7 Discussion

We have presented the first results of our deep XMM-Newton observation of the globular cluster $\omega$ Cen, emphasizing the general properties of the population of faint X-ray sources present in the cluster. We have detected 11 and 27 faint X-ray sources within the core and half-mass radii respectively. We have estimated that $4\pm1$ and $9\pm2$ of these objects could be unrelated background sources. Comparing the Chandra ACIS-I and XMM-Newton EPIC observations, we have found that 63 XMM-Newton sources have a Chandra counterpart. Fifteen sources that lie within the half mass radius have shown variability between the two observations. Intensity variations for several sources were also found within the XMM-Newton observation and between the XMM-Newton and previous ROSAT observations. We have also presented the first X-ray spectra of the brightest and peculiar objects in the field, in particular the proposed quiescent neutron star low mass X-ray binary for which the EPIC spectrum strengthens its classification. We have shown that the spectra of the two brightest core sources strongly support the proposal that they are CVs. We have found objects with similar spectra in the cluster. In the following, we briefly discuss what appears to be the main implications of our observation.

First of all, what is striking from our data is the excess of sources located in the vicinity of the cluster (just outside the half mass radius). This has already been noted by Cool et al. (2002) and Verbunt (2002a). For instance between 1 and 2 half-mass radii, there is an excess of $\sim$25 sources, over the expected number of background sources. Obviously, some of them may be foreground stars. However, some may also belong to the cluster, and may have already been found as active binaries (RS CVn). Two of the OGLEGC sources located at relatively large off-axis (more than 2.5 core radii) were detected by Chandra (OGLEGC15 and OGLEGC22, Cool et al. 2002). Another one (OGLEGC30) has a position coincident with our XMM-Newton source 29. All are proposed to be RS CVn stars (Kaluzny et al. 1996; Kaluzny et al. 2002). Their X-ray luminosity is not atypical of such systems $\sim$ $1{-}3 \times 10^{31}$ ergs s^-1, though on the bright end of their luminosity distribution (Dempsey et al. 1997).

Due to mass segregation effects, binaries which are more massive than lonely stars are expected to lie close to cluster center (e.g. Meylan & Heggie 1997). In $\omega$ Cen  the mass segregation is very low, and that might explain the large population of binaries outside the core (e.g. Verbunt & Johnston 2000). Some of these binaries might also have been ejected outside the cluster through three-body interactions (an encounter of a binary with a single star). Another possibility could be that the potential well of the cluster was recently disrupted by the accretion of a discrete component of another stellar system. The recent discovery of a metal-rich stellar population in the cluster (Pancino et al. 2000) with a coherent bulk motion with respect to the other stars (Ferraro et al. 2002) was explained by the accretion of an independent stellar system by $\omega$ Cen. The presence of faint X-ray sources in the vicinity of $\omega$ Cen might thus be a consequence of an unusual dynamical evolution of the cluster.

The second most striking feature of our data is the lack of soft X-ray sources and the large number of sources showing long term variability. In Figs. 2 and 3, there is an obvious clustering of sources around the two previously identified CVs (which are spectrally hard). There are only two sources in the soft area: the star identified by Cool et al. (1995) and the proposed quiescent neutron star binary (note that these two objects are also among the brightest, see Fig. 3). The sources clustering below the two CV candidates (sources 2 and 5, represented by two filled squares) in the hardness intensity plot (see Fig. 3) have luminosities in the range $2 \times 10^{31}$ to $6 \times 10^{32}$ ergs s^-1. The spectra of four of these objects are shown in Fig. 8. Similar spectra and luminosities are observed from disk and globular cluster CVs (e.g. Pooley et al. 2002a). CVs are well known to be variable, and might thus account for some of the variable sources present in the cluster (note that one core CV, source 5, showed variability by a factor of $\sim$1.7 between the XMM-Newton and Chandra observations).

However, we note that one of the proposed RS CVns (OGLEGC30, Kaluzny et al. 1996, the counterpart of the XMM-Newton source 29) has colors similar to the two CVs (the statistic was unfortunately too poor to fit its spectrum). This source was not detected by Chandra and must have varied by at least a factor of $\sim$3-4 between the two observations. The two OGLEGC sources (RS CVn candidates) detected by Chandra were not detected by XMM-Newton: the Chandra luminosity of OGLEGC 15 was below the XMM-Newton sensitivity threshold, on the other hand OGLEGC 22 should have been detected. Another example of variability is given by the ROSAT source R20, associated with a BY Dra (Cool et al. 2002). This source was detected by Chandra but not by XMM-Newton. These objects are variable, and will be preferentially detected during flaring outbursts, due to their low quiescent X-ray luminosities. The large number of variable sources in $\omega$ Cen (15 within the half mass radius between the Chandra and XMM-Newton observations) is also suggestive of a large population of RS CVns (including BY Dra) in the cluster. During snapshot X-ray observations, only a fraction could be seen. These sources could account for the population of the lower luminosity ($\sim$10³¹ ergs s^-1) sources found in $\omega$ Cen.

Some of the faint and persistent sources of the hardness-intensity diagram ($\sim$10, see Fig. 3) have colors consistent with power law like spectra, with indices of the order of $\sim$2. Such power laws could result from magnetospheric emission of millisecond pulsars (Becker & Trümper 1998; Webb et al. 2002b). However, these sources are unlikely to be millisecond pulsars, because no such radio pulsars are presently known in the cluster (Freire 2002). Furthermore, Grindlay et al. (2002) have recently shown that the emission of the millisecond pulsars detected in 47 Tuc is dominated by the thermal emission from the polar caps of the neutron star. Such emission is much softer than the magnetospheric emission. These sources with soft thermal X-ray emission should lie in Fig. 3 between vertical lines passing through the two CV candidates and the quiescent neutron star binary. There are no sources in that region. Our observation would thus support the idea that $\omega$ Cen lacks millisecond pulsars. Obviously the difficulties in retaining neutron stars in a globular cluster and the low collision frequency of $\omega$ Cen could provide an explanation (see e.g. Pfahl et al. 2002; Verbunt 2002a).

In the disk, quiescent neutron star binaries have luminosities in the range 10³²-10³³ erg s^-1 (Narayan et al. 2002). Furthermore, they all have extremely soft X-ray spectra (Rutledge et al. 2000). In our observation, there is one single object with these characteristics. If globular cluster quiescent neutron star binaries behave similarly to those in the disk, then our observation should provide a complete census of the content of such objects in $\omega$ Cen (a similar conclusion was derived by Rutledge et al. 2002, from the Chandra observations). Some of these objects have also been found in other clusters (Edmonds 2002; Grindlay et al. 2001; Pooley et al. 2002b). In globular clusters, these systems are certainly formed from a close encounter between a neutron star with a single star or with a binary (see Verbunt 2002b, for a recent review). The presence of one such system in $\omega$ Cen is consistent with its lower collision frequency compared to other clusters (Verbunt 2002b). The presence of this object far away from the core remains somewhat puzzling in that regard. It might have been ejected from the core during a three-body interaction (Verbunt 2002b). The presence of a single candidate quiescent neutron star binary together with the apparent lack of millisecond pulsars in the cluster makes $\omega$ Cen clearly different from clusters like 47 Tuc in which a large population of binaries with neutron star primaries are being discovered (e.g. Grindlay et al. 2001; Camilo et al. 2000). This difference, if confirmed, should help us in understanding how neutron stars form and evolve in globular clusters.

Finally, some of the faint sources could be quiescent black hole binaries (one black hole may have just been discovered in the globular cluster M15, and some are being found in globular clusters in other galaxies, Gerssen et al. 2002; Verbunt 2002b). Most quiescent black hole binaries within the galactic disk have been observed with luminosities of $\sim$10³¹ erg s^-1 (Kong et al. 2002; Hameury et al. 2003). Their spectra are well fitted with power laws, with spectral indices of 1.5-2 (Kong et al. 2002). Such spectra correspond to hardness ratios in the range 0.4-0.9 in Fig. 3. About 10 objects have colors and luminosities consistent with a quiescent black hole binary nature. This hypothesis is however poorly constrained from our X-ray observations alone, as many faint sources could be background active galactic nuclei.