3 Cross-correlation with a previous Chandra observation

The mean statistical error on the positions of the sources detected by XMM-Newton is of the order of $\sim$4 $^{\prime\prime}$ (see Tables 1 and 2). To get the final positional error, one must add quadratically, the systematic error on the pointing direction of the XMM-Newton satellite, which is about 4 $^{\prime\prime}$ (Jansen et al. 2001). This means that on average the position error will be around 6 $^{\prime\prime}$.

Figure 4: The Chandra field of view (delimited by the square) overlaid with the XMM-Newton image. The blue open circles denote XMM-Newton sources and the red points indicate the Chandra sources. The sources detected both by XMM-Newton and Chandra appear as red points surrounded by a blue circle. The core and half mass radii are indicated with a solid and dashed line circle respectively.

Figure 5: The contour images of the extended emission. The image on the left provides an overview of the cluster region. The two large circles indicate the core radius (solid line circle) and the half mass radius (dashed line circle). The Chandra sources are indicated by crosses. The right image provides a zoom on the extended emission. The XMM-Newton sources are indicated by filled squares if the source is also detected by Chandra, or a filled circle otherwise. For both these images, the contour levels are 4.5, 5, 6, 7 and 8 $\sigma$ levels.

However, this can be improved by using the most accurate positions provided by Chandra. For this purpose, we have analyzed the publicly available Chandra observation. The Chandra observation took place on 2000 January 24-25, in imaging mode using the ACIS-I detector placed in the focus of the telescope. The observation was $\approx$70 kiloseconds long. These data have already been presented by Cool et al. (2002) and Rutledge et al. (2002).

We retrieved the data from the archives and used CIAO version 2.2.1 and CALDB version 2.12 to calibrate the event files (using the CIAO task acis_process_events). During the Chandra observation, the background was low and stable. We then filtered the events file for non astrophysical events using the ASCA grades 0, 2, 3, 4, 6, and for good time intervals using the provided GTI. We also rejected events with energies below 0.2 keV. We made a basic detection scheme, using the CIAO task wavdetect. As recommended in the CIAO detect manual, we used a conservative spurious detection threshold of 10^-5 as input to wavdetect. As for XMM-Newton, sources were searched in the 0.5-5.0 keV range.

We found 129 sources within the ACIS-I field of view, to be compared to 146 in Cool et al. (2002) (Rutledge et al. 2002, used a sliding box algorithm and a very conservative detection threshold and reported 40 sources only). Even if we take into account a 1 $^{\prime\prime}$ systematic error in the attitude reconstruction of Chandra, the mean error on the source positions remains very small, typically $\sim$1.5 $^{\prime\prime}$.

The Chandra positions were then used to compute the astrometric correction for the XMM-Newton observation. We have selected three XMM-Newton sources (sources 2, 8, and 15, see Tables 1 and 2), properly spaced within the field of view, far away from CCD gaps and bad columns, and among the brightest sources (i.e. with a small statistical error on their position). These sources are also clearly detected by Chandra. The astrometric correction was then computed with the three reference positions provided by Chandra, using the starast Interactive Data Language tools of the astrolib library. The positions listed in Tables 1 and 2 reflect this correction.

With this correction applied, 11 and 27 sources are detected by XMM-Newton within the core and half mass radii. For comparison, Chandra detected 22 and 46 sources within the same regions.

This correction further allows us to get rid of the systematic error and to cross-correlate the Chandra and XMM-Newton source positions. The positions of the Chandra sources found within the statistical error box of the XMM-Newton sources are given in Table 3. Sixty three XMM-Newton sources have a Chandra counterpart. We present in Fig. 4 the Chandra field of view and sources overlaid with the XMM-Newton image.

It is beyond the scope of this paper to investigate the error box content of each XMM-Newton source, and we have focussed the present paper on sources for which a previous identification has been reported. Two EINSTEIN sources were detected by both XMM-Newton and Chandra (sources B and C), and due to its larger field of view, XMM-Newton also detected the EINSTEIN sources A and D (these ones were missed by Chandra). EINSTEIN sources A and D (XMM-Newton sources 3 and 7) are associated with foreground M dwarfs (Cool et al. 1995). The source EINSTEIN C was resolved into sources R9a and R9b in ROSAT (Verbunt & Johnston 2000). These two core sources which are detected as XMM-Newton sources 2 and 5 are the two CV candidates (Carson et al. 2000; Cool et al. 2002). The ROSAT source R20 (detected by Chandra and proposed to be a BY Dra system, Cool et al. 2002) is not detected by XMM-Newton. The proposed quiescent neutron star binary is detected as source 4 by XMM-Newton. Finally, the source identified by Verbunt & Johnston (2000) as HD116789 is detected as source 28 (see Table 3).

We have estimated the limiting count rate of the XMM-Newton observation as the count rate needed for a detection of a source placed at a mean off-axis angle of 7.5$^\prime$ which is half the radius of the EPIC-PN field of view. In the 0.5-5.0 keV band, the limiting count rate is $1.4 \times 10^{-3}$ counts s^-1. For comparison the limiting count rate of a source on-axis is $1.0 \times 10^{-3}$ counts s^-1. For Chandra, the limiting count rate estimated with the same method is $1.5 \times 10^{-4}$ counts s^-1 for an on-axis source and $1.6 \times 10^{-4}$ counts s^-1 for a source at 7.5$^\prime$.

The limiting count rates have been converted into limiting fluxes using two spectral models; a blackbody of 0.6 keV and a thermal Bremsstrahlung of 3 keV. This gives $3.8\times10^{-15}$ ergs s^-1 cm^-2 and $3.5\times10^{-15}$ ergs s^-1 cm^-2 respectively for XMM-Newton and $1.2\times 10^{-15}$ ergs s^-1 cm^-2 and $1.3\times10^{-15}$ ergs s^-1 cm^-2 for Chandra. At the distance of 5.3 kpc, these fluxes translate to 0.5-5.0 keV bolometric luminosities of $\sim$ $1.3 \times 10^{31}$ ergs s^-1 and $\sim$ $4.2\times 10^{30}$ ergs s^-1 for XMM-Newton and Chandra respectively for the blackbody model.

As said above XMM-Newton detected 146 sources, and 63 of them have a Chandra counterpart. Of the 83 remaining XMM-Newton sources, 55 were outside the field of view of Chandra. Of the remaining 28, 2 were missed by ACIS-I because of CCD gaps and noisy columns (one is in the core). This leaves a total of 26 sources detected by XMM-Newton and not detected by Chandra. We have reprocessed the Chandra data with a less conservative spurious detection threshold (10^-4) to search for fainter objects. The above number decreases from 26 to 23. Since the Chandra observations were more sensitive than the XMM-Newton one, one needs to investigate why 26 sources seen by XMM-Newton were not detected by Chandra. Part of the discrepancy resides in the presence of a region of extended emission.