Wd1-9: Rare Massive Binary in Westerlund 1

• Wd1-9 is a rare massive binary in Westerlund 1, defined by strong emission lines, a dusty torus, and colliding-wind driven mass loss.
• Multiwavelength observations reveal a dense circumstellar environment with quasi-Keplerian disc features, ionized outflows, and extreme X-ray emission.
• Its 14-day orbital period and rapid mass transfer indicate a brief evolutionary phase transitioning from pre-interaction to a colliding-wind WR+OB system.

Wd1-9, also designated W9 and Ara C, is a highly obscured massive emission-line object in the Galactic cluster Westerlund 1 that has historically been classified as a supergiant B[e] star. In the classical observational sense, that classification reflects strong Balmer emission, permitted low-ionization metal emission, forbidden lines including [Fe II] and [O I], and a strong infrared excess from circumstellar dust. The modern picture is more specific: multiwavelength work has progressively reinterpreted Wd1-9 as a short-period, interacting massive binary embedded in dense circumbinary material, with extreme mass loss, a dusty torus, a slow dense ionized outflow, and hard thermal X-ray emission characteristic of colliding winds (Clark et al., 2013, Fenech et al., 2016, Anastasopoulou et al., 23 Jul 2025).

1. Classification and place within Westerlund 1

Wd1-9 lies within the massive stellar population of Westerlund 1 and is not treated as an isolated field B[e] object with unconstrained age or luminosity. Earlier work placed it in a coeval post-main-sequence cluster environment with characteristic initial masses of order $\sim 35$–$40\,M_\odot$, while later studies emphasized that both the age and distance of Westerlund 1 remain actively debated (Clark et al., 2013, Beasor et al., 2021, Castellanos et al., 27 Feb 2026).

The source is exceptionally rare in Galactic terms. It has been described as one of only two confirmed cluster sgB[e] stars in the Milky Way, and its rarity has been connected to the brevity of the relevant evolutionary phase (Anastasopoulou et al., 23 Jul 2025). Earlier census-based discussion similarly noted only two confirmed cluster or association sgB[e] stars among $\sim 68$ young massive Galactic clusters and associations containing $\sim 600$ post-main-sequence massive stars, with Wd1-9 as the sole sgB[e] representative in Westerlund 1 (Clark et al., 2013).

The classification itself has changed in meaning. Older studies used “sgB[e]” as a phenomenological description tied to the optical and infrared appearance, whereas the 2025 EWOCS X-ray analysis argued that the label is probably not the fundamental physical identity of the star. In that interpretation, Wd1-9 is “most likely classified as sgB[e] due to the optical emission originating entirely from its dense circumbinary material,” so the designation is largely phenomenological rather than a direct statement about an unobscured stellar photosphere (Anastasopoulou et al., 23 Jul 2025).

2. Optical and infrared circumstellar structure

The optical and near-infrared spectrum of Wd1-9 is dominated by circumstellar emission and shows no stellar photospheric absorption features. Across roughly $5800$–$9300$ Å, the spectrum contains very strong hydrogen and He I emission, extremely strong O I $8446$ Å, numerous permitted lines of N I, O I, Mg I, Mg II, Fe II, Ca II, and Si II, and forbidden lines of [O I], [O II], [N II], [Fe II], [Ni II], [Ni III], [S III], and [Ar III]. The only absorption features are interstellar Na I, K I, diffuse interstellar bands, and telluric bands. This directly indicates that the central source is veiled by circumstellar material rather than directly exposed (Clark et al., 2013).

Hydrogen and helium lines demonstrate both high density and kinematic complexity. H$\alpha$ has ${\rm EW}({\rm H}\alpha) = -640 \pm 40~{\rm \AA}$, a relatively narrow core with ${\rm FWHM}\sim 125~{\rm km\,s^{-1}}$, and electron-scattering wings extending to about $40\,M_\odot$0. The He I lines have broad bases extending to at least $40\,M_\odot$1. Permitted Ca II and [Ni II] show clear double peaks with separation $40\,M_\odot$2 and $40\,M_\odot$3, respectively, consistent with a quasi-Keplerian disc or torus. At the same time, the broader He I bases and higher-excitation forbidden lines indicate an additional outflowing or shocked component (Clark et al., 2013).

Line diagnostics were interpreted as requiring multiple physically distinct zones: a cool dense hydrogen-neutral region, a hotter dense transition zone with Ly$40\,M_\odot$4-pumped Fe II and Ly$40\,M_\odot$5-pumped O I $40\,M_\odot$6, a hot H II region producing H and He lines, lower-density forbidden-line regions, and a diffuse post-shock component responsible for the highest-excitation fine-structure lines such as [O IV]. The ratio

$40\,M_\odot$7

was specifically used to infer Ly$40\,M_\odot$8 fluorescence in a dense, predominantly neutral transition zone (Clark et al., 2013).

Mid-infrared spectroscopy and SED modeling established an oxygen-rich dusty environment. An illustrative continuum fit used a $40\,M_\odot$9 kK stellar source plus dust components at approximately $\sim 68$0 and $\sim 68$1. A flared-disc model yielded $\sim 68$2, $\sim 68$3, $\sim 68$4, and $\sim 68$5; with a gas-to-dust ratio of 100, the corresponding torus mass is $\sim 68$6 (Clark et al., 2013).

Variability in the line spectrum was comparatively limited before the deep X-ray timing work. No convincing reflex radial-velocity motion was detected in the strong optical lines, and the absence of such motion was attributed to the fact that the line-forming region does not reveal the central binary directly because the stars are almost completely veiled by circumstellar material (Clark et al., 2013).

3. Radio and millimetre properties

Radio observations first established Wd1-9 as the outstanding radio source in Westerlund 1. At 8.6 GHz it has $\sim 68$7, making it the brightest radio source in the cluster and one of the most luminous radio stars known. The emission is spatially resolved into a compact source surrounded by extended emission. The compact component has $\sim 68$8, consistent with thermal emission from a stellar wind, while the extended region has $\sim 68$9, consistent with optically thin thermal emission from an earlier mass-loss episode (Dougherty et al., 2010).

That two-component morphology led to explicit two-epoch mass-loss estimates. For the current wind,

$\sim 600$0

and for the older shell-like component,

$\sim 600$1

Those rates are about an order of magnitude above the $\sim 600$2 values inferred for most other transitional massive stars in the cluster, and they were taken as direct evidence for enhanced envelope loss (Dougherty et al., 2010).

ALMA 3 mm observations refined the geometry and kinematics. At 100 GHz, Wd1-9 resolves into a bright compact core and lower-brightness extended emission. A 2D Gaussian fit to the core gives a deconvolved size of

$\sim 600$3

with position angle $\sim 600$4 and integrated flux density $\sim 600$5. The in-band ALMA spectral index at the peak is $\sim 600$6, interpreted as thermal free-free emission transitioning from partially optically thick at centimetre wavelengths to more optically thin at millimetre wavelengths. The extended emission is strongly non-spherical, oriented almost north-south, and spans $\sim 600$7 at 5 kpc (Fenech et al., 2016).

The same ALMA dataset yielded serendipitous H41$\sim 600$8 radio recombination-line emission at 92.060 GHz. The line is spatially unresolved and coincident with the compact core, with $\sim 600$9, centroid velocity $5800$0, and a centroid gradient of about $5800$1 along $5800$2, roughly perpendicular to the north-south extended continuum structure. These measurements were interpreted as consistent with a disc wind by analogy with MWC349A (Fenech et al., 2016).

Under spherical assumptions, the millimetre work inferred

$5800$3

with $5800$4. If a similar aspherical geometry to MWC349A is adopted, the rate becomes

$5800$5

These values, together with the compact core, extended aspherical nebula, and H41$5800$6 kinematics, were taken to indicate a compact dusty torus, a slow dense ionized wind, and a larger-scale ejection nebula from prior mass loss (Fenech et al., 2016).

4. X-ray phenomenology and orbital period

The deepest X-ray study to date transformed the interpretation of Wd1-9 from strongly suspected binary to system with a measured period. Using 44 Chandra observations—36 EWOCS ACIS-I datasets and 8 archival ACIS-S datasets—totalling about 1.1 Ms, Wd1-9 was identified as CXO J164704.1-455031 at RA 16:47:04.13, Dec $5800$7, and as the third brightest X-ray source in Westerlund 1, with $5800$8 net counts in the 0.5–8 keV band (Anastasopoulou et al., 23 Jul 2025).

Wd1-9 shows significant long-term variability, and a Lomb-Scargle analysis detected a strong periodic signal at $5800$9 d. The periodogram exceeded a false-alarm-probability threshold of 0.01%, and an independent String Length Method recovered the same solution. The X-ray study interpreted this $9300$0-day modulation as the orbital period, making it the first period determination for the system (Anastasopoulou et al., 23 Jul 2025).

The spectrum is thermal, hard, and line-rich. Two absorbed thermal models were used: $9300$1 and

$9300$2

Both require a soft component around $9300$3–$9300$4 keV and a hard component around $9300$5 keV. The most important line diagnostic is Fe XXV at $9300$6 keV, detected for the first time in Wd1-9, together with Si XIII, S XV, and Ar XVII. The spectrum also shows phase-dependent changes near the iron complex, including Fe K$9300$7 fluorescent emission near 6.4 keV around the hard-flux maximum (Anastasopoulou et al., 23 Jul 2025).

For the $9300$8 model, the fit gives $9300$9, $8446$0, $8446$1, $8446$2, and $8446$3. For the $8446$4 model, the fit gives $8446$5, $8446$6, $8446$7, and the same absorbed luminosity within uncertainties. The ISM-corrected luminosity lies at $8446$8, depending on model choice (Anastasopoulou et al., 23 Jul 2025).

This X-ray behavior is unlike that of a normal single sgB[e] star and was explicitly argued to be unlike a typical supergiant HMXB as well. Instead, the hard thermal spectrum, Fe XXV emission, circumbinary material, and 14-day periodic modulation were taken as the hallmarks of a colliding-wind binary, closely resembling the bright Wolf-Rayet binaries in Westerlund 1 (Anastasopoulou et al., 23 Jul 2025). A separate EWOCS study of the cluster WR population did not include Wd1-9 directly, but it reinforced the broader cluster-level association between hard thermal X-rays, Fe XXV emission, and colliding-wind binarity, so the analogy is physically specific rather than generic (Anastasopoulou et al., 2024).

5. Evolutionary interpretation

The evolutionary interpretation of Wd1-9 has converged on binary interaction, but the exact phase assignment has shifted with the data. The 2013 multiwavelength synthesis argued that the “most complete explanation” was a massive interacting binary currently undergoing, or having recently exited, rapid Roche-lobe overflow. In that picture, the radio-derived mass-loss rates of $8446$9, the dusty torus, and the thermal X-ray luminosity $\alpha$0 together pointed to rapid binary-mediated envelope stripping rather than a normal line-driven sgB[e] wind (Clark et al., 2013).

The ALMA study sharpened the analogy with MWC349A and concluded that Wd1-9 and MWC349A are interacting binaries evolving away from the main sequence and undergoing rapid case-A mass transfer. In that view, the compact torus, slow dense ionized wind, and older ejection nebula are products of systemic mass loss during an active transfer phase, and the wider class of sgB[e] stars may contain objects caught during similar episodes (Fenech et al., 2016).

The 2025 EWOCS X-ray analysis modified that evolutionary label while retaining the interacting-binary framework. It argued that Wd1-9 has “all the hallmarks of a binary—likely a WR+OB—that recently underwent early Case B mass transfer,” and suggested that the hidden donor may already be a newly formed WN star. Its binary toy models found that combinations involving only O/B supergiants generally fail to reproduce the observed $\alpha$1 keV, whereas more plausible solutions require a WN7 or WN8 primary with an underluminous hydrogen-rich OB mass gainer. A second WR component was considered unlikely at this stage (Anastasopoulou et al., 23 Jul 2025).

Taken together, these studies imply a strong continuity in the physical picture even where nomenclature differs. The stable elements are extreme mass loss, a compact torus or circumbinary envelope, colliding-wind X-ray emission, and recent non-conservative transfer. The main uncertainty is whether the relevant transfer episode is best described as rapid case-A or early Case B. A plausible implication is that Wd1-9 occupies a very short-lived transition linking pre-interaction O+O binaries to the Wolf-Rayet binary population of Westerlund 1 (Fenech et al., 2016, Anastasopoulou et al., 23 Jul 2025).

6. Cluster context, adopted parameters, and remaining uncertainties

Several global properties of Westerlund 1 remain non-unique across the literature, and those differences propagate into any quantitative interpretation of Wd1-9. Early Wd1-9 studies adopted $\alpha$2 kpc, and their linear scales and mass-loss expressions were explicitly written with $\alpha$3 scaling. Later EWOCS work adopted $\alpha$4 kpc, while a Gaia DR2 Bayesian analysis inferred a cluster parallax of $\alpha$5 mas, corresponding to $\alpha$6 kpc, and stated that 5 kpc is ruled out with 99% confidence (Fenech et al., 2016, Anastasopoulou et al., 23 Jul 2025, Aghakhanloo et al., 2019).

The cluster age is similarly contested. One reanalysis of the cool supergiants argued for $\alpha$7 Myr and proposed that Westerlund 1 cannot be described by a single-age paradigm, whereas infrared spectroscopy of the main sequence later yielded $\alpha$8 Myr and argued for moderate coevality and a single burst of star formation. This means that Wd1-9’s membership does not by itself fix a unique progenitor mass or timescale without specifying which cluster-age framework is adopted (Beasor et al., 2021, Castellanos et al., 27 Feb 2026).

Survey selection effects also matter. A VVVX/VIRAC2 study of the hidden intermediate-mass population mentioned Wd1-9 only in the bibliography, supplied no object-level photometric or variability measurements, and emphasized that the brightest central massive stars are saturated in that survey. This suggests that the absence of Wd1-9 from some NIR variability catalogues is not physically diagnostic in itself (Ordenes-Huanca et al., 26 Jun 2026).

The most stable conclusion across the direct Wd1-9 literature is therefore not a unique distance or age, but a morphological and physical one: Wd1-9 is a heavily obscured, line-dominated, dusty, radio-bright, X-ray-hard massive binary whose sgB[e] appearance is produced by dense circumstellar or circumbinary material rather than by an unobscured single supergiant. The present best-supported formulation is a short-period colliding-wind WR+OB-like system, embedded in the ejecta of recent rapid mass transfer, and observed at a rarely sampled stage of massive-binary evolution (Clark et al., 2013, Fenech et al., 2016, Anastasopoulou et al., 23 Jul 2025).