The star-forming complex LMC-N79 as a future rival to 30 Doradus
Bram B. Ochsendorf, Hans Zinnecker, Omnarayani Nayak, John Bally, Margaret Meixner, Olivia C. Jones, Remy Indebetouw, Mubdi Rahman
TThe star-forming complex LMC-N79 as a future rival to 30Doradus
Bram B. Ochsendorf , ∗ , Hans Zinnecker , , Omnarayani Nayak , John Bally , Margaret Meixner , ,Olivia C. Jones , Remy Indebetouw , & Mubdi Rahman Department of Physics and Astronomy, The Johns Hopkins University, 3400 North Charles Street,Baltimore, MD 21218, USA, [email protected] Deutsches SOFIA Institut (DSI), University of Stuttgart, Pfaffenwaldring 29, D-70569, Germany Universidad Autonoma de Chile, Santiago de Chile, Chile Astrophysical and Planetary Sciences Department, University of Colorado, UCB 389 Boulder,Colorado 80309, USA Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA Department of Astronomy, University of Virginia, PO 400325, Charlottesville, VA 22904, USA National Radio Astronomy Observatory, 520 Edgemont Rd, Charlottesville, VA 22903, USA
Within the early Universe, ‘extreme’ star formation may have been the norm rather than theexception
1, 2 . Super Star Clusters (SSCs; M (cid:63) (cid:38) M (cid:12) ) are thought to be the modern-dayanalogs of globular clusters, relics of a cosmic time ( z (cid:38)
2) when the Universe was filled withvigorously star-forming systems . The giant HII region 30 Doradus in the Large MagellanicCloud (LMC) is often regarded as a benchmark for studies of extreme star formation . Here,we report the discovery of a massive embedded star forming complex spanning ∼
500 pc inthe unexplored southwest region of the LMC, which manifests itself as a younger, embeddedtwin of 30 Doradus. Previously known as N79, this region has a star formation efficiencyexceeding that of 30 Doradus by a factor of ∼ (cid:46) The LMC is the prototypical ‘Barred Magellanic Spiral’, a population of galaxies with anasymmetric, sometimes off-centered stellar bar, a single spiral arm, and often a large star formingcomplex at one end of the bar . More recently, evidence of multiple arm-like features extendingfrom the outer disc of the LMC were obtained with high-resolution H I maps , thought to originatefrom tidal interactions with both the Galaxy and Small Magellanic Cloud (SMC) (Fig. 1a). Inparticular, at heliocentric velocities of ∼
255 - 270 km s − (Fig. 1b) the LMC resembles a barredspiral galaxy with two prominent, opposing arms extending from the eastern (Arm E) and western(Arm W) part of the H I disk.Arm E culminates in the ‘south-eastern H I overdensity’ located at the leading edge of theLMC’s motion through the Galactic Halo . At the tip of the south-eastern H I overdensity andArm E lies 30 Doradus, harboring the largest H II region in the Local Group. Ionized gas tracesmassive star populations with a median age of ∼ . In addition, with the use of sensitive,1 a r X i v : . [ a s t r o - ph . GA ] O c t M to Bridge & SMC B a r to LA (a) N7930 Dor A r m E A r m W N11 (b) (c)
N44
Figure 1:
Large-scale structure of the LMC. (a) : H I map of the LMC. Black and orange mark-ings highlight locations of prominent features in the LMC and the larger-scale Magellanic complex,such as the assymetrical, off-centered optical stellar bar (see Fig. 2), and tidal arms E and W. ArmE culminates in the south-eastern H I overdensity, splits at the LMC tidal radius, and subsequentlyleads to the Magellanic Bridge/SMC and the Leading Arm (LA). The proper motion (PM) of theLMC through the Galaxy is also indicated. In orange we highlight the rotation of the LMC disk and the locations of several prominent star forming regions. (b) : Same as (a) , but here we overlay ∼ (c) : SameH I map, but only showing the velocity range 257 - 270 km s − to accentuate the ‘barred spiral’appearance of the LMC. Here, we overlay the subset of massive YSOs (MYSOs; M > M (cid:12) ).Luminous, embedded, and extremely young, these MYSOs offer a snapshot to the massive starformation activity of the LMC averaged over the past (cid:46) .galaxy-wide IR surveys of the LMC with Spitzer and
Herschel , some ∼ massive YSOs (MYSOs) that are well characterized by YSO models. ThisMYSO catalogue is tested to be complete for massive (
M > M (cid:12) ), young ( (cid:46) most recent massive star formation activity of theLMC .In Fig. 1b, we overplot the H I map of the LMC with the location and clustering of YSOcandidates found across the galaxy. Several obvious clusterings stand out: the stellar bar (whichlikely contains many false positives; see Methods), and the well-known star forming regions 30Doradus, N11 and N44. In addition, Fig. 1b reveals a star forming complex in the relativelyunexplored southwest region of the LMC, which coincides with the N79 H II region . Figure1c plots the subsample of MYSOs over the H I gas at a velocity range 257 - 270 km s − , whichhighlights the tidal arms of the LMC. Most interestingly, both 30 Doradus and N79 are perched on2 α30 Dor N79(a) B a r
24 μm8 μmHα
500 pc
N79-S N79-W(b)
N79-EN11 l ead i ng edge s p i r a l a r m (c) Figure 2:
Dissecting N79. (a) : H α image of the LMC. Highlighted are N79 (grey box), 30 Do-radus, N11, and the stellar bar. (b) : Blow-up of the N79 region in a three-color image showingH α ( blue ), Spitzer/IRAC 8 µ m ( green ), and Spitzer/MIPS 24 µ m ( red ). White contours show COclouds from the MAGMA survey, where we identify three main CO sub-complexes: N79-South(S), N79-East (E), and N79-West (W) (c) : The CO-based ( filled black contour ) and dust-based( grayscale ) molecular mass in N79, overplotted with the location of MYSOs ( inverted red tri-angles ; size reflects source luminosity). Also plotted is the H I gas ( blue contours ). While COpeaks in distinct regions, harboring apparent clusterings of MYSOs, the entire complex is bridgedthrough molecular gas as traced by dust, which is sensitive to the extended, more diffuse envelopesof GMCs .the leading edges or ‘tips’ of the opposite tidal arms E and W, respectively.The N79 H II region pales in comparison to optically bright star forming regions such as N11or 30 Doradus (Fig. 2a). Hence, N79 has not been the subject of any prior high-resolution study.However, our IR observations trace the younger, more embedded phase of massive star formationand unveil that the N79 region is a highly efficient star forming engine, exceeding the star formationefficiency of 30 Doradus and N11 by a factor of ∼ Spitzer and
Herschel dissect the structure of the complex, spanning roughly 500 pc, and har-boring three main CO complexes: N79-South, N79-East, and N79-West (Fig. 2b). The CO (1-0)tracer is known to probe a limited range in volume densities of molecular gas because of criticaldensity, depletion, opacity, and photo-chemical effects. In addition, at the reduced metallicity ofthe LMC, a significant part of H may be in a ‘CO-dark’ phase . By combining far-infrared dustemission and H I one can circumvent these limitations and estimate the H distribution . Thedust-based molecular material (Fig. 2c) shows that the entire N79 region consist of one singlemolecular structure of ∼
500 pc. This unusual large size may be the result of gas accumulation andcompression at the tip of Arm W. Star formation concentrates within the molecular material, withapparent clusterings in the CO-emitting clouds (Fig. 2c).At the heart of the large-scale N79 complex lies an extremely luminous object (Fig. 3a), whichimmediately draws parallels to the central cluster of 30 Dor, R136 (Nayak et al., in prep). Thissource has been catalogued as HSOBMHERICC J72.971176-69.391112, but will be referred toas ‘H72.97-69.39’. At L IR (cid:39) × L (cid:12) (Fig. 3b), H72.97-69.39 is more luminous than any3 b) H72.97-69.39: proto-R136? 1’15 pc (a)
Figure 3:
H72.97-69.39. (a) : The immediate environment of the compact luminous object at theheart of N79, H72.97-69.39, possibly a precursor to the R136 cluster in 30 Doradus (Nayak etal., in prep). (b) : Spectral energy distribution of H72.97-69.39 compiled from various groundand space-based surveys (see Methods). A two-temperature modified blackbody yields an infraredluminosity of L IR = 2.2 × L (cid:12) .MYSO or compact H II region discovered with large-scale IR surveys of the LMC and MilkyWay . This luminosity is equivalent to more than three O3V stars of M ∼
70 M (cid:12) or a single verymassive star of ∼
160 M (cid:12) , using the mass-luminosity relation for upper-main sequence stars , L ∝ M . (at M (cid:38)
70 M (cid:12) ).We measure the star formation characteristics of N79 within an aperture of increasing sizecentered on H72.97-69.39 and compare this with 30 Doradus and N11. The total SFR
MYSO isobtained by counting MYSOs and using an initial mass function (IMF) and characteristic age( t (cid:63) (cid:46) MYSO with the SFR measured through H α and24 µ m emission, SFR H α (see Methods), which allows us to compare the average SFR over the past ∼ ∼ MYSO of N11, while being a factor of ∼ and shows SFR MYSO ≈ SFR H α , thus sustaining its average SFR over thepast ∼ MYSO < SFR H α , consistent with its inferred star formationhistory, which dramatically accelerated roughly ∼ . Conversely, the SFR in N79 has significantly increased over the pastfew Myr (SFR MYSO > SFR H α ) and has yet to reach its peak star formation activity. N79 may4 .2 1.4 1.6 1.8 2.0 2.2 2.4log[ R (pc)]3.03.54.04.55.0 l o g [ S F R ( M fl M y r − )] SFR
N7930DorN11 R (pc)]3.02.52.01.51.00.5 l o g [ S F R m Y S O / M H ( M y r − )] SFE = SFR/ M gas N7930DorN11 R (pc)]3.02.52.01.51.00.5 l o g [ S F R m Y S O / M H + H I + H II ( M y r − )] SFE = SFR/ M gas N7930DorN11
Figure 4:
Star formation properties: N79 versus N11 and 30 Doradus. (a) : The SFR as mea-sured by MYSO counting, SFR
MYSO , in apertures of radius R centered on H72.97-69.39 (RA =72.972, DEC = -69.391), N11 (RA = 74.227, DEC = -66.368), and R136 (RA = 84.633, DEC= -69.092). The asterisks marks the SFR of the regions as measured by H α , SFR H α , where thesize R of the H II regions is defined to enclose 90% of the total flux from the central clusters .Error bars are dominated by multiplicity (SFR MYSO ) or stochastic sampling of the IMF (SFR H α ;see Methods). (b): The inverse of the gas depletion time, SFR/ M gas , where the gas includes onlythe molecular component. The error bar shows the absolute uncertainty, dominated by systematicuncertainties in determining the molecular gas mass. However, the relative uncertainties are ex-pected to be lower (see Methods). (c): Same as (b), but now including the molecular, neutral, andionized components.therefore be in a similar accelerating star formation phase 30 Doradus was ∼ M gas . This quantity providesa measure of the timescale to exhaust the available gas reservoir at the current SFR (assuming allgas would be converted into stars). While the molecular clouds in the LMC are associated withH I envelopes , it is unclear which fraction of M H I will eventually be available for star formation.Therefore, we consider two cases. First, we take M gas = M H , i.e., we only take into account themolecular (dust-based) material (Fig. 4b). Second, we assume M gas = M H + M H I + M H II . Bycombining the molecular (dust-based), neutral, and ionized gas, we attempt to estimate an upperlimit to the available gas reservoir for star formation, while tracing gas which may have alreadybeen disrupted/dissociated by the ionizing radiation of massive stars. In both cases, it becomesapparent that N79 is the most efficient site of current massive star formation, exceeding N11 and30 Doradus by a factor of ∼ × M (cid:12) Myr − versus 2.6 × M (cid:12) Myr − measured through H α (see Methods). A percentage of 18%, 9%, and 7% of the total SFR MYSO originates from a ∼ I disk ). These numbers will likely increasefor N79, and decrease for both 30 Doradus and N11 (see above). While the absolute SFR MYSO of579 and N11 do not differ significantly at R (cid:38)
50 pc (Fig. 4), the star formation efficiency of N79(through MYSOs) and 30 Doradus (through H α ) are elevated compared to N11. This may suggestthat the location of 30 Doradus and N79 on the leading edges of Arm E and Arm W positivelyinfluences the local star formation efficiency.Could the central object in N79, H72.97-69.39, eventually evolve into a SSC like R136? Thetotal luminosity of R136, L tot ∼ × L (cid:12) , is currently at least an order of magnitude higherthan H72.97-69.39, L tot ∼ × L (cid:12) (Fig. 3b). With a formation period of 5 - 10 Myr , anaverage SFR of ∼ × M (cid:12) Myr − is needed to create a 10 M (cid:12) stellar cluster, which is afactor of ∼ t (cid:63) may be lower than 0.5 Myr , which would increase our SFR MYSO estimates throughSFR
MYSO ∝ t − (cid:63) (see Methods). The properties of the surrounding gas reservoir also play a role indeveloping H72.97-69.39. If we assume that the (molecular) gas in N79 is gravitationally collaps-ing together with a formation timescale of 5 - 10 Myr, a star formation efficiency per free-fall time (cid:15) ff ∼ (cid:15) ff >
9, 21, 22, 32 , while extraordinary high star formationefficiencies have been quoted for more distant SSCs . However, we note that observed values of (cid:15) ff of individual GMCs extend over several orders of magnitude. Plus, stellar feedback may disruptthe cluster formation process, although the exact effects of feedback on massive protoclusters re-main unclear . Finally, the formation timescale of SSCs may be much smaller than our assumed5 -10 Myr . All of these effects may limit the final cluster mass. In this regard, detailed follow-upobservations with the Atacama Large Milimeter Array (ALMA) and the upcoming
James WebbSpace Telescope (JWST) are needed to establish if H72.97-69.39 could evolve into a SSC likeR136 ( ∼ M (cid:12) ), or a less-massive counterpart similar to the Arches and Quintuplet clusters nearthe Galactic center ( ∼ M (cid:12) ).The formation of very massive stars and SSCs is poorly understood . In this regard, ourfindings on N79 and H72.97-69.39 highlights the importance of high-resolution IR observations tounveil the earliest phases of extreme star formation. The unique location of 30 Doradus and N79suggest that the crossroads of spiral arms and galactic bars-ends may provide the right physicalconditions to create massive clusters . However, other factors that may play a role are the area-normalized SFR of a galaxy , accretion flows , or tidal interactions : observations suggest thatR136 formed after a recent collision of distinct H I flows, which were initially induced by the lastLMC - SMC interaction ∼ . Because of the proximity and face-on orientation of theLMC, ALMA and JWST will allow to spatially resolve the formation of this candidate SSC downto (cid:46) et al. Submillimetre-wavelength detection of dusty star-forming galaxies at highredshift.
Nature , 248–251 (1998). astro-ph/9806317 .2. Turner, J. L. Extreme Star Formation.
Astrophysics and Space Science Proceedings , 215(2009). . 6. Kruijssen, J. M. D. Globular clusters as the relics of regular star formation in ‘normal’ high-redshift galaxies. MNRAS , 1658–1686 (2015). .4. Walborn, N. R. The Starburst Region 30 Doradus. In Haynes, R. & Milne, D. (eds.)
TheMagellanic Clouds , vol. 148 of
IAU Symposium , 145 (1991).5. de Vaucouleurs, G. & Freeman, K. C. Structure and dynamics of barred spiral galaxies, inparticular of the Magellanic type.
Vistas in Astronomy , 163–294 (1972).6. Staveley-Smith, L., Kim, S., Calabretta, M. R., Haynes, R. F. & Kesteven, M. J. A new lookat the large-scale HI structure of the Large Magellanic Cloud. MNRAS , 87–104 (2003). astro-ph/0210501 .7. Bekki, K. & Chiba, M. Dynamical Influences of the Last Magellanic Interaction on the Mag-ellanic Clouds.
PASA , 21–29 (2007). astro-ph/0603812 .8. de Boer, K. S., Braun, J. M., Vallenari, A. & Mebold, U. Bow-shock induced star formationin the LMC? A&A , L49–L52 (1998). astro-ph/9711052 .9. Murray, N. Star Formation Efficiencies and Lifetimes of Giant Molecular Clouds in the MilkyWay.
ApJ , 133 (2011). .10. Ochsendorf, B. B., Meixner, M., Chastenet, J., Tielens, A. G. G. M. & Roman-Duval, J. TheLocation, Clustering, and Propagation of Massive Star Formation in Giant Molecular Clouds.
ApJ , 43 (2016). .11. Henize, K. G. Catalogues of H α -emission Stars and Nebulae in the Magellanic Clouds. ApJS , 315 (1956).12. Madden, S. C., Poglitsch, A., Geis, N., Stacey, G. J. & Townes, C. H. [C II] 158 MicronObservations of IC 10: Evidence for Hidden Molecular Hydrogen in Irregular Galaxies. ApJ , 200–209 (1997).13. Jameson, K. E. et al.
The Relationship Between Molecular Gas, H I, and Star Formation inthe Low-mass, Low-metallicity Magellanic Clouds.
ApJ , 12 (2016). .14. Seale, J. P. et al.
Herschel Key Program Heritage: a Far-Infrared Source Catalog for theMagellanic Clouds. AJ , 124 (2014).15. Mottram, J. C. et al. The RMS Survey: The Luminosity Functions and Timescales of MassiveYoung Stellar Objects and Compact H II Regions.
ApJ , L33 (2011). .16. Zinnecker, H. & Yorke, H. W. Toward Understanding Massive Star Formation.
ARA&A ,481–563 (2007). .17. Walborn, N. R. & Parker, J. W. Two-stage starbursts in the Large Magellanic Cloud - N11 asa once and future 30 Doradus. ApJ , L87–L89 (1992).78. Cignoni, M. et al.
Hubble Tarantula Treasury Project. II. The Star-formation History of theStarburst Region NGC 2070 in 30 Doradus.
ApJ , 76 (2015). .19. Fukui, Y. et al.
Molecular and Atomic Gas in the Large Magellanic Cloud. II. Three-dimensional Correlation Between CO and H I.
ApJ , 144–155 (2009). .20. Malumuth, E. M. & Heap, S. R. UBV stellar photometry of the 30 Doradus region of the largeMagellanic Cloud with the Hubble Space Telescope. AJ , 1054–1066 (1994).21. Lee, E. J., Miville-Deschˆenes, M.-A. & Murray, N. W. Observational Evidence of DynamicStar Formation Rate in Milky Way Giant Molecular Clouds. ApJ , 229 (2016). .22. Vutisalchavakul, N., Evans, N. J., II & Heyer, M. Star Formation Relations in the Milky Way.
ApJ , 73 (2016). .23. Turner, J. L. et al.
Highly efficient star formation in NGC 5253 possibly from stream-fedaccretion.
Nature , 331–333 (2015). .24. Ginsburg, A. et al.
Toward gas exhaustion in the W51 high-mass protoclusters.
A&A ,A27 (2016). .25. Crowther, P. A. et al.
The R136 star cluster dissected with Hubble Space Telescope/STIS.I. Far-ultraviolet spectroscopic census and the origin of He II λ MNRAS , 624–659 (2016). .26. Figer, D. F. et al.
Hubble Space Telescope/NICMOS Observations of Massive Stellar Clustersnear the Galactic Center.
ApJ , 750–758 (1999). astro-ph/9906299 .27. Krumholz, M. R. The Formation of Very Massive Stars. In Vink, J. S. (ed.)
Very MassiveStars in the Local Universe , vol. 412 of
Astrophysics and Space Science Library , 43 (2015). .28. Athanassoula, E. The existence and shapes of dust lanes in galactic bars.
MNRAS , 345–364 (1992).29. Johnson, L. C. et al.
Panchromatic Hubble Andromeda Treasury. XVIII. The High-mass Trun-cation of the Star Cluster Mass Function.
ApJ , 78 (2017). .30. Fukui, Y. et al.
Formation of the young massive cluster R136 triggered by tidally-drivencolliding H i flows.
PASJ , L5 (2017). .31. van der Marel, R. P. & Kallivayalil, N. Third-epoch Magellanic Cloud Proper Motions. II.The Large Magellanic Cloud Rotation Field in Three Dimensions. ApJ , 121 (2014). . 82. Ochsendorf, B. B., Meixner, M., Roman-Duval, J., Rahman, M. & Evans, N. J., II. What Setsthe Massive Star Formation Rates and Efficiencies of Giant Molecular Clouds?
ApJ , 109(2017). .33. Lopez, L. A. et al.
The Role of Stellar Feedback in the Dynamics of H II Regions.
ApJ ,121 (2014). . Author correspondence.
Correspondence and request for materials should be directed to B. B.Ochsendorf.
Author contributions.
B. B. O. performed the analysis, coordinated collaboration, and wrote themanuscript. O. N. helped characterizing H72.97-69.39. M. M. and O. C. J. helped with the creationof the MYSO catalog and estimates of source contamination. H. Z., J. B., R. I., and M. R. providedhelp with the interpretation of the results and implications.9 ethodsLMC surveys.
In this work we have made use of various galaxy-wide surveys of the LMC:1.
Atomic gas:
21 cm data from the Australian Telescope Compact Array and Parkes 64 mradio Telescope map .2. Molecular gas: CO (1-0) data from the Magellanic Mopra Assessment (MAGMA) DataRelease 2 (resolution 45”).3. Ionized gas: H α from the Southern H-Alpha Sky Survey Atlas (SHASSA) was used forcalculating the ionized gas mass. The H α image displayed in Figure 2a stems from the Mag-ellanic Clouds Emission Line Survey (MCELS), which has higher resolution compared toSHASSA but is not calibrated nor continuum subtracted.4. Infrared: µ m mid-IR data from Spitzer’s Surveying the Agents ofa Galaxy’s Evolution (SAGE) and 70, 160, 250, 350, and 500 µ m far-IR data from theHerschel Inventory of the Agents of Galaxy Evolution (HERITAGE). MYSO selection & completeness.
We have compiled a catalog of (highly) probable YSOs bycombining the results of galaxy-wide searches of YSO candidates
14, 40–42 using SAGE and HER-ITAGE data. The creation of the catalogue is explained in detail elsewhere , but the essentialpoints are discussed here as well.YSO candidates are identified through careful selection criteria (e.g., color-magnitude cuts,morphological inspection) tailored to minimize contamination from sources such as planetary neb-ulae, evolved stars, and background galaxies. Contamination estimates range from ∼ to ∼ . This means that in regions of high source density (such as the stellar bar), a relativelylarge amount of false source candidates can be expected (see Figure 1b). However, contaminationlevels vary between the faint and bright end of the YSO distribution, as faint YSOs overlap morewith the aforementioned contaminants in color-magnitude space compared to their luminous (i.e.,higher-mass) counterparts. For high mass YSO candidates, the contamination from evolved starsand background galaxies is shown to be (cid:46) . However, in star-forming regions the contami-nation becomes negligible ( <
14, 40, 41 andsubsequently fit their spectral energy distributions with YSO models . These models (2 × intotal) cover a wide range of physical parameters for different stages in the YSO evolutionary path,often divided in Stage 1 (least evolved), 2, and 3 (most evolved). The stringent color cuts used toseparate out YSOs from fore- and background contaminations renders our census of Stage 2 andStage 3 sources incomplete. However, these sources are largely irrelevant to this work since weaim to probe youngest population of YSOs, i.e., the earliest stages of (massive) star formation.The age of Stage 1 MYSOs is estimated at 0.5 Myr, which is the most recent value ob-tained for the observationally-derived ‘Class 1’ low-mass sources (which largely overlap with thetheoretically-based ‘Stage 1’ sources
45, 46 ) in the Gould’s Belt . It is not clear whether this valueapplies to massive stars; the absolute durations of the starless and active star-forming phases for10assive protostars is highly uncertain . In addition, the accreting phase for massive protostarsmay decrease with luminosity, possibly reaching 0.1 Myr for a 10 L (cid:12) star . Indeed, massivestars are expected to evolve more quickly than their lower-mass counterparts, and the assumed agemay therefore represent an upper limit to the age of these systems. A younger age would impactour results by increasing our SFR through SFR ∝ t − (cid:63) , where t (cid:63) is the age of the YSO population.Completeness of the YSO catalogues has been evaluated through false source extraction testsfor both the SAGE and HERITAGE data, and conclude that our catalogue of YSOs shouldbe complete for Stage 1 MYSOs of M > (cid:12) ( L (cid:38) L (cid:12) ). We set the photometric errors to10% in the 2MASS, IRAC, and MIPS bands, which allows to account for multiple sources oferror (systematic, calibration, variability, photon-counting) . However, as PAH emission is notincorporated into the SED models (which will alter the emission in the 3.6, 5.8, and 8.0 µ m bandscompared to the models), we relax our constraints and adjust the error bars in these bands to20%, 30%, and 40%, respectively, corresponding to the intrinsic strengths of the PAH bands .For the HERITAGE data, other sources of uncertainty were considered as well (background, PSFshapes) : typical uncertainties reported in the HERITAGE catalog are of order 5% - 20%. Weonly consider ‘well-fitted’ sources, i.e. those yielding reduced chi square of χ ≤ . Therefore, we stress that a poor fit does not necessarily mean that the object is nota YSO. We ultimately end with a catalogue of 693 Stage 1 MYSOs ( M > (cid:12) ) across the LMC.
Spectral energy distribution of H72.97-69.39.
The spectral energy distribution of the central lu-minous source H72.97-69.39 was compiled from the InfraRed Survey Facility (IRSF) , WISE ,SAGE , and HERITAGE . Its exceptional brightness and extended morphology causes the YSOmodel to severely underestimates its far-IR flux . Instead, we use a simple two-temperature modi-fied blackbody (MBB) function, where the temperature T , spectral index β , and scaling parametersare left as free parameters . The temperature of the hot dust component, peaking at ∼ µ m, ra-diates at T = 300 K (with β = 0.8), while the cold component peaking at ∼ µ m has T = 60 K(with β = 0.8). From this, we obtain a total infrared luminosity of L IR = 2.2 × M sun . Mass determination.
The mass in N79, N11, and 30 Doradus in the various phases of the ISM(Fig. 4) is estimated in the following ways:1.
Neutral atomic mass : assuming the H I gas is optically thin, the column density is estimatedthrough N HI = X HI W HI , where W HI is the integrated H I intensity and X HI = 1.82 × H cm /(K km s − ) is the proportionality constant . This can subsequently be converted togas surface density in M (cid:12) pc − with Σ H I = 0.8 × − N H I . We note that optically thickand/or cold H I gas emits disproportionately compared to optically thin H I . From absorptionspectra it is known that the 21 cm line may be optically thick in the LMC . Many studieshave attempted to estimate the optical depth correction to the H I mass, with differences of10% to 30% reported compared to optically thin gas
56, 57 . Thus, we adopt an uncertainty of0.1 dex for the H I mass. 11 .0 1.2 1.4 1.6 1.8 2.0 2.2 2.4log[ R (pc)]3.03.54.04.55.05.56.06.57.0 l o g [ M H , C O ( M fl )] Molecular gas (CO)
N7930DorN11 R (pc)]3.03.54.04.55.05.56.06.57.0 l o g [ M H , du s t ( M fl )] Molecular gas (dust-based)
N7930DorN11 R (pc)]3.03.54.04.55.05.56.06.57.0 l o g [ M H I ( M fl )] Neutral atomic gas
N7930DorN11 R (pc)]3.03.54.04.55.05.56.06.57.0 l o g [ M H II ( M fl )] Ionized gas
N7930DorN11
Figure 5:
ISM properties of N79, N11, and 30 Doradus.
The molecular (CO-based and dust-based), neutral, and ionized gas in apertures of radius R centered on N79 (RA = 72.972, DEC =-69.391), 30 Doradus (RA = 84.633, DEC = -69.092), and N11 (RA = 74.227, DEC = -66.368).2. CO-based molecular mass: estimated through M = α CO L CO , where L CO is the CO lumi-nosity and α CO = 8.6 (K km s − pc ) − is the proportionality constant appropriate for theLMC . The α CO factor is expected to be accurate within ∼ .3. Dust-based molecular mass: obtained by subtracting from the gas surface density, based onfar-infrared dust emission (modeled with a single temperature blackbody modified by a bro-ken power-law emissivity ), the surface density of atomic hydrogen. The creation of thismap is discussed in detail elsewhere , but the main caveats are reiterated here. First, theoptical thin limit was used to convert H I intensity to column density (see above). Second, it12as assumed that the gas-to-dust ratio in the diffuse and atomic gas is the same as in molecu-lar regions. However, there is mounting evidence that this gas-to-dust ratio changes betweendifferent phases of the ISM, with lower gas-to-dust ratios in the dense phase compared tothe diffuse phase . Both optically thick H I and a decrease in the gas-to-dust ratio may leadto an overestimation of the dust-based molecular gas. These effects introduce a systematicuncertainty of ∼ . However, given that weare focussing on (dense) star forming regions only, these uncertainties will propagate simi-larly for 30 Doradus, N11, and N79, and therefore the relative uncertainties are expected tobe smaller.4. Ionized gas : H II column density in cm − can be obtained by estimating electron densi-ties in different brightness regimes , which can then be converted to gas surface densityin M (cid:12) pc − through Σ H II = 0.8 × − N H II . The systematic uncertainty is estimated at ∼ α intensity to emission measure,which depends on the assumed electron temperature, which can vary within a factor of ∼
2, leading to a ∼
50% difference ( ∼ α intensity to emissionmeasure .Figure 5 shows the molecular (CO and dust-based), neutral, and ionized gas mass. Note thatthe total CO-based molecular mass in 30 Doradus and N79 are very similar within 100 pc. How-ever, the dust-based molecular mass in 30 Doradus exceeds the CO-based material by almost anorder of magnitude, indicating that the bulk of molecular material around 30 Doradus resides inthe ‘CO-dark’ phase, possibly through the local intense radiation field from R136. The ionized gascontent around the clusters both reflect the mass (i.e., ionizing photon budget) of the central clusterand the evolutionary state (i.e., embeddedness) of the region. Virial analysis.
The virial parameter, α vir = 5 σ R /( GM ) , where σ is the luminosity-weighted(one-dimensional) CO velocity dispersion, G is the gravitational constant, and M the CO mass,can be used to determine whether a cloud (complex) is bound and can undergo collapse, or isunbound, and may expand and dissolve back into the ISM. The critical virial parameter is α cr (cid:39) α cr ≤
2. However, lower values for α cr are possible in the case of strongmagnetic fields .Figure 6 shows that the molecular gas in the entire N11 region contains extreme high α vir ,which is to be expected given the evolved state of the region, with an expanding ring of materialmoving outward from the central ‘hole’, where an earlier generation of massive stars appear to havebeen born (note that individual cloud fragments may still be collapsing). 30 Doradus also con-tains elevated α vir ; higher-resolution studies show that the CO gas in 30 Doradus has elevated COlinewidths, probably due to the highly energetic environmental conditions within 30 Doradus .Conversely, N79 reveals sub-critical α vir throughout the majority part of the N79 cloud complex.This indicates that the N79 cloud complex is bound, and may be in a collapsing state. Star formation rates (SFR) from MYSOs.
Young Stellar Objects (YSOs) can be used to obtaina direct measure of SFR through SFR = N (YSOs) × (cid:104) M (cid:63) (cid:105) / t (cid:63) . Here, (cid:104) M (cid:63) (cid:105) ≈ .0 1.2 1.4 1.6 1.8 2.0 2.2 2.4log[ R (pc)]0.00.51.01.52.02.5 l o g [ α v i r ] Virial parameter
N7930DorN11
Figure 6:
Virial analysis.
The virial parameter α vir plotted versus radius R in N79, N11, and 30Doradus.mass for a fully-sampled IMF . Whereas counting YSOs has previously been applied in studiesof nearby molecular clouds, in the LMC we are limited to bright objects. Here, we assume that theluminosity of an MYSO is dominated by a single source. This assumption is motivated by obser-vations of nearby star clusters (e.g., the Orion Trapezium ). Subsequently, we use YSO modelsto estimate the mass of each individual source and multiply the source mass with an IMF toaccount for completeness. We choose 0.5 Myr for t (cid:63) (but see above). With these assumptions, thecompleteness limit ( M > M (cid:12) ) translates to a lower limit of SFR MYSO ∼ M (cid:12) Myr − whichwe can detect in our observations of the LMC.Many of the sources in our MYSO sample will break into small clusters when observed at highresolution
68, 69 . We estimate the uncertainty in our SFR
MYSO measurement associated with mul-tiplicity as follows. The reprocessed IR luminosity of the Orion Trapezium cluster would appearas a compact source at the resolution of our LMC IR maps. In the Orion Trapezium, the mainionizing source θ Ori C emits (cid:39)
50% of the total luminosity of the cluster . The IR luminositywould thus overestimate the luminosity of a single most massive source by a factor of ∼ ∼ . Therefore, we adopt 0.1 dex as our systematicuncertainty in SFR MYSO ; note that this assumes that the evolutionary tracks used in the YSO mod-els are correct . Star formation rates (SFR) from H α + 24 µ m. We convolve the 24 µ m map (from SAGE ) tothe resolution of our H α map (resolution 0.8’; from SHASSA ), and correct the H α emission forextinction using the 24 µ m emission. We then transform the combined H α and 24 µ m luminosity,14 (H α ) and L (24 µ m), to a SFR : SFR H α ( M (cid:12) yr − ) = 5 . × − [ L (H α ) + 0 . L (24 µ m)] . (1)We note that Eq. 1 assumes a fully-sampled IMF, which can be attained by averaging over largespatial scales such that each phase of star formation is probed. When studying star formation onsmaller scales these assumption may break down, which introduces stochastic effects that mainlyaffect the high-end part of the IMF. To account for this, we use the tool ‘Stochastically LightingUp Galaxies’ (SLUG ) to estimate to which extend stochastic sampling of the IMF affect ourmeasured SFR. This process is fully described elsewhere . The total SFR in the LMC measuredthrough H α equals × M (cid:12) Myr − . Star formation efficiency per free-fall time.
Can H72.97-69.39 become a 10 M (cid:12) star cluster?The current state of the (molecular) gas in N79 reveals that it is gravitationally bound and prone tofurther collapse (Figure 6). We thus assume that the gas collapses on its gravitational timescale, thefree-fall time τ ff = (cid:112) π/ Gρ , where ρ = M cloud /(4/3 πR ) is the mean density, M cloud the cloudmass, R cloud the cloud radius, and G the gravitational constant. We adopt a formation timescaleof 5 - 10 Myr, motivated by the star formation history of the NGC2070 region (including R136).In this timescale, we calculate that a total of 1.3 - 3.7 × of mass (molecular and atomic) cancollapse to the centre of the N79 cloud from R cloud (cid:46)
30 - 60 pc. Thus, to create a 10 M (cid:12) stellarcluster, a star formation efficiency per free-fall time (i.e., the fraction of mass that is transformedinto stars during τ ff ) of (cid:15) ff ∼ M (cid:12) cluster is massive enough to eventually become a globu-lar cluster ( (cid:38) M (cid:12) ) is not clear, given that young stellar clusters may lose a significant amountof mass through supernovae, stellar winds, and stripping . Data availability statement. et al.
A Neutral Hydrogen Survey of the Large Magellanic Cloud: Aperture Synthesisand Multibeam Data Combined.
ApJS , 473–486 (2003).35. Wong, T. et al.
The Magellanic Mopra Assessment (MAGMA). I. The Molecular Cloud Pop-ulation of the Large Magellanic Cloud.
ApJS , 16 (2011). .36. Gaustad, J. E., McCullough, P. R., Rosing, W. & Van Buren, D. A Robotic Wide-Angle H α Survey of the Southern Sky.
PASP , 1326–1348 (2001). astro-ph/0108518 .37. Smith, R. C. & MCELS Team. The UM/CTIO Magellanic Cloud emission-line survey.
PASA , 163–64 (1998). 158. Meixner, M. et al. Spitzer Survey of the Large Magellanic Cloud: Surveying the Agents ofa Galaxy’s Evolution (SAGE). I. Overview and Initial Results. AJ , 2268–2288 (2006). astro-ph/0606356 .39. Meixner, M. et al. The HERSCHEL Inventory of The Agents of Galaxy Evolution in theMagellanic Clouds, a Herschel Open Time Key Program. AJ , 62 (2013).40. Whitney, B. A. et al. Spitzer Sage Survey of the Large Magellanic Cloud. III. Star Formationand ˜1000 New Candidate Young Stellar Objects. AJ , 18–43 (2008).41. Gruendl, R. A. & Chu, Y.-H. High- and Intermediate-Mass Young Stellar Objects in the LargeMagellanic Cloud. ApJS , 172–197 (2009). .42. Seale, J. P. et al.
The Evolution Of Massive Young Stellar Objects in the Large MagellanicCloud. I. Identification and Spectral Classification.
ApJ , 150–167 (2009). .43. Jones, O. C. et al.
The SAGE-Spec Spitzer Legacy program: The life-cycle of dust andgas in the Large Magellanic Cloud. Point source classification III.
ArXiv e-prints (2017). .44. Robitaille, T. P., Whitney, B. A., Indebetouw, R., Wood, K. & Denzmore, P. InterpretingSpectral Energy Distributions from Young Stellar Objects. I. A Grid of 200,000 YSO ModelSEDs.
ApJS , 256–285 (2006). astro-ph/0608234 .45. Heiderman, A. & Evans, N. J., II. The Gould Belt ’MISFITS’ Survey: The Real Solar Neigh-borhood Protostars.
ApJ , 231 (2015). .46. Heyer, M. et al.
The rate and latency of star formation in dense, massive clumps in the MilkyWay.
A&A , A29 (2016). .47. Dunham, M. M. et al.
Young Stellar Objects in the Gould Belt.
ApJS , 11 (2015). .48. Battersby, C., Bally, J. & Svoboda, B. The Lifetimes of Phases in High-mass Star-formingRegions.
ApJ , 263 (2017). .49. Carlson, L. R. et al.
A Panchromatic View of NGC 602: Time-resolved Star Formation withthe Hubble and Spitzer Space Telescopes.
ApJ , 78 (2011). .50. Kato, D. et al.
The IRSF Magellanic Clouds Point Source Catalog.
PASJ , 615–641 (2007).51. Wright, E. L. et al. The Wide-field Infrared Survey Explorer (WISE): Mission Descriptionand Initial On-orbit Performance. AJ , 1868–1881 (2010). .52. Sewiło, M. et al. The youngest massive protostars in the Large Magellanic Cloud.
A&A ,L73 (2010). . 163. Gordon, K. D. et al.
Dust and Gas in the Magellanic Clouds from the HERITAGE HerschelKey Project. I. Dust Properties and Insights into the Origin of the Submillimeter Excess Emis-sion.
ApJ , 85 (2014). .54. Spitzer, L.
Physical processes in the interstellar medium (1978).55. Marx-Zimmer, M. et al.
A study of the cool gas in the Large Magellanic Cloud. I. Propertiesof the cool atomic phase - a third H i absorption survey.
A&A , 787–801 (2000).56. Dickey, J. M., Mebold, U., Stanimirovic, S. & Staveley-Smith, L. Cold Atomic Gas in theSmall Magellanic Cloud.
ApJ , 756–772 (2000).57. Lee, M.-Y., Stanimirovi´c, S., Murray, C. E., Heiles, C. & Miller, J. Cold and Warm AtomicGas around the Perseus Molecular Cloud. II. The Impact of High Optical Depth on the HIColumn Density Distribution and Its Implication for the HI-to-H Transition.
ApJ , 56(2015). .58. Bolatto, A. D., Wolfire, M. & Leroy, A. K. The CO-to-H Conversion Factor.
ARA&A ,207–268 (2013). .59. Roman-Duval, J. et al. Dust and Gas in the Magellanic Clouds from the HERITAGE HerschelKey Project. II. Gas-to-dust Ratio Variations across Interstellar Medium Phases.
ApJ , 86(2014). .60. Paradis, D. et al.
Spitzer Characterization of Dust in the Ionized Medium of the Large Magel-lanic Cloud.
ApJ , 6 (2011). .61. Shaver, P. A., McGee, R. X., Newton, L. M., Danks, A. C. & Pottasch, S. R. The galacticabundance gradient.
MNRAS , 53–112 (1983).62. Dickinson, C., Davies, R. D. & Davis, R. J. Towards a free-free template for CMB fore-grounds.
MNRAS , 369–384 (2003). astro-ph/0302024 .63. Bertoldi, F. & McKee, C. F. Pressure-confined clumps in magnetized molecular clouds.
ApJ , 140–157 (1992).64. Kauffmann, J., Pillai, T. & Goldsmith, P. F. Low Virial Parameters in Molecular Clouds:Implications for High-mass Star Formation and Magnetic Fields.
ApJ , 185 (2013). .65. Rosado, M. et al.
Formation of the nebular complex N11 in the Large Magellanic Cloud.
A&A , 588–600 (1996).66. Nayak, O. et al.
Studying Relation Between Star Formation and Molecular Clumps on Sub-parsec Scales in 30 Doradus.
ArXiv e-prints (2016). .67. Kroupa, P. On the variation of the initial mass function.
MNRAS , 231–246 (2001). astro-ph/0009005 . 178. Vaidya, K., Chu, Y.-H., Gruendl, R. A., Chen, C.-H. R. & Looney, L. W. A Hubble Space Tele-scope View of the Interstellar Environments of Young Stellar Objects in the Large MagellanicCloud.
ApJ , 1417–1426 (2009). .69. Stephens, I. W. et al.
Stellar Clusterings around ”Isolated” Massive YSOs in the LMC.
ApJ , 94 (2017). .70. Vacca, W. D., Garmany, C. D. & Shull, J. M. The Lyman-Continuum Fluxes and StellarParameters of O and Early B-Type Stars.
ApJ , 914 (1996).71. O’dell, C. R., Valk, J. H., Wen, Z. & Meyer, D. M. Identification of velocity systems in theinner Orion nebula.
ApJ , 678–683 (1993).72. Sim´on-D´ıaz, S., Herrero, A. & Esteban, C. The Trapezium Stars. Preliminary Results onDetailed Atmosphere Modeling. In Reyes-Ruiz, M. & V´azquez-Semadeni, E. (eds.)
RevistaMexicana de Astronomia y Astrofisica Conference Series , vol. 18 of
Revista Mexicana deAstronomia y Astrofisica, vol. 27 , 123–125 (2003).73. Robitaille, T. P. SED Modeling of Young Massive Stars. In Beuther, H., Linz, H. & Henning,T. (eds.)
Massive Star Formation: Observations Confront Theory , vol. 387 of
AstronomicalSociety of the Pacific Conference Series , 290 (2008). .74. Gaustad, J. E., McCullough, P. R., Rosing, W. & Van Buren, D. A Robotic Wide-Angle H α Survey of the Southern Sky.
PASP , 1326–1348 (2001). astro-ph/0108518 .75. Calzetti, D. et al.
The Calibration of Mid-Infrared Star Formation Rate Indicators.
ApJ ,870–895 (2007). .76. Krumholz, M. R. et al.
Star Cluster Formation and Feedback.
Protostars and Planets VI .77. Kruijssen, J. M. D. & Longmore, S. N. An uncertainty principle for star formation - I. Whygalactic star formation relations break down below a certain spatial scale.
MNRAS , 3239–3252 (2014). .78. Kennicutt, R. C., Jr., Bresolin, F., Bomans, D. J., Bothun, G. D. & Thompson, I. B. Largescale structure of the ionized gas in the magellanic clouds. AJ , 594–604 (1995).79. Krumholz, M. R. & McKee, C. F. A General Theory of Turbulence-regulated Star Formation,from Spirals to Ultraluminous Infrared Galaxies. ApJ , 250–268 (2005). astro-ph/0505177 .80. Portegies Zwart, S. F., McMillan, S. L. W. & Gieles, M. Young Massive Star Clusters.
ARA&A , 431–493 (2010). .81. D’Ercole, A., Vesperini, E., D’Antona, F., McMillan, S. L. W. & Recchi, S. Formation anddynamical evolution of multiple stellar generations in globular clusters. MNRAS , 825–843(2008). . 182. Bekki, K. Secondary star formation within massive star clusters: origin of multiple stellarpopulations in globular clusters.
MNRAS , 2241–2259 (2011). .83. Schaerer, D. & Charbonnel, C. A new perspective on globular clusters, their initial massfunction and their contribution to the stellar halo and the cosmic reionization.
MNRAS ,2297–2304 (2011).1101.1073