The star formation rates of active galactic nuclei host galaxies

0.44 MB·

by  Sara L. Ellison

Mon.Not.R.Astron.Soc.000,000–000(0000) Printed14January2016 (MNLATEXstylefilev2.2)
The star formation rates of active galactic nuclei host galaxies.
Sara L. Ellison1, Hossen Teimoorinia1, David J. Rosario2, J. Trevor Mendel2
1DepartmentofPhysics&Astronomy,UniversityofVictoria,FinnertyRoad,Victoria,BritishColumbia,V8P1A1,Canada.
2Max-Planck-InstitutfurExtraterrestrischePhysik,Giessenbachstrasse,D-85748Garching,Germany.
6
1
14January2016
0
2

n ABSTRACT
a Using artificial neural network (ANN) predictions of total infra-red luminosities (L ), we
J IR
compare the host galaxy star formation rates (SFRs) of ∼ 21,000 optically selected active

3 galactic nuclei (AGN), 466 low excitation radio galaxies (LERGs) and 721 mid-IR selected
1 AGN. SFR offsets (∆ SFR) relative to a sample of star-forming ‘main sequence’ galaxies

(matchedinM ,zandlocalenvironment)arecomputedfortheAGNhosts.Opticallyselected
] (cid:63)
A AGNexhibitawiderangeof∆SFR,withadistributionskewedtolowSFRsandamedian∆
SFR=−0.06dex.TheLERGshaveSFRsthatareshiftedtoevenlowervalueswithamedian
G
∆ SFR = −0.5 dex. In contrast, mid-IR selected AGN have, on average, SFRs enhanced by
h. a factor ∼ 1.5. We interpret the different distributions of ∆ SFR amongst the different AGN
p classes in the context of the relative contribution of triggering by galaxy mergers. Whereas
- theLERGsarepredominantlyfuelledthroughlowaccretionratesecularprocesseswhichare
o
notaccompaniedbyenhancementsinSFR,mergers,whichcansimultaneouslyboostSFRs,
r
t mostfrequentlyleadtopowerful,obscuredAGN.
s
a Keywords: galaxies:active,galaxies:Seyfert,galaxies:interactions
[

1
v
9
1 INTRODUCTION magneticspectrumfromtheX-rayandopticaltothemid-infrared
4
(IR) and radio (e.g. see Alexander & Hickox 2012 for a review).
3
3 Nuclearactivityisprevalentinstarforminggalaxies,indicatingthat AGN selected at different wavelengths can exhibit very different
0 a ready supply of matter for the production of stars is linked to properties in their accretion rates, environments and host galaxy
. thefuellingofactivegalacticnuclei(AGN).However,thelevelof properties(e.g.Kauffmannetal.2003;Tasseetal.2008;Hickox
1
star formation in AGN hosts, relative to galaxies that do not ex- etal.2009;Smolcic2009;Best&Heckman2012).Thedifferent
0
hibitsignsofnuclearactivity,iscontentious.StudiesofAGNhost techniquesusedinstudiesofdifferentAGNsamplesmakesitdif-
6
1 galaxies have variously shown that they exhibit elevated (Silver- ficult to fairly compare the SFRs throughout the AGN family. A
: man et al. 2009; Santini et al. 2012; Juneau et al. 2013; Rosario more complete understanding of the SFRs in AGN therefore re-
v etal.2015;Magliocchettietal.2016),normal(Hatziminaoglouet quires both large, sensitive samples for which the distribution of
i
X al. 2010; Harrison et al. 2012; Rosario et al. 2013; Stanley et al. SFRs can be fully parametrized, and a homogeneous comparison
2015;Xuetal.2015;Lanzuisietal.2015)orsuppressed(Salimet betweenmulti-wavelengthselectedAGN.
r
a al.2007;Mullaneyetal.2012;Hardcastleetal.2013;Gurkanet Oneofthefundamentalchallengesinthepursuitofahomo-
al.2015;Shimizuetal.2015;Leslieetal.2016)levelsofstarfor- geneous, multi-wavelength assessment of star formation in AGN
mation,eitherthrough‘direct’measuresofthestarformationrate, isthelackoflargesampleswithuniformlyandrobustlymeasured
SFR,orindirectlythroughgalaxycolour. SFRs.SincetheAGNitself‘contaminates’theopticalrecombina-
Atleastpartoftheapparenttensionintheliteraturemaybeas- tion lines that are commonly used to calibrate SFRs, these stud-
sociatedwithobservationalmethodsandbiasesinthedata.Forex- ieshavehadtoadoptalternativemethodstodeterminetheSFRs.
ample,sincestarformationrateisknowntocorrelatecloselywith Twocommonapproacheshaveemergedinrecentyears.Thefirst
stellarmass,andAGNtendbemoreprevalentinhighermassgalax- methodappliestheanti-correlationofD4000andspecificSFRmea-
ies,asimplecomparisonofAGNandnon-AGNwillbebiasedto- sured in star-forming galaxies (e.g. Brinchmann et al. 2004) and
wards high SFRs in the former. Comparing SFRs at fixed M is assumes that the same anti-correlation applies in AGN. The ad-
(cid:63)
therefore essential (e.g. Shimizu et al. 2015; Leslie et al. 2016). vantageoftheD4000 SFRcalibrationisthatitisreadilyappliedto
Somedisagreementbetweenstudiesmayalsoariseduetotheneed galaxieswithopticalspectroscopy,andSFRscanhencebedeter-
to either stack or average the data in small samples (Mullaney et minedforlargenumbersoflocalAGNhosts,suchasthoseinthe
al.2015).Furthermore,thenatureofstarformationinAGNmay SDSS.ThemaindisadvantageoftheD4000techniqueisthatitsun-
evolveasafunctionofredshift(Magliocchettietal.2015).Amore certaintiesarerelativelylarge(Rosarioetal.2016).
subtle point is that AGN may be selected over a broad range of MorerobustSFRsforlocalAGNmaybedeterminedfromfar
luminositiesandviaanumberofdiagnostics,spanningtheelectro- IRluminositieswhicharegenerallyconsideredtobelesscontam-
(cid:13)c 0000RAS
2 Ellisonetal.
inatedbyAGNthanothertracers(e.g.Netzeretal.2007;Buatet from the SDSS for star-forming galaxies. In this paper, we addi-
al.2010;Hatziminaoglouetal.2010;Mullaneyetal.2011).How- tionallyutilizetheL valuesoftheEllisonetal.(2016)catalogto
IR
ever,theuseofIRSFRdiagnosticstowidelycharacterizethelo- determineSFRsforAGNhostgalaxies;indeed,L SFRsareused
IR
cal galaxy population has so far been relatively limited. Until re- throughout this paper, for both star-forming and AGN dominated
cently, the main resource for large area studies of galactic IR lu- galaxies,unlessotherwisestated.Throughoutthispaperwerequire
minositiesweretheallskysurveysperformedbytheInfraredAs- σ (the uncertainty on the ANN predicted L ) to be less than
ANN IR
tronomical Satellite (IRAS) (Neugebauer et al. 1984) and AKARI 0.1dex(247,137galaxeis)andconvertL toSFRusingequation
IR
(Murakamietal.2007).However,thesesurveyssufferfrompoor 1.
angularresolutionandshallowsensitivity,leadingtoproblemsin InordertocomparetheSFRsofAGNtothoseofstar-forming
confusionanddetectionofonlythehighestSFRgalaxies.TheHer- (non-AGN)galaxies,wedefinetheSFRoffset(∆SFR)ofagiven
schelSpaceObservatory(hereafter,simplyHerschel,Pilbrattetal. galaxy(SFR )relativetoasetofcomparisonstarforming(asde-
gal
2010),hasrecentlyprovidedasignificantstepforwardintermsof finedbyKauffmannetal.2003andwithS/N>3)galaxiesthatare
bothangularresolutionandsensitivity.Severallargesurveyshave matchedinstellarmass,redshiftandlocalgalaxydensity(environ-
beenperformedwithHerschel,suchastheHerschelMulti-Tiered ment), also selected from the SDSS. Local density is defined as
Extragalactic Survey (HerMES, Oliver et al. 2012) and the Her- Σ = 5 ,whered istheprojecteddistanceinMpctothe5thnear-
5 πd2 5
schelStripe82survey(Vieroetal.2014).However,eventhelargest estneig5hbourwithin±1000kms−1.Normalizeddensities,δ ,are
5
oftheHerschelextra-galacticsurveys,theHerschelAstrophysical computedrelativetothemedianΣ withinaredshiftslice±0.01.
5
TerahertzLargeAreaSurvey(H-ATLAS,Ealesetal.2010),cov-
Thebaselinetoleranceusedformatchingis0.1dexinM ,0.005
eredonly∼550deg2,whichismuchsmallerthantheareascov- (cid:63)
inzand0.1dexinδ .Werequireatleast5comparisongalaxiesin
5
eredbyopticalspectroscopicsurveyssuchastheSDSS.Therefore,
thematchedsample;ifthisisnotachievedthenthemass,redshift
asignificantchallengeinstudyingtheSFRsofAGNhasbeenas-
andlocaldensitytolerancesaregrowninfurtherincrementsof0.1
semblingsamplesthatarebothlarge,andforwhichrobust(e.g.IR)
dex,0.005and0.1dexrespectively,untiltheminimumsizecrite-
SFRsareavailable.Largesamplesarevitalforthemeasurementof
rionof5ismet.TheSFRofthecomparisonstarformingsample
SFRdistributions,ratherthanbinnedaverageswhichcanbedomi-
(SFR ) is taken as the median of the matched control sample.
comp
natedbyoutliers(Mullaneyetal.2015).InEllisonetal.(2016)we TheSFRoffsetisthendefinedas:
presented a catalog of ∼ 330,000 infra-red luminosities (L ) for
IR
galaxiesintheSDSS,derivedbyartificialneuralnetwork(ANN) ∆SFR=logSFRgal−logSFRcomp. (2)
techniques.Inthecurrentwork,weusetheL catalogofEllison
IR
etal.(2016)toexploretheSFRsofAGNinthelocalUniverse.
3 RESULTS
3.1 OpticallyselectedAGN
2 METHODOLOGY Baldwin,Phillips&Terlevich(1981)pioneeredtheuseofemission
lineratiostodistinguishgalaxiesdominatedbyvariousphotoion-
Full details of the L determinations used in this work are pro-
IR izing processes. Diagnostics separating star-forming from AGN
vided in Ellison et al. (2016); only a brief summary is presented
dominated galaxies are now commonly referred to as ‘BPT’ dia-
here.TheHerschelStripe82Survey(Vieroetal.2014)coversato-
grams, in reference to these authors. In this paper, we make use
talof79deg2inanequatorialstripoftheSDSSfootprint.Rosario
of three of the most commonly used diagnostics which incorpo-
etal.(2016)cross-matchedthe250µmdetectedgalaxiesfromHer-
ratetheratiosof[NII]/Hαand[OIII]/Hβ,byKewleyetal.(2001),
schelwiththeSDSSDR7MainGalaxySamplewhoseextinction
Kauffmannetal.(2003)andStasinskaetal.(2006),hereafterK01,
correctedPetrosianr-bandmagnitudesarebrighterthan17.77and
K03andS06respectively.S06identifiesgalaxieswithevenasmall
whoseredshiftsareintherange0.04<z<0.15,usingapositional
AGNcontribution.K03isslightlymorerelaxedinitsselectionof
matchingtoleranceof5arcsecs.TheHerschel-SDSSmatchedcata-
AGNbutstillincludesgalaxiesthatarecompositesofstarforma-
logcontains3319galaxies,whichwerefurthercross-matchedwith
tionandAGN.AGNidentifiedbyK01arethosedominatedbythe
theWideFieldIRSkyExplorer(WISE,Wrightetal.2010)catalog
AGN.
to yield photometry spanning the mid- to far-IR. Infra-red lumi-
We begin our analysis with a single definition of ‘AGN’,
nosities were derived by Rosario et al. (2016) by fitting the mid-
adoptingK03asourfiducialclassification(althoughourresultsare
and far-IR photometry with the spectral energy distribution tem-
qualitativelysimilarwithanyofthe3diagnostics),withaS/N>5
plates of Dale & Helou (2002). AGN are expected to contribute
requirement to minimize the contribution by Low Ionization Nu-
negligibly to the L of this sample (see Rosario et al. 2016 for
IR clearEmissionlineRegions(LINERs;seeLeslieetal.2016foran
moredetails).
assessmentofSFRoffsetsintheLINERpopulation).Theseselec-
TheL isestimatedfromanartificialneuralnetworkthatuses
IR tioncriteriayield20,926AGNwithrobust(σ < 0.1dex)L
1136Herscheldetectedgalaxieswith23physicalparameters,such ANN IR
predictions.Fig.1showsthelocationofthemainsequenceofstar
asstellarmasses(Mendeletal.2014),emissionlinestrengthsand
forminggalaxiesasfilledbluecontours.TheSFRsofAGNderived
photometry(Simardetal.2011),measuredfromSDSSdata.Based
fromtheANNL areshowninredcontours.Forcomparison,we
onthepopularconversionpresentedbyKennicutt(1998),theL IR
IR alsoshowtheaperturecorrectedAGNSFRsderivedfromtheSDSS
maybeconvertedtostarformationrates(foraChabrierinitialmass
spectraviatheD method.
function)using 4000
AsfoundbypreviousauthorsusingtheD methodtomea-
4000
logL (erg/s)=logSFR(M /yr)+43.591. (1) sureSFRs(e.g.Salimetal.2007;Shimizuetal.2015;Leslieetal.
IR (cid:12)
2016), the optically selected AGN clearly have L SFRs that lie
IR
TheSFRsthatresultfromtheANNpredictionswereshown belowthemainsequence,i.e.AGNhavelowSFRsfortheirM ,
(cid:63)
by Ellison et al. (2016) to be in excellent agreement with those typicallyby0.05–0.1dex,withamedian∆SFR=-0.06.Themain
(cid:13)c 0000RAS,MNRAS000,000–000
SFRsofAGN 3
8.5 9.0 9.5 10.0 10.5 11.0
0.5
1.5 0.4
1.0
)
yr 0.3
/ 1.0
fl
M 0.2
R ( 0.5 Hβ 0.5 0.1
g SF 0.0 OIII]/ 0.0
Lo 0.5 og [ 0.0 0.1
L
0.2
0.0
R 0.1 0.5 0.3
F
S 0.2
∆ 0.3 0.4
0.4
8.5 9.0 9.5 10.0 10.5 11.0 1.0 0.5
1.5 1.0 0.5 0.0 0.5
Log (M /M ) Log [NII]/Hα
fl
Figure2.BPTdiagramcolourcodedby∆SFRforAGN.The3blacklines
Figure1.Upperpanel:Acomparisonofstarformationrates(fromANN showtheAGNdemarcationsof(fromtoptobottom),K01,K03andS06.
LIR estimates) in star forming galaxies (filled blue contours) and AGN Arandomsampleof5000outof∼33,300AGNselectedbyS06withLIR
(open contours) classified according to Kauffmann et al. (2003). Lower basedSFRsisshown.
panel:TheoffsetbetweenSFRsinAGN(asmeasuredfromLIR)relative
tothemainsequence.Inbothpanels,redandgreencoloursrefertoSFRs
al.(2015)haveinvestigatedtheeffectofstarformation‘dilution’,
derivedfromANNLIRpredictionsandfromD4000measurements,respec-
tively. wherebyhighSFRscanmoveevenmoderatelypowerfulAGNonto
thestarformingbranchoftheBPTdiagram.Thedilutioneffectis
mostproblematicforlowmassgalaxies.However,asshowninthe
lowerpanelofFig.1themedianvalueof∆SFRinopticallyse-
sequenceoffsetsmeasuredusingD extendedtosomewhatlower
4000 lectedAGNdoesnotdependstronglyonM ,indicatingthatany
SFRsthantheL derivedvalues.Thereareatleasttwopossible (cid:63)
IR bias against the identification of highly star-forming AGN is not
reasonsforthis.First,theSFRsderivedfromtheANNarelimited
drivingthetendencytowardssuppressedSFRs.
toabout0.3M yr−1,duetothedetectionlimitinthetrainingset.
(cid:12) Fortheremainderofthispaper,weusetheK03classification
It is therefore likely that there is a further tail of low SFR AGN
asourfiducialdefinitionof‘optical’AGN.
that are not included in our sample due to an effective ‘detection
limit’intheANNL predictions(e.g.Leslieetal.2016).Asec-
IR
ondpossiblereasonforthemoreextendedgreencontoursisthat,as
3.2 AGNselectedintheIRandradio
shownbyRosarioetal.(2016)andEllisonetal.(2016),theD
4000
basedSFRsaremuchmoreuncertainthanthosederivedfromL , Thephysicalproperties,suchasstellarmassdistribution,morpho-
IR
sothelowerSFRsintheformermaybearesultofabroadererror logicalstructureandaveragecolours,ofAGNhostgalaxiesdepend
disribution. stronglyonhowtheAGNareselected(e.g.Hickoxetal.2009).In
Inordertoinvestigatethedependenceof∆SFRontherela- thissub-section,wedetermineSFRoffsetsfromthemainsequence
tivecontributionoftheAGN,inFig.2weplottheBPTdiagramfor fortwoothersamplesofAGN,oneselectedbasedonradioemis-
AGNcolour-codedby∆SFR.Inordertoexplorethedependenceof sion,andtheotherbasedonmid-IRcolourselection.Thenovelty
∆SFRoverthemaximumrangeontheBPTdiagram,we‘relax’the ofourworkisthatwecanmakeadirectcomparisonofSFRoffsets
definitionofAGNbyadoptingthecriteriaofS06,butalsoshowing forthesethreepopulationsbasedonahomogeneousanalysis.
thedemarcationlinesfoK01andK03.Thereare∼33,300galax- The sample of low luminosity radio-selected AGN, specifi-
ies that are classified as AGN by the S06 criterion for which we callythelowexcitationradiogalaxies(LERGs),istakenfromthe
haveANNL predictionsandforwhich∆SFRcanbedetermined. SDSScatalogcompiledbyBest&Heckman(2012).Wedetermine
IR
To avoid crowding on Fig. 2 we plot a random sample of 5000 the SFRs of the Best & Heckman LERG catalog from our ANN
AGN. For reference, all three commonly used diagnostics (K01, LIRpredictions;thereare466LERGsforwhichσ <0.1.
ANN
K03,S06)areshowninFig.2asblacklines.Negativevaluesof∆ The sample of mid-IR selected AGN is identified by cross-
SFR(lowSFRs)dominatethroughouttheAGN‘wing’.Theexcep- matchingtheWISE AllSkySurvey(Wrightetal.2010)withthe
tionisveryclosetotheS06demarcation,wherethereappeartobe SDSS and requiring a maximum angular separation of 6 arcsecs.
fewergalaxieswithsuppressedSFRs.Despitethegeneralprefer- We use the W1 and W2 band profile magnitudes, with a mini-
enceforAGNtoexhibitlow∆SFRs,thereisawidevarietyofSFR mumS/Nrequirementof5andidentifyAGNasgalaxiesexceeding
enhancementsatmostlocations.Fig.2demonstratesthat,although W1−W2>0.8(Sternetal.2012).721oftheseWISEselectedAGN
thereisatrendtowardsmoresuppressedSFRsaswemovealong haveANNL withσ <0.1.
IR ANN
theAGNbranch(seealsoLeslieetal.2016),evengalaxiesthatare InFig.3weshowtheSFRoffsetsofthethreeAGNsamples
dominated by AGN can sometimes exhibit strong SFR enhance- (colouredhistograms),aswellasthedistributionof∆SFRamongst
ments(seealsothemiddlepanelofFig.3).Animportantcaveatto star-forminggalaxies(blackline).Bydefinition,thislatterdistribu-
theseconclusionsisthatsamplesofopticallyselectedAGNmaybe tionshouldbecentredat∆SFR=0,sincethestarforminggalax-
inherentlybiasedagainstgalaxieswithveryhighSFRs.Trumpet iesdefinethemainsequencefromwhichoffsetsarecomputed;the
(cid:13)c 0000RAS,MNRAS000,000–000
4 Ellisonetal.
inmid-IRselectedAGNenhancedrelativetothemainsequence,
whereasLERGshaveSFRsthatarealmostuniversallybelowex-
2.5 N(AGN)=466 SF pectationsfortheirM (seealsoGurkanetal.2015)?Hardcastle
(cid:63)
2.0 LERG AGN etal.(2013)posedasimilarquestionafterfindingthathighexcita-
1.5
tionradiogalaxies(HERGs)hadenhancedSFRs,incontrasttothe
1.0 lowSFRsinLERGs.Hardcastleetal.(2013)proposedthisdiffer-
0.5
encewasduetotheenvironmentsinwhichLERGsanHERGsare
0.0
) typicallylocated(Best&Heckman2012).However,sinceourstar-
ed 2.5 N(AGN)=20,926 SF formingcomparisonsampleismatchedinδ ,thisseemsunlikely
aliz 2.0 Opt. AGN tobetheexplanationforthedifferent∆SFR5distributionsseenin
m 1.5
Fig.3.
or 1.0
n Our finding that IR, optical and radio-selected AGN form a
N ( 0.5 sequence in ∆ SFR is consistent with an evolutionary picture in
0.0
which an abundant gas supply triggers enhanced star formation
2.5 N(AGN)=721 SF
in an obscured AGN phase. As the fuel supply declines, so does
2.0 mid-IR AGN
the star formation rate and the obscuration. A similar picture has
1.5
been proposed by Hickox et al. (2009), based on a comparison
1.0
between the host galaxy and clustering properties of a sample of
0.5
IR,X-rayandradioselectedAGN.Hickoxetal.(2009)foundthat
0.0
1.5 1.0 0.5 0.0 0.5 1.0 1.5 AGNselectedinthesewaysformedasequenceindecreasingstar
formation and accretion rate, and increasing halo mass. The evo-
∆ SFR
lutionarysequencebetweenstarformationandobscurationisfur-
ther supported by the observation that, for samples of X-ray se-
lectedAGN,starformationratesaresignificantlyenhancedforthe
Figure3.Distributionsof∆SFRforvariousclassesofAGN.Inallpanels, obscured/absorbed AGN, but are normal or suppressed for unob-
thedistributionof∆SFRinstarforminggalaxiesisshowninblack;by scured AGN (Stevens et al. 2005; Juneau et al. 2013). Moreover,
definition,thisdistributioniscentredaround∆SFR=0. theX-rayabsorbedAGNfractionincreasesasgalaxieslieprogres-
sivelyabovethestarformingmainsequence(Juneauetal.2013).
width of the histogram simply shows the scatter in the main se- Galaxy mergers are a natural mechanism to explain such an
quence.IntheupperpanelofFig.3weconfirmtheresultofGurkan evolutionarysequence,asfirstproposedbySandersetal.(1988),
et al. (2015), who studied LERGs in the H-ATLAS sample, that sincemergersareknowntobecapableoftriggeringbothstarfor-
LERGs are offset to lower SFRs than the star forming main se- mation(e.g.Scudderetal.2012)andAGNactivity(e.g.Ellisonet
quence.TheopticallyselectedAGN(K03)areshownintheblue al.2011).Inthisscenario,themergerremnantofagas-rich,ma-
histogramofthemiddlepanelofFig.3.Bycomparingthetopand jor merger first undergoes a significant starburst, possibly mani-
middlepanelsofFig.3itcanbeseenthat,whilstboththeLERGs festingasaluminous(orultra-luminous)IRgalaxyandhighaccre-
andtheopticallyselectedAGNhaveamodal∆SFRthatisneg- tionrateAGN(possiblyaquasar).Asthestarburstfades,theAGN
ative,thedistributionsarequitedifferent.Onaverage,theLERGs transitionsfromobscuredtounobscuredandthegalaxyeventually
are shifted to much lower values of ∆ SFR than the optically se- evolves to an early type (e.g. see Fig. 6 in Alexander & Hickox
lectedAGN;themedianoffsetis∆SFR=−0.5.Whilst65percent 2012foraschematicviewofthisprocess).
of optically selected AGN lie within a factor of two of the main ThedependenceofstarformationrateenhancementonAGN
sequence,andcanevenexhibitenhancedSFRs,theLERGsrarely obscurationduringthepost-mergerphaseissupportedbytheobser-
attainthemainsequenceSFRfortheirstellarmass. vationthatcertainclassesofAGNaremoreprevalentamongston-
InthelowerpanelofFig.3weshowthedistributionofSFR goingmergers.Forexample,inclosepairsandrecentpost-mergers,
offsetsforAGNidentifiedinthemid-IR.Forthefirsttime,wecan opticallyselectedAGNareover-abundant(relativetotheAGNfre-
nowcomparetheSFRoffsetsinahomogeneouswayforthethree quencyinisolatedgalaxies)byafactorof2–3(Ellisonetal.2013),
AGN populations. In contrast to the optically selected AGN and comparedtoafactorof5–20inmid-IRselectedAGN(Satyapalet
theLERGs,themedian∆SFRofmid-IRselectedAGNis+0.17 al.2014).Mid-IRselectedAGNhavealsobeenshowntoexhibit
dex,indicatingthatthetypicalpowerful,dust-obscuredAGNhas a high dual black hole fraction (Secrest et al., in preparation), in
a factor of ∼ 1.5 excess SFR compared with the main sequence. contrast to optical searches which have a low success rate in de-
However,thetailofthe∆SFRdistributionextendsupto+1dex, tectingbinaries(Fuetal.2011;Muller-Sanchezet.al2015).The
indicatingthatsomemid-IRselectedAGNhaveSFRsafactorof connectionbetweenobscurationandmergingisalsofoundinthe
10abovetheircomparisonsample.SuchelevatedSFRsareabsent X-ray,wherethemostabsorbedAGNarefoundmostfrequentlyin
intheLERGsample,andrareamongsttheopticallyselectedAGN. mergers(Kocevskietal.2015;Lanzuisietal.2015).
In Ellison et al. (2016) we showed that the ANN predictions of IncontrasttotheconsistentlinkbetweenobscuredAGNand
L wereaccurateevenwhentheSFRwaselevatedtotheselevels mergers,theresultsfortheunobscuredpopulationhasbeenmore
IR
abovethemainsequence. controversial.WhereasanexcessofAGNismeasuredinclosepairs
ofgalaxies(Ellisonetal.2011;Silvermanetal.2011),demonstrat-
ingthatmergersmaytriggerAGN,themorphologicalclassification
ofAGNindicatesthatmergersdonotappeartodominatethetrig-
4 DISCUSSIONANDCONCLUSIONS
geringofunobscuredAGN(e.g.Gaboretal.2009;Kocevskietal.
Whatdrivesthedifferencebetweenthe∆SFRdistributionsofthe 2012).TheseresultsareconsistentwiththedistributionofSFRsin
3typesofAGNshowninFig.3?Inparticular,whyaretheSFRs opticallyselectedAGN(Fig3,middlepanel),inwhichthemajor-
(cid:13)c 0000RAS,MNRAS000,000–000
SFRsofAGN 5
ityofgalaxieshavenormal-to-lowSFRs,andarehenceunlikelyto Ellison,S.L.,Patton,D.R.,Mendel,J.T.,Scudder,J.M.,2011,MNRAS,
belinkedtomergers.ThesmallnumberofopticalAGNwithen- 418,2043
hanced SFRs may be those that are merger triggered, a postulate Ellison,S.L.,Teimoorinia,H.,Rosario,D.J.,Mendel,J.T.,2016,MNRAS,
thatisconfirmedthroughvisualinspectionoftheSDSSimages. 455,370
Fu,H.,Myers,A.D.,Djorgovski,S.G.,Yan,L.,2011,ApJ,733,103
In contrast to optical and mid-IR selected AGN, there is no
Gabor,J.M.,etal.,2009,ApJ,691,705
excessofLERGsinmergersamples,indicatingthattheyarefuelled
Gurkan,G.,etal.,2015,MNRAS,452,3776
bysecularprocesses,wherelowlevelsofaccretion(notconnected
Hardcastle,M.J.,etal.,2013,MNRAS,429,2407
withastarburst)canbemaintainedfromacombinationofstellar
Harrison,C.M.,etal.,2012,ApJ,760,L15
andexternalsources(Ellison,Hickox&Patton2015).Ratherthan Hatziminaoglou,E.,etal.,2010,A&A,518,L33
an ‘evolutionary’ (i.e. time sequence) scenario, our observations Hickox,R.C.,etal.,2009,ApJ,696,891
maythereforealsobeexplainedbydifferentrelativecontributions Juneau,S.,2013,ApJ,764,176
ofmergerstothe3AGNclasses.I.e.LERGshavethelowestSFRs Kauffmann,G.,etal.,2003,MNRAS,346,1055
becausethereisnoboostfrommergers,whereasahigherfraction Kennicutt,R.C.1998,ARA&A,36,189
of IR selected AGN are linked to mergers and have experienced Kewley,L.J.,Dopita,M.A.,Sutherland,R.S.,Heisler,C.A.,Trevena,J.,
2001,ApJ,556,121
triggered star formation during the interaction. Put another way,
Kocevski,D.,etal.2012,ApJ,744,148
mergersmayleadtohighSFRsandluminous(obscured)AGN,but
Kocevski,D.,etal.2015,ApJ,814,104
adistinct,secularpathwayisresponsibleforlowerluminosityAGN
Lanzuisi,G.,etal.,2015,A&A,573,137
withnormalorlowSFRs(e.g.Shaoetal.2010).
Leslie,S.K.,Kewley,L.J.,Sanders,D.B.,Lee,N.,2016,MNRAS,455,
Theideathatthedistributionof∆SFRinthedifferentAGN L82
classesisbydrivenbytherelativecontributionsofmergersissup- Magliocchetti,M.,Lutz,D.,Santini,P.,Salvato,M.,Popesso,P.,Berta,S.,
ported by differences in their merger fractions. We compute the Pozzi,F.,2016,MNRAS,456,431
mergerfractionforeachAGNclassasthefractionofgalaxiesclas- Mendel,J.T.,Palmer,M.J.D.,Simard,L.,Ellison,S.L.,Patton,D.R.,
sifiedasamergerinGalaxyZoo,requiringamergervotefraction> 2014,ApJS,210,3
0.4(e.g.Dargetal.2010).Mergerfractionsforthedifferenttypes Mullaney,J.R.,Alexander,D.M.,Goulding,A.D.,Hickox,R.C.,2011,
MNRAS,414,1082
of AGN are: radio - 1 per cent, optical - 3 per cent, mid-IR - 7
Mullaney,J.R.,etal.,2012,ApJ,419,95
percent.Thehighermergerfractionamongstthemid-IRsampleis
Mullaney,J.R.,etal.,2015,MNRAS,453,L83
despitearedshiftdistributionskewedtowardsslightlyhigherval-
Muller-Sanchez,F.,Comerford,J.M.,Nevin,R.,Barrows,R.S.,Cooper,
uesthantheopticalAGNandLERGsamplesmakingitpotentially
M.C.,Greene,J.E.,2015,ApJ,813,103
harder to identify mergers therein. Although these merger frac- Murakami,H.,etal.,2007,PASJ,59S,369
tions should not be considered as absolute, due to the limitations Netzer,H.,etal.,2007,ApJ,666,806
of visual classification in shallow ground-based images, without Neugebauer,G.,etal.,1984,ApJ,278,L1
removalofprojectedcompanions,theirrelativevaluesareindica- Oliver,S.J.,etal.,2012,MNRAS,424,1614
tivethatmergersaremostprevalentamongstIRselectedAGN,and Pilbratt,G.L.,etal.,2010,A&A,518,1
leastcommonamongstLERGs.Acompleteassessmentofmergers Rosario,D.J.,etal.,2013,A&A,560,72
Rosario,D.J.,etal.,2015,A&A,573,85
inthissampleisbeyondthescopeofthispaper,butwenotethat
Rosario,D.J.,Mendel,J.T.,Ellison,S.L.,Lutz,D.,Trump,J.R.,2016,
theexcessofmergersamongstmid-IRselectedpersistsforstricter
MNRAS,inpress
cutsonmergervotefraction.
Salim,S.,etal,2007,ApJS,173,267
In conclusion, our main result is that the SFRs of AGN are
Sanders, D. B., Soifer, B. T., Elias, J. H., Madore, B. F., Matthews, K.,
criticallydependentonhowanAGNisselected.IRselectedgalax- Neugebauer,G.,Scoville,N.Z..,1988,ApJ,325,74
iesintheSDSSshowamedianSFRenhancementofafactorof∼ Santini,P.,etal.,2012,A&A,540,109
1.5,comparedtoanunder-abundanceofstarformationinoptically Satyapal,S.,Ellison,S.L.,McAlpine,W.,Hickox,R.C.,Patton,D.R.,
selectedAGN(by∼25percent)andlowluminosityradio-selected Mendel,J.T.,2014,MNRAS,441,1297
AGN(byafactorof3).Weproposethatthesedifferencesmaybe Scudder,J.M.,Ellison,S.L.,Torrey,P.,Patton,D.R.,Mendel,J.T.,2012,
explainedbyarelativelyhighfractionofmergersamongstIRse- MNRAS,426,549
Shao,L.,etal.,2010,A&A,518,L26
lectedAGN,alowermergerincidenceamongstopticallyselected
Shimizu,T.T.,Mushotzky,R.F.,Melendez,M.,Koss,M.,Rosario,D.J.,
AGN,andthedominationofsecularfuellingprocessesforLERGs,
2015,MNRAS,452,1841
whicharehostedprimarilybymassivegalaxieswithlowSFRs.
Silverman,J.D.,etal.,2009,ApJ,696,396
Simard,L.,Mendel,J.T.,Patton,D.R.,EllisonS.L.,McConnachie,A.W.,
2011,ApJS,196,11
Smolcic,V.,2009,ApJ,699,L43
REFERENCES
Stanley,F.etal.,2015,MNRAS,453,591
Alexander,D.M.,Hickox,R.C.,2012,NewAR,56,93 Stasinska,G.,CidFernandes,R.,Mateus,A.,Sodre,L.,Asari,N.V.,2006,
Baldwin,J.A,;Phillips,M.M.,Terlevich,R.,1981,PASP,93,5 MNRAS,371,972
Best,P.N.,Heckman,T.M.,2012,MNRAS,421,1569 Stern,D.,etal.,2012,ApJ,753,30
Brinchmann,J.,Charlot,S.,White,S.D.M.,Tremonti,C.,Kauffmann,G., Stevens,J.A.,Page,M.J.,Ivison,R.J.,Carrera,F.J.,Mittaz,J.P.D.,Smail,
Heckman,T.,Brinkmann,J.,2004,MNRAS,351,1151 Ian,McHardy,I.M.,2005,MNRAS,360,610
Buat,V.,etal.,2010,MNRAS,409,1 Trump,J.R.,etal.,2015,ApJ,811,26
Dale,D.A.,&Helou,G.,2002,ApJ,576,159 Tasse,C.,Best,P.N.,Rottgering,H.,LeBorgne,D.,2008,A&A,490,893
Darg,D.W.,etal.,2010,MNRAS,401,1552 Viero,M.P.,etal.,2014,ApJS,210,22
Eales,s.,etal.,2010,PASP,122,499 Wright,E.L.,etal.,2010,AJ,140,1868
Ellison,S.L.,Mendel,J.T.,Patton,D.R.,Scudder,J.M.,2013,MNRAS, Xu,L.,Rieke,G.H.,Egami,E.,Pereira,M.J.,Haines,C.P.,Smith,G.P.,
453,3627 2015,ApJS,219,18
Ellison,S.L.,Patton,D.R.,Hickox,R.C.,2015,MNRAS,451,L35
(cid:13)c 0000RAS,MNRAS000,000–000