Using the Atacama Large Millimeter/submillimeter Array, we have made the first high spatial and spectral resolution observations of the molecular gas and dust in the prototypical blue compact dwarf galaxy II Zw 40. The and emission is clumpy and distributed throughout the central star-forming region. Only one of eight molecular clouds has associated star formation. The continuum spectral energy distribution is dominated by free–free and synchrotron; at 870 μm, only 50% of the emission is from dust. We derive a CO-to-H2 conversion factor using several methods, including a new method that uses simple photodissocation models and resolved CO line intensity measurements to derive a relationship that uniquely predicts for a given metallicity. We find that the CO-to-H2 conversion factor is 4–35 times that of the Milky Way (18.1–150.5 ). The star formation efficiency of the molecular gas is at least 10 times higher than that found in normal spiral galaxies, which is likely due to the burst-dominated star formation history of II Zw 40 rather than an intrinsically higher efficiency. The molecular clouds within II Zw 40 resemble those in other strongly interacting systems like the Antennae: overall they have high size–linewidth coefficients and molecular gas surface densities. These properties appear to be due to the high molecular gas surface densities produced in this merging system rather than to increased external pressure. Overall, these results paint a picture of II Zw 40 as a complex, rapidly evolving system whose molecular gas properties are dominated by the large-scale gas shocks from its ongoing merger.
The high star formation rate surface densities and low metallicities found in blue compact dwarf galaxies represent one of the most extreme environments for star formation in the local universe, more akin to that found in high redshift galaxies than in local spirals (Cardamone et al. 2009; Izotov et al. 2011). These global properties result in increased disruption of the interstellar medium by newly formed young massive stars (Krumholz et al. 2006; Murray et al. 2010), higher and harder radiation fields (Madden et al. 2006), and reduced dust content (Rémy-Ruyer et al. 2013), all of which may significantly change how the molecular gas within these galaxies transforms into stars. To date, however, the molecular gas fueling the starbursts within blue compact dwarfs remains poorly understood due to the intrinsically faint emission from the most common molecular gas tracers (CO and dust continuum). By quantifying the properties of molecular gas in blue compact dwarfs, we can determine how the physical conditions in these galaxies influence their molecular gas, and thus the formation of young massive stars, as well as gain insight into star formation in high-redshift galaxies, where detailed observations are difficult.
Previous low-resolution studies of molecular gas in low-metallicity galaxies, including blue compact dwarfs, have shown that these galaxies have extremely high star formation rates compared to their CO luminosity—a key tracer of the bulk molecular gas—and that this ratio increases as the metallicity of a galaxy decreases (Taylor et al. 1998; Schruba et al. 2012). Taken by itself, this trend suggests that low-metallicity galaxies either have increased molecular star formation efficiency (i.e., less molecular gas is necessary to form a given amount of stars) or that CO emission is not as effective a tracer of molecular gas because of the decreased dust shielding and reduced abundance of molecules in low metallicity environments (i.e., less CO emission for a given amount of molecular gas). Distinguishing between these two scenarios, however, requires higher spatial resolution to directly measure the CO-to-H2 conversion factor (and thus determine the total molecular gas mass) and to link the young massive stars within the galaxy to the giant molecular clouds from which they presumably form (although see Krumholz 2012 and Glover & Clark 2012 for arguments that neutral hydrogen may play an increasing role in star-forming clouds at low metallicity).
Resolved giant molecular cloud observations in low-metallicity galaxies have historically been very difficult because of the faint nature of the CO emission in these systems. One of the few resolved studies of giant molecular clouds in a sample of dwarf galaxies showed that the giant molecular clouds in these galaxies have similar sizes, linewidths, and CO-to-H2 conversion factors to more massive spiral galaxies like the Milky Way, M33, and M31, which supports the idea of higher star formation efficiencies (Bolatto et al. 2008). In contrast, estimates of the CO-to-H2 conversion factor from resolved dust observations find systematically higher values for low-metallicity galaxies, suggesting that dwarf galaxies have lower CO luminosities for a given amount of molecular gas (Leroy et al. 2011). These two apparently contradictory sets of observations can be reconciled if the reduced dust shielding for CO in low metallicity environments pushes the CO emission to the densest portion of the molecular cloud, while the H2 remains distributed throughout the molecular cloud because it can self-shield. Therefore, the CO observations only trace the central regions of the molecular clouds, while the infrared observations trace the dust, which is well-mixed with the surrounding envelope of molecular gas (Bolatto et al. 2008; Leroy et al. 2011; Sandstrom et al. 2013).
The sensitivity of the previous generation of millimeter interferometers limited the sample in the most comprehensive study to date (Bolatto et al. 2008) to nearby galaxies ( Mpc) with relatively high metallicities; only one galaxy has a metallicity less than 12+log(O/H) = 8.2. Therefore, it is not surprising that these authors see little variation in the sizes, linewidths, and CO-to-H2 conversion factors of low metallicity galaxies. Intriguingly, they do see hints that the lowest metallicity galaxy included in their sample (the Small Magellanic Cloud (SMC)) may deviate from the fiducial trends seen in normal galaxies, although the deviations are relatively weak. This result suggests that expanding resolved molecular gas studies to lower metallicities and more extreme systems than possible with the previous generation of millimeter interferometers may uncover more variations in molecular cloud properties. Fortunately, today we have access to the Atacama Large Millimeter/submillimeter Array (ALMA), whose excellent sensitivity and resolution allow us to do just this.
The blue compact dwarf galaxy II Zw 40 represents a key test case for understanding how the properties of molecular clouds vary with metallicity and star formation rate surface density: this galaxy bridges the gap between ultra-low metallicity () starburst galaxies like SBS 0335-052 and I Zw 18 and starbursting galaxies with normal solar metallicities. Although II Zw 40 has only a moderate metallicity (12+log(O/H) = 8.09; Guseva et al. 2000), roughly comparable to the SMC, its central star-forming region has an extraordinarily high star formation rate surface density of 520 (Kepley et al. 2014), comparable to that found in more massive starburst galaxies. Crucially for our purposes, this galaxy is also only 10 Mpc away (Tully & Fisher 1988), which is two to five times closer than other comparable galaxies like I Zw 18 and SBS0335-052. At this distance, we can resolve the giant molecular cloud fueling the starburst within II Zw 40 using only moderate angular resolution—05 corresponds to 24 pc linear resolution—allowing us to quantify the resolved properties of its molecular gas. By comparing these properties to those in other starburst galaxies of varying metallicity, we can begin to disentangle the relative effects of metallicity and high star formation rate surface density on the star formation efficiencies and CO luminosities in blue compact dwarf galaxies.
In this paper, we present new, high spatial and spectral resolution ALMA observations of the molecular gas and dust content of II Zw 40. Our goal is to understand how the interplay of intense star formation and low metallicity within this galaxy shapes its molecular gas and dust, and whether dust and molecular gas in this galaxy differs in key ways from that in other metal-rich star-forming galaxies. To do this, we measure the properties of the dust and giant molecular clouds within II Zw 40. We use these measurements to derive the CO-to-H2 conversion factor ()—which underlies most of what we know about star formation beyond the Local Group—in II Zw 40 and compare it to other observational and theoretical estimates for this factor. Then we compare the molecular cloud properties in II Zw 40 to the cloud properties in other systems to see if there are systematic differences in cloud properties with metallicity and/or star formation rate surface density. Finally, we use these observations to place the properties of the molecular gas and star formation within II Zw 40 in the larger context of star formation within galaxies.
This paper presents ALMA Cycle 1 (proposal 2012.1.00984.S; PI: A. Kepley) observations of the CO and continuum emission from the central starburst region of the prototypical blue compact dwarf galaxy II Zw 40 at 3, 1 mm, and 870 μm. We describe the data calibration and imaging below. Tables 1 and 2 summarize the observations and spectral window configurations.
Table 1. Observation Summary
|3 mm||1 mm||870 μm|
|Date||2014 Aug 30||2014 Jun 14||2013 Oct 09|
|Number of Antennas||35||33||19|
|Mean Wind Speed ()||9||7||6|
|Mean Precipitable Water Vapor (mm)||3.38||0.51||0.42|
The 870 μm and 1 mm data were calibrated in CASA (McMullin et al. 2007) using scripts provided by the North American ALMA Science Center (NAASC). First, the water vapor radiometer corrections and system temperatures were calculated and applied to the data, and the flux density of the flux calibrator was set. Then the bandpass solutions were calculated using the bandpass calibrator. Next the amplitudes and phases of the calibration sources were derived. The calibration sources had per-integration phases and per-scan amplitudes derived and applied and were flux-scaled. The selected 870 μm phase calibrator was faint (~0.06 Jy), so we combined all four spectral windows and both polarizations to increase the signal-to-noise of the solutions for that band. Finally, we applied the per-scan phases, per-scan amplitudes, and fluxscale for the phase calibrator to the science target. The 3 mm data were calibrated using the CASA pipeline version Cycle2-R1-B, which follows a similar calibration procedure.
All of the observations used quasars instead of planets to calibrate the flux density scales, which introduces additional uncertainty into the flux density scale. For the 870 μm observations, the flux of the "flux" calibrator (J0510+180) decreased by 30% between the two flux measurements bracketing our observations: its measured flux density was 1.090 ± 0.06 Jy on 2013 September 29, but had dropped to 0.710 ± 0.04 Jy by 2013 October 20 according to the ALMA Quasar Catalogue (Fomalont et al. 2014). Extrapolating between these two values gives a flux density for J0510+180 of 0.75 Jy. We used our bandpass calibrator (J0421-120), which had a more stable flux density value, to check the accuracy of our flux density scale. Adopting the extrapolated flux density for J0510+180 yields a flux density within 5% of its measured flux. For the 1 mm observations, the flux of J0510+180 at 1 mm was measured 15 days after the observations to be 1.5 Jy. We used this measurement to set our flux density scale. Unfortunately, the bandpass and phase calibrators did not have recently measured flux densities, so we could not use them to cross-check our flux density scale. At 3 mm, the flux calibrator J0510+180 was observed within two days of our observations and had a flux of 1.56 Jy. We conservatively assume that the flux density of our observations is accurate to within 10% for the 870 μm data and within 20% for the 3 mm and 1 mm data.
The data were not self-calibrated because the line and continuum emission from the source was too faint. Self-calibration typically requires a signal-to-noise of approximately 10 on an individual baseline for times on the order of a few integrations. For both data sets, the signal-to-noise on timescales of a few integrations was less than 4 for the line and continuum data.
We imaged the data in CASA. Since we are interested in measuring and comparing the CO cloud properties, we imaged the line data sets with robust = 0 to maximize angular resolution while retaining some sensitivity to larger-scale emission. The line images were restricted to the overlapping region of uv-space between the two data sets and smoothed to the same beam size to ensure that we are comparing emission on the same size scales for both lines. We also created a naturally weighted image without any uv restrictions, which will be the most sensitive to emission on large angular scales, to compare to previous single-dish observations by Albrecht et al. (2004). In addition to 12CO, there were several additional isotopologues of CO in this data set (see Table 2). We note that the spectral window intended for missed the line and thus we did not produce a image.
Table 2. Spectral Window Setup
|Central Sky Frequency||Rest Frequency||Bandwidth||Bandwidth||Channel Width|
We also created a set of continuum images from both the ALMA data presented in this paper and the VLA data from Kepley et al. (2014). We optimized for flux recovery and matching uv distributions over resolution because the three different continuum emission components—synchrotron, free–free, and dust—will have different extents. We imposed a inner uv plane limit of to match the largest angular scale and tapered the naturally weighted data to a resolution and restored the data with a common 1'' beam. This procedure provided the most accurate match to the uv coverage and resolution between the relatively heterogenous uv sampling at the different wavelengths, although it is lower resolution than our line data cubes. For the VLA data, the fluxes in these images are within 10% of the fluxes measured in Kepley et al. (2014): −7% for the 4.86 GHz image, −6% for the 8.45 GHz image, and 1.5% for the 22.46 GHz image. Based on the single-dish continuum measurements of Klein et al. (1991), we only recover ~50% of the emission from this galaxy and thus are resolving out continuum emission on larger angular scales than we probe here (124). See the analysis in Section 4.1 of Kepley et al. (2014) for a detailed discussion of this point.
Table 3. Line Image Properties
|Rest Frequency||FOV||u–v Range||Beam||PA||Velocity Resolution||Noise|
|345.79599||18.2||16.6–496||robust = 0||28||2||2.6||0.1|
|230.53800||27.3||16.6–496||robust = 0||28||2||3.74||0.32|
|115.27120||54.7||16.6–496||robust = 0||28||2||5.2||1.8|
|330.58797||19.1||16.6–496||robust = 0||27||2||2.4||0.1|
3.1. Overview of the Data
The blue compact dwarf galaxy II Zw 40 consists of a central starburst region, visible as a compact (~8'' = 390 pc) knot in the optical, embedded within a large neutral hydrogen envelope characterized by extended tidal tails (Figure 1). Given its morphology, the starburst is the likely result of a major merger between two dwarf galaxies (Baldwin et al. 1982; van Zee et al. 1998). The burst hosts at least three young massive clusters brighter than 30 Doradus (Kepley et al. 2014), but little is known about the molecular gas that is presumably fueling this extraordinary episode of star formation.
To quantify the properties of the molecular gas fueling the central starburst within II Zw 40, we used ALMA to observe , , and as well as the continuum at 3, 1 mm, and 870 μm in a single field centered on the central starburst region, indicated by a magenta circle in the rightmost panels of Figure 1. Using high-resolution radio continuum observations, the star formation rate surface density in this region was found to be 520 (Kepley et al. 2014): five orders of magnitude higher that of the solar neighborhood Milky Way value (0.003 ) and the same order of magnitude as star formation rates in more massive starburst systems like LIRGs and ULIRGs (100–1000 Kennicutt 1998). This region also has high H i surface densities () consistent with the threshold for formation in low metallicity galaxies (McKee & Krumholz 2010; Wong et al. 2013) and is located near an H i velocity reversal (see the middle panel on the right-hand side of Figure 1).
Our observations detected and emission from the central region as well as continuum emission at all three wavelengths. We also have a global detection of when we average over the main ridge of CO emission (Figure 2). The emission generally has a higher brightness temperature than the emission, but the observations have better brightness temperature sensitivity (), leading to higher signal-to-noise for the data (Figure 4). The line was not detected. However, given the higher noise in the image (in K), this non-detection is consistent with the ratio between the and emission seen in II Zw 40 in other lower-resolution observations (0.5; Albrecht et al. 2004).
3.1.1. CO Emission in II Zw 40
The observed and emission is clumpy and distributed throughout the central star-forming region of II Zw 40 (Figure 3). This CO distribution can be compared to distribution of young massive clusters traced by the 1.3 cm radio continuum emission from Kepley et al. (2014). These deep, high-resolution observations detect clusters with masses as low as anywhere within the field of view.6 We find that only the western end of the main CO ridge is coincident with 1.3 cm continuum emission indicating recent star formation. The other 70% of the CO emission has no coincident 1.3 cm emission.
The morphology of the molecular gas emission in II Zw 40 strongly resembles that found in the nearby dwarf starburst galaxy NGC 1569 (Taylor et al. 1999), which is at a distance of 3.36 Mpc (Grocholski et al. 2008). In that galaxy, there are several well-separated molecular gas clouds. The molecular gas is not spatially associated with the two older (>7 Myr) super star clusters found in the center of NGC 1569 (Hunter et al. 2000). However, the tip of one cloud is associated with an H ii region. This molecular gas distribution is in contrast to that found in observations with similar resolution of more massive (and metal-rich) starburst galaxies such as M82 (Keto et al. 2005) and the Antennae (Wei et al. 2012). In these galaxies, the molecular gas emission is much more extensive and complex, although in the case of M82 this may be more a result of the high inclination of this galaxy rather than the intrinsic molecular gas distribution. Separation of the molecular gas into individual molecular clouds for both the Antennae and M82 is much more difficult, and individually identified clouds frequently show signs of having multiple components along the line of sight (Keto et al. 2005; Wei et al. 2012). In both galaxies, the star formation appears to be concentrated on the edges of the molecular gas clouds (Keto et al. 2005; Wei et al. 2012).
The absence of 1.3 cm continuum emission from most of the CO clouds is unlikely to be the effect of extinction. Once emitted the 1.3 cm emission is unaffected by dust. We note, however, that dust can still absorb photons before they ionize the surrounding gas and produce free–free emission. This effect is roughly on the order of 30% for the Large Magellanic Cloud (LMC) (Inoue 2001) and decreases as a function of metallicity. It is also unlikely to be star formation below the 1.3 cm detection limit. The least massive CO cloud in our sample (Cloud E in Figure 12) has a mass of , assuming the minimum derived for II Zw 40 (see Section 3.2.2 for details). For a typical star formation efficiency of 3% (Heiderman et al. 2010), this implies a stellar mass of which is at the 1.3 cm detection limit. Higher values for II Zw 40, which are likely, would imply even larger cluster masses. All this evidence suggests that the current star formation within II Zw 40 is associated only with the western edge of the main CO ridge and that the rest of the molecular gas does not appear to be forming stars.
Given GMC-scale (~24 pc) resolution of the and data, we do not expect all the molecular clouds within II Zw 40 to have associated star formation. On large (~kpc) scales, the star formation rate surface density is strongly correlated with the molecular gas surface density (Wong & Blitz 2002; Leroy et al. 2013). This relationship breaks down on smaller GMC-sized (~10–75 pc) scales because the observations are averaging over fewer molecular clouds, accentuating the evolutionary state of individual clouds (Chen et al. 2010; Onodera et al. 2010; Schruba et al. 2010).
Using the classification system of Kawamura et al. (2009), we can classify the molecular clouds within II Zw 40 into three different types. Type I clouds are the youngest molecular clouds and show no sign of current star formation. Type I clouds evolve into Type II clouds, which have H ii regions generated by the ionizing radiation from young massive clusters, but no visible optical clusters. Type III clouds represent the final evolutionary stage in a molecular cloud's life. They have both H ii regions and visible optical clusters and are being destroyed by their embedded star formation. For the clouds in the LMC, Kawamura et al. (2009) estimated lifetimes for three types of clouds of 6, 13, and 7 Myr, assuming a constant star formation rate. This classification system is analogous to the scheme developed by Whitmore et al. (2014) to classify molecular clouds within the Antennae galaxies.
In II Zw 40, cloud C would be classified as a Type III cloud. The remainder of the molecular clouds within II Zw 40 are Type I clouds. Using optical data, Kepley et al. (2014) determined that the optical cluster associated with the western end of the molecular gas ridge is less than 5 Myr old, which is consistent with the lifetime of Type III clouds in the LMC estimated by Kawamura et al. (2009). The optical cluster to the south of the main star-forming region is 9.5 Myr old and shows no sign of associated molecular gas, which is again consistent with the timescale for Type III clouds in the LMC.
II Zw 40 has three times more Type I clouds than one would predict based on the relative numbers of clouds in the LMC (7 instead of 2). Although we are in the realm of small number statistics, this difference is still significant. This preponderance of Type I clouds has several possible causes. First, the lifetimes of the Type I clouds in II Zw 40 could be longer than similar clouds in the LMC. Second, the Type I clouds all formed at the same time but have not yet had time to form stars. Third, these clouds are stochastic, unbound, agglomerations of molecular gas that will disperse in the future. Finally, any Type II and III clouds could have been rapidly destroyed via feedback from the central starburst within II Zw 40. Clouds in the process of destruction could also be fainter than Type I clouds and thus may be below our detection limit. Given the evolutionary history of II Zw 40, we suggest that the latter three possibilities are more likely, although we do not have enough evidence to definitely say which mechanism is at work.
For the entire galaxy, the CO line brightness temperature ratios are consistent with an optically thick blackbody with an excitation temperature of 10 K, i.e., an value of 0.7, value of 0.54, and value of 0.74. From our 870 μm and 1 mm cubes, we find an ratio of 0.7 with an uncertainty of 30% for the entire region; the large uncertainty in this value is due mainly to the overall uncertainty in the flux density calibrations. Adjusting for the relative beam sizes, single-dish observations by Albrecht et al. (2004) and Meier et al. (2001) found values of 0.5 and 0.58 ± 0.16, respectively. Both of these values are within the error of the value we derive from our data. However, we note that the single-dish observations will sample different mean molecular gas conditions from the higher resolution observations presented here, so exact correspondence is not expected. In general, the line ratios are not consistent with optically thin gas in local thermodynamic equilibrium (LTE). In that case, the line ratios would be greater than one, except for extremely low excitation temperatures (~few K). The to ratio averaged over the main CO ridge is 15 ± 12, which is consistent with optically thin and optically thick .
For the above arguments, we have assumed that the gas is in LTE. However, the CO in II Zw 40 does not necessarily have to be in LTE. If the gas does not have sufficiently high density, we could have line ratios that mimic the ratios from optically thick emission (less than unity) because the upper levels are not populated. Given that all other evidence suggests we are observing dense star-forming emission cores, we view this as an unlikely scenario.
The line ratios within II Zw 40 are consistent with the observed line ratios derived from single-dish observations of a larger sample of dwarf starburst galaxies (including II Zw 40) by Meier et al. (2001). That work found a mean, error-weighted value of of 0.60 ± 0.06, which is more similar to that found in LIRGs (0.47; 0.1–0.7 Leech et al. 2010) and in the centers of galaxies (0.2–0.7 and 0.61, respectively Mauersberger et al. 1999; Mao et al. 2010) than in nearby normal star-forming galaxies (0.18 ± 0.02; Wilson et al. 2012).
Although the ratio is relatively constant across the face of the galaxy, there are two outliers: cloud G and the western edge of the main CO ridge. The line brightness temperature ratio for cloud G (0.53) is consistent with a blackbody with a lower excitation temperature of ~6 K, which implies an value of 0.34. The western edge of the main CO ridge has elevated ratios roughly a factor of two, indicating higher CO excitation, which is consistent with presence of star formation in this region (Figure 5).
The velocity field for II Zw 40 does not show strong evidence for rotation except for the main ridge of CO emission (Figure 6). The rest of the CO clouds appear to be relatively well separated in velocity space. Most of the clouds have velocity dispersions of ~2 , except for the two bright emission peaks, which have larger velocity dispersions of ~6 .
The velocity field overlaps, but is not completely coincident with the H i velocity field (Figure 7). In the plane of the sky, the CO emission lies between two H i peaks, which we refer to as H i-I and H i-II. In velocity space, the CO clouds overlap the velocities of both peaks, although the CO main ridge appears to be more closely associated with the neutral hydrogen peak H i-I and the high velocity disperion region with H i-II.
Although II Zw 40 has not had its large-scale kinematics modeled in detail, its H i velocity field bears a strong resemblance to that of the merging galaxy NGC 7525, which has been modeled in detail by Hibbard & Mihos (1995). In particular, the H i velocity reversal within the field of view of the CO observations is consistent with gas from the progenitor galaxies flowing inward (see Figure 6 in Hibbard & Mihos 1995, especially T = 32 and T = 40). It also agrees with the scenario presented in Brinks & Klein (1988) that the infalling gas from the progenitor systems may have led to the formation of molecular gas in and thus triggered the current burst of star formation in II Zw 40. High resolution N-body/SPH simulations of the merging galaxy Antennae have shown that a merger can produce a short-lived, off-center burst of star formation (Karl et al. 2010).
Approximately 30% of the total emission in II Zw 40 is resolved out by our ALMA observations. The integrated intensity for II Zw 40 from single-dish observations with the IRAM 30 m is 0.69 (Albrecht et al. 2004), while the integrated intensity for the ALMA cube smoothed to the same resolution (11'') is 27% lower (0.5 ; Figure 8).7 Although both the single-dish and ALMA profiles have similar central velocities, the ALMA profile is missing emission in the line wings, suggesting that there is diffuse CO emission at non-systemic velocities. This result is consistent with CO observations in galaxies, in particular the Local Group dwarf galaxy IC 10 (Leroy et al. 2006).
3.1.2. Continuum Emission in II Zw 40
Dust is key to our picture of how CO emission changes with metallicity. As mentioned in the introduction, dust shields CO from disassociation, while is able to self-shield. The reduced dust content found in low-metallicity galaxies means less dust is available to shield CO, while the molecular gas itself is unaffected because H2 self-shields. To explore the interplay of dust and CO emission in II Zw 40, we compile a submillimeter-to-centimeter continuum spectral energy distribution (SED) based on the ALMA data presented in this paper and archival centimeter VLA observations from Kepley et al. (2014). This SED contains synchrotron emission, free–free emission, and thermal emission from dust. Matched uv-coverage and resolution observations at multiple wavelengths are required to separate these components and characterize their properties.
Figure 9 shows matched beam and uv-coverage images for II Zw 40 from 6.2 cm through 870 μm. The 6.2, 3.5, and 1.3 cm data were originally presented in Kepley et al. (2014) and have been re-imaged to match the uv-coverage and beam size of the ALMA observations. See Section 2 for details on the imaging process. The 3 mm continuum emission is compact and coincident with the free–free dominated 1.3 cm emission (Kepley et al. 2014). The emission is slightly extended at the highest (1 mm and 870 μm) and lowest wavelengths (6.2 cm), indicating the presence of emission from dust and synchrotron emission, respectively. We note that the morphology of the extended 870 μm emission is similar to that of the emission (overlaid as contours), while the 6.2 cm emission extends further south toward the older star-forming region (SSC-south using the terminology of Kepley et al. 2014). These images suggest that the millimeter and submillimeter continuum emission from this galaxy is dominated by free–free emission and that the dust does not dominate the continuum emission from II Zw 40 even at 870 μm.
To quantify this further, we have extracted the continuum SED in a diameter aperture centered at (5:55:42.63161, 3:23:31.2387). The measured flux densities are given in Table 5. We fit a simple two-power law model to these values:
where S is the flux density at frequency ν compared to the flux density (So) at a fiducial frequency (), pth is the thermal fraction emission, and is the spectral index of the non-thermal emission. The aperture was selected to completely encapsulate the continuum emission at all frequencies. For our fit, we excluded the 1 mm and 870 μm data since these data points contain a contribution from dust and set to 4.86 GHz. The fit is shown in Figure 10. The best fit thermal fraction is 0.34 with a range of 0–0.7 and the best fit non-thermal spectral index is −0.3 with a range of −0.21 to −0.87. The reduced value for the best fit is 0.18. This fit shows that, up to 100 GHz, the emission from II Zw 40 is a combination of free–free and non-thermal emission. Only at frequencies greater than 100 GHz do we see any indications of dust emission, and the contribution at those frequencies is relatively low: even at 870 μm, the dust emission is only 50% of the total. We note that the present fit contains a greater contribution from synchrotron emission than the fit to only the VLA data points shown in Kepley et al. (2014). This difference is due to the different uv-coverage and resolution of the images. Here we have emphasized sensitivity to large-scale flux at the expense of resolution, making the synchrotron emission more pronounced because it is typically on larger scales. We note that the spectral index of the synchrotron emission is consistent between the old and new fits.
To further explore the dust content of II Zw 40, we have added the mid- and far-infrared Herschel photometry and resulting fit from Rémy-Ruyer et al. (2013) to the ALMA and VLA data points to better capture the peak of the dust emission (Figure 11). To obtain an estimate of the dust mass from these points, we attempted to fit a modified blackbody to the dust emission:
where Mdust is the dust mass, is the dust absorption coefficient at the reference wavelength , D is distance to the object, λ is the observed wavelength, β is the dust emissivity index, and is the Planck function. We assume a value for of 4.5 . This value is based on the models of Zubko et al. (2004) and is the same value used in Rémy-Ruyer et al. (2013). All attempts to fit a modified blackbody through both the ALMA and Herschel data points yielded poor fits. The failure of these fits is most likely due to either large-scale dust emission being resolved out by ALMA or to the large difference between the apertures used for the Herschel and ALMA photometry: 132'' versus 9''. Although we could increase the aperture size for the ALMA data, the Herschel aperture is larger than our ALMA field of view (see Table 4).
Table 4. Continuum Image Properties
Table 5. Radio/Submillimeter Continuum Spectrum of II Zw 40
|6.2 cm||4.86||9.4 ± 1.9|
|3.5 cm||8.46||7.5 ± 1.5|
|1.3 cm||22.46||7.1 ± 1.4|
|2.7 mm||112.70||4.7 ± 0.9|
|1.3 mm||233.38||5.9 ± 1.2|
|870.0 μm||344.88||7.7 ± 1.5|
Notes.aFlux densities derived from matched beam and uv-coverage images in a 9'' diameter aperture. See Figure 9 for the location of the aperture. The errors are estimated to be 20% at all wavelengths except for the 870 μm band, based the accuracy of the flux density scales. The error is estimated to be 10% at 870 μm because we were able to cross-check the flux calibrator brightness. See Section 2 for more details.
Given our inability to fit both the ALMA and Herschel points simultaneously, we derive an estimate of the dust mass using only the ALMA data by setting β and the temperature to the values derived by Rémy-Ruyer et al. (2013): 1.71 and 33 K. We obtain a dust mass of , which is a factor of 10 less than the dust mass estimated by Rémy-Ruyer et al. (2013): . We note that fixing the β value only has a minor effect on the derived mass. Changing β from 1.71 to 2.0 changes the mass by less than 10%. In addition, since our observations are on the Raleigh–Jeans tail of the dust distribution, our derived dust mass is only minimally affected by the assumed temperature: the mass-weighted temperature would need to have a factor of two uncertainty to produce a factor of two uncertainty in the mass.
Armed with a dust mass for II Zw 40, we can calculate its gas-to-dust ratio (), which links the gas and dust content within the galaxy. The gas-to-dust ratio is used as an alternative way to trace the molecular gas content of a galaxy (e.g., Israel 1997; Leroy et al. 2011; Sandstrom et al. 2013). Here we use archival VLA neutral hydrogen data (van Zee et al. 1998) and the molecular gas measurements from Section 3.2.2 to calculate the gas-to-dust ratio in II Zw 40 and compare it to theoretical estimates of the gast-to-dust ratio in low metalllicity systems.
The gas-to-dust ratio is defined as
where is the CO-to-H2 conversion factor in , is the CO luminosity in , and MH i and Mdust are the neutral hydrogen and dust masses, respectively, in . Here we are defining in terms of masses, instead of surface densities, which requires us to match the apertures used to measure each quantity. In the same 9'' aperture as we used to measure the continuum SED, we measure an MH i of and a luminosity of , which corresponds to a luminosity of using an value of 0.54. From Section 3.2.2, we have estimates from that range from 18.1 to 150.5 . These values give a range for of 720–3500. We note that, if we are resolving out a significant amount of continuum emission, our will be overestimated (because we will have underestimated the dust content). The lower end of our estimated range range is similar to the Rémy-Ruyer et al. (2014) predicted value (645–660, depending on assumptions about ). It is greater than, but still within an order of magnitude of, the predicted by Leroy et al. (2011) (330) and the range of values given for blue compact dwarfs of similar metallicity by Hunt et al. (2014). We do not see any evidence for the dichotomy seen in Hunt et al. (2014) for the ratios for the very low metallicity galaxies I Zw 18 and SBS0335-052, although the metallicity of II Zw 40 is ten times higher than in both of those systems.
3.2. The Molecular Clouds Fueling the Starburst within II Zw 40
Observations of low-metallicity galaxies with the spatial and spectral resolutions necessary to resolve individual molecular clouds are rare because of the faint nature of the CO emission in these galaxies. Previously, only a dozen or so galaxies had bright enough CO to obtain high-resolution observations with sufficient signal-to-noise, e.g., Table 1 in Bolatto et al. (2008). These galaxies were typically limited either to very nearby galaxies or to those with relatively high metallicities (and thus brighter CO). The sensitivity and resolution of ALMA observations, like those presented here for II Zw 40, allow us to measure for the resolved molecular cloud properties in fainter and more extreme low metallicity systems, and compare them to the properties of molecular clouds in other galaxies.
Originally motivated by observations in the Milky Way (Larson 1981; Solomon et al. 1987), the so-called Larson's laws are commonly used to quantify and compare the giant molecular cloud properties in galaxies. The first relationship states that the linewidth of a molecular cloud scales with its size and is generally referred to as the size–linewidth relationship. The second relationship shows that the CO luminosity of a molecular cloud is correlated with its virial mass. This relationship is typically used to infer the conversion factor between the observed CO luminosity and the molecular gas mass of a cloud, under the assumptions that virial equilibrium holds and that the CO traces the full extent of the molecular gas. The final relationship, which is a consequence of the first two, states that molecular clouds all have approximately the same surface density. These relationships have been shown to be sensitive to the resolution and sensitivity limits of the data and the methods used to measure the cloud properties (e.g., Wong et al. 2011; Hughes et al. 2013). However, they continue to provide the basic framework for understanding giant molecular cloud populations.
In this section, we identify individual molecular clouds within II Zw 40, quantify their properties, and compare these properties to those of molecular clouds in other environments. Our comparison sample was selected to span a wide range of cloud properties and environments. It includes cloud samples from the Milky Way disk (Heyer et al. 2009), the center of the Milky Way (Oka et al. 2001), the nearby low metallicity irregular LMC (Wong et al. 2011), the Local Group dwarf starburst IC 10 (Leroy et al. 2006), the major merger referred to as the Antennae (Whitmore et al. 2014, A. Leroy et al. 2016, in preparation), and the nuclear starburst galaxy NGC 253 (Leroy et al. 2015). To mitigate the effects of varying sensitivity and resolution, we have focused on comparison cloud samples whose original observations have relatively high ( pc) resolution and good surface brightness sensitivity and where the algorithms used to determine the cloud properties were similar to those used here, i.e., variants on the algorithm in Rosolowsky & Leroy (2006).
3.2.1. Measuring the Cloud Properties
We used the cprops algorithm (Rosolowsky & Leroy 2006) to identify individual molecular clouds within II Zw 40 and calculate their properties. This algorithm identifies significant emission within a cube using a dilated mask and associates this emission with individual clouds using a modified watershed algorithm. The data cube was used to identify the clouds because it has the highest signal-to-noise out of all our data sets. To identify significant emission, we used a three-dimensional noise cube and required the emission to be greater than 5σ in two adjacent channels to generate an initial mask. All pixels connected to this initial mask down to a 2σ level were then added to the mask. To decompose the emission into individual clouds, we used a box of 035 by 035 by 6 ; 035 is 17pc at the adopted distance of II Zw 40. We rejected regions with a contrast less than 1.5σ and areas less than the FWHM of the beam. We note that emission that can be associated with more than one cloud is not included in the mask. The measured cloud properties have been extrapolated down to 0 K and corrected for the angular and velocity resolution of the data as described in Rosolowsky & Leroy (2006).
Our results do not depend strongly on the cube used for the decomposition, decomposition algorithm, or decomposition parameters. The cube decompositions for both the and both identify the brightest clouds. The main difference between the two cloud identifications is that the emission spans a larger continuous area at high signal-to-noise; the emission appears to be moderately clumpier because only the brightest peaks remain at high signal-to-noise. An alternative decomposition algorithm—clumpfind (Williams et al. 1994)—produces similar results when applied to our data to those generated by the cprops algorithm. Finally, changing the decomposition parameters only affects how many clouds the main molecular ridge is separated into. Depending on the parameters chosen, the ridge can either be identified as a single large cloud or up to four different clouds. We have tuned these parameters to match the number of clumps that can be distinguished by eye in the position–velocity plot of the emission (Figure 12).
The cloud identifications are shown in Figure 12 overlaid on the data and the cloud properties in Table 6. Since the clouds are largely Gaussian, the final cloud sizes and luminosities (and thus their associated properties like mass, etc.) were taken from the Gaussian fit to the cloud sizes with the beam and channel width deconvolved. For clouds that were unresolved, we derive an upper limit on the radius of the cloud by calculating the diameter of a circle that has the same area as the area at half-maximum covered by the cloud.
Table 6. Properties of Giant Molecular Clouds within II Zw 40
|A||5:55:42.268||3:23:27.45||769.0||3600 ± 400||1100 ± 200||6600 ± 3400||3.08 ± 0.02||<11.4||<1.1||0.9||1.0|
|B||5:55:42.572||3:23:31.57||738.5||3000 ± 300||1900 ± 400||5600 ± 2900||4.19 ± 0.04||<12.8||<2.3||0.5||0.8|
|C||5:55:42.620||3:23:32.13||754.7||18700 ± 2000||18300 ± 3700||34500 ± 17600||5.58 ± 0.07||<7.2||<2.3||1.3||2.5|
|D||5:55:42.654||3:23:32.30||746.1||22900 ± 2400||39500 ± 7900||42200 ± 21500||4.92 ± 0.05||34.6 ± 1.2||4.6 ± 0.2||1.4||2.8|
|E||5:55:42.733||3:23:30.18||742.1||4100 ± 400||2900 ± 600||7500 ± 3800||2.28 ± 0.02||15.9 ± 3.8||0.4 ± 0.1||0.7||1.5|
|F||5:55:42.790||3:23:33.56||730.1||6100 ± 600||4600 ± 900||11300 ± 5800||2.99 ± 0.03||38.8 ± 0.7||1.9 ± 0.1||0.7||1.1|
|G||5:55:42.808||3:23:32.61||763.8||20700 ± 2100||26900 ± 5400||60900 ± 31100||8.79 ± 0.04||32.5 ± 0.5||13.7 ± 0.3||0.9||2.2|
|H||5:55:42.891||3:23:32.03||761.4||4300 ± 400||3000 ± 600||7900 ± 4000||3.13 ± 0.03||24.4 ± 0.9||1.3 ± 0.1||0.7||1.2|
Note. Column (1) Cloud label. Columns (2) and (3) R.A. and decl. of the luminosity weighted center of the cloud in J2000. Column (4) Mean luminosity-weighted velocity of cloud in the LSRK frame. Column (5) Measured luminosity. Column (6) Measured luminosity. Column (7) Estimate of luminosity based on observed luminosity and ratio (see Section 3.2.1 for details). Column (8) Linewidth of cloud derived using data. Column (9) Size of cloud derived using data. (see Section 3.2.1 for details). Column (10) Virial mass of cloud derived using data (see Section 3.2.1 for details). Column (11) and Column (12) Peak and temperature of the cloud.
Monte Carlo simulations were used to determine the errors on the measured properties. We added Gaussian noise with the same properties as the noise cube used to mask the emission to the image cube and calculated the properties of the resulting clumps. After 100 iterations, we calculated the standard deviation of the resulting clump properties and used this value as the error. For flux-related values, we added the errors on the derived flux (10% at 870 μm and 20% at 1 mm; see Section 2) in quadrature with the errors derived from the Monte Carlo simulations.
In the literature, the transition is the most common tracer of molecular gas. To compare our data to the data in the literature, we scale our emission for all clouds, except cloud G, to the expected value using a 10 K blackbody, which is consistent with the line ratios for the entire galaxy (see Section 3.1). Since cloud G has a significantly different value of , we scale its emission to the expected value using a 6K blackbody, which is consistent with the ratio seen in that cloud (again see Section 3.1).
3.2.2. The CO-to-H2 Conversion Factor
Since CO is used as the primary tracer of molecular gas beyond the Local Group, quantifying the relationship between CO emission and the amount of molecular gas in different environments is key for understanding the role of molecular gas in star formation throughout cosmic time. In the Milky Way, this conversion factor is typically 4.3 (Bolatto et al. 2013). However, it is a factor of ~5 lower in starburst galaxies like LIRGs and ULIRGs and several times higher in low-metallicity systems (Bolatto et al. 2013). As detailed in the introduction, the high values in the latter systems are due to reduced dust shielding for the CO (e.g., Bolatto et al. 2008; Leroy et al. 2011). In contrast, the lower values of in starburst galaxies are due to increased CO luminosity due to higher gas temperatures and broader velocity widths (e.g., Narayanan et al. 2012).
Assuming that the clouds in II Zw 40 are in virial equilibrium, we can derive for molecular clouds within II Zw 40 by comparing the virial masses of the clouds with their CO luminosities. For II Zw 40, we derive a value for of 18.1 ± 0.5 by fitting a slope to the resolved giant molecular cloud population in –Mvir space (Figure 13). This value is approximately four times higher than the Milky Way value, but is consistent with values in other dwarf starburst galaxies derived using resolved CO observations (Bolatto et al. 2008). Two exceptions to this trend are Clouds E and C. Cloud E, the faintest cloud with virial mass estimate in our sample, is a more diffuse molecular cloud located south of the main star-forming region. However, given the errors on the associated luminosity, the value for this cloud is broadly consistent with the derived value above. Cloud C is associated with the central starburst. This cloud has the lowest ratio of virial mass to CO luminosity, suggesting that its value may be more like that of a starburst galaxy. However, given that we only have an upper limit on its virial mass, this remains a suggestion rather than a definite statement.
Our estimate is dependent on our assumed value of . If is higher, i.e., the clouds are hotter, then the estimated values will be smaller and the fitted factor higher. If is lower, i.e., the clouds are cooler, then the estimated values will be larger and the fitted factor lower. We note that we would require an value of 0.12, implying an extremely cold cloud temperature (), for the value derived from our data to be equal to the Milky Way value. This value is clearly at odds with the line ratios seen in II Zw 40 (Section 3.1.1). We include the statistical errors on (see Section 3.1.1 for a dicussion) in our luminosities in Figure 13. Based the errors for ranges, it is unlikely that our derived would be off by more than the error in (~30%), although the unresolved points only have upper limits for their virial masses, and thus could drive the factor lower.
In low-metallicity systems, estimates of the CO-to-H2 conversion factor derived using virial masses are systematically lower than the estimates derived using IR observations (e.g., Leroy et al. 2011). This offset is thought to be caused by the reduced dust content of low-metallicity systems: dust is the primary shielding mechanism for CO, but H2 molecules are able to self-shield. At solar metallicity, the H2 and CO emission have similar physical extents. As the metallicity decreases, the amount of dust decreases, which decreases the shielding available for CO, driving it toward the centers of the molecular clouds (Bolatto et al. 2008). The molecular gas not traced by CO is commonly referred to as the "dark" molecular gas (Wolfire et al. 2010).
The systematic differences between CO and IR-based estimates suggests that the estimate for II Zw 40 derived here may underestimate the true value for II Zw 40. Ideally, we would compare the CO-based estimate for II Zw 40 to a measured IR-based value of using far-IR dust maps and neutral hydrogen (Leroy et al. 2011; Sandstrom et al. 2013). However, the spatial resolutions of the best possible neutral hydrogen and far-IR observations are far too low to resolve the central star-forming region of II Zw 40: the maximum feasible resolution of neutral hydrogen observation is , while far-IR observations from Herschel have resolutions ranging from to . Although the 870 μm continuum data presented here could be used to provide an estimate for the dust emission, the extended dust emission is only seen in the heavily tapered, low-resolution image, not in an image with uv-coverage and resolution matched to that of the observations.
We can use the total CO luminosity and derived dust mass along with an estimate of the gas-to-dust ratio () to estimate a value of for the central star-forming region of II Zw 40. Rémy-Ruyer et al. (2013) provide an empirical estimate of the as a function of metallicity. For the metallicity of II Zw 40 (8.09; Guseva et al. 2000), the estimated is between 645–660. Here we adopt a mid-range value of 650. The total estimated luminosity in this region (adopting the typical value of in Section 3.2.1) is , the dust mass for the central star-forming region is (see Section 3.1.2), and the neutral hydrogen mass for the same region is . These values give an estimate of of 14 , which is consistent with the CO-based estimate. However, this value is a lower limit, given that we may be underestimating the dust mass because we are resolving out continuum emission.
Finally, we can use the simple photodissociation model developed by Wolfire et al. (2010) to estimate . In this model, depends strongly on the column density of the gas, which for resolved observations depends on the metallicity and integrated CO line intensity. Thus, with our resolved observations, we can independently derive an estimate for based only on these two readily observed quantities. In the
Our resolved observations provide the rare opportunity to compare our cloud-based estimates of to model predictions of global values in low-metallicity environments. Narayanan et al. (2012) uses the results of Wolfire et al. (2010) to model populations of molecular clouds within galaxies. From their simulations, they derive a relationship between the surface brightness of CO and the metallicity of a galaxy. With our cloud-scale resolution, we are in a unique position to test the veracity of these models at low metallicity. Using their Equation (7), the metallicity of II Zw 40 and the average surface brightness of its clouds from Figure 14 (1.8 ), we derive an estimate of of 17.5, which is close to value we derive from the virial mass estimate. However, we note that the value derived here is in the low CO surface brightness limit where, according to Narayanan et al. (2012), the properties of the molecular clouds depend less on environment. Given the observational evidence for larger values in low-metallicity galaxies, we suggest that the values derived from these models may be underestimates for low-metallicity systems.
The preceding discussion of estimates has ignored any time-variable effects on the value, which may be essential for starbursting systems like II Zw 40. Bursts of star formation like those found in II Zw 40 may increase the dichotomy between the dust-derived and virial-derived estimates by increasing CO photodissociation. In pre-starburst systems, these estimates may be closer together due to the reduced flux of ionizing photons.
3.2.3. The Molecular Star Formation Efficiency
With our estimates for in hand, we can now derive a value for the molecular star formation efficiency within II Zw 40, which characterizes how well this galaxy is able to turn molecular gas into stars.8 For the entire central region, is , which corresponds to an of using the appropriate value for (see Section 3.1.1). To obtain an upper limit on the star formation efficiency, we use our largest estimate for : 150.5 . The star formation rate in the central region of II Zw 40 is 0.34 (Kepley et al. 2014). These values yield a molecular star formation efficiency for II Zw 40 of , which is approximately 10 times higher than the average star formation efficiencies found in nearby spirals (Leroy et al. 2008).
If we use the virial-based estimate, we derive an even higher molecular star formation efficiency: . This value is two orders of magnitude higher than in normal spirals and is similar to the star formation efficiencies seen in starburst galaxies (Daddi et al. 2010). However, by using the virial-derived measurement, we are only including the molecular gas emission from deep inside the molecular cloud rather than averaging the molecular gas over kpc regions as in Leroy et al. (2008) and Daddi et al. (2010). Therefore, the molecular star formation efficiency calculated using a virial-based estimate may be expected to be higher because it includes only the densest molecular gas regions where stars are more likely to form, while the values given in Leroy et al. (2008) and Daddi et al. (2010) average over many such regions and also include diffuse molecular gas.
The molecular star formation efficiency values derived above do not take into account the star formation history of II Zw 40 and the effect of its young massive clusters on their surrounding molecular gas. In more massive galaxies, the star formation rate is roughly constant with time over kpc-sized regions or we would not see the observed level of agreement between different star formation rate tracers (Murphy et al. 2011, 2012; Leroy et al. 2012). A continuous star formation rate ensures a steady conversion of gas into stars, allowing the present star formation in these systems to be linked to the remaining gas supply. However, in the case of interacting systems like II Zw 40, their star formation varies with both position and time and their bursts of star formation dissociate the leftover molecular gas. These time-dependent effects mean that one cannot directly link the present-day star formation with the past molecular gas supply.
We suggest that the high molecular star formation efficiency seen in II Zw 40 is due to its rapidly changing, merger-driven star formation history, not an intrinsically high molecular star formation efficiency. For example, in II Zw 40, the massive cluster associated the molecular gas (SSC-N) is young (5 Myr; Kepley et al. 2014). The cluster immediately to the south (SSC-S) is much older (9.5 Myr) and shows no associated molecular gas. We can also see clear signs that SSC-N is destroying the surrounding molecular gas. As we saw in Section 3.1.1, the ratios are elevated here, indicating hotter gas. Higher-resolution observations of the ionized gas within cloud C show filamentary structures, suggesting that ionizing radiation from SSC-N is destroying its molecular envelope (Kepley et al. 2014). Therefore, understanding the observed molecular star formation efficiencies in galaxies like II Zw 40 requires careful modeling of the star formation histories in these systems and the effects of the young massive clusters on the surrounding interstellar medium.
3.2.4. CO Surface Brightnesses and Mass Surface Densities
As discussed in Section 3.2, there is increasing evidence that the third Larson's relation—molecular clouds have constant mass surface densities—is the result of the sensitivity and resolution limits of earlier surveys (Wong et al. 2011; Hughes et al. 2013). However, there do appear to be systematic offsets in the surface density of clouds between galaxies that have been attributed to differences in pressure (Hughes et al. 2013). In this section, we explore whether the molecular clouds within II Zw 40 have constant surface brightnesses and thus constant mass surface densities and compare these quantities to the surface brightnesses and surface densities in our comparison sample.
Figure 14 compares CO luminosity as function of cloud size for both II Zw 40 and the comparison sample. The left panel includes the measured values for both the Antennae and II Zw 40 while the right panel has the estimated values for II Zw 40 and the measured values from galaxies in the comparison sample. Lines of constant surface brightness are shown as dot--dashed lines and the sensitivity and resolution limit of the II Zw 40 data is shown as a dotted line.
These plots show that, for the most part, the clouds in II Zw 40 have lower CO surface brightnesses than clouds in the Galactic Center, NGC 253, and the Antennae. These surface brightnesses are comparable to clouds found in the LMC and IC 10. Therefore, the globally low CO luminosity in II Zw 40 is due to low overall CO surface brightnesses, not a few high surface brightness clouds with very low filling factors. If II Zw 40 is representative of dwarf starburst galaxies as a whole, then we can expect that they will also have low overall CO surface brightnesses. The exception to the low-surface brightness trend is Cloud C: it has a comparable surface brightness to clouds in the Antennae, the Galactic Center, and NGC 253. This cloud is also the only cloud in II Zw 40 with active star formation.
Figure 14 also shows that the molecular clouds within a particular galaxy tend to have similar CO surface brightness, i.e., the clouds fall along a diagonal line in versus radius plots. For our data, however, any systematic variations in the CO surface brightness could also be due to differences in , not just differences in the mass surface density of clouds, To account for the difference values, we plot the CO-derived cloud mass versus radius for II Zw 40 and the comparison sample as a function of size in Figure 15. This plot shows that the CO-derived mass surface densities of molecular clouds within II Zw 40 are on average higher than those in the LMC or IC 10 and are comparable to the CO-derived mass surface densities of clouds within the Antennae and the Galactic Center. Based on these results, the low CO luminosities seen in II Zw 40 are the result of higher factors rather than low surface density molecular gas, showing that the clouds within II Zw 40 have normal mass surface densities compared to clouds in other galaxies.
Systematic uncertainities with the values of and will impact the location of the II Zw 40 clouds on these plots. Smaller values of , which imply colder molecular gas, will move the clouds to higher values in the right panel of Figure 14. For the mass versus radius plots, the and affect the result in opposite directions: increasing leads to lower values and thus lower mass surface densities, while increasing leads to higher mass surface densities. Given that the values we derive for and are consistent with values derived in similar galaxies (see Sections 3.1.1 and 3.2.2), it is relatively unlikely that we have major systematic errors in our derivations of and .
To explore the trend of mass surface density with metallicity further, we plot the CO surface brightness as a function of metallicity in the left panel of Figure 16. For this plot, we have selected comparison galaxies with CO observations at comparable spatial resolution (~30 pc) with a range of metallicities including the Milky Way (Heyer et al. 2009), the LMC (Wong et al. 2011), IC 10 (Leroy et al. 2006), NGC 1569, NGC 205, and SMC (Bolatto et al. 2008), M31 (A. Schruba et al. 2016, in preparation), M33 (Rosolowsky et al. 2007), the Antennae (Whitmore et al. 2014; A. Leroy et al. 2016, in preparation), and NGC 253 (Leroy et al. 2015). As expected, we find that the CO surface brightness for galaxies with metallicities less than 12+log(O/H) = 8.6 is less than or equal to 10 . However, when the proper values are applied to the data, we find that at least two of the galaxies (II Zw 40 and the LMC) have larger mass surface densities than expected if we applied a Galactic value to a CO surface brightness limit of 10 . This result shows that the low CO surface brightnesses seen in low-metallicity galaxies do not necessarily mean that the molecular gas surface densities in these galaxies are low.
3.2.5. The Size–Linewidth Relationship
The last fundamental molecular cloud relationship we will investigate is the size–linewidth relationship, which is thought to arise from the turbulent cascade of energy within a molecular cloud. The location of clouds within this plot reflects the combined effects of virial equilibrium, turbulence, and external pressure. For viralized clouds, the normalization of this relationship is set by cloud surface density. As we saw in Section 3.2.4, the surface brightnesses of clouds in the Milky Way (Solomon et al. 1987; Heyer et al. 2009) and in high-resolution studies of nearby galaxies (Bolatto et al. 2008; Wong et al. 2011) are similar and thus their clouds fall along the same size–linewidth relationship. Higher surface density clouds like those seen in the nuclear starburst NGC 253 (Leroy et al. 2015) and in the Antennae (Whitmore et al. 2014; A. Leroy et al. 2016, in preparation) have higher normalization factors due to their high gas surface densities. Finally, clouds can be elevated beyond the nominal normalization factor derived from their surface densities by higher external pressures, such as those found in the center of the Milky Way (Oka et al. 2001).
Figure 17 compares the sizes and linewidths of clouds within II Zw 40 to those in our comparison samples. We find that the molecular clouds within II Zw 40 have larger than average linewidths for their size compared to clouds in the Milky Way, suggesting that external pressure and/or high gas surface densities play a significant role in shaping the molecular clouds within this galaxy. We note that IC 10, one of the few other dwarf starbursts with measured molecular cloud properties, displays similar properties to II Zw 40, but that clouds in the LMC generally have smaller linewidths compared to those in II Zw 40. Deep neutral hydrogen observations suggest that IC 10 is either an advanced merger or has an infalling gas cloud (Ashley et al. 2014). The molecular clouds in this galaxy also lie largely on the edges of super-shells produced by stellar winds and supernovae generated by its ongoing burst of star formation (Leroy et al. 2006).
To further distinguish between the different samples, we compare the distribution of the square of the coefficient of the size–linewidth relationship (), which is proportional to the molecular gas surface density if the cloud is in virial equilibrium, in the different samples (Figure 18). This plot makes clear that the II Zw 40 clouds have the largest overlap with the clouds seen in the Antennae, although they do also overlap the upper end of the Milky Way molecular clouds and the lower end of the molecular clouds in NGC 253 and the inner Galaxy. Again, the clouds in II Zw 40 and the nearby dwarf starburst galaxy IC 10 have similar values. This result suggests that the clouds in II Zw 40 have higher surface densities than found in the disk of the Milky Way (if they are in virial equilibrium).
To investigate whether high external pressures may be playing a role in II Zw 40, we compare the size–linewidth coefficient squared to the surface density of the molecular gas determined using CO in Figure 19. Clouds in virial equilibrium will fall along the solid line in this diagram, while clouds requiring high external pressures will fall above the line. The dashed lines indicate the required external pressure to maintain virial equilibrium. However, since we have assumed virial equlibrium to determine (see Section 3.2.2), our clouds will fall along the virial equilibrium line by design.
To show the effects of different assumptions, we include in Figure 19 points using our CO-based measurements (large magenta points) as well as a Milky Way value and the dust-based estimate derived in Section 3.2.2 (small magenta points to the left and right of the virial equilibrium line, respectively). Using a Milky Way value would imply that our clouds require external pressures ranging over two orders of magnitude from to , while a dust-based estimate would imply that all of our clouds are sub-virial and collapsing. Although these effects could arise in strong turbulent flows, we argue that the more likely origin of the large linewidths with II Zw 40 is high surface densities generated by the large scale gas shocks driven by its ongoing merger since our CO-derived factor is consistent with CO-based measurements of in other galaxies with similar metallicity (see Section 3.2.2). Given this, we do not find any strong evidence that the clouds within II Zw 40 require high external pressures to maintain their large linewidths.
The two clouds with the greatest deviations from the size–linewidth relationship—clouds C and G—occupy interesting regions of this diagram. Cloud C lies below the virial equilibrium line even using our adopted value. This cloud appears to have already collapsed and formed multiple young massive clusters. High-resolution radio continuum images find at least three embedded sources with ionizing photon fluxes comparable to 30 Doradus—the closest example of a starburst—associated with cloud C (Kepley et al. 2014). The optical cluster near cloud C has an age of less than 5 Myr, which supports the idea that cloud C recently collapsed. Cloud G, which also has line ratios consistent with colder CO, would require the highest external pressures () to maintain virial equilibrium, suggesting that it could be a progenitor of a super star cluster (Johnson et al. 2015).
A related key parameter for quantifying the clouds within II Zw 40 is the turbulent Mach number (Leroy et al. 2015):
where is the one-dimensional velocity dispersion and cs is the sound speed. We adopt a value of 0.2 for cs, which assumes an isothermal molecular gas with a temperature of 10 K. We note that the gas temperature may be higher closer to the star-forming region. The value used here reflects the average temperature of the molecular clouds throughout the central region. The median Mach number of clouds in II Zw 40 is ~36 with values ranging from 20 to 76. Comparing these values with those for giant molecular clouds in the Milky Way and in NGC 253 (see Table 4 in Leroy et al. 2015), we find that the median Mach number within II Zw 40 is higher than that found in molecular clouds in the Milky Way disk (11), but below that of the clouds in the nuclear starburst galaxy NGC 253 (85). The Mach number within II Zw 40 is consistent with the increased Mach numbers necessary to produce virialized clouds.
Overall, these results suggest that the large-scale gas shocks due to II Zw 40's ongoing merger are driving high molecular gas surface densities in II Zw 40, leading to elevated linewidths. We caution that II Zw 40 is clearly a rapidly evolving system. It is possible that some (or all) of the molecular gas structures we see here are transient, unbound features that will not collapse to form stars. However, our estimates of are consistent with other virial equilibrium based estimates of this value at low metallicity (see Section 3.2.2 for a discussion), which suggests that the II Zw 40 clouds identified here are most likely bound objects.
4. SUMMARY AND CONCLUSIONS
In this paper, we present high-resolution (~24 pc) ALMA observations of and emission and 3 mm, 1 mm, and 870 μm continuum emission in the prototypical blue compact dwarf galaxy II Zw 40. Our goal is to characterize the molecular gas content within this unique, low-metallicity and high star formation rate surface density environment and compare it to the molecular cloud properties seen in other environments to gain insight into the fundamental physical processes driving star formation in galaxies.
Due to the combination of relatively bright line emission and excellent surface brightness sensitivity, our highest signal-to-noise CO detection is in ALMA Band 7. The distribution of the CO emission tracing the molecular clouds and various star formation tracers is consistent with a picture of merger-driven star formation within II Zw 40. The CO emission within this galaxy is clumpy and distributed throughout the central star-forming region of II Zw 40 and its kinematics are similar, but not identical, to that of the galaxy's neutral hydrogen reservoir. Only one of the molecular clouds traced by the CO emission has associated star formation, suggesting that the other clouds, in particular the relatively cold cloud G, will serve as future sites for star formation or are transient molecular features.
The centimeter through submillimeter continuum SED of II Zw 40 is dominated by free–free and synchrotron emission. Dust emission only begins to play a role at 1 mm and is only 50% of the total emission at 870 μm. We derive a total dust mass of and a gas-to-dust ratio of between 770 and 3500. The lower end of this range is consistent with other estimates based on far-IR data. The gas-to-dust ratio derived here may be an overestimate if we have resolved out significant continuum emission.
Using the high spatial resolution provided by ALMA, we have used the data to measure the giant molecular cloud sizes, line widths, and luminosities within II Zw 40 and compared their properties with populations of clouds in other fiducial star-forming environments including the Galactic Center (Oka et al. 2001), the LMC (Wong et al. 2011), the Local Group low-metallicity irregular IC 10 (Leroy et al. 2006), the nuclear starburst NGC 253 (Leroy et al. 2015), and the merging Antennae system (Whitmore et al. 2014, A. Leroy et al. 2016, in preparation).
Using our CO and continuum data, we find that the CO-to-H2 conversion factor in II Zw 40 is at least four times higher than in the Milky Way (18.1 ) and may be as much as 35 times higher (150 ). The lower limit is based on an average virial equilibrium estimate for the resolved clouds in II Zw 40. This value is comparable to CO-based virial equilibrium measurements in other low-metallicity galaxies. The upper limit is based on a new method that uses simple photodissociation models by Wolfire et al. (2010) and the resolved line intensity to uniquely predict . We note that the CO-to-H2 conversion factors based on the numerical models of Narayanan et al. (2012) may underestimate the conversion factor, especially in low-metallicity systems.
We use the derived conversion factor to estimate the molecular star formation efficiency of II Zw 40. We find that even using our largest estimate of the CO-to-H2 conversion factor the molecular star formation efficiency of II Zw 40 is still larger than the typical value by a factor of 10. This high value could be due to truly high molecular star formation efficiencies in these systems or, more likely, time-dependent evolutionary effects due to II Zw 40's ongoing merger. We also find that the low CO surface brightnesses in II Zw 40 and other low-metallicity galaxies from the literature do not necessarily imply that these systems have low molecular gas surface densities. The low CO surface brightness in these galaxies appears to be due to reduced dust shielding for CO, not to intrinsically low surface density molecular gas.
The properties of the clouds within II Zw 40 reveal that the large-scale gas shocks produced by its ongoing merger and subsequent star formation may play a key role in shaping the molecular clouds within this galaxy. On average, the clouds in II Zw 40 lie above the Milky Way size–linewidth relationship and have higher size–linewidth coefficients. Their size–linewidth coefficients, which are proportional to the virial mass surface densities, are comparable to those found the Antennae, which is a late-stage merger, and IC 10, whose neutral hydrogen distribution is consistent with either an advanced merger or an infalling gas cloud and whose molecular gas lies on the edges of large-scale shells due to ongoing star formation. They are lower than those in the nuclear starburst NGC 253 or the Galactic Center. Increased surface densities leading to higher linewidths appear to the origin of the elevated size–linewidth coefficients in II Zw 40, rather than large external pressures. We note that this result does depend on the adopted value.
Using II Zw 40 as a template, we can extend our results to make several inferences about properties of molecular gas in blue compact dwarfs. First, the properties of the molecular gas in II Zw 40 are driven by the large-scale gas shocks induced by its ongoing merger, not by its metallicity. The low metallicity of this galaxy only affects the observability of the most commonly used molecular gas tracer CO. II Zw 40 is not a unique case; many blue compact dwarf galaxies appear to be the result of mergers or interactions (Martinez-Delgado et al. 2012; Lelli et al. 2014; Stierwalt et al. 2015). Indeed, unresolved, single-dish and observations of a sample of eight dwarf starburst galaxies have also found that the line ratios and derived densities and temperatures of the molecular gas in their sample have more in common with that of starburst galaxies than of low-metallicity galaxies (Meier et al. 2001).
Second, as II Zw 40 shows, the star formation histories of blue compact dwarfs are complex and vary rapidly with time, unlike larger spiral galaxies which more or less continuously form stars at a low rate. In other words, blue compact dwarfs are more like firecrackers than the slow-burning bonfires of spiral galaxies. Therefore, comparing present-day gas content to current star formation to derive molecular star formation efficiencies and other related quantities in blue compact dwarfs is misleading at best. The high molecular gas star formation efficiencies seen here in II Zw 40 and claimed for other galaxies (Turner et al. 2015) are likely the result of the massive burst of star formation—which establishes a source as a blue compact dwarf galaxy—using up a majority of the molecular gas in a galaxy. Building up larger samples of molecular gas in dwarf galaxies in different evolutionary stages—now possible with the power of ALMA—is key to understanding the properties of the molecular gas and star formation in these sources.
Based on our results, we suggest two avenues for future studies of the molecular gas in these unique systems. First, CO detection experiments for low-metallicity galaxies like II Zw 40 should focus on the or transitions, which are significantly brighter than the ground-state transition. In particular for ALMA, Band 7 observations of should have the highest signal-to-noise, assuming that other blue compact dwarf galaxies have similar CO excitations.
Second, although the sensitivity and resolution of ALMA allow us to extend detailed CO studies to galaxies like II Zw 40, the use of other molecular gas tracers in these galaxies needs to be explored. One commonly used tracer of molecular gas is dust continuum emission. Although this is a powerful tracer, as we have demonstrated here, the dust emission in these galaxies is faint even at 870 μm and blended with free–free emission. Higher resolution () far-infared and neutral hydrogen observations would be another way to probe the molecular gas content of these galaxies, but require the development of a high resolution far-IR telescope and something like the Square Kilometer Array. Another promising tracer might be C i (Glover & Clark 2016b) or C ii (Glover & Clark 2016a), which may be prevalent throughout molecular cloud envelope, but may also extend beyond the cloud.
The authors would like to thank Dr. Liese van Zee for generously sharing her neutral hydrogen data and the referee for his or her thoughtful comments. AAK thanks the Collective Agency (Portland, OR) for hosting her while a portion of this work was written. AAK would also like to thank John Hibbard, Mark Krumholz, and Jennifer Donovan Meyer for helpful conversations. KEJ gratefully acknowledges support provided in part by NSF through award 1413231. The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. This paper makes use of the following ALMA data: ADS/JAO.ALMA# 2012.1.00984.S. ALMA is a partnership of ESO (representing its member states), NSF (USA) and NINS (Japan), together with NRC (Canada) and NSC and ASIAA (Taiwan), in cooperation with the Republic of Chile. The Joint ALMA Observatory is operated by ESO, AUI/NRAO and NAOJ. This research also made use of APLpy, an open-source plotting package for Python hosted at http://aplpy.github.com.
APPENDIX: DERIVATION OF PREDICTED CO-to-H2 CONVERSION FACTOR AS A FUNCTION OF METALLICITY AND INTEGRATED CO LINE INTENSITY
In the Wolfire et al. (2010) models of the dark molecular gas, the estimated CO-to-H2 conversion factor depends strongly on the average extinction through the giant molecular cloud, which is turn depends on both the metallicity of the system and the column density of the giant molecular cloud. For extragalactic observations that do not resolve molecular clouds, the column density must be assumed and is commonly set to values similar to those found in Milky Way clouds. For resolved observations like those presented in this paper, the observed CO brightness of the cloud adds an additional constraint
where WCO has units of and has units of . Thus, in the resolved case, the estimated CO-to-H2 conversion factor depends both on the metallicity, the CO line intensities, and itself. In this Appendix, we solve for the estimated CO-to-H2 conversion factor as a function of metallicity and CO line intensity.
where is the metallicity of the galaxy relative to solar, is the value of at metallicity , is the value of at solar metallicity ( Bolatto et al. 2013), is the difference in extinction between the CO-dark layer of gas and the CO-emitting layer, and is the mean extinction through the giant molecular cloud.9 This is given by the expression
where is the dust-to-gas ratio relative to the local Galactic value and is the mean hydrogen column density of the cloud in units of . For , the quantity equals one. Substituting Equations (5) and (7) into Equation (6), we obtain
which we rearrange to obtain
We use Equation (9) to calculate how the value of changes with the WCO and . The term has only a weak dependence on metallicity, so we adopt the mean value for from Wolfire et al. (2010): 0.7. The solution to this equation is of the form
where is the Lambert W function, x is , and c is the right-hand size of Equation (9). The value for the ratio is plotted in Figure 20 as a function of integrated line intensity (WCO) and metallicity ().
Note that Figure 2 in Kepley et al. (2014) does not show the entire field of view for these observations.
The size of the single dish beam is 11'', which is comparable to the largest angular scale observed by ALMA in this image.
We use the extragalactic definition for the molecular star formation efficiency: SFE = SFR/.
The original expression in Bolatto et al. (2013) is in terms of XCO. However, since scales linearly with XCO, this equation also gives the ratio of .
- Fomalont E., van Kempen T., Kneissl R. et al 2014 Msngr 155 19
- Israel F. P. 1997 A&A 328 471
- Klein U., Weiland H. and Brinks E. 1991 A&A 246 323
- Mauersberger R., Henkel C., Walsh W. and Schulz A. 1999 A&A 341 256
- McMullin J. P., Waters B., Schiebel D., Young W. and Golap K. 2007 ASP Conf. Ser. 376, Astronomical Data Analysis Software and Systems XVI ed R. A. Shaw, F. Hill and D. J. Bell (San Francisco, CA: ASP) 127
- Taylor C. L., Hüttemeister S., Klein U. and Greve A. 1999 A&A 349 424
- Tully R. B. and Fisher J. R. (ed) 1988 Catalog of Nearby Galaxies (Cambridge: Cambridge Univ. Press)
Physical conditions of the molecular gas in metal-poor galaxies
L. K. Hunt et al 2017 Astronomy & Astrophysics 606 A99
The final data release of ALLSMOG: a survey of CO in typical local low-M∗ star-forming galaxies
C. Cicone et al 2017 Astronomy & Astrophysics 604 A53
ALMA Reveals Potential Localized Dust Enrichment from Massive Star Clusters in II Zw 40
S. Michelle Consiglio et al. 2016 The Astrophysical Journal Letters 833 L6