Suzaku Observations of Near-Relativistic Outflows in the Bal Quasar Apm 08279+5255.
We present results from three Suzaku observations of the gravitationally lensed broad absorption line quasar APM 08279+5255. We detect strong and broad absorption at rest-frame energies of 2 keV (low-energy) and 7–12 keV (high-energy). The detection of these features confirms the results of previous long-exposure (80–90 ks) Chandra and XMM-Newton observations. The low and high-energy absorption is detected in both the back-illuminated (BI) and front-illuminated (FI) Suzaku XIS spectra (with an -test significance of 99%). We interpret the low-energy absorption as arising from a low-ionization absorber with log 23 and the high-energy absorption as due to lines arising from highly ionized (; where is the ionization parameter) iron in a near-relativistic outflowing wind. Assuming this interpretation we find that the velocities in the outflow range between 0.1 and 0.6. We constrain the angle between the outflow direction of the X-ray absorber and our line of sight to be 36. We also detect likely variability of the absorption lines (at the 99.9% and 98% significance levels in the FI and BI spectra, respectively) with a rest-frame time scale of 1 month. Assuming that the detected high-energy absorption features arise from Fe xxv, we estimate that the fraction of the total bolometric energy injected over the quasar’s lifetime into the intergalactic medium in the form of kinetic energy to be .
Subject headings:cosmology: observations — X-rays: galaxies — galaxies: active — quasars: absorption lines
Recent observations of nearby galaxies indicate a – relation (e.g., Ferrarese & Merritt, 2000; Gebhardt et al., 2000), where is the mass of the central black hole and is the velocity dispersion of the stars in the bulge of the host galaxy. The presence of a – relation suggests that a feedback mechanism exists regulating the co-evolution between the massive black hole at the center of a galaxy and the formation of its bulge. A possible mechanism of feedback is quasar outflows. Recent theoretical models demonstrate that quasar feedback can serve as a fundamental ingredient in structure formation and galaxy mergers (e.g., Granato et al., 2004; Hopkins et al., 2005; Springel et al., 2005). Quasar outflows could possibly provide an important source of feedback during the growth of the super-massive black-holes (SMBHs) in galactic bulges (e.g., Fabian, 1999). Another possible mechanism of feedback may be linked to powerful jets apparently driven by magnetohydrodynamic forces. As observations indicate, these powerful jets are predominantly present in radio-loud (RL) AGNs,111Radio-quiet (RQ) AGN in general do not contain large (i.e. kpc) scale collimated jets, although pc-scale jets have been found in some RQ AGNs (e.g., Blundell et al., 1996). Also a fraction (40%) of radio-quiet AGN could have kpc radio-structures possibly indicating the presence of an “aborted jet” (Gallimore et al., 2006). which show a tendency to be found in massive galaxies and dense environments (e.g., Best et al., 2005). The importance of jets as a feedback mechanism has been demonstrated with recent Chandra observations of cavities in clusters of galaxies and giant elliptical galaxies (e.g., McNamara & Nulsen, 2007, and references therein). The injection of power into the Intergalactic Medium (IGM) by radio jets is a promising feedback mechanism that may explain the suppression of cooling flows in the centers of clusters of galaxies (e.g., Fabian et al., 2000; McNamara et al., 2000; Schindler et al., 2001; Heinz et al., 2002). It is not clear, however, if radio jets also contribute to the feedback process in field galaxies, especially ones in the redshift range of where the number density of galaxy mergers is thought to peak. Most clusters of galaxies are not formed until as inferred from observations (e.g., Hilton et al., 2007) and as predicted in theories that consider a low-density () Universe (e.g., Bahcall & Fan, 1998; Younger et al., 2005). In addition, the fraction of radio-loud AGNs (RLF) appears to evolve with redshift (e.g., Peacock et al., 1986; Schneider et al., 1992; Jiang et al., 2007) and luminosity (e.g., La Franca et al., 1994; Jiang et al., 2007). In particular, the RLF tends to increase with luminosity and decrease with redshift. For example, for luminous AGNs (; where is the absolute magnitude at rest-frame 2500Å) it is expected that the RLF declines from 24.3% to 4.1% as the redshift increases from 0.5 to 3 (Jiang et al., 2007). 222As in Jiang et al. (2007) RLF can be written in the form of , where is the absolute magnitude at rest-frame 2500 Å, , , and .
Quasar outflows present a promising mechanism of feedback in high-redshift quasars and possibly in both radio-quiet and radio-loud AGNs. Powerful winds are observed in Broad Absorption Line (BAL) quasars, which show deep and broad absorption features from highly ionized ultraviolet (UV) transitions. BAL quasars are also commonly detected to be X-ray weak as a result of high intrinsic absorption column densities () typically in the range of (1–50) cm (e.g., Gallagher et al., 2002, 2006). We note, however, that a recent survey of BAL quasars obtained from the cross correlation of SDSS and 2XMM cathalogs by Giustini et al. (2008) finds no or lower than typical intrinsic X-ray neutral absorption from that found in optically selected BAL quasar samples. In the orientation-based BAL model (e.g., Weymann et al., 1991) quasar winds exist in most quasars; however, because of the relatively small opening angles of these outflows only a fraction of radio-quiet quasars have detectable BAL features in their UV and/or optical spectra. Models based on numerical simulations and observations suggest that the winds of BAL quasars are nearly equatorial (e.g., Murray et al., 1995; Elvis, 2000; Proga et al., 2000); however, there are a few observed cases of BAL quasars with outflowing absorbers in the polar direction (e.g., Zhou et al, 2006). Recent studies indicate that BAL quasars comprise 15–40% of the quasar population (e.g., Chartas, 2000; Hewett & Foltz, 2003; Gibson et al., 2008; Dai et al., 2008).
Our current understanding of AGN physics suggests that the most
likely mechanisms to explain the origin of outflows in AGN are
thermal driving, radiation driving (line and continuum), and
magnetic driving. Thermal driving will produce slow winds (with
speeds similar to the sound speed) at large radii (; where is the Schwarzchild radius) and with a
relatively small mass-loss rate ()
(e.g., Begelman et al., 1983; Krolik et al., 1986). Therefore, in AGNs thermal driving
is not expected to produce fast and massive winds and consequently
it is likely not a major contributor to feedback.
Given the typical low temperatures of AGN accretion disks ( K) and the large gas densities at the base of winds we expect that initially the absorbing material will have a relatively low ionization parameter. For such conditions radiation-driving can lead to significant acceleration of the absorber. Magnetic driving could also be present in strong AGN winds, through the action of MHD (magnetohydrodynamic) forces (e.g., Everett, 2005). In general, we expect MHD and radiation-pressure forces to act jointly with the contribution of radiation pressure becoming increasingly important in sources with higher (e.g., Everett, 2005, 2007). Dust in the outflow could also boost the radiation pressure depending on the spectral energy distribution (SED) and column density of the material surrounding the AGN (Laor & Brandt, 2002; Fabian et al., 2008). At the moment, evidence for the presence of near-relativistic outflows in AGN is accumulating (e.g., Chartas et al., 2002; Reeves et al., 2003; Pounds et al., 2003; Dadina & Cappi, 2004; Chartas et al., 2007a; Zheng & Wang, 2008)333A recent paper by Vaughan & Uttley (2008) suggests that some of the claimed near-relativistic outflows, especially in cases with narrow absorption lines, are detected at moderate significance levels and may be spurious. We note, however, that the statistical significance of the blushifted broad X-ray absorption features detected in APM 08279+5255 and PG 1115+080 is not disputed. ; however, there is no satisfactory model that can produce outflows with the near-relativistic velocities observed (e.g., Murray et al., 1995; Proga et al., 2000; Everett, 2005). We note that recent studies (e.g., Chelouche & Netzer, 2003; Everett, 2005) indicate that with the appropriate shielding, initial density of the wind, AGN SED and , the efficiency of the outflows can be significantly increased and the outflow velocities may approach near-relativistic values.
Due to their high intrinsic absorption, many BAL quasars appear as
faint X-ray sources (e.g., Green & Mathur, 1996; Gallagher et al., 1999). Partly because
of this faintness, it is difficult to detect BALs in X-ray
spectra, and as a consequence, there are only a few cases where
X-ray BALs have been detected in gravitationally lensed BAL
quasars where the magnification effect results in increased
signal-to-noise ratio spectra. Observations in X-rays of the BAL
quasar APM 08279+5255, the mini-BAL quasar PG 1115+080, and perhaps the
low-ionization BAL quasar H 1413117 have suggested the presence
of near-relativistic outflows of X-ray absorbing material in these
objects (Chartas et al., 2002, 2003, 2007a, 2007b). The reported variability
of the high-energy absorption features is over rest-frame
time-scales of 1.8 weeks in APM 08279+5255 (significant detection
of variability) and 6 days in PG 1115+080 (marginal detection of
variability). The analysis of these high-redshift quasars implied
that outflows should have a significant impact in shaping the
evolution of their host galaxies and in regulating the growth of
the central black hole. These observations are particularly
important because they allow us to probe quasar winds at times
close to the peak of the comoving number density of luminous
In this paper we describe the analysis of three recent Suzaku observations of the lensed BAL quasar APM 08279+5255. A 100 ks observation of APM 08279+5255 was performed starting on 2006 October 12 (OBS1), a 100 ks observation was performed starting on 2006 November 01 (OBS2), and a 120 ks observation was performed starting on 2007 March 24 (OBS3).
Unless stated otherwise, throughout this paper we use CGS units, the errors listed are at the 1- level, and we adopt a flat -dominated universe with Mpc, , and .
2. Data Analysis
For the reduction and analysis of our observations we used the Suzaku software version 7, which is included in HEASOFT version 6.4. To analyze data from the X-ray Imaging Spectrometer (XIS) and the Hard X-ray Detector (HXD) we used calibration files that are part of the Suzaku CALDB database released on 2008 April 01. 444CALDB version 20080401.
2.1. XIS data analysis
|Date||OBS ID||Telescope||Instrument||Exposure||Net exp||Net counts|
|2002-02-24||Cha02||Chandra||ACIS BI||88.8 ks||…||572376||4.3|
|2002-04-28||Has02||XMM-Newton||EPIC pn||100.2 ks||…||12928136||4.0|
|2006-10-12||701057010||Suzaku||XIS FI||102.3 ks||71.3 ks||776088||4.20.4|
|2006-10-12||701057010||Suzaku||XIS BI||102.3 ks||71.3 ks||304655||3.50.5|
|2006-11-01||701057020||Suzaku||XIS FI||102.3 ks||67.9 ks||712184||3.80.3|
|2006-11-01||701057020||Suzaku||XIS BI||102.3 ks||67.9 ks||285578||3.50.4|
|2007-03-24||701057030||Suzaku||XIS FI||117.1 ks||86.4 ks||6059104||4.00.3|
|2007-03-24||701057030||Suzaku||XIS BI||117.1 ks||86.4 ks||383388||3.90.3|
Our data reduction followed the procedures recommended by the Suzaku team for Spaced-Row Charge Injection (SCI) data. The data reduction was performed on the event files of each XIS instrument (XIS 0, 1, 2, and 3), and began with recalculating the PI555 Each event has a measured “Pulse Height Amplitude” (PHA). A calculated “PHA Invariant” (PI) value is obtained using the PHA in combination with the instrumental calibration and gain drift. For the XIS, the PI column name is “PI”, which takes values from 0 to 4095. The PI vs. energy relationship is the following: . values of the unfiltered event files using the XISPI routine. Once the event files were reprocessed, we used the XSELECT software to apply the standard screening criteria (see the Suzaku ABC guide666http://heasarc.gsfc.nasa.gov/docs/suzaku/analysis/abc/) and obtain “cleaned” event files. The data-screening criteria include selecting events corresponding to ASCA grades 0, 2, 3, 4, and 6, Earth elevation angles greater than 5 (ELV5), Earth day-time elevation angles greater than 20 (DYE_ELV20), exclusion of passages through or close to the South Atlantic Anomaly (SAA), and cut-off rigidity criteria of 6 (COR6). As a final step in screening the data we removed hot-flickering pixels through the use of the SISCLEAN routine in XSELECT. The total exposure time of each XIS chip decreased by 20% after the above screening criteria were applied. Using the clean event files we extracted events in a circular region centered on the source with a radius of 150 pixels (2.5). Background events were extracted in an annulus centered on the source with an inner radius of 230 pixels (3.8) and an outer radius of 430 pixels (7.1). Our selected background region excludes APM 08279+5255 and the calibration sources located near the corners of the CCDs. The response matrix files (RMFs) and ancillary response files (ARFs) were generated using the XISRMFGEN and XISSIMARFGEN routines which include the correction for the hydrocarbon contamination777The XISSIMARFGEN routine incorporates the XISCONTAMICALC routine which is used to correct the observation for the XIS optical blocking filter (OBF) contamination. The absorption due to these contaminants depends on the X-ray energy, time, detector ID and location on the detector. on the optical blocking filter.
For the front-illuminated (FI) XIS chips (XIS 0, 2, 3) we considered events with energies lying in the range 0.6–10 keV, while for the back-illuminated (BI) XIS 1 chip we considered events with energies lying in the range 0.4–8 keV. Due to calibration uncertainties near the CCD Si K absorption edge at 1.84 keV, events with energies lying in the range 1.7–1.95 keV were ignored in the analysis of all four XIS chips. In order to assess systematic uncertainties in the response files, we fitted the Ni K (7.470 keV) calibration line of each instrument. We found similar positive shifts in the inferred energies of the calibration lines of each XIS chip ranging from 10 to 20 eV. These shifts in energy were not large enough to cause any significant impact on our analysis, and therefore we did not attempt to correct them. The net source count rate for each XIS chip and each observation was 0.04 counts s, with a background of 30 % of the source rate. The spectra obtained on the FI chips were combined using the routine ADDSPEC (in HEASOFT FTOOLS) in order to increase their signal-to-noise ratio. In Table 1 we have included information relevant to the XIS data analysis. Specifically, this table lists the observation ID, exposure time, net exposure time (after the screening process), net counts (for the FI and BI chips) and flux in the 2–10 keV observed-frame (for the FI and BI chips) using the best-fitted absorbed power-law model (model 2; §3). We also have included in Table 1 information from two previous deep X-ray observations of APM 08279+5255. These observations correspond to an 88.8 ks Chandra exposure (see Chartas et al., 2002) and to a 100.2 ks XMM-Newton exposure (see Hasinger et al., 2002). The counts collected by the XIS FI chips for each of our observations are comparable to those obtained in the Chandra observation.
2.2. HXD data Analysis
Similarly to the XIS case the clean event files were obtained from the unfiltered event files following the instructions in the Suzaku ABC guide. The screening criteria are similar to those applied to the XIS instruments, specifically, we used ELV5, DYE_ELV20, exclusion of passages close to the SAA, and COR6 (units of ). The HXD-PIN spectrum was extracted from the cleaned events file described above. We extracted the source spectra from the clean files XSELECT. In order to estimate non X-ray background (NXB) events, we used version 2 of a time-dependent instrumental background event file (referred to as the PIN background event file) which was provided by the Suzaku team. The PIN background event file was generated with a count rate that is ten times larger than the real instrumental PIN background. Therefore, we increased the effective exposure time of our observed PIN background spectra by a factor of ten. The exposure time was corrected for dead time using the HXDDTCOR task, leaving an effective exposure time of 90% of the original exposure time. The effective exposure time of each observation, together with the count rates (10–40 keV) of the source and NXB are presented in Table 2. The NXB does not include the contribution from cosmic X-ray background (CXB). Therefore the CXB counts (see Table 2) have been estimated from a fake spectrum generated using the FAKEIT command of XSPEC with the following model (e.g., Boldt, 1987):
|Epoch||Net exposure||10–40 keV count rate (cts )|
|aafootnotetext: The CXB counts have been estimated from a fake spectrum generated using the FAKEIT command of XSPEC with the model given in equation (1).|
The response file used to fit the PIN spectra was obtained from the Suzaku CALDB calibration files. The HXD spectral analysis was performed in the 10–40 keV energy range.
3. Spectral Analysis
In this section we fit the Suzaku spectra of APM 08279+5255 with a variety of models using the software tool XSPEC version 12. We also fit the spectra with more realistic models based on the photoionization code XSTAR. In all spectral models we assume a Galactic column density of 4.1 (Kalberla et al., 2005). Most of this section concentrates on the analysis of the XIS spectra; however, in the last paragraph we present results from the spectral analysis of the PIN spectra of APM 08279+5255.
3.1. XIS spectral fits.
3.1.1 XSPEC spectral fits.
Each observation of APM 08279+5255 provides spectra obtained with the single BI chip (XIS1) and the FI chips (XIS 0, 2, and 3). Since the responses of the FI chips are similar we co-added the FI spectra from each observation. We note that there is no XIS 2 spectrum of APM 08279+5255 for our third epoch (OBS3) due to the failure of the XIS2 chip.888On 2006 November 9, about 2/3 of the imaging area of XIS2 became suddenly unusable (http://heasarc.gsfc.nasa.gov/docs/suzaku/news/xis2.html). To fit the spectra using statistics we grouped each XIS spectrum with a sufficient number of counts. The minimum number of counts per bin was also chosen to maximize the signal-to-noise ratio in each bin without losing the features in the spectra and to keep a similar number of spectral bins in each spectrum (70). The minimum number of counts per bin chosen for the BI chip was 40 for epochs OBS1 and OBS2 and 50 for epoch OBS3. The grouping for the FI chips was 100 counts per bin for epochs OBS1 and OBS2 and 80 counts per bin for epoch OBS3. Note that for epoch OBS3 we have increased the binning of the BI spectra due to the longer exposure and decreased the binning of the FI spectra to compensate for the loss of XIS2.
|FI SPECTRUM||BI SPECTRUM|
|Model||Parameter||Values OBS 1||Values OBS 2||Values OBS 3||Values OBS 1||Values OBS 2||Values OBS 3|
We fit the spectra of APM 08279+5255 with the following models: 1)
power-law (PL; XSPEC model wabs*pow) , 2) absorbed power-law (APL;
XSPEC model wabs*zwabs*pow), 3) ionized-absorbed power-law (IAPL;
XSPEC model wabs*absori*pow), 4) partially covered absorbed
power-law (PAPL; XSPEC model wabs*zpcfabs*pow), 5) absorbed
power-law with a notch (APL+No; XSPEC model wabs*zwabs*notch*pow),
6) absorbed power-law with an absorption edge (APL+Ed; XSPEC model
wabs*zwabs*zedge*pow), and 7) absorbed power-law with two
absorption lines (APL+2AL999We note that if we replace the
APL+2AL model by the XSPEC absorption-line multiplicative model
, we obtain similar results for the
fitted energies and equivalent widths of the absorption features
found at energies 7–12 keV in the rest-frame. All the results described in
this paper using the APL+2AL can be reproduced using this
multiplicative model.; XSPEC model ).
The results of the FI and BI fits with the models described are listed in Table 3. The error bars of the fitted parameters are given at the 68% level (). For models 2 to 7 we assume an intrinsic absorber with a redshift of 3.91 (Downes et al., 1999). The fits using a power-law model (model 1) are not acceptable in a statistical sense. We next fit the spectra of APM 08279+5255 with the absorbed power-law model (model 2; Table 3) assuming an intrinsic absorber. The -test indicates that fits with model 2 result in a significant improvement at the 99% and 99.9% confidence levels in the FI and BI spectra, respectively, compared to fits using model 1. Fits with model 2 indicate significant intrinsic absorption in APM 08279+5255 with a column density of log 23. We also fit the spectra of APM 08279+5255 with more complex models that included an ionized and partially covered absorber (models 3 and 4), however, these fits did not result in a significant improvement (-test significance 95%) compared to the simpler model 2.
Fits to the spectra of APM 08279+5255 with models (models 5–7) that account for the absorption found between 7–12 keV in the rest-frame result in significant improvements (the -test indicates improvements at 99.9% and 99% confidence in the FI and BI spectra, respectively) compared to fits with models that do not include this high-energy absorption. We note that the absorption feature at 7–12 keV in the rest-frame corresponds to a significant detection following the criteria described in §3 of Vaughan & Uttley (2008). Specifically, we find the ratio of the total equivalent width101010The equivalent width (EW) is defined as , where is the continuum flux and is the flux in the absorber. of the absorption features to their uncertainty to be in every observation (see models 5–7 in Table 3).
To illustrate the presence of the high-energy absorption features, we fit the spectra from observed-frame energies of 3.6–10 keV with a power-law model and extrapolated this model to the energy ranges not fit (see Figure 1). The lower panels in Figure 1 show the residuals (difference between the measured counts and model) between the best-fit power-law model and the FI and BI data, respectively. The best-fit values of the photon indices in all observations with this model were consistent with at the 1- level. For the purpose of comparing the absorption residuals between epochs the photon indices for all observations were set to = 2.0. From these fits we notice that the residuals show an absorption feature centered near a rest-frame energy of 8 keV and a possible second absorption feature near a rest-frame energy of 10 keV. We fit the high-energy absorption features with the models listed in Table 3. From these fits we found that adding to the APL model an absorption edge (APL+Ed) or two absorption lines (APL+2AL) improves the fits at the 99% confidence level in the two sets of spectra (FI and BI) and in each observation. The -test indicates that we cannot distinguish between the (APL+Ed) and (APL+2AL) models for fits performed to the spectra of APM 08279+5255 in epochs OBS2 and OBS3, since both models fit equally well the 7–12 keV rest-frame absorption during these epochs. 111111We only find marginal improvements in fits to the spectra of APM 08279+5255 taken in epochs OBS2 and OBS3 using model 7 (APL+2AL) compared to fits using model 6 (APL+Ed). Specifically, in epoch OBS2 these improvements are at the 61% and 91% significance levels in the FI and BI, and in epoch OBS3 they are at the 68% and 50% significance levels. However, fits to the FI and BI spectra of epoch OBS1 using the APL+2AL model provide a significant improvement at the 95% and 98% confidence levels, respectively, compared to fits that use the APL+Ed model. It is important to note that the APL+2AL model was clearly favored over the APL+Ed model in a previous 88.8 ks Chandra observation (Chartas et al., 2002). Fits to the spectra in epoch OBS1 with a model that includes an absorption notch (see Table 3) also provide a significant improvement compared to ones using the APL+Ed model. These -test improvements are at the 98% and 99% levels of significance in the FI and BI spectra, respectively. We note, however, that when we compare the quality of the spectral fits that use the APL+Ed model with fits that use either the APL+2AL or APL+notch models, the -test may not be a reliable tool. The reason for the non-reliability of the -test is that we are not comparing nested models (see Protassov et al. 2002 for details). In order to check the reliability of the -test for these cases, we performed Monte Carlo simulations of 10,000 fake spectra (using the FAKEIT command of XSPEC) assuming an APL+Ed model. In these simulations, the energy and optical depth of the absorption edge are assumed to be normally distributed around their fitted values for epoch OBS1 (see model 6 of Table 3), with a standard deviation given by the error bars of the fits. All other parameters of the APL+Ed model were set to their best-fitted values (epoch OBS1 and model 6 of Table 3). The results of our Monte Carlo analysis are presented in Table 4. In each simulation we have fitted the data with the null model (APL+Ed) and the alternative model (either APL+2AL or APL+notch). We then calculated the value of the -statistic between the null model and the alternative model. In Table 4 we show that the -value, which represents the fraction of simulated cases with values of the -statistic higher than the actual value obtained from our real data, is similar to the null probability of the -test. We therefore conclude that our -test values are reliable and are approximately representative of the improvement of the alternative model (either APL+2AL or APL+notch) with respect to the null model (APL+Ed).
|Alternative model||Spectrum||-statistic/null probability|
|APL+2AL||FI||2.59 / 4.4 10||5.2 10|
|APL+2AL||BI||3.05 / 2.3 10||3.1 10|
|APL+notch||FI||5.32 / 2.4 10||6.8 10|
|APL+notch||BI||7.15 / 0.9 10||3.9 10|
The results of the spectral fits shown in Table 3
indicate a change (greater than 1- ) of the energies of
the best-fit values of the first absorption line (abs1; model 7),
and in the absorption-edge energy (; model 6)
between epochs OBS2 and OBS3 in both the FI and BI spectra. This
change is also suggested by the residuals in Figure 1,
where we have marked with an arrow the best-fitted energies of the
first absorption feature of model 7 for epochs OBS2 and OBS3. The
shift in the energy of the first absorption line is indicative of
possible variability of the outflow. This change can be seen more
clearly in Figure 2 where we show the
confidence contours of the best-fit energies of the first
absorption line (APL + 2AL model) versus its line-flux
normalization in epoch OBS2 (solid line) and in epoch OBS3 (dotted
line).121212We note that for the FI spectra of APM 08279+5255 during
epoch OBS2, the first absorption feature (abs1) falls near the Si
K edge where events with energies lying in the range 1.7–1.95 keV
were ignored in the analysis. Although the loss of these data
points adds a larger statistical error to the best-fitted
parameters of abs1, it does not significantly affect the analysis.
Specifically, this error in the best-fitted energies is less than
the value of and more likely close to
Table 3). In the upper and lower panels of
Figure 2, we show the confidence contours
of the FI and the BI spectra, respectively. The confidence
contours touch at the 99% level of significance for the FI
spectra and at the 95% level of significance for the BI
spectra. The probabilities that the flux-energy parameters of the
first absorption line (model 7; Table 3) are the same
between OBS2 and OBS3 (null probabilities) are
and in the
FI spectra and BI spectra, respectively.131313The square of
the probabilities of being outside the confidence contours that
barely touch (see Figure 2) is an upper limit to the
null probabilities. To take into account possible sampling
effects caused by the number of trials used in our variability
analysis we multiply the null probabilities by six. This factor
corresponds to the number of absorption lines (two) times the
number of observations (three). We conclude that the variability
of the first absorption line is significant at the 99.9%
and 98% levels in the FI and BI spectra, respectively.
We note that the slight differences ( 68% significance) between the FI and BI confidence contours may possibly be associated with differences in the responses, variations in the signal-to-noise ratios of the two detectors, and statistical noise.
|Values FI||Values BI||Values FI||Values BI||Values FI||Values BI|
3.1.2 XSTAR spectral fits.
The spectral analysis presented in §3.1.1 indicates that the intrinsic X-ray absorbing medium of APM 08279+5255 is complex and contains absorbers with different properties (see models 5–7 in Table 3). We identified a low-ionization absorber with a column density of log 23 and an ionization parameter of log 0 141414The spectral fits did not show an improvement using a warm-absorber model; however the 4 keV absorption was not well constrained since the Suzaku spectra start at rest-frame energies of keV. (see models 2–4 in Table 3). This low-ionization absorber is required to model the absorption detected below 2 keV (observed-frame). An additional complex absorber is required to fit the broad absorption features with rest-frame energies between 7–12 keV. Given the range of energies and variability of the broad absorption features, we interpret this absorption as a blend of highly ionized (2.75log 4) iron absorption lines blueshifted by an outflow. This explanation is consistent with recent models that attempt to simulate X-ray BALs in quasars (e.g., Schurch & Done, 2007). To test this interpretation we next employ more complex, but more realistic, models to fit the APM 08279+5255 spectra.
As a first attempt we fit the low and high-energy absorption of
APM 08279+5255 in epoch OBS1 with a model that includes a power-law (with
Galactic absorption) and one warm absorber (model XSTAR1,
Table 5). The warm-absorber model is calculated using
the XSTAR code (see, e.g., Kallman & Bautista, 2001; Kallman et al., 1996). XSTAR
calculates the physical conditions and absorption-emission spectra
of photoionized gases with variable abundances. In the current
analysis we use a recent implementation of the XSTAR code called
WARMABS that can be used as a model within XSPEC. For the WARMABS
model we assume turbulent velocities
(default velocity of the model).151515Since a warm absorber
with is expected to have a temperature of
K (e.g., Chelouche & Netzer, 2003), we do not expect a thermal
broadening higher than 100 km s. For the fits we assumed solar
abundances, a redshift of 3.91 for the warm absorber, and we left
the column density and ionization parameter of the warm absorber
free to vary in the fit. We note that spectral fits using model
XSTAR1 attempt to fit both the the low and high-energy absorption
of APM 08279+5255 with a single warm absorber.
To constrain the iron abundance (), we allowed this parameter to vary in the spectral fits in the model XSTAR2. The only difference between models XSTAR1 and XSTAR2 is that is fixed in model XSTAR1 and free to vary in model XSTAR2. We find that allowing to vary in our spectral fits does not lead to a significant improvement in the fits, and the best-fit values of in fits with model XSTAR2 are consistent with no iron over-abundance. Our results do not confirm an apparent iron over-abundance (2) claimed by Hasinger et al. (2002) and Ramírez (2008) based on their analyses of previous observations of APM 08279+5255.
We next assumed the spectral model XSTAR3 consisting of a power-law, one stationary ionized absorber with a turbulent velocity of and a second outflowing ionized absorber with a turbulent velocity of (the maximum value allowed by the model161616The high-enengy absorption features modeled with Gaussian absorption lines in model 7 could be the result of one or more highly ionized absorbers. The Doppler broadening velocities of each absorption line component of model 7 are (where =1,2 indicates the component). From Table 3 these Doppler broadening velocities are at first order comparable to the assumed values of .). We allowed the ionization parameters of both ionized absorbers and the redshift of the second ionized absorber to vary in the fit (model XSTAR3; Table 5). We find that the best fitted redshift of the second warm absorber is in both the FI and BI spectra. The -test indicates an improvement in the fits of OBS1 with model XSTAR3, that assumes two ionized absorbers, compared to fits with models XSTAR1 and XSTAR2, that assume a single ionized absorber, at the 99.5% of significance level in the FI and BI spectra. We conclude that a single warm-absorber model cannot accurately fit both the low and high-energy absorption in APM 08279+5255.
We finally assumed the spectral model XSTAR4 consisting of an absorbed power-law and two outflowing ionized absorbers. We assumed turbulent velocities of for the first and second outflowing ionized absorber (model XSTAR4; Table 5). The main difference between models XSTAR3 and XSTAR4 is that the redshift of the first warm-absorber is fixed in model XSTAR3 to the systemic redshift of the quasar and free to vary in model XSTAR4 and that model XSTAR4 includes a neutral absorber. For fits using model XSTAR4, we allow the redshifts, column densities, and ionization parameters of the absorbers to vary. The values for these fits are similar to those found for model 7 (see Table 5 and 3). We find on average best-fit redshifts of () and (), and column densities of and , where and correspond to the two warm absorbers. We confirm the results of model 7, by finding a significant change in the best-fitted redshift of the first warm-absorber component in model XSTAR4 between epochs OBS2 and OBS3 (see Table 5). Even though the two warm-absorber model results in acceptable fits, the best-fit parameters should only be considered as basic estimates of the wind properties since the kinematic and ionization structure of the outflow are likely to be more complicated. Our spectral fitting results of models that include ionized absorbers (see Table 5) indicate that both models XSTAR3 and XSTAR4 provide acceptable fits to the spectra of APM 08279+5255 for epochs OBS1 and OBS3, however, model XSTAR4 provides a better fit to the data for epoch OBS2 than model XSTAR3. The -test indicates that in epoch OBS2, spectral fits using model XSTAR4 provide an improvement over fits using model XSTAR3 at the level of significance in the FI and BI spectra.
3.2. PIN spectral analysis.
We also examined the spectrum of APM 08279+5255 in the 10–40 keV energy band using the PIN-HXD data. Unfortunately, no signal from the source above the background in the NXB was found. The background-subtracted source spectrum 171717The background-subtracted source spectrum is obtained by subtracting the non-X-ray-background and an estimate of the X-ray background from the total detected PIN spectrum was found to be within 5% of the non-X-ray background (NXB) spectrum provided by the Suzaku team (e.g., Mizuno et al., 2007). We arrive at a similar conclusion from the count rates presented in Table 2. The non-detection with the PIN provides an upper limit on the flux density at 20 keV of APM 08279+5255 of photons s. This limit is consistent with an extrapolation of the XIS spectrum.
The X-ray spectrum of APM 08279+5255 is known to contain absorption features at rest-frame energies above 7 keV (Chartas et al., 2002; Hasinger et al., 2002). These features have been interpreted in the past in two different ways. The first interpretation by Chartas et al. (2002) was based primarily on the analysis of the 2002 Chandra observation of APM 08279+5255 and posits that the absorption features are due to highly blueshifted Fe XXV K and/or Fe XXVI K absorption lines. An XMM-Newton observation of APM 08279+5255 performed 1.8 weeks (proper-time) after the Chandra observation showed significantly different high-energy absorption structure which was interpreted by Chartas et al. (2003) to imply variability of the absorption features over timescales of the order of weeks. The second interpretation by Hasinger et al. (2002) is based on the 2002 XMM-Newton observation of APM 08279+5255 and proposes that the high-energy absorption feature arises from an iron absorption edge produced by a metal enriched (Fe/O 2–5 Fe/O) ionized absorber. One important conclusion from section 3.1.2 is that the absorption feature found at 7–12 keV rest-frame can be fitted with two highly ionized blue-shifted warm absorbers that do not require super-solar metallicities. We note that Hasinger et al. (2002) and Ramírez (2008) had claimed iron over-abundance (2) based on their analyses of previous observations of APM 08279+5255. In support of the two-component “iron-blend-outflow” scenario we mention that the two absorption-line model and the notch model (models 7 and 5; Table 3) provide significantly better fits to the absorption feature between 7–12 keV rest-frame than an absorption-edge model in epoch OBS1 (see §3). We also note that, in the 2002 90 ks Chandra observation analyzed in Chartas et al. (2002), the model containing two absorption lines successfully fits the 7–12 keV rest-frame feature whereas an absorption-edge model did not provide an acceptable fit. The absorption described either by two absorption lines or a notch, may crudely represent absorption through an outflow with a large velocity gradient along the flow. Variability of the kinematic and ionization state of the outflow may explain why, depending on the observation, these absorption features could be modelled by either a notch, absorption lines, or an edge (see figure 4 of Schurch & Done, 2007). We conclude that a time-variable outflow provides a plausible explanation for all the past X-ray observations of the absorption features of APM 08279+5255 (Chartas et al., 2002; Hasinger et al., 2002) including those analyzed here. In §4.1 we provide plausible explanations for the observed variability of the high-energy absorption. In §4.2 we use the results of our spectral analysis to place constraints on the kinematics of the outflow and in §4.3 we provide estimates of the mass-outflow rate and efficiency of the outflow of APM 08279+5255.
4.1. Origin of the Variability of the High-Energy Absorption Feature.
Assuming the first interpretation of the origin of the high-energy absorption features in APM 08279+5255 the observed shift in the energy of the first absorption line between epochs OBS2 and OBS3 is likely due to a change in the outflow velocity of the absorber. Two alternative explanations of the shift are a change in the direction of the outflow (with respect to the line of sight) and a variation in the ionization parameter of the absorber. A change in the direction of the outflow is expected to show a shift in energy of both components of model 7 (Table 3). We only find variability in one component (abs1, see §3); however this picture could still be valid if each outflow component is driven independently. A change in the ionization parameter seems to be a less probable scenario. We checked this by fitting the spectra of epochs OBS2 and OBS3 simultaneously with model XSTAR4, keeping the redshift of abs1 as the only common parameter between the fits. In the case where a change in the ionization parameter produced the detected variability of the first absorption line abs1, we expect that the fits to the spectra of epochs OBS2 and OBS3 will not be improved by allowing the redshift of abs1 to vary independently in these fits. For the simultaneous fit to the spectra of epochs OBS2 and OBS3, where the redshift of abs1 was kept a common parameter, we find , 0.06, and . In the case where we fit the spectra of epochs OBS2 and OBS3 independently using model XSTAR4 (see Table 5) we find and .181818These values are obtained by summing the and degrees of freedom of epochs OBS2 and OBS3 in Table 5. The improvement based on the -test of fitting the spectra of epochs OBS2 and OBS3 independently, compared to keeping a common redshift of abs1, is at the 99% and 99.9% level of significance in the FI and BI spectra. We conclude that the variability of the energy of the first absorption line abs1 is likely not driven by changes in the ionization parameter of abs1. Suzaku cannot resolve the images of APM 08279+5255, however, the time-delays between the two brightest images A and B of APM 08279+5255 is estimated to be of the order of a few hours (e.g., Munoz et al., 2001), much shorter than the observed variability of the high-energy absorption feature. We therefore do not expect the combined X-ray spectrum of all images of APM 08279+5255 to differ from that of the individual images within the time-delay.
4.2. Constraints on the Kinematics of the Outflow.
Under the premise of the outflow interpretation to explain the absorption at rest-frame 7–12 keV we expect a continuous distribution of outflow velocities. This range of velocities leads to the the Doppler shift of the energies of the resonance absorption lines. The absorption-line rest-frame energies will thus be shifted to the observed energies according to
where is the Lorentz factor, is the angle
between the wind and our line of sight (l.o.s), and .
The minimum and maximum projected velocities (, ) of the outflow are estimated from the minimum and maximum energy ranges () of the high-energy absorption features in APM 08279+5255. We obtained and from our spectral fits assuming first the two absorption-line (APL+2AL) model and second assuming the notch (APL+No) model. Specifically, based on the best-fit values of the APL+2AL model (model 7; Table 3), we obtain and . From the best-fit values of the APL+No model (model 5; Table 3) we have and . The values of and are presented in Table 6. In this table the values of and are shown separately for the FI and BI spectra and for the two different models used to obtain them (model 7APL+2AL; model 5 APL+No).
The relative strengths of the iron resonance absorption lines will depend on the ionization state of the outflowing medium. To demonstrate this effect in a basic way we have performed several simulations using the warm-absorber model (WARMABS) of XSTAR. In Figure 3 we show a simulated absorbed spectrum in the 6–8 keV rest-frame energy range for an absorbing medium with solar composition, log =23, and having six different values of the ionization parameter (). The two strongest iron lines for this highly ionized absorbing medium have rest (or laboratory) energies of 6.70 keV (Fe xxv K; ) and 6.97 keV (Fe xxvi K ). In general the Fe xxv K line will be stronger than the Fe xxvi K line for a medium with . Therefore the absorption at the lower end of the 7–12 keV rest-frame range is most likely associated with the Fe xxv K line. Based on our interpretation of the high-energy absorption features we estimate assuming that the absorption at the low end of the X-ray BAL is due to a line arising from highly blueshifted Fe xxv K ( keV). On the other hand to estimate we make the conservative assumption that the absorption at the high end of the X-ray BAL is due to a line arising from highly blueshifted Fe xxvi K ( keV). The estimated values of and obtained through the procedure outlined above are presented in Table 6. These velocities are obtained using equation (2) assuming that our line-of-sight makes an angle of with the velocity of the outflow (Chartas et al., 2007a). A change of 10 in will introduce a variation of 10% in our estimates of the velocities. The velocities of the outflow range between 0.1 and 0.6. The mean value of from all CCDs and observations is 12.040.22 keV corresponding to a mean value of = 0.550.02. This maximum value of the outflow velocity constrains the angle between the outflow direction and our line of sight to be 36.191919The Doppler-shift formula (equation 2) predicts that given a fixed ratio of / the maximum angle between our line of sight and the wind direction is given by . This relatively small angle is consistent with outflow models that posit that BAL quasars are viewed through collimated outflows.
As argued in Chartas et al. (2002, 2003, 2007a) we are likely observing the X-ray absorbers as they are accelerated near their launching radii. We use the following equation to describe the basic dynamics of a radiation-driven outflow (e.g., equation 1 of Chartas et al., 2002):
where is the outflow velocity in units of , is the force multiplier, is the Eddington luminosity, is the radius (units of ) at which the wind is launched from the disk, and is the distance (units of ) from the central source. The expression for the dynamics of the outflow can be simplified by defining .
In Figure 4 we plot wind velocity versus radius (left panel) and wind velocity versus time (right panel) for an outflow launched at radii of (solid lines) and (dashed lines). In Figure 4 the units of velocity, distance, and time are , , and , respectively. Assuming (Irwin et al., 1998), and (e.g., Arav et al., 1994; Laor & Brandt, 2002), we have . is obtained using the vs. correlation found in RQ quasars (e.g., Wang et al., 2004; Shemmer et al., 2006, 2008) for . Therefore with (Irwin et al., 1998; Reichers et al., 2008)202020We note that an estimation of based on the optical and UV spectra should be more precise than an extrapolation of based on X-ray luminosities as it is done in Ramírez (2008). , where (Egami al., 2000)212121See, however, Reichers et al. (2008) that find a magnification of . Reichers et al. (2008) also use the observed width of the CIV line to obtain a black-hole mass of is the lens magnification factor. Equation 3 could be modified using a reliable SED describing the central source, adding relativistic corrections, and calculating the force multiplier at every point of the trajectory of the outflow. We stress, however, that our simplified approach is sufficient to provide first order approximations to the launching radius and the time scales involved in the dynamics of the outflow. Equation 3 can be written as ; therefore for . The latter expression can be used to obtain first order approximations of the launching radius given the velocity of the outflow.
Assuming a radiation-driven wind, it is expected that the time required to accelerate the outflow to fractions of is of the order of 10 (see Figure 4). For the black-hole mass of APM 08279+5255 of we estimate that the time to accelerate an absorber to near-relativistic velocities is weeks (rest-frame). We have reported in this work probable variability of the high-energy absorption features over a time-scale of 1 month (rest-frame). This short time-scale variability is consistent with the expected variability timescale of a radiation-driven wind.
4.3. Constraints on Mass-Outflow Rate and Efficiency of the Outflow.
Based on our estimated values of the outflow velocities, column densities, and launching radii we present constraints on the mass-outflow rates and outflow efficiency associated with the outflowing X-ray absorbers of APM 08279+5255. The efficiency is defined as the ratio of the rate of kinetic energy injected into the ISM and IGM by the outflow to the quasar’s bolometric luminosity, i.e.,
where is the covering fraction, is the column density, is the radius, and is the thickness of the absorber. To estimate the efficiency we use the two absorption-line model (APL+2AL; model 7 of Table 3). We calculate the bulk velocities of each outflow component based on the energies of the absorption lines in model 7 and through the use of equation (2) with =6.7 keV and . As in Chartas et al. (2002, 2003) we interpret the high-energy absorption features as being due to highly ionized Fe (Fe xxv K) in a gas with solar abundances, and we estimate log using a curve-of-growth analysis. In Table 7 we present the outflow velocities, , the column densities, log , the mass-outflow rates, , and the outflow efficiencies, , of the two modeled absorbers of the outflow. We note that the values of the column densities and velocities in Table 7 are consistent with those found using the photoionization code XSTAR (see Table 5).222222The velocities obtained in Table 5 assume the redshifts of the absorbers in model XSTAR4 are due to the relativistic Doppler effect (see equation 2).
|OBS||Instr.||log (abs1)||(abs1)||(abs1)||log (abs2)||(abs2)||(abs2)|