.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

Part of the measured thermal emission may also come from OB stars, which are expected to be abundant in SB regions and whose thermal emission is characterized by $kT \sim 0.2$ keV, $L \sim 10^{33}$ erg s^-1 (see PR02).

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...2003)

The two quoted limits correspond to Eddington luminosities for, respectively, a $1.5 ~M_\odot$ neutron star and a $8~ M_\odot$ black hole (BH) which is the limiting BH mass obtainable via ordinary stellar evolution. ULXs are called "super-Eddington sources''.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... HMXBs

In a few cases the optical counterparts of ULX have been identified as O stars (Liu et al. 2002; Roberts et al. 2002), thus suggesting a relation between ULXs and HMXBs, which of course are young objects that trace the current SFR.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... BH

Qualitatively, the inverse correlation between L_2-10 and $\Gamma$ (i.e., flatter spectra for higher luminosities) can be understood within the standard accretion model devised to explain the high $L_{\rm x}$ of interacting binaries. The increase in $L_{\rm x}$ is driven by an increase of $\dot M$ , which in turn means a piling up of material around the emission region that leads to a higher Compton scattering optical depth, and hence to a flatter spectrum. E.g., according to Sunyaev & Titarchuk (1980) and Shapiro et al. (1976), if a source of photons is embedded in a plasma cloud of optical depth $\tau$ and temperature T, the escaping radiation has a high-energy photon index

$\begin{displaymath}\Gamma ~=~ \biggl[ {9 \over 4} ~+~ {\pi^2 m_{\rm e} c^2 \over 3 ~ (\tau + 2/3)^2 ~ kT} \biggr]^{1/2} ~-~ {1 \over 2} \cdot \end{displaymath}$

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.