Particle-Accelerating Colliding-Wind Binaries (PACWBs)

• PACWBs are massive star systems with colliding supersonic winds that form shock regions capable of accelerating particles via diffusive shock acceleration.
• Observations using radio interferometry and spectropolarimetry reveal complex, phase-dependent wind-collision structures often obscured by free-free absorption.
• Advanced MHD simulations highlight that turbulence, magnetic field amplification, and reconnection govern the efficiency of particle acceleration in these binaries.

Particle-accelerating colliding-wind binaries (PACWBs) are massive non-degenerate stars in binary or higher multiplicity systems whose winds collide and form a wind-wind interaction region bounded by two shocks and a contact discontinuity. In the PACWB subset, those shocks produce a population of relativistic particles, most conspicuously revealed by non-thermal radio synchrotron emission, and in a few cases by hard X-rays or $\gamma$ rays. The designation displaced the older expression “non-thermal radio emitter” because radio synchrotron can be hidden by free-free absorption even when particle acceleration is present (Becker et al., 2013).

1. Definition, census, and astrophysical scope

The defining ingredients of a PACWB are a multiple system of massive stars, colliding supersonic winds, and observational evidence for relativistic particles. In the 2013 catalogue of De Becker and Raucq, PACWBs were defined as “massive non-degenerate stars in binary or higher multiplicity systems, leading to a wind-wind interaction region characterized by two shocks likely responsible for the acceleration of particles through the diffusive shock acceleration (DSA) mechanism,” and the catalogue contained 43 systems (Becker et al., 2013). That compilation emphasized O-type, Wolf-Rayet, Of/WN-transition, and LBV-like systems, with orbital periods from days to decades and wind kinetic powers spanning roughly $10^{34}$ to $10^{37}\,\mathrm{erg\,s^{-1}}$ (Becker et al., 2013).

Multiplicity is central to the class. In the 2013 catalogue, at least 37 out of 43 systems, or $\ge 85\%$, were confirmed or strongly suspected binaries or multiples, a much larger fraction than the quoted binary fraction for the general massive-star population (Becker et al., 2013). Subsequent work discussed “about 40 objects,” then “more than 50 systems are currently catalogued,” and an 18-year Fermi-LAT population study used 61 confirmed PACWBs in list A, showing that the census has expanded and been revised as multiplicity constraints and radio diagnostics improved (Becker et al., 2017, Blanco et al., 21 May 2026, Son, 6 Jul 2026).

The known systems do not occupy a narrow stellar subtype. The catalogue and later reviews explicitly state that PACWBs span essentially all O and WR subtypes, with no clear privileged subclass, while the deficit of B-type systems and late O-type main-sequence systems is attributed to observational bias rather than a demonstrated physical cutoff in acceleration capability (Becker et al., 2017). This suggests that particle acceleration in colliding winds is not an anomaly confined to a few exotic binaries, but an outcome that becomes observable only under favorable combinations of wind power, geometry, separation, opacity, and instrumental sensitivity.

2. Wind-collision structure and orbital control

In a colliding-wind binary, each star launches a radiatively driven, supersonic wind. Their interaction produces two shocks and a contact discontinuity, and the fundamental geometric parameter is the wind-momentum ratio

$$\eta = \frac{\dot{M}_2 V_{\infty,2}}{\dot{M}_1 V_{\infty,1}}.$$

This ratio fixes the location of the stagnation point and the opening angle of the interaction region: for $\eta \ll 1$ the shock cone wraps around the weaker-wind star, whereas $\eta \approx 1$ yields a more symmetric structure (Becker et al., 2013).

Whether a given wind-collision region is adiabatic or radiative depends primarily on separation and wind parameters. In wide systems, the winds typically reach terminal velocity before collision, shocks are strong, and the post-shock flow can remain largely adiabatic. In short-period systems, the collision region may lie deep inside dense winds, pre-shock velocities may be below $v_\infty$, and free-free opacity can obscure the synchrotron source even if acceleration is operating (Becker et al., 2017). WR 133 illustrates the compact extreme: it is a 112.4-day WN5+O9I system, yet the WR radio photosphere at 5.5 and 9 GHz is much larger than the orbit, so synchrotron from the inner wind-collision region is expected to be heavily absorbed (Becker et al., 2019).

Resolved imaging of HD 93129A provides a direct geometric benchmark. The system comprises HD 93129Aa, an O2 If$^+$ primary, and HD 93129Ab, an O3.5 V secondary, at a distance of about $2.5\,\mathrm{kpc}$. Long-baseline radio imaging at 2.3 GHz revealed an extended arc-shaped or “comma-shaped” non-thermal source between the stars, exactly where a wind-collision region is expected; the arc is curved around Ab, consistent with the weaker wind belonging to the secondary (Benaglia et al., 2015). From the resolved geometry the ratio $10^{34}$0 implies $10^{34}$1, while the opening half-angle $10^{34}$2 gives $10^{34}$3, and analytic contact-discontinuity shapes with $10^{34}$4–0.6 reproduce the observed curvature (Benaglia et al., 2015). The same study used HST/FGS astrometry to conclude that Aa and Ab are gravitationally bound, with a preliminary orbit suggesting $10^{34}$5 and $10^{34}$6 (Benaglia et al., 2015).

Orbital motion modulates both intrinsic shock conditions and observability. In hierarchical triples such as HD 167971, the wide wind-collision region can dominate the synchrotron emission while a close inner colliding pair dominates much of the soft thermal X-ray budget (Becker, 2015). In long-period eccentric WC+O systems such as WR 125, multi-year infrared and X-ray variations are tied to periastron passages, with the paper proposing an orbital period of 28–29 years from recurrence of dust formation and X-ray low states (Arora et al., 2021). PACWB phenomenology is therefore intrinsically phase dependent.

3. Particle acceleration, turbulence, and magnetic fields

The standard acceleration picture in PACWBs is DSA at the two shocks. Reviews and catalog papers consistently treat the shocks as the site where electrons, and likely hadrons, are accelerated to relativistic energies, yielding synchrotron radio emission and, in favorable conditions, inverse-Compton or hadronic high-energy emission (Becker et al., 2013, Becker et al., 2017). Modern semi-analytic modelling has made that standard picture more specific by including global shock obliquity, finite scattering-centre speed, magnetic-field amplification, and back-reaction. In a model applied to WR 146, the Mach number, Alfvénic Mach number, and transverse field strength vary strongly along the shocks, so acceleration efficiency varies non-linearly across the arc, and roughly 30 per cent of the wind power at the shocks is channelled into non-thermal particles (Pittard et al., 2021).

At the same time, multidimensional MHD work has shown that the shocked layer is not generally a smooth laminar structure. Three-dimensional simulations with radiative cooling find that nonlinear thin-layer instability and related instabilities disrupt the contact discontinuity, generate strong turbulence, and produce magnetic-field amplification much larger than one-dimensional Rankine-Hugoniot expectations (Falceta-Goncalves, 2015). In that framework, first-order Fermi acceleration may be inefficient in some systems, while second-order Fermi acceleration in turbulence and acceleration in turbulent reconnection regions can become dominant; for the $10^{34}$7 Car-like setup examined there, the paper quotes $10^{34}$8–10 TeV and argues that stochastic processes are likely primary (Falceta-Goncalves, 2015).

Test-particle calculations coupled to 3D MHD reinforce the importance of turbulence and magnetic complexity. In one such study, shocks in colliding-wind binaries produced a non-thermal population amounting to nearly 1% of the total particles, reaching from a few tens of GeV up to a few TeV depending on magnetization, with the fastest acceleration loci associated with turbulent plasma and amplified magnetic field (Kowal et al., 2021). A later high-resolution MHD plus test-particle analysis found that for moderate magnetization, $10^{34}$9, colliding-wind binaries can accelerate hadrons to hundreds of TeV or even PeV energies, with more than 1% of particles reaching the very-high-energy range, and explicitly concluded that turbulence and magnetic-field complexity dominate the acceleration while classical DSA plays a limited role (Cordeiro et al., 20 Aug 2025).

The magnetic field is thus indispensable but not necessarily as a strong ordered stellar dipole. A spectropolarimetric survey of nine PACWBs found no magnetic signature, all longitudinal field measurements compatible with 0 G, and for several stars derived upper limits on polar field strengths of order 200 G, excluding strong or moderate stellar magnetic fields typical of magnetic massive stars while still allowing weak fields (Neiner et al., 2014). Dedicated MHD calculations nonetheless show that if stellar polar fields are of order 50–200 G, magnetic confinement can modify the wind-collision structure, shock obliquity, and instability development significantly (Kissmann et al., 2016). A plausible implication is that PACWB acceleration requires magnetization, but not necessarily strong globally ordered surface fields.

4. Radiative signatures and observational diagnostics

Radio observations remain the principal identification method. The standard spectral convention is

$10^{37}\,\mathrm{erg\,s^{-1}}$0

For steady, spherically symmetric thermal winds, $10^{37}\,\mathrm{erg\,s^{-1}}$1, while synchrotron emission from relativistic electrons typically has a negative spectral index; mixed spectra can be flatter or curved (Becker et al., 2013, Becker et al., 2019). Additional radio indicators include high brightness temperature, typically $10^{37}\,\mathrm{erg\,s^{-1}}$2–$10^{37}\,\mathrm{erg\,s^{-1}}$3 for non-thermal emitters versus $10^{37}\,\mathrm{erg\,s^{-1}}$4 for pure thermal winds, flux variability correlated with orbital phase, and, in a small number of systems, VLBI localization of the emission to the wind-collision region (Becker et al., 2013).

High-angular-resolution imaging has become especially decisive. In 2015 only five wind-collision regions had been resolved at radio frequencies on milliarcsecond scales (Benaglia et al., 2015). Subsequent European VLBI Network observations at 1.67 GHz imaged HD 167971 and resolved HD 168112 for the first time. For HD 167971, the measured $10^{37}\,\mathrm{erg\,s^{-1}}$5 flux density was $10^{37}\,\mathrm{erg\,s^{-1}}$6, whereas the expected summed thermal wind contribution was only about $10^{37}\,\mathrm{erg\,s^{-1}}$7; for HD 168112 the observed flux was $10^{37}\,\mathrm{erg\,s^{-1}}$8, compared with a thermal estimate of about $10^{37}\,\mathrm{erg\,s^{-1}}$9, and the source was elongated over $\ge 85\%$0, consistent with a colliding-wind origin (Becker et al., 2024). These observations were used explicitly to argue that snapshot VLBI is an efficient means of confirming PACWB status.

Free-free absorption is a recurrent complication. In WR 133, a non-thermal or composite spectrum reported in 1993 contrasts with purely thermal spectra in 2007 and 2014–2015; the paper concludes that the simple short-period WR+O binary cannot account for the earlier non-thermal detection because the inner wind-collision region is deeply buried in the WR wind, and it proposes a triple-system scenario with a wider outer orbit as a consistent alternative (Becker et al., 2019). In WR 147, a recent phenomenological study compared a classical foreground free-free absorption model with an internal free-free absorption model and found that the foreground-only description fails at low frequencies, while internal free-free absorption reproduces the spectral energy distribution down to 610 MHz; the 150 MHz upper limit suggests two turnovers, indicating that both internal and foreground absorption shape different parts of the spectrum (Tasseroul et al., 14 Aug 2025).

Linear polarization, expected for synchrotron emission in an ordered field, has also turned out to be elusive. VLA observations of WR 147 and HD 167971 in L and C bands detected no polarization in Stokes $\ge 85\%$1 or $\ge 85\%$2, even in narrow high-frequency sub-bands intended to mitigate bandpass depolarization; the most conservative upper limit on the polarization degree is of order 1% for both targets (Blanco et al., 21 May 2026). The paper attributes the suppression to turbulent magnetic fields, Faraday depolarization, beam depolarization due to unresolved geometry, and thermal dilution from unpolarized free-free emission (Blanco et al., 21 May 2026). PACWBs therefore differ from idealized synchrotron sources in being composite, absorbing, and geometrically complex.

X-rays and $\ge 85\%$3 rays probe the same shocks through different channels. Soft X-rays are usually thermal emission from the hot shocked plasma, and long-term XMM-Newton monitoring of HD 168112 and HD 167971 showed that particle acceleration may remain efficient even when a wide-orbit colliding-wind region contributes little to the total thermal X-ray luminosity (Becker, 2015). Inverse-Compton emission scales against synchrotron as $\ge 85\%$4, so the balance between magnetic and photon energy densities is fundamental for high-energy expectations (Becker et al., 2013, Romero, 2019). Among PACWBs, $\ge 85\%$5 Car is the prototype high-energy case: it is the first binary system detected in high-energy $\ge 85\%$6 rays without a compact object, with modelling requiring a few percent of the shock mechanical power in relativistic particles and permitting proton energies up to $\ge 85\%$7 near periastron (Balbo et al., 2019).

5. Representative systems and phenomenological diversity

System	Configuration	Key lesson
HD 93129A	O2 If$\ge 85\%$8 + O3.5 V	Resolved arc-shaped wind-collision region; very massive O-star PACWB (Benaglia et al., 2015)
WR 147	WN8 + early B/O companion	Bright synchrotron source shaped by internal and foreground free-free absorption; no detected linear polarization (Tasseroul et al., 14 Aug 2025)
HD 167971 / HD 168112	Hierarchical triple / O+O candidate binary	Snapshot VLBI confirms or resolves synchrotron regions at mas scales (Becker et al., 2024)
WR 133	WN5 + O9I	Epoch dependence and free-free absorption can mimic absence of particle acceleration (Becker et al., 2019)
$\ge 85\%$9 Car	LBV + O/WR-like companion	Prototype $$\eta = \frac{\dot{M}_2 V_{\infty,2}}{\dot{M}_1 V_{\infty,1}}.$$0-ray PACWB and hadronic/leptonic benchmark (Balbo et al., 2019)

HD 93129A is a prototypical very massive O-star PACWB because it combines gravitationally bound binarity, a steep negative radio spectrum, and a spatially resolved radio arc between the stars (Benaglia et al., 2015). The source is especially important because the resolved morphology ties the non-thermal emission directly to the stagnation region rather than to either stellar wind separately. That removes much of the ambiguity that persists in unresolved systems.

WR 147, by contrast, is the archetype of a wide WR PACWB in which the synchrotron source is bright but radiative-transfer effects are intricate. The system is unresolved by the VLA in polarization experiments, yielding sub-percent polarization upper limits, while broad-band spectral modelling shows that low-frequency behaviour cannot be explained by foreground free-free absorption alone (Blanco et al., 21 May 2026, Tasseroul et al., 14 Aug 2025). WR 147 therefore demonstrates that the absence of observed polarization or a simple negative spectral index does not imply the absence of relativistic electrons.

HD 167971 and HD 168112 exemplify the value of VLBI in O-star PACWBs. HD 167971 is a hierarchical triple with a 3.32-day inner O+O pair and a $$\eta = \frac{\dot{M}_2 V_{\infty,2}}{\dot{M}_1 V_{\infty,1}}.$$1-year outer orbit, and its wide collision region dominates the observed synchrotron output (Becker et al., 2024). HD 168112, long recognized as non-thermal from its radio variability and spectral indices, had its synchrotron region resolved for the first time at 1.67 GHz as a slightly elongated source (Becker et al., 2024). These systems also show that strong radio synchrotron emission need not correlate with equally conspicuous thermal X-ray signatures from the same wind-collision zone (Becker, 2015).

WR 133 provides an instructive counterexample to a persistent misconception. PACWB status is not equivalent to a permanently negative radio spectral index. New JVLA observations over two 112.4-day orbits yielded constant, thermal radio emission, yet earlier data showed a non-thermal or composite spectrum; the paper interprets this as evidence that synchrotron identification is strongly epoch dependent and likely affected by multiplicity and line-of-sight opacity (Becker et al., 2019). The observational boundary between CWB and PACWB is therefore partly a function of when, where, and at what frequency the system is observed.

Finally, $$\eta = \frac{\dot{M}_2 V_{\infty,2}}{\dot{M}_1 V_{\infty,1}}.$$2 Car remains singular in the current class. Its extreme winds, highly eccentric 2024-day orbit, and secure GeV-to-TeV detection establish that colliding stellar winds alone can produce a genuine $$\eta = \frac{\dot{M}_2 V_{\infty,2}}{\dot{M}_1 V_{\infty,1}}.$$3-ray binary (Balbo et al., 2019). At the same time, its extremity warns against generalizing its high-energy efficiency to the broader PACWB population.

6. Selection effects, population limits, and evolving boundaries

PACWB identification is strongly biased by opacity, angular resolution, cadence, and wavelength. Radio synchrotron is the most efficient discovery channel, but the same dense winds that power strong shocks also produce free-free absorption that can hide the non-thermal component, especially in short-period WR systems (Becker et al., 2017, Becker et al., 2019). This is why catalog papers stress that the known sample almost certainly underrepresents weaker-wind O and B systems and under-samples heavily absorbed binaries (Becker et al., 2017).

Population-level high-energy results sharpen that point. An 18-year Fermi-LAT stacking analysis of 61 confirmed PACWBs removed systems coincident with bright catalogue $$\eta = \frac{\dot{M}_2 V_{\infty,2}}{\dot{M}_1 V_{\infty,1}}.$$4-ray sources and obtained a clean, mutually isolated sample of six objects. Their stack was consistent with control fields, with $$\eta = \frac{\dot{M}_2 V_{\infty,2}}{\dot{M}_1 V_{\infty,1}}.$$5, giving no evidence for collective GeV emission; the resulting 95% upper limit on the mean per-source flux was

$$\eta = \frac{\dot{M}_2 V_{\infty,2}}{\dot{M}_1 V_{\infty,1}}.$$6

implying a $$\eta = \frac{\dot{M}_2 V_{\infty,2}}{\dot{M}_1 V_{\infty,1}}.$$7-ray production efficiency

$$\eta = \frac{\dot{M}_2 V_{\infty,2}}{\dot{M}_1 V_{\infty,1}}.$$8

about two orders of magnitude below $$\eta = \frac{\dot{M}_2 V_{\infty,2}}{\dot{M}_1 V_{\infty,1}}.$$9 Car at the sample median distance (Son, 6 Jul 2026). The same paper showed that retaining PACWBs whose lines of sight coincide with bright catalogue sources creates a spurious $\eta \ll 1$0 excess, making source confusion in the Galactic plane a central methodological caveat (Son, 6 Jul 2026).

Several active problems follow directly from these results. One is the sparse inventory of spatially resolved wind-collision regions, despite the clear success of mas-scale radio imaging for systems such as HD 93129A, HD 167971, and HD 168112 (Benaglia et al., 2015, Becker et al., 2024). Another is the unsettled microphysics of acceleration: DSA remains the default language of the field, but multidimensional MHD work increasingly emphasizes turbulence, magnetic amplification, and reconnection (Falceta-Goncalves, 2015, Cordeiro et al., 20 Aug 2025). A third is the relation between stellar magnetic fields and accelerator performance, given the non-detection of strong organized fields in spectropolarimetry and the equally clear requirement for magnetized shocks (Neiner et al., 2014).

A later conceptual extension widened the PACWB framework beyond the classical “massive non-degenerate stars only” definition by proposing supercritical colliding-wind binaries, in which a super-Eddington black-hole disk wind collides with the wind of an early-type companion and produces radio-to-GeV non-thermal emission at luminosities of $\eta \ll 1$1–$\eta \ll 1$2 (Abaroa et al., 2023). Whatever terminology is adopted for that extension, the classical PACWB problem remains: to determine under which combinations of wind power, separation, magnetization, and opacity a colliding-wind system becomes an observable particle accelerator. The current literature indicates that the answer is controlled less by a single canonical mechanism than by the coupled dynamics of shocks, turbulence, radiation fields, and line-of-sight transfer.