A VLT/FLAMES survey for massive binaries in Westerlund 1: VII. Cluster census
aa r X i v : . [ a s t r o - ph . S R ] A ug Astronomy&Astrophysicsmanuscript no. census˙final c (cid:13)
ESO 2019August 16, 2019
A VLT/FLAMES survey for massive binaries in Westerlund 1: VII.Cluster census ⋆ J. S. Clark , B. W. Ritchie , , and I. Negueruela School of physical sciences, The Open University, Walton Hall, Milton Keynes, MK7 6AA, United Kingdom Lockheed Martin Integrated Systems, Building 1500, Langstone, Hampshire, PO9 1SA, UK Departamento de F´ısica Aplicada, Facultad de Ciencias, Universidad de Alicante, Carretera San Vicente s / n, E03690, San Vicentedel Raspeig, SpainPreprint online version: August 16, 2019 ABSTRACT
Context.
The formation, properties, and evolution of massive stars remain subject to considerable theoretical and observational un-certainty; impacting on fields as diverse as galactic feedback, the production of cosmic rays, and the nature of the progenitors of bothelectromagnetic and gravitational wave transients.
Aims.
The young massive clusters many such stars reside within provide a unique laboratory for addressing these issues, and in thiswork we provide a comprehensive stellar census of Westerlund 1, potentially the most massive Galactic example of such an aggregate,to underpin such e ff orts. Methods.
We employ optical spectroscopy of a large sample of early-type stars to determine cluster membership for photometrically-identified candidates, characterise their spectral type, and identify new candidate spectroscopic binaries.
Results.
69 new members of Westerlund 1 are identified via I − band spectroscopy. Together with previous observations, they illustratea smooth and continuous morphological sequence from late-O giant through to OB supergiant. Subsequently, the progression bifur-cates, with one branch yielding mid- to late-B hypergiants and cool super- / hypergiants, and the other massive blue stragglers, prior toa diverse population of H-depleted Wolf-Rayets. We identify a substantial population of O-type stars with very broad Paschen serieslines, a morphology that is directly comparable to known binaries in the cluster. In a few cases additional low-resolution R − bandspectroscopy is available, revealing double-lined He I profiles and confirming binarity for these objects; suggesting a correspondinglyhigh binary fraction amongst relatively unevolved cluster members. Conclusions.
Our current census remains incomplete, but indicates that Westerlund 1 contains at least 166 stars with initial massesestimated to lie between ∼ M ⊙ to ∼ M ⊙ , with more massive stars already lost to supernova. Our data is consistent with thecluster being co-eval, although binary interaction is clearly required to yield the observed stellar population, which is characterisedby a uniquely rich cohort of hypergiants ranging from spectral type O to F, with both mass-stripped primaries and rejuvenated secon-daries / stellar mergers present. Future observations of Wd1 and similar stellar aggregates hold out the prospect of characterising bothsingle- and binary- evolutionary channels for massive stars and determining their relative contributions. This in turn will permit thephysical properties of such objects at the point of core-collapse to be predicted; of direct relevance for understanding the formation ofrelativistic remnants such as the magnetars associated with Wd1 and other young massive clusters. Key words. stars: evolution - stars: supergiant - stars: Wolf Rayet
1. Introduction
Even before the detection of gravitational waves from coalescingneutron stars and black holes, the case for fully understandingthe evolutionary cycle of massive stars was compelling. They,and their explosive endpoints contribute to the secular evolutionof galaxies via radiative and mechanical feedback and the depo-sition of dust and the products of nuclear burning into the widerinterstellar medium, facilitating subsequent generations of starformation. Upon core-collapse they yield a rich variety of ener-getic electromagnetic transients, with their relativistic remnantssubsequently forming both X- and γ -ray binaries; phenomenathat have also been associated with the production of cosmicrays.Rich young stellar clusters and associations potentially serveas powerful laboratories for investigating the properties and life-cycles of massive stars, as illustrated by Massey et al. (2000, ⋆ This work is based on observations collected at the EuropeanSouthern Observatory, Paranal Observatory under programme IDs ESO081.D-0324, 383.D-0633, 087.D-0440 and 091.D-0179 ffi ciently categorised and, as a consequence, evo-lutionary models observationally tested. Moreover, observationsof multiple clusters at di ff erent ages means that we can recon-struct the evolutionary pathways of stars sampling a wide rangeof initial masses.The e ffi cacy of this methodology is evidenced by the inves-tigation of the 30 Dor star formation region initiated by Evanset al. (2011). However, as part of this multi-work campaign,Schneider et al. (2018) reveal a potential impediment to this ap-proach; specifically 30 Dor is found to have a complex, extendedformation history resulting in multiple spatially unsegregatedstellar populations. Such a situation complicates the interpre-tation of evolutionary sequences, which are better constrainedin simple, co-eval stellar populations. Moreover, at a distance of ∼ ffi cult to apply. J. S. Clark et al.: Stellar census of Wd1
Consequently, the identification of such stellar aggregateswithin the Galaxy would be of considerable value. As potentiallythe most massive open cluster within the Galaxy, Westerlund 1(henceforth Wd1; Westerlund 1961, 1987) is a compelling targetand, although heavily reddened, it remains accessible in the R − and I − bands. With an age of ∼ / hypergiants in addition to significant num-bers of OB supergiants, Wolf-Rayet (WR) stars and the magne-tar CXO J164710.2-455216 (Clark et al. 2005, Crowther et al.2006a, Muno et al. 2006a, Negueruela et al. 2010, Kudryavtsevaet al. 2012).Wd1 is spectacular in appearance across the electromagneticspectrum, with point source (stellar) and di ff use emission ob-served at: – X-ray energies due to emission from the magnetar, shocks inthe winds of single stars, the wind collision zones of massivebinaries, (unresolved) pre-main sequence low-mass stars andthe hot intracluster medium (Clark et al. 2008, 2019c, Munoet al. 2006b). Spatially extended, highly energetic γ -rays(GeV and TeV; Ohm et al. 2013 and Abramowski et al. 2012,respectively) have also been associated with Wd1 and as aconsequence it has also been implicated in the production ofcosmic rays. Both phenomena have been attributed to the (in-teractions between the) winds of massive single and binarystars, the cluster wind and supernovae (e.g. Aharonian et al.2018, Bednarek et al. 2014, Bykov et al. 2015, Cesarsky &Montmerle 1983). – IR emission is due to warm dust associated with interact-ing O-star and Wolf-Rayet binaries and the ejection neb-ulae of cool super- / hypergiants, which is subsequently en-trained within the intracluster medium (Clark et al. 1998,2013, Crowther et al. 2006a, Dougherty et al. 2010). – mm- and radio-continuum emission is attributed to stellarwinds from early-type and WR stars, circumstellar ejecta as-sociated with the cool stellar cohort, wind collision zonesin interacting binaries and the ionised intracluster medium(Dougherty et al. 2010, Fenech et al. 2017, 2018, Andrewset al. 2018).These phenomena are all ultimately driven by the underlyingmassive stellar population of the cluster in both life and death;hence if we are to understand them we are required to constructan accurate stellar census for Wd1. Since many are also man-ifestations of binary interaction, any such e ff orts should alsoserve to constrain the nature of this population, noting that multi-wavelength (X-ray, IR and radio) and multi-epoch photometricobservations imply a high binary fraction (Clark et al. 2019c,Crowther et al. 2006a, Dougherty et al. 2010 and Bonanos 2007,respectively). Such a rigorously determined population census isalso essential in order to constrain the bulk properties of the clus-ter such as age, mass, initial mass function, degree of mass seg-regation, dynamical state and mode of formation (e.g. Andersenet al. 2017, Brandner et al. 2008, Cottaar et al. 2012, Gennaroet al. 2011, 2017, Kudryavtseva et al. 2012, Lim et al. 2013,Negueruela et al. 2010).Consequently between 2008-2013 we undertook a largemulti-epoch I − band spectroscopic survey of massive starswithin Wd1 with the Fibre Large Array Multi ElementSpectrograph (FLAMES; Pasquani et al. 2002) mounted on theVery Large Telescope (VLT). These observations were designedto (i) constrain the binary population of Wd1 via radial veloc-ity (RV) monitoring, (ii) provide an expanded stellar census en-compassing the fainter cohort of massive evolved stars and (iii) supplement extant spectroscopic and photometric data in orderto permit quantitative model atmosphere analysis of individualcluster members. To date these observations have allowed usto investigate the incidence of binarity amongst the supergiants(Ritchie et al. 2009a, 2011), provide tailored analyses of a num-ber of interacting and post-interaction systems (Clark et al. 2011,2014b, 2019a, Ritchie et al. 2010) and relate the X-ray propertiesof cluster members to their underlying stellar properties (Clark etal. 2019c). In this work we present an enlarged census for Wd1,comprising 166 massive evolved stars, and discuss the implica-tions of this population for both single- and binary-star evolu-tionary channels, and the nature of the cluster - both in isolationand in comparison to other young massive stellar aggregates.
2. Observations & data reduction
As described above the original, and primary, science goal of ourprogramme was the identification and characterisation of mas-sive binaries within Wd1. In order to accomplish this multipleobservations of four target fields were made in service modeduring 2008 and 2009 using the FLAMES-GIRAFFE multi-fibre spectrograph with setup HR21 to cover the 8484-9001Årange with R = λ/ ∆ λ ∼ bright field,containing 22 spectroscopically- and photometrically-selectedtargets (detailed in Ritchie et al. 2009a) and three faint fieldscontaining 17 spectroscopically-confirmed cluster members and63 photometrically-selected candidates. With three exceptions -Wd1-57a, Wd-71 and WR K - the faint lists contain only starswith previously-known spectral types no later than B0 Iab, orphotometry consistent with lower-luminosity O9-B0 stars justevolving towards the supergiant phase (cf. Fig. 1). Integrationtimes were typically 2 × bright and 3 × faint configurations. Given the appearance of some targetsin more than one configuration - motivated by the desire to max-imise fibre allocation - and the di ff ering frequency with whicheach configuration was executed, individual stars were ordinar-ily observed between five and eleven times.Subsequently, additional observations were made in 2011and 2013 utilising the same configuration. Target selection wasoptimised to include (i) follow up of spectroscopically and / orphotometrically identified binary candidates and (ii) initial ob-servations of previously unclassified luminous candidate clustermembers and / or X-ray bright stars. Due to the observing runsremaining incomplete and the constraints imposed by fibre allo-cation, the final number of observations (and hence integrationtime) for individual stars varies considerably; a full observinglog will be provided in Clark et al. (in prep.) where we analysethe full RV dataset.The FLAMES-GIRAFFE pipeline and Common PipelineLibrary were used to bias subtract, flat field and wavelength cal-ibrate the data, while subsequent processing made use of IRAF for the extraction individual spectra, rectification and heliocen-tric velocity correction; full details of data reduction are given inRitchie et al. (2009a).Fortuitously, R − band spectra of a number of our targetswere acquired on the nights of 12 and 13 June 2004 usingVLT / FORS2 in longslit and mask exchange unit (MXU) modes,and are used to supplement the limited wavelength coverage ofthese targets provided by our single VLT / FLAMES setup. The IRAF is distributed by the National Optical AstronomyObservatories, which are operated by the Association of Universitiesfor Research in Astronomy, Inc., under cooperative agreement with theNational Science Foundation.. S. Clark et al.: Stellar census of Wd1 3
10 11 12 13 14 15 16 14 15 16 17 18 m I m R New cluster membersWesterlund (1987)
Fig. 1.
Comparison of R − and I − band photometric magni-tudes from Clark et al. (2005) for cluster members listed byWesterlund (1987; open squares) and new confirmed clustermembers listed in Table 2 (solid squares).G1200R grism was employed, yielding a nominal dispersion of0.38Å / pixel over the spectral range 5750-7310Å. A 0.3” slit wasutilised for the longslit mode to obtain a resolution of ∼ ∼ . ′′ ×
4k E2V CCD detectorand an AAO2 CCD controller. Given the very high reddening tothe targets, the blue arm did not produce useful spectra. The redarm used the 1700D grating, providing a resolution R ∼
11 000in the region surrounding the Ca ii triplet. The spectra cover a500Å wide range centred on 8700Å, but the projection of thespectrum from each fibre on the CCD depends on its position onthe plate, displacing the range limits up to ∼
3. New cluster members
We are able to identify a total of 69 new cluster members via ourspectroscopy. Foreshadowing Sect. 3.1-3.5, the resultant stellarcensus for Wd1 is summarised in Table 1, while the co-ordinates,photometry and spectral classifications for individual membersare presented in Table 2. Once ordered by increasing right as-cension, we apply a W xxx designation for these new sources .In addition to these stars, eight further objects were ob-served. Seven of these eight targets were found to be late-type We suggest that this naming convention supercede that employedfor the subset of these stars described in previous works (e.g. Ritchieet al. 2009a, Fenech et al. 2018). To enable cross-correlation, whereappropriate, we give the old nomenclature in Table 2.
Table 1.
Stellar demographics of Wd1
Spectral Cluster New toType population this workO9-9.5 III 27 25O9-9.5 II,II-III 11 11O9-9.5 Iab,Ib 31 11B0-0.5 Ia,Iab,Ib 23 7B1-1.5 Ia,Iab 10 2B2-4 Ia 7 0O4-8 Ia + + / WNVL 4 1B5-9 Ia + + RSG 10 0sgB[e] 1 0OB SB2 12 10OeBe 1 1WN5-8 14 0WC8-9 8 0Total 166 69Sub-panels in the table comprise, from the top: giants and supergiants,hypergiants, transitional stars, binaries of uncertain spectral type and / orluminosity class (e.g. Wd1-36, W1001, W1003, and the supergiant B[e](sgB[e]) star Wd1-9), OeBe stars and Wolf-Rayets. Transitional starscomprise luminous blue variables (LBVs), yellow hypergiants (YHGs),and red supergiants (RSGs). We include the O9-9.5 I-III stars W1033and -1040 in the O9-9.5 Iab,Ib category (Sect. 3.4); thus explainingthe slight discrepancy between this table and the version in Clark etal. (2019c). stars , while the final object is a B-type star of luminosity classV. The RVs measured for the majority of these indicate thatthey are foreground objects, but two objects, F7 and F8, haveRVs consistent with membership of Wd1 (Clark et al. 2014b, inprep.). However, the pronounced DIB at ∼ bona fide cluster members (cf. Sect. 3.1) appearsweak or absent for both stars, which are also spatial outliers tothe east and north of the cluster respectively. Given their appar-ent classification as mid-K giants we conclude that these are alsointerlopers. Co-added spectra of (apparently) non-binary objects are used asstandards for classification of the newly-identified stars in Wd1,with an appropriate I − band stellar sequence shown in Figs. 2 and3. Following guidelines to classifying O9–B9 stars, taken fromClark et al. (2005) and Negueruela et al. (2010), the followingdiagnostics are used: – The comparative weakness of the higher Paschen linesand the strength of C iii + C iii / Pa-15 ≫ Three of these stars - F1 (see Ritchie et al. 2009a), F2 and F4 -show a TiO 8860Å bandhead (Ramsey 1981) with a strength implyingthey are early-M giants. The remaining objects show spectra typical oflate-G or early-K stars, with strong Ca ii and Mg i N o r m a li z ed f l u x ( + o ff s e t ) Wavelength (Å) Paschen seriesDIBC III He I N I W36 (OB bin)W1015 (O9 III)W74 (O9.5 Iab)W373 (B0 Iab)W8b (B1.5 Ia)W71 (B2.5 Ia)C (2−0) 0.75 1 1.25 1.5 1.75 2 2.25 8500 8550 8600 8650 8700 8750 8800 8850 8900 N o r m a li z ed f l u x ( + o ff s e t ) Wavelength (Å) Paschen seriesDIBC III He I N I W36 (OB bin)W1015 (O9 III)W74 (O9.5 Iab)W373 (B0 Iab)W8b (B1.5 Ia)W71 (B2.5 Ia)C (2−0) 0.75 1 1.25 1.5 1.75 2 2.25 8500 8550 8600 8650 8700 8750 8800 8850 8900 N o r m a li z ed f l u x ( + o ff s e t ) Wavelength (Å) Paschen seriesDIBC III He I N I W36 (OB bin)W1015 (O9 III)W74 (O9.5 Iab)W373 (B0 Iab)W8b (B1.5 Ia)W71 (B2.5 Ia)C (2−0) 0.75 1 1.25 1.5 1.75 2 2.25 8500 8550 8600 8650 8700 8750 8800 8850 8900 N o r m a li z ed f l u x ( + o ff s e t ) Wavelength (Å) Paschen seriesDIBC III He I N I W36 (OB bin)W1015 (O9 III)W74 (O9.5 Iab)W373 (B0 Iab)W8b (B1.5 Ia)W71 (B2.5 Ia)C (2−0) 0.75 1 1.25 1.5 1.75 2 2.25 8500 8550 8600 8650 8700 8750 8800 8850 8900 N o r m a li z ed f l u x ( + o ff s e t ) Wavelength (Å) Paschen seriesDIBC III He I N I W36 (OB bin)W1015 (O9 III)W74 (O9.5 Iab)W373 (B0 Iab)W8b (B1.5 Ia)W71 (B2.5 Ia)C (2−0) 0.75 1 1.25 1.5 1.75 2 2.25 8500 8550 8600 8650 8700 8750 8800 8850 8900 N o r m a li z ed f l u x ( + o ff s e t ) Wavelength (Å) Paschen seriesDIBC III He I N I W36 (OB bin)W1015 (O9 III)W74 (O9.5 Iab)W373 (B0 Iab)W8b (B1.5 Ia)W71 (B2.5 Ia)C (2−0) Fig. 2.
Montage of I -band spectra of classification standards for apparently single stars of spectral type O9-B2.5. The spectrum ofthe 3.18-day eclipsing binary Wd1-36 is included to indicate the appearance of SB2 systems of comparable spectral type withinWd1; note the shallow, flat-bottomed Paschen series lines (see Sect. 3.5). – At B0, Pa-15 and Pa-16 lines are of approximately equalstrength due to the increasing contribution of C iii i i – B0.5-2 spectral types show a rapid increase in the strengthof both the Paschen series and the He i lines (Negueruelaet al. 2010), with all objects later than B0 showingPa-15 / Pa-16 >
1. A wealth of N i lines between ∼ − ∼ B2 and are clear by B2.5 – For spectral types B2.5 and later both the Paschen series andN i photospheric lines continue to increase in strength, whilethe He i lines drop out by ∼ B5. At the same point, compari-son of the strength of the emerging Ca ii ∼ B9 to be distinguished.With our limited spectral coverage and paucity of lines suit-able for classification, distinguishing between O9 and O9.5 ischallenging. At high S / N He i S / N of manyfaint targets, and C iii N i available classifier, while the width and profile of the wings ofPa-11 serve as a guide to the luminosity class. However, as dis-cussed in Negueruela et al. (2010), these features will also be af-fected by rotation and abundance anomalies, and classificationsgiven here are therefore uncertain to ∼ half a spectral type.A further potential di ffi culty in classifying O-type stars isintroduced by binarity. The brevity of (post-MS) evolutionaryphases means that a system containing a luminous supergiantprimary will appear single-lined unless the mass ratio is nearunity – such that both objects are in a comparable state – andwill therefore display a spectrum that is very similar to that ofa single star. However systems that consist of an ∼ O9–9.5 IIIprimary and massive OB secondary are more likely to displaycomposite spectra. Consideration of synthetic spectra suggeststhat the relatively broad Paschen series lines in a short-periodO9 III + O9 III system would remain heavily blended even atquadrature. We might therefore expect such systems to displaya spectrum with broad, flat-bottomed Paschen series lines andanomalously broad, weak C iii ffi cultto classify. An exampler of this phenomenon, albeit of greaterluminosity, is the 3.18-day eclipsing binary Wd1-36 (Bonanos2007); as a consequence of the strong line-blending we are onlyable to prove a generic OB Ia + OB Ia classification for thissystem. This also serves to illustrate the significant morpholog- . S. Clark et al.: Stellar census of Wd1 5
Fig. 3.
New I -band spectra of W1049 and W1069 plotted against a montage of cluster B0-B9 hypergiants in order to permit classi-fication (with the highly luminous B4 Ia star Wd1-57a included for completeness). The spectrum of W1069 has been degraded inresolution (red dashed line) to enable direct comparison to the spectra of B5-9 hypergiants.ical di ff erences between the spectra of such systems and thoseof stars following the empirical O9 III - B2.5 Ia evolutionary se-quence illustrated by Fig. 2; we discuss this and similar systemsin greater detail in Sect. 3.5.A number of interstellar features are also visible in our spec-tra. The strong, well-defined ∼ σ RV(8620) = . − and varia-tions in FWHM of less than 2%. A broad DIB at ∼ S / N spec-tra. A number of weaker interstellar features are also present,with the C Phillips (2–0) band system (e.g. van Dishoeck & deZeeuw 1984) prominent on the red wing of the Pa-12 line. The R lines largely overlap the core of Pa-12 and cannot be mea-sured with accuracy, but in very high S / N co-added spectra ofWd1-71 (total integration time ∼ Q lines are verystrong and may be distinguished to Q(18), while a strong R(0)line, moderate P(2) and P(8) lines and weak R(10), R(12) mayalso be distinguished, with P(4) and P(6) blended with the wingsof He i − . ± . − , very close to typical systemic velocitiesfor the stellar population of Wd1, suggesting that the absorbingmaterial is local to the cluster. The centre of the strong ∼ We turn first to W1049 and W1069; two new reddened hyper-giant candidates identified via our AAT observations (Figs. 3 &4). W1069 has only been observed in the I − band, but it appearsto be an excellent match to the B5 Ia + cluster members Wd1-7 and -33; the strength of the Ca ii lines precluding an earlier( ≤ B4; Wd1-57a) or later ( ∼ B9; Wd1-42a) spectral type. By con-trast the I − band spectrum of W1049 appears earlier, matchingthat of both the apparently single B1.5 Ia supergiant Wd1-8b(Fig. 2) and Wd1-5 (WN10-11h / B0.5 Ia + ; Crowther et al. 2006a,Negueruela et al. 2010).Unlike W1069 an R − band spectrum is available and, as withWd1-5, W1049 appears to di ff er from the normal supergiantswithin Wd1 (Fig. 4). Specifically both the strength and pro- J. S. Clark et al.: Stellar census of Wd1
Fig. 4. R -band spectrum of W1049 plotted with comparable spectra of highly luminous cluster super- / hypergiants of spectral typeB0-B9.file of the H α line is atypical for such stars, while weak emis-sion in the C ii ii + / WN10-11h stars Wd1-5 and -13, while the H α profile- comprising a strong, narrow central peak and broad wings - isalso reminiscent of these stars and, indeed, the B5 Ia + stars Wd1-7 and -33 (Fig. 4 and Ritchie et al. 2010). A similar H α profileis observed for Cyg OB2 + ) and ζ Sco (B1.5 Ia + );albeit with the C ii lines in absorption in both stars and a promi-nent P Cygni component characterising the latter star (such thatit resembles the much cooler Wd1-42a; Clark et al. 2012).The corresponence between W1049, Wd1-5, and -13 is notexact, with P Cygni profiles of the He i Nine of the newly classified cluster members present spectraconsistent with early-B supergiants. Only two stars appear tohave spectral types later than B0; W1039 and W1048 resem- ble Wd1-78 (B1 Ia; Neguerueula et al. 2010) and Wd1-8b (B1.5Ia; Fig. 2) respectively, and hence we adopt such classificationsfor them. The other seven objects - W1005, -1009, -1053, -1055,-1065, -1067, and -1068 - all appear to be B0 supergiants ofluminosity class Iab or Ib. Of these W1068 shows discrepantCa ii and Mg i lines that are most probably due to blending witha foreground late-type star. W1055 is of note since it presentsanomalously strong C iii α and C ii M init ∼ − M ⊙ - which wouldpresent as anomalously photometrically faint objects with broadbut sharp troughed Paschen series absorption profiles.However, a tenth early-type star, W1004, displays Pa-11. . . 16 lines in emission with double peaks characteristic of aclassical OeBe star. Unfortunately, contamination of the sparsediagnostic features by emission from the circumstellar disc pre- . S. Clark et al.: Stellar census of Wd1 7 N o r m a li z ed f l u x Wavelength (Å)
Wavelength (Å)
Fig. 5.
Comparison of the Pa11 line from W1045 (O9.5 II, leftpanel) and W1015 (O9 III, right panel) to synthetic model at-mosphere spectra for 30kK and 32kK respectively. Note the in-fluence of He i ′ ′′ to the south-west (3 . d / ∼ ∼ ∼ Finally we turn to the majority of the new cluster memberswhich, pre-empting the following discussion, we find to be O-type stars of luminosity class III–Ib. Given the limitations of thespectroscopic diagnostics present in the I − band we need to makeexplicit use of extant photometry and, where available, R − bandspectroscopy in order to obtain the most precise classificationsfor this cohort. Despite the possibility of a systematic o ff set inthe classifications we arrive at in comparison to those obtainedfrom more traditional methodologies (e.g. employing blue-endspectral diagnostics) we are confident our technique robustlyclassifies such stars relative to other cluster members. Indeed itreinforces a central empirical finding of our survey; that there isa smooth and continuous progression in spectral morphologiesfrom the earliest O giants through to the mid-B supergiants andindeed the mid- to late-B hypergiants (cf. Figs. 2 and 3).Nevertheless, while further spectroscopic and photometricobservations of the newly-identified cluster members are re-quired to fully categorize this population, in general the newcluster members fall into four distinct groups: – Stars with spectra and photometry comparable to Wd1-1 and-74 (O9.5 Iab; Negueruela et al. 2010). As a result these aregiven an identical classification. – Stars with spectra similar to Wd1-74, but with slightlybroader Paschen series lines and (when photometry is avail-able) I − band magnitudes roughly a magnitude fainter thanthe B0 supergiants listed in Table 2. These objects are clas-sified as O9.5 II. – Stars with broader, weaker Paschen series lines than theO9.5 II stars, often with strong C iii . In the few cases wherephotometry is available, these objects are 2–3 magnitudesfainter than O9.5–B0 supergiants. These objects are classi-fied as O9-9.5 III. Comparison with synthetic spectra gen-erated with the non-LTE model atmosphere code CMFGEN(Hillier & Miller 1998b) suggests that such a classification isgenerally reasonable, with non-tailored fits at 30kK (O9.5)and 32kK (O9) shown in Fig. 5. – Stars with very broad, weak Paschen series lines that are gen-erally flat-bottomed. C iii is also weak and broad. R − bandspectra, when available, show shallow and possibly double-lined He i and infilled H α , while the objects may be unex-pectedly luminous for the implied early spectral type. Theseobjects have very similar I-band morphologies to known bi-naries in Wd1 and are described in more detail in Sect. 3.5. Seven stars appear to be bona fide late-O supergiants on the ba-sis of their I − band spectra; we therefore adopt classifications ofO9.5Iab for these objects (Table 2). W1036 has a R − band spec-trum (Fig. 6) which, demonstrating strong, narrow H α and He i photospheric absorption, is entirely consistent with this conclu-sion.However, a number of cluster members do not fit neatly intothis scheme. W1057 would also be classified as O9.5 Iab on thebasis of its I − band spectrum. However it is one of the more lu-minous O-type stars in our sample, with an R − band spectrumshowing infilled H α and weak wind lines of C ii (Fig. 6); thecombination of these properties suggest that it could instead bea rapidly-rotating B0 Iab supergiant. As such we adopt a classi-fication of O9.5–B0 Iab for W1057, noting that other O9.5Iabstars, such as W1024 and -1034, are similarly luminous (al-though no R − band data are available for them). The spectra ofW1033 and W1040 ( = C07-X5 and C07-X3 respectively; Clarket al. 2008, 2019c) are hard to interpret, with Paschen serieslines that suggest a classification of O9 III, but weak C iii sug-gesting a later O9.5 (or possibly B0) spectral type and photo-metric magnitudes that are consistent with supergiants in Wd1.Low-resolution R − band spectroscopy of W1040 , showing in-filled H α (Fig. 6), again supports a supergiant identification. Wetherefore assign an intermediate classification of O9-9.5 I-III forboth stars. Finally W1041 is classified as O9.5Iab but presentsanomalously broad Paschen lines; we discuss this further in Sect.3.5. Nine cluster members are classified as O9.5 II due to theirspectroscopic similarity to Wd1-74 (O9.5 Iab), with the impor-tant diagnostic exceptions of broader wings to the Paschen se-ries lines - indicative of a higher surface gravity - and I − bandmagnitudes that are somewhat lower than the O9.5 Iab super-giants. Low-resolution FORS2 / MXU R − band spectra are avail-able for W1022, -1047 and -1050 (Fig. 6), all of which are con- W1018, -1024, -1027, -1030, -1034, -1036, and -1064. W1008, -1022, -1037, -1042, -1045 -1047, -1050, -1056, and -1060 J. S. Clark et al.: Stellar census of Wd1 no r m a li z ed f l u x ( + o ff s e t ) wavelength (Å)W1014*W1022W1028*W1032*W1036W1040* He IHe IH a no r m a li z ed f l u x ( + o ff s e t ) wavelength (Å)W1014*W1022W1028*W1032*W1036W1040* He IHe IH a no r m a li z ed f l u x ( + o ff s e t ) wavelength (Å)W1014*W1022W1028*W1032*W1036W1040* He IHe IH a no r m a li z ed f l u x ( + o ff s e t ) wavelength (Å)W1014*W1022W1028*W1032*W1036W1040* He IHe IH a no r m a li z ed f l u x ( + o ff s e t ) wavelength (Å)W1014*W1022W1028*W1032*W1036W1040* He IHe IH a no r m a li z ed f l u x ( + o ff s e t ) wavelength (Å)W1014*W1022W1028*W1032*W1036W1040* He IHe IH a a a a a a a Fig. 6.
Montage of select FORS2 / MXU R -band spectra of newly-identified cluster members. Objects marked with a ‘*’ show verybroad Paschen series lines indicative of a SB2.sistent with such a classification, demonstrating narrow single-lined photospheric He i lines with H α in absorption.W1043 is photometrically fainter than this sample, withPaschen series lines that also appear slightly broader; as suchwe assign a provisional classification of O9.5 II–III. An R − bandspectrum is available (Fig. 6); He i absorption lines are com-parable to other O9.5 II stars, although H α appears somewhatbroader and infilled. W1002 also has a luminosity similar to theO9.5 II systems discussed above but displays very broad, dilutePaschen series lines that are morphologically similar to knownbinaries in Wd1; given this we suspect this system may containan O9–9.5 II primary and an O-type secondary, which we dis-cuss further in Sect. 3.5.Finally, Bonanos (2007) reported that both W1002 and -1022display rapid ( P < δ Scutivariables. However our spectral classifictions are clearly incon-sistent with such a suggestion and are also earlier than expectedfor β Cephei-type variables (cf. Pamyatnykh 1999).
After accounting for the above stars, 35 cluster members remainto be classified. Of these 12 stars fulfill the preceding observa-tional criteria for classification as O9-9.5 III stars (cf. W1015;Figs. 2 & 5). Unfortunately no R − band spectra are available forany of these objects. A further 13 stars have spectra and, whereavailable, photometry consistent with an O9-9.5 III classifica-tion - which we adopt - albeit with notably broader photosphericPaschen series lines. We discuss these stars as well as (i) thehandful of more luminous stars previous highlighted as showingthe same phenomenon and (ii) ten further, but more extremeexamples - which, as a consequence defy precise determina-tion of spectral type and / or luminosity class (cf. Tables 1 & 2) -immediately below. . S. Clark et al.: Stellar census of Wd1 9 W71
W265 W1044W1028 R e l a t i v e w i d t h ( W = ) I-band magnitude
W1014W1012W1017W1062W1006
Fig. 7.
Relative Pa-11 line width as a function of I − band lumi-nosity for targets with photometry from Clark et al. (2005). Forclarity, the box in the top left of this figure presents an expandedview of the points within the small dotted region. While the spectral sequence shown in Figs. 2 and 3 providesa good general template for classification, many of the faintestobjects display a distinct I-band morphology, with weak, verybroad, and flat bottomed Pa-11 and Pa-12 lines, almost no traceof Pa-13. . . 15, and broad, weak C iii I − band mag-nitude for objects with photometry from Clark et al. (2005).Cluster members show a linear increase in the width of the Pa-11 line towards lower luminosity classes, but many of the faintestobjects display substantially broader lines than this progressionwould suggest. In Fig. 8 we plot two representative examplesof the broad lined population - W1020 and -1046 - against theO9 III classification standard W1015, the binary systems Wd1-36 and -53 ( P ∼ .
18d and ∼ .
3d respectively; Bonanos 2007)and the post-binary interaction massive blue straggler Wd1-30a(O4-5Ia + ; Clark et al. 2019a). It is clear from these data that thebroad Paschen series lines of W1020 and -1046 are not a goodmatch to the sharp bottomed photospheric profiles of W1015 andWd1-30a (where the spectrum of the later is dominated by theearly hypergiant component).Instead, given the strong morphological similarities that ex-ist between W1020 and -1046 and the binaries Wd1-36 and -53,we suggest that many, if not all , of the observed broad-lined sys-tems are SB2 binaries, with a composite spectrum that includescontributions from a ∼ O9 III primary and OB III–V secondary.Support for this conjecture is provided by: – The additional presence of comparable I − band spectral mor-phologies in the photometrically identified binaries Wd1-6a(Fig. 7) and W1021 ( P ∼ .
20d and ∼ .
43d respectively; W1007, -1015, -1023, -1026, -1031, -138, -1051, -1052, -1058, -1059, -1063, and -1066. W1006, -1012, -1014, -1016, -1017, -1019, -1021, -1028, -1029,-1032, -1035, -1044, and -1061. W1002, -1040, -1041 and -1055 W1001, -1003, -1010, -1011, -1013, -1020, -1025, -1046, -1054,and -1062.
Bonanos 2007); four such broad-lined systems are thus un-ambiguously binary. – Despite our I − band observations being poorly-suited to de-tecting radial velocity changes in such broad-lined systems(Sect. 3.1), three stars - W1028, -1032, and -1061 - show ra-dial velocity shifts indicative of reflex binary motion ( σ rv ∼ − − ; Clark et al. in prep.). – The appearance of anomalously broad and / or doubletroughed profiles in the photospheric H α and He i / MXU spec-troscopy (Fig. 6). Specifically W1046 clearly shows adouble-lined He i α and He i lines that arebroad and shallow compared to both (apparently) single stars(e.g. W1036; O9.5 Ib) and binaries in which the primary issubstantially more luminous than the secondary and hencedominates the spectrum (e.g. W1065; B0 Ib).As a consequence we provisionally apply generic classifi-cations of O9-9.5III bin? and O + O? to those stars itemised infootnotes 8 and 10, respectively. The more luminous binariesWd1-36 and -53a are likewise assigned an OB + OB classifica-tion while, as described previously, more precise classificationsare possible for W1002, -1040, -1041 and -1055 (Table 2).If this distinctive I − band morphology is indeed indicative ofbinarity then the implied binary fraction amongst the late-O pop-ulation in Wd1 appears substantial. However, considerable cau-tion must be exercised before accepting this assertion. An obvi-ous selection e ff ect that must be allowed for is that binaries withan appreciable contribution from a massive secondary will bemore luminous than an equivalent isolated star, and hence morelikely to meet our photometric criteria for selection. Moreoverin a subset of cases the broad-lined morphology might reflectrapid rotation rather than binarity. As a consequence, confirmingthe binary nature of the majority of these broad-lined systemsis clearly a priority. Unfortunately this is technically challeng-ing since it will require high-resolution observations targetingthe He i / FLAMESconfiguration yields spectra of insu ffi cient S / N to accomplishthis goal in a reasonable timeframe (cf. Clark et al. in prep.).
4. Discussion
Our new observations yield a total of 69 new cluster membersfor Wd1, bring the total census to 166 massive evolved stars.Understandably, given the fainter nature of the majority of tar-gets (Fig. 1), we find that they are predominantly biased to-wards less evolved O stars of lower luminosity than the clus-ter OB supergiants previously classified (Table 1). We highlightthat spectral classification for all objects has been uniformly ac-complished via a combination I − band and optical photometry,supplemented with R − band spectroscopy where available (Sect.3 and Negueruela et al. 2010); as a consequence we consider thisprocess to be internally robust.A particular problem with the quantitative interpretation ofthese data is that Wd1 su ff ers from heavy di ff erential reddeningand the form of the extinction law is ill-constrained, which inturn leads to a large range of possible bolometric luminositiesfor individual cluster members (Clark et al. 2019a). As a resultwe refrain from updating the semi-empirical cluster HR diagramintroduced by Negueruela et al. (2010) at this time; instead sum- N o r m a li z ed f l u x ( + o ff s e t ) Wavelength (Å) Paschen seriesDIBC III W30 (O+O)W36 (3.18d)W53 (1.3d)W1046W1020W1015 (O9 III)C (2−0) 0.75 1 1.25 1.5 1.75 2 2.25 8500 8550 8600 8650 8700 8750 8800 8850 8900 N o r m a li z ed f l u x ( + o ff s e t ) Wavelength (Å) Paschen seriesDIBC III W30 (O+O)W36 (3.19d)W53 (1.3d)W1046W1020W1015 (O9 III)C (2−0) 0.75 1 1.25 1.5 1.75 2 2.25 8500 8550 8600 8650 8700 8750 8800 8850 8900 N o r m a li z ed f l u x ( + o ff s e t ) Wavelength (Å) Paschen seriesDIBC III W30 (O+O)W36 (3.19d)W53 (1.3d)W1046W1020W1015 (O9 III)C (2−0) 0.75 1 1.25 1.5 1.75 2 2.25 8500 8550 8600 8650 8700 8750 8800 8850 8900 N o r m a li z ed f l u x ( + o ff s e t ) Wavelength (Å) Paschen seriesDIBC III W30 (O+O)W36 (3.19d)W53 (1.3d)W1046W1020W1015 (O9 III)C (2−0) 0.75 1 1.25 1.5 1.75 2 2.25 8500 8550 8600 8650 8700 8750 8800 8850 8900 N o r m a li z ed f l u x ( + o ff s e t ) Wavelength (Å) Paschen seriesDIBC III W30 (O+O)W36 (3.19d)W53 (1.3d)W1046W1020W1015 (O9 III)C (2−0) 0.75 1 1.25 1.5 1.75 2 2.25 8500 8550 8600 8650 8700 8750 8800 8850 8900 N o r m a li z ed f l u x ( + o ff s e t ) Wavelength (Å) Paschen seriesDIBC III W30 (O+O)W36 (3.19d)W53 (1.3d)W1046W1020W1015 (O9 III)C (2−0) Fig. 8.
Comparison the I − band spectra of known and newly-identified candidate binaries with the O9 III star W1015. Wd1-30a isa strong X-ray source indicative of a colliding-wind binary, while Wd1-53 displays a 1.3-day periodic modulation in its light curveand double He i lines, and Wd1-36 is a 3.18-day eclipsing binary.marising the data via a colour / magnitude plot (Fig. 9 and Sect.4.1).Nevertheless a number of features are obvious from the cur-rent analysis: – The presence of a smooth progression in spectral morphol-ogy / stellar classification of evolved stars from late-O giantsthrough to early-mid B supergiants and ultimately late B hy-pergiants, with the supergiant population dominated by spec-tral types O9-B1 (60 examples). – An unprecedentedly rich population of both hot and cool hy-pergiants with, uniquely, spectral types ranging from O4-5Ia + (Wd1-30a) to F8 Ia + (Wd1-8a) and, arguably, M5 Ia(Wd1-20). – A population of WRs exhibiting a diverse array of spectralsub-types. – A large number of spectroscopic binary candidates, includ-ing a number of interacting / post-interaction systems.We discuss the implications of these findings in more detailbelow, in the context of both stellar evolution and the bulk prop-erties of Wd1. The extreme mass inferred for Wd1 (Clark et al. 2005, Limet al. 2013, Andersen et al. 2017) implies that even veryrapid / rare phases of massive stellar evolution may be repre-sented. Conversely at an apparent age of ∼ M init & M ⊙ to have been lost to core-collapse ; as a consequence it is also of interest to review thestellar subtypes that are not present. Considering first the leastevolved members and, as might be expected, very early O2-4 I-III stars - found within the youngest ( . + and O7-8 Ia + hypergiantsWd1-30a and Wd1-27, which appear to be the products of binaryevolution (Sect. 4.4; Clark et al. 2019a). Discounting these starson this basis, the earliest ‘canonical’ supergiants are of spectraltype ∼ O9Ib...Iab (e.g. Wd1-15 and -17).The presence of a rich population of late-O to mid-B super-giants has long been recognised (Clark et al. 2005, Negueruelaet al. 2010). In contrast to such earlier spectroscopic studies, our Groh et al. (2013) predict that for an age of 4Myr non-rotating (ro-tating) stars of ∼ M ⊙ ( M init ∼ M ⊙ ) will currently be undergingcore-collapse and at 5Myr M init ∼ M ⊙ ( ∼ M ⊙ ).. S. Clark et al.: Stellar census of Wd1 11 observations include large numbers of lower luminosity objectsin Wd1, with target selection based on photometry consistentwith heavily-reddened OB stars. These criteria would not distin-guish between O9 III stars evolving to become such luminousOB supergiants - which are seen in large numbers - and earlyB0.5–2 Ib...II stars from an older population of stars with lowerinitial masses currently evolving to the RSG phase. Such starswould be readily distinguished from the O9 III population bythe presence of strong He i (and possibly N i ) lines and the ab-sence of C iii . Critically they are entirely absent from our survey;the only B-type objects are super- / hypergiants of su ffi cient lu-minosity that they are consistent with the wider evolved stellarpopulation of Wd1 (Sect. 4.2).Examples of every transitional evolutionary phase betweenH-rich OB supergiants and H-depleted WRs are found withinWd1 (e.g. BHG, LBV, sgB[e], YHG, and RSG). Surprisingly,only one LBV is identified (Wd1-243; Clark & Negueruela2004), compared to large numbers of confirmed and candi-date systems within, for example, the Quintuplet (Clark et al.2018b) . Intriguingly, Wd1-243 also has a rather moderatemass-loss rate (Ritchie et al. 2009b, Fenech et al. 2018) in com-parison to other (candidate) LBVs such as P Cygni (Najarro etal. 2001), AG Car (Groh et al. 2009) and HDE 316285 (Hillier etal. 1998a) where, unlike Wd1-243, the wind densities are su ffi -cient to drive the Paschen series into emission. One might appealto the comparatively low temerature of Wd1-243 to explain thisdiscrepancy; however the Paschen series is seen in emission inthe YHG IRAS 18457-0604, which appears to be of comparablespectral type to Wd1-243 (Clark et al. 2014a). Expanding on thispoint, and it is noteworthy that the population of YHGs withinWd1 also lack the pronounced emission line spectra of other ex-amples, such as IRAS 18457-0604 and IRC +
10 420 (Clark etal. 2014a). Apparently comparatively low mass loss rates are afeature of the the current population of LBVs and YHGs withinWd1, although the presence of extended mm- / radio nebulae as-sociated with a number of these stars is indicative of extensivemass loss in the recent past (Dougherty et al. 2010, Andrews etal. 2018, Fenech et al. 2018).Finally we turn to the WR content of Wd1. A wide rangeof WN sub-types are present, although the earliest (WN2-4) ex-amples are not represented (Table 4). No WN5-9ha stars are ob-served; consistent with the expectation that they descend fromexceptionally massive stars (e.g. Lohr et al. 2018a, Schnurr etal. 2008, Bonanos et al. 2004). Three late WN9-11h stars areidentified (Wd1-5, -13 and -44); however all are potentially theproducts of binary interaction (Ritchie et al. 2010, Clark et al.2014b, in prep.). In contrast to the WN stars, the WC stars areuniformly of late (WC8-9) spectral subtype; neither transitionalWN / WC nor WCE stars are identified; we return to this in Sect.4.3. Lastly, Groh et al. (2013) predict that non-rotating (rotating)stars of M init & M ⊙ ( & M ⊙ ) present as WRs of WO subtypeimmediately prior to core-collapse. Once again no examples arefound within Wd1; however with predicted optical magnitudes & . / magnitude diagram given in Fig. 9, constructed with thedataset presented in Clark et al. (2005); inevitably this excludesthe 62 stars with spectral classifications but lacking suitable pho-tometry (Table 2). Nevertheless, remaining cluster members areclearly delineated from field stars by their excessive redden- We note that Gvaramadze (2018) suggest that the LBV MN44 is arunaway from Wd1. ing, with the detection threshold set by the V − band magnitude.Stars assigned both a spectral types and luminosity class arefurther indicated; unfortunately a subset of binaries cannot bedesignated on this basis (Wd1-9, -36, -53a and those in foot-note 10). The progression in both magnitude and colour withspectral classification is broadly as expected with, for exam-ple, the cool super / hypergiants being systematically redder andbrighter than the OB (super-)giants. The breadth of the regionoccupied by cluster members in the diagram is indicative of sig-nificant di ff erential reddening, with Negueruela et al (2010) find-ing ∆ E ( V − I ) ∼ . − . ∼ . ± . V − I ) ∼ .
6; as shown in Fig. 9, with the excep-tion of a single outlier this indeed serves as a suitable discrim-inant. Uncertainties in spectral classification and the e ff ect ofbinarity likewise contribute to the overlapping regions occupiedby, for example, O9-B0.5 stars of luminosity classes I and II...III(cf. Sect. 3.4.1 and 3.5). Finally the diagram reveals the resid-ual incompleteness of our current spectroscopic survey. This isparticularly prevalent amongst the fainter cohort of members,although this shortcoming also extends to stars within the clus-ter core with luminosities suggestive of a supergiant nature, butwhich still await spectral classification (e.g. Wd1-12b, -39a and40a) and, in some cases, photometric observations. Previous studies attempting to determine the age of Wd1have focused on both high- and low-mass stellar components.Placement of the OB super- / hypergiant population on a semi-empirical HR-diagram suggests a high degree of coevality forthis cohort, with a comparison to evolutionary tracks imply-ing an age of ∼ < . ∼ . − ff erentgroups fail to converge on a consistent set of physical propertiesfor this phase (cf. comparison in Britavskiy et al. 2019; see alsoDavies et al. 2018 for a comparison to observations); potentiallycomplicating the employment of such a diagnostic.Unfortunately a combination of saturation and blending ofthe Wd1 RSGs at IR wavelengths precludes accurate determina- Fig. 9.
Colour magnitude plot of the 5’ ×
5’ field-of-view centred on Wd1 utilising the photometry of Clark et al. (2005). Clustermembers with both spectral types and luminosity classes are denoted as follows: RSGs and YHGs - red and yellow diamonds;B5-9 Ia + - cyan diamonds; B1-4 Ia,Iab,Ib - cyan squares; O9-B0.5 Ia,Iab,Ib - blue squares; O9-B0.5 II,III - blue circles; earlyBHG / WNVLh - purple diamonds; WN5-8 & WC8-9 - purple circles. The vector shows the e ff ect of di ff erential reddening across thecluster (cf. Negueruela et al. 2010) assuming an ‘o ff -the-peg’ Cardelli et al. (1989) prescription. A colour cut for cluster membershipof ( V − I ) ∼ . L bol / L ⊙ & . . S. Clark et al.: Stellar census of Wd1 13 fer the assembly of an empirical HR diagram - and consequentcomparison of the cluster population to theoretical isochrones- until this issue is resolved. As such we are currently unableto improve on the quantitative age estimate of Negueruela etal. (2010). Notwithstanding this, the newly classified cohortof early-type stars appears qualitatively consistent with the as-sertion that the (massive) stellar population of Wd1 is essen-tially co-eval. Specifically, the super- / hypergiant evolutionarysequence described by Negueruela et al. (2010) can now be ex-tended to lower-luminosity giants of systematically earlier (O9-9.5) spectral type, yielding a remarkably homogeneous popula-tion of OB stars within Wd1. In the (current) absence of an iden-tifiable population of core H-burning stars of luminosity class V,the sole examples of O stars of even earlier ( < O9) spectral typeappear to be binary products, while no B supergiants descend-ing from stars of lower initial masses have been distinguished.We therefore conclude that, subject to completeness (Sect. 4.1),our current stellar census provides no qualitative evidence foryounger or older stellar populations within Wd1.
Utilising a combination of stellar evolution and non-LTE model-atmosphere codes, Groh et al. (2014) and Martins & Palacios(2017) calculated synthetic spectra for single non-rotating mas-sive stars, from which empirical spectral classifications can bederived for comparison to real data. Both sets of calculationsimply that stars with M init ∼ M ⊙ fail to yield the early-midB supergiants present within Wd1; indeed at an age of > M init ≤ M ⊙ pathways. This leaves a 20 M ⊙ < M init < M ⊙ window which appears to yield OB stars of the spectraltype and luminosity class found within Wd1, with the WR co-hort deriving from the upper reaches of this range. The eclips-ing binary Wd1-13 provides direct support for this supposition,with Ritchie et al. (2010) deriving current dynamical massesof ∼ . + . − . M ⊙ and ∼ . + . − . M ⊙ for the B0.5 Ia + / WNVLhand OB supergiant components, respectively. Under highly non-conservative late-Case A / Case B mass transfer, a pre-interaction M init ∼ M ⊙ is inferred for the primary. Similarly high progeni-tor masses are suggested by quantitative analyses of Wd1-5, -27and -30a (Clark et al. 2014b, 2019a). Although subject to sys-tematic uncertainties in both cluster distance and and extinctionlaw the semi-empirical HR diagram of Negueruela et al. (2010)is also fully consistent with this hypothesis.Outside of Wd1, few dynamical mass-estimates for late-O / early-B stars of luminosity class I-III are available; being lim-ited to CC Cas (O8.5 III, 35 . ± M ⊙ ; Gies 2003), the secondaryin the detatched binary HD 166734 (O9 I(f), ∼ . + . − . M ⊙ ;Mahy et al. 2017) and the evolved, post-interaction primaryin Cyg OB2-B17 (O9 Iaf, 45 ± M ⊙ ; Stroud et al. 2010).Nevertheless these are consonant with both values derived fromWd1 and also the the mass range inferred for field B0-3 Ia stars( ∼ − M ⊙ ) by Crowther et al. (2006b).Following from the above we may attempt to infer an inte-grated mass for Wd1 from the massive star census, assuming aMaschberger (2013) IMF with standard parameters, including anupper-mass slope of 2.3 (ie. Salpeter). If we take a highly con-servative limit that the 166 spectrally classified cluster membersspan a range of 20–50 M ⊙ then the current mass of Wd 1 (ie. themass of stars between 0.1 and 50 M ⊙ ) is ∼ × M ⊙ (with a ∼ ⊙ , the current mass of Wd 1 is ∼ × M ⊙ (again with a ∼ ⊙ increases the total mass by 50 per cent,showing the crucial importance of accurately determining thelower mass limit. Conversely, the upper mass limit is much lessimportant; for example if we were to specify a range of massesof 20–45 M ⊙ , the estimated integrated mass would increase byless than 10 per cent.We refrain from specifying a contribution from higher massstars since it would require an extrapolation of the mass func-tion to stars already lost to core-collapse and likely comprises acomparatively minor contribution to the integrated cluster mass.Moreover, we caution that if the complete initial mass functiondeparts from a Salpeter form - for example if it is top-heavy - theabove masses would be over-estimates. At this juncture it is instructive to compare the stellar populationof Wd1 to other young massive clusters. Both NGC3603 andR136 are su ffi ciently young ( ∼ − ff erentfrom Wd1 and so we do not consider them further here. Of morerelevance are the three clusters in the central molecular zone ofthe Galaxy. Both the Arches and Quintuplet are expected to beyounger than Wd1 ( ∼ − . ∼ − . ± .Once di ff ering ages are taken into account, similarities areapparent between the evolutionary sequences of H-rich clus-ter members of the Arches, Quintuplet and Wd1 (with theGalactic Centre population insu ffi ciently sampled for compar-ison). Specifically, stars in the three clusters demonstrate asmooth progression in spectral morphologies extending fromOB supergiants (O4-6 Ia, O7-B0 Ia and O9-B4 Ia respectively)through to hypergiants (O4-8 Ia + , B0-3 Ia + and B5-9 Ia + ). Whileabsent from the Arches due to its youth, more advanced evolu-tionary phases, such as LBVs and the related WN9-11h stars,are found within the Quintuplet, with the YHGs and RSGs play-ing an analogous role in Wd1. Moreover, both Wd1 and theQuintuplet contain a cohort of stars with anomalously earlyspectral types , of which a number would fit seemlessly intothe Arches cluster; potentially indicative of binary interactionand rejuventation (Sect. 4.4 and Clark et al. 2018a, 2018b).However the WR cohorts of the four clusters look very dif-ferent from one another (Table 4). The Arches appears too youngfor any H-free examples to be present, with a population solelyconfined to WN7-9ha stars (Clark et al. 2019b). Conversely, Such a large age spread results from uncertainties in both pop-ulation synthesis and the quantitative modeling of cluster members(Paumard et al. 2006, Martins et al. 2007). However at most three RSGsmay be physically associated with the Galactic Centre cluster (Blum etal. 1996) implying a much more extreme WR / (YHG + RSG) ratio thanfound for Wd1 (33:3 versus 24:10); potentially indicative of a youngerage (Davies et al. 2009). Such an assertion is further supported by therather high dynamical masses inferred for the components of the binaryGCIRS 16SW ( ∼ × M ⊙ ; Martins et al.2006a). qF256 (WN8-9ha), -274 (WN8-9ha) and -406 (O7-8 Ia + ) andLHO01 (O7-8 Ia + / WNLha) and -54 (O7-8 Ia + ) within the Quintupletand Wd1-5 (WN10-11h / B0.5 Ia + ), -13 (B0.5 Ia + ), -27 (O7-8 Ia + ), -30a(O4-5 Ia + ), -44 (WN9h), and W1049 (B1-2Ia + ).4 J. S. Clark et al.: Stellar census of Wd1 WN sub-types appear surprisingly under-represented within theQuintuplet, which contains just three WN9-11h, two WNLha(footnote 14) and a single WN6 star - despite Groh et al. (2014)predicting a WNE phase lasting ∼ × longer than the WNLphase. Instead the cluster is dominated by WC8-9 stars (Table4), many of which appear to be binaries identifiable by excessIR emission from hot dust (Clark et al. 2018b and refs. therein);again none of the WCE stars anticipated by Groh et al. (2014)are present. In stark constrast, the WN:WC star ratio is reversedfor Wd1, with all spectral sub-types from WN5 to WN11h rep-resented, albeit with an apparent (weak) bias towards earlier(WN5-7) stars.The WN:WC ratio for the Galactic Centre cluster is moreevenly balanced although, unlike Wd1, the WN populationappears biased towards later spectral sub-types (Table 4). Aswith the Quintuplet and Wd1, the WC population is dominatedby WC8-9 sub-types, with the exception of a single WC5 / → WN8 → WN / WC → WC8-9for stars within this cluster. No such simple scheme canbe constructed for Wd1 due to an absence of transitionalWN / WC stars and the range of WN spectral sub-types observed.Similarly the almost complete lack of H-free WN stars withinthe Quintuplet complicates interpretation of evolution from theBHG / LBV / WN9-11h to WC phase at very high initial masses( & M ⊙ ; Clark et al. 2018b); we return to these observationalfindings in Sect. 4.4.We may also extend this comparison beyond core-collapse.One of the defining features of Wd1 is the presence of the mag-netar CXO J164710.2-455216 (Muno et al. 2006a). Intriguingly,a magnetar has also been associated with the Galactic Centrecluster (SGR J1745-29; Kennea et al. 2013, Mori et al. 2013),while PSR J1746-2850 - either a high B − field pulsar or tran-sient magnetar - is located in close proximity to the Quintuplet( ∼ pc ; Deneva et al. 2009, Dexter et al. 2017) although, giventhe population of isolated massive stars in the central molecularzone, an unambiguous physical association has yet to be proven.Finally, in terms of stellar content - and hence age - the hostcluster of SGR1806-20 appears to be directly comparable to theQuintuplet (Figer et al. 2005, Bibby et al. 2008). These obser-vations strongly suggest that clusters in the ∼ − ffi cient sites for their formation, despite a naiveexpectation that under a single star evolutionary channel suchmassive progenitors should instead form black holes. In turn thisimplies that either the physical process by which neutron starsform from massive progenitors favours the production of mag-netars (cf. Clark et al. 2014a) or that (counter-intuitively) a largefraction of neutron stars are born as magnetars; a conclusion re-cently reached by Beniamini et al. (2019). Two key findings of this work are the unprecedented range ofhypergiants within Wd1 - not reproducible for a co-eval stel-lar population under single star evolution (e.g. Ekstr¨om et al.2012, Brott et al. 2011) - and the apparently rich binary popu-lation amongst the earliest O stars present. Specifically, of the27 O9-9.5 III stars within Wd1 (Table 2), one is a confirmedphotometric binary (W1021) and 13 have anomalously broadPaschen lines apparently indicative of binarity (of which threeare RV variable; Sect. 3.4.3 & 3.5 and footnote 8). Moreover, a further 10 objects have Paschen lines su ffi ciently broad to pre-clude precise classification but otherwise appear observation-ally consistent with an O + O designation (footnote 10). This ap-pears indicative of a remarkably high binary fraction, althoughwe caution that with both components of comparable luminos-ity - as shown by their composite spectra - such systems will besystematically brighter than their single siblings, introducing apotentially significant bias into their identification. Irrespectiveof this, we also identify similar candidate spectroscopic bina-ries amongst brighter, more evolved stars (footnote 9), while weare able to confirm that a number of photometric binaries fromBonanos (2007) are indeed luminous OB cluster members (e.g.Wd-36, -53, and W1048).These results are consonant with previous observationswhich imply a large number of massive binaries within Wd1.Examples include (i) a binary fraction of &
70% inferred forWR stars via IR and X-ray observations (Crowther et al. 2006a,Clark et al. 2008, Clark et al. 2019b) and (ii) the embodimentof very short-lived phases of binary interaction such as Wd1-9,which is thought to be undergoing rapid case-A mass transfer(Clark et al. 2013).Given the wealth of observational diagnostics, we defer aquantitative analysis of the full binary population of Wd1 to afuture paper, where we present a synthesis of multiple datasets,including results from our full RV spectroscopic survey (Clarket al. in prep.). Nevertheless, just considering the data presentedhere provides compelling observational evidence for the role thatbinary-driven evolution plays within Wd1.Following Negueruela et al. (2010) we may presume asingle star evolutionary channel broadly progressing:late-O giants → early-B supergiants → mid to late-Bhypergiants → YHG / RSGsbefore looping back bluewards to H-depleted WRs, poten-tially via an LBV phase (cf. Wd1-243). However neither themid-O nor the early-B hypergiants within Wd1 are consistentwith such a scenario and hence we must suppose two futher,binary-modulated channels to yield these stars (cf. Clark et al.2019a).The first is uncontroversial and presumably occurs as the pri-mary evolves towards the supergiant phase and fills its Rochelobe, with the resultant binary-driven mass stripping yield-ing undermassive but overluminous and chemically peculiarWNLh / BHG stars such as Wd1-5, -13, -44 and, presumably,W1049 (Petrovic et al. 2005, Ritchie et al. 2010, Clark et al.2014b). Secondly one might suppose that such binary interac-tion may also lead to the secondary accreting significant quanti-ties of mass or, in extreme cases, merger. In this case one wouldexpect a very luminous and massive blue straggler to form; ex-amples being Wd1-27 and -30a (de Mink et al. 2014, Schneideret al. 2014, Clark et al. 2019a). Such a channel is more contro-versial since it requires the secondary to be able to accrete largequantities of mass without spinning it up to critical rotation viathe transfer of angular momentum (which would quickly haltaccretion; Petrovic et al. 2005, de Mink et al. 2014); physicalmechanisms that might facilitate this include angular momen-tum loss via an accretion disc or tidal interaction between bothcomponents.Unfortunately, the relative prevalence or weighting ofboth channels is currently unclear for a number of reasons.Evolutionary codes including all relevant physical processeshave yet to be constructed, with the e ffi ciency of the accretion ofmass and angular momentum essentially treated as a free param- . S. Clark et al.: Stellar census of Wd1 15 eter at this point. This uncertainty also means that subsequent,post-interaction evolution is opaque; given the wide range ofWN sub-types present within Wd1 it would appear highly likelythat some result from a binary evolutionary channel (cf. Clarket al. 2019c), but we lack theoretical predictions to validate thisconclusion (cf. Groh et al. 2014). Observationally, one might ex-pect that in many realisations of mass-stripping the secondarydominates the emergent, post-interaction spectrum, preventingidentification and characterisation of the primary (cf. G¨otberg etal. 2018).Focusing specifially on Wd1 and since the secondaryin Wd1-13 superficially resembles other cluster supergiants(Ritchie et al. 2010), one might suppose that a number ofapparently single OB stars within Wd1 are also in a post-interaction phase. However extensive and systematic quantita-tive model-atmosphere analysis will be required to identify suchbinary products via their rapid rotation and / or anomalous chem-ical abundances. Indeed we caution that while the mass ra-tio has reversed in Wd1-13, we cannot empirically constrainthe quantity of mass the secondary has accreted at this time,and hence whether the system evolved via (quasi-)conservativemass-transfer.As a consequence we are simply left with the conclusion thatboth channels must operate in parallel in Wd1, with Wd1-27 and-30a indicating that mass-transfer may be very e ffi cient (Clarket al. 2019a). Conversely, the extreme mass loss rate exhibitedby the sgB[e] star and interacting binary Wd1-9 reveals that incertain instances much of the mass stripped from the primary islost from the system rather than accreted by the secondary (Clarket al. 2013, Fenech et al. 2017, 2018).
5. Concluding remarks and future prospects
In this work we present classifications of a further 69 members ofWd1, producing a current census of 166 massive, evolved stars.As expected, given the photometric selection criteria adopted,the majority of these are late-O stars of luminosity type I-III,which smoothly extend the morphological sequence of hot andcool super- / hypergiants identified by Negueruela et al. (2010;Sect. 4.4) to higher temperature and less evolved objects. Whilea handfull of B-type supergiants and two new hypergiants havealso been identified, no lower luminosity B stars - which wouldbe indicative of an older stellar population - have been discov-ered. Likewise we find no evidence for a younger, more massivepopulation. We conclude that on this basis our current stellarcensus is consistent with the hypothesis that Wd1 is co-eval.Unfortunately, given the limitations of these data and un-certainties in both the distance and extinction towards Wd1 itis premature to construct an HR diagram for the cluster and sowe may not improve on previous quantitative age estimates viacomparison to theoretical isochrones, or provide initial massesfor individual stars. However comparison to the the stellar pop-ulation of the ∼ − . ff erences, most noticably in the spectral typedistributions of OB super- / hypergiants, that indicate that Wd1 isolder. Indeed the updated census of massive stars presented hereis consistent with previous age estimates of ∼ ff erent methodologies (Sect. 4.2.1).Moreover, qualitative comparison of the observations to thesingle star synthetic spectra generated by Martins & Palacios(2017) suggests that, collectively, the cluster members sampledevolved from 20 M ⊙ < M init < M ⊙ progenitors (Sect. 4.2),with more massive stars having already been lost to SNe (Grohet al. 2013). This would in turn imply that we may expect to see an evolutionary turn-o ff within Wd1 around O7-8 V, correspond-ing to stellar masses of ∼ M ⊙ (Martins & Palacios 2017, Clarket al. 2019a).Based on spectral morphology and classification we mayconstruct an apparent evolutionary sequence for H-rich singlestars within Wd1 up to the hypergiant phase that is analogousto that inferred for the Quintuplet (Sect. 4.3 & 4.4). However,this progression cannot be extended to the H-free WR cohortfor either cluster, with that of Wd1 found to be unexpectedy di-verse (Table 4). Moreover a population of massive blue strag-glers, with properties inconsistent with these evolutionary path-ways, may also be identified within both aggregates (footnote14). The simplest explanation for these phenomena is the onsetof binary interaction as stars evolve beyond luminosity class V.Consonant with both this hypothesis and previous observa-tional studies, our data suggests a large binary fraction for Wd1for stars at every evolutionary stage. These comprise both pre-interaction (W1021), interacting (Wd1-9) and post-interactionsystems (Wd1-5, -13, -27, -30a and WR A and B). The prop-erties of the latter cohort are indicative of two distinct but re-lated evolutionary channels producing low-mass stripped pri-maries and potentially rejuvenated, mass-gaining secondaries -the extreme mass-loss associated with the sgB[e] star Wd1-9indicating that in some instances mass lost from the primaryis ejected from the system rather than accreted. Unfortunatelyour current observations only identify the most extreme exam-ples of massive blue stragglers resulting from the latter pathway.Determining the relative prevalence of both evolutionary chan-nels will require quantitative analysis of the full cohort of OBsuper- / hypergiants in order to identify the subset which are po-tential binary products via their rapid rotation, anomalous sur-face abundances and / or mass luminosity ratios.Allowing for both incompleteness and the presence of mas-sive binary companions, a conservative estimate suggests thatWd1 currently contains >
166 stars with M init & M ⊙ .Assuming a standard Maschberger IMF this would imply arather extreme integrated cluster mass of ∼ × M ⊙ (Sect.4.2).As a consequence of their large integrated masses, the richstellar populations of Wd1 and the Arches and Quintuplet clus-ters provide a unique opportunity to study the post-main se-quence evolution of stars with M init ∼ M ⊙ to & M ⊙ inthe ∼ − ∼ & M ⊙ ; Melena et al. 2008, Zeidler et al. 2016, Bonanoset al. 2004, Schnurr et al. 2008). Conversely, older aggregatessuch as Berkeley 51 & 55 and the RSG dominated clusters at thebase of the Scutum-Crux arm ( ∼ − ∼ M ⊙ - delineating the onset of core-collapse - to be explored. Of particular interest is the prospect ofcharacterising the properties of stellar subtypes which have hith-erto been poorly studied due to the brevity of the evolutionaryphase; a possibility exmplified by the uniquely rich populationof hot and cool hypergiants within Wd1.One may also extend such e ff orts to core-collapse and be-yond. Groh et al. (2013) predict that non-rotating (rotating) starsof M init ∼ M ⊙ (28 M ⊙ ) will undergo core-collapse at an ageof 5.81Myr (7.92)Myr; thus depending on the admixture onewould expect all 166 stars within this census to be lost within . . that the immediate progenitors of such a process will be presentat this time and, if identifiable, ameanable to quantitative analy-sis. This has clear implications for the nature of the resultantrelativistic remnant; particularly interesting since clusters suchas Wd1 appear to function as highly e ffi cient factories for theproduction of magnetars (Sect. 4.4). Intriguingly the anticipatedmagnetic decay timescale for magnetars ( ∼ yr; Beniaminiet al. 2019 and refs. therein) is directly comparable to the SNrate within Wd1, suggesting that many core-collapse events mustproduce such objects. This raises the prospect of delineating theformation channel for magnetars (Clark et al. 2014b) - of partic-ular interest given that they have been hypothesised to power su-perluminous SNe (Thompson et al. 2004, Woosley 2010, Kasen& Bildsten 2010). Moreover it immediately raises the questionof how black holes of masses such as those found in high-massX-ray binaries (e.g. Cyg X-1 with M BH ∼ . ± . M ⊙ ; Oroszet al. 2011), and the more extreme merging black holes detectedvia gravitational waves, may form, if neutron star formationis strongly favoured for such comparatively massive progenitorstars. . S. Clark et al.: Stellar census of Wd1 17 Table 2.
The stellar population of Westerlund 1
ID RA (J2000) Dec (J2000) B V R I Spectral Type NotesW1 16 46 59.28 −
45 50 46.7 21.9 18.37 16.09 13.65 O9.5 IabW2a 16 46 59.71 −
45 50 51.1 20.4 16.69 14.23 11.73 B2 Ia RV binary? , H α variable W4 16 47 01.42 −
45 50 37.1 18.7 14.47 11.80 9.15 F3 Ia + W5 16 47 02.97 −
45 50 19.5 21.4 17.49 14.98 12.48 WN10-11h / B0.5 Ia + WR S, stripped primary W6a 16 47 03.04 −
45 50 23.6 22.2 18.41 15.80 13.16 B0.5 Iab P(2.20d) , H α variable W6b 16 47 02.93 −
45 50 22.3 23.6 20.20 17.91 15.25 O9.5 IIIW7 16 47 03.62 −
45 50 14.2 20.0 15.57 12.73 9.99 B5 Ia + H α variable, Pulsator? W8a 16 47 04.79 −
45 50 24.9 19.9 15.50 12.64 9.89 F8 Ia + W8b 16 47 04.95 −
45 50 26.7 − − − −
B1.5 Ia Pulsator? W9 16 47 04.14 −
45 50 31.1 21.8 17.47 14.47 11.74 sgB[e] Interacting binary W10 16 47 03.32 −
45 50 34.7 − − − −
B0.5 I + OB SB2W11 16 47 02.23 −
45 50 47.0 21.2 17.15 14.52 11.91 B2 IaW12a 16 47 02.21 −
45 50 58.8 22.0 16.94 13.54 10.42 F1 Ia + W13 16 47 06.45 −
45 50 26.0 21.1 17.19 14.63 12.06 B0.5 Ia + + OB E(9.27d) , W14c 16 47 06.07 −
45 50 22.6 − − − −
WN5o WR RW15 16 47 06.63 −
45 50 29.7 22.8 18.96 16.38 13.75 O9 IbW16a 16 47 06.61 −
45 50 42.1 20.5 15.89 12.82 9.90 A5 Ia + H α variable W17 16 47 06.25 −
45 50 49.2 22.7 18.87 16.19 13.56 O9 IabW18 16 47 05.71 −
45 50 50.5 21.2 17.32 14.81 12.27 B0.5 IaW19 16 47 04.86 −
45 50 59.1 22.6 18.22 15.21 12.37 B1 Ia H α variable W20 16 47 03.09 −
45 52 18.8 − − − −
M5 IaW21 16 47 01.10 −
45 51 13.6 22.5 18.41 15.56 12.74 B0.5 Ia Pulsator? W23a 16 47 02.57 −
45 51 08.7 22.1 17.85 14.91 12.07 B2 Ia + B I? H α variable, Pulsator? W24 16 47 02.15 −
45 51 12.4 23.0 18.71 15.96 13.24 O9 Iab Pulsator? W25 16 47 05.78 −
45 50 33.3 21.9 17.85 15.22 12.61 O9 IabW26 16 47 05.40 −
45 50 36.5 22.1 16.79 12.63 9.19 M2 ↔ W27 16 47 05.15 −
45 50 41.3 21.5 17.94 15.35 12.80 O7-8 Ia + Merger remnant? W28 16 47 04.66 −
45 50 38.4 20.9 16.87 14.26 11.64 B2 Ia H α variable W29 16 47 04.41 −
45 50 39.8 22.6 18.66 16.02 13.38 O9 IbW30 16 47 04.11 −
45 50 39.0 22.4 18.45 15.80 13.20 O4-5 Ia + RV binary? H α variable W31 16 47 03.78 −
45 50 40.4 − − − −
B0 I + OBW32 16 47 03.67 −
45 50 43.5 − − − −
F5 Ia + W33 16 47 04.12 −
45 50 48.3 20.0 15.61 12.78 10.04 B5 Ia + H α variable W34 16 47 04.39 −
45 50 47.2 22.1 18.15 15.40 12.69 B0 IaW35 16 47 04.20 −
45 50 53.5 22.7 18.59 16.00 13.31 O9 IabW36 16 47 05.04 −
45 50 55.3 22.8 18.89 16.09 13.38 OB Ia + OB Ia E(3.18d) , SB2 , very broad Pa linesW37 16 47 06.01 −
45 50 47.4 22.8 19.11 16.40 13.65 O9 IbW38 16 47 02.86 −
45 50 46.0 23.2 19.10 16.47 13.81 O9 IabW41 16 47 02.70 −
45 50 56.9 21.3 17.87 15.39 12.78 O9 IabW42a 16 47 03.25 −
45 50 52.1 − − − −
B9 Ia + H α variable, Pulsator? W43a 16 47 03.54 −
45 50 57.3 22.8 18.05 15.22 12.26 B0 Ia RV(16.27d) , SB1, H α variable W43b 16 47 03.52 −
45 50 56.5 − − − −
B1 IaW43c 16 47 03.76 −
45 50 58.3 20.4 18.35 16.18 13.66 O9 IbW44 16 47 04.20 −
45 51 06.9 22.6 18.86 15.61 12.52 WN9h: WR L, RV binary? W46a 16 47 03.91 −
45 51 19.5 23.0 18.55 15.46 12.46 B1 IaW46b 16 47 03.61 −
45 51 20.0 − − − −
O9.5 IbW47 16 47 02.64 −
45 51 17.6 22.7 19.95 16.36 13.68 O9.5 IabW49 16 47 01.90 −
45 50 31.5 22.6 18.76 16.30 13.80 B0 IabW50b 16 47 01.17 −
45 50 26.7 22.8 19.66 17.21 14.69 O9 IIIW52 16 47 01.84 −
45 51 29.2 21.8 17.48 14.68 11.94 B1.5 Ia P(6.7d) W53 16 47 00.48 −
45 51 32.0 22.9 18.51 15.80 13.13 OB Ia + OB Ia P(1.30d) , SB2, very broad Pa linesW54 16 47 03.06 −
45 51 30.5 − − − −
B0.5 IabW55 16 46 58.40 −
45 51 31.2 21.6 17.67 15.25 12.67 B0 IaW56a 16 46 58.93 −
45 51 48.8 21.7 17.46 14.81 12.15 B1.5 IaW56b 16 46 58.85 −
45 51 45.8 22.8 18.88 16.36 13.76 O9.5 IbW57a 16 47 01.35 −
45 51 45.6 20.7 16.54 13.83 11.13 B4 Ia Pulsator? W57c 16 47 01.59 −
45 51 45.5 − − − −
WN7o WR PW60 16 47 04.13 −
45 51 52.1 22.8 18.50 15.96 13.28 B0 IabW61a 16 47 02.29 −
45 51 41.6 21.2 17.16 14.62 12.01 B0.5 Ia H α variable Table 2. continued.
ID RA (J2000) Dec (J2000) B V R I Spectral Type NotesW61b 16 47 02.56 −
45 51 41.6 22.7 18.59 16.00 13.31 O9.5 IabW62a 16 47 02.51 −
45 51 37.9 − − − −
B0.5 IbW63a 16 47 03.39 −
45 51 57.7 22.6 18.56 16.20 13.68 B0 IabW65 16 47 03.89 −
45 51 46.3 22.9 18.73 16.27 13.68 O9 IbW66 16 47 03.96 −
45 51 37.5 − −
45 50 49.6 21.2 16.88 14.10 11.29 B3 Ia H α variable W71 16 47 08.44 −
45 50 49.3 21.5 17.01 14.06 11.16 B2.5 Ia H α variable, Pulsator? W72 16 47 08.32 −
45 50 45.5 − W74 16 47 07.08 −
45 50 13.1 − − − −
O9.5 IabW75 16 47 08.93 −
45 49 58.4 − − − −
M4 IaW78 16 47 01.54 −
45 49 57.8 21.0 17.06 14.54 12.04 B1 Ia Pulsator? W84 16 46 59.03 −
45 50 28.2 21.3 17.82 15.60 13.63 O9.5 IbW86 16 46 57.15 −
45 50 09.9 22.9 18.76 16.43 14.00 O9.5 IbW228b 16 46 58.05 −
45 53 01.0 − − − −
O9 IbW232 16 47 01.41 −
45 52 34.9 21.3 17.53 15.25 12.85 B0 IabW237 16 47 03.09 −
45 52 18.8 22.8 17.49 13.00 9.19 M3 Ia Spec. variable W238 16 47 04.41 −
45 52 27.6 21.4 17.47 14.98 12.45 B1 IabW239 16 47 05.21 −
45 52 25.0 21.7 17.86 15.39 12.90 WC9d WR F, RV(5.05d) , SB1W241 16 47 06.06 −
45 52 08.3 − − − −
WC9 WR E, RV binary? W243 16 47 07.55 −
45 52 28.5 − − − −
LBV Spec. variable, pulsator W265 16 47 06.26 −
45 49 23.7 22.0 17.05 13.62 10.54 F1 ↔ + Spec. variable , pulsator W373 16 46 57.71 −
45 53 20.1 − − − −
B0 IabWR B 16 47 05.36 −
45 51 05.0 − , SB1WR C 16 47 04.40 −
45 51 03.8 − − − −
WC9dWR D 16 47 06.24 −
45 51 26.5 − − − −
WN7oWR G 16 47 04.01 −
45 51 25.2 22.7 20.87 17.75 14.68 WN7oWR H 16 47 04.22 −
45 51 20.2 − − − −
WC9dWR I 16 47 00.88 −
45 51 20.8 − − − −
WN8oWR J 16 47 02.47 −
45 51 00.1 − − − −
WN5hWR K 16 47 03.25 −
45 50 43.8 − − − −
WC8WR N 16 46 59.9 −
45 55 26 − − −
45 52 35.9 − − − −
WN6oWR Q 16 46 55.55 −
45 51 35.0 − −
45 47 58 − − − −
WC9dWR U 16 47 06.55 −
45 50 39.0 − − − −
WN6oWR V 16 47 03.81 −
45 50 38.8 − − − −
WN8oWR W 16 47 07.58 −
45 49 22.2 − − − −
WN6hWR X 16 47 14.1 −
45 48 32 − − − −
WN5o1001 16 46 49.20 −
45 53 10.0 − − − − O + O? very Pa broad lines1002 16 46 49.68 −
45 52 53.0 − O9-9.5 II + O? P(0.144d) , broad Pa lines1003 16 46 52.32 −
45 52 03.4 − − − −
O9-9.5 bin? broad Pa lines1004 16 46 53.52 −
45 53 00.2 − OeBe star1005 16 46 54.24 −
45 51 54.7 23.6 18.93 16.08 13.26 B0 Iab W30021006 16 46 54.48 −
45 53 30.1 − −
O9-9.5 III bin? P(0.127d) , broad Pa lines1007 16 46 54.96 −
45 50 06.0 − − − −
O9-9.5 III1008 16 46 55.44 −
45 51 54.4 23.1 19.73 17.11 14.40 O9.5 II1009 16 46 55.92 −
45 51 41.4 22.9 19.07 16.74 14.08 B0 Ib W20021010 16 46 55.92 −
45 52 10.2 − − − − O + O? very broad Pa lines1011 16 46 56.86 −
45 52 04.4 22.5 19.14 17.10 14.79 O + O? very broad Pa lines1012 16 46 56.95 −
45 50 56.0 23.8 20.36 17.78 14.97 O9-9.5 III bin? broad Pa lines1013 16 46 57.60 −
45 52 30.7 − − − − O + O? very Pa broad lines1014 16 46 57.81 −
45 51 19.8 23.3 19.58 17.40 15.03 O9-9.5 III bin? broad Pa, H α / He I double?1015 16 46 57.96 −
45 51 40.7 23.1 19.24 16.77 14.17 O9 III1016 16 46 58.08 −
45 52 46.9 − − − −
O9-9.5 III bin? broad Pa lines1017 16 46 58.23 −
45 50 33.9 23.5 19.91 17.44 14.94 O9-9.5 III bin? broad Pa lines1018 16 46 58.32 −
45 50 56.8 − − − −
O9.5 Iab CXO164658.2-4550561019 16 46 58.36 −
45 51 48.8 23.8 20.33 17.86 15.32 O9-9.5 III bin? broad Pa lines1020 16 46 58.48 −
45 52 27.1 21.8 18.45 16.38 14.19 O9-9.5 + O? broad Pa lines . S. Clark et al.: Stellar census of Wd1 19
Table 2. continued.
ID RA (J2000) Dec (J2000) B V R I Spectral Type Notes1021 16 46 58.78 −
45 54 31.9 − O9-9.5III bin P(4.43d), SB2, broad Pa lines1022 16 46 59.93 −
45 50 25.4 23.0 19.48 17.08 14.82 O9.5 II P(0.1703d) −
45 51 10.4 − − − −
O9 III1024 16 47 00.72 −
45 51 01.8 23.3 19.38 16.72 14.04 O9.5 Iab W20111025 16 47 00.76 −
45 52 04.6 22.9 19.91 17.65 15.26 O + O? very broad Pa lines1026 16 47 00.96 −
45 49 48.7 − − − −
O9-9.5 III1027 16 47 00.96 −
45 50 06.7 − − − −
O9.5 Iab CXO164701.0-4550061028 16 47 01.32 −
45 51 38.2 23.9 20.76 17.98 15.29 O9-9.5 III bin? broad Pa lines, H α / He I double?1029 16 47 01.44 −
45 49 50.2 − − − −
O9-9.5 III bin? broad Pa lines1030 16 47 01.69 −
45 52 57.8 − −
O9.5 Iab W30051031 16 47 01.92 −
45 50 56.4 − − − −
O9 III1032 16 47 02.27 −
45 50 17.6 24.0 21.18 18.01 15.38 O9-9.5 III bin? broad Pa lines, H α / He I double?1033 16 47 02.40 −
45 52 34.3 22.6 18.96 16.43 13.81 O9-9.5 I-III C07-X5, skewed lines?1034 16 47 02.52 −
45 51 48.2 22.5 18.69 16.34 13.85 O9.5 Iab1035 16 47 02.64 −
45 51 51.1 23.5 19.56 17.15 14.54 O9-9.5 III bin? broad Pa lines1036 16 47 02.77 −
45 52 12.4 23.5 19.29 16.69 14.09 O9.5 Iab C07-X4, W20171037 16 47 02.84 −
45 50 06.3 24.0 19.62 17.00 14.41 O9.5 II1038 16 47 03.60 −
45 48 57.2 − − − −
O9 III1039 16 47 03.60 −
45 49 21.7 − − − −
B1 Ia W30191040 16 47 04.52 −
45 50 08.8 21.9 18.89 16.37 13.76 O9-9.5 I-III bin? W2019, C07-X3, broad Pa lines1041 16 47 04.56 −
45 51 09.4 − − − −
O9.5 Iab bin? CXO164704.4-455109, broad Pa lines1042 16 47 04.56 −
45 52 06.6 22.8 19.17 16.80 14.27 O9.5 II1043 16 47 04.57 −
45 50 59.3 23.3 20.17 17.39 14.53 O9.5 II-III1044 16 47 05.56 −
45 49 51.5 24.1 20.28 17.85 15.36 O9-9.5 III bin? broad Pa lines, H α infilled?1045 16 47 05.82 −
45 51 54.9 23.2 19.77 17.20 14.49 O9.5 II1046 16 47 06.00 −
45 49 56.9 22.5 18.87 16.59 14.19 O + O? very Pa broad lines, He I double?1047 16 47 06.11 −
45 52 32.1 22.1 18.93 16.58 14.03 O9.5 II1048 16 47 06.26 −
45 51 03.9 23.8 20.50 16.47 13.80 B1.5 Ia P(5.2d) −
45 47 38.5 22.1 17.70 14.81 12.05 B1-2 Ia + −
45 49 55.3 23.1 19.52 16.97 14.44 O9.5 II1051 16 47 06.96 −
45 49 40.1 − −
O9 III1052 16 47 06.96 −
45 52 55.9 − − − −
O9 III1053 16 47 07.44 −
45 48 50.0 − − − −
B0 Ib1054 16 47 07.68 −
45 51 41.0 − − − −
O9-9.5 bin? broad Pa lines1055 16 47 07.92 −
45 51 47.9 − −
B0 Ib ( + O?) W3024, broad Pa lines1056 16 47 08.64 −
45 51 01.4 24.2 20.54 17.31 14.30 O9.5 II1057 16 47 08.67 −
45 50 47.1 23.3 19.39 16.60 13.71 O9.5-B0 Iab W2028, H α infilled, C II em.1058 16 47 08.88 −
45 51 24.5 − − − −
O9 III1059 16 47 09.12 −
45 53 20.8 − − − −
O9 III?1060 16 47 09.18 −
45 50 48.3 23.3 20.09 17.09 14.31 O9.5 II1061 16 47 09.74 −
45 50 40.2 23.9 20.91 18.16 15.42 O9-9.5 III bin? broad Pa lines1062 16 47 10.62 −
45 50 46.4 23.6 20.03 17.48 14.90 O + O? very broad Pa lines1063 16 47 10.80 −
45 49 47.6 − − − −
O9 III1064 16 47 11.52 −
45 49 59.5 − − − −
O9.5 Iab CXO164711.5-4550001065 16 47 11.60 −
45 49 22.4 23.1 19.08 16.19 13.28 B0 Ib W3003, RV(11.12d) , SB11066 16 47 12.72 −
45 50 55.3 − − − −
O9 III1067 16 47 13.39 −
45 49 10.5 23.2 18.93 15.88 12.98 B0 Iab W30041068 16 47 16.56 −
45 51 41.0 − − − −
B0Ib blended1069 16 47 24.24 −
45 53 29.0 − − − −
B5 Ia + Compilation of stellar classifications for members of Westerlund 1. Column 1 lists the primary optical identifier for the sourcesand columns 2 and 3 their co-ordinates. Columns 4-7 present B − , V − , R − and I − band photometry derived from the dataset de-scribed in Clark et al. (2005) and, if not available, Bonanos (2007; italics). Column 8 presents the spectral classification and whereappropriate column nine presents other designations (including the W xxx and W xxx designations previously employed but su-perceded by this work) and notes regarding spectral appearance and variability, with E clipsing and P eriodic photometric variablesand R adial V elocity spectroscopic variables listed with their relevant periods. The vast majority of classifications derive from thiswork, Clark et al. (2005), Crowther et al. (2006a) and Negueruela et al. (2010), with exceptions highlighted. Additional data usedin the construction of this table from Bonanos (2007), Ritchie et al. (2009a), Ritchie et al (2010), Ritchie et al. (2011), Clark etal. (2010), Clark et al. (2011) Koumpia & Bonanos (2012), Clark et al. (2013) Clark et al. (2014b) and Clark et al. (2019a).
Table 3.
Field stars
ID Spectral Type RA (J2000) Dec (J2000)F2 M0–2 bin 16 46 51.36 -45 50 10.3F3 B V 16 46 54.00 -45 53 06.4F4 M0–2 16 46 54.24 -45 49 42.6F5 G5–K5 16 46 55.92 -45 52 18.8F6 G5–K5 16 46 57.12 -45 51 36.0F1 M2 II-III 16 46 57.84 -45 52 18.5F7 K0–K5 16 47 07.44 -45 48 42.8F8 K0–K5 16 47 18.00 -45 49 43.7
Table 4.
Summary of the WR populations of Wd1 and theArches, Quintuplet and Galactic Centre clusters
Wd1 Arches Quint. Gal Cen.Sub-typeWN8-9ha 0 13 2 0WN5 3 0 0 0WN6 4 0 1 1WN7 5 0 0 3WN8 2 0 0 5WN9-11h 2 0 3 8WN / WC 0 0 0 2WC5 / / Acknowledgements.
We are indebted to Prof. Simon Goodwin for calculatingthe integrated cluster masses discussed in Sect. 4.2. This research is partiallysupported by the Spanish Government under grants AYA2015-68012-C2-2-Pand PGC2018-093741-B-C21 (MICIU / AEI / FEDER, UE), and made use of theSIMBAD database, operated at CDS, Strasbourg, France.
References
Abramowski, A., Acero, F., Aharonian, F. et al. 2012, A&A, 537, A114Aharonian, F. Yang, R., de ˜Ona Wilhelmi., E. et al. 2018, arXiv:1804.02331Andersen, M., Gennaro, M., Brandner, W. et al. 2017, A&A, 602, A22Andrews, H., Fenech, D., Prinja, R. K., Clark, J. S. & Hindson, L. 2018,MNRAS, 477, L55Beasor, E. R., Davies, B., Smith, N., & Bastian, N. 2019, MNRAS, 486, 266Bednarek, K., Pabich, J. & Sobczak, T. 2014, PhRvD, 90, 3008Beniamini, P., Hotokezaka, K. van der Horst, A. & Kouveliotou, C. 2019,arXiv:1903.06718Bibby, J. L., Crowther, P. A., Furness, J. P. & Clark, J. S. 2008, MNRAS, 386,L23Blum, R. D., Sellgren, K. & Depot, D. L. 1996, AJ, 112, 5Bonanos, A.Z., Stanek, K. Z., Udalski, A. et al. 2004, ApJ, 611, L33Bonanos, A.Z. 2007, AJ, 133, 2696Brandner, W., Clark, J. S., Stolte, A. et al. 2008, A&A, 478, 137Britavskiy, N., Lennon, D. J., Patrick, L. R. et al., 2019, A&A, 624, A128Brott, I., de Mink, S. E., Cantiello, M. et al. 2011, A&A, 530, A115Bykov, A. M., Ellison, D. C., Gladilin, P. E., & Osipov, S. M. 2015, MNRAS,453, 113Cardelli, J. A., Clayton, G. C. & Mathis, J. S., 1989, ApJ, 345, 245Cesarsky, C. J. & Montmerle, T. 1983, SSRv, 36, 173Clark, J. S., Fender, R. P., Waters, L. B. F.M., et al. 1998, MNRAS, 299, L43Clark, J. S. & Negueruela, I. 2004, A&A, 413, L15 Clark, J.S., Negueruela, I., Crowther, P.A. & Goodwin, S. 2005, A&A, 434, 949Clark, J.S., Muno, M.P., Negueruela, I., et al. 2008, A&A, 477, 147Clark, J. S., Negueruela, I., Davies, B. et al. 2009, A&A, 498, 109Clark, J.S., Ritchie, B.W. & Negueruela, I. 2010, A&A, 514, A87Clark, J.S., Ritchie, B.W.& Negueruela, I., et al. 2011, A&A, 531, A28Clark, J. S., Najarro, F., Negueruela, I., et al. 2012, A&A, 541, A145Clark, J. S., Ritchie, B. W. & Negueruela, I. 2013, A&A, 560, A11Clark, J. S., Negueruela, I. & Gonz´alez-Fern´andez, C. 2014a, A&A, 561, A15Clark, J. S., Ritchie, B. W., Najarro, F. Langer N. & Negueruela, I. 2014b, A&A,565, A90Clark, J. S., Lohr, M. E., Najarro, F., Dong, H. & Martins, F. 2018a, A&A, 617,A66Clark, J. S., Lohr, M. E., Patrick, L. R., et al. 2018b, A&A, 618, A2Clark, J. S., Najarro, F., Negueruela, I. et al. 2019a, A&A, 623, A83Clark, J. S., Lohr, M. E., Patrick, L. R. & Najarro, F. 2019b, A&A, 623, A84Clark, J. S., Ritchie, B. W. & Negueruela, I. 2019c, A&A, in pressCottaar, M., Meyer, M. R., Andersen, M. & Espinoza, P. 2012, A&A, 539, A5Crowther, P. A., Hadfield, L. J., Clark, J. S., Negueruela, I. & Vacca, W. D. 2006a,MNRAS, 372, 1407Crowther, P.A., Lennon, D. J. & Walborn, N. R. 2006b, A&A, 446, 279Crowther, P. A., Caballero-Nieves, S. M., Bostroem, K. A., et al. 2016, MNRAS,458, 624Davies, B., Figer, D. F., Kudrtizki, R.-F. et al. 2007, ApJ, 671, 781Davies, B., Figer, D. F., Kudrtizki, R.-F. et al. 2009, ApJ, 707, 844Davies, B., Crowther, P. A. & Beasor, E. R. 2018, MNRAS, 478, 3138de Mink, S.E., Sana, H., Langer, N., Izzard, R. G., Schneider, F. R. N. 2014, ApJ,782, 7Deneva, J. S., Cordes, J. M. & Lazio, T. J. W. 2009, ApJ, 702, L177Dexter, J., Degenaar, N., Kerr, M. et al. 2017, MNRAS, 468, 1486Dougherty, S.M., Clark, J.S., Negueruela, I., Johnson, T. & Chapman, J.M. 2010,A&A, 511, A58Ekstr¨om, S., Georgy, C., Eggenberger, P. et al. 2012, A&A, 537, A146Evans, C.J., Taylor, W.D., H´enault-Brunet, V., et al. 2011, A&A, 530, A108Fabregat, J. & Torrej´on, J.M. 2000, A&A, 357, 451Fenech, D. M., Clark, J. S., Prinja, R. K., 2017, MNRAS, 464, L75Fenech, D. M., Clark, J. S., Prinja, R. K. et al. 2018, A&A, 617, A137Figer, D. F., Najarro, F., Geballe, T. R., Blum, R. D. & Kudritzki, R. P. 2005,ApJ, 622, L49Figer, D. F., MacKenty, J. W., Robberto, M., et al. 2006, ApJ, 643, 1166Gennaro, M., Brandner, W., Stolte, A. & Henning, Th. 2011, MNRAS, 412, 2469Gennaro, M., Goodwin, S. P., Parker, R. J., Allison, R. J. & Brandner, W. 2017,MNRAS, 472, 1760Gies, D. R. 2003, in A Massive Star Odyssey: From Main Sequence toSupernova, ed. K. van der Hucht, A. Herrero, & C. Esteban, IAU Symp.,212, 91Gonz´alez-Fer´andez, C., Dorda, R., Negueruela, I. & Marco, A. 2015, A&A, 578,A3G¨otberg, Y., de Mink, S. E., Groh, J. H. et al. 2018, A&A, 615, A78Groh, J. H., Hillier, D. J., Damineli, A. et al. 2009, ApJ, 698, 1698Groh, J. H., Meynet, G., Georgy, C. & Ekstr´om, S. 2013, A&A, 558, A131Groh, J. H., Meynet, G., Ekstr¨om, S. & Georgy, C. 2014, A&A, 564, A30Gvaramadze, V. V. 2018, RNAAS, 2, 214Hillier, D. J., Crowther, P. A., Najarro, F. & Fullerton, A. W. 1998, A&A, 340,483Hillier, D.J. & Miller, D.L. 1998b, ApJ, 496, 407Hosek, M. W., Lu, J. R., Anderson, J. et al. 2018, ApJ, 855, 13Jenniskens, P. & Desert, F.-X. 1994, A&AS, 106, 39Kasen, D. & Bildsten, L. 2010, ApJ, 717, 245Kennea, J. A., Burrows, D. N., Kouvelioutou, C. et al. 2013, ApJ, 770, L24Koumpia, E. & Bonanos, A.Z. 2012, A&A submitted [arXiv:1205.1369]Kudryavtseva, N., Brandner, W., Gennaro, M., Rochau, B., Stolte, A., et al. 2012,ApJ, 750, L44Lim, B., Chun, M., Sung, H. et al. 2013, AJ, 145, 46Lohr, M. E., Clark, J. S., Najarro, F. et al. 2018a, A&A, 617, A66Lohr, M. E., Negueruela, I., Tabanero, H. M., et al. 2018, MNRAS, 478, 3825Mahy, L., Damerdiji, Y., Gosset, E. et al. 2017, A&A, 607, A96Martins, F., Trippe, S., Paumard, T. et al. 2006a, ApJ, 649, L103Martins, F. & Plez, B. 2006b, A&A, 457, 637Martins, F., Genzel, R., Hillier, D. J. et al. 2007, A&A, 468, 233Martins, F. & Palacios, A. 2017, A&A, 598, A56Maschberger, T. 2013, MNRAS, 429, 1725Massey, P., DeGioia-Eastwood, K. & Waterhouse E. 2000, AJ, 119, 2214Massey, P., DeGioia-Eastwood, K. & Waterhouse E. 2001, AJ, 121, 1050Mauerhan, J. C., Muno, M. P., Morris, M. R., Stolovy, S. R. & Cotera, A. 2010,ApJ, 710, 706Melena, N. W., Massey, P., Morrell, N. I. & Zangari, A. M. 2008, AJ, 135, 878Meynet, G. & Maeder, A. 2003, A&A, 404, 975Mori, K., Gotthelf, E. V., Zhang, S. et al. 2013, ApJ, 770, L23 . S. Clark et al.: Stellar census of Wd1 21
Muno, M. P., Clark, J. S., Crowther, P. A. et al. 2006a, ApJ, 636, L41Muno, M. P., Law, C., Clark, J. S., et al. 2006b, ApJ, 650, 203Najarro, F., 2001, ASPC, 233, 133Negueruela I., Steele, I. A. & Bernabeu, G. 2004, 325, 749Negueruela, I., Clark, J. S. & Ritchie, B. W. 2010, A&A, 516, A78Negueruela, I. & Marco, A. 2012, ApJ, 143, 46Ohm, S., Hinton, J. A. & White, R. 2013, MNRAS, 434, 2289Orosz, J. A., McClintock, J. E., Aufdenberg, J. P. et al. 2011, ApJ, 742, 84Palmer, D.M. 2009, ApJ, 695, 496Pamyatnykh, A.A. 1999, AcA, 49, 119Pasquani, L., Avila, G., Blecha, A., et al. 2002, The Messenger, 110, 1Paumard, T., Genzel, R., Martins, F. et al. 2006, ApJ, 643, 1011Petrovic, J., Langer, N. & van der Hucht, K.A. 2005, A&A, 435, 1013Ramsey, L.W. 1981, AJ, 86, 557Ritchie, B.W., Clark, J.S., Negueruela, I. & Crowther, P.A 2009a, A&A, 507,1585Ritchie, B.W., Clark, J.S., Negueruela, I. & Najarro, F. 2009b, A&A, 507, 1597Ritchie, B.W., Clark, J.S., Negueruela, I. & Langer, N. 2010, A&A, 520, A48Ritchie, B.W., Clark, J.S. & Negueruela, I. 2011, BSRSL, 80, 628Roman-Lopes, A., Franco, G. A. P. & Sanmartim, D. 2016, ApJ, 823, 96Sana, H., de Mink, S. E., de Koter, A. et al. 2012, Science, 337, 444Schneider, F. R. N., Izzard, R. G., de Mink, S. E., et al. 2014, ApJ, 780, 117Schneider, F. R. N., Ram´ırez-Agudelo, O. H., Tramper, F. et al. 2018, A&A, 618,A73Schnurr, O. Mo ffff