A Connection between Obscuration and Star Formation in Luminous Quasars

1.1 MB·English

by  Chien-Ting J. Chen

ACCEPTEDFORPUBLICATIONINAPJ
PreprinttypesetusingLATEXstyleemulateapjv.5/2/11
ACONNECTIONBETWEENOBSCURATIONANDSTARFORMATIONINLUMINOUSQUASARS
CHIEN-TINGJ.CHEN1,RYANC.HICKOX1,STACEYALBERTS2,3,CHRISM.HARRISON4,DAVIDM.ALEXANDER4,ROBERTOASSEF
5,MICHAELJ.I.BROWN6,AGNESEDELMORO4,WILLIAMR.FORMAN7,VAROUJANGORJIAN8,ANDREWD.GOULDING7,9,KEVIN
N.HAINLINE1,CHRISTINEJONES7,CHRISTOPHERS.KOCHANEK10,STEPHENS.MURRAY11,ALEXANDRAPOPE2,EMMANOUEL
ROVILOS4,ANDDANIELSTERN8
AcceptedforpublicationinApJ
5
1 ABSTRACT
0 Wepresentameasurementofthestarformationpropertiesofauniformsampleofmid-IRselected,optically
2 unobscuredandobscuredquasars(QSO1sandQSO2s)intheBoötessurveyregion. Weuseaspectralenergy
distribution(SED)analysisforphotometricdataspanningopticaltofar-IRwavelengthstoseparatetheAGN
n
a andhostgalaxycomponents. We findthatwhencomparedto a matchedsampleof QSO1s, the QSO2shave
J roughlytwicethehigherfar-IRdetectionfractions,far-IRfluxesandinfraredstarformationluminosities(LSF).
0 Correspondingly,weshowthattheAGNobscuredfractionrisesfrom0.3to0.7between(4- 40)×1011L⊙.IWRe
2 alsofindevidenceassociatingX-rayabsorptionwiththepresenceoffar-IRemittingdust.Overall,theseresults
areconsistentwithgalaxyevolutionmodelsinwhichquasarobscurationisassociatedwithadust-enshrouded

] starburstgalaxies.
A
Subjectheadings:galaxies: active— quasars:general— galaxies: starburst— infrared: galaxies— X-rays:
G
galaxies
.
h
p 1. INTRODUCTION adusty“torus”surroundingtheSMBH(e.g.Urry&Padovani
-
1995; Antonucci 1993). This model predicts no difference
o Quasars, the most luminous active galactic nuclei (AGNs
r 12), have been linked to galaxies with active star formation in host galaxy properties between obscured and unobscured
t AGNs. To date, it is still a matter of debate whether the
s (SF) ever since the discovery of their observationalconnec-
a tion to ultra-luminous infrared galaxies (ULIRGs, galaxies obscuration in luminous quasars can be explained solely by
[ the orientation-based unification model or if it is also en-
more luminous than 1012L⊙, e.g. Sandersetal. 1988). One hancedduetodustonlargerscalesthroughoutthehostgalaxy.
1 well-studiedscenarioformassivegalaxyevolutionpositsthat
Severalstudieshaveshownresultssupportinga scenariode-
v gas-rich galaxy major mergers trigger both rapid supermas-
parting from the unification model, such as the enhanced
9 sive black hole (SMBH) accretion and intense SF. This sce-
SFactivity(e.g.Canalizo&Stockton2001;Pageetal.2004;
5 narioassociatesthedust-enshroudedstarburstwithstrongnu-
Hineretal. 2009;Brusaetal. 2008;Shan&Chen2012) and
9 clearobscurationthatislaterexpelledbythepowerfulAGN,
the more disturbed structure (e.g. Lacyetal. 2007b) of the
4 implying an evolutionary link between unobscured (type 1)
hostgalaxiesofdust-obscuredquasarswhencomparedtoun-
0 and obscured (type 2) quasars (e.g. DiMatteoetal. 2005;
1. Hopkinsetal.2006;Gillietal. 2007;Somervilleetal. 2008; oalbssocushreodwqnutahsaatrso.bCscluusrteedrinAgGoNfsdiafrfeermenotrteypstersonogflAyGcNlussthearevde
0 Treisteretal.2009).
thanunobscuredAGNs(e.g.Hickoxetal.2011;Donosoetal.
5 On the other hand, the “unification model” of AGN as-
2014; DiPompeoetal. 2014). However, other studies have
1 cribesobscurationofAGNstodifferentlinesofsightthrough
foundno significant difference between obscured and unob-
:
v scuredAGNpopulationsintheirmorphologicalandSFprop-
1Department of Physics and Astronomy, Dartmouth College, 6127
i erties (e.g. Sturmetal. 2006; Zakamskaetal. 2006, 2008;
X WilderLaboratory,Hanover,NH03755,USA;[email protected].
2Department of Astronomy, Amherst, University of Massachusetts, Mainierietal. 2011; Schawinskietal. 2012; Merlonietal.
r Amherst,MA01003,USA 2014), and the host galaxy star formation rate (SFR) does
a 3StewardObservatory,UniversityofArizona,Tucson,AZ85721,USA not distinguish X-ray selected AGNs with different obscur-
4Department of Physics, Durham University, South Road, Durham, ing column densities (e.g. Rovilosetal. 2012; Rosarioetal.
DH13LE,UnitedKingdom
2012;Merlonietal.2014).
5NúcleodeAstronomíadelaFacultaddeIngeniería,UniversidadDiego
Portales,Av.EjércitoLibertador441,Santiago,Chile Nonetheless,themeasurementsofSFpropertiesforquasar
6SchoolofPhysics,MonashUniversity,Clayton3800,Victoria, Aus- host galaxies still suffer from selection biases that are often
tralia. differentamongvariousquasarpopulations.Inparticular,op-
7Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, ticalandX-rayselectedquasarsamplesmighthavedifferent
Cambridge,MA02138.
completenessinobscuredandunobscuredsourcesduetothe
8Jet Propulsion Laboratory, California Institute of Technology, 4800
OakGroveDr.,Pasadena,CA91109,USA attenuationofopticalandX-rayradiationbydustandgas. In
9Princeton University, Department of Astrophysical Sciences, Ivy fact, some studieshave suggestedthat opticalsurveysmight
Lane,Princeton,NJ08544,USA miss∼50%oftheAGNpopulationduetoboththeobscura-
10DepartmentofAstronomy,OhioStateUniversity,140West18thAv-
tionandhostgalaxycontaminationin low-luminosityAGNs
enue,Columbus,OH43210
11DepartmentofPhysics&Astronomy,TheJohnsHopkinsUniversity, (e.g. Goulding&Alexander 2009; Gouldingetal. 2010).
3400N.CharlesStreet,Baltimore,MD21218. While current X-ray observations probing photons with en-
12 Inthiswork,weuseAGNsasageneralnomenclatureforallenerget- ergyat∼10keVarebelievedtobelessaffectedbymoderate
i∼ca1ll0y45reelregvasn- t1,SwMeBreHfes.rtoFothreAmGaNssquwaistharsbo(QloSmOestr)i.c luminositylarger than levelsofobscuringmaterials(NH <1024cm- 2 ),asignificant
2 CHENETAL.
fractionofX-rayAGNs(∼20- 50%,e.g.Donleyetal.2005; relation between star formation and AGN accretion for the
Guainazzietal.2005;Parketal.2010;Alexanderetal.2011; mid-IR QSOs. In galaxy evolution models which suggest
Georgantopoulosetal. 2013; Wilkesetal. 2013) might still that the concurrent growth of AGN and host galaxy occurs
be missed due to obscuration that is Compton-thick (N ∼ during the dust enshrouded galaxy merger phase, different
H
1024cm- 2). AlthoughveryhardX-rayphotonscan penetrate types of quasars represent different stages of galaxy evolu-
Compton-thick obscuration, the previous high energy X-ray tion. To date, observational studies on the connection be-
surveys are limited to local sources due to the shallow flux tweenSFRandAGNaccretionratehavenotyetshowndefini-
limits (e.g. Swift/BAT, Burlonetal. 2011 and INTEGRAL, tive conclusions. Positive, strong correlations have been
Sazonovetal. 2012). While the NuSTAR (Harrisonetal. found in studies of optically selected type I quasars (e.g.
2013) mission opened a new window of high energy X-ray Serjeant&Hatziminaoglou 2009) and type II quasars (e.g.
upto80keV,therecentstudiesthatusedNuSTARtoobserve Netzer2009);whiledifferentconclusionshavebeenfoundin
heavily obscured AGNs have found that obscuration is still X-rayselectedquasarsathigherredshift(e.g.Silvermanetal.
a non-negligible effect even for X-ray photons at such high 2009;Mullaneyetal. 2012;Rosarioetal. 2012). The differ-
energy(Sternetal.2014;Lansburyetal.2014). ence between observed correlations could be driven by var-
In contrast, mid-IR observations of the reprocessed emis- ious limitations and observational constraints such as sam-
sion from the obscuring dust can detect heavily obscured ple size (Pageetal. 2012; Harrisonetal. 2012) or AGN in-
AGNs. A number of studies have shown that large pop- trinsic variability(e.g. Chenetal. 2013; Hickoxetal. 2014).
ulations of AGNs can be selected using the power-law In addition, it is also important to note that AGNs selected
SED shapeof AGNs at mid-IRwavelengths(e.g.Lacyetal. withdifferentcriteriaarehostedbyverydifferenthostgalaxy
2004; Sternetal. 2005; Hickoxetal. 2007; Donleyetal. populations (e.g. Hickoxetal. 2009; Griffith&Stern 2010;
2012; Assefetal. 2010b). Although these mid-IR color Gouldingetal. 2014) in which the galaxyand SMBH might
selection criteria cannot avoid some star forming galaxy followdifferentevolutionarypaths. Thesampleinthiswork
interlopers and might miss AGNs accreting at lower ac- consistsofalargenumberofluminousQSOsincludingboth
cretion rates (e.g. Hickoxetal. 2009; Donleyetal. 2012; the unobscured and heavily obscured populations, thus bi-
Mateosetal.2013;Hainlineetal.2014;Chungetal.2014)or asesduetosmallnumberstatisticsandtheexclusionofdust-
AGNswithcomplicatedsilicatefeaturesintheirmid-IRSED enshroudedAGNsare small. The biasinthe observedSFR-
(Kirkpatricketal. 2013), they are effective in selecting both AGN accretion rate relation in AGN host galaxies due to
obscuredandunobscuredAGNswithsimilarcompletenessin short-termAGNaccretionratestochasticitymayalsobeless
mid-IRwavelengths. Recentstudieshaveshownthatmid-IR significantin luminousquasars (Hickoxetal. 2014). There-
selectedAGNscanalsobeseparatedintoobscuredandunob- fore,the SFR-AGN accretionrate correlationforthe mid-IR
scuredpopulationsusingasimpleopticaltomid-IRcolorse- selected quasar populations may shed light on the origin of
lectioncriterion(e.g.Hickoxetal.2007;Donosoetal.2014; theconcurrentgrowthofSMBHandgalaxy.
DiPompeoetal.2014),whichcaneasilybeexplainedbythe This paper is organizedas follows: in §2 we describe the
differentlevel of extinction in the optical emission from the multi-wavelengthdata and the propertiesof the quasar sam-
nucleus (e.g. Hickoxetal. 2007, H07 hereafter). Therefore, ple.In§4,wediscusstheSEDfittingproceduresthatweused
a direct comparison of the host galaxy SF properties of ob- todisentangletheAGNandthehostgalaxycontributions. In
scured and unobscured sources in a mid-IR selected quasar §3, we explore details of the observed far-IR properties of
sample can provide insights to the origin of obscurations in QSO1sandQSO2s. A comparisonof the SF luminositybe-
rapidlyaccretingSMBHs. tweenQSO1sandQSO2sislaidoutat§5;Thex-rayproper-
In this work, we adopt the mid-IR selected quasar sam- tiesfortheQSO2sarediscussedin§6andtheAGNobscured
ple from Hickoxetal. (2011, H11 hereafter), which is com- fractionasafunctionofSFluminosityisdiscussedin§7. A
prisedofunobscuredandobscuredquasarswithsimilardistri- discussionandasummaryaregivenin§8.Throughoutthepa-
butionsinquasarproperties(e.g.redshiftandquasarluminos- per,weusetheVegamagnitudesystemandassumeaΛCDM
ity),thusmakingitanexcellentsampletostudytheconnec- cosmologywithΩm=0.3,ΩΛ=0.7andH0=70kms- 1.
tion between host galaxy properties and quasar obscuration.
2. THEQUASARSAMPLE
Thequasarsamplestudiedinthisworkconsistsof546unob-
scured quasars (QSO1s) and 345 obscuredquasars (QSO2s) Inthissection,wediscussthedatafromtheBoötessurvey
selectedusingSpitzermid-IRobservationsintheBoötessur- regionaswellasthequasarselectionandclassificationcriteria
veyregion.WeutilizetheopticalspectroscopyfromtheAGN weadopted.
andGalaxyEvolutionSurvey(AGES,Kochaneketal.2012)
2.1. Data
and the XBoötes Chandra X-ray observations(Murrayetal.
2005), along with the 250 µm data from the Spectral and The sample studied in this work comes from the 9 deg2
PhotometricImagingReceiver(SPIRE,Griffinetal.2010)on BoötessurveyregioncoveredbytheNOAODeepWide-Field
board the Herschel Space Observatory. With the inclusion Survey (NDWFS, Jannuzi&Dey 1999). Boötes is unique
of the far-IR photometry in which AGN contributions have amongextragalacticsurveysbecauseofitslargeareaandthe
been shown to be much smaller than that of the host galaxy excellentmulti-wavelengthcoveragefromspace-andground-
(e.g. Netzeretal. 2007; Lacyetal. 2007a; Kirkpatricketal. basedtelescopes,whichmakepossiblestatisticalstudyofthe
2012;Mullaneyetal.2012;Chenetal.2013;DelMoroetal. rareluminousAGNs.
2013; Drouartetal. 2014), we can use the wealth of multi- In this work, we use the multiwavelengthphotometrycat-
wavelength observationsin the Boötes field to obtain robust alog from Brown et al. (private communication), which the
measurements of SFR for luminous quasars and test if they same the one used in Chungetal. (2014). This catalog cov-
areassociatedwithobscuration. ers the optical to mid-IR bands including the NDWFS opti-
We also take advantage of this sample to study the cor- calobservationsinBw,R,I bands,near-IRphotometryfrom
the NOAO NEWFIRM survey (J, H and Ks, Gonzalezetal.
3
1.5 60
40
(a) QSO-1 (b)
20
QSO2
FIR-QSO 0
1.0 QSO1 FIR-QSO
QSO2
)
a
g 4
e
V
)
( a
5] 0.5 eg
4. V 6
]-[ ](
6 5
3. 4.
[ [
0.0 R- 8
10
-0.5
0 1 2 3 45.0 45.5 46.0 46.5 47.0 0 25 50 75
[5.8]-[8.0] (Vega) log L (erg/s)
AGN
FIG.1.—(a)IRACcolor-colordiagramshowingtheselectionofthequasarsamplesusingthecriteriaofSternetal.(2005).Thegray-scaleshowsthedensity
ofsourcesdetectedat>5σsignificanceinallfourbandsinSDWFSdata. BluestarsandredcirclesshowtheQSO1andQSO2samples,respectively. The
Sternetal.(2005)color-colorselectionregionisshownbythedashedline. Inaddition,theQSOswithfar-IRdetections(FIR-QSO)areenclosedwithorange
circles. Someofthemid-IRQSOsfalloutoftheselectionwedgeduetotheupdatedIRACphotometryandaperturecorrections(see§2.3). (b)Illustrationof
theoptical-IRcolor-selectioncriteriafordividingtheIR-selectedQSOsampleintounobscured(QSO1)andobscured(QSO2)subsamples. Shownisobserved
R- [4.5]colorversusbolometricluminosity,calculatedasdescribedin§4. ContoursshowthedistributionforalltheHickoxetal.(2007)IR-selectedquasars,
whilebluestarsandredcirclesshowtheQSO1andQSO2subsamplesat0.7<z<1.8usedinH11andthisanalysisasdescribedin§2.Thedistributioninthe
R- [4.5]colorandLAGNarealsoshownashistogramsinthesidepanels.QSO1sareshownasthebluesolidlines,QSO2sareshownasthereddashedlinesand
theFIR-QSOsareshownastheorangedash-dottedlines.Thecontoursandcolorhistogramsshowthatasimplecutinoptical-IRcolorclearlyseparatestheQSO
samplesintotwopopulations.
2010) and the Spitzer Deep Wide Field Survey (SDWFS, missedbythestandardpipelinereduction.Wealsoconvolved
Ashbyetal.2009)ofthe4bandsofmid-IRobservationsfrom the raw maps with a matched filter (see Chapinetal. 2011),
theSpitzerInfraredArrayCamera(IRAC)at3.6,4.5,5.8and whichaidedinsourceextractionbyloweringtheoverallnoise
8µm. In addition, the mid-IR photometry from the Spitzer andde-blendingsources. Fromthis,wegeneratedamatched
Multi-bandImagingPhotometer(MIPS)at24µmisalsoin- filtercatalogwitha5σdetectionthreshold.Inthiscatalog,we
cluded(IRSGTOteam,J.Houck(PI),andM.Rieke). Inthis considerSPIREsourceswithfluxeslargerthan20mJyasun-
catalog,theopticaltonear-IRphotometrywasextractedfrom ambiguously detected. Completeness simulations show that
a matched aperture for each band. For Bw, R, I, H and Ks thesecatalogsare95%completeintheinnerregionand69%
bands,thephotometrywasmeasuredfromimagessmoothed complete in the outer regions above a flux of 20 mJy. We
to a common point spread function (PSF) with a 1.′′35 full also find minimal flux boosting for low SNR sources above
widthhalfmaximum(FWHM);whileforJband,thephotom- thisfluxcutoff(seeS2.2inAlbertsetal.2013,forthedetails
etrywasmeasuredfromimagessmoothedtoacommonPSF of the completeness simulation). We match the positions of
with a 1.′′60FWHM. To ensure consistencyin the photome- the SPIRE catalog to the I-band positions with a matching
try,weuse6′′aperturephotometryfromopticalbandsthrough radiusof5′′. Wetestedtherateofspuriousmatchesbyoffset-
theIRACbandsinthemid-IRtoaccountforthelargeIRAC ting the SPIRE source positionsby 1′ in a randomdirection
beam size. For the MIPS data, we use PSF photometry in- andmatchingtherandomlyshiftedcatalogtotheBoötescata-
steadofaperturephotometryduetothestilllargerbeamsize log.Wefoundthatwitharadiusof5′′,ourmatchingbetween
of MIPS. The 5σ flux limits ofthe opticalto near-IRbroad- the SPIRE and Boötes catalog only yielded <2% spurious
bandphotometryare25.2,23.9,22.9,21.1,20.1,18.9(Vega matches.
magnitudes)fortheBw,R,IJ,H,Ksbands,respectively. For Inaddition,wealsomatchthepositionsoftheBrownetal.
themid-IRwavelengths,the5σfluxlimitsare6.4,8.8,51,50 (2007) catalog to the publicly available Wide-field Infrared
and170µJyforthe3.6,4.5,5.8,8.0and24µmbands,respec- SurveyExplorer(WISE)All-skycatalog(Wrightetal.2010)
tively. A extensive description of the multiband photometry and obtain the profile-fit photometrymagnitudes in the W1,
extractioncanbefoundinBrownetal.(2007). W2,W3andW4bands(3.4,4.6,12and22µm).
We also make use of the far-IR observations from the As a complementary measurement of the AGN accretion
publiclyavailableHerschelMulti-tieredExtragalacticSurvey rate and absorption by gas, we utilize X-ray data from the
(HerMES, Oliveretal. 2012). We re-reduced and mosaiced XBoötessurvey,whichisamosaicof126short(5ks)Chan-
the Boötes SPIRE observations (Albertsetal. 2013), which dra ACIS-I images (Murrayetal. 2005; Kenteretal. 2005)
include a deep ∼2 deg2 inner region near the center of the covering the entire NDWFS. XBoötes contains 3,293 X-ray
fieldandashallower∼8.5deg2outerregion.Wespecifically point sources with at least four counts in the AGES survey
focusedonremovingstriping,astrometryoffsets,andglitches region. The conversion factors from count rates (in counts
4 CHENETAL.
s- 1)toflux(inergs- 1)fortheXBoötesare6.0×1012ergs- 1 the later SDWFS survey. While the Sternetal. (2005) AGN
count- 1inthe0.5–2keVbandand1.9×1011ergs- 1 count- 1 selectioncanbecontaminatedbySFgalaxiesindeepmid-IR
inthe2–7keVband,whicharederivedfora5kson-axisob- surveys(e.g.Donleyetal. 2012), thecontaminationisnegli-
servation and assuming a canonicalunabsorbedAGN X-ray gibleattheshallowfluxlimitsofISS(seeHickoxetal.2007;
spectrum(see§3.3ofKenteretal.2005). Assefetal.2010a,2011).
Forthisstudy,weusespectroscopicredshifts(spec-zs)from H07 also showed that the distribution of the optical (R)
AGES when possible. For the sources without a spec-z, we band to mid-IR (4.5µm) colors is bimodal for the luminous
adopt the photometric redshifts (photo-zs) calculated using mid-IR quasars. This bimodality can be easily explained
techniquescombining artificial neural network and template by the difference between the SEDs of unobscured and ob-
fitting algorithms (Brodwinetal. 2006). The uncertainty of scuredquasars. Formid-IRselectedquasars,theAGNdom-
this set of photo-z is σ = 0.06(1+z) for galaxies and σ = inates the mid-IR wavelengths, while for obscured quasars,
0.12(1+z)forAGNs. WenotethattheuncertaintyforAGNs the nuclear emission at optical wavelengths is heavily ab-
are dominatedby the differencebetween photo-zsand spec- sorbed. Thusthe SED inthe R bandforobscuredquasarsis
zsforthetypeIAGNs. Theuncertaintiesinthephoto-zsfor similar to those of normalgalaxies (e.g. Pollettaetal. 2006;
QSO2sare difficultto estimate accuratelyasonly9%of the Hickoxetal. 2007, and Hickox et al. 2014, in preparation).
QSO2shavespec-zmeasurements. Forthese32QSO2s, the H07hasshownthatthemid-IRselectedquasarscanbesepa-
uncertaintyofthephoto-zsisσ=0.06(1+z),whichisconsis- ratedintotwodistinctpopulationsofquasarswithanempiri-
tentwiththeuncertaintyforthegalaxypopulation. Recently calcolorcutatR- [4.5]=6.1. Most(∼80%)ofthequasars
spectroscopic follow-up of 35 WISE-selected, optically ob- inH07withR- [4.5]<6.1arespectroscopicallyconfirmedas
scuredquasarsbyHainlineetal.(2014)alsofoundthatphoto- type1quasars,thereforeinthisworkwerefertotheseunob-
zsderivedfromtemplate-basedalgorithm(Assefetal.2010a) scuredquasarswithR- [4.5]<6.1asQSO1s. Asforquasars
haveasimilaraccuracy. Thisisnotsurprisingsincetheopti- withR- [4.5]>6.1,duetothelimiteddepthofAGES,only
calSEDfortheheavilyobscuredAGNsisdominatedbythe ∼5%havespectroscopicmeasurementsandcanbeclassified
hostgalaxywhichhasmorespectralfeaturesandshouldallow as type 2 quasarsspectroscopically. However,H07 have ex-
foramoreaccuratephoto-zmeasurement. tensively studied the quasars with R- [4.5]>6.1 and have
An upper limit for the QSO2 photo-zuncertaintiescan be shownthatthesequasarsareconsistentwithbright,obscured
estimatedbasedondifferentapproaches.H07obtainedanun- X-rayAGNswithN >1022cm- 2andwithAGNbolometric
H
certaintyofσz =0.25(1+z)bycomparingthe Brodwinetal. luminositiesLAGN>1045 ergs- 1 . Therefore,werefertothe
(2006)photo-zstothephoto-zsestimatedbyfitting3different obscuredquasarswithR- [4.5]>6.1asQSO2s.
galaxytemplatestotheBw,R,Iphotometry. Sincetheaccu- Thisempiricalclassificationbasedonthequasarmid-IRto
racyofthephoto-zsbasedonfitting3galaxytemplateto3op- optical color has been shown to be broadly consistent with
ticalphotometricobservationsis limited, theσz =0.25(1+z) both the spectroscopically classified sources (Donosoetal.
uncertaintyisaveryconservativeupperlimit. Forthisstudy, 2014)andX-rayhardnessratio(definedasHR=(H- S)/(H+
weadoptthephoto-zuncertaintyupperlimitofσz=0.25(1+z) S) where H and S are the photon counts in the 2- 7 keV
from H07. However, this is a very conservative estimate band and 0.5–2 keV band, respectively) classification crite-
sincethephoto-zuncertaintiesfortheQSO2swithhostgalaxy ria (Hickoxetal. 2007; Usmanetal. 2014). Since we lack
dominatedopticalSEDsarelikelytobemuchsmaller. Afull spectroscopicconfirmationforthe majorityofthe QSO2s, it
discussionofthephoto-zuncertaintiesisgivenin§8. ispossiblethattheQSO2saredifferentthantheopticaltype
2quasars. Recently, Lacyetal.(2013) havealso studiedthe
2.2. AGNidentificationandclassification
optical and near-IR spectroscopy of mid-IR AGNs selected
TheAGNsinthisworkisdrawnfromthequasarsampleof usingtheDonleyetal.(2012)IRACcolorcriteria,andfound
H11,whichisasubsetofquasarsfromtheH07mid-IRAGN thatasignificantfraction(∼33%)oftheAGNswithevident
sample. The H07 AGNs were identified based on the IRAC mid-IR power-law continuum have no significant emission
color-colorselection criterion of Sternetal. (2005). For the linesconsistentwiththeopticalemissionlinediagnostics(e.g.
redshiftrangeoftheH07sample(0.7<z.3.0),theIRACfil- Baldwinetal. 1981), suggesting that these mid-IR obscured
tersprobewavelengthsatwhichthecharacteristicpower-law AGNs are more deeply obscured than the optically selected
continuumfromthereprocessedAGNradiationstartstodom- type 2 AGNs. This has also been suggested by the recent
inate. Atthesewavelengths,thelightfromoldstellarpopula- studyofHainlineetal.(2014).
tionsischaracterizedbyaRayleigh-Jeanstailofablackbody
radiation with temperature higher than 2500K, which peaks
2.3. Thefinalsample
at wavelengths different than both the AGN accretion disk
and the reprocessed AGN emission. In addition, dust emis- Tominimizetheuncertaintyinthemeasurementsofquasar
sion at near-sublimation temperature heated by the AGN is andhostgalaxypropertiesduetothescatteringinthephoto-
significantly strongerthan that heated by massive stars, thus zs, we focus on the H11 quasar sample, which is a subsam-
AGNs can be easily identified with their mid-IR color (e.g. ple from H07. The H11 quasar sample focuses on the red-
Lacyetal.2004;Sternetal.2005;Donleyetal.2012).These shiftrange0.7<z<1.8andlimitstheQSO1stothosewith
reprocessedphotonsatmid-IRwavelengthsare less affected broadopticalemissionlinesandrobustspec-zmeasurements
byobscurationthantheopticalandtheX-rayphotonsdueto from AGES. For the QSOs in this redshift range, all of the
theirmuchsmallerabsorptioncrosssection. Therefore,even sources have I <18, and the spectroscopic QSO1 sample is
heavilyobscuredAGNscanbeidentifiedusingmid-IRselec- highlycomplete.FortheQSO2sinthisredshiftrange,only32
tioncriteria. (∼9%)oftheQSO2shavespec-zmeasurements.Fortherest
An important feature of the H07 catalog is that the ofQSO2s,weadoptthephoto-zsfromBrodwinetal.(2006).
AGNs were selected from the IRAC Shallow Survey (ISS, ThissubsetofQSOshavebeenusedinthestudyofclustering
Eisenhardtetal. 2004), which has shallower flux limits than propertiesofQSO1sandQSO2sin H11, which showedthat
5
intheredshiftrange0.7<z<1.8,thesampleishighlycom-
100
pleteattheshallowfluxlimitsofISS.Themid-IRselectionof
thisparticularsampleisshowntosufferlessthan<20%con- 80
tamination from star-forming galaxies (Hickoxetal. 2011). y)
J
Sincethe purposeofthisworkisto comparethestar forma- m 60
(
tsiuornepthroaptearntyiesdiofffeQreSnOce1sinanSdFQpSroOp2esr,tiietsisisimnoptodrtraivnetntobmyathkee S 250 40
presenceofstarburstgalaxyinterlopers.Also,evenforbright
mid-IRquasarsthestarburstcontributionintheMIRmaybe 20
non-negligible.Therefore,werelyonSEDdecompositionsto
60
disentangle the AGN and starburst components. The details
oftheSEDdecompositionarelaidoutat§3. OurSEDfitting
40
resultsshowthatall ofthe H11QSOs havea non-negligible
(>20%)AGNcomponent. BasedontheSEDfittingresults,
∼3%oftheH11QSOswouldhaveL lessthantheQSO 20
AGN
criterion(L >1045ergs- 1). Wethereforelimitourfocus
AGN
to thesampleof 546QSO1sand345QSO2swhichsatisfies 0
theQSOluminositycriterionandcomprise∼97%oftheH11 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
sample. z
As a demonstration of the sample used in this work, we
showthedistributionoftheIRAC[3.6]- [4.5]to[5.8]- [8.0] FIG.2.— Top:The250µmdistributionsforQSO1sandQSO2swithdirect
colors and the R- [4.5] to L distributions of the quasar SPIREdetections. QSO1sareshownasthebluestars.The5far-IRdetected
AGN QSO2swithspec-zmeasurements areshownasthefilledcircles. Therest
sampleinFig.1. We notethatinFig.1a, asmallnumberof
ofthefar-IRdetectedQSO2swithphoto-zsareshownasopencircles. Bot-
the H07 sourceshave [3.6]- 4.5and [5.8]- [8.0]colorsthat tom: TheredshiftdistributionsofQSO1sandQSO2sareshowninblueand
lieoutsidetheSternetal.(2005)AGNselectionwedge. The redsolidlines. AlsoshownaretheredshiftdistributionsoftheQSO1sand
Sternetal. (2005) AGN selection wedge and the H07 sam- QSO2sdetectedat250µm. Thehistogramsforfar-IRdetectedQSO1sand
QSO2sareshownastheblueandreddash-dottedlinesandfilledinorange.
ple were definedusing the originalISS catalog in which the
ThisplotshowsthattheredshiftdistributionsofQSO1sandQSO2saresim-
mid-IR photometry was derived using the standard aperture ilarandQSO2shavemore250µmdetectionsthanQSO1s.
correction,whichisdifferentthanthePSFprofilefittingcor-
icallyderivedtemplatestocreateAGNandhostgalaxytem-
rectionsforindividualsourcesintheBrownetal.(2007)cat-
platescoveringthewavelengthrangefromopticaltofar-IRin
alog. Moreover, the updated IRAC photometry comes from
ourdata(see§3.1).
SDWFS,whichis2timesdeeperthantheISS.Thus,itisnot
surprising that a small fraction (∼4%) of the original H07
3.1. SEDtemplates
sampledoesnotmeettheoriginalSternetal.(2005)criterion.
However,theQSOsampleinFig.1ashowsatight[5.8]- [8.0] OurSEDfittingproceduresutilizethefourempiricalSED
to [3.6]- [4.5] distribution, which is not seen in the similar templates spanning 0.03- 30 µm described in Assefetal.
Fig. 1a ofH11whichuse theoriginalISSphotometry. This (2010a, A10 hereafter). A10 have shown that the non-
suggests that the majority of our sample does have power- negativecombinationsofthreedifferentgalaxytemplatesand
law like mid-IR SED with the high-precisionphotometryof a single AGN template, with the addition of extinction to
SDWFS(e.g. see Fig. 22in H07). Forthe purposeofcom- the AGN component only, can robustly describe the opti-
pletenesswedonotexcludethesourcesoutsideoftheupdated cal to mid-IR SEDs of a wide variety of AGNs selected us-
AGN selectionwedge, sincethese sourcesmightstill havea ing Spitzer IRAC color (Assefetal. 2010a) or WISE color
non-negligible AGN component which can be identified by (Assefetal.2010b;Chungetal.2014). However,thesetem-
the SED fits. However, we note the exclusion of this small plates do not include the far-IR wavelengths that are crucial
fraction of sources has little effecton the average properties forthiswork.
oftheQSOs. Wealsoshowtheredshiftandfluxdistributions To extend the A10 AGN template to the far-IR, we create
of the sample in Fig. 2, which shows thatthe redshiftdistri- adhoc AGN templatesby replacingthe hot-dustcomponent
butionsoftheQSO1sandQSO2saresimilar. of the A10 AGN template with the average infrared quasar
templatefromNetzeretal.(2007)andthethreeinfraredAGN
3. SEDDECOMPOSITION templates(forlow-,mean-,andhigh-luminosityAGNs)from
Although it is now well-established that far-IR fluxes Mullaneyetal. (2011), which cover a wide range of AGNs
at wavelengths longer than 60 µm in AGN host galaxies with different mid-IR to far-IR properties. This is based on
are mostly dominated by the emission related to SF (e.g. theassumptionthattheSEDofthehotaccretiondiskaround
Netzeretal.2007;Mullaneyetal.2011;Rosarioetal.2012), theSMBHisthesameforalltheradiativelyefficientAGNs,
cautionisstillrequiredwhenestimatingtheSFpropertiesof andthatthedifferencesareduetovariableextinctionanddif-
powerfulquasars.HereweuseSEDdecompositionstoensure ferentspectralshapesinthemid-andfar-IRwavelengthsdue
thatwehavereliableestimatesoftheintrinsicAGNluminos- to different distributions of dust. To account for the extinc-
ity(L )andstarformationluminosity(LIR13.) tionattheAGNtemplates,we usetheDraine(2003)extinc-
AGN SF
To properlydisentanglethe emission fromstars and AGN tion law with RV =3.1, which mainly attenuates the SED at
accretion,wefitalloftheavailablebroadbandphotometryin λ≤30µmandalsoproducesthesilicateabsorptionfeatures
ourdatasetswithSEDtemplates. Wecombineseveralempir- atmid-IRwavelengthsthatarecommonamongAGN.Theex-
tinctionstrengthistreatedasafreeparameterwitharangeof
13 Defined as the integrated 8- 1000µmluminosity ofthe hostgalaxy 0<Av<48.
componentonly. For the host galaxy templates, we consider two different
6 CHENETAL.
components: the contribution from the stellar population of thanthe20mJy. SinceoursampleisselectedtohaveAGN-
the galaxy, which accounts for the optical to near-IR emis- dominatedmid-IRSED,thefar-IRphotometryisessentialto
sion; and a starburst component, which represents the mid- constraintheaverageSFproperties. Forthesourceswithdi-
to far-IR dust emission from re-processed stellar light. For rectfar-IRdetections,wefittheirmultiwavelengthphotome-
the stellar population component, we follow the approach trywiththeSEDtemplatesdescribedintheprevioussection.
described in A10 by assuming that the stellar population in Forthe731sourceswithoutdirectfar-IRdetections,weusea
any galaxyis comprised of the non-negativecombinationof stackinganalysistoconstraintheiraveragefar-IRflux.
thethreeempiricalgalaxytemplateswithpopulationsofstar- WestacktheQSOswithoutdirectfar-IRdetectionsinbins
burst (Im), continuous star-forming (Sbc) and old stars (el- ofR- [4.5](see§4.2)andbinsofL (derivedinH11from
AGN
liptical), respectively (Assefetal. 2008, 2010a). Unlike the interpolations between the IRAC photometry and the MIPS
ellipticaltemplate,theSbcandImtemplatesbothcontainhot 24µm photometry). The uncertainty of the stacked flux is
dust components in addition to the stellar population SEDs. determinedbybootstrapresampling. Wecreated10,000ran-
Sincethedustcomponentdoesnotextendtothefar-IRwave- dom samples by drawing objects from the original samples
lengthswerequireandwillbetakenintoaccountinthestar- withreplacementuntilthenumberofobjectsineachrandom
bursttemplateswhichwewillchooselateron,wereplacethe sampleisthesameasthenumberintheoriginalsample. The
dustemission(λ>4.9µm)intheSbcandImtemplateswith uncertaintyofthe stackedfluxis thevariationinthe stacked
the SED identicalto the elliptical galaxy to create empirical fluxesoftherandomsamples.Thedetailsofthefar-IRstack-
stellar populationtemplateswith no dustemission. Atthese ing analysis can be found in §3.1.1 of Albertsetal. (2013).
wavelengthsthestellarSEDsaredominatedbylowmassstars TheresultsofthestackinganalysisaregiveninTable1.
similar to the elliptical template. For our mid-IR selected To estimate the average SF luminosity of the far-IR non-
quasar sample, the star light contributionis negligibleat the detected QSOs, we assume that their 250 µm fluxes and
wavelengthswherehotdustemissiondominates. 250 µm flux uncertainties are equal to the averages found
Forthestarburstcomponent,weusethe105starbursttem- fromthestackinganalysisforthesourceswithsimilarvalues
plates from Chary&Elbaz (2001) and the 64 starburst tem- ofbothR- [4.5]andL .Wefindthatforthemajorityofthe
AGN
platesfromDale&Helou(2002). TheChary&Elbaz(2001) SPIRE-nondetectedsample,thebest-fittingSEDreproduces
and Dale&Helou (2002) starburst templates cover a wide theobservedaveragefar-IRfluxwell.
range of SEDs for various prototypical local star forming FromSEDfittingresultswithfar-IRfluxesstackedinbins
galaxies, with LIR in the rangebetween108- 1013.5L⊙ . For if R- [4.5] and LAGN, we find that ∼ 85% of the quasars
SFgalaxiesat z>1, these templateshavebeenshownto be haveaprominentAGNhotdustcomponent(AGNcomponent
reliablewhenusedtoestimateSF-relatedL withmonochro- >50%) atmid-IRwavelengthscoveredby the IRACbands.
IR
maticSPIREobservations(e.g.Elbazetal.2011). However, Allofthequasarshaveanon-negligible(>20%)AGNcom-
aspointedoutbyKirkpatricketal.(2012),thestarburstgalax- ponent, confirming that the H11 sample consists of power-
ies at higher redshift have different dust temperatures and ful quasars. However, the AGN dominates the quasar SED
SEDshapes.Althoughtheeffectofthedifferentdusttemper- evenat24µmformorethanahalfofoursample. SotheSF
atureontheestimationofLSF issmallwhenthewavelengths propertiesofthesepowerfulquasarsisprimarilyconstrained
IR
oftheSPIREbandsprobethepeakofthecolddustemission, bytheirfar-IRphotometry. Forindividualsourceswithonly
itisstillimportanttoincludehighredshiftstarbursttemplates the stacked SPIRE flux to anchor the starburst component,
to accommodate the possibly different SED shapes of high this approach of using stacked far-IR flux can lead to inac-
redshiftstarbursts. Therefore,we also includethe z∼1 and curate LSF. However, the main purpose of this work is not
IR
z∼2averagestarburstSEDtemplatesfromKirkpatricketal. to determine the LSF for individual QSOs. Instead, we take
IR
(2012) in our SED fitting analysis. We thus adopta total of thisapproachtoconstraintheAGNcontaminationintheav-
171starbursttemplatesinourSEDfittinganalysis. eragefar-IRSF luminositywith the well-fittedmid-IRAGN
GiventheseSEDtemplatesdescribedabove,wefittheob- SED, which can robustly determine the average LSF for dif-
IR
served photometry with an iterative χ2 minimization algo- ferentQSOpopulations. We notethatwhenwecomparethe
rithm(Levenberg-Marquardt)tominimizethefunction. stacked fluxes from either the R- [4.5] bin or the L bin
AGN
with the best-fitting AGN component, we find that the aver-
nfilters ageAGNcontributionat250µmisonly5%;andnoneofthe
χ2=X QSOsinoursamplehaveanAGNcomponentthatcontributes
morethan40%at250µm. SincetheaverageintrinsicSEDs
i=0 (1)
F - aF - bext(E(B- V))F - cF forAGNsareexpectedtofallmorerapidlyinmoreluminous
obs,i star,i AGN,i starburst,i. systems (Mullaneyetal. 2011), a starburst componentis es-
σi2 sentialtoreproducetoaveragefluxat250µm.
With the best-fitting SEDs, we calculate the total infrared
Here F is the observed flux in the ith band and σ is its
obs,i i luminosityforeachQSO(Ltot )byintegratingthebest-fitting
uncertainty. Fgalaxy,i,FAGN,i andFstarburst,i arethefluxesofthe SEDsfrom8- 1000µm. FoIRreachsourceinoursample, we
stellar(optical),AGN(opticaltofar-IR)andstarburst(far-IR)
alsocalculatetheL ofthestarburstcomponentbyintegrat-
templates for filter i. The function ext gives the extinction IR
ing the host galaxy component of the best-fitting SED over
oftheAGNfluxinbandigiventhecolorexcessE(B-V).We
optimizea,b,candE(B-V)tominimizetheχ2. the same wavelength range (LSIRF hereafter) 14. We find that
for our sample, the average ratio between LSF and Ltot is
IR IR
3.2. Results
While uniform NUV to mid-IR photometric coverage is 14Forclarification,weuseLIRasageneralterminologyforanyintegrated
available for all of the sources in our sample, only ∼18% 8- 1000µmluminosity.LtIoRt istheintegratedLIRofthebest-fittingSEDand
of them have a SPIRE 250µm detection with a flux larger LSIRFistheLIRofthehostgalaxycomponentonly.
7
QSO1 QSO2
1046 E( Bc-2V=1)=108..2009 1046 E( Bc-2V=1)=711..0608
z=1.31 1045 z=1.05
1045 1044
1043
1046 E(B-V)=0.19 1046 E(B-V)=1.34
c 2=15.55 1045 c 2=84.35
z=1.46 z=0.70
1045 1044
1043
1044
1046 1047
E(B-V)=0.24 E(B-V)=0.81
-1s) c 2=137.72 -1s) 1046 c 2=134.04
g z=1.57 g z=1.07
(er 1045 (er 1045
Ln Ln 1044
n n
1046
E(B-V)=0.09 E(B-V)=0.72
1046 c 2= z1=218..4502 1045 c 2 =z=111..6475
1045 1044
E(B-V)=0.06 E(B-V)=0.58
1046 c 2=88.49 1046 c 2=73.61
z=1.79 1045 z=1.23
1045
1044
1044
0.1 1.0 10.0 100.0 1000.0 0.1 1.0 10.0 100.0 1000.0
Wavelength (m m) Wavelength (m m)
FIG.3.—Examplesofthebest-fittingSEDs(solidline)intherest-frameofeachsource. ThesesourcesarefittedusinganAGNcomponentwithaDraine
(2003)extinctionlaw(dottedcurve),anempiricalstellarcomponent(dash-dotted)andempiricalstarbursttemplates(dashed). Wefindthatalmostallsources
haveanSEDdominatedbythestarburstcomponentinthefar-IR,see§3fordetails.
∼55%, indicative of a non-negligible AGN contribution to
13.2
Ltot inthesemid-IRluminousquasars. Thisisalsoconsistent
IR
withtheKirkpatricketal.(2012)resultsofthestudycombin- 12.5
ingSpitzerIRSspectraandmulti-bandHerschelphotometry ]LO • 12.0
forz∼1andz∼2ULIRGs. However,thefar-IRfluxprobed [R
bytheSPIRE250µmbandisdominated(>70%)bythestar- LI 11.5
burstcomponentfor ∼90%of our sample. This result con-
11.0
firmsthatquasarSEDsarestilldominatedbyastarburstcom-
ponent at rest-frame wavelengths longer than 100µm (e.g. 47.8
Netzeretal.2007; Mullaneyetal. 2011;Rosarioetal. 2012; s)
mDeidl-MIRorqouaestaarls.2c0an13s)t.ilTlhbeerreofobures,tltyhemaevaesruargeedLwSIRFitohftphoewinecrlfuu-l (erg/N 47.0
G 46.0
sionoffar-IRphotometry. LA
We nextderivethe bolometricAGNluminosity(L ) by g
AGN lo 45.0
directly integrating the de-absorbed AGN component. We
comparetheL fromourSEDfittingwiththeL derived 0.8 1.0 1.2 1.4 1.6
AGN AGN
z
inH11,andfoundthatourL ishigherthantheH11L
AGN AGN
byan averageof∼0.1dex. Since the derivationofL in
H11didnotcorrectfortheweakbutnon-negligibleduAsGtNob- FIG.4.—Theredshiftdistributionsofthe(8- 1000µm)LtIoRtalandtheAGN
bolometricluminosityderivedfromSEDfittingandfar-IRstackingfornon-
scurationandSFgalaxycontaminationinthemid-IR,itisnot detectedsourcesdescribedin§3.2.Thesymbolsandlinesrepresentthesame
surprisingthatourL isdifferent. However,thedifference subsetsofobjectsshowninFig.1,wherebluestarsandredcirclesareQSO1s
AGN
is small due to the fact thatmost of the quasarsin this sam- andQSO2s,respectively.Individuallyfar-IRdetectedsources(FIR-QSO)are
enclosedwithorangecircles.
plearedominatedbytheAGNandthedustobscurationonly
weaklyattenuatestheSEDatmid-IRwavelengths. Weshow
theredshiftdistributionsofLtot andL inFig.4. inated by the cold dust emission heated by young stars
IR AGN (e.g. Netzeretal. 2007; Lacyetal. 2007a; Kirkpatricketal.
2012;Mullaneyetal.2012). Althoughalternativecaseshave
4. FAR-IROBSERVATIONSANDQUASAROBSCURATION been reported in some recent studies (e.g. Hineretal. 2009;
Several studies have pointed out that even for AGNs with Daietal. 2012), the cold temperatureof the far-IR emission
quasar-like luminosites, their far-IR SEDs are often dom- stillrequirestheAGN-heateddusttoresideatalargedistance
8 CHENETAL.
0.4 DuetothefluxlimitoftheSPIREcatalog, f onlyreflects
250
QSO1-average (a) the fractionof QSOs hosted by galaxieswith very luminous
QSO2-average
far-IRemission. Toextendthisanalysis,weestimatetheav-
on 0.3 erage250µmfluxinbinsofR- [4.5]totestifthefar-IRflux
acti ofourquasarsamplealsoevolveswithR- [4.5].Themeasure
r
n f the average 250µm flux (S250) for all QSOs, we first esti-
ectio 0.2 mSPaItReEthoebasveerrvaagtieo2n5s0usµimngflthuexsftoarckthinegsoaunracleyssiws idtihsocuutssdeidreicnt
det §3.2. Next, the stacked fluxes is combine with the sources
m 0.1 individually detected in far-IR to calculate the average S250
m0 forallQSOs. Weshowtheaverage250µmflux(S )inthe
5 250
2 same R- [4.5] bins used for Figure 5, with results given in

0.0 Table1.
4 5 6 7 8 Fromthestackinganalysis,wefindthatforQSOswithout
R-[4.5] (Vega) directfar-IRdetections,thereisnosignificantdependenceof
theaverageS onR- [4.5]. However,themeanS ofthe
FIG.5.— The250µmdetection fraction(f250)asafunction oftheR- individuallyd2e5t0ectedfar-IRsourcesshowsaR- [4.525]0depen-
[4.5]color. The f250 forallQSO1sisshownasthebluestar,andthe f250 dencesimilartothatof f . Therefore,drivenbytheincreas-
forallQSO2s(weightedtohavearedshiftandAGNbolometricluminosity 250
distributionsimilartothatofQSO1s)isshownastheredcircles.Thisfigure ing fraction of far-IR luminous sources, the average S250 of
showsthatQSO2sare2.4timesmorelikelytohaveafar-IRdetection. theentireQSOsampleiscorrelatedwithR- [4.5].
This result shows that for powerful mid-IR quasars with
from the central SMBH (Sandersetal. 1988). In addition, moredustattenuationatopticalwavelengths,theaveragefar-
powerful molecular outflows might also produce strong far- IRemissionis stronger. Thisimpliesthatatleastpartofthe
IR emission (e.g. Sunetal. 2014). However, the number of obscuration seen in the obscured QSOs might be due to the
AGNsandstarburstgalaxieswithconfirmedwarmmolecular far-IRemittingdust.
outflowsfromhigh-resolutionfar-IRorsub-mmobservations
is limited. Nonetheless, strong far-IR emission for quasars 5. THEAVERAGELSIRFOFQSO1SANDQSO2S
implies the existence of dust extending to a large distance In§3,wefoundthatQSO2shavehigher f andS than
250 250
well beyond the putative obscuring torus, and a comparison QSO1s.Inthissection,weexaminewhethertheobserveddif-
of far-IR properties between QSO1s and QSO2s could de- ferenceintheaveragefar-IRfluxesofQSO1sandQSO2scan
termine that whether the quasar obscurationis related to the beassociatedwiththedifferenceinthestarformationproper-
large-scaledust. tiesoftheirhostgalaxies,aspredictedbythegalaxyevolution
modelswithaconnectionbetweenquasarobscurationandstar
4.1. SPIREdetectionfraction formation.
We first calculate the median star formation luminosity
We begin with a very simple test by measuring the far-IR
by averaging the LSF derived from SED fitting (§4). We
detection fraction above a flux limit of 20 mJy in the Her- IR
schel SPIRE 250 µm filter (f250 hereafter) for QSO1s and find that for QSO1s, the mean LSIRF is 1045.29±0.03 erg s- 1
QSO2s separately. In our survey region, the shallowness of (1011.71L⊙); and for QSO2s, the mean LSIRF is 1045.59±0.04
the SPIRE observation implies that any quasars in our sam- ergs- 1 (1012.01L ). We find that similar to f and the av-
⊙ 250
ple with a far-IRdetection are at least as brightas luminous erageS , the medianLSF forQSO2sissignificantlyhigher
infraredgalaxies(LIRGs,whicharedefinedasgalaxieswith than tha2t50of QSO1s by 0I.R30 dex. The higher LSF in QSO2s
LIR(8- 1000µm)>1011L⊙ ) at the same redshifts. At first confirmstheresults from§3 thatQSO2sare assIoRciatedwith
glance,wefindthatamongthe546QSO1s,only68(12%)of
host galaxies with more star-forming cold dust which might
them are detected in the SPIRE 250µm filter, while 102 of beobscuringthenuclearemission.
the345(29%)QSO2shavefar-IRdetections.
We nextuse R- [4.5]as a roughproxyof the obscuration 5.1. LSF vs. R- [4.5]
IR
strengthonthenuclearemissionandstudywhether f isre-
250 TofurtherstudytherelationbetweenAGNobscurationand
latedtotheextinctionattheRband. Theprimarygoalofthis LSF, we divide our QSO sample into bins of R- [4.5] as a
analysis is to study whether the star formation properties of IR
proxyfor AGN obscuration. To be consistent with §3.1 and
thequasar-hostinggalaxiesarerelatedtotheobservedobscu-
avoid the uncertaintiesby the lack of spec-zs of QSO2s, we
ration.Therefore,itisimportanttomakesuretheQSO1sand opttouseR- [4.5]insteadoftheE(B- V)derivedfromSED
QSO2sarematchedinkeyproperties,i.e. redshiftandAGN
fitting as the proxyfor nuclear obscuration, since the uncer-
bolometricluminosity(L ,estimatedfromtheSEDfitting
AGN taintiesofthephoto-zlimittheaccuracyofourabilitytomea-
analysisdiscussedin§3).Wedivideoursampleintofourbins sureE(B- V). WecalculatethemedianAGNcontributionat
ofR- [4.5]andcalculate f foreachbinbyapplyingstatis-
250 observedframeRand[4.5]bandsusingthebest-fittingSEDs
ticalweightsto eachquasarso thatthe samplesare matched in each bin. We find that towards redder R- [4.5] colors,
in redshift and L in each bin. We find that the weighted
AGN theAGNcontributes89%,79%,29%,12%attheRbandand
f increasesrapidlywithR- [4.5]forQSO1s. ForQSO2s,
250 74%,59%,43%,60%atthe[4.5]band.Thissuggeststhatthe
f250 is only weakly correlated with R- [4.5], but the f250 of empiricalR- [4.5]colorisindeedagoodproxyofAGNob-
QSO2sishigherthanthatofQSO1sbyafactorof2.7±0.2.
scurationatopticalwavelengths,andthattheQSOclassifica-
Weshowthe f250 toR- [4.5]relationinFig.5. tionbasedontheR- [4.5]colorisreliable.Wethencalculate
theaverageLSF(hLSFi)inbinsofR- [4.5].Weagainestimate
4.2. TheaverageSPIRE250µmfluxes IR IR
theuncertaintyinhLSFiusingbootstrapresampling.
IR
9
TABLE1
RESULTSOFFAR-IRSTACKINGANALYSIS
Average f250inbinsofR-4.5
R- [4.5](Vega) 4.7 5.4 6.6 7.5
S250(mJy)(ND) 6.58±0.73 7.93±0.59 7.39±0.75 8.57±1.00
S250(mJy)(All) 8.04±0.84 11.4±0.77 16.4±1.50 18.0±0.21
hzi(all) 1.31 1.20 1.19 1.31
Average f250inbinsofLAGN
loghLAGNi[ergs- 1] 45.2 45.6 46.1 46.5
S250(QSO1)(mJy) 3.76±0.82 5.49±0.70 5.21±0.75 7.70±0.95
S250(QSO2)(mJy) 3.56±0.94 8.84±1.40 5.18±0.66 8.44±2.70
z(average) 0.98 1.17 1.36 1.49
NOTE. —Resultsfromthefar-IRstackinganalysisdescribedin§4.2. ThetophalfofthetableshowstheresultsforsourcesbinnedinR- [4.5]. Thefirst
rowshowstheresultsforthefar-IRnon-detected(ND)sourcesonly,andthesecondrowshowstheresultfortheentireQSOsample.Theaverageredshiftforall
QSOsineachbinisalsogiven.ThebottomhalfofthetableshowstheresultsfortheentireQSO1andQSO2samplesbinnedintheirAGNbolometricluminosity.
TheaverageredshiftforeachR- [4.5]andLAGNbinisalsolisted;wenotethatthereisonlya<0.02averageredshiftdifferencebetweentheQSO1sandQSO2s.
46.2 46.0 12.6
46.0 QQ SSFSOOItaR12c-k d((iaaenlltglle))cted 45.8 QQASSllOO Q12SO Hickox et al. 2014 12.4
s] 45.8 s] 12.2
g/ g/ 45.6
SF g L [er IR 45.6 SF [erg L IR 45.4 main-sequence galaxy 12.0log LO •
o 45.4 o 11.8
l l
45.2
45.2 11.6
45.0 45.0 11.4
4 5 6 7 8 45.5 46.0 46.5 47.0
R-[4.5] (Vega) log L [erg/s]
AGN
FIG.6.—(a)Left:TherelationshipbetweenLSIRFandR- [4.5]forthemid-IRselectedQSOs.Far-IRdetectedQSOsareshownbytheorange,opensymbolsand
thestackingresultsforthefar-IRnon-detectedQSOsareshownbythegray,opensymbols. Inbothcases,thereisonlymarginalincreaseinLSFwithR- [4.5].
IR
However,drivenbythemuchhigherfar-IRdetectionfractioninQSO2s,theaverageLSF forQSO2sishigherthantheaverageLSFforQSO1. (b)Right: the
IR IR
averageLSIRFforQSOsinbinsofLAGN.TheresultsforQSO1sareshownasthefilled,bluestars;whiletheresultsforQSO2sareshownasthefilledredcircles.
FortheentireQSOpopulation,wealsoshowtheiraverageLSIRFinfourbinsofLAGNastheopen,greensquares.Forcomparison,weshowtheHickoxetal.(2014)
modelevaluatedat0.7<z<1.8astheshadedregion.WenotethattheaverageredshiftineachbinincreaseswithLAGN;andtherelationshipbetweenhLSIRFiand
LAGNisconsistentwiththeredshiftevolutionformain-sequenceSFgalaxieswithM⋆=1011M⊙.WeshowtheredshiftevolutionofMSgalaxiesasthedashed
line.TheLSFforallQSOscoincideswiththeevolutionofMSgalaxies,suggestingaconnectionbetweenQSOhostgalaxiesandMSstar-forminggalaxies(see
IR
§5.2).
We show the results in Fig. 6a. Even though the LSF dif- slopes,withlogLSF∝0.25logL forQSO1sandlogLSF∝
IR IR AGN IR
ferencebetween QSO1s and QSO2s is large, hLSFi doesnot 0.27logL for QSO2s. However, the hLSFi for QSO2s is
IR AGN IR
showthestrongdependenceonR- [4.5]seenfor f andthe higherthanthatofQSO1sineachL binby0.28,0.25,0.32
250 AGN
averageS (Fig.6a.) withinQSO1sandQSO2s. Also, the and0.28dex.
250
LSIRFdifferencebetweenthefar-IRdetectedQSO1sandQSO2s Forcomparison,we also showthe LAGN- LSIRF modelfrom
andthefar-IRnon-detectedQSO1sandQSO2sareonly0.14 Hickoxetal.(2014)whichassumesadirect,linearcorrelation
dexand0.17dex,whicharenotassignificantasthe0.30dex betweentheaverageLAGNandLSIRFwhiletheobservedLAGNis
difference between all QSO1s and QSO2s. Thus the higher modulatedbyrapidvariability.WecalculatetheHickoxetal.
hLSFioftheQSO2sampleismainlydrivenbyitshigherfrac- (2014) relation spanning the 0.7<z<1.8 redshift range in
IR
tionoffar-IRluminousSFgalaxies. Fig. 6 as the shaded region. We find that for all QSO2s be-
sidesthoseinthemostluminousL bin,ourresultiscon-
AGN
sistentwiththeH14model,andtheQSO1shaveLSF slightly
5.2. LSIRF vs. LAGN lowerthanthe modelpredictionsthroughouttheenIRtireLAGN
ToexploretheconnectionsbetweenAGNandhostgalaxy range.However,forbothtypesofQSOs,theslopeoftheLSF-
IR
growthratesindifferentQSOpopulations,we separatelydi- L relation appearsto be shallower than the prediction of
AGN
vide the QSO1s and QSO2s into bins of L , and mea- the simple modelif we take the differentaverageredshiftin
AGN
sure the hLSIRFi for each bin. The uncertainties are again eachLAGN binsintoaccount.
estimated by bootstrapping as discussed in §5.1. We plot As was shown by the results in Table 1, the average red-
the L -LSF relations for QSO1s and QSO2s in Fig. 6b. shiftsformoreluminousQSOsarehigher. Althoughourfar-
AGN IR
We find that both the QSO1 sample and the QSO2 sample IRstackingapproachcanreliablymeasuretheaveragefar-IR
have a positive LSF-L correlation with similarly shallow fluxesandhencethe averageLSF, itis importanttonotethat
IR AGN IR
10 CHENETAL.
thedifferenceintheaverageredshiftbetweenthelowestand rionofourQSOsample. Wenextcombinethesenormalized
the highest L bins is 0.5, and it is possible that the ob- AGN templates with two archetypical starburst galaxy tem-
AGN
served correlation between LSF and L is partially driven plates,M82andArp220fromPollettaetal.(2007). Wefind
IR AGN
bythecosmicevolutionofstarformationandAGNaccretion. that to have a star forming galaxy falling out of the AGN
Eventhoughrecentstudiesoftheevolutionofcosmicinfrared identification wedge while having an L ∼ 1045 erg s- 1,
AGN
luminosity density have shown that for the redshift range of its LSF must exceed 1047 erg s- 1. This suggests that an SF
IR
thefourLAGN binsinFig.6b,thecosmicinfraredluminosity galaxymustbeahyper-luminousinfraredgalaxies(HyLIRG,
isonlyweaklycorrelatedwithredshift(ρIR∝(1+z)- 0.3±0.1for LIR >1013L⊙) to be able to hide an underlying quasar-like
1.1<z<2.85,Gruppionietal.2013),thereisstillasubstan- AGN component. According to the far-IR luminosity func-
tialevolutioninthespecificSFR (SFRperunitstellarmass) tionofGruppionietal.(2013),therewouldonlybelessthan
of“mainsequence(MS)”SFgalaxies(e.g.Elbazetal.2011). 2HyLIRGsintheredshiftrangeofoursamplegiventhesize
A simple estimate of the effect of MS galaxy SFR red- oftheBoötessurveyregion. Becauseofthelargenumberof
shift evolution can be made by calculating the average LSF sourcesinthelowestL binofoursample,theinclusionof
IR AGN
for MS galaxies with M =1011M at the redshift range of the2additionalQSOshiddeninHyLIRGswouldonlyaffect
⋆ ⊙
our QSO sample. From the theoretical framework of dark theaverageLSFby0.15dex,whichisstillwithinthe2σrange
IR
matterhaloabundancematchingmethods(e.g.Behroozietal. oftheoriginalL .
SF
2013), 1011M is the typical galaxy stellar mass hosted by We can also estimate biases in LSF due to the sources not
⊙ IR
dark matter haloes of mass M = 1013.1[h- 1M ], which selected in the AGN identification wedge by examining the
halo ⊙
is the halo mass reported in recent clustering studies of SEDs of the sources not classified as mid-IR QSOs. In the
mid-IR QSOs (e.g. Hickoxetal. 2011; Donosoetal. 2014; redshift range of our sample, there are 77 SPIRE-detected
DiPompeoetal. 2014). At the redshift of each QSO in our sourcesoutsidetheSternetal.(2005)wedge.Whenwefitthe
sample, the LSF for a 1011M MS galaxy can therefore be SEDsofthesesourceswefindthatmostofthesepowerfulSF
evaluatedusinIgRtheredshift-de⊙pendentSFR-M relationfrom galaxieshavenosignificantAGNcontributioninthemid-IR.
⋆
Whitakeretal.(2012)andaKennicuttrelation,SFR=1.09× For the sources without a mid-IR AGN component, we as-
fLuSIRnFc/tLio⊙n()K. Wenenificnudttt1h9a9t8th,emaovdeirfiaegdefLoSrFafCorhMabSriegrailnaxitiieaslmevaasls- smuamtiectlhuamtiAnoGsNitycoatnt6riµbmuteass<an1u0p%petorltihmeitm.iOd-nIlRy4moofnothcehr7o7-
IR haveL satisfyingtheL >1045ergs- 1 criterion. Ifwe
uated using this approach is consistent with the average LSF AGN AGN
IR addthese foursourcesinto the lowestL bin, the average
in each L bin. This implies that for mid-IR quasars in AGN
AGN LSFwouldonlyincreaseby0.05dex.Therefore,weconclude
thisredshiftrange,theLSF andinstantaneousL mightnot IR
IR AGN that the L - L correlation observed in Fig. 6 is not bi-
be directly connected but still follow similar redshift evolu- SF AGN
ased by the exclusionof heavily obscuredquasarshiddenin
tion. Thissuggeststhatacommonphysicalparameter(e.g. a
powerfulSFgalaxies.
common gas supply) might be driving the evolution of both
SMBHsandgalaxies. Thisisalso consistentwiththerecent
6. X-RAYPROPERTIESOFMID-IRQSOS
study of Rosarioetal. (2013), which showed that the LSF-
IR In §5, we have shown that for mid-IR selected QSOs, the
L correlationfor broad emission line QSOs is consistent
AGN
AGNobscurationatopticalwavelengthsandtheAGNmid-IR
with a scenario in which quasars are hosted by normal star-
luminositycanbothbeconnectedtothestarformationoftheir
forminggalaxies.
hostgalaxies. HereweuseanalternativeAGNaccretionrate
For mid-IR selected AGNs, contamination from star-
indicator,theX-rayemission,tostudytheinterplaybetween
forming galaxy interlopers is also an issue that might intro-
thefar-IRemittingdustandtheX-rayemission.
ducebiasestothe measurementsofAGNandstar formation
Wefirstcountthenumberofdetectionsinthefar-IRandthe
luminosities. While our SED fitting results show that 95%
X-rayobservations. We note thatthe countsof X-raydetec-
ofthesourcesinoursampleareindeedpowerfulQSOswith
tionscan be affected by the varyingsensitivity of the Chan-
AGN-dominatedmid-IRSED,itisstillveryimportanttover-
ifythattheobservedLSF- L relationinFig.6isnotcaused dra observations across the field (e.g. Mendezetal. 2013).
IR AGN However,theXBoötesobservationsarerelativelyuniformin
by the incompleteness of our AGN selection criterion. In
depth (∼4- 8×10- 15 ergs- 1 , Kenteretal. 2005) and this
particular, a very powerful SF galaxy can harbor a heavily
effectshould be equivalentfor both QSO types. The results
obscured AGN and still have starburst-like mid-IR SED. In
of the detection fractions are summarized in Table 2. First,
suchcase,theIRACcoloroftheSFgalaxywouldfalloutof
QSO1s have an X-ray detection fraction of 65%, which is
theSternetal.(2005)selectionwedge(e.g.Kirkpatricketal.
muchhigherthanthe12%far-IRdetectionfractionforQSO1s
2012;Assefetal.2010b;Chungetal.2014), andthesample
(see §3). For QSO2s, the X-ray detection fraction is only
selectedwiththeSternetal.(2005)wedgewouldshowabi-
asedLSF- L relation. 34%whilethefar-IRdetectionfractionis27%. Wealsofind
IR AGN thatthepresenceoffar-IRemissionisassociatedwith lower
By happenstance,themid-IRcoloroftypicalstar-forming
X-ray detection fractions in both QSO1s and QSO2s. For
galaxy templates occupies the lower-left corner of the
QSO1s, 91% of the X-ray detected AGNs are not detected
Sternetal. (2005) wedge at the redshift range of this sam-
inthefar-IR;whileforQSO2s,73%oftheX-rayAGNshave
ple (e.g. Donleyetal. 2012, Fig. 17). Therefore, any AGN
no detectable far-IR emission. These number counts imply
contributioninadditiontotheSFgalaxytemplateswouldeas-
thatlargescale dustmightalso playan importantroleinthe
ily promotea SF galaxyinto the AGN identification wedge.
absorptionofX-rays.
Thisis particularlytrue for oursample of luminousquasars.
Sincealargefraction(∼76%)ofthefar-IRdetectedQSO2s
We test this effect by first normalizing the 6µm luminosity
have no direct X-ray detection, these far-IR bright QSO2s
for AGN templates we used in the SED fitting described in
§4 to L =1045 ergs- 1 to meet the minimum L crite- might not be included in an X-ray selected AGN sample.
AGN AGN However, since both the far-IR and the X-ray observations