NOEMA Observations of a Molecular Cloud in the Low-metallicity Galaxy Kiso 5639

Kiso 5639 (Miyauchi-Isobe et al. 2010) is a dwarf galaxy with a kpc-size starburst at one end, giving the system a tadpole, or cometary, shape (Elmegreen et al. 2012). The rotation speed of ∼35 km s⁻¹ (Sánchez Almeida et al. 2013) combined with a radius of 1.2 kpc in the bright part of the disk implies that the dynamical mass there is 3 × 10⁸/sin² i M_⊙, which is a factor of ∼6 larger than the stellar mass of 5 × 10⁷ M_⊙ from Sloan Digital Sky Survey (SDSS) photometry (Elmegreen et al. 2012) and comparable to the total H i mass of ∼3 × 10⁸ M_⊙ (Salzer et al. 2002).

The galaxy is a member of our spectroscopic survey of 22 low-metallicity dwarfs where the metallicity in the starburst "head" appears to be less than in the rest of the galaxy (the "tail"); 16 others in this survey have the same metallicity drop, three do not, and two are ambiguous (Sánchez Almeida et al. 2013, 2014, 2015). This peculiar pattern of metallicity suggests that the starbursts in these systems were triggered by accreting gas with lower metallicity than in the rest of the galaxy. Other examples of metallicity drops were reported in Levesque et al. (2011), Haurberg et al. (2013), and Lagos et al. (2018).

Hubble Space Telescope (HST) observations of Kiso 5639 (Elmegreen et al. 2016) in six UV-optical and Hα filters were used to resolve the head and derive the star formation properties. The head contains 14 young star clusters more massive than 10⁴ M_⊙ and an overall clustering fraction for star formation of 25%–40%. The Hα luminosity of the core region of the head is (8.8 ± 0.16) × 10³⁹ erg s⁻¹ inside an area of 3.6 × 3.6 square arcsec. The corresponding star formation rate is ∼0.04 M_⊙ yr⁻¹. This rate is based on a conversion factor of SFR = 4.7 × 10⁻⁴²L(Hα) M_⊙ yr⁻¹(erg s⁻¹)⁻¹ that is appropriate for low metallicity, which introduces a factor of 0.87 times the standard value (Hunter et al. 2010; Kennicutt & Evans 2012). For a distance of 24.5 Mpc (Elmegreen et al. 2012), the corresponding area is 0.18 kpc² (430 × 430 square pc), and the star formation rate per unit area is 0.23 M_⊙ pc⁻² Myr ⁻¹.

This is a high rate for a small galaxy, and it suggests that there is a large reservoir of dense gas in the head. For the conventional molecular gas consumption time of ∼2 Gyr (Bigiel et al. 2008), the molecular mass surface density would be 450 M_⊙ pc⁻², which is ∼40 times higher than the average stellar surface density and ∼7 times higher than the average dynamical surface density in the bright part of the disk. For the Kennicutt–Schmidt relation in whole galaxies from Figure 11 in Kennicutt & Evans (2012), the required gas surface density would be ∼200 M_⊙ pc⁻². These high gas surface densities suggest that Kiso 5639 is lopsided and recently accreted at least ∼10⁸ M_⊙ of gas on one side. Simulations of such a process can be found in Verbeke et al. (2014) and Ceverino et al. (2016).

The detection of CO emission from the head of Kiso 5639 would help to clarify the situation. The metallicity of this galaxy, determined with the 2.5 m Nordic Optical Telescope from the [N ii]λ6583 to Hα line ratio in Sánchez Almeida et al. (2013), was $12+\mathrm{log}({\rm{O}}/{\rm{H}})\sim 7.48\pm 0.04$. CO is highly underabundant compared to H₂ at low metallicity (Leroy et al. 2011). The lowest metallicities in regions detected in CO, i.e., 7.5 in Sextens A and DDO 70 (Shi et al. 2015, 2016) and 7.8 in WLM (Elmegreen et al. 2013; Rubio et al. 2015), occur in dwarf irregular galaxies where the CO clouds are small and in the centers of large H i and H₂ clouds. For the large molecular mass expected in Kiso 5639 and in other galaxies of its type, CO emission should be more evident.

To search for CO in Kiso 5639, we observed CO(1–0) in the head region with NOEMA in D configuration. The result was a detection corresponding to a gas surface density similar to that estimated above, depending on the assumed α_CO conversion factor. Because the metallicity is important for this α_CO, we re-observed Kiso 5639 in several optical lines, including [O ii]λ3727 and [O iii]λ4363, using the 10 m Gran Telescopio Canarias (GTC) to get the metallicity again. The observations, results, and implications are discussed in the next three sections.

2.1. Metallicity

2.2. CO Observations

We obtained observations of Kiso 5639 on 2017 May, July, and August with the Northern Extended Millimeter Array (NOEMA) at Plateau de Bure, in the CO (J = 1–0) line, at 114.572 GHz redshifted to the velocity of Kiso 5639. The observations were carried out with either seven or eight antennas in Configuration D, with baselines between 15 and 175 m. We used the Widex correlator, with a total bandwidth of 3.6 GHz and a native spectral resolution of 1.95 MHz (5.1 km s⁻¹).

Data reduction, calibration, and imaging were performed with CLIC and MAPPING softwares of GILDAS,⁸ using standard procedures. Images were reconstructed using the natural weighting, resulting in a synthesized beam of 28 × 37 (PA = 26°). The rms noise in the CO cube is 14.3 mK (1.6 mJy beam⁻¹) in a 5.1 km s⁻¹ channel.

The resulting data cube was integrated between local standard of rest (LSR) velocities of −62.3 and −26.5 km s⁻¹, where CO emission was observed. The rms was measured in the integrated image and a mask for pixels with emission greater than 2σ was made. The average spectrum of Kiso 5639 was obtained within this mask. A 1D Gaussian fit to this spectrum gives V_LSR = −42.1 ± 2.2 km s⁻¹, FWHM = 35.2 ± 5.3 km s⁻¹, peak emission 32.1 ± 4.5 mK, and average integrated line profile, 1203 ± 247 mK km s⁻¹, as summarized in Table 1.

Table 1.  CO Cloud Properties

Cloud	VLSR	Tpeak	FWHM	Ave. Line Int.	Radius	Mass	Surface Density	Virial Ratio
	km s−1	mK	km s−1	K km s−1	arcsec	×106 M⊙	M⊙ pc−2	5Rσ2/(GM)
Total	−42.1 ± 2.2	32.1 ± 4.5	35.2 ± 5.3	1203 ± 247	1.7	29	120	1.8
East	−35.2 ± 3.8	29.8 ± 4.3	20.8 ± 7.1	659 ± 245	<1.6	16	66	<1.1
West	−54.3 ± 3.8	24.7 ± 7.3	17.8 ± 6.5	469 ± 219	<1.6	11	47	<1.1

Download table as:  ASCIITypeset image

Figure 1 shows a map of the main emission and the integrated spectrum. The velocity resolution is 5.11 km s⁻¹, the half-power beam size is 285 × 372, and the rms is 1.6 mJy/beam. The equivalent radius of the source is 235 (280 pc), which corresponds to the area of the detection contour, 2.45 × 10⁵ pc². The emission has two peaks separated by 2'', which is 240 pc. Spectra determined for each peak are shown in Figure 2. The blue histogram is from one of the peaks and the green histogram is from the other; the red double peak histogram is the composite. Each histogram has a Gaussian fit, as indicated by the red lines and the black line. The integrated line fits to these spectra and the derived quantities are in Table 1.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Several other regions inside the primary beam of the NOEMA telescope were also suspected of containing CO emission, but the brightness temperatures were less than 20% of the peak shown in the figure, and we do not consider them to be definitive detections of molecular gas. These include the contoured regions to the southwest and northwest of the main emission in Figure 1.

Conversion from the observed CO flux to a molecular mass depends on the conversion factor, α_CO. Hunt et al. (2015) suggested an extrapolation of α_CO from the solar neighborhood value to low-metallicity Z as α_CO = 4.3(Z/Z_⊙)⁻² M_⊙ pc⁻²(K km s⁻¹)⁻¹, where Z_⊙ is the solar metallicity corresponding to $12+\mathrm{log}({\rm{O}}/{\rm{H}})\,=8.69$ (Asplund et al. 2009), and Z is the metallicity for Kiso 5639 corresponding to $12+\mathrm{log}({\rm{O}}/{\rm{H}})=7.83$. The local conversion factor is α_CO,⊙ = 4.3 M_⊙ pc⁻² (K km s⁻¹)⁻¹, including He and heavy elements (Bolatto et al. 2013). Similarly, Amorín et al. (2016) suggested an extrapolation as Z to the power −1.5. In the first case, the result for Kiso 5639 would be α_CO = 225 in these units, and in the second case it would be 84. Compared to the WLM galaxy, where α_WLM ∼ 124 ± 60 (Elmegreen et al. 2013) and $12+\mathrm{log}\,{({\rm{O}}/{\rm{H}})}_{\mathrm{WLM}}=7.8$, α_CO = 108 and 112 for the same power-law scalings. On the other hand, if we extrapolate between the low-Z α_CO values determined by Shi et al. (2016), we would get α_CO ∼ 600 and 900 at Z = 7.8, respectively, which results in a much larger molecular mass. Here we take α_CO ∼ 100 as a conservative estimate.

For the observed CO average integrated line profile, 1.2 ± 0.2 K km s⁻¹, estimated α_CO = 100 M_⊙ pc⁻²(K km s⁻¹)⁻¹, and CO source area, 2.45 × 10⁵ pc², the H₂ mass is the product of these, 2.9 × 10⁷ M_⊙, with the largest uncertainty in the value of α_CO, which is probably a factor of 2. The mass divided by the total emitting area is the molecular surface density, Σ_mol = 120 M_⊙ pc⁻².

In this same region, the star formation rate is ∼0.04 M_⊙ yr⁻¹ and the star formation surface density is 0.23 M_⊙ pc⁻² Myr ⁻¹, as given above. These imply that the molecular gas consumption time is 0.5–0.7 Gyr, which is slightly less than in normal spiral galaxy disks (Bigiel et al. 2008; Schruba et al. 2011). This time is about the same as in the SMC, where the metallicity is also low (Z ∼ 0.2Z_⊙) and α_CO is high, ∼220 M_⊙ pc⁻²(K km s⁻¹)⁻¹ (Bolatto et al. 2011).

Following the same method, the masses of the two components of the emission in Figures 1 and 2 can be determined (see Table 1). The component in the east has a velocity of −35 km s⁻¹ and an average integrated line profile of 0.66 ± 0.25 K km s⁻¹, and the component in the west has a velocity of −54 km s¹ and an average integrated line profile of 0.47 ± 0.22 K km s⁻¹. We assume that each has the same total area. Then the masses are these line integrals multiplied by the areas and the α_CO factor, which are 1.6 × 10⁷ M_⊙ in the east and 1.1 × 10⁷ M_⊙ in the west, with surface densities of 66 M_⊙ pc⁻² and 47 M_⊙ pc⁻², respectively. The velocity difference between the components, 19 km s⁻¹, is nearly half of the rotation velocity of the galaxy, 34.7 ± 6.2 km s⁻¹ (Sánchez Almeida et al. 2013), suggesting a catastrophic event.

The masses, radii, and velocity dispersions can be combined to determine the virial ratio, 5Rσ²/(GM), where σ = 0.42 × FWHM. These ratios require a deconvolved radius, R, which is (2.35²–2.85 × 3.73/4)^0.5 = 169 for the total cloud (using the source and beam sizes from above). The radii of the components are taken to have upper limits equal to the average beam size of 163. The results are in Table 1. The virial ratios are in the range <1.1 to 1.8, which suggests that the whole cloud and the two separate parts are gravitationally self-bound.

Figure 3 shows the HST image of Kiso 5639 in Hα, V, and B bands with superimposed CO contours at values of 2σ, 3σ, and 4σ. The main molecular cloud is 22 = 260 pc to the east of the bright star formation region. This offset means that comparisons to the Kennicutt–Schmidt relation are not exactly appropriate. Still, the molecular gas reservoir for the starburst appears to be of sufficient mass to explain the observed young stars.

Zoom In Zoom Out Reset image size

Figure 4 shows the same CO contours on the HST Hα image of the whole galaxy. An expanded view of the head region is on the left, where the black contours are 3%, 10%, and 20% of the peak Hα emission. Offset positions are as in Figure 1.

Zoom In Zoom Out Reset image size

Subject to uncertainties about the conversion factor between CO emission and molecular gas mass, our observations suggest that a 2.9 × 10⁷ M_⊙ molecular cloud with a surface density of ∼120 M_⊙ pc⁻² spanning a region 560 pc in diameter is associated with the starburst head of the tadpole galaxy Kiso 5639, where HST observations previously suggested that the star formation rate is 0.04 M_⊙ yr⁻¹ and the star formation surface density is 0.23 M_⊙ pc⁻² Myr⁻¹.

The molecular surface density is comparable to that in giant molecular clouds in local galaxies, which average 100–200 M_⊙ pc⁻² (Heyer & Dame 2015), but the extinction through the molecular cloud in Kiso 5639 should be much less. Considering the local conversion factor between color excess and H i column density (Bohlin et al. 1978), A_V = N/(1.87 × 10²¹ cm⁻²), for a ratio of total to selective extinction R = 3.1, and using a mean molecular weight of 1.36 times the hydrogen mass, the extinction through a cloud is related to its surface density by

$$\begin{eqnarray}&&{{\rm{\Sigma }}}_{\mathrm{gas}}=20.2{A}_{{\rm{V}}}({Z}_{\odot }/Z)\,{M}_{\odot }\,{\mathrm{pc}}^{-2}.\end{eqnarray} \tag{ 1 }$$

For Kiso 5639, Z_⊙/Z = 7.2, so the observed Σ_gas = 120 M_⊙ pc⁻² corresponds to A_V = 0.8 mag. In the solar neighborhood, this is comparable to the threshold for CO formation (Pineda et al. 2008; Glover & Clark 2012). The same extinction threshold was obtained for the WLM galaxy at a metallicity of $12+\mathrm{log}({\rm{O}}/{\rm{H}})=7.8$, where pc-size CO clouds in the core of a giant H i and H₂ envelope had a total shielding column density equivalent to ∼1.5 mag visual extinction (Rubio et al. 2015, see also Schruba et al. 2017).

The assumed value of α_CO = 100 M_⊙ pc⁻² (K km s⁻¹)⁻¹ implies that the ratio of invisible molecular hydrogen to observed CO is high, ∼23 times higher than in the solar neighborhood. A similar situation arises for the low-metallicity galaxy WLM, where the CO clouds resolved by the Atacama Large Millimeter Array (ALMA) are pc-scale inside resolved H i and dust clouds that are ∼200 pc in size. If the peak CO antenna temperature of ∼30 mK in Kiso 5639 corresponds to a beam-diluted thermal temperature of ∼30 K, which is not unreasonable for a molecular region of intense star formation (e.g., Glover & Clark 2012), then the beam-dilution factor of the total CO is ∼10⁻³, and their individual radii would be $\sim 0.03/\sqrt{N}$ of the overall cloud radius, or $\sim 9/\sqrt{N}$ pc for N cores.

Such small CO cores would presumably have a collective motion that is observed as the CO emission Gaussian linewidth of σ = 15 km s⁻¹. Considering again a typical dense cloud temperature of 30 K, these motions would have a Mach number of ∼46 and a compression ratio in the shocked regions of approximately the square of this, which would be ∼2000. The average compressed density is then the compression ratio times the average density. The average density is the surface density divided by the cloud thickness. For a self-gravitating slab, the thickness is $2{\sigma }^{2}/(\pi G{{\rm{\Sigma }}}_{\mathrm{gas}})$, which in our case is 140 pc. Then the average density is ∼7.6 cm⁻³. If this average density is compressed by shocks, then the density in the compressed regions, where the CO might actually be located, is 1.6 × 10⁴ cm⁻³. For a temperature of ∼30 K, the thermal pressure would be 4.8 × 10⁵ k_B. This derivation of pressure is the same as what we would get from the equation for self-gravitational binding pressure, $(\pi /2)G{{\rm{\Sigma }}}_{\mathrm{gas}}^{2}$, using Σ_gas ∼ 120 M_⊙ pc⁻². Any additional contribution to the surface density from atomic gas in an envelope around the molecular cloud would increase the pressure.

The derived pressure inside the Kiso 5639 molecular cloud is comparable to the pressures in local giant molecular clouds, corresponding to similar surface densities. Pressure has been proposed to control the fraction of mass in the form of bound clusters in the star formation process (Kruijssen 2012). Kiso 5639 has a relatively high clustering fraction, 30%–45% (Elmegreen et al. 2016) compared to other dwarf galaxies (Billett et al. 2002), perhaps because of its relatively high pressure. Its clustering fraction is about the same as in spirals (Adamo et al. 2015) where the pressure is the same.

High pressure might also be necessary to form more massive clusters (Elmegreen & Elmegreen 2001). The present observation of molecular clouds close to the extinction threshold for molecule formation (and in Rubio et al. 2015; Schruba et al. 2017) suggest that the surface densities (and therefore pressures) of star-forming clouds will be higher at lower metallicity. That is, low metallicity and the corresponding low dust opacity per column of gas could diminish the ability of self-gravitating clouds to shield themselves against background starlight. This would delay the formation of molecules and cold thermal temperatures until the cloud surface density is high. Then the pressure, which depends only on the surface density, would become high too. The result would be a greater likelihood of gravitationally bound massive clusters in low-metallicity galaxies. Massive concentrations of molecular gas as in Kiso 5639 would be needed too.

We thank R. García-Benito for deriving the abundance of oxygen from the SDSS spectrum. We are grateful to the anonymous referee for comments. M.R. wishes to acknowledge support from CONICYT(CHILE) through FONDECYT grant No. 1140839 and partial support from project BASAL PFB-06. M.R. is a member of UMI-FCA, CNRS/INSU, France (UMI 3386). A.O.G. thanks Fundación La Caixa for financial support in the form of a PhD contract, and S.A. and C.M.T. acknowledge MINECO for funding through the project AYA2016-79724-C4-2-P. This work is based on observations carried out under project number S17AP with the IRAM NOEMA Interferometer. IRAM is supported by INSU/CNRS (France), MPG (Germany) and IGN (Spain). Based on observations made with GTC, in the Spanish Observatorio del Roque de los Muchachos of the IAC, under DDT.