The eclipsing, double-lined, Of supergiant binary Cyg OB2-B17
V. E. Stroud, J. S. Clark, I. Negueruela, P. Roche, A. J. Norton, F. Vilardell
aa r X i v : . [ a s t r o - ph . S R ] F e b Astronomy&Astrophysicsmanuscript no. 12123 c (cid:13)
ESO 2018October 24, 2018
The eclipsing, double-lined, Of supergiant binary Cyg OB2-B17
V. E. Stroud , , , J. S. Clark , I. Negueruela , P. Roche , , , A. J. Norton , and F. Vilardell Faulkes Telescope Project, School of Physics and Astronomy, Cardi ff University, Cardi ff , CF24 3AA, United Kingdom e-mail: [email protected] Department of Physics and Astronomy, The Open University, Walton Hall, Milton Keynes MK7 6AA, United Kingdom Division of Earth, Space and Environment, University of Glamorgan, Pontypridd, CF37 1DL, United Kingdom Departamento de F´ısica, Ingenier´ıa de Sistemas y Teor´ıa de la Se˜nal, Universidad de Alicante, Apdo. 99, 03080 Alicante, SpainPreprint online version: October 24, 2018
ABSTRACT
Context.
Massive, eclipsing, double-lined, spectroscopic binaries are not common but are necessary to understand the evolution of massivestars as they are the only direct way to determine stellar masses. They are also the progenitors of energetic phenomena such as X-ray binariesand γ -ray bursts. Aims.
We present a photometric and spectroscopic analysis of the candidate binary system Cyg OB2-B17 to show that it is indeed a massiveevolved binary.
Methods.
We utilise V band and white-light photometry to obtain a light curve and period of the system, and spectra at di ff erent resolutions tocalculate preliminary orbital parameters and spectral classes for the components. Results.
Our results suggest that B17 is an eclipsing, double-lined, spectroscopic binary with a period of 4 . ± . Key words. binaries: eclipsing – binaries: spectroscopic – stars: early-type – stars: fundamental parameters – stars:evolution – stars: individual:Cyg OB2 B17
1. Introduction
O stars are amongst the most massive and intrinsically lumi-nous stellar objects found in galaxies. Since they are the onlydirect way to measure the masses and radii of stars, binariesare the perfect testbeds for studying the physical properties andevolution of such stars. Unfortunately, as reported by Bonanos(2009), less than 20 O stars have accurate ( ≤ ff erent evolutionary states(e.g., Gies 2002) - the mass luminosity relation and theoreticalevolutionary tracks of massive stars ( M ≥ M ⊙ ) are currentlypoorly constrained by observations.In order to address this shortfall, much e ff ort has been ex-pended to identify further examples, utilising photometric andspectroscopic observations of young massive clusters such asthe Arches (Martins et al. 2008), Quintuplet (Figer et al. 1999),and Westerlund 1 (Clark et al. 2005; Ritchie et al. 2009). Theseobservations indicate that the binary fraction is potentially very high ; Kobulnicky & Fryer (2007) inferred it to be ≥ ≥ M ⊙ + M ⊙ ; Rauw et al. 2004, Bonanos et al.2004) and the newly discovered WN6ha binary NGC3603-A1(116 ± M ⊙ + ± M ⊙ ; Schnurr et al. 2008). This in turn hasimplications for the determination of the Initial Mass Function(IMF) for the clusters in question and, by extension, the empir-ically determined maximum mass possible for a star (cf. Figer2005). Moreover, massive close binaries are the progenitors ofsuch diverse energetic phenomena as supernovae, γ -ray burstsand X-ray binaries (Ribas 2006). Clearly the properties of theprogenitor binary population must be known to constrain theirformation channels.Cyg OB2 is one of the most massive and richest asso-ciations in the Galaxy. It is ∼ A large amount of observations and patience are necessary to de-termine the true binarity of a sample of stars in an open cluster asobserved in Sana et al. (2008) V. E. Stroud et al.: The eclipsing, double-lined, Of supergiant binary Cyg OB2-B17 (Negueruela et al. 2008), its proximity and accessibility to op-tical studies have made it the focus of numerous observationalcampaigns to determine the properties of its massive stellarpopulation (e.g. Massey & Thomson 1991; Kn¨odlseder 2000;Hanson 2003; Comer´on et al. 2002; Kiminki et al. 2007, 2008).Cyg OB2 B17 (Comer´on et al. 2002; henceforth B17 andalso known as V1827Cyg, 2MASS J20302730 + α = h m s , δ =+ o ′ ′′ ; V = Br γ emission, confirming it to be anevolved massive star. Follow up observations were made byNegueruela et al. (2008), who found it to stand out from therest of the members due to its variability and strong emissionlines. They classified it as an Ofpe star and suggested it was astrong binary candidate.This paper reports the first results of an extensive multi-epoch photometric and spectroscopic observational cam-paign on the binary candidate B17. Section 2 gives a de-scription of the photometric and spectroscopic observations.Photometrically, the system was found to be variable and wereport the analysis of the light curve in Section 3. More spectro-scopic data were obtained permitting preliminary spectral andluminosity classification of the system; this analysis is shownin Section 4, along with descriptions of the long and shorttimescale variations of the spectra. The light and radial velocitycurve modelling is presented in Section 5. A discussion of thesystem, including its evolutionary status, is found in Section 6and a summary is presented in Section 7.Note that the central goals of this manuscript are to verifythe binary hypothesis and present a preliminary spectral clas-sification. A full analysis of the system, consisting of the de-convolution of an expanded spectral data set and subsequentmodel atmosphere analysis to determine the fundamental stel-lar parameters of the system will be presented in a future paper(Stroud et al. in prep).
2. Observations and Data Reduction
The North Sky Variability Survey (NSVS; W´ozniak et al.2004) is a record of the sky at declinations higher than δ = − ◦ over the optical magnitude range 8 to 15.5. It containslight curves of over 14 million objects. The data were takenbetween 1999 April and 2000 March by the first-generationRobotic Optical Transient Search Experiment (ROTSE-I) atLos Alamos National Observatory, New Mexico. The telescopeconsisted of four unfiltered Canon 200mm telephoto lenseswith f / . d × . d
2. These wereequipped with AP10-cameras and Thomson TH7899M CCDs.The lenses had a typical point-spread function with a full widthhalf maximum of ∼ ′′ . In a median field, the bright unsat-urated stars had a point-to-point photometric scatter of ∼ ′′ . The calibrated images werepassed through SExtractor software (Bertin & Arnouts 1996),reducing them to object lists. The data were accessed through the Sky Database for Objects in Time-Domain (SkyDOT) atLos Alamos National Laboratory. A total of 186 observationswere obtained for B17.Additional V band photometry was obtained by amateur as-tronomers Pedro Pastor Seva (observer 1) and Manuel M´endezMarmolejo (observer 2) between 2007 April 03 and 2007 June11. Observatory 1 is located in Muchamiel (Alicante, Spain).The telescope used was an 8 inch Vixen VISAC Schmidt-Cassegrain telescope. It has a field of view of 24 ′ x20 ′ anda focal ratio of f / ′ x12 ′ (as a smaller SBIG ST7-XME chip was used) and a focal ra-tio of f / ∼ and AIP4WINv2 packages. The apparent magni-tudes were obtained using di ff erential photometry with respectto a set of reference stars in the image with known magnitudes.In both cases the precission for the individual meassures is al-ways better than 0.01 mag. The spectra were obtained from several telescopes during thecourse of ∼ three years. Table 1 lists the full set of observa-tions. The first set was observed with the 1.52-m G. D. Cassinitelescope at the Loiano Observatory (Italy) during the night of2004 July 18. The telescope was equipped with the BolognaFaint Object Spectrograph and Camera (BFOSC) and an EEVcamera. Grism 3 was used, which covers 3300-5800Å witha resolution of ∼ / pixel (the resolution element is approximately 3pixels in the blue and 2 pixels in the red).The system was also observed with the 2.5m Isaac NewtonTelescope (INT) in La Palma on 2006 September 10-11.The telescope was equipped with the Intermediate DispersionSpectrograph (IDS), fitted with a R632V grating and EEV10CCD. Another 12 spectra were obtained during a dedicatedrun on 2007 August 21-22 at the WHT. The system was ob-served in the blue arm with grating R1200B (nominal dis-persion of ∼ / pixel). All the spectra were reduced withthe Starlink packages ccdpack (Draper et al. 2000) and figaro (Shortridge et al. 1997) and analysed using figaro and dipso (Howarth et al. 1998). Mira Pro software is published by Mirametrics Inc., which has noconnection with the Monterey Institute for Research Astronomy AIP4WINv2 is published by Willmann-Bell, Inc.. E. Stroud et al.: The eclipsing, double-lined, Of supergiant binary Cyg OB2-B17 3
Table 1.
Log of Spectroscopic Observations.
Date Telescope Nominal Wavelength Phase / Instrument Dispersion Range (Å)(Å / pix)04 / /
18 Cassini / BFOSC 6 3800-6400 0.43704 / /
18 Cassini / BFOSC 6 6100-8200 0.45006 / /
11 WHT / ISISB 0.86 3200-5200 0.03606 / /
11 WHT / ISISR 0.93 5400-8000 0.03606 / /
18 WHT / ISISB 0.86 3200-5200 0.92906 / /
18 WHT / ISISR 0.93 5400-8100 0.93106 / /
10 INT / IDS 0.9 3900-5000 0.60006 / /
11 INT / IDS 0.9 3900-5000 0.84006 / /
11 INT / IDS 0.9 3900-5000 0.89107 / /
21 WHT / ISISB 0.86 3600-5300 0.73807 / /
21 WHT / ISISR 0.93 5400-8300 0.73807 / /
21 WHT / ISISR 0.93 5400-8300 0.74007 / /
21 WHT / ISISR 0.93 5400-8300 0.74107 / /
21 WHT / ISISB 0.23 3900-4750 0.38007 / /
21 WHT / ISISB 0.23 3900-4750 0.38507 / /
21 WHT / ISISB 0.23 3900-4750 0.39707 / /
21 WHT / ISISB 0.23 3900-4750 0.41607 / /
21 WHT / ISISB 0.23 3900-4750 0.43607 / /
21 WHT / ISISB 0.23 3900-4750 0.45307 / /
22 WHT / ISISB 0.23 3900-4750 0.62907 / /
22 WHT / ISISB 0.23 3900-4750 0.63407 / /
22 WHT / ISISB 0.23 3900-4750 0.64807 / /
22 WHT / ISISB 0.23 3900-4750 0.66407 / /
22 WHT / ISISB 0.23 3900-4750 0.68307 / /
22 WHT / ISISB 0.23 3900-4750 0.701
3. Photometry
The
Starlink software period (Dhillon et al. 2001) was used onboth the NSVS and amateur photometry to search for a mod-ulation period in the photometric data, using phase dispersionminimisation, χ of sine fit vs frequency and string-length vsfrequency methods. The resultant periods were consistent andfavoured an orbital period of 4 . ± . ff erent methods employed). When both data sets were foldedtogether they were found to be in phase; therefore, given thatthere appears to be no change of shape or shift in period in the7 years between the NSVS and amateur observations, we areconfident that the period determined is accurate to within theerrors quoted.Fig. 1 shows the periodogram obtained using Period withthe reduced- χ technique (top) and the V band light curvefolded on a 4.0217 day period, along with the phases of thespectra obtained (bottom).The shape of the light curve suggests that the system is asemi-detached binary; Both minima are narrow and demon-strate di ff erent eclipse depths, although the 0 . − . ff erent luminosi-ties (in a contact system, the temperature of both stars shouldbe the same) There appears to be an asymmetry on the lightcurve between phases φ = . − .
75 which is observed in bothsets of data. Hilditch et al. (2005) found similar depressions ineclipsing binaries in the Small Magellanic Cloud and Bonanos
Fig. 1.
Upper panel: Periodogram for B17 data using thereduced- χ technique. Lower panel: Light curve folded on a4.0217 day period. The red triangles show the phases for whichspectra were obtained.(2009) found a similar depression in the light curve for the bi-nary LMC-SC1-105. They attributed these asymmetries to thepresence of a mass-transfer stream. Linder et al. (2009) alsoobserved this asymmetry in the overcontact binary Cyg OB2 ff ect - where themaxima are of di ff erent brightnesses - of ∼ JD = , , . ± . minI = . + . E (JD) (1)where E is the number of orbital cycles after the given epoch.We note that our period determination is in agreement witha value recently established by Otero (2008) using the samedata from NSVS as used in this paper. V. E. Stroud et al.: The eclipsing, double-lined, Of supergiant binary Cyg OB2-B17
4. Spectroscopy
The spectrum of B17 is found to be highly variable on bothshort ( < day) and long ( ∼ year) timescales, with the most promi-nent features in the spectra being H, He i , He ii and N iii linesin absorption and He ii and N iii lines in emission. Wavelengthshifts are observed for di ff erent lines, some of which becomedouble in some spectra suggesting that it is a binary systemwith at least one hot component, as revealed by the presence ofHe ii and N iii . Fig. 2 shows blue-violet spectra (4000-4900Å) obtained at ran-dom phases between 2004 July and 2007 August. The mostprominent elements observed are H, He i , He ii and N iii in ab-sorption and He ii λ iii λλ iii and He ii emis-sion lines; in 2004 July 18 these are of a similar strength but by2006 June 11 the He ii line is almost three times the intensityof the N iii lines. The N iii profile is also variable; it appears astwo separate peaks in some of the spectra (2006 June 11, 2006September 10-11, 2007 July 21) and it is almost completelyblended in the spectra taken on 2004 July 18 and 2006 August18, although this in part could be due to the low resolution ofthe spectrum from 2004 July 18.The hydrogen lines are also highly variable; this is partic-ularly evident for H γ when compared to the di ff use interstellarband (DIB) at λ β shows a P-Cygni profile varying inboth width and strength suggesting the stellar wind is highlyvariable. Finally, the He i λ / He ii λ i λ i interstellar lines (Fig. 3). When comparedto the nearby DIBs at 5780-5800Å, the absorption trough inHe i line was observed to vary by a factor of 4 in intensity rel-ative to the nearby DIB features, with the emission componentalso varying in strength by a factor of 2. Other lines in the yel-low spectra include C iii λ iv around λ α line which also demonstrates a highly vari-able P-Cygni profile which is blended with He ii absorption fea-tures in all the red spectra, complicating analysis of the profile.It has a double peak profile which varies in intensities withthe redder peak being between 2 and 4 times stronger thanthe bluer peak. The depth of the He ii line, which presumablycontributes to the double peaked morphology is also variable(Fig. 4). Preliminary calculations for the velocities of the blueedge of the H α profile ( v edge ) have a range of 2260 - 2625kms − with an average 2440 ±
50 kms − . The v edge is relatedto the terminal velocity of the wind; in the extreme ultravio-let, the terminal velocities for OB stars are 15%-20% smallerthan the edge velocities (Prinja et al. 1990). Variations in thewind profiles of OB stars are thought to be a consequence ofhighly structured winds on multiple scales, and while it appears Fig. 2.
Blue-violet spectra of B17 obtained at di ff erent phasesover a three year period, with the most prominent lines labeled.likely that the variations observed for B17 arise due to a highlyanisotropic circumstellar envelope observed at di ff ering lines ofsight throughout the orbital period, the current limited data seto ff ers little prospect of a more explicit physical interpretation.Unfortunately, no spectra with the same orbital phases havebeen obtained at di ff erent epochs and so it has not been possibleto search for unambiguous long term secular variability in thecurrent data set.In Fig. 5 we plot the EW of the He ii λ ∼ constant over the segments of the orbitalperiod sampled by the observations, implying that the changein EW is primarily due to dilution by the variable continuum. The set of 12 high resolution spectra obtained during the nightsof the 2007 August 21-22 cover the phases φ = φ = ii ab-sorption lines lines compared to the spectra from the secondnight. During the first night, the He i λ / He ii λ i λ . E. Stroud et al.: The eclipsing, double-lined, Of supergiant binary Cyg OB2-B17 5 Fig. 3.
B17 yellow spectra showing variability at di ff erentphases taken over a period of three years. Prominent lines arelabeled.A second, blueshifted absorption component appears on theshoulder of the He i λ δ profile. Initially,the N iii λ φ = ii λ i λ ii λ γ linesdo not vary significantly throughout the first night.During the 17hrs between the last observation of night one( φ = φ = i λ ii λ γ lines arebroader and non symmetrical. On the second night (Fig. 6 lowerpanel), the He i λ / He ii λ > φ = i λ δ line appears to be the result of the blending of twolines, showing a wide profile with two troughs. The bluertrough is stronger than the redder during the second night. TheN iii λ ii λ Fig. 4.
B17 red spectra showing H α variability at di ff erentphases taken over a period of three years. Prominent lines arelabeled. Fig. 5.
Upper panel: Light curve folded on a 4.0217 day period.Lower panel: Equivalent widths of the He ii λ V. E. Stroud et al.: The eclipsing, double-lined, Of supergiant binary Cyg OB2-B17 ens throughout the night. We note that (as mentioned in Section3) the light curve demonstrates an asymmetry - possibly due tomass transfer - between phases φ ∼ Given that the current spectroscopic dataset does not span anentire orbital cycle we have not attempted a formal deconvo-lution of the blended spectra. Instead we have simply used thespectra closest to the primary and secondary eclipses to per-form a preliminary classification of the components of the sys-tem, using the stellar atlas of Walborn & Fitzpatrick (1990) andWalborn & Howarth (2000).The spectrum at φ = ii λ / He i λ ii λ iii λλ iv λ Walborn (2001) mentions that a similar e ff ect is found inspectra in the Small Magellanic Cloud where no Si iv λ + for this component.The spectrum at φ = ii λ / He i λ ii λ iii λλ iv φ = / HeII4541 ratio less than unity (see Section 4.2), itsuggests there is emission in-filling of HeI4471.
The high resolution blue spectra show systematic night to nightvariations in the H, He and N lines, revealing significant ra-dial velocity shifts. Radial velocities were determined for theprimary lines which appear not to be blended (He ii λ ii λ γ λ dipso emission line fittingcommand ( elf ; the results are discussed in Sec. 5). The linesfor the secondary appear to be blended and were not measured.The radial velocities for the secondary will be measured in thefuture paper after the spectra have been dissentangled. We note that while this line is not present in the spectrum of theO7 supergiant Sanduleak 80, given the reduced metalicity appropriatefor a SMC star it is not clear it provides a valid comparison.
Fig. 7.
Upper panel: B17 spectrum obtained with the WHT onthe night of 2007 August 21, close to the secondary eclipse,compared with Of supergiant spectra from the Digital Atlas ofStellar Classification (Walborn & Fitzpatrick 1990). The spec-trum is most compatible with an O7Iaf classification. Doublelines are observed even though the system is at eclipse whichwe currently cannot account for and will be further studied withthe spectra dissentangling. Lower panel: B17 spectrum ob-tained with the WHT on the night of 2006 June 11, close to theprimary eclipse, compared with Of supergiant spectra from theDigital Atlas of Stellar Classification (Walborn & Fitzpatrick1990). The spectrum is more comparable to the O9 classifica-tion.
5. Light and radial velocity curve analysis
In order to reproduce the observed characteristics of the photo-metric light curve, we analysed it with the 2003 version of theWilson & Devinney (W-D; 1971) code. For the analysis, theROTSE (with an e ff ective wavelength similar to Johnson R )and amateur V light curves were modeled as di ff erent datasets.In all cases, detailed reflection-model and proximity-e ff ect cor-rections were included. Considering the spectroscopic analy-sis, the temperature of the primary ( T Pe ff ) was fixed to 35 000 K,and the bolometric albedo and gravity brightening coe ffi cientswere set to unity, as generally found for stars with radiativeenvelopes. In addition, a circular orbit was adopted, as sug-gested by the equal separation between both (primary and sec- . E. Stroud et al.: The eclipsing, double-lined, Of supergiant binary Cyg OB2-B17 7 ondary) eclipses, and a rotation rate synchronized with the or-bital period was assumed for both components. The fitting pro-cess was carried out iteratively until three consecutive solutionsprovided di ff erential corrections for all the parameters smallerthan twice their internal errors.Considering the characteristics in the light curves describedabove, the first runs in the modelling with W-D assumed a semi-detached configuration. Numerous attempts were performedwith several mass ratios and with either the primary or thesecondary component filling the Roche lobe. However, in allcases, the fits provided solutions where both stars tended to bein contact . Therefore, despite the properties of the lightcurvesuggesting a semi-contact configuration we finally attemptedan over-contact solution - the results of which are describedhere - where each component could have a di ff erent tempera-ture, as observed from the di ff erent depths of the eclipses. Forthis model, the mass ratio was set to q = .
75 - as might be ex-pected for O7 and O9 super-giant components - noting that themass ratio cannot be smaller than 0.4 . The time of minimum( t min ), the period ( P ), the orbital inclination ( i ), the temperatureof the secondary component ( T Se ff ), the surface potential ( Ω P )and the luminosity of the primary component ( L P ) were all leftas free parameters in the fit.The V light curve (which has smaller photometric errors)clearly reveals that one quadrature is ∼ ff ect, one with an equatorial hot spot (30%hotter than the photosphere) on the primary component and an-other one with the spot being on the secondary component. Thesize and position of the spot were left as free parameters andalso fitted at each run, although we emphasise that the size andtemperature of the spot are strongly correlated. The fits with thespot on the primary component were finally adopted, since theyprovided slightly smaller errors. In addition, the position of thespot (oriented roughly towards the secondary component) canmore easily be explained as the interaction of the stellar winds.Together with the light curve analysis, the radial velocitieswere also used to constrain the solution. Radial velocities werefitted separately from the light curve to avoid the larger num-ber of photometric observations dominate in the solution; theparameters fitted being the semi-major axis ( a ) and the sys-temic velocity ( γ ). The rms of the fit is 9.5 km s − for the radialvelocities, 0.023 mag for the V light curve and 0.048 mag forthe ROTSE light curve. Unfortunately, the accuracy of the solu-tion obtained is strongly dependent on the mass ratio adopted.Nevertheless, Fig. 8 and Fig. 9 show the light curve and radialvelocity curve with the best fit and Table 2 gives the parametersobtained from this solution.Despite our e ff orts we are still not completely satisfied withthis fit. In particular, we are still unable to adequately fit theegress from secondary minimum nor the subsequent lightcurvebetween phases 0.6-0.9. Moreover it is expected that the tem-peratures of stars in over-contact binaries should be ∼ equalwhere a ratio of 0.85 was found between the secondary and For a mass ratio of 0.4, the implied mass of the primary -considering the minimum amplitude of the radial velocity curve- isover 120 M ⊙ , regardless of the light curve analysis. Table 2.
Results from the analysis of the light and radial veloc-ity curves: The errors shown should be considered internal er-rors of the fit. Any possible systematic errors are not included.
Parameter ValueT (MJD) 4272.534 ± P ± i ± e q ± R ⊙ ) 50 ± ± − Parameter Primary SecondaryMass ( M ⊙ ) 60 ± ± R ⊙ ) 22 ± ± ± ± − ) 257 ± ± ff (K) 35000 (Fixed) 29900 ± Ω ) 3.14 ± Bol -9.8 ± ± primary. As such we regard the parameters presented in Table2 as provisional at present. We anticipate that the determina-tion of radial velocities for the secondary component, whichwill directly constrain the binary mass ratio will greatly clarifythe fundamental properties of the components and the configu-ration of this EB system.Finally, the lightcurve modelling, allows us to address thedistance to B17, and by extention the Cyg OB2 association.Adopting the bolometric corrections from Martins et al. (2005)we may use the bolometric luminosities determined above (SeeTable 2) to calculate the absolute V magnitudes of both com-ponents. The V band reddening was then calculated by follow-ing a similar procedure used by Negueruela et al. 2008 (seeSection 6) With the absolute V magnitudes, and the V bandreddening for B17, along with the V band values for the twominima in the lightcurve, the distance modulus was calculatedto be 10.9-11.3 which corresponds to a distance of about 1.5-1.8 kpc. While we regard these values as provisional due to thedi ffi culties in modelling described above, these are consistentwith the commonly adopted distance estimate of 1.7 kpc. (e.g.Hanson 2003, Torres-Dodgen et al. 1991 and Kiminki et al.2007). Therefore they do not agree with the distance estimate of900-950 pc obtained by Linder et al. (2009) from the lightcurvemodelling of Cyg OB2 ∼ half thosecurrently derived from the modelling. The derived radii scalelinearly with the semi-major axis and the semi-major axis inturn is strongly dependant on the assumed mass ratio. For thissystem, a semi-major axis a factor of 2 smaller would imply amass ratio of 6. We consider this to be extremely unlikely, butnot impossible. V. E. Stroud et al.: The eclipsing, double-lined, Of supergiant binary Cyg OB2-B17
Fig. 8.
The V band and R band light curves of B17 with the bestfit model overlaid. The panels below the light curves show thedeviations from the model. Fig. 9.
The RV curve for the primary component obtained fromthe WHT data, with the best fit overlaid (solid line). The panelbelow shows the residuals between the observed data and themodel.
6. Discussion
Given the di ffi culties in determining a unique model fit to thedata (Sect. 5), we regard the parameters presented in Table 2to be provisional. As a check on consistency we have cal-culated a preliminary bolometric magnitude for the compo-nents, following a similar procedure the that employed byNegueruela et al. (2008). Assuming that both components areof similar colour , we adopted the 2MASS value for ( J − K S )used in Negueruela et al. (2008), the e ff ective temperature T e ff and bolometric correction (BC) calibrations of Martins et al. The colour di ff erences between an O7 and an O9 supergiant are( J − V ) = .
09 and ( K − V ) = .
11 (Wegner 1994) (2005), and the intrinsic ( J − V ) and ( K − V ) colour calibra-tions of Wegner (1994). Using the 2MASS observed ( J − K S ),we derived E ( J − K S ). The reddening of the system was cal-culated using the relation A K S = . E ( J − K S ), due to thereddening in the association being close to standard (Hanson2003). The reddening in the V band was calculated using therelation 0 . A V ≃ A K (Rieke & Lebofsky 1985). The absolute V band magnitude for the primary was then calculated, usingthe V band value of the secondary minimum and adopting thedistance modulus of 11.3, obtained by averaging spectroscopicdistances (Kiminki et al. 2007). A semi-observational M bol wasthen calculated to be − . ± . − . ± . V − K ) and the BC to the V , as-suming an uncertainty of one spectral type. With this value, theluminosities were calculated to be log( L / L ⊙ ) = . ± . L / L ⊙ ) = . ± .
1, noting that the main source of erroris likely to be the uncertainty in the spectral type and hencetemperature and BC. We regard these estimates as upper limitssince it assumes that both components are completely eclipsedduring the minima.The positions of the two components of B17 in the HR di-agram (Fig. 10) are consistent with other known members ofCyg OB2 suggesting that it too is a bona fide member, with anage of ∼ / WNL + O6.5-7 binary Cyg OB2 / WNL star being slightly more evolved than the O9 Iafsecondary in B17. The slightly longer period of Cyg OB2 ⊙ , suggesting the following evolutionarypathways (also see Crowther et al. 1995): • ∼ ⊙ : O → Ofpe / WN9 ⇋ LBV → WN8 → WN / C • ∼ ⊙ : O → Of → WNL + abs → WN7With an age of ∼ + WN6ha bi-nary WR20a (m = m = M ⊙ Rauw et al. 2004; Bonanos et al.2004) we suspect that it will instead evolve to resemble CygOB2 / LBV candi-dates ( P orb = .
45 days; Martins et al. 2006), with stellar massloss resulting in an eventual lengthening of the orbital periodof B17. Indeed, assuming the system avoids merger during theLBV phase and, remaining bound, receives a favourable SNekick to reduce the orbital separation it might briefly form a high . E. Stroud et al.: The eclipsing, double-lined, Of supergiant binary Cyg OB2-B17 9 mass X-ray binary with a WR mass donor prior to the secondSN.Irrespective of its ultimate fate, B17 is amongst the bright-est / most massive systems in Cyg OB2 and adds to theincreasing number of massive binaries identified within it(Kiminki et al. 2009, see Table 3). Similar trends for botha high binary fraction and the multiplicity of the bright-est / most evolved cluster members have been observed forPismis 24 (Ma´ız Apell´aniz 2008), NGC3603 (Schnurr et al.2008), Westerlund 1 (Clark et al. 2008; Ritchie et al. 2009) andpotentially the Arches (Clark et al. 2009). If these trends con-tinue it will have significant implications for the formationchannels and relative production rates of both low and highmass X-ray binaries and systems comprising of two relativisticobjects (Kobulnicky & Fryer 2007). Fig. 10.
Updated semi-observational HR diagram fromNegueruela et al. (2008), based on published spectral classesand 2MASS
JHK S photometry and a DM of 11.3. The continu-ous lines are non-rotating isochrones for log t = .
2, 6 . . t = .
7. Summary
Using photometric and spectroscopic data, we have demon-strated that B17 is an eclipsing, double lined spectroscopic bi-nary comprising two supergiants with preliminary classifica-tions of O7Iaf and O9Iaf. The spectra are highly variable, andwith a subset revealing features from both stars, raise the pos-sibility of achieving a dynamical mass determination for bothcomponents. Utilising both the photometric lighturve and our limited RV dataset we attempted to determine an initial orbitalsolution for the binary.Despite the morphology of the lightcurve indicating a semi-contact configuration we were unable to to achieve conver-gence for such a hypothesis and hence were forced to adoptedan over-contact configuration. In the absence of a full RV curvefor both system components we were forced to fix the binarymass ratio, and had to include the presence of a star spot toaddress the observed asymmetries in the lightcurve (which arelikely due to the e ff ects of binary mass transfer). However, wewere still unable to fully fit both secondary eclipse and thelightcurve between orbital phase ∼ / LBV phase into a longperiod WR + WR binary configuration as mass loss via stellarwinds increases the orbital separation. In combination with therecent work of Kiminki et al. (2009) and Kobulnicky & Fryer(2007) the results of our analysis provides additional evidencethat Cyg OB2 has a very high fraction of massive binary stars.Such an observational constraint needs to be considered whendetermining the initial mass function of the association as itmay both influence the slope of the relationship and also leadto a population of artificially massive stars, resulting in the in-flation of a putative high mass cut-o ff to the IMF. Acknowledgements.
We thank Pedro Pastor and Manuel M´endez forhaving obtained and reduced the photometric data. Amparo Marcofor help with the 2004 observing run and Miriam Garc´ıa for assis-tance with the 2006 INT run. We thank Dan Kiminki, Fraser Lewisand Chris Evans for useful discussions and reading of the manuscript.We also thank the referee for his guidance in completing the pa-per for publication. The Faulkes Telescope Project is an educationaland research arm of the Las Cumbres Observatory Global TelescopeNetwork (LCOGT). VS acknowledges support from the Dill FaulkesEducational Trust. This research is partially supported by the SpanishMinisterio de Ciencia e Innovaci´on undergrants AYA2008-06166-C03-03 and Consolider-GTC CSD2006-70. The G.D. Cassini tele-scope is operated at the Loiano Observatory by the OsservatorioAstronomico di Bologna. The WHT is operated on the island of LaPalma by the Isaac Newton Group in the Spanish Observatorio delRoque de Los Muchachos of the Instituto de Astrof´ısica de Canarias.The 2006 observations were taken as part of the service programme(programme SW2005A20).
References
Bertin, E., & Arnouts, S. 1996, A&A, 117, 393
Table 3.
Evolved massive binaries in Cyg OB2 from Kiminki et al (2009).
Star Sp Types Period (days) ReferencesMT05 O9 III & mid B early B & early B < / WN9 6.6 Wilson (1948), Wilson & Abt (1951), Miczaika (1953)Walborn (1973), Contreras et al. (1997), Rauw et al. (1999)Schulte 8a (
O8?