The Astrophysical Journal, 611:L33–L36, 2004 August 10 ൴ ᭧ 2004. The American Astronomical Society. All rights reserved. Printed in U.S.A.

WR 20a IS AN ECLIPSING BINARY: ACCURATE DETERMINATION OF PARAMETERS FOR AN EXTREMELY MASSIVE WOLF-RAYET SYSTEM1 A. Z. Bonanos and K. Z. Stanek Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138; [email protected], [email protected] and A. Udalski, L. Wyrzykowski,2 K. Z˙ ebrun´ , M. Kubiak, M. K. Szyman´ ski, O. Szewczyk, G. Pietrzyn´ ski,3 and I. Soszyn´ ski Warsaw University Observatory, Al. Ujazdowskie 4, PL-00-478 Warsaw, Poland; [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected] Received 2004 May 18; accepted 2004 June 24; published 2004 July 8

ABSTRACT We present a high-precision I-band light curve for the Wolf-Rayet binary WR 20a, obtained as a subproject of the Optical Gravitational Lensing Experiment. Rauw et al. have recently presented spectroscopy for this system, M, for the component 3.8 ע and 68.8 4.0 ע strongly suggesting extremely large minimum of70.7 of the system, with the exact values depending strongly on the period of the system. We detect deep of about 0.4 mag in the light curve of WR 20a, confirming and refining the suspected period ofP p 3.686 days and 2Њ.0 . Using these photometric data and the data of Rauw ע deriving an inclination angle ofi p 74Њ.5 ,M, . Therefore 5.0 ע and 82.0 5.0 ע et al., we derive the masses for the two components of WR 20a to be83.0 WR 20a is confirmed to consist of two extremely massive stars and to be the most massive binary known with an accurate determination. Subject headings: binaries: eclipsing — binaries: spectroscopic — stars: fundamental parameters — stars: individual (WR 20a) — stars: Wolf-Rayet Online material: machine-readable table

1. INTRODUCTION 7.3 M, (Rauw et al. 1996; Schweickhardt et al. 1999), and Plaskett’s with a minimum primary mass of 51 M, (Bag- Measuring accurate masses for the most massive stars in our nuolo et al. 1992). However, WR 20a, a Wolf-Rayet (W-R) and beyond is important for constraining star formation binary in the compact cluster , seems to be the and theories, which have indirect implications new record holder. Rauw et al. (2004) obtained spectroscopy for many objects that are not well understood, such as super- for WR 20a and measured extremely large minimum masses ע ע novae, gamma-ray bursts, and Population III stars. The most of70.7 4.0 and 68.8 3.8 M, for the components. The massive candidates in the are LBV 1806Ϫ20 (Ei- final masses strongly depend on both the period and the in- kenberry et al. 2004) and the Pistol Star (Figer et al. 1998), clination of the binary, which can only be measured from the light curve. They derived a period of 3.675 days from the which have inferred masses up to ∼200 M,. HDE 269810 in absolute values of the radial velocity differences and assumed the LMC is another massive candidate ∼150 M, (Walborn et al. 2004). Finally, h Carinae, one of the most studied massive a circular to derive the masses of the components. In their discussion of WR 20a, Rauw et al. (2004) stressed the impor- stars, has an estimated mass of ∼120 M,, yet it remains un- known whether it is a rapid rotator or a close binary system tance and necessity of photometric monitoring of this binary. (see review by Davidson & Humphreys 1997). However, all Shara et al. (1991) were the first to make the discovery that these masses are only indirect estimates and thus have large WR 20a was a W-R star, by obtaining spectroscopy of the uncertainties associated with them. The only direct way of system. Moffat et al. (1991) obtained UBV photometry; how- measuring accurate masses of distant stars is through double- ever, as Rauw et al. (2004) suggest, it is likely affected by lined spectroscopic binary systems with eclipses present in their photometric eclipses. WR 20a was classified as a candidate light curves. The limited number of existing mass measure- binary, along with 30 other W-R stars in the seventh catalogue ments for massive stars (120 M,) can be explained by their of galactic Wolf-Rayet stars (van der Hucht 2001), because of intrinsic rarity and the observationally demanding process of its comparatively weak emission lines, which were thought to discovering them in eclipsing binary systems. be diluted by the continuum light of an OB-type companion. Until recently, the most massive stars ever weighed in bi- Rauw et al. (2004) spectroscopically surveyed these candidate naries were R136-38 (O3 VϩO6 V) in the LMC with a primary W-R binaries, and WR 20a clearly emerged as an interesting M, (Massey et al. 2002), WR 22 object, which they followed up. Due to its favorable location 0.6 ע mass of 56.9 in the sky, this object was included in the target list covered ע WN7ϩabsϩO) with a minimum primary mass of 55.3) by the Optical Gravitational Lensing Experiment (OGLE) 1 Based on observations obtained with the 1.3 m Warsaw telescope at Las photometric survey. We have quickly obtained a precise light Campanas Observatory, which is operated by the Carnegie Institute of curve of WR 20a and thus confirm and refine, by measuring Washington. the period and inclination, the remarkable masses of the WR 2 School of Physics and Astronomy and Wise Observatory, Tel Aviv Uni- versity, Tel Aviv 69978, Israel. 20a components. 3 Universidad de Concepcio´n, Departamento de Fı´sica, Casilla 160-C, Con- In this Letter, we present photometry for WR 20a, which cepcio´n, Chile. clearly shows the presence of eclipses. In§2wedescribe the L33 L34 BONANOS ET AL. Vol. 611

TABLE 1 TABLE 2 I-Band Photometry of WR 20a Light-Curve Parameters for WR 20a

UT HJD I jI Parameter Value days 0.01 ע May 1 ...... 2,453,126.57440 10.932 0.01 Period, P ...... 3.686 2004 2,453,126.57581 10.931 0.01 Primary , Tprim ...... 2453124.569 2Њ.0 ע Inclination, i ...... 74Њ.5 0.01 10.920 2,453,126.58207 2,453,126.58350 10.925 0.01 Eccentricity, e ...... 0 (fixed) 2004 May 2 ...... 2,453,127.57488 10.657 0.01 , Teff1 ...... 42,000 K (fixed) ע 2,453,127.57883 10.658 0.01 Effective temperature, Teff2 ...... 40,300 1000 K ע p Surface potential (Q12Q ) ...... 3.92 0.03 ע -Note.—Table 1 is published in its entirety in the elec Radius, rpole ...... 18.7 0.3 R, ע tronic edition of the Astrophysical Journal. A portion is Radius, rpoint ...... 22.0 0.3 R, ע .shown here for guidance regarding its form and content Radius, rside ...... 19.3 0.3 R, ע Radius, rback ...... 20.4 0.3 R, observations, in§3wepresent the light curve and the analysis, and in § 4 we discuss the results. thus confirming and refining the of WR 20a and 2. OGLE OBSERVATIONS the remarkable masses of its components. WR 20a (pSMSP 2 p THA 35-II-036) is located in the In addition, we can derive an inclination angle i for the ata p 10hms 23 58.0 , d p Ϫ57Њ45 49 system from our well-sampled light curve. To first order, the (J2000.0). Observations were carried out with the 1.3 m Warsaw value of the inclination angle is given by the depth of the ∼ telescope at Las Campanas Observatory, Chile, which is operated eclipses ( 0.4 mag), and we find that the exact value of i only by the Carnegie Institute of Washington. The telescope is weakly depends on the details of the model fit or even on which equipped with a mosaic CCD camera with8192 # 8192 pixels. eclipsing binary model we use to fit. To demonstrate this, we WR 20a is very bright,I ∼ 11.0 , so very short exposures of first fitted the light curve with a simple model of two spherical 10 s each were used to avoid saturating the CCD. The obser- stars with limb darkening (J. Devor 2004, private communi- vations presented here started on 2004 May 1 and consist of 83 cation), with the more complex Eclipsing Binary Orbit Program measurements obtained on 17 consecutive nights. (EBOP) model (Nelson & Davis 1972; Popper & Etzel 1981), The absolute calibration of our photometry was obtained on and finally with the Wilson-Devinney (WD) code (Wilson & one photometric night with a couple of standards from Landolt Devinney 1971; Wilson 1979; van Hamme & Wilson 2003) (1992). The uncertainty in the zero points of the photometry for modeling distorted stars. We find that indeed the inclination is about 0.03 mag. angle is insensitive to the model, and the best-fit value with Њ ע p Њ The data were reduced using the standard OGLE-III data the WD code isi 74.5 2.0 . pipeline, as described by Udalski (2003). The photometry was We ran the WD code in the overcontact mode (mode 3), fixing p derived using difference imaging analysis. This method is the Teff1 42,000 K (since WR 20a is intermediate between Mrk current state of the art for photometric accuracy in crowded 42 and WR 47c; see Rauw et al. 2004 and Crowther & Dessart fields (Alard 2000). Our analysis resulted in a set of time-series 1998), using linear limb-darkening coefficients from van Hamme photometry in the I band, presented in Table 1. The formal & Wilson (2003), values for gravity-darkening exponents and error in the photometry is as low as 0.001–0.002 mag, but we albedos from theoretical values for radiative envelopes, and a have adopted a conservative error of 0.01 mag for the differ- mass ratio determined from the radial velocity curve (see below). ential light curve. V-band images were also obtained on one We fitted for the inclination i,Teff2 , the of the primary, p night during maximum. We measured the following magnitudes and the surface potential (Q12Q ), and we defined convergence and I p to be reached after five consecutive iterations for which the 0.025 ע for WR 20a at maximum:V p 13.45 corrections for all adjusted parameters were smaller than their .0.05 ע 10.65 respective standard (statistical) errors. The results of the fit are 3. LIGHT-CURVE AND RADIAL VELOCITY CURVE ANALYSIS given in Table 2. In Figure 1 we show the result of the I-band light-curve model fit for WR 20a. The uneven eclipse depths The first goal of our observations was to confirm and suggest slightly different effective temperatures and, thus, dif- refine the orbital period of the WR 20a binary. As discussed ferent spectral types for the components. Using differential pho- extensively by Rauw et al. (2004), besides the main period tometry, we can measure a difference between the two stars as ofP p 3.675 days, their spectroscopic data also allowed for small as 1%–2%, which is not possible with spectral decom- a much shorter period of 1.293 days, which would result in position. However, multiband photometry is necessary to resolve reduced minimum masses of about 25 M, for each star in the effective temperature–radius degeneracy. Finally, we derive the system. A longer period of 4.419 days was also possible, the following best-fit fractional radii for both stars: polar radius p p resulting in increased minimum masses of above 80 M, for rpole0.34, radius to the inner Lagrangian pointr point 0.40 , p p each star. Using the analysis of the variance technique of rside0.35, andr back 0.37 . Schwarzenberg-Czerny (1989), the most significant period Armed with the refined period and the exact value of the days, where the inclination angle, we reanalyzed the radial velocity data of Rauw 0.005 ע derived from our data isP p 1.843 error is a conservative 3 j estimate. However, we know that et al. (2004). Fixing the eccentricity to 0, as the light curve the system has to have two eclipses, and this results in a confirms, and the period to 3.686 days, we fitted the radial ve- true orbital period that is twice as long. We find an ephemeris locity data by applying equal weights and rejecting points with for the primary (deeper) eclipse of radial velocity measurements smaller than 80 km sϪ1. These measurements were made by fitting Gaussians to lines that are p ϩ Tprim 2,453,124.569 3.686E (JD), (1) not resolved and that thus include errors much larger than 10– 15 km sϪ1 and are also subject to non-Keplerian effects. Our fit No. 1, 2004 ECLIPSING BINARY WR 20a L35

Fig. 1.—Wilson-Devinney model fit of a near contact binary to the I-band Fig. 2.—Radial velocity curve fit to measurements of WR 20a made by Rauw light curve of WR 20a. The period is 3.686 days, the eccentricity is 0, and p Њ et al. (2004). The filled circles correspond to the primary (more massive) star. the inclinationi 74.5 . p Ϫ1 p Ϫ1 The new semiamplitude values areK12362 km s andK 366 km s , and 33p p the masses arem1 sin i 74.3 Mm, and2 sin i 73.4 M, , which, combined p Њ is shown in Figure 2, and the best-fit values with their statistical withi 74.5 , yield mass values for the components of 83.0 and 82.0 M,. errors are presented in Table 3. We note that due to the refined period, as Rauw et al. (2004) caution, their two earliest radial Space Telescope (Whitney et al. 2004) contain measurements velocity epochs are now “flipped.” We derive slightly larger of the binary. Belloni & Mereghetti (1994) have obtained velocity semiamplitudes than Rauw et al. (2004) do, which in ROSAT observations of RCW 49, tentatively identifying WR p 20a with one of their point sources. However, further obser- turn yield larger masses. Specifically, the values are K1 Ϫ1 p Ϫ1 p vations are needed to determine the extinction to WR 20a. 362.2 km s andK2 366.4 km s , the mass ratio q p p Ϫ1 Analogs of WR 20a, possibly with even higher mass com- m 21/m 0.99, the systemic velocityg 8kms , the semimajor axisa sin i p 53 R, , and finally the masses ponents, are bound to exist in nearby , such as M31 33p p and M33. There they would first be found by variability m1 sin i 74.3 Mm, and2 sin i 73.4 M, . Knowing the .searches such as DIRECT (Stanek et al. 1998; Bonanos et al ע inclination, we calculate the masses of the stars to be 83.0 M,. These errors include the error in the 2003) as bright eclipsing binaries (about 18th magnitude for 5.0 ע 5.0and 82.0 p Ϫ minimum mass and the error in the inclination. MV 6.5), which can then be followed by spectroscopy to obtain radial velocities. In our experience (A. Z. Bonanos et al. 2004, in preparation), accurate radial velocities can be ob- 4. DISCUSSION tained for such systems with the existing 6.5–10 m telescopes. As discussed at length by Rauw et al. (2004), the extreme The advantage of studying eclipsing binaries in these galaxies mass values of WR 20a make it a “cornerstone system for would be their well-known distances and low reddening, since future investigations of massive star evolution.” We confirm systems with high reddening would be too faint to follow up. and refine these extreme values, by measuring the masses of We conclude that WR 20a truly is a “cornerstone system” M,.Ifthe that deserves detailed additional studies. Studies of the most 5.0 ע and 82.0 5.0 ע the components to be83.0 WN6ha spectral type is confirmed, WR 20a will be one of the massive binaries will not only constrain models but will also few known W-R binaries with a mass ratio near unity. Thus, help us to understand other problems, such as the nature of it becomes very important now to obtain high-resolution spectra mass loss and of winds of early-type stars, and the connection and multiband photometry of WR 20a, corrected for eclipses. these early-type stars have to supernovae and gamma-ray bursts Such future observations will allow one to take an accurate (Stanek et al. 2003; Maund et al. 2004). WR 20a now overtakes reddening measurement and, furthermore, to make a direct, independent and accurate distance determination of the system. TABLE 3 This will resolve the discrepancy in the distance to Westerlund Radial Velocity Curve Parameters 2, which ranges from 2.5 to 8 kpc (see discussion in Churchwell for WR 20a et al. 2004). Several multiband measurements of WR 20a al- Parameter Value ready exist. As measured by us (§ 2), WR 20a has V p 5.0kmsϪ1 ע p Systemic velocity, g ...... 8.0 Ϫ1 ע andI 10.65 at maximum. Our V is different from the 13.45 Semiamplitude, K1 ...... 362.2 8.0kms Ϫ1 ע -V p 13.58 measured by Moffat et al. (1991), but their mea Semiamplitude, K2 ...... 366.4 8.0kms ,R 1.0 ע surement could have been affected by the eclipses. The Two Semimajor axis, a sin i ...... 53.0 ע p Micron All Sky Survey (Skrutskie et al. 1997) has measured Mass ratio, q m21/m ...... 0.99 0.03 ע 3 p p p Mass, m1 sin i ...... 74.3 4.0 M, J 8.86,H 8.08 , andK 7.59 for WR 20a. Recent in- 3 ,M 4.0 ע Mass, m sin i ...... 73.4 frared observations of the H ii region RCW 49 with the Spitzer 2 L36 BONANOS ET AL. Vol. 611 the previous record holders WR 22 (Rauw et al. 1996; comments. A. Z. B. and K. Z. S. were partially supported by Schweickhardt et al. 1999) and R136-38 (Massey et al. 2002) HST grant HST-GO-09810.06-A. Support for Proposal HST- in the most massive star competition, by becoming “the most GO-09810.06-A was provided by NASA through a grant from massive star ever weighed.” the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., We thank G. Torres for a careful reading of and comments under NASA contract NAS5-26555. Partial support to the on the manuscript and D. Sasselov and J. Devor for useful OGLE project was provided by the following grants: Polish discussions. We also thank the referee, Eric Gosset, for his KBN grant 2P03D02124 to A. Udalski and NSF grant AST- prompt and careful reading of the manuscript and his useful 0204908 and NASA grant NAG5-12212 to B. Paczyn´ski.

REFERENCES

Alard, C. 2000, A&AS, 144, 363 Schwarzenberg-Czerny, A. 1989, MNRAS, 241, 153 Bagnuolo, W. G., Gies, D. R., & Wiggs, M. S. 1992, ApJ, 385, 708 Schweickhardt, J., Schmutz, W., Stahl, O., Szeifert, Th., & Wolf, B. 1999, Belloni, T., & Mereghetti, S. 1994, A&A, 286, 935 A&A, 347, 127 Bonanos, A. Z., Stanek, K. Z., Sasselov, D. D., Mochejska, B. J., Macri, Shara, M. M., Smith, L. F., Potter, M., & Moffat, A. F. J. 1991, AJ, 102, 716 L. M., & Kaluzny, J. 2003, AJ, 126, 175 Skrutskie, M. F., et al. 1997, in ASSL Vol. 210, The Impact of Large Scale Churchwell, E., et al. 2004, ApJS, in press Near-IR Sky Surveys, ed. F. Garzo´n et al. (Dordrecht: Kluwer), 25 Crowther, P. A., & Dessart, L. 1998, MNRAS, 296, 622 Stanek, K. Z., Kaluzny, J., Krockenberger, M., Sasselov, D. D., Tonry, J. L., Davidson, K., & Humphreys, R. M. 1997, ARA&A, 35, 1 & Mateo, M. 1998, AJ, 115, 1894 Eikenberry, S. S., et al. 2004, ApJ, in press (astro-ph/0404435) Stanek, K. Z., et al. 2003, ApJ, 591, L17 Figer, D. F., Najarro, F., Morris, M., McLean, I. S., Geballe, T. R., Ghez, Udalski, A. 2003, Acta Astron., 53, 291 A. M., & Langer, N. 1998, ApJ, 506, 384 van der Hucht, K. A. 2001, NewA Rev., 45, 135 Landolt, A. U. 1992, AJ, 104, 372 van Hamme, W., & Wilson, R. E. 2003, in ASP Conf. Ser. 298, GAIA Spec- Massey, P., Penny, L. R., & Vukovich, J. 2002, ApJ, 565, 982 troscopy: Science and Technology, ed. U. Munari (San Francisco: ASP), Maund, J. R., Smartt, S. J., Kudritzki, R. P., Podsiadlowski, P., & Gilmore, 323 G. F. 2004, Nature, 427, 129 Walborn, N. R., Morrell, N. I., Howarth, I. D., Crowther, P. A., Lennon, Moffat, A. F. J., Shara, M. M., & Potter, M. 1991, AJ, 102, 642 D. J., Massey, P., & Arias, J. I. 2004, ApJ, 608, 1028 Nelson, B., & Davis, W. D. 1972, ApJ, 174, 617 Whitney, B. A., et al. 2004, ApJS, in press (astro-ph/0406100) Popper, D. M., & Etzel, P. B. 1981, AJ, 86, 102 Wilson, R. E. 1979, ApJ, 234, 1054 Rauw, G., et al. 2004, A&A, 420, L9 Wilson, R. E., & Devinney, E. J. 1971, ApJ, 166, 605 Rauw, G., Vreux, J.-M., Gosset, E., Hutsemekers, D., Magain, P., & Rocho- wicz, K. 1996, A&A, 306, 771