SUZAKU OBSERVATIONS OF HARD X-RAY-SELECTED SEYFERT 2 GALAXIES

A number as high as 50% of Seyfert 2 galaxies in the nearby universe are obscured in the X-ray band by column densities on the order of or larger than the inverse of the Thomson cross section (N_H ⩾ σ⁻¹_T ≃ 1.5 × 10²⁴ cm⁻²), hence dubbed Compton thick (CT). If the optical depth (τ = N_H σ_T) for Compton scattering does not exceed values on the order of "a few," X-ray photons with energies higher than 10–15 keV are able to penetrate the obscuring material and reach the observer. For higher values of τ, the entire X-ray spectrum is depressed by Compton down scattering and the X-ray photons are effectively trapped by the obscuring material irrespective of their energy. The former class of sources (mildly CT) can be efficiently detected by X-ray instruments sensitive above 10 keV, while for the latter (heavily CT) their nature may be inferred through indirect arguments, such as the presence of a strong iron Kα line over a flat reflected continuum. The search for and the characterization of the physical properties of CT AGNs are relevant for understanding the evolution of accreting supermassive black holes (SMBHs). In particular, mildly CT AGNs are the most promising candidates to explain the so far largely unresolved spectrum of the X-ray background around its 30 keV peak (Worsley et al. 2005; Treister & Urry 2005; Gilli et al. 2007). According to the Gilli et al. (2007) X-ray background (XRB) synthesis model, their integral contribution to the hard X-ray background is on the order of 25%–30%. This fraction has been estimated under simplified hypotheses. In particular, the same luminosity function and cosmological evolution of unobscured and Compton thin AGNs is assumed; moreover, the number density of "heavily" CT AGNs is the same as "mildly" CT. While these assumptions are not inconsistent with the present observational framework (Risaliti et al. 1999; Guainazzi et al. 2005), it should be noted that absorption column densities in excess of 10²⁴ cm⁻² were measured or inferred for about a few tens of nearby AGNs (Comastri 2004; Della Ceca et al. 2008), and only a handful of them are known beyond the local universe (Norman et al. 2002; Iwasawa et al. 2005). CT AGNs may represent a significant fraction of the accretion power in the universe and indeed a correction for the CT contribution to the estimates of the SMBH local mass density has to be included in the calculations (Marconi et al. 2004).

An unbiased census of extremely obscured AGNs would require one to survey the hard X-ray sky above 10 keV with good sensitivity. Such an argument is one of the key scientific drivers of the NuSTAR (Harrison et al. 2010) and Astro-H (Takahashi et al. 2010) missions, which will be launched in the next few years, and of the nhxm mission study (Pareschi et al. 2010).

For the time being, one has to rely on the observations obtained by the high-energy detectors on board BeppoSAX, Integral, Swift, and, more recently, Suzaku. All-sky surveys were performed using both the IBIS coded-mask telescope on board Integral and the BAT detector on board Swift. Though limited to bright and thus low-redshift sources, they have proven to be quite successful. More than one hundred AGNs are reported in both the Integral (Beckmann et al. 2009) and Swift catalogs (Tueller et al. 2010; Cusumano et al. 2010).

Hard X-ray selection is less biased against absorption and thus provides a useful benchmark for studying the column density distribution of obscured AGNs and the fraction of CT sources in the local universe. The large majority of Integral and Swift sources were observed by XMM-Newton and Chandra and their broadband X-ray spectra were discussed by Winter et al. (2008). While the already known nearby CT AGNs were recovered by Swift and Integral surveys, the fraction of "newly" discovered CT AGNs is surprisingly low and apparently inconsistent, by about a factor of 2, with that predicted by Gilli et al. (2007) in the local universe. On the basis of these findings, it has been proposed (Treister et al. 2009) that the contribution of CT AGNs to the hard X-ray background may be significantly lower (by a factor of 2 to 3) than previously thought, with important implications for the evolution of the accretion power.

Since absorption column densities are often measured on relatively poor-quality X-ray spectra taken from non-simultaneous observations, above and below about 10 keV, and combining different instruments, the CT fraction and, more in general, the absorption distribution in the local universe may still be subject to several uncertainties. Good quality simultaneous spectra extending over the 0.5–100 keV energy range are needed for a robust measurement of absorption column densities, especially in the CT regime. Moreover, nearby obscured AGNs always show excess emission above the extrapolation of the obscured nuclear spectrum (e.g., Turner et al. 1997). Soft X-ray spectroscopy is a powerful tool for studying the origin of this component which is plausibly related to warm gas photoionized by the nuclear continuum (Guainazzi & Bianchi 2007). The X-ray detectors on board the Japanese satellite Suzaku are well suited to this purpose, at least as far as "local" AGNs are concerned. We have conceived a program with Suzaku to observe nearby, relatively X-ray bright (>10⁻¹¹ erg cm⁻² s⁻¹) AGNs selected by requiring (1) a significant detection above 10 keV by either Swift or Integral and (2) evidence for significant X-ray absorption at lower energies from archival observations. The immediate science objective is the characterization of the broadband X-ray spectra of hard X-ray-selected, heavily obscured AGNs and a first step toward a better census of CT absorption in the nearby universe.

Five sources (NGC 4992, NGC 5728, ESO 137–G34, ESO 263–G13, and ESO 323–G37), spectroscopically classified as Seyfert 1.9–2.0 galaxies, were originally selected from the Integral/IBIS (Beckmann et al. 2006) and Swift/BAT (Markwardt et al. 2005) catalogs. The column densities, as inferred from archival (Chandra and XMM-Newton) observations, are on the order of 10²³⁻²⁴ cm⁻², though affected by large errors. By construction, the sample is biased toward obscured AGNs.

The extremely bright flux of NGC 4992 in the Integral observation, coupled with the lack of a detection in the RASS (Voges et al. 1999), is consistent with an absorbed spectrum, as confirmed by a snapshot Chandra observation (Sazonov et al. 2005).

Chandra revealed a heavily obscured nucleus in NGC 5728 (Zheng et al. 2006). Although the column density is not well constrained due to the limited response at high energies, the presence of a strong iron line (EW ∼ 1 keV) suggests that the source might be obscured by CT gas.

ESO 137–G34 was detected by Integral and subsequently observed with XMM-Newton (Malizia et al. 2009). The combined XMM-Newton and Integral spectra can be well fitted by both a transmission and a reflection model as well as by a more complex absorption distribution. In all cases, there is evidence for CT absorption.

ESO 323–G32 and ESO 263–G13 were selected on the basis of a significant (>5σ) Integral/IBIS detection, unambiguous optical classification as Seyfert 2 galaxies, and the lack of archival observations in the 2–10 keV energy range. Both are detected by ROSAT with a soft X-ray spectrum. The extrapolation of the best-fit Integral spectrum in the soft ROSAT band suggests obscuration in excess of 10²³ cm⁻².

The fifth Japanese X-ray satellite Suzaku (Mitsuda et al. 2007) carries four X-ray telescopes with a CCD camera, the X-ray imaging spectrometer (XIS; Koyama et al. 2007) in the focal plane of each of them, and a non-imaging Hard X-ray Detector (HXD; Takahashi et al. 2007), consisting of silicon PIN photodiodes and GSO scintillation counters. They simultaneously cover the 0.2–12 keV and 10–700 keV energy ranges, respectively, although with different sensitivities.

The XIS cameras were operating with the Space-Row Charge Injection in the standard mode. The data were taken by the three XISs (two FI CCD cameras—XIS0 and XIS3—and one BI CCD camera—XIS1) and the HXD PIN. For the two sources— NGC 5728 and NGC 4992—observed at the beginning of the mission, data from the FI CCD XIS2 are also included. All the sources were observed in the "HXD" nominal position on the detector plane which is optimized for the maximum throughput of the HXD rather than the XIS.

The data from version 2 processing⁵ were used for the analysis. The cleaned event files produced by data processing with the standard selection criteria were used. The non-X-ray background of the PIN detector was estimated by the "tuned" background model. The effective exposure times and the background subtracted count rates in the XIS and PIN detectors for each object are reported in Table 1.

Spectral fits are performed with XSPEC version 11.3. Front illuminated CCD were co-added, while the back illuminated chip was discarded after verifying that its spectrum does not significantly improve the quality of spectral fits.

The data from XIS and PIN detectors were fitted together. The intercalibration constant (PIN/XIS) is 1.18.⁶ The broadband Suzaku spectra in a νF_ν representation of the five sources are shown in Figure 1. In Figure 2, we report the XIS spectra.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

For the sake of clarity and to ease comparison with some of the previous Suzaku results on heavily obscured AGNs (Ueda et al. 2007; Eguchi et al. 2009), we consider two baseline spectral models for the broadband Suzaku spectra, namely a transmission dominated (TD) model: wa_Gal(apec + pexrav + zpcafbs(cutoffpl+2×zgauss)) in XSPEC notation and a reflection dominated (RD) model: wa_Gal(apec + pow + pexrav + 2×zgauss). The spectral components of the TD model are: a power law with an exponential cutoff fixed at 200 keV (Gilli et al. 2007) for the primary spectrum plus intrinsic absorption, a reflection component from neutral gas (pexrav; Magdziarz & Zdziarski 1995), and iron K emission (both Kα and Kβ, whenever required by the spectral fits). The photon index of the reflection component is linked to that of the primary continuum and no absorption on the reflected spectrum is considered. The RD model includes the same spectral components except for the absorbed primary power law. The photon index of the RD model is left free to vary. The intensity of the reflection component R in TD model is defined as the ratio between the normalization of the reflected and primary (i.e., power law) spectrum. The inclination angle of the reflection component is fixed to cos i = 0.9, which is equivalent to assuming a rather face-on geometry for the reflecting material, likely to be associated with the inner walls of the obscuring torus.

Soft X-ray emission in excess of the extrapolation of the absorbed nuclear continuum at low energy is observed in all the sources of our sample. This emission may be due to hot gas associated with the host galaxy and is modeled with an updated version of the original Raymond & Smith (1977) code for thermal plasmas (apec in XSPEC), and with AGN primary continuum scattered on the line of sight or leaking through the nuclear absorber, represented by a partial covering (zpcfabs in XSPEC) in the TD model, and with a power law in the RD model. In both models, the power-law slope is assumed to be the same as the intrinsic hard X-ray continuum, with the exception of one object (see Section 4.4). The scattering fraction f can be defined only for the TD model, because the level of the intrinsic continuum in RD fits is unknown. In order to compare the intensity of the soft X-ray power law in TD and RD models, we report the monochromatic flux densities at 1 keV in Table 2.

We have also considered an alternative possibility for the origin of soft X-ray emission in obscured AGNs in terms of emission from a photoionized plasma modeled in XSPEC as wa_Gal(pow + N×zgauss + wa(cutoffpl + 2×zgauss)) for TD sources and as wa_Gal(pow + N×zgauss + pexrav + 2×zgauss)) for RD AGNs (Table 3). N is the number of fitted lines in the soft X-ray band which varies from source to source (Table 4). The additional zgauss component accounts for Kα and Kβ lines. Given that the XIS CCD energy resolution is not appropriate to fit a photoionization code model (Iwasawa et al. 2003), the expected soft X-ray spectrum of a photoionized plasma is modeled with a power law with photon index free to vary (Table 3), plus intrinsically narrow (setting σ = 0 in the zgauss XSPEC model) Gaussian lines at the energy expected (Table 4) for the strongest features observed in high-resolution RGS spectra of obscured AGNs (see Guainazzi & Bianchi 2007). The presence of the line is evaluated on the basis of a visual inspection of the residuals obtained by fitting a single power law. The best-fit energies, fluxes, and associated 1σ errors of the various lines for each source are reported in Table 4.

The best-fit spectral parameters of the continuum along with observed X-ray fluxes and 2–10 keV intrinsic luminosities are summarized in Table 2 for the fits with a partial covering and a thermal spectrum for the soft X-rays, and in Tables 3 and 4 when the soft X-rays are modeled with a power law plus Gaussian lines. The shape of the hard X-ray continuum is rather independent of the model adopted for the soft X-ray spectrum. Therefore, fluxes and luminosities are reported only in Table 2. Galactic absorption using Morrison & McCammon (1983) cross sections (model wa_Gal in XSPEC) is fixed at the values measured by Dickey & Lockman (1990).

The detailed results of spectral fits for each source are summarized in the following.

4.1. NGC 4992

4.2. NGC 5728

4.3. ESO 263–G13

4.4. ESO 137–G34

Both an RD model (χ²/dof≃ 169/139, see Table 3) and a TD model (χ²/dof≃ 160/138) provide an acceptable fit to the broadband spectrum of ESO 137–G34. In the TD fit, the primary hard X-ray continuum (Γ≃ 1.8) is obscured by a column density of ∼1.2 × 10²⁴ cm⁻². The intensity of the scattering fraction (f_scatt≃ 5%) is higher than that measured in the sources best fitted with a TD model, while the reflection component is relatively weak (R∼ 0.3). Even though neither of the two models can be preferred on the basis of purely statistical arguments, the individual parameters of the TD model are highly degenerate and cannot be reliably constrained. As a consequence, an RD model is preferred. It is worth noting that the source would be classified as Compton thick in both cases. The strong Kα line (EW ∼ 1.5 keV) is accompanied by a line at ∼ 7 keV. The best-fit line intensity ratio (∼ 0.2) is slightly higher, but not inconsistent within the errors, than that expected by a Kβ line origin for the ∼ 7 keV feature. The soft X-ray spectrum can be fitted with a thermal plasma (kT∼ 0.8 keV) and a power-law slope which is significantly steeper than that of the hard X-ray continuum. Some residual line like emission, around 1 keV and 1.8 keV, is still present, suggesting a more complex spectrum. A fit with a power law plus several Gaussian lines (Table 4) provides an acceptable solution in terms of χ² statistics and smooth residuals (see Figure 3).

Zoom In Zoom Out Reset image size

4.5. ESO 323–G32

4.6. Summary of Fitting Results

Thanks to the broadband energy range and sensitivity of the X-ray detectors on board Suzaku it has been possible to study, with good accuracy, the spectral properties of a sample of five hard-X-ray-selected type-2 AGNs. The presence of cold obscuring gas has been unambiguously established in all the sources. Although this result was expected, given the sample selection criteria, the X-ray spectra show a high degree of spectral complexity. In particular, the absorption column density and the relative intensity of the various spectral components differ from object to object.

The primary emission is seen, in three out of five sources, through column densities in the range ∼10^23.5−24 cm⁻². The fraction of the intrinsic flux scattered into the line of sight and parameterized by f_scatt in the partial covering fits is relatively low and the reflection component intensity weak. On the basis of Suzaku observations of six Swift-selected AGNs, thus an almost identical selection criterion, Eguchi et al. (2009) found that sources with a low scattering fraction (f_scatt< 0.5%) also have a relatively strong reflection component (R ≳ 1). They dubbed these objects as "new"-type AGNs (see also Ueda et al. 2007) and proposed that they may represent the tip of the iceberg of a hitherto uncovered population of AGNs obscured by a geometrically and optically thick torus (Levenson et al. 2002).

The analysis of the present data does not support the "new"-type AGN interpretation. A possible explanation may be due to the strong degeneracy in the fit parameters. For example, if the reflection component is also absorbed (like in the Eguchi et al. 2009), a solution in terms of a stronger reflection intensity and lower normalization of the scattered component is found, and results similar to Eguchi et al. (2009) are also recovered in our sample (Comastri et al. 2009). However, we believe that the quality of our data does not allow us to constrain absorption on the reflection component; moreover, our best-fit solution has the advantage of having a lower number of free parameters. It is also worth pointing out that the intensity of the reflection spectrum from CT gas strongly depends on the geometry of the system. The reflection fraction, obtained from fitting disk-reflection models to CT AGNs, cannot be directly associated with a solid angle (Murphy & Yaqoob 2009). A better understanding of the nature of "new"-type AGNs and whether there are two distinct classes of Seyfert 2, possibly related to the geometry of the obscuring medium or a continuous distribution of spectral parameters, requires a larger sample of objects with a good counting statistic to allow for a detailed spectral analysis.

The observational evidence of obscured AGNs with a low scattering fraction is at variance with the average values adopted for the above parameters in the Gilli et al. (2007) XRB synthesis models. We have tested the impact of different normalizations for the scattering component in absorbed (N_H > 10²² cm⁻²) AGNs. The fit to the XRB spectrum and source counts are basically unaffected, unless f_scatt exceeds values of the order of 10%. High f_scatt values appear to be the exception, rather than the rule, and are definitely ruled out among highly obscured Swift-selected AGNs.

The possibility that photoionized gas is responsible for the soft X-ray emission in obscured AGNs is consistent with the present observations. It is interesting to note that the residuals in the soft X-ray band of the NGC 5728 Chandra spectrum (Zheng et al. 2006) and ESO 137–G34 XMM-Newton spectrum (Malizia et al. 2009), both fitted with thermal emission, suggest the presence of unaccounted-for emission lines at energies close to those reported here. Unfortunately, the physical status of the ionized gas cannot be constrained by the present observations. Only a limited number of emission lines are individually detected above the 3σ level. It is, however, reassuring that the best-fit energies reported in Table 4 are consistent with those expected by ionized metals observed with the XMM-Newton and Chandra gratings observations of much brighter Seyfert 2 galaxies (Brinkman et al. 2002; Kinkhabwala et al. 2002; Guainazzi & Bianchi 2007). The ∼0.9 keV feature identified with Ne ix appears to be ubiquitous, possibly with the exception of NGC 4992 which is, however, the faintest soft X-ray source in the sample.

It is worth noting that the three brightest soft X-ray sources in the sample, all of them CT (Figure 1), show strong [O iii] ionization cones in optical images (e.g., Wilson et al. 1993). Moreover, extended soft X-ray emission almost spatially coincident with [O iii] cones is detected by Chandra imaging observations of NGC 5728 (Zheng et al. 2006). Furthermore, ESO 137–G34 and ESO 323–G32 are hosted by S0 galaxies which are expected to have neither intense star formation activity nor thermal X-ray emission from hot gas.

Although it may be premature to discard a thermal origin for the soft X-ray emission on the basis of the present X-ray observations, we stress that the above described arguments would support the presence of photoionized plasma, which is also favored by high-resolution rgs observations of bright obscured AGNs (Guainazzi & Bianchi 2007).

Long-look observations of selected samples of obscured AGNs are clearly needed to elucidate the nature of the soft X-ray emission in these objects, though firm conclusions can be drawn only by high-resolution X-ray observations (with either gratings or calorimeters).

We thank Giovanni Miniutti for extremely useful discussions. Support from the Italian Space Agency (ASI) under contracts ASI-INAF I/088/06/0 and ASI-INAF I/23/05 is acknowledged.