ALMA Deep Field in SSA22: Blindly Detected CO Emitters and [CII] Emitter Candidates

TL;DR

This study uses ALMA to detect and analyze millimeter line emitters in the SSA22 field, identifying high-redshift [CII] candidates and providing insights into early universe star formation.

Contribution

First detection of multiple millimeter line emitters in SSA22, including high-redshift [CII] candidates, with implications for star formation history at z~6.

Findings

Detected four line emitters with >6 sigma significance.

Identified two likely high-redshift [CII] emitters at z=6.0 and 6.5.

Estimated star formation rates of 10-20 solar masses per year.

Abstract

We report the identification of four millimeter line emitting galaxies with the Atacama Large Milli/submillimeter Array (ALMA) in SSA22 Field (ADF22). We analyze the ALMA 1.1 mm survey data, with an effective survey area of 5 arcmin$^2$, a frequency range of 253.1--256.8 and 269.1--272.8 GHz, angular resolution of 0".7 and RMS noise of 0.8 mJy beam$^{-1}$ at 36 km s$^{-1}$ velocity resolution. We detect four line emitter candidates with significance levels above $6 \sigma$. We identify one of the four sources as a CO(9-8) emitter at $z = 3.1$ in a member of the proto-cluster known in this field. Another line emitter with an optical counterpart is likely a CO(4-3) emitter at $z = 0.7$. The other two sources without any millimeter continuum or optical/near-infrared counterpart are likely to be [CII] emitter candidates at $z = 6.0$ and $6.5$. The equivalent widths of the [CII] candidates are consistent with those of confirmed high-redshift [CII] emitters and candidates, and are a factor of 10 times larger than that of the CO(9-8) emitter detected in this search. The [CII] luminosity of the candidates are $4-7 \times 10^8~\rm L_\odot$. The star formation rates (SFRs) of these sources are estimated to be $10-20~\rm M_{\odot}~yr^{-1}$ if we adopt an empirical [CII] luminosity - SFR relation. One of them has a relatively low-S/N ratio, but shows features characteristic of emission lines. Assuming that at least one of the two candidates is a [CII] emitter, we derive a lower limit of [CII]-based star formation rate density (SFRD) at $z~\sim~6$. The resulting value of $> 10^{-2}$ $\rm M_\odot yr^{-1} Mpc^{-3}$ is consistent with the dust-uncorrected UV-based SFRD. Future millimeter/submillimeter surveys can be used to detect a number of high redshift line emitters, with which to study the star formation history in the early Universe.

Tables

Table 1. Table 1: Properties of datacube of Four Spectral Windows.

SPW	${\nu }_{\mathrm{obs}}\text{<sup>(1)</sup>}$	$z_{\left[\mathrm{CII}\right]}$(2)	d$v$(3)	angular resolution(4)	RMS of original data(5)	# of net	# of matched	max positive	max negative
ID	[GHz]		[km s-1]	[arcmin]	[mJy/beam]	clumps(6)	clumps	S/N ratio	S/N ratio
0	253.12 – 254.83	6.458 – 6.508	18.3	0.67, 0.53, 1.09	0.7, 0.9, 1.2,	25 /18	9 /10	7.77$\sigma$ (10(7))	5.70$\sigma$ (2)
1	255.14 – 256.83	6.400 – 6.449	18.1	0.68, 0.54, 1.11	0.7, 0.8, 1.1,	6 /4	5 /3	5.73$\sigma$ (6)	5.81$\sigma$ (0)
2	269.14 – 270.84	6.017 – 6.062	17.2	0.62, 0.49, 1.02	0.8, 1.0, 1.4,	18/14	7 /6	6.51$\sigma$ (21)	6.05$\sigma$ (6)
3	271.14 – 272.84	5.966 – 6.009	17.1	0.62, 0.49, 1.01	0.9, 1.2, 1.6,	10/21	5 /10	5.99$\sigma$ (21)	6.30$\sigma$ (8)

Table 2. Table 2: Measured and derived candidate properties.

ADF22 ID	Name	${\nu }_{\mathrm{peak}}$	Speak/N(1)	Smoothing(2)	$S_{1.1\mathrm{mm}}$(3)	$S_{\mathrm{line}}$(4)	FWHM(5)	EWobs	$z$(6)	log${}_{10}{}^{}L_{\mathrm{line}}^{}$(7)	log${}_{10}{}^{}L_{\mathrm{FIR}}^{}$(8)
	(J2000)	[GHz]	[$\sigma$]	[km s-1]	[mJy]	[Jy km s-1]	[km s-1]	[$\mu m$]		[$L_{\odot }$]	[$L_{\odot }$]
ADF22-LineA	ALMAJ221737.43+001710.7	253.79	6.5	220 (12)	0.2	0.4$\pm 0.1$	220$\pm 40$	$>$ 8.6	6.489	8.6${}_{}{}^{+0.2}{}_{-0.2}{}^{}$	11.4
ADF22-LineB	ALMAJ221731.95+001820.3	269.92	6.2	258 (15)	0.2	0.7$\pm 0.1$	220$\pm 45$	$>$ 14.6	6.041	8.8 ${}_{}{}^{+0.2}{}_{-0.3}{}^{}$	11.4
ADF22-LineC	ALMAJ221736.97+001820.8	253.49	7.8	220 (12)	2.0$\pm 0.1$	0.5$\pm 0.1$	140$\pm 20$	1.1 ${}_{-0.2}{}^{}{}_{}{}^{0.2}$	3.091	8.0 ${}_{}{}^{+0.2}{}_{-0.2}{}^{}$	12.6 ${}_{-0.1}{}^{}{}_{}{}^{+0.2}$
ADF22-LineD	ALMAJ221733.07+001718.8	269.70	6.5	361 (21)	0.7$\pm 0.2$	1.0$\pm 0.1$	240$\pm 40$	7.3${}_{-2}{}^{}{}_{}{}^{+5}$	0.709	6.7 ${}_{}{}^{+0.2}{}_{-0.3}{}^{}$	11.9 ${}_{-0.2}{}^{}{}_{}{}^{+0.2}$

Table 3. Table 3: Photometry.

ADF22 ID	$u^{\ast }$	$B$	$V$	$R$	$i^{\prime }$	$z^{\prime }$	$J$
ADF22-LineA,B	$>$ 26.6	$>$ 27.0	$>$ 27.1	$>$ 27.2	$>$ 26.9	$>$ 26.2	$>$ 24.2
ADF22-LineD	23.84$\pm 0.05$	23.46$\pm 0.03$	22.86$\pm 0.01$	22.08$\pm 0.01$	21.47$\pm 0.01$	21.13$\pm 0.01$	19.79$\pm 0.01$

Equations

$$\frac{L_{[\mathrm{CII}]}}{L_{\odot }}=1.04\times 10^{-3}{\left(\frac{D_{\mathrm{L}}}{\mathrm{Mpc}}\right)}^2\frac{{\nu }_{\mathrm{o}}}{\mathrm{GHz}}\frac{S^{\mathrm{v}}}{\mathrm{Jy}\mathrm{km}{\mathrm{s}}^{-1}}.$$

$$\frac{\mathrm{SF}{\mathrm{R}}_{[\mathrm{CII}]}}{{\mathrm{M}}_{\odot }\mathrm{y}{\mathrm{r}}^{-1}}=10^{-5.73\pm 0.32}{\left(\frac{L_{[\mathrm{CII}]}}{{\mathrm{L}}_{\odot }}\right)}^{0.80\pm 0.05}.$$

$$\frac{\mathrm{SF}{\mathrm{R}}_{\mathrm{UV}}}{{\mathrm{M}}_{\odot }\mathrm{y}{\mathrm{r}}^{-1}}$$

$$\frac{\mathrm{SF}{\mathrm{R}}_{\mathrm{IR}}}{{\mathrm{M}}_{\odot }\mathrm{y}{\mathrm{r}}^{-1}}$$

Full text

1 Introduction

The cosmic star-formation history in the early Universe has been studied in optical/near-infrared (NIR) wavelengths, which trace ultraviolet (UV) radiation in rest-frame at high redshifts (e.g., [Madau & Dickinson (2014)]). The UV star formation rate density (SFRD) does not account for all components of star-forming galaxies (e.g., [Bouwens et al. (2012), Bouwens et al. (2016)]). Recent studies suggest that far-infrared (FIR) SFRD contributes more than half of the total at $z=0-4$ (e.g., [Blain et al. (1999), Barger et al. (2012), Burgarella et al. (2013), Gruppioni et al. (2013), Swinbank et al. (2014)]). Millimeter/submillimeter (mm/submm) galaxy surveys would be, in principle, efficient to probe the dust-obscured component of SFRD at high-redshift (Takeuchi et al., 2005; Burgarella et al., 2013; Chen et al., 2016; Carniani et al., 2015; Fujimoto et al., 2016; Aravena et al., 2016a; Dunlop et al., 2017; Umehata et al., 2017). The advantage of such observations in mm/submm is the well-known negative $k$-correction; the continuum flux of a typical star-forming galaxy of fixed SFR remains approximately constant with increasing redshift (Blain et al., 2002). However, it is often difficult to estimate redshifts for very faint and dusty sources (e.g., Simpson et al. (2014)).

Strong emission lines such as [Cii]158$\rm\mu$m or [Oiii] 88$\rm\mu$m lines can be used to study the SFR and gas properties of high-$z$ star-forming galaxies as well as to determine their spectroscopic redshifts (e.g., Colbert et al. (1999); Maiolino et al. (2005); Brauher et al. (2008); Swinbank et al. (2012); Venemans et al. (2012); De Looze et al. (2014); Inoue et al. (2014); Willott et al. (2015); Maiolino et al. (2015); Inoue et al. (2016); Carniani et al. (2017)). Interestingly, Capak et al. (2015) report that Lyman-break galaxies (LBGs) at $z=5-6$ show enhancement of [Cii] emission relative to the FIR continuum compared with mm/submm-selected galaxies. They also serendipitously detected a [Cii] emitter which is faint in both the rest-UV and FIR continuum. Combining observations in rest-UV, FIR and mm/submm emission lines appears to be essential to understand the physical properties of galaxies at high redshifts (e.g., Bouwens et al. (2016); Aravena et al. (2016a); Dunlop et al. (2017)).

One of the brightest submm emission lines is [Cii] (e.g., Maiolino et al. (2005, 2009); Iono et al. (2006); Venemans et al. (2012); Swinbank et al. (2012); Willott, Omont, & Bergeron (2013); Willott et al. (2015); Maiolino et al. (2015); Capak et al. (2015); Díaz-Santos et al. (2016); Pentericci et al. (2016)). Carbon in the interstellar medium is largely in a singly ionised state in a variety of environments, from Hii regions to molecular clouds, because the ionization potential of atomic carbon is 11.3 eV, lower than that of hydrogen and close to dissociation energy of CO of 11.1 eV (e.g., Wolfire et al. (2010); Carilli & Walter (2013)). The critical density of [Cii] emission is about $3\times 10^{3}$ cm*-3*, and thus [Cii] emission can arise even in a molecular cloud with temperature around 92 K (Hollenbach & McKee, 1989). Therefore [Cii] radiative cooling often dominates in regions with a wide range of densities (e.g., Wolfire et al. (1995); Kaufman et al. (1999)). Finally, [Cii] emission is thought to be a potential tracer of SFR because of its main origin of photo-dissociated region associated with young, massive stars (e.g., De Looze et al. (2011, 2014); Sargsyan et al. (2012); Kapala et al. (2015)). An important observational advantage is that [Cii] line emission at $z>4$ is redshifted to wavelengths with low atmospheric absorption and thus it is possible to detect [Cii] line emission even from galaxies at $z=7$ (e.g., Venemans et al. (2012); Aravena et al. (2016b); Pentericci et al. (2016)).

A number of high-redshift [Cii] emitters are expected to be detected with forthcoming high sensitivity observations with the Atacama Large Millimeter/submillimeter Array (ALMA) (e.g., Geach & Papadopoulos (2012); da Cunha et al. (2013); Matsuda et al. (2015); Aravena et al. (2016b)). In this paper, we present a blind search for [Cii] emitters using ALMA Cycle 2 data (Umehata et al., 2017). We briefly introduce the observations in §2. The details of our data analysis is described in §3. Then we show the results in §4 and discuss the implications for cosmic star formation history in §5. We summarize the results and discussions in §6. Throughout the paper, we adopt the standard $\rm\Lambda$CDM cosmology with the matter density $\Omega_{\rm M}=0.3$, the cosmological constant $\Omega_{\rm\Lambda}=0.7$, the Hubble constant $h=0.7$ in the unit of $H_{0}=100~{}\rm km~{}s^{-1}Mpc^{-1}$. All magnitudes are given in the AB system, unless otherwise noted. We calculate SFR assuming Chabrier initial mass function (IMF) (Chabrier, 2003), with an integration range from 0.08 $\rm M_{\odot}$ to 100 $\rm M_{\odot}$. When needed, we use the conversion factor of 1.8 from the Chabrier IMF to the equivalent Salpeter IMF (Salpeter, 1955) and 1.1 from the Chabrier IMF to the Kroupa IMF (Kroupa, 2001).

2 Observation

We analyze data from the ALMA Deep Field survey of SSA22 (ADF22) observed in Band 6 in ALMA Cycle 2 in June 2014 and April 2015 (Proposal ID 2013.1.00162.S, PI: H. Umehata). The details of the observation are described in Umehata et al. (2017).

ADF22 is a survey field with an area of $2^{\prime}\times 3^{\prime}$ centered on a $z=3.09$ proto-cluster; RA (J2000) = 22h17m34s, Dec (J2000) = +00∘17*′00′′* consisting of 103 pointing fields. The field was observed using four 1.875 GHz spectral windows (SPW) with the central frequency of 263 GHz, which corresponds to the [Cii] redshift of $z=6.2$.

The typical angular resolution of a combined data is $0^{\prime\prime}.72\times 0^{\prime\prime}.62$ corresponding to $\sim 6$ kpc at $z=6.2$. The on-source time per pointing in the fields is 4.5 min. The data observed in 2014 and 2015 have angular resolution of $0^{\prime\prime}.54\times 0^{\prime\prime}.49$ and $1^{\prime\prime}.24\times 0^{\prime\prime}.87$, and on-source time per pointing in the fields of 2.5 min. and 2.0 min. for 2014 and 2015, respectively.

The four SPWs have root-mean-square (RMS) noise level of 0.7, 0.7, 0.8 or 0.9 mJy beam*-1* at a 36 km s*-1* velocity resolution. The RMS of each SPW at 36 km s*-1* resolution of combined, 2014 and 2015 data as a function of the observed frequency are shown in Figure 1, where no significant atmosphere absorption is seen. Other properties of the data are listed in Table 1.

In order to search faint emission line sources, we use high sensitivity data of 80 pointing fields; Field 1 - Field 80 and search in a rectangle area of $\sim$ 5 arcmin2; (RA (J2000), Dec (J2000)) = (22h17m31.86s, +00∘15*′25.46′′) to (22h17m38.17s, +00∘18′35.05′′*), and a frequency coverage of 253.1–272.8 GHz (Table 1). The effective survey area corresponds to about 29 comoving Mpc2 and the effective survey volume is $\sim$ 2.2 $\times$ $10^{3}$ comoving Mpc3 at $z=6.2$.

3 Method

The flowchart of our source selection method is shown in Figure 2. The data are analyzed with Common Astronomy Software Application (casa) ver. 4.5.3 (McMullin et al., 2007). We make continuum-subtracted datacube by using uvcontsub and clean. We first spectrally smooth the data to obtain high signal-to-noise (S/N) ratios. The top-hat spectral smoothing window is set to be 0, 2, 4, …, 12, 15, 18, …, 21 slices, with a slice width corresponding to $\sim$18 km s*-1*. We use the spectral smoothing function “boxcar” so that the velocity sampling of the output data is kept constant. As each spectral data slice has a different RMS value as shown in Figure 1, we normalise each slice by its RMS. We call a datacube thus-generated as “S/N cube”.

We use clumpfind (Williams, de Geus, & Blitz, 1994) to search emission line sources in the S/N cube. We search for sources with a threshold value “low” of clumpfind of $\geq$ 4.5. We then do ‘matching’ of the clumps detected at the same position between the S/N cubes in the same SPW with different resolutions and retain the clump that has maximum S/N ratio (see also Table 1). We select clumps that have the S/N ratio larger than $6.0\sigma$ and also larger than the maximum negative S/N ratio measured in the inverted S/N cube in each SPW (see also Figure 2), in order to avoid contamination by spurious sources (e.g., Hatsukade et al. (2016)). We also check line spectral features of the detected clumps (sources) in the datacube separately for those observed in 2014 and 2015.

For the detected sources, we search for their counterparts in $u^{*}$ band taken with the Canada France Hawaii Telescope/MegaCam obtained by archival data (Kousai, 2011), $B$, $V$, $R$, $i^{\prime}$, $z^{\prime}$, NB912, $J$, $H$, $K$ band taken with the Subaru Telescope (Hayashino et al., 2004; Nakamura et al., 2011; Suzuki et al., 2008; Uchimoto et al., 2012), 3.6 $\rm\mu$m, 4.5 $\rm\mu$m, 5.8 $\rm\mu$m, 8.0 $\rm\mu$m, 24 $\rm\mu$m taken with the Spitzer Space Telescope/IRAC and MIPS (Hainline et al., 2009; Webb et al., 2009) 0.5 keV, 2 keV and 8 keV taken with the Chandra X-Ray Observatory (Lehmer et al., 2009).

[Cii] line emitting galaxies at $z=6.0-6.5$ are likely to be detected only longward of $z^{\prime}$ band and/or in narrow-band NB912 if they are LBGs or Ly$\rm\alpha$ emitters (LAEs) (e.g., Nakamura et al. (2011)), although the available $z^{\prime}$ band and NB912 data could be too shallow for high-redshift [Cii] emitters in our blind search. For the sources with counterparts, we estimate either their photometric redshift by means of spectral energy distribution (SED) fitting or spectroscopic redshift by assuming their line species. SED fitting is calculated by using hyperz software (Bolzonella et al., 2000). In §4.3, we also use the equivalent width and the source number density to consider if the detected [Cii] emitter candidates are other line emitters.

4 Result

4.1 Source Detection

We detect four line emitter candidates. Hereafter, we call the two sources without optical, NIR and FIR counterparts ADF22-LineA and ADF22-LineB. Those with counterparts are dubbed as ADF-LineC and ADF-LineD. The peak S/N ratio are $6.5\sigma$, $6.2\sigma$, $7.7\sigma$ and $6.5\sigma$, for ADF22-LineA, B, C and D, respectively. The first moment images of the candidates are shown in Figure A, and their properties are shown in Table 2.

Figure 3 shows the cumulative number of positive and negative clumps as a function of S/N ratio. Although the S/N ratios of ADF22-LineA, B and D are below $6\sigma$ at the original spectral sampling, the lines are detected at $\geq 6\sigma$ in the smoothed S/N cubes. We compare the spectral line features of the emitter candidates in different observation epochs, 2014 and 2015 (Figure 4). Overall, the contiguous positive signals over a velocity range of $\geq$ 180 km s*-1* and the line features commonly seen suggest that the candidates are likely real sources.

We note here that we also detect one clump with $6.3\sigma$ in the inverted S/N cube, and thus we would naively be concerned that one candidate with $6.2\sigma$ could be a spurious source. However, the most-negative clump is actually detected in SPW 3, where none of our four candidates is located. We also find that datacubes with higher RMS value have higher maximum negative detection (Table 1). Since the datacubes in different SPWs have different properties, the existence of the high-$\sigma$ negative clump in SPW 3 does not immediately impacts the confidence of our line emitter candidates. ADF22-LineB has a lower S/N ratio than ADF22-LineA, whereas it has non-negative $z^{\prime}$ band counterparts with $<3\sigma$ (see also Figure 5). Velocity-gradient is also seen around ADF22-LineB (see also Figure A).

4.2 Line Identification

Figure 5 shows the images of the four candidates in $B$, $V$, $z^{\prime}$, 3.6 $\rm\mu$m and 1.1 mm wavebands. We plot SED and model fit for ADF22-LineD in Figure B, and the measured photometry in the detected bands are given in Table 3. The photometric redshift is estimated by using hyperz (Bolzonella et al., 2000). We fit the SED templates by Bruzual & Charlot (1993) to the spectral coverage from UV to 8 $\rm\mu m$, assuming a Calzetti dust extinction law (Calzetti et al., 2000). We also use SED templates from swire library (Polletta et al., 2007).

ADF22-LineA and B: We do not find any secure counterpart nor close sources within 2*′′* of the sources. Therefore we regard LineA and LineB as good [Cii] emitter candidates.

ADF22-LineC: LineC very likely arises from the galaxy ADF22.4 reported in Umehata et al. (2017), whose redshift is determined to be $z=3.091$ from far-infrared spectroscopic follow up observations (Umehata et al.  in prep.). Thus we identify ADF22-LineC as CO(9-8) line emission at $z=3.091$. In addition, the optical component near ADF22-LineC is a known galaxy at $z=0.55$ (Kubo et al., 2015), but we exclude possibility of ADF22-LineC to be at $z=0.55$ because there is no obvious line species observed at 1.1 mm. LineC is also detected in X-ray (Lehmer et al., 2009), which may indicate that ADF22-LineC is an AGN-host galaxy. Further details of the galaxy will be discussed in Umehata et al. (in prep.).

ADF22-LineD: LineD is spatially consistent with the position of a tentatively detected continuum source ADF22.21 reported in Umehata et al. (2017). The result of SED fitting shows that reduced $\chi^{2}$ values becomes minimum at $z\sim 0.6-0.8$ (Figure B left). Interestingly, the SED is well fitted by that of Arp220 placed at $z\sim 0.7$ (Figure B right). By searching for possible lines in this redshift range, we conclude that ADF22-LineD is likely a CO(4-3) emitter at $z=0.71$.

4.3 Possibility of other line emissions

Besides the [Cii] line emission, there are also possibilities that ADF22-LineA and LineB are other emission line sources, such as 12CO line emission at $z\leq 3.1$, H2O at $z\sim 1.9$ or $2.8$, [Nii]205$\rm\mu m$ at $z\sim 4.6$, [Oi]145$\rm\mu m$ at $z\sim 6.9$, [Nii]122$\rm\mu m$ at $z\sim 8.5$ or [Oiii]88$\rm\mu m$ at $z\sim 12$ (Swinbank et al., 2012; Tamura et al., 2014; Ono et al., 2014; Decarli et al., 2016a; Aravena et al., 2016b).

If ADF22-LineA and LineB are 12CO emitters, the number density is consistent with the result of ASPECS survey (Decarli et al., 2016a) and with semi-analytical/empirical predictions referred to the article (Lagos et al., 2012; Popping et al., 2016; Vallini et al., 2016). Thus we cannot exclude the possibility of 12CO emitters by the discussion of detectability.

We compare the equivalent widths (EWs) in observed frame of the four candidates. The estimated EWs are $>$ 8.6, $>$ 14.6, 1.1 and 7.3 $\rm\mu m$ for ADF22-LineA, B, C and D, respectively, assuming $3\sigma$ continuum flux limit. ADF22-LineA and B have higher EW than the blindly detected 12CO emitters in our survey. The left and middle panels of Figure 6 also show the distribution of the EWs in 0.9–1.3 mm observed frame of the four candidates, high-redshift [Cii] emitting LBGs and LAEs (Capak et al., 2015; Pentericci et al., 2016), [Cii] emitter candidates detected in ASPECS (Aravena et al., 2016b), 12CO emitter candidates detected in band 6 in ASPECS (Decarli et al., 2016b). The EWs of ADF22-LineA, B and other high-redshift [Cii] emitter/candidates are comparable. Given these information, we argue that ADF22-LineA and B are more likely [Cii] emitters at $z=6.5$ and $6.0$, rather than CO emitters at $z\leq 3.1$. EW values of ADF22-LineA and B are comparable to those of the blindly detected CO emitters. Further consideration by using forthcoming follow up observation and theoretical study will be needed to yield any insight about the trends of EW distributions. As with ADF22-LineA and B, blindly detected line emitter candidates are expected to have often no counterpart (Aravena et al., 2016b). Thus it is important to study the EWs of a large sample of CO/[Cii] emitters. We note that H2O molecular lines are expected to have similar line flux to CO line emission in the submm band (e.g, Rangwala et al. (2011); Omont et al. (2013)), and thus can be distinguished from high-redshift [Cii] emitters by comparing their EWs.

[Cii] luminosity, $L_{\rm[CII]}$ of ADF22-LineA and B is calculated by using luminosity distance $D_{\rm L}$, observed frequency $\nu_{\rm o}$, velocity-integrated flux $S^{\rm v}$ (e.g., Carilli & Walter (2013)),

[TABLE]

The estimated $L_{\rm[CII]}$ of $4-7\times~{}10^{8}~{}L_{\odot}$ is consistent with the values of normal star-forming galaxies in the local universe (e.g., Swinbank et al. (2012)), thus we do not consider the effect of [Cii] line deficit (e.g., Graciá-Carpio et al. (2011); Díaz-Santos et al. (2013)). We then derive [Cii] luminosity function by using SFR-[Cii] luminosity relation (De Looze et al., 2014) and SFR function at $z=6$ (Smit et al., 2012). The right panel of Figure 6 shows that the detection of one [Cii] emitter candidate in the survey area is roughly consistent with the expected [Cii] number counts, if we use the SFR-$L_{\rm[CII]}$ relation by De Looze et al. (2014) that is calibrated from observations of nearby low-metallicity dwarf galaxies (see also §5):

[TABLE]

We also plot the predicted number counts of [Nii] 122$\rm\mu$m and [Nii] 205$\rm\mu m$ from the model of Orsi et al. (2014). The predicted number count of [Oiii] 88$\rm\mu$m emission at $z\sim 12$ are lower than the [NII] 122$\rm\mu$m emission (Orsi et al., 2014). It is expected that such line emitters will not be found in our survey area. From the discussion above, we assume ADF22-LineA and B to be [Cii] emitter candidates.

5 Discussion

In order to discuss the cosmic star formation history, we derive the SFRs of ADF22-LineA and ADF22-LineB assuming that they are [Cii] emitters at $z=6$. We calculate total SFR by summing up the dust-uncorrected SFRUV and SFRIR, SFRUV+IR(e.g., Buat et al. (2010)), by using the following equations (Kennicutt, 1998);

[TABLE]

where Lν refers to the UV luminosity density in the wavelength range 1500-2800 Å, and LIR refers to the IR luminosity integrated over 8-1000 $\rm\mu$m. We estimate $L_{\rm IR}$ of ADF22-LineA, B and D from the observed 1.1 mm continuum fluxes by using the SED fitting method of Chary & Elbaz (2001). Continuum upper limits of LineA and B are assumed to be $3\sigma$. $L_{\rm IR}$ of ADF22-LineC is referred to estimation of Umehata et al. (2017). Upper limit of UV luminosity is estimated in Nakamura et al. (2011). The obtained SFRUV+FIR is $<$ 30 $\rm M_{\odot}yr^{-1}$, being consistent with the SFR-$L_{\rm[CII]}$ relation of De Looze et al. (2014) that is calibrated by local low-metallicity dwarf galaxies (Figure 6 left). We note that the relation calibrated by high-$z$ galaxies is considered to be applicable to bright [Cii] emitters with $>~{}10^{9}\rm L_{\odot}$ (De Looze et al., 2014) and exceeds the SFRUV+FIR upper limit for the candidates in this survey. The estimated SFR[CII] from the low-metal dwarf relation are 13${}^{+4}_{-5}$ and 20${}^{+9}_{-10}$ M⊙/yr for ADF22-LineA and LineB, respectively (see also Table 3).

We estimate the [Cii] luminosity function (LF) at $z\sim 6.2$ from only one source, because one of the two [Cii] candidates has a relatively low-$\sigma$ and thus could possibly be a spurious source (see §4.1). We show this result in Figure 7 and compare it to [Cii] LFs from previous studies. The estimated [Cii] LFs at $z=0.0$ (Swinbank et al., 2012; Hemmati et al., 2017) are derived from the follow up observation of the IRAS sources (Brauher et al., 2008) or samples from the Great Observatories All-sky LIRG Survey (Díaz-Santos et al., 2013). We indicate the upper limit at $z=4.4$ derived by Matsuda et al. (2015) using ALMA Cycle 0 archive data, and the lower limit at $z=4.4$ based on two serendipitous detection in ALESS survey (Swinbank et al., 2012). The estimation at $z=5-6$ is derived from follow up observation by Capak et al. (2015). We also indicate the estimation of over-dense region at $z=6$ by Miller et al. (2016) . The constraint for [Cii] LF at $z=6-8$ is provided by result of ASPECS (Aravena et al., 2016b), which is based on an assumption that all [Cii] candidates are real [Cii] emitters. As discussed In §4.2, We also derive the simple model of [Cii] luminosity function at $z$ = 6 by using SFR-$L_{\rm[CII]}$ relation (De Looze et al., 2014) and SFR function at $z$ = 6 (Smit et al., 2012). Our [Cii] LF model is close to our own observational result and the other studies, whereas the estimated LFs based on the empirical relations for high-redshift, and for all galaxies (De Looze et al., 2014), do not match the observational result at $z>4$. We note that if completeness of the detection is lower than unity, the estimated [Cii] LF represents the lower limit.

We calculate a conservative limit of [Cii] SFRD from the mean of the SFR[CII] of the two sources divided by the survey volume (Figure 8). The derived [Cii] SFRD is $>7.5\times 10^{-3}\rm M_{\odot}yr^{-1}Mpc^{-3}$. Interestingly, this is close to the dust-uncorrected UV SFRD at $z=6.2$. The contribution of the only one [Cii] emitter with faint UV and dust emission to the cosmic SFRD might already constitute a major contribution. The result may imply the existence of the untraceable component of the SFRD by rest-UV. In order to confirm the truth of this, the estimation of a faint end slope of the [Cii] LF would be crucial.

In figure 8, we also derive upper limits of SFRDs at $z=7-12$ from the non-detections of [Oi] 145$\rm\mu$m, [Nii] 122$\rm\mu$m [OIII] 88$\rm\mu$m lines in our search as discussed in §4.3. The SFRs are calculated from line luminosities by using observational relations estimated by Farrah et al. (2013). This result demonstrates that line survey enables us to estimate SFRDs at multiple redshifts at once.

There are a few possible mechanisms for the [Cii] line emission to be particularly intense relative to FIR and UV emission. For example, it can be caused by high far-UV radiation from massive, young stars in the early universe (e.g., Wolfire et al. (1995)). The environment of a low metallicity and a low dust-to-gas ratio can also cause enhancement of [Cii] radiative cooling (Wolfire et al., 1995; Capak et al., 2015). In particular, the low dust-to-metal environment may not only enhance [Cii] line emission but also weaken dust continuum emission (Inoue, 2003; Asano et al., 2014). Hot dust dominates the short-wavelength portion of the SED (Casey et al., 2014; Zhou et al., 2016), making the dust continuum at long-wavelengths to be relatively suppressed. The size distribution of dust grains also affect faint FIR continuum (Takeuchi et al., 2003, 2005). Altogether, observations in the submm-band can provide invaluable information on the physical properties of high-redshift galaxies. Future deep submm surveys will enable us to understand the formation of galaxies and to probe the early cosmic star-formation history.

Summary

We search millimeter line emitters by using 1.1 mm ADF22 survey data taken in ALMA Cycle2. Our newly constructed method for line search worked for detecting two CO emitters at $z=0.7$ and $3.1$ and two [Cii] emitter candidates at $z=6.0$ and $6.5$ with $>6\sigma$. [Cii] emitter candidates are faint in all counterparts. The line species of the CO emitters are identified by SED fitting or spectral follow up observation. For [Cii] emitter candidates, the possibility of other line emissions are excluded by discussion about number counts, line ratio and EWs. Since one spurious source is possibly contaminated with the candidates, we assume at least one of the two candidates to be a real [Cii] emitter. We constrain $z=6$ [Cii] LF for one source and found that the [Cii] LFs at $z>4$ show good agreement with the predicted LF by using SFR-$L_{\rm[CII]}$ relation calibrated by local metal poor dwarfs. We also found that estimated [Cii]-based SFR are consistent with upper limit of total SFR if we use the SFR-$L_{\rm[CII]}$ relation for local metal poor dwarfs. We estimate a conservative limit of [Cii] SFRD at $z=6.2$ for one source, which is close to the dust-uncorrected UV SFRD at $z=6.2$. The results might be imply that mm/submm line survey is a powerful probe to estimate untraceable SFRD component from rest-UV observation at high-redshift. The constrain for faint end slope of [Cii] LF from further line survey and FIR/UV follow-up observation will give us the truth of such implication and detailed picture of cosmic star-formation history.

Funding

This work was supported by the ALMA Japan Research Grant of National Astronomical Observatory of Japan (NAOJ) Chile Observatory, NAOJ-ALMA-0071 and NAOJ-ALMA-0160. NHH was supported by the grant of NAOJ Visiting Fellow Program supported by the Research Coordination Committee, National Astronomical Observatory of Japan (NAOJ) and by funding from Foundation for Promotion of Astronomy. IRS acknowledges support from STFC (ST/L00075X/1), the ERC Advanced Grant DUSTYGAL (321334) and a Royal Society Wolfson Research Merit Award.

Acknowledgements

We are grateful to the referee R. Maiolino for his useful comments and suggestions. The authors wish to thank A. Sternberg, E. Seaquist, L. Yao, M. Oguri, H. Nagai, I. Shimizu, K. Mawatari, T. Saito. This paper makes use of the following ALMA data: ADS/JAO.ALMA#2013.1.00162.S. Data analysis ware carried out on common use data analysis computer system at the Astronomy Data Centre, ADC, of the NAOJ. IRAC data was reduced and provided by J. Huang. ALMA is a partnership of ESO (representing its member states), NSF (USA) and NINS (Japan), together with NRC (Canada) and 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.