Table Of Content

DRAFTVERSIONJANUARY4,2013 PreprinttypesetusingLATEXstyleemulateapjv.5/2/11 ASOFTX-RAYREVERBERATIONLAGINTHEAGNESO113-G010 E.M.CACKETT1,A.C.FABIAN2,A.ZOGBHI3,E.KARA2,C.REYNOLDS3,ANDP.UTTLEY4 1DepartmentofPhysics&Astronomy,WayneStateUniversity,666W.HancockSt,Detroit,MI48201 2InstituteofAstronomy,UniversityofCambridge,MadingleyRd,Cambridge,CB30HA,UK 3DepartmentofAstronomy,UniversityofMaryland,CollegePark,MD20742,USAand 4AstronomicalInstitute‘AntonPannekoek’,UniversityofAmsterdam,Postbus94249,1090GEAmsterdam,theNetherlands DraftversionJanuary4,2013 3 ABSTRACT 1 Reverberationlagshaverecentlybeendiscoveredinahandfulofnearby,variableAGN.Here,weanalyzea 0 ∼100ksecarchivalXMM-NewtonobservationofthehighlyvariableAGN,ESO113- G010inordertosearch 2 forlagsbetweenhard,1.5– 4.5keV, andsoft, 0.3–0.9keV,energyX-raybands. Atthe lowestfrequencies n availableinthelightcurve(.1.5×10- 4Hz),wefindhardlagswherethepower-lawdominatedhardbandlags a thesoftband(wherethereflectionfractionishigh). However,athigherfrequenciesintherange(2- 3)×10- 4 J Hzwefindasoftlagof- 325±89s. Thegeneralevolutionfromhardtosoftlagsasthefrequencyincreases 3 issimilartootherAGNwheresoftlagshavebeendetected. We interpretthissoftlagasduetoreverberation fromtheaccretiondisk,withthereflectioncomponentrespondingtovariabilityfromtheX-raycorona. Fora E] blackholemassof7×106M⊙ thiscorrespondstoalight-crossingtimeof∼9Rg/c,however,dilutioneffects meanthattheintrinsiclagislikelylongerthanthis. Basedonrecentblackholemass-scalingforlagproperties, H thelagamplitudeandfrequencyaremoreconsistentwithablackholeafewtimesmoremassivethanthebest h. estimates,thoughflux-dependenteffectscouldeasilyaddscatterthislarge. p Subjectheadings:accretion,accretiondisks—galaxies:active—galaxies:nuclei—X-rays:galaxies - o r 1. INTRODUCTION wheretheyhavebeendetected,ageneralshapeemerges,with t s hardlagsatthelowestfrequencies,transitioningtosoftlagsat a Since the discovery of the first X-ray reverberation lag [ in the AGN 1H 0707- 495 (Fabianetal. 2009), X-ray lags higherfrequencies,interpretedasviscouspropagationsdom- between lightcurves from the soft excess and power-law inating on long timescales and reverberation dominating on 2 shortertimescales.Thisisnottheonlyproposedmodelforthe dominated regions of the spectrum (‘soft’ lags) have now v lags–Milleretal.(2010)suggestanalternativeinterpretation been observed in over a dozen AGN (Zoghbi&Fabian 4 involving reflection from distant absorbers homogeneously 2011; Emmanoulopoulosetal. 2011; Tripathietal. 2011; 7 distributedalong our line of sight, somethingthatrequiresa DeMarcoetal. 2011, 2012; Fabianetal. 2012) and also in 8 the black hole X-ray binary GX 339- 4 (Uttleyetal. 2011). specificgeometry. While sucha modelcouldpotentiallyex- 7 plainonesource,the presenceofsoftlagsina largenumber Recently,thefirstdiscoveryofabroadFeKlaghasnowbeen 0. reportedinNGC4151(Zoghbietal.2012),withlargeXMM- of AGN rules out this model in general (further arguments 1 Newton datasets from 1H 0707- 495and IRAS 13224- 3809 againstthismodelaregiveninZoghbietal.2011). 2 also revealing an Fe K lag (Karaetal. 2012a,b). Such lags In this manuscript, we present an X-ray timing anal- 1 are naturally predicted in the reflection paradigm where a ysis of ESO 113- G010 (z = 0.0257), using XMM- : Newtonobservations, which we find to show the same lag- v corona (the power-law component) irradiates the accretion frequency evolution as has been observed in other AGN. Xi disk, leading to a reflected componentconsisting of fluores- ESO 113- G010is a Seyfert 1.8 displaying a strong soft ex- centemissionlinesandscatteredcontinuumemissionthatare cess and X-ray variability (Pietschetal. 1998; Porquetetal. r broadened and skewed by the dynamical and relativistic ef- a 2004,2007) whicharethetwo basiccomponentsneededfor fectspresentintheaccretiondiskclosetotheblackhole(see tracingnegativereverberationtimedelays. Reynolds&Nowak2003;Miller2007,forreviewsofreflec- tion). The lags can then be ascribed to the light travel time 2. DATAANALYSIS between the direct power-law component and the reflection ESO 113- G010 was observed by XMM-Newton on components (see Reynoldsetal. 1999, for a theoretical dis- 2005/11/10foratotalof104ksec, ObsID:0301890101(see cussiononX-rayreverberation). Porquetetal.2007,forapreviousanalysisofthesedata).We Lags in the opposite sense (‘hard’ lags where the soft analyzedtheXMM-NewtondatausingSASversion11.0.0and band leads the hard) are also seen on longer timescales. themostrecentcalibrationfiles. Calibratedeventsfileswere These hard lags have been seen in both X-ray binaries and generated from the Observation Data Files using the epproc AGN (e.g. Miyamoto&Kitamoto 1989; Nowaketal. 1999; andemproctoolsfortheEPIC/pnandEPIC/MOSdetectors, Papadakisetal. 2001; Arévaloetal. 2006), and are thought respectively. We searchedforbackgroundflaresin the usual toariseduetoviscouspropagationofmassaccretionfluctua- manner, looking at the lightcurves from the entire detector tionsinthedisktransmittedtothecorona(Kotovetal.2001; above 10 keV for the MOS, and in the range 10 – 12 keV Arévalo&Uttley 2006; Uttleyetal. 2011). Comparing the for the pn. We found a significantly elevated background evolutionofthelagswithFourierfrequencyfromalltheAGN count rate during the beginning and end of the observation, which after we filtered out yielded a continuously sampled [email protected] datastream(importantfortiminganalysis)of92ks. 2 Cackettetal. 10 TABLE1 SPECTRALFITSTOESO113- G010 V−1 1 ke Parameter Value unts s −1 0.1 NEeHd(g1e0(2k0eVcm)- 2) 20..7846±(fix0e.0d1) alized co 0.01 σEτmlliinnaxee((kkeeVV)) 060...05226(+-fi±00x..00e035d.0)1 orm Nline (1.7±0.5)×10- 6 N Eline(keV) 6.97(fixed) 10−3 σline(keV) 0.0(fixed) Nline (2.4±0.5)×10- 6 P.L.index,Γ 2.36±0.01 1.1 P.L.norm (1.51±0.01)×10- 3 Ratio 1 SIEnpmcilniisnspaiavtirioatynm,ienited(dre,exag) 6086..94±+-9008..219- 0.002 0.9 Feabundance(Fe/solar) 1.0(fixed) 0.8 0.5 1 2 5 RIoenflizioantixonnopramraamlizeatetiro,nξ(ergcms- 1) (18.5.0+- 00±..811.6)×10- 5 Energy (keV) χ2ν,dof 1.03,154 FIG.1.— XMM-Newton/pn spectrum of ESO 113- G010 fitted with a power-law (blue, dashedline), blurredreflection(dotted, redline)andtwo NOTE.—Wefitthemodelphabs*zedge*(power-law+zgauss+zgauss+ narrowGaussianemissionlines(green,dot-dashedline). kdblur⊗relionx),andfixz=0.0257inallrelevantmodelcomponents. Spectra and lightcurves were extracted from an 800-px 2.2. Frequency-dependentlags radius circular region, with the background extracted from nearby source-free region of the same size to the NE of the We first look for lags between the soft and hard bands source forthe pn and to the SE for the MOS detectors. The as a function of Fourier-frequency in a similar manner as responses were generated with rmfgen and arfgen, and the has been done for all the other sources (see references in spectragroupedtoaminimumof25countsperbinandamin- section 1). The lightcurves exhibit significant variability imumbinwidthof1/3oftheFWHMspectralresolutionata (see Porquetetal. 2007, and their figure 1). The lags be- givenenergyusingtheSAS‘specgroup’tool. tweenthesoftandhardbandsaredeterminedfromthecross- spectrumfollowingthestandardFouriertechniquedetailedin 2.1. Spectralanalysis Nowaketal.(1999). Briefly,wecalculatethecross-spectrum foreachpairofsoftandhardlightcurves,averagingthecross- We fit the pn spectrum in XSPEC v12, in order to deter- spectrum in frequency bins and also averaging the cross- minetheenergieswherethepower-lawandreflectedcompo- spectra from multiple detectors. At a given frequency, f, nentsaremostdominant.Suchaprocedureallowsustomax- the argument of the cross-spectrum gives the phase differ- imizethesensitivitytofindingalagbetweenthesetwocom- ence between the Fourier transforms of the two lightcurves. ponents by isolating lightcurves where the relative strength Thisphasedifferencecanthenbeconvertedto atime lagby ofeachcomponentisgreatest. Themodelconsistsofanab- dividing by 2πf. We therefore determine time lags in each sorbed power-law plus a blurred reflection model using the frequencybinbydividingtheaveragephaseinthebinbythe relconv convolution kernel (Dauseretal. 2010) and reflionx middle-frequencyofthelogarithmicbin.Wedetermineerrors reflectionmodel(Ross&Fabian2005). Weuseanunbroken inthelagsfollowingequation16ofNowaketal.(1999). power-lawemissivityin relconv. Inadditiontothis, thedata The lags between the soft and hard bands are shown as a alsorequireanabsorptionedgeat0.86keV(restenergy),con- functionofFourierfrequencyinFigure2. Thelagisdefined sistentwith an OVIIIedgeand two narrowemission linesat such that a positive lag refers to the hard lightcurve lagging approximately6.5and6.97keV(seePorquetetal.2007,fora the soft(hence‘hard’lag), and a negative(or‘soft’ lag) im- detaileddiscussionoftheFelinecomplexinthisobject). We plies the soft lightcurve lagging the hard. We find a lag of show the best-fitting modelfor the pn spectrumin Figure 1, - 325±89 s in the frequency range (2- 3)×10- 4 Hz. The andseefigure7ofPorquetetal.(2007)foraratiotoasimple evolutionof the lags with frequencyshows a positive (hard) power-law.Thespectrumdisplaysastrongsoftexcess,which canbewell-fitbytheblurredreflectionmodel(seeTable1for lagatthelowestfrequencies(.1.5×10- 4 Hz)transitioning spectral fitting parameters). The best-fitting spin parameter to a negative (soft) lag at higher frequencies, (2- 3)×10- 4 indicatesa maximally spinningblack hole. While the broad Hz, and then becoming a zero lag at the highest frequen- Felineinthisobjectisnotespeciallyconstrainingonthespin, cies (&4×10- 4 Hz). This evolution follows the same gen- itisthesmoothsoftexcessthatrequiresamaximallyspinning eralshapeashasbeenseenintheothersourceswhereasoft blackhole. laghasbeendetected. Onebinatapproximately4×10- 4Hz Based on the shape of the spectrum, we choose the 0.3 – shows a return to a positive lag. However, this bin is only 0.9keVenergyrangeforthesoftbandlightcurve,wherethe 2.2σfromzero,andslightlylargerfrequencybinsreducethis reflectioncomponentisstrongest(relativetothepower-law), significancefurther. and the 1.5 – 4.5 keV range for the hard band lightcurve, We model the frequency-dependent lags using two sim- where the spectrum is dominated by the power-law compo- ple transfer functions - a top-hat transfer function to fit the nent. Lightcurvesintheseenergybandswereextractedusing soft (negative) lags and a power-law transfer function to fit evselectandepiclccorr,and20sbinning. the hard (positive) lags, in a similar fashion to Zoghbietal. LagsinESO113-G010 3 FIG.2.— Frequency-dependent lags inESO113- G010 between the 0.3 – 0.9 keV and 1.5 – 4.5 keV lightcurves. The lag is defined such that a positivelagmeansthatthehardlightcurveleadsthesoft.Thesolidblackline isthebest-fitting transferfunction, comprisedoftwocomponents –atop- hattransferfunction(bluedashedline)andpower-lawtransferfunction(red dashedline). FIG.4.— (a)Energy-dependentlagsforESO113- G010forthefrequency range(2- 3)×10- 4Hz,calculatedusingthepndataonly.Thelagsarearbi- trarilyshiftedsothattheminimumlagiszero. Itistherelativelagbetween eachbandthatisimportant.(b)Unfoldedcovariancespectrum(pndataonly) forESO113- G010inthesamefrequencyrange((2- 3)×10- 4 Hz)asthe energy-dependentlagsinthetoppanel. 2.3. Energy-dependentlagsandthecovariancespectrum FIG.3.—CoherenceforESO113- G010betweenthe0.3–0.9keVand The energy-dependenceof the lags at a given Fourier fre- 1.5– 4.5keV lightcurves. The lightcurves are coherent to approximately quencycanhelpidentifythenatureandoriginofthelags(e.g. 8×10- 4Hz. Zoghbietal. 2011; Karaetal. 2012a,b). Here, we follow a similarproceduretoZoghbietal.(2011). Briefly,weextract lightcurves in narrow energy bands using the pn data only. (2011)andEmmanoulopoulosetal.(2011). Wefindthatthis Lags are then calculated between each narrow energy band model readily reproduces the lags. Our best-fitting model and the reference band. For the reference band we use the is shown in Fig. 2, which has the top-hat transfer function lightcurvefromtheentireenergyrange(0.2–10keV)exclud- extending from 0 to 2390 seconds. The power-law trans- ingthecurrentnarrowenergyband(toavoidcorrelatedPois- fer function is not well constrained as the slope and nor- sonnoise).Thisensuresahighsignal-to-noiseratiointheref- malization are degenerate. However, any combination of erenceband.AsdiscussedbyZoghbietal.(2011)theslightly those parameters gives the same parameters for the top-hat differentreferencelightcurveused for eachband leadsto an transfer function. In lag-frequency space the model for the insignificant systematic error in the lag. In Figure 4(a) we hard lags is consistent with a power-law with slope of ap- showthelagspectrumforthefrequencyrange(2- 3)×10- 4 proximately - 1.7. This is slightly steeper than the slope of Hz wherethe soft lag is detected. Note thatit is the relative - 1 seen in other AGN (Vaughanetal. 2003; McHardyetal. lagsbetween the bandsthatare important(the lags are arbi- 2004; Arévaloetal. 2008), though a slope of - 1 would im- trarilyshiftedsothattheminimumlagseeniszero). Aswas plyalagofapproximately100sat10- 3 Hz,lyingwithinthe showninthefrequency-dependentlags,itcanbeseenthatthe 95.4%confidenceintervalderivedfromthedata. averagelaginthesoftbandislaggingbehindthehardband.It We checkthereliabilityofthe lagsbystudyingthe coher- alsoshowsthatthelagislargestinthe0.4–0.8keVrange.If enceasafunctionoffrequency,followingthemethodologyof wemodifytheenergybandsforthesoftlightcurvesslightlyto Vaughan&Nowak(1997). ThecoherenceisshowninFig.3, 0.4–0.8keVthisincreasesthemagnitudeofthelagobserved andremainshighoverthefrequencyrangewherethelagsare inthe(2- 3)×10- 4Hzrangeslightlyto- 403±83s. detected,uptoapproximately8×10- 4Hz,wherethePoisson It is interesting to note that the general shape of the lag- noise levelin the hard band lightcurveis reached. The high energy spectrum is quite similar to both 1H 0707- 495 and coherenceindicates thata high fractionof the lightcurvesin IRAS 13224- 3809(see Karaetal. 2012a, for a comparison eachbandiscorrelatedandhenceonelightcurvecanbepre- ofthese two objects),wherethe lagsare seento peakwhere dictedfromtheother. thesoftexcessisstrongest,andthendropdowntoaminimum 4 Cackettetal. at3–4keVandincreaseagainattheFeKline.Obviously,we give the Hα line FWHM for ESO 113- G010 as 2000 km donothavethesignal-to-noiseinordertodetectasignificant s- 1. FromtheiropticalspectrumofESO113- G010,weesti- increaseinthe lagabove4keV (thelagmeasuredin the5 – matethefluxdensityat5100Åas fλ=1×10- 15ergcm- 2s- 1 1E0SOke1V1r3a-nGge01is03is05th±e5sa2m2es.),but,thegeneralshapeweseein Å(5-110.0UÅs)in=g7z.5=×0.10024527eragnds- s1t.anWdeartdhecnosumseoltohgeyG,wreeengeet&λHLλo We also calculate the covariance spectrum (see Wilkinson&Uttley 2009) for ESO 113- G010 from the (2005) relationship between L(5100Å) and LHα (their equa- pn data, following the methodology of Uttleyetal. (2011). tion1)alongwiththeirupdatedscalingwithHαFWHMfrom Greene&Ho (2007) (their equationA1) to estimate a black The covariance spectrum shows the strength of correlated variability between a given energy band and the reference hole mass of approximately7×106M⊙, completelyconsis- band, and hence gives the spectral shape of the correlated tentwiththeestimatefromPorquetetal.(2007). variability. We calculate it in a givenfrequencyrange using The lag of - 325 s corresponds to 9.4 Rg/c for a mass of the frequency-averaged cross spectrum in that range. The 7×106M⊙,thoughdilutionofthelagbyanyhardlagcompo- resultant covariance spectrum is proportional to the photon nent,andthezerolagcomponentthatarisesduetoreflection count rate, and so we rebin the pn response to match the andpower-lawcomponentsbeingpresentin bothbands(see energy binning of the lightcurves. Using the rebinned further discussion of dilution effects in Zoghbietal. 2011; response, we can fit it in XSPEC. The covariance spectrum Karaetal.2012a,b),meansthattheintrinsicreverberationlag (unfoldedusingapower-lawwithindex0andnormalization will be larger than this. Determining the dilution effects is of 1) is shown in Fig. 4(b). A simple power-law with index non-trivialgiventhatitdependsontherelativestrengthsofthe Γ = 2.1±0.2 fits it well, though given the signal-to-noise variablecomponents,however,fittingofthelagswithtransfer (the average fractional uncertainty is 35%), we cannot rule functions,suggeststhattheintrinsiclagmaybecloserto800s outmorecomplexmodels. Notethatthe statistics in the 5 – (see the blue line in Fig. 2). Note, however, an added com- 10 keV range are too low to calculate the covariance, as is plicationisthattheobservedlagsarenotadjustedforgravita- alsodemonstratedbythe largeuncertaintyforthelag inthis tional(Shapiro)timedelayswhichwouldactintheopposite energyrange. sense to dilution effects. Comparing with the soft lag mag- nitude – black hole mass relation of DeMarcoetal. (2012), 3. DISCUSSION thelagliessignificantlyabovethescalingrelationandwould We have founda lag of - 325±89secondsin the variable be more consistent with a black hole mass a factor of a few AGN, ESO 113- G010in the frequencyrange (2- 3)×10- 4 higher. DeMarcoetal.(2012)alsopresentthescalingofthe Hz. Moreover, the evolution of the lag over the full fre- frequencywherethesoftlagisobservedwithblackholemass quencyrangeofthelightcurveshowshardlagsonthelongest (the frequency decreases with increasing black hole mass). timescales changing to soft lags on intermediate timescales Thefrequencywherethesoftlagisobserved(approximately and zero lag on the shortest timescales. This evolutionary 2.5×10- 4Hz)isslightlylowerthanexpectedfromthescal- trendisthesameasinthegrowingnumberofAGNwheresoft ing,againconsistentwithablackholeafewtimesmoremas- lagshavebeenobserved.Itsuggestscommonphysicalmech- sive. However,itisimportanttonotethatscatterisexpected anismsarepresentinalltheseAGNleadingtohardandsoft in these scaling relations, especially given that one source, lagsondifferenttimescales. Thesoftlagcanbeinterpretedas IRAS13224- 3809hasasoftlagthatisseentobothincrease duetoreverberationfromtheinneraccretiondisk,wherethe inmagnitudeanddecreaseinfrequencybyafactorofapprox- lagarisesduetolight-traveltimebetweenthesourcepower- imately3inthehighfluxstatecomparedtothelowfluxstate lawemissionandtheaccretiondiskwherethereflectedcom- (Karaetal.2012a). Thus,flux-dependencecouldaddsignif- ponent arises. Such a model predicts that the soft lag mag- icantscattertotheselagscalingrelations. Moreover,several nitude and frequencywhere it is observedshould both scale of the soft lags detected in DeMarcoetal. (2012) occur at with black hole mass, as the characteristic size-scale for the thelowestfrequenciessampledbythelightcurveandthereal system is simply set by GM/c2. Recently, DeMarcoetal. minimummaywelloccuratevenlowerfrequenciesnotsam- (2012) observedsuchscaling with black holemass using15 pled by the currentdata. This too can add scatter to the lag AGNwheresignificantsoftlagsweredetected. scalingrelations. We can compare the soft lag observed in ESO 113- G010 Another cause of scatter in the DeMarcoetal. (2012) re- withwhatisexpectedfromtheDeMarcoetal.(2012)black lations is the black hole spin. Black hole spin changes the hole mass scaling. But first, we should consider the black locationoftheinnerdiskradius(Bardeenetal.1972;Thorne holemassestimateforESO113- G010. Porquetetal.(2007) 1974), thus objects with a maximally-spinning Kerr black estimate the black hole mass in ESO 113- G010 using the hole will have shorter lags than objects with a non-spinning mass-luminosity-timescalerelation (McHardyetal. 2006) to (Schwarzschild)blackhole(assumingtherestofthegeome- be in the range (0.4- 1.0)×107 M⊙, with the range due to tryisunchanged).Therefore,non-spinningblackholescould theuncertaintyinthebolometricluminosity.Asanadditional beabovetheblackholemass-lagscalingrelation,withoutre- check on the black hole mass estimate, we use the optical quiringamoremassiveblackhole. However,ourspectralfit- spectrum of Pietschetal. (1998) to get a separate estimate. tingofESO113- G010implyaninnerdiskradiusconsistent Opticalreverberationmappinghasestablishedascalingrela- with a maximallyspinningblack hole, suggestingthatblack tionship between the broad-line region radius and AGN lu- holespinisnottheoriginoftheoffsetfromthescalingrela- minosity (Kaspietal. 2000; Bentzetal. 2006, 2009) which tions. allows single-epoch black hole mass estimates when com- binedwiththebroadlinewidth. Whiledeterminedusingthe Hβ emission line, this method has been modified for vari- ACF thanks the Royal Society for support. We thank the ous other emission lines too, for instance using the stronger refereefor helpfulsuggestionsthat have helped improvethe Hα line (Greene&Ho 2005, 2007). Pietschetal. (1998) manuscript. LagsinESO113-G010 5 REFERENCES Arévalo,P.,McHardy,I.M.,&Summons,D.P.2008,MNRAS,388,211 McHardy,I.M.,Koerding,E.,Knigge,C.,Uttley,P.,&Fender,R.P.2006, Arévalo,P.,Papadakis,I.E.,Uttley,P.,McHardy,I.M.,&Brinkmann,W. Nature,444,730 2006,MNRAS,372,401 McHardy,I.M.,Papadakis,I.E.,Uttley,P.,Page,M.J.,&Mason,K.O. Arévalo,P.,&Uttley,P.2006,MNRAS,367,801 2004,MNRAS,348,783 Bardeen,J.M.,Press,W.H.,&Teukolsky,S.A.1972,ApJ,178,347 Miller,J.M.2007,ARA&A,45,441 Bentz,M.C.,Peterson,B.M.,Netzer,H.,Pogge,R.W.,&Vestergaard,M. Miller,L.,Turner,T.J.,Reeves,J.N.,&Braito,V.2010,MNRAS,408, 2009,ApJ,697,160 1928 Bentz,M.C.,Peterson,B.M.,Pogge,R.W.,Vestergaard,M.,&Onken, Miyamoto,S.,&Kitamoto,S.1989,Nature,342,773 C.A.2006,ApJ,644,133 Nowak,M.A.,Vaughan,B.A.,Wilms,J.,Dove,J.B.,&Begelman,M.C. Dauser,T.,Wilms,J.,Reynolds,C.S.,&Brenneman,L.W.2010,MNRAS, 1999,ApJ,510,874 409,1534 Papadakis,I.E.,Nandra,K.,&Kazanas,D.2001,ApJ,554,L133 DeMarco,B.,Ponti,G.,Cappi,M.,etal.2012,MNRAS,submitted, Pietsch,W.,Bischoff,K.,Boller,T.,etal.1998,A&A,333,48 arXiv:1201.0196 Porquet,D.,Reeves,J.N.,Uttley,P.,&Turner,T.J.2004,A&A,427,101 DeMarco,B.,Ponti,G.,Uttley,P.,etal.2011,MNRAS,417,L98 Porquet,D.,Uttley,P.,Reeves,J.N.,etal.2007,A&A,473,67 Emmanoulopoulos,D.,McHardy,I.M.,&Papadakis,I.E.2011,MNRAS, Reynolds,C.S.,&Nowak,M.A.2003,Phys.Rep.,377,389 416,L94 Reynolds,C.S.,Young,A.J.,Begelman,M.C.,&Fabian,A.C.1999,ApJ, Fabian,A.C.,Zoghbi,A.,Ross,R.R.,etal.2009,Nature,459,540 514,164 Fabian,A.C.,etal.2012,MNRAS,submitted,arXiv:1208.5898 Ross,R.R.,&Fabian,A.C.2005,MNRAS,358,211 Greene,J.E.,&Ho,L.C.2005,ApJ,630,122 Thorne,K.S.1974,ApJ,191,507 —.2007,ApJ,670,92 Tripathi,S.,Misra,R.,Dewangan,G.,&Rastogi,S.2011,ApJ,736,L37 Kara,E.,Fabian,A.C.,Cackett,E.M.,Miniutti,G.,&Uttley,P.2012a, Uttley,P.,Wilkinson,T.,Cassatella,P.,etal.2011,MNRAS,414,L60 MNRAS,submitted Vaughan,B.A.,&Nowak,M.A.1997,ApJ,474,L43 Kara,E.,Fabian,A.C.,Cackett,E.M.,etal.2012b,MNRAS,inpress, Vaughan,S.,Fabian,A.C.,&Nandra,K.2003,MNRAS,339,1237 arXiv:1210.1465 Wilkinson,T.,&Uttley,P.2009,MNRAS,397,666 Kaspi,S.,Smith,P.S.,Netzer,H.,etal.2000,ApJ,533,631 Zoghbi,A.,&Fabian,A.C.2011,MNRAS,418,2642 Kotov,O.,Churazov,E.,&Gilfanov,M.2001,MNRAS,327,799 Zoghbi,A.,Fabian,A.C.,Reynolds,C.S.,&Cackett,E.M.2012, MNRAS,422,129 Zoghbi,A.,Uttley,P.,&Fabian,A.C.2011,MNRAS,412,59