ALMA detects CO(3–2) within a super star cluster in NGC 5253
We present observations of CO(3–2) and CO(3–2) emission near the supernebula in the dwarf galaxy NGC 5253, which contains one of the best examples of a potential globular cluster in formation. The 03 resolution images reveal an unusual molecular cloud, “Cloud D1”, coincident with the radio-infrared supernebula. The -pc diameter cloud has a linewidth, , that reflects only the gravitational potential of the star cluster residing within it. The corresponding virial mass is . The cluster appears to have a top-heavy initial mass function, with -2 . Cloud D1 is optically thin in CO(3–2) probably because the gas is hot. Molecular gas mass is very uncertain but constitutes 35% of the dynamical mass within the cloud boundaries. In spite of the presence of an estimated O stars within the small cloud, the CO appears relatively undisturbed. We propose that Cloud D1 consists of molecular clumps or cores, possibly star-forming, orbiting with more evolved stars in the core of the giant cluster.
Jean L. Turner
0000-0003-4625-2951]Jean L. Turner
0000-0002-0214-0491]S. Michelle Consiglio
0000-0002-5770-8494]Sara C. Beck
Giant young star clusters, with masses , contain thousands of massive stars within the space of only a few parsecs. Given the rapid rate of evolution of O stars, how other stars can form in their presence to build a large star cluster remains an outstanding problem. The closest young massive clusters are in nearby galaxies; at these distances subarcsecond resolution is required to study star formation on cluster scales. The Atacama Large Millimeter/Submillimeter Array (ALMA) can now provide images of gas at subarcsecond resolution corresponding to cluster scales in local galaxies. The J=3–2 line of CO is bright and easily excited in dense () gas, and can be used to estimate gas masses and kinematics to study the star formation process and how feedback occurs within massive clusters.
NGC 5253 is a local (3.8 Mpc: 1″= 18.4 pc) dwarf spheroidal galaxy (Caldwell & Phillips, 1989; Martin, 1998) with many young star clusters (Caldwell & Phillips, 1989; Meurer et al., 1995; Gorjian, 1996; Calzetti et al., 1997; Harris et al., 2004; Tremonti et al., 2001; de Grijs et al., 2013; Calzetti et al., 2015) and an infrared luminosity of (Vanzi & Sauvage, 2004; Hunt et al., 2005). Its stellar mass is (Martin, 1998, the dark matter mass could be ten times larger). At least one-third of the galaxy’s infrared luminosity originates from a giant star-forming region, a compact ( pc) radio source known as the “supernebula” (Beck et al., 1996; Turner et al., 1998; Gorjian et al., 2001; Turner & Beck, 2004; Rodríguez-Rico et al., 2007). Extinction toward the supernebula is high, -18 (Calzetti et al., 1997; Turner et al., 2003; Martín-Hernández et al., 2005; Calzetti et al., 2015). The supernebula is coincident with a bright infrared source (Gorjian et al., 2001; Turner et al., 2003; Alonso-Herrero et al., 2004) that may coincide with a visible red cluster, # 11 (Calzetti et al., 2015; Smith et al., 2016). Based on CO observations, there appears to be little molecular gas within NGC 5253; most of the CO emission is found in a streamer along the prominent minor axis dust lane (Turner et al., 1997; Meier et al., 2002; Miura et al., 2015). A bright CO(3–2) source was detected near the supernebula in 4″ observations with the Submillimeter Array; comparison to lower J CO images suggest very warm gas, K, in the central regions (Turner et al., 2015).
We present ALMA observations of CO and CO J=3–2 emission at 03 resolution (5.5 pc) within the central region of NGC 5253. Requiring gas of density for collisional excitation, CO(3–2) traces the dense gas typically associated with star forming cores (e.g., Myers, 1985; Lada et al., 2010). The observations reveal a number of dense clouds within the central pc starburst region identified as Cloud D by the CO(2-1) analysis of Meier et al. (2002); one cloud stands out in terms of its unusual properties. We present here an analysis of the molecular cloud coincident with the supernebula, henceforth denoted “Cloud D1”.
NGC 5253 was observed in Band 7, a Cycle 1 program (ID = 2012.1.00105.S, PI = J. Turner) executed in Cycle 2 on 2015 June 4 and 5. The pointing shown here is centered at 13:39:44.911910, -31:38:26.49657 (J2000). The full mosaic of NGC 5253 including CO(3–2) is presented elsewhere (Consiglio et al., 2017). Spectral windows have a total bandwidth of 937.500 MHz, with 244.141 kHz per channel. Velocities are barycentric, in radio convention. Bandpass and phase were calibrated with J1427-42064 and J1342-2900 respectively. Titan was the flux calibrator. Calibration was done with CASA, pipeline 4.2.2 by the Joint ALMA Observatory. Absolute flux calibration is to within 10% for Cycle 2 Band 7 data (Lundgren, 2013). Imaging was done with CASA pipeline 4.5.0 by the authors. The synthesized beam for the CO(3-2) maps is 0.33″ 0.27″ p.a. -90°. The conversion to brightness is 1K 9-17 mJy, with the smaller value for point sources, and the larger for sources filling the beam. A continuum map was constructed from offline channels in the band and subtracted from the (u,v) data by the authors before making line maps. The shortest baselines in the image are 25-100 k; emission more extended than 4″ (75 pc) can be poorly represented in these maps; from comparison with SMA data we estimate that 50% of the emission is in such faint, extended structure (Consiglio et al., 2017). The RMS noise in the individual 1 line maps is 2.7 mJy/bm for CO(3-2) at 345.796 GHz. The integrated intensity map (Moment 0 map) was made by summing emission greater than in the cube.
3 Cloud D1 and the supernebula
The ALMA image of CO(3-2) integrated line emission, in red and contours, is shown overlaid on an HST H image in Figure 1. At ten times higher spatial resolution than previous CO maps, the ALMA CO(3–2) image reveals that what was previously identified as Cloud D (Meier et al., 2002; Turner et al., 2015) actually comprises many molecular clouds. “Cloud D1” a compact and bright CO(3–2) source labeled in the figure, is coincident with the core of the radio/infrared supernebula. Fifteen to twenty parsecs to the southwest of D1 is a separate extended cloud with stronger CO(3–2) emission that is redshifted with respect to Cloud D1. There are numerous other clouds within the central region that are discussed elsewhere (Consiglio et al., 2017). In this paper we focus on the unusual molecular Cloud D1.
NGC 5253 is a galaxy known for bright nebular emission as well as for the presence of dust (Burbidge & Burbidge, 1962; Kleinmann & Low, 1970). The radio and mid-infrared emission in NGC 5253 is dominated by the supernebula, a source of radius pc (Turner et al., 1998, 2000; Gorjian et al., 2001; Turner & Beck, 2004) located near luminous and young star clusters (Meurer et al., 1995; Calzetti et al., 1997). From radio continuum fluxes (Turner et al., 1998, 2000; Turner & Beck, 2004; Meier et al., 2002; Rodríguez-Rico et al., 2007), and radio recombination line fluxes (Rodríguez-Rico et al., 2007; Bendo et al., 2017), a Lyman continuum rate of is indicated (corrected for direct dust absorption of UV photons following Inoue (2001), Turner et al., 2015). is computed for an ionization-bounded nebula, and will be larger if there is photon leakage. Starburst99 models, described below, indicate that this Lyman continuum rate corresponds to 1400-1800 O stars. The supernebula lies close to objects identified as Clusters 5 and 11 by Calzetti et al. (2015); from their colors both appear to contain stars 1 Myr in age (Calzetti et al., 2015; Smith et al., 2016).
Cloud D1 is located within 0.6 pc in projection of the embedded supernebula (Table 1; Turner & Beck, 2004) The CO(3-2) line centroid, v = 387.60.5 , agrees to 2 with the main 10.5m [SIV] line from the ionized gas (Beck et al., 2012) and to 3 with the H30 radio recombination line (Bendo et al., 2017) from the supernebula. Cloud D1 is thus coincident with the supernebula both in projection and in velocity, and it is similar if not identical in size.
|RA(2000)||13 39 55956100004|
|Assumed distance||3.8 Mpc|
|V(CO3-2)aaLine centroid, barycentric, radio definition.|
|V(CO(3–2))bbFWHM; Gaussian fit to line in CASA.|
|S(CO 3–2)ccIntegrated line fluxes are for 05 aperture centered on D1. Uncertainty in the CO(3–2) flux is dominated by 10% calibration uncertainty; the CO(3–2) flux uncertainty is due to signal-to-noise. See Observations.||Jy|
|S(CO3–2)ccIntegrated line fluxes are for 05 aperture centered on D1. Uncertainty in the CO(3–2) flux is dominated by 10% calibration uncertainty; the CO(3–2) flux uncertainty is due to signal-to-noise. See Observations.||Jy|
|RadiusddDeconvolved from beam in CASA assuming a Gaussian source profile. See text.||03 005|
|eeVirial mass based on CO linewidth and size. For ; for the mass is 30% less.|
|ffLyman continuum rate from the literature, primarily (Turner et al., 1998; Bendo et al., 2017), and corrected for 30% direct absorption of uv photons by dust (Turner et al., 2015) following the procedure of Inoue (2001). Corrected to 3.8 Mpc. Assumes ionization-bounded nebula and as such is a lower limit to the true rate.|
Cloud D1 is small. A Gaussian fit to the integrated intensity image (Fig. 1) gives a size of 21833mas 10454mas, p.a. 2.2°77°, FWHM, deconvolved from the beam. Fits of D1 within the 24 individual channel maps within the FWHM give sizes of 03 005. We adopt a size of 03 005, or r=2.8 pc. Excitation of CO(3-2) requires minimum molecular gas densities of , so the molecular gas in Cloud D1 is dense.
Cloud D1 is optically thin in the CO(3–2) line. In Figure 2, spectra are plotted for both CO(3-2) and CO(3-2) lines within a 05 region centered on Cloud D1. The integrated intensity of CO(3–2) is for CO(3–2) is - The line ratio of 50 for Cloud D1 is close to the abundance ratio of [CO]/[CO]70 for the solar neighborhood, and the [C]/[C]40-50 estimated for the LMC (Heikkilä et al., 1999). The Cloud D1 ratio is significantly higher than the - observed in the nearby clouds, including the large cloud to the southwest of D1 (Consiglio et al., 2017), which is closer to the value of 13 observed in the Orion molecular cloud in optically thick, cool CO gas (Schilke et al., 1997). Warm gas can produce low optical depths in CO and this is what we propose for D1. For 300 K, as estimated from modeling of the CO(3-2)/CO(1-0) line ratio (Turner et al., 2015), the CO partition function is 100, and the populations of the low J levels are correspondingly low. Thus the low optical depth in the CO(3-2) line from Cloud D1 is additional evidence that the gas is in close proximity to the super star cluster.
A dynamical mass within Cloud D1 can be obtained from the width of the CO line, (FWHM). For an mass profile, as in the Galactic cloud W49N, (r2.8 pc). (An profile giving 30% less mass is included in the uncertainty.) This mass estimate assumes a virialized cloud; it is unclear if this is valid for such a dynamic young source. However, most other sources of velocity, such as protostellar outflows or stellar winds, would if anything broaden the line over the gravitational value. Moreover the mass is close to that expected from the star cluster based on its radio free-free flux. Hence we adopt a gravitational mass of . The mean mass density within the cloud region, most of which is in stars, is then 2700 pc, corresponding to , and surface density , in the range observed for super star clusters (Tan et al., 2014). These would have been the original gas values at the onset of star formation in the cloud.
It is exceedingly difficult to determine a molecular gas mass for Cloud D1. The metallicity of the region is unclear. The galaxy overall is metal-poor, , but there is evidence for localized enrichment near the young clusters (Walsh & Roy, 1989; Schaerer et al., 1997; Kobulnicky et al., 1997; López-Sánchez et al., 2007; Turner et al., 2015). From the CO(3-2) line intensity of , and for K (Turner et al., 2015), an uncertain number, the mass in CO molecules alone is for optically thin emission; then for a Galactic [CO]/[H] the molecular gas mass is (includes He; for T=50K, the mass would be 4 times lower.) However, the Galactic [CO]/[H] abundance may not be appropriate in this environment. An alternate method using an empirical CO “conversion factor”, , gives (including He). Since this relation applies to gas-dominated, optically thick clouds, it should overestimate the gas mass in D1. The mass of ionized gas in the supernebula is (Turner & Beck, 2004); the estimated HI mass within D1 based on the 21 cm absorption column (Kobulnicky & Skillman, 2008) toward the supernebula is minimal (200-500 ; larger HI columns will become molecular). The combined molecular-plus-ionized gas mass within Cloud D1 is 6000-60,000 . We estimate that gas constitutes % of the mass within the boundaries of Cloud D1.
Based on the dynamical mass from the CO linewidth, which is dominated by stars, we can compare the mass, , indicated for the cluster, and the observed Lyman continuum rate. We find /. We use Starburst99 models to model the cluster age and IMF. For the calculated and (Table 1), Starburst99 (Leitherer et al., 1999, 2014) models were run for Padova and Geneva models for metallicities of Z=0.004 (closest to the mean metallicity of NGC 5253) and 0.008 with Kroupa initial mass functions (IMFs) with exponent 2.3. Statistical evidence (Weidner & Kroupa, 2004; Oey & Clarke, 2005; Kroupa et al., 2013) suggests an upper mass limit of 150 for stars. Starburst99 models with an upper IMF mass cutoff of 150 require a lower IMF mass cutoff 2 to reach the observed for a cluster of . It has been suggested that stars of 200 or more may exist in R136 and NGC 5253 (Crowther et al., 2010; Smith et al., 2016), perhaps formed from binary mergers (Banerjee et al., 2012). Such supermassive stars could increase the Lyman continuum rate without adding much mass. Even for a 200 cutoff, Starburst99 models indicate a lower mass cutoff 1 for the D1 cluster. If there is leakage of uv photons beyond the supernebula, which would increase , the lower mass cutoffs for the IMF are even higher. Another explanation for the low mass-to-luminosity ratio may be interacting binary stars (Stanway et al., 2016).
4 The internal structure of Cloud D1
Cloud D1 is located within the supernebula/cluster core, a harsh environment for molecules, where the mean separation between O stars is only 0.1 pc. The CO properties suggest that Cloud D1 is composed of many pockets of dense molecular gas, which may be in the form of protostellar disks, hot molecular cores surrounding individual stars, or residual dense molecular clumps. The estimated molecular gas mass of -, including He, predicts a column density of -, and gas volume density of -. Comparison with the CO(3–2) critical density of indicates a volume filling factor of 3-50%. We regard the lower value, corresponding to the optically thin mass, as more likely, and that the volume filling factor, .
The smooth and near-Gaussian CO line profile and the similarity in shape of D1 across the line also suggests many clumps. Figure 2 shows the CO(3–2) line profile and the Gaussian fit. Departures from smoothness can be used to estimate numbers of clumps, even if the line is not perfectly Gaussian (Beck, 2008). Channel-to-channel variations for the central 24 channels are typically 20%, implying 25 clumps per channel and 600 for clumps with individual thermal linewidths of 1 . If instead Cloud D1 is composed of CO protostellar disks or molecular cores surrounding individual stars, with expected linewidths of 1-10 , fewer disks/cores are needed to produce a smooth line. Spatial variations in the centroid and deconvolved size of the emission from Cloud D1 across the 24 line channels are also consistent with no variation (centroids coincident to 20 mas and in size, to 50 mas), consistent with the many-clump hypothesis. The slight nonGaussianity would be consistent with randomly distributed substructures. The blueshifted side of the line is smoother than the redshifted side, which has no clear explanation; the redshifted side may be related to the southeastern extension of the cloud.
Are these small clumps or cores consistent with the radiation shielding necessary for the existence of CO? There is a minimum size expected for CO-emitting clouds, since CO is chemically sustained only at , or cloud column (Hollenbach & Tielens, 1999; Bisbas et al., 2015). At a density of (for T=300K), the minimum cloud column corresponds to a distance of 0.04 pc = 8000 AU, for a Galactic ratio. The corresponding minimum clump mass is 0.1 . It is thus plausible that Cloud D1 could accommodate these minimum clumps both in linear dimension and in mass.
5 Feedback and star formation within a young, massive cluster
Cloud D1 coincides with the supernebula, which is an ultracompact HII region, and nearby evidence of Wolf-Rayet stars (Walsh & Roy, 1989; Schaerer et al., 1997; Kobulnicky et al., 1997; López-Sánchez et al., 2007) that are typically 3-4 Myr in age, but sometimes less (Smith et al., 2016). Wolf-Rayet stars can lose copious amounts of metal-enriched mass. Thus the presence of molecular gas may not in itself give a good indication of cluster age. Some super star clusters appear to be actively dispersing gas in winds, as appears to be occurring in NGC 253 (Sakamoto et al., 2006; Bolatto et al., 2013). However, if the cluster is sufficiently massive, theory suggests that the evolution of the HII region can be affected by gravity (Kroupa & Boily, 2002; Murray et al., 2010), so that in some cases the enriched products of stellar mass loss may be retained by the cluster (Silich & Tenorio-Tagle, 2017). Cloud D1 could perhaps survive after a supernova, since simulations suggest that gas can backfill into the cluster after the explosion (Tenorio-Tagle et al., 2015). We note that the CO linewidth reflecting the core cluster mass within the supernebula is almost precisely equal to the thermal linewidth of the ionized hydrogen; this may not be coincidence.
To put Cloud D1 in perspective, it is instructive to compare to its closest Galactic analog, W49N. Of similar extent (r3 pc) with ultracompact HII regions (Dreher et al., 1984; De Pree et al., 2000), W49N appears to be at a similar evolutionary stage to the supernebula in NGC 5253. W49N has a luminosity of (Buckley & Ward-Thompson, 1996), 100 times less that of the supernebula. Yet W49N has a larger, perhaps significantly larger, molecular gas mass, (Galván-Madrid et al., 2013). Even including the cloud to the southwest of the supernebula, of similar mass to the W49N cloud (Consiglio et al., 2017), the supernebula region is 50 times more efficient than W49N at forming stars in terms of .
An alternative to high star formation efficiency is that gas has already dispersed from the young cluster. Based on the CO linewidth, dispersal of gas from the cluster does not appear to be occurring at present. Given the young age of the cluster, any gas that has previously been dispersed by the cluster cannot be far away. Gas expelled at 20 , the overall velocity dispersion within the region (Consiglio et al., 2017) would still be within 20-60 pc of the cluster given its age; the clouds will still lie within Figure 1.
The continued presence of dense molecular gas within the supernebula cluster and its relative quiescence suggests that negative star formation feedback effects are minimal on the molecular gas at this stage of cluster evolution of the supernebula, perhaps because the gas is in the form of compact clumps or cores. The high density indicated by the CO(3–2) emission suggests that Cloud D1 may still be forming stars. The CO(3-2) emission may arise in dense hot molecular cores around young stellar objects, which would be consistent with the warm, temperature suggested by lower resolution CO line ratios (Turner et al., 2015), and the optically thin CO(3–2) emission reported here. For this temperature, and given that the CO cores are embedded in an HII region of density (Turner & Beck, 2004), the Bonnor-Ebert mass of Cloud D1 is 13 . If stars are still forming within the core of the cluster embedded in Cloud D1, they will be massive.
We present ALMA observations of CO(3–2) and CO(3–2)
at 03 resolution of the region surrounding the
supernebula in NGC 5253.
We identify an unusual cloud, “Cloud D1”, that is precisely coincident with the supernebula/
embedded cluster in
space and velocity.
Cloud D1 has a radius of r2.8 pc, nearly the same size as the supernebula.
1) spatial coincidence to pc in projection with the supernebula,
2) velocity coincidence to within - with mid-IR nebular and radio recombination lines from the HII region and
3) the fact that the cloud is optically thin, probably because it is hot
we conclude that the molecular cloud is mixed in with the super star cluster, which contains 1500 O stars, and the compact HII region.
The CO linewidth of 21.7 indicates that the CO gas in Cloud D1 is relatively undisturbed in spite of its location within this dense cluster. The linewidth indicates a dynamical mass of . This gives for the cluster, implying a top-heavy IMF, with lower cutoff of 1–2 , and more if there is photon leakage beyond the supernebula.
A CO(3–2)/CO ratio of 50 indicates the the emission from Cloud D1 is bright because the gas is optically thin and not because there is a high gas column. Estimates of gas mass based on this line are uncertain, but suggest that the molecular gas mass is 60,000 . The star formation efficiency appears to be 50 times that of W49N, the closest Galactic analogue. The smoothness of the CO line profile and its near Gaussianity suggests Cloud D1 consists of many, up to hundreds, of molecular clumps or cores. Given the high ambient pressure and temperature within Cloud D1, if the cloud is indeed hot as previously estimated (Turner et al., 2015) only massive stars 13 can form at present. We propose that Cloud D1 is composed of many hot molecular clumps or cores orbiting within the cluster potential with the stars of the super star cluster, and may yet be capable of forming stars.
- Alonso-Herrero et al. (2004) Alonso-Herrero, A., Takagi, T., Baker, A. J., et al. 2004, ApJ, 612, 222
- Banerjee et al. (2012) Banerjee, S., Kroupa, P., & Oh, S. 2012, MNRAS, 426, 1416
- Beck (2008) Beck, S. C. 2008, A&A, 489, 567
- Beck et al. (2012) Beck, S. C., Lacy, J. H., Turner, J. L., et al. 2012, ApJ, 755, 59
- Beck et al. (1996) Beck, S. C., Turner, J. L., Ho, P. T. P., Lacy, J. H., & Kelly, D. M. 1996, ApJ, 457, 610
- Bendo et al. (2017) Bendo, G. J., Miura, R. E., Espada, D., Nakanishi, K., Beswick, R. J., D’Cruze, M. J., Dickinson, C., & Fuller, G. A., MNRAS, submitted.
- Bisbas et al. (2015) Bisbas, T. G., Haworth, T. J., Barlow, M. J., et al. 2015, MNRAS, 454, 2828
- Bolatto et al. (2013) Bolatto, A. D., Warren, S. R., Leroy, A. K., et al. 2013, Nature, 499, 450
- Buckley & Ward-Thompson (1996) Buckley, H. D., & Ward-Thompson, D. 1996, MNRAS, 281, 294
- Burbidge & Burbidge (1962) Burbidge, E. M., & Burbidge, G. R. 1962, ApJ, 135, 694
- Caldwell & Phillips (1989) Caldwell, N., & Phillips, M. M. 1989, ApJ, 338, 789
- Calzetti et al. (1997) Calzetti, D., Meurer, G. R., Bohlin, R. C., et al. 1997, AJ, 114, 1834
- Calzetti et al. (2015) Calzetti, D., Johnson, K. E., Adamo, A., et al. 2015, ApJ, 811, 75
- Consiglio et al. (2017) Consiglio et al., submitted.
- Cresci et al. (2005) Cresci, G., Vanzi, L., & Sauvage, M. 2005, A&A, 433, 447
- Crowther et al. (2010) Crowther, P. A., Schnurr, O., Hirschi, R., et al. 2010, MNRAS, 408, 731
- de Grijs et al. (2013) de Grijs, R., Anders, P., Zackrisson, E., & Östlin, G. 2013, MNRAS, 431, 2917
- De Pree et al. (2000) De Pree, C. G., Wilner, D. J., Goss, W. M., Welch, W. J., & McGrath, E. 2000, ApJ, 540, 308
- Dreher et al. (1984) Dreher, J. W., Johnston, K. J., Welch, W. J., & Walker, R. C. 1984, ApJ, 283, 632
- Galván-Madrid et al. (2013) Galván-Madrid, R., Liu, H. B., Zhang, Z.-Y., et al. 2013, ApJ, 779, 121
- Gorjian (1996) Gorjian, V. 1996, AJ, 112, 1886
- Gorjian et al. (2001) Gorjian, V., Turner, J. L., & Beck, S. C. 2001, ApJ, 554, L29
- Harris et al. (2004) Harris, J., Calzetti, D., Gallagher, J. S., III, Smith, D. A., & Conselice, C. J. 2004, ApJ, 603, 503
- Heikkilä et al. (1999) Heikkilä, A., Johansson, L. E. B., & Olofsson, H. 1999, A&A, 344, 817
- Hollenbach & Tielens (1999) Hollenbach, D. J., & Tielens, A. G. G. M. 1999, Reviews of Modern Physics, 71, 173
- Hunt et al. (2005) Hunt, L., Bianchi, S., & Maiolino, R. 2005, A&A, 434, 849
- Inoue (2001) Inoue, A. K. 2001, AJ, 122, 1788
- Kleinmann & Low (1970) Kleinmann, D. E., & Low, F. J. 1970, ApJ, 159, L165
- Kobulnicky & Skillman (2008) Kobulnicky, H. A., & Skillman, E. D. 2008, AJ, 135, 527
- Kobulnicky et al. (1997) Kobulnicky, H. A., Skillman, E. D., Roy, J.-R., Walsh, J. R., & Rosa, M. R. 1997, ApJ, 477, 679
- Kroupa & Boily (2002) Kroupa, P., & Boily, C. M. 2002, MNRAS, 336, 1188
- Kroupa et al. (2013) Kroupa, P., Weidner, C., Pflamm-Altenburg, J., et al. 2013, Planets, Stars and Stellar Systems. Volume 5: Galactic Structure and Stellar Populations, 5, 115
- Lada et al. (2010) Lada, C. J., Lombardi, M., & Alves, J. F. 2010, ApJ, 724, 687
- Leitherer et al. (2014) Leitherer, C., Ekström, S., Meynet, G., et al. 2014, ApJS, 212, 14
- Leitherer et al. (1999) Leitherer, C., Schaerer, D., Goldader, J. D., et al. 1999, ApJS, 123, 3
- López-Sánchez et al. (2007) López-Sánchez, Á. R., Esteban, C., García-Rojas, J., Peimbert, M., & Rodríguez, M. 2007, ApJ, 656, 168
- Lundgren (2013) Lundgren, A. ALMA Cycle 2 Technical Handbook Version 1.1, ALMA
- Martin (1998) Martin, C. L. 1998, ApJ, 506, 222
- Martín-Hernández et al. (2005) Martín-Hernández, N. L., Schaerer, D., & Sauvage, M. 2005, A&A, 429, 449
- Meier et al. (2002) Meier, D. S., Turner, J. L., & Beck, S. C. 2002, AJ, 124, 877
- Meurer et al. (1995) Meurer, G. R., Heckman, T. M., Leitherer, C., et al. 1995, AJ, 110, 2665
- Miura et al. (2015) Miura, R. E., Espada, D., Sugai, H., Nakanishi, K., & Hirota, A. 2015, PASJ, 67, L1
- Murray et al. (2010) Murray, N., Quataert, E., & Thompson, T. A. 2010, ApJ, 709, 191
- Myers (1985) Myers, P. C. 1985, Protostars and Planets II, 81
- Oey & Clarke (2005) Oey, M. S., & Clarke, C. J. 2005, ApJ, 620, L43
- Rodríguez-Rico et al. (2007) Rodríguez-Rico, C. A., Goss, W. M., Turner, J. L., & Gómez, Y. 2007, ApJ, 670, 295
- Schaerer et al. (1997) Schaerer, D., Contini, T., Kunth, D., & Meynet, G. 1997, ApJ, 481, L75
- Sakamoto et al. (2006) Sakamoto, K., Ho, P. T. P., Iono, D., et al. 2006, ApJ, 636, 685
- Schilke et al. (1997) Schilke, P., Groesbeck, T. D., Blake, G. A., Phillips, & T. G. 1997, ApJS, 108, 301
- Silich & Tenorio-Tagle (2017) Silich, S., & Tenorio-Tagle, G. 2017, MNRAS, 465, 1375
- Smith et al. (2016) Smith, L. J., Crowther, P. A., Calzetti, D., & Sidoli, F. 2016, ApJ, 823, 38
- Stanway et al. (2016) Stanway, E. R., Eldridge, J. J., & Becker, G. D. 2016, MNRAS, 456, 485
- Tan et al. (2014) Tan, J. C., Beltrán, M. T., Caselli, P., et al. 2014, Protostars and Planets VI, 149
- Tenorio-Tagle et al. (2015) Tenorio-Tagle, G., Muñoz-Tuñón, C., Silich, S., & Cassisi, S. 2015, ApJ, 814, L8
- Tremonti et al. (2001) Tremonti, C. A., Calzetti, D., Leitherer, C., & Heckman, T. M. 2001, ApJ, 555, 322
- Turner & Beck (2004) Turner, J. L., & Beck, S. C. 2004, ApJ, 602, L85
- Turner et al. (2003) Turner, J. L., Beck, S. C., Crosthwaite, L. P., et al. 2003, Nature, 423, 621
- Turner et al. (2000) Turner, J. L., Beck, S. C., & Ho, P. T. P. 2000, ApJ, 532, L109
- Turner et al. (1997) Turner, J. L., Beck, S. C., & Hurt, R. L. 1997, ApJ, 474, L11
- Turner et al. (1998) Turner, J. L., Ho, P. T. P., & Beck, S. C. 1998, AJ, 116, 1212
- Turner et al. (2015) Turner, J.L., Beck, S.C., Benford, D.J., et al. 2015, Nature, 519, 7543
- Vanzi & Sauvage (2004) Vanzi, L., & Sauvage, M. 2004, A&A, 415, 509
- Walsh & Roy (1989) Walsh, J. R., & Roy, J.-R. 1989, MNRAS, 239, 297
- Weidner & Kroupa (2004) Weidner, C., & Kroupa, P. 2004, MNRAS, 348, 187