3 Analysis of the X-ray sources

Figures 1-3 show DSS images superimposed on the smoothed X-ray contour plots (data from EPIC-MOS2 camera). The D₂₅ ellipse is also shown for comparison. Here and in the following, if not explicitly indicated, the data on the host galaxies are obtained from the catalog of Ho et al. (1997a).

The X-ray properties of the ULX are given in Table 3. For eight of these sources the statistics are good enough (at least 80 counts) to perform a spectral fitting with simple models. To extract the spectrum we selected source regions with radius 20'' - 30'', depending on the presence or absence of nearby sources. In all cases, the flux was corrected according to the encircled energy fraction (Ghizzardi 2001).

Data from MOS1, MOS2, and PN were fitted simultaneously. We used the following models: power law (PL), black body (BB), thermal bremstrahlung (BR), unsaturated Comptonization (CST) by Sunyaev & Titarchuk (1980), and the multicolor black body accretion disk (MCD) by Mitsuda et al. (1984). The latter two correspond to the compST and diskbb models in xspec. The unsaturated Comptonization model has two free parameters, the temperature and the optical depth. The multicolor disk has one free parameter, i.e. the temperature at the inner disk radius.

In Table 3 are summarized the results for the power-law model, while in the following we discuss other cases. We consider acceptable only the fits with reduced $\chi^2$ less than 2.

3.1 NGC 1058

Two ULX have been detected and both of them appear to have no evident optical counterpart in the DSS. ULX1 has sufficient counts to be fitted. The best model is the power law, even though with an unusual low $\Gamma = 1.1\pm 0.3$ ( $\chi^2=41.1$, $\nu=33$). The 0.5-10 keV flux is calculated as $8.5\times 10^{-14}$ erg cm^-2 s^-1, which corresponds to a luminosity of $1.1\times 10^{39}$ erg s^-1. Other models give worse fits; for example, the black body model with $kT=0.7\pm 0.2$ keV gives $\chi^2=53.0$ for $\nu=33$, and we get an upper limit for the temperature of the inner disk in the MCD model of kT<4 keV ( $\chi^2=43.8$, $\nu=33$).

3.2 NGC 3185

This source is the farthest in our sample, and inside its small angular size ( D₂₅=2.34'), we find one clear ULX, for which the low statistics do not allow a spectral fit. No clear optical counterpart is visible in the DSS.

3.3 NGC 3486

One ULX has been detected at 23'' from the optical centre. The statistics are sufficient to perform a spectral fitting. The best-fit model is obtained with a simple power law model with a photon index of $2.2\pm 0.5$ ( $\chi^2=7.7$, $\nu=14$). The 0.5-10 keV flux is $8.3\times 10^{-14}$ erg cm^-2 s^-1, which corresponds to a luminosity of $5\times 10^{38}$ erg s^-1. The fits are worse, but still acceptable, with the BB ( $\chi^2=13.3$, $\nu=14$) and MCD ( $\chi^2=11.9$, $\nu=14$) models. The former gives a temperature of $kT=0.26\pm 0.05$ keV, while the latter gives a temperature of the inner disk of $kT=0.4\pm 0.2$ keV.

After the analysis of 41.3 ks of ASCA/SIS0 data, Pappa et al. (2001) found, for this galaxy, an observed flux of $5\times 10^{-14}$ erg cm^-2 s^-1 (in the energy band 0.8-10 keV), using a power law with $\Gamma=1.9$ absorbed by an additional column density of $3.2\times 10^{21}$ cm^-2. They suggested that NGC 3486 may be an obscured Seyfert 2 galaxy. However, from XMM-Newton data, in a circle of 10'' centered in the optical centre of NGC 3486, the count rate is $(2.1\pm 0.7)\times 10^{-3}$ s^-1, which corresponds to a flux of $(9\pm 3)\times 10^{-15}$ erg cm^-2 s^-1 (given the conversion factor used in this paper). The luminosity is $(6\pm 2)\times 10^{37}$ erg s^-1 in the 0.5-10 keV energy band. These data are compatible with those of Chandra (Ho et al. 2001) and the upper limit of ROSAT/HRI ( $6.1\times 10^{-14}$ erg cm^-2 s^-1 in the 0.1-2.4 keV band; Halderson et al. 2001).

Since the ASCA flux was calculated by extracting photons in wider regions (1.5'), because of the low angular resolution, this suggest the source observed by ASCA was not the nucleus, but the ULX, at only 23'' ($\sim$825 pc) from the optical centre of NGC 3486.

3.4 NGC 3941

Only one ULX has been detected in this galaxy at 38'' from the centre. The spectrum is well fitted with a PL model with $\Gamma=1.9\pm 0.2$ ( $\chi^2=17.8$, $\nu=19$) or a BR model with $kT=4\pm 2$ keV ( $\chi^2=24.4$, $\nu=19$). In the first case, the flux is $1.7\times 10^{-13}$ erg cm^-2 s^-1 and the luminosity is $7.4\times 10^{39}$ erg s^-1.

3.5 NGC 4138

No ULX was detected inside the D₂₅ ellipse (figure not shown). The flux limit is about 10^-14 erg cm^-2 s^-1 for this observation.

3.6 NGC 4168

The XMM-Newton observation shows one clear ULX in this galaxy, which appears to have a possible correlation with a point source in the 2MASS survey. No spectral fitting was possible.

3.7 NGC 4501

We clearly detected two ULX in this galaxy, one of them just on the border of the D₂₅ ellipse. ULX2 has sufficient counts for a spectral fitting. The best fit ( $\chi^2=26.3$, $\nu=26$) is a power law with $\Gamma = 2.3\pm 0.4$, which corresponds to a flux of $1.0\times 10^{-13}$ erg cm^-2 s^-1 and a luminosity of $3.7\times 10^{39}$ erg s^-1. Other acceptable models are BB ( $\chi^2=40.9$, $\nu=26$) with $kT=0.28\pm 0.09$ keV and MCD ( $\chi^2=35.6$, $\nu=26$) with $kT=0.5\pm 0.3$ keV. For BR we obtain only an upper limit for kT<4 keV ( $\chi^2=30.2$, $\nu=26$).

3.8 NGC 4565

We find seven sources inside the D₂₅ ellipse, but since this galaxy is edge-on, there could be additional sources projected along the minor axis above the plane of the disk. Examination of Fig. 3 suggests that this indeed might be the case, although interestingly none of the sources seem to coincide with known globular clusters (Kissler-Patig et al. 1999).

For three ULX (2, 4, and 6) it is possible to perform spectral fitting. Specifically, ULX4 has the highest counts in all the present catalog. The best fit is still the power law with photon index $1.9\pm 0.1$( $\chi^2=43.7$, $\nu=46$). The flux and the corresponding luminosity are $2.2\times 10^{-13}$ erg cm^-2 s^-1 and $2.5\times 10^{39}$ erg s^-1, respectively.

It is worth noting that, with the exception of the BB model, all of the other models give an acceptable spectral fit for ULX4. In addition, it is also the only source that is fitted well with the unsaturated Comptonization model. Specifically, the BR model is fitted with $kT=3.4\pm 0.8$ keV ( $\chi^2=48.4$, $\nu=46$), the MCD with $kT=0.8\pm 0.1$ keV ( $\chi^2=83.1$, $\nu=46$), and the CST has $kT=1.8\pm 0.6$ keV with optical depth $\tau_{\rm e} = 16\pm 7$ ( $\chi^2=42.8$, $\nu=45$).

ULX2 is best fitted with the black body model at $kT=0.9\pm 0.1$ keV ( $\chi^2=17.7$, $\nu=13$). The flux and luminosity for this model are $7.5\times 10^{-14}$ erg cm^-2 s^-1 and $8\times 10^{38}$ erg s^-1, respectively. Other reasonable fits are obtained with PL (see Table 3) and with MCD, with the temperature of the inner disk of $2\pm 1$ keV ( $\chi^2=20.8$, $\nu=13$).

ULX6 also is best fitted with BB, but with $kT=0.52\pm 0.07$ keV ( $\chi^2=13.5$, $\nu=10$). In this case the flux is $3.9\times 10^{-14}$ erg cm^-2 s^-1 and the luminosity is $4\times 10^{38}$ erg s^-1. Yet other statistically acceptable models are the PL (see Table 3) and the MCD, with temperature of the inner disk of $1.0\pm 0.3$ keV ( $\chi^2=15.1$, $\nu=10$).

Four of the ULX sources have already been detected by ROSAT (Vogler et al. 1996), namely RXJ1236.2+2600 (ULX2), RXJ1236.2+2558 (ULX4), RXJ1236.3+2600 (ULX5), RXJ1236.4+2557 (ULX6). For ULX4, ROSAT observations suggest an additional intrinsic absorption, with values from 3.4 to $3.6\times 10^{20}$ cm^-2, in addition to the Galactic column density ( $1.3\times 10^{20}$ cm^-2). Instead, the best fit from ASCA data gives an upper limit of $2\times 10^{20}$ cm^-2 (Mizuno et al. 1999). In our sample, only ULX2 indicates a possible extra absorption in addition to the Galactic $N_{\rm H}$, with 95% significance (cf. Table 3). Additional absorption in the fitting of ULX4 and ULX6 has low significance (68% and 82%, respectively). The values of $N_{\rm H}$ are < $1\times 10^{21}$ cm^-2 and $2\times 10^{21}$ cm^-2, respectively.

It is useful to note that Mizuno et al. (1999) suggest, on the basis of ASCA data, that this galaxy has no X-ray nucleus and that the twin bright sources in the middle of the galaxy could be two ULX. XMM-Newton has sufficient angular resolution to separate the two sources, one of them being ULX4; this is consistent with the ROSAT results. We identify the other source with the active nucleus.

3.9 NGC 4639

We find two ULX, but one of them (ULX2) was detected with the MOS, but not with the PN, because its position fell into a gap between the PN chips. In both cases, the statistics were not sufficient to perform spectral fitting.

3.10 NGC 4698

We find one ULX with sufficient statistics to perform spectral fitting. The best fit is a power law with $\Gamma = 2.0\pm 0.2$ ( $\chi^2=18.3$, $\nu=24$) that gives a flux of $8.6\times 10^{-14}$ erg cm^-2 s^-1 and a luminosity of $3\times 10^{39}$ erg s^-1.

Other statistically acceptable models are BB with $kT=0.35\pm 0.05$ keV ( $\chi^2=34.5$, $\nu=24$), BR with $kT=2\pm 1$ keV ( $\chi^2=19.0$, $\nu=24$), and MCD with $kT=0.6\pm 0.1$ keV ( $\chi^2=23.7$, $\nu=24$).

Figure 4: Examples of spectra. Combined spectrum (MOS1, MOS2, PN) of the source NGC 4565-ULX4 ( left) and the source NGC 4698-ULX1 ( right). Both were fitted with a single power-law model. Refer to Table 3 for more details. The lower window of the two panels shows the ratio between the data and the model.