Reionization Era Bright Emission Line Survey: Selection and Characterization of Luminous Interstellar Medium Reservoirs in the z > 6.5 Universe

Over the last few years, it has become increasingly clear that many of the brightest known ISM-cooling lines at z ≳ 5.5 are found in especially UV-bright galaxies (M_UV,AB ≲−22 mag: ≥2 ${L}_{\mathrm{UV}}^{* }$). Examples include the 10 z ∼ 5.5 sources targeted by Capak et al. (2015), two z ∼ 6.1 galaxies targeted by Willott et al. (2015), CR7 by Matthee et al. (2017), B14-65666 by Hashimoto et al. (2019), VR7 by Matthee et al. (2019), and the three z ∼ 6.1–6.2 galaxies targeted by Harikane et al. (2020). All show bright (≳2 mJy km s⁻¹) [C ii]_{158 μm} lines. Meanwhile, follow-up of z ≥ 6 galaxies with redshifts from Lyα have frequently revealed only moderate- to low-luminosity (≲1 mJy km s⁻¹) [C ii]_{158 μm} lines (Ouchi et al. 2013; Ota et al. 2014; Pentericci et al. 2016; Bradač et al. 2017; Carniani et al. 2018, 2020), strongly suggesting that there may be an inverse correlation between Lyα equivalent width (EW) and the mass of the ISM reservoirs (Figure 18 from Harikane et al. 2020). Given such, it seems the best way to obtain a representative sample of massive galaxies at z > 6 is to identify galaxies on the basis of their UV luminosity (rather than Lyα emission alone).

To quantify the relationship between the UV luminosity M_UV of galaxies and their ${L}_{{[{\rm{C}}\,{\rm\small{II}}]}_{158\,\mu {\rm{m}}}}$, we present a compilation of some recent results at z > 4 (Capak et al. 2015; Smit et al. 2018; Matthee et al. 2019; Béthermin et al. 2020; Schouws et al. 2022b) in Figure 1. A conversion factor of 7.1 × 10⁻²⁹ L_ν [M_⊙ yr⁻¹/(erg s⁻¹ Hz⁻¹)] (M. Stefanon et al. 2022, in preparation) is used to transform the UV luminosities of galaxies into SFRs. What is particularly striking here is the strong correlation we observe between ${L}_{{[{\rm{C}}\,{\rm\small{II}}]}_{158\,\mu {\rm{m}}}}$ and the unobscured SFR_UV, with a particularly clear correlation around 10 M_⊙ yr⁻¹. The fraction of sources with [C ii]_{158 μm} luminosities in excess of 2 × 10⁸ L_⊙, the typical sensitivity limit achieved in REBELS observations, increases from 35% at 10 M_⊙ yr⁻¹ to 80% at 20 M_⊙ yr⁻¹.

Because of the strong correlation between ${L}_{{[{\rm{C}}\,{\rm\small{II}}]}_{158\,\mu {\rm{m}}}}$ and M_UV, one can very efficiently derive redshifts for galaxies that are bright in the UV by scanning for the [C ii]_{158 μm} ISM-cooling line with ALMA, as demonstrated in the introduction by both pilot programs to REBELS (Smit et al. 2018; Schouws et al. 2022b).

In constructing a sample of 40 UV-bright galaxies to follow up with the REBELS LP, we made use of deep wide-area optical + near-IR observations over the ∼2 deg² COSMOS/UltraVISTA field (Scoville et al. 2007; McCracken et al. 2012), the ∼5 deg² UKIDSS/UDS + VIDEO/XMM-LSS fields (Lawrence et al. 2007), and a wide range of HST search fields, including CANDELS, CLASH, and the BoRG/HIPPIES pure parallel fields.

A significant fraction of the UV-bright targets for REBELS are drawn from the ∼2 deg² COSMOS/UltraVISTA field. In constructing our source catalogs for this field, we made use of the very sensitive Y, J, H, and K_s near-IR observations (both data release 3 and 4) from the UltraVISTA program (McCracken et al. 2012), the deep optical Subaru Suprime-Cam BgVriz observations from Taniguchi et al. (2005), the deep optical ugriyz observations from the CFHT Deep Legacy survey (Erben et al. 2009; Hildebrandt et al. 2009), and the ugrizy Subaru Hyper Suprime-Cam (HSC) observations (Aihara et al. 2018b, 2018a). Use of the Spitzer/IRAC 3.6 μm and 4.5 μm observations from SCOSMOS (Sanders et al. 2007), Spitzer Large-Area Survey with HSC (SPLASH: Steinhardt et al. 2014), SMUVS (Caputi et al. 2017; Ashby et al. 2018), COMPLETE (Labbé et al. 2016), and COMPLETE2 (Stefanon et al. 2018) programs were also made.

The bulk of the remaining targets are drawn from the ∼5 deg² UKIDSS/UDS and VIDEO/XMM-LSS area. The UDS/XMM-LSS area includes sensitive near-IR observations from both the UKIDSS/UDS (Lawrence et al. 2007; J, H, and K bands) and VISTA Deep Extragalactic Observations (VIDEO) programs (Jarvis et al. 2013; Y, J, H, and K_s ), sensitive optical observations from the Subaru Hyper Suprime-Cam deep and wide-area programs (Aihara et al. 2018a, 2018b), and Spitzer/IRAC observations from the SPLASH and Spitzer Extragalactic Representative Volume Survey (SERVS; Mauduit et al. 2012) programs.

We also considered including UV-bright z > 6.5 galaxies from the CANDELS, CLASH, and BoRG/HIPPIES fields as targets in the REBELS program. In total, the CANDELS, CLASH, and BoRG/HIPPIES fields that we used for selecting targets covered an area of 0.2 deg² (excluding those HST CANDELS fields within COSMOS and UDS).

From each of these search fields, we considered a range of different catalogs in arriving at our final target list. Here, we provide a brief summary of each sample we consider:

Bowler et al. (2014, 2017) (z ∼ 7): The Bowler et al. (2014) z ∼ 7 selection was constructed based on deep z, Y, J, H, and K_s observations obtained over the COSMOS/UltraVISTA and UKIDSS/UDS fields with Subaru Suprime-Cam + VISTA and Subaru, VISTA, and UKIRT, respectively. The deep Y + J and J band images over UltraVISTA/COSMOS and the UKIDSS/UDS fields, respectively, were used as the detection image in constructing source catalogs for the search. Then, after removing all sources detected at 2σ in the optical imaging data, the lephare photometric redshift software (Arnouts et al. 1999; Ilbert et al. 2006) was run, using Bruzual & Charlot (2003) models with a range of exponentially declining star formation histories, metallicities of 0.2 Z_⊙ and Z_⊙, a range of dust attenuation (A_V < 4), Lyα EWs to 240 Å, and Madau (1995) IGM absorption. Then, Bowler et al. (2014) compared their lephare fit results with similar fits to stars from the SpecX library (Burgasser 2014) to identify and exclude any possible low-mass stars from their selection. Bowler et al. (2017) refined the Bowler et al. (2014) selection, taking advantage of subsequent WFC3/IR F140W imaging they obtained over the fields with HST, identifying two additional and fainter z ∼ 7 galaxies in the neighborhood of their bright candidates and removing three crosstalk artifacts from their catalogs (Section 3.2 of Bowler et al. 2017). In total, 22 bright z ∼ 7 galaxies were identified as part of the Bowler et al. (2014) and Bowler et al. (2017) studies, with M_UV luminosities ranging from −20.7 to −23.2 mag.

Stefanon et al. (2017b, 2019a) (z ∼ 8–9): The selection of z ∼ 8–9 galaxies in Stefanon et al. (2017b, 2019a) was performed by first creating a large parent catalog over the COSMOS/UltraVISTA fields based on the Y, J, H, and K_s images and then applying an optical nondetection and two-color Lyman-break galaxy-like criteria. Stefanon et al. (2019a) then made use of the mophongo package (Labbé et al. 2006, 2013, 2015) to do careful optical, near-IR, and Spitzer/IRAC photometry for each source in the parent catalog, modeling and subtracting the flux from nearby neighbors to improve the robustness of the flux measurements. Candidate z ∼ 8–9 galaxies were then selected by running the eazy photometric redshift code (Brammer et al. 2008) on the optical+near-IR+Spitzer/IRAC photometry that had been derived. Eighteen z ∼ 8–9 candidate galaxies were identified by Stefanon et al. (2017b, 2019a) over the ∼2 deg² COSMOS field. Through similar fits of the source photometry to the SpecX dwarf star spectral libraries (Burgasser 2014), Stefanon et al. (2019a) explicitly verified that all sources were better fit (Δχ² > 1) by galaxy SED templates than dwarf star templates.

Bowler et al. (2020): Candidate z ∼ 8–11 galaxies were identified using the very deep ground-based optical and near-IR observations over ∼6 deg² in the COSMOS/UltraVISTA and VIDEO/XMM-LSS + UKIDSS/UDS fields. Candidate z ∼ 8 galaxies were required to be detected at 5σ in either than J or H bands, while candidate z ∼ 9 galaxies were required to show a 5σ detection in either the H or K_s /K bands. No detection (<2σ) was allowed for sources in any of the deep optical bands, as well as the Y band for z ∼ 9 candidates. The reality of sources showing 5σ detections in the UDS JHK data was tested by looking for similar 2σ detections in the VISTA data. The redshift likelihood distributions for candidate z ∼ 8–11 galaxies were then computed using a range of exponentially declining star formation histories, metallicities of 0.2 Z_⊙ and Z_⊙, a range of dust attenuation (A_V < 6), and the Madau (1995) IGM absorption. Bowler et al. (2020) identified 27 z ∼ 8–10 galaxies with UV luminosities M_UV ranging from −23.7 to −21.2 mag.

Endsley et al. (2021a): Candidate z ∼ 7 galaxies were identified over the deep ∼2 deg² COSMOS/UltraVISTA and ∼1 deg² UKIDSS UDS fields. Sources were detected from χ² detection images (Szalay et al. 1999) constructed from the deep yYJHK_s HSC/UltraVISTA and yYJHK HSC/VIDEO/UKIDSS imaging observations. Endsley et al. (2021a) then applied the following color criteria: (1) z − y > 1.5, (2) z − Y > 1.5, (3) NB₉₂₁ − Y > 1.0, and (4) y − Y < 0.4. Sources were required to be detected at 5σ in at least one of the y, Y, and J bands and detected at 3σ in all three and undetected at <2σ in the HSC g and r bands. Possible T dwarf contaminants were removed by requiring either Y − J < 0.45 or both J − H > 0 and J − K_s > 0. Endsley et al. (2021a) identified 50 sources over the COSMOS/UltraVISTA and UDS fields.

Schouws et al. (2022b): Candidate z ∼ 7 galaxies in Schouws et al. (2022b) were selected from the ∼2 deg² UltraVISTA observations over the COSMOS field and required to show a >6σ detection in a stack of the Y, J, H, and K_s images and a ${\chi }_{\mathrm{opt}}^{2}\lt 4$ (Bouwens et al. 2011). Best-fit photometric redshifts and redshift likelihood distributions were then derived with eazy using photometry of sources derived from the available CFHT + Subaru optical imaging data, UltraVISTA YJHK_s near-IR observations, and Spitzer/IRAC observations from SCOSMOS, SPLASH, and SMUVS programs. For inclusion in the Schouws et al. (2022b) z > 6.5 sample, sources were required to have an integrated z > 6 probability of >50% and to be better fit by a galaxy SED template than one of the SED templates from the SpecX dwarf star spectral library. Schouws et al. (2022b) identified some 30 bright z ∼ 7 galaxies as part of their study, with M_UV ranging from −23.0 mag to −21.4 mag.

This Paper (see also M. Stefanon et al. 2022, in preparation): Finally, we considered separate selections of z ∼ 7–8 galaxies drawn from the ∼5 deg² VIDEO/XMM-LSS + UKIDSS/UDS fields generated using procedures similar to those described in Stefanon et al. (2019a). The detection was performed on the combination of the J, H, and K_S mosaics, either from the UKIDSS/UDS or VIDEO programs. Candidates at z ∼ 7 were selected through the z − Y > 0.7 mag and 〈YJ〉 − 〈JH〉 < 0.5 mag color criteria, where 〈..〉 denotes the average flux density in the two indicated bands. To identify the samples at z ∼ 8, we imposed Y − J > 0.7 mag and 〈JH〉 − 〈HK_S〉 < 0.5 mag for z ∼ 8. These criteria were coupled with the requirement of nondetection ( < 2σ) in all bands (CFHTLS, HSC, and SUBARU) bluer than the nominal Lyman break, together with a 5σ detection in the combined J, H, and K_S bands, to increase the robustness of the detection. The selection was refined computing the photometric redshifts with eazy (Brammer et al. 2008), after adding the flux densities in the 3.6μm and 4.5μm Spitzer/IRAC bands from the SWIRE, SpUDS, and SEDS programs. Finally, the image stamps of the candidates were visually inspected, to remove contaminants and sources whose flux measurements are contaminated by poorly subtracted neighbors. Sources with better χ² from the templates of Burgasser (2014) were excluded as likely brown dwarf contaminants. This led to the identification of ∼40 candidates at z ∼ 7 and ∼25 at z ∼ 8 brighter than J ∼ 25 − 26 mag.

We then created a master list of 60 UV-bright z = 6.5–10 candidates from the above analyses, as well as five other particularly bright z > 6.5 galaxies that had been identified over various HST legacy fields (MACS0429Z-9372034910, BORG2229-0945-394, Super8-1, Super8-4, and Super8-5; see Bradley et al. 2012; Smit et al. 2014, 2015; Bouwens et al. 2015, 2019, Morishita et al. 2018; Bridge et al. 2019; Salmon et al. 2020). Combining the above analyses, the approximate area probed in the COSMOS/UltraVISTA field, UKIDSS/UDS + VIDEO/XMM-LSS fields, and various HST archival fields (not included in the COSMOS or the UDS fields) is 1.82 deg², 4.97 deg², and 0.2 deg², respectively, or 6.99 deg² in total.

For each of the 60 candidates identified over the UltraVISTA/COSMOS, VIDEO/XMM-LSS, and UKIDSS/UDS fields, measurements of the optical, near-IR, and Spitzer/IRAC fluxes were made by three members of our team (Mauro Stefanon, Rebecca Bowler, and Ryan Endsley) applying their own photometric procedures to the imaging data sets each had compiled of the fields. We refer to these flux measurements as PHOT1, PHOT2, and PHOT3, respectively. Very briefly, PHOT1 is based on neighbor-subtracted aperture photometry of sources in 12 diameter and 18 diameter apertures in the optical/near-IR and Spitzer/IRAC data using mophongo (e.g., see Stefanon et al. 2019a), PHOT2 is based on aperture photometry in 18 diameter apertures and neighbor-subtracted aperture photometry in 28 diameter apertures with T-PHOT (Merlin et al. 2015) for the optical/near-IR and Spitzer/IRAC data, respectively (see, e.g., Bowler et al. 2020), and PHOT3 is based on 12 diameter aperture photometry for the optical/near-IR data, and in 28 diameter apertures after subtracting neighbors using a mophongo/T-PHOT-like algorithm (see Endsley et al. 2021a). For each of these photometric catalogs, aperture measurements are corrected to total based on the curve of growth. More details on these photometric procedures will be provided in M. Stefanon et al. (2022, in preparation). For sources found within various HST legacy fields, use was made of published photometry, except in the case of the Super8 sources where deeper IRAC imaging observations (Holwerda et al. 2018) were utilized.

Each team member doing photometry (M.S., R.B., and R.E.) then made use of a separate photometric redshift code (eazy, lephare, and beagle [Chevallard & Charlot 2016], respectively) to derive a redshift likelihood distribution for each candidate on the basis of their photometry (PHOT1, PHOT2, and PHOT3, respectively). A fourth member of the REBELS team (Rychard Bouwens) then derived a fourth, fifth, and sixth photometric redshift distribution for each source by running the eazy photometric redshift code a second time on the photometry derived by R.B., R.E., and M.S., but using a different set of SED templates SED4. In particular, instead of using the EAZY_v1.0 template set augmented by the Binary Population and Spectral Synthesis code (BPASS; Eldridge et al. 2017) v1.1 for subsolar metallicity (Z = 0.2Z_⊙), which include nebular emission from cloudy (Ferland et al. 2017), and 2Gyr old passively evolving systems with varying amounts (A_V = 0–8 mag) of dust extinction adopting the Calzetti et al. (2000) law, i.e., SED template set SED1, the second eazy runs use the EAZY_v1.0 template set augmented by SED templates from the Galaxy Evolutionary Synthesis Models (GALEV; Kotulla et al. 2009), i.e., SED template set SED4. Nebular continuum and emission lines were included in the latter templates according to the prescription provided in Anders & Fritze-v. Alvensleben (2003), a 0.2Z_⊙ metallicity, and scaled to a rest-frame EW for Hα of 1300 Å.

The mean of the six redshift likelihood distributions was then taken and a single distribution derived (Figure 2). Care was taken to explicitly verify that none of the candidates considered resulted from the VISTA/VIRCAM electronic crosstalk artifact Bowler et al. (2017) identified. Bowler et al. (2017) found that such artifacts occurred at multiples of 128 pixel separations from saturated stars in the VISTA/VIRCAM data. Similar crosstalk artifacts are known to occur in the UKIDSS/UDS observations (Dye et al. 2006; Warren et al. 2006), and as such, care was taken to confirm the reality of sources based on multiple data sets, e.g., with VISTA/VIRCAM, UKIDSS/UDS, and Spitzer/IRAC data. Any source suspected to correspond to an artifact was excluded from consideration.

Sources were then ordered in terms of the likelihood of detecting an ISM-cooling line in each. In computing these likelihoods, an SFR_UV-to-${L}_{{[{\rm{C}}\,{\rm\small{II}}]}_{158\,\mu {\rm{m}}}}$ conversion factor of 2 × 10⁷ L_⊙/(M_⊙ yr⁻¹) was assumed (De Looze et al. 2014), with a 0.3 dex scatter. Additionally, z = 6.5–7.2 candidate sources, z = 7.2–7.7 candidate sources, and z = 7.7–9.5 candidate sources were assumed to have an allocation of 2, 3, and 6 tunings, respectively. To ensure a good sampling of luminous ISM reservoirs as a function of redshift, 20 z = 6.5–7.2 targets, 16 z = 7.2–8.5 targets, and 4 z = 7.8–9.4 targets were chosen.

The final set of 40 targets for the REBELS program is presented in Table 1. Along with the R.A. and decl. for individual targets in our program, we also include the original source names used in the ALMA observations, UV luminosities, and photometric redshifts. UV-continuum slopes β, stellar mass estimates, [O iii]_4959,5007+Hβ EWs, and SFRs are presented in Table 2. UV luminosities for sources and photometry are based on flux measurements made within 24 diameter apertures. Fiducial stellar masses and [O iii]_4959,5007+Hβ EWs for sources in REBELS are derived using the beagle photometric+spectroscopic stellar population fitting code (Chevallard & Charlot 2016), while our UV-continuum slopes β are derived based on power-law fits to the best-fit SED (as performed in Stefanon et al. 2019a). Our procedures for quantifying the UV-continuum slopes β, stellar masses, and [O iii]_4959,5007+Hβ EWs for all 40 sources in the REBELS sample will be detailed in M. Stefanon et al. (2022, in preparation). Unobscured and obscured SFRs for targets are derived from the rest-frame UV luminosities and IR luminosities using prescriptions given in Section 4.2.

Table 2. UV-continuum Slopes, Stellar Population Parameters, and SFRs for Bright z > 6.5 Candidate Galaxies Targeted by the REBELS Program

REBELS ID	Redshift	β	log10 M* (M⊙) a	EW([O iii]+Hβ) (Å) a	SFRUV [M⊙ yr−1] b	SFRIR [M⊙ yr−1] c
REBELS-01	7.177	$-{2.04}_{-0.20}^{+0.24}$	${10.02}_{-0.53}^{+0.41}$	${2.90}_{-0.12}^{+0.28}$	45 ± 5	<35
REBELS-02	${6.65}_{-0.13}^{+0.18}$	$-{2.24}_{-0.36}^{+0.44}$	${9.04}_{-0.49}^{+0.50}$	${3.07}_{-0.16}^{+0.14}$	22 ± 4	<30
REBELS-03	6.969	$-{2.14}_{-0.46}^{+0.63}$	${9.13}_{-0.88}^{+0.64}$	${3.05}_{-0.22}^{+0.33}$	16 ± 4	<34
REBELS-04	${8.57}_{-0.09}^{+0.10}$	$-{2.15}_{-0.38}^{+0.20}$	${8.72}_{-0.68}^{+1.03}$	${3.25}_{-0.31}^{+0.33}$	23 ± 2	${59}_{-36}^{+20}$
REBELS-05	6.496	$-{1.29}_{-0.44}^{+0.36}$	${9.16}_{-1.00}^{+0.85}$	${3.12}_{-0.32}^{+0.30}$	14 ± 3	${40}_{-16}^{+23}$
REBELS-06	${6.79}_{-0.11}^{+0.13}$	$-{1.24}_{-0.35}^{+0.67}$	${9.50}_{-0.79}^{+0.45}$	${2.96}_{-0.18}^{+0.27}$	15 ± 4	${49}_{-19}^{+28}$
REBELS-07	${7.15}_{-0.14}^{+0.20}$	$-{2.39}_{-0.43}^{+0.37}$	${8.69}_{-0.76}^{+0.74}$	${3.18}_{-0.17}^{+0.33}$	20 ± 5	<34
REBELS-08	6.749	$-{2.17}_{-0.58}^{+0.58}$	${9.02}_{-0.68}^{+0.64}$	${3.07}_{-0.22}^{+0.23}$	16 ± 6	${64}_{-24}^{+48}$
REBELS-09	${7.58}_{-0.13}^{+0.27}$	$-{2.66}_{-0.53}^{+0.93}$	${8.65}_{-0.43}^{+0.43}$	${3.77}_{-0.02}^{+0.05}$	49 ± 14	<41
REBELS-10	${7.42}_{-0.91}^{+0.23}$	$-{1.34}_{-0.83}^{+0.48}$	${10.16}_{-0.32}^{+0.31}$	${2.95}_{-0.09}^{+0.10}$	37 ± 10	<41
REBELS-11	${8.24}_{-0.37}^{+0.65}$	$-{1.60}_{-1.15}^{+0.17}$	${9.36}_{-0.56}^{+0.52}$	${3.08}_{-0.19}^{+0.21}$	39 ± 8	<77
REBELS-12 d	7.349	$-{1.99}_{-0.76}^{+0.48}$	${8.94}_{-0.70}^{+0.93}$	${3.26}_{-0.25}^{+0.29}$	30 ± 8	${62}_{-28}^{+38}$
REBELS-13	${8.19}_{-0.50}^{+0.84}$	$-{1.08}_{-0.65}^{+0.59}$	${9.80}_{-0.44}^{+0.43}$	${2.98}_{-0.09}^{+0.13}$	44 ± 9	<72
REBELS-14	7.084	$-{2.21}_{-0.47}^{+0.41}$	${8.73}_{-0.70}^{+0.80}$	${3.21}_{-0.20}^{+0.35}$	35 ± 13	${41}_{-17}^{+24}$
REBELS-15	${6.78}_{-0.09}^{+0.11}$	$-{2.18}_{-0.50}^{+0.52}$	${8.81}_{-0.50}^{+0.50}$	${3.73}_{-0.54}^{+0.10}$	33 ± 9	<44
REBELS-16	${6.74}_{-0.09}^{+0.09}$	$-{1.70}_{-0.76}^{+0.48}$	${9.47}_{-0.36}^{+0.34}$	${3.02}_{-0.10}^{+0.10}$	12 ± 1	<44
REBELS-17	${6.66}_{-0.22}^{+0.16}$	$-{1.70}_{-0.47}^{+0.33}$	${9.07}_{-0.63}^{+0.58}$	${3.04}_{-0.19}^{+0.21}$	15 ± 3	<48
REBELS-18	7.675	$-{1.34}_{-0.32}^{+0.19}$	${9.49}_{-0.73}^{+0.56}$	${3.00}_{-0.17}^{+0.22}$	31 ± 4	${41}_{-16}^{+23}$
REBELS-19	7.369	$-{2.33}_{-0.64}^{+0.45}$	${8.79}_{-0.69}^{+0.69}$	${3.11}_{-0.22}^{+0.21}$	14 ± 3	${52}_{-23}^{+31}$
REBELS-20	${7.07}_{-0.08}^{+0.10}$	$-{2.59}_{-0.60}^{+0.57}$	${8.59}_{-0.63}^{+0.63}$	${3.17}_{-0.11}^{+0.15}$	16 ± 2	<52
REBELS-21	${6.63}_{-0.12}^{+0.09}$	$-{2.15}_{-0.24}^{+0.42}$	${10.38}_{-0.42}^{+0.25}$	${2.84}_{-0.12}^{+0.28}$	18 ± 4	<32
REBELS-22	${7.31}_{-0.10}^{+0.11}$	$-{2.23}_{-0.30}^{+0.21}$	${9.65}_{-0.76}^{+0.42}$	${2.91}_{-0.14}^{+0.31}$	23 ± 2	<34
REBELS-23	${6.68}_{-0.09}^{+0.12}$	$-{1.57}_{-0.45}^{+0.28}$	${9.11}_{-0.61}^{+0.54}$	${3.03}_{-0.16}^{+0.19}$	14 ± 7	<47
REBELS-24	${8.35}_{-0.51}^{+0.66}$	$-{1.56}_{-0.83}^{+0.56}$	${8.97}_{-0.89}^{+0.89}$	${3.13}_{-0.21}^{+0.19}$	20 ± 4	<38
REBELS-25	7.306	$-{1.85}_{-0.46}^{+0.56}$	${9.89}_{-0.18}^{+0.15}$	${2.79}_{-0.06}^{+0.21}$	15 ± 3	${185}_{-64}^{+101}$
REBELS-26	${6.64}_{-0.13}^{+0.22}$	$-{1.92}_{-0.25}^{+0.19}$	${9.54}_{-0.82}^{+0.52}$	${2.98}_{-0.21}^{+0.27}$	17 ± 2	<58
REBELS-27	7.090	$-{1.79}_{-0.45}^{+0.42}$	${9.69}_{-0.34}^{+0.25}$	${2.89}_{-0.11}^{+0.27}$	20 ± 4	${35}_{-13}^{+19}$
REBELS-28	${6.82}_{-0.13}^{+0.13}$	$-{1.95}_{-0.36}^{+0.29}$	${8.61}_{-0.51}^{+0.70}$	${3.26}_{-0.20}^{+0.49}$	29 ± 8	<43
REBELS-29 d	6.685	$-{1.61}_{-0.19}^{+0.10}$	${9.62}_{-0.19}^{+0.19}$	${2.90}_{-0.08}^{+0.12}$	25 ± 3	${35}_{-14}^{+20}$
REBELS-30	6.983	$-{1.95}_{-0.22}^{+0.15}$	${9.28}_{-0.61}^{+0.45}$	${3.06}_{-0.13}^{+0.20}$	27 ± 2	<36
REBELS-31	${6.65}_{-0.06}^{+0.10}$	$-{2.27}_{-0.33}^{+0.18}$	${9.21}_{-0.37}^{+0.37}$	${3.05}_{-0.06}^{+0.06}$	27 ± 4	<52
REBELS-32	6.729	$-{1.50}_{-0.30}^{+0.28}$	${9.55}_{-0.37}^{+0.35}$	${3.01}_{-0.11}^{+0.11}$	14 ± 2	${37}_{-17}^{+23}$
REBELS-33	${6.68}_{-0.12}^{+0.12}$	$-{2.04}_{-0.71}^{+0.24}$	${9.39}_{-0.51}^{+0.40}$	${2.90}_{-0.11}^{+0.27}$	13 ± 2	<49
REBELS-34	6.633	$-{2.02}_{-0.15}^{+0.07}$	${9.33}_{-0.34}^{+0.33}$	${3.03}_{-0.06}^{+0.07}$	31 ± 2	<46
REBELS-35	${6.98}_{-0.10}^{+0.10}$	$-{2.07}_{-1.12}^{+0.27}$	${8.91}_{-0.65}^{+0.66}$	${3.18}_{-0.18}^{+0.30}$	31 ± 3	<47
REBELS-36	7.677	$-{2.57}_{-0.47}^{+0.48}$	${9.40}_{-0.95}^{+0.76}$	${2.99}_{-0.16}^{+0.25}$	24 ± 4	<32
REBELS-37	${7.75}_{-0.17}^{+0.09}$	$-{1.24}_{-0.27}^{+0.16}$	${8.58}_{-0.71}^{+0.74}$	${3.25}_{-0.17}^{+0.32}$	28 ± 3	${74}_{-41}^{+18}$
REBELS-38	6.577	$-{2.18}_{-0.42}^{+0.45}$	${9.58}_{-1.27}^{+0.74}$	${3.01}_{-0.25}^{+0.35}$	18 ± 4	${98}_{-35}^{+54}$
REBELS-39	6.847	$-{1.96}_{-0.28}^{+0.30}$	${8.56}_{-0.57}^{+0.57}$	${3.58}_{-0.37}^{+0.17}$	37 ± 6	${52}_{-20}^{+30}$
REBELS-40	7.365	$-{1.44}_{-0.36}^{+0.29}$	${9.48}_{-0.99}^{+0.45}$	${2.98}_{-0.20}^{+0.32}$	17 ± 1	${35}_{-14}^{+20}$

Notes.

^a Estimated using beagle (Chevallard & Charlot 2016) assuming a constant star formation history (CSFH). See M. Stefanon et al. (2022, in preparation) for details. Note that alternate estimates of the EWs are provided assuming a delayed star formation history in Table 4 of Appendix B. ^b Derived from the measured UV luminosity using the prescription given in Section 4.2. ^c Derived from the measured IR luminosity L_IR (Inami et al. 2022) using the prescription given in Section 4.2. Note that no use of this SFR is made for Figure 5. Instead, for that figure, SFR_IR is taken to be (10^0.19 − 1)SFR_UV as found for the average luminous z ∼ 7 source by Schouws et al. (2022a). ^d Estimates of the stellar mass presented in Fudamoto et al. (2021) for REBELS-12 and REBELS-29 are higher by ∼0.7–1.0 dex than those quoted here using BEAGLE and assuming a constant star formation history, and are more in line with stellar mass estimates from Prospector relying on nonparametric star formation histories (M. Stefanon et al. 2022, in preparation; Topping et al. 2022). Estimates for the obscured SFR estimates from Fudamoto et al. (2021) are almost identical to what we present here, while the UV SFRs are ∼0.1 dex higher.

Download table as:  ASCIITypeset image

In Figure 4, we present the UV luminosities M_UV of sources in the 40 source REBELS selection versus their photometric redshift. For context, the light blue circles are included to show the UV luminosities and redshifts of sources from various legacy fields covering an area of ∼0.3 deg² (Bouwens et al. 2021). The light green hexagons indicate the UV luminosities and redshifts for other z > 6 sources with spectroscopic redshift determinations from ALMA. The solid and open blue circles are sources with redshift measurements from ≥7σ and <7σ Lyα lines, respectively (Vanzella et al. 2011; Schenker et al. 2012; Shibuya et al. 2012; Ono et al. 2012, Pentericci et al. 2011; Jiang et al. 2013; Oesch et al. 2015; Sobral et al. 2015; Zitrin et al. 2015; Song et al. 2016; Hoag et al. 2017; Stark et al. 2017; Larson et al. 2018; Pentericci et al. 2018; Fuller et al. 2020; Jung et al. 2020; Endsley et al. 2021a; Laporte et al. 2021; Pelliccia et al. 2021).

Figure 5 shows the approximate SFRs of targets in the REBELS program versus their inferred stellar masses (solid red circles). These results are shown (shaded gray region) relative to the main sequence at z ∼ 7 implied by the Speagle et al. (2014) star-forming main-sequence fitting formula ${\mathrm{log}}_{10}\mathrm{SFR}({M}_{* },t)\,=(0.84\pm 0.02-(0.026\pm 0.003)t){\mathrm{log}}_{10}{M}_{* }-(6.51\,\pm \,0.24\,-(0.11\pm 0.03)t)$ where t is the age of the universe in Gyr. Also shown for comparison is the main-sequence relation shifted to match the specific star formation results from Stefanon et al. (2022) at z ∼ 7.8 (shaded blue region). Overall, the REBELS targets are distributed both above and below the main-sequence relation (particularly that derived by Stefanon et al. 2022). Our assuming a constant star formation history may result in a significant underestimate of the stellar masses for sources with bursty star formation histories (Topping et al. 2022). Topping et al. (2022) will provide an improved discussion of where the REBELS targets fall on the SFR versus stellar mass relation.

Zoom In Zoom Out Reset image size

Figure 17 from Appendix B shows the distribution of the REBELS targets in stellar mass and compares this distribution with that inferred (Faisst et al. 2020) for the z = 4–6 ALPINE sample. The REBELS sample spans a range in stellar mass similar to that of ALPINE. The median stellar masses for the REBELS sample inferred from the beagle and Prospector stellar population fitting codes (assuming constant and nonparametric star formation histories, respectively) are 10^9.25 M_⊙ and 10^9.79 M_⊙, ∼0.4 dex lower and ∼0.1 dex higher than inferred for ALPINE. Stellar mass estimates for sources in our sample from Prospector will be presented in M. Stefanon et al. (2022, in preparation). Thanks to the similarities between the samples, ALPINE provides us with a convenient z = 4–6 comparison sample for assessing evolution in the galaxy population with redshift.

The distribution of REBELS targets in redshift, UV luminosity, UV-continuum slope β, and [O iii]_4959,5007+Hβ EW is illustrated in Figure 18 of Appendix B. The latter two distributions appear to be completely consistent with that found for the z ∼ 7 galaxy population as a whole, suggesting that results derived from REBELS should be representative of the general population of massive star-forming galaxies at z > 6.5. Despite the similar stellar mass characteristics of the two samples, REBELS does not include targets that extend as faint in UV luminosities as ALPINE does (as Figure 4 makes clear).

Galaxies with the most luminous ISM reservoirs shine very brightly in both the 157.74 μm [C ii] and 88.36 μm [O iii] ISM-cooling lines, and both lines are readily detectable with ALMA over a significant fraction of the redshift range between z ∼ 6 and z ∼ 10. As a result, ALMA has already been successful in scanning for the [C ii]_{158 μm} line in five sources at z ∼ 6–7 (Smit et al. 2018; Schouws et al. 2022b) as well as uncovering serendipitous galaxies in the immediate neighborhood of z ∼ 6–7 QSOs (Decarli et al. 2017; Venemans et al. 2019, 2020). Similarly, spectral scans for the [O iii]_{88 μm} line have been shown to allow for efficient redshift determinations for galaxies at z > 8 (e.g., Hashimoto et al. 2018; Tamura et al. 2019).

As such, we could potentially target either transition in searching for bright ISM-cooling lines in z > 6 galaxies. To help decide which line would be more efficient, we calculated the approximate break-even luminosity ${L}_{{[{\rm{O}}\,{\rm\small{III}}]}_{88\,\mu {\rm{m}}}}$/${L}_{{[{\rm{C}}\,{\rm\small{II}}]}_{158\,\mu {\rm{m}}}}$ ratio for [O iii]_{88 μm} versus [C ii]_{158 μm} spectral scans as a function of redshift. In calculating this ratio, we expressly made use of the atmospheric transmission, using Equation (9.8) from the ALMA Technical Handbook ²⁷ and the specified inputs to this equation from chapter 9 of the handbook to derive sensitivities for a given integration time (Cortes et al. 2020). Observations were assumed to be conducted in a precipitable water vapor (PWV) octile consistent with that shown in Figure 9.1 of the ALMA Technical Handbook, i.e., with band 6 observations being conducted in the fifth octile weather and band 7 observations being conducted in generally better weather but depending on the observational frequency. As a check on the results from our sensitivity calculations, use was made of the ALMA sensitivity calculator. Also, account was made for the line scans with [O iii]_{88 μm} covering a 1.8× smaller range in Δz than [C ii]_{158 μm} (due to the higher frequency of [O iii]_{88 μm}). The FWHM of the [C ii]_{158 μm} and [O iii]_{88 μm} lines was assumed to be the same, i.e., 250 km s⁻¹, for this calculation. Such an FWHM is fairly typical for ISM-cooling lines found for UV-bright z > 6 galaxies (e.g., Matthee et al. 2019; Harikane et al. 2020; Schouws et al. 2022a) and is consistent with theoretical expectations (Kohandel et al. 2019, 2020). We emphasize that other FWHMs should yield essentially identical results.

The results are shown in Figure 6. For the typical luminosity ratios observed for z ∼ 6–8 galaxies, i.e., ∼4 (e.g., Bakx et al. 2020; Carniani et al. 2020; Harikane et al. 2020; Inoue et al. 2016), not only is [C ii]_{158 μm} clearly the most efficient line to use for our spectral scans in the redshift range z ∼ 6.0 to z ∼ 8.5, but also is not subject to significant gaps in redshift coverage. For sources at z > 8.5, by contrast, [O iii]_{88 μm} is the most efficient line to use for spectral scans. The [O iii]_{88 μm} scans do become significantly less efficient at a redshift z ∼ 9.4 due to the H₂O 5₁₅ − 4₂₂ line in the Earth's atmosphere at 325 GHz, but [C ii]_{158 μm} is also not especially efficient at that redshift, due to the H₂O 3₁₃ − 2₂₀ line at 183 GHz.

Zoom In Zoom Out Reset image size

Here, we describe our strategy for setting up the tunings for our spectral scans. For each target, we aim to optimally cover the redshift likelihood distribution derived combining the results from three independent sets of photometry (Figure 2).

For bright sources in the redshift range z = 6.5–7.2, the redshifts of sources can be accurately estimated thanks to the Lyman-break and high-EW [O iii]_4959,5007 doublet lying very close to edge of multiple spectral elements (z_Subaru, z_HSC, Y_VISTA, y_HSC, [3.6], and [4.5]) with deep coverage over the COSMOS/UltraVISTA, VIDEO/XMM-LSS, and UKIDSS/UDS fields. For z = 6.5–7.2 sources over the COSMOS/UltraVISTA field, the median Δz uncertainty is ± 0.11 while over the VIDEO/XMM-LSS + UKIDSS/UDS fields, the median Δz uncertainty is ± 0.15. For most z = 6.5–7.2 sources over the COSMOS/UltraVISTA field, we find that we can use [C ii]_{158 μm} scans to cover ∼89% of the likelihood distribution with two redshift tunings, each covering a contiguous 5.375 GHz frequency range (and a 10.75 GHz contiguous range in total). REBELS-08 in Figure 15 of Appendix A is one example of a source where we used this tuning strategy.

From z = 7.2 to z = 7.7, the redshifts of sources are largely constrained based on the position of the Lyman break in the Y-band filter. Nevertheless, since there are fewer passbands with flux measurements providing direct constraints on the redshifts of sources, the median uncertainty on the redshifts of sources is larger, with a value of ±0.15. For sources in this redshift range, we find that we can use [C ii]_{158 μm} scans to cover the 91% of the redshift likelihood distribution in three tunings covering a contiguous frequency range of 20.375 GHz (5.375 + 7.5 + 7.5 GHz). REBELS-01 in Figure 15 of Appendix A is one example of a source where we used this tuning strategy.

For sources with photometric redshifts in excess of 7.7, the redshift uncertainties increase substantially, due to the Lyman break falling between the ground-based Y and J bands. For these sources, the redshift likelihood distribution typically extended from z ∼ 7.7 to 9.3, with a median uncertainty of ±0.44. To execute a spectral scan for the ISM-cooling line over this range, we made use of [C ii]_{158 μm} scans in cases where such a scan had already begun as part of a cycle-6 program (2018.1.00236.S, PI: Stefanon) probing the dust continuum. Otherwise, sources required six tunings in band 7 to search for [O iii]_{88 μm} (probing the redshift range z = 8.10 to 9.39) and approximately three tunings in band 6 to search for [C ii]_{158 μm} (probing the range z = 7.37–8.00). ²⁸

We present the spectral scan windows we use for the 40 targets in the REBELS LP in Figures 15 and 16 of Appendix A. In total, 16 of the targets from the program required 2 tunings to cover the redshift likelihood distribution, 17 targets required 3 tunings, 1 target required 4 tunings, 2 targets required 5 tunings, 1 target required 6 tunings, and 1 target required 8 tunings. For 2 targets, only 1 tuning was allocated to extend scans that had already started in our second pilot program (Schouws et al. 2022b). In total, the number of targets × tuning windows for the REBELS program is equal to 91 for the [C ii]_{158 μm} line and 22 for the [O iii]_{88 μm} line. Archival ALMA observations contributed an additional 12 tunings to our [C ii]_{158 μm} scans (see Appendix A).

Our sensitivity requirements for REBELS relies on our experience with searches in our pilot programs (Smit et al. 2018; Schouws et al. 2022b). There, the detected [C ii]_{158 μm} lines had peak fluxes of 1.5–4.0 mJy. To guarantee the selection of similar sources at z ∼ 7 with REBELS, we required that the peak flux sensitivity be ∼340 μJy in a 66 km s⁻¹ channel such that sources with a peak flux of 1 mJy can be detected at 5σ when combining multiple channels. This is equivalent to a 5σ limiting point-source luminosity of ∼ 3 × 10⁸ L_⊙ at z ∼ 7 (assuming a line width of 250 km s⁻¹), which requires ∼20 minutes of integration time per tuning. Reaching the same limiting luminosity at z ∼ 8 requires a 300 μJy sensitivity and ∼30–40 minutes of integration time in band 5 or 6 (for z > 8 and z < 8 [C ii]_{158 μm} searches, respectively). In scanning for the bright [O iii]_{88 μm} lines in our z ∼ 9 targets, the equivalent sensitivity requirement is 450 μJy, assuming an [O iii]_{88 μm}/[C ii]_{158 μm} luminosity ratio of 3.5.

For the purposes of illustration, the limiting [C ii]_{158 μm} and [O iii]_{88 μm} luminosities probed as a function of source redshift for the spectral scans planned for sources in REBELS are provided in Figure 7. This is for the requisite sensitivity specified for the REBELS program. In practice, the sensitivity achieved is typically ∼1.4× better than specified (S. Schouws et al. 2022, in preparation), allowing us to detect even lower-luminosity ISM-cooling lines (indicated with the lighter shading in Figure 7). For context, we have included the [C ii]_{158 μm} luminosities and redshifts of detected sources from the literature (gray circles) and our pilot programs (filled stars). As should be clear from the figure, REBELS will detect line emission from sources if their [C ii]_{158 μm} luminosities exceed ∼ 3 × 10⁸ L_⊙ and ∼2 × 10⁸ L_⊙ (requested and typical sensitivities, respectively), and for z > 8.5 targets, if their [O iii]_{88 μm} luminosities exceed ∼ 1.1 × 10⁹ L_⊙ and 0.8 × 10⁹ L_⊙ (required and typical sensitivities, respectively).

Zoom In Zoom Out Reset image size

To maximize the sensitivity of the REBELS spectral scan observations and not overresolve the [C ii]_{158 μm} line, observations were conducted in the lowest spatial resolution configurations (C43-1 and C43-2), with ∼12–16 FWHM. The REBELS observations were obtained in frequency domain mode (FDM) at a spectral resolution of 488 MHz and then spectrally averaged in bins of size 16, giving the output data a spectral resolution of 7.813 MHz, equivalent to a velocity resolution of ∼9 km s⁻¹ for [C ii]_{158 μm} line at z ∼ 7. Given that [C ii]_{158 μm} lines in luminous z ∼ 7 galaxies have been found to have a minimum FWHM of ∼80 km s⁻¹ and more typically 250 km s⁻¹, this was expected to be more than sufficient to study the kinematic structure of sources revealed by the program.

Of the z ∼ 7 galaxies targeted with observations from our pilot programs, 63% yielded lines with luminosities in excess of 2 × 10⁸ L_⊙. Given that similar selection criteria are used for the REBELS large program, we would expect a similar detection rate of [C ii]_{158 μm} for the REBELS program, suggesting we will detect lines in 25 out of 40 targets. Combining the expected results with previous [C ii]_{158 μm} and [O iii]_{88 μm} line detections, we expected ≥35 z > 6.5 galaxies with ALMA line detections once the program was completed. This is a sufficient number for a detailed physical characterization of massive galaxies at z ≳ 6.5 and a study of their evolution.

The detection of dust-continuum emission from star-forming galaxies at z ∼ 5–8 with ALMA has been found to be much more difficult than detection of the [C ii]_{158 μm} or [O iii]_{88 μm} ISM-cooling lines. This basic difference in detectability was already evident in z ∼ 5.5 results by Capak et al. (2015), who were able to detect all 10 luminous z = 5.2–5.7 galaxies they targeted in [C ii]_{158 μm}, but were only able to detect 4 of the 10 sources in the dust continuum. Other results available on z > 5 galaxies have been similar, with [C ii]_{158 μm} being detected in a much larger fraction of sources than the dust continuum (e.g., Maiolino et al. 2015; Willott et al. 2015; Inoue et al. 2016; Matthee et al. 2017, 2019; Béthermin et al. 2020; Harikane et al. 2020).

The relative detectability of ISM-cooling lines like [C ii]_{158 μm} or [O iii]_{88 μm} and the dust continuum can be quantified on the basis of their measured ${L}_{{[{\rm{C}}\,{\rm\small{II}}]}_{158\,\mu {\rm{m}}}}$ -to-L_IR ratios or ${L}_{{[{\rm{O}}\,{\rm\small{III}}]}_{88\,\mu {\rm{m}}}}$ -to-L_IR ratios. Compiling previous results from Capak et al. (2015), ALPINE (Béthermin et al. 2020; Faisst et al. 2020; Le Fevre et al. 2020), Smit et al. (2018), Schouws et al. (2022a), and Schouws et al. (2022b), we can calculate the relative integration time required to detect sources in the dust continuum and the integration time required to detect sources in [C ii]_{158 μm}. In performing this calculation, we assume a line width of 235 km s⁻¹ (FWHM) for both the [C ii]_{158 μm} and [O iii]_{88 μm} lines, and we assume a modified blackbody form for the dust-continuum SED with a dust temperature of 50 K and dust emissivity index β of 1.6, which lies intermediate between the lowest and highest dust temperature measurements at z > 7 (e.g., Knudsen et al. 2017; Bakx et al. 2020). We present results in Figure 8.

Zoom In Zoom Out Reset image size

Given the much longer exposure times required to detect sources in the dust continuum than in either [C ii]_{158 μm} or [O iii]_{88 μm}, our use of multiple tuning windows for spectral scans really does provide an advantage in allowing us to probe the dust continuum. We have illustrated the approximate IR luminosities we are able to probe with the REBELS LP in Figure 9. Two spectral scan windows (10.75 GHz bandwidth) are assumed at z = 6.5–7.2, three spectral scan windows (20.375 GHz bandwidth) at z = 7.2–7.7, and six spectral scan windows (45 GHz bandwidth) at z > 7.7.

Zoom In Zoom Out Reset image size

The REBELS dust-continuum probe also allows for a very valuable assessment of incompleteness in our spectral scan results. This is because incompleteness can arise as a result of (1) scan range not extending over a wide enough range in frequency to find the relevant ISM-cooling line or (2) the relevant ISM-cooling line being fainter than the 5σ detection limit for the spectral scan. Given the strong correlation between dust-continuum luminosities of galaxies and the luminosities of [C ii]_{158 μm} and [O iii]_{88 μm}, the detection of the dust continuum in a source strongly suggests that the associated ISM cooling is sufficiently bright to have been detected in our spectral scans. If the ISM line is not found but the dust continuum is, it strongly suggests the spectral scan did not extend broadly enough in frequency. In cases where neither the line nor the dust continuum is detected, it may mean that the line is fainter than the sensitivity limits of the scans (or in the worst case at lower redshift, but our careful selection suggests that the number of such targets in the REBELS program is small).

The REBELS LP observations began on 2019 November 15, when ALMA was in the C43-2 configuration, and continued until 2020 January 10, while ALMA was in configurations C43-1 and C43-2. Thus far, 60.6 hr of ALMA observations have been acquired, with 8 hr remaining to be observed.

Thirty-four targets from the program have now been fully observed. Observations are still incoming for REBELS-04, REBELS-06, REBELS-11, REBELS-16, REBELS-24, and REBELS-37. Two of the sources with incoming observations are part of our 33-target z = 6.5–7.7 sample. The remaining four are part of our seven-target z = 7.7–9.4 sample. The REBELS spectral scans for our z = 6.5–7.7 and z = 7.7–9.4 targets are therefore 94% and 43% complete, respectively.

Following the first year of observations from the REBELS program and successful detection of [C ii]_{158 μm} in REBELS-18 and REBELS-36, several tunings from REBELS-18 and REBELS-36 were shifted to REBELS-06, REBELS-16, and REBELS-37 to extend the spectral scan range for [C ii]_{158 μm} in those sources.

The remaining observations appear likely to be executed in March 2022 when the configuration of ALMA is in C43-1 or C43-2, according to the current JAO schedule.

Here, we provide a very brief summary of the procedures used to reduce and calibrate ALMA observations from the REBELS program. More details will be given in S. Schouws et al. (2022, in preparation) and Inami et al. (2022).

Reduction and calibration of data from the program were performed using the standard ALMA calibration pipeline as implemented in the Common Astronomy Software Applications package (CASA) version 5.6.1. Data cubes were reimaged with the tclean task using a natural weighting to maximize our sensitivity for detecting ISM-cooling lines in ALMA band-5/6/7 observations obtained from the program. Cleaning was done down to 2σ in producing the data cubes.

Line searches were performed using three different line-search algorithms on the reduced data cubes. Searches for >5σ lines were performed within 1" of the target center in the rest UV. Line widths from 80 to 600 km s⁻¹ were considered in searching for lines throughout our data cubes. S. Schouws et al. (2022, in preparation) will provide a detailed description of our line search procedures and catalogs, while carefully quantifying both the completeness and purity of the line searches.

As an illustration of the effectiveness of the spectral scans employed in the REBELS program to date, we include in Figure 10 the 18 [C ii]_{158 μm} lines detected with a significance of >7σ. These lines are also presented in Table 3. Redshifts of the detected [C ii]_{158 μm} lines range from z = 6.496 to 7.677. The value of 7σ was adopted as the detection threshold in this paper to keep the focus on the brightest and most significant lines found in the survey. Details on the purity of our ISM line searches, characteristics of the [C ii]_{158 μm} line detections, as well as several additional lower S/N [C ii]_{158 μm} lines will be presented in S. Schouws et al. (2022, in preparation).

Zoom In Zoom Out Reset image size

Even if we consider only the >7σ line detections in the REBELS data obtained thus far, the program is already having a substantial impact on the number of spectroscopic redshifts derived with ALMA at z > 6.5. Figure 11 (top panel) shows how the sources with ≥7σ ISM-cooling lines are distributed as a function of redshift and UV luminosity. Also shown in the top panel of Figure 11 are sources from the two pilot programs to REBELS (Smit et al. 2018; Schouws et al. 2022a, 2022b).

Zoom In Zoom Out Reset image size

As the lower two panels of Figure 11 demonstrate, the number of spectroscopic redshift measurements from ALMA is already fairly similar to that available from similarly significant Lyα lines at z > 7, if one makes use of the most significant ISM-cooling line detections from the REBELS program and from the two pilot programs. This is especially the case for the brightest and most massive sources at z > 6.5, where the number of redshifts from ALMA already exceeds that available from Lyα by a factor of two.

Simultaneous with the REBELS spectral scans for bright ISM-cooling lines, we are able to probe dust continuum emission from sources in REBELS. Figure 12 shows the continuum observations available for the 18 targets from REBELS with ≥7σ ISM-cooling line detections, and it is clear that the majority of these [C ii]_{158 μm} -detected sources (13/18) show nominal 3.3σ detections in the dust continuum. These dust-continuum images were generated using the tclean task in CASA. Any channels containing emission from the [C ii]_{158 μm} line were excluded when producing the continuum maps. More details on our current dust-continuum results from REBELS will be presented in Inami et al. (2022).

Zoom In Zoom Out Reset image size

We can already use the results from the REBELS first-year data to assess the effectiveness of the spectral scan strategy. Perhaps the most relevant variable to utilize in gauging the effectiveness of the scans is in terms of the SFRs of individual targets in our program.

For our SFR estimates for individual sources in the REBELS program, we make use of the luminosities of sources in both the rest-UV and far-IR continuum. The unobscured SFR_UVs for sources are estimated based on source UV luminosities using the relation SFR_UV = 7.1 × 10⁻²⁹ L_ν [erg s⁻¹ Hz⁻¹], while the obscured SFR_IRs of sources are estimated based on source IR luminosities using the relation SFR_IR = 1.2 × 10⁻¹⁰ L_IR /L_⊙. The inferred SFRs are ∼14% lower for a given UV/IR luminosity than the calibrations adopted for ALPINE (e.g., Schaerer et al. 2020). A detailed motivation and discussion of these relations will be provided in M. Stefanon et al. (2022, in preparation), Topping et al. (2022), and Inami et al. (2022).

In computing the obscured SFR_IR of sources, the IR luminosities of sources L_IR are taken to equal 14.2${}_{-4.7}^{+7.6}$ ν L_ν where ν is the frequency of the [C ii]_{158 μm} line (Inami et al. 2022; Sommovigo et al. 2022). Using this scaling, sources in REBELS have L_IR luminosities ranging from 3 × 10¹¹ L_⊙ to 1 × 10¹² L_⊙ (equivalent to obscured SFRs of 36 M_⊙ yr⁻¹ to 120 M_⊙ yr⁻¹, respectively). The precise scaling we use here is equivalent to that found from a modified blackbody SED with a dust temperature of 50 K and a dust opacity index β of 1.6. This is a slightly higher dust temperature than ALPINE use in the analysis of their z = 4–6 sample (Béthermin et al. 2020), but lower than is found by Bouwens et al. (2020) in modeling the dust temperature versus redshift measurements available at the time. It is also consistent with the general range of dust temperatures found by semi-analytical models (Sommovigo et al. 2022) and numerical simulations (R. Schneider et al. 2022, in preparation) for REBELS-like sources. A more extensive motivation for these conversion factors will be provided in Inami et al. (2022).

Using the above scalings to estimate the SFRs of individual sources, we present the number of sources showing prominent ISM-cooling line detections as a function of the total SFR of sources in Figure 13. We only include in the analysis the 32 sources for which our ISM-cooling line scans are complete and which appear to be securely at z > 6. ²⁹ Since only one of our [O iii]_{88 μm} spectral scans is complete to present in the first-year data, it makes sense to frame these search results in terms of our spectral scans for [C ii]_{158 μm}.

Zoom In Zoom Out Reset image size

Looking over the results, our spectral scans for [C ii]_{158 μm} show a dramatic increase in efficiency above 28 M_⊙ yr⁻¹. Fifteen of the 19 sources with SFRs in excess of 28 M_⊙ yr⁻¹, i.e., 79%, show prominent 7σ ISM-cooling lines in the observations taken to date. Meanwhile, below an SFR of 28 M_⊙ yr⁻¹, only three of 13 sources show a prominent 7σ [C ii]_{158 μm} detection. We note that the 7σ limiting luminosity for [C ii]_{158 μm} scans in the REBELS program (Figure 7), 2.8 × 10⁸ L_⊙ yr⁻¹, is equivalent to an SFR threshold of 28 M_⊙ yr⁻¹, using the z ∼ 0 relation from De Looze et al. (2014). This result is suggestive of only minimal evolution in the [C ii]_{158 μm}–SFR relation to z ∼ 0 (see also Matthee et al. 2017, 2019; Carniani et al. 2018, 2020; Harikane et al. 2020; Schaerer et al. 2020).

We also note that some sources with SFRs higher than 28 M_⊙ yr⁻¹ remain undetected in [C ii]_{158 μm} , while some sources below 28 M_⊙ yr⁻¹ are detected. This is suggestive of there being some intrinsic scatter in the ${L}_{{[{\rm{C}}\,{\rm\small{II}}]}_{158\,\mu {\rm{m}}}}$ versus SFR relation at z ∼ 7, similar to the 0.27 dex scatter observed at z ∼ 0 by De Looze et al. (2014). We will further characterize both the evolution and the scatter in the ${L}_{{[{\rm{C}}\,{\rm\small{II}}]}_{158\,\mu {\rm{m}}}}$ versus SFR relation in S. Schouws et al. (2022, in preparation).

It is also interesting to present the efficiency of our [C ii]_{158 μm} scans in terms of the unobscured SFRs for sources. Figure 14 shows the number of sources where [C ii]_{158 μm} is prominently detected (>7σ: red histogram) as a function of the unobscured SFRs, and also the number of sources where no prominent [C ii]_{158 μm} emission is found (gray histogram). While the [C ii]_{158 μm} detection fraction does show some dependence on the unobscured SFR_UVs, the dependence is significantly less steep than on the total SFR_UV+IRs (Figure 13). This demonstrates the essential value ALMA observations have for characterizing the ISM of star-forming galaxies at z > 6.5.

Zoom In Zoom Out Reset image size

We emphasize that only the most prominent [C ii]_{158 μm} detections from the REBELS program are included here in evaluating the efficiency of the spectral scans. There are indeed a larger number of [C ii]_{158 μm} detections in REBELS, but these detections require a more careful demonstration of their robustness. These line detections will be presented in detail in S. Schouws et al. (2022, in preparation).