Numerical counterparts of GRB host galaxies

0.86 MB·English

by  Stephanie Courty

Mon.Not.R.Astron.Soc.000,000–000 (0000) Printed5February2008 (MNLATEXstylefilev2.2)
Numerical counterparts of GRB host galaxies
S. Courty1,2, G. Bj¨ornsson & E. H. Gudmundsson
1Science Institute, Universityof Iceland, Dunhaga 3, IS–107 Reykjavik, Iceland
2Laboratoire de l’Univers et de ses Th´eories, CNRSUMR 8102,
Observatoire de Paris-Meudon,5 place Jules Janssen, 92195 Meudon, France
7 e-mail: courty, gulli, [email protected]
0
0
2
5February2008

n
a
J
ABSTRACT

6
2 Weexploregalaxypropertiesingeneralandpropertiesofhostgalaxiesofgamma-

ray bursts (GRBs) in particular, using N-body/Eulerian hydrodynamic simulations
1
andthestellarpopulationsynthesismodel,Starburst99,toinferobservableproperties.
v
We identify simulated galaxiesthat have optical star formation rate (SFR) and SFR-
2
to-luminosity ratio similar to those observed in a well-defined sample of ten host
6
7 galaxies.Eachofthenumericalcounterpartsarefoundincatalogsatthesameredshifts
1 as the observed hosts. The counterparts are found to be low-mass galaxies, with low
0 mass-to-lightratio,recentepochofformation,andhighratiobetweentheSFRandthe
7 averageoftheSFR.Whencomparedtotheoverallgalaxypopulation,theyhavecolors
0 much bluer than the high-mass star-forming galaxy population. Although their SFRs
/ spanarangeofvalues,the specific rates ofthe numericalcounterpartsareequalto or
h
p higherthan the median valuesestimated atthe different redshifts.We alsoemphasize
- the strongrelationshipsbetweenthe specificstarformationrate(SFR)andquantities
o known to reflect the star formation history of galaxies, i.e. color and mass-to-light
r
ratio: At intermediate redshift, the faintest and bluest galaxies are also the objects
t
s withthehighestspecificrates.TheseresultssuggestthatGRBhostgalaxiesarelikely
a
to be drawn from the high specific SFR sub-population of galaxies, rather than the
:
v high SFR galaxy population. Finally, as indicated by our catalogs, in an extended
Xi sample, the majority of GRB host galaxies is expected to have specific SFRs higher
than found in the magnitude-limited sample studied here.
r
a
Key words: cosmology: large-scale structure of Universe – galaxies: formation –
galaxies: evolution – gamma rays: bursts
1 INTRODUCTION and/ortransformationofnewaccretedgas.Thespecificrate
should therefore give some insight into the star formation
history of galaxies, as do other properties, such as colors.
Galaxies with a wide variety of properties are observed in
The specific SFR has been estimated in a number of ob-
the universe and galaxy sub-populations contribute in a
servational studies: Guzman et al. (1997) compare the spe-
different way to the overall galaxy properties at different
cificSFRsofcompactbluegalaxieswithothergalaxypopu-
redshifts. Importantquestionsin present daycosmology in-
lations, Brinchmann et al. (2004) discuss the star-forming
clude investigations into how the different sub-populations
galaxies of the SDSS, Bell et al. (2005) combine infrared
can be characterized, how sub-populations formed at high
data from the Spitzer Space Telescope with optical data,
redshift evolve into galaxies in the local universe and how
and Feulneret al. (2005) estimate the specific SFR up to
these populations contribute to various properties of the
z=5.
overall galaxy population. In this paper we use a numer-
ical approach to extend our investigation of galaxy prop- As in Paper I, we here focus on the properties of a
erties (Courty et al. 2004, hereafter referred to as Paper particular population of galaxies, namely the host galax-
I) that focused on the specific star formation rate. While ies of long-duration gamma-ray bursts. The nature of the
the star formation rate is a ’snapshot’ of the stellar activ- hosts and their evolution with redshift are still open ques-
ity in a galaxy, the specific rate is an indicator of how the tions. The emphasis over the last few years has been on
galaxy forms its stellar mass relative to thetotal mass that the host galaxies of long-duration GRBs, which are seen
has been assembled through its entire lifetime, via mergers as a powerful tracer of massive star formation in the uni-
2 Courty, Bjo¨rnsson & Gudmundsson
verse. It is now well established that at least some long- servable properties, the results of Paper I for the specific
duration GRBs occur at the death of massive stars (e.g. SFR suggest that a sub-population of faint and blue galax-
Hjorth et al. 2003; Stanek et al. 2003) and in a cosmolog- ies, some of the characteristics of theGRB hosts, are likely
ical context massive stars are very short-lived. Because of tobelongtothehighspecificSFRgalaxypopulation,rather
their extreme brightness, long-duration GRBs are an effec- than thehigh-SFR population.
tive way of locating distant galaxies, most of which are so Inthispaper,weextendthediscussionoftheproperties
faint that they would go undetected in galaxy surveys. In ofGRBhostgalaxiesinPaperI,bycombiningtheresultsof
fact, host galaxies fainter than magnitude 29 have already the same simulations as in that paper with the stellar pop-
beendetected(Jaunsen et al.2003).Inaddition,GRBshave ulation synthesis (SPS) code, Starburst99 (V´azquez et al.
been detected out to a redshift higher than z =6, and will 2005), to infer observable properties. SPS codes allow usto
likely be detected to even higher redshifts. Rather detailed compute the spectral energy distributions of the simulated
studiesofthelocalstarformingregionsinthehostsisthere- galaxiesatdifferentredshifts.Weidentifysimulatedgalaxies
forepossible(Berger et al.(2005),Berger et al.(2005)).The that have both similar rest-frame ultraviolet SFRs and ra-
most distant burst to date, GRB 050904 at z = 6.29 tiosbetweenthisSFRandtheB−bandluminosityastheten
(Tagliaferri et al. 2005; Kawai et al. 2005), has already re- observed hosts of a well-defined sample (Christensen et al.
vealed a number of interesting properties of the interstellar (2004), hereafter Chr04). The numerical counterparts of
medium of thehost (Totani 2005),while information about theseobserved hostsare characterized byavariety of prop-
galaxy formation and evolution is yet to be explored in de- erties,estimatedeitherdirectlyfromthesimulation orfrom
tail, mostly awaiting the increase of the currently modest thecomputationoftheSEDs.Inparticular,foreachgalaxy
sample size (around 50 hosts). we determine the mass, the ratio between the SFR and
¿From individual studies of host galaxies of GRBs the average of the SFR, SFR∗/hSFRi, the specific SFR,
(Fruchteret al.1999;Fynboet al.2003),comparisonofhost the epoch of formation, the R− K color, and the mass-
sampleswithothersourcesdetectedinvariousdeepsurveys to-light ratio. The properties of the counterparts are then
(Le Floc’h et al. 2003), and statistical stellar population compared to the overall galaxy population, that is charac-
synthesis of the optical and near-infra-red host properties terized through the close relationships between the specific
(Sokolov et al. 2001; Chary et al. 2002; Christensen et al. SFRandthecolor indexand mass-to-light ratio. Thiscom-
2004), indications are that host galaxies have particular parison is, however, limited by the fact that the observed
characteristics: These galaxies tend to be optically sub- samplestillonlyincludes10hosts,spanningalargeredshift
luminous,low-mass,blue,star-burstinggalaxies,withyoung range.
stellar populations, a modest activity of optical star forma- Although fairly small, the sample of Chr04 is the only
tion,andperhapslow-metalicityandmodestamountofdust available homogeneous sample that estimates the ratio be-
obscuration,althoughthislastfeaturestillneedstobefirmly tween the optical SFR and the luminosity. Information on
established. Le Floc’h et al. (2003) compare a large sample otherhostsdoesexistintheliteratureandotherstudies,e.g.
of host galaxies of GRBs, observed in the near infra-red, Sollerman et al.(2005),usedifferentstarformation ratees-
with various galaxy surveys and find that the observed K timatorssuchasSFRHα andSFROII,ratherthantheUV-
andR−bandmagnitudesofthehostsarecomparabletothe based indicator adopted in Chr04, but only for a couple of
field sources selected in optical/N-IR deep surveys, butdif- hosts. A comprehensive study of a large host sample using
fersignificantly from luminousanddustystarburst galaxies all available information awaits future studies. In Section
observed with ISO and SCUBA. Moreover they show blue 3 we also briefly discuss the counterparts of GRB 000911
R−K colors typical of the faint blue galaxy population in and 030329, whose SFRUV and MB values are available
thefieldatz =1.Also,SCUBAsub-millimeterobservations inMasetti et al.(2005)andGorosabel et al.(2005),respec-
ofGRBhostgalaxiesperformedbySmith et al.(2005)sug- tively.
gest that most hosts are not luminous dusty star-forming The paper is organized as follows: In Section 2, we de-
galaxies. Furthermore, Christensen et al. (2004) have stud- scribe briefly the simulation and use of the SPS code. Sec-
ied a magnitude-limited sample of hosts and estimate the tion 3 contains the main results of the paper. It starts by
ratio between the rest-frame UV star formation rate and discussing the observed host sample we use (section 3.1)
thehost opticalluminosity.Theysuggest thatthehostsare and the procedure used to identify the numerical counter-
similar to those HDF galaxies that have the highest SFR- parts (section 3.2). We discuss the observational properties
to-luminosity ratios. of the counterparts in section 3.3 and compare their prop-
In Paper I, we showed that among the population still erties with those of the galaxy population in section 3.4.
activelyformingstarsatlowredshift,thehigh-massgalaxies Section 4 concludes thepaper.
have much lower specific SFR than the low-mass galaxies.
The non star-forming galaxies, that span the whole galaxy
mass range, are old galaxies while most of the stellar pop-
2 NUMERICAL PROCEDURE
ulations in the high specific SFR galaxies formed recently.
At high redshift the trend of increasing specific SFR with We repeat the simulation used in Paper I in order to
decreasinggalaxymassisalsoseen,butaninterestingpoint obtain galaxy catalogs at the same redshifts as that of
is that the cosmological evolution is much stronger for the the observed GRB hosts in the sample of Chr04. We
high-mass than the low-mass galaxies. These trends agree briefly recall the numerical method and we refer to Pa-
in general well with the aforementioned observational esti- perIfor detailsregarding thesimulation. Thethreedimen-
mates, to theextent that we concentrate on thequalitative sional N-body/hydrodynamical code couples a PM scheme
behavior of the specific SFR. Although not based on ob- for computing gravitational forces with an Eulerian ap-
Numerical counterparts of GRB host galaxies 3
proachforsolvingthehydrodynamicalequations.Thedom- itslatestversion(V´azquezet al.2005),toderiveobservable
inant processes relevant for galaxy formation are included: properties.Themain changein thisversion istheintroduc-
Gravitation, hydrodynamical shocks and radiative cool- tion of the Padova stellar evolutionary tracks, allowing the
ing processes. Collisional ionization equilibrium is not as- computation of stellar evolution for old and low-mass stars
sumed and the cooling rates are explicitly computed from as well as high-mass stars.
the evolution of a primordial composition hydrogen-helium WestartbydefiningaSEDtemplate:weconsiderasin-
plasma. The cosmological scenario adopted is a Λ−cold gle stellar population of 106 M⊙, assumed to form instan-
dark matter model with the following parameters: H = taneously and evolve passively over a maximal time of 15
0
70 km s−1 Mpc−1, Ω = 0, Ω = 0.3, Ω = 0.7, Ω = Gyr.TheSEDiscomputedat1221pointsbetween91˚Aand
K m Λ b
0.02h−2 with h = H /100, σ = 0.91. The comoving size 160 µm. We select in Starburst99 the “Padova AGB” evo-
0 8
of the computational volume is 32 h−1Mpc and the simu- lutionarytracks,selection ofthe1992−1994 Padovatracks
lation has 2563 dark matter particles and an equal number with thermally pulsing AGB stars added, with a metalic-
of grid cells. Galaxy formation is introduced using a phe- ity of Z = 0.004 = 0.2Z⊙, where Z⊙ ≃ 0.02. We choose
nomenological approach. At each time step, in cells whose a Salpeter IMF, dP/dm=m−2.35, with low and high-mass
gassatisfiesgivencriteria,afractionofthegasisturnedinto cut-offsat0.1and100M⊙,respectively.Theotherinputpa-
astellarparticleofmassm∗ andepochofformation a∗,the rametersadopted(supernovacut-offmass,blackholecut-off
scale-factorcorrespondingtothecosmictime,t∗.Thecrite- mass,windmodel,interpolation inmassmethod,modelat-
riaarethefollowing:Thecoolingtimemustbelessthanthe mosphere),arethestandardones.Wereferthereadertothe
dynamical time, t <t ; the baryonic density contrast, Starburst99 website1 for a full explanation.
cool dyn
(1+δ ), must be higher than a threshold (1+δ ) =5.5; The total SED of a galaxy at a given redshift is the
B B s
the gas must be in a converging flow, ∇·~v < 0; and the sumoftheSEDsofallthestellarpopulationspresentinthe
sizeofthecellmustbelessthantheJean’slength,givenby galaxy. The SED of a stellar particle, at a given redshift, is
λJ = cs(π/Gρ)1/2. The mass m∗ is given by mB(t0)∆t/t∗, theSEDtemplateevolvedonthetimet(z)−t∗,inorderto
where ∆t is the timestep, the characteristic time is taken takecosmological expansionintoaccount,andscaled tothe
to be t∗ = max(tdyn,108 yr), and mB(t0) is the baryonic massofthestellarpopulationwiththefactorm∗/(106M⊙).
mass enclosed within the grid cell. Galaxy-like objects are Toproperlyincludethecontributionofthosestellarparticles
then defined, at any redshift, by grouping the stellar parti- thatformed atthesimulation outputtime,t(z),theSEDis
cleswithafriend-of-friendalgorithm. Asimulatedgalaxyis evolved over 104 yr. Note that this procedure is applied to
therefore a collection of stellar particles of different masses thesimulatedgalaxiesafterthesimulationrun,andnotdur-
formed at different epochs. ing it. Nevertheless, the inferred SED of a galaxy accounts
At any redshift, galaxies are characterized by their fortheformation allalongthesimulationofdifferentstellar
mass, M, defined as the sum of the mass of all the stel- populations.Acaveatoftheprocedureisthatthesimulation
lar particles the galaxy is composed of; an epoch of forma- does not take into account any evolution of the gas metal-
tion, defined to be the mass-weighted average of the epoch icity, and our derived observable properties are based on a
of formation of all its stellar particles; and a star forma- single template. Due to the higher UV-luminosities of low-
tion rate, SFR∗, defined as the amount of stellar material metalicitystars,assuminghighermetalicitytrackswouldde-
formed intheprevious108 yr.Wealsoestimatethespecific crease the total UV-luminosity. However, in the catalog at
star formation rate, ǫ ≡ log(1011 yr SFR∗/M), a quantity z=1almost allofthenonstar-forminggalaxies havelumi-
thatmeasurestheefficiencyoftheconversionofthegasinto nosities more than twice the luminosity computed with so-
stellar material relative to the galaxy mass. We have, as in larmetalicitytracks,whereasthereareonly19star-forming
Paper I, only considered galaxies with a mass higher than galaxies with such a large luminosity.
5·108 M⊙ inthesimulatedcatalogs.WerefertoPaperIfor StarformationratescanbederivedfromtheUVcontin-
an illustration of the different sub-populations of galaxies uumortheHαluminosityusingappropriateproportionality
obtained. coefficients. These coefficients are derived from stellar pop-
Consideringeachstellarparticleasahomogeneousstel- ulation synthesis models, involving an initial mass function
larpopulation,thetotalspectralenergydistribution(SED) (IMF) and star formation history, by looking for a propor-
of a galaxy can be computed from synthesis codes. Evolu- tionality relation assuming that the star formation rate is
tionary synthesis codes combine stellar evolutionary theory constant in the later phases on given timescales. These are
to describe the time evolution of model stars, stellar atmo- roughly 107 yr for Hα and 108 yr for the UV. These coef-
sphere theory to transform quantities from the theoretical ficients therefore depend on the IMF and its slope, the up-
space to the observational one, and the stellar birth rate, perandlowermasscut-off,andthemetalicity(Boselli et al.
giving thenumberof stars with agiven initial mass formed 2001). Sinceone of thepurposes of this paperis to find the
at a given time. The stellar birth rate comes from the star numericalcounterpartsofobservedhostgalaxies,weusethe
formation history, which gives the number of stars born in same indicator of the star formation rate as Chr04, and es-
a given time, and the initial mass function (IMF), which timate it from the UV continuum at 2800 ˚A, although the
givestherelativenumberofstarsbornasafunctionofmass SFRsofthetwohostsatthehighestredshiftsareestimated
(see Cervino & Luridiana (2005) for a discussion regarding using the UV continuum at 1500 ˚A in Chr04. The calibra-
the uncertainties of the synthesis codes). In our procedure, tion factor used to convert the luminosity at 2800 ˚A into
the star formation history of a galaxy comes directly from SFRis 1.4×10−28, in unitsof (M⊙ yr−1)(erg s−1 Hz−1)−1
the formation epoch and mass of all its stellar populations
as recorded in the simulation. We use the stellar popula-
tion synthesis model, Starburst99 (Leitherer et al. 1999) in 1 http://www.stsci.edu/science/starburst99/
4 Courty, Bjo¨rnsson & Gudmundsson
Figure 1. Comparison of the star formation rate SFR∗ with
SFRUV obtained from the synthetic spectrum for galaxies at
z = 1. Non-star forming galaxies are for display purposes plot-
ted at SFR∗ = 0.02 M⊙ yr−1. The diagonal dashed line shows
SFRUV =SFR∗.
(Kennicutt 1998). In our procedure the luminosity at 2800
˚A is obtained by averaging the spectrum over 20 ˚A around
2800 ˚A. Note that, although our spectral energy distribu-
tions are determined at sub-solar metalicity, we adopt the
usual calibration to convert the UV luminosity into SFR
thatisestimatedatsolarmetalicity.Atlowermetalicitythe
calibration increases due to the higher luminosity of low-
metalicity stars. The SFR of low-metalicity galaxies when
estimatedwithacalibrationatsolarmetalicityisthusover-
Figure2.Themass-magnitudediagrams(upperpanel:K−band,
estimated. The evolution of the calibration associated with lower-panel:B−band)forthesimulatedgalaxiesatz=1.Inboth
differentSFRindicatorsisestimatedinSullivan et al.(2001) panels, dots denote star-forming galaxies while the small dots
and Bicker & Fritze-v.Alvensleben (2005), for instance. At denote the non star-forming galaxies. The estimate of M∗ =
Ks
Z = 0.2Z⊙, the latter paper found that the SFR is overes- −24.7around z =1 fromthe K20galaxy survey (Pozzetti etal.
timated bya factor 1.4, compared to our calibration. 2003)ismarkedbythedotted horizontalline.
In Figure 1, we compare the SFR∗, computed directly
from thesimulation, with theSFR estimated from thesyn-
thetic spectrum of the simulated galaxies at z = 1. Non-
star forming galaxies (i.e. with SFR∗=0) are for display characteristic magnitude of the galaxy luminosity function
purposes plotted at SFR∗ = 0.02 M⊙ yr−1. At this red- in Cole et al. (2001), havea mass of ∼1011 M⊙.
shift,thegalaxypopulationincludes1927objects,with1148 Inthefollowingsectionsweconsiderotherquantitiesto
star-forming galaxies (as expected,thenumberof non star- characterize galaxies. The lifetime of a galaxy is estimated
forming galaxies is larger at z = 0). The diagonal dashed atagivenredshift,t(z)−t ,wheret istheepochof
form form
lineshowsSFR =SFR∗.Despitethescatter inthedia- the formation of the first stellar population. From this the
UV
gram, it is clear that the two different estimates are of the average of the star formation rate, hSFRi, at a given red-
sameorder.Theestimatefromthesimulationsisafactorof shiftisdeterminedastheamountofstellarmaterialformed
about2lowerthanthatobtainedfromthesyntheticspectra. overthelifetimeofthegalaxy.TheSFR∗ iscomparedwith
Figure 2 shows the B− and K−band magnitudes as the average through the ratio SFR∗/hSFRi, that may be
functions of the galaxy mass for the simulated catalog at seen as analogous to the so-called birth-rate parameter, b,
z=1.TheB andK−bandmagnitudesarecomputedusing (Kennicuttet al. 1994). The mass-to-light ratios in the B
the“Buser’s B3” filterand the“IRK filter+Palomar 200 and K−bands are estimated using the solar magnitudes of
IR detectors + atmosphere 0.57” from the Galaxev pack- 5.45 and 3.3, respectively, and the color R−K using the
age(Bruzual & Charlot2003).Differentsymbolsdistinguish “Cousins R” filter from the Galaxev package. Finally, the
star-forming and non star-forming galaxies. Notethat both ratio between the rest-frame ultraviolet SFR and the lumi-
arestronglycorrelatedwiththemass.Thereismuchlessdis- nosity,SFR /(L /L∗),usesthecharacteristicmagnitude
UV B B
persioninM sincetheB−bandluminositiesaredominated or magnitude at the break of the local galaxy luminosity
K
bythefluxfromyoungmassivestarswhereasL isabetter function, M∗ = −21. This ratio is sometimes called the
K B
tracer of the overall stellar population. The corresponding specificSFR ase.g., in Chr04, butwe will refer toit by the
plotsatz =0wouldshowthesametrendalthoughquantita- term SFR-to-luminosity ratio and reserve the term specific
tively different. For instance, the M −mass relation shows SFRfortheratiobetweentheSFRandthegalaxymass,as
K
slightly more dispersion and galaxies at M∗ = −24.2, the is customary in the literature.
K
Numerical counterparts of GRB host galaxies 5
Table 1.Properties of the numerical counterparts corresponding to the GRB hosts listed incolumn one. Following the redshiftof the
simulatedcatalogincol.2,thepercentageerrorsinSFRUV andtheSFR-to-luminosityratio,∆X and∆Y definedinthetext,aregiven
incols.3and4.Column5showsthemassofthecounterpart, whilecol.6shows theepoch ofitsfirstformedstellarpopulation, tform
(seeSection2fordefinition)relativetotheHubbletime.Themass-to-lightratiosintheK andB−bands,respectively,aregivenincols.
7and8, followedbythecorrespondingabsolute magnitudes. Thencomes thecolorindex, R−K,andthenexttolastcolumnliststhe
R(AB)apparentmagnitude. Finally,areferencenumberforeachhostisgiveninthelastcolumn.
GRB z ∆X ∆Y M(M⊙) 1−tform/tH(z) M/LK M/LB MK MB R−K R(AB) #
000926 2.036 0.7 0.5 1.16×1010 0.76 0.36 0.36 −22.97 −20.83 2.61 24.41 1
990123 1.592 5.6 3.2 2.41×1010 0.81 0.47 0.59 −23.47 −21.08 3.50 24.27 2
000418 1.113 5.6 8.8 8.27×109 0.86 0.35 0.32 −22.63 −20.57 2.47 23.24 3
980703 0.963 1.3 0.6 3.68×1010 0.88 0.52 0.66 −23.82 −21.41 2.87 22.18 4
000210 0.843 6.5 5.7 5.16×109 0.89 0.47 0.54 −21.79 −19.50 2.61 23.64 5
970508 0.832 3.7 0.3 7.20×108 0.89 0.30 0.25 −20.14 −18.18 2.22 24.84 6
991208 0.704 1.2 0.9 1.64×109 0.90 0.41 0.45 −20.71 −18.46 2.54 24.24 7
970228 0.695 10.8 0.5 1.60×109 0.90 0.49 0.56 −20.49 −18.19 2.51 24.41 8
010921 0.449 1.9 1.1 8.97×109 0.92 0.68 1.01 −22.01 −19.42 2.58 21.93 9
990712 0.429 3.3 2.1 1.42×109 0.92 0.38 0.36 −20.62 −18.55 2.07 22.71 10
3 THE NUMERICAL COUNTERPARTS OF z=1.06 andderiveaB−bandmagnitudeofaround −18.4,
OBSERVED GRB HOST GALAXIES giving a L /L∗ = 0.09 with our M∗ (see section 2), and
B B B
In this section we attempt to identify numerical counter- an extinction-corrected SFRUV of 2.7 M⊙ yr−1. Assuming
an extincted SFR to be a factor of 2 lower2, gives a
partstothehost galaxies oftheCh04sample, exploretheir UV
SFR-to-luminosityratioaround15. Thesecond host isthat
properties,andcomparethehostcandidatepopulationwith
of GRB 030329 at z = 0.168. Gorosabel et al. (2005) de-
tehvaenotvperraolplegratileasxyofptohpeuloabtsioernv.eWd esafimrpstlesuamndmoaurirzeaptphreoareclh- arivSeFaRSF/R(ULV /oLf∗0).1=7 1M0.⊙6.yNr−o1teatnhdatMthBis=va−lu1e6i.s5,cogmivpinag-
to identifying the numerical counterparts. We then discuss UV B B
rable to the SFR-to-luminosity ratios derived in the Chr04
their observed properties.
sample. The counterparts of these two hosts are discussed
near theend of this section.
3.1 The observed sample
The sample presented in Chr04 is homogeneous and con-
3.2 Counterpart Identification
sists of 10GRBhosts with redshifts between z=2.037 and
z=0.433. Thesample ismagnitudelimited (R<25.3) and We generated catalogs of galaxies at the same redshifts as
ensures that hosts are bright enough to make multi-color theobservedhosts,andlookedineachcatalogforsimulated
photometry possible. The first column in Table 1 lists the galaxies that have both SFR and SFR /(L /L∗)
UV UV B B
hosts under consideration. The observationally determined nearest to the corresponding values for each observed host.
valuesofSFRUV andSFRUV/(LB/L∗B)aregiveninTable We are able to find a numerical counterpart to each of
4and5ofChr04andwenotethattheSFRUV enteringthe the 10 observed host of the Chr04 sample, although de-
SFR-to-luminosityratioisnotcorrectedforinternalextinc- partures from the observationally inferred values may be
tion,generallyfoundtobemoderateorlow.TheSFRspans large in some cases. Thepercentage errors in Table1quan-
awiderange(between0.8and13M⊙ yr−1),buthostswith tify how close from these values the counterpart is found.
redshifts higher than 0.9 have the highest SFRUV, more They are defined by ∆X = |SFRUobVs −SFRUnuVm|/SFRUobVs
than ∼ 6 M⊙ yr−1, whereas hosts at lower redshifts have and ∆Y =|ǫoLbs−ǫnLum|/ǫoLbs. In this last definition, ǫL de-
SFRUV <2.5 M⊙ yr−1. notes the SFR-to-luminosity ratio. The largest departures,
Various observational studies of individual hosts of the when at least one of these quantities is larger than 5%, are
Chr04 sample result in slightly different estimates of the seen for counterparts #3, 5 and 8. The first two are noted
same properties (Gorosabel et al. (2003) for GRB 000418, in Chr04 as having the least acceptable fits to their spec-
Gorosabel et al.(2003)forGRB000210,Bloom et al.(1999) tral energy distributions. The host of GRB 000418 (#3) is
for GRB 990123, for instance). We have not attempted extensively discussed in Gorosabel et al. (2003) who ana-
to collect all the data existing in the literature, and have lyze its spectral energy distribution using a variety of syn-
only considered the SFRUV and SFRUV/(LB/L∗B) given thetic spectral templates. The adopted SFR does of course
in Chr04. All the hosts in Chr04 are also discussed in depend on the adopted SED. Also, Gorosabel et al. (2003)
Le Floc’h et al. (2003), giving either their observed R and find the host galaxy of GRB 000210 (#5) to be brighter
absolute B−band magnitudes or their observed K magni- thantheChr04 estimate: TheyfindthattheB−bandmag-
tudes and colors, R− K. In addition to the Chr04 sam-
ple, we also consider two hosts for which the SFRs based
on the rest-frame ultraviolet flux as well as the B−band
magnitudes, are available. We can examine such hosts in a 2 The host of GRB 991208 has roughly similar B−band mag-
similar way as for the Chr04 sample. Masetti et al. (2005) nitude and AV and has a ratio of un-extincted/extincted SFR
discuss theproperties of thehost galaxy of GRB 000911 at withinafactorof2(Chr04).
6 Courty, Bjo¨rnsson & Gudmundsson
rate SFR∗, the specific SFR, the galaxy mass, the ratio
SFR∗/hSFRi, the formation epoch of the first stellar pop-
ulation to form in the galaxy, and the epoch of formation
ofthegalaxy asdefinedinsection 2.Fromthecomputation
of theSEDs we determine the star formation rate SFR ,
UV
the SFR-to-luminosity ratio, the magnitudes in the B and
K−bands,theapparentRmagnitudeandthecolor,R−K.
The mass-to-light ratios in the B and K−bands are then
a combination of the primary simulation outputs and the
observable properties.
3.3 ’Observed’ properties of the numerical
counterparts
From Table 1 we note the following: Low-redshift hosts
(z <0.9) are low B−band luminosity galaxies, M >−20.
B
We also see that the counterparts are fainter than −22 in
the K−band at z <0.9. The 10 counterparts are low-mass
galaxies with M < 4×1010 M⊙ (the highest mass in the
simulation at z = 1, being 7.4×1011 M⊙). The B−band
magnitudes and mass of the counterparts can also be seen
inFig.3discussedbelow,whereintheupperpanelthedot-
ted curves denote values of constant M = −18 to −23
B
(from left to right) and in the lower panel the dashed lines
indicatevaluesofconstantmassof109 to1011M⊙ (fromleft
toright).ThecounterpartshaveL /L∗ between0.074 and
B B
1.46andL /L∗ between0.028and0.84(withM∗ =−24),
K K K
making most of them sub-luminous galaxies. The mass-to-
light ratios in Table 1 are relatively small and more typical
oflate-typeanddwarfgalaxiesthanlargespiralsorelliptical
galaxies.
Figure3.SFR-to-luminosityratioversusSFRUV (upperpanel)
andspecificSFRversusSFR∗ (lowerpanel)forthestar-forming The counterparts have apparent R-band magnitudes
galaxy population at z =0.832 (dots). Thedotted curves inthe between 22 and 24.4 and comparison between those and
upper panel denote constant MB-values from −18 to −23 (left the N-IR observational data reported in Le Floc’h et al.
to right). In the lower panel the dotted lines indicate constant (2003) for the same hosts (here referred to as the common
mass of 109, 1010 and 1011 M⊙ (left to right). Although at dif- sample) shows that the candidate hosts #6 and #9 are
ferentredshifts,the numerical counterparts of the observed host the faintest and the brightest, respectively, both of our
galaxies (listed in Table 1) are displayed in both panels (solid
sample and of the common sample. The apparent R-band
symbols): In decreasing order of SFRUV, GRB 980703, 000926, magnitudes of the counterparts are, however, generally
000418,990123, 000210,010921, 990712,970508,991208, 970228
(the order is the same in the bottom panel with SFR∗, except slightly brighter than in Le Floc’h et al. (2003). The bluest
of the 9 counterparts (#10) for which the R−K colors of
thatthehostofGRB970228istotheleftofGRB991208).Inthe
upper panel, for each counterpart a short linesegment indicates the corresponding hosts were estimated in Le Floc’h et al.
wherethecorrespondingobservedhostgalaxywouldbefound.In (2003), is also the bluest of these 9 observed hosts. The
some cases the difference is too small for the linesegment to be observed host with the highest color index corresponds
visible.Thetwodiamondsdenotethecounterpartsatthehighest the counterpart with the second highest color index
redshifts. (after the least blue #2). Globally, we do find that 9 coun-
terpartshaveR−Kcolorsbetween2and2.9and3.5for#2.
nitudeis−20.16,giving,withtheirSFRUV =2.1M⊙ yr−1, InFig.3wesuperposethenumericalcounterpartsona
a SFRUV/(LB/L∗B)=4.5 (10.7 in Chr04). 3 plot showing the simulated star-forming galaxy population
Each numericalcounterpart ischaracterized bya num- atz =0.832inadiagram ofSFR-to-luminosityratioversus
berofpropertiesthatareeithertabulatedinTable1ordis- SFR (upperpanel)andinadiagramofspecificSFRver-
UV
played in Fig. 3 and 5. The properties that are estimated susSFR∗ (bottompanel).Wehavechosenz=0.832asitis
directly from the simulation output are the star formation roughlythemedianredshiftofthesample.Rigorouslyspeak-
ing,onlythecounterpartofthehostatz=0.832 shouldbe
compared directly with the simulated catalog, since galaxy
3 Insteadofadoptingthenearestcounterpartineachcatalog,we
propertiesevolvewithredshift.Itmayneverthelessbeuseful
couldhavelookedforthesecondbestclosestcounterparts orthe
toplotthewholesampleonasinglediagram.Wehaveused
counterpartswithinagivenerrorboxaroundeachobservedhost.
different symbols (diamonds) to mark the highest two red-
Because of the uncertainties in the observationally determined
values andthe smallnumber ofobserved hosts inthe samplewe shifts, and will return to the issue of evolution later in this
consider here, we do think that searching only for the closest section. Each counterpart in the top panel is accompanied
counterpartisthemostreasonablethingtodo. bya straight line, sometimes smaller than thesymbol,that
Numerical counterparts of GRB host galaxies 7
indicates how well the properties of that numerical coun-
terpart match those of the observed host (the difficulty in
obtainingagoodcounterpartforhost#3isclearlyseen).As
noted above, the hosts have various SFRs, be it either the
SFR ortheSFR∗,buttheyhavehighSFR-to-luminosity
UV
ratio and high specific rate, as theyall lie in theupperhalf
of both panels in Fig. 3. The specific SFR and the SFR-to-
luminosityratiowouldbeevenbettercorrelatedifthelatter
quantity is based on the K−band luminosity, according to
themagnitude-mass diagrams in Fig. 2.
3.4 Comparison with the overall galaxy
population
Evenifsomeaspectsofthesimulationmaybeincomplete,it
provides us with catalogs of galaxies at various redshifts in
whichdifferentgalaxypopulationsmaybedistinguished.Af-
terdefiningaparticularsub-populationofgalaxiese.g.sim-
ilar to the host galaxies of GRBs, an important issue is
to compare this particular population to the overall galaxy
population.Suchacomparisonisinthepresentcaselimited
bythefactthattheobservedsampleincludesonly10hosts,
spanningawideredshiftrangewherecosmologicalevolution
cannot beneglected.
3.4.1 The relationship between the specific SFR and the
color index
Wefirstconductaqualitativecomparisonbetweenthecoun-
terparts and the overall galaxy population at z = 1, focus-
ing on the relationship between the specific SFR and the Figure 4. Mass-to-light ratio in solar units in the K−band as
mass-to-light ratio and color index. By presenting observ- a function of mass (upper panel) and color-magnitude diagram
able properties at this redshift we follow-up on and extend (bottom panel) for the simulated galaxy population at z = 1.
theresults and discussion of Paper I. Forclarity,only1000objectsrandomlyselectedfromthecatalog,
Figure 4 presents the mass-to-light ratios and color in- areplotted. Thedots indicatenon star-forming,low-mass(M <
dex of the whole galaxy population distinguished accord- 5·1010 M⊙) galaxies; the filled circles non star-forming, high-
ing to the specific rate (ǫ < 1, 1 < ǫ < 1.3 and ǫ > 1.3). mass (M > 5·1010 M⊙) galaxies; the open circles star-forming
We also plot the low-mass (M < 5·1010 M⊙) and high- galaxies with ǫ<1; the open squares star-forminggalaxies with
1<ǫ<1.3;thefilledsquaresstar-forminggalaxieswithǫ>1.3.
mass (M >5·1010 M⊙),non-starforming galaxies. Weuse Galaxieswithamassaround1011 M⊙,correspondinginFig.2to
the same threshold, ǫ = 1.3, as in Paper I. It corresponds thecharacteristicmagnitudeM∗ =−24.7(orL∗ =1.6×1011
roughly to the peak of the probability density function of Ks Ks
L⊙),haveamass-to-lightratiooflog(M/LK)∼−0.2.
thespecificSFRatz=1.Inaddition,galaxieswithspecific
rates below or above this value contribute about equally to
thetotalstarformation rateatthatredshift.Thetoppanel
inFig.4showsthelargevariation ofthemass-to-lightratio of old elliptical galaxies at z & 1. In contrast, the majority
with respect to the galaxy mass, but there is a clear cor- ofthestar-forming galaxieshavecolorsaround3.Thecolor
relation with the specific SFR (i.e. ǫ). High-mass non star- properties correlate with the specific SFR: The bluest ob-
forming galaxies havelarge M/L , up to about 1, whereas jects are also faint galaxies with the highest specific rates.
K
theminimalvalueforthelow-mass high-specificrategalax- Comparing the data from Table 1 with Fig. 4 shows that
iesis∼0.15.Themass-to-lightratiointheB−band,M/L , the counterparts are clearly bluer than the high-mass star-
B
ranges between ∼0.09 and ∼3, for the overall population. forminggalaxy population,with colorindexlowerthan∼3.
Recalling the results in Table 1, we note that the major- Recalling the correlation between mass and magnitude dis-
ity of the counterparts have log(M/L ) between −0.5 and cussed in the previous section, we note that the colors and
K
−0.3, typical of our low-mass, high specific SFR galaxies, mass-to-light ratios are also tightly related (Bell & deJong
although it should bekeptin mind that Fig. 4is plotted at 2001). There is also a strong relation, albeit with some dis-
a single redshift (z=1). persion between mass and SFR, resulting in intermediate
The color-magnitude diagram in the bottom panel in colorsofthehigh-SFRgalaxies.Thereforetheblueandfaint
Fig. 4 shows a large variation of the color index R−K, galaxies, typically characterized by high specific rates, are
from 1.5 up to 4.7, consistent with that seen in the recent not the objects that havethehighest SFRs.
K20 galaxy survey (Pozzetti et al. 2003). The high-mass, Properties like the mass-to-light ratio and color index
nonstar-forming galaxies, thatarealso oldobjects (seePa- are generally considered to provide information on the star
perI)havethehighestcolorindex,closetothetypicalvalue formation history of galaxies or how they assemble their
8 Courty, Bjo¨rnsson & Gudmundsson
given redshift of a galaxy to its average activity since the
objectstartedtoformitsstellarpopulations.Bothdistribu-
tions peak around unity, but the high-redshift distribution
includes more galaxies with SFR∗/hSFRi > 1, meaning
thathigh-redshiftgalaxiesaremoreactivethanpresent-day
galaxies.Allhostsexcept#2haveSFR∗/hSFRiclosetoor
aboveunity (0.85 for #9).
The lower panel shows the epoch of formation normal-
ized to the Hubble time. Comparing the distributions at
the two redshifts shows that the low-redshift one is more
extended but includes a non-negligible amount of objects
with recent epochs of formation. All candidate hosts have
ages within 40% of the age of the universe. We remind the
reader that the distributions are constructed only from the
star-forming galaxies and do not include the old, non star-
forminggalaxies.Theepochofformationindicatestheepoch
at which thegalaxy was themost active and is expected to
be quite different from the epoch of formation of the first
stellar populations, t . Table 1 shows that the epoch
form
of formation of the counterparts are indeed different from
t . This particular time is relatively close to 1 (as de-
form
finedin Table 1),meaning that thecandidate hostsinclude
an early-formed stellar population, although not dominant
since they are at thesame time young objects.
That the major fraction of the mass of most counter-
partswasassembledintherecenttimesisconsistentlyshown
by the specific rate, the ratio SFR∗/hSFRi and the epoch
offormation. Wenotethatthethreehostswith thehighest
SFR∗/hSFRiratios(counterparts#3,6and10),alsohave
the highest specific rates although their SFR∗ and magni-
tudes differ widely (GRB 000418, 990712 and 970508, see
Figure5.ProbabilitydensityfunctionsoftheSFR∗/hSFRi(up-
Fig.3).Theyareamongthehostswiththeyoungestepochs
per panel) and the epoch of formation EoF, when normalized
offormation,thebluestR−Kcolors,andthelowestmass-to-
to the Hubble time (bottom panel), for the star-forming galaxy
populations at z = 2.036 (dotted) and z = 0.832 (dashed). The lightratios(Table1),pointingagaintothetightcorrelation
counterparts of the hosts in Table1 are superposed at arbitrary between these galaxy properties as discussed above and to
ordinate,butinorderofdecreasingredshift(fromtoptobottom). theconsistency with theGRB host galaxy properties.
3.5 Discussion
mass. They also tend to be more tightly correlated with
the specific rate than the SFR. The SFR∗/hSFRi, to be Here,weemphasizeourmainresultbycomparingtheSFR-
to-luminosityratiosandspecificSFRsofthenumericalcoun-
further discussed below, is another property that strongly
terpartstothemedianvaluesofthesequantitiesinthecat-
correlates with the specific rate. As already pointed out in
alogs they are selected from. These are displayed in Fig. 6.
PaperI, butnowconfirmed usingthecalculated observable
Themedian valueof thespecificSFRclearly increases with
properties, we note the consistency between the observed
redshift. Note that at redshifts below z ∼ 0.7, the mini-
properties of the GRB host galaxies: If host galaxies are
malspecificSFRtendstoincreasewith decreasingredshift:
blue and faint, they are expected not to be high-SFR and
This is due to the fact that the star formation of massive
early-formedgalaxies. Thepropertieswehavefoundforthe
galaxies slows down as the redshift decreases and eventu-
counterparts confirm this conclusion.
allyceasesinanincreasingnumberofthem.Theytherefore
disappear from the star-forming galaxy population. Com-
paring the specific SFRs and the SFR-to-luminosity ratio
3.4.2 The star-forming activity of the counterparts
of the counterparts to themedian values at similar redshift
Figure 5 compares the SFR∗/hSFRi and the epoch of for- shows that the counterparts in all cases except one have
mationofthecounterpartstothatofthestar-forminggalaxy values higher than the median. The only exception is the
population at z =0.832 (dashed lines) and z =2.036 (dot- counterpart to GRB 990123 (#2). The SFR /(L /L∗)
UV B B
ted lines), the highest redshift of the host sample. In each ofthisparticularhostisthelowestoftheChr04sampleand
panel the counterparts are superposed on the distributions its observed spectral energy distribution was best fit by a
at arbitrary ordinate values, but with a decreasing redshift star-forming Sa galaxy type, whereas all of the other hosts
(from top to bottom). Strictly speaking only two hosts, #1 were fitbystarburst templates. The B−bandmagnitudeof
and#6,shouldbecompareddirectlytotheirrespectivedis- this host could however befainter (Bloom et al. 1999), giv-
tributions. ing a higher SFR-to-luminosity ratio. Moreover the R−K
The top panel compares the star-forming activity at a color of this host is as blue as most of the hosts studied in
Numerical counterparts of GRB host galaxies 9
we find for the host of GRB 030329 has a lower SFR-to-
luminosity ratio than the observed estimate (∆X = 1.7%
but ∆Y ∼ 19%). The candidate host is a low-mass ob-
ject (M = 1.4×109 M⊙), with blue color R−K = 2.2,
a specific rate of ǫ = 0.7 and an epoch of formation of
1−EoF/t =0.45. Compared to the overall population
H(z)
thesetwocandidatehostshaveSFR-to-luminosityratiosand
specific SFRs well above or similar to the median values at
thesame redshifts.
The fact that the Chr04 sample is magnitude limited
and includes bright hosts may explain why the numerical
counterpartsarenotfoundamongtheobjectswiththehigh-
est possible specific SFRs. The counterpart of GRB 970508
(#6)isanexample.FromHSTobservationsandotherstud-
ies, this host may be described as a blue compact dwarf
galaxy(Fruchteret al.2000).Couldthishostbeaprototype
of the general host GRB galaxy population? Interestingly,
the counterpart of this host could be found in the sample
of compact galaxies at z ∼ 0.7 analyzed in Guzman et al.
(1997). The counterpart #6 (z = 0.83) with an absolute
magnitudeMB =−18.2,amasscloseto109 M⊙ andaspe-
cificrateofǫ∼1.8fallsintothefaintestandhighestspecific
SFRpopulationofcompactgalaxiesinGuzman et al.(1997)
(seetheirfigures5and7).Moreover,Sollerman et al.(2005)
discuss three GRB host galaxies and show that they have
similar properties as a sample of compact blue galaxies in
thelocal universe.
4 CONCLUSIONS
Figure6.Themedianvalues(crosses)oftheSFR-to-luminosity
ratio(top)andspecificSFR(bottom)forthestar-forminggalaxy In this paper, we have discussed cosmological galaxy prop-
populations attheredshiftsoftheGRBhosts.Therangeofval- ertiesandthepropertiesofhostgalaxiesofGRBsinpartic-
ues for each catalog is shown by the length of the vertical lines. ular, using fully hydrodynamic simulations of galaxy for-
Diamonds refer to the numerical counterpart in each case. The mation and the stellar population synthesis (SPS) code,
twosquaresarethecounterpartsofGRB000911and030329hosts Starburst99, to infer observable properties. An important
thatarebrieflydiscussedattheendofSection3. feature of the numerical procedure is that the star forma-
tion history of galaxies entering the SPS code, comes di-
rectlyfromthesimulation,witheachstellarpopulationcon-
Le Floc’h et al. (2003), whereas it is the reddest one of our tained in a given galaxy treated as a homogeneous popula-
sample, as seen above(see Table 1). tion, that forms instantaneously. We identify objects in the
It is interesting to note that Chary et al. (2002) esti- simulation that have optical star formation rate, SFR ,
UV
mate thespecific SFRsof hosts #2, 4, 6, 7 and 8, although and SFR-to-luminosity ratio, SFR /(L /L∗), similar to
UV B B
their observational study is based on extinction-corrected those estimated in a well-defined sample of ten observed
SFRs,and thustheirderived specific rates are muchhigher host galaxies (Christensen et al. 2004), the only available
than our results. The observed hosts #4, 6, 7, and 8 are homogeneoussamplefocusingonthisratio.Each numerical
among those with the highest specific SFR of the whole counterpartisselectedfromasimulatedcatalogatthesame
Chary et al. (2002) sample. They are also found to have redshift as the corresponding observed host, and is charac-
higherspecificratesthanlocal starburstshave.Thenumer- terizedbyanumberofproperties:TheSFR∗,mass,specific
ical counterparts to #6, 7 and 8 have specific rates well SFR, SFR∗/hSFRi, epoch of formation of the first stellar
above the median values (the first three objects on the left population, and epoch of formation of the galaxy object.
in Fig. 3). Theseareobtaineddirectlyfromthesimulation.Inaddition,
The two squares in Fig. 6 show the SFR-to-luminosity B and K−band luminosities, R-band apparent magnitude,
ratio and the specific SFR of the counterparts of the hosts R−K color,andmass-to-light ratiosareobtainedfrom the
of GRB 000911 and 030329, discussed at the beginning of SPSorcombinationofbothresults.Itshouldbeemphasized
this section. These counterparts were searched for in cat- thatsomeoftheseproperties(e.g.massandSFR∗)areesti-
alogs at slightly different redshifts from the host redshifts, mateddirectlyfromthesimulations,makingtheirdefinition
z =1 and z =0.133, respectively. The counterpart of GRB inherentlydifferentfrom thosecommonly adoptedinobser-
000911 (∆X =0.3% and ∆Y =5.2%) is a low-mass, young vations.
object(M =1.1×109 M⊙,1−EoF/tH(z)=0.37)withahigh The Christensen et al. (2004) sample includes host
specificrate(ǫ=1.87), bluecolor R−K =2.12 and amass- galaxies with redshifts in the range 0.43 < z < 2.03. The
to-light ratio M/L = 0.36. The closest counterpart that sampleismagnitude-limited(R<25.3),withestimatedab-
K
10 Courty, Bjo¨rnsson & Gudmundsson
solute B−band magnitudes between −21.4 and −18.1. Our host galaxies to the overall population, and whether hosts
counterparthostsarelow-mass galaxies (M <4·1010 M⊙), are a part of the average normal star-forming galaxy pop-
with low mass-to-light ratios (M/L around 0.5); most of ulation or a sub-population of this one, with blue and very
B
themareblue(R−K <2.9)andyounggalaxies,withepochs low-luminosity objects.
offormation(orages),within40%oftheageoftheuniverse
at the different redshifts. Although the SFR∗ of the coun-
terparts varies between ∼ 0.4 and 8 M⊙ yr−1, the specific
SFRisequaltoorhigherthanthemedianvaluesestimated ACKNOWLEDGMENTS
forthedifferentcatalogs, with thelowest valuebeingǫ∼1.
WethankJensHjorthandJohanFynboformanyinteresting
Because of its strong correlation with the specific rate, the
SFR∗/hSFRi also hashigh values,around unityor higher. discussion sessions on GRB galaxy hosts and their proper-
ties.Wealsothanktheanonymousrefereeforcommentsthat
Tooutlinetheconsistencyofsuchanensembleofproperties,
helped us improve the presentation. This research was sup-
we compare the counterparts to the overall galaxy popula-
portedinpartbyaspecialgrantfromtheIcelandicResearch
tion at intermediate redshift and discuss the strong rela-
Council. Thenumericalsimulations used in thispaperwere
tionshipsbetweenthespecificSFRandquantitiesknownto
performed on NEC-SX6 of the High Performance Comput-
reflect the star formation history of galaxies, i.e. color and
ing CenterStuttgart (HLRS)of theUniversityof Stuttgart
mass-to-light ratio. Indeed,thebluest objects are also faint
(Germany).
galaxies with the highest specific rates.
Our identification of simulated galaxies with observed
hostshassomelimitations,bothfromobservationalandnu-
merical points of view. Some of the observed hosts may, in
REFERENCES
otherobservationalstudies,befoundtohaveslightly differ-
ent SFRs or magnitudes, making their SFR-to-luminosity Bell E. F., de Jong R. S., 2001, ApJ, 550, 212
ratios uncertain. Although the general agreement between Bell E. F., Papovich C., Wolf C., Le Floc’h E., Caldwell
thesimulation andtheobservationsisfairly good,itshould J. A. R., Barden M., Egami E., McIntosh D.H., Meisen-
be kept in mind that the moderate resolution and some- heimerK.,P´erez-Gonz´alezP.G.,RiekeG.H.,RiekeM.J.,
what limited number of physical processes included in the RigbyJ. R., Rix H.-W.,2005, ApJ,625, 23
simulation may affect the relative number of galaxies in Berger E., Penprase B. E., Cenko S. B., Kulkarni S. R.,
each sub-population. For instance, these simulations only FoxD.B.,SteidelC.C., ReddyN.A.,2005, arXiv:astro-
account for a limited number of “extreme”-type galaxies, ph/0511498
such as extreme starbursts or massive star-forming galax- Berger E., Penprase B. E., Fox D. B., Kulkarni S. R., Hill
ies. Furthermore, we do not include the effects of dust and, G., Schaefer B., Reed M., 2005, arXiv:astro-ph/0512280
as in Christensen et al. (2004), we only consider extincted Bicker J., Fritze-v.Alvensleben U., 2005, A&A,443, L19
SFRUV intheestimateoftheSFRUV/(LB/L∗B)ratio.That Bloom J. S., Odewahn S. C., Djorgovski S. G., Kulkarni
maynotbeaseriousdrawbacksinceourfocusisonlow-mass S.R.,HarrisonF.A.,KoreskoC.,NeugebauerG.,Armus
galaxiesthatonaveragesufferlessattenuationthanthemas- L.e. a., 1999, ApJL,518, L1
siveones.Inaddition,theamountofdusthasbeenshownto Boselli A., Gavazzi G., Donas J., Scodeggio M., 2001, AJ,
be limited in most host galaxies of GRBs (Le Floc’h et al. 121, 753
2006). The main limitation of this study is the wide red- Brinchmann J., Charlot S., White S. D. M., Tremonti C.,
shiftrangeofthe10observedhostgalaxies, withonlyafew Kauffmann G., Heckman T., Brinkmann J., 2004, MN-
galaxies in each narrow redshift bin. The inferences made RAS,351, 1151
therefrom need to be confirmed once the sample size has Bruzual G., Charlot S., 2003, MNRAS,344, 1000
grown by a factor of 5-10, with a number of hosts at sim- Cervino M., Luridiana V., 2005, astro-ph/0510411
ilar redshifts. We should expect to see still higher specific Chary R., Becklin E. E., ArmusL., 2002, ApJ, 566, 229
ratesin low-mass objects, asthesimulated catalogs tendto Christensen L., Hjorth J., Gorosabel J., 2004, A&A, 425,
show. The evolutionary effects displayed in Fig. 6 support 913 (Chr04)
this conclusion. Cole S., Norberg P., Baugh C. M., Frenk C. S., Bland-
Comparing an expanded host galaxy sample based on Hawthorn J., Bridges T., Cannon R., Colless M. e. a.,
Swiftdatawiththesimulatedcatalogsandotherwell-known 2001, MNRAS,326, 255
galaxy sub-populations, such as compact bluegalaxies may Courty S., Bj¨ornsson G., Gudmundsson E. H., 2004, MN-
beextremelyuseful.Hostsgalaxies shouldprovideaclearer RAS,354, 581 (Paper I)
view into the formation and evolution of galaxies and the Feulner G., Gabasch A., Salvato M., Drory N., Hopp U.,
role ofdifferentsub-populationstherein.Inparticular, they BenderR., 2005, ApJL, 633, L9
will help in investigating the faint end of the galaxy lumi- Fruchter A. S., Pian E., Gibbons R., Thorsett S. E., Fer-
nosityfunction.Ourresults,obtainedusinganumericalap- gusonH.,PetroL.,SahuK.C.,LivioM.e.a.,2000,ApJ,
proach,areconsistent withandconfirmthepictureofGRB 545, 664
hostgalaxiesthathasemergedlately:Theytendtohavelow FruchterA.S.,ThorsettS.E.,MetzgerM.R.,SahuK.C.,
mass, be blue in color and have relatively high specific star PetroL.,LivioM.,FergusonH.,PianE.e.a.,1999,ApJL,
formation rates. Host galaxies may then belong to thehigh 519, L13
specific SFR galaxy population, rather than the high SFR FynboJ. P.U., Jakobsson P., M¨oller P.,Hjorth J., Thom-
population. High-resolution simulations are required to de- senB.,AndersenM.I.,FruchterA.S.,Gorosabel J.e.a.,
termine,inamorequantitativeway,thecontributionofthe 2003, A&A,406, L63