SXDF-ALMA 1.5 arcmin^2 deep survey. A compact dusty star-forming galaxy at z=2.5
Ken-ichi Tadaki, Kotaro Kohno, Tadayuki Kodama, Soh Ikarashi, Itziar Aretxaga, Stefano Berta, Karina I. Caputi, James S. Dunlop, Bunyo Hatsukade, Masao Hayashi, David H. Hughes, Rob Ivison, Takuma Izumi, Yusei Koyama, Dieter Lutz, Ryu Makiya, Yuichi Matsuda, Kouichiro Nakanishi, Wiphu Rujopakarn, Yoichi Tamura, Hideki Umehata, Wei-Hao Wang, Grant W. Wilson, Stijn Wuyts, Yuki Yamaguchi, Min S. Yun
aa r X i v : . [ a s t r o - ph . GA ] A ug Draft version August 8, 2018
Preprint typeset using L A TEX style emulateapj v. 5/2/11
SXDF-ALMA 1.5 ARCMIN DEEP SURVEY. A COMPACT DUSTY STAR-FORMING GALAXY AT Z=2.5.
Ken-ichi Tadaki , Kotaro Kohno , Tadayuki Kodama , Soh Ikarashi , Itziar Aretxaga , Stefano Berta ,Karina I. Caputi , James S. Dunlop , Bunyo Hatsukade , Masao Hayashi , David H. Hughes , Rob Ivison ,Takuma Izumi , Yusei Koyama , Dieter Lutz , Ryu Makiya , Yuichi Matsuda , Kouichiro Nakanishi , WiphuRujopakarn , Yoichi Tamura , Hideki Umehata , Wei-Hao Wang , Grant W. Wilson , Stijn Wuyts ,Yuki Yamaguchi , Min S. Yun Max-Planck-Institut f¨ur extraterrestrische Physik (MPE), Giessenbachstr., D-85748 Garching, Germany; [email protected] Institute of Astronomy, The University of Tokyo, 2-21-1 Osawa, Mitaka, Tokyo 181-0015, Japan Research Center for the Early Universe, The University of Tokyo, 7-3-1 Hongo, Bunkyo, Tokyo 113-0033, Japan National Astronomical Observatory of Japan, 2-21-1 Osawa, Mitaka, Tokyo 181-8588, Japan Department of Astronomical Science, SOKENDAI (The Graduate University for Advanced Studies), Mitaka, Tokyo 181-8588, Japan Kapteyn Astronomical Institute, University of Groningen, P.O. Box 800, 9700AV Groningen, The Netherlands Instituto Nacional de Astrof´ısica, ´Optica y Electr´onica (INAOE), Luis Enrique Erro 1, Sta. Ma. Tonantzintla, Puebla, Mexico Institute for Astronomy, University of Edinburgh, Royal Observatory, Blackford Hill, Edinburgh EH9 3HJ, UK European Southern Observatory, Karl-Schwarzschild-Strasse 2, Garching bei Munchen, Germany Joint ALMA Observatory, Alonso de C´ordova 3107, Vitacura 763-0355, Santiago de Chile Department of Physics, Faculty of Science, Chulalongkorn University, 254 Phayathai Rd, Pathumwan, Bangkok 10330, Thailand Kavli Institute for the Physics and Mathematics of the Universe (WPI), Todai Institute for Advanced Study, University of Tokyo,5-1-5 Kashiwanoha, Kashiwa, 277-8583, Japan Institute of Astronomy and Astrophysics, Academia Sinica, Taipei, Taiwan Canada-France-Hawaii Telescope, HI, USA and Department of astronomy, University of Massachusetts, Amherst, MA 01003, USA
Draft version August 8, 2018
ABSTRACTWe present first results from the SXDF-ALMA 1.5 arcmin deep survey at 1.1 mm using Ata-cama Large Millimeter Array (ALMA). The map reaches a 1 σ depth of 55 µ Jy/beam and covers 12H α -selected star-forming galaxies at z = 2 .
19 or z = 2 .
53. We have detected continuum emissionfrom three of our H α -selected sample, including one compact star-forming galaxy with high stellarsurface density, NB2315-07. They are all red in the rest-frame optical and have stellar masses of log( M ∗ /M ⊙ ) > . M ∗ /M ⊙ )=10.0-10.8 areexceedingly faint, < µ Jy (2 σ upper limit). We also find the 1.1 mm-brightest galaxy, NB2315-02,to be associated with a compact ( R e = 0 . ± . +20 − % or 37 +25 − %) and high star formation rate surface density (126 +27 − M ⊙ yr − kpc − ),the concentrated starburst can within less than 50 +12 − Myr build up a stellar surface density matchingthat of massive compact galaxies at z ∼
2, provided at least 19 ±
3% of the total gas is convertedinto stars in the galaxy center. On the other hand, NB2315-07, which already has such a high stellarsurface density core, shows a gas fraction (23 +8 − %) and is located in the lower envelope of the starformation main-sequence. This compact less star-forming galaxy is likely to be in an intermediatephase between compact dusty star-forming and quiescent galaxies. Subject headings: galaxies: evolution — galaxies: high-redshift — galaxies: ISM INTRODUCTION
One of the fundamental questions in galaxy forma-tion is when and how massive galaxies are formed athigh redshift. Massive quiescent galaxies (QGs) alreadyexisted at z ≥ R e ∼ slow and fast modes. In the slow mode,new, larger quenched galaxies continuously add to qui-escent population at 0 < z < z ∼ R e ∼ z ∼ RA (J2000) D e c ( J ) Fig. 1.—
12 H α -selected SFGs in the SXDF-ALMA 1.5 arcmin deep field. Red and blue squares indicate 1.1 mm sources and non-detection, respectively. A white solid and dashed line show theareas where the primary beam correction is below 20% and 50%,respectively. core whose size is four times smaller than their stellarcomponent (Simpson et al. 2015, see also Ikarashi et al.2014). Moreover, massive compact SFGs have been dis-covered at z ∼ z ∼
2, along with the discovery of a galaxy with an ex-tremely compact, dusty star-forming component. Theseare the first results from the SXDF-ALMA 1.5 arcmin deep survey, which is unconfused (FWHM=0 ′′ .5) and un-biased (i.e., blind) observations with Atacama Large Mil-limeter Array (ALMA). The ALMA observations targeta part of the SXDF-UDS-CANDELS field, where HubbleSpace Telescope (HST) high resolution images are avail-able (Grogin et al. 2011; Koekemoer et al. 2011). Weonly focus on 12 SFGs identified by a H α narrow-bandimaging survey with Subaru Telescope (Tadaki et al.2013; Kodama et al. 2013). Survey design and sourcecatalog of the ALMA observations are described in de-tail in Kohno et al. in preparation. We assume theChabrier initial mass function (IMF; Chabrier 2003) NB2315-02 NB2315-07 NB2315-11 NB2315-12NB2315-13 NB2315-15 NB2315-16 NB2315-19NB2315-21 NB2315-32 NB2315-34 NB209-39
Fig. 2.— α -selectedsample, superimposed on three-color images at ACS/ F W -,WFC3/ F W - and F W -band (3 ′′ × ′′ ). Contours are plottedevery 1 σ , starting at 2 σ , except for NB2315-02 and 12 (every 5 σ ).Dashed contours denote negative fluxes. The synthesized beamsizeis shown at bottom left of the NB2315-02 panel. and adopt cosmological parameters of H =70 km s − Mpc − , Ω M =0.3, and Ω Λ =0.7. SAMPLE AND DATA
The SXDF-ALMA 1.5 arcmin deep survey consistsof 19 pointings (Figure 1). The band 6 receivers wereused in frequency range of 255–259, 271–275 GHz ( ∼ ′′ .53 × ′′ .41 and 55 µ Jy/beam, respectively.The SXDF-ALMA 1.5 arcmin deep field covers 11 H α -selected SFGs at z = 2 .
53 and one at z = 2 .
19 (Figure 1).As they are carefully identified on the basis of the narrow-band excess and the broad-band colors for the line sep-aration between H α and [O iii ], the redshift ranges arestrongly-constrained by the width of narrow-band filters,∆ z = ± .
02. X-ray detected AGNs or galaxies with apower-low spectral energy distribution (SED) at IRACfour bands are not included in our sample (Tadaki et al.2013).First, we model their SEDs to measure the rest-frame U − V and V − J color ( U V J ) using the 3D-HST cata-log with photometries at 18 bands (Skelton et al. 2014)and the
EAZY code (Brammer et al. 2008). Next, we esti-mate the stellar mass and the amount of dust extinctionfor our sample by fitting the photometric SEDs with thestellar population synthesis model of Bruzual & Charlot(2003). The SED fitting is done using the
FAST code(Kriek et al. 2009) with a solar metallicity, exponentiallydeclining star formation histories (SFHs), and dust at-tenuation law of Calzetti et al. (2000). Following therecipe presented by Wuyts et al. (2011b), we adopt aminimum e -folding time of 300 Myr. The uncertainties ofthe derived stellar mass are estimated from the 68% con-fidence interval. While the stellar mass tends to be themost robust parameter in SED modeling as long as singleexponentially declining SFHs are assumed, the mass-to-light ratios could become smaller in 2-component SFHsXDF-ALMA 1.5 arcmin deep survey 3 . . . . . (V - J) rest [mag] . . . . . . ( U - V ) r e s t [ m ag ] quiescentstar-forming A V =1 magNB2315-02NB2315-07 NB2315-12 . . . . . log M ∗ [M ⊙ ] . . . . . . . . l og S F R H α [ M ⊙ y r − ] MS × Fig. 3.—
Left: Rest-frame U − V versus V − J color for our sample. Red circles and blue squares indicate 1.1 mm-detected andnon-detection sources, respectively. Gray circles show the parent sample of H α -selected SFGs. The red shaded area indicates a regionwhere quiescent galaxies dominate (Whitaker et al. 2011). Right: H α -based star formation rates plotted against stellar masses. The blacksolid line indicates the main-sequence of SFGs at z = 2 . − . with recent burst (Wuyts et al. 2007). The differencesare smaller for redder galaxies because a new burst makesgalaxies blue in the rest frame optical. In section 3.2, thestellar masses are used for determine dust temperatures.A higher stellar mass leads to a lower dust temperatureand results in a higher gas mass, but does not signifi-cantly change it for red galaxies.Star formation rates (SFRs) are derived from H α lu-minosities (Kennicutt 1998) and SED-based extinc-tions, accounting for extra extinction toward H ii regions(Wuyts et al. 2013). The [N ii ] contribution has been cor-rected on the basis of the measured equivalent width ofH α +[N ii ] emission lines (Tadaki et al. 2013). We havealso applied UNIMAP (Piazzo et al. 2015) and extrac-tion methods as described in Lutz et al. (2011) to thearchival Herschel-PACS data. We detect one sourceNB2315-02 at S µ m = 7 . ± . L IR /L ⊙ ) = 12 . z = 2 .
53 from PACS 160 µ m to L IR , based on the Wuyts et al. (2011b) template.The PACS 160 µ m-based SFR is 414 ± M ⊙ yr − whichis consistent with the H α -based SFR of 495 ± M ⊙ yr − . RESULTS
Detections in 1.1 mm continuum map
As the 1.1 mm continuum emission of SFGs at z ∼ ′′ .5 aper-tures and estimate the noise from 2000 random apertures(1 σ = 125 µ Jy before the primary beam correction). Weadopt a 2 σ detection criterion in the same manner asScoville et al. (2014). The probabilities where a nega-tive 2 σ signal is detected by chance in the ALMA map is2.2%. Three SFGs are associated with a 1.1 mm source with a > . σ significance (Figure 2). NB2315-02 andNB2315-12 are robustly detected above 9 σ , and NB2315-07 are marginally done at 2.6 σ . The spurious detectionrates of 2.6 σ sources is 0.5%. Table 1 lists their measuredflux densities along with stellar masses and SFRs. Theother 9 SFGs do not have a significant 1.1 mm continuumemission and their 2 σ upper limit is S . < µ Jyafter the primary beam correction.Our deep H α narrowband selection covers wide rangesin colors and stellar mass of SFGs from blue, less mas-sive to red, massive ones. The 1.1 mm-detected SFGsare redder in the rest-frame optical and more massivethan the non-detected sample (Figure 3). The U V J di-agram is useful for distinguishing between dusty star-forming, unobscured star-forming, and quiescent galaxies(e.g., Wuyts et al. 2007). The three 1.1 mm sources arefound to be dusty SFGs while most of SFGs without 1.1mm detection fall in the unobscured star-forming regime,where also the bulk of galaxies satisfying the Lymanbreak criterion lies. One galaxy, which lies in the regimewhere quiescent galaxies dominate, is not detected above2 σ . NB2315-07 is also close to this regime compared toNB2315-02 and NB2315-12. Their lower position withrespect to the main sequence supports that they are lessstar-forming galaxies. We stack the unconfused 1.1 mmmaps in the positions of 8 non-detected objects excludingone less star-forming galaxy to investigate the existenceof faint emission for typical main-sequence SFGs withlog( M ∗ /M ⊙ )=10.0-10.8. However, the deep stacked im-age does not show a significant emission and gives the 2 σ upper limit of S . < µ Jy.
Gas mass estimates
Measurements of 1.1 mm continuum flux densities, S ν ,allow us to derive gas masses of galaxies although thereare uncertainties associated with dust temperature vari- Tadaki et al. TABLE 1Properties of 1.1 mm-detected SFGs
ID R.A. Decl. z NB M ∗ SF R H α S . , aper M gas f gas (J2000) (J2000) (10 M ⊙ ) ( M ⊙ yr − ) (mJy) (10 M ⊙ ) (%)NB2315-02 02 17 40.53 −
05 13 10.7 2.53 ± +4 . − . ±
95 2.06 ± +1 . − . +20 − NB2315-07 02 17 42.67 −
05 13 58.4 2.53 ± +0 . − . ±
10 0.38 ± +1 . − . +8 − NB2315-12 02 17 41.11 −
05 13 15.2 2.53 ± +1 . − . ±
20 1.34 ± +1 . − . +5 − ations (e.g., Scoville et al. 2014; Genzel et al. 2015). Weestimate the gas mass of our sample using a modifiedblackbody radiation model as, M gas = M dust /δ dgr = S ν d κ ISM B ν ( T dust )(1 + z ) , (1)where δ dgr is dust-to-gas ratio, κ ISM is the dust opacityper unit mass of interstellar medium ( ∝ ν β ), B ν is thePlank function, T dust is the dust temperature, and d L is the luminosity distance. Here, κ ISM , µ m = 4 . × − g − cm and β = 1 . ± κ ISM ) is not straightforward, observations of dustemission still have a big cost advantage over CO obser-vations. Genzel et al. (2015) present scaling relationsof dust temperature, gas fraction, and gas depletiontimescale, by compiling CO and dust continuum data foreach ∼
500 galaxies over 0 < z <
3, and find dust temper-atures to only change slowly with specific SFR (sSFR)in the narrow range around the main sequence (see alsoMagnelli et al. 2014). This scaling relation of dust tem-peratures is used for better estimates of the rest-frame850 µ m fluxes from the observed 1.1 mm fluxes. The de-rived dust temperatures are 33 +3 − , 28 ±
2, and 30 ± f gas = M gas / ( M gas + M ∗ ), are summarizedin Table 1.The 2 σ upper limit of our 1.1 mm survey correspondsto a gas mass of log ( M gas /M ⊙ ) = 10 . T dust = 30 K at z = 2 .
53. Surprisingly, we do not detect the 1.1 mm emis-sion from relatively less massive SFGs around the main-sequence although their gas mass is expected to be log( M gas /M ⊙ ) = 10 . − . κ ISM (= κ dust δ dgr )where the term of δ dgr is canceled out on the right side ofthe equation (1). SFGs with lower metallicity should befainter at 1.1 mm due to a lower dust-to-gas ratio thanmetal-rich ones, log δ dgr = − . × (12+log(O/H)-8.67), (Leroy et al. 2011) even if they have the samegas mass. The mass-metallicity relation at z ∼ δ dgr = − . M ∗ /M ⊙ )=10(Wuyts et al. 2014). Moreover, although the used dustopacity has been calibrated mainly by using outliersabove the main-sequence such as local ultraluminous in-frared galaxies and bright SMGs at z ∼ z ∼
2. Diffuse dust distribu-tions within galaxies could lead to a low dust opacity A m p [ m Jy ] J0215-0222 A m p [ m Jy ] NB2315-02 gaussian (R e =0.66 kpc)point source0 100 200 300 400 500 600 UV Distance [k λ ] A m p [ m Jy ] NB2315-12 R e =0.81 kpcpoint source Fig. 4.—
The visibility amplitudes averaged over uv distancesfor the phase calibrator, J0215–0222 (top), NB2315-02 (middle),and NB2315-12 (bottom). A blue solid line and a shaded regionindicate the best-fitting circular Gaussian component and the 1 σ error, respectively. A red dashed line presents the best-fitting pointsource model as a reference. (Dunne et al. 2003), resulting in faint 1.1 mm emissionat fixed gas mass. Also, dust SEDs of SFGs are bet-ter described by a multi-temperature model includingcool dust in diffuse ISM (dominating a dust mass) andwarm dust in birth clouds (dominating a total infraredluminosity). Local normal SFGs have a cold componentwith T dust ∼
20 K in a two-temperature model whilea single-temperature model shows higher dust tempera-tures (Dunne et al. 2011).Assuming log δ dgr = − . κ dust , µ m = 0 . − cm (Dunne et al. 2003), and T dust = 20 K, the 2 σ gas masslimit of our observations would become log ( M gas /M ⊙ ) =11 .
0. As the non-detections for less massive galaxies canbe explained by a combination of these factors, it is dif-ficult to determine whether the gas mass is actually assmall as log ( M gas /M ⊙ ) = 10 . Size measurements of dust emission
We measure the size of the dust emission for two 1.1mm bright galaxies (NB2315-02 and NB2315-12) by fit-XDF-ALMA 1.5 arcmin deep survey 5ting the visibility data in the uv plane. Their signal-to-noise ratios in the 1.1 mm map are >
10 in the flux den-sity per synthesized beam to reliably constrain the size asthe visibility coverage is similar as those by Ikarashi et al.(2014) and Simpson et al. (2015). Then, we assumetwo models, a circular Gaussian component and a pointsource. The uvamp task in MIRIAD (Sault et al. 1995) isused for calculating the visibility amplitudes averaged inannuli according to uv distance after shifting the phasecenter to the center position measured in the image planeand subtracting a clean component of another source inthe primary beam with the uvmodel task. The phasecalibrator, J0215-0222, shows constant amplitudes as afunction of uv distance, suggesting a point source in theused antenna configuration (Figure 4). On the otherhand, two SFGs of our sample seem to be resolved at >
300 k λ . The gaussian fitting shows reduced chi-squarevalues of 0.61 for NB2315-02 and 1.62 for NB2315-12while the horizontal fitting (point source) does 5.76 and2.60, respectively. Therefore, we adopt the gaussianmodel which is the same approach as previous studies(Ikarashi et al. 2014; Simpson et al. 2015). The best-fitresults are FWHM=0 ′′ .16 +0 . − . ( R e = 0 . ± +0 . − . kpc)for NB2315-02 and FWHM=0 ′′ .20 +0 . − . ( R e = 0 . +0 . − . kpc) for NB2315-12. NB2315-02 surely has a compact,dusty star-forming component with R e < σ sig-nificance. For a gaussian source with R e = 0 . ± +0 . − . kpc, 80 +7 − % (SFR H α =396 +84 − M ⊙ yr − ) of star forma-tion traced by H α is occurring within 1 kpc region. Then,the SFR surface density could be 126 +27 − M ⊙ yr − kpc − .It is also 105 +19 − M ⊙ yr − kpc − in the case of usageof the PACS 160 µ m-based SFR. The gas surface den-sity is similarly estimated from the total gas mass to be(2 . ± . × M ⊙ kpc − , which probably causes anextremely red color due to strong extinction (Figure 3). DISCUSSION
We find NB2315-02 to have a high SFR surface den-sity, 126 +27 − M ⊙ yr − kpc − , corresponding to SFR=396 +84 − M ⊙ yr − within a region of 1 kpc radius. The cen-tral stellar surface density will become comparable withcompact SFGs/QGs, log( M ∗ /R . e ) > . M ⊙ kpc − . ](Barro et al. 2013), provided that the current star forma-tion is maintained for another 50 +12 − Myr in the galaxycenter. This is plausible since it would need only 19 ± M gas =10.3 +1 . − . × M ⊙ ) beingconverted into stars. Then, the gas depletion timescale, τ depl ≡ M gas /SF R H α , is 207 +78 − Myr. Barro et al.(2014b) have also estimated dynamical masses of 13 com-pact SFGs from line widths of nebular emission (H α or[O iii ]) by near-infrared slit spectroscopy and derivedsimilar gas depletion timescales ( τ depl =230 +110 − Myr).The agreement of the gas depletion timescales would sup-port that the compact dusty SFG can be an immediateprogenitor of the high stellar surface density SFGs.In the WFC3/ F W -band image, NB2315-02 has asub-component with log ( M ∗ /M ⊙ ) = 10 .
67, which is seen5 kpc east of the main component with the H α peak(Figure 2). The compact starburst region appears to belocated in between the two components. Given the highgas surface density of (2 . ± . × M ⊙ kpc − in the compact dusty SFG, the rest optical morphology maybe severely affected by strong attenuation. One likelyinterpretation is that the two components would consti-tute a single large SFG with a central dusty star forma-tion. The gas fraction becomes 37 +25 − % in this systemincluding the companion. A physical process to reduceangular momentum is required in order to explain a con-centrated gas distribution in a center of extended SFGs.In gas-rich disks at z ∼
2, gravitational torques and dy-namical friction due to clumps can drive angular mo-mentum out and cause mass inflow towards galaxy cen-ters (Dekel & Burkert 2014; Zolotov et al. 2015). Then,if the inflow timescale is shorter than the star formationtimescale, a gas-rich, central starburst may form. Themeasured high gas fraction of the compact dusty SFGcould support the possibility of such a dissipational pro-cess. Another possible explanation is we are witnessinga late-stage merger with 1:3 stellar mass ratio immedi-ately before the final coalescence. Numerical simulationsdemonstrate that gas-rich mergers produce an instanta-neous starburst at the final coalescence although it de-pends on merger parameters such as orbits of the twogalaxies (Hopkins et al. 2013).As compact SFGs are commonly defined by not highgas/dust surface densities but stellar ones (Barro et al.2013, 2014a), they are likely to have already com-pleted most of morphological transformation from ex-tended disks to compact spheroids while compact dustySFGs have not done yet. NB2315-07 of our sampleshows a compact morphology with a circularized, effec-tive radius of R e = 1 . ± .
03 kpc at F W -band(van der Wel et al. 2014) and satisfies the conditions forcompact SFGs, log( M ∗ /R . e ) = 10 . +0 . − . > . M ⊙ kpc − . ] and log ( SF R H α /M ∗ ) = − . ± . > -1.0[Gyr − ] (Barro et al. 2014a). The 1.1 mm-based gas frac-tion is 23 +8 − %, which is smaller than that of the compactdusty SFG at 2 σ significance. Given the observed lowergas fraction and less star-forming properties shown inFigure 3, this compact less star-forming galaxy is likelyto be already in a late stage of their evolutionary pathto compact QGs.In the fast quenching mode for galaxy evolution, we arelooking a various compact objects in different evolution-ary phases from dusty to quiescent through star-forminggalaxies. Barro et al. (2014a) also present a diversity ofcompact SFGs from highly obscured (they are similar tocompact dusty SFGs but already have a compact stellarcore) to less star-forming ones (to which NB2315-07 be-longs). The spatial extent of gas remaining within galax-ies can provide key information about the subsequentevolution of compact SFGs. If gas is still concentrated ina galaxy center, compact SFGs are expect to exhaust allgas by a nuclear starburst or a feeding to a super mas-sive black hole and then quench star formation. Deepand high-resolution submillimeter imaging with ALMAhas great potential to address this issue.We thank the anonymous referee who gave us anumber of comments, which improved the Letter.This paper makes use of the following ALMA data:ADS/JAO.ALMAsignificance. Given the observed lowergas fraction and less star-forming properties shown inFigure 3, this compact less star-forming galaxy is likelyto be already in a late stage of their evolutionary pathto compact QGs.In the fast quenching mode for galaxy evolution, we arelooking a various compact objects in different evolution-ary phases from dusty to quiescent through star-forminggalaxies. Barro et al. (2014a) also present a diversity ofcompact SFGs from highly obscured (they are similar tocompact dusty SFGs but already have a compact stellarcore) to less star-forming ones (to which NB2315-07 be-longs). The spatial extent of gas remaining within galax-ies can provide key information about the subsequentevolution of compact SFGs. If gas is still concentrated ina galaxy center, compact SFGs are expect to exhaust allgas by a nuclear starburst or a feeding to a super mas-sive black hole and then quench star formation. Deepand high-resolution submillimeter imaging with ALMAhas great potential to address this issue.We thank the anonymous referee who gave us anumber of comments, which improved the Letter.This paper makes use of the following ALMA data:ADS/JAO.ALMA