Super-accreting X-ray Binaries

• Super-accreting X-ray binaries are systems where the mass transfer rate approaches or exceeds the Eddington limit, leading to advection-dominated flows and powerful outflows.
• They involve diverse accretion channels such as wind capture, Roche-lobe overflow, and degenerate-donor transfer, each imparting unique spectral and timing signatures.
• Observational diagnostics including luminosity profiles, spectral curvature, and timing modulations provide insights into the extreme feedback processes and evolutionary pathways of these systems.

Super-accreting X-ray binaries are accreting compact binaries in which the mass-transfer rate is near or above the Eddington limit of the compact object, so that the flow departs from the standard thin-disc or weakly perturbed wind-accretion regime and enters a feedback-dominated state with advection, outflows, geometrical collimation, or magnetospheric regulation. In the high-mass X-ray binary literature, the term is also used in a broader sense for systems that reach very high instantaneous accretion rates and X-ray luminosities, often near the neutron-star Eddington luminosity, even when the source is not persistently ultraluminous. In the ultraluminous X-ray source literature, the strict observational benchmark is typically $L_X \gtrsim 10^{39}\,\mathrm{erg\,s^{-1}}$, a regime now known to be accessible to both black holes and neutron stars (Dage et al., 2024, Chaty, 2015).

1. Definition and parameter regimes

A compact object of mass $M$ has a classical Eddington luminosity

$L_{\rm Edd} = \frac{4\pi G M m_p c}{\sigma_T},$

or, in a commonly used numerical form,

$$L_{\rm Edd} \simeq 1.3\times10^{38}\left(\frac{M}{M_\odot}\right)\,\mathrm{erg\,s^{-1}}.$$

For a $1.4\,M_\odot$ neutron star this is of order $2\times10^{38}\,\mathrm{erg\,s^{-1}}$, whereas a $10\,M_\odot$ black hole has $$L_{\rm Edd}\sim 1.3\times10^{39}\,\mathrm{erg\,s^{-1}}$$ (Dage et al., 2024, Paul et al., 2011).

The corresponding accretion-rate language is usually expressed through

$\dot m \equiv \frac{\dot M}{\dot M_{\rm Edd}},$

with “super-Eddington” meaning $\dot m>1$. For hydrogen-poor accretion onto a neutron star, one formulation adopts

$M$0

which is relevant for ultracompact systems transferring He-rich material (King, 2011).

This parameter space encompasses several observationally distinct subclasses. Wind-fed supergiant HMXBs commonly radiate at $M$1, obscured systems can reach $M$2, Roche-lobe-filling supergiant systems can reach $M$3, Be/X-ray binaries in giant outbursts can reach $M$4 or more, and ULXs extend to $M$5 (Chaty, 2015, Paul et al., 2011, Dage et al., 2024). A plausible implication is that “super-accreting” is best treated as a physical regime rather than a single taxonomic class.

2. Accretion channels, feedback, and flow structure

Three mass-transfer channels recur across the literature: wind capture from massive donors, Roche-lobe overflow through $M$6, and degenerate-donor transfer in ultracompact systems. In wind-fed supergiant HMXBs, the donor launches a radiatively driven wind with a standard velocity law

$M$7

with typical OB-supergiant parameters

$M$8

The compact object then accretes in a Bondi–Hoyle–Lyttleton-like regime in which the effective accretion radius increases sharply as the relative velocity decreases (Manousakis et al., 2013).

X-ray feedback is central. In hydrodynamic models of supergiant HMXBs, the ionization parameter is

$M$9

and line driving is suppressed when

$L_{\rm Edd} = \frac{4\pi G M m_p c}{\sigma_T},$0

Above this threshold, “most of the elements responsible for the wind acceleration are fully ionized and hence the radiative acceleration force vanishes.” The resulting slowdown of the wind increases the local density, the effective accretion radius, and the accretion rate, creating a positive feedback loop between accretion and wind inhibition (Manousakis et al., 2013).

In supercritical discs, the inner flow is no longer thin. A widely used picture places the onset of the thick, outflowing region at the spherization radius $L_{\rm Edd} = \frac{4\pi G M m_p c}{\sigma_T},$1; inside that radius, radiation pressure launches strong winds and the escaping luminosity follows

$L_{\rm Edd} = \frac{4\pi G M m_p c}{\sigma_T},$2

For sufficiently high $L_{\rm Edd} = \frac{4\pi G M m_p c}{\sigma_T},$3, the emission is geometrically beamed through a funnel. One empirical prescription gives

$L_{\rm Edd} = \frac{4\pi G M m_p c}{\sigma_T},$4

so isotropic-equivalent luminosities can exceed $L_{\rm Edd} = \frac{4\pi G M m_p c}{\sigma_T},$5 for stellar-mass accretors (Dage et al., 2024).

A related channel is wind Roche-lobe overflow. In this regime the donor underfills its Roche lobe, with filling factor

$L_{\rm Edd} = \frac{4\pi G M m_p c}{\sigma_T},$6

but the wind is slow enough, and the donor close enough to Roche-lobe filling, that the Roche potential beams the outflow into the orbital plane and toward the compact object. Ballistic calculations give captured fractions $L_{\rm Edd} = \frac{4\pi G M m_p c}{\sigma_T},$7 up to $L_{\rm Edd} = \frac{4\pi G M m_p c}{\sigma_T},$8, often an order of magnitude above the Bondi–Hoyle estimate, and imply that WRLOF-driven ULXs require $L_{\rm Edd} = \frac{4\pi G M m_p c}{\sigma_T},$9 (Mellah et al., 2018).

3. Wind-fed supergiant systems and fast transients

Supergiant HMXBs are among the clearest laboratories for high-efficiency wind accretion. They typically host a neutron star orbiting an OB supergiant at a separation $$L_{\rm Edd} \simeq 1.3\times10^{38}\left(\frac{M}{M_\odot}\right)\,\mathrm{erg\,s^{-1}}.$$0, and hydrodynamic simulations reveal persistent bow shocks and accretion wakes. In obscured systems such as IGR J17252–3616, reproducing the orbital dependence of the column density requires a very slow wind terminal velocity,

$$L_{\rm Edd} \simeq 1.3\times10^{38}\left(\frac{M}{M_\odot}\right)\,\mathrm{erg\,s^{-1}}.$$1

together with a neutron-star mass

$$L_{\rm Edd} \simeq 1.3\times10^{38}\left(\frac{M}{M_\odot}\right)\,\mathrm{erg\,s^{-1}}.$$2

This wind speed is 2–3 times lower than expected for an OB supergiant of that type and produces $$L_{\rm Edd} \simeq 1.3\times10^{38}\left(\frac{M}{M_\odot}\right)\,\mathrm{erg\,s^{-1}}.$$3, about ten times larger than in classical systems (Manousakis et al., 2013).

Vela X-1 illustrates a different high-accretion phenomenon. In two-dimensional VH-1 simulations, low-density bubbles form behind the bow shock, reach sizes $$L_{\rm Edd} \simeq 1.3\times10^{38}\left(\frac{M}{M_\odot}\right)\,\mathrm{erg\,s^{-1}}.$$4 times the accretion radius, and collapse on a timescale comparable to the free-fall time at $$L_{\rm Edd} \simeq 1.3\times10^{38}\left(\frac{M}{M_\odot}\right)\,\mathrm{erg\,s^{-1}}.$$5 cm, yielding a quasi-periodic modulation at $$L_{\rm Edd} \simeq 1.3\times10^{38}\left(\frac{M}{M_\odot}\right)\,\mathrm{erg\,s^{-1}}.$$6 s. The simulated and observed luminosity distributions are all well fit by log-normal distributions, with $$L_{\rm Edd} \simeq 1.3\times10^{38}\left(\frac{M}{M_\odot}\right)\,\mathrm{erg\,s^{-1}}.$$7 for INTEGRAL, $$L_{\rm Edd} \simeq 1.3\times10^{38}\left(\frac{M}{M_\odot}\right)\,\mathrm{erg\,s^{-1}}.$$8 for RXTE, and $$L_{\rm Edd} \simeq 1.3\times10^{38}\left(\frac{M}{M_\odot}\right)\,\mathrm{erg\,s^{-1}}.$$9 for the simulation. In this interpretation, off-states arise without invoking clumpy winds or magnetospheric gating; instead, the accretion flow exhibits self-organized criticality (Manousakis et al., 2013).

Supergiant Fast X-ray Transients extend the same wind-fed physics into a far larger dynamic range. Their defining properties are bright flares lasting from a few minutes to a few hours, with $1.4\,M_\odot$0, quiescent states at $1.4\,M_\odot$1, and more frequent intermediate states at $1.4\,M_\odot$2. The dynamic range can reach $1.4\,M_\odot$3, and only the brightest flares are typically detected by INTEGRAL above 17 keV (Sidoli, 2011).

The physical interpretation remains debated. Clumpy-wind models attribute flares to the capture of dense wind clumps; magnetic or centrifugal gating models invoke transitions across the magnetospheric or corotation radii; equatorial-wind scenarios explain systems with strictly phase-locked outbursts; and hydrodynamic models emphasize intrinsic instabilities in the accretion stream (Sidoli, 2011, Chaty, 2014). In the “grand unification” picture for supergiant systems, obscured sgHMXBs, intermediate SFXTs, and classical SFXTs differ primarily by orbital geometry and the radius at which the neutron star samples the structured wind (Chaty, 2014).

4. Disc-dominated super-accretors and ultracompact systems

Disc-mediated transfer remains the most direct route to sustained near-Eddington accretion in HMXBs. Roche-lobe-filling supergiant systems are the classical bright sgHMXBs, with matter flowing through the inner Lagrangian point to an accretion disc and outbursts reaching $1.4\,M_\odot$4. In this regime, the compact object approaches the neutron-star Eddington limit in a comparatively steady fashion (Chaty, 2015).

Be/X-ray binaries provide a transient disc-fed analogue. Type I outbursts are regular, periastron-linked events with $1.4\,M_\odot$5, whereas Type II outbursts are giant, not tied to orbital phase, last weeks to months, and reach

$1.4\,M_\odot$6

These giant outbursts occur when a large fraction of the Be decretion disc is accreted and are among the clearest HMXB examples of near-Eddington accretion (Paul et al., 2011).

Ultracompact X-ray binaries show that super-Eddington accretion is not confined to massive donors. For a white-dwarf donor approximated as an $1.4\,M_\odot$7 polytrope, the period–mass relation is

$1.4\,M_\odot$8

and gravitational-radiation-driven mass transfer follows

$1.4\,M_\odot$9

Young UCXBs with $2\times10^{38}\,\mathrm{erg\,s^{-1}}$0 can therefore exceed the hydrogen-poor Eddington rate of

$2\times10^{38}\,\mathrm{erg\,s^{-1}}$1

and appear as ULXs with mild or strong beaming depending on $2\times10^{38}\,\mathrm{erg\,s^{-1}}$2 (King, 2011).

This UCXB interpretation has been used to explain the brightest globular-cluster X-ray sources without invoking black holes. For H-poor accretion onto a neutron star, the apparent luminosity can be written

$2\times10^{38}\,\mathrm{erg\,s^{-1}}$3

with $2\times10^{38}\,\mathrm{erg\,s^{-1}}$4 for $2\times10^{38}\,\mathrm{erg\,s^{-1}}$5. For $2\times10^{38}\,\mathrm{erg\,s^{-1}}$6, the apparent luminosity is $2\times10^{38}\,\mathrm{erg\,s^{-1}}$7 with $2\times10^{38}\,\mathrm{erg\,s^{-1}}$8; for HLX-1, an extreme UCXB interpretation would require $2\times10^{38}\,\mathrm{erg\,s^{-1}}$9, $10\,M_\odot$0, $10\,M_\odot$1 min, and $10\,M_\odot$2 (King, 2011).

5. Ultraluminous X-ray binaries and supercritical accretion

ULXs are extragalactic, non-nuclear X-ray binaries with

$10\,M_\odot$3

and they constitute the clearest observational realization of super-accreting X-ray binaries. The current picture no longer requires intermediate-mass black holes for most ULXs: some are powered by accreting neutron stars, as demonstrated by pulsating ULXs, and others by stellar-mass or moderately massive black holes accreting above the Eddington limit (Dage et al., 2024).

For neutron-star ULXs, a specific evolutionary scenario links non-pulsating ULXs and ultra-luminous supersoft sources to a super-critical propeller regime, and pulsating ULXs to a later super-critical accretor phase. In that framework, the disc is supercritical when $10\,M_\odot$4, the propeller regime requires

$10\,M_\odot$5

and the accretor regime requires

$10\,M_\odot$6

The luminosity can then be powered either by direct accretion,

$10\,M_\odot$7

or by spin-down power,

$10\,M_\odot$8

with apparent luminosity enhanced by a beaming factor $10\,M_\odot$9 in the illustrative models (Erkut et al., 2018).

Population-level X-ray spectroscopy supports this supercritical interpretation. Using 200 luminous HMXB candidates in 27 nearby galaxies, the intrinsic collective spectrum per unit star-formation rate is

$$L_{\rm Edd}\sim 1.3\times10^{39}\,\mathrm{erg\,s^{-1}}$$0

equivalent to a photon index $$L_{\rm Edd}\sim 1.3\times10^{39}\,\mathrm{erg\,s^{-1}}$$1. That spectrum is dominated by ULXs with $$L_{\rm Edd}\sim 1.3\times10^{39}\,\mathrm{erg\,s^{-1}}$$2, with hard sources dominating above $$L_{\rm Edd}\sim 1.3\times10^{39}\,\mathrm{erg\,s^{-1}}$$3 keV and soft and supersoft sources dominating at lower energies. It has been interpreted as the angle-integrated emission of near- and super-critically accreting stellar-mass black holes and neutron stars (Sazonov et al., 2017).

Detailed source studies reinforce the same point. In NGC 5643 ULX1, the high-quality XMM-Newton spectra are better described by a broad thermal component, an advection-dominated disc with $$L_{\rm Edd}\sim 1.3\times10^{39}\,\mathrm{erg\,s^{-1}}$$4 and $$L_{\rm Edd}\sim 1.3\times10^{39}\,\mathrm{erg\,s^{-1}}$$5 keV, or an optically thick Comptonising corona with $$L_{\rm Edd}\sim 1.3\times10^{39}\,\mathrm{erg\,s^{-1}}$$6 keV and $$L_{\rm Edd}\sim 1.3\times10^{39}\,\mathrm{erg\,s^{-1}}$$7, rather than by a simple hard-state power law. The source is therefore interpreted as a stellar-origin black hole accreting at super-Eddington rates, plausibly a $$L_{\rm Edd}\sim 1.3\times10^{39}\,\mathrm{erg\,s^{-1}}$$8 black hole radiating at $$L_{\rm Edd}\sim 1.3\times10^{39}\,\mathrm{erg\,s^{-1}}$$9 times the Eddington limit (Pintore et al., 2016).

Binary-evolution plus multiwavelength SED modelling leads to similar masses for several ULXs. Tracks computed for case-A Roche-lobe overflow from donors up to $\dot m \equiv \frac{\dot M}{\dot M_{\rm Edd}},$0 onto black holes up to $\dot m \equiv \frac{\dot M}{\dot M_{\rm Edd}},$1 show that NGC4559 X-7, NGC 5204 X-1, Holmberg II X-1 and NGC 5907 ULX-2 are consistent with accretion onto black holes in the range $\dot m \equiv \frac{\dot M}{\dot M_{\rm Edd}},$2, with optical counterparts often dominated by the accretion disc rather than the donor star (Ambrosi et al., 2021).

6. Observational diagnostics, evolutionary consequences, and open problems

Super-accreting X-ray binaries are identified through a combination of luminosity, spectral curvature, variability, absorption, and orbital modulation. In supergiant systems, high $\dot m \equiv \frac{\dot M}{\dot M_{\rm Edd}},$3, strong orbital-phase dependence, eclipses, off-states, and quasi-periodic modulations trace the structure of the wind and accretion wake (Manousakis et al., 2013). In transient systems, flare durations, recurrence, pulse periods, and column-density variations discriminate between clumpy-wind accretion, magnetospheric gating, and equatorial outflow scenarios (Sidoli, 2011). In ULXs, coherent pulsations, cyclotron features, broadband curvature, soft excesses, high-energy tails, and large nebulae or radio bubbles are all signatures of supercritical accretion, outflows, and, in some cases, neutron-star accretors (Dage et al., 2024).

Long-timescale timing provides an additional structural diagnostic. Across ULX pulsars, disc-fed supergiant HMXBs, and Be/X-ray binaries, the super-orbital and orbital periods follow a nearly linear relation,

$\dot m \equiv \frac{\dot M}{\dot M_{\rm Edd}},$4

with $\dot m \equiv \frac{\dot M}{\dot M_{\rm Edd}},$5 for the combined ULX + Roche-lobe-filling SGXB + BeXRB sample excluding wind-fed SGXBs. This has been interpreted as the orbital modulation of precessing hot spots or density waves in an accretion or circumstellar disc, and it places ULX pulsars on a phenomenological continuum with other HMXB pulsars (Townsend et al., 2020).

The evolutionary consequences are substantial. Supergiant HMXBs are likely progenitors of double neutron star systems and are connected to common-envelope evolution and short/hard gamma-ray bursts (Chaty, 2015). Super-Eddington accretion in neutron-star X-ray binaries can, in detailed binary-evolution models, drive the accretor to the maximum neutron-star mass and produce mass-gap black holes through accretion-induced collapse; the outcome depends strongly on the neutron-star magnetic field, donor metallicity, and the binary bifurcation period (Gao et al., 2022). A plausible implication is that some short-period black-hole X-ray binaries may be descendants of super-accreting neutron-star systems rather than direct core-collapse remnants.

At still larger scales, the integrated luminosity and outflow power of these systems matter beyond binary astrophysics. The intrinsic collective HMXB spectrum per unit star-formation rate provides a reference spectrum for calculations of X-ray preheating of the Universe by the first generations of X-ray binaries (Sazonov et al., 2017). Galactic super-accreting X-ray binaries with kinetic luminosities exceeding $\dot m \equiv \frac{\dot M}{\dot M_{\rm Edd}},$6 have also been proposed as Super-PeVatrons: in that picture, trans-relativistic jets and winds can accelerate protons to energies above several PeV and produce ultra-high-energy $\dot m \equiv \frac{\dot M}{\dot M_{\rm Edd}},$7-ray halos and large-scale nebulae (Wang et al., 28 Jul 2025).

Several controversies remain open. One concerns the relative roles of clumpy winds, gating, transient discs, and self-organized criticality in wind-fed systems (Sidoli, 2011, Manousakis et al., 2013). Another is whether the slow winds inferred for obscured sgHMXBs are intrinsic to the donor or largely a product of X-ray feedback (Manousakis et al., 2013). In ULXs, the proportion of neutron-star versus black-hole accretors, the degree of geometric beaming, and the role of magnetar-strength fields are still unsettled (Dage et al., 2024, Erkut et al., 2018). These uncertainties do not alter the central conclusion: across HMXBs, UCXBs, and ULXs, super-accreting X-ray binaries are systems in which extreme mass transfer drives the compact object, its disc or magnetosphere, and the donor environment into a coupled, non-linear regime that cannot be described by sub-Eddington accretion physics alone.