ALMA Deep Field in SSA22:

Intense star-forming activity in a growing galaxy is often obscured by dust in the early Universe. The absorption of stellar emission and reradiation by dust make the galaxy apparently faint at optical-to-near-infrared wavelengths even though they have star-formation rates of $\sim$ 100-1000 $M_{\odot}$ yr^-1 (Blain et al. 2002; Casey et al. 2014 for reviews). Consequently it is challenging to recognize a fraction of star-forming galaxies selected at (sub)millimeter even in a deep near-infrared image. Such a ”near-infrared-dark” population was first identified among bright submillimeter galaxies (SMGs) (e.g., Dey et al. 1999; Smail et al. 1999; Frayer et al. 2004). The faintness makes it difficult to determine spectroscopic redshift $z_{\rm spec}$ using traditional spectroscopic observations at (observed) optical-to-near infrared wavelengths, which has been a major obstacle in studying these populations. As an alternative, bright rest-frame far-infrared (FIR) lines such as [C ii] and CO lines has been utilized to determine $z_{\rm spec}$, which undoubtedly shows that they are heavily dust-obscured star-forming galaxies at high redshift. For instance, Wang et al. (2009) placed a strong upper limit ($K_{s}$=27.3 AB at 3$\sigma$) on GN10 at $z=5.303$ (the redshift has been reported by Pavesi et al. 2018), and Walter et al. (2012) found that HDF850.1, which has no counterpart on an HST image, is located at $z=5.183$.

Following these works in the era when the Atacama Large Millimeter/submillimeter Array (ALMA) had not been available, the advent of interferometers such as ALMA has allowed us to detect and precisely locate fainter and more abundant (sub)millimeter populations. One avenue of research is follow-up observations of SMGs that were originally discovered by a single-dish telescope. For instance, Simpson et al. (2014) found ten SMGs which were detected at 3.6 $\mu$m, but they were blank at shorter wavelengths. While obtaining a spectroscopic redshift of these types of near-infrared-dark SMGs is still not easy, a handful of these galaxies are now placed at $z\sim 4-5$ via molecular and atomic line detections at (sub)millimeter and the number is increasing (e.g., Swinbank et al. 2012; Riechers et al. 2017; Oteo et al. 2018; Casey et al. 2019; Jin et al. 2019; Jiménez-Andrade et al. 2020; Riechers et al. 2020).

ALMA deep surveys, which use many pointings to create a contiguous field, are another avenue of research to measure dust-obscured star-forming activity across the cosmic time (e.g., Tadaki et al. 2015; Umehata et al. 2015; Walter et al. 2016; Dunlop et al. 2017). Some of the galaxies identified by these types of surveys are known to be blank at the H-band (“HST-dark”; e.g., Franco et al. 2018) and Ks-band (e.g., Umehata et al. 2015; Yamaguchi et al. 2016; 2019). The existence of the near-infrared-dark ALMA galaxies has also been uncovered using another approach. Wang et al. (2019) report the identifications of ALMA sources which were originally selected as “H-dropout” galaxies, which are IRAC sources without HST/WFC3 F160W counterparts. They suggest that these types of “ALMA-detected H-dropouts” are the bulk populations of massive ($M_{*}\gtrsim 10^{10.3}$ $M_{\odot}$) star-forming galaxies at $z\approx 3-6$ and that they dominate the total star-formation rate density at such a massive regime in the era (see also Dudzevičiūtė et al. 2020).

These works suggest that some fraction of galaxies could have been missed in previous studies on the basis of the optical-to-near-infrared selection. Charting the nature of the near-infrared-dark ALMA galaxies is necessary for a comprehensive view of galaxy formation and evolution in the early Universe. While the faintness of the optical-to-near-infrared emission makes the spectroscopy in these types of bands quite challenging, the search for molecular and atomic lines at the (sub)millimeter wavelengths can be a solution. In this letter, we report a secure redshift identification of a near-infrared-dark galaxy discovered by an ALMA deep survey. We derive properties of the interstellar medium (ISM) and discuss the role of these types of galaxies in the galaxy formation history. Throughout this letter, we adopt a cosmology with $\Omega_{\rm m}=0.3,\Omega_{\Lambda}=0.7$, and H₀=70 km s^-1 Mpc^-1.

Our target, ADF22.A2, was originally discovered through our ALMA deep survey in the SSA22 field. This field represents a significant proto-cluster at $z\approx 3.1$ (e.g., Steidel et al. 1998; Hayashino et al. 2004; Yamada et al. 2012), and the ALMA mosaic was designed to observe a $6^{\prime}\times 3^{\prime}$ region located at the proto-cluster core at 1.1 mm (ALMA deep field in SSA22 or ADF22; Umehata et al. 2015; 2017; 2018; 2019; Hayatsu et al. 2017). ADF22.A2 is the second brightest source in a $2^{\prime}\times 3^{\prime}$ subregion among ADF22 (ADF22A) with a flux density of $S_{\rm 1.1\leavevmode\nobreak\ mm}=2.02\pm 0.02$ mJy (Umehata et al. 2017). ADF22.A2 was detected at 3.6 $\mu$m and 4.5 $\mu$m, while it was completely invisible even at the Ks-band or shorter wavelengths (Table. 1). In this work, we refer to ALMA-identified galaxies, which have no detectable counterpart at optical-to-near-infrared wavelengths, as “near-infrared-dark” galaxies. ADF22.A2 is then an example of the population. While this practical selection allows us to shed light on the populations which are missed in optical-to-near infrared surveys on the basis of an available data set, it is not straightforward to compare galaxies between different works since there are a wide variety of data sets, including bands and depths as well as the brightness of galaxies. Considering this ambiguity, we compared ADF22.A2 with other near-infrared-dark ALMA galaxies (Wang et al. 2009; Swinbank et al. 2012; Simpson et al. 2014; Franco et al. 2018; Yamaguchi et al. 2019) and a systematical sample of ALMA-identified SMGs (Dudzevičiūtė et al. 2020) in Fig. 1, where the Ks-band magnitude and a flux ratio (Log ($S_{1.2\rm mm}/S_{3.6\mu\rm m}$)) are displayed. ADF22.A2 and the majority of the near-infrared-dark populations are characterized as faint at the $Ks$-band ($\gtrsim 24$ AB) and have a low $S_{\rm 1.2mm}$/$S_{\rm 3.6um}$ flux ratio ($\gtrsim 3$), which naively suggests that they are at a higher redshift and/or have a lower stellar mass among the ALMA-identified galaxies (see also Simpson et al. 2014).

Band 3 observations of ADF22.A2 were performed at two frequency set-ups. The first was carried out as a part of project 2015.1.00212.S and 2016.1.00543.S (PI. H. Umehata for both) at 84.1–87.8 GHz and 96.1–99.8 GHz (Umehata et al. 2019). The 36–44 available antennas were used and observations were executed in two periods; the former one was between 2016 July 21 and 2016 August 19, and the latter one was between 2017 May 7 and 2017 July 4. These resulted in the total on-source time of 13.9 hours for a 13-point mosaic, covering the 7 arcmin² area which is nearly cospatial with the 1.1 mm mosaic in ADF22A. J2226+0052 and J2148+0657 were observed for bandpass and phase calibration. To set the absolute flux scale, Pallas, J2148+0657, J2232+1143 and Titan were observed. Subsequently, ADF22.A2 was observed as a part of project 2017.1.00602.S (PI. N. H. Hayatsu) at 91.4–95.2 GHz and 103.4–107.2 GHz. Observations were carried out with 43-45 usable 12 m antennas between 2018 January 18 and 20. The total on-source time was 263 min. The quasar J2148+0657 was utilized for the bandpass, phase, and flux calibration.

The ALMA data were reduced using the Common Astronomy Software Application (casa) versions 4.7.0, 4.7.2, 5.1.0, and 5.4.0 (McMullin et al. 2007). The data were processed using the standard casa pipeline. The calibrated visibility data were used for imaging with the casa task tclean. We created dirty cubes with a 100 km s^-1 velocity bin, applying uvtaper, which yields the synthesized beam sizes of $\approx 1^{\prime\prime}.3$. The dirty cubes were utilized to identify emission lines. For the identified lines, cleaning was done down to $2\sigma$. As we show, we detected a bright CO(4-3) line, and hence we also repeated the clean process without any tapering for the spectral window which encompasses the line. A band 3 continuum map was also created using line-free channels without tapering, which have the synthesized beam $0^{\prime\prime}.90\times 0^{\prime\prime}.71$ (P.A.$=-74$ deg) and a typical root mean square (RMS) noise of 44 $\mu$Jy beam^-1.

The robust identification of its redshift, $z=3.991$, and the suite of multi-wavelength data sets pave the way to identify an example of near-infrared-dark galaxies in the history of galaxy assembly and understand the physical and chemical properties of these types of populations. The left panel of Fig. 5 shows the relation between the stellar mass and SFR for ADF22.A2, compared with two other samples of near-infrared-dark galaxies and the “main-sequence” relation at $z=4$ derived by Schreiber et al. (2017) on the basis of ALMA band 7 observations. The panel suggests that ADF22.A2 is characterized by massive dusty star-forming galaxies, which are located around (or above) the main-sequence at $z\sim 4$. ADF22.A2 is also located at a region where $z>3.5$ SMGs selected by an ALMA Survey of Submillimetre Galaxies in the Extended Chandra Deep Field South (ALESS, da Cunha et al. 2015) and ”H-dropout” populations (Wang et al. 2019), which tend to be fainter at (sub)millimeter than the SMGs originally sampled as a single-dish source overlap, which indicates that ADF22.A2 shares the properties with both populations. This would be reasonable, considering the fact that ADF22.A2 was first identified by an ALMA deep survey (Umehata et al. 2015) and successively identified by SCUBA2 (Geach et al. 2017), while it was missed by the previous AzTEC/ASTE survey (Umehata et al. 2014).

Since ADF22.A2 was found in an ALMA deep survey which covers a significant cosmic volume contiguously, one can estimate the cosmic volume of the near-infrared-dark ALMA galaxies. Wang et al. (2019) estimated the number density of the H-dropout galaxies $\approx 2\times 10^{-5}$ Mpc^-3 on the basis of photometric redshifts. If we consider a similar redshift range ($z\approx 3.4-5.4$) as was adopted in Wang et al. (2019), the cosmic volume covered by the $2^{\prime}\times 3^{\prime}$ region of ADF22A is $\approx 3.5\times 10^{4}$ Mpc³, and the expected number of these types of populations is around unity. Considering the fact that the detection limit is comparable between the works (Umehata et al. 2017; Wang et al. 2019) and that all of the 18 ADF22A sources now have spectroscopic redshifts (Umehata et al. 2019), both estimates are consistent. This gives another clue supporting the prevalence of the near-infrared-dark ALMA galaxies in the early Universe. We also note that there are some caveats, however. The sample in Wang et al. (2019) is for H-dropout galaxies, and the selection criteria are not perfectly equal to ours. Most of the spectroscopic redshifts of the H-dropout galaxies are still lacking; hence, they leave uncertainties as to the cosmic volume.

The detection of the [C i](1–0) and CO(4–3) lines, together with infrared luminosity, gives us a clue as to the nature of the ISM. We derived the gas density and the strength of the far-ultraviolet radiation field on the basis of the PDR model of Kaufman et al. (1999) and Kaufman et al. (2006). While dust continuum and [C i](1–0) emission are generally optically thin, CO emission is optically thick. We increase the observed CO luminosity by a factor of two to incorporate this effect. Fig. 6 shows the comparison between the observed ratios and the theoretical models. The PDR models suggest the atomic gas density of $n=(5.6_{-2.5}^{+4.6})\times 10^{4}$ cm^-3 and the UV radiation field $(1.8_{-0.8}^{+7.8})\times 10^{3}\leavevmode\nobreak\ G_{0}$ for ADF22.A2 (Fig. 6). The right panel of Fig. 5 shows the ratios of $L_{\rm[CI](1-0)}/L_{\rm IR}$ and $L_{\rm[CI](1-0)}/L_{\rm CO(4-3)}$ compared with $z\sim 4$ lensed SMGs (Bothwell et al. 2017; Cañameras et al. 2018;Nesvadba et al. 2019) and $z\sim 3$ SMGs (Alaghband-Zadeh et al. 2013). The consistency of the line ratios suggests that these characteristics, the UV radiation field strength and gas density of the ISMs within ADF22.A2, are similar to those of other dusty starburst galaxies selected from traditional bright SMG surveys. More samples of near-infrared-dark ALMA galaxies, which have measurements of [C i](1–0) and CO(4–3) lines, will pave the way to further characterize ISM properties of the population.

We also placed ADF22.A2 in the context of galaxy evolution across the cosmic time. Recently an increasing number of quiescent galaxies has been identified at $z\sim 3-4$ (e.g., Schreiber et al. 2018; Tanaka et al. 2019; Valentino et al. 2019). These findings invoke the prevalent galaxy populations which experience an active star-forming phase at higher redshifts. The measurements of ADF22.A2 suggest a gas-mass fraction of $\sim 35\%$, which are defined as $f_{\rm gas}=M_{\rm gas}$/($M_{\rm gas}+M_{*}$). Together with a significant amount of stellar mass $M_{*}\approx 10^{11}$ $M_{\odot}$, the gas mass fraction indicates that ADF22.A2 is a relatively evolved system. The gas depletion time of ADF22.A2 is $\tau_{\rm dep}=M_{\rm gas}/SFR\approx 0.1-0.2$ Gyr; hence, ADF22.A2 would evolve into a massive (log($M_{*}/M_{\odot}\approx 10^{11}$)) quiescent system at $z=3.6-3.8$ on the simple assumption that the galaxy simply consumes the estimated molecular gas reservoir with the constant star-formation rate. These results support the scenario that the near-infrared-dark ALMA populations are progenitors of quiescent galaxy populations at $z\approx 3-4$.

ADF22.A2 demonstrates that ALMA deep surveys are a powerful tool to obtain an unbiased picture of galaxy formation and evolution, exemplifying an intriguing galaxy which was identified and characterized by ALMA observations. Further systematic census will help us understand the general nature of the near-infrared-dark ALMA galaxies.