Quasi-stellar objects in the ALHAMBRA survey

17 Pages·2012·1.33 MB·English

A&A542,A20(2012) Astronomy
DOI:10.1051/0004-6361/201118111 &
(cid:2)c ESO2012 Astrophysics
(cid:2)
Quasi-stellar objects in the ALHAMBRA survey
I. Photometric redshift accuracy based on 23 optical-NIR filter photometry
I.Matute1,I.Márquez1,J.Masegosa1,C.Husillos1,A.delOlmo1,J.Perea1,E.J.Alfaro1,A.Fernández-Soto2,
M.Moles1,3,J.A.L.Aguerri4,T.Aparicio-Villegas1,N.Benítez1,T.Broadhurst5,J.Cabrera-Cano1,6,F.J.Castander7,
J.Cepa4,8,M.Cerviño1,D.Cristóbal-Hornillos1,3,L.Infante9,R.M.GonzálezDelgado1,V.J.Martínez10,11,
A.Molino1,F.Prada1,andJ.M.Quintana1
1 InstitutodeAstrofísicadeAndalucía(CSIC),GlorietadelaAstronomías/n,18008Granada,Spain
e-mail:[matute;isabel;pepa;cesar;chony;jaime;emilio;benitez;mcs;rosa;amb;fprada;quintana]@iaa.es
2 InstitutodeFísicadeCantabria(CSIC-UC),39005Santander,Spain;e-mail:[email protected]
3 CentrodeEstudiosdeFísicadelCosmosdeAragón(CEFCA),44001Teruel,Spain;e-mail:[moles;dch]@cefca.es
4 InstitutodeAstrofísicadeCanarias,LaLaguna,Tenerife,Spain;e-mail:[email protected]
5 SchoolofPhysicsandAstronomy,TelAvivUniversity,Israel;e-mail:[email protected]
6 FacultaddeFísica.DepartamentodeFísicaAtómica,MolecularyNuclear,UniversidaddeSevilla,Sevilla,Spain
e-mail:[email protected]
7 InstitutdeCiènciesdel’Espai,IEEC-CSIC,Barcelona,Spain;e-mail:[email protected]
8 DepartamentodeAstrofísica,FacultaddeFísica,UniversidaddelaLaguna,Spain;e-mail:[email protected]
9 DepartamentodeAstronomía,PontificiaUniversidadCatólica,Santiago,Chile;e-mail:[email protected]
10 Departamentd’AstronomíaiAstrofísica,UniversitatdeValència,Valencia,Spain;e-mail:[email protected]
11 ObservatoriAstronòmicdelaUniversitatdeValència,Valencia,Spain
Received16September2011/Accepted7February2012
ABSTRACT
Context.Eventhespectroscopiccapabilitiesoftoday’sgroundandspace-basedobservatoriescannotkeepupwiththeenormousflow
ofdetections(>105 deg−2)unveiledinmoderncosmological surveysas:i)wouldberequiredenormous telescopetimetoperform
thespectroscopic follow-upsandii)spectraremainunattainable forthefainterdetectedpopulation. Inthepastdecade, thetypical
accuracy ofphotometric redshift (photo-z) determination hasdrasticallyimproved. Nowdays, ithas becomeaperfect complement
tospectroscopy, closingthegapbetweenphotometricsurveys andtheirspectroscopic follow-ups. Thephoto-zprecision foractive
galacticnuclei(AGN)hasalwayslaggedbehindthatforthegalaxypopulationowingtothelackofpropertemplatesandtheirintrinsic
variability.
Aims.Our goal is to characterize the ability of the Advanced Large, Homogeneous Area Medium-Band Redshift Astronomical
(ALHAMBRA) survey in assigning accurate photo-z’s to broad-line AGN (BLAGN) and quasi-stellar objects (QSOs) based on
their ALHAMBRA very-low-resolution optical-near-infrared (NIR) spectroscopy. This will serve as a benchmark for any future
compilationofALHAMBRAselectedQSOsandthebasisforthestatisticalanalysisrequiredtoderiveluminosityfunctionsupto
z∼5.
Methods.WeselectedasampleofspectroscopicallyidentifiedBLAGNandQSOsandusedalibraryoftemplates(includingtheSEDs
ofAGNandbothnormalandstarburstgalaxies,aswellasstars)tofitthe23photometricdatapointsprovidedbyALHAMBRAin
theopticalandNIR(20medium-bandopticalfiltersplusthestandardJHKs).
Results.WefindthattheALHAMBRAphotometryisabletoprovideanaccuratephoto-zandspectralclassificationfor∼88%ofthe
170spectroscopicallyidentifiedBLAGN/QSOsover2.5deg2indifferentareasofthesurveyandbrighterthanm =23.5(equivalent
678
tor ∼24.0).Thederivedphoto-zaccuracyisbelow1%andiscomparabletothemostrecentresultsinothercosmologicalfields
SLOAN
that use photometric information over a wider wavelength range. The fraction of outliers (∼12%) is mainly caused by the larger
photometricerrorsforthefaintestsourcesandtheintrinsicvariabilityoftheBLAGN/QSOpopulation.Asmallfractionofoutliers
mayhaveanincorrectlyassignedspectroscopicredshift.
Conclusions.ThedefinitionoftheALHAMBRAsurveyintermsofthenumberoffilters,filterproperties,arealcoverage,anddepth
isabletoprovidephotometricredshiftsforBLAGN/QSOswithaprecisionsimilartoanyprevioussurveythatmakesuseofmedium-
bandopticalphotometry.Inagreementwithpreviousliteratureresults,ouranalysisalsorevealsthat,inthe0<z<4redshiftinterval,
veryaccuratephoto-zcanbeobtainedwithouttheuseofNIRbroadbandphotometryattheexpenseofaslightincreaseintheoutliers.
Theimportance of NIRdataisexpected toincreaseat higher z(z > 4). Theseresultsarerelevant for thedesign of futureoptical
follow-upsofsurveyscontainingalargefractionofBLAGN,suchasmanyX–rayorradiosurveys.
Keywords.galaxies:active–cosmology:observations–quasars:general–galaxies:evolution–galaxies:distancesandredshifts
1. Introduction
(cid:2) Based on observations collected at the German-Spanish
Astronomical center,Calar Alto(Almeria,Spain), jointlyoperated by Theroleofactivegalacticnuclei(AGN)intheformationofthe
theMax-Planck-InstitutfürAstronomieatHeidelbergandtheInstituto earlystructuresandtheirlaterevolutionhasbeenreviewedover
deAstrofísicadeAndalucía(CSIC). thepast15years,becomingakeyingredientofgalaxyevolution
ArticlepublishedbyEDPSciences A20,page1of17
A&A542,A20(2012)
models (e.g. Cattaneo 2002; Menci et al. 2003; Croton et al. probethefainterdetectedpopulation,whichisdifficulttoaccess
2006; Hopkins et al. 2010, and references therein). Evidence using current ground-basedspectroscopic observatories. In ad-
shows that many, if not all, massive galaxies harborsupermas- dition,photo-zareusedtovalidateuncertainspectroscopicred-
sive black holes (SMBHs; e.g. Kormendy & Richstone 1995). shiftstypicallyobtainedforspectraoflowsignal-to-noiseratio
The close interactionbetween the formationand growth of the (S/N)orlimitedwavelengthcoverage(e.g.Fernández-Sotoetal.
SMBH and the evolution of its host galaxy were initially re- 2001). Several computationalmethods have been developed to
vealed by: i) the tight correlations between the masses of cen- derive photometric redshifts with increasingly high precision
tral SMBHs and the velocity dispersions and luminosities of (BPZ, HyperZ, LePhare, ZEBRA, AnnZ, EAzY, among others).
the bulges of many galaxies (Tremaine et al. 2002); ii) the re- Only recently have photo-z for AGN (Salvato et al. 2009; Luo
markable similarities in the redshift at which starburst and ac- et al. 2010; Cardamone et al. 2010) reached accuracies similar
cretionactivitiesocurredandiii)theobservationoftheso-called to those computed for normal and starburst galaxies (∼1–2%;
downsizing effect not only for the galaxy population but also e.g.Ilbertetal2009).Adescriptionofthecurrentstateoftheart
for AGN, i.e. the most massive galaxiesappear to have assem- photo-zcomputationaswellasadetailedperformancecompari-
bledthemajorityoftheirstarsearlierthanlowermassgalaxies sonofvariousphoto-zcodeswasprovidedbyHildebrandtetal.
(Cowieetal.1996;Zhengetal.2007),whilethedensityoflow- (2010).
luminosityAGN peaksat lower z than the more powerfulones In this context,we presentthe analysis of photometricred-
(e.g.Hasingeretal.2008,andreferencestherein).Therefore,the shiftsolutionsfoundforapopulationofspectroscopicallyiden-
measureofthespacedensityofAGNwithcosmictimenotonly tified QSOs using the optical and near-infrared (NIR) multi-
providesinformationabouttherelativeimportanceofaccretion band catalog of the ALHAMBRA survey. The ALHAMBRA
activitytotheglobalenergyoutputintheuniversebutalsoplaces survey was designed with an optimal filter combination in or-
important constraints on early structure formation and galaxy der to provide one of the most homogeneous,large, deep, and
evolution(e.g.DiMatteoetal.2005;Hopkinsetal.2010). accurate photometricsurveys. Given the proposeddepth of the
Quasi-stellar objects (QSOs) are the members of the AGN ALHAMBRA filters of AB ∼ 24.5−25, we expect to sample
family that have particularly high intrinsic luminosities allow- the whole QSO LF up to z ∼ 4.2 and up to z ∼ 6 for sources
ingthemtobedetectedatlargedistancesandtoprovideunique with M > −24.2.Atthe currentstage,thesurveyhasmapped
B
inside into the early history of the AGN-host galaxy interac- ∼2.5deg2 oftheskyinsevendifferentregions.Theresultspre-
tion.Moreover,QSOsarepotentialcontributorstotheultraviolet sented here, and the comparison with existing data from other
(UV)ionizingbackground(Cowieetal.2009)andhaveproba- cosmologicalsurveys,willprovethecapabilitiesofthesurveyto
bly played a non-negligiblerole in the reionization of the uni- deriveaccuratephotometricredshiftsfortheBLAGN/QSOpop-
verse(Fanetal.2006;Wangetal.2010). ulation. Furthermore, this test will potentially identify redshift
The optical selection of QSOs has been performed mainly ranges for which QSO photo-z estimation maybe unreliable or
with follow-up spectroscopic observations of color–color se- QSOswithatypicalSEDsthatwouldthenbesuitableformore
lected candidates (e.g. SDSS, Richards et al. 2002; and 2dF, detailedstudy.
Croom et al. 2004). These observations use slitless or prism This paper is structured as follows. In Sect.2, we describe
spectroscopic surveys and poorly efficient flux–limited spec- the current photometric catalog from the ALHAMBRA sur-
troscopic surveys (e.g. VIMOS–VLT Deep Survey, Gavignaud vey as well as the ancillary data available in each of the
etal. 2006;Bongiornoet al. 2007).A noveltechniquewith re- ALHAMBRAfieldsfromothercosmologicalsurveys.Thissec-
specttopreviousselectioncriteriawasintroducedbytheCADIS tion also introduces the QSO sample selection. The method-
(Meisenheimeretal.1998)andCOMBO–17surveys(Wolfetal. ology followed during the photo-z determination is discussed
2003). These photometric surveys used several optical broad– in Sect.3 while in Sect.4 we quantify the precision of our
and medium-band filters to characterize the nature of the de- photo-z estimates by comparing them with previous results
tected population and derive its photometric redshift (photo-z) for this type of sources. Finally, Sect.5 discusses the im-
viathespectralenergydistributions(SEDs).Thefluxesreached plications of our results and planned future analysis. A de-
by the surveyhave allowed the study of the high-zQSO popu- tailed QSO catalog will be presented in a forthcoming pa-
lation,thusovercomingtheproblemofQSOincompletenessin per. Throughoutour analysis, we assume a ΛCDM cosmology
the redshiftinterval2.2 ≤ z ≤ 3.6 (Richardset al. 2002).This with H0 =70kms−1Mpc−3, ΩΛ =0.7, and ΩM = 0.3. Unless
redshift range is important because it corresponds to the peak otherwisespecified,allmagnitudesaregivenintheABsystem.
andthe turnoverofthe observedQSO spacedensity (e.g.Wolf
etal.2004).
Over the past decade, a clearer understanding of the QSO 2. Dataset
evolution has been achieved thanks to a more accurate charac-
2.1.Photometricdata:theALHAMBRAsurvey
terizationoftheirdifferentSEDs,toamoreprecisetreatmentof
theirvariability,andtoasignificantimprovementintheirphoto-z The ALHAMBRA1 (Advanced, Large, Homogeneous Area,
determination.Thisadvancehasbeenencouragedbytheconcep- Medium-Band Redshift Astronomical) survey provides a
tionofmoderncosmologicalsurveysandnewlyavailablespace- photometric dataset over 20 contiguous, equal-width, non-
based observing facilities (e.g. HST, XMM-Newton, Chandra, overlapping, medium-band optical filters (3500−9700 Å) plus
Spitzer,andHerschelamongothers)thathavebeenabletodetect 3 standard broad-band NIR filters J, H, and Ks over 8 dif-
alargeamountofsources(≥106).Inparticular,whenagivensci- ferent regions of the northern hemisphere (Moles et al. 2008).
entificgoaldoesnotrequiredetailedknowledgeof thespectral The survey aims to understand the evolution of the structures
propertiesofindividualobjects,properlydesignedphotometric and the different families of extragalactic sources throughout
surveyscanprovideahighlyreliablephoto-zandspectralclassi- cosmic time by samplinga largeenoughcosmologicalfraction
ficationforeachsource.Thesephoto-z’sareanessentialcomple- of the universe. This requires precise photometric redshifts for
menttotheusuallysmallfractionofsourceswithspectroscopic
redshifts in major extragalactic surveys and to more reliably 1 http://alhambra.iaa.es:8080
A20,page2of17
I.Matuteetal.:Quasi-stellarobjectsintheALHAMBRAsurvey.I.
Table1.ALHAMBRAfields.
Field Alpha(J2000) Delta(J2000) Area(deg2) Obs.period Surveys E(B−V) Spectro-QSO Source
(i) (ii) (iii) (iv) (v) (vi) (vii) (viii) (ix)
ALH-2 022832.0 +004700 0.50 Sep.05–Nov.09 DEEP2 0.030 30/29 1,2
ALH-3 091620.0 +460220 0.25 Dec.04–May09 SDSS 0.015 2/2 1
ALH-4 100028.6 +021221 0.25 Dec.04–May09 COSMOS 0.018 81/77 3
ALH-5 123500.0 +615700 0.25 May05–Jun.09 HDF-N 0.011 18/15 1,4,5
ALH-6 141638.0 +522505 0.19 Aug.04–Aug.09 EGS-AEGIS 0.011 35/33 1,2,6
ALH-7 161210.0 +543000 0.50 Aug.04–Jul.09 SWIRE/ELAIS-N1 0.007 11/11 1,4,7
ALH-8 234550.0 +153450 0.50 Aug.04–Aug.09 SDSS 0.024 3/3 1
TOTALQSOs(a) 2.44 180/170
Notes. (i) = ALHAMBRA field name; (ii, iii) = central coordinates of the field; (iv) = area covered by each field; (v) = period between the
beginning and the end of the observations in a given field; (vi) = name of the cosmological survey for which a particular ALHAMBRA field
overlaps;(vii)=meanGalacticreddeningalongthelineofsightderivedbySchlegeletal.(1998)fromtheIRAS 100μmdata;(viii)=totalnumber
ofspectroscopicallyidentifiedQSOsinthefieldandthosewithinourselectioncriteria;(ix)=sourceoftheQSOclassificationandspectroscopic
redshift:(1)=Schneideretal.(2010);(2)=DEEPwebpage,Davisetal.(2003);(3)=Brusaetal.(2010);(4)=Veron-Cetty&Veron(2010);
(5) = Barger et al. (2008); (6) =Willmer (priv. comm.); (7) = Rowan-Robinson et al. (2008); (a) Sum over all theALHAMBRA fieldsof the
total/selectedspectroscopicQSOs.
several hundreds of thousands objects and, therefore, a survey
withhighphotometricaccuracyaswellasdeepandwidespec-
tralcoverageoveralargearea.ThesimulationsofBenítezetal.
(2009), relating the image depth and photo-z accuracy to the
number of filters, have demonstrated that the filter–set chosen
for ALHAMBRA can achieve a photo-z precision, for normal
andstar-forminggalaxies,thatisthreetimesbetterthana clas-
sical 4–5 optical broad-bandfilter set. The final survey param-
eters and scientific goals, as well as the technicalpropertiesof
the filter set were described by Moles et al. (2008). The sur-
vey has collected its data for the 20+3 optical-NIR filters in
the 3.5m telescope, at the Calar Alto observatory2, using the
wide-fieldcameraLAICAintheopticalandtheOMEGA–2000
camera in the NIR. The full characterization, description, and
performance of the ALHAMBRA optical photometric system
was presented by Aparicio-Villegas et al. (2010). The strategy
of ALHAMBRA for each run has been to observe the fields
with the lowest airmasses trying to complete the requested in-
tegrationtimeforeachfilterinordertoreachtheplanneddepth Fig.1. Wavelength coverage of the ALHAMBRA filter set for the
(ABmag∼25).Inconsequence,anddependingalsoonthetele- LAICAcameraCCD1(thincontinuouslines).Thepositionofthemost
scope/instrument downtime and weather, the time to complete prominentQSOandBLAGNemissionlines,plustheLymanlimit,are
eachfilterineachfieldcanvaryfrommonthstoseveralyears3. shownevolvingwithredshiftasthicklines.Allfiltershavebeennormal-
Therefore,althoughALHAMBRAphotometryallowsustode- izedtounity(seeAparicio-Villegasetal.2010,fortheirtrueefficiency).
tect variability, the lack of a common band(s) taken during all Dottedlinestraceeachfiltercentralwavelength(Table2).
theobservingrunsdoesnotallowustocorrectforitseffect.The
deep NIR counts in one of the ALHAMBRA fields (ALH-8),
over ∼0.5 deg2, which has a 50% detection efficiency depth of
J ∼ 22.4,H ∼ 21.3,andK ∼ 20.0(Vega),havebeenanalyzed
s
byCristóbal-Hornillosetal.(2009).Theirresultshelpedtocon-
The photometricdata pointsused in thiswork are givenby
straindifferenttype-dependentgalaxyevolutionarymodels.
theMAG_AUTOmeasurebySExtractor(Bertin&Arnouts1996).
In this work, we used the seven ALHAMBRA fields for Toavoidtheexcessiveweightofsomepointsinthecomputation
whichdatahavebeencurrentlyobservedandreduced.Thecen- ofaphoto-z,weadoptedaminimumphotometricerrorofδm =
tral coordinatesof these 7 fields, the area coveredby each, the 0.05 (i.e. if a photometric error is smaller than 0.05, it is set
observingepoch,theircoincidencewithothercosmologicalsur- to 0.05)forthe methoddescribedin Sect.3. In agreementwith
veys,andtheirmeanGalacticextinctionaredetailedinTable1. the results of other authors (e.g. Bolzonella et al. 2000), we
Thetotalnumberof spectroscopicallyidentifiedQSOs, aswell found that there is no gain, or even that we obtain poorer re-
as the fraction selected for our analysis, are given in Col. 7 of sultsforsomeobjects,whenweconsiderδm<0.05.Thephoto-
Table1.Figure1andTable2detailthegeneralcharacteristicsof metricdatapointsofeachobjectwerecorrectedforinterstellar
theALHAMBRA23filterset. extinctionusingthevaluesofE(B−V)providedbythemapsof
Schlegeletal.(1998),whicharebasedonIRAS100μmdata4.
2 http://www.caha.es
3 Furtherdetailsoftheobservationswillbeprovidedinacompanion
paper. 4 http://irsa.ipac.caltech.edu/applications/DUST/
A20,page3of17
A&A542,A20(2012)
Table2.ALHAMBRAfiltercharacteristics. surveys shows that source detections with percent-weight
≥0.70 are highly reliable and that a negligible fraction of
Name λ FWHM AB Offset (cid:6)m(cid:7) (cid:6)σ(m)(cid:7) themarespurious.
mean
(μm) (μm) corr. (AB) – Thesourcemustbewithintheveryhighconfidencemagni-
(1) (2) (3) (4) (5) (6) (7) tude intervalof the survey.The chosen magnitudeof refer-
A366M 0.3661 0.0279 0.96 –0.033 21.81 0.08
enceis A678M filter centeredon6789Å andthe intervalis
A394M 0.3941 0.0330 0.02 –0.210 21.60 0.05
definedby17.0≤ A678M ≤23.5.Thebrightmagnitudecut
A425M 0.4249 0.0342 –0.13 –0.081 21.65 0.05
ensuresthatnosourcesaturatesanyfilter,whilethefaintcut
A457M 0.4575 0.0332 –0.18 –0.011 21.62 0.07
A491M 0.4913 0.0356 –0.05 –0.065 21.54 0.06 avoidssourceswithphotometricerrorslargerthan(cid:9)0.2mag.
A522M 0.5224 0.0326 –0.04 –0.054 21.48 0.06
A551M 0.5510 0.0297 0.01 0.003 21.48 0.07 We decided against the inclusion of a stellarity criteria, as
A581M 0.5809 0.0324 0.07 –0.001 21.39 0.05 the precision of the one derived by the SExtractor (Bertin &
A613M 0.6134 0.0320 0.13 0.009 21.35 0.06 Arnouts 1996) package was valid only for the brighter part of
A646M 0.6461 0.0357 0.23 0.006 21.33 0.08 theQSOsample(m ≤22).
A678M 0.6781 0.0314 0.24 –0.046 21.20 0.06 678
A708M 0.7078 0.0332 0.29 –0.055 21.14 0.05
A739M 0.7392 0.0304 0.34 0.007 21.16 0.06 2.3.Spectroscopicdata
A770M 0.7699 0.0354 0.39 0.000 21.11 0.06
A802M 0.8020 0.0312 0.44 0.002 21.06 0.07 To assess the quality and accuracy of the photo-z determi-
A829M 0.8294 0.0296 0.48 0.007 21.00 0.08 nation for the ALHAMBRA database, we compiled all the
A861M 0.8614 0.0369 0.54 –0.023 20.91 0.05 published or publicly available spectroscopic information for
A892M 0.8918 0.0303 0.50 0.022 20.93 0.08 BLAGNs/QSOs. The online services and public spectroscopic
A921M 0.9208 0.0308 0.48 0.028 20.83 0.10 catalogs included the Sloan Digital Sky Survey (SDSS5) DR7
A948M 0.9482 0.0319 0.52 0.077 20.65 0.15
(Schneider et al. 2010), the Deep Extragalactic Evolutionary
J 1.2094 0.2471 0.87 0.104 20.65 0.06
Probe (DEEP/DEEP26; Davis et al. 2003), the All-wavelength
H 1.6482 0.2665 1.38 0.186 20.41 0.08
Ks 2.1409 0.3040 1.83 0.155 20.20 0.09 ExtendedGrothstripInternationalSurvey(AEGIS7;Davisetal.
2007),theCOSMOS8 XMMsourcecatalog(Brusaetal.2010),
Notes. Columns: (1) filter name; (2) filter mean wavelength; (3) fil- theGOODS-NorthredshiftcompilationbyBargeretal.(2008),
ter FWHM; (4) AB-Vega magnitude correction: mAB = mVega + the SWIRE9 spectroscopic catalog by Rowan-Robinson et al.
AB_correction;(5)offsetsappliedtoeachfilterasmfinal =mfilter+offset (2008) and the 13th edition of the Veron-Cetty& Veron QSOs
duringthephotometricredshiftdetermination(seeSect.3.3fordetails); catalog (2010; VERONCAT10 hereafter).We verified the qual-
(6) mean magnitude in each filter band for the spectroscopic sample
ityoftheidentificationsaccordingtothefollowingcriteria:i)all
(Sect.2.3);(7)meanmagnitudeerrorsineachfilterbandforthespec-
spectrafromthe DR7 SDSS catalogwere visuallyinspectedto
troscopicsample.
determinewhethertheycontainedbroad-lineemission;ii)spec-
tra with a high-quality classification flag (flag ≥ 3) were se-
2.2.Sampleselection lected from the DEEP/DEEP2 and the AEGIS database and
visually inspected to confirm that they displayed broad-line
Broad-line AGNs (BLAGN) and QSOs are powerful emitters
emission; iii) as neither a spectral classification nor a redshift
over the entire electromagnetic spectrum. They show signifi-
quality were given by Barger et al. in the GOODS-N field,
cantspectralfeaturesintheformofintenseemissionlines(with validBLAGN/QSOcandidateswereselectedbasedontheirhard
EWrangingfromseveraltenstoseveralthousandsofÅ)inthe X-rayluminosity(L [2–8kev])beingbrighterthat1043ergs−1;
X
rest-frameUV,optical,andNIRregime.Thesepropertiesmake iv) in the COSMOS field, we selected the high-quality public
QSOs easily detectable out to very high redshifts (z ∼ 6) and spectraofBLAGN(flags1113,14,18,213,214,218;Lillyetal.
perfectcandidatestohelpimproveourunderstandingoftheac- 2007), while we considered as bona fide BLAGN/QSO the re-
cretionmechanismswithinSMBHs(MBH >106M(cid:8)).Theyalso mainingofXMMsourceswithoutpublicdatareportedbyBrusa
probethe distributionof large-scalestructuresandthe physical etal.(2010)as“bl” basedonMMTandIMACSspectroscopy;
conditions of the intergalactic medium (IGM). The strong fea- v)allSWIREsourceshavehighqualityspectra(Pérez-Fournón,
turesthatcharacterizetheQSOopticalemissionspectrumallow priv.comm.);andvi)allthesourcesfromtheVERONCATwere
us to test the ALHAMBRA photometry and its ability to pro- consideredasbonefideBLAGN/QSOs.Nolowerredshiftorab-
ducevery low resolutionspectra. Thiswould providea correct solutemagnitudecutoffwasincludedinthesourceselectionas
spectralclassificationandahigh-precisionredshiftestimatefor our goal is to test the efficiency of our method and photome-
theexpectedpopulationofseveralthousandsofQSOs. try as good redshift estimators not only for the most powerful
WeselectedourinitialQSOcandidatesfromasubsampleof BLAGNs and QSOs but also for the low redshiftSeyfert1 nu-
thecurrentALHAMBRAcatalog(v3),whichwascreatedusing clei, which may provide an important contribution to the total
thefollowingphotometriccriteria: lightof their host galaxies.In all cases, the match between the
– A surveyqualityflag ≥0.7.EachALHAMBRA sourcewas 5 http://www.sdss.org/
flagged with a parameter (“percent-weight” in the catalog) 6 http://deep.ucolick.org
that takes into account the total exposure time of a given 7 http://aegis.ucolick.org/
sourcerelativetothemaximumforagivenfield.Alowvalue 8 http://cosmos.astro.caltech.edu/
ofthisflag(<0.70)indicatesthatthesourceiseitherwithin 9 http://swire.ipac.caltech.edu/swire/swire.html
aregionstronglyaffectedbytheditheringprocessduringthe 10 http://heasarc.gsfc.nasa.gov/W3Browse/all/veroncat.
observation,containsbad pixels/artifacts,or is locatednear html
a bright (masked) source. A detailed comparison with the 11 Flags18and218refertoaspectroscopicredshiftcomputedwitha
deeperphotometrydataavailableforsomefieldsfromother singleline.
A20,page4of17
I.Matuteetal.:Quasi-stellarobjectsintheALHAMBRAsurvey.I.
ALHAMBRA photometry and the spectroscopic catalogs was
performedusingaonearcsecsearchradiusandalwaysidentified
auniquecounterpart.ThesourcesoftheBLAGN/QSOspectro-
scopicredshiftsforeachoftheALHAMBRAfieldsare:
– ALH-2: This ∼0.5 deg2 field partially overlaps the deep
strip of SDSS whose DR7 version provides redshifts for
23QSOs.Thecommonregionofthisareawithfield-4ofthe
DEEP/DEEP2surveyyields7additionalsourcesfromtheir
dataproductrelease3(DR312)oftheDEEP2spectroscopic
catalog.Ofthe30spectroscopicredshiftsforBLAGN/QSO
availableinthefield,29(6fromDEEP2and23fromSDSS)
complywithourphotometriccriteria.
– ALH-3:This consists of a ∼0.25deg2 field area which par-
tially overlaps with that of the SDSS. The matching be-
tween ALHAMBRA and SDSS DR7 spectroscopy yields
2 QSO redshifts, for which both sources verify our photo-
metriccriteria.
– ALH-4:Thisisa∼0.25deg2fieldincludedintheCOSMOS Fig.2. Magnitude-redshift distribution of the selected spectroscopic
survey area. The common area contains a total of sample.
81 BLAGN/QSO redshifts of which 77 (7 from SDSS and
70fromCOSMOS)complywithourphotometriccriteria. 1.56±0.88(DEEP2),1.56±0.63(AEGIS),1.42±0.69(SWIRE),
– ALH-5:Thisfieldcoversa∼0.25deg2areaoverlappingthat and1.24±0.78(VeronCat).
of the GOODS-N. There are a total of 18 BLAGN/QSOs
with spectroscopic redshifts (9 from SDSS, 6 from Barger
3. QSOphoto-zdetermination
et al. 2008, and 3 from the VERONCAT). Of these,
15 (7 from SDSS, 6 from GOODS-N, and 2 from the We used the publiclyavailable templatefitting codeLePhare13
VERONCAT)complywithourphotometriccriteria. (Arnoutsetal.1999;Ilbertetal.2006)toestimateredshiftsfor
– ALH-6: This ∼0.19deg2 field is centered on the GROTH our selected QSOs. The code matches the photometric data of
strip and therefore overlaps with the DEEP2, AEGIS, and each ALHAMBRA QSO source to a library of available tem-
SDSSsurveys.Intotal,thereare35spectroscopicallyiden- plates providing the best-fit, spectral classification, and photo-
tified BLAGN/QSO in this field, of which 33 sources metricredshiftbymeansofaχ2minimizationprocess.Themin-
(6 SDSS, 6 DEEP2 and 21 AEGIS) comply with our pho- imizationprocessaccepts the inclusionof user-suppliedpriors,
tometriccriteria. different extinction laws, and the possibility to apply system-
– ALH-7: A ∼0.5deg−2 field centered on the ELAIS-N1 of aticoffsetstothedifferentphotometricbandsinordertoachieve
the SWIRE survey. Sources with spectroscopic redshifts thebestmatchbetweenthecolorsofthesampleandthosepro-
and classifications are provided by the catalog of Rowan- videdbythetemplatelibrary.Thefullcapabilitiesandpossibili-
Robinson et al. (2008), the SDSS spectroscopy and the tiesoftheLePhareminimizationcodewasextensivelydiscussed
VERONCAT. All 11 BLAGN/QSO found in this field by Ilbert et al. (2006, 2009). Our final selection of templates,
(1 from SDSS, 8 from SWIRE, and 2 from VERONCAT) adopted reddening laws, priors, and systematic offsets are dis-
complywithourphotometriccriteria. cussedinthefollowingsections.
– ALHAMBRA-8: The 3 QSO spectroscopic identifications
in this ∼0.5deg2 field are provided exclusively by the
QSO DR7 catalogofthe SDSS. All3 sourcescomplywith 3.1.Templateselection
ourphotometriccriteria.
Thelistofextragalactictemplatesusedinthisworkaredetailed
in Table3 and Fig.3. The selection includes SEDs for QSOs,
The final spectroscopic catalog of the ALHAMBRA fields
includes 94% (170/180) of the total numbers of sources Seyferts, starburst, normal galaxies, and stars. To include low
spectroscopically identified in the different fields. The ALH4- luminosity BLAGN that are partially or completely dominated
COSMOS field contains∼44%of the sources, followedby the bytheirhostgalaxylight,weadoptedthehybridtemplates(con-
ALH2andALH6fields(DEEP2/AEGIS)whichcorrespondsto sisting of a mixtureof QSO andhostgalaxySEDs) introduced
∼17%and19%oftheIDs,respectively.Table1detailsthenum- bySalvatoetal.(2009).Thevarietyoftemplatesisjustifiedby
theneedtotesttheabilityoftheALHAMBRAsurveytodiffer-
ber of identified sources in each field, while Fig.2 shows the
entiatebroad-lineAGNemissionfromthatofotherextragalac-
redshiftdistributionoftheselectedspectroscopicsample.
ticsourcesorstars,andenableustodoablindsearchforthese
Althoughthe surveysfromwhichthe spectroscopicsample
sources(Matuteetal.,inprep.).
isextractedencompassawiderangeofselectioncriteria(color
InTable3andFig.3,thetemplatesareorganizedas:
selectionintheSDSS,opticalflux-limitedinzCOSMOS_bright,
X-rayselectedforIMACSandMMTspectroscopy,etc.),wefind – Non-active and starburst galaxies:this includes3 elliptical
theirredshiftdistributioncompatiblewithintheerrors.Themean templatesofdifferentages(2,5,and13Gyr;#1–3),thestar-
redshift and 1σ dispersion for the sources extracted from the burst galaxiesArp220, M82, NGC6240,IRAS20551,and
differentspectroscopiccatalogsare:1.61±0.63(SDSS),1.99± IRAS2249114 (#4–8)and7spirals(S0throughSd;#9–15).
0.68(zCOSMOS_faint),1.92±0.64(zCOSMOS_bright),1.24±
0.45 (MMT), 1.55±0.66 (IMACS), 1.81±0.81 (GOODS-N), 13 http://www.oamp.fr/people/arnouts/LE_PHARE.html
14 MaycontainanAGNresponsiblefor20%ofthebolometricflux(e.g.
12 http://deep.berkeley.edu/DR3/dr3.primer.html Veilleuxetal.2009).
A20,page5of17
A&A542,A20(2012)
Table3.Extragalactictemplatelibrary.
Index SEDname Class
1 Ell2 Elliptical(5Gyrold)a
2 Ell5 Elliptical(2Gyrold)a
3 Ell13 Elliptical(13Gyrold)a
4 Arp220 Starbursta
5 M82 ”a
6 IRAS20551–4250 ”a
7 IRAS22491–1808 ”a
8 NGC6240 ”a
9 S0 S0a
10 Sa Saa
11 Sb Sba
12 Sc Sca
13 Sdm Sdma
14 Sd Sda
15 Spi4 Spirala
16 Sey18 Seyfert1.8a
17 Sey2 Seyfert2a
18 IRAS19254-7245 Seyfert2a
19 QSO2 QSO2a
20 hyb1_gal10_agn90 Hybrid10%S0+90%QS02b
21 hyb1_gal20_agn80 ...
22 hyb1_gal30_agn70 ...
23 hyb1_gal40_agn60 ...
24 hyb1_gal50_agn50 ...
25 hyb1_gal60_agn40 ...
26 hyb1_gal70_agn30 ...
27 hyb1_gal80_agn20 ...
28 hyb1_gal90_agn10 Hybrid90%S0+10%QS02
29 hyb2_gal10_agn90 Hybryd10%I22491+90%TQSO1b
30 hyb2_gal20_agn80 ...
31 hyb2_gal30_agn70 ... Fig.3.Thecompleteextragalactic(galaxy+AGN)templatedatabase
32 hyb2_gal40_agn60 ... usedinthiswork.Thesourceandspectralclassforeachtemplategiven
33 hyb2_gal50_agn50 ... inTable3anddescribedinSect.3.1.
34 hyb2_gal60_agn40 ...
35 hyb2_gal70_agn30 ...
36 hyb2_gal80_agn20 ... obscuredsourcesare representedbya type-2QSO (QSO2)
37 hyb2_gal90_agn10 Hybryd90%I22491+10%TQSO1 and the BALQSO Mrk231 templates (Polletta et al. 2007;
38 QSOL QSO(low-luminosity)b #19 & #50). We also added the hybrid template library of
39 QSOH QSO(high-luminosity)b Salvatoetal. (2009)definedby9 differentcombinationsof
40 TQSO1 QSOcompositea aS0andaQSO2template(#20–28).
41 QSO1 ”a – QSO and hybrid-QSO templates: here we considered both
42 syth_qso–0.25 QSOsyntheticc thehighandlowluminositySDSScomposites(indices#38,
43 syth_qso–0.50 ” #39), two templates from Polletta et al. (2007; QSO1 and
44 syth_qso–0.75 “ TQSO1 with indices #40 and #41), the Cristiani & Vio
45 syth_qso–1.00 ” QSO SED (#47), the VVDS mean QSO SED (Gavignaud
46 syth_qso–1.25 “ et al. 2006; #48), and the mean QSO from Vanden Berk
47 QSO_Cristiani QSOd et al. (2001; #49) also based on SDSS data. Hybrid
48 QSO_VVDS QSOcompositee templates include 9 different combinations of the star-
49 QSO_vandenBerk QSOcomposite f burst/ULIRG IRAS22491 template and a QSO1 template
50 Mrk231 BALQSOa (Salvatoetal.2009; #29–#37).As the quality and accuracy
of the fit improved in several cases, we completed the list
References.(a)Pollettaetal.(2007).(b)Salvatoetal.(2009).(c)LePhare
withasetof5syntheticQSOtemplates(#42–46),covering
template database. (d) Cristiani & Vio (1990). (e) Gavignaud et al. continuum slopes (να) from α = −0.25 to α = −1.25 (in
(2006).(f)VandenBerketal.(2001).
stepsof0.25)below1μmandfixedatα=−0.7above1μm
(LePharetemplatedatabaseandreferencestherein).
Finally,the stellartemplatedatabaseincludes131spectrafrom
They are all part of the SED library published by Polletta thePickels(1998)stellarlibraryplus4spectraofwhitedwarfs
etal.(2007). fromBohlin et al. (1995) and 19 additionaltemplatesfromthe
– Obscured BLAGN: includes the Polletta et al. (2007) com- LePharestellarlibrary.Stellartemplateswerealsoincludedbe-
posite templates of a Seyfert1.8 and a Seyfert2 and cause white dwarfs and F/G stars QSOs have similar colors
the Seyfert-2 IRAS19254 (#16–18). The high-luminosity to F/G stars in the z = [2–3] redshift interval. Thus, our final
A20,page6of17
I.Matuteetal.:Quasi-stellarobjectsintheALHAMBRAsurvey.I.
database contains 204 templates (50 extragalactic + 154 stel- “adapt” the templates to provide a better fit the observed pho-
lar).Duringtheminimizationprocess,Galacticandextragalactic tometry.Therefore,alternativetemplatesorobjectselectioncan
templateswereusedseparately. (and will) lead to different offsets (e.g. Table 1 by Ilbert et al.
2006).
We used the spectroscopic sample described in Sect.2 and
3.2.Extinction
the colors in the filters A457M, A646M, and A829M to com-
putethesesystematicoffsetsusinganiterativeapproachoffind
Thephotometryforseveralofthesourcesanalyzedinthiswork
the best-fit SED for each source, deriving the mean deviations
defines a continuum that strongly deviates from single or even
foreachfilter,applyingoffsets,re-computingbest-fitSEDs,etc.
doublepower-laws,mostprobablybecauseofdustobscuration.
The iteration process was halted when the variation in χ2 be-
For manyof these typesof sources, thereis an ongoingdebate
tweeniterationsdropsbelow2%.In general,theproceduredid
about whether (some of) these red QSOs are obscured by ei-
notrequiremorethan4iterationstoconverge.Wefoundthatthe
therdustoranintrinsicallyredcontinuum(Richardsetal.2003;
offsets to be applied are small and agree with the typical pho-
Younget al. 2008). Thus, our multibandtemplate fit takes into
tometric error for each band in the sample. Table 2 reports the
account the possibility of intrinsic dust obscuration within the
valuesofthesecorrectionsaswellasthetypicalphotometricer-
source.WeadoptedtheSmallMagellanicCloud(SMC)extinc-
rorforeachbandinCols.5and7,respectively.
tionlaw(Prevotetal.1994),whichhasbeenshowntoreproduce
theobservedreddeningformildlyobscuredQSOsatz<2.2,for
whichtherearenoindicationsoftheGalacticfeatureat2175Å 3.4.Priors
(Hopkins et al. 2004; Richards et al. 2003; York et al. 2006).
Galleranietal.(2010)appearedtomeasuresomedeviationfrom Theintroductionofimportantaprioriinformationintothered-
theSMCextinctionlawforhigherredshiftsourceswhichisone shift probability distribution function (Pdz) based on Bayesian
reasonforadoptingalternativeextinctionlaws(seebelow).The probability can in many cases improve the quality of the solu-
attenuationdue to dust (A ) is givenas a functionof the color tionsbyfavoringaparticularredshiftbasedonknownredshifts
excessE(B−V)asA =RV × E(B−V).WeassumedR =3.1 and/orcolordistributions(e.g.Benítez2000).Ouranalysisonly
V V V
andacolorexcessintherange[0,1]. makesuseofaparticularluminosityandredshiftrangepriorand
doesnotincludeanyredshiftdistributionorcolorinformationof
Furthermore, as our spectroscopic sample includes lower
known BLAGN/QSO populations. We restricted the permitted
redshiftSeyfert1 nuclei, which may have a stronghost-galaxy
absolutemagnitudesintheA457M(λ4575Å)filterbetween−17
contribution, we considered alternative extinction laws as the
dust present in different galaxy types follow extinction curves and −28. Absolute magnitudesin this filter are consistent with
thecommonlyusedbroad-bandstandardfilter B.Thisincludes
thatdeviatefromthatoftheSMC.Thesedeviationsincludevari-
notonlythetypicalrangewheremostBLAGN/QSOsarefound
ationsofthesteepnessintheattenuationcurveasinthestarburst
extinctionlawderivedbyCalzettietal.(2000),orthepresenceof (−28 ≤ M457 ≤ −20) but also the range for host-dominated
BLAGNandnormalgalaxies(Salvatoetal.2009;Pollettaetal.
abroadbumparound2175Å asfoundfortheMilkyWay(MW;
2007;Rowan-Robinsonetal.2008).
Seatonetal.1979;Cardellietal.1989)ortheLargeMagellanic
Cloud(LMC;Fitzpatrick1986).Therefore,toreproducenormal
galaxyandstarburstspectrawe alsoconsideredintheχ2 mini- 3.5.Photo-zdeterminationsummary
mizationsolutionsbasedontheLMC, MW, Calzetti’slaw,and
Calzetti’s law plus the absorption feature around 2175Å. In WeusedthecodeLePharetoestimatephotometricredshiftsfor
thiscase,theminimizationprocesstakesintoaccountallpossi- 170 spectroscopically identified BLAGN and QSOs with high
bleSEDsandextinctionlawssimultaneously,choosingthebest qualityALHAMBRAphotometry.Fortheχ2minimizationpro-
suitedtoeachsource.Theseadditionalextinctionlawsallowus cess,weconsideredthefollowing:
toprobetheirrelevancetotheaccuracyoftheresults.
– Adatabaseof204templates:154stellarand50extragalactic.
The light attenuation by the inter-galactic medium (IGM)
– Several extinction laws: MW, LMC, SMC, and Calzetti’s
was taken into account internally by LePhare following the
starburst laws with a color-excess range of E(B − V) =
opacitycurves,binnedinto redshiftintervalsof Δz = 0.1,pub-
[0.0,1.0].
lishedbyMadau(1995).
– Givenourtemplatelibrary,wemadeacorrectiontothezero-
point of the filters that show a non-zero average deviation
3.3.Systematicoffsets betweentheobservedandbest-fitpredictedmagnitudes.
– Asimpleluminositypriorof−17 ≤ M ≤−28.
A457M
Photometric redshifts depend strongly on the precision of the – A redshift space interval of 0 ≤ z ≤ 6 binned in redshifts
photometryand the capabilities of the template database to re- intervalsofδz=0.04.
produce the colors of the source population as a function of z.
If we were to assume that the selected template database is Figure4showsanexampleoftheexcellentagreementbetween
representative of our source population, then for a given filter thedataandthefittedtemplatefor6sourceswithawiderange
the average deviation between the observed flux and the best- ofmagnitudes(20 ≤ mA678M ≤ 23.5),redshifts(0.7 ≤ z ≤ 2.3),
fitpredictedfluxshouldbezerofornormallydistributeduncer- andintrinsicextinctions(0.0≤E[B−V]≤0.2).
tainties. If this is not the case, a zero-point offset must be ap-
pliedtothephotometryderivedfromthetemplatedatabasewhen
thereisanon-zeroaveragedeviationbetweentheobservedand 4. Resultsanddiscussion
predicted fluxes. The ALHAMBRA photometric calibration is
4.1.Photo-zaccuracy
basedonaselectionofNGSLstarsfollowingthemethodology
discussed in Aparicio-Villegaset al. (2010). We note here that The efficiency of the photo-z determination is quantified by
wehavenotmodifiedthiscriteria.Thecomputedoffsetsinstead, comparing the spectroscopic redshifts (hereafter spectro-z) of
A20,page7of17
A&A542,A20(2012)
Fig.4.Examplesofbest-fitsolutionsassumingaSMCextinctionlawfor6sourcescoveringawiderangeofmagnitudes(∼20≤ m ≤∼ 23.5)
678
and spectroscopic redshifts (0.7 ≤ z ≤ 2.3). Each panel includes the observed photometry, associated errors, and FWHM for each of the 23
ALHAMBRAfilterset(blackdots,verticalandhorizontalerrorbarrespectively).Photometricupperlimitsareindicatedbyarrows.Thecontinuous
lineshowsthebest-fitsolution,whiletheopencirclesgivetheexpectedmagnitudefromthemodelcorrectedfromsystematicoffsets.Additional
infofor each source includes: model name, reduced χ2, amount of extinction, thenormalized probability distributionasa function ofz(Pdz),
thespectro-z(anditssourcecatalog),andthebest-fitphoto-zsolution.ThetitleofeachpanelislabeledwiththesourceIDintheALHAMBRA
catalogandthemeasuredmagnitudeinthem filter(greendot).
678
170sourcesinourBLAGN/QSOsample.Photometricredshifts whiletheoutlierfraction(η)isdefinedasthefractionofsources
are generally characterized by both their accuracy and outlier with catastrophic solutions (i.e. solutions that are inconsistent
fraction. The accuracy is defined as the standard deviation of with the measuredspectro-z).In ouranalysis, we assumed that
Δz/(1 + zspec), denoted σΔz/(1+zspec), where Δz = zspec − zphot, a source is an outlier if |Δz|/(1 + z) ≥ 0.15. This value was
A20,page8of17
I.Matuteetal.:Quasi-stellarobjectsintheALHAMBRAsurvey.I.
selected a priori to be compatible with the cutoff of similar
studies (e.g. Luo et al. 2010; Salvato et al. 2009; Ilbert et al.
2009;Rowan-Robinsonetal.2008).Analternativeaccuracyes-
timate that has been used by several authors (e.g. Ilbert et al.
2006; Brammer et al. 2008) is the normalizedmedian absolute
deviation(NMAD)definedas
(cid:2) (cid:2)
(cid:2) (cid:2)
σ =1.48×median(cid:2)(cid:2)(cid:2)Δz−median(Δz)(cid:2)(cid:2)(cid:2)·
NMAD (cid:2) 1+z (cid:2)
spec
Thisparametercanbedirectlycomparedtothestandarddevia-
tionofΔz/(1+z )inthecaseofnormaldistributionsandhas
spec
the advantageof being less sensitive to outliers. From now on,
weuseσ asourestimateofthephoto-zaccuracy.
NMAD
We nowdiscuss ourresultsbased on the numberof extinc-
tionlawsconsideredinthecomputation,namelyeitherasingle
(SMC) extinction law (SEL) or multiple (SMC, LMC, Milky-
Way,andCalzetti)extinctionlaws(MEL).Table4describesthe
solutions found for the two sets of extinction laws considered.
In the case of a SMC extinction law, we obtained an accuracy
of σ = 0.009 with a fraction of outliers of η ∼ 12% (21
NMAD
outof170sources).Acomparisonbetweenthederivedphoto-z
andthespectro-zisshowninthetoppanelofFig.5.Thenarrow
scatterpresentforthegoodfits(greendots)ishighlightedbythe
distributionofthesourceredshiftaccuracythatliesintherange
|Δz|/(1+z )≤0.15(i.e.nooutlierregion)andshowninFig.6.
spec
Thisdistributioniswell-representedbyaGaussianwithnomea-
surable bias (centered at −0.001) and a σ of ∼0.006. Identical
resultswerefoundwhenwe consideredseveralextinctionlaws
during the minimization process (σ = 0.009, η ∼ 12%)
NMAD
but, as we see in the following paragraphs, the MEL approach
isabletomoreaccuratelyreproducetheSEDdistributionofthe
BLAGN/QSOpopulation.
Besidestheabilitytoprovideprecisephoto-z’s,ouranalysis
allows us to recover the correct SED for most of the sources.
Figure7 presentsthe distributionof the templates for the MEL
best-fitsolutions.Wedidnotfindsourceswithstellartemplates
that have best-fit solutions (i.e. χ2 < χ2 ) and the ma-
stellar gal−QSO
jority of the sources are best-fitted by pure type-1 QSO tem-
plates or QSO hybrid templates (ULIRG/QSO1 template in-
dices #29 and above; Sect.3.1). When we did not take into
accounttheoutlierfractionofthesources,wefoundthat95.3%
of them (142/149) have either a QSO or hybrid-QSO best-fit
template, 1.3% (2/149) are fitted by a QSO2 or hybrid-QSO2
template,and3.4%(5/149)arefittedwithanormalorstarburst
template.Five sourceshave best-fit solutiontemplatescompat-
iblewith anon-activeSED.A closerlookattheALHAMBRA
Fig.5. Photo-z efficiency using several extinction laws (the MEL so-
photometry,thebestfitsolution,andtheobservedspectra(when
lution).Top:comparison betweenthebestfitphoto-zsolutionandthe
available)revealedthat:i)onesourcemightbeincorrectlyclas-
measuredspectro-zshowsagoodagreementbetweentheboth.Thecon-
sified as BLAGN since both the spectra and the ALHAMBRA tinuous line gives the z = z relation while the dashed line rep-
photometry point to an early-type galaxy; ii) one source be- phot spec
resent the boundary between good solutions (green dots) and outliers
laonndgsthetoasthsoecifaatiendteerrrpoarrst(oΔfmth>e p0o.2p)ulhaatvioend(ilmu6te78d a=ny2p3o.s3s3i-) t(rriebdu,tiinodneoxfeΔdzd/o(1ts)+azndd)eafisnaedfuansct|Δiozn|/o(1ft+hezsApeLc)H>AM0.1B5R.ACemntarganl:itduidse-
spec
bleBLAGNsignatureintheALHAMBRAphotometry;iii)the m .Themeanmagnitudeerror,permagnitudeintervalofΔm=1,of
678
other3sourceshaveopticalspectracompatiblewithgalaxytem- thefilterA678Misindicatedbythecontinuouslines.Theaccuracyper
plateswithdifferentdegreesofstarformingandpost-starforming magnitudeinterval(Δm=1)ishighlightedbyagreyshadedarea.This
activities,i.e.nosignsofbroademissionandacontinuumwitha accuracyshowsasmallcorrelationwithapparent magnitude. Bottom:
thispanel showsthecontribution ofeach magnitude bin(Δm = 1) to
well-defined4000Å break,butwithALHAMBRAphotometry
the outlier population as filled diamonds connected by a dashed line.
showingsomedegreeofAGNactivity(bluecontinuumandindi-
Themagnitudeintervalsconsideredandtheassociatederrorsareindi-
cationsoffaint,typicalBLAGNline-emissionnamelyofMgII, catedbythelargediamonds.Thefractionofoutlierswithrespecttothe
CIII] or CIV at the spectroscopic redshift of the source). The numberofsourcesinthesamemagnitudeintervalsaregivenbyaster-
incorrectsolutionsfoundforthese3sourcesareprobablyacon- isksconnected by a continuous line.The magnitude intervals and the
sequenceoftheuncertaintiesexpectedfromthemethod,partic- associatederrorsareindicatedbythelargeverticalandhorizontallines.
ularly regardingthe chosen template database and the absence Inbothcases,errorsareassumedtobePoissonianandwerecalculated
ofanyvariabilitycorrectionofthephotometry. followingGehrels(1986).
A20,page9of17
A&A542,A20(2012)
For this small fraction of sources (2%; 3/149), the incorrect
spectralclassification,usingthemethoddescribedhere,willbe
takenintoaccountinanystatisticalanalysisoftheBLAGN/QSO
populationdetected in the ALHAMBRA fields to be presented
inaforthcomingpaper(Matuteetal.,inprep.).
Furthermore, we note that the photo-z determination de-
scribedhereisabletorecovertheredshiftofthesourcesinthe
interval 2.2 ≤ z ≤ 3.6, which has been traditionally biased
against the selection of QSOs because of their similar colors
to F/G stars. Hence, the photometry and the method described
herecouldprovideanefficientwayofbothclassifingandderiv-
ingareliablephotometricredshiftsforBLAGN/QSOcandidates
pre-selected,for example,by their X-ray flux. The catalog and
derived luminosity functions for BLAGN/QSO selected purely
based on the ALHAMBRA photometry will be presented in a
companionpaper(Matuteetal.,inprep.).
The dependence of our results on redshift, apparent mag-
Fig.6.Uncertaintydistribution,Δz/(1+z),forthe170BLAGN/QSOsin nitude of the source, and the systematic offsets applied during
oursampleconsideringtheMELapproach.Non-outlierandoutliersare
thephoto-zcomputationaredescribedinthefollowingsections.
representedbyfilledandopenhistograms,respectively.Thecontinuous
Thesedependencesaresimilarinthetwosetsofextinctionlaws
linerepresentsthebestGaussianfittotheobserveddistributionofnon-
unlessotherwisespecified.
outliers.Thenumberofnon-outliers(#),thecenter,andσofthebest-fit
Gaussiandistributionareindicated.
4.1.1. Dependenceonredshift
Theaccuracyofthephoto-zresultsisratherindependentofthe
redshift with the exception of the interval z = [0.9, 1.4] (light-
grey square in Fig.5a for the SMC results). The presence of
onlyoneprominentline(MgII)inthisintervalintroducessome
aliasing that dependson the intensity of the line and how well
is sampledbythe ALHAMBRA filters. Thissmall degradation
of the solution occurs when the peak of the MgII line falls
within two ALHAMBRA filters. The distributionofthe outlier
fraction of the sources (Fig.5, red dots) follows a bimodalbe-
havior around z ∼1.4. Below this redshift (z ≤1.4), the
spec spec
minimization process tends to overestimate the photo-z solu-
tions, while it underestimates them at higher redshifts (z >
spec
1.4). This effect can be explained by i) the QSO color/redshift
degeneracy (i.e. the degree of similarity between the colors
at different redshifts; e.g. Richards et al. 2001) and ii) a line
misidentification(Croomet al. 2004). These degeneracies,still
Fig.7.SpectralenergydistributionfortheMELbest-fitsolutions.Open present in the ALHAMBRA data but to a much lesser extent
histogramtakesintoaccount allsources, whiletheshadedhistograms than for broadbandphotometry,are highlightedas grey shaded
consider only the sources with good photo-z solutions. We find that areasanddot-dashedlinesinFig.5forthecolor–colorandline
noneoftheALHAMBRABLAGN/QSOsiswell-representedbyastel- misidentification degeneracies. Further details of the origin of
lartemplate.Ofalltheextragalactictemplatesconsidered,themajority
these degeneraciesare given in Sect.4.3 where we explore the
(95.3%; 142 out of 149) of the sources with good solutions (i.e. no
natureoftheoutlierfractionofsources.
outliers)havebest-fittemplatescompatiblewithapureQSOorhybrid
QSO/ULIRGtemplate.
4.1.2. Dependencywithapparentmagnitude
As a test of the degeneracy introduced by the chosen tem-
plate database, we considered an alternative database of only AshighlightedinthecentralpanelofFig.5,wedonotfindany
QSOs and hybrid QSO/ULIRG templates (see indices #29–39 dependence of the accuracy on the apparent magnitude of the
inTable3andFig.3)forthe5sourceswithanormalgalaxyor source butonly a clear degradationof the solutionsis foundat
starburstbest-fitsolution.Wefoundthat:i)for2sourceswewere faintermagnitudes(m ≥ 22.0)causedbytheslightlynoisier
678
unabletorecoverinthiscasethephoto-z,castingsomedoubtson photometry(Δm ∼ 0.2atm = 23asindicatedbythecon-
678 678
the spectral classification of the sources or an incomplete tem- tinuouslineinthecentralpanelofFig.5).Ontheotherhand,the
plate database; ii) a correct photo-z was found for the other 3 outlierfractionshowsamoderatecorrelationwithapparentmag-
sourceswherehybridtemplateswithaweakerBLAGNcompo- nitude(bottompanelofFig.5),where∼62%oftheoutliershave
nent (10–20%) were selected by the best-fit solution. This last m ≥21.0.Nevertheless,althoughsomeoutliersmightbepro-
678
case highlightsthe degeneracyintroducedby the selected tem- ducedbynoisierphotometry,otherfactorsmightalsocontribute
platedatabaseforcertainhostandBLAGNluminosityregimes. tothecatastrophicfailures(seeSect.4.3).
A20,page10of17