1. New insights into ultraluminous X-ray sources from XMM-Newton/EPIC observations Tim Roberts Ann-Marie Stobbart, Bob Warwick, Mike Goad, Leigh Jenkins (Leicester) Martin Ward (Leicester/Durham) Jörn Wilms (Warwick) Phil Uttley, James Reeves (GSFC) Luminous X-ray sources in M101 (from Jenkins et al. 2004)

2. Tuesday 12th April 2005Tim Roberts - New insights into ULXs 2 ULXs and IMBHs ULXs – discrete X-ray sources with L X > 10 39 erg s -1. But at these luminosities L X > L Edd for a ~ 10 M  black hole – a new class of ~ 100 – 10 4 M  intermediate-mass black holes (IMBHs) required? Supporting evidence from “soft excess” in XMM-Newton ULX spectra (e.g. Miller et al. 2003). Now 10+ examples. NGC 1313 X-1 Power-law + diskbb T  M -0.25 kT in ~ 0.15 keV cf. kT in ~ 1 – 2 keV for stellar BHs

3. Tuesday 12th April 2005Tim Roberts - New insights into ULXs 3 But…… Multiple ULXs (10+) are found in Starburst galaxies – e.g. Cartwheel galaxy (Gao et al. 2003). Ongoing star formation  ULXs are intrinsically short-lived. Requires an infeasibly large underlying population of IMBHs (King 2004). Alternative: are ULXs in Starbursts high-mass X-ray binaries (HMXBs)? NB – no comparable population in ellipticals (Irwin et al. 2004). From Gao et al. (2003)

4. Tuesday 12th April 2005Tim Roberts - New insights into ULXs 4 Stellar-mass BHs as ULXs Possible mechanisms for breaking Eddington limit: – Beaming by relativistic jets (e.g. Körding et al. 2002). – Anisotropic radiation patterns (King et al. 2001). – True super-Eddington discs (Begelman 2002, Ebisawa et al. 2003). Podsiadlowski et al. (2003), Rappaport et al. (2005) – super- Eddington mass transfer rates in HMXBs – account for most ULXs. Blue stellar counterparts to several ULXs. At least three stellar mass BHs (albeit LMXBs) in our galaxy have been seen to reach super-Eddington luminosities – GRS1915+105 does so frequently (McClintock & Remillard 2003). Some ULXs do have stellar-mass disc temperatures (~1 – 2 keV). But not much recent observational evidence from ULX X-ray diagnostics…….

5. Tuesday 12th April 2005Tim Roberts - New insights into ULXs 5 The NGC 55 ULX See Stobbart, Roberts & Warwick, 2004, MNRAS, 351, 1063. Combined XMM-Newton EPIC image DSS image with EPIC contour

6. Tuesday 12th April 2005Tim Roberts - New insights into ULXs 6 XMMU J001528.9-391319 10 39 erg s -1 Source exhibits temporal variability including dipping. Dips most prominent at high energies. Γ ~ 4 kT in ~ 0.9 keV

7. Tuesday 12th April 2005Tim Roberts - New insights into ULXs 7 A problematic spectrum Dominance of power-law continuum at soft energies not seen before in Galactic systems – though e.g. GRS 1915+105 high state extrapolates below 2 keV to this spectrum. Disc parameters extreme (high kT in, low R in ) but plausible for slim disc accretion onto a stellar-mass (or slightly bigger) BH. Problem is dominant soft power-law – cannot be disc- Comptonisation (too few photons below peak in disc emissivity). Unlikely to be jet – too soft ( Γ ~ 3 – 4 vs ~ 1.5 – 2 for jet). Power-law form is not consistent with thermal emission from outflow/wind. Insurmountable problem? Possible explanation: greater spectral complexity. “3-layer” model of Zhang et al. (2000) – based on the Solar atmosphere - cold inner disc, warm & optically-thick accretion disc atmosphere, much hotter optically-thin corona. From Zhang et al. (2000) kT ~ 0.2 – 0.5 keVkT ~ 1 – 1.5 keV, τ ~ 10 kT ~ 100 keV NGC 55 ULX VHS of GX 339-4

8. Tuesday 12th April 2005Tim Roberts - New insights into ULXs 8 Other examples of “new” spectrum This spectrum is seen in second L X ~ 10 39 erg s -1 ULX – M33 X-8 (Foschini et al. 2004). More luminous (L X ~ 5 × 10 39 erg s -1 ) NGC 5204 X-1 data well fit by both “IMBH model”, i.e. cool accretion disc (kT in ~ 0.2 keV) + hard power-law continuum ( Γ ~ 2), and “non-standard” description ( Γ ~ 3.3, kT in ~ 2.2 - 2.8 keV). kT in ~ 1.16 keV Γ ~ 2.5 From Roberts et al. (2005) M33 X-8NGC 5204 X-1

9. Tuesday 12th April 2005Tim Roberts - New insights into ULXs 9 A sample of bright ULXs How prevalent is the “new” spectrum in ULXs? Particularly in comparison to an IMBH spectrum? Select 13 (predominantly archival) ULXs observed by XMM-Newton/EPIC with (a) ~20 ks or more EPIC exposure, and (b) > 10 ct/ks in ROSAT HRI. Expect ~ few thousand counts per source. Full range of L X covered (10 39 – few × 10 40 erg s -1 ). Uniform reduction to produce clean spectra for comparison of empirical models and state-of-the-art physical models. Analysis ongoing…

10. Tuesday 12th April 2005Tim Roberts - New insights into ULXs 10 Empirical models Absorbed multi-colour disc blackbody spectrum ( diskbb in XSPEC ) rejected at high significance for all data. Absorbed power-law continuum not rejected at 95% confidence in only 4 datasets (including 3 lowest quality). IMBH model produces “good” ( χ 2 ν ~ 1) fits in 7 sources (“better” in 2 more). Find kT in ~ 0.1 – 0.25 keV, Γ ~ 1.6 – 2.5. Masses circa. 1000 M  for IMBH. Problem: Γ too small? Theory and observations show Γ > 2.5 for high-state black hole accretion discs.

11. Tuesday 12th April 2005Tim Roberts - New insights into ULXs 11 Other four datasets much prefer “new” spectrum. But this description not rejected in 6/7 IMBH candidates, and as equally plausible as IMBH in 4. Key discriminator: curvature in 2-10 keV regime. Broken power-law versus power-law fits over 2-10 keV: 5 significant improvements (> 3 σ via F-test), 2 marginal. Correspond with preference for “new” spectrum. 2-10 keV curvature IMBH model “New” model From Roberts et al. (2005)

12. Tuesday 12th April 2005Tim Roberts - New insights into ULXs 12 Physical models (1) Slim disc model (e.g. Watarai et al. 2001; Ebisawa et al. 2004); XSPEC parameterisation courtesy of K. Ebisawa. At ~L Edd expect advection-dominated optically-thick discs – differences to “standard accretion disc, e.g. R in decreases below ISCO as Mdot increases. Provides poor fits in most cases; problems with degeneracy between α and M BH, Mdot. M BH typically 10 – 50 M  and < 100 M  in all but one case. Gives Mdot in 0.1 – 10 in Eddington units.

13. Tuesday 12th April 2005Tim Roberts - New insights into ULXs 13 Physical models (2) Physically self-consistent accretion disc + comptonisation model: diskpn+eqpair in XSPEC. Fits well to 5 datasets; a further 5 have good fits, or only moderately worse than, other (empirical) models. But only two fits look like IMBHs: cool disc, low optical depth (kT ~ 0.3 keV, τ ~ 1). Other fits have cool discs (kT ~ 0.2 keV) but are optically thick ( τ ~ 6 – 10). INCONSISTENT WITH IMBHs! cf. Zhang’s 3-layer model… kT ~ 0.2 – 0.5 keV kT ~ 100 keV kT ~ 1 – 1.5 keV, τ ~ 10

14. Tuesday 12th April 2005Tim Roberts - New insights into ULXs 14 Archetypal luminous ULX (L X > 10 40 erg s -1 ). Deep XMM-Newton observation (80 ks, though > 50% spoiled by bad space weather). X-ray spectrum: not well-fit by any model. Best empirical description: IMBH model (kT in ~ 0.2 keV, Γ ~ 2.6). But diskpn+eqpair provides best overall fit with kT ~ 0.2 keV, τ ~ 6.6. Also – lack of variability puzzling. PDS – red noise, only seen above poisson noise level at < 10 mHz. Not consistent with ~1000 solar mass BH in high state. Holmberg II X-1

15. Tuesday 12th April 2005Tim Roberts - New insights into ULXs 15 Conclusions Detailed spectroscopy – some ULXs just don’t look like IMBHs with “standard” accretion disc + corona spectra extrapolated from Galactic BHs (2 – 10 keV curvature/flat Γ ). Highly compton-thick layer may be key evidence – ionised surface of bloated accretion disc fed by super- Eddington inflow of material from high-mass secondary. BH mass few 10s of M . Lack of short-term variability supports Compton-thick layer. However, only conclusive means of ending this debate is to derive a dynamical mass limit on the BH from orbital dynamics…and that’s another story!