A&A 483, 415-424 (2008)
M. Krumpe1 - G. Lamer1 - A. Corral2 - A. D. Schwope1 - F. J. Carrera2 - X. Barcons2 - M. Page3 - S. Mateos4 - J. A. Tedds4 - M. G. Watson4
1 - Astrophysikalisches Institut Potsdam, An der Sternwarte 16, 14482 Potsdam, Germany
2 - Instituto de Física de Cantabria (CSIC-UC), Avenida de los Castros, 39005 Santander, Spain
3 - Mullard Space Science Laboratory, Holmbury St. Mary, Dorking, UK
4 - Department of Physics & Astronomy, University of Leicester, Leicester LE1 7RH, UK
Received 6 November 2007 / Accepted 3 March 2008
Aims. We present the results of the X-ray spectral analysis of an XMM-Newton-selected type II QSO sample with and 0.5-10 keV flux of erg/s/cm2. The distribution of absorbing column densities in type II QSOs is investigated and the dependence of absorption on X-ray luminosity and redshift is studied.
Methods. We inspected 51 spectroscopically classified type II QSO candidates from the XMM-Newton Marano field survey, the XMM-Newton-2dF wide angle survey (XWAS), and the AXIS survey to set-up a well-defined sample with secure optical type II identifications. Fourteen type II QSOs were classified and an X-ray spectral analysis performed. Since most of our sources have only 40 X-ray counts (PN-detector), we carefully studied the fit results of the simulated X-ray spectra as a function of fit statistic and binning method. We determined that fitting the spectra with the Cash-statistic and a binning of minimum one count per bin recovers the input values of the simulated X-ray spectra best. Above 100 PN counts, the free fits of the spectrum's slope and absorbing hydrogen column density are reliable.
Results. We find only moderate absorption ( cm-2) and no obvious trends with redshift and intrinsic X-ray luminosity. In a few cases a Compton-thick absorber cannot be excluded. Two type II objects with no X-ray absorption were discovered. We find no evidence for an intrinsic separation between type II AGN and high X-ray luminosity type II QSO in terms of absorption. The stacked X-ray spectrum of our 14 type II QSOs shows no iron K line. In contrast, the stack of the 8 type II AGN reveals a very prominent iron K line at an energy of 6.6 keV and an keV.
Key words: surveys - X-rays: galaxies - galaxies: active - galaxies: quasar: general
The unification model for active galaxy nuclei (AGN) is based on the assumption that the Seyfert type I/type II dichotomy in AGN is a result of varying orientation relative to the line of sight of similar objects (Antonucci 1993). All AGN consist of a super-massive black hole with an accretion disk. This central engine is surrounded by optically thick, toroidally concentrated dusty material. An observer can either view the central engine directly (type I AGN) or through the optically thick torus (type II AGN). Direct observation of the broad line region reveals a strong UV continuum and broad permitted emission lines while the optical spectra of type II AGN show narrow permitted and forbidden emission lines. Hao et al. (2005) provided a formal separation criterion by finding a bimodal distribution in the H-FWHM with a significant dip at FWHM(H km s-1.
Kleinmann et al. (1988) discovered the first type II quasi stellar object (QSO). QSOs show higher intrinsic luminosity than AGN. In the optical the conventional dividing line between Seyfert galaxies and QSOs is . Despite their high intrinsic luminosities, type II QSOs are very hard to identify. A significant fraction of the emitted power is absorbed by the optically thick torus. Furthermore, their lack of emission lines in a wide optical wavelength range hampers their identification. About 150 optically selected type II QSOs from the Sloan Digital Sky Survey were studied in Zakamska et al. (2003). However type II QSOs are more efficiently found in follow-up observations of X-ray surveys (Szokoly et al. 2004; Mainieri et al. 2002; Krumpe et al. 2007; Barcons et al. 2007; Tedds et al. 2008).
Type II QSOs play an important role in understanding the X-ray universe. Since they show significant absorption, type II QSOs/AGN are considered to be a main contributor to the hard X-ray background and their existence in considerable numbers is needed for the synthesis of the X-ray background (Gilli et al. 2007). Possible evolution of the absorption with intrinsic luminosity and/or redshift is a matter of intensive debate and would essentially influence the X-ray background synthesis models.
Many papers discuss the fraction of absorbed AGN as a function of luminosity and/or redshift, and this is a highly controversial subject (Dwelly & Page 2006; Tozzi et al. 2006; Akylas et al. 2006). However, there are only a few studies (e.g. La Franca et al. 2005) which investigate the evolution of absorbing column densities in type II AGN up to high redshift. Other studies focus on the local universe, where the existence of so-called Compton-thick absorbed objects is well known (Maiolino et al. 1998; Bassani et al. 1999; Risaliti et al. 2000). A Compton-thick object is absorbed by column densities of cm-2, so that the cross section of Compton scattering overcomes the photo-electric absorption. Hence, the reflected X-ray component is observable.
There is evidence for a large fraction of Compton-thick absorbed quasars at high redshifts (Martinez-Sansigre et al. 2007), but only one candidate has been reported so far at high redshift (Norman et al. 2002), and even in this case the column density is quite uncertain. Most of the X-ray sources from deep Chandra and XMM-Newton surveys are absorbed by column densities , with hard X-ray luminosities of (Comastri 2004; Szokoly et al. 2004; Mainieri et al. 2002; Mateos et al. 2005). These values are also found in Compton-thin absorbed ( ) Seyfert II galaxies at low redshifts.
Table 1: Observed properties of type II QSO candidate sample.
The definition of type II QSOs is somewhat arbitrary. Zakamska et al. (2003) define a type II QSO only based on its optical properties. They select AGN with narrow permitted emission lines ( km s-1) and classify objects with [O III] 5008 line luminosity of as type II QSO. Mainieri et al. (2002) introduce the term ``type II QSO region''. Although their type I/II classification is based on optical spectra, a type II QSO has to have an intrinsic X-ray luminosity of erg/s (0.5-10 keV) and an absorbing hydrogen column density of cm-2. Throughout the paper we use the definition of a type II QSO to be an object with narrow forbidden and narrow permitted emission lines in the optical spectrum, as well as a de-absorbed intrinsic X-ray luminosity erg/s (0.5-10 keV band). However, the Seyfert type II classification of our sample objects is strictly based on the optical spectra.
The present work focuses on properties of the high X-ray luminosity type II QSOs found in medium deep surveys performed by us. The paper is organised as follows. In Sect. 2 we describe how we selected our type II QSO candidate sample and we study the properties in Sect. 3. The X-ray data and the extraction of the X-ray spectra are described in Sect. 4. Since X-ray spectra with small numbers of counts have to be fitted, we studied different fitting and binning methods in Sect. 5 and then describe the analysis of the X-ray spectra. The results are discussed in Sect. 6. Finally, our conclusions are outlined in Sect. 7.
Unless mentioned otherwise, all errors refer to a confidence interval. We assume H0=70 km s-1 Mpc-1, , and .
We began with a list of 51 spectroscopically classified type II QSO candidates which are associated with X-ray sources. All optical counterparts have and observed X-ray luminosities (not corrected for intrinsic absorption) of log ( . The type II QSO candidates were taken from three different X-ray surveys, the XMM-Newton Marano field survey (Krumpe et al. 2007), the XMM-2dF wide angle survey (XWAS, Tedds et al. 2008), and from ``An XMM-Newton International Survey'' (AXIS, Barcons et al. 2002,2007; Carrera et al. 2007). To obtain a well-defined sample with secure optical type II identifications and X-ray counterparts we applied a two stage process. First, we visually inspected the optical spectra of all type II QSO candidates from all X-ray surveys and determined the FWHM of the emission lines. The type II QSO candidate classification was only based on the optical spectra. The spectral resolution of the optical setup was 1050 km s-1 for the Marano field survey, 700 km s-1 for XWAS, and 300-600 km s-1 for AXIS. Therefore, an intrinsic line of 1200 km s-1 was observed with 1250-1600 km s-1. We defined emission lines with a measured resolution (including instrumental resolution) of km s-1as narrow.
The optical spectra were only used to determine a secure Seyfert type of the objects. They were not used to separate between AGN and QSOs. None of the 51 type II QSO candidates showed obvious broad emission lines in the optical spectra. However, since we aimed for a very strict type II QSO sample, we excluded many objects that showed only the low excitation O II line. The optical spectrum had to comprise several emission lines that allowed us to establish an optically secure classification of the Seyfert type. In addition the signal-to-noise ratio had to be appropriate in order to verify the non-existence of broad permitted emission line. We excluded all doubtful cases.
We introduced two different categories of type II QSO candidates based on the reliability of the type II classification. Objects that have optical spectra with high excitation lines and at least one permitted AGN emission line that is detected but narrow (e.g. Ly-, C IV, H) are marked with the optical flag ``secure'' (see Table 1) - signifying a secure optical identification. Their optical classification is very robust.
Objects with less secure type II identification belong to the tentative sample (optical flag ``tentative''). The classification as tentative object can be due to the following reasons:
As a second step to setting up a well-defined type II QSO sample,
we verified that the spectroscopically identified counterpart is
associated with the X-ray source. For a significant number of objects much
deeper imaging data have become available subsequent to the epoch when the
spectra were obtained. For example, the XWAS optical counterparts were originally selected from
SuperCosmos optical imaging survey data (Hambly et al. 2001).
For the majority of the objects we now have additional, deeper imaging data in different bands.
These imaging data were obtained with the Wide Field Imager (WFI, 2.2 m
telescope at La Silla) or the Wide Field Camera (WFC, 2.5 m Isaac Newton telescope
on La Palma).
We visually investigated the best available
imaging data (WFI, WFC, SuperCosmos) for additional
optical counterparts and rejected doubtful counterpart identifications.
Furthermore, we computed the probability that the optical object with the
is associated with the X-ray source. This was based on
After the X-ray counterpart verification our sample consists of 22 sources including 13 secure and 9 tentative objects. The separation between AGN and QSOs was made after the determination of the de-absorbed intrinsic X-ray luminosity (see Sect. 5).
Comments on the optical spectra of tentative objects:
Marano 32A - all narrow emission lines have underlying broad (mainly blue-shifted) components.
Marano 47A - the signal-to-noise ratio of the optical spectrum does not allow the exclusion of the presence of a weak, broad H emission line.
Marano 50A - H and [O III] not covered by the spectrum, however, most likely a type II object since no Mg II but [O II] and [Ne V] emission lines.
Marano 51A - H and [O III] covered by the spectrum but H coincides with the edge of the atmospheric A-Band. However, a strong broad H emission line can be ruled out and [O II], [Ne V] emission lines are visible.
Marano 66A - SNR of the optical spectra does not exclude the presence of a weak, broad H emission line ([O II] and [Ne V] emission).
Marano 116A - type II object/X-ray bright optical normal galaxy (XBONG) - galaxy spectrum with prominent [O II] emission line, no Mg II, very weak [O III] emission lines, H coincides with the edge of the A-Band.
Marano 133A - some narrow emission lines have underlying broad, blue-shifted components.
Marano 253A - type II object/XBONG - no Mg II and H emission, weak [O III] emission, the second [O III] line falls into the atmospheric A-band.
sds1b-014 - high SNR spectrum, H and [O III] not covered by the spectrum, however most likely a type II object since no Mg II but [O II] and [Ne V] emission lines.
In Table 1 we summarise the observed properties of the objects. We list the name of the spectroscopically identified counterpart, optical coordinates, distance between spectroscopically identified counterpart and X-ray source, the WFI-R-band magnitude (unless otherwise mentioned), redshift, optical flag, X-ray counterpart probability, count rate, 0.5-10 keV flux, and the Galactic absorption along the line of sight.
The objects cover a redshift range of . Since the lack of emission lines in a wide optical wavelength range hampers the identification of type II objects, we only have one type II QSO candidate in the redshift interval of z=1-2. Figure 1 shows the R-band magnitude histogram of the selected objects. As a simplification we consider the r-band (SDSS) magnitude to be equal to the R-band magnitude, although we are aware that shifts of 0.5 mag between R and r (SDSS) may occur. The majority of our sources have 21<R<24. The 0.5-10 keV flux in the sample ranges from erg/s/cm2.
|Figure 1: R-band magnitude histogram of all 22 type II QSO candidates (black solid line) and of the 13 secure type II QSO candidates (grey filled histogram).|
|Open with DEXTER|
|Figure 2: Observed 0.5-10 keV X-ray flux vs. R-band magnitude. Large symbols represent secure type II QSO candidates while small symbols illustrate tentative type II QSO candidates. Dashed lines indicate values of 0.1, 1, and 10.|
|Open with DEXTER|
To calculate values we derived the optical fluxes in a band centred at 7000 Å with a width of 1000 Å using the equation (Zombeck 1990). X-ray fluxes were calculated for the 0.5-10 keV energy range. The R-band magnitude vs. X-ray flux plane (Fig. 2) clearly shows AGN activity in all selected sources, since they have X-ray-to-optical flux ratios of . In accordance with Mainieri et al. (2002) we consider objects with as AGN. Almost half of the selected type II QSO candidates have noticeable high X-ray-to-optical flux ratios ( ). Szokoly et al. (2004) mentioned that type II AGN/QSOs cluster at higher X-ray-to-optical flux ratios than type I AGN. The majority of the spectroscopically classified type II AGN/QSOs in the Lockman Hole also show high values (Mainieri et al. 2002). The observed 0.5-10 keV X-ray luminosities (not corrected for intrinsic absorption) of our objects range from erg/s.
All data were processed with the SAS version 7.0 (Science Analysis Software, Gabriel et al. 2004) package and corresponding calibration files. The epchain and emchain tasks were used for generating linearised event lists from the raw PN and MOS data. For all sources the effects of photon pile-up was negligible.
The XMM-Newton data reduction for the Marano field survey is described in detail in Krumpe et al. (2007). For the XWAS and AXIS sources we downloaded all available X-ray data from the XMM-Newton archive up to and including July 2007. Periods of high background were excluded from the analysis of all relevant data sets in the standard way.
Circular or box-shaped source and background regions were manually determined for all contributing observations. Sources at large off-axis angles in the contributing observations were not considered as follows. The largest offaxis angles were 720 arcsec for the PN and 820 arcsec for the MOS detectors, respectively. The lower PN area was a result of the enhanced background contamination near the edges of the PN detector. Additional X-ray sources in the background regions were masked out. The auxiliary response file (arf) was computed for each source and observation individually. For the Marano field sources, we used appropriately ``canned'' response matrices from the XMM-Newton calibration homepage (XMM-revolution 110 and pattern 0-12). The PN y-coordinate of an X-ray source was determined to link the relevant PN response matrix file (rmf for single and double events, version 6.8) to the X-ray spectrum of the source. For the XWAS and AXIS sources the response matrix files were computed on a case-by-case basis. Where multiple X-ray spectra of a given source were added, the mean rmf was computed as a weighted mean. For details of the procedure see Page et al. (2003). The two MOS spectra were always added to form a single MOS spectrum.
The X-ray source of the secure type II object phl5200-001 is surrounded by diffuse X-ray emission. For the reduction of the X-ray spectrum we used a smaller extraction radius for the point source and the diffuse X-ray emission as the background region.
The X-ray spectral analysis was performed with XSPEC (Arnaud 1996) version 12.3.0. Although we have a few objects with several hundreds of net PN source counts in the 0.2-8 keV range, the distribution peaks at 40 net PN source counts. An appropriate X-ray spectral analysis for the low count regime has to be found. Although we only refer to the net PN counts in our simulations and in Table 4, the fit uses both the PN and MOS data. Hence, the total number of counts used by the fit is typically twice that given in Table 4.
A problem notorious to X-ray astronomy is proper fitting of spectra with relatively few counts. Tozzi et al. (2006) approached this problem for the X-ray sources in the Chandra Deep Field South by running simulations for two different fitting procedures: Cash-statistics (unbinned) and classic -statistics with a binning of 10 counts per bin (min 10). They concluded that the unbinned Cash-statistic fits recovered the input values better for X-ray spectra with less than 50 counts.
We carried out a much more extensive study of the fit results as a function of fit statistic and binning method. This investigation can be used to study the error distribution of the intrinsic column density () and to determine how many PN net counts are required to perform free fits of and photon index with acceptable errors in both parameters.
We assumed emission from the AGN to be described as a power law with photon index , modulated by Galactic foreground absorption and further intrinsic cold absorption at the redshift of the AGN. Following Mainieri et al. (2002) and Mateos et al. (2005), we simulated X-ray spectra with an input value of . All simulated X-ray spectra included Galactic absorption of cm-2 and considered all possible parameter combinations of Table 2.
The X-ray spectra were normalised to reach the desired 0.2-8 keV PN net counts with a deviation of up to 5%. We added Poisson noise to the X-ray spectra. We used a typical representative background file (Marano 9A). As for the real data we added MOS1 and MOS2 spectra to form a single MOS spectrum. The same source was simulated 1000 times for each set of specific parameters.
The simulated X-ray spectra were grouped by using different binning methods (Table 3). Cash and -statistic were applied to recover the input value. The redshift and the Galactic absorption were set to the input value. The nominal initial guess for the intrinsic absorption was cm-2, but other initial guesses up to cm-2 were tested as well.
Table 2: Set of input parameters for the simulated X-ray spectra.
Table 3: Fit statistics and binning methods for the simulated X-rayspectra.
|Figure 3: Recovered distribution for an X-ray source with 40 net PN counts at z=2. The input cm-2 is indicated by the dashed line. Fit method used: fixed , Cash-statistic, binning: min 1; 1000 simulations.|
|Open with DEXTER|
|Figure 4: Recovered distribution of an unabsorbed X-ray source with 40 net PN counts at z=1. The input cm-2 is indicated by the dashed line. cm-2 is represented by the lowest bin. Fit method used: fixed , Cash-statistic, binning: min 1; 1000 simulations.|
|Open with DEXTER|
|Figure 5: Diagnostic vs. redshift plot. The plotted contours contain 68% (grey) and 90% (black) of the simulated X-ray sources with 40 net PN counts in the 0.2-8 keV band (NOTE: The fits used more than twice as many counts since both the PN and MOS data are used). The solid lines are the limits for intrinsically unabsorbed sources. The limits for an intrinsic absorption of cm-2 are shown by dashed lines while dotted lines represent cm-2.|
|Open with DEXTER|
Following our analysis in Sect. 5.1 we grouped the extracted X-ray spectra in bins of at least one count per bin (0.2-8 keV) and used the Cash-statistic to determine . The initial guess was set to . We corrected for Galactic absorption and performed free fits in and for spectra with more than 100 net PN counts. For X-ray spectra with less than 100 net PN counts we fixed and only fit . In these cases we give two errors for the value of in Table 4. The first is the 1 error of based on the fit with fixed , while the second takes into account the systematic shift for different photon indices (deviation in for a fixed and ). The observed (not corrected for intrinsic absorption) and the intrinsic 0.5-10 keV X-ray luminosities were computed by either a free fit in and for the bright sources (net PN counts >100) or a fixed fit for the faint sources.
Due to contamination by soft detector background we only used the 0.3-8 keV band for the X-ray spectral analysis of X-ray source X03246_092 (24 net PN counts in 0.3-8 keV band). The soft energies in the X-ray spectra of Marano 51A and Marano 610A could not be well fitted by an single absorbed power law. In these cases a fit of an absorbed power law plus a soft excess component reproduce the X-ray data better. Therefore, we used a two component fit, an absorbed power law and an unabsorbed power law. No evidence for soft excess was found in the X-ray spectra of the brightest objects. The best fit models of the X-ray data are shown in Appendix A.
Table 4 shows the spectral parameters. In the columns we list the name of the spectroscopically identified counterpart (1), the XMMU source name (2) of the XMM-Newton X-ray source, the observed (3) and intrinsic (4) X-ray luminosity, the PN count number (5) in the 0.2-8 keV band, the absorbing hydrogen column density (6), the photon index (7), and the quality for the used fit (8) when the X-ray data are modelled by a power law plus intrinsic absorption model, and finally, the photon index (9) and the quality for a purely reflection-dominated model (10). If no error is given for the photon index, this parameter was fixed for the fitting.
Table 4: Computed properties of type II QSO candidate sample.
In Fig. 6 we show the fitted intrinsic distribution of the sources. A column density peak at cm-2 is found. We find moderate absorption in the majority of our objects. The significance of the absorption exceeds 2 in most of the cases (see the confidence contours in Appendix A). Two type II objects are consistent with being unabsorbed X-ray sources, one of which is from the secure type II sample.
|Figure 6: Intrinsic distribution of all 22 type II objects (black) and of the 14 type II QSOs (grey filled histogram).|
|Open with DEXTER|
The determination of the intrinsic also allows us to compute the de-absorbed intrinsic X-ray luminosity of our sources. As mentioned in Sect. 1, we define a type II object by the detection of narrow emission lines in the optical spectrum. Figure 7 shows the intrinsic column density vs. de-absorbed intrinsic X-ray luminosity plane. The dividing line of erg/s is used to distinguish between type II AGN and QSOs. Ten type II objects from the secure sample and four tentative objects are identified as QSOs. We detected one unabsorbed type II QSO but the object belongs to the tentative sample. No obvious trend of absorption in type II QSOs with intrinsic X-ray luminosity is found.
Most of the type II QSOs fall into the same region of the diagram where previous studies have also found type II QSOs (Mainieri et al. 2002; Szokoly et al. 2004; La Franca et al. 2005; Ptak et al. 2006). The additional criterion ( cm-2, Mainieri et al. 2002) makes only a small difference to our sample selection (two more secure objects).
|Figure 7: Intrinsic vs. de-absorbed intrinsic 0.5-10 keV X-ray luminosity. Large symbols represent optical secure type II QSO candidates while small symbols illustrate the tentative sample. Open symbols indicate X-ray sources that have less than 40 net PN counts in the 0.2-8 keV band; filled symbols 40 net PN counts. The vertical solid line at log marks the dividing line between AGN and high luminosity QSOs. Objects with log and cm-2 (upper right corner) fall in the ``type II QSO region'' as defined by Mainieri et al. (2002).|
|Open with DEXTER|
|Figure 8: Intrinsic vs. redshift diagram for type II QSOs ( top panel) and type II AGN ( lower panel). Labels as in Fig. 7. The dashed and dotted lines represent the values which correspond to a 30% and 50% decrease in the 0.5 keV flux at the given redshift, respectively.|
|Open with DEXTER|
The -redshift plane for type II QSOs and type II AGN is shown in Fig. 8. We found no obvious differences between the distribution of AGN and QSOs. Although a tentative anticorrelation of vs. redshift below z<1 and a direct correlation above z=1 is indicated, when all data is taken together there is no significant trend in with redshift. Considering the typical flux and column densities found for our sources we expected all z>1 type II objects to be classified as QSOs. At lower redshifts 40% of the type II QSO candidates are actually type II QSOs.
Statistical fluctuations in the X-ray spectrum can lead to high values of spuriously measured values at high redshifts (e.g. Akylas et al. 2006). However, Fig. 5 show that objects with intrinsic of several 1022-1023 cm-2 are not significantly influenced by systematic trends in redshift. The fit method is also able to pick up much higher absorptions than found in our sources. The scatter in the mentioned range at z>2.5 is consistent with an intrinsic absorption of cm-2 for all sources.
As a further test of the impact of statistical fluctuations, we plot in Fig. 8 the amount of that is needed to reduce the 0.5 keV flux by 30% and 50%. The 30%-line agrees well with the 90% contours of unabsorbed sources in Fig. 5. Consequently, even for the sources at z>2.5 the fitted values are unlikely to be caused by statistical fluctuations.
The previous conclusions depend on not having misinterpreted Compton-thick absorbed objects ( cm-2) as moderately absorbed objects. Increasing column densities cause a hardening of the X-ray spectrum. This trend stops when the material becomes optically thick to electron scattering. The X-ray spectrum of a Compton-thick absorbed source is completely dominated by the reflection component. Their X-ray spectra show soft X-ray radiation but with a much lower photon index ( , Maiolino et al. 1998; Bassani et al. 1999; Risaliti et al. 2000; Iwasawa et al. 2001; Comastri 2004). We simulated an X-ray spectrum with as observed in the 3-12 keV band for NGC 1068. When we fit the spectrum with a fixed , we recover values for absorption of cm-2, cm-2 and cm-2 for redshifts of z=1,2,3 respectively. Hence, moderately absorbed sources are found if intrinsically Compton-thick absorbed sources are studied. However, we are able to exclude a general misinterpretation of the spectra based on the following arguments.
First we tested if the X-ray spectra are better modelled with a purely reflection-dominated model (pexrav). We used solar abundances and a cutoff energy of 100 keV. The ratio between the reflected and direct component (fit parameter: scaling factor) was set to 100 in order to obtain pure reflection. The cosine of the inclination was left as a free parameter. Again, we fitted the photon index as a free parameter only in the case the X-ray source had more than 100 PN net counts.
Only the data of X-ray source X01135_126 and X03246_092 are slightly better represented by the reflection model in comparison to a single power law with intrinsic absorption. Further four X-ray sources (Marano 51A, 63A, 66A, 610A) can be equally well modelled with a reflection model or a power law with intrinsic absorption (see Table 4). However, most of these objects have PN net counts of <50. Only X-ray source X01135_126 has . Low count spectra have the obvious problem that their X-ray data can be well fitted by different models. Nevertheless, the majority of our objects is only adequately fitted by a model that includes intrinsic absorption.
Secondly the reflected component of Compton-thick absorbed sources is 50-150 times weaker than the de-absorbed intrinsic X-ray luminosity (Brandt & Hasinger 2005). Hence they are usually found at low values of (Bassani et al. 1999). For the majority of our objects the [O III] line is redshifted out of our spectral range. As a second best estimator we adopt rather than . Figure 2 shows that almost all objects have high which is inconsistent with Compton-thick absorbed objects. Only X00851_154 has a rather low ratio, but the best X-ray spectral fit is a moderately absorbed power law with . Under the assumption of Compton-thick absorption, all studied objects would be classified as QSOs and our sample would contain the most X-ray luminous QSOs ever studied ( ).
|Figure 9: Stacked X-ray spectrum of all 14 type II QSOs ( upper panel) and 8 type II AGN ( lower panel). The QSOs stack contains 1600 counts in the shown 4-8.2 keV energy range, the AGN stack 300 counts. The squares represent the ratio between the flux of the stacked spectrum and the simulated continuum spectrum. The solid line shows the 1 detection error while the dotted line represents the 2 detection error. The solid grey line at a ratio of one illustrates the continuum flux.|
|Open with DEXTER|
Another indication for Compton-thick absorption is the detection of the iron line which is normally outshone by the direct component. Due to the reflection of the X-ray radiation from cold material in the torus, an iron K fluorescence line with an equivalent width keV is expected (Turner et al. 1997; Risaliti 2002).
The low count numbers in our X-ray spectra do not allow the study of individual spectral features such as an iron line. Based on the assumption that the intrinsic X-ray properties in the sample are similar, a stack of the individual spectra can reveal the presence of an iron line. Since the number of individual X-ray spectra increases the SNR in the stacked spectrum, we decided to include all 14 type II QSOs regardless of the optical flag. The assumption is justified because the secure and tentative sample do not show differences in (Fig. 2) and absorption values (Fig. 6).
Corral et al. (2008) explain the stacking process in detail. Here we outline the essential components of the process. Using the best-fit parameters, unfolded spectra (eufspec in XSPEC) from the ungrouped observed X-ray spectra (MOS and PN data) are extracted. The unfolded spectra are corrected for Galactic foreground absorption, shifted to the rest-frame and normalised to the same rest-frame flux in the 2-5 keV band. The flux is rebinned into a common energy grid for all spectra. The stacked spectrum is binned to a final energy grid that contains at least 100 counts per energy bin. The final errors of the stacked spectrum are based on Gaussian propagation of the errors in the individual spectra. To distinguish between real spectral features and artifacts from the averaging process the underlying continuum has to be modelled. Each source was simulated 100 times using the same model (absorption plus powerlaw) with the same 2-8 keV flux as observed in the real spectra. The individual simulated X-ray continuum spectra from all different X-ray sources are stacked exactly as the observed X-ray spectra which are used to determine the 1 and 2 contours (68% and 95% of all simulated continuum spectra).
The 0.5-10 keV final stacked spectrum of the QSOs contains 4300 counts. Together with the 1 and 2 contours the ratio of the observed spectrum to the average simulated continuum is shown in Fig. 9, upper panel. The stacked type II QSO spectrum shows no residuals around 6-7 keV which would correspond to an iron K emission line. The SNR in the stacked spectrum is sufficient to detect an iron line with keV. The stacked spectrum does not provide any evidence for Compton-thick absorbed sources based on a line emission and supports the assumption that our type II QSOs are actually moderately absorbed sources.
Interestingly, the stacked spectrum of our 8 type II AGN does show a very prominent emission line (Fig. 9, lower panel), although in none of the individual AGN X-ray spectra a significant excess at 6-7 keV is recognised (see the X-ray spectra in the Appendix A). Since only a few objects are included in the AGN stack we used energy bins containing 30 counts instead of 100 counts. We fitted the positive residuals in the stacked AGN spectrum with a Gaussian line profile. A line is detected with a significance above 2 at a line energy of E=6.6-0.08+0.07 keV with a eV. The equivalent width of the fit is EW=2.0-0.7+0.6 keV.
In principle there are two possible explanations for the detection of a broad iron fluorescence line.
The selection of both optical, narrow high excitation emission lines and intrinsic X-ray luminosities erg/s yielded 14 type II QSOs.
Since the distribution of net PN counts peaks at 40, the X-ray spectral analysis has to account for this very low numbers of counts. We extensively simulated and studied such low count number X-ray spectra. A binning with at least one count per bin combined with the Cash-statistic recovered the input values best. We proved that the method is able to find absorption up to cm-2.
However, we discover only moderately absorbed type II QSOs. One QSO is consistent with being unabsorbed but it belongs to the tentative sample. Compton-thick absorbed sources may be detected as moderately absorbed but we can exclude this scenario for the majority of our sources. The X-ray data are not well fitted by reflection models. The values of the objects and the non-detection of a broad iron line in the stacked type II QSOs spectrum give evidence that we have not misclassified Compton-thick absorbed sources. In contrary to the QSO stack, the stack of 8 type II AGN revealed a very prominent iron line with an keV. However, the shape of the single X-ray spectra and the modelling of the X-ray data showed that the majority of our AGN are most probably moderately absorbed X-ray sources.
The column density distribution found by us agrees well with those in deep Chandra and XMM-Newton surveys (Mainieri et al. 2002; Szokoly et al. 2004; Mateos et al. 2005; Ptak et al. 2006). These authors also reported a few cases of unabsorbed type II AGNs, as well as evidence for some heavily absorbed sources ( cm-2). La Franca et al. (2005) studied column density trends with X-ray luminosity in different redshift bins. They classified type II AGN as all sources that do not show any emission lines with km s-1. Independent of X-ray luminosity and redshift bins, they discovered type II AGN with an average absorption of cm-2, slightly above the column density peak of our survey ( cm-2). The present study of strictly classified type II AGN/QSOs, based on the optical spectra, verified that there is no obvious trend of absorbing column density with redshift or X-ray luminosity.
Our results apparently contradict studies of the local universe. Bassani et al. (1999) and Risaliti et al. (2000) found 75% of their sources with high absorption ( cm-2) and 25-45% with cm-2. Either the column density properties change dramatically from the local to distant universe or the majority of the heavily absorbed distant sources are missed in our surveys. Indeed even the most luminous Compton-thick absorbed sources in the local universe, e.g. NGC 6240, could not have been detected by our survey if they were at .
In summary, the column densities in our survey show no trend in X-ray luminosity and no clear trend in redshift. If we compare our results with samples of AGN in the local universe, our sample does not contain a significant fraction of heavily absorbed sources ( cm-2). Our survey of objects with is limited to observed X-ray luminosities in excess of erg/s. We can only expect to find QSOs intrinsically more luminous than erg/s, if they are Compton-thick absorbed. Norman et al. (2002) claim that for the most distant type II QSO ever detected (z=3.7), there is strong evidence for heavy or even Compton-thick absorption. However, our survey rules out large numbers of Compton-thick absorbed sources with X-ray luminosities of erg/s. Hence, potential Compton-thick absorbed objects at high redshifts are likely to have similar X-ray luminosities to Compton-thick absorbed objects in the local universe. In order to find a supposed, rare population of very luminous, Compton-thick absorbed QSOs a larger survey area is needed. The 2XMM catalogue (Watson et al. 2008) with a survey area of 360 deg2 could provide a valuable source to reveal such a population.
Mirko Krumpe is supported by the Deutsches Zentrum für Luft- und Raumfahrt (DLR) GmbH under contract No. FKZ 50 OR 0404. Georg Lamer acknowledges support by the Deutsches Zentrum für Luft- und Raumfahrt (DLR) GmbH under contract no. FKZ 50 OX 0201. Amalia Corral, Francisco J. Carrera, and Xavier Barcons acknowledge financial support by the Spanish Ministry of Education and Science, through projects ESP2006-13608-C02-01. M.P., S.M., J.T. and M.W. thank STFC for financial support.
Optical spectra: Optical atmospheric absorption corrected, wavelength and
flux calibrated spectrum for the optical X-ray counterpart is shown. All spectra
flux units of 10-18 erg cm-2 s-1 Å-1. The black solid
line represents the spectrum, the green solid line the error spectrum (not
available in all spectra). Red markers indicate possible emission lines,
blue markers absorption lines. There are exceptions in a few spectra.
Spectral features at 5580 Å are spurious due to incomplete subtraction
of a night sky line.
The X-ray spectra show the PN (black) and combined MOS (red) data, as well as the best fit model (foreground Galactic absorption, power law with intrinsic absorption at the object's redshift). In the case of Marano 51A and Marano 610A the shown best fit model consists of foreground Galactic absorption, an unabsorbed power law, and an power law with intrinsic absorption at the object's redshift to account for the soft excess in the X-ray data.
For objects that have less than 100 PN source
counts in the 0.2-8 keV band we fixed the photon index
(). Otherwise, free fits in
The fit parameters are given in Table 4. For illustration
purposes the X-ray data have been rebinned to different signal-to-noise ratio,
after grouping to a minimum of one count per bin (min 1).
Confidence contours: Confidence contours of the absorbing column density in units
cm-2 vs. photon index
of the X-ray
spectral fits. Confidence contours are plotted for 68% (black), 90% (red) and 99%
(green). The contours are based on free
fits for all objects independent of PN net counts.
Comments: Below every set of optical & X-ray spectra and X-ray contour a comment for the objects shown is given. This comment includes the classification number of the optical counterpart and the corresponding redshift. Furthermore, a short fit description is given. For more details on the properties of the sources and the X-ray spectral fits see Tables 1, 4.