Annu. Rev. Astron. Astrophys. 2006. 44: 323-366
Copyright © 2006 by . All rights reserved

Next Contents Previous


The most widely used observational definition of ULXs is that of sources detected in the X-ray observing band-pass with luminosities of at least 1039 erg s-1, implying bolometric luminosities clearly in excess of this limit. ULXs (also named intermediate luminosity X-ray objects - IXOs; Colbert & Ptak 2002) were first detected with Einstein (Long & Van Speybroeck 1983; see the review by Fabbiano 1989). These sources were dubbed super-Eddington sources, because their luminosity was significantly in excess of the Eddington limit of a neutron star (~ 2 × 1038 erg s-1), suggesting accreting objects with masses of 100 Modot or larger. Because these masses exceed those of stellar black holes in binaries (which extend up to ~ 30 Modot; Belczynski, Sadowski & Rasio 2003), ULXs could then be a new class of astrophysical objects, possibly unconnected with the evolution of the normal stellar population of a galaxy. They could represent the missing link in the black-hole mass distribution, bridging the gap between stellar black holes and the supermassive black holes found in the nuclei of early-type galaxies. These "missing" black holes have been called intermediate mass black holes (IMBH), and could be the remnants of the collapse of primordial stars in the early universe (Madau & Rees 2001; Volonteri, Haardt & Madan 2003; Volonteri & Perna 2005; see the review by Bromm & Larson 2004), or they could be forming in the core collapse of young dense stellar clusters (e.g., Miller & Hamilton 2002).

Conversely, ULXs could represent a particularly high-accretion stage of X-ray binaries, possibly with a stellar black-hole accretor (King et al. 2001), or even be powered by relativistic jets, as in Doppler-beamed microquasars (Koerding, Falke & Markoff 2002). Other models have also been advanced (very young SNRs, e.g., Fabian & Terlevich 1996; young Crab-like pulsars, Perna & Stella 2004), but cannot explain the bulk of the ULXs, given their spectral and time variability characteristics that point to accretion systems (see below).

Given these exciting and diverse possibilities it is not surprising that ULXs have generated a large amount of both observational and theoretical work. An in-depth discussion of all this work is beyond the scope of the present review. Among the recent reviews on ULXs presenting different points of view are those of Fabbiano (2004), Miller & Colbert (2004), Mushotzky (2004), and Fabbiano & White (2006). Two recent short articles in Nature and Science (McCrady 2004 and Fabbiano 2005) are also useful examples of different perspectives on this subject: McCrady argues for the IMBH interpretation of ULXs, whereas Fabbiano instead concludes that although a few very luminous ULXs are strong candidates for IMBHs, the majority may be just sources at the upper luminosity end of the normal XRB population.

Here, I will discuss the main points of the current debate on ULXs, as they pertain to the discourse on X-ray populations, quoting only recent and representative work. In Section 6.1 I discuss the association of ULXs with star formation, in Section 6.2 their spectral and time variability, suggesting the presence of accreting binaries, and in Section 6.3 identification with optical and radio objects. In Section 6.4 I give a summary of the current theoretical debate on the nature of ULXs.

6.1. Association of ULXs with Active Star-Forming Stellar Populations

From a population point of view it is useful to see where we find ULXs. The heightened recent interest in ULXs has spurred a number of studies that have sought to take a systematic view of these sources. These include both works using the Chandra data archive and those revisiting the ROSAT data and the literature. From a mini-survey of 13 galaxies observed with Chandra, including both ellipticals and spirals, Humphrey et al. (2003) suggested a star-formation connection on the basis of a strong correlation of the number of ULXs per galaxy with the 60-µm emission and a lack of correlation with galaxy mass. Swartz et al. (2004) published spectra, variability, and positions for 154 ULXs in 82 galaxies from the Chandra ACIS archive, confirming their association with young stellar populations, especially those of merging and colliding galaxies. This conclusion is in agreement with that of Grimm, Gilfanov & Sunyaev (2003), based on a comparison of XLFs of star-forming galaxies (see Section 4.2). The strong connection of ULXs with star formation is also demonstrated by the analysis of a catalog of 106 ULXs derived from the ROSAT HRI observations of 313 galaxies (Liu & Bregman 2005). Liu & Mirabel (2005) instead compile a catalog of 229 ULXs from the literature, together with optical, IR, and radio counterparts, when available; they observe that the most luminous ULXs (those with LX > 1040 erg s-1), which are the most promising candidates for IMBHs, can be found in either intensely star-forming galaxies or in the halo of ellipticals (the latter, however, are likely to be background QSOs, see below). The association of ULXs with high SFR galaxies is exemplified by the discovery of 14 of these sources in the Antennae galaxies, the prototype galaxy merger (Figure 10).

Figure 10

Figure 10. Left: Chandra ACIS image of the Antennae from two years of monitoring (4.8' side box; from the web page; credit NASA/CXC; Fabbiano et al. 2004a). The regions of most intense emission, where most of the X-ray sources are clustered, correspond to regions of intense star formation. Note that the region of most intense star formation is obscured in X ray (blue). Right: Spitzer IR (3.6-8-µm in red) and optical composite image from the web page Media/releases/ssc2004-14/ssc2004-14a.shtml (Credit: NASA/JPL-Caltech/Z. Wang).

As discussed in Section 3.3, the XLFs of E and S0 galaxies are rather steep, i.e., the number of very luminous sources in these LMXB populations is relatively small, especially in comparison with star-forming galaxies; however, sources with luminosities in excess of 1039 erg s-1 exist (see an earlier discussion of this topic in Fabbiano & White 2005; see also Section 3.7).

Several authors have considered the statistical association of ULXs with early-type galaxies (E and S0s, old stellar populations). Swartz et al. (2004) find that the number of ULXs in early-type galaxies scales with galaxy mass and can be explained with the high luminosity end of the XLF (see Gilfanov 2004 and discussion in Section 3.3). They also point out that ULX detections in early-type galaxies are significantly contaminated by background AGNs, in agreement with the statistical works of Ptak & Colbert (2004) and Colbert & Ptak (2002), which are based on ROSAT HRI (5" resolution) observations of galaxies. Irwin, Bregman & Athey (2004) find that sources in the 1 × 1039 erg s-1 -2 × 1039 erg s-1 luminosity range are likely to belong to the associated galaxies and have spectra consistent with those of Galactic black-hole binaries (Irwin, Athey & Bregman 2003). The sample of sources more luminous than 2 × 1039 erg s-1 (if placed at the distance of the associated galaxy) is instead consistent with the expected number and spatial distribution of background AGNs.

This growing body of results demonstrates that ULXs are associated with the star-forming population. The presence of ULXs in early-type galaxies has been debated, but there is no strong statistical evidence for the existence of a population of sources with LX > 2 × 1039 erg s-1 in these galaxies. In the following I will only discuss ULXs in star-forming galaxies.

6.2. Spectra and Time Variability from Chandra and XMM-Newton

Chandra and XMM-Newton work has confirmed that ULXs are compact accreting sources, building on the more limited observations of nearby ULXs with ASCA (Makishima et al. 2000, Kubota et al. 2001). Flux-color transitions have been observed in a number of ULXs, suggesting the presence of an accretion disk (in the Antennae - Fabbiano, Zezas & Murray 2001; Fabbiano et al. 2003a, b, 2004a; Zezas et al. 2006; M101 - Jenkins et al. 2004; NGC7714 - Soria & Motch 2004; M33 - LaParola et al. 2003; Dubus, Charles & Long 2004; Foschini et al. 2004; Ho II X-1 - Dewangan et al. 2004; and a sample of 5 ULXs in different galaxies monitored with Chandra - Roberts et al. 2004). Some of these spectra and colors are consistent with or reminiscent of those of black-hole binaries (see above references and Colbert et al. 2004, Liu et al. 2005). A recent spectral survey with XMM-Newton finds different spectral types, suggesting either spectral variability or a complex source population (Feng & Kaaret 2005).

Shorter-term time variability is also consistent with the presence of X-ray binaries and accretion disks. In particular a ULX in NGC253 has recently been shown to be a recurrent transient (Bauer & Pietsch 2005). Moreover, features in the power density spectra have been used to constrain the mass of the accreting black hole (Strohmayer & Mushotzky 2003; Soria et al. 2004). In the very luminous M82 ULX (Lx > 1040 erg s-1, Lbol ~ 1041 erg s-1), which is the most compelling IMBH candidate, Strohmayer & Mushotzky (2003) detect a 55mHz QPO, also confirmed by Fiorito & Titarchuck (2004).

The most statistically significant spectra are those obtained with XMM-Newton in nearby very bright ULXs where confusion with unresolved emission in the detection area is not severe. In several cases, a composite power-law plus very soft accretion disk is the simplest model that optimizes the fit to the observed spectra (there are exceptions, e.g., for the 1041 erg s-1 ULX in NGC2276, where a multicolored disk model is preferred, Davis & Mushotzky 2004). A very soft component was first reported by Kaaret et al. (2003) for the ULX NGC5408 X-1, and soon after by Miller et al. (2003) for the ULX NGC1313 X-2, with temperatures of ~ 110 and 150 eV, respectively. These soft components would be consistent with the emission of an accretion disk surrounding an IMBH of nearly 1000 Modot (but see a more recent estimate of 100 Modot for NGC1313 X-2, Zampieri et al. 2004). Similar soft components were found in other ULXs (Miller, Fabian & Miller 2004a; Miller et al. 2004; Jenkins et al. 2005; Roberts et al. 2005).

Unfortunately, these results are not the smoking gun that one may have hoped for to conclusively demonstrate the presence of IMBHs in ULXs. Two other models have been proposed that fit the data equally well, but are consistent with normal stellar black-hole masses. One is the slim disk model (e.g., Watarai et al. 2005, Ebisawa et al. 2004, advanced to explain the emission of an accretion disk in a high accretion mode; see Foschini et al. 2005, Roberts et al. 2005). The second model is a physical Comptonized disk model (Kubota, Makishima & Done 2004). Although both models are significantly more complex than the power-law + soft-component model, nature can easily be wicked, and the models are physically motivated. The controversy is raging, given the tantalizing possibility of proving the discovery of IMBHs (see Fabian, Ross & Miller 2004; Miller, Fabian & Miller 2004b; Wang et al. 2004; Goad et al. 2006).

The recurrent variable very soft ULX in M101 provides an excellent case study to illustrate the difficulty of reaching a firm conclusion on the presence of an IMBH. Given their very soft spectra, SSSs and QSSs in the ULX luminosity range are IMBH candidates (in Sombrero - Di Stefano et al. 2003; M101 - Mukai et al. 2003, 2005; Kong, Di Stefano & Yuan 2004; the Antennae - Fabbiano et al. 2003b). These sources are too luminous to be explained in terms of hot white dwarfs, unless the emission is beamed, which is unlikely (e.g., Fabbiano et al. 2003b).

The expanding black-hole photosphere of a stellar black hole was first suggested to explain the M101 very soft ULX (Mukai et al. 2003), but the subsequent detection of a hard power-law component and low/hard-high/soft spectral variability pointed to a Comptonized accretion disk in a black-hole binary (Kong, Di Stefano & Yuan 2004; Mukai et al. 2005); but what kind of black hole?

Based on the XMM-Newton spectrum, which can be fitted with an absorbed blackbody, and implies outburst luminosities in the 1041 erg s-1 range, Kong, Di Stefano & Yuan (2004) advanced the IMBH candidacy. Mukai et al. (2005), instead, argue for a 20-40 Modot stellar black-hole counterpart. Their main point is that the high LX derived in the previous study results from the adoption of an emission model with a considerable amount of line-of-sight absorption; the colors of the optical counterpart, instead, are consistent with very little absorption; moreover, if the obscuring material were close to the black hole it would be most likely ionized (warm absorber). Adopting an accretion disk plus emission line model, Mukai et al. (2005) obtain luminosities in the 1039 erg s-1 range. They also use the variability power density spectrum of the source to constrain the emission state, and with the luminosity, the mass of the black hole. This is another example where a considerable amount of ambiguity exists in the choice of the X-ray spectral model, and X-ray spectra alone may not give the conclusive answer. The luminous optical counterpart makes this source an obvious candidate for future studies aimed at obtaining the mass function of the system.

6.3. Counterparts at Other Wavelengths

As shown by the example at the end of Section 6.2, identification of ULXs may be crucial for understanding their nature. Three main classes of counterparts have been discussed in the literature: stellar counterparts, ionized or molecular nebulae, and radio sources. Stellar counterparts tend to have very blue colors, suggesting early-type stars, although the colors could also arise from the optical emission of the accretion disk (Kaaret, Ward & Zezas 2004; Liu, Bregman & Seitzer 2004; Soria et al. 2005; Zampieri et al. 2004; Rappaport, Podsiadlowski & Pfahl 2005; see also Fabbiano & White 2006 for earlier references). These counterparts would point to the high accretion rate model of ULXs, if they were indeed early-type stars (e.g., King et al. 2001; Rappaport, Podsiadlowski & Pfahl 2005). However, even ignoring the uncertainty on the nature of the optical counterpart, these results cannot firmly constrain the nature of the compact object. A recent paper by Copperwheat et al. (2005) proposes a model, including irradiation by X-rays of both the accretion disk and the companion star, which when supplemented by variability data and IR photometry, could be used to constrain both the nature of the companion star and the mass of the accreting black hole.

Nebular counterparts suggest isotropic emission in some cases, and therefore a truly large LX, thus arguing against a substantial amount of beaming and pointing to fairly massive black holes (Roberts et al. 2003; Pakull & Mirioni 2003; Kaaret, Ward & Zezas 2004); radio counterparts have alternately been found consistent with either beamed sources or IMBHs (Kaaret et al. 2003; Neff, Ulvestad & Campion 2003; Miller, Mushotzy & Neff 2005; Koerding, Colbert & Falke 2005). Optical variability studies of the stellar counterparts are needed to firmly measure the mass of the system. The new generation of large-area, high-resolution, optical telescopes are likely to solve the nature of these ULXs.

In more distant systems, like the Antennae (D ~ 19 Mpc - Zezas et al. 2002a, b, 2005) or the Cartwheel galaxies (D ~ 122 Mpc - Gao et al. 2003, King 2004, Wolter & Trinchieri 2004), where spectacular populations of ULXs are detected, individual stellar counterparts cannot be detected. However, comparison with the optical emission field also provides very interesting results. In the Antennae, ULXs tend not to coincide with young star clusters, suggesting that either the system has been subject to a supernova formation kick to eject it from its birthplace (thus implying a normal HMXB with a stellar mass black hole; Zezas et al. 2002b; Sepinsky, Kalogera & Belczynski 2005), or that the parent cluster has evaporated, in the core collapse model of IMBH formation (e.g., Zwart et al. 2004). However, a recent paper suggests that some of these displacements may be reduced with better astrometric corrections (Clark et al. 2005). In the Cartwheel, the ULXs are associated with the most recent expanding star-formation ring, setting strong constraints to the IMBH hypothesis and favoring the high accretion HMXB scenario (King 2004). It must be said, however, that given the distance of this galaxy, and the lack of time monitoring, it cannot be excluded that the ULXs may represent clumps of unresolved sources.

Higher red-shift galaxy and QSO counterparts to ULXs have also been found in some cases (Masetti et al. 2003; Arp, Gutierrez & Lopez-Corredoira 2004; Gutierrez & Lopez-Correidora 2005; Burbidge et al. 2004; Galianni et al. 2005; Clark et al. 2005; see also H. Arp & E.M. Burbidge, submitted; Ghosh et al. 2005). Although at this point these identifications are still few and consistent (within small number statistics) with chance coincidences with background AGN, some of the above authors (H. Arp, E.M. Burbidge, G. Burbidge and collaborators) have raised the hypothesis of a physical connection between the QSO and the parent galaxy; clearly this possibility cannot be extended to the entire body of ULXs, given the results of other identification campaigns (see above).

6.4. Models of ULX Formation and Evolution, and Paths for Future Work

I will summarize here some of the more recent theoretical work on ULXs, and refer the reader to the reviews cited earlier for details on earlier work. As I've already noted, the two principal lines of thought are: (a) most ULXs are IMBHs; (b) most ULXs are luminous X-ray sources of "normal" stellar origin and the IMBH explanation should be sought only for the ULXs with Lbol > 1041 erg s-1 (the M82 ULX - Kaaret et al. 2001, Matsumoto et al. 2001; the NGC2276 ULX - Davis & Mushotzky 2004; the most luminous ULX in the Cartwheel galaxy - Gao et al. 2003, Wolter & Trinchieri 2004; and the variable ULX in NGC7714 - Soria & Motch 2004; Smith, Struck & Nowak 2005).

The stellar evolution camp was originally stimulated by the abundance of ULXs in star-forming galaxies (King et al. 2001; King 2004) and by the apparently universal shape of the XLF of the star-forming population (Grimm, Gilfanov & Sunyaev 2003; see Section 4.2). The variability and spectra of these systems (see Section 6.2) point to accretion binaries. In this paradigm, the problem is to explain the observed luminosities. Both relativistic (Koerding, Falke & Markoff 2002) and nonrelativistic beaming (King et al. 2001), and super-Eddington accretion disks (Begelman 2002; both spectra and observed variability patterns can be explained, M. Belgelman, private communication) have been suggested as a way to explain the source luminosities inferred from the observations. With the exception of relativistic beaming, these mechanisms can account for a factor of 10 enhancement of the luminosity above the Eddington value. If black-hole masses of a few tens solar masses exist (Belczynski, Sadowski & Rasio 2004), most or all the ULXs could be explained this way. For example, Rappaport, Podsiadlowski & Pfahl (2005) have combined binary evolution models and binary population synthesis, finding that for donors with M geq 10 Modot, accretion binaries can explain the ULXs, with modest violation of the Eddington limit.

The IMBH camp has generated a larger volume of papers. IMBHs may be remnants of collapse in the early universe (e.g., Heger & Woosley 2002; Islam, Taylor & Silk 2004; Van der Marel 2004), or may result from the collapse of dense stellar clusters (e.g., Gurkan, Freitag & Rasio 2004; Zwart et al. 2004). In the cosmological remnant options, one would expect IMBHs to be particularly abundant in the more massive elliptical galaxies, contrary to the observed association with star-forming galaxies (Zezas & Fabbiano 2002). However, IMBHs would not be visible unless they are fueled, and fuel is more readily available in star-forming galaxies, in the form of dense molecular clouds (Schneider et al. 2002; Krolik 2004). Accretion from a binary companion is an efficient way of fueling an IMBH, and consequently a number of papers have explored the formation of such binaries via tidal capture in GCs. In this picture, the ULX may not be still associated with the parent cluster because of cluster evaporation (Hopman, Zwart & Alexander 2004; Li 2004; Zwart, Dewi & Maccarone 2004; Zwart et al. 2004). A twist to the cosmological hypothesis is given by the suggestion that the very luminous ULXs, with Lbol > 1041 erg s-1 such as the M82 ULX, may be the nuclei of satellite galaxies, switching on in the presence of abundant fuel (King & Dehnen 2005).

Some of this work has resulted in predictions that can be directly compared with the data, and complement the tests based on the study of the optical and multiwavelength counterparts discussed in Section 6.3. In particular, the slope and normalization of the high-luminosity XLFs of star-forming galaxies have been reproduced in both IMBH (Islam, Taylor & Silk 2004; Krolik 2004) and jet models (Koerding, Colbert & Falke 2004). Zezas & Fabbiano (2002) discuss the effect of either a beamed population of ULXs or a population of IMBHs in the context of the XLF of the Antennae. Gilfanov, Grimm & Sunyaev (2004b) predict a change of slope in the LX-SFR relation, where LX is the total X-ray luminosity of a galaxy, if a new population of IMBHs is present at the higher luminosities (see Section 7).

Other properties have also been investigated, including the X-ray spectral distribution (Section 6.2), the presence of radio emission from IMBHs and the comparison of radio and X-ray properties with those of AGN and stellar black-hole Galactic binaries (e.g., Merloni, Heinz & Di Matteo 2003), and time variability-based tests. The latter include studying the QPO frequency (which may be a function of back-hole mass, Abramowicz et al. 2004), the observation of long-term transient behavior (expected from IMBH binaries, whereas thermal-timescale mass transfer onto stellar black holes would produce stable disks; Kalogera et al. 2004) and the detection of eclipses (expected more frequently in stellar black-hole binaries than in IMBHs, Pooley & Rappaport 2005). These time variability tests require long-term monitoring of ULXs and future larger X-ray telescopes.

Next Contents Previous