Analysis of X-ray spectral variability and black hole mass determination of the NLS1 galaxy Mrk 766
aa r X i v : . [ a s t r o - ph . H E ] N ov Analysis of X-ray spectral variability and black hole massdetermination of the NLS1 galaxy Mrk 766.
S. Giacchè , R. Gilli and L. Titarchuk Max-Planck-Institut für Kernphysik, Saupfercheckweg 1, 69117 Heidelberg,Germany Dipartimento di Fisica, Università di Ferrara, Via Saragat, 1, 44100 Ferrara, Italy INAF – Osservatorio Astronomico di Bologna, Via Ranzani 1, 40127 Bologna,ItalyNovember 04, 2013
Abstract
We present an
XMM-Newton time-resolved spectral analysis of the Narrow Line Seyfert 1galaxy Mrk 766. We analyse eight available observations of the EPIC-pn camera taken betweenMay 2000 and June 2005 in order to investigate the
X-ray spectral variability as producedby changes in the mass accretion rate. The 0 . − keV spectra are extracted in time binslonger than 3 ks to have at least 3 × net counts in each bin and then accurately trace thevariations of the best fit parameters of our adopted Comptonisation spectral model. We testa bulk-motion Comptonisation (BMC) model which is in general applicable to any physicalsystem powered by accretion onto a compact object, and assumes that soft seed photons areefficiently up-scattered via inverse Compton scattering in a hot and dense electron corona.The Comptonised spectrum has a characteristic power-law shape, whose slope was found toincrease for large values of the normalisation of the seed component, that is proportional tothe mass accretion rate ˙ m (in Eddington units). Our baseline spectral model also includes awarm absorber lying on the line of sight and radiation reprocessing from the accretion disk orfrom outflowing matter in proximity of the central compact object. Our study reveals that thenormalisation-slope correlation, observed in Galactic Black Hole sources (GBHs), also holdsfor Mrk 766: variations of the photon index in the range Γ ∼ . − . m , as observed in X-ray binary systems. We finally applieda scaling technique based on the observed correlation to estimate the BH mass in Mrk 766.This technique is commonly and successfully applied to measure masses of GBHs, and this isthe first time it is applied in detail to estimate the BH mass in an AGN. We obtain a valueof M BH = 1 . +1 . − . × M ⊙ that is in very good agreement with that estimated by thereverberation mapping. The study of the
X-ray spectral properties of accreting compact objects is crucial in modernastronomy to understand the physics of the accretion process and to investigate the distinctivefeatures of these objects. The main questions which need to be answered are whether thereare remarkable differences in the accretion onto Black Holes (BHs) and Neutron Stars (NSs)in
X-ray
Galactic Binary Systems, whether there is a unified accretion scheme involving both alactic Black Hole sources (GBHs) and Active galactic Nuclei (AGNs) and whether, in thisscheme, AGNs show the same variability patterns observed in GBHs.Focusing on the accretion mechanism onto GBHs, the observed phenomenology is usually de-scribed in terms of BH state classification. A BH transient going into outburst leaves the quiescent state and enters a low-hard state , i.e. a low-luminosity state whose energy spectrumis dominated by a Comptonisation component combined (convolved) with a weak thermalcomponent. The source might then evolve towards a high-soft state or very high-soft state ,characterised by high luminosity and dominant thermal component . For Galactic Black Holesources (GBHs) in X-ray
Binary Systems, the relationship between timing, spectral and massaccretion properties has been intensively studied in the
X-ray energy window [35, 36, 32] witha generic Comptonisation model (BMC) which consistently convolves a black body (BB) spec-trum originating from the accretion disk and the Green function of the Comptonisation process(which is a broken power-law) that takes place in the hot and dense electron corona (ComptonCloud, CC) surrounding the central black hole. Despite the name (bulk motion Comptonisa-tion) that could sound a little bit confusing, the BMC broken power-law is a generic kernelwhich is valid for any kind of up-scattering, both thermal or bulk motion [13]. Accordingto this Comptonisation model, the spectral evolution undergone by objects powered by theaccretion mechanism is driven by the variations in the mass accretion rate ˙ m , expressed inunits of the Eddington rate ˙ M E = L E /c . The power emitted by the accreting system viathe accretion mechanism is then L = ˙ M c η ( r ) ∝ ˙ mmη ( r ), where m is the mass of the compactobject in solar mass units and η ( r ) is the radiative efficiency at distance r to it. The increaseof ˙ m implies a rise of the soft photon supply from the innermost regions of the accretiondisk (emitting BB-like radiation) that efficiently cools down and shrinks the CC via Comptonscattering, softening in turn the resulting spectrum. For ˙ m ≫ − N BMC observed in GBHs, where Γ is the intrinsicphoton index of the emitted power-law and N BMC is the normalisation of the BMC model.In particular for BH sources, the correlation sometimes shows a plateau at low values of bothΓ and accretion rate. For higher values of ˙ m , a slope becomes steeper which is dictated bythe soft photons cooling efficiency when ˙ m increases ([45, 31]). Thus the spectral transitionfrom the low-hard state to the high-soft state takes place, which is followed by a saturationof the photon index for large values of ˙ m . In fact, the photon index is an inverse of theComptonisation parameter Y = κN sc that describes the efficiency of the Compton scattering,where κ is the mean energy gain per scattering and N sc is the number of scattering events( κ = 4 kT /m e c for normal thermal Comptonisation). [36] showed that when the bulk-motionis established κ ∝ /τ and N sc ∝ τ , where τ is the optical depth. Because of this the Comp-tonisation parameter becomes constant and so does α ∝ /Y (Γ = α + 1, see equation B7 in[36], hereafter ST09, for details). The final saturation ( very high-soft state ) usually occurs inthe range Γ ∼ . − −
250 keV [37]. Given that thebulk-motion and the related saturation of the photon index can only establish in presence ofthe event horizon of black hole sources, the saturation plateau in the correlation pattern is aconclusive piece of evidence on the nature of the central compact object. The photon indexsaturation has been observed for example for the GBHs Cyg X-1, XTE J1550-564 and XTE here we described only the main observed spectral states, for a complete review see [29]. we follow the definition used in [44] and [48]. Note that this differs from the definition given in [34], commonlyused in AGN literature, by a factor of η . Therefore in our notation λ = L/L E = η ˙ m . X-ray observation. The final purpose of our analysis isinstead to use a long
X-ray monitoring of an extragalactic source to verify whether the highvariability, shown by AGNs on every observed time-scale (down to few hundreds seconds), isactually driven by changes in ˙ m . In addition, we aim to perform an estimate of the mass ofthe SMBH powering an AGN measuring the Γ − N BMC correlation and scaling it to the samerelation obtained for a reference GBH.In order to carry out this analysis we selected Mrk 766, a low mass ( M BH ∼ . × M ⊙ , [3])Narrow Line Seyfert 1 galaxy that has undergone an intense X-ray monitoring by the
XMM-Newton observatory from May 2000 to June 2005, with eight observations lasting at least30 ks giving an overall monitoring lasting ∼ X-ray variability, steep
X-ray spectra and mass accretionrates close to the Eddington value [12]. Thus, they constitute a perfect target to sample avariety of spectral states and possibly see the saturation of the intrinsic photon index. Inaddition, the large exposure time and the high throughput in the 0 . −
10 keV energy bandof
XMM-Newton provide a good photon statistics to well constrain the spectral parametersof the applied models. This is crucial to perform a consistent analysis given the complexityof the AGNs spectra. Usually, for GBHs the BMC model is able to account for almost thewhole 1 −
100 keV spectrum. However, the power-law description of AGN spectra is only goodto the first order, since these sources usually show additional components in the 0 . −
10 keVenergy band, such as a soft excess, a spectral hardening at E & ComptonHump ), as well as absorption and emission features.The aim of the present work is to sample the intrinsic
X-ray spectral changes in order topopulate the Γ − N BMC correlation.
Given the limited band available with XMM-Newton , itwill not be possible to study the high energy cut-off expected in the BMC scenario and thus wewill just neglect the behavior of this spectral feature in the present work . The paper is organizedas follows. The
XMM-Newton observations and the analysis of the light curves, namely thecount rate in specific energy bands as a function of time, are described in Sect. 2, Sect. 3 dealswith the criteria that led to the composition of the time intervals and with the subsequenttime resolved spectral analysis that we performed on our sample. In Sect. 4 we discuss theresults obtained with the baseline spectral model and we present the results of the scaling ofthe mass of Mrk 766, while in Sect. 5 we draw our conclusions.
We only consider EPIC-pn data. All observations were performed in small-window mode.The EPIC-pn data reduction was performed using the Science Analysis Software (SAS) version11.0, following the standard pipeline suggested by the
XMM-Newton
Science Operation Centre(SOC) in the SAS ’threads’ .The pipeline comprises several steps, from the filtering from flaring particles backgroundto the generation of the spectra. The flaring particles background (in particular protonswith E p .
100 keV) affects the observation towards the end of the orbit, when the satelliteapproaches the radiation belt. Such kind of noise is usually very small ( ∼ × − counts/s )compared with the average source count rate that is ∼ counts/s , but a sudden bump usuallyappears at the end of the runs, reaching also ∼ counts/s . In order to discard time intervalsaffected by high flaring levels we set a threshold at the suggested value of 0 . counts/s .Once the event list has been filtered from flares, we selected the extraction radius of thelight curve that maximizes the signal-to-noise ratio (S/N). This procedure entailed in each http://xmm.esac.esa.int/sas/current/documentation/threads/ XMM-Newton observation log of Mrk 766.revolution starting time total netnumber [yyyy-mm-dd] exposure [s] exposure [s]0082 2000-05-20 58835 255400265 2001-05-20 129906 881700999 2005-05-23 95510 652741000 2005-05-25 98910 588821001 2005-05-27 98918 591791002 2005-05-29 95514 614151003 2005-05-31 98918 509521004 2005-06-03 35017 20090The net exposure time already accounts for the corrections due to the detector live time (71% ofthe total frame time in the small-window mode) and to the filtering of flaring background periods. observation an extraction radius of ∼ arcsec , that is slightly larger than the one chosen byother authors for Mrk 766 (for example see [30]). By applying the SAS task ’epatplot’, weverified that none of our observations has to be corrected for pile-up problems, not even thoseperformed during orbit 0256 (see Fig. 1), when the count rate reached the maximum value of48 counts/s .We extracted the light curve in three different energy ranges: the 0 . − . − −
10 keV range (hard band). Thisseparation is useful to study the behavior of the different spectral components in connectionwith the others. In fact, the soft band flux, hereafter F(0.2-0.8), should be related to thethermal emission of the accretion disk and the medium band flux, hereafter F(2-5), should bedominated by the power-law component arising from the soft-photons up-scattered in the hotcorona, while the hard band flux, hereafter F(6-10), is in principle related to the radiationfraction reprocessed in the AGN environment. The study of these three components revealedthat the larger contribution to the total 0 . −
10 keV count rate, hereafter F(0.2-10), alwayscomes from the soft band, that is usually almost one order of magnitude larger than thecontribution of the medium band. In particular, the average values of the count rate for thespecific case of orbit 1001 (plotted in Fig. 1) are 10 − counts/s for F(0.2-0.8), 1 − counts/s for F(2-5) and 0 . − . counts/s for F(6-10), resulting in 17 − counts/s in the total F(0.2-10).Following the procedure presented in ST09, we studied the hardness-ratio light curves F(2-5)/F(0.2-0.8) and F(6-10)/F(2-5). As can be seen in Fig. 1, Mrk 766 underwent remarkableluminosity variations on a time-scale of few hundreds seconds, accompanied by small spectraloscillations around an almost stable state on comparable times. Nevertheless, during orbit0082, as the luminosity increases from ∼ counts/s to ∼ counts/s , we notice an overallsoftening of the spectrum, implying a total variation of the order of 30% of F(2-5)/F(0.2-0.8),while the change in F(6-10)/F(2-5) is less evident. A similar behavior, even of smaller strength,can be observed during orbit 0265 run in 2001, but in the opposite direction: an overall decreaseof the luminosity, that passed from ∼ counts/s to ∼ counts/s in ∼ ks , is accompaniedby an overall spectral hardening visible in the F(2-5)/F(0.2-0.8) hardness-ratio light curve.The situation is quite different during the observations performed between May and June 2005.The luminosity variations mainly occur at almost constant spectral shape. From the lowerpanel in Fig. 1, we notice that both the ’medium-to-soft’ hardness-ratio and the ’hard-to-medium’ hardness-ratio usually oscillate around ∼ .
18. Nevertheless, we noticed the suddendrops of F(2-5)/F(0.2-0.8) down to ∼ .
12 and the unusual peaks of F(6-10)/F(2-5) up to ∼ .
46. It is worth noting that the most remarkable drops and peaks are related to the lowest . - k e V c oun t r a t e [ c oun t s / s ] time [x10 s] s 12.64x10 s 7.53x10 s 7.59x10 s 8.20x10 s 7.53x10 s15.4x10 s orbit 0082 orbit 0265 orbit 0999 orbit 1000 orbit 1001 orbit 1002 orbit 1003 orbit 10043.15x10 s 12.64x10 s 7.53x10 s 7.59x10 s 8.20x10 s 7.53x10 s15.4x10 s orbit 0082 orbit 0265 orbit 0999 orbit 1000 orbit 1001 orbit 1002 orbit 1003 orbit 10043.15x10 s 12.64x10 s 7.53x10 s 7.59x10 s 8.20x10 s 7.53x10 s15.4x10 s orbit 0082 orbit 0265 orbit 0999 orbit 1000 orbit 1001 orbit 1002 orbit 1003 orbit 1004 F ( - ) / F ( . - . ) time [x10 s] F ( - ) / F ( - ) Figure 1: Light curve (upper panel) and hardness-ratio light curves (lower panel) extracted fromthe whole
XMM-Newton observation set. The time bin-size in both panels is 1000 s to make theplots clearer. The vertical dotted lines and the related labels indicate the time gap between oneobservation and the next one. We plot in red the data corresponding to the occultation eventspresented in [30] (see text for discussion). Note that the time is measured from the beginning ofthe XMM-Newton mission. 5 uminosity states observed in orbits 0999 and 1000.
The low luminosity episodes, combined with the behaviour of the hardness-ratio light curves,seem to indicate a lack of photons in the 2 − keV range lasting about 20% of the XMM-Newton observing time. It is likely that this phenomenon is due to Broad Line Region (BLR)clouds crossing our line of sight [30]. The resulting behaviour of the hardness-ratio light curves(see red curves in Fig. 1) is probably caused by the photo-electric absorption occurring in theseclouds, that produces a steepening of the observed spectral shape in the soft band.Since this work is not devoted to the study of the occultation episodes, we limited ourselvesto consider this possible scenario and to neglect the time intervals affected by the eclipses inthe study of the Γ − N BMC correlation.
The study of the hardness-ratio light curves is crucial in determining the time intervals inwhich the spectra must be extracted. The acquisition of one single spectrum integrated overeach observation, as it has been done in [11], would have basically averaged several differentspectral states. On the other hand, this analysis revealed that Mrk 766 have passed throughdifferent states during the
XMM-Newton observations: during orbits 0082 and 0265 it un-derwent overall and slow intrinsic softening and hardening episodes respectively, while during2005 the luminosity changes occurred at almost constant spectral shape, besides the occulta-tion episodes that only caused the variation of the observed spectrum. The distinction betweenintrinsic and observed spectral slope is crucial, in fact for our study we are only interestedin spectral variations that involve the radiation produced in the core of the system. Giventhe fact that it can be cumbersome to provide a correct and unambiguous parametrization ofthese episodes (see [30] for more details) we decided to neglect the BLR clouds eclipses.Thus, we split the whole useful observing time ( ∼ ks ) in different intervals, trying first ofall to sample the overall hardness variations observed in 2000 and 2001, and then to collect asmany different spectral states as we could from the spectral oscillations in 2005. In addition,we aimed to separate the good time intervals from those affected by the occultations episodes.Since the spectral analysis involves models with several free parameters, we need a good pho-ton statistics in each time slice to obtain restrained relative errors (corresponding to the 90%confidence level). Given the brightness of the source, F [2 − keV ] ∼ − erg s − cm − , theprevious criterion is matched with a minimum duration of 3 ks , that entails a minimum of3 × photons in each time interval, the only exception being the final slice extracted fromorbit 1003. The end of orbit 1003 is probably affected by a BLR cloud occultation, that wetried to separate from the rest of the observation creating a dedicated time slice. Unfortu-nately, simultaneously to the eclipse, there is a strong background flare reaching ∼ count/s ,that in that specific time interval constituted more than 10% of the total count rate, so weexcluded this bad time interval from the subsequent analysis (time slice 47 in Table 3 in theAppendix).The procedure we applied led us to the composition of forty-nine time intervals (fifty includ-ing slice 47), at least three from each observation, in which the spectrum was extracted. There-binning of the spectra was performed in order for each energy channel to contain at leasttwenty photons: this allowed us to use the χ statistics to fit the models to the data and toestimate the goodness of the fits. The duration and the total counts of each time intervalare collected in Table 3 in the Appendix. The time resolved spectral analysis consisted in astep-by-step procedure, from the simplest to the most complicated model, and was performedwith the XSPEC fitting package [1], version 12.7.1. .2 2-10 keV power-law fit First of all we fit each time slice with a simple power-law in the 2 −
10 keV range to comparethe observed photon index Γ obs with the photon index measured with other
X-ray missions.Here we refer to Γ as observed because the power-law basically average all the spectral featuresacross the energy range and moreover it does not account for the very first part of the non-thermal emission. In this specific energy range this simple description is on average good, withreduced chi-square χ ν ∼
1. As shown in Fig. 2, the Γ obs variation between 0.95 and 2.1 andthe flux variation in the 2 −
10 keV range correlate with each other in the flux interval between F [2 −
10 keV] ∼ . × − erg s − cm − and F [2 −
10 keV] ∼ . × − erg s − cm − . It is worth I nd e x , Γ ob s Log F [erg cm -2 s -1 ] orbit 0082orbit 0265orbit 0999orbit 1000orbit 1001orbit 1002orbit 1003orbit 1004 Figure 2: Observed photon index Γ obs plotted against the flux in the 2 −
10 keV range. Notice thatthe slices with the lower flux, corresponding to the occultation events (20% of the
XMM-Newton observing time) are also characterised by the flatter power-law Γ obs . . noting that the photon index becomes larger and larger as the flux increases. The upper partof the plot, i.e. for Γ obs > .
6, is in good agreement with the measures of the photon indexperformed in other periods. In fact, [15] found Γ obs ∼ . − . obs ∼ . − .
25 with
Beppo
SAX (1997) and [41], hereafter SP09, found Γ obs ∼ . − . F [2 −
10 keV] ∼ − − × − erg s − cm − with RXTE (2001-2008). In particular theupper part of the plot of Fig. 2 constitutes the low flux tail of the trend found in SP09.On the other hand, the points related to the low-luminosity episodes during 2005
XMM-Newton observations correspond to the occultations described in Sect. 2.1 and differ from the averagebehaviour. In particular, these eight time slices are characterised by low fluxes, consistentwith F [2 −
10 keV] ∼ − erg s − cm − , and by flat power-law slopes Γ obs . .
5. The fact thatΓ obs < . Next we applied the BMC model [44, 48] in the full 0 . −
10 keV range to the forty-one timeslices unaffected by the occultation events. All the parameters of the BMC model were leftfree to vary in all phases of our spectral analysis.
It is worth noting that even though theBMC theory predicts the high energy cut-off due to recoil effect, the BMC model implementedin XSPEC does not include any parameter accounting for this phenomenon. In this sense,it can be applied in an energy range where the recoil effect is negligible, as it is the case forXMM-Newton bandpass.
We also included in this modeling the absorption associated to theinterstellar medium in our own Galaxy with the WABS component. The value of the galacticcolumn density in the direction of Mrk 766 has been fixed to N H = 1 . × cm − [6]. An xample of this fit is shown on the upper panel of Fig. 3. In the 0 . −
10 keV energy rangeboth the ’soft-excess’ below ∼ ∼ . − . −
10 keV band the description of the observed photon distributionis quite good, with χ ν = 1112 . / ∼ . ∼ . − .
74 keV, observed also in MCG-6-30-15 [18], is an open issue and ithas been widely debated in the literature. [15] identified an absorption feature at ∼ .
74 keVconsistent with an absorption K-edge of O V II . On the other hand, [27] claimed that at the red − shift of Mrk 766 ( z = 0 . O V II would be found at 0 .
73 keV,but they found no evidence of such a feature in the Reflection Grating Spectrometer (RGS) onboard
XMM-Newton . Rather, they found an absorption feature at ∼ . O V II
K-edge, it would originate in an absorbing material red − shifted bymore than 10000 km s − , that appears to be in contrast with the current physical scenariosfor a warm absorber gas, such as an outflowing wind. Last, [19] ascribed the feature to arelativistically broadened Ly α emission line of the H-like O V III . In our sample the featureappears with a typical energy E edge = 0 . +0 . − . keV and an optical depth within the range0 . − .
34, thus our analysis does not provide any further evidence to confirm or discardone of the suggested scenarios and so we limit ourselves to include the absorption edge tothe model. The addition of this component improves the fit significantly, with ∆ χ ≫ . − α line using ASCA data, whereas [20] found no strong evidence ofFe emission line in
BeppoSAX data. Thus, we just add a Gaussian emission line in the range6 . − .
97 keV when it is required to improve the quality of the fit. We also try to model theintrinsic absorption of the galaxy hosting the AGN by adding one more WABS componentand leaving the column density parameter free to vary. In all the spectra in which we use thisapproach, the equivalent column density drops below the Galactic column density, thus the fitresults completely insensitive to this parameter and we conclude that the intrinsic absorptionis negligible. This sort of the phenomenological model provides a good description ( χ ν ∼ ∼ . − .
24, indicating thatthe BMC power-law slope is only slightly steeper (∆Γ ∼ .
2) than the observed power-lawfound in the 2 −
10 keV range (this only refers to the points above the dotted line in Fig. 2).The BB color temperature we obtain from the fit oscillates between kT = 7 . × − keVand kT = 9 . × − keV about the average value kT = 8 . × − keV and seems to beunrelated to the variations of the photon index. We then included in our model a Compton reflection component using the PEXRAV model [17]as follows. We only considered the reflected component provided by the PEXRAV model Theother parameters were left bound to their default values, but for the normalisation N pex thatis the only extra free parameter. We fixed the value of the incident power-law to the photonindex resulting from the BMC model, so that the reprocessed photons are those emergingfrom the CC in the BMC scenario. The addition of the PEXRAV component produces onaverage a significant enhancement of the quality of the fit, with ∆ χ > . kT =5 eV ) with respect to the previous modeling, and the power-law steeper (∆Γ ∼ . this is achieved by fixing the parameter rel-refl to minus one. . −
10 keV range extracted from time slice5 in orbit 0265 in 2001 (see Table 4 in the Appendix). Upper panel: the fit model comprises justthe BMC and the galactic absorption. Lower panel: the fit model is improved by the additionof the PEXRAV and the ZXIPCF components. A Gaussian emission line is also included whoseparameters are E = 6 . ± .
19 keV, σ = 0 . +0 . − . keV. contrary, the BMC normalisation remained essentially constant in the whole sample. It isworth noting that when one calls the PEXRAV model, one of the most important parametersis the covering factor R = Ω / π , that describes the portion of sky that is covered by thereflecting/reprocessing medium with respect to the non-thermal radiation source (CC). Inour analysis we fixed it to a negative value to obtain just the reflected component, since theincident one consisted in the BMC power-law. Usually we have that for a geometrically thinaccretion disk the covering does not exceed half of the sky (Ω = 2 π , R = 1). In the mostdiffuse picture of the AGNs structure this holds true unless we are in one of the followingcases:(a) the thickness of the disk increases with the distance to the centre of the system;(b) the strong gravitational field causes the photons traveling in proximity of the eventhorizon (not in the direction of the observer) to be deflected and impinge on the accretiondisk and to be reflected [10]. somewhat different and maybe complementary approach to explain covering factors R > τ &
2) and cold ( T ∼ K) material traveling in theoutward direction from the central object at some fraction of the speed of light is likely tobe responsible for the modification of the source spectrum between ∼
10 keV and ∼
100 keV.Detailed analytical work and Monte Carlo simulations have been carried out by [47] and by[14] respectively, and satisfactory results have been obtained for the microquasar Cyg X-3 andfor the AGN MCG-6-30-15. Similar results have been obtained by [40], according to whichthis process can originate the so-called Compton Hump observed in a number of AGNs [10].In a sample of ten time intervals unaffected by the eclipses, we checked whether the coveringfactor exceeded the critical value. This was done by generating in XSPEC a diagonal responsein the 1 − keV range and running the PEXRAV model with all parameters pegged to thedefault values, but for the photon index Γ, which was fixed to the best-fit BMC index in thecorresponding slice, and the covering factor R and the PEXRAV normalisation N pex . We thentuned R and N pex to obtain the same 1 −
10 keV flux measured with the true (BMC+reflection)applied model. It turned out that
R > N pex & × − photons keV − cm − s − at 1 keV in the ’true’ model. This holdsin particular when Mrk 766 is in the brighter and softer states found in the XMM-Newton observations. The interpretation of this behavior is quite difficult because several processescan contribute to increase the measured covering factor, as we mentioned above. Nonetheless,the energy and profile of the Fe emission line (see Fig. 3) could be a further evidence in favorof the presence of outflowing material. In fact, a broad, weakly red-skewed and blue-shiftedFe emission line is predicted by [38] and [40] to arise in an obscuring wind due to scatter-ing of line photons and fluorescence recombination. This interpretation is alternative to the’relativistic blurring’ occurring in the innermost layers of the accretion disk, whose effects onthe out coming radiation would be completely smeared out by the large number of scatteringevents. The occultation events suggest that nearby the central engine, at least on a scale offew parsec (inferred dimension of the BLR), there is some outflowing material, probably inthe form of clouds or ’comets’, that moves with the speed of the order of some percent of thespeed of light [30]. This in principle could be the environment where the downscattering of
X-ray photons takes place ([47]). Nonetheless, further analysis is required to ascribe with nodoubts the behavior of the parameter R to the down scattering in outflowing wind rather thanto other physical mechanisms.
The last step of our spectral analysis is to apply a self-consistent absorption model to thescenario comprising the BMC model and the reflection/reprocessing mechanism. Besides thestrong feature at 0 . −
10 keV spec-trum of Mrk 766. We replace the absorption edge with the ionized absorption model ZXIPCF[22] which, as opposed to the ABSORI component [8], also accounts for the absorption linesbesides the absorption edges. Again, we fix the value of the incident photon index to theΓ provided by the BMC spectral component and we set the covering factor of the absorberto one. In our idea both the power-law emerging from the Compton Cloud and the photonscoming from the accretion disk (either the thermal radiation or the Compton reflected com-ponent) must pass through the warm absorber before reaching the observer at infinite. Fromthe statistical point of view, we just substitute the two free parameters of the absorption edge(energy and optical thickness) with the column density and the ionization degree of the newcomponent. On the lower panel of Fig. 3 the unfolded spectrum and the resulting final best-fitmodel are shown for slice 5 extracted during orbit 0265. The improvement of the goodness f the fit is relevant in all slices and from the residuals plot the good quality of the fit isevident. The final step of our analysis leads to an average slight increase of the normalisationsof both the BMC and PEXRAV components, as a consequence of the subtraction of flux dueto the absorption. In addition, the intrinsic photon index suffers a further average steepening∆Γ ∼ .
05 and the BB color temperature increases on average by ∆ kT = 5 eV, giving a meanvalue of kT = 8 . +0 . − . × − keV, in good agreement with the results of the BB fit to Mrk766 soft excess performed by [24] and [4].Hence, the final modeling comprises the Galactic absorption, the BMC model, the reflec-tion/reprocessing scenario and a warm absorber. A Gaussian Fe K α emission line is alsoincluded when it is required. The best-fit values of the main parameters of our spectral anal-ysis are collected in Table 4 in the Appendix. Our general results for
X-ray spectral fitting are in good agreement with the ones found inthe literature. Besides the few exceptions constituted by the occultation episodes in 2005,where the source is a factor of ∼ − keV energyrange is F [2 −
10 keV] ∼ − erg s − cm − , corresponding to a luminosity of L [2 −
10 keV] ∼ × ergs − , is the same presented in [15], [4] and [30]. Similarly, in most cases we foundΓ obs ∼ . − .
15, that overlaps the range presented in [41], and confirms the results of [15]and [20], except for those occulted intervals in which flatter photon indexes Γ obs . .
55 areobserved. In our analysis we do not build a model to parametrize the occultation episodesand we thus neglect the related times intervals (see slices 13-20 in table 3 in the Appendix).Our final baseline model provides a satisfactory physical picture of Mrk 766 spectra and con-firms the presence of a warm absorber that was strongly suggested by [51]. In addition, therange spanned by the intrinsic photon index Γ ∼ . − . XMM-Newton observing time allowed us to extractforty-one ’good’ time intervals that describe a number of different spectral states of the source.In Fig. 4 we present the Γ − N BMC correlation obtained from these intervals (data pointsare re-binned in thirteen bins of N BMC in order to underline the magnitude of the correla-tion): a positive correlation is observed, in particular a variation of the BMC normalisation I nd e x , Γ Log BMC normalisation (L/10 erg s -1 )(10 kpc/d) Figure 4: Γ − N BMC correlation obtained excluding the eight time intervals affected by the occul-tation events. We re-binned in thirteen bins the N BMC range corresponding to the forty-one ’good’points to make the correlation between the photon index and the BMC normalisation clearer. of ∆ N BMC ∼ × − implies a change of the intrinsic spectral index of ∆Γ ∼ .
5. We then rgue that the small oscillations observed in the hardness-ratio light curves (see Fig. 1) aredriven by the same physical mechanism producing the long term spectral transitions in binarysystems, namely the variations of the mass accretion rate. We recall, in fact that the luminos-ity, entering the definition of the normalisation of the BMC model, is proportional to ˙ m , asdiscussed in Sect. 1. It is evident from the plot that as the normalisation increases, the spec-trum becomes softer, as observed in a number of galactic X-ray binaries ([45] among others).Such a behavior was somewhat anticipated by our study of the light curves and hardness-ratiolight curves. We recall that the overall increase of the luminosity occurred during orbit 0082is accompanied by an overall spectral softening (see Fig. 1) that is exactly what we wouldexpect in the BMC scenario: as ˙ m increases the soft photon supply from the accretion diskbecomes larger and larger and efficiently cools down the Compton Cloud, that shrinks. Thenumber of up-scattered photons in the hot corona decreases and the spectrum becomes softer.The opposite occurred in orbit 0265: an overall decrease of the luminosity, manifestation ofthe decrease of ˙ m , is related to a spectral hardening since the CC puffs out and the number ofefficiently Comptonised photons increases. This correlation was in a sense anticipated by thebehavior of the photon index as a function of the 2 − keV flux plotted in Fig. 2 and a similarcorrelation has been previously found for Mrk 766 by [41] with RXTE data. Nevertheless, onemust be cautious and use this relation just as an indication of this expected behavior, sincethe existence of a correlation has been proven only between the intrinsic photon index and thenormalisation of the BMC model.When we compared the correlation we found for Mrk 766 to the Γ − N BMC diagram ob-tained for galactic black hole sources (GBHs), we realized that what we work out for Mrk766 is probably just a fraction of the entire spectral transition of our target (see Fig. 5).This seems to be justified by the fact that in systems powered by the mass accretion pro-cess, the physical properties, such as dimensions and time-scales, are ruled by the mass of thecentral object. As a consequence, given that a complete spectral evolution from a low-hardstate to a high-soft state or vice versa for a GBH ( M BH ∼ M ⊙ ) lasts ∼ −
100 days(see ST09) the corresponding variation of the spectral index for a Super Massive Black Hole( M BH ∼ − M ⊙ ) occurs at least in 10 years. In this plausible assumption, with theobservations we have at hand we are essentially investigating the small spectral oscillationsdue to changes of ˙ m about a putative high-soft state of Mrk 766. We also notice that thetransition, towards a harder or softer state, is not smooth. In fact, in 2000 (orbit 0082) wefound N BMC ∼ × − ( L/ erg s − )(10 kpc/d ) and Γ ∼ .
15; in 2001 (orbit 0265) thesource was brighter and softer ( N BMC ∼ × − ( L/ erg s − )(10 kpc/d ) and Γ ∼ . X-ray spectralanalysis of these oscillations.Another important point to be taken into account is the magnitude of the reflected/reprocessedspectral component described by the PEXRAV model. As we mentioned before, the coveringfactor R increases and overcomes the limiting value 1 when Mrk 766 is in its brighter and softerstates, in particular during orbit 0265. Given that the spectrum resulting from the downscat-tering process is quite similar in shape to the one originating from the reflection in an ionizedmedium ([14, 40]), we are tempted to say that we obtain
R >
R > m [28, 12, 7] that is likely to occurin the softer states of our target. The ionization degree of the absorbing material is not highenough (see Table 4 in the Appendix) to hamper the formation of such a wind [49] whosecontribution, that could constitute the ionized absorbing material described by the ZXIPCFspectral component, is strongly suggested by [51] to account for the spectral variability of Mrk
66. The presence of this material would not only explain the spectral rising above ∼ ∼ . ∼ α emission line of O V III at0 . XMM-Newton no conclusive evidence can be provided in favorof one given model, because we are able to observe only the very first part of this phenomenon.It is worth noting that, given the physical characteristics of the BLR clouds absorbing materialas column density and ionization degree, the 0 . − . − X-rays , rather than the source of soft
X-rays . In other words,it seems plausible that the soft photons source (the accretion disk) is more extended than thehard photons source (the Compton Cloud).
The mass scaling technique is completely based on the shape of the Γ − N BMC correlation,that according to ST09 is fit by a functionΓ( N BMC ) = A − B · ln (cid:26) exp (cid:20) − (cid:16) N BMC N tr (cid:17) β (cid:21) + 1 (cid:27) (1)where the meaning of the parameters is the following: coefficient A is a value of the saturationof the intrinsic photon index, B is related to the lower value achievable by Γ, N tr indicates theBMC normalisation value at which Γ starts growing and β provides the slope of the correlation(see Fig. 5). The crucial assumption for this technique to be applied is that different sourcespresenting the same Γ − N BMC correlation undergo the same kind of spectral evolution, the onlydifference being the BH mass to distance squared ratio
M/d that determines the horizontalshift on the correlation plot. In particular, the appropriate reference source must be selectedaccording to the slope of the correlation. In fact, the steepness of the correlation is tightlyconnected to the underlying physical process leading to the saturation of the photon index,namely it is the signature of the temperature of the converging flow. A steep slope testifiesfor an efficient cooling of the Compton Cloud from the soft-photon supply coming from theaccretion disk and vice versa. The assumption that these two sources behave in the same wayonly holds if the two correlations are as similar as possible. Thus, to scale the mass M of atarget source we need to select an appropriate reference source, whose mass and distance areknown, and compare its BMC normalisation N BMC with the one of the target at the samevalue of the intrinsic photon index. If the previous conditions are matched, a simple scalingrelation can be worked out N t N r = M t d t d r M r f − G ⇒ M t = M r N t N r (cid:16) d t d r (cid:17) f G (2)where t stands for the target, r stands for the reference and f G = cosθ r /cosθ t is a geometricalfactor that depends on the respective inclination angles θ of the accretion disks with respectto the line of sight. This factor has to be considered when the accretion process is assumed tooccur in disk-like geometry, while it can be neglected if spherical accretion is assumed. Another I nd e x , Γ Log BMC normalisation (L/10 erg s -1 )(10 kpc/d) Mrk 766J1550-R98aJ1550-R98b4U1630-Ra4U1630-Rb
Figure 5: The forty-one points of the Γ − N BMC correlation obtained for Mrk 766 are plotted alongwith the four reference patterns that are more likely to describe Mrk 766 (reference patterns fromST09 and [33]). point to be stressed is the following: usually one must compare rise (decay) transitions withrise (decay) transitions in order to be sure that the spectral evolution occurs in response toan increase (drop) of the mass accretion rate.As reference sources we selected the objects presented in ST09, which are all
X-ray binariescontaining a BH of known mass and distance. Specifically, for each reference we have athand two possible transition patterns. The available patterns for XTE J1550-564 are tworise transitions occurred in 1998, while for GRO J1655-40 we have a rise episode and a decayepisode, both of them occurred in 2005. For GX 339-4 the patterns are extracted from a decaytransition undergone in 2003 and from a rise transition in 2004. On the contrary, the availablepatterns for 4U 1630-47 are obtained from observations performed between 1996 and 2004 (inFig. 5 the label 4U 1630-Ra refers to
Beppo
SAX data, while 4U 1630-Rb refers to
RXTE data). In Fig. 5 we plotted the forty-one points extracted from Mrk 766 spectra unaffectedby occultation episodes along with the four reference patterns that are more likely to describeour target source. In fact, the correlations of both GRO J1655-40 and GX 339-4 saturate attoo low values of Γ, namely Γ sat = 2 . sat = 2 .
02 for GRO J1655-40 rise 2005 and decay2005, and Γ sat = 2 .
08 and Γ sat = 2 .
14 for GX 339-4 decay 2003 and rise 2004 respectively.This means that the final temperature of the Compton Cloud for these two sources is largerthan for Mrk 766. In principle some problem can arise since we do not know whether theΓ − N BMC correlation for Mrk 766 testifies of a rise or a decay transition. In any case, someeducated guess can solve this issue. We should point out that even if we do not see a completespectral evolution for the target, from 2000 to 2001 Mrk 766 spectrum underwent a softeningfrom Γ ∼ .
13 to Γ ∼ .
35, while from 2001 to 2005 the spectrum essentially hardened backto almost the initial value. Then it seems that the rise and decay patterns of our target arequite similar. According to this remark we fit function (1) to Mrk 766 points with the QDP’ftool’ . Unfortunately, since the spectral sample only covers a small fraction of the correlationpattern we could not leave all the four parameters of the function free to vary. The only wayto get a converging fit was to leave one single parameter free and constrain all the othersto the values of the parameters of the reference patterns. We thus fixed the upper and lowersaturation levels and the slope of the correlation to the values A, B and β of the four referencesrespectively. In addition, in order to better constrain the parameters, we fit function (1) tothe re-binned points presented in Fig. 4, for which the oscillation of both Γ and N BMC areless pronounced. The results of this procedure are provided in Table 2, where it is clear, from http://heasarc.gsfc.nasa.gov/docs/software/ftools/others/qdp/node3.html he value of χ ν that the best reference to be compared with Mrk 766 is 4U 1630-Rb. Table 2: Parameters of the fit performed on Mrk 766 points.reference A B N tr [ × − ] β χ /dofXTE J1550-R98a 2 .
86 1 .
50 1 . +0 . − . .
50 33 . / .
55 1 .
21 2 . +0 . − . .
00 93 . /
94U 1630-Ra 2 .
88 1 .
29 2 . +0 . − . .
64 39 . /
94U 1630-Rb 2 .
40 1 .
29 0 . +0 . − . .
43 16 . / In Fig. 6 we plot Mrk 766 data points with its best fit curve along with the 4U 1630-Rbreference pattern: the black arrow stresses the horizontal shift due to the different mass todistance squared ratio. The small box contains the re-binned points (green) and the averagevalues of the Γ − N BMC correlation computed in 2000, 2001 and 2005 respectively (blue). Mrk766 seems to be properly described by the selected reference pattern and the relatively large χ value ( χ /dof = 16 . /
9) is likely to be due to the oscillations around the average valuesand to the small range of the correlation covered by our data points.Once we select the suitable reference we can proceed with the estimate of the BH mass of Mrk766 with formula (2) as follows (see also Fig. 6). The mass of 4U 1630-47 is estimated to be M r = 9 . ± . M ⊙ by [33]. Then, as far as the distances are concerned, we use d r = 10 kpc for4U 1630-47 [33], while for our target we choose d t = 57 . ± . Mpc provided by NED . Themost uncertain term of the previous relation is the geometrical factor f G . For the referencesource we use θ r = 67 ± ◦ , but for Mrk 766 this parameter is not known with suitableprecision. [50] derived θ ∼ ◦ from the analysis of the Fe K α line profile. Furthermore,the type 1 activity shown by Mrk 766 points towards an almost face-on situation. Hence, toaccount for the lack of a precise measure of the inclination angle, we perform the scaling ofthe BH mass for θ t ≤ ◦ . We used function (1) with the previously mentioned parameters tocompute the BMC normalisation of the reference source N r for Γ = 1 . − .
4, i.e. the valuesof the intrinsic photon index that we find for Mrk 766. Then we used formula (2) for eachpoint and we average over the sample, getting one evaluation of the mass for each value of theinclination angle θ t . The scaling technique that we apply provides an estimate of the centralBlack Hole mass for Mrk 766 of M BH = 1 . +0 . − . × M ⊙ for θ t = 15 ◦ , to be compared with M BH = 1 . +1 . − . × M ⊙ computed with the Reverberation Mapping method by [3]. Theuncertainty of the inclination angle results in an upper limit of M BH = 1 . +0 . − . × M ⊙ for θ t = 30 ◦ , and in a lower limit of M BH = 1 . +0 . − . × M ⊙ for a face-on situation. Theentire confidence range we obtain for the mass of Mrk 766 is then M BH = 1 . +1 . − . × M ⊙ ,where the larger contributions to the uncertainty are given by the errors on the accretiondisks inclination angles (33% and 15% given by θ r and θ t respectively). Besides the issueconcerning the uncertainty, the BH mass estimates performed with the scaling technique andthe reverberation mapping method are in good agreement. The fact that our measure is inthe lower part of the confidence range found by [3] confirms what is usually expected forNarrow Line Seyfert 1 Galaxies, i.e. relatively small BH masses and large values of the massaccretion rate. In particular, using the bolometric correction k [2 −
10 keV] = 38 . +5 . − . presentedin [52], the average luminosity measured in our sample L [2 −
10 keV] ∼ × erg s − and theEddington luminosity L E = 1 . × (cid:16) M BH M ⊙ (cid:17) erg s − we obtain L bol ∼ . × erg s − and λ = L bol /L E ≈
1, that is slightly smaller than λ ∼ . λ support the hypothesis that in the circum-nuclear regions of Mrk 766 a strong radiation-pressure driven outflowing wind rises in response to the increase of the mass accretion rate [7].From Fig. 6 we can also infer that the saturation level of the photon index for Mrk 766, http://ned.ipac.caltech.edu/ I nd e x , Γ Log BMC normalisation (L/10 erg s -1 )(10 kpc/d) Mrk 766fit to Mrk 7664U 1630-Rb
Figure 6: The comparison between the Γ − N BMC correlations for Mrk 766 and for 4U 1630-Rb isshown. The black arrow stresses the fact that the two sources seem to behave the same way, theonly difference being the gap in the BMC normalisation, due to the different value of the M BH /d ratio. In the small box the re-binned points (green) and the average points computed from 2000,2001 and 2005 observations (blue) are plotted. signature of the establishment of the full bulk-motion onto a BH, is Γ ∼ .
4. Unfortunately,the
XMM-Newton observations do not provide any point on the plateau that would conclusivelyprove the index saturation. Among the other
X-ray missions which observed Mrk 766 since1992, ROSAT measured F [0 . − . = 1 . × − erg s − cm − during the ROSAT All SkySurvey [25]. This flux value is larger by a factor of ∼ . − . F [0 . − . = 5 . × − erg s − cm − , corresponding to F [0 . −
10 keV] = 7 . × − erg s − cm − from slice 5, see Table 4 in the Appendix ), butROSAT neither provides the sufficient photon statistics nor it is endowed with the suitableenergy window to perform such an analysis and measure with a good precision the intrinsicphoton index Γ. Nevertheless, this single ROSAT run testifies that Mrk 766 can reach evenhigher luminosity states than observed by XMM-Newton . In addition, the source proved tobe strongly variable in the
X-ray band on a time-scale of 100 ks (see orbit 0265 in Fig. 1) sothat a dedicated campaign of pointed observations providing at least ∼ counts per eachobservation could in principle find the index saturation for Mrk 766. The definition of soft excess corresponds to the increase of flux measured above the underlyingpower-law continuum at E . keV that is usually observed in AGNs spectra. On one handthe nature of this feature can be ascribed to the thermal emission from the innermost layersof the accretion disk that emit a modified BB spectrum ([7] among others). Basic and simplearguments show that the BB temperature is proportional to M − / . If we assume that mostof the disk luminosity is a perfect BB emission coming from within a distance r = r ∗ R g fromthe centre, where r ∗ ∼ − R g = GM/c is the gravitationalradius, according to the Stefan-Boltzmann law we get L (cid:27) πr σ b T (3)where σ b = 5 . × − erg cm − s − K − is the Stefan-Boltzmann constant and 2 πr is thesurface of the two-sided disk. Hence, provided that the luminosity equals some fraction λ of the 0 . − . dummyrsp command in XSPEC. he Eddington luminosity L E , the expected disk temperature turns out to be T exp (cid:27) (cid:16) λL E πr σ b (cid:17) / ∝ M − / λ / . (4)As far Mrk 766 is concerned, this physical scenario is plausible. In fact, performing an order ofmagnitude estimate by considering the central value of our confidence range for the BH massin Mrk 766, the previously estimated λ ≈ r ∗ ∼ −
7, we obtain kT exp ∼ . − . × − keV, that partially overlaps the range spanned by the BB colour temperature provided by ourbaseline model kT fit = 8 . +0 . − . × − keV (see Sect. 3.5).Unfortunately for this interpretation, some problem arises comparing kT exp with the disk tem-perature kT fit resulting from a BB spectral fit of the soft excess for AGNs hosting BHs whosemass is ∼ − M ⊙ : the expected value spans the range kT exp ∼ −
40 eV, whereas theusual fit value is kT fit ∼ . X-ray
Compton reflectionand photo-ionisation on the accretion disk [17], sometimes combined with relativistic blurringof emission lines below 1 keV [10]. The combination of these two phenomena could producea bump above the underlying power-law. [19] claimed that this very scenario produces goodresults when applied to Mrk 766. The recent discovery of soft/negative time lags in Mrk 766spectra by [9] and [5] seems to support the reverberation scenario, as the variations in the soft
X-ray band (0 . − . X-ray band (1 . − t ∼ ks ), whereas it is clear fromthe previous papers that for longer characteristic time-scales ( t & ks ) the soft band drivesthe changes in the hard band. This is consistent with the scenario presented in this paper.In fact, the short time-scales soft/negative lags would be justified by the energy released inthe accretion disk by the X-ray radiation: energetic photons traveling in the inward directionlose energy in the dense medium via photo-absorption and Compton recoil. Basically, theydeposit energy in the accretion disk, indeed increasing its temperature. This energy is thenre-emitted, mainly in the UV and soft
X-ray bands. This mechanism was suggested by [2]to explain the emission from the surface of a normal star in a binary system, but it seemsplausible that it works also in this situation. To strengthen this conclusion, in Fig. 7 we plotthe BB disk temperature against the flux in the 2 − r = 0 .
57 which denotes a strong linear correlation between thetwo quantities. In principle, this mechanism also explains the difference between kT exp and BB t e m p e r a t u r e [ k T / k e V ] Log flux [2-4 keV] [erg cm -2 s -1 ] Figure 7: BB color temperature (in keV) plotted against the flux in the 2 − r = 0 .
57 indicates a strong linear correlation. kT fit that we obtain for Mrk 766. In fact, the simple estimate we performed using equation (4) oes not account for the contribution to the disk temperature given by the energy depositedby the X-ray radiation.Thus, in this plausible scenario, the short-term variations ( t ∼ ks , [5]) in the soft X-ray band, which are dictated by the impingement of high energy photons on the accretion disk,determine the large spread of the points in the Γ − N BMC correlation (see red points in Fig.6). On the other hand, the long-term variations are driven by the soft
X-ray band, which inturn is ruled by the propagation of the mass accretion rate through the disk [16]. This givesrise to the correlation plotted in Fig. 6 (specifically green and blue points).As a future development of the present work, we aim to extend this analysis to other NLS1galaxies, to check whether the correlation we find for Mrk 766 between the photon index Γ andthe normalisation of the BMC model N BMC holds for other AGNs with M BH ∼ M ⊙ . Fur-thermore, the extension of the sample of spectral transitions, both for GBHs and AGNs, wouldbe the key to understand whether there is a finite or infinite number of possible transition pat-terns, which could provide some constrain on the physics at work in accreting compact objects. This section is devoted to a very brief review of some alternative models used in the literatureto describe the spectral variability which is the main focus of this paper. For a complete dis-cussion we address the interested reader to the papers we quoted and to the references therein.[41], hereafter SP09, also find the correlation of the photon index Γ with the dimensionlessmass accretion rate ˙ m combining the results for different AGN sources. In particular, theydemonstrate this correlation using a phenomenological model, power-law plus line plus edgeapplying RXTE data for a number of the AGN sources (Mkr 766, NGC 3227, NGC 5548 NGC5506 and NGC 3516). It is worth noting that SP09 define ˙ m as ratio of the average fluxin the energy range from 2 to 10 keV, F −
10 keV to the Eddington luminosity L Ed which is1 . × M/M ⊙ erg s − for an AGN with a BH mass M . Although the flux from the disk as asource of the soft photons in an AGN should be calculated in the energy range much lower than2 keV, see a typical disk temperature and its dependence on the energy flux in 2-4 keV range inFig. 7. SP09 emphasize that the index vs mass accretion correlation reflects a true/intrinsic correlation between the photon index of the power-law component and accretion rate. Thisstatement is similar to that which we claim in our presented paper. SP09 also point out that itis widely believed that hard X-rays from AGN are produced by the thermal Comptonisation.SP09 claim that the enhancement factor η comp due to thermal Componization depends on thegeometry of the accretion flow while [42] demonstrate that η comp is determined by the prop-erties of the hot plasma and seed photons, namely the spectral index α = Γ −
1, the plasmatemperature kT e and the seed photon temperature kT s . SP09 also suggest the index vs massaccretion, ˙ m correlation can be explained if η comp is proportional to ˙ m . Finally, SP09 concludethat the “observed ”Γ − F −
10 keV ” or Γ − ˙ m , can be explained using the reflection effect withthe constant reflection amplitude R = Ω / π = 1, where Ω is the solid angle constant coveredby the cold material as viewed from the X-ray source if one can assume that the power-lawcontinuum varies in flux and shape.[51], hereafter T07, investigate the origin of the high variability of Mrk 766 with two differ-ent models, both of them relying on a power-law in the range 1 − keV that is constant inslope and variable in normalisation. In the first model a constant scattered component and anionised reflection component are assumed to play an important role and the observed spectrumis the result of the relative strength of directly viewed and reflected/scattered components.The second model mainly ascribes the spectral variability to the presence of complex layersof absorbing material partially covering the cxentral source of radiation. These layers of gas,perhaps arising from an outflowing wind, are free to vary both in covering fraction, ionisationdegree and column density. T07 claim that the most robust description of the spectral vari- bility is given in terms of directly viewed and scattered or absorbed fractions of flux, eventhough it is difficult to understand whether the continuum drives the variations of the absorb-ing material or whether the continuum is intrinsically constant and the observed variabilityis simply a consequence of the changes in the covering fraction. In addition, the degeneracybetween the ionisation degree and the column density makes impossible to understand whichof these parameters is actually responsible for the spectral changes.Another physical process often addressed to explain the spectral variability in AGNs is Comp-ton reflection from ionised or partially ionised material (see for example [10]). As alreadymentioned in Sect. 3.4, according to this scenario the putative accretion disk or the dustytorus are illuminated by the hard X-rays giving rise to a ’reflection’ spectrum dominated byfluorescent K α line from the most abundant elements, particularly iron. One possible draw-back of this model is the fact that sometimes the fraction of sky R = Ω / π occupied by theaccretion disk with respect to the source of X-rays results to be larger than the maximumexpected value R = 1, as we discussed in the previous section. The issue can be solved takinginto account general relativistic effects taking place in proximity of the central BH. In fact,depending on the height of the hard X-ray source above the disk [23], the light bending phe-nomenon deflects photon that would travel to infinite causing them to be intercepted by thedisk, increasing de facto the factor R , or to fall into the hole event horizon. [46] showed that the outflow can be launched from the accretion disk if the local mass accretionrate ˙ M loc is higher than the Eddington one. The disk works a filter that does not allow tosupply ˙ M loc higher than a certain critical value. The resulting Thomson optical optical depthof the outflow τ W can be higher than 1 using typical parameters of the wind and disk. [43]studied an effect of outflow on the emergent spectra from compact objects (NS and BH). Theydemonstrated solving analytically the radiative Fokker-Planck equation that the emergent ironline profile formed in the outflow of the optical depth of order of 1 expanding with the outflowvelocity of 0.05-0.1 of the speed of light c leads to the formation of a broad red-shifted andskewed line feature. Later this result was confirmed by [14] who used Monte Carlo simulationsto investigate the iron line profile formed in the outflow. On the other hand, using XMM- Newton monitoring of Mkr 766, [51] and [30] found an outflow component of the velocityspanning from 0.01 to 0.05c which can lead to the formation of the broad red-skewed iron linesobserved in many Galactic and extragalactic sources (see e.g. a review by [21]).
We study the timing, spectral and accretion properties of the NLS1 galaxy Mrk 766 by exploit-ing an intense
XMM-Newton monitoring from May 2000 to June 2005, for an overall observingtime of ∼ ks ( ∼ ks of effective monitoring). We study the light curves and the hardness-ratio light curves to isolate time intervals corresponding to different spectral states and studythe resulting spectral transition pattern. This led to the selection of forty-nine time sliceslasting at least 3 ks and containing at least 3 × photons where the spectra were extracted.From the time resolved spectral analysis it emerges that Mrk 766 spectrum is satisfacto-rily described by a model comprising the Galactic absorption, a simple Comptonisation model(bulk-motion Comptonisation model BMC currently used for GBHs), a reflection/reprocessingcomponent, and a warm absorber. A Gaussian iron emission line is also included when sta-tistically required. The average 2 −
10 keV measured flux is F [2 −
10 keV] ∼ − erg s − cm − ,corresponding to a luminosity of L [2 −
10 keV] ∼ × erg s − . Twenty per cent of the wholeobserving time is probably affected by BLR clouds occultations occurring during the obser-vations performed in 2005 [30], for which the just mentioned spectral description was notcompletely physically reliable. Hence, we neglected the time intervals related to the eclipses.We observed luminosity increases lasting few thousands seconds that proved to be related to eneral intrinsic softening of the spectrum, as well as luminosity drops connected to spectralhardening in the photon index, in agreement with the theoretical expectations. The patternof the correlation between the slope and the normalisation of the model we work out testifiesof spectral changes about a putative high-soft state of Mrk 766 rather than a real spectraltransition like those observed for GBHs. This seems to be consistent with the fact that forlarge systems like AGNs the expected time-scales for a complete spectral transition ( ∼ years) are much longer than for stellar BHs ( ∼ −
100 days). Nevertheless, the shape ofthe correlation is a conclusive evidence that those small spectral changes are indeed driven bythe same physical process that causes the complete spectral evolutions seen in
X-ray binarysystems, namely the variations of the mass accretion rate ˙ m . We used the reference Γ − N BMC correlation of the GBH 4U 1630-47, for which both mass and distance are known, to derive themass of Mrk 766. The obtained value of M BH = 1 . +1 . − . × M ⊙ is in perfect agreementwith the mass estimate performed with the reverberation mapping method, proving that thisscaling technique is a powerful and reliable tool to estimate the mass of SMBHs in AGNs,provided the suitable quality of X-ray data and a moderate knowledge of the inclination angleof the accretion disk with respect to the line of sight.In addition, the mass estimate allowed us to compare the disk color temperature providedby our baseline model ( kT fit = 8 . +0 . − . × − keV ) with the one computed with a sim-ple model involving a prefect BB emission from the innermost regions of the accretion disk( kT exp ∼ . − . × − keV ), that is slightly smaller but still comparable to the best fitvalue. This would also justify the discovery of soft/negative time lags on t ∼ ks time-scales.The comparison of our target diagram with the reference source also pointed out that thepossible saturation level of the intrinsic photon index for Mrk 766 is at Γ ∼ .
4. The indexsaturation would provide an important piece of evidence on the nature of the compact objectsitting in the centre of AGNs. Such a measure is within reach of a dedicated observationscampaign to Mrk 766.
Acknowledgements
We thank Guido Risaliti and Chris Done for stimulating discussions and suggestions andPiero Ranalli for his support in the installation of the SAS tool. Elena Seifina is gratefullyacknowledged for providing fundamental data on 4U 1630-47. SG would like to thank theIMPRS for Astronomy and Cosmic Physics for the financial support. RG acknowledges supportfrom the Italian Space Agency (ASI) under the contract ASI-INAF I/009/10/0.
A Appendix: details of spectral analysis
We provide the table collecting the fifty time intervals extracted from the
XMM-Newton ob-servations of Mrk 766 from May 2000 to June 2005. For each time slice the initial and finaltime (expressed as detector time), the net exposure and the total amount of counts are given.Furthermore, we present the best-fit values of the main parameters of the applied baselinemodel, along with the values of the reduced χ stating the goodness of the fit. The eighttime intervals affected by the occultation events (slices 13 through 20) and slice 47 have beenneglected. References [1] K. A. Arnaud. XSPEC: The First Ten Years. In G. H. Jacoby and J. Barnes, editors,
Astronomical Data Analysis Software and Systems V , volume 101 of
Astronomical Societyof the Pacific Conference Series , page 17, 1996.
2] M. M. Basko, R. A. Sunyaev, and L. G. Titarchuk. Reflection and reprocessing of X-ray source radiation by the atmosphere of the normal star in a binary system.
A&A ,31:249–263, March 1974.[3] M. C. Bentz, J. L. Walsh, A. J. Barth, N. Baliber, V. N. Bennert, G. Canalizo, A. V.Filippenko, M. Ganeshalingam, E. L. Gates, J. E. Greene, M. G. Hidas, K. D. Hiner,N. Lee, W. Li, M. A. Malkan, T. Minezaki, Y. Sakata, F. J. D. Serduke, J. M. Silverman,T. N. Steele, D. Stern, R. A. Street, C. E. Thornton, T. Treu, X. Wang, J.-H. Woo, andY. Yoshii. The Lick AGN Monitoring Project: Broad-line Region Radii and Black HoleMasses from Reverberation Mapping of H β . ApJ , 705:199–217, November 2009.[4] T. Boller, R. Keil, J. Trümper, P. T. O’Brien, J. Reeves, and M. Page. Detection ofan X-ray periodicity in the Narrow-line Seyfert 1 Galaxy Mrk 766 with XMM-Newton.
A&A , 365:L146–L151, January 2001.[5] B. De Marco, G. Ponti, M. Cappi, M. Dadina, P. Uttley, E. M. Cackett, A. C. Fabian,and G. Miniutti. Discovery of a relation between black hole mass and soft X-ray timelags in active galactic nuclei.
MNRAS , 431:2441–2452, May 2013.[6] J. M. Dickey and F. J. Lockman. H I in the Galaxy.
ARAA , 28:215–261, 1990.[7] C. Done, S. W. Davis, C. Jin, O. Blaes, and M. Ward. Intrinsic disc emission and thesoft X-ray excess in active galactic nuclei.
MNRAS , 420:1848–1860, March 2012.[8] C. Done, J. S. Mulchaey, R. F. Mushotzky, and K. A. Arnaud. An ionized accretion diskin Cygnus X-1.
ApJ , 395:275–288, August 1992.[9] D. Emmanoulopoulos, I. M. McHardy, and I. E. Papadakis. Negative X-ray reverberationtime delays from MCG-6-30-15 and Mrk 766.
MNRAS , 416:L94–L98, September 2011.[10] A. C. Fabian and G. Miniutti. The X-ray spectra of accreting Kerr black holes.
ArXivAstrophysics e-prints , July 2005.[11] M. Gliozzi, L. Titarchuk, S. Satyapal, D. Price, and I. Jang. Testing a Scale-independentMethod to Measure the Mass of Black Holes.
ApJ , 735:16, July 2011.[12] S. Komossa. Narrow-line Seyfert 1 Galaxies. In
Revista Mexicana de Astronomia yAstrofisica Conference Series , volume 32 of
Revista Mexicana de Astronomia y Astrofisica,vol. 27 , pages 86–92, April 2008.[13] P. Laurent and L. Titarchuk. The Converging Inflow Spectrum Is an Intrinsic Signaturefor a Black Hole: Monte Carlo Simulations of Comptonization on Free-falling Electrons.
ApJ , 511:289–297, January 1999.[14] P. Laurent and L. Titarchuk. Effects of Downscattering on the Continuum and LineSpectra in a Powerful Wind Environment: Monte Carlo Simulations, Analytical Results,and Data Analysis.
ApJ , 656:1056–1074, February 2007.[15] K. M. Leighly, R. F. Mushotzky, T. Yaqoob, H. Kunieda, and R. Edelson. The X-RaySpectral Variability of Markarian 766.
ApJ , 469:147, September 1996.[16] Y. E. Lyubarskii. Flicker noise in accretion discs.
MNRAS , 292:679, December 1997.[17] P. Magdziarz and A. A. Zdziarski. Angle-dependent Compton reflection of X-rays andgamma-rays.
MNRAS , 273:837–848, April 1995.[18] K. O. Mason, G. Branduardi-Raymont, P. M. Ogle, M. J. Page, E. M. Puchnarewicz,E. Behar, F. A. Córdova, S. Davis, L. Maraschi, I. M. McHardy, P. T. O’Brien, W. C.Priedhorsky, and T. P. Sasseen. The X-Ray Spectrum of the Seyfert I Galaxy Markarian766: Dusty Warm Absorber or Relativistic Emission Lines?
ApJ , 582:95–104, January2003.[19] K. O. Mason, G. Branduardi-Raymont, P. M. Ogle, M. J. Page, E. M. Puchnarewicz,E. Behar, F. A. Córdova, S. Davis, L. Maraschi, I. M. McHardy, P. T. O’Brien, W. C.Priedhorsky, and T. P. Sasseen. The X-Ray Spectrum of the Seyfert I Galaxy Markarian766: Dusty Warm Absorber or Relativistic Emission Lines?
ApJ , 582:95–104, January2003.
20] G. Matt, G. C. Perola, F. Fiore, M. Guainazzi, F. Nicastro, and L. Piro. The BeppoSAXobservation of Mrk 766.
A&A , 363:863–868, November 2000.[21] J. M. Miller. Relativistic X-Ray Lines from the Inner Accretion Disks Around BlackHoles.
ARAA , 45:441–479, September 2007.[22] L. Miller, T. J. Turner, J. N. Reeves, I. M. George, S. B. Kraemer, and B. Wingert.The variable X-ray spectrum of Markarian 766. I. Principal components analysis.
A&A ,463:131–143, February 2007.[23] G. Miniutti and A. C. Fabian. A light bending model for the X-ray temporal and spectralproperties of accreting black holes.
MNRAS , 349:1435–1448, April 2004.[24] S. Molendi and T. Maccacaro. The variable soft X-ray excess of MKN 766.
A&A , 291:420–424, November 1994.[25] S. Molendi, T. Maccacaro, and S. Schaeidt. Variability of the Seyfert Galaxy MARKAR-IAN:766 in the ROSAT All-Sky Survey.
A&A , 271:18, April 1993.[26] K. Nandra, I. M. George, R. F. Mushotzky, T. J. Turner, and T. Yaqoob. ASCA Obser-vations of Seyfert 1 Galaxies. II. Relativistic Iron K alpha Emission.
ApJ , 477:602, March1997.[27] M. J. Page, K. O. Mason, F. J. Carrera, J. Clavel, J. S. Kaastra, E. M. Puchnarewicz,M. Santos-Lleo, H. Brunner, C. Ferrigno, I. M. George, F. Paerels, K. A. Pounds, and S. P.Trudolyubov. The variable XMM-Newton spectrum of Markarian 766.
A&A , 365:L152–L157, January 2001.[28] D. Proga and T. R. Kallman. Dynamics of Line-driven Disk Winds in Active GalacticNuclei. II. Effects of Disk Radiation.
ApJ , 616:688–695, December 2004.[29] R. A. Remillard and J. E. McClintock. X-Ray Properties of Black-Hole Binaries.
ARAA ,44:49–92, September 2006.[30] G. Risaliti, E. Nardini, M. Salvati, M. Elvis, G. Fabbiano, R. Maiolino, P. Pietrini,and G. Torricelli-Ciamponi. X-ray absorption by broad-line region clouds in Mrk 766.
MNRAS , 410:1027–1035, January 2011.[31] E. Seifina and L. Titarchuk. On the Nature of the Compact Object in SS 433: Observa-tional Evidence of X-ray Photon Index Saturation.
ApJ , 722:586–604, October 2010.[32] E. Seifina and L. Titarchuk. On the Constancy of the Photon Index of X-Ray Spectra of4U 1728-34 through All Spectral States.
ApJ , 738:128, September 2011.[33] E. Seifina, Lev Titarchuk, and Nikolai Shaposhnikov. Black Hole mass determination in
X-ray binary system 4U 1630-47 using scaling of spectral and variability characteristics. submitted to ApJ , April 2013.[34] N. I. Shakura and R. A. Sunyaev. Black holes in binary systems. Observational appear-ance.
A&A , 24:337–355, 1973.[35] N. Shaposhnikov and L. Titarchuk. Determination of Black Hole Mass in Cygnus X-1 byScaling of Spectral Index-QPO Frequency Correlation.
ApJ , 663:445–449, July 2007.[36] N. Shaposhnikov and L. Titarchuk. Determination of Black Hole Masses in Galactic BlackHole Binaries Using Scaling of Spectral and Variability Characteristics.
ApJ , 699:453–468,July 2009.[37] N. Shaposhnikov and L. Titarchuk. Implications of the Power Law Cutoff Evolutionin XTE J1550-564 to the Comptonization Regimes. In
American Astronomical SocietyMeeting Abstracts , volume 216 of
American Astronomical Society Meeting Abstracts ,page
MNRAS , 388:611–624, August 2008.
39] S. A. Sim, D. Proga, R. Kurosawa, K. S. Long, L. Miller, and T. J. Turner. SyntheticX-ray spectra for simulations of the dynamics of an accretion flow irradiated by a quasar.
MNRAS , 426:2859–2869, November 2012.[40] S. A. Sim, D. Proga, L. Miller, K. S. Long, and T. J. Turner. Multidimensional modellingof X-ray spectra for AGN accretion disc outflows - III. Application to a hydrodynamicalsimulation.
MNRAS , 408:1396–1408, November 2010.[41] M. A. Sobolewska and I. E. Papadakis. The long-term X-ray spectral variability of AGN.
MNRAS , 399:1597–1610, November 2009.[42] R. A. Sunyaev and L. G. Titarchuk. Comptonization of low-frequency radiation in accre-tion disks Angular distribution and polarization of hard radiation.
A&A , 143:374–388,February 1985.[43] L. Titarchuk, D. Kazanas, and P. Becker. Broad Redshifted Line as a Signature ofOutflow.
ApJ , 598:411–418, November 2003.[44] L. Titarchuk, A. Mastichiadis, and N. D. Kylafis. X-Ray Spectral Formation in a Con-verging Fluid Flow: Spherical Accretion into Black Holes.
ApJ , 487:834, October 1997.[45] L. Titarchuk and E. Seifina. Discovery of Photon Index Saturation in the Black HoleBinary GRS 1915+105.
ApJ , 706:1463–1483, December 2009.[46] L. Titarchuk, N. Shaposhnikov, and V Arefiev. Power Spectra of Black Holes and NeutronStars as a Probe of Hydrodynamic Structure of the Source: Diffusion Theory and ItsApplication to Cygnus X-1 and Cygnus X-2 X-Ray Observations.
ApJ , 660:556–579, May2007.[47] L. Titarchuk and C. Shrader. Downscattering Due to Wind Outflows in Compact X-RaySources: Theory and Interpretation.
ApJ , 623:362–371, April 2005.[48] L. Titarchuk and T. Zannias. The Extended Power Law as an Intrinsic Signature for aBlack Hole.
ApJ , 493:863, January 1998.[49] F. Tombesi, M. Cappi, J. N. Reeves, R. S. Nemmen, V. Braito, M. Gaspari, and C. S.Reynolds. Unification of X-ray winds in Seyfert galaxies: from ultra-fast outflows to warmabsorbers.
MNRAS , 430:1102–1117, April 2013.[50] T. J. Turner, L. Miller, I. M. George, and J. N. Reeves. Evidence for orbital motion ofmaterial close to the central black hole of Mrk 766.
A&A , 445:59–67, January 2006.[51] T. J. Turner, L. Miller, J. N. Reeves, and S. B. Kraemer. The variable X-ray spectrumof Markarian 766. II. Time-resolved spectroscopy.
A&A , 475:121–131, November 2007.[52] R. V. Vasudevan, R. F. Mushotzky, L. M. Winter, and A. C. Fabian. Optical-to-X-ray emission in low-absorption AGN: results from the Swift-BAT 9-month catalogue.
MNRAS , 399:1553–1575, November 2009.[53] S. Vaughan and A. C. Fabian. The high frequency power spectrum of Markarian 766.
MNRAS , 341:496–500, May 2003.[54] R. Walter and H. H. Fink. The Ultraviolet to Soft X-Ray Bump of SEYFERT-1 TypeActive Galactic Nuclei.
A&A , 274:105, July 1993.
Slice t astart t astop net total orbit Time t astart t astop net total orbitexposure b counts slice exposure b counts1 0 . . .
57 134019 0082 26 2 . . .
59 89374 10012 0 . . .
41 125636 0082 27 2 . . .
94 85746 10013 0 . . .
56 170706 0082 28 2 . . .
31 89651 10014 1 . . .
45 245568 0265 29 2 . . .
31 105675 10015 1 . . .
11 326828 0265 30 2 . . .
22 204204 10016 1 . . .
11 289675 0265 31 2 . . .
18 152048 10017 1 . . .
82 266728 0265 32 2 . . .
31 70649 10018 1 . . .
06 283484 0265 33 2 . . .
91 139757 10019 1 . . .
12 288279 0265 34 2 . . .
11 58107 100210 1 . . .
61 277075 0265 35 2 . . .
77 256078 100211 1 . . .
12 267828 0265 36 2 . . .
41 164053 100212 1 . . .
77 267617 0265 37 2 . . .
41 150947 100213 c . . .
50 62289 0999 38 2 . . .
68 143367 100214 c . . .
31 30118 0999 39 2 . . .
01 139289 100215 c . . .
54 63629 0999 40 2 . . .
59 134918 100216 c . . .
52 76954 0999 41 2 . . .
54 92120 100217 c . . .
64 57031 0999 42 2 . . .
68 180264 100318 c . . .
76 46840 0999 43 2 . . .
11 152568 100319 c . . .
16 30153 1000 44 2 . . .
42 246411 100320 c . . .
44 41207 1000 45 2 . . .
01 190676 100321 2 . . .
53 135942 1000 46 2 . . .
81 94548 100322 2 . . .
41 102879 1000 47 d . . .
59 19094 100323 2 . . .
51 151659 1000 48 2 . . .
36 44136 100424 2 . . .
58 130706 1000 49 2 . . .
46 124369 100425 2 . . .
41 83773 1000 50 2 . . .
27 84421 1004
The net exposure accounts for the detector live time and the filtering from flaring background.The values of t start and t stop are measured from the begenning of the XMM-Newton mission. a In units of 10 s . b In units of 10 s . c Affected by occultation episodes. d Excluded because not matching of the sampling criteria.24able 4: Best-fit parameters of the baseline applied model.
Slice kT a N bbmc Γ c N dpex N eH ξ f F g χ /dof1 8 . +0 . − . . +0 . − . . +0 . − . . +1 . − . . +0 . − . .
71 3 .
62 916 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +4 . − . .
35 947 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +3 . − . .
34 998 . . +0 . − . . +0 . − . . +0 . − . . +1 . − . . +0 . − . . +2 . − . .
89 1055 . . +0 . − . . +0 . − . . +0 . − . . +2 . − . . +0 . − . . +2 . − . .
32 1189 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . .
78 1264 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +3 . − . .
74 1153 . . +0 . − . . +0 . − . . +0 . − . . +1 . − . . +0 . − . . +1 . − . .
71 1151 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . .
77 1141 . . +0 . − . . +0 . − . . +0 . − . . +1 . − . . +0 . − . . +2 . − . .
36 1201 . . +0 . − . . +0 . − . . +0 . − . . +1 . − . . +0 . − . . +2 . − . .
29 1127 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +1 . − . .
78 1342 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +3 . − . .
11 984 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +4 . − . .
86 903 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +3 . − . .
32 981 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +3 . − . .
46 902 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +5 . − . .
27 757 . Slice kT a N bbmc Γ c N dpex N eH ξ f F g χ /dof26 8 . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +5 . − . .
08 815 . . +0 . − . . +0 . − . . +0 . − . . +1 . − . . +0 . − . . +3 . − . .
89 773 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +3 . − . .
24 743 . . +0 . − . . +0 . − . . +0 . − . . +1 . − . . +0 . − . . +3 . − . .
63 709 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +2 . − . .
10 1054 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +3 . − . .
24 1021 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +1 . − . .
61 710 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . − . .
92 1206 . . +0 . − . . +0 . − . . +0 . − . . +1 . − . . +0 . − . . +4 . − . .
40 608 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +2 . − . .
46 1127 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . .
26 935 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . .
08 943 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . .
42 994 . . +0 . − . . +0 . − . . +0 . − . . − . . +0 . − . . +1 . − . .
48 974 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +3 . − . .
65 977 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +4 . − . .
40 879 . . +0 . − . . +0 . − . . +0 . − . . +1 . − . . +0 . − . . +0 . − . .
26 985 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +2 . − . .
91 955 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . .
75 1170 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . − . .
10 1055 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +1 . − . .
32 824 . . +0 . − . . +0 . − . . +0 . − . . +1 . − . . +0 . − . . +7 . − . .
31 591 . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +3 . − . .
83 948 . . +0 . − . . +0 . − . . +0 . − . . +1 . − . . +0 . − . . +3 . − . .
80 812 . The missing time intervals (13-20,47) refer to the occultation episodes which have been neglected. a BB colour temperature in unit s of 10 − keV . b Normalisation of the BMC component in units of 10 − (cid:16) L erg s − (cid:17) × (cid:16) kpcd (cid:17) . c Intrinsic photon index. d Normalisation of the PEXRAV component in units of 10 − photons keV − cm − s − at 1 keV . e Column density of the ZXIPCF component in units of 10 cm − . f Ionisation degree of the ZXIPCF component in units of erg cm s − . g . − keV flux in units of 10 − erg s − cm −2