Modelling the variable broad-band optical/UV/X-ray spectrum of PG1211+143: Implications for the ionized outflow
AAstronomy & Astrophysics manuscript no. pg1211 +
143 c (cid:13)
ESO 2018November 5, 2018
Modelling the variable broad-band optical/UV/X–ray spectrum ofPG1211+143: Implications for the ionized outflow
I. E. Papadakis , , F. Nicastro , , , and C. Panagiotou Department of Physics & Institute of Theoretical & Computational Physics, University of Crete, PO Box 2208, GR-710 03 Herak-lion, Crete, Greece IESL, Foundation for Research and Technology-Hellas, GR-71110 Heraklion, Crete, Greece Osservatorio Astronomico di Roma-INAF, Via di Frascati 33, I-00040 Monte Porzio Catone, RM, Italy Harvard-Smithsonian Center for Astrophysics, Cambridge, MA 02138, USAReceived .. ...... ; accepted .. ......
ABSTRACT
Context.
We present the results from a detailed analysis of the 2007
Swift monitoring campaign of the quasar PG1211 + Aims.
We study its broad-band optical / UV-X–ray spectral energy distribution and its variations, with the use of physically motivatedmodels.
Methods.
We constructed broad-band, optical / UV-X–ray spectral energy distributions over three X–ray flux intervals, and we fittedthem with a model which accounts for the disc and the X–ray coronal emission. We also added a spectral model component to accountfor the presence of the warm absorber which has been well established from past observations of the source.
Results.
We detected no optical / UV variations over the two-month period of the monitoring campaign. On the other hand, the X–raysare highly variable in a correlated way in the soft and hard X–ray bands with an amplitude larger than has been commonly observedin nearby Seyferts, even on longer time scales. The three flux spectra are well fitted by the model we considered. The disc innertemperature remains constant at ∼ × km / s (3 σ upper limit), and has a column density of log N H ∼ .
2. Itsionization parameter varies by a factor of 1.6, and it is in photo-ionizing equilibrium with the ionizing flux. It is located at a distanceof less than 0.35 pc from the central source, and its relative thickness, ∆ R / R , is less than 0.1. The absorber’s ionization parametervariations can explain the larger than average amplitude of the X-ray variations. Conclusions.
The absence of optical / UV variations are consistent with the high black hole mass estimate of ∼ M (cid:12) for this object,which implies variability time scales longer than the period of the Swift observations. It argues against the presence of inward prop-agating fluctuations in the disc as the reason for the flux variability in this source and against the hypothesis that X–ray illuminationsingificantly a ff ects the disc emission. This is consistent with the low ratio of X–ray over the bolometric luminosity of ∼ −
35 inthis source. Based on the properties of the ionized outflow, we estimate an upper limit for the mass outflow of ∼ (cid:12) per year, whichis ∼ . + ∼ . × ergs s − . Asa result, the outflow cannot deploy significant mechanical energy to the surrounding ISM of the quasar’s host galaxy, but is su ffi cientto heat the ISM to 10 K and to produce a fast decline to the star formation rate of the galaxy.
Key words.
Galaxies: active – Galaxies: Seyfert – quasars: individual: PG1211 +
143 – X-rays: galaxies
1. Introduction
It is generally accepted that active galactic nuclei (AGN) arepowered by the accretion of matter onto supermassive blackholes (BHs) situated at the centres of galaxies in the form ofa geometrically thin, optically thick disc. Assuming black-bodyemission of the locally dissipated energy in the disc, it is ex-pected that the optical / UV (opt / UV) spectra of AGN are dom-inated by a broad quasi-thermal feature with an inner disc ra-dius temperature of the order of ∼ − K (e.g. Shields,1978). This feature has been detected in the opt / UV part of thebroad-band energy distributions of bright quasars (e.g. Elvis etal. 1994), and is commonly referred to as the AGN “big bluebump”. In addition to this emission component, X–ray emissionis also a ubiquitous feature of the AGN spectra. Variability argu-ments and microlensing observations (e.g. Mosquera et al. 2013)indicate that the X–ray source size is less than ∼ − r g (where r g is the gravitational radius), and is probably located very close to the central BH where most of the gravitational energy is liber-ated. The main process for the production of the X-rays is gen-erally thought to be the Comptonization of soft photons (like theones emitted by the disc) by electrons of a hot (kT e ∼ − / UV and X–ray emis-sion should be somehow related in AGN. For example, if theseed photons for Comptonization are those emitted by the disc,their variations should a ff ect the cooling / heating of the X-raycorona, and hence result in X-ray spectral variations. Further-more, given that the AGN emitted power is provided by matteraccretion, one expects variations of the accretion matter rate toa ff ect both the opt / UV and the X–ray region. If the X–ray sourceis located in the innermost region in these objects, the variationsare expected to originate at longer wavelengths (optical) and thenmove to shorter ones (UV, X-rays). Finally, a fraction of the X-ray emission may be intercepted and reprocessed by the materialin the disc, serving as an external heating source for the disc. In
Article number, page 1 of 9 a r X i v : . [ a s t r o - ph . GA ] M a y & A proofs: manuscript no. pg1211 + this case, the X–ray variations are expected to preceed those ob-served in the opt / UV band with a delay proportional to the lighttravel time from the X–ray source to the disc region responsiblefor the opt / UV emission.An additional physical component, namely absorbing mate-rial which lies along the line of sight towards the central en-gine of AGN, is revealed by the detection of absorption linesin the UV and soft X–ray spectra of AGN. Approximately 50%of Seyferts and quasars show absorption signatures in the UVband (e.g. Crenshaw et al. 1999) with a similar fraction of AGNshowing warm absorbing material signatures in their X-ray spec-tra (e.g. George et al. 1998; Piconcelli et al. 2005). Both the UVand the soft X–ray band absorbers appear to outflow. The warmabsorber velocities are rather low, of the order of a few hundredkm / s, and their kinetic power is less than 1% of the AGN bolo-metric luminosities (Blustin et al. 2005; Krongold et al. 2007).However, over the last few years highly ionized, high-velocityoutflows (with a velocity greater than 10 km / s) have been de-tected in a few AGN. They are manifested as blue-shifted K-shellabsorption lines from Fe XXV and Fe XXVI at energies E > Swift mon-itoring data of the quasar PG1211 + z = . β line is 1832 ± / s (Kaspi et al. 2000). This has led to the classification of thisobject as a Narrow Line Seyfert 1 (NLS1) galaxy. PG1211 + ∼ . c has been proposed (Pounds et al. 2003).Kaspi & Behar (2006) have analysed both the CCD and the RGSdata of the same 2001 XMM-Newton observation and report thepresence of an outflow component of about 3000 km / s, in con-trast with the ultra-fast velocity reported by Pounds et al. (2003).Further evidence of the fast ionized outflow has been reported byPounds & Reeves (2007, 2009) using additional XMM-Newtondata. Zoghbi et al. (2015) did not detect absorption features in-dicative of a fast outflow in a recent NuSTAR observation. Ina recent paper, Pounds (2014) has reported the results from anew analysis of the XMM-Newton RGS data, which suggeststhe presence of two absorbers with outflow velocities of ∼ . . Swift observed PG1211 +
143 several times over a period ofjust over two months, in March–May 2007. Bachev et al. (2009)have already studied the
Swift data of the source from the moni-toring campaign in 2007. They found that the source was highlyvariable in X–rays but showed little variations in the opt / UVbands. They also found that its X–ray spectrum can be explainedby either a partial covering absorber or by X-ray reflection ontothe disc, but they were not able to distinguish between the twoscenarios.We model in detail the broad-band (optical / UV / X–ray) spec-tral energy distribution (SED) of the source. Our main aim is toinvestigate whether a simple physical model, consisting of anaccretion disc, a hot corona, and a warm absorber, can fit thebroad-band SED of the source. The PG1211 +
143 data from the2007
Swift monitoring are particularly useful for this purpose asthe X–ray flux was highly variable during the campaign, whilethe optical / UV flux remained constant at all flux intervals. Thisobservational fact can put strong constraints on the modelling ofthe broad-band SED, simultaneously, in all the di ff erent X–rayflux intervals of the source. Table 1.
Swift observations log. T start T exp (s)(Date / UT) (V,B,U) (UVW1) (UVW2) XRT1 09-03 / / / / / / / / / / / / / / / / / / / /
2. Observations and data analysis
PG1211 +
143 was observed by
Swift twenty times from 9 March2007 to 20 May 2007. At first,
Swift observed the source daily;after March 19, Swift observed PG1211 +
143 every 5 to 9 days.A summary of these observations is given in Table 1. Columns 1and 2 list the number, the date, month, and the start time (UT) ofeach observation. The next four columns list the exposure timeof each UVOT filter and XRT in seconds. We note that the sourcewas not observed in the V − band during the last observation (20May 2007).The Swift
XRT (Burrows et al. 2005) observations were per-formed in Photon Counting mode. The event file of the observa-tion was created by using the
Swift analysis tool xrtpipeline .Source photons were extracted from a circle with a radius of20 pixels centred on the source. The background was selectedfrom a nearby, source-free region with a radius of 100 pixels.Only single to quadruple events in the energy range of 0.3-10keV were selected for further analysis. Source and backgroundspectra were extracted from the event file by using XSELECTversion 2.4b. The auxiliary response files were created by theSwift tool xrtmkarf . We used the response matrix version 012with a grade selection 0-12.In addition to the X-ray data, we also obtained photometrywith the UV / Optical Telescope (UVOT; Roming et al. 2005) inthe V, B, U, UVW1, and UVW2 filters.
Swift observed the sourcewith the UVM2 filter as well. Since its e ff ective area overlapsthe e ff ective area of the UVW1 and UVW2 filters, the respectivesource’s flux will be heavily correlated with the flux measuredwith the UVW1 and UVW2 filters. For this reason, we do notconsider the data from the UVM2 filter. Photons were extractedfrom a circular region with r = uvotsource wasused to determine the magnitudes and fluxes (Poole et al. 2008;Breeveld et al. 2010). The fluxes were corrected for Galacticreddening (E(B-V) = Article number, page 2 of 9. Papadakis et al.: Opt / UV / X–ray variability of PG1211 + F l u x ( x - e r g s s - c m - H z - ) C oun t R a t e V UBUVW1UVW2
Fig. 1.
PG1211 +
143 UVOT and X–ray light curves (top and bottompanels, respectively) during the 2007
Swift monitoring campaign. dard reddening correction curves by Cardelli et al. (1989), asdescribed by equation 2 in Roming et al. (2009).
3. Optical/UV and X–ray light curves
The top panel in Fig. 1 shows a plot of the PG1211 +
143 lightcurves in all the UVOT bands (a value of 0.5 × − erg s − cm − Hz − has been added to the V filter’s fluxes for clarity). The bot-tom panel shows the X–ray light curves in the 0.3–2 and 2–8 keVband (soft and hard X–rays, respectively). Time on the x- axis ismeasured in days starting from the date of the first observation.The UVOT band light curves exhibit variations which arewithin the observational errors. Application of the χ -test con-firms that the the V, B, U, and UVW1 light curves are consistentwith the hypothesis of a constant flux (the probability, p , is largerthan 0.22 in all cases). The probability of constant flux is some-what lower for the UVW2 light curve ( p = . F rms )is 0 . ± .
01 (the errors correspond to the measurement errorsonly, and are computed as described in Vaughan et al. 2003).The X–ray light curves are highly variable. We observe the typ-ical, erratic variations we observe in the long term light curvesof other AGN as well. The observed variations show a soft andhard band min-to-max variability amplitude of the order of ∼ ∼ .
5, respectively. We found that F rms = . ± .
02 and0 . ± .
03 in the soft and hard X–ray bands, respectively.Given that we do not detect significant variations in the op-tical / UV bands, it is meaningless to search for X–ray / opt / UVcorrelations. On the other hand, the soft and hard band X–ray . - k e V band c oun t r a t e ( c / s ) Fig. 2. / s. variations are highly correlated. This is demonstrated in Fig. 2,where we plot the soft vs the hard band count rate. We fitteda straight line (i.e. y = a + bx ) to the data using the subrou-tine fitexy of Press et al. (1992) in order to take into accountthe errors on both count rates. A straight line fits the data rea-sonably well: χ = . /
18 degrees of freedom ( p null = . y − intercept turned out to be negative ( a = − . ± . / s ( χ = . y − intercept is consistent with zero: − . ± . / s. This is the case (good fit and best-fit y − intercept con-sistent with zero) irrespective of the upper limit on the 2 − / s.The red dashed line in Fig. 2 indicates the best-fit line to thedata with 2–8 keV count rate smaller than 0.04 counts / s. Clearly,the highest X-ray flux points are not consistent with this line.They lie above the extrapolation of the best-fit line to highercount rates. Our results suggest that the relation between the softand hard band count rates is linear, but at higher fluxes the softband varies with a larger amplitude than the hard band. The valueof the upper count rate of 0.04 c / s may seem arbitrary; however,as we discuss in §5.3, this behaviour (i.e. non-linear above a cer-tain X–ray luminosity level) could arise in the case of changes inthe opacity of a warm absorber, which would a ff ect mainly thesoft band photons.
4. Flux resolved spectroscopy
In order to study the broad-band spectral energy distribution ofthe source (SED), we defined three di ff erent flux intervals basedon the observed 0.3-2 keV light curve: (a) a high flux-interval(HF: 3 Swift observations with a soft band count rate > . − ), (b) a medium flux-interval (MF: 4 Swift observationswith count rates in the range 0.07-0.2 ct s − ), and (c) a low flux-interval (LF: 14 Swift observations with count rates < .
07 cts − ). The horizontal dashed lines in the bottom panel of Fig. 1 Article number, page 3 of 9 & A proofs: manuscript no. pg1211 + indicate the flux limits for these intervals. The three HF obser-vations are those which do not follow the linear trend indicatedby the dashed line in Fig. 2. They were taken in the last 25 daysof the monitoring campaign. The MF state data correspond toobservations that were taken throughout the observing Swift run.The LF state data include most of the observations that weremainly taken during the first intensive monitoring period of the
Swift campaign. We note that these intervals refer to distinct
X–ray flux intervals of the source and not to its overall flux outputwhich, given the lack of opt / UV variations, remains almost con-stant throughout the
Swift observing run.To create the total X–ray spectrum for each state, we used mathpha to combine the individual source and background spec-tra of all the observations, within the respective state. The X–rayspectra were then rebinned with grppha (v. 3.0.1) to have atleast 15 photons per bin. We computed the mean flux in eachUVOT band using the deredenned mean flux measurements ofthe individual observations within each state. We then used the flx2xsp command to transform the flux measurements to countrates, and hence to create the respective opt / UV spectrum. Forthe spectral analysis we used the fitting package
Sherpa (Free-man et al. 2001), part of the
Chandra
Interactive Analysis ofObservation software (CIAO; Fruscione et al. 2006).
We first modelled the broad-band SED (extending over fourorders of magnitude in energy) with a single spectral com-ponent, namely xsnthcomp in Sherpa (Zdziarski, Johnson &Magdziarz 1996; Zycki, Done & Smith, 1999), attenuated (onlyin the X-rays) by the amount of Galactic neutral absorptionalong the line of sight to PG 1211 + N H = . × cm − (Kalberla et al. 2005). This model provides a description of thecontinuum shape from thermal Comptonization. The model pa-rameters are (a) the electron temperature, T e , of the hot Comp-tonizing corona, which determines the high-energy cut-o ff of theX-ray spectrum; (b) the photon index, Γ , of the ComptonizedX–ray powerlaw; (c) the redshift, z , of the emitting source; (d)the model normalization, N ; and (e) the soft-photon tempera-ture, T BB . We chose the seed photons to have a multicolour, discblack-body energy distribution in order to simultaneously fit theopt / UV and the X–ray data with the same model.In practice, we did this by fitting the six opt / UV and X-ray spectra (two spectra for each of the HF, MF, and LF inter-vals) with six distinct xsnthcomp model components. For eachof the three flux-state spectra, all parameters of the three X-ray xsnthcomp components are linked to the respective parametersof the associated UV xsnthcomp components. For all spectra,the high-energy cut-o ff of the Comptonized X-ray power law(to which our data are insensitive with the current UVOT + XRTdata) is frozen to T e =
100 keV . The only three parameters freeto vary in the fit are therefore T BB , Γ , and N . Since we do notobserve any significant opt / UV flux variability during the
Swift campaign, we force T BB to be the same for all three flux-statespectra, while N and Γ are left free to vary independently in thedi ff erent flux intervals (hereafter Model A).Table 2 summarizes the best-fitting parameter values andstatistics, for Model A. Figure 3 shows the HF, MF, and LF spec-tra (black, red, and blue points, respectively). The linesaboveeach spectrum indicate the best-fit Models A (top panel). The Zoghbi et al. (2015) have recently reported a lower limit of 124 keVon T e . Our results remain una ff ected even if we choose a value as largeas 200 keV for T e Table 2.
Best-fitting Model A results ( χ r / dof = / Spectrum T BB Γ Norm. a (eV)HF 2 . ± .
06 2 . ± .
01 18 . ± . b . ± .
02 7 . ± . b . ± .
01 3 . ± . Notes. ( a ) in 10 − ph s − cm − keV − b ) Parameter value linked to thevalue for the HF spectrum
Fig. 3.
Broad-band spectra of PG1211 +
143 in its HF (black), MF (red),and LF (blue) flux; super-imposed are the corresponding best-fittingmodel A lines (top panel) and the residuals (in standard deviations) be-tween the data and their best-fit models (bottom panel). plot in the bottom panel shows the residuals (in standard devia-tions) between the data and their best-fit models. The model fitsthe opt / UV part of the spectra well; however, it fails to accuratelymodel the X–ray data, yielding a global reduced χ ( χ r ) of 2.5for 218 degrees of freedom (dof). For all three flux intervals, theresiduals in the X-rays show clear systematic deviations; thereare two pronounced excesses between 0.3-0.7 keV and 2-6 keV,and a deficit in between (Fig. 3, bottom panel). This is a clearsignature, in low-resolution CCD-like spectra, of the presenceof an ionized absorber, attenuating the primary X-ray contin-uum mainly at energies around 1 keV where the residual opacitydue to highly ionized metals is stronger. This is not a surprisingresult, given the numerous previous reports for the presence ofsuch material in PG1211 + We therefore added a dust-free ionized absorber component (i.e.that does not a ff ect the opt / UV data) to our Model A by multiply-ing the continua of the three flux-state spectra by three distinct phase components (our PHoto-ionized Absorber Spectral En-gine model, Krongold et al. 2003, implemented as a user-modelin
Sherpa ), one for each flux interval (hereafter Model B). Theparameters of a phase component are (a) the ionization parame-
Article number, page 4 of 9. Papadakis et al.: Opt / UV / X–ray variability of PG1211 + Table 3.
Best-fitting Models B and C: χ r / do f = . /
208 and χ r / do f = . / Model BSpec. Γ Norm. a log U logN H f c υ out F Ionb F − b (cm − ) (km s − )HF 2 . ± .
01 4 . ± . . + . − . . + . − . . ± . − + − . ± .
02 0 . ± . . ± .
01 3 . ± . . + . − . . ± .
01 0 . + . − . L-HF 0 . ± .
03 0 . ± . . ± .
01 1 . ± . . + . − . . + . − . . ± .
01 L-HF 0 . ± .
02 0 . ± . Γ Norm. log U logN H f c v out F Ion F − (cm − ) (km s − )HF 2 . ± .
01 4 . ± . . + . − . . + . − . . ± . − . ± .
02 0 . ± . . ± .
01 3 . ± . . + . − . L-HF L-HF L-HF 0 . ± .
03 0 . ± . . ± .
01 1 . ± . . + . − . L-HF L-HF L-HF 0 . ± .
02 0 . ± . Notes. ( a ) in 10 − ph s − cm − keV − at 1 keV ; ( b ) in 10 − erg s − cm − ter U , defined as the ratio between the volume density of ionizingphotons (i.e. with energy ≥ . n phot , over the electron volume densityin the cloud, n e : U = [ Q ion / (4 π cR )] / n e = n phot / n e (where Q ion is the total number of ionizing photons); (b) the equivalent Hcolumn density of the cloud along our line of sight, N H ; (c) theinflow / outflow radial velocity of the absorber along the line ofsight, υ r ; and (d) the fraction of ionizing source covered by theabsorber along the direction of the observer, f c .We first left all three of the relevant parameters, U , N H , and f c , free to vary independently between the three phase compo-nents, while the radial velocity of the absorber was kept linked toa single value in all three spectra and initially left free to vary inthe fit. As before with Model A, T BB was forced to be the samefor all three flux-state spectra, while N and Γ were left free tovary independently in the di ff erent flux-state spectra.Table 3 summarizes the best-fit parameter values and statis-tics for Model B, which provides an excellent description of boththe UV and X-ray data in all three flux intervals. The best-fit soft-photon temperature is T BB = . ± .
03 eV. In columns 8 and 9we list the model ionizing flux (i.e. the unabsorbed 0.0136-100keV continuum flux) and the model unabsorbed X–ray flux inthe 2–10 keV band.The model results indicate that neither the column densitynor the covering factor varies (within the errors) between thethree flux-state spectra. Moreover, while the sign of the line ofsight velocity of the absorber indicates that the gas is outflow-ing (even at our low resolution we rule out positive velocitiesat (cid:39)
95% confidence level; see Fig. 4), our low-resolution X-ray spectra cannot constrain this velocity to better than ∼ υ r = − (3000 ± − reported by Kaspi & Behar(2006), based on the high-resolution XMM- Newton -RGS spec-trum of PG 1211 + σ level) with the v r ∼ − − value reported byPounds (2014) for the mildly relativistic outflow in this object(see Fig. 4). Our 0.3-6 keV spectra are not sensitive to the evenhigher velocity v r (cid:39) . c outflow first reported by Pounds &Page (2006).We therefore linked the column densities and covering fac-tors of the absorber to a single value in all three flux-state spec-tra, we froze the outflow velocity of the absorber to v r = − − (Model C), and we repeated the fit. The best-fitting pa-rameters and statistics of Model C are given in Table 3, while z abs0 0.02 0.04 0.06 0.08 f c Fig. 4.
Contours at 1, 2, and 3 σ confidence levels for the best-fittingredshift of the ionized absorber of Model B, which translates into theradial velocity of the absorber through the relationship υ r = ( z abs − z sys )* c, where z abs and z sys are respectively the redshift of the absorber andthe systemic redshift of PG1211 + Fig. 5 shows the spectra with the corresponding best-fitting mod-els (top panel) and residuals (bottom panel). Model C fits the UVand X-ray data in all three flux intervals as accurately as ModelB ( χ r / do f = . / BB = . ± .
03 eV. The residuals are flat over the entireenergy range covered by our data (Fig. 5, bottom panel).We note that both the ionization parameter of the photo-ionized outflow and F
Ion decrease monotonically from the HF tothe LF spectrum. Figure 6 shows the measured number of ion-izing photons per steradian emitted by the central source, Q ion (divided by 4 π c ), plotted against the best-fit ionization parame-ter U , for the three flux-interval spectra, in log-log space. Clearlythe two quantities are linearly correlated (as shown by the best-fit line in the same figure). The ionized absorber is thus con-sistent with being always in photo-ionization equilibrium withthe ionizing flux during the entire Swift campaign. This allowsus to infer an accurate estimate on the quantity ( n e R ). Usingthe definition of U , it can be shown that log( Q ion / π c ) = log U + log( n e R ). Therefore, the intercept of the best-fit line shownin Fig. 6 is a measure of the product between the electron vol- Article number, page 5 of 9 & A proofs: manuscript no. pg1211 + Fig. 5.
Same as Fig. 3, for the Model C best-fit results.
20 30 40 50
Fig. 6.
Best-fitting number Q ion of ionizing photons at the illuminatedface of the absorbing cloud of gas (divided by the quantity 4 π c ), vsthe best-fitting ionization parameter, U , for the three flux-interval spec-tra HS, MS and LS. Both the x and y − axis are in logarithmic scale.The straight line indicates the best-fitting linear relationship betweenlog( Q ion ) and log( U ), whose intercept provides an estimate for the quan-tity n e R ) (see §4.2). ume density in the absorbing gas and the square of the distanceof the cloud from the ionizing source. In this way, we find that n e R = × cm − . The photo-ionization equilibrium scenario also allows us to re-move the degeneracy between n e and R , by setting a lower limit on n e . If the PG1211 +
143 absorber is in photo-ionization equi-librium during the entire campaign, the absorbing gas must beable to relax to photo-ionization equilibrium over a time scaleshorter than the shortest time interval between two consecutivedi ff erent X-ray flux intervals of the quasar.The time needed for a cloud of gas to reach equilibrium withits ionizing radiation depends inversely on the gas volume den-sity, both during increasing and decreasing phases of the ioniz-ing radiation. It is typically significantly longer during recom-bination than during ionization phases (e.g. Nicastro et al. 1999;Krongold et al. 2007). For a 3-dominant-ion condition (i.e. whenthe ion fractions of a given element are mostly distributed among3 consecutive species) this equilibration time can be analyticallyapproximated as t X i , X i + eq ( t → t + dt ) ∼ (cid:34) α rec ( X i , T e ) eq n e (cid:35) (cid:16) α rec ( X i − , T e ) α rec ( X i , T e ) (cid:17) eq + (cid:18) n Xi + n Xi (cid:19) eq t + dt , where t X i , X i + eq is the equilibration time scale relative to the transi-tion from the ions i and ( i +
1) of the element X , as the source fluxchanges over the time interval from t to ( t + dt ); α rec ( X i , T e ), n e ,and n X i are the radiative-recombination coe ffi cient of the ion X i at an electron temperature T e , the gas electron volume density,and the relative fraction of the ion X i , respectively. The subscript eq indicates quantities at equilibrium at the time ( t + dt ).In our case, the 3-dominant-ion hypothesis is verified forseveral light elements. For example, our Model C best-fit re-sults show that the three highest ionization ions of oxygen (He-like, H-like, and fully-stripped) have three-digit approximatedfractions of f OVII = . f OVIII = . f OIX = . U = . f OVII = . f OVIII = . f OIX = .
671 (log U = . f OVII = . f OVIII = . f OIX = .
892 (log U = .
51: HS), and that their sum is equalto unity in all cases. During the
Swift campaign, the shortest timeinterval for a transition from one flux-interval to a another is thatbetween the fourth (flux-interval MS) and fifth (flux-interval LS)observations, separated by ∆ t = U (LS) = .
317 (and logN H = .
2, in cm − ) is T (LS) = . × K. This gives α rec ( OV III ; 2 . × ) = . × − cm s − , α rec ( OV II ; 2 . × ) = . × − cm s − (coef-ficients from Shull & van Steenberg, 1982), and ( n OIX / n OVIII ) = .
22. Thus, t OVIII , OIXeq (4 th → th observation ) ∼ . × n − e < s (1)or n e > . × cm − . From ( n e R ) = × cm − , we thenget R < . × cm = .
35 pc (or 1.1 lt-years). This radius isfully consistent with the broad line region (BLR) radius in thisobject, which is of the order of ∼ . ∆ R . This is simply equal to the ratio of the equiv-alent H column density over the H volume density, n H . For afully ionized gas (which is a good approximation here), n H isequal to n e / .
23. Therefore, using the best-fit N H estimate andthe upper limit on n e that we estimated above, we find that ∆ R = N H / n H (cid:39) . N H / n e < . × cm = .
035 pc (0.11lt-years). We can also provide an upper limit on the relative
Article number, page 6 of 9. Papadakis et al.: Opt / UV / X–ray variability of PG1211 + thickness of the out flow as follows: ( ∆ R / R ) = ( N H / n H ) / R = . N H ( n e R ) − / ( n e ) − / . Using our n e R estimate (see Section4.3), our best-fit N H value, and the lower limit on n e , we find that ∆ R / R < .
5. Discussion and conclusions
We present the results of a detailed study of the data from the
Swift monitoring campaign of PG1211 +
143 in 2007. This is abright quasar, with a reverberation BH mass estimate of ∼ . × M (cid:12) (Peterson et al. 2004). Based on our best-fit Model Cresults, the 1 µ m −
100 keV luminosity of PG1211 +
143 (whiche ff ectively measures the bolometric luminosity, L bol ) is of theorder of 5 × erg s − . This is ∼ . solar mass BH.We studied the broad-band spectra of the source by con-structing opt / UV / X–ray spectra in three di ff erent X–ray flux in-tervals. Our main aim was to describe the broad-band SED withthe simplest physical model possible, and to use the observa-tionally strong fact of highly variable X–ray emission in the ab-sence of opt / UV variations to constrain the model as much aspossible. We found that the three SEDs are modelled accuratelyby two physically motivated, spectral components. The first ac-counts for the e ff ects of the warm absorber in this object. Thesecond accounts for the opt / UV / X–ray emission. We find that asingle nthcomp model, with a constant temperature of the in-put soft photons, can fit the opt / UV / X-ray SED well in all threeX–ray flux levels. From a physical point of view, this result cor-responds to the emission from an accretion disc consisting ofmultiple black-body components and Comptonization from a hotcorona which covers the disc.The configuration of the inner region in PG1211 +
143 couldbe similar but not identical to what has been observed in otherAGN as well. For example, Petrucci et al. (2013) presented theresults from the study of the broad-band SED of Mrk 509. Thephysical picture in PG1211 +
143 could be similar to the pictureshown in their Fig. 10. The deep layers of the accretion disc radi-ate like a black body at T BB ∼ +
143 appears to be much hotterthan the
WARM corona in Mrk 509 (as Petrucci et al. called it),and could be the
HOT corona itself.A soft-excess spectral component is not necessary to fit thepresent data set. In fact, evidence that this component is missingin PG1211 +
143 is quite strong and is independent of our spectralmodel fit assumptions. This evidence results from the fact thatthe soft and the hard band count rates vary proportionally mostof the time, and the soft-vs-hard counts plot is well fitted by astraight line with an intercept consistent with zero. Noda et al.(2011, 2013) have detected significant positive o ff sets in similarsoft-vs-hard “count-coubt” plots in a handful of X–ray quasars.They have interpreted them as evidence of an extra soft-excesscomponent which is less variable than the X–ray continuum. Thelack of a similar non-positive o ff set in our case argues againstthe presence of an extra, less-variable, soft-component in thissource.Our results strongly suggest that both the soft and hard bandX–ray photons are produced by the same component, hence thelinear relation between the two. At the highest X–ray flux in-tervals the lineal relation breaks. The soft X–ray count rates in-crease by a factor that is larger than in the hard band. As wedemonstrate and discuss in §5.3, this is most probably due to strong opacity changes, caused by higher ionization of the warmabsorber, which a ff ect mainly the 0.3–2 keV energy band, andnot the X–ray photons at higher energies. We do not detect any significant optical / UV variability duringthe 80-day observations. The lack of opt / UV variations is con-sistent with the high BH mass measured in this object, as thevarious disc time scales are expected to be long in this case. Forexample, the viscous time scale for an accretion disc around aBH is given by the relation t visc ∼ α − . R / M ( r / h d ) s, where α . is the viscosity parameter (in units of 0.1), R is the disc ra-dius (in units of 3 Schwarzschild radii, R S ), M is the BH mass(in units of 10 M (cid:12) ), and r / h d is the ratio of radius over thedisc thickness (Czerny, 2006). For a thin disc (i.e. a disc where h d / r ∼ . ∼
120 daysfor M ∼ R S (i.e. the innermost radius fora stable circular orbit in a non-rotating BH) assuming a value of0.1 for the viscosity parameter. This time scale is longer than theduration of the Swift monitoring campaign of PG1211 + / UV variations is not consistentwith the hypothesis of inwards propagating disc accretion ratefluctuations. The absence of opt / UV variations argues againstthe operation of accretion rate fluctuations at larger disc radii.Therefore, the long term X–ray variations cannot be due to prop-agating fluctuations that happened at larger radii. Of course, thepropagation time scale may be longer than the time span of the2007
Swift observations, and the X–ray variations we observemay be due to fluctuations at larger radii which happened beforethe start of the observations. It is possible that we were some-what unlucky and that disc fluctuations at large radii simply didnot occur during the 2007 observations, although we consider itrather unlikely.Furthermore, X–ray illumination does not seem to signifi-cantly a ff ect the disc emission in PG1211 +
143 (as again, wewould expect the opt / UV flux to be variable). This is not sur-prising, given the small fraction of the X–ray over the bolometricluminosity (which is mainly determined by the disc flux). The ra-tio between the F ion and F − fluxes (listed in Table 3) is of theorder of 18–36 in the HF and LF intervals (given the steep X–rayspectrum, the F − flux is representative of the total X–ray fluxin this object). Therefore, energetically, the X–rays are not pow-erful enough to influence the opt / UV emission of PG1211 + According to our results the temperature of the inner disc radius,T BB , remains constant and equal to ∼ ∼ . × K. As-suming that the opt / UV emission in PG1211 +
143 is producedby a multicolour accretion disc, the hottest temperature in thedisc should be ∼ . GM ˙ M / π R σ ) / (Pringle, 1981). As-suming a BH mass of ∼ M (cid:12) , R in = R S , an accretion rate of ∼ . M Edd , and an e ffi ciency of 0.1, we find that the maximumdisc temperature should be of the order of ∼ . ∼ ∼ R S .The accretion flow at smaller radii may be filled by the hotX–ray corona. The X–ray spectral slope we found ( ∼ . − .
6) israther steep, but is in agreement with the measurement of Zoghbi
Article number, page 7 of 9 & A proofs: manuscript no. pg1211 + et al. (2015). We cannot constrain the temperature of the X–ray corona in PG1211 + ∆Γ ∼ .
15 steepening of the spectral slopein the 2–10 keV band, as the flux decreases by a factor of 3,from the high to the low flux state. Interestingly, this “steeperwhen fainter” is uncommon in Seyferts, where we usually ob-server a “steeper when brighter” behaviour (i.e. Sobolewska &Papadakis, 2009). This is further evidence for the decoupling be-tween the opt / UV and the X–ray emission. A variable opt / UVsoft photon input would result in spectral slope – flux relation op-posite to what we observe. The nthcomp model does not providean estimate for the optical depth of the hot plasma. We comparedthe nthcomp best-fit spectral models with those from the compps model (Poutanen & Svensson, 1996), assuming a slab geometryand the same T e and T BB values. We found that the observedspectral slopes correspond to an optical depth of τ ∼ . − . nthcomp does not include an explicit factor forthe relative power between the disc and the corona emission. Wecan get a rough estimate of the ratio of the energy released ineach phase following Appendix B of Petrucci et al. (2013). Us-ing their equation (B.4), we find that the soft photon count rate, n s , is roughly equal to the observed photon count rate (since τ ismuch smaller than unity). The observed count rate can by com-puted by the best-fit nthcomp model (we assume here an aver-age Γ of 2.5, an average Norm of 3 × − , and kT e =
100 keV).Knowing n s and the temperature of the disc modified black-bodyspectrum, we can estimate the soft photon luminosity, L s . Usingequation (B10) of Petrucci et al. (2013), we can then estimatethe heating power, L h , liberated in the corona to Comptonizethe soft photon field. In this way, we found that the amplifica-tion ratio, A = L h / L s , is roughly equal to 2. This is the casewhen all the gravitational power is liberated in the hot medium(Haardt & Maraschi, 1991). However, if this were the case, wewould expect the disc emission to be entirely due to X–ray re-processing, and hence, to be variable as well, which is not thecase. If the X–ray emission region is outflowing, so that the pho-tons Comptonized in the corona are only emitted upward, then L h ≈ L obs − L s and A ≈
1, which is the case when the gravitationalpower is released almost entirely in the disc. Consequently, ourresults suggest the case of an outflowing corona in PG1211 + The X–ray fractional variability amplitude is of the order of ∼ . − .
85, which is quite large. For comparison, the maxi-mum observed fractional variability amplitude of nearby brightSeyferts is of the order of ∼ . − . +
143 over a period of just ∼ ∼
3. This is not enough toexplain the observed max-to-min variability by a factor of ∼ × km / s (upper 3 σ limit) and has a column densityof log N H ∼ .
2. It is in photo-ionizing equilibrium with the
Fig. 7.
Ratios between the best-fitting transmitted models (at high spec-tral resolution of ∆ E = / HF (black), LF / HF (green),and LF / MF (magenta) spectra in the 0.3-2 keV spectral range (see textfor details). ionizing flux and is located at a distance of less than 0.35 pcfrom the central source. The relative thickness of the absorber, ∆ R / R , is less than 0.1. Therefore, the PG1211 +
143 outflow, asseen along our line of sight is significantly thinner than its dis-tance from the central ionizing source. This suggests, at least atthe location where our line of sight crosses the flow, that the out-flow is not propagating outwards radially but rather transversallywith respect to our line-of-sight perspective.The opacity of an ionized absorber to X-ray photons de-pends critically on the degree of ionization of the absorber itself.During the
Swift campaign, the ionized absorber varies betweenlog U (cid:39) . − .
5. We expect both the He- and H-like ions of theelements lighter than S, as well as the Fe unresolved transitionarray (UTA) and the higher ionization Fe L transitions, to playan important role in modifying the transmitted spectral shape ofthe continuum between E (cid:39) . − ∆ E = / HF (black), LF / HF (green), and LF / MF (magenta) inthe 0.3-2 keV spectral range. The straight lines are the identityratios for the respective cases (the three ratios have been multi-plied by factors of 200 (black), 40 (green), and 0.5 (magenta) toseparate them and make them clearly visible in a single panel.)The changes in opacity between the three flux-state a ff ect theentire 0.3-2 keV spectral range, and clearly extend even at ener-gies higher than 2 keV. The strongest variations are between theHF and the LF spectra (green curves), where the high-ionizationFe L opacity changes mimic a flattening of the spectral slope atE (cid:38) . (cid:38) Article number, page 8 of 9. Papadakis et al.: Opt / UV / X–ray variability of PG1211 + linear soft vs hard X-ray count rate relation at high X-ray fluxes(Fig. 2). Based on our best-fitting Model C estimates of the physical(log(N H ) = . v r = − − ), and ge-ometrical (thin flow with R < .
35 pc and ∆ R / R < .
1) pa-rameters of the ionized outflow of PG 1211 + M out (cid:39) . π m p N H v r R < . × gs − (cid:39) (cid:12) yr − (assuming a vertical disc wind and an incli-nation angle of 30 degrees). This upper limit is about 2.3 timeslarger than the Eddington accretion rate for PG1211 +
143 (whichis about 2.2 M (cid:12) yr − , assuming L bol = × ergs / s, and a 10%radiative e ffi ciency). If the mass outflow rate is indeed that high,then outflows like this are probably short-lived episodes in thequasar lifetime.Finally, for the kinetic power of the outflow we obtain P out = / M out v r < . × erg s − , i.e. only about 7% of the radia-tive power of the quasar. At this power, it would require about2.3 Gyrs for the outflow to deploy enough mechanical energy( (cid:39) ergs) into the galaxy and the surrounding inter-galacticmedium (IGM) to control their evolution. This is probably fartoo long for a plausible wind lifetime. However, it would only re-quire about 3 Myrs for such an outflow to deposit enough energy(about 10 ergs, assuming an average ISM density of n H = . − ; e.g. Krongold et al. 2007 and references therein) in the100 kpc radius, 1 kpc thick disc of its host galaxy, and heat theISM to its evaporation temperature of (cid:39) K. This is probablysu ffi cient to produce a fast decline in the star formation rate ofthe galaxy. Acknowledgements.
This work was supported by the “AGNQUEST” project,which is implemented under the "Aristeia II" Action of the "Education and life-long Learning" operational programme of the GSRT, Greece.
References