CHANDRA DETECTS THE RARE OXYGEN-TYPE WOLF–RAYET STAR WR 142 AND OB STARS IN BERKELEY 87

WR 142 is detected by Chandra as an X-ray source at J202144.35+372230.7 (Figure 1). Table 2 summarizes some properties of the X-ray emission from WR 142. These coordinates are the source centroid as determined by wavdetect using an energy-filtered image (0.3–8 keV). Without further astrometric calibration, the Chandra ACIS-I absolute position has an accuracy of ≈04 (68% limit).¹⁰ The X-ray position of WR 142 is in exact agreement with the Two Micron All Sky Survey (2MASS) position of J202144.35+372230.7. Additionally, the Chandra position has offsets from optical counterparts of only 020 from the USNO-B1.0 source at J202144.35+372230.9 and 023 from the Hubble Space Telescope (HST) GSC 2.3.2 source at J202144.36+372230.5. All offsets are well within the positional accuracy of Chandra.

Results from the CIAO tool srcextent indicate that WR 142 is not extended. Use of a PSF file specifically tailored for our observation with the WR 142 spectrum and the position of WR 142 on the ACIS-I detector yields a source size of only 052 (041–064, 90% confidence), equivalent to about one ACIS pixel, and a PSF size of 085 (066–103, 90% confidence). Thus, the X-ray emission of WR 142 is not extended at the 90% confidence level.

The X-rays from WR 142 are weak but extremely hard, with a mean photon energy of 4.31 keV (Table 2). Surprisingly, WR 142 does not exhibit any significant emission below 2 keV and there is no evidence of a soft component (Figures 2 and 3). We find only one event below 2 keV, with an energy of 1.89 keV. This is in contrast to a faint soft component at E ⩽ 1 keV in the XMM-Newton EPIC MOS spectrum of WR 142 reported by Oskinova et al. (2009). Assuming that no variability occurred in the WR 142 source spectrum or N_H between the two observations, the most likely explanation for this difference would be the much lower effective area of Chandra ACIS-I at low energies E ⩽ 1 keV.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We tried to fit the spectrum of WR 142 (Figure 3) with several different models: a 1T (one temperature) APEC optically thin plasma model and a power-law model. The APEC model requires a very high temperature and thus a thermal bremsstrahlung model is similar. Because of low counts, the spectral fit results were somewhat sensitive to binning strategy. Representative spectral fits are shown in Table 3, based on fits of unbinned spectra. Between the different models, there is little variation in the quality of fit using C-statistics (Cash 1979). However, determining whether a particular model is statistically acceptable using C-statistics is not as straightforward as with χ² statistics (Heinrich 2003).

Allowing the abundances to vary from solar displays no significant improvement to the APEC fits, nor does the use of representative WO abundances (Table 3). This is largely because we do not detect any emission lines in the spectrum from elements such as O which are highly non-solar in WO stars. Additionally, all of the O and C lines that lie in the Chandra bandpass are below 1 keV where we do not detect any signal.

The data show a slight preference for a model that includes a Gaussian line for an iron line in the ≈6.4–6.67 keV energy range (Table 3). The counts from WR 142 are not sufficient to distinguish between the possible iron lines. The event energies closest to the possible iron emission, 6.24, 6.40, 6.60, and 6.63 keV, show two events within 0.1 keV of the iron K line complex at 6.67 keV (Fe xxv He-like triplet). If a weak Fe line is present, it is likely to be the 6.67 keV iron K line, which has been detected in other WR stars.

We have also experimented with fits of binned spectra using χ² statistics in order to estimate confidence bounds on important fit parameters (F_X, N_H, kT, Γ_pow). Given the limited number of counts (46), binning at desired levels of >10 counts per bin for fitting with χ² statistics was not feasible. We thus consider spectra binned to a minimum of 6 counts per bin. Results indicate that if the plasma is thermal, as the APEC models assume, then the temperature kT must be very high, with a lower 90% confidence bound of kT > 3.6 keV (Table 3). The Chandra data do not place a useful upper bound on kT.

In determining the X-ray flux, the unbinned spectral fits display fairly consistent results. The fit with the lowest C-statistic shown in Table 3 gives an observed (absorbed) X-ray flux of F_X = 1.8 [0.8–2.5] × 10⁻¹⁴ erg s⁻¹ cm⁻² (0.3–8 keV, 1σ confidence intervals). The error bounds reflect the range of fluxes measured for the different models in Table 3. The best-fit models give X-ray luminosities of log L_X= 30.72–30.82 (erg s⁻¹), which are derived from unabsorbed fluxes (0.3–8 keV) resulting from different unbinned spectral fits in Table 3.

All indications are that the absorption toward WR 142 is high, with N_H around (4–5) × 10²² cm⁻². This is about 3–4 times higher than the absorption estimated from the observed optical extinction of 1.3 × 10²² cm⁻² obtained using A_V ≈ 6.1 (van der Hucht 2001; Gorenstein 1975 conversion). Spectral fits of WR 142 with N_H fixed at the value determined from optical extinction give much larger fit residuals than the fits with larger absorption given in Table 3.

The source shows no obvious large-scale variations nor flares in its X-ray light curve (Figure 4). As for short-term variability, KS statistics and the GL algorithm give a probability of constant count rate of 0.77 and 0.40, respectively. Thus, neither test implies variability at high confidence levels (>90%) on timescales of hours during the <1 day Chandra observation.

Zoom In Zoom Out Reset image size

For purposes of comparing the observed X-ray flux by Chandra and the previously published XMM-Newton X-ray flux of F_X = (4 ± 2) × 10⁻¹⁴ erg cm⁻² s⁻¹ from 0.25 to 12 keV (Oskinova et al. 2009), we measure F_X = (3 ± 1) × 10⁻¹⁴ erg cm⁻² s⁻¹ (0.25–12 keV) using an average value of the fluxes from the extended energy range from the three models shown in Table 3. Thus, Chandra and XMM-Newton X-ray fluxes agree to within the uncertainties. Based on the above flux comparison, any variability over the approximate one year interval between the Chandra and XMM-Newton observations was a factor of ∼3 or less.

Based on the Chandra data discussed above, the X-ray properties of WR 142 can be summarized as follows: (1) source structure is consistent with a point source at Chandra's angular resolution; (2) no significant variability is detected over a timescale of ≈1 day; (3) strong X-ray absorption below 2 keV, with an absorption column density (N_H) that is a factor of ≈3–4 greater than expected from published optical estimates; (4) no evidence for a soft X-ray component below 2 keV; (5) a spectrum dominated by hard X-rays above 2 keV that can be fit equally well using a very high temperature thermal plasma model or a power-law model; and (6) a relatively low X-ray luminosity log L_X(0.3–8 keV) = 30.7–30.8 erg s⁻¹ (at d = 1.23 kpc), which is ≈10 times less than found for two X-ray bright OB stars in Berkeley 87 (Section 4). We discuss these properties in more detail below.

This star was classified as O9V by TF82 (Berkeley 87-25) and was the only object explicitly identified as an O-type star in their catalog of 105 likely members of Berkeley 87. The spectral type was later refined to O8.5 III by Massey et al. (2001), who also estimated the mass to be M_* = 39 M_☉. BD+36 4032 lies ≈3' northwest of WR 142.

X-ray characteristics are listed in Table 2. The X-ray spectrum is soft (Figure 3) with almost all of the emission emerging below 2 keV, which is dramatically different than the hard emission from WR 142. Reasonably good fits can be obtained using a solar abundance 1T APEC model with a cool plasma component at kT ≈ 0.6 keV (Table 4). The 1T APEC fit can be improved slightly by adding a second temperature component between 1 and 2 keV, but almost all of the emission measure still resides in the cool component.

The best-fit N_H given in Table 4 equates to A_V = 5.1 [4.4–6.5] mag (Gorenstein 1975). This is in good agreement with the optically determined value A_V = 4.8 mag using E(B − V) = 1.60 and A_V = 3E(B − V) (TF82). Thus, the excess X-ray absorption detected in WR 142 is not present toward this O star.

This B supergiant has V = 8.71 mag and was the brightest member of Berkeley 87 listed by TF82 (Berkeley 87-3). They assigned a spectral type B2 Iabe but Massey et al. (2001) classified it as B1 Ia.

The Chandra spectrum of HD 229059 is nearly identical to that of BD+36 4032 (Figure 3). One noticeable difference is that HD 229059 shows stronger emission in the 1.8–1.9 keV range, which is likely due to stronger line emission from the Si xiii triplet (1.84–1.87 keV).

Spectral fits of HD 229059 give nearly identical results to BD+36 4032. Table 4 summarizes the best-fit 1T APEC model. No significant fit improvement was obtained by adding a second temperature component. The N_H from Table 4 gives A_V = 5.2 [4.5–5.6] mag, in excellent agreement with the optically determined value A_V = 5.1 mag (TF82). Thus, we find no excess X-ray absorption toward HD 229059.

This B0.2 III star lies ≈51 west of WR 142 and is object number 4 in the list of likely cluster members compiled by TF82. It was included in the study of Massey et al. (2001) who obtained a mass estimate M_* = 23 M_☉. Berkeley 87-4 was the third brightest Chandra X-ray detection of the OB stars in Berkeley 87 (Table 2). There are sufficient counts for a crude spectral fit, and a 1T APEC model gives N_H and kT values that are similar to the brighter OB stars BD+36 4032 and HD 229059 but a lower X-ray luminosity (Table 4). The X-ray-derived N_H agrees to within the model uncertainties with that expected from optical reddening (TF82).

This variable star is also known as MWC 1015 and is object number 15 in the list of likely cluster members compiled by TF82. V439 Cyg has anomalous colors and is displaced well to the redward side of the zero-age main-sequence in the cluster color–magnitude diagram (TF82). The star's spectrum is peculiar and evidently variable, with most recent studies assigning an early B spectral type of B0e (Polcaro & Norci 1998) or B[e] (Massey et al. 2001). V439 Cyg shows some similarities to an LBV (Polcaro & Norci 1998).

V439 Cyg lies ≈32 NW of WR 142 and is a bright near-IR source with 2MASS data giving K_s = 7.59 mag and position 2MASS J202133.57+372451.57. There is no X-ray source at or near this position in our Chandra image. A circular region of radius 196 (90% encircled energy at E = 2 keV ¹¹) centered on the V439 Cyg 2MASS position encloses only one photon in the ACIS-I image (0.3–8 keV range).

To compute an X-ray upper limit, we assume a conservative 4 count ACIS-I detection threshold and an intrinsic thermal stellar spectrum similar to the two OB stars discussed above. Using N_H = 1.0 × 10²² cm⁻² corresponding to E(B − V) = 1.53 (TF82) and a Raymond–Smith solar-abundance thermal plasma model with kT = 0.6 keV, the Chandra PIMMS¹² simulator gives an unabsorbed flux upper limit F_X(0.3–8 keV) ⩽ 3.9 × 10⁻¹⁵ erg cm⁻² s⁻¹. If V439 Cyg is a cluster member, then d = 1.23 kpc gives log L_X ⩽ 29.85 (erg s⁻¹).

Given the unusual nature of V439 Cyg, further consideration of its X-ray non-detection is warranted. In this regard, it is important to note that Chandra detected only the most luminous B stars in the cluster having absolute magnitudes m_V ⩽ −2.1 (Section 4.5). Several less-luminous early B stars were undetected (Table 2 notes). This is likely a consequence of the known L_X ∝ L_bol trend for O and early B-type stars (Berghöfer et al. 1997). The absolute magnitude of V439 Cyg is not well known due to uncertain reddening, but it appears to lie at or near the m_V cutoff for our Chandra detections (Figure 7 of TF82). Thus, V439 Cyg may have insufficient L_bol to produce an X-ray luminosity above the Chandra detection threshold.

Excess absorption toward V439 Cyg may have also partially contributed to its non-detection. The anomalous colors of V439 Cyg imply excess reddening, and it may be surrounded by a shell of material that was ejected during a previous mass-loss phase (Polcaro & Norci 1998). A gas-rich shell would increase the line-of-sight photoelectric absorption to lower energy X-rays (E < 1 keV).

If V439 Cyg is indeed an LBV (Polcaro & Norci 1998), it could be evolving into the WN phase. For stars of initial masses ∼40–75 M_☉, the evolutionary sequence is (Crowther 2007): O → LBV → WN_H-poor. Some putatively single WN7–9 stars (also known as WNL stars) have so far gone undetected in X-rays (Figure 5). But a few recent WNL detections have been reported, including an unambiguous detection of the apparently single WN9ha star WR79a (Skinner et al. 2010b), a possible faint detection of WR 16 (WN8h) in a recent XMM-Newton observation (Figure 5), and a new Chandra detection of the WN8 component of the close WR+OB binary system WR 147 (Zhekov & Park 2010). Although the nature and evolutionary status of V439 Cyg remain somewhat of a mystery, its X-ray non-detection would not be in discord with an evolved star that is approaching the WNL phase, given that a few other WNL stars have also gone undetected in X-rays.

Zoom In Zoom Out Reset image size

In addition to the OB stars discussed above, Chandra provides faint X-ray detections of four other B stars in Berkeley 87. These are Berkeley 87 No. 13 (B0.5 III), No. 24 (B1 Ib), No. 26 (B0.5 Iab), and No. 32 (B0.5 III). Their basic properties are summarized in Table 2, but they lack sufficient counts for spectral analysis. We also report non-detections of seven other B stars in Berkeley 87 (Table 2 notes).

The Chandra results show a noticeable trend in that all of the B star detections have high luminosities (m_V ⩽ −2.1; TF82 photometry). Most, but not all, of the undetected B stars are less luminous (m_V ⩾ −2.1). However, there are a few interesting exceptions. The B1 V star Berkeley 87-31 (m_V = −2.7; TF82) lies nearly on-axis at an offset of 187 from WR 142 but was not detected. However, it does lie near an ACIS-I CCD gap. Two other luminous but undetected B stars, No. 9 (B0.5 V) and No. 38 (B2 III), lie further off-axis (39 and 50, respectively).

Chandra spectral fits of WR 142 (Table 3) give an absorption column density of N_H≈ (4–5) × 10²² cm⁻². In contrast, optical studies yield N_H ≈ 1 × 10²² cm⁻² (TF82; Barlow & Hummer 1982; van der Hucht 2001; Gorenstein 1975 conversions) for Berkeley 87 cluster members. Our X-ray analysis of the cluster members BD+36 4032 and HD 229059 (Section 4) also yields N_H ≈ 1 × 10²² cm⁻². Thus, the X-ray spectra clearly show that excess absorption is present toward WR 142 that is not seen toward two other massive cluster members. This could arise either in the dense WO wind or perhaps in cold circumstellar gas.

Excess X-ray absorption seen as a disagreement between absorption based on optical extinction and the X-ray fit N_H has been seen in other WR stars as well, such as γ² Vel (Skinner et al. 2001) and recently detected WN stars (Skinner et al. 2010b).

Inhomogeneous extinction, although present, does not account for the discrepancy in X-ray and optical N_H values for WR 142. The spread in extinction values for cluster members is small, from N_H = (0.8–1.3) × 10²² cm⁻² (TF82; Gorenstein 1975 conversion). Additionally, the smallest foreground extinction is near the south and WR 142 is ≈26 southeast of the cluster center (TF82), indicating the excess N_H is due to local absorption associated with WR 142.

If the excess absorption is due to the wind, we can estimate the radius at which the X-rays emerge using a wind optical depth calculation. From the observed Chandra WR 142 spectrum, the X-rays are entirely absorbed below ≈2 keV. Using generic WO abundances (van der Hucht et al. 1986) and assuming the mass-loss parameters in Table 1, the radius of optical depth unity at E = 2 keV is R(τ = 1, 2 keV) ≈1500 R_☉. Clumping in the WR winds may reduce the mass-loss rate by a factor of 2–4 (Crowther 2007) and would reduce R(τ = 1, 2 keV) by the same factor. R(τ = 1) is well outside the wind acceleration zone and the wind will already have reached terminal speed, assuming a standard β = 0.5–1.0 wind velocity law. It should be kept in mind that this is a minimum radius for 2 keV X-rays to escape.

Another possibility regarding excess absorption is the presence of cold circumstellar gas. Nebulosity around WR 142 has been detected (Miller & Chu 1993), though a ring nebula is not present. Polcaro et al. (1997) note excess reddening surrounding WR 142 of at least N_H = 2.7 × 10²¹ cm⁻² (Gorenstein 1975 conversion) suggesting dense material, possibly a nebula from the stellar wind. Alternatively, Lozinskaya (1991) detected a distant IR shell (∼23 pc from WR 142) whose dynamical age suggests that it formed before, and therefore independently, of the WO wind. Though it is apparent that some circumstellar material is present, our X-ray spectral fits have not distinguished between wind absorption and absorption by circumstellar gas far from the star. Both the wind and local circumstellar material could contribute to the excess X-ray absorption.

There is no evidence from ACIS images (or other observations) that WR 142 is a binary. For an on-axis point source, the Chandra HRMA and ACIS PSF encircle 70% of the source energy for an angular source radius of ≈12 at 2 keV and ≈13 at 6 keV.¹³ If two X-ray sources are present, their separation must be <1''.

The WR 142 X-ray luminosity from this Chandra observation is much less than that of known WR+O binaries, which are generally very luminous in the X-rays. Due to plasma formed in colliding wind shocks, these systems can have luminosities approaching L_X ≈ 10³³ erg s⁻¹, such as γ² Vel (WC8+O7.5III; Skinner et al. 2001; Schild et al. 2004). WR 142 is at least 2 orders of magnitude less luminous.

However, the possibility of an unseen companion remains. But, a compact companion seems to be unlikely. Theoretical models of the WR winds accreting onto neutron stars predict high luminosity (log L_X>34 erg s⁻¹; HD 50896 = EZ CMa; Stevens & Willis 1988). Additionally, a neutron star would be required to spin fast enough to throw off most of the accreting material to remain undetected (propeller effect; e.g., Lipunov 1982). If an optically faint lower mass normal stellar companion (rather than a compact object) lies close to WR 142, it could easily escape detection.

Currently, the binary frequency for galactic WR stars is ∼40% (van der Hucht 2001). Additional searches for binarity are needed for WR 142. These could include high-resolution infrared imaging or radio interferometry, though they would only be capable of detecting a companion down to separations of a few tenths of an arcsecond. More promising approaches would be to search for periodic photometric or spectroscopic variability that could signal a closer spectroscopic companion. Spectroscopically, an X-ray emitting companion orbiting in the WR wind could reveal itself through phase variability of the WR star ultraviolet lines (Hatchett & McCray 1977).

Low counts in the WR 142 ACIS-I spectrum result in nearly identical fits by either thermal APEC models or power-law models. As such, the WR 142 spectrum cannot sufficiently distinguish between a very high temperature thermal plasma and non-thermal processes, e.g., inverse Coulomb scattering due to relativistic electrons accelerated in wind shocks (Chen & White 1991a). In general, other WR stars with higher signal-to-noise spectra show detections of emission lines indicative of thermal emission. While the goodness of fit for spectral models slightly improves with the addition of the Fe K line for WR 142, there are not enough line counts to qualify as a definite line detection. There is no obvious evidence of other thermal emission lines, such as the S line at 2.46 keV. Theoretical models do exist that attempt to account for the hard X-ray emission from some early-type stars by non-thermal processes (Chen & White 1991a). However, if WR 142 were a non-thermal X-ray source, it would be the first putatively single WR star in that class. The possibility of non-thermal emission has been previously mentioned for other WR stars such as WR 110 but the evidence for non-thermal emission was not conclusive (Skinner et al. 2002).

Thermal X-ray emission is possible through several processes. Often radiative wind shocks are considered, yet with no emission below 2 keV detected from WR 142, a line-driven flow instability model predicting soft X-rays from radiative wind shocks that is typically assumed for O stars (Gayley & Owocki 1995) cannot explain the hard emission. We discuss wind kinetic energy conversion, colliding wind shocks, magnetically confined wind shocks (MCWSs), and coronal emission as possible thermal emission processes.

5.4.1. Kinetic Energy of the Wind

5.4.2. Colliding Wind Shocks

5.4.3. Magnetically Confined Wind Shocks

5.4.4. Coronal Emission

Electrons in the stellar wind accelerated to relativistic energies by the Fermi mechanism and trapped in weak magnetic fields could scatter stellar UV photons to produce X-ray emission and γ-ray emission (Chen & White 1991a, 1991b), as suggested for WR stars by Pollock (1987). The accelerated electrons undergo adiabatic cooling between shocks and thus can only lose heat through non-thermal processes, such as inverse Compton, non-thermal bremsstrahlung, or synchrotron radio emission (Chen & White 1991a). Such processes may be at work in WR 142: X-ray spectra of WR 142 can be fitted with a power law and Berkeley 87 lies close to an extended region of high-energy γ-ray emission in Cygnus (Abdo et al. 2007, 2009; Smith et al. 2005), though the primary source of the γ-ray emission has not yet been proven to be Berkeley 87 itself.

For OB stars, the radiation will produce a power-law electron spectrum where the X-ray flux from inverse Compton emission should scale with energy as F_X ∝ E^−1/2 for electron index n = 2 (Chen & White 1991a). Spectral fits in Table 3, when the given photon index Γ_pow = 1 is converted to an energy index, give F_X∝ E⁰, though the photon index was poorly constrained. If the photon index is frozen to Γ = 1.5 so as to produce Chen & White's energy power-law index, the fit is reasonable but exhibits slightly more absorption and higher X-ray flux and luminosity (Table 3 footnotes).

The X-ray emission is predicted to be hard without large-scale variations, and if present, variations only on timescales of hours to days (Chen & White 1991a). However, the Chen & White (1991a) model is tuned to OB stars, such that inverse Compton is the assumed emission mechanism whereas bremsstrahlung may be just as important or even dominate inverse Compton for WR stars. Additionally, other factors, such as the Coulomb effects in the dense WR wind, have not been fully evaluated (Chen & White 1991a). Still, the non-thermal emission process is plausible for some WR stars, especially where non-thermal radio emission is detected (Pollock 1987). WR 142, is as of yet, undetected in the radio. Cappa et al. (2004) determined an upper limit of <0.90 mJy for the 3.6 cm flux density from WR 142. Despite the lack of detection of non-thermal radio emission at 3.6 cm, radio observations at longer wavelengths where non-thermal emission could lead to higher flux densities would be useful. The hard and essentially featureless X-ray spectrum of WR 142 could be non-thermal, but a higher signal-to-noise X-ray spectrum will be needed to distinguish between thermal and non-thermal emission.