The Atomic Gas Mass of Green Pea Galaxies

The nature of "Green Pea" galaxies, the low-redshift (z ≲ 0.3) extreme emission-line galaxies identified by the Galaxy Zoo project (Cardamone et al. 2009), has been of much interest over the last decade. Their low metallicity and dust content, strong nebular lines, compact or interacting morphology, and intense star formation activity are all reminiscent of high-z Lyα emitters (e.g., Izotov et al. 2011; Yang et al. 2017; Jiang et al. 2019). Indeed, for Green Peas studied at ultraviolet (UV) wavelengths, the Lyα equivalent width distribution is similar to that of Lyα emitters at z ≳ 2.8 (Yang et al. 2016), while the Lyα and UV continuum sizes are similar to those of Lyα emitters at z ≈ 3 − 6 (Yang et al. 2017). Green Peas show a high [O iii]λ5007/[O ii]λ3727 luminosity ratio, similar to many high-z star-forming galaxies, indicating optically thin ionized regions (e.g., Jaskot & Oey 2013; Nakajima et al. 2020). Perhaps most interesting, and unlike most galaxies in the low-z universe, Green Peas have been found to commonly show leakage of Lyman-continuum radiation, with escape fractions of ≈2.5%–73% (Izotov et al. 2016, 2018a, 2018b). Such Lyman-continuum radiation escaping from star-forming galaxies is expected to have been the prime cause of the reionization of the universe, at z ≳ 6 (e.g., Fan et al. 2006); however, the dependence of the escape fraction on local conditions is still not understood today. Galaxies like the Green Peas that show strong Lyman-continuum leakage are the best low-z analogs of the galaxies that drove cosmological reionization, and offer the exciting possibility of understanding this critical process in the nearby universe.

While detailed optical and UV imaging and spectroscopic studies have characterized the stellar, nebular, and star formation properties of the Green Peas (e.g., Amorín et al. 2010; Izotov et al. 2011, 2018b; Jaskot & Oey 2014; Yang et al. 2016, 2017; Lofthouse et al. 2017; Jiang et al. 2019), little is known about the primary fuel for star formation in these galaxies, the neutral atomic, or molecular gas. As such, the cause of the intense starburst activity in the Green Peas remains unclear. Further, there is a natural tension between requiring cold neutral gas to fuel the starburst activity and having a sufficiently low Hi column density to allow the resonantly scattered Lyα and Lyman continuum to escape. This suggests that the Hi column density distribution in Green Peas may be highly non-uniform, with Hi porosity playing a key role (but see Henry et al. 2015).

At present, only two Green Peas have published searches for Hi 21 cm emission, both yielding upper limits on the Hi mass of the galaxy (Pardy et al. 2014; McKinney et al. 2019). We report here Arecibo Telescope (hereafter, Arecibo) and Green Bank Telescope (GBT) Hi 21 cm spectroscopy of a large sample of Green Peas at z ≈ 0.02 − 0.1, which allow us to measure the atomic gas mass of these galaxies for the first time. ⁶

Jiang et al. (2019) have compiled the most comprehensive Green Pea galaxy sample to date, consisting of approximately 1000 galaxies at 0.01 ≲ z ≲ 0.41, identified from the Sloan Digital Sky Survey (SDSS) spectroscopic Data Release 13. We used the correlation between B-band luminosity and Hi mass (e.g., Dénes et al. 2014) to pre-select Green Peas from the above sample with Hi masses high enough to show detectable Hi 21 cm emission with Arecibo and the GBT in reasonable integration time (few hours). Our targets span a wide range of absolute B-band magnitudes ( − 20.0 ≤ M_B ≤ − 16.1) and gas-phase metallicities (7.6 ≤12+[O/H] ≤8.35). We also carried out two-sample Kolmogorov–Smirnov tests to compare the distributions of metallicity, stellar mass, and absolute B-magnitude in our target sample with those of the parent Green Pea sample of Jiang et al. (2019). We find that the data are consistent with the null hypothesis that the two samples are drawn from the same distribution, in all three parameters.

We used Arecibo and the GBT to carry out a search for Hi 21 cm emission from 44 Green Peas, at z ≈ 0.02–0.1 (proposals GBT/19A-301: PI Malhotra; Arecibo/A3302: PI Rhoads), between 2019 February and August. To use the complementary strengths of Arecibo and the GBT, we observed lower-redshift targets (z ≲ 0.05) with higher expected Hi 21 cm line flux densities over the entire northern and equatorial sky using the GBT. With Arecibo, we broadened the selection to include Green Peas with lower expected Hi 21 cm line flux densities and higher redshifts (z ≲ 0.1), within the region of sky accessible to the telescope.

The Arecibo observations used the L-wide receiver, the WAPP backend, two orthogonal polarizations, and a 25 MHz band sub-divided into 4096 spectral channels and centered on the redshifted Hi 21 cm line frequency. The GBT observations used the L-band receiver with the VEGAS spectrometer as the backend, two polarizations, and a 23.44 MHz bandwidth sub-divided into 8192 channels, and centered on the redshifted Hi 21 cm line frequency. Position-switching, with 5 m On and Off scans, was used to calibrate the system bandpass, while the system temperatures were measured using a blinking noise diode at the GBT, and a separate noise diode, switched on and off for 10 s, at Arecibo. Online Doppler tracking was not used. The total time on each source ranged from 0.75 to 4.5 hr, depending on the galaxy redshift, radio frequency interference (RFI) conditions, and observing exigencies.

All data were analyzed in the IDL package, following standard procedures, with the package gbtidl used for the GBT data. Each On/Off pair was initially calibrated and the final spectrum, for each polarization, shifted into the barycentric frame. Each spectrum was then inspected for the presence of RFI or systematic effects in the spectral baseline; spectra showing non-Gaussian behavior within ≈ ± 200 km s⁻¹ of the expected redshifted Hi 21 cm line frequency were removed from the analysis. For each source, the remaining spectra, from both polarizations, were median-averaged together, with the median used to obtain a more conservative (i.e., less sensitive to outliers) estimate of the average. For four sources, two from each telescope, all spectra were affected by RFI around the expected redshifted line frequency, and the data were essentially unusable.

Hi 21 cm emission was detected from 19 Green Peas at ≥5σ significance (two of which, in J0844+0226 and J1010+1255, have ≈5σ significance and hence should be viewed as tentative detections); the Hi 21 cm spectra of these galaxies are shown in Figure 1. Twenty-one galaxies showed no clear signature of Hi 21 cm emission. Table 1 summarizes the results of the Arecibo and GBT observations; we also include the relevant galaxy properties of each Green Pea, derived from the optical imaging and spectroscopy (e.g., Jiang et al. 2019). The upper limits are computed assuming a Gaussian line profile with a full width at half maximum (FWHM) of 50 km s⁻¹, typical of dwarf galaxies (the Hi mass limits are mostly ≲10⁹ M_⊙, i.e., in the dwarf galaxy range; e.g., Begum et al. 2008). We note that the errors quoted on the Hi 21 cm line flux densities and the Hi masses are statistical errors, and do not include the uncertainty in the flux scale; we estimate this uncertainty to be typically ≈10%.

Zoom In Zoom Out Reset image size

Our Arecibo and GBT Hi 21 cm spectroscopy of Green Pea galaxies has yielded an ≈50% detection rate, with 19 detections of Hi 21 cm emission at redshifts z ≈ 0.023–0.091. These are the first measurements of the atomic gas content of Green Pea galaxies. The Hi masses of the detected galaxies lie in the range ≈ (4–300) × 10⁸ M_⊙, with a median value of 2.6 × 10⁹ M_⊙. For the non-detections, the 3σ upper limits on the Hi mass lie in the range (0.6–32) × 10⁸ M_⊙, with a median upper limit of 5.5 × 10⁸ M_⊙. Note that the large primary beams of Arecibo and the GBT imply that we cannot rule out the possibility that some of the Hi 21 cm emission in the detections may arise from companion galaxies.

Figure 2(A) plots the Hi-to-stellar mass ratio f_Hi ≡ M_Hi /M_⋆ against the stellar mass M_⋆ of the 40 Green Peas of our sample. We used the xGASS sample as the comparison sample, as this is a stellar mass-selected (M_⋆ ≥ 10⁹ M_⊙) sample of nearby galaxies, with Hi 21 cm emission studies (Catinella et al. 2018). The dark green stars indicate the median value of f_Hi (treating the 3σ upper limits to f_Hi as detections) in two stellar mass bins, while the filled blue circles indicate the median values of f_Hi in galaxies in different stellar mass bins in the GALEX Arecibo SDSS Survey (xGASS) sample (Catinella et al. 2018), with the dashed blue line connecting the xGASS values. It is clear that the median value of f_Hi for Green Peas in the higher M_⋆ bin is in excellent agreement with the median value for xGASS galaxies at the same M_⋆, while the median f_Hi in the lower M_⋆ bin appears to lie close to the extrapolated xGASS relation (Catinella et al. 2018). It thus appears that the Hi content of Green Pea galaxies, relative to their stellar mass, is in excellent agreement with that of "normal" galaxies in the nearby universe.

Zoom In Zoom Out Reset image size

The atomic gas depletion timescale τ_dep ≡ M_Hi /SFR gives the timescale for which a galaxy can continue to form stars without replenishment of its Hi reservoir. Lower values of τ_dep would imply that a galaxy's star formation activity would be regulated by the availability of Hi; for example, Chowdhury et al. (2020) argued that the cause of the decline of the star formation activity in the universe at z < 1 is because the Hi reservoirs in star-forming galaxies are not sufficient to support their star formation activity for more than ≈ 1–2 Gyr. However, at z ≲ 0.35, the Hi depletion timescale has been found to be relatively long in main-sequence galaxies, ≈5–10 Gyr (e.g., Saintonge et al. 2017; Bera et al. 2019). Figure 2(B) plots the τ_dep values of our Green Pea galaxies against stellar mass; the dashed green line shows the median value of the sample, τ_{dep, med} ≈ 0.58 Gyr (conservatively treating the upper limits on M_Hi as detections). For comparison, the median value of τ_dep in the xGASS sample (again treating upper limits to M_Hi as detections), shown by the dashed blue line in the figure, is ≈6 Gyr (Saintonge et al. 2017; Catinella et al. 2018), larger by an order of magnitude. It appears that the starburst activity in the Green Peas will exhaust their atomic fuel on very short timescales, far shorter than in most other galaxies in the nearby universe.

The depletion time of star-forming material could be longer than the Hi depletion timescale when the H₂ depletion timescale is taken into account. However, in star-forming galaxies at all redshifts, the H₂ depletion timescale is typically ≲1 Gyr (e.g., Saintonge et al. 2017; Tacconi et al. 2020), far shorter than the Hi depletion timescale. As such, star formation in such galaxies is not limited by the depletion of Hi, as there is a long timescale on which the Hi can be replenished from the circumgalactic medium. However, the very short Hi depletion time in Green Peas implies that Hi depletion could itself act as a bottleneck for star formation (as has been seen in main-sequence galaxies at z ≈ 1; Chowdhury et al. 2020).

Figure 3(A) plots the Hi mass of the Green Peas against their absolute B-magnitude M_B ; the dashed line indicates the M_Hi –M_B relation of galaxies in the local universe, with the dotted lines indicating the ±0.6 dex (≈ 2σ) spread around the local relation (e.g., Dénes et al. 2014). While the majority of the Green Peas are seen to lie within the spread of the local M_Hi –M_B relation, it is interesting that nine of the 40 galaxies of our sample (i.e., ≈ 22%) lie ≳ 0.6 dex above it. This suggests that a significant fraction of Green Peas are gas-rich for their optical luminosity, possibly due to recent gas accretion from the circumgalactic medium or via a minor merger, or due to a gas-rich companion galaxy within the relatively large GBT or Arecibo beam.

Zoom In Zoom Out Reset image size

Conversely, five of the non-detections and two of the detections of Hi 21 cm emission lie ≳0.6 dex below the local M_Hi –M_B relation. Further, most of the detections of Hi 21 cm emission lie above the local relation, while most of the non-detections lie below it. This may suggest bimodality in the Hi properties of Green Pea galaxies, with one group having exhausted its neutral gas in the starburst (which may have been itself triggered by a recent gas acquisition via infall or a merger), and the other having only consumed a fraction of its neutral gas in the starburst. We note that a caveat to the above result is that the M_Hi –M_B relation of Dénes et al. (2014) is based on an Hi-selected galaxy sample from the all-sky HI Parkes All Sky Survey (HIPASS; Zwaan et al. 2005), and thus may be biased toward Hi-rich galaxies. As such, objects lying below the M_Hi –M_B relation of Dénes et al. (2014) may not necessarily be Hi-poor galaxies.

Despite the above caveat, it is tempting to identify the first group of galaxies above with the objects that are likely to show leakage of Lyα and Lyman-continuum radiation (i.e., to show Lyα emission). Eight of the Green Peas of our sample have Lyα spectroscopy, with seven detections of Lyα emission and one (J1448-0110) showing net Lyα absorption (McKinney et al. 2019). Interestingly, five of the detections of Lyα emission are not detected in Hi 21 cm emission, as expected from the above argument. However, two of the Lyα-emitting galaxies, J0213+0056 and J1200+2719, do show detections of Hi 21 cm emission, and with relatively high Hi masses, ≈ 3.2 × 10⁹ M_⊙ (J0213+0056) and ≈ 1 × 10¹⁰ M_⊙ (J1200+2719). Further, both these galaxies are "gas-rich" systems in Figure 3(A). Hi 21 cm mapping studies are needed to test whether the Green Pea galaxy itself is Hi-rich, or if it might have a gas-rich companion. Such Hi 21 cm mapping studies are also critical to directly determine the Hi column density distribution within the Green Peas, to test for the presence of Hi holes through which the Lyα and Lyman-continuum photons might escape. At any event, at the present time, no clear trend is apparent between the gas richness of the above eight Green Peas and their Lyα escape fraction, with high Lyα escape fractions obtained at both high and low Hi masses (and gas richness) in the relatively small current sample (McKinney et al. 2019). Deeper Hi 21 cm emission studies would be needed to test the possibility of bimodality in the gas content of Green Pea galaxies.

We also examined the dependence of the Hi mass, Hi-to-stellar mass ratio, and Hi depletion time, on the metallicity (12+[O/H]) of the Green Peas of our sample, finding no evidence of a dependence of any of these properties on the metallicity.

Jaskot & Oey (2013) argued that the high luminosity ratio O32 ≡ [O iii]λ5007/[O ii λ3727 observed in a number of Green Pea galaxies at z ≈ 0.1–0.3 makes them excellent candidates for the escape of ionizing Lyman-continuum radiation. A high Lyman-continuum leakage was indeed later found in galaxies with high O32 values, both at high redshifts (e.g., Nakajima & Ouchi 2014; Nakajima et al. 2016; Fletcher et al. 2019) and low redshifts (including Green Pea galaxies; e.g., Izotov et al. 2016, 2018a, 2018b, 2020), for typical O32 values ≳ 10. One would expect easier leakage of Lyman-continuum photons from galaxies with a lower average Hi column density, and also with a lower Hi mass. We hence examined the Hi properties in our Green Peas as a function of their O32 value; Figure 3[B] shows the measured Hi mass for the 40 Green Peas plotted against the O32 values; the median O32 value is ≈5.5, indicated by the dashed vertical line. Among the 20 Green Peas with O32 values below the median, there are 12 detections of Hi 21 cm emission, with an average Hi mass of 5.6 × 10⁹ M_⊙, while for Green Peas with O32 values above the median there are seven detections and an average Hi mass of 3.2 × 10⁹ M_⊙. Further, there is only a single detection of Hi 21 cm emission in the 11 Green Peas with O32 ≥10 (i.e., a detection fraction of ${0.091}_{-0.02}^{+0.21}$), and 18 detections in the 29 Green Peas with O32 < 10 (i.e., a detection fraction of ${0.62}_{-0.14}^{+0.18}$). Thus, although the numbers are still small, both the detection rate and the average Hi mass appear to be significantly lower in galaxies with O32 ≳ 10, consistent with the expected high Lyman-continuum leakage.

Tilvi et al. (2009) modeled star formation in Lyα emitters by assuming that the accretion of gas rapidly results in star formation, to obtain a star formation efficiency (f_⋆) of ≈ 2.5%. This is similar to the estimate of f_⋆ ≈4 − 8% obtained by Baldry et al. (2008), by comparing the cosmic stellar mass density to the cosmic baryon density (see also Fukugita et al. 1998). Assuming f_⋆ ≈ 2.5% yields a median star formation timescale of ${\tau }_{\mathrm{SF}}\,\approx {f}_{\star }\times {\left({M}_{{\rm{H}}{\rm\small{I}}}/\mathrm{SFR}\right)}_{\mathrm{med}}\equiv {f}_{\star }\times {\tau }_{\mathrm{dep},\mathrm{med}}\,\approx 15\,\mathrm{Myr}$. Interestingly, this is similar to the age of the young stellar population that dominates the starlight of the Green Peas of our sample (≈3–8 Myr, with a median age of ≈4 Myr; Jiang et al. 2019). We note, however, that the above f_⋆ estimates (Baldry et al. 2008; Tilvi et al. 2009) are for all baryonic material, including the ionized gas. The agreement between the star formation timescale and the age of the young stellar population in Green Peas might then suggest that the timescale of conversion from ionized gas to neutral gas is short in these galaxies.

We report an Arecibo and GBT search for Hi 21 cm emission from a large sample of Green Pea galaxies at z ≈ 0.02 − 0.1, obtaining 19 detections of Hi 21 cm emission and 21 upper limits to the Hi mass, and yielding the first estimates of the gas content of these starbursting systems. The Hi properties of the majority of the Green Peas appear similar to those of galaxies in the local universe, in terms of the Hi-to-stellar mass ratio and the M_Hi –M_B relations. However, a significant fraction of the Green Peas (≈22%) have an Hi mass that is ≳ + 0.6 dex (i.e., ≳ 2σ) above the local M_Hi –M_B relation, indicating either recent gas accretion or a gas-rich companion galaxy. A similar fraction lie ≳ 0.6 dex below the local relation, suggesting possible bimodality in the gas properties of Green Peas. This large fraction of outliers (≈30%) from the M_Hi − M_B relation and the young ages of the stellar populations are indicative of a possible "boom and bust" nature of star formation in Green Peas. Further, the Hi depletion times in Green Peas are an order of magnitude lower than values in local galaxies, indicating that the starburst activity will consume their Hi on timescales less than a Gyr. The detection rate of Hi 21 cm emission appears low in galaxies with the highest O32 values, O32 ≥ 10, consistent with the high Lyman-continuum leakage expected from these galaxies.

This Letter is dedicated to the Arecibo Observatory and its people.

De estas calles que ahondan el poniente, Una habrá (no sé cual) que he recorrido, Ya por última vez, ... ⁷

N.K. acknowledges support from the Department of Science and Technology via a Swarnajayanti Fellowship (DST/SJF/PSA-01/2012-13). This work was supported by the Department of Atomic Energy, under project 12-R&D-TFR-5.02-0700. The Arecibo Observatory is a facility of the National Science Foundation operated under cooperative agreement by the University of Central Florida and in alliance with Universidad Ana G. Mendez, and Yang Enterprises, Inc. The Green Bank Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc.