Astronomical Science

The VLT FLAMES Tarantula Survey

Chris Evans1 7 School of Physics, University of Exeter, high quality survey of extragalactic William Taylor2 United Kingdom massive stars ever assembled, and Hugues Sana3 8 ESO is already providing exciting new Vincent Hénault-Brunet2 9 Harvard–Smithsonian Center for Astro­ insights into their evolution, multiplicity Tullio Bagnoli3 physics, Cambridge, USA and formation. Nate Bastian4 10 Department of Astronomy, University of Joachim Bestenlehner5 Vienna, Austria Alceste Bonanos6 11 Argelander-Institut für Astronomie der Massive stars and their descendants Eli Bressert7, 8, 9 Universität Bonn, Germany dominate the dynamics and chemical Ines Brott10 12 Department of Physics and Astronomy, enrichment of young star-forming gal­ Michael Campbell2 The Open University, Milton Keynes, axies, through their intense winds, radia­ Matteo Cantiello11 United Kingdom tion fields and dramatic deaths as core- Giovanni Carraro8 13 Departamento de Astronomia, Univer­ collapse supernovae. An overarching Simon Clark12 sidad de Chile, Santiago, Chile goal for studies of stellar evolution and Edgardo Costa13 14 Department. of Physics & Astronomy, resolved stellar populations is to develop Paul Crowther14 University of Sheffield, United Kingdom realistic tools to analyse integrated-light Alex de Koter3, 15 15 University of Utrecht, the Netherlands observations of distant star clusters Selma de Mink16, 17 16 Space Telescope Science Institute, and galaxies; if we can understand the Emile Doran14 Baltimore, USA properties and behaviour of the stars Philip Dufton18 17 Hubble Fellow on our own doorstep, we can be more Paul Dunstall18 18 Department of Physics & Astronomy, confident of an accurate interpretation of Miriam Garcia19, 20 Queen’s University Belfast, Northern unresolved populations far away. Mark Gieles21 Ireland, United Kingdom Götz Gräfener5 19 Instituto de Astrofísica de Canarias, Artemio Herrero19, 20 La Laguna, Tenerife, Spain Background and motivations Ian Howarth22 20 Departamento de Astrofísica, Universi­ Rob Izzard11 dad de La Laguna, Tenerife, Spain A vibrant area of research has been the Karen Köhler11 21 Insitiute of Astronomy, University of role of environment (metallicity) on the Norbert Langer11 Cambridge, United Kingdom evolution of massive stars, often focus­ Daniel Lennon23 22 Department of Physics & Astronomy, sing on the properties of individual stars Jesús Maíz Apellániz24 University College London, United in the nearby metal-poor Magellanic Nevena Markova25 ­Kingdom Clouds compared to their Galactic cous­ Paco Najarro26 23 European Space Agency, Space Tele­ ins. These efforts culminated in a previ­ Joachim Puls27 scope Science Institute, Baltimore, ous Large Programme, the VLT FLAMES Oscar Ramirez3 USA Survey of Massive Stars (FSMS), which Carolina Sabín-Sanjulián19, 20 24 Instituto de Astrofísica de Andalucía- analysed tens of O-type stars and hun­ Sergio Simón-Díaz19, 20 CSIC, Granada, Spain dreds of early B-type stars, from ob­ Stephen Smartt18 25 Institute of Astronomy with NAO, servations centred on open clusters in Vanessa Stroud12, 28 ­Smoljan, Bulgaria the Galaxy and in the Clouds (Evans et Jacco van Loon29 26 Centro de Astrobiología, CSIC-INTA, al., 2008). Jorick S. Vink5 Madrid, Spain Nolan Walborn16 27 Universitäts-Sternwarte München, A major result from the previous survey ­Germany was empirical evidence for weaker winds 28 Faulkes Telescope Project, University of in massive O-type stars at lower metal­ 1 United Kingdom Astronomy Technol­ Glamorgan, Wales, United Kingdom licities, in agreement with theoretical pre­ ogy Centre, STFC, Royal Observatory, 29 School of Physical & Geographical Sci­ dictions. This was a crucial test because ­Edinburgh, United Kingdom ences, Keele University, United Kingdom evolutionary models employ theoretical 2 Scottish Universities Physics Alliance, scaling laws to account for the loss of Institute for Astronomy, University of mass and angular momentum in metal- Edinburgh, United Kingdom We introduce the VLT FLAMES Taran- poor stars. In lower metallicity systems 3 Astronomical Institute Anton tula Survey, an ESO Large Programme stars therefore require a larger initial mass ­Pannekoek, University of Amsterdam, from which we have obtained optical to progress to the Wolf–Rayet (W–R) the Netherlands spectroscopy of over 800 massive stars phase than in the Galaxy, and will lose 4 Excellence Cluster Universe, Garching, in the spectacular 30 Doradus region less angular momentum over their life­ Germany of the Large Magellanic Cloud. A key times. Consequences of this include 5 Armagh Observatory, Northern Ireland, feature is the use of multi-epoch obser- reduced feedback of material from their United Kingdom vations to provide strong constraints winds and potentially different types of 6 National Observatory of Athens, Greece on the binary fraction. This is the largest core-collapse explosions.

The Messenger 145 – September 2011 33 Astronomical Science Evans C. et al., The VLT FLAMES Tarantula Survey

The O-type stars span a wide range of A thriving stellar nursery to the Giraffe spectrograph. Three of mass (upwards of ~ 20 MA), effective the standard Giraffe settings were used: temperatures (~ 30 000 to 50 000 K) and The 30 Doradus nebula is more than just LR02 (3960–4564Å, R = 7000), LR03 wind properties, with a diverse range a beautiful extragalactic H ii region — it (4499–5071Å, R = 8500), HR15N (6442– of morphological sub-groups. To under­ harbours one of the richest populations 6817Å, R = 16 000). This provided inter­ stand the most massive stars as a popu­ of massive stars in the local Universe and mediate resolution spectroscopy of the lation, including the evolutionary con­ is our only opportunity to resolve the blue optical lines commonly used in nections between those sub-groups, a components of a young, small-scale star­ ­classification and quantitative analysis of much larger, homogeneous sample burst. The nebula spans 15 arcminutes massive stars, combined with higher of high quality data was required. In par­ on the sky, equivalent to over 200 par­ ­resolution observations of the Hα line to ticular, rotationally-induced mixing is secs (pc) and comparable in scale to provide a diagnostic of the stellar wind thought to modify the surface chemistry regions of intense star formation observed intensity. These settings also include of massive stars on the main sequence, in high-redshift galaxies. The engine at important strong nebular lines (e.g., [O iii] with the most dramatic enhancements its core is the dense stellar cluster R136, and [S ii]), giving a useful tracer of the predicted for their nitrogen abundances. home to some of the most massive stars gas velocities along each line of sight. In

This is an important test of evolutionary known (with masses in excess of 150 MA; addition, a small number of selected models, where the benefits of large sam­ Crowther et al., 2010). When combined ­targets were observed at greater spectral ples are highlighted by the puzzling with its well-constrained distance (by resolution with the fibre-feed to UVES results for some of the B-type stars from ­virtue of being in the LMC) and low fore­ (see Figure 1). the FSMS (e.g., Brott et al., 2011; and ref­ ground extinction, 30 Doradus provides erences therein). Moreover, the progres­ us with a unique laboratory in which to One of the primary motivations of the sion of increasing wind strength along compile a large spectroscopic sample of VFTS was to detect massive companions the evolutionary sequence of O–Of–Of/ massive stars. WN–WN types is not well known, yet these are the stars which truly dominate The VFTS observations primarily em­ Figure 1. The FLAMES ARGUS integral field unit in terms of mechanical feedback and ion­ ployed the Medusa mode of FLAMES, in images (and extracted sources) overlaid on an HST WFC3 F555W image of R136. The location of ising photons. which 132 fibres are positioned within a Medusa and UVES targets in the central region are 25-arcminute field of view to feed targets also shown. To add to these factors, the catalyst for a new and ground-breaking survey was provided by recent results highlighting ARGUS the prevalence of binarity in massive stars MEDUSA (e.g., Sana & Evans, 2011). Parts of the UVES extragalactic community have begun to explore the consequences of this, with massive binaries argued to account for some of the integrated-light properties of low-redshift star-forming galaxies. How­ ever, an empirical description of the bi­ N nary fraction of massive systems (and their distribution of mass ratios) has, to date, been poorly constrained in models of star formation and cluster evolution.

The VLT FLAMES Tarantula Survey (VFTS; Evans et al., 2011) was conceived E to address these outstanding issues relating to the evolution, multiplicity and eventual fate of massive stars. By using the multi-object capability of FLAMES, 10ǥ combined with the light-gathering power of the VLT, we targeted the massive- star population of the Tarantula Nebula (30 Doradus, NGC 2070), the largest star- forming complex in the Local Group of galaxies, located in the Large Magellanic Cloud (LMC).

34 The Messenger 145 – September 2011 the LMC. We now summarise three dis­ –68°56' coveries which highlight the huge poten­ tial of the VFTS data.

Discovery of a very massive star outside –69°00' R136

There have been numerous narrowband imaging surveys in the Clouds to look for W–R stars and it is often presumed that we have a complete census of them. 04' However, the VFTS has discovered a new hydrogen-rich W–R star, VFTS 682, ) located 29 pc northeast of R136 (Fig­ 000

2 ure 3; Evans et al., 2011). The spectrum (J

c of VFTS 682, classified as WN5h, resem­

De bles those of the very luminous stars in 08' the core of R136 (Crowther et al., 2010). No radial velocity shifts are seen between the individual epochs, suggesting that the star is single to a high level of confidence.

From analysis of the spectra and availa­ 12' ble optical/infrared photometry the star appears to be very luminous, with log(L/

LA) = 6.5 ± 0.2 and with a current mass of ~ 150 MA (Bestenlehner et al., 2011). It is optically faint (V ~ 16) due to a line- 16' of-sight visual extinction of ~ 4.5 magni­ tudes, explaining why it had not been 40m49m 38m5h37m ­discovered by past narrowband imaging. RA (J2000) To date, such massive stars have only been found in the cores of dense clusters Figure 2. Combined YJKs-band image of 30 Doradus part of the VFTS was to investigate if R136 (e.g., R136 and the Arches), whereas from the VISTA Magellanic Clouds (VMC) Survey is dynamically stable (i.e. in virial equilib­ VFTS 682 appears to have either formed (Cioni et al., 2011). The O-type (red circles) and W–R stars (blue diamonds) observed by the VFTS are rium) via a determination of its velocity in relative isolation (asking important overlaid. dispersion. Rejecting detected binaries questions of theories of high-mass star (which would otherwise inflate the result), formation) or was ejected from R136, preliminary results indicate a stellar veloc­ which would pose an exciting challenge to if they are present. This search for “bina­ ity dispersion of less than 7.5 km/s, sug­ models of cluster dynamics. Tanta­li­singly, ries” (both true binaries and multiple gesting that the cluster is stable (Hénault- its high effective temperature (52 200 ± ­systems) shaped the observational strat­ Brunet et al., in preparation). 2500 K) might be as a result of chemi­ egy, which employed repeat observations cally-homogeneous evolution, a process at the LR02 setting to monitor for radial In total, the VFTS has observed over 300 suggested as an evolutionary channel for velocity variations. The majority of the O-type stars and 20 W–R (and Of/WN) long-duration gamma-ray bursts. data were obtained over the period Octo­ emission-line stars (see Figure 2) — a ber 2008 to February 2009, with con­ unique extragalactic sample in the high- straints to ensure reasonable gaps (of 28 mass domain of the Hertzsprung–Russell R139 revealed as an eccentric binary days or more) for the repeat LR02 obser­ diagram. The VFTS has also observed vations. A final epoch of LR02 data was over 500 B-type stars, which span the full One of the most luminous objects in obtained in October 2009, which helps luminosity range of main sequence 30 Doradus is R139 (VFTS 527; V ~ 12). significantly with the detection of both dwarfs to bright supergiants. This gives a Located one arcminute to the north of intermediate- and long-period binaries. sample comparable in size to the B-type R136 (Figure 3), it has been suggested in stars from the previous survey, but now the past as the most massive star outside A similar multi-epoch strategy was used with the benefit of multi-epoch data to R136. In a conference proceedings from to observe five pointings in the central investigate multiplicity. Lastly, there are 2002, Virpi Niemela reminded us of an region of R136 with the ARGUS integral 90 cooler-type stars with radial velocities old adage that “the brightest star in each field unit (Figure 1). The objective of this consistent with them being members of open cluster is (at least) a binary”

The Messenger 145 – September 2011 35 Astronomical Science Evans C. et al., The VLT FLAMES Tarantula Survey

Figure 3. The central part of the Tarantula Nebula showing the locations of VFTS 682 and R139 with respect to the massive cluster R136. The image is a combination of the YJKs images from the VMC ­Survey with V- and R-band images from the MPG/ ESO 2.2-metre telescope.

–69°00'00"

(­Niemela, 2002). Her words ring true in 30 Doradus, as the VFTS data have revealed R139 as a double-lined binary comprising two luminous supergiants (Figure 4; Taylor et al., 2011), classified 30" R139 as O6.5 Iafc and O6 Iaf for the primary VFTS682 and secondary, respectively. )

00 This discovery was only possible due to 20

(J the high quality multi-epoch spectra c from the VFTS. While the data are an ex­­ De cellent resource to detect binarity, char­ 01'00" acterisation of orbital parameters will necessitate additional spectroscopy (in the majority of cases) to determine peri­ ods and amplitudes. From follow-up observations of R139 with the MPG/ESO R136 2.2-metre telescope, Magellan and the 30" VLT, its orbit appears to be highly eccen­ tric with a period of 153.9 days. This gives lower mass limits for the compo­ 3 nents of M1 sin i = 78 ± 8 MA and M2 3 sin i = 66 ± 7 MA, making R139 one of the most massive binary systems known, and the most massive containing two 02'00" O-type supergiants (i.e., both stars are 44s40s 36s 32s 5h39m28s RA (J2000) relatively evolved).

1.4 Figure 4. Illustrative FLAMES Giraffe spectra of R139 at minimum separation (lower spectrum) and R139 with well-separated components; each is labelled 06.5 laf + 06 laf 54810.216 with the modified Julian Date (–2 400 000) of the 1.3 exposure.

1.2 54837.134 x flu

1.1

54868.056 1.0 Normalised

0.9 42 0.8 71 45 44

4485 4505

I II

IV IV He S He S 0.7 4460 4480 4500 4520 4540 4560 4580 Wavelength/Å

36 The Messenger 145 – September 2011 Mass estimates from comparisons with significant differential velocity when com­ used to explore the diverse populations in evolutionary models are in good agree­ pared to nearby stars and gas suggest the region (Walborn et al., in preparation). ment with the lower limits from the orbital that it is a runaway star, ejected from its analysis. This suggests a large inclination formation site by dynamical interactions In the VFTS sample of 352 O-type stars angle for the system, such that we might in a cluster or by a kick from a supernova (which includes some B0-type spectra), expect photometric eclipses. As part of explosion in a binary system. Spectral there are 125 with indications of binarity, a monitoring programme, we obtained analysis indicated a large evolutionary giving a lower limit to the binary fraction

54 V-band images with the Faulkes Tele­ mass of ~ 90 MA, suggesting interactions of 36%. With informed assumptions scope South over an 18-month period, with the even more massive stars in about the intrinsic distribution of periods, but found no evidence of photometric R136. Considering R136 is thought to eccentricities and mass ratios, the prob­ variations. An intensive photometric study be too young to have undergone a super­ ability of detecting a spectroscopic binary near to the predicted periastron would nova explosion, if VFTS 016 originated (of a given period) from the VFTS ob­ determine conclusively if there is an from there it would be one of the clearest servations can be calculated. Once cor­ eclipse, giving valuable constraints on the cases to date for this ejection mecha­ rected for completeness, the spectro­ inclination of the system. nism. Further investigation of the three- scopic binary fraction rises to between dimensional dynamics of VFTS 016 and 47 and 55% depending on the input other candidate runaways identified by period distribution (Sana et al., in prepa­ Identification of a massive runaway the survey requires high quality proper ration). This fraction is roughly compa­ motions; a recently approved imaging rable to results in young Galactic clusters A massive O2-type star, VFTS 016, on programme with the Hubble Space Tele­ (e.g., Sana & Evans, 2011). Compared the western fringes of 30 Doradus pro­ scope (HST; PI: Lennon) will provide to previous surveys of giant H ii regions vides an excellent illustration of how the the last piece of the puzzle in this regard. (e.g., Carina Nebula, NGC 346, etc.) the multi-epoch VFTS data is also powerful VFTS sample represents an order-of- in the case of non-detections of binarity. magnitude increase in the number of First observed with the 2-degree Field Massive single stars are in the minority O-type stars in a single environment. The (2dF) instrument at the Anglo–Australian significant number of B-type spectra Telescope, its spectrum (classified as These three results serve to illustrate the from the VFTS have also been investi­ O2 iii-if*) had a radial velocity 85 km/s power of the multi-epoch approach. The gated for radial velocity variations. Prelim­ lower than the local systemic velocity. This full sample of O-type spectra from the inary results from cross-correlation analy­ could indicate a large-amplitude binary, survey has now undergone rigorous vari­ sis point to a binary fraction of greater but no variations were seen in any of the ability analysis, in which a target is con­ than 33% (prior to correction for detec­ FLAMES data, from which a massive sidered a spectroscopic binary if it dis­ tion sensitivities), providing good evi­ companion with a period of less than one plays statistically-significant radial velocity dence for a comparably large binary pop­ year was excluded at the 98% level (Evans variations (with an amplitude in excess ulation in the lower-mass B-type regime et al., 2010). of 15 km/s) or evidence of double-lined (Dunstall et al., in preparation). profiles. Absolute radial velocities have VFTS 016 is at a projected distance of been obtained from this analysis for the These preliminary values argue strongly 120 pc from R136, and 70 pc from apparently single stars which, when com­ that the majority of the most massive the less massive cluster NGC 2060. Its bined with spectral classifications, will be stars are in binary systems, and that the

Figure 5. Evolutionary onset Roche-lobe overflow detectable as binary by radial velocity variations phases of a binary 18 (initial masses of 20 and 15 MA) as a function of initial orbital period and 16 M1 = 20 Msun # M = 15 M age (de Mink et al., 2011). 14 4. Single main-sequence star 2 sun 3. 12 Main-sequenc ) s yr e 10 er star with naked helium star companion (M rg er Ongoingansf 8 Me 2. tr

Time 1. mass 6

4 Two detached main-sequence stars 2 0 12345 10 1001000 10000 Initial orbital period (days)

The Messenger 145 – September 2011 37 Astronomical Science Evans C. et al., The VLT FLAMES Tarantula Survey

effects of binary evolution have to be every massive star should be associated of these results, ­enabling tests of evolu­ explored thoroughly in the context of stel­ with a cluster of (lower-mass) stars. In tionary models which include the physical lar evolutionary tracks and population models of monolithic collapse, the stellar processes for both single stars and synthesis models. Indeed, the develop­ masses are set by the gas core from binary systems. ment of new evolutionary models includ­ which they form, and star formation will ing the effects of rotation and binarity is trace the gas such that some form in a The physical parameters for each star an integral part of the VFTS collaboration. highly clustered distribution, whereas oth­ will ultimately be combined to investigate Extensive grids of evolutionary models ers can form in relative isolation (mean- the radiative and mechanical feedback of individual massive stars, including the ing that their formation is not strongly beyond the confines of 30 Doradus, and effects of rotation, are being extended to influenced by the gravitational potential of will enable empirical tests of spectral include masses of up to 300 MA (Köhler other nearby stars). ­synthesis techniques used to analyse dis­ et al., in preparation). In parallel, rapid tant, unresolved clusters/massive H ii evolutionary codes for population synthe­ The spatial distribution of high-mass regions. The final piece of the picture will sis studies are being used to investigate stars can also be used to constrain come from a new HST spectroscopic the effects of binarity and stellar rotation, ­scenarios about how the stellar initial programme (PI: Crowther) to determine and how binaries can provide tests of mass function is sampled, with a broad the physical properties of the large num­ the uncertain physics of binary interac­ range of implications, from determina­ ber of massive stars in the dense central tions. For example, Figure 5 (de Mink et tions of the star formation rate and mass parsec of R136, which can only be al., 2011) shows the effects of binarity, of galaxies, to the chemical evolution observed from space or with adaptive depicting the variety of stages (including of the Universe. The question is whether optics. mergers and spun-up accretors) during the mass of the most massive star is which one or both stars are on the main set by the mass of the host cluster, or if Remarkably, longer wavelength observa­ sequence. Figure 5 also illustrates the the mass of each star is simply set by tions show that 30 Doradus is at the observational challenge posed by the stochastic sampling? The “isolated” O northern end of a large column of molec­ apparently single stars that may well have stars from the VFTS survey (Bressert et ular gas which extends south for over experienced an interaction with a com­ al., submitted) suggest the latter in at 2000 pc. With such a reservoir of poten­ panion in the past, e.g., they may be the least some situations. Only with proper tial star-forming material, the region product of mergers or companions left motions (HST and Gaia) and high resolu­ seems destined to become an even more after the explosion of the primary. tion mm/sub-mm observations (e.g., impressive complex over the next few ALMA) will the issue be settled for each million years. candidate. Candidates for isolated high-mass star formation References This is only the beginning… Bestenlehner J. et al. 2011, A&A, 530, L14 There were a number of O-type stars out­ Brott, I. et al. 2011, A&A, 530, A116 side of R136 (and the lower-mass cluster The core set of results from the survey Cioni, M.-R. et al. 2011, A&A, 527, A116 NGC 2060) which appear to be single are yet to come — considerable work is Crowther, P. et al. 2010, MNRAS, 408, 731 and have radial velocities consistent with now underway to analyse the W–R and de Mink, S., Langer, N. & Izzard, R. 2011, Bulletin Société Royale des Sciences de Liège, 80, 543 the local systemic value. Rejecting stars OB-type stars from the VFTS to obtain Evans, C. et al. 2008, The Messenger, 131, 25 which might have an associated cluster full physical parameters (i.e., effective Evans, C. et al. 2010, ApJ, 715, L74 (from inspection of deep HST and VLT temperatures, gravities, mass-loss rates, Evans, C. et al. 2011, A&A, 530, A108 imaging), 15 of these appear to be unre­ rotational velocities, chemical abun­ Niemela, V. 2002, Extragalactic Star Clusters, IAU Symp. 207, ASP, p202 lated to a cluster, leading to a similar dances, and more). Given the scale of the Sana, H. & Evans, C. 2011, Active OB stars, IAU question as that posed by VFTS 682, i.e. dataset and the complexity of the theo­ Symp. 272, arXiv:1009.4197 were these stars formed in situ or ejected retical model atmospheres we are using a Taylor, W. et al. 2011, A&A, 530, L10 from a cluster as runaways (in the plane combination of techniques, employing of the sky)? the genetic algorithm and the grid- searching methods developed in the The two main theories of star formation, course of the previous survey. The results competitive accretion and monolithic of these quantitative analyses will pro- ­collapse, predict different formation sites vide us with an unparalleled view of the for the most massive stars in a young wind parameters for the full range of ­ system. In the former, a cluster of low- O- and WN-type stars, and enable the mass stars provides a potential well to first study of the effects of rotation on bring surrounding gas into the cluster, surface abundances in the most massive allowing some stars to accrete mass and stars. Knowledge of binarity will add a to grow to become O-type stars — i.e. valuable parameter to the interpretation

38 The Messenger 145 – September 2011