Identification of a Population of X-ray Emitting Massive Stars in the Galactic Plane
G. E. Anderson, B. M. Gaensler, D. L. Kaplan, B. Posselt, P. O. Slane, S. S. Murray, J. C. Mauerhan, R. A. Benjamin, C. L. Brogan, D. Chakrabarty, J. J. Drake, J. E. Drew, J. E. Grindlay, J. Hong, T. J. W. Lazio, J. C. Lee, D. T. H. Steeghs, M. H. van Kerkwijk
aa r X i v : . [ a s t r o - ph . H E ] N ov A CCEPTED TO
The Astrophysical Journal
Preprint typeset using L A TEX style emulateapj v. 11/26/03
IDENTIFICATION OF A POPULATION OF X-RAY EMITTING MASSIVE STARS IN THE GALACTIC PLANE G EMMA
E. A
NDERSON , B. M. G
AENSLER , † D AVID
L. K
APLAN , ‡ B ETTINA P OSSELT , P ATRICK
O. S
LANE , S TEPHEN
S. M
URRAY , J ON C. M
AUERHAN , R OBERT
A. B
ENJAMIN , C RYSTAL
L. B
ROGAN , D EEPTO C HAKRABARTY , J EREMY
J. D
RAKE , J ANET
E. D
REW , J ONATHAN
E. G
RINDLAY , J AESUB H ONG , T. J
OSEPH
W. L
AZIO , J ULIA
C. L EE , D ANNY
T. H. S
TEEGHS , M ARTEN H. VAN K ERKWIJK , Accepted to
The Astrophysical Journal
ABSTRACTWe present X-ray, infrared, optical and radio observations of four previously unidentified Galactic plane X-ray sources, AX J163252–4746, AX J184738–0156, AX J144701–5919 and AX J144547–5931. Detectionof each source with the
Chandra X-ray Observatory has provided sub-arcsecond localizations, which weuse to identify bright infrared counterparts to all four objects. Infrared and optical spectroscopy of thesecounterparts demonstrate that all four X-ray sources are extremely massive stars, with spectral classificationsOfpe/WN9 (AX J163252–4746), WN7 (AX J184738–0156 = WR121a), WN7–8h (AX J144701–5919) andOIf + (AX J144547–5931). AX J163252–4746 and AX J184738–0156 are both luminous, hard, X-ray emitterswith strong Fe XXV emission lines in their X-ray spectra at ∼ . Subject headings: stars: winds, outflows – stars: Wolf-Rayet – supergiants – X-rays: binaries – X-rays:individual (AX J163252–4746, AX J184738–0156, AX J144701–5919, AX J144547–5931)– X-rays: stars INTRODUCTION
Wolf-Rayet (WR) stars and their O-type supergiant progenitors (Of) evolve from the most massive stars in our Galaxy, with initialmasses & M ⊙ . These evolved stars, particularly WR, have extremely strong stellar winds, experiencing high mass-loss ratesof ˙ M ∼ - M ⊙ yr - and, in some cases, have luminosities > L ⊙ (Crowther 2008). Their short lifetimes make them very rare; <
400 WR stars are known in our Galaxy (van der Hucht 2006; Martins et al. 2008; Shara et al. 2009; Mauerhan et al. 2009,2010), and are usually only found in the Galactic plane.Massive stars have historically been discovered through optical and infrared observations and are classified based on their spectralcharacteristics in these wavebands. However, X-ray observations are now becoming a newly recognized technique for discoveringmassive stars and are also a powerful tool in assessing their physical environments (e.g. Mauerhan et al. 2010), allowing us todetermine if they are isolated or in a high-mass X-ray binary (HMXB) or colliding-wind binary (CWB) system. By discoveringmore of these massive stars and understanding their emission mechanisms, we can determine how mass-loss drives the differentstages of stellar evolution.The most accepted model for X-ray generation in a single hot star is the instability-driven wind-shock picture, which attributesthe production of soft, thermal X-ray emission, with temperatures of kT < Sydney Institute for Astronomy, School of Physics A29, The University of Sydney, NSW 2006, Australia: [email protected] Hubble Fellow; Kavli Institute for Theoretical Physics, Kohn Hall, University of California, Santa Barbara, CA 93106, USA Harvard-Smithsonian Center for Astrophysics, Cambridge, MA 02138, USA Spitzer Science Center, California Institute of Technology, Pasadena, CA 91125, USA Department of Physics, University of Wisconsin, Whitewater, WI 53190, USA National Radio Astronomy Observatory, Charlottesville, VA 22903, USA MIT Kavli Institute for Astrophysics and Space Research and Department of Physics, Massachusetts Institute of Technology, Cambridge, MA 02139, USA Centre for Astrophysics Research, STRI, University of Hertfordshire, Hatfield AL10 9AB, UK Naval Research Laboratory, Washington, DC 20375, USA Department of Physics, University of Warwick, Coventry CV4 7AL, UK Department of Astronomy and Astrophysics, University of Toronto, Toronto, ON M5S 3H4, Canada † Australian Research Council Federation Fellow ‡ Current address: Department of Physics, University of Wisconsin, Milwaukee, WI 53201-0431, USA
Anderson et al.& White 1980; Lucy 1982). More exotic models of X-ray generation in isolated massive stars include the magnetically channeledwind-shock mechanism, which was first explored in detail by Babel & Montmerle (1997a,b). In this case a radiatively drivenstellar wind is magnetically channeled from the two hemispheres of the star. These streams collide at the magnetic equator andthe rapid deceleration causes shock heating resulting in high levels of hard thermal X-ray emission. There is also the possibilityof isolated OB supergiant stars producing intrinsic hard non-thermal X-rays through inverse Compton scattering (Chen & White1991).Alternatively, the X-ray emission may not be completely intrinsic to the massive star but created through a binary interaction.X-rays can be generated in high-mass X-ray binaries (HMXBs) through gravitational accretion onto a compact object such as aneutron star (NS) or black hole (BH). The two main classes of HMXBs are the Be X-ray binary systems (BeX) and the supergiantX-ray binaries (SGXB). BeXs are accretion fed by a disk around a Be star and are often transient X-ray sources whereas SGXBsare wind-fed, and persistent, X-ray sources (McClintock & Remillard 2006).Colliding-wind binary (CWB) systems are another class of massive stellar binaries marked by extreme wind loss and high-energy emission. Originally predicted by Prilutskii & Usov (1976) and Cherepashchuk (1976), the supersonic winds from thetwo massive stars in a binary produce shock-heated gas (Stevens et al. 1992), resulting in hard, thermal X-ray emission (Pittard& Parkin 2010) and possibly γ -rays, likely produced by inverse Compton scattering (e.g. Benaglia & Romero 2003; Pittard &Dougherty 2006; Reimer et al. 2006; De Becker 2007). The detection of this high-energy emission allows us to probe the natureof these shocks and provides a laboratory for investigating particle acceleration in a very different density regime to that insupernova remnants.In this Paper we discuss our classification of four previously unidentified Galactic X-ray sources; AX J163252–4746, AXJ184738–0156, AX J144701–5919 and AX J144547–5931. These sources have been observed with the Chandra X-ray Ob-servatory as part of the “ChIcAGO” (
Chasing the Identification of ASCA Galactic Objects ) project (Anderson et al. in prep.),a survey designed to localize and classify the unidentified X-ray sources discovered during the
ASCA
Galactic Plane Survey(AGPS; Sugizaki et al. 2001).
ASCA (the Advanced Satellite for Cosmology and Astrophysics) surveyed the inner region of theGalactic plane, detecting 163 X-ray sources with fluxes between 10 - and 10 - erg cm - s - in the 0 . - . ASCA
X-ray telescope’s ∼ ′ spatial resolution, only a third of these X-ray sources have been properly characterizedand little is known about the remaining unidentified objects. This unidentified population should contain at least some rare classesof X-ray objects, as modeling by Hands et al. (2004) and Grindlay et al. (2005) has demonstrated that the Galactic populations ofcataclysmic variables, bright X-ray binaries, and background Active Galactic Nuclei cannot account for the entire observed fluxdistribution of X-ray sources detected in the AGPS. In ChIcAGO, we are combining the sub-arcsecond localization capabilitiesof Chandra with a detailed multi-wavelength follow-up program, with the goal of classifying the >
100 unidentified sources inthe AGPS.We chose to begin our investigation with AX J163252–4746, AX J184738–0156, AX J144701–5919 and AX J144547–5931(listed in order of decreasing X-ray flux) as they are all highly absorbed with bright infrared counterparts (8 µ m magnitude < Chandra observations and follow-up studies are described in § § OBSERVATIONS AND RESULTS
Chandra and Archival XMM Newton Data
Our
Chandra observations of AX J163252–4746, AX J184738–0156, AX J144701–5919 and AX J144547–5931 were short(between 1 - ∼
100 counts per target so as to fulfill the primary aim to localize eachsource, while providing limited spectral information. We observed AX J184738–0156, AX J144701–5919 and AX J144547–5931 with the Advanced CCD Imaging Spectrometer (ACIS; Garmire et al. 2003) in the ACIS-S mode. As AX J163252–4746has a predicted ACIS count-rate > . - , we used the High Resolution Camera (HRC; Murray 2000) in the HRC-I modein order to avoid the positional degradation associated with pile-up (Davis 2001) that would arise from an ACIS observation.These data were reduced using the Chandra Interactive Analysis of Observation software (CIAO), version 4.1, following thestandard reduction recipes given in the online CIAO 4.1 Science Threads . Further details on the Chandra observations will bepublished by Anderson et al. (in prep). For each of the four AGPS targets, a
Chandra source with 50 -
150 counts was detectedwithin 3 ′ of the published ASCA position. The need for
Chandra observations to localize the AGPS sources is illustrated inFigure 1, where each
Chandra detection clearly shows a much more precise source position (white contour) when compared tothe original
ASCA detection (black contours). The
Chandra instruments and exposure times used for the observations can befound in Table 1. The resulting
Chandra position for each source is listed in Table 2.AX J163252–4746 and AX J184738–0156 were also detected off-axis in archival
XMM-Newton data (observational ID’s0201700301 and 0203850101, respectively), with > XMM-Newton
Serendipitous Source catalogue(2XMMi: Watson et al. 2009). The details on these observations can also be found in Table 1. (AX J163252–4746 was also de-tected off-axis in ten other
XMM observations. However, we chose to concentrate our analysis on observational ID 0201700301as it has the longest integration and therefore the greatest number of counts detected.) http://cxc.harvard.edu/ciao/threads/ New Population of X-ray Emitting Massive Stars 3We chose to model the
XMM spectra of AX J163252–4746 and AX J184738–0156, rather than their
Chandra data, as the HRCinstrument does not provide any spectral information for AX J163252–4746 and the
XMM observations detected many moresource counts for AX J184738–0156. We fit the
XMM spectra of AX J163252–4746 and AX J184738–0156 with absorbedRaymond & Smith (1977) models as both sources show a strong emission line between 6 - XMM spectra of both AX J163252–4746 and AX J184738–0156 are shown inFigure 2. Using a Gaussian profile to model the emission line in the
XMM spectra, the line was found to be unresolved for bothAX J163252–4746 and AX J184738–0156. In each case the equivalent width (EW) is ∼ . + . - . keV, at energies 6 . ± . . ± .
01 keV (90% confidence) for AX J163252–4746 and AX J184738–0156, respectively, where the
XMM energyresolution is FWHM ∼ We identify both emission lines as Fe
XXV , which has an approximateenergy of 6.7 keV. The absorbed Raymond-Smith modeling of both AX J163252–4746 and AX J184738–0156 includes smallerbumps in the spectral fits, as seen in Figure 2, which could indicate the presence of other spectral lines. However, the error barsin the data points are too large for the existence of other ionic species to be confirmed.There is no evidence for short-term variability within our
Chandra and archival
XMM observations between ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ Chandra and
XMM observations of each source are within a factor of two from the original ASCA flux,also derived from a powerlaw fit, published by Sugizaki et al. (2001). Using the
Chandra and
XMM datasets we also determinedthat there is no evidence for periodicity between ∼ ∼ ∼ ∼ Chandra spectra of AX J144701–5919 and AX J144547–5931we compared the
XMM and
Chandra spectra of AX J184738–0156. We used the
XMM best-fit absorbed Raymond-Smith valuesto fit the
Chandra spectrum of AX J184738–0156, allowing the normalization parameter to vary. The resulting fit is shownin blue in Figure 2(b) demonstrating that the
Chandra spectrum is compatible with the
XMM fits. An independent absorbedRaymond-Smith fit to the
Chandra spectrum results in best-fit values that have very large uncertainties. It is therefore difficult tostart from the low-statistics of the
Chandra data and then constrain the parameters of the source’s spectrum. We therefore limitour interpretation of the AX J144547–5931 and AX J144701–5919
Chandra spectra by using absorbed Raymond-Smith fits onlyto estimate the unabsorbed X-ray flux. With this caveat, the Raymond-Smith spectral fit parameters for all four sources aresummarized in Table 2.
Multi-wavelength DataSurvey and Catalog Comparisons
Comparison of the
Chandra positions of AX J163252–4746, AX J184738–0156, AX J144701–5919 and AX J144547–5931 tooptical and infrared catalogs allow us to discover any multi-wavelength counterparts to these X-ray sources. AX J144701–5919and AX J144547–5931 are both faintly detected at optical wavelengths (R magnitude > .
8) in the US Naval Observatory(USNO) B catalog, version 1.0 (Monet et al. 2003). In both the Two Micron All Sky Survey (2MASS) (Skrutskie et al. 2006)and the Galactic Legacy Infrared Mid-Plane Survey Extraordinaire (GLIMPSE) (Benjamin et al. 2003) all four AGPS sourcesare shown to have clear counterparts at near-infrared and mid-infrared wavelengths. The corresponding 2MASS and GLIMPSEsource names and magnitudes are listed in Table 3.The GLIMPSE data also show that AX J163252–4746 is surrounded by a diffuse shell of 8 µ m emission (see Figure 1(a)). Thisemission is likely due principally to polycyclic aromatic hydrocarbons (PAHs) excited by soft ultraviolet radiation, detectedwithin the broadband 8 µ m filter (Watson et al. 2008). Thus, such shells are signposts for hot stars with powerful stellar windsthat are impacting the interstellar medium (Churchwell et al. 2006, 2007).Examination of the 24 µ m mosaic images from the MIPSGAL Survey (Carey et al. 2009) shows that AX J163252–4746, AXJ144701–5919 and AX J144547–5931 are also detected at this wavelength. AX J184738–0156 is situated within the H II regionW43 (Lester et al. 1985). W43 appears very bright and diffuse at 24 µ m, preventing us from identifying a possible counterpart toAX J184738–0156 at this wavelength. Targeted Infrared Observations
AX J184738–0156 is coincident with 2MASS J18473666–0156334, which Blum et al. (1999) resolved into a dense stellar clusterembedded within W43. We re-observed this cluster in the J , H and K s bands using the Persson’s Auxiliary Nasmyth InfraredCamera (PANIC: Martini et al. 2004; Osip et al. 2008) on the Baade, 6.5m, Magellan telescope, located at Las CampanasObservatory, on June 25, 2007. As shown in Figure 1(b), the X-ray source AX J184738–0156 is coincident with the brightest The pipeline-produced archival
XMM spectra were fit using
XSPEC (Dorman & Arnaud 2001). We note that in the
XMM observation of AX J163252–4645,many scatter arcs from the nearby bright X-ray binary 4U 1630–47 are present. This could lead to poor background subtraction in the
XMM reduction pipeline.However, the other
XMM archival observations of AX J163252–4746, which do not have scatter arcs, show very similar spectral shapes. We also included a Gaussian profile component to the powerlaw fit of the
XMM data in order to model the flux in the 6.7 keV Fe
XXV emission line. The
Chandra spectra were fit using the
CIAO spectral fitting tool
Sherpa . Anderson et al.member of the star cluster. Blum et al. (1999) obtained a K -band spectrum of this star at a spectral resolution of R ≈ K -band 2 . - . µ m spectra of the counterparts to AX J163252–4746, AX J144701–5919 and AX J144547–5931on June 15 and 16, 2008 and July 20, 2009, respectively, at the 4.1m Southern Observatory for Astrophysical Research (SOAR)Telescope, located on Cerro Pachon in Chile. The Ohio State Infrared Imager/Spectrometer (OSIRIS; Depoy et al. 1993) was usedto observe AX J163252–4746 and AX J144701–5919 in the high-resolution longslit mode, which provides a spectral resolutionof R ≈ K -band. Stellar spectra were acquired in a slit-scan sequence of 5 positions separated by 5 ′′ each. We usedthe same instrument to observe AX J144547–5931 in the cross-dispersed mode at a resolution of R ≈ ′′ slit scanwith 4 repetitions. Standard reduction procedures, including telluric correction, were applied to the data.The K -band spectrum of AX J163252–4746, shown in Figure 3(a), exhibits an extremely strong emission line of He I λ µ m, as well as strong emission features from the λ µ m complex, composed of He I , N III , C
III and O
III , andBr γ /He I λ µ m. The lower panel of Figure 3(a) is a magnified view of this star’s spectrum showing the presence of weakeremission features from N III λ µ m, He II λ µ m and possibly N V at λ µ m.The K -band OSIRIS spectra of AX J144701–5919 and AX J144547–5931 are shown in Figure 3(b). While both AX J144701–5919 and AX J144547–593 show clear emission lines, the spectrum of the latter is poorer, as high clouds and vibrations dueto strong wind led to a high background and significant flux variation between the exposures. The K -band spectrum of AXJ184738–0156 can be found in Figure 5 of Blum et al. (1999). In all three cases the λ µ m and Br γ /He I λ µ m are seen in emission. AX J144701 - II emission line at λ µ m.While AX J184738-0156 shows He I λ µ m in emission, for AX J144701–5919 this line has a P-Cygni profile, with a weakemission, but strong absorption component. AX J144547–0593 displays the C IV λ µ m emission line complex. Theatomic transitions and center wavelengths of these emission lines can be found in Morris et al. (1996). Targeted Optical Observations
A low signal-to-noise optical spectrum of AX J144547–5931 was obtained with the Low Dispersion Survey Spectrograph (LDSS-3) (Osip et al. 2008) on the 6.5m Clay Magellan telescope, on June 23, 2008. This source was observed using the VPH_ALL(400 lines/mm) grism, with a center slit, resulting in a wavelength range of 3000 - α in emission,with C III λ III λ Radio Observations
AX J163252–4746, AX J144701–5919 and AX J144547–5931 are coincident with radio emission in the first and second epochMolonglo Galactic Plane Surveys (MGPS1 and MGPS2 respectively; Green et al. 1999; Murphy et al. 2007). The MGPS datasets are made up of mosaic images, taken at 843 MHz with the Molonglo Synthesis Telescope (MOST).AX J144701–5919 appears to be associated with an unresolved radio source. Molonglo observed this source in 1994 and 1998,with a measured flux density of 15 ± ± II region W43, as seen at1.4 GHz in the Multi-Array Galactic Plane Imaging Survey (MAGPIS: Helfand et al. 2006).We used the Australia Telescope Compact Array (ATCA) to follow-up the counterpart to AX J144701–5919 and resolve out thediffuse radio emission surrounding AX J163252–4746, observing each source for one hour at each of 1.4, 2.4, 4.8 and 8.6 GHzwith a 6km baseline configuration on January 21, 2008 (AX J144701–5818) and April 11, 2008 (AX J163252–4746). Theradio counterpart to AX J144701–5919 was detected as an unresolved point source in all 4 ATCA frequency bands, with a fluxthat decreased with frequency according to S total ν ≈ ν α with α = - . ± .
09. AX J163252–4746 was only detected at 4.8 and8.6 GHz resulting in a spectral index of α = + . ± .
02. The observed ATCA radio fluxes and upper limits for AX J163252–4746 and AX J144701–5919 are listed in Table 3. Further high-resolution radio follow-up is required to determine whether AXJ184738–0156 and AX J144547–5931 have compact radio counterparts. DISCUSSION
Classification of Stellar Counterparts
The infrared colors of AX J163252–4746, AX J184738–0156, AX J144701–5919 and AX J144547–5931 are reasonably con-sistent with the Hadfield et al. (2007) and Mauerhan et al. (2009) selection criteria for WR stars, i.e., GLIMPSE colors[3 . - [8 . > . . - [4 . > . K - [0 . > .
7. This color selection criteria takes ad-vantage of the free-free excess emission that is generated within the strong, dense, ionized winds of WRs. The colors of our fourtarget, thus, indicate the presence of such winds. The PAH ring surrounding AX J163252–4746 (Figure 1(a)) is also consistentwith a strong wind being generated by the central source. During the observation of AX J163252–4746 the antenna providing the 6km baseline was not operational at 1.4 and 2.4 GHz. This resulted in 3km beingthe longest baseline during the observations of this source at these frequencies.
New Population of X-ray Emitting Massive Stars 5The K -band spectrum of AX J184738–0156 obtained by Blum et al. (1999) and our OSIRIS K -band spectra of AX J163252–4746, AX J144701–5919 and AX J144547–5931 all show strong emission lines from the λ µ m complex and Br γ /He I λ µ m. The presence of such emission lines are indicative of these stars being massive ( > M ⊙ ), from either the late-typeWR stars in the nitrogen sequence (WN) that are hydrogen-rich or their O-type supergiant progenitors, particularly of the OIf + variety (Martins et al. 2008). WR is the name traditionally given to hydrogen-poor, emission line stars, however, there is a subsetof the nitrogen sequence WR stars that have hydrogen in their spectra. Smith & Conti (2008) proposed a new designation forthis subset, WNH, as they are different from classical WR stars in that they are still undergoing hydrogen-core burning and aretherefore at an earlier stage in their evolution. WNH stars are some of the most massive in our Galaxy, with a mass distributionthat peaks around ∼ M ⊙ , but which can reach as high as 120 M ⊙ (Smith & Conti 2008), making them very short-lived andrare. They can have luminosities as high as 2 × L ⊙ and large mass-loss rates of ˙ M > - M ⊙ yr - due to fast extended winds(Martins et al. 2008). OIf + stars are also extremely massive with very similar properties (Martins et al. 2008). Both WNH andOIf + stars are known X-ray emitters (e.g. Mauerhan et al. 2010).WNH and OIf + stars also produce other spectral line features allowing us to distinguish between the different subtypes. The K -band spectrum of AX J163252–4746 (Figure 3(a)) is remarkable, in that it exhibits an ultra-strong He I λ µ m emissionline. The total equivalent width is ≈ ≈
74 A (1080 km s - ). The spectrum also exhibits strong emissionfeatures from the λ µ m complex (EW ≈
80 A) and Br γ /He I λ µ m (EW ≈
90 A). The relative strengths of theemission lines, especially the very strong He I emission, are similar to those of the Ofpe/WN9 stars WR 122 and WR 85a (Morriset al. 1996; Figer et al. 1997), although the individual line strengths are significantly stronger for AX J163252–4746 than forthese other sources. Ofpe/WN9 stars, also known as “slash stars,” have been described as both WNHs or luminous blue variables(Smith & Conti 2008). We assign AX J163252–4746 the spectral type Ofpe/WN9.The zoomed view of AX J163252–4746, seen in the lower panel of Figure 3(a), show fainter emission features that could indicatethe presence of a massive companion. There is a spectral hint of broad He I λ µ m emission under the narrower line ofthe brighter star and a broad spectral feature from He II λ µ m. Though He II λ µ m is a known emission featureof Ofpe/WN9 stars, the broadness of this line and the He I feature are both reminiscent of an early-type WN star (WNE) of theWN4–6 variety (Figer et al. 1997). If these features are real, their weakness may be due to the fact that WNE stars are up totwo magnitudes fainter than Ofpe/WN9 stars, intrinsically (Crowther et al. 2006). The questionable weak feature near λ µ m, speculated to be N V emission, could also indicate the presence of an underlying WNE spectrum since this line requiresthe associated high temperatures of a WNE star. The weak N III λ µ m doublet could be from either star, though is moreordinarily seen in late-type WNs, which include the WN7–9 subtypes (Figer et al. 1997; Martins et al. 2008). Finally, as a separatespeculation, we note that the shape of the peak in the He I λ µ m line, seen in the top panel of Figure 3(a), appears to have adouble peaked profile, as if it were a blend of two velocity-shifted components having comparable line strengths. However, noneof the other strong lines exhibit this profile. Given the current data this is all very speculative. We defer any definitive conclusionsable the weak spectral features or the potential contribution of a companion spectrum until further observations are conducted.Higher resolution spectroscopic measurements over several epochs will be required to determine if the weak, underlying, broadfeatures are real, and to identify this source as a spectroscopic binary.As noted in § . .
2, AX J184738–0156 is coincident with WR 121a, which was identified by Blum et al. (1999) as a star of subtypeWN7. The K -band spectrum of this star, seen in Figure 5 of Blum et al. (1999), is similar to WN7–8h stars, which are WNHsubtypes (e.g. see Martins et al. 2008; Mauerhan et al. 2010). X-ray emission has not been connected with WR 121a prior to thispaper. However, there is evidence to suggest that WR 121a and its surrounding star-formation region, W43, may be associatedwith the extended (intrinsic rms size of 0 . ◦ ± . ◦ ) very-high-energy (E > γ -ray source HESS J1848–018 (Chaveset al. 2008) and its possible counterpart, 0FGL J1848.6–0138, detected with the Large Area Telescope (LAT) onboard the FermiGamma-ray Space Telescope (Tam et al. 2010). W43 is also host to the bright H II region G30.8–0.2 and a giant molecular cloud,which partially overlaps with HESS J1848–018, their centroids separated by ∼ . ◦ . WR 121a lies within the region of this HESSsource’s extended emission, ∼ . ◦ from its centroid (Chaves et al. 2008). A similar offset exists between HESS J1023–575 andthe young open cluster, Westerlund 2, host to WR 20a (Aharonian et al. 2007) yet the extended nature of this HESS source, andtherefore HESS J1848–018, argues against a single star origin for these VHE γ -ray sources (Aharonian et al. 2007; Chaves et al.2008).For AX J144701–5919 the Br γ /He I emission line in the K -band spectrum (Figure 3(b)) is stronger than the λ µ memission line complex indicating a spectral type later than WN6. As the He II λ µ m feature is also seen in emission weexpect a spectral type earlier than WN8-9 as this feature would otherwise appear in absorption or would have a P-Cygni profile(e.g., see Martins et al. 2008). The ratio of equivalent widths of the detected lines can be used to quantitatively determine theWN subtype. We measure EW ( λ µ m)/EW( λ µ m) = 0.4 and EW ( λ µ m)/EW( λ µ m) = 0.3, which aremost consistent with WN7–8 subtypes (Figer et al. 1997; Martins et al. 2008). We therefore classify the counterpart to AXJ144701–5919 as a WNH star of subtype WN7–8h.The K -band spectrum of AX J144547–5931 (Figure 3(b)) shows He II λ µ m in absorption, which is indicative of either aWN8–9 or an OIf + star (Martins et al. 2008). As the λ µ m complex and the Br γ /He I λ µ m emission line areof similar strength it is more likely an OIf + subtype (Martins et al. 2008). The LDSS-3 optical spectrum of AX J144547–5931(Figure 3(c)) is missing the He I λ µ m emission line seen very strongly in optical spectra of WN8–9 stars (e.g., see Corradiet al. 2010). It does have a very broad H α line, as well as C III λ III λ + stars identified by DeBecker et al. (2006, 2009). Further spectroscopic follow-up is required to constrain the OIf + subtype of AX J144547–5931. This Anderson et al.could be achieved by the detection of Si IV emission at λ + stars (Walborn 1971). Distances and X-ray Luminosities
Stars with the spectral classification of AX J163252–4746, Ofpe/WN9, have a narrow range in absolute K -band magnitude ( M K ),which can be combined with an estimate of the extinction along the line of sight to determine the approximate distance to thesource. This method is outlined and tabulated by Hadfield et al. (2007) and Mauerhan et al. (2009), and adopts intrinsic colorsfrom Crowther et al. (2006). Assuming the colors and M K value for a WN10–11 star (aka Ofpe/WN9) from Crowther et al. (2006)we derive an approximate distance of 4.9 kpc. Using red clump stars to measure the reddening as a function of distance along theline of site (López-Corredoira et al. 2002) we can also calculate a lower limit of & . ∼ . L x ≈ . × erg s - in the 0 . - . This is also consistentwith the red clump lower limit of & . L x ≈ . × erg s - in the 0 . - . ∼ ′ and are coincident with a star formationcomplex. There are close surrounding H II regions, G316.808–0.037, G317.037+0.300 and G317.291+0.268, for which Caswell& Haynes (1987) used radio recombination line velocities to calculate a near and a far kinematic distance to each. As Shaveret al. (1981) present radio and infrared observations indicating that G316.808–0.037 is located at the closer kinematic distanceof ∼ . ∼ . ∼ . & . K -band magnitude.By averaging together the M k values of two WN7–8h stars in the Arches cluster (Martins et al. 2008) and using the Crowtheret al. (2006) colors for WN7–9 stars to estimate the extinction along the line of sight, we further refine our distance estimate to ∼ . L x ≈ . × erg s - in the 0 . - . II regions.As we do not have a strong enough constraint on the OIf + luminosity class we cannot apply the same method to AX J144547–5931. We therefore derive a range of possible unabsorbed X-ray luminosities based on both the kinematic and red clump distances,2 . - . L x ≈ . - . × erg s - in the 0 . - . AX J163252–4746 and AX J184738–0156
AX J163252–4746 and AX J184738–0156 are the two most luminous sources out of the four investigated in this paper. Wediscuss these sources together as they are both associated with WNH stars and have similar X-ray luminosities and characteristicsas well as comparable observational information. There are three possibilities as to how these WNH stars can produce the X-rayemission we observe. These include X-rays that are intrinsic to the massive star, X-rays produced through gravitation accretionin a HMXB or X-rays from shock-heated gas in a CWB. We now discuss each possibility in detail.Many WN stars have been shown to have thermal X-ray emission created through instability-driven wind-shocks with typicaltemperatures around kT ≈ . XMM spectra of AX J163252–4746 and AX J184738–0156(Figure 2) show that few X-ray photons were detected with energies < kT = 3 . kT ∼ L x ≈ erg s - ; 0 . -
10 keV) associated with the O5.5 V star θ Orionis C. However, there are problemsapplying this same model to WR stars as they have far stronger wind momenta than O stars. Mauerhan et al. (2010) showedthat the WR stars and O supergiants in their sample, which have similar X-ray properties and luminosities to the four sourcesdiscussed in this paper, would require ≈ <
10 keV, any non-thermal X-rays are likely to make an Revised for a modern distance of ∼ New Population of X-ray Emitting Massive Stars 7insignificant contribution to the overall stellar X-ray spectrum (Pittard & Parkin 2010). We assume that this is also the case forWNH stars as they are only slightly more evolved than massive O stars (Smith & Conti 2008).Given the above arguments it is unlikely that the majority of X-ray emission from either AX J163252–4746 or AX J184738–0156 is generated through the instability-driven or magnetically channeled wind-shock mechanisms, or through inverse-Comptonscattering. The X-rays are therefore not intrinsic to the WNH counterparts, so further scenarios need to be considered.The radio spectrum of a single WR star with an optically thick stellar wind is thermal and has a spectral index of α ≈ + . ± . < + . α ≈ + . L x ∼ - erg s - (Sguera et al. 2006). Currently we only know of four HMXBswith WR companions, three of which are SGXB BH candidates and only one, Cygnus X–3, is found in the Milky Way (Szosteket al. 2008). The other, OAO 1657–415, which is also the AGPS source AX J170047–4139 (Sugizaki et al. 2001), is an X-raypulsar with an Ofpe/WN9 companion (Mason et al. 2009). This source has an X-ray luminosity of L x ∼ - erg s - andappears to be transitioning from a wind-fed SGXB into a BeX disk-fed system (Mason et al. 2009). “Classical” SGXBs andsystems such as OAO 1657–415 are therefore too luminous to explain the X-ray emission seen in AX J163252–4746 and AXJ184738–0156.However, the unabsorbed X-ray luminosities of AX J163252–4746 and AX J184738–0156 are far more consistent with thequiescent luminosities of a subclass of SGXBs that have recently been discovered with the Integral satellite, called supergiantfast X-ray transients (SFXTs: Sguera et al. 2006). These systems are neutron star HMXBs that produce extremely bright, rapid X-ray bursts with luminosities reaching L x ∼ erg s - (Chaty 2010) and have quiescent luminosities at or below L x ∼ - erg s - (Sguera et al. 2006). Currently very few sources with X-ray bursts have been confirmed as SFXTs, most of which containX-ray pulsars (Negueruela et al. 2006; Sguera et al. 2006). The X-ray luminosities of AX J163252–4746 and AX J184738–0156,listed in Table 2, are at least an order of magnitude brighter than the quiescent luminosities of SFXTs and the X-ray flux observedfor each source during the ASCA , XMM and
Chandra observations shows no evidence for long term variability. A thoroughinvestigation of the
XMM and
Chandra observations also show no evidence of short term variability, flaring or pulsation. SFXTsalso have hard, nonthermal X-ray spectra that are well described by a power law with a photon index of Γ ∼ XMM spectra of AX J163252–4746 and AX J184738–0156 are much steeper than this with photonindices of Γ ∼ XMM spectra of AX J163252 - - XXV emission line at ∼ . I - Fe XII ) have been seen in many HMXB X-rayspectra (e.g., Tomsick et al. 2009), created by cool accretion disks irradiated by a high-energy source (Caballero-García et al.2009), it is hard to explain the presence of the ∼ . XXV in the X-ray spectra of these two sources. This line can onlybe created in an environment capable of producing a very highly ionized plasma (Reig 1999). Given this fact, combined with theother X-ray properties of AX J163252–4746 and AX J184738–0156, a HMXB interpretation does not seem viable.We now consider a CWB interpretation for these two sources. The X-ray luminosities from WNH CWBs are usually between L x ∼ - erg s - (Mauerhan et al. 2010). Our estimates of the unabsorbed luminosities of AX J163252–4746 and AXJ184738–0156 are consistent with this range. The very hot thermal plasma ( kT > XMM spectra of AX J163252–4746 and AX J184738–0156 indicates binarity, where the hard X-ray emission could be generatedby colliding supersonic winds in a CWB (Prilutskii & Usov 1976; Cherepashchuk 1976; Usov 1992).The interface between the winds of two massive stars is capable of generating the hot thermal plasma necessary to produce the6.7 keV Fe
XXV emission line seen in the
XMM spectra of AX J163252–4746 and AX J184738–0156. Raassen et al. (2003)argue that for WR 25, another WNH star classified as WN6h, the 6.7 keV Fe
XXV emission line seen in its X-ray spectrum canonly be created in a wide binary, made up of this star and a massive WR or O star companion, with their winds thus colliding atspeeds close to their terminal velocities of over 1000 kms - (van der Hucht 2001). It should be noted here that this might not betrue for all WR CWBs (see Zhekov & Park 2010). As both AX J163252–4746 and AX J184738–0156 show a lack of variabilityand periodicity in X-rays, have thermal X-ray plasma temperatures and luminosities similar to known WNH CWBs, and alsodisplay the 6.7 keV Fe XXV emission line in their X-ray spectra, we classify these two sources as colliding-wind binaries.The same relationship that we see between the X-ray and bolometric luminosities of single O stars, attributed to instability-drivenwind-shocks, has been shown to hold for WN + O and O + O binaries, where log( L x / L bol ) ≈ - < . L x / L bol ) > -
7. We calculated a range of log( L x / L bol ) for AX J163252–4746 and AX J184738–0156 by averaging the L bol values of the Ofpe/WN9 stars listed in Oskinova (2005) and Mauerhan et al.(2010) for AX J163252–4746 and the WN7–8h stars in Mauerhan et al. (2010) for AX J184738–0156. For both sources we findlog( L x / L bol ) & - .
4, as listed in Table 2. There are very few examples of WN CWBs where log( L x / L bol ) > - L x / L bol ) ∼ - . L x / L bol ) ∼ - . AX J144701–5919 and AX J144547–5931
AX J144701–5919 and AX J144547–5931 are much fainter and somewhat softer than AX J163252–4746 and AX J184738–0156but all four sources have similar massive counterparts. We now address the isolated massive star, HMXB and CWB scenarios forAX J144701–5919 and AX J144547–5931.Due to the low number of counts in their
Chandra observations we cannot obtain a reliable spectral fit or identify any X-rayemission lines in the X-ray spectra of AX J144701–5919 and AX J144547–5931. Instead we explore the hardness of their X-rayspectra by seeing how the 50% and 75% photon energy ( E and E , respectively), the energy below which 50% and 75% of thephotons energies are found, of these two sources compare to the same values for AX J163252–4746 and AX J184738–0156. The E value indicates whether a given source is hard ( E > E < E keV grades the level of hardnesswithin the hard category. In all four cases E is between 2 and 3 keV. The E value of AX J163252–4746, AX J184738–0156,AX J144701–5919 and AX J144547–5931 are ∼ . ∼ . ∼ . ∼ . § .
3, we calculated a range of log( L x / L bol ) for bothsources, listed in Table 2. In each case log( L x / L bol ) > - . L x / L bol ) ≈ - L x / L bol ) > -
7) putatively single WN and O stars, as well as CWBs, withlog( L x / L bol ) ≈ - .
5, have been detected (e.g. Oskinova 2005; Sana et al. 2006; Antokhin et al. 2008; Mauerhan et al. 2010).There is no evidence to suggest long term variability between the original
ASCA and recent
Chandra observations of AX J144701–5919 and AX J144547–5931. This, combined with the lack of short term variability or periodicity down to 6.4s seen in the
Chandra observations, is more supportive of a single star or CWB scenario rather than a SFXT scenario. However, longer X-rayobservations are required to confirm this interpretation.The radio spectral index of AX J144701–5919 is negative ( α ≈ - .
5) and therefore dominated by non-thermal emission. Thisspectral index predicts a flux of ∼
38 mJy at 843 MHz, which is much higher than the fluxes measured by MOST at thisfrequency ( ∼ Chandra observations of AX J144701–5919 and AX J144547–5931 show these sources to have similar X-rayluminosities to those seen in putatively single massive stars, CWBs and quiescent SFXTs. As log( L x / L bol ) > - . kT > CONCLUSIONS
We have characterized the multi-wavelength properties of four previously unidentified
ASCA
Galactic plane survey sources,AX J163252–4746, AX J184738–0156, AX J144701–5919 and AX J144547–5931, identifying them as X-ray emitting massivestars. Our data are most consistent with colliding-wind binary classifications for AX J163252–4746 and AX J184738–0156.AX J144547–5931 and AX J144701–5919 require longer X-ray observations to distinguish between a colliding-wind binary, ahigh-mass X-ray binary or an extremely X-ray luminous isolated massive star.In the full ChIcAGO survey we have so far observed 95
ASCA
Galactic Plane Survey sources with
Chandra , ∼
10% of whichemit hard X-rays (median energies > L x ∼ . × erg s - (Motch et al. 2010). It ispossible that the other AGPS sources with similar X-ray and infrared properties to the four discussed in this paper could be moreX-ray emitting CWBs or X-ray binary objects. New Population of X-ray Emitting Massive Stars 9The counterparts to AX J163252–4746, AX J184738–0156, AX J144701–5919 and AX J144547–5931 as well as AX J183116–1008, are all extremely massive stars, making them short-lived and therefore rare, absorbed and only found in the Galactic plane.This makes the ASCA
Galactic Plane Survey and our follow-up ChIcAGO project, an efficient way of locating and identifyingmassive stars in our Galaxy.Special thanks goes to Michael Muno for his encouragement, expertise and participation in this project. We also thank SeanFarrell, Stan Owocki and Nathan Smith for their advice and help with this research, and the referee for their constructive responseand suggestions. G.E.A acknowledges the support of an Australian Postgraduate Award. B.M.G. acknowledges the support of aFederation Fellowship from the Australian Research Council through grant FF0561298. D.L.K. was supported by NASA throughHubble Fellowship grant
Chandra
Award Number GO9-0155X issued by the CXC, which is operated by the Smithsonian Astrophysical Observatory for and onbehalf of NASA. This research makes use of data obtained with the
Chandra X-ray Observatory , and software provided by theCXC in the application packages
CIAO and
Sherpa . OSIRIS is a collaborative project between Ohio State University and CTIO.Observing time on the 6.5m Clay and Baade Magellan Telescopes, located at Las Campanas Observatory, was allocated throughthe Harvard-Smithsonian Center for Astrophysics. The ATCA, part of the Australia Telescope, is funded by the Commonwealthof Australia for operation as a National Facility managed by CSIRO. This publication makes use of data products from the secondcatalogue of serendipitous X-ray sources (2XMMi) from the European Space Agency’s (ESA)
XMM-Newton observatory. Thesedata were accessed through the Leicester Database and Archive Service at Leicester University, UK. 2MASS is a joint projectof the University of Massachusetts and the IPAC/Caltech, funded by the NASA and NFS. GLIMPSE survey data are part of theSpitzer Legacy Program. The
Spitzer Space Telescope is operated by the JPL/Caltech under a contract with NASA. This researchhas made use of NASA’s Astrophysics Data System.
Facilities:
ASCA, ATCA, CXO (ACIS, HRC), Magellan:Baade (PANIC), Magellan:Clay (LDSS-3), Molonglo, SOAR (OSIRIS),XMM (EPIC)0 Anderson et al.
New Population of X-ray Emitting Massive Stars 11 F IG . 1.— Infrared images of the regions surrounding AX J163252–4746, AX J184738–0156, AX J144701–5919 and AX J144547–5931. The white contoursrepresent the smoothed Chandra detection of AX J163252–4746 and AX J184738–0156 (each at 95% of the peak count-rate), and AX J144701–5919 and AXJ144547–5931 (each at 40% of the peak count-rate). In each case the infrared counterpart to the X-ray source is clearly detected. The black contours showthe original
ASCA detection, in the 0 . - . µ m image ofAX J163252–4746. b) PANIC H -band image of the star cluster in which AX J184738–0156 is embedded. The Chandra contour is clearly coincident with thebrightest member of the cluster. The field of view of this image is too small to overlay the
ASCA contours. c) 2MASS K -band image of AX J144701–5919. d)2MASS K -band image of AX J144547–5919. A nd e r s on e t a l . T ABLE
1. X-
RAY O BSERVATIONS OF
AX J163252–4746, AX J184738–0156, AXJ144701–5919
AND
AX J144547–5931Source X-ray Telescope Obs ID Date (yyyy-mm-dd) Instrument Exp Time (s) Total Counts a AX J163252 - Chandra ± XMM ± ± ± - Chandra ± XMM ± ± ± - Chandra ± - Chandra ± a Total number of counts in the 0 . - . Chandra energy range (0 . - . . - . XMM . N e w P opu l a ti ono f X -r a y E m itti ng M a ss i v e S t a r s T ABLE
2. X-
RAY P OSITIONS AND S PECTRAL M ODELING OF
AX J163252–4746, AXJ184738–0156, AX J144701–5919
AND
AX J144547–5931Source
Chandra
Position a Telescope Distance Raymond-Smith Fit Parameters c AX RA(J2000) Dec(J2000) Obs ID b (kpc) kT (keV) N H Abundances d F x , unabs log( L x / L bol ) e χ red F x , abs L x , unabs J163252–4746 16:32:48.548 -47:45:06.20
XMM ∼ . . ± . . ± . . ± . . + . - . ∼ -5.4 – -5.10201700301 0.9 3 . ± . ∼ . XMM ∼ . . ± . . ± . . ± . . + . - . ∼ -5.4 – -5.00203850101 0.8 2 . ± . ∼ . Chandra > .
9) 10 . + . - . > . Chandra ∼ . . + . - . . + . - . ∼ ∼ -6.4 – -6.08233 0.2 0 . + . - . ∼ . Chandra ∼ . . + . - . . + . - . ∼ ∼ -6.5 – -6.38240 0.3 0.3 ( < . ∼ N OTE . — In the cases where one of the 90% errors for a given spectral fit parameter was incalculable we quote the best fit parameter followed by a 90% confidence upper or lower limit. The
XMM spectra were used rather than the
Chandra spectrum to calculate the unabsorbed flux and luminosity of AX J184738–0156 as the model fits to the
XMM data are far more statistically significant. a All position errors are within a 0 . ′′ b X-ray telescope and corresponding observation identification used in the spectral modeling. c The best fit Raymond-Smith model parameters including temperature kT (keV), absorption column density N H (10 cm - ), reduced chi-sqaure χ red from ChiGehrels statistics and abundances relative to the solar value given by Anders & Grevesse (1989). The absorbed X-ray flux, F x , is in the 0 . - . . - . Chandra and
XMM fits, respectively. The unabsorbed X-ray flux, F x , unab (10 - erg cm - s - ) and the unabsorbed X-ray luminosity, L x , unab , (10 ergs s - ) are calculated from the Raymond-Smith fits over the 0 . - . d The abundance parameter was prevented from varying during the fitting process of the
Chandra spectra. The parameter was set to the number listed, which isrelative to the solar value given by Anders & Grevesse (1989). The abundance parameter was allowed to vary during the fitting of the
XMM spectra. e A range of possible log( L x / L bol ) calculated using the bolometric luminosity, L bol (ergs s - ), of sources with similar spectral types taken from Table 3 ofMauerhan et al. (2010) and Table 5 of Oskinova (2005). This ratio was calculated using WN7–8h stars for AX J184738–0156 as its K -band spectral morphology(see Figure 5 in Blum et al. 1999) is similar to stars of this subtype seen in Martins et al. (2008) and Mauerhan et al. (2010). In the case of AX J144547–5931 weused the value of L bol that Mauerhan et al. (2010) calculated for X-ray source CXOGC J174617.0–285131, which is classified as an O6If + star. A nd e r s on e t a l . T ABLE
3. I
NFRARED M AGNITUDES AND R ADIO F LUXES OF COUNTERPARTS TO
AXJ163252–4746, AX J184738–0156, AX J144701–5919
AND
AX J144547–5931Source 2MASS Magnitudes a GLIMPSE Magnitudes b ATCA Fluxes (mJy) c AX J H K s m . m . m . m . ± > . > .
27 8.18 ± ± ± ± < . < . ± ± ± ± ± ± ∗ ± ± ± ± ∗ ± ∗ ± ± ± ± ± ± ± ± ± ± ± ± ± N OTE . — AX J184738–0156 was detected in the GLIMPSE 3.6, 4.5 and 8.0 µ m bands but no magnitudes were obtained as the source was either saturated or the surrounding region too confused. a J, H and K s counterpart magnitudes are taken from the 2MASS All-Sky Catalog of Point Sources (Skrutskie et al. 2006). Due to blending with nearby sourcesonly 95% confidence lower limits are quoted for the H and K magnitudes of AX J163252–4746. b m . - m . refers to the counterpart magnitudes at 3 . - . µ m taken from the GLIMPSE I DR5 Catalog. Magnitudes labeled with ∗ are from the GLIMPSE I DR5 Archive, which is described as more complete but less reliable than the GLIMPSE I DR5 Catalog. c ATCA radio observations were only obtained for AX J144701–5919 and AX J163252–4746. The 1 σ flux errors are quoted. No radio counterpart to AXJ163252–4746 was detected at 1.4 and 2.4 GHz so 5 σ flux upper-limits are quoted. New Population of X-ray Emitting Massive Stars 15 F IG . 2.— X-ray spectra of AX J163252–4746 and AX J184738–0156. The XMM
EPIC-MOS1 (black) and EPIC-MOS2 (red) spectra of these sources (datapoints) are overlaid with an absorbed Raymond-Smith fit (solid lines). The bottom panel in both a) and b) show the residuals of the best-fit Raymond-Smithplasma for each detector. a)
XMM spectra of AX J163252–4746. b)
XMM spectra of AX J184738–0156 overlaid with the
Chandra (blue) spectrum of this source.The solid blue line is the
Chandra spectral fit using the best-fit parameters from the absorbed Raymond-Smith fit to the
XMM data but allowing the normalizationto vary. The