ebook img

NASA Technical Reports Server (NTRS) 20150002691: The Suzaku View of Highly Ionized Outflows in AGN. 1; Statistical Detection and Global Absorber Properties PDF

1.2 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview NASA Technical Reports Server (NTRS) 20150002691: The Suzaku View of Highly Ionized Outflows in AGN. 1; Statistical Detection and Global Absorber Properties

MNRAS Advance Access published January 21, 2013 MNRAS(2013) doi:10.1093/mnras/sts481 The Suzaku view of highly ionized outflows in AGN – I. Statistical detection and global absorber properties Jason Gofford,1‹ James N. Reeves,1,2 Francesco Tombesi,3,4 Valentina Braito,5,6 T. Jane Turner,2 Lance Miller7 and Massimo Cappi8 1AstrophysicsGroup,KeeleUniversity,KeeleST55BG,UK 2DepartmentofPhysics,UniversityofMarylandBaltimoreCounty,Baltimore,MD21250,USA 3X-rayAstrophysicsLaboratoryandCRESST,NASA/GSFC,Greenbelt,MD20771,USA 4DepartmentofAstronomy,UniversityofMaryland,CollegePark,MD20742,USA 5INAF-OsservatorioAstronomicodiBrera,ViaBianchi46,I-23807Merate,Italy 6DepartmentofPhysicsandAstronomy,UniversityofLeicester,UniversityRoad,LeicesterLE17RH,UK 7DepartmentofPhysics,OxfordUniversity,DenysWilkinsonBuilding,KebleRoad,OxfordOX13RH,UK D 8INAF-IASFBologna,ViaGobetti101,I-40129Bologna,Italy o w n lo a d e Accepted2012November22.Received2012November22;inoriginalform2012October14 d fro m h ttp ABSTRACT ://m WepresenttheresultsofanewspectroscopicstudyofFeK-bandabsorptioninactivegalactic n ra nuclei (AGN). Using data obtained from the Suzaku public archive we have performed a s.o statisticallydrivenblindsearchforFeXXVHeαand/orFeXXVILyαabsorptionlinesinalarge xfo sample of 51 Type 1.0−1.9 AGN. Through extensive Monte Carlo simulations we find that rdjo u statisticallysignificantabsorptionisdetectedatE(cid:2)6.7keVin20/51sourcesatthePMC ≥ rna 95percentlevel,whichcorrespondsto∼40percentofthetotalsample.Inallcases,individual ls.o absorption lines are detected independently and simultaneously amongst the two (or three) arg/ available X-ray imaging spectrometer detectors, which confirms the robustness of the line t N A detections. The most frequently observed outflow phenomenology consists of two discrete S A absorptiontroughscorrespondingtoFeXXVHeαandFeXXVILyαatacommonvelocityshift. G o FromxstarfittingthemeancolumndensityandionizationparameterfortheFeKabsorption dd a components are log(N /cm−2) ≈ 23 and log(ξ/ergcms−1) ≈ 4.5, respectively. Measured rd H S outflow velocities span a continuous range from <1500 km s−1 up to ∼100000 km s−1, pa c withmeanandmedianvaluesof∼0.1cand∼0.056c,respectively.Theresultsofthiswork e F areconsistentwiththoserecentlyobtainedusingXMM–Newtonandindependentlyprovides ligh sfrtarocntigonevoifdebnocthefroadritoh-eqeuxieistteanncderoadfivoe-rlyouhdigAhlGyNioinniztehdecloirccaulmunniuvcelresaer.materialinasignificant t Ctr on A p Keywords: line:identification–galaxies:active–galaxies:nuclei–X-rays:galaxies. ril 5 , 2 0 1 3 gravitational radii (R = GM/c2) in size (Morgan et al. 2008; 1 INTRODUCTION g Chartas et al. 2009; Dai et al. 2010) and so the spectral analysis Observational evidence for outflows and winds in active galactic of absorption features imprinted on the X-ray continuum by cir- nuclei (AGN) is seen in multiple energy regimes, ranging from cumnuclear material in an AGN is a powerful diagnostic of the the prominent radio-jets seen in radio-loud sources, to the broad physicalconditionsoftheenvironmentinthevicinityofthecen- absorptionlinesobservedinbroadabsorptionline(BAL)quasars, tralsuper-massiveblackhole(SMBH),ofthedynamicsandkine- through to the photoionized ‘warm absorber’ which is frequently matics of the outflowing material, its chemical composition and observed in the soft X-rays (e.g. Crenshaw, Kraemer & George its ionization state. Understanding how such winds are formed, 2003;Blustinetal.2005;McKernan,Yaqoob&Reynolds2007). their physical characteristics and their energetic output are of vi- Gravitational micro-lensing studies have shown that the primary tal importance when it comes to studying how the interplay be- X-ray emission region in AGN is of the order of a few tens of tweentheaccretionandejectionflowsatsmallradiicanaffectthe hostgalaxyonlargerscales(e.g.Ferrarese&Merritt2000;King & Pounds 2003; Sazonov et al. 2005; Ciotti, Ostriker & Proga (cid:4)E-mail:[email protected] 2009). (cid:5)C 2013TheAuthors PublishedbyOxfordUniversityPressonbehalfoftheRoyalAstronomicalSociety 2 J.Goffordetal. IntheX-rayregimethepresenceofphotoionizedmaterialinAGN sorptionfeatures–pointtoanoriginmorelikelyassociatedwitha iswellestablished,withatleast50percentofobjectsshowingdirect windwhichislaunchedfromthesurfaceoftheaccretiondiscitself spectroscopicevidencefordiscreteabsorptionlinesandphotoelec- (e.g.Poundsetal.2003b;Reevesetal.2009;Goffordetal.2011; tric edges in their soft-band (E < 3 keV) spectra (e.g. Reynolds Tombesietal.2012).Inthisscenariotheinferredmassoutflowrates 1997;Crenshawetal.2003;Blustinetal.2005),withtypicalline- fordiscwindsareoftencomparabletothoseofthematterwhich of-sightvelocitiesrangingfromafewhundredtoaroundathousand accretesontothecentralblackholeandtheconsequentmechani- kms−1(Blustinetal.2005;McKernanetal.2007).Detailedmod- calpowercanalsobeasizeablefraction(i.e.≥fewpercent)ofan ellingathighspectralresolutionwithphotoionizationcodessuchas AGN’sbolometricluminosity(e.g.Chartasetal.2002;Poundsetal. xstar(Kallmanetal.2004)oftenfindsthesoft-bandabsorberto 2003b;Gibsonetal.2005;Reevesetal.2009;Goffordetal.2011; bedescribedbycolumndensitiesandionizationparameters1inthe Tombesietal.2012)makingthemapromisingmeansoflinkingthe range of log(N /cm−2)∼20−23 and log(ξ/ergcms−1)∼0−3, small-andlarge-scaleprocessesatplayingalacticevolutionover H respectively.Theseparametersimplythatthewarmabsorbersare cosmictime. typically located on fairly large distances from the central black hole,andperhapsassociatedwithawindoriginatingfromthepu- 1.1 WhySuzaku? tative parsec scale torus (Blustin et al. 2005) or with the latter stagesofanaccretiondiscwindwhichhaspropagatedouttolarger InthisworkweusearchivalSuzakuobservationsofalargesampleof D radii(Proga&Kallman2004).Byvirtueoftheirlowoutflowve- AGNtocharacterizethepropertiesofhighlyionizedoutflowsinthe ow locitiesthesoftX-raywarmabsorbersarethoughttoonlyhavea FeKband.Todate,themostcomprehensivestudyoftheseoutflows nlo a weakfeedbackeffectintheirhostgalaxy.Indeed,themechanical hasbeenconductedbyTombesietal.(2010a,hereafterT10A)who de d plooww,etryipmicpaalrlyted(cid:3)b0y.0i1ndpievridceunatlwofaramnaAbGsoNrp’stiboonlcoommeptroicnelnutmsiinsovseirtyy p4e2rsfoourmrceeds(awsiytshte1m01atoibcsnearrvraotwio-nbsa)nodb(tai.ien.e3d.5fr–o1m0.t5hkeeXVM)Man–aNlyeswistoonf from (thLebo∼l)0(e.5.gp.eBrcluesnttinofetLbaoll.t2h0o0u5g)h,twnheiccehssiasrysigfonrififceaendtblyaclkowtoeraftfheacnt atirncuhuivme,inustihnegFaesKimbpalnedb.aTsheelinbeasmeloindeelmtooddeelsccorinbseistthedeoAfGaNpocwoner- http://m the host galaxy (Hopkins & Elvis 2010). However, Crenshaw & law, narrow Gaussians and, where required, neutral absorption to n Kraemer(2012)haverecentlyshownthatthis∼0.5percentthresh- approximateforanyspectralcurvature.Thisphenomenologicalap- ras.o oldcanbeexceededprovidedthemechanicalpowerisintegrated proachyieldedafittothe4–10keVenergybandofmostsources xfo oovfearfaelwlUmVodaenrdatXe--lruamyianbossoitryptlioocnalcAomGpNo.nents,atleastinthecase awnhdicFhewXXasVIsLufyfiαciaebnstotropteinoanbllienethsewsiythstoeumtanteiecdsienagrctohtoafkFeeadXdXiVtiHoneaαl rdjourn ferMedobreyrXecMenMtl–yNtehwethoinghaenrdthSruozuagkhupautthaingdhelarrgXe-rraeyffeecnteivrgeieasre(aEo∼f- sappepcrtoraaclhcroemsupllteexditiyncinotnotiancucuomunpta.rMamoreeteorvsewr,hTic1h0Awefroeulnadrgtehlaytctohnis- als.o rg 5–10keV)haveshownthatabsorption,specificallyintheformof sistentwiththosefoundbyauthorswhoconductedamorethorough a/ very highly ionized resonant absorption lines associated with the fit using the entire XMM–Newton bandpass. However, while this t N K-shell(1s→2p)transitionsofFeXXVandFeXXVI,isalsomani- approachissuitableformostsourcesitisimportanttonotethatin AS A festedinthehardX-rayspectrumofasignificantfractionoflocal thosewhichhavemorecomplexX-rayspectra,e.g.thosewhichare G AGN.Whileevidenceforsuchabsorptionlineswasinitiallycon- veryheavilyabsorbedorthosewithstronghardexcesses,usinga od d fined to detailed studies of individual sources (e.g. Pounds et al. narrow-band fit can yield a model which is a poor representation ard 2003b;Reeves,O’Brien&Ward2003;Risalitietal.2005;Turner ofthedatawhenextrapolatedtoconsiderthehigherandloweren- S p etal.2008;Cappietal.2009;Giustinietal.2011)therecentsys- ergies.Theonlywaytoovercomethislimitationisbyperforming ac e (te2m01a0tiac,a2r0c1h1ivaa,l2X01M2M)h–aNseswhtoownnsttuhdaytFceonXdXuVc–tXeXdVbIyabTsoomrpbtieosnielitnaels. aobdseetravialetodrbyrowahdi-cbhanodffaenrsalayssius.ffiScuizeanktluyisbrcouardrenbtalnydtphaesson(liy.e.X0-r.6ay– Fligh ainrethperelsoecnatliunntihveeXrs-era(yzs<pe0c.t1r)a.oMfo(cid:2)r4eo0vpeerr,cseuncthoofurtafldoiow-squhiaevteAaGlsNo 5ex0ckeessV)antodathlleowCofomrptthoenerfefflecetcstioofnscooftm-bpaonndenatbstoorpbteiocno,ntshterasinoeftd- t Ctr on been detected in a small sample of local broad line radio galax- simultaneously. This makes it the ideal instrument to confidently A p ies (BLRGs; Tombesi et al. 2010b, hereafter T10B) which thus constraintheunderlyingcontinuumandtorobustlyassessforthe ril 5 suggeststhattheymayrepresentanimportantadditiontothecom- presenceofhighlyionizedoutflowsinAGN. , 2 0 monlyheldAGNunificationmodel(e.g.Antonucci1993;Urry& 1 3 Padovani1995). 2 SAMPLE SELECTION In comparison to the soft-band absorbers these hard X-ray absorbers generally have much more extreme parameters, with The sample was selected from the Data Archive and Transmis- log(NH/cm−2) ≈ 23–24 and log(ξ/ergcms−1) ≈ 3–6, and their sion System2 (DARTS; Tamura et al. 2004) which contains all outflowvelocitiesrelativetothehostgalaxycanreachmildlyrela- publiclyavailableSuzakuobservationscategorizedbytheclassifi- tivisticvalues.Whilealternativeexplanationshavebeenpositedin cation ofthetargetsource.AninitialsampleofAGN was drawn the literature, e.g. resonant absorption by highly ionized material bypositionallycross-correlatingthetargetedpointingcoordinates inacorotatingandopticallythinplasmaabovetheaccretiondisc ofallpubliclyavailable(asoftheendof2011December)Suzaku (Gallo & Fabian 2011), the large inferred velocities – combined observations of extragalactic compact sources against the known withtheshorttime-scalevariabilitysometimesexhibitedbytheab- positions of sources contained in the VERONCAT catalogue of Quasars&AGN(Ve´ron-Cetty&Ve´ron2010).TheVERONCAT hasanextensivelistoflocalAGN,allofwhichhavebeenquan- 1Theionizationparameterisdefinedasξ =Lion/nR2 (Tarter,Tucker& titatively classified based upon their optical properties using the Salpeter1969),whereLion isthe1−1000Rydbergionizingluminosity,n istheelectrondensityandRisthedistanceoftheionizingsourcefromthe absorbingclouds. 2http://darts.jaxa.jp/astro/suzaku/ TheSuzakuviewofhighlyionizedoutflowsinAGN 3 criteria introduced by Winkler (1992). Observations of non-AGN which were matched by virtue of their having a similar position toaknownAGNonthesky,suchasthoseofextragalacticultra- luminousX-raysources(ULXs;Feng&Soria2011)orX-raybright supernovae,weresystematicallyexcluded.Onlythoseobservations with exposures long enough to ensure that the net source counts between2and10keVinthesourcerest-frameexceeded∼15000 wereretained(typicalexposureswere(cid:2)50ks)suchthatanarrow (i.e.EW=30eV)absorptionfeaturewasdetectableat95percent fromMonteCarlosimulationsatsourcerest-frameenergiesofup to8–9keV(seeSection4.4).Inthecaseofthehighred-shiftquasar APM08279+5255,whichshowsevidenceforabsorptionlinesat rest-frameenergiesofE>10keV(e.g.Chartasetal.2009;Saez, Chartas&Brandt2009;Saez&Chartas2011),themeasureofnet sourcecountswasinsteadtakenfortheentireX-rayimagingspec- trometer (XIS) bandpass (i.e. 0.6–10 keV in the observer frame) Figure1. Histogramshowingthedistributionofsourcesincludedinthis D becausetheFeKfeaturesareshiftedtolowerenergiesduetothe work. The high red-shift QSO APM 08279+5255 (z = 3.91) has been ow highred-shiftoftheAGN. omittedforscalingpurposes. nlo a Ashighlyionizedoutflowsarethoughttooriginateatrelatively d e d sTmomalbledsiisteatnacle.s20fr1o1ma,t2h0e1c2e,nhterraelanftuecrleTu1s1(aen.gd.TG1o2f,forersdpeetctailv.e2ly0)1,1i;t from isimportantthattheprimarycontinuumemissionfromthecentral http nucleus,ratherthanthatwhichisreprocessed/scatteredbycircum- ://m nuclear material out of line of sight, is directly observed so that n suchoutflowscanbedetected.TothisendweexcludeallType2 ras .o sourcestomakesurethatallsourceswereopticallythintoX-rays x fo below 10 keV. Therefore only those sources with a Type 1.0–1.9 rd jo orientation to the line of sight, as per the classifications listed on u rn the NASA/IPAC Extragalactic Database (hereafter NED), in the a ls VERONCATcatalogueitselfandthroughliterarysourceswerein- .o rg cluded. Note that we conservatively include the radio-quiet AGN a/ ESO 103-G035 (z = 0.01329) which, despite being classified as t N A Type2.0intheVERONCATandonNED,isoftenregardedasa S A Type1.9Seyfertintheliterature(e.g.Warwick,Pounds&Turner G 1988)byvirtueofthepresenceofamoderatelybroadHαlineinits od d opticalspectrum(Phillipsetal.1979).Observationdetailsforall Figure2. Histogramshowingthelogarithmofthetotalfront-illuminated ard sourcesincludedintheheterogeneouslyselectedsamplearelisted XIS2–10keVcountsinallfittedspectra.Instackedspectra(seeTableA1) S p inAppendixAintheonlinesupportinginformation.Thereare99 thetotalco-addedcountsareconsideredhere,ratherthanthecountsineach ac observations of 51 AGN spanning a wide range of spectral types individualsequence. e F andradio-properties.AsshowninTable1thesampleisdominated ligh bylow–moderateinclinationSeyfert1.0–1.5galaxies(28/51;34/51 0.1040), PKS 0558−504 (z = 0.1372) and PBC J0839.7−1214 t C ifNarrowLineSeyfert1sareincluded)andcontainscomparatively (z=0.198).ThegravitationallylensedquasarAPM08279+5545 tr o n fewhigh-inclination(Type1.8–1.9)Seyferts(6/51).Therearesix is by far the most distant object in the sample with a red-shift A radio-loudsourcesinthesample,includingallfiveoftheBLRGs tohfezfi=tted3.9sp1e.cAtrsasshpoawnntwino oFridge.r2s othfemtoatganlit2u–d1e0bkuetVarecoaupnptsrofxoir- pril 5 includedintheT10Boutflowcasestudy,andfiveQSOs.Thedistri- , 2 mately normally distributed about the peak and mean number of 0 butionsofsourcered-shiftandtotal2–10keVcounts(fortheXIS 1 105counts. 3 front-illuminateddetector)areshowninFigs1and2,respectively. TheAGNarepredominantlylocal,with∼90percentofthesam- plebeinglocatedatared-shiftofz(cid:3)0.1,butalsoincludesmore 3 DATA REDUCTION distantobjectssuchasPDS456(z=0.1840),1H0419−577(z= 3.1 XISreduction Table1. Sourceclassifications. TherearefourXISCCDcamerasaboardSuzakuwhichcoverthe energy range of 0.3–12.0 keV. The XIS 0, XIS 2 and XIS 3 are Sourceclassification Number front-illuminated,whiletheXIS1isback-illuminatedandprovides Sy1.0–1.2 17 superiorsensitivitybelow∼1keVbutalowereffectiveareaanda Sy1.5 11 higherbackgroundatharderX-rayenergies.TheXIS2suffereda Sy1.8–1.9 6 chargeleakin2006November3 andisthereforeonlyavailablein NLSy1 6 observationstakenbeforethisdate.Allspectrawerereducedusing BLRG 6 QSO 5 Total 51 3http://heasarc.gsfc.nasa.gov/docs/suzaku/news/xis2.html 4 J.Goffordetal. v6.11oftheHEASoft4 suitefollowingtheprocessoutlinedinthe fitstatistic.HardX-rayfaintand/orhighbackgroundobservations SuzakuABCDataReductionGuide.5 Eventswereselectedusing wheresourcecountrateswere<4percentofthePINtotalwerenot ASCAX-raygrades0,2,3,4and6,adoptingthestandardscreening consideredinouranalysis.OnlyAPM08279+5255andPDS456, criteriasuchthatdatawereexcludediftaken:(1)within436sof amounting to a total of five observations, meet this criterion (see passagethroughtheSouthAtlanticAnomaly(SAA),(2)withinan TableA1).AswiththeXISdata,thefinalPINexposuresandthe Earth elevation angle (ELV) <5◦ and/or (3) with Earth day-time totalbackground-subtractedrest-frame15–50keVcountsarelisted elevation angles <20◦. Hot and flickering pixels were removed inTableA1. fromtheXISimagesusingtheCLEANSISscript.Sourcespectrawere extractedfromwithincircularregionsofradius1.5≤r<3.0arcmin 4 ANALYSIS to ensure good coverage of the source events, while background spectra were typically extracted from offset circles of the same 4.1 Spectralfitting radiuswithcaretakentoavoidthechipcornerscontainingtheFe55 calibrationsources.Forintrinsicallyfaintsourcesthebackground Adetailedbroad-bandspectralanalysisofallsourceswasconducted spectrawereextractedfromcircleslargerthanthoseofthesource usingversion12.6.0qoftheXSPECspectralfittingpackage.Allspec- andtheratioofsource/backgroundareawasaccountedforwithan tralmodelsaremodifiedbytheappropriatecolumndensityofGalac- appropriatebackgroundscalingfactor.Redistributionmatricesand ticabsorptionusingvaluestakenfromtheLeiden/Argentine/Bonn D ancillaryresponsefilesforeachobservationweregeneratedusing SurveyofGalacticHI(Kalberlaetal.2005)whichwereobtained ow n thetasksXISRMFGENandXISSIMARFGEN,respectively. fromtheon-lineversionoftheNH FTOOL.6 ThevaluesofGalactic loa Wherepossiblespectraobtainedfromthefront-illuminatedXIS absorptionforeachsourcearelistedinTableA1.XIS-FIdatawere de d detectorswerecombinedintoasinglesourcespectrum(hereafter typicallyconsideredbetween0.6and10.0keV,whileXIS-BIwere fro XIS-FI)usingMATHPHAinordertomaximizesignal-to-noiseratio only included between 0.6 and 5.0 keV due todecreasing S/Nin m (S/N)intheFeKband.Mostoftheobservationsinthesamplehave theFeKbandwhichcouldhamperabsorptionlinedetection.All http dataforatleasttheXIS0andXIS3withafurther22alsohaving XISspectrawereexcludedbetween1.6and2.1keVduetouncer- ://m datafortheXIS2(seeTableA1).ForSWIFTJ2127.4+5654(OB- taintiesincalibrationassociatedwiththeSiKdetectoredge.Where n ra SID702122010)onlydataforthefront-illuminatedXIS3camera available,PINdatawereincludedtocoveratleastthe15–40keV s.o areavailablebecausetheXIS0wasnotoperationalduringtheob- energyrange. xfo servation.Spectrafortheback-illuminatedXIS1(hereafterXIS-BI) A constant multiplicative factor was included in all models to rd jo werereducedusingthesameprocessasoutlinedaboveandwere accountfortheXIS/PINcross-normalizationwhosevaluedepends urn analysedsimultaneouslyasaseparateinputspectrum.AllXISspec- not only on the nominal pointing of the observation, but also the als traweregroupedtohaveaminimumof50countsperenergybin versionoftheSuzakupipelinewithwhichthedatahavebeenpro- .org toenabletheuseoftheχ2 fitstatistic.NetXISexposuresandthe cessed.ThecurrentXIS/PINratiossuitableforversion2processed a/ total background-subtracted 2–10 keV count rates (in the source dataare1.16(1.18)fortheXIS(HXD)nominalpointingpositions.7 t N A rest-frame),forboththeco-addedXIS-FIandtheXIS-BIdetectors, However,thecross-normalizationwasupto∼5–6percentlowerfor S A arelistedinTableA1. dataprocessedwithversion1ofthepipeline,correspondingtoan G o XIS/PIN ratio of 1.09(1.13)7 in the XIS (PIN) nominal pointing d d a positions.Whilethedifferenceisonlysmalltheadditionaluncer- rd 3.2 HXD/PINreduction taatinEty>ca1n0hkaveVe,apcaorntisciudlearralbyleinefhfeacrdtoXn-trhaeycbornigtihntusuomurpcaersamweittheras Space Spectra from the Hard X-ray Detector (HXD) Positive Intrinsic high PIN count rate where the model can become driven by the Flig Negative(PIN)siliconediodeswerealsoreducedusingthemethod hardX-rayband.Toaccountforanysystematiceffectsassociated h t C oacuctloirndeidnginttohethSeuzsackrueeDnaintagRcerdituecritaiondeGscuriidbeedanpdr,eavgiaoiuns,lpy.roAcessstehde wstaitnhttphaerainmsetrtuemriennetaalcchromsos-dcealltiobrvaatiroyn±w5eptehrecreefnotraebaolulotwthtehveacluoens- tr on A HXD/PINisacollimatingratherthananimaginginstrument,the suggestedbytheSuzakuteamtotakeintoaccountanysystematic p contributionofboththeinstrumentalnon-X-raybackground(NXB) errors. ril 5 andthecosmicX-raybackground(CXB)needstobeindependently Thereare20AGNinthesamplewhichhavetwoormoreSuzaku , 2 0 accounted for when estimating the total background. The NXB observations. In APM 08279+5255, IC 4329A, MCG -6-30-15, 13 wasgeneratedusingtheappropriateresponseandtunedeventfiles Mrk 841, NGC 5506, NGC 5548 and PKS 0558−504 the differ- (backgroundmodelD)foreachobservation.Commongoodtime ent observations are similar in spectral shape which allows them intervals (GTIs) in the event and background files were selected tobeco-addedusingtheappropriaterelativeweightingfactorsto usingMGTIMEandspectralfileswereextractedusingXSELECT.PIN takeintoaccountdifferencesinindividualexposures,withthere- detectordead-timewasaccountedforusingtheHXDDTCORtaskand sultant time-averaged spectra for these sources being used in all theNXBbackgroundexposurewasincreasedbytentimestoreduce subsequentanalyses.1H0419−577andNGC2992bothhadob- theeffectsofphotonnoise.TheCXBcontributionwassimulated servationswhichweretakenindifferentSuzakunominalpointing usingtheformofBoldt(1987),combinedwiththeNXBtoforma positionswhichcouldinfluenceco-adding.However,effectivearea totalbackgroundspectrumusingMATHPHA,whichwassubsequently wasalwaysconsistenttowithin±10percentandsothespectrawere subtractedfromthesourcespectrumwithinXSPEC. stillco-addedusingthemeanoftheresponsefiles.Anyadditional Allbackground-subtractedPINspectrawerebinnedtoatleastthe systematicuncertaintyintroducedwasadequatelyaccountedforby 3σ levelabovebackground(typically>5σ)toenabletheuseofχ2 6http://heasarc.gsfc.nasa.gov/cgi-bin/Tools/w3nh/w3nh.pl 4http://heasarc.nasa.gov/lheasoft/ 7http://www.astro.isas.jaxa.jp/suzaku/doc/suzakumemo/suzakumemo- 5ftp://legacy.gsfc.nasa.gov/suzaku/doc/suzaku_abc_guide.pdf 2007-11.pdf TheSuzakuviewofhighlyionizedoutflowsinAGN 5 thevariableXIS/PINcross-normalization.In3C120andMrk509, listed in Tables D2 and D3 for single- and multi-epoch spectra, whichbothshowednotablevariabilitybetweenobservations,spec- respectively.Adescriptionoftheunderlyingmodellingassumptions tra were jointly fitted depending on the extent of their spectral and associated caveats is presented in Appendix C in the online variability between epochs. In 3C 120 we followed the analysis supportinginformation. methodofKataokaetal.(2007)andT10Bbyco-addingOBSIDs 700001020, 70001030 and 70001040, which are all of a similar spectralshapeandfluxlevel,andjointlyfittedthemwithOBSID 4.2.1 WarmabsorptionI:fullcovering 700001010 which has a more prominent underlying soft-excess. Dependingonthepropertiesoftheinterveningmaterial(suchasits InMrk509,whichiswellknownforhavingastrongsoft-excess ionizationstateandcolumndensity)absorptionbycircumnuclear (Mehdipouretal.2011),OBSID701093010,thestackedOBSIDs materialcanaddsignificantspectralcurvaturetotheobservedX-ray 701093020,701093030,701093040andOBSID705025010were spectrumandcanthereforehaveadirecteffectonthecontinuum fittedsimultaneouslytoaccountfortheobservedvariabilityinthe andlineparametersmeasuredinbroad-bandmodels.Inthiswork softX-rayband. wemodelwarmabsorptioncomponentsusingasuiteofxstar(v. The remaining nine sources with more than one observation 2.2.1bc)tableswhichareallgeneratedassuminginputvalueswhich (3C111,Fairall9,Mrk766,NGC1365,NGC3227,NGC3516, are ‘typical’ for local Seyfert galaxies8 (e.g. T11). The resultant NGC3783,NGC4051andPDS456)showedstrongspectralvari- xstar tables cover a wide range of parameter space in terms of D ability in both the shape of the spectrum and/or drastic changes column densities [1018 < log(N /cm−2) ≤ 1024], and ionization ow in flux state which made co-adding impractical. In these sources parameter[−3<log(ξ/ergcmsH−1)≤6]whichmakesthemwell nlo a the available spectra were fitted simultaneously and a model was d suitedforaccountingforallmannersofwarmabsorption. e d cfeownsatrdudcitteiodnatolfdreeescpraibraemtheteerosbasserpvoesdsibsplee.cItnraNl GvaCria1b3i6li5ty,OwBitShIDass whFeurell-nceocveesrsinargywtoaramchaiebvseoraptgioonodzofintetsoathreeisnocfltuXde-drayinbmanodd;eilns from 702047010and705031010aredominatedbyverydeepFeKab- somecasesmorethanoneabsorptionzoneisneeded.Inthesecases http sorption lines. These lines are not present in OBSID 705031020, thecolumndensityandionizationparameterofeachzoneareal- ://m pdoomssiinbalyteddusetateto(steheeMsaoiuorlcineodertoaplp.in2g010intfooradeqtauialssio-sfctahtetervianrgi-- lwohwiecdhtcoovnasirsytsinodfempeunldtiepnletlylaaynedrsreopfrgesaesn.tAatnthabeseonrpertgioynrgeesoomluetitoryn nras.o ability patterns in this source). For simplicity we therefore only x of the XIS CCDs the bound–bound absorption lines required to fo simultaneously fitted OBSID 702047010 and OBSID 705031010 constraintheoutflowvelocitiesofindividualsoftX-rayabsorption rdjo inTNhGeCχ1236m5i,naimndizfiatttieodnOteBcShnIDiqu7e05i0s3u1s0e2d0tsherpoaurgahteoluyt. this work, componentsareunresolvedandallabsorptionzonesaretherefore urna fittedasstationaryinthesourcerest-frame(i.e.fixedoutflowve- ls w((cid:7)itχh2al=ls2ta.7ti1stifcoarloenrreoprsarqaumoetetedrtoofthinete9r0epste)r.cWenhtecreontfihedesntacteislteivceall sloocfittiXes-roayfvaobusto=rbe0rktmo vsa−r1y).aAlwllaoywsinhgasthaenoeugtlfligoiwbleveelfofceicttyoonftthhee a.org/ significanceofcomponentsisquotedintermsofa(cid:7)χ2 valuethe reportedFeKabsorptionlineparameters. t N A componentinquestionhasbeenremovedfromthemodelandthe S A datarefittedtoensurethattheorderinwhichcomponentsareadded G tothemodelhasnoinfluenceonthederivedstatistics.Whenrefer- 4.2.2 WarmabsorptionII:partialcovering od d orifntghteofistt,awtishtiilcealacnheagnagteivseto(cid:7)aχfi2tianpdoicsaittievsea(cid:7)sχta2tidsetincoatleismapwroovresmeneinntg. Wealsoconsiderthepossibilitythatthesightlinetoasourceispar- ard S Positiveoutflowvelocitiescorrespondtoanetblue-shiftrelativeto tiallycovered.Inthisabsorptiongeometryafractionfcov<1ofthe pac thesystematicofthehostgalaxy,whileanegativevelocityindicates sourcefluxisabsorbedwiththeremaining1−fcovleakingthrough e F anetred-shift. theabsorptionlayer.Forsimplicityweaccountforpartial-covering ligh absorptionlayersusingacustomizedversionofthezxipcfmodel9 t C whichmodelsthepartial-coveringabsorptionbypartiallyionized tr o 4.2 Modelconstruction gas(seeReevesetal.2008a)withoutneedingtousecomplicated n A All spectra were first fitted with a power law modified solely nraemsteetderpsoiwneorulrawcusstaonmdizxesdtatarblteabmleosd.eAlsarienczoxluimpcnfdethnesitfyre(eNpa)-, pril 5 by Galactic absorption. Additional components were added to ionization parameter (logξ), covering fraction (f ) and red-shHift , 20 the model and retained provided their significance exceeded cov 1 the>99percentconfidencelevelbytheF-test.Anyemissionlines relativetotheobserver(z).Themodelisbasedonthesametables 3 inthesoftX-raybandwerefittedwithnarrowGaussians(σ∼5eV). asdiscussedinSection4.2.1andthereforehasaslightlylowertur- bulentvelocitythanzxipcf,atv =100kms−1,butcoversthe Insourceswithmultiplespectraweinitiallyfoundabroad-bandfit turb same parameter space in terms of column density and ionization to the observation with the highest flux; later observations were parameter. thenaddedsequentiallyandthecontinuum/absorptionparameters Partial-covering absorption can have a strong effect on the ob- wereallowedtotovaryindependentlyuntilasimultaneousfittoall servedcontinuumwithmoderatecolumndensitiesofmaterial(N ∼ spectrahadbeenachievedusingasfewfreeparametersaspossible. H 1023 cm−2)addingconsiderablespectralcurvatureatE<10keV Once a statistically acceptable fit to the broad-band spectrum of (Reevesetal.2004;Risalitietal.2005;Braitoetal.2011;Turner eachsourcehadbeenfound,wefittedanynecessaryemissioncom- ponentsandthenperformedasystematicandmethodicalsearchfor FeKabsorptionlinesbetween5and10keVusingenergy–intensity 8Absorptiongridsaredescribedbyanilluminatingphotonindexof(cid:8) = planecontourplots(Section4.3)anddetailedMonteCarlosimula- 2.1,agasdensityofn=1010cm−3,amicro-turbulentvelocityofvturb= tions(Section4.4). 100kms−1andanintegratedmodelluminosityofL=1044ergs−1between Themodellingapproachforindividualspectralcomponentsare 1and1000Rydbergs. outlined below, with the continuum parameters for each source 9Thestandardzxipcfusesaspecificgridwiththefollowingparameters: being noted in Table D1 and those for the warm absorber being (cid:8)=2.2,n=1010cm−3,vturb=200kms−1,L=1044ergs−1. 6 J.Goffordetal. etal.2011).Bound-freetransitionsinsimilarcolumndensityma- 4.2.4 Lowlyionizedreflection terialcanalsofitbroadresidualemissionprofilesintheFeKband Coldreflectionfromlargecolumndensitiesofneutralorlowlyion- (Inoue&Matsumoto2003;Miller,Turner&Reeves2008,2009; izedmaterialoutsideofthesightlinecanhaveastronginfluence Tatumetal.2012),andpartialcoveringbyCompton-thickmaterial (N (cid:2)1024cm−2)canreducetheobservedemissionbelow10keV ontheobservedX-rayspectrum.Thestrongestobservationalchar- H acteristicsofsuchreflectionincludetheCompton-reflectionhump withthetrueintrinsiccontinuumonlybecomingapparentathigher at∼30−40keVandthealmostubiquitousFeKα andFeKβ flu- energies as a ‘hard excess’ of emission relative to that expected orescencelinesat∼6.4and∼7.06keV(Nandraetal.2007;Shu, fromstandardreflectionmodels(Reevesetal.2009;Risalitietal. Yaqoob&Wang2010),respectively.Comptondown-scatteringof 2009;Turneretal.2009;Tatumetal.2013). both Kα line photons and high energy continuum emission also We include partially-covering absorbers in our models if and gives rise to a ‘Compton shoulder’ at ∼6.2 keV (e.g. Matt 2002; whentheyarerequiredbythedataattheP ≥99percentconfidence F Yaqoob & Murphy 2011) and resonant line emission in the soft levelandasatisfactoryfittothedatacouldnotbeachievedusing X-ray band (e.g. Ross & Fabian 2005; Garc´ıa & Kallman 2010) solely fully-covering absorption; several sources appear to need whichcanfurthercomplicatetheemergentspectrum. multiple partially-covering absorption zones which suggests the Naturally, owing to the important effect it can have on the presence of a clumpy stratified absorber along the line of sight. observed X-ray spectrum there are numerous models available Again, all absorber parameters are listed in Tables D2 and D3 in for modelling the reflection component (e.g. pexrav/pexriv, D AinpspoemndeicxirDcuminsttahneceosntlhineenesuedppoofratinhgighincfoorlummatniodne.nWsiteypnaortteiatlhlya-t Mreafgldizoianrxz, R&osZsd&ziaFrasbkiian1929050;5;pxeixlmlovne,r,N10anGdarrac´ıaet&alK.al2l0m0a7n; ownloa coveringcomponentcanbecontingentonthemeanswithwhichthe d 2010;MYTorus,11Murphy&Yaqoob2009).Inthisworkweusea ed udnegdeenrleyriancgierse.flHecotwioenvecro,maspowneendtisicsumssodinelAlepdpleenaddiixngCt,orseogmaredlmesosdoefl combinationofreflionxandpexrav,bothofwhicharepublicly from whetherthehardX-raydataarefittedwithreflectionornot,partial afovraitlhaibsleitatnhdatebxetecnasuisveelryeufsleidoinnxthinetleirtperoaltauteres.tThheeopbrsiemrvaerydrreeaflseocn- http covering has little measurable effect on the parameters measured tionspectrumfromapre-generatedgridoftablemodelsitissignifi- ://m foranyhighlyionizedabsorptionlinesystems. cantlyfasteratfittingspectrathan,forexample,pexrav,pexrivor nra s pexmon,whichanalyticallycalculatetheComptonreflectionspec- .o x trumontheflyateachstepofthefittingprocess,orMYToruswhich fo rd 4.2.3 Thesoft-excess requiresthemodeltobetailoredforeachindividualsource.Using jo u reflionxasourprimarymeansoffittingthereflectionspectrum rn Relative to the low energy extrapolation of the power-law con- a ensurestheleasttime-consumingMonteCarlosimulationswhich ls tinuum in the 2–10 keV band the X-ray spectrum of AGN often .o &shoPwosunadssm19o8o8th;Pinocrrqeuaesteeitnale.m2i0tt0e4d).flTuhxebtehleorwma∼l1temkepVera(tTuurrenoefr ittshinaiutmucpmoonrwstiadintehtroewudthietnhnethdsiesiamwliunolgrtakwn.eSiotehucsoancldoalrnygs,etwrashianemtnpofilfetttiohnfegoFrbeejfleKecctαstilosiunncechothnaes- at NArg/ this‘soft-excess’suggeststhatitisunlikelytobethedirectemis- reflection fraction, R, reported by pexrav (and pexriv) can be- SA s(Sioonbofrloemwsakast&andDaorndea2c0cr0e7t;ioDnodniescetwailt.h2o0u1t2a)d.dAitlitoenrnalatrievperoexcpeslasninag- comedegeneratewiththephotonindexoftheprimarypowerlaw, God withahardeningreflectioncomponentcompensatedforbyasofter d tionspositthatthesoft-excessmaybeduetoanincreaseinoptical primarycontinuum.Bysimultaneouslyfittingthereflectioncontin- ard dsietpiothnsaasstoEci(cid:3)ate0d.7wkiethVcwirhcuicmhncuacnleeanrhOanVcIeI–eVitIIhIearntdheFetraLn-ssmheiltltetrdaonr- uum,theFeKαlineandthesoftX-rayresonancelines,reflionx Spac reflectedfluxalongthesightlinethroughsmearedabsorption(e.g. isabletoovercomethesemodellingdegeneracieswhichleadstoa e F CDhoenvea2ll0ie0r7)eotrabl.lu2r0re0d6;reDfloecnteio&n(Ne.agy.aCkrsuhminm2y0e0t7a;l.S2o0b0o6l;ewBrseknan&e- ctoonthfiedoenbtsecrovnesdtrsapinetctornumth.erceofnltriiobnutxioanlsoofhthasetrheefleacdtdioitniocnoanltainduvuanm- light C manetal.2011;Nardinietal.2011;Nardini,Fabian&Walton2012) tageofallowingtheionizationstateofthereflector,ξ,tobeafree tr o effects.Furthermore,insomesources(e.g.Mrk766,Milleretal. parameterwhichenablesittomodelchangesintheFeKαemitted n A 2007;MCG-6-30-15,Milleretal.2008)the‘excess’couldsimply flizuaxtioanndstFaeteK.W-sehesltlreesdsg,ehporwoefivleera,stshoactieaqteudivwaliethntthreesruelfltseacrtoerfsoiuonnd- pril 5 bopeaacpitryodaruocutnodfc∼o1m–p2lekxeVab.sorptionandjustbethemanifestationof forthedetectedFeKabsorbersifpexmonisusedinstead,butwith , 201 theresultantMonteCarlosimulationstakingsignificantlylongerto 3 Regardless of the true physical origin we take a purely phe- completewhicheffectivelyprohibitsitsuniformusethroughoutthe nomenologicalapproachwhenfittingthesoft-excessinthiswork, sample. andpredominantlyusethebbodymodelwhichrepresentstheemis- Weinitiallyfittedallsourceswithpexravtodeterminethepa- sionfromaconstanttemperatureblackbody.Whilenotnecessarily rameters of the Fe Kα and Fe Kβ lines. pexrav was then re- physicallymotivated,modellingthesoft-excessinthismannerof- placed with reflionx, and a systematic search for additional fersasimpleparametrizationofany‘excess’softX-rayemission atomic features in the Fe K band was conducted. A total of 11 whichissufficienttogetagoodhandleontheunderlyingcontin- sources have best-fitting reflector Fe abundances which are non- uum parameters. For completeness we investigate the effects that solar, with 4 requiring a slight over-abundance (MCG+8-11-11, otherwaysofmodellingthesoft-excesscanhaveonanyFeK-band NCG4593,NCG7213,NCG7469)and7withanunder-abundance absorptionlinesinAppendixC.Roughlyhalfofthesourcesinthe (4C +74.26, Fairall 9, IC 4329A, IGR J16185−5928, Mrk 335, sample (24/51; ∼47percent) show evidence for a soft-excess, of which 22/24 are fitted with a bbody component. In 3C 120 and Mrk 509 the soft-excess is very broad and extends beyond that 10The xillver reflection model is not currently available in the public whichcanbefittedwithasimpleblackbody.Inthesesourceswe domainandisthereforenotconsideredforuseinthiswork. insteadfittheexcesswithasecondpowerlawwithasofterphoton 11The MYTorus model and documentation are publicly available at: index. www.mytorus.com TheSuzakuviewofhighlyionizedoutflowsinAGN 7 Mrk359,Mrk841).Theseabundancesaremostlikelyabyprod- diskline12 dependingontheasymmetryoftheobservedprofile, uctofreflionxassumingaface-onreflectiongeometry,andisa andasecondintermediatecontourplotwasgeneratedtodetermine caveatwhichisdiscussedingreaterdetailinAppendixC.PDS456 whether any further components were needed by the model. As andAPM08279+5255arenotfittedwithareflectioncomponent before, all other model parameters were allowed to vary freely becausetheylacksufficientcountsintheHXD/PIN. during this process. If there was no evidence for a broad profile wedidnotgenerateanadditionalcontourplotandinsteadmoved directlyontothenextstep; (iv) wherenarrowemissionprofilesweredetectedwitharesolved 4.3 SearchingforFeKabsorption confidencecontourof>99percenttheywerefittedwithunresolved (σ =10eVfixed)Gaussianprofilesprovidedtheywererequired Onceastatisticallyacceptablefittothebroad-bandcontinuumhad bythedataatP >99percent(correspondingto(cid:7)χ2 >9.21for been found, i.e. including all necessary absorption regions, soft F twoparametersofinterest); X-ray emission lines and continuum components, we performed (v) onceallemissionprofileshadbeenaccountedforweagain a thorough search for additional spectral features between 5 and checkedforthepresenceofanyblue-shiftedFeKabsorptionlines 10 keV. The method consists of inspecting the |(cid:7)χ2| deviations atE>6.7keV.Ifnoabsorptionlineswerepresentweendedthe fromthebest-fittingcontinuummodelusinginvertedcontourplots of the energy–intensity plane in the Fe K band. The method of search at this step, reported the best-fitting continuum and Fe K D emissionlineparametersintherelevanttablesofAppendixD,and ow calculatingthecontourplotswasadaptedfromthemethodoutlined n inT10A,andwascarriedoutasfollows: movedontothenextobservationinthesample; loa (vi) otherwise, where absorption troughs were clearly detected de (i) anunresolved(σ =10eV)Gaussianwassteppedacrossthe wliniteh(s)aucsoinngfidinenvecretecdonGtaouusrsioafns>w9i9thpeσr-cwenidtthwieniptiaarlalymefitxriezdedattehie- d from entire5–10keVenergybandofthebaselinecontinuummodelin 25eVintervals,withnormalizationallowedtoadoptbothpositive ther 10, 30 or 100 eV depending on which provided the greatest http andnegativevaluestoprobeforspectrallinesinbothemissionand improvementto(cid:7)χ2.Notethatwhileinitiallyfixed,thelinewidths ://m absorption.Alloftheotherspectralcomponents wereallowed to wereallowedtovarywhereappropriateleadingtofoursourceswith nra befree; resolvedabsorptionprofiles.Thekeyparametersofalldetectedab- s.o (ii) after each step the (cid:7)χ2 deviation was recorded generating sorptionlinesarereportedinTable2. xfo a χ2 distribution of the entire Fe K band relative to the baseline Thisprocesswascarriedoutoneachindividuallyfittedspectrum rdjo u continuummodel; in the sample, including those which were included in simulta- rn cor(diiiin)gctoon(cid:7)fidχe2ncdeevcioantitoonusrsoffo−rt2h.e3g,−rid4.o6f1χ,2−v9a.l2u1e,s−w1e3re.8p2l,o−tte1d8.a4c2- napeopulisedfittos.AInrkF1ig2s03(OaBnSdID4 7w0e20s1h4o0w10e)x,awmhpilcehssohfowthsisevpirdoecnecses als.org from the baseline model, which correspond to confidence in- for a broad asymmetric emission profile and no absorption lines, at N/ tervals of 68percent, 90percent, 99percent, 99.9percent and andMrk766(OBSID701035020),whichisdominatedbyFeXXV A 9to9.a99(cid:7)pχer2c=ent+, r0e.s5pewcotirvseelyfi.tAiscaolnsofidpelnoctteedcownhtoicuhr cisorirnetsepnodneddintgo sHheoαwatnhderFeesidXuXaVlIsLwyhαicahbrseomrpatiinoni.nTthheeFtoepKpbaannedlswohfebnoathllfiatgoumreics SA Go denoteanapproximatelevelforthecontinuum. lines have been removed from the best-fitting continuum model dda Theenergy–intensitycontourplotsproducedwiththismethodpro- athnedptrheesernecfleecotfiotnheconmeuptroanleFnet iKsαfi/tFteedKwβitflhupoerexsrcaenvcetolihniegsh.liTghhet rd Spa vsoidrpetiaopnocwoemrfpuolnmenetasnpsroefsesnetarinchtihnegFfoerKadbdaintidonwahlielemaislssoionviosuraalbly- cwohnetnoutrhepFloetsKsαhoawndtFheeKsiβgnliifinceasnhcaevseobfeethnefitrteemdawiniitnhgrerefslidiuoanlxs ce Flig assessing their energy, intensity and rough statistical requirement h and a narrow Gaussian line. The continuous outer contour cor- t C relTathiveeretoarteheaunnudmebrleyrinogfactoonmtiincufuematumreosdeble.tween6.0and8.0keV responds to the (cid:7)χ2 = +0.5 residual as mentioned previously. tr on Fromoutertoinnertheclosedsignificancecontourscorresponding A wfehatiuchrecsainnccloumdepliiocnaitzeetdheemdeitsesciotinonlinoefsabfrsoomrptFioenXXliVneansydsFteemXsX.VSIuecxh- 9to0(cid:7)peχrc2einmt)p,r−ov9e.2m1e(nPts=of9−92p.e3rc(PenFt)=,−6183p.8e2rc(ePnt)=, −994..96p1e(rPcFen=t) pril 5 epdegcteecdoamtp∼le6x.6a3t−∼67..71kaenVd,∼a6n.d97ankyebVr,oraedspFeecltiinveelpyr,otfihleedFueeKto-sGheelnl- tainnduu−m18m.4o2de(lPaFre=s9h9oF.w99npienrrceedn,t)grreeleant,ivbelutoe,thcyeFabnesatn-dfittminaggecnotna-, , 2013 eralRelativity(GR)orscatteringeffects.Itisimportanttoaccount respectively. fortheseemissionresiduapriortosearchingforabsorptionlines, Thereareclearresidualprofilespresentinbothsources,particu- particularly in sources which show evidence for a broad residual larlyArk120wherethepositiveasymmetricprofileandadditional asthebroadnessoftheprofilecaneffectivelymaskthepresenceof FeXXVIemissionareapparentinthemiddleandbottompanelsof low-velocityabsorptionsystemsintherawdata.Foragivencontin- Fig.3,respectively.ThetwoabsorptionprofilesinMrk766–which uummodelwesearchedforhighlyionizedabsorptionlinesinthe have previously been identified as Fe XXV Heα and Fe XXVI Lyα, FeKbandusingthefollowingsteps: respectively,byMilleretal.(2007)–arestatisticallydistinguish- able at >99.99percent confidence and are self-consistently fitted (i) wefirstgeneratedanenergy–intensitycontourplotusingthe withasinglehighlyionizedregionofphotoionizedabsorption(see methodoutlinedabove; Section4.6).Energy–intensitycontourplotsforallfittedspectrain (ii) wetheninspectedthecontourplottodeterminewhetherthere wereanyintenseionizedemissionandabsorptionlinespresentin thedatawithconfidencecontoursof>99percent; 12Forsimplicity,thedisklinewasfittedwithanassumedrest-frameenergy (iii) wheretherewasevidenceforabroademissionresidualatan ofE=6.4keVandatypicalemissivityprofileofq=−2.5(e.g.Patrick F-testsignificanceofatleastPF>99percenttheywerefittedwith etal.2011a).Theinneremissionradius(Rin),outerradius(Rout)anddisc eitherabroadenedGaussian(withσ-widthafreeparameter)ora inclination(i)wereleftfreetovary. 8 J.Goffordetal. Table2. GaussianparametersfordetectedFeKbandabsorptionlines.Notes:(1)sourcename;(2)SuzakuobservationID;(3)absorptionline identification;(4)measuredlineenergyinthesourcerest-frame,inunitsofkeV;(5)measuredlineenergywidth,inunitsofeV.Unresolvedlines werefittedwithwidthsfixedat10,30or100eVdependinguponwhichyieldedthebeststatisticalimprovementtothefit;(6)lineintensity,in unitsof×10−6photonscm−2s−1;(7)lineEW,inunitsofeV;(8)changeinfitstatistic(anddegreesoffreedom)whenthelineisremovedfrom thebest-fittingmodel;(9)linesignificanceaccordingtotheF-test,inpercent;(10)MonteCarlosignificanceofindividuallines,inpercent; (11)MonteCarlosignificanceofobservingapairoflineswithenergyseparationcorrespondingtoaFeXXV–XXVIpair(seethetextfordetails), inpercent.AbsorptioncomplexesdetectedatPMC≥99percentconfidencearelistedinbold. Source OBSID LineID Energy σ-width Intensity EW (cid:7)χ2/(cid:7)ν PF PM1C PM2C (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) 3C111 703034010 Lyα 7.24+−00..0044 10f −5.6+−22..11 −26+−99 16.4/2 >99.99 99.8 705040020 Lyα 7.76+−00..0074 10f −15.2+−55..11 −20+−1100 11.1/2 99.60 97.8 3C390.3 701060010 Lyα 8.08+−00..0065 10f −5.6+−22..66 −21+−1100 12.1/2 >99.99 98.7 4+74.26 702057010 Lyα 8.38+−00..0057 10f −7.2+−33..44 −29+−1144 12.4/2 99.80 96.6 APM08279+5255 stacked Heα–Lyα(1) 7.76+−00..1111 100f −6.9+−22..53 −24+−88 23.8/2 >99.99 99.7 Heα–Lyα(2) 10.97+−00..1166 100f −4.4+−11..98 −30+−1155 18.4/2 >99.99 97.3 D CBS126 705042010 Lyα 7.04+−00..0045 30f −3.3+−11..00 −77+−2222 27.5/2 >99.99 >99.9 ow EMMMSRrCOkG22127-056931--3-G1070-18355 s777t000ac434k000e335d115[000a111ll000] FeHLLLLyyy–eααααHeα 78767.....0827558625+−+−+−+−+−0000000000..........00000011005521222266 110911143000+−+−fff65861542 −−−−−1142752.....79815+−+−+−+−+−4444111111..........8111222200 −−−−−4371102384+−+−+−+−+−1111333333440000 3822151565.....06791/////32232 >>>>9999999999.....9999959999 >>>9999999999.....99961 >99.9 httpnloaded from Mrk766 701035010 Heα 7.11+−00..0077 10f −2.8+−11..66 −27+−1155 7.0/2 96.27 71.0 99.8 ://m Lyα 7.42+−00..0066 10f −3.9+−11..66 −40+−1177 15.3/2 99.92 98.4 nra 701035020 Heα 6.80+−00..0022 10f −7.7+−11..88 −60+−1144 50.0/2 >99.99 >99.9 >99.9 s.ox NGC1365 702047010 HLyeαα 76..1727+−+−0000....00004411 6150+−f88 −−295..90+−+−1111....6699 −−14546+−+−117788 111283..89//23 >>9999..9999 >9999..89 >99.9 fordjou Lyα 6.97+−00..0033 65(cid:4) −8.7+−11..77 −69+−1133 77.0/3 >99.99 >99.9 rna Heβ 7.94+−00..0011 65(cid:4) −26.1+−11..55 −164+−1100 964.5/3 >99.99 >99.9 ls.o 704031010 HLyeβα 86..3781+−+−0000....00003322 3605(cid:4)f −−1106..61+−+−1122....8844 −−8677+−+−11115500 19344..39//32 >>9999..9999 >>9999..99 >99.9 at Nrg/ Lyα 7.00+−00..0022 30(cid:4) −13.4+−22..00 −85+−1144 90.7/2 >99.99 >99.9 AS Heβ 7.87+−00..0088 30(cid:4) −4.7+−22..55 −34+−2200 7.2/2 96.77 – A G NGC3227 703022010 HLyeβα 86..3679+−+−0000....11001144 1300(cid:4)f −−59..39+−+−2233....7777 −−4251+−+−227744 187..55//22 9998..9293 >99.–9 >99.9 oddard Lyα 6.95+−00..0033 10f −15.5+−33..77 −38+−99 42.4/2 >99.99 >99.9 Sp NNNGGGCCC334570185631 777177000000343010000000262330232134000000315111000000 HHHHHHLLLyyyeeeeeeααααααααα 777666666.........000777767054616497+−+−+−+−+−+−+−+−+−000000000000000000..................000000000000000000345577332222442244 111111111000000000fffffffff −−−−−−−−−111104865962.........000568102+−+−+−+−+−+−+−+−+−223333333322111111..................882011556688667700 −−−−−−−−−222142212214915060+−+−+−+−+−+−+−+−+−1111999899556656351100 1146635996215606.........202672739/////////222222232 >>>>>999999999999999989.........999999763499999449 >>>>>999999998999999619.........999998284 >>>>99999999....9999 ace Flight Ctr on April 5, 2013 Lyα 7.08+−00..0033 10f −3.3+−11..11 −38+−1133 22.7/2 >99.99 >99.9 703023010 Heα 6.82+−00..0033 10f −4.8+−11..33 −22+−66 26.4/2 >99.99 >99.9 >99.9 Lyα 7.11+−00..0022 10f −6.4+−11..33 −33+−77 57.5/2 >99.99 >99.9 NGC4151 701034010 Lyα 7.17+−00..0044 92+−5411 −15.9+−44..73 −28+−88 43.7/3 >99.99 >99.9 NGC4395 702001010 Heα 6.63+−00..0075 10f −2.2+−11..11 −35+−1199 10.4/2 >99.99 74.5 >99.9 Lyα 6.91+−00..0066 10f −2.9+1.11.1 −55+−2200 18.3/2 >99.99 86.8 NGC5506 stacked[all] Lyα 9.23+−00..0066 10f −11.1+−44..77 −16+−55 16.2/2 >99.99 99.8 PDS456 701056010 Heα–Lyα(1) 9.07+−00..0066 100f −2.5+−00..88 −108+−3355 29.5/2 >99.99 >99.9 Heα–Lyα(2) 9.57+−00..0099 100f −2.0+−00..88 −99+−4411 15.7/2 >99.99 99.8 705041010 Heα–Lyα(1) 8.58+−00..0099 100f −2.8+−11..00 −118+−4466 19.5/2 >99.99 96.2 Heα–Lyα(2) 9.03+−00..0099 100f −3.5+−11..11 −169+−5511 78.6/2 >99.99 >99.9 SJ2127.4+5654 702122010 Lyα 9.04+−00..0055 10f −10.0+−44..77 −60+−2288 11.9/2 99.78 98.9 findicatesthataparameterwasfrozenduringspectralfitting. (cid:4)indicatesthataparameterwastiedduringspectralfitting. TheSuzakuviewofhighlyionizedoutflowsinAGN 9 D o w n lo a d e d fro m h ttp ://m n ra s .o x fo Figure4. Toppanel:asinthecaseofFig.3,butthistimeforMrk766 rd jo (OBSID701035020).Bottompanel:asinthecorrespondingpanelofFig.3. u There are no ionized emission lines required in Mrk 766; however, two rna highlysignificant(eachrequiredatPF>99.99percent)absorptionprofiles ls.o aarnedcFleeaXrlXyVdIeLtyeαct,erde.sTpheceteivneelryg,yoouftflthoewsienlginaetsviosucto∼ns6i0st0e0ntkwmitsh−F1e.XXVHeα at Nrg/ A S significances.Indeed,ashasbeenpointedoutbynumerousauthors A G in the literature (e.g. Protassov et al. 2002; Porquet et al. 2004; o Figure 3. Top panel: ratio plot showing the residuals remaining in the d Fe K band of Ark 120 (OBSID 702014010) once all atomic lines have Markowitz,Reeves&Braito2006;T10A)theF-testalonemight dard beenremovedandreflectionfittedwithpexrav.Middlepanel:confidence notbeanadequatestatisticaltesttodeterminethedetectionsignifi- S p contoursshowingdeviationsfromthebest-fittingmodelwhentheFeKα canceofatomiclinesincomplexspectralmodelsasitneithertakes ac e and Fe Kβ lines have been fitted with reflionx and a narrow Gaus- into account the number of energy resolution bins over a given F sian, respectively. The closed significance contours corresponding (from energy range, nor the expected energy of a given atomic line. In lig h outertoinner)to(cid:7)χ2 improvementsof−2.3(PF =68percent),−4.61 caseswherethereisnoapriorijustificationforexpectingaspectral t C (PF=90percent),−9.21(PF=99percent),−13.82(PF=99.9percent) line at a particular energy and the line search is done over what tr o and −18.42 (PF = 99.99percent) relative to the best-fitting continuum isessentiallyanarbitraryenergyrange,asisoftenthecasewhen n A modelareshowninred,green,blue,cyanandmagenta,respectively.The p muaalgiesnstaigcnoifintcoaunrtsainttthheem>i9d9d.l9e9ppaenrecleinndtilceavteelthanatdththeabtraonadaedmdiitsisoinoanlrceosimd-- dsoemalienwghwatiothvesrtprorendgilcyttbhleued-estheicfttieodnapbrosobrapbtiiloitnylainnedsw,thheenFco-tmesptacreadn ril 5, 2 toextensivesimulations.Forthesereasonsallsuspectedabsorption 0 ponentisrequiredinthemodel.Bottompanel:theremainingconfidence 1 lines which have an F-test significance of P > 99percent were 3 contoursoncethebroadprofilehasbeenfitted.Thebroadresiduaareno F longerdetectedbutanadditionalFeXXVIemissionlineispresentatE∼ followedupwithextensiveMonteCarlosimulationswhichallows 6.97keVwhichissignificantat>99percent.Inallpanelsthedashedverti- theirdetectionsignificancetobeassessedagainstrandomfluctua- callinesindicatetheexpectedrest-frameenergiesof(fromlefttoright)the tionsandphotonnoiseinthespectrum. FeKαfluorescenceline,FeXXVHeαandFeXXVILyα,respectively. 4.4 MonteCarlosimulations thissampleareincludedinFigsB1andB2inAppendixB,inthe onlinesupportinginformation.Wenotethatthe(cid:7)χ2ofindividual Such Monte Carlo simulations have been used extensively in the linesandtheircorrespondingsignificancesaccordingtotheF-test literature to overcome the limitations of the F-test (e.g. Porquet as listed in Table 2 are taken from a direct spectral fitting of the etal.2004;Markowitzetal.2006;Miniutti&Fabian2006;T10A; lineprofilesthemselves,andarenotdetermineddirectlyfromthe T10B)andenablethestatisticalsignificanceofaspectrallinetobe contourplots. robustly determined independently of spectral noise and detector It is also important to note that while we use the F-test as a effects.ThemethodofMonteCarlosimulationweusedfollowsthe roughinitialgaugerofthesignificanceofanyline-likeprofilesin sameprocesswhichwasfirstoutlinedbyPorquetetal.(2004),and the Fe K band we do not claim to robustly detect any absorption isalmostidenticaltothatusedbyT10A.Theprocesswascarried linesbasedsolelyontheirmeasured(cid:7)χ2andcorrespondingF-test outwiththefollowingsteps. 10 J.Goffordetal. (i) From the null hypothesis model (i.e. the broad-band con- minimizedpriortosearchingforabsorptionlines,suchthatthere- tinuum model with all narrow absorption lines in the Fe K band ducedχ2is≈1.0inallcases.Moreover,andasshowninAppendix removed)wesimulatedbothXIS-FIandXIS-BIspectrausingthe ,wenotethatthevastmajorityofsuspectedabsorptionresiduals fakeitcommandinXSPECandsubtractedtheappropriatebackground aremanifestedthroughdiscretenarrowdipsinthespectrumrelative files.Thesimulatedspectrahadthesameexposureastheoriginal tothebest-fittingmodel,whereasanysystematicresidualswould dataandusedthesamespectralresponsefiles. usuallybebroaderthantheinstrumentresolution. (ii) WethenfittedthesimulatedXIS-FI(XIS-BI)spectrabetween Thesecondcaveatisassociatedwithspectralcomplexityinthe 0.6and10.0keV(0.6–5.0keV)withthenullhypothesismodelto FeKbandwhichcancomplicatebothlineidentificationandspectral produceanewandrefinednullhypothesiswhichtakesintoaccount interpretation.Therecanbesignificantatomiccomplexitybetween theuncertaintyinthenullhypothesismodelitself.Allcontinuum ∼5and7keV,e.g.thenarrowFeKαandFeKβ fluorescencelines, parameters bar the photon index of the primary power law and broad underlying Fe K lines and ionized Fe XXV–XXVI emission itsnormalization,andthenormalizationofanybbodycomponent, lines, and the detection of absorption in this regime can depend were frozen to their best-fitting parameters taken from the real stronglyonhowthesefeaturesaremodelled.Therearethreesources data to prevent degeneracies between model components during (4C+74.26,MGC-6-30-15,SWIFTJ2127.4+5654;seeFig.B2)in re-fitting.AnybroadprofileatFeKhaditswidthfixedtothebest- thesampleinwhichbothabroadunderlyingprofileandatleastone fittingvalue but was allowed tovary in bothcentroid energy and absorptionlinehavebeendetected.Inboth4C+74.26andSWIFT D normalization. J2127.4+5654theabsorptionlinesaredetectedathighenergies(i.e. ow (iii) Fromtherefinednullhypothesismodelwegeneratedasec- E>8keVinthesourcerest-frame)andtheeffectofthebroadFe nlo a ondsetofsimulatedXIS-FIandXIS-BIspectraandsubtractedthe lineonthelinedetectionisnegligible.InMCG-6-30-15thelines d e d athpepnrofiptrtieadtewbiatchktghreounnudllfihleysp.oTthheessiessmecoodnedlsainmdultahteedresspuelctatrnatwχe2re hetavale.2a0ls0o5;bMeeinllecroentfiarlm.2ed00w9)itwhhoitchhersuXgg-reasytsotbhsaetrtvhaetoprrieessen(Yceouonfga from val(uive)wAansurencroersdoelvde.d(σ=10eV)Gaussianwasaddedtothemodnuelll bTrhouasdmFeodlienlelindgoeosfnaobtrinotardodeumciessainoynsliignneifidcoaenstnmotodapelpseyasrtteomaaftfieccst. http://m at5keVinthesourcerest-framewithintensityinitiallysettozero thedetectionsofFeKabsorptionlinesinthesecases. n butleftfreetovary between bothpositiveand negative values to ras .o probe for both emission and absorption lines. The Gaussian line x 4.5 Consistencychecks fo wasthensequentiallysteppedbetween5and9.5keV(rest-frame) rd in 25 eV increments. After each step the |(cid:7)χ2 | was recorded To further test the robustness of the absorption lines detected in jou relativetoχ2 . noise theco-addedXIS-FIspectraweperformedaseriesofconsistency rna (v) ThispnruollcesswascarriedoutforT=1000simulatedspectra checks with the individual XIS detectors. If a line is detected in ls.o per observation yielding a distribution of |(cid:7)χn2oise| under the null two(ormore,iftheXIS2isalsopresent)detectors,thelineisvery arg/ hypothesiswhichessentiallymapsthestatisticalsignificanceofany likelytobearealfeatureintrinsictothesourcespectraratherthanan t N A deviationsfromthenullhypothesismodelwhicharesolelydueto artefactofbackgroundsubtraction,systematicnoiseoranassociated S A randomphotonnoiseinthespectrum. detectoreffect.TheXIS-BIisnotsuitableforuseasaconsistency G (vi) Themeasuredsignificanceofthelineintherealspectrum check as it tends to have a much lower S/N above around 5 keV od d |(cid:7)χl2ine|wasthencomparedtothevaluesinthe|(cid:7)χn2oise|distribution owing to its lower effective area and higher background. Where ard toassesshowmanysimulatedspectrahadarandomfluctuationwith absorption has been detected at the P ≥ 95percent level we S MC p a detection significance over this threshold value. If N simulated fittedeachindividualbackground-subtractedXIS0,XIS2(where ac e spectrahave|(cid:7)χ2 |≥|(cid:7)χ2 |,thentheestimatedconfidencein- present)andXIS3spectrumwiththebest-fittingmodeltotheco- F tervalforthe(cid:2)obsneor(cid:3)isveedlineflrionemMonteCarlosimulationsisthen addedXIS-FIspectrum.WethenfittedaGaussianabsorptionline, ligh PLine=1− NLine .Moreover,iftherearetwoabsorptionprofiles with rest-frame energy and normalization left free to vary, at the t C coMnCsistentwithaTFeXXV–XXVIpair,wecaninferthenullprobability energywheretheabsorptionlineisdetectedintheXIS-FIspectrum, tr on ofbothlinessimultaneouslybeingafalsedetectionbymultiplying refittedthedataandnotedtheresultantlineparametersforeachXIS A p theprobabilitiesofeachindividualline. detector. We note that in SWIFT J2127.4+5654 this consistency ril 5 check could not be conducted as only the XIS 3 spectrum was , 2 0 MonteCarlosignificancesforallabsorptionlinesdetectedinthis available. 1 3 workarelistedincolumn10ofTable2.Absorptioncomplexeswith InallsourcestheresidualsdetectedintheindividualXISspectra atotalP ≥99percentareconservativelyidentifiedasrobustly haveGaussianparametersconsistentwiththosefoundfortheco- MC detected,whilethosewith95percent≤P <99percentareonly addedFIspectrum.InFig.5weshowacomparisonofabsorption MC listedasmarginaldetections.Theoveralldetectionrateandglobal line equivalent width (EW), along with the 90percent error bars, significanceofFeK-bandabsorptionlinesarefurtherdiscussedin as measured independently with the XIS 0 and XIS 3 detectors. Section5.1. ThedasheddiagonallinerepresentsthepositionwheretheEWis TherearetwopossiblecaveatsassociatedwiththeaboveMonte equal in each of the detectors; where the EW is consistent at the Carloprocesswhichbothwarrantfurtherdiscussion.Firstandfore- 90percent level the error bars will overlap this line. In this plot most,theMonteCarlosimulationsimplicitlyassumethatthenull- the Fe XXV Heα, Fe XXVI Lyα and blended Fe XXV–XXVI lines are hypothesismodelisthecorrectrepresentationofthecontinuumin plottedseparately,eveninthecaseswherethelinesarefoundtobe agivensource,andthereforetheMonteCarloprobabilitiesdonot part of a Fe XXV pair. We find that the mean EWs measured with accountforthepossibilityofcontinuummis-modelling.Giventhis theXIS0andXIS3detectorsareconsistentatthe90percentlevel possibilityitisimportanttonotethatwehaveattemptedtoachievea (i.e.theerrorbarscrossthediagonallineinFig.5)inallsources. statisticallyacceptablerepresentationofthebroad-bandspectrum, WherepossiblethisanalysiswasalsocarriedoutusingtheXIS2 sothatnoobviousbroadresidualsarepresent.Carehasbeentaken spectraversusbothXIS0andXIS3.Again,inallcasesthemean in the Fe K band in particular such that any broad residuals are parameters of the lines were always consistent at the 90percent

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.