UvA-DARE (Digital Academic Repository) Delving …...Delving into the Dragons Den: The host galaxies...

UvA-DARE is a service provided by the library of the University of Amsterdam (http://dare.uva.nl)

UvA-DARE (Digital Academic Repository)

Delving into the dragons den : the host galaxies of gamma-ray bursts

Wiersema, K.

Link to publication

Citation for published version (APA):Wiersema, K. (2007). Delving into the dragons den : the host galaxies of gamma-ray bursts. Amsterdam.

General rightsIt is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s),other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).

Disclaimer/Complaints regulationsIf you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, statingyour reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Askthe Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam,The Netherlands. You will be contacted as soon as possible.

Download date: 12 Mar 2020

https://dare.uva.nl/personal/pure/en/publications/delving-into-the-dragons-den--the-host-galaxies-of-gammaray-bursts(3c9ff3a9-628e-44b0-afd4-5d78da86aeef).html

Delving into the Dragons Den:

The host galaxies of γ-ray bursts

Delving into the Dragons Den:

The host galaxies of γ-ray bursts

Academisch Proefschrift

Ter verkrijging van de graad van doctoraan de Universiteit van Amsterdamop gezag van de Rector Magnificus

prof. dr. J. W. Zwemmerten overstaan van een door het college voor promoties ingestelde

commissie, in het openbaar te verdedigen in deAgnietenkapel der Universiteit

op

donderdag 13 september 2007, te 14:00 uur

door

Klaas Wiersema

geboren te Kampen

Promotiecommissie

Promotor Prof. Dr. R. A. M. J. Wijers

Overige leden Prof. Dr. N. R. TanvirProf. Dr. N. LangerDr. A. de KoterProf. Dr. E. P. J. van den HeuvelProf. Dr. L. KaperProf. Dr. M. van der Klis

Faculteit der Natuurwetenschappen, Wiskunde en InformaticaUniversiteit van Amsterdam

Paranimfen: Dr. A. J. van der Horst; Dr. E. Rol

Cover: In a gamma-ray burst a massive star core-collapses, or a binary merges,forming a black hole and ejecting two relativistic jets. These slam into the in-terstellar medium, producing the afterglow. We observe these phenomena usingtelescopes from the ground and in space. When a satellite detects a gamma-rayburst, it rapidly sends the relevant information to the ground, where it is dis-tributed within seconds to observers. We receive this information on our mobilephones, frequently on friday evening, in the middle of the night or other incon-venient times.The yellow colour of the cover will be instantly familiar to anybody who hasvistited the Astronomical Institute, and has been chosen for maximum visibilitywhen observing at night.

Reports that say that something hasn’t happened are always interesting to me,because as we know, there are known knowns; there are things we know we

know. We also know there are known unknowns; that is to say we know thereare some things we do not know. But there are also unknown unknowns - the

ones we don’t know we don’t know.

Feb. 12, 2002, Secretary of Defense Donald Rumsfeld,Department of Defense news briefing

Soo laag in ’t stof te zyn geseeten,En’s hoogen heemels loop te meeten,

Schynt veel: Maar ’t is van veel meer nut,Den loop des leevens naa te speuren,En wat’ er Eind’ling staat te beuren,Op dat men ’t Eeuwich Onheil schut

(Seated in the dust, so low, from whereTo measure high heaven ’s course up there

Seems much: but a far better way to goIs on the course of life to ponder;

On our final lot to wonder;So as t’ avert eternal woe)

Jan Luycken, “Spiegel van het menselijk bedrijf” (1694),accompanying “The Astrologist” (Vermeer)

Contents

Contents i

1 Introduction 11.1 Flashes in the sky . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1.1 Making gamma-ray bursts . . . . . . . . . . . . . . . . . 21.1.2 Progenitors . . . . . . . . . . . . . . . . . . . . . . . . . 41.1.3 The era of Swift afterglow observing . . . . . . . . . . . 7

1.2 The dragon’s den: GRB host galaxies . . . . . . . . . . . . . . . 91.3 The progenitor - host galaxy connection . . . . . . . . . . . . . . 12

1.3.1 Position information . . . . . . . . . . . . . . . . . . . . 141.3.2 Abundances . . . . . . . . . . . . . . . . . . . . . . . . 151.3.3 ISM and CSM properties in absorption . . . . . . . . . . 191.3.4 Extinction . . . . . . . . . . . . . . . . . . . . . . . . . 201.3.5 Massive stars in GRB hosts . . . . . . . . . . . . . . . . 211.3.6 Redshift distribution . . . . . . . . . . . . . . . . . . . . 221.3.7 Masses and ages . . . . . . . . . . . . . . . . . . . . . . 231.3.8 Afterglow dynamics . . . . . . . . . . . . . . . . . . . . 241.3.9 Relative rates . . . . . . . . . . . . . . . . . . . . . . . . 251.3.10 The local Universe . . . . . . . . . . . . . . . . . . . . . 261.3.11 Looking through sunglasses . . . . . . . . . . . . . . . . 28

1.4 Outline of this thesis . . . . . . . . . . . . . . . . . . . . . . . . 29

2 Spectroscopy and multiband photometry of the afterglow of intermediateduration γ-ray burst 040924 and its host galaxy 332.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342.2 Observations and analysis . . . . . . . . . . . . . . . . . . . . . . 35

2.2.1 Afterglow optical photometry . . . . . . . . . . . . . . . 352.2.2 Host galaxy spectroscopy . . . . . . . . . . . . . . . . . 352.2.3 Radio observations . . . . . . . . . . . . . . . . . . . . . 372.2.4 Supernova search and host galaxy . . . . . . . . . . . . . 37

2.3 The afterglow . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382.4 The host galaxy . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

2.4.1 Star formation rate . . . . . . . . . . . . . . . . . . . . . 422.4.2 Metallicity . . . . . . . . . . . . . . . . . . . . . . . . . 432.4.3 [Ne III] in GRB hosts . . . . . . . . . . . . . . . . . . . . 452.4.4 Host SED . . . . . . . . . . . . . . . . . . . . . . . . . . 46

2.5 An associated supernova . . . . . . . . . . . . . . . . . . . . . . 462.6 Discussion and conclusions . . . . . . . . . . . . . . . . . . . . . 47

i

ii Contents

3 Gas and dust properties in the afterglow spectra of GRB 050730 513.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523.2 The optical afterglow spectra . . . . . . . . . . . . . . . . . . . . 52

3.2.1 Observations . . . . . . . . . . . . . . . . . . . . . . . . 523.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

3.3 The absorbed X-ray afterglow . . . . . . . . . . . . . . . . . . . 543.3.1 Observations . . . . . . . . . . . . . . . . . . . . . . . . 543.3.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

3.4 Discussion and conclusions . . . . . . . . . . . . . . . . . . . . . 563.4.1 Host galaxy properties . . . . . . . . . . . . . . . . . . . 563.4.2 The neutral hydrogen column . . . . . . . . . . . . . . . 58

4 Gamma-Ray Burst afterglows as probes of environment and blastwavephysics: absorption by host galaxy gas and dust 614.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624.2 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 644.3 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

4.3.1 Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 684.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

4.4.1 Notes on individual sources . . . . . . . . . . . . . . . . 764.4.2 Comparison with previous studies . . . . . . . . . . . . . 85

4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 864.5.1 Approaches to measuring absorption in the host galaxies . 87

4.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

5 Probing Cosmic Chemical Evolution with Gamma-Ray Bursts:GRB 060206 at z = 4.048 935.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 935.2 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 945.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 955.4 Discussion and conclusions . . . . . . . . . . . . . . . . . . . . . 97

6 The host of GRB 060206: dissecting kinematics of a massive galaxy 1016.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1016.2 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1036.3 Absorption line analysis . . . . . . . . . . . . . . . . . . . . . . 1036.4 Velocity components in the host galaxy . . . . . . . . . . . . . . 104

6.4.1 Comparison with QSO-DLAs . . . . . . . . . . . . . . . 1096.4.2 Fine structure levels . . . . . . . . . . . . . . . . . . . . 1116.4.3 Extinction along the line of sight . . . . . . . . . . . . . . 1136.4.4 The nature of the components . . . . . . . . . . . . . . . 114

6.5 The host in emission . . . . . . . . . . . . . . . . . . . . . . . . 1166.6 Intervening systems . . . . . . . . . . . . . . . . . . . . . . . . . 117

Contents iii

6.6.1 The z = 1.4787 system . . . . . . . . . . . . . . . . . . . 1186.6.2 The z = 2.2599 system . . . . . . . . . . . . . . . . . . . 1186.6.3 Identifying the absorber in emission . . . . . . . . . . . . 119

6.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

7 The nature of the dwarf starforming galaxy associated with GRB 060218 /SN 2006aj 1237.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1247.2 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1257.3 General host galaxy properties . . . . . . . . . . . . . . . . . . . 128

7.3.1 Metallicity . . . . . . . . . . . . . . . . . . . . . . . . . 1287.3.2 Relative element abundances . . . . . . . . . . . . . . . . 1327.3.3 Star formation . . . . . . . . . . . . . . . . . . . . . . . 134

7.4 Discrete velocity components in emission and absorption . . . . . 1367.4.1 Emission line profiles . . . . . . . . . . . . . . . . . . . 1367.4.2 Absorption lines in the circumburst medium . . . . . . . . 137

7.5 Secondary metallicity calibrators . . . . . . . . . . . . . . . . . . 1427.6 Massive stars and progenitors . . . . . . . . . . . . . . . . . . . . 1437.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

8 GRB 051022: physical parameters and extinction of a prototype dark burst1498.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1508.2 Observations and data reduction . . . . . . . . . . . . . . . . . . 151

8.2.1 X-ray observations . . . . . . . . . . . . . . . . . . . . . 1518.2.2 Optical and near infra-red observations . . . . . . . . . . 1528.2.3 Radio observations . . . . . . . . . . . . . . . . . . . . . 153

8.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1548.3.1 Broadband modeling . . . . . . . . . . . . . . . . . . . . 1548.3.2 The non-detection of the optical afterglow . . . . . . . . . 1588.3.3 The host galaxy of GRB 051022 . . . . . . . . . . . . . . 160

8.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1618.4.1 The burst environment . . . . . . . . . . . . . . . . . . . 1618.4.2 Energetics . . . . . . . . . . . . . . . . . . . . . . . . . . 163

8.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1648.6 Appendix : Interstellar scintillation in the radio modeling . . . . . 165

Samenvatting in het Nederlands / Summary in Dutch 167

List of publications 177

Thank you! 183

Bibliography 185

1Introduction

This thesis is about the connection between the progenitor objects of gamma-raybursts (GRBs) and the properties of their host galaxies from an observational per-spective. In the scientific chapters of this thesis several methods to probe GRB hostgalaxies are used, which I will introduce in this chapter. In the first four sectionsI will give a broad introduction to GRBs, their progenitor objects and the natureof their hosts. In the sections after that, I will give a more detailed and thoroughoverview of the different probes of GRB environments and specifically how theycan be used to derive information on the progenitors of GRBs.

1.1 Flashes in the sky

Gamma-ray bursts are intense, brief flashes of highly energetic radiation, that occurat unpredictable times, and unpredictable locations on the sky. For brief momentsthese flashes outshine the entire Universe in gamma-rays, becoming the brightestobjects in the sky at these frequencies. Their first discovery was in the 1960’s(Klebesadel et al. 1973) when the Vela satellites designed to monitor the sky withinthe framework of the Nuclear Test Ban Treaty recorded short bursts of gamma-raysthat seemed to have an extra-terrestrial origin. The unpredictability of these fleetingevents makes it very difficult to pinpoint their location on the sky, which is key tolearning whether these sources are Galactic or extra-Galactic in nature, and wouldenable follow-up in other wavelengths. For many years the origin and behaviour ofGRBs could therefore only be studied through spectral and light curve analysis ofthe gamma-ray emission. The Burst And Transient Source Experiment (BATSE) onthe Compton Gamma Ray Observatory made important breakthroughs in this field:from the large number of BATSE detected bursts (over 2000) it is very clear thatthere are in fact two distinct populations of GRBs, that can be categorised basedon duration (with the dividing value around 2 seconds) and spectral hardness, andare commonly termed long and short GRBs (Kouveliotou et al. 1993, Figure 1.1).GRB light curves are very diverse, no two light curves being exactly the same, butare all dominated by non-thermal spectra, which can be detected all the way up toGeV energies.

The isotropic sky distribution of the detected GRBs made the idea of a Galac-tic nature of GRBs less likely than an extra-Galactic one (Paczynski 1995), butdirect identification of progenitors would require better positions than gamma-ray

1

2 Introduction

detectors can provide.Rapid availability of a reasonably accurate (few arcminutes) position for the

GRB on the sky makes it possible to point ground- and space-based telescopes atthe position, and search for transient phenomena at wavelengths other than gamma-rays. The highly successful coded-mask wide field X-ray cameras onboard theItalian-Dutch satellite BeppoSAX could provide just that, leading to the discoveryof the so-called afterglows of long GRBs in 1997, at X-ray (Costa et al. 1997),optical (Van Paradijs et al. 1997) and radio wavelengths (Frail et al. 1997). Thedetection of afterglows meant that the debate on the distance scale of GRBs couldbe closed, as redshifts could be directly determined through optical spectra of af-terglows and hosts (Metzger et al. 1997). It was now clear that long GRBs occurat cosmological distances, and that they are the most energetic phenomena in theUniverse.

1.1.1 Making gamma-ray bursts

The 1997 afterglow revolution in GRBs has led to a rough picture of the origin andcause of GRBs (specifically, of the long bursts, which have duration longer than 2seconds): they are explosions of massive stars, likely accompanied by the forma-tion of a black hole (BH), according to the so-called ‘collapsar model’ (MacFadyen& Woosley 1999). In this catastrophic event, the rapidly rotating core of the star isnot able to directly fall into the black hole, but instead forms an extremely densetorus of material accreting rapidly onto the black hole. A large amount of energyis produced in this central engine, somehow giving rise to highly relativistic jetsoutflowing from the central source, likely along the rotational axis. Jet productionis something that is observed in several other types of accreting black hole systems(e.g. Active Galactic Nuclei and X-ray binaries), making this assumption quite alogical one. The highly relativistic jet pierces its way out of the stellar envelope,typically of order 1010 − 1011 cm in size. This outflow of relativistic material fromthe central engine is often assumed to be variable, consisting of several shells eachhaving a different (bulk) Lorentz factor. These shells propagate outwards and ex-pand adiabatically, faster shells slamming into slower ones, converting part of theirkinetic energy into radiation (the internal shock model). This produces the promptgamma-ray emission, typically at distances of 1012 − 1013 cm from the source (fora review see Piran 1999). As the fireball is optically thick before this phase, we aregenerally not able to study the properties of the jets at much earlier times.

Finally the relativistic outflow will collide with the medium around the star(e.g. the circumstellar or interstellar medium (CSM/ISM) or perhaps the remnantsof the stellar wind of the progenitor). This decelerates the blastwave, and producesa forward shock, giving rise to the ‘afterglow’ of X-ray to radio emission, producedat a distance of order 1015−1017 cm from the central engine (Sari 1997). This after-glow is observable for hours to weeks, i.e. much longer than the prompt emission.The afterglow radiation is likely formed by synchrotron radiation (e.g. Meszaros

1.1 Flashes in the sky 3

Figure 1.1 — Plot showing the hardness (vertical) versus duration (horizontal) of Swiftbursts (circles) (from Sakamoto et al. 2006) up until February 2006 with their 90% errorbars. Small dots show the BATSE detected bursts. The duration is expressed as T90: thetime interval during which 90% of the total observed counts have been detected. Thestart of this interval is defined by the time at which 5% of the total counts have beendetected, and the end when 95% of the total counts have been detected. It is clear thatSwift is able to detect short bursts, but preferentially detects those with softer spectra.Through target of opportunity triggering, Swift has also been successful at observingshort GRBs detected by e.g. Konus-Wind.

& Rees 1992, 1997; Wijers et al. 1998; Wijers & Galama 1999; Meszaros & Rees1999), which can be characterized by a series of connected power laws, which dis-play characteristic break frequencies and fluxes (Figure 1.2). It is believed that thisradiation comes from electrons that are accelerated to very high (relativistic) veloc-ities in the shock, though the detailed mechanism is difficult to study and requiresextensive numerical simulations.

We can measure the spectral breaks and slopes of the afterglow synchrotronspectrum as well as the power law decay slopes of the afterglow brightness bycovering the spectral energy distribution (SED) with a good time resolution, andobtain a wealth of information on the GRB phenomenon (Meszaros & Rees 1999):the jet energy, the energy distribution of the radiating electrons, the density and thestructure of the circumburst medium and the jet structure to name just a few (e.g.Wijers & Galama 1999, PhD thesis A. van der Horst 2007). At some point the jetblastwave has decelerated to a bulk Lorentz factor roughly equal to the inverse of

4 Introduction

the initial half-opening angle of the jet. At this time the sideways expansion ofthe jet becomes significant, and also the visible opening angle (due to relativisticbeaming) becomes larger than the jet opening angle. This means that the afterglowno longer looks like a spherical one, and a (smooth) achromatic break will appearin the light curve (Rhoads 1997, 1999), after which the flux decays more rapidly.Detection of this break would directly give information on the jet opening angle, butit is difficult to detect, requiring confirmation of its achromaticity (for an examplesee Curran et al. 2007b). A jet break has been confirmed in several sources, albeitmainly through optical data (perhaps the best case being GRB 990510, Harrisonet al. 1999).

Most of the observed light curves of afterglows obtained before 2005 (mainlyoptical data) can be well explained by the models above, barring a few exceptions.It is important to note here that afterglows of short GRBs proved rather more dif-ficult to find than the ones of long GRBs - their afterglows were only discoveredby the Swift satellite (Gehrels et al. 2005), but the way in which the afterglow isproduced appears, in general, similar to long bursts.

1.1.2 Progenitors

The connection of long GRBs with supernovae (SNe) was established observation-ally in 1998 (SN 1998bw, Galama et al. 1998b), when a peculiar, highly energeticsupernova of type Ib/c was found in the error box of the position of GRB 980425.The association of the two was not conclusive, as no afterglow of this burst wasseen. The clear confirmation came in 2003, when the very nearby and brightGRB 030329 showed signatures of a supernova very similar to SN 1998bw in itslate time afterglow data (Figure 1.3, Hjorth et al. 2003b). This burst did have abright afterglow, making it the best sampled GRB to date. The most recent GRB-SN was detected by Swift in 2006, showing a bright supernova (GRB 060218 /SN 2006aj).

The strongest constraints on progenitor objects of long GRBs are likely to befound by detailed spectroscopic and photometric studies of the supernovae that ac-company these GRBs (see for the case of GRB 060218 e.g. Mazzali et al. 2006;Maeda et al. 2007; Tominaga et al. 2007). In all the cases described above the su-pernovae are very energetic (‘hypernovae’), and belong to the class of SNe Ib/c,consistent with the predictions of the collapsar model: if a large (hydrogen) enve-lope had been present, the relativistic jet would have difficulty piercing through thestellar envelope, and perhaps not make it to the stage of producing a GRB. Whilemost bursts are at too high a redshift to study their SNe in detail (say z � 0.3), thecontribution of the supernova to the afterglow light curve can still be distinguishedin some cases, aided by a change in colour when the SN light starts to dominateover the afterglow and host. This leads to so-called supernova-bumps, that can beidentified for GRBs out to a redshift of z ∼ 0.9.

Recently, two very nearby GRBs (GRB 060505 and GRB 060614 at z = 0.089


Figure 1.2 — The radio to X-ray spectrum of GRB 970508 (Galama et al. 1998a) show-ing the characteristic synchrotron shape. From the values of the three break frequenciesin the spectrum (the cooling frequency νc, the peak frequency νm, and the self-absorptionfrequency νa), we can determine the physical parameters of the blast wave, such as themagnetic field strength, the explosion energy, and ambient density. Acquiring high qual-ity data over such a wide frequency range requires large international collaborations andobserving programmes at a large number of observatories.

and 0.125, respectively, Fynbo et al. 2006b) showed no sign of a supernova-bump,to limits ∼ 5 magnitudes deeper than a redshifted, K-corrected SN 1998bw. Dustextinction cannot be the cause of a lack of detected supernova, as no line of sightextinction was seen in the afterglow SEDs. However, already before the associa-tion with SNe was established for long GRBs, it was predicted that some GRBsshould be SN-less: while the direct black hole formation results in the productionof SNe with GRBs, black holes formed through fallback could produce GRBs with-out making a bright supernova (Heger et al. 2003; Fryer & Heger 2005), as in thefallback case less 56Ni may be produced and expelled. It is worth noting here thatthe long duration nature of these two bursts is debated.

In the case of long GRBs, the requirement of a large amount of angular momen-tum in the stellar core - enough to give sufficient rotational energy to the black holeat core-collapse, so via an accretion disk relativistic jets can be launched - narrowsdown the family of progenitor models considerably. It has proven to be difficult toretain the angular momentum in the star, and still produce a type Ib/c supernova,

6 Introduction

Figure 1.3 — Spectra of the afterglow of GRB 030329 obtained with the ESO Very LargeTelecope (VLT) (Hjorth et al. 2003b). As the afterglow becomes fainter, the supernovasignature is revealed, showing broad absorption and emission lines, and no hydrogen.Most GRBs have a redshift too high to observe their associated supernova spectroscop-ically.

which has lost its hydrogen envelope through e.g. a stellar wind. The difficultylies with the balance of having to lose the envelope through a wind (which takesaway angular momentum) whilst keeping the core of the star rapidly rotating. Thetheoretically predicted and observed relation between metallicity and wind mass-loss (Vink et al. 2001; Mokiem et al. 2007a) plays an important role in this: asthe metallicity increases, the mass loss through winds increases as the winds areline-driven (dominated by iron, which has very many line transitions), and whenthe mass loss increases, the star loses angular momentum.


The most recent and highly advanced numerical stellar evolution models in-cluding the effects of magnetic fields (Yoon & Langer 2005; Woosley & Heger2006) consistently point to rapidly rotating, low-metallicity massive stars as GRBprogenitors, through a rather exotic evolutionary channel involving very efficientchemical mixing within the star. This causes the whole star to become chemicallyhomogeneous if the initial rotational velocity is sufficiently high, especially at lowmetallicity (Yoon & Langer 2005). In principle this low metallicity requirementcan be observationally tested (Section 1.3.2). The models mentioned above de-scribe single star evolution, but it is also possible to form a massive star with highangular momentum in the core and a small envelope through binary channels (seee.g. Fryer & Heger 2005; Fryer et al. 2007; Cantiello et al. 2007; Van den Heuvel &Yoon 2007). These binary channels could relax the requirement of low metallicitysomewhat.

In the case of short bursts, the afterglows have only recently been discovered,and deep searches for an associated supernova (e.g. Gehrels et al. 2005; Hjorthet al. 2005; Soderberg et al. 2006a) have ruled out the presence of a supernova inall cases, making a massive star origin for short GRBs unlikely. There are sev-eral proposed models for the progenitors of short bursts, some of which have beenaround since the BATSE days. One of the most favoured is the merger of a compactobject binary: two neutron stars (NS-NS) or a NS-BH system (see Lee & Ramirez-Ruiz 2007, and references therein). These mergers can be modelled numerically,showing that they do indeed have potential as progenitors of GRBs (Rosswog et al.2003). Recently, Tanvir et al. (2005) found that the positions of a fraction of allshort bursts detected by BATSE have a significant correlation with the positions oflow-redshift galaxies, indicating that between 10 and 25 % of short GRBs originateat low redshifts (z < 0.025). The extremely bright giant flare of Soft Gamma Re-peater SGR 1806-20 that was observed on 27 December 2004 (e.g. Gaensler et al.2005; Palmer et al. 2005) provided a clear example of a possible origin for these“local” short GRBs.

1.1.3 The era of Swift afterglow observing

The advent of the Swift satellite (Gehrels et al. 2004) in November 2004 revolu-tionised the study of GRBs once more. This satellite carries three instruments: awide field of view gamma-ray imaging instrument (Burst Alert Telescope or BAT),a narrow field of view X-ray telescope (XRT) and a narrow field of view UV/opticalimager (UVOT). The BAT covers approximately one sixth of the sky, and is verysensitive in the 10 - 150 keV range, detecting and localising approximately 100bursts per year to an accuracy of ∼ 3 arcminutes (a dramatic improvement fromthe ∼20 bursts per year that its predecessor the High Energy Transient Explorer II(HETE-II) detected). When BAT detects a GRB during its monitoring of the sky(using both sky images and flux rate monitoring), the satellite’s onboard computerderives the location and the entire satellite performs a rapid, autonomous slew to

8 Introduction

the GRB position, which generally takes between 30 and 70 seconds. The XRT andUVOT telescopes will then perform a brief snapshot observation, which togetherwith the gamma-ray data are rapidly sent to the ground via the TDRSS network ofsatellites. This data helps to refine the location of the burst from the few arcminutesthat BAT provides to a few arcseconds from the XRT image to less than an arcsec-ond when a UV/optical counterpart is detected. These positions are determinedon the fly and GCN Alerts are sent within seconds. Meanwhile Swift performs anautomated sequence of observations in X-rays and UV/optical. For virtually allGRBs an X-ray afterglow is detected, though some of them can be very faint (seethe case of GRB 050509b, for which only eleven photons were detected, Gehrelset al. 2005). For most GRBs a position with an accuracy of a few arcseconds isavailable within two minutes after the burst.

This rapid slewing and rapid dissemination of coordinates enables us to alsoacquire multiwavelength data during the prompt emission and the onset of the af-terglow, which was not possible in the pre-Swift days, when alerts generally tookbetween 10 and 30 minutes to arrive. Necessary ingredients in this are opticaland infrared data. For this purpose several highly successful robotic telescopeswere developed, varying from small aperture instruments (e.g. ROTSE, TAROT,PROMPT) to large aperture robots (e.g. Liverpool Telescope, Faulkes Telescopes).All these instruments are in principle able to react to alerts within seconds to min-utes. However, larger telescopes require human interaction, which is achieved byrouting alert messages to the mobile phones of observers through SMS. The smallerror circles produced by Swift for relatively bright bursts have also enabled theuse of robotic (script) triggering of large telescopes. The most outstanding exam-ple is the ESO Very Large Telescope (VLT) Rapid-Response Mode (RRM), whichenables observers to directly trigger an 8 m VLT telescope by submitting the nameof a pre-registered Observing Block and relevant coordinates through ftp. The tele-scope slews to the source directly upon receipt of the file. The service astronomerwill be able to recognize an optical afterglow by comparison with survey data (e.g.Digitized Sky Survey (DSS), Sloan Digital Sky Survey (SDSS)), overlay the slitof a spectrograph and acquire data within minutes after the burst. The choice ofinstrument and instrument mode depends on instrument availability and brightnessof the afterglow, as determined by the service astronomer. At several visitor modetelescopes scripts are installed to alert the observer when a GRB has gone off, gen-erating observing scripts that the visitor can run to ensure very fast data acquisition,for example at the William Herschel Telescope of the ING, by our group. Late timefollow-up (more than 1 hour after the burst) is generally manually triggered. Whendata come in, rapid analysis is crucial, and time critical results are reported rapidlyin Circulars.

Despite best efforts with large aperture telescopes and rapid observations, thereis still a significant fraction of bursts for which no optical afterglow is discovered.This may be due to excessive extinction in the host, an unfortunate position of thesynchrotron break frequencies or simply intrinsically faint afterglows (for example

1.2 The dragon’s den: GRB host galaxies 9

due to a low circumburst density or high redshift). The nature of these dark bursts oreven their exact definition is not yet clear (Jakobsson et al. 2004; Rol et al. 2005b).It is important to note that in many cases where an optical afterglow has not beenfound, this is due to an unfortunate position on the sky, making it impossible toobtain early data.

Likely one of the most important breakthroughs provided by the Swift satellitewas the localisation and detection of afterglows for several short GRBs. AlthoughSwift is not detecting many short bursts, the number of localisations steadily riseswith a few per year. The energy range and sensitivity of Swift BAT is quite differentfrom the BATSE detectors. This is readily evident from the detection of bursts thatBATSE would have classified as “classic” short bursts, being spectrally hard andwith duration less than 2 seconds, that in the Swift BAT data seem to have extendedtails of soft emission lasting tens of seconds (Norris & Bonnell 2006, and referencestherein). As the properties of prompt emission may be used as probes of the centralengines, understanding the nature of these bursts is very important. In any caseit has become clear that a separation of classes of bursts purely based on (T90)duration values may not be sufficient, and futher parameters of prompt emission,afterglow or host galaxy need to be used as well. In the following I will still use theBATSE working definition of a short burst to avoid confusion.

1.2 The dragon’s den: GRB host galaxies

Deep observations performed several months after long GRBs have shown that formany a host galaxy is detected at the afterglow position (Figure 1.4). Long GRBhosts show prominent emission lines, they are sub-luminous, under-massive andblue (e.g. Djorgovski et al. 1998; Fruchter et al. 1999; Chary et al. 2002; Chris-tensen et al. 2004; Fruchter et al. 2006, PhD thesis P. Vreeswijk; all Chapters),tend to have low metallicities and are generally dwarf irregular galaxies. There is ahandful of cases where the GRB occurred in a spiral galaxy (see Figure 1.4 for anexample). In two of those cases we can analyze the properties of the host galaxywith spatially resolved spectroscopy (e.g. Hammer et al. 2006; Thone et al. 2007),and it is evident that the site where the GRB exploded has a lower metallicity andhigher excitation than the host as a whole. The luminosity function of long GRBhosts is broad (the host of GRB 060218 has 0.008L∗B but the host of GRB 020127 is5L∗B), and appears not to be significantly different from the local sample of metal-poor actively star forming emission line galaxies.

For short GRBs the picture is more complicated. There are a few establishedcases where short GRBs could be conclusively linked to host galaxies with mucholder dominant stellar populations than the average long GRB host (Berger et al.2005b; Barthelmy et al. 2005), but also clear cases of late type hosts (Hjorth et al.2005), as seen in the selection of short burst fields in Figures 1.4 and 1.5. ShortGRBs are deected at relatively low redshifts (Fox et al. 2005, but see also Berger

10 Introduction

050813

Figure 1.4 — Upper six pictures: a selection of long GRB host galaxies imaged by theHubble Space Telescope (HST; Fruchter et al. 2006), where the afterglow position isnoted by a cross. With the exception of the host of GRB 990705, which is a grand-design spiral, these are faint blue irregular galaxies. In some cases there may be signs ofinteraction (e.g. GRB 020903), possibly triggering the star formation that produced theGRB. Lower six pictures: the likely host galaxies of several short-duration GRBs showa large diversity (Hjorth et al. 2006). Circles indicate the best afterglow position and itserror, whilst GRB 051103 had no detectable afterglow, and for GRB 050906 we showonly a portion of the afterglow error circle. The probable host of 050509B is a largeelliptical galaxy.

1.2 The dragon’s den: GRB host galaxies 11

Figure 1.5 — The optical afterglow of short burst GRB 050724 as observed with WHT(Wiersema et al. 2005). Left are two R band observations obtained with the Aux-PortImager, at 0.424 days (top) and 3.405 days (bottom) after the burst. The top right imageis a PSF-matched image subtraction of the two epochs, using the ISIS code (Alard &Lupton 1998). The subtraction shows a bright point source (the white dot near themiddle) which has faded between the first and second epochs. This optical transienthad a magnitude of R ∼22.1 at first epoch, and is located near the centre of a brightelliptical galaxy, which made it necessary to use image subtraction methods to searchfor an afterglow. The connection of short bursts with galaxies of various Hubble typeshas been established over the last two years, thanks to the X-ray afterglow positions ofseveral short bursts. Note that the Galactic reddening in the direction of GRB 050724 ishigh, with AV = 2.03 magnitudes.

et al. 2006a), and there are subtle indications that there is a very local contributionto short GRBs through bright magnetar flares in nearby galaxies (out to approxi-mately 100 Mpc, Tanvir et al. 2005). As optical and X-ray afterglows of short burstsare on average fainter than those of their long cousins, it can be difficult to detectthem. This makes it difficult to be certain about the association of bright galaxiesin the errorcircle of the prompt emission with the GRB (illustrated by the case ofGRB 060912A; Levan et al. 2007).

The observation that GRB afterglows radiate synchrotron emission (Figure 1.2)is very advantageous, as this spectrum can be accurately described by a set of sim-ple power laws. The slopes of these power laws and the time dependence of theposition of the breaks in the spectrum are determined by only a handful of param-eters, describing the microphysics within the blastwave as well as the global prop-erties of the blastwave (see Section 1.1.1). In the optical and X-ray regimes this

12 Introduction

means that every deviation from power law behaviour must be due to absorption orextinction caused by the material between us and the GRB (Starling et al. 2006).As long GRBs reside in star forming regions in small galaxies, we can thereforeaccurately study the properties of the interstellar medium in dwarf galaxies at highredshift in absorption, see Figure 1.6. Routinely large numbers of metal absorptionlines are observed in afterglow spectra. Frequently large numbers of fine structurelines of sulphur, silicon, iron and oxygen are detected, which have never been ob-served in Quasi-Stellar Object (QSO or quasar) intervening absorbers (Vreeswijket al. 2004; Prochaska et al. 2007a). From these lines and the continuum shape, wecan measure a great number of parameters of the line of sight towards the GRB: themetallicity and abundance pattern, the dust depletion pattern, the dust extinctionparameters, the kinematics in the host galaxy (and even the matter close to the star,see e.g. Starling et al. 2005b; Chen et al. 2006b), the (molecular) chemistry in thehost and so on.

When the GRB afterglow has faded away, we can study the properties of thesehost galaxies in emission (at least of the brighter hosts), combining the line of sightproperties we measure from the afterglow with information from the emission lines(Chapters 6, 7). There is no other method with which we can obtain such detailedinformation about star forming regions that far away (the most distant GRB at thetime of writing this Chapter is GRB 050904 at z = 6.3, Kawai et al. 2006).

1.3 The progenitor - host galaxy connection

The prospects of ever directly identifying the progenitor of a GRB after the fact,by using previously made images of nearby galaxies (see e.g. Maund et al. 2005,for the case of SN 2005cs in M51), seem bleak due to the high average redshiftand the extreme scarcity of GRBs in the local Universe (“ordinary” core collapsesupernovae outnumber GRBs by a factor ∼ 103 − 104). However, in the case ofshort GRBs the situation is slightly different in the sense that if the compact binarymerger model is applicable to a subset of short bursts, we may expect to see theirgravitational wave signals generated upon spiral-in prior to the merger using e.g.LIGO.

As we can only analyze the supernovae for the nearest few GRBs, we needother sources of information about the progenitor population, evolution and theenvironment in which GRBs occur. Fortunately, the properties of GRB host galax-ies (single sources and large samples) may be used as pointers to the nature andevolution of the progenitor object(s). Properties observed in emission and (lineof sight) absorption of both GRB hosts and the circumstellar environment aroundGRBs can be directly correlated to progenitor properties, but often a more statisti-cal approach is necessary. Many of the different observational characteristics andtheir relevance for constraining progenitor evolution models are coupled together,and we may need to approach the problem from all possible sides. In the following

1.3 The progenitor - host galaxy connection 13

Figure 1.6 — Spectra obtained of the afterglow of GRB 060206 (Chapters 5, 6) withthe WHT using the Intermediate dispersion Spectrograph and Imaging System (ISIS).This burst had a redshift of z = 4.047, but the afterglow nevertheless reached a verybright peak magnitude of R ∼ 16. Spectra with WHT could already be obtained at∼1.5 hours after the burst. The ISIS spectrograph has a red and a blue arm (split by adichroic), whose spectra are shown at the top and bottom, respectively. The continuumis normalised and the error spectrum is plotted in grey. Several of the metal lines used toderive the metallicity in Chapter 5 are labelled, enabling us to measure the metallicity at[S/H] = -0.84 ± 0.10. The blue spectrum clearly shows the Lyman forest and the broadLy β, Ly γ and Ly δ lines of the host.

subsections I aim to give a summary of the observational properties of GRB hostsand afterglows and their use in our understanding of progenitor objects, referringto the appropriate chapters in this thesis.

I will start by showing the possibilities that high-resolution space based imag-ing offers, in Section 1.3.1. More quantitative information can be found throughspectroscopy, providing abundances and a detailed look on the ISM in the host(Sections 1.3.2, 1.3.3) but also information required to understand the dust proper-ties of GRB hosts (Section 1.3.4). This is in turn important to make corrections fordust depletion and reddening, needed to correctly calculate abundances. As longGRBs are formed by massive stars, a detailed knowledge of the massive star popu-

14 Introduction

lation of GRB hosts is required (Section 1.3.5). These massive stars live for a shorttime, and therefore trace star formation well, unlike short GRBs (Section 1.3.6).This can be further explored through the study of individual host galaxies, where inlow redshift cases the age of the dominant stellar population can be measured (Sec-tion 1.3.7). The afterglows themselves also provide clues to the properties of theexploding star, particularly the properties of its stellar wind (Section 1.3.8). Thesestudies can in some cases be combined with the properties we measure from spec-troscopy of afterglows. In the more local universe we are able to measure propertiesof individual stars, and perhaps analogues of GRB progenitors or their environmentin more detail (Section 1.3.10). This may help us considerably in understanding thedata of high redshift GRBs. I close the list of methods by naming one that couldnot be explored in my PhD period: polarimetry of afterglows (Section 1.3.11). Po-larimetry provides information that can not be found through spectroscopy or SEDmonitoring (e.g. Rol et al. 2003, 2000; Greiner et al. 2003; PhD thesis E. Rol 2004).Unfortunately obtaining the required datasets is very difficult. Most of the meth-ods described below to derive constraints to GRB progenitor models are connectedwith one another, so the list is quite long. The best constraints on progenitors comefrom those cases where multiple methods can be combined for individual sourcesor large samples.

1.3.1 Position information

A direct diagnostic for the type of stars involved in making GRBs is the location ofthe GRBs within their hosts (e.g. Bloom et al. 2002). To get reliable measurementsalso for high redshift bursts it is necessary to use space-based instrumentation suchas the Hubble Space Telescope. Fruchter et al. (2006) have compared locationsand host morphologies of a sample of GRBs with the sample of core-collapse su-pernovae found in the “Hubble Higher z Supernova Search” programme. Whileboth objects have similar progenitors (high mass stars), their locations differ sig-nificantly: long GRBs appear more concentrated on the (UV) brightest regions oftheir hosts. The host morphologies also differ in the sense that long GRB hosts areon average fainter and have a more irregular morphology. The authors concludethat this strengthens the association of GRBs with the highest mass stars and low-metallicity environments. The case of the very lowest redshift (z = 0.0085) spiralhost of GRB 980425 makes this view less straightforward: the GRB went off in avery small H II region, with an extremely big and bright H II region 0.8 kpc awayfrom the GRB position, which fully dominates the massive star population and UVradiation of the host (Hammer et al. 2006). At just slightly higher redshift the GRBwould have been linked to this region, rather than to its real environment. Derivingpositional information requires the detection of an optical transient (to get a posi-tion with accuracy � 0.2 arcsec) and high resolution host imaging (as most hostsare comparable in size to the typical seeing at ground-based observatories), makingspace based observations a necessity to avoid observational biases.


If the majority of short GRBs come from a compact binary merger, the kick thatsuch a binary gets from the supernovae producing the compact objects may causethe system to leave its host galaxy in the time it takes to merge, which is ∼ 106−1010

years. In this case the association of the galaxy with the GRB will become moredifficult to infer. However, it has been predicted that some NS-NS binaries receivea much lower kick velocity than previously expected (Van den Heuvel 2004, 2005,2006, 2007; Dewi et al. 2005), showing that this assumption is uncertain, and betterstatistics on NS-NS binaries are required.

1.3.2 Abundances

As long GRBs are thought to be formed by low-metallicity massive stars (at leastin the single star progenitor scenario, see Section 1.1.2) and short bursts do notshare this bias, metallicity determinations of a large sample of bursts may showthis preference, though perhaps complicated by the fact that the average galaxymetallicity drops at increasing redshift.

Abundances can generally be determined using two methods:(i) by taking afterglow spectra to measure column densities of hydrogen and(preferably non-depleted) metals along the line of sight towards GRBs (Chapters 3,5, 6),(ii) by taking very deep host galaxy spectra to find the abundances in the dominantH II region(s) through nebular and (temperature sensitive) auroral emission lines.

Afterglow spectroscopy

The first method requires the GRB to be at a redshift sufficient to detect the Lyα lineat 1215 Å (z ∼ 2.3 or larger, to sufficiently probe the wings of the Ly α line) andthe afterglow to be bright enough to use medium or high resolution spectroscopyto accurately fit the metal line profiles to determine column densities. The latteris a hard requirement to meet, as it requires a rapid response with very large in-struments. The average redshift of Swift bursts is high enough (Jakobsson et al.2006b) that the sample of bursts with afterglow line of sight spectroscopy includ-ing the Lyα transition grows rapidly. Most GRB afterglow spectra show a largehydrogen column density, causing the Lyα line to develop very strong dampingwings (Chapters 3, 5), which in turn makes it easy to determine the hydrogen col-umn density even in the case of low resolution spectroscopy. For a large fraction ofbursts the hydrogen column density is in the range of Damped Lyman Absorbers(log NH I ≥ 19 cm−2). The distribution of these hydrogen column densities is some-what puzzling, as it is not fully consistent with that expected from GRBs randomlyplaced in large molecular clouds (e.g. Jakobsson et al. 2006a). In fact there areseveral cases where the measured hydrogen column density seems to be higherthan appears possible from single clouds (Schaye 2001, Watson et al. 2006 findlog NH I = 22.6 for GRB 050401). It is likely that the hydrogen is not all co-located

16 Introduction

with the burst, but may be further out in the galaxy. The same holds for the metallines, which are frequently split into several velocity components (Chapter 5, 6).Some of these velocity components are due to kinematics within the host, othersmay arise from galactic outflows or even intervening galaxies (see Section 1.3.3).To derive abundances in the host we simply add the columns of the lines togetherand divide by the total hydrogen column. In many cases several elemental abun-dances can be measured this way, which in turn provides a depletion measure (e.g.Savaglio et al. 2003). To derive the “metallicity” of the host we use the elementwhich is least likely to be significantly depleted onto dust grains. Depending onthe redshift and spectral range of the instrument this is usually sulphur or zinc(Chapters 3, 5; Savaglio 2006). This method can be compared with QSO absorbers(Chapter 5), although QSO absorbers probe other regions of the intervening galax-ies than GRBs probe in their hosts (e.g. Vreeswijk et al. 2004; Prochter et al. 2006;Prochaska et al. 2006b; Fynbo et al. 2006a): GRBs are found near regions of starformation, and show in their spectra numerous fine-structure lines that are commonin star forming regions, but are never found in QSO absorbers.

Emission line spectroscopy

This method is used when the afterglow has faded away and the host galaxy canbe studied with optical spectroscopy (Chapters 2,7). The main advantage over thefirst method is that it provides information about the dominant H II regions (andthat is where we may expect to find the long GRBs, see Section 1.3.1, but seealso Hammer et al. 2006, and Thone et al. 2007 for the cases of GRB 980425 andGRB 060505), rather than just the line of sight. The resulting measurements canbe compared with catalogues of local galaxy spectra, such as the SDSS, and deepsurvey spectroscopy, such as the Gemini Deep Deep Survey (Savaglio 2006). Inthe case of long bursts the host spectra are dominated by bright emission linesof [O II] λ3727, [O III] λλ4959,5007 and the Hα and H β Balmer lines. As hostgalaxies are faint, quite often these are the only lines that are detectable in a 3 hourspectrum with the VLT Focal Reducer and low dispersion Spectrograph (FORS)in cases where the redshift is sufficiently low ([O III] λ4959 is redshifted out ofthe optical window beyond z � 0.9). When the host is brighter than R∼25 and theredshift is sufficiently low, the lines of [Ne III] λ3869 and higher order Balmer linesshow up.

In rare cases the hosts are brighter than R ∼ 20 − 22 (GRBs 980425, 031203,020903 and 060218), so we can achieve enough signal to noise to directly detectthe temperature sensitive auroral line [O III] λ4363. The strength of this line com-pared with that of [O III] λλ4959,5007 gives a direct measurement of the electrontemperature (Te) in the hot parts of the H II region, where [O III] is the dominantoxygen ion. The electron density can be calculated from the [O II] doublet, andthe electron temperature in the lower temperature regions from the singly ionizedline fluxes, which provides enough information to calculate the oxygen abundance


to high accuracy (see Chapter 7 for detailed calculations). The abundances of ni-trogen, neon and sulphur (and sometimes argon) can then be calculated using thedensities and temperatures found for oxygen.

In most cases, however, the GRB hosts are too faint to detect [O III] λ4363within a reasonable exposure time, and secondary methods need to be used. Guidedby nebular theory, several bright line ratios (involving [O II] λ3727, [O III] λ5007,Hβ, Hβ, [N II] λ6583) etc.) are empirically found to correlate with metallicity. Thesituation is slightly complicated, as both metallicity and ionisation parameter playa role. The most commonly used is the so-called R23 method (Pagel et al. 1979),which uses the brightest lines: R23 = [O II] λ3727 + [O III] λλ4959,5007 / Hβ. R23

has a double-valued correlation with the oxygen abundance, which can be brokenby constraining the other free parameter, the ionisation parameter, through otherlines. This diagnostic has its roots in nebular theory, and its calibration to Te de-termined metallicities is complicated, showing that this is probably not the best(though it is the easiest!) secondary calibrator to use (Bresolin et al. 2004). Othersecondary calibrators often include the [N II] lines, which are used because thenitrogen abundance as a function of metallicity stays constant at low metallicity(Chapter 7), making it a useful indicator of a possible low metallicity environment.Their use for GRB hosts is limited, as the [N II] lines are redshifted out of the usefuloptical window when z � 0.45.

Luminosity - metallicity

For hosts without spectroscopic metallicities we may still be able to get a rough ideaof their metallicity through imaging: there is a clear correlation between galaxy lu-minosity and metallicity (e.g. Garnett 2002; Lamareille et al. 2004), which is likelycaused by a more fundamental correlation between the mass and the metallicity ofgalaxies (the stronger gravity of the more massive galaxies helps to keep the metalsin the galaxy). The behaviour of the mass- and luminosity-metallicity (MZ, LZ) re-lations is well studied in the local Universe (e.g. Lamareille et al. 2004), but it canbe extended to a larger range of redshifts (z � 1, Kobulnicky et al. 2003; Kobul-nicky & Kewley 2004), where the metallicities of galaxies are determined throughemission line spectroscopy. This is particularly important as at around z ∼ 1 thestar formation rate of galaxies is on average high. There is a dependence on red-shift of the LZ and MZ relations, found in several different datasets by independentgroups (e.g. Savaglio et al. 2005), indicating that the galaxies become more metal-poor for their luminosity as redshift increases. However, it becomes increasinglyhard to verify the behaviour of the LZ/MZ relations at high redshift, as only thevery brightest of sources in the samples can be used to acquire spectra, introduc-ing complicated selection biases. It appears, however, that the trend of a lowermetallicity for a given luminosity with increasing redshift continues to z ∼ 3 (e.g.Shapley et al. 2004; Moller et al. 2004). This could be of benefit to analyzing thesample of GRB hosts with just one detected emission line, or with only photometric

18 Introduction

redshifts, though the scatter in the LZ relation is very large. Whether or not GRBhosts follow the local LZ/MZ relations is subject to debate (Savaglio 2006; Kewleyet al. 2007).

Results

Through all the methods described above we come to the following conclusions:long GRB hosts at redshifts z < 1 have on average significantly lower metallicitythan the local galaxy population (i.e. compared to SDSS; Savaglio 2006); afterglowspectroscopy shows that line of sight metallicities are on average slightly higherthan those of QSO absorbers (as QSOs probe on average less dense environmentsthan GRBs), but are all significantly sub-Solar (Chapter 5), and extremely low val-ues can be found (Chapter 3). The sample of host galaxies with a Te determinedmetallicity all show low to very low metallicity environments for GRBs (Chap-ter 7), consistent with the GRB production scenarios involving Wolf-Rayet (WR)stars with rapidly rotating cores and small envelopes, but also not inconsistent withseveral binary models (e.g. Fryer et al. 2007). When abundance patterns can beestablished in emission through Te methods, a possibly significant nitrogen over-abundance with respect to the oxygen abundance is visible for three of the four hostgalaxies (Chapter 7). The case of GRB 980425, where spectra could be obtainedat three different locations in the host, shows that this nitrogen overabundance isonly present at the site where the GRB occurred, suggesting that there is a possiblerelation between the overabundance and the GRB. To verify this, more high signalto noise spectra need to be obtained for host galaxies, which requires more brighthost identifications.

The metallicities of short burst host galaxies are much harder to determine:many of these hosts do not show emission lines at all, and afterglow spectra havenot yet been successfully obtained. Modelling of their spectra and broadband SEDsgives some constraints to their metallicities, but uncertainties are large (Prochaskaet al. 2006a).

Caveats

While Te methods provide highly accurate values for the abundance of several el-ements, unfortunately the nebular lines of iron are too faint to detect in most GRBhosts. As iron is the main driver in the stellar winds, direct measurements wouldbe very helpful. Through afterglow spectra iron abundances can be measured ina few cases, but iron is strongly depleted onto dust, easily ionised to undetectablespecies, and iron lines are quickly saturated, making abundance determination verycomplicated.

I also note that in all these methods reddening plays an important role. In thecase of galaxy emission line spectra, accurate correction for reddening in our Milky


Way, reddening in the H II region in the host and the underlying stellar Balmer ab-sorption in the host (the latter two determined through fitting of the Balmer decre-ment) is necessary. Long GRB hosts generally have young populations, so wemay expect the effects of stellar Balmer absorption to be small on average (Chapter7, also see Thone et al. 2007). Short GRB hosts usually have much older pop-ulations, and underlying Balmer extinction is clearly present in at least one case(GRB 050709; Covino et al. 2006), requiring high signal to noise and good spec-tral resolution to accurately derive metallicities. In the case of line of sight studiesthe way in which elements deplete onto dust plays a role in calculating the abun-dance that stars may have had when formed. For these reasons it is important tounderstand dust properties in hosts (Section 1.3.4).

1.3.3 ISM and CSM properties in absorption

An underlying assumption when deriving line of sight abundances is that the ab-sorption lines do not vary significantly as a function of time and that the derivedabundances and ISM properties are somewhat representative of the conditions nearthe progenitor. The real situation may not be so simple: in the majority of opticalafterglow spectra, discrete velocity components are visible with differing physicalconditions (Chapters 5, 6, 7). In most cases these components can be identified withbound kinematics of the host galaxy (i.e. rotation) or outflows from the host (Chap-ters 5, 6). Also, intervening absorbers (galaxies) are commonly found in thesespectra, creating absorption lines with large velocity offsets. In a few rare cases,however, lines are found at several discrete velocities that closely resemble the ve-locities expected from interactions of a progenitor WR star wind with the slowerwind components produced in earlier phases of the massive progenitor’s evolution(Starling et al. 2005b; Van Marle et al. 2005, 2007). There are some caveats in theidentification of absorption systems at high velocities (a few thousand km s−1) assignatures of the circumstellar medium (CSM, see e.g. Chen et al. 2006b). TheGRB and afterglow produce hefty amounts of highly-energetic radiation, and itis not directly evident that it is possible for the measured ionic species (C IV andSi IV) to escape ionisation (but see Starling et al. 2005b). If indeed these absorptionfeatures are produced by the progenitor wind remnants, strong constraints on theprogenitor can be obtained (Van Marle et al. 2007).

A strong influence of the GRB on its environment is almost inevitable, whichhas various effects on the derived line of sight values. The most revealing ex-ample regarding optical ISM lines has been obtained by Vreeswijk et al. (2007).For the relatively nearby GRB 060418 a series of VLT UV-Visual Echelle Spectro-graph (UVES) spectra were obtained in Rapid Response Mode (RRM), beginningvery rapidly at 11 minutes after the burst. Apart from common ISM lines, a largenumber of unusual lines were detected that could be identified with meta-stabletransitions. Decay of these transitions to stable states was observed in the spectra.From these lines it was evident that UV pumping is the mechanism exciting these

20 Introduction

states, and that the distance of the ISM cloud from the GRB is ∼ 1.7 kpc. No sig-nificant amount of Fe II or Ni II is observed at distances smaller than this, perhapsbecause of ionisation by the GRB and early afterglow radiation. Hydrogen has aneven lower ionisation potential than these species, implying that the line of sightmetallicities that we measure are generally not representative of gas close to theburst, but further out in the host. The effects of GRBs on their surroundings canalso be seen in X-rays (Section 3), where changes in the absorbing columns areseen following X-ray flares (though other reasons have been suggested). Though itseems a done deal that GRBs influence line ratios in their line of sight, whether ornot this influences the emission line diagnostics is very much an open question: in amulti-year spectroscopy campaign no sign of the expected emission line variabilityin GRB host spectra has been seen (Kupcu Yoldas et al. 2006).

The assumption that the metallicity local to the GRB is equal or close to the af-terglow derived value is hard to either defend or reject, as more detailed 3D knowl-edge of the host is needed, clearly demonstrated by the cases of GRBs 980425 and060505 (Hammer et al. 2006; Thone et al. 2007).

1.3.4 Extinction

The issue of the role of dust extinction in the lines of sight towards long GRBsis still very much an open one. While clear signs of dust depletion are seen inseveral afterglow spectra, the AV values that are predicted from these depletionmeasures are generally much higher than the observed ones, that can be found bymeasuring deviations from the power law shape of the continuum (Savaglio & Fall2004). Homogeneous analysis of 10 BeppoSAX long GRB afterglows for X-rayand optical extinction including the effects of metallicity show different dust togas ratios than Galactic and Magellanic Cloud values (Chapter 4; Starling et al.2006). Comparison of neutral hydrogen columns and metallicities of afterglowlines of sight with X-ray extinction values (Watson et al. 2007) showed that theabsorption probed by these two wavelength regimes is generally located at differentpositions in the host. In all these cases there may be significant biases against burstswith low apparent magnitudes, preventing optical spectroscopy, which are hard toquantify. The origin of dark bursts (i.e. the cause of their optical faintness) plays animportant role in these issues. Currently accurate X-ray positions can be obtainedfor all GRBs through Swift, allowing identification of their hosts and measurementof the X-ray extinction. In the cases of dark bursts it can be difficult to obtainline of sight optical extinction measurements, but where limits can be set throughmultiwavelength observations at early times, it is possible to separate bursts withlow optical restframe luminosity from highly extincted sources. This enables usto sample the distribution function of optical extinction in GRB lines of sight in amore unbiased fashion, allowing us to establish in a more quantitative way whichenvironments are probed by these GRBs, and with which local analogies (i.e. the30 Dor region in the Large Magellanic Cloud, or perhaps more dense molecular


clouds) we need to compare GRB line of sight information.Important input to solve these problems can come from the detection of

molecules such as H2 or CO. There are various possible mechanisms to produceanomalous dust to gas ratios, and different AV values from those obtained via deple-tion fitting, and some of these have implications for the presence of molecules in theline of sight. In Chapter 5 we show our tentative detection of hydrogen moleculesin GRB 060206, but note that no other GRB afterglow has shown signs of molecu-lar lines (Tumlinson et al. 2007). Interestingly, GRB 060206 has the most massivehost with the highest metallicity at this redshift. In low redshift afterglow spectrait may be possible to detect diffuse interstellar bands (for an extragalactic examplein a supernova spectrum see Sollerman et al. 2005a), which give constraints on thecarbon chemistry in the lines of sight (PhD thesis N. Cox 2006). These bands canonly be detected at low redshift (the deepest bands are in the optical) and at rela-tively high signal to noise, and may be much easier to find in sightlines with highAV , which hinders afterglow detection.

Accurate measurement of extinction in short GRB hosts, via optical to X-rayafterglow broadband data, has not been very successful so far due to the lack ofbright optical afterglows with sufficient broadband coverage. When such datasetsare available, this technique will be a useful tool for probing the short burst envi-ronments.

1.3.5 Massive stars in GRB hosts

The sub-L∗ and blue nature of long GRB hosts already suggests the presence ofactive star formation, but spectroscopically the signs are obvious. The Hα line is agood tracer of massive star formation, as it traces the UV flux. Generally this lineis very bright in long GRB host galaxy spectra (see Chapter 7), indicating a highstar formation rate, consistent with the idea that long GRBs are caused by massivestars. The opposite is the case for short GRB hosts: while some of them showcelar signs of star formation (e.g. Hjorth et al. 2005; Covino et al. 2006), thereare several short GRB hosts where no star formation has occurred recently (e.g.the case of GRB 050724, Berger et al. 2005b). A further clue to the presence ofmassive stars comes from the frequent detection of bright lines of higher ionisationspecies, such as [Ne III], requiring the presence of O stars in the hosts. A highratio of [Ne III] / [O II] may be used as an estimator of the ionisation parameter(and therefore may serve as a rough indicator of metallicity as well), and can beeasily compared with local galaxies. In Chapter 2, we show that for GRB hostsup to z ∼ 1 this ratio is similar to the sample of SDSS emission line galaxieswith accurate metallicities. This indicates that the correlations derived from SDSSsamples are also valid for GRB hosts.

A valuable diagnostic of the massive star population is the He II λ4686 line,which appears as a broad line and/or as a nebular line in some H II galaxies(Schaerer et al. 1999), and is a direct sign of (unusually) high excitation levels

22 Introduction

caused by the presence of WR stars (especially WC- and WO-type WR stars).Together with, amongst others, N III λ4640, C IV λ4658, [Fe III] λ4658 and[Ar IV] λ4711 lines, this line forms the so-called blue WR bump (from ∼4650 –4700 Å) in low resolution spectra, which is often accompanied by a C IV λ5808 line(the red bump). In a recent very deep spectroscopic search in the brightest nearbyGRB host galaxies, Hammer et al. (2006) have detected the He II line and resolvedseveral other components of a WR bump in the spectra of the hosts of GRB 980425,020903 and 031203 (Figure 1.7). From the relative intensities of the H β and He IIλ4686 lines or WR bump (measured flux ratios H β /He II λ4686 generally rangefrom ∼0.01-0.02) one can estimate the ratio of WR to O stars. The required signalto noise is very high, so that in these spectra also the auroral [O III] λ4363 canbe detected, allowing accurate metallicity measurements as well. In the hosts ofGRBs 980425 and 020903, Hammer et al. (2006) find values of WR /O ∼ 0.05 and∼ 0.14−0.2, respectively (for the LMC this ratio is ∼0.04, for the SMC ∼0.02). Wenote that no metallicity correction has been applied to these values, which wouldincrease the number of WR stars from the WR bump flux, as the studied hosts havesub-Galactic metallicity (Crowther & Hadfield 2006). The value for GRB 020903is much higher than expected (Chapter 7).

There is a prediction that the fast rotational mixing thought to be present inGRB progenitors may lead to high WR /O ratios (higher than expected when as-suming a continuous star formation history and taking into account the metallicity),as stars keep rotating rapidly at low metallicity due to reduced mass loss rates (Yoonet al. 2006). This could provide us with a new tool to test current GRB progenitorevolution models (Chapter 7), and requires accurate measurement of metallicity,the star formation history (starburst or continuous) and the WR /O star ratio (fromthe blue WR “bump”), as WR /O is predicted to depend on metallicity in the chem-ically homogeneous scenario. Particularly important is the host of GRB 060218, asits very low metallicity would dramatically extend the metallicity range of WR /Omeasurements. I further discuss star formation histories in Section 1.3.7.

1.3.6 Redshift distribution

Long GRBs are linked to massive stars, and are therefore linked to the star forma-tion rate of the Universe, and GRB production is expected to follow the global starformation rate convolved with the cosmic metallicity evolution (Langer & Nor-man 2006). Several observational biases complicate this scenario. It is startingto become clear that there is a group of long GRBs which can be classified aslow-luminosity bursts, having low integrated gamma-ray fluxes, and often faint af-terglows. These sources can only be detected in the local Universe. At least twoof these bursts (GRB 060218 and GRB 980425) may not even have a bona fide af-terglow. If these had been at a redshift larger than ∼ 0.8, a redshift measurementwould have been impossible because the SN would not have been detected, and noX-ray position would have been available to find a host galaxy. A further obser-


Figure 1.7 — On the left the detection of the signatures of WR stars in the host galax-ies of GRB 980425 and GRB 020903 (from Hammer et al. 2006). The right panelis a zoomed-in view of the WR bump in a bright starforming region in the host ofGRB 980425, with the lines that make up the bump.

vational problem is the existence of a so-called redshift desert: when a GRB has aredshift between ∼1.4 and ∼ 2.1 the redshift is more difficult to determine as the[O II] host galaxy emission line is redshifted out of the optical window, and the Lyαbreak is not yet detectable, so one has to rely on much fainter lines. Spectrographslike VLT X-shooter may be able to see host lines in the near-infrared, producingmore redshifts in this desert. As these redshifts are close to the expected redshift ofthe peak of cosmic star formation, it is vital to detect many bursts in this range. It isof importance to note that several families of progenitors involving binaries have arather different redshift distribution than massive single star progenitors (Fryer et al.2007), but a substantially larger sample of (unbiased) GRB redshifts is required tobe able to quantify this observationally.

Short GRBs have more complicated predicted redshift distributions, and theobserved redshift distribution is more complicated to interpret. A subsample ofshort GRBs comes from the local Universe (possibly magnetar flares), and binarymergers are hardly connected at all to the star formation history.

1.3.7 Masses and ages

The ages of the dominant stellar populations in GRB hosts and particularly the ageof the region where the GRB originated can give limits on the age and mass ofthe progenitor. Host galaxy SEDs provide the main tool with which to estimatethese numbers: when enough data are present the host fluxes can be fit using var-

24 Introduction

ious galaxy templates, fitting for redshift, star formation rate, extinction and age(Figure 1.8). In the case where observations in the radio and particularly submil-limetre (submm) are present, the broadband SED can be fit (Castro Ceron et al.2006). This also provides much stricter values for the star formation that is highlyextincted in the optical, but produces significant radio emission. The specific starformation rates (SFR per luminosity or mass) are high (as expected for galaxiesactively forming stars), but are generally not inconsistent with continuous star for-mation histories. Data in the radio, submm and far-infrared are crucial for thesestudies, but often not available. The advent of the new radio array LOFAR mayhelp: at radio wavelengths star formation produces a power law spectrum givinghigher flux with increasing wavelength. The extremely low frequencies at whichLOFAR observes (10 - 240 MHz) are well suited to obtain very good limits on thestar formation rate for all GRB hosts within z ∼ 1 − 2.

Most short GRB hosts show SEDs characterized by old populations, and no orlittle current star formation (though statistics are small). The lifetime of a binarysystem which evolves into a double neutron star or neutron star black hole binaryplus the added merger timescale is significantly longer than the lifetime of a col-lapsar progenitor object. In a sense this is also suggested by the observations, as allhosts of short bursts show distinct old populations, while most long GRB hosts arefully dominated by young stars.

1.3.8 Afterglow dynamics

The kinematic evolution of the relativistic blastwave appears, at least for the lateafterglow, well understood (see e.g. PhD thesis A. van der Horst 2007). However,to derive constraining parameters, the afterglows of GRBs have to be sampled withrelatively good time resolution and using a large range of wavelengths. When op-tical and X-rays are combined, values for the particle number density profile ofthe medium in which the blastwave propagates (parametrized as n(r) ∝ r−k) canbe derived (Starling et al. 2007) from the spectral slopes, and from the temporaldecay. In Figure 1.9 the values derived for k for ten long GRBs are shown. Thecase k = 2 would be a freeflowing stellar wind, and k = 0 a homogeneous medium.It is clear that of the five cases with small errors, four are consistent with a stellarwind density profile, and one with a homogeneous medium. This is important interms of the progenitor models: the density structure through which the blastwavepropagates is determined by the characteristics of the stellar wind and the ambientmedium star, as the wind of the progenitor star propagates into the surroundingmedium (Pe’er & Wijers 2006). Together with the metallicity and ISM densities(determined from afterglow spectroscopy, but see Section 1.3.3) and the density ofthe medium in which the afterglow propagates as derived from high quality SEDmodelling, it may be possible to provide constraints on the wind behaviour of theprogenitor. A further important parameter is the source size of the radiating surfaceof the blastwave. It is clear from high-redshift surveys that there is a very large pop-


Figure 1.8 — An example of an optical to near-infrared host galaxy SED (BVRIJHK)and a fit using the HyperZ code. This is the host galaxy of GRB 051022, a dark burstwith no detected optical afterglow (Chapter 8). This host galaxy has a redshift of z = 0.8,the dominant stellar population is young with ∼ 20 Myr, and the host is best fit with anirregular galaxy template.

ulation of metal-rich, highly compact, C IV absorbing pockets of gas present in theintergalactic medium (Schaye et al. 2007). If one or more of these small absorbersis present in the line of sight, this could influence abundance measurements.

If a similar broadband analysis could be done for short bursts, this may provehighly constraining. A measurement of k could possibly differentiate between dif-ferent progenitor models, at the very least between magnetar flares and binarymergers. A further defining characteristic could be the circumburst density. Sev-eral authors (e.g. Soderberg et al. 2006a) claim that the circumburst density in shortburst afterglows is several orders of magnitude below that of long bursts, but this isbased on small amounts of data, and far from secure.

1.3.9 Relative rates

A key factor in modelling GRB progenitors is to determine how rare they really are:the more rare GRBs are in the Universe, the more exotic the progenitor models, toprevent making too many of them. In this argument it is of vital importance to havea good figure for the ratio of long GRBs to “normal” type Ib/c core-collapse su-pernovae. There are many uncertain factors in the determination of this ratio, mostimportantly in the determination of the redshift distribution of SNe and GRBs, and

26 Introduction

Figure 1.9 — For ten long GRBs detected with BeppoSAX the X-ray to optical afterglowSEDs were fit at a common epoch in count space, including the effects of metallicity andincorporating various local extinction laws (see Chapter 4; Starling et al. 2007, 2006).Optical and X-ray extinction values are measured, and from the spectral and temporalslopes we derive values for the density structure, k, of the medium into which the blast-wave propagates, where we parametrize the density structure in which the blastwavepropagates as n(r) ∝ r−k (see Section 1.3.8). The GRBs are plotted in date order alongthe horizontal axis.

the determination of the fraction of events that we miss due to e.g. beaming of theGRB jet away from us, excessive reddening in the host, etcetera. Work by Fruchteret al. (2006), for example, has shown that there is a significant difference in the lo-cations of core-collapse supernovae and those of long GRBs within their hosts (seeSection 1.3.1), and indeed in the global host properties. This may reflect a differ-ent metallicity environment bias in the formation of these two classes of transients.As metallicity is a vital input parameter in GRB progenitor modelling (see Section1.1.1), a careful survey of metallicities of type Ib/c SNe host galaxies (Modjaz et al.2007) may be able to give the ratio of SNe Ib/c versus long GRBs as a function ofmetallicity. A combination of such a survey with the supernova parameters (thenickel production, outflow velocities, etc.) would produce meaningful constraintson models.

1.3.10 The local Universe

A large number of input parameters of progenitor models need to come from ob-servations of local sources. It would be impossible to list all of these here, but toname just a few:

* Measuring rotation rates of WR stars is extremely complicated, hindered


Figure 1.10 — One of the most actively star forming regions in the Galactic neigh-bourhood is the 30 Doradus region in the Large Magellanic Cloud, better known as theTarantula nebula, of which a small part is shown here. These images I made using theDanish 1.5m telescope using the DFOSC instrument, taking 2 x 1 minute exposures innarrow-band filters centred around the emission lines of Hα (top left), [O III] (top right)and [S II] (bottom left). In terms of age, stellar content, excitation and star formationrate this region is expected to be quite similar to long GRB hosts. The images clearlyshow filamentary structure with different temperatures and densities (visible by differ-ences between the images taken with different filters), over large as well as small spatialscales. Star forming regions in GRB hosts may have similar structure, complicatinginterpretation of measurements of host properties through line of sight afterglow spec-troscopy. 30 Dor is one of the largest H II regions in the local Universe. If it were asnear as the Orion star forming complex M42, it would take up 30 degrees of sky andcast dark shadows, being several times brighter than Venus.

by their strong, opaque winds. The rotational properties and the metallicitydependence of massive star winds has recently been analyzed for a largesample of bright young O and early B stars in the LMC and SMC, usingVLT FLAMES and accurate parameter fitting methods (Mokiem et al. 2006,2007b). Interestingly the initial rotational velocity distribution derived doesnot disagree with the rotational velocity needed for a long GRB progenitor.

* The destruction of dust is frequently mentioned as a method to explain the“anomalous” dust to gas ratios in GRB afterglow lines of sight. This wouldlikely mean that there should be variation in line ratios like Mg/S (easilydepleted elements versus not-easily depleted elements). This is however not

28 Introduction

observed. A useful local analogy would be to look at dust destruction in localsupernovae and variable energetic jet sources interacting with their ISM (i.e.X-ray binaries like GRO J1655-40).

* It is clear that the dust to gas ratio measured in different GRB lines of sightdiffers from the value we expect from their best fitting extinction curve. Re-cently, Padoan et al. (2006) analysed the dust to gas ratio towards the Taurusmolecular cloud complex, using CO measurements and near-infrared deepimaging. While averages over a large spatial scale resulted in the well knownaverage gas to dust ratio for our Milky Way, at increasingly small scales thefluctuations in the dust to gas ratio dramatically increase (illustrated by Fig-ure 1.10). This may also play an important role in GRB star forming regions,potentially explaining the discrepancy between depletion derived AV ’s andtheir measured values and the fact that X-ray and optical extinction ratiosdeviate from the average values expected from their best fitting extinctioncurves.

* There is likely to be a population of massive stars that (almost) fails to makea GRB: the jet does not make it to the surface of the star, or is subrelativisticwhen exiting the star. Likewise there must be a number of bright GRBs thatwe miss because the jets are pointed too far away from us. Detecting mem-bers of these two classes is a big challenge. In the first case, polarimetryof type Ib/c supernovae to look for asymmetry may be an option, or un-usual abundance patterns in the supernova spectrum. The second case hasbeen tackled through very deep searches for afterglow-like phenomena in ra-dio supernova light curves. Orphan afterglows (afterglows with no detectedGRB for a variety of reasons) have been searched for extensively, throughprogrammes that image large parts of the sky but to shallow limits or imagesmaller parts of the sky going much deeper (Rykoff et al. 2005; Malacrinoet al. 2007). Wide field very deep surveys like the ones at VST, LSST andPan-STARRS may find some of these objects.

* Giant magnetar flares could provide a significant fraction of all short gamma-ray bursts. These flares are rare in our Galactic environment (three havebeen seen so far). A better knowledge of the physics of these flares andparticularly the fireballs they produce may help us understand some shortGRBs.

1.3.11 Looking through sunglasses

It is clear that in the interpretation of the GRB line of sight results (e.g. the survivalof GRB 021004 WR wind lines, the influence of GRB radiation on the surroundingmedium, the fraction of GRBs we do not detect because of beaming) our knowl-edge of the structure of the jet is important. To give an example: if GRB jetscontain most of their energy in the central fraction of the opening angle, the chanceof low ionisation atomic species surviving the blast increases with the angle fromthe central jet axis. Studies of bright afterglows from radio to X-ray with good

1.4 Outline of this thesis 29

temporal resolution can in rare cases provide enough information to differentiatebetween different jet models, but a powerful tool to combine with these SEDs is awell sampled high signal to noise (linear) polarisation light curve. The magneticfield present in the blastwave can be probed by circular polarimetry, and variabil-ity in geometry by linear polarisation. This has been done for a few very brightafterglows of long bursts, but as these afterglows are only very weakly polarised(generally less than a few percent, e.g. GRB 030329 Greiner et al. 2003), a veryhigh signal to noise is necessary, requiring exposure times in excess of an hour onan 8 m telescope even when doing imaging polarimetry of the brightest afterglows.These low polarisation values combined with the observed synchrotron spectrumpoint to the fact that the magnetic field is created in the shock in the blastwave,rather than a powerful ambient magnetic field from e.g. the host galaxy. Since 2004I have run the GRB polarimetry programme at ESO VLT, but no afterglows werebright enough to trigger during nights available to my programme. In the case ofshort bursts, polarimetry may perhaps be the most powerful tool to find the progen-itor: in the binary merger model, (one of) the neutron stars is a non-recycled pulsar,having a powerful magnetic field. This could perhaps be captured in the blastwave,producing polarisation values higher than those of long GRBs. This is extremelydifficult to measure, as a it requires high signal to noise (∼500) measurements atfour retarder angles before the afterglow fades away. Ideally one would also needto measure polarisation quasi-simultaneously in a different band, to be able to cor-rect for polarisation induced by scattering by dust in the host galaxy. Short burstafterglows that are sufficiently bright do exist (the afterglow of GRB 050724 wasbright enough a few hours after burst; Figure 1.5), but they tend to be discoveredtoo late after the GRB event. The advent of larger telescopes with imaging polari-sation capabilities could potentially enable us to follow-up more sources (Yoon &Langer 2005).

1.4 Outline of this thesis

This thesis presents work I did over the past few years, mainly focussed on obser-vations of GRB afterglows, associated supernovae and host galaxies. The emphasisis on long GRBs and their relationship to their hosts. The central question can beformulated as: how can we observationally study GRB progenitors if we are neverable to resolve the hosts of GRBs into individual stars?

Chapter 2 includes a study of the afterglow, supernova and host galaxy proper-ties of GRB 040924. This is a GRB from the pre-Swift era, detected by the HETE-IIsatellite, which has now ceased regular operations. There are no X-ray observationsfor this source, but good radio limits, and a well sampled optical afterglow. Thisburst has a short duration, and in the pre-Swift days had been claimed by severalauthors to be a short burst. We discuss the properties of the host and the light curveof the afterglow, and find them to be very similar to the sample of long bursts.

30 Introduction

Furthermore, a supernova has been observed for this burst, making it a standardcollapsar GRB.

A high signal to noise spectrum of the afterglow of GRB 050730 is presented inChapter 3. We acquired this spectrum using the WHT ISIS, starting observationsapproximately 3 hours after the burst. This is a classic GRB afterglow spectrum,with a very high hydrogen column density of log NH I = 22.1 ± 0.1 - the highestknown hydrogen column density at the time of observation. We determined theredshift at z = 3.968. Several fine structure lines are visible, indicating an originin star forming regions. We find an upper limit on the sulphur abundance of [S/H]< −2.0, one of the lowest line of sight abundances seen so far in the Universe. TheX-ray spectrum shows a further surprise: a decrease in the X-ray absorbing columndensity (commonly expressed in terms of an equivalent hydrogen column density)is observed following X-ray flares.

Chapter 4 shows how well sampled multiband optical light curves can be com-bined with X-ray data for a sample of afterglows. The optical to X-ray data arefit together in count space: to convert an X-ray spectrum to flux space one has tofit for the intrinsic model, the instrument response and extinction simultaneously,so a fit in flux space would have meant a prior assumption about the model. Thisprovides accurate values for the optical and X-ray extinction. It is clear that the bestfitting optical extinction model is the SMC model, and the ratios of X-ray to opticalextinction are often higher than expected. While this may point to the destructionof dust, it may also result from other effects.

Both Chapters 5 and 6 deal with the same source, GRB 060206. In a coordi-nated effort with the Nordic Optical Telescope (NOT) and WHT we acquired lowand high resolution spectroscopy of the afterglow of this long burst, using the AL-FOSC and ISIS instruments. The WHT spectra are displayed in Figure 1.6. InChapter 5 we describe the spectral features, and calculate the sulphur abundance.We compare this abundance to that in a sample of GRBs with afterglow spectra, andfind that this host galaxy, at redshift z = 4.048 has a high metallicity for its redshift.It is interesting to compare the spectrum of this burst with that of GRB 050730 inChapter 3, which has a very similar redshift. The most remarkable discovery ofthis paper is the tentative detection of hydrogen molecules in this spectrum. Chap-ter 6 describes a thorough Voigt profile fit to all the detected lines, with the aimof discovering the origin of the various velocity systems we see in the afterglowspectrum. Using a deep pointing with Gemini, the host galaxy of this burst hasbeen detected, which allows us to put some constraints on the existence of a mass-metallicity relationship at these very high redshifts.

Chapter 7 presents a thorough analysis of a high spectral resolution VLT UVESspectrum of SN 2006aj, the supernova associated with GRB 060218. We combinedthis spectrum with a deep, low resolution VLT FORS spectrum obtained by us-ing all spectra taken in our observing campaign of the supernova. A series of hostgalaxy emission lines are clearly visible, with very high signal to noise. Also visiblein the FORS spectrum is the [O III] λ4363 auroral line. Using this we calculate the

1.4 Outline of this thesis 31

metallicity using electron temperature methods, for which we find 12 + log(O/H)= 7.54+0.17

−0.10, the lowest GRB host metallicity found so far from emission lines. Thisis also only the fourth host for which the metallicity could be determined throughTe methods. We determine abundances for several more elements, with the partic-ularly notable result of a possible nitrogen overabundance. This has been noted fortwo more hosts, and may provide a connection with the GRB event. A big surpriseis the detection of two discrete velocity components in the [O III] emission lines,and more so the identification of two absorption systems of Ca II and Na I, at al-most the same two velocities (separation ∼ 22 km s−1). The two components havedifferent widths in the emission lines, and have different ratios in the column den-sities of Ca II and Na I. We speculate that these components are two separate starforming regions in the host. The very low metallicity of this host makes it a veryinteresting candidate from which to obtain limits on the WR /O star ratio as it formsa sequence in metallicity with the hosts of GRB 980425 and 020903, the only twohosts for which WR lines have been measured so far. This is the best studied GRBsupernova, with detailed photometric and spectroscopic data of the supernova, theprompt emission and even the shock breakout, so a thorough understanding of thehost for this burst is crucial in providing sufficient data for numerical modelling ofthe progenitor.

Chapter 8 presents the case of a dark burst, GRB 051022, that is clearly highlyextincted rather than intrinsically faint. This is one of the very few bursts for whichno optical afterglow was found, but that did show a radio afterglow, and for whicha redshift could be measured. From the radio, submm and X-ray data a broad-band SED can be constructed. It is very clear that the optical flux limits we obtainthrough observations with a variety of filters and telescopes are far below the fluxesexpected from the SED. This is likely caused by extinction. The host galaxy how-ever appears to be a very typical host, and does not require very high amounts ofdust to model its SED.

2Spectroscopy and multiband photometry of theafterglow of intermediate duration γ-ray burst

040924 and its host galaxy

K. Wiersema, A. J. van der Horst, D. A. Kann, E. Rol, R. L. C. Starling,P. A. Curran, J. Gorosabel, A. J. Levan, J. P. U. Fynbo, A. de Ugarte Postigo,

R. A. M. J. Wijers, A. J. Castro-Tirado, S. S. Guziy, A. Hornstrup, J. Hjorth,M. Jelınek, B. L. Jensen, M. Kidger, F. Martın-Luis, N. R. Tanvir, P. Tristram,

P. M. VreeswijkAstronomy and Astrophysics submitted; arXiv:0706.1345

Abstract We present optical photometry and spectroscopy of the afterglow andhost galaxy of gamma-ray burst 040924. This GRB had a rather short durationof T90 ∼ 2.4s, and a well sampled optical afterglow light curve. We aim to usethis dataset to find further evidence that this burst is consistent with a massive starcore-collapse progenitor. We combine the afterglow data reported here with datataken from the literature and compare the host properties with survey data. We findthat the global behaviour of the optical afterglow is well fit by a broken power-law,with a break at ∼0.03 days. We determine the redshift z = 0.858 ± 0.001 from thedetected emission lines in our spectrum. Using the spectrum and photometry wederive global properties of the host, showing it to have similar properties of otherlong GRB hosts. We detect the [Ne III] emission line in the spectrum, and comparethe fluxes of this line of a sample of 15 long GRB host galaxies with survey data,showing the long GRB hosts to be comparable to local metal-poor emission linegalaxies in their [Ne III] emission. We fit the supernova bump accompanying thisburst, and find that it is similar to other long GRB supernova bumps, but fainter. Allproperties of GRB 040924 (the associated supernova, the spectrum and SED of hostand afterglow) are consistent with an origin in the core-collapse of a massive star.

33

34 Spectroscopy and multiband photometry of gamma-ray burst 040924

2.1 Introduction

The large sample of γ-ray bursts (GRBs) detected by the BATSE instrument onboard the CGRO satellite clearly shows that GRBs can be classified in two distinctclasses, based on spectral hardness and duration (Kouveliotou et al. 1993), thoughthe possibility of three classes has also been proposed (see e.g. Horvath et al. 2006,see also Hakkila et al. 2003). The difference in duration and spectral hardness mayreflect important differences in the production mechanism of relativistic jets by theprogenitor object, and therefore may be an important clue to the nature of the pro-genitor. The great difficulty in obtaining accurate γ-ray positions for the short-hardburst population made searches for afterglows unsuccessful for a long time. Withthe arrival of Swift and its ability to rapidly identify X-ray afterglows also for faintand short duration bursts, this situation changed dramatically, resulting in severallocalisations of short bursts by both Swift and HETE-2, and the detection of X-rayand optical afterglows for a considerable fraction of these (see the review by Nakar2007 and references therein). With the different sensitivity and spectral range thatis covered by these satellites with respect to BATSE, the classification of bursts inthe classes of short and long burst became more ambiguous in several cases, forexample due to long lasting soft tails detected in the prompt emission of bursts thatBATSE would have classified as members of the short-hard class (see e.g. Norris& Bonnell 2006, and references therein). The question of the progenitors of shortbursts versus long bursts can now also be approached through other angles, suchas the presence of a supernova, the nature of the host galaxies of a large sample ofbursts, the density in which the afterglow propagates and its gradient, the circum-burst environment through afterglow spectroscopy to name just a few (see Lee &Ramirez-Ruiz 2007; Nakar 2007, and references therein). The distinction betweenshort and long GRBs is however under strong debate. A classification just by theduration of the prompt emission has proved to be weak to identify progenitor mod-els (see also the recent detection of two long bursts without supernovae Fynbo et al.2006b; Gal-Yam et al. 2006; Della Valle et al. 2006; Gehrels et al. 2005). A largesample of datasets covering prompt emission, afterglow and host galaxies of GRBsat both sides of the classical 2 second limit is required. In this paper we focus onthe properties of GRB 040924. This burst has a duration very close to the classical2 second long-short divide, and has been considered a short burst candidate (e.g.Fan et al. 2005).

GRB 040924 was localised by the Fregate and WXM instruments onboard theHETE-2 satellite on 2004 September 24 at 11:52:11 UT (Fenimore et al. 2004),and also detected by satellites within the IPN network (Golenetskii et al. 2004).The burst was classified as an X-ray rich burst, based on the ratio of the fluencesof the burst in the 7-30 keV and the 30-400 keV bands. Donaghy et al. (2006)report a T90 duration of 2.39 ± 0.24 seconds from the HETE-2 data, and show thatthe duration alone would make this burst a good candidate for membership of theshort GRB class. However, a clear spectral lag is present and the spectrum is soft,

2.2 Observations and analysis 35

favouring a long burst class membership (Donaghy et al. 2006). As such it is aninteresting burst to study.

In this paper we present our dataset on the afterglow and host GRB 040924,supplemented with observations from literature. In §2 we describe our observa-tions. In §3 we analyze the afterglow data, and in §4 the data on the host galaxy.In §5 we derive general properties for the associated supernova in relation to otherGRB related supernovae, and in §6 we present our conclusions and discussion.

2.2 Observations and analysis

2.2.1 Afterglow optical photometry

Due to the quick localisation by HETE-2, the Palomar 60 inch telescope was able torapidly find a candidate for the optical afterglow associated with this burst (Fox &Moon 2004; Soderberg et al. 2006c). We observed the afterglow of GRB 040924 atseveral different observatories. All data were reduced with standard IRAF and MI-DAS packages. Cosmic ray hits were removed using the L.A.Cosmic software byVan Dokkum (2001). Astrometric calibration was done using the USNO-B1.0 cat-alogue. Photometry was done using the IRAF DAOPHOT and APPHOT packages.The V, R and I band images were calibrated relative to the photometric sequenceprovided by Henden (2004). The log of photometric observations and their resultsare shown in Table 2.1.

We imaged the error box with the 0.6 m telescope at Mt. John Observatory(+ MOA camera) using a wide R-band filter. A 300-s single image was obtainedthrough clouds at 0.087 days after burst, but did not detect the afterglow. Furtherimaging at 0.2 days did marginally detect the afterglow. Two epochs of V, R and Iband imaging of the afterglow position were done with the Very Large Telescope(VLT) in Cerro Paranal (Chile), using the Focal Reducer - Low Dispersion Spec-trograph (FORS1) on UT2 (Kuyen). The conditions during the first epoch wereexcellent (seeing ∼0.6 arcsec), but seeing increased in the second epoch. Two af-terglow observations were done with the 2.56-m Nordic Optical Telescope (NOT)at La Palma. Observations were carried out using the Andalucıa Faint Object Spec-trograph and Camera (ALFOSC). We performed an H-band observation using thenear-infrared MPI fur Astronomie General-Purpose Infrared Camera (MAGIC) onthe 1.5m telescope at Observatorio de Sierra Nevada, which we calibrated using2MASS field stars.

2.2.2 Host galaxy spectroscopy

Attempts at measuring the redshift from afterglow spectroscopy using the WilliamHerschel Telescope (WHT) on La Palma on September 25 and 26 unfortunatelyproved unsuccessful due to bad weather decreasing the signal to noise ratio. There-fore we obtained a dark time spectrum aimed at the host galaxy redshift using VLT


Figure 2.1 — Cut from the first epoch VLT FORS2 I band observation, showing theafterglow marked by tickmarks. The star marked by G was used to align the FORS2 slitfor the host galaxy spectroscopy. The moving object marked by A is a main belt asteroidwhich was present in the first epoch V, R and I band observations.

UT1 (Antu) with the FORS2 instrument in long slit spectroscopy (LSS) mode onOctober 10, approximately 15.75 days after burst. A bright star marked as G inFig. 2.1 was used to position the slit, 27 arcseconds away from the host galaxy;the position angle was approximately 25.1 degrees. Two spectra of 1800 secondsexposure time each were taken, using the 300V and 300I grisms, to increase wave-length coverage. We used a 1 arcsec slit. Both exposures were done without anorder sorting filter to increase sensitivity, which means that the red end of the 300Vis contaminated by the second order (that has twice the resolution and wavelength).As there are no lines visible in the bluest region and the continuum is undetected,this has no influence on the measured emission line fluxes. Both spectra havebeen reduced in a standard fashion using the data reduction package IRAF. TheL.A.Cosmic program (specifically the lacos spec routine) written by Van Dokkum(2001) was used to remove both point- and irregular shaped cosmic ray hits fromthe science and standard images. The wavelength resolution is estimated by mea-suring the FWHM of the arc lines at 10.2 Å for the 300V, and 9.6 Å for the 300Ispectrum. The spectra were flux calibrated by using the standard star LTT1788(Hamuy et al. 1992, 1994). Atmospheric extinction correction was done by ap-plying a mean extinction curve for Paranal. A Galactic extinction correction wasperformed by using the E(B−V) value of 0.058 mag (Schlegel et al. 1998), assum-ing a Galactic extinction law Aλ/AV expressed as RV = AV/E(B−V) (Cardelli et al.1989). We make the standard assumption RV = 3.1 (Rieke & Lebofsky 1985).

The combined spectra and the associated error spectrum are shown in Figure

2.2 Observations and analysis 37

2.2. The continuum level of the host galaxy is faint, with a mean signal to noiseratio of∼2 per pixel. We find several emission lines in the spectra, which we presentin Table 2.2. All of these lines are consistent with a redshift z = 0.858 ± 0.001. Noother faint, galaxy-like sources, visible in the WHT imaging data near the GRBhost galaxy, fell in the slit.

Figure 2.2 — The averaged VLT FORS2 300V and 300I spectrum of the host galaxy(upper panel), with corresponding error spectrum (lower panel). The prominent [O II]line is labelled. The spectrum is smoothed by 3 pixels for displaying purposes.

2.2.3 Radio observations

The radio observations reported in this paper are all done using the WesterborkSynthesis Radio Telescope (WSRT) in the Netherlands. Three observations at theposition of the afterglow were made, all at 4.9 GHz. Data reduction was performedusing the MIRIAD software (Sault et al. 1995). No significant detection of anafterglow was found in any of the three epochs, see Table 2.1. To search for a radiohost galaxy, we combine the data of all three observations. Again no radio sourceis found on the position of the afterglow, with corresponding 3σ upper limit of 63μJy. A formal flux measurement for a point source at the location of the opticalafterglow gives a value of 12 ± 21 μJy for the combined data.

2.2.4 Supernova search and host galaxy

The observations reported in this section were aimed at observing a possible super-nova associated with GRB 040924, which was later identified through much deeper


HST imaging (Soderberg et al. 2006c). A redshifted, K-corrected extrapolation ofsupernova 1998bw would yield expected SN magnitudes of R ∼24.6 and I ∼23.8,see section 2.5, making it possible for us to detect it with ground based telescopes.Observations were done with the Auxiliary Port Imaging Camera (Aux-Port Im-ager) mounted at the auxiliary port of the Cassegrain A&G box of the WilliamHerschel Telescope (WHT). The first epoch was done on October 23, consisting of4 x 15 minutes in I and 4 x 15 minutes in R. Whilst at the start of observations theconditions were good, the seeing greatly deteriorated during the exposures, which,combined with an increasing airmass, led to a relatively shallow limiting magnitudeand low image quality of the combined R band exposure.

A second epoch of I band imaging was performed on January 10, aimed atobserving the host galaxy. The observations were done in the same manner as thefirst epoch, consisting of 4 × 15 minutes exposure time. The host galaxy is clearlydetected. A second epoch of R band imaging was performed on July 16 2005,consisting of 3 × 15 minutes.

In order to get a broadband colour estimate of a potential supernova, two K′band observations were done near the supernova peak with the Omega2000 near-infrared wide field camera on the prime focus of the 3.5m telescope at the CalarAlto observatory, on 2004 October 26 and 27, with exposure times 2760 and 6390seconds respectively. The host and SN are not detected.

We note here that the photometric calibration of the WHT data are hamperedby the very small field of view (the Aux Port imager has a circular field of view of1.8 arcmin diameter, but dithering required for fringe correction reduced the usefulfield of view to approx 1.2 arcmin). No Henden (2004) calibration stars can be usedin R band to calibrate, as there are only two of these in the field of view, and theyare saturated. We use the NOT data to calibrate field objects, but most objects nearthe GRB host that are in both NOT and WHT R band data are galaxies. The I bandWHT data provided a better calibration due to the fact that more unsaturated Hen-den stars were present in the WHT I band image due to slightly different pointingand dithering. We therefore point out that particularly the R band magnitudes ofthe WHT data should be treated with caution.

2.3 The afterglow

We gathered published data from the literature (Terada et al. 2004b,a; Silvey et al.2004; Huang et al. 2005; Soderberg et al. 2006c) and combined these with our data,also adding data from the RTT 150 telescope (I. Khamitov, priv. comm.1 Khamitovet al. 2004). We corrected the afterglow magnitudes for Galactic extinction whereneeded. The resulting light curve is shown in Fig. 2.3.

We used the fitting method described in Curran et al. (2007a) which fits the datafrom all available bands simultaneously, assuming that each band has the same

1http://hea.iki.rssi.ru/∼rodion/040924

2.3 The afterglow 39

Δt (d) Magnitude, Error (1σ) Band / frequency Telescope0.204 21.3 ± 0.5 R MOA0.568 Upper limit 18.2 (3σ) H OSN MAGIC0.666 22.97 ± 0.09 V VLT FORS20.674 22.56 ± 0.13 R VLT FORS20.679 21.95 ± 0.19 I VLT FORS20.731 22.59 ± 0.15 R NOT ALFOSC0.871 23.27 ± 0.08 V VLT FORS20.878 22.85 ± 0.13 R VLT FORS20.890 22.08 ± 0.19 I VLT FORS22.97 23.38 ± 0.29 R NOT ALFOSC28.69 24.1 ± 0.30 R WHT Aux28.64 23.3 ± 0.2 I WHT Aux31.72 Upper limit 19.8 (3σ) K′ CAHA Omega200032.58 Upper limit 19.6 (3σ) K′ CAHA Omega2000108.39 23.6 ± 0.2 I WHT Aux295.69 24.39 ± 0.30 R WHT Aux0.663 Flux < 96 μJy (3σ) 4.9 GHz (6cm) WSRT5.541 Flux < 81 μJy (3σ) 4.9 GHz (6cm) WSRT32.48 Flux < 75 μJy (3σ) 4.9 GHz (6cm) WSRT

Flux < 63 μJy (3σ) 4.9 GHz (6cm) WSRT combined data

Table 2.1 — Log of observations of GRB 040924 reported in this paper. We note thatthe last NOT exposure is close to the limiting magnitude.

power-law decay but leaving free the offset between bands to account for differ-ences in filter curves and calibration. We refer to Curran et al. (2007a) for details onthe fitting method and error analysis. Our fit using all available data shows a brokenpower-law, with values of the afterglow decay index of α1 ∼ 0.31, α2 = 1.25±0.02and a break time of ∼ 0.03 days, where the break smoothness is fixed at a values = 10, with a best fit χ2/dof = 1.2. We note that the pre-break decay (α1) and thebreak time are not well constrained, the χ2 is fairly insensitive to changes in theseparameters. The R band afterglow shows a low significance wiggle around the bestfit power-law, most noticeably around ∼ 0.5 days. This is commonly seen in wellsampled afterglow lightcurves, and does not affect our afterglow fit noticeably.

Using the data set presented in this paper, Kann et al. (2006) found that theVRIK′ spectral energy distribution (SED) of the afterglow of GRB 040924 is fitwell by a power-law plus SMC-like extinction, though the data are also well fitwith no extinction. The extinction-corrected spectral slope is β = 0.63 ± 0.48and the host galaxy extinction is AV = 0.16 ± 0.44, agreeing with the value ofthe spectral slope β = 0.61 ± 0.08 reported by Silvey et al. (2004), who do notfit for extinction. A comparison with other intrinsic GRB afterglow luminosities


Figure 2.3 — Light curves of the afterglow of GRB 040924, fitted with an achromaticbroken power-law. The top curve is the I band lightcurve and the middle curve is R band.The lower curve shows the V band curve where 1 magnitude is added for clarity. Dataare from this paper and sources given in §2.3. The I band fit includes the contribution ofa supernova.

(Kann et al. 2006) shows that the afterglow of GRB 040924 was the faintest in theirsample of pre-Swift GRB afterglows with observed optical counterparts in multi-ple bands and enough sampling to derive temporal and spectral indices. Soderberget al. (2006c) have previously noted the break from a shallow decay to a more com-mon temporal decay index, and our decay slopes and break time agree with theirs,but improve on accuracy particularly for α2. This break is not consistent with ajet break, as the difference in temporal decay index is not as expected from a jetbreak, the break being very early, and both pre- and post-break decay slopes beingtoo shallow (cf. Zeh et al. 2006, for a large sample of optical afterglow parame-ters). Huang et al. (2005) reach similar results qualitatively, but have very differentvalues due to a much smaller dataset. The early data showing this break have onlybeen taken in R band so it is unclear whether this break is achromatic, but afterthe early break multi-color data are available. The case of the synchrotron coolingand peak frequency redward of the observed optical frequency after the break, i.e.νm < νc < νopt, is ruled out, as the spectral and temporal slope can not be acco-modated together in the standard blastwave model (e.g. Zhang & Meszaros 2004).The case for νm < νopt < νc using a homogeneous density medium or a 1/r2 (stellar

2.4 The host galaxy 41

wind) medium would give p ∼ 2.7 for a wind medium and p ∼ 2.1 for a homoge-neous medium from the temporal decay. The spectral slope gives p ∼ 2.2 in thiscase. The distinction between the homogeneous and wind medium can be made byassuming n(r) ∝ r−k, and calculating k using β and α (we refer to Starling et al.2007, for details). For the post-break values we find k = 1.6+0.3

−0.5, which agrees withan interpretation of a 1/r2 wind medium in which the blastwave propagates, thoughthe uncertainties are quite large.

The change in decay slope of Δα ∼ 0.9 rules out the possibility that the breakis caused by the passage of the cooling break νc, for which Δα = 0.25 wouldbe expected. Possibly this shallow decay phase and break can be attributed toenergy injection, a mechanism frequently invoked to explain shallow decay phasesin afterglows (e.g. Nousek et al. 2006; Pandey et al. 2006). Using a parametrizationE ∝ tq (Nousek et al. 2006), we find q ∼ 0.7 in the case of a homogeneous medium,and q ∼ 1.1 in the case of a wind medium.

2.4 The host galaxy

We detect several emission lines in the VLT FORS2 spectra of the host (see figure2.2), though the detections are of low significance due to the faintness of the host.The analysis of the emission lines was done using the Starlink DIPSO spectralfitting package, using mainly the ELF (emission line fitting) routines, and comparedwith results using the IRAF splot package. The results were in agreement withinerrors. We note that several of the emission lines (particularly the [O III]λ5007) areaffected by skylines and show differing fluxes in the 300V and 300I spectra, causedby the relative success of skyline subtraction. We have not included these lines infurther analysis. In the case where an emission line is found with similar flux in the300I as well as in the 300V spectrum, the average value is given.

Line ID Line flux×10−17 erg s−1 cm−2

[O II] λ3727.4 2.31 ± 0.14[Ne III] λ3868.8 0.47 ± 0.17H β λ4861.3 0.44 ± 0.20[O III]λ4958.9 0.94 ± 0.24

Table 2.2 — Spectroscopy results for the 300V and 300I grism. We list conservative 1σuncertainties. For lines that were detected in both grism spectra, a weighted average ofthe fluxes was used. Spectra are corrected for Galactic extinction.

The emission lines found in our host galaxy spectra allow us to securely deter-mine the redshift at z = 0.858 ± 0.001. The widths of the emission lines appear notto be significantly broadened, but are consistent with the instrumental resolution,


as is expected from spectroscopy of other host galaxies and the spectral resolutionthat was used .

The ratios of the measured fluxes [OIII]/Hβ and [OII]/Hβ compared to thosefrom samples of AGN and normal galaxies (Kennicutt 1992) show that the hostgalaxy spectrum is likely not dominated by (non-thermal) AGN emission. We donot significantly detect the Hγ line, with an upper limit lower than the expected fluxbased on the flux of the H β line and case B recombination, so we are not able toderive Balmer decrement values, and apply no host galaxy reddening correction tothe spectrum. This is consistent with the absence of apparent absorption affectingthe afterglow SED.

2.4.1 Star formation rate

To estimate the star formation rate we use the expression from Kennicutt (1998),

SFR[OII] = 1.4 × 10−41L[OII] M� yr−1,

where L[OII] = 4πd2l f[OII],obs, with dl the luminosity distance. Using cosmological

parameters H0 = 70 km s−1/Mpc, ΩM = 0.3 and ΩΛ = 0.7 (dl = 5.47 Gpc), we findSFR[OII] = 1.15 ± 0.08 M� yr−1. The real uncertainty in the SFR is much larger:the SFR[OII] has been calibrated on the Hα SFR through empirical methods, pro-viding an additional uncertainty through the scatter in this relation; and the effectsof dust obscuration is uncertain (but likely negligible; see Sect. 3 and 4). This esti-mate for the non-extincted star formation is very typical of long GRB host galaxies(Christensen et al. 2004).

The radio continuum flux of a normal (ie. non-AGN hosting) galaxy is formedby synchrotron emission of accelerated electrons in supernova remnants and byfree-free emission from HII regions, see Condon (1992). It is expected that theradio continuum flux is a particularly good tracer of the recent SFR, due to theshort expected lifetime of the supernova remnants, which is � 108 yr. We usethe methods by Vreeswijk et al. (2001) and Berger et al. (2003a) to calculate anupper limit on the unobscured star formation rate. The three WSRT observations(see Table 2.1) combined give a 3σ upper limit to the radio continuum flux of thehost galaxy of 63 μJy (see Section 2.2.3), and from this limit we find a 3σ upperlimit on the unobscured star formation of 137 M� yr−1 (using the prescription inBerger et al. 2003a). This is a limit deeper than that obtained for most GRB hostgalaxies at similar redshift (Berger et al. 2003a). We note that the conversion ofradio luminosity to star-formation rate fails in cases where there is an appreciablecontribution to the radio flux from an AGN component, though no association of aGRB to an AGN harbouring host galaxy has been seen so far.


2.4.2 Metallicity

As the metallicity of the progenitor object plays a crucial role in our understandingof the evolutionary process leading to GRBs (e.g. Yoon & Langer 2005), it is vitalto secure metallicity measurements for as many GRB host galaxies as possible. Incases where this is not possible due to redshift or signal to noise constraints, we maybe able to at least ascertain the possibility that the host has significantly subsolarmetallicity. The difference in progenitor properties between short and long burstsmay also be reflected in the distribution of parameters of the host galaxies of GRBs,such as their metallicity (but see Levan et al. 2007).

Metallicities of GRB host galaxies are in general hard to determine: the aver-age redshift of GRBs is high, and the hosts tend to be faint. For the high redshiftsample a line of sight value for the metallicity towards (long) GRBs can be obtainedby observing the hydrogen column through the Lyman line(s) and the columns oflines that are not easily depleted in dust (e.g. sulphur or zinc). This method reg-ularly provides very accurate values of the columns of elements (e.g. Fynbo et al.2006a; Price et al. 2007; Prochaska et al. 2007a), but as the data are obtained overa line of sight and GRBs influence their environments profoundly, the metals andhydrogen are not necessarily co-located (e.g. Watson et al. 2007). High qualitymeasurements of oxygen abundance using electron temperatures (Te), determinedthrough the [O III]λ4363 line, can generally only be obtained for the very lowestredshift subsample of GRB hosts, and has so far only been possible for four cases(Prochaska et al. 2004; Hammer et al. 2006; Wiersema et al. 2007). For GRB hostsat higher redshift, this line is often too faint to detect in a reasonable exposuretime (except in cases of very high ionization parameter), and secondary indica-tors of metallicity are required. Extensive studies have been performed to findmetallicity diagnostics that depend solely on bright emission lines and on their cal-ibration with Te-derived measurements. Generally the requirements of a secondarymetallicity indicator are that the lines that are used are bright; that the correlation(based on local galaxies) with Te-determined metallicities has a low dispersion;that the line fluxes can preferably be obtained in a single spectrum (ie lines thatare close together in wavelength) and that preferably both high and low ionisationlines are used so the ionization parameter can be found. One of the most frequentlyused metallicity diagnostics using the brightest lines is the R23 method (Pagel et al.1979). This method gives a low and a high metallicity solution for a value of R23.This degeneracy can be broken through other lines, that are often fainter and moredifficult to detect, and the ionization parameter needed to get a good value throughR23 can be hard to determine in the case of high redshift and low signal to noisedata. In the case of the host of GRB 040924, the [O III]λ5007 is located right on askyline, which makes it impossible to measure its flux. Instead we use a fixed ratioof [O III] λ5007 to λ4959 of 3 (Izotov et al. 2006b) to estimate the [O III] λ5007at ∼2.8 ×10−17 erg s−1 cm−2. We find R23 ∼ 13.8 and using the Kewley & Dopita(2002) formulas we find that the host flux ratio is located near the connection of


-2.0 -1.5 -1.0 -0.5 0.0 0.5log([Ne III]/[O II])

0

5

10

15

20

25

Num

ber

Figure 2.4 — Histogram showing the distributions of log ([Ne III]/[O II]): a dot-dashedline shows the SDSS spectrophotometric catalogue values of sources with a determinedTe metallicity from Izotov et al. (2006b), 125 sources. With a solid line the values for14 GRB host galaxies are shown. In both cases we only use sources where both linesare detected. The filled squares show the values for the three regions in the host galaxyof GRB 980425 as determined by Hammer et al. (2006), where the highest value isthe bright star-forming region where many WR stars are present, and the middle valueis taken at the GRB location, see Hammer et al. (2006) for details. GRB hosts usedare 970228 (Bloom et al. 2001a), 970508 (Bloom et al. 1998a), 970828 (Djorgovskiet al. 2001), 980613 (Djorgovski et al. 2003), 990712 (Vreeswijk et al. 2001), 000418(Bloom et al. 2003), 020405 (Price et al. 2003), 020903 (Hammer et al. 2006), 030329(Gorosabel et al. 2005), 031203 (Prochaska et al. 2004), 040924 (this work), 050824(Sollerman et al. 2007), 060218 (Wiersema et al. 2007), 060505 (GRB position Thoneet al. 2007) and 060912A (Levan et al. 2007, priv. comm.).

the low and high metallcity branches, having a metallicity 12 + log(O/H)∼ 8.5, butis badly constrained due to the large uncertainty in R23, which is dominated by theuncertainty in H β. It has been noted by several authors that the R23 can give resultsdiscrepant from direct (Te) measurements, and several empirical calibrations havebeen offered (e.g. Nagao et al. 2006). We find a value 12 + log(O/H)∼ 8.1 from theR23 value of the host of 040924 when using the Nagao et al. calibrations.

More accurate secondary methods (e.g. involving the [N II] and [S II] lines) cannot be used for this host, as the diagnostic lines are redshifted out of the spectralrange.


2.4.3 [Ne III] in GRB hosts

GRB host galaxies can be found up to high redshift, and in several cases spectraof host galaxies with redshifts even higher than z ∼ 1 have been obtained. Atthese redshifts, the bright Hα and [O III]λ5007, 4959 are shifted out of the (use-ful) range of (low-resolution) optical spectra, making any secondary calibrator thatrequires these lines or an accurate correction for reddening impossible to use. Ithas been noted (e.g. Bloom et al. 1998a, 2001a; Djorgovski et al. 2001) that the[Ne III]λ3869 line is frequently detected in GRB host spectra, and its strength canbe used to infer the presence of very massive stars. The ionization potentials of[Ne III] and [O III] are very similar (41.1 eV and 35.1 eV), so we may expect to beable to use [Ne III] lines as a substitute for [O III] lines when the redshift is too high.An added advantage is the fact that the [Ne III]λ3869 is close in wavelength to the[O II]λ3727 line, making uncertainties in the reddening correction less pronouncedwhen using ratios of these blue lines. Nagao et al. (2006) have gathered spectro-scopic data from several samples of local galaxies with a broad range of metallici-ties, to derive unbiased samples of line fluxes and metallicities, in order to calibratecommonly used secondary metallicity indicators. Their analysis shows that at agiven metallicity the ionisation parameter has a low dispersion, and that it dependsstrongly on metallicity. Based on their sample of galaxies Nagao et al. propose touse the ratio of the [Ne III] and [O II] fluxes as metallicity estimators in the absenceof other, brighter high ionisation lines such as [O III] due to high redshift. Thisrelation is of particular importance to long GRB hosts, as their redshift distributionpeaks at z ∼ 2.8 (Jakobsson et al. 2006b), long GRB hosts tend to have high ioniza-tion parameters and because the range up to z ∼ 1 is most suitable to compare withother deep galaxy surveys (Savaglio 2006, and references therein). We compile alist of GRB host galaxies with published fluxes of [O II] and [Ne III], and comparethe log ([Ne III]/[O II]) values of the hosts with those of the SDSS (DR3) sampleof galaxy spectra with Te derived oxygen abundances of Izotov et al. (2006b), i.e.a sample of emission line galaxies with detected [O III]λ4363. This line becomesundetectable when the metallicity increases above 12 + log(O/H)∼ 8.5 (as the H IIregions are too cool), so this is essentially a sample of metal-poor emission linegalaxies. Figure 2.4 shows the resulting distributions. A Kolmogorov-Smirnovtest shows that the GRB host sample is not significantly inconsistent (probabilityP = 0.23) with the SDSS sample of galaxies with Te metallicity determinations.This result indicates that the sample of low redshift GRB host galaxies does notsignificantly differ from metal poor emission line galaxies as a whole in a similarredshift range in terms of the their integrated massive star populations, though thecurrent GRB host spectroscopy sample is sparse and more data is required to becertain. The values of the host of GRB 980425 show that conclusions can not bedrawn about the site where the GRB occurred, but only on the integrated light of thegalaxy, as has been shown in much more detail in Hammer et al. (2006) and Thoneet al. (2007). The value log ([Ne III]/[O II]) ∼ −0.7 for the host of GRB 040924 is


not an extreme value for a GRB host, and using the samples of Nagao et al. (2006)we find that the host has a metallicity of 12 + log(O/H)∼ 8.1 from this flux ratio,matching the value from R23 (Section 4.2).

2.4.4 Host SED

The host galaxy of GRB 040924 is clearly detected in deep HST pointings (Soder-berg et al. 2006c; Wainwright et al. 2007). We use the detections in the F775 andF850LP filters, and add our R and I band detections from WHT, and upper limits inK′ and H band from GCNs. The R band and HST bands straddle the 4000 Å break,and can therefore be used to determine some parameters of the host. We fit galaxytemplates using the HyperZ program2 developed by Bolzonella et al. (2000). Wefit using the eight synthetic galaxy templates provided within HyperZ, using thespectroscopic redshift. We find that the host is best fit by a starburst template, andhas MB ∼ −18.7, which confirms the HST results reported by Wainwright et al.(2007). The age of the stellar population is young (best fit ∼0.12 Gyr) with littleor no reddening, but these parameters are not well constrained as the R band mag-nitude is uncertain and bluer bands are not available, which makes the distinctionbetween reddening and age hard to establish quantitatively. The best fit parametersare very similar to the sample of long GRB hosts analysed by Christensen et al.(2004). The absolute B band magnitude of the host and the estimated metallicityof 12 + log(O/H)∼ 8.1 fits well in the local luminosity-metallicity relation (Salzeret al. 2005), though uncertainties are large. One characteristic of long GRB hosts isthe relatively high specific starformation rate (Christensen et al. 2004). AssumingM∗B = −21.1 and using the SFR from [O II], we find that the host of GRB 040924has a SSFR ∼10 M� yr−1 (L/L∗)−1. This value matches well with the range of SSFR∼ 5 − 12 found for a sample of 10 long GRB hosts analyzed by Christensen et al.(2004).

2.5 An associated supernova

We search for a supernova associated with GRB 040924 in the images from WHT(see section 2.2.4) by searching for a bump in the late time light curve. Soderberget al. (2006c) use image subtraction with HST ACS data to derive the parametersof the supernova. They acquired three epochs, of which the first and third arein filters F775W and F850LP, and for the second epoch only F775W was used.Through image subtraction they find a rebrightening of the afterglow of more thana magnitude, a clear signal of a supernova bump.

We use the values reported in Soderberg et al. (2006c) and Wainwright et al.(2007) for host and supernova measurements, and add to that our ground-based datafrom WHT and CAHA. We fit the full light curve of the afterglow + host galaxy +

2See http://webast.ast.obs-mip.fr/hyperz/

2.6 Discussion and conclusions 47

supernova using the method described in Zeh et al. (2004), using SN1998bw as thetemplate supernova. As we only have good measurements of the SN bump in the Iband, but the afterglow is determined well only in the R band, we use the R bandafterglow as a reference light curve. The host galaxy magnitudes in I band deter-mined from WHT and HST agree well with each other. We find a luminosity factork = 0.203 ± 0.202 and a stretch factor s = 1.371 ± 0.971. Essentially, the SN bumpis only marginally detected through this analysis. We take k < 0.4 (equivalent toa peak magnitude which is one magnitude fainter than SN 1998bw) as a conser-vative upper limit and thus confirm the finding of Soderberg et al. (2006c) that thesupernova associated with this burst is fainter than a K−corrected, redshifted SN1998bw. This can not be caused solely by extinction in the host: the analysis ofthe afterglow SED reveals a negligible amount of extinction along the line of sight.Even correcting for the possible small amount of extinction, this is still the faintestGRB-associated SN detected thus far (see Ferrero et al. 2006 and Sollerman et al.2007 for the complete sample). It is most similar to the SN associated with GRB970228 (Galama et al. 2000), but we caution that this SN was significantly brightereven without a correction for the extinction in the host galaxy. As the photometriccalibration uncertainty adds a large amount to the WHT data uncertainties, we per-form differential photometry with respect to bright field stars present in all WHTepochs, and find brightening with 2.2σ in I and 1.8σ in R, showing that while aSN 1998bw at this redshift would have easily be detected by WHT, a GRB-SN thismuch fainter requires higher signal to noise data.

2.6 Discussion and conclusions

It is clear that the standard differentiation between short and long bursts using solelytheir duration is not a clear enough way to also separate out the properties of theirprogenitors as a whole; the distributions of long and short bursts overlap signifi-cantly (Donaghy et al. 2006). The same goes for the properties of their host galax-ies: there are clear associations of short GRBs with elliptical galaxies devoid ofrecent starformation (e.g. the case of GRB 050724 Berger et al. 2005b), but thereare also cases of actively starforming short burst host galaxies (e.g. Hjorth et al.2005; Covino et al. 2006). The lack of an optical afterglow for many short bursts(and therefore a sub-arcsecond position) makes the association of a galaxy witha burst rather complicated (Levan et al. 2007), making the diagnostic power of agalaxy association less strong by selecting only the optically luminous cases. Possi-ble progenitor models comprise several mechanisms (e.g. compact binary mergers,SGR giant flares), each contributing to the observed overall short burst population.A collection of parameters that can each on its own not decisively separate outdifferent progenitor models, could in combination be used for population synthe-sis. In a sense these parameters are not much different from the parameters thatconstrain long GRB progenitors. These include (but are not limited to): the lumi-


nosity function of both GRB and afterglow, requiring the redshift distribution andjet-break determinations; the age of the dominant stellar population, determinedfrom host galaxy SED fitting; the quantity and age of the most massive stars, deter-mined through WR signatures and high-excitation lines; the offset of GRBs fromthe hosts / starforming regions; the abundance structure of the host galaxies, foundthrough afterglow and emission line spectroscopy; the circumburst density and itsstructure, found from afterglow SED modelling; the jet opening angle distribution,etc. The sample of bursts that seem to find themselves close to the classic 2 secondduration division are important in these studies, as they likely contain bursts fromdifferent progenitors and allow for easy comparison with the samples of long andshort bursts.

GRB 040924 clearly finds its origin in the collapse of a massive star. The prop-erties of the host galaxy fall within the range of the sample of long GRB hostgalaxies: the host is blue, is actively forming stars, and has significantly sub-solarmetallicity. The afterglow, albeit faint with respect to other pre-Swift long GRBs,behaves similarly to other long GRBs. The supernova could have been missed ifthis burst had occured at significantly higher redshift, but although there are shortbursts that share one or two of the host/afterglow properties with this burst (see e.g.Prochaska et al. 2006a, for an overview of short burst host properties such as metal-licities, luminosities etc.), the combination of prompt emission, afterglow and hostproperties makes it clear that the progenitor was a collapsar. The part of promptemission parameter space where both long and short bursts can be found (i.e. thepart of the hardness - duration diagram where the two burst distributions overlap)is interesting to explore, as we may expect to find contributions from at least twoprogenitor scenarios.

Acknowledgements This project is partly based on observations obtained at theESO VLT under ESO programme 073.D-0465(B), PI R.A.M.J. Wijers. We expressour gratitude to ESO for granting our request for spectroscopy. We are very gratefulto the observers that performed the observations reported here, and we thank particu-larly Ian Skillen for his assistance in the planning of GRB observations. We would liketo thank Arne Henden for taking a deeper photometric calibration exposure. KW andPAC thank NWO for support under grant 639.043.302. This study is supported by Span-ish research programmes ESP2002-04124-C03-01 and AYA2004-01515. This paper isbased on observations made with the William Herschel Telescope operated on the is-land of La Palma by the Isaac Newton Group in the Spanish Observatorio del Roque delos Muchachos of the Instituto de Astrofisica de Canarias; on observations made withthe Nordic Optical Telescope, operated on the island of La Palma, jointly by Denmark,Finland, Iceland, Norway and Sweden, in the Spanish Observatorio del Roque de losMuchachos of the Instituto de Astrofısica de Canarias; and on observations collected atthe German-Spanish Astronomical Center, Calar Alto, operated jointly by Max-PlanckInstitut fur Astronomie and Instituto de Astrofısica de Andalucıa (CSIC). The Wester-bork Synthesis Radio Telescope is operated by ASTRON (Netherlands Foundation for


Research in Astronomy) with support from the Netherlands Foundation for ScientificResearch (NWO).

The authors acknowledge benefits from collaboration within the Research TrainingNetwork ‘Gamma-Ray Bursts: an enigma and a tool’, funded by the EU under contractHPRN-CT-2002-00294.

3Gas and dust properties in the afterglow spectra

of GRB 050730

R. L. C. Starling, P. M. Vreeswijk, S. L. Ellison, E. Rol, K. Wiersema, A. J. Levan,N. R. Tanvir, R. A. M. J. Wijers, C. Tadhunter, J. R. Zaurin, R. M. Gonzalez

Delgado & C. KouveliotouAstronomy and Astrophysics, 442, L21 (2005)

Abstract We present early WHT ISIS optical spectroscopy of the afterglow ofgamma-ray burst GRB 050730. The spectrum shows a DLA system with the highestmeasured hydrogen column to date: N(H I) = 22.1 ± 0.1 at the third-highest GRBredshift z = 3.968. Our analysis of the Swift XRT X-ray observations of the earlyafterglow show X-ray flares accompanied by decreasing X-ray absorption. Fromboth the optical and the X-ray spectra we constrain the dust and gas properties ofthe host galaxy. We find the host to be a low metallicity galaxy, with low dustcontent. Much of the X-ray absorbing gas is situated close to the GRB, whilst theH I absorption causing the DLA is most likely located further out.

51

52

3.1 Introduction

Gamma-ray bursts (GRBs) have proven to be excellent probes of the distant Uni-verse. High luminosity GRB afterglows allow absorption line studies of the ISMat high redshift to at least z = 4.5 (see Andersen et al. 2000). The launch andsuccessful operation of the Swift satellite means more GRBs are being localisedand afterglows studied. Subsequently, the number of high redshift bursts suitablefor host galaxy spectral studies has dramatically increased. Deep observations ofafterglow positions have detected host galaxies in almost all cases (Conselice et al.2005). Most hosts are compact, actively star-forming galaxies and, where the rele-vant data are available, are found to have low metallicity and low intrinsic extinction(e.g. Berger et al. 2003a; Tanvir et al. 2004; Christensen et al. 2004). However, ina few cases, radio/submm observations of hosts give a star-formation rate (SFR)which is of order a few to ∼100 times larger than rates derived from optical es-timators such as the line luminosities of Hα and [O II] or the 2800 Å restframeUV continuum flux (e.g. Berger et al. 2003a). This may be caused by strong dustobscuration, but neither spectra nor colours of hosts show strong internal extinc-tion. Afterglow spectroscopy provides a unique window on the near environmentof GRBs (e.g. GRB 021004 Schaefer et al. 2003; Fiore et al. 2005; Starling et al.2005b), allowing us to probe the absorbing dust and gas properties in more detail.In this Letter we present optical and X-ray spectra of GRB 050730, discovered bySwift on July 30th 2005, 19:58:23 UT (Holland et al. 2005) and lying at a redshiftof z = 3.97 (Chen et al. 2005b; Rol et al. 2005a), in which we study the circumburstgas and dust properties.

3.2 The optical afterglow spectra

3.2.1 Observations

During the afterglow phase of GRB 050730, we acquired spectra using theIntermediate-dispersion Spectroscopic and Imaging System (ISIS) on the WilliamHerschel Telescope. The R316R and R300B grisms were used on the red and bluearms respectively. Two observations were done sequentially, at the parallactic an-gle, with exposure times of 1260 and 1800 seconds. The first observation started at22:57 UT at airmass ∼2.73 (midpoint 0.132 days after burst), the second at 23:19UT (midpoint 0.145 d) and airmass ∼3.4. The seeing quality at the high airmassesrequired the slit width to be widened to 2.5 arcsec. Conditions during the observa-tions were not photometric. These factors mean that our absolute flux calibrationis not reliable, but the relative calibration should not be affected. Both spectra havebeen reduced using the data reduction package IRAF following standard proce-dures. A Galactic extinction correction of E(B − V) = 0.049 (Schlegel et al. 1998)was applied. The wavelength resolutions of blue- and red-arm spectra respectively

3.2 The optical afterglow spectra 53

Figure 3.1 — The WHT ISIS combined, normalised spectrum of the afterglow (midpt0.14 days), and 1σ error spectrum (lower curve). Overlaid is the best-fitting DLA profile(solid line) and its errors (dashed lines). All significant lines (3σ) are indicated above forz = 3.969 (solid), z = 3.565 (dashed), z = 1.773 (dot-dashed) and unidentified (dotted)systems; see on-line table for details. Lines used in further analysis are labelled.

are 8.7 and 8.1 Å. The signal to noise per pixel, measured at 6800 Å, is 27 in thefirst and 17 in the second spectrum.

3.2.2 Results

The spectrum, shown in Fig. 3.1, is rich in line features at z = 3.97, 3.56 and1.77. A strong Damped Lyman-Alpha absorption system (DLA) is present; herewe focus on this and a selection of metal lines presumed to originate in the GRBhost galaxy. We fitted a power law continuum corrected for Galactic extinction tothe ∼6500–7500 Å region of each spectrum, excluding the absorption lines, andfind an epoch averaged slope of β = −1.34 ± 0.21 (2σ formal fit error). We testedfor any departure from a pure power law due to host-galaxy extinction: fitting MW,LMC and SMC extinction curves (Pei 1992) all result in epoch averaged AV = 0.01.The optical/IR spectral slope from published BVRIJ photometry extrapolated to acommon epoch using a temporal decay slope of 0.89 (Haislip et al. 2005; Holmanet al. 2005; Cobb & Bailyn 2005; Blustin et al. 2005) gives β ∼ −1, consistent withthe spectral analysis but not very constraining.

Despite the moderate dispersion of the ISIS 300 grisms, the damping wingsof the host galaxy DLA are clearly visible. In fact, the determination of N(H I) inDLAs based on long slit spectra is considerably simpler than for echelles. Since thedamped profile may extend over many spectral orders in a typical echelle, accuratecombination and flux calibration can be troublesome. Using the Starlink softwareDIPSO, we determine log N(H I) = 22.1 ± 0.1 (see Fig. 3.1). Taking Lyβ intoaccount did not lead to a more accurate determination of N(H I), and the error on ourfit is dominated by uncertainties in the determination of the power law continuum.

54

The N(H I) value is consistent with, although slightly lower than, that reportedby Chen et al. (2005a). This high value (the first DLA to break the 1022 atomscm−2 barrier) continues the trend amongst GRB DLAs towards very high neutralhydrogen columns (e.g. Jensen et al. 2001; Hjorth et al. 2003a; Vreeswijk et al.2004).

Although our spectra do not enable as detailed a study of the metal lines asis possible via echelle observations (e.g. Chen et al. 2005a), we briefly commenton a selection of these. Detection limits are quoted at the 3 σ level. Althoughwe detect both S IIλ, λ 1253, 1259, both lines are likely to be at least partiallysaturated. In addition, the weaker S II λ 1253 line which potentially offers a betterlimit on N(S II) is blended with another (unidentified) feature (H.-W. Chen, 2005,private communication). We determine an upper limit of [S/H] < −2.0 based onthe absence of the weaker S II λ 1250 Å line1, in good agreement with Chen et al.2005a). Similarly, from the Fe II λ1608 line which is partially saturated and theundetected Fe II λ1611, we determine −2.9 < [Fe/H] < −1.9.A search for variability in line features between our two spectra, separated by ∼25mins, revealed no significant changes (see on-line table for details), neither did theN(H I) column vary.

3.3 The absorbed X-ray afterglow

3.3.1 Observations

We have analysed the early Swift XRT data, to look for evidence of intrinsic ab-sorption in the X-ray spectrum. The XRT data consist of Windowed Timing (WT)mode data for the first orbit (133 to 793 seconds after the trigger) and start of thesecond orbit, and Photon Counting (PC) mode data for later orbits. The data werereduced using the standard pipeline for XRT data within the HEADAS 6.0 package(Swift software version 2.0). WT mode data was extracted using a rectangular re-gion centred on the source, and a similar area in a source-free region of the sameimage to determine the background level. PC mode data was extracted using a cir-cular aperture, except for orbits 2 to 4 which show evidence of pile-up (count rates� 0.8 counts s−1) and were extracted using an annular region centred on the sourceand filtered on grade 0 only. The light curve was obtained between channels 30 and1000 (spanning ∼0.3–10 keV). Spectral analysis was done using XSpec 11.3, withthe standard Ancillary Response Function (ARF) files, which estimate the effectivetelescope area, for PC mode data, and with ARFs based on ray-tracing (‘physical’ARFs) for WT mode data which should provide a better calibration at low energies.

1S abundance of log (S/H) + 12 = 7.20 (Grevesse & Sauval 1998

3.3 The absorbed X-ray afterglow 55

0

1

2

3

4

NH

(×10

22cm

2)

NH

(×10

22cm

2)

100 200 500 1000Time since trigger (seconds)Time since trigger (seconds)

1.25

1.5

1.75

2

ΓΓ

10

20

50R

ate

(cou

nts/

sec)

Rat

e(c

ount

s/se

c)

Figure 3.2 — Evolution of the 0.3–10 keV count rate (top), 0.2–10 keV power law slope,Γ, (middle, note that Γ = 1−β) and additional equivalent hydrogen column, NH,int, at theredshift of the host galaxy (bottom) during the first ∼800 s of the Swift XRT observations(90% errors). The individual spectra are shown with filled circles; the two combinedspectra pre- and post-500 s, for which better constraints are obtained, are shown withdiamonds.

3.3.2 Results

The first orbit shows several flares in the light curve, first reported for this afterglowby Grupe et al. (2005). We have performed a detailed analysis of the spectralevolution of the early-time data. The fitted model consists of a power law plusGalactic absorption (fixed at 3.05 × 1020 cm−2 Dickey & Lockman 1990) and avariable Galactic-like absorption component with Solar metallicity and z = 3.97.Errors are quoted at the 90% confidence level for 1 interesting parameter. We findevidence for a change in power law photon index, from Γ = 1.52 ± 0.04 at the startof the first orbit to 1.79 ± 0.06 at the end of the orbit (note that Γ = 1 − β). We alsofind evidence for an excess absorption column, which at the redshift of the burstamounts to an intrinsic column of NH,int = (1.4±0.3)×1022 cm−2. However, around500 seconds post trigger, the absorption column abruptly changes, becoming lower

56

by about a factor of 4: NH,int = (3.4 ± 2.7) × 1021 cm−2. In the late-time PC modespectrum the intrinsic column cannot be constrained, setting an upper limit of onlyNH,int ≤ 1.0×1022 cm−2, and the power law photon index remains stable at Γ ∼ 1.77.We have checked for a possible correlation between the intrinsic NH,int and Γ in thefit. Contour plots for the intervals 133–503 s and 503–793 s post trigger showno evidence for any correlation, confirming the reality of both the drop in NH,int

and increase in Γ (Fig. 3.2). Interestingly, this happens directly after the peak ofthe second visible flare, where the light curve intensity has increased by a factorof 3. Given the host galaxy metallicity we measure in the optical spectrum, weadjust the X-ray absorption model accordingly. Using Z = Z�/100 for all theelements heavier than He included in the zvphabs X-ray absorption model, therequired intrinsic equivalent hydrogen column increases by a factor of ∼10 in bothcases to NH,int = 9.5+2.3

−2.1 ×1022 cm−2 (first 400 s) and NH,int = 2.6+1.9−1.6 ×1022 cm−2 (≥

500 s post trigger) with approximately the same goodness of fit.A preliminary analysis of published optical photometry together with the PC

mode XRT spectrum has shown the X-ray and optical slopes at 0.19 days (βX ∼−0.7 to −0.8, βopt ∼ −1.0 to -1.5) to likely be incompatible with a position ofthe cooling break between the optical and xrays, and might suggest the presenceof an inverse Compton component; we await the availability of further optical/IRphotometry for a full analysis.


3.4.1 Host galaxy properties

There is a well known relationship between galaxy luminosity and metallicity (e.g.Garnett 2002; Lamareille et al. 2004) which spans 6 orders or more of magnitude inMB. Tremonti et al. (2004) have recently demonstrated that this relation is driven byan underlying relation between mass and metallicity. The cause of the relationship,they argue, is due to the increased gravitational potential of massive galaxies whichenhances metal retention. In the absence of a detected host for GRB 050730 at thetime of writing, it is in principle possible to use the luminosity-metallicity (LZ)and mass-metallicity (MZ) relations to predict the MB and stellar mass of the host.Both of these relations are best determined locally (e.g. Lamareille et al. 2004),although sizeable datasets have now investigated the LZ relation up to z ∼ 1 (e.g.Kobulnicky et al. 2003; Kobulnicky & Kewley 2004). There is clear evidence forevolution in the LZ relation, in the sense that galaxies are more metal-poor for theirluminosity at higher z (although see caveats in Kewley & Ellison in prep.). Thistrend appears to continue both for the LZ and MZ relations up to z ∼ 3 (e.g. Shap-ley et al. 2004; Moller et al. 2004, Erb et al. in prep.), although only the highestmass/luminosity galaxies are bright enough to be included in spectroscopic sam-ples. The lowest metallicity bin in the fitted MZ relation of Erb et al. (in prep.) is


Table 3.1 — Lines detected above 3σ in the first and second epoch WHT spectra a.

λ Wobs(I) Wobs(II) ID z4683.0b Lyβ λ1025.722 3.56565090.6b Lyβ λ1025.722 3.96295549.2b Lyαλ1215.668 3.56476031.5b Lyαλ1215.668 3.96156230.8 1.42 ± 0.24 1.19 ± 0.38 S IIλ1253.805 3.96956261.9 3.64 ± 0.25 3.50 ± 0.41 Si IIλ1260.422 3.9681

blended with S IIλ1259.518 3.97176283.9 3.95 ± 0.23 3.45 ± 0.33 Si II∗ λ1264.738 3.96856364.2 0.51 ± 0.17 0.78 ± 0.35 C I? λ1280.135 3.97156470.8 1.95 ± 0.17 2.43 ± 0.39 O Iλ1302.169 3.96926482.8 2.57 ± 0.14 2.87 ± 0.28 O I∗ λ1304.858 3.9682

blended with Si IIλ1304.370 3.97006489.8 0.50 ± 0.11 0.82 ± 0.20 O I∗∗ λ1306.029 3.96916504.6 1.86 ± 0.16 2.17 ± 0.35 Si II∗ λ1309.276 3.9681

blended with? Fe IIλ2344.214 1.77476544.6 0.44 ± 0.13 0.47 ± 0.176583.9 0.42 ± 0.10 0.55 ± 0.186588.9 0.55 ± 0.11 0.57 ± 0.176597.1 0.82 ± 0.12 1.33 ± 0.266607.0 1.24 ± 0.14 1.58 ± 0.23 Fe IIλ2382.765 1.77286633.4 5.12 ± 0.19 5.07 ± 0.31 C IIλ1334.532 3.9706

blended with C II∗ λ1335.663 3.96646721.0 1.05 ± 0.17 0.73 ± 0.206924.1 2.43 ± 0.15 2.42 ± 0.32 Si IV λ1393.760 3.96796934.0 0.88 ± 0.09 1.18 ± 0.246968.7 2.67 ± 0.15 2.32 ± 0.28 Si IV λ1402.773 3.9678

blended with Si IIλ1526.707 3.56456989.4 1.44 ± 0.17 1.28 ± 0.227000.9 0.52 ± 0.09 0.86 ± 0.24 Si II∗ λ1533.432 3.56557066.1 0.42 ± 0.12 0.20 ± 0.19 C IV λ1548.204 3.56417077.7 0.32 ± 0.12 0.48 ± 0.18 C IV λ1550.781 3.56407107.4 0.62 ± 0.14 0.61 ± 0.227172.6 0.71 ± 0.14 0.92 ± 0.25 Fe IIλ2586.650 1.77297209.0 1.46 ± 0.17 0.86 ± 0.20 Fe IIλ2600.173 1.77257586.4c 2.53 ± 0.18 2.48 ± 0.26 Si IIλ1526.707 3.9691

58

Table 3.1 — – continued

7628.5c Al IIλ1670.789 3.56587692.9 3.79 ± 0.14 3.86 ± 0.29 C IV λ1548.204 3.96897705.3 3.66 ± 0.20 4.22 ± 0.36 C IV λ1550.781 3.96877754.9 4.18 ± 0.34 5.15 ± 0.63 Mg IIλ2798.743 1.77097775.5 2.28 ± 0.25 2.67 ± 0.36 Mg IIλ2803.532 1.77357994.3 3.14 ± 0.54 3.27 ± 0.74 Fe IIλ1608.451 3.9716

a Blueward of Lyα the low resolution and Lyα forest hampers secure iden-tification of metal lines, which we therefore do not list.

b Due to the uncertain continuum level, we do not attempt to measure thewidths of the Lyα and Lyβ lines.

c This equivalent width measurement is seriously affected or made impossi-ble by the atmospheric absorption band from 7584–7675Å.

Z ∼ Z�/3 corresponding to a stellar mass log(M�/M�) ∼ 9.5. The metallicity mea-sured from absorption lines in the optical afterglow considered here is Z ∼ Z�/100,which indicates that the host is not a massive, luminous Lyman break galaxy (LBG),although Jakobsson et al. (2005a) argue that GRB hosts follow the same UV lumi-nosity function as the faint LBGs. We do note that the MZ relation is based onemission lines. However, HST imaging has shown that GRBs occur in regions ofstrongest star formation (e.g. Fruchter et al. 2006), justifying our assumption thatthe absorption lines are formed in the same regions as the higher wavelength emis-sion lines. Combining the measured N(H I) with the metallicity and assuming anSMC gas-to-reddening ratio (Bouchet et al. 1985), we can estimate the extinctionassociated with the GRB host galaxy and compare this to the values obtained fromthe optical continuum fits. Negligible E(B − V) is determined by both methods,consistent with the similarly small amounts of dust seen towards intervening DLAs(e.g. Murphy & Liske 2004; Ellison et al. 2005).

3.4.2 The neutral hydrogen column

GRB 050730 has the strongest DLA seen in a GRB afterglow spectrum, with ahydrogen column density of log N(H I) = 22.1 ± 0.1. The X-ray absorption at latetimes scaled to Z�/100 yields a comparable log NH,int = 22.4+0.2

−0.4 (assuming theNH,int measured at ∼500–800 s post burst in the WT mode XRT spectrum can beextrapolated to a few hours post burst). The NH,int we measure in the early-timeX-ray spectra covering ∼133–500 s post trigger is about ten times higher than that


measured at t > 500 s. The change in X-ray absorbing column could be caused byionisation by the gamma-ray jet, or by the X-ray flares which are suggested to becaused by prolonged central engine activity (Burrows et al. 2005; King et al. 2005).

It should be noted that what is measured in the X-ray models is an equivalenthydrogen column, since primarily metal edges contribute to the X-ray absorption atthe redshift of GRB 050730, and that this is highly dependent upon the metallicityassumed (see e.g. Wilms et al. 2000). There will be a contribution to the X-ray ab-sorption from intervening systems, which cannot be disentangled from absorptionin the host, particularly given that we do not know the metallicity of the closest in-tervening system observed in this spectrum (z = 1.77). In principle, a lower columnvery close to the observer could have a similar effect on the spectrum as a largecolumn at high redshift.

The observed X-ray column variability does, however, lead us to conclude thatmost of the X-ray absorbing gas in GRB 050730 is located close to the GRB. Theoptical H I column remained stable over the ∼25 mins between our ISIS spectra,taken at 0.132 days since burst, well after the observed X-ray flaring (although theoccurrence of X-ray flares at later times cannot be ruled out owing to low countrates). The H I creating the DLA is likely to be located much further away from theGRB, unaffected by the GRB radiation. We would expect to observe destructionby the GRB of dust co-located with the X-ray absorbing gas. Our spectra imply avery low extinction in the host at ∼3 hrs post burst. Future prompt optical spectra,in conjunction with X-ray observations, are required to investigate this further.

Acknowledgements We thank J. P. U. Fynbo for comments on and improvements tothe manuscript and P. A. Curran, A. J. van der Horst, K. L. Page and S. Vaughan for use-ful discussions. This work is based on observations made with the WHT operated on theisland of La Palma by the Isaac Newton Group in the Spanish Observatorio del Roquede los Muchachos at the Instituto de Astrofisica de Canarias - we thank N. O’Mahonyfor excellent support. The authors acknowledge support from and collaboration withinthe EU-funded Research Training Network ‘Gamma-Ray Bursts: an enigma and a tool’(HPRN-CT-2002-00294).

4Gamma-Ray Burst afterglows as probes of

environment and blastwave physics: absorption byhost galaxy gas and dust

R. L. C. Starling, R. A. M. J. Wijers, K. Wiersema, E. Rol, P. A. Curran, C.Kouveliotou, A. J. van der Horst and M. H. M. Heemskerk

Astrophysical Journal, 661, 787 (2007)

Abstract We use a new approach to obtain limits on the absorbing columnstowards an initial sample of 10 long Gamma-Ray Bursts observed with BeppoSAXand selected on the basis of their good optical and nIR coverage, from simultaneousfits to nIR, optical and X-ray afterglow data, in count space and including theeffects of metallicity. In no cases is a MW-like extinction preferred, when testingMW, LMC and SMC extinction laws. The 2175Å bump would in principle bedetectable in all these afterglows, but is not present in the data. An SMC-likegas-to-dust ratio or lower value can be ruled out for 4 of the hosts analysed here(assuming SMC metallicity and extinction law) whilst the remainder of the samplehave too large an error to discriminate. We provide a more accurate estimate ofthe line-of-sight extinction and improve upon the uncertainties for the majority ofthe extinction measurements made in previous studies of this sample. We discussthis method to determine extinction values in comparison with the most commonlyemployed existing methods.

61

62 GRB afterglows as probes of environment

4.1 Introduction

The accurate localisation of Gamma-Ray Bursts (GRBs) through their optical andX-ray afterglows has enabled detailed studies of their environments. Selection bythe unobscured gamma-ray flash alone has allowed the discovery of a unique sam-ple of galaxies which span an enormously wide range of redshifts from z ∼ 0.009(GRB 980425, e.g. Tinney et al. 1998; Galama et al. 1998b) to 6.3 (GRB 050904Kawai et al. 2005, 2006). The subset of long-duration (> 2 s) GRBs are almost cer-tainly caused by the collapse of certain massive stars to black holes, confirmed byobservations of supernova components in the late-time afterglows of a number oflong-GRBs (Woosley & Bloom 2006; Kaneko et al. 2007) and by the observed lo-cation of GRBs in UV-bright regions within their host galaxies (Bloom et al. 2002;Fruchter et al. 2006). GRBs are located in host galaxies which are generally small,faint, blue and highly star-forming (e.g. Chary et al. 2002; Fruchter et al. 1999; LeFloc’h et al. 2003). Hence, detailed and extensive host galaxy observations pro-vide a wealth of information on the gas and dust properties of star-forming galaxiesthroughout cosmological history.

Accurately measuring the dust content of these galaxies is of great importancein, to name one example, the determination of their unobscured star formation rateswhere uncertainty in the correction for dust can easily dominate the errors on themeasured star formation rates for high redshift galaxies (e.g. Pettini et al. 1998;Meurer et al. 1999). Absorption within our own Galaxy along a particular line ofsight can be estimated and removed, but absorption which is intrinsic to the GRBhost galaxy as a function of wavelength is unknown, and is especially difficult todetermine given its dependence on metallicity and the possible existence of dustyintervening systems whose extinction curves cannot be disentagled from those ofthe host galaxy. Afterglow spectroscopy and/or photometry can be used to pro-vide an estimate of the total extinction along the line-of-sight to the GRB. If thehost galaxy itself is bright and extended enough to be observed once the afterglowhas faded, different lines-of-sight may be probed besides the off-centre UV-brightregions within which GRBs are generally situated.

Extinction in the optical/UV regime due to dust grains is typically modelled us-ing either Milky Way (MW or Galactic), Large Magellanic Cloud (LMC) or SmallMagellanic Cloud (SMC) extinction curves (e.g. Pei 1992) because these curvescan be measured and so are well known, or with the Calzetti extinction law derivedempirically from UV observations of starburst and blue compact galaxies (Calzettiet al. 1994). It appears that dust content in most GRB hosts produces an SMC-likeextinction law (e.g. Galama & Wijers 2001; Vreeswijk et al. 2004; Stratta et al.2004; Kann et al. 2006, Schady et al. in preparation), owing to an observed lack ofthe 2175Å feature thought to be caused by carbonaceous dust grains (Draine & Lee1984). This feature has, however, been clearly observed in GRB afterglow spec-tra where the line of sight between us and the GRB is intercepted by interveningsystems: the best example to date, in which the extinction curve of an interven-

4.1 Introduction 63

ing system could for the first time be disentangled from that of the host galaxy, isGRB 060418 (Ellison et al. 2006).

In general, low amounts of optical extinction are found towards GRBs, unex-pected if GRBs are located in dusty star-forming regions, whilst the X-ray spectrareveal a different picture (first noted by Galama & Wijers 2001). At X-ray wave-lengths absorption is caused by metals in both gas and solid phase, predominantlyoxygen and carbon (see e.g. Wilms et al. 2000), and we often measure high val-ues for the absorbtion columns. However, these absorption edges are shifted outof the X-ray observing window for high redshifts (z � 2) beyond which only largecolumns can be measured and there is a degeneracy between redshift and X-ray col-umn density (e.g. Watson et al. 2002). The GRB host metallicity is observed to below (compared with the Milky Way) in measurements via optical spectroscopy ofa dozen or so afterglows. Host metallicities can reach values as low as 1/100 Solar(GRB 050730 Starling et al. 2005a; Chen et al. 2005a, but see also Prochaska 2006for potential caveats) - even lower than found for the SMC (see Figure 3 of Fynboet al. 2006a for an overview). This only increases any measured X-ray column,which is expressed as an equivalent hydrogen column density, NH. But here wenote that metallicities are not generally obtainable for lower redshift GRBs (z � 2)due to the hydrogen Lyman-α line lying in the far UV outside typical observingwindows.

The apparent discrepancy between optical and X-ray extinction resulting inhigh gas-to-dust ratios in GRB host galaxies (often far higher than for the MW,LMC or SMC, e.g. GRB 020124, Hjorth et al. 2003a, but see Schady et al. inpreparation) is not satisfactorily explained, though the suggestion that dust destruc-tion can occur via the high energy radiation of the GRB (e.g. Waxman & Draine2000) could possibly account for the discrepancy. It is thought that circumburst dustmay be destroyed by sublimation of dust grains due to UV emission (Waxman &Draine 2000), sputtering (Draine & Salpeter 1979) or dust grain heating and charg-ing (Fruchter et al. 2001). Alternative models for the extinction by dust grains,including skewing the dust grain size distribution towards larger grains, have beeninvestigated, and in fact such a grain size distribution may result from exposure ofthe dust to the GRB radiation field since destruction of small grains is more efficientthan for larger grains (Perna et al. 2003). Attempts to model the process of dust de-struction have been made by e.g. Perna & Lazzati (2002). However, such modelshave not replaced SMC-like (low metallicity) extinction as the best description ofmost GRB environments (e.g. Stratta et al. 2004; Kann et al. 2006).

Traditionally the optical and X-ray spectra have been treated seperately in ex-tinction studies. Since the underlying spectrum is likely a synchrotron spectrum(power law (pl) or broken power law (bknpl), e.g. Wijers & Galama (1999), ex-tending through both wavelength regions, it would be most accurate to performsimultaneous fits. More recently such fits have been made, either by fitting theX-ray spectrum individually and thereby transforming the model counts to flux tocreate a spectral energy distribution (SED) with the optical data in flux space (e.g.


Stratta et al. 2005; Watson et al. 2006), or by using the Swift XRT X-ray and UVOTUltraviolet (UV) and U,B,V band data together in a fit to the count spectrum (e.g.Blustin et al. 2006, Schady et al. in preparation). The Swift UVOT data can beloaded directly into the xspec spectral fitting package and treated in count spaceowing to its calibration, which is generally not true for ground-based data. Wepresent here an alternative method, which makes use of simultaneous fits in countspace extending from near-infrared (nIR, in this case K band) to X-ray (10 keV) toobtain the most accurate possible measurements of both the underlying continuumspectrum and the extinction.

In Section 4 we fit the broad-band SEDs (from nIR through X-ray) of a sub-sample of the BeppoSAX sample of GRB afterglows to better measure the extinctionproperties of their host galaxies - a sample chosen for its availability of suitable datawell studied in the seperate optical and X-ray band passes. Section 2 outlines thedata sample and reduction techniques. Section 3 describes the method we use tomodel the broad-band SEDs and Section 4 presents the results of our fitting throughdiscussion of individual bursts and comparison with previous studies. In Section 5we discuss the implications of our findings for galaxy extinction curves, and sum-marise and compare the various methods now available to measure extinction inthe hosts. We conclude by summarising our method and findings in Section 6. Ananalysis of the blastwave parameters and density profiles for the circumburst mediaobtained from these fits will be presented in a forthcoming paper (Starling et al.2007, paper II).

4.2 Observations

This sample of 10 long GRBs observed with the BeppoSAX Narrow Field Instru-ments (approx. 0.1–10 keV) is chosen for the good availability (3 bands or more)of optical/nIR photometry (Tables 1 and 2). The optical/nIR bands available foreach source and their references are listed in Table 4.3. As these GRBs are all pre-viously studied, overlapping with the samples studied by Galama & Wijers (2001)and Stratta et al. (2004), this constitutes a good sample on which to first adopt thismethod of simultaneous SED fitting.

All X-ray observations are taken from the BeppoSAX data archives, using theLECS and MECS instruments raw data within the energy ranges 0.1–4 and 1.0–10 keV respectively. Data have been reduced using the SAXDAS routines. Wecombined data from the MECS2 and MECS3 instruments (except in the case ofGRB 970228, where we use the MECS3 instrument only, see Stratta et al. (2004)),including a gain equalisation. We then combined multiple observations for eachsource and instrument type among the narrow field instruments, omitting the lastobservation if it was ≥ 3 days later than the previous one, before extracting spectra.Background X-ray spectra were taken from blank fields and count rates checkedagainst the local background finding no adjustments necessary: the net count rates

4.3 Method 65

Table 4.1 — BeppoSAX data table for combined datasets. We have measured the MECS–LECS offset values from the combined X-ray spectra with respect to the MECS instru-ment.

GRB obs start obs end obs midpt(log) ton−source (s) MECS–LECS(days since trigger) MECS LECS offset

970228 0.344 0.693 0.520 1.43×104 5.611×103 1.45+0.67−0.42

970508 0.434 3.091 1.679 5.90×104 2.36×104 1.09+0.50−0.43

971214 0.274 2.528 1.362 1.01×105 4.62×104 1.57+0.67−0.56

980329 0.294 2.026 1.148 6.38×104 2.49×104 0.68+0.58−0.36

980519 0.406 1.468 0.930 7.82×104 2.31×104 1.00+0.94−0.64

980703 0.827 1.902 1.333 3.92×104 1.66×104 0.99+0.36−0.34

990123 0.242 2.573 1.245 8.20×104 2.80×104 0.77±0.9990510 0.334 1.850 1.067 6.79×104 3.17×104 0.86+0.13

−0.12000926 2.003 2.466 2.234 1.96×104 5.027×103 0.70+2.13

−0.67010222 0.376 2.703 1.511 8.84×104 5.11×104 1.43+0.14

−0.13

of the two types of field agree on average to within 0.0001 counts s−1. The latestcanned arf and rmf files were used with MECS data, but were created for LECSobservations at their off-axis source positions (listed in Table 3 of Stratta et al.2004). We group all spectra such that a minimum of 20 counts are in each binin order to use the χ2 statistic. However, in the case of GRB 000926 there werevery few counts in the X-ray spectrum so we have required only 10 counts perbin. There is a known offset between the normalisations of the LECS and MECSinstruments. We fit for this offset in the X-ray spectra only, adding a constant-value free parameter to the model and adopting MECS as the reference for theLECS instrument. We fix the offset values in the SED of each GRB to these values(Table 4.1).

All temporal decay slopes, both for X-ray and optical lightcurves, have beentaken from the literature and are listed together with Galactic extinction correctionsin Table 4.2. Optical and nIR photometry was taken from the literature and from ourown nIR observations of GRB 990510 (described in Curran et al. in preparation).

4.3 Method

Per source all data are fitted simultaneously, assuming wherever possible no priormodel. This is achieved by fitting in count space (as is traditional in the X-rayregime where one fits for the emission model, extinction and instrumental responsesimultaneously): the optical and nIR magnitudes are converted to flux and then tocounts. For the magnitude to flux conversion we use the zero points and effectivebandwidths of each optical band (Johnson for U, B, V , R, I, J, K, 2MASS forH and Ks and Bessel for Vc, Rc, Ic). In the small number of cases for which the


Table4.2

—G

RB

known

properties:G

alacticabsorption

(columns

2-3,×10

22cm−

2,andcolum

n4),redshift,opticaltem

poraldecayslope(s)

and(jet)

breaktim

ein

dayssince

triggerandX

-raytem

poralslope.

GR

BG

al.N

H(1)

Gal.

NH

(2)E

(B−V

)Gal

redshiftzα

1 ∗α

2 ∗tbk ∗

αx+

9702280.165

0.1340.203

0.6950±0.0003

a1.46±

0.15-

-1.3±

0.2970508

0.05260.0485

0.0500.835±

0.001b

1.24±0.01

--

1.1±0.1

9712140.0167

0.01280.016

3.418±0.010

c1.49±

0.08-

-1.6±

0.1980329

0.09180.0916

0.0733.6

d0.85±

0.12-

-1.5±

0.2980519

0.1830.189

0.267-

1.50±0.12

2.27±0.03

0.48±0.03

1.83±0.3

9807030.0579

0.04980.057

0.9661±0.0001

e0.85±

0.841.65±

0.461.35±

0.940.9±

0.2990123

0.02130.0165

0.0161.600±

0.001f

1.24±0.06

1.62±0.15

2.06±0.83

1.44±0.11

9905100.0924

0.08150.203

1.619±0.002

g0.92±

0.022.10±

0.061.31±

0.071.4±

0.1000926

0.02650.0220

0.0232.0379±

0.0008h

1.74±0.03

2.45±0.05

2.10±0.15

1.7±0.5

0102220.0163

0.01750.023

1.4768±0.0002

i0.60±

0.091.44±

0.020.64±

0.091.33±

0.04

(1)taken

fromD

ickey&

Lockm

an(1990)(resolution

of∼1 ◦)

(2)taken

fromthe

Leiden

/Argentine/B

onnG

alacticH

ISurvey,K

alberlaetal.(2005)(resolution

of∼0.6 ◦)

∗taken

fromZ

ehetal.(2006),w

hereuncertainties

are1σ

+taken

fromG

endre&

Boer

(2005)and

in’t

Zand

etal.

(1998)(G

RB

980329,M

EC

S2–10

keVdata),

Nicastro

etal.

(1999)(G

RB

980519)

aB

loometal.(2001a)

bB

loometal.(1998a)

cK

ulkarnietal.(1998)d

photometric

redshiftonly,Jaunsenetal.(2003)

eD

jor-govskietal.(1998)

fKulkarnietal.(1999a)

gV

reeswijk

etal.(2001)h

Castro

etal.(2003)iM

irabaletal.(2002)

4.3 Method 67

specific band is not stated, we assume the appropriate Johnson filter. These fluxesare then converted to photons cm−2 s−1 per bin (bin width = effective bandwidth ofthe filter) within the ISIS spectral fitting program (Houck & Denicola 2000) whichis equivalent to the X-ray units of counts cm−2 s−1 per bin since total number ofcounts is conserved.

Since we fit in count space, we need not first assume a model for the X-rayspectrum to convert the counts to flux. Herein lies one advantage of using thismethod. The second advantage comes through the multiwavelength approach. ThenIR, optical and X-ray spectra are related since we assume the broadband spectrumis caused by synchrotron emission, hence a simultaneous fit provides greater ac-curacy and consistency between the parameters. Inclusion of nIR data and R bandoptical data together with the 2–10 keV X-ray data, regions over which extinctionhas the least effect, allows the underlying power law slope to be most accuratelydetermined.

The X-ray data typically have much longer exposure times than the individualoptical/IR measurements, particularly given that where sensible we have combinedX-ray datasets from different epochs to increase the signal to noise. We have chosento omit the last X-ray observation of any GRB which occurs more than 3 days afterthe previous observation where this contributes more to the background noise thanthe signal and skews the observation time midpoint. We note that structure in the X-ray lightcurves and spectral changes at early times have often been reported for thebetter and earlier sampled Swift GRB afterglows (e.g. Nousek et al. 2006), whichmay affect the BeppoSAX afterglows with fairly early coverage, i.e. those beginningat 0.2–0.3 days since burst (see Table 4.1). All optical data are scaled to a commontime, which corresponds to the midpoint of the X-ray observations, calculated inlog-space. This is done by extrapolation according to the decay rates and lightcurvebreak times determined in a thorough analysis of pre-Swift burst optical lightcurvesby Zeh et al. (2006), Table 4.2).

We fit for the X-ray and optical extinction at the redshift of the GRB. ForGRBs 980329 and 980519 the redshift is not known, and therefore absorption atthe source cannot be measured. However, a photometric redshift has been madefor GRB 980329 of z ∼ 3.6 (Jaunsen et al. 2003) which we adopt here. ForGRB 980519 we adopt the mean redshift for this sample, z ∼ 1.58, to make anestimate of the intrinsic absorption required.

Flux is depleted in the bluemost optical bands for high redshift bursts due tothe Lyman absorption edges and the resulting transmission is calculated for eachband per burst (see Curran et al. in preparation for further details of this calculation,which we note involves assuming a spectral slope for the optical flux over a givenband and uses parameters from Madau 1995 and Madau et al. 1996). However,this has an effect only on the U through R band magnitudes of GRBs 971214 and980329 and a minor effect on the U band magnitudes of GRBs 980519, 990123,990510 and 000926.

The errors on the optical magnitudes are taken from the photometric errors in


Table 4.3 — Range of optical and nIR magnitudes used in this study.

GRB bands used refs

970228 VRcIc [1]970508 UBVRcIcKs [2,3]971214 VRIJKs [4,5,6,7]980329 RIJK [8]980519 UBVRcIc [9]980703 RIJHK [10]990123 UBVRIHK [11]990510 BVRIJHKs [12,13]000926 UBVRIJHK [14]010222 UBVRIJ [15]

1 Galama et al. (2000) 2 Galama et al. (1998a) 3 Chary et al. (1998) 4 Halpernet al. (1998) 5 Diercks et al. (1997) 6 Tanvir et al. (1997) 7 Ramaprakash et al.(1998) 8 Reichart et al. (1999) and references therein (their Table 1) 9 Jaunsenet al. (2001) 10 Vreeswijk et al. (1999) 11 Galama et al. (1999) 12 Stanek et al.(1999) 13 Curran et al. in prep. 14 Fynbo et al. (2001a) 15 Masetti et al. (2001)

the literature, and are set to 0.1 for cases where the literature reports a smallererror to account for systematic uncertainties. The conversion from magnitudes toflux has an associated error of up to 5 % (Fukugita et al. 1995). We note that byextrapolating magnitudes to different times we introduce a possible random error,since there is an uncertainty in the measured decay indices and break times, whichwe can allow for in an offset between the optical and X-ray data. This error is notincluded in the individual data points because this would introduce an artificiallylarge error in the optical slope which is not in fact present since the relative errorsbetween the bands do not change.

4.3.1 Models

Data are fitted within the spectral fitting package ISIS (Houck & Denicola (2000))using both models written for use within xspec (Arnaud 1996) and models writtenfor ISIS. Models consist of either a single or a broken power law, to allow for apossible cooling break in between the optical and X-rays. Should the break in thepower law be due to cooling, the difference in slope is Δβ = 0.5 (e.g. Wijers &Galama 1999), which we fix in the broken power law model.On its way to us, the intrinsic power law is absorbed by optical extinction at thehost/burst redshift (this could be local to the GRB itself or another location withinthe host galaxy). The extinction curves used for intrinsic optical extinction in this

4.4 Results 69

study are Galactic-, SMC- and LMC-like (Figure 4.1) following the prescriptionsof Cardelli et al. (1989) and Pei (1992) respectively. We do not use the Calzettiextinction curve (Calzetti et al. 1994) because it has a larger error associated with it,being constructed from fewer measurements than those for the nearby MagellanicClouds. There is also absorption in the X-ray regime predominantly by metals,e.g. the oxygen edge, for which we use a photoelectric absorption model. We referthe reader to Wilms et al. (2000) for a detailed description of the X-ray absorptionmodels. The X-ray absorption model can be computed for various metallicitiesby simple scaling of the Solar abundances by a constant factor. We adopt firstlySolar metallicity and secondly the metallicity assumed in the optical extinctionmodel: using SMC-like absorption one would adopt Z=1/8 Z� and for LMC-likeabsorption Z=1/3 Z� (Pei 1992), for self-consistency.

The flux is then corrected for Galactic absorption (Table 4.1). In the X-rayregime these values are fixed at the NH values given in Dickey & Lockman (1990),which are averages over 1 degree at the positions as given in the Simbad cataloguesor from the BeppoSAX Narrow-field instruments. For completeness and compari-son, we also list in Table 4.1 extinction measurements from the newer and slightlyhigher resolution Leiden/Argentine/Bonn Galactic H I Survey (resolution of ∼0.6◦Kalberla et al. 2005). These new values are not significantly different than those ofDickey & Lockman though appear to be generally lower, and to date all previousstudies use the values from Dickey & Lockman which we will also use here. Forthe optical extinction we use E(B−V)Gal values given by Schlegel et al. (1998) fromtheir full-sky 100 μm map together with the Galactic extinction curve of Cardelliet al. (1989) with RV = AV/E(B − V) = 3.1. The Schlegel et al. (1998) maps havea best resolution of 6.1’: for each source we use the best resolution available, butin some cases we must use an average over 1 square degree centred on the sourcecoordinates.The fit statistic calculated is χ2, using a Levenberg-Marquardt fit minimisationmethod. Errors in the LECS-MECS offsets (except for GRB 970508), optical decayslopes and redshifts are not propagated through the fitting routines. These valuesare simply fixed from Tables 1 and 2 respectively. We also do not include uncer-tainties on the zero points on our photometric data points (see previous Section).

4.4 Results

The SEDs and the results of fits to the SEDs for all GRBs in the sample are listed inTable 4.4, and best fitting models are shown overlaid on the data in Figure 4.2. Fig-ure 4.3 shows a comparison of the absorption measurements with Galactic, LMCand SMC gas-to-dust ratios, which we discuss in the following section. This plothas been constructed in a number of previous works (e.g. Galama & Wijers 2001;Stratta et al. 2004; Kann et al. 2006, Schady et al. in preparation) but here we showthe observed distribution of E(B − V) and NH for the first time derived simultane-


Table4.4

—R

esults(m

ainparam

eters)of

fitsto

thespectral

energydistributions.

Foreach

GR

Bw

efitted

theSE

Dw

iththe

power

lawand

thebroken

power

lawm

odelsw

ithall

3extinction

models

(MW

,LM

Cand

SMC

).X-ray

column

densitiesgiven

inbrackets

areforthe

appropriatem

etallicities(L

MC

orSMC

),otherwise

Solarmetallicity

isassum

ed.The

LE

CS-M

EC

Soff

setsw

erefixed

atthecentralvalues

listedin

Table2.For

thebroken

powerlaw

models

Γ1=Γ

2-

0.5.Where

thebreak

energiesare

unconstrained(indicated

bythe

letterU

)w

egive

inbrackets

thecentralvalue

derivedand

anylim

itsset.

Allerrors

arequoted

atthe90

%confidence

level(or1.6σ

).T

heF-testprobability

givesthe

probabilitythatthe

resultisobtained

bychance,therefore

asignificantim

provementin

thefit

when

addingone

extrafree

parameter

isindicated

bya

lowprobability.For

GR

B970508,w

heretw

oF-testprobabilities

arelisted

thesecond

isfor

thecom

parisonbetw

eenfits

with

theoptical–X

-rayoff

setfreeand

fixed.Plots

ofthe

dataoverlaid

with

theirbest-fitting

models

areshow

nin

Figure2.

GR

B/m

odelN

H,int (×

1022

atoms

cm−

2)E

(B−V

)intΓ

(2)E

bk(keV

)χ

2/dofF-testprob.

GR

B970228

pl+m

w0+

0.58

−0

0+

0.20

−0

1.72+

0.11

−0.03

11.6/14

pl+lm

c0+

0.57

−0

(0.001+

1.40

−0.001 )

0+

0.17

−0

1.72+

0.10

−0.03

11.6/14

pl+sm

c0+

0.57

−0

(0.01+

2.48

−0.01 )

0+

0.17

−0

1.72+

0.09

−0.03

11.6/14

bknpl+m

w0.54

+0.68

−0.46

0+

0.19

−0

2.06+

0−

0.03

0.32+

0.09

−0.32

7.6/13

2.1×10 −

2

bknpl+lm

c0.54

+0.68

−0.41

(1.22+

1.65

−0.97 )

0+

0.16

−0

2.06+

0.23

−0.03

0.32+

0.09

−0.32

7.6/13

2.1×10 −

2

bknpl+sm

c0.54

+0.68

−0.41

(2.01+

2.96

−1.66 )

0+

0.16

−0

2.06+

0−

0.03

0.32+

0.09

−0.32

7.6/13

2.1×10 −

2

4.4 Results 71

Tabl

e4.

4—

–co

ntin

ued

GR

B97

0508

pl+

mw

1.57+

1.33

−0.8

50+

0.00

8−0

1.94+

0.02

−0.0

138

.7/3

1

pl+

lmc

1.57+

1.33

−0.8

5(3

.83+

3.45

−2.1

4)

0+0.

007

−01.

94+

0.02

−0.0

138

.7/3

1

pl+

smc

1.57+

1.33

−0.8

5(6

.84+

6.72

−4.0

1)

0+0.

007

−01.

94+

0.02

−0.0

138

.7/3

1

bknp

l+m

w1.

93+

0.09

−0.9

00.

043+

0.01

5−0.0

432.

09+

0.19

−0.0

1U

(<0.

27)

31.2/3

01.

2×10−2

bknp

l+lm

c1.

93+

0.05

−0.9

00.

040+

0.01

4−0.0

402.

09+

0.18

−0.0

1U

(<0.

25)

31.7/3

01.

5×10−2

bknp

l+sm

c2.

08+

0.58

−1.2

80.

035±

0.03

52.

12+

0 −0.0

6U

(<0.

26)

32.2/3

02.

1×10−2

optic

al:X

-ray

offse

tfr

ee(s

eete

xt)

pl+

mw

0.71

8+1.

39−0.7

180.

004+

0.06

0−0.0

041.

76+

0.07

−0.2

432

.1/3

01.

9×10−2

pl+

lmc

0.75

6+1.

34−0.6

26(1

.80+

3.41

−1.5

2)

0+0.

054

−01.

78+

0.05

−0.2

332

.1/3

01.

9×10−2

pl+

smc

0.75

6+1.

07−0.6

26(3

.50+

6.36

−2.8

9)

0+0.

052

−0(0

.005+

0.02

7−0.0

05)

1.78+

0 −0.0

132

.1/3

01.

9×10−2

bknp

l+m

w0.

63+

0.49

−0.6

30.

032+

0.07

1−0.0

322.

14+

0.14

−0.2

9U

(3.4

2)29

.7/2

90.

236,

0.13

7

bknp

l+lm

c0.

72+

0.05

−0.7

20.

021+

0.02

2−0.0

212.

17+

0.14

−0.3

25U

(3.4

8)30

.1/2

90.

224,

0.17

6

bknp

l+sm

c0.

88+

0.09

−0.8

80.

007+

0.09

0−0.0

072.

23+

0.08

−0.3

4U

(3.5

6)30

.3/2

90.

188,

0.20

0


Table4.4

—–

continued

GR

B971214

pl+m

w10.88

+15.97

−10.16

0.045+

0.038

−0.042

1.60±0.04

37.6/44

pl+lm

c11.47

+16.16

−10.33

(31.70+

45.78

−28.42 )

0.036+

0.021

−0.023

1.60±0.04

34.4/44

pl+sm

c11.48

+16.15

−10.33

(68.77+

104.3

−61.26 )

0.031±0.018

1.60±0.03

33.2/44

bknpl+m

w28.37

+20.45

−26.53

0.056±0.051

2.04+

0.09

−0.14

U(1.56)

35.8/43

0.149

bknpl+lm

c29.78

+20.30

−25.94

0.044+

0.057

−0.027

2.04+

0.09

−0.14

U(1.40)

31.7/43

6.2×10 −

2

bknpl+sm

c22.12

+26.61

−15.38

0.058+

0.027

−0.040

1.85+

0−

0.12

U(0.048)

30.1/43

4.1×10 −

2

GR

B980329

pl+m

w0.0012

+9.74

−0.0012

0.358+

0.088

−0.050

1.88+

0.07

−0.09

38.4/27

pl+m

w+

zfree

0.001+

9.52

−0.001

0.247+

0.039

−0.042

1.81+

0.04

−0.05

32.7/26

4.3×10 −

2

pl+lm

c0.001

+8.6

−0.001

(0.001+

24.2

−0.001 )

0.210+

0.047

−0.019

1.84+

0.06

−0.07

35.4/27

pl+sm

c0+

8.2

−0

(0.001+

51.3

−0.001 )

0.178+

0.039

−0.030

1.82+

0.04

−0.07

34.3/27

bknpl+m

w3.42

+21.9

−3.42

0.346+

0.148

−0.120

2.34+

0.08

−0.34

U(2.39:

<6.89)

35.4/26

0.150

bknpl+lm

c8.00

+15.7

−8.00

(21.7+

43.9

−21.7 )

0.211+

0.080

−0.064

2.26±0.31

U(1.09:

<5.79)

31.7/26

9.3×10 −

2

bknpl+sm

c7.50

+15.4

−7.50

(42.5+

95.9

−42.5 )

0.179+

0.067

−0.051

2.25+

0.10

−0.31

U(1.08:

<5.54)

30.2/26

7.2×10 −

2

4.4 Results 73

Tabl

e4.

4—

–co

ntin

ued

GR

B98

0519

pl+

mw

0+5.

4−0

0.00

8+0.

015

−0.0

081.

97+

0.05

−0.0

318

.9/2

3

pl+

lmc

0+5.

8−0

(0.0

6+15.1

2−0.0

6)

0.01

4+0.

024

−0.0

041.

98±0

.04

18.0/2

3

pl+

smc

0.00

5+5.

9−0.0

05(0

.36+

29.6

−0.3

6)

0.01

3+0.

019

−0.0

091.

98+

0.03

−0.0

417

.7/2

3

bknp

l+m

w0.

84+

1.96

−0.8

40.

012+

0.04

5−0.0

122.

44+

0.07

−0.1

0U

(1.7

6:>

0.26

)17

.4/2

20.

182

bknp

l+lm

c1.

27+

0.86

−1.2

7(3

.22+

21.9

−3.2

2)

0.01

9+0.

055

−0.0

192.

44+

0.07

−0.1

0U

(1.5

6)16

.0/2

20.

111

bknp

l+sm

c1.

39+

0.63

−1.3

9(6

.52+

43.0

−6.5

2)

0.01

7+0.

038

−0.0

172.

43+

0.07

−0.1

0U

(1.4

6)15

.6/2

29.

9×10−2

GR

B98

0703

pl+

mw

0.55+

1.02

−0.5

50.

302±

0.05

91.

92±0

.03

30.2/2

7

pl+

lmc

0.54+

1.02

−0.5

4(1

.33+

2.53

−1.3

3)

0.27

5±0.

054

1.92±0

.03

30.0/2

7

pl+

smc

0.53+

1.01

−0.5

3(2

.35+

4.64

−2.3

5)

0.28

7+0.

057

−0.0

561.

92±0

.03

29.8/2

7

bknp

l+m

w1.

35+

1.47

−1.0

60.

31+

0.09

−0.0

62.

38+

0.06

−0.2

41.

40+

1.84

−1.3

822

.9/2

68×

10−3

bknp

l+lm

c1.

34+

1.47

−1.0

6(3

.28+

3.70

−2.6

0)

0.28+

0.08

−0.0

62.

37+

0.05

−0.2

41.

40+

1.83

−1.3

822

.6/2

67×

10−3

bknp

l+sm

c1.

33+

1.46

−1.0

5(5

.80+

6.87

−4.6

4)

0.30+

0.08

−0.0

62.

37+

0.05

−0.2

41.

40+

1.81

−1.3

822

.3/2

67×

10−3


Table4.4

—–

continued

GR

B990123

pl+m

w0+

0.11

−0

0.006+

0.019

−0.002

1.61±0.01

191/121

pl+lm

c0+

0.11

−0

0.004+

0.013

−0.004

1.61±0.01

191/121

pl+sm

c0+

0.11

−0

0.004+

0.018

−0.004

1.61±0.01

191.3/121

bknpl+m

w0.61

+0.51

−0.49

0.01+

0.03

−0.01

2.01+

0−

0.04

0.67+

1.74

−0.49

115.5/120

7×10 −

15

bknpl+lm

c0.59

+0.52

−0.37 (1.43

+1.33

−1.42 )

0.01+

0.02

−0.01

2.00+

0−

0.04

0.55+

1.85

−0.40

115/120

7×10 −

15

bknpl+sm

c0.59

+0.51

−0.37 (2.54

+2.43

−2.54 )

0.014+

0.024

−0.014

1.99+

0−

0.04

0.55+

1.85

−0.40

115/120

6×10 −

15

GR

B990510

pl+m

w0+

0.286

−0

0+

0.003

−0

1.855+

0.010

−0.007

129/78

pl+lm

c0+

0.344

−0

0+

0.003

−0

1.854+

0.009

−0.010

129/78

pl+sm

c0+

0.340

−0

0+

0.003

−0

1.855+

0.007

−0.010

129/78

bknpl+m

w0.12

+0.75

−0.12

0+

0.01

−0

2.03+

0−

0.01

0.018±0.002

83.3/77

7×10 −

9

bknpl+lm

c0.13

+0.02

−0.13

(0.34+

1.88

−0.34 )

0+

0.01

−0

2.03+

0.07

−0.01

0.018±0.002

83.3/77

7×10 −

9

bknpl+sm

c0.13

+0.02

−0.13

(0.74+

3.26

−0.74 )

0+

0.01

−0

2.03+

0.07

−0.01

0.018±0.002

83.3/77

7×10 −

9

4.4 Results 75

Tabl

e4.

4—

–co

ntin

ued

GR

B00

0926

pl+

mw

0fix

ed0.

166+

0.02

1−0.0

251.

80+

0.09

−0.0

527

.5/1

6

pl+

lmc

0fix

ed0.

119±

0.01

51.

77+

0.07

−0.0

411

.7/1

6

pl+

smc

0fix

ed0.

100+

0.01

4−0.0

151.

76+

0.08

−0.0

515

.0/1

6

bknp

l+m

w0

fixed

0.16

7+0.

026

−0.0

252.

30±0

.09

U(4

.36:>

0.59

)27

.4/1

50.

818

bknp

l+lm

c0

fixed

0.12

2+0.

044

−0.0

172.

25+

0.08

−0.2

8U

(2.2

3:>

0.02

)11

.1/1

50.

382

bknp

l+sm

c0

fixed

0.10

2+0.

017

−0.0

162.

25+

0.09

−0.1

1U

(2.9

0:>

0.38

)14

.7/1

50.

588

GR

B01

0222

pl+

mw

0.63+

0.30

−0.2

40.

063±

0.03

31.

87±0

.03

97.2/1

37

pl+

lmc

0.60+

0.29

−0.2

3(1

.44+

0.72

−0.5

66)

0.04

3+0.

022

−0.0

211.

86+

0.02

4−0.0

2595

.5/1

37

pl+

smc

0.58+

0.28

−0.2

2(2

.39+

1.28

−0.9

87)

0.03

5+0.

019

−0.0

181.

85+

0.02

1−0.0

2296

.3/1

37

bknp

l+m

w1.

35+

0.57

−0.4

70.

087±

0.05

12.

07+

0.09

−0.0

8U

(0.0

3)96

.9/1

360.

518

bknp

l+lm

c1.

15+

0.53

−0.3

6(2

.68+

1.33

−0.8

13)

0.07

6+0.

025

−0.0

292.

02+

0.09

−0.0

40.

01+

0.05

−0.0

186

.5/1

362×

10−4

bknp

l+sm

c1.

15+

0.54

−0.4

0(4

.57+

2.37

−1.5

4)

0.06

0±0.

023

2.02+

0.09

−0.0

50.

017+

0.06

3−0.0

1786

.3/1

361×

10−4


Figure 4.1 — The well known extinction curves for the Milky Way (MW), Large andSmall Magellanic Clouds (LMC and SMC respectively, Pei 1992). The transmission ofoptical/UV intrinsic flux with energy is shown for an object at redshift z = 0 and with alarge optical extinction of E(B − V) = 0.2.

ously from a fit to X-ray, optical and nIR data. We find an excess in absorptionabove the Galactic values particularly significant in two sources: GRBs 000926(E(B − V) only) and 010222 (Figure 4.4), whilst no significant intrinsic absorptionis necessary in GRBs 970228 and 990510. The cooling break can be located inthree of the afterglows: GRBs 990123, 990510 and 010222 and to all other SEDsa single power law is an adequate fit. Details are given below for each individualafterglow.

4.4.1 Notes on individual sources

GRB 970228

No significant absorption is measured for GRB 970228. We find only a singlepower law is required and it is possible to pin down the power law slope relativelywell (χ2/dof = 11.6/14).

4.4 Results 77

GRB 970228 GRB 970508

GRB 971214 GRB 980329

GRB 980519 GRB 980703

Figure 4.2 — Data (crosses) and best fitting models (solid lines, see Table 4.4) for eachof the GRBs in the sample. Data (nIR and optical photometry and BeppoSAX LECSand MECS X-ray spectra) and models are shown in count space. The bin size (effectivebandwidth) of the optical data points can be seen in the model fits. The lower panelsshow the deviation from the model for individual data points, in units of their measure-ment error.

GRB 970508

A well-defined temporal decay slope starting 1.9 days after trigger (Zeh et al. 2006).Preceding this time there is an increase in flux followed by an apparent flattening.The time of the X-ray observation log midpoint occurs a little before the 1.9-day


GRB 990123 GRB 990510

GRB 000926 GRB 010222

Figure 4.2 — – continued

break at 1.679 days, so we have extrapolated the optical data post 1.9 days back to1.9 days and then assume that the evolution of the lightcurve is flat back to 1.679days after trigger. To allow for a different behaviour before 1.9 days we include aconstant value offset in the model between the optical and X-ray data, which weboth fix at 1.0 and leave as a free parameter. The improvement in the fits when theoffset is a free parameter is somewhat marginal. An F-test indicates that the free-parameter model is better at a 98% level, with the offset increasing to 3. Given theuncertain extrapolation, however, we use the fits with the offset as a free parameterin our further analysis. The best fitting model for GRB 970508 is a single powerlaw with relatively low intrinsic X-ray absorption at the level of NH ∼ 1021 cm−2.Kann et al. (2006) et al. (2006) found a best fit with MW-like dust, but as wemeasure no significant optical extinction we cannot distinguish between differentextinction laws.

GRB 971214

GRB 971214 is the highest redshift source in the sample at z = 3.418, and we notethat the source was faint, particularly as seen by LECS, crucial for the low X-rayenergies. We measure an optical-to-X-ray spectral index αox of 0.6 and find that a

4.4 Results 79

Figure 4.3 — Intrinsic absorption in optical/nIR (E(B − V)) and X-rays (log NH) mea-sured for the GRB sample. We compare the measurements with three different opticalextinction laws overlaid with solid curves: Galactic (top panel Predehl & Schmitt 1995),LMC (middle panel Koornneef 1982, see also Fitzpatrick 1985) and SMC (lower panelMartin et al. 1989). Appropriate metallicities are adopted for LMC (1/3 Z�) and SMC(1/8 Z�) calculations (diamonds), and stars mark the centroids of the Solar metallicityfits. For GRB 000926 the data were too sparse to fit for NH, so we plot the E(B − V)range at log NH = 17.0 for clarity. Error bars are 90 per cent confidence.

single power law is an acceptable fit to these data. The intrinsic X-ray absorptionappears to be extremely large whilst the optical extinction is moderately large, butwe note that the high redshift of the GRB makes measurement of the X-ray ex-tinction more difficult. The curvature in the optical part of the spectrum has beenpreviously interpreted as a cooling break (Wijers & Galama 1999) and as extinctionby either SMC-like extinction (Stratta et al. 2004) or by a presently unknown, morecomplex extinction law (Halpern et al. 1998; Ramaprakash et al. 1998). WhilstSMC extinction is the best fitting law of the 3 used here, it is not sufficient to ac-curately reproduce the shape of the optical SED, reflected in the large errors on NH

and E(B − V).


Figure 4.4 — Spectral energy distributions of GRBs 000926 (upper panel) and 010222(lower panel) with best fitting models overlaid with a solid line. Overlaid with a dashedline is the unabsorbed source flux, demonstrating that for these two sources extinc-tion significantly affects the observed optical to X-ray emission, in complete contrastto GRBs 970228 and 990510 where extinction in both optical and X-ray regimes arenegligible.

4.4 Results 81

GRB 980329

In the absence of an accurate redshift determination for GRB 980329 we adopt thephotometric redshift of z = 3.6 (Jaunsen et al. 2003), hence all results must be takenwith caution. The most striking feature of this SED is the apparent flux deficit inthe R band, which is present even after correction for Galactic absorption and theprobable high redshift. We tested for the possibility that the R-band flux deficit isdue to the 2175Å feature in the Milky Way extinction curve by fitting the pl+MWmodel whilst leaving the redshift as a free parameter lying between z = 1.2 and z =4.2 (Jaunsen et al. 2003) (we note that at z = 3.6 the 2175Å bump would lie betweenthe I and J bands if extinction is MW-like, which is not observed). We obtain a fitwith χ2/dof = 32.7/26 and z = 1.95+0.29

−0.31. The other best-fitting parameters, as listedin Table 5 for standard fits, do not differ greatly from those found in the z = 3.6 fitand the optical curvature is no better fitted. We also tested for the possibility that theR-band deficit is caused by a break in the power law, by allowing the break energyto reside within the optical regime (and adopting z = 3.6 with SMC extinction). Inthis fit the break energy could not be well constrained, and the power law slopesare not well described by the difference of 0.5 as expected for a cooling break (thesecond slope is steeper), hence we rule out this possibility. However, it is alsopossible that we are seeing an I band excess rather than an R band deficit. Two latetime I band points taken at 1–10 days since burst (Yost et al. 2002) appear not toagree with all of the I band data used here (Reichart et al. 1999), which could bethe result of overestimation of the early I magnitudes, underestimation of the later Imagnitudes or the occurrence of color evolution. Among the models applied to thewhole GRB sample here, a single power law unabsorbed in the X-rays is the bestfit, with a moderate E(B − V) of 0.178 and a slight preference for SMC extinction.

GRB 980519

We caution that all the results are based on a redshift estimate equal to the meanof the sample spectroscopic redshifts of z = 1.58. A few attempts to constrain theredshift have not been very precise, amounting to 0.5< z <3.6 or 1.5< z <3.6(lower limits from Jaunsen et al. (2003) and upper limit from the fact that we detecta U band counterpart). We find that a single power law with a small E(B − V) andX-ray absorption consistent with zero but with a large error is sufficient to modelthis afterglow. We note that the Galactic X-ray extinction towards GRB 980519 isthe highest for this sample which, together with the lack of known redshift hampersa good measurement of NH for this source.

GRB 980703

A single power law provides an acceptable fit to the spectrum, when absorbed bya large amount at both X-ray and optical/UV wavelengths. In this afterglow we


measure the largest E(B − V) value in the sample of 0.29±0.05 or AV∼0.85 at 1.33days after burst assuming SMC extinction (which is marginally preferred). This isconsistent with the value found by Bloom et al. (1998b) of 0.9 ± 0.2, 5.3 days afterthe burst.

A great deal of work has been done on the host galaxy properties of this burstbecause the host is bright, with optical extinction measurements by 5 groups. Thereis a hint that the optical extinction may be decreasing with time (see e.g. Hollandet al. 2001) since measurements of AV at different times are inconsistent: AV∼2.2 at0.9 days, Castro-Tirado et al. (1999); AV∼1.5 at 1.2 days, Vreeswijk et al. (1999);AV∼0.3 at 4.4 days, Djorgovski et al. (1998); ∼0.9 at at 5.3 days, Bloom et al.(1998b), and there appears to be a discrepancy between the measured optical spec-tral and temporal slopes when assuming AV is constant. However, we note thatthe optical spectral slope was taken to be βOA = -2.71±0.12 from Vreeswijk et al.(1999), and in this study we obtain a lower value of βOA = -0.92±0.03 which wouldbe completely consistent with βOX = (1+2α)/3 = 0.9 using α = 0.85 from Zeh et al.(2006), noting their α value has a very large associated error - Table 4.1). We usethe Vreeswijk et al. (1999) optical data here and scale it from 1.2 days to 1.33 daysafter trigger. Combining the optical and X-ray data when fitting provides us witha different estimate for the extinction than was obtained by Vreeswijk et al. for theoptical data alone. We note that we have made only a minimal extrapolation of theoriginal optical data used in this analysis, from ∼1.2 days to 1.3 days after trigger.

GRB 990123

GRB 990123 does not have significant X-ray absorption above the Galactic value,and the optical extinction is consistent with zero. We set an upper limit to the latterwhich is ten times lower than the value found by Savaglio et al. (2003) from fits toZn II and H I in the optical spectrum. A single power law fit to both optical and X-rays results in a spectral slope of β=0.61±0.01 at 1.24 days since burst, comparableto the βOX=0.67±0.02 at about the same time since the burst found by Galamaet al. (1999). The latter authors note that the cooling break must lie at or aboveX-ray frequencies at that time. However, we find an improved fit with a brokenpower law, constraining the cooling break to 0.15 < νc < 2.4 keV, within the X-rayspectrum, also found by Stratta et al. (2004), Corsi et al. (2005) and Maiorano et al.(2005). We note, however, that the X-ray spectrum comprises flux accumulatedbetween 0.2 and 2.6 days since burst, including the proposed jet break time of 2days (Table 4.2). We tested for the possibility that the offset between optical andX-ray data was incorrect, but this made very little difference to our overall goodnessof fit. It is possible that the cooling break has entered the X-ray band during theseobservations, since this break is expected to decrease in frequency with time. Ifthis is the cooling break, the spectral slopes above and below the break differ bythe expected factor of 0.5 when left free. We therefore took only the data fromthe first X-ray observation, before the jet break (Zeh et al. 2006), together with

4.4 Results 83

scaled optical/nIR data, which has a log midpoint of ∼0.7 days since trigger. Werepeated all our fits. We find acceptable fits again only with the broken powerlaw model, with the break energy lying somewhere between the optical and ∼2keV (preferring central values just below X-ray frequencies, in line with our X-rayonly analysis in which a single power law is a good fit to all spectra). We notethat an excess of flux at high energies (as seen by the BeppoSAX PDS instrumentnot used here) is reported by Corsi et al. (2005). They attribute this to an InverseCompton component (though we note that this remains inconsistent with the radiodata Kulkarni et al. 1999b). Since the X-ray spectrum is adequately fit with a singlepower law we assume that the tail of any such component is not significant below10 keV.

GRB 990510

Fits with Galactic, LMC- and SMC-like extinctions show that E(B−V) is very lowin this source and we can only determine upper limits. The low amount of extinc-tion makes it impossible to pin down the extinction curve shape, hence a fit usingGalactic-like extinction is sufficient here and LMC- and SMC-like extinctions givesimilar results. There is considerable improvement in the χ2 when allowing fora break in the power law, noted by previous authors, which we find is located at0.016–0.020 keV at ∼1.06 days since burst (of the order of the value of ∼0.029 keVat ∼1 day since burst given by Pian et al. 2001). The slope change is as expectedfor a cooling break in the slow cooling regime when leaving both power law slopesfree. Our measured spectral slope in the optical regime agrees well with that mea-sured in the Very Large Telescope (VLT) spectra of β = 0.6±0.1 by Vreeswijk et al.(2001).

We note that Kuulkers et al. (2000) analysed these X-ray data in several timebins and found no spectral evolution, hence the cooling break remains outside theX-ray frequencies during the observations. We constrain the optical extinction tobe E(B − V) ≤ 0.01, and the X-ray equivalent hydrogen column to be ≤ 0.15×1022

cm−2 for SMC extinction and ≤ 0.87×1022 cm−2 for MW extinction. From the X-ray spectrum alone a higher column of NH = 2.1 ± 0.6×1021 cm−2 was measured byKuulkers et al. (2000). Optical spectra have provided a lower limit on the amountof neutral hydrogen towards GRB 990510 of log N(H I) ≥ 19.7 cm−2 (Vreeswijket al. 2001). These authors obtain an approximate estimate for the metallicity fromthe optical spectra using Fe/H, and find 12+ log [Fe/H] = -1.5 ± 0.5 or 0.01–0.1times the Solar value. This range approximately covers the metallicity of both theLMC (0.33, Pei 1992) and the SMC (0.125, Pei 1992) so we also fitted the datawith these two metallicities. As there is no substantial absorption observed, the fitsdo not change significantly.


GRB 000926

The optical and IR SED of this burst is very well sampled, but the X-ray afterglowwas very faint at the time of observation. For this reason these X-ray data have beengrouped to have a maximum of only 10 counts per bin, and strictly speaking thismeans Gaussian statistics should be treated with caution. However, we do use theχ2 statistic as a goodness of fit for comparison with the rest of the sample. Given thepoor quality of the X-ray data, and the lack of sufficient counts in low X-ray energybins, we have fixed the X-ray column density at zero and fit only for the opticalextinction. We find a large amount of intrinsic extinction is necessary to describethe flux deficit with respect to a single power law, consistent with the value foundby most other studies (Stratta et al. 2004; Fynbo et al. 2001a; Harrison et al. 2001;Price et al. 2001). A large extinction was also derived from an optical spectrumby Savaglio et al. (2003), and the AV found in that study is approximately twicethe value found here. However, Savaglio et al. use the spectral line measurementsto first fit for the depletion pattern and then infer an extinction. Harrison et al.(2001) interpreted the optical flux deficit as indicating that a significant fractionof the X-ray flux was in fact Inverse Compton emission - later also suggested forGRB 990123 (see above). This is the only afterglow for which an LMC extinctionlaw is (marginally) preferred. Fynbo et al. (2001a) report a tentative H I columndensity measurement of N(H I) ∼2×1021 cm−2 which leads to a relatively highmetallicity with [Zn/H] = -0.13. This metallicity is between the LMC and the MWvalues, which may explain the preference here for LMC-like extinction if no 2175Åbump is present.

GRB 010222

A good dataset for GRB 010222 allows the spectral properties to be well con-strained. We find that whilst a single power law is a reasonable fit to these data,a broken power law significantly improves the fit. The break energy lies around0.01 keV, above the frequency of the last optical band in our SED. Optical extinc-tion is clearly non-zero, with E(B − V) = 0.06±0.02, consistent with that found byLee et al. (2001) for an SMC extinction law but about three times lower than thatinferred from the spectral lines by Savaglio et al. (2003). X-ray absorption is alsorequired with an effective hydrogen column of NH = 1.15+0.54

−0.39 × 1022 cm−2.Panaitescu & Kumar (2002) in their fits to the multiwavelength afterglow of

this source find significant reddening of the optical spectrum of AV = 0.21 withan SMC extinction curve, which is consistent with our value. But they find a largefitting error and attribute this to 8 outlying points suggesting that either some re-ported observations have underestimated uncertainties or there are short timescalefluctuations in the afterglow emission (Cowsik et al. 2001). Their jet model re-quires the cooling break to pass through the X-ray band at about 1 day, which theyfind incompatible with their observations. Our analysis places the cooling break at

4.4 Results 85

optical/UV wavelengths at 1.51 days since burst.

4.4.2 Comparison with previous studies

We can compare our results directly with those of previous studies of samples over-lapping with this BeppoSAX subsample. In general, we are finding similar centralvalues for extinction as all previous studies, and are improving upon the uncer-tainties, fitting all afterglows in the same consistent manner allowing for directcomparison.

Galama & Wijers (2001) performed the first systematic study of line-of-sightNH and E(B − V) with a sample of 8 afterglows, consisting of all but the two mostrecent bursts in our sample. From fits to the X-ray spectra they found intrinsicNH amounting to 1022-1023 cm−2, ruling out the possibility that some hosts haveno X-ray column at all. They noted that these values lie in the range of Galacticgiant molecular clouds (estimating cloud sizes of 10–30 pc) - a conclusion alsorecently arrived at when including Swift bursts (Campana et al. 2006b) and whenmeasuring H I from damped Lyman alpha absorption in GRB optical afterglowspectra (Jakobsson et al. 2006a). They used a simple extinction law AV ∝ ν witha smoothly broken power law. Comparing their dust to gas ratios with that of theMilky Way they obtain an optical extinction 10–100 times smaller than expected.Their finding of generally low AV was attributed to previously predicted dust de-struction by the GRB. Of the bursts common to both their and our samples andhaving a known redshift we obtain consistent extinction values within the 2σ un-certainties in all cases except for GRB 971214 for which the two optical extinctionsare only consistent at the 3σ level, our central value lying four times lower than theGalama & Wijers measurement. We obtain smaller uncertainties in our extinctionmeasurements in all cases except for the X-ray column of GRB 980703.

Stratta et al. (2004) have also measured NH and AVr in the optical and X-raydata seperately, and later plotted the combined data in flux space (after assumingthe X-ray model to be correct) in order to judge the position of the cooling break.To derive optical spectral slopes, Stratta et al. first adopt the p-value derived fromthe X-ray fits (the input electron energy index, see Paper II for our fits for thisparameter), use this to fix the optical spectral index and fit for AV . Our resultsfor E(B − V) with SMC-like extinction are consistent in all cases to within the 90% confidence limits, adopting the cooling break positions found from this work ifknown (the values for GRB 970508 are only consistent if νo > νc). Even for the twoGRBs for which we have assumed different redshifts we find consistency in extinc-tion estimates (Stratta et al. adopt z = 1 for both 980329 and 980519). In 7 of the 10cases we derive a better constrained value or upper limit to the extinction. The X-ray absorbing columns we measure are also generally consistent with those foundby Stratta et al. in a fit to the X-ray data only (these are of course the same dataused in our study, except for differences in number of observations combined forsome sources, and we include GRB 000926 as well). Our method obtains improved


estimates for NH for 7 of the 9 sources common to both studies. For GRB 990510Stratta et al. find a value higher than our derived NH using Solar metallicity, and forGRB 980329 our result for NH is less accurate at the 90 % confidence level, thoughwe note the different redshifts assumed hence direct comparison is not possible.

Kann et al. (2006) fitted only the optical SEDs for a sample of pre-Swift burstsincluding 8 analysed here (not including GRBs 970228 and 980329 and using z =1.5 for GRB 980519). We find similar central values of AV and improve upon theirmean error by a reduction of 5–10 %. Values disagree at the 90 % confidence levelonly for GRB 970508, where we find an upper limit to the extinction which is 2.4times lower than the lower limit of Kann et al. but would be consistent with theirestimate at the 3σ level.

4.5 Discussion

For half the afterglows the best-fitting model to the SED includes SMC-like ex-tinction (as opposed to LMC or MW) and in one case LMC-like extinction. In nocases is there a preference for MW-like extinction. We are sensitive to the 2175Åbump (MW) in the redshift range z = 0.46–9.9, covering all our selected GRBs, butclearly we do not detect any such feature. We find a wide spread in central valuesfor the gas-to-dust ratios, and for 4 GRBs the gas-to-dust ratios are formally incon-sistent with MW, LMC and SMC values at the 90 % confidence limit assuming theSMC metallicity (Figure 4.5). In these 4 cases the ratio is several orders of mag-nitude higher than the SMC value of 4.4±0.7×1022 cm−2 mag−1 (Koornneef 1982;Bouchet et al. 1985) and must mean that either gas-to-dust ratios in galaxies canspan a far larger range than thought from the study of local galaxies, or the ratios aredisproportionate in GRB hosts because the dust is destroyed by some mechanisms(likely the GRB jet), or that the lines of sight we probe through GRBs tend to bevery gas-rich or dust-poor compared with random lines of sight through galaxies.Finally, a dust grain size distribution which is markedly different than consideredhere may also affect these ratios.

In fact, a recent study has shown that for the LMC the former is true. A recentobservation of four core-collapse supernova remnants (SNRs) in the LMC withSpitzer has shown IR emission associated with the supernova blastwave (Williamset al. 2006). This is interpreted as dust with an LMC-like grain size distributionwhich has been collisionally heated by the X-ray emitting plasma. The observationsrequire that some fraction of the small dust grains has been destroyed by sputteringby high energy ions in fast shocks. Dust destruction is known to occur in SNRshock fronts (Jones 2004), and we will return to this issue later in the section. Thederived gas-to-dust ratios are several times higher than the LMC ratio, as we see inthe line-of-sight measurements of GRBs and has been observed in other types ofsupernova (e.g. Borkowski et al. 2006), the cause of which is not known.

We measure a large amount of intrinsic absorption in some of the sample (Fig-

4.5 Discussion 87

ure 4.4), and can state that absorption is insignificant in others, as seen for examplein the contrast between GRBs 010222 and 990510. We have tested for the possibil-ity that the Galactic column density in the GRB direction is affecting our intrinsiccolumn derivations by plotting NH,int vs. NH,Gal, and found no correlation. We as-sumed therefore that all the measured absorption lies within the host galaxy. Wenote that a few of these afterglows have spectra: for GRB 990510 no interveningsystems were clearly identified, an intervening system is measured at 168 km s−1

from the host redshift for 000926 (assumed to lie within the host galaxy Castro et al.2003), 2 intervening systems are found for GRB 010222 (Jha et al. 2001; Mirabalet al. 2002) and at the host redshift there are two components separated by 119 kms−1 (Mirabal et al. 2002). None of these intervening systems are close enough to usas the observer to significantly affect measurement of the intrinsic host extinction.

4.5.1 Approaches to measuring absorption in the host galaxies

In this study we provided a thorough, uniform study of both optical and X-rayextinction along the lines of sight towards a sample of 10 GRB afterglows. Thewell known spectral shape and relative brightness of the afterglow emission makeGRB afterglows a powerful line-of-sight probe of high redshift extinction. This isone of several approaches to measuring absorption in GRB host galaxies. One canglobally divide the studies of extinction in the field of GRBs in two categories: line-of-sight extinction studies and studies of extinction of the integrated host galaxy orparts thereof.

Line-of-sight studies

Line-of-sight studies generally involve fitting the afterglow spectral energy distri-butions in optical and/or X-rays with template extinction models (i.e. MW, LMC,SMC or more parametrized models) as we have done here. The standard conver-sions between X-ray extinction and optical extinction for the Milky Way and thetwo Magellanic Clouds are generally in disagreement with the column densitiesmeasured through this method, but these skewed gas-to-dust ratios are also beingfound in other astrophysical situations as discussed above for SNRs in the LMC,and the destruction of dust can go some way to alleviating the mismatch.

A further way to probe line-of-sight extinction properties is through opticalspectroscopy of the afterglow. In this case the careful measurement of columndensities of heavy elements can be used to study the dust depletion pattern along theline of sight (e.g. Savaglio & Fall 2004). The measured metal column densities incombination with the best-fitting depletion pattern and the empirically determinedconversion between AV and the dust column, can provide a prediction of the dustextinction along the line of sight to a GRB (for a detailed explanation see e.g.Savaglio & Fall 2004). Savaglio & Fall (2004) show that the extinction derivedfrom the dust depletion method is significantly higher than the value derived from


direct fitting to the continuum of the afterglow spectrum, e.g. for GRB 020813they find an overestimation by at least five times. We show that this conclusionholds when fitting the afterglow continuum emission over a much larger wavelengthrange and can quantify that overestimation factor for GRBs 990123, 000926 and010222 as approximately � 11, � 2 and � 3 times overestimated respectively. Thereasons for this apparent discrepancy may be two-fold: firstly the fitted extinctionprofiles to the afterglow SED are likely poor approximations to the true extinctionprofile, and secondly the host dust depletion chemistry may well differ from theMilky Way chemistry. In addition, the GRB or afterglow may preferentially destroysmall dust grains, skewing the extinction profile towards larger grains, resulting ina “grey” extinction curve. This would alter the derived extinction from SED fitting,and possibly bring estimates from dust depletion methods and SED fitting closertogether. Whist a grey extinction curve was the best fitting extinction curve toGRB 020405 (Stratta et al. 2005), we note that grey extinction curves have beenfitted to samples of afterglow SEDs with no conclusive improvement in fit (e.g.Stratta et al. 2004).

Integrated host galaxy studies

One can also study the extinction properties of host galaxy as a whole, and thereare again several methods to do this. Whilst it has been shown that GRBs occurin starforming regions in the host (Bloom et al. 2002; Fruchter et al. 2006), manyhost galaxies are small and mixing timescales may be short, enabling global prop-erties to be measured. The host and afterglow have similar, moderate reddening inGRB 000418 (Gorosabel et al. 2003b) which is taken as evidence that the ISM iswell mixed. But more extreme values of reddening are also seen, such as the ex-tremely red afterglow and host of GRB 030115, in which the host is an ExtremelyRed Object (ERO Levan et al. 2006).

One of the most common methods of integrated host galaxy studies is fitting ofthe broadband optical and near-infrared SEDs of the hosts themselves (e.g. Chris-tensen et al. 2004). Galaxy templates can be fit to the data, using photometricredshift programmes such as HyperZ (Bolzonella et al. 2000) providing values forthe photometric redshift, the age of the dominant stellar population and the extinc-tion, by fitting a series of galaxy templates. The extinction measured this way isthe extinction by the ISM of the galaxy on the stellar light, E(B − V)s, in whichthe geometry of the dust in the galaxy can play an important role (e.g. dustlanes asopposed to an homogeneous dust distribution). A study of a large sample of hostshas been performed through HyperZ template fitting by Christensen et al. (2004),who find that GRB hosts generally exhibit little extinction, and have young stellarpopulations. The dependence on metallicity and assumed initial mass function issmall (Gorosabel et al. 2003b,a; Christensen et al. 2004). One of the difficultiesfaced here is that the galaxies are often very faint.

4.5 Discussion 89

Emission-line spectroscopy

In low redshift (z � 1) cases, an optical spectrum of the host galaxy can be taken,typically showing several nebular and Balmer emission lines. The Balmer lines canbe used to derive values for the reddening by calculating their deviation from case Brecombination values expected in typical starforming region conditions (Osterbrock1989). The derived reddening E(B − V)g is the reddening of the ionized gas in thesource, i.e. the dominant starforming region(s) producing the Balmer emissionlines. In general the reddening found from the Balmer decrement is low (see e.g.Prochaska et al. 2004), to very low (Wiersema et al. 2007).

The two host galaxy extinction estimates E(B − V)g and E(B − V)s may becorrelated for most galaxies (see Calzetti 2001 for a review). The increasing datavolume on nearby GRB host galaxies will allow a test of these correlations, provid-ing further insight into effective GRB host galaxy extinction curves.

On rare occasions it is possible to obtain high-resolution spectra of an afterglowthat also shows host galaxy emission lines, allowing one to obtain a simultaneousview of the extinction along the line of sight and of the Balmer decrement. In thecase of GRB 060218, both absorption lines and emission lines are detected at highresolution using the UVES spectrograph on the VLT (Wiersema et al. 2007). Thespectrum shows asymmetric emission lines that are well fitted with two Gaussiansseparated by 22 km s−1. The same two velocity components can be seen in absorp-tion in Ca II and Na I, and have different chemical properties. These two systemscan be interpreted as two seperate starforming regions, through which the light ofthe afterglow shines. A broadband measure of the extinction either from the after-glow or from template fitting of the host would not have been able to seperate outthe contributions of the two individual systems.

The longer wavelengths

Yet another way to detect the presence of dust, is the detection of GRB host galax-ies in the far-infrared or sub-mm (see e.g. Barnard et al. 2003; Tanvir et al. 2004),where the UV radiation from massive stars is reprocessed by dust and reradiatedin the far infrared. Detection of hosts in the far infrared as well as optical canseverely constrain their SED (Le Floc’h et al. 2006), providing estimates for theunobscured star formation rate of GRB hosts. In a few cases values up to hundredsof solar masses per year have been reported, while optical indicators give muchlower values, indicating a lot of dust-obscured star formation. A different probeof unobscured star formation is the radio continuum flux, which is thought to beformed by synchrotron emission from accelerated electrons in supernova remnantsand by free-free emission from H II regions (Condon 1992). Berger et al. (2003a)performed a survey of host galaxies at radio and sub-mm wavelengths, finding thata significant fraction of GRB host galaxies have a much higher radio-derived starformation rate than optical methods indicate, pointing again to significant dust ex-


Figure 4.5 — Gas-to-dust ratios, NH/E(B − V), derived from best fits to the SEDs as-suming SMC metallicity, plotted for each sample GRB in date order from left to right(GRB 970228 first), excluding GRB 000926 where we did not fit for X-ray column den-sity. We include the ratios for 3 GRBs taken from the literature for comparison (squaredata points): GRBs 000301c (Jensen et al. 2001), 000926 (Fynbo et al. 2001a) and thelower limit on 020124 (Hjorth et al. 2003a). The solid, dotted and dot-dashed linesshow the measured gas-to-dust ratios and their errors for the Milky Way (Diplas & Sav-age 1994), LMC and SMC (Koornneef 1982; Bouchet et al. 1985). Error bars are at the90 % confidence level.

tincted star formation but based on only a few detections.

Our method compared

All methods considered, SED fitting of the afterglow as described and carried outin this paper is the most broadband view of line-of-sight properties. Given the verywell known underlying continuum shape, the extinction curve can in principle be

4.5 Discussion 91

extremely well modelled using this method. As discussed in Section 1, the effectsof dust extinction in the optical are better measured at higher redshifts whereas theX-ray absorption is best measured at lower redshifts, so this method is most reliablefor some middle range of redshifts centred around z = 2 which approximately cor-responds to the current Swift median redshift (e.g. Jakobsson et al. 2006b). As withany line-of-sight method, the measured columns may not be representative of thehost galaxy as a whole, so comparison with the integrated host galaxy methods isimportant. The general drawbacks of this method lie in the inability to disentanglethe locations of absorbers, both intervening (between us and the host) and within thehost itself. Intervening absorbers may be identified using afterglow spectroscopy(Ellison et al. 2006). Distribution of absorbers within the host may be tackled fromthe point of view of searches for variability in the absorption which may indicatedestruction of dust or ionisation by the GRB and/or afterglow, and comparisonof the measured columns with known absorbers such as Molecular clouds. Cam-pana et al. (2006b) find that the X-ray absorptions of 17 bursts (including most ofthis sample) are consistent with the GRBs lying along random sightlines towardsGalactic-like Molecular clouds. Our sample is too small to make such a statisticalcomparison, but we note that the Campana et al. sample includes the results of theStratta et al. (2004) study of the GRBs presented here for which we obtain largelyconsistent extinction values.

Dust destruction is of course impossible to measure in a single SED, but maybe seen with multiple epochs of data. Destruction of grains by for example theafterglow UV/X-ray emission would in principle be observable, because depletionindicative line ratios would vary on observable timescales as metals are releasedfrom grains into the ISM. Variable line emission has been searched for in afterglowswith multiple optical spectra and has been seen in only two cases, GRB 020813, inwhich the Fe II λ2396 transition equivalent widths decreases by at least a factorof five over 16 hours (Dessauges-Zavadsky et al. 2006) and GRB 060418 in whichseveral transitions of Fe II and Ni II are seen to vary in observations covering 11to 71 minutes post-burst (Vreeswijk et al. 2007). This line emission is thought toarise at 50-100 pc from the GRB site, possibly within range of the GRB ionisingflux. A variable AV has only been suggested for GRB 980703 (described in Section4.1.6). There is evidence for time variable X-ray absorption in a small number ofGRBs ranging from tentative to moderately strong, in particular for GRBs 011121(Piro et al. 2005), 050730 (Starling et al. 2005a) and 050904 (Boer et al. 2006;Campana et al. 2006a; Gendre et al. 2007) and 060729 (Grupe et al. 2006). Thisimplies ionisation of the line-of-sight gas, probably by the high energy GRB jet(though we note that column density changes and early-time spectral evolution canhave similar effects on the spectrum).

Dust destruction can, however, only affect the immediate environment arounda burst, (perhaps out to 10 pc Waxman & Draine 2000), and not all the dust in thatregion need be destroyed since this depends on the efficiency of the mechanismand on the dust grain size distribution. Kann et al. (2006) noted a possible tenta-


tive correlation between AV and star formation rate as measured via host galaxyemission in the sub-mm (for a sample of 7 GRBs). If this turns out to hold whentested against larger samples it would demonstrate that a substantial amount of thedust we see comes from star forming regions located throughout the whole hostgalaxy rather than very close to the burst where it could be destroyed. However,this correlation is unlikely to be real given the very low resolution of the sub-mmobservations which make it difficult to differentiate between host galaxy emissionand field galaxy emission. This differentiation can be made with Spitzer and lowerstar formation rates in comparison with the sub-mm derived values have been foundfor some hosts including those of GRBs 980703 and 010222 (Le Floc’h et al. 2006).

4.6 Conclusions

Here we have demonstrated the advantages of simultaneous fitting of broadbanddata using a subsample of BeppoSAX GRB afterglows. In no cases is a MW-likeextinction preferred, when testing MW, LMC and SMC extinction laws. The 2175Åbump would in principle be detectable in all these afterglows, but is not present inthese data. An SMC-like gas-to-dust ratio (or lower value) can be ruled out for 4 ofthe hosts analysed here (assuming SMC metallicity and extinction law) whilst theremainder of the sample have too large an error to discriminate.

We discuss the various methods employed to derive host galaxy extinctions,and compare our results with previous works. We find that this method providessimilar central values of E(B− V) and NH to previous works in which extinction orabsorption is determined through afterglow continuum fitting, and in the majorityof cases we obtain tighter constraints. We confirm that, with respect to continuumfitting methods such as this, optical extinction is overestimated with the depletionpattern method, and quantify this for a small number of cases.

Swift, robotic telescopes and Rapid Response Mode on large telescopes suchas the William Herschel Telescope on La Palma and the Very Large Telescopesin Chile now allow earlier and higher quality data to be obtained, which will helpimmensely in discriminating between the different extinction laws at work in thehost galaxies.

Acknowledgements We are grateful to Mike Nowak for his assistance with ISIS. Wethank Wim Hermsen, Jean in ’t Zand, Erik Kuulkers, Tim Oosterbroek and Nanda Reafor advice on the BeppoSAX data reduction, and the referee for useful comments. Thisresearch has made use of SAXDAS linearized and cleaned event files produced at theBeppoSAX Science Data Center. The authors acknowledge benefits from collaborationwithin the Research Training Network ‘Gamma-Ray Bursts: An Enigma and a Tool’,funded by the EU under contract HPRN-CT-2002-00294. RLCS and ER acknowledgefunding from PPARC.

5Probing Cosmic Chemical Evolution with

Gamma-Ray Bursts: GRB 060206 at z = 4.048

J. P. U. Fynbo, R. L. C. Starling, C. Ledoux, K. Wiersema, C. C. Thone,J. Sollerman, P. Jakobsson, J. Hjorth, D. Watson, P. M. Vreeswijk, P. Møller,

E. Rol, J. Gorosabel, J. Naranen, R. A. M. J. Wijers, G. Bjornsson,J. M. Castro Ceron, P. Curran, D. H. Hartmann, S. T. Holland, B. L. Jensen,

A. J. Levan, M. Limousin, C. Kouveliotou, G. Nelemans, K. Pedersen,R. S. Priddey, N. R. Tanvir

Astronomy and Astrophysics, 451, L47 (2006)

Abstract We present early optical spectroscopy of the afterglow of the gamma-ray burst GRB 060206 with the aim of determining the metallicity of the GRB ab-sorber and the physical conditions in the circumburst medium. We also discuss howGRBs may be important complementary probes of cosmic chemical evolution. Ab-sorption line study of the GRB afterglow spectrum. We determine the redshift ofthe GRB to be z = 4.04795±0.00020. Based on the measurement of the neutral hy-drogen column density from the damped Lyman-α line and the metal content fromweak, unsaturated S II lines we derive a metallicity of [S/H]= −0.84 ± 0.10. This isone of the highest metallicities measured from absorption lines at z ∼ 4. From thevery high column densities for the forbidden Si II*, O I*, and O I** lines we infervery high densities and low temperatures in the system. There is evidence for thepresence of H2 molecules with log N(H2)∼17.0, translating into a molecular frac-tion of log f ≈ −3.5 with f = 2N(H2)/(2N(H2)+ N(H I)). Even if GRBs are onlyformed by single massive stars with metallicities below ∼ 0.3 Z�, they could still befairly unbiased tracers of the bulk of the star formation at z > 2. Hence, metallicitiesas derived for GRB 060206 here for a complete sample of GRB afterglows will di-rectly show the distribution of metallicities for representative star-forming galaxiesat these redshifts.

5.1 Introduction

Long gamma-ray bursts (GRBs) are now established to be caused by the deaths ofmassive stars (e.g. Hjorth et al. 2003b; Stanek et al. 2003) and due to their bright-ness they can be observed throughout most of the observable Universe (e.g. Kawaiet al. 2006). Given these facts it has long been realized that GRBs could be ideal

93

94 Probing Cosmic Chemical Evolution with Gamma-Ray Bursts

probes of star-formation activity throughout the history of the Universe (e.g. Wi-jers et al. 1998). However, previous GRB missions detected too few rapidly well-localised GRBs to build statistically interesting samples and hence really capitalizeon this potential. The Swift satellite (Gehrels et al. 2004) has increased the detectionrate of rapidly well-localised GRBs by roughly an order of magnitude compared toprevious missions. Moreover, its significantly deeper detection limit (e.g. Band2006) means that Swift detects more distant bursts than previous missions (Jakobs-son et al. 2006b).

In this Letter we present optical spectroscopy of GRB 060206, focussing onthe measurement of the metallicity of the GRB absorption system, the density inthe circumburst environment and the presence of molecules. We then discuss howGRBs are important for understanding the build-up and distribution of metals ingalaxies (e.g. Pei & Fall 1995), which is currently not fully understood (e.g. Ferraraet al. 2005).

5.2 Observations

GRB 060206 was discovered by the Burst Alert Telescope (BAT) aboard the Swiftsatellite on February 6 04:46:53 UT. The burst exhibited a slow rise and a fasterdecline, with a T90 of 7 ± 2 s (Palmer et al. 2006). The X-ray Telescope (XRT)slewed promptly to the location and began taking data at Δt = 58 s, where Δt is thetime from the onset of the burst. Due to entry into the South Atlantic Anomaly,XRT could only observe the BAT error circle briefly, and therefore no fading X-ray source was immediately localised (Morris et al. 2006a). Observations with theUV/Optical Telescope (UVOT) began at Δt = 57 s but failed to reveal an opticalafterglow (OA) candidate in the initial data products (Morris et al. 2006a).

We observed the GRB 060206 BAT error circle in the R-band with the An-dalucıa Faint Object Spectrograph and Camera (ALFOSC) on the Nordic Opti-cal Telescope (NOT) starting at Δt ≈ 15 min. A point-like object (R ∼ 17.3) notpresent in the Digitized Sky Survey was detected. The detection was confirmed byre-analysis of the XRT and UVOT data (Boyd et al. 2006; Morris et al. 2006b).

Starting at Δt ≈ 48 min we obtained a 1800 s spectrum with a low resolution(LR) grism and a 1.3 arcsec wide slit covering the spectral range from about 3500Å to 9000 Å at a resolution of 14 Å. The airmass during the spectroscopic observa-tion was very low, starting at 1.01. We also obtained medium resolution (MR) spec-troscopic data with the Intermediate-dispersion Spectroscopic and Imaging System(ISIS) on the William Herschel Telescope (WHT) at Δt = 1.61 hours. We tookspectra of exposure times 2×900 s with both the the 600B grating (covering 3800–5300 Å) and and the 1200R grating (covering 6100–7200 Å) both with a 1.0 arcsecwide slit. The mean airmass was 1.036 and the observations were carried out at theparallactic angle. The resolutions of blue- and red-arm spectra are 1.68 and 0.82 Å,respectively.

5.3 Results 95

Figure 5.1 — Upper panel: Section of the LR afterglow spectrum showing the DLA lineat the GRB redshift, zabs = 4.048. Overlaid is the best fitting DLA profile, correspondingto log N(H I) = 20.85 ± 0.10. Lower panel: Fits to the O I, O I*, O I**, Si II, Si II*, andS II lines in the MR spectrum. The zero-point for the velocity scale is zabs = 4.048.

5.3 Results

The spectra show absorption both at the GRB redshift of z = 4.04795 ± 0.00020and from two intervening systems at redshifts z = 1.48 and z = 2.26. In the


Figure 5.1 — – continued The likely H2 lines at the GRB redshift in the MR spectrum.

analysis presented here we focus on the metallicity of the GRB absorption system.From a fit of the damped Lyα line at z = 4.048, we measure a neutral hydrogencolumn density of log N(H I) = 20.85 ± 0.10 (Fig. 5.1), well above the definitionfor Damped Lyα Absorbers (DLAs, log N(H I) ≥ 20.3). This value is consistentwith fits to the higher order Lyman lines in the blue MR spectrum. To derive themetal column density we use the unsaturated S II λ1250, 1253, 1259 lines shownin Fig. 5.1. The profile of the GRB absorption systems consists of at least fourcomponents spread over ∼500 km s−1 in velocity space. On the right hand side ofthe middle panel of Fig. 5.1, a four-component fit of S II and Si II* is shown. Onthe left hand side, we have only fitted the redmost component at 175 km s−1 wherewe can derive a column density for Si II, O I* and O I**. We fit the lines using theFitLyman package in MIDAS; the measured column densities for all componentsare given in Table 5.1. Note that for components 2 and 3 we fixed the turbulentbroadening parameter value to 15 km s−1. Summing over the four components ofthe system, we get log N(S II) = 15.21± 0.03 which leads to [S/H] = −0.84 ± 0.10.

We also find log N(Si II*) = 14.42±0.02 and therefore [Si II*/Si II]= −1.15, as-suming [S/Si] = 0. In the fourth component, we have [Si II*/Si II] = −0.34, whichis high compared to the ratios found in other GRB host galaxies (twice as highas the value found for GRB 050505 by Berger et al. (2006b) and 20 times higherthan the value found for GRB 030323 by Vreeswijk et al. (2004)). Fine-structurelevels can be populated through collisions, photo-excitation by IR photons, and/orfluorescence (Bahcall & Wolf 1968). Assuming that the former mechanism is dom-inant (but see Berger et al. 2006b; Prochaska et al. 2006b), we can estimate the H Ivolume density, using the calculations by Silva & Viegas (2002). Assuming anelectron fraction ne ∼ 10−4nHI (see Vreeswijk et al. 2004; Berger et al. 2006b) anda temperature of 1000 K, we find nHI ∼ 105 cm−3 for the fourth component, andnHI ∼ 104 cm−3 for the mean ratio of all components (see Fig. 8 of Silva & Viegas2002). For the fourth component, we can actually constrain the kinetic temperature


Table 5.1 — Ionic column densities in individual components of the GRB system atzabs = 4.048.

Ion Transition log N ± σlog N b ± σb

lines used (km s−1)zabs = 4.0441S II 1250, 1253, 1259 14.36±0.06 14±7Si II* 1309 < 13.10a 14±7zabs = 4.0480S II 1250, 1253, 1259 14.79±0.05 ∼ 15Si II* 1309 13.54±0.08 ∼ 15zabs = 4.0490S II 1250, 1253, 1259 14.60±0.06 ∼ 15Si II* 1309 13.77±0.06 ∼ 15zabs = 4.0509S II 1250, 1253, 1259 14.59±0.06 14±2Si II* 1309 14.22±0.05 14±2O I* 1304 15.03±0.16 14±2O I** 1306 14.25±0.05 14±2Si II 1304 14.56±0.10 14±2a 3σ upper limit.

and volume density, as the observed ratio [O I*/O I**] = 0.78 ± 0.17 can only bereached below 320 K (within the 1σ errors), while [Si II*/Si II] = −0.34 ± 0.11 re-quires a temperature above 240 K. This temperature range corresponds to a volumedensity of nHI = 1 − 3 × 107 cm−3.

Finally, our data show the first evidence for H2 molecules in a GRB absorber.Two consistent features are seen at the location of the W1-0 R(0), W1-0 R(1) andW1-0 Q(1) lines at zabs = 4.04793 (bottom panel in Fig. 5.1), with column den-sities of log N(H2) ∼ 17.0 for the J = 1 rotational level, and log N(H2) < 16.7for J = 0. The corresponding H2 molecular fraction is log f ∼ −3.5 withf = 2N(H2)/(2N(H2)+ N(H I). This is the second highest redshift at which theseH2 lines have been detected (Ledoux et al. 2006b).


In Table 5.2, we have compiled metallicity measurements for z > 2 GRB absorptionsystems from the literature. As seen, despite its high redshift, the GRB 060206system has one of the highest metallicities measured for a GRB absorption systemand one of the highest metallicities measured from QSO absorption lines at z > 4.


Table 5.2 — Table of published absorption metallicities for GRBs. References: [1]Savaglio et al. (2003); [2] Vreeswijk et al. (2006); [3] Vreeswijk et al. (2004); [4] Watsonet al. (2006); [5] Berger et al. (2006b); [6] Starling et al. (2005a); [7] Chen et al. (2005a);[8] Ledoux et al. (2005); [9] Kawai et al. (2006); [10] this work.

GRB Metallicity Redshift Ref000926 [Zn/H]= −0.13 ± 0.25 2.038 1011211 [Si/H]= −0.90 ± 0.5 2.142 2030323 [S/H]= −1.26 ± 0.2 3.372 3050401 [Zn/H]= −1.0 ± 0.4 2.899 4050505 [S/H]� −1.2 4.275 5050730 [S/H]= −2.0 ± 0.2 3.968 6,7050820 [Si/H]= −0.6 ± 0.1 2.615 8050904 [S/H]= −1.3 ± 0.3 6.295 9060206 [S/H]= −0.84 ± 0.10 4.048 10

We also note that the metallicity of GRB 060206 is about 15 times higher than thatof GRB 050730 which has a similar redshift (z = 3.968). This shows that GRBsdo not only occur in very low metallicity environments, but also in environmentscovering a broad range of metallicities at a given redshift.

Fine-structure lines are ubiquitous in GRB absorbers (Vreeswijk et al. 2004;Berger et al. 2006b; Chen et al. 2005a). As discussed above, for GRB 060206 wederive high densities and low temperatures, consistent with the possible detectionof H2 in the spectrum, with a molecular fraction of log f ∼ −3.5. Molecules arealso detected in QSO-DLAs (Ledoux et al. 2003) in cold gas with T� 300 K, butwith lower densities (� 400 cm−3 Srianand et al. 2005). The difference in densi-ties along random (cross-section selected) and GRB sightlines like the one towardsGRB 060206 must reflect the higher than average density of the star forming regionin which the progenitor star was located. Such regions must have a cross-sectionwhich is less than a few percent of the total cross-section for QSO-DLAs at red-shifts z = 2–4 in order to explain that similar fine-structure lines have not yet beenseen in QSO-DLAs.

In the collapsar models GRBs can only be formed by massive single stars witha metallicity below ∼0.3 Z� (Hirschi et al. 2005; Woosley & Heger 2006). Sucha bias is not present in all models, most noteably not in models involving a binaryprogenitor (Fryer & Heger 2005). To gauge what a metallicity bias could meanfor the completeness of GRBs as cosmological probes it is important to know thepresent mass-weighted mean metallicity and how it declines with redshift. Zwaanet al. (2005) find a present-day mean metallicity in the gas phase of Z ≈ 0.44 Z�and a slope between −0.25 and −0.30 dex per unit redshift. This means that at z ≈ 1and earlier the mean metallicity of the gas is below the cut-off value above which


0 1 2 3 4 5 6Redshift

-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0[M

/H]

Figure 5.2 — Metallicity as a function of redshift for different classes of objects. Theblack circles are the measurements for GRBs from Table 5.2 (GRB 060206 is alsomarked with ×). The open triangles show the metallicity of three low-z GRB host galax-ies (Sollerman et al. 2005b). The squares and the dashed line represent the columndensity weighted metallicity evolution derived by Zwaan et al. (2005), their Fig. 22).The small dots with no error-bars are measurements for 121 DLAs from Prochaska et al.(2003). The hatched region indicates the metallicity above which GRBs cannot form inthe collapsar models.

single massive stars in the collapsar models do not make GRBs. It is thereforelikely that GRBs at z > 2 will be fairly unbiased tracers of star-formation, whilethey become increasingly biased at z < 1 if there is a low-metallicity bias.

Metallicities can be measured from optical spectroscopy of the afterglow or thehost galaxy. Fig. 5.2 shows that most metallicities for GRB absorption systemsfall below 0.3 Z�, but one (GRB 000926, Z = 0.7 Z�, Savaglio et al. 2003) hasa metallicity well above the threshold for a single massive star to produce GRBsin the collapsar models. The metallicity of the host galaxy of GRB 980425 bySollerman et al. (2005b) is also higher than this limit (0.8 Z�). This indicates thatcollapsars resulting from single massive stars are not the only progenitors to longGRBs or that massive stars with Z > 0.3 Z� can also produce long GRBs. The GRBmetallicities fall within the range spanned by QSO-DLAs, but most are above themean curve derived by Zwaan et al. (2005). The mean offset is 0.49 and the scatter0.38 dex (excluding GRB 050904 which is at very high redshift where there areno QSO-DLA data). This offset may reflect that DLAs are cross-section selected,whereas GRBs represent sightlines to more central regions in their hosts (Bloomet al. 2002). A metallicity gradient of −0.09 dex kpc−1 as assumed in the study


by Zwaan et al. (2005) (and observed in the Galaxy) seems sufficient to explainthe offset. Note that the current sample of QSO-DLAs is not likely to be stronglybiased against high metallicity sightlines (Akerman et al. 2005).

Furthermore, in the derivation of the cosmic star-formation density, and hencethe production of metals, the shape of the luminosity function is important.Roughly, two thirds of the UV light from LBGs is emitted by galaxies that arefainter than the actual flux limit of the ground based LBG survey of Steidel et al.(2003). Using only LBGs to derive the total star-formation density means that anextrapolation to the poorly determined faint end of the luminosity function is un-avoidable. GRBs allow us to probe this faint end since the selection is not limitedby the brightness of the host. We refer to Jakobsson et al. (2005b) for a quantita-tive analysis. Finally, GRBs allow the measurement of metallicities at very highredshifts (z > 6 Kawai et al. 2006), which is inaccessible to QSO-DLAs.

In conclusion, we measure a metallicity of [S/H] = −0.84 ± 0.10 forGRB 060206 at z = 4.048. We also find the first evidence for molecular lines from aGRB absorber. We have argued that GRBs can be used to measure the metallicitiesand luminosities of typical star forming galaxies at z > 2, making GRBs promisingcomplementary probes of chemical evolution at high redshift.

Acknowledgements The authors acknowledge the indispensable assistance givenby both observers and staff at WHT and NOT. We also thank C. Peroux, S. Ellison,and J. Andersen for helpful discussions. The Dark Cosmology Centre is funded bythe Danish National Research Foundation. KW and PC thank NWO for support un-der grant 639.043.302. AL, NRT and ER thank PPARC for support. The researchof JG is supported by the Spanish Ministry of Science and Education through pro-grammes ESP2002-04124-C03-01 and AYA2004-01515. JMCC acknowledges partialsupport from IDA and the NBI’s International Ph.D. School of Excellence. We also ac-knowledge benefits from collaboration within the EU FP5 Research Training Network“Gamma-Ray Bursts: An Enigma and a Tool” (HPRN-CT-2002-00294).

6The host of GRB 060206: dissecting kinematics of

a massive galaxy

C. C. Thone, K. Wiersema, C. Ledoux, J. P. U. Fynbo, R. L. C. Starling, A. Levan,J. Gorosabel, E. Rol, A. de Ugarte Postigo, N. R. Tanvir, R. A. M. J. Wijers,

P. Curran, A. J. van der HorstTo be submitted

Abstract The spectra of afterglows often give us detailed information on theline of sight towards high redshift γ-ray bursts (GRBs), allowing us to use GRBafterglows as sensitive probes of interstellar matter in GRB host galaxies, and thecircumstellar material around the progenitor star. In this paper we present earlyNOT/ALFOSC and WHT/ISIS optical spectroscopy of the afterglow of the gamma-ray burst GRB 060206 at z = 4.048 and deep imaging in r′ and H band. The res-olution and wavelength range of the spectra, the clear detection in emission of thehost and the bright afterglow facilitate a detailed study of the circumburst and hostgalaxy environment and the nature of the intervening systems that are seen in ab-sorption in the afterglow spectra. We detect several discrete velocity systems inthe host, all containing fine structure lines, best explained by shells within and/oraround the host, possibly created by starburst winds. We show how the combina-tion of afterglow spectra and detection of GRB hosts in emission can give valuableinsight into the evolutionary stages of galaxies at high redshift.

6.1 Introduction

In recent years, GRBs have been detected out to very high redshifts with the mostdistant so far having a redshift of z = 6.3 (Haislip et al. 2006; Kawai et al. 2006).Theory suggests, and observations agree, that ∼10% of Swift-detected gamma-raybursts (GRBs) will originate at redshifts z � 5 (Bromm & Loeb 2006) and ∼3%at redshifts z � 6 (Daigne et al. 2006). Furthermore, observed flux is not expectedto fade significantly with increasing redshift, hence GRBs should be detectableout to z � 10 (Ciardi & Loeb 2000) making them the most distant observableobjects in the Universe. As such, these events can teach us something about galaxyevolution and the early Universe: the occurence of GRBs in star forming regions inhigh redshift galaxies allows us to probe lines of sight to these galaxies with high

101

102 The host of GRB 060206

spectral resolution, by exploiting the brightness and simple spectral shape of GRBafterglows (e.g. Prochaska et al. 2006b; Starling et al. 2006).

Through studies of absorption lines in the spectra of high redshift GRBs, prop-erties of these distant galaxies can be determined, which is very hard to achieveotherwise. One key piece of information that can be derived from resonant metalabsorption lines which are shifted into the optical above z ∼ 2 is the metallicity inthe host galaxies. At low redshifts, GRB hosts have lower emission line metallicitythan the mean of the SDSS distribution (Savaglio 2006), but at high redshifts after-glow spectroscopy shows metallicities above the mean QSO absorber metallicity asa function of redshift (Fynbo et al. 2006a; Prochaska et al. 2007a).

QSO absorption systems have been used for a long time to study distant galax-ies. However, an association of the absorbing system with a galaxy in emissionis not always possible. In addition, QSO sightlines seem to probe the less denseouter regions of these galaxies, whereas GRBs reside in star forming regions in-side galaxies, so their sighlines allow us to probe the inner parts of the galaxyand the immediate environment of the burst. It is becoming apparent that all GRBafterglow spectra of bursts at sufficient redshift show fine structure lines such asSi II*,Fe II*, O I* and even higher levels up to Fe II**** have been detected (Bergeret al. 2006b). These lines have not been observed yet in QSO absorption spectra,and suggest the presence of massive stars. They are populated by photon pumpingor collisional excitation (in at least a few cases likely by the strong UV radiationfield of the GRB; Prochaska et al. 2006b) and may therefore be closely connectedto the GRB itself.

In our first paper on the dataset presented here (Chapter 5, Fynbo et al. 2006a,hereafter Paper I) we showed first results of the WHT and NOT spectra of the after-glow of GRB 060206, focussing on the metallicity, which we compared to a largesample of QSO absorber metallicities and the sample of known GRB afterglowmetallicities. We also reported our tentative detection of hydrogen molecules atthe host redshift in Paper I. In this second paper we analyze the full set of absorp-tion lines, and add a more detailed analysis of the properties of the host from thephotometric detection of the host galaxy and broadband afterglow properties.

The paper is organized as follows: In §2 we describe the observations and themeasurements of the absorption lines. In §3 we present the absorption line mea-surement method and results. In §4 we analyse the discrete velocity components inthe host through the detected (fine structure) lines and the line of sight extinction.We present the properties of the host galaxy in emission in §5. In §6 we discuss theabsorption line properties of the intervening absorbers, and evaluate the possibilitythat a galaxy close to the afterglow location is responsible for one of the interveningabsorber systems.

Throughout this paper we use the cosmological parameters H0 = 70 kms−1/Mpc, ΩM = 0.3 and ΩΛ = 0.7.

6.3 Absorption line analysis 103

6.2 Observations

We make extensive use in this paper of the dataset described in Paper I, to whichwe refer for details on the data and their reduction. For clarity, we reiterate here theimportant steps. We obtained a medium resolution spectrum with the Intermediate-dispersion Spectroscopic and Imaging System (ISIS) on the William Herschel Tele-scope (WHT) on 2006 February 6, starting 1.61 hours after the burst. We took twospectra with exposure times of 900 s each, with in between just the read-out time ofthe CCD. The spectrograph has a blue and a red arm, providing resolutions of 1.68and 0.82 Å, respectively. A redshift of z = 4.048 was determined from several ab-sorption systems within the host (Paper I, see also Section 6.3). At this redshift, theWHT spectra cover the restframe wavelength ranges of approximately 753 – 1030Å and 1228 – 1387 Å for the blue and red arm, respectively. The signal to noiseratio of the spectra is high, strongly aided by the sudden, strong brightening of theafterglow just prior to and during the spectroscopic observations (see Monfardiniet al. 2006; Curran et al. 2007b).

The optical afterglow of this burst proved bright, and owing to its fortunatesky location was observed with a large range of observatories (e.g., Wozniak et al.2006; Monfardini et al. 2006; Stanek et al. 2007, Curran et al. 2007b). A detaileddiscussion of photometric afterglow observations and the light curve physics canbe found in Curran et al. (2007b). We use these observations together with SwiftXRT data to analyse the broadband optical to X-ray SED to find the extinction (seeSection 6.4.3).

We observed the location of the afterglow on February 16, 2007, over a yearafter the burst, using the GMOS-N instrument on Gemini-North with the aim todetect the host and intervening absorber galaxies in emission. Observations wereperformed in the r′ band, which is similar to the SDSS r′ band, and consisted of9 × 300 seconds with a seeing of ∼1 arcsec and an airmass of ∼ 1.1. Photometrywas calibrated using standard star observations. The host galaxy is clearly detectedat the position of the afterglow. Close to the GRB host, there are several galaxiesthat might be responsible for the two intervening systems detected in the spectra,see Section 6.6. We observed the position of the host galaxy in the infrared on June3 2007 using the Omega2000 instrument on the 3.5m telescope at the Calar AltoAstronomical Observatory. Observations consisted of 50 × 60 seconds in H band.The host was not detected, and the 3σ detection limit at the coordinates of the hostis H = 20.6.

6.3 Absorption line analysis

We reported equivalent widths and column densities for a selection of absorptionlines given in Paper I, with the main aim of deriving a metallicity for the line ofsight material within the host. In this paper we present a full analysis of all absorp-


tion lines in the host galaxy and intervening systems. For the host galaxy, we detectfour main different velocity systems at z ∼ 4.0508, 4.0489, 4.0475and4.0442. Inaddition, there are (at least) two more velocity components with smaller columndensities. We also find intervening systems at z = 1.4790 and 2.2599, where thefirst one has two components, at z = 1.4790 and 1.4787.In order to fit the different components, we use the FITLYMAN package in MIDAS(Fontana & Ballester 1995); the fitting is complicated, however,by the strong blend-ing in a number of lines. The “cleanest” transition that has all velocity componentspresent is O Iλ1302, which we use as a template for the position of the differentcomponents. As shown in Fig. 6.1, S IIλ1260, C IIλ1334, S IIλλ1250,1253 andmost likely S IIλ1304 have all four strong velocity components, and added to thatare weaker components. S II and S IIλ1259 are blended, C II is blended with itsfine structure line C II* and S IIλ1304 is blended with O I*. N Vλλ1238,1242 andNi IIλ1317 appear not to be blended, but are not detected at all four velocity sys-tems: N V only at component 1, Ni II only at component 4. In those cases where weare able to fit multiplets with at least one unblended component, the column den-sity of the unblended line is given, and we can deduce the column of the blendedline from the column density of the entire blended system (as in case of S IIλ1259).From Si II*λ1309, the only unblended fine structure line, we see that the fine struc-ture lines are only present in the three redmost velocity components where the sec-ond and third components are weaker. Due to the blending, we fitted only the tworedmost components for O I**. The intrinsically blended lines Si II*λλ1264,1265were fitted together. For all host component lines, the turbulent b parameter(bturbulent), has been fixed to ∼14 km s−1 (see Paper I where this value was found

from S II lines). The total b parameter is then b =√

b2turbulent + b2

thermal. The inter-vening systems are unblended and the multiplet fits for the Mg IIλλ2796,2803 andFe IIλλ 2586,2600 doublets give consistent results for the column densities. Thedetections of Zn IIλ2062.6 and Cr IIλ2062.2 are rather weak, but the two lines ap-pear unblended.In Table 6.2 we present our measurements of column densities, doppler parametersand identifications of the lines detected and their different velocity systems and Ta-ble 6.2 gives the line fits for the intervening systems. Fig. 6.1 shows the red end ofthe WHT spectrum, Figs. 6.2 and 6.3 show the fit of the different components fordifferent lines in the host galaxy and the two intervening systems.

6.4 Velocity components in the host galaxy

The velocity components in the absorption lines span a range of about 500 km s−1.Zero velocity is chosen arbitrarily as the midpoint of the entire system as the ex-act redshift of the host galaxy is unknown. The components at -167 km s−1 and228 km s−1 (components 1 and 4) are relatively narrow with b around 15 km s−1,whereas some transitions of the two components in between have higher values of

6.4 Velocity components in the host galaxy 105Ta

ble

6.1

—Fi

tof

the

velo

city

com

pone

nts

inth

eho

stga

laxy

abso

rptio

nsy

stem

s.

Com

p.ID

λre

stλ

mea

s.z

log

Nb r

est

[km

s−1]

note

slo

gN

tot

1S

II12

50.5

8463

16.5

14.

0508

(5)

14.5

0±0.

1115

15.2

1±0.

382

6314

.41

4.04

91(7

)14

.79±

0.04

253

6312

.71

4.04

78(1

)14

.72±

0.07

254

6308

.13

4.04

41(5

)14

.30±

0.10

151

SII

1253

.811

6332

.81

4.05

08(5

)14

.50±

0.11

1515

.21±

0.38

263

30.7

04.

0491

(7)

14.7

9±0.

0425

363

29.0

04.

0478

(1)

14.7

2±0.

1025

463

24.4

14.

0441

(5)

14.5

0±0.

1015

1S

II12

59.5

1963

61.6

44.

0508

(5)

14.5

0±0.

1115

blen

ded

with

SiIIλ

1260

15.2

1±0.

382

6359

.52

4.04

91(7

)14

.79±

0.04

25bl

ende

dw

ithSi

IIλ

1260

363

57.8

14.

0478

(1)

14.7

2±0.

0725

blen

ded

with

SiIIλ

1260

463

53.2

04.

0441

(5)

14.3

0±0.

1015

1Si

II12

60.4

221

6366

.20

4.05

08(5

)14

.41±

0.09

1515

.23±

0.47

1a63

65.2

84.

0501

(2)

12.6

4±0.

1320

263

63.7

04.

0488

(6)

14.6

6±0.

0525

363

62.3

74.

0478

(1)

14.7

9±0.

0725

3a63

61.1

94.

0468

(8)

13.7

1±0.

0815

463

57.7

64.

0441

(5)

14.5

2±0.

0815

1Si

II13

04.3

702

6588

.18

4.05

08(5

)14

.41±

0.09

15bl

ende

dw

ithO

I*15

.23±

0.47

1a63

87.2

24.

0501

(2)

12.6

4±0.

1320

blen

ded

with

OI*

265

85.5

94.

0488

(6)

14.6

6±0.

0525

blen

ded

with

OI*

365

84.2

14.

0478

(1)

14.7

9±0.

0725

3a65

83.0

04.

0468

(8)

13.7

1±0.

0820

465

79.4

44.

0441

(5)

14.5

2±0.

0815

1O

I13

02.1

685

6577

.06

4.05

08(5

)16

.02±

0.53

152

6574

.86

4.04

91(7

)15

.72±

0.24

2516

.33±

0.72

365

73.1

04.

0478

(1)

15.5

2±0.

4725

3a65

71.8

84.

0468

(7)

14.9

1±0.

1320

465

68.3

34.

0441

(5)

15.1

4±0.

1815

106 The host of GRB 060206Table

6.1—

–continued

Com

p.ID

λrest

λm

eas.z

logN

brest [km

s −1]

noteslog

Ntot

1C

II1334.5323

6740.524.0508(5)

16.40±0.39

15<

16.851a

6739.554.0501(2)

13.67±0.42

202

6738.284.0491(7)

15.10±0.40

253

6736.474.0478(1)

<16.50

253a

6735.224.0468(7)

14.69±0.18

204

6731.584.0441(5)

16.09±0.24

151

NV

1238.8216257.46

4.0511(4)13.73±

0.0625

13.73±0.06

1N

V1242.804

6277.584.0511(4)

13.73±0.06

2513.73±

0.061

SiII*1264.7377

6388.004.0508(5)

14.17±0.04

15blended

with

SiII*λ1265

14.38±0.43

1a6387.08

4.0501(2)12.89±

0.1420

blendedw

ithSiII*

λ12652

6385.874.0491(7)

13.72±0.07

25blended

with

SiII*λ1265

36384.16

4.0478(1)13.41±

0.0525

blendedw

ithSiII*

λ12653a

6382.974.0468(7)

12.73±0.08

201

SiII*1265.0020

6389.334.0508(5)

14.17±0.04

1514.38±

0.431a

6388.414.0501(2)

12.89±0.14

20blended

with

SiII*λ1264

26387.21

4.0491(7)13.72±

0.0725

blendedw

ithSiII*

λ12643

6384.164.0478(1)

13.41±0.05

25blended

with

SiII*λ1264

3a6384.21

4.0468(7)12.73±

0.0820

blendedw

ithSiII*λ1264

1SiII*

1309.27576612.95

4.0508(5)14.17±

0.0415

14.38±0.43

1a6612.00

4.0501(2)12.89±

0.1420

26610.75

4.0491(7)13.72±

0.0725

36608.98

4.0478(1)13.41±

0.0525

3a6607.75

4.0468(7)12.73±

0.0820

1O

I*1304.8576

6590.644.0508(5)

14.86±0.07

1514.89±

0.402

6588.444.0491(7)

13.47±0.06

25blended

with

SiII3

6586.674.0478(1)

13.42±0.27

25blended

with

SiII1

OI**

1306.02866596.55

4.0508(5)14.15±

0.0915

14.37±0.26

26594.36

4.0491(7)13.98±

0.1225

1C

II*1335.7077

6746.464.0508(5)

15.91±0.22

1515.92±

0.521a

6745.494.0501(2)

13.08±0.38

202

6744.214.0491(7)

13.97±0.06

25

6.4 Velocity components in the host galaxy 107

Tabl

e6.

2—

Abs

orpt

ion

line

fittin

gof

the

two

inte

rven

ing

syst

ems

atz=

1.47

9an

dz=

2.25

9.T

hez=

1.47

9sh

ows

two

velo

city

com

pone

nts.

Com

p.ID

λre

stλ

mes.

zlo

gN

b res

t[k

ms−

1]

log

Nto

t

1Fe

II25

8664

12.7

21.

4791

(6)

13.2

7±0

.07

19.9±3

.413

.78±0

.22

225

8664

11.5

11.

4786

(7)

13.6

2±0

.07

141

FeII

2600

6446

.24

1.47

91(6

)13

.27±0

.07

19.9±3

.413

.78±0

.22

226

0064

45.0

31.

4786

(7)

13.6

2±0

.07

141

Mg

II27

9669

32.6

01.

4791

(6)

13.4

9±0

.08

19.9±3

.413

.95±0

.25

227

9669

31.3

01.

4786

(9)

13.7

7±0

.11

141

Mg

II28

0369

50.4

01.

4791

(6)

13.4

9±0

.08

19.9±3

.413

.95±0

.25

228

0369

49.0

91.

4786

(9)

13.7

7±0

.11

141

Zn

II20

62.7

6724

.06

2.25

98(9

)13

.00±0

.06

22.0

1C

rII

2062

.267

22.6

52.

2598

(9)

13.1

8±0

.11

22.0


Figure 6.1 — Fits to the different velocity components in the host absorption lines wherethe normalized flux is plotted against velocity. The dashed lines indicate velocity com-ponents, and the dotted line the adopted zero velocity. O Iλ1302 serves as the templateline for the absorption systems. Note that some components of S IIλ1259 are blendedwith S IIλ1260 and the fit shown is derived from a multiplet fitting using the S IIλ1250lines as reference. Note that for N V the position of the components deviates slightlyfrom the ones in O I, which might point to a slightly different spatial origin in the host.O I*λ1304 and O I**λ1306 are strongly blended, and not all components could be reli-ably fit. The Si II*λλ1264,1265 lines are strongly blended and have been fit together.

b. Not all lines show all four main velocity components, which indicates differentconditions in the different velocity systems. Ni IIλ1317 is only found in the lowestredshift component, and N Vλλ1238, 1242 only in the reddest velocity component,and their redshifts are slightly different from the redshifts derived from the tem-plate O Iλ1302 line system. The fine structure lines are strongest in the redmostcomponent, considerably weaker in the second and third component and absent inthe bluest one.

In the highest redshift component (though slightly offset in velocity), the highionisation N V line is detected, together with the low ionisation lines. N V hasbeen detected in many QSO spectra, but also in lines of sight towards stars in theLMC (e.g. Caulet 1996), Galactic sightlines towards early type stars (e.g. Welsh &Lallement 2005; Savage et al. 2001), in WR star winds and in galactic outflows. Itshigh ionization potential (77.5 to 97.9 eV) means it is found in rather hot, highlyionized interstellar regions, and in the solar neighbourhood often at the (conductive)interfaces between evaporating hot and cooler interstellar gas (Slavin 1989; Welsh


Figure 6.2 — The normalised WHT ISIS red arm spectrum. The metal absorption linesof the host and the intervening systems are indicated.

& Lallement 2005), e.g. at the boundary of a hot bubble formed by O and B starswith the neutral ISM, where one would expect to find both high ionization andlow ionization species at comparable velocities. The observed high b of N V isconsistent with this picture (Savage et al. 2001).

6.4.1 Comparison with QSO-DLAs

To compare the kinematics in the afterglow spectrum with QSO-DLAs we followthe procedure of Ledoux et al. (2006a) and calculate the line profile velocity width,Δv, as c [λ(95%) − λ(5%)] /λ0, where λ(5%) and λ(95%) are the wavelengths cor-responding to, respectively, the five per cent and 95 per cent percentiles of theapparent optical depth distribution, and λ0 is the first moment (the average) of thisdistribution (see Fig. 1 of Ledoux et al. 2006a). We choose the S IIλ1253 transitionas it is a low ionisation transition and the line appears not saturated. The apparentoptical depth for the line at the derived velocity width is shown in Fig. 6.4. We infera velocity width of 296 km s−1 consistent with the distribution of DLAs with simi-lar metallicities (Ledoux et al. 2006a, their Fig. 2; see also Prochaska et al. 2007b).This is important as it indicates that the kinematics of GRB absorbers is similar to


Figure 6.3 — Fit to the intervening systems in the line of sight at redshift z = 1.4, wherenormalised flux is plotted against velocity. The two Mg II lines and the two Fe II lineshave been fit as a doublet, and show two velocity systems indicated by dashed lines. Thedotted lines denotes the zero velocity.

that of QSO-DLAs and hence that the immediate environment of the GRB is notstrongly dominating the kinematics and metallicity of the system. This is consistentwith the recent findings of Vreeswijk et al. (2007) and Watson et al. (2007) that theGRB absorbers probe gas that can be more than several kpc away from the GRBsite.


Figure 6.4 — Apparent optical depth distribution of the S IIλ1253 velocity system in thehost galaxy, with full width W2. The top panel shows the integration of the apparent op-tical depth starting at wavelength λ1 and ending at λ2, indicated also by the two verticaldashed lines. The horizontal bar shows the measured velocity width (W1) when 5% ofthe total apparent optical depth is avoided at both edges of the profile, see Section 6.4.1.

6.4.2 Fine structure levels

In Paper I we used the detected fine structure lines to determine the density andtemperature for the highest redshift component, assuming that the lines were ex-cited by collisional excitation. It has been shown through the VLT UVES RRMcampaign on GRB 060418 (Vreeswijk et al. 2007) that at least some of the absorp-tion complexes in GRB afterglows are excited by indirect UV pumping from thebright GRB and its afterglow, rather than by collisional excitation, which was alsoput forward by Prochaska et al. (2006b) who analyzed the spectra of GRB 050730and GRB 051111. Vreeswijk et al. (2007) measure strong variation with time of thecolumns of absorption lines, and use this to obtain a direct distance determinationof the absorbing material with respect to the GRB, by fitting the columns and theirtime dependence with indirect UV pumping models. Our spectra of GRB 060206do not have the required time-base and spectral resolution for such an analysis. Wecan, however, determine approximate properties for the absorbers assuming that in-direct UV pumping or collisional excitation is the way that the fine structure levelsare populated. GRB 060206 has a complex spectrum: the velocity components weobserve have a rather large spread in velocities (e.g. GRB 050730 has componentswith a spread up to ∼ 100km s−1, while for 060206 we observe a velocity spread of∼500 km s−1), and the combined WHT spectrum was taken when a significant opti-cal rebrightening occured in the afterglow light curve (see the light curve by Curran


Table 6.3 — Properties of the four velocity systems observed in the WHT spectrum, seeSection 6.4.2. The diagnostic ratios log(O I*/O I**), log(S II/Si II*), and log(O I/O I*)have larger errors and are not as constraining as the Si lines, and are therefore not listedin this table. Electron densities ne have been derived using an ionization fraction of 10−2.Note that G0 = 1.6 × 10−3 erg cm−2 s−1 (Prochaska et al. 2006b).

Comp. Rel. velocity log(Si II/Si II*) ne G/G0

[km s−1] [cm−3]1 −167 0.24 ± 0.10 ∼ 103 ∼ 5 × 105

1a -0.25 ± 0.19 � 103 ∼ 106

2 −62 0.94 ± 0.09 ∼ 102 ∼ 105

3 10 1.38 ± 0.09 ∼ 30 ∼ 104

3a 0.98 ± 0.11 ∼ 102 ∼ 105

4 228 > 1.4 � 30 < 104

et al. 2007b). UV photons from this rebrightening may have played a significantrole in the indirect UV excitation of the transitions.

The highest redshift component shows the high ionisation N V line as wellas the highest occupation of fine structure levels. The O I*λ1305 line is likelysaturated (see also D’Elia et al. (2007)). As we have only one usable fine structuretransition for components 2,3 and 4 (Si II*), we can not solve for both (electron)density and temperature for these systems in the case of collisional excitation.

We can use absorption line ratios to estimate the UV radiation field needed toexcite the levels, following Prochaska et al. (2006b), assuming that UV pumping isthe dominant excitation mechanism for this component. The column density ratiosSi II*/Si II λλ 1309, 1304 and O I**/O I*λλ1305, 1302 indicate a high radiationfield intensity G/G0 (see Table 6.3). Note that the diagnostic line ratios of otherelements and transitions have too large uncertainties to constrain the parameters.

The high G value required for the highest redshift component is similar tothat found for the component with the second-highest redshift in the spectrum ofGRB 050730 (D’Elia et al. 2007), and is somewhat lower than the value foundfor GRB 051111 (Prochaska et al. 2006b). We can in this spectrum directly use[Si II/Si II*] for all redshift components, but note that it is possible to use S II in-stead of Si II in cases where the blending is too severe, as indeed [S/Si] = 1 (Table6.2). If all the velocity components in the host are excited through indirect UVpumping by photons of the GRB and afterglow, it is evident that the highest red-shift component is likely closest to the burst, and the lowest the furthest: we findthat the required photon field drops significantly, for the three lower redshift ab-sorbers.

The required electron densities in the case of collisional excitation are 103 cm−3

and less for the three lower velocity systems (see Table 6.3), where we use the


prescription in Prochaska et al. (2006b), and a ionization fraction of 10−2. Thesevalues are comparable to the electron densities found in star forming regions andmolecular clouds in our Galaxy and several GRB host galaxy star forming regions(e.g. Wiersema et al. 2007; Thone et al. 2006 and references therein).

We note that for component 2 the columns N(Si II*) > N(O I*), whichProchaska et al. (2006b) note is not possible for either collisonal or indirect UVpumping excitation, except in the case of collisonal excitation in a predominantlyionized medium, so rather different from the 10−2 assumed earlier.

The three lowest redshift systems can be explained with collisional excitationas the required electron densities are feasible, but the highest redshift componentis a better candidate for indirect UV pumping. Prochaska et al. (2006b) providein their Figure 13 a comparison between the predicted columns of the first excitedstates of Fe II and O I through collisional and indirect UV pumping mechanismswith respect to Si II. Unfortunately we have no detection of Fe II* lines due to ourlimited wavelength range. Our uncertainties in the column of O I* are too large todecide between the two excitation models.

At the required high densities of the highest redshift component in the case ofcollisional excitation, we expect to see neutral carbon, iron and magnesium in ab-sorption (see Prochaska et al. 2006b. In the highest redshift component we do notfind C I, and neutral Fe and Mg lines are not in our wavelength range. No neutralspecies with ionisation potentials lower than hydrogen are observed in the highestredshift absorber (O I has an ionisation potential over 1 Ryd and can therefore bescreened by hydrogen), so we cannot set a constraining lower limit to the distanceto the GRB for this system through ionisation analysis. When we assume that indi-rect UV pumping by the afterglow of the GRB is exciting the levels in the highestredshift system, we can derive an order of magnitude estimate of the distance to theGRB (see Prochaska et al. 2006b) using the value of G found above. We use theafterglow light curve (Curran et al. 2007b) to calculate the flux received at Earthat the time of the spectrum, using the spectral index derived in Section 6.4.3 andthe restframe frequency of 1.7 × 1015 Hz as in Prochaska et al. (2006b). We find adistance of the highest redshift system to the GRB of order one kiloparsec, whichis similar to the accurate distance found for the absorbing gas through absorptionline variability in GRB 060418 (Vreeswijk et al. 2007).

6.4.3 Extinction along the line of sight

To obtain limits on the absorption along the line of sight to GRB 060206, we fit thenear-infrared to X-ray spectral energy distribution (SED) in count space (Starlinget al. 2006), and using the metallicity as determined through the sulphur lines inPaper I of [S/H] = -0.84. The SED is created using RJHKS photometry and a SwiftXRT X-ray spectrum centred at ∼ 1 × 104 s after burst, detailed in Curran et al.(2007b). We note that this is close in time to our WHT spectra, which were takenat 0.67 ×104 s after burst. We perform fits using an absorbed single power law, and


then a broken power law to model a possible cooling break.The SED is well fit with a single power law with β = 0.93 ± 0.01 and χ2/dof

= 1.05. A broken power law does not provide a significant improvement in the fitaccording to the F-test. We find no evidence for intrinsic (z = 4.048) optical orX-ray extinction above the Galactic values. At this high redshift of z = 4.048 themetal edges that dominate the X-ray extinction are shifted out of the XRT energyrange and are not sensitively probed. The near-infrared and optical data at thisredshift, however, probe restframe UV and blue bands, and are therefore sensitiveto extinction. We estimate a 3σ upper limit for the intrinsic optical/UV extinctionof E(B−V) < 0.01. This shows that the intervening systems as well as the velocitysystems in the host have low dust content, which can be used to distinguish betweendifferent origins of these systems.

Figure 6.5 — The SED of the afterglow 1 × 104 seconds after burst using RJHKS bands(Curran et al. 2007b), fit using a SMC extinction curve and a metallicity of [S/H] =–0.84. The SED is shown unfolded in counts space. No extinction above Galacticextinction is required in the fit.

6.4.4 The nature of the components

Several explanations can be found for the nature of these components, and sim-ilar absorption features have been discovered in other GRB spectra such asGRB 020813 (Fiore et al. 2005), GRB 021004 (Fiore et al. 2005), GRB 030329


(Thone et al. 2006), GRB 050730 (Starling et al. 2005a; Chen et al. 2005a) andGRB 051111 (Penprase et al. 2006; Prochaska et al. 2006b). The velocity range ofthese absorption systems span a wide range from 200 to 3000 km s−1 and mighthave different origins according to their widths.A first possibility would be that these features are somehow connected to the pro-genitor of the GRB which is assumed to be a massive Wolf-Rayet (WR) star inmost long-duration GRB progenitor scenarios, with considerable mass loss beforecollapse. Van Marle et al. (2005) numerically evaluate a scenario in which the pro-genitor star wind interacts with circumstellar material, including the mass loss ofearlier stages in the progenitor evolution, forming shells of material propagatingoutwards at a range of velocities. The velocity components of up to 500 km s−1

we observe in our spectrum match very well with the models predictions by vanMarle et al. (2005), but there should also be high velocity components (2000–3000km s−1) from the WR wind present, which we do not detect. In the spectra ofGRB 021004, the observed lines in these high velocity components were the highionization C IV and Si IV, which are outside the usable range of our WHT spec-trum for GRB 060206. We are therefore unable to make a direct comparison withGRB 021004, or the sample of high resolution GRB afterglow spectra that wassearched for high velocity C IV absorbers by Chen et al. (2006b). Another issue isthe detection of strong low ionization lines in all components, which disfavours anorigin very close to the burst as the radation field from the GRB should ionize mostof its surrounding material.

If a connection to the GRB immediate environments can be excluded, the ab-sorption features likely originate in the ISM of the host galaxy. Absorption systemswith small velocity spreads (v < 250 km s−1) have been found in a number of highresolution spectroscopic studies of QSO absorbers using mainly Mg II and C IVsystems (e.g. Nestor et al. 2005; Churchill et al. 2003; Boksenberg et al. 2003).These systems are often connected to strong Lyman-α absorption. They usuallyconsist of a strong, saturated central profile and some weaker subsystems at highervelocities, similar to GRB 060206. Churchill et al. (2003) and Ellison et al. (2003)suggested a rotating disc together with a halo to explain the different systems. Thiscould also be a possible scenario for what we observe in GRB 060206, where thehost has been found to be rather massive (see Section 6.6). This would imply thatthe GRB is somewhere behind the galaxy, shining through regions with differentdensities, with the densest region possibly being the central region of the galaxy.It would then also explain the differing properties between the different velocitysystems. However, as shown in Section 6.4.3, the extinction in the line of sight tothe GRB is consistent with zero, which makes it unlikely that the GRB was situatedbehind most of the disc material of a massive galaxy. Also, the hydrogen columndensity of log N = 20.85 ± 0.1 we find in the NOT ALFOSC spectrum (Paper I) isquite low for a GRB-DLA (Jakobsson et al. 2006a), which might imply a locationfor the GRB inside or on the edge of the galaxy.

Another possible explanation for the velocity components in QSOs and GRBs


is a supergalactic wind in the line of sight (Bond et al. 2001). The cumulativesupernova explosions within a star forming region can expel gas out into the haloof the galaxy which fragments and can produce clumps of material outside the mainpart of the galaxy. These galactic winds have been directly observed in a range ofnearby galaxies (e.g. M82, NGC 3079 and NGC 4945). The inner parts of thesesuperbubbles are relatively hot and show X-ray emission and optical emission lines,surrounded by a warm transition phase with highly ionized metal absorption linessuch as O VI and C IV. The outer shells, however, are colder and show low ionizedabsorption systems (see e.g. Heckman 2001). Highly ionised plasma has beencommonly observed in absorption towards QSO sightlines (Adelberger et al. 2005),reaching velocities of several hundreds km s−1. In many cases these systems canbe explained through outflows, but are mainly seen either in Lyman-α emission orC IV and Si IV absorption. Low ionization absorption features in GRB spectra thatcan be explained through galactic winds have been seen in GRB 030329 (Thoneet al. 2006) and GRB 051111 (Penprase et al. 2006) and it is also the favouritescenario for the absorption systems towards QSOs (Bouche et al. 2006).

These galactic winds are considered to be the main cause for the metal enrich-ment of the intergalactic medium. If this is the scenario for the absorption lineswe see in the spectra here, from a burst at redshift z = 4.048, it shows that thesemechanisms began quite early in the history of the Universe. To conclusively dis-criminate between the possibility of a galactic outflow and absorption in differentregions within the galaxy, we require the exact redshift of the galaxy from emissionlines as was possible for GRB 030329, Thone et al. (2006), for which the absorp-tion systems of Mg I and Mg II spanned a velocity range of 260 km s−1 and weremostly blueshifted compared to the host emission lines.

6.5 The host in emission

In the deep r′ band Gemini images of 2700 seconds exposure time, we detect ex-tended emission at the position of the afterglow (see Fig. 6.6). The total magnitudeis r′ = 25.0 ± 0.2. It is likely that some or all of this emission is from the hostgalaxy of GRB 060206, but spectroscopy is required to firmly establish this. Wealso detect a few other galaxies very close to the host candidate which might beassociated with the two intervening absorption systems discussed in Sect. 6.6.

At a redshift of z = 4.048, r′ = 25.0 corresponds to an absolute magnitude of−22.8. This corresponds to L∗ in the Schecter function fit to the luminosity functionof z = 4 Lyman-break galaxies (LBG, Steidel et al. 1999). Such a bright galaxy isquite exceptional for a long GRB host galaxy (see, e.g. Fruchter et al. 2006). If longGRBs are intrinsically biased towards low metallicity it is, however, natural that an∼L∗ host is found at relatively high redshift when the metallicity of L∗ galaxieswere lower.

At z = 4.048, the r′ filter with center wavelength 6300Å and width 1400Å

6.6 Intervening systems 117

Figure 6.6 — Gemini r′ band image, with the host candidate for GRB 060206 indicatedby tickmarks. Indicated by an arrow is the galaxy likely causing one of the interveningabsorption systems, North of the host. The vertical bar denotes a 10 arcsecond distance.Note that the orientation is North up, East left.

corresponds to restframe central wavelength of 1250 Å and a width of 277 Åcorresponding to a intermediate width filter almost centred on the Lyα emissionline. Several GRB hosts have been probed for Lyα emission, both through narrow-band imaging and through spectroscopy of the host or the afterglow (e.g. Fynboet al. 2003; Jakobsson et al. 2005a), and in most cases the hosts were found to bebright Lyα emitters, with restframe equivalent widths up to ∼145 Å (GRB 030323;Vreeswijk et al. 2004). Therefore, a significant part (up to approximately 50%) ofthe flux of the r′ band emission from the likely host is therefore possibly from theLyα emission line.

A detection of a GRB host galaxy at such high redshift is relatively rare. Sofar, only three hosts for a long GRB with redshift z > 3 have been detected:GRB 971214 (z = 3.42, Kulkarni et al. 1998), GRB 000131 (z = 4.50, Ander-sen et al. 2000; Fruchter et al. 2006) and GRB 030323 (z = 3.372, Vreeswijk et al.2004) despite the fact that 26 GRBs are known at z > 3 so far.

6.6 Intervening systems

In addition to the systems intrinsic at the GRB redshift, we detect two interveningabsorbers. The one at z = 1.4787 contains two velocity systems of the two doubletsMg IIλλ2796, 2803 and Fe IIλλ2586, 2600 with a separation of ∼50 km s−1. Thesecond absorber is found at z = 2.2599 and we detect Zn IIλ2062 and Cr IIλ2062.


6.6.1 The z = 1.4787 system

Hao et al. (2007) claim significant variability in the equivalent widths of the Mg IIdoublet and the Fe IIλ2600 line, in a time interval between 4.13 hrs and 7.63 hoursafter the burst, covered by 7 low resolution spectra. They fit the lines profiles withGaussian profiles, and report variation in the equivalent widths of the lines, as wellas in the relative velocities of the lines. The two WHT spectra we present in thispaper and Paper I (each 900 seconds exposure taken consecutively) have been takenearlier than the ones from Hao et al. with mid-points 1.73 and 2.01 hours after burst.Our spectral resolution and signal to noise are significantly better than the spectraused by Hao et al., so we may expect to measure values that continue the trend inequivalent width and velocity variation claimed by these authors.

We find that both the Fe II and Mg II lines can be split into two seperate veloc-ity components which will be unresolved in low resolution spectra, but we measureequivalent widths of these two systems combined to allow for fair comparison withthe Hao et al. values. We detect no significant variation in equivalent width be-tween our two exposures, see Table 6.4. The trend reported by Hao et al. (2007)of a rapidly decaying Fe II and Mg II equivalent width until ∼5-6 hours after burstis clearly in disagreement with our data, and our values agree within ∼2 σ witha weighted average of the Hao et al. (2007) values. We note that the explanationthat Hao et al. put forward for the variability of these lines, namely the compa-rable angular size of the GRB blastwave and the absorber, would also hold forabsorbing material in (the halo of) GRB host galaxies themselves, but variation ofstable atomic transitions in host galaxies can be excluded with high levels of con-fidence from the UVES RRM time-series observations reported in e.g. Vreeswijket al. (2007), and longer time-base monitoring of other bursts (e.g. GRB 021004,030329, 050730).

Churchill et al. (2003) find that the column densities of Mg II and Fe II corre-late linearly. The column densities we find for both velocity systems fall within 1sigma of the linear correlation of Churchill et al (2003). The b values we find arevery high for both components (compared with the sample of Churchill et al.), ei-ther pointing to very high temperatures or that there are more velocity componentspresent that we can not resolve at our intermediate resolution. This last possibilityseems backed up by the correlation that Churchill et al. find between the restframeequivalent width of the Mg IIλ2796 line and the number of velocity components inthe absorber, which predicts a higher number of velocity components (∼ 10) thanthe two we find through our analysis.

6.6.2 The z = 2.2599 system

The z = 2.2599 absorber is detected in both Zn IIλ2062.7 and Cr IIλ2062.2 lines,commonly observed in QSO absorbing systems. Both Cr IIλ2066 and Zn IIλ2026lines can not be measured in this absorber due to blending with other lines and

6.6 Intervening systems 119

Table 6.4 — Measurements of the lines reported to vary by Hao et al. 2007, in the two900 second WHT spectra. Uncertainties are conservative 1σ errors. Equivalent widthsare as observed.

Spectrum EW Mg II 2796 EW Mg II 2803 EW Fe II 26001 2.39 ± 0.09 2.22 ± 0.09 1.23 ± 0.092 2.59 ± 0.13 2.23 ± 0.13 1.12 ± 0.11Combined 2.40 ± 0.08 2.19 ± 0.08 1.13 ± 0.08

residuals from skylines. The ratio of the column densities of these two speciesprovides a measure of the dust depletion in the material that is probed by the line ofsight. We assume a ratio (Cr/Zn)� = 10.72 (Anders & Grevesse 1989) and use thenotation [Cr/Zn] = log(Cr/Zn) − log(Cr/Zn)�. QSO DLA absorbers show typicalvalues of Cr/Zn of [Cr/Zn] ∼ −1.0 − 0 (e.g. Akerman et al. 2005). We find that[Cr/Zn] = −0.85 ± 0.2, indicating that a significant fraction of Cr in this absorberhas been depleted onto dust. The non-detection of dust in the afterglow indicatesthat the quantity of dust is low.

6.6.3 Identifying the absorber in emission

Detecting intervening galaxies in emission is complicated in the case of QSOs, butthe transience of GRBs makes these more suitable to detect intervening systems inemission once the afterglow has faded. While intervening systems can produce veryclear absorption systems, the identification of specific absorbers with individualgalaxies is difficult (e.g. Vreeswijk et al. 2003; Ellison et al. 2006). The deepGemini image shows several galaxies within a radius of 5 arcseconds, with themost promising candidate a mere ∼2.5 arcsec from the host, a distance of ∼ 21kpc at z = 1.479, and ∼ 20 kpc at z = 2.259. Photometry of this galaxy yieldsa magnitude of r′ = 23.9 ± 0.1. We can also not exclude the possibility that thehost candidate, which has r′ = 25.0 ± 0.2, is in fact one of the intervening systems,obscuring the real host galaxy. The position of this galaxy right on the afterglowposition makes this scenario less likely.

At z = 1.479, the Gemini r′ band (with center wavelength 6300 Å and width1400 Å) corresponds to restframe 2540 Å (width 560 Å), at z = 2.259 this is 1933Å (width 430 Å). In both cases restframe UV emission is probed, and in the caseof z = 1.479 this filter is close to 2800 Å. We estimate the flux of the galaxynorth of the host at 2800 Å therefore at ∼1.2μJy, which corresponds to a UV starformation rate of ∼ 25 M�yr−1 (using the conversion from Kennicutt 1998). Weuse the Madau et al. (1996) relation to estimate the SFR for the galaxy were it atz = 2.259, and find a star formation rate ∼ 66 M�yr−1.

Correlations between parameters of the Mg II absorption lines and the galaxythat produces them is a topic of much study. Recently, Zibetti et al. (2007) used


the large dataset of the SDSS to correlate the Mg II absorber properties with thelight of SDSS galaxies at a range of impact parameters (from 10 – 200 kpc) andfound that stronger absorbers can be related to bluer galaxies. The rest equivalentwidth of the Mg IIλ2796 clearly places it in the category of the strong Mg II ab-sorbers (though the definition of a strong absorber varies from author to author, butwe assume W0(2796) � 1Å), which are generally associated with relatively brightgalaxies, with luminosities from ∼0.1 to several L∗. At the redshift interval con-sidered by Zibetti et al. (2007), z ∼ 0.4 − 1, the best fitting absorber template isan intermediate type galaxy (Sbc spirals), with negligible SED evolution over theconsidered range in redshift, though as the strength of the Mg II line increases, theblue colour of the galaxy seems to get stronger. It is also apparent that the sample ofaverage luminosity of sources increases with redshift. This is not inconsistent withan identification of the galaxy closest to the host as the z = 1.4787 absorber, thoughthe redshift of this intervening absorber is too high for quantitative comparison withthe SDSS sample. The non-detection of the host in the H band is consistent withthe intervening absorber candidate being a blue star forming galaxy.

Ellison (2006) show that the velocity spread of the Mg II absorber in combi-nation with the restframe equivalent width (the D parameter) supplies an efficientmeans to select possible DLAs. We find for our absorber D ∼ 8, making it likelythat this absorber contains a DLA. The blue, luminous and rapidly star forminggalaxy very nearby the GRB host fits in this picture.

We note that comparison with QSO absorber samples may be further compli-cated by the recently established over-abundance of Mg II absorbers on GRB linesof sight compared to QSO sightlines (Prochter et al. 2006), for which the cause isyet unclear. This difference is not present in line of sight detections of C IV, forreasons not fully understood (Sudilovsky et al. 2007; Tejos et al. 2007).

6.7 Conclusions

In this paper we have presented the full set of detected absorption lines in themedium resolution spectrum from WHT, as well as host galaxy observations withdeep imaging in r′ and H band, using Gemini North and the CAHA 3.5m tele-scope. The spectra show at four main velocity components accompanied buy sev-eral weaker absorbers, spread over a range of ∼ 500 km s−1. For each we detecta series of absorption lines, including several species of fine structure lines thatare not found in QSO absorbers, but seem abundant in GRB lines of sight (e.g.Vreeswijk et al. 2004; Prochaska et al. 2006b; Fynbo et al. 2006a; Kawai et al.2006; Starling et al. 2005a; Penprase et al. 2006). We measure the column densi-ties through Voigt profile fitting. Not all lines are measureable due to the blendingof the lines of different velocity systems, but for every velocity component a di-agnostic set of absorption lines can be measured. The four velocity componentsshow decreasing occupation of their excited fine structure levels as the redshift get

6.7 Conclusions 121

smaller. The column density ratios of the Si II lines indicate that lines in the com-ponent with the highest redshift are either collisionally excited in a dense mediumwith nHI ∼ 105 cm3, or are excited through indirect UV pumping by a strong pho-ton field equivalent to ∼ 5 × 105G/G0, produced by the GRB and afterglow. Thepowerful rebrightening of the afterglow (Curran et al. 2007) may play an importantrole in this excitation. If indeed indirect UV pumping through the afterglow is thesource of excitation, the highest redshift absorber is of the order a kiloparsec distantfrom the burst.

The other three velocity systems that have a lower redshift require a signifi-cantly lower density in the case of collisional excitation, or a lower photon field inthe case of indirect UV pumping. If the latter is the case, than these componentsare progressively further away from the burst. We are unable to make a decisivedistinction between the two excitation mechanisms, as no other fine structure linesare within our spectral range. The similarities of the highest redshift componentwith the values for GRB 050730 and GRB 051111 and the direct detection of ab-sorption line variability in GRB 060418 make an origin of the lines through indirectUV pumping likely. While the fine structure level analysis helps to understand therelative positions of the velocity components with respect to the GRB, their natureis not clear, but could be clarified through high resolution (space based) imagingand emission line spectroscopy of the host.

The host galaxy is detected at r′ = 25.0 ± 0.2, and an upper limit of H = 20.6(3 σ) is achieved, indicating a bright (L � L∗) host galaxy. The line of sightmetallicity of the host is relatively high at [S/H] = -0.84 ± 0.10, and we fit thevelocity width of the S IIλ1253 line to investigate if this host obeys the possiblemass-metallicity relation found through this method in QSO absorbers (Ledouxet al. 2006a). We find a velocity width of 296 km s−1, consistent with the QSOdistribution, indicating that the detection of GRB hosts at such a high redshift canprovide valuable information on correlations between galaxy mass, luminosity andmetallicity, as GRBs probe different environments than QSOs (e.g. Prochaska etal. 2007).

In the Gemini image we find several galaxies close to the afterglow position,with the closest one within 2.5 arcseconds. In the spectrum we see several inter-vening absorbers, at z = 1.4787 and z = 2.2599. We analyse the absorption linesdetected of these systems, and find that the bright galaxy is likely to cause thestrong Mg II and Fe II seen at z = 1.4787. This galaxy is not detected in H band,consistent with it being a spiral or irregular galaxy, which is also suggested fromthe absorption line properties. Hao et al. (2007) claim significant variability in theequivalent width of the Mg II doublet and the Fe IIλ2600 line in the z = 1.4787 ab-sorber. Our spectroscopy at higher signal to noise and spectral resolution has beentaken much earlier after the burst, and does not confirm the variability in equivalentwidth of these lines or the trend in the variability as a function of time.

We note that the high resolution spectroscopy of GRB 060206 has been takenwith WHT, demonstrating that these detailed line analyses are clearly possible with


4m class telescopes when a bright afterglow presents itself and intermediate or highresolution spectroscopy is possible.

Acknowledgements We thank the observers and ING staff for performing the re-ported observations. The Dark Cosmology Centre is funded by the Danish National Re-search Foundation. KW, PC and RW thank NWO for support under grant 639.043.302.The authors acknowledge benefits from collaboration within the EU FP5 ResearchTraining Network “Gamma-Ray Bursts: An Enigma and a Tool” (HPRN-CT-2002-00294).

Based on observations made with the Nordic Optical Telescope, operated on theisland of La Palma jointly by Denmark, Finland, Iceland, Norway, and Sweden, andwith the William Herschel Telescope, in the Spanish Observatorio del Roque de losMuchachos of the Instituto de Astrofısica de Canarias. Based on observations obtainedat the Gemini Observatory, which is operated by the Association of Universities forResearch in Astronomy, Inc., under a cooperative agreement with the NSF on behalfof the Gemini partnership: the National Science Foundation (United States), the Parti-cle Physics and Astronomy ResearchCouncil (United Kingdom), the NationalResearch-Council (Canada), CONICYT (Chile), the Australian Research Council (Australia),CNPq (Brazil), and CONICET (Argentina).

7The nature of the dwarf starforming galaxy

associated with GRB 060218 / SN 2006aj

K. Wiersema, S. Savaglio, P. M. Vreeswijk, S. L. Ellison, C. Ledoux, S.-C. Yoon,P. Møller, J. Sollerman, J. P. U. Fynbo, E. Pian, R. L. C. Starling, R. A. M. J. Wijers

Astronomy and Astrophysics, 464, 529 (2007)

Abstract We present high resolution VLT UVES and low resolution FORSoptical spectroscopy of supernova 2006aj and its host galaxy, associated with thenearby (z = 0.03342) gamma-ray burst GRB 060218. This host galaxy is a uniquecase, as it is one of the few nearby GRB host galaxies known, and it is only thesecond time high resolution spectra have been taken of a nearby GRB host galaxy(after GRB 980425). The resolution, wavelength range and S/N of the UVES spec-trum combined with low resolution FORS spectra allow a detailed analysis of thecircumburst and host galaxy environments. We analyse the emission and absorp-tion lines in the spectrum, combining the high resolution UVES spectrum with lowresolution FORS spectra and find the metallicity and chemical abundances in thehost. We probe the geometry of the host by studying the emission line profiles. Ourspectral analysis shows that the star forming region in the host is metal poor with12 + log(O/H) = 7.54+0.17

−0.10 (∼ 0.07 Z�), placing it among the most metal deficientsubset of emission-line galaxies. It is also the lowest metallicity found so far for aGRB host from an emission-line analysis. Given the stellar mass of the galaxy of∼ 107M� and the SFRHα = 0.065 ± 0.005 M� yr−1, the high specific star formationrate indicates an age for the galaxy of less than ∼ 200 Myr. The brightest emissionlines are clearly asymmetric and are well fit by two Gaussian components separatedby ∼ 22 km s−1. We detect two discrete Na I and Ca II absorption components atthe same redshifts as the emission components. We tentatively interpret the twocomponents as arising from two different starforming regions in the host, but highresolution imaging is necessary to confirm this.

123

124 The host galaxy of GRB 060218

7.1 Introduction

Long-duration gamma-ray bursts (GRBs) are widely accepted to be related to core-collapse supernovae: clear supernova signatures are seen in the afterglow spectraof low redshift GRBs (e.g. Stanek et al. 2003; Hjorth et al. 2003b; Pian et al. 2006).Dedicated surveys of GRB hosts suggest that GRBs occur preferentially in lowmass, subluminous, blue star-forming galaxies (e.g. Chary et al. 2002; Fruchteret al. 1999; Le Floc’h et al. 2003). The GRBs are often located within UV-brightparts of their hosts, where star formation takes place (Bloom et al. 2002; Fruchteret al. 2006), and are shown to be more concentrated towards the brightest regionsof their hosts than are, in general, core-collapse supernovae (Fruchter et al. 2006).

GRB host galaxies are not selected through their luminosities or colours, butmerely by the fact that a GRB has been detected. This could potentially providean unbiased sample of starforming galaxies, which may be used to study star for-mation in the Universe (see e.g. Jakobsson et al. 2005b). The significant increasein detection rate and localisation of long GRBs through the successful operationof Swift in principle permits the study of a large and uniformly selected sample ofGRB host galaxies. However, to date the sample of spectroscopically studied GRBhost galaxies is small (∼30 cases) and may suffer from several selection biases, dueto their faintness.

One of the key properties needed to understand GRB progenitors and their en-vironments is the metallicity distribution (e.g. Langer & Norman 2006; Yoon et al.2006): the popular collapsar model for long GRBs requires low metallicity pro-genitors. Metallicities for GRB hosts can be determined through afterglow spec-troscopy and through host galaxy spectroscopy. Absorption lines of H I and heavymetals have provided metallicities along GRB sight-lines in the redshift interval2 < z < 6.2.

Host galaxy spectroscopy can provide metallicities for the galaxy as a whole,but relies on the detection of the nebular emission lines, which is difficult atz � 1. There are only a few GRB host galaxies that are bright enough to per-mit a direct abundance analysis through an electron temperature (Te) determination(GRBs 980425, 020903, 031203 Prochaska et al. 2004; Hammer et al. 2006). It isonly possible to study the metallicities of higher redshift and fainter hosts throughsecondary metallicity indicators using bright (nebular) lines (e.g. the R23 method),using empirical correlations between the fluxes of certain emission lines and metal-licity.

On the 18th of February 2006 a bright, nearby GRB was discovered by Swift.The proximity (z ∼ 0.0334 Mirabal & Halpern 2006, the second closest GRB) ofthis GRB triggered a large follow-up campaign by several groups, which provideda unique opportunity to unveil the nature of a faint nearby galaxy associated witha GRB event. GRB 060218 was found to be unusually long in duration (T90 =

2100 ± 100 s Campana et al. 2006a) and of relatively low luminosity (Eγ,iso =

6.2 ± 0.3 × 1049 erg Soderberg et al. 2006b). Its prompt emission was soft, placing

7.2 Observations 125

it in the class of X-ray flashes (XRFs). A bright supernova (designated SN 2006aj)was clearly associated with this event which was studied with very high spectraland time resolution over a wide range of wavelengths, from X-rays to radio (e.g.Campana et al. 2006a; Pian et al. 2006; Mazzali et al. 2006; Soderberg et al. 2006b).The host was found to be a small galaxy with an irregular morphology.

In this paper we study the host of GRB 060218 through a high resolution VLTUltraviolet and Visual Echelle Spectrograph (UVES Dekker et al. 2000) spectrum,taken around the peak magnitude of the supernova. This spectrum shows a varietyof well resolved emission lines, associated with ionized gas in the starforming re-gions in the host. The neutral gas of the ISM is probed by the detection of a fewnarrow absorption lines. Low resolution FORS spectra (described in Pian et al.2006) are used to study the fluxes of emission lines that fall outside the spectralcoverage of the UVES spectrum.

Several papers have already been published on GRB 060218, its associated su-pernova, and the host. However, on the metallicity of the host there is a wide rangeof reported values. Modjaz et al. (2006) report a metallicity of 0.15 Z�, Mira-bal et al. (2006) derive 0.46 Z�, while Sollerman et al. (2006) mention that theabundance is below Solar, but that the exact value is unconstrained from the strongemission lines. These values are derived from different spectra, but the spread islargely due to internal scatter in empirical, secondary calibrators that are used, andthe limited range of metallicities and diagnostic emission line ratios for which thesesecondary calibrators are valid (see e.g. Ellison & Kewley 2006). Since the derivedmetallicities are frequently used to draw conclusions on important issues such asprogenitors (e.g. Sollerman et al. 2005b) or the rate of nearby GRBs (Stanek et al.2006; Wolf & Podsiadlowski 2007), it is clearly of great interest to pin down theseuncertainties.

The paper is organized as follows: In Section 2 we describe the observations.In Section 3 we calculate the metallicity of the dominant starforming region(s) inthe host galaxy and relative abundances of Ne and N, as well as the star formationrate from optical and radio fluxes. In Section 4 we analyse the discrete velocitycomponents in the host through emission and absorption lines. We discuss theuse of metallicity values found from secondary metallicity calibrators in Section5. In Section 6 we discuss the implications that a future detection of Wolf-Rayetstar signatures in the host of GRB 060218 may have on single star GRB progenitormodels.

Throughout this paper we use the cosmological parameters H0 = 70 kms−1/Mpc, ΩM = 0.3 and ΩΛ = 0.7.

7.2 Observations

GRB 060218/SN 2006aj was observed with ESO VLT Kueyen (UT 2) on March 4,2006, roughly at the time of maximum light of the supernova (around March 1,


e.g. Sollerman et al. 2006). The UVES observation started at 00:30 UT, for a totalexposure time of 2100 seconds. The magnitude of SN 2006aj was V ∼ 17.6 at thetime of observation. The airmass was high, averaging 2.5, and the seeing measuredfrom the 2D spectrum is 1.1 arcsec. The UVES-setup spectral resolution is R ∼46000 (FWHM � 6.5 km s−1). The slit width was set at 1 arcsec, corresponding toabout 7 kpc physical size at the distance of the GRB. The exposure was performedat the parallactic angle. The spectrum (wavelength ranges 3285 - 4527 Å, 4621 -5598 Å and 5676 - 6651Å) was reduced in the standard fashion using MIDAS andIRAF routines.

The L.A.Cosmic program (specifically the lacos spec routine Van Dokkum2001) was used to remove both point- and irregular shaped cosmic ray hits fromthe 2D spectrum before extraction. The spectrum was dispersion corrected, andflux calibrated by archived response functions. An air to vacuum conversion anda heliocentric correction (+28 km s−1) were applied to the spectrum. The result-ing spectrum was compared with a (quasi)-simultaneous FORS spectrum taken 20minutes after the UVES spectrum, which is flux calibrated through a standard starobservation and through simultaneous photometry in B,V,R at VLT FORS2 (Pianet al. 2006), using the magnitude to flux conversions from Fukugita et al. (1995).To the UVES spectrum a multiplication factor of 1.625 was applied to match thewell calibrated FORS flux values (compensating for the slit loss between the FORSand UVES spectra).

We find a good match between the UVES and FORS spectra and photometry inthe red end (above λ ∼4500 Å), as shown in Figure 7.1. In the blue end (λ � 4500Å) the UVES continuum flux is slightly higher than the FORS continuum flux.The prime reason is likely the very high airmass at which both spectra have beentaken, making the flux calibration of both the UVES and FORS spectrum at theblue end more uncertain. We decide to not alter the flux calibration, but warn thatthe fluxes of emission lines below ∼4500 Å have a small additional uncertainty.This uncertainty does not significantly affect the metallicity results of this paper,because the electron temperature uncertainty is dominated by the uncertainty in theflux of [O III] λ4364 from the FORS spectra. The host galaxy emission line fluxratios found in the UVES spectrum agree, within the errors, with those found fromthe combination of the FORS spectra.

A Galactic extinction correction was performed using E(B − V)MW = 0.142(Schlegel et al. 1998), assuming a Galactic extinction law Aλ/AV expressed as RV =

AV/E(B−V) (Cardelli et al. 1989), and RV = 3.1. This value is slightly higher thanthe extinction derived by Guenther et al. (2006), who used the Galactic sodiumlines to find E(B−V)MW = 0.127. Given that the systematic error in the conversionof the equivalent widths of Na to E(B − V) is poorly known, we choose to useE(B − V)MW = 0.142 mag.

A selection of lines from the spectrum and the associated error spectra areshown in Figures 7.2 and 7.5. Although the spectrum is dominated by the SN spec-trum (V ≈ 17.6 vs the host V ≈ 20.2), several bright emission lines from the host

7.2 Observations 127

Figure 7.1 — The UVES spectrum (black), with overlaid the low resolution FORS spec-trum. For presentation purposes the UVES spectrum has been smoothed to the pixelscaleof the FORS spectrum. The points denote B,V,R VLT FORS photometry from the samenight (Pian et al. 2006). The widths of the broadband filters are not shown.

galaxy are detected and resolved in the UVES spectrum (Fig. 7.2, Table 7.3). On theother hand, the detected absorption lines, from both the host galaxy and Milky Way,are narrower than the UVES resolving power. The highest signal-to-noise emissionlines provide a heliocentric mean redshift z = 0.03342(2) (Pian et al. 2006), wherewe adopt a conservative error on the redshift due to poorly known systematic ef-fects of the spectrograph (e.g. the centering of the object in the slit). The analysisof the emission lines was done using the IRAF software packages, mainly usingthe splot routines. As the emission lines are clearly asymmetric (see Section 4.1),fluxes are measured using the numerical integration method (e in splot), and not thestandard Gaussian fitting. The errors in the line fluxes are generally dominated bythe uncertainty in the continuum level and are given at the 1σ level.


Figure 7.2 — Emission and absorption lines detected in the UVES spectrum that wereused to derive the host properties. The error spectra are plotted with dotted lines. Thevertical dashed lines denote the mean redshift of the host. The bottom right panel showsthe [O III] λ4364 line from the FORS spectrum.

7.3 General host galaxy properties

7.3.1 Metallicity

Reddening

Before we can use the emission lines in the FORS and UVES spectra to derive thehost properties, we need to correct for the dust extinction intrinsic to the host. Weuse the Balmer line fluxes to measure a value for the Balmer decrement, assumingcase B recombination (e.g. Osterbrock 1989; Izotov et al. 1994). The detection ofseveral of the Balmer lines in the UVES spectrum, see Table 7.3, provides a goodconstraint on the extinction value. We assume intrinsic Balmer line ratios at 20000K, and compute flux ratios of the detected Balmer lines. We caution that Hδ andHε are located in the blue end of the UVES spectrum, where we see a slight dis-crepancy in flux calibration between the FORS spectrum of March 4 and the UVESspectrum (the UVES spectrum having a slightly higher mean continuum flux). Wefind that the Balmer line ratios from the UVES spectrum are all (Hβ/Hγ to Hβ/Hε)

7.3 General host galaxy properties 129

consistent with the theoretical case B recombination values. The FORS spectra aretaken at low resolution (R ∼ 300), making the extinction derivation based on theHα/Hβ ratio from these data less reliable. We test various intrinsic stellar Balmerabsorption strengths, but find no evidence for internal reddening from the Balmerdecrement in the UVES spectrum, with upper limit E(B − V) � 0.03. Howeverthis value may be influenced by the uncertain flux calibration of the blue end ofthe spectrum, see section 7.2. The Hβ and Hγ lines are located in an area of thespectrum where the FORS and UVES continuum fluxes are more in agreement.The ratio of the Hβ and Hγ lines is consistent with case B recombination values,E(B − V) � 0.04 (1 sigma). In Guenther et al. (2006) a reddening of E(B − V)= 0.042 ± 0.003 is derived from the strengths of the sodium absorption lines inthe host, and used by Pian et al. (2006) to account for reddening in the host. Wechoose to use the more common method of accounting for extinction by using theBalmer decrement. The resulting net extinction (Galactic + host) is close to thevalue applied by Pian et al. (see also Sollerman et al. 2006 for a discussion) anddoes not affect further analysis, as the uncertainties in the line ratios are dominatedby the uncertainties in their fluxes. We further note that the high electron tempera-ture seen in this source (see section 7.3.1) can make collisional excitation of neutralhydrogen important, which can mimic reddening and affect the Balmer decrementderived reddening. We can not reliably evaluate this effect, as we do not measure asignificant Balmer decrement.

We use various nebular line flux ratios to evaluate the possibility of an ActiveGalactic Nucleus (AGN) contribution to the excitation of the nebular lines. Kauff-mann et al. (2003) refine the popular line ratio diagnostic [O III] λ5008 /Hβ vs[N II] /Hα, by analyzing a large sample of galaxies from the SDSS. They empir-ically define the demarcation between starburst galaxies and AGN as follows: agalaxy is AGN dominated if

log([O III]λ5008/Hβ) > 0.61/(log([N II]/Hα) − 0.05) + 1.3. (7.1)

Kauffmann et al. (2003) use extinction corrected fluxes, but note that these line ra-tios are relatively insensitive to extinction effects. We use the flux values measuredfrom the FORS spectra, [O III] λ5008 /Hβ = 4.55 ± 0.28 and [N II] /Hα = 0.06 ±0.01. These flux ratios place the host galaxy comfortably in the locus of activelystarforming galaxies. Hence we assume that the host galaxy nebular emission lineexcitation is not dominated by (non-thermal) AGN emission.

Electron density and temperature

We make the standard two zone model assumption for the H II region(s) from whichthe emission lines originate, consisting of a low and intermediate temperature re-gion (e.g. Osterbrock 1989). The [O II] λλ3726 and 3729 Å lines (Figure 7.2) areclosely linked to collisional excitation and de-excitation at the typical temperatures


of star forming regions (∼ 104 K). The electron density (ne) in the low temperatureregion can be derived from the ratio of the fluxes of these two lines (Osterbrock1989). The ratio of the line fluxes λ3726 / λ3729 approaches 4

6 in the ne → 0limit. In the limit ne → ∞ the flux ratio approaches 2.86 (collisional excitation andde-excitation dominate over radiative transitions, forming a Boltzmann populationratio). As shown in Fig. 2, in the UVES spectrum the [O II] doublet is clearly re-solved, and its flux ratio is 0.62 ± 0.05. This number is consistent with the lowdensity limit, implying that collisional de-excitation is not important for the fluxesof the forbidden lines. The resolved [O II] doublet has also been observed in twoother GRB host galaxy spectra, GRBs 990506 and 000418, which have values of[O II] λ3726 /λ3729 of 0.57 ± 0.14 and 0.75 ± 0.11, respectively (Bloom et al.2003). These two detections imply low values for ne similar to GRB 060218.

A different doublet used frequently as a density diagnostic is the [S II] λλ6717,6731 doublet. The line ratio [S II] λ6717 / λ6731 approaches 1.5 when ne → 0,and 0.44 above ne ∼ 104 cm−3. These lines are redshifted out of the UVES wave-length range, but are detected in the FORS spectra. We find [S II] λ6717 / λ6731 =1.0 ± 0.6, which is therefore not useful to discriminate between the high and lowdensity regimes. Prochaska et al. (2004) find a ratio of 1.19 ± 0.09 for the host ofGRB 031203, corresponding to ne ∼ 300 cm−3. We assume the relatively low val-ues of ne = 100 cm−3 for our analysis, following e.g. Skillman et al. (1994); Izotovet al. (2006a).

We estimate the electron temperature (Te) in the intermediate temperature re-gion from the ratio of [O III] nebular and auroral fluxes. The auroral [O III] λ4364is not significantly detected in the UVES spectrum - it is located in a noisy regionclose to the gap in between the wavelength ranges, see Section 2.

From the FORS spectrum we find the flux ratio [O III] (λ4959 + λ5008) /λ4364= 29.1 ± 6.3. The relatively large error is mainly due to the rather large uncertaintyin the [O III] λ4364 flux value. We use the electron density assumed above, whichwe take to be the same in both the low- and intermediate temperature regions (seee.g. Osterbrock 1989), and find an electron temperature Te(O2+) in the intermedi-ate temperature region of 2.48 +0.5

−0.3 ×104 K. This is high when compared to that ofthe host of GRB 031203 (Te(O2+) ∼ 13400 K (Prochaska et al. 2004), as shown inTable 1. Comparably high values of Te(O2+) have been observed in the recent dis-covery of two extremely low metallicity galaxies by Izotov et al. (2006a). This veryhigh temperature and low density suggests a low oxygen abundance for the host ofGRB 060218, since the main nebular cooling is done through oxygen forbidden lineemission.

Especially at low metallicity there can be large differences in electron temper-ature between the low and high temperature zones. Due to a lack of detected linesthat can be used as temperature indicators in the low temperature region (the [O II]λλ7320, 7331 are redshifted out of the UVES coverage and are too faint for the


FORS spectrum), we follow the recipe by Izotov et al. (2006b),

Te(O+)= −0.577 + Te

(O2+)×(2.065 − 0.498Te

(O2+)), (7.2)

where Te is in units of 104K. We find Te(O+) = 1.5+0.1−0.2 × 104 K, using the ap-

proximation for a low metallicity environment (12 + log O/H ∼ 7.2, see Izotovet al. 2006b). When we assume an intermediate metallicity (12 + log O/H ∼7.6), we find Te(O+) = 1.3+0.2

−0.5 × 104 K. We adopt the low metallicity value fornow, and note that the prescription by Pagel et al. (1992) gives a similar value ofTe(O+) = 1.66+0.1

−0.08 × 104 K. The conversion above from Te

(O2+)

to Te (O+) isderived through sequences of photoionization models (Stasinska & Izotov 2003).Izotov et al. (2006b) show by comparing this conversion with direct measurementsof Te

(O2+)

and Te (O+) that the models agree with the measurements. There is asignificant scatter of direct measurements of Te(O+) and Te(O2+) with respect tothe model predictions, which is probably mainly due to the large uncertainties inthe measurements of Te(O+) (Izotov et al. 2006b). We do not include an additionaluncertainty in the following analysis to account for this, but note that the small er-rors reported here should not be overinterpreted. We assume the Te(O+) to also bevalid for N II and S II, which have comparable ionisation potentials.

Oxygen abundance

We use the equations from Izotov et al. (2006b) to derive ionic abundances from thederived electron densities and temperatures. To find the oxygen abundance, O/H,we sum the O2+ /H+ and O+ /H+ ion abundances, assuming the O3+ abundance isnegligible since no high temperature lines (e.g. He II) were detected. We take theline fluxes from the UVES spectrum, and find O2+ /H+ = 1.72 ± 0.45 × 10−5 andO+ /H+ = 1.78+1.2

−0.32 × 10−5. The errors include the uncertainties on the line flux ra-tios and electron temperature. The ratio O+ / (O+ + O2+) > 0.1, confirming that theexcitation is too low for O3+ to be important, as shown by photoionization modelsby Izotov et al. (2006b). The large uncertainty in the O+ /H+ abundance is causedby the relatively large error in the Te(O+). The total oxygen abundance is O /H =O2+ /H+ + O+ /H+ = 3.50 +1.65

−0.77 × 10−5, or 12 + log(O/H) = 7.54 +0.16−0.1 , or � 0.07Z�,

assuming log(O/H)� + 12 = 8.69 (Allende Prieto et al. 2001). This value places thehost of GRB 060218 among the most metal deficient subset of emission line galax-ies in the local universe (e.g. Izotov et al. 2006b; Lee et al. 2006b). It is also thelowest metallicity found so far for a GRB host from emission line analysis (absorp-tion line metallicities from several afterglows show lower line of sight metallicities,down to ∼0.01 Z� for GRB 050730 Starling et al. 2005a; Chen et al. 2005a). Wenote that in an independent analysis Kewley et al. (2007) found a similar value forthe metallicity of the host of GRB 060218 of 12 + log(O/H) ∼ 7.6.


7.3.2 Relative element abundances

Accurate emission-line abundances have been derived for a small sample of GRBhost galaxies. A notable example is the spectrum of the host of GRB 031203, forwhich a solar abundance pattern was established (Prochaska et al. 2004). We usethe host emission lines measured in the UVES and FORS spectra to gain an insightinto the abundance pattern in the host of GRB 060218.

The detection of the forbidden [Ne III] lines allows us to derive a Ne abun-dance, using the values for ne and Te(O2+) found in Section 3.1. We use the[Ne III] λ3869 /Hβ flux ratio from the UVES spectrum and find Ne2+ /H+ =2.9 ± 1.2 × 10−6, and Ne2+ /O2+ = 0.17 ± 0.11, which is consistent with other lowmetallicity H II regions, e.g. in I Zw 18 (where Ne2+ /O2+ ∼ 0.13, see Table 7.1).To derive the Ne/H abundance we need an ionisation correction function (ICF) forwhich we use the parametrization by Izotov et al. (2006b). We find ICF (Ne2+) =1.11 and Ne /H = 3.3 ± 1.3 ×10−6.

The [S III] λ6312 line is not detected in the UVES or FORS spectra and [S III]λλ9532, 9069 are redshifted out of both the UVES and FORS coverage. We willtherefore only derive a value for the ionic S+ abundance, and give an upper limit forthe total sulphur abundance. We use Te(O+) and find S+ /H+ = 2.7+2.0

−1.0 × 10−7. Forthe limit on S2+ /H+ we transform Te(O2+) to Te(S2+) through the recipe of Izotovet al. (2006b), and use the upper limit on the [S III] flux from the UVES spectrumto find S2+ /H+ < 2.2 × 10−6. We calculate an ICF(S++ S2+) of ∼1 which allowsus to set the not particularly constraining limit log(S/O) < −1.1 (Solar value islog(S/O) = −1.50 Lodders 2003), which is consistent with the observed trend for Sto follow Solar (S/O) ratios independent of (O/H) for low metallicity H II regions.

The [N II] λ6584 line is redshifted out of the UVES range, but is detected inthe FORS spectrum. The weaker [N II] λ6548 line is not significantly detected inthe FORS spectrum, with a 3σ upper limit on the flux of ∼ 8× 10−17 erg s−1 cm−2.We use the fixed flux ratio λ6584 / λ6548 = 2.9 (Osterbrock 1989) and Te(O+), andfind N+ /H+ = 1.4+0.8

−0.4 × 10−6.To calculate N/H from N+ /H+ we correct for ionisation using ICF (N+) = 1.98,

and find N/H = 2.8 +1.6−0.8 × 10−6, see Figure 7.3. The ratio N+ /O+ = 0.08 +0.07

−0.05is comparable to, though slightly above, the ratio for I Zw 18 of N+ /O+ ∼ 0.03.We note that here we compare the abundances of two elements using two differentspectra (UVES and FORS) taken under different conditions, and the uncertainty onlog(N/O) is likely underestimated. Nevertheless the N/O ratio does not significantlydeviate from the observed trend of low metallicity galaxies to have log(N/O) ∼ −1.5(e.g. Lopez & Ellison 2003; Izotov et al. 2006b).

Hammer et al. (2006) observed the host of GRB 980425 / SN 1998bw withsignificant spatial resolution (owing to the large spatial extent of this host galaxy),and found a high ratio log(N/O) = −0.6 from a Te abundance analysis at the regionat which the GRB / SN took place, which corresponds to almost twice the Solarvalue. This is unexpected at the measured metallicity, see Figure 7.3. Prochaska


Figure 7.3 — Measurements of log(N/O) and 12 + log(O/H) from a sample of galax-ies from the SDSS-DR3 (Izotov et al. 2006b) are shown with open squares, togetherwith points from Lopez & Ellison (2003) (small points), for which typical uncertaintiesare indicated in the bottom right corner. The blue dashed lines show the tracks whereprimary and secondary nitrogen dominate the N/O ratio. Published GRB host galaxymeasurements are overplotted: GRB 060218 (this work), GRB 031203 (Prochaska et al.2004), GRB 980425 (the dashed line connects the value for the SN region with thehigher log(N/O) and the general host value, see Section 3.2, Hammer et al. 2006) andGRB 020903 (Hammer et al. 2006).

et al. (2004) find a similarly high value of log(N/O) = −0.74 ± 0.2 dex in theirspectrum of the host of GRB 031203, at a metallicity of log(O/H) + 12 = 8.1.GRB 020903 shows a value more in line with expectation from its metallicity (seeFig.3). In the case of GRB 060218 the uncertainty on the (N/O) ratio is too highto exclude a deviation from the expected value at the metallicity of the host. Apossible high N/O value can be explained by a variety of reasons. If Te has beenhighly overestimated the metallicity would decrease, moving the points in Fig. 7.3to the left. Physical reasons for an enhanced N/O ratio may be e.g. a contributionof shock heating to the line emission or a chemical evolution effect: Hammer et al.(2006) explain the higher N/O ratio at the locus of GRB 980425 by a larger N yieldof a GRB progenitor or SN remnants.

The hosts of GRBs 980425, 031203, 020903 and 060218 form a sequence in


Property GRB 060218 GRB 031203 I Zw 18 NW I Zw18 SETe(O III) 2.48 +0.5

−0.3 × 104 K 13400 ± 2000 19780 ± 640 19060 ± 610Te(O II) 1.5+0.1

−0.2 × 104 K 12900 15620 ± 470 15400 ± 460O2+/H+ 1.72 ± 0.45 × 10−5 - 1.216 ± 0.09 × 10−5 1.106 ± 0.082 × 10−5

O+/H+ 1.78+1.2−0.32 × 10−5 - 0.179 ± 0.014 × 10−5 0.403 ± 0.031 × 10−5

O/H 3.50 +1.65−0.77 × 10−5 - 1.465 ± 0.092 × 10−5 1.523 ± 0.088 × 10−5

12 + log (O/H) 7.54 +0.16−0.1 8.02 ± 0.15 7.166 ± 0.027 7.183 ± 0.025

N+/H+ 1.4+0.8−0.4 × 10−6 - - 1.074 ± 0.084 × 10−7

ICF (N) 1.98 - - 3.78log (N/O) −1.1 ± 0.4 -0.74 ± 0.2 - −1.574 ± 0.06Ne2+/H+ 2.9 ± 1.2 × 10−6 - 1.91 ± 0.16 × 10−6 1.83 ± 0.15 × 10−6

ICF (Ne) 1.11 - 1.20 1.28log (Ne/O) −0.8+0.2

−0.4 -0.85 ± 0.2 −0.803 ± 0.053 −0.781 ± 0.051

Table 7.1 — Table of properties of the hosts of GRB 060218 (this work) andGRB 031203 (Prochaska et al. 2004) and a comparison to the northwest and southeastregions of I Zw 18 (Izotov et al. 1999).

metallicity, from the metallicities where we expect both primary and secondarynitrogen production to play an active role (the host of GRB 980425) to where pri-mary N is expected to dominate (as in the host of GRB 060218). This makes theGRB 060218 host an interesting candidate for deep spectroscopy to obtain a moreaccurate N/O when the SN has fully faded.

The measured nitrogen, oxygen and neon abundances derived for the hostgalaxy of GRB 060218 are shown in Table 7.1. We note that these are not spa-tially resolved. Izotov et al. (1999) noted that in the case of I Zw 18 a gradientin electron temperature can be seen, with the highest temperatures in the regionswhere WR stars are found. These differences in temperature are associated withsignificant differences in (oxygen) abundance (with factors up to ∼1.4 Izotov et al.1999). This gradient may be due to oxygen enrichment by starforming clusters,and incomplete mixing in the galaxy. Without spatial information we can not checkabundance gradients in most GRB hosts, and assume the oxygen abundances foundare representative for the galaxy as a whole, including the progenitor locus. How-ever, Hammer et al. (2006) and Sollerman et al. (2005b) have shown that in the caseof the host of SN 1998bw strong differences in Te and abundances are observed aswell.

7.3.3 Star formation

The detection of the bright Hα λ6563 emission line in the FORS spectrum allowsus to accurately measure star formation in the host: the Hα line luminosity onlyweakly depends on the physical conditions of the ionized gas. We use

SFRHα = 7.9 × 10−42LHα, (7.3)

as found by Kennicutt (1998). Moustakas et al. (2006) assess the accuracy of thisexpression through a direct comparison of extinction corrected LHα SFR values and


the SFR derived from LIR measurements of a sample of IRAS detected galaxies.They find that when the Hα flux is extinction corrected, the IR and Hα SFRsagree without systematic offset with a precision of ∼70%. The extinction correctedflux of Hα can be found from the FORS spectra, yielding SFRHα = 0.065 ± 0.005M� yr−1. Due to slit losses the true Hα flux is likely to be higher, and the SFRHα

can be interpreted as a lower limit.

Radio and submillimetre observations do not suffer from dust extinction. Theradio continuum flux of a normal galaxy (i.e. non-AGN hosting) is thought to beformed by synchrotron emission by accelerated electrons in supernova remnantsand by free-free emission from H II regions (Condon 1992). It is expected that theradio continuum flux is a particularly good tracer of the recent SFR, due to theshort expected lifetime of the supernova remnants, which is � 108 yr. We use themethod described by Vreeswijk et al. (2001) and Berger et al. (2003a) to calculatean upper limit to the full star formation rate (ie not influenced by any form ofdust extinction). We use the deepest 6 cm (4.9 GHz) Westerbork Synthesis RadioTelescope (WSRT) flux limit, i.e. when the initial radio afterglow has faded beyonddetection limit, and find a 3σ limit of 72 μJy (formal flux measurement 8 ± 24 μJyKaneko et al. 2007) at a 12h full synthesis on April 1 2006. This leads to a 3σ SFRupper limit of SFRradio < 0.15 M� yr−1, which excludes a large amount of obscuredstar formation when compared to SFRHα. However, we note that despite manysimilarities, GRB hosts studied to date are a diverse population: a handful of hostsshow clear indications of much higher star formation rates than seen from opticalindicators (comparable to the submillimetre galaxies at several hundred M� yr−1)from their submm fluxes (Berger et al. 2003a; Barnard et al. 2003) and in onecase (GRB 980703 at z = 0.97 Berger et al. 2003a) its radio flux. However, theradio derived SFRs may systematically overestimate star formation through a non-negligible contribution from AGN activity to the total radio continuum flux.

One of the properties that sets GRB hosts apart from field galaxies is the spe-cific star formation rate (SSFR): the star-formation rate per unit mass. This quantityis shown to be high in GRB hosts (e.g. Christensen et al. 2004; Courty et al. 2004).Sollerman et al. (2006) find that the host of GRB 060218 has L = 0.008L∗B, im-plying a SSFR of ∼ 8 M� yr−1(L/L∗)−1. This value is comparable to the SSFRsfor a sample of GRB hosts determined through SED fitting by Christensen et al.(2004). Given the multi band SDSS photometry of the host, a low stellar mass isexpected. After modeling the SED of the host, Savaglio, Glazebrook & Le Borgne(in preparation) find M∗ = 107.2±0.3 M�. The measured metallicity and stellar masstherefore place this galaxy on the mass-metallicity relation found recently in localdwarf galaxies by Lee et al. (2006a).


7.4 Discrete velocity components in emission and ab-sorption

7.4.1 Emission line profiles

The resolution of the UVES spectrum is sufficient to search for structure in theemission line profiles. Figure 7.4 shows that the brightest emission lines of [O III]λλ4959, 5008 significantly deviate from a single Gaussian line profile, and areskewed towards the blue. We can rule out an instrumental effect as no such ef-fect was seen in the arc or sky line profiles. Due to a lower signal to noise, weare not able to verify quantitatively whether other emission lines share the sameline profile. We fit Gaussian components to the lines, where the number of com-ponents and the position and width of the components are free parameters. The[O III] λλ4959, 5008 lines are fit simultaneously with a fixed flux ratio between theλ4959 and λ5008 components of 1:3. We use both a fit by eye with IRAF splotand a quantitative deblending using VPFIT1. Results agree, and we find a good fit(reduced χ2 ∼ 0.72) using two Gaussian components, shown in Fig. 7.4, where thecomponents are bluewards and redwards of the average host galaxy redshift. Forthe red component we find a FWHM of 49 ± 5 km s−1 and redshift 0.033453 ±0.000019; and for the blue component a FWHM of 35 ± 3 km s−1 and redshift0.033379 ± 0.000005. This gives Δz = 7.4 ±2.4 × 10−5, or ∼21.6 km s−1 velocityseparation. These values are similar to those found from spatially resolved highresolution spectroscopy of emission line regions (i.e. the 30 Doradus nebula, seeMelnick et al. 1999 for a Hα study). Arsenault & Roy (1986) show that ≈ 66% ofall giant extragalactic H II regions show symmetrical Hα lines in their integratedspectra; the remaining profiles show asymmetries similar to the ones in the host ofGRB 060218. When asymmetric line profiles are observed (i.e. in high resolutionstudies of galactic and extragalactic H II regions) the explanation of their profilesis usually only possible by using a high spatial resolution and correlation of thespectra with images. As an example, the integrated Hα line profile of 30 Doradusshows a broad and narrow Gaussian component (Melnick et al. 1999), whose cen-tral wavelengths coincide, while spectra taken at multiple positions in the nebulashow a large variety of line profiles and Gaussian components, which generallycan be associated with filaments in the nebula. In the case of this host galaxy, weobserve the integrated profile over 7 kpc.

There are several possible explanations for the occurence of two lines, with thebroader one redshifted with respect to the blue component. One possibility is thatwe see two separate star forming regions in the host. An alternative explanationis that the components are caused by an expanding shell (bubble) around the star-forming region; or by infalling gas onto the H II region (e.g. ionised gas ejectedby perhaps SN shocks or stellar winds, that falls back). Neither of the last two

1See http://www.ast.cam.ac.uk/∼rfc/vpfit.html

7.4 Discrete velocity components in emission and absorption 137

scenarios seem plausible: an expanding shell is not likely to have the measured ve-locity width; and the infalling gas would have to be very highly excited to generatethe required [O III] luminosity, which makes it difficult to have co-exisiting Na I(see Section 4.2). We tentatively interpret the two components as arising from twodifferent starforming regions in the host. High resolution imaging would be neces-sary to confirm this. There are several GRB hosts where HST images resolve thehost into multiple starforming regions with similar brightnesses (see e.g. Fruchteret al. 2006). We note that this is the first identification of resolved emission linecomponents in a GRB host galaxy spectrum.

Figure 7.4 — The [O III] λ4959 (top) and λ5008 (bottom) emission lines. The con-tinuum is normalized. The error spectrum is shown by a dotted line. The lines aresimultaneously fit using fixed flux ratios between λ4959 and λ5008 components of 1:3.The two seperate Gaussian components are shown with dash-dotted lines, and the totalemission with a dashed line.

7.4.2 Absorption lines in the circumburst medium

In the UVES spectrum we detect absorption lines at ∼ 4070 Å, ∼ 4100 Å and∼ 6090 Å, associated with Ca II λλ3934, 3969 and Na I λλ5891, 5897 in the fore-ground gas of the GRB /SN, and within the host galaxy. Figure 7.5 shows at leasttwo discrete velocity components, separated in velocity by ∼24 km s−1 (systems 1and 2 in Table 7.2; see also Guenther et al. 2006). The Ti II absorption lines are notdetected because they are in a very noisy region of the spectrum (λ < 3500 Å).

The observed lines have been fitted with Voigt profiles using the MIDAS pack-age FITLYMAN (Fontana & Ballester 1995). Figure 7.5 shows the fitting results,and column densities and Doppler parameters are listed in Table 7.2. The lines are


Table7.2

—A

bsorptionlines

inthe

UV

ES

spectrumof

GR

B060218

SystemL

ineλ

oz

Wo

logN

blog

(Na

I/Ca

II)(Å

)(Å

)[cm

−2]

(kms −

1)1

Na

Iλ5891

6088.230.033378

0.091±0.008

11.79±

0.04

6.3±

1.0

−0.54±

0.08

Na

Iλ5897

6094.410.049±

0.007C

aIIλ3934

4066.050.033363

0.093±0.029

12.33±

0.07

13.3±

3.2

Ca

IIλ3969

4102.030.064±

0.0282

Na

Iλ5891

6088.730.033462

0.071±0.006

12.22±

0.22

1.1±

0.2

0.01±

0.23

Na

Iλ5897

6094.900.064±

0.006C

aIIλ3934

4066.420.033458

0.075±0.025

12.21±

0.08

4.6±

1.8

Ca

IIλ3969

4102.400.054±

0.029G

alaxyN

aIλ5891

5891.750.0000285

0.3312.68±

0.02

6.4±

0.2

0.27±

0.06

Na

Iλ5897

5897.730.25

Ca

IIλ3934

3934.930.0000381

0.1312.41±

0.06

(5)C

aIIλ3969

3969.740.08


Figure 7.5 — The Na I and Ca II absorption lines detected in GRB 060218. The twovelocity components are marked by the dashed lines (more precisely, these mark theposition of the two Na I absorption lines). The smooth line is the result of the best fitVoigt profile (reported in Table 7.2). The bottom panel shows the [OIII] λ5008 emissionline for comparison. The dotted spectrum is the errorspectrum.


Figure 7.6 — Na I and Ca II column densities for the two systems in the host ofGRB 060218 (filled dots) and lines of sight in the Milky Way (open squares; typicalerror for these cases is < 0.1 dex Hunter et al. 2006).

barely resolved, i.e. the line widths are at the level of the UVES spectral resolutionFWHM � 6.5 km s−1.

The measured FWHM corresponds to an instrumental PSF of 3.9 km s−1, ifexpressed in terms of Doppler parameter, which can be translated into an upperlimit on the gas temperature (derived from the lightest element Na, assuming ther-mal broadening) of T � 2 × 104 K. Narrow metal lines are often detected in GRBafterglows when high resolution spectra are acquired (see for instance Chen et al.2006a).

There are indications that Na I and Ca II are not tracing each other. The po-sitions of the two Ca II components are shifted by a few km s−1 with respect toNa I, which could be consistent with uncertainties on the fitted parameters (moresevere for the Ca II doublet). However, the line broadenings and Ca II/Na I ratiosare different in the two components. We note that in the Galactic ISM Ca II andNa I are found in regions with different physical conditions. The former is found tobe generally broader than the latter, indicating that it traces warmer, more turbulent,and/or larger gas clouds (Welty et al. 1996).

Remarkably, the positions of the two absorption systems are consistent withthe redshifts of the two emission-line components in the H II regions derived inde-pendently (see Section 3.5 and the lower panel of Figure 7.5). The difference inredshift between emission and absorption is < 1 km s−1 and ∼ 2 km s−1 for the


blueshifted and redshifted systems, respectively. The relative velocity for the twoemission components is 21 km s−1, close to the 24 km s−1 measured from the ab-sorption systems. However, the broader component in absorption is at the lowestredshift, whereas the opposite is true for the emission.

The Ca II column density was measured in another two GRB afterglows(Savaglio & Fall 2004; Savaglio 2006) with a total column density of nearly 1014

cm−2 in each of them. Na I absorption in GRBs is reported here for the first time,basically due to a lack of suitable data in past GRB observations (Na I is redshiftedinto the NIR for z > 0.7). The Na I and Ca II abundances have been studied inthe Galaxy and LMC (Hunter et al. 2006; Vladilo et al. 1993). Beyond the LocalGroup, Na I and Ca II have been detected in 2 damped Lyman-α systems (DLAs)in QSO spectra, at z = 1.062 and 1.181 (Petitjean et al. 2000; Kondo et al. 2006).Other QSO absorption line studies report upper limits for Na I (Boksenberg et al.1978, 1980). GRB 060218 is the third source outside the Local Group where Na Iis measured 2.

It is rather complicated to interpret the detection of Na I and Ca II inGRB 060218 in terms of relative abundances. The ionization potentials of the twoions are quite different (5.1 eV and 11.9 eV for Na I and Ca II, respectively). Na IIis likely the dominant ion in a neutral-gas environment (the H I ionization poten-tial is 13.6 eV), whilst Ca III dominates the calcium species. To derive the totalabundance of calcium, a significant ionization correction can therefore be neces-sary. Moreover, Ca is an α element, while Na is not. Hence in low metallicitysystems, like GRB 060218, an α-element enhancement is expected. Sodium wasfound to trace iron in stars with metallicities close to Solar, but it can have lowerabundances for lower metallicities (Timmes et al. 1995). Our expectations are fur-ther complicated by the rather different refractory properties of the two elements:Na is little depleted on to dust grains, whereas Ca can be 99% depleted (Welty et al.1994; Savage & Sembach 1996). This problem may be somewhat mitigated by thelikely negligible dust depletion in the gas, as suggested by the small dust extinctionderived in the H II regions of GRB 060218 (see Section 3.1).

Nevertheless, we compare Na I and Ca II in GRB 060218 with what is typicallyobserved in the ISM of the Milky Way (Figure 7.6). The two systems in the hostof GRB 060218 lie in the bottom left corner of the distribution. Is the observed Naand Ca behaving like the neutral gas of the Milky Way? If the absorption lines arearising in a neutral region of the ISM, and if we consider the empirical relation thatlinks H I to Na I (derived by Hunter et al. 2006) we would expect an H I columndensity along the GRB sight line of the order of log NH I = 20.6 cm−2. However,if the metallicity in the neutral gas is similar to the H II regions probed by theemission lines, a much lower total H I column density is more likely: log NH I ∼19.4, suggesting that perhaps the Na I-H I relation found for the Milky Way may not

2More Na I lines are detected in low z SDSS QSO spectra, but no column density measure-ments are reported for these systems (V. Wild, private communication.)


be applicable for the gas around GRB 060218 or that the metallicity in the galaxy isnot uniform. An H I column density of the order of 1019.4 cm−2 is relatively large,considering that the stellar mass of this GRB host is more than 1000 times smallerthan that of the Milky Way. However, a large reservoir of gas is expected given thehigh SFR per unit stellar mass estimated for the host (see section 7.3.3).

7.5 Secondary metallicity calibrators

The association of GRB 060218 with a low mass, low metallicity, high excitationhost galaxy follows the trend seen from the other nearby GRBs (e.g. Sollermanet al. 2005b). In fact, the very low metallicity of the host significantly extendsthe known metallicities of GRB hosts through emission line spectroscopy. If theredshift of the galaxy had been substantially higher, the [O III] λ4364 line wouldhave gone undetected, making a direct determination of the abundance throughTe impossible, and the metallicity inferred from secondary calibrators would haveplaced the host at significantly higher metallicity (e.g. Mirabal et al. 2006; Mod-jaz et al. 2006). The sample of GRB host galaxies with a measure of abundancesthrough Te is limited to just four nearby sources due to the faintness of the [O III]λ4364 line, where GRB 020903 has the highest redshift with z = 0.25. For all othergalaxies we are forced to use secondary, empirical methods to calculate metallic-ity. The most common is the R23 calibrator (see e.g. Kobulnicky & Kewley 2004),which uses the bright [O III], Hβ and [O II] lines. This method produces a de-generate metallicity solution, which can be broken through other emission lines(e.g. [N II] and Hα), but in many cases these other lines are not available (dueto e.g. redshift or insufficient S/N). The R23 method is calibrated through photo-ionization models, which have limitations at the low and high metallicity ends.It has been shown that R23 metallicities have an offset with respect to Te metal-licities for metallicities close to Solar (e.g. Bresolin et al. 2004). Modjaz et al.(2006) use the R23 method to find 12 + log(O/H) = 8.0 ± 0.1 for the host ofGRB 060218. An alternative method that is calibrated on a sample of H II regionswith Te determined metallicities, is the ratio of the nebular [O III] and [N II] lines(e.g. Pettini & Pagel 2004). The N2 index (N2 = log([N II]λ6583/Hα)) has beenproposed particularly for low metallicity galaxies. However, for GRB 060218, theN2 calibration overestimates the metallicity of the host by more than a factor of 3,yielding 12 + log(O/H) ∼ 8.2, but the scatter in this relation is large (90% of thefitted data in Pettini & Pagel (2004) falls within a range log(O/H) = ±0.4). TheO3N2 = log (([O III]λ5008/Hβ)/([N II]λ6583/Hα)) ratio has significantly lessscatter but above O3N2 � 1.9 this indicator breaks down and cannot be reliablyused (Pettini & Pagel 2004). Our host has O3N2 = 1.88 ± 0.11, which may explainthe overestimate of the metallicity through this indicator of 12 + log(O/H) ≈ 8.1.

Concluding, we can state that an analysis using a variety of different secondarymethods (e.g. R23, O3N2) would not have found the true metallicity, but may still

7.6 Massive stars and progenitors 143

have identified this host as a very low metallicity candidate.

7.6 Massive stars and progenitors

Long GRB progenitors are likely massive Wolf-Rayet (WR) stars. We can gain agreater understanding of the evolutionary paths of such massive stars towards GRBsby detecting WR populations within GRB host galaxies. A valuable diagnosticon the massive star population is the He II λ4686 line, which appears as a broadline and/or as a nebular line in some H II galaxies (Schaerer et al. 1999), and isa direct sign of (unusually) high excitation levels caused by the presence of WRstars (especially WC- and WO-type WR stars). Together with, amongst others,N III λ4640, C IV λ4658, [Fe III] λ4658 and [Ar IV] λ4711 lines, this line formsthe so-called blue WR bump (from ∼4650 – 4700 Å) in low resolution spectra,which is often accompanied by a C IV λ5808 line (the red bump). In a recentdeep spectroscopic search in nearby GRB host galaxies, Hammer et al. (2006) havedetected the He II line and accompanying WR bump in the spectra of the hosts ofGRB 980425, 020903 and 031203. From the relative intensities of the H β andHe II λ4686 lines or WR bump (measured flux ratios H β /He II λ4686 generallyrange from ∼0.01-0.02) one can estimate the ratio of WR to O stars.

In Crowther & Hadfield (2006) the effect of metallicity on the derivedWR / (WR + O) ratio is investigated, by comparing the fluxes of SMC, LMC andGalactic WR stars with atmosphere models. They find a lower WR line luminosityat decreasing metallicity. In the hosts of GRBs 980425 and 020903, Hammer et al.(2006) find values of WR /O ∼ 0.05 and ∼ 0.14−0.2, respectively. We note that nometallicity correction has been applied to these values, which would increase thenumber of WR stars as the studied hosts have sub-Galactic metallicity.

These high WR /O ratios are of particular interest as their metallicities areaccurately known from Te analyses: 0.5 Z� and 0.19 Z� for GRB 980425 andGRB 020903, respectively (Hammer et al. 2006). At these low metallicities the highmeasured WR /O ratios may be regarded as evidence for very recent star bursts, butit is difficult to determine the ages of the dominant stellar populations. Optical andnear-infrared SED fitting has shown that the dominant stellar populations in GRBhosts are young at typically ∼0.1 Gyr (Christensen et al. 2004). In the case of thehost of GRB 980425 the entire galaxy is well fitted with a continuous star forma-tion history (Sollerman et al. 2005b), although Hammer et al. (2006) find a youngpopulation from the equivalent widths of the emission lines. No strong evidencefor a very young starburst is apparent in the host spectrum of GRB 020903, butthe merger morphology of the host and the emission line EWs may suggest thatsome recent star formation is present. However, the WR/O ratio of 0.14 – 0.2 inthe host of GRB 020903 is particularly remarkable, as such abundant productionof WR stars at the low metallicity of 0.19 Z� can only be explained by invokinginstant star bursts with peculiar initial mass function within the standard star burst


model by Schaerer & Vacca (1998) (e.g. Fernandes et al. 2004).Recent stellar evolution models indicate that the effects of rotation may be, in

part, responsible for observed high WR /O ratios in galaxies. According to Meynet& Maeder (2005), including the effects of rotation significantly enhances the massloss rates of massive stars during the giant phase compared to the non-rotating case,as the shear instability due to the strong degree of differential rotation betweenthe core and the envelope induces fast chemical mixing. Their models predict aWR /O ratio of about 0.02 at Z = 0.004 for a constant star formation, and in prin-ciple WR /O ∼ 0.2 might be achieved at the given metallicity for instant star burstseven with a standard initial mass function. However, such rotating models predictGRB / SN Ibc ratios that are too high. Their prediction of spin rates of young neu-tron stars is also inconsistent with observations (Hirschi et al. 2005; see also Hegeret al. 2000).

More recent models that include magnetic torques for the angular momentumtransport in the star suggest another way to produce WR stars at low metallicity.Although in magnetic models the chemical mixing induced by the shear instabilityduring the giant phase is negligible as the magnetic torque tends to keep the starrigidly rotating, the mixing by meridional circulation can be very fast even on themain sequence (Maeder & Meynet 2005). This may even cause the whole starto become chemically homogeneous if the initial rotational velocity is sufficientlyhigh, especially at sub-solar metallicity (Yoon & Langer 2005; Woosley & Heger2006).

Formation of WR stars can thus be induced not only by mass loss but also bychemical mixing, and Woosley & Heger (2006) found that some WR stars formedthrough such chemically homogeneous evolution can actually retain enough an-gular momentum necessary for GRB production. The predicted GRB /SN ratiothrough such evolutionary channels also turns out to be consistent with observa-tions, when the observationally derived initial rotational velocity distribution ofmassive stars by Mokiem et al. (2006) is adopted (Yoon et al. 2006). Interestingly,as stars may be kept rotating rapidly at low metallicity due to lowered mass lossrates (e.g. Vink et al. 2001), this new type of evolution by fast rotational mixingcan lead to high WR /O ratios even at very low metallicity (Yoon et al. 2006). Italso predicts rather large delay times for WR production from the star formation(i.e., from several 106 yrs to a few ∼ 107 yrs Yoon et al. 2006) compared to thosefrom mass-loss induced WR formation ( < a few 106 yrs).

In this regard, an estimate of the WR /O ratio in the host galaxy of GRB 060218might provide a valuable test case for the new GRB progenitor scenario by Yoon &Langer (2005) and Woosley & Heger (2006). That is, if the WR /O ratio in this hostturns out to be high despite its very low metallicity, it may support the chemicallyhomogeneous evolution scenario for GRB progenitors and extend the metallicity –WR /O star ratio relation for GRB hosts down to lower metallicity. This relationis well known for the Milky Way, LMC and SMC, and a deviation of GRB hostsaway from that trend gives strong input for progenitor modelling.

7.7 Conclusions 145

The high resolution UVES spectrum of GRB 060218 should be able to resolvethe components comprising the WR bump, and resolve a He II line into a broad anda nebular component. However, no nebular He II line or WR bump has been sig-nificantly detected. We measure the flux upper limit on the WR bump in the UVESspectrum by summing the flux in the WR bump wavelength region (restframe 4650- 4686Å). Following the method of Schaerer & Vacca (1998) we set an upper limitof WR / (WR + O) � 0.4, which is not a constraining limit, owing to the fact thatthe SN outshines the possible He II line. To reliably detect the WR bump in thishost we need to detect the host continuum with reasonable S/N, which means thesupernova has to fade below this level before more sensitive searches are feasible.

7.7 Conclusions

We present a VLT UVES high resolution spectrum of SN 2006aj, associated withthe nearby GRB 060218 at heliocentric redshift z = 0.03342(2). We use the emis-sion lines of the UVES spectrum as well as the line measurements from our FORSspectroscopic campaign to derive properties of the host. We find that the electrondensity is low, and the electron temperature is high, Te(O2+) = 2.48+0.5

−0.3 × 104K,as shown in Table 1. We find a low host metallicity of 12 + log(O/H) = 7.54+0.17

−0.10,placing it among the most metal deficient subset of emission-line galaxies. It is alsothe lowest metallicity found so far for a GRB host from an emission line analysis.The metallicity we find lies considerably below the values derived using secondarycalibrators, e.g. the metallicity 12 + log(O/H) = 8.0 as derived from the R23 cali-brator (e.g. Modjaz et al. 2006). The mass of the galaxy is low, and matches whatis expected from the mass-metallicity relation for dwarf galaxies. We measure arelatively high value for log(N/O) with respect to the metallicity, which is also seenin a few other GRB hosts. As our uncertainty on log(N/O) is relatively high, deeperspectroscopy is needed to confirm this overabundance.

The bright emission lines show strong evidence for asymmetry, and a singleGaussian provides a poor fit to the profiles of the bright [O III] emission lines. Atwo Gaussian model provides a satisfactory fit with the two components separatedby ∼22 km s−1. We find the same two velocity components in absorption throughthe Ca II and Na I absorption lines in the host. We tentatively interpret these twovelocity components to be due to two star forming regions in the host galaxy. How-ever, to unravel their true identity, high spatial resolution imaging is needed.

The dust content of the galaxy is low, based on the Balmer line decrement. Thisis also evident from the low limit on the obscured star formation rate we set througha 3σ upper limit on the flux at 6 cm of SFRradio < 0.15 M� yr−1, compared to theoptical star-formation rate SFRHα = 0.065 ± 0.005 M� yr−1.

This host galaxy is an interesting target for future spectroscopy targeted at theWR bump, as the low metallicity of the host will significantly extend the presentsample of GRB hosts with known WR star content and metallicity. We show that


a measure of these two quantities for a sample of GRB hosts may provide furtherinsight into the nature of GRB progenitors.

The absolute magnitude of the host (MB = −15.9, e.g. Sollerman et al. 2006)is such that this galaxy would not have been detected in any survey at a redshift ofz ∼ 1, let alone at the mean Swift GRB redshift of z ∼ 2.8 (Jakobsson et al. 2006b),which makes this host an important object to study in the context of larger redshiftGRB hosts.

Acknowledgements We thank the observers and Paranal staff for performing thereported observations at ESO VLT. We are very grateful to R.B.C. Henry, H. Lee andN. Tanvir for helpful discussions. We thank the anonymous referee for helpful com-ments. KW thanks NWO for support under grant 639.043.302. The Dark CosmologyCentre is funded by the Danish National Research Foundation. SCY is supported bythe VENI grant (639.041.406) of the Netherlands Organization for Scientific Research(NWO). The authors acknowledge benefits from collaboration within the EU FP5 Re-search Training Network “Gamma-Ray Bursts: An Enigma and a Tool” (HPRN-CT-2002-00294).

Table 7.3 — Table of emission lines used in this paper. Fluxes and EWs are as observed.Fluxes from the FORS spectrum are taken from Pian et al. (2006).

UVES spectrumID EW Flux

(Å) (×10−17 erg s−1 cm−2)[O II] λ3727 1.59 ± 0.09 75.23 ± 3.42[O II] λ3729 2.62 ± 0.12 121.1 ± 4.0[Ne III] λ3869 0.75 ± 0.08 34.06 ± 3.51[Ne III] λ3969 0.26± 0.06 11.80 ± 2.64Hβ λ4862 2.17 ± 0.06 92.92 ± 1.93Hγ λ4341 0.90 ± 0.07 47.99 ± 3.46Hδ λ4102 0.61 ± 0.07 27.83 ± 3.04Hε λ3971 0.29 ± 0.07 13.03 ± 3.02H8 λ3889 0.27 ± 0.07 12.25 ± 2.90H9 λ3835 0.18 ± 0.07 7.65 ± 2.92[O III] λ4960 2.88 ± 0.06 139.0 ± 2.08[O III] λ5008 8.34 ± 0.09 426.2 ± 2.48He I λ5877 0.24 ± 0.04 9.55 ± 1.17

FORS spectrumID Flux

(×10−17 erg s−1 cm−2)[O II] λ3727, 3729 190 ± 50(Doublet unresolved)[Ne III] λ3869 31 ± 5[Ne III] λ3969 18 ± 3Hγ λ4341 37 ± 6[O III] λ4364 21 ± 4Hβ λ4862 101 ± 4[O III] λ4960 150 ± 7[O III] λ5008 460 ± 10He I λ5877 5.5 ± 1.0Hα λ6563 315 ± 25[N II] λ6584 19 ± 2[S II] λ6717 14 ± 2[S II] λ6731 14 ± 2

8GRB 051022: physical parameters and extinction

of a prototype dark burst

E. Rol, A. J. van der Horst, K. Wiersema, S. K. Patel, A. Levan, M. Nysewander,C. Kouveliotou, R. A. M. J. Wijers, N. Tanvir, D. Reichart, A. S. Fruchter,

J. Graham, Jan-Erik Ovaldsen, A. O. Jaunsen, P. Jonker, W. van Ham, J. Hjorth,R. L. C. Starling, P. T. O’Brien, J. P. U. Fynbo, D. N. Burrows, R. Strom

Astrophysical Journal accepted, arXiv:0706.1518

Abstract GRB 051022 was undetected to deep limits in early optical obser-vations, but precise astrometry from radio and X-ray showed that it most likelyoriginated in a galaxy at z ≈ 0.8. We report radio, optical, near infra-red and X-rayobservations of GRB 051022. Using the available X-ray and radio data, we modelthe afterglow and calculate the energetics of the afterglow, finding it to be an orderof magnitude lower than that of the prompt emission. The broad-band modelingalso allows us to precisely define various other physical parameters and the mini-mum required amount of extinction, to explain the absence of an optical afterglow.Our observations suggest a high extinction, at least 2.3 magnitudes in the infrared(J) and at least 5.4 magnitudes in the optical (U) in the host-galaxy restframe. Suchhigh extinctions are unusual for GRBs, and likely indicate a geometry where ourline of sight to the burst passes through a dusty region in the host that is not directlyco-located with the burst itself.

149

150 GRB 051022

8.1 Introduction

Dark gamma-ray bursts (GRBs) — at the most basic level those without opticalafterglows — are a long-standing issue in GRB observations. Although in manycases the non-detection of an afterglow at optical wavelengths may simply be dueto an insufficiently deep search, or one which takes place at late times (e.g. Fynboet al. 2001b), a subset of GRBs with bright X-ray afterglows remains undetecteddespite prompt and deep optical searches (e.g. Groot et al. 1998) and directly im-plies suppression of the optical light.

There are several plausible explanations for this, the most likely being that theburst is at high redshift, such that the Ly-alpha break has crossed the passband inquestion, or that there is high extinction in the direction of the GRB. Examples ofboth have been found, with a small number of GRBs at z > 5 appearing as V andR band dropouts (e.g. Jakobsson et al. 2006b; Haislip et al. 2006) and some GRBafterglows appearing very red at lower redshift, due to effects of extinction (e.g.Levan et al. 2006; Rol et al. 2007).

Identification of GRBs at very high redshifts is the key to using them as cos-mological probes. The proportion of bursts exhibiting high dust extinction is alsointeresting from the point of view of estimating the proportion of star formation thatis dust enshrouded, as well as understanding the environments which favor GRBproduction (Trentham et al. 2002; Tanvir et al. 2004).

The detection and follow-up of dark bursts at other wavelengths is essential,as it enables 1) the modeling of the afterglow, deriving estimates of the extinctionand energies involved, potentially providing information about the direct burst en-vironment, 2) pinpointing the burst position in the host, to enable late-time highresolution imaging and the detection of dust enhanced regions in the host, and 3)determination of the properties of the GRB host itself, such as the SFR and averagehost-galaxy extinction.

The High Energy Transient Explorer 2 mission (HETE-2; Ricker & Vander-spek 2003) detected and located an unusually bright gamma-ray burst (Tanakaet al. 2005) with its three main instruments, the French Gamma Telescope (FRE-GATE), the Wide field X-ray monitor (WXM) and the Soft X-ray Camera, (SXC),on October 22, 2005. A 2.5 arcminute localization was sent out within minutes,enabling prompt follow-up observations (e.g. Torii 2005; Schaefer 2005); a target-of-opportunity observation was also performed with Swift. Details of the HETE-2observations can be found in Nakagawa et al. (2006).

The Swift observations resulted in the detection of a single fading point sourceinside the SXC error region, which was consequently identified as the X-ray after-glow of GRB 051022 (Racusin et al. 2005a). However, optical and near infra-red(nIR) observations failed to reveal any afterglow to deep limits, while radio andmillimeter observations with the Very Large Array (VLA), the Westerbork Synthe-sis Radio Telescope (WSRT) and the Plateau de Bure Interferometer detected theradio counterpart (Cameron & Frail 2005; Van der Horst et al. 2005; Bremer et al.

8.2 Observations and data reduction 151

2005). The position coincides with its likely host galaxy (Berger & Wyatt 2005) ata redshift of z = 0.8 (Gal-Yam et al. 2005).

In this paper, we describe our X-ray, optical, nIR and radio observations ofGRB 051022. The outline of the paper is as follows: in Section 8.2 we describe ourobservations, data reduction and initial results. In Section 8.3, we analyze theseresults and form our afterglow picture, which is discussed in Section 8.4. Ourfindings are summarized in Section 8.5.

In the following, we have used F ∝ ν−βt−α in our definition of α and β. Weassume a cosmology with H0 = 71 kms−1Mpc−1, ΩM = 0.27 and ΩΛ = 0.73. Allquoted errors in this paper are 1 sigma (68%) errors.

8.2 Observations and data reduction

8.2.1 X-ray observations

X-ray observations were performed with the Swift X-Ray Telescope (XRT) and theChandra X-ray Observatory (CXO).

The XRT started observing the afterglow of GRB 051022 3.46 hours after theHETE-2 trigger, for a total effective integration time of 137 ks between October 22and November 6.

Observations were performed in Photon Counting (PC) mode, the most sensi-tive observing mode. We reduced the data using the Swift software version 2.6 inthe HEAsoft package version 6.2.0. Data were obtained from the quick-look siteand processed from level 1 to level 2 FITS files using the xrtpipeline tool in its stan-dard configuration. The first two orbits (until 2.1 × 104 seconds post burst) showpile-up and were therefore extracted with an annular rather than circular region,with an inner radius of 19 and 12′′ for orbits 1 and 2, respectively, and an outerradius of 71′′. Orbits 3 – 7 (2.4 × 104 – 4.9 × 104 seconds) were extracted witha circular region of 71′′ radius, and later orbits were extracted using a 47′′ radiuscircle instead. The data for the light curve were extracted between channels 100and 1000, corresponding to 1 and 10 keV, respectively; while the commonly usedrange is 0.3 – 10 keV, the large absorption prevents the detection of any data fromthe source below 1 keV. Otherwise, the procedure is similar to that described inEvans et al. (2007).

Observations with the CXO started on October 25, 2005, 21:14:20, 3.34 daysafter the HETE trigger, for a total integration time of 20 ks (Patel et al. 2005). Datawere reduced in a standard fashion with the CIAO package.

We performed astrometry by matching X-ray sources with an optical R-bandimage that was astrometrically calibrated to the 2MASS catalog. Our CXO positionis RA, Dec = 23:56:04.115, +19:36:24.04 (J2000), with positional errors of 0.33′′and 0.12′′ for the Right Ascension and Declination, respectively. This puts theafterglow within 0.5′′ of the center of its host galaxy.

152 GRB 051022

We modeled the XRT spectra with an absorbed power law in XSpec (Arnaud1996), using data from the first seven orbits. A good fit (χ2/d.o.f. = 87.2/99)was obtained with a resulting spectral energy index of β = 1.00 ± 0.12 andexcess absorption (at z = 0.8 and for assumed Galactic abundances) of NH =

(2.82 ± 0.46) × 1022 cm−2 on top of the estimated Galactic absorption at this posi-tion (NH = 4.06 × 1020 cm−2, Dickey & Lockman 1990). The CXO data are fullyin agreement with these values, showing no change in the spectrum over time be-tween 0.3 and 3.3 days after the burst. The absorption measured is far less than thatmeasured by the HETE team in their prompt data, NH = (8.8+1.9

−1.8)×1022 cm−2 (Nak-agawa et al. 2006). This could indicate a change in absorption between the early(prompt) measurements and those at the time of the XRT observations. For theprompt emission spectrum, however, the values found by Konus-Wind (Golenet-skii et al. 2005) are rather different than those found by HETE-2, and may be theresult of the lower energy cut-off for FREGATE compared to Konus-wind. Alter-natively, the fact that these spectra are an average over the whole emission periodmay also result in incorrect model parameters. In the two last cases, the NH in theprompt emission could be as low as the XRT value and still produce an equally wellfit, but with slightly different model parameters.

For the XRT data, Butler et al. (2005) and Nakagawa et al. (2006) find a valuesomewhat higher than our value (4.9 × 1022 cm−2 and 5.3 × 1022 cm−2 respectively,when scaled by (1+z)3, Gunn & Peterson 1965). This difference could be explainedby a different count-binning or an updated XRT calibration used in our modeling.

The XRT light curve count rates have been converted to 1–10 keV fluxes usingthe results from our spectral modeling and calculating the ratio of the flux and countrate at the logarithmic center of the orbits. The 1 – 10 keV CXO flux was derivedusing the actual spectral fit.

A broken power law fit to the X-ray light curve results in α1 = 1.16 ± 0.06,α2 = 2.14±0.17 and a break time of 110+21

−23 ks, or around 1.27 days. The differencebetween α1 and α2, and the fact that the spectral slope does not change acrossthe break (the CXO measurement is past the break), are highly indicative that theobserved break in the light curve is a jet break. In Section 8.3.1, we perform fullmodeling of the afterglow using the fireball model, indeed resulting in a jet-breaktime tj that agrees reasonably well with the break time as determined from only theX-rays. We point out that our value for tj is different than that cited in Racusin et al.(2005b), largely because their measurement of tj was based on a preliminary XRTlight curve.

8.2.2 Optical and near infra-red observations

Observations were obtained in Z and R-band with the William Herschel Telescope(WHT) using the Auxiliary Port and the Prime Focus Imaging Camera, respec-tively, in r′i′z′ with the Gemini South telescope using the GMOS instrument, inJHKs with the Wide Field Camera on the United Kingdom InfraRed Telescope

8.2 Observations and data reduction 153

(UKIRT), in BVRI with the DFOSC instrument on the Danish 1.54m telescopeand in J and Ks with the Southern Astrophysical Research (SOAR) telescope usingOSIRIS. The optical data were reduced in a standard fashion using the ccdprocpackage within the IRAF software (Tody 1986), whereas the SOAR data were re-duced using the cirred package within IRAF. The UKIRT data were reduced usingthe standard pipeline reduction for WFCAM.

Photometric calibration was done using the calibration provided by Henden(2005) for Johnson-Cousins filters. For the r′i′z′ GMOS filters, we converted themagnitudes of the calibration stars provided by Henden to the Sloan filter systemusing the transformations provided by Jester et al. (2005), and verified by the pub-lished GMOS zero points. The WHT Z-band was calibrated using the spectroscopicstandard star SP2323+157. Calibration of the infrared JHK magnitudes was doneusing the 2MASS catalog (Skrutskie et al. 2006).

No variable optical source was found at the position of the X-ray and radioafterglow. For the early epoch images (< 1 day post burst), we estimated a limitingmagnitude by performing image subtraction between this and a later image usingthe ISIS image subtraction package (Alard 2000). To this end, artificial low signal-to-noise sources were added onto the images, with a Gaussian PSF matched in sizeto the seeing (some artificial sources were added on top of existing sources, e.g.galaxies, some on the background sky). We determined our upper limit to be thepoint where we could retrieve 50% of the artificial sources in the subtracted image.This assumes that the change in brightness of any point source on top of the hostgalaxy is sufficient to be seen in such a subtracted image. With the difference intime between the epochs, this seems a reasonable assumption (for example, for asource fading with a shallow power law like slope of F ∝ t−0.5, the magnitudedifference between the two WHT Z-band observations is ≈ 0.6 magnitudes).

Photometry of the host galaxy has been performed using aperture photometry,with an aperture 1.5 times the seeing for each image, estimated from the measuredFWHM of the PSF for point sources in the images.

Table 8.1 shows the log of our optical/nIR observations, while Table 8.2 showsthe upper limits for any optical/nIR afterglow.

8.2.3 Radio observations

Radio observations were performed with the WSRT at 8.4 GHz, 4.9 GHz and 1.4GHz. We used the Multi Frequency Front Ends (Tan 1991) in combination with theIVC+DZB back end1 in continuum mode, with a bandwidth of 8x20 MHz. Gainand phase calibrations were performed with the calibrators 3C 286 and 3C 48, al-though at one 8.4 GHz measurement 3C 147 was used. Reduction and analysiswere performed using the MIRIAD software package2. The observations are de-

1See sect. 5.2 at http://www.astron.nl/wsrt/wsrtGuide/node6.html2http://www.atnf.csiro.au/computing/software/miriad

154 GRB 051022

Start date ΔT (average) exposure time filter seeing telescope & instrument(days) (seconds) (arcsec)

2005-10-22T23:25:14 0.4287 1800 Z 0.8 WHT + API2005-10-23T00:22:33 0.4684 1620 J 1.2 SOAR + OSIRIS2005-10-23T00:56:00 0.4917 1620 Ks 1.3 SOAR + OSIRIS2005-10-23T00:48:03 0.5144 1920 i′ 0.6 Gemini South + GMOS2005-10-23T01:07:53 0.5288 1920 r′ 0.6 Gemini South + GMOS2005-10-23T01:27:46 0.5426 1920 z′ 0.5 Gemini South + GMOS2005-10-23T06:31:03 0.7525 720 J 1.4 UKIRT +WFCAM2005-10-23T06:36:39 0.7526 360 H 1.3 UKIRT +WFCAM2005-10-23T06:47:59 0.7604 360 K 1.3 UKIRT +WFCAM2005-10-23T21:15:57 1.3389 1200 Z 1.0 WHT + API2005-10-24T09:35:10 1.8467 720 K 0.3 UKIRT +WFCAM2005-10-25T01:34:03 2.5181 1602 Ks 1.3 SOAR + OSIRIS2005-10-25T02:13:18 2.5454 720 J 1.2 SOAR + OSIRIS2005-10-25T02:22:02 2.5698 1920 r′ 1.1 Gemini South + GMOS2005-10-25T02:39:59 2.5792 1440 z′ 1.2 Gemini South + GMOS2005-10-26T00:36:58 3.4785 1800 R 1.4 WHT+PFIP2005-10-26T02:48:06 3.5695 600 Gunn i 1.4 DK1.54m + DFOSC2005-10-26T03:23:35 3.5942 600 R 1.9 DK1.54m + DFOSC2005-10-27T01:01:04 4.4952 600 B 2.3 DK1.54m + DFOSC2005-10-27T02:59:20 4.5773 600 R 1.6 DK1.54m + DFOSC2005-10-27T02:00:48 4.5367 600 V 1.8 DK1.54m + DFOSC2005-10-28T02:18:38 5.5491 600 i 1.4 DK1.54m + DFOSC2005-10-30T02:32:59 7.5590 600 B 1.8 DK1.54m + DFOSC2005-10-30T04:18:30 7.6323 600 U 1.8 DK1.54m + DFOSC2005-10-30T01:33:57 7.5180 600 V 1.4 DK1.54m + DFOSC2005-10-31T03:19:05 8.5910 600 B 1.0 DK1.54m + DFOSC2005-10-31T01:03:40 8.4970 600 R 1.0 DK1.54m + DFOSC2005-10-31T02:10:02 8.5431 600 V 1.0 DK1.54m + DFOSC2005-11-01T01:52:57 9.5312 600 R 0.9 DK1.54m + DFOSC2005-11-02T02:04:47 10.539 600 V 1.2 DK1.54m + DFOSC2005-11-03T01:10:34 11.502 600 B 1.2 DK1.54m + DFOSC2005-11-07T01:25:30 15.512 600 Gunn i 1.4 DK1.54m + DFOSC2005-11-08T01:40:48 16.523 600 Gunn i 1.4 DK1.54m + DFOSC

Table 8.1 — Overview of optical observations

tailed in Table 8.3. In our modeling described in section 8.3.1 we have also usedthe VLA radio detection at 8.5 GHz from Cameron & Frail (2005).

8.3 Analysis

8.3.1 Broadband modeling

We have performed broadband modeling of the X-ray and radio measurements,using the methods presented in Van der Horst et al. (2007). In our modeling weassume a purely synchrotron radiation mechanism.

The relativistic blastwave causing the afterglow accelerates electrons to rela-

8.3 Analysis 155

filter limiting magnitudea ΔT (average) frequency fluxb

Ks > 20.0 0.5035 1.40 · 1014 < 6.82J > 20.3 0.4800 2.40 · 1014 < 12.3Z > 22.9 0.4443 3.43 · 1014 < 2.66z′ > 23.5 0.5426 3.36 · 1014 < 1.53r′ > 25.3 0.5288 4.76 · 1014 < 0.305

Table 8.2 — Limiting magnitudes. a See text for the definition of the limiting magnitude.b Specific fluxes have been corrected for a Galactic extinction value of EB−V = 0.04(Schlegel et al. 1998), and converted from magnitudes using the calibration by Tokunaga& Vacca (2005) for the JKs filters; the other filters are on the magnitude AB-system (Oke& Gunn 1983)

Start date ΔT (average) integration frequency specific flux(days) time (hours) (GHz) (μJy)

2005-11-04T18:14:24 13.37 4.0 8.5 38 ± 1322005-11-08T14:19:41 17.19 7.0 8.5 28 ± 972005-10-23T15:20:10 1.19 5.0 4.9 281 ± 322005-10-24T15:17:17 2.22 6.2 4.9 342 ± 342005-10-25T15:12:58 3.30 5.4 4.9 143 ± 302005-10-28T18:33:08 6.40 8.5 4.9 91 ± 282005-10-30T18:00:00 8.32 5.8 4.9 138 ± 282005-11-01T18:00:00 10.38 8.9 4.9 169 ± 282005-11-04T17:31:12 13.37 4.6 4.9 70 ± 342005-10-25T15:56:10 3.33 5.4 1.4 8 ± 78

Table 8.3 — Overview of WSRT radio observations

tivistic velocities, which gives rise to a broadband spectrum with three character-istic frequencies: the peak frequency νm, corresponding to the minimum energyof the relativistic electrons that are accelerated by the blastwave, the cooling fre-quency νc, corresponding to the electron energy at which electrons lose a significantfraction of their energy by radiation on a timescale that is smaller than the dy-namical timescale, and the self-absorption frequency νa, below which synchrotronself-absorption produces significant attenuation. The broadband spectrum is furthercharacterized by the specific peak flux Fν,max and the slope p of the electron energydistribution.

The dynamics of the relativistic blastwave determine the temporal behavior ofthe broadband synchrotron spectrum, i.e. the light curves at given frequencies. Atfirst the blastwave is extremely relativistic, but is decelerated by the surroundingmedium. When the Lorentz factor Γ of the blastwave becomes comparable to θ−1

j ,

156 GRB 051022

where θj is the opening angle of the jet, the jet starts to spread sideways. At thattime, tj, the temporal behavior of the broadband spectrum changes (see e.g. Rhoads1997).

We fit our data to six parameters: νc, νm, νa, Fν,max, p and tj. From these pa-rameters and the redshift of the burst, z = 0.8, we can find the physical parametersgoverning the blastwave and its surroundings: the blastwave isotropic equivalentenergy Eiso, the jet opening angle θj, the collimation corrected blastwave energyEjet, the fractional energy densities behind the relativistic shock in electrons andin the magnetic field, εe and εB respectively, and the density of the surroundingmedium. The meaning of the latter parameter depends on the density profile of thesurrounding medium. For a homogeneous circumburst medium, we simply deter-mine the density n. For a massive stellar wind, where the density is proportional toR−2 with R the distance to the GRB explosion center, we obtain the parameter A∗,which is the ratio of the mass-loss rate over the terminal wind velocity of the GRBprogenitor.

Our modeling results are shown in Table 8.4, for both the homogeneous ex-ternal medium and the stellar wind environment. The light curves for the best fitparameters are shown in Figure 8.1. We have performed Monte Carlo simulationswith synthetic data sets in order to derive accuracy estimates of the best fit param-eters, which are also given in the table. It is evident from the results that our sixfit parameters are reasonably well constrained in both cases for the circumburstmedium. The derived physical parameters are also well constrained, except for εe

and εB. The values we find for both the isotropic and the collimation correctedenergy, are similar to those found for other bursts; this is also true for p. See e.g.Panaitescu & Kumar (2001) and Yost et al. (2003). The jet opening angle and thedensity of the surrounding medium are quite small, but both not unprecedented.The jet-break time tj is somewhat smaller than estimated in Section 8.2.1, but bothestimates have relatively large errors, likely because of the lack of (X-ray) dataaround the jet-break time.

With the absence of optical light curves, it is not possible to discriminate be-tween the two different circumburst media. This is mainly due to the fact that theX-ray band lies above both νm and νc, in which case the slopes of the light curvesdo not depend on the density of the circumburst medium (even at 0.15 days, back-extrapolating νc from Table 8.4 results in its value being below the X-ray band).The χ2

red is somewhat better for the stellar wind case, but the homogeneous casecannot be excluded. From the X-ray light curve, however, one can conclude thatthe density profile of the medium does not change between approximately 0.15 and12 days after the burst. If there were a transition from a stellar wind to a homoge-neous medium, the X-ray flux has to rise or drop significantly, unless the densitiesare fine-tuned at the transition point (Pe’er & Wijers 2006). From the fact that themedium does not change during the X-ray observations, one can draw conclusionson the distance of the wind termination shock of the massive star: if one assumesthat the medium is already homogeneous at ≈ 0.15 days, the wind termination

8.3 Analysis 157

shock position is at Rw � 9.8 · 1017 cm (0.32 pc); if the circumburst medium is astellar wind up to ≈ 12 days after the burst, Rw � 1.1 · 1019 cm (3.7 pc).

Parameter Homogeneous Stellar wind

νc(tj) (1.45+1.12−0.23) · 1017 Hz (2.84+0.32

−1.30) · 1017 Hz

νm(tj) (3.50+2.26−1.47) · 1011 Hz (2.90+2.03

−1.15) · 1011 Hz

νa(tj) (4.56+2.85−3.08) · 109 Hz (2.68+2.17

−1.60) · 109 Hz

Fν,max(tj) 888+52−109 μJy 694+30

−240 μJy

p 2.06+0.19−0.05 2.10+0.08

−0.09

tj 0.96+0.40−0.28 days 1.06+0.41

−0.11 days

θj 3.39+2.02−2.27 deg 2.30+1.09

−0.85 deg

Eiso (5.23+1.13−1.69) · 1052 erg (28.2+31.0

−10.4) · 1052 erg

Ejet (0.917+0.655−0.512) · 1050 erg (2.27+2.25

−0.79) · 1050 erg

εe 0.247+1.396−0.212 0.0681+0.3951

−0.0348

εB (7.63+42.57−6.30 ) · 10−3 (8.02+28.18

−7.17 ) · 10−3

n (1.06+9.47−1.04) · 10−2 cm−3

A∗a (2.94+6.98−2.11) · 10−2

χ2red 1.9 1.5

Table 8.4 — Results of broadband modeling for both a homogeneous external mediumand a massive stellar wind. The best fit parameters are shown together with accuracyestimates from Monte Carlo simulations with synthetic data sets. The characteristicfrequencies of the synchrotron spectrum and the specific peak flux are given at tj. a Theparameter A∗ is a measure for the density in the case of a stellar wind environment, beingthe ratio of the mass-loss rate over the terminal wind velocity, and here given in units of10−5 Solar masses per year divided by a wind velocity of 1000 km/s (see Van der Horstet al. 2007).

158 GRB 051022

Figure 8.1 — Fit results for a homogeneous circumburst medium (left panel) and a mas-sive stellar wind (right panel). The solid and dash-dotted lines are the best model fits,and the dotted and dashed lines indicate the predicted rms scatter due to interstellar scin-tillation; see the appendix for further details. Also included in the figure (and modeling)is the reported VLA 8.5 GHz detection (Cameron & Frail 2005, left-most point in the8.5 GHz subplot).

8.3.2 The non-detection of the optical afterglow

It is quickly seen that GRB 051022 falls into the category of the so-called “darkbursts”. Using, for example, the quick criterion proposed by Jakobsson et al.(2004), we find βOX < −0.05 at 12.7 hours after the burst using the Gemini r′band observation, well below the proposed limit of βOX < 0.5. A more precise cri-terion would combine the available spectral and temporal parameters of the X-rayafterglow, allow all valid combinations, and from that infer the range of possibleoptical magnitudes from the X-rays (see e.g. Rol et al. 2005b). This is, in fact im-plied in our previous modeling: the modeled specific fluxes corresponding to theband and epoch of our optical and nIR upper limits are listed in Table 8.5 (see alsoTable 8.2).

While the values in this table are given for local extinction, not K-correctedto z = 0.8, it is immediately obvious that our K-band observations put a stringentconstraint on the required extinction directly surrounding the burst.

8.3 Analysis 159

filter upper limit homogeneous density profile stellar wind density profilemodeled flux extinction EB−V modeled flux extinction EB−V

(μJy) (μJy) (mag.) μJy) (mag.)Ks < 6.82 93.1 2.84 7.74 57.2 2.31 6.29J < 12.3 117 2.44 2.71 74.1 1.95 2.16Z < 2.66 103 3.97 2.58 67.8 3.52 2.29r′ < 0.305 74.5 5.97 2.17 44.4 5.41 1.97z′ < 1.53 87.7 4.40 2.97 51.9 3.83 2.59

Table 8.5 — Upper limits compared to model specific flux calculations. The inferredlower limits on the extinction are given in the observers frame. The EB−V values aregiven for a Galactic extinction curve (RV = 3.08), and are for illustrative purposes; seethe comments at the end of Section 8.3.2.

To estimate the amount of local extinction in the host galaxy, we have modeledthe nIR to X-ray spectrum around 0.5 days after the burst, considering 3 differentextinction curves: those of the Milky Way (MW), the Large Magellanic Cloud(LMC) and the Small Magellanic Cloud (SMC), from Pei (1992), with RV of 3.08,3.16 and 2.93, respectively.

For this, we used the unabsorbed XRT flux obtained from the spectral fit toorbits 3 – 7 (which do not contain piled-up data), and fixed the energy spectralslope in the X-rays at β = 1 (also from the X-ray spectral fit). The optical specificfluxes were scaled to the logarithmic mid-observation time of the X-ray observa-tions with an assumed α = 1.16 decline. This estimated optical decay is derivedfrom the pre-break X-ray decay value, allowing for the cooling break between thetwo wavelength regions, and averaging the two possible values for αX − αopt (-0.25and 0.25). We can further put the most stringent constraint on the broken power lawspectral shape, by setting the spectral break just below the X-rays, at 1.8× 1017 Hz,which follows from our previous broad-band modeling. Our results indicate that,for the aforementioned extinction curves, a local extinction of EB−V ≈ 7 (for allthree extinction curves) is necessary to explain the K-band upper limit.

We can relate the resulting NH from our X-ray spectral fits to any local EB−V , us-ing the relations found in Predehl & Schmitt (1995), Fitzpatrick (1985) and Martinet al. (1989) for N(HI)/EB−V , and adjusting the metallicity in our X-ray absorptionmodel accordingly. We obtain EB−V = 7.5, 1.54 and 0.84 for a MW, LMC and SMCextinction curve respectively, with the MW value showing the best agreement withour findings for optical extinction. This, obviously, depends on the assumption thatthe MW (or otherwise, LMC or SMC) extinction curves are valid models to com-pare with our observed data here. Since these data happen to originate from justone sight line in a galaxy, this may very well not be the case. Further, even if theextinction curve is correct, the actual value of RV may be rather different for thehost galaxy. Finally, the EB−V – NH relations show a rather large scatter, especiallyat higher column densities, nor is the NH always derived using X-ray spectroscopy.

160 GRB 051022

Our above results are therefore approximations, which are useful to compare withother (GRB host) studies, but should be taken with the necessary caution.

8.3.3 The host galaxy of GRB 051022

filter magnitude magnitude errorK 18.40 0.04Ks 18.36 0.09H 19.42 0.09J 19.92 0.05

Za 21.41 0.05z′ 21.30 0.04i′ 21.77 0.01r′ 22.04 0.01R 21.84 0.09V 22.30 0.04B 22.75 0.02U > 21.3 b

Table 8.6 — Measured host galaxy magnitudes. a AB magnitude b 5-σ upper limit

Using the optical data described above, we fit the SED of the host ofGRB 051022 using the HyperZ program3 developed by Bolzonella et al. (2000).The photometry of the host has been performed using apphot within IRAF, in anaperture 1.5 times the estimated seeing in the different exposures. The results arereported in Table 8.6 (see also Ovaldsen et al. 2007). The range of photometricmagnitudes reported in this paper provides one of the most complete broadbandoptical datasets of a GRB host galaxy to date. We fit using the eight syntheticgalaxy templates provided within HyperZ at the redshift of the host, and find thatthe host galaxy is a blue compact galaxy of type irregular, with a dominant stellarpopulation age of ≈ 20 Myr, similar to other long GRB hosts (Christensen et al.2005). A moderate amount of extinction of AV ≈ 1 mag is required to fit the SED,with an SMC-type extinction curve providing a best fit, and the luminosity of thehost is approximately 1.5 L∗ (assuming M∗,B = −21); these findings are in fullagreement with Castro-Tirado et al. (2006). The amount of extinction in the lineof sight towards the GRB required to suppress the optical light of the afterglow tothe observed limits is clearly higher than the AV value found from the host SED:AV = 4.4 magnitudes towards the GRB, estimated from blueshifting our measured(observer frame) z′ band extinction to z = 0.8. The host galaxy SED extinction is,however, an average value derived from the integrated colors of the host.

3See http://webast.ast.obs-mip.fr/hyperz

8.4 Discussion 161

The host of GRB 051022 is located in a field crowded with galaxies of variousHubble types. We perform photometry on several galaxies close to the GRB host(within 1 arcminute) to investigate the possibility that the high star formation rateseen in the optical (Castro-Tirado et al. 2006 report an SFR of ≈ 20M�yr−1) isinduced by a recent interaction with one of the neighboring galaxies. As forma-tion of high mass stars has also been observed to occur in dusty regions in mergingsystems (see e.g. Lin et al. 2007), this could help to explain the excess optical ex-tinction towards GRB 051022. We performed HyperZ fits to these galaxies, andfind that none of them is well fit by a photometric redshift of z ≈ 0.8. Particularlythe two galaxies closest to the GRB host galaxy are not compatible with a redshift0.8, and show best fits with photometric redshifts of z ≈ 0.2 – 0.25. Out of the sam-ple of six galaxies close to the GRB host we find that four have best-fit photometricredshifts in the range 0.20 – 0.25, making it unlikely that a possible overdensity ofgalaxies near the host galaxy is due to a cluster or galaxy group at the host redshift.

8.4 Discussion

The issue of non-detected (“dark”) GRB afterglows has received significant in-terest ever since the discovery of the first GRB afterglow, starting with the non-detection of GRB 970828 to very deep limits (Groot et al. 1998; Odewahn et al.1997). For this particular afterglow, its non-detection has been attributed to a dust-lane in its host galaxy (Djorgovski et al. 2001). Dust extinction as the cause of thenon-detection of the optical afterglow has been inferred in the case of several otherGRBs, notably those with a precise X-ray or radio position, where one can pinpointthe afterglow position on top of its host galaxy (e.g. GRB 000210, Piro et al. 2002).

Optical drop-outs due to high redshift will also result in dark bursts, but areharder to confirm, since it would require at least one detection in a red band, todetect the Lyα break. Otherwise, it becomes indistinguishable from dust extinction.

Other explanations of afterglow non-detections include the intrinsic faintnessof the afterglow. For HETE-2 detected GRBs, this has been inferred for e.g.GRB 020819 (Jakobsson et al. 2005c). For Swift bursts, where rapid and accurateX-ray positions are often available, this is a major cause of non-detections (Bergeret al. 2005a), largely attributed to a higher average redshift.

In our case here, the host galaxy has been detected at a relatively modest red-shift, which almost automatically points to the dust extinction scenario. The radioand X-ray detections even allow us to accurately model the necessary amount ofextinction between us and the GRB.

8.4.1 The burst environment

The issue of the role of dust extinction in the lines of sight towards GRBs is stillvery much an open one. While clear signs of dust depletion are seen in several af-

162 GRB 051022

terglow spectra, the AV values that are predicted from these depletion measures aregenerally much higher than the observed ones, that can be found from the contin-uum shape (Savaglio & Fall 2004). Recently, selected samples of GRB afterglowswere homogeneously analyzed for X-ray and optical extinction, showing dust togas ratios different from Galactic and Magellanic cloud values (Starling et al. 2006;Schady et al. 2007). Galama & Wijers (2001) and Stratta et al. (2004) had alreadyfound dust (optical) to gas (X-ray) ratios to be lower than the Galactic value (in allcases, however, there is a bias in these samples to optically and X-ray detected af-terglows). Comparison of neutral hydrogen columns and metallicities of afterglowlines of sight with X-ray extinction values (Watson et al. 2007) showed that the ab-sorption probed by these two wavelength regimes is generally located at differentpositions in the host. In all these cases there may be significant biases against burstswith low apparent magnitudes, preventing optical spectroscopy, which are hard toquantify.

In the case of GRB 051022 there is a significant discrepancy between the ex-tinction for the host as a whole and that along the line of sight to the burst, or atleast along our line of sight towards the burst. This is perhaps not too surprising ifone assumes, for example, that the burst occurred inside a Giant Molecular Cloud(GMC). Jakobsson et al. (2006a) compared the GRB N(HI) distribution to that ofmodeled GRBs located inside Galactic-like GMCs. They found that the two distri-butions are incompatible, and possibly GRBs are more likely to occur inside cloudswith a lower N(HI), or alternatively, outside the actual GMC. (Note that their studyconcentrates on bursts with z > 2, where the Ly-α absorption is visible in the opticalwavebands; it is also biased towards optically detected afterglows). A GMC couldtherefore actually be positioned in front of the GRB, where the required opticaland X-ray extinction is easily achieved. This agrees with the findings by Prochaskaet al. (2007a), who analyzed several GRB-Damped Lyman Alpha spectra and fromobserved depletion levels infer that the gas is not located directly near the GRB (e.g.its molecular cloud) but further out. The specific case of GRB 060418 confirmedthis through time-resolved high resolution spectroscopy, showing that the observedmetal lines originate past 1.7 kpc from the burst itself (Vreeswijk et al. 2007). Infact, dust destruction (Waxman & Draine 2000; Fruchter et al. 2001; Draine & Hao2002) could easily destroy grains out to 100 pc and permit the afterglow radiationto penetrate the surrounding molecular cloud. Dust extinction is therefore likely tooccur further out, perhaps to several kiloparsecs.

It is interesting to find a non-SMC type of extinction curve from the combi-nation of X-ray and optical absorption (though not completely ruled out): in mostcases modeled, an SMC extinction curve fits the optical–X-ray spectra best (Star-ling et al. 2006; Schady et al. 2007), presumably attributable to the absence of the2175 Å feature (Savage & Mathis 1979) and the low dust to gas ratio. Our findingsindicate that the extinction along the line of sight to the GRB will generally be dif-ferent than one of the three assumed extinction curves. Local small scale densityvariations in clouds, such as found by from infrared studies in the Taurus region

8.4 Discussion 163

and from simulations (Padoan et al. 2006), could cause this fairly easily.

8.4.2 Energetics

Our modeling provides us with a detailed set of parameters of the afterglow ener-getics, including Ejet, the energy of the afterglow. For the prompt emission energy,we use the data from the Konus-Wind measurements (Golenetskii et al. 2005). Wecalculate a prompt isotropic energy of 4.39+0.29

−0.18 × 1053 erg in the 20 keV – 20 MeVobserver frame, and, by applying a K-correction (as in e.g. Bloom et al. 2001b),Ep,iso = 10.4+0.7

−0.4 ×1053 erg in the 1 – 105 keV rest frame. The collimation correctedenergy depends on the assumed structure of the surrounding medium: for a homo-geneous medium, we obtain Ep,jet = 18.2 × 1050 erg, and for a wind-like medium,Ep,jet = 8.38 × 1050 erg. With Epeak = 918+66

−59 keV in the burst rest frame, we findthat the Epeak – Ep,jet relation (Ghirlanda et al. 2004) somewhat underestimates theEpeak when calculated from Ep,jet: Epeak ≈ 740 keV for a homogeneous medium,and ≈ 430 keV for a wind medium (the difference between our chosen cosmologyand that used by Ghirlanda et al. 2004 amounts to only a 0.3% difference in Eiso).These estimates, however, come with a few caveats: 1) the Epeak from the Konus-Wind data is calculated using an exponential cut-off model, not the Band function(Band et al. 1993). Since the Band function includes the case of an exponential cut-off model (with β = −∞, this should, however, pose no problem in estimating theactual Epeak), 2) our break time, and therefore the jet-opening angle, are calculatedfrom the full modeling of the afterglow, which effectively means derived from theavailable X-ray and radio data. Further, the original Ghirlanda relation was derivedusing optical break times. Recent efforts show that estimating jet-break times fromX-ray light curves may not lead to the same results (e.g. Panaitescu et al. 2006), and3) the relatively large error on the jet opening angle estimate allows for a relativelylarge range in collimation corrected energies. We have simply used here our bestvalue, but an Epeak value of 1498 keV derived from Ejet can still be accommodatedwithin our errors. (We note that, with a different Epeak estimate and an incorrectvalue for the jet-break time, Nakagawa et al. 2006 still found their results to lie onthe Ghirlanda relation). The break time problem can be avoided by looking only atthe Epeak – Ep,iso relation (Amati et al. 2002; Amati 2006). From this, we estimateEpeak ≈ 924 keV, nicely in agreement with the value found directly from the spectrafit.

Comparing the prompt emission energy (Ep,jet) and afterglow blast wave kineticenergy (Ejet), we find their ratio to be Ep,jet/Ejet = 3.7 in the case of a wind-likecircumburst medium, while for a homogeneous medium, Ep,jet/Ejet = 20. Theseratios are similar to those found for other bursts (e.g. Berger et al. 2003b, Figure 3).

GRB 051022 is also one of the brightest bursts observed by HETE, with aprompt 30–400 keV fluence of S = 1.31 × 10−4 erg cm−2 (Nakagawa et al. 2006).In fact, compared to the sample of 35 FREGATE bursts analyzed by Barraud et al.(2003), GRB 051022 has the largest fluence, even topping the relatively close-by

164 GRB 051022

GRB 030329 (Vanderspek et al. 2004, S = 1.2 × 10−4 erg cm−2; note that forGRB 051022, its redshift is close to the median redshift of HETE-2 detected GRBsand therefore distance effects will play a very minor role). Rol et al. (2005b) notedthis potential correlation of fluence with the non-detection of a GRB afterglow forthe small subset of genuinely dark bursts in their sample: the truly dark bursts allhave a much higher than average fluence (although this is for a relatively small sam-ple only). Potentially, this could point to an external origin for the prompt emission,instead of being due to internal shocks: a large amount of dust may result in morematter that will radiate, while at the same time the radiation will be suppressedat UV and optical wavelengths. This would indicate an origin of the extinctionquite close to the burst instead, in contrast to previous findings for other bursts,as discussed in Section 8.4.1. These latter bursts, however, were all optically se-lected to obtain spectroscopy, and may therefore show different surroundings thanGRB 051022. Unfortunately, with the small sample size of genuine dark bursts afirm conclusion on this correlation is not possible, but remains something to watchfor in future dark bursts.

8.5 Conclusions

GRB 051022 is a prototypical dark burst, with the local extinction exceeding 2.3magnitudes in J and 6 magnitudes in U, in the host-galaxy restframe. The extinc-tion curve derived from an X-ray – optical spectral fit points towards a Galactic typeof extinction curve, although it is likely that this is more or less a coincidence: thehost galaxy itself is best modeled with an SMC-like extinction curve, with a modestamount of extinction, AV ≈ 1 mag. The large optical absorption towards the after-glow of GRB 051022 is therefore probably the effect of an unfortunate position inthe host where the line of sight crosses dense regions within the host.

The X-ray and radio afterglow data allow for a full solution of the blast-wave model, although we unfortunately cannot distinguish between the densityprofile (homogeneous or wind-like) of the circumburst medium. We estimate acollimation-corrected energy in the afterglow emission of 0.92 – 2.3 ×1050 erg,while the energy in prompt emission (1 – 105 keV rest frame) is 8.4 – 18 ×1050 erg.Aside from the large optical extinction, the afterglow otherwise appears as an av-erage afterglow, with no outstanding properties. The potentially interesting pointhere is that the 30-400 keV fluence of the prompt emission is one of the largest everdetected in the HETE-2 sample.

In the era of Swift GRBs, dust-extincted bursts can actually be found in op-tical/nIR thanks to the rapid availability of precise positions: examples are foundwhere the burst is relatively bright early on at optical/nIR wavelengths, while theafterglow proper (post few hours) often can go undetected (e.g. Oates et al. 2006;Perley et al. 2007). This allows targeted follow-up of such dark bursts, i.e. de-termining the host galaxy (and the bursts precise position therein) and a redshift

8.6 Appendix : Interstellar scintillation in the radio modeling 165

measurement. In our case, a precise CXO and radio position pinpointed the hostgalaxy, but such data may not always be available. High resolution late-time obser-vations of the host, at the location of the GRB, may then reveal whether the burstindeed occurred inside a dense host region.

8.6 Appendix : Interstellar scintillation in the radiomodeling

The 4.9 GHz measurements show scatter around the best fit light curve, which canbe accounted for by interstellar scintillation (ISS). In Figure 8.1 we have indicatedthe predicted rms scatter due to ISS. We have calculated the scattering measurefrom the Cordes & Lazio (2002) model for the Galactic distribution of free elec-trons: S M = 2.04 · 10−4 kpc /m−20/3. The radio specific flux will be modulatedwhen the source size is close to one of the three characteristic angular scales, i.e.for weak, refractive or diffractive ISS. From Walker (1998), we calculate the transi-tion frequency between weak and strong ISS, ν0 = 9.12 GHz, and the angular sizeof the first Fresnel zone, θF0 = 0.994 μas. Our measurements were all performedat frequencies below ν0, i.e. in the strong ISS regime, which means that only re-fractive and diffractive ISS modulate the specific flux significantly. We calculatethe evolution of the source size in the extreme relativistic phase (θs = R/Γ) andafter the jet-break (θs = Rθj), and compare this source size with the diffractive an-gular scale θd = θF0(ν0/ν)

−6/5 = 0.0701 · ν6/5GHz μas and the refractive angular scaleθr = θF0(ν0/ν)

11/5 = 128 · ν−11/5GHz μas to calculate the modulation index mp. In the

case of diffractive ISS the modulation index is 1, and in the case of refractive ISSmp = (ν0/ν)−17/30 = 0.286 · ν17/30

GHz . Because of the expansion of the blastwave theangular source size exceeds one of the characteristic angular scales at some pointin time. Then the modulation will begin to quench as mp(θd/θs) in the case ofdiffractive ISS, and as mp(θr/θs)7/6 in the case of refractive ISS.

Acknowledgements We thank the referee for a careful reading of the manuscriptand constructive comments. We thank Kim Page and Andy Beardmore for useful dis-cussions regarding the XRT data analysis. ER and RLCS acknowledge support fromPPARC. KW and RAMJW acknowledge support of NWO under grant 639.043.302.The authors acknowledge funding for the Swift mission in the UK by STFC, in the USAby NASA and in Italy by ASI. The Dark Cosmology Centre is funded by the Danish Na-tional Research Foundation. The William Herschel Telescope is operated on the islandof La Palma by the Isaac Newton Group in the Spanish Observatorio del Roque de losMuchachos of the Instituto de Astrofısica de Canarias. The United Kingdom InfraredTelescope is operated by the Joint Astronomy Centre on behalf of the U.K. ParticlePhysics and Astronomy Research Council. The data reported here were obtained as partof the UKIRT Service Programme. The Westerbork Synthesis Radio Telescope is op-

166 GRB 051022

erated by ASTRON (Netherlands Foundation for Research in Astronomy) with supportfrom the Netherlands Foundation for Scientific Research (NWO). Support for this workwas provided by the National Aeronautics and Space Administration through ChandraAward Number 1736937 issued by the Chandra X-ray Observatory Center, which is op-erated by the Smithsonian Astrophysical Observatory for and on behalf of the NationalAeronautics Space Administration under contract NAS8-03060. This publication makesuse of data products from the Two Micron All Sky Survey, which is a joint project of theUniversity of Massachusetts and the Infrared Processing and Analysis Center/CaliforniaInstitute of Technology, funded by the National Aeronautics and Space Administrationand the National Science Foundation. This research has made use of data obtained fromthe High Energy Astrophysics Science Archive Research Center (HEASARC), providedby NASA’s Goddard Space Flight Center.

Samenvatting in het Nederlands

Inleiding

Als je op een donkere, heldere nacht omhoog kijkt, dan zie je een grote hoeveelheidsterren. Ook zie je meestal de Maan en enkele planeten. Als het echt goed donkeris, dan zul je recht boven je hoofd een vage, lichte band zien lopen: de Melkweg.De Melkweg is in feite niets anders dan een grote verzameling sterren die in eenplatte schijf ronddraaien, waar onze Zon deel van uitmaakt. Als je nu een jaar laterweer buiten omhoog kijkt, dan is er vrij weinig veranderd. De Maan zal op een watandere plek staan, en ook de planeten zullen van positie zijn veranderd. Behalvedat zullen de sterren onveranderd lijken. Ik zeg met opzet “lijken”, want er is weldegelijk het een en ander veranderd. Wellicht het meest opvallende voorbeeld is dester Mira. Deze ster is een zogenaamde veranderlijke ster, en veranderd over eenperiode van zo’n 330 dagen sterk van helderheid van onzichtbaar tot een van dehelderste sterren aan de hemel.

De meeste sterren aan de hemel zijn variabel, maar meestal zijn de helderheids-verschillen veel te zwak om met het oog waar te nemen. Nu zijn onze ogen alleengevoelig voor zichtbaar licht, maar als we het heelal in kijken met telescopen dieandere soorten straling kunnen waarnemen (bijvoorbeeld rontgenstraling), dan zienwe heel andere objecten maar ook daar een heel dynamisch heelal waarin op korteen lange tijdschalen van alles verandert. Niets demonstreert beter het dynamischheelal dan gammaflitsen, de meest extreme veranderlijke objecten en het onderwerpvan dit proefschrift.

Gammaflitsen

De ontdekking van gammaflitsen vond plaats in de 60-er jaren, in de gegevens vande Vela satellieten. Deze satellieten waren gebouwd door de VS, om in de gatente houden of de USSR zich wel hield aan het verdrag dat atoombom proeven in dedampkring en in de ruimte verbood, de Nuclear Test Ban Treaty. Deze satellietenkonden goed gammastraling detecteren: een heel energierijke vorm van straling dieonze ogen niet kunnen waarnemen. Deze straling wordt goed geabsorbeerd door deaardatmosfeer, en moet daarom voornamelijk vanuit de ruimte worden waargeno-men. Wanneer een atoombom ontploft komt er een korte flits gammastraling vrij,en dat kon gedetecteerd worden door deze satellieten. En jawel, flitsen van gam-mastraling werden inderdaad gezien. Gelukkig waren dit geen bommen, maar dewerkelijke oorzaak van deze flitsen was lange tijd niet vast te stellen. Naarmate ermeer en meer werden gezien (een paar per jaar), werd wel duidelijk dat deze flitsenin ieder geval niet van de Aarde afkomstig konden zijn, maar waar dan wel vandaan

167

168 Samenvatting in het Nederlands

Figuur 8.2 — Kaartje van de hemel in gammastraling, gemaakt door het EGRET in-strument aan boord van de CGRO satelliet. De gammastralingsbronnen zijn de zwartevlekken. In het centrum van onze Melkweg zitten veel van die bronnen, dat is de concen-tratie van punten in het midden van het plaatje. Wanneer er een gammaflits plaatsvindt,is die gedurende enkele seconden helderder dan alle andere gammastralingsbronnen bijelkaar opgeteld. (Credit: EGRET/NASA)

kon niet worden vastgesteld. Deze flitsen van gammastraling werden gammaflitsen(in het Engels “gamma-ray bursts”) genoemd.

Vanaf 1991 bracht een nieuwe satelliet meer duidelijkheid. Deze satelliet(Compton Gamma-Ray Observatory, CGRO) had acht gevoeliger detectoren speci-aal ontworpen voor gammaflitsen aan boord. Deze satelliet kon hierdoor veel meergammaflitsen per jaar vinden (zie Figuur 8.3). Bovendien was deze satelliet in staatom veel nauwkeuriger de positie aan de hemel vast te stellen, iets wat erg moeilijkis voor gammastralingsbronnen. Deze betere posities lieten duidelijk zien dat gam-maflitsen willekeurig over de hemel waren verspreid. Dat laatste betekende dat zeniet uit ons melkwegstelsel kwamen maar van ver weg in het heelal.

Door de grote hoeveelheid gedetecteerde gammaflitsen werd duidelijk dat ertwee verschillende soorten gammaflitsen zijn: er zijn korte gammaflitsen (die durengemiddeld korter dan 2 seconden), en er zijn lange gammaflitsen (deze duren langerdan twee seconden en kunnen tot zo’n kwartier lang actief zijn).

Een volgende grote revolutie kwam door de Italiaans-Nederlandse satel-liet BeppoSAX. Deze satelliet had behalve Italiaande telescopen voor zach-te rontgenstraling en gammadetectoren ook twee in Nederland ontwikkeldegroothoek-rontgencamera’s aan boord. Het idee hierachter is dat wanneer er veelgammastraling geproduceerd wordt, er ook veel rontgenstraling (een minder ener-

Samenvatting 169

Figuur 8.3 — Een collectie gammaflitsen: de helderheid is uitgezet tegen de tijd in se-conden. Het is duidelijk te zien dat gammaflitsen er allemaal anders uitzien. De meestrechtse gammaflits op de onderste rij is een korte gammaflits. (Credit: J.T. Bonnell(NASA/GSFC))

gierijke vorm van straling) vrijkomt, waarvan makkelijker de positie aan de hemelgemeten kan worden.

Op 11 januari 1997 was het voor het eerst raak: een rontgenbron werd gede-tecteerd vlak na een gammaflits. De positie aan de hemel kon hierdoor nauwkeurigworden bepaald. Deze posities werden snel doorgegeven aan teams op de grond,die op hun beurt optische telescopen (die kijken naar zichtbaar licht) konden rich-ten op die positie om te zien of er een tegenhanger in zichtbaar licht te zien was.Er werd echte rniets gevonden. Op 28 februari 1997 was het wederom raak, en ditkeer vielen alle stukjes van de puzzel op hun plek: Amsterdamse sterrenkundigenslaagden erin om voor het eerst een nieuw lichtpuntje aan de hemel te vinden op de


plek van een gammaflits (Figuur 8.4), dat razendsnel zwakker werd. De zwakkerwordende straling na een gammaflits wordt met een passende naam de nagloeiergenoemd (in het Engels “afterglow”).

De detectie van nagloeiers is heel belangrijk, omdat uit de combinatie van derontgen- en optische helderheid veel duidelijk wordt over de fysica van deze na-gloeiers. Ook kan via een nagloeier in zichtbaar licht makkelijker direct de afstandgemeten worden.

Figuur 8.4 — Twee opnames gemaakt in zichtbaar licht van een klein stukje van dehemel waar een gammaflits is afgegaan. Duidelijk is te zien dat op de opname links eenhelder object staat, dat op de rechter opname (anderhalve week later gemaakt) volledigis verdwenen. Dit is de optische nagloeier van de gammaflits die op 28 februari 1997afging (Van Paradijs et al. 1997), de allereesrte nagloeier ooit ontdekt.

Op latere, langer belichte opnames, was duidelijk te zien dat op de plek van denagloeier zich een klein, zwak en ver weg gelegen melkwegstelseltje bevond. Hetwas hiermee duidelijk geworden dat gammaflitsen in melkwegstelsels zeer ver weggebeuren.

Het waarnemen van gammaflitsen

Om meer te weten te komen over gammaflitsen en hun nagloeiers, is het belangrijkom er zoveel mogelijk te detecteren in zoveel mogelijk golflengtes (zichtbaar licht,rontgen-, maar ook radiogolflengtes). Dit is geen gemakkelijke zaak omdat denagloeiers snel zwakker worden. De nieuwe satelliet Swift (gelanceerd eind 2004;Figuur 8.5) maakt dit heel veel makkelijker. Deze satelliet kan een groot deel vande hemel in de gaten houden met zijn gammastralingsdetector. Deze detector iszeer gevoelig, waardoor er zo’n 100 gammaflitsen per jaar worden gedetecteerd.

Als er een gammaflits wordt gedetecteerd door het BAT instrument, dan kande computer aan boord van de satelliet zelf de positie uitrekenen, en de satelliet

Samenvatting 171

snel naar de positie van de gammaflits toedraaien. Aan boord van Swift bevindenzich namelijk ook een rontgen- en een ultraviolettelescoop die maar een klein deelvan de hemel kunnen zien. Zodra de satelliet gedraaid is (dit duurt meestal min-der dan een minuut), beginnen deze camera’s de gammaflits en zijn nagloeier waarte nemen. Hierdoor kan veel nauwkeuriger de positie worden bepaald. De eersteposities die de satelliet uitrekent uit de gamma- en rontgenstraling worden binnenenkele seconden door een netwerk van communicatiesatellieten naar de Aarde ge-stuurd. Daar worden die berichten opgevangen en verder verspreid. Binnen enkeleseconden komen ze terecht bij waarnemers (bijvoorbeeld bij mij) via email en SMS.Zodra we zo’n positiebericht ontvangen kunnen we snel via internet ervoor zorgendat grote telescopen op Aarde de positie van de gammaflits gaan waarnemen. Om-dat deze waarnemingen direct gedaan moeten worden, kunnen we niet eerst naarde telescoop toe reizen, en moeten we het waarneemprogramma van de waarnemerter plekke onderbreken (dit wordt ook wel een “Target of Opportunity override”genoemd, een onderbreking voor een belangrijk, tijdelijk zichtbaar object).

De opnames die de telescopen maken kunnen we vaak binnen een uur al viainternet ontvangen. Deze data moeten dan eerst bewerkt worden, en de helderheidvan de nagloeier kan dan worden bepaald. De gegevens worden dan in de vorm vaneen kort rapport (een Circular) rondgestuurd naar alle andere waarnemers. Dit moetsnel, omdat de helderheid van een nagloeier snel afneemt. Omdat je van tevorenniet weet waar een gammaflits af zal gaan, moeten we ervoor zorgen dat we opgrote telescopen over de hele Aarde verspreid kunnen “inbreken”. Hiervoor moetenwe (meestal twee maal per jaar) de beheerders van deze telescopen overtuigen dathet heel belangrijk is om ons die status te geven, en werken we vaak met groteinternationale collaboraties, wat wel duidelijk is aan de lange lijst van auteurs vande hoofdstukken.

Behalve metingen van de helderheid van nagloeiers en hoe snel die afneemt,maken we ook spectra. Spectroscopie is het uiteenrafelen van het licht van een ob-ject in de verschillende kleuren (golflengtes) waaruit het is opgebouwd (zo maakteen regendruppel een spectrum van het zonlicht: een regenboog). Nu laten chemi-sche elementen (bijvoorbeeld ijzer, zuurstof) sporen achter in zo’n spectrum: speci-fieke kleuren licht worden geabsorbeerd. We weten precies welk element bij welkegolflengtes absorbeert, en kunnen daardoor de chemische eigenschappen meten ennauwkeurig de afstand tot de gammaflits bepalen. Voor spectroscopie moet meest-al de nagloeier wat helderder zijn dan voor een helderheidsmeting, dus hebben wehier vaak de grootste telescopen ter wereld voor nodig.

Welke sterren zijn veroorzaken gammaflitsen?

De afstanden tot gammaflitsen zijn heel erg groot, en op zo’n grote afstand kunnenwe niet in detail zien waar de flits vandaan komt. Ook kunnen we niet op eerdergemaakt opnames van iemand anders zien of we een object zien dat ontplofte in


XRT

BAT

UVOT

Figuur 8.5 — De Swift satellliet. De drie hoofd-instrumenten staan aangege-ven: het gammastraling instrument (BAT) dat als eerste de gammaflits ontdekt, derontgentelescoop XRT en de ultraviolettelescoop UVOT.

een gammaflits. Gelukkig helpt moeder natuur ons een handje: heel af en toe vindter toevallig een gammaflits plaats die wat dichter bij ons is. Bij dichtbije langegammaflitsen hebben we duidelijk kunnen zien dat de nagloeier na een paar dagenineens weer helder werd, in plaats van zwakker. In sommige gevallen kunnen weduidelijk zien dat het hier gaat om een bijzonder type supernova. Van dit soortsupernovae weten we heel goed hoe ze ontstaan:

Sterren bevatten een heleboel massa, en hebben de neiging om onder hun eigengewicht in elkaar te storten door de zwaartekracht. Dit wordt tegengegaan doordatsterren in hun kern energie opwekken, wat de druk binnenin vergroot en weerstandbiedt tegen de zwaartekracht. Deze energie wordt opgewekt doordat sterren in hunbinnenste waterstof omzetten in helium, in een proces dat kernfusie heet. Op eengegeven moment raakt de brandstof in de kern op, en zal de ster geen weerstandmeer kunnen bieden aan de zwaartekracht en in elkaar storten. Hierbij komt enormveel energie vrij, en wordt het object gedurende korte tijd (enkele maanden) ex-treem helder: een supernova. Sterren met een grotere massa moeten meer energieopwekken in hun kern om de grotere zwaartekracht te weerstaan, en raken dus ooksneller door hun brandstof heen. De zwaardere sterren leven dus het kortst. Enhet zijn nu net de allerzwaarste sterren die een supernova maken van het type datwe detecteren bij lange gammaflitsen. Het is daarmee duidelijk dat lange gamm-aflitsen ontstaan bij de dood van de zwaarste sterren. Toch komt er niet bij elkesupernova een gammaflits voor, dus kennelijk zijn er nog meer eigenschappen dieeen rol spelen. We denken dat bij het instorten van zo’n zware ster door zijn eigenzwaartekracht, er vanuit de kern gedurende korte tijd een bundel van materie wordtuitgestoten met snelheden dicht bij de lichtsnelheid. Het is deze bundel die de gam-mastraling en de latere nagloeier produceert. Het uitstoten van zo’n bundel is ietswaar de ster waarschijnlijk heel specifieke eigenschappen voor moet hebben.

Het verhaal voor de korte gammaflitsen is wat anders: daarvoor zien we geen

Samenvatting 173

supernovae. We denken dat deze gammaflitsen voornamelijk ontstaan bij de bot-sing van twee zware, maar heel compacte sterren. Bij zo’n botsing komt ook veelenergie vrij en kan er op dezelfde manier als bij de lange gammaflitsen een gamm-aflits ontstaan.

Figuur 8.6 — Sommige nagloeiers van gammaflitsen kunnen kortstondig helderder zijndan magnitude 19, en zijn dus helder genoeg om te worden vastgelegd door amateurtele-scopen en camera’s. Gemiddeld worden er over de wereld zo’n 5 nagloeiers per jaar dooramateurs gedetecteerd. Deze opname is van de supernova die gammaflits 060218 ver-gezelde, gemaakt met de 25cm ABT internet-telescoop van AWSV Metius te Alkmaar(Wiersema & Nieuwenhout 2006). Op het tijdstip van deze opname was de supernovaop zijn helderst met een magnitude 17.4.

De melkwegstelsels

Nu we weten dat de lange gammaflitsen gemaakt worden door zware sterren, kun-nen we proberen meer te weten te komen over welke eigenschappen deze sterrenprecies hebben. Hier kunnen we gebruik maken van de melkwegstelsels waar degamnmaflitsen vandaan komen. Voor bijna alle lange gammaflitsen geldt dat nadatde nagloeier te zwak is geworden om te zien, er een heel zwak klein melkwegstel-seltje kan worden gevonden: hierin heeft de gammaflits plaats gevonden (in hetEngels een “host galaxy”). Deze melkwegstelseltjes kunnen we bestuderen vanafde grond met de grootste telescopen op Aarde, en we kunnen bijvoorbeeld nagaanof deze melkwegstelseltjes verschillen van andere melkwegstelsels er vlak bij. Ookkunnen we meten hoe helder het stelsel is in verschillende kleuren licht, waaruit wekunnen bepalen of er veel sterren worden gevormd in het stelsel, en of er veel zwa-re sterren aanwezig zijn. Een andere handige truc is om gebruik te maken van denagloeier zelf: de nagloeier schijnt voor een deel door het melkwegstelseltje heen


waar hij in plaatsvindt. Hierbij absorbeert het materiaal in het stelseltje bepaaldekleuren licht heel efficient, waaruit we heel precies de chemische samenstelling vandat melkwegstelsel kunnen meten. Via dit type onderzoek kunnen we zien dat demelkwegstelsels van lange gammaflitsen allemaal veel jonge sterren en veel zwa-re sterren bevatten en vrij arm zijn aan metalen (sterrenkundigen noemen behalvewaterstof en helium alle elementen metalen). Deze gegevens helpen ons om tebegrijpen welke sterren het zijn die de lange gammaflitsen maken.

De melkwegstelsels waarin korte gammaflitsen plaatsvinden zijn veel diverserdan die van lange gammaflitsen: ze hebben ook veel oude sterren, ze bevatten meermetalen en maken niet veel nieuwe sterren meer.

Waar gaan de hoofdstukken over?

In mijn onderzoek heb ik me bezig gehouden met de connectie tussen de eigen-schappen van de melkwegstelsels waar gammaflitsen in plaatsvinden en de sterrendie gammaflitsen produceren. Hiervoor heb ik vooral veel nagloeiers van gamm-aflitsen waargenomen, vooral in zichtbaar licht, en veel melkwegstelsels.

Hoofdstuk 2

In mijn eerste jaar als promovendus was Swift nog niet gelanceerd, en werden demeeste gammaflits detecties gedaan door de satelliet HETE-2, zo’n 20 per jaar.Van een van die gammaflitsen, 040924, hebben we toen waarnemingen gedaan meteen serie telescopen in Chili, Spanje en de Canarische Eilanden. We hebben denagloeier duidelijk kunnen meten en ook de afstand tot deze gammaflits bepaald.Ook konden we het melkwegstelsel detecteren in diepe opnames. Deze gamm-aflits is interessant omdat zijn duur tussen de lange en korte gammaflitsen in ligt.Via onze analyse komen we tot de conclusie dat de meeste eigenschappen van de-ze gammaflits (nagloeier, supernova en de eigenschappen van het melkwegstelsel)overeenkomen met die van lange gammaflitsen.

Hoofdstuk 3

Op 30 juli 2005 vond er een gammaflits plaats met een heldere nagloeier. Wekonden snel reageren met een grote telescoop op La Palma, en konden daardoor eenmooi spectrum maken. We vonden dat het melkwegstelsel waar deze gammaflitsin gebeurde zeer arm is aan metalen (een factor 100 minder dan de Zon), en dat ervariatie zit in de absorptie van rontgenstraling door het melkwegstelsel. Dit komtmisschien doordat het absorberende materiaal beinvloed wordt door de straling vande gammaflits. De hoeveelheid waterstof in de gezichtslijn naar deze gammaflitstoe was de hoogste ooit gemeten op het moment dat dit artikel werd gepubliceerd.

Samenvatting 175

Hoofdstuk 4

De Nederlands-Italiaanse satelliet BeppoSAX heeft van een aantal gammaflitsende rontgennagloeier kunnen meten. Door die gegevens te combineren met de hel-derheid van de nagloeier in zichtbaar licht, kunnen we nauwkeurig meten wat deinvloed precies is van het melkwegstelselmateriaal dat tussen ons en de gammaflitsinzit. Het optische licht ondervindt het meeste hinder van stof, maar rontgenstralinghet meest van metalen (vooral zuurstof) in de gasfase. We vinden uit deze analysedat de verhouding van gas en stof in deze melkwegstelsels anders is dan in onzeMelkweg, en dat deze stelsels nog het meest lijken op een metaal-arm buur-stelselvan onze Melkweg, de kleine Magelhaense wolk. het belang van dit soort metin-gen is dat we nu beter weten hoe we rekening moeten houden met de effecten vandit materiaal wanneer we eigenschappen van het melkwegstelsel van de gammaflitsafleiden.

Hoofdstuk 5 en 6

Op 6 februari 2006 vond er wederom een gammaflits plaats met een heldere na-gloeier. Binnen twee uur na de gammaflits hebben we hiervan heel gedetailleerdespectra genomen. Hieruit konden we afleiden dat het melkwegstelsels waar dezegammaflits in plaatsvond rijker is in metalen dan we hadden vermoed op basis vanzijn zeer grote afstand. We vermoeden in het spectrum de sporen te zien van water-stofmoleculen, iets dat nog nooit eerder is waargenomen in het melkwegstelsel vaneen gammaflits.

Hoofdstuk 7

Dit hoofdstuk gaat over een gammaflits die plaatsvond op 18 februari 2006. Hetwas direct duidelijk dat dit geen gewone gammaflits was, want de nagloeier gedroegzich anders dan andere nagloeiers. Al snel bleek dat dit een zeer nabije langegammaflits was, en dat de nagloeier in feite de zeer heldere supernova was. Toende supernova het helderste was, hebben we met een zeer grote telescoop in Chilieen spectrum gemaakt met een hele hoge resolutie. Hierin zijn vele elementen tezien van het melkwegstelsel waar de gammaflits in plaatsvond. Deze gegevens hebik gebruikt om een gedetailleerd onderzoek te doen naar dit melkwegstelsel. Hetblijkt dat dit een ongewoon melkwegstelsel is: zeer klein, heel zwak en met eenheel lage hoeveelheid metalen.

Hoofdstuk 8

Soms hebben we pech en kunnen we ondanks een snelle reactie met een grote te-lescoop en lang belichte opnames geen nagloeier vinden in zichtbaar licht. Ditsoort gammaflitsen worden wel donkere gamaflitsen (“dark bursts” in het Engels)


genoemd. De gammaflits waar dit hoofdstuk over gaat heeft een heldere nagloeierin rontgen- en radiostraling, maar niet in zichtbaar licht. Door de helderheid vande rontgen- en radiostraling kunnen we afleiden dat dit waarschijnlijk wordt ver-oorzaakt doordat er te veel stof in het melkwegstelsel in de weg zit. Waar dit stofprecies zit in het stelsel is niet helemaal duidelijk, maar wat wel duidelijk is, is dathet stelsel zelf niet noemenswaardig verschilt van andere melkwegstelsels die langegammaflitsen huisvesten.

In dit proefschrift heb ik van een aantal gammaflitsen de eigenschappen onderzocht,met name gericht op de eigenschappen van hun melkwegstelsels. Hiermee zijn na-tuurlijk de raadsels rond het ontstaan van gammaflitsen nog niet opgelost: hiervoorzijn gedetailleerde gegevens nodig voor vele tientallen tot honderden objecten. Ge-lukkig zal in de komende jaren het aantal gedetecteerde gammaflitsen snel blijventoenemen!

List of publications

Refereed publications

* GRB 051022: physical parameters and extinction of a prototype dark burst,E. Rol, A. J. van der Horst, K. Wiersema, S. K. Patel, A. Levan, M. Ny-sewander, C. Kouveliotou, R. A. M. J. Wijers, N. Tanvir, D. Reichart,A. S. Fruchter, J. Graham, R. L. C. Starling, P. T. O’Brien, J. Hjorth, J. Fyn-bo, P. Jonker, W. van Ham, D. N. Burrows, J.-E. Ovaldsen, A. O. Jaunsen,R. StromApJ accepted, arXiv:0706.1518

* Spectroscopy and multiband photometry of the afterglow of intermediate du-ration γ-ray burst 040924 and its host galaxy,K. Wiersema, A. J. van der Horst, D. A. Kann, E. Rol, R. L. C. Starling,P. A. Curran, J. Gorosabel, A. J. Levan, J. P. U. Fynbo, A. de Ugarte Postigo,R. A. M. J. Wijers, A. J. Castro-Tirado, S. S. Guziy, A. Hornstrup, J. Hjorth,M. Jelınek, B. L. Jensen, M. Kidger, F. Martın-Luis, N. R. Tanvir, P. Tri-stram, P. M. VreeswijkA&A submitted, arXiv:0706.1345

* GRB 060206 and the quandary of achromatic breaks in afterglow light cur-ves,P. A. Curran, A. J. van der Horst, R. A. M. J. Wijers, R. L. C. Starling,A. J. Castro-Tirado, J. P. U. Fynbo, J. Gorosabel, A. S. Jarvinen, D. Male-sani, E. Rol, N. R. Tanvir, K. Wiersema, M. R. Burleigh, S. L. Casewell,P. D. Dobbie, S. Guziy, P. Jakobsson, M. Jelınek, P. Laursen, A. J. Levan,C. G. Mundell, J. Naranen, S. PiranomonteMNRAS accepted, arXiv:0706.1188

* On the nature of the short duration GRB 050906,A. J. Levan, N. R. Tanvir, P. Jakobsson, R. Chapman, J. Hjorth, R. Priddey,J. P. U. Fynbo, K. Hurley, B. Jensen, R. Johnson, J. Gorosabel, A. J. Castro-Tirado, M. Jarvis, E. Ramirez-Ruiz, D. Watson, K. WiersemaMNRAS submitted, arXiv:0705.1705

* A case of mistaken identity? GRB 060912A and the nature of the long – shortGRB divide,A. J. Levan, P. Jakobsson, C. Hurkett, N. R. Tanvir, J. Gorosabel, P. Vrees-wijk, E. Rol, R. Chapman, N. Gehrels, P. T. O’Brien, J. P. Osborne,R. S. Priddey, C. Kouveliotou, R. L. C. Starling, D. Vanden Berk, K. Wier-semaMNRAS 378, 1439, 2007

177

178 Publications

* Spatially resolved properties of the GRB 060505 host: implications for thenature of the progenitor,C. C. Thone, J. P. U. Fynbo, G. Ostlin, B. Milvang-Jensen, K. Wiersema,D. Malesani, D. Della Monica Ferreira, J. Gorosabel, D. A. Kann, D. Watson,M. J. Michalowski, J. Hjorth, A. S. Fruchter, J. SollermanApJ submitted, astro-ph/0703407

* Gamma-Ray Burst afterglows as probes of environment and blastwave phy-sics II: the distribution of p and structure of the circumburst medium,R. L. C. Starling, A. J. van der Horst, E. Rol, R. A. M. J. Wijers, C. Kouve-liotou, K. Wiersema, P. A. Curran, P. WeltevredeApJ accepted, arXiv:0704.3718

* Gamma-Ray Burst afterglows as probes of environment and blastwave phy-sics I: absorption by host galaxy gas and dust,R. L. C. Starling, R. A. M. J. Wijers, K. Wiersema, E. Rol, P. A. Curran, C.Kouveliotou, A. J. van der Horst, M. H. M. HeemskerkApJ 661, 787, 2007

* The nature of the dwarf starforming galaxy associated with GRB 060218 /SN 2006aj,K. Wiersema, S. Savaglio, P. M. Vreeswijk, S. L. Ellison, C. Ledoux, S.-C. Yoon, P. Moller, J. Sollerman, J. P. U. Fynbo, E. Pian, R. L. C. Starling,R. A. M. J. WijersA&A 464, 529, 2007

* H I column densities of z > 2 Swift gamma-ray bursts,P. Jakobsson, J. P. U. Fynbo, C. Ledoux, P. M. Vreeswijk, D. A. Kann,J. Hjorth, N. R. Tanvir, D. Reichart, J. Gorosabel, S. Klose, R. S. Priddey,D. Watson, J. Sollerman, A. S. Fruchter, A. de Ugarte Postigo, K. Wierse-ma, G. Bjornsson, C. C. Thone, K. Pedersen, B. L. JensenA&A 460, L13, 2006

* Probing cosmic chemical evolution with gamma-ray bursts: GRB 060206 atz = 4.048,J. P. U. Fynbo, R. L. C. Starling, C. Ledoux, K. Wiersema, C. C. Thone, J.Sollerman, P. Jakobsson, J. Hjorth, D. Watson, P. M. Vreeswijk, P. Moller, E.Rol, J. Gorosabel, J. Naranen, R. A. M. J. Wijers, G. Bjornsson, J. M. CastroCeron, P. Curran, D. H. Hartmann, S. T. Holland, B. L. Jensen, A. J. Levan,M. Limousin, C. Kouveliotou, G. Nelemans, K. Pedersen, R. S. Priddey, N.R. TanvirA&A, 451, L47, 2006

* A log N(HI) = 22.5 DLA in a dark gamma-ray burst: the environment of GRB050401,

Publications 179

D. Watson, J. P. U. Fynbo, C. Ledoux, P. Vreeswijk, J. Hjorth, A. Smette, A.C. Andersen, K. Aoki, T. Augusteijn, A. P. Beardmore, D. Bersier, J. M. Cas-tro Ceron, P. D’Avanzo, D. Diaz-Fraile, J. Gorosabel, P. Hirst, P. Jakobsson,B. L. Jensen, N. Kawai, G. Kosugi, A. Levan, J. Masegosa, J. Naranen, K. L.Page, K. Pedersen, A. Pozanenko, J. N. Reeves, V. Rumyantsev, T. Shahbaz,D. Sharapov, J. Sollerman, R. L. C. Starling, N. Tanvir, K. Torstensson, K.WiersemaApJ, 652, 1011, 2006

* Gas and dust properties in the afterglow spectra of GRB 050730,R.L.C. Starling, P.M. Vreeswijk, S.L. Ellison, E. Rol, K. Wiersema, A.J.Levan, N.R. Tanvir, R.A.M.J. Wijers, C. Tadhunter, J.R. Zaurin, R.M. Gon-zalez Delgado, C. KouveliotouA&A, 442, 21, 2005

* The search for the host galaxy of the gamma-ray burst GRB 000214,S. Guziy, J. Gorosabel, A.J. Castro-Tirado, A. de Ugarte Postigo, M. Jelınek,M.D. Perez Ramırez, J.M. Castro Ceron, S. Klose, E. Palazzi, K. WiersemaA&A, 441, 975, 2005

* Discovery of the magnetic field in the pulsating B star β Cephei,H. F. Henrichs, J. A. de Jong, E. Verdugo, R. S. Schnerr, C. Neiner, J.-F. Do-nati, C. Catala, S. L. S. Shorlin, G. A. Wade, P. M. Veen, J. S. Nichols,A. Talavera, G. M. Hill, L. Kaper, A. M. Tijani, V. C. Geers, K. Wiersema,B. Plaggenborg, and K. L. J. RyglA&A submitted

* Magnetic field measurements and wind-line variability of OB-type stars,R. S. Schnerr, H. F. Henrichs, C. Neiner, E. Verdugo, J. de Jong, V. C. Geers,K. Wiersema, B. van Dalen, A. Tijani, B. Plaggenborg, and K. L. J. RyglA&A submitted

* Discovery of a magnetic field in the Slowly Pulsating B star zeta Cassiopeiae,C. Neiner, V. C. Geers, H. F. Henrichs, M. Floquet, Y. Fremat, A.-M. Hubert,O. Preuss, K. WiersemaA&A, 406, 1019, 2003

Circulars

* GRB 070612A: Second Epoch WSRT Radio Observations,A. J. van der Horst, R. A. M. J. Wijers, K. Wiersema, E. Rol, GCN6576

* GRB 070612A: Possible WSRT Radio Detection, A. J. van der Horst,R. A. M. J. Wijers, K. Wiersema, E. Rol, GCN 6549

180 Publications

* GRB 070411: break in the optical light curve, D. Malesani, E. Rol, L. A.Antonelli, J. P. U. Fynbo, P. A. Curran, K. Wiersema, A. J. Levan, N. R.Tanvir, V. Testa, E. Palazzi, A. O. Jaunsen, C. C. Thone, J. Hjorth, P. M.Vreeswijk, GCN 6343

* GRB 070223: confirmation of nIR and optical afterglow, E. Rol, N. Tanvir,N. Mirabal, K. Wiersema, J. P. Halpern, A. Levan, R. Chapman, A. Melan-dri & D. Pinfield, GCN 6221

* GRB 060901: Candidate optical afterglow, K. Wiersema, E. Rol & C. C.Thone, GCN 6188

* INT+WFC observations of GRB 061126, E. Rol, K. Wiersema & P. Prema,GCN 5876

* GRB 060908: Detection of a possible host, C. C. Thone, C. Henriksen & K.Wiersema, GCN 5674

* GRB 060807: Detection of a possible host galaxy, C. C. Thone, C. Henriksen& K. Wiersema, GCN 5672

* GRB060906: Danish/DFOSC optical observations, K. Wiersema & E. Rol,GCN 5557

* GRB 060908: Danish/DFOSC optical observations, K. Wiersema, C. C.Thone, E. Rol, GCN 5552

* GRB060901: Optical observations, K. Wiersema & C. C. Thone, GCN5501

* GRB 060708: VLT spectroscopy, P. Jakobsson, P. Vreeswijk, S. Ellison, J.Gorosabel, N. Tanvir, J. P. U. Fynbo, B. L. Jensen, J. Hjorth, D. A. Kann, K.Wiersema, R. A. M. J. Wijers, GCN 5319

* GRB 060319: WHT K-band candidate afterglow, N. R. Tanvir, E. Rol, K.Wiersema, R. Starling, N. O’Mahoney, GCN 4897

* GRB 060218/SN 2006aj: ABT observations, K. Wiersema & F. Nieuwen-hout, GCN 4834

* GRB060203: WHT Z-band observations, E. Rol, A. Levan, K. Wiersema,P. Dobbie, D. Boyce, N. Tanvir, GCN 4680

* GRB050802: Tentative absorption redshift, J. Fynbo, J. Sollerman, B. Jen-sen, J. Hjorth, D. Watson, J. Castro Ceron, P. Moller, T. Augusteijn, J. Goro-sabel, K. Wiersema, GCN 3749

Publications 181

* GRB050730: confirmation of redshift, E. Rol, R. Starling, K. Wiersema,P. Vreeswijk, A. Levan, N. O’Mahony, C. Tadhunter, J. Rodriguez, R. M.Gonzalez Delgado, GCN 3710.

* GRB 050724: WHT optical observations, K. Wiersema, E. Rol, R. Starling,N. Tanvir, D. S. Bloomfield, H. Thompson, GCN 3699.

* GRB 050509B: WSRT Radio Observations, A. J. van der Horst, K. Wierse-ma, R. A. M. J. Wijers, GCN 3405.

* GRB 050502A: WSRT Radio Observations, A. J. van der Horst, R. A. M. J.Wijers, K. Wiersema, GCN 3341.

* GRB 050408: Mercator Optical Observations, P. Curran, K. Wiersema, K.Lefever, H. van Winckel, C. Waelkens, O. van Braam, Y. Grange, R. deRooij, A. de Vries, L. Waters, G. Bourban, G. Burki, F. Carrier, E. Rol, GCN3211.

* GRB 050408: Mercator Optical Observations, K. Wiersema, P. Curran, K.Lefever, H. van Winckel, C. Waelkens, O. van Braam, Y. Grange, R. deRooij, A. de Vries, L. Waters, GCN 3200.

* GRB 050401: VLT spectroscopic redshift, J. Fynbo, B. Jensen, J. Hjorth,K. Wiersema, R. Starling, P. Vreeswijk, E. Rol, A. Levan, S. Ellison, N.Masetti, GCN 3176.

* GRB 040924: VLT spectroscopy, K. Wiersema, R. Starling, E. Rol, P.Vreeswijk, R. Wijers, GCN 2800.

Non-refereed publications

* Gamma-ray Burst host galaxy gas and dust,R.L.C. Starling, R.A.M.J. Wijers and K. Wiersema,Proceedings of the Eleventh Marcel Grossmann Meeting on General Rela-tivity, held July 2006 in Berlin, edited by H. Kleinert, R.T. Jantzen and R.Ruffini, World Scientific, Singapore, 2007

* Spectroscopy of the optical afterglow of GRB 021004: origin of the blueshif-ted hydrogen lines,R.L.C. Starling, R.A.M.J. Wijers, M.A. Hughes, N.R. Tanvir, P.M. Vrees-wijk, E. Rol, K. Wiersema and I. Salamanca 2005, in the proceedings of’Gamma-ray Bursts in the Afterglow Era: 4th Workshop’, Rome, Eds L.Piro, L. Amati, S. Covino and B. Gendre

182 Publications

* Searching for Magnetic fields in A type Supergiants,E. Verdugo, A. Talavera, A. I. Gomez de Castro, H. F. Henrichs, V. C. Geers,K. Wiersema,EAS, 9, 271, 2003

* Magnetic Fields and Winds in A-type Supergiants,E. Verdugo, A. Talavera, A. I. Gomez de Castro, H. F. Henrichs, V. C. Geers,K. Wiersema,ASPC, 305, 364V, 2003

* The Short Gamma-Ray Burst Revolution,J. Hjorth, A. Levan, N. Tanvir, R. Starling, S. Klose, C. Kouveliotou, C. Fe-ron, P. Ferrero, A. Fruchter, J. Fynbo, J. Gorosabel, P. Jakobsson, D. Kann,K. Pedersen, E. Ramirez-Ruiz, J. Sollerman, C. Thone, D. Watson, K. Wier-sema, D. XuESO Messenger 126, 16

Thank you!

The very first person I would like to thank is my esteemed promotor RalphWijers. Ralph, you were an amazing PhD supervisor, offering exactly the rightmix of independence and supervision, introducing me to the wonderful world ofgamma-ray bursts. The direction in which my research developed was somewhatdifferent from what you originally had in mind, but we did good science. I am veryglad you offered me this job!I am grateful to the members of the (reading) committee for taking the time to readmy thesis and give very useful comments.Most of my scientific work involves operating in large scientific collaborations.It takes a while to establish your position in such large consortia, but I foundworking in these large groups a nice experience (it sure involves writing a hugeamount of proposals!). Therefore I would like say a big thank you to all co-authorson papers and all co-I’s on proposals for all the nice collaborations. In particularI would like to say thanks to Evert Rol, Nial Tanvir, Johan Fynbo, Jens Hjorth,Cedric Ledoux, Paul Vreeswijk, Sandra Savaglio, Andrew Levan, Christina Thone,Daniele Malesani, Elena Pian, Sara Ellison, Javier Gorosabel, Jesper Sollerman,Sung-Chul Yoon, Alex Kann, and many many more people. I also had theopportunity to travel during my PhD, and I’d like to thank Jens Hjorth for invitingme to Copenhagen, Julian Osborne for having me over in Leicester and JohanFynbo for giving me the opportunity to go observing at La Silla.

I think it is important to see the broader picture in astronomical research andexplore the “boundaries” between different fields of research, applying thingsyou learn in one field of research to another field. It’s all astronomy after all. Ihave been fortunate enough to be able to participate in several projects/papers thatdo not involve GRBs, but different (sometimes even non-exploding!) objects. Iwould like to warmly thank the people that have invited me to join their research,particularly Huib, Roald, Marc, Rudy, Dipankar, PG and Dave.

The Astronomical Institute of the UvA is a very unique place. I have never seen aplace where the atmosphere is so good, which is evidently reflected in the qualityof the science done. So a big thanks to all my friends at the API! I have been atthe institute for a long time, so I have seen many people come and go. ThereforeI’m more than likely to forget a few names, but since it’s always nice to see yourname in someone’s thesis, I’m going to attempt to make a list of names of thepeople most often at the East and other social events. So thank you (in randomorder): Evert, Alexander (Big Al), Martin (Smartie), Patrick, Peter, Gemma &Cees, James (Jimbo), Phil (Dr.), Casey (the Keessies), John (the Ripper), Eduardo,Joe, Nathalie, Marc, Arjan, Arjen, Arjan ((dancing/singing) Dikkiedik), Annique,

183

184 Thank you!

Wing-Fai (the french spy), Rob, Nick & Carlijn (nickienick), Roy, Thomas, Stefan,Ammar, Tom (macaroni), Onno (onnionio), Diego, Manu, PG, Dipankar (1,2,3),Rudy, Elena, Simone, Rien (the Rienenator), Rohied (Smith), Brechtje, Michiel(ieniemienie), Jason, Atakan (Ataturk), Dave (djplomme), Dave, Thijs, Joke, ThePolish (Anna 0 - 3), Kazi, Vincent G, Stratos (stratovarius) Lianne, Asaf (Pavarotti)& Keren, Roald, Sera, Mike, Alessia, Valeriu, Alessandro, Evghenii, Hendrik,Hylke, Mieke, Alex (the king of gossip), Carsten, Ben and of course Sinterklaas,who every year came to the Institute with a load of interesting surprises...I will never forget the great times I had at the koffietafel, cola-break and in the East/ Brouwerij ’t IJ / Gollem and the lessons I learned about Rule #1, sharing, no-flyand other valuable facts of life.And of course the Institute would grind to a halt without Minou, Eva, Fieke,Annemiek, Nicole and Lide. A special thanks also to Ton, for always beinginterested in the night sky, and ready to have a chat about that.

Ondanks het feit dat we de Academische Jaarprijs niet hebben gewonnen, wil ikhet Einsteinflits-team enorm bedanken voor de geweldige tijd. Eigenlijk haddenwe die 100.000 gewoon moeten krijgen!!

Ik begon serieus interesse in natuur- en sterrenkunde te krijgen in de lagereklassen van de middelbare school, en daarom wil ik heel graag de leden van M 111bedanken, maar vooral Willem-Jan Braakman. Ik zal nooit de waarneemavonden(bijvoorbeeld voor Ikeya-Zhang, Hyakutake, Hale-Bopp en talloze maansverduis-teringen) en raketlanceringen vergeten. Ook wil ik graag de leden van AWSVMetius te Alkmaar bedanken, en dan met name de leden van de ABT groep, dewerkgroep waarnemingen en theoretische sterrenkunde. Ik heb enorm veel vanjullie geleerd, en hoop nog vele jaren actief lid te zijn!

A special word of thanks for all my friends who do not share so much in myenthusiasm for astronomy. Thanks for putting up with my nerdiness!A special thanks to Pier, Thijs, Saskia and Evert.

Tenslotte wil ik hier graag mijn familie en ouders bedanken. Pap en mam: bedanktdat jullie altijd mijn interesse in natuur- en sterrenkunde hebben aangemoedigd, ikben jullie heel erg dankbaar. Superbedankt!! I am also very grateful to my familyin the UK, who have always made me feel so much at home. Thank you so muchfor your wonderful support!

I would like to finish these last lines of my thesis by thanking you, my sweet darlingRhaana, for your endless love and all the wonderful support you give me. You arethe most brilliant star in the Universe!

Bibliography

Adelberger, K. L., Shapley, A. E., Steidel, C. C., et al. 2005, ApJ, 629, 636Akerman, C. J., Ellison, S. L., Pettini, M., & Steidel, C. C. 2005, A&A, 440, 499Alard, C. 2000, A&AS, 144, 363Alard, C. & Lupton, R. H. 1998, ApJ, 503, 325Allende Prieto, C., Lambert, D. L., & Asplund, M. 2001, ApJ, 556, L63Amati, L. 2006, MNRAS, 372, 233Amati, L., Frontera, F., Tavani, M., et al. 2002, A&A, 390, 81Anders, E. & Grevesse, N. 1989, Geochimica et Cosmochimica Acta, 53, 197Andersen, M. I., Hjorth, J., Pedersen, H., et al. 2000, A&A, 364, L54Arnaud, K. A. 1996, in Astronomical Society of the Pacific Conference Series, Vol. 101,

Astronomical Data Analysis Software and Systems V, ed. G. H. Jacoby & J. Barnes,17

Arsenault, R. & Roy, J.-R. 1986, AJ, 92, 567Bahcall, J. N. & Wolf, R. A. 1968, ApJ, 152, 701Band, D., Matteson, J., Ford, L., et al. 1993, ApJ, 413, 281Band, D. L. 2006, ApJ, 644, 378Barnard, V. E., Blain, A. W., Tanvir, N. R., et al. 2003, MNRAS, 338, 1Barraud, C., Olive, J.-F., Lestrade, J. P., et al. 2003, A&A, 400, 1021Barthelmy, S. D., Chincarini, G., Burrows, D. N., et al. 2005, Nature, 438, 994Berger, E., Cowie, L. L., Kulkarni, S. R., et al. 2003a, ApJ, 588, 99Berger, E., Fox, D. B., Price, P. A., et al. 2006a, ArXiv Astrophysics e-printsBerger, E., Kulkarni, S. R., Fox, D. B., et al. 2005a, ApJ, 634, 501Berger, E., Kulkarni, S. R., Pooley, G., et al. 2003b, Nature, 426, 154Berger, E., Penprase, B. E., Cenko, S. B., et al. 2006b, ApJ, 642, 979Berger, E., Price, P. A., Cenko, S. B., et al. 2005b, Nature, 438, 988Berger, E. & Wyatt, P. 2005, GCN Circular, 4148Bloom, J. S., Berger, E., Kulkarni, S. R., Djorgovski, S. G., & Frail, D. A. 2003, AJ,

125, 999Bloom, J. S., Djorgovski, S. G., & Kulkarni, S. R. 2001a, ApJ, 554, 678Bloom, J. S., Djorgovski, S. G., Kulkarni, S. R., & Frail, D. A. 1998a, ApJ, 507, L25Bloom, J. S., Frail, D. A., Kulkarni, S. R., et al. 1998b, ApJ, 508, L21Bloom, J. S., Frail, D. A., & Sari, R. 2001b, AJ, 121, 2879Bloom, J. S., Kulkarni, S. R., & Djorgovski, S. G. 2002, AJ, 123, 1111Blustin, A. J., Band, D., Barthelmy, S., et al. 2006, ApJ, 637, 901Blustin, A. J., Holland, S. T., Cucchiara, A., White, N., & Hinshaw, D. 2005, GCN

Circular, 3717Boer, M., Atteia, J. L., Damerdji, Y., et al. 2006, ApJ, 638, L71Boksenberg, A., Carswell, R. F., Smith, M. G., & Whelan, J. A. J. 1978, MNRAS, 184,

773Boksenberg, A., Danziger, I. J., Fosbury, R. A. E., & Goss, W. M. 1980, ApJ, 242, L145Boksenberg, A., Sargent, W. L. W., & Rauch, M. 2003, ArXiv Astrophysics e-prints

185

186 Bibliography

Bolzonella, M., Miralles, J.-M., & Pello, R. 2000, A&A, 363, 476Bond, N. A., Churchill, C. W., Charlton, J. C., & Vogt, S. S. 2001, ApJ, 562, 641Borkowski, K. J., Williams, B. J., Reynolds, S. P., et al. 2006, ApJ, 642, L141Bouche, N., Murphy, M. T., Peroux, C., Csabai, I., & Wild, V. 2006, MNRAS, 371, 495Bouchet, P., Lequeux, J., Maurice, E., Prevot, L., & Prevot-Burnichon, M. L. 1985,

A&A, 149, 330Boyd, P., Morris, D., Marshall, F., & Gehrels, N. 2006, GCN Circular, 4722Bremer, M., Castro-Tirado, A. J., & Neri, R. 2005, GCN Circular, 4157Bresolin, F., Garnett, D. R., & Kennicutt, Jr., R. C. 2004, ApJ, 615, 228Bromm, V. & Loeb, A. 2006, ApJ, 642, 382Burrows, D. N., Romano, P., Falcone, A., et al. 2005, Science, 309, 1833Butler, N. R., Ricker, G. R., Lamb, D. Q., et al. 2005, GCN Circular, 4165Calzetti, D. 2001, New Astronomy Review, 45, 601Calzetti, D., Kinney, A. L., & Storchi-Bergmann, T. 1994, ApJ, 429, 582Cameron, P. B. & Frail, D. A. 2005, GCN Circular, 4154Campana, S., Mangano, V., Blustin, A. J., et al. 2006a, Nature, 442, 1008Campana, S., Romano, P., Covino, S., et al. 2006b, A&A, 449, 61Cantiello, M., Yoon, S.-C., Langer, N., & Livio, M. 2007, A&A, 465, L29Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245Castro, S., Galama, T. J., Harrison, F. A., et al. 2003, ApJ, 586, 128Castro Ceron, J. M., Michałowski, M. J., Hjorth, J., et al. 2006, ApJ, 653, L85Castro-Tirado, A. J., McBreen, S., Jelınek, M., et al. 2006, in American Institute of

Physics Conference Series, Vol. 836, Gamma-Ray Bursts in the Swift Era, ed. S. S.Holt, N. Gehrels, & J. A. Nousek, 79–84

Castro-Tirado, A. J., Zapatero-Osorio, M. R., Gorosabel, J., et al. 1999, ApJ, 511, L85Caulet, A. 1996, in Science with the Hubble Space Telescope - II, ed. P. Benvenuti, F. D.

Macchetto, & E. J. Schreier, 481Chary, R., Becklin, E. E., & Armus, L. 2002, ApJ, 566, 229Chary, R., Neugebauer, G., Morris, M., et al. 1998, ApJ, 498, L9+Chen, H.-W., Prochaska, J. X., & Bloom, J. S. 2006a, in American Institute of Physics

Conference Series, Vol. 836, Gamma-Ray Bursts in the Swift Era, ed. S. S. Holt,N. Gehrels, & J. A. Nousek, 534–539

Chen, H.-W., Prochaska, J. X., Bloom, J. S., & Thompson, I. B. 2005a, ApJ, 634, L25Chen, H.-W., Prochaska, J. X., Ramirez-Ruiz, E., et al. 2006b, ArXiv Astrophysics e-

printsChen, H.-W., Thompson, I., Prochaska, J. X., & Bloom, J. 2005b, GCN Circular, 3709Christensen, L., Hjorth, J., & Gorosabel, J. 2004, A&A, 425, 913Christensen, L., Hjorth, J., & Gorosabel, J. 2005, ApJ, 631, L29Churchill, C. W., Vogt, S. S., & Charlton, J. C. 2003, AJ, 125, 98Ciardi, B. & Loeb, A. 2000, ApJ, 540, 687Cobb, B. E. & Bailyn, C. D. 2005, GCN Circular, 3708Condon, J. J. 1992, ARA&A, 30, 575Conselice, C. J., Vreeswijk, P. M., Fruchter, A. S., et al. 2005, ApJ, 633, 29Cordes, J. M. & Lazio, T. J. W. 2002, ArXiv Astrophysics e-prints

Bibliography 187

Corsi, A., Piro, L., Kuulkers, E., et al. 2005, A&A, 438, 829Costa, E., Frontera, F., Heise, J., et al. 1997, Nature, 387, 783Courty, S., Bjornsson, G., & Gudmundsson, E. H. 2004, MNRAS, 354, 581Covino, S., Malesani, D., Israel, G. L., et al. 2006, A&A, 447, L5Cowsik, R., Prabhu, T. P., Anupama, G. C., et al. 2001, Bulletin of the Astronomical

Society of India, 29, 157Crowther, P. A. & Hadfield, L. J. 2006, A&A, 449, 711Curran, P. A., van der Horst, A. J., Beardmore, A. P., et al. 2007a, A&A, 467, 1049Curran, P. A., van der Horst, A. J., Wijers, R. A. M. J., et al. 2007b, ArXiv e-prints, 706Daigne, F., Rossi, E. M., & Mochkovitch, R. 2006, MNRAS, 372, 1034Dekker, H., D’Odorico, S., Kaufer, A., Delabre, B., & Kotzlowski, H. 2000, in Presented

at the Society of Photo-Optical Instrumentation Engineers (SPIE) Conference, Vol.4008, Proc. SPIE Vol. 4008, p. 534-545, Optical and IR Telescope Instrumentationand Detectors, Masanori Iye; Alan F. Moorwood; Eds., ed. M. Iye & A. F. Moorwood,534–545

D’Elia, V., Fiore, F., Meurs, E. J. A., et al. 2007, A&A, 467, 629Della Valle, M., Chincarini, G., Panagia, N., et al. 2006, Nature, 444, 1050Dessauges-Zavadsky, M., Chen, H.-W., Prochaska, J. X., Bloom, J. S., & Barth, A. J.

2006, ApJ, 648, L89Dewi, J. D. M., Podsiadlowski, P., & Pols, O. R. 2005, MNRAS, 363, L71Dickey, J. M. & Lockman, F. J. 1990, ARA&A, 28, 215Diercks, A., Deutsch, E. W., Wyse, R., et al. 1997, IAU Circ., 6791, 1Diplas, A. & Savage, B. D. 1994, ApJ, 427, 274Djorgovski, S. G., Bloom, J. S., & Kulkarni, S. R. 2003, ApJ, 591, L13Djorgovski, S. G., Frail, D. A., Kulkarni, S. R., et al. 2001, ApJ, 562, 654Djorgovski, S. G., Kulkarni, S. R., Bloom, J. S., et al. 1998, ApJ, 508, L17Donaghy, T. Q., Lamb, D. Q., Sakamoto, T., et al. 2006, ArXiv Astrophysics e-printsDraine, B. T. & Hao, L. 2002, ApJ, 569, 780Draine, B. T. & Lee, H. M. 1984, ApJ, 285, 89Draine, B. T. & Salpeter, E. E. 1979, ApJ, 231, 438Ellison, S. L. 2006, MNRAS, 368, 335Ellison, S. L., Hall, P. B., & Lira, P. 2005, AJ, 130, 1345Ellison, S. L. & Kewley, L. J. 2006, in Proceedings contribution for ”The Fabulous

Destiny of Galaxies; Bridging the Past and Present”Ellison, S. L., Mallen-Ornelas, G., & Sawicki, M. 2003, ApJ, 589, 709Ellison, S. L., Vreeswijk, P., Ledoux, C., et al. 2006, MNRAS, 372, L38Evans, P. A., Beardmore, A. P., Page, K. L., et al. 2007, A&A, 469, 379Fan, Y. Z., Zhang, B., Kobayashi, S., & Meszaros, P. 2005, ApJ, 628, 867Fenimore, E. E., Ricker, G., Atteia, J.-L., et al. 2004, GCN Circular, 2735Fernandes, I. F., de Carvalho, R., Contini, T., & Gal, R. R. 2004, MNRAS, 355, 728Ferrara, A., Scannapieco, E., & Bergeron, J. 2005, ApJ, 634, L37Ferrero, P., Kann, D. A., Zeh, A., et al. 2006, A&A, 457, 857Fiore, F., D’Elia, V., Lazzati, D., et al. 2005, ApJ, 624, 853Fitzpatrick, E. L. 1985, ApJ, 299, 219

188 Bibliography

Fontana, A. & Ballester, P. 1995, ESO Messenger, 80, 37Fox, D. B., Frail, D. A., Price, P. A., et al. 2005, Nature, 437, 845Fox, D. B. & Moon, D.-S. 2004, GCN Circular, 2734Frail, D. A., Kulkarni, S. R., Nicastro, L., Feroci, M., & Taylor, G. B. 1997, Nature,

389, 261Fruchter, A., Krolik, J. H., & Rhoads, J. E. 2001, ApJ, 563, 597Fruchter, A. S., Levan, A. J., Strolger, L., et al. 2006, Nature, 441, 463Fruchter, A. S., Thorsett, S. E., Metzger, M. R., et al. 1999, ApJ, 519, L13Fryer, C. L. & Heger, A. 2005, ApJ, 623, 302Fryer, C. L., Mazzali, P. A., Prochaska, J., et al. 2007, ArXiv Astrophysics e-printsFukugita, M., Shimasaku, K., & Ichikawa, T. 1995, PASP, 107, 945Fynbo, J. P. U., Jakobsson, P., Moller, P., et al. 2003, A&A, 406, L63Fynbo, J. P. U., Starling, R. L. C., Ledoux, C., et al. 2006a, A&A, 451, L47Fynbo, J. P. U., Watson, D., Thone, C. C., et al. 2006b, Nature, 444, 1047Fynbo, J. U., Gorosabel, J., Dall, T. H., et al. 2001a, A&A, 373, 796Fynbo, J. U., Jensen, B. L., Gorosabel, J., et al. 2001b, A&A, 369, 373Gaensler, B. M., Kouveliotou, C., Gelfand, J. D., et al. 2005, Nature, 434, 1104Gal-Yam, A., Berger, E., Fox, D. B., et al. 2005, GCN Circular, 4156Gal-Yam, A., Fox, D. B., Price, P. A., et al. 2006, Nature, 444, 1053Galama, T. J., Briggs, M. S., Wijers, R. A. M. J., et al. 1999, Nature, 398, 394Galama, T. J., Groot, P. J., van Paradijs, J., et al. 1998a, ApJ, 497, L13+Galama, T. J., Tanvir, N., Vreeswijk, P. M., et al. 2000, ApJ, 536, 185Galama, T. J., Vreeswijk, P. M., van Paradijs, J., et al. 1998b, Nature, 395, 670Galama, T. J. & Wijers, R. A. M. J. 2001, ApJ, 549, L209Garnett, D. R. 2002, ApJ, 581, 1019Gehrels, N., Chincarini, G., Giommi, P., et al. 2004, ApJ, 611, 1005Gehrels, N., Sarazin, C. L., O’Brien, P. T., et al. 2005, Nature, 437, 851Gendre, B. & Boer, M. 2005, A&A, 430, 465Gendre, B., Galli, A., Corsi, A., et al. 2007, A&A, 462, 565Ghirlanda, G., Ghisellini, G., & Lazzati, D. 2004, ApJ, 616, 331Golenetskii, S., Aptekar, R., Mazets, E., et al. 2004, GCN Circular, 2754Golenetskii, S., Aptekar, R., Mazets, E., et al. 2005, GCN Circular, 4150Gorosabel, J., Christensen, L., Hjorth, J., et al. 2003a, A&A, 400, 127Gorosabel, J., Klose, S., Christensen, L., et al. 2003b, A&A, 409, 123Gorosabel, J., Perez-Ramırez, D., Sollerman, J., et al. 2005, A&A, 444, 711Greiner, J., Klose, S., Reinsch, K., et al. 2003, Nature, 426, 157Grevesse, N. & Sauval, A. J. 1998, Space Science Reviews, 85, 161Groot, P. J., Galama, T. J., van Paradijs, J., et al. 1998, ApJ, 493, L27+Grupe, D., Gronwall, C., Wang, X.-Y., et al. 2006, ArXiv Astrophysics e-printsGuenther, E. W., Klose, S., Vreeswijk, P., Pian, E., & Greiner, J. 2006, GCN Circular,

4863Gunn, J. E. & Peterson, B. A. 1965, ApJ, 142, 1633Haislip, J., Kirschbrown, J., Reichart, D., et al. 2005, GCN Circular, 3712Haislip, J. B., Nysewander, M. C., Reichart, D. E., et al. 2006, Nature, 440, 181

Bibliography 189

Hakkila, J., Giblin, T. W., Roiger, R. J., et al. 2003, ApJ, 582, 320Halpern, J. P., Thorstensen, J. R., Helfand, D. J., & Costa, E. 1998, Nature, 393, 41Hammer, F., Flores, H., Schaerer, D., et al. 2006, A&A, 454, 103Hamuy, M., Suntzeff, N. B., Heathcote, S. R., et al. 1994, PASP, 106, 566Hamuy, M., Walker, A. R., Suntzeff, N. B., et al. 1992, PASP, 104, 533Hao, H., Stanek, K. Z., Dobrzycki, A., et al. 2007, ApJ, 659, L99Harrison, F. A., Bloom, J. S., Frail, D. A., et al. 1999, ApJ, 523, L121Harrison, F. A., Yost, S. A., Sari, R., et al. 2001, ApJ, 559, 123Heckman, T. M. 2001, in Astronomical Society of the Pacific Conference Series, Vol.

240, Gas and Galaxy Evolution, ed. J. E. Hibbard, M. Rupen, & J. H. van Gorkom,345

Heger, A., Fryer, C. L., Woosley, S. E., Langer, N., & Hartmann, D. H. 2003, ApJ, 591,288

Heger, A., Langer, N., & Woosley, S. E. 2000, ApJ, 528, 368Henden, A. 2004, GCN Circular, 2811Henden, A. 2005, GCN Circular, 4184Hirschi, R., Meynet, G., & Maeder, A. 2005, A&A, 443, 581Hjorth, J., Levan, A., Tanvir, N., et al. 2006, The Messenger, 126, 16Hjorth, J., Møller, P., Gorosabel, J., et al. 2003a, ApJ, 597, 699Hjorth, J., Sollerman, J., Møller, P., et al. 2003b, Nature, 423, 847Hjorth, J., Watson, D., Fynbo, J. P. U., et al. 2005, Nature, 437, 859Holland, S., Fynbo, J. P. U., Hjorth, J., et al. 2001, A&A, 371, 52Holland, S. T., Barthelmy, S., Burrows, D. N., et al. 2005, GCN Circular, 3704Holman, M., Garnavich, P., & Stanek, K. Z. 2005, GCN Circular, 3727Horvath, I., Balazs, L. G., Bagoly, Z., Ryde, F., & Meszaros, A. 2006, A&A, 447, 23Houck, J. C. & Denicola, L. A. 2000, in Astronomical Society of the Pacific Confer-

ence Series, Vol. 216, Astronomical Data Analysis Software and Systems IX, ed.N. Manset, C. Veillet, & D. Crabtree, 591

Huang, K. Y., Urata, Y., Filippenko, A. V., et al. 2005, ApJ, 628, L93Hunter, I., Smoker, J. V., Keenan, F. P., et al. 2006, MNRAS, 367, 1478in ’t Zand, J. J. M., Amati, L., Antonelli, L. A., et al. 1998, ApJ, 505, L119Izotov, Y. I., Chaffee, F. H., Foltz, C. B., et al. 1999, ApJ, 527, 757Izotov, Y. I., Papaderos, P., Guseva, N. G., Fricke, K. J., & Thuan, T. X. 2006a, A&A,

454, 137Izotov, Y. I., Stasinska, G., Meynet, G., Guseva, N. G., & Thuan, T. X. 2006b, A&A,

448, 955Izotov, Y. I., Thuan, T. X., & Lipovetsky, V. A. 1994, ApJ, 435, 647Jakobsson, P., Bjornsson, G., Fynbo, J. P. U., et al. 2005a, MNRAS, 362, 245Jakobsson, P., Bjornsson, G., Fynbo, J. P. U., et al. 2005b, MNRAS, 362, 245Jakobsson, P., Frail, D. A., Fox, D. B., et al. 2005c, ApJ, 629, 45Jakobsson, P., Fynbo, J. P. U., Ledoux, C., et al. 2006a, A&A, 460, L13Jakobsson, P., Hjorth, J., Fynbo, J. P. U., et al. 2004, ApJ, 617, L21Jakobsson, P., Levan, A., Fynbo, J. P. U., et al. 2006b, A&A, 447, 897Jaunsen, A. O., Andersen, M. I., Hjorth, J., et al. 2003, A&A, 402, 125

190 Bibliography

Jaunsen, A. O., Hjorth, J., Bjornsson, G., et al. 2001, ApJ, 546, 127Jensen, B. L., Fynbo, J. U., Gorosabel, J., et al. 2001, A&A, 370, 909Jester, S., Schneider, D. P., Richards, G. T., et al. 2005, AJ, 130, 873Jha, S., Pahre, M. A., Garnavich, P. M., et al. 2001, ApJ, 554, L155Jones, A. P. 2004, in Astronomical Society of the Pacific Conference Series, Vol. 309,

Astrophysics of Dust, ed. A. N. Witt, G. C. Clayton, & B. T. Draine, 347Kalberla, P. M. W., Burton, W. B., Hartmann, D., et al. 2005, A&A, 440, 775Kaneko, Y., Ramirez-Ruiz, E., Granot, J., et al. 2007, ApJ, 654, 385Kann, D. A., Klose, S., & Zeh, A. 2006, ApJ, 641, 993Kauffmann, G., Heckman, T. M., Tremonti, C., et al. 2003, MNRAS, 346, 1055Kawai, N., Kosugi, G., Aoki, K., et al. 2006, Nature, 440, 184Kawai, N., Yamada, T., Kosugi, G., Hattori, T., & Aoki, K. 2005, GCN Circular, 3937Kennicutt, Jr., R. C. 1992, ApJ, 388, 310Kennicutt, Jr., R. C. 1998, ARA&A, 36, 189Kewley, L. J., Brown, W. R., Geller, M. J., Kenyon, S. J., & Kurtz, M. J. 2007, AJ, 133,

882Kewley, L. J. & Dopita, M. A. 2002, ApJS, 142, 35Khamitov, I., Aslan, Z., Yesilyaprak, C., et al. 2004, GCN Circular, 2749King, A., O’Brien, P. T., Goad, M. R., et al. 2005, ApJ, 630, L113Klebesadel, R. W., Strong, I. B., & Olson, R. A. 1973, ApJ, 182, L85Kobulnicky, H. A. & Kewley, L. J. 2004, ApJ, 617, 240Kobulnicky, H. A., Willmer, C. N. A., Phillips, A. C., et al. 2003, ApJ, 599, 1006Kondo, S., Kobayashi, N., Minowa, Y., et al. 2006, ApJ, 643, 667Koornneef, J. 1982, A&A, 107, 247Kouveliotou, C., Meegan, C. A., Fishman, G. J., et al. 1993, ApJ, 413, L101Kulkarni, S. R., Djorgovski, S. G., Odewahn, S. C., et al. 1999a, Nature, 398, 389Kulkarni, S. R., Djorgovski, S. G., Ramaprakash, A. N., et al. 1998, Nature, 393, 35Kulkarni, S. R., Frail, D. A., Sari, R., et al. 1999b, ApJ, 522, L97Kupcu Yoldas, A., Greiner, J., & Perna, R. 2006, A&A, 457, 115Kuulkers, E., Antonelli, L. A., Kuiper, L., et al. 2000, ApJ, 538, 638Lamareille, F., Mouhcine, M., Contini, T., Lewis, I., & Maddox, S. 2004, MNRAS, 350,

396Langer, N. & Norman, C. A. 2006, ApJ, 638, L63Le Floc’h, E., Charmandaris, V., Forrest, W. J., et al. 2006, ApJ, 642, 636Le Floc’h, E., Duc, P.-A., Mirabel, I. F., et al. 2003, A&A, 400, 499Ledoux, C., Petitjean, P., Fynbo, J. P. U., Møller, P., & Srianand, R. 2006a, A&A, 457,

71Ledoux, C., Petitjean, P., & Srianand, R. 2003, MNRAS, 346, 209Ledoux, C., Petitjean, P., & Srianand, R. 2006b, ApJ, 640, L25Ledoux, C., Vreeswijk, P., Ellison, S., et al. 2005, GCN Circular, 3860Lee, B. C., Tucker, D. L., Vanden Berk, D. E., et al. 2001, ApJ, 561, 183Lee, H., Skillman, E. D., Cannon, J. M., et al. 2006a, ApJ, 647, 970Lee, H., Skillman, E. D., & Venn, K. A. 2006b, ApJ, 642, 813Lee, W. H. & Ramirez-Ruiz, E. 2007, New Journal of Physics, 9, 17

Bibliography 191

Levan, A., Fruchter, A., Rhoads, J., et al. 2006, ApJ, 647, 471Levan, A. J., Jakobsson, P., Hurkett, C., et al. 2007, MNRAS, 378, 1439Lin, L., Koo, D. C., Weiner, B. J., et al. 2007, ApJ, 660, L51Lodders, K. 2003, ApJ, 591, 1220Lopez, S. & Ellison, S. L. 2003, A&A, 403, 573MacFadyen, A. I. & Woosley, S. E. 1999, ApJ, 524, 262Madau, P. 1995, ApJ, 441, 18Madau, P., Ferguson, H. C., Dickinson, M. E., et al. 1996, MNRAS, 283, 1388Maeda, K., Kawabata, K., Tanaka, M., et al. 2007, ApJ, 658, L5Maeder, A. & Meynet, G. 2005, A&A, 440, 1041Maiorano, E., Masetti, N., Palazzi, E., et al. 2005, A&A, 438, 821Malacrino, F., Atteia, J.-L., Boer, M., et al. 2007, A&A, 464, L29Martin, N., Maurice, E., & Lequeux, J. 1989, A&A, 215, 219Masetti, N., Palazzi, E., Pian, E., et al. 2001, A&A, 374, 382Maund, J. R., Smartt, S. J., & Danziger, I. J. 2005, MNRAS, 364, L33Mazzali, P. A., Deng, J., Nomoto, K., et al. 2006, Nature, 442, 1018Melnick, J., Tenorio-Tagle, G., & Terlevich, R. 1999, MNRAS, 302, 677Meszaros, P. & Rees, M. J. 1992, MNRAS, 257, 29PMeszaros, P. & Rees, M. J. 1997, ApJ, 476, 232Meszaros, P. & Rees, M. J. 1999, MNRAS, 306, L39Metzger, M. R., Djorgovski, S. G., Kulkarni, S. R., et al. 1997, Nature, 387, 878Meurer, G. R., Heckman, T. M., & Calzetti, D. 1999, ApJ, 521, 64Meynet, G. & Maeder, A. 2005, A&A, 429, 581Mirabal, N. & Halpern, J. P. 2006, GCN Circular, 4792Mirabal, N., Halpern, J. P., An, D., Thorstensen, J. R., & Terndrup, D. M. 2006, ApJ,

643, L99Mirabal, N., Halpern, J. P., Kulkarni, S. R., et al. 2002, ApJ, 578, 818Modjaz, M., Kewley, L., Kirshner, R. P., et al. 2007, ArXiv Astrophysics e-printsModjaz, M., Stanek, K. Z., Garnavich, P. M., et al. 2006, ApJ, 645, L21Mokiem, M. R., de Koter, A., Evans, C. J., et al. 2007a, ArXiv e-prints, 704Mokiem, M. R., de Koter, A., Evans, C. J., et al. 2007b, A&A, 465, 1003Mokiem, M. R., de Koter, A., Evans, C. J., et al. 2006, A&A, 456, 1131Moller, P., Fynbo, J. P. U., & Fall, S. M. 2004, A&A, 422, L33Monfardini, A., Kobayashi, S., Guidorzi, C., et al. 2006, ApJ, 648, 1125Morris, D., Barbier, L., Barthelmy, S., et al. 2006a, GCN Circular, 4682Morris, D., Burrows, D., Gehrels, N., Boyd, P., & Angelini, L. 2006b, GCN Circular,

4689Moustakas, J., Kennicutt, Jr., R. C., & Tremonti, C. A. 2006, ApJ, 642, 775Murphy, M. T. & Liske, J. 2004, MNRAS, 354, L31Nagao, T., Maiolino, R., & Marconi, A. 2006, A&A, 459, 85Nakagawa, Y. E., Yoshida, A., Sugita, S., et al. 2006, PASJ, 58, L35Nakar, E. 2007, Phys. Rep., 442, 166Nestor, D. B., Turnshek, D. A., & Rao, S. M. 2005, ApJ, 628, 637Nicastro, L., Amati, L., Antonelli, L. A., et al. 1999, A&AS, 138, 437

192 Bibliography

Norris, J. P. & Bonnell, J. T. 2006, ApJ, 643, 266Nousek, J. A., Kouveliotou, C., Grupe, D., et al. 2006, ApJ, 642, 389Oates, S. R., Mundell, C. G., Piranomonte, S., et al. 2006, MNRAS, 372, 327Odewahn, S. C., Djorgovski, S. G., Kulkarni, S. R., et al. 1997, IAU circ., 6735, 1Oke, J. B. & Gunn, J. E. 1983, ApJ, 266, 713Osterbrock, D. E. 1989, Astrophysics of gaseous nebulae and active galactic nuclei (Re-

search supported by the University of California, John Simon Guggenheim Memo-rial Foundation, University of Minnesota, et al. Mill Valley, CA, University ScienceBooks, 1989, 422 p.)

Ovaldsen, J.-E., Jaunsen, A. O., Fynbo, J. P. U., et al. 2007, ApJ, 662, 294Paczynski, B. 1995, PASP, 107, 1167Padoan, P., Cambresy, L., Juvela, M., et al. 2006, ApJ, 649, 807Pagel, B. E. J., Edmunds, M. G., Blackwell, D. E., Chun, M. S., & Smith, G. 1979,

MNRAS, 189, 95Pagel, B. E. J., Simonson, E. A., Terlevich, R. J., & Edmunds, M. G. 1992, MNRAS,

255, 325Palmer, D., Barbier, L., Barthelmy, S., et al. 2006, GCN Circular, 4697Palmer, D. M., Barthelmy, S., Gehrels, N., et al. 2005, Nature, 434, 1107Panaitescu, A. & Kumar, P. 2001, ApJ, 560, L49Panaitescu, A. & Kumar, P. 2002, ApJ, 571, 779Panaitescu, A., Meszaros, P., Burrows, D., et al. 2006, MNRAS, 369, 2059Pandey, S. B., Castro-Tirado, A. J., McBreen, S., et al. 2006, A&A, 460, 415Patel, S., Kouveliotou, C., & Rol, E. 2005, GCN Circular, 4163Pe’er, A. & Wijers, R. A. M. J. 2006, ApJ, 643, 1036Pei, Y. C. 1992, ApJ, 395, 130Pei, Y. C. & Fall, S. M. 1995, ApJ, 454, 69Penprase, B. E., Berger, E., Fox, D. B., et al. 2006, ApJ, 646, 358Perley, D. A., Bloom, J. S., Butler, N. R., et al. 2007, ArXiv Astrophysics e-printsPerna, R. & Lazzati, D. 2002, ApJ, 580, 261Perna, R., Lazzati, D., & Fiore, F. 2003, ApJ, 585, 775Petitjean, P., Aracil, B., Srianand, R., & Ibata, R. 2000, A&A, 359, 457Pettini, M., Kellogg, M., Steidel, C. C., et al. 1998, ApJ, 508, 539Pettini, M. & Pagel, B. E. J. 2004, MNRAS, 348, L59Pian, E., Mazzali, P. A., Masetti, N., et al. 2006, Nature, 442, 1011Pian, E., Soffitta, P., Alessi, A., et al. 2001, A&A, 372, 456Piran, T. 1999, Phys. Rep., 314, 575Piro, L., De Pasquale, M., Soffitta, P., et al. 2005, ApJ, 623, 314Piro, L., Frail, D. A., Gorosabel, J., et al. 2002, ApJ, 577, 680Predehl, P. & Schmitt, J. H. M. M. 1995, A&A, 293, 889Price, P. A., Harrison, F. A., Galama, T. J., et al. 2001, ApJ, 549, L7Price, P. A., Kulkarni, S. R., Berger, E., et al. 2003, ApJ, 589, 838Price, P. A., Songaila, A., Cowie, L. L., et al. 2007, ArXiv e-prints, 705Prochaska, J. X. 2006, ApJ, 650, 272Prochaska, J. X., Bloom, J. S., Chen, H.-W., et al. 2006a, ApJ, 642, 989

Bibliography 193

Prochaska, J. X., Bloom, J. S., Chen, H.-W., et al. 2004, ApJ, 611, 200Prochaska, J. X., Chen, H.-W., & Bloom, J. S. 2006b, ApJ, 648, 95Prochaska, J. X., Chen, H.-W., Dessauges-Zavadsky, M., & Bloom, J. S. 2007a, ArXiv

Astrophysics e-printsProchaska, J. X., Chen, H.-W., Wolfe, A. M., Dessauges-Zavadsky, M., & Bloom, J. S.

2007b, ArXiv Astrophysics e-printsProchaska, J. X., Gawiser, E., Wolfe, A. M., Castro, S., & Djorgovski, S. G. 2003, ApJ,

595, L9Prochter, G. E., Prochaska, J. X., Chen, H.-W., et al. 2006, ApJ, 648, L93Racusin, J., Burrows, D., & Gehrels, N. 2005a, GCN Circular, 4141Racusin, J., Kennea, J., Fox, D., et al. 2005b, GCN Circular, 4169Ramaprakash, A. N., Kulkarni, S. R., Frail, D. A., et al. 1998, Nature, 393, 43Reichart, D. E., Lamb, D. Q., Metzger, M. R., et al. 1999, ApJ, 517, 692Rhoads, J. E. 1997, ApJ, 487, L1+Rhoads, J. E. 1999, ApJ, 525, 737Ricker, G. R. & Vanderspek, R. K., eds. 2003, American Institute of Physics Conference

Series, Vol. 662, Gamma-Ray Burst and Afterglow Astronomy 2001: A WorkshopCelebrating the First Year of the HETE Mission

Rieke, G. H. & Lebofsky, M. J. 1985, ApJ, 288, 618Rol, E., Osborne, J. P., Page, K. L., et al. 2007, MNRAS, 374, 1078Rol, E., Starling, R., Wiersema, K., et al. 2005a, GCN Circular, 3710Rol, E., Wijers, R. A. M. J., Fynbo, J. P. U., et al. 2003, A&A, 405, L23Rol, E., Wijers, R. A. M. J., Kouveliotou, C., Kaper, L., & Kaneko, Y. 2005b, ApJ, 624,

868Rol, E., Wijers, R. A. M. J., Vreeswijk, P. M., et al. 2000, ApJ, 544, 707Rosswog, S., Ramirez-Ruiz, E., & Davies, M. B. 2003, MNRAS, 345, 1077Rykoff, E. S., Aharonian, F., Akerlof, C. W., et al. 2005, ApJ, 631, 1032Sakamoto, T., Barbier, L., Barthelmy, S., et al. 2006, in American Institute of Physics

Conference Series, Vol. 836, Gamma-Ray Bursts in the Swift Era, ed. S. S. Holt,N. Gehrels, & J. A. Nousek, 43–47

Salzer, J. J., Lee, J. C., Melbourne, J., et al. 2005, ApJ, 624, 661Sari, R. 1997, ApJ, 489, L37+Sault, R. J., Teuben, P. J., & Wright, M. C. H. 1995, in Astronomical Society of the Pa-

cific Conference Series, Vol. 77, Astronomical Data Analysis Software and SystemsIV, ed. R. A. Shaw, H. E. Payne, & J. J. E. Hayes, 433

Savage, B. D. & Mathis, J. S. 1979, ARA&A, 17, 73Savage, B. D., Meade, M. R., & Sembach, K. R. 2001, ApJS, 136, 631Savage, B. D. & Sembach, K. R. 1996, ARA&A, 34, 279Savaglio, S. 2006, New Journal of Physics, 8, 195Savaglio, S. & Fall, S. M. 2004, ApJ, 614, 293Savaglio, S., Fall, S. M., & Fiore, F. 2003, ApJ, 585, 638Savaglio, S., Glazebrook, K., Le Borgne, D., et al. 2005, ApJ, 635, 260Schady, P., Mason, K. O., Page, M. J., et al. 2007, MNRAS, 377, 273Schaefer, B. E. 2005, GCN Circular, 4132

194 Bibliography

Schaefer, B. E., Gerardy, C. L., Hoflich, P., et al. 2003, ApJ, 588, 387Schaerer, D., Contini, T., & Pindao, M. 1999, A&AS, 136, 35Schaerer, D. & Vacca, W. D. 1998, ApJ, 497, 618Schaye, J. 2001, ApJ, 562, L95Schaye, J., Carswell, R. F., & Kim, T.-S. 2007, ArXiv Astrophysics e-printsSchlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525Shapley, A. E., Erb, D. K., Pettini, M., Steidel, C. C., & Adelberger, K. L. 2004, ApJ,

612, 108Silva, A. I. & Viegas, S. M. 2002, MNRAS, 329, 135Silvey, J., Allen, D., Bayliss, M., et al. 2004, GCN Circular, 2833Skillman, E. D., Televich, R. J., Kennicutt, Jr., R. C., Garnett, D. R., & Terlevich, E.

1994, ApJ, 431, 172Skrutskie, M. F., Cutri, R. M., Stiening, R., et al. 2006, AJ, 131, 1163Slavin, J. D. 1989, ApJ, 346, 718Soderberg, A. M., Berger, E., Kasliwal, M., et al. 2006a, ApJ, 650, 261Soderberg, A. M., Kulkarni, S. R., Nakar, E., et al. 2006b, Nature, 442, 1014Soderberg, A. M., Kulkarni, S. R., Price, P. A., et al. 2006c, ApJ, 636, 391Sollerman, J., Cox, N., Mattila, S., et al. 2005a, A&A, 429, 559Sollerman, J., Fynbo, J. P. U., Gorosabel, J., et al. 2007, A&A, 466, 839Sollerman, J., Jaunsen, A. O., Fynbo, J. P. U., et al. 2006, A&A, 454, 503Sollerman, J., Ostlin, G., Fynbo, J. P. U., et al. 2005b, New Astronomy, 11, 103Srianand, R., Petitjean, P., Ledoux, C., Ferland, G., & Shaw, G. 2005, MNRAS, 362,

549Stanek, K. Z., Dai, X., Prieto, J. L., et al. 2007, ApJ, 654, L21Stanek, K. Z., Garnavich, P. M., Kaluzny, J., Pych, W., & Thompson, I. 1999, ApJ, 522,

L39Stanek, K. Z., Gnedin, O. Y., Beacom, J. F., et al. 2006, Acta Astronomica, 56, 333Stanek, K. Z., Matheson, T., Garnavich, P. M., et al. 2003, ApJ, 591, L17Starling, R. L. C., van der Horst, A. J., Rol, E., et al. 2007, ArXiv e-prints, 704Starling, R. L. C., Vreeswijk, P. M., Ellison, S. L., et al. 2005a, A&A, 442, L21Starling, R. L. C., Wijers, R. A. M. J., Hughes, M. A., et al. 2005b, MNRAS, 360, 305Starling, R. L. C., Wijers, R. A. M. J., Wiersema, K., et al. 2006, ArXiv Astrophysics

e-printsStasinska, G. & Izotov, Y. 2003, A&A, 397, 71Steidel, C. C., Adelberger, K. L., Giavalisco, M., Dickinson, M., & Pettini, M. 1999,

ApJ, 519, 1Steidel, C. C., Adelberger, K. L., Shapley, A. E., et al. 2003, ApJ, 592, 728Stratta, G., Fiore, F., Antonelli, L. A., Piro, L., & De Pasquale, M. 2004, ApJ, 608, 846Stratta, G., Perna, R., Lazzati, D., et al. 2005, A&A, 441, 83Sudilovsky, V., Savaglio, S., Vreeswijk, P., et al. 2007, ArXiv e-prints, 705Tan, G. H. 1991, in Astronomical Society of the Pacific Conference Series, Vol. 19, IAU

Colloq. 131: Radio Interferometry. Theory, Techniques, and Applications, ed. T. J.Cornwell & R. A. Perley, 42–46

Tanaka, K., Ricker, G., Atteia, J.-L., et al. 2005, GCN Circular, 4131

Bibliography 195

Tanvir, N., Wyse, R., Gilmore, G., & Corson, C. 1997, IAU Circ., 6796, 3Tanvir, N. R., Barnard, V. E., Blain, A. W., et al. 2004, MNRAS, 352, 1073Tanvir, N. R., Chapman, R., Levan, A. J., & Priddey, R. S. 2005, Nature, 438, 991Tejos, N., Lopez, S., Prochaska, J. X., Chen, H.-W., & Dessauges-Zavadsky, M. 2007,

ArXiv e-prints, 705Terada, H., Akiyama, M., & Kawai, N. 2004a, GCN Circular, 2750Terada, H., Akiyama, M., & Kawai, N. 2004b, GCN Circular, 2745Thone, C. C., Fynbo, J. P. U., Oestlin, G., et al. 2007, ArXiv Astrophysics e-printsThone, C. C., Greiner, J., Savaglio, S., & Jehin, E. 2006, ArXiv Astrophysics e-printsTimmes, F. X., Woosley, S. E., & Weaver, T. A. 1995, ApJS, 98, 617Tinney, C., Stathakis, R., Cannon, R., et al. 1998, IAU Circ., 6896, 1Tody, D. 1986, in Presented at the Society of Photo-Optical Instrumentation Engineers

(SPIE) Conference, Vol. 627, Instrumentation in astronomy VI; Proceedings of theMeeting, Tucson, AZ, Mar. 4-8, 1986. Part 2 (A87-36376 15-35). Bellingham, WA,Society of Photo-Optical Instrumentation Engineers, 1986, p. 733., ed. D. L. Craw-ford, 733

Tokunaga, A. T. & Vacca, W. D. 2005, PASP, 117, 1459Tominaga, N., Maeda, K., Umeda, H., et al. 2007, ApJ, 657, L77Torii, K. 2005, GCN Circular, 4130Tremonti, C. A., Heckman, T. M., Kauffmann, G., et al. 2004, ApJ, 613, 898Trentham, N., Ramirez-Ruiz, E., & Blain, A. W. 2002, MNRAS, 334, 983Tumlinson, J., Prochaska, J. X., Chen, H.-W., Dessauges-Zavadsky, M., & Bloom, J. S.

2007, ArXiv Astrophysics e-printsVan den Heuvel, E. P. J. 2004, in ESA Special Publication, Vol. 552, 5th INTEGRAL

Workshop on the INTEGRAL Universe, ed. V. Schoenfelder, G. Lichti, & C. Winkler,185

Van den Heuvel, E. P. J. 2005, in NATO ASIB Proc. 210: The Electromagnetic Spectrumof Neutron Stars, ed. A. Baykal, S. K. Yerli, S. C. Inam, & S. Grebenev, 191

Van den Heuvel, E. P. J. 2006, Advances in Space Research, 38, 2667Van den Heuvel, E. P. J. 2007, ArXiv e-prints, 704Van den Heuvel, E. P. J. & Yoon, S. . 2007, ArXiv e-prints, 704Van der Horst, A. J., Rol, E., & Wijers, R. A. M. J. 2005, GCN Circular, 4158Van der Horst, A. J., Wijers, R. A. M. J., & Van der Horn, L. 2007, A&A submittedVan Dokkum, P. G. 2001, PASP, 113, 1420Van Marle, A. J., Langer, N., & Garcıa-Segura, G. 2005, A&A, 444, 837Van Marle, A. J., Langer, N., & garcia-Segura, G. 2007, ArXiv e-prints, 704Van Paradijs, J., Groot, P. J., Galama, T., et al. 1997, Nature, 386, 686Vanderspek, R., Sakamoto, T., Barraud, C., et al. 2004, ApJ, 617, 1251Vink, J. S., de Koter, A., & Lamers, H. J. G. L. M. 2001, A&A, 369, 574Vladilo, G., Centurion, M., & Cassola, C. 1993, A&A, 273, 239Vreeswijk, P. M., Ellison, S. L., Ledoux, C., et al. 2004, A&A, 419, 927Vreeswijk, P. M., Fender, R. P., Garrett, M. A., et al. 2001, A&A, 380, L21Vreeswijk, P. M., Galama, T. J., Owens, A., et al. 1999, ApJ, 523, 171Vreeswijk, P. M., Ledoux, C., Smette, A., et al. 2007, A&A, 468, 83

196 Bibliography

Vreeswijk, P. M., Møller, P., & Fynbo, J. P. U. 2003, A&A, 409, L5Vreeswijk, P. M., Smette, A., Fruchter, A. S., et al. 2006, A&A, 447, 145Wainwright, C., Berger, E., & Penprase, B. E. 2007, ApJ, 657, 367Walker, M. A. 1998, MNRAS, 294, 307Watson, D., Fynbo, J. P. U., Ledoux, C., et al. 2006, ApJ, 652, 1011Watson, D., Hjorth, J., Fynbo, J. P. U., et al. 2007, ApJ, 660, L101Watson, D., Reeves, J. N., Osborne, J. P., et al. 2002, A&A, 395, L41Waxman, E. & Draine, B. T. 2000, ApJ, 537, 796Welsh, B. Y. & Lallement, R. 2005, A&A, 436, 615Welty, D. E., Hobbs, L. M., & Kulkarni, V. P. 1994, ApJ, 436, 152Welty, D. E., Morton, D. C., & Hobbs, L. M. 1996, ApJS, 106, 533Wiersema, K. & Nieuwenhout, F. 2006, GCN Circular, 4834Wiersema, K., Rol, E., Starling, R., et al. 2005, GCN Circular, 3699Wiersema, K., Savaglio, S., Vreeswijk, P. M., et al. 2007, A&A, 464, 529Wijers, R. A. M. J., Bloom, J. S., Bagla, J. S., & Natarajan, P. 1998, MNRAS, 294, L13Wijers, R. A. M. J. & Galama, T. J. 1999, ApJ, 523, 177Williams, B. J., Borkowski, K. J., Reynolds, S. P., et al. 2006, ApJ, 652, L33Wilms, J., Allen, A., & McCray, R. 2000, ApJ, 542, 914Wolf, C. & Podsiadlowski, P. 2007, MNRAS, 375, 1049Woosley, S. E. & Bloom, J. S. 2006, ARA&A, 44, 507Woosley, S. E. & Heger, A. 2006, ApJ, 637, 914Wozniak, P. R., Vestrand, W. T., Wren, J. A., et al. 2006, ApJ, 642, L99Yoon, S.-C. & Langer, N. 2005, A&A, 443, 643Yoon, S.-C., Langer, N., & Norman, C. 2006, A&A, 460, 199Yost, S. A., Frail, D. A., Harrison, F. A., et al. 2002, ApJ, 577, 155Yost, S. A., Harrison, F. A., Sari, R., & Frail, D. A. 2003, ApJ, 597, 459Zeh, A., Klose, S., & Hartmann, D. H. 2004, ApJ, 609, 952Zeh, A., Klose, S., & Kann, D. A. 2006, ApJ, 637, 889Zhang, B. & Meszaros, P. 2004, International Journal of Modern Physics A, 19, 2385Zibetti, S., Menard, B., Nestor, D. B., et al. 2007, ApJ, 658, 161Zwaan, M. A., van der Hulst, J. M., Briggs, F. H., Verheijen, M. A. W., & Ryan-Weber,

E. V. 2005, MNRAS, 364, 1467

So far as hypotheses are concerned, let no one expectanything certain from astronomy, which cannot furnish it, lest he acceptas the truth ideas conceived for another purpose, and depart from thisstudy a greater fool than when he entered it.

Andreas Osiander in the foreword of De Revolutionibus Orbium Coelestium byNicolaus Copernicus

UvA-DARE (Digital Academic Repository) Delving …...Delving into the Dragons Den: The host galaxies...

Documents

Transcript of UvA-DARE (Digital Academic Repository) Delving …...Delving into the Dragons Den: The host galaxies...