A&A584,A106(2015) Astronomy DOI:10.1051/0004-6361/201526859 & (cid:2)c ESO2015 Astrophysics Pan-STARRS1 variability of XMM-COSMOS AGN (cid:2),(cid:2)(cid:2) I. Impact on photometric redshifts T.Simm1,R.Saglia1,2,M.Salvato1,R.Bender1,2,W.S.Burgett5,K.C.Chambers3,P.W.Draper4,H.Flewelling3, N.Kaiser3,R.-P.Kudritzki3,2,E.A.Magnier3,N.Metcalfe4,J.L.Tonry3,R.J.Wainscoat3,andC.Waters3 1 MaxPlanckInstituteforExtraterrestrialPhysics,Giessenbachstrasse,Postfach1312,85741Garching,Germany e-mail:[email protected] 2 UniversityObservatoryMunich,Ludwig-MaximiliansUniversitaet,Scheinerstrasse1,81679Munich,Germany 3 InstituteforAstronomy,UniversityofHawaiiatManoa,Honolulu,HI96822,USA 4 DepartmentofPhysics,DurhamUniversity,SouthRoad,DurhamDH13LE,UK 5 GMTOCorporation,251S.LakeAve.,Suite300,Pasadena,CA91101,USA Received30June2015/Accepted11September2015 ABSTRACT Aims.Upcoming largeareasky surveyslikeEuclidand eROSITA,which arededicated tostudying theroleof darkenergy inthe expansionhistoryoftheUniverseandthethree-dimensional massdistributionofmatter,cruciallydepend onaccuratephotometric redshifts.Theidentificationofvariablesources,suchasactivegalacticnuclei(AGNs),andtheachievableredshiftaccuracyforvary- ingobjectsareimportantinviewofthesciencegoalsoftheEuclidandeROSITAmissions. Methods.WeprobeAGNopticalvariabilityforalargesampleofX-ray-selectedAGNsintheXMM-COSMOSfield,usingthemulti- epochlightcurvesprovidedbythePan-STARRS1(PS1)3πandMediumDeepFieldsurveys.Toquantifyvariabilityweemployed asimplestatistictoestimatetheprobabilityofvariabilityandthenormalizedexcessvariancetomeasurethevariabilityamplitude. Utilizingthesetwovariabilityparameters,wedefinedasampleofvaryingAGNsforeveryPS1band.Weinvestigatedtheinfluenceof variabilityonthecalculationofphotometricredshiftsbyapplyingthreedifferentinputphotometrysetsforourfittingprocedure.For eachofthefivePS1bandsg ,r ,i ,z ,andy ,wechoseeithertheepochsminimizingtheintervalinobservingtime,themedian P1 P1 P1 P1 P1 magnitudevalues,orrandomlydrawnlightcurvepointstocomputetheredshift.Inaddition,wederivedphotometricredshiftsusing PS1photometryextendedbyGALEX/IRACbands. Results.Wefind that the photometry produced by the 3π survey issufficient toreliably detect variable sources provided that the fractionalvariabilityamplitudeisatleast∼3%.ConsideringthephotometricredshiftsofvariableAGNs,weobservethatminimiz- ing the time spacing of the chosen points yields superior photometric redshifts in terms of the percentage of outliers (33%) and accuracy(0.07), outperformingtheothertwoapproaches. Drawingrandompointsfromthelightcurvegivesrisetotypically57% ofoutliersandanaccuracyof∼0.4.AddingGALEX/IRACbandsfortheredshiftdeterminationweakenstheinfluenceofvariabil- ity.Althoughtheredshiftqualitygenerallyimproveswhenaddingthesebands,westillobtainnotlessthan26%ofoutliersandan accuracyof0.05atbest,thereforevariablesourcesshouldreceiveaflagstatingthattheirphotometricredshiftsmaybelowquality. Key words. catalogs – methods: data analysis – techniques: photometric – galaxies: active – galaxies: distances and redshifts – X-rays:galaxies 1. Introduction The Euclid mission aims to map the geometry of the dark Universe by accurately gauging distortions of galaxy shapes Understanding the expansion history of the Universe is one of mediated by weak lensing effects and constraining the pattern the fundamentalquestionsof modernastrophysics.Thisis par- of baryonic acoustic oscillations (BAO) from galaxy cluster- ticularlytrueforthenatureofdarkenergyanddarkmatter,the ingmeasurements.Applyingthesetwoindependentcosmologi- presumed agents behind cosmic acceleration and cosmological calprobes,Euclidwillsurveythethree-dimensionaldistribution structure formation.Unveiling the dark Universe, which repre- of structures with unparalleled accuracy out to redshift z ∼ 2, sents 96% of the cosmic matter-energy content, allows setting thereby covering the entire period of the accelerated expan- major constraints on the past, present, and future evolution of sion of the Universe that is driven by dark energy. Observing theUniverseandpromisestoprovideinsightintoradicallynew 15000deg2oftheextragalacticsky,Euclidwillprobethegrowth physics. Significant progress in our understanding is expected ofcosmicstructureintomographicbins.Thiswillbedetermined to be delivered by current and upcoming surveys, such as the throughphotometricredshifts(photo-z’s)thatneedto beasac- DarkEnergySurvey(DES;DePoyetal.2008;Mohretal.2008), curateasσ /(1+z) < 0.05atI ≤ 24.5(Bordoloietal.2010, Euclid(Laureijsetal.2011),andeROSITA(Predehletal.2007; z AB 2012) and as unbiased as possible. The mission will deliver Cappellutietal.2011). photo-z’sforanunprecedentedlargenumberofabouttwobillion galaxies and a million active galactic nuclei (AGNs; Laureijs (cid:3) Appendicesareavailableinelectronicformat et al. 2011; Amendola et al. 2013). The photometric redshifts http://www.aanda.org willbecomputedfromopticalandnear-infrared(NIR)photome- (cid:3)(cid:3) ThecataloguesofvariableAGNsareonlyavailableattheCDSvia anonymousftptocdsarc.u-strasbg.fr(130.79.128.5)orvia try,withEuclidprovidingtheNIRY,J,H-bandsandopticalob- http://cdsarc.u-strasbg.fr/viz-bin/qcat?J/A+A/584/A106 servationscollectedfromground-baseddeepwideareasurveys. ArticlepublishedbyEDPSciences A106,page1of22 A&A584,A106(2015) Complementary informationabout the large scale structure The surveyscarriedoutwith the Pan-STARRS1 instrument willbeprovidedbytheeROSITA all-skysurvey,observingthe deliver multi-epoch light curves for 3π of the sky and ten ad- hot X-ray Universe. The mission is expected to detect a very ditionalfields observedwith highercadencein five opticaland large sample of ∼105 galaxy clusters, about three million ob- NIR bands, thus providing variability information for millions scuredandun-obscuredAGNsand∼500000stars(Merlonietal. of AGNs. Motivated bythe aforementionedissues we used the 2012). This unique data set will allow the co-evolution of su- light curves of the PS1 3π and Medium Deep Field surveys to permassive black holes and their host galaxies to be studied definea sample of variableAGNsin each PS1 band.The sam- within the cosmic structure in unprecedented detail, provided ple is drawn from the well-characterized source list of X-ray- that accurate redshifts can be obtained for the point-like and selectedAGNsfromtheXMM-COSMOSsurvey.Utilizingthis extended X-ray sources. This requires precise identification of sample ofvariableobjects, we studyhow multibandvariability therespectiveopticalcounterpartsandsufficientmultibandpho- affects the quality of photometric redshifts in detail and assess tometry for the photo-z computation, again to be supplied by the achievable redshift accuracy using solely PS1 photometry deep wide area surveys. Various sophisticated methods have andPS1 photometryplusGALEX/IRACbands.In a follow-up beendevelopedtoderivephoto-z’s,whicheitherfollowanem- paper we will then use the same sample to search for correla- pirical approach by exploring the possible color-redshift com- tions between optical variability and physicalparameters, such binations of galaxies with the help of a spectroscopic training asblackholemass,luminosity,Eddingtonratio,andredshift. set(Csabaietal. 2003;Collister & Lahav2004;Vanzellaetal. This work is organized as follows. In Sect. 2 we describe 2004; Gerdes et al. 2010; Wolf 2009; Carliles et al. 2010) or the observationsandcharacterizeour data set, the sample defi- by applying template fitting (Giallongo et al. 1998; Bolzonella nitionandthedetectionofvariabilityforourAGNsisdepicted et al. 2000; Benítez 2000; Bender et al. 2001; Babbedge et al. inSects.3and4,thefittingtechniqueandthemethodsofstudy- 2004; Padmanabhan et al. 2005; Feldmann et al. 2006; Ilbert ing the effects of variability on photo-z calculations are intro- etal.2006;Finlatoretal.2007;Mobasheretal.2007;Brammer duced in Sect. 5, the photometric redshift results for variable et al. 2008; Assefetal.2008; Kotulla et al. 2009; Pelló et al. AGNs are presentedin Sect. 6, and Sect. 7 summarizesthe re- 2009;Barro et al. 2011;Dahlen et al. 2010,2013;Saglia et al. sults.ThroughoutthepaperweuseABmagnitudesandassume 2012).Althoughmodernphoto-zcodescomfortablyreachaccu- a ΛCDM cosmology with H = 70 kms−1Mpc−1, Ω = 0.3, 0 m raciesbetterthan5%forinactivegalaxies(Gabaschetal.2004; andΩΛ =0.7. Wolf et al. 2004; Grazian et al. 2006; Ilbert et al. 2006, 2009; Cardamoneet al. 2010),photometricredshiftsofsimilar preci- 2. Observationaldataset sionforAGNsrequiremuchmoreeffortandareavailablesolely for well-studied sky fields with extensive multiband coverage 2.1.ThePan-STARRS13πandMediumDeepFieldsurveys (Salvatoetal.2009,2011;Luoetal.2010;Hsuetal.2014). The observationaldata used in this work are based on the sur- The difficulties related to SED fitting of AGNs are mainly veyscarriedoutbythePan-STARRS1ScienceConsortiumcov- ering a period of about four years from November 2009 to driven by the fact that the spectrum is a superposition of the March 2014. The Pan-STARRS1 instrument is a single wide- AGNcorecomponentandthehostgalaxylight,plusthestrong fieldtelescopedesignedforsurveymodeoperationandislocated intrinsicvariabilityofAGNsacrosswavelength.While thefor- merdifficultycanbetackledbycollectinghighqualityempirical at the Haleakala Observatoryon the island of Maui in Hawaii. The f/4.4 optical system, comprising a 1.8 m primary mirror templatesofarepresentativesubsampleoftheAGNpopulation, anda0.9msecondary,generatesa3.3◦ fieldofviewincombi- the actualuncertaintiesintroducedby multibandvariability are nation with the PS1 gigapixelcamera (GPC1). The 1.4 Gpixel currentlynotknown.Thelattermay,however,introducefatalbi- ases into the photo-z accuracy for AGNs, which in turn affect detector is composed of a mosaic of 60 CCD chips each of 4800×4800pixelswithone10μmpixelmapping0.258arcsec the ability to study the evolutionof the X-ray luminosityfunc- ofthesky.ThePS1systemperformsimagingthroughfivemain tionandAGNclustering,aswellasthestarformationandstel- broadbandfiltersdenotedasg ,r ,i ,z ,y coveringopti- larpopulationpropertiesofAGNhostgalaxies(Airdetal.2010; P1 P1 P1 P1 P1 caltoNIRspectralregimeswithrespective“pivot”wavelengths Rosarioetal.2013). of481,617,752,866,and962nmanda“wide”filterw ,used P1 For this reason, it is paramount to reliably detect variable for large depth solar system observations(Hodapp et al. 2004; sourcesintheentireextragalacticsky.ThatAGNsexhibitstrong Kaiseretal.2010;Tonry&Onaka2009).ThePS1photometric variabilityinawidespectralrange,coveringradio,UV/optical, system is described in Tonry et al. (2012b), whereas passband X-ray,andγ-raywavelengths(Ulrichetal.1997)allowsidenti- shapesaredetailedinStubbsetal.(2010). ficationofAGNsonthebasisoftheirvariabilityproperties.The AmongtheseveralsurveysthatPS1accomplished,twoma- onsetofwide-areamassivetime-domainopticalimagingsurveys jor ones, the 3π survey and the Medium Deep Field (MDF) triggeredamultitudeofAGNvariabilitystudieswiththeaimof survey, are of primary importance for extragalactic studies. characterizingthe opticalvariability and establishing a method The 3π survey observed the three-quarter of the sky north of for quasar selection (Kelly et al. 2009, 2011, 2013; Kozłowski −30◦ declination in the five main filters officially starting in etal.2010,2011,2012,2013;MacLeodetal.2010,2011,2012; May 2010 and lasting until March 2014.By completionof the Schmidt et al. 2010, 2012; Palanque-Delabrouille et al. 2011; surveymission,eachobservablefieldshouldideallybeimaged Butler & Bloom 2011;Kim et al. 2011;Ruan et al. 2012;Zuo 12timesperfilter in six differentobservingnightswithtypical etal.2012;Andraeetal.2013;Zuetal.2013;Morgansonetal. exposure times of 30–60 s. Based on the requirements of the 2014;Sunetal.2014;Grahametal.2014;DeCiccoetal.2015; variousscience projects,the observationsfollowa complicated Falocco etal. 2015;Kokuboet al. 2014;Kokubo2015).These operatingscheduledictatingthateachindividualfieldisvisited investigations confirmed the general picture that AGNs show twice perobservingnightin a single filter with a temporalgap non-periodic,stochasticfluxvariabilityoccurringontimescales of20−30min.Thisenablesthedetectionofmovingobjectslike ofseveralmonthstoseveralyearswithafractionalamplitudeof asteroidsandnearearthobjects(Magnieretal.2013;Chambers typically∼10−20%. 2014). A106,page2of22 T.Simmetal.:Pan-STARRS1variabilityofXMM-COSMOSAGN.I. The MDF survey provides deeper multi-epoch data by re- (Brusa et al. 2010) from the XMM-COSMOS survey, which peatedly exposing a set of ten selected fields, with observa- havebeenobservedinthe 0.5–2keV, 2–10keV, and5–10keV tionsofeachfielddistributedthroughouttheperiodoftheyear energybandsfora totalof ∼1.5Ms in55 XMM-Newtonpoint- thatallowsfor1.3airmasspointingsatleast.Thescheduledca- ings (Hasinger et al. 2007; Cappelluti et al. 2009). The sur- dencecomprisesobservationsineachnightperiodicallyrunning vey reaches a depth of ∼5 × 10−16, ∼3 × 10−15, and ∼7 × throughthe five PS1 bands. One cycle starts with 8 × 113 s in 10−15ergs−1cm−2Hz−1inthesementionedbands. the g and r bands in the first night, followed by 8 × 240 s In thisworkwewantto focusonQSOs, so we firstlimited P1 P1 in the i band the second night, and finishing with 8 × 240 s the sample to those 495 X-ray detected sources that have a se- P1 in the z band the third night. Afterwards the next cycle be- cure opticalcounterpart(Brusa et al. 2010)and that are classi- P1 gins by again integrating 8 × 113 s in the g and r bands. fiedaspointlikeonthebasisofthemorphologicalanalysisper- P1 P1 Additionally for each of the three nights on either side of full formedbyLeauthaudetal.(2007)usingdeepHST/ACSimages. Moon,8×240sinthey bandareobtained(Sagliaetal.2012; In addition, to ascertain that our photometry is not influenced P1 Tonryetal.2012a).Thelargenumberofexposurestakeninthe byblendingeffects,wecross-matchedtheCOSMOS-ACScata- course of the MDF survey deliver very deep stack images and logue(Leauthaudetal.2007)onthepositionsofourobjectsand the observing strategy produces light curves permitting exten- removedeverysourcefromoursamplethathasanearbyobject sivevariabilityinvestigations. within1.5arcsec1.Thisreducedthefinalsampleto384sources. TherawscienceframesexposedwiththePS1telescopeare ThroughoutthisworkweusePSFphotometryfortheseobjects. reducedbythePS1ImageProcessingPipeline(IPP)conducting Outofthe384sources,249havereliablespectroscopicred- standard proceduresof image calibration, source detection, as- shifts (Lilly et al. 2009; Trump et al. 2007). For the rest, the trometry,andphotometry.Theresultingobjectcataloguescanbe availabilityofdeepandhomogeneousphotometryin 31bands, accessedviathePublishedScienceProductsSubsystem(PSPS) includingintermediate-andnarrow-bandfilters(Taniguchietal. database (Heasley 2008). Amongst the various data products 2007), allowed computing high quality photometric redshifts storedinthePSPSdatabaseinviewofvariabilitystudies,theob- with an accuracy of 0.015 with only a handful of outliers jectanddetectiontablesareveryimportant.Theobjecttablelists (Salvatoetal. 2011).Ourfinalsample also contains47objects thecollectedinformationaboutallsourcesidentifiedasanastro- thatareclassifiedasstarsbytheirspectralfeatures.Inthefollow- nomicalobject in multiple detections, such as sky coordinates, inganalysisweincludeboththeAGNsandthestarsinorderto mean and stack magnitudesin all bands, and summary proper- beabletocomparethePS1observationaldataforthesedifferent ties obtained from model fits like the PS1 star/galaxy separa- objecttypes. tor.Thedetectiontablecontainsallavailableinformationabout theindividualdetectionsofeachobjectcomprisinginstrumental fluxes, zeropoints, exposure times, and the PSF model fit pa- 3. Sampledefinition rameters, to name but a few. Magnitudes in the “AB system” 3.1.Limitingmagnitudes (Oke&Gunn1983)foreachbandpasscanbeobtainedfromthe instrumental flux Finstr in the considered filter and the respec- To identify these objects within the PS1 3π and MDF sur- tivezeropointzpstoredinthedetectiontableunderthetermsof veys again, with the latter including XMM-COSMOS in the magAB = −2.5log10(Finstr)+zp. The PS1 IPP providesinstru- MDF04field,wematchedthepositionsofthecounterpartsofthe mentalfluxes computedfrom PSF modelfits suitable for point X-raysourcestothePS1cataloguesandrecovered285sources sourcesandKronfluxes(Kron1980),givinga meaningfulflux within the 3π survey and 313 within the MDF04 survey, here- estimation forextendedsourceslike galaxies.The Kron flux is after referredto as the “3π sample” and the “MDF04 sample”. definedasthefluxwithintheKronradius,withthelattergiven TheangularseparationoftheXMM-COSMOSandPS1coordi- by 2.5times the first radialmomentof the flux in the PS1 IPP. nates is less than 0.25 arcsec for all of these sources. We note From the AB magnitudes calibrated fluxes in units of 3631Jy that within the photometry errors, none of the 285 objects of maybeobtainedaccordingto the 3π sample has a median magnitude exceeding the 5σ me- (cid:2) dianlimitingmagnitudesforindividual3πsurveyexposuresof F =10−0.4magAB = (cid:2) 363f1ν(Jhyν()h−ν1)A−1(νA)(dνν)dν, (1) 2m2o.1re(gthPa1n), ∼210..91(mrPa1g),(2M1.o6rg(aiPn1s)o,n20e.t9a(lz.P12)0,1a4n)d. T1h9e.9re(fyoPr1e)wbye do not apply a further magnitude cut within the 3π sample. with1Jy=10−23ergs−1cm−2Hz−1.TherightpartofEq.(1)fol- However,amongthe313AGNsoftheMDF04sample,wefind a number of sources that are considerably fainter than the ex- lowsfromthedefinitionofthe“band(cid:3)passABmagnitud(cid:4)e”,where pected5σlimitingmagnitudesforMDFsingleexposures.Since ν denotesthe photonfrequency, f ergs−1cm−2Hz−1 the flux ν weobservefromtheMDF04detectiontablethattheindividual density, h the Planck constant, and A(ν) the capture cross sec- MDF04 exposure times are on average a factor of two longer tion(Oke&Gunn1983;Tonryetal.2012b).Thecapturecross than the single 3π exposuretimes andsince the signal-to-noise sectionmeasurestheprobabilityofreleasinganelectronperin- ratio(S/N√)isproportionaltothesquarerootoftheexposuretime comingphotonwithinthedetector.Inthecourseofthiswork,the S/N ∝ t ,weexpectanaverageincreaseinthe5σlimiting exp √ variabilityparametersdefinedinSect.4.1arecomputedfromthe magnitudesoftheMDF04surveyof|−2.5log 2| ∼ 0.4mag fluxescalculatedafterEq.(1)andthecorrespondingfluxerrors. 10 compared to the 3π survey. Adding this correction of 0.4 mag Throughoutthiswork,weusePS1datafromtwoprocessingver- to the respective 3π survey values quoted above, the approx- sions,PV1.2forthe3πsurveyandPV2fortheMDFsurvey. imate 5σ median limiting magnitudes for single detections of the MDF04 survey become 22.5 (g ), 22.3 (r ), 22.0 (i ), P1 P1 P1 2.2.XMM-COSMOS 1 This is a reasonable value considering that 75% of the PS1 The initial sample of objects building the starting point of our frames have a FWHM below 1.51, 1.39, 1.34, 1.27, 1.21 arcsec for studies is a catalogue of 1674 X-ray selected point sources g ,r ,i ,z ,y (Magnieretal.2013). P1 P1 P1 P1 P1 A106,page3of22 A&A584,A106(2015) 21.3 (z ), and 20.3 (y ). We applied a magnitude cut in each within one night. In contrast, a typical MDF04 light curve is P1 P1 band by discarding everyobject with median magnitudelarger dividedintoseveralobservingblockslastingaboutthreetofour thantheselimitingmagnitudes. monthswithahighsamplingrateofabouteightobservationsper nighttakenapproximatelyeveryonetothreedays.Theindivid- ual observing blocks are separated by gaps of around seven to 3.2.Removaloffalsedetections ninemonthswithnoobservations.Thefulllightcurvecoversa periodofaboutfouryears. Since thePS1 GPC1 is a prototypecameraconstructedforfast Prior to performing a variability analysis, it is instructive readout consisting of almost 4000 CCD cells there are sev- eraldifferentdefectsand a huge numberof detector edgesthat to visually inspect the light curves in order to identify possi- ble problems. Looking at a large number of light curves from can lead to false detections (Metcalfe et al. 2013). Moreover, there can be reflections that lead to ghost images, diffrac- the 3π and MDF04 surveys, we discovered a significant num- ber of measurements that imply variability of up to several tion spikes of bright stars, or masked pixels, potentially re- tenthsofamagnitudewithinoneobservingnight.Thismustbe sultinginspuriousdetectionsandmisleadingphotometricmea- compared with typical intra-nightoptical variability, termed as surements. To reduce the contamination by “bad” and “poor” micro-variability,of∼0.01–0.1magfornormalAGNs.Onlyex- detections we downloaded only those detections from the tremeobjectslikeblazarsoroptically-violentlyvariable(OVV) PSPS database with none of the following flags set: FITFAIL, objects may show micro-variability of a few 0.1 mag within SATSTAR, BADPSF, DEFECT, SATURATED, CR LIMIT, onenight(Gopal-Krishnaetal. 2003;Gaskell& Klimek2003; MOMENTSFAILURE,SKYFAILURE,SKYVARFAILURE, Stalinetal.2004,2005;Gupta&Joshi2005;Carinietal.2007). SIZE SKIPPED, POORFIT, PAIR, BLEND, MOMENTS SN, Consideringthatthenightlyobservationsofthe3πandMDF04 BLEND FIT, ON SPIKE, ON GHOST, and OFF CHIP. In addition we removed detections suffering from very bad see- surveys are separated by ∼30–60 min at most, which corre- spondsto even shortertime intervals in the AGN rest frame, it ing or focus shifts by excluding PSF model fits with ps- is very likely that the observed micro-variability is not physi- fWidMajor > 6 arcsec and extremely elliptic model fits with psfWidMinor/psfWidMajor < 0.65.Tominimizetheeffectsof callyfounded,butratherstemsfromlowqualitymeasurements withunderestimatederrorbars.Moreover,wedetectthisshort- pixel masking on the measurements, we only kept detections term variability not only in AGN light curves, but also in the with psfQf > 0.85 and psfQfPerfect >0.85, i.e. PSF model fits stellarlightcurvesofoursample.Inadditionanumberoflight withfewerthan15%maskedpixelsweightedbythePSF.Finally curvesof the MDF04 surveyexhibitfew fatal outlier measure- toexcludeveryfaintmeasurements,weworkedwith5σdetec- ments, sometimes deviating from the bulk of data points by tionsaccordingtopsfFlux/psfFluxErr >5. several magnitudes. These problems are illustrated in Fig. 1, We note that the vast majority of the detections in the 3π showingthe raw lightcurvesof six AGNs in the left-handcol- and MDF04 samples have zeropointerrorsthat are more accu- umnand of six objectsclassified as stars in the right-handcol- rate than 10 millimag from the “Ubercal” photometric calibra- umn for both the 3π and MDF04 surveys. The light curves of tion (Schlafly et al. 2012). All detections of the MDF04 sam- the AGNs exhibit clear signs of variability on timescales of plehavespecifiedzeropointerrors,buta substantialfractionof months to years, whereas the stellar light curves are compara- the detections of the 3π sample have Δzp = −999. For these bly flat. However, the intra-night variations in both the stellar unspecified zeropoint errors, we have assumed a conservative andAGNlightcurvesareessentiallyundistinguishable.Thisis value of Δzp = 0.07 in the calculation of the photometric er- also true considering the occurrence of catastrophic outliers in rors. Saturation for individual detections with typical 3π sur- theMDF04lightcurves. vey exposure times sets in at g , r , i ∼ 13.5, z ∼ 13.0, P1 P1 P1 P1 Toprobethequalityofthebulkofthemeasurementsinthe and y ∼ 12.0 (Magnier et al. 2013).Althoughwe do not ex- P1 MDF04 lightcurves,we re-computedthe photometryfrom the pect any of our objects to be affected by saturation in any of rawimagesfora 0.4◦×0.4◦ fieldusingthe MunichDifference the PS1 bands, since none of our sources has z (Subaru) < 17, ImagingAnalysis (MDIA)pipeline described in Koppenhoefer fewPS1detectionsexistthataresignificantlybrighterthanthese etal.(2013).Comparingtheresultinglightcurveswiththeones saturation limits, and even negative magnitude values occur. created by the PS1 IPP, we observe that the overall trends of Because these bright detections have very likely been wrongly variability,visiblein thelightcurvesegments,arethesame for associated with our sources, we excluded every detection with bothpipelines.Thisimpliesthatthevastmajorityoftheretained g ,r ,i ,z ,y < 14.0. Furthermore,even after applyinga P1 P1 P1 P1 P1 PS1 IPP detections exhibit sufficient quality. However, the oc- 5σcut,weobservedafewmagnitudevaluesofsingleexposures currence of fatal outliers is much higher in the PS1 IPP light amongstoursampledetectionsthatliewellabovethe5σlimit- curves,suggestingthatthesemeasurementsareindeednotcred- ingmagnitudesoftwo-yearstackimagesfromtheMDF04sur- ible.Theoriginofthefataloutliersmaybespuriousdetectionsin vey(forreferencemag (g ) ∼ 24.5,Sagliaetal.2012),with lim P1 thevicinityofoursources,whichhavebeenwronglyassociated some being as faint as g ∼ 40. For this reason we addition- P1 withthelatter2.Itisclearthatthepresenceofthesecatastrophic ally removed every detection with g ,r ,i ,z ,y > 23.5, P1 P1 P1 P1 P1 outliers means that any variability measurement would be sig- 23.5,23.5, 22.5,21.5, therebydiscarding these extremelyfaint nificantly biased towards very large variability amplitudes. On measurements. thesegroundswedecidedtoremovethefewfataloutliersfrom the(PS1IPP)MDF04lightcurves. 3.3.Lightcurvetreatment Finally, in view of the non-negligible number of detec- tionsshowingconsiderableshort-termvariabilityontimescales The light curves of the 3π survey suffer from extreme sparse sampling because they consist of pairs of observations carried 2 Thisissupportedbyover-plottingallsingledetectionsremainingaf- out within ∼30 min in one night, followed by large temporal ter thestepsdescribed inSect.3.2onto amuchdeeper MDF04 stack gapsofseveralmonthsuntilthenextobservation.Onlythedata image,revealingthepresenceofanumberoffalsedetectionsthatcan- acquiredatthe endof 2009comprisesupto eightobservations notbeassociatedwithanopticalcounterpart. A106,page4of22 T.Simmetal.:Pan-STARRS1variabilityofXMM-COSMOSAGN.I. AGNs Stars Δmag = 0.592 Δmag = 0.116 21.8 16.95 21.6 16.90 mag [AB] 22220111....8024 oooooooo oooo oo oooo oo mag [AB]111666...788505 oooooooo oooo oo oo oo 16.70 20.6 16.65 20.4 5.52 5.54 5.56 5.58 5.60 5.52 5.54 5.56 5.58 5.60 10+4 10+4 MJD MJD Δmag = 0.624 Δmag = 0.196 18.3 21.2 21.0 18.2 mag [AB] 22220000....2468 oooooooo ooo oo oo mag [AB] 111788...901 oooooooo oo oo oo oo 20.0 17.8 5.52 5.54 5.56 5.58 5.60 5.52 5.54 5.56 5.58 5.60 10+4 10+4 MJD MJD Δmag = 0.797 Δmag = 0.108 21.6 19.55 21.4 19.50 21.2 mag [AB] 2222200001.....24680 ooooo oooo oo oooo oo mag [AB]11119999....33440505 oooooooo oooo oo o oo 19.25 20.0 19.8 19.20 5.52 5.54 5.56 5.58 5.60 5.52 5.54 5.56 5.58 5.60 10+4 10+4 MJD MJD Δmag = 3.451 Δmag = 3.103 23 23 22 22 o o B] 21 o B] 21 A A mag [ 112890 oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooo mag [ 1290 ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooooooooooo 18 17 17 5.55 5.60 5.65 5.55 5.60 5.65 10+4 10+4 MJD MJD Δmag = 2.525 Δmag = 3.347 23 23 22 mag [AB] 222012 oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooooooooo mag [AB] 11228901 ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooooo 19 17 18 16 5.55 5.60 5.65 5.55 5.60 5.65 10+4 10+4 MJD MJD Δmag = 2.012 Δmag = 1.704 22.0 22.5 21.5 22.0 mag [AB] 1122299001.....05050 ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooooooo mag [AB] 22220011....0505 ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo ooooooooooooooooooooooooooooooooooooooooooooo 18.5 19.5 18.0 19.0 5.55 5.60 5.65 5.55 5.60 5.65 10+4 10+4 MJD MJD Fig.1.Rawlightcurves(gP1band)ofsixAGNs(leftcolumn)andsixstars(rightco(cid:5)lumn(cid:6)).Theto(cid:5)pthr(cid:6)eepanelsshowdatafromthe3πsurvey,the bottomthreepanelsfromtheMDF04survey,respectively.ThevalueΔmag=max mag −min mag quotesthemaximumamountofvariability ineachlightcurve. A106,page5of22 A&A584,A106(2015) less than one hour, we calculated nightly averages from the Toevaluatethevariabilityamplitude,weemploythenormal- observationsofthe3πandMDF04surveys.Toassignaconser- izedexcessvariance(Nandraetal.1997)definedby vativeandmeaningfulerrortoeachaveragedfluxormagnitude ⎛ (cid:3) (cid:4) ⎞ vblaaarlrgusees,ehrcorowonrisnbigdaerlsarirbngugetsbncoeatgthtleitgrhiaebblpeorusetcsatehtnteecrem,owefaenp,toaaiksnetwsaeswlleirtahrsopsrmothianeltlslaewrrgriotehrr σ2rms =(cid:11)s2−σ2err(cid:12)/(cid:3)f¯(cid:4)2 = (cid:3)f1¯(cid:4)2 ⎜⎜⎜⎜⎜⎜⎜⎜⎝(cid:9)i=N1 (fNi−−f1¯)2 −(cid:9)i=N1 σN2err,i⎟⎟⎟⎟⎟⎟⎟⎟⎠ (7) ofthetwoestimates (cid:7)(cid:8) σ(cid:3)f¯(cid:4)= 1 (cid:9)n (cid:3)f − f¯(cid:4)2 (2) TwhitehnNo,rmfi,alainzdedσe2exrrc,iedsessvcarirbiainngceth(ehesraemafeteqrujaunsttiteiexsceasssinvaEriqa.n(c4e)). n(n−1) i depictstheresidualvarianceaftersubtractingtheaveragestatis- (cid:7)(cid:8)(cid:9)n i=1 ticalerrorσ2errfromthesamplevariances2ofthelightcurveflux. 1 Δf¯ = (Δf)2. (3) Sincetheexcessvarianceisnormalizedtothesquaredmeanof Gauss n i=1 i the flux, it(cid:19)essentially specifies the squared fractional variabil- (cid:3) (cid:4) ity Fvar = σ2rms (Edelsonet al. 1990).The excess varianceis Hereσ f¯ denotesthestandarderrorofthearithmeticmeancal- an estimatorofthe intrinsic fractionalvarianceofa sourceand culatedfrom n values f observedin one night,and Δf¯ the providesameaningfulmeasureofthevariabilityamplitudeeven i Gauss forsparselysampledlightcurves.Theexcessvariancehasbeen uncertaintyfollowingGaussianerrorpropagationoftheindivid- ualerrorsΔf. frequentlyusedinX-rayvariabilitystudiesandwasfoundtobe i correlatedwiththeblackholemass,X-rayluminosity,andX-ray spectralindexofAGNs(Nandraetal.1997;Turneretal.1999; 4. Detectionofvariability Leighly1999;Georgeetal.2000;Papadakis2004;O’Neilletal. 2005;Zhouetal.2010;González-Martínetal.2011;Pontietal. 4.1.Statisticalmethodstocharacterizevariability 2012; Lanzuisi et al. 2014). In this work we adopt the excess varianceasanestimatoroftheopticalvariabilityamplitudecal- Considering the tremendous data volumes provided by the culatedfromthePS1lightcurvesineachofthefivebands. PS1 surveys and future massive time-domain optical surveys, The uncertainty of σ2 caused by Poisson noise alone has e.g.LSST(Ivezicetal.2006),identifyinglargenumbersofvari- rms been determined by Vaughan et al. (2003). They performed able sources requires using variability estimators that can be obtained with low computationaleffort. To quantify variability Monte Carlo simulations thereby generating a random “red noise” light curve (power spectral density with logarithmic wethereforeutilizetwovariabilityparametersthatpossesswell slopesbetween−1and−2,seeAppendixAfordetails),adding knownstatistical properties,and owingto their fast evaluation, Poisson noise by drawing fluxes from the Poisson distribution theycanbeeasilyappliedtoverylargesamples. and then measuring the excess variance of the smeared light Toestimatetheprobabilitythatanobjectisactuallyvarying, curve.Thewidthoftheσ2 distributionresultingfrom104“ob- wefirstcalculatetheobservedχ2 givenby rms obs servations”ofthelightcurveisfoundtobewellfittedby (cid:3) (cid:4) χ2obs =(cid:9)i=N1 fiσ−2err,f¯i 2 (4) err(cid:3)σ2rms(cid:4)= (cid:7)(cid:20)(cid:20)(cid:20)(cid:20)(cid:20)(cid:8)⎛⎜⎜⎜⎜⎜⎜⎜⎜⎝(cid:21)N2 · (cid:3)σf¯2e(cid:4)r2r⎞⎟⎟⎟⎟⎟⎟⎟⎟⎠2+⎛⎜⎜⎜⎜⎜⎜⎜⎜⎜⎝(cid:22)σN2err · 2(cid:3)Ff¯v(cid:4)ar⎞⎟⎟⎟⎟⎟⎟⎟⎟⎟⎠2· (8) fromthelightcurveconsistingofNmeasuredfluxes f within- i dividualerrorsσ and arithmetic mean f¯. Then assuming an err,i Inthecaseofa lowintrinsicvariabilityamplitudeorveryfaint intrinsicallynon-varyingsource,wecomputetheprobabilitythat a χ2 larger than the observedone could just emergeby chance sources s2 ∼ σ2 , the excess variance will thus be small and err duetoPoissonnoisefollowing can even be negative. In this situation the first term of Eq. (8) (cid:10) dominates. In contrast, if the variability signal is strong, s2 (cid:10) (cid:3) (cid:4) ∞ (cid:3) (cid:4) P χ2 ≥ χ2obs = f χ2,N−1 dχ2 (5) σ2err, and the second term of Eq. (8) dominates(Vaughanet al. χ2 2003). It is well known that there are additional uncertainties obs (cid:3) (cid:4) connectedtoanexcessvariancemeasurement.Thesearerelated where f χ2,d.o.f. is the probability density function of to the stochastic nature of AGN variability and the light curve the χ2-distribution with N − 1 degrees of freedom (d.o.f.). samplingpattern.However,asdiscussedinAppendixA,thebias factorassociatedwiththeseuncertaintiesisexpectedtobeclose SubsequentlywedefinethevariabilityindexV accordingto tooneforourobservationaldata,sowedonotcorrectforthese (cid:3) (cid:4) V =−log P χ2 ≥ χ2 , (6) errors. 10 obs FollowingLanzuisietal. (2014),whenusingsolely the ex- cessvariancetoquantifyvariability,we regarditasa detection providingameasureofthestrengthoftheevidenceofvariabil- ofvariabilityif ity. This method is depicted in McLaughlin et al. (1996) and wassubsequentlyappliedbyPaolilloetal.(2004),Youngetal. (cid:3) (cid:4) (2012),andLanzuisietal.(2014).ValuesofV = 1.0,V = 1.3, σ2 −err σ2 >0. (9) rms rms andV = 2.0correspondinglyexpressthatwerejectthenullhy- pothesisofanintrinsicallynon-variablesourcewith90%,95%, Tobeabletodefinerobustsamplesofvaryingsources,wetake and 99% confidence. The V parameter is a useful tool for pre- advantageoftheinformationprovidedbybothoftheintroduced selecting variable objects, yet beyond that it contains no infor- variabilityparameters.Throughoutthiswork,unlessquoteddif- mationaboutthemagnitudeofthefluxvariations. ferently, we consider it as a detection of variability when the A106,page6of22 T.Simmetal.:Pan-STARRS1variabilityofXMM-COSMOSAGN.I. Table1.NumberofvariableAGNsfromthe3πsample. 3πsampleinourdatasetwiththemeannumberofobservations of 4.5 (g ), 3.9 (r ), 3.1 (i ), 3.4 (z ), and 3.5 (y ), which P1 P1 P1 P1 P1 Filter N >2 V >1.3 σ2 −err(σ2 )>0 (1)∧(2) mightexplaintheobservedlackofvariability.Furthermore,con- rms rms sideringthecorrespondingfractionsofvariablesourcesfromthe (1) (2) MDF04 sample, we find more variable objects in the i band g 151 107 92 90 P1 P1 thaninthez band.GiventhesamplingrateoftheMDF04sur- r 116 76 55 54 P1 P1 vey, essentially all AGNs in our sample show some amountof i 50 27 14 14 P1 z 95 38 39 37 variabilityduringthe nearlyfouryearsof repeatedmonitoring, P1 y 36 12 8 8 and the vast majority are variable in multiple bands. After av- P1 eragingthe MDF04 lightcurves, the mean numberof observa- Notes.Numberthatfulfiltheconditionsgiveninthecolumnheadings. tionsfor all bandsis givenby 69.1(g ), 70.5 (r ), 83.6 (i ), P1 P1 P1 N:numberofdetections 88.2 (z ), and 51.9 (y ). The MDF04 survey produced sig- P1 P1 nificantly fewer observations in the y band, while the g P1 P1 Table2.NumberofvariableAGNsfromtheMDF04sample. and r bands suffered the most from fatal outliers. This cata- P1 logue of well characterizedXMM-COSMOS sources therefore Filter N >2 V >1.3 σ2 −err(σ2 )>0 (1)∧(2) allows us to study the variability properties of ∼180 AGNs in rms rms (1) (2) the “blue” bandsand more than100 AGNsin the “red” bands. g 187 187 185 184 ThesecataloguesofvariableAGNsareavailableattheCDS(see P1 rP1 184 183 182 181 AppendixCfordetails). i 165 165 163 162 P1 As a summarizing example, Fig. 2 shows the nightly av- z 135 135 132 131 P1 eraged light curves of one AGN that is varying in all five y 76 76 74 74 P1 PS1bands,alongwiththelightcurvesofonestarthatdoesnot Notes.Numberthatfulfiltheconditionsgiveninthecolumnheadings. varyinanyband.WhereastheAGNlightcurvesexhibitapprox- N:numberofdetections imatelysimultaneousvariationsinallfivePS1bandswithsignif- icantamplitudesofabout∼0.5mag,thestellar lightcurvesare constantwithinthephotometricerrorswithonlytwooutliersap- probability for spurious variability is less than 5% and the ex- pearinginthey band,whichhoweverdonotcauseadetection cess varianceis greaterthan zerowithin its error,expressedby P1 ofvariabilityaccordingtocondition10. thecondition V >1.3∧σ2 −err(σ2 )>0. (10) rms rms 4.3.Comparisonofthe3πandMDF04variability We apply these two variability parameters to identify variable objectsusingthelightcurvesofthe3πandMDF04surveysby Since the 3π survey covers three-quarters of the sky, it allows testingforcondition10ineachofthefivePS1bandlightcurves. the propertiesof millions of AGNs to be investigated and pro- vides optical photometry for the sources to be observed with EuclidandeROSITA.Itisthereforeimportanttounderstandto 4.2.Cataloguesofvariableobjects whatextentwecanprobevariabilityusingthesparselysampled Toinvestigatevariabilityinthe3πandMDF04samples,weonly 3π light curves as compared to the much better sampled MDF consideredobjectswithmorethantwodetections(N >2).From light curves, which are howeveronly available for ten selected the nightlyaveragedflux lightcurvesof the point-likeand iso- skyfields.Toaddressthisquestionweperformedavisualcom- lated sources, we calculated the V parameter and the normal- parisonofourAGNlightcurvesofthetwosurveys.Examination izedexcessvariance.ThenumbersofAGNssatisfyingV >1.3, ofalargenumberofalllightcurvesrevealsthatthevastmajor- σ2 −err(σ2 ) > 0, or bothof these conditionsare listed for ityofthedetectionsofbothsurveysyieldverysimilarmagnitude rms rms each filter in Table 1 for the 3π sample and Table 2 for the values;i.e.,thenightlyaveragedlightcurvesofthe3πsurveyfit MDF04 sample, respectively. The numbers reveal that, when almost perfectly in the correspondingones of the MDF04 sur- estimating the probability of variability, the V parameter has a vey.ThatisillustratedinFig.3,showingthelightcurvesofthe tendencytoselectmoreobjectsasvariablethantheexcessvari- MDF04surveyinblack,over-plottedwiththerespective3πob- ance, with the latter quantifying the net amplitude of variabil- servationsinredforthreeAGNs.However,someofthe3πsur- ity.Nevertheless,theintersectionofthetwovariabilitydetection veylightcurvesarestillcontaminatedbyfataloutliersthatcan methods, given in the last column of Tables 1 and 2, is large; only be identified as such with the additional information pro- i.e., both methods are consistent for identification of variable vided by the MDF04 light curves also showing the long-term sources. Considering the numbers in this last column, 59.6% trendsinvariability.Suchacaseisvisibleinthebottompanelof (g ), 46.6% (r ), 28.0% (i ), 38.9% (z ), and 22.2% (y ) Fig.3,whereasthetwootherAGNlightcurvesofthesameplot P1 P1 P1 P1 P1 oftheAGNswith N > 2aredetectedasvariableinthe3πsur- agreeverywell. veyand98.4%(g ),98.4%(r ),98.2%(i ),97.0%(z ),and A morequantitativecomparisonofthetwo surveysinview P1 P1 P1 P1 97.4%(y )intheMDF04survey,respectively. of the variability amplitude can be done by contrasting the ex- P1 AftercomparingthedifferentPS1filters,thefractionofvari- cess variancesasmeasuredfromthe lightcurvesof the 3πand ableAGNsisfoundtobelargerforthe“bluer”bands,andyetthe MDF04survey.ThisisdisplayedinFig.4forthoseAGNswith i bandofthe3πsamplecomprisesfewervaryingsourcesthan a positive g and r band excess variance. Even though the P1 P1 P1 the “redder” z band. We stress, however, that one should be excessvarianceis calculatedfromonlyaboutsix pointsatbest P1 carefulwhencomparingthesefractionssincethe3πlightcurves inthecaseofthe3πsurvey,asopposedtotypically∼70points suffer from extreme sparse sampling, so that the ability to de- in the MDF04 survey, the two measurements yield similar es- tectvariabilitycruciallydependsonthenumberofobservations. timates for a large portion of the tested sample. Nevertheless, In fact the i bandhas on averagethe fewest detectionsin the the error of the excess variance is considerably larger for the P1 A106,page7of22 A&A584,A106(2015) AGN(XID1) Star(XID60462) Fig.2.NightlyaveragedMDF04lightcurvesshowingallfivePS1bandsoftheAGNwithXID1(top)andthestarwithXID60462(bottom). A106,page8of22 T.Simmetal.:Pan-STARRS1variabilityofXMM-COSMOSAGN.I. 20.8 20.6 20.4 mag [AB]mag [AB] 122900...802 oooooooooooooooooo ooooo ooooooooooooooooooooo ooooooooooooooooo 19.6 19.4 5.52 5.54 5.56 5.58 5.60 10+4 MMJJDD 20.8 20.6 ag [AB]ag [AB] 222000...024 oooooooooooooooooo ooooo oooooooooooooooooooooo mm 19.8 ooooooooooooooooo 19.6 19.4 19.2 5.52 5.54 5.56 5.58 5.60 10+4 MMJJDD 21.4 21.2 21.0 mag [AB]mag [AB] 22220000....2468 ooooooooooo oo ooooooooooooooooooooo ooooooooooooo 20.0 19.8 5.52 5.54 5.56 5.58 5.60 10+4 MMJJDD Fig.3.MDF04lightcurves(g band)ofthreeAGNsover-plottedwith P1 thecorresponding3πlightcurvesinred. 3πsample,sothevariabilitysignalcannotbedetectedaswellas Fig.4.ExcessvariancecalculatedfromtheMDF04lightcurvesversus withintheMDF04survey. the respective value computed from the 3π light curves for all AGNs The differences in the σ2 measurements are particularly rms withσ2 >0intheg band(top)andr band(bottom).Theblackline largeforobjectswhose3πlightcurvessuggestaconstantsource rms P1 P1 representstheone-to-onerelation,andtheerrorbarsshowtheaverage just because they miss the variability occurringin between the valueoferr(σ2 ). rms observations, which is however visible in the MDF04 light curves. Such light curves give rise to the dramatic outliers ap- in each band. In this way we are unaffected by different num- parent in the top left region within each panel of Fig. 4. For bers of available bands per object. This means that from the thesereasonsthefractionsofvariableAGNsreportedinTable1 samples defined in Sect. 4.2, we are left with 40 type-1 AGNs aresignificantlylowerthanthecorrespondingfractionsobtained fromthe 3πsurveyand 75type-1AGNsfromthe MDF04sur- usingtheMDF04survey.Forexample,welose39%inthevari- vey, for which we can compute photometric redshiftsfrom the ability detection for the g band and 52% for the r band as five PS1 bands. In the followingphoto-zanalysis, we focuson P1 P1 comparedtotheMDF04survey.Nonetheless,wepointoutthat the results obtainedwith the MDF04 sample,since it is almost althoughtheuncertaintiesintheexcessvariancemeasurements twiceaslargeasthe3πsample,andthesamplingpatternofthe are large, it is possible to obtain a reasonable variability am- MDF04lightcurvesallowsforamorethoroughinvestigationof plitude estimation by utilizing the light curves of the 3π sur- the effects of variability on photo-z calculations. Amongst the vey for a large number of our sources. Considering Fig. 4, 75 AGNs from the MDF04 sample, 72 sources vary in all five we may assume that all objects with logσ2 (MDF04) > −3 PS1 bands,with three sourcesvaryingin onlythree bands. We rms and logσ2 (3π) > −3, i.e., all sources varying at least at the point out that although our variability detection threshold de- rms 3%level,haveawell-estimatedvariabilityamplitudeevenwhen finedinEq.(10)correspondstoa1σdetectionregardingtheex- using3πsurveylightcurves.Whenassumingthisvariabilitycut, cessvariance,72(g ),72(r ),72(i ),71(z ),and61(y ) P1 P1 P1 P1 P1 theexcessvariancevaluesofbothsurveysaresimilarfor91%of of the 75 sources satisfy σ2 − 3err(σ2 ) > 0. The redshift rms rms thegP1bandobjectsand89%oftherP1bandobjects.Thismeans distribution of these sources is shown in Fig. 5. For compari- thatthe3πsampleofvariableobjectsispurebutnotcompleteat sonreasonsacleansampleofnon-varyingAGNswouldbevery the 3% level of variability,therefore the observationsprovided useful. We stress, however,thatwe are unableto define such a bythe3πsurveyallowvariableobjectstobeselectedforthree- sample because the vast majority of our AGNs vary in at least quarters of the sky, at least as long as the intrinsic variability oneband,andthefewnon-varyingsourceslackingphotometry amplitudeislarge. inseveralbands. 4.4.Definitionofthephoto-zsample 5. PhotometricredshiftsofvariableAGNs Thesampleofvaryingsourcesforourphoto-zanalysisisdrawn 5.1.Multibanddata by selecting only those AGNs from the 3π and MDF04 sam- ples,whicharedetectedasvariableaccordingtocondition10in TostudytheeffectsofAGNopticalvariabilityonthecalculation atleastoneof thePS1 bandsandhaveat leastoneobservation ofphotometricredshiftsin detail,we usedthe PSF photometry A106,page9of22 A&A584,A106(2015) Table4.TemplateSEDsusedinthiswork. ModelID Modelname 1 I22491_60_TQSO1_40 2 I22491_50_TQSO1_50 3 I22491_40_TQSO1_60 4 pl_I22491_30_TQSO1_70 5 pl_I22491_20_TQSO1_80 6 pl_I22491_10_TQSO1_90 7 pl_QSOH 8 pl_QSO 9 pl_TQSO1 Fig.5.Redshiftdistributionofthe75AGNsfromtheMDF04sample usedinthephoto-zanalysis. Notes.SamemodelnamesasinTable2ofS09. Table3.Photometriccoverageusedfortheredshiftcomputation. a χ2 minimization, comparing the observed flux with the tem- Filter Telescope (cid:3)λeff(cid:4) FW(cid:3)H(cid:4)M platefluxineachbandtodeterminethemostlikelyredshift,SED Å Å template,andintrinsicextinction.Whenaimingtocalculatered- FUV GALEX 1546 234 shiftsforAGNs, it is ofprimaryimportanceto utilize a library NUV GALEX 2345 795 ofSEDtemplatescoveringthevarietyofpossiblesuperpositions gP1 PS1 4900 1149 of the AGN and hostgalaxy emission components.To account rP1 PS1 6241 1398 for this, we use the well-tested model set employed in Salvato i PS1 7564 1292 P1 et al. (2009, 2011; hereafter S09 and S11, respectively). This z PS1 8690 1039 P1 library comprises hybrid templates with varying contributions y PS1 9645 665 P1 (90:10,80:20,...,20:80,10:90)ofseveralhostgalaxytypesand IRAC1 Spitzer 35634 7412 differenttypes of AGNs (type-1, type-2,QSO1, QSO2). These IRAC2 Spitzer 45110 10113 templatesaredescribedindetailinS09.Sinceallofoursources Notes.ValuescalculatedfromthetransmissioncurveswiththeLePhare areluminouspoint-liketype-1AGNs,weonlyconsiderthenine code. templateswithasignificantQSO-fraction,listedinTable4.This comparablysmallnumberofmodelshelpstoreducedegeneracy ofthefivebroadbandPS1filtersforwhichwehavevariability betweentemplatesandredshifts.Moreover,sincewearedriving information.Wedeterminedphotometricredshiftsobtainedwith the fitting routine towards QSO-dominated templates, we min- justthesefivePS1bandstoestimatetheachievablephoto-zac- imize the color-redshift degeneracy that is produced by AGNs curacyforaphotometrysetconsistingexclusivelyofwavelength andgalaxiesoccupyingsimilarregionsincolorspaceforcertain bandsthatshowstrongvariability. redshifts(Richardsetal.2002;Wolfetal.2004). In addition, we derived photometric redshifts by extend- ToaccountforGalacticextinction,wecorrectedeachofour ing the photometry set with the near-UV (NUV) and far-UV photometricmeasurementsintheopticalPS1bandsbythecorre- (FUV) bands of the Galaxy Evolution Explorer (GALEX) and spondingtotalabsorptionA inmagnitudes.Theextinctionval- λ the IRAC1/IRAC2 mid-infrared(MIR) Spitzer bands. We used ueswereobtainedfromtheNASA/IPACExtragalacticDatabase the GALEX-COSMOS catalogue of Zamojski et al. (2007), (NED) and are based on the extinction maps of Schlafly & which provides de-blended, PSF-fitted NUV, and FUV mag- Finkbeiner (2011). The GALEX photometry was corrected for nitudes in the AB system, to find the nearest object within Galactic extinction by subtracting 8.612 × 0.0167 from the 0.25arcsectotheCOSMOScoordinatesofeachofoursources. NUV magnitudes and 8.290 × 0.0167 from the FUV magni- Amongthe75AGNsfromtheMDF04sample,fiveobjectslack tudes, respectively. These A values were calculated with the λ GALEXphotometry.RegardingtheIRACphotometry(Sanders LePharecodeusingtheGalacticextinctionlawofCardellietal. et al. 2007; Ilbert et al. 2010) we searched for the closest (1989)asa functionofcolorexcess E(B−V),assuming A = V counterpartwithin1.0arcsec tothe opticalcoordinatesofeach R ×E(B−V) with R = 3.1. We did not performa Galactic V V of our objects. However, only five IRAC counterparts deviate extinction correction for the IRAC bands, because the extinc- by more than 0.25 arcsec from the corresponding optical co- tionintheMIRwavelengthrangeistypicallymuchlessthanthe ordinates. From the COSMOS-IRAC catalogue, we then ex- photometricerrorsoftheobservations. tractthe 1.9 arcsec aperturefluxesof the IRAC1/IRAC2 bands ToobtainarepresentativelibraryofexpectedintrinsicSEDs and obtained total fluxes by dividing the aperture fluxes by inthe ABphotometricsystem ofthePS1 bands,we performed 0.765 (IRAC1) and 0.740 (IRAC2), following the instruc- the following steps. First we multiply the template SEDs with tions given in the readme file attached to the catalogue (see the filter transmission curves of the used bands and integrate also Surace et al. 2005). The total fluxes were finally trans- overthewavelengthrangecoveredbythelatter.ThentheSEDs formed from μJy to AB magnitudes according to magAB = areredshiftedwithinarangeofz = 0.02–5,applyingabinsize −2.5logFtot +23.9.Thetotalwavelengthcoverageoftheused of Δz = 0.01. Subsequently, we create a grid of redshift and bandsislistedinTable3. hostextinctionvaluesbyallowingforarangeofE(B−V)val- uesbetween0and0.5withstepsof0.05totakecareofthein- trinsic reddening caused by the AGN host galaxy. For the lat- 5.2.Fittingtechnique ter we apply the SMC extinction law of Prevot et al. (1984), TheSED fitting is realizedwiththe publiclyavailable LePhare which was found to produce the best photo-z results for the code (Arnouts et al. 1999; Ilbert et al. 2006), which performs XMM-COSMOSsourcesinS09.TheLy absorptionproduced α A106,page10of22