The SEDs and Host Galaxies of the dustiest GRB afterglows

Abstract

(Abridged) Until recently the information inferred from gamma-ray burst
follow-up observations was mostly limited to optically bright afterglows,
biasing all demographic studies against sight-lines that contain large amounts
of dust. Here, we present GRB afterglow and host observations for a sample of
bursts that are exemplary of previously missed ones because of high visual
extinction along the sight-line. This facilitates an investigation of the
properties, geometry and location of the absorbing dust of these
poorly-explored host galaxies, and a comparison to hosts from
optically-selected samples. The hosts of the dustiest afterglows are diverse in
their properties, but on average redder, more luminous and massive than the
hosts of optically-bright events. We hence probe a different galaxy population,
suggesting that previous host samples miss most of the massive,
chemically-evolved and metal-rich members. This also indicates that the dust
along the sight-line is often related to host properties, and thus probably
located in the diffuse ISM or interstellar clouds and not in the immediate GRB
environment. Some of the hosts in our sample, are blue, young or of small
stellar mass illustrating that even apparently non-extinguished galaxies
possess very dusty sight-lines due to a patchy dust distribution. The presented
observations establish a population of luminous, massive and correspondingly
chemically-evolved GRB hosts. This suggests that GRBs trace the global
star-formation rate better than studies based on optically-selected host
samples indicate, and the previously-claimed deficiency of high-mass host
galaxies was at least partially a selection effect.

Full text

Page 1

Astronomy & Astrophysics manuscript no. v2˙ref1
August 4, 2011
c ? ESO 2011
The SEDs and Host Galaxies of the dustiest GRB afterglows?
T. Kr¨ uhler1,2,3, J. Greiner1, P. Schady1, S. Savaglio1, P. M. J. Afonso1,4, C. Clemens1, J. Elliott1, R. Filgas1, D. Gruber1,
D. A. Kann5, S. Klose5, A. K¨ upc¨ u-Yoldas ¸6, S. McBreen7, F. Olivares E.1, D. Pierini??, A. Rau1, A. Rossi5,
M. Nardini1,8, A. Nicuesa Guelbenzu5, V. Sudilovsky1, and A. C. Updike9,10
(Affiliations can be found after the references)
ABSTRACT
Context. The afterglows and host galaxies of long gamma-ray bursts (GRBs) offer unique opportunities to study star-forming galaxies in the high-z
Universe. Until recently, however, the information inferred from GRB follow-up observations was mostly limited to optically bright afterglows,
biasing all demographic studies against sight-lines that contain large amounts of dust.
Aims. Here we present afterglow and host observations for a sample of bursts that are exemplary of previously missed ones because of high visual
extinction (AGRB
V
? 1mag) along the sight-line. This facilitates an investigation of the properties, geometry and location of the absorbing dust of
these poorly-explored host galaxies, and a comparison to hosts from optically-selected samples.
Methods. This work is based on GROND optical/NIR and Swift/XRT X-ray observations of the afterglows, and multi-color imaging for eight
GRB hosts. The afterglow and galaxy spectral energy distributions yield detailed insight into physical properties such as the dust and metal content
along the GRB sight-line as well as galaxy-integrated characteristics like the host’s stellar mass, luminosity, color-excess and star-formation rate.
Results. For the eight afterglows considered in this study we report for the first time the redshift of GRB 081109 (z = 0.9787 ± 0.0005), and the
visual extinction towards GRBs 081109 (AGRB
V
= 3.4+0.4
GRB afterglows. Combined with non-extinguished GRBs, there is a strong anti-correlation between the afterglow’s metals-to-dust ratio and visual
extinction.
The hosts of the dustiest afterglows are diverse in their properties, but on average redder (?(R − K)AB? ∼ 1.6mag), more luminous (?L? ∼ 0.9L∗)
and massive (?log M∗[M?]? ∼ 9.8) than the hosts of optically-bright events. We hence probe a different galaxy population, suggesting that previous
host samples miss most of the massive, chemically-evolved and metal-rich members. This also indicates that the dust along the sight-line is often
related to host properties, and thus probably located in the diffuse ISM or interstellar clouds and not in the immediate GRB environment. Some
of the hosts in our sample, are blue, young or of small stellar mass illustrating that even apparently non-extinguished galaxies possess very dusty
sight-lines due to a patchy dust distribution.
Conclusions. The afterglows and host galaxies of the dustiest GRBs provide evidence for a complex dust geometry in star-forming galaxies. In
addition, they establish a population of luminous, massive and correspondingly chemically-evolved GRB hosts. This suggests that GRBs trace the
global star-formation rate better than studies based on optically-selected host samples indicate, and the previously-claimed deficiency of high-mass
host galaxies was at least partially a selection effect.
−0.3mag) and 100621A (AGRB
V
= 3.8 ± 0.2mag), which are among the largest ever derived for
Key words. Gamma-ray burst: general, ISM: dust, extinction, Galaxies: star formation
1. Introduction
Long gamma-ray bursts (GRBs, see e.g., Zhang 2007; Gehrels
et al. 2009, for reviews) are linked to core-collapse supernovae
and hence star-formation via the death of massive stars (e.g.,
Galama et al. 1998; Hjorth et al. 2003). At high redshifts,
where a significant fraction of the star-formation is thought to
be dust-obscured (e.g., Adelberger & Steidel 2000; Chapman
et al. 2005), GRBs and their host galaxies offer an independent
track towards a better understanding and full census of the star-
formation in the early Universe: GRBs, having luminous emis-
sion in a simple power-law spectrum provide the ideal back-
ground light to illuminate dust-enshrouded star-forming regions
which would otherwise remain unexplored, while at the same
time pinpointing their host galaxies.
However,theextenttowhichGRBhostsprovideanunbiased
picture of the formation of high-mass stars, and whether they
?Based on observations made with GROND at the MPI/ESO 2.2 m
telescope and with telescopes at the European Southern Observatory at
LaSilla/Paranal, Chile under program 086.A-0533 and obtained from
the ESO/ST-ECF Science Archive Facility from programs 177.A-0591
and 078.D-0416.
??Visiting Astronomer at MPE
preferentially occur in low-metallicity environments remains a
much debated issue (e.g., Le Floc’h et al. 2003; Fynbo et al.
2003; Tanvir et al. 2004; Fruchter et al. 2006; Kocevski et al.
2009). In single progenitor models, metal-poor stars are favored
by theory (Woosley 1993; MacFadyen & Woosley 1999), as they
would in principle be able to keep more angular momentum at
the time of stellar collapse due to smaller wind pressures and
losses throughout their evolution (e.g., Yoon & Langer 2005;
Mokiem et al. 2007). However, binary progenitor channels could
also play an important role in the formation of long GRBs (e.g.,
Fryer et al. 1999), having somewhat relaxed metallicity con-
straints relative to single star progenitors (Fryer et al. 2007).
Observations of GRB hosts are hence not only important in a
cosmological context, but provide relevant clues to the exact na-
ture of GRB progenitors.
A fundamental limit of hitherto available GRB host galaxy
samples is the incompleteness which arises from the non-
detection of the optical afterglow of a GRB (e.g., Groot et al.
1998; Fynbo et al. 2001). These optically dark bursts could be
caused by either high-redshift (e.g., Greiner et al. 2009; Tanvir
et al. 2009; Salvaterra et al. 2009; Cucchiara et al. 2011), large
columns of dust (e.g., Klose et al. 2000, 2003; Tanvir et al. 2008;
arXiv:1108.0674v1 [astro-ph.CO] 2 Aug 2011

Page 2

2T. Kr¨ uhler et al.: The SEDs and Host Galaxies of the dustiest GRB afterglows
Perley et al. 2011b) or an intrinsically fainter optical afterglow
as compared to the extrapolation of X-ray data when using syn-
chrotron emission theory, i.e., a decoupled optical/X-ray after-
glow light-curve (e.g., Panaitescu et al. 2006; Ghisellini et al.
2009; Nardini et al. 2010). New afterglow samples became avail-
able through the recent advent of dedicated afterglow follow-up
campaigns on medium-to-large aperture telescopes (e.g., Fynbo
et al. 2009a; Cenko et al. 2009; Greiner et al. 2011). These new
afterglow samples reach completeness levels of ∼90% (Greiner
et al. 2011) and settled the dark-burst issue: Around three quar-
ters of dark bursts are the result of a dusty afterglow line of sight
(e.g., Perley et al. 2009; Greiner et al. 2011). Accurate positions
from afterglow observations are necessary to unambiguously as-
sociate galaxies to GRBs. The lack of optical/NIR afterglows for
dark GRBs therefore implies an inherent bias against the associ-
ated host galaxies.
The available host population is not as complete as the most
recent afterglow samples. Instead, it is largely selected from op-
tically bright afterglows and shows a prevalence of young and
vigorously star-forming galaxies with sub-L∗luminosities and
masses around 109M?(e.g., Bloom et al. 1998; Le Floc’h et al.
2003;Christensenetal.2004;Fruchteretal.2006;Savaglioetal.
2009, referenced as SGL09, hereafter). However, it is an open
question whether this is a physical consequence of GRBs prefer-
ring low-metallicity environments, or merely a selection effect:
Host galaxies of dark GRBs were typically not identified, and
hence are under-represented in the available host sample.
Whether the physical characteristics of hosts of optically
dark and bright GRBs are distinct is also the subject of discus-
sion. Previous sample studies (e.g., Berger et al. 2003; Le Floc’h
et al. 2003; Perley et al. 2009) have not revealed strong evidence
of a significant difference. A handful of single dark GRBs were
however hosted by red and dusty galaxies with high metallic-
ities and stellar masses over 1011M?(e.g., Levan et al. 2006;
Berger et al. 2007; Levesque et al. 2010; Hashimoto et al. 2010;
K¨ upc¨ u Yoldas ¸ et al. 2010; Chen et al. 2010). Recently Perley
et al. (2010a, 2011a) suggested, that the general galaxy popula-
tion hosting dark bursts is redder and more luminous and hence
suggestive of a higher metallicity than those selected via opti-
cally bright afterglows.
In this paper, we study the nature of GRB hosts that previ-
ously escaped detection due to the dust bias, and are hence ex-
emplary of those missing from demographic studies. We avail of
dedicated GRB afterglow campaigns with high completeness to
preselect the GRB hosts for this study. These afterglow data do
not only provide accurate positions for host identifications, but
for the first time allow us to directly select dust extinguished
(and not only optically faint) GRBs via well-sampled broad-
band (NIR to X-ray) afterglow observations.
After selecting afterglows with visual extinctions AGRB
ceeding unity, we search for the associated hosts with the
Gamma-Ray Optical and Near-Infrared Detector (GROND,
Greiner et al. 2008) as well as the ESO Very Large and New
Technology Telescopes (VLT and NTT, respectively) and the
Ultra-Violet Optical Telescope (UVOT, Roming et al. 2005) on-
board the Swift satellite (Gehrels et al. 2004).
The obtained data allow us to study the physical properties
of the hosts of high-AGRB
V
GRBs in detail, and to investigate the
bias against dust in GRB host samples. As an ultimate conse-
quence they address the role of GRB host galaxies as tracers
of galaxy formation and evolution. Furthermore, they link after-
glow diagnostics, i.e., detailed information about a single sight
line, to host-integrated properties, and in this combination di-
V
ex-
012345
AfterglowAV[mag]
100
101
102
Numberof bursts
PreviousAfterglows
ThisSample
Fig.1. Histogram of visual extinction values from previous af-
terglows and those selected for this study. We note that five of
the afterglows in this work (GRBs 070802, 080605, 080607,
080805, 090926B) were already discussed in previous sample
studies, but are only included in one of the histograms for clar-
ity.
rectly probe the nature of dust and its properties in high-redshift,
star-forming galaxies.
Throughout this work, we adopt the convention that the flux
density of the afterglow Fν(ν,t) can be described as Fν(ν,t) ∝
ν−βt−α, and concordance (ΩM = 0.27, ΩΛ = 0.73, H0 =
71km/s/Mpc)ΛCDMcosmology.Allerrorsaregivenat1σcon-
fidence unless indicated otherwise. All magnitudes and colors
are given or converted into the AB system.
2. Sample Selection
The host galaxy sample presented in this work is based on a
direct measurement of large visual extinction along the GRB
line of sight (AGRB
V
? 1mag) from multi-color (NIR to X-
ray) afterglow observations. Specifically, eight GRB afterglows
(GRBs 070306, 070802, 080605, 080607, 080805, 081109,
090926B and 100621A) fulfill the selection criterion and define
our initial host sample. Our host sample is a direct result of af-
terglow observations. The selection itself is hence not limited by
galaxy brightness, nor introduces a bias towards luminous galax-
ies. Afterglow measurements for the initial selection have been
obtained from the literature or by analyzing photometric opti-
cal/NIR data from the GROND archive. In the latter case, they
are detailed in Sect. 3.1.
Our eight GRBs have a median redshift of ?zAV? = 1.5.
This is significantly lower than the published mean of Swift
GRBs with measured redshifts (?zSwift? = 1.9, Fynbo et al.
2009a1), but larger than the one of the previous host sample
(?zSGL09? = 0.96 from SGL09), which includes a large number
of pre-Swift events. Within the selected eight afterglows, four of
them show a clear 2175Å dust feature (GRBs 070802, 080605,
080607, 080805), which is with the exception of GRB 080603A
(Guidorzi et al. 2011; Kann et al. 2011) the full sample where
a significant detection of this feature has been reported to date
(e.g., Zafar et al. 2011). In fact, the 2175 Å feature could not
1Including updates from http://www.mpe.mpg.de/˜ jcg/grbgen.html
and http://www.raunvis.hi.is/˜pja/GRBsample.html

Page 3

T. Kr¨ uhler et al.: The SEDs and Host Galaxies of the dustiest GRB afterglows3
have been detected in the rest of the sample. This arises from
the combination of a large dust column, and not deep and rapid
enough follow-up in the case of GRBs 070306 and 090926B, or
no observational wavelength coverage at 2175Å×(1+z) (GRBs
081109, 100621A).
While observationally challenging, the requirement of a dust
measurement has several obvious advantages over a selection
based on the optical-to-X-ray flux ratio (βoX, see e.g., Jakobsson
et al. 2004; Rol et al. 2005; van der Horst et al. 2009). Most im-
portantly, the selection is the result of a measurement rather than
an extrapolation, and provides a clean selection of dusty GRBs:
Our afterglows are chosen accord to their visual extinction, in-
stead of only their optical faintness. The latter events could of
course also be at high-z or have different emission mechanisms
in the optical and X-ray regime.
We note, that our selection is still somewhat model-
dependent, specifically that the afterglow emits synchrotron ra-
diation in the optical/NIR and X-ray regime. Despite the lack
of conclusive alternatives to the standard synchrotron-fireball
model, there are still puzzling features in well-sampled multi-
color light curves, such as chromatic2breaks, optical and/or X-
ray flares or plateaus (e.g., Panaitescu et al. 2006; Evans et al.
2009; Kr¨ uhler et al. 2009; Oates et al. 2011, also illustrated in
Fig. 3). The apparently decoupled optical and X-ray light curve
for some bursts results in a strong dependence of βoXon the time
of the observation (which in fact is directly observed in some
cases, see e.g., Filgas et al. 2011), and hence the dark burst def-
inition depends also on observational constraints and not only
physical properties.
In addition, the required optical and/or NIR afterglow detec-
tion yields the GRB position to sub-arcsec accuracy, and hence
negligible chance-coincidence probabilities for field galaxies,
which is particularly important for small sample sizes as in this
work. For all events a spectroscopic redshift is available from
the literature, or could be obtained via host galaxy spectroscopy.
This enables quantitative studies, such as the comparison be-
tween specific sight-lines against host-integrated properties, as
well as an investigation of the relation between the dusty hosts
to the host sample of SGL09.
Due to the inherent difficulties to accurately localize high-
AGRB
V
afterglows, most previous afterglow samples are biased
towards small visual extinctions. This is illustrated in Fig. 1,
whichshowsthevisualextinctionofGRBsselectedforthiswork
as compared to previously compiled AGRB
were taken from Kann et al. (2006, 2010); Schady et al. (2010);
Greiner et al. (2011) and Zafar et al. (2011).
V
values. The latter
3. Observations
3.1. Afterglows
Optical and near-infrared measurements of the afterglows of
GRBs 070306, 070802, 080605, 080607, 080805 and 090926B
or results thereof are taken from Jaunsen et al. (2008), Kr¨ uhler
et al. (2008), Perley et al. (2011b), Greiner et al. (2011), and
Zafar et al. (2011), respectively. GROND observations of the af-
terglows of GRB 081109 and GRB 100621A are not presented
elsewhere and are shortly described in the following.
2Chromatic light-curve breaks are breaks in the X-ray regime not
associated with contemporanous breaks in the optical/NIR bands, and
vice versa.
10
100
1000001e+061e+07
19
20
21
22
1 10 100
Fν [µJy]
Brightness [magAB]
Time since GRB trigger [s]
Time since GRB trigger [d]
I
H
Fig.2. GROND H-band light curve of the afterglow of
GRB 081109. Highlighted with a grey shaded area and labeled
with I is the time interval which has been used to extract a si-
multaneous, host subtracted afterglow SED from the GROND
and Swift/XRT data.
3.1.1. GRB 081109
Swift triggered on GRB 081109 (Immler et al. 2008), and X-ray
and NIR measurements of the afterglow were rapidly reported
by Beardmore et al. (2008) and D’Avanzo et al. (2008). GROND
observations in seven optical/NIR filters (g?r?i?z?JHKs) simulta-
neously, started 17.1 hr after the GRB trigger (Clemens et al.
2008b) and a preliminary analysis of the spectral energy distri-
bution (SED) already revealed significant reddening of the op-
tical/NIR afterglow (Clemens et al. 2008a). GROND continued
to observe the transient at 2, 3, 6 and 378 days after the trig-
ger, where the host brightness was derived from the last epoch.
GROND afterglow and host measurements are given in Table 1,
and Tables 3 and 4, respectively. A possible host galaxy was also
reported in the UVOT white filter (Kuin & Immler 2008). High-
energy prompt and afterglow data, early NIR imaging and a light
curve and X-ray spectral analysis of this burst are presented by
Jin et al. (2009).
3.1.2. GRB 100621A
GROND reacted immediately (Updike et al. 2010) to the Swift
trigger of GRB 100621A (Ukwatta et al. 2010b) taking the
first images at 230 s after the burst. Simultaneous imaging in
g?r?i?z?JHKscontinued for 3.05 h, and was resumed on night 2,
4 and 10 after the burst. Analysis of the Swift/X-ray data and
further detections of the NIR afterglow were reported by Stratta
et al. (2010) and Naito et al. (2010), respectively. Early GROND
dataalreadyrevealedevidenceforsubstantialreddeningandhost
emission dominating the flux in the bluest filters, which was
also seen by Swift/UVOT (Oates & Ukwatta 2010). Afterglow
and host measurements are again shown in Tables 1, 3 and 4.
The temporal evolution of the optical/NIR afterglow is complex
with a very steep increase in brightness of around 1.5 mag in
the J band from 3.5 to 4.5 ks after the trigger. The light curve
is hence very similar the one of GRB 081029 (Nardini et al.
2011), where an analogous behavior could be associated with
the intrinsic properties of the GRB and not to changes in the in-
tervening dust content. For completeness, the J-band light curve
of GRB 100621A is shown in Fig. 3, but its detailed modeling

Page 4

4T. Kr¨ uhler et al.: The SEDs and Host Galaxies of the dustiest GRB afterglows
10
100
1000
100010000 1000001e+06
16
17
18
19
20
21
0.01 0.1 1 10
Fν [µJy]
Brightness [magAB]
Time since GRB trigger [s]
Time since GRB trigger [d]
I II
J
Fig.3. GROND J-band light curve of the afterglow of
GRB 100621A. Highlighted with grey shaded areas and labeled
with I and II are the time intervals which have been used to ex-
tract a simultaneous, host subtracted afterglow SED from the
GROND and Swift/XRT data.
and interpretation is beyond the scope of this paper, and will be
discussed in a future work.
3.2. Hosts
Once the sample based on afterglow host extinction was defined,
late follow-up observations were initiated first with GROND,
and in case of non-detections in individual filters, were con-
tinued with telescopes of successively increasing aperture size,
specifically with EFOSC/SOFI at the NTT (4m class) and
FORS2/HAWKI at the VLT (8m class). In one case without
published redshift (GRB 081109), the photometric imaging was
complemented by low-resolution spectroscopy with FORS2.
Public VLT data for known hosts (GRBs 070306 and 070802)
were obtained from the ESO archive. Ground-based data were
complemented by Swift/UVOT imaging for GRBs 081109 and
100621A. In the case of GRB 080607 all host measurements
were taken from Chen et al. (2010, 2011).
4. Data reduction and SED fitting
4.1. Swift/XRT & UVOT data
X-ray data have been retrieved from the HEASARC archive3,
and reduced with the xrtpipeline. Spectra have been grouped to
yield a minimum of 20 counts per bin, while the light curve was
taken from the Swift/XRT light curve repository (Evans et al.
2007, 2009). Swift/UVOT photometry has been obtained follow-
ing Poole et al. (2008) and is provided in Table 2.
4.2. Ground-based optical/NIR photometry
All optical (GROND, EFOSC, FORS1/2) and near-infrared
(GROND, SOFI, HAWK-I, ISAAC) imaging was reduced in a
standard manner using pyraf/IRAF (Tody 1993) similar to the
procedure outlined in Kr¨ uhler et al. (2008). For afterglow and
host photometry, point-spread function (PSF) fitting and aper-
ture photometry was used, respectively. The aperture diameter
3http://heasarc.gsfc.nasa.gov/cgi-bin/W3Browse/swift.pl
for individual hosts ranges between typically 1.0??to 2.5??, cor-
responding to values between 1.5 to 4 times the PSF FWHM. It
has been chosen sufficiently large to include the largest fraction
of host flux, and given the typical extent of these galaxies (? 1??,
i.e., ? 8.5kpc at z ∼ 1.5), the fraction of missed low-surface
brightness emission is very likely not a major contribution to the
presented measurements.
PSF and aperture photometry were then flux calibrated
against GROND observations of SDSS fields (Abazajian et al.
2009) taken immediately before or after the GRB field for the
optical g?r?i?z?filters. BVRI photometry was obtained by cre-
ating a set of 20 − 30 secondary standards from the GROND
photometry of field stars and the color terms from Lupton4. U-
band photometry was tied to observations of Landolt standard
stars (Landolt 1992) taken during the same night at different air-
masses, which allowed a reliable correction of the atmospheric
extinction to be applied.
NIR photometry was measured against 20−60 point sources
from the 2MASS catalog (Skrutskie et al. 2006) in the 10?× 10?
field of view of GROND. The zeropoint for the NIR imagers
with smaller field of views (in particular SOFI and ISAAC) was
then derived using 3 − 10 secondary standards common in the
GROND and SOFI/ISAAC/HAWK-I frames. The HAWK-I Y-
band imaging was calibrated against the z?and J measurements
from field stars using the color term
Y = 0.05 + 0.463 × (z − J) + J
where all magnitudes are in the AB system. This color term
has been obtained from synthetic photometry of stars with the
templatesofPickles(1998)andChabrieretal.(2000),andyields
an rms residual scatter for individual stars of 0.07 mag.
This procedure resulted in typical absolute accuracies of 2-
5% for the optical (U to z?) filters and 4-8% in the NIR (YJHKs),
which have been added in quadrature to the error introduced
by photon noise. All data used in the analysis have been cor-
rected for the expected Galactic foreground extinction according
to Schlegel et al. (1998) with RV= 3.08. For the selected fields,
this correction is small, i.e., EB−V ? 0.15mag in all cases. The
uncertainty of order of 10% in the Galactic foreground correc-
tion is hence not going to affect the overall results of this work.
Ground-based photometric measurements of the afterglows and
hosts are shown in Tables 1, 3 and 4, respectively.
4.3. Long-slit spectroscopy
In addition to the photometric observations, the host of
GRB 081109 was also observed spectroscopically with the VLT
equipped with FORS2. In total 2×1200 s spectra were obtained
with the grisms 300V and 300I and a long-slit width of 1.6??.
Acquisition images were taken in the V and I filter. The spec-
troscopic data were obtained at airmasses of ∼ 1.1 and see-
ing of 0.9??, which results in a line-spread function of approx-
imately 1.9 nm at 570.0 nm. The data were reduced using stan-
dard procedures in pyraf/IRAF, with the wavelength solution ob-
tained against an HeHgCd arclamp exposures with 25 lines leav-
ing residuals of around 0.07 nm rms. Flux-calibration was per-
formed against the spectro-photometric standard BPM162745.
The wavelength- and flux-calibrated spectrum was corrected for
Galactic foreground extinction and renormalized to the available
photometry.
4http://www.sdss.org/dr7/algorithms/sdssUBVRITransform.html
5http://www.eso.org/sci/observing/tools/standards/spectra/bpm16274.html

Page 5

T. Kr¨ uhler et al.: The SEDs and Host Galaxies of the dustiest GRB afterglows5
Table 1. Afterglow photometric measurements
GRB
081109
100621A
Notes: ∆t is the mean time of the observation after the GRB trigger. All magnitudes are in the AB system and uncorrected for Galactic foreground
reddening. Values in brackets correspond to photometric errors in units of valid digits. Upper limits are given at 3σ confidence.
∆t [ks]
65
7.6
g?
r?
i?
z?
JHKs
>24.53
21.56(17)
>24.75
19.92(08)
23.6(4)
18.67(06)
22.8(3)
17.71(04)
21.24(17)
16.02(04)
19.78(11)
15.02(05)
18.83(08)
14.32(06)
Table 2. Swift/UVOT UV (uvw2 to u) photometric measurements of GRB hosts
Hostuvw2
>24.1
22.31(04)
uvm2
23.6(3)
22.23(06)
uvw1
23.4(3)
22.20(07)
u
GRB 081109
GRB 100621A
Notes: All measurements in the AB system and uncorrected for the Galactic foreground reddening. Values in brackets correspond to photometric
errors in in units of valid digits. Upper limits are given at 3σ confidence.
>22.9
21.95(06)
Table 3. Optical (U to z?) photometric measurements of GRB hosts
HostU
—
—
—
—
g?
V
—
—
—
r?
Ri?
Iz?
GRB 070306
GRB 070802
GRB 080605
GRB 080805
GRB 081109
GRB 090926B
GRB 100621A
Notes: All magnitudes in the AB system and uncorrected for the Galactic foreground reddening. Values in brackets correspond to photometric
errors in units of valid digits. Upper limits are given at 3σ confidence. Data for the host of GRB 080607 were taken from Chen et al. (2010, 2011),
and are not shown in the table.
22.90(09)
—
23.15(07)
—
23.07(07)
23.31(07)
21.86(06)
23.08(09)
—
22.82(07)
—
22.74(07)
22.96(06)
21.48(06)
23.00(09)
25.20(09)
—
25.5(2)
—
—
—
22.81(13)
25.5(3)
22.81(08)
25.7(4)
22.01(08)
22.92(12)
21.15(06)
—
—
—
—
22.86(17)
—
22.76(11)
—
21.99(09)
22.44(10)
21.46(06)
25.7(2)
22.85(06)
—
—
23.23(14)
23.71(13)
21.95(10)
21.96(09)
—
—
4.4. SED fitting
4.4.1. Afterglows
For the afterglow SED analysis, X-ray (0.3 − 10 keV) and opti-
cal/NIR data were fit together under the assumption that the un-
derlying continuum emission is well represented by synchrotron
emission (e.g., Sari et al. 1998; Galama & Wijers 2001; Schady
et al. 2007; Rossi et al. 2011a).
Rest-frame soft X-ray photons are absorbed by metals, pre-
dominantly α-chain elements, while UV over optical to NIR
wavelengths are decreasingly affected by dust absorption. Good
coverage above 1 keV combined with NIR observations allows
for an accurate determination of the continuum, and hence good
constraints on the dust abundance (represented by the absorp-
tion in the rest-frame, optical V-band, AGRB
and total metal content (converted into a Hydrogen-equivalent
column density NH,Xassuming solar abundances from Anders
& Grevesse 1989) along the line of sight. Specifically, single
and broken power-law continua were used, where in the latter
case the two power-law slopes β1and β2were fixed to yield
β1+ 0.5 = β2as expected for the cooling break of synchrotron
emission in the slow cooling regime (e.g., Sari et al. 1998;
Granot & Sari 2002).
TheNIRtoX-raySEDswerefitinX-spec(Arnaud1996)us-
ing extinction laws for the Milky Way (MW), and Small (SMC)
and Large (LMC) Magellanic Clouds from Pei (1992). For the
GROND photometry, measurements from the time frame indi-
cated in Fig. 2 and 3 were used, where we chose interval II for
GRB 100621A due to a better signal-to-noise ratio. Data taken
at interval I yield consistent results for AGRB
X-ray regime, where spectral changes in the late evolution of an
afterglow are typically very moderate or in most cases even com-
pletely absent, the full XRT data set with a constant hardness ra-
V
), the extinction law
V
and NH,X. In the
tio was used to create a time-averaged spectrum. The latter was
then rescaled to the flux value at the time of the GROND ob-
servations derived from fitting the XRT light-curve with simple
afterglow models. The early steep decay of the X-ray light curve
and epochs of flaring activity were excluded from the combined
spectrum as well as the light-curve fitting.
4.4.2. Hosts
UV/optical/NIR photometry of the hosts of the selected GRBs
were analyzed in a standard way using stellar population synthe-
sis (SPS) techniques to convert luminosities into stellar masses
M∗(e.g., Bell et al. 2003; Ilbert et al. 2010) within LePhare6. In
detail, 3 × 106galaxy templates based on models from Bruzual
& Charlot (2003) with a universal IMF (Chabrier 2003) and dif-
ferent ages, star formation histories, extinction laws, reddening
values and metallicities were fit to the data. In addition, emission
lines are taken into account by converting the de-reddened UV
luminosity into a star formation rate, and hence line strengths
of Ly-α, Hα, Hβ, [OII] and [OIII] following Kennicutt (1998)
and Ilbert et al. (2009). In particular for vigorously star-forming
galaxies such as GRB hosts, this effect is a significant contribu-
tion to even broad-band photometry (e.g., Watson et al. 2011)
and reaches values of up to ∼ 0.2 − 0.3mag (Ilbert et al. 2009).
For a direct comparison with results published in the literature
(e.g., Fontana et al. 2006; Marchesini et al. 2009; Ilbert et al.
2010, SGL09), the attenuation law from Calzetti et al. (2000)
derived for starburst galaxies is used, unless different reddening
laws provide a better fit to the host data at 90% confidence. We
caution that access to the rest-frame NIR, which is the best tracer
of the stellar mass of the galaxy, is somewhat limited for part of
the sample. However, all hosts are detected in at least one filter
6http://www.cfht.hawaii.edu/∼arnouts/LEPHARE

Page 6

6T. Kr¨ uhler et al.: The SEDs and Host Galaxies of the dustiest GRB afterglows
Table 4. NIR (Y to K) photometric measurements of GRB hosts
HostY
—
—
—
—
JGROND
21.9(4)
—
21.9(2)
—
21.40(17)
22.3(4)
21.22(10)
JHGROND
21.5(4)
—
22.3(3)
—
21.5(4)
> 21.6
21.18(14)
HKGROND
> 21.1
—
> 21.1
—
> 20.6
> 20.9
> 21.1
K
GRB 070306
GRB 070802
GRB 080605
GRB 080805
GRB 081109
GRB 090926B
GRB 100621A
Notes: All measurements are given in the AB system and are uncorrected for the Galactic foreground reddening. Values in brackets correspond to
photometric errors in units of valid digits. Upper limits are given at 3σ confidence.
21.62(08)
24.5(3)
—
23.6(2)
21.37(06)
21.88(13)
21.43(06)
21.20(12)
—
—
—
—
21.9(3)
—
21.38(10)
23.4(3)
—
23.1(2)
21.05(08)
21.44(19)
21.23(11)
21.63(08)
—
21.10(06)
redwards of the 4000 Å break, which allows a reasonable esti-
mate of M∗of a galaxy (e.g., Glazebrook et al. 2004; Ilbert et al.
2009, SGL09).
Systematic uncertainties of up to an average of 0.2−0.3 dex
on M∗are present due to the specific details of the stellar popula-
tionmodelsandtheassumedattenuation/extinctionlaw(Pozzetti
et al. 2007; K¨ upc¨ u Yoldas ¸ et al. 2007; Kajisawa et al. 2009; Ilbert
et al. 2010). Despite the small sample size of GRB hosts, there
is evidence for an offset of around 0.2 dex between the presented
method and the one of SGL09 as shown in Fig 4. In the follow-
ing the recalculated stellar masses of the long GRB hosts with
the photometric data compiled in SGL09 are used as a compari-
son sample.
The SPS fit returns not only the luminosity and mass of the
galaxy, but also other physical properties of the host, such as the
age of the dominant stellar population τ, its color-excess E(B−V)
and the star-formation rate (SFR) derived from the rest-frame
UV flux. Reported physical host properties are the median of
the probability distribution of the total grid over all galaxy tem-
plates at the fixed spectroscopic redshift. Errorbars represent the
maximum and minimum value of the respective parameter in the
global χ2-distribution of the multi-dimensional parameter grid.
Typically, errors are asymmetric and dominated by the uncer-
tainty of the color excess in the host galaxy, which results in
logarithmic errors on all galaxy parameters. Absolute magni-
tudes and masses are compared against the redshift dependent
galaxy luminosity functions (Willmer et al. 2006; Marchesini
et al. 2007), and stellar mass function (Marchesini et al. 2009;
Ilbert et al. 2010)
5. Results
5.1. Afterglow extinction and metals-to-dust ratios
Out of the total of eight afterglows in the sample, the extinc-
tion properties for five of them (GRBs 070802, 080605, 080607,
080805, 090926B) are extensively discussed in previous works
(Kr¨ uhler et al. 2008; El´ ıasd´ ottir et al. 2009; Perley et al. 2011b;
Greiner et al. 2011; Zafar et al. 2011). In these cases, also the
published afterglow analysis is comparable to the approach of
this work, and it is therefore not repeated here, with results from
the literature being summarized in Table 5. For the remaining
three events (GRBs 070306, 081109, 100621A), we present ei-
ther new data and their modeling (GRBs 081109, 100621A), or
a new analysis (GRB 070306) in the following.
The afterglow of GRB 070306 was discussed in Jaunsen
et al. (2008). As their analysis is significantly different from
our approach, we refit the available afterglow data following
Sect. 4.4.1. The broad-band SED in Fig. 5 is reasonably well
fit (χ2= 123/108 d.o.f.) with a single power-law continuum
with spectral index β = 1.00 ± 0.07, an AGRB
V
= 5.5+1.2
−1.0mag and
789 10 11
log MSGL09
∗
[M?]
7
8
9
10
11
log MBC03
∗
[M?]
−0.50.0
∆(logM)
0.51.0
0
2
4
6
8
10
12
14
Numberof galaxies
Fig.4. Comparison of stellar masses from the host sample of
SGL09. The plot shows host masses which were derived from
the photometry compiled in SGL09 and following Sect. 4.4.2
with models from Bruzual & Charlot (2003) against values di-
rectly taken from SGL09. Error bars are the maximum and mini-
mum value of the stellar mass in the global χ2-distribution of 3×
106galaxy templates (see Sect. 4.4.2). The solid lines represents
equality, and the dashed line the median offset. Increasingly
grey shaded areas show dispersions of 0.2, 0.5 and 1.0 dex re-
spectively. The inset shows the distribution of mass differences,
which has a median of around 0.2 dex.
NH,X= 2.5+0.3
metals-to-dust ratio of NH,X/AGRB
the redshift of z = 1.496, and the sparse wavelength coverage
in the NIR (probing the rest-frame optical red-ward of 400 nm,
wherethereislittledistinctionbetweenlocalextinctionlaws),all
local dust models provide equally good fits to the data of course,
and within errors compatible values of β, AGRB
strong statements can be made either with respect to a possible
presence of a break between NIR and X-ray data. We adopt the
model with the least number of free parameters (single power-
law continuum, which also yields the lowest χ2), but note that in
the case of a break between the two wavelength regimes (as seen
in most early GRB afterglows, Greiner et al. 2011) the fit is of
comparable quality (∆χ2= 3), and yields an best-fit AGRB
would be significantly lower (AGRB
errors consistent with the single power-law values.
The afterglow SED for GRB 081109 at z = 0.979 has been
constructed from GROND and Swift data and is shown in Fig. 6.
After subtraction of the late host epoch, no residual flux is de-
tected in the two bluest GROND filters g?and r?. The after-
−0.2× 1022cm−2at 90% confidence, which implies a
V
= 4.4×1021cm−2/mag. Given
V
, and NH,X. No
V
which
V
= 4.3+1.1
−1.0mag), but within

Page 7

T. Kr¨ uhler et al.: The SEDs and Host Galaxies of the dustiest GRB afterglows7
Table 5. Afterglow properties
Afterglow Redshift(∗)
T(∗∗)
90
[s]
210(a)
16.4(d)
20(e)
79(g)
78(h)
190(i)
110(j)
63.6(k)
β(∗∗∗)
opt
βX
AGRB
V
[mag]
5.5+1.2
−1.0
NH,X
χ2/ d.o.f.
1022cm−2
2.5+0.3
−0.2
2.0+0.7
1.01+0.09
2.7+0.8
1.0+0.6
1.10+0.13
2.2+0.5
1.62+0.15
GRB 070306
GRB 070802(b,c)
GRB 080605(b,c)
GRB 080607(c,f)
GRB 080805(b,c)
GRB 081109
GRB 090926B(b)
GRB 100621A
(∗)Redshifts from Jaunsen et al. (2008), El´ ıasd´ ottir et al. (2009), Fynbo et al. (2009a), Fynbo et al. (2009b) and Milvang-Jensen et al. (2010).
(∗∗)T90is the duration in which the GRB emits from 5% to 95% of its γ-rays, and is used to discriminate between short and long bursts. Typically,
long GRBs have T90> 2 s (Kouveliotou et al. 1993). All GRBs in this work are hence unambigously long events.
(∗∗∗)βoptis tied to βXin the fitting, and hence has the same error.
Notes: Afterglow measurements were taken from the reference denoted with the superscript in the first line. When two references are given, we
quote both values for AGRB
V
and NH,X, but values for β and χ2only from the first one for the sake of clarity.
References: (a) Barthelmy et al. (2007), (b) Greiner et al. (2011), (c) Zafar et al. (2011), (d) Cummings et al. (2007), (e) Cummings et al. (2008), (f)
Perley et al. (2011b), (g) Stamatikos et al. (2008), (h) Palmer et al. (2008), (i) Markwardt et al. (2008), (j) Baumgartner et al. (2009), (k) Ukwatta
et al. (2010a)
1.496
2.452
1.640
3.036
1.505
0.979
1.24
0.542
1.00
0.60
0.67
0.96
0.47
1.10
0.73
0.79
1.00 ± 0.07
1.10+0.14
0.67 ± 0.01
0.96+0.05
0.97 ± 0.05
1.12+0.02
0.73+0.09
1.29+0.11
123 / 108
15 / 14
428 / 327
70 / 39
23 / 18
48 / 63
33 / 31
138 / 124
−0.12
1.23+0.18
0.47 ± 0.03 / 1.20+0.09
3.3 ± 0.3 / 2.33+0.46
1.01+0.19
3.4+0.4
−0.3
1.4+1.1
−0.6
3.8+0.2
−0.2
−0.16/ 1.19 ± 0.15
−0.8/ < 2.9
−0.08/ 0.71 ± 0.08
−0.7/ 3.8+0.2
−0.4/ 1.2+0.4
−0.12
−0.4
−0.15
−0.10
−0.43
−0.06
−0.2
−0.5
−0.14/ 1.53 ± 0.13
−0.02
−0.07
−0.10
10−3
10−2
Observedenergy[keV]
10−1
100
101
10−3
10−2
10−1
100
101
102
Fν[µJy]
1101001000
18
20
22
24
26
28
30
Brightness[magAB]
Observedwavelength[nm]
Fig.5. NIR to X-ray SED of the afterglow of GRB 070306 ob-
tained approximately 120 ks after the trigger in the observer’s
frame. The dashed line shows the unabsorbed synchrotron con-
tinuum emission while the best-fit model (including dust and
metal absorption) is shown by solid lines. Upper limits are
shown by downward triangles.
glow is detected in all five redder bands, which implies an ex-
tremely red color of (R − K)AB ? 6mag and βoX < 0.44.
The combined GROND and XRT data are well fit with a sin-
gle power-law continuum, indicating that both the optical/NIR
and the X-ray regimes probe the same part of the synchrotron
spectrum. The obvious curvature in the GROND data is accu-
ratly described with either of the local dust models, with best-
fit parameters of AGRB
V
NH.X = (1.1 ± 0.1) × 1022cm−2at 90% confidence and a χ2of
48.1 for 63 d.o.f. in a MW-like extinction law. SMC and LMC
models yield within errors comparable parameters and provide
equally good fits to the data (χ2s of 48.2 and 48.8, respectively).
The SED of the afterglow of GRB 100621A at z = 0.542
(Milvang-Jensen et al. 2010) is shown in Fig. 7. Similarly to the
SED of GRB 081109 there is strong curvature and obvious red-
dening in the optical/NIR part of the SED. The inferred ultra-red
= 3.4+0.4
−0.3mag, β = 1.12 ± 0.02, and
10−3
10−2
Observedenergy[keV]
10−1
100
10−2
10−1
100
101
102
Fν[µJy]
1 101001000
18
20
22
24
26
28
30
Brightness[magAB]
Observedwavelength[nm]
Fig.6. Same as Fig. 5 for the afterglow of GRB 081109 obtained
approximately 60 ks after the trigger.
color of (R− K)AB∼ 5.8 mag, and the βoXvalue of 0.39 provide
evidenceforstrongdustextinction.Thebestfitisobtainedwitha
brokenpower-law withspectral indicesβ1= β2−0.5 = 0.79+0.11
as well as AGRB
V
= 3.8 ± 0.2mag for an LMC-like extinction
law, and NH.X = (1.62 ± 0.15) × 1022cm−2at 90% confidence
(χ2of 138 for 124 d.o.f.). Given the rest-frame coverage of
∼ 300−1500 nm all local dust models return comparable values
with AGRB
V
values of 3.8±0.2mag for an SMC- and 4.0±0.2mag
for a MW-type extinction law. All data bluewards and including
the r?filter are consistent with this extinction laws, while the
g?-band photometry is somewhat (2 − 3σ) brighter than the best
fit predicts. This could indicate a discrepancy between the de-
tails of the specific dust extinction law and local models similar
as observed e.g., in GRBs 070802 or 080607 (El´ ıasd´ ottir et al.
2009; Perley et al. 2011b).
The visual extinctions for GRBs 070306, 081109 and
100621A are among the largest ever measured directly along
GRB sight-lines, and further imply metals-to-dust ratio of
NH,X/AGRB
V
−0.10,
∼ (3 − 5) × 1021cm−2/mag. Compared to previous

Page 8

8T. Kr¨ uhler et al.: The SEDs and Host Galaxies of the dustiest GRB afterglows
10−3
10−2
Observedenergy[keV]
10−1
100
10−1
100
101
102
103
104
Fν[µJy]
110 1001000
14
16
18
20
22
24
26
28
Brightness[magAB]
Observedwavelength[nm]
Fig.7. Same as Fig. 5 for the afterglow of GRB 100621A ob-
tained approximately 7.6 ks after the trigger.
afterglow measurements, these NH,X/AGRB
low, and within a factor of 2-3 similar to NH/AVas observed in
the SMC, LMC and MW (see also Sect. 6.4). The results of af-
terglow SED fitting, as well as values taken from the literature,
are summarized in Table 5.
V
values are relatively
5.2. Host properties
5.2.1. The host of GRB 070306
The galaxy hosting the strongly extinguished GRB 070306 at
z = 1.496 was previously discussed in Jaunsen et al. (2008).
In addition to the public VLT imaging data (FORS R, ISAAC
JsHKs), the host of GRB 070306 was observed with GROND
(g?r?i?z?JHKssimultaneously), and its SED (Fig. 8) is further
complemented by published u and I-band data (Jaunsen et al.
2008). The host is bright (r?= 23.1mag), mildly red7with (R −
K)AB∼ 1.5mag and shows evidence of a 4000Å break. The data
are well fit with a non-extinguished (A?
and yield an absolute AB magnitude of MB= −22.4 ± 0.1mag,
which, at z ∼ 1.5 corresponds to ∼ 1.7L∗. The stellar mass of
log(M∗[M?]) = 10.4 ± 0.2 puts the galaxy among the most mas-
sive hosts compared to the sample of SGL09. The star-formation
rate estimate from the rest-frame UV flux is 13+11
gives a specific star formation rate (SSFR = SFR/M∗) of ∼ 0.5
Gyr−1, or growth timescale (i.e., 1/SSFR) of 2 Gyr. The SFR
is in reasonable agreement with the one derived from the [OII]
emission line (Jaunsen et al. 2008). We note, that the ISAAC
H-band host image was obtained only 2.5 days after the GRB,
and hence very likely contains a significant fraction of afterglow
light, explaining the blue H − K color. The physical properties
of the host, however, are comparable if fit with or without the
ISAAC H filter.
V< 0.6)8host template,
−4M?/yr, which
7We compare GRB host colors against the median (R − K)AB =
0.8mag color from the SGL09 sample
8For galaxies, we use A?
and effective reddening, since these quantities depend for example on
the topology of the ISM and galaxy geometry (Pierini et al. 2004), and
on galaxy scales are different from the corresponding values of a given
extinction law.
V, or E?
B−Vto indicate measured attenuation
5.2.2. The host of GRB 070802
The host of GRB 070802 at z = 2.452 was discovered in deep
FORS R and ISAAC K band imaging (El´ ıasd´ ottir et al. 2009).
The afterglow SED is characterized by a prominent 2175 Å dust
feature, and significant dust in the range of AV∼ 1mag (Kr¨ uhler
et al. 2008; El´ ıasd´ ottir et al. 2009). To construct the optical/NIR
host SED, the R ∼ 25.2mag host is further observed with
EFOSC/NTTiniandHAWK-I/VLTinthe J-band.Thegalaxyis
moderatelyred((R−K)AB∼ 1.8mag)anditsSED(Fig.8)shows
a 4000 Å break, but the age of the dominant stellar population is
not well constrained by the available data (? 1 Gyr). There is no
strong evidence for internal reddening, and the best-fit absolute
magnitude is MB= −21.4±0.2mag, which is ∼ 0.6L∗at z ∼ 2.5
with log(M∗[M?]) = 9.7+0.2
rived from the galaxy model fitting, an estimate for the SFR of
10+30
−0.3. Using the rest-frame UV flux de-
−7M?/yr, and the SSFR of ∼ 2 Gyr−1can be derived.
5.2.3. The host of GRB 080605
The host of GRB 080605 at z = 1.640 was discovered in late
GROND follow-up observations of the burst field 22 days af-
ter the GRB trigger. The afterglow SED shows evidence for
a 2175 Å dust feature, and significant AV in the range of ∼
0.5 − 1.3mag (Greiner et al. 2011; Zafar et al. 2011), and will
be further discussed in Nicuesa et al. (in preparation). The host
is bright (r?∼ 22.8mag), and blue with a flat g?− z?color, and
(R−K)AB∼ 0.5mag as estimated from the best fit galaxy model
(Fig. 8). The SED fit further yields MB= −22.6±0.2mag, which
is ∼ 2.1L∗at z ∼ 1.5, and log(M∗[M?]) = 9.6+0.3
nant stellar population of the host is young (τ = 0.06+0.18
and there is no evidence for reddening at the 2σ level (E?
0.22mag). The host galaxy is vigorously star-forming with a
SFR of 40+40
−0.2. The domi-
−0.03Gyr),
B−V?
−20M?/yr and a SSFR of ∼ 10 Gyr−1.
5.2.4. The host of GRB 080607
The afterglow of GRB 080607 is heavily reddened (AV ∼
3mag), has a modest 2175 Å dust feature and is characterized by
astrongneutralhydrogenabsorberwithroughlysolarmetallicity
and molecular gas (Prochaska et al. 2009; Perley et al. 2011b).
Data from Chen et al. (2011) show a R ∼ 27mag, very red host
with a synthetic color of (R − K)AB∼ 3mag (Fig. 8). The host
is well described with an extinguished galaxy template (A?
1 − 2mag), and physical parameters of MB= −21.1 ± 0.3mag,
which is ∼ 0.3L∗at z ∼ 3, log(M∗[M?]) = 9.9+0.4
tion corrected SFR of 40+60
These values are in good agreement with previously published
ones (Chen et al. 2011).
V∼
−0.6, an extinc-
−26M?/yr, and SSFR of ∼ 8 Gyr−1.
5.2.5. The host of GRB 080805
GRB 080805 had a very red afterglow, where both an SED
and spectral analysis showed a large dust column (AV ∼ 1.0 −
1.5mag), and evidence for a 2175 Å dust feature (Fynbo et al.
2009a; Greiner et al. 2011; Zafar et al. 2011). Its host is dis-
covered in late EFOSC/HAWK-I imaging in five filters (VRiJK,
see Fig. 8), is relatively bright (R ∼ 25.5mag) and red ((R −
K)AB = 2.5mag), with best-fit physical parameters of MB =
−20.4±0.2mag (∼ 0.3L∗at z ∼ 1.5) and log(M∗[M?]) = 9.7+0.2
The remaining galaxy properties are not well constrained by
the available data, yielding a limit on the galaxy reddening
−0.2.

Page 9

T. Kr¨ uhler et al.: The SEDs and Host Galaxies of the dustiest GRB afterglows9
200 200300 400300 400 600
Wavelength[nm]
1000100020002000
100
101
102
Fν[µJy]
600
19
20
21
22
23
24
25
Brightness[magAB]
HostGalaxyof GRB070306
3003004004006006001000100020002000
Wavelength[nm]
10−1
100
101
Fν[µJy]
21
22
23
24
25
26
27
Brightness[magAB]
HostGalaxyof GRB070802
300 400300 4006006001000 1000 20002000
Wavelength[nm]
100
101
102
Fν[µJy]
18
19
20
21
22
23
24
Brightness[magAB]
HostGalaxyof GRB080605
300 400300 400 600 6001000
Wavelength[nm]
200020004000 60004000 6000
10−2
10−1
100
101
Fν[µJy]
1000
20
21
22
23
24
25
26
27
28
29
30
Brightness[magAB]
HostGalaxyof GRB080607
300 300400400600600 1000100020002000
Wavelength[nm]
10−1
100
101
Fν[µJy]
20
21
22
23
24
25
26
27
Brightness[magAB]
HostGalaxyof GRB080805
150 200150 200300 400 300 400 600600 10001000 20002000
Wavelength[nm]
100
101
102
Fν[µJy]
18
19
20
21
22
23
24
25
Brightness[magAB]
HostGalaxyof GRB081109
200200300 400300 400600
Wavelength[nm]
1000100020002000
100
101
102
Fν[µJy]
600
18
19
20
21
22
23
24
25
Brightness[magAB]
HostGalaxyof GRB090926B
150 200150 200 300 400300 400 600 60010001000 20002000
Wavelength[nm]
101
102
Fν[µJy]
19
20
21
22
23
Brightness[magAB]
HostGalaxyof GRB100621A
Fig.8. SEDs of the host of in this sample and the best-fit galaxy model (solid line) in the observer’s frame. Filled black circles
represent photometric measurements, while downward triangles denote 3σ upper limits.

Page 10

10T. Kr¨ uhler et al.: The SEDs and Host Galaxies of the dustiest GRB afterglows
Table 6. Host galaxy properties
Hostzspec
zphot
R.A.(a)
J2000
09:52:23.31
02:27:35.72
17:28:30.05
20:56:53.43
22:03:09.63
03:05:13.91
21:01:13.08
Decl.(a)
J2000
+10:28:55.4
-55:31:39.0
+04:00:56.0
-62:26:39.3
-54:42:39.9
-39:00:22.6
-51:06:22.2
Offset(b)
arcsec
? 0.??2
0.??15 ± 0.??08
0.??23 ± 0.??11
? 0.??3
? 0.??2
? 0.??6
0.??12 ± 0.??08
Probability(c)
Instruments(d)
Filter
GRB 070306
GRB 070802
GRB 080605
GRB 080805
GRB 081109
GRB 090926B
GRB 100621A
Notes: Redshifts from Jaunsen et al. (2008), El´ ıasd´ ottir et al. (2009), Fynbo et al. (2009a), Fynbo et al. (2009b) and Milvang-Jensen et al. (2010).
The host of GRB 080607 is not shown in this table. All measurements were directly taken from Chen et al. (2010, 2011)
(a)Host position derived after tying the astrometric solution to the USNO-B1 catalog (Monet et al. 2003). Typical absolute uncertainties are ≈ 0.??3.
(b)Relative offset calculated by registering the host images against astrometric templates derived from afterglow images with a typical precision of
40 mas rms
(c)Estimated chance coincidence probability following Bloom et al. (2002) and Perley et al. (2009)
(d)G is GROND at the 2.2 m MPG/ESO telescope, E/S are EFOSC/SOFI at the NTT, and F/H/I are HAWK-I, FORS1/2 and ISAAC at the VLT,
and U UVOT onboard Swift respectively.
(f)Further complemented by the u and I band magnitudes in Cool et al. (2007) and Jaunsen et al. (2008).
1.496
2.452
1.640
1.505
0.979
1.24
0.542
1.64+0.35
2.3+1.5
−0.6
1.7+0.7
−0.2
1.7+0.6
−0.7
0.84+0.16
1.43+0.62
0.50+0.14
−0.13
0.002
0.011
0.002
0.030
0.002
0.018
0.0006
FGI
EFHI
G
EH
EFGHU
EFGS
EGHU
g?r?Ri?z?JGJHK(f)
RIJK
g?r?i?z?JH
VRIJK
Ug?Vr?i?Iz?YJGJHK
Ug?r?i?z?JHK
w2m2w1uUg?r?i?z?YJHK
−0.07
−0.26
−0.02
Table 7. SPS Host galaxy fitting results
HostRedshiftMB
Age
Gyr
1.6+1.6
−1.0
0.38+0.66
0.06+0.18
0.16+0.53
0.51+0.74
0.24+0.22
0.14+0.60
0.05+0.07
E?
mag
B−V
log(M∗)
M?
10.39+0.19
9.7+0.2
−0.3
9.6+0.3
−0.2
9.9+0.4
−0.6
9.7+0.2
−0.2
9.82+0.09
10.1+0.4
8.98+0.14
log SFR
M?/yr
1.1+0.3
−0.2
1.0+0.6
−0.5
1.6+0.3
−0.3
1.6+0.4
−0.5
0.8+0.7
−0.8
1.5+0.2
−0.2
1.9+0.4
−0.5
1.13+0.15
log SSFR
yr−1
−9.3+0.4
−8.7+1.0
−8.0+0.5
−8.2+0.6
−8.9+0.9
−8.4+0.3
−8.1+0.6
−7.9+0.2
L
L∗
χ2/Nfilt
magAB
−22.4 ± 0.1
−21.4 ± 0.2
−22.6 ± 0.2
−21.1 ± 0.1
−20.4 ± 0.2
−21.27 ± 0.09
−21.5 ± 0.1
−20.68 ± 0.08
B
GRB 070306
GRB 070802
GRB 080605
GRB 080607
GRB 080805
GRB 081109
GRB 090926B
GRB 100621A
1.496
2.452
1.640
3.036
1.505
0.979
1.24
0.542
< 0.16 (2σ)
< 0.42 (2σ)
< 0.22 (2σ)
0.35+0.25
< 0.65 (2σ)
0.24+0.06
0.35+0.08
0.14+0.03
−0.15
−0.3
−0.6
−0.6
−0.7
−0.9
−0.3
−0.9
−0.3
1.7 ± 0.2
0.6 ± 0.2
2.1 ± 0.4
0.3 ± 0.1
0.3 ± 0.1
0.9 ± 0.1
0.9 ± 0.1
0.6 ± 0.1
11.8/12
2.3/4
3.3/6
3.9/5
0.3/5
11.2/14
7.3/9
16.5/14
−0.31
−0.03
−0.12
−0.40
−0.11
−0.09
−0.03
−0.09
−0.04
−0.05
−0.04
−0.09
−0.5
−0.10
−0.20
E?
a SFR of 6+25
At a distance of 2.5??north-east of the afterglow/host posi-
tion, there is another R ∼ 25mag, and even redder ((R − K)AB∼
4mag) galaxy, a plausible candidate for the strong MgII absorb-
ing system at z = 1.20 reported in Fynbo et al. (2009a).
B−V? 0.7mag,anageofthestellarpopulationof0.53+0.76
−5M?/yr, and SSFR of ∼ 1 Gyr−1.
−0.42Gyr,
5.2.6. The host of GRB 081109
GRB 081109 is the only burst in the sample where no spectro-
scopic redshift was available in the literature. However, the host
is bright (r?∼ 22.7mag), moderately red ((R− K)AB= 1.6mag)
and our host spectrum (see Fig. 9) reveals a single emission line
above a well-detected continuum. Within the wavelength cov-
erage of the spectrum (∼ 370 − 950 nm), this emission line is
interpreted as [OII][λ3727] at z = 0.9787 ± 0.0005. If it were
any of the other prominent nebular lines (Hβ, [OIII], Hα), we
would have expected to detect [OII][λ3727] in the spectrum as
well. At this redshift, there is further spectroscopic evidence for
a Balmer break, and a tentative absorption of the CaII HK dou-
blet (see Fig. 9).
The host SED is shown in Fig. 8 and well fit with a young
(τ = 0.18+0.19
population. The absolute magnitude of MB= −21.27±0.09mag
corresponds to ∼ 0.9L∗at z ∼ 1. The stellar mass obtained from
the SPS fit is log(M∗[M?]) = 9.80+0.09
M?/yr, which together yields a SSFR of ∼ 5 Gyr−1. The emis-
sion line flux of [OII][λ3727] is (1.8 ± 0.2) × 10−16erg/cm2/s,
which includes a systematic error contribution from the abso-
−0.07Gyr) and reddened (A?
V= 1.0 ± 0.2mag) stellar
−0.09, with a SFR of 33+19
−13
500600700800900
Wavelength[nm]
0
5
10
15
20
Fν[µJy]
[OII]
K H
Fig.9. Wavelength and flux-calibrated FORS2 300V spectrum
of the host of GRB 081109 in black. The thin grey line shows the
sky spectrum. The red line is the best-fit galaxy model obtained
fromtheavailable photometryandredpointsarethe photometric
measurements.
lute flux normalization. This corresponds to a dust un-corrected
star-formation rate of 13 ± 4 M?/yr (Kennicutt 1998), or 48+18
M?/yr when using the upper A?
agreement with the SFR from the UV continuum.
−16
Vmeasurement, which is in good

Page 11

T. Kr¨ uhler et al.: The SEDs and Host Galaxies of the dustiest GRB afterglows 11
5.2.7. The host of GRB 090926B
Promptly after the trigger, Fynbo et al. (2009b) reported the de-
tection of the host galaxy of GRB 090926B based on a promi-
nent [OII] emission line in a VLT/FORS spectrum. The host
is also imaged in later GROND and NTT observations with a
brightness of r?∼ 23.0mag, and a mildly red color ((R−K)AB∼
1.5mag) as shown in Fig. 8. The host SED is well fit with an ex-
tinguished host model (A?
(∼ 0.9L∗at z ∼ 1.3) and log(M∗[M?]) = 10.1+0.6
based on the extinction corrected UV flux is 80+110
is among the highest ever measured from optical data, but not
well constrained given the uncertainty on the dust extinction
properties. The constraints on the age of the stellar population
and SSFR are weak, with values of τ = 0.14+0.60
SSFR of ∼ 7 Gyr−1, respectively.
V= 1.4+0.3
−0.2mag) with MB= −21.5±0.2
−0.5. The SFR
−50M?/yr which
−0.09Gyr and the
5.2.8. The host of GRB 100621A
The host of GRB 100621A was reported very early after the
trigger as a DSS2 source, providing a constant contribution to
the afterglow in the UV/blue light curve (Updike et al. 2010;
Oates & Ukwatta 2010). In fact, the redshift of z = 0.542 of
GRB 100621A is based on emission lines from a bright host
(Milvang-Jensen et al. 2010). The SED of the r?∼ 21.5 mag
host is well-sampled from the UV to the NIR and shown in
Fig. 8. In strong contrast to the extremely red afterglow (R −
K)AB ∼ 5.8 mag, the host is blue with an inferred color of
(R − K)AB ∼ 0.3mag and (uvw2 − K)AB ∼ 0.9mag. The
SPS host fit returns an intrinsic extinction of A?
for a very young stellar population of age τ = 0.05+0.07
This host has the lowest stellar mass in the presented sample
with log(M∗[M?]) = 8.98+0.14
MB= −20.68±0.08mag, which is ∼ 0.6L∗at z ∼ 0.5. The SFR
and SSFR are 13+6
V= 0.6+0.1
−0.2mag
−0.03Gyr.
−0.10, and an absolute magnitude of
−5M?/yr and ∼ 14 Gyr−1, respectively.
6. Discussion
6.1. Dust reddening in GRB afterglow SEDs
The visual extinction measured from X-ray-to-NIR SED-fitting
towards GRBs 070306 (AGRB
V
3.4+0.4
V
the largest ever derived from optical/NIR data for GRB after-
glows (e.g., Savaglio et al. 2003; Kann et al. 2006; Greiner
et al. 2011). They clearly show that a large dust column can
be the dominating feature in a GRB afterglow SED. The dust
properties inferred from afterglow measurements are well rep-
resented with local models in the rest-frame 300-1100 nm, and
at the resolution of broad-band imaging (see also e.g., Fynbo
et al. 2001; Watson et al. 2006; Kann et al. 2006; Schady et al.
2007; Starling et al. 2007), while noteworthy exceptions exist in
the literature (e.g., Savaglio & Fall 2004; Perley et al. 2010b;
Clemens et al. 2011). The good fit provided by local dust ex-
tinction laws further suggests an abundance of small dust grains
comparable to the MW/LMC or SMC. There is hence no di-
rect evidence that the dust towards these GRBs through their
hosts is different than observed along local sight-lines. A differ-
ent dust grain size distribution would have been expected if the
dust were located in the immediate vicinity (R ? 1019cm) of
the GRB and shaped through its intense radiation, i.e., through
dust destruction (Waxman & Draine 2000; Fruchter et al. 2001;
Draine & Hao 2002). In addition, the metals-to-dust ratios for
these afterglows are only a few times the Galactic value of
= 5.5+1.2
= 3.8 ± 0.2mag) are among
−1.0mag), 081109 (AGRB
V
=
−0.3mag) and 100621A (AGRB
NH/AV ≈ 2 × 1021cm−2/mag (Predehl & Schmitt 1995). For
un-extinguished GRB sight-lines this ratio is generally found to
be a factor of 10-100 times larger than those of the Magellanic
Clouds or Milky Way (e.g., Galama & Wijers 2001; Stratta et al.
2004). This is suggestive of a dependence of the metals-to-dust
ratio on the amount of visual extinction. We will return to this
issue in Sect. 6.4.
6.2. The hosts of the dustiest afterglows
The general properties of the selected GRB host galaxies are di-
verse. They have (R − K)ABcolors ranging from flat and blue
(R − K)AB ∼ 0mag to extremely red (R − K)AB ∼ 3mag with
an average color of ?(R − K)AB? = 1.6mag, and host extinction
values between A?
lar mass and absolute magnitude distributions are broad, with
values between log(M∗[M?]) = 9.0 to log(M∗[M?]) = 10.4
(?log(M∗[M?])? = 9.8 ± 0.4) and MBbetween −20.3 mag and
−22.6 mag (?MB? = −21.3 ± 0.6mag). These absolute bright-
nesses are in a range between several tenths to few L∗(?L? =
0.9L∗) as compared to the general field galaxy population at the
same redshift. The average SFR and SSFR are about 30 M?/yr
and ?logSSFR[yr]? ∼ −8.3, respectively. The average growth
time is ∼ 0.2 Gyr, which illustrates that not only optically-
selected hosts, but also those of highly-reddened afterglows are
very efficient in producing stars. A rough estimate on the metal-
licity of the hosts can be obtained if these GRB hosts follow
the fundamental plane as defined from nearby SDSS galaxies
(Mannucci et al. 2010a). With a given stellar mass and SFR, the
host galaxies in this sample are expected to have metallicities in
a range between 12 + log(O/H) ∼ 8.2 and 12 + log(O/H) ∼ 8.9,
with an average of 12 + log(O/H) ∼ 8.6. We caution that the
SFRs were derived using the rest-frame UV flux, which is quite
sensitive to the dust extinction properties.
Although not well-constrained in all cases, the average
luminosity-weighted effective reddening inferred from host pho-
tometry is typically smaller or equal to that measured from af-
terglow observations. This is not a particularly surprising result,
given that the sample selection was based on high visual extinc-
tions of the afterglow SEDs in the first place. It directly indicates
some variation in the dust distribution of the hosts, which again
is not a surprising result, given the differences in extinction prop-
erties along different sight-lines through the diffuse ISM to Giant
Molecular Clouds in the Local Group (e.g., Gordon et al. 2003;
Fitzpatrick 2004), and the geometrical differences between a sin-
gle sight-line and an extended distribution of star-light and dust
(e.g., Gordon et al. 1997; Silva et al. 1998).
One intriguing case is the host of GRB 100621A. Although
having one of the most extinguished afterglows ever detected
(even in the presented sample), its host shows very blue colors,
and is one of the youngest and the least massive galaxy in this
work. This particular example provides evidence for a patchy
dust component where the geometry of the dust distribution and
not the properties of the host galaxy makes the single GRB sight-
line dust-enriched.
V∼ 0mag and A?
V∼ 2mag. Also, their stel-
6.3. Comparison to previous GRB host samples
One key result of this study is the success rate of the discovery
of the selected hosts. Out of eight hosts, which were selected
given their afterglow properties (hence a selection independent
on host properties, in particular galaxy brightness), all are lu-
minous enough to detect in optical ground-based imaging. This

Page 12

12T. Kr¨ uhler et al.: The SEDs and Host Galaxies of the dustiest GRB afterglows
fraction is significantly larger than expected from a host sam-
ple based on XRT detections (Fynbo et al. 2008; Malesani et al.
2009). The effect of an increased detection rate is even stronger
intheNIR:Sevenoutofeightaredetectableinthe K-band,while
this fraction is only ∼35% for the general host population of
Swift/GRBs (Malesani et al. 2009). Partially, this is the result of
the lower average redshift of the selected hosts (?zAV? = 1.5) as
compared to all Swift GRBs with ?zSwift? ∼ 1.9.
Thelowerredshiftishowevernottheonlyreasonforthehigh
detection rate. The selected hosts are on average redder and, as
shown in Fig. 10, have typically higher luminosities and stel-
lar masses than the (sensitivity-limited) SGL09 sample which
has ?(R − K)AB? = 0.8mag, ?MB? = −19.6 ± 1.5mag and
?log M∗[M?]? = 9.1 ± 0.6. A two-sample K.-S. test returns p-
values of 0.002 for the stellar mass, and 0.006 for the absolute
magnitude distributions respectively, which is tentative evidence
that both distributions are not drawn from the same parent sam-
ple. However, given the small sample size of only eight high-AV
events, larger samples are required to statistically establish the
existence of a difference at higher significance. Of course, both
distributions are drawn from the same physical parent sample
(GRB hosts), indicating that the different selection criteria probe
different host properties.
A possible explanation of the different host properties would
bethenowon-averagehigherredshiftrelativetotheSGL09sam-
ple, where star-formation was driven by more massive galax-
ies as compared to the more nearby Universe (e.g., Cowie et al.
1996; Hopkins 2004). To test this hypothesis, we selected a sub-
sample from SGL09 with a median redshift comparable to the
hosts in this work. This essentially removes all z < 1 SGL09
hosts and leaves only 13 events for comparison (see histograms
in Fig. 11). Despite the small number statistics, the M∗and MB
valuesareagainplacedatthehigh-massandhigh-luminosityend
of their respective distribution, and a K.-S. test is also marginally
suggestive of a difference (p-values of 0.001 and 0.034 for the
masses and absolute magnitudes).
We conclude that by selecting extinguished afterglows
we are very likely probing a more luminous, massive and
chemically-evolved population of GRB hosts.
As it is clear that these were largely missing from previous
samples due to their poor localizations, there is a selection bias
and the host population is missing most of its massive, evolved
and metal-rich members. As a direct consequence, GRB hosts
trace the global SFR closer than indicated in studies which are
based on host samples of optically selected GRB afterglows, and
the apparant deficiency of high-mass host galaxies is at least par-
tially a selection effect.
Similar conclusions apply for all galaxies hosting afterglows
that show a significant 2175 Å dust feature in their SED. Four
out of five currently known afterglows are within the presented
sample, which argues for a direct connection between large dust
columns and the presence of the UV bump (see e.g., Greiner
et al. 2011; Zafar et al. 2011). Their, on average, more mas-
sive and luminous hosts suggest a qualitative relation between
the stellar mass of a galaxy and the presence of a 2175 Å fea-
ture, where the latter is only present in fairly massive and
metal-enriched galaxies (see also e.g., Noll & Pierini 2005).
Conversely, a strong 2175 Å feature in an afterglow SED is also
very likely a good proxy for the stellar mass and luminosity of
the GRB host.
7.58.08.5
log(StellarMass[M?])
9.09.5 10.010.5
0
2
4
6
8
10
12
Numberof galaxies
−22
−20
−18
−16
MB[magAB]
0
2
4
6
8
10
12
14
Numberof galaxies
Fig.10. Distribution of stellar masses and luminosities of the
hosts of highly extinguished afterglows (blue) and the host sam-
ple from SGL09 (grey).
8.59.0
log(StellarMass[M?])
9.5 10.010.5
0
1
2
3
4
5
Numberof galaxies
M∗
0.1M∗
0.01M∗
?z? ∼ 1.5
−23
−22
−21
−20
−19
−18
MB[magAB]
0
1
2
3
4
5
6
Numberof galaxies
L∗
0.1L∗
?z? ∼ 1.5
Fig.11. Distribution of stellar masses and luminosities of the
hosts of highly extinguished afterglows (blue) and a subsample
of SGL09 (grey) with ?z? ∼ 1.5.
6.4. Metals-to-dust ratios in context
The ratio between the line-of-sight extinction and the total metal
column for GRB afterglows has been investigated in a num-
ber of papers (e.g., Galama & Wijers 2001; Stratta et al. 2004;
Savaglio & Fall 2004; Kann et al. 2006; Schady et al. 2007,
2010; Greiner et al. 2011; Zafar et al. 2011) where ratios typ-
ically much higher than the ones observed in the Local Group
were derived. Measurements for different Galactic sight-lines
(e.g., Predehl & Schmitt 1995; G¨ uver &¨Ozel 2009) show an
almost universal value of around NH/AV ≈ 2 × 1021cm−2/mag,
while the matter probed by afterglows can yield metals-to-dust
ratios up to and sometimes even above 100-times higher (e.g.,
Watson et al. 2007; Rau et al. 2010).

Page 13

T. Kr¨ uhler et al.: The SEDs and Host Galaxies of the dustiest GRB afterglows 13
10−1
100
NH,X/AV[1022cm−2/mag]
101
102
10−2
10−1
100
101
AV[mag]
GRB080607
GRB080605
AV> 1
Greineretal.(2011)
Schadyet al.(2010)
Zafaretal.(2011)
Perleyetal.(2011)
Thiswork
1 10100
NH,X/AV[GalacticUnits]
Fig.12.Metals-to-dustratioversussight-lineextinctionforGRB
afterglows. The horizontal dashed lines marks the selection cri-
terion for GRBs to enter this sample. Vertical dashed lines il-
lustrate metals-to-dust ratios of 1, 10 or 100 times the Galactic
value. Solid lines show the toy model of two physically inde-
pendent absorbers, where one is fully devoid of dust with a H-
equivalent metal column of NH,X= 1021.5cm−2and represented
by the dotted line, while the other is neutral and has metals-to-
dust ratios of 1, 2 or 3 times the Local Group value. Two in-
dividual cases (GRBs 080605 and 080607) illustrate the scatter
between the analysis of different data sets.
6.4.1. An anti-correlation between metals-to-dust ratio and
sight-line extinction
Figure 12 shows the NH,X/AGRB
terglows and illustrates its dependence on the sight-line dust ex-
tinction. With the afterglows in this work, there is now for the
first time reasonable coverage in the AGRB
Intriguingly, the metals-to-dust ratio is strongly anti-correlated
with AGRB
V
, confirming the tentative trend reported by Perley
et al. (2009). A Spearman rank-order correlation analysis for the
combined sample in Figure 12 returns a correlation coefficient
ρ = −0.63, with a two-tailed p-value of 3 × 10−7, in strong con-
trast to a constant, universal NH,X/AGRB
There are two straight forward ways to reconcile this result:
One is a dependence of the metals-to-dust ratio on the specific
environment such that evolved and dust-enriched hosts are more
efficient in forming dust out of their metals (and in fact we do
observe on average larger stellar masses for the hosts of high
AGRB
V
afterglows). The other is the presence of two physically
independent absorbers, where the first dominates the total metal
column, the other the visual extinction measurements. This triv-
ially produces a non-correlation between NH,X and AGRB
consequently an NH,X/AGRB
V
Some outliers of the metals-to-dust anti-correlation (Fig. 12)
might be explained with difficulties of measuring the respective
physical parameters. This is also illustrated by the example of
two individual events (GRBs 080605 and 080607) where differ-
ent values have been published in the literature. Assumptions on
the continuum emission and the extinction law, notably the total-
to-selective reddening RV, but also its parametrization can affect
the AGRB
V
measurement. In addition, there is the possibility that
the cooling break is located close to or within the range of the
UV/optical/NIR measurements. In a standard analysis, the intro-
V
ratio for a large number of af-
V
∼ 1 − 5mag range.
V
ratio.
V
, and
to AGRB
V
anti-correlation.
100
101
NH,X/AV[1022cm−2/mag]
8.5
9.0
9.5
10.0
10.5
log(StellarMass)[M?]
GRB 100621A
GRB 070306
GRB 081109
110
NH,X/AV[GalacticUnits]
Fig.13. Total metals-to-dust ratios for GRB afterglows versus
stellar mass of their host galaxies. Black data are galaxies host-
ing a highly-extinguished afterglow using the first values of
Table 5, where the three hosts with the highest AGRB
beled. The shaded area indicates the probability distribution for
optically-selected GRBs, represented by a log-normal distribu-
tion of NH,X/AGRB
V
based on Schady et al. (2010) and a Gaussian
distribution in log(M∗[M?]) based on SGL09.
V
are la-
duced curvature caused by the spectral break is then interpreted
asanincreaseddustcolumn(Kr¨ uhleretal.2011).The NH,Xmea-
surements are prone to errors as well: spectral variation intrinsic
to the afterglow can lead to wrong estimates on the soft X-ray
absorption (e.g., Butler & Kocevski 2007).
6.4.2. Metals-to-dust ratio compared to host mass
As shown in Sect. 6.2, the hosts of dusty afterglows are on av-
erage more massive and luminous than their non-extinguished
counterparts, but there is a broad range of galaxy properties and
the only common feature between all afterglows/hosts in this
work is hence the dusty line of sight. In particular, if the envi-
ronment were responsible for the observed NH,X/AGRB
extinction anti-correlation, we would expect the metals-to-dust
ratio for GRB 100621A to be comparable to the bulk of optically
bright afterglows. It is, however, one of the lowest ever observed
for GRB afterglows and a factor of 5 lower than the median for
afterglows with hosts of similar mass (see Fig. 13). Although we
note that number counts are still too low to derive strong con-
straints with high statistical significance, this suggests that the
specific host environment is not responsible for the observed de-
pendence on the metals-to-dust ratio to the visual extinction.
V
to visual
6.4.3. A system of two absorbers
In the second scenario, two to first order physically independent
(one neutral, host-galaxy related, one ionized, circumburst spe-
cific) columns of material contribute to the observed absorption,
denoted as NH,neutral, NH,ionin the following. NH,Xmeasures the
sum of both, whereas the AGRB
V
column would only be associated
with the NH,neutralabsorber. Such a circumburst environment is
not unexpected: The intense afterglow radiation should not only
photo-ionize the vicinity of the burst, but also destroy the asso-
ciated dust to large amounts (Waxman & Draine 2000; Draine &
Hao 2002; Perna et al. 2003), albeit with different effective radii.

Page 14

14T. Kr¨ uhler et al.: The SEDs and Host Galaxies of the dustiest GRB afterglows
For AGRB
results in a large metals-to-dust ratio. With an increasing AGRB
column of around 1mag, the NH,neutralabsorber contributes sig-
nificantly to the total metal column (NH,neutral ≈ NH,ion): this is
illustrated in Fig. 12 by the solid lines, where the NH,X/AGRB
tio asymptotically reaches 1, 2 or 3 times the Local Group value.
Even larger AGRB
V
columns than present in this work can test
this hypothesis. For an AGRB
V
? 10mag sight-line, for example,
NH,neutralis much larger than NH,ionand the expected metals-to-
dust ratio would be comparable to its intrinsic value, and to the
Local Group value (if universal).
For the bulk of standard, un- or mildly extinguished after-
glows, the large column of ionized metals with an equivalent
NH,ion∼ 1021−22cm−2in the circumburst material dominates the
total NH,X measurement, whereas the visual extinction is very
likely caused by dust further out, either in the diffuse ISM, or
localized in interstellar clouds (see also Sect. 6.5). Hence, for
optically bright GRBs the total metal absorption as probed by
the soft X-ray absorption is neither a good measure nor a di-
rect tracer of the dust extinction along the line of sight. Or, in
other words, an NH,Xcolumn even as large as 1022cm−2does
not necessarily imply a significant visual extinction (see also,
e.g., Galama & Wijers 2001; Kann et al. 2006; Schady et al.
2007).
Also, the shape of the extinction law, which is typically con-
sistent with sight-lines through galaxies of the Local Group ar-
gues against an association of the absorbing dust with the im-
mediate vicinity of the burst. The steepness of the UV rise of
the extinction law as probed by GRB afterglows is comparable
to the one of the LMC (Schady et al. 2011b), generally sugges-
tive of an abundance of small dust grains. These grains, however,
are expected to be destroyed first and the grain-size distribution
function would be skewed to larger grains. This would result in
flat, or grey extinction properties, in contrast to the ultra-red af-
terglows of GRB 081109, or GRB 100621A, for example.
We thus conclude that the anti-correlation between metals-
to-dust ratio and sight-line extinction indicates the presence of
two absorbing systems, which are to first order physically inde-
pendent. One of them is dusty, the other ionized and dust-free,
where the former is probed by the optical/NIR data and the latter
typically dominates the NH,Xmeasurement.
V
∼ 0.1mag sight-lines, NH,neutral? NH,iondirectly
V
V
ra-
6.5. Location and geometry of the absorbing dust column
A natural question about the nature of the absorbing dust, gas
and metal columns detected in the afterglow SEDs and spectra
is their locations, and if they are directly related to the burst en-
vironment. A number of previous studies have already revealed
some clues about the geometry of the absorbing matter, which
have been derived quite exclusively from sight-lines with low to-
tal dust content: The distance of the cold-neutral material, linked
to the DLA and the low-ionization metal absorption lines has
been constrained to few hundreds of pc to few kpc (Prochaska
et al. 2006; Vreeswijk et al. 2007). In contrast, the high metal
column densities as derived from soft X-ray absorption were as-
sociated with a fully ionized circumburst medium up to few to
several tens of pc (Watson et al. 2007; Schady et al. 2011a). For
bursts with largely unextinguished afterglows, most of the met-
als along the sight-line are typically in a highly ionized state and
only ?10% of the absorbing gas is neutral (Schady et al. 2011a).
In addition, there is no statistically significant correlation be-
tween the soft X-ray absorption and the dust column as inferred
from optical/NIR data nor the metallicity of the neutral material
(e.g., Schady et al. 2007; Zafar et al. 2011), nor the darkness
of a GRB (Campana et al. 2010). There is however a trend of
higher visual extinctions with larger neutral metal columns, an
anti-correlation between gas-to-dust ratio and metallicity (Zafar
et al. 2011), and dark bursts have stronger neutral metal absorp-
tion lines in their optical spectrum (Christensen et al. 2011).
The dustiest afterglows and their hosts add two further hints.
Firstly,thereisthepreviouslydiscussedanti-correlationbetween
metals-to-dust ratio and sight-line extinction, albeit with a large
scatter for individual events. And secondly, their hosts are on av-
erage redder, more luminous and massive, and supposedly also
more evolved, dust- and metal-rich than their low-AGRB
parts (e.g., Kobulnicky & Kewley 2004; Savaglio et al. 2005).
There is hence a relation between the dust along the sight-line
towards the GRB and physical host properties. Such a relation is
expected if the dust probed by the afterglow is located at large
enough distances to be fairly representative of the size of the
host galaxy, its global dust enrichment and chemical state. In
contrast, dust in the surrounding environment of a GRB, even if
present and able to survive the intense afterglow or progenitor
radiation, would rather be related to the very specific details of
its circumburst environment or the GRB’s birth cloud.
An overlap with optically selected hosts clearly exist: very
young, blue and low-mass galaxies possess sight-lines through
dusty regions as demonstrated through the afterglows and hosts
of, e.g., GRBs 080605 and 100621A. For this kind of events
the data are clearly not consistent with a uniform dust-shield. In
a few cases the observations rather indicate a patchy dust dis-
tribution where the dust is located in clumps of small enough
covering factor and large enough extinction to be negligible in
the integrated host light distribution. The variations on the vi-
sual extinction in some cases are therefore a geometrical effect
(see also, e.g., Berger et al. 2003; Perley et al. 2009).
This raises questions about the validity of star-formation es-
timates obtained for young and blue galaxies. Afterglow obser-
vations show that there is extinguished star-formation ongoing
even in apparently unextinguished galaxies. Here, either the cov-
ering factor of the dusty clump is just too small to remove signif-
icantly from the host light, or the clump completely absorbs all
UV light from the star-forming region, thus having a negligible
effectonthehostgalaxycolors.Inthelattercase,theUV-derived
SFRs would strictly represent lower limits on the ongoing star-
formation in these galaxies. Far-infrared and sub-mm observa-
tions with Herschel and ALMA, respectively, would enable to
measure the galaxies SFR, dust mass and temperature and would
help to clarify this issue.
A coherent picture of highly extinguished afterglows in com-
bination with their diverse, but on average redder and more mas-
sive hosts could be obtained within a complex dust distribution
made out of several constituents related to extinction in the dif-
fuseISM,extendedinterstellarclouds,orlocalizedinfairlycom-
pact and dense regions such as giant molecular clouds.
The dust extinction is hence very likely not directly associ-
ated with the GRB environment but plausibly with the neutral
absorber at distances of a few hundred pc to one kpc. We stress,
however, that the effective radii of dust destruction and photo-
ionization will shape the detailed gas-to-dust and metals-to-dust
ratios, adding further complexity to the absorbing system(s) in
front of GRB afterglows. Furthermore, the dust distribution in
high-z galaxies could be even more complex due to dusty galaxy
outflows, and reflect the absence of a uniform chemical enrich-
ment on scales up to one kpc (e.g., Noll et al. 2009).
V
counter-

Page 15

T. Kr¨ uhler et al.: The SEDs and Host Galaxies of the dustiest GRB afterglows 15
7. Conclusions
The afterglows of GRBs 081109 (AGRB
100621A (AGRB
V
= 3.8 ± 0.2mag) join the growing sample of
highly extinguished events. Their continuum emission is well-
constrained by the combination of X-ray and NIR data, and the
optical observations provide a detailed measurement of the dust
properties along the sight-lines. While some diversity in their
extinction properties, particularly dust abundance, clearly ex-
ists, GRBs 081109 and 100621A provide compelling evidence
that a highly obscured afterglow is also a highly reddened one,
and that extinction laws derived from local sight-lines accurately
estimate the dust properties towards even highly extinguished
GRBs.
Having a large enough sample of coeval afterglows with
multi-wavelength data would ideally enable to advance from sin-
gle sight-line, pencil beam investigations to a statistically sym-
metric geometry where each GRB afterglow represents a differ-
ent sight-line through its host galaxy. In analogy to the case stud-
ied by Witt & Gordon (1996), this could provide a good descrip-
tion of the structure and evolution of the absorbing medium and
helpconstraintheopacityandfillingfactorsofthedustgeometry
and clumps from the distribution of AGRB
galaxies out to very high redshift.
The hosts of the dustiest afterglows provide a different pic-
ture of GRB host galaxies compared to the hosts of optically-
selected bursts. Although both samples are overlapping in their
properties, the galaxies in this work have typical luminosities of
around L∗and stellar masses of M∗∼ 1010M?, more luminous
and massive when compared to the hitherto discovered faint and
blue hosts. Although the number counts are still low, this work
indicates that a selection based on a large AGRB
erentially the more massive and chemically-evolved GRB hosts,
which is in qualitative agreement with searches for dark GRB
hosts (Perley et al. 2010a, 2011a; Rossi et al. 2011b).
This suggests that the properties of complete GRB host sam-
ples are diverse, and complex selection biases are still present:
not only are the very faintest GRB hosts missing due to inherent
sensitivity limits, but also some of the brightest, most luminous,
and chemically evolved ones, because they have not been lo-
calized accurately enough. Fairly large and massive, dusty, and
metal-rich galaxies are able to host GRBs, and the trend of low-
metallicity GRB hosts is not as significant as claimed in previous
studies, and possibly a selection effect of the young galaxy pop-
ulation dominating the global SFR at low-z (e.g., Berger et al.
2007; Mannucci et al. 2010b). This has substantial implications
for the feasibility of tracing the star-formation history with GRB
hosts and also for the progenitor channels of GRB production as
a result of their metallicity dependence. In the former case, this
work indeed indicates that the deficiency of high-mass GRB host
galaxies in previous studies was at least partially due to a selec-
tion bias. The latter case, however, depends quite strongly on the
assumption that the host-inferred metallicities are representative
of the composition of the progenitor star, while different sight-
lines through a GRB host can show a dispersion in metallicity of
around a factor 100 (Pontzen et al. 2010).
Intriguingly, all GRBs with AGRB
dust ratios significantly below what is typically measured for
GRB afterglows, and more in line with measurements from the
Local Group. In addition, there is a strong anti-correlation be-
tween the metals-to-dust ratio and the visual extinction along
the GRB sight-line. This effect seems independent on the spe-
cific host properties and can be interpreted as evidence of two
physically independent absorbers: dust-free, ionized metals in
V
= 3.4+0.4
−0.3mag) and
V
values in star-forming
V
picks up pref-
V
? 4mag have metals-to-
the circumburst environment (and probed by soft X-ray absorp-
tion), and in contrast a dusty absorber further out (and probed by
reddening measurements in the UV/optical/NIR).
A dust column independent of the immediate circumburst
environment is further supported by the relation between af-
terglow AGRB
V
and host properties, in particular the on-average
higher stellar mass and redder colors. Coupled with the blue
and very young hosts of, e.g., GRBs 080605 or 100621A it in-
dicates a complex dust geometry with different constituents in
the diffuse ISM and in localized patches, which are plausibly
associated with the cold-neutral absorber detected in rest-frame
UV/optical spectra.
Further advances can now be made by getting direct observa-
tional access to more dusty sight-lines including AGRB
events at an increasing redshift interval. Similar observations for
a large enough sample would investigate the dependence of the
global dust enrichment with cosmic evolution, and constrain the
fraction of dust-enshrouded star-formation out to very high red-
shifts. A sophisticated observational strategy coupled with state-
of-the-art instrumentation makes such a challenging study fea-
sible: A rapid response of the order of several minutes by a
NIR imager at an 8m-class telescope would have enabled the
detection of the afterglows of GRB 100621A (z ∼ 0.5) up to
AGRB
V
and GRB 070802 (z ∼ 2.5) up to AGRB
quently could have been followed-up using NIR spectroscopy.
Once an accurate localization as well as detailed information
about the GRB sight-line is available, their hosts should be read-
ily accessible for multi-wavelength surveys via large ground-
and space-based facilities, yielding information about otherwise
fully extinguished environments and unprecedented insights into
the conditions of star-forming galaxies throughout the Universe.
V
? 10mag
≈ 30mag, GRB 081109 (z ∼ 1) up to AGRB
V
≈ 20mag,
V
≈ 10mag, which subse-
Acknowledgements. We thank the referee for valuable comments. TK acknowl-
edges support by the DFG cluster of excellence ’Origin and Structure of the
Universe’ and support by the European Commission under the Marie Curie IEF
Programme in FP7. Part of the funding for GROND (both hardware as well as
personnel) was generously granted from the Leibniz-Prize to Prof. G. Hasinger
(DFG grant HA 1850/28-1). The Dark Cosmology Centre is funded by the
Danish National Research Foundation. PS acknowledges support by DFG grant
SA 2001/1-1. SS acknowledges support through project M.FE.A.Ext 00003 of
the MPG. SK, DAK & ANG acknowledge support by DFG grant Kl 766/16-1.
ARo acknowledges support from the BLANCEFLOR Boncompagni-Ludovisi,
n´ ee Bildt foundation. SMB acknowledges support of a European Union Marie
Curie European Reintegration Grant within the 7th Program under contract num-
ber PERG04-GA-2008-239176. This work made use of data supplied by the UK
Swift Science Data Centre at the University of Leicester.
References
Abazajian, K. N., Adelman-McCarthy, J. K., Ag¨ ueros, M. A., et al. 2009, ApJS,
182, 543
Adelberger, K. L. & Steidel, C. C. 2000, ApJ, 544, 218
Anders, E. & Grevesse, N. 1989, Geochim. Cosmochim. Acta, 53, 197
Arnaud, 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
Barthelmy, S. D., Barbier, L., Cummings, J., et al. 2007, GCN, 6173
Baumgartner, W. H., Barthelmy, S. D., Cummings, J. R., et al. 2009, GCN, 9939
Beardmore, A. P., Page, K. L., & Immler, S. 2008, GCN, 8506
Bell, E. F., McIntosh, D. H., Katz, N., & Weinberg, M. D. 2003, ApJS, 149, 289
Berger, E., Cowie, L. L., Kulkarni, S. R., et al. 2003, ApJ, 588, 99
Berger, E., Fox, D. B., Kulkarni, S. R., Frail, D. A., & Djorgovski, S. G. 2007,
ApJ, 660, 504
Bloom, J. S., Djorgovski, S. G., Kulkarni, S. R., & Frail, D. A. 1998, ApJ, 507,
L25
Bloom, J. S., Kulkarni, S. R., & Djorgovski, S. G. 2002, AJ, 123, 1111
Bruzual, G. & Charlot, S. 2003, MNRAS, 344, 1000
Butler, N. R. & Kocevski, D. 2007, ApJ, 663, 407
Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ, 533, 682