An Astrophysical Laboratory: Understanding and Exploiting the Young Massive Cluster Westerlund 1
Simon Clark1 sient high-energy phenomena in the Uni- on the interplay of these parameters. At Ignacio Negueruela2 verse: supernovae, gamma-ray bursts a given stellar mass, do these factors Ben Ritchie1 and X-ray binaries. It is therefore some- combine to drive sufficient mass loss to Paco Najarro3 what unsettling that the mechanisms of permit a type Ibc rather than a type II SN? Norbert Langer4 their birth, the processes governing their Is the pre-SN core rotating rapidly Paul Crowther5 subsequent evolution on and beyond the enough to permit a gamma-ray burst? Liz Bartlett6 main sequence (MS) and the nature of Indeed, is the core massive and compact Danielle Fenech7 their deaths and ultimate fate — neutron enough to form a black hole rather than Carlos González-Fernández8 star or black hole — are shrouded in a neutron star, and, if the latter, is it pow- Simon Goodwin5 uncertainty. ered by the extraction of rotational or Marcus Lohr1 magnetic energy? Raman Prinja7 Nevertheless, our current understanding is sufficient to enable us to identify the physical agents driving massive stellar Observational tests 1 Department of Physics and Astronomy, evolution. A central theoretical tenet is The Open University, Milton Keynes, that the evolutionary pathway of a star How, then, may one quantify the roles United Kingdom is dependent on its initial mass, but, for played by these various evolutionary 2 Departamento de Física, Ingeniería the most massive stars, powerful stellar agents? Ideally, one would like to study de Sistemas y Teoría de la Señal, winds and, in their latter evolutionary a homogeneous stellar population where Universidad de Alicante, Spain stages, violent instabilities act to continu- the competing physical effects might 3 Departamento de Astrofísica, Centro de ously strip their hydrogen-rich outer be distinguished. An obvious solution is Astrobiología, (CSIC-INTA), Madrid, layers away. The efficiency of such pro- to employ massive stellar clusters or Spain cesses is likely to contribute to the associations as laboratories, since both 4 Argelander Institut für Astronomie, dichotomy between type II (H-rich) and the age and metallicity of the stars may Bonn, Germany type Ibc (H-poor/depleted) supernovae be expected to be well constrained and 5 Department of Physics and Astronomy, (SNe). Since stellar winds are driven by furthermore they furnish a statistically University of Sheffield, United Kingdom radiation pressure, this in turn introduces robust sample size. 6 Astrophysics, Cosmology and Gravity an additional explicit metallicity depend- 5 Centre (ACGC), Astronomy Department, ence into massive stellar evolution. With a mass likely > 10 MA, the University of Cape Town, South Africa 30 Doradus star-forming complex (R136) 7 Department of Physics & Astronomy, Moreover, the internal dynamics of such within the Large Magellanic Cloud is a University College London, United King- stars, and hence the degree of mixing compelling target (Figure 1, bottom), more dom of chemically processed and unpro- so because the low foreground extinction 8 Institute of Astronomy, University of cessed material they experience, will also permits observations of traditional “blue- Cambridge, United Kingdom play a profound role in their evolution. end” (400–500 nm) spectral diagnostics Since it is likely that both the (differential?) for massive stars. These factors drove rotation of such stars and the presence the development and implementation of Westerlund 1 provides a unique oppor- (or absence) of an internal magnetic field the VLT/FLAMES Tarantula survey (Evans tunity to probe the physics of massive will play key roles in determining the effi- et al., 2011), which encompasses ~ 800 O stars, from birth to death and beyond, ciency of such processes, these too must and B stars and, to date, has resulted as well as the formation and evolution be incorporated into theoretical models. in about 20 refereed publications on the of a super star cluster that appears properties and evolution of its members. destined to evolve into a globular clus- Finally, it has recently been recognised Given this undoubted success, what is ter. We highlight the result of current that a large proportion of massive stars the motivation for additional studies? studies of this cluster, its diverse stellar are born in binary systems, and that in constituents and immediate environ- many cases these are close enough to We may identify a number of factors. ment, concluding with a summary permit interactions on or beyond the Most importantly one would wish to of future research avenues enabled by MS (e.g., Sana et al. 2013; de Mink et al., study differing metallicity environments in ESO facilities. 2014). Since this can lead to the removal order to determine the effect of chemical of the outer layers of a mass donor, the composition on stellar evolution. Sec- rejuvenation/growth of the mass recipient ondly, star formation across the ~ 200 pc Massive stars play a central role in the and, in extreme cases, their merger, it extent of 30 Dor appears to have pro- evolution of galaxies, from the early is apparent that both the binary fraction ceeded in multiple bursts over ~10 Myr Universe through to the current epoch, as well as the physical properties (e.g., (Walborn & Blades, 1997). Therefore, via their intense radiation fields and the mass ratio and orbital separation) also identifying the co-eval populations ideally deposition of chemically enriched gas profoundly impact stellar evolution. required to most effectively investigate and solid-state material into the interstellar stellar evolution is non-trivial, especially medium. Moreover they are ultimately One can thus readily envisage how the since the most likely locations to host responsible for the most energetic, tran- endpoints of the stellar lifecycle depend them — dense individual clusters such as
30 The Messenger 159 – March 2015 Figure 1. Top: Optical image of West- R136 — are difficult to probe due to erlund 1 (~ 7.3 pc on a side) taken with crowding. Despite this extended star for- the Wide Field Imager (WFI) on the MPG/ESO 2.2-metre telescope, with mation history, 30 Dor appears to lack BVR filters, showing the position of the cool hypergiants which help drive the both the magnetar and putative binary extensive mass loss required to facilitate companion Wd1-5. the formation of H-depleted Wolf–Rayet Middle: Wide-field optical image (~ 175 pc on a side) of the field centred stars (WRs). Finally one might also wish on Wd1 (the fuzzy orange blob in the to explore whether the environment in centre). which the stars form affects their life Bottom: HST composite-colour image cycle, e.g., via the dynamical formation, of the star- forming complex 30 Dora- dus in the Large Magellanic Cloud modification and disruption of binaries. (size 185 pc by ~ 146 pc), from a com- bination of Hubble Space Telescope Historically, it was thought that the mas- Advanced Camera for Surveys/Wide sive stellar aggregates required to enable Field Channel (ACS/WFC) and Wide Field Camera 3 (WFC3/UVIS) images such studies were absent from the Gal- (i-band) and WFI ([O III] and Hα narrow axy. However the advent of near- and band filters). mid-infrared surveys revealed an ever- expanding population of heavily obscured young massive clusters, alleviating this apparent deficiency. Nevertheless, given the effort expended in uncovering this population, it is frustrating that possibly DSS the most massive cluster in the Galaxy appears to have been hidden in plain sight: in 1961 Bengt Westerlund simply characterised his discovery as a very young “heavily reddened” cluster, with Westerlund 1 (Wd1; Figure 1, top and middle) languishing in relative obscurity for the next forty years.
In retrospect the detection of an unprece- dented cohort of cool, short-lived yellow hypergiants (YHGs) and red supergiants (RSGs) in the earliest observations of Wd1 obviously required a substantial population of massive progenitors; a clear indication that the cluster was worthy of follow-up observations. Motivated instead by the unusual radio properties of Wd1, the discovery that it hosted unprece- dented numbers of both blue super- and hypergiants (BSG/BHGs) and WRs came as a complete surprise (Clark et al., 2005). Analysis of these populations implied a distance of ~ 5 kpc, a radius of < 2 pc (in comparison to the 200 pc extent of 30 Dor; Figure 1, bottom) and an apparent 5 integrated mass of ~10 MA, revealing that Wd1 was the first direct Galactic
NASA, ESA and Lennon D. (ESA/STScI) analogue of the super star clusters that characterise the young stellar population of starburst galaxies.
Intriguingly, the simultaneous presence of both WRs and RSGs was unexpected, suggesting that we either view Wd1 at a privileged point in its evolution, or that it comprises two or more stellar
The Messenger 159 – March 2015 31 Astronomical Science Clark S. et al., Understanding and Exploiting the Young Massive Cluster Wd1
Figure 2. Montage of representative red-end optical
"l spectra of evolved OB stars in Wd1 showing the smooth progression in both spectral type and lumi- 6.!AHM nosity class expected for a coeval stellar population (from Clark et al., in prep.). 6.((( of photometrically selected targets, S
RD but no such population was identified.
EE Entirely unexpectedly, Wd1 appears to
N 6. K@A have formed in splendid isolation. Given
TW "((( k both its coevality and the absence of 6!K@A additional, contaminating populations it thus serves as a “gold-standard” labora- tory for the analysis of massive stellar
-NQL@KHRDC 6A! K@ evolution. 'D( This, however was not the only surprise. 6! K@ Ostensibly designed to search for binary -( candidates, a multi-epoch radial velocity #(! (RV) survey of the OB star population of /@RBGDMRDQHDR Wd1 was undertaken with VLT/FLAMES (Ritchie et al., 2009). However the result- ant dataset also permitted a determina- 6@UDKDMFSGÄ tion of the cluster velocity dispersion, once the intrinsic RV variables, which populations. In order to investigate this beyond, appear to show an additional would artificially inflate this measurement, issue, follow-up observations of the level of hierarchy, whereby star clusters were removed from the sample. Such an evolved OB star population were made themselves form in larger complexes analysis of the pre-2010 dataset returns with the suite of Very Large Telescope embedded within the remains of their a velocity dispersion of < 4.6 km s–1 (VLT) instrumentation. These revealed a natal giant molecular clouds. Exemplars (Clark et al., 2014). This implies that Wd1 highly uniform population, with the spec- include 30 Doradus, which hosts a num- is sub-virial, which is difficult to explain tral type and luminosity class of stars ber of clusters distributed throughout its since, in conjunction with its radial extent, evolving continuously from O9 III to confines (Figure 1, bottom), and the G305 one would instead expect it to be dynam- B2.5 Ia (see Figure 2; Negueruela et al. complex, comprising a number of appar- ically relaxed and hence in virial equilib-  and Clark et al. in prep.). This ent proto-clusters embedded on the rium. One potential solution to this appar- behaviour is exactly that expected for a periphery of a wind-blown bubble driven ent paradox might be that Wd1 formed, simple homogeneous and co-eval stellar by the young massive clusters Danks 1 or is in the process of forming, via the population at an age of ~ 5 Myr, exactly and 2 (Clark & Porter, 2004; Davies et al., merger of a number of discrete sub- the time for which one might expect the 2012). A further, diffuse population of clumps. However no statistical evidence co-existence of WRs and RSGs. Critically, massive stars is distributed across both for such spatially distinct co-moving these observations would have detected regions, which taken together are indica- groups was found within the RV data. a population of either younger or older tive of ongoing star formation activity over supergiants if present within Wd1, the several Myr. Our new observations appear to raise absence of such stars implying that it more questions than answers. Why is formed in a near instantaneous starburst. However wide-field optical imaging of Wd1 currently in a sub-virial state? How Such a conclusion was independently Wd1 reveals no such ongoing star forma- was so much mass accumulated in such reached via observations at the other tion activity, nor satellite clusters (Fig- a small volume of space? What was the extreme of the mass function, with an ure 1, top and middle); a conclusion sup- nature of the physical agent that led to age spread of < 0.4 Myr suggested for ported by inspection of near- and mid- its apparently instantaneous formation in Wd1 via the properties of stars at the MS infrared data (e.g., the Two Micron All Sky an otherwise unassuming region of the turn-on (Kudryavtseva et al., 2012). Survey [2MASS] and the Spitzer Galactic Galaxy before quenching any further Legacy Infrared Mid Plane Survey activity? The pronounced differences [GLIMPSE] and Multi Band Imaging between Wd1 and other massive galactic How did Wd1 form? Photometer Galactic [MIPSGAL] surveys). stellar aggregates demand answers Mindful that it might be surrounded by to such questions, even before one con- The deduction of a nearly instantaneous a diffuse halo of isolated massive stars siders that it represents the outcome burst immediately begs the question of (either OB stars or RSGs; c.f., Negueruela of the dominant mode of star formation how Wd1 formed. Observations of other et al., 2011), we obtained AAOmega in starburst galaxies and, that, given its star-forming regions in the Galaxy, and multi-object spectroscopic observations sub-virial nature, we are witnessing the
32 The Messenger 159 – March 2015 formation of a proto-globular cluster in contrasting the homogeneous population cation of binarity, the latter group forms the local Universe. of OB giants and supergiants to the a natural extension of the sequence of remarkably diverse cohort of further- B0-2Ia supergiants as predicted by evo- evolved transitional and WR stars within lutionary theory. Conversely the properties Binary evolution in action Wd1 (Clark et al., 2005; Crowther et al., of the former are consistent with binary- 2006). Binary interaction in compact driven mass loss; indeed RV data for Despite the uncertainty in the mode of systems has the effect of prematurely both Wd1-13 and -44 reveal them to be its formation, the highly coeval nature removing the outer H-rich mantle of the short-period systems (Ritchie et al., 2010 of Wd1 makes it an ideal laboratory to primary and hence modifying evolution, and in prep.). study the effects of binarity on stellar by both preventing a subsequent tran evolution. Binary systems may be identi- sition through a cool hypergiant phase The WC9+O star binary Wd1-239 (Porb fied via a number of methodologies, and initiating the extreme mass-loss that ~ 5.05 days; Clark et al., 2011) is conse- from the RV survey described above and characterises WRs earlier than antici- quently of interest since its properties periodic (eclipsing or ellipsoidal) photo- pated. Subject to limits imposed by its are consistent with expectations for the metric modulation, to indirect diagnostics rapid spin-up, the secondary accretes pre-SN endpoint of such a pathway. such as hard, over-luminous X-ray emis- some of the material lost by the primary, Under this scenario, the early onset of sion and/or the presence of host dust, while some is lost to the system. This the WR phase leads to significant addi- both forming in the wind collision zones is well illustrated by Wd1-9, where the tional mass loss, a low pre-SN core of massive binaries (Clark et al., 2008). A central binary appears completely veiled mass and the production of a neutron synthesis of these disparate criteria has by a dense dusty circumstellar torus, star. However the current compact con- to date resulted in the identification of resulting in the rich emission-line spec- figuration of Wd1-239 accommodates an over 70 confirmed and candidate binaries trum and infrared excess that character- even more exciting scenario: the rapid within Wd1 (Ritchie et al., in prep.), with ises such supergiant B[e] stars (Figure 3; rotation induced in massive, tidally locked the Monte Carlo simulations required to Clark et al., 2013). binaries naturally leads to highly efficient return an unbiased binary fraction under- internal mixing and hence chemically way. The subset of early (Wd1-5, -13 and -44) homogeneous evolution. This halts the and late (Wd1-7, -33 and -42a) BHGs expansion of the primary and, as a con- The central role that binarity plays in apparently represents the immediate out- sequence, it remains within its Roche stellar evolution is clearly illustrated by come of both pathways. Showing no indi- lobe, preventing binary-driven mass loss.
(< DS -( : ER _ ' NE
TW (< Ek D( D( -( : ' ' ( "( 6C
DS ER ( "@KK
NE K W
DK Figure 3. Optical spectra of the % E kT K supergiant B[e] binary Wd1-9 and the DK
% 6C blue hypergiant Wd1-5, a post binary interaction product. A best-fit synthetic spectrum derived from a non-local
-NQL@KHRDC /@RBGDMRDQHDR thermal equilibrium model atmosphere for Wd1-5 is overplotted in red. Major 6C RXMSGDSHBRODBSQTL transitions are indicated; weaker emission features in the spectrum of Wd1-9 are from low excitation metals 6@UDKDMFSGÄ such as Fe II.
The Messenger 159 – March 2015 33 Astronomical Science Clark S. et al., Understanding and Exploiting the Young Massive Cluster Wd1
second interactive phase in which its outer layers are also ejected, revealing l the chemically processed core. A final phase of reverse mass transfer from this newborn WR star pollutes the primary, yielding carbon-enhanced chemistry, before the secondary is lost to the type Ibc SN in which the magnetar forms and the primary is unbound, which we observe today as Wd1-5. Encouragingly, @ A such a binary pathway naturally leads to rapid rotation in the magnetar progeni-
A @ tor, as well as physical conditions in ) which a significant magnetic field may be born, which are both pre-requisites of @ @S HN M current theories of magnetar formation. #DBKHM Future prospects @ In the preceding text we have highlighted some recent results from our ongoing efforts to fully exploit the opportunity pre- sented by Wd1, but where do we go from here? In the near future our immediate goals are to fully characterise the binary population in terms of frequency, sepa ration and period. Tailored quantitative analyses of individual systems will run in parallel to this; the first aim of this strand 1HFGS@RBDMRHNM) being the determination of the evolution- ary status of the cool BHG population — Figure 4. Optical image of Westerlund 1 (greyscale) i.e., are they the progenitors or descend- with both the 3 cm radio map (red contours; ants of the RSGs? Dougherty et al., 2010) and ALMA pointings (blue circles) overplotted. Our current and archival data will also permit us to identify massive blue strag- In contrast to the former scheme this Universe (gamma-ray bursts and super glers and other binary products within in turn leads to a high pre-SN core mass luminous SNe). Wd1 that have been predicted on theo- and resultant black hole (BH) formation. retical grounds (de Mink et al., 2014). This It is therefore reassuring that observa- is of importance for both the verification Clearly then, the evolutionary history of tional support for this paradigm has been of our putative magnetar formation sce- a given star — e.g., single versus binary provided by the discovery of a candidate nario and also because Schneider et al. and chemically homogeneous versus for the pre-SN binary companion of the (2014) suggest that such a route may inhomogeneous — plays a critical role magnetar — the runaway BHG Wd1-5 produce some of the most massive stars in the “choice” of post-SN relativistic (Figures 1 and 3; Clark et al., 2014). Tai- known — essentially suggesting that
remnant. With an initial mass > 35 MA lored spectral analysis reveals a carbon- massive star formation may be a two- (c.f., Ritchie et al., 2010), one would rich chemistry that can only have arisen stage process, and in the process modi- expect that any single star undergoing via binary evolution, a conclusion but- fying the primordial cluster initial mass a SN within Wd1 at this epoch would tressed by its over-luminosity for its spec- function. result in the formation of a BH. Hence, troscopic mass. We employed the cur- following the arguments above, the dis- rent properties of Wd1-5 to infer a pre-SN The advent of the K-band Multi Object covery of a young, highly magnetised history for the putative Wd1 magnetar Spectrograph (KMOS) enables the effi- neutron star within Wd1 — a magnetar progenitor system whereby the primary cient accumulation of spectra sampling
(Muno et al., 2006) — provides a power- in a compact (Porb < 8 days) 41 + 35 MA the chemical and rotational diagnostics ful motivation for validating these sce binary transfers its outer layers to the essential for the model-atmosphere narios, more so since the birth of such secondary. As a result it rapidly spins up analysis of the lower luminosity, less- magnetars has been implicated in some and becomes so massive that it evolves evolved stars; hitherto inaccessible due of the most energetic explosions in the more rapidly than the primary, initiating a to interstellar extinction. Novel approaches
34 The Messenger 159 – March 2015 Figure 5. Representative spectral energy distribution LL for the B0 Ia star ε Ori showing the submillimetre excess introduced by wind clumping (from Blomme et al., 2002). NCDK L SG (MEQ@QDC 1@CHN RLNN W trum, with its youth, mass and proximity %KT conspiring to create a unique environ-
%KTW ment for the study of massive star and star cluster formation and evolution. Complementing ongoing studies in the 2LNNSGVHMCLNCDK low-metallicity environment of 30 Dor, it naturally falls along a sequence of Galactic clusters with ages ranging from ƅLƅL ƅL LL BL BL 1–3 Myr (e.g., the Arches and NGC 3603) 6@UDKDMFSG through to 15–20 Myr (the RSG-domi- nated clusters at either end of the Galactic bar) allowing us to address the to understand the nature of these stars RSGs to be made via the dust content evolution of stars from ~10 to > 100 MA, may also be employed. For instance cur- of their winds. In combination these while providing fresh insights into the rent observations indicate that pulsational approaches will better constrain the rate recent star formation history of the Milky variability is ubiquitous across all stellar at which mass is lost throughout the Way. Moreover its integrated mass and populations within Wd1 (Clark et al., full evolutionary cycle of massive stars — stellar density allow us to probe the phys- 2010); theoretical modelling by Saio et al. critical input physics for theoretical mod- ics of the unresolved super star clusters (2013) suggests that the periodicities of elling. that characterise external starburst gal- such pulsations provide important infor- axies through cosmological time and, mation on the evolutionary state of the Very Large Telescope (VLTI) observations given its apparent sub-virial nature, the pulsators. A synthesis of such data would of the hypergiant population permit the birth throes of a future globular cluster. allow the accurate construction of an direct resolution of their photospheres enhanced Hertzsprung–Russell diagram and circumstellar environments. This is for the cluster, potentially allowing us to essential in order to quantify the effects References break temperature/luminosity degenera- of the harsh ultraviolet environment of Arroyo-Torres, B. et al. 2015, A&A, in press, cies and determine whether, for example, Wd1 on such distended stars — do they arXiv:1501.01560 a given BSG or B-/YHG is in a pre- or experience surface ablation and/or dust Blomme, R. et al. 2002, A&A, 382, 921 post-RSG phase. destruction, and are the extended CO Clark, J. & Porter, J. 2004, A&A, 427, 839 and water layers (called MOL-spheres; Clark, J. et al. 2005, A&A, 545, 949 Clark, J. et al. 2008, A&A, 477, 147 Moving beyond multi-object spectros- Arroyo-Torres et al., 2015) that charac Clark, J., Ritchie, B. & Negueruela, I. 2010, A&A, copy, the proximity of Wd1 permits the terise isolated RSGs precluded? Similar 516, A78 resolution and detection of individual observations will also allow constraints Clark, J. et al. 2011, A&A, 531, A28 stars at submillimetre and radio wave- to be placed on the geometry and prop- Clark, J., Ritchie, B. & Negueruela, I. 2013, A&A, 560, A11 lengths, a feat not yet replicable for erties of the circumstellar torus of Clark, J. et al. 2014, A&A, 565, A90 30 Dor. Figure 4 shows the coverage of Wd1-9 and hence on the physics of the Crowther, P. et al. 2006, MNRAS, 372, 1407 extant radio observations (Dougherty et common-envelope binary evolution. Davies, B. et al. 2012, MNRAS, 419, 1871 al., 2010) and approved Atacama Large Dougherty, S. et al. 2010, A&A, 511, A58 de Mink, S. et al. 2014, ApJ, 782, 7 Millimeter/submillimeter Array (ALMA) Binary interactions may also be probed Evans, C. et al. 2011, A&A, 530, A108 observations (PI: Fenech). At long wave- with the ~ 350 ks of archival Chandra and Kudryavtseva, N. et al. 2012, ApJ, 750, L44 lengths, the submillimetre and radio XMM-Newton observations that have Muno, M. et al. 2006, ApJ, 636, L41 observations will allow the degree of sub- accumulated via studies of the magnetar Negueruela, I., Clark, J. & Ritchie, B. 2010, A&A, 516, 78 structure (clumping) present in the winds (Bartlett et al., in prep.), a six-fold Negueruela, I. et al. 2011, A&A, 528, A59 of the OB star cohort to be quantified, increase in integration time compared to Ohm, S., Hinton, J. A. & White, R. 2013, MNRAS, and refinement of the mass-loss esti- previous studies (Clark et al., 2008). 434, 2289 mates, which are currently uncertain by Determining the origin of the high-energy Ritchie, B. et al. 2009, A&A, 507, 1585 Saio, H., Georgy, C. & Meynet, G. 2013, MNRAS, a factor of ~ 5 or more due to their emission of Wd1 is a particularly exciting 433, 1246 dependence on this parameter. Figure 5 prospect, given that it has recently been Sana, H. et al. 2013, A&A, 550, A107 shows as an example the effect of wind recognised to be both a GeV and TeV Schneider, F. et al. 2014, ApJ, 780, 117 clumping compared to the smooth wind source (Ohm et al., 2013). Walborn, N. & Blades, J. 1997, ApJS, 112, 457 model for a nearby B0 supergiant. Simi- larly, the ALMA observations will permit To summarise, Wd1 is an enticing target estimates of mass loss for the YHGs and across the whole electromagnetic spec-
The Messenger 159 – March 2015 35