Star formation in z>1 3CR host galaxies as seen by Herschel

4.1 MB·English

by  P. Podigachoski

Astronomy&Astrophysicsmanuscriptno.Herschel3CR (cid:13)cESO2015
January8,2015
Star formation in z > 1 3CR host galaxies as seen by Herschel∗
P.Podigachoski1,P.D.Barthel1,M.Haas2,C.Leipski3,B.Wilkes4,J.Kuraszkiewicz4,C.Westhues2,S.P.Willner4,
M.L.N.Ashby4,R.Chini2,D.L.Clements5,G.G.Fazio4,A.Labiano6,C.Lawrence7,K.Meisenheimer3,
R.F.Peletier1,R.Siebenmorgen8,andG.VerdoesKleijn1
1 KapteynAstronomicalInstitute,UniversityofGroningen,9747ADGroningen,TheNetherlands
e-mail:[email protected]
2 AstronomischesInstitut,RuhrUniversität,D-44801Bochum,Germany
3 Max-PlanckInstitutfürAstronomie(MPIA),D-69117Heidelberg,Germany
5
4 Harvard-SmithsonianCenterforAstrophysics,Cambridge,MA02138,USA
1
5 AstrophysicsGroup,ImperialCollegeLondon,BlackettLaboratory,PrinceConsortRoad,LondonSW72AZ,UK
0
6 InstituteforAstronomy,DepartmentofPhysics,ETHZurich,Wolfgang-Pauli-Strasse27,CH-8093Zurich,Switzerland
2
7 JetPropulsionLaboratory,CaliforniaInstituteofTechnology,4800OakGroveDrive,PasadenaCA91109,USA

n 8 EuropeanSouthernObservatory,Karl-Schwarzschild-Str.2,85748Garchingb.München,Germany
a
J Received;accepted

7
ABSTRACT
]
A
WepresentHerschel(PACSandSPIRE)far-infrared(FIR)photometryofacompletesampleofz>13CRsources,fromtheHerschel
G GTprojectTheHerschelLegacyofdistantradio-loudAGN(PI:Barthel).CombiningthesewithexistingSpitzerphotometricdata,we
performaninfrared(IR)spectralenergydistribution(SED)analysisoftheselandmarkobjectsinextragalacticresearchtostudythe
.
h starformationinthehostsofsomeofthebrightestactivegalacticnuclei(AGN)knownatanyepoch.Accountingforthecontribution
p froman AGN-powered warmdust component totheIR SED,about 40%of our objects undergo episodes of prodigious, ULIRG-
- strength starformation, withratesof hundreds of solar massesper year, coeval withthegrowthof thecentral supermassive black
o
hole. Median SEDs imply that the quasar and radio galaxy hosts have similar FIR properties, in agreement with the orientation-
r
t basedunificationforradio-loudAGN.Thestar-formingpropertiesoftheAGNhostsaresimilartothoseofthegeneralpopulationof
s equallymassivenon-AGNgalaxiesatcomparableredshifts,thusthereisnostrongevidenceofuniversalquenchingofstarformation
a
(negativefeedback)withinthissample.Massivegalaxiesathighredshiftmaybeformingstarsprodigiously, regardlessofwhether
[
theirsupermassiveblackholesareaccretingornot.

1
Keywords. galaxies:active–galaxies:high-redshift–galaxies:starformation–infrared:galaxies
v
4
3
41. Introduction the duration and frequency of AGN accretion and host galaxy
1 starformationphases.
0The understanding that most (if not all) galaxies in the Uni- Highredshiftradio-loudAGN(P >1025 WHz−1 and
1.verse host a supermassive black hole (SMBH) is among the z > 1) provide a unique opportun1it.4yGtHozprobe the interplay
most important findings of modern astronomy. The growth of
0 between the growth of the black hole and the hosting stel-
aSMBHthroughmassaccretiongenerateslargeamountsofen-
5 lar bulge. They are invariably associated with massive galax-
ergyduringaphasein theevolutionofthegalaxyknownasan
1 ies having M & 1011 M (Bestetal. 1998; Seymouretal.
active galactic nucleus (AGN) phase. Although there is a dif- stellar ⊙
: 2007;DeBreucketal.2010),andhaveedge-brightened,double-
vference of a factor of ∼ 109 in their physical size scales, the
i lobed, FRII morphologies (Fanaroff&Riley 1974) that permit
XSMBHs and their host galaxies exhibit strong scaling relations estimatesofthedurationoftheepisodeofstrongAGNactivity.
(e.g.Magorrianetal.1998;Tremaineetal.2002;Gültekinetal.
r In addition to being used in studies of massive galaxy evolu-
a2009),suggestingalinkbetweenthegrowthoftheSMBHsand
tion, radio-loud AGN are being used extensively in unification
that of their host galaxies. Moreover, both these processes are
studies, where, distant radio-loud galaxies and quasars are be-
thoughttopeakatredshiftsz∼2(e.g.Hopkins&Beacom2006;
lieved to make up one and the same population(Barthel 1989;
Alexanderetal. 2008). The symbiosisofblack holeand global
Antonucci 1993; Urry&Padovani 1995), hence have equally
galaxygrowthisintriguingbecauseofthepossiblefeedbackef-
massivehosts.Ultravioletorvisibleopaquecircumnucleardust
fects: positive (AGN inducing star formation) and/or negative
isanessentialelementofthisscenario;distant3CRquasarsand
(AGN quenching of star formation).These feedback processes
radio galaxies are indeed luminous mid-infrared (MIR) emit-
are of paramount importance for our understanding of galaxy
ters (Siebenmorgenetal. 2004; Haasetal. 2008; Leipskietal.
formation (e.g. Crotonetal. 2006; Hopkinsetal. 2008). How-
2010).
ever,neitherthefeedbackmechanismsnortheoverallimpactof
It has long been suspected that hosts of powerful high-
feedbackonthehostgalaxiesisknown.Otherbigunknownsare
redshiftradio-loudAGNundergoepisodesofvigorous(dustob-
∗ HerschelisanESAspaceobservatorywithscienceinstrumentspro- scured) star formation (e.g. Archibaldetal. 2001). Huge reser-
videdbyEuropean-ledPrincipalInvestigatorconsortiaandwithimpor- voirs of molecular gas, have been deduced in several objects
tantparticipationfromNASA. fromsubmillimetre(submm)studies(e.g.Reulandetal. 2004).
Articlenumber,page1of30
A&Aproofs:manuscriptno.Herschel3CR
Such studies were mainly based on one submm flux measure-
mentandwerelimitedto thehighestredshiftobjectsforwhich
theobscurednewbornstarradiationre-emittedbytheubiquitous
cold(30-50K)dustisredshiftedtosubmmwavelengths.How-
ever,quantificationof the cold dustemission (e.g. constraining
thecolddusttemperature)requiressamplingthefullrest-frame
infrared-submmspectralenergydistribution(SED)of the stud-
ied objects. Earlier far-infrared (FIR) studies failed to provide
strongconstraintsontheFIRpropertiesforrelativelylargesam-
ples of powerfulradio-loudAGN because of their small detec-
tionfractionsandonlylimitedrest-frameFIRwavelengthcover-
age (Heckmanetal. 1992; Hesetal. 1995; Meisenheimeretal.
2001; Siebenmorgenetal. 2004; Haasetal. 2004; Clearyetal.
2007).
TheHerschelSpaceObservatory(Pilbrattetal.2010),with
itsunprecedentedFIRsensitivityandwavelengthcoverage,ex-
plored terra incognita (caelum incognitum...) allowing studies
which have revolutionized the understanding of the connec- Fig.1.Observedradio(178MHz)luminosityasafunctionofredshift
tion between AGN and star formation activity. Several stud- forthe3CRsampleconsideredinthiswork.Circles(red)indicateradio
ies utilizing deep X-ray and Herschel data revealed that the galaxies, and squares (blue) indicate quasars. Theplus symbols mark
hosts of moderately luminous radio-quiet AGN out to z ∼ 3 theobjectsdetectedinatleastthreeHerschelbands(typicallythetwo
form stars at rates comparable to the general non-AGN pop- PACSandtheSPIRE250µmbands.
ulation (Shaoetal. 2010; Mullaneyetal. 2012; Rosarioetal.
2012).Forhigh(radio-quiet)AGNluminosities(L > 1044 erg
X Spinradetal.1985).Theextremelyhighluminosities(Fig.1)of
s−1), Pageetal. (2012) reported suppression of star formation,
thesedouble-lobedradiogalaxies(RGs)andquasars(QSRs)are
consistentwith theexpectationsfromtheoreticalmodels,while
producedbysomeofthemostpowerfulaccretingSMBHs.The
Harrisonetal.(2012)foundnoclearevidenceofsuppressionof
low-frequency(178 MHz) radio selection ensures no bias with
starformationbyextendingtheanalysestosampleslargerbyan
respect to orientation: the steep-spectrum lobes of radio-loud
orderofmagnitude.Moreover,atthehighestAGNluminosities
AGN emit optically thin and isotropic synchrotron radiation,
(in excess of 1046 erg s−1), recent studies based on decompo-
making the 3CR sample ideal for testing the orientation-based
sition of the IR emission to AGN and star formation contribu-
unificationscenarioofradio-loudAGN.AsshowninFig.1,both
tions, have shown star formation rates (SFRs) of the order of
the RGs and QSRs are homogeneously distributed in redshift.
severalhundredsolarmassesperyearinthehostsofsomeofthe
Thez > 13CRobjectsshowmostlyFRIImorphologies;thisis
most powerful radio galaxies (Bartheletal. 2012, - B12 here-
well-establishedfrom,forinstance,high-resolutionVLAmaps.
after;Seymouretal.2012;Drouartetal.2014)and(radio-quiet)
Compact, presumably young, morphologies (O’Dea 1998) are
quasars(Leipskietal.2013,2014).
alsofoundwithinthesample.
In order to quantify the energetics of AGN at the peak of
The z > 1 3CR sample is (spectroscopically) completely
their activity as well as their star formation characteristics, we
identified using 3 to 5 m-class telescopes in the 1960s-1980s
obtainedfive-bandHerschelphotometryofthe3CRsampleus-
(Spinradetal. 1985). The objects in the sample almost univer-
ingthePhotodetectorArrayCamera(PACS)at70and160µm
sally accrete at high Eddington rate, i.e. in quasar-mode (e.g.
andtheSpectralandPhotometricImagingReceiver(SPIRE)at
Best&Heckman2012).Thetotalnumberofobjectsinthez>1
250,350,and500µmon-boardtheHerschelSpaceObservatory.
3CRsample is641. The highestredshift3CRsourceis 3C257
Thefirstresults,dealingwith3archetypalobjectsofthatsample
(z = 2.47). Two z > 1 3CR sources, 3C 287 and 3C 300.1,
werepresentedinB12.HereweanalysetheFIRpropertiesofthe
havebeenobservedinotherHerschelobservingmodes,andthus
complete(flux-limited)sample of objectsspanningthe redshift
havebeendroppedfromthiswork.Theremaining62sources,37
range1 < z< 2.5.Thispaperisorganizedasfollows.Section2
RGsand25QSRs,comprisetheHerschelsamplestudiedinthis
describesthe sample selection,the data obtained,and the steps
work.AnoverviewofselectedpropertiesisprovidedinTable1.
usedformeasuringfluxdensitiesinthefiveHerschelbands.Sec-
Thehigh-z3CRsamplehasbeenobservedwithmanyspace
tion3addressestheprocedureforfittingtheobservedIRSEDs
telescopes(includingHubble, Spitzer, and Chandra;Bestetal.
oftheobjects.ResultsarethenpresentedanddiscussedinSect.4
1998;Haasetal. 2008;Leipskietal. 2010;Wilkesetal. 2013):
andSect.5,respectively,andthepaperisbrieflysummarizedin
objectsfromthissamplerepresentlandmarksinourstudyofac-
Sect.6.ThroughoutthispaperweuseaflatcosmologywithH
0 tivegalaxiesthroughcosmictime.
=70kms−1Mpc−1andΩ =0.7,andwefollowtheconversion
Λ
inKennicutt(1998)(whichassumesaSalpeter1955initialmass
function)whenderivingSFRs. 2.2.Herschelphotometry
The data for this work were obtained as part of our Herschel
2. Data GuaranteedTimeprojectTheHerschelLegacyofdistantradio-
loud AGN (PI: Barthel, 38 hoursof observations).Five objects
2.1.Sampleselection (see Table 1) were observed as part of another Herschel pro-
gramme(PI:Seymour).Therawdatafortheseobjectswerere-
Withthisstudy,wetargetthewellknown,completeflux-limited
sampleofthebrightest(F178MHz >10Jy),high-redshift(z>1) 1 Thesampleincludestwo4Cobjects,4C13.66and4C16.49,which
radio-loudAGNsampleinthenorthernhemisphere:theRevised formallymatchthe selectioncriteriaof, and areincluded in, the 3CR
Third Cambridge Catalogue of radio sources (hereafter 3CR; sample.
Articlenumber,page2of30
P.Podigachoskietal.:Starformationinz>13CRhostgalaxiesasseenbyHerschel
trievedfromthe HerschelScienceArchive(HSA), andthe data 6-7mJybeam−1,asestimatedfromdeepextragalacticobserva-
reductionwasperformedasdetailedbelow. tions (Nguyenetal. 2010). Our adopted procedure for the de-
terminationofthephotometricuncertaintiesintheSPIREmaps
isfullydescribedbyLeipskietal.(2013),whichinturnfollows
2.2.1. PACS theprocedurespresentedbyElbazetal.(2011)andPascaleetal.
(2011).Initially,anartificialsourcefreemapwascreatedbyre-
Photometric observations were carried out with PACS
movingtheextractedsourcesinaSPIRE mapfromthemapit-
(Poglitschetal. 2010) in the scan-map observational mode,
self. Then, the pixel-to-pixel rms in this source free map was
both in the blue (70 µm, 5′′ angular resolution) and in the red
calculated using a box centred on the nominal position of the
(160 µm, 11′′ angular resolution) bands. A concatenated pair
targetobject. The size of the boxwas chosen as a compromise
of coextensive scan maps at two different orientations was
betweenavoidingthenoisyedgesoftheSPIREmapandobtain-
obtainedforeachsource.Datareductionwasperformedwithin
ingproperstatisticsoftheimmediateenvironmentofthetarget
the Herschel Interactive Processing Environment (HIPE, Ott
object.
2010, version 11.0.0), following the standard procedures for
As indicated in Table 2, three SPIRE detections are for-
deep field observations. Maps were created by employing the
mallybelowtheestimated3σvalues.Theseparticularmeasure-
high-passfilteringmethod,usinganappropriatesourcemasking
ments were included in the subsequentanalyses because, upon
steptoavoidsignificantfluxlossesduetothehigh-passfilter.A
visual inspection, they showed obvious emission at the known
first data reductionresulted in a preliminarymap, created after
positionofthetargetobject.Theavailabilityofancillarymulti-
combiningtheindividually(foreachscanorientation)processed
wavelengthdataatshorterwavelengths(tocheckforsourcecon-
scan maps. Source masking was performedby hand, using the
fusion), and the understanding of the overall shape of the ob-
preliminarycreated map as an input.This method in particular
ject’s SED, furthersupportthe inclusionof these flux densities
allowed us to minimize the flux losses of the observedsources
inthesubsequentanalyses.Whiletheformalsignal-to-noisera-
(Popessoetal. 2012). The final data reduction and mosaicking
tio of one of these three detections is very close to three, the
were then performedusing the mask generated in the previous
othertwodetectionsdonotreachthisratioonlybecausetheas-
step.
sociatedSPIREmapsarelesscleanthanothermapsinthesam-
Photometry(usingappropriateaperturecorrections)wasper-
ple,leadingtosignificantlylargerphotometricuncertainties.The
formedwithin HIPE, using the annularSkyAperturePhotometry
SPIRE 500 µm photometry should be considered tentative be-
task. Aperturesof 6′′ and 10′′ radiusfor PACS blue andPACS
causethebeamatthisparticularwavelengthislarge,andunde-
red, respectively, were centred on the known radio core posi-
tectedsourcesintheregionsurroundingtheAGNmaycontribute
tionoftheobjectinthemap.PACSmapssufferfromcorrelated
tothemeasuredfluxdensity.TheSPIREphotometricuncertain-
noise, thus pixel-to-pixel variations cannot yield robust photo-
tiesprovidedinTable2donotincludethe4%uncertaintyonthe
metric uncertainties. Instead, we opted for the well-established
absolutefluxcalibration(Bendoetal.2013).Postagestampsof
procedure of placing apertures at random positions on the sky
the resulting SPIRE maps, centred on the radio position of the
(Lutzetal. 2011; Popessoetal. 2012). Following Leipskietal.
objects,areincludedinAppendixE.
(2013), we placed 500 apertures of 6′′ (for blue) and 10′′ (for
red) radii at locationsavoiding the noisy edges of the map, re-
quiringthatthecentralpixeloftherandomaperturehasatleast 2.3.SupplementaryData
75%oftheintegrationtimeofthatofthesourceofinterest.The
The FIR photometry of all 3CR sources in our work was
resultingdistributionofthefluxdensitiesmeasuredinthese500
supplemented with MIR photometry obtained with the Spitzer
apertures was then fitted with a Gaussian, and the sigma value
Space Telescope (Werneretal. 2004) during three Spitzer GT
oftheGaussianwastakentobethe1σphotometricuncertainty
observing programmes (PI: G. Fazio) in six bands, using the
ofthemap.Measuredfluxdensitiesandassociated1σuncertain-
instruments IRAC (Fazioetal. 2004), IRS-16 peak-up array
ties,togetherwith3σupperlimitsforthenon-detectionsarepro-
(Houcketal. 2004), and MIPS (Riekeetal. 2004). Details on
videdinTable2.ThePACSphotometricuncertaintiesprovided
theSpitzerdatareductionandphotometryhavepreviouslybeen
inTable2donotincludethe5%uncertaintyontheabsoluteflux
published by Haasetal. (2008). Table 2 lists the Spitzer pho-
calibration (Balogetal. 2013). Postage stamps of the resulting
tometry.Whenavailable,additional850µmdatawerecollected
PACSmaps,centredontheradiopositionoftheobjects,arein-
fromtheliterature.The850µmemissioninquasarscanbeheav-
cludedinAppendixE.
ilycontaminatedbysynchrotroncontribution(seeSect.3.3).The
quasar850µmthermalfluxdensitiesutilizedinthisworkwere
2.2.2. SPIRE taken from Haasetal. (2006). To obtain the 850 µm thermal
fluxdensitiesofquasars,Haasetal.(2006)extrapolatedthesyn-
SPIRE (Griffinetal. 2010) photometric observationswere car- chrotroncontributionat 850 µm usingthe measuredradio core
ried out in small scan-map observational mode, at 250 (18.2′′ flux densities, and subtracted it from the total flux density at
angularresolution),350(24.9′′ angularresolution)and500µm 850µm.Table3liststheradiogalaxyandquasarthermalsubmm
(36.3′′ angular resolution). Data reduction was performed in fluxdensitiesusedinthiswork.
HIPE following standard procedures for SPIRE data. Source
extraction on the fully reduced map was performed using the
sourceExtractorSussextractortask(Savage&Oliver2007).Ex- 3. Spectralenergydistributions
tractedsourceslocatedwithinhalftheFullWidthatHalfMax-
3.1.Fittingcomponents
imum (FWHM) of the given SPIRE array (measured from the
radiocorepositionofthesources)wereselectedastentativede- Theestimationofphysicalpropertiesfortheactivegalaxieswas
tections. performedusinganSEDfittingtechnique.Ourfittingroutineis
While SPIRE does notsuffer fromcorrelated noise, SPIRE basedonacombinationofseveraldistinctcomponents,responsi-
observationsare dominatedby confusionnoise, of the orderof blefortheemissionfromactivegalaxiesindifferentwavelength
Articlenumber,page3of30
A&Aproofs:manuscriptno.Herschel3CR
Observed wavelength (µm) Observed wavelength (µm)
10 100 1000 10 100 1000
10−11 −11
RG 3C454.1 host QSR 3C205 1300K
torus torus
z = 1.84 SF z = 1.53 SF
−2m) 10−12 −12
c
−1s
g
er
F (ν 10−13 −13
ν
10−14 −14
1 10 100 10001 10 100 1000
Rest wavelength (µm) Rest wavelength (µm)
Fig.2.IRspectralenergydistributions(SEDs,solidblack)fortworepresentativeobjectsfromthiswork.Opencirclesshowthephotometricdata.
Errorbarscorrespondto1σphotometricuncertainties.Arrowsindicate3σupperlimits.Leftpanel:3C454.1,aradiogalaxyatz=1.84.Thethree
componentsusedtofittheSEDsofradiogalaxiesaccountforemissionfromhostgalaxy(old)stars(dash-dottedred),fromanAGN-heatedtorus
(dottedgreen),andfromdustheatedbystarformation(dashedyellow).Rightpanel:3C205,aquasaratz=1.53.Themulti-componentapproach
usedtofitquasarsaccountsforemissionfromhot(1300K)dust(long-dashedblue),fromtheAGN-heatedtorus(dottedgreen)andfromthestar
formationheateddust(dashedyellow).
regimes.Below,we describethismulti-componentapproachof FIRemission(inexcessoftheemissionoftheAGN-heateddust)
fittingtheobservedIRSEDs. asbeingpoweredbystarformation,andwerepresentitwithan
The presence of circumnuclear dust surrounding the broad opticallythinmodifiedblackbodycomponent,i.e.,a blackbody
line regions in AGN and blocking their UV/visible emission modifiedbyfrequency-dependentemissivity,givenby
is central to orientation-based unification of powerful FRII
S ∝ B (T)νβ. (1)
radio galaxies and quasars (Barthel 1989; Antonucci 1993). ν ν
Given its proximity to the AGN, the emission from this AGN-
Wereducethenumberoffreeparametersinthiscomponentby
illuminated warm dust peaks at rest-frame MIR (10-40 µm)
fixing the dust emissivity index β to a typical value of 1.6 as
wavelengths (e.g. Rowan-Robinson 1995). Spitzer photometric
foundinstudiesofhigh-zAGN(e.g.Beelenetal.2006).There-
and spectroscopic data have shown that the majority of high-
mainingtwofreeparametersherearethecolddusttemperature
z 3CR objects are luminous MIR emitters (Haasetal. 2008;
and the flux normalization of the modified blackbody compo-
Leipskietal. 2010), with observed MIR luminosities L
15µm nent. The use of a modifiedblackbody,as opposedto starburst
much higherthan 8 × 1043 ergss−1, the value separatinghid- templates(e.g.Drouartetal.2014),mightslightlyunderestimate
den quasars from mid-IR weak radio galaxies at intermediate thestarformationluminositiesbecauseonemissesthestarburst
redshifts (Ogleetal. 2006). There exists broad agreement that MIRemission, but offersthe uniquepossibility of constraining
theAGN-heatednucleardustismainlylocatedinclumpswhich thecolddusttemperatures.
are distributedin a toroidalpatternaltogetherreferredtoas the ThetwocomponentsdescribedabovefeatureintheSEDfit-
AGNtorus(e.g.Nenkovaetal.2002;Kuraszkiewiczetal.2003; ting of both radio galaxies and quasars. We include additional
Hönigetal. 2006; Schartmannetal. 2008). To account for the SEDcomponentstothefittingdependingonthetypeofthestud-
emission from the AGN heated dust, we chose the library of iedobject.Forradiogalaxies,weaddedablackbodycomponent
torus modelsfrom Hönig&Kishimoto (2010). The parameters peakinginthenear-IRtoaccountfortheemissionfromtheold
consideredwhencreatingthislibraryincludethe(1)radialdust stellar population in the AGN host (e.g. Seymouretal. 2007).
distributionofdustclumps;(2)geometricthicknessofthetorus; Thetemperatureoftheblackbodyanditsfluxnormalizationare
(3) numberof clumpsalongan equatorialline of sight; (4)op- thetwofreeparametersforthisSEDcomponent.Forquasars,we
ticaldepthoftheindividualclumps;and(5)outerradiusofthe addeda blackbodycomponenttoaccountforthehot(graphite)
torus.Thereare240setsofparametersinthelibrary,eachcom- dust close to the sublimation temperature. This component is
putedforsevenviewing-anglesrangingfromface-on(i= 0◦)to often empirically required to fit the observed SEDs of quasars
edge-on(i = 90◦) in steps of 15degrees,resulting in a total of (e.g. Mor&Netzer 2012; Leipskietal. 2013). Following these
1680torusmodels.Detailedinformationonthemodelparame- authors, we fixed the temperature of the blackbody to 1300 K,
tersandtheadoptedstrategyingeneratingthetoriSEDsispro- leavingitsfluxnormalizationastheonlyfreeparameterduring
videdbyHönig&Kishimoto(2010).Inadditiontotheparame- thefitting.Suchacomponentisalsoneededinthefittingofthe
ters listed above,the overallflux normalizationofthe modelis SEDs of a few radio galaxies (see also B12) whose observed
anotherfreeparameterthroughoutthefittingprocedure(outlined photometryintheNIR/MIRcouldnotbewellrepresentedwith
below). thecomponentsdescribedabove.Theseradiogalaxies,indicated
The rest-frame FIR emission (40-500µm) is largely gener- inTable4,mightbeviewedalonglinesofsightatwhichthenu-
atedbycolddust,heatedbystarformationonkpc-scalesinthe clearregionisonlypartlyobscured,thusresultinginsomewhat
AGN host(e.g. Rowan-Robinson1995; Schweitzeretal. 2006; elevated MIR luminosities. The inclusion of the hot dust com-
Netzeretal. 2007). Following these authors, we interpret any ponentto the SEDs of some of the radio galaxies might lower
Articlenumber,page4of30
P.Podigachoskietal.:Starformationinz>13CRhostgalaxiesasseenbyHerschel
theestimatesofthemassoftheevolvedstellarpopulations,but For objects with fewer than three Herschel detections, we
this is outside the scope of this work. For the quasars we also estimatedupperlimitsfortheIRstarformationandIRAGNlu-
consideredanadditionalpower-lawcomponentrepresentingthe minositiesusingtwodifferentapproaches.Inthefirstapproach,
emissionfromanaccretiondiskintheUV/visible.However,as we fitted the (longer wavelength) Herschel upper limits using
demonstrated in Appendix A, the inclusion of this power-law a modified blackbody with fixed β = 1.6 value and fixed cold
componenthadlittle influenceon the resultsobtainedfromthe dusttemperature(T = 37K),typicallyfoundfromthefitsof
dust
FIR partof the SED, thereforeit was excludedfromthe fitting theobjectsdetectedinatleastthreeHerschelbands(seebelow).
procedure. We thenintegratedunderthe blackbodycomponenttoestimate
upper limits for the star formation luminosities. In the second
approach,we took the 70/160/250µm upper limits as tentative
3.2.Fittingprocedure detections,andestimatedupperlimitsforthestarformationlu-
minositiesusingthe procedureadoptedfortheobjectsdetected
Whileourphysicallymotivatedfittingapproachresultsinaclose
inatleastthreeHerschelbands.Bothapproachesyieldedsimi-
approximationoftheobservedSEDsofthesampleobjects,itis
larresults(within10%)forthestar formationluminosities,but
not primarily designed to yield precise models of their SEDs.
we retained the second approach because it allowed us to es-
In particular, we are not interested in constraining the proper-
timate the IR AGN luminosities for objects with only PACS
ties of the dusty torus with our multi-wavelength broad-band
70 µm detections. Best-fit SEDs for objects with fewer than
photometry. This kind of analysis remains challenging even at
three Herschel detections, together with images centred on the
lower redshifts (e.g. RamosAlmeidaetal. 2009). Torus mod-
radiopositionsoftheAGN,arepresentedinAppendixDandin
els are used to separate AGN-heated dust emission (peaking
AppendixE,respectively.SystematicallydemandingthreeHer-
in MIR) from star-formation-heated dust emission (peaking in
scheldetectionswhenfittingtheobservedphotometryregardless
FIR), and to determine for the first time the star-formation-
ofredshiftmeansthattheFIRresultsfortheobjectsdetectedin
dominatedFIRenergeticsofthehigh-z3CRsources.Whenfit-
onlythe two PACSbandsare treatedasupperlimits. However,
ting the observed SEDs of our objects, we used a chi-square
dependingontheobject’sredshift,thePACS160µmbandalone
minimizationtechniquebasedontheMPFITroutine(Markwardt
may probe the peak of the cold dust emission, allowing robust
2009).Inpractice,westartedwithatorusmodelfromthelibrary
constraintsforthephysicalparametersestimatedinthiswork.
ofHönig&Kishimoto(2010)andaddedalinearcombinationof
theremainingSEDcomponents(dependingontheobjecttype)
to minimize the overall chi-square. We repeated the procedure 3.3.Synchrotroncontribution
foreachtorusmodelinthelibrary.
Earlier submillimetre and millimetre studies of high-redshift
Example best-fit SEDs, along with their individual SED
3CRsourcespresentedclear evidencefora synchrotroncontri-
components, are shown in Fig. 2. At the redshifts of our sam-
bution to the observed flux densities (vanBemmeletal. 1998;
ple, the PACS 70 µm band is crucial in our adopted fitting ap-
Archibaldetal. 2001; Willottetal. 2002; Haasetal. 2006).
proach, as it strongly constrains the longer wavelengths of the
Whilenegligibleforradiogalaxies,extrapolationfromcorera-
torus emission. On the other hand, the SPIRE 250 µm band is
dio data shows that synchrotron emission can account for up
the most important measurement for constraining the compo-
to 80% of the observed submm flux densities from quasars
nent representing the cold dust emission. Therefore,our fitting
(Haasetal.2006).However,thelongestrest-framewavelengths
approachwas appliedto all objectsthat are detected in at least probedbytheSPIRE 500µmbandinourstudyare∼ 200µm:
three Herschel photometric bands (typically the three shortest
emissionatthesewavelengthsiscompletelydominatedbydust
Herschel bands). These objects are homogeneously distributed
andthereforefreefromanysynchrotroncontribution.Theonly
intheredshiftrange(1 < z < 2.5)studiedinthiswork(Fig.1).
exception is 3C 418. This source appears to be flat-spectrum-
Best-fitSEDsfortheseobjects,togetherwithimagescentredon
core-dominated (in an otherwise steep-spectrum selected sam-
the radio positions of the AGN, are presented in Appendix C
ple), and as such its FIR SED is clearly dominated by non-
andinAppendixE,respectively.OccasionalSEDmismatchesat
thermal (synchrotron) radiation from its core. The power-law
observed-frame16 µm and/or 24 µm are most probablydue to
likeIRSEDofthissourceisshowninAppendixC.3C418was
luminouspolycyclicaromatichydrocarbonemission and/orthe
removedfromthesubsequentanalyses.
10µmsilicateabsorption(Haasetal.2008;Leipskietal.2010).
AtFIRwavelengths,thefixedβapproachmaybethemainrea-
sonbehindthefailureofthefittingroutinetoexactlyreproduce
4. Results
someSPIREdatapoints.
The physical parameters constrained by our fitting method 4.1.Detectionstatistics
includeIRstarformationandIRAGNluminosity,andthetem-
TheHerscheldetectionratethroughoutoursamplerangesfrom
peratureandmassofthecolddustcomponent(Fig.3,Table4).
67%inthePACS70µmbandto13%intheSPIRE500µmband.
Uncertainties in the derived parameters were calculated by re-
Inparticular,7objectshaverobustdetectionsinallfiveHerschel
sampling the observed SEDs, allowing the individual photo-
bands.Furthermore,atotalof24objectsaredetectedinatleast
metric measurements to vary within their 1σ ranges of uncer-
three Herschel bands, most importantly in the SPIRE 250 µm
tainty. More precisely,we generated100mock observedSEDs
band,whichforthehighestredshiftofoursample(3C257:z =
andstudiedthedistributionsoftheparametersderivedfromthe 2.47)correspondsto ∼ 70µm rest-frameemission2. Excluding
corresponding best fits. We inspected the best fits to the mock
3C418fromthese24objects,resultsin13radiogalaxiesand10
SEDstoconfirmtheiroverallqualityinthedifferentwavelength
regimes.From the distributions,we retainedthe medianvalues 2 For the median redshift of our sample, (z =1.38), the SPIRE
med
as the best estimates of the parameters, and the 16th-84th per- 250µmbandsamplesthepeak(∼100µm)ofthetypicalcolddustSED,
centilerangesastheirassociateduncertainties(whichincaseof allowingstrongconstraintsonthemodifiedblackbodycomponentused
aGaussiandistributionwouldcorrespondto±1σvalues). inthefittingapproach.
Articlenumber,page5of30
A&Aproofs:manuscriptno.Herschel3CR
Table3.Objectswithsignificantthermalsubmillimetrefluxdensities. shows that the RGs and QSRs occupy different ranges in the
distributionsof L , with the distributionforQSRs shifted to
AGN
Object ThermalF Reference higher values compared to that for RGs. Haasetal. (2008) al-
850µm
(mJy) ready established this result by investigating the (energetically
important)rest1.6-10µmwavelengthrangeforthehigh-z3CR
3C191 2.95 1,2
objects, finding the QSRs to be, on average, 3-10 times more
3C257 5.40 3
luminousthanRGs.Weconfirmtheirfindingbyincludingrest-
3C280.1 2.48 1,2 framewavelengthslongerthan10µm.ThemedianIRAGNlu-
3C298 7.25 1,2 minositiesoftheRGsandQSRsplottedinFig.3are3.7×1012
3C368 4.08 3 L⊙and1.1×1013L⊙,respectively.
Asoutlinedinprevioussections,weattributetheFIRemis-
3C432 6.33 1,2
sion in excess of the AGN-powered dust emission to emission
3C470 5.64 3
fromstar-formation-heateddust.Figure4showsL asafunc-
SF
4C13.66 3.53 3 tionofL forallobjectsinoursample.Thepresence/absence
AGN
of correlation between these two parameters depends on both
References.(1)Willottetal.(2002);(2)Haasetal.(2006);
redshift and AGN luminosity, and is still debated in the litera-
(3)Archibaldetal.(2001).
ture(Lutz2014).Giventhedata,andtakingintoaccountonlythe
FIR-detectedobjects,thetwoplottedparametersshowatmosta
weakcorrelation,in partintroducedbythe dependenceof both
quasars.TheSpitzerdetectionratethroughoutoursampleranges
L andL onredshift.Moreover,thenumerousupperlimits,
from94%intheIRS16µmandMIPS24µmbandsto100%in SF AGN
together with the fact that both parameters span only a limited
theIRAC5.8µmband.Commentsonselectedindividualobjects
range (∼ 2 orders of magnitude), make it difficult to establish
areincludedinAppendixB.
anysuchcorrelation(orlackof)inoursample.Nevertheless,we
observearangeofL fromweak(ifnotabsent)toverystrong,
SF
4.2.PhysicalpropertiesobtainedfromtheSEDfitting coevalwiththegrowthoftheblackhole.Figure4alsoshowsthat
thehostsofeventhestrongestAGNcanhavesignificantstarfor-
Studies investigating the cold dust temperatures in high-z ob-
mationactivity,unlikethetrendsfoundbyPageetal.(2012)for
jectspriortoHerschelwereoftenuncertainbecausetheyrelied
radio-quietAGN.Ingeneral,thetotalIRemissionfromthe3CR
heavilyonobservationsinasinglephotometricbroadband(e.g.
AGNispredominantlyAGNpowered,despitethefrequentlyac-
Benfordetal. 1999; Beelenetal. 2006). In our current study,
companyingstrongstarformationactivity.
cold dust temperatures estimated for the FIR-detected objects
WeestimatedthemassoftheFIRemittingdustcomponent,
range from ∼ 25 to ∼ 45 K (Fig. 3, Table 4). Radio galaxies
M ,using
andquasarsspanthesamerangeincolddusttemperatures,This dust
rangeissimilartothatobtainedforz> 5quasars(Leipskietal. S D2
2014) and to that estimated for distant submm galaxies (e.g. M = 250µm L , (2)
dust κ B (250µm,T )
Magnellietal. 2012). Similarly to Leipskietal. (2013), the in- 250µm ν dust
clusion of the 1300 K hot dust component in the SED fitting
where S is the flux at 250 µm rest-frame found from the
lowerstheestimatesofthecolddusttemperaturesby∼5K.By 250µm
best-fit,D istheluminositydistance,κ isthedustabsorp-
includingthehotdustcomponent,wepreferentiallyselecttorus L 250µm
tion coefficientat250 µm (κ = 4 cm2g−1 fromthe models
models that emit more of their energy at longer wavelengths. 250µm
ofDraine2003),andB (250µm,T )isthevalueofthePlanck
Asaconsequence,thecolddustcomponentsarealsoshiftedto ν dust
function at the corresponding rest-frame wavelength and tem-
coldertemperatures.
perature.ResultsareshowninTable4.GiventhattheRGs and
Central to the subsequent discussion are the star formation
QSRsin oursamplecoverroughlythe sameredshiftrangeand
luminosities,L ,whichwecomputed(orestimatedupperlim-
SF showsimilarstarformationproperties,itisnosurprisethattheir
its) by integrating the best-fit modified blackbody components
colddustmassesarecomparableaswell.Moreinterestingly,the
from8µmthrough1000µm.Figure3showsthedistributionof
masses of the cold dust component in the hosts of radio-loud
L forthe radiogalaxiesand quasarsdetectedinat leastthree
SF AGNdetectedinatleastthreeHerschelbandsarecomparableto
Herschelbands.Bothtypesofobjectsshowsimilarlybroaddis-
thoseobtainedforSMGsatredshiftssimilartothoseofoursam-
tributions,withmanyobjectshavingL >1012L ,characteriz-
SF ⊙ ple(e.g.Santinietal.2010).Thedustmassesofradio-loudAGN
ingthemasultra-luminousinfraredgalaxies(ULIRGs)(seealso
provide clues to the triggering of the starburst event (and also
Table4).ThemedianstarformationluminositiesoftheRGsand
thatoftheblackholeactivity,Tadhunteretal.2014).Giventhat
QSRsplottedinFig.3are2.0×1012 L and2.6×1012 L ,re-
⊙ ⊙ the high-z SMGs are likely undergoing strong merger-induced
spectively.ConvertingthestarformationluminositiesintoSFRs
starburst events (e.g. Kartaltepeetal. 2012), their similar dust
usingthecalibrationderivedbyKennicutt(1998)gives100M
⊙ content suggests that that FIR-luminous radio-loud AGN also
yr−1 <SFR<1000M yr−1,consistentwithSFRsobtainedfor
⊙ builduptheirstellarmassinmajorgas-richmergers.
typical submm galaxies (SMGs) at comparable redshifts (e.g.
Magnellietal.2012).
TocomputetheIRluminositiesofthecomponentspowered 4.3.MedianSEDs
by the AGN, L , we integrated the best-fit torus component
AGN
Thebest-fitSEDsfortheobjectsdetectedinatleastthreeHer-
fortheRGsandthesumofthebest-fithotdustandtoruscom-
ponent for the QSRs3 between 1 µm and 1000 µm. Figure 3 schel bands are shown in Appendix C. The SEDs show a con-
siderable range of shapes and absolute scaling, with all QSR
3 AfewRGsalsorequireanadditionalhotdustcomponenttobetterfit (andafewRG)SEDspeakingatwavelengthsaround20µm,and
theirobservedphotometry.ThecomputationofL fortheseobjects mostRGSEDspeakingatlongerwavelengths.Furthermore,the
AGN
takesintoaccountthiscomponentaswell. SEDsshowthattheAGN-poweredandstar-formation-powered
Articlenumber,page6of30
P.Podigachoskietal.:Starformationinz>13CRhostgalaxiesasseenbyHerschel
Fig.3.DistributionsofindividualphysicalparametersobtainedfromtheSEDfits,forradiogalaxies(red)andquasars(blue).Thevaluesplotted
arethosefortheobjectsdetectedinatleastthreeHerschelbands.Upperleftpanel:temperatureofthecolddustcomponentemittingintheFIR,
T .Upperrightpanel:massoftheFIRemittingcolddustcomponent,M .Lowerleftpanel:AGN-poweredIRluminosity,L .Lowerright
dust dust AGN
panel:star-formation-poweredIRluminosity,L .TheverticallinesinthelowerpanelscorrespondtotheaveragevaluesoftheFIR-detected(solid
SF
lines)andnon-detected(dashedlines)stackedsubsamplesdiscussedinSect.4.4.
dustemissionswitchdominancetypicallyat50µm,(butthiscan framewavelengthsbetween10and40µm,themedianRG and
happenatallwavelengthsbetween35and65µm).Giventhatall QSRSEDsstillshowaconsiderableanisotropy,withtheQSRs
SEDs were computed on the same rest-frame wavelength grid, being a factor of two more luminous compared to the RGs at
wecreatedmedianSEDsforthetwotypesofobjectstoprevent 20µm.Atrest-frameFIRwavelengths(&40µm),however,the
the most extreme objects dominatingthe average SEDs. When medianRGandQSRSEDsappeartoberemarkablysimilarboth
creatingthemedianSEDs,werefrainedfromapplyinganynor- inshapeandabsolutescale,arguingfor,onaverage,similarstar
malization,inordertopreservetheabsoluteluminositiesofthe formationpropertiesforthehostsofbothtypesofAGN.
individualSEDs.Themedian4SEDsareshowninFig.5.
ThemedianSEDsofRGs andQSRs differstronglyatrest- Figure 5 also shows the average SED for the subsample of
frame3-10µm,withtheQSRsbeingafewtimesmoreluminous the z ∼ 2 star-forming galaxies from Kirkpatricketal. (2012).
than the RGs. Such an observational difference in this wave- ThisaverageSEDiscomposedof30(U)LIRGgalaxies,initially
lengthregimehasalreadybeenexplainedbyHaasetal. (2008) selectedbasedontheir24µmfluxdensity(F24µm &100µJy),for
inthecontextofunificationbyorientation.Inthisscenario,the which good multi-wavelengthcoverageis available throughout
observed luminosity differences result from viewing the QSRs muchof the IR regime.Representingthe star-formation-heated
and RGs along differentangles, such that the hotinner regions colddustemissionwithamodifiedblackbody,Kirkpatricketal.
of the dusty torus are observed directly in the case of QSRs (2012) estimate the average SFR and cold dust temperature of
butare obscuredin thecase ofRGs. AsfoundbyLeipskietal. thesegalaxiestobe344±122M⊙yr−1and28±2K,respectively.
(2010),similarobservationaldifferencescorrelatewithorienta- Comparing these (U)LIRG numbers to those obtained for the
tion indicators, such as the radio core dominance. The median FIR-detected3CRsinourwork,wefindcomparableSFRsbuton
SED of RGs can therefore be viewed as the median SED of averagehighercolddusttemperatures.Furthermore,Fig.5con-
reddenedQSRs (Haasetal. 2008; Leipskietal. 2010). At rest- firmsthemarkeddifferencebetweenthe(U)LIRGsand3CRsin
theNIR/MIRluminosity:whilethez ∼ 2star-forminggalaxies
4 The QSRs median SED is given only for rest-frame wavelengths arecharacterizedbypronouncedpolycyclicaromatichydrocar-
longerthan2µmbecausetheemissionatwavelengthsshorterthan2µm bon(PAH)features,thepowerfulemissionfromthewarmdusty
isdominatedbythehotaccretiondisk,whichwasnotincludedinour torus completely outshines these features in the MIR SED of
SEDfitting(asexplainedinAppendixA). the 3CR host galaxies. Finally, the average stellar mass of the
Articlenumber,page7of30
A&Aproofs:manuscriptno.Herschel3CR
Table5.ResultsfromtheSEDfittingofthestackedsubsamples.
Subsample log(L (L )) log(L (L ))
AGN ⊙ SF ⊙
FIR-detectedRGs 12.6 12.2
non-detectedRGs <12.2 <11.7
FIR-detectedQSRs 13.2 12.1
non-detectedQSRs 12.7 <11.6
4.4.Stackingofnon-detectedobjects
Inadditionto probingthe generalpropertiesof theobjectsthat
are individually detected in at least three Herschel bands, we
attempt a stacking analysis to extract an average signal for the
objects which lack significant emission in the Herschel bands.
Forconvenience,werefertotheformerobjectsasFIR-detected,
andto the letter asnon-detected.Ouraim is to discussaverage
Fig. 4. IR emission from star-formation-heated cold dust, L , versus
SF
IRemissionfromAGN-powered dust,L ,for radiogalaxies(filled propertiesforeachradio-loudAGNtype,thereforewesplitthe
AGN
red circles) and quasars (filled blue squares). Upper limits have been non-detectedobjectsintoRGsandQSRs, respectively.Inorder
estimatedasexplainedinthetext.Thelargeemptysymbolscorrespond toretaindecentnumberstatisticsinthetwosubsamples,wede-
tothesubsamplesofFIR-detectedRGs(circle)andQSRs(square),and cided not to further divide the RGs and QSRs in redshift bins,
non-detectedRGs(triangle)andQSRs(diamond)discussedinSect.4.4. despite the considerable redshift range of our sample. In prac-
ThedashedlinemarksLSF=LAGN. tice, we take the non-detected objects to be those that are de-
tectedinatmostone,namelythePACS70µmband,whichgiven
theredshiftsofoursourcesoftenprobesthepeakemissionfrom
thedustytorus.SeveralRGsandQSRswerenotincludedinthe
stackinganalysisowingtothepresenceofpotentiallyconfusing
sourcesinthecorrespondingmaps,closetotheknownradiopo-
sitionoftheobjectinquestion.Furthermore,twoRGs(3C252
and 3C 267) with a strong detection in only the PACS 70 µm
bandwerenotincludedinthe stackingbecausetheyareincon-
sistentwith formingasinglepopulationwiththeobjectswhich
havenoHerscheldetections.Asaresult,stackingwasperformed
on subsamples of 12 RGs and 6 QSRs with median redshifts
of z =1.44 and z =1.27, respectively. The objects entering
med med
thesubsamplesareflaggedinTable4.Forcomparison,wealso
selected two subsamples from the FIR-detected objects: a sub-
samplewith 13RGs (z =1.34)andanotheronewith 8 QSRs
med
(z =1.52;3C298and3C318,sourceswithstronglyemitting
med
nearbyobjects,werenottakenintoconsideration).
We stacked equal areas extracted from the individual Her-
schel maps, centred on the known radio position. Photometry
on the stacked map was performed following the same proce-
duresasadoptedinthecaseoftheindividuallydetectedobjects
Fig.5.Medianspectralenergydistributions(SEDs)fortheradiogalax- (Sect.2).Weexaminedthediversitywithinthegivensubsample
ies(dashedred)andquasars(dottedblue)detectedinatleastthreeHer- by bootstrapping with 1000 realizations. In practice, from the
schel bands, and for z ∼ 2 star-forming galaxies (solid yellow) from originalsubsamplesidentifiedabove,foreachbootstrappingre-
Kirkpatricketal.(2012).Shadesareas,redforradiogalaxiesandblue alizationweselectedarandomsubsample(withthesamenum-
forquasars,correspondtotheassociated16th-84thpercentileranges.
ber of objects, allowing for repetitions), stacked the Herschel
maps,andperformedthephotometry.Thecentroidandthedis-
persion of the resulting distribution were taken to be the mean
z ∼ 2 star-forminggalaxiesis a factor of∼ 3 smaller than that fluxdensityoftheon-sourcestackanditsassociateduncertainty.
oftheRGhostsasseenfromtheoffsetintherespectiveSEDsat Additionally,westackedrandompositionsinordertoinspectthe
around1.6µm.WhileAGNcontaminationtothe1.6µmisnot overallsignificanceoftheon-sourcestackedsignal.Ifthemean
accountedforin thisdiscussion,the verysteep medianSED of valueoftheon-sourcestackdistributionwasatleastthreetimes
radiogalaxiesbetween4and6µmsuggeststhatanycontribution larger than the mean value of the corresponding background
fromhotdusttothe1.6µmfluxisnegligible.AGNcontamina- stackdistribution,weconcludedthattheon-sourcesignalissig-
tionat1.6µmlikelyispresentinindividualcases,butdoesnot nificant.Incasesofanon-significantsignal,wetookthreetimes
appear to be crucialwhen discussing averageproperties.Over- the mean value of the background stack distribution to be our
all, the comparisonbetween the 3CR and the Kirkpatricketal. on-sourcestackupperlimitvalue.
(2012) sample shows that massive galaxies in the high-z Uni- The mean stacks for the four different subsamples selected
versemay beactivelyformingstarsregardlessofwhethertheir in our study are presented in Fig. 6. The non-detected quasars
supermassiveblackholesareaccretingornot. haveasignificantstackedsignalinonlythePACS70µmband,
Articlenumber,page8of30
P.Podigachoskietal.:Starformationinz>13CRhostgalaxiesasseenbyHerschel
FIR-detected RGs FIR-detected RGs FIR-detected RGs FIR-detected RGs FIR-detected RGs
70µm 160µm 250µm 350µm 500µm
non-detected RGs non-detected RGs non-detected RGs non-detected RGs non-detected RGs
70µm 160µm 250µm 350µm 500µm
FIR-detected QSRs FIR-detected QSRs FIR-detected QSRs FIR-detected QSRs FIR-detected QSRs
70µm 160µm 250µm 350µm 500µm
non-detected QSRs non-detected QSRs non-detected QSRs non-detected QSRs non-detected QSRs
70µm 160µm 250µm 350µm 500µm
Fig.6.Stacked imagesinall Herschel bands, for thefour subsamples discussed here.Fromlefttoright: PACS70 µm,PACS160 µm,SPIRE
250µm,SPIRE350µm,andSPIRE500µmbands,respectively.Fromtoptobottom:FIR-detectedradiogalaxies,non-detectedradiogalaxies,
FIR-detectedquasars,andnon-detectedquasars.Eachstackedimageshownherehasdimensionsof2x2arcminandiscentredontheknownradio
positionsoftheobjectsenteringthestack.
detectedquasarsandradiogalaxieshavesignificantstackedsig-
nalsinall, andinallbuttheSPIRE 500µm band,respectively.
In order to study the full IR SED of a stacked subsample, we
FIR−detected QSRs, N=8, zmed=1.52 calculated mean flux densities in the Spitzer bands, where vir-
non−detected QSRs, N=6, z =1.27
med tually all objects have been individuallydetected. All obtained
10−11 FIR−detected RGs, N=13, z =1.34
med averagefluxesareintheobservedframe:theseweretransferred
−2m) non−detected RGs, N=12, zmed=1.44 totherest-frame,usingthemedianredshiftofthecorresponding
c
−1s 10−12 subsample. The SEDs of the stacked subsamples are shown in
g Fig.7.
er
F (ν
ν The1-10µmSEDsofradiogalaxiesarecomposedofemis-
10−13
sionfromanevolvedstellarpopulationandfromaheavilyred-
dened AGN (Haasetal. 2008). At the rest-frame wavelengths
(given the comparable median redshifts of the various sub-
10−14
samples) probed by the two shortest wavelength Spitzer bands
1 10 100 1000
Rest wavelength (µm) (< 2 µm), the FIR-detected and non-detected subsamples of
RGs are similar. Thus, it is likely that the stellar masses of
Fig. 7. Spectral energy distributions of stacked subsamples selected the host galaxies for the two RG subsamples are, on average,
from the hosts of the 3CR sources studies here. The four subsamples similar as well. Identifying differences in the host galaxy stel-
presented are: FIR-detected radio galaxies (red circles), non-detected lar masses requires additional NIR photometry, and is outside
radiogalaxies(redtriangles),FIR-detectedquasars(bluesquares),and the scope of this work. At wavelengths between 2 and 10 µm,
non-detectedquasars(bluediamonds).Arrowsmark3σupperlimitsas the SEDsof the two RG subsamplesdiffer fromeachother. At
explainedinthetext. thesewavelengths,AGN-heatedhotdustemissionappearstobe
present/absentintheSEDoftheFIR-detected/non-detectedsub-
sample,respectively,reflectingdifferentlevelsof dustobscura-
whereasthenon-detectedradiogalaxieshavenosignificantsig- tion,asinferredfromthediversityintheindividualMIRSEDs
nal in any of the Herschel bands. On the other hand, the FIR- of RGs (Haasetal. 2008). Particularly interesting is the non-
Articlenumber,page9of30
A&Aproofs:manuscriptno.Herschel3CR
significant stacked signal for non-detected RGs at rest-frame quenchedthestarformationinthehostgalaxy,whichisatodds
30 µm, which couldindicatethat the two subsampleshavedif- with results presented by Pageetal. (2012). A series of coeval
ferentintrinsicAGN luminosities.Indeed,fittingthe SEDsfol- episodes of strong star formation and black hole activity may
lowing the routine taken in Sect. 3 gives (at least) a factor of have formed the massive host galaxiesand their massive black
twodifferencebetweentheestimatedL inthetwoRGssub- holes(seeB12).
AGN
samples(Table5,Fig. 4).AcleardifferencebetweentheSEDs The SFRs estimated for the FIR-detected radio-loud AGN
of the two subsamples of RGs is observed at rest-frame wave- arecomparableto those obtainedforSMGs atsimilar redshifts
lengthslongerthan60µm,suggestingmarkeddifferencesinthe (e.g. Magnellietal. 2012). Other parameters, such as the tem-
starformationpropertiesofthetwosubsamples.Thisresultsin peraturesandmassesofthecolddustcomponentandtotalstel-
(atleast)afactorofthreedifferencebetweentheestimatedSFRs lar masses agree with each other, at least on average, as well
in the two RGs subsamples (Table 5, Fig. 4). The SEDs of the (Santinietal. 2010; Michałowskietal. 2012; Swinbanketal.
twoQSRsubsamplesgenerallyfollowthesametrendsasthose 2014).Itiswidelythoughtthathigh-zSMGsformstarsinstar-
observedfortheRGsubsamples.Whilethenon-detectedQSRs burstevents,inducedasaresultofavarietyofdifferentprocesses
could overall be at the fainter end of the QSR population, the including mergers and tidal interactions. Radio-loud AGN are
moststrikingdifferenceintheQSRSEDsisatFIRwavelengths, known to be at the centres of over-densitiesin the high-z Uni-
with the FIR-detected QSRs having SFRs (at least) a factor of verse (e.g. Venemansetal. 2007; Wylezaleketal. 2013). Thus,
three higher than the non-detectedQSRs. While more analysis the mergerscenario appearsto be an attractive way of produc-
isrequiredtopindownthepotentialdifferencesamongthesub- ing at least some of the extremely high SFRs estimated in our
samples in wavelengthsotherthan FIR, it is beyonddoubtthat study.Asimilar conclusionhasbeendrawnina studyofafew
theformationofnewstarscanbeprodigiousinsome,butmodest high-zradiogalaxiesusingbothHerschelandCO observations
orweakinotherradio-loudAGNhosts. (Ivisonetal.2012).
Jet-inducedstarformation,alsoknownaspositivefeedback,
isyetanotherpossiblewaytotriggerhighSFRs(e.g.Deyetal.
5. Discussion 1997; vanBreugeletal. 1998) in hosts of radio-loud AGN. In
this scenario, the outgoingradio jet shocks the surroundingin-
5.1.StarformationinhostsofpowerfulAGN
terstellarmaterial,whichsubsequentlycoolsdowntoformnew
Starbursts powering the FIR emission in some hosts of radio- stars.Recently,basedonUV-to-submmtemplatesbuiltwiththe
loudAGNhavebeenfoundatlow-to-intermediateredshifts(e.g. evolutionary code PEGASE.3, Rocca-Volmerangeetal. (2013)
Dickenetal. 2010). In the high-z Universe, high levels of star showed that the star formation timescales from stellar popula-
formation in some radio-loud AGN have been estimated using tionsynthesisagreewellwiththeagesoftheradioepisodesfor
severalobservationalindicators.Theseincludetheusageofrest- twohigh-zradiogalaxies.Similar studiesoflargersamplesare
frame UV spectroscopy to detect huge Lyα halos surrounding requiredtobetter understandthe detailsof jet-inducedstar for-
theAGN(e.g.Villar-Martínetal.1999),submmphotometryto mation.
probethecooldustandmoleculargascontentofAGNhosts(e.g. Despiteevidenceforstrongstarformationinthe3CRhosts
Archibaldetal. 2001; Reulandetal. 2004), and Spitzer MIR detectedinatleastthreeHerschelbands,∼60%of3CRsources
spectroscopytodetectstrongPAHfeatures(e.g.Rawlingsetal. are detected in fewer than three Herschel bands, arguing for
2013). The latter, however,are seen only in rare cases because significantlylowerstar formationactivityin these hosts. While
the AGN-powered hot/warm dust emission in powerful radio- the upper limits in the SPIRE bands still allow energetically
loudAGNhostsusuallyoutshinesthePAHemission. significant SFRs of up to 300 M yr−1 in some of these hosts,
⊙
UV/visible data have been used to infer SFRs for several the overwhelmingmajority have SFRs of at most 100 M yr−1.
⊙
3CR radio galaxies studied in this work (Chambers&Charlot Thisraisesanimportantquestion:whatleadstothesignificantly
1990).Theseinclude3C065,3C068.2,3C266,3C267,3C324, lower star formation activity in these hosts? If the star forma-
3C 356 and 3C 368. The ratio between the SFR obtained us- tion activity is indeed merger-induced, one possible answer to
ing the approach taken in this work and the UV/visible SFR theabovequestionisthatoccurrencesofminor-and/orgas-poor
ranges between 4 and 40, therefore we conclude that the star mergersarelikelytoleadtorelativelylowlevelsofstarforma-
formation is often strongly obscured in the UV/visible. This tion activity. The relevance of the merger scenario in the con-
comparison demonstrates that rest-frame FIR data are crucial textofthetriggeringofstarbursts,butalsooftheAGNactivity,
in quantifying the dust-enshrouded star formation in the hosts hasbeendiscussedbyTadhunteretal.(2011).Theyfindthatthe
of these sources, especially the ones hosting quasars because stronglystarburstingradiogalaxiesintheirintermediate-redshift
theemissionfromtheaccretiondiskcomplicatestheSFRsesti- sample have opticalmorphologicalfeaturesconsistentwith the
matesfromUV/visibledata.UsingHerscheldata,Seymouretal. idea that they are triggered in major mergers. Another possi-
(2011)foundmeanSFRsfor1.2 < z < 3.0radio-selectedAGN bility is that we are observing some hosts after the star forma-
to range between 80 and 581 M yr−1. Similarly, Drouartetal. tionhasbeenquenched(e.g.Farrahetal.2012).IftheAGNac-
⊙
(2014) studied a sample of ∼ 70 1 < z < 5 radio galaxies, es- tivity is responsible for the quenching, then we argue that this
timatingSFRsofafewhundredtoafewthousandsolarmasses negative feedback is not universal, even if it acts over a very
per year. While the authors suggest that the highest values re- shorttimescale.First,giventhestrongstarformationinthe3CR
ported in their study are likely overestimated and should be hostsdetectedinatleastthreeHerschelbands,itisunlikelythat
treated as upper limits, the overall idea that the hosts of high- the quenching of star formation occurs before the radio-loud
z radio-loud AGN can be prodigious star-formers is consistent AGN phase. Second, quenching taking place during the radio-
with our findingsfor the hosts of the powerful3CR objects. A loudphaseisnotsupportedwiththefindinginSect.4.4thatthe
crucialpointintheseanalysesisthatthehighstarformationlu- <2µmrest-framestackedSEDsofthetwoRGsubsamplesare
minosities obtained are coeval with the growth of the SMBHs similar. In particular,if star formationin the non-detectedRGs
residing in the nuclei of the host galaxies. As such, these re- isquenched,thentheyareexpectedtobebrighterthantheFIR-
sultsargueforascenariowherebyonaveragetheSMBHhasnot detected RGs because the star formation in the latter is heav-
Articlenumber,page10of30