ALMA CO(3-2) Observations of Star-forming Filaments in a Gas-poor Dwarf Spheroidal Galaxy

Super star clusters (SSCs) are the birth places of most massive stars; these immense concentrations of hot stars have the potential to have an enormous impact on their host galaxies. Understanding the formation and evolution of SSCs is critical to understanding the energy budget, feedback, and regulation of star formation in galaxies and the environments that produce them.

NGC 5253 is a local (D = 3.8 Mpc; Sakai et al. 2004) dwarf spheroidal galaxy undergoing an intense starburst that has created a large population of massive star clusters over the course of a Gyr (Meurer et al. 1995; Gorjian 1996; Calzetti et al. 1997, 2004, 2015; Tremonti et al. 2001; Alonso-Herrero et al. 2004; Cresci et al. 2005; Martín-Hernández et al. 2005; de Grijs et al. 2013; Smith et al. 2016). NGC 5253 has a stellar mass of only $\sim 1.5\times {10}^{8}\,$M_⊙ (Martin 1998), and a dark matter mass that is likely 8–9 times higher (Mac Low & Ferrara 1999). This is a galaxy with dispersion-dominated kinematics and little rotation (Atherton et al. 1982; Caldwell & Phillips 1989). The abundant atomic hydrogen gas, comparable to the stellar mass at $\sim (2\mbox{--}3)\,\times$ 10⁸ M_⊙ (Kobulnicky & Skillman 1995, 2008; López-Sánchez et al. 2012), resides largely outside the galaxy, in filaments and streamers extending to ∼3–5 kpc (∼2–3 optical radii).

The dominant radio/IR H ii region in NGC 5253, the supernebula, is excited by a massive, young SSC (Turner et al. 2000; Gorjian et al. 2001; Meier et al. 2002; Turner & Beck 2004). The supernebula has an ionizing rate of ${N}_{\mathrm{Lyc}}=3.3\times {10}^{52}\,{{\rm{s}}}^{-1}$ (Turner et al. 2015; Bendo et al. 2017 corrected for dust following Inoue 2001), corresponding to an IR luminosity of $\sim 5\times {10}^{8}\,{L}_{\odot }$ (Gorjian et al. 2001), contributing to a total galactic infrared luminosity of $1.6\times {10}^{9}$ L_⊙ (Vanzi & Sauvage 2004; Hunt et al. 2005). Many investigators find that there are two massive clusters in the nucleus (e.g., Alonso-Herrero et al. 2004; Calzetti et al. 2015), one associated with the supernebula and one associated with the Hα peak; however, extinction within the central starburst region is very high, making it difficult to match radio, infrared, and optical observations (Calzetti et al. 1997; Turner et al. 2003; Martín-Hernández et al. 2005). Abundant molecular gas associated with such a luminous H ii region in the ultracompact stage is expected, but the CO(1-0) and (2-1) emission is weak. Roughly three-quarters of this CO emission, as well as the H i, is found along the minor axis, outside the central regions (Turner et al. 1997; Meier et al. 2002; Miura et al. 2015).

Previous observations of CO(3–2) from the Submillimeter Array (SMA) indicate that the star formation in the CO cloud (Cloud D) near the supernebula is very efficient, having converted >50% of its gas mass into stars (Turner et al. 2015). The unexpectedly bright CO(3–2) emission in the central starburst region, as compared to lower J CO lines, indicates that this central cloud, Cloud D (Meier et al. 2002), is warm, $T\sim 300$ K. However, the ∼4'' ×2'' beam of the SMA images does not separate the dense gas at the cluster scale.

With higher resolution and sensitivity, the Atacama Large Millimeter/Submillimeter Array (ALMA) can image NGC 5253 at the ∼5 pc cluster scale. We have been able to image the structure of its remarkable starburst. Here we present 2015 ALMA observations of the J = 3–2 rotational transition of CO in NGC 5253 with 03 (∼5.5 pc) resolution.

NGC 5253 was observed in ALMA Band 7 as a Cycle 1 (Early Science) program executed in Cycle 2 (ID = 2012.1.00105.S: PI = J. Turner) on 2015 June 4 and 5. Two fields, each with an 18'' field-of-view centered on 13:39:56.62–31.38.33.5 and 13:39:55.91–31.38.26.5 (ICRS), were observed simultaneously with 8383 seconds on both sources. The uv range covers ∼24.5–900 kλ; the largest structures sampled are 4''–8''. Spectral resolution of 244.14 kHz, or 1 km s⁻¹ per channel, resolves the CO(3–2) lines. Bandpass, flux, and phase were calibrated with J1427-42064, Titan, and J1342-2900, respectively. Absolute flux density calibration is estimated to be better than 10% in Band 7 Cycle 2 (Lundgren et al. 2013). Data calibration was performed using the code produced by the Joint ALMA Observatories with Common Astronomy Software Applications (CASA). Continuum emission was subtracted in the (u, v) plane before creating the line maps. The beam size is 033 × 027 at p.a. = −898 for ¹²CO(3–2) and 033 × 029 at p.a. = 890 for ¹³CO(3–2). The conversion to brightness is such that 1 K corresponds to ∼10–17 mJy, depending on the source geometry. The rms in a single 1 km s⁻¹ channel is 2.7 mJy/bm for ¹²CO(3-2) and 3.2 mJy/bm for ¹³CO. Integrated intensity images (MOM0) were created by summing up all of the emission that was greater than $\pm 1\sigma$ in all of the channels; for the intensity-weighted mean velocity map (MOM1) the cutoff was $\pm 5\sigma$.

The single dish integrated flux density reported by Meier et al. (2001) with the 22'' CSO beam was 3.4 K km s⁻¹, corresponding to ∼155–170 Jy $\mathrm{km}\,{{\rm{s}}}^{-1}$, depending on the assumed source geometry. The integrated flux density recovered in the lower resolution SMA synthesis map of Turner et al. (2015) was 110 ± 20 Jy $\mathrm{km}\,{{\rm{s}}}^{-1}$. We detect ≈50% of the expected integrated flux density from the SMA maps in the form of individual clumps; the remaining flux density is extended and much has been resolved out because of the lack of short spacings in the array. Although the individual dense clouds that emit CO(3-2) are not likely to be large, collections of such clouds with low (≲10%) volume filling may lie below even the ALMA sensitivity limit.

3.1. The Dense Gas is Clumpy and the Clumps Form Filaments

The CO(3-2) integrated line intensity, ${S}_{\mathrm{CO}}=\int {I}_{\mathrm{CO}}{dv}$, in NGC 5253 is shown in Figure 1, superimposed on an archival Hubble Space Telescope/Advanced Camera for Surveys (HST/ACS) image of the galaxy. The J = 3 state of CO has an excitation energy of 33 K, requiring a critical density of $n\gtrsim {\rm{20,000}}\,{\mathrm{cm}}^{-3}$ for excitation, thus the J = 3 state of CO tends to trace warm and dense star-forming molecular clouds (Heiderman et al. 2010; Lada et al. 2010).

Zoom In Zoom Out Reset image size

Within the ∼30'' by 60'' region covered by the two ALMA pointings, CO is detected over an area ≈20'' or 370 pc. We identify five regions, labeled in Figure 2, whose properties are given in Table 1. In addition to Cloud D (Meier et al. 2002; Turner et al. 2015), which has the brightest CO(3–2) clumps, there is the Streamer, which lies along the optical dust lane to the southeast, Cloud E, as observed in Meier et al. (2002), northwest of the center, Cloud F from Turner et al. (2015) to the southwest, and finally, the Southern filament, a group of clumps due south of the center. Individual filaments have characteristic widths of only a few clumps, ∼05, but typical lengths of ∼5'' or 90 pc. Individual clumps at these densities have gravitational collapse free-fall timescales of ${t}_{{ff}}\lesssim 3.6\,\times {10}^{5}$ years.

Zoom In Zoom Out Reset image size

Table 1.  Source ¹²CO and ¹³CO Fluxes, Ratio, and Masses

Source	ICRS R.A.a	ICRS Decl.a	vsysb	12CO(3-2)c	13CO(3-2)c	$\tfrac{{}^{12}\mathrm{CO}}{{}^{13}\mathrm{CO}}$	12CO Massd	13CO Masse
	13h39m	−31°38'	(km s−1)	(Jy km s−1 )	(Jy km s−1 )	ratio	(103 ${M}_{\odot }$)	(103 ${M}_{\odot }$)
D1	55.965	24.36	385	2.4	0.06 ± 0.05	${40}_{-20}^{+30}$	100	3.0
D2	56.12	23.1	394	0.8	0.07 ± 0.02	12 ± 4	33	0.6
D3	56.03	22.9	395	1.7	0.28 ± 0.04	6 ± 1	70	2.6
D4	55.91	25.0	402	8.4	1.2 ± 0.07	7 ± 0.4	345	11
D5	55.79	25.6	384	0.87	0.12 ± 0.02	7 ± 1	36	1.1
D6	55.88	27.0	409	1.8	0.25 ± 0.04	7 ± 1	74	2.3
Squiggle (D7)	56.0	27	403	4.1	0.5 ± 0.1	9 ± 2	170	4.4
Center Sum of D Clouds	55.95	25	385	20	2.4 ± 0.1	8 ± 0.5	823	24.7
Streamerf	56.5	33	430	8.79	0.86 ± 0.2	10 ± 2	360	8.1
Cloud E	55.5	19	400	12.55	0.8 ± 0.2	16 ± 4	510	7.5
Cloud F	55.8	31	368	6.82	0.9 ± 0.1	8 ± 1	280	8.2
Southern Filament	55.9	367	413	2.45	0.54 ± 0.1	5 ± 1	100	5.1
Overall Sum	56.3	30	380	51	5.5	9	2073	53.6

Notes. All measurements are made from velocity-selected, primary beam corrected MOM0 integrated flux density maps. As ¹³CO emission is weak, ¹²CO was used to define the velocity regions for each cloud, and all flux was summed up for that velocity range.

^aPrecision in coordinates depends on the compactness and structure of the source. ^b±5 km s⁻¹. ^cUncertainties in ¹²CO flux density are 10% from flux density calibration uncertainties unless otherwise stated. Uncertainties in ¹³CO flux density, which is mostly signal-to-noise limited, are computed from the individual channel rms times $\sqrt{{N}_{\mathrm{chan}}{N}_{\mathrm{beam}}}$. ^dMass derived assuming optically thick emission, using a conversion factor of ${\alpha }_{\mathrm{CO}}=4.7\times {10}^{20}$ cm⁻² (K kms⁻¹)⁻¹, and an estimated CO(3–2)/CO(1–0) ratio of 7. This mass will overestimate the cloud mass for optically thin, warm clouds and for clouds in which enclosed stellar mass is significant. Mass includes He. ^eMass derived assuming optically thin ¹³CO flux density and 20 K, a Galactic value of [CO]/[H₂] = 8.5 × 10⁻⁵ and an isotopic ratio of 40, except for Cloud D1, where T = 300 K is assumed. The Rayleigh–Jeans approximation is not used. This mass will underestimate the true cloud mass for optically thick emission or if the clouds are warmer than 20 K. Mass includes He. ^fAll integrated line intensities, particularly ¹²CO, are subject to the missing short spacing data, and therefore this image is unable to reconstruct structures >4'' in size This will not affect small clouds such as the Cloud D individual clumps. The Streamer, however, appears to be missing flux on the whole. See the text.

Download table as:  ASCIITypeset image

Previous observations of NGC 5253 established that lower J CO emission is located primarily in the Streamer (Turner et al. 1997; Meier et al. 2002; Miura et al. 2015; Turner et al. 2015), following the prominent dust lane to the east along the minor axis of the galaxy. Optical emission lines, probably from the surface of the streamer, are also present in this feature (Graham 1981; Martin 1998; Zastrow et al. 2011). At the 10 times higher spatial resolution of the ALMA images, the Streamer splits up into clumps roughly a beam size (∼03, 5 pc) in extent or less, arranged in filamentary structures that are ∼25–200 pc in extent.

The properties and kinematics of the clumps within the filaments are discussed in this section. Individual clouds in the central starburst region, within Cloud D, are discussed further in Section 5.

3.2. The Clumps Form Filamentary Structures that are Distinct in Velocity

3.3. Conditions in the Filament Clumps

The central ∼150 pc region surrounding the supernebula in NGC 5253 (Cloud D of Meier et al. 2002) contains an extraordinary concentration of young stars. Here we use ALMA CO(3-2) images and previous lower resolution observations of CO(2–1) and CO(3–2) to understand the properties of dense gas and related star formation.

The CO(3–2) integrated line intensity is shown with two tracers of star formation, Hα and radio continuum, in Figures 3 and 4. The CO(3–2) clumps in the ALMA image and the Hα emission in NGC 5253 do not show an obvious correlation (Figure 3). In fact on ∼5 pc scales these properties appear to be anti-correlated, with CO(3–2) appearing in dust lanes in the Hα image. The correspondence of CO with dust lanes is not surprising: CO forms at ${A}_{V}\gtrsim 2$ (e.g., Bisbas et al. 2015), corresponding to an overall cloud column of ${A}_{V}\gtrsim 4$. The bright central H ii region for which NGC 5253 is so well known (Burbidge & Burbidge 1962) is centered on the heavily extincted supernebula (Turner et al. 2003; Alonso-Herrero et al. 2004; Turner & Beck 2004). Loops and arcs in the Hα suggest that the supernebula is responsible for not only the large scale leakage of photons, but also possibly for winds. Both effects suggest patchiness of extinction in the supernebula core, consistent with the kinematics and brightness of Cloud D1 (Turner et al. 2017).

Zoom In Zoom Out Reset image size

Radio continuum emission traces ionizing photons through free–free emission; like Hα, it traces emission measure, but is far weaker than Hα. Radio continuum is the most reliable tracer of star formation in regions of high extinction. In Figure 4, a VLA 6 cm image (Turner et al. 1998) is overlaid on the ALMA CO(3–2) map. Except for an isolated supernova remnant to the east, the remaining 6 cm emission is thermal free–free emission from the H ii regions. The ∼2'' resolution of this VLA image is much lower than the ALMA images, but an excellent spatial correlation of the free–free continuum with CO(3–2) is apparent. The dominant radio continuum source in this image is the supernebula, associated with Cloud D1. Radio continuum peaks are also associated with Cloud F and the Southern Filament, located $\sim 140\mbox{--}180$ pc to the southwest of the supernebula, as well as clouds to the northwest within the Cloud D supernebula complex. The Streamer also shows radio continuum emission, presumably from the same gas producing the "ionization cone" seen in [O iii] λ5007, Hα, [S ii], and [S iii] $\lambda \lambda 6716$, 9069 (Graham 1981; Martin 1998; Zastrow et al. 2011).

Zoom In Zoom Out Reset image size

The relation of gas, star formation, and molecular gas depletion time can be estimated from the radio continuum and CO(3–2). Because free–free emission is weak and difficult to detect at subarcsecond scales, we quantify the relation of the 6 cm free–free and CO using the VLA maps with larger 1''–2'' (20–40 pc) beams (Turner et al. 1998) and images of CO(2–1) from the Owens Valley Millimeter (OVRO) Array (Meier et al. 2002) for gas masses. The Streamer has the brightest emission in low J CO images, and contains most of the molecular mass in NGC 5253. The estimated gas mass of the streamer based on CO(2-1) is $\sim 1.7\times {10}^{6}$ M_⊙ (corresponding to Cloud C of Meier et al. 2002, corrected to ${X}_{\mathrm{CO}}=4.7\times {10}^{20}\,{\mathrm{cm}}^{-2}{({\rm{K}}{\mathrm{km}}^{-1})}^{-1}$). The gas surface density based on their 96×48 beam is ∼90 M_⊙ pc⁻², comparable to the surface densities inferred for giant molecular clouds (GMCs) in the galaxy (Lombardi et al. 2010). Using Starburst 99 models and a full Kroupa initial mass function (IMF), the relation between the young stellar mass and the Lyman continuum rate is $\mathrm{log}[(M/{M}_{\odot })/({N}_{\mathrm{Lyc}}/{{\rm{s}}}^{-1})]=-46.6$, with ${N}_{\mathrm{Lyc}}=1.25\times {10}^{50}\,{{\rm{s}}}^{-1}$ ${(T/{10}^{4}K)}^{-0.507}{(\nu /100\mathrm{GHz})}^{0.118}$ ${D}_{\mathrm{Mpc}}^{2}\,{S}_{3.3\mathrm{mm}}(\mathrm{mJy})$. For a mean star-formation timescale of 5 Myr for the star formation, and for a radio continuum flux of ∼1 mJy for Cloud C (Meier et al. 2002), which contains most of the Streamer emission, the estimated mass depletion timescale is ${\tau }_{\mathrm{dep}}={M}_{\mathrm{gas}}$/star-formation rate (SFR) $\approx \,0.3\,\mathrm{Gyr}$ for the Streamer. We estimate that this number is uncertain to factors of 3–5, due to the unknown star-formation timescale and the uncertain gas mass.

The central 1'' (∼18.4 pc) region surrounding the supernebula, Cloud D, is forming stars at a SFR of ∼ 0.1 ${M}_{\odot }$ yr⁻¹ (Turner et al. 2015; Bendo et al. 2017). CO(2–1) images indicate that ${M}_{{{\rm{H}}}_{2}}\,\sim$ 8 × 10⁵ ${M}_{\odot }$ within Cloud D. If this gas is all associated with the current star formation, the gas depletion timescale is ${\tau }_{\mathrm{SF}}\sim 8\,\mathrm{Myr}$. However, these ALMA maps reveal that Cloud D actually consists of many clumps (Section 5), and it is unclear if all of them are related to the currently forming massive cluster. If some of this gas is not associated with the current star formation as suggested by the gas kinematics Turner et al. (2017), then this is an upper limit to ${\tau }_{\mathrm{SF}}$.

The molecular gas depletion time in the Streamer, ${\tau }_{\mathrm{dep}}=0.3\,\mathrm{Gyr}$, is less than the median value of 0.8 Gyr for local low-metallicity galaxies as determined by Schruba et al. (2011), but within the observed spread. While CO-dark H₂ may be partly responsible for the low value, presumably the same holds for the general low-metallicity sample as a whole. It may also be that the ionized gas of the Streamer is caused by UV photons from the nuclear starburst, and not from star formation within the streamer, which would increase the value of ${\tau }_{\mathrm{dep}}$.

Cloud D has a very short molecular gas depletion time of ${\tau }_{\mathrm{dep}}=8\,\mathrm{Myr}$, a factor of ∼100 times smaller than that of the typical low-metallicity galaxy in the Schruba et al. (2011) sample. We caution that the central gas mass is uncertain, and that the star-formation rate depends on our assumption of 5 Myr for the age; the latter is probably uncertain to factor two, although there is an additional uncertainty in the spread in star-formation age. However, these uncertainties of a factor of a few are not expected to be large enough to explain the two order of magnitude shorter depletion time in Cloud D1. This star formation is unusually efficient.

Lada et al. (2013) point out that individual GMCs in the galaxy do not obey a Kennicutt–Schmidt law; the star-formation rate surface density varies more than two orders of magnitude among Galactic GMCs. They propose that the Schmidt–Kennicutt law for galaxies (Kennicutt 1989, 1998) is a function of relative dense gas fraction and beam dilution on kpc scales. With these high-resolution ALMA images we resolve individual GMCs in NGC 5253, and thus it is not surprising that we see a spread in star-formation surface densities similar to that observed in Galactic GMCs. However by Galactic standards, Cloud D is extraordinarily efficient at forming stars.

Images of the ¹²CO(3–2) and ¹³CO(3–2) for the central ∼100 pc region corresponding to Cloud D are shown in Figure 5. With the high ∼5 pc resolution of the ALMA images, the central Cloud D detected in the CO(3–2) images from the Submillimeter Array (Turner et al. 2015), breaks into a number of smaller clouds with diameters of $\sim 5\mbox{--}20$ pc.

Zoom In Zoom Out Reset image size

The larger CO structures are labeled and their characteristics are detailed in Table 1. ¹³CO(3–2) emission provides valuable additional information on the optical depth of the CO and the gas mass in the central star-forming region of NGC 5253. All of the clouds in the central region are at least weakly detected in ¹³CO. ¹³CO(3–2) emission is also detected for the brighter clouds in the Streamer, not shown here; but this emission is very faint ($\lesssim 2\sigma$).

Clouds D3, D4, D5, D6, and Cloud F have ¹²CO(3-2)/¹³CO(3-2) flux ratios of $\sim 6\mbox{--}8$. These values are somewhat lower than the value of ∼13 observed in Orion (Schilke et al. 1997). We infer that, as in Orion, these clouds are optically thick in ¹²CO(3-2). In fact, half of the measured 2.4 Jy km s⁻¹ of ¹³CO(3-2) emission arises in Cloud D4, an apparently massive cloud in the central region located about 15–20 pc southwest of the supernebula that does not host obvious star formation and has an optically thick ¹²CO/¹³CO ratio of 7. Cloud D4 also appears in H₂ λ2.12 μm images of the region (Cresci et al. 2005), suggesting the presence of shocks (turbulence) or fluorescence; given the extinction implied by the CO, it is likely that the H₂ emission arises on the near surface of the cloud.

The highest ¹²CO(3–2)/¹³CO(3–2) ratio is ∼40 for Cloud D1, the small (r ∼ 2.8 pc) cloud coincident with the giant central H ii region known as the supernebula. This value is uncertain because the ¹³CO(3–2) is so weak, but is unlikely to be < 20. Star formation may still be ongoing within Cloud D1, which appears to coexist with the $\sim 2.5\times {10}^{5}$ M_⊙ SSC and H ii region (Turner et al. 2017). Based on the high ¹²CO(3–2)/¹²CO(2–1) ratios of lower resolution maps, Turner et al. (2015) suggested that the larger Cloud D is optically thin. The ¹²CO(3–2)/¹³CO(3–2) ratio for Cloud D1 is consistent with the suggestion. This situation would be unusual; Galactic GMCs are optically thick in these lines. Cloud D1 may be optically thin because the gas is hot (Turner et al. 2015) and the molecules spread out over the rotational ladder. At 300 K, the partition function is 100, such that only 1% of the CO molecules will be in the ground state and ∼3% in J = 1. It is also possible that Cloud D1 could also have selective photodissociation or that the ¹²CO(3–2) and ¹³CO(3–2) could probe different physical environments due to temperature gradients in cloud surfaces due to the presence of strong radiation fields in the cluster environment. Cloud D1 is also bright in ${{\rm{H}}}_{2}\,\lambda 2.12\,\mu {\rm{m}}$ emission (Cresci et al. 2005); given that Cloud D1 is likely to consist of multiple star-forming clumps, the H₂ emission may arise on the surfaces of these individual clumps. (However Clouds D4 and D6, which are optically thick, are also bright in the H₂ line).

Cloud D2, located ∼40 pc to the northeast of Cloud D1, may also be optically thin with a ¹²CO/¹³CO ratio of ∼12, which is higher than average for this galaxy. As this cloud appears in radio continuum emission at cm wavelengths (Beck 2017, private communication) and in ${{\rm{H}}}_{2}\,\lambda 2.12\,\mu {\rm{m}}$ emission (Cresci et al. 2005), Cloud D2 appears similar to Cloud D1, if smaller in scale. Thus, Cloud D2 appears to be a young, compact cluster that has warm molecular gas, although significantly smaller than that of the massive cluster within D1.

If ¹³CO(3-2) is optically thin in some clouds, we can use the ¹³CO integrated line intensities to estimate optically thin masses. These masses are given in Table 1. For Cloud D1 we adopt a temperature of 300 K, based on RADEX models of the CO(3–2)/CO(2–1) line ratio (Turner et al. 2015), and for clouds besides D1 we adopt 20 K. The optically thin masses are not very sensitive to temperature for this line until temperatures exceed ∼100 K, although if the clouds are actually very warm, these masses will be underestimates. If we adopt a Galactic [CO]/[${{\rm{H}}}_{2}]$ ratio of $8.5\times {10}^{-5}\,$ and a [¹²CO]/[¹³CO] abundance ratio of 40, the minimum indicated by the Cloud D1 line ratio, we obtain cloud masses ${M}_{\mathrm{Cl}}\,\sim$ 0.6–11 $\times {10}^{3}{M}_{\odot }$ (Table 1), with the most massive cloud being the apparently quiescent Cloud D4.

The masses based on ¹²CO(3-2), assuming a conversion factor (¹²CO Masses), and those based on ¹³CO(3-2). assuming optically thin emission (¹³CO Masses), differ by an order of magnitude. There are many possible reasons for the differences. For optically thin clouds, D1, D2, the values of [CO]/[${{\rm{H}}}_{2}$], and the [¹³CO]/[¹²CO] abundance ratios are sources of systemic uncertainty, since they may differ significantly from Galactic values in this low-metallicity galaxy. For optically thick clouds, the conversion factor cannot account for clouds that are unusually warm; or that have significant stellar mass contained within the cloud, such as Clouds D1 and D2; or for clouds that have significant CO-dark gas, although this is not likely to be a major mass contributor in these dense clouds (Langer et al. 2014).

The CO(3-2) critical density is aligned with the gas densities expected for star-forming gas. Even the highest estimates of cloud mass indicate that the gas masses of these clouds are fairly small. The largest cloud, D4, has an estimated mass of $3.4\times {10}^{5}\,{M}_{\odot }$, a relatively low mass considering the $\sim {10}^{5}\,{M}_{\odot }$ star clusters that NGC 5253 has been forming. The current suite of clouds does not seem capable of forming a massive SSC, unless mergers of clouds take place, perhaps along accreting filaments, and if the star formation is extraordinarily efficient.

We present new ALMA images of ¹²CO(3–2) and ¹³CO(3–2) emission from the nearby (3.8 Mpc) dwarf galaxy NGC 5253. The spatial resolution is ∼03, ten times smaller in diameter than that of the best previous maps. The images reveal dense ($\sim {\rm{20,000}}\,{\mathrm{cm}}^{-3}$) molecular clumps ∼5 pc in diameter with estimated masses of a few thousand ${M}_{\odot }$.

While atomic and ionized filaments are a feature of NGC 5253 on the largest scales (Kobulnicky & Skillman 2008), these observations are the first to detect clear, kinematically distinct filaments in the molecular gas. The inner 200 pc of NGC 5253 holds four filaments, composed of small (∼5 pc diameter) clouds. These filaments extend to ∼50–150 pc in length. The Streamer filament, detected previously in CO(1-0) and CO(2-1) in lower resolution images, is aligned along the prominent dust lane to the east of the galaxy. The Streamer is redshifted with respect to the systemic velocity, and thus appears to be falling into the galaxy. Also redshifted and in the foreground, thus apparently infalling, is the weaker Southern filament to the south of the galaxy center. Cloud F, a filament southwest of the center, is blueshifted with respect to the galaxy but shows no apparent extinction and may be on the far side of the galaxy. We propose that the CO filaments form the inner position of H i filaments that are falling into the galaxy from outside. This model particularly explains the Streamer along the prominent dust lane (redshifted) and Cloud F (blueshifted, with no apparent foreground extinction). The clearest filaments have length-to-width aspect ratios of ≈20. Dense clumps in the filaments are likely to be overpressured relative to ambient gas, and probably gravitationally supported; they may be likely sites of future star formation.

The central cloud surrounding the supernebula, Cloud D, previously detected as a bright CO(3–2) source, breaks up into multiple clouds. Some of these clouds are optically thin in CO(3-2), as indicated by high ¹²CO/¹³CO flux ratios of ∼12–40, higher than the average cloud ratio of ∼6–8. Cloud D1 is coincident with the giant H ii region and supernebula; its flux ratio of ${}^{12}\mathrm{CO}{/}^{13}\mathrm{CO}\sim 40$ suggests that the [¹²CO]/[¹³CO] ratio in NGC 5253 is $\gtrsim 40$. Cloud D2 appears to contain a smaller young cluster. We propose that Clouds D1 and D2 are most likely optically thin in CO(3–2) because the molecular gas is hot, due to the proximity of young and massive star clusters, although it is also possible that optical depth effects in the presence of high temperature gradients could produce the high observed ratios. These clouds are likely to be brighter than would be expected based on their gas columns alone.

From radio continuum flux densities and lower J CO lines, we derive average gas depletion timescales of 0.3 Gyr for the Streamer and 8 Myr for Cloud D1 coincident with the supernebula. The masses we obtain for the individual clumps from the ALMA observations, from assumption of optically thin emission, for the thin clouds, and from ¹²CO(3–2) emission, for the thick clouds, differ by about an order of magnitude. Uncertainties in the H₂ masses are due to unknown [CO]/[H₂] for this low metallicity, high radiation environment, unknown CO(3–2)/CO(1–0) ratio, and uncertainty in the conversion factor applicability. However, even the highest estimates of mass for the clumps,$\sim {10}^{4}\mbox{--}{10}^{5}\,{M}_{\odot }$, are small compared to the masses of the star clusters that are currently forming. This suggests that filament mergers and/or highly efficient star formation is taking place in NGC 5253. Further high-resolution CO observations may clarify these issues and allow for improved masses. Even with these uncertainties, it appears that the star formation in the central Cloud D1 associated with the supernebula is extraordinarily efficient.

This paper makes use of the following ALMA data: ADS/JAO.ALMA# 2012.1.00105.S. ALMA is a partnership of the ESO (representing its member states), NSF (USA), and NINS (Japan), together with NRC (Canada), MOST, ASIAA (Taiwan), and KASI (Republic of Korea), in cooperation with the Republic of Chile. The Joint ALMA Observatory is operated by the ESO, AUI/NRAO, and NAOJ. The National Radio Astronomy Observatory (NRAO) is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. Support for this work was provided by the NSF through award GSSP SOSPA2-016 from the NRAO to S.M.C. and grant AST 1515570 to J.L.T., and by the UCLA Academic Senate through a COR seed grant. The authors wish to thank W. M. Goss, P. T. P. Ho, and the anonymous referee for their helpful comments.

Facility: ALMA - Atacama Large Millimeter Array.