NASA Technical Reports Server (NTRS) 20140010551: Hubble Space Telescope Hx Imaging of Star-forming Galaxies at z approximately equal to 1-1.5: Evolution in the Size and Luminosity of Giant H II Regions

1.5 MB·English

by  NASA Technical Reports Server (NTRS)

Mon.Not.R.Astron.Soc.427,688–702(2012) doi:10.1111/j.1365-2966.2012.21900.x
α (cid:2) .
Hubble Space Telescope H imaging of star-forming galaxies at z 1–1 5:
evolution in the size and luminosity of giant HII regions
(cid:2)
R. C. Livermore,1 T. Jones,2 J. Richard,1,3 R. G. Bower,1 R. S. Ellis,2
A. M. Swinbank,1 J. R. Rigby,4 Ian Smail,1 S. Arribas,5 J. Rodriguez-Zaurin,6
L. Colina,5 H. Ebeling7 and R. A. Crain8
1InstituteforComputationalCosmology,DurhamUniversity,SouthRoad,DurhamDH13LE
2AstronomyDepartment,CaliforniaInstituteofTechnology,MC249-17,Pasadena,CA91125,USA
3CRALObservatoiredeLyon,9AvenueCharlesAndre´,69561Saint-Genis-Laval,France
4NASAGoddardSpaceFlightCenter,Code665,Greenbelt,MD20771,USA
5CentrodeAstrobiologia,DepartamentodeAstrof´ısica,CSIC-INTA,Ctra.deAjalvirkm.4,28850Torrejo´ndeArdoz,Madrid,Spain
6InstitutodeAstrof´ısicadeCanarias(IAC),C/ViaLa´cteas/n,E38205,LaLaguna,Tenerife,Spain
7InstituteforAstronomy,UniversityofHawaii,2680WoodlawnDrive,Honolulu,HI96822,USA
8LeidenObservatory,LeidenUniversity,POBox9513,2300RALeiden,theNetherlands
Accepted2012August8.Received2012July13;inoriginalform2012May23
ABSTRACT
We present Hubble Space Telescope/Wide Field Camera 3 narrow-band imaging of the Hα
emissioninasampleofeightgravitationallylensedgalaxiesatz=1–1.5.Themagnification
causedbytheforegroundclustersenablesustoobtainamediansourceplanespatialresolution
of 360pc, as well as providing magnifications in flux ranging from ∼10× to ∼50×. This
enablesustoidentifyresolvedstar-formingHIIregionsatthisepochandthereforestudytheir
Hαluminositydistributionsforcomparisonswithequivalentsamplesatz∼2andinthelocal
Universe.WefindevolutioninthebothluminosityandsurfacebrightnessofHIIregionswith
redshift.ThedistributionofclumppropertiescanbequantifiedwithanHIIregionluminosity
function,whichcanbefitbyapowerlawwithanexponentialbreakatsomecut-off,andwe
find that the cut-off evolves with redshift. We therefore conclude that ‘clumpy’ galaxies are
seen at high redshift because of the evolution of the cut-off mass; the galaxies themselves
followsimilarscalingrelationstothoseatz=0,buttheirHIIregionsarelargerandbrighter
andthusappearasclumpswhichdominatethemorphologyofthegalaxy.Asimpletheoretical
argument based on gas collapsing on scales of the Jeans mass in a marginally unstable disc
showsthattheclumpymorphologiesofhigh-zgalaxiesaredrivenbythecompetingeffectsof
higher gas fractions causing perturbations on larger scales, partially compensated by higher
epicyclicfrequencieswhichstabilizethedisc.
Keywords: gravitationallensing:strong–galaxies:high-redshift–galaxies:starformation.
inwhichthegasisaccretedsmoothlyalongfilaments.Thesecold
1 INTRODUCTION
flowsarelessdisruptivethanamajormerger,andhenceofferaroute
Observationsofstar-forminggalaxiesathigh-zhaveshownthata tomaintainmarginallystablediscs(ToomreparameterQ∼1)with-
significantfractionofthepopulationhasturbulent,clumpy,rotating out disrupting the structure and dynamics. Cold-flow accretion is
discswithclumpmassesof∼108−9M(cid:3),afactorof∼100×thetypi- expected to be a dominant mode of mass assembly above z (cid:4) 1,
calgiantmolecularcloud(GMC)locally(e.g.Cowie,Hu&Songaila andthusaccountsfortheubiquityoflargeclumpsathighredshift
1995;Elmegreenetal.2004,2009;Elmegreen&Elmegreen2005; (e.g.Bournaud&Elmegreen2009;Dekel,Sari&Ceverino2009;
Fo¨rsterSchreiberetal.2009).Theclumpsarethoughttoformfrom Bournaudetal.2011).
gravitationalinstabilitiesingas-richdiscs(Elmegreenetal.2007, Inthispicture,theclumpsareconsideredtobetransientfeatures,
2009;Genzeletal.2008;Bournaudetal.2010). forminginmarginallyunstablediscsathigh-zandfedbysmoothac-
Somerecentnumericalsimulationshavesuggestedthatthema- cretionofgasontothegalaxy.Clumpygalaxiesthereforerepresent
jorityofmassive,high-zgalaxiesaccretetheirgasvia‘coldflows’, aphaseintheevolutionofpresent-dayspiraldiscs.
Thereisaneedtotesttheinternalphysicalpropertiesofthein-
terstellarmedium(ISM)observationallytodeterminewhetherthe
(cid:2)
E-mail:[email protected] clumpsarescaled-upanaloguesoflocalHIIregionsorrepresenta
(cid:5)C 2012TheAuthors
MonthlyNoticesoftheRoyalAstronomicalSociety(cid:5)C 2012RAS
Evolution in the properties of giant HII regions 689
different‘mode’ofstarformation,andwhethertheycanexplainthe Gemini/Near-Infrared Integral Field Spectrometer (NIFS;
strongevolutionofstarformationrate(SFR)densitywithredshift. Swinbank et al. 2009) and Very Large Telescope
However,sufficientspatialresolutionisrequiredtoresolvetheISM (VLT)/Spectrograph for Integral Field Observations in the Near
onthescalesofstar-formingregions.Evenwiththeuseofadaptive Infrared (SINFONI; Fo¨rster Schreiber et al. 2009). These allow
optics, spatially resolved studies of high-redshift galaxies to date detailed mapping of the nebular emission lines, but at lower
have been limited to a resolution of ∼1.5kpc (e.g. Genzel et al. sensitivity than is achievable with imaging. An alternative means
2006;Fo¨rsterSchreiberetal.2009);usingtheHubbleSpaceTele- ofidentifyingstar-formingregionswithhighsensitivityistotake
scope(HST),onlythelargeststarburstcomplexescanberesolved, imaging through narrow-band filters. The Wide Field Camera 3
on scales of ∼1kpc (Elmegreen et al. 2007). On these scales, it (WFC3) on the HST presents an opportunity to study the star
ispossibletoprobethedynamicsofgalaxiesonlargescales,and formationingalaxiesatz∼1and∼1.5,astherearenarrow-band
Genzel et al. (2011) found evidence that Q < 1 in the regions of filters available which correspond to the wavelength of the Hα
galaxieswhereclumpsarefound,lendingobservationalsupportto emissionlineattheseredshifts.Combiningthesensitivityandhigh
the theory that the clumps form from internal gravitational insta- spatial resolution of HST/WFC3 with the magnification afforded
bilities.Inordertostudytheclumpsindetail,weneedtoresolve by gravitational lensing by foreground clusters, we can map the
high-redshiftdiscsonthescalesofindividualstar-formingregions; internal star formation distribution and so identify the frequency
inthelocaluniverse,thisis∼100pc. andpropertiesofgiantHIIregions.
The required spatial resolution can currently only be achieved Inthispaper,wethereforestudythestarformationmorphologies
by exploiting gravitational lensing. By targeting galaxies that lie ofeightgalaxiesatz∼1–1.5.WepresentthesampleinSection2,
behindforegroundclusterlenses,itispossibletobenefitfromlin- presentthepropertiesofthegalaxiesandtheirstar-formingclumps
earmagnificationfactors(alongonedirection)ofupto50×(e.g. inSection3,discusstheimplicationsinSection4andpresentour
Swinbanketal.2007,2009;Jonesetal.2010),andtoisolateHII conclusions in Section 5. Throughout, we adopt a (cid:5) cold dark
regionsoforder∼100pcouttoz∼5(Swinbanketal.2009).Re- matter ((cid:5)CDM) cosmology with H0 = 70kms−1Mpc−1, (cid:6)(cid:5) =
gionswerefoundwithstarformationsurfacedensities(cid:4)SFR∼100× 0.7 and (cid:6)m = 0.3. SFRs are calculated from Hα luminosity LHα
higherthanthosefoundlocally(Swinbanketal.2009;Jonesetal. usingtheprescriptionofKennicutt(1998a)adjustedtoaChabrier
2010). These regions of dense star formation are comparable to (2003)initialmassfunction(IMF).
the most intensely star-forming interacting systems in the local
Universe(Bastianetal.2006),yetappeartobeubiquitousinnon-
interactinggalaxiesathighredshift. 2 SAMPLE AND OBSERVATIONS
Itisnotknownwhatdrivestheseregionsofintensestarforma-
Our sample comprises eight lensed galaxies, each with spectro-
tionathigh-z,althoughJonesetal.(2010)suggestacombinationof scopicallyconfirmedredshiftsintherange1<z<1.5suchthat
highergasdensity,increasedstarformationefficiencyandshorter the Hα emission line falls within the high-transmission region of
starformationtime-scales.Inaddition,theirdatagivetheappear-
thenarrow-bandfiltersonWFC3.Theassociatedclusterlensesare
anceofabimodaldistributionofHIIregionsurfacebrightnesses, massive systems from the X-ray selected Brightest Cluster Sam-
althoughthereisnoknownphysicalprocessthatmightdrivethis.
ple(BCS)andMassiveClusterSurvey(MACS)samples(Ebeling
In order to understand this result further, we require a sample at
etal.1998,2007,2010;Ebeling,Edge&Henry2001)withwell-
intermediateredshift(z∼1–1.5)withwhichwecanprobetheevo-
constrained mass models (see references in Table 1), so that the
lutionofstarformationdensitywithredshiftathighersensitivityso
effectsoflensingcanbeaccountedfor.
thatregionscomparabletothoseatz=0aredetectable.
ThepositionsandpropertiesofthesamplearegiveninTable1.
Previous work on high-z clumps has made use of integral Weobservedeachtargetinthenarrow-bandfiltercoveringHαfor
field units such as Keck/OH-Suppessing Infrared Integral Field
a typical exposure time of 6ks (2 orbits), using a 3- or 4-point
Spectrograph (OSIRIS; Jones et al. 2010; Wisnioski et al. 2012), linearditheringpatternof±5arcsecinbothdirectionstoimprove
Table1. Propertiesoftheredshift-selectedsample.Lensingmagnifiestheimagebyafactorμxatapositionangle(PA),withatransversemagnificationμy.
ThetotalmagnificationfactorμiscalculatedfromtheamplificationofHαflux,andtheresolutiongivenisthehighestachievablealongthemostmagnified
direction,calculatedfromtheFWHMofastarunderthesamelensingtransformationasthatappliedtothegalaxy,asdescribedinthetext.Allobservations
wereobtainedunderProgram12197(Cycle18,PI:Richard)unlessotherwisestated.
Targetcluster Arcposition z Hαflux Magnification Resolution Broad-band Narrow-band Lens
RA Dec. (intrinsic) μx×μy(PA) μ (pc) filter filter model
(J2000) (J2000) (10−18ergs−1cm−2) reference
Abell611 08:00:57.30 +36:03:37.0 0.908 30±5 10.4×2.7(1◦) 28±5 338 F125Wa F126N [1]
Abell2390 21:53:34.55 +17:42:02.4 0.912 39±6 5.5×2.3(73◦) 12.6±1.9 435 F125Wb F126Nb [2]
Abell773 09:17:58.80 +51:43:42.3 1.010 274±49 7.0×1.0(61◦) 7±1 336 F110W F132N [1]
F160Wc
MACSJ0947.2+7623 09:47:15.26 +76:23:02.9 1.012d 8.5±2.2 3.0×17.7(51◦) 53±14 172 F125W F132N [3]
Abell68 00:37:04.91 +09:10:21.0 1.017 119±11 3.0×1.7(41◦) 5.1±0.5 615 F110W F132N [4]
F160We
MACSJ0159.8−0849 01:59:04.68 -34:13:03.4 1.488e 7±1 10.7×3.0(111◦) 32±4 592 F160W F164N [3]
MACSJ1149.5+2223 11:49:35.30 +22:23:45.8 1.490f 11±2 4.5×3.5(140◦) 15±3 315 F160Wb F164N [5]
MACSJ1133.2+5008 11:33:14.31 +50:08:39.7 1.550 8±1 1.1×12.7(67◦) 14±2 68 F160W F167N [6]
aProgram12065-9;bProgram11678;cProgram11591;dEbelingetal.(2010);eEbelingetal.(inpreparation);fEbelingetal.(2007).
References:[1]Richardetal.(2010);[2]Pelloetal.(1991);[3]Richardetal.(inpreparation);[4]Richardetal.(2007);[5]Smithetal.(2009);[6]Sandetal.
(2005).
(cid:5)C 2012TheAuthors,MNRAS427,688–702
MonthlyNoticesoftheRoyalAstronomicalSociety(cid:5)C 2012RAS
690 R. C. Livermore et al.
the detection and removal of cosmic ray hits and bad pixels. At Wederivetotalmagnificationfactorsbycomparingthetotallu-
the same time, three of the targets (MACSJ0947, MACSJ0159 minositiesoftheimage-andsourceplaneHα excessimages.The
andMACSJ1133),whichdidnothaveWFC3datainthearchive, intrinsic Hα luminosities are in the range 0.45–15 × 1041ergs−1
were observed in the corresponding broad-band filter using the correspondingtoSFRsof0.4–12M(cid:3)yr−1.Theseareatthefaint
samesequenceofobservationsastheircorrespondingnarrow-band endoftheHαluminosityfunction(LF)forthisredshiftrange(see
data, for a total of 3ks (1 orbit). The narrow-band data and new Fig.3),andprobefaintergalaxiesthanthez∼2sampleofJones
broad-bandobservationswereobtainedinCycle18underProgram et al. (2010), which covers the range 2.5–32 × 1041ergs−1, al-
12197(PI:Richard),withtheexception ofAbell2390,forwhich though the two samples overlap in luminosity. Because of the
thebroad-bandandnarrow-banddataweretakeninCycle17under increased sensitivity provided by the lensing magnification, both
Program11678(PI:Rigby).Theremainingbroad-banddatawere of the lensed samples cover a lower range of intrinsic Hα lumi-
obtainedunderCycle17Program11591(PI:Kneib)orCycle18 nosities than the sample of Spectroscopic Imaging Survey in the
Program12065-9(PI:Postman)asindicatedinthenotestoTable1. Near-Infrared with SINFONI (SINS) galaxies studied by Fo¨rster
AlloftheWFC3datawerereducedusingtheMULTIDRIZZLEsoft- Schreiberetal.(2011b),whichwereselectedtohavebrightHαand
ware (Koekemoer et al. 2002) under PyRAF to perform a cosmic lieintherange28–43×1041ergs−1,makingthemrare,intensely
ray rejection, sky subtraction and drizzling on to an output pixel star-forminggalaxies.Thus,byharnessinggravitationallensingwe
scaleof0.05arcsec.Thenarrow-bandandbroad-bandimagesofthe areabletoprobethemore‘normal’star-formingpopulation.
sameclusterwerealignedusingthelocationof∼20brightstars.A Sincegravitationallensingcanpreferentiallyshearonedirection,
narrow-bandexcessimagewasconstructedbydirectpixel-to-pixel weestimatetheeffectivesourceplaneresolutionbyreconstructing
subtraction between the narrow-band and broad-band images, in- theimageofastarfromthefieldrepositionedtolieatthecentre
cludinganarbitraryscalingfactor.Wecalibratedthisscalingfactor ofthetarget.Themaximumlinearresolution,derivedfromthefull
bycheckingthatallbrightclustermembers,whicharefeatureless width at half-maximum (FWHM) of the reconstructed star in the
elliptical galaxies with no emission lines in the respective filters, directionofgreatestmagnification,is68–615pcwithamedianof
became consistent with the background in the excess image. For 360pc,sufficienttoresolvegiantHIIregions.
Abell773 and Abell68, the broad-band images available in the
archive did not directly overlap the Hα emission line, so an esti- 2.1 Comparisonsamples
mateofthebroad-bandcontinuumwasmadebylinearinterpolation
In order to interpret our high-z data, we exploit the Hα narrow-
betweentheadjacentF110WandF160Wfilters.
band imaging from the Spitzer Infrared Nearby Galaxies Survey
ThefluxcalibrationofeachimagewasverifiedusingTwoMicron
(SINGS; Kennicutt et al. 2003), which comprises Hα imaging of
All Sky Survey (2MASS) stars in the fields, and in all cases was
75galaxieswithcorrectedSFRsofupto11M(cid:3)yr−1.Weusethe
foundtoagreetowithin15percent,whichissufficientprecision
publiclyavailablecontinuum-subtractedHαnarrow-bandimaging
forourpurposes.
and restrict the sample to those with Hα detections with signal-
Colour HST images of the clusters are shown in Fig. 1, with
to-noiseratioof>5thathavenosignificantdefectsinthegalaxy
thecriticallinesattheredshiftofthetargetarcoverlaid.Weusethe
images(determinedbyvisualinspection).ThisrestrictstheSINGS
transformationbetweenimageandsourceplanemappingfromthe
sampleto41galaxieswithSFR>4×10−4M(cid:3)yr−1.
best-fitting cluster mass models (for details of the mass models,
Toensureafaircomparison,werebintheSINGSimagessothat
seereferencesinTable1)withLENSTOOL(Kneib1993;Julloetal.
theresolutioniscomparabletothehigh-zdataandthenthresholdto
2007)toreconstructtheimagesinthesourceplane,andshowthese
themediansurfacebrightnesslimitofthez∼1–1.5observations.It
in Fig. 2. In order to reconstruct the source plane morphology,
isworthnotingthatthresholdingtheimagesinthismannerexcludes
LENSTOOL uses the mapping between the image and source planes
10–50percentofthetotalstarformation.Thisshouldnotaffectthe
onacluster-by-clusterbasisandraytracesthegalaxyimage.The
comparisonbetweensampleswhichhavethesamesurfacebright-
lensingeffectistostretchthegalaxyimage–inmostcasesalong
ness limit, but may serve as an indication of the fraction of star
onedirection–andsothereconstructioncannot‘create’newHII
formationmissedinhigh-zobservations.
regions,butratherthelensinghasactedtoextendthem.Assurface
To provide a comparison to local galaxies which are more ac-
brightnessisconservedbylensing,wethenapplythisconservation
tivelystarforming,weusetheVisibleMulti-ObjectSpectrograph
toobtaintheintrinsicsourceplaneflux.Thetotalmagnificationis
(VIMOS) Hα imaging spectroscopy of Rodr´ıguez-Zaur´ın et al.
thensimplytheratiooftheimage-tosourceplaneflux.Toobtain
(2011),whichincludes38luminousinfraredgalaxies(LIRGs)and
theerrorsonthemagnification,weusethefamilyofbest-fittinglens
ultraluminousinfraredgalaxies(ULIRGs)atz<0.13withspatial
modelswhichadequatelydescribetheclusterpotential,derivedby
resolutionof130–1.2kpcandSFR(cid:2)25M(cid:3)yr−1.
sampling the posterior probability distribution of each parameter
Wealsocomparethez∼1–1.5sampletothez∼2lensedarcs
of the model (see Richard et al. 2010 for more details). For each
ofJonesetal.(2010),whichwereobservedwithKeck/OSIRIS.In
acceptable lens model, we reconstruct the arc and remeasure the
amplification.Wegivetheresultingmagnificationfactors,μ,and order to provide a fair comparison, we have constructed narrow-
bandimagesbysummingtheOSIRIScubesover100Åeitherside
associatederrorsinTable1.
oftheredshiftedHαemissionline,matchingthewidthoftheWFC3
In cases where the target is multiply imaged, the images were
narrow-band filters. The resulting images are then corrected for
reconstructedseparatelyandthenadjustedforsmalldifferencesin
lensing using the same image-to-source plane mapping as Jones
positionandorientationbeforebeingcombined.ForMACSJ0159,
etal.(2010)inordertoobtaintheintrinsicgalaxyproperties.
which consists of five images, only the first three were used due
to the high magnification gradients in the fourth and fifth images
2.2 Determinationofgalaxyproperties
resultinginhighdistortioninthesourceplanereconstructions.In
thecaseofAbell611,weuseonlythenorthernmostarcduetohigh The total Hα luminosities of the galaxies in all samples are de-
distortionbyaforegroundgalaxylyingclosetothelineofsightof termined by summing all pixels in sky-subtracted images with
thesouthernarc. signal-to-noise ratio of >3. In the case of the SINGS galaxies,
(cid:5)C 2012TheAuthors,MNRAS427,688–702
MonthlyNoticesoftheRoyalAstronomicalSociety(cid:5)C 2012RAS
Evolution in the properties of giant HII regions 691
Figure1. HST/ACSandWFC3three-colourimagesoftheobservedclusterswiththecriticallineattheredshiftofthetargetarcoverlaid,showingthepositions
ofthetargetarcs.ThearcsarecontainedwithinthewhitedashedboxeswhichdenotetheregionsextractedinFig.2.
eachimagewascheckedbyvisualinspectionandanyforeground WeconvertHα luminositytoSFRusingtheKennicutt(1998a)
sourcesanddefectsmasked.Theresultingluminositieswerethen prescription, corrected to a Chabrier (2003) IMF, which reduces
compared to the published values and found to agree to within the SFR by a factor of 1.7×. As we do not have constraints
∼20percent. on the dust extinction, we adopt an estimate of AHα = 1 in all
(cid:5)C 2012TheAuthors,MNRAS427,688–702
MonthlyNoticesoftheRoyalAstronomicalSociety(cid:5)C 2012RAS
692 R. C. Livermore et al.
Figure2. Hαexcessimagesintheimageplane(left)andreconstructedinthesourceplane(right).Theimagescalesareinarcsecintheimageplaneandin
kpcinthesourceplane.Identifiedclumpsareindicatedinthesourceplaneimagesbyblackcrosses,andthemagentaellipseshowstheFWHMoftheeffective
sourceplanePSF,asdescribedinthetext.
samples.Thisassumptioniswidelyusedintheliteraturealthough increase by factors of 1.3–1.8×, and the z ∼ 1 galaxies Abell68
it is the subject of some disagreement. Garn et al. (2010) sug- andAbell773;theformerwouldbeafactorof1.8×lower,while
gestaluminosity-dependentAHα ismoreappropriate;wereweto thelatterwouldbeunchanged.Qualitatively,thereisnosignificant
adopttheirrelation,wewouldobtainAHα=0.7–1.6withamedian impactonourresults,asadoptingeitherluminosity-dependentex-
AHα=1.15.However,wealsonotethatrecentworkbyDom´ınguez tinctionrelationwouldservetoincreasetheevolutionweobservein
et al. (2012) suggests that galaxies with LHα (cid:2) 4×1041ergs−1 Section3.2.2.Forsimplicityandreproducibility,weadoptAHα=1
maybeconsistentwithhavingAHα=0,andthatabovethisthresh- throughout.
old extinction increases in a luminosity-dependent way. Had we Wedefinethesizesofthegalaxiesastwicethehalf-lightradius.
adopted this correction instead, the SFRs of the majority of our Thehalf-lightradiusisdeterminedusingthecontinuumimagesto
galaxies would be reduced by a factor of 2.5×. The exceptions findtheshape(i.e.thecentreandmajortominoraxisratioofanel-
are the three brightest z ∼ 2 galaxies, in which the SFRs would lipsethatbestfitsthegalaxy),andthenadjustingthesemimajoraxis
(cid:5)C 2012TheAuthors,MNRAS427,688–702
MonthlyNoticesoftheRoyalAstronomicalSociety(cid:5)C 2012RAS
Evolution in the properties of giant HII regions 693
–i.e.thepointsatwhichthelinestendtozero–wecanseethatthe
samplesaredifferent,withthez∼2galaxieshavingpeaksurface
brightnessesofaroundanorderofmagnitudehigherthanthelower
zsamples.Similarly,thez∼1–1.5sampleissystematicallybrighter
thantheSINGSsample,withtheexceptionofMACSJ1133,which
issimilartothefainterz=0galaxies,MACSJ0947whichissimilar
to the median of the z = 0 sample, and Abell773 which appears
similartothez∼2galaxies.
Asastatisticalmeasureoftheclumpinessofgalaxies,weinves-
tigateusingtheGinicoefficient,G,whichisusedineconomicsto
measuretheinequalityofwealthinapopulation(Gini1912).Ithas
valuesfrom0to1,whereattheextremesG=0foracompletely
uniformdistribution,andG=1ifthereisonlyonenon-zerovalue.
FollowingFo¨rsterSchreiberetal.(2011a),weuseittoquantifythe
distributionof fluxin animage, soavalue close toone indicates
thattheprofilehasasinglepeak(inthecaseofG=1,allofthe
flux would be in a single pixel), a galaxy with multiple clumps
wouldhavealowerG,andattheextreme,agalaxywithcompletely
uniformsurfacebrightnesswouldhaveG=0.
Inthez∼1–1.5sample,wefindanarrowrangeof0.25≤G≤
0.39withamedianofG=0.34.Thez∼2sampleismarginally
higher, with 0.42 ≤ G ≤ 0.56 and a median of G = 0.43. The
z=0SINGSsamplehasasimilarmedianG=0.45butamuch
wider range of 0.05 ≤ G ≤ 0.82, and the low-z (U)LIRGs have
0.38≤G≤0.85withthehighestmedianG=0.70.Onthebasis
Figure3. IntrinsicHαluminositiesofthehigh-zsamplescomparedtoHα of the Gini coefficient there are no clear differences between the
LFsfromHigh-zEmissionLineSurvey(HiZELS;Sobraletal.2012).Also samples.ComparingtheGinicoefficientswiththevisualappearance
shownistherangeofHαluminositiesoftheFo¨rsterSchreiberetal.(2011b) of the galaxies, the lack of distinction reflects the fact that a low
samplefromtheSINSsurveyatz∼2.Thetwolensedsamplesoverlapin Gini coefficient may arise from either a smooth distribution of
luminosityandarebothatthefaintendoftheLF,withthemedianofthez∼ star formation or from star formation that is concentrated into a
1–1.5WFC3samplelowerthanthatofthez=1.6−2.6OSIRISsampleby large number of distinct clumps. Furthermore, we find no strong
afactorof6.6×,whiletheunlensedSINSgalaxiescoverarangeofhigher
correlations between G and any of the properties of the galaxies.
Hαluminosities.
Clearly,toprogressfurtherwewillhavetocomparetheproperties
of individual clumps. In particular, we will show that the clump
oftheellipseuntilitencompasseshalfofthetotalHα luminosity LF provides a good means of distinguishing different galaxy star
calculatedinthemannerdescribedabove.Thegalaxy-averagedstar formationmorphologies.
formationsurfacedensity,(cid:4) ,isdefinedfromthetotalluminosity
SFR
enclosedwithintwohalf-lightradiiperunitarea.
3.2 Propertiesofstar-formingclumps
3 RESULTS AND ANALYSIS 3.2.1 Definitionofclumps
StudiesofHIIregionsorstar-formingclumpshaveusedavariety
3.1 Thespatialdistributionofstarformation
of methods to define and separate clumps from the background
Acommonthemeintherecentliteratureisthathigh-redshiftgalax- emissionofthegalaxy.Usuallyanisophoteisdefinedat3σ above
iesare‘clumpier’thangalaxiesinthelocalUniverse.Thisconcept thebackgroundnoise(e.g.GonzalezDelgado&Perez1997;Jones
originatesfromthefrequentappearanceof‘chain’galaxiesinthe etal.2010).However,thismethodisclearlydependentonthenoise
high-redshift universe (e.g. Cowie et al. 1995; Elmegreen et al. propertiesoftheimage,andthusisproblematicwhencomparing
2004;Elmegreen&Elmegreen2005).Evenwithoutlookingatthe localandhigh-redshiftobservations.Inparticular,ashigh-redshift
properties of individual star-forming regions, it is interesting to galaxyimagestendtohavehighrelativenoiselevelsandlowdy-
compare the morphologies of the star-forming regions across the namicrange,thechoiceofisophotetendstoselectonlythebright-
samples. estregionsinthegalaxy,neglectinganylower surfacebrightness
Fromvisualinspection,itisclearthattherearesignificantdif- clumpsandunderestimatingtheirsizes.
ferencesbetweenthesamples.Inparticular,thesurfacebrightness An alternative is the IRAF task DAOFIND as employed by Fo¨rster
distributionsofthegalaxiesshowdistinctdifferencesinthediffer- Schreiberetal.(2011b),whichisdesignedtolocatepointsources
ent samples. In Fig. 4, we show the fraction of star formation in inimages.However,wefoundthatitdidnotperformwellonour
pixelsaboveagiven(cid:4)SFRforthez∼1–1.5andz∼2samples,with sample.ThisislikelytobebecauseDAOFINDrequiresanexpected
theinterquartilerangeofthethresholdedSINGSsampleshownfor size of features to look for. As the clumps of Fo¨rster Schreiber
comparison. etal.(2011b)arelargelyunresolved,theywereabletousethepoint
Toallowforthedifferingsurfacebrightnesslimitsofthesamples, spreadfunction(PSF)oftheirobservationsastheexpectedsize.As
weonlyshowstarformationaboveasurfacebrightnessof(cid:4) = ourclumpsareresolved,theroutinedoesnotworkreliably.Forthis
SFR
0.001M(cid:3)yr−1kpc−2.Thisenablesustocomparethestarformation paper,wethereforeusethe2DversionofCLUMPFIND(Williams,de
occurringinbrightregionsinaconsistentmanner.Fromthepeaks Geus&Blitz1994),whichusesmultipleisophotestodefineclumps.
(cid:5)C 2012TheAuthors,MNRAS427,688–702
MonthlyNoticesoftheRoyalAstronomicalSociety(cid:5)C 2012RAS
694 R. C. Livermore et al.
Figure4. Thefractionofstarformationwithineachgalaxyoccurringaboveagivensurfacebrightness,forthez∼1–1.5andz∼2samples.Theshaded
regionistheinterquartilerangeoftheSINGSz∼0sample.Therearetwogalaxies,MACSJ1133andMACSJ0947,fromthez∼1–1.5samplewithsimilar
surfacebrightnessestothez=0sample,andtheremainderaresystematicallybrighter.Thez∼2samplehassignificantlyhighersurfacebrightnesses.Hence,
thereisclearevolutioninthesurfacebrightnessesofgalaxieswithredshift.
Wedefinedthecontourlevelswithrespecttothermsnoiseinthe Becauseofthemannerinwhichclumpsare‘grown’,theirsizes
image,startingat3σ andincreasingin1σ intervalsuntilthepeak returned by CLUMPFIND tend to be larger than those obtained by
value of the image is reached. The data are first contoured at the othermethods.Asacomparison,wealsofita2DellipticalGaus-
highestleveltolocateclumps,andthealgorithmthenworksdown sianprofiletoeachpeakandmeasuretheFWHM.Acomparison
inbrightnessthroughthecontourlevels.Anyisolatedcontoursare of the clump radii found by the two methods is shown in Fig. 5.
defined as new clumps, while others extend existing clumps. If The rms difference between the two radii is ∼100pc, and on av-
a contour surrounds one existing peak, they are allocated to that eragewefindthatCLUMPFINDoutputssizes25percenthigherthan
clump, and any which enclose two or more are divided using a theFWHM.Wisnioskietal.(2012)notethatsizesdefinedthrough
‘friends-of-friends’algorithm.Theadvantagesofthisapproachare isophotes can be unreliable due to the level of ‘tuning’ required
thatitenablesaconsistentclumpdefinitiontobeappliedtomultiple to select an appropriate isophote level in a given galaxy. This is
datasets,lowersurfacebrightnessclumpsarenotexcludedandthere lesssignificantwithCLUMPFINDbecausethistuningisnotrequired;
isnoassumptionmadeabouttheclumpprofile. the use of multiple isophote levels in all galaxies allows the lev-
The clumps identified by CLUMPFIND were all confirmed by vi- els to be defined in a consistent way across a large sample. We
sualinspectiontoremoveanysourcenotassociatedwiththetarget thereforefindmuchlowerscatterbetweentheisophotalsizesout-
galaxy,ofparticularimportanceinthecaseoftheSINGSimages put by CLUMPFIND and the clump FWHM than they do in their
whereforegroundsourceslieclosetooroverlapthetargetgalax- sample. Throughout this work, we use the CLUMPFIND size for all
ies.TheareaAoftheclumpisthenobtainedfromthenumberof samples, and give error bars that encompass the FWHM of the
pixelsassignedtoit,multipliedbythesourc√eplanepixelscale,and clumps.
fromthiswedefinetheeffectiveradiusr = A/π.Weonlyaccept ThesizesandHα-derivedSFRsofthez∼1–1.5clumpsaregiven
clumpswhere2rislargerthantheFWHMofthePSF,soallclumps inTable2.Weanalysethesepropertiesincomparisontotheother
areresolved. samplesbelow.
(cid:5)C 2012TheAuthors,MNRAS427,688–702
MonthlyNoticesoftheRoyalAstronomicalSociety(cid:5)C 2012RAS
Evolution in the properties of giant HII regions 695
Table2. Propertiesofclumpsidentifiedinthez∼1–1.5sample,
determinedasdescribedinSection3.2.1.
Clump Radius(pc) SFR(M(cid:3)yr−1)
MACSJ0947-1 350±56 0.054±0.010
MACSJ0947-2 324±22 0.0328±0.0086
MACSJ0947-3 384±48 0.045±0.012
MACSJ0947-4 334±39 0.0350±0.0092
MACSJ0947-5 318±38 0.0339±0.0089
MACSJ0947-6 376±6 0.0347±0.0091
MACSJ0947-7 311±17 0.0261±0.0068
MACSJ0947-8 149±9 0.0050±0.0013
MACSJ0159-1 402±89 0.282±0.060
MACSJ0159-2 370±77 0.203±0.043
MACSJ0159-3 530±130 0.355±0.076
MACSJ0159-4 468±13 0.170±0.036
Abell611-1 730±180 0.370±0.083
Abell611-2 560±160 0.181±0.041
Abell611-3 630±140 0.200±0.045
Abell611-4 390±59 0.065±0.015
Abell68-1 378±26 0.081±0.016
Abell68-2 132±8 0.0075±0.0015
Abell68-3 375±24 0.076±0.015
Figure5. ComparisonoftheclumpsizerclumpfindoutputbyCLUMPFINDwith Abell68-4 354±25 0.070±0.014
thesizerFWHMobtainedbytakingtheFWHMofa2DGaussianprofilefit. Abell68-5 509±31 0.114±0.023
Onaverage,CLUMPFINDoutputssizes25percentlargerthantheFWHM.For Abell68-6 337±33 0.062±0.013
consistency,weadopttheisophotalsizeoutputbyrclumpfindinallsamples. Abell68-7 386±61 0.099±0.020
Abell68-8 205±42 0.0273±0.0055
Abell68-9 348±44 0.069±0.014
3.2.2 Clumpproperties Abell68-10 299±89 0.066±0.013
Abell68-11 328±31 0.057±0.012
Onewayofquantifyingthe‘clumpiness’ofagalaxyistoconsider
Abell68-12 312±15 0.0476±0.0097
the fraction of a galaxy’s total Hα luminosity contained within Abell68-13 412±60 0.100±0.020
clumps.Wefindmediansof31percentinSINGS,36percentfor Abell68-14 293±16 0.0429±0.0087
the z < 0.13 ULIRGs, 50 per cent for the z ∼ 1–1.5 sample and Abell68-15 263±5 0.0302±0.0061
68percentforthez∼2sample.Thus,asexpected,thehigherz Abell68-16 280±50 0.0454±0.0092
galaxiesareclumpierthantheirlocalcounterparts. Abell68-17 169±7 0.0132±0.0027
Abell68-18 195±33 0.0215±0.0044
We now consider the properties of the clumps themselves, and
Abell68-19 224±6 0.0199±0.0040
firstcomparetheHα-derivedSFRtotheclumpradius,asshownin Abell68-20 239±51 0.0344±0.0070
Fig.6.Locally,thereisawell-definedrelationshipbetweenthese Abell68-21 135±5 0.0069±0.0014
properties, as found by Kennicutt (1988) who found almost con- Abell68-22 163±23 0.0142±0.0029
stantsurfacebrightnessinlocalHIIregions,exceptinmergingand Abell68-23 112±2 0.00489±0.00099
interacting systems (Bastian et al. 2006). The situation at high-z, Abell68-24 171±31 0.0161±0.0033
Abell68-25 98±5 0.00416±0.00085
though,appearsdifferent;Swinbanketal.(2009)andJonesetal.
Abell68-26 122±8 0.0066±0.0013
(2010)foundclumpswithSFRsof∼100×higheratagivensize Abell2390-1 352±59 0.115±0.025
thanfoundlocally,insystemswithnoevidenceofinteractions. Abell2390-2 404±68 0.136±0.030
Fig. 6 is an updated version of one presented in Jones et al. Abell2390-3 409±6 0.086±0.019
(2010),wherewehavere-analysedthez∼2andSINGSgalaxies Abell2390-4 366±26 0.067±0.015
usingCLUMPFINDsothatclumpsaredefinedconsistentlyacrossall Abell2390-5 463±9 0.111±0.024
samples, and we have added the results from our new z ∼ 1–1.5 Abell2390-6 347±13 0.059±0.013
data set and the z < 0.13 ULIRGs as well as the z = 1–2 results Abell2390-7 470±1 0.093±0.020
Abell2390-8 341±73 0.069±0.015
from SHiZELS (Swinbank et al. 2012) and WiggleZ (Wisnioski Abell773-1 1040±200 5.6±1.3
et al. 2012). We show lines of median surface brightness in the Abell773-2 1430±180 9.6±2.2
samples,andverticaloffsetsfromtheselinesrepresentdifferences MACSJ1133-1 1120±100 0.118±0.025
in the surface density of star formation, (cid:4) , in the clumps. We MACSJ1133-2 890±160 0.068±0.015
SFR
willexploretherelationoftheseoffsetstoglobalgalaxyproperties MACSJ1133-3 790±280 0.057±0.012
MACSJ1133-4 835±4 0.0334±0.0072
inSection4.
MACSJ1149-1 174±34 0.084±0.020
WenotethattheclumpsweidentifyintheSINGSgalaxiesare
derivedfromimageswhichhavebeendegradedtocomparablereso-
lutiontothehigh-zdata,andwefindtheeffectofthisistodecrease
the surface brightness by a factor of ∼2×, as the size increases from the local relation by a factor of ∼100×. This confirms
morethantheluminosity.ThepointsinFig.6movealongthevector the large differences between the local and high-redshift popu-
labelled‘A’.Definingclumpsinthez=0sampleinthiswayensures lation already noted by Swinbank et al. (2009) and Jones et al.
thefairestpossiblecomparisonwiththehigh-zdata. (2010).
Uponre-analysisusingCLUMPFIND,wefindsomelower(cid:4)SFRre- Ournewz∼1–1.5samplefitsinbetweentheSINGSandz∼2
gionsintheJonesetal.(2010)sample,buttheyallremainseparated samples, with the exception of the two regions from the most
(cid:5)C 2012TheAuthors,MNRAS427,688–702
MonthlyNoticesoftheRoyalAstronomicalSociety(cid:5)C 2012RAS
696 R. C. Livermore et al.
Figure6. HαSFRforextractedHIIregionsasafunctionofsize,comparedtothelensedz∼2sampleofJonesetal.(2010),high-zunlensedsamplesfrom
SHiZELS(Swinbanketal.2012)andWiggleZ(Wisnioskietal.2012),low-z(U)LIRGsfromRodr´ıguez-Zaur´ınetal.(2011)andthez=0SINGSgalaxies
(Kennicuttetal.2003).SFRsarecalculatedusingtheKennicutt(1998a)prescriptionadjustedforaChabrierIMFwithadustextinctionAHα=1inallsamples,
andtheerrorbarsofthehigh-zlensedsourcesaredominatedbytheuncertaintyinthelensingmagnification.Dashedlinesshowthemediansurfacebrightnesses
intheSINGS,z∼1–1.5andz∼2samples.Theblackdottedlineindicatesthesensitivitylimitofthez∼2OSIRISobservations.Thearrowindicatesthe
effectofdegradingtheimageresolution,asdiscussedinthetext.Thefourlowestsurfacebrightnessclumpsinthez∼1–1.5samplecomefromonegalaxy
(MACSJ1133),andthetwobrightestregionsarefromAbell773,themostcompactgalaxyinthesample.Theremaininggalaxieshaveclumpswithsurface
brightnessesinbetweenthoseofthez=0andz∼2samples,similartolocal(U)LIRGs.
compact source Abell773, which have (cid:4) similar to the z ∼ a much tighter sequence with all the clumps sharing a common
SFR
2sample,andthefourregionsfromMACS1133,whicharesimilar surfacebrightness,particularlyinthelow-redshiftsample.Thusthe
to z = 0 clumps. This indicates clear evolution in clump surface spreadinclumppropertiesinFig.6appearstobedrivenbyglobal
brightness,(cid:4) ,withredshift. differencesinthegalaxies.
SFR
The surface brightness limit of the z ∼ 2 data means that we Wethereforenextcomparetheclump(cid:4) tothepropertiesof
SFR
cannot identify the low SFR clumps in that sample. We show a theirhostgalaxiesinFig.7.Intheleft-handpanel,wecorrelatethe
dottedlinerepresentingthelowerlimitatwhichwedefineclumpsin clumppropertieswiththetotalSFRofthegalaxy.Forclarity,we
thez∼2galaxies.Itislikelythatthereareadditionalclumpswhich plotthemedianclump(cid:4) ineachindividualgalaxy,andtheerror
SFR
liebelowthislimitandareundetected;however,suchclumpsmake barsencompassthecentral68percentofclumpswithineachgalaxy
only a small contribution to the total SFR, as we shall discuss in (i.e.1σiftheyfollowaGaussiandistribution).Thereisevidencefor
Section3.3. correlationbetweentheclump(cid:4) andthegalaxyHαluminosity
SFR
Selection effects have no impact on the lack of high surface (whichweassumetobeproportionaltothetotalSFR);wefinda
brightnessregionsinthelowerredshiftsamples,however.Thein- Spearmanrankcorrelationcoefficientρ=0.69,representinga5.8σ
tensestar-formingregionsareclearlymorecommoninhigh-zgalax- deviationfromthenullhypothesisofnocorrelation.Thissuggests
ies;theyarefoundonlyinextremesystemssuchastheAntennae thatthestarformationinthehigh-zsamplefollowsasimilartrend
locally,butexistinallfiveofthez∼2galaxiesandoneoftheeight tothelocalsample,andthatthedifferencesseeninFig.6mayarise
z∼1–1.5sample. fromthehighertotalSFRsofthehigh-redshiftgalaxies.
As noted in Section 2, the z ∼ 2 galaxies have ‘normal’ SFRs Forthemajorityofthesamples,anevenstrongerrelationarises
fortheirredshift,belowthekneeoftheHαLF.Theoffsetsseenin ifwecomparetheclump(cid:4) tothegalaxy-averaged(cid:4) .This
SFR SFR
thefigureemphasizetheimportanceofanalysingclumpsinterms is shown in the right-hand panel of Fig. 7, and has a correlation
oftheirsurfacebrightness.Thisisevenmoreevidentiftheclumps coefficientρ=0.79with6.6σ significance.Theratioofclump-to-
belongingtoasinglegalaxyareexaminedseparately.Ratherthan average(cid:4) canbethoughtofasameasureofthe‘clumpiness’of
SFR
beingdistributedacrosstheplotatrandom,individualgalaxiesform thegalaxy.
(cid:5)C 2012TheAuthors,MNRAS427,688–702
MonthlyNoticesoftheRoyalAstronomicalSociety(cid:5)C 2012RAS
Evolution in the properties of giant HII regions 697
Figure7. Comparisonsbetweenthestarformationsurfacedensity(cid:4)SFR ofstar-formingclumpswithineachgalaxyandtheintrinsicHα luminosityand
galaxy-averaged(cid:4)SFR.Theclump(cid:4)SFRshownisthemedianforeachgalaxy,witherrorbarsencompassingthefullrangeof(cid:4)SFRforallclumpswithineach
galaxy.Thesolidlineisthebestfittothedata,andthedashedlineillustratestheclump(cid:4)SFRexpectedfromtheory,discussedinSection4.Wefindthatboth
arecorrelatedatthe5σlevel,implyingthatwefindmorehigh(cid:4)SFRclumpsathighredshiftbecausetherearemorehighSFRand(cid:4)SFRgalaxiesatthisepoch.
We conclude that the properties of star-forming clumps in a SFRofthehostgalaxies,andshowthisintheright-handpanel.The
galaxy are strongly dependent on the global (cid:4)SFR of the galaxy. similarityofalltheHIIregionLFsisnowclear.
Galaxieswithhigheroverall(cid:4) havehigherclumpsurfaceden- There is a striking similarity between the LF of the z ∼ 1–1.5
SFR
sitiesandarecorrespondinglyoffsetintheclumpsize–SFRrelation. sample and that of the highest SFR galaxies in the low-z SINGS
WhilethisaccountsforsomeofthedifferencesseeninFig.6,itis sample. The excess of very bright regions (L ∼ 1041ergs−1) is
alsoclearthattherearemorebrightclumpsinthehigherredshift downtoonegalaxy,Abell773,whichisthesamecompactgalaxy
galaxies.Wequantifythisbelow. for which we found the clump surface brightnesses to be more
typical of the highest redshift galaxies. The low-luminosity slope
oftheLFtendsalsotobeflatterthanthatseeninthelow-redshift
3.3 HIIregionluminosityfunctions galaxies,butitishardtoquantifythisdifferencewithoutdirectly
A quantitative measure of the clump brightness is to construct a comparable surface brightness limits and is likely to be affected
LFofHIIregions.InthelocalUniverse,theHIILFispresentedin byunresolvedregionswhichareexcluded.Inanycase,thesefaint
Kennicutt,Edgar&Hodge(1989)andGonzalezDelgado&Perez regionscontributelittletothetotalflux.
(1997). They demonstrate that the LF can be fitted by a broken Inbothpanels,theHIIregionLFforthehighestredshiftgalaxies
powerlaw,orbyapowerlawwithanexponentialbreak.Inorderto isstronglyoffsetfromtherelationseeninthelow-redshiftSINGS
beconsistentwithourdefinitionsofclumpsizes,were-analysethe sampleandfromthesampleatz∼1–1.5,butissimilartothelow-
localdatainordertoconstructourownLF.Theresultsareshown redshiftULIRGs.Althoughthedatadonotprobethelow-luminosity
in Fig. 8. The left-hand panel of the figure shows the cumulative slopeoftheLF,thesegalaxieshavemuchbrighterregionsthanare
numberofregionspergalaxyasafunctionofHαluminosity.The seenatlowerredshift.Theright-handpanelemphasizesthatthisis
normalizationofeachbintakesintoaccountthedifferentsurface notbecausetheycontainmanymoreregionsoverall.
brightnesslimitofthegalaxies,witherrorbarscomputedfromthe In order to compare our data to models of mass functions, we
Poissonerrorincountingregions.Theslopeofthepower-lawpartof must relate the measured Hα luminosities to model clump mass,
themassfunctionis∼−0.75,sothatalthoughtheLFappearssteep M. As an estimate, we use the Hα-derived SFR and adopt SFR
inthisrepresentation,mostofthetotalluminosityiscontributedby (M(cid:3)yr−1) = 4.6±2.6×10−8M(cid:3) (Lada, Lombardi & Alves
thebrightestHIIregions. 2010). This empirical relation is based on local molecular clouds
SolidblackpointsshowtheaverageofallSINGSgalaxies.How- andappliestothehigh-densitygaswhereAK >0.8mag.However,
ever,sincewewillbecomparingthegalaxiescoveringarangeof wenotethatthisrelationisconsistentwiththefar-infrared-derived
luminositiesandredshifts,wehaveseparatedthegalaxiesfromthe SFRandCO-derivedgasmassesofstar-formingclumpsreportedin
SINGSsampleintothreetotalHαluminositybins.Atafixedlumi- alensedz=2.3galaxybySwinbanketal.(2011),butclearlymore
nosity,galaxieswithlowertotalemissionhavefewerregions,but high-resolutionCOobservationsofhigh-zgalaxiesarerequiredto
theshapeoftheLFissimilar.Inordertoemphasizethesimilarity confirmthis.Asaguide,weincludethisconversionontheupper
ofthemassfunction,wenormalizeeachofthecurvesbythetotal axisofFig.8.
(cid:5)C 2012TheAuthors,MNRAS427,688–702
MonthlyNoticesoftheRoyalAstronomicalSociety(cid:5)C 2012RAS