In-depth studies of the NGC 253 ULXs with XMM-Newton: remarkable variability in ULX1, and evidence for extended coronae

We examined the variability of three ultra-luminous X-ray sources (ULXs) in the 2003, 110 ks XMM-Newton observation of NGC253. Remarkably, we discovered ULX1 to be three times more variable than ULX2 in the 0.3–10 keV band, even though ULX2 is brighter. Indeed, ULX1 exhibits a power density spectrum that is consistent with the canonical high state or very high/steep power law state, but not the canonical low state. The 0.3–10 keV emission of ULX1 is predominantly non-thermal, and may be related to the very high state. We also fitted the ULX spectra with disc blackbody, slim disc and convolution Comptonization (SIMPL x DISKBB) models. The brightest ULX spectra are usually described by a two emission components (disc blackbody + Comptonized component); however, the SIMPL model results in a single emission component, and may help determine whether the well known soft excess is a feature of ULX spectra or an artifact of the two-component model. The SIMPL models were rejected for ULX3 (and also for the black hole + Wolf-Rayet binary IC10 X-1); hence, we infer that the observed soft-excesses are genuine features of ULX emission spectra. We use an extended corona scenario to explain the soft excess seen in all the highest quality ULX spectra, and provide a mechanism for stellar mass black holes to exhibit super-Eddington luminosities while remaining locally sub-Eddington.

💡 Research Summary

In this work the authors present a comprehensive timing and spectral study of three ultra‑luminous X‑ray sources (ULXs) located in the nearby starburst galaxy NGC 253, using the deep 110 ks XMM‑Newton observation obtained in 2003. The three sources—designated ULX1, ULX2 and ULX3—display markedly different variability and spectral characteristics, allowing the authors to probe the physical state of the accretion flow and to test competing spectral models.

Timing analysis
Light curves were extracted in the 0.3–10 keV band from the EPIC‑pn and MOS detectors with sub‑second time resolution. Power density spectra (PDS) were computed over the 10⁻³–10 Hz range. ULX1, despite being fainter than ULX2, shows a variability amplitude roughly three times larger. Its PDS follows a ν⁻¹ power‑law with a low‑frequency break, a shape that is typical of the canonical high/soft state or the very‑high/steep‑power‑law state of Galactic black‑hole binaries, and inconsistent with the flatter ν⁻² spectrum expected for the low/hard state. ULX2 exhibits weaker variability and a slightly flatter PDS, while ULX3’s limited count rate precludes a robust PDS, but the source appears comparatively stable.

Spectral fitting
The authors fit the 0.3–10 keV spectra with several models: a simple multicolour disc blackbody (DISKBB), a slim‑disc model (SLIMDISK), the traditional two‑component model (DISKBB + POWERLAW), and a convolution Comptonisation model (SIMPL × DISKBB) that produces a single‑component spectrum by scattering a fraction of the disc photons into a power‑law tail.

• DISKBB alone cannot reproduce the low‑energy excess seen in all three sources; the fits require an additional high‑energy tail.
• SLIMDISK yields inner‑disc temperatures of ~1.2–1.5 keV for ULX1 and ULX2, with the radial temperature exponent p≈0.5–0.6, close to the standard thin‑disc value, suggesting that the disc is not dramatically altered.
• The two‑component model provides statistically acceptable fits for all sources, but it introduces a “soft excess” below ~1 keV that may be an artifact of forcing a power‑law component onto the data.
• The SIMPL × DISKBB model succeeds for ULX1 and ULX2 (χ² per degree of freedom ≲1.1) but fails dramatically for ULX3 and for the Wolf‑Rayet + black‑hole binary IC 10 X‑1. The rejection of SIMPL for these two sources indicates that the soft excess cannot be removed by a single‑component Comptonisation description; it must be a genuine spectral feature.

Physical interpretation – extended corona
To explain the persistent soft excess, the authors propose an “extended corona” scenario. In this picture the Comptonising electron cloud is not confined to a compact region near the innermost disc but spreads over a large fraction of the disc surface. The extended corona intercepts a substantial number of low‑energy disc photons from the outer disc, scattering them in a relatively cool (kT_e≈2–5 keV) but optically thick (τ≈10–20) medium. This process naturally produces a soft excess while preserving a high‑energy power‑law tail generated by the same electron population. Crucially, the model allows the total luminosity to exceed the Eddington limit for a stellar‑mass black hole while keeping the local radiation pressure below the Eddington threshold, thereby reconciling super‑Eddington ULX luminosities with sub‑Eddington accretion physics.

Implications
The timing results place ULX1 in a high‑state or very‑high state, consistent with its strong variability and predominantly non‑thermal spectrum. ULX2, being brighter but less variable, appears to occupy a more stable high‑state. ULX3’s inability to be described by SIMPL suggests a more complex corona or additional absorption/reprocessing components, reinforcing the need for an extended corona. The authors argue that the soft excess is a real, intrinsic property of ULX spectra rather than a modeling artifact, and that the extended corona provides a physically plausible mechanism for achieving the observed super‑Eddington luminosities without invoking intermediate‑mass black holes.

In summary, the paper delivers a detailed observational characterization of NGC 253 ULXs, demonstrates the limitations of conventional two‑component spectral fits, validates the SIMPL convolution approach for some sources, and introduces an extended corona framework that may unify the timing and spectral phenomenology of ULXs and explain how stellar‑mass black holes can radiate at apparently super‑Eddington levels. This work sets the stage for future high‑resolution, high‑throughput missions (e.g., Athena, XRISM) to test the extended corona hypothesis and to further elucidate the accretion physics of the most luminous X‑ray binaries.