DISCOVERY OF X-RAY EMISSION FROM THE WOLF–RAYET STAR WR 142 OF OXYGEN SUBTYPE

Stars of the Wolf–Rayet (WR) type have a highly peculiar chemical composition and very strong stellar winds. The WR spectra are sorted into three subclasses: WN, WC, and WO, according to the dominance of nitrogen, carbon, and oxygen emission lines, respectively. These spectral types correspond to evolutionary stages (Conti et al. 1983). The largest C + O abundance and the fastest stellar winds are observed in WO-type stars that represent the final evolutionary stage of a massive star prior to its explosion as a type Ic supernova or γ-ray burst (Hirschi et al. 2005).

The WR stars continue to challenge our understanding of line-driven winds. Schaerer (1996) pointed out the importance of the iron opacity for the acceleration of WR winds. The first hydrodynamical model for a WR wind was presented by Gräfener & Hamann (2005). In their simulation, the mass loss is initiated at high optical depth by the so-called iron bump in the opacity. It was thus demonstrated that WR-type winds can be driven by radiation pressure.

It has long been known that line-driven winds are subject to an instability that can lead to strong shocks (Lucy & White 1980). These shocks are thought to explain the X-ray emissions from O star winds, as predicted by time-dependent hydrodynamic modeling (Owocki et al. 1988; Feldmeier et al. 1997) and largely confirmed by observations (Kramer et al. 2003; Oskinova et al. 2006; Zhekov & Palla 2007; Waldron & Cassinelli 2007). The growth of instability in WR winds was investigated by Gayley & Owocki (1995). They found that despite the damping effects due to the multiline scattering, the instability remains effective. Therefore, X-ray emission from wind shocks could, in principle, be expected in WR winds, a conjecture that has not been yet tested by time-dependent hydrodynamic simulations.

Significant observational effort has been made to study the X-ray emission of WR stars. White & Long (1986), Pollock (1987), Pollock et al. (1995), and Oskinova (2005) presented X-ray observations of Galactic WR stars. A survey of X-ray emission from WR stars in the Magellanic Clouds was conducted by Guerrero & Chu (2008a, 2008b). Ignace et al. (2000) and Oskinova (2005) demonstrated that X-ray properties of single WR stars differ from those of O stars. Whereas O stars display a trend in which the ratio of the X-ray to the bolometric luminosity L_X/L_bol has a typical value of 10⁻⁷ (Long & White 1980; Berghoefer et al. 1997; Sana et al. 2006), this trend is not observed in the case of WR stars.

Observations with the XMM-Newton and Chandra X-ray telescopes established that some bona fide single WN stars are X-ray active (Skinner et al. 2002a, 2002b; Ignace et al. 2003; Oskinova 2005), while others are apparently not (Oskinova 2005; Gosset et al. 2005). Oskinova et al. (2003) found that no single WC star had been conclusively detected at X-ray energies, a result that continues to hold. Among the WR subclasses, only the WO-type stars have not been observed in X-rays so far. In this Letter, we present the XMM-Newton observations of the closest WO-type star WR 142. The star is introduced in Section 2, and its XMM-Newton observations are described in Section 3. The implications of the XMM-Newton detection of WR 142 are discussed in Section 4.

WR 142, also named Sand 5 and St 3, has a spectrum characteristic for the spectral subtype WO2 (Barlow & Hummer 1982; Kingsburgh et al. 1995). The optical spectrum of WR 142 was discussed by Polcaro et al. (1997), who also noticed some line variability.

Figure 1 shows the spectral energy distribution (SED) of WR 142 together with the model calculated with the Potsdam Wolf–Rayet (PoWR) model atmosphere code (Gräfener et al. 2002). Photometric infrared (IR) measurements are plotted together with our optical spectrum and the Spitzer IRS mid-IR spectrum. We have adopted parameters typical for a WO star: stellar temperature T_* = 150 kK, "transformed radius" (see Gräfener et al. 2002) R_t = 2 R_☉, and a composition of 40% carbon, 30% oxygen, and 30% helium (by mass). WR 142 is assumed to be a member of the open cluster Berkeley 87 at a distance of d = 1.23 kpc (Turner et al. 2006). Based on this preliminary model, the fit of the photometric observations requires an interstellar reddening of E_B−V = 1.7 mag and a stellar luminosity of log L_bol/L_☉ = 5.35, implying a stellar radius of only R_* = 0.6 R_☉. The corresponding mass-loss rate is about 10^−5.1 M_☉ yr⁻¹ for an adopted microclumping volume filling factor of 0.1.

Zoom In Zoom Out Reset image size

The adopted model does not provide an entirely satisfactory fit to the line spectra. For example, the model does not match the huge O vi emission at 3811, 3834 Å, a problem also experienced by Crowther et al. (2000) while reproducing these lines for Sand 2 with the cmfgen code. We also fitted the SED shown in Figure 1 with a model that has a mass-loss rate lower by a factor of two, 10^−5.4 M_☉ yr⁻¹, and higher bolometric luminosity, log L_bol/L_☉ = 5.65. This model fits the SED in Figure 1 equally well. Figure 2 shows the radius of unity optical depth plotted as a function of wavelength in the X-ray range for both models. The stronger wind is opaque even to hard X-rays, but the thinner wind is largely transparent, because its higher ionization reduces the X-ray absorbing ions. The same could happen in denser, but hotter models. Thus, our ability to predict the influence of wind photoabsorption is somewhat limited, owing to ambiguities in the ionization state of metals and uncertainty in the mass-loss rate.

Zoom In Zoom Out Reset image size

The profiles of the emission lines in the spectrum of WR 142 are very broad. Assuming that the line widths correspond to the wind terminal velocity, the velocity of v_∞ ≈ 5500 km s⁻¹ would be deduced. However, the profile shapes of almost all lines are much more round than the roughly Gaussian shapes usually seen in WR spectra. Such round profiles cannot be reproduced by the standard models. It is tempting to reproduce these profiles by convolution with the semi-ellipse for rotational broadening, albeit rotating stellar winds certainly require a more sophisticated treatment which has not been accomplished yet. If rotation is the cause for the round profiles, the projected rotation speed must be comparable to v_∞, i.e., the star would rotate near to breakup with v_rotsin i ≲ 4000 km s⁻¹ (see the inset in Figure 1). Interestingly, a similar suggestion has been made for the hottest and the most compact Galactic WN star, WR 2 (T_* = 140 kK, R_* = 0.89 R_☉, v_rotsin i ≈ 2000 km s⁻¹; Hamann et al. 2006).

WR 142 was observed by XMM-Newton during two consecutive satellite orbits (ObsId 0550220101 and ObsId 0550220201). The data were merged and analyzed using the latest versions of software sas 8.0.0. After the high background level time intervals have been rejected, the combined exposure time of all detectors was ≈100 ks; an EPIC image of WR 142 and its surroundings is shown in Figure 3. WR 142 is detected with a 5σ confidence level in all XMM-Newton EPIC detectors using standard source-detection algorithms. We define a "hardness ratio (HR)" as HR = (N_hard − N_soft)(N_hard + N_soft)⁻¹, where N_soft is the number of counts in the 0.25–2.0 keV band and N_hard in the 2.0–12.0 keV band. For WR 142 we find HR ≈0.57. For comparison, the HR of the WN star WR 1, which has a reddening similar to WR 142, is HR(WR 1) ≈−0.9, while more reddened WR 110 has HR(WR 110) ≈−0.4 and less reddened WR 6 has HR(WR 6) ≈−0.6. Thus, the X-ray emission from WR 142 is the hardest among putatively single WR stars. The count rates in the 0.25 keV–12.0 keV band are (1.89 ± 0.34) × 10⁻³ counts s⁻¹ for EPIC MOS1+2 and (3.80 ± 0.84) × 10⁻³ counts s⁻¹ for EPIC PN cameras. Assuming a two-temperature thermal plasma model (kT₁ = 0.3 keV,  kT₂ = 10 keV), the observed X-ray flux of WR 142 is F_X = 4 ± 2 × 10⁻¹⁴ erg s⁻¹ cm⁻². The reddening toward WR 142 is known from the analysis of its optical spectrum, and the distance is known from its cluster membership. The X-ray luminosity of WR 142 is thus L_X ≈ 7 × 10³⁰ erg s⁻¹, or log L_X/L_bol ≈ −8.1.

Zoom In Zoom Out Reset image size

The angular resolution of XMM-Newton is ≲6''. To exclude the potential confusion with a source in close vicinity of WR 142, we inspected optical and IR images with higher angular resolution. According to the USNO-B1.0 catalog (Monet et al. 2003), the closest object to WR 142 is located 8'' away. The optical monitor (OM) on board of XMM-Newton provides images with an angular resolution of ≈1''. The OM images in the four filters did not detect any objects closer than 8'' around WR 142. The Spitzer Space Telescope (SST) has observed the Berkeley 87 cluster in the IR. With an angular resolution of ≲1'', the Spitzer Infrared Array Camera (IRAC) camera took images in the four channels. No point sources within 8'' from WR 142 were detected in any of the IRAC channels. It is extremely unlikely that any X-ray source is not detectable in either the optical or IR. We therefore rule out any serendipitous X-ray sources within ≈1'' of WR 142 that might get confused as our target.

In addition, there is no evidence to suggest that the observed X-rays originate from a wind-blown bubble around WR 142. Although theoretically expected, there is a dearth of WR stars with detected diffuse X-ray emission from their wind-blown bubbles (Chu et al. 2003; Wrigge et al. 2005). The only two detected hot bubbles show a limb-brightened morphology and are extended on the scale of parsecs. On the contrary, we confidently detect a point source at the position of WR 142.

The spectra of WR 142 were extracted from the 15'' region. The small number of counts obtained in our observation does not allow for quantitative spectral analysis, but interesting conclusions can be made about the gross energy distribution. Figure 4 shows the X-ray spectrum of WR 142 before a background subtraction, together with the spectrum of a nearby background region normalized to the same area. The emission from WR 142 dominates over the background at 2–7 Å (1.7–6 keV), while longward it dives below the background until about ≈20 Å (0.6 keV) where it rises above the background again. The presence of "a dip" between ≈ 7 and 12 Å (0.8–1 keV) is also seen in the background-subtracted PN and MOS-2 spectra shown in Figure 5. The appearance of the dip may be due to unresolved strong emission lines. However, the wavelength of the dip coincides well with the K-shell edges of oxygen (Verner & Yakovlev 1995), where our stellar wind models predict an absorption maximum (see Figure 2). Therefore, it is tempting to attribute the observed dip to the oxygen K-shell absorption—to confirm this identification, data of higher quality will be required. Assuming the presence of K-shell of oxygen, the thermal plasma model fit to our low signal-to-noise ratio (S/N) data indicates the presence of plasma with temperatures spanning from 1 MK to 100 MK.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The confident detection of weak hard X-rays from WR 142 prompts us to reconsider the previous nondetections of X-ray emission from WR stars. The X-ray luminosity of WR 142 at L_X ≈ 7 × 10³⁰ erg s⁻¹ is comparable to present upper limits for the nondetections (Oskinova et al. 2003; Oskinova 2005; Gosset et al. 2005; Skinner et al. 2006). Perhaps some WR stars that are as yet undetected in X-rays may be weak sources similar to WR 142. This raises the question, what is the mechanism responsible for the generation of X-rays in WR stars? Usual suspects are wind shocks and magnetic fields.

The strong shock condition predicts the peak temperature in the shocked gas, T_X, as

$$\begin{equation} kT_{\rm X}^{\rm max}=\frac{3}{16} \mu m_{\rm H} U^2, \end{equation} \tag{ 1 }$$

where U is the velocity with which the gas rams into the shock (i.e., preshock velocity relative to the shock), and μ is the mean molecular weight per particle in the postshock gas. As the shock temperatures are very high, abundant species will be nearly entirely ionized. Therefore, μ depends only weakly on chemical compositions. For example, $\mu \approx \frac{1}{2}$ for ionized H, while for fully ionized carbon, $\mu \approx \frac{2}{3}$. Thus, differences in wind molecular weight between O and WR stars cannot strongly affect the temperatures of the shocked plasma in stars of these types.

Note that Equation (1) gives an upper limit to the temperature at the shock. The radiative cooling will restrict the hottest temperatures to a limited part of the total emission measure. Moreover, some fraction of the energy will be consumed by ionization processes. To roughly estimate the energy required to ionize WO wind material in the post-shock zone we consider oxygen and assume that the leading ion in the pre-shock material is O vi, while in the post-shock plasma it is O ix. A WO wind may contain 30% oxygen. The ionization potentials are IP(O vi)  ≈ 0.14 keV, IP(O vii)  ≈ 0.74 keV, and IP(O viii)  ≈ 0.87 keV (Cox 2000). The specific energy for ionization is = ΣN_iIP_i, where N_i = X_i(A_im_H)⁻¹ with X_i being the mass fraction of the element, A_i its atomic weight, and m_H the atomic mass unit. Inserting these numbers, the full ionization of oxygen requires ≈ 3 × 10¹³ erg g⁻¹. This is small compared to the kinetic energy of the wind (5 × 10¹⁵ erg g⁻¹ for typical wind speed 1000 km s⁻¹). Thus the ionization is not a significant cooling process even in metal-rich winds.

In the time-dependent hydrodynamic simulations by Feldmeier et al. (1997) the velocity jump U depends on the ratio between the period of the perturbations at the wind base, T_c, and the flow time, T_flow = R_*/v_∞. The former was estimated by the acoustic cutoff period, T_c = aH⁻¹, where a is thermal speed and H is the pressure scale height. Assuming that the perturbations are seeded in the hydrostatic layers within the star, T_c/T_flow ∝ a v⁻¹_esc. The thermal speed is lower in hydrogen-deficient WR atmospheres, while v_esc can be larger than in O stars, as it is the case for our program star WR 142. Therefore, the velocity jumps in WR wind shocks are expected to be rather smaller than larger, compared to O-star winds.

These qualitative considerations indicate that if the same shock mechanism were in operation in WR- and O-star winds, the X-ray spectra of WR stars would be softer than those of O stars. However, observations show the contrary—the handful of single WR stars with available spectral measurements all show X-ray emission harder than typically found in O stars (Skinner et al. 2002a, 2002b; Ignace et al. 2003). This trend is confirmed by our new data on WR 142.

Relatively hard X-ray emission can occur in a binary system; however, optical spectra of WR 142 show no contamination from an early-type companion. A low-mass coronal companion to this initially very massive star is extremely unlikely (Lucy 2006).

Babel & Montmerle (1997) proposed that if a stellar wind is magnetically confined it can be strongly heated. The magnetic field locally dominates the bulk motions if the magnetic energy density exceeds the wind kinetic energy density, B²/μ₀ > ρv². At characteristic distance of 1 R_* from the photosphere, the velocity is v ≈ 0.5v_∞, density $\rho =\dot{M}/4\pi v(r)^2 r \approx 5\times 10^{-11}$ g cm⁻³. Therefore, B(r = 2R_*)>7 kG is required to control the WR 142 wind. For a dipole field, this implies that the magnetic field strength at the surface is larger than 50 kG. A field of such strength is not unrealistic, since the radius of WR 142 is ∼50 times smaller than that of OB supergiants, where surface fields of 100 G have been observed (Bouret et al. 2008). One may speculate that even less strong magnetic fields may lead to hard X-rays due to some type of reconnection and heating processes as invoked to explain the high X-ray temperatures observed in O-type stars (Waldron & Cassinelli 2007).

It is difficult to test where in the wind the observed X-rays have been produced. Since no high-resolution X-ray spectrum of a single WR star exists, the f–i–r emission-line diagnostic that has been used for O-star winds has not been applied to WR stars. X-rays produced close to the stellar core must travel through the bulk of photoabsorbing wind. Tentative evidences for absorption edges are found in the X-ray spectra of WR 1 (Ignace et al. 2003) and WR 142.

In the wind of WR 142, the radius at which the photoabsorption to X-rays is equal to unity is plotted in Figure 2 against photon energy. The two models shown reproduce the stellar SED equally well. One of the models is thick to X-rays for thousands of R_*, and the observed X-rays would have to emerge from the far outer regions of the wind. The origin of X-rays far out in the wind is favored in the alternative scenario of X-ray emission from O stars (Pollock 2007). However, a reduction in the mass-loss (a poorly constrained parameter) by a factor of only 2 and/or a higher effective stellar temperature results in a higher degree of wind ionization. In this case, a fraction of X-rays formed deep in the wind could freely escape. Wind clumping can further reduce wind attenuation (Feldmeier et al. 2003).

There is suggestive evidence that WR 142 is a fast rotator, as is the case for the WN2 star WR 2 (see Section 2). The detection of strong and hard (log L_X ≈ 10³² erg s⁻¹, T_X ≫ 10 MK) X-rays in WR 2 (Oskinova 2005; Skinner et al. 2008) likely indicates that a similar mechanism operates in these hot, compact, fast rotating stars. The lower X-ray luminosity of WR 142 compared to WR 2 can reflect the higher absorption of X-rays in its denser wind. Fast rotation can lead to a slow equatorial wind with enhanced density, possibly forming an outflowing disk, and a faster, thinner polar wind (Ignace et al. 1996). If the round shape of the line profiles found in both, WR 142 and WR 2, is due to rotational broadening, the inclination angle must be large.

Summarizing, we have reported the detection of weak, but hard X-rays from WR 142 and speculate that their origin is connected with magnetic activity of this far-evolved, compact WR star. The specific shape of its optical line profiles signals that WR 142 may be rotating at nearly breakup velocity. In this case, the spherical symmetry will be broken, potentially affecting the X-ray production and absorption. Observations of better quality and progress in modeling are certainly needed to understand fully the high-energy processes in the winds of massive stars at late stages of their evolution.

This study is based on observations obtained with XMM-Newton, an ESA science mission with instruments and contributions directly funded by ESA Member States and NASA. This research has made use of NASA's Astrophysics Data System Service and the SIMBAD database, operated at CDS, Strasbourg, France. Special thanks to Robert Gruendl for providing the IR measurements. The authors thank the referee, M. De Becker, for important comments that improved the paper. Funding for this research has been provided by NASA grant NNX08AW84G (Y.-H.C. and R.I.) and DLR grant 50 OR 0804 (L.M.O.).