4 Properties of the X-ray-mid-IR matched sources

Optical colours, redshifts and spectral classifications are available for many of the galaxies emitting in X-ray and mid-IR bands. In the case of the Lockman Hole, the best known galaxies are those already detected with ROSAT (see Lehmann et al. 2000, 2001), which constitute approximately half of our sample. On the contrary, redshifts are known for all but two of the galaxies of the HDF-N sample due to the great efforts made in this area (e.g. Hogg et al. 2000; Cohen et al. 2000), although only few galaxies are classified as AGNs or starburst galaxies according to their spectral features. In this Section we analyse the sources both detected in the mid-IR and X-ray from the point of view of their optical, X-ray and infrared emissions. By comparing their properties with those of local template galaxies we are able to classify these sources as starburst - or AGN-dominated.

Figure 5: Optical colour diagrams for X-ray-emitting galaxies in the part of HDF-N (left) and Lockman Hole fields (right) surveyed in X-ray and mid-IR. Triangles, diamonds and circle refer to type-1 AGNs, type-2 AGNs and unknown type galaxies, respectively. X-ray sources detected in the mid-IR are marked with full symbols. Dashed and solid lines show colours of spiral and elliptical galaxies, respectively, as a function of the redshift. The curves are computed using the PEGASE2 code (Fioc & Rocca-Volmerange 1997). Sources with $L_{\rm X} < 10^{41}$ erg s-1 in the HDF-N are indicated with smaller symbols and lie around the lines of normal galaxies.

4.1 Optical properties

The HDF-N survey, which is the deepest we consider here, allows one also to detect very faint sources and thus starburst and nearby galaxies (see Hornschemeier et al. 2001; Elbaz et al. 2002). Only a small part of the sources detected both in the mid-IR and X-rays are optically classified as AGN (20%). Also the redshift distribution of these galaxies reflects this situation. The median redshift of 0.5 is typical of the mid-IR galaxy population (see Fadda et al. 2002; Flores et al. 2002), while the median redshift of the galaxies classified as AGN is 1. On the contrary, the sources detected in the Elais-S1 survey are almost exclusively AGN at high redshift. Excluding a normal galaxy detected at z=0.3, all the other sources lie at z>0.4 with a median value of z=2.

Finally, among the sources detected in the Lockman Hole almost half of the sample is classified as AGN while the rest is up to now of unknown type. The median redshift of the sources is z=1 and all the sources lie at z>0.4. Therefore, this population of galaxies differs from the bulk of the galaxies detected in the Lockman Hole, which lie at a redshift of 0.6 (see Fadda et al. 2002). We can learn something more about the spectrally unclassified galaxies by looking at the optical colour diagrams (see Fig. 5). As expected, the type-1 AGNs cluster in a region of blue colors while the type-2 AGNs are in general redder and less clustered on the diagram. Many of the galaxies with unknown type lie in the region occupied by type-2 AGNs, suggesting that they are highly extincted objects and probably most of them are type-2 AGNs.

To aid in the interpretation of the diagrams, we overlay two galaxy tracks corresponding to elliptical and spiral templates. These models were produced with the PEGASE2.0 code (Fioc & Rocca-Volmerange 1997) assuming a Salpeter initial mass function with standard cutoff (0.1-120 $M_{\odot}$). For the elliptical track we adopt a star formation timescale of 1 Gyr, observed at 6 Gyr, without extinction and nebular emission. For the spiral track we consider a star formation timescale of 5 Gyr, observed at 3 Gyr, extinction with disk geometry and no nebular emission. Tracks are labelled with representative redshifts over the range 0<z<3, which corresponds to the redshift range of the galaxies observed. Few galaxies appeared clustered around these lines. In particular, five galaxies detected in the HDF-N with low X-ray luminosities have colours typical of normal galaxies. Most of the galaxies are scattered over the diagram, but there are almost no galaxies which follow the track of the elliptical galaxies with z>1.

Figure 6: X-ray diagnostic diagrams based on hardness ratios (see Hasinger et al. 2001). Triangles, diamonds and circle refer to type-1 AGNs, type-2 AGNs and unknown-type galaxies, respectively. Galaxies inside the common X-ray and mid-IR area are shown. The cross indicates the median error bar of the points. Only points with error less than 0.1 are plotted. The galaxies with mid-IR emission are marked with full symbols. For these galaxies HR1, HR2 and HR3 values are reported in Cols. 16-18 of Table 3. The grid gives the expected hardness ratios for power-law models with different values of the photon index $\Gamma$ and of the neutral hydrogen absorption $\log N_{\rm H}$ (in the observed frame).

Finally, we note that our sample of XMM-ISO matched sources contains five highly obscured sources that are extremely red objects (i.e. EROs, according to the definition $R-K \ge 5$). Objects of this type are claimed to constitute about 30% of the optically faint X-ray sources in the deep Chandra survey of the HDF-N (Alexander et al. 2001a). On the other hand, Pierre et al. (2001) showed that is possible to select this kind of objects using mid-IR observations. This sample of objects will be studied in more detail by Franceschini et al. (2001). In the Lockman field, another four EROs were detected by XMM-Newton and not by ISOCAM. As discussed in Franceschini et al. (2001), the expected 15 $\mu$m fluxes of these objects fall below the detection limit of the survey (0.3 mJy).

Figure 7: 2-10 keV rest-frame luminosity versus redshift (left) and R - K colour (right) for the X-ray mid-IR matched sources. Open symbols, grey symbols and crosses refer to the HDF-N, Lockman and Elais surveys, respectively. Type-1 and type-2 AGNs are marked with triangles and squares, respectively, while circles identify unclassified sources. In the left figure, the three dashed lines show the sensitivity limits of the three X-ray surveys. The horizontal lines trace the X-ray luminosity of the template galaxies discussed in the text. The luminosity distance is computed according to Carroll et al. (1992) assuming a cosmology of H0= 70 km s-1 Mpc-1, $\Omega _\lambda =0.7$ and $\Omega _{\rm M}=0.3$.

4.2 X-ray diagnostic

Thanks to the large energy range which can be explored with XMM-Newton it is possible to construct colour-colour X-ray diagrams and to classify sources on the basis of their X-ray spectra alone (Hasinger et al. 2001). Figure 6 shows X-ray spectral diagnostic diagrams based on the hardness ratios computed using four independent energy bands. The hardness ratios are obtained with the formula HR = (H-S)/(H+S), where H and S correspond to the counts in the harder and softer energy bands, respectively. HR1, HR2 and HR3 compare the 0.2-0.5 vs. 0.5-2 keV, 0.5-2 vs. 2-4.5 keV, and 2-4.5 vs. 4.5-10 keV bands, respectively. A grid representing the expected hardness ratios for power-law models with different values of photon index ($\Gamma$) and hydrogen absorption ( $\log N_{\rm H}$) computed in the observed frame is superimposed on the data. The populations of type-1 and type-2 AGNs occupy different regions in these diagrams. In particular, AGN-1 galaxies populate a limited portion of the diagrams in the soft range (and a particularly narrow HR2 range) while the new XMM-Newton galaxies and known AGN-2 type galaxies have harder spectra than those of AGN-1 galaxies and occupy a larger area (see discussion in Hasinger et al. 2001). Also in this case, most of new XMM-Newton galaxies detected in the mid-IR lie in a clearly separated region with respect to the type-1 AGNs.

If we admit that unclassified galaxies are all type-2 AGNs, we detect at 15 $\mu$m at the 3$\sigma$ level 7 AGN-1 galaxies and 15 AGN-2 galaxies (only three of these are classified as AGN-2). Although the statistics are poor, the fraction of AGN source types matches that found in the CFRS field 1415+52 using a multi-wavelength method to classify the galaxies. In this case, studying a sample of 19 ISOCAM sources, Flores et al. (1999) classified two sources as AGN-1 and three as AGN-2. For two other sources the classification as AGN-2 or starburst galaxies is equally probable.

4.3 Comparison with local templates

Detailed spectral energy distributions (SEDs) have been obtained in the hard X-ray band (with ASCA and Beppo-SAX) and in mid- to far-IR (with ISO) for few local galaxies which are representative of the classes of objects found in our samples.

Before analysing X-ray luminosities and X-ray to mid-IR spectral indices of the galaxies of our samples, we discuss the template galaxies which will be compared with our data.

Type-1 AGNs. Mrk 509 and NGC 4593 have been chosen as typical Seyfert 1 galaxies (data from Clavel et al. 2000; Perola et al. 2000; Guainazzi et al. 1999), while PG 1613+658 has been taken as representative of radio-quiet quasars (data from Haas et al. 2000; Lawson & Turner 1997).

Type-2 AGNs. This class of objects is expected to be easily detected by combined hard X-ray and mid-IR surveys, since almost all the UV and soft X-ray emission of the nucleus is reprocessed into infrared light. We consider four examples with different column densities. NGC 1068, the archetypal object for the class of Seyfert 2 galaxies (data from Sturm et al. 2000; Matt et al. 1997), has an extreme Compton-thick nucleus ( $N_{\rm H} > 10^{25}$ cm^-2; Matt et al. 1997). Due to this fact, it has a mid-IR to X-ray flux spectral index which is more typical of starburst galaxies than type-2 AGNs (see Fig. 8). We consider two moderately Compton-thick Seyfert 2s: Circinus and NGC 6240 ( $N_{\rm H} \sim 4.3 \times 10^{24}$ cm^-2, Matt et al. 1999a, and $N_{\rm H} \sim 2 \times 10^{24}$ cm^-2, Vignati et al. 1999, respectively). Circinus is a Seyfert 2 object with a reflection-dominated spectrum in the 2-10 keV range and a transmitted component above 10 keV (data from Sturm et al. 2000; Siebenmorgen et al. 1997; Matt et al. 1999; Sambruna et al. 2001). NGC 6240 (data from Charmandaris et al. 1999; Vignati et al. 1999), according to Vignati et al. (1999) is dominated by the AGN and not from star formation, as deduced by Genzel et al. (1998) on the basis of the ISO spectrum. Finally, we show the Compton-thin luminous IRAS source IR 23060+0505 (data from Brandt et al. 1997 and from the ISO archive) which has $N_{\rm H} \sim 10^{22}$ cm^-2 (Brandt et al. 1997).

Starbursts. M 82 and NGC 253, two of the nearest starburst galaxies, are assumed as typical templates for galaxies with active star formation (data from Sturm et al. 2000; Cappi et al. 1999).

Ultraluminous galaxies. Galaxies of this class, which emit large parts of their bolometric luminosity in the infrared, are known to be powered mainly by star formation, although a small fraction of the emission is probably due to AGN activity (e.g. Lutz et al. 1998; Tran et al. 2001). We chose Arp 220 as an example of an ultraluminous starburst galaxy (data from Sturm et al. 1996; Charmandaris et al. 1999; Iwasawa et al. 2001).

Figure 8: On the left: distribution of the observed mid-IR to hard X-ray spectral indices $\alpha _{\rm IX}$ as a function of redshift. Open, grey and black symbols refer to objects detected in the HDF-N, Elais-S1 and Lockman Hole surveys, respectively. Triangles, diamonds and circles represent type-1 AGNs, type-2 AGNs and unknown types, respectively. HDF-N galaxies with $L_{\rm X} \le 10^{40}$ erg s-1 are drawn with small symbols. Templates derived from various types of known active and starburst galaxies are shown (see the text). The horizontal band shows the $\alpha _{\rm IX}$ of the cosmic background. On the right: $\alpha _{\rm IX}$ versus the 2-10 keV flux. Median values in the flux ranges assumed to compute the AGN contribution to the mid-IR background are drawn.

4.4 X-ray luminosities

Since we have spectroscopic and photometric redshifts for almost all the galaxies of our samples, it is possible to compute X-ray luminosities of these galaxies and compare them with those of local templates.

To compute the luminosity distance we assume a cosmology of H₀ = 70 km s^-1 Mpc^-1, $\Omega _\lambda =0.7$, and $\Omega _{\rm M}=0.3$ using the formula in Carroll et al. (1992). As we can note in Fig. 7, the sources in the Lockman Hole sample have 2-10 keV rest-frame luminosities between 10^42.5 erg s^-1 and 10⁴⁵ erg s^-1, which are typical of luminous type-2 AGNs and normal type-1 AGNs. Sources detected in the Elais-S1 have luminosities typical of type-1 AGNs. Finally, among the sources detected by Chandra in HDF-N, we find low luminosities sources at low redshift which are probably starburst galaxies, few galaxies in the luminosity range populated by Lockman sources, and a population of galaxies with intermediate luminosities which could be ultraluminous infrared galaxies or low-luminosity type-2 AGNs.

In the same Fig. 7 we plot also the hard X-ray luminosity versus the R - K colour. This allows one clearly to segregate normal galaxies which are faint X-ray sources and $R-K \sim 2$ and type-1 AGNs, which populate the left upper corner of the diagram. It is still difficult to distinguish type-2 AGNs from ultraluminous galaxies.

4.5 Mid-IR to X-ray spectral index

A way to combine the information coming from mid-IR and X-ray fluxes is to compute the mid-IR to X-ray spectral index $\alpha _{\rm IX}$ assuming a power law spectral energy distribution: $F_{\nu}\propto \nu^{-\alpha_{\rm IX}}$. Values reported in Tables 3 and 4 are computed using the observer frame flux densities at 15 $\mu$m and 5 keV. The flux densities at 5 keV have been derived from the observed 2-10 keV fluxes (2-8 keV in the case of HDF-N Chandra data) and spectral indices.

In Fig. 8, which gives $\alpha _{\rm IX}$ as a function of redshift, we report all the sources detected in the Lockman, HDF-N and Elais-S1 surveys. We also show the values of $\alpha _{\rm IX}$ as a function of redshift for the aforementioned local templates.

Galaxies dominated by star formation (starburst and ultraluminous galaxies) have high values of $\alpha _{\rm IX}$ at any redshift. On the contrary, type-1 AGNs have quasi-constant values between 1 and 1.2. Between these two envelopes of curves we find the templates of moderately Compton-thick and Compton-thin type-2 AGNs. Only the type-2 Seyfert NGC 1068, which has an extreme Compton-thick nucleus, lies in a region of the diagram occupied by starburst-dominated galaxies. In fact, if the column density exceeds 10²⁵ cm^-2the nuclear radiation is heavily obscured also in the hard X-ray band (see e.g. Matt et al. 2000).

Most of the galaxies detected in the Lockman Hole survey populate the region of the diagram delimited by moderately Compton-thick and Compton-thin type-2 AGNs. The galaxies detected in the Elais-S1 surveys lie around the type-1 AGN curves.

The HDF-N survey, due to its high sensitivity, is able to detect also non-active galaxies with high $\alpha _{\rm IX}$ index. In fact, half of the HDF-N sources lie just below the curves of starburst and ultraluminous galaxies, while the other half have $\alpha<1.4$. Combined with the information on the X-ray luminosity, we will use this diagram to discriminate between HDF-N sources whose emission is dominated by AGN or star formation activity.

It is interesting to remark that a large part of the type-1 AGNs detected in the Lockman area have an $\alpha _{\rm IX}$ index greater than those of the local templates. Except for one case which is an absorbed type-1 AGN, as revealed by the X-ray hardness ratio diagrams (#79 in Table 3), the most probable explanation is that star formation of the host galaxies contribute a large fraction of the mid-IR flux. Hence, their $\alpha _{\rm IX}$ values should differ significantly from those of local templates, for which we can easily discriminate between the host galaxy and AGN.