Table Of Content

DRAFTVERSIONJANUARY16,2014 PreprinttypesetusingLATEXstyleemulateapjv.04/17/13 AREX-RAYEMITTINGCORONAEAROUNDSUPERMASSIVEBLACKHOLESOUTFLOWING? TENGLIU(刘腾),JUN-XIANWANG(王俊贤),HUANYANG(杨欢),FEI-FANZHU(朱飞凡),YOU-YUANZHOU(周又元) CASKeyLaboratoryforResearchinGalaxiesandCosmology,DepartmentofAstronomy,UniversityofScienceandTechnologyofChina,Hefei,Anhui 230026,China;[email protected],[email protected] DraftversionJanuary16,2014 ABSTRACT 4 Hard X-ray emission in radio-quiet active galactic nuclei (AGNs) is believed to be produced via inverse 1 Comptonscatteringby hotand compactcoronaenear the supermassive blackhole. Howeverthe originand 0 physicalpropertiesof the coronae, includinggeometry,kinematicsand dynamics, yet remainpoorlyknown. 2 Inthiswork,taking[OIV]25.89µmemissionlineasanisotropicindicatorofAGN’sintrinsicluminosity,we comparethe intrinsic coronaX-ray emission between Seyfert1 and Compton-thinSeyfert2 galaxies, which n areviewedatdifferentinclinationsaccordingtotheunificationscheme.Wecompileasampleof130Compton- a J thinSeyfertgalaxieswithboth[OIV]25.89µmlineluminositiesmeasuredwithSpitzer-IRSandX-rayspectra observed by XMM-Newton, Chandra, Suzaku or Swift. Known radio-loud sources are excluded. We fit the 5 X-ray spectra to obtain the absorption-corrected2 – 10 keV continuumluminosities. We find that Seyfert1 1 galaxiesareintrinsicallybrighterinintrinsic2–10keVemissionbyafactorof2.8+−00..54(2.2+−00..93inSwift-BAT ] 14– 195keVemission), comparingwith Compton-thinSeyfert2galaxies. TheSeyfert1 andCompton-thin E Seyfert 2 galaxies follow a statistically identical correlation between the absorption-corrected 2 – 10 keV H luminosityandthe14–195keVluminosity,indicatingthatourabsorptioncorrectiontothe 2–10keV flux is sufficient. The difference in X-ray emission between the two populations is thus unlikely due to X-ray . h absorption,andinsteadimpliesanintrinsicanisotropyinthecoronaX-rayemission. Thisstrikinganisotropy p ofX-rayemissioncanbeexplainedbyabipolaroutflowingcoronawithabulkvelocityof∼0.3−0.5c. This - wouldprovideanaturallinkbetweentheso-calledcoronaeandweakjetsinthesesystems. Otherconsequences o ofoutflowingcoronaearealsodiscussed. r t Subjectheadings: galaxies: active–galaxies: nuclei–galaxies: Seyfert–galaxies: jets–X-rays: galaxies– s a relativisticprocesses [ 1 1. INTRODUCTION 2000;Molinaetal. 2009). Meanwhile,similartosolarcoro- v nae,theX-raycoronaeinblackholesystemsmightbeheated 4 StronghardX-rayemissionhasbeendetectedinblackhole 6 accreting systems spanning from stellar mass (∼10M⊙) to throughmagnetic processes and could produce magnetic in- 5 super massive scales (∼105−1010M ). Such systems are flow/outflow above the disk (Beloborodov1999). Bulk mo- ⊙ tionoftheflow,iffastenough,wouldproducebeamedX-ray 3 the dominantpopulationin the X-raysky. Inradio-quietsu- emission. The X-ray emission is also likely producedin the . per massive black hole (SMBH) accretion systems, the pri- 1 jetbase, whichisphysicallysimilarto an outflowingcorona mary hard X-ray emission is power-law shaped in spectrum 0 (Markoffetal. 2005). Lightbendingeffectoftheblackhole 4 with a high energycutoffat &100 keV (Molinaetal. 2013; could also produce anisotropy in X-ray emission, by reduc- 1 Riccietal.2011). ItisnowwidelyacceptedthatX-rayemis- ingtheX-rayfluxescapingupward(Miniutti&Fabian2004; : sioninradio-quietAGNscomesfrominverseComptonscat- v Chenetal. 2013). These arguments suggest that the corona tering of the seed photons from the accretion disks through i X-rayemissioninAGNsislikelyanisotropic,althoughweak X hotand compactplasma, named corona(Haardt&Maraschi if there is any as no observationalevidence has yet been re- 1991, 1993). Such coronae are also required in radio-loud r ported. However, measuring the level of the anisotropy or a AGNs, in which strong relativistic jets could produce extra putting strong constraint to it is essential to understand the emission from radio to X-ray and even gamma-ray, and in originandphysicalpropertiesofthecorona. stellar mass black hole accretion systems (e.g. Plotkinetal. In the unification scheme of AGNs, type 1 sources 2012). However,westillknowverypoorabouttheoriginand are viewed face-on while type 2 sources viewed edge-on physicalnatureofthecorona. (Antonucci 1993), providing an opportunity to measure the Being free from dust extinction, hard X-ray emission has anisotropy of the corona X-ray emission. In this work we been widely used to represent the intrinsic power of AGNs, presentanobservationalinvestigationtocomparethecorona e.g.,tomeasuretheratiooftype1totype2AGNsandtode- X-rayemissionintype1andtype2 AGNsatgivenintrinsic rive the AGN bolometric luminosity (Meléndezetal. 2008; blackholeaccretionpower.Todoso,weneedanindependent Rigbyetal. 2009; Burlonetal. 2011; Maliziaetal. 2009; AGNluminosityindicatorandsamplesofAGNswithdiffer- Runnoeetal. 2012). However, we still know little observa- entinclinationangles. tionallytowhatlevelthecoronaX-rayemissionisisotropic. Inspired by recent studies (Meléndezetal. 2008; While a slab-like corona could produce weaker X-ray emis- Rigbyetal. 2009; Diamond-Stanicetal. 2009; Weaveretal. sionathigherinclinationduetoprojectioneffect(e.g.Zhang 2010; LaMassaetal. 2010; Liu&Wang 2010), we opt to 2005; Chenetal. 2013), the suggested patchy structure and use[OIV] 25.89µmline emission, a forbiddenline produced smallopacityofthecoronaindicatetheanisotropyproduced in the so called narrow line region (NLR), as an intrinsic this way is rather weak (Haardtetal. 1994; Zdziarskietal. luminosityproxyofAGNs. Comparingwiththewidelyused [OIII] 5007Å line, [OIV] line is significantly less attenuated ofopticalspectroscopiccatalogsofAGNswithexisting by dust extinction (AV ∼ 39 corresponds to A25.89µm ∼ MBHestimatesand[OIII]5007Ådetectionsareinvolved 0.06-0.18;Goulding&Alexander2009),andhasrelatively in matchingwith allthe IRS targets. The finalsample higherionization potential(54.9eV) thus is less affected by consistsof81AGNswithresolved[SIV],[NeIII],[OIV] contamination from star formation in the host galaxy. The or[NeV]lines. validityofusing[OIV]asintrinsicAGNluminosityindicator has been confirmed observationally (Diamond-Stanicetal. E: Acollection(fromliterature,archive,andtheauthor’sown 2009; Meléndezetal. 2008; Rigbyetal. 2009). Particu- observing programs) of Spitzer-IRS high-resolution larly, Diamond-Stanic et al. (2009) showed that the [OIV] spectra of 426 galaxies, including quasars, Seyferts, luminosity distributions are indistinguishable for obscured LINERs and HII galaxies (Pereira-Santaellaetal. and unobscured AGNs in a well-defined galaxy-magnitude- 2010). limited sample, while [OIII] luminosities are systematically lower in obscured sources. However we note that unlike These samples somehow overlap with each other. High- [OIII], measurements of [OIV] line fluxes are available in resolution data are adopted with higher priority when avail- muchfewerAGNs(see§2). able.High-resolution[OIV]fluxmeasurementsfromdifferent We select to compare corona X-ray emission of type 1 subsamples for overlapped sources are generally consistent AGNs with that of Compton-thin type 2 AGNs. Only se- witheachother. Inafewofthem(7sources,5%ofthefinal curely identified Compton-thin sources (with X-ray obscu- sample),thedifferencesareaslargeasbyafactorof>1.5.We ration column density N < 1024 cm−2) are included, be- H simplyadopttheaverage[OIV]fluxesinlogarithmspacefor cause for Compton-thick ones reliable correction to X-ray theseoverlappedsources. The[OIV]fluxesareconvertedinto obscuration is extremely hard or even impossible. Known luminositiesadoptingthesamecosmologicalparameters1. radio-loud sources are also excluded to avoid contamina- We then match the composite [OIV] sample to XMM- tions from known strong jet emission. 2 – 10 keV intrin- Newton data archive to search for reliable X-ray spectra. sic(i.e.,absorption-correctedcontinuum)luminositiesareob- For a few sources without XMM observations, we supple- tainedthroughspectralfitting. ment the X-ray spectra collection with Chandra, Suzaku or In §2, we describe the samples, and the processing of X- Swift data (see Table 1). We perform literature search and raydata. In§3, wecomparetheintrinsicX-rayemissionbe- independent X-ray spectral fittings to get the classifications tweentype1andtype2subsamplesatgiven[OIV]luminosity, of their X-ray obscuration nature. Possible Compton-thick showingrelativelyweakerintrinsicX-rayemissionintype2 sourcesareexcluded,andonlysecurelyidentifiedCompton- sources. In §4, we present discussions on the robustness of thin ones are included. The criteria we adopted to iden- theanisotropyofcoronaX-rayemission,andfinallyinterpret tify secure Compton-thin sources are: 1) X-ray spectra fit- itintermsofcoronaoutflowing.§5presentsthesummary. ting could provide reliable measurements to X-ray obscura- tionsat N <1024 cm−2, if with highqualityspectra; 2)for 2. SAMPLEANDDATAREDUCTION H those sources without good-enough X-ray spectra, sources 2.1. [OIV]SampleCompilation with T ratio <1 (T ratio = f2−10keV/f[OIII], where f2−10keV To gather a large sample of AGNs with both istheobservedX-rayflux,and f thedustextinctioncor- [OIII] [OIV] flux measurements and X-ray observations, we rectedlineflux)ornarrowFeKαlineEW>600eVareex- first combine five major samples in literature with cludedascandidatesofCompton-thicksources(Bassanietal. [OIV] 25.89µm line fluxes measured by Spitzer In- 1999). This is a rather conservative approach since some frared Spectrometer (IRS, Diamond-Stanicetal. 2009; known Compton-thin sources also show T ratio <1 or line Weaveretal. 2010; Tommasinetal. 2010; Dasyraetal. EW>600eV (Bassanietal. 1999). We exclude19 sources 2011; Pereira-Santaellaetal. 2010). The five samples are (15ofwhicharetype2)withtoofewX-raycounts(<100in brieflydescribedasfollows: 2–10keV)inXMMdata,forwhichreliablemeasurementsto theX-rayobscurationsarehard.Toavoidcontaminationfrom A: Asubsampleof91SeyfertgalaxieswithSpitzer-IRShigh- strongrelativisticjettotheX-rayemission,knownradio-loud resolutionspectroscopicobservations(Tommasinetal. sourcesareexcluded. InTable1wepresentthesamplewith 2010) from the 12µm flux limited sample of Seyfert referencesto their X-rayidentifications. Thefull sample in- galaxies(116sources,Rushetal.1993). cludingpossibleCompton-thicksourceswillbepresentedin afuturepaper. B: Spitzer-IRS low-resolution spectroscopic observations We also excluded5 X-rayun-obscuredSeyfert2galaxies, of a spectroscopically selected, galaxy-magnitude- i.e. those with X-ray obscuration column densities N < H limited sample of Seyfert galaxies from the revised 1022 cm−2 but without any evidence of broad line regions Shapley-Ames (RSA) catalog (Diamond-Stanicetal. (BLRs) detected, which could be physically different from 2009). The RSA Seyfert sample, containing 18 normal Seyfert 2 galaxies (Brightman&Nandra 2008). In- type 1 and 71 type 2 Seyfert galaxies, is a well cludingthemintoanalyseswouldhoweverslightlystrengthen studied galaxy-magnitude-limited complete sample ourconclusioninthiswork. (Maiolino&Rieke1995;Hoetal.1997). Finally,thecompositesampleincludes84type1AGNs(in- cludingSeyfert1, Seyfert1.2andSeyfert1.5galaxies,here- C: A subsample of 79 AGNs with high-resolution Spitzer- after Sy1s) and 46 Compton-thin type 2 AGNs (including IRS spectroscopy from the Swift-BAT hard X-ray se- Seyfert1.8,Seyfert1.9andSeyfert2,hereafterSy2s). lectedlocalAGNsample(Weaveretal.2010). D: Spitzer-IRShigh-resolutionspectroscopyobservationsof 1H0=70kms−1Mpc−1,Ωm=0.27,andΩλ=0.73,adoptedthrough- a composite AGN sample (Dasyraetal. 2011). A list outthispaper. 2.2. X-rayDataReductionandSpectralFitting To obtain the intrinsic 2 – 10 keV luminositiesof sources inoursample,wefirstcollectdatafromliterature(seeTable 1 and Appendix A). For a major fraction of the sample we needtoperformourownspectralfittingtomeasuretheintrin- sicX-rayfluxesaseithertheX-raydataarenotpublishedyet, or no intrinsic 2 – 10 keV fluxesare givenin literature. For sources with multiple XMM observations, only the observations duringwhich the sources were Compton-thinare kept, and exposure-time weighted mean intrinsic luminosities are calculatedinlogarithmspace. TheXMM-Newton data areprocessedwith the XMM-SAS packageinthestandardway.Thesourceregionstoextractthe spectra are optimizedwith the “eregionanalyse”task. Back- groundextractionregionsareafewmanuallyselectedcircles withproperradiiaroundthesourceregion,beingcentralsym- metric as much as possible, and kept away from the CCD edges, the out-of-timeeventsstrips and other sources. Each exposure is checked for pile-up with the task “epatplot”. A corecircleregionwithasmallradiusisexcludedinthespec- tra extracting in case of pile-up. The circle radius is manu- ally selected by increasing step by step until the pile-up ef- fectnolongershowsupinthe“epatplot”result. Wegenerate FIG.1.— The relations between the 2 – 10 keV intrinsic X-ray emis- thesourceandbackgroundspectratogetherwiththeappropri- sion(absorption-correctedpower-lawemission)andthe[OIV]25.89µmline ate redistributionmatrix and ancillaryresponsefile from the emissionforSy1s(type1Seyfertgalaxies,bluedots)andCompton-thinSy2s source and backgroundregionsfor each exposure, using the (type2Seyfertgalaxies,redsquares). Sy2saredividedintotwotypeswith task “especget”. The Chandra and Swift-XRT data of a few NHlower(solidsquares)orhigher(opensquares)than1023cm−2.Acouple ofupperlimits to[OIV]luminosities areplotted inarrows. Linesplotthe sourcesarealsoprocessedwiththeirstandardpipelines. best-fitcorrelations (through simplelinear regression), andshaded regions Tomeasuretheabsorption-correctedpower-lawfluxinthe arethe1σconfidencebandsofthefits.Thetypical(median)errorsinL[OIV] 2–10keVband,wefitthe0.5–10keVspectrain“Xspec” andL2−10keV are5%and4%respectively,andarenotplottedforsimplicity. withapower-lawabsorbedbybothGalacticandintrinsicab- Inthelowerpanelweplottheresidualsofthedatapointstothebest-fitlineof Sy1s(alongthey-axis)forbothSy1sandSy2s. Sy2sareobviouslyweaker sorbers,togetherwiththreeadditionalemissioncomponents– inintrinsic2–10keVemission. asoftexcess,acoldreflectioncomponentandanarrowFeKα lineat6.4keV.Weareonlyinterestedindecomposingthead- Sy2s:y=(0.86±0.08)×(x−40.99)+42.52±0.09 ditionalcomponentsfromthepower-law,butnotinthecom- Remarkably weaker intrinsic X-ray emission at given [OIV] plexoriginsofsoftexcessandreflection. ForSy1s,wefitthe luminosityintype2AGNsisseeninthisfigure. Weperform softexcesswiththetraditionallyusedblackbodymodel,while K-Stesttoexaminewhethertheresidualsofdatapointsfrom forSy2swithanextrasoftpower-lawmodel,becausethesoft Sy1sandSy2stothebest-fitslopeofSy1s(lowerpanelofFig. excessdisplaysdifferentoriginsforunabsorbedandabsorbed 1) are extracted from the same population. At a confidence AGNs (Corraletal. 2011). In some cases, additional black- levelof99.94%theyarenot.Thedeviationofresidualsofthe bodyorpower-laworionizedgasemissionisneededtomodel two subsamples shows that Compton-thin type 2 AGNs are thesoftexcess.Becausethespectralqualitiesofmostsources are notgood enoughto put reasonableconstrainson the pa- fainterinintrinsic2–10keVemissionbyafactorof2.8+−00..54, rametersof the reflectioncomponent(“pexrav”in “Xspec”), comparing with Sy1s. We note that orthogonal distance re- wefixthecutoffenergyofthepower-lawat200keV,andthe gression(ODR,Isobeetal.1990)yieldsdifferentcorrelation reflection fraction R at 0.5. Adopting different values of R slopes,butdoesnotalterotherresultsinthiswork. Thebest- does not alter the results in this work (see §4.4). In a few fittedODRlinesare: cases,abroadFeKαlineisalsoneededtoimprovethefitting Sy1s:y=(1.00±0.06)×(x−41.17)+43.15±0.06 statistics. Sy2s:y=(0.99±0.08)×(x−40.99)+42.56±0.09 3. COMPARISONBETWEENTYPE1ANDTYPE2AGNS Inourstatisticalanalyses,weomit5sources(4Sy1sand1 In Fig. 1, we plot the intrinsic 2 – 10 keV X-ray lumi- Sy2, plotted in Fig. 1) with only [OIV] upper limits. These nosities versus [OIV] luminositiesof our sample. Clear cor- [OIV] upperlimits couldbe taken into accountwith survival relations between intrinsic 2 – 10 keV X-ray and [OIV] line analysesbyeithertakingtheintrinsic2–10keVluminosityas emission,bothofwhichsomehowreflecttheintrinsicpower theindependentvariable,ortreatingtheupperlimitsto[OIV] oftheSMBHaccretion,aredetectedforbothtype1andtype luminositiesas lower limits to 2 – 10 keV emission instead. 2 AGNs, confirmingpreviousstudies (Meléndezetal. 2008; Both approachesyield consistentresults. Here we adoptthe Rigbyetal. 2009; Diamond-Stanicetal. 2009; Weaveretal. morestraightforwardapproachbyexcludingthe[OIV]upper 2010). Simple linear regression (Isobeetal. 1990, taking limitsandtaking[OIV]astheindependentvariable. [OIV] as the independentvariable)is performedon the Sy1s Hard X-ray photons above 10 keV are insensitive to andSy2ssubsamples.Thebest-fittedlinesare: Compton-thin obscuration. We match our sample to Swift- BAT 70 month catalog (Baumgartneretal. 2013) to extract Sy1s:y=(0.88±0.05)×(x−41.17)+43.13±0.06 their14–195keVX-rayfluxes. ForSwift-BATnon-detected FIG.2.—Swift-BAT14–195keVversusintrinsic2–10keVluminosities. Sy1sandCompton-thinSy2sfollowastatisticallyidenticaltightcorrelation. Arrows plotthe upperlimits toBATnon-detected sources. Lines plot the best-fitcorrelations,andshadedregionsthe1σconfidencebandsofthefits. sources,weadoptanupperlimitof1.34×10−11ergcm−2s−1 to their 14 – 195 keV fluxes, which is the 5σ sensitiv- ity limit of Swift-BAT all sky survey for 90% of the sky (Baumgartneretal. 2013). We plot 14 – 195 keV versus intrinsic 2 – 10 keV luminosities in Fig. 2, and perform FIG.3.—SimilartoFig. 1,butreplacingtheintrinsic2–10keVX-ray luminositywiththeSwift-BAT14–195keVluminosity.Thetypicalerrorin Buckley-James linear regression on type 1 and type 2 subsamples,takingaccountoftheupperlimits. Bothtype1and L14−195keVisplottedatthelowerrightcornerintheupperpanel.Arrowsplot theupperlimitstoBATnon-detectedsources. type2sourcesfollowastatisticallyidenticaltightcorrelation. 10 – 200 keV emission comparing with Sy1s by a factor Thebest-fittedlinesinFig. 2are: of 1.9±0.5. This decrement in 10 – 200 keV emission in Sy1s:y=(0.93±0.03)×(x−43.23)+43.63±0.03 Compton-thinSy2s,althoughstatisticallyinsignificant(91%), isingoodagreementwithours.Howeveritishardtoattribute Sy2s:y=(0.92±0.05)×(x−42.63)+43.14±0.04 it to obscuration since it would require an average obscura- Thisconfirmsthatbothquantitiesmeasuretheintrinsiccorona tion of logNH = 24.3±0.1 cm−2, which is too high com- emission. It also proves that our absorption corrections to paring with the median column density for their Compton- the2–10keVfluxesaresufficient, otherwisewe wouldsee thinsample(logN =23.0cm−2)measuredbasedon2–10 H weaker2–10keVemission(relativeto14–195keVemis- keVspectralfitting. We notethatthesmaller samplesize of sion)inSy2s. Rigbyetal. (2009) and the fact that a significant fraction of Comparing 14 – 195 keV emission with [OIV] emission theirsourcesonlyhaveupperlimitsto10–200keVemission (Fig. 3)wellconfirmsthepatternshowninFig. 1. Residuals mighthavepreventedthemtodetectastatisticallysignificant for Sy1s and Sy2s are calculated in the same way as above differencebetween Sy1s and Compton-thinSy2s. A similar (lowerpanelinFig. 3) andthencomparedwith logranksur- trendwasalsoseeninLiu&Wang(2010). Byutilizing[OIV] vivalanalysis. Wefindasignificantdifferenceataconfidence emissionasintrinsicluminosityindicator,theyfoundthatthe level of 99.1%, and a flux decrementin type 2 sources by a monochromatic6.4keVcontinuuminSy1sis4.2±1.6times factorof2.2+0.9. Thebest-fitted(Buckley-Jamesregression) strongerthanCompton-thinSy2s,whichonaveragerequires linesinFig.−30a.3re: aN =8.6±1.9×1023cm−2 inCompton-thinSy2s,similar H tothevaluereportedbyRigbyetal.(2009). Sy1s:y=(0.83±0.06)×(x−41.17)+43.50±0.06 Sy2s:y=(0.71±0.10)×(x−40.99)+43.06±0.09 4.2. OntheIsotropyof[OIV]25.89µmLineEmission Although [OIV] 25.89µm emission line is believed to Sy2s tend to have a slightly flatter correlation slope (though be a good isotropic indicator of AGN intrinsic lumi- statisticallyinsignificant)comparingwithSy1s. Itispossible nosity (Diamond-Stanicetal. 2009; Meléndezetal. 2008; that the difference between Sy1s and Sy2s is more signifi- Rigbyetal. 2009), below we discuss possible factors which cantathighluminosities. Howevermuchlargersamplesare mightaffectitsisotropy. neededtodemonstratethispostulation. Zhangetal.(2008)foundthatSy1shavesmaller[NII]/Hα ratioson the BPT diagramcomparingwith Sy2s. They pro- 4. DISCUSSION posedthatthe innerNLRofAGNislikely heavilyobscured 4.1. HintsinPreviousStudies in type 2 sources by the extendingregionof the torus. This It is widely assumed that the corona emission in AGNs is effect, if confirmed, could yield weaker [OIV] 25.89µm and isotropic.However,lowtomoderatelevelofanisotropycan’t [OIII]emissioninSy2sduetoheavyobscurations. be ruled out based on previous studies. Actually, in several Another factor is that, AGN with smaller covering factor studieshintsforanisotropicX-rayemissionarevisible. of the obscuring torus (and thus larger open-cone angle of Rigbyetal.(2009)comparedthe[OIV]25.89µmemission the NLR), is more likely to be viewed as Type 1, than that with the 10 – 200 keV X-ray emission for the RSA Seyfert with larger torus covering factor (Turneretal. 2009). Nat- sample. They found that Compton-thin Sy2s are weaker in urally, Sy1s tend to have stronger NLR emission than Sy2s Sy2s within the two samples, considering a few upper lim- Sy1 its to [OIV] luminosities. We find no statistical differencein Sy2 [OIV] luminosity distributions between Sy1s and Sy2s, con- firming the isotropy of [OIV] line emission as reported by Diamond-Stanicetal. (2009). However,if correctingthe as- 101 sumed 3.2 times increment in the [OIV] fluxes of Sy2s, the er distributionsbecomesignificantlydifferentbetweenSy1sand b um Sy2s(ataconfidencelevelof99.6%forthe12µmsampleand N 96.6% for the RSA sample respectively). We conclude that Sy2s unlikely have relatively stronger [OIV] emission com- paringwithSy1s. 4.3. OntheSampleIncompletenessandBias 100 0.00 0.05 0.10 0.15 0.20 The sample we complied is based on five [OIV] subsam- z ples, thus is not homogeneous. We test the robustness of FIG.4.—RedshiftdistributionsofSy1sandCompton-thinSy2s. our results with the two complete samples used above – the 12µm Seyfert sample and the RSA Seyfert sample, both of (Lawrenceetal.2013). whicharehighlycompletein[OIV]measurements. TheirX- Possibleslitlossin[OIV]measurementsmightalsoleadob- ray completenesses are also high, with available X-ray data servationalbiasif Sy1sandSy2shave differentredshiftdis- for91outofthe11312µmsourcesand75outofthe89RSA tributions. InFig. 4weseethatSy2stendtohavelowerred- sources. Also, the X-ray incompleteness is more severe to shifts in our composite sample, thus slit loss could be more type2AGNsasobscuredsourcesaremuchfainterin X-ray, severeinSy2sthaninSy1s. ButnoteslitlossinSpitzerspec- therefore correcting such incompleteness (if possible in the tracouldberatherweakanywayduetothelargeSpitzer-IRS future) would further strengthen our finding. For the 12µm slitwidth(4.7–11.1′′forobservationsadoptedinthiswork). sampleonly,we findSy2sarefainterinintrinsic2–10keV Furthermore,the[OIV]emissionregionwasfoundtobemore emission at given[OIV] luminosity comparingwith Sy1s by compactthat[OIII](Dasyraetal.2011). afactorof2.5+0.7 withaconfidencelevelof95.4%(38Sy1s −0.5 However, all three possible factors above lead to higher versus 18 Compton-thin Sy2s). Considering only the RSA [OIV]emissioninSy1scomparingwithSy2s,thereforecould sample,weobtainafactorof2.2+0.6 withaconfidencelevel noteasetherelativestrongerintrinsichardX-rayemissionin of89.7%(22Sy1sversus17 Com−0p.t5on-thinSy2s). Combin- Sy1s that we have detected. Instead, our findings would be ingthe12µmsampleandtheRSAsampleweobtainafactor furtherstrengthened. of 2.4+0.5 with a confidencelevelof 97.9%(44 Sy1s versus Onthecontrary,couldthe[OIV]emissionbeinfactintrin- −0.5 29Compton-thinSy2s)Consideringtheremainingsourcesin sically stronger in Sy2s than in Sy1s? Based on the large ourcompositesampleweobtainstatisticallyconsistentnum- dispersion in the ratios of [OIII] to hard X-ray luminosities, bers,afactorof3.2+0.9 withaconfidencelevelof96.6%(36 Trouille&Barger (2010) postulated that some AGNs may −0.8 Sy1sversus16Compton-thinSy2s).Thereforeourresultsare havelow[OIII]luminositiesbecauseofthecomplexityofthe independentoftheselectionofsubsamples. NLRstructure. Theyalsosuggestedthat[OIV]emissionmay Another issue is the possible bias in the optical identi- be similarly affected by the NLR structure complexity. Is it fications of Sy1s and Sy2s due to strong nuclei variations possiblethattheNLRofSy2sabsorbalargerfractionofion- in some sources. In other words, a Sy1 could be mis- izationemission–becauseofmoregasinSy2s,eitheraround identifiedasSy2ifitsnucleiactivitysignificantlyreducedre- thenucleioratpc–kpcscalesinthehostgalaxies–thuspro- cently,yieldingmuchweakerbroademissionlinesandX-ray vide stronger narrow emission lines? Diamond-Stanicetal. (2009)showedthatthe[OIV]luminositydistributionsarein- emission but relatively stronger [OIV] emission, since [OIV] takes a longer time to respond to nuclei variation. How- distinguishableforobscuredandunobscuredAGNsinacom- ever, sources with such state transitions would more likely plete AGN sample – the RSA Seyfert sample. This pro- be identified as Compton-thick in X-ray, due to the signifi- vides a strong support to the isotropy of [OIV] emission in cantweaknessofthecentralX-rayemission(Guainazzietal. thesensethatSy1sandSy2shaveintrinsicallysimilar[OIV] 2005), which would have been excluded from our sample. emission. Below we extend such test to examine whether To further address this issue, we re-divide our whole sam- weakanisotropyin[OIV]emissioncouldbetoleratedbycur- ple into two subsamples with N <1022 cm−2 (66 sources) rent data. Considering the best-fitted slope of Sy1s in the H intrinsic 2 – 10 keV luminosity – [OIV] luminosity distri- and NH >1022 cm−2 (59 sources) respectively, independent of their optical identifications. Following the method as de- bution (0.88, see §3), a 2.8 times decrement in the X-ray emission of Sy2s can be explainedby a 3.2 times increment scribed in §3, we find that Seyfert galaxies with 1022 cm−2 in their [OIV] emission. We test this possibility by com- < NH < 1024 cm−2 have weaker intrinsic X-ray emission paring the [OIV] luminosity distributions of Sy1s and Sy2s at given [OIV] luminosity than sources with lower NH (by a withintwocompleteAGNsamples–the12µmSeyfertsam- factor of 1.70+0.30, with a confident level of 99.5%). This −0.25 ple (Rushetal. 1993; Tommasinetal. 2008, 2010) and the demonstrates that the possible bias in optical identifications RSASeyfertsample(Maiolino&Rieke1995;Hoetal.1997; is unableto erase the observeddifferencebetween Sy1sand Diamond-Stanicetal.2009),asmentionedin§2.1.Bothsam- Sy2sinoursample. ples are highly complete in [OIV] line measurements in our 4.4. OntheAbsorptionCorrection composite[OIV] sample, with 113outof 116sourcesof the 12µmsampleandall89oftheRSAsample. Weperformlo- In§3,wehaveshownthatourabsorptioncorrectiontothe2 granktestsonthe[OIV]luminositydistributionsofSy1sand –10keVfluxisgenerallysufficient,asSy1sandSy2sfollow a statistically identical correlation between intrinsic 2 – 10 keVluminosityand14–195keVluminosity. Wealsofound that both intrinsic 2 – 10 keV emission and 14 – 195 keV R=0 emissionare relativelyweakerin Sy2sthan in Sy1satgiven 101 Sy1 [OIV] luminosity. We note previous studies also reported Sy2 that Sy2s have relatively weaker hard (> 10 keV) X-ray emission than Sy1s (Rigbyetal. 2009; Weaveretal. 2010; 100 0.6 0.8 1.0 1.2 1.4 LaMassaetal. 2010), and interpreted this result in terms of heavy X-ray absorptions. Particularly, Rigbyetal. (2009) R=1 foundthatknownCompton-thinSy2sareweakerin14–195 101 Sy1 keVemissioncomparingwithSy1sbyafactorof1.9±0.5, Sy2 andattributedittoapossibleobscurationwhichismuchheav- ber100 ierthanvaluesobtainedthroughspectralfittingandcouldat- um 0.6 0.8 1.0 1.2 1.4 N tenuate even 14 – 195 keV X-ray emission. However, such R=2 interpretation is unlikely for Compton-thin Sy2s, otherwise 101 Sy1 peculiarthickabsorberswhichareessentiallyimpenetrableto Sy2 X-rays would be required in most (if not all) Compton-thin 100 Sy2s, andsuchabsorberscan notbeaccountedforwith cur- 0.6 0.8 1.0 1.2 1.4 rentcommonX-rayabsorptionfittingprocedures. One may suspect that the absorption correction might be 101 R=3 Sy1 inadequateforheavilyobscuredSy2s(i.e. thosewith1023< Sy2 N <1024 cm−2),astheirobscurationsmighthavebeensig- H nificantly underestimated. However, in Fig. 1 we show that 100 0.6 0.8 1.0 1.2 1.4 Sy2s with 1023<NH <1024 cm−2 (24 out of 45) are statis- Flux/FluxR=0.5 tically indistinguishablefrom Sy2s with N <1023 cm−2 in H FIG.5.—Distributionsoftheratiosoftheintrinsic2–10keVfluxesmea- termsoftheirintrinsic2–10keVemission(relativeto[OIV]), suredwithR=0,1,2,3tothosemeasuredwithR=0.5. rulingoutsuchpossibility. Below we present further discussions on issues related to oursample. absorptioncorrection. Tombesietal. (2010) reported detections of highly ion- We use the traditional photoelectric absorption model ized X-ray obscurations in local AGNs with N ∼ 1022− H “wabs” in “Xspec” in spectra fitting. Using a more mod- 1024 cm−2,whichcan’tbecorrectedthroughourspectralfit- ern obscuration model “tbabs” instead of “wabs” does not ting. However, most of the ionized absorbers are not thick affect the absorption-corrected fluxes. The mean ratios of enough to significantly attenuate the 2 – 10 keV continuum theabsorption-correctedfluxesusingtwodifferentmodelsare throughComptonscattering. Furthermore,thesimilardetec- 99.97%and99.76%forSy1sandSy2srespectively. tionratesofsuchabsorbersintype1andtype2AGNssuggest Duringspectrafitting,wefixthereflectionfractionRofthe thattheyhave nofavoredinclination. Theirlargeskycover- pexrav model at 0.5. We show in Fig. 5 that adopting dif- age,ifconfirmedwithhighS/Ndetection,alsoindicatesthat ferentR valuesonlyproducesminorchangesto the intrinsic theywouldnotproduceinclinationdependentbiasestoX-ray 2–10keVfluxes,comparingwiththedifferencewefindbe- fluxes. tween Sy1s and Sy2s. Also, it was suggested that reflection inSy2sislikelystrongerthaninSy1s(Riccietal.2011),thus 4.5. OutflowingCorona adoptinga uniformR for both Sy1s and Sy2s would lead to It was proposed that there could be projection effect in asystematic(althoughratherweak)overestimationofthein- corona’s X-ray emission, as if the corona is flat and opti- trinsicX-rayemissioninSy2s. Correctiontothiseffectcould cally thick, it will produce weaker radiation at larger incli- furtherstrengthenourresult. nation angle (Zhang 2005). However, studies have shown Inacoupleoftype1AGNs,hardX-rayexcesseswerede- that the corona can only cover a minor fraction of the disk, tectedabove20keV,suggestingtheyarepartiallyobscuredby and should be patchy instead of flat/disk-like (Haardtetal. Compton-thickclouds(Turneretal.2009;Reevesetal.2009; 1994).Also,observationsshowedthatcoronaopacityissmall Risalitietal. 2009a), and such partially covering absorber (Zdziarskietal. 2000; Molinaetal. 2009). Therefore, the couldbecommonintype1AGNs(Tatumetal.2013). X-ray projectioneffectshouldberatherweakinX-raycoronaemis- spectral fitting at <10 keV thus could have underestimated sion.Lightbendingeffectcausedbythestronggravityaround theintrinsic2–10keVemissioninAGNs. Thiseffect,how- theSMBHcouldalsoproduceanisotropy(Fabian&Vaughan ever doesnot affectthe major results in this work, since the 2003; Chenetal. 2013) in X-ray emission, however in the partiallycoveringabsorber,ifubiquitous,producesconsistent waycontrarytoourfinding. biasestotype1andtype2sources,sincewehaveshownthey Thereareotherfactorsthatmaycontroltherelativestrength followanidenticalcorrelationbetween14–195keVandin- of X-ray emission in AGNs, i.e., luminosity and Edding- trinsic 2 – 10 keV emission (see Fig. 2). The biases could tonratio(normalizedaccretionrate),inthesensethatAGNs havebeencanceledwhile we comparebetweenthem. Actu- with higher luminosity and/or Eddington ratio tend to have ally,theaverageratioof14–195keVfluxtointrinsic2–10 weaker X-ray emission relative to the accretion power (e.g. keV flux in our sample is consistentwith a power-lawspec- Fanalietal.2013).Ouranalysesareperformednormalizedto trumwithaphotonindexΓ=1.78+0.18(Γ=1.87+0.17assum- −0.37 −0.38 [OIV]luminosity,thushavecorrectedthepossibleeffectoflu- ingapower-lawspectrumwithareflectionfractionR=0.5), minosity. Sy1sand Sy2s generallyshow consistentphysical suggesting that partial covering absorption effect is weak in parameters, including black hole mass, luminosity and Ed- dingtonratio(Singhetal.2011). X-rayspectralslopesΓalso measuresthe Eddingtonratio(Fanalietal. 2013). Our spectral fittings yield consistent Γ between Sy1s and Sy2s (Fig. 7), indicatingthattwosubsampleshaveonaveragethesame Eddington ratio. Therefore the observed relatively stronger intrinsic hard X-ray emission in Sy1s than in Sy2s are not causedbydifferentEddingtonratiosbetweenthetwosubsam- ples. Under the scheme of AGN unification model, the difference we find between Sy1 and Compton-thin Sy2 indicates anisotropyofthe X-rayemission, whichcan be attributedto thebeamingeffectofcoronaoutflowingalongtheaxisofthe accretionsystem. Below we calculate the bulk outflowing velocity of the corona required to produce the observed anisotropy. Under theschemeofAGNunificationmodel,weattributetheclassi- 4 ficationofSy1,Compton-thinSy2andCompton-thickSy2to differentrangesofinclinationangles. Based onstudyof the 3.5 12µm galaxy sample, an obscured fraction of AGN of 62% andaCompton-thickfractionofAGNof∼20%werereported 3 (Brightman&Nandra2011). Accordingto these values, we FSy2 define0◦−52◦and52◦−78◦astherangesofinclinationan- F/Sy12.5 glesofSy1sandCompton-thinSy2s(seeFig. 6). Assuminganoptical-thincoronawithanoutflowingveloc- 2 ity of β =v/c, the beaming factor is defined as the relative intensityobservedfromaninclinationanglei: 1.5 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 b=(γ−1(1−βcosi)−1)3+α Bulkvelocityβ whereγ =(1−β2)−1/2,αistheenergyindexofthepower- FIG.6.—Intheupperpanel,weplottheprobabilitythataSeyfertgalaxy couldappearasSy1orSy2asafunction oftheinclination anglei(orthe law. Here a typical value of α=0.8 is taken. We calculate solidangle1-cosi)inourtoymodels.Thefilledblueandredregionsmark theratioofthebeamingfactorsaveragedwithintherangesof thedistinctinclinationanglerangesof0◦−52◦forSy1sand52◦−78◦for Sy1sandCompton-thinSy2satdifferentbulkvelocities,and Compton-thinSy2s. Withtheseassumptions, therequiredbulkoutflowing we find that a flux decrement of 2.8+0.5 times of Compton- velocitytoexplaintheaveragedifferenceintheintrinsicX-rayemissionbe- −0.4 tweenSy1sandCompton-thinSy2s(FSy1/FSy2)isplottedwithasolidlinein thin Sy2s comparing with Sy1s indicates an outflowing ve- thelowerpanel. Thehachuredregionsundertheblue(Sy1)andred(Sy2) locity of0.47±0.05c (see Fig. 6). Thiscorrespondsto the curvesintheupperpanelplotacomparisonmodel. Insuchatoymodel,the escape speed at ∼7−12 GM/c2, and the electron thermal pfuronbctaiboinlitoyfothfeasSoelyidfearntgglael,axwyhbicehinrgedvuiecwesedfraosma1Sya1t0(b◦lutoe0cu.5rvaet)5i2s◦a,cionstihnee speedinplasmawithtemperatureT∼64keV(E=kT).The sensethatattheotherwiseassumedtransitionangleofi=52◦betweenSy1s factor of 1.70+0.30 (if simply comparingsources with N > andSy2s,aSeyfertgalaxyhasequalpossibilitytoappearasaSy1oraSy2. −0.25 H Underthissituation,therequiredbulkoutflowingvelocityishigher,asshown 1022cm−2withthosewithsmallerNH,regardlessoftheirop- bythedashedlineinthelowerpanel. ticalidentifications,see§4.3)canbetranslatedtoanoutflow- ingvelocityof0.29+−00..0068c,andthefluxdecrementof2.2+−00..93 accurate calculations rely on future exact knowledge on the in14–195keVemission(seeFig. 3)correspondstoanout- inclinationsofSy1sandCompton-thinSy2s. flowingvelocityof0.38+0.12c. In radio-loudAGNswhich constitute only ∼10−20%of −0.06 ThedifferenceininclinationanglebetweenSy1sandSy2s theAGNpopulation(Kellermannetal.1989),theradioemis- is likely not as distinct as we have assumed, as suggested sionoriginatesfrompowerfulrelativisticjetsuptokpc–Mpc by the revised unification model (Elitzur 2012), the clumpy scales,thelaunchofwhichisanoutstandingpuzzleinastro- torus model (Nenkovaetal. 2008), the possibility that some physics(Nemmenetal.2012).Inradio-quietones,jet-likera- Sy2s are due to large scale obscurationsin the hostgalaxies diomorphologiesarealsooftendetected(Panessa&Giroletti (Matt 2000; Malkanetal. 1998; Rutkowskietal. 2013), and 2013), though much weaker and more compact. Learning the misalignmentbetweenthe axisof jet and the axisof ob- how weak and compact jets are launched can help to un- scuration in some AGNs (Rabanetal. 2009). For example, derstand the nature of more powerful ones. Strong corre- theclumpytorusmodelcouldallowforSy1sviewededge-on lations between X-ray and radio emission were detected in and Sy2s viewed face-on, although with very low probabil- radio-quiet AGNs (Panessaetal. 2007; Laor&Behar 2008; ities. These effects would reduce the difference in inclina- Wuetal. 2013) and in X-ray binaries (Merlonietal. 2003), tion between two populations (Sy1s and Sy2s), and require suggestinga close corona-jetcoupling. How couldthishap- even higher outflowing velocities. To illustrate such effect, penisstillunclear.Populartheoreticalinterpretationsinclude: we adopta simple toy modelby assumingthat the probabil- bothX-rayandcoreradioemissionoriginatefromstatic(non- ity of a Seyfert galaxy appearing as Sy1 gradually reduces outflowing) coronal activity (Laor&Behar 2008); magneti- fromface-ontoedge-on(asacosinefunctionofthesolidan- callydominatedcoronaproducesjet/outflowthatisresponsi- gle,seetheupperpanelofFig. 6andthecaptionfordetails), bleforradioemission(Merloni&Fabian2002);abortedjets andcomparetheoutflowingvelocityrequiredinthiscasewith (Ghisellinietal. 2004) produce both radio and X-ray emis- thatin the abovecase (see the lowerpanelof Fig. 6). More sion; or the role of X-ray corona itself could be subsumed 1.5 1.5 2.0 2.5 3.0 Although this pattern is statistically insignificant yet due to thelimiteddataquality,thispointsadirectiontofutureobser- 1.0 vations. 0.5 Another consequence of bipolar outflowing corona model al is relativelyweaker reflection componentfromthe accretion esidu 0.0 disk beneath the corona comparing with the static corona R−0.5 model (Beloborodov 1999; Miniuttietal. 2010), because of −1.0 Fixed the relativistic beaming effect. An outflowing velocity of Sy1 0.4c would reduce the relative strength of disk reflection by −11.65 Sy2 a factor of ∼ 2 – 8 comparing with a static corona ( view- Number1112468024 SSyy12 iddneigstkeicnitnciolsinnoamotfieosnAtrGdoenNpgsenr(edefl.egen.ct,TtiBaonnealockboaomertpoaodlno.ev1n91t99f59r;o9mR).eHythnoeowladecvscer1re9,ti9toh7ne; 0 1.5 2.0 2.5 3.0 Nandraetal. 1997), such as the broad Fe Kα line emission, Γ does not necessarily argue against the model of outflowing FIG.7.—Theresiduals inFig. 1asafunctionofX-rayspectralphoton corona. Inadditiontotheinclinationangleeffect,itispossi- indexΓ(upperpanel),whichisaniceindicatorofEddingtonratioinAGNs. ble that the corona outflowing velocity in AGNs is not uni- ThetrendofweakerX-rayemissioninSy2sisvisiblefromlowtohighΓ form. It could be very low in some AGNs, producing no values. ForacoupleofsourceswithtoofewX-raycounts,Γisfixedat1.8 (greensquares).Themedian(mean)ΓofSy1sandSy2sare1.81(1.86)and beaming effect, thus could explain (likely together with the 1.82(1.83)respectively. K-Stestalsoshowsthereisnostatisticaldifference light bending effect) the detected strong broad Fe Kα lines. betweenthetwodistributionsofΓ(lowerpanel). Higher outflowing velocity leading to stronger beaming effect, could also naturally explain the non-detectionof broad bythejetbase(Markoffetal.2005). Ourfindingremarkably Fe Kαlines inmanyAGNs(e.g.Nandraetal. 2007). Alter- matches the scenario that the role of corona could be sub- natively,therecouldbebothanoutflowingcorona,andanon- sumedbythebaseofjetinAGNs(Markoffetal.2005),and outflowing (or even inflowing like magnetic loops) corona couldnaturallyexplainthemysteriouscorrelationbetweenra- (Wilkins&Fabian2012). Inthisscheme,thedirectobserved dioandX-rayemissioninradio-quietAGNs. Twooutstand- continuum is dominated by the emission from the outflow- ingpuzzlesin astrophysics,thelaunchesofcoronaeandrel- ingcorona, while the disk is mainly illuminatedby the non- ativistic jets, could thereforebe directly bound, and even be outflowing(orinflowing)corona.Asthevariationsoftheout- merged.Thesamepicturecouldbeappliedtootherblackhole flowingcoronaandthenon-outflowing(orinflowing)corona accretionsystems,suchasX-raybinaries. do not necessarily correlate, this diagram may naturally ex- plainthepuzzlingconstantdisk reflectionstrengthwhilethe 4.6. FinalRemarks continuumemission is highlyvariablein a few sources(e.g. We have foundthat X-ray coronaemission is intrinsically Fabianetal. 2002), andthe non-uniformvariationpattern of weaker in type 2 AGNs than in type 1 AGNs. Studies tak- the disk reflection relative to X-ray continuum in different inghardX-rayemissionasisotropicluminosityproxytocal- sources (e.g. Iwasawaetal. 1996; Wangetal. 1999, 2001; ibrate emission in other bandscould thereforehave been bi- Shuetal.2010). ased. ThisalsoimpliesthathardX-raysurveys(evenat>10 Theoreticalcalculations(Malzacetal.2001)showthatout- keV) are biased against not only Compton-thick AGNs, but flowingcoronaeproducehardX-rayspectrawhicharemuch also Compton-thin type 2 sources, and the obscured frac- flatterthanobservedvaluesinAGNs(Γ∼1.4versus∼1.9) tion of AGNs yielded from hard X-ray surveys could have with a bulk-outflowing velocity of ∼ 0.4c. However, this been underestimated. Interestingly, by taking the anisotropy relies on the assumption that the disk reprocessed radia- ofhardX-rayemission(althoughinterpreteddifferently)into tion is the main cooler of the coronal plasma, while soft account,onecantackletheinconsistenceintheobservedob- radiation produced viscously from the accretion disk and scured AGN fraction and its dependence to luminosity be- bremsstrahlung/cyclo-synchrotron emission in the corona, tweenX-raysamplesandoptical,radioandIRselectedones have been neglected (Malzacetal. 2001). Such assumption (Zhang 2005; Mayo&Lawrence 2013), providing an inde- couldonlybevalidifthecoronaisinformofstrongconcen- pendentsupporttotheanisotropyofcoronaX-rayemission. trated flares and locally the reprocessed flux from the disk Outflowingcoronaemayproducesystematicallyhighercut- dominates the soft seed radiation (Haardtetal. 1994), and offenergiesinthehardX-rayspectraintype1AGNsthanin maynotholdinreality. type2sourcesduetoDopplershift.Assumingtheviewingan- WenotethelargedispersionsinthecorrelationbetweenX- glestotype1AGNsspanfrom0◦−52◦and52◦−90◦totype ray and [OIV] luminosities. The standard deviations of the 2AGNs(Brightman&Nandra2011),theexpecteddifference residuals (as defined in §3, see Fig. 1) are 0.5 and 0.7 for incut-offenergybetweentwopopulationsisbyafactorof1.4, Sy1sandSy2srespectively,both(slightly)largerthantheav- adoptingabulkoutflowingvelocityof0.47c. Thisdifference eragedifferencebetweenSy1sandSy2s(log2.8∼0.45).The ishowevertoosmalltobedetectedwithcurrentobservations largescattercouldbecausedbymanyfactors,includingX-ray consideringthe large uncertaintiesand huge intrinsic scatter variation, the structureof NLR, the spectral energydistribu- in the measured cut-off energies (Molinaetal. 2013). Nev- tionofAGNs, etc. Interestingly,adispersionintheoutflow- ertheless, it could be an efficient approach to measuring the ingvelocityofthecoronainAGNscouldalsoproducealarge coronaoutflowinthefuture.Riccietal.(2011)comparedthe scatterin thecorrelationbetweenX-rayand[OIV] luminosi- composite INTEGRAL spectrum of Sy2s with that of Sy1s. ties. Further investigationsare requiredto interpretboth the Interestingly, from their Fig. 7 we can see a likely drop at large scatter and the offset in the correlations between Sy1s above 80 keV in the ratio between the composite spectra of andSy2ssimultaneously. Sy2sandSy1s,suggestingSy2shavelowercut-offenergies. 5. SUMMARY 3: Under the scheme of AGN unification model, this result suggests the coronae in AGNs are bipolar outflowing To investigate whether the corona emission in radio-quiet withavelocityof∼0.3−0.5c. Suchoutflowingcoro- AGNsisisotropic,wecompilealargesampleofAGNswith naecouldnaturallybelinkedtothebasesofweakjets both[OIV]25.89µmemissionlinefluxmeasurementsandX- insuchsystems. ray spectra. Only secularly identified Compton-thinsources are included. Known radio-loud sources are also excluded. 4: Wediscusstheimplicationsofoutflowingcoronae,includ- Taking [OIV] luminosity as a proxy of AGN intrinsic lumi- ing the relative strength of disk reflection component nosity,wecomparetheabsorption-corrected2–10keV(also (i.e. the broad Fe Kα line), the selection bias in hard Swift-BAT14–195keV)luminositiesofSy1sandCompton- X-ray surveys, the validity of using hard X-ray emis- thinSy2s. Ourresultsaresummarizedasfollows: sion as AGN luminosity proxy, the obscured fraction of AGNs in X-raysurveys, and a particularprediction 1: At given [OIV] luminosity, the absorption-corrected 2 – that Sy2sshould have smaller X-raypower-law cutoff 10 keV X-ray emission is stronger in Sy1s than in energiescomparingwithSy1s,whichcouldbetestified Compton-thin Sy2s at a confidence level >99.9% by withfuturehardX-rayobservations. a factorof2.8+0.5. ConsistentpatternisseeninSwift- −0.4 BAT14–195keVemission. This workissupportedby Chinese NSF throughgrant 2: Basedoncarefulanalyses,wearguethatthedifferencecan 10825312&11233002. JXWacknowledgessupportfrom not be attributed to sample selection, insufficient ab- Chinese Top-notch Young Talents Program. We thank sorption corrections, or anisotropy in [OIV] line emis- TingguiWang,FengYuan,AndrzejZdziarskiandFuguo sion. This is the first solid detection of moderate Xiefordiscussions,andtheanonymousrefereeforhelpful anisotropyinAGN’scoronaX-rayemission. comments. APPENDIX SHORTNOTESONINDIVIDUALSOURCES A short note is given to each source whose intrinsic 2 – 10 keV flux is taken from literature. Generally their X-ray spectra areofhighqualityandhavebeennicelyinterpretedinpreviousstudies. Forsourceswithmultipleobservations,weadopttheir exposure-timeweightedmeanintrinsicluminositiesinlogarithmspace. NGC3516: Turneretal.(2005)performeddetailedspectralanalysisonthreeXMM/ChandraobservationsofNGC3516,revealed threedistinctabsorbingcomponents. 2MASXJ05580206-3820043(H0557-385): Longinottietal. (2009) foundwith XMM data that 2MASX J05580206-3820043 is absorbed by both ionized gas and partially covering neutral gas. It showed extreme flux variation by a factor of 10, entirelyduetointerveningline-of-sightclouds. ESO434-G040(MCG-5-23-16): Braitoetal.(2007)revealedwithXMM-Newton-RGSspectrumthatthesoftX-rayemission of ESO 434-G040 is likely dominated by several emission lines superimposed on an unabsorbed scattered power-law continuum.ThisSy1.9hasabroadFeKαline. UGC03973(Mrk79): Galloetal.(2011)fittedthespectraofUGC03973(XMM/Suzaku)withapower-lawwithionizedreflec- tionandabsorption,plusabroadFeKαline. NGC3227: Markowitzetal.(2009)modeledthetime-averagedspectrumofNGC3227(observedbyXMM)asamoderatelyflat power-lawwithtwoionizedabsorbers. NGC3783: Krongoldetal.(2003)analyzedtheChandraHETGSspectrumofNGC3783.Theirmodelconsistsofatwo-phase ionizedabsorber. NGC4388: Beckmannetal. (2004) presented INTEGRAL and XMM observationsof NGC 4388, and foundheavy absorption (2.7×1023cm−2). NGC4395: Nardini&Risaliti (2011) analyzed the XMM and Suzaku spectra of the dwarf Seyfert NGC 4395, attributing its spectralvariationtovaryingpartialcoveringabsorber. NGC4507: Braitoetal.(2013)presentedtheSuzakuobservationofNGC4507,foundavariableabsorberandastrongreflected component. ESO323-G077: Jiménez-Bailónetal.(2008)analyzedtheXMMspectraofESO323-G077,foundaneutralabsorber(withN H =5.82×1022cm−2),andtwoionizedabsorberplusabroadFeKαline. NGC1365: Risalitietal.(2009b)detectedanabsorptionwithN ∼3.5×1023cm−2 crossingthelineofsightbasedona60ks H XMMobservation.AbroadFeKαlineisalsodetected. FAIRALL0049(IRAS18325-5926): Moczetal.(2011)detectedablue-shiftedionizedabsorberintheChandraHETGSspec- trumofIRAS18325-5926. NGC5506: Guainazzietal.(2010)foundabroadFeKαlineintheX-rayobscuredNLS1NGC5506usingXMMdata. MRK0273: The ULIRG Mrk 273 is mergingwith an unabsorbedSy2 Mrk 273x, which lies 1.2 arcmin away. Balestraetal. (2005)usedthreecollisionallyionizedplasmacomponentsinadditiontoanabsorbedpower-lawtofittheXMMspectrum ofMrk273.Theintrinsicfluxis22.1×10−13. MCG-03-34-064: Miniuttietal. (2007) analyzed the XMM spectrum of MCG-3-34-64 (IRAS 13197-1627), recovering Compton-thinabsorptionandabroadFeKαemissionline. NGC4151: Lubin´skietal.(2010)presentacomprehensivespectralanalysisofNGC4151,usingtheINTEGRAL,RXTE,XMM, SwiftandSuzaku,consideringionizedabsorber,partialcoveringabsorber,andreflection. MRK0463E: Bianchietal. (2008)analyzedtheChandra,XMMandHSTdataofthedoublenucleusULIRGMrk463,which is consistof Mrk463EandMrk463Wwith a projectedseparationof3.83′′. Their2– 10keVintrinsic luminositiesare 1.5×10−43and3.8×10−42ergs−1respectively,asmeasuredwithChandradata.The[OIV]fluxofMrk463Eismeasured withSpitzerIRSLHmodule(slitwidth11.1′′),thusmaybepollutedbyMrk463W. NGC4253(Mrk766): The NLS1 Mrk766 is highlyvariablein the hardX-ray flux ontime-scales asshortas a few hundred seconds. Turneretal.(2007)attributesthespectralvariabilityofMrk766tothevariationsinacomplexandmulti-layered absorber. NGC6860: Winter&Mushotzky(2010)analyzedtheXMMandSuzakuobservationsofNGC6860,andfoundatwo-component warmionizedabsorberandabroadFeKαline. NGC0788,NGC6300,NGC7172,ESO103-G035: The2–10keVintrinsicluminositiesofthesesourcesaretakenfromthe broad-bandX-rayspectralanalysisbydeRosaetal.(2012)usingvariousavailableinstrumentsincludingXMM,Chandra, Suzaku,INTEGRALandBeppoSAX. FAIRALL0009: Lohfinketal. (2012) fitted the multi-epoch spectra of FAIRALL 9 as observed by XMM and Suzaku with a modelincludingarelativisticallyionizedreflectioncomponent. NGC1052: Brennemanetal. (2009) analyzedthe 101ksSuzakuspectrumofNGC 1052. The0.5–10keV continuumis well modeledbyapower-lawcontinuummodifiedbyGalacticandintrinsicabsorptionandasoft,thermalcomponentbelow1 keV.BroadFeKαlineisalsodetected. MRK335: MRK 335 was monitored by XMM with 200 ks exposure. Its spectra show warm absorber and blurred reflection (Galloetal.2013). UGC05025: Galloetal. (2005) fitted the Chandra spectrum of UGC 05025 (MRK 705) with a primary power-law and an additionalbrokenpower-lawrepresentingthesoftexcess. MRK1298: Giustinietal.(2011)foundmassiveionizedabsorbersalongthelineofsightofMRK1298(PG1126-041). IC5063: LaMassaetal.(2011)fittedtheChandraspectrumofIC5063withadoubleabsorbedpower-law. ESO383-G035: Milleretal.(2008)compiledalltheavailablelong-exposure,high-qualitydataforESO383-G035(MCG-6-30- 15):522ksofChandraHETGS,282ksofXMMPN/RGSand253ksofSuzakuXIS/PINdata.Avariablepartial-covering zoneplusabsorbedlow-ionizationreflection,distantfromthesource,providesacompletedescriptionofthevariableX-ray spectrum. NGC1097: WiththehighresolutionobservationofChandra,Nemmenetal.(2006)eliminatedthenearbyULXandsurrounding star-formingringofNGC1097,andpresentedthespectrumofthenucleusregion. IC4329A: Steenbruggeetal. (2005) detected seven distinct absorbing systems in the high-resolution X-ray spectrum of IC 4329AtakenwithXMM. NGC4051: Steenbruggeetal. (2009) fitted the Chandra LETGS high-resolutionspectrum of NGC 4051 with multiple warm absorbermodels. NGC1566,NGC4941: Kawamuroetal. (2013) presented broad band (0.5–195 keV) X-ray spectra of NGC 1566 and NGC 4941observedwithbothSuzakuandSwiftBAT.