5 Overall view and discussion

We have analyzed the X-ray data from XMM-Newton observations of 10 nearby Seyfert galaxies. The host galaxies are located between 7.4 and 21.3 Mpc, with 7/10 between 16.8 and 21.3 Mpc. Only one host galaxy (NGC 4168) is elliptical (E2), while all the remaining are spirals of various types (see Table 1).

We found ULX in 9 of the 10 galaxies. The contamination with background sources is very low. Indeed, from XMM-Newton observations of the Lockman Hole, Hasinger et al. (2001) found about 100 sources per square degree with flux higher than 10^-14 erg cm^-2 s^-1 in the energy band 0.5-2 keV, and about 200 sources per square degree in the energy band 2-10 keV. Assuming the same $\log N {-} \log S$ and considering the flux limit reached by our observations, we expect to find, in the worst case (NGC 4565), fewer than one (0.7) background source inside the D₂₅ ellipse. For all remaining host galaxies, the expected number of background objects is significantly less than one (0.2 for most of the cases). Therefore, we expect that, in the worst case, the overall sample contains fewer than 2 background objects.

The total number of ULX in the present catalog is 18. The mean value is 1.8 ULX per galaxy, with the maximum value in NGC 4565 with 7 ULX and the minimum in NGC 4138 with no detection. By omitting NGC 4565, we have a mean value of about 1.2 ULX per galaxy. With respect to the ROSAT/HRI survey by Roberts & Warwick (2000), we find that XMM-Newton allows a significative improvement in the number of ULX detections (see Table 4). In Table 4 we list also the luminosities in B and far-infrared bands of the host galaxies, the latter being a rough indicator of the star formation activity. At a first look, no evident correlation appears, but the present sample is very small. Although it appears that host galaxies with high  $L_{{\rm B}}$ and  $L_{{\rm FIR}}$ (NGC 4501 and NGC 4565) have a higher number of ULX, we caution that this effect could be spurious because our survey does not reach a uniform luminosity threshold to detected ULX. NGC 4565, for example, reaches a flux limit of $\sim$ $1\times 10^{-14}$ erg cm^-2 s^-1, which is deeper than the flux corresponding to the ULX luminosity limit of $2\times 10^{38}$ erg s^-1 (cf. Sect. 2). For NGC 4501, on the other hand, the flux limit is too shallow to reach this luminosity limit.

 

 

Table 4: Predicted and observed number of ULX. Columns: (1) Name of the host galaxy; (2) total absolute B magnitude, from Ho et al. (1997a); (3) luminosity (10¹⁰ $L_{\odot }$ erg s^-1) calculated from data in Col. 2; (4) expected number of ULX according to Roberts & Warwick (2000), who found a relationship between the number of ULX and $L_{{\rm B}}$; (5) number of ULX actually found in the present survey; (6) luminosity in the far-infrared (FIR, 42.5-122.5 $\mu$m) in units (10⁴² erg s^-1), calculated from data of Ho et al. (1997a).

Galaxy	$M_{{\rm B}}$	$L_{{\rm B}}$	Expected	Found	$L_{{\rm FIR}}$
(1)	(2)	(3)	(4)	(5)	(6)
NGC 1058	-18.25	0.15	0.1	2	1.9
NGC 3185	-18.99	0.30	0.2	1	5.1
NGC 3486	-18.58	0.21	0.1	1	2.7
NGC 3941	-20.13	0.87	0.6	1	5.3 $^{{\rm *}}$
NGC 4138	-19.05	0.32	0.2	0	2.0 $^{{\rm *}}$
NGC 4168	-19.07	0.32	0.2	1	0.37
NGC 4501	-21.27	2.5	1.7	2	48
NGC 4565	-20.83	1.7	1.2	7	10
NGC 4639	-19.28	0.40	0.3	2	4.1
NGC 4698	-19.98	0.70	0.5	1	1.5

$^{{\rm *}}$ Since no FIR data were available in Ho et al. (1997a), we calculate
the FIR luminosity according to the relationship $\log L_{{\rm FIR}}/L_{{\rm B}}=-0.792$
(Pogge & Eskridge 1993).

The luminosities observed are in the range $(2 {-} 74)\times 10^{38}$ erg s^-1, depending on the model considered. If we make the simplistic assumptions that the accretion is uniform and spherical, that the bolometric luminosity approximately equals the X-ray luminosity, and that the Eddington ratio is 1, these luminosities correspond to compact objects with masses between 1.5 and 57 $M_{\odot}$. However, the X-ray luminosity is generally only $30{-}40\%$ of the bolometric luminosity of the accreting sources (e.g., Mizuno et al. 1999). In addition, if the Eddington ratio is in the range of 0.1-0.01, as suggested by observations of Galactic black hole candidates (e.g., Nowak 1995), the mass range would shift toward 10³-10⁴ $M_{\odot}$. Unless the sources are very young, such high masses are difficult to explain for off-centre sources, because dynamical friction would tend to drag the objects toward the centre in less than the Hubble time (cf. Binney & Tremaine 1987).

Therefore, as proposed by several authors, alternative scenarios must be considered. For example, Makishima et al. (2000) proposed a Kerr black hole scenario: in this case, the luminosity produced by a spinning black hole can be up to 7 times larger than in a Schwarzschild black hole. On the other hand, King et al. (2001) suggested that the matter could accrete anisotropically: an anisotropic factor of 0.1-0.01 reduces the values of the mass to those typically observed in X-ray binaries in the Milky Way. Something similar has been suggested by Begelman (2002): in this case, the presence of inhomogeneities in radiation pressure-dominated accretion disks, as a consequence of photon-bubble instability, would allow the radiation to escape. Finally, Körding et al. (2002) and Georganopoulos et al. (2002) suggested the possibility of relativistic beaming due to the presence of jets coupled to an accretion disk. Both are based on the microquasar model by Mirabel & Rodriguez (1999).

The statistics of the present observations do not allow us to discriminate clearly between the different models, but we can infer some useful hints from the eight sources, which gave sufficient counts for a spectral fitting. In 5/8 cases the best-fit model is obtained with a simple power law with $\Gamma \approx 1.9{-}2.3$ (for an example of spectra, see Fig. 4). One of these five sources (NGC 1058-ULX1) presents an almost flat spectrum ( $\Gamma = 1.1\pm 0.3$). For the remaining sources (2/8), we obtained a best fit with the black body model with $kT\approx 0.5{-}0.9$ keV.

It is known that the emission expected from a black hole X-ray binary is variable: in the hard state, the spectrum is typically a power law with $\Gamma \approx 1.3 {-}1.9$, while in the soft state the spectral index increases up to about 2.5 and a soft component appears in the X-ray spectrum (e.g., Ebisawa et al. 1996). Therefore, our sources could be black hole X-ray binaries in a hard or soft state. Terashima & Wilson (2002) proposed the existence of two different populations of ULX, one characterized by soft thermal and the other by non-thermal X-ray emission. A possible key to distinguish between the available hypotheses can be to perform time variability studies, but current statistics are too low for such a study.

It is interesting to note that the MCD model, which has often been successful in the past for ULX (e.g., Colbert & Mushotzky 1999; Makishima et al. 2000) is never the best fit in our data. Even when we obtain a reasonable fit with MCD, a simple black body model is statistically better. This may be due to the low photon counts of the present spectra.

The unsaturated Comptonization (CST) model does not give acceptable fits, except for ULX4 in NGC 4565, for which it represents the second best fit, after the power law.

Our understanding of the nature of ULX is limited by the fact that, to date, counterparts at other wavelengths are quite rare (Roberts et al. 2001; Pakull & Mirioni 2002; Wang 2002). In our sample, we note that one source (NGC 4698-ULX1) is detected in the radio (6 cm). NGC 4168-ULX1 and NGC 4639-ULX2 both show a highly probable optical counterpart. In the second case, it is identified as a H II region, which is a type of counterpart frequently associated with ULX (Pakull & Mirioni 2002). It is worth noting the case of NGC 4656-ULX3 could be obscured by, rather than correlated with, a dust cloud. Finally, the probable counterparts of NGC 4168-ULX1 and NGC 4698-ULX1 appear to be considerably red objects, with B-R > 2 mag.