JamesAirdUniversityofCalifornia,SanDiego
KirpalNandra,EliseLaird,AntonisGeorgakakis,Ma@Ashby,PaulineBarmby,AlisonCoil,JiashengHuang,AntonKoekemoer,ChuckSteidel,ChrisWillmer
IntroducJon
ThefaintendoftheXLFatz≈4–5:preliminaryresults Conclusions
FormoreinfoseeAirdetal.2010,MNRAS,401,2531email:[email protected]
References:AdelbergerK.L.etal.,2004,ApJ,607,226;GeorgakakisA.etal.,2008,MNRAS,388,1205;LuoB.etal.,2010,ApJS,187,560;SteidelC.C.etal.,2003,ApJ,592,728
2)CorrecJngforincompleteness
3)Photometricredshibs 4)Rest‐frameUVcolourpre‐selecJonatz≈2–3
1)X‐rayfluxuncertainJesandsensiJvity
5)Bayesianmodelcomparison
0 2 4 6 8 10 12Count rate s
0.00
0.01
0.02
0.03
0.04
0.05
P(s
| N
,b)
-16 -15 -14 -13 -12 -11log(fX / erg s-1 cm-2)
10
15
20
25
30
R m
agnit
ude
(AB
)
(a)
log(fX /f
opt )=-1.0
0.0
+1.0
R=25.5
-2 -1 0 1 2log(fX/fopt)
0
20
40
60
80
100
120
Num
ber
(b)
0.1 0.2 0.5 1.0 2.0 5.0zspec
0.1
0.2
0.5
1.0
2.0
5.0
CF
HT
LS
z
phot
-0.5 0.0 0.5 1.0 1.5G-
-1
0
1
2
3
Un-G
R
QSO
GAL
NLAGN
LBG
BX
-0.5 0.0 0.5 1.0 1.5g’-r’
-1
0
1
2
3
u*-g
’
LBG
BX
-0.5 0.0 0.5 1.0 1.5r’-i’
-1
0
1
2
3
g’-
r’
z~4
0.0
0.2
0.4
0.6
0.8
1.0
P(z
)
GAL
NLAGN
QSO
u*g’r
’i’
1 2 3
0.0
0.2
0.4
0.6
0.8
P(z
)
BX LBG
1 2 3 Redshift (z)
BX LBG
1 2 3 4
BX LBG
UGR
• BayesianapproachusedtodeterminetheXLFthataccountsforthefullprobabilitydistribuJons(i.e.uncertainJes)inX‐rayfluxandredshibforeachsource,opJcalincompleteness,X‐raysensiJvityandthehigh‐zselecJonfuncJons.
• Keyfeature:Bayesianevidenceiscalculated(vianestedsamplingalgorithm),allowingrobustmodelcomparison‐penalisesmorecomplexmodels
• CompareourluminosityanddensityevoluJon(LADE)withsimplerPureLuminosityEvoluJon(PLE)andmorecomplexLuminosity‐DependentDensityEvoluJon(LDDE)
SystemaJcbias?
• Photo‐zessenJaltoobtainadequateredshibcompletenessforlow‐luminosityAGN
• Adoptphoto‐zfromCFHTLS(templatefilng,u*g’r’i’z’)andANNz(ArJficialNeuralNetworkstrainedatz<1.2)
• Photo‐zhavelargeerrorsandtheremaybemulJplepeaksintheredshibprobabilitydistribuJon(seeFig.3)
• p(z)distribuJonsareadoptedforeachsource(seealsobox5)
• Photo‐zsufferfromcatastrophicfailuresandmaybesystemaJcallybiasedatz≳1.2(seeFig.4).WethereforeadoptanalternaJveapproachathighredshibs(seebox4)
• Atz>1.2werestrictoursampletoobjectsthatsaJsfyrest‐frameUVcolourselecJoncriteriainUGRcolourspace
‐BX:z~2.3,LBGz~3,(Steideletal.2003,Adelbergeretal.2004)
• Suchsamplesarehighlyincomplete
• However,havewell‐definedselec;onfunc;ons(seeFig.6)thatallowustocorrectforthisincompleteness
• CalculatedbymodelingcolourdistribuJonofdifferentAGNhostsandsimulaJngtheobserveddata(seeFig.5)
Fig.5(above):ModeltrackswithredshibfordifferentopJcalclassificaJonsinUGRcolourspaceFig.6(le0):SelecJonfuncJonsforLBGandBXcriteriaforeachopJcalclassificaJon
GAL NLAGN QSO
high‐zUVcolourpre‐selecJon
Results
10-8
10-7
10-6
10-5
10-4
10-3
0.0 < z < 0.2
0.2 < z < 0.5
0.5 < z < 0.8
10-8
10-7
10-6
10-5
10-4
10-3
d!
/d(l
og L
X)
0.8 < z < 1.0
1.0 < z < 1.2
1.2 < z < 1.5
41 42 43 44 45 46
10-8
10-7
10-6
10-5
10-4
10-3
1.5 < z < 2.0
42 43 44 45 46log(LX / erg s-1)
2.0 < z < 2.5
42 43 44 45 46
2.5 < z < 3.5
TheXLFevolvesinluminosity…Movingtohigherluminosi9esathigherredshi0sThisstrong,posi3veevolu3onofthecharacteris3cluminosity,L*,,dominatestheevolu3onatz≲1.5
…ANDdensityDecreasinginoveralldensityasredshi0increasesAthighredshiDs(z≳1.5)theluminosityevolu3onslows,andtheevolu3onisdominatedbythisweakexponen3aldeclineindensitywithincreasingredshiD.
BUTretainsthesameshapeWefindnoevidenceforaflaNeningofthefaint‐endslopeathighredshiDs.TheXLFhasthesameshapeatallredshiDs.Wedonotrequireamorecomplexluminosity‐dependentdensityevolu3onmodel(LDDE).Previouses3matesmaybebiasedbythefailureofphotometricredshiDsatz≈1.5–3,thatcanimplyaflaNenedfaint‐endslope.
photo‐z(notusedinfilngatz>1.2)
LADEmodelevaluatedatz=0
• ManyX‐raysourcesdetectedwithfewcounts➔significantPoissonianuncertaintyintheflux
• WeaccountforthefullprobabilitydistribuJonforthefluxofeachsourcedirectlyinourXLFfilng
• AlsousedindeterminaJonoftheX‐raysensiJvityfuncJon(seeGeorgakakisetal.2008)
• AllowscorrecJonforthesignificanteffectsofEddingtonbias(seeFig1) high‐z
colourpre‐selecJon
Fig.1:ExamplePoissionprobabilitydistribuJonforX‐raycountrate(black)andthedistribuJonabercorrecJonforEddingtonbias(red)
N=5b=0.4
Fig.2:X‐rayfluxvs.opJcalmagnitudeforoursample
Fig.3:ExampleredshibprobabilitydistribuJon,p(z),foraCFHTLSphoto‐z
Fig.4:CFHTLSphoto‐zvs.spectroscopicredshibsforAEGIS‐X(200ks)sources
Fig.7:2‐10keVX‐rayluminosityfuncJonatarangeofredshibs
• LikelihoodRaJo(LR)methodisusedtoidenJfysecureopJcalcounterpartsofX‐raysources
• X‐raysourcesaresca@eredoverarangeofX‐rayflux–opJcalfluxraJos
• ForthefaintestX‐raysourceswesufferfromincompletenessassourceslackopJcalcounterparts
• IncompletenessiscorrectedforbydirectlymodelingthefX/foptrelaJoninthedeterminaJonoftheXLF,andcorrecJngforthefracJonofsourceswithopJcalcounterpartswithR>25.5
Incompleteness
• AddiJonal1.8MsofChandraJmeawardedinA09totake3oftheAEGIS‐XChandrapoinJngsto800ksdepth–AEGIS‐XD(eep)
• Providesaunique,largeareaofverydeepX‐raydata,providingasignificantsampleoflow‐LXAGNatz≳4
• X‐raysourcesarecross‐idenJfiedwithmulJwavelengthcounterpartsinopJcal,near‐ormid‐IRusingthelikelihoodraJotechnique=>97%idenJficaJonrate(seealsoLuoetal.2010)
• AperturemagnitudesextractedforopJcallyblanksources=>upperlimits
• Photo‐z’sdeterminedfrom10bandsfromUVto8.0µm
• PreliminaryresultsagreewellwithextrapolaJonofLADEmodel
41 42 43 44 45 46
10-8
10-7
10-6
10-5
10-4
10-3
d!
/d(l
og L
X)
3.5 < z < 4.5
42 43 44 45 46log(LX / erg s-1)
4.5 < z < 5.5
Fig.8:Preliminarymeasurementsofthe2‐10keVX‐rayluminosityfuncJonatz≈4andz≈5
LADEmodelatz=0
PreliminaryresultsusingAEGIS‐XDdata
ExtrapolaJonofLADEmodel
• TheluminosityfuncJonofAcJveGalacJcNuclei(AGN)isakeytracerofthedistribuJonandhistoryofaccreJonacJvityoverthelifeJmeoftheUniverse.AccuratemeasurementsouttothehighestpossibleredshibsareessenJaltoconstrainmodelsofsupermassiveblackholeformaJonandgrowth,thetriggeringandfuelingofAGN,andtheirco‐evoluJonwithgalaxies.
• X‐raysurveysprovideanefficientmethodofidenJfyingAGN,includingunobscuredandmoderatelyobscuredsources,andlow‐luminosityAGN.
• However,arangeofissuescanleadtoincompletenessinsamplesandpotenJallybiasmeasurementsoftheX‐rayluminosityfuncJon,especiallyatthefaintestluminosiJesandhighestredshibs.
• Wepresentmeasurementsofthe2—10keVX‐rayluminosityfuncJon(XLF)overawiderangeofredshibs(z≈0—3)
• UsedatafromCDF‐N(2Ms),CDF‐S(2Ms),AEGIS‐X(200ks),ASCALSSandASCAMSS
• Wemakeanumberofsignificantandimportantmethodologicalimprovements–seeboxes1—5
• WealsopresentpreliminaryresultsfromtheAEGIS‐XD(800ks)surveyatz≈4—5
• SophisJcatedmethodusedtodeterminetheevoluJonofthe2—10keVXLF,whichaccountsforuncertainitesinphotometricredshibs,thePoissoniannatureofX‐rayfluxesJmates,thefracJonofsourceswithcounterpartsbelowthemagnitudelimitsoftheopJcaldataandtheopJcalselecJonfuncJonsathighredshibs.
• WefindthattheXLFretainsthesameshapeatallredshibs,evolvingonlyinluminosityandoveralldensity.Thereisnoevidenceforafla@eningofthefaint‐endslopeathighredshibs.
• ThetotalluminositydensityofAGNpeaksatz=1.2±0.1,withamilddeclinetohigherredshibs.LowerluminosityAGNpeakinnumberdensityatlowerredshibs,butwefindasmallershibthanpriorstudies.
• TheseresultsindicatethatthesameprocessesareresponsiblefortriggeringandfuellingAGNatallredshiMs,butincreaseinoveralldensityfromtheearliest;mestoz≈1.Exhaus;onofgassuppliesinthemostmassivegalaxiesatlater;mesmayresultinthe“downsizing”ofAGNtolowerluminosi;es.
• Wefindthat>50%ofblackholegrowthtakesplaceatz>1,witharoundhalfinlow‐luminosityAGN.
