Multiwavelength Observations of the Runaway Binary HD 15137
M. Virginia McSwain, Michael De Becker, Mallory S. E. Roberts, Tabetha S. Boyajian, Douglas R. Gies, Erika D. Grundstrom, Christina Aragona, Amber N. Marsh, Rachael M. Roettenbacher
MMultiwavelength Observations of the Runaway Binary HD 15137 M. Virginia McSwain Department of Physics, Lehigh University, Bethlehem, PA 18015 [email protected]
Micha¨el De Becker
Institut d’Astrophysique et G´eophysique, Universit´e de Li`ege, FNRS, Belgium [email protected]
Mallory S. E. Roberts
Eureka Scientific, Inc., Oakland, CA 94602-3017 [email protected]
Tabetha S. Boyajian , , Douglas R. Gies Department of Physics and Astronomy, Georgia State University, Atlanta, GA 30302-4106 [email protected], [email protected]
Erika D. Grundstrom Physics and Astronomy Department, Vanderbilt University, Nashville, TN 37235 [email protected]
Christina Aragona , Amber N. Marsh , Rachael M. Roettenbacher Department of Physics, Lehigh University, Bethlehem, PA 18015 [email protected], [email protected], [email protected]
ABSTRACT
HD 15137 is an intriguing runaway O-type binary system that offers a rare opportunity toexplore the mechanism by which it was ejected from the open cluster of its birth. Here we presentrecent blue optical spectra of HD 15137 and derive a new orbital solution for the spectroscopicbinary and physical parameters of the O star primary. We also present the first
XMM-Newton observations of the system. Fits of the EPIC spectra indicate soft, thermal X-ray emissionconsistent with an isolated O star. Upper limits on the undetected hard X-ray emission placelimits on the emission from a proposed compact companion in the system, and we rule out aquiescent neutron star in the propellor regime or a weakly accreting neutron star. An unevolvedsecondary companion is also not detected in our optical spectra of the binary, and it is difficult toconclude that a gravitational interaction could have ejected this runaway binary with a low massoptical star. HD 15137 may contain an elusive neutron star in the ejector regime or a quiescentblack hole with conditions unfavorable for accretion at the time of our observations.
Subject headings: stars: individual(HD 15137), binaries: spectroscopic, X-rays: binaries a r X i v : . [ a s t r o - ph . S R ] D ec . Introduction While most O- and B-type stars are believedto form in open clusters and stellar associations,some are observed at high galactic latitudes andwith large peculiar space velocities (Gies & Bolton1986). These stars were likely ejected from theclusters of their birth, and there are two acceptedmechanisms to explain the origin of their run-away velocities. Close multi-body interactionsin a dense cluster environment may cause oneor more stars to be scattered out of the region(Poveda et al. 1967), and about 10% are expectedto be ejected as binary pairs (Leonard & Duncan1990). An alternative mechanism involves a su-pernova explosion within a close binary, ejectingthe secondary due to the conservation of momen-tum (Zwicky 1957). The resulting neutron star(NS) or black hole (BH) may remain bound to thesecondary if not enough mass is lost during theexplosion or if a retrograde kick velocity occurs,and a binary fraction of 20 −
40% is predicted forthese runaways (Portegies Zwart 2000). The ob-served fraction of binaries among runaways seemsconsistent with either scenario (5 − Visiting Astronomer, Kitt Peak National Observatory.KPNO is operated by AURA, Inc. under contract to theNational Science Foundation. Hubble Fellow Based partly on observations collected at the Observa-toire de Haute-Provence (France) and on observations ob-tained with XMM-Newton, an ESA science mission with in-struments and contributions directly funded by ESA Mem-ber States and NASA. none have been found (Philp et al. 1996; Sayer etal. 1996). X-ray emission is a better diagnosticsince stellar wind interactions with a NS producea measurable X-ray excess above the normal, in-trinsic emission from shocks in the stellar winds(Sana et al. 2006; Popov et al. 2000; Lamers et al.1976).HD 15137 is a runaway binary that offers therare opportunity to diagnose its cluster ejectionmechanism. Its spectral type is O9.5 III(n). Boy-ajian et al. (2005) found that this runaway single-line spectroscopic binary (SB1) system was likelyejected from the open cluster NGC 654, but thetravel time since its ejection, 10 Myr, presents anunusual paradox since it is longer than the ex-pected lifetime of an O-type star. Boyajian et al.(2005) and McSwain et al. (2007a) presented pre-liminary orbital solutions for the SB1, finding ahighly eccentric orbit with a period of about 30d. The O-type star has an effective temperature T eff = 29700 K and surface gravity log g = 3 .
50, islocated at a distance of 2.420 kpc, and has a fastpeculiar space velocity V pec = 63 km s − (Mc-Swain et al. 2007a). Such a value of V pec placesthis SB1 system among the class of runaway Ostars.The very low mass function suggests a low masscompanion in HD 15137, yet so far Boyajian et al.(2005) and McSwain et al. (2007b) have tried un-successfully to identify the nature of the compan-ion star. Using doppler tomographic separationwith their preliminary orbital solution, Boyajianet al. (2005) were unable to identify an unevolvedsecondary companion in their red optical spectra.Furthermore, we simulated 4-body interactions ofHD 15137 with the STARLAB package and founda very low probability that dynamical interactionscould have ejected a low mass, unevolved star withthe O-type primary (McSwain et al. 2007b). Itsrelatively short period, eccentric orbit led Boya-jian et al. (2005) to suggest that the orbit waswidened during a supernova event and that tidalcircularization has not yet occurred. The systemis not a known X-ray source, so it has been pro-posed to be a high mass X-ray binary (HMXB)candidate. Current upper limits of the X-ray fluxrule out a stellar wind-accreting NS in the system,and we performed a radio pulsar search with nullresults (McSwain et al. 2007b).In this work, we present new blue optical spec-2ra of HD 15137 in §
2. Based on new radial veloc-ities of the system, we argue that the orbital pe-riod is somewhat longer than we have previouslydetermined, and we present an improved orbitalsolution for this SB1. In §
3, we present the stellarphysical parameters determined by comparing ouroptical spectra to model spectra. We describe thefirst X-ray detection of HD 15137 with the
XMM-Newton observatory in §
4. From fits of the softX-ray spectra, we place upper limits on the hardX-ray flux from a potential NS companion. Wediscuss these results further in §
5, and summarizethe key results in §
2. Optical Observations and Radial Veloc-ity Measurements
We obtained 44 spectra of HD 15137 at the Ob-servatoire de Haute-Provence (OHP) during sev-eral observing runs from 2005 October to 2007November, including 23 spectra obtained over 30consecutive nights in fall 2007. We used the 1.52mtelescope with the Aur´elie spectrograph with grat-ing × µ m. These spec-tra cover a wavelength range between 4460–4890˚A with a resolving power of R = λ/ ∆ λ ≈ R ≈ . Finally, thecoud´e feed spectra were rectified to a unit contin-uum using line-free regions and interpolated ontoa log wavelength scale using a common heliocen-tric wavelength grid.The recent optical spectra of HD 15137 showevidence of low amplitude, high frequency line pro-file variations, especially during the long staresthat were performed in each of the runs describedabove. The variations in the He I λ V r , variations of eachabsorption line. Thus HD 15137 appears to bea single-line spectroscopic binary (SB1) that alsoexhibits line profile variations.To measure V r of HD 15137, we used a cross-correlation procedure similar to that described inMcSwain et al. (2007a). However, we used a meanspectrum created from each data set, rather thanany single observation, as a reference spectrum forthe cross correlation process. This choice allowedus to minimize the effects of small-scale line pro-file variations and identify the bulk V r shifts inthe data instead. Using this mean spectrum, wefit the core of each absorption line with a parabolato determine its absolute radial velocity. The re-maining spectra were then cross correlated withthis template to determine V r for each line. We IRAF is distributed by the National Optical AstronomyObservatory, which is operated by the Association of Uni-versities for Research in Astronomy, Inc., under cooperativeagreement with the National Sciences Foundation V r , and rest wavelengths for eachline were taken from the NIST Atomic SpectraDatabase . In the CF data set, the lines used tomeasure V r were H γ , He I λλ , , II λ β , He I λλ , II λλ , V r and standard deviation, σ , are listed in Table 1. In many cases, the mea-sured σ probably underestimate the true errorin V r due to the line profile variations. Duringboth of our long stares at HD 15137, we observedchanges in V r of about 10 km s − over only a fewhours. We recommend adding this error of ± − in quadrature with σ for a more represen-tative error in V r .In our previous spectroscopic studies of HD15137 (Boyajian et al. 2005; McSwain et al.2007a), we proposed an orbital period P ∼
30d for the system. However, our new V r measure-ments exclude such a value of P , especially usingour data from two observing runs that each tookplace over ∼
30 consecutive nights. We performeda new period search on all available V r measure-ments of HD 15137 using a version of the discreteFourier transform and CLEAN deconvolution al-gorithm of Roberts, Leh´ar, & Dreher (1987) (writ-ten in IDL by A. W. Fullerton). In addition, weused the phase dispersion minimization (PDM)algorithm (Stellingwerf 1978) to compare with theCLEANed power spectrum. The PDM method isbetter suited for eccentric orbits that are stronglynon-sinusoidal.There was no one clear signal that stands outfrom any of the resulting periodograms, so we in-spected each candidate frequency carefully. Weused each proposed period as input into the non-linear, least-squares fitting program of Morbey &Brosterhus (1974) to solve for the resulting orbitalelements. After ruling out all resulting V r curveswith poor fits and extremely large scatter (andalso any P <
35 d), we settled upon two accept-able solutions with nearly equally good fits andfrequencies 0.018 d − and 0.015 d − , correspond-ing to P = 55 . P = 65 . The NIST Atomic Spectra Database is available online athttp://physics.nist.gov/PhysRefData/ASD/index.html. IDL is a registered trademark of Research Systems, Inc.
Since our data are often spaced at intervals onthe order of 30 d, this produces a sampling fre-quency f s = 0 .
033 d − , which is also a signalfound by CLEAN (and also corresponds to our ear-lier proposed period). Our sampling rate inducesan alias frequency f a = f s − f , where f is thetrue signal frequency. Since no significant alias-ing would occur if the Nyquist condition is satis-fied, f < f s /
2, we attribute the lower frequencyas the alias and adopt the true orbital frequency f = 0 .
018 d − . Thus we adopt an orbital period P = 55 . − and0.007 d − found by CLEAN are the result.As further verification of our adopted orbitalperiod, we tried manipulating our collection of V r data in several ways to test whether the periodsearch results remained consistent. We performeda period search with the CF dataset omitted, find-ing two potential periods of 57 d and 67 d. Whenwe instead omitted the OHP data from the col-lection, a period of 52 d stands alone as the mostprominent signal. Finally, we repeated the periodsearch an additional ten times, randomly removing20% of all of the V r points in each trial. Promi-nent signals remain consistently at 55 d and 65d. These results provide further support that thetimescales of 55 d and 65 d are good candidatesfor the orbital period, with the former preferred.However, we recommend that both values be con-sidered with caution since low frequency trendsalso appear to contribute to the temporal behaviorof the radial velocities. The corresponding orbitalparameters should be treated with caution as well.Finally, we inspected the complete V r datasetfor systematic velocity differences by repeating theorbital fit for each set separately, allowing only V to vary. The V r offset was determined by bringingeach V into agreement with V from the CF set.Thus we offset the OHP measurements by − . − , measurements from Boyajian et al. (2005)by +2 .
7, and measurements from McSwain et al.(2007a) by +2 .
4. Note that the unaltered
OHPmeasurements are presented in Table 1. We thenrepeated the period search and orbital fit with thecorrected V r , which improved our orbital solutionsignificantly. Figure 2 shows the results from thefinal period searches. Note that for the CLEANmethod, peaks in the signal indicate likely frequen-4ies, while for the PDM method the minima indi-cate more probable frequencies. Our systematic V r correction removed f a in the CLEAN search,but it remains present in the PDM periodogram.We present the final orbital solution in Table2, and the corresponding radial velocity curve inFigure 3. The mean spectrum of HD 15137 fromour CF and OHP runs, shifted to its rest wave-length according to our orbital solution, is shownin Figure 4. The final orbital solution of HD15137 indicates a highly eccentric ( e = 0 .
62) bi-nary with a low velocity semiamplitude ( K = 13 . − ). These orbital elements are very similarto our preliminary orbit (McSwain et al. 2007a),although the period is quite different. While theO9.5 III primary is likely a massive star, the verylow mass function of this binary suggests a lowmass companion. Based on the runaway nature ofthe binary, the eccentric orbit, and the low massfunction, Boyajian et al. (2005) proposed a super-nova ejection scenario and a NS companion in HD15137, even though the system was not a knownX-ray source. They proposed that HD 15137 maybe a “quiet” HMXB, too widely separated for theNS to accrete a significant mass of stellar windsto produce the bright X-ray flux commonly as-sociated with X-ray binaries (Liu et al. 2006). Toinvestigate this quiet HMXB scenario, we describe XMM-Newton observations of the system below in §
3. Physical Parameters from SpectralModels
In McSwain et al. (2007a), we measured thestellar parameters of HD 15137 using the TlustyOSTAR 2002 grid of line-blanketed, non-LTE,plane-parallel, hydrostatic atmosphere modelspectra for O-type stars (Lanz & Hubeny 2003).The OSTAR2002 grid uses a microturbulent ve-locity V t = 10 km s − , and we assumed solarabundances for the star. In our earlier paper wefound V sin i = 234 km s − , T eff = 29700 K, andlog g = 3 . II line profiles. Given that we have obtained higherresolution blue spectra for a new comparison, werepeated the spectral modeling using the TlustyOSTAR2002 grid (Lanz & Hubeny 2007).We measured V sin i by comparing the ob-served He I λλ , , , V sin i values by minimizing the mean square ofthe deviations, rms . The formal error, ∆ V sin i ,is the offset from the best-fit value that increasesthe rms by 2 . / N , where N is the number ofwavelength points within the fit region. We mea-sured a new value of V sin i = 258 ±
20 km s − ,consistent with our earlier result.We measured T eff by comparing the observedHe II :He I equivalent width ratios to the equiva-lent width ratios in the broadened model spectra(Walborn & Fitzpatrick 1990). The OSTAR2002grid overlaps with the Tlusty BSTAR2006 gridover the range of T eff and log g appropriate fora late O-type star, although the BSTAR2006 griduses a lower V t (2 km s − for our values of log g ;Lanz & Hubeny 2007). To consider whether lowervalues of V t may be appropriate for HD 15137, wealso compared our spectra to both grids over theappropriate range in T eff and log g . We measuredequivalent widths, W λ , using numerical integra-tions over each line profile, and we estimate anerror of 10% in each observed W λ due to errorsin the continuum placement and intrinsic noise.By fixing log g at each of three possible values(log g = 3 .
25, 3.5, and 3.75), we determined anominal value of T eff from our observed W λ ratiosby interpolation with the model W λ ratios. Theresulting values of T eff and log g are listed in Ta-ble 3. Using the mean T eff for each value of log g ,we then compared the broadened model spectrumto our mean observed spectra over the full wave-length range and determined a reduced χ fromthe difference. The resulting values of reduced χ for each fit are listed in Table 3. Based uponour renewed inspection of the He I and He II linestrengths, our new measurements for T eff = 29700K and log g = 3 . T eff to be 1700K based on the standard deviation of the threemeasured values, and we estimate an error in log g to be 0.25 dex due to the spacing of the grids ofmodel spectra.The final broadened model spectrum from theOSTAR2002 grid is presented in Figure 4. Com-pared to the model spectrum, HD 15137 has a5lightly weak C III λ III λλ ,
4. XMM-Newton Observations
We observed HD 15137 with the
XMM-Newton observatory on 2008 August 3, observation ID0553810201, for approximately 20 ks. Based onour proposed orbital solution, these observationstook place at orbital phase φ = 0 .
69. The fieldof view was centered on HD 15137 for a direct on-axis view. The three EPIC cameras were operatedin Full Frame mode with the medium optical fil-ter, and the Optical Monitor was turned off duringthe observation due to the brightness of the star.The RGS instruments were operated in StandardSpectroscopy mode, but due to the faintness of thesource the RGS data were not used.The EPIC Observation Data Files (ODFs) andevent lists were provided by the standard XMMPipeline Processing System. Using the XMM-Science Analysis System (SAS) version 7.1.2, wefiltered the event lists using the standard cutoffof 0.35 counts s − for the MOS cameras and 5counts s − for the pn camera using the evselect command to exclude times of high particle back-ground. The resulting effective exposure times(good time intervals) for each camera are listedin Table 4.We extracted the source spectra from the eventlists using a circular region with radius of 25 (cid:48)(cid:48) forthe MOS cameras and 32 (cid:48)(cid:48) for the pn camera. Toextract the background spectrum for each camera,we used a partial annulus region with inner andouter radii of 80 (cid:48)(cid:48) and 110 (cid:48)(cid:48) , respectively, centeredon the source. Because another possible weak X-ray source was detected in this annular region,we excluded the portion of the annulus near thatsource. The resulting count rates for the sourceand background regions are listed in Table 4. Asa weak X-ray source, we found no indication ofpile-up in the observation.The resulting EPIC spectra of HD 15137 havelow signal-to-noise (S/N), but they are consistentwith a soft thermal source typical of isolated O- type stars (Sana et al. 2006). We fit the twoMOS spectra simultaneously, over the range 0.5–2.3 keV, using a variety of warm absorbed ( wabs ;Morrison & McCammon 1983), single temperature(1-T) and two temperature (2-T) thermal mod-els available with Xspec version 11.3.2ag, includ-ing Raymond-Smith (Raymond & Smith 1977),MEKAL (Mewe et al. 1985, 1986; Liedahl et al.1995), APEC (Smith et al. 2001), bremsstrahlung(Karzas & Latter 1961; Kellogg et al. 1975), andplain blackbody models. We repeated our fits forthe pn spectrum using the same group of models.A sample 1-T fit of the pn spectrum is shown inFigure 5. The resulting fits were equally good forthe 1-T and 2-T models, but a statistical F-test re-veals that the second temperature component doesnot significantly improve the fits. We also cannotdistinguish between the quality of the various 1-Tthermal model fits due to the low S/N. We weaklyconstrain the temperature to 0 . ≤ kT ≤ . . × ≤ nH ≤ . × atoms cm − . Thereis no evidence of any hard X-ray photons detectedin our data.Our measured value of nH for HD 15137 canbe compared to its reddening, determined fromthe ultraviolet, optical, and near-infrared spectralenergy distribution (McSwain et al. 2007a). Inthat work, we measured reddening E ( B − V ) =0 .
43 and a ratio of total-to-selective extinction, R = 3 .
18. The optical extinction is thus A V = R × E ( B − V ). Several authors have published nH − A V relations (Bohlin et al. 1978; Wolk etal. 2006; G¨uver & ¨Ozel 2009) that predict nH =2 . − . × atoms cm − based on the measuredreddening of HD 15137. Our measured range in nH is consistent with A V , although our warm ab-sorbed 1-T models also allow values of nH thatexceed the expected value by as much as a factorof 3.Unfortunately, interpreting our X-ray spectrais made somewhat ambiguous since the emergentspectra of cooling NSs is soft and nearly ther-mal, very similar to a blackbody (Treves et al.2000). We performed fits of the EPIC spectrausing a model of the hydrogen atmosphere of aNS ( nsa ; Pavlov et al. 1991; Zavlin et al. 1996).As recommended by those authors, we fixed theNS mass to M NS = 1 . M (cid:12) and the radius to R NS = 10 km. We used a fixed magnetic field6trength fixed to B = 10 G and included warmabsorption fixed to the predicted nH = 3 × atoms cm − based on the observed A V . Fittingthe two MOS spectra simultaneously results in anunredshifted effective temperature of the NS of T eff , NS = 437000 +239000 − K and a normalization of1 . × − . From the fit of the pn spectrum, weobtain T eff , NS = 955000 +394000 − K and a normal-ization of 7 . × − . The reduced χ values areequally good as the 1-T models described above,although the NS temperatures from the MOS andpn spectral fits are very inconsistent with eachother.In order to place an upper limit on any hardpower law component that may originate from anaccreting compact companion, we repeated thewarm absorbed, 1-T models with an additionalpower law component with photon index Γ = 2(fixed). All of the best fit parameters from the 1-T fits, including nH, kT , and their normalizations,were fixed in Xspec. We then refit each model, al-lowing only the normalization of the power lawcomponent to vary. In every case, the best fit nor-malization for the power law component was zero.We then used the steppar routine to investigatethe 90% confidence limit for the power law nor-malization. Upon removing the thermal compo-nent from the models, we used the fixed Γ and theupper limit for its normalization in the remainingabsorbed power law model to determine the up-per limit for the X-ray flux, F X , of the putativecompact object. Fits of the pn and MOS spectraindicated unabsorbed F X (cid:46) − erg cm − s − (over the range 0.2-10 keV). At a distance d = 2 . L X (cid:46) erg s − for any hardpower law component.
5. Discussion
We can loosely constrain the mass of the com-panion in HD 15137 using the mass function, f ( m ), from our new orbital fit. Assuming theO9.5 III star has a mass of 20 . M (cid:12) (Martins et al.2005), and noting that there is an 87% probabilitythat the binary inclination i ≥ ◦ , we constrainthe companion mass to 1 . ≤ M ≤ . M (cid:12) . Aplot of allowable values of M and M is shownin Figure 6. The most probable range in M cer-tainly suggests that a low mass NS or BH could present in the system. The kick velocity from aprior supernova event is a reasonable explanationfor the current high peculiar space velocity (63 ± − ; McSwain et al. 2007a) for the system. Awhite dwarf companion is unlikely due to the ini-tial mass of the supernova progenitor.HD 15137 has a measured mass loss rate,log ˙ M = − . M (cid:12) yr − (Howarth & Prinja 1989)and a terminal wind velocity, v ∞ = 1690 km s − (McSwain et al. 2007b). Using these wind param-eters and our new orbital solution, we can pre-dict the X-ray luminosity emitted from a possiblewind-accreting NS in the system using a simpleBondi-Hoyle wind accretion model (Lamers et al.1976). With the O star mass and a NS compan-ion with mass 1 . M (cid:12) , we should expect to ob-serve an unabsorbed F X ∼ × − erg cm s − (corresponding to an unabsorbed X-ray luminos-ity log L X = 33 . − ) at φ = 0 . L X . However, the XMM-Newton obser-vations rule out such a bright X-ray source byat least two orders of magnitude, clearly rulingout an accreting BH companion. O-type starstypically exhibit order-of-magnitude variations intheir mass-loss rates (McSwain et al. 2004), so ifthe mass loss rate has decreased with time sincemeasured by Howarth & Prinja (1989), the ex-pected wind accretion luminosity should decreaseproportionately. However, variability in ˙ M is notsufficient to explain the discrepancy between thepredicted F X and our observed upper limit.The dynamical age of the putative NS in HD15137 also allows us to discriminate against thecooling NS model. A young NS cools to a tem-perature ∼ K after about 10 –10 yr after itsbirth, and it is expected to remain at that roughlyconstant temperature for ∼ yr (Treves et al.2000). Since the binary was ejected 10 yr agofrom the open cluster of its birth (Boyajian et al.2005), then it should have cooled to temperaturesfar below our measured T eff , NS .If there is a NS or BH companion present inHD 15137, it is more likely in a quiescent state.Quiescent NSs in the propellor regime have beenobserved with hard X-ray spectra with Γ ∼ . (cid:46) log L X (cid:46) . − (Campana et al. 2002). Our upper limitson the hard X-ray flux of HD 15137 clearly rule7ut such a NS in the propellor regime. NSs in theejector regime are potentially observed as radiopulsars, but our pulsar search did not detect anysuch candidate (McSwain et al. 2007b). We can-not firmly rule out a NS companion in HD 15137,but our observations suggest that a companion isunlikely to be present. However, a quiescent BHis quite possible. Black hole X-ray binaries spendmost of their lives in quiescence with an X-rayluminosity as little as 10 − of an active accretor(Pszota et al. 2008; Coriat et al. 2009), consistentwith our upper limits on L X .If HD 15137 has a “twin” companion with massratio q = M /M ∼ .
95 (Pinsonneault & Stanek2006; Lucy 2006), the system must have an im-probably low i ∼ ◦ . The late O-type or early B-type companion would have spectral lines heavilyblended with the primary star, making it difficultto detect from V r variations but possibly identi-fiable using doppler tomography. Boyajian et al.(2005) failed to detect an optical companion in HD15137 using doppler tomography, but we repeatedsuch an effort here with the advantage of a new or-bital solution for the binary. We assumed a widerange of 0 . ≤ q ≤ .
95 in an effort to detect amassive or low mass secondary. The reconstructedspectra from doppler tomography do a poor jobof matching realistic H Balmer line profiles, andall possible reconstructed spectra maintain con-stant He I :He II line strength ratios. These lineratios would be expected to differ from the pri-mary star in any cooler optical companion, so wecannot claim any detection of a possible opticalsecondary star with the tomography.Boyajian et al. (2005) noted that the travel timesince the ejection of HD 15137 presents an un-usual paradox since it is longer than the expectedlifetime of an O-type star. Perets (2009) offersa possible explanation: the close binary may havebeen ejected from its parent cluster by a dynamicalejection, while later mass transfer rejuvenated themassive star and extended its lifetime. The orbitalperiod and probable high mass ratio of HD 15137are consistent with the strong mass transfer sce-nario (Perets 2009). However, with a total binarymass almost certainly > M (cid:12) , the binary mighthave originally been composed of two ∼ M (cid:12) stars. Perets predicts a very small parameterspace for such massive binaries that will experi-ence strong mass transfer; most of the massive bi- naries in their simulations experience weak masstransfer or mergers. While rejuvenation throughmass transfer is possible, it seems an unlikely ex-planation for the current properties of HD 15137.On the other hand, it is also possible that HD15137 experienced a supernova after the binary’sejection via dynamical processes. In such case,the age of the compact companion would not cor-respond to the dynamical age found by Boyajianet al. (2005), and a young, cooling NS compan-ion might be present. However, such a two-stepevolutionary scenario would obscure the true dy-namical history of HD 15137; any supernova kickvelocity would alter the system’s trajectory andrender any measurement of its travel time unre-liable. It is impossible to draw any conclusionsregarding the dynamical ejection and subsequentevolutionary history of HD 15137.
6. Summary and Conclusions
We have searched for a compact companionin the massive runaway binary HD 15137, to noavail.
XMM-Newton
EPIC spectra of the systemare consistent with a soft thermal source typical ofisolated O-type stars (Sana et al. 2006). The dy-namical age of HD 15137 implies the system is tooold to contain a cooling NS with T eff , NS ∼ − K, although our X-ray spectra are consistent withsuch a NS. Our lack of detection of hard X-rayphotons is inconsistent with the expected emis-sion from a quiescent NS in the propellor regimeor a stellar wind-accreting NS, but a quiescent BHremains a possibility.Similarly, we have been unable to identify anoptical secondary in the system. Doppler tomog-raphy based on our updated orbital solution wasunsuccessful, and spectral modeling of the primaryis consistent with a single O9.5 III star. Weak Nenrichment and C depletion are observed in the op-tical spectra, possible evidence of CNO processedgas at the stellar surface. The companion proba-bly has a mass in the range 1–3 M (cid:12) , implying avery low mass ratio of q < .
2. For a MS compan-ion, this would correspond to a late B star whichwould contribute a negligible flux compared to thebright O-type star. Such an optical companionwould be difficult if not impossible to detect inthe close, eccentric orbit of HD 15137.Although it might be tempting to rule out a8ompact companion based on the lack of its de-tection in our
XMM-Newton observation, it is dif-ficult to imagine how the binary might have beenejected as a runaway with a low mass stellar com-panion in a close gravitational encounter. HD15137 bears many similarities with the class ofBe/X-ray biniaries (BeXRB): a moderately longorbital period, high eccentricity, low mass func-tion, and a rapidly rotating primary star. Al-though Boyajian et al. (2005) reported no H α emission from HD 15137, the rapid rotation andpossible nonradial pulsations make the star a goodcandidate for a transient Be star that experi-ences sporadic disk disappearances (McSwain etal. 2009). Many BeXRB are X-ray transients thatare invisible in X-rays except at times when theNS encounters higher density disk gas (there areno known BH companions in BeXRB systems; Bel-czynski & Ziolkowski 2009). X-ray outbursts mayhappen either when a Be star disk outburst oc-curs and the disk density increases close to theNS, or when the NS crosses the plane of the disk(in cases where the orbital and disk planes are dif-ferent). The XMM-Newton observations may haveoccurred when conditions were unfavorable for ac-cretion.We thank the referee, Hagai Perets, for sug-gestions that improved this manuscript. We aregrateful to the staff at KPNO for their hard workto schedule and support these observations. MDwould like to thank Drs. G. Rauw and N. Lin-der for taking some of the OHP spectra, thestaff of OHP for the technical support duringthe various observing runs, and the Minist`ere del’Enseignement Sup´erieur et de la Recherche dela Communaut´e Fran¸caise de Belgique for finan-cial support for the OHP observing runs. MVMand CA would also like to thank the WyomingInfrared Observatory for providing time to finishthis manuscript and for a renewed appreciationof the small things in life. This work is sup-ported by NASA DPR numbers NNX08AX79Gand NNG08E1671 and an institutional grant fromLehigh University.
Facilities:
KPNO:CFT, OHP:Aur´elie, XMM:EPIC.
REFERENCES
Bally, J., & Zinnecker, H. 2005, AJ, 129, 2281 Belczynski, K., & Ziolkowski, J. 2009, ApJ, inpress (astro-ph/0907.4990)Boyajian, T. S., Beaulieu, T. D., Gies, D. R.,Huang, W., McSwain, M. V., Riddle, R. L.,Wingert, D. W., & De Becker, M. 2005, ApJ,621, 978Bohlin, R. C., Savage, B. D., & Drake, J. F. 1978,ApJ, 224, 132Campana, S., Stella, L., Israel, G. L., Moretti, A.,Parmar, A. N., & Orlandini, M. 2002, ApJ, 580,389Clarke, C. J., & Pringle, J. E. 1992, MNRAS, 255,423Conti, P. S., Leep, E. M., & Lorre, J. J. 1977, ApJ,214, 759Coriat, M., Corbel, S., Buxton, M. M., Bailyn, C.D., Tomsick, J. A., K¨ording, E., & Kalemci, E.2009, MNRAS, 400, 123De Becker, M., Linder, N., & Rauw, G. 2008, In-formation Bulletin on Variable Stars, 5841, 1Dray, L. M., Dale, J. E., Beer, M. E., Napiwotzki,R., & King, A. R. 2005, MNRAS, 364, 59Gies, D. R., & Bolton, C. T. 1986, ApJS, 61, 419Gillet, D., et al. 1994, A&AS, 108, 181Gosset, E., Royer, P., Rauw, G., Manfroid, J., &Vreux, J.-M. 2001, MNRAS, 327, 435G¨uver, T., & ¨Ozel, F. 2009, MNRAS, in press(arXiv:0903.2057)Heck, A., Manfroid, J., & Mersch, G. 1985, A&AS,59, 63Hoogerwerf, R., de Bruijne, J. H. J., & de Zeeuw,P. T. 2001, A&A, 365, 49Howarth, I. D., & Prinja, R. K. 1989, ApJS, 69,527Karzas, W. J., & Latter, R. 1961, ApJS, 6, 167Kellogg, E., Baldwin, J. R., & Koch, D. 1975, ApJ,199, 299Lamers, H. J. G. L. M., van den Heuvel, E. P. J.,& Petterson, J. A. 1976, A&A, 49, 3279anz, T., & Hubeny, I. 2003, ApJS, 146, 417Lanz, T., & Hubeny, I. 2007, ApJS, 169, 83Leonard, P. J. T., & Duncan, M. J. 1990, AJ, 99,608Liedahl, D. A., Osterheld, A. L., & Goldstein,W. H. 1995, ApJ, 438, L115Liu, Q. Z., van Paradijs, J., & van den Heuvel,E. P. J. 2006, A&A, 455, 1165Lucy, L. B. 2006, A&A, 457, 629Martins, F., Schaerer, D., & Hillier, D. J. 2005,A&A, 436, 1049Mason, B. D., Gies, D. R., Hartkopf, W. I., Bagn-uolo, W. G., ten Brummelaar, T., & McAlister,H. A. 1998, AJ, 115, 821McSwain, M. V., Boyajian, T., Grundstrom, E.,& Gies, D. R. 2007a, ApJ, 655, 473McSwain, M. V., Gies, D. R., Huang, W., Wiita,P. J., Wingert, D. W., & Kaper, L. 2004, ApJ,600, 927McSwain, M. V., Huang, W., & Gies, D. R. 2009,ApJ, 700, 1216McSwain, M. V., Ransom, S. M., Boyajian, T.,Grundstrom, E., & Roberts, M. S. E. 2007b,ApJ, 660, 740Mewe, R., Gronenschild, E. H. B. M., & van denOord, G. H. J. 1985, A&AS, 62, 197Mewe, R., Lemen, J. R., & van den Oord, G. H. J.1986, A&AS, 65, 511Morbey, C. L., & Brosterhus, E. B. 1974, PASP,86, 455Morrison, R., & McCammon, D. 1983, ApJ, 270,119Pavlov, G. G., Shibanov, I. A., & Zavlin, V. E.1991, MNRAS, 253, 193Perets, H. B. 2009, ApJ, 698, 1330Philp, C. J., Evans, C. R., Leonard, P. J. T., &Frail, D. A. 1996, AJ, 111, 1220Pinsonneault, M. H., & Stanek, K. Z. 2006, ApJ,639, L67 Popov, S. B., Colpi, M., Prokhorov, M. E., Treves,A., & Turolla, R. 2000, ApJ, 544, L53Portegies Zwart, S. F. 2000, ApJ, 544, 437Poveda, A., Ruiz, J., & Allen, C. 1967, Bol. Obs.Tonantzintla Tacubaya, 4, 86Prochaska, J. X., et al. 2006, ApJ, 642, 989Pszota, G., Zhang, H., Yuan, F., & Cui, W. 2008,MNRAS, 389, 423Raymond, J. C., & Smith, B. W. 1977, ApJS, 35,419Roberts, D. H., Leh´ar, J., & Dreher, J. W. 1987,AJ, 93, 968Sana, H., Rauw, G., Naz´e, Y., Gosset, E., &Vreux, J.-M. 2006, MNRAS, 372, 661Sayer, R. W., Nice, D. J., & Kaspi, V. M. 1996,ApJ, 461, 357Smith, R. K., Brickhouse, N. S., Liedahl, D. A., &Raymond, J. C. 2001, ApJ, 556, L91Stellingwerf, R. F. 1978, ApJ, 224, 953Treves, A., Turolla, R., Zane, S., & Colpi, M. 2000,PASP, 112, 297Walborn, N. R., & Fitzpatrick, E. L. 1990, PASP,102, 379Wolk, S. J., Spitzbart, B. D., Bourke, T. L., &Alves, J. 2006, AJ, 132, 1100Woosley, S. E., & Bloom, J. S. 2006, ARA&A, 44,507Zavlin, V. E., Pavlov, G. G., & Shibanov, Y. A.1996, A&A, 315, 141Zwicky, F. 1957, Morphological Astronomy(Berlin: Springer)
This 2-column preprint was prepared with the AAS L A TEXmacros v5.2. able 1Radial Velocity Measurements of HD 15137 HJD V r σ ( − − ) (km s − )53652.642 . . . OHP − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . able 1— Continued
HJD V r σ ( − − ) (km s − )54419.412 . . . OHP − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . − . able 1— Continued
HJD V r σ ( − − ) (km s − )54781.654 . . . CF − . − . − . − . − . − . − . − . − . − . − . Table 2Orbital elements of HD 15137
Element Value P (d) 55 . ± . T (HJD – 2,400,000) 54421 . ± . K (km s − ) 13 . ± . V (km s − ) − . ± . e . ± . ω ( ◦ ) 152 . ± . f ( M ) ( M (cid:12) ) 0 . ± . a sin i ( R (cid:12) ) 11 . ± . − ) 4.2813 able 3He Line Ratios and Effective Temperature T eff (K) T eff (K) T eff (K)log W λ (HeII) W λ (HeI) (log g =3.25) (log g =3.50) (log g =3.75)Using OSTAR2002 grid:He II λ I λ − .
656 27041 27814 28719He II λ I λ − .
295 28695 30153 30971He II λ I λ − .
122 30099 31202 32390Mean T eff . . . . . . . . . . . . . . . . . . . . . . . · · · χ from CF spectrum . · · · χ from OHP spectrum · · · II λ I λ − .
656 27764 28889 29954He II λ I λ − .
295 29983 31527 33532He II λ I λ − .
122 31640 33798 36645Mean T eff . . . . . . . . . . . . . . . . . . . . . . . · · · χ from CF spectrum . · · · · · · · · · Reduced χ from OHP spectrum · · · · · · · · · Table 4Journal of
XMM-Newton observations of HD 15137
Time of Mid Exposure Performed Effective Source Count Background CountCamera (JD − − ) Rate (cnt s − )MOS1 54681.56704 19.667 15.90 0 . ± . . ± . . ± . . ± . . ± .
003 0 . ± . I λ f s , alias frequency f a , and true sig-nal frequency f are also marked in the top panel. Fig. 3.— Radial velocity curve of HD 15137.The single point from Conti et al. (1977) is plot-ted as an open square, points from Boyajian etal. (2005) as open triangles, points from McSwainet al. (2007a) as open diamonds, points from theOHP in this work as filled circles, and points fromthe CF in this work as filled squares. Our longstares were performed at φ ≈ . φ ≈ .
3. Atypical error bar, assuming σ = 7 km s − in addi-tion to the intrinsic V r error of ± − due torapid line profile variations, is also shown.15ig. 4.— Mean spectrum of HD 15137 from ourCF observations ( λ < λ > V sin i = 258 km s − , T eff = 29700 K,and log g = 3 .
5, is shown for comparison. . . . . . N O R M A L I Z E D C O UN T R A T E ( c n t/ s e c / k e V ) CHANNEL ENERGY (keV)
Fig. 5.—
XMM-Newton pn spectrum of HD15137. The spectrum has been fit in XSPEC us-ing a warmly absorbed thermal model, wabs(apec) ,with nH = 6 . × atoms cm − , kT = 0 . . × − . Thereduced χ = 1 .
33 for this fit.Fig. 6.— Mass diagram of HD 15137 is plottedfor a range of inclination angles (solid lines). Theexpected M1