Cold Dust Emission from X-ray AGN in the SCUBA-2 Cosmology Legacy Survey: Dependence on ...

23 Pages·2015·0.9 MB·English

Mon.Not.R.Astron.Soc.000,000–000(0000) Printed2September2015 (MNLATEXstylefilev2.2)
Cold Dust Emission from X-ray AGN in the SCUBA-2 Cosmology
Legacy Survey: Dependence on Luminosity, Obscuration & AGN
Activity
5
1
0
Manda Banerji1,2⋆, R. G. McMahon1,2, C. J. Willott3, J. E. Geach4, C. M. Harrison5, S.
2
g Alaghband-Zadeh1, D. M. Alexander5, N. Bourne6, K. E. K. Coppin4, J. S. Dunlop6, D.
u
Farrah7, M. Jarvis8,9, M. J. Michałowski6, M. Page10, D. J. B. Smith4, A. M. Swinbank5,
A
M. Symeonidis10, P. P. Van der Werf11
1
3 1InstituteofAstronomy,UniversityofCambridge,MadingleyRoad,Cambridge,CB30HA,UK.
2KavliInstituteforCosmology,UniversityofCambridge,MadingleyRoad,Cambridge,CB30HA,UK.

] 3HerzbergInstituteofAstrophysics,NationalResearchCouncil,5071WestSaanichRd,Victoria,BCV9E2E7,Canada.
A
4CentreforAstrophysicsResearch,UniversityofHertfordshire,CollegeLane,Hatfield,HertfordshireAL109AB,UK.
G 5CentreforExtragalacticAstronomy,DepartmentofPhysics,DurhamUniversity,SouthRoad,DurhamDH13LE,UK.
6InstituteforAstronomy,UniversityofEdinburgh,RoyalObservatory,BlackfordHill,EdinburghEH93HJ,UK.
.
h 7DepartmentofPhysics,VirginiaTech,Blacksburg,VA24061,USA
p 8OxfordAstrophysics,DepartmentofPhysics,KebleRoad,OxfordOX13RH,UK.
- 9PhysicsDepartment,UniversityoftheWesternCape,Bellville7535,SouthAfrica.
o 10MullardSpaceScienceLaboratory,UniversityCollegeLondon,HolmburyStMaryDorking,SurreyRH56NT,UK.
tr 11LeidenObservatory,LeidenUniversity,POBox9513,NL-2300RALeiden,theNetherlands.
s
a
[

1 2September2015
v
8
1
ABSTRACT
0
0
Westudythe850µmemissioninX-rayselectedactivegalacticnuclei(AGN)inthe∼2
0
deg2COSMOSfieldusingnewdatafromtheSCUBA-2CosmologyLegacySurvey.Wefind
.
9 nineteen 850µm bright X-ray AGN in a “high-sensitivity” region covering 0.89 deg2 with
0
flux densities of S850=4-10 mJy. The 19 AGN span the full range in redshift and hard X-
5
ray luminosity covered by the sample - 0.7 . z . 3.5 and 43.2 . log10(LX) . 45. We
1
reporta highlysignificantstacked 850µmdetectionof a hardX-rayflux-limitedpopulation
:
v of 699 z > 1 X-ray AGN - S850=0.71±0.08mJy.We explore trends in the stacked 850µm
i flux densities with redshift, finding no evolution in the average cold dust emission over the
X
redshiftrangeprobed.ForType1AGN,thereisnosignificantcorrelationbetweenthestacked
r
850µmfluxandhardX-rayluminosity.However,inType2AGNthestackedsubmillimeter
a
fluxisafactorof2higherathighluminosities.WhenaveragingoverallX-rayluminosities,
nosignificantdifferencesarefoundinthestackedsubmillimeterfluxesofType1andType2
AGNaswellasAGNseparatedonthebasisofX-rayhardnessratiosandoptical-to-infrared
colours.However,atlog10(L2−10 / ergs−1 > 44.4),dependencesin averagesubmillimeter
flux on the optical-to-infraredcolours become more pronounced. We argue that these high
luminosityAGNrepresenta transitionfromaseculartoamerger-drivenevolutionaryphase
where the star formationrates and accretionluminosities are more tightly coupled.Stacked
AGN 850µm fluxes are compared to the stacked fluxes of a mass-matched sample of K-
bandselectednon-AGNgalaxies.Wefindthatat10.5<log10(M∗/M⊙)<11.5,thenon-AGN
850µmfluxesare1.5−2×higherthaninType2AGNofequivalentmass.Wesuggestthese
differences are due to the presence of massive dusty, red starburst galaxies in the K-band
selectednon-AGNsample,whicharenotpresentinopticallyselectedcataloguescoveringa
smallerarea.
Keywords: galaxies:active,submillimetre:galaxies,X-rays:galaxies
⋆ E-mail:[email protected]
(cid:13)c 0000RAS
2 M. Banerjiet al.
1 INTRODUCTION crucial information for the most luminous, high-redshift galaxies
and AGN. Recently, the SCUBA-2 bolometer on the JCMT has
Studyingtheconnectionbetweenstarformationingalaxiesandac- providedasubmillimetermapat850µmovertheentire2deg2field
cretion onto the supermassive black-hole is of prime importance aspart of theSCUBA-2Cosmology LegacySurvey (Geachetal.
in understanding galaxy formation (e.g. see Alexander&Hickox 2013). The 850µm data offers several advantages over Herschel
2012forarecentreview).Thediscoverythatsupermassiveblack- observations over the same region. Firstly, the superior angular
holesareubiquitouswiththeirmassconnectedtightlytothestel- resolution and smaller beam size relative to Herschel mean that
larbulgemassofgalaxies(Magorrianetal.1998;Kormendy&Ho sourceblendingislessofanissuewheninferringtheaveragesub-
2013) has led to the realisation that these active galactic nuclei millimeterpropertiesofAGN(seeFig.1).Secondly,thenegative
(AGN)playafundamentalroleinregulatingthegrowthofgalaxies. k-correctionatthesewavelengthsatz > 1(Blainetal.2002),en-
Sophisticatedgalaxyformationsimulationsnowincorporatefeed- suresthatwearesensitivetoroughlyconstantdustluminosityfor
backfromtheAGN,whichisshowntobecriticalinordertoex- essentiallyallhighredshiftgalaxiesinan850µmflux-limitedsam-
plaintheobservednumbersofmassivegalaxies(e.g.Springeletal. ple.Furthermore,evenintheHerschel-SPIREbands,therecanbe
2005; Crotonetal. 2006; DiMatteoetal. 2005). In these simu- asignificantcontributionfromAGNdustheatinginthecaseofthe
lations, massive galaxies are assembled through gas-rich mergers mostluminousquasars(e.g.Rosarioetal.2012).Thesubmillime-
whichinduceluminousburstsofobscuredstarformation.Starfor- ter dust emission at 850µm is much less prone to such contami-
mationandblack-holeaccretionarefuelledbyacommongassup- nationfromtheAGN.Althoughitcouldpotentiallybemorecon-
plyandthestarformationisinitiallydust-obscured.Asthegasand taminatedbysynchrotronemission,thisisunlikelytobeanissue
dustareexpelledfromthegalaxye.g.viaAGN-drivenwinds,the forthemajorityofAGNwhichareradio-quietThedustemission
accretingblack-holeemergesasanX-ray/UVluminousquasar.In at850µmcanthereforemoresafelybeattributedtoheatingbystar
suchapictureofgalaxyformation,luminousstarburstgalaxies,ob- formation.Atshorter wavelengthsitbecomes necessarytodisen-
scuredandunobscuredAGNarelinkedthroughawell-definedevo- tangle the relative contributions of starburst and AGN heating to
lutionarysequence.WithinunifiedAGNmodels(Antonucci1993; thetotalinfraredfluxandlinkingthetwophenomena istherefore
Urry&Padovani 1995), the obscuration of the central accreting morecomplicated,withpotentialbiasesarisingtoderivedstarfor-
sourcecanalsobeexplainedpurelyasaresultoforientationeffects mation rates due to different assumptions regarding the intrinsic
andthedistinctionbetweenunobscuredType1AGNandobscured AGNmodel.
Type2AGNisattributedtothelineofsightthroughwhichtheyare The aim of this paper is to study the 850µm properties of
viewed. bothunobscured(type1)andobscured(type2)AGNintheCOS-
Obscuration in AGN almost certainly results from a com- MOS field where statistically significant numbers of these popu-
bination of orientation and evolution driven factors. One way to lationsnow exist. Welook for dependences of thissubmillimeter
distinguish between these is to look for evidence for obscured emissionontheAGNhostgalaxypropertiesandcomparetonon-
AGN having excess star formation in their host galaxies rela- AGNgalaxies.Assuch,ouranalysisisintendedtoprovidethefirst
tive to their unobscured counterparts. Several early studies of X- SCUBA-2 Cosmology Legacy Survey 850µm view of the X-ray
ray luminous AGN using the SCUBA bolometer on the James AGNpopulationathighredshift.Throughoutthispaperweassume
Clerk Maxwell Telescope (JCMT) showed this to be the case aflatconcordancecosmologywithH =70kms−1Mpc−1.
0
(Pageetal. 2001, 2004; Stevensetal. 2005). More recently, new
data at far infrared wavelengths from the Herschel Space Ob-
servatory, has been used to study the connection between star
formation in galaxies and accretion onto the central AGN (e.g. 2 DATA
Bonfieldetal.2011;Lutzetal.2008;Shaoetal.2010;Pageetal.
2.1 SCUBA-2850µm
2012; Harrisonetal. 2012; Mullaneyetal. 2012; Rovilosetal.
2012; Rosarioetal. 2012; Santinietal. 2012; Azadietal. 2014; OurprimarydatasetusedinthisanalysisistheSCUBA-2850µm
Stanleyetal.2015).Lutzetal.(2010)havealsoconductedastudy mapintheCOSMOSfieldobtainedaspartoftheSCUBA-2Cos-
of the 870µm properties of z ∼ 1 AGN using LABOCA. There mologyLegacySurvey(S2CLS;Geachetal.2013).Thewide-field
isnowagrowingconsensusthatmoderateluminosityAGNtendto mapcoversatotalareaof∼2deg2 imagedusinga4×45′ PONG
occupystar-forminghostgalaxiesandthatthestarformationrateis scanningpattern.Thebasicprincipleofthedatareductionistoex-
independentofAGNX-rayluminosityinthisregime.AGNactiv- tract astronomical signal from the timestreams recorded by each
ityingalaxiesoccursonverydifferentspatialaswellastemporal bolometerintheSCUBA-2arrayandmapthemontoatwodimen-
scalesrelativetostarformation,whichmeansthattheinstantaneous sional celestial projection. We have used the Dynamical Iterative
starformationrateingalaxiesisunlikelytobetightlycoupledtothe Map-Maker(DIMM)withintheSub-MillimetreCommonUserRe-
instantaneousaccretionrate(Hickoxetal.2014).However,athigh ductionFacility(SMURF;Chapinetal.2013).Wereferreadersto
luminosities, major-merger driven evolution can lead to a tighter Chapinetal.(2013)foradetailedoverviewofSMURF,anddetailed
couplingbetweenAGNactivityandstarformation. datareductionstepswillbegiveninGeachetal.(inpreparation)
StudyingtheconnectionbetweenstarformationandAGNac- butdescribethemainstepshere.
tivityathighluminositiesrequiresmulti-wavelengthdatafromthe Flat-fieldsareappliedtothetime-streamsusingflatscansthat
X-ray to submillimeter wavelengths over reasonably wide fields. bracket each observation, and a polynomial baseline fit is sub-
TheCOSMOSfieldisamong themost well-studiedextragalactic tracted from each bolometer’s time-stream. Each time-stream is
fields witha plethora of multi-wavelength photometric and spec- cleanedforspikesandDCstepsareremovedandgapsfilled.After
troscopicdata.X-raydataisavailablefromboththeXMM-Newton cleaning, the DIMMentersaniterativeprocessthataimstofitthe
andChandraobservatoriesandthefieldhasbeenimagedatmany datawithamodelcomprisingacommonmodesignal,astronomical
otherwavelengths,inparticularwithSpitzerandHerschel.More- signalandnoise.Next,afilteringstepisperformedintheFourier
over,thewideareacoveredbymanyofthesedatasetsnowprovides domain,whichrejectsdataatfrequenciescorrespondingtoangular
(cid:13)c 0000RAS,MNRAS000,000–000
Cold DustEmissionfromX-RayAGNin S2CLS 3
Herschel-SPIRE250µm Herschel-SPIRE500µm SCUBA-2850µm
S / mJy S / mJy S / mJy
250 500 850

14 5.4 1.2
40 13 40 5.35 40 1
12
20 11 20 5.3 20 0.8
∆ Dec / arcsec−200 8910 ∆ DEC / arcsec−200 555...12255 ∆ DEC / arcsec−200 00..46
7 0.2
6 5.1
−40 −40 −40 0
5 5.05
4 −0.2
−6−06 0 −40 −20 0 20 40 −6−06 0 −40 −20 0 20 40 −6−06 0 −40 −20 0 20 40
∆ RA / arcsec ∆ RA / arcsec ∆ RA / arcsec
Figure1.Stacked250µm(left),500µm(middle)and850µm(right)imagesof2865radiosourcesintheCOSMOSfieldfromSchinnereretal.(2010).The
superiorspatialresolutionandangularscaleoftheSCUBA-2dataisapparentrelativetoHerschel.Forthe850µmstack,weconfirmthatthepeakemission
isclearlycomingfromthecentralpixelsdemonstratingthattheastrometryofthemapsisaccurateatthesub-pixellevel.Thenegative“bowling”aroundthe
brightcentralsourceseenintheSCUBA-2850µmimage,isanartefactofthefilteringprocedureemployedtogeneratethemap.
scalesθ>150′′andθ<2′′.Thenextstepistoestimatetheastro- 2.2 AGNCatalogue
nomicalsignal,whichissubtractedfromthedata.Finally,anoise
The aim of this paper is to study the submillimeter properties of
modelisestimatedforeachbolometerbymeasuringtheresidual,
a statistically significant and homogenous sample of AGN. The
whichisthenusedtoweightthedataduringthemappingprocess
COSMOSfield(Scovilleetal.2007)providesthelargestandmost
inadditional steps. Theiterativeprocessabove runsuntil conver-
well characterised AGN sample overlapping the new SCUBA-2
genceismet.Inthiscase,weexecuteamaximumof20iterations,
CLSdata.TheAGNparentsampleisacatalogueofXMM-Newton
orwhenthemaptolerancereaches0.05.
detected X-ray point sources, which goes down toa fluxlimit of
9.3e-15 erg/s/cm2 in the 2-10 keV band over 90% of the area.
Thefinal850µmmapintheCOSMOSfieldhasnon-uniform
dσe8p50th∼w4ithmJtyypiincatlheRMothServahlaulefsanodf σσ885500 ∼∼12mmJJyyininthoenecenhtarlafl, Tarhee5fleu-1x6liemrgi/ts/fcomr2thaendfai2n.t5ees-t1s5ouerrgc/ess/cdme2tecintedthein0t.h5e-2ckaetaVloagnude
2-10keVbandsrespectively.WeonlyconsiderX-raysourceswith
region.An850µmcataloguecontaining360sourceshasbeenpro-
robust multi-wavelength optical and infrared cross-identifications
ducedfromthismapbysearchingforallsourcesabove3.5σinre-
fromBrusaetal.(2010),whichcorrespondsto∼98%oftheX-ray
gionswithσ850 <2mJy(Geachetal.inpreparation).This850µm
catalogue covers an area of 0.89 deg2. Our search for AGN that sources. We make use of the latest spectroscopic and photomet-
ric redshifts for these AGN (Bongiornoetal. 2012; Salvatoetal.
areindividuallydetectedat850µmcoversthissmallerarea,higher
2011) together with the corresponding spectroscopic and photo-
sensitivityregion.However,whenstudyingthestackedproperties
metric classifications1. The Type 1 and Type 2 samples specifi-
oftheAGN,weutilisethefullareainordertogetthelargestAGN
callyconstituteAGNwithbroademissionlinesinthecaseofthe
samplespossible.Asstackedfluxesareweightedbythenoise(see
spectroscopicType1sandnarrowemissionlinesinthecaseofthe
Section 4), AGN in lower RMS regions are automatically down-
spectroscopic Type 2s. For the AGN with photometric redshifts,
weightedwhencalculatingaverage850µmfluxes.
Type1AGNareconsidered tobethosefitbytemplates19-30 in
Salvatoetal. (2011) whereas Type 2 AGN are those fit by tem-
Webeginbycheckingtheastrometryofthis850µmmapby
plates1-18or>100(whichcorrespondtogalaxytemplates).The
stacking a sample of 2865 radio sources in the COSMOS field
(Schinnereretal.2010)inthismap.Thestackingprocedureisde- photometricredshiftaccuracyisingeneralverygoodσ∆z/(1+z)
scribedinmoredetailinSection4andthestacked850µmimagefor =0.014foriAB < 22.5and0.015foriAB < 24.5(Salvatoetal.
2009).Photometryatmidinfra-redwavelengthsisfromtheSpitzer
theseradiosourcesisshowninFig.1.Thepeak850µmemission
hasanoffsetof1.4′′ relativetotheradioposition.TheSCUBA-2 S-COSMOSSurvey(Sandersetal.2007).Thereareatotalof1797
850µmpixelscaleis2′′.Thisconfirmsthatthemapshavethecor- AGN with multi-wavelength cross-identifications in these cata-
logues.Theobservedhard-band(2-10keV)X-rayluminosityver-
rectastrometryatthesub-pixellevelandcanthereforebeusedin
sus redshift for both the Type 1 and Type 2 AGN can be seen in
ourstackinganalysis.Wealsoshowthestackedimagesofthesame
Fig. 2. These X-ray luminosities have been calculated from the
radiosourcesintheHerschel-SPIRE250and500µmbandsinFig.
hard-band fluxes from Brusaetal. (2010) assuming a photon in-
1.TheHerschel-SPIREdataintheCOSMOSfieldistakenfromthe
dexofΓ=1.8.InthecaseoftheType1AGNthataredetectedin
HerschelMulti-tieredExtragalacticSurvey(HerMES;Oliveretal.
thesoft(0.5-2keV)bandbutnotinthehardband,thesoftX-ray
2012).Mapshavebeenproduced usingtheLevel2dataproducts
fluxesfromBrusaetal.(2010)areusedtoestimatethehardX-ray
fromtheESAarchiveasdetailedinSwinbanketal.(2014).While
flux,onceagainassumingΓ=1.8.Wenotethatnoabsorptioncor-
thestacked250µm emissionappearstobewellcentredatthera-
rection has been applied to these luminosities as the X-ray data
dio position, the 500µm emission shows an offset relative to the
for the entiresample isin general of insufficient quality toallow
radioposition.Thesuperiorspatialresolutionandpixelscaleofthe
full spectral fitting. The Type 1 AGN have negligible absorption
SCUBA-2datarelativetoHerschel-SPIRE,areclearlyapparentin
this plot. The longer wavelength SCUBA-2 850µm observations
arethereforeextremelycomplementarytoalreadypublishedHer-
schelobservationsofAGNinthisfield. 1 http://www2011.mpe.mpg.de/XMMCosmos/xmm53 release/
(cid:13)c 0000RAS,MNRAS000,000–000
4 M. Banerjiet al.
andinthecaseoftheType2COSMOSAGNthataredetectedin of850µmfluxdensitiesonawiderangeofAGNandhostgalaxy
thehardX-ray,Lussoetal.(2011)estimatetheaverageshiftinthe properties.TheseSEDfitsincorporatephotometryatoptical,near
hardX-rayluminosityintroducedbyanabsorptioncorrectiontobe infraredaswellasmidinfraredwavelengths. Around80percent
< ∆log(L2−10) >= 0.04±0.01. In Appendix A we compare ofthesourcesaredetectedintheSpitzerMIPS24µmband.Inad-
thehardX-rayluminositiesderivedinthispaper tothosederived dition, several of the AGN are also detected in the MIPS 70µm
byBrightmanetal.(2014)forasubsetoftheCOSMOSAGNwith band as well as in the Herschel-PACS 100µm and 160µm bands
Chandradataandwherecomplexabsorptionspectralmodelscan (Bongiornoetal. 2012). All of this multi-wavelength photometry
befit,demonstratingthattherearenosignificantbiasesintheX-ray isusedintheSEDfittingandbothpureAGNandpuregalaxytem-
luminositiesusedinthispaper. platesareallowedinthefits.Thegalaxytemplatesconstituteaset
In the majority of the analysis presented in this paper we of Bruzual&Charlot (2003) stellar population synthesis models
consider the average stacked submillimeter fluxes of AGN as with a Chabrier (2003) initial mass function and assuming expo-
a function of various properties. For these stacking studies, we nentially declining star formation histories with varying amounts
construct flux-limited populations of AGN restricted to sources of dust reddening. The AGN template is the mean quasar SED
with S>2.5×10−15 erg s−1 cm−2 in the hard (2-10 keV) band fromRichardsetal.(2006),againwithvaryingamountsofredden-
(Lussoetal.2011).Aspreviouslymentioned,Type1AGNthatare ing.ThischoiceofAGNtemplateisappropriatefortheluminous
detectedinthesoftband butnot inthehardband, havehad their XMM-COSMOSAGNinvestigatedinthiswork.Bongiornoetal.
hardbandfluxesestimatedfromthesoftX-rayfluxusingaΓ=1.8 (2012)havealreadydiscussedtheaccuracyofthestellarmassesti-
power-lawSED.Type2AGNthataredetectedinthesoftbandonly matesforboththeType1andType2AGNsamplesandwhencon-
havebeendiscarded forthemajorityofthestackinganalysis due sideringthepropertiesoftheAGNasafunctionofstellarmass,we
to their essentially unknown dust columns and therefore intrinsic onlyconsider thoseAGNthathaverobuststellarmassesasspec-
luminosities. These soft band detected Type 2 AGN are however ifiedby theflags parameter. Asexpected, > 99.5 per cent of the
includedwhensearchingforAGNthatareindividuallydetectedat Type2AGNhaverobuststellarmassestimates.
850µminSection3.Thehard-bandX-rayfluxlimitisillustratedin
Fig.2andcorrespondstoLX &1043 erg/satz >1,wheretheX-
rayluminosityshouldbecompletelydominatedbyAGNactivity.
WealsorestrictourselvestoonlythoseAGNatz > 1forallthe 2.2.1 LimitationsoftheX-rayAGNCatalogue
stackinganalysis.Atz>1,the850µmfluxisessentiallyinvariant
Inthisanalysisweconsiderthesubmillimeterpropertiesofahard
withredshiftforagivendustluminosityandthereforestarforma-
X-ray flux limited catalogue of AGN. However, before proceed-
tionrate,allowingustoaveragethesubmillimeterfluxesofAGN
ing it is important to highlight the selection biases and incom-
overrelativelywideredshiftbinsinordertoimprovethestatistical
pletenessesofanAGNcatalogueselectedinthisway.Donleyetal.
significanceofourresults.The850µmdatadoesnothowevertrace
(2012) have used a mid infrared IRAC selection for AGN candi-
thepeakofthedustSEDattheseredshiftsandtherecouldpoten-
datesintheCOSMOSfielddemonstratingthatonly∼38%ofthese
tiallybebiasesinhoweffectivelythissubmillimeterfluxistracing
AGNcandidateshaveX-raycounterpartsduetotherelativelyshal-
starformationduetovariationsinthetemplateSEDsoverthisrest-
lowfluxlimitoftheXMM-COSMOShardX-raydata.Hence,we
framewavelengthrange.InSection4.1wewillthereforeexplicitly
maypotentiallybemissingasignificantpopulationofheavilyob-
checkthedependenceof850µmfluxonstarformationrates.
scured to mildly Compton thick AGN in this analysis. Recently
Finally,thereisaconcernthatthe850µmfluxcouldbecon-
Lacyetal.(2015)havealsousedmidinfra-redselectedAGNsam-
taminated by synchrotron emission for populations of radio-loud
plestodemonstratethatX-raysurveysbegintobecomeincomplete
AGN.ThefractionofradioloudAGNinX-rayselectedAGNsam-
atz &1.6whereadeficitintheX-rayluminosityfunctionisseen
plesisexpectedtobesmall(.5%forType1AGNfromHaoetal.
relativetothemidinfra-redluminosityfunctioneveninthehighest
2014).Nevertheless,wematchtheX-rayAGNsampletotheVLA
luminositybins.Wereferreaderstothesepapersforadetailedcon-
radio catalogue (Schinnereretal. 2010) and remove all sources
siderationoftheincompletenessinhardX-rayselectedsamplesof
with S1.4GHz >0.5 mJy. This removes 10 Type 1 and 7 Type 2
AGN at z > 1 from the sample, which represents ∼2% of both AGN.
theType1andType2samples2.Theradiofluxlimitimposedcor-
responds toa synchrotron 850µm fluxof <0.01 mJy assuming a
synchrotron spectral index of 0.7. Therefore we can be confident
2.3 Non-AGNGalaxyCatalogue
thatthe850µmstacksthatwestudyarecompletelydominatedby
thermal dust emission. Our final z > 1AGNsample usedinthe Finally,asacontrolsampletocomparetheAGNpropertiesto,we
stacking analysis, totals 699 AGN. These are comprised of 428 alsoutiliseagalaxysamplefromtheUltraVISTAsurveyDataRe-
Type1AGN(314 spectroscopic, 114photometric) and271 Type lease 1 (McCrackenetal. 2012) over the same region. We work
2AGN(46spectroscopic,225photometric). with the K-band flux limited sample from Muzzinetal. (2013),
Bongiornoetal.(2012)haveusedspectralenergydistribution which essentially represents a mass-limited sample of galaxies.
(SED) fits to the multi wavelength COSMOS photometry, in or- Corresponding photometric redshifts, stellar masses and other
der toestimatestellarmasses, star formationratesand other host SED-fitting parameters are also taken from Muzzinetal. (2013).
galaxyparametersfortheseType1andType2AGN.Bothgalaxy An advantage of using a galaxy catalogue that is selected in the
andAGNcomponentsarefitallowingustostudythedependence redderK-band,isthatthesecataloguesarepotentiallymoresensi-
tivetodustygalaxiesthatwouldbedetectedat850µmcompared
togalaxycataloguesconstructedusingdataatopticalwavelengths.
2 Including these radio-detected AGN in the 850µm stacking does not Wenotethatseveralpreviousstudiesthatwecompareourresults
changetheaverage850µmfluxderivedforoursampleinSection4towithin tointhispaper,haveusednon-AGNgalaxysamplesselectedusing
0.1mJy flux-limitedsurveysatbluer(optical)wavelengths.
(cid:13)c 0000RAS,MNRAS000,000–000
Cold DustEmissionfromX-RayAGNin S2CLS 5
Type 1 Type 2
45.5 45.5
45 45
44.5 44.5
−1s) 44 −1s) 44
g g
er 43.5 er 43.5
L / 2−10 43 L / 2−10 43
g(1042.5 g(1042.5
o o
l l
42 42 Spectroscopic
Spectroscopic Photometric
41.5 Photometric 41.5
41 41
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
Redshift Redshift
Figure2.Redshiftversusobservedhard(2-10keV)bandluminosityforType1AGN(left)andType2AGN(right).Bothspectroscopicandphotometric
AGNareshown.Thecirclesmarkactualdetections whereasthedownwardtrianglesrepresentupperlimitsinthecaseoftheType2AGNandhard-band
luminositiesestimatedfromthesoft-bandfluxinthecaseoftheType1AGN.Thesolidlinesdenotethefluxlimitappliedtothedatatogeneratehomogenous
populationsofthetwopopulationsforourstackinganalysis.Theverticaldashedlinescorrespondtoz >1whichistheredshiftcutappliedbeforestacking
suchthatthe850µmfluxisinvariantwithdustluminosityasafunctionofredshift.
3 850µmDETECTEDAGN 1%) of being chance associations (see Table 1). The 19 sources
arecomprisedof6spectroscopicallyconfirmedbroad-lineAGN,1
We begin by looking for AGN that are detected at >3.5σ in the
spectroscopically confirmed narrow-line AGN, 3 AGN withpho-
SCUBA-2CLS850µmcatalogues. Asstatedabove, onlyregions
tometric redshifts that are best fit by Type 1 AGN templates and
with σ850 < 2mJy were used in the construction of the 850µm 9 AGN with photometric redshifts that are best fit by Type 2
catalogue,whichthereforedoesnotcovertheentirefield.Amatch-
AGN or galaxy templates. Thus out of a sample of 360 850µm
ingradiusof7.5′′ isusedtoassociate850µm cataloguedsources bright sources, we find that ∼5% are associated with X-ray lu-
withourAGNsample.TheAGNpositionsinallcasescorrespond
minous AGN. The XMM-COSMOS data used in this work cor-
totheopticalpositionsfromBrusaetal.(2010),whichhavemuch
responds to relatively bright AGN X-ray flux limits as can be
smalleruncertaintiesthantheXMM-NewtonX-raypositions.Given
seeninFig. 2. Wangetal. (2013) identify theX-raycounterparts
theerrorsontheopticalpositionsarelikelytobenegligible,wefol-
toALMA detectedsubmillimeter bright galaxiesintheExtended
lowIvisonetal.(2007)andassumethatforasignal-to-noiseratio
Chandra Deep Field South (E-CDFS). Chandra observations are
(SNR)of3.5,thepositionaluncertainty,σpos=0.6×FWHM/SNR. usedtogodowntoX-rayluminositiesof∼7×1042 erg/sandthe
For a FWHMof 14.5′′ at 850µm, σpos is therefore 2.5′′ and the authors find a larger AGN fraction of ∼17% among the submil-
7.5′′ matching radius corresponds to 3σpos and is therefore ex- limeter galaxies. However as can be seen in their figure 10, this
pectedtoencompass>99.7%oftruecounterparts.
AGN fraction decreases as a function of increasing X-ray lumi-
Thereareatotalof19850µmsourcesthatliewithin7.5′′ of
nosity.DowntosimilarX-rayluminositiesasWangetal.(2013),
anX-rayAGNandtherearenoinstancesofmultiplesubmillimeter
Symeonidisetal.(2014)findanAGNfractionof18%amongin-
sourcesmatchedtothesameX-rayAGN.All19matchesarepre-
fraredluminous galaxies intheCDF-N andCDF-Sfieldsat z <
sentedinTable1andweestimatethecorrectedPoissonprobability,
1.5. Alexanderetal. (2005) have used ultra-deep X-ray observa-
dpeconrsritoyfotfheaslelXbe-rinaygAchGanNcethaastshoacviaetiflounxsebsabseridghotnerththeabnatchkegrtoarugnedt tions inthe CDF-N tostudy the AGN fraction in S850 & 4 mJy
sub-mm bright galaxies at z > 1, finding that the majority host
beingconsidered.Specifically:
AGNactivity.However,theX-raydatafromChandrausedinthat
work, is an order of magnitude fainter than our XMM-COSMOS
praw =1−exp πr2N(>S) sample. In Fig 3 we plot the redshift versus hard (2-10 keV) X-
pdet =1−exp πr2ma(cid:0)xN(>Slim)(cid:1) (1) arasywleulmlainsoAsilteyxafnodretrheetsaul.b(m2i0l0li5m)eatnerddWetaencgteedtaAl.G(N201in3)t.hWisewaolsrko
pcorr =1−exp praw 1(cid:0)+ln praw (cid:1) showherethesubmillimeterdetectedX-rayabsorbedquasarsfrom
(cid:18) (cid:18) (cid:18)pdet(cid:19)(cid:19)(cid:19) Stevensetal.(2005),whichwereselectedusingX-raysurveyswith
where r is the separation between the submillimeter source and ROSAT, XMM-Newton and Chandra and followed up at submil-
the X-ray AGN, rmax is the maximum search radius and N(>S) limeter wavelengths using SCUBA. The Wangetal. (2013) ob-
and N(>Slim) represent the background surface density of AGN served 0.5-8 keV luminosities have been converted to 2-10 keV
brighter than the source being considered and brighter than the luminositiesassumingtheintrinsicphotonindicesderivedforeach
flux limit of the survey respectively. We estimate N(>S) us- object in that paper whereas conversions from 0.5-2 keV fluxes
ing the source counts in the soft and hard X-ray bands from inStevensetal. (2005) assume Γ = 1.8. Our work can immedi-
Cappellutietal. (2009) and assume S =5e-16 erg/s/cm2 and atelybeplaceinthecontextofthesepreviousstudies.Whileboth
lim
2.5e-15 erg/s/cm2 in the 0.5-2 keV and 2-10 keV bands respec- Alexanderetal.(2005)andWangetal.(2013)probesignificantly
tively. Given the low surface density of X-ray AGN and submil- fainterintermsofX-rayluminosity,oursamplefillsintheluminos-
limeter sources, all 19 matches have very small probabilities (. itygapbetween thesestudiesandtheindividual follow-upobser-
(cid:13)c 0000RAS,MNRAS000,000–000
46
Alexander+05
Wang+13
45.5
Stevens+05
This Work
45
1)
−
s
g 44.5
er
/ 0 44
1
−
L2
(043.5
1
g
o
l
43
42.5
42
0.5 1 1.5 2 2.5 3 3.5 4
Redshift
Figure3.RedshiftversusobservedhardX-ray(2-10keV)luminosityfor
theX-rayAGNthatareindividuallydetectedatsubmillimeterwavelengths
in this work and compared to individually detected submillimeter bright
X-rayAGNfromAlexanderetal.(2005)andWangetal.(2013)andsub-
millimeterdetectedquasarsfromStevensetal.(2005).Downwardtriangles
denoteupperlimits.ThehardX-rayfluxlimitandz>1redshiftlimitap-
pliedpriortoourstackinganalysis,aredenotedbythedashedlines.Both
Alexanderetal. (2005)and Wangetal. (2013)probe fainter X-ray lumi-
nosities comparedtoourwork,whiletheX-rayAGNfromStevensetal.
(2005)thatwereindividuallyfollowedup,probeveryhighX-rayluminosi-
ties.Ourstudyisusefulinbridgingthegapbetweenthesedifferentdatasets.
vationsofX-raybrightquasarsbyStevensetal.(2005).TheX-ray
brightest sub-mm AGN in our sample isbrighter than any of the
X-rayAGNdetectedinthesub-mmbyAlexanderetal.(2005)and
Wangetal. (2013). It corresponds to XID18 and is an extremely
redAGNwithevidenceforoutflowinggaswithvelocitiesof∼300
km/s (Brusaetal. 2015). This class of objects will be discussed
laterinSection4.4.
Herschelfluxesforall850µmdetectedsourcesareextracted
fromthecataloguesofSwinbanketal.(2014)bymatchingtheop-
ticalpositionsoftheAGNinourcataloguetothe24µmpositionsin
theHerschelcatalogues,usingamatchingradiusof3′′.TheHer-
schel catalogues have already been corrected for blending using
the 24µm positional priors (Swinbanketal. 2014). These fluxes
arepresentedinAppendixBtogetherwithotherderivedproperties
forthesesubmillimeterbrightX-rayAGN.AstheHerschelfluxes
havebeenextractedbySwinbanketal.(2014)usingthe24µmpo-
sitionalpriors,all24µmdetectedsourceshavemeasuredHerschel
fluxeswhereassourcesnotdetectedat24µmarenotintheHerschel
catalogues.
Table1.SummaryofX-rayAGNwithpotential850µmbrightidentificationsintheCOSMOSfield.ThecorrectedPoissonprobabilityofachanceassociation
isquotedascalculatedfromthesoftX-rayfluxandthehardX-rayfluxrespectively.Sourceswithphotometricredshiftsaremarkedwithapz.Sourceswith
hardX-rayluminositiesmarkedwithanasteriskareType1AGNthatareonlydetectedinthesoftbandandwherethehardbandluminosityhasbeencalculated
assumingΓ=1.8.
XID RA Dec Redshift log10(L2−10) Separation S0.5−2keV p0.5−2keV S2−10keV p2−10keV S850
ergs−1 ′′ ergs−1cm−2 ergs−1cm−2 mJy
13 150.00924 2.2755119 0.850 43.95 3.42 2.11e-14 0.0006 2.9e-14 0.0009 4.5±1.1
18 150.13304 2.3032850 1.598 44.92 5.92 1.80e-14 0.002 6.04e-14 0.0009 4.4±1.2
139pz 150.04184 2.6294839 0.739 43.69 3.30 9.52e-15 0.001 2.22e-14 0.001 7.4±1.7
160 150.15839 2.1396031 1.825 44.28 1.48 2.39e-15 0.001 1.02e-14 0.0008 8.1±1.2
246pz 150.05100 2.4938550 2.342 44.56 3.65 8.81e-16 0.008 1.09e-14 0.003 6.4±1.7
250 150.06455 2.3290511 2.446 44.14∗ 3.04 2.41e-15 0.004 <3.78e-15 – 4.0±1.1
270pz 150.10612 2.0144781 1.883 <44.13 1.83 1.17e-15 0.003 <6.64e-15 – 9.8±1.7
278pz 150.09308 2.1014100 2.269 <44.46 2.61 1.46e-15 0.004 <9.39e-15 – 5.0±1.4
353 150.08438 2.2904881 1.112 43.70 2.04 2.34e-15 0.002 8.57e-15 0.002 4.7±1.0
402pz 150.25226 2.2619131 1.078 43.57 2.86 2.22e-15 0.003 6.82e-15 0.003 5.7±1.5
415pz 149.96981 2.1834869 1.558 43.33∗ 2.90 1.04e-15 0.005 <8.31e-15 – 5.5±1.4
469pz 150.09685 2.0215000 3.362 <44.45 2.22 1.08e-15 0.004 <3.73e-15 – 6.6±1.7
10675 150.19262 2.2198489 3.090 43.92∗ 0.28 8.45e-16 0.0001 <1.73e-15 – 5.7±1.2
10809pz 150.20758 2.3816111 1.288 <44.15 1.79 1.12e-15 0.003 <1.7e-14 – 6.5±1.5
30182pz 150.12541 2.6978931 0.742 43.23 6.00 6.26e-15 0.005 7.52e-15 0.009 6.9±1.7
53922 150.09454 2.7029131 0.850 43.38 1.81 2.25e-15 0.002 7.74e-15 0.001 8.6±1.8
54440pz 150.06308 1.9446778 1.947 43.91 4.39 1.13e-15 0.009 <3.74e-15 – 7.7±1.7
60070pz 150.22719 2.2324781 2.066 <44.01 1.93 7.33e-16 0.003 <4.11e-15 – 5.8±1.4
60490pz 150.10544 2.1852681 1.089 43.42 5.33 <7.57e-16 – 4.76e-15 0.011 6.9±1.1
4 AVERAGE850µmEMISSIONFROMSTACKED
SUB-SAMPLES
ThemajorityoftheX-rayAGNarenotindividuallydetectedinthe
850µm catalogue, which goes down to a 3.5σ flux limit of ∼3.5
mJy in the highest sensitivity regions. To understand the average
submillimeter properties of the AGN we therefore have to study
stacked 850µm maps. We begin by constructing an inverse vari-
ance weighted stacked image at 850µm for all 699 z > 1 AGN
(428Type1and271Type2)thatwillbeusedfromhereoninour
stackinganalysis.Specifically:
Sij = ΣNkΣ=1NkP=1ijP×i,jf/ik(jσ/ik(jσ×ikjσ×ikjσ)ikj) (2)
whereSijrepresentsthestacked850µmimage,Pij isthepointre-
sponsefunctionofSCUBA-2at850µmgeneratedbystackingall
>10σ sourcesinthemap,fij arethe40×40pixel(80×80′′)flux
imagesofeachoftheN sourcesgoingintothestacksandσij are
thecorrespondingRMSimages.Stackedimagesgeneratedinthis
way can be seen inFig. 4 for both the AGN sample and a set of
random points in the map, where any random points within7.5′′
ofacatalogued850µmsource,havebeenremoved.Fig.4clearly
demonstratesthepresenceofastatisticallysignificant850µmsig-
nalfromthez > 1X-rayAGNrelativetotherandompoints.The
typicalRMSnoiseinthemapsiswellabovetheconfusion limit.
Asstatedin Section2.1, the noiseisnon-uniform over thefull 2
deg2 map. Out of the699 z > 1 sources being stacked, 80% lie
in regions with RMS noise <3.5 mJy. From the 850µm source
counts presented in Caseyetal. (2013), we estimate that the in-
tegral source count at thisRMS noise level isonly ∼630 deg−2.
Assuming a beam FWHMof 14.5′′ for SCUBA-2at 850µm, the
beam density is∼78,000 deg−2. Hence only 1in 120 beams are
expectedtobeaffectedbyconfusionatthisRMSnoise.Inthelow-
estRMSregionsinourmap(σ=1mJy),theintegralsourcecounts
reach∼3000deg−2 and1in25beamsareaffectedbyconfusion,
whichisstillasmalleffect.
Before investigating the dependence of the 850µm emission
ontheAGNproperties,wefirstdescribeourmethodologyforcal- Figure4.80×80′′ weighted,stacked850µmimagescentredontheposi-
culating the stacked 850µm fluxes. Throughout this analysis the tionsofthe699z > 1X-rayAGNinourfluxlimitedsample(top)and
stacked 850µm flux ismeasured asthe median or mean value of centredonexactlythesamenumberofrandompoints(bottom).TheAGN
the central pixel in 100 bootstrapped inverse variance weighted stackincludes428Type1AGNand271Type2AGN.Thecolourscaling
isthesameinbothimages.Thecontoursareat1-5σandthesolidcontours
stacked images (see Eq. 2). We have already checked in Section
correspond to positive fluxes whereas thedashed contours correspond to
2.1thattheastrometryoftheSCUBA-2mapisaccuratetowithin
negativefluxes.Weclearlyobserveastatisticallysignificant850µmdetec-
apixelsothiscentralpixelfluxprovidesthemostconservativeand
tionfromtheAGNrelativetotherandompoints.
unbiasedestimateoftheaverageflux.Thecentralpixelfluxismea-
suredtobe0.71±0.08mJyforall699AGN.Themedianredshift
ofthesampleisz = 1.66andthemedianhardX-rayluminosity
islog10(L2−10)=44.2.WeemphasisethattheseX-rayluminosi- wefindthattheaveragefluxfallsto0.60±0.07mJy,whichisstill
tiescorrespondtothemostluminousAGNinthe“quasar”regime. a>8σdetection.
Forcomparison, theaverage870µmfluxoftheLutzetal.(2010) Wehavecheckedthatthemedianandmeanstackedfluxesare
AGN is 0.49±0.04mJy. The Lutzetal. (2010) sample has both a verysimilarinbothcases.Theerroronthefluxiscalculatedusing
lowermedianredshift(z = 1.17)andalowermedianhardX-ray 100bootstrapsamples.Wealsocalculatestackedfluxesusingthe
luminosity-log10(L2−10)=43.6calculatedfromthemedianhard publiclyavailablecodeSIMSTACK(Vieroetal.2013).Thecodeis
X-rayfluxinthatpaperandusingourassumedvalueofΓ = 1.8 designed tocorrectlyaccount for clusteringof sourceswithinthe
asdescribedinSection2.2.Giventhesedifferencesinthemedian scale of the SCUBA-2 beam which is assumed to be 14.5′′. The
propertiesofthetwosamplesaswellasthecorrectionfrom870µm methodessentiallyinvolvesperformingaregressionofthetrueflux
to850µmforatypicalsingletemperaturegreybody SEDandthe mapwitha‘hits’map,wherethe‘hits’mapiscreatedbycounting
statisticalerrorsinbothmeasurements,ourresultsarebroadlycon- thenumberofsourcestobestackedthatcontributetoeachpixelin
sistentwiththoseofLutzetal.(2010).Excludingtheindividually themap.Westackthe699flux-limitedz>1AGNwiththez<1
detected850µmX-rayAGNdescribedinSection3fromthestacks, AGN.Ifthephotometricredshiftsareaccurate,thereshouldbeno
clusteringbetweenthehighandlowredshiftAGNpopulations.The Inluminous,unobscuredType1AGN,thebigbluebumptypically
stacked850µmfluxforthe699z >1AGNis0.82±0.02mJyfrom seeninquasarspectramaymimictheUVbumpfrommassivestars
SIMSTACKwhichisslightlyhigherthanourcentralpixelfluxmea- sostarformationratesbasedonSEDfittingarenolongerreliable.
surement.Thehigher SIMSTACKfluxislikelyduetothefactthat HenceType1AGNarenotconsideredinthistest.
thepeak850µmemissioninFig.4isoffsetfromthecentralpixel. We split the Type 2 AGN into two SFRSED bins at
The SIMSTACKerrorsreflect the regression errorsand do not ac- SFR>10M⊙yr−1andSFR<10M⊙yr−1.Themedianredshiftsare
count for anyerrorsduetosamplevariance. Our bootstraperrors z =1.36andz=1.77forthelowandhighSFRSEDsub-samples
are therefore more conservative. The SIMSTACK results demon- respectively.Themedianstarformationratesare1.5M⊙yr−1 and
stratethatthecentralpixelfluxrepresentsaconservativeestimate 33M⊙yr−1 while the mean star formation rates are 1.1M⊙yr−1
and is not biased up due to clustering. For the rest of the analy- and 46M⊙yr−1 in the two bins. The average 850µm fluxes for
sis,wethereforeusethecentralpixelfluxtomeasuretheaverage thesesub-samplesare0.3±0.1mJyand1.0±0.2mJyrespectively
submillimeter emission for all the sub-samples we construct. As and are therefore clearly higher in the higher SFR bin. The ex-
afinalcheck,giventhatsomeregionsofthe850µmwideS2CLS cess 850µm flux in the high SFR bin is significant at the ∼3σ
COSMOSmapsaresignificantlynoisierthanothers(Section2.1), level.Severalpreviousstudieshaveinsteadusedthe250µmfluxes
wealsorestrictourstacksonlytotheregionswithσ850 <2mJyto asaproxyforSFR.WethereforealsocalculateaverageHerschel
mitigateagainsttheeffectsoffluxboosting.Thereare341z>1X- 250µmfluxesfortheType2AGNinthesetwoSFRbinsalthough
rayAGNinourflux-limitedsampleinthesehighsensitivityregions we caution that the bins encompass AGN over a relatively broad
andthemeanfluxisS850=0.69±0.10mJy,consistentwiththere- range of redshifts, which could be adding scatter to the 250µm
sultsderivedusingtheentire850µmwidemap.Wewilltherefore fluxes. The250µm fluxesarecalculated byinjectingtheAGN in
proceed withusing the 850µm data over thefull COSMOSwide each SFRbin in turninto the Herschel SPIRE250µm map from
areaforallthestackinganalysisthatfollows. Swinbanketal.(2014),whichhashadall24µm detectedsources
Beforecontinuing,itisimportanttohighlighttheassumptions removed. The maps are then stacked at the AGN position and
inherent in our work. Throughout this paper, we assume that the the average flux is once again measured as the value at the cen-
850µm flux is tracing the total amount of cool dust emission in tral pixel. Note that we have also checked the astrometry of the
theAGNhost galaxiesandthat thiscool dust isprimarilyheated 250µm map in Section 2.1. The 250µm fluxes in these two SFR
bystarformationandhasanegligiblecontributionfromtheAGN. bins are 2.0±0.3 mJy and 5.3±0.6 mJy respectively. In order to
Giventhatthe850µmfluxisinvariantwithredshiftatz > 1,we check whether the 250µm and 850µm stacks are consistent with
canaveragethesefluxesoverrelativelybroadredshiftbinswithout each other, we fit a single temperature greybody with β=2.0 to
bias. Note that this is not the case for Herschel fluxes for exam- thesedata points inthe twoSFRbins. Thedust temperatures are
ple,whichtracethepeakofthedust SEDatz ∼ 1−3,andare 24±6Kand27±5Krespectively,consistent withasingletemper-
expectedtobevaryingmorestronglywithredshift.Aswedonot atureinbothSFRbins.Fromtheintegralunderthegreybody, we
havelargeenoughAGNsamplestobeabletosplitourAGNinto canthencalculatethefarinfraredluminosity(between60–300µm
narrow redshift bins, werestrictour analysistothe850µm prop- toeliminateanycontributionsfromtheAGN)andaveragestarfor-
erties only rather than constructing full SEDs from the Herschel mation rate for each stack. These are log10(LFIR/L⊙)=10.9±0.3
and 850µm data. The 850µm emission can also only be used as correspondingtoSFR=12±161M⊙yr−1 forthelowSFRstackand
a proxy for the star formation rate provided the dust temperature log10(LFIR/L⊙)=11.5±0.2correspondingtoSFR=60±3200M⊙yr−1
is not varying significantly between AGN sub-samples, and also forthehighSFRstack.Inbothcasesthestarformationratesde-
assuming thatthedust heatingatthesewavelengths isdominated rivedfromtheHerscheland850µmdataarehigherthanthemean
byrecentstarformationratherthanbeingdominatedbycolddust starformationratesfromBongiornoetal.(2012)althoughtheer-
heatedbyoldstarsandtheinterstellarmedium(e.g.Bourneetal. rorbarsandscatterinbothaverageSFRmeasurementsarelarge.
2013). We therefore begin by testing how the 850µm flux varies In Appendix B where we have shown best-fit far infrared SEDs
withstarformationrate. for our X-ray AGN that are individually detected at 850µm, we
alsoshow that thefar infraredluminosity and star formation rate
asmeasuredfromthesebest-fitSEDs,correlateswellwithboththe
4.1 StarFormationRate
250 and 850µm fluxes (Fig. B2). In Appendix B, the dust mass
Bongiornoetal. (2012) have estimated the star formation rates fromfullSEDfittinghasalsobeenplottedagainstboththe250and
for the COSMOS AGN using SED-fitting to the available multi- 850µmfluxesandingeneralshowsaweakercorrelationwiththe
wavelength photometry (Brusaetal. 2010) - SFR hereafter. infra-redandsub-mmfluxdensitiescomparedtothefarinfra-red
SED
Thesecanbeusedtocheckthedependenceofthe850µmfluxden- luminosity.Wewillthereforeproceedbyassumingthatthestacked
sityon star formation rate. As described inSection 2.2, theSED 850µmfluxinourz >1AGNisgivingusadirectestimateofthe
basedstar-formationratesincludephotometryintheSpitzerIRAC averagestarformationrate,withthecaveat thatforgalaxieswith
aswellasMIPS24µmbandsinmostcases.Insomecases,theSED lowlevelsofstarformation,theremayalsobeacontributiontothe
fitsalsoincludelonger wavelengthphotometryatMIPS70µmas dustheatingat850µmfromoldstarsandtheinterstellarmedium.
well as at Herschel 100 and 160µm. We select only the Type 2
z > 1 AGN with robust star formation rate estimates as speci-
4.2 Redshift
fied by the flags parameters in Bongiornoetal. (2012). We note
that soft X-raydetected Type 2AGN that are not detected inthe Thesizeofourcurrentsamplenecessitatesstackingoverrelatively
hardbandarealsoincludedinthisanalysistoimprovethestatis- broadredshiftbins.Evolutionaryeffectscouldthereforeinfluence
tics. Hereweare primarilyconcerned withchecking whether the any derived results and introduce differences in stacked 850µm
850µmfluxescorrelatewithotherindependently derivedstarfor- fluxes between different sub-samples. Before looking for trends
mationrateconstraintsandknowledgeofthetruedustcolumnsand withotherAGNpropertieswethereforefirstconsidertheredshift
therefore intrinsicX-rayluminositiesof the AGN isnot relevant. evolutionofthe850µmfluxesoverthefullredshiftrangeconsid-
1.2
Type 1 1.6 Type 1
1.1 Type 2
Type 2
1 1.4
0.9
1.2
y 0.8 y
J J
m m 1
/ 00.7 / 0
5 5
S80.6 S80.8
0.5
0.6
0.4
0.4
0.3
0.2 0.2
1.2 1.4 1.6 1.8 2 2.2 2.4 43.8 44 44.2 44.4 44.6 44.8
Redshift log (L / erg s−1)
10 2−10
Figure5.Average850µmfluxversusredshiftforbothType1andType2 Figure6.Average850µmfluxversushardX-rayluminosityforbothType
AGN. 1andType2AGN.
eredinourstackedanalyses.Inordertodothis,wesplitthez >1 AGNinthreeX-rayluminositybins-43.5< log10(L2−10) < 44,
Type1andType2AGNintothreeredshiftbinsandconsidertheir 44.0<log10(L2−10)<44.4andlog10(L2−10)>44.4.TheAGN
stacked 850µm fluxes. The results are illustrated in Fig. 5. Both that are individually detected at 850µm (see Section 3) are also
theType1andType2AGNsubmillimeterfluxesshowlittleevolu- included inthestack.Thereareonly 7X-rayAGNout of the19
tionwithredshiftovertheredshiftrangeprobedinthiswork.Inall thatsatisfyboththehard X-rayfluxlimitand thez > 1redshift
threeredshiftbins,theType1andType2AGN850µmfluxesare cut and we have checked that inclusion of these AGN does not
consistentwitheachothergiventheerrorbars.Whenconsidering change the stacked fluxes quoted withinthe error bars. Note that
trends in AGN submillimeter fluxes with luminosity, obscuration thelowestX-rayluminositybinisincompleteatz & 2sothere-
andAGN activityinthefollowing sections, wewillthereforeas- sultshereshouldbeinterpretedwithcaution.Theaverage850µm
sume that the 850µm fluxes do not evolve with redshift over the fluxesarepresentedinTable2andillustratedinFig.6.Whilethe
redshiftrangeconsidered. Type1submillimeterfluxappearstobefairlyindependentofX-ray
luminosity, there is some evidence that the most X-ray luminous
Type2AGNhavehighersubmillimeterfluxes.TheType2AGNat
4.3 X-rayLuminosity
log10(L2−10)>44.4have850µmfluxesthatare∼2σhigherthan
Wenowexploretrendsinthese850µmfluxeswiththehardX-ray in the previous X-ray bin although the numbers of Type 2 AGN
luminosity, which can be used as a proxy for the energy emitted at such high X-ray luminosities are small so the results could be
bythecentralaccretingpowersource.Severalrecentstudieshave affectedbysmallnumberstatistics.
foundareasonablyflatrelationshipbetweenAGNluminosityand In the highest X-ray luminosity bin, the AGN are expected
star formation rate (e.g. Harrisonetal. 2012; Rosarioetal. 2012; to have a relatively narrow Eddington ratio distribution based on
Azadietal.2014;Stanleyetal.2015).AtmodestX-rayluminosi- themodelsofAirdetal.(2013).TheseAGNthereforelikelycorre-
tieswherelargesamplesofX-rayAGNwithfarinfraredandsub- spondtohigh-massblack-holeshostedinmassivegalaxies.IfAGN
millimeterdatanow exist,theseresultsarenow robust.However, werepreferentiallyfoundinstar-forminggalaxiesonthemainse-
the situation is less clear at very high X-ray luminosities. While quence, thesemassivegalaxieswouldbeexpectedtohavehigher
Pageetal.(2012)havesuggested that thefar infra-redfluxesand starformationrates.ThehighX-rayluminositybinsmaytherefore
therefore star formation rates of AGN withhigh X-ray luminosi- bedominated byalargerfractionofquiescent hostsfor theType
ties,aresuppressedrelativetoAGNoflowerluminosity,Lutzetal. 1 AGN, which would naturally explain the lower 850µm fluxes.
(2010)andRosarioetal.(2012)insteadfindanincreaseinstarfor- X-rayluminous Type1AGNcould alsorepresent askewed pop-
mationrateathighX-rayluminosities,withtheresultsbeingmore ulation of highly accreting black-holes seen during a brief phase
pronouncedforlowredshiftAGNinRosarioetal.(2012).There- of black-hole growth. Assuch, their host galaxies maybe under-
sultsfromrecenthydrodynamicalsimulationssuggestthatthereisa massive relative to their black-hole masses, and this could once
largescatterintheL -SFRrelationoverthedynamicrangeprobed againexplainthelowerdustluminositiesandthereforestarforma-
X
by current surveys (Sijackietal. 2014),which may explain some tionratesinthesesources.
ofthediscrepanciesbetweenpreviousstudies. The Type 2 AGN show a more marked increase in average
The SCUBA-2 photometry presented in this paper now al- submillimeterfluxwithX-rayluminositywiththemostX-raylu-
lowsanindependentinvestigationoftherelationbetweenstarfor- minousAGNhavinghigherstarformationrates.Ascanbeseenin
mation and AGN luminosity with the advantage that the submil- Fig.2,foraflux-limitedsample,themedianredshiftincreaseswith
limeterfluxislesscontaminated byAGNemissionthantheHer- increasingX-rayluminosity. However,wehaveshowninSection
schel fluxes in the case of the most luminous quasars. As illus- 4.2thatthe850µmfluxdoesnotevolvesignificantlywithredshift
tratedinFig.2,atz > 1,theCOSMOSfieldisonlycompleteat in the Type 2 AGN population. Therefore the trend observed in
relatively high X-ray luminosities. We select Type 1 and Type 2 Fig.6isunlikelytobedrivenbyredshift.Evidenceofhigherstar