ZFOURGE catalogue of AGN candidates: an enhancement of 160{\mu}m-derived star-formation rates in active galaxies to $z$ = 3.2

2.3 MB·

by  Michael J. Cowley

MNRAS000,1–16(2015) Preprint12January2016 CompiledusingMNRASLATEXstylefilev3.0
ZFOURGE catalogue of AGN candidates:
an enhancement of 160µm-derived star-formation rates in
active galaxies to z = 3.2 (cid:63)
Michael J. Cowley,1,2† Lee R. Spitler,1,2 Kim-Vy H. Tran,3 Glen A. Rees,1,4
Ivo Labb´e,5 Rebecca J. Allen,2,6 Gabriel B. Brammer,7 Karl Glazebrook,6
Andrew M. Hopkins,2 St´ephanie Juneau,8 Glenn G. Kacprzak,6
6
1 James R. Mullaney,9 Themiya Nanayakkara,6 Casey Papovich,6 Ryan F. Quadri,5
0
2 Caroline M. S. Straatman,5 Adam R. Tomczak,3,10 and Pieter G. van Dokkum11

n
a 1Department of Physics and Astronomy, Macquarie University, NSW 2109, Australia
J 2Australian Astronomical Observatory, PO Box 915, North Ryde, NSW 1670, Australia
3George P. and Cynthia W. Mitchell Institute for Fundamental Physics and Astronomy, Department of Physics and Astronomy, Texas
8
4CSIRO Australia Telescope National Facility, PO Box 76, Epping, NSW 1710, Australia

A&M University, College Station, TX 77843, USA
]
A 5Leiden Observatory, Leiden University, PO Box 9513, 2300 RA Leiden, The Netherlands
6Centre for Astrophysics and Supercomputing, Swinburne University, Hawthorn, VIC 3122, Australia
G
7Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA
. 8CEA-Saclay, DSM/IRFU/SAp, F-91191 Gif-sur-Yvette, France
h
9Department of Physics and Astronomy, The University of Sheffield, Hounsfield Road, Sheffield, S3 7RH, UK
p
10Department of Physics, University of California Davis, One Shields Avenue, Davis, CA 95616, USA
-
o 11Department of Astronomy, Yale University, New Haven, CT 06520, USA
r
t
s
a AcceptedXXX.ReceivedYYY;inoriginalformZZZ
[

1
v ABSTRACT
6 We investigate active galactic nuclei (AGN) candidates within the FourStar Galaxy
1 EvolutionSurvey(ZFOURGE)todeterminetheimpacttheyhaveonstar-formationin
0 theirhostgalaxies.Wefirstidentifyapopulationofradio,X-ray,andinfrared-selected
2 AGN by cross-matching the deep K -band imaging of ZFOURGE with overlapping
s
0 multi-wavelengthdata.Fromthis,weconstructamass-complete(log(M /M )≥9.75),
∗ (cid:12)
1. AGN luminosity limited sample of 235 AGN hosts over z = 0.2−3.2. We compare
0 the rest-frame U − V versus V − J (UVJ) colours and specific star-formation rates
6 (sSFRs) of the AGN hosts to a mass-matched control sample of inactive (non-AGN)
1 galaxies.UVJdiagnosticsrevealAGNtendtobehostedinalowerfractionofquiescent
: galaxies and a higher fraction of dusty galaxies than the control sample. Using 160µm
v
i Herschel PACSdata,wefindthemeanspecificstar-formationrateofAGNhoststobe
X elevatedby0.34±0.07dexwithrespecttothecontrolsampleacrossallredshifts.This
r offset is primarily driven by infrared-selected AGN, where the mean sSFR is found
a
to be elevated by as much as a factor of ∼5. The remaining population, comprised
predominantly of X-ray AGN hosts, is found mostly consistent with inactive galaxies,
exhibiting only a marginal elevation. We discuss scenarios that may explain these
findings and postulate that AGN are less likely to be a dominant mechanism for
moderating galaxy growth via quenching than has previously been suggested.
Key words: galaxies: active – galaxies: evolution – galaxies: high-redshift – X-rays:
galaxies – infrared: galaxies – radio continuum: galaxies
(cid:63) Thispaperincludesdatagatheredwiththe6.5meterMagellan
† E-mail:[email protected]
TelescopeslocatedatLasCampanasObservatory,Chile.
(cid:13)c 2015TheAuthors
2 M. J. Cowley et al.
of rest-frame UVJ colours to distinguish quiescent galaxies
fromstar-forminggalaxies.Togaugestar-formationactivity,
we employ deep far-infrared (FIR) data (160µm) from the
Herschel Space Observatory. Our principal aim is to com-
1 INTRODUCTION
pare AGN hosts with a mass-matched sample of inactive
There is mounting evidence demonstrating that supermas- galaxies,beforediscussingtheimplicationsofourresultsfor
sive black holes (SMBHs) play a fundamental role in the understanding the connection between star-formation and
formation and evolution of galaxies over cosmic time. Pre- AGN activity, as well as the impact AGN has on galaxy
vious work has found the mass of a SMBH is tightly cor- evolution.
related with various properties of its host’s hot spheroidal Thispaperisstructuredasfollows.InSections2and3,
bulge,includingitsluminosity(e.g.Kormendy&Richstone we describe the ZFOURGE and multi-wavelength data sets
1995; Graham 2007; Sani et al. 2011) mass (e.g. Magorrian and AGN sample construction, while in Section 4 we out-
et al. 1998; Marconi & Hunt 2003; Beifiori et al. 2011) and lineourmethodologytoconstructamass-matchedsampleof
velocitydispersion(e.g.Gebhardtetal.2000;Gu¨ltekinetal. inactive galaxies. In Section 5, we present our comparative
2009; Graham et al. 2011). During periods of rapid accre- analysis,beforediscussingtheresultsandtheirimplications
tion, the galactic nuclei of these systems can also release in Section 6. Finally, we summarise our findings in Section
animmenseamountofenergyintothesurroundingenviron- 7.
ment of the host galaxy (e.g. Kormendy & Richstone 1995; Throughout this paper, we use an AB magnitude sys-
Magorrian et al. 1998). As a result, theoretical simulations tem, a Chabrier (2003) IMF, and assume a ΛCDM cosmol-
commonly invoke feedback from these active galactic nuclei ogy with H0 = 70 km s−1 Mpc−1, ΩM = 0.3, ΩΛ = 0.7.
(AGN) outflows to regulate the star-formation activity of
galaxies(e.g.Ciotti&Ostriker1997;Silk&Rees1998;Cro-
tonetal.2006).Theinclusionofanegativefeedbackmecha- 2 ZFOURGE AND ANCILLARY DATA SETS
nismhelpsresolvetheoverproductionofmassivegalaxiesin
2.1 Galaxy Catalogues
simulations by heating or driving out gas to suppress star-
formation. While observational evidence supports negative Our parent sample is comprised of galaxies identified in
feedbackviaAGN-drivenoutflows(e.g.Nesvadbaetal.2006; the ZFOURGE1 survey, which covers three 11’ x 11’ point-
Feruglio et al. 2010; Fischer et al. 2010), recent studies also ingsintheCDFS(Giacconietal.2002),COSMOS(Scoville
pointtothepossibilityofAGNproducingpositivefeedback, et al. 2007) and UDS (Lawrence et al. 2007) legacy fields.
whereby AGN outflows trigger star-formation by compress- ZFOURGE uniquely employs deep near-IR imaging taken
ing cold dense gas. (e.g. Silk & Norman 2009; Elbaz et al. withfivemedium-bandfiltersontheFourStarimager(Pers-
2009; Zinn et al. 2013). son et al. 2013) mounted on the 6.5m Magellan Baade tele-
In order to reconcile these contradictory outcomes, the scope.Theimagingreaches5σpoint-sourcelimitingdepths
complexinterplaybetweenAGNactivityandstar-formation of ∼26 AB mag in J ,J ,J and ∼25 AB mag in H, H, K
1 2 3 s l s
mustbeexamined.Earlystudies,whichtriedtoachievethis, (Spitler et al. 2012). For galaxies at redshifts z = 1.5−4,
relied on optical spectra to select AGN from large parent these filters bracket the rest-frame 4000˚A/Balmer breaks,
samples of galaxies. The main drawback of this approach resulting in well-constrained photometric redshifts within
was the restriction of low redshifts (z < 0.3; Ho 2005; Kim σ(z)/(1+z) ≈ 1−2% (e.g. Kawinwanichakij et al. 2014).
et al. 2006; Salim et al. 2007). With cosmic AGN activity ZFOURGE is supplemented with existing data from CAN-
peakingatasimilarepochtocosmicstar-formation(z∼2), DELSHST/WFC3/F160W(Groginetal.2011;Koekemoer
thesestudiespotentiallymissakeyphaseofAGNevolution. etal.2011;Skeltonetal.2014)andSpitzer/IRAC,aswellas
More recent studies have pushed to higher redshifts by other ground-based imaging, to generate multi-wavelength
taking advantage of X-ray emission, which is an effective catalogues spanning 0.3−8 µm. Fluxes at wavelengths of
probe of AGN activity. Upon comparing X-ray AGN hosts the Infrared Array Camera (IRAC; Fazio et al. 2004), 3.6,
to mass-matched reference galaxies, these studies yield re- 4.5, 6.8, and 8.0 µm are measured using the deblending ap-
sultssuggestingonlyminorornodifferenceinstar-formation proach described in Labb´e et al. (2006). For further details
activity between the two samples (Xue et al. 2010; Santini on the acquisition, data reduction, and bands used to con-
etal.2012;Mullaneyetal.2011;Rosarioetal.2014).How- structtheZFOURGEcatalogues,seeTomczaketal.(2014)
ever, by relying on X-ray selected AGN, these studies may and Straatman et al. (2015, submitted).
also miss a key phase when AGN are hosted in dust-rich,
X-ray obscured galaxies (Sanders et al. 1988).
In this paper, we expand on this work by investigating 2.2 Radio Data
the empirical connection between AGN activity and star-
Following Rees et al. (2015), we cross-match ZFOURGE
formation by selecting and analysing a diverse sample of
withpublishedradiosourcesbasedonoverlappingdatafrom
AGN across a broad range of obscuration levels over z =
the Very Large Array (VLA). We use the VLA 1.4 GHz
0.2−3.2.Ourparentsampleisthedeep K-bandimagingof
s Survey of the Extended Chandra Deep Field South: Second
ZFOURGE (Straatman et al. 2015, submitted), which not
Data Release of Miller et al. (2013) for the ZFOURGE-
onlygrantsusaccesstoallgalaxiestypes,butalsoallowsus
CDFS field, the VLA-COSMOS Survey IV Deep Data
to probe to lower stellar masses and higher redshifts.
and Joint catalogue of Schinnerer et al. (2010) for the
To identify AGN, we cross-match the K-band imaging
s
with radio, X-ray and infrared (IR) datasets to allow the
use of standard AGN selection techniques, and make use 1 http://zfourge.tamu.edu
MNRAS000,1–16(2015)
Comparison of active and inactive galaxies in ZFOURGE 3
ZFOURGE-COSMOS field, and the Subaru/XMM-Newton CDFS.PhotometryfromthisdataareproducedusingMulti-
Deep Field-I 100 µJy catalogue of Simpson et al. (2006) for Resolution Object PHotometry oN Galaxy Observations
the ZFOURGE-UDS field. The minimum root-mean-square (MOPHONGO) code written by I. Labb´e (for further de-
(RMS)sensitivityforeachsurveyis6,10and100µJy/beam, tails,seeLabb´eetal.2006;Fumagallietal.2014;Whitaker
respectively.Uponcorrectingforsystematicastrometricoff- et al. 2014).
sets in each field, radio sources are cross-matched within a
radiusof1(cid:48)(cid:48) oftheir K-bandcounterparts.Ofthe286radio
s
sources that overlap with the ZFOURGE fields, 264 were
cross-matched with a K-band counterpart. We visually in-
s 2.5 Photometric Redshifts, Rest-frame Colors,
specttheremaining22sourcesandfind2intheZFOURGE-
Stellar Masses and Star-Formation Rates
COSMOS field were missed due to confusion from complex
extendedstructures(i.e.radiojets),withtheirrecordedpo- The photometric redshifts and rest-frame colours of galax-
sitionoffsetfromthegalaxycore.Theremaining20sources iesinZFOURGEarecalculatedusingthepublicSED-fitting
areconsideredcandidateIRfaintradiosources(IRFS;Nor- code,EAZY(Brammeretal.2008).EAZYusesadefaultsetof5
risetal.2006),withavisualinspectionyieldingnoidentifi- templatesgeneratedfromtheP´EGASElibrary(Fioc&Rocca-
ablecounterpartsintheKs-bandimages.Consideringthis,a Volmerange 1997), plus an additional dust-reddened tem-
total of 266 radiocounterparts are found in the ZFOURGE plate from Maraston (2005). Linear combinations of these
Ks-bandimages(∼92%ofalloverlappingradiosources),with templates are fit to the observed 0.3−8 µm photometry for
119 in CDFS, 116 in COSMOS, and 31 in UDS. estimatingredshifts.Stellarmassesarecalculatedbyfitting
Bruzual&Charlot(2003)stellarpopulationsynthesismod-
elsusingFAST(Krieketal.2009),assumingsolarmetallicity,
2.3 X-Ray Data
aCalzettietal.(2000)dustextinctionlaw(withA =0−4),
V
We cross-match ZFOURGE with published X-ray sources aChabrier(2003)initialmassfunction(IMF)andexponen-
based on overlapping data from the Chandra and XMM- tially declining star-formation histories of the form SFR(t)
Newtonspaceobservatories.WeusetheChandraDeepField- ∝e−t/τ, where t is the time since the onset of star-formation
South Survey: 4 Ms Source catalogue of Xue et al. (2011) and τ (varied over log[τ/yr] =7−11) modulates the declin-
fortheZFOURGE-CDFSfield(X11henceforth),theChan- ing function. Star-formation rates (SFRs) are calculated by
dra COSMOS Survey I. Overview and Point Source cata- considering both the rest-frame UV emission from massive
logueofElvisetal.(2009)fortheZFOURGE-COSMOSfield unobscuredstarsandthere-radiatedIRemissionfromdust
(E09 henceforth), and the Subaru/XMM-Newton DeepSur- obscuredstars.ThecombinedUVandIRluminosities(L
UV
veyIII.X-RayDataofUedaetal.(2008)fortheZFOURGE- and L ) are then converted to SFRs (Ψ) using the calibra-
IR
UDSfield(U08henceforth).Theon-axislimitingfluxinthe tionfromBelletal.(2005),scaledtoaChabrier(2003)IMF:
soft and hard bands for each survey is 9.1×10−18 and 5.5×
10−17 erg cm−2 s−1,1.9×10−16 and7.3×10−16 erg cm−2 s−1,and
6.0×10−16 and 3.0×10−15 erg cm−2 s−1, respectively. Upon ΨIR+UV[M(cid:12) yr−1]=1.09×10−10(3.3LUV+LIR) (1)
correctingforsystematicpositionoffsetsineachfield,X-ray
sources are cross-matched within a radius of 4(cid:48)(cid:48) of their K- where L =νL is an estimate of the integrated 1216-
s UV ν,2800
band counterparts. Of the 683 X-ray sources that overlap 3000˚Arest-frameUVluminosity,derivedfromEAZY,andL
IR
withtheZFOURGEfields,545(∼80%)arefoundwithin1(cid:48)(cid:48) is the bolometric 8-1000µm IR luminosity calculated from
ofaK-bandcounterpart.Afurther47sources(∼7%)at>1(cid:48)(cid:48) a luminosity-independent conversion (Wuyts et al. 2008,
s
areaddedafteravisualinspectionofboththeX-rayandK- 2011)usingPACS160µmfluxes.Forstackedmeasurements,
s
bandimagingconfirmednoconfusionfrommultiplesources we consider all sources, including those with zero or nega-
withinthematchingradius.Theremaining91sourcesyield tive 160µm fluxes. This ensures our samples are not biased
nofurthermatcheswithnovisiblecounterpartsidentifiable. against quiescent galaxies or those with low SFRs. A com-
Consideringthis,atotalof592X-raycounterpartsarefound parisonofour160µmfluxestothatofthePACSEvolution-
in the ZFOURGE K-band images (∼87% of all overlapping aryProbesurvey(Lutzetal.2011)revealsgoodcorrespon-
s
X-ray sources), with 422 in CDFS, 93 in COSMOS, and 77 dence, with a median offset of ∆mag ∼0.20. The quality of
in UDS. otherderivedgalaxyparametersareexploredinmoredepth
intheZFOURGEsurveypaperStraatmanetal.(2015,sub-
mitted).
2.4 Far-infrared Data
Forallgalaxies,whetheractiveorinactive,weuse‘pure’
We make use of overlapping Spitzer/MIPS and Her- galaxy templates in our SED fits, without consideration of
schel/PACS FIR imaging. The data used in this study are anAGNcomponent.Somestudiesadoptasinglepower-law
from 24 and 160 µm photometry. We use imaging from the template in an effort to decompose the combined SED into
GOODS Spitzer Legacy program (PI: M. Dickinson) and AGN and host galaxy components (e.g. Hao et al. 2005;
GOODS-H (Elbaz et al. 2011) for the ZFOURGE-CDFS Bongiornoetal.2013;Rovilosetal.2014).Thoughpopular,
field, S-COSMOS Spitzer Legacy program (PI: D. Sanders) it is unknown if such a broad approach would be effective
and CANDELS-H (Inami et al. 2015, in prep) for the on our diverse sample of AGN. We acknowledge potential
ZFOURGE-COSMOSfield,andSpUDSSpitzerLegacypro- contamination from AGN and adopt various tests to check
gram(PI:J.Dunlop)andCANDELS-HfortheZFOURGE- for the effects on photometric redshifts (Section 2.6) and
UDS field. The median 1σ flux uncertainties for each sur- otherderivedgalaxypropertieswhenpresentingourresults
vey is ∼10 µJy in COSMOS and UDS, and 3.9 µJy in (Section 5.3).
MNRAS000,1–16(2015)
4 M. J. Cowley et al.
the efficient way the ZFOURGE medium-band filters trace
Radio AGN the 4000˚A/Balmer breaks, which is driven by stellar light.
3.0 X-ray AGN Despitethis,itremainspossiblethatrestframeopticalAGN
Infrared AGN emission can increase the uncertainty of the photometric
7.40% outliers redshifts. For obscured (i.e. Type-2) AGN, several studies
2.5
ft σNMAD = 0.023 have demonstrated contamination to host galaxy proper-
hi N = 136
s ties is negligible (Silverman et al. 2009; Schawinski et al.
d
e 2.0 2010;Xueetal.2010).However,theAGNpopulationinthis
R
work may also contain luminous, unobscured (i.e. Type-1)
ric σ=0.183 AGN, which may impact SED fits. To quantify how many
et 1.5 80 of these might be in our sample, we search for objects (at
m
o 60 all redshifts) with rest-frame UVJ colours ±0.5 mag around
hot 1.0 40 a SWIRE Type-1 QSO template (Polletta et al. 2007). We
P find 23 sources (∼4% of the parent AGN population) with
20 these colours. A visual inspection of their SEDs reveal a
0.5 0 soundfittophotometry,resultinginaphotometricredshift
−0.2 0 0.2
with low error. Given this, we retain these sources in the
Δz/(1+z )
spec parentAGNpopulation.Forourcomparativeanalysis(Sec-
0
tion5.1),weselectamassandluminositylimitedsubsample
0 0.5 1.0 1.5 2.0 2.5 3.0
fromtheparentAGNpopulation(Section4).Only1ofthe
Spectroscopic Redshift
23 sources with QSO colours is selected in this subsample.
Figure1.Comparisonofphotometricandspectroscopicredshifts
forourradio(greendiamonds),X-ray(bluesquares)andIR(red 3 MULTI-WAVELENGTH AGN SELECTION
dciarschleesd)lAinGeNs ahroestzpsh.oTt =he0s.0o5li±d(l1in+ezisspetch)eaznpdhotth=ezsdpeocttreedlatliinones,tahree ThediverseandcomplexinteractionsbetweenanAGNand
zphot=0.15±(1+zspec).AGNHostsoutsideofthedottedlinesare its host galaxy make constructing a thorough and unbiased
definedasoutliers. sampleaformidabletask.Variationsinluminosity,morphol-
ogy,orientationanddustobscurationdictatetheneedfora
multi-wavelength, multi-technique approach. For example,
2.6 Reliability of AGN Photometric Redshifts
whileopticalandX-rayselectiontechniquesarebothhighly
AGN emission is known to complicate the computation of efficient, they break down when AGN hosts are heavily ob-
photometric redshifts (e.g. MacDonald & Bernstein 2010), scured by large amounts of gas and dust (e.g. Lacy et al.
which can ultimately impact the derivation of rest frame 2006; Eckart et al. 2009). On the other hand, radio and IR
coloursandstellarpopulationproperties.Inordertotestthe selection techniques are relatively immune to dust extinc-
accuracy of our AGN sample (see Section 3 for AGN clas- tion, but galaxies with copious amounts of star-formation
sifications), we compare the sample’s photometric redshifts can contaminate a sample (e.g. Condon et al. 2002; Donley
from ZFOURGE to a secure sample of publicly available etal.2005).Inthissection,wedescribeourapproachtomin-
spectroscopic redshifts sourced from the compilation of the imise such bias by constructing a robust AGN sample from
3D-HST (Skelton et al. 2014) and ZFIRE (Nanayakkara et multi-wavelength data. We restrict the sample to sources
al.2015,submitted)surveys.WeusetheNormalisedMedian over z = 0.2−3.2 with clean photometric detections in the
Absolute Deviation (NMAD) to calculate scatter: ZFOURGEcatalogues(e.g.nearstarandlow-signal-to-noise
(cid:32)|∆z−median(∆z)|(cid:33) flags).FurtherdetailsoftheZFOURGEqualitycontrolflags
σ =1.48×median (2) will be presented in Straatman et al. 2015, submitted.
NMAD 1+z
spec
where ∆z = z −z . From the 500 AGN hosts identified
phot spec 3.1 Radio AGN Selection
inZFOURGE,wefind136cross-matcheswithreliablespec-
troscopicredshifts.Figure1showsarelativelysmallnumber The accretion of material onto a supermassive black hole
ofAGNhostswithphotometricredshiftsverydifferentfrom isknowntoproducenuclearradioemission,collimatedinto
thespectroscopicvalue.Theseoutliers(definedheretohave relativistic jets that propagate perpendicular to the plane
(cid:16) (cid:17)
|∆z|/ 1+z >0.15) make up 7.40% of our sample and are of the accretion disc. While the detection of such radio-
spec
subsequently ejected. Assuming the remainder of the AGN emitting jets unmistakably implies the presence of AGN,
population has a similar outlier fraction, there is potential radio emission can also be caused by star-formation. As a
for an additional 27 AGN in our sample to have unreliable result,radiodetectionsatredshiftsbeyondtheobservablejet
redshifts.Indeed,wevisuallyinspecttheSEDsofthoseAGN structurerequirealternativemeansfordiscriminatingAGN
lacking a spectroscopic counterpart and manually eject 14 hosts from inactive, star-forming galaxies. To achieve this,
(3.85%) with questionable fits. The accuracy of photomet- weusetheRadio-AGNActivityIndexofReesetal.(2015).
ric redshifts for our AGN hosts is σ = 0.023, which is Briefly,thistakesadvantageofthetightcorrelationobserved
NMAD
onlyslightlyhigherthanthegeneralZFOURGEpopulation between a galaxy’s radio (synchrotron) and IR (thermal)
(σ = 0.018; Tomczak et al. 2014). emissions(Helouetal.1985),whichMori´cetal.(2010)found
NMAD
The strong correspondence between the photometric holds for a diverse range of galaxies over a broad redshift.
and spectroscopic redshifts in ZFOURGE is attributed to TheexceptionwasradioAGN,whichpresentedadiscernible
MNRAS000,1–16(2015)
Comparison of active and inactive galaxies in ZFOURGE 5
103 thencalculatedusingthecalibrationfromBell(2003),scaled
to a Chabrier (2003) IMF:
Radio Source
Radio AGN Ψ [M yr−1]=3.18×10−22L (5)
Rees et al. (2015) RADIO (cid:12) RADIO
x As shown in Figure 2, the Radio-AGN Activity Index leads
e 102
d totheidentificationof67radiosourcesdominatedbyAGN
n
I activity in ZFOURGE, with 20 in CDFS, 32 in COSMOS,
y
t and 15 in UDS.
vi
ti
Ac 101
3.2 X-ray AGN Selection
N
G
A While radio surveys pioneered the way for AGN research

o (e.g. Baade & Minkowski 1954; Schmidt 1963; Schmidt &
adi 1 Matthews 1964), the launch of Chandra and XMM-Newton
R heralded in a new era of sensitive, deep X-ray surveys, of-
fering an effective alternative to select AGN. These surveys
havefoundthatX-rayemissionfromsourcesathighGalac-
tic latitudes are predominantly AGN (e.g. Watson et al.
10−1
0 0.5 1.0 1.5 2.0 2.5 3.0 2001)androutinelyoutshinethehigheststar-forminggalax-
ies (∼1042 erg s−1; e.g. Moran et al. 1999; Lira et al. 2002).
Redshift
While this provides an excellent discriminator for AGN se-
lection, heavy obscuration by dense circumnuclear gas can
proveproblematic.One way toaccountforthis is byexam-
Figure 2. The Radio AGN Activity Index (see equation 3) for
all radio sources in ZFOURGE. The evolution of the Wuyts ining the hardness ratio (HR) of a source, which is defined
etal.(2008)averagestar-formingSEDtemplate,calculatedfrom as the normalised difference of counts in the soft and hard
160µm fluxes, is shown by the red line. The grey shaded region X-raybands,(hard-soft)/(hard+soft).TheHRallowsan
represents the 3σ 0.39 dex scatter found in the local radio-FIR estimate of absorption in the X-ray band, where obscured
correlation(Mori´cetal.2010).Reesetal.(2015)adoptaconser- AGN are expected to exhibit a harder spectrum than un-
vative cut above this region (SFRRAD/SFRIR+UV >3; crosshatched obscured AGN due to the absorption of soft X-rays by ob-
region)toselectradioAGN(greendiamonds).Sourcesthatlack
scuringgas(Szokolyetal.2004).Consideringthis,weselect
areliable(>3σ)160µmdetectionaregiven3σlimits(arrows).
X-ray AGN using both the X-ray luminosity and HR of a
source.
offsetfromthecorrelation.TheRadio-AGNActivityIndex, We first start with the cross-matched photometric red-
which operates in SFR space, exploits this offset by assum- shiftsfromZFOURGEandapplyX-rayK-correctionstoes-
ing SFRRADIO=SFRIR+UV if 100% of the radio emission origi- timate rest-frame luminosities using:
natesfromstar-formation.SourceswithexcessSFRRADIO are L [erg s−1]=4πd2(1+z)Γ−2f (6)
identified as radio AGN: X l x
whered istheluminositydistanceincm, f istheobserved
SFRRADIO/SFRIR+UV=Radio AGN Activity Index>3 (3) X-ray fllux in erg cm−2 s−1, and Γ is the phoxton index of the
The inclusion of UV emission accounts for the possibility X-rayspectrum,whichwasfixedtoatypicalgalaxyphoton
of radio star-forming galaxies with low dust, which would index3 of Γ = 1.4. For sources in the X11 catalogue, the
otherwise produce an excess in SFR and be misclassified intrinsic flux is derived from counts in the 0.5-8 keV full
RAD
as radio AGN. To calculate radio SFRs, we first make use band, while for the E09 and U08 catalogues it is derived
ofthecross-matchedphotometricredshiftsfromZFOURGE from the sum of the counts in the relevant bands over 0.5-
and apply radio K-corrections to estimate rest-frame radio 10 keV. We adjusted flux values calculated in the E09 and
luminosities using: U08cataloguestoalignwiththefullbandpassvaluesofthe
X11 catalogues (0.5-10 → 0.5-8 keV) assuming a power-law
LRADIO[W Hz−1]=4πdl2(1+z)−(α+1)fRADIO (4) model of Γ = 1.4 (i.e. E09 and U08 fluxes are multiplied
by a factor of 0.95). We then use the selection technique of
where d is the luminosity distance in cm, f is the ob-
l RAD
servedradiofluxinWm−2 Hz−1,andαistheradiospectral Szokoly et al. (2004) to select X-ray AGN:
index2,whichwefixtoα=−0.3asfoundintheWuytsetal. L ≥1041erg s−1 & HR>−0.2
X
(2008)averagestar-formingSEDtemplate.Whilethisspec-
tralindexisflatterthanthestandardα=−0.7,itisadopted LX≥1042erg s−1 & HR≤−0.2 (7)
toensureconsistencywiththeWuytsetal.(2008)SEDtem- Theluminositythresholdislowerforsourceswithastronger
plate, which is also used to derive IR SFRs. The difference HR on account of substantial absorption. In the absence
betweenthetwoindexvaluesisonelesssourceidentifiedas of a HR measurement, we only select sources with L ≥
x
a radio AGN under α=−0.7. 1042erg s−1. As shown in Figure 3, this approach leads to
Usingtherest-frameradioluminosities,radioSFRsare the identification of 270 X-ray sources dominated by AGN
2 Theradiospectralindex,αisdefinedfromSν∝να,whereS is 3 Thephotonindex,Γisrelatedtothenumberofincomingpho-
themeasuredfluxdensityandνistheobserver’sframefrequency tonsasafunctionofenergyE,dN(E)/dE∝E−Γ
MNRAS000,1–16(2015)
6 M. J. Cowley et al.
andrapidlyintroducecontaminantsintotheselection.Mes-
1045 sias et al. (2012) investigated this and found by extending
the use of IRAC to additional wavebands, they could reli-
ably select AGN over a broader redshift range. Specifically,
UDS
1044 theauthorsproposedtwocolourdiagnostics, K +IRACat
s
lowerredshifts(z=0−2.5)andIRAC+24µmathigherred-
shifts(z=1−4).Weadoptthesediagnosticswiththeadded
)
-1s1043 COSMOS bcoannddistitoonrseoduurcceesschaattveer.aT5oσmdaettcehcttihoenrleimdsihtifitnbailnlsruelseevdanint

rg CDFS our analysis (see Section 4.1), we select IR AGN based on
(e1042 the following constraints:
x 
L z<1.8[K4s.5−][−4.[58].0>]>0 0 (8)
1041

XXS--zorrkaaoyy lySA oGeuNt rcael. (2004) z>1.8[[88..00]]−−[[2244]]>>20..95×([4.5]−[8.0])+2.8 (9)
1040
0 0.5 1.0 1.5 2.0 2.5 3.0 As shown in Figure 4, this approach leads to the identi-
fication of 234 IR sources dominated by AGN activity in
Redshift
ZFOURGE, with 66 in CDFS, 50 in COSMOS, and 118 in
UDS.
Figure 3. X-ray rest-frame luminosity as a function of redshift
forallX-raysourcesinZFOURGE.Allsourcesabove1042ergss−1
(uppercrosshatchedregion;Szokolyetal.2004)areidentifiedas 3.4 Summary of AGN Samples
AGN(bluesquares),whileonlysourceswithaHR>-0.2downto
1041 ergss−1 (lowercrosshatchedregion;Szokolyetal.2004)are WeillustratetherelativesizeandoverlapbetweentheAGN
identified as AGN. The approximate luminosity limits for each samples in Figure 5 (right panel). Overlap arises from the
fieldareindicatedbythereddashedcurves. complexandbroademissionofAGNspectraandemphasises
thatoursamplesarenotwhollyindependentandnotsimply
relegatedtoeitheraradio,X-rayorIRselectionbin.Despite
activity in ZFOURGE, with 187 in CDFS, 57 in COSMOS, this,therelativesizeoftheoverlapiscomparabletoprevious
and 26 in UDS. studies that have performed multi-wavelgth AGN selection
(Hickox et al. 2009; Juneau et al. 2013). Like these studies,
we find the overlap between radio and X-ray AGN hosts is
3.3 Infrared AGN Selection
low, while the overlap between IR and X-ray AGN hosts is
DespitetheefficiencyofAGNselectioninX-raysurveys,an significantlylarger.Ofthe500AGNidentified,54arefound
imbalance in the cosmic X-ray background budget suggests tooverlapinoneormorewavebands,with5identifiedinall
an additional population of heavily obscured AGN are be- three. For this work, overlapping AGN are treated as inde-
ingmissed(Comastrietal.1995;Gillietal.2001,2007).IR pendentsources(i.e.5sources:aradio,X-rayandIRAGN)
observationsofferaneffectivewaytoidentifytheseAGNby unless measurements are made on the combined AGN sam-
virtue of dust radiating the reprocessed nuclear emission in ple, in which case they are treated as a single source. We
the mid-IR regime (Sanders et al. 1988, 1989). Such emis- summarisethecolumnsofthecompleteAGNdatasetinTa-
sion is evident by the changing shape of a galaxy’s SED, ble 4, which provides all host galaxy parameters used to
where an increase in AGN activity also leads to a dilution select AGN in ZFOURGE. In Figure 5, we display the stel-
inthestrengthofPolycyclicAromaticHydrocarbon(PAH) lar mass and Ks-band distributions, along with the popula-
emissions features formed by ultraviolet excitation typical tion numbers by way of a Venn diagram. This dataset acts
in star-forming regions (Brandl et al. 2006). The mid-IR as a complementary catalogue to the primary ZFOURGE
is then dominated by the thermal continuum (e.g. Neuge- catalogues. An amended version will be made available at
baueretal.1979;Heisler&DeRobertis1999).Anumberof http://zfourge.tamu.edu upon the full public release of
IRACcolour-colourdiagnosticshavebeendesignedtoselect ZFOURGE.
AGN by taking advantage of this process (e.g. Lacy et al.
2006; Stern et al. 2005; Donley et al. 2012). The choice of
diagnostic depends on the science being conducted as each
4 MASS-LIMITED SAMPLE
has a particular level of completeness and reliability, with
one often dominating in favour of the other (e.g. Barmby Inthissection,weextractAGNhostsfromthecatalogueof
et al. 2006; Donley et al. 2007; Messias et al. 2012). Un- candidatesselectedinSection3withthegoalofconstructing
fortunately, with increasing redshift, the IRAC bands be- amass-matched,inactivesampleofgalaxies(controlsample)
gin to probe shorter rest-frame wavelengths and eventually tocomparestar-formationactivitybetweenAGNhostsand
tracethe1.6µmstellarbumpofagalaxy’sSED,whichcan inactivegalaxies.Selectionisbasedonredshift,stellarmass
mimic the AGN thermal continuum. As a result, diagnos- and luminosity limits, with the goal of minimising bias on
tics limited to IRAC colours become ineffective at z (cid:38) 2.5 host galaxy properties. Given the shallow X-ray and radio
MNRAS000,1–16(2015)
Comparison of active and inactive galaxies in ZFOURGE 7
4
2 Infrared Source z=0.2-1.8 z=1.8-3.2
Infrared AGN
Messias et al. 2012 3
)
) 1 B 2
B A
A (
(
]
] 4
5 2 1
4. [
[ 0 -
-
0] 0
Ks 8. Starburst (SB)
[ Spiral
Elliptical
Hybrid (AGN+SB)
1
1 QSO1
QSO2
Tracks z > 1.8
Tracks z < 1.8
2
1 0 1 2 1 0 1 2
[4.5] - [8.0] (AB) [4.5] - [8.0] (AB)
Figure 4.TheMessiasetal.(2012) Ks +IRAC(left)andIRAC+24µm(right)infraredcolour-colourspaceforallIRsourceswithin
ZFOURGE.Sourcesthatfallwithinthecross-hatchedregionsareconsideredAGNdominated(redcircles).Overplottedaretheredshift-
dependentspectraltracksforaselectionofactive(hybrid,QSO1andQSO2)andinactive(starburst,spiralandelliptical)galaxiesfrom
theSwireTemplates(Pollettaetal.2007).Thedashedportionofthetracksrepresentsz=0−1.8andthesolidportionz=1.8−4.
2015), as shown in Figure 5 (left panel). We apply further
Table 1.Luminositylimitsofmass-limitedAGNsample
restrictionsbysplittingtheAGNsampleintothreeredshift
binsofz=[0.2-0.8],[0.8-1.8],[1.8-3.2],eachwithvaryinglu-
Waveband L1.4GHz LX LI1R zmin zmax NAGN2 minosity limits based on the luminosity thresholds of their
(WHz−1) (ergs−1) respective wavebands (i.e. L , L and L ). These lim-
1.4GHz X IR
its are summarised in Table 1 and while they reduce AGN
Radio 1.0×1023 - - 0.2 0.8 10
numbersandrestrictcomparisonacrossredshifts,theymin-
6.0×1023 - - 0.8 1.8 11 imise potential luminosity biases by ensuring a consistent
1.9×1024 - - 1.8 3.2 5 luminosity-completeness within each redshift bin.
X-ray - 4.0×1041 - 0.2 0.8 31
- 2.0×1042 - 0.8 1.8 60
4.2 Control Sample of inactive Galaxies
- 7.0×1042 - 1.8 3.2 50
Tight correlations exist between the physical properties of
Infrared - - 6.0×1027 0.2 0.8 7
galaxies and their stellar mass (e.g. Tremonti et al., 2004,
- - 3.0×1028 0.8 1.8 39 mass-metallictyandNoeskeetal.,2007,mass-star-formation
- - 1.0×1027 1.8 3.2 22 rate).Thismakesconstructingamass-matchedcontrolsam-
pleofinactivegalaxiesanessentialcomponentforourcom-
1 LIR = L8µm atz=0.2−1.8and L24µm atz=1.8−3.2 parative analysis. Without this consideration, even a mass-
2 NumberofAGNhostswithinthespecifiedlimits limited sample would be dominated by galaxies just above
themassthreshold,potentiallybiasinganycomparison.We
construct our mass-matched control sample by binning in-
datausedtoselectAGNhostsinZFOURGE-UDS,thisfield
activegalaxiesintonarrowmassintervalsof∆M =0.2dex.
will be excluded from the comparative analysis. ∗
For each AGN host, we randomly select an inactive
galaxy from the same redshift bin (z = [0.2-0.8], [0.8-1.8]
or [1.8-3.2]) and of similar mass, within ∆M . For example,
∗
a z = 0.74 radio AGN host with log(M /M ) = 10.87 has
4.1 Redshift, Mass and Luminosity Cuts ∗ (cid:12)
112 inactive analogues from which to draw from. We then
To overcome the potential bias associated with K-band se- calculateandrecordameanvalueforvariousphysicalprop-
s
lected galaxies, we limit our sample of AGN hosts to a erties of the selected control inactive galaxy (i.e. rest-frame
stellar-mass cut of log(M /M ) ≥ 9.75, which sits above colour, stellar mass and star-formation rate) and repeat for
∗ (cid:12)
the 80% completeness limit of ZFOURGE (Papovich et al. thenextAGNhostuntilwehaveacontrolsamplewiththe
MNRAS000,1–16(2015)
8 M. J. Cowley et al.
12
Infrared AGN
25
Radio AGN
11 24
23 180 3 47
/M)⊙10 AB) 22 5 12
M* ( 21 46
( S
log 9 K 20
207
19
8 All Sources All Sources
Radio AGN 18 Radio AGN
X-ray AGN X-ray AGN
Infrared AGN 17 Infrared AGN
7
0 1 2 3 0 1 2 3 X-Ray AGN
Redshift Redshift
Figure 5. Stellar mass (left) and Ks-band magnitude (middle) as a function of redshift for our radio (green diamonds), X-ray (blue
squares)andIR(redcircles)AGNhosts,alongwiththeparentsamplefromZFOURGE(greycircles).Thereddottedlineintheleftplot
representsthe80%mass-completenesslimitinZFOURGE,whiletheblackdashedlineisthestellarmasscutoflog(M∗/M(cid:12))≥9.75.The
Venndiagram(right)showstherelativenumberofAGNidentifiedinradio(green),X-ray(blue)andIR(red)wavebands.Theoverlapping
regionsbetweensamplescorrespondtotherelativenumbersselectedinmultiplewavebands.Notethatthesenumberscorrespondtothe
completeAGNcandidatecataloguedetailedinSection3.Forclarity,only1/3rd oftheparentsampleisplotted.
same number of galaxies as the AGN sample. We gener- When comparing the distribution of UVJ colours be-
ate 100 such independent control samples, which we use to tweenAGNhostsandthecontrolsample,thetwoarefound
compute a final mean control value for each physical prop- tobequalitativelysimilaratallredshifts,withslightdiffer-
erty. The distribution of various physical properties for the encesinthepeakoftheirdistributions.Toaccentuatethese
mass-limitedsampleofAGNandcontrolsampleofinactive differencesandexaminetheirimpact,wecomparethequies-
galaxies is shown in Figure 6. centfraction(f =N /(N +N ))anddustystar-formerfrac-
q q q sf
tion (f = N /N ) of both samples in Figure 7 (lower
dusty dusty sf
panels). While low numbers in the radio AGN population
hindertheabilitytoproducestatisticallysignificantresults,
5 RESULTS offsets are observed between the IR and X-ray AGN hosts
and their respective control samples. For both populations
5.1 Comparison of Rest-frame Colours
overallredshifts,thedustyfractionisfoundtobeslightlyel-
Examining the rest-frame UVJ colours of galaxies has be- evatedoverthecontrolsamples,whilethequiescentfraction
come a common approach to distinguish a quiescent pop- is lower.
ulation from a star-forming one, including those exhibiting Together, all panels in Figure 7 reveal no significant
heavy extinction (e.g. Labb´e et al. 2005; Wuyts et al. 2007; differencesbetweentheUVJcoloursofourAGNandcontrol
Williamsetal.2009).ReferringtothetoppanelinFigure7, samples, with the exception that the AGN hosts tend to
quiescentgalaxiesoccupytheupperleftregion,delimitedby be more dusty and hosted in a lower fraction of quiescent
thevertices(V−J,U−V)=(−∞,1.3),(0.85,1.3),(1.6,1.95), galaxies. In the following section, we explore these results
(1.6,+∞), while the vertical dashed-line (V−J = 1.2) sepa- in further detail by quantitatively gauging the difference in
ratesnon-dusty(lowerleft)fromdustystar-forminggalaxies star-formation activity between both samples.
(Spitler et al. 2014).
Withinthisfigure,weexaminetheUVJ colourspaceof
our mass-limited AGN hosts and control sample of inactive
5.2 Comparison of Star-Formation Activity
galaxies.Inthelowestredshiftbin(z=0.2−0.8),wefindthe
UVJ colours of each subsample of AGN, identified in radio, We now focus on the star-formation activity in our mass-
X-rayorIR,tobeconsistentwithadistinctgalaxypopula- limited AGN hosts and how they compare to the control
tion.IRAGNarefoundexclusivelyinstar-forminggalaxies, sample of inactive galaxies. We use specific star-formation
radio AGN in quiescent galaxies, and X-ray AGN in both rate (sSFR) as a measure of the relative strength of star-
quiescent (29.0%±8.2%) and star-forming hosts. However, formation activity, which is a galaxy’s SFR normalised by
at higher redshifts (z>0.8), the trend weakens and the dis- the mass of its stars (ΨIR+UV/M∗). In Figure 8, we show the
tributionofUVJ coloursscattertothepointwhereallAGN meansSFRagainststellarmassforourAGNhostsandcon-
arepredominantlyfoundinthecolourspaceofstar-forming trol sample in bins of redshift. The mean sSFR is found to
hosts(radioAGN;57.1%±13.2%,X-rayAGN;79.0%±4.1%, decrease with increasing stellar mass for all sources, with
IRAGN;91.2%±3.8%),mirroringthebehaviourofthecon- slight offsets observed between the AGN hosts and control
trol sample of galaxies. sample. AGN hosts exhibit an elevation over the control
MNRAS000,1–16(2015)
Comparison of active and inactive galaxies in ZFOURGE 9
0.4
ZFOURGE Galaxies
Control Sample
0.3
AGN Hosts
t
o
T
N 0.2
/
N
0.1
0.4
0.3
t
o
T
N 0.2
/
N
0.1
−2 2
0 1 2 3 10 11 12 10 1 10
Redshift Log(M [M ]) SFR
∗ ⊙ 160µm
Figure 6.Theredshift(left),stellarmass(middle)andSFR(right,limitedtopositivefluxes)distributionsfortheparentpopulationof
galaxies(toprow,hatched),controlsampleofinactivegalaxies(bottomrow,hatched),andluminositylimitedAGNhosts(solidorange
line)inZFOURGE.
sample,withanaveragelogarithmicoffset(linearaverageof tion of UV and IR luminosities and may contain a mixed
the difference between the logarithmic sSFRs) for the com- contribution of light from stars and AGN. We first exam-
binedmassbinsof0.26±0.14dexatz=0.2−0.8,0.37±0.10 inetheimpacttotheUVbyremovingtheUVcontribution
dex at z=0.8−1.8, and 0.38±0.10 dex at z=1.8−3.2 (see to the SFRs of the AGN sample and recalculating our re-
Table 2 for more details). sults. We find the offsets increase an average of 0.01 dex
To better understand the source of this offset, we split in each redshift bin, suggesting there is negligible impact
the AGN population by detection technique (i.e. radio, X- from AGN contamination in the UV regime. If we assume
ray and infrared). In Figure 9, the mean sSFR of each sub- contamination to the IR regime wholly explains the eleva-
sample of AGN hosts, along with their respective control tionofstar-formationactivityobservedinourAGNsample,
sample is shown. It can be seen that each subsample ex- the contribution from AGN emission would need to be in
hibits an elevated level of sSFR over their control samples, excess of ∼25%. However, the FIR regime is thought to be
withtheexceptionoflow-redshiftradioAGNhosts.Foreach mostlyimmunetotheeffectsofAGN(e.g.Netzeretal.2007;
subsample, the elevation is found to increase with redshift. Mullaney et al. 2011), which is the primary motivation for
This elevation is found to be consistently high and signifi- employing PACs-based SFRs.
cant for IR AGN (0.48±0.21 dex; z = 0.2−0.8, 0.50±0.12
The other potential impact is AGN contamination to
dex;z=0.8−1.8,0.72±0.13dex;z=1.8−3.2),butlowerand
stellar masses. Ciesla et al. (2015) inspected this by omit-
insignificant for X-ray AGN (0.15±0.13 dex; z = 0.2−0.8,
tinganAGNcomponentwhileperformingSEDfittingona
0.21±0.13 dex; z = 0.8−1.8, 0.25±0.16 dex; z = 1.8−3.2).
range of Type-I, intermediate type, and Type-II AGN and
While high redshift radio AGN hosts (z > 0.8) also present
comparingthemeasuredstellarmasstothetruevalue.Their
anelevatedsSFRoverthecontrolsample,lownumberstatis-
results showed contamination from a Type-I AGN can lead
tics impact its significance (−0.53±0.20 dex; z = 0.2−0.8,
toanoverestimationinmassbyasmuchas150%.Thecon-
0.57±0.20dex;z=0.8−1.8,0.55±0.32dex;z=1.8−3.2(see
taminationfromintermediateandType-II,believedtodom-
Table 3 for more details).
inatethesampleinthisstudy,wasoverestimatedby∼50%.
We examine the most extreme of these cases (150% overes-
timation) and how it impacts our results. We first reduce
5.3 AGN Contamination
the mass of our AGN population and then re-sample our
AsdiscussedinSection2.5,thereispotentialforAGNcon- mass-matched control sample of inactive galaxies. We find
taminationtoimpactgalaxypropertiesusedinthisanalysis. the total average logarithmic offset between active and in-
The first is our SFRs, which are derived from a combina- activegalaxiestodecreasefrom0.34±0.07dexto0.25±0.07
MNRAS000,1–16(2015)
10 M. J. Cowley et al.
2.5
Radio AGN
] X-ray AGN
g
a 2.0 Infrared AGN
M
[
Quiescent
V 1.5
-
U

e
m 1.0
a Dusty
r
f Star-Forming Star-
t
es 0.5 Forming
R
0.2<z<0.8 0.8<z<1.8 1.8<z<3.2
0
0 0.5 1.0 1.5 2.0 0 0.5 1.0 1.5 2.0 0 0.5 1.0 1.5 2.0
Restframe V-J [Mag] Restframe V-J [Mag] Restframe V-J [Mag]
1.0 1.0
z=0.2-0.8 .
c
z=0.8-1.8 a
z=1.8-3.2 r
0.8 f 0.8
c.
a er
fr m
0.6 r 0.6
t o
n f
e -
sc 0.4 ar 0.4
uie st
q 0.2 y 0.2
t
s
u
d
0 0
IR X-RAY RADIO IR X-RAY RADIO
Figure 7. (Top) The rest-frame UVJ colour classification of galaxies in bins of redshift (z = 0.2−0.8; left, z = 0.8−1.8; middle and
z=1.8−3.2; right). The points represent the mass-limited (log(M∗/M(cid:12)) ≥9.75) AGN hosts selected via radio (green diamonds), X-ray
(blue squares) and IR (red circles) techniques. A representation of the control sample is shown by the grey-scale density plot in each
panel.Thesolidlinedividesthepopulationintoquiescentandstar-forminghosts,whilethedashedlinefurtherdividesthestar-forming
population into dusty and non-dusty galaxies. (Lower-left) The quiescent fraction (Nq/(Nq+Nsf)) and (lower-right) dusty star-former
fraction (Ndusty/Nsf) for the mass-limited AGN hosts (closed markers) and the control sample (open markers) at z=0.2−0.8 (diamond
markers),z=0.8−1.8(circlemarkers)andz=1.8−3.2(squaremarkers).ValuesarederivedfromtheUVJ colourclassification.Vertical
errorbarsindicatethe1σClopper-Pearsonconfidenceintervals.Unlessshown,errorbarsaresmallerthantheplottingsymbolsforthe
controlsample.
Table 2.MeansSFR(Gyr−1)valuesbyredshiftbin(rows)forAGNhostsandthecontrolinbinsofstellarmass(cols)
AGNHosts ControlSample
Redshift 109.75−10.25M(cid:12) 1010.25−10.75M(cid:12) 1010.75−11.25M(cid:12) 109.75−10.25M(cid:12) 1010.25−10.75M(cid:12) 1010.75−11.25M(cid:12)
z=0.2−0.8 0.79±0.39 0.70±0.25 0.06±0.02 0.50±0.02 0.29±0.01 0.06±0.01
σ=0.19 σ=0.21 σ=0.01 σ=0.11 σ=0.06 σ=0.02
z=0.8−1.5 2.68±0.64 1.48±0.52 0.43±0.13 1.33±0.03 0.58±0.01 0.20±0.01
σ=1.39 σ=0.60 σ=0.14 σ=0.28 σ=0.09 σ=0.05
z=1.5−2.5 8.42±2.83 5.57±1.54 1.09±0.34 3.55±0.20 2.05±0.05 0.93±0.03
σ=2.52 σ=2.40 σ=0.49 σ=1.16 σ=0.39 σ=0.19
Notes.Uncertaintiesarefromabootstrapanalysis.Dispersionsaroundthemean(σ)onquantitiesaremedianabsolutedeviations
(MAD).
MNRAS000,1–16(2015)