ALMA twenty-six arcmin$^2$ survey of GOODS-S at one-millimeter (ASAGAO): Near-infrared-dark faint ALMA sources

TL;DR

This study uncovers faint, dust-obscured millimeter sources at high redshift in the GOODS-S field using ALMA, revealing a significant contribution to early galaxy formation missed by optical/near-infrared surveys.

Contribution

First deep wide-field ALMA survey detecting near-infrared-dark faint millimeter sources, highlighting their role in early galaxy formation and cosmic star formation history.

Findings

Detected two near-infrared-dark millimeter sources with no optical counterparts.

Estimated their contribution to cosmic star formation rate density at z~3-5.

Uncovered additional candidates, emphasizing the importance of unbiased ALMA surveys.

Abstract

We report detections of two 1.2 mm continuum sources ($S_\mathrm{1.2mm}$ ~ 0.6 mJy) without any counterparts in the deep $H$- and/or $K$-band image (i.e., $K$-band magnitude $\gtrsim$ 26 mag). These near-infrared-dark faint millimeter sources are uncovered by ASAGAO, a deep and wide-field ($\simeq$ 26 arcmin$^2$) Atacama Large Millimeter/submillimeter Array (ALMA) 1.2 mm survey. One has a red IRAC (3.6 and 4.5 $\mu$m) counterpart, and the other has been independently detected at 850 and 870 $\mu$m using SCUBA2 and ALMA Band 7, respectively. Their optical to radio spectral energy distributions indicate that they can lie at $z \gtrsim$ 3-5 and can be in the early phase of massive galaxy formation. Their contribution to the cosmic star formation rate density is estimated to be ~ 1 $\times$ 10$^{-3}$ $M_\odot$ yr$^{-1}$ Mpc$^{-3}$ if they lie somewhere in the redshift range of $z$ ~ 3-5. This value can be consistent with, or greater than that of bright submillimeter galaxies ($S_\mathrm{870\mu m}>$ 4.2 mJy) at $z$ ~ 3-5. We also uncover 3 more candidates near-infrared-dark faint ALMA sources without any counterparts ($S_\mathrm{1.2mm}$ ~ 0.45-0.86 mJy). These results show that an unbiased ALMA survey can reveal the dust-obscured star formation activities, which were missed in previous deep optical/near-infrared surveys.

Tables

Table 1. Table 1: K 𝐾 K -dropout ASAGAO sources

ID	RA	Dec.	$S_{\mathrm{ALMA}}$	S/Npeak	PB coverageaaThe primary beam (PB) coverage values in Table 1 look smaller than typical values ($>$0.5), but this does not mean that these are sources outside nominal FoVs; this is simply caused by the non-uniform PB coverage of the ASAGAO final map (Figure 1, right panel), which was produced by combining three different ALMA programs including the HUDF data (Dunlop et al., 2017), and the GOODS-S ALMA data (Franco et al., 2018) as well as the ASAGAO data. If we exclude the HUDF data, which cause the non-uniformity (the orange dashed region in Figure 1), the PB coverage value of ID17, ID20, ID22, ID24, and ID25 is 0.62, 0.48, 0.79, 0.97, and 0.69, respectively.	$z_{3.6\mu m/1.2\mathrm{mm}}$	$z_{5.0\mathrm{cm}/1.2\mathrm{mm}}$	counterpart?
(ASAGAO)	(deg.)	(deg.)	(mJy)
17	53.206042	$-$27.819166	0.564 $\pm$ 0.090	6.078	0.403	3.93${}_{-0.30}{}^{}{}_{}{}^{+0.43}$	$>$4.14	Y
20	53.120445	$-$27.742093	0.614 $\pm$ 0.109	5.565	0.317	$>$5.52	$>$4.39	Y
22	53.171662	$-$27.817153	0.612 $\pm$ 0.101	5.446	0.483	$>$5.41	$>$4.26	–
24	53.183284	$-$27.755207	0.446 $\pm$ 0.082	5.022	0.572	$>$4.48	$>$3.70	–
25	53.201002	$-$27.789483	0.858 $\pm$ 0.223	5.020	0.438	$>$5.92	$>$4.93	–

Full text

galaxies: evolution — galaxies: high-redshift — galaxies: star formation — submillimeter: galaxies

1 Introduction

The advent of the Atacama Large Millimeter/sub-millimeter Array (ALMA), which offers high sensitivity and angular resolution capabilities, has enabled us to uncover faint (sub-)millimeter populations (observed flux densities, $S_{\mathrm{obs}}$ $\simeq$ 0.1–1 mJy, corresponding to infrared luminosity of $L_{\mathrm{IR}}\lesssim$ 1012 $L_{\odot}$111Rest-frame 8–1000 $\mu$m.). Recently, several blind surveys using ALMA have been performed in the SXDF (e.g., Tadaki et al., 2015; Kohno et al., 2016; Hatsukade et al., 2016; Yamaguchi et al., 2016; Wang et al., 2016) and the GOODS-S field (Aravena et al., 2016; Dunlop et al., 2017; Ueda et al., 2018; Fujimoto et al., 2018; Franco et al., 2018; Hatsukade et al., 2018, Yamaguchi et al. submitted to ApJ) to detect and characterize the faint (sub-)millimeter galaxies (hereafter, faint SMGs). These studies suggest that they are primarily “typical” or“the main-sequence” star-forming galaxies at $z$ = 1–4 (e.g., da Cunha et al., 2015; Aravena et al., 2016; Yamaguchi et al., 2016; Dunlop et al., 2017, Yamaguchi et al. submitted to ApJ), based on the cross-matching of the ALMA-selected sources and optical to near-infrared-selected sources with reliable photometric redshifts and stellar mass estimates.

Here, we focus on the ALMA-selected galaxies which are not well characterized by such a cross-matching technique, i.e., faint SMGs without significant counterpart seen in the optical and near-infrared (near-IR) wavelengths. The existence of optical/near-IR-dark SMGs have already reported by using pre-ALMA interferometers (e.g., Yun et al., 2008; Wang et al., 2009; Tamura et al., 2010). In the ALMA era, Simpson et al. (2014) found that a significant fraction (17 out of 96) of the bright ALMA sources in ECDF-S, which are originally selected by the LABOCA on APEX survey at 870 $\mu$m, are too faint in the optical/near-IR bands to obtain reliable constraints on their photometric redshift, arguing that such “near-IR-dark” SMGs tend to lie at higher redshift than the typical SMGs based on the Herschel stacking. Similarly, ALMA follow-up observations of SCUBA2-selected SMGs in UDS revealed that 4 bright ALMA sources out of 23 does not have significant near-IR counterparts (Simpson et al., 2015). And in fact, such trend extends to the faint SMGs purely selected by ALMA. For instance, Fujimoto et al. (2016) suggest that $\simeq$ 40% of faint ALMA sources ($S_{\mathrm{1.2mm}}$ = 0.02–1 mJy) uncovered in the ALMA archival images of various fields (the total coverage is $\sim$9 arcmin2) have no counterparts at optical/near-IR wavelengths (the 5$\sigma$ limiting magnitude of $\sim$ 27–28 mag at optical wavelengths and $\sim$ 25–26 mag at near-IR wavelengths). Yamaguchi et al. (2016) find that one out of five ALMA sources in the 2 arcmin2 survey of SXDF ($S_{\mathrm{1.1mm}}$ = 0.54–2.0 mJy) are faint at $H$-band ($\simeq$ 25.3 mag) and not detected at wavelengths shorter than $\sim$ 1.3 $\mu$m. All these studies strongly motivate us to conduct a systematic search for near-IR-dark faint SMGs in the fields where the deepest near-IR images to date are available.

In this paper, we report detections of near-IR-dark, faint ALMA sources ($S_{\mathrm{1.2mm}}$ = 0.45–0.86 mJy), which do not have any significant counterparts in the ultra-deep $H$- and/or $K$-band images, based on the ALMA twenty Six Arcmin2 survey of GOODS-S At One-millimeter (ASAGAO; Project ID: 2015.1.00098.S, PI: K. Kohno). This paper is structured as follows. Section 2 presents ALMA observations and ALMA source identification. In section 3, we describe the properties of the near-IR-dark faint ALMA sources detected by ASAGAO. Then, we put constraints on their physical properties such as redshifts and stellar masses in Section 4. Finally, we estimate their contribution to the cosmic star formation rate density (SFRD) in the high redshift universe (Section 5). Throughout this paper, we assume a $\Lambda$ cold dark matter cosmology with $\Omega_{\mathrm{M}}$ = 0.3, $\Omega_{\mathrm{\Lambda}}$ = 0.7, and $H_{0}$ = 70 km s*-1* Mpc*-1*. All magnitude are given according to the AB system. We adopt the Chabrier Initial Mass Function (IMF; Chabrier, 2003) when necessary to compute the SFR in galaxies in this paper.

2 ALMA source catalog and identifications of near-IR-dark faint ALMA candidates

We examined 25 secure ALMA sources with signal-to-noise ratio (S/N) $>$ 5 in the 26 arcmin2 map of the ASAGAO (Hatsukade et al., 2018) to search for near-IR dark faint ALMA sources. Here we adopt peak S/N values, rather than the spatially integrated S/N values, to conduct source extraction. The details of the ALMA observations and the source catalog creation are given in Hatsukade et al. (2018). Here we provide a brief overview. The 26 arcmin2 map of the ASAGAO field was obtained at 1.14 mm and 1.18 mm (two tunings) to cover a wider frequency range, whose central wavelength was 1.16 mm. To obtain the best ALMA image of this field, we also include ALMA archival data toward the same field (Project ID: 2015.1.00543.S, PI: D. Elbaz and Project ID: 2012.1.00173.S, PI: J. S. Dunlop). After adopting a 250-k$\lambda$ taper, which gives an optimal combination of the sensitivity and angular resolution, the final map reached a typical rms noise of 30 $\mu$Jy beam*-1* at the central $\sim$ 4 arcmin2 and $\sim$ 70 $\mu$Jy beam*-1* at the remaining area with the synthesized beam 0*′′.59 $\times$ 0′′*.53 (PA = $-$83∘).

Yamaguchi et al. (submitted to ApJ) report that 20 of 25 sources candidates have been listed in $K$-band selected sources catalog by the FourStar galaxy evolution survey (ZFOURGE; Straatman et al. 2016; the 5$\sigma$ limiting magnitude of $K_{s}$ = 26.0 mag at the 80% completeness levels). The ASAGAO sources are cross-matched against the ZFOURGE catalog, after correcting for a systematic offset with respect to the ALMA image ($-$0*′′.086 in right ascension and $+$0′′.282 in declination), which is calibrated by the positions of stars in the Gaia Data Release 1 catalog Gaia Collaboration et al. (2016). Here, we adopt the search radius $=$ 0′′*.5 for point-like sources, which is comparable with the synthesized beam of the final ALMA map. Considering the number of ZFOURGE sources within the ASAGAO field ($\sim$ 3,000), the likelihood of random coincidence is estimated to be 0.03 (this likelihood is often called the $p$-value; Downes et al., 1986). In the case that a counterpart is largely extended in the $K$-band image, we allow a larger positional offset, up to half-light radius of $K$-band emission. However, we still have 5 candidates without ZFOURGE counterparts with S/N $>$ 5, which are undetected at $K$-band (Figure 1). We summarize the ASAGAO candidates without ZFOURGE counterparts in Table 1 and we show the multi-wavelength postage stamps of these 5 near-IR-dark faint ALMA candidates in Figure 2.

We check the reliability of these near-IR-dark faint ALMA candidates using two independent methods. First, we apply the same source finding algorithm to the negative map in order to estimate the degree of contamination by spurious detections. The semi-analytical model by Casey et al. (2018) suggests that the contamination rate is small in the range of S/N $>$ 5.0. There is no negative detections with S/N $>$ 5.2 to be compared with the 23 positive detections with S/N $>$ 5.2. In the 5.0 $<$ S/N $<$ 5.2 bin, we find one negative detections and two positive detections (i.e., ID24 and ID25; see Table 1). Therefore, the negative fraction in the S/N bin is 0.5 (see also Figure 15 in Hatsukade et al. 2018). Second, we split the ASAGAO visibilities into two polarization components (i.e., XX and YY polarization images) and create two XX and YY images, which are purely independent. With these two images, we find that all 5 candidates are detected with S/N $\sim$ 3–5 in both XX and YY. This is the behavior expected for $>5\sigma$ detections. Based on these tests, we suggest that two highest S/N near-IR-dark ALMA sources seem to be secure, whereas remaining 3 sources may contain false detections. To test the reality of these sources further, we then consult with other deep images available in the next section.

3 $K$-dropout ASAGAO candidates

In this section, we describe ASAGAO candidates without ZFOURGE counterparts (hereafter, $K$-dropout ASAGAO candidates) individually. First, we perform a stacking analysis for each 5 $K$-dropout ASAGAO candidates using optical/near-IR images obtained by the 3D-HST survey (Grogin et al., 2011; Koekemoer et al., 2011; Skelton et al., 2014). This technique is often used to check reliability of extremely high redshift ($z\gtrsim$ 7) Lyman break galaxies (e.g., Bouwens et al., 2013). We use the Advanced Camera for Survey (ACS; Ford et al., 1998)/F435W, F606W, F775W, F850LP, F814W, and the Wide Field Camera 3 (WFC3; Kimble et al., 2008)/F125W, F140W, F160W222These images are available at the 3D-HST website; https://3dhst.research.yale.edu/Data.php. In the stacking analysis, the Point Spread Functions (PSFs) of the HST images are matched to the WFC3/F160W image ($\simeq$ 0*′′*.16). We show the results of the stacking in Figure 3. We find no significant detections even in these ACS/WFC3 stacked images. Nevertheless, we find that two of the K-dropout ASAGAO candidates have independent detections in longer wavelengths as follows:

• –

ID17: This object is detected at 3.6 and 4.5 $\mu$m bands of Spitzer/InfraRed Array Camera (IRAC; Fazio et al., 2004) by the Spitzer-Cosmic Assembly Deep Bear infrared Extragalactic Legacy Survey (S-CANDELS; PI G.Fazio; Ashby et al., 2015, see Figure 2). Its apparent magnitudes at 3.6 $\mu$m and 4.5 $\mu$m are 25.38 $\pm$ 0.30 and 25.00 $\pm$ 0.27 mag, respectively (Ashby et al., 2015).

• –

ID20: This object is detected at JCMT/SCUBA2 and ALMA Band 7 (Cowie et al., 2018). The observed flux density is 1.35 $\pm$ 0.24 mJy at 870 $\mu$m (Cowie et al., 2018). This source is recognized as ID68 in Cowie et al. (2018).

Considering the multi-wavelength information, two of the five $K$-dropout ASAGAO candidates with multi-wavelength counterparts (i.e., ID17 and ID20) must be real (secure detections), while we suggest that the rest of three candidates without multi-wavelength counterparts should remain “candidates”, which shall be verified by further follow-up observations.

4 Physical properties

These extremely red colors can be reproduced by the high-redshift sources or highly-reddened low-redshift sources (e.g., Caputi et al., 2012). We plot optical to radio Spectral Energy Distributions (SEDs) of these $K$-dropout ASAGAO sources (including 3 candidates) in Figure 4. As a comparison, we also show the average SED of ALESS333The ALMA follow-up observation of the LABOCA Extended Chandra Deep Field South Survey (e.g., Hodge et al., 2013; da Cunha et al., 2015; Swinbank et al., 2014) sources with visual extinction ($A_{\mathrm{V}}$) $>$ 3.0 (the reddest case; hereafter we call this SED as the average SED of ALESS SMGs) obtained by da Cunha et al. (2015). As shown in Figure 4, all sources can lie at $z\gtrsim$ 3–5, even though we assume the highly-reddened SED. The relation between the flux ratio between 3.6 $\mu$m and 1.2 mm ($S_{\mathrm{3.6\mu m}}/S_{\mathrm{1.2mm}}$; the left panel of Figure 5) also prefer high redshift cases. For ID17 which is detected at 3.6 $\mu$m, the ratio indicates that it can lie at $z$ = 3.93${}^{+0.43}_{-0.30}$, when we assume the average SED of ALESS sources. The redshift error is attributed to the error in the $S_{\mathrm{3.6\mu m}}/S_{\mathrm{1.2mm}}$ ratio. On the other hand, as shown in the left panel of Figure 5, variation between SEDs are quite large and some degeneracy between the reddened-color and redshift is still unresolved at 3.6 $\mu$m.

They are not detected by the Kerl G. Jansky Very Large Array (JVLA) C band (5 cm) deep observation ($\sigma\simeq$ 0.35 $\mu$Jy beam*-1*; Rujopakarn et al. 2016; Rujopakarn et al. in preparation). As suggested by Carilli & Yun (1999), the flux ratio between radio and (sub-)millimeter wavelengths can be a redshift indicator. In the left panel of Figure 5, we show the redshift dependence of the flux ratio at radio and millimeter wavelengths ($S_{\mathrm{5cm}}/S_{\mathrm{1.2mm}}$). We show the upper limits of the flux ratio of $K$-dropout ASAGAO sources including 3 candidates. As comparisons, we also plot the redshift dependence of $S_{\mathrm{5cm}}/S_{\mathrm{1.2mm}}$ of IR bright sources. The result suggests that their flux ratio are roughly consistent with the estimated redshifts (i.e., $z\gtrsim$ 3–5) when we assume the average SED of ALESS sources. In Table 1, we show the estimated lower limits of photometric redshifts in this case.

In the high-redshift case (i.e., $z\sim$ 4), the 3$\sigma$ upper limits of stellar masses of $K$-dropout ASAGAO sources are estimated to be $\log(M_{*}/M_{\odot})$ $\lesssim$ 10.4 using Spitzer/IRAC 8.0 $\mu$m, i.e., rest-frame $H$-band data (3$\sigma$ limiting magnitude of 24.3 mag; Dickinson et al., 2003) if they lie at $z\sim$ 4. Here, we assume a mass-to-light ratio obtained in the rest-frame $H$-band luminosity (e.g., Hainline et al., 2011). Hainline et al. (2011) estimated the mass-to-light ratio of dusty sources $M_{*}/L_{H}$ = 0.17 and 0.13 $M_{\odot}\ L_{\odot}^{-1}$ for constant and single-burst star formation histories, respectively. In this paper, we adopt the average value (i.e., $M_{*}/L_{H}$ = 0.15 $M_{\odot}\ L_{\odot}^{-1}$) of those two extreme case. We also estimate its IR luminosity by integrating the SED presented in Figure 4 and find $\log(L_{\mathrm{IR}}/L_{\odot})\sim$ 12.0 corresponding to $\log(\mathrm{SFR/[M_{\odot}\ yr^{-1}]})\sim$ 2. Here, we assume the average SED of ALESS sources at $z\sim$ 4. When we consider the $M_{*}$–SFR relation at $z\sim$ 4 (e.g., Schreiber et al., 2017), they show starburst-like features. As discussed in Wang et al. (2016), this source can represent the early phase of formation of massive galaxies, which are difficult to be observed using rest-frame ultraviolet (UV) selected galaxies such as Lyman-$\alpha$ emitters or Lyman break galaxies.

In the low-redshift case (i.e., $z\sim$ 2), we can estimate the stellar mass of ID17 to be $\log(M_{*}/M_{\odot})$ $\simeq$ 9.4, because it is detected at Spitzer/IRAC 4.5 $\mu$m data, which delivers the rest-frame $H$-band light at $z\sim$2. According to Straatman et al. (2016), a completeness limit of the ZFOURGE survey is $\log(M_{*}/M_{\odot})$ $\sim$ 9.0 at $z$ = 2, which implies that ID17 prefers the high-redshift case rather than the low-redshift case. For other 4 $K$-dropout ASAGAO sources including 3 candidates, their 3$\sigma$ upper limits of stellar masses are estimated to be $\log(M_{*}/M_{\odot})$ $\lesssim$ 8.8 when we consider the 3$\sigma$ limiting magnitude of S-CANDELS (26.5 mag)444For ID20, it is difficult to use Spitzer/IRAC photometries because of heavy confusions (Figure 2). These upper limits are consistent with their non-detections at $K$-band. Thus, we can not exclude the low-redshift case for these 4 sources. If they lie at $z\sim$ 2, their IR luminosities are estimated to be $\log(L_{\mathrm{IR}}/L_{\odot})\sim$ 11.6 when we assume the SED template of Dale & Helou (2002) with $T_{\mathrm{dust}}$ = 25 K. Therefore, in this case, they seem to be extremely low-mass starburst galaxies, which have been missed in previous deep surveys at optical/near-IR wavelengths.

5 Contribution to the cosmic SFRD

Many previous studies predict that the contribution of dust-obscured star-forming activities to the cosmic SFRD have a peak level at $z\simeq$ 2–3 and decline toward $z\gtrsim$ 3–4 based on, for example, IR luminosity functions obtained by the Herschel (e.g., Burgarella et al., 2013) or dust attenuation-corrected UV observations (e.g., Bouwens et al., 2015). On the other hand, Rowan-Robinson et al. (2016) predict that the contribution seems to be constant at $z$ = 1–5 based on the integrated SFR functions estimated by Hershel/SPIRE-500 $\mu$m sources.

According to Simpson et al. (2014), their optical/near-IR-dark SMGs are located in the redshift range of $z\sim$ 3–5. Thus, in this section, we assume the case that all of $K$-dropout ASAGAO candidates lie somewhere in the redshift interval of $z$ $\sim$ 3–5. When we use the average SED of ALESS sources, their contribution to the cosmic IR SFRD is estimated to be $\rho_{\mathrm{SFR}}\sim$ 1–3 $\times 10^{-3}$ $M_{\odot}\ \mathrm{yr^{-1}\ Mpc^{-3}}$, which corresponding to $\sim$ 10–30% of previous works (e.g., Madau & Dickinson, 2014). Here, we simply sum up the SFRs of $K$-dropout ASAGAO sources and divide them by the co-moving volume. The uncertainty of their contributions to the cosmic IR SFRD in Figure 6 are attributable to the relativity of $K$-dropout ASAGAO candidates. If only 2 secure sources with counterparts (i.e., ID17 and ID20) are real, their contribution is expressed by the solid horizontal dark-red line in Figure 6 ($\rho_{\mathrm{SFR}}\sim$ 1 $\times 10^{-3}$ $M_{\odot}\ \mathrm{yr^{-1}\ Mpc^{-3}}$). On the other hand, in the case that all 5 sources are real, their contribution is shown by the dark-red dashed horizontal line ($\rho_{\mathrm{SFR}}\sim$ 3 $\times 10^{-3}$ $M_{\odot}\ \mathrm{yr^{-1}\ Mpc^{-3}}$).

We also consider uncertainty attributed to different assumed SEDs. If we estimated SFRs of $K$-dropout ASAGAO candidates using SED templates presented in Figure 5, their contributions to the cosmic IR SFRD can vary by $\sim$ $\pm$ 0.3 dex, which dose not affect following our conclusion significantly.

As shown in Figure 6, their contributions to the cosmic SFRD can be comparable with, or greater than that of bright ALESS SMGs ($S_{\mathrm{870\mu m}}$ $>$ 4.2 mJy; Swinbank et al. 2014). Therefore, the non-negligible contribution of dust-obscured star formation activities to the cosmic SFRD at high redshift could have been missed in previous surveys. This result shows the importance of ALMA deep contiguous survey to study the evolution of the cosmic SFRD.

We thank the referee for the comments, which improved the manuscript. This paper makes use of the following ALMA data: ADS/JAO.ALMA#2015.1.00098.S, 2015.1.00543.S, and 2012.1.00173.S. ALMA is a partnership of ESO (representing its member states), NSF (USA), and NINS (Japan) together with NRC (Canada), NSC and ASIAA (Taiwan), and KASI (Republic of Korea) in cooperation with the Republic of Chile. The Joint ALMA Observatory is operated by ESO, AUI/NRAO, and NAOJ. Data analysis was partly carried out on the common-use data analysis computer system at the Astronomy Data Center (ADC) of the National Astronomical Observatory of Japan. Y. Yamaguchi is thankful for the JSPS fellowship. This study was supported by the JSPS Grant-in-Aid for Scientific Research (S) JP17H06130 and the NAOJ ALMA Scientific Research Number 2017-06B.