X-ray-selected broad absorption line quasi-stellar objects

Abstract

1 INTRODUCTION

Broad absorption line (BAL) quasi-stellar objects (QSOs) are a subset of QSOs which show blue-shifted absorption lines with widths >2000 km s⁻¹ in their rest-frame ultraviolet (UV) spectra. Most are identified by absorption due to C iv, observed to the blue of the C iv 1550 Å broad emission line. BALQSOs represent between 5 and 20 per cent of the QSO population (Trump et al. 2006; Gibson et al. 2009; Scaringi et al. 2009), with the fraction somewhat dependent on the criteria used to define BALQSOs. It was realized during the 1990s that BALQSOs were systematically fainter in the X-ray band than QSOs without BALs (Green et al. 1995; Green & Mathur 1996). Since the launches of XMM–Newton and Chandra, considerable advances have been made in characterizing the X-ray properties of optically selected BALQSOs (e.g. Gallagher et al. 2006; Morabito et al. 2014), with strong X-ray absorption appearing to be an ubiquitous characteristic of optically selected BALQSOs.

As might be expected, given their X-ray faintness, BALQSOs are rarely discovered through X-ray sky surveys. Exceptions include one BALQSO found in the Chandra Deep Field-South (CDF-S; Giacconi et al. 2001), and two discovered during the AXIS survey (Barcons et al. 2002). The largest sample of X-ray-selected BALQSOs comes from the 1^H Deep Field, which yielded four such objects (Blustin et al. 2008). The properties of the four 1^H BALQSOs and the CDF-S BALQSO were studied in detail by Blustin et al. (2008). They found that in terms of their rest-frame 2500 Å to 2 keV spectral slopes (α_ox), the X-ray-selected BALQSOs were a little less X-ray faint than optically selected BALQSO samples, but still faint in the X-ray compared to X-ray-selected non-BAL QSO samples. From the X-ray spectral shapes, Blustin et al. (2008) inferred that the X-ray-selected BALQSOs had similar levels of X-ray absorption to those seen in optically selected samples of BALQSOs.

Somewhat contradictory results were presented by Giustini, Cappi & Vignati (2008), who obtained a sample of 54 X-ray-detected BALQSOs by cross-correlating the Sloan Digital Sky Survey Data Release 5 QSO catalogue (Schneider et al. 2007) with the Second XMM–Newton Serendipitous X-ray Source Catalogue (2XMM; Watson et al. 2009). Giustini et al. (2008) found that one-third of the BALQSOs in their sample showed little or no X-ray absorption in their X-ray spectra, and that the distribution of α_OX of their sample was indistinguishable from that of non-BAL QSOs, implying that the BALQSOs in their sample are not X-ray weak.

Subsequently, Streblyanska et al. (2010) compiled a sample of 88 X-ray-detected BALQSOs by cross-correlating the 2XMM catalogue with BALQSOs in the NASA Extragalactic Database. They found that when the X-ray absorption is modelled as ionized gas, as opposed to the neutral absorber model considered by Giustini et al. (2008), 90 per cent of the X-ray-detected BALQSOs showed X-ray absorption. Streblyanska et al. (2010) did not examine the α_OX distribution of their BALQSO sample.

It is notable that the X-ray properties found for samples of BALQSOs may depend on the criterion which was used to identify the BALQSOs from their rest-frame UV spectra. A quantitative criterion for classifying an object as a BALQSO was introduced in the form of the balnicity index (BI) by Weymann et al. (1991). For a full definition of BI, see appendix A of Weymann et al. (1991), but, in simple terms, it is approximately the equivalent width in km s⁻¹ of continuous troughs of absorption, with outflow velocities between 3000 and 30 000 km s⁻¹ to the blue of C iv λ1550, and from which the first 2000 km s⁻¹ velocity interval of each absorption trough is excluded from the calculation. Thus, BI >0 implies at least one absorption trough with a width of >2000 km s⁻¹. According to Weymann et al. (1991), BI > 0 is required to classify an object as a BALQSO. An alternative criterion, known as the absorption index (AI), was introduced by Trump et al. (2006), based on an earlier proposal by Hall et al. (2002). In essence, the AI is also an equivalent width measurement, but includes the contribution of absorption troughs with outflow velocities below 3000 km s⁻¹, and requires a minimum absorption trough width of only 1000 km s⁻¹ to yield a positive AI. Thus, BI >0 is a more conservative criterion for the classification of BALQSOs than AI > 0. Prior to 2006, BALQSOs studied in the X-ray were usually selected by BI, as was the sample studied by Blustin et al. (2008). However, the BALQSO sample of Giustini et al. (2008) was based on AI. Streblyanska et al. (2010) selected the sample on AI, but also presented BI values for their sample. They found that while the sources in their sample with BI > 0 were almost always absorbed, the sources in their sample with BI = 0 typically did not have significant X-ray absorption.

In this paper, we present a new sample of X-ray-selected BALQSOs, which is larger and brighter in X-rays than the sample presented by Blustin et al. (2008). We use the new sample to investigate the X-ray properties of such objects, in particular their optical to X-ray spectral indices and the incidence of X-ray absorbing gas. Throughout, we use the phrase ‘X-ray-selected BALQSOs’ specifically for objects which were discovered, first, as X-ray sources, and then found to be BALQSOs in the subsequent optical spectroscopic follow-up. We define power-law spectral indices α such that f_ν ∝ ν^−α. We have assumed cosmological parameters H₀ = 70 km s⁻¹ Mpc⁻¹, Ω_Λ = 0.7, and Ω_m = 0.3. Unless stated otherwise, all uncertainties are given at 1σ.

2 SAMPLE AND UV ABSORPTION PROPERTIES

The sample is drawn from the XMM–Newton Wide Angle Survey (XWAS; Esquej et al. 2013), an optical identification programme for X-ray sources, carried out using the 2dF spectrograph (Lewis et al. 2002) on the Anglo-Australian Telescope. The first stage of sample selection involved a visual inspection of XWAS QSOs to identify those which have deep absorption lines bluewards of C iv λ1550. Next, absorption features and parts of the spectra shortwards of N v λ1250 were masked, before cross-correlating with the QSO template spectrum of Vanden Berk et al. (2001), to obtain redshifts of the objects. Note that for these strong-absorption objects, we expect the redshifts obtained by this process to be more accurate than those derived in Esquej et al. (2013). The sample was then refined by measuring the BI of each candidate object. Following Weymann et al. (1991), we consider any object with BI > 0 to be a BALQSO, and only such objects were included in the final sample which consists of six objects. Their properties are summarized in Table 1, and their optical spectra, as measured in XWAS, are shown in Fig. 1.

Figure 1.

Rest-frame UV spectra of the X-ray-selected BALQSOs. The dashed line marks the rest-frame wavelength of C iv, 1550 Å.

Open in new tabDownload slide

3 X-RAY DATA AND CONSTRUCTION OF X-RAY SPECTRA

X-ray spectra were constructed for the BALQSOs by using all the XMM–Newton European Photon Imaging Camera (EPIC; Strüder et al. 2001; Turner et al. 2001) observations available in the XMM–Newton Science Archive. The observations are listed in Table 2. The observations were reduced using standard tasks in the XMM–Newton Standard Analysis Software (sas) version 14.0,¹ with periods of high background filtered out. Spectra were accumulated from each observation using elliptical source regions with the major axis sized according to the off-axis angle and oriented perpendicular to the off-axis vector to match the point spread function. Background was taken from an annular region around the source, from which circular regions around other sources were excised. Response and effective area files were then constructed for each source and for each EPIC camera, in each observation, using the standard sas tasks rmfgen and arfgen. Then, the source spectra, background spectra, responses and effective areas for all EPIC cameras and all observations were combined to produce a single source spectrum, background spectrum, response and effective area file for each BALQSO, following the procedure described in Page, Davis & Salvi (2003). Spectra were grouped to a minimum of 20 counts per bin using grppha.²

4 RESULTS

4.1 X-ray spectral analysis

The X-ray spectra of the six BALQSOs are shown in Fig. 2. Analyses of the X-ray spectra were carried out in the spectral fitting package spex version 2.02.04.³ As a first step, the spectra were fitted using a power-law model, with energy index α free to vary, and photoelectric absorption from a fixed column density of cold Galactic gas, determined from the Leiden/Argentine/Bonn H i survey (Kalberla et al. 2005). The results are given in Table 3. The best-fitting slopes α range from 0.49 to 0.50, significantly below the mean 〈α〉 = 0.98 for X-ray-selected QSOs (Mateos et al. 2005), indicating that the BALQSOs have harder spectral shapes than are usual for QSOs. For five of the six BALQSOs, the simple power-law fit provides an acceptable goodness of fit (null hypothesis probability >1 per cent), but the fit is poor for XWAS J005007.

Figure 2.

XMM–Newton EPIC spectra of the six BALQSOs. The spectra have been flux calibrated by dividing by the product of the effective area and Galactic transmission, and are shown as EF_E. A power law with α = 1 would be a horizontal line. The dark stepped lines show the power-law model (model PL in Table 3), while the light-grey stepped lines show the model, including an ionized absorber (model PL × IA in Table 3).

Open in new tabDownload slide

A deficiency of soft X-rays is commonly a sign of absorption, so we have considered an alternative model consisting of a power law with ionized absorption as well as absorption from cold Galactic material. As before, the Galactic N_H was fixed. This time, we fixed the power-law slope to α = 0.98. For the ionized absorption, we used the xabs model in spex, which predicts both bound–bound and bound–free absorption from an equilibrium photoionized plasma for which the ionization state is characterized by the ionization parameter ξ. The xabs model assumes an ionizing continuum shape based on the type 1 active galactic nucleus (AGN) NGC 5548 (Steenbrugge et al. 2005) and a Gaussian velocity dispersion for the absorber. With respect to the rest frame of the QSO, we assume that the absorber seen in the X-ray band has a similar outflow velocity and velocity dispersion as the UV absorber. Therefore, appropriate parameters were determined from Gaussian fits to the main C iv absorbing trough in each rest-frame UV spectrum, and are given in Table 1.

The power-law and ionized absorber model provides acceptable fits for all six BALQSO spectra. Confidence contours for log ξ and log N_H of the absorbers, as derived from the X-ray fits, are shown in Fig. 3. The ionization parameters of the absorbers are poorly constrained from the fits, and somewhat correlated with the column densities. Nonetheless, robust lower limits on the column densities of between 10²² and 10²³ cm⁻² can be inferred from Fig. 3 for all of the objects except XWAS J022645, which has the poorest quality X-ray spectrum.

Figure 3.

Confidence contours for ξ and N_H in the ionized absorber model fits to the XMM–Newton spectra. The small cross indicates the best-fitting parameter values. Solid, dashed and dot–dashed contours correspond to the 68, 95 and 99.73 per cent confidence for two interesting parameters.

Open in new tabDownload slide

4.2 X-ray-to-optical ratio

Following the usual convention in AGN studies, we have parametrized the X-ray-to-optical flux ratio of our BALQSOs as α_OX, the power-law slope which would connect the flux densities at 2500 Å and 2 keV in the rest frame of the source. The 2 keV flux measurements were derived from the power-law model fits to the X-ray spectra without ionized absorbers (i.e. model PL in Table 3), similar to the approach taken by Gallagher et al. (2006) and Blustin et al. (2008). The rest-frame 2500-Å flux was derived from R-band magnitudes measured by the SuperCOSMOS Sky Survey (Hambly et al. 2001a) scans of the UK Schmidt Telescope R plates, converted to monochromatic flux densities assuming the template spectrum of Vanden Berk et al. (2001). Uncertainties on the R magnitudes are taken from table 6 of Hambly, Irwin & MacGillivray (2001b). Uncertainties on α_OX were derived by propagating the statistical uncertainties on the X-ray model fits and R magnitudes. The values of α_OX are given in Table 4, together with broad-band fluxes and magnitudes.

5 DISCUSSION

The X-ray spectra of the BALQSOs show significant evidence for absorption. Although the majority of the X-ray spectra can be successfully fitted with power-law models without intrinsic absorption, the derived spectral indices (−0.5 < α_X < 0.5) are significantly lower than the norm for non-BAL QSOs, which cluster around a mean of 〈α_X〉 = 0.98 with a dispersion of σ_α = 0.27 (Mateos et al. 2005, 2010; Scott et al. 2011), indicating that the BALQSOs have significantly harder shapes than non-BAL QSOs. All of the spectra can be fitted successfully with a model in which a typical non-BAL QSO spectrum passes through an ionized absorber. The six sources presented here therefore have X-ray spectral properties which are consistent with those of the five X-ray-selected BALQSOs studied by Blustin et al. (2008).

On the other hand, the observed UV to X-ray spectral slopes α_OX of our six X-ray-selected BALQSOs are quite consistent with those expected for non-BAL QSOs. There is a well-known relation between UV luminosity and α_OX in non-BAL QSOs (Strateva et al. 2005; Steffen et al. 2006; Just et al. 2007). The relation holds for both optically selected and X-ray-selected non-BAL QSO samples (Lusso et al. 2010). Using the expression from Just et al. (2007), α_OX = 0.14(log L_ν) + 2.705, where L_ν is the monochromatic luminosity at 2500 Å, we have calculated the predicted α_OX values (see Table 4). Taking into account the observed scatter around this relation of σ_α = 0.14 (Just et al. 2007), we would predict a mean of 〈α_OX〉 = 1.64 ± 0.06 for our sample of objects if they were non-BAL QSOs, which is quite consistent with the observed 〈α_OX〉 = 1.69 ± 0.05. In contrast, optically selected samples of BALQSOs usually show much larger values of α_OX than our X-ray-selected sample. For example, the majority of the sample of 36 BALQSOs from the Large Bright Quasar Survey studied by Gallagher et al. (2006) have α_OX > 2, and their 〈α_OX〉 is larger by 0.52 than the mean which would be expected, given their UV luminosities.

At first sight, finding that our BALQSOs have α_OX values in the normal range for non-BAL QSOs seems at odds with the presence of significant X-ray absorption in their X-ray spectra because the X-ray absorption would be expected to lead to increased α_OX compared to a typical QSO without significant X-ray absorption. An explanation can be gleaned by examining how the absorption should affect α_OX. For each of our BALQSOs, we have recalculated the α_OX using the intrinsic (i.e. unabsorbed) rest-frame 2 keV model continuum from the ionized absorber model fits in Table 3. These absorption-corrected values are given in Table 4 in the column headed ‘α_OX corrected’. In all cases, the absorption-corrected α_OX values are lower than those predicted by the Just et al. (2007) relation by between one and three times the scatter (0.14) found by Just et al. (2007). Therefore, intrinsically, these objects occupy the lowest 15 per cent of the α_OX distribution.

In constructing an X-ray-selected BALQSO sample, we have naturally selected BALQSOs which are at the X-ray-bright end of the distribution. As such, we would expect to find the part of the population with the lowest levels of X-ray absorption, and/or the smallest intrinsic values of α_OX. We do not find any examples of BALQSOs without strong X-ray absorption, but all of our sample have intrinsic α_OX values which are low compared to the majority of QSOs at similar UV luminosities.

Consistent with our findings, Blustin et al. (2008) also found evidence for strong X-ray absorption in the X-ray spectra of all of their X-ray-selected BALQSOs. The values of α_OX found by Blustin et al. (2008) are a little higher than for the six objects in our sample, though still much lower than typically found for optically selected BALQSOs. Observationally, the sample of Blustin et al. (2008) was selected from deep X-ray surveys, whereas ours is drawn serendipitously from moderate exposure (<50 ks) XMM–Newton observations. The mean 0.5–2 keV flux of the Blustin et al. (2008) sample is five times lower than the mean 0.5–2 keV flux of the six sources studied here, and the mean 0.5–2 keV luminosity of the Blustin et al. (2008) sample is seven times lower than the mean 0.5–2 keV luminosity of the six sources studied here. By selecting BALQSOs at a higher X-ray flux limit than Blustin et al. (2008), we have been able to push further into the X-ray-bright end of the BALQSO population, and so are able to draw stronger conclusions regarding their X-ray properties. Our findings suggest that X-ray absorption is a universal property of BALQSOs but imply that X-ray weakness (i.e. large α_OX) is not.

In the study of Streblyanska et al. (2010), there were two notable correlations concerning the X-ray absorbers, which Streblyanska et al. (2010) modelled as ionized absorbers. First, they found a correlation between the ionization parameter ξ and N_H, and secondly, they found a correlation between N_H and the BI of the UV absorption. Our sample is too small to permit such a correlation analysis, but our spectral fitting may provide some insights into these correlations. It is clear from Fig. 3 that with the current statistics in our X-ray spectra, the ionization parameters of the absorbers are not constrained, and there is a significant covariance between  log ξ and log  N_H in the individual spectral fits. The six BALQSOs in our sample show a consistent trend in correlation between log  ξ and log  N_H to that found by Streblyanska et al. (2010), but it can be accounted for entirely by this covariance. The ionized absorption model we have used, xabs in spex, is more sophisticated than that used by Streblyanska et al. (2010), absori in xspec (Arnaud 1996), because xabs includes bound–bound transitions as well as bound–free transitions, while absori includes only bound–free transitions. However, we would expect a similar level of covariance between log  ξ and log  N_H in the two models.

Most of the BALQSOs in the sample of Streblyanska et al. (2010) for which spectral fitting has been performed are brighter in X-rays than those in our sample, so the degeneracy between ξ and N_H may be mitigated to some extent by the improved photon statistics compared to our sample, but it is nonetheless a concern that some or all of the correlation found by Streblyanska et al. (2010) between ξ and N_H may be a consequence of the covariance of these two parameters.

With regard to the correlation found by Streblyanska et al. (2010) between BI and N_H, we can neither confirm nor refute this correlation with our small sample. Some insight into the relation between UV C iv absorption and X-ray absorption in QSOs may be drawn by comparing our sample of BALQSOs with the X-ray-absorbed QSOs studied by Page et al. (2011). Comparison of Fig. 3 with the corresponding fig. 4 in Page et al. (2011) (and disregarding the object RX J124913 from their sample, which is itself a BALQSO), we find that the X-ray-absorbed QSOs have confidence contours in (log  ξ, log  N_H) which fall systematically below those of our BALQSO sample. Despite the degeneracy between ξ and N_H in both samples, the X-ray-selected BALQSOs have systematically more X-ray absorption than the X-ray-absorbed QSOs, which were also X-ray selected. The X-ray-absorbed QSOs of Page et al. (2011) also have lower C iv equivalent widths than the BALQSOs. Thus, there appears to be a continuous spectrum of QSO absorption properties, from the majority of QSOs which have C iv equivalent widths <5 Å and minimal X-ray absorption, through the X-ray-absorbed QSOs which typically have C iv equivalent widths in the range of 5–20 Å and moderate X-ray absorption, to the BALQSOs which have C iv equivalent widths >20 Å and the strongest X-ray absorption. A similar correlation between C iv equivalent width and α_OX has long been established (Brandt, Laor & Wills 2000). As X-ray absorption appears to be a universal property of BALQSOs but large α_OX is not, the connection between C iv absorption and X-ray absorption may be considered more fundamental than that between C iv and α_OX.

Significant X-ray absorption, with log  N_H > 10^22.5 cm⁻², appears to be a ubiquitous property amongst X-ray-selected BALQSOs, defined according to the classical BI > 0 criterion, as presented here and by Blustin et al. (2008). The same conclusion was reached by Streblyanska et al. (2010) for BALQSOs identified by correlating optical catalogues with the 2XMM catalogue. We are thus led to the same conclusion as Streblyanska et al. (2010) that the rather different absorption properties found by Giustini et al. (2008) for X-ray-detected BALQSOs probably relate to their use of the AI definition of BALQSOs. The AI criterion used by Giustini et al. (2008) includes objects with lower C iv equivalent widths than the classical BI criterion, and this may account for the comparatively weak X-ray absorption found by those authors for some of their sample. The apparent difference in X-rays between BALQSOs defined by BI and those defined by AI (and not BI) is interesting in the context of the bimodality of the AI distribution demonstrated by Knigge et al. (2008) and their interpretation that the bimodality corresponds to two distinct subpopulations of QSOs. It appears that the two subpopulations selected by AI have distinct distributions in X-ray absorption, as well as in their UV absorption-line properties.

6 CONCLUSIONS

We have presented a sample of six X-ray-selected BALQSOs from XWAS. Examination of their X-ray spectra indicates that all have hard spectral shapes compared to the majority of QSOs, suggesting absorption. An ionized absorption model applied to a typical AGN template X-ray spectrum is able to reproduce the X-ray spectral shapes, and implies that these are intrinsically rather X-ray bright for QSOs, given their UV luminosities. X-ray absorption appears to be ubiquitous in classically defined (i.e. BI > 0) BALQSOs, and this is true for X-ray-selected BALQSOs as well as for optically selected samples. In contrast, X-ray weakness (i.e. high α_OX) is not a universal property of BALQSOs. We have argued that there is a continuous spectrum of X-ray-absorption properties in QSOs which are related to the equivalent width of UV C iv absorption, with BALQSOs forming the most heavily absorbed end of the distribution.

Acknowledgments

This research was based on observations obtained with XMM–Newton, an ESA science mission with instruments and contributions directly funded by ESA Member States and the National Aeronautics and Space Administration. This research was also based on observations made at the Anglo-Australian Telescope. MJP acknowledges financial support from the UK Science and Technology Facilities Council. FJC and SM acknowledge financial support through grant AYA2015-64346-C2-1-P (MINECO/FEDER). MTC acknowledges support by the Spanish Programma Nacional de Astronomia y Astrofisica under grant AYA2009-08059. MK acknowledges support by DFG grant KR 3338/3-1. The NASA/IPAC Extragalactic Database is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration.

REFERENCES