5.1. Probing X-ray weak AGNs
As discussed in Sect. 1most of the former hard X-ray spectroscopic surveys of Sy2s were seriously biased for X-ray bright sources. This is illustrated in Fig. 4. The observed 2-10 keV luminosity in the Sy2 (and Sy1.9) galaxies surveyed by GINGA, most of which are reported in Smith & Done (1996), has an average . Turner et al. (1997a, 1997b) studied lower luminosity Sy2s by using ASCA data: their sample has a mean , but it includes some objects with luminosity lower than erg/sec.
Our sample was not selected according to the X-ray flux. Also, we avoided objects for which hard X-ray spectroscopic data were already available. As a consequence, our survey samples Sy2s with X-ray fluxes significantly lower than in former studies, although at similar distances. Indeed, in our sample .
These X-ray weak AGNs have spectral properties similar to the bright ones: a relatively flat continuum and the presence of the Fe K emission line blend, generally dominated by the neutral component at 6.4 keV. The main difference between the X-ray properties of the sources in our sample and bright AGNs is that the former have observed spectra that are generally flatter (see Appendix) and are characterized by Fe lines with larger EW; as discussed in Sect. 4.4, these differences are very likely due to heavy obscuration along our line of sight for the sources in our sample.
Although these sources are characterized by apparently low luminosities, most of them are Compton thick, i.e. their intrinsic luminosity must be much higher than observed (by perhaps two orders of magnitude). Yet, NGC 4941 is Compton thin and its intrinsic luminosity (obtained by correcting for the estimated NH) is only erg/sec, similar to some LINER nuclei.
It should be noted that about half of the Seyferts in the Maiolino and Rieke's (1995) sample have [OIII] luminosities lower than NGC 4941, and have not been studied in the X rays. Therefore, the hard X-ray spectral properties of the lowest luminosity AGN population have still to be probed.
5.2. The distribution of absorbing column densities
The most remarkable result of our survey is that all the objects are heavily obscured with NH 1023.6cm-2 and, in particular, 6 out of 8 objects are Compton thick. More specifically NGC2273, NGC3393, NGC4939, NGC5643 and MCG-05-18-002 were identified as Compton thick with NH 1025cm-2, NGC1386 is Compton thick with NH 1024cm-2, while NGC3081 and NGC 4941 were identified as Compton thin with NH cm-2. However, the nature of NGC4939, NGC4941 and MGC-05-18-002 is still questionable.
This result has to be compared with former spectral surveys. Smith & Done (1996) studied the spectra of a sample of type 2 and 1.9 Seyferts observed with Ginga, that were probably selected amongst bright X-ray sources. The distribution of NH in their sample is shown in Fig. 5. The average absorbing column density is about cm-2. Turner et al. (1997a, 1997b) analyzed the spectra of a sample of Sy2s from the ASCA archive. The latter sample includes a larger fraction of weak Sy2s with respect to the Ginga survey (see discussion in the former section). As a consequence, it contains a larger fraction of heavily absorbed Sy2s, as shown in Fig. 5. 1
Our sample was selected by means of an isotropic indicator of the nuclear luminosity, the dereddened [OIII] flux, and therefore is not biased against obscuration on the pc scale. Thus, it contains a larger portion of heavily obscured Sy2s. In particular, our survey doubles the number of known Compton thick Seyferts: before this study only 6 sources were surely known to be Compton thick, namely NGC 1068, Circinus, NGC4945, NGC 6240, NGC 6552, NGC7674 (the latter identified by means of BeppoSAX as well, Malaguti et al. 1998).
It is interesting to compare the properties of our sources with respect to former surveys also in terms of their location on the NH vs. diagram, as shown in Fig. 6 2. As expected (see discussion in Sect. 4.3), NH correlates with the ratio. The shaded region indicates the expected correlation by assuming that LX (2-10 keV) is absorbed by the NH reported on the Y-axis, starting from the average ratio observed in Sy1s (the width of the shaded region reflects the dispersion of the distribution in Sy1s) and by assuming a 1% reflected component. The observed distribution is in agreement with what expected from the simple model outlined above, especially if allowance is made for an efficiency of the reflected component different from what assumed. Two objects, namely NGC4968 and NGC5135, are clearly out of the correlation; however, as discussed in detail in Bassani et al. (in prep.), these two sources are very likely Compton thick with NH 1024cm-2. All of the Sy2s in our sample (filled circles) lie in the high-NH and low- region of the distribution, while most of the Sy2s in former ASCA (open squares, Turner et al. 1997a) and GINGA (open triangles, Smith & Done 1996) surveys are in the low-NH and high- region.
One of the consequences of our result is that the average absorbing column density in Sy2s turns out to be higher than what was estimated in the past. However, the NH distribution in our sample alone does not necessary reflect the real distribution. Indeed, our sample is not biased against heavily absorbed nuclei, but is probably biased against little absorbed Sy2s, since we avoided Sy2s already observed in previous surveys (i.e. generally X-ray bright sources). The issue of the real distribution of NH is tackled in Bassani et al. (in prep.), where the spectra of several Sy2s (including the ones in this paper) are collected and analyzed.
5.3. Warm scattering versus cold reflection
In Compton thick Seyferts the nuclear continuum can be observed only when scattered by a warm, highly ionized mirror, or Compton reflected by a cold neutral medium (possibly the molecular torus). Since our survey has significantly enlarged the number of known Compton thick Sy2s, it is now possible to statistically assess the relative importance of the cold versus warm scattering in this class of objects.
By merging our sample with Compton thick sources previously reported in the literature, we found that most of the Compton thick Seyferts are cold reflection dominated. More specifically, out of 12 Compton thick Seyferts 7 are cold reflection dominated (namely NGC 1386, NGC 2273, NGC 3393, NGC 4939, Circinus, NGC 6552, NGC 7674), in 3 sources warm scattered and Compton reflected components contribute to a similar extent (namely NGC 1068, NGC 6240, and NGC4945), and only one source (NGC 5643, though to be confirmed) appears to be warm scattering dominated; the nature of the scatterer in MGC-05-18-002 is still uncertain.
The apparent overabundance of cold reflection dominated Compton thick sources must probably be ascribed to the low efficiency of the free electron scattering process. Indeed, in the latter case the fraction of scattered light is given by
where is the free electron column density and is the solid angle subtended by the warm mirror. The average opening angle of the light cones in Sy2s is (Maiolino & Rieke 1995). If we assume that the warm mirror in Sy2s is the analogous of the warm absorber observed in Sy1s, then cm-2 (Reynolds 1997). According to Eq. 1this implies that the scattering efficiency of the warm mirror is only , i.e. significantly lower than the typical efficiencies observed in Compton thick sources. The Compton cold reflection efficiency can be as high as 16% in the 2-10 keV range, though geometry dependent absorption effects (e.g. torus self shielding) can lower this value to a few percent.
Those (few) Compton thick objects whose spectrum shows evidence for a warm scattered component could either be characterized by an anomalously high column density of the warm mirror, or the Compton reflected component could be significantly absorbed along our line of sight; the latter case might occur if the putative torus is edge-on.
5.4. The soft excess
Most of the objects in our sample show an emission below keV in excess of the extrapolation of the high energy spectrum. Such soft X-ray excesses are commonly observed in type 2 Seyferts and have been studied by several authors. This component is usually not observed in type 1 Seyferts, presumably because the low absorption affecting them makes their flat power law component overwhelming. The nature of the soft excess in type 2 Seyferts is not clear yet. There are two possible origins: 1) thermal emission from hot gas in the Narrow Line Region, or 2) emission associated to starburst activity in the host galaxy (X-ray binaries, supernova remnants, starburst driven superwinds). The possibility that the observed soft excess is due to the primary AGN radiation scattered by a warm medium is not favored based on energy budget arguments (Wilson & Elvis 1997).
Our data are not well suited to study the emission below 3 keV, since at low energies other satellites (ROSAT, ASCA) are much more sensitive. However, the wide spectral coverage of our BeppoSAX data enables us to determine what fraction of the soft X-ray emission is contributed by the flat AGN component extrapolated to low energies, and what fraction, instead, is actually due to an extra component. In other words, by means of our fits in Sect. 4.4we can derive the flux of the black body component alone.
In these low luminosity Seyfert nuclei most of the Far-IR luminosity can be ascribed to star formation in the host galaxy (star forming activity is much more effective than the AGN in powering the FIR emission). So, the FIR luminosity can be used to estimate the level of star formation, and to determine the contribution of the latter to the observed soft excess. David et al. (1992) derived an empirical relationship between and LFIR in starburst and normal galaxies. This relation is shown in Fig. 7 by means of a thick solid line, while the shaded region indicates the 90% confidence area. The same figure shows the location of the objects in our sample (NGC3081 and MGC-05-18-002 do not have available IRAS data, while NGC5643 does not show evidence of a soft X-ray excess). NGC 2273 seems to follow the relation for SB and normal galaxies, thus in this object most of the excess can be completely ascribed to the starburst. Most of the other objects show indications of 0.5-4.5 keV emission in excess of what would be expected from the starburst activity alone, although NGC3393, NGC 1386 and NGC4941 are also consistent with the upper limit of the SB/normal galaxies' distribution, when allowance is made for uncertainties. NGC4939 (which also has the highest signal-to-noise spectrum) is the object for which is strongest the evidence for an extra contribution to the observed soft X-ray excess, i.e. emission related to the AGN.
We should mention that, although weak, these Seyfert nuclei might contribute some of the FIR luminosity. In the latter case the observed FIR luminosity provides an upper limit to the contribution from the starburst. Maiolino et al. (1995) addressed the issue of the relative importance of the IR emission from AGNs and from star forming activity in their host galaxies. Two of the objects in our sample, namely NGC 2273 and NGC 1386, were also studied in Maiolino et al. (1995). By using their results we could provide lower limits for the starburst contribution to the FIR luminosity in these two objects; they are indicated by points connected by dotted lines in Fig. 7. In the case of NGC 2273 the lower limit indicates that the soft X-ray excess is fully consistent with it being originated by the starburst activity. In the case of NGC 1386 it provides further support to the idea that a fraction of the soft excess comes from the AGN (NLR) component.
© European Southern Observatory (ESO) 1998
Online publication: September 17, 1998