6 Spectral analysis of the brightest objects

We have extracted spectra for the brightest sources; those with a total number of counts exceeding $\sim$100 in the most sensitive EPIC-PN camera. There are 26 sources satisfying this criterion. In this paper, we limit the spectral analysis to the 16 sources lying within twice the half mass radius.

To accumulate spectra, we chose an extraction radius of $\approx$0.7$^\prime$, except when another source was closer than 1.5$^\prime$ (two extraction radii). We extracted the background using an adjacent area of the same surface, at the same off axis angle on the same CCD. We generated ancillary response files and redistribution matrix files with the SAS tasks arfgen and rmfgen of the 5.3.3 release.

Whenever possible, we binned the spectra to contain at least twenty net counts in each bin, in order to use $\chi^2$ statistics. Otherwise we used the Cash statistics. For the spectra with the largest number of counts, we have left the interstellar column density as a free parameter of the fit. Note however that in all but one case, the fitted $N_{\rm H}$ is consistent within error bars with the value expected from the optical extinction in the direction of the cluster. We used XSPEC v11.1 (Arnaud 1996) to fit the spectra. The limited statistics does not allow us to use spectral models more sophisticated than thermal Bremsstrahlung, blackbody, and power law.

6.1 Sources within the core and the half mass radii

There are three sources in the core for which a spectrum can be extracted (sources 2, 5 and 24), two more between the core and half mass radii (sources 9 and 20) and the proposed quiescent neutron star binary which is just at the border of the half mass radius (source number 4).

Figure 7: The EPIC-PN unfolded spectra of the two CV candidates. These two sources are located within the core radius. These spectra are shown with a best thermal Bremsstrahlung fit. The spectra measured by XMM-Newton strengthen the CV classification for these two sources. These are to date the highest quality spectra of faint core globular cluster sources.

The three core sources are the two proposed CV candidates (Carson et al. 2000) and source 24 which was found to be variable within the XMM-Newton observation (see Sect. 5 and Fig. 5). The best fit spectral results are listed in Table 5. For the two CV candidates, the spectra can be accurately fitted with a thermal Bremsstrahlung (or alternatively with power laws of index $\sim$1.4). Such spectra are expected from such systems (Richman 1996). Thus our spectral observations reinforce the classification of these two objects as CVs. Their unfolded spectra are shown in Fig. 7. These are to date the highest quality spectra ever measured from faint globular cluster X-ray sources.

For the third variable source, its spectrum is also consistent with a power law, but given the limited statistics it could be also fitted with a thermal bremsstrahlung. Despite the source faintness, we have searched for spectral variations within the observation. Two spectra were extracted, one during its steady state and another one during the flaring state. As can be seen in Table 5, the two spectra are consistent. Best fit results for the two additional sources found between the core and half mass radii are also listed in Table 5.

6.2 The quiescent neutron star binary

Figure 8: The unfolded EPIC-MOS and EPIC-PN spectra of the quiescent neutron star binary candidate source 4. This spectrum is accurately fitted with a pure hydrogen neutron star atmosphere model. The filled squares, diamonds and circles refer to the MOS1, MOS2 and PN data respectively.

Rutledge et al. (2002) showed that the Chandra spectrum of the quiescent neutron star binary candidate can be fitted with a pure hydrogen neutron star atmosphere model (Pavlov et al. 1992; Zavlin et al. 1996). This source is also clearly detected in our observation, as the fourth brightest source in the field of view. Taking advantage of the better statistics of the XMM-Newton spectrum, we have also fitted its spectrum with the same neutron star atmosphere model. The parameters of the latter model are the temperature, the radius, the mass of the neutron star and its distance. The temperature and the radius were derived as measured by an observer at infinity. The best fit result with the mass of the neutron star and the source distance frozen are listed in Table 6.

 

 

Table 6: Spectral fit parameters of the quiescent neutron star binary candidate source 4, using the spectra of the EPIC-PN and EPIC-MOS cameras. The model used was a pure hydrogen neutron star atmosphere model (Pavlov et al. 1992; Zavlin et al. 1996). Parameters between parenthesis were frozen during the fit. The errors are also given at the 90% confidence level. A mass of 1.4 $M_{\odot }$ was assumed for the neutron star. The luminosities are given in units of 10³² ergs s^-1. Parameters obtained by Rutledge et al. (2002) are also listed.

Radius	Temp.	Distance	$N_{\rm H}$	$\chi^{\scriptscriptstyle 2}_{\scriptscriptstyle \nu}$	0.1-5.0 keV	Reference
$R_{\infty}$ (km)	$T_{\rm eff,\infty}$ (eV)	(kpc)	(1020 cm-2)		luminosity
14.3 $\pm$ 2.1	$66^{\scriptscriptstyle +4}_{\scriptscriptstyle -5}$	(5)	(9)	...	$5 \pm 2$	Rutledge et al. (2002)
13.6 $\pm$ 0.3	$67^{\scriptscriptstyle +2}_{\scriptscriptstyle -2}$	(5.3)	9.0 $\pm$ 2.5	1.00	3.2 $\pm$ 0.2	This work

There is a remarkable consistency between the results reported in Rutledge et al. (2002) and ours, but thanks to the improved statistics the error bars on the fitted parameters are much smaller with the present data. We did not find any evidence for the presence of a power law tail. Assuming a power law with photon index of 2, an upper limit of 10% of the total flux (90% confidence limit) can be derived for such a power law tail. The 0.1-5.0 keV bolometric luminosity measured by XMM-Newton is consistent within error bars with the Chandra value. The unfolded combined EPIC-PN and MOS spectrum is presented in Fig. 8. This is one of the best spectra of a quiescent neutron star binary obtained so far. Our observation thus strengthens the quiescent neutron star binary hypothesis for this object.

6.3 Notes on remaining sources within the field of view

In Fig. 9, we show the unfolded spectra of four more sources whose positions are between one and two half mass radii. All spectra are relatively hard, corresponding to colors in Fig. 3 similar to the colors measured from the CVs. Their luminosities, just around $\sim$10³² are also consistent with the CV hypothesis. Note however that some of them may also be background sources.

Figure 9: The unfolded spectra of four of the brightest objects in the EPIC-PN field of view. These sources are located outside the half mass radius, but within twice this radius. The spectrum of source 6 is shown with a thermal Bremsstrahlung fit, whereas for the others their spectra are shown with a power law fit. The four spectra are relatively hard and are consistent with those observed from the two core CVs, both in shapes and luminosities.

Figure 10: The unfolded spectrum of the brightest object within the field of view. This source is very absorbed. Its spectrum is presented with a power law fit. It is presently unidentified, but clearly deserves some follow-up investigations.

In addition, there are two other sources in the field of view which deserve some attention. The first one is the brightest object (it was detected by both Chandra and ROSAT, see Table 3). Its unfolded spectrum together with its best power law fit is presented in Fig. 10. The high $N_{\rm H}$ derived from the fit and its relatively large angular distance from the cluster center (about 8.5$^\prime$) calls into question its membership of $\omega$ Cen. However, its unusual properties (time variability and hard spectrum) make it an interesting target for follow-up investigations. There are no counterparts listed in the USNO A2.0 catalog within 2'' of the Chandra position.

The second object is the third brightest object in the field (source 3). It has an M dwarf counterpart (USNO-A2 0375-18249604, Cool et al. 1995). Its spectrum is well fitted by a 2 temperature Raymond-Smith model (2T) expected from such a system (Singh et al. 1996). The best fit parameters of all these objects are also given in Table 5.