The central WNh stars of R136

VLT/SINFONI time-resolved spectroscopy of the central, luminous, H-rich WN stars of R136thanks: Based on observations collected at the European Organisation for Astronomical Research in the Southern Hemisphere, Chile, under program ID 076.D-0563, and on observations made with the Hubble Space Telescope obtained from the ESO/ST-ECF Science Archive Facility

O. Schnurr, A.-N. Chené, J. Casoli, A. F. J. Moffat, and N. St-Louis
Dept. of Physics and Astronomy, University of Sheffield, Hicks Building Hounsfield Road, Sheffield S3 7RH, United Kingdom
Dépt. de Physique , Université de Montréal, C. P. 6128, succ. centre-ville, Montréal (Qc) H3C 3J7, and Centre de Recherche
en Astrophysique du Québec, Canada
Herzberg Institute of Astrophysics, 5071 West Saanich Road, Victoria (BC) V9E 2E7, Canada
Ecole Normale Supérieure, 45, rue d’Ulm, 75230 Paris Cédex 05, France
Version 18 May 2009

Using the Very Large Telescope’s Spectrograph for INtegral Field Observation in the Near-Infrared (VLT/SINFONI), we have obtained repeated AO-assisted, NIR spectroscopy of the six central luminous, Wolf-Rayet (WR) stars in the core of the very young ( Myr), massive and dense cluster R136, in the Large Magellanic Cloud (LMC). We also de-archived available images that were obtained with the Hubble Space Telescope’s Space Telescope Imaging Spectrograph (HST/STIS), and extracted high-quality, differential photometry of our target stars to check for any variability related to binary motion.

Previous studies, relying on spatially unresolved, integrated, optical spectroscopy, had reported that one of these stars was likely to be a 4.377-day binary. Our study set out to identify the culprit and any other short-period system among our targets. However, none displays significant photometric variability, and only one star, BAT99-112 (R136c), located on the outer fringe of R136, displays a marginal variability in its radial velocities; we tentatively report an 8.2-day period. The binary status of BAT99-112 is supported by the fact that it is one of the brightest X-ray sources among all known WR stars in the LMC, consistent with it being a colliding-wind system. Follow-up observations have been proposed to confirm the orbital period of this potentially very massive system.

binaries: general – stars: evolution – stars: fundamental parameters
pagerange: VLT/SINFONI time-resolved spectroscopy of the central, luminous, H-rich WN stars of R136thanks: Based on observations collected at the European Organisation for Astronomical Research in the Southern Hemisphere, Chile, under program ID 076.D-0563, and on observations made with the Hubble Space Telescope obtained from the ESO/ST-ECF Science Archive FacilityReferencespubyear: 2009

1 Introduction

30 Doradus (NGC 2070) is an active star-forming, giant Hii region in the Large Magellanic Cloud (LMC). At the center of 30 Dor is located the very massive star cluster R136 (HD 38268), which is regarded as the closest visible example of a “super star cluster”, excluding its Galactic clone NGC 3603, which is not surrounded by a very massive cluster halo.

The inner arcsecond of R136, denoted R136a, is unresolvable by conventional ground-based telescopes, and it was suspected that R136a is in fact a single stellar object of more than 3000 M (Cassinelli et al. 1981, Feitzinger et al. 1980). However, careful ground-based work (both direct imaging: Moffat & Seggewiss 1983; Moffat, Seggewiss & Shara 1985; and speckle interferometry: Weigelt & Baier 1985) showed that R136a was not a single star. High-resolution imaging by the Hubble Space Telescope (HST) confirmed that R136a indeed consists of individual, hot and luminous stars (Campbell et al. 1992), while the region immediately surrounding R136 contains many dozens of massive O stars, many of them of spectral type O3, the hottest of all known O-type stars at the time (Massey & Hunter 1998).

It has been theorized that high stellar densities in the cores of very massive proto-clusters are a prerequisite for the formation of very massive stars (Bate 2002). Moreover, there is empirical evidence that the mass of the most massive cluster member correlates with the total mass of the cluster (Weidner & Kroupa 2006). Therefore, it can be expected that the most massive stars known are found in the core regions of the most massive, and densest, unevolved clusters known. Given its core density of Mpc (e.g. Moffat et al. 1985), R136a is hence a prime candidate for harboring extremely massive stars.

It is a very remarkable fact that even the hottest and most massive O-type stars seem not to exceed M (Lamontagne et al. 1996; Massey et al. 2002). More massive stars are invariably members of a very luminous and hydrogen-rich subtype of Wolf-Rayet (WR) stars of the nitrogen-rich sequence, the so-called WN5-7h stars. Studies using model-atmospheres indicate that these WN5-7h stars are not classical, core-helium burning objects, usually identified with the WR phase, but rather core-hydrogen burning objects on the the main sequence; their WR-like appearance is due to their very high luminosities, ; de Koter et al. 1997; Crowther & Dessart 1998) which drive fast stellar winds, whose high densities give rise to the emission-line spectrum. As has been confirmed by weighing WN5-7h stars with Keplerian orbits in eclipsing binaries, the very high luminosities of WN5-7h stars indeed correspond to very high masses: The present record holder, NGC3603-A1, tips the balance at 116 M (Schnurr et al. 2008a).

Following the spectral classification of Crowther & Dessart (1998), R136a contains three hydrogen-rich WN5h stars, and one O3f/WN6 star which is a transition type between the hottest Of stars and the least extreme WN5-7h stars (Walborn 1986) and thus weaker-lined than the WN5h stars. Slightly off-centered are a cooler, evolved WN9h star (R136b), and another WN5h star (R136c). From ground-based, spatially unresolved (i.e. integrated), optical spectroscopy of R136a, Moffat & Seggewiss (1983) reported a 4.377-day WNh binary in R136a, with a diluted radial-velocity amplitude kms; the subsequent study by Moffat et al. (1985) confirmed this finding. The situation in R136a seems thus very similar to that in NGC 3603, where from unresolved spectroscopy of the central arcsecond, too, Moffat & Niemela (1984) had identified a 3.77-day binary among the three WN6h stars (this finding was confirmed by Moffat et al. 1985 as well). Follow-up observations of NGC 3603 confirmed these results (Moffat et al. 2004; Schnurr et al. 2008a), revealing a second close binary, C, in NGC 3603, with an orbital period days. Hence a more detailed investigation of the six WNh stars in R136a was warranted, since it offered the potential to increase considerably the number of known, very massive binary systems.

We have therefore obtained, for the first time, repeated, spatially resolved spectroscopy of the six WNh stars in R136, in order to single out the 4.377-day binary and to identify any further short-period system among our targets. To search for eclipsing systems (photospheric or atmospheric) among the stars, additional optical HST photometry of our target stars was extracted from publicly available, archival imaging data.

Our study also complements that of Schnurr et al. (2008b), who had surveyed the 41 of the 47 known, late-type WN stars in the LMC according to BAT99 located outside R136a, and thereby concludes the efforts of the Montréal hot-star group to monitor every WR star in the Magellanic Clouds to establish their binary status.

In the present paper, we describe the data acquisition and reduction (Section 2), and the data analysis (Section 3). We discuss our findings in Section 4, and close with the summary and conclusion in Section 5.

2 Observations and data reduction

2.1 Near-infrared spectroscopy

Targets and their relevant properties are listed in Table 1, giving both the numbers from the BAT99 catalogue (Breysacher et al. 1999) and the older Radcliffe numbers (e.g. Feast et al. 1960).

Observations were carried out in service mode at the Very Large Telescope (VLT) with Unit Telescope 4 under Program-ID P076.D-0563, between November 13 and December 5, 2005. The observations thus cover a time span of 22 days. We obtained repeated K-band (1.95 to 2.45 m) spectroscopy using the Spectrograph for INtegral Field Observation in the Near-Infrared (SINFONI) (Eisenhauer et al. 2003; Bonnet et al. 2004) with adaptive-optics (AO) correction to obtain the highest possible spatial and spectral resolution. BAT99-111 (R136b) served as AO reference star. The field of view was with a “spaxel” scale of 12.5 mas 25 mas. Our six target WNh stars were observed with four pointings (with R136a1, a2 and a5 all in one field), defined as telescope offsets from the AO guide star BAT99-111. Figure 1 shows a montage of the four fields as seen on the sky, reconstructed from the SINFONI data cube. Stars here are separated at least as well as on HST images.

Figure 1: Montage of the four fields as seen on the sky, reconstructed from the SINFONI data cube. Target stars are identified by the BAT99 numbers. North is up and East is left.
BAT99 other spec
name type mag mag mag mag
106 R136a3 WN5h 13.1 +0.20 0.28 -6.6
108 R136a1 WN5h 12.9 +0.18 0.28 -6.8
109 R136a2 WN5h 13.1 +0.21 0.28 -6.7
110 R136a5 O3If*/WN6-A 14.0 +0.21 0.28 -5.7
111 R136b WN9h 13.4 -6.5
112 R136c WN5h 13.6 -6.3
Table 1: Target list of the observations. For easier identification, both BAT99 and Radcliffe numbers are given together with the spectral types (based on optical spectra). colors and were directly adopted from Crowther & Dessart (1998), while their Johnson- magnitudes were used to estimate narrow-band -band magnitudes, applying mag to take into account the contribution of emission lines.

Total exposure times were 150s per star and per visit, each organized in 2 detector integration times (DITs). Given the brightness of our targets ( 11–12mag) and the AO deployment, no dedicated sky frames were taken. Other calibrations (dark and flat-field frames, and the telluric standard star) were provided by the ESO baseline calibration.

For most of the data reduction steps, ESO’s pipeline was used (cf. Abuter et al. 2006). Standard reduction steps were taken. The two-dimensional spectra produced by each illuminated slitlet were individually extracted using iraf, and combined into one wavelength-calibrated spectrum per star and visit. A main-sequence B-type star was used for telluric corrections. To remove the B star’s Br absorption which coincides with the Br/Hei 2.166 m emission blend of the target WNh stars, a Lorentzian was fitted to the absorption line and subtracted from the B star’s spectrum. Residuals were very small, and proved to be harmless in the subsequent analysis. Finally, science spectra were rectified by fitting a low-order spline function to the stellar continuum. The final uniform stepwidth of the spectra was 2.45 Å/pixel, resulting in a conservative three-pixel resolving power of 3000, and a velocity dispersion of 33 kms/pixel.

2.2 Optical photometry

We have de-archived optical imaging of R136 that was obtained in Cycle 8 as part of the HST program GO-8217 (PI: Philip Massey). Using the Space Telescope Imaging Spectrograph (STIS), 30 pairs of short (1.1-1.3s) images were obtained at various roll angles through the long-pass filter (F28X50LP). The image size was and the spatial scale was 0.05 per pixel. For more details on the data, we refer the reader to Massey et al. (2002). Standard pipeline data reduction was carried out on the fly by the HST archive software, and no further data handling was done on our part.

3 Data analysis and results

3.1 Spectroscopy

Figure 2: Montage of average spectra of our six target stars, arranged by spectral type and limited to the useful spectral region. The two strongest emission lines, Br/Hei 2.166 m and Heii 2.188 m are indicated. For clarity, the spectra have been shifted by 0.25 flux units, starting with the second from the bottom.

The mean spectra of the six WNh stars are shown in Figure 2. The systemic velocity of each star was measured by fitting Gaussians to the reasonably symmetric, top third of the Heii m emission line (below that, the line quickly becomes too asymmetric for this). For the WN5h stars, which have the strongest emission lines, this was done for each individual spectrum of the time series. The standard deviation of the series was adopted as error, , and the error of the mean of the systemic velocity was calculated as , with the number of spectra.

For the two weaker-lined stars, BAT99-110 (O3f/WN6) and 111 (WN9ha), the fit could only be carried out for the averaged spectrum since the S/N ratio was better. Also these two stars display obvious P Cygni profiles in Heii m, hence two Gaussians (one absorption, one emission) plus a linear continuum were fitted to the entire line. The resulting errors on the fits are comparably large, in particular for the faint O3f/WN6 star, because its very weak lines remained buried in the noise even in the mean spectrum. Since only one spectrum was used, no error of the mean was calculated. Values are listed in Table 2.

BAT99 other Spectral eom comment
name type (kms) (kms) (kms)
106 R136a3 WN5h 380 11 3.7 14 RV reference
108 R136a1 WN5h 374 19 6.3 19 RV reference
109 R136a2 WN5h 392 17 5.7 13 RV reference
110 R136a5 O3f/WN6 288 63 em. 33 very weak P Cyg
52 16 abs.
111 R136b WN9ha 348 36 em. 26 weak P Cyg
157 6 abs.
112 R136c WN5h 389 21 7.0 28 variable?
Table 2: Systemic velocities, standard deviations , and errors of the mean () of our program stars, obtained by fitting a single Gaussian to the Heii m line. For BAT99-110 and 111, two Gaussians (emission and absorption) were fitted to the P Cygni profile to the mean spectrum; therefore, both RVs are given, but no error of the mean was calculated. The listed RV scatter was obtained from cross-correlation for each individual star. Those stars used as RV reference are flagged accordingly.

All four WN5h stars display systemic velocities that are redder (larger) than the systemic velocity expected from the LMC, kms (e.g. Kim et al. 1998). This redshift is most likely of the same nature as that reported for the optical Heii line in other WN and O3f/WN6 stars in the LMC (e.g. Schnurr et al. 2008b), i.e. due to radiative-transfer effects (Hillier 1989). The mean systemic velocity of the four WN5h stars is kms (standard deviation), whose dispersion is comparable to their individual errors of the mean; these stars thus have systemic velocities which are statistically consistent with each other.

For BAT99-110, the emission part of the Cyg profile for Heii m surprisingly displays a significantly smaller redshift than the Heii m emission of WN5h stars. In fact, despite the large error, BAT99-110’s systemic velocity is in excellent agreement with that of the LMC. Either near-infrared Heii emission lines of O3f/WN6 stars do not display the intrinsic redshift of their optical counterparts (possibly because of optical-depth effects, since their wind is thinner than that of, say, WN5h stars) or BAT99-110 is indeed travelling at relatively high differential velocity along the line of sight towards the observer. In this case, it could be that 110 is not actually in the core, but rather ejected from it towards the observer, i.e. along the line of sight (also see below). Thus, from a point of view of cluster dynamics and cluster evolution, it would be very interesting to carry out a more detailed RV study of the massive stars in the periphery of R136a to check for ejected cases (see also Brandl et al. 2007 for a more detailed discussion of this topic).

Figure 3: Relative radial velocities as obtained from cross-correlation, normalized so that the average kms. Error bars () shown in the upper left of each panel are adopted from the reference stars, kms. Zero velocity is shown by the dashed line. Left graphs show RVs plotted against time, right graphs show the data folded into the phase corresponding to the period of P = 4.377 days of the putative binary (with arbitrary zero phase). BAT99 numbers are given in the upper right corner of each panel. No coherent RV variation can be seen in the data, with the possible exception or BAT99-112 (R136c; top left row). See text for more details.

To identify the variable stars in our sample, relative radial velocities (RVs) were measured by cross-correlation, limited to the region from 2.134 to 2.214 m which comprises the two strongest emission lines, Br/Hei m and Heii m. The iterative approach of Schnurr et al. (2008b) was followed: A high-S/N mean spectrum of the time series acted as cross-correlation template for each star. Then, all spectra of the time series were shifted by their respective RVs, to generate a better mean template spectrum, and the cross-correlation was repeated with the new template. Figure 3 shows time-plots of these relative RVs for each star.

Even without this iteration scheme, the resulting RV scatter of each individual star, is surprisingly small; final values are listed in Table 2. Following the approach of Schnurr et al. (2008b) to derive a posteriori error bars, we proceeded to construct a RV reference star by using the least variable stars in our sample, the three WN5h stars in R136a. For each of these stars, relative RVs obtained by cross-correlation were normalized to to correct for different velocities of the cross-correlation template, and combined. For the resulting 33 data points in total, the overall standard deviation kms was obtained. This value was adopted as measurement error for each individual data point.

To determine whether or not the RV variability of a given star is statistically significant, we adopt as a rough estimate of the threshold above which a star can be considered to be variable at the 99.9% level (see Schnurr et al. 2008b for a more detailed discussion). We obtain kms from our three reference stars, and find that R136a5, R136b, and R136c exceed this threshold. However, the large scatter for the first two stars can be explained by very weak lines and thus relatively lower S/N, which greatly affects the precision of the RV measurement. In contrast to this, R136c is a stronger-lined WN5h star, and well isolated, hence its large RV scatter is likely intrinsic.

Including R136c in our RV reference would yield a somewhat larger error, kms, and accordingly kms. Hence, R136c would have exceeded the cut-off value in any case (if just), thus regarded as variable, and been removed from the reference group.

A period search was carried out on the RV data of R136c, constrained to the range from 2 to 44 days. The lower limit of this range is set by the mean sampling frequency, while the upper limit is twice the total time covered by the observations, 22 days. However, due to the low level of variability, no significant (at or above ) period can be found. Furthermore, the scarcity of data points leads to many empty phase bins if a phase-dispersion method is applied. Thus, more and better data are required to confirm the variability of R136c, and eventually its binary status.

However, if a sine wave is fitted to the RV data of R136c (i.e. assuming zero orbital eccentricity), the (marginally) best fit is obtained for a period days, with an amplitude kms (see Figure 4). For this solution, kms, which is unrealistically small, and essentially produced by only one data point (around ), which does not lie on the curve. If this point is excluded from the fit, then kms, which is an indication for an ill-constrained fit (effectively six independent data points for four parameters) rather than real. While it is not impossible that SINFONI delivers such good data, it sheds some doubt onto the reality of the orbital solution. We therefore consider R136c as a marginal case for supporting a binary, subject to future confirmation.

Figure 4: Orbital solution for BAT99-112 (R136c) if a sine wave (i.e. a circular orbit) is forced to the RV data obtained through cross-correlation.

3.2 Photometry

Crowded-field, quasi-white light photometry was extracted from each of the STIS frames using the daophot ii package with the additional allstar and allframe subroutines (Stetson 1992; Stetson 1994). The point-spread function (PSF) was iteratively fitted using a Moffat function () whose full-width at half-maximum (FWHM) was typically pixels. The fitting radius around stars was set to 4 pixels and the extraction radius of the aperture was 1.5 FWHM pixels. Sky was determined from an annulus between 7 and 9 pixels from the center of the star. In order to separate all the stars in the crowded center of the cluster, profile-fitting was done iteratively. That way, we were able to obtain reliable photometry of BAT99-108, 109 and 100, as well as 3 other fainter stars in the crowded field around BAT99-109.

To carry out differential photometry, a single adjustment to the photometric zero point was applied as in Massey et al. (2002). For each star, the dispersion around the time-averaged magnitude is plotted against its instrumental magnitudes (Figure 5). Indicated are both our program stars and the binaries and variable stars reported by Massey et al. (2002). Since the exposures were very short ( sec), our target stars, which are the brightest objects in the field, remain well below the non-linearity limit.

Figure 5: Photometric dispersion of the times-series of each source in the STIS field-of-view, plotted against its instrumental magnitude. The solid line is the curve, the dashed line the curve according to the model described in the text. Our target stars are identified by their respective BAT99 number. We have also identified eclipsing binaries (squares) and other variables (triangles) that were reported by Massey et al. (2002), using their nomenclature (R136-nnn).

These very short exposures times have repercussion on the photometric error statistics. If there were no exterior noise sources, photometric accuracy would be simply described by Poissonian statistics, and related to the stellar intensity I such that . In reality, however, read-out noise becomes quickly very important for the photometric accuracy of fainter stars. The photometric precision of very bright stars, on the other hand, is limited by the quality of the flat-field frame. To obtain a reliable error statistics, we therefore have applied a more detailed noise model, taking into account the following noise sources: ) The Poisson noise of the stellar intensity I, which is simply . ) The read-out noise , which for STIS is 4 e per pixel (gain = 1). Note that the mean read-out level , since there is no flux associated with a read-out operation, only an error on the count rate. The extraction radius of the PSFs are pixels (see above), hence PSFs cover a total area of pixels; hence the total read-out noise sums to e. ) The noise of the flat-field frame, which we assume to be normalized to unity prior to dividing the science exposures by it, so that the mean flat-field factor .

Since the HST is in space and exposures are too short for a significant stray-light contribution, there is no sky background and its associated noise that need to be considered. We further note that the flat-field noise is multiplicative (it is proportional to the intensity recorded by a pixel), whereas the read-out noise is purely additive. Hence, the photometric precision for bright stars is limited by the quality of the flat-field frame, whereas faint stars are more affected by read-out noise.

Thus, from the usual relation

where is a magnitude zero point, error propagation for independent values of F, I, and R then yields the total error on the measured magnitude,

(with the Euler number), which reduces to

Using (which is consistent with the typical S/N of a flat-field frame), and , this noise model fits the data very well. Both the and curves, based on this model, are shown in Figure 5. As can be seen, even the brightest stars in our sample do not yet reach the noise floor that is determined by the flat-field noise.

As can clearly be seen, our target stars do not display any significant photometric variability; in fact the dispersion around their respective mean magnitudes is very small (see Table 3 for individual values). Since we were perfectly able to identify the eclipsing binaries and other variable stars reported by Massey et al. (2002), we have to conclude that our target stars appear to be photometrically constant. This is especially relevant for BAT99-112 (R136c), which despite its short spectroscopic period does not show any indications of phase-dependent variability.

BAT99 other STISmag
name (mag)
106 R136a3 12.850 0.013
108 R136a1 12.233 0.011
109 R136a2 12.863 0.016
110 R136a5 13.778 0.017
111 R136b 13.124 0.015
112 R136c 13.072 0.013
Table 3: Mean instrumental magnitudes (through the long-pass filter) and standard deviation around the mean for our target stars as obtained from HST-STIS images using daophot ii.

4 Discussion

With our data, we cannot confirm the existence of the 4.377-day binary that was reported in optical emission lines by Moffat & Seggewiss (1983) and Moffat et al. (1985). Moreover, we cannot clearly identify any short-period binary among our six target stars; there only is one binary candidate, BAT99-112 (R136c). This star was already known to be a very bright X-ray source. Several studies have reported a very high X-ray luminosity of ergs s for BAT99-112 (Portegies Zwart et al. 2002; Townsley et al. 2006; Guerrero & Chu 2008), in fact the second-brightest object in the greater R136 area after BAT99-116 (Mk 42). It is very likely that both stars are colliding-wind binaries (cf. Usov 1992). Interestingly, Schnurr et al. (2008b) reported significant RV variability for BAT99-116 as well, but were unable to establish a periodicity. While it is thus very likely that both are long-period systems, we note that short-period systems, i.e. with a small orbital separation, are not ruled out by the high X-ray luminosities. NGC3603-C is an 8.9-day binary with one of the highest X-ray fluxes known among all WR stars (Moffat et al. 2002; Schnurr et al. 2008a), so it seems that self-absorption of X-ray photons in the dense WN wind is not too much of a problem even in relatively close systems.

Thus, the following question arises: If one or more of our target stars in R136 are indeed short-period binaries, could their non-detection be the result of low orbital inclination angles? After all, very low inclination angles could comfortably explain why neither clear RV nor photometric variability (i.e., eclipses or ellipsoidal variations) are found, despite any possible enhancement of short-period systems (see above). To investigate this in more detail, we have calculated the RV scatter of the primary (WN-type) component that would be expected from the continuous sampling of a circular binary with orbital periods ranging from 1 to 100 days. In this case study, the masses of the two components have been fixed to 90 for the WN5h primary and 30 M for the presumed O-type secondary, i.e. the system has a total mass of 120 M (the exact value is not critical), and a mass ratio . The large value is reasonably pessimistic, given that such large ratios are observed even in very massive systems: In many reported cases, the companion is too faint to be (easily) detected (e.g. Schweickhardt et al. 1999; Schnurr et al. 2008a,b; Schnurr et al. 2009).

The expected value for different values for the inclination angle, to in steps of , are shown in Figure 6. Also indicated are, with two vertical, dotted lines, the position of the expected 4.4-day binary, and the longest period we can reasonably hope to find in our data set, twice the time coverage of our observations, i.e. 44 days. The dash-dotted, horizontal line marks the observed RV scatter for BAT99-112, kms.

From Figure 6 it becomes clear that if BAT99-112 were indeed a short-period binary, it would have to be seen under a very low inclination angle, . However, in a sample with randomly distributed inclination angles, there would be

times more binaries between and than between and ; thus, it is very unlikely that there is a significant number of such low-inclination systems in our sample.

Figure 6: RV scatter of the WN component in a fictuous, continuously sampled, circular binary, with masses fixed at 90 M for the WN5h primary and 30 for the O-type secondary, as a function of the orbital period of the system. Curves show different values of the orbital inclination angle, , from top. The dot-dashed, horizontal line indicates the measured scatter for R136c (BAT99-112), kms. The two dotted, vertical lines indicate the period of the putative binary, 4.4 days, and the upper limit of periods we can reasonably hope to detect in our 22-day long campaign, 44 days.

We note here that Moffat & Seggewiss (1983) and Moffat et al. (1985) had reported a RV amplitude of kms for R136a, which seems surprisingly consistent with kms. However, Moffat & Seggewiss (1983) and Moffat et al. (1985) only had integrated, i.e. spatially unresolved spectroscopy, and they reported the binary to be member of R136a, i.e. one of the three WN5h stars. (The fourth star, BAT99-110, is O3f/WN6 and hence too faint both in magnitude and emission-line strength to significantly have contributed to the integrated signal.) Our SINFONI observations, however, have individually resolved the stars; hence one would expect an RV amplitude at least three times higher, i.e. kms. Even if one allows for the fact that due to non-continuous sampling, the measured could be somewhat smaller, such large RV scatter is simply not seen among the WN5h stars in R136a.

Hence, we have to conclude that the large scatter in Moffat & Seggewiss’s (1983) and Moffat et al.’s (1985) data was not due to the binary motion of one of the target stars, but of different origin. A possible explanation could be that during the observations, the slit was not located at exactly the same position each time, e.g. due to difficulties introduced by variable seeing conditions, and that thus the light took different optical paths through the instrument. This could lead to an increased RV scatter, while well-isolated single stars are much less affected by this problem (see e.g. Schnurr et al. 2008b). It thus seems that with the exception of the 8.2-day (but possibly longer-period) binary candidate BAT99-112, there is no binary among our six program stars.

For BAT99-111 (R136b), this result does not come unexpectedly: Crowther & Dessart (1998) re-classified this star as WN9ha, and there seems to be a dichotomy between WN6,7 and WN8,9 stars when it comes to their binary status: No WN8,9 star is known to reside in a short ( days) WR+O binary system (Milky Way: Moffat 1989; LMC: Schnurr et al. 2008b). Also, if BAT99-111 is indeed a WN9ha star, then it is a descendent of an O8Iaf star (Crowther & Bohannan 1997), i.e. less massive but slightly more evolved than the O3f/WN6 and WN5h stars in R136. This could be an indication that BAT99-111 is not a member of R136a’s relatively unevolved stellar generation, but rather from an older population of massive stars outside R136, and simply located in the line of sight.

If, on the other hand, BAT99-111 is indeed an O4Iaf star, as classified by Massey & Hunter (1998), then both the spectral type and the fainter visual magnitude are consistent with BAT99-111 being a less extreme version of the brighter O3f/WN6 stars in and around R136, and a member of R136a after all. Unfortunately, our data do not allow us to settle this classification issue.

For the other five, less evolved objects in our sample, in particular for the three very crowded WN5h stars in the core of R136a, the null-result is somewhat surprising. We had expected to find at least one short-period binary in R136a, similar to the situation in its almost perfect, Galactic twin NGC3603, where Schnurr et al. (2008a) reported that of the three central WNh stars, two are binaries with days. Also, Schnurr et al. (2008b) report that among the nine O3f/WN6 and WN5-7h stars in 30 Dor, but outside R136, only two (20%) are confirmed spectroscopic binaries. This value (at least) should have been recovered in the core of R136 as well, since in the present study, we are using data of almost identical quality.

Considering the small number of studied stars, however, our non-detection of systems is fully consistent with a frequency of 20% of binaries with days. In fact, combining our results with that of Schnurr et al. (2008b), we can confirm that the short-period binary frequency among very massive stars around R136 within 30 Dor is not very high: Out of the total of fifteen O3f/WN6 and WN5-7h stars now studied, only two are confirmed binaries with days, i.e. 2/15 = 13%. If we include the candidate BAT99-112 (R136c) as a positive detection, we obtain 3/15 = 20% binary frequency. If the distribution of orbital periods were flat in log P (or proportional to 1/P), something which is usually assumed, we would find 13% (20%) binaries in each period bin [1;10], [10;1000], and [100;1000], i.e. the total frequency of binaries with orbital periods shorter than 1000 days thus would be 40% (60%).

Very recently, Bosch et al. (2009) have reported on the results of a spectroscopic monitoring campaign of absorption-line OB stars in the 30 Dor region; for 52 stars, six to seven spectra per star were obtained over a time span of 500 days. While this small number of spectra does not allow one to establish an orbital solution, Bosch et al. (2009) were able to identify binaries from their significantly large RV scatter compared to single stars and binaries with much longer periods; among the 52 stars monitored, 25 stars showed RV variability. If we suppose that with their method, Bosch et al. (2009) were able to detect binaries with periods up to 1000 days (i.e. twice the covered time span), the binary frequency is thus %. Taking into account all statistical errors, our null-result for days reported here is perfectly consistent with this value. Remarkably, Monte Carlo simulations by Bosch et al. (2009) report that their observational results are consistent with a binary frequency of 100%. Although our observations were only designed to identify the reported 4.4-day culprit, and we have not yet enough data points to identify systems much beyond days, we can rule out such a high binary frequency for our stars with strong winds, because only BAT99-112 shows an X-ray excess.

5 Summary and Conclusion

We have obtained for the first time, spatially-resolved, repeated, low-spectral resolution, near-infrared spectroscopy of the most luminous stars in the core of R136 in an attempt to identify the 4.377-day binary that was reported by Moffat & Seggewiss (1983) and Moffat et al. (1985). Additional archival HST-STIS imaging was used to extract photometry of our target stars.

For none of the studied stars could significant photometric variability be found. Furthermore, we cannot confirm the presence of a binary system with the reported 4.4-day period, nor do we identify any other short-period ( days) system among the most luminous stars in R136. One star, however, BAT99-112 (R136c), shows small, marginally significant RV scatter. A forced sine-fit to the data yields a best period days. While it is not entirely impossible, it is not very likely to see a binary under very low inclination angle (), which would be required to reconcile BAT99-112’s 8.2d binary nature with the observed, low RV scatter. Thus, it is more probable that BAT99-112 is a longer-period system that the limited time coverage of our observations could not detect, as is suggested by the fact that the star is one of the brightest X-ray sources among all known WR stars in the LMC.

To settle this issue, long-term monitoring of BAT99-112 will be required, but the possibility that it is indeed binary clearly harbors the potential to weigh one of the most massive stars known, and to obtain a clearer picture of the very upper main sequence in high-density, high-mass environments like R136.

The question as to why R136 appears to have a dearth in its short-period binary content, compared to its Galactic counterpart NGC 3603, remains to be explained, if not simply due to small numbers. A key difference, though, is the fact that NGC 3603 is not surrounded by a massive stellar halo as is R136 (i.e. 30 Dor). Another difference is the factor-of-two lower metallicity in R136. How these or other as-yet unknown factors play out remains a mystery. In fact, ultimately, long-term monitoring of all the luminous stars in R136 is necessary to check for long-period systems.


OS is grateful for financial support by PPARC/STFC. ANC is grateful to Peter Stetson for invaluable help with the daophot software package. AFJM and NSL are grateful for financial aid to NSERC (Canada) and FQRNT (Québec).


  • Abuter et al. (2006) Abuter R., Schreiber J., Eisenhauer F., Ott T., Horrobin M., Gillesen S., 2006, New Astronomy Review, 50, 398
  • Bate (2002) Bate M. R., 2002, in Shara M. M., ed., Stellar Collisions, Mergers and their Consequences Vol. 263 of Astronomical Society of the Pacific Conference Series, pp 111+
  • Bonnet et al. (2004) Bonnet H., Abuter R., Baker A., Bornemann W., et al., 2004, The Messenger, 117, 17
  • Bosch et al. (2009) Bosch G., Terlevich E., Terlevich R., 2009, AJ, 137, 3437
  • Brandl et al. (2007) Brandl B. R., Portegies Zwart S. F., Moffat A. F. J., Chernoff D. F., 2007, in St.-Louis N., Moffat A. F. J., eds, Massive Stars in Interactive Binaries Vol. 367 of Astronomical Society of the Pacific Conference Series, pp 629+
  • Breysacher (1981) Breysacher J., 1981, A&AS, 43, 203
  • Breysacher et al. (1999) Breysacher J., Azzopardi M., Testor G., 1999, A&AS, 137, 117
  • Campbell et al. (1992) Campbell B., Hunter D. A., Holtzman J. A., Lauer T. R., Shayer E. J., Code A., Faber S. M., Groth E. J., Light R. M., Lynds R., O’Neil Jr. E. J., Westphal J. A., 1992, AJ, 104, 1721
  • Cassinelli et al. (1981) Cassinelli J. P., Mathis J. S., Savage B. D., 1981, Sci, 212, 1497
  • Crowther & Bohannan (1997) Crowther P. A., Bohannan B., 1997, A&A, 317, 532
  • Crowther & Dessart (1998) Crowther P. A., Dessart L., 1998, MNRAS, 296, 622
  • de Koter et al. (1997) de Koter A., Heap S. R., Hubeny I., 1997, ApJ, 477, 792
  • Eisenhauer et al. (2003) Eisenhauer F., Abuter R., Bickert K., Biancat-Marchet F., et al., 2003, in Iye M., Moorwood A. F. M., eds., Proc. SPIE, 4841, 1548
  • Feast et al. (1960) Feast M. W., Thackeray A. D., Wesselink A. J., 1960, MNRAS, 121, 337
  • Feitzinger et al. (1980) Feitzinger J. V., Schlosser W., Schmidt-Kaler T., Winkler C., 1980, A&A, 84, 50
  • Guerrero & Chu (2008) Guerrero M. A., Chu Y.-H., 2008, ApJS, 177, 216
  • Hillier (1989) Hillier D. J., 1989, ApJ, 347, 392
  • Kim et al. (1998) Kim S., Staveley-Smith L., Dopita M. A., Freeman K. C., Sault R. J., Kesteven M. J., McConnell D., 1998, ApJ, 503, 674
  • Lamontagne et al. (1996) Lamontagne R., Moffat A. F. J., Drissen L., Robert C., Matthews J. M., 1996, AJ, 112, 2227
  • Massey & Hunter (1998) Massey P., Hunter D. A., 1998, ApJ, 493, 180
  • Massey et al. (2002) Massey P., Penny L. R., Vukovich J., 2002, ApJ, 565, 982
  • Moffat (1989) Moffat A. F. J., 1989, ApJ, 347, 373
  • Moffat et al. (2002) Moffat A. F. J., Corcoran M. F., Stevens I. R., Skalkowski G., Marchenko S. V., Mücke A., Ptak A., Koribalski B. S., Brenneman L., Mushotzky R., Pittard J. M., Pollock A. M. T., Brandner W., 2002, ApJ, 573, 191
  • Moffat & Niemela (1984) Moffat A. F. J., Niemela V. S., 1984, ApJ, 284, 631
  • Moffat et al. (2004) Moffat A. F. J., Poitras V., Marchenko S. V., Shara M. M., Zurek D. R., Bergeron E., Antokhina E. A., 2004, AJ, 128, 2854
  • Moffat & Seggewiss (1983) Moffat A. F. J., Seggewiss W., 1983, A&A, 125, 83
  • Moffat et al. (1985) Moffat A. F. J., Seggewiss W., Shara M. M., 1985, ApJ, 295, 109
  • Portegies Zwart et al. (2002) Portegies Zwart S. F., Pooley D., Lewin W. H. G., 2002, ApJ, 574, 762
  • Schnurr et al. (2008a) Schnurr O., Casoli J., Chené A.-N., Moffat A. F. J., St-Louis N., 2008a, MNRAS, 389, L38
  • Schnurr et al. (2008b) Schnurr O., Moffat A. F. J., St-Louis N., Morrell N. I., Guerrero M. A., 2008b, MNRAS, 389, 806
  • Schnurr et al. (2009) Schnurr O., Moffat A. F. J., Villar-Sbaffi A., St-Louis N., Morrell N. I., 2009, ArXiv e-prints
  • Schweickhardt et al. (1999) Schweickhardt J., Schmutz W., Stahl O., Szeifert T., Wolf B., 1999, A&A, 347, 127
  • Stetson (1992) Stetson P. B., 1992, in Worrall D. M., Biemesderfer C., Barnes J., eds, Astronomical Data Analysis Software and Systems I Vol. 25 of Astronomical Society of the Pacific Conference Series, pp 297+
  • Stetson (1994) Stetson P. B., 1994, PASP, 106, 250
  • Townsley et al. (2006) Townsley L. K., Broos P. S., Feigelson E. D., Garmire G. P., Getman K. V., 2006, AJ, 131, 2164
  • Usov (1992) Usov V. V., 1992, ApJ, 389, 635
  • Walborn (1986) Walborn N. R., 1986, in de Loore C. W. H., Willis A. J., Laskarides P., eds, Luminous Stars and Associations in Galaxies Vol. 116 of IAU Symp. Proc., pp 185–196
  • Weidner & Kroupa (2006) Weidner C., Kroupa P., 2006, MNRAS, 365, 1333
  • Weigelt & Baier (1985) Weigelt G., Baier G., 1985, A&A, 150, L18
Comments 0
Request Comment
You are adding the first comment!
How to quickly get a good reply:
  • Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
  • Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
  • Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
The feedback must be of minimum 40 characters and the title a minimum of 5 characters
Add comment
Loading ...
This is a comment super asjknd jkasnjk adsnkj
The feedback must be of minumum 40 characters
The feedback must be of minumum 40 characters

You are asking your first question!
How to quickly get a good answer:
  • Keep your question short and to the point
  • Check for grammar or spelling errors.
  • Phrase it like a question
Test description