Multicone Survey AGN1 Sample

Extensive optical and near-infrared identification programmes using many tens of nights on the largest telescopes of the world were performed on the deep Chandra and XMM-Newton survey fields. Because of the average faintness of the optical counterparts, typical exposure times had to be in excess of 4-5 h, but still the faintest sources escape spectroscopic diagnostics and have to be identified using photometric redshift techniques. For cosmological studies well-defined flux-limited samples of active galactic nuclei have been chosen in [36], with flux limits and survey solid angles ranging over five and six orders of magnitude, respectively. To be able to utilize the massive amount of optical identification work performed previously on a large number of shallow to deep ROSAT surveys, the analysis was restricted to samples selected in the 0.5-2 keV band. In addition to the ROSAT surveys already used in [56], data from the recently published ROSAT North Ecliptic Pole Survey (NEPS) [60], from an XMM-Newton observation of the Lockman Hole [49], as well as the Chandra Deep Fields South (CDF-S) [50,85,98] and North (CDF-N) [3] were included.

Based on deep surveys with Chandra and XMM-Newton, the X-ray log(N) -log(S) relation has now been determined down to fluxes of 2.4 x 10-17,2.1 x 10-16, and 1.2 x 10-15 erg cm-2 s-1 in the 0.5-2, 2-10, and 5-10 keV band, respectively [5]. Figure 25.5 shows the normalized cumulative source counts N(> Sx14)SX54

CO A

Fig. 25.5 Cumulative number counts N(> S) for the total sample (upper blue thin line), the AGN-1 subsample (lower black thick line), the AGN-2 subsample (red dotted line), and the galaxy sub-sample (green dashed line). From [36]

Fig. 25.5 Cumulative number counts N(> S) for the total sample (upper blue thin line), the AGN-1 subsample (lower black thick line), the AGN-2 subsample (red dotted line), and the galaxy sub-sample (green dashed line). From [36]

for the total soft X-ray selected sample, as well as for the subsamples of AGN-1, AGN-2, and normal galaxies. For simplicity we define SX14 = 10-14 erg cm-2 s-1. Euclidean source counts would correspond to horizontal lines in these graphs. A broken power law fitted to the differential source counts yields power law indices of ab = 2.34 ± 0.01 and af = 1.55 ± 0.04 for the bright and faint end, respectively, a break flux of SX14 = 0.65 ± 0.10, and a normalisation of dN/dSX14 = 103.5 ± 5.3 deg-2 at SX14 = 1.0 with a reduced x2=151. We see that the total source counts at bright fluxes, as determined by the ROSAT All-Sky Survey data, are significantly flatter than Euclidean, consistent with the discussion in [31]. Moretti et al. [58], on the other hand, have derived a significantly steeper bright flux slope (ab « 2.8) from ROSAT HRI pointed observations. This discrepancy can probably be attributed to the selection bias against bright sources, when using pointed observations where the target area has to be excised.

The ROSAT HRI Ultradeep Survey had already resolved 70-80% of the extra-galactic 0.5-2 keV XRB into discrete sources, the major uncertainty being in the absolute flux level of the XRB. The deep Chandra and XMM-Newton surveys have now increased the resolved fraction to 85-100% [58,92]. Above 2 keV the situation is complicated by the uncertain normalization of the background measurements. The 30% higher normalization discussed above would require that the resolved fractions above 2 keV have to be scaled down correspondingly. On the other hand, the 2-10 keV band has a large sensitivity gradient across the band. A more detailed investigation, dividing the recent 770 ks XMM-Newton observation of the Lockman Hole and the Chandra Megasecond surveys into finer energy bins, comes to the conclusion that the resolved fraction decreases substantially with energy, from over 90% below 2 keV to less than 50% above 5 keV [92]. The flux resolved into discrete sources is indicated in Fig. 25.3b.

To avoid systematic uncertainties introduced by the varying and a priori unknown AGN absorption column densities, only unabsorbed (type-1) AGN classified by optical and/or X-ray methods were selected. Hasinger et al. [36] are using a definition of type-1 AGN, which is largely based on the presence of broad Balmer emission lines and small Balmer decrement in the optical spectrum of the source (optical type-1 AGN, e.g., the ID classes a, b, and partly c in [73], which largely overlaps the class of X-ray type-1 AGN defined by their X-ray luminosity and unabsorbed X-ray spectrum [85]. However, as Szokoly et al. show, at low X-ray luminosities and intermediate redshifts the optical AGN classification often breaks down because of the dilution of the AGN excess light by the stars in the host galaxy, so that only an X-ray classification scheme can be utilized. Schmidt et al. [73] have already introduced the X-ray luminosity in their classification. For the deep XMM-Newton and Chandra surveys in addition the X-ray hardness ratio was used to discriminate between X-ray type-1 and type-2 AGN.

Type-1 AGN are the most abundant population of soft X-ray sources. For the determination of the AGN-1 number counts we include a small fraction of unidentified sources (6%), which have hardness ratios consistent with AGN-1. Figure 25.5 shows that the break in the total source counts at intermediate fluxes is produced by type-1 AGN, which are the dominant population there. Both at bright fluxes and at the faintest fluxes, type-1 AGN contribute about 30% of the X-ray source population. At bright fluxes, they have to share with clusters, stars, and BL-Lac objects, and at faint fluxes they compete with type-2 AGN and normal galaxies (see Fig. 25.5 and [5]). A broken power law fitted to the differential type-1 AGN source counts yields power law indices of ab = 2.55 ± 0.02 and af = 1.15 ± 0.05 for the bright and faint end, respectively, a break flux of SX14 = 0.53 ± 0.05, consistent with that of the total source counts within errors, and a normalization of of dN/dSX14 = 83.2±5.5deg~2 at SX14 = 1.0 with areduced%2 = 126. The AGN-1 differential source counts normalized to a Euclidean behavior (dN/dSX14 x ^X54) is shown with filled symbols in Fig. 25.5.

0 0

Post a comment