X-RAY EMISSION FROM NITROGEN-TYPE WOLF–RAYET STARS

The Chandra observations are summarized in Table 2. Exposures were obtained using the ACIS-S (Advanced CCD Imaging Spectrometer) imaging array in faint timed-event mode. A reduced 1/4 subarray size was used for the brighter source WR 134 to minimize any photon pileup. The WR star target was placed at the nominal aim point on CCD S3 of the array. The CCD pixel size is 0492. Approximately 90% of the encircled energy at 1.49 keV lies within 2'' of the center pixel of an on-axis point source. More information on Chandra and its instrumentation can be found in the Chandra Proposer's Observatory Guide (POG).⁷

The Level 2 events file provided by the Chandra X-ray Center (CXC) was analyzed using standard science threads in CIAO version 4.1.2.⁸ The CIAO processing used calibration data from CALDB version 4.1.2. Source detection was carried out using the CIAO wavdetect tool, which correlates the input image with "Mexican Hat" wavelet functions of different scale sizes. We ran wavdetect on full-resolution images with a pixel size of 0492 using events in the 0.3–8 keV range to reduce the background. The wavdetect threshold was set at sigthresh = 1.5 × 10⁻⁵ and scale sizes of 1, 2, 4, 8, and 16 were used. The wavdetect tool provides source centroid positions and net source counts (background subtracted) inside the computed 3σ source region (Table 3).

CIAO specextract was used to extract source and background spectra for the WN stars along with source-specific response matrix files (RMFs) and auxiliary response files (ARFs). We used the 3σ source ellipses from wavdetect to define the extraction regions and background was extracted from adjacent source-free regions. Background is negligible, contributing <1 count (0.3–8 keV) inside the source extraction regions over the duration of each observation. Spectral fitting was undertaken with the HEASOFT Xanadu⁹ software package including XSPEC vers. 12.4.0. X-ray light curves were extracted using the CIAO tool dmextract and checks for source variability were carried out on energy-filtered source event files using the Kolmogorov–Smirnov (K–S) test (Press et al. 1992) and the Bayesian-method CIAO tool glvary (Gregory & Loredo 1992, 1996).

Table 2 summarizes the XMM-Newton observation of WR 79a. Data were acquired with the European Photon Imaging Camera (EPIC), which provides CCD imaging spectroscopy from the pn camera (Strüder et al. 2001) and two nearly identical MOS cameras (MOS1 and MOS2; Turner et al. 2001). The observation was obtained in full-window mode using the thick optical blocking filter. The EPIC cameras provide energy coverage in the range E ≈ 0.2–15 keV with energy resolution E/ΔE ≈ 20–50. The MOS cameras provide the best on-axis angular resolution with FWHM ≈43 at 1.5 keV.

Data were reduced using the XMM-Newton Science Analysis System (SAS vers. 7.1) using the latest calibration data. Event files based on pipeline processing carried out at the XMM-Newton Science Operations Center were filtered to select good event patterns. Further filtering based on event energies was done to reduce background emission. No time filtering was required.

Spectra and light curves were extracted from circular regions of radius R_e = 15'' centered on WR79a, corresponding to ≈68% encircled energy at 1.5 keV. This approach reduces the number of background counts while capturing most of the source counts. Total source counts (Table 3) were measured in a larger circular region of radius r = 45'' corresponding to ≈90% encircled energy at 1.5 keV. Background spectra and light curves were obtained from circular source-free regions near the source. The SAS tasks rmfgen and arfgen were used to generate source-specific RMFs and ARFs for spectral analysis. The data were analyzed using the HEASOFT Xanadu software package.

We have examined Chandra and ROSAT archive data for selected WN stars which are not known to be binaries. Our archive analysis is not exhaustive and was restricted to a few high-interest objects for which either good CCD spectra or useful upper limits were obtained. Analysis of Chandra archive data was carried out using the same procedures described above (Section 2.1). ROSAT images were analyzed using the Ximage image analysis tool, which is part of the HEASOFT Xanadu package. Results of the archive analysis are summarized in Section 3.3 and in Tables 3 and 5.

Table 3 summarizes the X-ray properties of the four newly detected WN stars and three WN stars from the Chandra archive (WR 20b, WR 24, and WR 136). There is very good agreement between the X-ray positions and the optical positions of the WN stars given in the HST Guide Star Catalog (GSC). Radial offsets between the X-ray and optical positions are ⩽03 for the new detections, being within the 1σ range of Chandra¹⁰ and XMM-Newton¹¹ positional accuracy. Similar offsets were obtained for the archive detections. There is thus no reason to doubt that the X-ray sources are the optically identified WR stars.

No significant X-ray variability was detected in any of the targets, but sampling in the time domain spans less than ≈10 hr for the four new detections and ≲1 day for the archive sources. Thus, variability on longer timescales exceeding one day cannot yet be ruled out.

The X-ray spectra of the four new detections are shown in Figure 5 and the archive spectra in Figure 6. All spectra show some absorption below ≈0.8 keV as well as prominent emission lines or line complexes indicative of thermal emission. The most prominent lines are the unresolved Mg xi He-like triplet at 1.33–1.35 keV (maximum line power temperature log T_max = 6.8 K) and the unresolved Si xiii triplet at 1.84–1.86 keV (log T_max = 7.0 K). The higher temperature S xv line at 2.46 keV (log T_max = 7.2 K) is also seen in all stars, but is very faint in WR 2. The high-temperature line from He-like Ar xvii (log T_max = 7.3 K) is detected in several stars and is quite strong in WR 20b. Faint emission near 3.9 keV in the WN6h stars WR 20b and WR 136 may be due to He-like Ca xix (log T_max = 7.5 K), but we classify this line only as a possible detection. The Fe Kα complex (Fe xxv; 6.67 keV log T_max = 7.6 K) is detected in WR 20b, WR 134, and WR 136. This line is a signature of very hot plasma, which is undoubtedly present in these latter three WN6h stars.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We have attempted to fit the spectra with various emission models, including the XSPEC variable-abundance optically thin plasma model vapec (Smith et al. 2001) and the constant temperature plane-parallel shock model vpshock (Borkowski et al. 2001, and references therein). All models included a photoelectric absorption component (XSPEC model wabs), which was used to determine the equivalent neutral H column density N_H.

Element abundances are expected to deviate strongly from solar composition in WN stars as the result of advanced nucleosynthesis. H is depleted along with C and O, while He and N are enriched (van der Hucht et al. 1986; hereafter VCW86). Because of the moderate CCD spectral resolution of ACIS-S and EPIC and limited total source counts, we are not able to place tight constraints on element abundances using the X-ray spectra. We have thus adopted the generic WN abundances of VCW86 (Table 4) as starting values in spectral fits using the vapec and vpshock models. Selected elements were allowed to deviate from the VCW86 values to improve the fits, as detailed in Tables 4 and 5. The total X-ray absorption likely includes contributions from solar abundance material in the interstellar medium and stellar wind absorption from material enriched with heavy elements. The CCD spectra do not provide sufficient information to disentangle solar and nonsolar absorption, and we have thus modeled X-ray absorption using solar abundances (XSPEC wabs; Anders & Ebihara 1982).

Our main spectral analysis results can be summarized as follows: (1) single-temperature optically thin plasma models (1T vapec) do not provide acceptable fits, even if abundances are allowed to vary, (2) two-temperature models (2T vapec) are acceptable but the goodness-of-fit statistic (χ²) is sensitive to abundances, (3) all stars show both a cool (kT₁ < 1 keV) and hot (kT > 2 keV) plasma component, (4) most stars show X-ray absorption in excess of that expected from their optical extinction A_V, and (5) fits using isothermal plane-parallel shock models are not as good as 2T vapec (WR 20b excepted).

3.2.1. Optically Thin Plasma Models (vapec)

3.2.2. Plane-Parallel Shock Models (vpshock)

WR 2 (WN2): WR 2 is a new Chandra detection and is the earliest WN subtype in our sample. The V = 12.73 mag object (HST GSC J010522.98+602505.1) lying 139 south of WR 2 was not detected by Chandra. WR 2 lies in a neutral hydrogen void (Arnal et al. 1999). No significant spectroscopic variability was found for WR 2 in the recent study of St.-Louis et al. (2009). However, unusual round-shaped emission line profiles have been reported for WR 2 that were interpreted as signaling rapid rotation near breakup (Hamann et al. 2006).

As Table 3 shows, it is the softest X-ray source of the detected WN stars as judged by its median photon energy E₅₀ = 1.04 keV. Spectral analysis confirms that the X-ray absorption toward WR 2 is less than the other three stars, resulting in the detection of more soft photons below 0.7 keV and a lower median photon energy. Events were detected in WR 2 down to energies of ≈0.3 keV as well emission at higher energies up to ≈3 keV.

The Chandra spectrum is interesting because it shows an emission feature near 0.5 keV that likely includes a contribution from the low-temperature hydrogen-like N vii Ly α line (log T_max = 6.3 K). The intermediate temperature Ne x line at 1.02 keV (log T_max = 6.7 K) is also detected, as well as faint emission from the higher temperature Si xiii and S xv lines. Thus, cool and hot plasma features are present in the same star. Two issues related to 2T vapec fits are worthy of note. This model is able to reproduce most of the broad emission feature near 0.5 keV with the N vii Lyα line (E_lab = 500.3 eV). But, there is some residual at slightly higher energies of 530–540 eV that is not reproduced by this line. The O vii resonance line is not obviously detected and in any case it lies at a higher energy (574 eV) and cannot account for this residual. However, other ions with lab energies slightly above that of N vii could be contributing to the residual including N vi (532.6 eV), S xiv (538.9 eV), and Ca xvi (546.8 eV). The 2T vapec model also has difficulty reproducing all of the flux in the faint feature near 2.46 keV (S xv) unless a very high S abundance is assumed.

WR 16 (WN8h): this star is of interest because of its later WN8h subtype and moderate extinction A_V = 1.8 mag (vdH01). WR 16 was not detected in ROSAT PSPC image rp200715n00, which has an exposure livetime of 7465 s. Our analysis of the PSPC data gives a 3σ upper limit on the total band count rate ⩽1.8 counts ksec⁻¹. The PSPC is sensitive at lower energies of ≈0.15–2.5 keV. This equates to a conservative 3σ upper limit log L_X(0.3–8 keV) ⩽ 31.7 ergs s⁻¹, assuming a distance d = 2.37 kpc (vdH01). The conversion from PSPC rate to unabsorbed flux and L_X was carried out using the Portable Interactive Multi-Mission Simulator (PIMMS)¹³ assuming N_H = 4 × 10²¹ cm⁻² corresponding to A_V = 1.8 mag (Gorenstein 1975) and a Raymond–Smith thermal plasma model with emission measure equally distributed between a cool component at kT₁ = 0.6 keV and a hotter component at kT₂ = 3.0 keV.

WR 18 (WN4): this WN4 star is associated with the asymmetric nebula NGC 3199, which has been modeled as a wind-blown bubble by Dyson & Ghanbari (1989). Cold circumstellar molecular gas has been reported by Marston (2001, 2003). It is a new Chandra detection and its spectrum shows a prominent Si xiii line and the hotter S xv line.

WR 20b (WN6h): this star was detected as a bright source in a previous Chandra observation of the massive stellar cluster Westerlund 2 (ObsId 6411; Nazé et al. 2008). Its X-ray emission is heavily absorbed and exceptionally hard with a median photon energy E₅₀ = 3.04 keV. We were unable to obtain acceptable fits with a 1T vapec model but a 2T vapec model is satisfactory and requires a hot plasma component at kT₂ ≈ 4–5 keV. The isothermal plane-parallel shock model vpshock provides a slightly better fit with an inferred shock temperature kT_shock = 3.9 keV. The X-ray absorption is greater than for any other WN star in our sample. The absorption determined from both vapec and vpshock models yields a ratio of X-ray to optically determined absorption N_H/N_H,vis ≈ 3. The ACIS-I spectrum shows high-temperature emission lines including Ar xvii (3.13 keV) and Fe xxv (6.67 keV). The distance to WR 20b is quite uncertain. Values range from d = 2.27 kpc (vdH01) up to 8 kpc (Nazé et al. 2008). Even at the low end of this range, its X-ray luminosity is very high for a WN star (log L_X > 33.5 ergs s⁻¹). Its high L_X and hard spectrum are reminiscent of colliding wind binaries, suggesting that this object may not be a single star (Section 4.6).

WR 24 (WN6ha): this star was detected as a bright off-axis X-ray source in an archival Chandra guaranteed time observation (GTO) of the Carina region (GTO ObsId 9482). Its ACIS-I spectrum shows a remarkably strong S xv emission line at 2.46 keV, as well as the high-temperature Ar xvii (3.13 keV) line and possible faint Fe line emission in the 6.4–6.7 keV range. A 2T vapec model provides a satisfactory fit with a hot plasma component at kT₂ ≈ 4 keV. The absorption N_H = 0.76 [0.44–1.05] × 10²² cm⁻² is several times larger than would be expected from A_V = 0.56 mag (vdH01). At an assumed distance d = 3.24 kpc (vdH01), a high unabsorbed X-ray luminosity log L_X = 33.0 ergs s⁻¹ is inferred, only slightly less than for WR 20b. Because of its high L_X and high L_bol, WR 24 is another candidate unresolved binary system (Section 4.6).

WR 78 (WN7h): this WNL star is similar to WR 16, having moderate extinction A_V = 1.5 mag and is relatively nearby at d = 1.99 kpc (vdH01). A previous tabulation of WN star X-ray observations by Oskinova (2005) listed WR 78 as a ROSAT PSPC detection. Our analysis of archival ROSAT PSPC images rp200716n00 does not confirm this. We report a non-detection in the total band and soft band images. Weak signal may be present in the hard band PSPC image, but the signal-to-noise ratio, S/N ≈ 2, is not sufficient to classify it as a detection. We obtain a 3σ upper limit on the total band PSPC count rate ⩽2.97 counts ksec⁻¹ based on an exposure livetime of 10,121 s. The corresponding 3σ luminosity upper limit from PIMMS is log L_X(0.3–8 keV) ⩽ 31.7 ergs s⁻¹ (at d = 1.99 kpc). This assumes N_H = 3.3 × 10²¹ cm⁻² corresponding to A_V = 1.5 mag (Gorenstein 1975) and a 2T Raymond–Smith thermal plasma model as for WR 16 above.

WR 79a (WN9ha): this is a new and important XMM-Newton detection, being the latest WN subtype in our sample and the first unambiguous X-ray detection of a WNL star in the WN7-9 range. There is a fainter near-IR source (2MASS J165458.09−410903.6) at an offset of 48 from WR 79a (see also Mason et al. 1998), but the X-ray positional accuracy is capable of distinguishing between this object and WR 79a. Radio continuum emission has been detected (Bieging et al. 1989) and follow-up observations suggest a nonthermal radio component may be present (Cappa et al. 2004). WR 79a (= HD 152408) was formerly classified as O8:Iafpe but was subsequently reclassified as WN9ha based on the analysis of its optical, UV, and IR spectra by Crowther & Bohannan (1997). Its EPIC spectrum reveals a strong Si xiii line and higher temperature S xv and Ar xvii lines. Since this star resembles an O8 supergiant, we have considered solar abundance spectral fits as well as the fit based on WN abundances given in Table 4. The solar abundance fit gives comparable temperatures to the WN abundance fit but yields a slightly higher N_H and L_X (Table 4 notes). Regardless of the reference abundances used, the L_X of WR 79a is at the low end of the range found here for WN stars.

WR 134 (WN6h): this is a new Chandra detection and a luminous X-ray source with a very hot component. It is surrounded by a ring nebula (Chu et al. 1983; Miller & Chu 1993). Because of its unusual 2.25 day periodic spectroscopic variability, it has been the subject of numerous previous studies (e.g., Howarth & Schmutz 1992; Schulte-Ladbeck et al. 1992; Morel et al. 1999; St.-Louis et al. 2009). The possibility that the 2.25 day periodicity might be due to a compact companion has been debated but there are difficulties with this interpretation (Morel et al. 1999). We discuss the compact companion hypothesis further in Section 4.4.

WR 136 (WN6h): WR 136 is surrounded by the ring nebula NGC 6888 (Gruendl et al. 2000) which is a source of diffuse X-ray emission (Bochkarev 1988; Gruendl et al. 2003; Chu et al. 2006). The WR star was detected at ≈8' off-axis in a deep 93 ksec Chandra exposure of NGC 6888 (ObsId 3763). The ACIS-S spectrum is faint, but quite remarkable. The spectrum is dominated almost entirely by emission lines, including the high-temperature S xv line (2.46 keV), possible faint emission from Ca xix (3.9 keV), and a feature near 6.67 keV that is likely Fe XXV. Attempts to fit the spectrum with a 2T vapec model using a single absorption component were unstable and resulted in high runaway temperatures for the hot component. A modified 2T vapec model that allows for different absorptions toward the cool and hot components is stable and provides a good fit (Table 5). The absorption derived for the cool component is consistent with that expected for previous extinction estimates A_V = 1.73 ± 0.32 mag (vdH01). However, the best-fit absorption of the hot component is a factor of ≈8 larger. The physical picture that emerges from such a model is that the cool X-ray emission originates far from the star and incurs little or no wind absorption. In contrast, the hot component originates deep within the wind and much closer to the star and is subject to strong attenuation by the wind. The inferred X-ray luminosity is sensitive to the temperature of the cool component but most fits yield values of log L_X(0.3–8 keV) ≈ 31.5 ergs s⁻¹, assuming the Hipparcos distance d = 1.64 kpc. Thus, WR 136 seems to be of lower X-ray luminosity than other WN6 stars in the sample.

The visual extinction A_V for the WN stars in our sample is dominated by interstellar material along the line of sight. There is insufficient dust in the winds of WN stars to significantly affect A_V, in contrast to some WC stars whose dusty winds can absorb and re-radiate stellar light (van der Hucht et al. 1987). The A_V values in Table 1 can be used to calculate an equivalent neutral H absorption column density along the line of sight using the conversion N_H,vis = 2.22 × 10²¹A_V cm⁻² (Gorenstein 1975). Assuming that the winds of WN stars have a negligible effect on A_V, the resulting value of N_H,vis is a measure of interstellar absorption. In Tables 4 and 5, we give the ratio of the X-ray to optically derived absorption column densities N_H/N_H,vis. This ratio is greater than unity for all WN stars except WR 2. That is, the X-ray absorption is greater than expected based on visual extinction estimates. The largest absorption excesses are found for WR 20b and WR 134. For the double-absorption model of WR 136, there is no significant absorption excess for the cool component but a large excess is found for the heavily absorbed hot component.

Most of the excess X-ray absorption likely occurs in the metal-rich WR wind, but in some cases cold circumstellar molecular gas may also contribute. It may not be a coincidence that WR 2 has little or no excess absorption given that it has the highest effective temperature of the four newly detected stars (T_* = 141 kK; Hamann et al. 2006) and the lowest wind density parameter $\dot{M}$/v_∞. Higher temperatures (along with higher ionization) plus lower densities would increase wind transparency to X-rays.

If the X-rays are formed in the wind then some wind absorption is expected, especially for lower energy photons. For an ideal spherical wind, the radius of optical depth unity R_τ=1 beyond which X-ray photons can escape (the "exospheric approximation") depends on several factors including photon energy, mass-loss parameters, photoelectric absorption cross sections, and wind abundances (e.g., Owocki & Cohen 1999). In Figure 7, we plot R_τ=1 as a function of photon energy for the four new detections. We have assumed an ideal spherical, homogeneous, constant-velocity wind with the mass-loss parameters as given in Table 1, canonical WN abundances (VCWH86), and the photoelectric absorption cross sections of Balucińska-Church & McCammon (1992, hereafter BM92). It is apparent that under the above assumptions, soft photons with E < 1 keV emerge at large distances from the star R_τ=1 ≳ 1000 R_☉. Taking WR 134 as a specific example, the softest photons visible in its spectrum (Figure 5) at E ≈ 600 eV have R_τ=1(600 eV) = 9262 R_☉ ≈ 1750 R_*, assuming R_* = 5.29 R_☉ (Hamann et al. 2006).

Zoom In Zoom Out Reset image size

The above estimates, which are based on the assumption that the wind is homogeneous and spherically symmetric, suggest that the softest X-ray photons detected in the WN star spectra emerge at large distances of thousands of stellar radii or more from the star. These are only rough estimates because wind abundances and mass-loss parameters are poorly known. Furthermore, there is accumulating evidence that the winds of WR stars are clumped rather than homogeneous (Moffat et al. 1988; Robert 1994; Lépine et al. 2000), in which case the X-rays could escape from smaller radii closer to the star. Even if the wind is the dominant absorber, other factors such as cold dense circumstellar gas could also contribute. Such gas has been detected around WR 18 and may be a vestige of mass loss in previous evolutionary phases (Marston 2001, 2003).

The radiative wind shock picture cannot explain the high-temperature plasma seen in the X-ray spectra of WN stars, but the temperature of the cooler plasma kT₁ ≈ 0.3–0.6 keV is compatible with radiative wind shocks. Of the four stars observed, WR 2 shows the coolest emission. Assuming its cool plasma originates in wind shocks, we equate the best-fit temperature of the cool component with the shock temperature, that is kT₁ = 0.32 [0.24–0.45] keV = kT_s. For an adiabatic shock, the maximum shock temperature is kT_s = (3/16)$\bar{\rm m}$v²_s, where v_s is the shock velocity perpendicular to the shock front (Luo et al. 1990). For a helium-rich WN wind $\bar{\rm m}$ = (4/3)m_p where m_p is the proton mass. An equivalent form of this relation that is easier to apply is kT_s = 2.61[v_s/1000 km s⁻¹]² keV.

The inferred shock speed for WR 2 is then v_s = 350 [304–416, 90% confidence range] km s⁻¹. This value is plausible, being only about one-fifth the terminal wind speed (Table 1) and within the range of shock velocity jumps found in numerical simulations of radiatively driven winds in O-type stars (Owocki et al. 1988; Feldmeier et al. 1997). The slightly higher values kT₁ ≈ 0.6 keV for the other three stars give shock speeds v_s ≈ 480 km s⁻¹.

The unabsorbed flux of the cool X-ray component in WR 2 (Table 4) gives a cool-component luminosity log L_X,1 = 32.47 (ergs s⁻¹) and log (L_X,1/L_wind) = −4.94. So there is sufficient wind kinetic energy to power the soft X-ray emission even at a rather low conversion efficiency from kinetic to radiative energy. But, numerical models show that forward shocks decay with radius (Feldmeier et al. 1997) and it remains to be demonstrated by detailed simulations that X-ray emitting wind shocks can persist as coherent structures in WR winds out to large distances of hundreds to thousands of stellar radii (Section 4.1).

We now consider another star, WR 20b, which is an interesting case because of the good fit obtained with an isothermal plane-parallel shock model. The best-fit shock temperature kT_s = 3.86 [2.97–6.53] keV is too high to explain by radiative wind shocks, but is plausible for colliding wind shocks. If the wind of WR 20b is shocking onto an unseen companion, then the inferred shock velocity perpendicular to the interface is v_s = 1216 [1067–1582] km s⁻¹. This estimate is for an adiabatic shock and assumes a pure He wind for the WN star. The above value is reasonable, being comparable to the terminal wind speeds of WN6 stars (Crowther 2007; Hamann et al. 2006).

There are at present no detections of magnetic fields in WR stars to our knowledge. But, if magnetic fields of sufficient strength exist then they could confine the ionized wind into two oppositely directed streams which collide near the magnetic equator and form a magnetically confined wind shock (MCWS). This mechanism is of interest because it is capable of producing hot X-ray plasma over a range of temperatures up to several keV (Babel & Montmerle 1997) and could potentially explain the hotter plasma seen in the WN star spectra.

A rough estimate of the field strength needed to confine the wind can be obtained using the confinement parameter

$$\begin{equation*} \eta = B_{\rm{eq}}^2R_{*}^2\big/\dot{M}v_{\infty }, \end{equation*}$$

where $B_{\rm{eq}}$ is the field strength at the magnetic equator (ud-Doula & Owocki 2002). For η ≈ 1, the wind is marginally confined by the B-field; for η ≪ 1, the wind overpowers the field and is not confined; and for η ≫ 1, the wind is strongly confined. Assuming η ≈ 1 (marginal confinement) and adopting the mass-loss parameters and stellar radii in Table 1, one obtains magnetic field strengths in the range $B_{\rm{eq}}$ ≈ 3.5 kG (WR 134) up to 17 kG (WR 2). These fields are for marginal confinement and larger fields would be needed for strong confinement. Without concrete evidence for strong B fields in WR stars, this mechanism must be considered speculative.

Our four new detections and archive sources were selected on the basis that they are so far not known to be close (spectroscopic) binary systems. Thus, any attempt to explain their X-ray emission by invoking an unseen companion is not strongly motivated on observational grounds. However, it is worth keeping in mind that close companions at subarcsecond spacings are difficult to detect, especially when in low-inclination orbits.

Of the WN stars in our sample, WR 134 has been regarded as a possible close binary system because of its known 2.25 day periodic line profile changes. This issue has been previously discussed by Morel et al. (1999). However, binarity has not yet been demonstrated and a rotationally modulated wind is an alternative explanation for the variability (St.-Louis et al. 2009). A close companion, if present, could potentially explain the hard X-ray emission in WR 134. Hard X-rays could be produced by the WR wind accreting onto a compact object or by the wind shocking onto the surface of a normal (nondegenerate) stellar companion. But this interpretation encounters some difficulties.

The most compelling example known so far of a WR star accreting onto a compact object is Cyg X-3. It has a high X-ray luminosity L_X ∼ 10³⁸ ergs s⁻¹ and the companion is thought to be either a black hole or a neutron star (van den Heuvel & De Loore 1973; Schmutz et al. 1996; Lommen et al. 2005). If the putative companion of WR 134 were a neutron star, then the predicted X-ray luminosity from wind accretion is log L_X,acc ∼ 10³⁷ ergs s⁻¹ (Morel et al. 1999). This exceeds the observed value for WR 134 (Table 4) by more than 4 orders of magnitude. This interpretation is thus untenable unless some mechanism is operating to inhibit wind accretion (see Davidson & Ostriker 1973 or Lipunov 1982 for details on such mechanisms). A similar luminosity mismatch has also been noted for the WN star WR 6 (EZ CMa) by Skinner et al. (2002b). It also undergoes periodic optical variability (P = 3.76 day; Firmani et al. 1980) of unknown origin and a neutron star companion has been suggested as a possible explanation. In the case of WR 134, there are other obstacles to overcome if the putative companion is a neutron star of typical mass M_ns ≈ 1.4 M_☉. If P = 2.25 day is interpreted as an orbital period, then Kepler's third law gives a separation a ≈ 20 R_☉≈ 4 R_*. Here, we have adopted $M_{\rm{WR}}$ = 19 M_☉ and radius R_* = 5.3 R_☉ for WR 134 (Hamann et al. 2006). Unless the wind is inhomogenous, X-rays at energies E ≈ 2–3 keV (obviously present in the spectrum) could not escape from such small radii. Specifically, if we use the mass-loss parameters for WR 134 in Table 1 and canonical WN abundances (VCW86), then R_τ=1(2 keV) ≈ 89 R_*. The slightly different mass-loss parameters for WN6 stars in Crowther (2007) give R_τ=1(2 keV) ≈ 36 R_*.

Given the luminosity mismatch obtained in the neutron star companion model, it seems more likely that any unseen companion is a normal (nondegenerate) star. It would have been difficult for a massive companion (M_comp ≈ $M_{\rm{WR}}$) to have escaped detection, but a lower mass star ($M_{\rm comp} \ll M_{\rm{WR}}$) would be more difficult to detect. The possibility of a low-mass companion was previously considered in our analysis of the similar WN-type star WR 6 (Skinner et al. 2002b). Likewise, for WR 134 it can be shown that the hard-component X-ray luminosity could be accounted for by the WR wind impacting the surface of a lower mass star of subsolar radius (Usov 1992). Furthermore, the plasma temperature of the hot component of WR 134 is compatible with maximum values expected from a shocked wind at or near the terminal wind speed of WR 134. Specifically, v_∞ ≈ 1700–1960 km s⁻¹ (St.-Louis et al. 2009) gives kT_s ≈ 7.5–10. keV. However, such a low-mass companion would be located very close to WR 134 if P = 2.25 day is the orbital period. Separations of <5 R_*(WR) are inferred from Kepler's third law, so again the escape of X-rays through the wind is potentially problematic. This problem is mitigated if the wind is clumped or if an unseen companion is present at a larger separation than inferred above, resulting in a lower WR wind density at or near the companion surface.

The sample of single WN stars for which X-ray parameters have been reliably determined from good-quality CCD spectra is not yet large enough to establish any clear correlations between stellar and X-ray properties. Other factors such as uncertain distances, bolometric luminosities, and mass-loss parameters hinder correlation studies and can mask dependencies.

However, X-ray observations now span the full range of spectral subtypes from WN2 to WN9 and a few notable trends are beginning to emerge. As Figure 8 shows, there is an apparent falloff in L_X toward later WN7-9 subtypes. The WN9ha star WR 79a has the lowest L_X of any detected single WN star so far. And, the WN8h star WR 40 was undetected in a previous XMM-Newton observation at an upper limit log L_X < 31.6 ergs s⁻¹ (Gosset et al. 2005). This upper limit is a factor of 3.5 below the L_X determined from the detection of WR 79a. Figure 8 also includes upper limits for WR 16 (WN8h) and WR 78 (WN7h) based on short-exposure ROSAT PSPC archive images. Since ROSAT was not sensitive to hotter plasma above ≈2.5 keV, which is likely present, the PSPC upper limits should be considered tentative until observations spanning a broader X-ray bandpass can be obtained.

Zoom In Zoom Out Reset image size

The apparent decrease in L_X for WN7-9 subtypes is surprising because it is in the opposite sense of L_bol, which is larger for WN7-9 stars than WN2-6 stars (Table 1; Hamann et al. 2006; Crowther 2007). We plot L_X versus L_bol in Figure 9. This figure clearly illustrates that even though the WN and WC stars observed to date have similar L_bol, the WC stars (all undetected) are at least 1–2 orders of magnitude less luminous in X-rays. Considering only the detected WN stars, one might be tempted to argue for a weak increase in L_X with L_bol. But, the absence of X-ray detections for three WNL stars with high L_bol (WR 16, 40, 78) is clearly at odds with this picture. This result is contrary to the trend seen in OB stars, for which L_X generally increases with L_bol, albeit with large scatter (Berghöffer et al. 1997). Thus, some other stellar parameters besides L_bol are critical for determining X-ray luminosity levels in these WN stars, and wind properties may play a key role.

Zoom In Zoom Out Reset image size

The classical theory of hot star winds in the exospheric approximation predicts that L_X ∝ ($\dot{M}$/v_∞)^1+s where s = 1 for optically thin winds and −1 < s < 0 for optically thick winds (Owocki & Cohen 1999). The existing WN star data do not show any clear increase in L_X with $\dot{M}$/v_∞. On the contrary, WR 79a is the least luminous X-ray source amongst the new detections, but has the largest $\dot{M}$/v_∞.

From the standpoint of observations, an increase in L_X with wind kinetic energy would seem to be more promising. In Figure 10, we plot L_X versus L_wind = (1/2)$\dot{M}$v²_∞ for several WN stars for which good X-ray CCD spectra are available. Despite uncertainties in $\dot{M}$, v_∞, and distances, there is a noticeable increase in L_X toward higher values of L_wind for the stars considered in this study.

Zoom In Zoom Out Reset image size

A notable outlier is WR 136, which is underluminous in X-rays by at least a factor of ∼5 compared to the other WN6 stars (Figure 8) and also has a low L_X for its wind luminosity (Figure 10). We have assumed d = 1.64 kpc for WR 136 based on its Hipparcos parallax. Other distance estimates are less (e.g., d = 1.26 kpc; vdH01), so a distance underestimate is not likely to be responsible for its lower L_X. One way to correct for the low L_X is to postulate very cool undetected plasma (kT < 0.2 keV) that is masked by absorption. If such plasma is present, then values log L_X > 32.0 ergs s⁻¹ are possible.

Equally interesting is the non-detection of WR 40 in X-rays. This is an important result because it has a slow terminal wind speed v_∞ = 650–840 km s⁻¹ (Hamann et al. 2006; Gosset et al. 2005) and thus a high average wind density. This may be a clue that X-ray emission ceases to be efficient (or becomes totally absorbed by the wind) in WN stars with slow dense winds.

X-ray observations of a larger sample of single WN stars spanning a broader range of mass-loss parameters are needed to firmly establish any dependencies between L_X and wind parameters. Observations of more WNL stars with slow winds (v_∞ < 1000 km s⁻¹) would be particularly useful.

Previous X-ray studies have shown that several WR+OB or WR+WR binaries have very high X-ray luminosities with typical values log L_X ≳ 33.0 ergs s⁻¹. Early evidence for this was found from Einstein observations of WR stars. The median X-ray luminosity of seven WN binaries detected by Einstein was log L_X(0.2–4 keV) = 32.92 ergs s⁻¹, with a range of 32.76–33.53 ergs s⁻¹ (Pollock 1987). More recent work has confirmed high X-ray luminosities in other WN binary systems. For example, XMM-Newton observations of the WN8+OB system WR 147 gave log L_X(0.5–10 keV) = 32.83 ergs s⁻¹ (Skinner et al. 2007). A Chandra observation of the massive WN binary system WR 20a in Westerlund 2 yielded log L_X = 33.69 ergs s⁻¹ based on 1T spectral fits (Nazé et al. 2008). Our analysis of the WR 20a Chandra data with 2T ${\it vapec}$ models gives even higher values log L_X(0.3–8 keV) > 34.0 ergs s⁻¹ for d > 2.37 kpc. The closely spaced WN5+O6 binary V444 Cyg (= WR 139) is also worthy of mention because of its short ≈4.2 day orbital period. ASCA observations show that its X-ray luminosity varies with orbital phase, reaching maximum values log L_X(0.7–10 keV) > 33.1 ergs s⁻¹ for an assumed distance of 1.7 kpc (Maeda et al. 1999).

Thus, there is no doubt that some WR binaries are X-ray luminous. But, are all WR binaries X-ray luminous? We do not yet know the answer to that question because it can be very difficult to distinguish a closely spaced binary from a single WR star. It is very likely that some WR stars currently classified as single objects are in fact close unresolved binaries. Also, a high X-ray luminosity is not strictly required in a colliding wind WR binary system. For an adiabatic colliding wind system, the luminosity scales roughly as L_X ∝ $\dot{M}^2$v^−3.2D⁻¹ where $\dot{M}$ and v are the mass-loss rate and wind-speed of the dominant component and D is the binary separation (Luo et al. 1990; Stevens et al. 1992). Binaries with lower mass-loss rates or wide separations are expected to have lower L_X, all other factors being equal.

Based on the above, which WN stars in our sample are good binary candidates? WR 24 would appear to be a very good candidate because of absorption lines in its optical spectrum (vdh01) and its high L_X and L_bol. In fact, its position in the (L_bol, L_X) diagram is almost identical to that of the known binary WR 147 (Figure 9). WR 20b is also a good candidate because of its exceptionally high X-ray luminosity, even when computed using the lowest current distance estimate of d = 2.37 kpc. Searches for close companions in these two stars are certainly justified.

Apart from WR 20b and WR 24, there are no other stars in our sample having high log L_X ≳ 33.0 ergs s⁻¹ that could signal binarity. A particularly interesting case is WR 79a (WN9ha). This star would seem to be a good candidate for binarity based on absorption lines in its spectrum and the possible presence of nonthermal radio emission (Cappa et al. 2004). But its X-ray luminosity log L_X = 32.14 ergs s⁻¹ is at the low end of the detected WN stars. If WR 79a is indeed a binary then its L_X is much less than other known WN binaries, or its distance has been underestimated.

WR 134 is another intriguing example. Based on its known 2.25 day spectroscopic variability, it would seem to rank as a strong binary candidate. As noted in Section 4.4, the high temperature of its hot X-ray component is indeed consistent with that expected if the wind of WR 134 were shocking onto a companion at terminal speed. But there are questions as to whether the X-rays could escape if produced in a closely spaced colliding wind binary system, and the X-ray luminosity of WR 134 (log L_X = 32.66 ergs s⁻¹) is not as high as other known WN binaries. Thus, the existing X-ray data for WR 134 do not provide an indisputable case for binarity. Further observations of WR 134 aimed at determining whether its X-ray emission is modulated at the 2.25 day optical period would be useful since colliding wind binaries such as γ² Velorum and WR 140 do show orbital X-ray modulation (Willis et al. 1995; Zhekov & Skinner 2000; Pollock et al. 2005).

Our pilot survey has so far detected all four WN stars observed. In contrast, X-ray observations at similar sensitivity levels did not detect any of the four WC stars in our sample, and a fifth went undetected in another XMM-Newton program (Section 1). Why is it that most WN stars are detected as X-ray sources, but WC stars are not?

An important clue to solving this puzzle may lie in the excess X-ray absorption seen in the spectra of most of the WN stars (Section 4.1). Such excess absorption has been seen in other WN stars such as WR 6 (WN4; Skinner et al. 2002b). The wind is likely responsible for most of this excess. X-rays produced within the wind must overcome the wind opacity to escape, and the opacity increases toward lower photon energies, all other factors being equal. Thus, softer X-ray photons will be more heavily absorbed and the softest photons detected will have emerged from the wind at large distances from the star.

The wind opacity in WC stars is much higher than in WN stars due to metal-enrichment. For a given set of mass-loss parameters, X-rays of a given energy will have more difficulty escaping a WC wind than a WN wind. As a specific example, we consider a generic WR star with a typical mass-loss rate of $\dot{M}$ = 2 × 10⁻⁵ M_☉ yr⁻¹ and v_∞ = 1800 km s⁻¹. Adopting canonical WN and WC abundances (VCW86) and BM92 cross sections, the radius of optical depth unity for an X-ray photon of energy E = 1 keV is ∼13 times greater for a WC star than for a WN star. Even at higher energies E = 3 keV, the R_τ=1 ratio is still a factor of ∼11. Thus, unless X-rays of moderate energy E ≈ 1–3 keV are formed very far out in the wind of a WC star (at thousands of solar radii) they will be absorbed. This is perhaps the most plausible explanation for the absence of WC star X-ray detections at present. But a larger sample of WC stars needs to be observed in order to determine if hard X-rays might escape through the wind in some cases. The best candidates for such observations would be WC stars with higher T_eff and lower wind densities (lower $\dot{M}$ and higher v_∞), which would conspire to reduce wind opacity.