Constraining the population of isolated massive stars within the Central Molecular Zone
J. S. Clark, L. R. Patrick, F. Najarro, C. J. Evans, M. E. Lohr
aa r X i v : . [ a s t r o - ph . GA ] F e b Astronomy&Astrophysicsmanuscript no. arxiv © ESO 2021February 17, 2021
Constraining the population of isolated massive stars within theCentral Molecular Zone. ⋆ J. S. Clark , L. R. Patrick , , F. Najarro , C. J. Evans , and M. Lohr Department of Physics and Astronomy, The Open University, Walton Hall, Milton Keynes, MK7 6AA, United Kingdom Instituto de Astrof´ısica de Canarias, E-38205 La Laguna, Tenerife, Spain DFISTS, EPS, Universidad de Alicante, Carretera San Vicente del Raspeig s / n, E-03690 San Vicente del Raspeig, Spain Departamento de Astrof´ısica, Centro de Astrobiolog´ıa, (CSIC-INTA), Ctra. Torrej´on a Ajalvir, km 4, 28850 Torrej´on de Ardoz,Madrid, Spain UK Astronomy Technology Centre, Royal Observatory Edinburgh, Blackford Hill, Edinburgh EH9 3HJ, UKPreprint online version: February 17, 2021
ABSTRACT
Context.
Many galaxies host pronounced circumnuclear starbursts, fuelled by infalling gas. Such activity is expected to drive thesecular evolution of the nucleus and generate super winds, while the intense radiation fields and extreme gas and cosmic ray densitiespresent may act to modify the outcome of star formation with respect to more quiescent galactic regions.
Aims.
The centre of the Milky Way is the only example of this phenomenon where, by virtue of its proximity, individual stars may beresolved. Previous studies reveal that it hosts a rich population of very massive stars; these are located within three clusters, with anadditional cohort dispersed throughout the Central Molecular Zone. In this paper we investigate the size and composition of the lattercontingent.
Methods.
We utilised the VLT + KMOS to obtain homogeneous, high signal-to-noise ratio observations of known and candidatemassive stars suitable for spectral classification and quantitative analysis.
Results.
Including previously identified examples, we find a total of 83 isolated massive stars within the Galactic Centre, which arestrongly biased towards objects supporting powerful stellar winds and / or extensive circumstellar envelopes. No further stellar clusters,or their tidally stripped remnants, were identified, although an apparent stellar overdensity is found to be coincident with the Sgr B1star forming region. Conclusions.
The cohort of isolated massive stars within the Central Molecular Zone is directly comparable in size to that of theknown clusters and, due to observational biases, is likely highly incomplete at this time. Combining both populations yields & ∼ Key words. stars:evolution - stars:massive - Galaxy:nucleus
1. Introduction
While many galaxies host prominent (circum-)nuclear star-bursts, the physics governing their formation and subsequentcontribution to the wider galactic ecology and energy budgetis currently opaque due to our inability to resolve individualstars in such environments at extragalactic distances. Indeedthere is only one example of this phenomenon - the centralregion of our own Galaxy - where this is currently possible.Multiwavelength observations have revealed that physical con-ditions in the Galactic Centre (GC) are particularly extreme withrespect to the disc, bearing close resemblance to those antici-pated for high redshift starburst galaxies (Kruijssen & Longmore2013), hence the hope that the GC will act as a template for suchobjects. Specifically, the mean temperature, density, pressure,and velocity dispersion of molecular material; the magnetic fieldstrength; the cosmic ray density; and the ionisation rate are sig-nificantly greater than those found in the Galactic disc, in somecases by orders of magnitude. As such, one might anticipate that ⋆ Based on observations made at the European Southern Observatory,Paranal, Chile under programmes ESO 093.D-0168 processes such as star formation proceed in a di ff erent mannerthan in more quiescent regions of the Milky Way.It is important to determine if this is the case. Encompassingthe inner ∼ ∼
10% of the molecular mass of the Galaxy( ∼ − × M ⊙ ; Morris & Serabyn 1996), which in turn fu-els the most extreme star forming region within the Milky Way.Observations from sub-mm to radio wavelengths suggest this ac-tivity is occurring at multiple locations within the H ii regionspopulating the GC; within ∼ ∗ (Yusef-Zadeh etal. 2009), Sgr B1 + ∼ ii region issuggestive of cluster formation (Yusef Zadeh et al. 2009); a con-clusion that is buttressed by sub-mm observations of the Sgr B2 J. S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone. region, which imply a combination of clustered and distributedstar formation (Ginsburg et al. 2018a, Ginsburg & Kruijssen2018b).However, despite this vigorous activity and presence of co-pious molecular material, estimates of the star formation ratefor the CMZ suggest that it is at least an order of magnitudelower than expected based on observations of nearby regions(Longmore et al. 2013). The physical cause of this discrepancyis uncertain (Barnes et al. 2017), but it raises the possibility thatthe resultant stellar population(s) may also show an environmen-tal dependance, possibly characterised by an anomalous initialmass function (IMF). As a consequence, much e ff ort has beenexpended in attempts to characterise the young stellar popula-tion within the CMZ, with near-IR observations revealing a richpopulation of massive stars located within clusters - the Arches,Quintuplet, and Galactic Centre (e.g. Figer et al. 1999, Paumardet al. 2006) - and distributed throughout the CMZ in apparentisolation (Cotera et al. 1996, 1999, Muno et al. 2006a, Mauerhanet al. 2007, 2010b, 2010c, Dong et al. 2015).Beyond constraining the star formation physics operatingwithin the extreme conditions of the GC, a determination of theproperties of this stellar cohort is of considerable importancefor a number of other astrophysical topics. With a subset bornwith masses M init > M ⊙ (Lohr et al. 2018), they provide vi-tal observational data on the lifecycle of the most massive starsthat form in the local Universe, up to and including the pointof core-collapse. Constraining a robust evolutionary scheme forsuch stars is essential if we are to predict both the nature and pro-duction rate of relativistic remnants from this population, notingthat a rich cohort of young neutron stars and black holes deriv-ing from such stars appears present within the GC (Deneva et al.2009, Kennea et al. 2013, Hailey et al. 2018).It is anticipated that massive stars also play an important rolein driving the evolution of the GC via the feedback of ionisingradiation, mechanical energy and chemically enriched material.Of particular interest is their role in shaping the emergent highenergy spectrum of the GC, which recent observations suggestextends from soft X-rays ( kT ∼ − γ -rays ( kT > ff use, low energy X-ray emission arises from a combination of unresolved low masspoint sources (pre-MS stars and Cataclysmic Variables), thewinds of massive stars and their supernova (SN) endpoints - act-ing both individually and in concert in massive clusters such asWd1 (Muno et al. 2006b) - and pulsar wind nebulae (Ponti et al.2015).The γ -ray component is thought to derive from the cosmicrays that permeate the GC (Aharonian et al. 2006, H.E.S.S. col-laboration 2016, 2018). An exceptional cosmic ray density maybe inferred by the abundance of H + (produced via the ionisationof H ) and has been suggested to play an important role in reg-ulating the temperature of the warm molecular material that suf-fuses the CMZ (Le Petit et al. 2016, Oka et al. 2019). Plausiblesources for the production of cosmic rays are the supermassiveblack hole Sgr A ∗ and massive stars - the latter via the interac-tion between their winds, cluster driven outflows and supernovae(e.g. Aharonian et al. 2018, Bednarek et al. 2014, Bykov et al.2015, Cesarsky & Montmerle 1983). Unfortunately, in the ab-sence of a full stellar census the relative contributions of thesechannels is currently uncertain.Nevertheless both physical agents have been implicated inthe initiation of mass outflows - thought to be driven by a com-bination of cosmic ray and thermal gas pressure (cf. Everettet al. 2008, Yusef-Zadeh & Wardle 2019) - that originate in the GC. These range in size from the ∼ ∼
2. Data acquisition, reduction, and classification
The VLT-KMOS (Sharples et al. 2013) data for this paper wereobtained under ESO programme 093.D-0306 (PI: Clark), withobservations made between 2014 August 02-13. KMOS is amulti-object, integral field spectrograph, which has 24 config-urable integral field units (IFUs) positioned within a 6.7 ar-cminute field of view. The spectral resolution of the observationsis a function of rotator angles and the IFUs used (e.g. Patrick etal. 2015), varying between ∆ λ/λ ∼ − ×
30 s exposures in an ABA observing pat-tern, where the first observation of each field used the more rig-orous 24-arm telluric standard star approach and all subsequentobservations of the same field used the standard 3-arm telluricapproach. The standard stars used for these observations wereHIP 84846 (A0V), HIP 91137 (A0V), and HIP 3820 (B8V).The data reduction methodology is identical to that of theKMOS data presented in Clark et al. (2018b). Science and stan-dard star observations were calibrated, reconstructed and com-bined using the KMOS / esorex pipeline (Davies et al. 2013), em-ploying the standard set of calibrations delivered by the tele-scope. Clark et al. (2018b) detail this procedure and discuss themodifications made to the standard processes.In the K-band, telluric correction is a fundamentally impor-tant part of the data reduction process. Since the majority ofthe useful diagnostic lines for these targets lie in regions of theK-band that are highly contaminated by telluric absorption, weimplemented a rigorous correction routine adapted from Patricket al. (2015, 2017) and further detailed in Clark et al. (2018b).Given the intrinsic shape of the telluric spectrum, the continuumplacement is vital to accurately recover the shape of the sciencespectrum. This is typically done empirically, by selecting multi- . S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone. 3 ple continuum points from the science and standard star spectrathroughout the entire spectral range and is highly non-linear.For targets with particularly broad spectral features, such asthe WN and WC stars, identification of the continuum is a com-plicated process. This problem is compounded when broad emis-sion features coincide with strong telluric absorption, as seenfor the He ii µ m and 2.3799 µ m and He i µ m featuresin the WN5-7 stars. For such stars continuum placement wasguided by comparison with a combination of published spec-tra - in particular that of qF353E (WN6; Steinke et al. 2016) -and synthetic examples computed with from the CMFGEN code(Hillier & Miller 1998, 1999). A number of publications have been dedicated to the classifica-tion of post-MS massive stars in the near-IR window: specifi-cally O stars (Hanson et al. 1996, 2005), B-hypergiants (Clark etal. 2012, 2018b), luminous blue variables (LBVs; Morris et al.1996, Clark et al. 2011) supergiant B[e] stars (sgB[e]; Oksala etal. 2013) and Wolf-Rayets (WRs; Figer et al. 1997, Crowther etal. 2006, Crowther & Walborn 2011 and Rosslowe & Crowther2018).We have employed - and expanded upon - these classifica-tion criteria in our study of the Arches and Quintuplet (Clarket al. 2018a, 2018b). We follow an identical methodology here,referring the reader to these works for details beyond those sum-marised in the relevant sections below. As in previous works,given the uncertainty in the parameterisation of the spatially in-homogeneous interstellar extinction along sightlines towards theGC, we prioritise spectral rather than photometric data. Howevergiven that we are unable to utilise cluster membership to locatetarget stars within the GC, we are forced to employ the photo-metric datasets and analysis of Dong et al. (2012) to identifylikely foreground interlopers. In doing so we choose to onlymake use of ground based photometry to avoid issues of cali-brating such data with space-based observations, given the sig-nificant issues accounting for convolving very di ff erent filter re-sponses for intrinsically red photometric sources (cf. Dong et al.2012).For candidate massive stars previously identified in the liter-ature but without KMOS observations we used published spec-tra to reappraise their classifications in light of this methodol-ogy. Where stars are reclassified on this basis, if we were unableto obtain the relevant spectra we provide the appropriate figurenumber in addition to the formal reference in the following dis-cussion.In the remaining sections for conciseness, we abbreviate the[DWC2011] xxx designation for stars in the primary list of Pa α emitters presented in Dong et al. (2011) to a simple P xxx . Norecognised nomenclature exists for those stars derived from thesecondary list of Pa α emitters from this paper; hence we chooseto designate these simply as S xxx sources.
3. Results
Observations were made of a total of 82 candidate and confirmedmassive stars derived from the list of Pa α excess sources of Donget al. (2011) and other literatures sources (Mauerhan et al. 2007,2010b, 2010c). Details of each target, including previous andnew classifications are provided in Table 1. The sample containsa diverse group of objects including pre- and post-main sequencemassive stars as well as a large number of foreground high- and low-mass interlopers. Below we break down this population byspectral sub-types and location along the sightline to the GC,including discussion of relevant examples not included in oursample in order to provide the basis for the construction of acomprehensive stellar census. Spectra of OB supergiants are presented in Fig. 1. We are ableto identify three new candidates; S73, S152, and P95. The firsttwo stars are clearly mid-O supergiants given the presence ofHe ii µ m absorption and C iv + µ m emission.The lack of a pronounced He i ∼ . µ m absorption feature inthe He i + N iii + C iii blend of either star indicates comparativelyearly spectral types (O4-5), while the relatively weak Br γ pho-tospheric line signals significant mass loss, though not su ffi cientto drive the line into emission as is seen in hypergiants (Fig.2). Three additional mid-O supergiants - CXOGC J174628.2-283920, 174703.1-285354, and 174725.3-282523 - were ob-served by Mauerhan et al. (2010c). These were not observedwith KMOS, but we include their published classifications inTable 1 in order to compile a comprehensive census of massivestars in the GC region.Moving to later spectral sub-types and the presence of nar-row photospheric absorption in He i µ m,Br γ and He ii µ m indicates that P95 is a new ∼ O9 super-giant via close similarity to the spectral template HD154368(Hanson et al. 2005); however the narrower Br γ profile andstronger He i µ m emission suggests a stronger wind thanthis classification standard. Given the S / N of our spectrum ofCXOGC J174537.3-285354 around 2.19 µ m, we may not im-prove on the previous O9-B0Ia classification, nor reassess thenature of P50 from the spectrum presented in Mauerhan et al.(2010c). Geballe et al. (2019; their Fig. 6) identify 2MASSJ17444501-2919307 as B2-3 Ia + ; we prefer a slightly more con-servative B0-3 Ia classification. Finally De Witt et al. (2013) pro-pose a generic O star classification for XID 947; given the lowS / N of the published spectrum we are unable to improve on this.
Consideration of the spectra of members of the Arches revealsthe close evolutionary and morphological similarities between ∼ O4-8 hypergiants and WN7-9ha Wolf-Rayets (Martins et al.2008, Clark et al. 2018a). The former are delineated by system-atically weaker Br γ emission and a P Cygni absorption com-ponent in the ∼ . µ m He i + C iii + N iii + O iii emission blend atlater (O6-8) spectral subtypes. In contrast no absorption compo-nent is present in the ∼ . µ m feature of any WNLha star, whileHe ii µ m is in emission in the earlier ( < WN7-8) spectralsubtypes.As a consequence we discuss both types of star here; pre-senting spectra of three new examples in Figs. 2 and 3. Of theseP15 and S131 are clearly new O hypergiants by virtue of broad,pronounced Br γ emission and He ii µ m absorption. StrongC iv emission and a lack of He i µ m absorption indicatesthat P15 is an early O4-5 Ia + star (Clark et al. 2018a, Hansonet al. 2005). Conversely the presence of He i µ m absorp-tion and weak C iv emission in the spectrum of S131 indicates alater (O7-8 Ia + ) subtype; we examine the possible causes of thedouble peaked Br γ emission line profile below.Of those stars with previous classifications, the detection ofBr γ emission in P36 and 114 marks them out as hypergiants J. S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone.
Fig. 1.
Montage of spectra of OB supergiants. Template spectra with classifications from Hanson et al. (2005) and Clark et al.(2018a) overplotted in red. It is important to note that hot pixels around 2.080 µ m - nearly coincident with the C iv µ m lines- in the spectrum of CXOGC J174537.3-285354 were artificially removed; we suspect the anomalously narrow component of theBr γ photospheric profile is also spurious. Likewise the broad absorption feature centred on ∼ . µ m in the spectrum of P95 is alsoartificial.rather than supergiants and of early (O4-5) spectral type, giventhe lack of He i µ m absorption and consequent similarityto the O4-5 Ia + Arches star F27 (Fig. 2). The Br γ emission linein the spectrum of P36 shows a central reversal, which is alsopresent in the O4 Ia + spectroscopic template HD15570 and otherhypergiants considered here (see below); unfortunately the pres-ence of a narrow emission component of uncertain origin pre-vents interpretation of the corresponding line profile of P114.The similarity of P100 & 107 to Arches F10 suggests a revi-sion to slightly later spectral subtypes (O4-6 to O7-8; Fig. 2);the strength of Br γ emission in both stars suggest they are closeto transitioning to a WNLha evolutionary phase. In terms of thestrength of C iv emission P97 is intermediate between these starsand P15; we assign an O6-7 subtype by comparison to ArchesF15 (Fig. 2).The similarity of P23 to P97 suggests a comparable classi-fication, although the former demonstrates stronger Br γ emis-sion, suggesting it too is close to becoming a WNLha star.As with S131 its Br γ profile is strongly double peaked, whileC iv µ m and the emission component of the ∼ . µ mHe i + C iii + N iii + O iii blend are unexpectedly broad. Regardingthe Br γ line, the only comparator we are aware of is theQuintuplet member LHO 001 which is spectroscopically vari-able and demonstrates a similarly double peaked profile at someepochs. Clark et al. (2018b) suggest that LHO 001 is a massivebinary system and such an explanation is also attractive for P23as well; it is not obvious that a physically justifiable combina-tion of He-abundance, mass-loss rate, wind clumping factor and velocity field for a single star can replicate the Br γ line profileobserved.The Br γ profile of P75 also appears double peaked, al-though the blue peak is less pronounced than in P23, beingmore comparable to S131. Other notable features include strong,broad emission with hints of substructure in the ∼ . µ mHe i + C iii + N iii + O iii blend, an absence of C iv emission and,uniquely, weak N iii µ m emission. The latter two obser-vational features are characteristic of the WN8-9ha stars ratherthan the mid-O hypergiants within the Arches (Clark et al.2018a), although the reverse is true for the He i µ m absorp-tion component also exhibited by P75. We suggest this uniquehybrid morphology is due to strong helium and nitrogen en-hancement (with the former yielding pronounced He i µ memission in the blue wing of Br γ ) and C depletion with respect tonormal mid-O hypergiants as the star enters the WNLha phase.As such we revise the classification of P75 to WN9ha / O6-7Ia + ; noting that further multi-epoch observations and quantita-tive analysis are required to confirm the nature of stars such asP23, P75 and S131, which demonstrate double peaked Br γ pro-files.Next we turn to the WNLha stars (Fig. 3). Comparison ofS120 to Arches F16 suggests that it is a new WN8-9ha star; thestrength of Br γ emission and lack of He i µ m absorptiondistinguishing it from an O hypergiant, while He ii µ m fullyin absorption suggests a late spectral sub-type. Neverthelessthere are incongruities; the Br γ line and the He i + C iii + N iii + O iii ∼ . µ m emission blend both appear anomalously narrow in . S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone. 5 comparison to other isolated examples and the cohort withinthe Arches cluster (Clark et al. 2018a). Moreover, the emissionfeatures exhibit a significant displacement from their rest wave-lengths ( ∆ RV & − ); possibly indicative of binary reflexmotion or a runaway nature.Assigned a generic O If + classification (Muno et al. 2006a,Dong et al. 2015) P35 is of particular interest since it is spa-tially coincident with the H ii region H2, which Dong et al.(2017) associates with an apparent overdensity of bright stars.Inspection of our spectrum reveals exceptionally strong, narrowand asymmetric He i µ m and Br γ emission, the latter witha rather broad base (Fig. 3). Such a morphology is not char-acteristic of O super- / hypergiants or WNLha stars. Conversely,weak C iv µ m emission, He ii µ m absorption and astrong broad pure emission profile in the He i + C iii + N iii + O iii ∼ . µ m blend is reminiscent of early-mid O hypergiants andweak-lined WN8-9ha stars. We therefore assign such a classifi-cation to P35, assuming that there is significant contamination ofthe He i µ m and Br γ profiles by nebular emission from theH2 H ii region - as suggested by the inflection in the red flankof both lines. Finally, as discussed in Clark et al. (2019b) P96closely resembles the WN7-8ha Arches member F4 (Table 1);the resolution and S / N of the published spectra of the remainingcandidates (cf. Mauerhan et al. 2010b, 2010c) being insu ffi cientto allow any further refinements to current classifications.Unlike the OB supergiants there exists su ffi cient quasi-homogeneous ground based photometry (Mauerhan et al. 2010b,Dong et al. 2011) to enable a comparison of the properties of iso-lated WN7-9ha and O hypergiants to those in the Arches clus-ter. Fig. 4 indicates that such stars within the Arches exhibit arange in both K − band ( ∼ . −
11) and ( H − K ) colour indice( ∼ . − . / bolometric correctionand continuum emission from the stellar wind) and extrinsicproperties (binarity, di ff erential stellar reddening); indeed Lohret al. (2018) find [FGR2002] 2 to be an exceptionally luminousstar with its apparent magnitude due to extreme reddening.Comparison to the photometric properties of the correspond-ing population of isolated WN7-9ha and O hypergiants shows anencouraging co-location in the colour / magnitude plot, althoughwith an increased proportion of fainter, redder examples. This islikely indicative of greater interstellar reddening along the rele-vant lines of sight, although verification awaits a parameterisa-tion of the reddening law towards the GC. The newly identifiedWN8-9ha star S120 appears an exception to this trend, being thefaintest example ( K ∼ .
5) but with a rather moderate near-IRcolour (( H − K ) ∼ . We present the spectra of WN9-11h stars and early-B hyper-giants in Figs. 5 and 6. Given the large number of such starswithin the Quintuplet cluster compared to those known in thewider galaxy (particularly early-B hypergiants) we utilise theseas classification templates following the discussion in Clark etal. (2018b). In doing so we are able to revise the classifica-tion of P103 from generic P Cyg O-type supergiant to B0-1Ia + / WNLh (Fig. 5). Interpreting the spectrum of P56 - plottedagainst the WN11h star LHO71 in Fig. 6 - is more di ffi cult.Both stars clearly show prominent electron scattering wings inthe P Cygni profile of He i µ m, while their Br γ emissionlines are broadly comparable. However the He i µ m dou-blet is in absorption in P56 but in emission in LHO71; given this discrepancy we suggest that a classification as either B0-1Ia + or WN11h would be appropriate (cf. P103). Likewise P98and P137 undergo less dramatic revisions to B1-2 Ia + / WNLhand WN10h respectively, with P19 found to be a twin of thebroad lined WN9h Quintuplet member LHO 158 (cf. Clark et al.2019b). Of these we note that Hankins et al. (2020) report P137is located within a mid-IR ring nebula, possible indicative of awind blown bubble or circumstellar ejecta (cf. the Pistol star).The combination of He i µ m and Mg ii / µ memission and narrow Br γ and He i µ m absorptionin the spectrum of S132 indicates that it is a new early- B hyper-giant (Fig. 5 and Table 1). Comparison of the spectrum of SSTUJ174523.11-290329.3 - which also demonstrates Mg ii emission(Mauerhan et al. 2007; their Fig. 4) - to those of S132 and similarobjects within the Quintuplet suggests that an identification as ahypergiant, rather than the previous supergiant classification, ismore appropriate. Finally, despite the low S / N and resolution ofthe spectrum of the previously unclassified 2MASS J17461292-2849001 (Geballe et al. 2019; their Fig. 6), it appears directlycomparable to the preceding two stars. As a consequence weadopt a similar B1-3 Ia + classification for it; further strength-ened by its close proximity to - and hence potential membershipof - the Quintuplet, which hosts a large number of such stars(Clark et al. 2018b).Fig. 4 illustrates the near-IR photometric properties ofboth isolated early-B hypergiants and WN9-11h stars andtheir counterparts within the Quintuplet cluster. The major-ity of Quintuplet members occupy a relatively compact regionof colour / magnitude space (9 . K .
10 and 1 . . ( H − K ) . . .Outliers include the faint ( K ∼ .
5) WN9h star LHO 158 andthe extremely red WN10h outlier LHO 67.The former is likelythe hottest of this cohort (and hence may require the largest bolo-metric correction) while an understanding of the latter - intrinsicIR excess and / or extrinsic reddening - awaits detailed quantita-tive analysis.While five of the isolated early-B hypergiants and WN9-11h stars are co-located with Quintuplet members in thecolour / magnitude plot, four are outliers. Both P19 (WN9h) andSSTU J174523.11-290329.3 (B0-2 Ia + ) appear rather faint andred and likely su ff er excess interstellar reddening (cf. Arches F2;Sect 3.2). Conversely the bright WN10h star P137 may either beseen through a window su ff ering reduced extinction or is a fore-ground object. Finally despite being over a magnitude brighterthan any other cluster or isolated early-B hypergiant observed todate, the ( H − K ) colour of S132 is unexceptional, suggestingthat it may be intrinsically highly luminous. Geballe et al. (2019; their Figs. 5 and 6) report on a cohortof ten stars with spectra dominated by strong Br γ and weakeremission in He i µ m and various Fe ii transitions. Of these2MASS J17452861-2856049 and J17462830-2839205 corre-spond to P35 (WN8-9ha + neb) and CXOGC J174628.2-283920(O4-6 Ia) respectively (Table 1), while 2MASS J17455154- An identical lower bound to the colour ( H − K ) index is also seenfor the Arches cluster. This is substantially in excess of the limit of( H − K ) & . H − K ) & . P56, 98 & 137, 2MASS J17461292-2849001 and WR102ka J. S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone.
Fig. 2.
Montage of spectra of isolated O hypergiants. Template spectra are overplotted in red along with identifications; we alsoshow the spectrum of F16, the WN8-9ha star with the weakest emission lines within the Arches (Clark et al. 2018a), for comparisonto P100 and 107. We note that a narrow emission feature of uncertain origin at line centre of Br γ in the spectrum of P114 wasremoved, leaving an artificially flat topped profile.2900231 is source D of Cotera et al. (1999; Sect 3.8). This leavesa total of seven objects of which, fortuitously, we have observa-tions of two - P112 and 141 ( = / N and reso-lution. These data are presented in Fig. 7, along with the spec-trum of a third object of similar morphology - P40 ( = H − K ) < .
0, with thelatter su ffi ciently blue that a foreground nature cannot be ex-cluded (Table 1). Five of the remaining objects cluster between( H − K ) ∼ . − . H − K ) ∼ .
6; values significantly in excess of those of otherearly-type massive stars in the GC (Table 1 and Fig. 4). Pre-empting the following discussion, in the absence of a classifica-tion yielding ‘photospheric’ colours for this cohort it is impos-sible to quantitatively decouple intrinsic and extrinsic contribu-tions to reddening, although it seems highly likely that a signifi-cant continuum contribution from a wind or dusty circumstellar disc is present in these six stars. Such a conclusion is supportedby mid-IR observations, with a number of stars appearing in-trinsically red upon consideration of the [4 . − [8 . > . ff ecting the near-IR photometry precludes us fromemploying the colour-magnitude and colour-colour plots of, forexample, Bonanos et al. (2009) in order to determine the physi-cal nature of these stars.Mindful of these issues we now turn to the spectra of P40,112 and 141. All three are dominated by strong, narrow Br γ emission. Critically, broad electron scattering wings are alsoseen in the profile of this transition in P40 and 112; indica-tive of a dense stellar wind and hence a massive star identi-fication. Weak He i µ m emission is also apparent in thespectra of both these stars, suggesting they are hotter than P141(where it is absent), although no trace of the He i µ m fea-ture is present in any of the three, nor are the spectral sig- P40, 112, & 141, 2MASS J17444319-2937526 and J17470940-2849235.. S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone. 7
Fig. 3.
Montage of spectra of WNLha stars. Template spectra with classifications overplotted in red (Clark et al. 2018a). We note thatS120 appears to have an anomalous RV redshift of ∼ + − ; the spectrum of F16 has been artificially shifted by a comparableamount to aid in comparison.natures of high excitation species such as He ii , N iii or C iv .Instead the remaining emission features present arise from lowexcitation metal transitions such as Fe ii µ m (P40 & 112)and 2.089 µ m (all stars), [Fe ii ] 2.045 µ m, 2.118 µ m, and 2.133 µ m(P141), Mg i + µ m (P112) and Na i + µ m(P40). At longer wavelengths the CO bandheads are seen inemission in P40 (Fig. 8); they and the Pfund series are absentfrom both P112 and P141. Finally, there is no indication of H emission in any of the stars, disfavouring a pre-MS classifica-tion.The narrow emission line spectra dominated by Br γ andlow excitation metals are reminiscent of both cool-phase LBVs(Clark et al. 2011, 2018b) and supergiant B[e] stars (sgB[e];Oksala et al. 2013). CO bandhead emission is present in a sub-stantial number of sgB[e] stars but appears absent from most, ifnot all, LBVs (Morris et al. 1996, Oksala et al. 2013). We presentthe spectra of the LBVs FMM362 and G24.73 + / magnitude plot (Fig. 4) andappear to show a comparable spectral morphology (subjectto the low resolution and S / N data) it is tempting to apply a similar classification to them. While the same is true for2MASS J17482472-2824313, we are more cautious in thiscase due to a possible association with a cold, dusty clump(Contreras et al. 2013) which could favour a pre-MS status.Likewise the comparatively blue ( H − K ) colour for 2MASSJ17470940-2849235 leaves open the possibility of a lower massforeground (post-AGB) object. Indeed, it is entirely possiblethat this cohort could be rather heterogeneous - comprisingstars of di ff erent luminosities and evolutionary status but similargross observational features. This would be analogous to starsexhibiting the B[e] phenomenon (cf. Lamers et al. 1998) which,as demonstrated by P40, these stars closely resemble.Finally we note that based on its spectral morphology theLBV G0.120-0.048 would be included in this cohort had not itsproximity to the Quintuplet cluster ( ∼ Eight isolated WN5-7 stars have been identified within the GC,of which two - P99 and P150 - are new discoveries (Table 1).A ninth - qF353 ( = P64; Steinke et al. 2016) - is located on theperiphery of the apparent wind blown structure encircling thenorth and east quadrants of the Quintuplet cluster; since it hashistorically been associated with this cluster, we do not includeit in this census. We note that P2 ( = [MCD2010] 17) - one of theeight stars considered here - is also proximate to the Quintuplet(Mauerhan et al. 2010b). J. S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone.
Fig. 4.
IR colour magnitude diagram for WN7-9ha and mid-Ohypergiants (blue symbols), WN9-11h and early-B hypergiants(black symbols) and putative cool blue hypergiants and super-giant B[e] stars (red symbols). Relevant members of the Arches(WN7-9ha and mid-O hypergiants) and Quintuplet (WN9-11hand early-B hypergiants) clusters are given by open blue andblack circles respectively, while isolated stars are represented byfilled squares of the appropriate colour. Axis scale chosen to mir-ror the comparable plot of WN and WC stars (Fig. 10).The spectra of the seven stars presented in Fig. 9 are all dom-inated by strong and broad emission in He i , He ii and N iii . Theline widths of P39 and 91 support a classification as broad linedsystems; the low resolution spectrum of P2 suggesting likewise(Mauerhan et al. 2010b; their Fig. 4).Steinke et al. (2016) identify qF353E as a WN6 star; in-formed by Rosslowe & Crowther (2018) we use this as a bench-mark to provide relative classifications for the remaining stars.The close similarity of P34 to qF353E suggests an identicalclassification for this star. In comparison to qF353E, the He ii µ m line in P2, 39, 91, and P99 is stronger relative to theother emission features in their spectra, suggesting a WN5-6identification for these stars (with the lack of N v µ m pre-cluding earlier sub-types); conversely the relative weakness ofthis transition in P109 and 147 implies a WN6-7 classification.Finally the combination of exceptionally strong He ii µ memission in P150 and the absence of the absorption componentdue to He i µ m that is present in all other examples suggestthat this is the hottest star observed and hence we assign a WN5sub-type.We present photometry for these stars in Fig. 10, noting thatthe newly discovered P99 and P150 are the faintest examplesidentified to date. Given their near-IR colours P150 appears in-trinsically rather faint while P99 could su ff er considerable ex-cess extinction along its line of sight. The broad lined WN5-6star P39 is also noteworthy in this regard, with the most extremevalue of the cohort ( H − K ∼ K ∼ . Inspection of the relevant literature (e.g. Mauerhan et al. 2010b,2010c, and associated errata; Geballe et al. 2019) reveals 13 iso-lated stars classified as WC stars . As with the Quintuplet cohort(Clark et al. 2018b) these are all WC8 and WC9 stars; indeedonly a single example of an earlier subtype - the WC5 / / Nand resolution, our spectra of P28, 49, 53, 101, and 151 do notnecessitate re-classification, although they do illustrate two es-sential features of this cohort: that stars of the same sub-typepresent spectra with unexpectedly diverse morphologies - pre-sumably indicative of di ff erences in both stellar and wind prop-erties - and that dilution / obscuration of emission features by asubstantial near-IR continuum excess due to the presence of hotdust is common.The spectrum of the final object, MP13 ( = CXOGCJ174519.1-290321; WC9d), exemplifies the latter phenomenon,with the weak emission line spectrum indicating the presenceof hot dust. Intriguingly, the He ii lines that are evident are sug-gestive of a WN7 classification (Fig. 9), with none of the weakerC ii - iv features that characterise WCL stars being visible, despitetheir presence in the spectrum of Mauerhan et al. (2010c; theirFig. 6). While we retain a WC9d classification for this object onthe basis of the latter work, we suspect that it may be intrinsi-cally variable, with greater dilution in our spectrum compared tothat of Mauerhan et al. (2010c) - although we cannot exclude thepossibility of a WN7 companion at this time.Within the GC, excess continuum emission from hot dustwas first recognised for five exceptionally bright near-IR sourceswithin the Quintuplet cluster. High spatial resolution near-IRimaging revealed these to be colliding wind binaries (Tuthill etal. 2006), with subsequent high S / N and resolution
JHK spec-troscopic observations identifying strongly diluted emission fea-tures characteristic of WCL stars (Najarro et al. 2017).Despite its featureless K − band spectrum Mauerhan et al.(2010c) assigned a WCLd classification to CXOGC J174645.2-281547 by analogy to the Quintuplet cohort and by virtue ofits hard X-ray emission. Geballe et al. (2019; their Fig. 2) re-port five further objects with very red, essentially featureless K − band spectra; however, despite the possible presence of weakHe i µ m emission in two examples they refrain from clas-sifying these as dusty WCL stars.We may question whether it is possible to provide a moredefinitive classification of CXOGC J174645.2-281547 and thefive sources from Geballe et al. (2019) via consideration ofphotometry and other observational data. Of these 2MASSJ17431001-2951460 appears associated with the methanolmaser MMB G358.931-0.030 (0.3” distant; Caswell et al. 1977,2010) which argues for an object in a pre-MS evolutionaryphase. 2MASS J17445461-2852042 ( = MGM 1 1, [MKN2009]7) is a large-amplitude photometric variable with an uncon- We note that the proximity of WR102ca and CXOGC J174617.7-285007 to the Quintuplet cluster suggest a physical association; hencewe do not consider these members of the isolated WC star cohort. The inclusion of CXOGC J174617.7-285007 brings the cluster totalat the time of writing to 14.. S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone. 9
Fig. 5.
Spectra of candidate isolated early-B hypergiants. Comparator spectra of Quintuplet cluster members and appropriate classi-fications given in red (Clark et al. 2018b).
Fig. 6.
Spectra of WN9-11h stars. Comparator spectra of Quintuplet cluster members and appropriate classifications given in red(Clark et al. 2018b).
Fig. 7.
Montage of spectra of isolated candidate late-B hypergiants / cool LBVs and sgB[e] stars. Spectra of the LBVs qF362 andG24.73 + Fig. 8.
K-band spectrum of the sgB[e] star P40 plotted with anincreased wavelength coverage to illustrate the presence of CObandhead emission.strained, but apparently long period ( ∆ H ∼ . ∆ K ∼ . / magnitude plot noting that it is most likelya Mira variable. This leaves three remaining candidate WCLdstars - 2MASS J17432173-2951430, J17432988-2950074, and2MASS J17460215-2857235 - which we plot along with the re-maining isolated WCL stars and, for context, the Quintuplet co-hort in Fig. 10. Comparison to other evolutionary groupings reveals that theWCLd stars are the most photometrically diverse, with K ∼ . − . H − K )colour index but an overall correlation - in the sense that brightersources are redder - is present. While multiple physical causesare clearly implied - such as intrinsic di ff erences in stellar andwind properties and di ff erential interstellar reddening - we sup-pose that an increasing contribution from hot circumstellar dustdominates this relationship; a conclusion supported by the factthat the brightest and reddest sources within the Quintuplet clus-ter are those with essentially featureless spectra due to dust dilu-tion.It is therefore encouraging that the spectroscopically con-firmed, isolated WCLd stars plotted in the colour / magnitudediagram are coincident with examples found within theQuintuplet, while CXOGC J174645.2-281547, 2MASSJ17432173-2951430, J17432988-2950074, and 2MASSJ17460215-2857235 seamlessly extend this co-location tobrighter, redder objects with featureless spectra . Moving tofainter K − band magnitudes and, of the three isolated WC starswith K ∼ − .
5, the emission lines in the spectra of the twostars with the largest ( H − K ) values - 2MASS J17444083-2926550 and J17463219-2844546 - are also very weak (Geballeet al. 2019; their Fig. 4). This is consistent with the presence ofsubstantial continuum veiling due to emission from hot dust, an The final star with a featureless spectrum from Geballe et al. (2019)- 2MASS J17431001-2951460 - is ∼ . H − K ) colour index than the preceding objects,bolstering the conclusion that it is in a pre-MS evolutionary phase givenits association with a methanol maser.. S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone. 11 P34P39P91P99P109P147P150WN6 (qF353)
Fig. 9.
Montage of spectra of isolated candidate WN5-7 Wolf-Rayets. The spectrum of the WN6 star qF353E is shown in red forcomparison. Prominent transitions are indicated; for completeness the unlabelled He i lines at 2.059 µ m and 2.1126 + µ m arethe 2p P o - 2s S and the 4s S - 3p P o + S - 3p P o transitions respectively.hypothesis strengthened by their extremely red ( J − K ) colourindices.Given this we conclude that, as assumed for CXOGCJ174645.2-281547, the three isolated objects from Geballe etal. (2019) with featureless K − band spectra are also bona fidedusty WCL stars. Indeed we may invert the argument: given thatthe Quintuplet clearly hosts such stars one would anticipate theirpresence in the isolated stellar cohort, so if these objects are notdusty WCL stars one would be need to explain their absence.In either eventuality we close this discussion by noting that thenumber of isolated dusty WCs within the CMZ - 13 or 16, de-pending on the nature of these sources - is directly comparable tothe number associated with the Quintuplet cluster (see footnotefive). The last homogeneous cohort that may be identified in our obser-vations are a group of six faint ( K ∼ − .
6) stars with spectradominated by strong Br γ emission and, in a subset, weak He i µ m emission (Fig. 12). He i and He ii photospheric lines -which might enable a temperature determination - are absent, as is emission in both high and low excitation metallic transi-tions (although the Mg ii + µ m doublet was identifiedin the spectrum of P105 presented by de Witt et al. 2013). Thelack of low excitation atomic (Na i , Ca i ) or molecular (CO band-head) absorption features disfavours the possibility that these areforeground cataclysmic variables. Moreover, the absence of CObandhead emission distinguishes these stars from the IR excessobjects within the Arches cluster, which Stolte et al. (2010) sug-gest are B-type stars surrounded by remnant protostellar discs.Instead they most closely resembles classical Oe / Be stars (Clark& Steele 2000); rapidly rotating non-supergiant late-O to early-A stars, characterised by gaseous, quasi-Keplerian circumstellardecretion discs that generate line emission in H i , He i and lowexcitation metallic transitions as well as a near-IR continuumexcess (cf. Porter & Rivinius 2003).We may ask whether the observational properties of this co-hort are consonant with such a classification. The equivalentwidths, full-width half maxima and line intensities of the Br γ and - where present - He i µ m emission lines are consistentwith the range expected for early-B stars exhibiting the Be phe-nomenon (Clark & Steele 2000). Likewise their broad, asymmet-ric emission profiles are a natural consequence of the one-armed Fig. 10.
IR colour magnitude diagram for candidate and con-firmed WN5-7 (red) and WCL stars (black). Members of theQuintuplet cluster and isolated examples given by open cir-cles and filled squares respectively. Stars with featureless K-band spectra are indicated by nested symbols, the anomalouslyfaint example being 2MASS J17431001-2951460 (with 2MASSJ17445461-2852042 excluded due to variability; Sect. 3.8).Ground-based photometry for the Quintuplet members given inCutri et al. (2003), Dong et al. (2012), and Hussmann et al.(2012). No photometry is available for qF235N while the lo-cation of qF76 & 309 ( K ∼ . , ( H − K ) ∼ . K ∼ . , ( H − K ) ∼ . i µ m line profile ofS124 clearly indicative of rotation. Finally such discs are tran-sient phenomena, leading to significant spectral variability onthe timescale of years; potentially explaining the disappearanceof Mg ii emission in the four years between the two observationsof P105.Turning to photometric properties and assuming a distanceto the GC of ∼ A K ∼ K ∼ . K ∼ . K & K − band magnitudes anticipated for Be stars withinthe CMZ is consonant with the stellar cohort considered here.Likewise, comparison of the intrinsic near-IR colours of Be stars(( H − K ) ∼ . − .
5; Lada & Adams 1992, Dougherty et al.1994) to the values exhibited by this cohort (( H − K ) ∼ . − . ∼ −
20% of B0-3 stars exhibit the Be phenomenon at Galacticmetallicities (Wisniewski & Bjorkman 2006) and, as demon- strated above, their properties (an IR excess and strong line emis-sion) favour their detection via surveys such as that of Dong etal. (2011). Consequently it would be surprising if none were tobe found in the GC - one would be forced to invoke a rather con-trived star formation history that limited the formation of starsof ∼ ⊙ over the past ∼ / massive young stellar objects(YSOs; cf. Bik et al. 2005, 2006) - at this time, we consider anidentification of this cohort as classical Be stars located withinthe CMZ to be the most compelling explanation for their spec-troscopic and photometric properties as reported here. Six sources do not fit into any of the above classifications, ofwhich we have new observations of three. Cotera et al. (1999) re-ported that the spectrum of the highly reddened (( H − K ) ∼ . ii region Sgr A-Dwas dominated by strong He i µ m and Br γ emission lines;on this basis they suggested that this indicated a nebular, ratherthan stellar origin. Despite the increased S / N and resolution ofour spectrum (Fig. 13), these remained the only identifiable fea-tures. With FWHM . − , the line profiles are narrowerthan expected for an origin in either the spherical wind of a hotstar or the disc of a classical Be star ; we therefore concur withCotera et al. (1999) that the spectrum is dominated by emissionfrom the UCH ii region.Both P57 and P58 show strong, single peaked emission inBr γ and the Pfund series up to at least Pf-30, with FWHM ∼ − in our new observations (Fig. 13). Emission in thePfund series is observed in massive YSOs (Bik et al. 2005,2006), sgB[e] stars (Oksala et al. 2013, Kraus et al. 2020) andLBVs (Oksala et al. 2013, Najarro et al. 2015). The line widthsof both stars are consistent with any of these possibilities; how-ever the ratio of (Br γ / Pf-20) ∼ . γ / Pf-20) ∼ µ m emission in P58 marks itout as a massive YSO; with ( H − K ) ∼ . . No H emission is seen in P57, while the marginal detec-tion of Mg ii µ m is consistent with either a massive YSOor a sgB[e] star; as a consequence we leave both options openat this time. However, in either eventuality the presence of pro-nounced He i µ m emission implies a rather hot source inorder to provide the requisite UV photons to drive the line intoemission.Of the remainder 2MASS J17431001-2951460 has alreadybeen briefly discussed in Sect. 3.6 where, despite the feature-less spectrum presented by Geballe et al. (2019; their Fig. 2),near-IR photometry is discrepant with expectations for a dustyWCL star; rather the presence of the methanol maser sourceMMB G358.931-0.030 only 0.3” away (Caswell et al. 1977, The Br γ lines of the six Be stars identified in Sect. 3.7 haveFHWM & − . Since pre-MS objects are typically intrinsically red (cf. 2MASSJ17431001-2951460 and J17470921-2846161; Table 1) it is possiblethat P58 is an interloper.. S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone. 13
Fig. 11.
Montage of spectra of isolated WC8-9 Wolf Rayet stars, with the Quintuplet WC8 star LHO 47 plotted in red for comparison(Clark et al. 2018b). Given the remarkable diversity of morphologies, no additional WC9 comparator spectra from the Quintupletare available. The emission features in MP13 are particularly weak; we therefore enhance these by a factor of five and overplot theresultant spectrum in blue for clarity.2010) is suggestive of a YSO classification. The presence ofCO bandhead emission associated with 2MASS J17470921-2846161 (Geballe et al. 2019; their Fig. 6) implies a dense, coolcircumstellar envelope; the addition of H molecular emissionfurther points towards a (M)YSO classification (Bik et al. 2005,2006). Unfortunately, given the potential for a near-IR contin-uum excess from the circumstellar envelope and uncertain inter-stellar extinction it is not possible to make any further inferencesas to the nature of the source from extant photometry at this time.Finally 2MASS J17444840-2902163 has a spectrum dom-inated by Br γ absorption and Na i and CO bandhead emission(Geballe et al. 2019; their Fig. 6). Comparison to the spectrumof the YHG ρ Cas presented in Yamamuro et al. (2007) suggesta similar classification for this star, although the near-IR proper-ties (Table 1) imply a rather moderate luminosity in comparisonto e.g. the YHG cohort of Westerlund 1 (Clark et al. 2005).
Given the crowded nature of the CMZ, the size of the IFUs andthe selection criteria employed and it is inevitable that interlop- ers will be included in our survey. The most common contami-nants were found to be cool, late-type stars in the line of sightwhich mirror the expected magnitudes and colours of early-typestars in the GC. Somewhat unexpectedly these were prevalentamongst targets selected from the list of putative Pa α emitters ofDong et al. (2011). Table 2 provides a list of such stars identi-fied amongst our primary targets, along with those with eitherfeatureless spectra or insu ffi cient S / N to attempt a classifica-tion. Where photometry was available, a combination of ( H − K )colour index and K > . the additional spectra wereexamined and found to be either featureless or those of coolstars.This leaves a handful of additional stars worthy of brief com-ment. Dong et al. (2012) classify both P38 and 140 as massive Those surrounding stars [DWC2011] 15, 25, 34, 40, 43, 47, 48, 49,57, 75, 78, 99, 105, 106, 135, 136, 143, and 147; S59, 64, 117, 123,131, 136, and 183; MP13.4 J. S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone.
Fig. 12.
Montage of spectra of possible isolated classical Be stars. We caution that the central region of the Br γ profile of P135 hasbeen artificially removed due to the presence of hot pixels. Fig. 13.
Montage of spectra of three sources dominated be narrow H i recombination lines (Sect. 3.8); the star embedded in theUCH ii region Sgr A-D, the MYSO / sgB[e] star P57 and MYSO P58. . S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone. 15 foreground objects on the basis of their near-IR colours. Basedon our new spectrum, we reclassify the X-ray bright P38 slightlyto O4 Ia. Mauerhan et al. (2010a; their Fig. 3) give P140 as B0-2Ia. However the spectrum presented by Geballe et al. (2019; theirFig. 6) shows a pattern of emission (including Fe ii and Na i ) andabsorption (bluewards of Br γ ) more reminiscent of a cool LBVcandidate; intriguingly, comparison of these spectra implies vari-ability between these observations (obtained in 2009 August and2016 September respectively). The third object, P102, likewisehas near IR colours indicative of a foreground object, while thedouble peaked Br γ profile suggest that it is a classical Be star.Lastly we turn to two objects for which we cannot yet ad-vance a classification. The spectrum of P25 is featureless savefor a single peaked Br γ line of moderate intensity ( I / I cont ∼ . K − band measurement is consistentwith the cohort of classical Be stars in the absence of colour in-formation we may not infer a distance to the object and hence ad-vance a classification. Finally P110 is of a comparable K − bandmagnitude and ( H − K ) colour to the candidate classical Be stars(Sect. 3.7) and shows emission coincident with the wavelengthof Br γ . However multiple apparently spurious emission featuresof comparable strength are also present in the spectrum, castingsome doubt as to the astrophysical origin for the putative Br γ emission; hence we refrain from classifying the star at this time. In total we have observed 82 di ff erent primary targets, derivedfrom the list of Pa α excess sources of Dong et al. (2011) andprevious reported candidates from the literature. Of these theIFUs of 26 sources contained one or more additional objects -yielding spectra of >
100 stars in total - although none of theseadditional serendipitous sources were found to be massive stars.Of the primary targets a total of 31 were either cool foregroundinterlopers, displayed apparently featureless spectra or were ofinsu ffi cient S / N to attempt classification (Table 2). Includingone literature classification, three objects were assessed as fore-ground massive stars while two further were unclassifiable fromour spectra (Sect. 3.9).This left a total of 47 objects which spectroscopy revealed tobe massive stars. Of these 17 are new identifications, while thespectra of a further 19 stars allowed improved spectral classifi-cations. The remaining 11 stars retained the same classificationas previously reported.Table 1 lists a total of 83 isolated massive stars - derived fromour observations and the literature - with near-IR photometryconsistent with a location in the GC and for which classificationhas proved possible . These comprise: • Five mid-O and four late-O / early-B supergiants. One fur-ther object, XID For completeness we provide a list of Pa α excess sources remainingto be observed in Appendix A. As noted in Sects. 3.3-3.7, the di ff use, extended nature of theQuintuplet cluster makes it di ffi cult to determine whether outlying mas-sive stars are physically associated with it or not. Historically the LBVG0.120-0.048 and WR stars WR102c ( = qF353E; WN6), WR102ca(WC8-9) and CXOGC J174617.7-285007 ( = MP14; WC8-9) have beenincluded as probable cluster members (cf. Clark et al. 2018b, 2020),despite their displacement from the central stellar concentration. As aconsequence, and in order to avoid double counting, we do not includethese in the census given here, but we highlight that their physical asso-ciation with the cluster is as yet unproven. • Eight mid-to late-O hypergiants, 11 WNLha stars and oneWN8-9ha / O6-7 Ia + hybrid. • Four WN9-11h stars, three early-B hypergiants and twohybrid early-B hypergiant / WNLh stars. Of these Geballe et al.(2019) highlight the close proximity of 2MASS J17461292-2839001 (B1-3Ia + ) to the Quintuplet; we retain it here for com-pleteness, although consider cluster membership highly likely. • Eight objects with spectra characterised by strong narrowBr γ and weak, low excitation metallic emission lines. We iden-tify one as a supergiant B[e] star on the basis of pronounced CObandhead emission. We tentatively classify the remaining starsas either late-B hypergiants or cool phase LBVs; further pho-tometric and spectroscopic monitoring being required to distin-guish between these possibilities. • One star, 2MASS J17444840-2902163, which we provi-sionally classify as a low-luminosity YHG and hence may ex-tend the preceding cohort to lower temperatures. • Eight WN5-7 stars, of which three are broad lined systems.Of these Mauerhan et al. (2010b) note the proximity of P2 to theQuintuplet cluster. •
16 WC8-9 stars, including the featureless sources CXOGCJ174645.2-281547, 2MASS J17432173-2951430, J17432988-2950074, and 2MASS J17460215-2857235 (Mauerhan et al.2010c, Geballe et al. 2019). • Six fainter stars with spectra dominated by strong Br γ emission that we classify as classical Be stars. • Three apparent (massive) YSOs and a further source dis-playing nebular emission associated with an UCH ii region (not-ing that a considerably larger population of (massive) YSOs havebeen identified on the basis of IR-radio continuum and mid-IR spectroscopic observations; An et al. 2011, Ginsburg et al.2018a, Yusef-Zadeh et al. 2009). • One star, P57, with a spectrum consistent with a classifi-cation as either a massive YSO or sgB[e] star.Given the nature of the surveys employed for target selectionwe emphasise that this census is highly likely to be incomplete.We discuss this limitation and the wider implications of this stel-lar population in the following sections. J . S . C l a r k e t a l . : C on s t r a i n i ng t h e popu l a ti ono f i s o l a t e d m a ss i v e s t a r s w it h i n t h e C e n t r a l M o l ec u l a r Z on e . Table 1.
Census of massive stars in the sightline towards the GC
ID RA Dec J H K Aliases Classification[DWC2011] (h m s) (d m s) (mag) (mag) (mag) Old NewOB supergiants50 17 45 02.89 -29 08 59.8 13.9 ± ± ± ± ± ± + [B0-3 Ia]17 45 29.89 -28 54 28.9 - 12.2 10.8 S152 - O4-5 Ia17 45 30.32 -28 52 07.0 - 13.4 11.9 S73 - O4-5 Ia17 45 37.30 -28 53 53.7 15.75 ± ± ± ± ± ± ± ± ± ± ± / O hypergiants15 17 46 03.21 -28 48 58.4 - 13.2 ± ± +
22 17 45 53.40 -28 49 36.9 15.4 ± ± ± ± ± ± +
35 17 45 28.62 -28 56 05.0 14.5 ± ± ± + WN8-9ha + neb36 17 45 31.50 -28 57 16.8 15.1 ± ± ± +
75 17 46 17.10 -28 51 31.5 15.0 ± ± ± + WN8-9ha / O6-7 Ia +
77 17 46 17.54 -28 53 03.5 15.0 ± ± ± ± ± ± / Of WN7-8ha97 17 45 47.72 -28 50 49.2 15.3 ± ± ± + O6-7 Ia +
100 17 45 42.32 -28 52 47.1 14.7 ± ± ± + O7-8 Ia +
107 17 45 39.34 -28 53 21.1 14.7 ± ± ± + O7-8 Ia +
111 17 45 36.12 -28 56 38.7 15.6 ± ± ± ± ± ± +
134 17 45 16.74 -28 58 25.1 16.7 ± ± ± +
17 46 20.87 -28 46 58.8 - 14.1 12.5 S120 - WN8-9ha17 46 56.36 -28 32 32.3 - 13.74 ± ± ± .
06 12.72 ± ± ± ± ± ± ± / early-B hypergiants19 17 45 48.61 -28 49 42.2 - 13.6 ± ± ± ± ± ± ± ± + B1-2 Ia + / WNLh103 17 46 01.65 -28 55 15.3 13.4 ± ± ± + / WNLh137 17 45 16.17 -29 03 14.7 11.6 ± ± ± / WN9 WN10h17 45 23.11 -29 03 29.3 16.7 ± ± ± + ]17 45 54.65 -28 47 44.9 - 9.5 7.9 S132 - B1-3 Ia +
17 46 12.93 -28 49 00.2 13.79 10.53 8.85 2MASS J17461292-2849001 - [B1-3 Ia + ]17 46 18.12 -29 01 36.6 12.97 ± ± ± . S . C l a r k e t a l . : C on s t r a i n i ng t h e popu l a ti ono f i s o l a t e d m a ss i v e s t a r s w it h i n t h e C e n t r a l M o l ec u l a r Z on e . Table 1. continued.
ID RA Dec J H K Aliases Classification[DWC2011] (h m s) (d m s) (mag) (mag) (mag) Old NewCool BHGs / sgB[e] stars40 17 45 24.06 -29 00 58.9 13.2 ± ± ± ± ± / LBV?141 17 45 09.29 -29 08 16.2 - 15.3 ± ± / LBV?17 44 43.20 -29 37 52.6 16.00 12.78 10.14 2MASS J17444319-2937526 uncl. [late BHG / LBV?]17 44 55.38 -29 41 28.5 16.72 12.63 10.14 2MASS J17445538-2941284 uncl. [late BHG / LBV?]17 45 02.41 -28 54 39.2 15.81 12.56 10.00 2MASS J17450241-2854392 uncl. [late BHG / LBV?]17 47 09.40 -28 49 23.6 12.66 10.8 9.46 2MASS J17470940-2849235 uncl. [late BHG / LBV?]17 48 24.73 -28 24 31.3 15.86 12.12 9.54 2MASS J17482472-2824313 uncl. [late BHG / LBV?]WN5-7 stars2 17 46 23.81 -28 48 10.8 16.3 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± J . S . C l a r k e t a l . : C on s t r a i n i ng t h e popu l a ti ono f i s o l a t e d m a ss i v e s t a r s w it h i n t h e C e n t r a l M o l ec u l a r Z on e . Table 1. continued.
ID RA Dec J H K Aliases Classification[DWC2011] (h m s) (d m s) (mag) (mag) (mag) Old NewClassical Be stars (cont.)17 45 52.16 -28 55 01.5 - 17.4 14.6 S69 - Be star17 46 12.95 -28 47 16.3 - 14.6 13.1 S124 - Be starMiscellaneous and uncertain classification57 17 46 29.90 -28 46 39.9 - 14.7 ± ± / sgB[e]58 17 46 31.85 -28 46 47.1 14.3 ± ± ± ± ± ± / ucH ii ucH ii Foreground (( H − K ) ≤
1) and unclassifiable25 17 45 58.31 -28 52 20.0 - - 13.4 ± ± ± ± ± ± ± ± ± ± ± ± xxx ) for their primary list of Pa α excess sources (column 1), noting that S69, 73, 120, 124, 131, 132,and 152 derive from their secondary list. Additional identifications derive from Mauerhan et al. (2010b; CXOGC sources), Mauerhan et al. (2010c; [MCD2010] xx ) de Witt etal. (2013 XID xxx ), Geballe et al. (2019, 2MASS sources) and Zhao et al. (2020; GCCR xxx ). In all cases photometry (and associated errors where available) and classificationswere obtained from these works, supplemented by Homeier et al. (2003), Mauerhan et al. (2007), Clark et al. (2009b), Oskinova et al. (2013), de Witt et al. (2013) and Dong etal. (2015) where required. . S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone. 19 Table 2.
Non massive star interlopers projected onto the CMZ [DWC2011] RA (J2000) Dec (J2000) Appearance3 17 46 03.60 -28 47 09.8 Cool16 17 45 54.85 -28 47 13.2 Cool29 17 46 09.60 -28 57 13.5 Cool30 17 46 08.10 -28 58 24.2 Cool31 17 46 06.27 -28 59 14.9 Cool43 17 45 13.95 -29 04 38.2 Cool47 17 45 15.05 -29 09 08.3 Featureless48 17 45 22.68 -29 10 56.5 Featureless78 17 45 48.22 -28 47 26.0 Cool104 17 46 07.30 -28 57 17.5 Cool113 17 45 33.69 -28 57 49.7 Cool133 17 45 36.86 -29 01 17.5 Featureless136 17 45 28.22 -29 03 27.1 Cool142 17 45 19.45 -29 10 33.9 Low S / N143 17 45 23.91 -29 10 24.7 Low S / N144 17 45 14.22 -29 11 41.7 Low S / N(S47) 17 46 31.07 -28 46 14.2 Cool(S52) 17 46 24.30 -28 46 53.1 Cool(S59) 17 46 01.64 -28 49 00.0 Cool(S64) 17 46 05.02 -28 52 06.7 Cool(S78) 17 45 53.36 -29 00 49.1 Low S / N(S113) 17 44 55.32 -29 11 54.8 Cool(S117) 17 46 26.56 -28 46 26.4 Cool(S119) 17 46 32.72 -28 46 08.1 Cool(S123) 17 46 29.95 -28 49 32.2 Low S / N(S136) 17 46 01.64 -28 51 10.3 Featureless(S139) 17 46 02.44 -28 53 13.7 Cool(S141) 17 46 06.14 -28 54 41.4 Low S / N(S146) 17 46 07.43 -28 59 18.5 Low S / N(S153) 17 45 35.44 -28 55 12.3 Low S / N(S183) 17 45 04.24 -29 09 26.7 CoolStars given in parentheses with an ‘S xxx ’ designation are from the sec-ondary list of Pa α emitters in Dong et al. (2011).
4. Discussion
The results described in the preceding section reveal that a largenumber of isolated and potentially very massive stars of diversenature appear distributed through the CMZ. These include veryrare phases - such as LBVs and sgB[e] stars - that are thought toplay an important role in the lifecycle of single and binary stars.As such characterising this population, including observationalbiases, will be critical to further understanding the evolution ofsuch extreme objects as well as the wider ecology and star for-mation history of the circumnuclear region of the Galaxy.
Centred on Sgr A ∗ , the HST Pa α excess survey has the small-est (asymmetric) footprint of those utilised in our study (39 × . ×
36 pc at a distance of 8kpc; Wang et al. 2010,Dong et al. 2011) and is expected to detect early-type stars viaemission from their ionised wind. As such detection probabilitywill be a function of both stellar mass and evolutionary phase,with mass loss rates greater for more massive stars and winddensities increasing through the stellar lifecycle . Since all threeyoung massive clusters are within the survey footprint we may Subject to the caveat that stars in a cool yellow hypergiant / red su-pergiant phase will be undetectable unless their winds are subject toexternal ionsiation (cf. Westerlund 1; Dougherty et al. 2010, Fenech etal. 2018) utilise their well defined stellar populations to empirically deter-mine its sensitivity to spectral type.All 13 WNLha stars and 5 / α excess, but only 3 /
30 of themid-O supergiants and none of the >
50 O5-9 stars of luminos-ity class III to V, although source blending / confusion may com-promise identification given the compact nature of the cluster(Dong et al. 2011, Clark et al. 2018a, 2019b). In the Quintuplet -which is less compact and so presumably less prone to blending- all three LBVs and 8 /
10 of the early-B HGs / WNLh are de-tected (with the two missing examples having the weakest He i /
16 of the WC stars and all 5 of the candidate ‘bluestragglers’ of spectral types WN8-9ha and O7-8Ia + . However,as with the Arches only 1 /
23 of the O7-B0 Ia supergiants are de-tected. Although apparently of lower luminosity (Martins et al.2007), the same pattern is repeated for stars within the GalacticCentre cluster, with 25 /
33 of the Wolf-Rayet cohort detected butonly one of the 26 OB supergiants present.Consistent with our findings for the isolated stellar cohort(Tables 1 and 3), the cluster detection demographics imply thateven stars as extreme as O4 supergiants may be routinely un-detectable via their excess Pa α emission. This is all the morestriking since such objects likely derive from very massive stars(M init & M ⊙ ; Groh et al. 2014, Martins & Palacios 2017, Clarket al. 2018a).The Chandra GC survey of Muno et al. (2009) covers a largerfield than the preceding study (2 o × . o or ∼ ×
112 pc at a dis-tance of 8kpc). In total Mauerhan (2009) suggest ∼ −
300 X-ray sources are coincident with near-IR sources with K < . K <
12) matchesto hard X-ray sources. However a number of targets were se-lected on the basis of their proximity to mid-IR structures indica-tive of the presence of massive stars (i.e. wind-blown bubblesand bow shocks), precluding a statistical analysis of the successrate deriving from the application of such observational criteria.Nevertheless, the distribution of the spectral types of the X-raybright massive stars essentially mirrors that derived from the Pa α survey (Table 1 and Mauerhan et al. 2010c); such a detection biasis unsurprising since both surveys are expected to be sensitive tostars supporting dense, high velocity winds.The photometrically selected survey of Geballe et al. (2019)has a footprint intermediate between the preceding studies(2 . o × . o or ∼ × ff use gas it em-ployed a mid-IR cut ([3.6] <
8) that is not optimised for identify-ing massive stars. This is evident in the detection rate, with ∼ candidate massive stars from over 500 spectroscopically sur-veyed; these being strongly biased towards sources potentiallyassociated with circumstellar dust such as cool LBVs / sgB[e] andWCLd stars . No WN5-7 or Be stars were detected, and thethree O hypergiants and WNLha stars identified were coinci-dent with extended mid-IR nebulae, suggesting they were stillassociated with their natal material or bow shocks (P35 and 114respectively; Dong et al. 2017).In conclusion the combination of di ff erent methodologies(and in some cases subjective criteria) means that we are unableto provide robust quantitative estimates for the detection thresh-olds and hence level of incompleteness of the surveys inform-ing our target selection. However it seem likely that all threeare insensitive to a large number of stars over a wide range ofinitial masses and evolutionary phases. Specifically, one wouldonly expect to detect comparatively low mass objects ( M init ∼ − M ⊙ ; Groh et al. 2013) at the end their lives, via dust emis-sion in an RSG phase or via a Pa α excess in a Be star episode.Stars above this threshold are expected to loop back to highertemperatures and hence one might also anticipate detecting themin a LBV / BHG or WR phase. The same is expected at still highermasses ( M init & M ⊙ ) - a regime in which stars remain at hightemperatures throughout their lives. However, evolutionary sim-ulations suggest that such stars will only support winds of su ffi -cient density to permit detection after a considerable proportionof their H-burning lifetime has elapsed. For example the simula-tions of Groh et al. (2014) show that a M init ∼ M ⊙ star has a life-time of ∼ . α excess - after ∼ . &
80% of its life.Informed by the stellar population of Westerlund 1, onemight expect binarity to aid detection via all survey method-ologies. Mass loss during the active interaction potentially leadsto the formation of sgB[e] stars (cf. Wd1-9; Clark et al. 2013;Kastner et al. 2006) which should be readily detectable as Pa α excess sources (cf. P40 and 57). Likewise stripped primaries en-tering a (proto-)WR phase (cf. Wd1-5; Clark et al. 2014) andmass-gainers / merger products (cf. Wd1-27 and 30a; Clark et al.2019a) both support pronounced emission line spectra whichare absent in their late-O supergiant progenitors. Moreover, dustproduction in CWBs containing WC stars enhances the like-lihood of their detection at mid-IR wavelengths, while X-rayemission via wind collision or, more rarely, accretion onto acompact companion may render identifiable otherwise unde-tectable stars. Various authors have suggested that massive stars may form incomparative ‘isolation’ in regions of low molecular and, subse-quently, stellar density (de Wit 2004, 2005, Parker & Goodwin2007). Indeed, observations of the Cyg OB2 association (Wrightet al. 2014, 2016) the 30 Dor star forming region (Bressertet al. 2012, Schneider et al. 2018) and the Small MagellanicCloud (Lamb et al. 2016) are consistent with such an hypoth-esis. However, determining the origin of isolated stars is noto- Intriguingly the sample of WCLd stars returned shows no overlapwith that selected via Pa α excess; presumably emission from the warmdust that allows their photometric detection dilutes emission featuresin the spectrum to the point at which they cannot meet the detectionthreshold for the survey of Dong et al. (2011). riously di ffi cult; the WN5h star VFTS 682 - which closely re-sembles the WNLha stars distributed through the CMZ - beinga case in point. Bestenlehner et al. (2011) highlight its locationin the outskirts of 30 Dor and the lack of an associated stellaraggregate as consistent with its formation in isolation, but areunable to exclude an origin in - and subsequent ejection from -the young massive cluster R136, some 29pc distant (which is it-self an analogue of the Arches; Crowther et al. 2016). Thereforebefore addressing the distribution of isolated stars through theCMZ it is instructive to consider the physical mechanisms thatmay redistribute stars from cluster to field. Two physical mechanisms are thought to give rise to the majorityof runaway stars in the Galactic disc; ejection via SN explosionsin binaries (Blaauw 1961) or dynamical interaction (Poveda etal. 1967, Banerjee et al. 2012, Fujii & Portegies Zwart 2011).Considerable e ff ort has been invested in determining the magni-tudes of SN kicks and their e ff ects on the survivability and mo-tion of binary systems (e.g Renzo et al. 2019 and refs. therein).With v ∼ − , the high mass X-ray binary Vela X-1 (B0 Ib + neutron star) suggests that a considerable velocity may be im-parted in at least some cases (Kaper et al. 1997). While this chan-nel appears inapplicable for the Arches given its age (2-3Myr;Clark et al. 2018a), it is likely viable for both the Quintupletand Galactic Centre clusters ( ∼ − . ∼ − ∼
3” of SgrA* (Kennea et al. 2013, Mori et al. 2013) and the young pulsarJ1746-2850I within 2’ of the Quintuplet (Deneva et al. 2009) aresuggestive of ongoing SNe activity in these regions.Turning to dynamical ejection and one example of a verymassive and high velocity runaway is VFTS 16 ( M ∼ M ⊙ , v ∼ − ), which Lennon et al. (2018) demonstrate to havea proper motion consistent with an origin in the LMC clusterR136; an aggregate that is too young to host SNe at this time(Crowther et al. 2016). This is of particular interest since R136appears similar to the Arches in terms of its youth, stellar densityand masses of constituent stars (Clark et al. 2018a, 2019b); sug-gesting that similar high velocity runaways might be expected tooriginate from the latter cluster.On larger physical scales, the expulsion of residual gas fromcompact clusters via stellar feedback has long been posited asa mechanism for driving their expansion and, in some cases,destruction - with the now supervirial velocities of the con-stituent stars dispersing them into the wider field (e.g. Goodwin& Bastian 2006; see also Park et al. 2018 for a discussion ofthis e ff ect in the context of the GC). Moreover, the tidal strip-ping and the eventual disruption of clusters may also distributemassive stars through the GC. Simulations of this phenomenonfor both the Arches and Quintuplet suggest that tidal arms ofseveral tens of parsecs may result from this process after onlya few Myr, although the extent of such structures is a sensitivefunction of both cluster age and distance from Sgr A ∗ (Habibi etal. 2014, Park et al. 2020). By comparison, assuming that bothdynamical interactions and SNe kicks may generate runaway ve-locities of up to ∼ kms − such stars may be displaced fromtheir natal clusters by up to ∼ ∼ . o at 8kpc) within10 yr - comfortably less than the age of either the Arches orQuintuplet. Combined with the bulk orbital motion of both ag-gregates (232 ± − and 167 ± − respectively; Stolteet al. 2008, 2014) one may anticipate a combination of thesethree processes potentially stripping massive stars from their na- . S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone. 21 tal clusters and distributing them across a significant fraction ofthe CMZ. We show the locations of the isolated massive stars acrossthe GC in Fig. 14. In constructing this plot we omitted bothOB dwarfs which, to date, have solely been identified withinthe Arches and Galactic Centre clusters (Clark et al. 2019b,Paumard et al. 2006) and red supergiants, which are present (e.g.Wollman et al. 1982, Cunha et al. 2007, Liermann et al. 2012)but di ffi cult to distinguish from cool interlopers with the data tohand.It is immediately apparent that there is a large overdensityof stars in the inner regions of the CMZ; we consider it likelythat this is in part due to the limited footprint of the Pa α survey(running from ∼ + ◦ ′ to ∼ ◦ ′ ; Wang et al. 2010) whichdrove placement of the KMOS IFUs. Given such observationalbiases and the potential redistribution of stars from their birthsites, we are limited in the conclusions that may be drawn fromthese data, especially for individual objects.Nevertheless, thanks to the sample size we are able to plotmassive, post-main sequence stars as a function of spectral typewhich, informed by cluster demographics, we may use as aproxy for stellar age. Comparison to the Arches cluster suggeststhat the mid-O super- / hypergiants and WNLha stars plotted inthe uppermost panel of Fig. 14 are likely to be the youngest andmost massive cohort present ( < M init & M ⊙ ; cf.Clark et al. 2018a). A cursory examination reveals this cohortis observed at low Galactic latitudes, as might be expected foryoung objects. More intriguingly, the entire population appearsto be distributed at positive galactic longitudes, with no exam-ples significantly westward of Sgr A ∗ (the right half of the panelin Fig. 14). Given that the entire region is covered by both thePa α and X-ray surveys, which are sensitive to such stars, it ap-pears di ffi cult to attribute this to observational bias.Historically it had been thought that the bulk of the densemolecular gas within the CMZ was located at positive galac-tic longitudes (so eastwards of Sgr A ∗ in Fig. 14) while themajority of young stars identified by mid-IR observations wereinstead found at negative longitudes (Morris & Serabyn 1996,Longmore & Kruijssen 2018). However, the distribution of mid-O super- / hypergiants and WNLha stars shows the situation isless clear cut, with a rich population of young, very massivestars coincident with active star forming regions. Habibi et al.(2014) suggested that a noteworthy association of six stars co-located with the Sgr B complex (RA ∼ ◦ ′ → ◦ ′ , δ ∼ ◦ ′ → − ◦ ′ ) could originate in a tidal arm associated withthe Quintuplet. However the cluster age and membership profileprecludes such an origin for these objects - and indeed for any ofthe mid-O super- / hypergiants and WNLha stars within the GCdistributed via any dispersal mechanism (cf. Clark et al. 2018b).The only cluster young enough to serve as a viable birthsite is theArches (Clark et al. 2018a); however predictions for the geome-try of a putative tidal tail suggest that it would not intersect withSgr B (Habibi et al. 2014). Therefore, given their apparent co-evality and proximity to one another we consider it likely thatthe sextet form a distinct, physically related group; we discussthis putative aggregate further in Sect. 4.2.3. CXOGC J174656.3-283232 (WN8-9ha), CXOGC J174703.1-283119 (O4-6 Ia), CXOGC J174711.4-283006 (WN8-9ha), CXOGCJ174712.2-283121 (WN7-8ha), CXOGC J174713.0-282709 (WN7-8ha) and CXOGC J174725.3-282523 (O4-6 Ia).
Conversely, as discussed in Clark et al. (2019b) the mid-O supergiants and WNLha stars adjacent to the Arches clustercould have originated there and been ejected via dynamical in-teractions or tidal stripping. A further grouping of four WNLhastars and mid-O hypergiants with small angular separations fromone another are located north east of Sgr A ∗ (RA ∼ ◦ ′ , δ ∼ ◦ ′ ); we revisit this possible assemblage in Sect. 4.2.3.We plot the locations of the late-O to late-B super- / hypergiants, LBVs and WN9-11h stars - so called ‘transitional’objects - in panel two and the H-depleted WN5-7 and WC8-9stars in panel three of Fig. 14. Comparison to the populations ofthe Quintuplet and GC suggest they derive from an older stellarpopulation ( ∼ − M init & M ⊙ ; cf. Paumard et al. 2006, Martins etal. 2007, Clark et al. 2018b). While both cohorts are again cen-trally concentrated there appears less evidence for segregationas a function of galactic longitude in comparison to the youngerstars. We note, however, that there is an absence of older starscoincident with Sgr B; on this basis one might speculate thatactive star formation in this region has proceded for . . While such systemscould arise from in situ formation, the tidal disruption of a clus-ter or dynamical ejection from such a site (Oh & Kroupa 2016)formation via a SN kick would appear disfavoured; future stud-ies to compare the binary properties of isolated stars and thosewithin clusters would be of particular interest in order to eluci-date their formation pathways. Simple visual inspection of the CMZ fails to reveal any compara-bly massive clusters to the Arches or Quintuplet. This absence isunlikely to be driven by stellar evolution; although the most mas-sive members will be lost to SNe the remnant population of anageing cluster will increasingly become dominated by IR-brightRSGs. For example, if placed in the GC, and assuming a rep-resentative extinction of A K ∼
3, the ten cool super- / hypergiantswithin Westerlund 1 would span K ∼ − The remaining dusty source, CXOGC J174645.2-281547 has al-ready been counted as an X-ray source.2 J. S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone.
Fig. 14.
Location of di ff erent classes of massive stars superimposed on a greyscale representation of Spitzer µ m continuum data.The location of the Galactic Centre, Arches and Quintuplet clusters are given by green crosses. Given their displacement from thenominal location of the Quintuplet cluster we have also plotted the positions of the LBV G0.120-0.048 and the WN6 star qF353Eeven though they are not included in the census on Table 1. Each panel covers ∼ × ∼ . . S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone. 23 through the CMZ they would be di ffi cult to distinguish fromforeground objects.We may ask whether there is any evidence for such a pro-cess occurring within the GC. Located within 6” of Sgr A ∗ , GCIRS 13E is a compact (d ∼ . . × . ∼
32% of these, no further candidates were found in ourobservations (Sect. 3.9).At larger angular scales the spatial coincidence of a num-ber of very massive, apparently co-eval stars with the Sgr Bstar forming complex is of considerable interest (Fig. 14 andSect. 4.2.2). Sgr B comprises three distinct regions that appearto be embedded within the same molecular cloud; Sgr B1, B2and G0.6-0.0, which bridges the gap between the two formerzones. Observations from mid-IR to radio wavelengths suggestthat Sgr B2 is actively forming stars at this time (e.g. Ginsburget al. 2018a), while Sgr B1 appears older, with bright rimmedand shell-like structures suggestive of the action of stellar winds(Mehringer et al. 1992, Hankins et al. 2020). Simpson et al.(2018) infer T e ff ∼ − / hypergiants co-located with it (Footnote 14;Martins et al. 2008, Lohr et al. 2018). Moreover, such stars gen-erate the powerful stellar winds and ionising radiation fields re-quired to sculpt and disperse the molecular material associatedwith Sgr B1; further strengthening the possibility of a physicalassociation between the two.Simpson et al. (2018) suggest that the exciting sources ofSgr B1 did not originate there, but instead are only now im-pinging on the molecular cloud by virtue of their orbital motion.While proper motion studies will be required to assess whetherthey were born in situ or not, either eventuality implies that theWNHLha and mid-O super- / hypergiants form a genuine, co-evalphysical association. This is particularly interesting since, lyingoutside of the Pa α survey footprint, all six were detected viatheir hard X-ray emission and hence are likely massive binaries.Despite over 50 stars of comparable spectral types being foundwithin the Arches (Clark et al. 2018a, 2019b) only four membersare known X-ray emitters (Wang et al. 2006). While this low de-tection rate may in part be a function of source confusion in thecrowded confines of the cluster, it raises the possibility that the‘Sgr B1’ cohort represents the ‘tip of the iceberg’ in terms of asimilarly rich massive stellar population in this region.Following on from the potential ‘Sgr B1 association’ and,as might be anticipated given the source density in the innerreaches of the GC (Fig. 14), there are multiple further exam-ples of massive stars with comparatively small angular separa-tions. Examples include (but not limited to): P100 (O4-6 Ia + )and P101 (WC8-9); P134 (WN7-8ha) and P135 (Be star); andP35 (WN8-9ha), P36 (O4-5 Ia + ), P111 (WN8-9ha), P112 (late-B HG / LBV) and P114 (O4-5 Ia + ). The spectral classifications -and hence relative ages - of the components in the first two pairsdisfavours a physical association, although this remains a possi-bility for the third group once P112 is excluded due to its evo-lutionary state. Of the remaining four stars only P35 and P114have been subject to further examination; the radial velocity ofthe former suggests that it is a runaway, while the latter appears to have formed formed in situ (and is possibly associated with astellar overdensity; Dong et al. 2017). This implies that P35 and114 are not part of a bound physical system; although in this re-gard it is interesting that P35 appears to be a member of a smallgroup of co-moving stars, possibly the remnant of a disruptedcluster (Shahzamanian et al. 2019). Given the limitations of thecurrent analysis no further conclusions may be drawn as to therelationship of either P36 or P111 to these stars.Finally, two isolated WCL stars - 2MASS J17443734-2927557 and J17444083-2926550 - are seen in close proximityto one another and along the line of sight to the Sgr C H ii re-gion (RA ∼ ◦ ′ → ◦ ′ , δ ∼ ◦ ′ → − ◦ ′ ; Fig. 14).The mid-IR morphology of Sgr C suggests ongoing star forma-tion (Hankins et al. 2020); if these stars are physically associatedwith one other and Sgr C it would extend the duration of such ac-tivity to at least ∼ We are now in a position to construct, and interpret, a censusof evolved massive stars in the GC from extant data, encom-passing both cluster members and isolated stars (Table 3). Giventhe sample size we are able to break down the supergiants, hy-pergiants and WRs by spectral sub-type in order to distinguishstars of di ff ering ages and initial masses. However, we explicitlyexclude cool evolved stars such as 2MASS J17444840-2902163(Table 1), IRS 7 (Wollman et al. 1982) and a number of lower lu-minosity stars of spectral type K-M along the line of sight to theQuintuplet ( ∼ − M ⊙ ; Liermann et al. 2012) since it is di ffi -cult to reliably distinguish these from the foreground populationof cool dwarfs and giants and hence deliver accurate populationstatistics.Observational biases associated with the census of isolatedstars have already been discussed (Sect. 4.1) but it is worthbriefly addressing the limitations of current cluster surveys. Ofthese we expect the population of WNLha stars and O super- / hypergiants within the Arches to be essentially complete (Table3; Clark et al. 2018a, 2019b); conversely the cohort of 42 O gi-ants and dwarfs ( M init & M ⊙ ) is likely to be increasingly in-complete as one moves to lower luminosities. As a consequencewe omit the latter from Table 3, emphasising that our observa-tions of the Quintuplet and the isolated stellar cohort are not suf-ficiently sensitive to identify such objects and hence place therespective populations in context With current spectral classifications only available for ob-jects within the central ∼ ×
40” region of the more di ff use For this reason we also exclude the cohort of isolated classical Bestars from the census at this time (Sect. 3.7), as well as those stars ofuncertain classification (Sect. 3.8).4 J. S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone.
Table 3.
Summary of the population of massive, evolved starswithin the GC broken down by location and spectral type.
Isolated Arches Quint. Gal Cen.Sub-type
Total
O4-6 Ia 5 35 0 0 40O7-8 Ia 0 0 19 0 19O9-B0 Ia 3 0 9 0 12OB Ia 1 0 10 26 37O4-5 Ia + + + + / / LBV / / WC 0 0 0 2 2WC5-6 0 0 0 1 1WC8-9 16 0 14 13 43Numbers derive from this work, Clark et al. (2018a, 2018b, 2019b) andPaumard et al. (2006). We expect the census for the Arches to be largelycomplete for WNLha stars and O hypergiants (Sect. 4.3); however fol-lowing from Bartko et al. (2010) the numbers for the GC cluster arelower limits, while it is almost certain that the same is true for both theQuintuplet cluster members and isolated stars (Sect. 4.3). Given the un-certainty in spectral types for supergiants within the GC cluster we as-sign them a generic OB Ia classification; see text for details. For brevitywe assign the isolated B0-3 Ia star 2MASS J17444501-2919307 to thiscohort and the hybrid star P75 to the WN7-9ha total. See footnotes 5and 11 regarding the totals derived for the Quintuplet; the generic OBIa stars listed for this cluster are the faint cohort identified by Clark etal. (2018b); m F205W > Quintuplet cluster - and being of insu ffi cient depth to reach starsof luminosity class III-V - the expectation is that the current cen-sus will be highly incomplete, even for very luminous sources(Table 3; Clark et al. 2019c). This appears borne out by the lowresolution and S / N spectra presented by Figer et al. (1999) whichimply that several outlying members appear to be additional late-O / early B hypergiants of extreme intrinsic luminosity.Interpretation of the GC cluster is more complex for a num-ber of reasons. Firstly, Bartko et al. (2010) report a further 28WR / O stars and 34 B-dwarfs (their notation) over those de-scribed by Paumard et al. (2006), for a total of 177 early-typestars. Unfortunately no break down of these revised totals byspectral sub-type was presented; hence we are forced to utilisethe incomplete census of the latter work in the construction ofTable 3. Secondly, since there is considerable observational un-certainty in the spectral subtypes assigned to supergiants - rang-ing from ∼ O7-B3 (Paumard et al. 2006) - we simply adopt ageneric OB supergiant classification for the 26 stars of lumi-nosity class I-II pending higher S / N data. While more preciseclassifications are available for the WN and WC stars, modellingsuggests that they are intrinsically fainter than comparable mem- bers of the Quintuplet, implying systematic di ff erences in initialmasses (Martins et al. 2007, Clark et al. 2018b); complicatingany direct juxtaposition of the two samples. Finally, Paumard etal. (2006) identify a cohort of lower luminosity and less evolvedobjects within the Galactic centre cluster, reporting three O7-9giants and 18 dwarfs stars with spectral types of O9 and later.Following the same line of reasoning applied to the comparableArches cohort, we exclude these from the breakdown of stars byspectral type presented in Table 3, as we do for the large numberof additional objects of uncertain spectral type and / or luminosityclass identified by these authors.Despite these limitations, Table 3 lists a total of 259 spec-troscopically classified massive, evolved stars located within theGC. It is immediately apparent that isolated stars contribute sig-nificantly, comprising ∼
26% of the currently identified popu-lation; therefore any discussion of the role of such stars in thewider ecology of the GC must account for their contribution.This is particularly important for the cohort comprising O super- / hypergiants and WN7-9ha stars which, given their stellar prop-erties, are likely to dominate mechanical and radiative feedback(Martins et al. 2008, Doran et al. 2013, Lohr et al. 2018).As such, it is striking that the number of WN7-9ha and O hy-pergiants within the Arches cluster ( M total & M ⊙ ) is directlycomparable to the isolated population of such objects (Table 3).While the current count of mid-O supergiants di ff ers betweenthe two settings, with only five isolated examples identified com-pared to 35 within the Arches, following the discussion in Sect.4.1 we attribute this discrepancy to observational incompletenessin the field population. Likewise, comparison of the numbers ofisolated early-B hypergiants, WN9-11h and WCL stars to thosefound within the comparably massive Quintuplet show that bothpopulations are essentially identical in size, with the dearth ofO7-8 and O9-B0 supergiants (three versus 28; Table 3) againattributable to incompleteness.The presence of a large number of isolated WN5-7 starspoints to an additional population that is under-represented inboth the Arches and Quintuplet (Table 3). Quantitative analysisof two examples - qF353 (WN6; Steinke et al. 2016) and IRS16SE2 (WN5 /
6; Martins et al. 2007) - reveal extreme temper-atures ( T e ff > / or late-B hypergiants is expected to be ratherheterogeneous in terms of initial mass, thus preventing directcomparison to the relevant subset of Quintuplet members.Based on these number counts we suggest that if the isolatedstellar population is drawn from the same (initial) mass functionas that of the Arches and Quintuplet clusters, then it should ri-val the combined stellar content of these aggregates. However,allowing for incompleteness and following suggestions that thequiescent star formation within the GC has proceeded at an es-sentially constant rate for the past 5-10Myr ( ∼ . M ⊙ yr − ;Barnes et al. 2017) it is likely that population of isolated massivestars dominates that of the currently identified stellar clusters. . S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone. 25 Even in the case that the isolated stars are drawn from an unex-pectedly top heavy mass function - unlikely given the presenceof isolated classical Be stars and low luminosity red and bluesupergiants - the large number of O super- / hypergiants, WN7-9ha and WN5-7 stars identified to date confirms that this cohortwill play an important role in the ecology of the GC via radiativefeedback.Nevertheless, upon consideration of all the above studieswe arrive at a total of 437 spectroscopically classified early-type / massive stars within the CMZ. These comprise 177 starswithin the central cluster (Bartko et al. 2010), 105 within theArches (Clark et al. 2019b), 72 within the Quintuplet (Clark etal. 2018b and footnote 11) and 83 isolated examples (Sect. 3.10).In the absence of reliable mass functions for both clustersand isolated stars, it is premature to employ these data to es-timate the global star formation rate for the CMZ, which willbe dominated by the low stellar mass component. However wemay attempt to place useful lower limits to the SN rate fromthese number counts. With masses of & M ⊙ expected for theO9.5 V stars within the Arches (Clark et al. 2018a, 2019b) and ∼ − M ⊙ for both the isolated classical Be stars and the B0-3 Vstars within the Galactic Centre cluster (Habibi et al. 2017) onewould expect a minimum of ∼
322 ( & ,implying a time-averaged rate of ∼ × − yr − . More realis-tic assumptions lead to higher rates; for example supposing the65 H-depleted WRs undergo core collapse in the next ∼ . ∼ × − yr − , with a similar number returnedif we also include the mid-O super- / hypergiants, early-B hyper-giants, WNLha, and WNLh stars and allow for a greater time un-til supernova (under the assumption that such stars derive from M init ∼ M ⊙ and will undergo core-collapse within ∼ & × − yr − ,which in turn is consistent with a SN rate of ∼ − yr − de-rived from, variously, radio observations (Lazio & Cordes 2008),SN remnant counts (Ponti et al. 2015) and simulations of massand energy flows through the innermost ∼
5. Conclusions and future prospects
We have presented the results of an extensive K − band spectro-scopic survey designed to characterise the population of isolatedmassive stars within the GC. The resultant dataset enabled theidentification of 17 new objects and the reclassification of an ad-ditional 19 known examples; a further 11 stars retained extantclassifications while a large number of candidates were found tobe cool, foreground sources. Including previous identificationsyields a total of 83 isolated massive stars, of which the vast ma-jority are evolved objects.Given the nature of the surveys utilised to construct the targetlist, the census is heavily biased towards objects with strong stel-lar winds - such as OB hypergiants and both H-rich and H-freeWRs - to the extent that even stars as extreme mid-O supergiants Exceptions, which are limited to the GC cluster, include five B4-9 V stars as well as the large number of early-type stars of uncertainclassification. are significantly under-represented. Cool supergiants do not fea-ture due to the di ffi culty in distinguishing them from foregroundobjects. Amongst the WRs, a large number of H-free WN5-7stars were identified, of particular interest since they are almostentirely absent from both the Arches and Quintuplet clusters.We also report the detection of classical Be stars; given that theBe phenomenon is limited to spectral type ∼ O9 III-V and later,these are likely to be the least massive stars ( M init . M ⊙ ) inour census. Rare transitional objects such as (candidate) sgB[e]and LBVs were also recognised; as a consequence this popula-tion will be invaluable for constraining massive stellar evolution- even for very rapid phases - a possibility exemplified by P75,which appears to be the first example of an O hypergiant transi-tioning to a WNLha stage. Intriguingly, a relatively large numberof stars appear to be candidate binaries by virtue of their multi-wavelength properties (Sect. 4.2.2), although the target selectioncriteria employed (i.e. X-ray or IR luminosity; Sect. 4.1) willpreferentially identify such systems.Isolated massive stars are found throughout the GC, althoughthe distribution of mid-O super- / hypergiants and WNLha stars -so the youngest and most massive cohort sampled - is signifi-cantly asymmetric as a function of galactic longitude. In partic-ular there is an overdensity of such objects spatially coincidentwith the young H ii complex Sgr B1; if physically associated theywould imply star formation has proceeded in this region for atleast 2Myr. Intriguingly all these stars were identified via theirhard X-ray emission; by comparison only four from fifty stars ofsimilar spectral types within the Arches cluster are X-ray bright;on this basis one might therefore anticipate that this region hostsan exceptionally rich OB association. No further clusters com-parable to the Arches or Quintuplet were identifiable, despite anexpectation that they should be visually prominent for at leastthe first ∼ / early-B hyper-giants and WCL stars and those within the Quintuplet. It is cur-rently uncertain whether the mode(s) of star formation that gen-erate the high mass clusters and apparently isolated massive stars(or sparse, low mass aggregates) yield a single, universal IMF;if this is the case it would imply an isolated stellar populationthat is, at a minimum, of comparable size to that of both clus-ters combined. In total ∼
437 massive cluster members and iso-lated stars have been identified spectroscopically within the GC;an unprecedented number in comparison to other resolved starforming regions. Of these &
74% will undergo core-collapse, implying a lower limit on the SN rate of & × − yr − ; in rea-sonable agreement with other direct and indirect estimates butlikely a significant underestimate given the incompleteness ofthe current surveys.Moving forward, our homogeneous, high S / N and resolutionspectroscopic dataset will permit tailored quantitative analysisfor individual stars in order to derive their underlying physi-cal parameters (cf. Clark et al. 2018a, 2018b); a significant by-product of which will be the determination of (di ff erential) inter-stellar reddening along multiple sightlines to the GC (cf. Geballeet al. 2019). Such a methodology will allow us to populate HRdiagrams for both cluster and isolated cohorts. This is an essen-tial first step in exploiting the potential of these data to improveour understanding of massive stellar evolution as well as deter-mining the bulk properties of the populations. Such an analysiswill enable recent star formation across the GC to be quantita-tively constrained (cf. 30 Dor; Schneider et al. 2018) and, criti-cally, will permit calibration of cluster luminosity functions de-rived from extant HST photometry in order to construct (initial)mass functions. In parallel abundance determinations - particu-larly of α -elements - will help determine whether historical ac-tivity proceeded via a mode characterised by the production ofa normal or top-heavy mass function (cf. Najarro et al. 2009).Combining the spectroscopic dataset with proper motions de-rived from HST observations will allow us to identify co-movingstellar groups as well as identify high velocity massive runawaysand, in turn, the relative proportions of massive stars that formedin true isolation.Nevertheless, additional observations are required in orderto elucidate the nature and yield of recent star formation activityin the GC. The list of candidates returned by both the Paschen α and X-ray surveys remain to be fully exploited (cf. AppendixA), while an extension of the former to include regions such asSgr B1 would be invaluable. Fortuitously, despite the incom-plete nature of the current census(es), observations of the 30Dor star forming complex (Doran et al. 2013) suggest that theobjects identified - and identifiable via such an approach - areexpected to dominate radiative feedback within the CMZ dueto their extreme luminosities (e.g. WNLha stars and O super- / hypergiants) and / or temperatures (e.g. WN5-7 stars). However,we are almost entirely insensitive to the expected population ofmoderately massive ( ∼ − M ⊙ ) stars that, for any reasonableIMF, will comprise the majority of core collapse candidates.Photometric pre-selection followed by a proper motion cutwill be necessary to construct a suitable candidate list for ex-haustive multi-object spectroscopic follow up. Although timeintensive, such an approach will be required in order to deter-mine the recent star formation rate within the GC - and conse-quently the production rate of relativistic remnants - as well asthe rate of feedback of mechanical energy and chemically en-riched material via stellar winds and SNe. This is essential toour understanding of the relative contributions of massive starsand the supermassive black hole Sgr A ∗ to the energy budget andecology of the GC, in terms of regulating the progression of starformation, the generation of very high energy (GeV) γ -ray emis-sion, the inflation of both X-ray and radio out-of-plane bubblescentred on the GC, and the production of PeV cosmic rays. Acknowledgements.
This research was supported by the Science andTechnology Facilities Council. F.N. acknowledges financial support throughSpanish grants ESP2017-86582-C4-1-R and PID2019-105552RB-C41(MINECO / MCIU / AEI / FEDER) and from the Spanish State Research Agency(AEI) through the Unidad de Excelencia “Mar´Ia de Maeztu”-Centro deAstrobiolog´Ia (CSIC-INTA) project No. MDM-2017-0737. L.R.P. ac-knowledges support from the Generalitat Valenciana through the grantPROMETEO / / References
An, D., Ramirez, S. V., Sellgren, K. et al. 2011, ApJ, 736, 133Aharonian, F., Akhperjanian, A. G., Bazer-Bachi, A. R. et al. 2006, Nature, 439,695Aharonian, F., Yang, R. & de Ona Wilhelmi, E. 2018 (arXiv:1804.02331)Banerjee, S., Kroupa, P. & Oh, S. 2012, ApJ, 746, 15Barnes, A. T., Longmore, S. N., Battersby, C., et al. 2017, MNRAS, 469, 2263Bartko, H., Martins, F., Trippe, S., et al. 2010, ApJ, 708, 834Bednarek, K., Pabich, J. & Sobczak, T. 2014, PhRvD, 90, 3008Bestenlehner, J. M., Vink, J. S., Gr¨afener, G., et al. 2011, A&A, 530, L14Bik, A., Kaper, L., Hanson, M. M. & Smits, M. 2005, A&A, 440, 121Bik, A., Kaper, L. & Waters, L. B. F. M. 2006, A&A, 455, 561Blaauw, A. 1961 Bull. Astron. Inst. Netherlands, 15, 265Bonanos, A. Z., Massa, D. L., Sewillo, M. et al. 2009, AJ, 138, 1002Borgman, J., Koornneef, J. & Slingerland, J. 1970, A&A, 4, 248Bressert, E., Bastian, N., Evans, C. J. et al. 2012, A&A, 542, A49Bykov, A. M., Ellison, D. C., Gladilin, P. E., & Osipov, S. M. 2015, MNRAS,453, 113Caswell, J. L., Haynes, R. F. & Goss, W. M. 1977, MNRAS, 181, 427Caswell, J. L., Fuller, G. A., Green, J. A., et al. 2010, MNRAS, 404, 1029Cesarsky, C. J. & Montmerle, T. 1983, SSRv, 36, 173Clark, J. S. & Steele, I. A. 2000, A&AS, 141, 65Clark, J. S., Negueruela, I., Crowther, P. A. & Goodwin, S. P. 2005, A&A, 434,949Clark, J. S., Negueruela, I., Davies, B. et al. 2009a, A&A, 498, 109Clark, J. S., Crowther, P. A. & Mikles, V. J. 2009b, A&A, 507, 1567Clark, J. S., Arkharov, A., Larionov, V. et al. 2011, BSRSL, 80, 361Clark, 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., Ritchie, B. W., Najarro, F., Langer, N. & Negueruela, I. 2014, A&A,565, A90Clark, J. S., Lohr, M. E., Najarro, F., Dong, H. & Martins, F. 2018a, A&A, 617,A65Clark, 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., & Najarro, F. 2019b, A&A, 623, A84Clark, J. S., Ritchie, B. W. & Negueruela, I. 2019c, 626, A59Clark, J. S., Ritchie, B. W., Negueruela, I. 2020, A&A, 635, A187Contreras, Y., Schuller, F., Urquhart, J. S. et al. 2013, A&A, 549, A45Cotera, A. S., Erickson, E. F., Colgan, S. W. J. et al. 1996, ApJ, 461, 750Cotera, A. S., Simpson, J. P., Erickson, E. F. et al. 1999, ApJ, 510, 747Crocker, R. M. 2012, MNRAS, 423, 3512Crowther, P. A., Hadfield, L. J., Clark, J. S., Negueruela, I. & Vacca, W. D. 2006,MNRAS, 372, 1407Crowther, P. A. & Walborn, N. R. 2011, MNRAS, 416, 1311Crowther, P. A., Caballero-Nieves, S. M., Bostroem, K. A. et al. 2016, MNRAS,458, 624Cunha, K., Sellgen, K., Smith, V. V. et al. 2007, ApJ, 669, 1011Cutri, R. M., Skrutskie, M. F., van Dyk, S. et al. 2003, VizieR Online DataCatalog: 2MASS All-Sky Catalog of Point SourcesDavies, B., Figer, D. F., Kudritzki, R.-P., et al. 2007, ApJ, 676, 1016Davies, B., Figer, D. F., Law, C. J. et al. 2008, ApJ, 676, 1016Davies, R. I., Agudo Berbel, A., Wiezorrek, E., et al. 2013, A&A, 558, A56Deneva, J. S., Cordes, J. M., & Lazio, T.J. W. 2009, ApJ, 702, L177de Wit, W. J., Testi, L., Palla, F., Vanzi, L. & Zinnecker, H. 2004, A&A, 425,937de Wit, W. J., Testi, L., Palla, F. & Zinnecker, H. 2005, A&A, 437, 247de Witt, C., Bandyopadhyay, R. M., Eikenberry, S. S., et al. 2013, AJ, 146, 109Diederik Kruijssen, J. M. & Longmore, S. N. 2013, MNRAS, 435, 2598Dong, H., Wang, Q. D., Cotera, A. et al. 2011, MNRAS, 417, 114Dong, H., Wang, Q. D. & Morris, M. R. 2012, MNRAS, 425, 884Dong, H., Mauerhan, J., Morris, M. R., Wang, Q. D. & Cotera, A. 2015,MNRAS, 446, 842Dong, H., Lacy, J. H., Sch¨odel, R. et al. 2017, MNRAS, 470, 561Doran, E. I., Crowther, P. A., de Koter, A. et al. 2013, A&A, 558, A134Dougherty, S. M., Waters, L. B. F. M., Burki, G. et al. 1994, A&A, 290, 609Dougherty, S., Clark J. S., Negueruela, Johnson, T. & Chapman, J. M. 2010,A&A, 511, A58Egan, M. P., Clark, J. S., Mizuno, D. R. et al. 2002, ApJ, 572, 288Everett, J. E., Zweibel, E. G., Benjamin, R. A. et al. 2008, ApJ, 7674, 258Fenech, D. M., Clark, J. S., Prinja, R. K. et al. 2018, A&A, 617, A137Figer, D. F., McLean, I. S. & Najarro, F. 1997, ApJ, 486, 420Figer, D. F., McLean, I. A. & Morris, M. 1999, ApJ, 514, 202Figer, D. F., MacKenty, J. W., Robberto, M. et al. 2006, ApJ, 643, 1166Fritz, T. K., Gillessen, S., Dodds-Eden, K. et al. 2010, ApJ, 721, 395Fujii, M. S. & Portegies Zwart, S. 2011, Science, 334, 1380Geballe, T. R., Najarro, F., Rigaut, F. & Roy, J.-R. 2006, ApJ, 652, 370 . S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone. 27
Geballe, T. R., Lambrides, E., Schlegelmilch, B. et al. 2019, ApJ, 872, 103Ginsburg, A., Bally, J., Barnes, A. et al. 2018a, ApJ, 853, 171Ginsburg, A. & Kruijssen, J. M. D. 2018b, ApJ, 864, L17Goodwin, S. P., & Bastian, N. 2006, MNRAS, 373, 752Groh, 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, A30Habibi, M., Stolte, A. & Harfst, S. 2014, A&A, 566, A6Habibi, M., Gillessen, S., Martins, F. et al. 2017, ApJ, 847, 120Hailey, C. J., Mori, K., Bauer, F. E. et al. 2018, Nature, 556, 70Hankins, M. J., Lau, R. M., Radomski, J. T. et al. 2010, arXiv:2001.05487Hanson, M. M., Conti, P. S. & Rieke, M. J. 1996, ApJS, 107, 281Hanson, M. M., Kudritzki, R.-P., Kenworthy, M. A., Puls, J. & Tokunaga, A. T.2005, ApJS, 161, 154H.E.S.S. Collaboration, 2016, Nature, 531, 476H.E.S.S. Collaboration, 2018, A&A, 612, A9Heywood I., Camilo, F., Cotton, W. D. et al. 2019, Nature, 573, 235Hillier, D. J. & Miller, D. L. 1998, ApJ, 496, 407Hillier, D. J. & Miller, D. L. 1999, ApJ, 519, 354Homeier, N. L., Blum, R. D., Pasquali, A., Conti, P. S., Damineli, A. 2003, A&A,408, 153Hussmann, B., Stolte, A., Brandner, W., Gennaro, M. & Liermann A. 2012,A&A, 540, A57Immer, K., Schuller, F., Omont, A. & Menten, K. M. 2012a, A&A, 537, A121Immer, K., Menten, K. M., Schuller, F. & Lis, D. C. 2012b, A&A, 548, A120Kaper, L., van Loon, J. Th., Augusteijn, T. et al. 1997, ApJ, 475, L37Kastner, J. H., Buchanan, C., Sahai, R., Forrest W. J. & Sargent, B. A. 2010, AJ139, 1993Kendrew, S., Ginsburg, A., Johnston, K. et al. 2013, ApJ, 775, L50Kennea, J. A., Burrows, D. N., Kouveliotou, C., et al. 2013, ApJ, 770, L24Kraus, M., Arias, M. L., Cidale, L. S. & Torres, A. F. 2020, MNRAS, 493, 4308Lada, C. J., & Adams, F. C. 1992, ApJ, 393, 278Lamb, J. B., Oey, M. S., Segura-Cox, D. M. et al. 2016, ApJ, 817, 113Lamers, H. J. G. L. M., Zickgraf, F.-J., de Winter, D., Houziaux, L. & Zorec, J.1998, A&A, 340, 117Lazio, T. J. W., & Cordes, J. M., 2008, ApJS, 174, 481Lennon, D. J., Evans, C. J., van der Marel, R. P. et al. 2018, A&A, 619, A78Le Petit, F., Rauad, M., Bron, E. et al. 2016, A&A, 585, A 105Libralato, M., Lennon, D. J., Bellini, A. et al. 2021, MNRAS, 500, 3213Liermann, A., Hamann, W.-R. & Oskinova, L. M. 2012, A&A, 540, A14Lohr, M. E., Clark, J. S., Najarro, F., et al. 2018, A&A, 617, A66Longmore, S. N., Bally, J., Testi, L. et al. 2013, MNRAS, 429, 987Longmore, S. & Kruijssen, J. M. 2018, Galax, 6, 55Martins, F., Genzel, R., Hillier, D. J., et al. 2007 A&A, 468, 233Martins, F., Hillier, D. J., Paumard, T. et al. 2008, A&A, 478, 219Martins, F. & Palacios, A. 2017, A&A, 598, A56Matsunaga, N., Kawadu, T., Nishiyama, S. et al. 2009, MNRAS, 399, 1709Mauerhan, J. C., Muno, M. P. & Morris, M. 2007, ApJ, 662, 574Mauerhan, J. C., Muno, M. P., Morris, M. R. et al. 2009, ApJ, 703, 30Mauerhan, J. C., Morris, M. R., Cotera, A., et al. 2010a, ApJ, 713, L33Mauerhan, J. C., Cotera, A., Dong, H. et al. 2010b, ApJ, 725, 188Mauerhan, J. C., Muno, M. P., Morris, M. R., Stolovy, S. R. & Cotera, A. 2010,ApJ, 710, 706Mehringer, D. M., Yusef-Zadeh, F., Palmer, P. & Goss, W. M. 1992, ApJ, 401,168Moneti, A., Glass, I. S., & Moorwood, A. F. M., 1992, MNRAS, 258, 705Moneti, A., Glass, I. S., & Moorwood, A. F. M., 1994, MNRAS, 268, 194Mori, K., Gotthelf, E. V., Zhang, S. et al. 2013, ApJ, 770, L23Morris, M. & Serabyn, E. 1996, ARA&A, 34, 645Morris, P. W., Eenens, P. R. J., Hanson, M. M., Conti, P. S. & Blum, R. D. 1996,ApJ, 470, 597Morris, M., Bagano ff , F., Mumno, M. et al. 2003, ANS, 324, 167Muno, M. P., Bower, G. C., Burgasser, A. J., et al. 2006a, ApJ, 638, 183Muno, M. P., Law, C., Clark, J. S. et al. 2006b, 650, 203Muno, M. P., Bauer, F. E., Bagano ff , F. K. et al. 2009, ApJS, 181, 110Najarro, F., Figer, D. F., Hillier, D. J., Geballe, T. R., & Kudritzki, R. P. 2009,ApJ, 691, 1816Najarro, F., de la Fuente, D., Geballe, T. R., Figer, D. F. & Hillier, D. J. 2015,IAUS, 307, 426.Najarro, F., Geballe, T. R., Figer, D. F. & de la Fuente, D. 2017, ApJ, 845, 127Negueruela, I., Gonz´alez-Fernandez, C., Marco, A., Clark, J. S. & Mart´ınez-N´u˜nez, S. 2010, A&A, 513, A74Oh, S., & Krupa, P. 2016, A&A, 590, A107Oka, T., Gebalkle, T. R., Goto, M. et al. 2019, ApJ, 883, 54Oksala, M. E., Kraus, M., Cidale, L. S., Muratore, M. F. & Borges Fernandes,M. 2013, A&A, 558, A17Oskinova, L. M., Steinke, M., Hamann, W.-R., et al. 2013, MNRAS, 436, 3357Park, S-M., Goodwin, S. P. & Kim, S. S. 2018, MNRAS, 478, 183Park, S-M., Goodwin, S. P. & Kim, S. S. 2020, MNRAS, 494, 325 Parker, R. J. & Goodwin, S. P. 2007, MNRAS, 380, 1271Patrick, L. R., Evans, C. J., Davies, B. et al. 2015, ApJ, 803, 14Patrick, L. R., Evans, C. J., Davies, B. et al. 2017, ApJ, 468, 492Paumard, T., Genzel, R., Martins, F et al. 2006, ApJ, 643, 1011Ponti, G., Morris, M. R., Terrier, R. et al. 2015, MNRAS, 453, 172Ponti, G., Hofmann, F., Churazov, E. et al. 2019, Nature, 567, 347Portegies Zwart, S. F., Makino, J., McMIllan, S. L. W. & Hut, P. 2002, ApJ, 565,265Porter, J. M. & Rivinius, T. 2003, PASP, 115, 1153Poveda, A., Ruiz, J. & Allen, C. 1967, Bol. Obs. Tonantzintla Tacubaya, 4, 860Renzo, M., Zapartas, E., de Mink S. E., et al. 2019, A&A, 624, A66Robitaille, T. P., Meade, M. R., Babler, B. L. et al. 2008, AJ, 136, 2413Rosslowe, C. K. & Crowther, P. A. 2018, MNRAS, 473, 2853Schneider, F. R. N., Ram´ırez-Agudelo, Tramper, F. et al. 2018, A&A, 618, A73Shahzamanian, B., Sch¨odel, R., Nogueras-Lara, F., et al. 2019, A&A, 632, A116Sharples, R., Bender, R. & Agudo Berbel, A., et al. 2013, Messenger, 151, 21Simpson, J. P., Colga, S. W. J., Cotera, A. S., Kaufman, M. J. & Solovy, S. R.2018, ApJ, 867, L13Steinke, M., Oskinova, L. M., Hamann, W.-R. et al. 2016, A&A, 588, A9Stolte, A., Ghez, A. M., Morris, M. et al. 2008, ApJ, 675, 1278Stolte, A., Morris, M. E., Ghez, A. M., et al. 2010, ApJ, 718, 810Stolte, A., Hußmann, B., Morris, M. R. et al. 2014, ApJ, 789, 115Su, M., Slayter, T. R. & Finkbeinder, D. P 2010, ApJ, 724, 1044Tuthill, P., Monnier, J., Tanner, A., et al. 2006, Science, 313, 935Wang, Q. D., Dong, H. & Lang, C. 2006, MNRAS, 371, 38Wang, Q. D., Dong, H., Cotera, A. et al. 2010, MNRAS, 2010, MNRAS, 402,895Wisniewski, J. P. & Bjorkman, K. S. 2006, ApJ, 652, 458Wright, N. J., Parker, R. J., Goodwin, S. P. & Drake, J. J. 2014, MNRAS, 438,639Wright, N. J., Bouy, H., Drew, J. E., et al. 2016, MNRAS, 460, 2593Wollman, E. R., Sith, H. A., Larson, H. P. 1982, ApJ, 258, 506Yamamuro, T., Nishimaki, Y., Motohara, K., Miyata, T. & Tanaka, M. 2007,PASJ, 59, 973Yusef-Zadeh, F., Hewitt, J. W., Arendt, R. G. et al. 2009, ApJ, 702, 178Yusef-Zadeh, F., Lacy, J. H., Wardle, M. et al. 2010, ApJ, 725, 1429Yusef-Zadeh, F. & Wardle, M. et al. 2019, MNRAS, 490, L1Zhao, J.-H., Morris, M. R. & Goss, W. M. 2016, ApJ, 817, 171Zhao, J.-H., Morris, M. R. & Goss, W. M. 2020, ApJ, in press. Appendix A: Incomplete exploitation of the Pa α excess survey After excluding clusters members and those isolated stars withspectroscopic observations, 27 objects from the primary list of152 Paschen α excess sources of Dong et al. (2011) remain unob-served. Of these six have colours indicative of foreground stars(( H − K ) < K <
13 which, excluding classical Be stars, characterise the ma-jority of massive stars in our survey (Sect. 3).The list of 189 secondary Pa α excess targets is much morepoorly characterised. 101 of the isolated stars from this rosterlack both spectroscopic and photometric observations so we maynot assign them to the CMZ, let alone classify them. Of the iso-lated stars currently lacking spectroscopic observations but withnear-IR photometric data, 31 appear to have colours consistentwith foreground objects (S48, 50, 51, 53, 54, 68, 70, 71, 72,76, 77, 83, 84, 86, 89, 92, 93, 95, 97,99, 101, 109, 118, 121,128, 145, 148, 149, 150, 154, and 170) leaving 33 possible CMZmembers (S47, 49, 52, 60, 64, 65, 78, 79, 88, 91, 103, 106, 110,113, 114, 117, 119, 123, 125, 136, 146, 151, 153, 158, 161, 168,169, 173, 178, 182, 183, 185, 187). Of the latter, ten have K <8 J. S. Clark et al.: Constraining the population of isolated massive stars within the Central Molecular Zone.