Mon.Not.R.Astron.Soc.000,1–10(2010) Printed6January2011 (MNLATEXstylefilev2.2) The extinction properties of long GRB host galaxies from H and HeI recombination lines⋆ 1 K. Wiersema1 † 1 1UniversityofLeicester,UniversityRoad,LeicesterLE17RH 0 2 n a Accepted04January2011 J 4 ABSTRACT ] O Inthispaperweshowhowaself-consistenttreatmentofhydrogenandheliumemissionline fluxesofthehostsoflonggamma-rayburstscanresultinimprovedunderstandingofthedust C propertiesinthesegalaxies.Inparticular,wefindthatevenwithmodestsignaltonoisespec- . h troscopywecandifferentiatedifferentvaluesforRV,theratiooftotaltoselectiveextinction. p TheinclusionofPaschenandBrackettlines,evenatlowsignaltonoise,greatlyincreasethe - accuracy of the derived reddening. This method is often associated with strong systematic o errors,causedbythe needformultipleinstrumentsto coverthewide wavelengthrange,the r t requirementto separatestellar hydrogenabsorptionfromthenebularemission,andbecause s of the dependancy of the predicted line fluxes on the electron temperature. We show how a [ these three systematic errors can be negated, by using suitable instrumentation (in particu- larX-shooterontheVeryLargeTelescope)andwidewavelengthcoverage.Wedemonstrate 1 thismethodusinganextensiveopticalandnear-infraredspectroscopiccampaignofthehost v galaxyofgamma-rayburst060218(SN2006aj),obtainedwithFORS1,UVESandISAACon 4 theVLT,coveringabroadwavelengthrangewithbothhighandlowspectralresolution.We 1 8 contrastourfindingsofthissourcewithX-shooterdataofastarformingregioninthehostof 0 GRB100316D,andshowtheimprovementoverexistingpublishedfluxesoflongGRBhosts. . 1 Keywords: gamma-rays:bursts 0 1 1 : v 1 INTRODUCTION main driver of stellar winds. Measuring the gas phase iron abun- i X dancefromemissionlinesofgalaxiesiscomplicated,astheseemis- Aslonggamma-raybursts(hereafterreferredtosimplyasGRBs) sionlinesarefaintandironiseasilyandrapidlydepletedintodust r are produced by massive stars, as evidenced by their associated a grains. Oxygen isagood alternative: itsabundance can be easily supernovae (e.g. Galama et al. 1998; Hjorth et al 2003; Pian et measured(thelinesarebright)eitherdirectly(usingoxygenlinesat al. 2006; Starling et al. 2010; but see also Fynbo et al. 2006), variousionisationstagesincombinationwithelectrontemperature thereisaclearlinkbetweenhighmassstarformationandGRBs. anddensitysensitivelines);orindirectlythroughthestrongforbid- This link is apparent in the properties of their host galaxies: the denlinescoupledwithradiationtransportsimulations.Inorderto hostsshowbrightnebularemissionlinesassociatedwithactivestar deriveaccurateabundances,oneneedstocorrectfortheextinction formation, and show spectral energy distributions and morpholo- experiencedbythehotgasthatemitsthedifferentemissionlines. giesgenerallyconsistentwithblue,metal-poor,starformingdwarf Besides deriving element abundances from emission lines, galaxies(e.g.Savaglio,Glazebrook&LeBorgne2009).Fromhigh therearefurtherreasonswhyweareinterestedintheextinctionin resolutionimagingitisevidentthatthepositionsoflongGRBsin thestarformingregion(s)inthehostgalaxies.Particularlyinterest- their hosts are strongly correlated with thelocation of starforma- ingisthecombinationofafterglowspectroscopyandphotometry, tionwithinthesehosts(e.g.Fruchter etal2006). Itisthereforea which probes the line of sight dust and gas properties within the reasonableassumptiontomakethatthepropertiesofthestarform- host,withhostgalaxyspectroscopyinemission.Thiscombination ingregionsinthehostalsoleadtoinsightinthepropertiesofthe mayconstrainthepossibledestructionofdustbytheGRB,itmay progenitor star:theprogenitor only livesfor ashort time.A con- showhowspecialtheregioninwhichtheGRBtookplaceiswith strainingparameterontheevolutionoftheprogenitorobjectisits respect to other star forming regions, and may shed light on the metallicity,ormorespecificallytheabundanceofiron,asitisthe (depletion)chemistryintheISMinGRBhostgalaxies. In this paper we focus on the extinction in star forming re- ⋆ BasedonobservationsmadewithESOTelescopesattheParanalObser- gions withinGRB hosts (perhaps better described asthe attenua- vatoryunderprogrammeID381.D-0723 tionofemissionlinesbydustwithinthehosts).Thehostgalaxyof † E-mail:[email protected] GRB060218isaparticularlyattractivetargettodothis.Thesuper- 2 K. Wiersema novaaccompanyingthisGRBiswellstudied,andthereforeacon- nectionwithhostgalaxyproperties(spectroscopic,spectralenergy distribution;seee.g.Savaglio,Glazebrook&LeBorgne2009)may providetightconstraintsontheprogenitorevolution.Inaprevious paper(Wiersemaetal.2007;hereafterW07)weperformedastudy of the metal abundances in this host and the ISM velocity struc- ture,bothobtainedthroughemissionandabsorptionlinesdetected inaVLTUVESspectrumofthesupernova(2006aj).Weobtained threemoredatasetsofthissource,whichwewillusetoprobethe extinctionpropertiesofthishostinthispaper.Thesedataarepart Figure1.Thisplotshowsthepositioninrestframewavelengthofthemost ofadedicatedVLTprogram(program381.D-0723,PIWiersema) usefulHandHeIlines.Thehydrogenlines(thehydrogenseriesiseasily aimed at a thorough understanding of the stellar population(s) of recognisablebyeye)areindicatedbyaredsquare,withfluxesnormalised thishostgalaxy,andconsistofopticalspectroscopyatmediumand toHβforcaseBrecombinationandTe =104Kandne =100cm−3.The HeIlinesareshownwithbluetriangles,theirfluxesarenormalisedtoHeI highresolution(usingtheFORS1andUVESinstruments)andlow resolutionKbandspectroscopy(usingtheISAACinstrument). λ5876,forthesameTeandneashydrogen.Thenominalrangesofthethree armsofX-shooter areindicated byhorizontal dotted lines. However, we Thispaper isorganizedasfollows: inSection2wedescribe notethattheNIRarmefficiencydropsveryrapidlyredwardoftheKband. thewayweuseHandHeIlinefluxratiostofittheextinctionprop- ItisclearfromthisplotthattheHeIlinesfillinimportantwavelengthgaps erties; inSection3 wedescribe the observations and theway we left bythe hydrogen series –amajority oftheX-shooter spectral orders measureemissionlinefluxesfromthespectra;inSection4wefit containsoneormoreemissionline. thedatausingourmodelsandinSection5wediscussourresults. Throughout the paper we adopt a cosmology with H = 71 0 kms−1Mpc−1,Ω =0.27,Ω =0.73. m Λ then proceed tomeasure theline fluxes and upper limitson non- detectedlines.HydrogenlinefluxesarenormalizedtoHβ,whereas fortheHeIlinesweuseHeIλ5876:theselinesaremostfrequently 2 METHOD detectedinlow-zhostsandarebright.Forthepredictedfluxratios R weuseafixedelectrondensityn =102cm−3:thevastmajority Weassumethattheforegroundabsorbingscreenofdustisuniform, p e ofGRBhostsandstarformingregionswithinthesehostsshowflux andexpresstheextinctionfunctioninA /A andtheobservedand λ V ratiosofthe[OII]λ3727,3729doubletorthe[SII]λ6716,6731 predictedratiosoftwoemissionlines,R andR atwavelengthsλ o p 1 doubletinthelowdensitylimit.Notealsothatatlowdensitiesthe andλ : 2 emissivities are fairly weak functions of the density (Osterbrock 2.5log(R /R ) 1989).WeuseaseriesofdifferentelectrontemperaturesT ,from A = o p . (1) e V (A −A )/A 0.5 to 2.5× 104 K, and obtain the predicted line flux ratios for λ2 λ1 V hydrogen fromHummer&Storey(1987) fortheseT valuesand e We now use the Cardelli, Clayton & Mathis (1989) electrondensity.HeIemissivitiesarestillsubjecttosomedebate, parametrizationoftheextinctioncurve, butgenerally thedifferencesinemissivitiesbetween different ap- proaches isconsiderably smaller thanthemeasurement errorswe A /A =a(λ)+b(λ)/R , λ V V canrealisticallyexpectfromeventhebrightestGRBhosts.Weuse withtheusualnotationR =A /E(B−V),forarangeofR values. theemissivitiesandtheinterpolationoverelectrontemperatureas V V V NotethatwewillusethisonlowredshiftGRBhosts,soweexpecta determinedbyPorteretal.(2005)andPorter,Ferland&MacAdam possible∼2200Åfeaturenottohaveaneffectontheanalysis.Any (2007;theirappendixA)tocomputeR forHeIusingthesameT p e lineratiowherethefluxratioscanbeaccuratelypredictedthrough gridandfixedn asforhydrogen. e atomicphysicscanbeused,butitismostcommontousehydrogen After the fluxes of the hydrogen and HeI lines in range are (HI) or helium (HeI or HeII) lines because of the large number measured(detailedinSection3),wearriveforeachsourceatagrid ofavailabletransitionsintheopticalwindow,theirbrightness,the ofvaluesofA /A (forarangeofR values)andthecorresponding λ V V lineemissivitiescanbeaccuratelymodelled,andbecausethereis 2.5log(R /R )values(forarangeofT ).Anexamplefortwodif- o p e nosensitivitytoabundance. ferentvaluesofR isshowninFigure7.Wenowfitastraightline V HeIIlinesaregenerallytoofaintinmostGRBhoststarform- tothepoints,foreachcombinationofR andT ,recordingtheχ2. V e ingregionstobeusefulinthiscontext,buttheHeIlinesdoprovide WedothisseperatelyforHandHeI.Theslopeofthebestfitting meaningfulconstraintsonA andR .Inaddition,theHeIlinesfill straight line gives the attenuation (it is equal to −A , see Eq. 1). V V V gapsinthewavelengthcoverageoftheHseries.Aswewilldemon- ThebestdescriptionofR andT isnowgivenbythemodelwith V e stratefurtherbelow,thisisparticularusefulforX-shooterspectra: thelowestχ2. mostordershaveaHorHeemissionlinein(atleastatthelowred- The use of hydrogen lines to derive R and A simultane- V V shiftsweconsider inthispaper).Figure1showsthisgraphically: ouslyiscommonlyusedingalacticemissionlineregions,seee.g. we plot the wavelength and flux of the relevant H and HI lines, Greve(2010)andPetersen&Gammelgaard(1997)andreferences wherethehydrogenfluxesarenormalisedtoHβandtheHeIfluxes therein.Theseauthorspointoutafurthersimplificationthatisuse- toHeIλ5876,assumingcaseBrecombination, A = 0,T = 104 fulwhenonlylimitedlinesareavailable:thefluxratiosofemission V e Kandn =100cm−3(seebelowfordetails). linesfromthesamespeciesthat comefromthesameupper level e WenowcomputeA /A forthewavelengthsλofallusefulH arefairlyinsensitiveton andT . λ V e e andHeIlinesinthespectrausingarangeofR valuesfrom2.5to Wenowproceedtodemonstratethestrengthsandweaknesses V 5.5(seealsoPetersen&Gammelgaard1997).ItisclearthatA /A ofthemethodoutlinedaboveforthecaseofstarformingregions λ V changes most strongly with changing R for the bluer lines. We inlongGRBhostgalaxies. V The extinctionpropertiesof GRBhostgalaxiesfromH andHeI 3 The datawere reduced using standard routines inIRAF,us- ingroutinesintheccdprocandspecredpackages,usingcalibration data(biasframes,flatfields,arclampframesandstandardstarex- H α posures)takenthesamenightsasthesciencedata.Fluxcalibration ofthedataofSep4wasachievedusingexposuresofFtypespec- trophotometricstandardstarLTT1788(Hamuyetal.1992,1994). ThestandardstarforthesecondnightofdatawasDA4starHZ4 (Oke1990).Theextracted,fluxcalibrated,individualspectrafrom thetwonightswerecombined throughaweightedaverage,using the signal to noise per bin as weight. As expected, the signal to noiseisdominatedbythesecondnightspectra,thathaveconsider- ablybetterseeingconditions.Weappliedameanextinctioncurve [O III] 5008 forParanal,anddereddenedthespectrafortheGalacticextinction valueof E(B−V) = 0.142andR = 3.1,seeforadiscussion Gal V W07.Publishedphotometryofthehost(Sollermanetal.2006)was usedtobringthespectrumtoanabsolutescale.Wemeasureacon- tinuumsignaltonoiseof∼50perpixelandaarclineFWHM(as ameasurefortheresolution)of5.0Å(at∼5000Å). Amultitudeofemissionlinesaredetected,seeFigure2,where -150 -100 -50 0 50 100 150 we have labelled lines commonly found in high signal to noise Velocity (km/s) spectraofbluecompactgalaxiesandWRgalaxies(manylabelled linesarenotdetected).Inthefollowingwewillonlydiscussthehy- Figure4.ThetoppanelshowstheHαemissionlinefromtheUVESdata drogenandHeIlines,themetallineswillbediscussedinafuture describedinSection3.1.3;thelowerpanelshowsthe[OIII]λ5008emission linefromW07.Inadditiontotheselines,severalotheremissionlinesare publication(Wiersemaetalinprep.). resolved. 3.1.2 ISAACspectroscopy 3 OBSERVATIONS WeacquireddeeplowresolutionISAACKbandspectroscopy,see 3.1 ThehostofGRB060218 Table 1. Data were reduced using the ESO ISAAC data reduc- tionsuite,version 5.7.0, using calibrationdata (darkframes, flat- ThedataofthehostgalaxyofGRB060218presentedinthispaper fields,arclampframes)takenthesamenight.Reductionincluded were taken with the aim of characterizing both the young stellar skysubtraction,flatfielding,wavelengthcalibration,spectralimage population that produced the GRB itself (e.g. the WR star popu- registrationandco-addition.Thedataweretakenusing30arcsec- lation) and apossible underlying old stellarpopulation (seeW07 ondnodthrows,anda10arcsecond jitterboxwidth.Wavelength forthemotivation).Inthispaperwewillrestrictourselvespredom- calibration from the arc line frames was refined using night sky inantly tothe hydrogen and helium linesand the inferred attenu- lines, and the host galaxy spectrum was extracted using tasks in ation properties, further physical properties of the source will be IRAF.ObservationsoftheB3VstarHip021575(HD29376)were discussedinfuturepublications. taken for telluricline calibration: telluriccorrection and flux cal- GeneraldetailsofourobservationsarelistedinTable1. ibration was achieved using the IDL / SpeXtool package (Cush- ing,Vacca&Rayner2004),inparticularthextellcor generaltask, whichusesahighqualityVegamodelempiricallyconvolvedwith 3.1.1 FORS1spectroscopy knowntelluricabsorptionfeaturestomatchtheentiretelluricstan- WetargetedthishostwithFORS1,asthisinstrumenthasbeenblue dardspectrumindetectorunits(therebyimplicitlysolvingforthe optimised (the similar FORS2 instrument is red optimised). The instrument response). The resulting model is then applied to the 600Bgrismwasused.Whilethisgrismdoesnotgiveuscoverage sciencetargetspectrum,producingatelluricfeaturecorrectedand attheredendofthevisualrange,itshigherresolutioncomparedto fluxcalibratedspectrum(exceptingslitlosses).Tobringthespec- e.g.the300Vgrism(astudyofthishostwiththisgrismispresented trum onto an absolute flux scale, one generally uses photometry inHanetal.2010)allowsmoreaccuratedeblendingofblueBalmer of the source. In the case of this source, Kocevski et al. (2007) linesandmoreaccuratefitsforstellarabsorption.Inaddition,the giveK =18.73±0.34,uncorrectedforGalacticextinction.How- s higherresolution,thoughstilllow,allowsamoreaccurateviewof ever,theKocevskietal.measurementsweredonewithpoorspatial theemissionlinesmaking uptheWRstar bumpclaimed byHan resolution. We therefore examine the reduced acquisition images et al. (2010). We selected a 1.3 arcsecond slit. The width of the from the second epoch of ISAAC data, which provide excellent slitwaschosenbasedonhighresolutionimagingfromtheHubble pixelscaleundergoodseeingconditions.Wecombinethethree10 SpaceTelescope(HST),seee.g.Starlingetal.(2010)foranHST second acquisitionexposures (K filter),andtietheastrometryto s imageofthehost,sothatwearemorelikelytogetallthegalaxy the FORS1 acquisition images. The host galaxy is detected, and lightintheslit. isresolved,butdetectionsignificanceislow.Weperformaperture Observationswereperformedattwoepochs,withpoorseeing photometrywithrespecttofour2MASSstarsinthefield.Wefind conditionsinthefirstepoch.Acquisitionwasdoneusingdirecton- K = 18.3±0.3 for the host, but because of the poor detection s sourceacquisitiononthesourcephoto-center,asthehostisbright rateandthesmallnumberof2MASSstarsinthefield,weusethe foraGRBhost(r =20.16±0.03;Sollermanetal.2006),thestar Kocevskietal.valuetobringthespectrumtoanabsolutescale. forming region(s) (see Section 3.1.3 and W07) are located at or Theslitpositionanglewaschosensuchthatanearby2MASS nearthecenterofthehostgalaxy. star fell into the slit, with the aim of providing a more accurate 4 K. Wiersema Table1.VLTobservationsofthehostofGRB060218.Theseeingisasmeasuredfromtheacquisitionimage(s). Inst.+grism+slitwidth(”) Date Exptime(s) Seeing(”) Airmass FORS1+600B+1.3” 2008-09-04 2×1345 1.5 1.3 2008-10-01 2×1345+1×900 1.1 1.3 Inst.+dichroiccenterwl(filters)+slitwidth(”) Date Exptime(s) Seeing(”) Airmass UVES+437and860nm(HER5+OG590)+1.2 2008-08-02 1400 0.9 1.4 Inst.+mode+centerwl(µm)+slitwidth(”) Date Exptime(s) Seeing(”) Airmass ISAAC+SWS1-LR+2.2+1.0 2008-07-17 12×60 0.8 1.7 2008-09-10 108×60 0.4 1.5 fluxcalibrationincasetheacquisitiondatawerenotdeepenough 3.2 ThecaseforX-shooter:sourceAinthehostof to detect the host, particularly since the value from Kocevski et GRB100316D al.(2007)hasalargeuncertainty.However,thespectrumandacqui- To demonstrate the advantages X-shooter offers compared to our sition images show that this object, 2MASS 03213974+1651305 observationswithmultipleinstrumentsofthehostofGRB060218, (K=12.55)isinfactablendoftwostars,separatedby0.9arcsec- wemakeuseofthespectraweobtainedofastarformingregionin onds,andwiththecomponentsnotfullycoveredbytheslit.Weare thehostgalaxyofGRB100316D. Thisstarformingregionisnot thereforeunabletofurtherfinetunetheabsolutefluxoffsetofour theonehostingtheGRB,butsimplyservesasanillustrationofthe spectrumusingthisobject. methodsused.Theobservations,thedatareductionandcalibration and the basic properties of this region are reported in Starling et al.(2010).Moredetailedpropertiesofthishost,includingthestar 3.1.3 UVESspectroscopy formingregionwheretheGRBoccured,willbereportedinafuture paper(Floresetal.inprep.). We acquired a high resolution spectrum of this source with the aimofverifyingthemultiplevelocitycomponentsobservedinboth emissionandabsorption([OIII]andNaI+CaII,respectively)in 3.3 Linefluxmeasurements UVES spectroscopy of SN2006aj, the supernova accompanying GRB060218 (W07). We selected the central wavelengths of the Besides the narrow nebular emission lines, the spectra also show spectra such that we would cover both the Hα emission line and absorption components. As demonstrated for Hδ in Figure 5, a the [OII] doublet in the red and blue arms, lines that will aid in clearabsorptioncomponentispresentunderneaththeemissionline, detecting changing physical parameters over the various velocity made by stellar atmosphere absorption. The strength (equivalent components(seealsoSection5.1). width) of the absorption component of the line is a strong func- Becauseofaspatialcoordinateoffseterroronlytheexposures tion of the age of the stellar population dominating the observed taken on Aug 2 can be used here. We reduce the data using the luminosity, the starformation history (starburst, continuous) and standardprescriptionswithintheESOUVESdatareductionpack- the metallicity (e.g. Gonzalez Delgado & Leitherer 1999). Most age,version4.2.4.Acrudefluxcalibrationisachievedusingspec- ofthesepropertiesarepoorlyconstrainedorcompletelyunknown trophotometricstandardstars(EG274;Hamuyetal.1992, 1994) for GRB hosts. Subtracting a constant equivalent widthof all in- takenonAug5andAug8,nospectrophotometricstandardswere tegrated hydrogen line fluxes to account for absorption (as done observedonthenightthedataweretaken.Becauseoftheshortex- inLevesqueetal.2010andforsomesourcesinSavaglio,Glaze- posuretime,nocontinuumemissionisdetected,butwedodetecta brook & Le Borgne 2009) will lead to inaccurate line fluxes and dozenemissionlines,includingthethreebrightestBalmerlinesand createinconsistencieswheninferringstellarpopulationproperties. the[OII]lines.Inthispaperweonlyusethevelocityinformationin MoreaccurateisthemethodemployedbyHanetal.(2010),who someofthelines,andfurthercalibrationofthelinefluxesthrough fit a linear combination of single stellar population models from theFORS1dataandthedatainW07willbedoneinfuturepapers. the evolutionary population synthesis code of Bruzual & Charlot TheUVESspectrumisconvertedintovacuumwavelengthsusing (2003).AsGRBstarformingregionsareoftenconsideredbursty, theIRAFdisptranstaskandbroughtontoaheliocentricwavelength metalpoorandthedominantstellarpopulationislikelyveryyoung scale,afterwhichwetransformthewavelengthsontothesolution (astheGRBsformfromthemassiveendoftheIMF),weprefernot ofW07bymeasuringthewavelengthoffsetsofsixlinesdetected torelyonmodels.Weperformfreefitsonthecontinuum,emission in both the current UVESdata and the UVES data inW07. This andabsorptionlines,acceptingthismaycomewithlargererrorson allowsustodirectlycomparetheprofilesofthedetectedemission theemissionlinefluxes.Thechoiceofspectralresolutionandre- lineswiththeonesintheW07UVESdataonafixedrelativeveloc- quiredsignaltonoiseforthe060218FORS1datawasmotivatedby ityscale.InFigure4weshowacomparisonoftheHαemissionline thenecessityoffittingabsorptionlinecomponentsinthismanner. fromthedatadescribedabovewiththe[OIII]λ5008lineasanal- The emission and absorption components are adequately fit ysedinW07.Thecomplexlinemorphologyseeninbothemission by Gaussian profiles (except in the UVES data of the host of andabsorptionislikelycausedbytwoormorestarformingregions GRB060218, seeSection5.1).Wefiteach line regionseparately and/orabrightstarformingregionwithoutflow(s)andgalaxydisk- using the ngaussfit task in the stsdas package within IRAF. We likeemission.TheHSTimagingdataarenotabletoresolvethis, startedthefittingwiththeHδline,whoseabsorptionandemission butthedetectionofseveralresolvedemissionlinesmayenableus components are easy to separate and have good signal to noise. toconstrainthesemodelssomewhat(Wiersemaetalinprep.). AsdemonstratedinFigure5wefitsimultaneouslythecontinuum The extinctionpropertiesof GRBhostgalaxiesfromH andHeI 5 Figure2.TheVLTFORS1600Bspectrum(seeSection3.1.1)ofthehostofGRB060218,plottedinrestframewavelengthscale.Indicatedbyverticallinesare strongandweak(nebular)emissionlines,aswellassomeabsorptionfeatures,commonlydetectedinstarforminggalaxies(notethatnotallthelabelledlines aresignificantlydetected).LineslabelledinitalicfontindicateGalacticabsorptionlines.Clearabsorptioncomponentscanbeseenunderneaththehydrogen andheliumemissionlines.Atthescaleofthisplot,theerrorspectrumisnotvisible(apartfromthelowestsignalarea,thedashedlinearound∼3200Å). (asafirstorderpolynomial);theemissionlinecenterwavelength, thesefreefitsontheHδlineswefound theabsorption andemis- Gaussianfullwidthathalfmaximum(FWHM)andamplitude;and sion line to be centered at the same wavelength to within errors. absorption line center wavelength, Gaussian FWHM and ampli- Inthefitsoflineswithrelativelypoorsignaltonoisewetherefore tude. Errors on these fitted parameters are calculated by resam- tietheemissionandabsorptionlinecenterstogetherinthefits.In pling,whichweuseaschecksonfitquality.Fromtheseparameters severalcases,mostnotablytheBalmerlinesbluewardofHδ,theH we compute the emission and absorption line fluxes and equiva- orHeIemissionlinesareblendedwithotheremissionlines,orthe lentwidths,andtheir1σuncertainties.AscanbeseeninFigure5, underlyinghydrogenandheliumabsorptionlinesareblendedwith theFWHMof theabsorption lineisconsiderably largerthanthat other,unrelated,absorptionlines.Inseveralcasesitwaspossibleto oftheemissionline,allowinginmostcasesreasonablefits.From deblendbyfittinganadditionalGaussianfixedatthewavelengthof 6 K. Wiersema Figure3.TheISAACspectrumtakenwiththeLRKgrism,plottedinrestframewavelengths.Thetopplotshowsthecombined2dimensionalspectrumbefore telluriclinecorrectionandsensitivitycalibration.Thelowergraphshowstheextracted,telluriclinecorrectedandfluxcalibratedspectrum.Severalemission andabsorptionlinesthatarecommonlydetectedin(ultracompact)HIIregions,starburstgalaxiesandWRstarsareindicated.Whereascontinuumemissionis detectedovernearlytheentirespectralrange,mostofthelinesindicatedarenotsignificantlydetected. theadditionalcomponent(forexampletheHǫtransitionisblended Ofthesourceswithfourdetectedlines,allbut020903havepoorfit with[NeIII]),butinsomecasestheresolutionoftheFORS1data χ2,inmostcasesdominatedbyasingleemissionline(oftenHδ). wasnotsufficienttoobtainaccurateemissionandabsorptionfluxes, Thefitsfor020903havegoodreducedχ2values,butnopref- andwedonotusetheselinesinfurtheranalysis.Wewillreportthe erenceforaR orT isclearfromthisdata(thereducedχ2ranges V e absorption equivalent widthsand theirimplicationsforthestellar from0.93to1.05)becauseofthelowA (intherangeA =0.23to V V populationsinafuturepaper. 0.33forthewholeR ,T grid)andthesmallA /A range.Itisclear V e λ V Inthenear-infraredwefurtherexcludelinesthatareaffected fromthesefitsthatwerequiresignificantlymorelinesandA /A λ V bysharpresidualsfromtelluriclinecorrections,or,inthecaseof range to constrain these parameters. In addition we note that the sourceAinthehostof GRB100316D, bybadpixels.Wefurther Hanetal.sampledealsmainlywithhost-integrated spectra(with excludelinesfromtheanalysisthathaveverylargeerrors(insome the exception of GRB060505), and since each host may contain casesthisisbecauseofproximitytotheedgesoftheechelleorders multiple bright star forming regions, the measured attenuation is wheresignaltonoiseoccasionallydropssteeply,orwhentheline asuperposition of these,and not easilycoupled toGRBenviron- isrightinanareawheretelluricabsorptionbringstheresponseto mentparameters(i.e.theinfluenceofWRstarsassociatedwiththe nearzero). starformationepisodeproducingtheGRBprogenitor).Aswewill showinSection5,thismayalsobeevidentinourdataonthehost ofGRB060218throughourhighresolutionspectroscopy. 4 FITSOFTHEEXTINCTION 4.1 LiteratureBalmerfluxdata 4.2 ThehostofGRB060218 Several optical spectroscopic studieshavebeen performed onthe Itisclearfromthefitsinsection4.1thatalarger A /A baseline λ V nearby(long)GRBhostgalaxypopulation,generallywiththeaim isbeneficial.Thedataof thehostof GRB060218 arewellsuited ofcomparingtheGRBhostpopulationwithothersamplesofgalax- todemonstratethepossibilities.Thishosthasbeenobservedwith ies,seeforexampleHanetal.(2010),Levesqueetal.(2010).Han HST(e.g.Starlingetal.2010)showingittobebeabluecompact etalfitstellarpopulationmodels,therebyaccountingforstellarab- galaxywhichappearstocontainonebrightmassivestarformation sorption,andlistthefluxesoftheHα,β,γandδlinesforeighthost region(butseeSection5.1andWiersemaetalinprep.). galaxies (for the host of GRB060505 both the values for the in- Figure7(top)showsanexampleoftwoofthefits,atR =3.1 V tegratedhost spectrumand theregionwhere theGRBtook place and 5.0, and T = 104 K, for the hydrogen and helium lines de- e are given; Tho¨ne et al. 2008). Han et al. (2010) compute the ex- tectedintheFORS1and ISAACdatadescribed above. Notethat tinctionbyassumingR =3.1andT =104 K,usingtheratiosof becauseoftherelativelylowresolutionoftheFORS1600Bgrism V e twolinesatatime(e.g.Hα/Hβ).Figure6showsourfitsonthese wearenotabletoderiveaccuratefluxesofalldetectedhydrogen emissionlinefluxesfortwovaluesofR andT =104K.Asinour linesbecauseofblendingwithotheremissionlines(e.g.Hǫ)and/or V e method thefluxesarenormalisedtoHβ,thefitstoallsources go absorptionlines(linesbluewardofH9),thisiseasilyvisibleinFig- roughlythroughthesamepoint,andtheslopeisequalto−A .Itis ure2.Thisisimportant:eventhoughthelinesbluewardofHδget V clearfromthisfigurethatmostGRBhostsintheHanetal.sample everclosertogetherinA /A space(seeFigure1),thispartofthe λ V havelow A .ThiscannotnecessarilybeextrapolatedtoallGRB A /A rangeismostsensitivetoR changes.Afurthercaveatisthe V λ V V hostsingeneral:thereisabiasagainstdustysightlinesinthede- fluxofthePaαline,whoseuncertaintyislargebecauseofthelarge tectionof afterglows(seee.g.Greiner etal. 2011), whichmayin uncertaintyinthehostKbandmagnitude. turntranslateinabiasagainstdustrichhostgalaxies. Usingthefluxesasdeterminedabovewefindacceptablefits Ofthesourcesinthissample,onlyfourhavefourBalmerline for the hydrogen lines for most models on our T ,R grid, with e V detections(thehostsofGRBs990712,020903, 031203,060218), the lowest χ2 for the T = 1.5 × 104 K and R = 4.5 model e V makingitimpossibleinmostcasestodistinguishtheeffectsofR (χ2 = 1.3). The χ2 distribution shows that the models with high V red andT throughfitχ2 (wehavetwofreeparameterstofitforeach R valuesperformbetterthanmodelswithlowR (. 3.5)forall e V V pointinourT ,R grid:theslopeandoffsetof thestraightline). temperaturesconsidered.ForallmodelswithR &3.1,highertem- e V V The extinctionpropertiesof GRBhostgalaxiesfromH andHeI 7 Figure7. AnexampleoftwoofthefitsintheTe,RVgridonthedataofthehostofGRB060218(topfourpanels)andsourceAinthehostofGRB100316D (lowerfourpanels).Theleftcolumnshowsafitusingthehydrogenlines,withTe=104K,displayedarethefitsforRV =3.1and5.Therighthandcolumn showsafitoftheneutralheliumlines,usingthesameparameters.IndicatedarethecorrespondingbestfitvaluesforAV.Thesefitsarenottheoneswiththe bestχ2foreitherofthesources,butservetodemonstratetheinfluenceofRV onthefit,andTe=104Kallowseasycomparisonwithliteraturevalues.Note thatthefitsdisplayedfor100316DusealsothebluestHeIlineinthefit. peratures(T >12500K)fitbetterthanlowerones.Broadlyspeak- 4.3 X-shooterdata e ingtheseresultsindicateapreferenceforahightemperature,high HSTimagingofthehostgalaxyofGRB100316D showsthatthe R attenuation, consistent withthefindings in W07and a“grey” V sourcewerefertoassourceA(seeStarlingetal.2010fordetails extinctioncurve.Thebestfittingmodelcorrespondstoanattenua- tionA =1.02±0.16magnitudes.WefittheHeIlinesseperately ofthissource)isasingle,largestarformingregionunrelatedtothe V GRB.Weusethedataofthissourceonlytoshowthepossibilities andfind,asexpectedfromthesmallnumberoflinesdetectedand X-shooteroffersinthestudyofdustinGRBhosts. thelargeerrorbars,thatwecannotreliablydistinguishR andT V e preferences: reduced χ2 values range from 0.2 to 0.4. We do see WefitthefluxesdeterminedinSection3.3.Inthiscase,wedo thattheA valuesfoundfromtheHeIlinefitsareconsistentwith detectHeIlinesatredwavelengths,andbecauseofthehigherres- V thehydrogen ones. Adopting thesame parametersasfor thebest olutionofX-shootercomparedtoFORS1600Bgrism,wecande- hydrogenfitwegetA =1.3±0.7. blendseveralofthebluerBalmerlines(butnotH8).Unfortunately, V While these fits show the possibility to distinguish models theredshift of thishost galaxy(z = 0.0591, Starlinget al. 2010) with different R ,T from a relatively small number of lines, we issomewhat unfortunate:several ofthebrighterlinesinthenear- V e note that the difference in χ2 between the best and worst fitting infrared fall on areas heavily affected by telluric absorption (e.g. model(R =2.5,T =2×104)isonly1.7.Alargernumberofde- Paγ),onbadpixels(e.g.Brǫ)orareredshiftedoutofthewindow V e tectedlines,particularlyaround A /A ∼ 0.5andbluewardofHδ, (wavelengthsredwardsof∼2.26µmarestronglyaffectedbyther- λ V canimprovethisconsiderably.Takingthebestfittingmodelatface mal noise and left out of the analysis, this affects e.g. Brγ). We value,thepreferenceforaR largerthanthemeanGalacticvalue furthernotethatnoinfraredimagingofsourceAwasavailableat V of3.1isinteresting.LargevaluesofR areindicativeofsightlines thetimeofwritingtoassessthequalityofthefluxcalibration. V dominated by relatively large dust grains. Dust is extremely sen- Thehydrogenlinesarewellfitbythemodels,seeFigure7for sitivetoenvironmental properties,asevidenced bythewidevari- twoexampleswhichclearlyshowasignificantlylowerattenuation ety of Galactic extinction law shapes (Valencic, Clayton & Gor- thaninthehostofGRB060218.Thislowattenuationmeansthere don2004),andfairlygreyextinctioncurvesmaypointtoregions islittlesensitivityto R and T .The χ2 has abroad minimumat V e withmassivestars(and/ordenseenvironmentswithshockeffects) χ2 =0.86forT =7500KandR =3.5.ForallvaluesofR the red e V V and/or fairly incomplete dust mixing and processing through the χ2 increaseswithincreasingelectrontemperature(toamaximum galaxy. of8.7at20000K,i.e.1.0abovethebestfit),allowingustocon- WecautionthattheUVESspectroscopygivessometentative cludetheelectrontemperatureinsourceAislikely. 104 K.For hints for more than one star forming region in this host (perhaps eachtemperaturethereisverylittlechangeinχ2withchangingRV, two,iftheGaussiancomponentsmakingupthelinearetwodistinct theentirerangeRV = 2.5−5.5onlychangesχ2 by0.1,whichis starformingregions),whichwasnotevidentinHSTimaging.This expectedwhenAV isverylow.TheAV foundforwhatisformally mayimplythatthefitsweperformhereareofthesuperpositionof thebestfittingmodel(Te=7500K,RV =4.5)isAV =0.19±0.06. thedustpropertiesofthesetworegions,withpotentiallydifferent TheHeIfluxesarealsowellfitwiththemodels(seeFigure7 R andA .Wethereforewillnotcomparethedustpropertiesfound foranexample).Wefitthedatasettwice,includingandexcluding V V abovefurtherwiththosefoundfromabsorptionspectroscopy. theHeIlineatrestframe3187 Å:thislineisobservedat3376Å 8 K. Wiersema Figure6.AnexampleofoneofthefitsontheBalmerHα,Hβ,HγandHδ emissionlinesfromaseriesoflongGRBhostsfromHanetal(2010).Both plotsshowafitusingTe = 104 K,withinthetopplotRV = 3.1,andthe lowerplotRV =5.TheAVvaluesthatcomefromthesefitsareshowninthe panels.Thesefitsarenotnecessarilytheoneswiththebestχ2,butnicely demonstratetheinfluenceofRVonthefit. agreementwiththeHresults,butsimilartotheHlinestheχ2does Figure 5. Fit of the Hδ lines of the X-shooter spectrum of source A in notchangemuch:fortheT =104 Kmodelstheχ2 changesfrom e the host of GRB100316D (top) and the FORS spectrum of the host of 3.2to3.9fromR = 2.5to5.5.Thebestfittingmodel(T = 104 V e GRB060218 (bottom). The weak emission line blueward of Hδ in both KandR =3.0)givesχ2 =0.8forA =0.43±0.31. spectra is[SII],whichisincluded inthefit.Thelineisclearly madeup V red V ofanemissioncomponent(formedbyhotgas)andanabsorptioncompo- nent (from stellar atmosphere absorption). Both emissionand absorption linesareadequatelyfitbyGaussianprofiles.Theresultingfitisshownin 5 DISCUSSION red,thebluedashedlineindicates thecontinuumlevel. TheHδemission 5.1 Caveats,implicitassumptionsandcalibration linefluxesforthesetwosourcesarevirtuallyidentical,whiletheabsorption equivalentwidthofthehostofGRB060218issignificantlylower.Thisplot requirements demonstratesthenecessityofadequatespectralresolution. In the list below we briefly outline a few important caveats, im- plicitassumptionsandcalibrationrequirementstoallowthefitsas describedabovetobesuccessful.Weillustratethesethroughthefit andthevalidityofthemeanParanalatmosphericextinctioncurve resultsofthetwosourcesabove. isuncertainattheseverybluewavelengths.Thefactthatthisline is even detected is quite remarkable, and will prove valuable for • Signal to noise: As can be seen from the poor results de- GRBhostsathigherredshifts,asitextendssensitivitybeyondthe scribedinSection4.1,werequiregoodsignaltonoisetodetectalso Balmerlimit(seeFigure1).Wefindthatincludingthisuncertain theweakerblueBalmerlinesandtheweakerPaschenandBrackett line flux in the fit increases the χ2 significantly (doubling it for lines.Theheliumlinesareevenfainterthanthehydrogenlines,but mostmodels),butweshowafitincludingthislineinFigure7for formanimportantadditionparticularlywheninfraredspectroscopy completeness.Excludingthisline,theHeIlinefluxesshowbroad isnotpossibleorwhentheredshiftisfairlyhigh(Figure1). The extinctionpropertiesof GRBhostgalaxiesfromH andHeI 9 • Redshift:thesignaltonoiserequirementsnaturallytranslate 5.2 TheadvantagesofX-shooter inaredshift requirement, but theredshift playsamoreimportant The list of caveats and requirements above shows the necessity roleinthepositionoftheemissionlines.Thefitsaboveshowthe of covering a wide (restframe) wavelength range but also the in- necessityofdetectionofoneormoreinfraredlines(e.g.Paα−δand herent problems of using multipleinstruments todo this. TheX- Brγ,δ),whichbecomeincreasinglyhardtodetectwithincreasing shooterspectrographontheESOVLTsolvedmanyoftheseissues. redshift. The(fixed)mediumresolutionofX-shooterissufficienttodeblend • Highresolutionimaging:theredshiftplaysaroleinthisre- manyoftheblueBalmerlines,providegoodfitsontheabsorption quirementtoo.Werequirehighresolution(i.e.spacebasedimag- components and to give good telluric line corrections. The very ing)toestablishexactlywhichregionsofthehostarecoveredby largewavelengthrange(inFigure1weshowthez=0wavelength theslit,andtomakesurewearenotaveragingmultiplestarforming coverageofthethreeX-shooterarms)meansthatalargenumberof regionstogether.Additionally,acomparisonwithdustpropertiesin linesarewithinreach,and,becauseoftheveryhighefficiencyof absorptionfromtheafterglowspectrawillrequireunderstandingof theinstrument,wedetectalsothefainterlines.Thearchitectureof thespatialpositionoftheGRBwithinthehost. theinstrument issuchthat acquisitionisdone simultaneously for • Spectralresolution:Asdiscussed inSection3.3andFigure allthreearms,meaningthattheslitscovertheexactsameregions 5,ahighspectralresolutionandsignaltonoiseisrequiredtoeffec- inthehost.Thesimultaneouscoverageofallthreearmsmeansthat tivelyfittheabsorptioncomponentsundertheHandHeIlines,and some consistency in flux calibration can be achieved through the to deblend these lines from other emission and absorption lines, continuumaswell.Incaseswherethereistroublebringingtherel- This is particularly important for the higher Balmer lines, which ativefluxcalibrationforeacharmontoanabsolutescale,thereare arevitaltodiagnoseR .Thereisanadditionaluseofhighresolu- V insome cases sufficient linesin each individual arm toallow fits tionspectra,asdemonstratedinFigure4:theresolved[OIII]lines foreacharmseparately. Usingothersources(stars)intheslitfor (confirmedbytheHαdata)canbeexplainedbytwostarforming additional flux calibration, as described in Section 3.1.2 will not regions (W07). The detection of two velocity components in ab- bepossibleduetotheshortslit.Thereisstillastrongrequirement sorptioninthesupernovaspectraindicatesthatinthisscenariothe for accurate pre-imaging, as the X-shooter acquisition CCD only GRBislocatedintheredmostcomponent. However,thereareal- operatesintheopticalwavelengthrange.Anadditionalbenefitof ternativescenariostoexplainthemultiplecomponents(Wiersema themethodsshownaboveisthatafitonthelinesintheUVBand etalinprep).InthecaseofouranalysisofthehostofGRB060218 VISarmsprovidesstrongenoughconstraintsontheattenuationpa- theUVESspectramayimplythatwearereallyseeingtheaverage rametersthattheresultingfitcanbeusedtorefinefluxcalibration propertiesoftwostarformingregionsintheanalysisinSection4.2. intheNIRarm,providingawaytogetadditionalfluxcalibration • Calibration: The wide wavelength range required for this fromthesciencedataitself. analysisoftenmeansacquiringdatawithmorethanoneinstrument The data above show some of these advantages already, (butsee5.2).Thisputsrequirementsontheabsolutefluxcalibra- though theexposure timefor this dataset isshort, and conditions tion:withinoneinstrument(orX-shooterarm)arelativefluxcali- werenotideal,asthedataweretakeninToOmode(seeStarlinget brationwillbesufficientasweconsiderfluxratios,butwhencon- al.2010).Highersignaltonoiseobservationsarelikelytoimprove sidering multiple instruments, an absolute calibration is required. theresultsinSection4.3. Eachwavelengthrangeoffersitsownproblems,inadditiontothe commonproblemswithabsolutefluxcalibration(spectrophotome- try),illustratedabove.Particularlyintheinfraredthiscanbevery 5.3 Reddening,extinctionandattenuationinGRBhosts difficult, as the telluric transmission has to be fit for in addition tothesensitivity,slitlossesandextinction,puttingadditionalcon- Longgamma-rayburstsareaccompaniedbybrightafterglowsand straintsontheobservingconditionsandtelluricstandardstars(e.g. inmostcasesbybrightsupernovae.ThismakesGRBsidealprobes theirspectraltype).Wenotethatahigherspectral resolutionalso of distant galaxies: spectra can be obtained of the afterglows to greatlyimprovesthequalityofthetelluriccalibration. study the host ISM in absorption, and, when the afterglow has Tofinalizethefluxcalibrationwerequiregoodphotometry,but fadedaway,spectraofthehostprobethehostgalaxyinemission. formanyGRBhoststhereisalackofhighqualityphotometryat DustpropertiesinGRBhostsareofparticularinterest,becauseof infraredwavelengths.ThisisillustratedinSection4.2.Therequire- thedegeneracyofdustreddeningwithotherparametersofinterest ment forgood multiband imagingalsoplays animportant rolein (e.g.deviationsfromsynchrotronmodels,searchesforhighredshift the acquisition process. GRB hosts are extended sources and the burststhroughLymanbreaksetc.),thepossiblemetallicitylimitson starformingregionsofinterestareblue,meaningthatdirectacqui- GRBprogenitors(e.g.theamountofmetalslockedupindust),the sitioncanbeproblematicwhen multipleinstrumentsfor different influenceofdustscatteringonafterglowpolarimetry,thenatureof wavelengthregionsareinvolved,asthephotocenteroftheregions darkbursts,etcetc. ofinterestiswavelengthdependent. Broadly speaking the dust properties of GRB hosts can be • Aftergloworsupernovaspectra:Tomaximizethescientific studiedusingtheaftergloworusingthehostgalaxyinemission.In returnitishelpfultohavehighresolutionspectroscopyoftheaf- thefirstcase,theamountofreddeningexperiencedbytheafterglow terglow or SN in question. The detection of dust-tracing absorp- andtheshapeoftheextinctioncurve(orthegrainsizedistribution) tionlines(suchasNaI,KIordiffuseinterstellarbands)willallow canbederivedthroughfitsontheafterglowspectralenergydistri- comparisonofthelineofsightextinctionproperties;anydetected bution(e.g.Galama&Wijers2001;Starlingetal.2007;Schadyet velocitystructureofabsorptionlineswillaidindeterminingthepo- al.2010andreferencestherein),throughstudiesoftherelativecol- sitionoftheGRBwithrespecttothestarformingregion(seeW07 umndensitiesofgasphasemetals(e.g.Savaglio&Fall2004)and andthediscussionoftheUVESdataabove);andknowledgeofthe the wavelength dependancy of afterglow polarisation (e.g. Klose SNbehaviourwillhelpinestimatinganyresidualsupernovacon- et al. 2004; Wiersemaet al in prep). The distance of the gasand tinuum or nebular phase contribution to the measured fluxes (see dust that is seen in absorption in afterglow spectra can be deter- alsoKu¨pcu¨ Yoldas¸,Greiner&Perna2006). minedthroughvariableexcitedmetastablefinestructurelines(e.g. 10 K. Wiersema Vreeswijketal.2007), excitedmolecular transitions(e.g.Sheffer Han,X.H.,Hammer,F.,Liang,Y.C.,Flores,H.,Rodrigues,M., etal.2009),ionisationstates(e.g.Prochaska,Chen&Bloom2006) Hou,J.L.,Wei,J.Y.2010,A&A514,24 andanysignsofdustdestructionbytheGRBphotonfield(i.e.re- HjorthJ.etal.,2003,Nature,423,847 lease of elements into the gas phase that were locked up in dust Hummer,D.G.&Storey,P.J.1987MNRAS,224,801 grainsleadstoobservablevariabilityofresonancelines). Klose,S.,Palazzi,E.,Masetti,N.,Stecklum,B.,Greiner,J.,Hart- Thestudyofdustinhostgalaxiescanbeapproachedbystudy- mann,D.H.,Schmid,H.M2004,A&A420,899 ingdustinthestarformingregionshostingGRBs,whichmeasures Ku¨pcu¨ Yoldas¸,A.,Greiner,J.&Perna,R.2006,A&A457,115 theinfluenceofdustonemissionlineemittinghotgas(inthispa- Kocevski,D.etal.2007,ApJ663,1180 per),orbySEDsofthegalaxies(measuringthereddeningexperi- Levesque,E.M.,Berger,E.,Kewley,L.J.,Bagley,M.M.2010, encedbystars).ItisclearthattheA valuesandextinctioncurves AJ139,694 V derived through all these methods are not probing the same dust Oke,J.B.1990,AJ99,1621 columnsandspatial distributionofdust, butprovidecomplemen- Osterbrock, 1989, Astrophysics of gaseous nebulae and active taryapproachestounderstandthepropertiesofdustinstarforming galacticnuclei,UniversityScienceBooks dwarfgalaxies. Petersen,L.&Gammelgaard,P.1997,A&A323,697 PianE.etal.,2006,Nature,442,1011 Porter, R. L.; Bauman, R. P.; Ferland, G. J.; MacAdam, K. B. 2005,ApJ622,L73 6 CONCLUSIONS Porter,R.L.,Ferland,G.J.&MacAdam,K.B.2007,ApJ,657, We have presented a spectroscopic dataset of the host of 327 GRB060218,which,togetherwithalreadypublisheddataforastar Prochaska,J.X.,Chen,H.-W.&Bloom,J.S.2006,ApJ648,95 formingtegioninthehostofGRB100316D,weusetogetfurther Savaglio,S.,Glazebrook,K.,LeBorgne,D.2009,ApJ691182 insightintotheattenuationof theemissionlinesbydust.Specifi- Savaglio,S.&Fall,S.M.2004,ApJ614,293 callyweshowhowhydrogenandheliumfluxesofmanyrecombi- Schady,P.etal.2010,MNRAS401,2773 nationtransitionscanbefitsimultaneously togetinsight intothe Schlegel,D.J.,Finkbeiner,D.P.&Davis,M.1998,ApJ,500,525 electrontemperatureandtheR ofthedust.X-shooteristheideal Sheffer,Y.,Prochaska,J.X.,Draine,B.T.,Perley,D.A.,Bloom, V instrumenttodothisforalargesampleofGRBhosts,aslongas J.S.2009,ApJ701,L63 accuratecalibrationcanbeachieved.Theresultingdustproperties Sollerman,J.,Jaunsen,A.O.,Fynbo,J.P.U.,etal.2006,A&A, fromrecombinationlinescanbecombinedwithinformationfrom 454,503 absorptionlinesofafterglowspectra;extinctioncurvesfromafter- Starling, R. L. C., Wijers, R. A. M. J., Wiersema, K., Rol, E., glowspectralenergydistributionsandreddeningfitsonintegrated Curran,P.A.,Kouveliotou,C.,vanderHorst,A.J.,Heemskerk, host galaxy spectral energy distributions to study dust formation M.H.M2007,ApJ661,787 anddestructioninlowmetallicitydwarfgalaxies. Starling,R.L.C.etal.2010,MNRASinpress,arXiv:1004.2919 Tho¨ne,C.C.,etal.2008,ApJ,676,1151 Valencic,L.A.,Clayton,G.C.&Gordon,K.D.2004,ApJ616, 912 ACKNOWLEDGMENTS Vreeswijk,P.M.etal.2007,A&A468,83 We thank Sandra Savaglio, Johan Fynbo, Cedric Ledoux, Palle Wiersema,K.,etal.2007,A&A,464,529 Møller, Sara Ellison, Sung-Chul Yoon, Elena Pian, Rhaana Star- ling,RalphWijers,NialTanvirandPaulVreeswijkforusefuldis- cussion, and the anonymous referee for helpful comments. We thank the ESO staff for obtaining the data discussed here. KW acknowledges support from STFC. The financial support of the British Council and Platform Beta Techniek through the Partner- shipProgrammeinScience(PPSWS005)isgratefullyacknowl- edged. REFERENCES Cardelli,J.A.,Clayton,G.C.&Mathis,J.S.1989,ApJ,345,245 Cushing,M.C.,Vacca,M.D.,&Rayner,J.T.2004,PASP116,362 Fruchter,A.S.,etal.,2006,Nature,441,463 Fynbo,J.P.U.etal.2006,Nature444,1047 GalamaT.J.etal.,1998,Nature,395,670 Galama,T.J.&Wijers,R.A.M.J.2001,ApJ549,L209 GonzalezDelgado,R.M.&Leitherer,C.1999,ApJS125,479 Greve,A.2010,A&A518,62 Greiner,J.etal.2011,A&A526,30 Hamuy,M.,Walker,A.R.,Suntzeff,N.B.,Gigoux,P.,Heathcote, S.R.,Phillips,M.M.1992,PASP104,533 Hamuy, M., Suntzeff, N. B., Heathcote, S. R., Walker, A. R., Gigoux,P.,Phillips,M.M.1994,PASP106,566