Discovery of a dark, massive, ALMA-only galaxy at z~5-6 in a tiny 3-millimeter survey
Christina C. Williams, Ivo Labbe, Justin Spilker, Mauro Stefanon, Joel Leja, Katherine Whitaker, Rachel Bezanson, Desika Narayanan, Pascal Oesch, Benjamin Weiner
DDraft version September 25, 2019
Typeset using L A TEX twocolumn style in AASTeX62
Discovery of a dark, massive, ALMA-only galaxy at z ∼ − Christina C. Williams,
1, 2
Ivo Labbe, Justin Spilker, Mauro Stefanon, Joel Leja,
6, 2
Katherine Whitaker, Rachel Bezanson, Desika Narayanan, Pascal Oesch,
10, 11 and Benjamin Weiner Steward Observatory, University of Arizona, 933 North Cherry Avenue, Tucson, AZ 85721, USA NSF Fellow Centre for Astrophysics & Supercomputing, Swinburne University of Technology, PO Box 218, Hawthorn, VIC 3112, Australia Department of Astronomy, University of Texas at Austin, 2515 Speedway, Stop C1400, Austin, TX 78712, USA Leiden Observatory, Leiden University, NL-2300 RA Leiden,Netherlands Harvard-Smithsonian Center for Astrophysics, 60 Garden St. Cambridge, MA 02138, USA Department of Physics, University of Connecticut, 2152 Hillside Road, Unit 3046, Storrs, CT 06269, USA Department of Physics and Astronomy and PITT PACC, University of Pittsburgh, Pittsburgh, PA, 15260, USA Department of Astronomy, University of Florida, 211 Bryant Space Science Center, Gainesville, FL 32611, USA Department of Astronomy, University of Geneva, 51 Ch. des Maillettes, 1290 Versoix, Switzerland International Associate, Cosmic Dawn Center (DAWN) at the Niels Bohr Institute, University of Copenhagen and DTU-Space,Technical University of Denmark
ABSTRACTWe report the serendipitous detection of two 3 mm continuum sources found in deep ALMA Band3 observations to study intermediate redshift galaxies in the COSMOS field. One is near a foregroundgalaxy at 1 . (cid:48)(cid:48)
3, but is a previously unknown dust-obscured star-forming galaxy (DSFG) at probable z CO = 3 . λ − . attenuation curve and results in significantlylarger stellar mass and SFR compared to a Calzetti starburst law, suggesting caution when relatingprogenitors and descendants based on these quantities. The other source is missing from all previousoptical/near-infrared/sub-mm/radio catalogs (“ALMA-only”), and remains undetected even in stackedultradeep optical ( > . > . SN R ∼ µ m, andVLA 3GHz, indicating the source is real. The SED is robustly reproduced by a massive M ∗ = 10 . M (cid:12) and M gas = 10 M (cid:12) , highly obscured A V ∼
4, star forming
SF R ∼
300 M (cid:12) yr − galaxy at redshift z = 5 . ± survey area implies a large yet uncertain contribution to thecosmic star formation rate density CSFRD(z=5) ∼ . × − M (cid:12) yr − Mpc − , comparable to allultraviolet-selected galaxies combined. These results indicate the existence of a prominent populationof DSFGs at z >
4, below the typical detection limit of bright galaxies found in single-dish sub-mmsurveys, but with larger space densities ∼ × − Mpc − , higher duty cycles 50 − INTRODUCTIONIn past decades, single dish sub-millimeter surveyshave identified populations of massive, dusty star-forming galaxies at z > < z < z > z >
4, but they trace only the very tipof the star-formation rate (SFR) distribution at early times (e.g. Cooray et al. 2014; Strandet et al. 2017;Marrone et al. 2018). The total contribution of dust-obscured star-formation, and therefore the census ofstar-formation in the early Universe, is unknown. De-spite the strongly negative k-correction allowing sourcesto be found to z = 10, the overwhelming majority of(sub)-mm selected galaxies continue to be confirmed at z < z > a r X i v : . [ a s t r o - ph . GA ] S e p Progress is hampered by the limited sensitivity and lowspatial resolution of single-dish sub-mm observationsand the difficulty of associating detections with coun-terparts in the optical-NIR. Ultradeep SCUBA surveysover moderate ∼
100 arcmin are now pushing intothe range of “normal” star formation rates (several 100M (cid:12) /yr, main sequence galaxies) (e.g., Koprowski et al.2016; Cowie et al. 2017; Cowie et al. 2018) and extend-ing to z >
4, but the analysis is often limited by theability to identify counterparts at other wavelengths andderive accurate redshifts.ALMA has opened an avenue to address this issuethrough surveys at superior sensitivity and spatial reso-lution. ALMA deep fields at ∼ < ). Progress has stillbeen limited likely because dust obscured star forma-tion preferentially occurs in massive galaxies (Whitakeret al. 2017), which are clustered and relatively rare ( ∼ − at log( M/M (cid:12) ) > z ∼
4; Davidzonet al. 2017). Wider (10’s arcmin ) and shallower ( ∼ µ Jy) ALMA surveys at ∼ z > > z > ∼ , and identified 6 continuum sources, all at z <
3. Larger archival studies of ALMA 3 mm observa-tions to find high-redshift candidates report some spec-troscopic confirmations, but like the 1-mm redshift dis-tribution, the majority lie at z < z > z >
3. We assumea ΛCDM cosmology with H =70 km s − Mpc − , Ω M =0.3, Ω Λ = 0.7, and a Kroupa (2001) initial mass function(IMF). METHODS2.1.
ALMA millimeter interferometry
The ALMA observations are part of a programtargeting CO(2-1) line emission in z ∼ . ∼
24 km s − ) channeliza-tion. Three additional 1.875 GHz bandwidth spectralwindows were placed at sky frequencies 96.8, 106.9, and108.8 GHz for continuum observations, each with 15.6MHz channelization. A total of 43 antennas were activereaching maximum baselines of 500m, for an angularresolution of ∼ µ Jy/beam at 101.9 GHz, and a typi-cal line sensitivity of 55-65 µ Jy/beam per 100 km s − channel. We imaged the data using natural weighting,creating a 101 GHz continuum image of the field fromall four spectral windows. We imaged the data using0.2” pixels and created a 500 ×
500 pixel image, yieldingimages 100” on a side. Given the ALMA primary beamat this frequency ( ∼
57” FWHM), these images extendto approximately the 0.05 response point of the primarybeam. 2.2.
ALMA source detection
Interferometric maps without correction for the pri-mary beam response have uniform, normally-distributednoise properties across the field, and source detectionsignificance is straightforward to measure from suchmaps. Two blind 3 mm continuum sources were appar-ent in this map, located 24.6” and 38.2” from the phasecenter, corresponding to primary beam response levelsof 0.57 and 0.29, respectively. Each source is detectedat a peak signal-to-noise ratio of ∼
8; the probability offinding a Gaussian noise fluctuation of this magnitudegiven the number of independent beams in the imagesis exceedingly low ( < − ). Both sources thus are real. ALMA 3mm Subaru Vgri UltraVISTA YJHKs IRAC 3.6 mIRAC 4.5 m MIPS 24 m SCUBA-2 850 m VLA 3 GHz
Figure 1.
Cutouts (25 (cid:48)(cid:48) x25 (cid:48)(cid:48) ) centered at 3-mm position of 3 MM -1 (red circle; 3 (cid:48)(cid:48) diameter). 3 MM -1 was not previouslydetected ( > σ ) at any shorter wavelength, including deep optical and near-IR stacks, Spitzer, Herschel , and S2COSMOSSCUBA2 850 µ m. Remeasuring with the ALMA position as prior reveals marginal 2 − σ measurements in IRAC 3.6+4.5 and850 µ m, consistent with heavy dust obscuration at z >
4. It is faintly detected at 3GHz (4 σ ) indicative of a moderate radioexcess due to a possible AGN. The 1.4 GHz image is excluded because neither source is significantly detected. 3 MM -2 is alsoidentified (blue circle), and is blended with a foreground galaxy at z = 0 .
95, 1.3” North (Muzzin et al. 2013; Laigle et al. 2016).
The 3 mm flux densities of these sources, corrected forthe primary beam response, are 155 ± µ Jy and 75 ±
10, hereafter referred to as 3
M M -1 and 3
M M -2, re-spectively. Neither source is spatially resolved, based ona comparison of the peak pixel values and the integratedflux densities.After finding both continuum sources in the combinedmap, we re-imaged the upper and lower sidebands of thedata separately in order to determine the spectral indexof each source at these frequencies. Thermal dust emis-sion on the Rayleigh-Jeans tail has a very steep power-law index with S ν ∝ ν β and β ∼ . −
2, while non-thermal synchrotron emission typically exhibits a nega-tive spectral index, S ν ∝ ν − . . Given their respectivefrequencies, we expect dust emission to be ∼ − − . 3 M M -2 contains a serendipitousemission line centered at 106.5 GHz, with an integratedflux density of 0.66 ± − . The spectrum isshown in Figure 2. A Gaussian fit indicates in a widthof 630 ±
70 km/s. Assuming the line is a transition of carbon monoxide, the possible redshifts are z =[0.08,1.16, 2.25, 3.33, 4.41, 5.49]. We find no significant emis-sion lines in the spectrum of 3 M M -1, and no evidencefor a line at the same frequency of 3
M M -2. A Gaussianfit restricted to the same frequency and width resultsin an integrated flux density of 0 . ± . − ,indicating no evidence for a line at that location.Both 3 M M -1 and 3
M M -2 were also contained withinthe field-of-view of an additional ALMA Band 3 pro-gram, 2015.1.00861.S. Both sources were again far outin the primary beam of these data, at approximately the0.25 and 0.1 response points, respectively. These dataare described in more detail in Silverman et al. (2018),and have non-overlapping frequency coverage with ourown data. We downloaded and imaged these data fol-lowing the same procedure as for our own data. Thenew images reach a continuum sensitivity at phase cen-ter of 10 µ Jy/beam at 93.5GHz and a line sensitivityof 120-170 µ Jy/beam per 100 km s − channel, approxi-mately a factor of two higher than in our data. Neithersource is detected in continuum in these data, and werenot expected to be detected given the sensitivity, effec-tive frequency, and position of our sources within theALMA primary beam. We additionally searched thesedata for blindly-detected CO lines as in our own data,but found no significant emission lines. The limited (pri-mary beam-corrected) sensitivity of these data preclude F l u x D e n s i t y [ m J y b e a m ] COS-3mm-2fitCO(4-3) at z=3.329800060004000200002000 Velocity offset [km/s] from z=3.3290.50.00.51.01.52.0 F l u x D e n s i t y [ m J y b e a m ] COS-3mm-1
Figure 2.
Portion of the observed ALMA Band 3 spectrum covering the detected CO line in 3 MM -2 at 106.5 GHz (upper sideband). The CO solution corresponding to CO(4-3) at z=3.329 is in excellent agreement with the photometric redshift measuredin Section 2.4. The line flux is 0.66 ± − and a Gaussian fit produces a width of 650 km/s. No line is found at thesame frequency in the spectrum of 3 MM -1 (lower panel) with formal SNR = 0 .
8. No lines are detected in the lower side bandfrom 94 < ν < . us from drawing strong conclusions about the redshiftsof either source.For convenience, we define an equivalent survey areaas the total area across all five ALMA maps at whicha source with the same S as 3 M M -1 would be de-tected at > σ . Taking into account the small varia-tions in the central frequency of each map (tuned to thespecific redshift of the target z ∼ µ Jy) due to the steep spectralindex of dust emission, we derive a total survey area of8.0 arcmin to 155 µ Jy (5 σ limit).2.3. Multi-wavelength photometry M M -1 has extremely deep coverage at all optical-to-sub-mm wavelengths, yet it has no counterpart withina radius of 3.3 arcseconds in the deepest publishedmultiwavelength catalogs in COSMOS to date, from0 . − µ m (Laigle et al. 2016; Muzzin et al. 2013; Aretxaga et al. 2011; Geach et al. 2017; Hurley et al.2017; Le Floc’h et al. 2009) or 3 − . M M -2is also missing from these catalogs, likely because it isblended with a bright neighboring galaxy ∼ . (cid:48)(cid:48) Ks and IRAC. It is possible thatthe ALMA source is simply a highly obscured region inthe low-redshift galaxy, which can be ruled out by spa-tially deblended SED analysis. We therefore proceeded original 4.5 μ m deblended 4.5 μ m 3.6 μ m 4.5 μ m COS-3mm-2COS-3mm-1
Figure 3.
Illustration of the deblending procedure using the mophongo software (Labb´e et al. 2015). Deblending is performedby simultaneously fitting the pixels of all sources using the deep optical images and the ALMA positions as priors and accountingfor differences in the PSFs. Top row: Deblending results for the 4.5 µ m band centered on the ALMA position of 3 MM -2(12 (cid:48)(cid:48) × (cid:48)(cid:48) ). Left panel shows the original 4.5 µ m image, middle panel shows 3 MM -2 after other modeled sources have beensubtracted, and the right panel a 3.6 µ m − . µ m color image clearly indicating a vastly different IRAC color at the location ofthe ALMA source. Bottom row: Higher contrast, zoomed-in panels centered on the ALMA position of 3 MM -1, showing a faint AB ∼ . to perform deblended photometry on both ALMA posi-tions using the following data sets.2.3.1. Optical, near-infrared, and Spitzer/IRAC
The optical data consists of the Subaru/Suprime-Cam B j , V j , g + , r + , i + and z + -imaging (Taniguchi et al.2007), with 5 σ limits of ∼ − . . (cid:48)(cid:48) g , r , i , z and y (Aihara et al. 2018a,b) imaging ( ∼ − . σ ). Ultradeep NIR coverage is provided by the 4th datarelease of the UltraVISTA survey (McCracken et al.2012), thanks to mosaics in the Y , J , H and K s filtersto ∼
25 mag (AB, 5 σ ). Remarkably, the coverage inthe K s band from DR4 is ∼ . M M -1 at NIR wavelengths. Stacked imageswere constructed using the optical imaging 0 . − . . − . Spitzer /IRAC 3 . µ m and 4 . µ m mosaics thatcombine data from the S-COSMOS (Sanders et al. 2007)and the Spitzer
Large Area Survey with HSC (SPLASH, PI: Capak) programs ( ∼ . σ in 1 . (cid:48)(cid:48) . µ m and 8 . µ m from the S-COSMOS program( ∼ . σ in 1 . (cid:48)(cid:48) M M -1 in the opticaland UltraVISTA bands in 1 . (cid:48)(cid:48) mophongo (Labb´eet al. 2013, 2015). This procedure carefully models thelight profiles of the sources using a higher resolution im-age as a prior, minimizing potential contamination bybright nearby objects (see Figure 3). In our analysis weadopted the HSC z -band image as prior for 3 M M -1 asit provides the best compromise between depth and res-olution and the F814W-band image for 3
M M -2, giventhe nearby bright neighbor at ∼ . (cid:48)(cid:48) . (cid:48)(cid:48) Far-infrared to submillimeter
Far-infrared and sub-millimeter
Herschel fluxes weremeasured for both galaxies by simultaneously fittingGaussian profiles at fixed prior locations in the
Herschel images specified by the ALMA locations and augmentedby the MIPS positions of all neighboring objects fromthe S-COSMOS data (Sanders et al. 2007; Le Floc’het al. 2009). The FWHM of the Gaussians were 7.7,12” (PACS 100, 160 µ m) and 18, 25 and 37” (SPIRE250, 350, 500 µ m), respectively. Uncertainties were com-puted by fitting gaussians at random locations withina 2 arcmin radius and computing the rms. Flux cali-bration was performed by comparing to the 24 µ m priorcatalog of Herschel
DR4 (Oliver et al. 2012; Hurley et al.2017). Comparison of the measured SPIRE fluxes withthose published showed a scatter of 30%; this was addedin quadrature to the flux uncertainties for those bands.The correction is minor considering the signal to noiseratio of (cid:46) µ monly, the flux of a neighboring z spec = 1 .
45 galaxy 5” tothe south was subtracted separately prior to deblend-ing, using its predicted IR emission based on the in-frared SED of Wuyts et al. (2011) and its SFR ,IR =65M (cid:12) /yr.The procedure for measuring photometry at 850 µ mfollows that of Herschel /PACS and SPIRE (250 and350 µ m) using the ALMA and MIPS positions as pri-ors. The sources lie in a region of shallower ( σ ∼ σ ∼ . Radio VLA 3Ghz and 1.4 Ghz
Neither source has a counterpart in either the VLA 3GHz 5 σ source catalog (Smolˇci´c et al. 2017) or the 1.4GHz deep survey 5 σ source catalog (Schinnerer et al.2010). Photometry on the 3GHz map using the ALMAposition as prior reveals 9.98 ± µ Jy/beam (4 σ ) pointsource for 3 M M -1, below the detection limit of theSmolˇci´c et al. (2017) catalog. Blindly detected sourcesat this flux density have a high probability ( > M M -2shows no significant flux to 3 σ limits of ∼ µ Jy/beam.Within 10 arcmin , the 1.4 GHz map has rms ∼ µ Jy/beam and no significant flux ( > σ ) is detectedfrom either galaxy.2.4. Spectral Energy Distribution Modeling
The deep photometry from λ = 0 . − µ m placesstrong constraints on the SEDs and allows us to model their stellar populations and redshift. It should be notedthat even if a source is not formally detected ( > σ ) atmost wavelengths, absence of flux can still provide usefulconstraints, in particular on the redshift. All fitting isperformed in linear fluxes and uncertainties and no up-per limits are enforced on measurements with low SNR.We use the Bayesian Analysis of Galaxies for PhysicalInference and Parameter EStimation ( bagpipes ) code(Carnall et al. 2018), which assumes the stellar popula-tion synthesis models of Bruzual & Charlot (2003) andimplements nebular emission lines following the method-ology of (Byler et al. 2017) using the cloudy photoion-ization code (Ferland et al. 2017). We select the flex-ible dust absorption model of Charlot & Fall (2000),with the exponent of the effective absorption ∝ λ − n as a free parameter and adopt the Draine & Li (2007)dust emission model under the assumption of energy bal-ance, such that dust-absorbed light is re-radiated in thefar-infrared. All stellar populations have this effectiveabsorption, while the youngest stars (defined as thosewith age < η ) to account for dusty birthclouds.The dust emission model is parameterized using thestarlight intensity U incident on the dust (translatinginto a distribution of dust temperatures), the amount ofPAH emission, and the fraction of dust at the coldesttemperature. We further assume a delayed exponen-tial star-formation history with τ and age left free andmetallicities ranging from 0 . − . Z (cid:12) . For each param-eter, uniform, diffuse priors are assumed (see Table 1).In general, we do not expect to constrain most free pa-rameters, but we will marginalize over those and limitthe discussion to the more important parameters includ-ing redshift, stellar mass, and SFR. Where relevant, wewill also quote results based on assumptions often usedin the literature, such as assuming a fixed Calzetti et al.(2000) dust attenuation model. RESULTSThe photometry for both galaxies is presented in theAppendix (Table 2). 3.1. 3
M M -1 M M -1 is undetected ( < σ ) in any individual opti-cal ( > > . > .
2) bands, and remains undetected in the stackedultradeep optical ( > . > . σ limits). The source is faintly detected (25.2AB ∼ σ ) in deep stacked IRAC 3.6+4.5 observations,indicating the source is likely real with extremely redcolors. Owing to the shallower depth, the source is un-detected in deep Spitzer/MIPS 24 micron and Herschel Observed Wavelength [ m]10 F l u x D e n s i t y [ J y ] COS-3mm-1
Median posterior modelModel + S
S/N < 3.0S/N 3.0 P ( z ) Observed Wavelength [ m]10 F l u x D e n s i t y [ J y ] COS-3mm-2
Median posterior model (CF00)Model + S
CalzettiS/N < 3.0S/N 3.0 P ( z ) Figure 4.
Left: Observed photometry of 3 MM -1 (points). For display purposes, we show 1 σ upper limits (downwardarrows) for data with SNR <
SNR > σ uncertainties but does not necessarily indicate a significantdetection. Light blue squares indicate photometry with SNR <
3, and dark blue circles indicate detections with SNR ≥
3. Shownare the median posterior spectrum (dark orange) and 16-84th percentile range (light orange) from Bayesian SED fitting with bagpipes assuming the Charlot & Fall (2000) dust model. Insets contain the posterior redshift distribution and the 16, 50, and84th percentiles (black lines). The deep FIR photometric limits at 24 < λ < µ m favor high redshift (90% probability at z > . ∝ ν − . expected from the total LIR (Tisani´c et al. 2019), suggesting a 3Ghz radio excess. Right: SED of 3 MM -2. As in the left panel, any photometry with SNR > σ uncertainties. The photometric redshift agrees with CO(4 −
3) at z CO =3.329 (red line). The SED fit (orange) assumes theCharlot & Fall (2000) dust model and the redshift fixed to z CO . Also shown is a fit assuming a Calzetti et al. (2000) starburstattenuation law (blue curve) a common assumption in high-redshift studies, which can drastically change the estimated stellarmass, LIR, and SFR. The blue curve has 8 × lower stellar mass and 2 × lower SFR, while still adequately fitting the observations. PACS+SPIRE 160 −
500 micron. The flux measuredwith SCUBA is S =3.5 ± SN R = 3 . > (cid:38) σ ). The observed SED is shown in Figure 4 (leftpanel). For clarity, photometry with S/N < σ rms level, and S/N > σ errorbars. The deep non-detections at allwavelengths between 24 − µ m and the extreme fluxratios ( S /S < × − , S /S mm < × − , S . /S < × − , S . /S mm < × − , see e.g.,Cowie et al. 2018; Yamaguchi et al. 2019), demand thatdust peak is highly redshifted z > . ± . M (cid:12) ,star forming SF R = 309 +241 − M (cid:12) /yr, highly obscured A V ∼ +1 . − . galaxy at very high redshift z ∼ . +1 . − . .The SED fitting results are shown in Figure 4 and theposterior values and priors are listed in Table 1. Theposterior distributions of all parameters are presented in Figure 7. Adopting a classical Calzetti dust law wouldincrease the stellar mass, SFR, and redshift by 60%,10%, and 5% respectively. These changes are within theuncertainties, so we elect to adopt our more conserva-tive values measured assuming Charlot & Fall (2000)as fiducial. We also experimented with fitting only themid-infrared to sub-millimeter SED, finding the sameresults. In all cases, the solutions require the galaxy tobe at high redshift, with z > . L IR = 4 × L (cid:12) (based on in-tegrating the median posterior spectrum). Adopting anSED typical for the most obscured A V > . (cid:12) /yr, but consistent within the 1 σ uncertainties from SED fitting.Calculating redshift using the radio-to-submm spec-tral index method (using 850 µ m and the upper limiton 1.4GHz) implies a lower limit to the photometricredshift of ∼ Observed Wavelength [ m]10 F l u x D e n s i t y [ J y ] COS-3mm-1
Median posterior modelModel + S
Arp 220 at z=5.5Arp 220 at z=3.329HDF850.1 at z=5.18S/N < 3.0S/N 3.0
Figure 5.
Comparison between the SED of 3 MM -1 with that of both HDF850.1 and Arp220. SED fitting results for 3 MM -1as in the left panel of Figure 4. The SED of Arp220 scaled to the 850 µ m and 3-mm photometry of 3 MM -1, at the photometricredshift of 3 MM -1, is remarkably similar to the 3 MM -1 SED. Forcing Arp220 to the redshift of 3 MM -2 violates the HerschelFIR+sub-mm constraints. The SED of 3 MM -1 is very similar to that of the bright z spec =5.18 dusty galaxy HDF850.1 (usingthe SWIRE SED of IRAS 20551-4250 as in Serjeant & Marchetti 2014) scaled to match the average observed 850 µ m and 3-mmflux). that this is very similar to the massive dusty galaxyHDF850.1 which had a redshift 4.1 according to this re-lation (Dunlop et al. 2004), and was later spectroscopi-cally confirmed at z=5.2 (Walter et al. 2012). The SEDof 3 M M -1 is very similar to that of HDF850.1; scal-ing its best-fit SED to the 850 µ m and 3-mm fluxes of3 M M -1 predicts its observed 3 GHz radio flux of 10 µ m(Figure 5).The observed 3 GHz radio flux (4 σ ) is in excess of thatexpected from empirical SED templates for obscuredstar-forming galaxies (e.g. SMGs; da Cunha et al. 2015)and recent calibrations of the redshift-dependent ratioof total infrared luminosity to rest-frame 1.4 GHz lu-minosity density (Delhaize et al. 2017; Tisani´c et al.2019). For typical assumptions of the radio spectralslope S ν ∝ ν − . (dashed line in Figure 4), the radioemission of 3 M M -1 is a factor of ∼ ± . × M M -1. Given the possible presence of an AGN, it is worthnoting that inferred parameters from SED fitting couldbe biased by AGN emission, in particular in the mid-infrared (e.g. Leja et al. 2018). Therefore, we investigateany possible contamination using the empirical AGN-dominated template published by (Kirkpatrick et al.2015). We use their mid-infrared template based ongalaxies with the largest mid-infrared luminosity con-tribution from AGN emission scaled to the 24 µ m flux of3 M M -1 and subtract the predicted AGN contributionfrom the observed SED. This represents the maximumAGN contribution that could be accommodated by thedata without violating the observed photometry. Wethen re-measure the SED fitted parameters. Subtract-ing this empirical AGN template reduces the SFR by ∼ M M -1 following the calibra-tion of sub-mm flux to gas mass (Scoville et al. 2016).Exploring the range of the photometric redshift PDF,and typical assumptions about the dust emissivity, tem-perature, and dust-to-gas ratio, we find that 3
M M -1 Table 1.
Stellar Population Properties of 3-mm continuum sourcesParameter 3 MM -1 a MM -2 3 MM -2 Range (uniform prior)Attenuation Curve CF00 CF00 CalzettiCoordinates 10:02:36.82 +02:08:40.60 10:02:36.30 +2:08:49.55Redshift 5.5 +1 . − . z CO =3.329 z CO =3.329 (0.0, 10.0)SFR [M (cid:12) yr − ] 309 +241 − +77 − +27 − Log Stellar Mass [M (cid:12) ] 10.8 +0 . − . +0 . − . +0 . − . (6.0, 14.0)Mass-weighted age [Gyr] 0.2 +0 . − . +0 . − . +0 . − . A V +1 . − . +0 . − . +0 . − . (0.0, 10.0) η c +0 . − . +0 . − . +0 . − . (1, 3.0)Umin d +7 . − . +6 . − . +3 . − . (1, 25.0)gamma e +0 . − . +0 . − . +0 . − . (0.01, 0.99)n f +0 . − . +0 . − . - (0.1, 2.0)qpah g +1 . − . +0 . − . +0 . − . (0.5, 4.0)Log LIR h [L (cid:12) ] 12.6 12.4 12.1Gas mass [M (cid:12) ] 0.5-1.5 × − × a Fitting priors are uniform with the range as defined in last column. b Redshift prior is a gaussian matching the redshift evolution at z > c Multiplicative factor producing extra attenuation for young stars d Starlight intensity on dust grains, related to dust temperature as T dust ∝ U / (Draine & Li 2007) e Fraction of stars at Umin f Power-law slope of the dust extinction law (for Charlot & Fall 2000) g PAH mass fraction h Total infrared luminosity (8-1000 µ m) in units from integrating the median posterior spectrum in Figure 4. likely has M gas ∼ . − . × M (cid:12) . This is indepen-dent evidence that the galaxy is massive, with a highinferred gas fraction ( ∼ M M -2 M M -2 is optically faint, but detected in K s,AB =24 . ∼
23 mag, withred optical-to-IRAC colors. There is an apparentbreak between the J s and H -band, consistent with aBalmer/4000˚A break at z ∼
3. The source is unde-tected at 24 −
870 micron, but with less extreme fluxratios ( S /S mm < . × − , S . /S mm < × − )compared to 3 M M -1, indicating a lower redshift.We follow the same procedure as before to fit theSED, finding a well-constrained photometric redshift of z = 3 . ± . . < z < . z = 0 .
95 foreground galaxy to thenorth. Considering the CO emission line detection (con-sistent with redshifts 2.25, 3.33, and 4.41), we determinethe line is likely CO(4 −
3) at z = 3 . . M (cid:12) , star forming SF R = 250 M (cid:12) /yr, highly obscured A V ∼ . ∼
375 M (cid:12) yr − . Thederived gas mass using the CO(4-3) line flux (assumingthe average CO excitation from Bothwell et al. 2013) isin the range ∼ − × M (cid:12) , given the factor of fourscatter in measured excitation for high-redshift dust ob-scured galaxies (e.g. Figure 1 in Narayanan & Krumholz2014). The 3 mm derived gas mass is 4 − × M (cid:12) . Both estimates indicate gas fractions in the range30 − M M -2 are listed inTable 1, the SED fit is shown in Figure 4, and posteriordistributions for the parameters are presented in Figure8. The best-fit power law index for the Charlot & Fall(2000) dust model is n = 0 . ± .
1, flatter than the n = 0 . E ( B − V ) = 0 .
3. The as-sumed attenuation model can have a large impact on thederived stellar mass. If instead of a flexible attenuationmodel a classical Calzetti starburst law is assumed thenthe fits result in a factor 8 +7 . − . lower stellar mass (10 . M (cid:12) ) and factor 2 . ± . × lower SFR (same answer if theinfrared-to-mm constraints are also removed). The largedifference in stellar mass is remarkable. The reason isthat the steeper reddening curve of Calzetti reproducesthe data with an intrinsically blue, young, low M/L ratiostellar population, driving the lower masses. The Char-lot & Fall (2000) model requires an older, redder, higherM/L population, while also implying larger attenuationat all wavelengths, including the near-infrared. We notethat these results all assume energy balance and a para-metric delayed − τ SFH. We explore the dependence onSFH by also fitting the SED with prospector (Lejaet al. 2017), which is capable of fitting a non-parametricSFH in logarithmic bins of age (Leja et al. 2019). Thesefits produce qualitatively similar results, with Calzettiyielding ∼ × lower SFR and stellar mass.An independent estimate of mass can be derived byestimating the dynamical mass from the 630 km/s widthof the CO emission line at 106.5 GHz (Figure 2). Fol-lowing the procedure used by Wang et al. (2013), wecompute V circ = 0 . F W HM ( CO ) /sin ( i ) and M dyn =1 . × V circ D , where we note that inclination angle i and disk diameter D (in kpc) are unknown. Adopt-ing the mean size R e = 3 +1 . − . kpc expected for massive10 M (cid:12) star forming galaxies at z = 3 (van der Welet al. 2014), we find M dyn sin ( i ) = 1 . +1 . − . × M (cid:12) .Assuming a disk geometry with an inclination angle of i = 45 ◦ would imply M dyn ∼ × M (cid:12) . Overall,the high dynamical mass appears to agree better withthe high stellar mass inferred from (Charlot & Fall 2000)dust model with flat n than with the Calzetti-based stel-lar mass, but the large uncertainties in dynamical andstellar mass prevents firmer conclusions.3.3. Redshifts
Given the close r ∼ (cid:48)(cid:48) separation on the sky ofthe two ALMA sources we consider the possibility thatthey are at the same redshift. The redshift of 3 M M -2 ( z = 3 . M M -1 (1%probability z < . z CO (4 − = 3 .
329 for 3
M M -2 is erroneous and thegalaxy is at z CO (5 − = 4 .
41, more consistent with3
M M -1. This is unlikely based on the SED fit of 3
M M -2, which is well constrained by the presence of both a Lyman Break and Balmer break. We find no sig-nificant emission line (
SN R >
3) in the spectrum of3
M M -1 which suggests it is unlikely at 3 . < z < . . < z < .
90. A Gaussian fit to the spectrumof 3
M M -1 at fixed 106 . SN R = 0 . M M -1 and 3
M M -2 occu-pying the same dark matter halo (DMH). Additionally,we scan the nearby velocity space in case both galaxiesare in a large scale structure filament, but we find noemission line
SN R > v = ± r ∼ (cid:48)(cid:48) on the sky. Using the 3-mm numbercounts of Zavala et al. (2018a), we simulate uniformlyrandom distributions of two sources which indicates an ∼
10% chance of finding a 75 µ Jy source within r < (cid:48)(cid:48) of a 155 µ Jy source. These odds are low, but not negli-gible. Overall, there is no conclusive evidence that thetwo sources are at the same redshift. We therefore pro-ceed and take the z phot = 5 . ± . M M -1 at facevalue. DISCUSSION4.1.
Implications from full optical-to-mm spectrumSED fitting
We consider the SED fitting analysis using a stellarpopulation model with self-consistent dust absorptionand emission under the assumption of energy balanceand constrained by high quality optical-to-millimeterobservations. In the case of the z ∼ . M M -1,which lacks strong detections at any wavelength otherthan 3 mm, it is remarkable that the Bayesian posteriorprobability distributions for key parameters such as red-shift, stellar mass, dust attenuation and star-formationrate are as constrained as they are (see Table 1 andFigure 7). Closer inspection indicates that the resultsare mostly driven by the combination of the high SNRALMA measurement with deep photometric limits inthe short wavelength infrared (
Spitzer /MIPS and
Her-schel /PACS 100 − µ m), which demand that the dustemission peak is highly redshifted ( z >
4) and the SFRis high. It is likely that the uncertainties on the stellarmass and other stellar population parameters are un-derestimated, because of the tight prior imposed by theparametric SFH (Carnall et al. 2019; Leja et al. 2019).We note that our choice of a rising SFH produces arelatively conservative (lower) estimate of stellar masscompared to constant or declining SFH.There is accumulating evidence that the attenuationlaw in very dusty galaxies can be flatter than Calzetti(Salmon et al. 2016; Leja et al. 2017; Salim et al. 2018),possibly caused by a more uniformly mixed geometryof both old and young stars with dust (Narayanan1et al. 2018). The flat inferred attenuation law λ − . for 3 M M -2 is not surprising in that regard. It is no-table and sobering however that modeling the optical-to-millimeter with a classical Calzetti starburst law in-stead (keeping everything else the same) results in signif-icantly lower SFR and stellar mass (10 . ), in apparenttension with the high dynamical mass (1 . − × M (cid:12) ).Detailed analyses of ULIRGs at z ∼ M M -1, which likely represents a population that isabsent from all current optical-IRAC selected galaxystudies, is below the nominal detection threshold ofdeep single-dish sub-mm surveys, and therefore deservesmore scrutiny. 4.2.
Number counts
A recent unbiased ALMA 3 mm archival searchfor continuum sources over 130 independent pointings(Zavala et al. 2018a) enabled the first 3 mm numbercounts. For archival searches, however, detection limitsand effective search area are necessarily very inhomoge-neous due to the strong variation of the ALMA primarybeam response, complicating a straightforward compar-ison of results. Restricting the comparison to 3
M M -1,we use their best-fit powerlaw to the cumulative num-ber counts, which considers effective selection area andincompleteness, predicting one source brighter thanS mm > µ Jy per 16 +16 − arcmin . This is consistentwith our finding one source in our effective survey areaof 8 arcmin (see § ∼ . M (cid:12) at z=5.5, we expect 3 M M -1 to occupya massive DMH ( ∼ M (cid:12) ) and to be strongly clus-tered. While our results are dominated by Poissonianuncertainties, this may imply that the uncertainties inpublished number counts are underestimates. Cluster-ing may affect source counts from studies over verysmall areas such as the ASPECS 4.6 arcmin surveyin the HUDF (Gonz´alez-L´opez et al. 2019), and eventhe counts in the multiple pointings of Zavala et al.(2018a), as these observations were generally targeting moderately-sized deep fields (COSMOS, CDFS, UDS).Clustering effects may be exacerbated by the fact thatdust obscuration preferentially occurs in massive galax-ies (Whitaker et al. 2017). We note that our results areunlikely to be biased by the original primary targets,which were all at low redshift ( z ∼ . Unbiased ALMA 3 mm selection ofhigh-redshift galaxies
Simulations of dust obscured galaxies in the early uni-verse predict 3 mm continuum selection optimizes thesensitivity to DSFGs at redshift z > z = 2 . z = 3 . z >
3, due tothe faintness of Lyman-alpha, nebular optical lines andlack of wavelength coverage > . z CO = 3 .
329 (based on a single CO line andcongruous z phot = 3 . ± .
15) and z phot = 5 . ± . z = 4 . z = 2 . µ m-selectedgalaxies (Simpson et al. 2014).A challenge in determining the selection function isthe difficulty in identifying counterparts and determin-ing redshifts. Neither of our two 3 mm sources hadcounterparts in previous deep optical-to-radio selectedcatalogs, raising some concern for analyses where thisis a critical step (e.g., photometric redshifts or spectro-scopic follow up in optical/NIR). This is particularlyproblematic for single-dish sub-mm observations, due tothe large beam size, but here we find it challenging evenwith the high spatial resolution F W HM = 2 (cid:48)(cid:48) and accu-rate astrometry offered by ALMA. In the case of 3
M M -2 the source is very close to a bright foreground galaxy(1.3” to the North), which was initially mistaken as thecounterpart. The optical/NIR faintness and low reso-lution of the Spitzer/IRAC data caused it to be miss-ing or blended in existing multiwavelength catalogs. Inthe case of 3
M M -1 the combination of high redshiftand high obscuration resulted in it being extremely faintand undetected at all optical-infrared wavelengths. It is2therefore possible that the identification of the highestredshift galaxies in existing (sub-)mm selected samplesis still incomplete. In addition, obtaining spectroscopicredshifts in the optical/NIR is unfeasible for the faintestsources. ALMA spectral scans targeting CO and [CII]are likely the only recourse until other facilities (e.g.,JWST/NIRSpec, LMT Redshift Search Receiver) be-come available.4.4.
Comparison to other “dark” galaxies
The observed SED of 3
M M -1 is different from thatof other optical/near-IR/IRAC-faint populations. TheALESS survey identified 9 IRAC-faint SMGs to 1 mag-nitude brighter IRAC limits (Simpson et al. 2014), butthese sources were much brighter in the far-infrared(peaking at 250 −
350 micron to ∼ z ∼ − M M -1 is fainter at all wavelengths than the so-calledH/K − dropout galaxies (e.g. Wang et al. 2016; Schreiberet al. 2017; Caputi et al. 2012), which are generally se-lected on H − . > × brighter in IRAC, andestimated to be at lower redshift ¯ z = 3 . ∼ . S = 2 − M M -1 at z > z > −
25 AB,no MIPS detection, and only a marginal 3 Ghz radiodetection, 3
M M -1-like objects would have likely defiedsecure identification, and its probable z = 5 − S . /S . mm and S /S flux ratios (indicativeof high redshift), but are generally fainter S . mm =0 . − . M M -1 ( S . mm ∼ . Contribution to CSFRD
The fact that 3
M M -1 was identified in a survey ofsuch small area suggests that similar galaxies are com-mon in the early Universe. Using the effective area of 8arcmin as derived in § . +0 . − . arcmin − , an order of magnitudehigher than the rare sub-mm selected starbursts at z > . − .
02 arcmin − ; e.g. Danielson et al. 2017; Mar-rone et al. 2018). If our results are representative, galax-ies like 3 M M -1 could represent the “iceberg under thetip” of the known extreme dust-obscured star-forminggalaxies in the early Universe.To estimate the space densities and contribution to theCSFRD requires estimating a selection volume, whichis difficult as we do not know the expected propertiesor true abundance of DSFGs at z >
4. Instead, we de-rive a simple estimate based on the derived properties of3
M M -1. Encouraged by the strong redshift constraintsfrom the long wavelength photometry, we set the lowerbound of the selection volume to the lower 10th per-centile redshift probability ( z > . HMF published by Murray et al. (2013) and assumingthe halo mass function of Behroozi et al. (2013b) and ahigh 30% baryon conversion into stars and (convertingM ∗ =10 . M (cid:12) into a conservative M halo ∼ M (cid:12) ) wecompute this upper bound to be z = 5 . . +6 . − . × − Mpc − . We find that the contribu-tion to the CSFRD by 3 M M -1 is ρ SFR . +2 . − . × − M (cid:12) yr − Mpc − (converted to a Chabrier 2003 IMF forcomparison with literature measurements). Assuminginstead the cumulative space density of massive halos( > ) at z ∼ ρ SFR . × − M (cid:12) yr − Mpc − for our derived SFR.With only a single object, the Poissonian uncertaintyis large and dominant (Gehrels 1986). Cosmic varianceis only on the order of ∼
30% based on the calculatorby Trenti & Stiavelli (2008), and is therefore not furtherincluded. No completeness corrections are applied, be-cause the true distribution is unknown. Figure 6 showscomparisons to various results at 0 < z <
10 from litera-ture that report dust-uncorrected UV-derived SFR andthe dust-obscured SFR (IR to millimeter derived).The contribution of 3
M M -1 is higher than inferredfor the two near-infrared dark ALMA 1.2 mm sources inYamaguchi et al. (2019), mostly owing to their smallerimplied total infrared luminosities. This study does notreport formal uncertainties or derive redshifts, whichprohibits a more quantitative comparison. Bright SMGs3 L o g S F R ( M y r M p c ) MD14 (UV+IR SFRD)Koprowski et al. 2017Yamaguchi et al. 2019Dust-Poor (Casey et al. 2018)Dust-Rich (Casey et al. 2018)COS-3mm-1Michalowski et al. 2017Magnelli et al. 2019Cowie et al. 2018Swinbank et al. 2014Dunlop et al. 2017Liu et al. 2018MD2014 IRMD2014 UV-dust-uncorrFinkelstein et al. 2015Bouwens16, T d = 35KMcLeod et al. 2016Oesch et al. 2017 Figure 6.
The cosmic star-formation history of the Universe. Blue shades are UV (dust un-corrected) and red shades areIR-to-millimeter derived SFRs. We add to the compilation of Madau & Dickinson (2014, ; blue circles) with more recent z > z > MM -1 is indicated by the black star and shaded region, where the redshift range indicates the estimatedselection volume as discussed in § beyond z > ∼ × − M (cid:12) yr − Mpc − (Swinbank et al. 2014; Micha(cid:32)lowski et al. 2017), aboutan order or magnitude lower than our best estimate. Re-sults from fainter SMGs found in the deepest SCUBAsurveys, with luminosities similar to 3 M M -1 (e.g. Ko-prowski et al. 2017; Cowie et al. 2018) show a decliningcontribution at 2 < z <
5, consistent with our estimates.Interestingly, if 3
M M -1 is representative, a popula-tion with similar properties could contribute as much tothe CSFRD as all known ultraviolet-selected galaxies atsimilar redshifts combined. This could even imply thatdust-obscured star-formation continues to dominate thecosmic star-formation history beyond z >
Contribution to Stellar mass density
The high stellar mass 10 . M (cid:12) and large space den-sity ∼ × − Mpc − of 3 M M -1 imply a considerablecosmic stellar mass density (CSMD) in similar objectsat z ∼ ρ ∗ = 1 . +4 . − . × M (cid:12) Mpc − , higher thanreported for bright (S > ≈ . × M (cid:12) Mpc − ; Micha(cid:32)lowski et al. 2017). Com-paring to the estimate of the CSMD based on HST- selected galaxies ρ ∗ ( z = 5) = 6 . × M (cid:12) Mpc − (Songet al. 2016), suggests that they could contribute a sig-nificant fraction (22 +25 − %) to the total and perhaps evendominate the high-mass end. The high-mass end of thegalaxy stellar mass function at high redshift z > ∼ . M (cid:12) galaxies at z ∼ − × − Mpc − dex − (Duncan et al. 2014;Grazian et al. 2015; Song et al. 2016; Davidzon et al.2017; Stefanon et al. 2017), comparable to the space den-sity derived for our ALMA-only galaxy 3 M M -1. Giventhat optical-IRAC dark galaxies are missing from theseprevious studies, it is therefore possible that about halfthe stellar mass density in high-mass galaxies at z ∼ Implications for the formation of massive galaxies
Bright sub-millimeter-selected galaxies at z > (cid:38) (cid:12) yr − are often hypothesized as progeni-tors of massive z ∼ < z < z > ∼ − × − Mpc − and Log (M/M (cid:12) ) (cid:38) . > z > ∼ . − × − Mpc − (Ivison et al. 2016; Micha(cid:32)lowski et al. 2017; Jin et al.2018). While their high SFRs indicate they will rapidlyform massive galaxies, the modest inferred gas massesindicate gas depletion timescales on the order of 10 − M M -1 on the other hand are already as large as those re-ported for the population of early quiescent galaxies at3 < z < ∼ M (cid:12) gas mass and modest SFR indicatesmuch longer depletion timescales ( ∼ −
500 Myr),half the age of the Universe at this epoch, and implies a ∼ − (cid:46) z (cid:46) M M -1 could therefore represent a moregradual path for massive galaxy growth compared to arapid and bursty formation as has been found in somebright merger-induced SMGs (e.g. Pavesi et al. 2018;Marrone et al. 2018). Overall, our results provide evi-dence for the existence of a sustained growth mode formassive galaxies in the early Universe.Finally, the large systematic difference in SFR andstellar mass in particular depending on the assumed at-tenuation model for 3
M M -2 suggests exercising cautionwhen relating progenitors and descendants galaxies. Of-ten such links are determined based on the capabilityof a progenitor population to produce adequate num-bers of sufficiently massive galaxies at later times (e.g.,Straatman et al. 2014; Toft et al. 2014; Williams et al.2014, 2015), or by the assumption that the rank-orderon stellar mass can be reliably determined (e.g. Bram-mer et al. 2011; Behroozi et al. 2013a; Leja et al. 2013).Such determinations could be more uncertain than hasbeen accounted for if the derived stellar masses of indi-vidual galaxies are off by factors of ∼ § z >
3, where veryfew galaxies have high enough SNR (sub-)mm observa-tions for meaningful constraints on the dust attenuationmodel. These results are obviously based on only a sin-gle galaxy and larger samples with full optical-to-mm photometry are needed to determine the scatter in stel-lar mass.4.8.
Future ALMA and JWST observations
The dark nature of the 3
M M -1, non-detection in verydeep stacked optical and near-IR data and the faintIRAC fluxes, suggests a prominent population of DS-FGs at z >
4. The high inferred redshift may point tothe efficacy of blind surveys at 2 − > − James Webb Space Telescope ( JWST ),ALMA alone can find and study these galaxies.Wide-area unbiased ALMA surveys covering 100’s ofarcmin are necessary to further constrain their promi-nence in the early universe. Such surveys are feasible at2 − JWST will enable systematic stud-ies of large samples of faint SMG galaxies. Large legacysurveys such as the
JWST
Advanced Deep Extragalac-tic Survey (JADES) will likely characterize ∼ − M M -1 (based on expected numberdensities from this work and Zavala et al. 2018a, and ob-servations described in Williams et al. 2018).
JWST willhave the capability to measure stellar population prop-erties and redshifts, and in combination with ALMAfar-infrared constraints, will enable a detailed investiga-tion into the star-formation, dust, and stellar populationproperties of massive galaxies in the early Universe.We are grateful to James Simpson for making partof S2COSMOS available for our analysis in advance ofpublication. We thank Jorge Zavala, Peter Behrooziand Pieter van Dokkum for helpful discussions. C.C.W.acknowledges support from the National Science Foun-dation Astronomy and Astrophysics Fellowship grantAST-1701546. J.L. is supported by an NSF Astronomyand Astrophysics Postdoctoral Fellowship under awardAST-1701487. This paper makes use of the follow-ing ALMA data: ADS/JAO.ALMA
M M -1 and 3
M M -2. In Table 2, we present the deblendedphotometry (measurements and 1 σ uncertainties) based on the ALMA priors for the two 3-mm sources that are used inthe SED fitting analysis presented in Section 2.4. In Figures 7 and 8 we present the posterior probability distributionsfor the galaxy properties of both sources measured using the Bayesian SED fitting with bagpipes .REFERENCES Aihara, H., Arimoto, N., Armstrong, R., et al. 2018a,PASJ, 70, S4, doi: 10.1093/pasj/psx066Aihara, H., Armstrong, R., Bickerton, S., et al. 2018b,PASJ, 70, S8, doi: 10.1093/pasj/psx081Aravena, M., Decarli, R., Walter, F., et al. 2016, ApJ, 833,68, doi: 10.3847/1538-4357/833/1/68Aretxaga, I., Wilson, G. W., Aguilar, E., et al. 2011,MNRAS, 415, 3831,doi: 10.1111/j.1365-2966.2011.18989.xBehroozi, P. S., Marchesini, D., Wechsler, R. H., et al.2013a, ApJ, 777, L10, doi: 10.1088/2041-8205/777/1/L10Behroozi, P. S., Wechsler, R. H., & Conroy, C. 2013b, ApJ,770, 57, doi: 10.1088/0004-637X/770/1/57B´ethermin, M., De Breuck, C., Sargent, M., & Daddi, E.2015, A&A, 576, L9, doi: 10.1051/0004-6361/201525718Bezanson, R., Spilker, J., Williams, C. C., et al. 2019,ApJL, 873, L19, doi: 10.3847/2041-8213/ab0c9cBothwell, M. S., Smail, I., Chapman, S. C., et al. 2013,MNRAS, 429, 3047, doi: 10.1093/mnras/sts562Bouwens, R. J., Oesch, P. A., Labb´e, I., et al. 2016, ApJ,830, 67, doi: 10.3847/0004-637X/830/2/67Brammer, G. B., Whitaker, K. E., van Dokkum, P. G.,et al. 2011, ApJ, 739, 24,doi: 10.1088/0004-637X/739/1/24Brisbin, D., Miettinen, O., Aravena, M., et al. 2017, A&A,608, A15, doi: 10.1051/0004-6361/201730558Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000,doi: 10.1046/j.1365-8711.2003.06897.xByler, N., Dalcanton, J. J., Conroy, C., & Johnson, B. D.2017, ApJ, 840, 44, doi: 10.3847/1538-4357/aa6c66Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ,533, 682, doi: 10.1086/308692Caputi, K. I., Dunlop, J. S., McLure, R. J., et al. 2012,ApJL, 750, L20, doi: 10.1088/2041-8205/750/1/L20Carilli, C. L., & Yun, M. S. 1999, ApJL, 513, L13,doi: 10.1086/311909 Carnall, A. C., Leja, J., Johnson, B. D., et al. 2019, ApJ,873, 44, doi: 10.3847/1538-4357/ab04a2Carnall, A. C., McLure, R. J., Dunlop, J. S., & Dav´e, R.2018, MNRAS, 480, 4379, doi: 10.1093/mnras/sty2169Casey, C. M., Hodge, J., Zavala, J. A., et al. 2018, ApJ,862, 78, doi: 10.3847/1538-4357/aacd11Casey, C. M., Narayanan, D., & Cooray, A. 2014, PhR,541, 45, doi: 10.1016/j.physrep.2014.02.009Casey, C. M., Capak, P., Staguhn, J., et al. 2019, arXive-prints. https://arxiv.org/abs/1903.05634Chabrier, G. 2003, PASP, 115, 763, doi: 10.1086/376392Charlot, S., & Fall, M. 2000, ApJ, 539, 718,doi: 10.1086/309250Chevallard, J., Charlot, S., Wandelt, B., & Wild, V. 2013,MNRAS, 432, 2061, doi: 10.1093/mnras/stt523Civano, F., Marchesi, S., Comastri, A., et al. 2016, ApJ,819, 62, doi: 10.3847/0004-637X/819/1/62Cooray, A., Calanog, J., Wardlow, J. L., et al. 2014, ApJ,790, 40, doi: 10.1088/0004-637X/790/1/40Cowie, L. L., Barger, A. J., Hsu, L.-Y., et al. 2017, TheAstrophysical Journal, 837, 139,doi: 10.3847/1538-4357/aa60bbCowie, L. L., Gonz´alez-L´opez, J., Barger, A. J., et al. 2018,ApJ, 865, 106, doi: 10.3847/1538-4357/aadc63da Cunha, E., Walter, F., Smail, I. R., et al. 2015, ApJ,806, 110, doi: 10.1088/0004-637X/806/1/110Daddi, E., Dannerbauer, H., Stern, D., et al. 2009, ApJ,694, 1517, doi: 10.1088/0004-637X/694/2/1517Danielson, A. L. R., Swinbank, A. M., Smail, I., et al. 2017,ApJ, 840, 78, doi: 10.3847/1538-4357/aa6cafDavidzon, I., Ilbert, O., Laigle, C., et al. 2017, A&A, 605,A70, doi: 10.1051/0004-6361/201730419Delhaize, J., Smolˇci´c, V., Delvecchio, I., et al. 2017, A&A,602, A4, doi: 10.1051/0004-6361/201629430Draine, B. T., & Li, A. 2007, ApJ, 657, 810,doi: 10.1086/511055 Table 2.
ALMA prior-based deblended photometryBand 3 MM -1 flux density RMS 3 MM -2 flux density RMS[ µ Jy] [ µ Jy]SUBARU B [ -1.3E-02 ] 7.8E-03 – –HSC g [ 3.6E-03 ] 1.5E-02 [ 2.3E-02 ] 1.1E-02SUBARU V [ -7.6E-03 ] 1.8E-02 – –HSC r [ -1.2E-02 ] 1.4E-02 9.3E-02 9.7E-03SUBARU rp [ 1.2E-03 ] 1.7E-02 – –SUBARU ip [ 2.0E-03 ] 2.6E-02 – –HSC i [ -1.0E-02 ] 2.1E-02 1.2E-01 1.5E-02HSC z [ 1.2E-02 ] 3.0E-02 1.3E-01 2.4E-02SUBARU zp [ -1.5E-01 ] 6.5E-02 – –HSC Y [ 2.2E-02 ] 7.4E-02 [ 4.9E-02 ] 5.8E-02UltraVISTA Y [ 9.7E-03 ] 8.9E-02 [ 1.6E-01 ] 8.2E-02UltraVISTA J [ -1.1E-01 ] 1.0E-01 [ 2.0E-01 ] 8.9E-02UltraVISTA H [ -5.9E-02 ] 1.3E-01 6.3E-01 1.1E-01UltraVISTA Ks [ 1.2E-01 ] 8.8E-02 8.4E-01 8.7E-02
Spitzer/IRAC 3.6 µ m [ 3.2E-01 ] 1.2E-01 1.7E+00 9.2E-02 Spitzer/IRAC 4.5 µ m [ 3.2E-01 ] 1.3E-01 3.0E+00 9.9E-02 Spitzer/IRAC 5.8 µ m [ 5.2E+00 ] 3.3E+00 8.8E+00 2.2E+00 Spitzer/IRAC 8 µ m [ 6.0E+00 ] 4.1E+00 [ 7.7E+00 ] 2.9E+00 Spiter/MIPS 24 µ m [ 1.1E+01 ] 1.5E+01 [ 2.6E+01 ] 1.5E+01 Herschel/PACS 100 µ m [ 7.1E+02 ] 1.4E+03 [ 1.5E+02 ] 1.4E+03 Herschel/PACS 160 µ m [ -2.4E+03 ] 3.2E+03 [ -1.9E+03 ] 3.2E+03 Herschel/SPIRE 250 µ m [ 3.7E+03 ] 3.9E+03 [ 1.0E+03 ] 3.9E+03 Herschel/SPIRE 350 µ m [ 4.3E+03 ] 4.9E+03 [ -5.0E+02 ] 4.9E+03 Herschel/SPIRE 500 µ m [ 5.5E+03 ] 5.8E+03 [ 4.4E+03 ] 5.8E+03SCUBA2 850 µ m 3.5E+03 1.1E+03 [ 4.0E+02 ] 1.1E+03ALMA 3mm 1.6E+02 2.2E+01 7.5E+01 1.0E+01VLA 3 GHz 1.0E+01 2.4E+00 [ 3.9E+00 ] 2.4E+00VLA 1.4 GHz [ 1.7E+01 ] 1.7E+01 [ 4.0E+01 ] 1.7E+01 Note —Subaru optical photometry for 3 MM -2 are not measured owing to cosmetic defects in the mosaics that cross the locationof the source. Photometric upper limits as indicated by downward arrows in Figure 4 are at the RMS values of photometricpoints in this table where S/N <
1. Non-significant photometric points (with SNR <
3) are indicated with brackets.Duncan, K., Conselice, C. J., Mortlock, A., et al. 2014,MNRAS, 444, 2960, doi: 10.1093/mnras/stu1622Dunlop, J. S., McLure, R. J., Yamada, T., et al. 2004,MNRAS, 350, 769, doi: 10.1111/j.1365-2966.2004.07700.xDunlop, J. S., McLure, R. J., Biggs, A. D., et al. 2017,MNRAS, 466, 861, doi: 10.1093/mnras/stw3088Ferland, G. J., Chatzikos, M., Guzm´an, F., et al. 2017,RMxAA, 53, 385. https://arxiv.org/abs/1705.10877Finkelstein, S. L., Ryan, Jr., R. E., Papovich, C., et al.2015, ApJ, 810, 71, doi: 10.1088/0004-637X/810/1/71 Franco, M., Elbaz, D., B´ethermin, M., et al. 2018, A&A,620, A152, doi: 10.1051/0004-6361/201832928Geach, J. E., Dunlop, J. S., Halpern, M., et al. 2017,MNRAS, 465, 1789, doi: 10.1093/mnras/stw2721Gehrels, N. 1986, ApJ, 303, 336, doi: 10.1086/164079Glazebrook, K., Schreiber, C., Labb´e, I., et al. 2017,Nature, 544, 71, doi: 10.1038/nature21680Gonz´alez-L´opez, J., Decarli, R., Pavesi, R., et al. 2019,arXiv e-prints, arXiv:1903.09161.https://arxiv.org/abs/1903.09161 Figure 7.
Posterior probability distributions for the fitted parameters of 3 MM -1 corresponding to the orange region inFigure 4, assuming Charlot & Fall (2000) dust attenuation with the slope left free. M del refers to the total stars formed(integral of the delayed star formation history) whereas M ∗ is the stellar mass excluding remnants and including mass loss dueto stellar evolution, and is the parameter referred to throughout the text. Parameters, their definitions, and their uncertaintiesare presented in Table 1.Grazian, A., Fontana, A., Santini, P., et al. 2015, A&A,575, A96, doi: 10.1051/0004-6361/201424750Hatsukade, B., Kohno, K., Yamaguchi, Y., et al. 2018,PASJ, 70, 105, doi: 10.1093/pasj/psy104Hodge, J. A., Swinbank, A. M., Simpson, J. M., et al. 2016,ApJ, 833, 103, doi: 10.3847/1538-4357/833/1/103Hurley, P. D., Oliver, S., Betancourt, M., et al. 2017,MNRAS, 464, 885, doi: 10.1093/mnras/stw2375Ivison, R. J., Lewis, A. J. R., Weiss, A., et al. 2016, ApJ,832, 78, doi: 10.3847/0004-637X/832/1/78Jin, S., Daddi, E., Liu, D., et al. 2018, ApJ, 864, 56,doi: 10.3847/1538-4357/aad4af Kennicutt, R. C., & Evans, N. J. 2012, ARA&A, 50, 531,doi: 10.1146/annurev-astro-081811-125610Kennicutt, Jr., R. C. 1998, ApJ, 498, 541,doi: 10.1086/305588Kirkpatrick, A., Pope, A., Sajina, A., et al. 2015, ApJ, 814,9, doi: 10.1088/0004-637X/814/1/9Koekemoer, A. M., Aussel, H., Calzetti, D., et al. 2007,ApJS, 172, 196, doi: 10.1086/520086Koprowski, M. P., Dunlop, J. S., Micha(cid:32)lowski, M. J., et al.2017, MNRAS, 471, 4155, doi: 10.1093/mnras/stx1843—. 2016, MNRAS, 458, 4321, doi: 10.1093/mnras/stw564Kroupa, P. 2001, MNRAS, 322, 231,doi: 10.1046/j.1365-8711.2001.04022.x Figure 8.
Posterior probability distributions for the fitted parameters of 3 MM -2 corresponding to the orange region in Figure4, assuming Charlot & Fall (2000) dust attenuation with the slope left free, and fixed to the spectroscopic redshift. Measuredparameters, their definitions and their uncertainties are presented in Table 1.Labb´e, I., Oesch, P. A., Bouwens, R. J., et al. 2013, ApJL,777, L19, doi: 10.1088/2041-8205/777/2/L19Labb´e, I., Oesch, P. A., Illingworth, G. D., et al. 2015,ApJS, 221, 23, doi: 10.1088/0067-0049/221/2/23Laigle, C., McCracken, H. J., Ilbert, O., et al. 2016, ApJS,224, 24, doi: 10.3847/0067-0049/224/2/24Le Floc’h, E., Aussel, H., Ilbert, O., et al. 2009, ApJ, 703,222, doi: 10.1088/0004-637X/703/1/222Leja, J., Carnall, A. C., Johnson, B. D., Conroy, C., &Speagle, J. S. 2019, ApJ, 876, 3,doi: 10.3847/1538-4357/ab133cLeja, J., Johnson, B. D., Conroy, C., & van Dokkum, P.2018, ApJ, 854, 62, doi: 10.3847/1538-4357/aaa8dbLeja, J., Johnson, B. D., Conroy, C., van Dokkum, P. G., &Byler, N. 2017, ApJ, 837, 170,doi: 10.3847/1538-4357/aa5ffe Leja, J., van Dokkum, P., & Franx, M. 2013, ApJ, 766, 33,doi: 10.1088/0004-637X/766/1/33Liu, D., Daddi, E., Dickinson, M., et al. 2018, ApJ, 853,172, doi: 10.3847/1538-4357/aaa600Lo Faro, B., Buat, V., Roehlly, Y., et al. 2017, MNRAS,472, 1372, doi: 10.1093/mnras/stx1901Madau, P., & Dickinson, M. 2014, ARA&A, 52, 415,doi: 10.1146/annurev-astro-081811-125615Magnelli, B., Ivison, R. J., Lutz, D., et al. 2015, A&A, 573,A45, doi: 10.1051/0004-6361/201424937Magnelli, B., Karim, A., Staguhn, J., et al. 2019, arXive-prints. https://arxiv.org/abs/1904.10006Marrone, D. P., Spilker, J. S., Hayward, C. C., et al. 2018,Nature, 553, 51, doi: 10.1038/nature24629 McCracken, H. J., Milvang-Jensen, B., Dunlop, J., et al.2012, A&A, 544, A156,doi: 10.1051/0004-6361/201219507McLeod, D. J., McLure, R. J., & Dunlop, J. S. 2016,MNRAS, 459, 3812, doi: 10.1093/mnras/stw904Micha(cid:32)lowski, M. J., Dunlop, J. S., Koprowski, M. P., et al.2017, MNRAS, 469, 492, doi: 10.1093/mnras/stx861Murray, S. G., Power, C., & Robotham, A. S. G. 2013,MNRAS, 434, L61, doi: 10.1093/mnrasl/slt079Muzzin, A., Marchesini, D., Stefanon, M., et al. 2013,ApJS, 206, 8, doi: 10.1088/0067-0049/206/1/8Narayanan, D., Conroy, C., Dav´e, R., Johnson, B. D., &Popping, G. 2018, ApJ, 869, 70,doi: 10.3847/1538-4357/aaed25Narayanan, D., & Krumholz, M. R. 2014, MNRAS, 442,1411, doi: 10.1093/mnras/stu834Nayyeri, H., Mobasher, B., Hemmati, S., et al. 2014, ApJ,794, 68, doi: 10.1088/0004-637X/794/1/68Oesch, P. A., Bouwens, R. J., Illingworth, G. D., Labb´e, I.,& Stefanon, M. 2018, ApJ, 855, 105,doi: 10.3847/1538-4357/aab03fOliver, S. J., Bock, J., Altieri, B., et al. 2012, MNRAS, 424,1614, doi: 10.1111/j.1365-2966.2012.20912.xPavesi, R., Riechers, D. A., Sharon, C. E., et al. 2018, ApJ,861, 43, doi: 10.3847/1538-4357/aac6b6Salim, S., Boquien, M., & Lee, J. C. 2018, ApJ, 859, 11,doi: 10.3847/1538-4357/aabf3cSalmon, B., Papovich, C., Long, J., et al. 2016, ApJ, 827,20, doi: 10.3847/0004-637X/827/1/20Sanders, D. B., Salvato, M., Aussel, H., et al. 2007, ApJS,172, 86, doi: 10.1086/517885Schinnerer, E., Smolˇci´c, V., Carilli, C. L., et al. 2007, ApJS,172, 46, doi: 10.1086/516587Schinnerer, E., Sargent, M. T., Bondi, M., et al. 2010,ApJS, 188, 384, doi: 10.1088/0067-0049/188/2/384Schreiber, C., Pannella, M., Leiton, R., et al. 2017, A&A,599, A134, doi: 10.1051/0004-6361/201629155Schreiber, C., Glazebrook, K., Nanayakkara, T., et al. 2018,A&A, 618, A85, doi: 10.1051/0004-6361/201833070Scoville, N., Sheth, K., Aussel, H., et al. 2016, ApJ, 820, 83,doi: 10.3847/0004-637X/820/2/83Serjeant, S., & Marchetti, L. 2014, MNRAS, 443, 3118,doi: 10.1093/mnras/stu1266Silverman, J. D., Rujopakarn, W., Daddi, E., et al. 2018,ApJ, 867, 92, doi: 10.3847/1538-4357/aae25eSimpson, J. M., Swinbank, A. M., Smail, I., et al. 2014,ApJ, 788, 125, doi: 10.1088/0004-637X/788/2/125Smolˇci´c, V., Novak, M., Bondi, M., et al. 2017, A&A, 602,A1, doi: 10.1051/0004-6361/201628704 Song, M., Finkelstein, S. L., Ashby, M. L. N., et al. 2016,ApJ, 825, 5, doi: 10.3847/0004-637X/825/1/5Spilker, J. S., Marrone, D. P., Aravena, M., et al. 2016,ApJ, 826, 112, doi: 10.3847/0004-637X/826/2/112Spilker, J. S., Aravena, M., B´ethermin, M., et al. 2018,Science, 361, 1016, doi: 10.1126/science.aap8900Stefanon, M., Bouwens, R. J., Labb´e, I., et al. 2017, ApJ,843, 36, doi: 10.3847/1538-4357/aa72d8Straatman, C. M. S., Labb´e, I., Spitler, L. R., et al. 2014,ApJL, 783, L14, doi: 10.1088/2041-8205/783/1/L14Strandet, M. L., Weiss, A., Vieira, J. D., et al. 2016, ApJ,822, 80, doi: 10.3847/0004-637X/822/2/80Strandet, M. L., Weiss, A., De Breuck, C., et al. 2017,ApJL, 842, L15, doi: 10.3847/2041-8213/aa74b0Swinbank, A. M., Simpson, J. M., Smail, I., et al. 2014,MNRAS, 438, 1267, doi: 10.1093/mnras/stt2273Taniguchi, Y., Scoville, N., Murayama, T., et al. 2007,ApJS, 172, 9, doi: 10.1086/516596Tisani´c, K., Smolˇci´c, V., Delhaize, J., et al. 2019, A&A,621, A139, doi: 10.1051/0004-6361/201834002Toft, S., Smolˇci´c, V., Magnelli, B., et al. 2014, ApJ, 782,68, doi: 10.1088/0004-637X/782/2/68Trenti, M., & Stiavelli, M. 2008, ApJ, 676, 767,doi: 10.1086/528674van der Wel, A., Franx, M., van Dokkum, P. G., et al. 2014,ApJ, 788, 28, doi: 10.1088/0004-637X/788/1/28Walter, F., Decarli, R., Carilli, C., et al. 2012, Nature, 486,233, doi: 10.1038/nature11073Wang, R., Wagg, J., Carilli, C. L., et al. 2013, ApJ, 773, 44,doi: 10.1088/0004-637X/773/1/44Wang, T., Elbaz, D., Schreiber, C., et al. 2016, ApJ, 816,84, doi: 10.3847/0004-637X/816/2/84Wardlow, J. L., Simpson, J. M., Smail, I., et al. 2018,MNRAS, 479, 3879, doi: 10.1093/mnras/sty1526Weiß, A., Kov´acs, A., Coppin, K., et al. 2009, ApJ, 707,1201, doi: 10.1088/0004-637X/707/2/1201Whitaker, K. E., Pope, A., Cybulski, R., et al. 2017, ApJ,850, 208, doi: 10.3847/1538-4357/aa94ceWilliams, C. C., Giavalisco, M., Cassata, P., et al. 2014,ApJ, 780, 1, doi: 10.1088/0004-637X/780/1/1Williams, C. C., Giavalisco, M., Lee, B., et al. 2015, ApJ,800, 21, doi: 10.1088/0004-637X/800/1/21Williams, C. C., Curtis-Lake, E., Hainline, K. N., et al.2018, ArXiv e-prints. https://arxiv.org/abs/1802.05272Wuyts, S., F¨orster Schreiber, N. M., Lutz, D., et al. 2011,ApJ, 738, 106, doi: 10.1088/0004-637X/738/1/106Yamaguchi, Y., Kohno, K., Hatsukade, B., et al. 2019,arXiv e-prints. https://arxiv.org/abs/1903.02744Zavala, J. A., Casey, C. M., da Cunha, E., et al. 2018a,ApJ, 869, 71, doi: 10.3847/1538-4357/aaecd20