2 XMM-Newton data analysis

The present ULX catalog is based on data from the European Photon Imaging Camera (EPIC) on board the XMM-Newton satellite. EPIC comprises three instruments: the PN-CCD camera (Strüder et al. 2001) and two MOS-CCD detectors (Turner et al. 2001).

To define a ULX source as an off-nuclear galactic object, we use three selection criteria:

• the object luminosity, in the energy band 0.5-10 keV, must be greater than $2\times 10^{38}$ erg s^-1;

• the object has to be located inside the D₂₅ ellipse (i.e., the dimension equal to 25 mag/arcsec²) of the host galaxy;

• the object must be sufficiently far from the optical centre, to avoid confusion with the galaxy's active nucleus.

Considering that the absolute location accuracy for XMM-Newton is about 4'' (Jansen et al. 2001), and that the uncertainty in the optical position of the host galaxy centre, from the Digitized Sky Survey (DSS, see Cotton et al. 1999) is less than 2.7'', we searched for off-nuclear sources at least 10'' away from the optical centre of the host galaxy. The characteristics of the sources in the present catalog are summarized in Table 2.

For the processing, screening, and analysis of the data we used the standard tools of XMM-SAS software v. 5.2 and HEAsoft Xspec (11.0.1). The images are prepared with DS9 v. 2.1, together with ZHTools v. 2.0.2.

Soft-proton flares affected the observations randomly, and in some cases it is necessary to filter the available data. Time intervals contaminated by flares have been excluded by extracting the background lightcurve in the 10-13 keV energy band. Periods with count rates higher than 0.2 s^-1 have been removed.

Figure 1: Images from Digitized Sky Survey (DSS) with XMM-Newton data superimposed. X-ray contours are calculated from MOS images in the 0.5-10 keV energy band. North is up and East to the left. The D25 ellipse is also shown for comparison.

To perform the detection, the EPIC-PN was selected, because of its larger effective area with respect to the MOS cameras, which allows more accuracy in the detection of faint X-ray sources. The detections were performed using the sliding box cell detection algorithm (eboxdetect of XMM-SAS). It uses a box, with dimension $5\times 5$ pixel as the detection cell, and $L=-\ln(P)=10$ as the minimum detection likelihood value, which in turn corresponds to a probability of Poissonian random fluctuations of the counts in the detection cell of $P=4.5 \times 10^{-5}$ (roughly $4\sigma$).

After the automatic procedure, each source inside the D₂₅ ellipse was carefully checked to exclude false or doubtful sources. Specifically, we noted that eboxdetect fails in some cases. Indeed, the software accumulates the source counts in a $5\times 5$ pixel box, while the background counts are accumulated in the region of $(9\times 9) {-} (5\times 5)$ pixels. This algorithm gives good results with uniform regions (both diffuse or with background only), but fails in border regions. By comparing visually the detections from EPIC-PN with data from the MOS cameras, it is possible to identify and exclude possible artifacts. It is worth noting that in the case of fake detection, eboxdetect gives an unusually poor point source location accuracy (PSLA) of 3-4'', to be compared with a PSLA of less than 0.1''for real and normal detections. (Note that the global positional uncertainty is given by the sum of the satellite pointing uncertainty of 4'' and the PSLA.)

After the detection run, we then extracted from the available list only those sources with X-ray luminosities higher than $2\times 10^{38}$ erg s^-1 in the energy band 0.5-10 keV (see Table 2).

The count rate calculated by eboxdetect has been converted into flux using a conversion factor of $3\times 10^{11}$ cnt cm²/erg. This has been calculated from the graphics available in the XMM-Newton User's Handbook (Ehle et al. 2001) and by assuming a power law with photon index 2 and an absorbing column density $N_{{\rm H}}=3.0\times 10^{20}$ cm^-2. This choice of the model parameters is consistent with results obtained from the spectral analysis of the brightest sources (see the next section).

We do not apply the correction for vignetting, because all the sources for which it is possible to extract the spectrum are close to the centre of the field of view (less than 2') and have most of their statistics below 5 keV (cf. Lumb 2002).

Figure 2: Images from Digitized Sky Survey (DSS) with XMM-Newton data superimposed. X-ray contours are calculated from MOS images in the 0.5-10 keV energy band. North is up and East to the left. The D25 ellipse is also shown for comparison.

Figure 3: Image from Digitized Sky Survey (DSS) with XMM-Newton data superimposed. X-ray contours are calculated from MOS images in the 0.5-10 keV energy band. North is up and East to the left. The D25 ellipse is also shown for comparison.