Extended source diagnostics

We now turn to the properties of the extended X-ray sources and their detection efficiency. Figure 6.20 provides a global summary of all extended X-ray sources detected with the single band detection scheme. Plotted are the maximum likelihoods for the extent of the source (EXT ML) versus the total source counts in the 0.35-2.4 keV band. The horizontal cut-off at EXT ML=5 indicates the lower significance threshold for the detection run.

Distant cluster candidates are represented by dark blue circles. The median number of total source counts for this population is 193, the median EXT ML is 11.7, corresponding to 5–6σ significance, and the median detection likelihood amounts to 36.6. The lighter blue circles indicate the population ofDSS-identified cluster candidates. Red symbols illustrate the distribution of sources classified as spurious extent (plus signs) or spurious detections (crosses). The extent maximum likelihood exhibits a decent correlation with the total photon number of the cluster candidates (blue symbols), with a scatter of about a factor of two. The extended sources classified as spurious show on average a lower EXT ML for a given total number of source counts. Even though spurious sources show a systematic offset in theirEXT ML, they still occupy almost the full parameter space populated by cluster sources, in particular at the important fainter levels.

The first important conclusion from this global view on the detected extended source population is that spurious sources cannot be automatically separated from real clusters by means of appropriate parameter cuts. This is to be attributed to the discussed system- atic (calibration) effects that dominate the spurious detections for the eboxdetect and

emldetect methods. The second preliminary conclusion is that the sensitivity goal for extended sources down to a total number of 100 counts and less seems to be achievable. The typical significance at which objects with ∼100 X-ray photons can be resolved into extended sources is about 4σ. The final survey sensitivity limit has to be quantified based on detailed simulations (see Sect. 9.1.3).

As a next step, we can compare the extent likelihoods of the sources for the three different detection schemes, which is shown in Fig. 6.21 for the distant cluster candidate sample. The EXT MLs of the standard scheme are plotted along the abscissa, whereas the corresponding source likelihoods for the single band scheme (dark squares) and spectral matched filter scheme (light blue circles) are represented by the ordinate.

The likelihoods for thesingle band schemeand standard schemescatter along the lower red line which indicates matching values. The corresponding comparison between the schemes with a wider spectral coverage exhibits a slightly tighter correlation about the upper red line, which indicates the average offset factor of 2.8 and corresponds to the

Figure 6.20: Extent likelihood versus X-ray source counts. Dark blue circles represent distant cluster candidates, light blue circles DSS-identified clusters, and red symbols sources that have been classified as spurious detections. The lower horizontal line indicates the minimum required extent likelihood, whereas the diagonal line shows the general correlation for the good sources. Sources that have been classified as spurious tend to have a lower extent likelihood compared to clusters candidates with the same number of counts. However, the populations are still largely overlapping which prevents an automated identification of false detection.

mean effective weight. The effective weighting factor of the spectral matched filter scheme is not fixeda priori, but depends on the spectral properties of the source. Below standard schemeEXT MLs of about 8, the distant cluster candidates appear to be systematically above the average correlation which points towards an achieved enhancement of the likelihoods at faint flux levels. The global statistics for thespectral matched filter detections confirm that the average likelihood offsets with respect to the standard scheme are increased by about 6% for distant cluster candidates compared to sources classified as spurious. However, no firm conclusion on the effectiveness of the spectral matched filter scheme can be drawn before a significant number of distant candidate clusters at lower EXT ML levels have been spectroscopically confirmed.

Last but not least important, the optimal detector area for serendipitous surveys is investigated. As discussed in Sect. 6.1, XMM-Newton’s grasp as survey instrument in- creases with the maximum off-axis angle Θ. On the other hand, the vignetting effect (Fig. 6.2) decreases the sensitivity towards the outskirts, and more importantly, the PSF shape (Fig. 6.3) broadens and becomes increasingly ill-behaved at large Θ, which hampers the detection of extended sources.

Figure 6.21: Extent likelihood comparison of the three detection schemes. The dark squares display the extent likelihoods of the single 0.35-2.4 keV detection band versus the EXT MLs of the standard detection schemefor distant cluster candidates with the lower red line indicating matching likelihoods. The top line and the lighter blue circles show the situation for thespectral matched filter schemewhich exhibits an offset of roughly a factor of 2.8 corresponding to the average effective weight.

Figure 6.22: Cluster detections per unit sky area as a function of off-axis angle normalized to the detections at 100_{off-axis angle (dashed red line). The detection rate is almost constant between 7}0 _and 120 from the optical axis. In the central region the detection rate rises owing to the increased effective area. Beyond 120 _{the increasing vignetting and the broadened PSF decrease the cluster detections per} unit area. The inner 120 (solid red line) are well-behaved and can be used for the core survey.

One way to address the issue of the maximum off-axis angle for a strictly selected survey sample is to ask at what point the cluster detections per unit sky area start dropping significantly. Figure 6.22 shows the results obtained from 120 fields without masked regions, i.e. no large foreground objects, and a total of 375 cluster sources (screen level 3 objects). The plot traces the normalized cluster detections per unit sky area as a function of off- axis angle, which is derived from the cumulative distribution N(< Θ) divided by the enclosed sky area Ω(<Θ) and normalized to the average value within 100 _{radius (dashed} red line). From Fig. 6.22 it can be seen that the cluster detections per unit area are almost constant between off-axis angles of 70_<

∼Θ∼<120. At larger radii the detection rate decreases

significantly. Hence, a maximum off-axis angle of Θmax=120 promises to be the best

compromise between survey grasp and completeness. With this cut-off radius for the final XDCP core sample, the XMM survey grasp isg12'204 cm2deg2, a 30% improvement with

respect to a 100 _constraint.