2 General properties of the faint X-ray sources in Omega Cen

2.1 The XMM-Newton observation

We observed $\omega$ Cen on 2001 August 13 with the XMM-Newton EPIC MOS (Turner et al. 2001) and PN (Strüder et al. 2001) cameras, using a full frame window mode and a medium filter. The observation was 40 kilosecond long with a low and stable background. The data were analyzed with the latest version  (5.3.3) of the XMM-Newton Science Analysis Software (SAS). We used the calibration chains of the EPIC cameras, emchain and epchain, using the embadpixfind task to flag bad pixels and bad columns. We filtered the event files produced for good time intervals and non astrophysical events (electronic noise, cosmic rays). We used the predefined patterns, keeping only patterns 0-12 for the MOS detectors and patterns 0-4 for the  PN, and we rejected all the events flagged as "bad'' by the calibration chains. Finally, we also rejected events with energies below 0.4  keV and above 10 keV, because of a high number of bad patterns.

2.2 Source detection

Sources were searched between 0.5 and 5.0 keV, a range which encloses the peak of the effective area of the EPIC cameras. A wavelet detection algorithm was used. It is better suited to crowded fields than the sliding box algorithm. Given the early development stage of the task, we used a conservative 4 $\sigma$ as the detection threshold. For each camera, the source list so obtained was used as an input to the task emldetect. emldetect computes for each source a maximum likelihood, taking into account the point spread function of the instrument. For each source, the task returns its best fit position, the statistical errors on this position ($1 \sigma$ or 68% confidence level), its count rate, and a maximum likelihood detection value. We used a maximum likelihood threshold of 12. In order to estimate the statistical error at the 90% confidence level for the source positions, we modified the public version of emldetect following the recommendation of the task author (for a two parameter fit, the 90% confidence limit level is given by likelihood + 6.18, Lamer, private communication). Three PN sources were removed because their best fit positions fell onto bad columns. The cleaned PN, MOS1 and MOS2 source lists were then correlated using a customized version of the task srcmatch. srcmatch returns the positions of the correlated sources weighted by their statistical errors as derived for each instrument separately. The maximum likelihood of a correlated source is the sum of the individual maximum likelihoods.

Figure 1: A false color image of the XMM-Newton field of view. It combines the EPIC-PN and MOS images. The color bands we used were 0.5-1.5 keV (red), 1.5-3.0 keV (green) and 3.0-10.0 keV (blue).

The majority of the correlated sources are detected by the PN camera because it is more sensitive than the MOS cameras. There are however some sources missed by the PN, due to CCD gaps, or bad columns and the smaller field of view of the PN compared to the MOS; these sources are marked MOS only in Tables 1 and 2.

146 sources were detected by the EPIC cameras; 59 are seen only by the PN, 9 are MOS only (and all of them are seen in MOS 1 and MOS 2), and the remaining 78 are detected in the PN and at least one MOS camera. We present a false color combined PN and MOS image of the field of view in Fig. 1. The positions and statistical errors given at the 90% confidence level, the 0.5-5.0 keV source count rate and associated error are listed in Table 1 for those sources lying within the half mass radius and in Table 2 for the remaining sources. In EPIC-PN, a count rate of 10^-2 counts s^-1 corresponds to an unabsorbed flux of $\sim$ $2.7\times 10^{-14}$ ergs s^-1 cm^-2 for a 0.6 keV blackbody model absorbed through the interstellar absorption derived from the optical extinction ( $\mbox{$N_{\rm H}$ }= 8.4\times 10^{20}$ cm^-2, Djorgovski 1993; Predehl & Schmitt 1995). Assuming a 3 keV thermal Bremsstrahlung and a power law of photon index 2, the corresponding fluxes are $\sim$ $2.5\times10^{-14}$ ergs s^-1 cm^-2 and $\sim$ $2.4\times10^{-14}$ ergs s^-1 cm^-2 respectively.

For the blackbody model, this flux translates to a luminosity of $\sim$10³² ergs s^-1 at the distance of $\omega$ Cen (5.3 kpc, Harris 1996; Thompson et al. 2001). Thus in Tables 1 and 2 one can see that the bulk of sources have luminosities in the range 10³¹-10³² ergs s^-1.

2.3 Background sources

Some of the sources listed in Tables 1 and 2 are extragalactic background sources unrelated to the cluster. In order to estimate their number, we used the statistical Log N-Log S relationship of extragalactic sources derived from the Lockman Hole XMM-Newton data (Hasinger et al. 2001). To account for the vignetting function of the XMM-Newton mirrors, we have computed limiting count rates (for source detection) within different annuli (all centered on the cluster center), using an approach similar to Cool et al. (2002). The radius of each annulus is computed such that the annulus contains a large ($\ge$25) number of sources (the radius varies between 2.5 and 4.2$^\prime$). This allows us to set the limiting count rate to the count rate of the weakest source detected in that annulus. The limiting count rate is a factor of $\sim$2 larger in the outer annulus than in the inner annulus. For direct comparison with the Log N-Log S curve of Hasinger et al. (2001), these count rates have then been converted into unabsorbed 0.5-2.0 keV fluxes using a power law model of index 2.0 absorbed through the cluster $N_{\rm H}$. Following this procedure, after a proper surface normalization, one expects 4, 9, 35 and 65 background sources within the core, half mass, twice the half mass and a 12.5$^\prime$ radii (the values so obtained were rounded to the nearest integer). Beyond 12.5$^\prime$, where we do not expect any cluster sources, the number of detected sources matched the one estimated with this procedure.

As an indication, we have computed the error on the above estimates assuming a 10% uncertainty on the limiting count rate estimate and a 10% uncertainty in the XMM-Newton calibration (note that the Log N-Log S relationship was derived from a processing of the Lockman hole data with the SAS prior to its first public release). This gives an error of 1 and 2 on the estimated number of background sources within the core and half mass radii.

2.4 Color-color and hardness-intensity diagrams

In order to investigate the general properties of the sources detected by XMM-Newton, we have computed X-ray color-color and hardness-intensity diagrams. For this purpose, we have produced PN images in three adjacent energy bands: 0.5-1.5 keV, 1.5-3.0 keV, 3.0-10.0 keV (similar bands were used by Grindlay et al. 2001). From these images, we have computed the net exposure corrected source count rate. To produce meaningful diagrams we have considered only sources detected with more than 3 counts in each band. There are 71 sources fitting this criteria. Two sets of diagrams have been computed: one for the sources found within a region of radius equal to twice the half mass radius and one for the whole field of view. They are presented in Figs. 2 and 3.

Figure 2: The color-color diagram of the sources detected by the EPIC-PN camera within twice the half mass radius (left) and within the field of view (right). The star identified by Cool et al. (1995), the quiescent neutron star binary candidate and the two CV candidates are represented by an open circle, an open diamond and two filled squares respectively. Unknown sources are represented by a filled circle, a filled star and an open star if the source lies within the core radius, within the half mass radius, or outside the half mass radius respectively. A representative error bar is shown. Each source is labeled according to Tables 1 and 2.

Figure 3: The hardness-intensity diagram of the sources detected by the EPIC-PN camera within twice the half mass radius (left) and within the whole field of view (right). The symbols refer to the same objects as Fig. 2. Each source is labeled according to Tables 1 and 2. The intensity is corrected for the vignetting of the mirrors. A representative error bar (source 42) is shown.