1 Introduction

Omega Centauri (NGC 5139, $\omega$ Cen) is one of the best studied objects of our galaxy. It is the most massive globular cluster ( $5.1 \times 10^{6}~M_{\odot}$, Meylan et al. 1995). It is characterized by large core and half mass radii (154.88 $\hbox{$^{\prime\prime}$ }$ and 250.8 $\hbox{$^{\prime\prime}$ }$ respectively, Harris 1996). Binaries are expected to be present in $\omega$ Cen  either as a result of the evolution of primordial binaries, or through close encounters between stars in the cluster (Di Stefano & Rappaport 1992, 1994; Davies & Benz 1995; Verbunt 2002a). Binaries such as CVs, low mass X-ray binaries either with a neutron star or a black hole, or active X-ray binaries (RS CVn or BY Dra systems) or millisecond pulsars could thus form. Some of these binaries have already been found as faint X-ray sources in $\omega$ Cen (Verbunt 2001).

Faint X-ray sources were first detected in $\omega$ Cen by the EINSTEIN X-ray satellite. EINSTEIN detected 5 faint point sources (one in the core, Hertz & Grindlay 1983) and a possible extended emission region, within the half mass radius (Hartwick et al. 1982). A decade after EINSTEIN, ROSAT detected 22 faint sources in the line of sight of $\omega$ Cen (Johnston et al. 1994; Verbunt & Johnston 2000). ROSAT confirmed the EINSTEIN sources, and resolved the core source into three components (Verbunt & Johnston 2000). However, ROSAT did not find any evidence for the diffuse emission seen by EINSTEIN (Johnston et al. 1994).

More recently, Chandra observed $\omega$ Cen and detected over 140 faint X-ray sources (Cool et al. 2002). From follow-up observations using the accurate Chandra positions, Cool et al. (2002) claimed that there were at least three classes of binaries present in the detected sample. Two of the three ROSAT core sources (ROSAT R9a and R9b) may be CVs (Carson et al. 2000). The third ROSAT core source (R20, Verbunt & Johnston 2000) was associated with a main sequence optical counterpart, showing weak H$\alpha$ emission, suggesting a BY Dra system rather than a CV. In addition, two more Chandra core sources were detected with HST/WFPC2, with properties matching those of RS CVn or active-corona binaries (Cool et al. 2002). Finally, Cool et al. (2002) found the X-ray counterparts of two variable binaries discovered far out from the cluster center by Kaluzny et al. (1996). Based on their light curve properties, these two systems are proposed to be RS CVn binaries (Kaluzny et al. 1996, 2002). Beside these three classes of binaries, by looking at the spectral characteristics of the Chandra sources, Rutledge et al. (2002) noticed that one relatively bright object had an extremely soft X-ray spectrum. This spectrum was found to be consistent with those observed from field quiescent neutron star binaries (Rutledge et al. 2002). X-rays would then come from the neutron star surface maintained at a high temperature by episodic mass accretion from a binary companion. In total, up to four different types of binaries may have already been found in the cluster.

We have initiated a survey of nearby globular clusters with XMM-Newton (M22, Webb et al. 2002a, M13, NGC 6366, Gendre et al., in preparation and $\omega$ Cen). Because of the limited angular resolution of XMM-Newton, we have selected nearby clusters with large core radii. Taking advantage of the large collecting area of XMM-Newton ($\sim$6 times that of Chandra), we wish to obtain the best possible spectral and timing information for the widest possible sample of faint X-ray sources.

In this paper, we present the first results of our deep XMM-Newton observation of $\omega$ Cen. We describe the general properties of the population of faint X-ray sources detected in the cluster (Sect. 2). Using the publicly available Chandra observation, we have correlated the XMM-Newton and Chandra data (Sects. 3 and 4) to improve the XMM-Newton positions and search for intensity variations between the two data sets (Sect. 5). We also present the spectra of the brightest objects in the field, with the emphasis on those for which an identification already exists (Sect. 6). We briefly discuss the implications of our findings in Sect. 7.