ULXs are also named super-Eddington sources (see Fabbiano 1989, 1995), super-luminous sources, and intermediate luminosity X-ray objects (IXOs) (Roberts & Warwick 2000; Colbert & Mushotzky 1999; Colbert & Ptak 2002). All these names aim to convey the fact that they are extremely luminous X-ray sources, emitting well in excess of the Eddington luminosity of a spherically accreting and emitting neutron star (~ 2 × 1038 ergs s-1). Usually, sources emitting at ~ 1039 ergs s-1 or above are included in this category. If these sources are emitting isotropically at the Eddington limit, masses in excess of those expected from stellar black holes are implied, up to in some cases, 100M (e.g. Fabbiano 1989, 1995; Makishima et al. 2000). Colbert & Mushotzky (1999) dubbed this type of black holes `intermediate mass black holes' (IMBH), to distinguish them from the stellar mass black holes found in Galactic black hole binaries, and also from the supermassive 107 -109 M found at the nuclei of galaxies that are responsible for AGNs.
3.1. Spectra and spectral variability
Although young supernova remnants may be responsible for ULX emission in some cases (e.g Fabian & Terlevich 1996), there is now sufficient evidence from spectral and variability data, to establish that the majority ULXs are indeed compact systems, most likely accreting binaries. ASCA X-ray spectra suggested accretion disk emission. These spectra, however, also require temperatures much larger than those expected from black holes of the mass implied by the luminosities of these sources, leading to the suggestion of rotating Kerr black holes (Makishima et al. 2000; Mizuno, Kubota & Makishima 2001). In The Antennae ULXs the Chandra spectra (Zezas et al. 2002a, b) tend to be hard, and their average co-added spectrum requires both a power law ( ~ 1.2) and a disk-blackbody component consistent with the ASCA results, with kT ~ 1.1 keV. A XMM-Newton survey of 10 galaxies reports ULX spectra consistent with black hole binaries in either high or low state (Foschini et al. 2002), but the data quality is too poor for detailed modelling. Similar general spectral results can be found in a Chandra survey of ULXs in different galaxies (Humphrey et al. 2003). Instead, XMM-Newton high quality spectra of two ULXs in NGC 1313 (X-1 and X-2) led to highly significant detections of soft accretion disk components, with temperatures of kT ~ 150 eV, consistent with accretion disks of IMBHs (Miller et al. 2003a; Fig. 8).
Figure 8. Bottom: the XMM-Newton image of NGC 1313, showing the position of the two ULXs. Top: X-ray spectrum of ULX-1, compared with best-fit model requiring a cool accretion disk component (Miller et al. 2003).
The XRB hypothesis is reinforced by observations of correlated luminosity-spectral variability similar to the `high/soft-low/hard' behavior of Cyg X-1 (e.g., in M81 X-9, La Parola et al. 2001, with a variety of X-ray telescopes, Fig. 9; and in two ULXs in IC 342, Kubota et al. 2001 with ASCA). However, more recently, Kubota, Done & Makishima (2002) argue that these power-law ULX spectra should not be identified with the low/hard state, but rather may be due to a strongly Comptonized optically thick accretion disk, analogous to the Comptonization-dominated `very high/anomalous state' in Galactic black-hole binaries. ASCA observations of one of the IC 342 sources in high state (disk-dominated) revealed a `high/hard-low/soft' low-level variability, with a possible 30-40 hr periodicity, as could be produced by a massive main sequence star orbiting a black hole (Sugiho et al. 2001).
With Chandra and XMM-Newton an increasing number of ULXs are being discovered and studied in galaxies. Variability in the Chandra observations of M82 established that the ULXs in this galaxy are likely to be accreting compact objects (Matsumoto et al. 2001). The Chandra observations of NGC 3628 (Strickland et al. 2001) show the re-appearance of the 1040 ergs s-1 variable ULX first discovered with ROSAT (Dahlem, Heckman & Fabbiano 1995). A new transient ULX was discovered in M74 (NGC 628) with (Soria & Kong 2002). Chandra observations of MF 16 in NGC 6946, formerly identified as an extremely luminous supernova remnant (Schlegel 1994), reveal instead a point-like source with the typical X-ray spectrum of a black-hole binary (Holt et al. 2003; Roberts & Colbert 2003). Similarly, M81 X-6, which is positionally coincident with a supernova remnant, is identified as a XRB by its X-ray spectrum (Swartz et al. 2003). Chandra observations of the nucleus of M33 have revealed a two-component (power-law and disk) spectrum and have established luminosity-spectral variability patterns in this ULX, reminiscent of the black hole binary LMC X-3 (La Parola et al. 2003; see also Long, Charles & Dubus 2002); Dubus & Rutledge (2002) compare this source with the Galactic microquasar GRS 1915+105.
High/hard-low/soft variability was found in M51 X-7, together with a possible 2.1 hr period (but the time coverage is scant) by Liu et al. (2002). Both Cyg X-1 like high/soft-low/hard as well as high/hard-low/soft variability was detected in the population of nine ULXs discovered with Chandra in the Antennae galaxies (Fabbiano ; Fig. 10). The latter type of variability can also be found in a few Galactic XRBs (1E 1740.7-2942, GRS 1758-258, GX 339-4, Smith et al. 2002; see also the XMM-Newton results on GRS 1758-258, Miller et al. 2002). This spectral variability may be indicative of the competition between the relative dominance of the accretion disk versus the innermost hot accretion flow; several scenarios for spectral variability are discussed in Fabbiano et al. 2003a and references therein.