB 090429B simultaneously in its (dichroic and filter defined) g r i z J H Ks filter set b eginning 14 minutes after the Swift discovery (Olivares et al. 2009). No source was detected at the X-ray afterglow p osition in any of the seven bands: the limits b eing shallower than usual due to the high airmass for this (northern) field. Nonetheless, the implied X-ray to optical sp ectral slop e of OX < 0.1 implied suppression of the optical flux relative to the X-ray, rendering GRB 090429B a "dark" burst under the definitions of Jakobsson et al. (2004) and van der Horst et al. (2009).

2.2.2.

VLT Observations

Deep R and z -band observations were made with the VLT/FORS-2 camera at 60 minutes p ost-burst. Once again no optical source was visible at the p osition of the X-ray afterglow, confirming that it was unusually dark, and thus a good candidate high-z GRB (D'Avanzo et al. 2009).

2.2.3.

Gemini-North Observations

Beginning roughly 2.5 hr after the burst trigger, we carried out a series of observations from Gemini-North. We gathered optical i z imaging with the Gemini Multi-Ob ject Sp ectrograph (GMOS; Hook et al. 2004) and NIR J H K imaging with the Near-Infared Imager (NIRI; Hodapp et al. 2003). GMOS observations consisted of five exp osures of 3 minutes each, p er filter; NIRI observations consisted of eight dithered p ositions of 60 s each. The Gemini GMOS and NIRI packages under the IRAF environment were used to sky-subtract, align, and combine the images. The NIRI images were also corrected for the small non-linearity effect seen in the detectors2 . Photometry was p erformed relative to SDSS stars for the GMOS data, and relative to secondary calibrators from GROND for the NIRI data (see Section 2.3). Our photometry is presented in Table 2. While no optical counterpart was present in our i or z images, we did identify a source within the X-ray localization in our NIR observations. The p osition of the source was RA=14h 02m 40.10s , Dec =+32 10 14. 20. Following this discovery we attempted sp ectroscopic observations from Gemini2

http://www.gemini.edu/sciops/instruments/niri/data-format-and-reduction/detector-linearization


-7- North, however, increasing summit winds forced the closure of the telescop e and meant that these were ab orted with <10 minutes of useable exp osure time and no trace is visible in the observations. We obtained a second ep och of K -band observations on April 30 UT revealed a clear fading of 1.2 mag of the identified source, confirming its transient nature (and corresp onding to a p owerlaw index of K = 0.53 + 0.10, shallower than the X-ray decay at that time). Figure 2 presents our Gemini imaging data, while the lower panel of Figure 1 shows our optical/NIR lightcurve. The resulting SED, from X-ray to IR is shown in Figure 3. No evidence of a host galaxy is present in our images. A deep r -band image of the field, taken again with GMOS under good conditions (0.4 seeing) at 14 days after the GRB, is shown in Figure 4. This allows us to place a 3 upp er limit on the host galaxy apparent magnitude of r > 27.07 mag. We also note in these images the presence of a massive elliptical galaxy, offset roughly 45 from the GRB location. This galaxy has absolute magnitude Mr -21.6 and MK -24.5 ( L ; Jones et al. 2006). It app ears to b e the central galaxy of a modest cluster at z = 0.0793 , It is likely that this foreground structure provides some lensing b oost to the observed flux of the burst, although the relatively large impact parameter suggests it will not b e a ma jor factor.

2.2.4.

HST Observations

We obtained late time observations of the field of GRB 090429B with the Hubble Space Telescope (HST). These were taken after the afterglow had faded, and had the goal of finding or constraining the host galaxy. We used b oth the Advanced Camera for Surveys (ACS) and the Wide Field Camera 3 (WFC3). Observations were obtained in F606W (broad V - R), F105W (broad Y - Z ), and F160W (H ): a log is given in Table 3. The data were reduced in the standard fashion using multidrizzle and the HST archive "on-the-fly" calibration. All the images were drizzled to a common pixel scale of 0. 05 pixel-1 . We ascertained the location of the burst on the HST images via relative astrometry b etween our first ep och K -band observations, and those obtained with HST. Doing so we used a total of 11 and 10 sources in common to each frame, for ACS and WFC3 resp ectively. The resulting astrometric accuracy is 0. 08 (F606W), 0. 07 (F105W) and 0. 06 (F160W) resp ectively. At the location of the afterglow we see no obvious host galaxy candidates in any of the images. To quantify the depths of these images we estimate the sky variance from a large numb er of background ap ertures (50) placed in the field around the target p osition, avoiding visible sources. We then measure the resultant flux at the target p osition in an ap erture of 0. 4 diameter, consistent with the approaches of many groups in searching for high-z galaxies (e.g. Bouwens et al. 2010, 2011). Our fluxes are shown in Table 3. In addition to the measured fluxes we also show the effective AB-magnitude limits at these locations, which are equal to the measured flux density + 3 , with an additional ap erture correction to account for light missing within our small measurement ap ertures. These corrections are small for ACS (0.18 mag
3

Redshift and r magnitude of galaxy obtained from the SDSS DR7 database; K magnitude from 2MASS.


-8- for F606W, Sirianni et al. 2005), but larger for the WFC3 images (0.31 and 0.54 magnitudes for F105W and F160W resp ectively 4 ).

2.2.5.

UKIRT Observations

We obtained observations with the United Kingdom Infra Red Telescop e (UKIRT), WideField Camera (WFCAM), b eginning April 29 at 09:18 UT, roughly 4 hr p ost-burst. Only a limited numb er of exp osures were p ossible due to high wind keeping the telescop e shut much of the night. These observations were not deep enough to reveal the afterglow, however b ecause the large field of view (13.6 arcmin square for each chip) includes many bright 2MASS stars these images allowed us to precisely determine the magnitudes of fainter stars, which was crucial for calibrating the NIRI images (see Section 2.3). Pip eline reductions were p erformed by the Cambridge Astronomical Survey Unit (CASU5 ).

2.3.

Precise Photometric Calibration and Uncertainties

Since our photometric redshift analysis will dep end critically on the accuracy of our photometry, we took particular care in b oth the calibration and estimates of photometric uncertainties. The Gemini-North/NIRI detections are crucial, but also difficult to analyze since the field of view is small (2 arcmin on a side) and there is only one 2MASS star (namely star B in Table 4, and Figure 4) that is in all the sub-exp osures of the nine-p oint dither pattern. There is another 2MASS star (A in Table 4) which app ears on two of the sub-exp osures, and we used this as a double check on the derived photometry. Both these stars are toward the faint end of the 2MASS catalog and have relatively large photometric uncertainties. To overcome this we used the wide-field UKIRT/WFCAM and ESO2.2m/GROND J H K images (b oth of which were obtained close in time to the Gemini observations), which were very precisely calibrated using many bright 2MASS stars, to obtain more accurate magnitudes for these reference stars. The two indep endent determinations were consistent with each other within their resp ective calculated errors (typically 0.01-0.02 mag), and we therefore formed a weighted average to obtain our b est estimates of the NIR magnitudes, as shown in Table 2. Magnitudes for the afterglow were measured relative to star B, although this procedure was further complicated by the fact that the p oint-spread-function (PSF) was found to change across the frame resulting in the core of the reference star b ecoming noticeably extended when it was close to the southern edge of the detector, as it was in some sub-exp osures. This precluded small ap erture
4 5

http://www.stsci.edu/hst/wfc3/phot zp lbn http://casu.ast.cam.ac.uk/


-9- (5 pixel radius, 0. 6 ) photometry for these exp osures, so in such cases we used a fainter star (C in Table 2) closer to the GRB p osition as a secondary reference, having determined its magnitude relative to star B using those frames where it was not near the edge. We note that a small ap erture was required to maximize the signal to noise for the afterglow, and that profile-fitting photometry was deemed inappropriate due to the small numb er of bright stars available to define the PSF. The magnitudes (and errors) for the afterglow in each band were then determined from an error weighted mean of the different sub-exp osures. Finally, we converted to flux density using a recent NIR sp ectrum of Vega (see Bohlin 2007) which resulted in values that are 2%-3% higher for our passbands than found using the conversion in Table 7 of Hewett et al. (2006). These are the flux densities rep orted in Table 2, although we note that when we come to the photometric redshift analysis (b elow) we fit in counts rather than flux, to allow for the different sp ectral shap es of the afterglow and comparison star. Since the optical observations provided only upp er limits the overall fit is not strongly sensitive to the precision of the optical photometry. However, in this case, the field of GRB 090429B lies fortuitously within the SDSS survey area, and our most constraining optical limits (from GMOS) are obtained in the same filter set. This allows a precise photometric calibration of these images. For our Very Large Telescop e (VLT) observations we calibrate the field using SDSS observations and the transforms of Jester et al. (2005). These latter values were confirmed as reasonable using archival zerop oints.

3.

Results and Discussion Temp oral Behavior

3.1.

Since our observations were taken over several hours, temp oral variations in the afterglow luminosity could effect our analysis. Unfortunately we have rather little handle on the variability at optical/IR wavelengths. Although most GRBs b egin p ower-law decline in luminosity fairly early, in some cases flat or even increasing luminosity can b e seen for a p eriod of time (e.g. Rykoff et al. 2009). A rapidly fading afterglow (similar to those commonly observed) would imply even more stringent upp er limits in our (earlier) blue band filters since the extrap olation to a common time would yield a more extreme limit on color index in e.g. z - H . Alternatively, the more unusual case of a rapidly rising afterglow would yield somewhat weaker constraints since the non-detections in the bluer bands could b e ascrib ed to the brightening of the afterglow in the time frame b etween the optical and IR observations. However, this is countered by the fact that such a rising afterglow would also imply an even bluer H - K color, more difficult to attain with extinction. In this scenario, the rising of the afterglow may offer some supp ort for a high-redshift scenario, since the time dilation at z 9 would result in a forward shock which takes a factor of 10 longer to reach maximum than at z 0.


- 10 - In an attempt to constrain the temp oral slop e of the optical afterglow we first p erform photometry on the individual NIRI frames. We find no statistically significant variation over the 10 minute time frame of these observations, implying that the afterglow is not varying esp ecially rapidly. Second, we utilize acquisition images taken prior to the ab orted NIRI sp ectroscopic observations. These suggest a minor brightening of the afterglow b etween 13000 s and 17000 s after the burst (0.3 + 0.2 mag, corresp onding to -1.0 + 0.7). However, these observations were obtained at a single dither p osition, and contained substantial p ersistence. Hence, we can accurately remove neither sky nor dark current and the resulting observations contain large variations in the sky on relatively short length scales. Thus, we caution against their use for detailed photometric work, aside from noting that suggest that the afterglow is neither rising, nor falling at an unusually rapid rate. We gain a much b etter handle on the decay b etween the first and second night observations, which gives 0.6 + 0.1, but of course this may not apply during the first few hours p ost-burst. On balance, then, we favor photometric fitting in which the observed magnitudes are assumed to b e constant over the p eriod of the early observations (i.e. we assume = 0). This is consistent with the relatively flat X-ray b ehavior b etween 1 and 3 hr p ost-burst (Figure 1). For completeness, we have included a single p ower-law temp oral decay as a p ossible parameter within these fits, and confirm that for any reasonable slop e -1 < < 1 our results are broadly insensitive to the assumed value of (see b elow). To avoid extrap olating over too wide a range of times, and to counter against unusual afterglow b ehavior (which is normally most notable in the first few hundred seconds in the rest frame) we also fit only data taken after 4000 s. The inclusion of earlier data would further strengthen our results if we assumed the afterglow to b e decaying, but would make minimal difference to the fit if we assumed a flat or rising light curve (since the early limits are shallower than those obtained at later times).

3.2.

Photometric Redshift Analysis

Here we attempt to derive the redshift of GRB 090429B via our broadband photometry of its afterglow. The absence of any detections within the optical window, if interpreted as the signature of high redshift, immediately implies z > 6.3. Similarly, if we interpret the red J - H color of the afterglow as indicative of a Lyman-break lying within the J -band, the inferred redshift is 8.0 < z < 10.5. To obtain stronger constraints on the redshift of GRB 090429B we p erformed the following analysis. We considered just the seven deep est observations, namely those obtained at the VLT and Gemini-North on the first night with filters redder than 6000 œ. We assume initially there is no A temp oral variation over the course of observations, although including plausible variability within our fits also confirms that our results are broadly insensitive to this assumption (see b elow). After correcting for Galactic foreground extinction (EB -V = 0.015; Schlegel et al. 1998) we fitted these flux density measurements with a grid of simple models for the SED of the afterglow. The errors are likely to b e Gaussian distributed, to good approximation, since the uncertainties are dominated


- 11 - by background subtraction (although we also included zero-p oint calibration uncertainties in the modelling), and therefore used minimum-2 fitting. Sp ecifically, the model was a simple p ower law, with the sp ectral index, O , and overall normalization as free parameters. The grid of models spanned a range in redshift of 0 < z < 12 and rest-frame V -band extinction of 0 < AV < 12. Beyond z 7 we are effectively fitting only three data p oints so there exists a degeneracy b etween extinction (AV ) and O . We therefore include a weak prior for the probability distribution of the value of O , (see Figure 5) which is modeled as a log normal with a p eak likelihood at O = 0.5 and a width such that the relative likelihood is 50% of the maximum from ab out 0.3 < O < 0.85. This is physically motivated since it allows values of O over a broad range, comparable to the range usually observed (e.g., Schulze et al. 2011), but in particular prefers O = X - 0.5 0.5, as would b e exp ected if there was a cooling break b etween the X-ray and optical regimes (e.g., Sari et al. 1998). The plausibility of such a cooling break is clear from Figure 3, while relaxing this assumption does not lead to any viable fits at low redshift. Added to this was absorption due to neutral hydrogen in the IGM (Madau 1995), (neutral hydrogen in the host was taken to have a typical column density of 1021 cm-2 , although the results are insensitive to the exact numb er assumed), and extinction due to dust. We exp erimented with several dust laws, from the Milky Way (MW), Large Magellanic Cloud (LMC) and Small Magellanic Cloud (SMC) (Pei 1992), as well as the extinction law of Maiolino et al. (2004). We also imp ose a weak prior on the intrinsic luminosity of the optical afterglow (Figure 5). Studies such as that of Kann et al. (2010), indicate that there is an upp er envelop e to the (broad) distribution of GRB optical afterglow luminosities. We therefore apply a prior which is flat b elow this envelop e and cuts off exp onentially at brighter luminosities, although the cut-off is slow enough to allow a reasonable probability that the luminosity could b e somewhat higher. In fact, rather than fitting directly to the flux densities, we integrated our model sp ectra over resp onse functions and compared the counts obtained by integrating an approximate sp ectrum of the comparison star. The resp onse functions were obtained from the measured filter transmission curves, multiplied by a typical atmospheric absorption curve generated by ATRAN6 . Going to these lengths effectively corrects for the small difference in the SED shap e of the afterglow from the reference star, although again the conclusions are not greatly affected. In Figure 6 we show the photometric data p oints and the b est-fit model for the afterglow sp ectrum assuming that the afterglow did not evolve temp orally during the first 3 hr (see b elow) and that the dust is similar to that in the SMC, which has frequently b een found to b e a good approximation to the dust laws along many other GRB sight lines (e.g., Schady et al. 2007, 2010). We also show the b est fit low-z model (as it happ ens z 0), which is formally ruled out at high confidence. In Figure 7 we plot contours of 2 over a grid of models spanning a range of redshift and rest-frame V -band extinction, AV . The red-cross shows the b est fitting model, which has z = 9.36 and extinction AV = 0.1, although the 99% confidence contour runs as low as z 7.7 if there is a modest amount of dust
6

http://atran.sofia.usra.edu/cgi-bin/atran/atran.cgi


- 12 - (rest-AV 0.5) in the host. Marginalizing the likelihood (which we define L exp(-2 /2)) over AV (assuming a flat prior) indicates a 90% likelihood range of 9.02 < z < 9.50. There is no solution at lower redshifts (z < 7) which is not ruled out at 99.9% level: and the b est fit at low redshift (z 0 as it happ ens, as shown by the blue cross) requires very high extinction of AV 10. In Figure 8 we show similar likelihood contours for fits spanning a broader range of models with different prior assumptions for the temp oral p ower-law decline index , and commonly used dust laws. Changing to +1 makes rather little difference, and in any case, as discussed ab ove, there is evidence to suggest the luminosity was not changing even as rapidly as this. Varying the dust law does have more effect, largely due to the 2175 œ feature in the MW, LMC (Pei 1992), and A Maiolino et al. (2004) laws producing the blue H - K color even at slightly lower redshifts, although in most cases the b est fit remains z 9. The Maiolino et al. (2004) dust law was determined from observations of a quasar at z = 6.2 and is argued to consistent with dust produced largely from early sup ernovae (note that this law is only defined up to 3200 œ in the rest frame, and we A therefore graft it to the SMC law at this p oint). This case is interesting as it does allow redshifts as low as z 6.5 at 99% confidence, although to date, only GRB 071025, with a photo-z 5, has shown evidence of requiring such a dust law (Perley et al. 2010). Finally, in Figure 9 we show the likelihood as a function of redshift for the SMC dust-law models having marginalized over b oth (assumed a flat prior b etween -1 and +1) and AV (assumed a flat prior b etween 0 and 12). The maximum likelihood is at z = 9.38, and 90% of the likelihood is b etween 9.06 < z < 9.52.

3.3.

Implications of the Absence of a Host Galaxy

Our late time date taken with Gemini and HST are p otentially extremely valuable, since we can use the absence of any host galaxy candidates to assess the plausibility of any lower-z solutions to our photometric redshifts (the HST images are shown in Figure 10). The detection of a host galaxy in the optical was used, for example, to show that GRB 060923A was z < 3 despite its afterglow b eing a K -band drop-out (Tanvir et al. 2008). In the case of GRB 090429B, the p ossible low-redshift scenarios seem to b e those with z < 1 and high dust extinction AV 10 (although we emphasize that such models remain formally ruled out). The limits these data provide on this are shown graphically in Figure 11, where we plot the absolute inferred magnitude of the host galaxy in the observed V , Y and H bands as a function of redshift. For completeness we cut each line at the p oint where 1216œ×(1 + z ) passes the central wavelength of the band. At z = 0.1 A close to the minimum of our lower redshift solution we obtain inferred absolute magnitude limits of MV > -10.6, MY > -9.9, MH > -10.5, these exceptionally deep limits are comparable to the luminosities of bright globular clusters, and significantly fainter than any known GRB or sup ernova host galaxy, indeed they place limits of 10-4 L (Blanton et al. 2003). Even at z = 1 the observed Y -band limits would imply MB > -15.1, or 0.001L (Ryan et al. 2007).


- 13 - In this regard it is worth noting that the lower redshift solutions are only viable in cases where the host galaxy extinction is high, whereas such faint galaxies typically have low metallicity, and little dust, and it is therefore extremely unlikely that one could create the extinction (AV 10) necessary to explain GRB 090429B. Any z < 3 solution would require the host of GRB 090429B to b e fainter than the large ma jority of GRB hosts currently known. Furthermore, our wide wavelength coverage would also allow us to uncover any very red dusty host galaxies, which would provide the necessary extinction, but would b e missing from optical only searches (e.g., Svensson et al. 2010; Levan et al. 2006) This offers strong supp ort for our high-z model, In these cases only the F160W observation yields p otential information as to the magnitude of the host galaxy. The inferred 1500 œ absolute A AB magnitude at z 9.4 is -19.95. This lies roughly in the middle of the observed absolute magnitude distribution of z 8 candidate galaxies found in deep HST ACS and WFC3 imaging (Bouwens et al. 2010), and thus the non-detection of a galaxy at z 9.4 is not unexp ected in observations of this depth.

3.4.

Other Indicators of High Redshift

In addition to the ab ove discussion there are additional lines of evidence which offer supp ort for the high-z interpretation of GRB 090429B. The b est z 5 solution is actually at z 0 (although it remains a very p oor fit to the available optical data) and requires AV 10, for an SMC extinction law, which would corresp ond to a foreground NH 1023 cm-2 , nearly two orders of magnitude larger than is observed. While the dust-to-gas ratios observed through the MW can show moderately large variations, such a large offset would b e unheard of, particularly in GRB afterglows where typically the ratio of dust extinction to X-ray determined gas column is actually less than is seen locally (Schady et al. 2010). Hence the observed X-ray sp ectrum seems to rule out any low redshift (z 1), high extinction scenario. This is illustrated in Figure 7 which shows that the b est low-redshift solutions are well ab ove the contours of AV inferred from the excess NH assuming typical GRB dust-to-gas ratios. A second line of evidence comes from observed high-energy correlations seen in many GRBs. In particular the relation b etween the p eak energy of the F sp ectrum (Ep ) and the burst isotropic energy (Amati et al. 2008). Although this relation has a significant scatter, it can also b e used to place some constraints on the burst redshift under the assumption that all long bursts should follow the relation. For GRB 090429B, the burst is only consistent with this relation at b etter than 3 if z > 1, implying that it also disfavors very low redshift models for the origin of GRB 090429B. This result couples with the lag-luminosity relation, which would imply an isotropic luminosity of Liso 1053 erg s-1 , similar to other long GRBs in the "silver sample" (Ukwatta et al. 2010).


- 14 - 3.5. Rest-frame Prop erties

The observed fluence of GRB 090429B is 3.1 × 10-7 erg cm-2 , comparable to that observed for GRB 090423, and implies an isotropic energy release in the 15 - 150 keV band of Eiso = 3.5 × 1052 erg at z 9.4 . Its absolute X-ray (rest-frame 3-100 keV) brightness at 1000 s of 2 × 1049 erg s-1 , and K -band luminosity of M/(1+z ) of -26.2(AB) are also similar b oth to GRB 090423 and the bulk of the long GRB p opulation. The rest-frame duration at first sight seems surprisingly short, with T90 /(1 + z ) 0.5 s. Interestingly, three of the four highest redshift GRBs discovered prior to this one have also had rather low values of T90 /(1 + z ) around 1-4 s (Ruiz-Velasco et al. 2007; Greiner et al. 2009; Tanvir et al. 2009; Salvaterra et al. 2009). A p ossible explanation for this tendency may b e that for high redshift sources the BAT is observing at rest frame MeV energies, where the light curves tend to b e more rapidly variable and shorter duration than at lower energies, and thus it is plausible that just a single p eak of emission is detected rising up ab ove the noise in these cases. An analysis obtained using the technique discussed in Ukwatta et al. (2010) reveals that this GRB presents a small value of lag31/(1 + z ) = 58 + 27 ms b etween the low-energy (15-25 keV) and the high-energy (50-100 keV) channels. The rest-frame correlations found by Ukwatta et al. (2010) indicate this corresp onds to a p eak isotropic luminosity Liso > 1052 ergs-1 , consistent with the observed fluence and short duration. Another interesting issue is that of the absorption inferred from the X-ray sp ectrum. Although the measurement is not highly significant, if taken at face value, the rest-frame column density is NH = 1.4+1.0 × 1023 cm-2 (90% confidence range). This would b e already high compared to -1.0 most other Swift observed GRBs, and would b e higher still if, as is very likely, the metallicity is substantially less than Solar (which by convention is often assumed in calculating NH ). Such a high-column density is no doubt surprising, although a similar value was found for GRB 090423 (Tanvir et al. 2009; Salvaterra et al. 2009). As in that case, it also raises the question of whether a high gas column would b e compatible with the low extinction indicated by the NIR afterglow. GRBs are exp ected to b e able to destroy dust to fairly large distances from their birth sites (e.g., Waxman & Draine 2000; Fruchter et al. 2001), but the good fit for many afterglow SEDs with a SMC extinction law suggests they have not generally b een highly modified by dust destruction, which would tend to produce "gray" extinction laws (e.g. Schady et al. 2010). In any event, the large error bar on this measurement and p otential systematic uncertainties in calibrating the soft resp onse of the XRT, taken together with the wide range of dust-to-gas ratios seen to other GRB sight-lines, makes the significance of this conflict hard to assess at the present time. On balance, it seems that in most resp ects the general prop erties of GRB 090429B do not stand far apart from the p opulation of long-duration GRBs, even at the inferred redshift of z 9.4 . In particular it shows no evidence that its progenitor is distinct from those of GRBs seen in the more local universe. This is of particular imp ortance at z 9.4 , since it is close to the redshift where WMAP observations imply the bulk of reionization of the universe occurred (z = 10.6 + 1.2,


- 15 - Komatsu et al. 2011). This reionization process is likely to have b een driven by the first generations of star formation, including Population I I I whose pristine H+He comp osition is exp ected to lead to generally more massive stars. It has b een prop osed that a consequence of this could b e that Population I I I stars produce particularly long duration, and energetic GRBs (e.g. M?szaros & Rees e? 2010). This is clearly not the case for GRB 090429B, and hence we conclude that its progenitor was more likely to b e a high-mass second generation (Population I I) star.

4.

Conclusions

We have presented our discovery and multi-wavelength observations of GRB 090429B and its NIR afterglow, and a deep late-time search for its host galaxy. The afterglow exhibited a strong sp ectral break in the J band, which coupled with the non-detection in the optical, and relatively blue H - K color, allows us to derive a b est-fit photometric redshift of z 9.4 . It is, of course, imp ortant to look carefully at the evidence against a lower redshift origin, since we know that Swift GRBs exist in much greater numb er at z < 4 than ab ove it. Our afterglow photometry allows us to exclude all z < 6 solutions with high confidence. A low redshift (z 1) is also effectively ruled out by our HST observations which would easily locate such dusty galaxies in either the optical or IR at z < 1, and also the relatively modest excess NH which would not b e consistent with a very high dust column. This immediately implies that GRB 090429B is one of the most distant ob jects yet discovered. The maximum-likelihood solution, with our preferred assumption of an SMC extinction law, is z = 9.38 with a 90% likelihood range of 9.06 < z < 9.52. The conclusions do not dep end sensitively on the priors adopted for other parameters, although a Maiolino et al. (2004) dust law would favor somewhat lower redshift, but still z > 7, at the exp ense of requiring fairly significant extinction (up to rest frame AV 2). Since all z > 6 bursts observed to date are consistent with having AV = 0 (Zafar et al. 2010, 2011), this suggests that z 9.4 provides a good estimate of the redshift of GRB 090429B. Our campaign shows again how rapid-resp onse multiband NIR observations play a crucial role in identifying candidate extreme-redshift afterglows. However, it also highlights the need for even more rapid observations and decisions to maximize the likelihood that sp ectroscopic observations can b e successfully obtained. In the future, additional dedicated ground-based optical/NIR multiband imagers such as GROND and RATIR (Farah et al. 2010) can b e exp ected to feed further such candidates directly to NIR sp ectrographs including X-Shooter on the VLT (D'Odorico et al. 2004), FIRE on Magellan (Simcoe et al. 2008), and GNIRS on Gemini (Elias et al. 2006); ultimately, such prompt sp ectroscopy of extreme-redshift candidates will not only resolve the nature of these events, but quite likely succeed in realizing the extraordinary promise of GRBs as prob es of the extreme-redshift universe. The Gemini data, acquired under the program ID GN-2009A-Q-26, are based on observations obtained at the Gemini Observatory, which is op erated by the Association of Universities for Re-


- 16 - search in Astronomy, Inc., under a coop erative agreement with the NSF on b ehalf of the Gemini partnership: the National Science Foundation (United States), the Science and Technology Facilities Council (United Kingdom), the National Research Council (Canada), CONICYT (Chile), the Australian Research Council (Australia), Minist?rio da Ci^ncia e Tecnologia (Brazil) and Minise e terio de Ciencia, Tecnolog?a e Innovaci?n Productiva (Argentina). Based on observations made i o with the NASA/ESA H ubbleS paceT elescope, obtained from the data archive at the Space Telescop e Institute. STScI is op erated by the association of Universities for Research in Astronomy, Inc. under the NASA contract NAS 5-26555. Data presented in this pap er is associated with programme GO-11189. Part of the funding for GROND (b oth hardware as well as p ersonnel) was generously granted from the Leibniz-Prize to Prof. G. Hasinger (DFG grant HA 1850/28-1). TK acknowledges supp ort by the DFG cluster of excellence "Origin and Structure of the Universe". AR acknowledges funding from the Science and Technology Funding Council. The Dark Cosmology Centre is funded by the Danish National Research Foundation. FOE acknowledges funding of his Ph.D. through the Deutscher Akademischer Austausch-Dienst (DAAD). KT and XFW acknowledge NASA NNX09AT72G and NNX08AL40G. We thank Paul Hewett for helpful discussions ab out absolute infrared flux calibration.

REFERENCES Abaza jian, K. N. et al. 2009, ApJS, 182, 543 Amati, L., Guidorzi, C., Frontera, F., Della Valle, M., Finelli, F., Landi, R., & Montanari, E. 2008, MNRAS, 391, 577 Barkana, R. & Loeb, A. 2004, ApJ, 601, 64 Barthelmy, S. D. et al. 2005, Space Science Reviews, 120, 143 Berger, E., Penprase, B. E., Cenko, S. B., Kulkarni, S. R., Fox, D. B., Steidel, C. C., & Reddy, N. A. 2006, ApJ, 642, 979 Berger, E., Penprase, B. E., Fox, D. B., Kulkarni, S. R., Hill, G., Schaefer, B., & Reed, M. 2005, Arxiv.org, astro-ph/0512280 Blanton, M. R. et al. 2003, ApJ, 592, 819 Bloom, J. S. et al. 2009, ApJ, 691, 723 Bloom, J. S., Starr, D. L., Blake, C. H., Skrutskie, M. F., & Falco, E. E. 2006, in Astronomical Society of the Pacific Conference Series, Vol. 351, Astronomical Data Analysis Software and Systems XV, ed. C. Gabriel, C. Arviset, D. Ponz, & S. Enrique, 751 Bohlin, R. C. 2007, in Astronomical Society of the Pacific Conference Series, Vol. 364, The Future of Photometric, Sp ectrophotometric and Polarimetric Standardization, ed. C. Sterken, 315-+


- 17 - Bouwens, R. J. et al. 2011, Nature, 469, 504 -- 2010, ApJ, 709, L133 Burrows, D. N. et al. 2005, Space Science Reviews, 120, 165 D'Avanzo, P., Levan, A. J., Malesani, D., Thoene, C. C., Fynb o, J. P. U., Hjorth, J., & Tanvir, N. 2009, GRB Coordinates Network, 9284, 1 D'Odorico, S. et al. 2004, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, ed. A. F. M. Moorwood & M. Iye, Vol. 5492, 220-229 Elias, J. H., Joyce, R. R., Liang, M., Muller, G. P., Hileman, E. A., & George, J. R. 2006, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 6269 Evans, P. A. et al. 2009, MNRAS, 397, 1177 -- 2007, A&A, 469, 379 Farah, A. et al. 2010, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, ed. Ian S. McLean and Suzanne K. Ramsay and Hideki Takami, Vol. 7735 Fruchter, A., Krolik, J. H., & Rhoads, J. E. 2001, ApJ, 563, 597 Fruchter, A. S. et al. 2006, Nature, 441, 463 Gehrels, N., Ramirez-Ruiz, E., & Fox, D. B. 2009, ARA&A, 47, 567 Greiner, J. et al. 2008, PASP, 120, 405 -- 2009, ApJ, 693, 1610 Haislip, J. B. et al. 2006, Nature, 440, 181 Hewett, P. C., Warren, S. J., Leggett, S. K., & Hodgkin, S. T. 2006, MNRAS, 367, 454 Hodapp, K. W. et al. 2003, PASP, 115, 1388 Hook, I. M., Jørgensen, I., Allington-Smith, J. R., Davies, R. L., Metcalfe, N., Murowinski, R. G., & Crampton, D. 2004, PASP, 116, 425 Jakobsson, P., Hjorth, J., Fynb o, J. P. U., Watson, D., Pedersen, K., Bjornsson, G., & Gorosab el, Å J. 2004, ApJ, 617, L21 Jakobsson, P. et al. 2006, A&A, 447, 897 Jester, S. et al. 2005, AJ, 130, 873 Jones, D. H., Peterson, B. A., Colless, M., & Saunders, W. 2006, MNRAS, 369, 25


- 18 - Kann, D. A. et al. 2010, ApJ, 720, 1513 Kawai, N. et al. 2006, Nature, 440, 184 Komatsu, E. et al. 2011, ApJS, 192, 18 Kruhler, T. et al. 2011, A&A, 526, A153+ Å Levan, A. et al. 2006, ApJ, 647, 471 Madau, P. 1995, ApJ, 441, 18 Maiolino, R., Schneider, R., Oliva, E., Bianchi, S., Ferrara, A., Mannucci, F., Pedani, M., & Roca Sogorb, M. 2004, Nature, 431, 533 McQuinn, M., Lidz, A., Zaldarriaga, M., Hernquist, L., & Dutta, S. 2008, MNRAS, 388, 1101 M?szaros, P. & Rees, M. J. 2010, ApJ, 715, 967 e? Olivares, F., Kup cu Yoldas, A., Greiner, J., Kruehler, T., & Yoldas, A. 2009, GRB Coordinates Network, 9283, 1 Pei, Y. C. 1992, ApJ, 395, 130 Perley, D. A. et al. 2010, MNRAS, 406, 2473 -- 2009, ApJ, 696, 1871 Racusin, J. L. et al. 2008, Nature, 455, 183 Roming, P. W. A. et al. 2005, Space Science Reviews, 120, 95 Ruiz-Velasco, A. E. et al. 2007, ApJ, 669, 1 Ryan, Jr., R. E. et al. 2007, ApJ, 668, 839 Rykoff, E. S. et al. 2009, ApJ, 702, 489 Salvaterra, R. et al. 2009, Nature, 461, 1258 Sari, R., Piran, T., & Narayan, R. 1998, ApJ, 497, L17 Schady, P. et al. 2007, MNRAS, 377, 273 -- 2010, MNRAS, 401, 2773 Schlegel, D. J., Finkb einer, D. P., & Davis, M. 1998, ApJ, 500, 525 Schulze, S. et al. 2011, A&A, 526, A23


- 19 - Simcoe, R. A. et al. 2008, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 7014 Sirianni, M. et al. 2005, PASP, 117, 1049 Starling, R. L. C. et al. 2005, A&A, 442, L21 Svensson, K. M., Levan, A. J., Tanvir, N. R., Fruchter, A. S., & Strolger, L.-G. 2010, MNRAS, 405, 57 Tanvir, N. R. et al. 2009, Nature, 461, 1254 Tanvir, N. R. & Jakobsson, P. 2007, Royal Society of London Philosophical Transactions Series A, 365, 1377 Tanvir, N. R. et al. 2008, MNRAS, 388, 1743 Totani, T., Kawai, N., Kosugi, G., Aoki, K., Yamada, T., Iye, M., Ohta, K., & Hattori, T. 2006, PASJ, 58, 485 Ukwatta, T. N. et al. 2010, ApJ, 711, 1073 van der Horst, A. J., Kouveliotou, C., Gehrels, N., Rol, E., Wijers, R. A. M. J., Cannizzo, J. K., Racusin, J., & Burrows, D. N. 2009, ApJ, 699, 1087 Waxman, E. & Draine, B. T. 2000, ap j, 537, 796 Zafar, T., Watson, D. J., Malesani, D., Vreeswijk, P. M., Fynb o, J. P. U., Hjorth, J., Levan, A. J., & Michalowski, M. J. 2010, A&A, 515, A94 Zafar, T., Watson, D. J., Tanvir, . R., Fynb o, J. P. U., Starling, R. L. C., & Levan, A. J. 2011, ArXiv:astro-ph/1101.1503

A This preprint was prepared with the AAS L TEX macros v5.2.


- 20 -

Table 1. X-ray Observations T - T0 (s) 158 291 368 452 561 651 708 768 838 927 4618 5636 6417 10316 10982 15921 120249 Flux Density (÷Jy) 0.53 0.80 1.18 1.19 1.69 1.37 1.35 1.18 1.00 1.31 0.099 0.111 0.126 0.089 0.087 0.0269 0.0021 Error (÷Jy) 0.12 0.18 0.27 0.27 0.36 0.31 0.30 0.27 0.22 0.21 0.026 0.029 0.033 0.023 0.023 0.0053 0.0006

Note. -- X-ray observations obtained by the XRT instrument onb oard the Swift satellite. The flux density is calculated at 2 keV, while the conversion factor from flux to Jy is 8.87 × 104 erg cm-2 s-1 Jy-1 (0.3-10 keV). The counts to flux conversion factor is 2.984 × 10-11 erg cm-2 s-1 count-1 . Uncertainties are 1 . For the b est fit X-ray sp ectrum parameters see Section 2.


- 21 -

Table 2. Log of Ground-based Optical/NIR Observations T - T0 (s) 990 990 990 990 990 990 990 3224 4017 5144 8135 9350 10611 11785 13280 95658 1.2 × 106 Magnitude > 23.08 > 22.86 > 22.03 > 21.87 > 21.06 > 20.50 > 19.84 > 24.5 > 25.9 > 23.6 > 25.7 > 24.5 22.80 + 0.16 21.41 + 0.05 21.12 + 0.04 22.42 + 0.16 > 27.07 Flux Density (÷Jy) Filter g r i z J H K B R z i z J H K K r Telescop e GROND GROND GROND GROND GROND GROND GROND VLT/FORS2 VLT/FORS2 VLT/FORS2 Gemini-N/GMOS Gemini-N/GMOS Gemini-N/NIRI Gemini-N/NIRI Gemini-N/NIRI Gemini-N/NIRI Gemini-N/GMOS

0.00 + 0.20 -0.08 + 0.08 0.27 + 0.34 0.02 + 0.06 0.02 + 0.18 2.82 + 0.44 10.21 + 0.50 13.26 + 0.51 4.0 + 0.6

Note. quoted in tion along are at the

-- Optical/NIR observations of GRB 090429B. Magnitudes are the AB system, and corrected for the exp ected Galactic extincthe line of sight, EB -V = 0.015. Quoted errors are 1 and limits 3 level.


- 22 -

Table 3: Log of HST Observations of the GRB 090429B Field Date 3 Jan 2010 10 Jan 2010 22 Feb 2010 24 Feb 2010 28 Feb 2010 Start Time 03:13 21:54 19:22 03:19 13:56 Inst/Filter ACS/F606W WFC3/F160W WFC3/F160W WFC3/F105W WFC3/F105W Exp time 2100 2412 2412 2412 2412 Limit > 27.6 > 27.5 > 28.3 Flux density (÷Jy) 0.005 + 0.008 0.007 + 0.005 -0.001 + 0.005

Note. -- A log of the H S T optical and NIR observations of the GRB 090429B field. Flux densities are given in the measured ap ertures and are not corrected for light outside the ap ertures. Errors are 1 and the limits are given in the AB-magnitude system at the 3 level (and do include ap erture corrections).

Table 4. Star A B C RA 14:02:35.05 14:02:40.60 14:02:38.11

Secondary Standards within the NIRI Field of View DEC J J H2M Hc K K

2M AS S

cal

AS S

al

2M AS S

cal

32:11:07.2 32:09:28.9 32:10:08.9

14.754 +0.034 15.419 +0.055

14.753 +0.006 15.451 +0.007 19.523 +0.035

14.411 +0.055 14.968 +0.076

14.407 +0.008 15.023 +0.009 19.001 +0.028

14.212 +0.075 14.846 +0.127

14.341 +0.010 14.944 +0.015 18.585 +0.032

Note. -- Vega magnitudes for our two secondary standard stars utilized in photometry of the afterglow of GRB 090429B. The 2MASS entries refer to the magnitudes contained within the 2MASS catalog, while those denoted cal refer to our improved values based on the UKIRT/WFCAM and ESO2.2/GROND observations. Uncertainties are 1 .


- 23 -

Fig. 1.-- X-ray (top) and optical/IR (b ottom) lightcurve of GRB 090429B, the left-hand and b ottom axis represent the observed time and flux/magnitude, while the top and right-hand axis show rest-frame time and luminosity, resp ectively. The solid p oints in the top panel show the observed XRT data, along with a solid line representing the model. The dashed line represents the b est fit model for GRB 090423 (Tanvir et al. 2009) overplotted as it would app ear at z 9.4 . The lower panel shows the optical lightcurve, along with a single p ower-law fit to the (red) K -band p oints. (H and J are shown as green and blue, resp ectively. For clarity we have shown only the iand z -band limits (cyan) in the optical). Additionally, the dashed line again shows the model of GRB 090423 at z 9.4 . As can b e seen, the luminosity and general b ehavior of GRB 090429B in b oth X-ray and optical is similar to that of GRB 090423.


- 24 -

Fig. 2.-- Discovery images of the GRB 090429B afterglow. The images are all obtained from Gemini-North, and show the deep non-detection in the z band (which agrees with similar observations in gr iz obtained at GROND, B , R, z obtained at the VLT, and an i-band image at Gemini), coupled with the relatively bright ob ject seen in H and K . At z 9.4 , Ly- lies within the J band, and explains the marginal detection at that wavelength.

Fig. 3.-- IR to X-ray sp ectral energy distribution at T0 + 104 s can b e explained by an intrinsic broken p ower-law sp ectrum. The green solid line extends the IR sp ectral slop e derived from the fit to the optical/NIR data, alb eit that the prior on O does essentially fix this value. The blue dot-dashed line extrap olates the unabsorb ed X-ray sp ectrum to lower frequencies, showing that a single p ower-law fails to fit the broadband SED at this time. The red dashed line shows the SED for the b est-fit extreme-redshift (z 9.4 ) model. z and i upp er limits are shown as black triangles.


- 25 -

Fig. 4.-- Wide-field image of the GRB 090429B field, obtained with Gemini-North 14 days after the burst. The location of the GRB is marked with a crosshair. Additionally, we mark the p ositions of the three comparison stars used to refine our photometry (note that star C is faint, and lies at the end of the marked arrow, just to the south of the galaxy), and the location of a large elliptical galaxy (G1), which is the central galaxy of a modest cluster at z 0.08, which may provide a modest lensing magnification. Note the silhouette of the guide prob e obscures part of the field.


- 26 -

Fig. 5.-- Input priors adopted for our photometric redshift fitting. [Left panel:] In the relativistic fireball model, the intrinsic sp ectral slop e in the optical should lie b etween X and X - 0.5 (plus the associated measurement errors). To achieve this we use a lognormal distribution centered at 0.5 (since there does app ear to b e a break b etween the optical and X-ray, see Figure 3). This is a relatively weak prior and simply avoids extreme values of . Right panel: The second prior is on the intrinsic optical afterglow luminosity, and impacts solutions that would result in an unreasonably bright luminosity (it is not b ounded at the faint end, and hence the low-redshift solutions are unaffected). It is therefore based on the empirically observed upp er envelop e of afterglow luminosities. The primary impact of this prior is to disfavor moderate (AV > 3) scenarios at high redshift (z > 7), where the burst would have b een more luminous than any other known afterglow.


- 27 -

Fig. 6.-- Sp ectral energy distribution of the GRB 090429B afterglow formed by extrap olating our observed photometry to 3 hr p ost-burst assuming the magnitude remains constant, i.e. = 0 (for varying fits see Figure 8). The vertical error bars represent 1 uncertainty, and the horizontal shaded bars illustrate the widths of the broadband filters. The b est fit model (2 /dof = 1.76/3) to the data p oints is shown as the solid red line, the parameters b eing redshift z = 9.36, restframe extinction AV = 0.10 and intrinsic p ower-law slop e O = 0.51. The inset simply replots the short wavelength part of the figure (indicated by a dotted b ox) on a logarithmic flux density scale, to more clearly show the constraints from the optical measurements. An alternative low-redshift (z 0), high extinction (AV = 10.6) model is shown as a dashed blue line, but in fact is formally ruled out at high significance (2 /dof = 26.2/4).


- 28 -

Fig. 7.-- Confidence contours on a parameter space of redshift and host galaxy extinction for the GRB 090429B afterglow, for our favored set of prior assumptions (green contours are 90%, 99% and 99.9% confidence). The gray scale shows likelihood down to much lower levels, formally 10-7 . All fits at z < 7.7 are ruled out at > 99% confidence, and while fits can b e found at z 0 they are markedly worse than the high-z solutions. The b est z < 5 solution (formally at z = 0) is marked with the blue cross and requires AV 10, and is also disfavored by the lack of any host galaxy to deep limits, and the inconsistency of the required AV with the hydrogen column density measured from the X-ray afterglow. To illustrate this the b est-fit NH from the X-ray sp ectrum is converted into AV and plotted onto the contour plot as the purple lines (dashed lines show the 90% error range, and the dotted lines show the limits of the systematic error due spanning the range of gas-to-dust ratios rep orted by Schady et al. 2010). As can b e seen the AV inferred from the X-ray, and that required from the photometric redshift fit are inconsistent at low redshift, but broadly consistent with the high-z fit.


- 29 -

Fig. 8.-- Results of SED fits with a range of p ossible values for the temp oral p ower-law index, and different reddening laws. The plots encompass the canonical reddening laws for the SMC, LMC and Milky Way (which are characterized by the increasing influence of the 2175 œ bump) as well as the A law of Maiolino et al. (2004) which is approximately flat ("gray") from 1800 to 3000 œ. As can A b e seen, the assumed temp oral index has only a minimal impact on our results, and our assumption of = 0 therefore does not affect our analysis. The ma jority of GRB afterglows are b est fit with SMC-like absorption, and we therefore adopt this as our choice model (e.g. Schady et al. 2007). Other laws can produce broader allowed redshift ranges, in particular extending as low as z 6.3 at 99% confidence in the Maiolino et al. (2004) case, but all rule out low-z (< 6) scenarios.


- 30 -

Fig. 9.-- Posterior likelihood plotted on b oth a linear (upp er) and log (lower) s assuming an SMC dust law, where we have marginalized over b oth (assumed b etween -1 and +1) and AV (assumed to have a flat prior b etween 0 and 12). shaded bands show the extent of the 90% and 99% enclosed likelihood regions

cale, for the models to have a flat prior The dark and light resp ectively.


- 31 -

Fig. 10.-- Our late time HST observations of the GRB 090429B field in the optical and NIR. No host galaxy is detected in any filter, supp orting a high-redshift origin, since a host with z < 1 would b e very unlikely to b e fainter than these limits, even if dusty. At F 160W the host remains undetected, but the observations reach limits which would uncover 50% of the z > 8 candidates in the Hubble Ultra-Deep Field (UDF). Hence, the non-detection of any host is fully consistent with our high-z model, but inconsistent with any lower redshift, high extinction scenario.


- 32 -

Fig. 11.-- Solid lines show the 3 absolute magnitude limits for GRB 090429B in each of our filters F 606W (blue), F 105W (green) and F 160W (red). The inferred absolute magnitudes (AB) of a sample of GRB host galaxies (Fruchter et al. 2006), as a function of redshift (plot modified from Perley et al. 2009). The known GRB hosts with H S T observations are plotted as red p oints, and are supplemented at high redshift by the observations of GRB 050904 by Berger et al. (2006). As can b e seen, all of these lines lie significantly b elow the ma jority of GRB hosts and offer supp ort for a high-redshift origin for GRB 090429B. The blue p oints at z 8 represents the Lyman break sample of Bouwens et al. (2010). As can b e seen, the limiting magnitude for GRB 090429B lies roughly at the median of this distribution, and so the non-detection in our observations would not b e unexp ected at z 9.4