3 Spectral analysis

3.1 Spectra extraction

The purpose of this work is to perform an X-ray spectral analysis of the sources in the Lockman Hole field, taking advantage of the large collecting area of the XMM-Newton satellite. This represent a step forward respect to the hardness ratios diagnostic diagrams and stacking techniques (Tozzi et al. 2001; Alexander et al. 2001a) in which the range of source redshifts will smear out the signature of absorption and other X-ray spectral features (e.g., the iron K$\alpha$ line).

We use an automated procedure to extract the X-ray spectra of the 98 sources. Firstly, a source catalog is constructed using the SAS detection algorithm (see Paper I for details on the detection process). We then perform the source detection using SExtractor (Bertin et al. 1996) on the same image ([0.5-7] keV band). SExtractor yields shape elliptical parameters for each source (the semi-major/minor axes and the orientation angle) which are added to the main SAS catalog by cross-correlating the two source lists.

Elliptical parameters for each source are used to define the appropriate region for the extraction of the spectrum, thus taking into account the broadening of the PSF at increasing off-axis angles. The background region is defined as an annulus around the source, after masking out nearby sources. The XSELECT tool is used to extract the spectrum, and the GRPPHA tool is used to bin the data so as to have at least 20 counts per bin. In this process, the background count rate is rescaled with the ratio of the source and background areas.

Figure 1: The power-law photon index ($\Gamma$) versus $\log(N_{\rm H})$. Filled circles show the Type-1 AGN spectroscopically identified in the ROSAT ultradeep HRI survey (Lehmann et al. 2001a) and in the on-going optical follow-up of the newly detected XMM-Newton sources (PI: Maarten Schmidt), open circles are Type-2 AGN, crosses are three still unidentified objects for which we have photometric redshift estimations (Lehmann et al. 2001a). The histogram shows the spectral index distribution (shaded area). For both parameters, error bars correspond to $90\%$ confidence level for one interesting parameter ( $\Delta \chi ^2=2.706$).

Figure 2: Intrinsic $N_{\rm H}$ distribution for sources inside an off-axis angle of 10 $\hbox {$^\prime $ }$ with more than 70 counts in the [0.5-7] keV band. The shaded part indicates the sources without redshift, for which the measured $N_{\rm H}$ value represents a lower limit.

   
3.2 Spectral parameters

We use XSPEC (v11.1) for the spectral fitting analysis. As a first approximation, a powerlaw model with an intrinsic absorption (wabsor zwabs if the redshift is known) is used. An additional photoelectric absorption component (wabs) fixed to the Galactic column density is also included in the model.

This fit yields the power-law photon index $\Gamma$, the intrinsic column density $N_{\rm H}$, and the X-ray luminosity in the [0.5-2] and [2-10] keV rest-frame bands. A clear soft excess is present in several sources (especially the absorbed ones). In order to reproduce this feature we add two separate components to the baseline model (wabs$\ast$zwabs$\ast$powerlaw): a blackbody or an extra powerlaw. Extra parameters measured from this composite fit (second power-law index or blackbody temperature) are not reported in Table 2.

Figure 3: X-ray diagnostic diagrams based on hardness ratios. Filled circles show the Type-1 AGN spectroscopically identified in the ROSAT ultradeep HRI survey (Lehmann et al. 2001a) and in the on-going optical follow-up of the newly detected XMM-Newton sources (PI: Maarten Schmidt). Type-2 AGN are marked with open circles and unidentified sources with crosses. The large filled circles are EROs ( $R-K^\prime \geq 5$). The sources inside a box are those with an intrinsic $\log(N_{\rm H}) > 21.5$. For clarity, only sources with hardness ratio errors less than 0.1 are plotted. The source $\char93 50$ is indicated, see Sect. 7.3 for more details.

In Fig. 1, we plot $\Gamma$ versus the column density $N_{\rm H}$ for the sources with known redshifts. This diagram suggests that the intrinsic slope of the X-ray spectrum is the same for all the objects whatever the absorption levels, with $\langle\Gamma\rangle \simeq 2$. Therefore the increasing hardness of the source spectra at fainter fluxes observed in the Chandra deep fields (Giacconi et al. 2001; Tozzi et al. 2001; Brandt et al. 2001) is probably due to intrinsic absorption and not to an intrinsically hard power law. Several teams (Della Ceca et al. 1999b; Akiyama et al. 2000; Fiore et al. 2001; Maiolino et al. 2001a; Page et al. 2001; Reeves et al. 2000; Risaliti et al. 2001) reported the existence of AGN optically classified as Type-1 but with an indication of absorption in their X-ray spectra. In our sample, there are seven objects with an intrinsic absorption between 10²¹ and 10²² cm^-2 which are optically classified as Type-1 AGN. Moreover, the source $\char93 96$ (ROSAT $\char93 39$) shows a high intrinsic absorption of $5^{+5}_{-3} \times 10^{22}$ cm^-2 while it was optically classified as an unabsorbed QSO at a redshift of 3.279. In these cases the optical classification is de-coupled from the X-ray classification. This could be due to a gas-to-dust ratio and/or a chemical composition different from those in Galactic interstellar gas (Akiyama et al. 2000; Maiolino et al. 2001b). The three sources with photometric redshifts have an absorption greater than 10^21.5 cm^-2 which, combined with the $R-K^\prime$ colours, suggest that they are probably obscured AGN (Lehmann et al. 2001a).

   
3.3 Observed N_H distribution

The $N_{\rm H}$ distribution and its cosmological evolution are key ingredients in the XRB synthesis models (Comastri et al. 1995; Gilli et al. 2001). In Fig. 2 we show the $N_{\rm H}$ distribution for the 38 sources with an off-axis angle <10 $\hbox {$^\prime $ }$. In this central region where the exposure time is approximately constant, our threshold of 70 counts in the [0.5-7] keV band corresponds to a flux of $1.6 \times 10^{-15}$ erg cm^-2 s^-1. The surface density of sources down to this flux limit, $\sim$1700 deg^-2, is in very good agreement with that derived from the $\log N-\log S$ relation given in Paper I. Therefore our sample can be regarded complete and the derived $N_{\rm H}$ distribution representative of the overal AGN population at the afore mentioned flux limit.

3.4 X-ray colour-colour diagrams

A useful method to constrain the nature of X-ray sources, in particular when the signal-to-noise ratio is not high enough for spectral analysis, is to use X-ray colour-colour diagrams (e.g. Della Ceca et al. 1999a; Paper I). In Fig. 3 we present two of these diagrams. We have used the energy bands: 0.2-0.5 keV (US), 0.5-2 keV (S), 2-4.5 keV (M) and 4.5-10 keV (H) to define three different hardness ratios: $\rm HR1=(S-US)/(S+US)$, $\rm HR2=(M-S)/(M+S)$, $\rm HR3=(H-M)/(H+M)$. We use different symbols to indicate Type-1 AGN, Type-2 AGN and unidentified sources. The sources with $\log(N_{\rm H}) > 21.5$ have special labels (square box); when the redshift is unknown, the derived column densities are only lower limits. We also highlight the sources with $R-K^\prime \geq 5$, usually called Extremely Red Objects (see Sect. 4.2 for a discussion of the properties of this class of objects).

In both diagrams, Type-1 AGN are confined in a small region, as opposed to Type-2 AGN and unidentified sources which are spread over a much broader area with high hardness ratios (see also Paper I). Moreover, using the additional information on the measured intrinsic absorption column density, it is now clear that the hardening of non-Type-1 sources is mainly due to the presence of intrinsic absorption with $\log(N_{\rm H}) > 21.5$ superimposed on relatively soft spectra (see Fig. 1), rather than intrinsically hard spectra (this is also consistent with the fact that Type-1 and Type-2 AGN span similar range in HR3). These diagrams also suggest that a large fraction of the unidentified sources (mainly newly detected XMM-Newton sources) are X-ray obscured AGN.

Figure 4: Color-magnitude diagram, $R-K^\prime$ versus R a) and $R-K^\prime$ versus $K^\prime$ b), of Lockman Hole sources. Symbols are as in Fig. 3, while field sources are marked with small dots. The four evolutionary tracks correspond to an unreddened QSO with $L=L^\ast _{\rm B}$ and $z=0\div 6$ (solid line), and to an unreddened elliptical, Sbc and irregular ${L}^\ast$ galaxies ($z=0\div 3$) from the Coleman et al. (1980) template library (dashed lines).