Astronomy&Astrophysicsmanuscriptno.ms c ESO2008 (cid:13) February2,2008 Is the anti-correlation between the X-ray variability amplitude and black hole mass of AGNs intrinsic? (ResearchNote) YuanLiu1 andShuangNanZhang1 8 0 PhysicsDepartmentandCenterforAstrophysics,TsinghuaUniversity,Beijing,100084,China 0 e-mail:[email protected], [email protected] 2 n ABSTRACT a J Aims.BoththeblackholemassandtheX-rayluminosityofAGNshavebeenfoundtobeanti-correlatedwiththenormalizedexcess 7 variance(σ2 )oftheX-raylightcurves.Weinvestigatewhichcorrelationwithσ2 istheintrinsicone. rms rms 1 Methods.Wedivideafullsampleof33AGNs(O’Neilletal.2005)intotwosub-samples.Theblackholemassesof17objectsin sub-sample1weredeterminedbythereverberationmappingorthestellarvelocitydispersion.Theblackholemassesoftheremaining ] 16objectswereestimatedfromtherelationshipbetweenbroadlineregionradiusandopticalluminosity(sub-sample2).Thenpartial h correlationanalysis,ordinaryleastsquaresregressionandK-Stestsareperformedonthefullsampleandthesub-samples,respectively. p Results.Wefindthatσ2 seemstobeintrinsicallycorrelatedwiththeblackholemassinthefullsample.However,thisseemstobe - rms o causedbyincludingthesub-sample2intotheanalysis,whichintroducesanextracorrelationbetweentheblackholemassandthe r luminosityandstrengthensanycorrelationwiththeblackholemassartificially.Therefore,theresultsfromthefullsamplemaybe t misleading.Theresultsfromthesub-sample1showthatthecorrelationbetweenσ2 andtheX-rayluminositymaybetheintrinsic s rms a oneandthereforetheanti-correlationbetweenσ2rmsandtheblackholemassisdoubtful. [ Keywords.X-rays:galaxies–galaxies:active–methods:statistical 1 v 8 1. Introduction black hole mass and σ2 , and some models have been subse- 9 rms quentlyconstructedtoexplainthiscorrelation. 5 X-rayemissionfromactivegalacticnuclei(AGNs)exhibitsvari- 2 abilityon time scalesfromminutesto days.Thisindicatesthat In this paper we revisit this problem,by studying the sam- . X-rays are likely to be emitted from the inner most regions of pleofO’Neilletal.(2005),whichincludes33AGNsanduses 1 0 AGNsandthevariabilitymayberelatedtotheimportantproper- nearlythesame timescale foralltheseobjects.Theblackhole 8 tiesofthecentralengine.Lawrence&Papadakis(1993)utilized massesof17objectsinthissampleweredeterminedbythere- 0 long term EXOSAT observations to investigate the power den- verberationmappingorthestellarvelocitydispersion(wedenote : sityspectraof12AGNs.Theyfoundthepowerdensityspectra theseobjectsassub-sample1inthefollowing).Theblackhole v could be described as a power law P ν α with a mean in- massesoftheremaining16objectswereestimatedfromtherela- i − X dex α = 1.55, and the amplitude was ∝anti-correlated with the tionshipbetweenbroadlineregionradiusandopticalluminosity r X-rayluminosity.Detailedstudiesofthepowerdensityspectra (wedenotetheseobjectsassub-sample2inthefollowing).We a have been performedusing RXTE and XMM-Newton data. The findthattheopticalluminosity(whichisusedindeterminingthe universal relation between the black hole mass and the ”break blackholemassforsub-sample2)hasobviouscorrelationwith time”inthepowerdensityspectrawasfoundbothinstellarmass theX-rayluminosityin2-10keVband,asshowninFigure1(a). andsupermassiveblackholes(e.g.Uttely&McHardy2005).A Thevaluesofthecorrelationcoefficientsbetweenopticallumi- tighterrelationwasdiscoveredwhenthebolometricluminosity nosity and X-ray luminosity for sub-sample 1 and sub-sample wasinvolved,i.e.T M2/L(McHardyetal.2006).However, 2 are 0.907and0.910,respectively.Evenif the datumofNGC B ∝ due to the limited observationdata, the accurate powerdensity 4395isexcludedfromsub-sample2,thevalueofthecorrelation spectra are only available for a small number of AGNs. As an coefficient is still 0.819. It is well known that there is a strong alternative,thenormalizedexcessvariance(σ2 )canbeeasily correlation between the optical luminosity and the black hole rms calculated,anditwasfoundthatσ2 isanti-correlatedwithX- mass(Kaspietal.2000).Thusifσ2 isintrinsicallycorrelated rms rms rayluminosity(e.g.Almainietal.2000). withoneofthem,anartificialcorrelationwiththeotherwillap- As a result of the progress in determining the black hole pear. However, there will be an additional artificial correlation massesinAGNs,therelationbetweenthevariabilityandblack forsub-sample2duetotheutilizationoftheopticalluminosity holemasshasalsobeeninvestigated.Lu&Yu(2001)foundthe indeterminingtheblackholemass,whereastheblackholemass anti-correlationbetweentheblackholemassandσ2 ,andsug- ofsub-sample1is independentlyobtainedbythereverberation rms gested this correlationwas an intrinsic one, rather than the ap- mappingorthestellar velocitydispersion.Therefore,theresult parentanti-correlationbetween the X-ray luminosity and σ2 . fromthefullsample(especiallyforthatfromthesub-sample2) rms SeveralauthorsalsoaddressedthisproblemfollowingLu&Yu’s may be misleading. Furthermore,due to the less reliable black work.(e.g. Bian & Zhao 2003;Papadakis2004;O’ Neill et al. holemassofsub-sample2,someunclearsystematicbiasesmay 2005).Theseauthorsconfirmedtheanti-correlationbetweenthe be introduced into the analysis. To find out which correlation 2 YuanLiuandShuangNanZhang:Anti-correlationbetweenAGNX-rayvariabilityamplitudeandluminosity(RN) 46 of them indicate σ2 is intrinsically correlated with the black rms −1λ L (erg s)λ44442345 °+ SSuubbssaammppllee 21 htthooeltehmemolraiemssri,eterlaidathbseliezresthuoabfn-tshtahemepplrlueems1einnstohssoiawtmys.pHtlheoe,wwceoevnectrro,anrthycelrueadsneualtlthysa.stiDstuhoeef log −44401 NGC 4395 a breestuwltesenoflusmubin-soasmityplean2diσs2rmdosuibsttfhuel ainntdrinmsaicyboeneth.eMcoorrererloabtiuosnt −1 conclusionwillbededucedwhenalargersampleisavailable. 2σ)rms−−12. 5 og(−2.5 2.2.Ordinaryleastsquaresregression l −3 −3.5 b c To verifythe resultsobtainedin 2.1,we performthe ordinary −4 40logL241−10 ke4V2 (erg43 s−1)44 5 6 log7( M)8 9 10 leastsquareregressiontothefull§sample,sub-sample1andsub- sample 2, respectively.The regression equationis, log(σ2 ) = Fig.1.(a):ThecorrelationbetweenX-rayluminosity(2-10keV) AlogM+BlogL+C. rms andopticalluminosity(λ=5100 A◦ ).Thepointintheleft-down Theresultsoftheregressionforthethreesamplesaresum- cornerisNGC4395,whichisadwarfSeyfertgalaxy.Thevalue marized in Table 2. In Figure 2, we show the comparison be- of the correlation coefficient is not significantly influenced by tween the valuesof σ2 predictedby the resultsof the regres- rms whetherthispointisincludedornot(seethetextfordetails).(b): sionandtheobservedones. ThedataofthenormalizedexcessvariancesandtheX-raylumi- TheresultsofFstatisticdemonstratethehighsignificanceof nosities(2-10keV).(c):Thedataofthenormalizedexcessvari- the linear correlation.However,the values of χ2 are still large, ancesandblackholesmasses.Thevaluesofblackholemasses especiallyforthefullsample.Weshouldnoticethatthedepen- are in units of M . The data are obtained from O’ Neill et al. dence on the black hole mass and the luminosity seems to be (2005)andtheref⊙erencestherein. different for the two sub-samples. For sub-sample 1, σ2 ap- rms pearsto dependweakly on the black holemass, whereasit de- pendsmorestronglyonthe X-rayluminosity.Due to thesmall is intrinsic, we perform the partial correlation analysis to sub- sample, the difference between values of A and B are not very sample1andsub-sample2separatelyin 2.1.Theordinaryleast significant(theyarecoincidencewithinthe95%confidencein- § squaresregressionresultsareshownin 2.2asanotherapproach terval).However,ifthesub-sample2isincluded,thedependence § tothisproblem.K-Stestsareperformedin 2.3totestwhether ontheblackholemassisstrengthenedandthegoodnessofthe § sub-sample1andsub-sample2aredrawnfromthesameparent regression decreases dramatically. The value of the total χ2 of population.In 3wediscussourresultsandmakeconclusions. thesub-samplesis139(27);therefore,theprobabilityoftheim- § provement by chance is only about 10 9 (obtained by F-test). − Thustheaboveresultsindicatethatthesub-sample1and2are 2. Dataanalysis likelytoobeydifferentcorrelationrelationshipsanditisnotap- 2.1.Partialcorrelationanalysis propriatetocombinethemintoonesample. Thepartialcorrelationanalysisisanappropriatemethodtodis- entanglethecorrelationbetweenvariables.Thedefinitionofthe 2.3.K-Stests firstorderpartialcorrelationcoefficientbetweenvariablesxand cyoinst(rKolelneddavlalr&iabSlteuaarntd1r977i)s,trhxey.zc=orr√el(ar1xt−yi−ro2xrznx)z(r1cz−yorez2yffi),cwiehnetrebeztiwsethene sTtwaomoinesvupebas-rtseiganmattdepilswetsrh.iebTtuhhteieorncth,uwemteuwlfiaortsisvtuepbe-drsifasomtrrmipbluethtsieoanr1eDfduKrnac-wtSinotnefssrtoomtof ttthhheee xy variablesxandy. twosamplesarecalculatedfirst.Thenthemaximumvalueofthe We adopt the data of the black hole mass, the X-ray lumi- absolute difference between two cumulative distribution func- nosityandσ2 fromO’Neilletal.(2005)andpresentthemin tionsisusedasthestatistictoobtainthesignificanceofthedif- rms Figure 1 (b) and (c) for clarity. The correlationanalysis is per- ference (see details of the K-S test in Press et al. [1992]).The formedonthefullsampleandthesub-samples,respectively.The significancesof the differencesare 89%, 25% and 96% for the resultsareshowninTable1. distributions of the black hole mass, the luminosity and σ2 , rms For the full sample, both the black hole mass and the X- respectively(thecumulativedistributionfunctionsareshownin rayluminosityshowstrongapparentanti-correlationswithσ2 Figure 3). Clearly exceptfor the X-rayluminositydistribution, rms (Figure 1 [b] and [c]). However, it appears that after the black boththe blackhole mass and σ2 for the two sub-samplesare rms hole mass is controlled, the correlation between the luminos- not likely drawn from the same parent population. There is no ity and σ2 is not significant. On the contrary, the correlation obviousreasonaccountingforthedifferences,thereforethisre- between trhmes black hole mass and σ2 is still significant after sultislikelytobeduetosomeunknownselectioneffects,which rms theluminosityiscontrolled.TheseresultsseemtosupportLu& shouldbeinvestigatedfurtherinthefuture. Yu’ssuggestionthatthecorrelationbetweentheblackholemass Since it seems visually that the difference between the dis- andσ2 istheintrinsicone.However,asdiscussedin 1,since tributions of sub-sample 1 and sub-sample 2 in Figure 1 (b) is rms § any correlation between the black hole mass may be strength- moresignificantthanthatinFigure1(c),weperformthe2DK-S ened artificially by the sub-sample 2, we should exclude them test to investigate this problem. The results of the 2D K-S test when investigating the intrinsic correlation. For sub-sample 1, show that the significance of the difference in Figure 1 (b) is thecorrelationbetweentheblackholemassandσ2 disappears 89.4%,whereassignificanceofthedifferenceinFigure1(c)is rms when the luminosity is controlled, whereas the correlation be- 97.6%.Thisunexpectedresultisduetotheexistenceofthepoint tweentheluminosityandσ2 isstillsignificant.Theresultsof of NGC 4395. After this pointis removed,the 2D K-S test re- rms sub-sample2areconsistentwiththoseofthefullsample.Both sultsofFigure1(b)and1(c)are96.5%and92.1%,respectively. YuanLiuandShuangNanZhang:Anti-correlationbetweenAGNX-rayvariabilityamplitudeandluminosity(RN) 3 Table1.Resultsofpartialcorrelationanalysis. r r r r Lσ Mσ Lσ.M Mσ.L Fullsample -0.636(>99.99%) -0.697(>99.99%) -0.277(87.6%) -0.452(99.1%) Sub-sample1 -0.856(>99.99%) -0.560(98%) -0.781(>99.99%) -0.003(0.8%) Sub-sample2 -0.605(98.7%) -0.784(>99.99%) 0.238(60.8%) -0.653(99.2%) Table 2. Resultsofordinaryleast squaresregression.Thesignificanceofthe linearcorrelationis obtainedbythe F statistic. The value of χ2 is calculated from the regression result to estimate the goodnessof the regression. The errorscorrespondingto 95% confidenceintervalsareshown. A B C Fstatistic χ2(dof) Fullsample -0.38 0.28 -0.22 0.30 10.4 11.4 16.6(>99.99%) 708.6(30) ± ± ± Sub-sample1 0.00 0.45 -0.71 0.33 28.3 12.2 19.2(>99.99%) 46.4(14) ± ± ± Sub-sample2 -0.56 0.39 0.19 0.45 -6.06 17.3 11.4(99.86%) 92.6(13) ± ± ± 0 thesub-sample2.IftheX-rayluminosityistheprimaryquantity, −0.5 a b c then this will artificially strengthen any correlation with black 2σ)og(rms−1−.15 hinoglethmeaisnst.riWnseicthceorrerfeolaretioshnowulidtheσxc2rmlus.dAecthceomrdiwnghetno itnhveersetsigualtts- d l−2 fromthesub-sample1,weconcludethatthecorrelationbetween e dict−2.5 σ2 andtheX-rayluminositymaybetheintrinsicone,whereas Pre−3 thremsapparentcorrelation between σ2 and the black hole mass rms −3.5 isdoubtful.OurK-Stestsalsosuggestthatsub-samples1and2 −−44 −3 −2 −1 −−44 −3 −2 −1 −−44 −3 −2 −1 0 arenotlikelydrawnfromthesameparentpopulation. Observed log(σ2 ) As discussed in Lu & Yu (2001),several mechanisms may rms be responsible for the correlation between σ2 and the X-ray rms Fig.2.Thecomparisonbetweenthevaluesofσ2 predictedby luminosity,suchasthehot-spotmodel,theobscurativevariabil- rms the resultsof the regressionandthe observedones. The results ity and so on. After the apparentcorrelation between σ2 and rms fromthe fullsampleare shownin (a). (b):The same as(a) but the black hole mass was discovered, some models accounting forthesub-sample1.(c):Thesameas(a)butforthesub-sample for this correlation were proposed (e.g. O’ Neill et al. 2005, 2. Pessah2007).However,itneedstobeverifiedwhetherthecorre- lationisintrinsic.Althoughtheblackholemassesofaboutthree 1 dozen AGNs have been determined by the reverberation map- 0.9 a b c ping method, the size of our sample is still limited due to the 0.8 lack of long enoughand highquality observationdata of these 0.7 0.6 objects. More conclusive results could be obtained when more F CD0.5 andhigherqualitydatabecomeavailable. 0.4 0.3 Acknowledgements. We thank the referee, Andy Lawrence, for valuable 0.2 comments and suggestions. SNZ acknowledges partial funding support by 0.1 DirectionalResearchProjectoftheChineseAcademyofSciencesunderproject 0 4 5 6log7(M)8 9 3l9ogL402−1410 ke42V (e43rg 4s4−−14) −3.5−3 l o−g2.5(σ−22r m −s1).5−1 −0.5 Nunod.eKrJpCroXje2c-tYnWo.-T100532a1n0d01b,y1t0h7e3N30at1i0onaanldN1a0tu7r2a5l3S1c3i.enceFoundationofChina Fig.3.Thecumulativedistributionfunctions(CDF)oftheblack hole mass(a), the X-ray luminosity(b)and σ2 (c). The solid References rms linesare the dataof the sub-sample1, andthe dashedlinesare Almaini,O.,etal.2000,MNRAS,315,325 thedataofthesub-sample2.Thevalueofblackholemassisin Bian,W.,&Zhao,Y.2003,MNRAS,343,164 unitsofM . Kaspi,S.,etal.2000,ApJ,533,631 ⊙ Kendall, M.,&Stuart, A.1977,TheAdvanced TheoryofStatistics (London: Griffin) Althoughthesignificanceofthedifferenceineachfigureishigh, Lawrence,A.,&Papadakis,I.1993,ApJ,414,L85 thevisualdifferencebetweentwothefiguresisnotsignificant. Lu,Y.,&Yu,Q.2001,MNRAS,324,653 McHardy,I.,etal.2006,Nature,444,730 O’Neill,P.,etal.2005,MNRAS,358,1405 Papadakis,I.2004,MNRAS,348,207 3. Discussionsandconclusions Pessah,M.2007,ApJ,655,66 Press, W., et al. 1992, Numerical Recipes in C, Cambridge University Press, In 2,wehaveperformedthepartialcorrelationanalysisandthe Cambridge § regression on the sample and found that the apparent intrinsic Uttely,P.,&McHardy,I.2005,MNRAS,363,586 correlationbetweenσ2 andtheblackholemassislikelytobe rms causedbyincludingthesub-sample2intotheanalysis.Because theblackholemassesofAGNsinsub-sample2wereestimated from their optical luminosity which in turn is positively corre- lated with their X-ray luminosity, an extra correlation between theblackholemassandX-rayluminositywillbeintroducedby