Resolving a dusty, star-forming SHiZELS galaxy at z=2.2 with HST, ALMA and SINFONI on kiloparsec scales
R. K. Cochrane, P. N. Best, I. Smail, E. Ibar, A. M. Swinbank, J. Molina, D. Sobral, U. Dudzeviciute
MMNRAS in press, 1–17 (2021) Preprint 17 February 2021 Compiled using MNRAS L A TEX style file v3.0
Resolving a dusty, star-forming SHiZELS galaxy at z = . HST ,ALMA and SINFONI on kiloparsec scales
R. K. Cochrane, , ★ P. N. Best, I. Smail, E. Ibar, C. Cheng, , , A. M. Swinbank, J. Molina, D. Sobral, U. Dudzeviči¯ut˙e Harvard-Smithsonian Center for Astrophysics, 60 Garden St. Cambridge, MA 02138, USA SUPA, Institute for Astronomy, Royal Observatory Edinburgh, EH9 3HJ, UK Centre for Extragalactic Astronomy, Department of Physics, Durham University, South Road, Durham DH1 3LE, UK Instituto de Física y Astronomía, Universidad de Valparaíso, Avda. Gran Bretaña 1111, Valparaíso, Chile Chinese Academy of Sciences South America Center for Astronomy, National Astronomical Observatories, CAS, Beijing 100101, China CAS Key Laboratory of Optical Astronomy, National Astronomical Observatories, Chinese Academy of Sciences, Beijing 100101, China Kavli Institute for Astronomy and Astrophysics, Peking University, 5 Yiheyuan Road, Haidian District, Beijing 100871, P.R. China Department of Physics, Lancaster University, Lancaster, LA1 4YB
Accepted 2021 February 10. Received 2021 February 10; in original form 2020 August 21
ABSTRACT
We present ∼ . (cid:48)(cid:48) spatial resolution imaging of SHiZELS-14, a massive ( 𝑀 ∗ ∼ M (cid:12) ), dusty, star-forming galaxy at 𝑧 = .
24. Our rest-frame ∼ 𝛼 emission line (from SINFONI/VLT), rest-frame UV continuum (from HST
F606W imaging), the rest-frame far-infrared(from ALMA), and the radio continuum (from JVLA). Although originally identified by its modest H 𝛼 emission line flux,SHiZELS-14 appears to be a vigorously star-forming (SFR ∼ (cid:12) yr − ) example of a submillimeter galaxy, probablyundergoing a merger. SHiZELS-14 displays a compact, dusty central starburst, as well as extended emission in H 𝛼 and therest-frame optical and FIR. The UV emission is spatially offset from the peak of the dust continuum emission, and appears totrace holes in the dust distribution. We find that the dust attenuation varies across the spatial extent of the galaxy, reaching apeak of at least 𝐴 H 𝛼 ∼ ∼ − (cid:12) yr − , andare particularly discrepant in the galaxy’s dusty centre. This galaxy highlights the biased view of the evolution of star-forminggalaxies provided by shorter wavelength data. Key words: galaxies: high redshift – galaxies: evolution – galaxies: starburst – galaxies: star formation – submillimetre: galaxies– infrared: galaxies
Galaxy surveys have long shown that star-formation rates within in-dividual galaxies increase towards high redshift. At a given stellarmass, typical star-formation rates increase by over an order of mag-nitude between the present day and the peak of cosmic star formationat 𝑧 ∼ 𝑧 = 𝐿 TIR > − L (cid:12) ), galaxies with typicalULIRG luminosities are more common around the peak of cosmicstar formation (Smail et al. 1997; Barger et al. 1998). Submillime- ★ E-mail: [email protected] ter galaxies (SMGs; Blain et al. 2002) are ULIRGs at high redshiftwith bright submillimeter fluxes that suggest star-formation rates(SFRs) of ∼ − (cid:12) yr − . Sustained star-formation rates ofthis magnitude have the potential to form massive galaxies (withstellar masses of ∼ M (cid:12) ) on sub-Gyr timescales (Simpson et al.2014; Dudzeviciute et al. 2020). Chapman et al. (2005) found thatthe volume density of SMGs increases by a factor of ∼ 𝑧 = 𝑧 = .
5, with the redshift distribution peaking at 𝑧 ∼ . − . < 𝑧 < ∼ −
30 per cent of thetotal comoving star-formation rate density at these redshifts (Swin-bank et al. 2014; Smith et al. 2017; Dudzeviciute et al. 2020). Evenin less far-infrared (FIR)-luminous high redshift galaxies, a signif-icant amount of star formation is obscured by dust. Dunlop et al.(2017) combined long- and short-wavelength data from two premierobservatories: the Atacama Large Millimeter Array (ALMA, prob-ing the dust continuum emission at 1.3mm) and the
Hubble Space © a r X i v : . [ a s t r o - ph . GA ] F e b R.K. Cochrane et al.
Telescope ( HST , Wide Field Camera 3, probing rest-frame UV), inthe well-studied Hubble Ultra Deep Field (e.g. Bouwens et al. 2010;Oesch et al. 2010; Illingworth et al. 2013; Dunlop et al. 2013). Thesecomplementary data enabled them to confirm that ∼
85 per cent ofthe total star formation at 𝑧 ∼ − 𝑀 ∗ ∼ × M (cid:12) , they suggest a ratio of obscuredto unobscured star formation of ∼
50. However, lower mass galaxiesare still affected, with the ratio decreasing to ∼ 𝑀 ∗ ∼ × M (cid:12) (see also Magnelli et al. 2020).While studies of wide areas are important in tracking the evolvingproperties of star-forming galaxies and the build-up of stellar massin the Universe, understanding the physical processes of star forma-tion within individual galaxies requires higher angular resolution.Until recently, resolved studies of distant star-forming (SF) galaxiestended to be based on observations from near-infrared integral fieldunit spectrographs, which probe rest-frame optical emission linessuch as H 𝛼 and [O III ] at 𝑧 ∼ HST at rest-frame UV wavelengths (e.g. Wuyts et al. 2012;Fisher et al. 2017). These claim a physical picture in which star for-mation takes place within massive clumps embedded in turbulentdisk structures (Genzel et al. 2008; Elmegreen et al. 2013; Genzelet al. 2013; Guo et al. 2015, 2017; Soto et al. 2017). Emission at theseshort wavelengths is, however, strongly attenuated by dust, and thesignificant global obscuration of star formation at 𝑧 < 𝑧 > ∼ − 𝑧 ∼ − 𝑀 ∗ ∼ M (cid:12) ; Hodgeet al. 2019; Dudzeviciute et al. 2020) has fuelled speculation that theSMGs detected at 𝑧 ∼ − 𝑧 = HST imaging(Chen et al. 2015; Barro et al. 2016; Hodge et al. 2015, 2016, 2019;Rujopakarn et al. 2019). In some cases, kpc-scale offsets have beenfound between the peaks of the FIR and UV emission (Hodge et al.2015; Tadaki et al. 2016; Chen et al. 2017; Calistro Rivera et al.2018). These offsets could potentially bias interpretations of global measurements (particularly for fits to photometry that focus solely onthe rest-frame optical to near-infrared, but also for ‘energy-balance’spectral energy distribution fitting). Indeed, Simpson et al. (2017)argue that attenuation in the dusty regions of SMGs is so great thatessentially all the co-located stellar emission is obscured at optical-to-near-infrared wavelengths; for ∼
30 per cent of their sample, thedata available at these wavelengths is insufficient to put constraintson photometric redshifts and stellar masses (see also work on ‘NIR-dark’ sources; e.g. Simpson et al. 2014; Franco et al. 2018; Wanget al. 2019; Dudzeviciute et al. 2020; Smail et al. 2020)Overall, it has become clear that drawing conclusions from single-wavelength surveys, especially in the rest-frame UV, is subject to sub-stantial bias and uncertainty, even where data is at high angular reso-lution. In this paper, we present multi-wavelength, 0 . (cid:48)(cid:48) -resolutionimaging of SHiZELS-14, a highly star-forming, H 𝛼 -selected galaxyat 𝑧 = .
24. Of the ALMA-studied SHiZELS parent sample (whichis presented in a companion paper, Cheng et al. 2020), SHiZELS-14is the most FIR luminous, with the largest of all H 𝛼 -derived effectiveradii (4 . ± . 𝛼 flux is modest, it displays SMG-like dust continuumemission. Our observations comprise matched-resolution imagingof the H 𝛼 emission line (from SINFONI/VLT), rest-frame UV andoptical continuum (from HST ), and the rest-frame far-infrared (fromALMA), as well as the radio continuum (from the Karl G. JanskyVery Large Array; JVLA). We find bright, extended structures in themulti-wavelength imaging, with clear clumps in H 𝛼 and extendeddust continuum emission. Given this extended structure and the highsignal-to-noise that results from its high SFR, we have been able toresolve star formation on kpc scales at multiple wavelengths.The structure of this paper is as follows. In Section 2, we providean overview of the data available for our study of SHiZELS-14. Wereview the high quality, but less well-resolved multi-wavelength dataavailable from imaging of the COSMOS field, and present the new0 . (cid:48)(cid:48) resolution imaging from SINFONI/VLT, HST , ALMA andJVLA. We discuss the astrometric alignment of these data in Section2.8. In Section 3, we present the global properties of SHiZELS-14that may be inferred from spectral energy distribution (SED) fitting.In Section 4, we present maps of the spatially-resolved SFRs in-ferred from different SFR indicators, and derive a spatially-resolveddust attenuation map. In Section 5, we compare the properties ofSHiZELS-14 to the submillimeter galaxy population. In Section 6,we summarise our results.We assume a Λ CDM cosmology with 𝐻 =
70 km s − Mpc − , Ω 𝑀 = . Ω Λ = .
7. We use a Kroupa (2002) initial massfunction (IMF).
The High-Redshift(Z) Emission Line Survey, HiZELS, used a com-bination of narrow-band and broad-band filters to select star-forminggalaxies via their emission line fluxes (Sobral et al. 2013a, 2015)in fields with high-quality multi-wavelength coverage (COSMOS,UDS & SA22). This survey has yielded thousands of H 𝛼 emittersat 𝑧 = .
4, 0 .
8, 1 .
47 & 2 .
23, providing sufficiently large samples toconstrain H 𝛼 luminosity functions, stellar mass functions and haloenvironments of typical star-forming galaxies around the peak ofcosmic star formation (Geach et al. 2008; Sobral et al. 2009, 2010,2014; Cochrane et al. 2017, 2018).As well as providing the sample sizes for population studies suchas these, HiZELS has also provided parent samples for more detailedfollow-up observations (Sobral et al. 2013b; Magdis et al. 2016; MNRAS in press, 1–17 (2021)
HiZELS-14: Resolving a dusty, star-forming galaxy at 𝑧 = . Stott et al. 2016; Molina et al. 2017, 2019; Gillman et al. 2019). Inparticular, by exploiting the wide area HiZELS coverage, a sampleof bright H 𝛼 emitters ( 𝑓 H 𝛼 > . × − erg s − cm − ) which bychance lie within 30 (cid:48)(cid:48) of bright natural guide stars ( 𝑅 <15) couldbe identified and targeted for IFU spectroscopy of the H 𝛼 line us-ing adaptive optics with the SINFONI Integral Field Unit on theVery Large Telescope (VLT). This campaign, known as SINFONI-HiZELS (SHiZELS), yielded high-resolution spectral maps for 20galaxies at 𝑧 = . 𝑧 = .
47 and 𝑧 = .
23 at ∼ . (cid:48)(cid:48) (rest-frame ∼ ∼ . (cid:48)(cid:48) resolution with ALMA (Band 6 or 7, dependingon redshift), to map the dust continuum emission (see Cheng et al.2020). UVIS Imaging in the rest-frame UV (F606W) and rest-frameoptical (F140W) filters obtained during HST
Cycle 24 completesthis dataset. We now have FIR-UV-H 𝛼 matched-resolution observa-tions of a small sample of HiZELS galaxies. Since these galaxiesare H 𝛼 -selected, they are likely to be a less biased sub-sample ofthe high-redshift star-forming galaxy population than UV-selectedsamples, which target the bluest and least dusty galaxies, at an epochwhere dust is important (see Oteo et al. 2015).Here, we present data for SHiZELS-14, which is the brightest,most extended and more extreme source in our sample. SHiZELS-14(10:00:51:6 +02:33:34.5) is a 𝑧 = .
24 galaxy, with high stellar mass( 𝑀 ∗ ∼ M (cid:12) ; Swinbank et al. 2012a; Laigle et al. 2016), anda star-formation rate of ∼ (cid:12) yr − These properties enable adetailed investigation of the multi-wavelength extended structures ofthis galaxy. In the following subsections, we provide details of thenew high-resolution imaging we have recently obtained as part of theSHiZELS campaign. We present new radio continuum imaging fromthe JVLA (at comparable angular resolution to the other new imag-ing), which were obtained only for SHiZELS-14. We also describethe existing data available for our multi-wavelength characterisationof this galaxy. H 𝛼 emission from SINFONI SINFONI observations of SHiZELS-14 took place in March 2010,in good seeing and photometric conditions ( ∼ . (cid:48)(cid:48) ), with total ex-posure time 12ks (each individual exposure was 600s). This yieldedthe sub-kpc resolution H 𝛼 map shown in the lower left-hand panelof Figure 1. SHiZELS-14 was the only 𝑧 ∼ . 𝑧 ∼ . ESOREX data reductionpipeline was used to perform extraction, flat fielding and wavelengthcalibration, and to create the data cube for each exposure. Thesedata cubes were then stacked and combined using an average witha 3 𝜎 clip, to reject cosmic rays and sky line residuals. Flux cal-ibration was performed using observations of standard stars takenimmediately before/after science exposures, which were reduced inthe same way. H 𝛼 and [Nii] 𝜆𝜆 , 𝜒 minimisation procedure. Thisyielded intensity, velocity, and velocity dispersion maps. An angularresolution of ∼ . (cid:48)(cid:48) was achieved. The spectral resolution of theinstrument is 𝜆 / Δ 𝜆 ∼ 𝛼 flux derived from the SINFONI observations of SHiZELS- 14 was 1 . ± . × − erg s − cm − . The H 𝛼 -derived effectiveradius is 4 . ± . 𝑉 rot / 𝜎 = . ± . ±
40 km s − ) peak-to-peak velocity gradient. Swinbanket al. (2012a) comment that the one- and two-dimensional velocityfields are consistent with an early-stage prograde encounter. Thissuggests that the disordered morphology and extreme star formationmay be related to a merger event. HST
SHiZELS-14 was observed over two
HST orbits during Cycle 24(Program 14719, PI: Best). One orbit (2700 s exposure) used theWFC3/UVIS F606W filter, and the other used the WFC3/IR F140Wfilter. Orbits were split into a 3-point dither pattern in the UVISchannel, as a compromise between maximising sensitivity and sub-sampling the point spread function (PSF). Since angular resolutionwas preferred over sensitivity in the IR channel, a 4-point ditherpattern was used for these orbits. At 𝑧 = .
24 (the redshift ofSHiZELS-14), the filters correspond to the rest-frame near-UV at ∼ ∼ HST images are made using standard
HST procedures and shown in the upper panels of Figure 1. We derivethe effective radius of the F140W image via a two-dimensional Sér-sic profile fit, obtaining effective radius along the semi-major axis 𝑅 maj 𝑒, opt = . ± . 𝛼 measure-ment) and axial ratio 𝑞 = .
64 (with a Sérsic index fixed at 𝑛 = 𝑛 = . SHiZELS-14 was observed with ALMA during August 2016 as partof ALMA Cycle 3 (project code 2015.1.00026.S, PI: Ibar). Our ob-servations, taken in configuration C36-5, used Band 6 (260 GHz,7 . 𝜇 m) emission ofSHiZELS-14 at ∼ . (cid:48)(cid:48) resolution.The image was manually cleaned down to 3 𝜎 (rms ∼ 𝜇 Jy beam − ) at the source position. We used Briggs (robust=0.5)visibility weighting, which assigns higher weights to longer base-lines, producing an image with higher angular resolution (see theimage in the lower right-hand panel of Figure 1). To investigate theimpact of visibility weighting on the reduced ALMA image, we re-imaged the ALMA data using a natural weighting, which weightsvisibilities only by the rms noise (see the left-hand panel of Figure2). This method minimises the noise level but provides poorer angu-lar resolution, given that the density of visibilities falls towards theoutskirts of the 𝑢𝑣 -plane and there is thus higher noise in the longer-baseline visibilities. Using the re-reduced, lower angular resolutionnatural-weighted image, we probe to slightly lower flux density perbeam. This will be used to assess the quality of our astrometric cali-bration in Section 2.8.SHiZELS-14 has an observed-frame 260 GHz flux density of2 . ± . ∼ MNRAS in press, 1–17 (2021)
R.K. Cochrane et al. D e c HST UV (F606W) h m s s s s RA D e c SINFONI H1kpc h m s s s s RA ALMA beamALMA 260GHzHST IR (F140W)
Figure 1: Astrometrically-calibrated, high-resolution observations of SHiZELS-14 in the rest-frame UV (
HST
F606W filter; upper left panel),rest-frame optical (
HST
F140W filter; upper right panel), H 𝛼 (SINFONI/VLT; lower left) and dust continuum (rest-frame 370 𝜇 m imagingfrom ALMA, reduced with Briggs weighting; lower right). The red contours on all panels outline the ALMA dust continuum emission at 50,200, and 300 𝜇 Jy beam − . The green contours outline the 3 𝜎 emission H 𝛼 emission from SINFONI as described in Section 2.1. Pale greycontours outline the peak of the F140W image. The emission imaged by SINFONI, ALMA and the HST
F140W filter span the same extendedregion, but display very different morphologies. The peaks of the short-wavelength emission are clearly offset from the peaks of the dustcontinuum emission. This is particularly striking for the F606W UV emission, which is concentrated in regions with little dust emission anddoes not extend down to the southern regions that are clearly probed by the other bands.emission. We derive this radius using multiple methods. First, wefit a Gaussian model with varying axis ratio in the 𝑢𝑣 -plane, using CASA ’s uvmodelfit task (see Figure A1). The effective radius alongthe semi-major axis, 𝑅 maj 𝑒 , is 4 . ± . 𝑞 = . ± .
01. A two-dimensional Sérsic profile fit in the imageplane yields 𝑅 maj 𝑒, FIR = . ± . 𝑞 = .
47 (with the Sérsicindex fixed at 𝑛 =
1; allowing this to vary gives 𝑛 = . 𝛼 and the HST rest-frame optical data. We make use of the deep existing radio observations in the COSMOSfield from the VLA-COSMOS surveys. The VLA-COSMOS LargeProject (Schinnerer et al. 2007) surveyed 2 square degrees in VLA A-and C-array configurations at 1 . ∼ − 𝜇 Jy beam − at angular resolution1 . (cid:48)(cid:48) × . (cid:48)(cid:48) . The VLA-COSMOS Deep project (Schinnerer et al.2010) added further A-array observations at 1 . . at a MNRAS in press, 1–17 (2021)
HiZELS-14: Resolving a dusty, star-forming galaxy at 𝑧 = . RA D e c ( I C R S ) ALMA : natural, BriggsHST F140W RA JVLA 6GHz RA JVLA 3GHz
Figure 2: Left: the
HST
F140W image, with contours of the 260 GHz ALMA data with two weightings overlaid. The image produced usingnatural weighting is shown with purple contours tracing 25 𝜇 Jy beam − . The red contours outline the Briggs-weighted image (50, 200, and300 𝜇 Jy beam − , as in Figure 1). The slightly lower angular resolution natural-weighted image shows flux towards the North East and the SouthWest, in the regions with extended F140W flux. This gives us confidence in the astrometric alignment of the images. Centre: 0 . (cid:48)(cid:48) imagingfrom the VLA-COSMOS 3 GHz Large Project (Smolčić et al. 2017). Right: new, ∼ . (cid:48)(cid:48) , 4 − ∼ . 𝜇 Jy beam − at 0 . (cid:48)(cid:48) angularresolution.SHiZELS-14 is one of the sources detected by these VLA surveys.The measured flux densities are 𝑆 . = ± 𝜇 Jy and 𝑆 = ± 𝜇 Jy. From these two flux densities, we derive a radio spectralindex of 𝛼 = − . ± .
16 (where 𝑆 𝜈 ∝ 𝜈 𝛼 ), in good agreement withmeasurements of star-forming galaxies (Condon 1992). The lowerangular resolution of the radio images limits our ability to proberesolved structure (see Figure 2, centre panel), but the source is stillextended at 0 . (cid:48)(cid:48) resolution. We will use the total flux density toestimate a star-formation rate later in the paper. We obtained C-band (4 − ∼ . (cid:48)(cid:48) spatial resolution). Observations took place during October 2019, aspart of Cycle 19A (project 19A-205). We used 3C147 for flux cal-ibration, and J1024-008 for phase calibration. The data presentedwere obtained during a 4 hr observing block, with around 3 hrs ofon-source time.We reduced the data using standard CASA calibration pipelines,and manually cleaned the images down to 2 𝜇 Jy beam − . We presentour Briggs-weighted image in Figure 2. We obtain a total continuumflux density of 20 ± 𝜇 Jy at 6 GHz. This is only roughly half theexpected flux, based on the 1 . 𝛼 = − . A wealth of lower resolution data exists for this galaxy due to its lo-cation within the well-imaged COSMOS field (Scoville et al. 2007).At NUV-optical wavelengths, COSMOS was imaged in the 𝑢 ∗ -bandfrom the Canada-France-Hawaii Telescope (CFHT/MegaCam), andin six broad bands ( 𝐵 , 𝑉 , 𝑔 , 𝑟 , 𝑖 , 𝑧 + ), 12 medium bands (IA427, IA464,IA484, IA505, IA527, IA574, IA624, IA679, IA709, IA738, IA767,and IA827), and two narrow bands (NB711, NB816), all from theCOSMOS-20 survey (Subaru Suprime-Cam; Taniguchi et al. 2007,2015). 𝑌 -band imaging was obtained with Hyper-Suprime-Cam onSubaru (HSC; Miyazaki et al. 2012). In the NIR, 𝑌 , 𝐽 , 𝐻 , & 𝐾 𝑠 data are provided by the UltraVISTA-DR2 release (McCracken et al.2015), which uses the VIRCAM instrument on the VISTA telescope.These are supplemented by 𝐻 and 𝐾 WIRCAM data from CFHT(McCracken et al. 2010). Mid-IR data are drawn from IRAC chan-nels 1, 2, 3 and 4 (3 . 𝜇 m, 4 . 𝜇 m, 5 . 𝜇 m and 8 . 𝜇 m), collected bythe Spitzer
Large Area Survey with HSC (SPLASH survey; Lin et al.2017; Capak et al. in prep). Laigle et al. (2016) collate these obser-vations and provide an NIR-selected photometric redshift catalogue.For consistency, we use their 3 (cid:48)(cid:48) diameter aperture fluxes extractedfor SHiZELS-14. We tabulate these measurements in Table B1 of theAppendix, along with the new measurements from this paper.In Figure 3 we show deeper imaging from more recent surveys: the 𝑢 ∗ band (from the CLAUDS survey on CFHT; Sawicki et al. 2019), 𝑖 -band (from HSC-DR2; Aihara et al. 2019), and in the 𝐻 and K s bands from UltraVISTA-DR4. These show interesting differences,with emission in the 𝑢 ∗ -band peaking to the North East compared tothe 𝐾 𝑠 -band emission (this is not driven by astrometric offsets in the 𝑢 ∗ -band data). MNRAS in press, 1–17 (2021)
R.K. Cochrane et al. D e c CFHT u * = 374.3nm HSC i= 771.1nm
UltraVISTA H= 1.645 m
UltraVISTA K s = 2.154 m Figure 3: NUV-NIR imaging of SHiZELS-14 from CFHT (u* from the CLAUDS survey; Sawicki et al. 2019), Subaru (HSC-DR2; Aihara et al.2019), and VISTA (UltraVISTA DR4; McCracken et al. 2015). These observations are seeing-limited, with angular resolution ∼ . − . (cid:48)(cid:48) . Weshow the typical angular resolution on the CFHT 𝑢 ∗ image. We overlay contours from our resolved imaging on relevant panels. Overplotted onthe HSC 𝑖 -band image are contours from HST
F606W imaging (blue). The contours on the UltraVISTA 𝐻 -band image are from HST
F140Wimaging (orange). Both SINFONI H 𝛼 (green) and ALMA dust continuum emission (black) contours are overplotted on the UltraVISTA 𝐾 𝑠 -band image. We draw data at mid-IR and far-IR wavelengths from
Spitzer and
Herschel imaging. We adopt the 24 𝜇 m flux density from the Spitzer
Multiband Imaging Photometer (MIPS; Rieke et al. 2004). The
Her-schel
Multi-tiered Extragalactic Survey (HerMES; Oliver et al. 2012)targeted COSMOS at 100 − 𝜇 m. The survey used Herschel -Spectral and Photometric Imaging Receiver (SPIRE) at 250 𝜇 m,350 𝜇 m and 500 𝜇 m and the Herschel -Photodetector Array Cam-era and Spectrometer (PACS) at 100 𝜇 m and 160 𝜇 m. One of themain aims of the Herschel
Extragalactic Legacy Project (HELP) was to develop the advanced statistical tools needed to de-blendthe low-resolution data from Herschel , in order to assign fluxes tocomponents (Hurley et al. 2017; Pearson et al. 2017). We use thesepublicly available, catalogued flux densities for SHiZELS-14 (seeTable B1).We also use the ALMA Band 7 flux density measured by Scovilleet al. (2014). The observed-frame continuum 350 GHz total flux den-sity is 4 . ± . . ± . − .SHiZELS-14 is also one of the sources in the catalogue of bright sub-mm sources detected by SCUBA-2 in the COSMOS field (Simpsonet al. 2019). The observed-frame 850 𝜇 m flux density measured thereis 5 . ± . Accurate astrometric alignment is critical when comparing multi-wavelength emission on small angular scales. However, due to thesmall fields of view of both the SINFONI ( ∼ (cid:48)(cid:48) × (cid:48)(cid:48) ) and ALMA( ∼ (cid:48)(cid:48) diameter) data, aligning the images is non-trivial. Here, wedescribe the alignment of the images.The ALMA image is expected to be tied to the International Ce-lestial Reference System (ICRS). Although calibration errors andself-calibration processes can lead to astrometric offsets, this is un-likely to be larger than the pixel level (0 . (cid:48)(cid:48) ). The JVLA data shouldalso be on the ICRS, and the spatially coincident emission seen by theJVLA and ALMA (Figure 2, right-hand panel), give us confidencein the astrometry of both. http://herschel.sussex.ac.uk We then align all other images to the ICRS. The SINFONI H 𝛼 im-age was aligned to the same reference frame as the main wide-fieldHiZELS survey images. We used a broad-band-subtracted narrow-band image from HiZELS, which had been aligned to the Two MicronAll-Sky Survey (2MASS), which itself uses the ICRS. We shifted theH 𝛼 image obtained from the SINFONI cube by sub-pixel quantities,and convolved down to the resolution of the broad-band image. Sub-tracting the images enabled a 𝜒 fit to define the optimal alignment.Based on these comparisons, we are able to achieve an accuracy onthe SINFONI image alignment of ∼ . (cid:48)(cid:48) .We calibrated the astrometry of the HST images by aligning toHSC-DR2 (Aihara et al. 2019), which inherits its astrometric accu-racy from
Gaia . Sources were extracted from the HSC 𝑌 and 𝑅 -bandimages, as well as the two HST images, using the SE
XTRACTOR soft-ware (Bertin & Arnouts 1996). We matched sources detected in theHSC 𝑌 -band and the HST
F140W (and then the 𝑅 -band and the HST
F606W), and constructed histograms of the small offsets betweentheir RA and Dec positions. The peaks of these histograms wereselected as the offset to be applied to both the
HST images (the sameprocedure was used in the companion paper, Cheng et al. 2020). Wealso performed this process using catalogued 2MASS sources, andderived essentially identical results. Based on this, and the narrowwidths of both histograms, we estimate that the alignment is cor-rect to within ∼ . (cid:48)(cid:48) . Inspecting our images, gives us confidence inthe alignment. As shown in the middle panel of Figure 2, there isfaint ALMA flux in the regions that show extended F140W flux. TheF140W image also aligns with the SINFONI image in terms of botharea covered and areas where the flux peaks. Figure 1 shows our four spatially resolved maps after these smallastrometric corrections were applied. The emission in all bands isaligned along the same axis. However, the peak of the dust emissionprobed by ALMA (and confirmed by the 4 − 𝛼 emission. Theseoffsets are far larger than the residual astrometric uncertainties (0 . − . (cid:48)(cid:48) ). The dust emission is centrally concentrated, whereas there area number of H 𝛼 peaks along the extended region where dust emissionis faint. There is a peak in the emission from both HST bands towardsthe North-East of the image, yet no detectable dust emission. This
MNRAS in press, 1–17 (2021)
HiZELS-14: Resolving a dusty, star-forming galaxy at 𝑧 = . Figure 4: Data presented in Table B1, fitted with
MAGPHYS (Da Cunha et al. 2008). The red points are the observational data, and the openblack circles show the model results. The blue line shows the intrinsic stellar SED for the best-fitting model, and the black line shows theSED after dust reprocessing. Residuals are shown in the lower panel. The fitting yields SFR = ±
30 M (cid:12) yr − , log 𝑀 ∗ / M (cid:12) = . ± . 𝑀 dust / M (cid:12) = . ± . 𝐴 𝑣 = . ± . ( 𝐿 TIR / L (cid:12) ) = . ± . 𝑢 ∗ -band emission (comparedto the longer wavelength bands) shown in Figure 3. Such offsets areseen in observed dusty galaxies (e.g. Chen et al. 2015, 2017), andalso in simulations (Cochrane et al. 2019). Before examining the resolved structures of SHiZELS-14 further,we place these into context by deriving the global properties of thegalaxy.
MAGPHYS
Spectral Energy Distribution (SED) fitting provides a powerful ba-sis for estimating galaxy properties from photometry. The
MAGPHYS energy balance SED fitting code (Da Cunha et al. 2008, 2015; Battistiet al. 2019) was used to derive physical parameters in Cheng et al.(2020). We provide details of the fitting here.
MAGPHYS employs anenergy balance method to match the attenuation of the stellar emis-sion in the UV/optical by dust, and the re-radiation of this energyin the far-infrared. The code uses the stellar population models ofBruzual & Charlot (2003), with a Chabrier IMF (Chabrier 2003)and metallicities between 0.2 and 2 times solar. The star-formationhistory (SFH) is parametrised as a continuous delayed exponentialfunction and to reproduce starbursts,
MAGPHYS also adds bursts tothe star-formation history. Dust attenuation is modelled using twocomponents following Charlot & Fall (2000). The code was ex-tensively tested with both observational constraints on SMGs andagainst model star-forming galaxies from the EAGLE simulation byDudzeviciute et al. (2020) and shown to perform well for these highlydust-obscured galaxies.We fit the photometry presented in Table B1. We estimatelog 𝑀 ∗ / M (cid:12) = . ± .
1, log 𝑀 dust / M (cid:12) = . ± .
1, andSFR = ±
30 M (cid:12) yr − (see Table 1). The estimated total infraredluminosity is log ( 𝐿 TIR / L (cid:12) ) = . ± .
01, and the estimateddust attenuation in the 𝑉 -band is 𝐴 𝑣 = . ± .
1. We obtain con-sistent results using another SED fitting code,
BAGPIPES (see ap-pendix). We have also used
BAGPIPES to experiment with different Basic property Measurement ReferenceRA (J2000) 10:00:51.6 Swinbank+12Dec (J2000) +02:33:34.5 Swinbank+12 𝑧 H 𝛼 𝑀 ∗ , SED / M (cid:12) . ± . 𝑀 gas / M (cid:12) . ± . 𝑀 dust / M (cid:12) . ± . 𝐿 TIR / L (cid:12) . ± .
01 This paperSFR
SED / M (cid:12) yr − ±
30 This paper 𝑅 𝑒, H 𝛼 / kpc 4 . ± . 𝑅 maj 𝑒, opt / kpc 4 . ± . 𝑅 maj 𝑒, FIR / kpc 4 . ± . 𝐴 𝑣 . ± . 𝑧 = . To assess the sensitivity of the derived
MAGPHYS parameters in thefar-infrared, we also fit the MIR-to-FIR SED of SHiZELS-14 sepa-rately, using only data from ALMA and
Herschel . We parametrise theemission from cold and warm dust using a simple two-body model: 𝑓 𝜈 ( mJy ) = A warm 𝜆 − 𝛽 warm B 𝜈 ( T warm ) + A cold 𝜆 − 𝛽 cold B 𝜈 ( T cold ) , (1)where 𝐴 warm and 𝐴 cold are normalisations and 𝐵 𝜈 ( 𝑇 ) is the Planckfunction, from dust grains radiating at rest frequency 𝜈 , at temper- MNRAS in press, 1–17 (2021)
R.K. Cochrane et al. / m [observed frame] S o b s / m J y T BBdust = 32 ± 2Klog M dust / M = 9.0 ± 0.1 IRACMIPSPACSSPIREALMAVLA
Figure 5: The dust SED of SHiZELS-14, constructed using col-lated archival data and the new ALMA 260GHz data. Error bars areplotted on the data points, but are small. A two grey body modelparametrisation (blue fitted curve) provides a good fit to both thecold and warm dust components. Integrating the 8 − 𝜇 m emis-sion gives log ( 𝐿 TIR / L (cid:12) ) = . ± .
02 and SFR − 𝜇 m = ±
50 M (cid:12) yr − . A single component model (grey fitted curve) isunable to fit the 100 𝜇 m PACS data. The characteristic dust temper-ature derived from this fit ( 𝑇 BBdust = ± 𝑀 dust / M (cid:12) = . ± . 𝛽 = EMCEE
MCMC python package (Foreman-Mackey 2016),with 300 walkers and 5000 steps. This yields posterior estimates:log 𝐴 warm = . ± . 𝑇 warm = ± 𝐴 cold = . ± . 𝑇 cold = ± 𝛽 cold and 𝑇 cold ,and a 5-parameter fit that allows 𝛽 cold to vary favours a higher 𝛽 cold and a lower 𝑇 cold .To derive the dust temperature in a consistent way to other studiesin the literature, we additionally fit a single modified black bodymodel to the data, also shown in Figure 5. We derive a character-istic dust temperature of 𝑇 BBdust = ± 𝜇 m PACS data point, and it is necessary to boost the errors onthat point artificially to get a good fit to the longer wavelength data.This suggests a contribution from hotter dust, perhaps indicative ofan obscured AGN. We will discuss this further in Section 3.6. Assuming the dust is optically thin at the rest-frame frequency, thedust mass is given by (e.g. James et al. 2002): 𝑀 dust = + 𝑧 𝑆 obs 𝐷 𝐿 𝜅 rest 𝐵 𝜈 ( 𝑇 BBdust ) , (2)where 𝑆 obs = . 𝐷 𝐿 is the luminosity distance, 𝜈 =
836 GHz is therest-frame frequency, 𝜅 rest is the mass absorption coefficient at thisfrequency, and 𝑇 BBdust = ± 𝜅 = . ± .
02 m kg − (James et al. 2002), which giveslog 𝑀 dust / M (cid:12) = . ± . MAGPHYS fits, log 𝑀 dust / M (cid:12) = . ± . ( 𝑀 dust / 𝑀 ∗ ) = − . ± .
2, which is comparable to the ratiosderived by Calura et al. (2017) for SMGs of stellar mass ∼ M (cid:12) at 𝑧 ∼ − − 𝜇 m within the MCMC fit (enabling us to foldin the correlations between fitted parameters), obtaining an estimatefor the total IR luminosity, log ( 𝐿 TIR / erg s − ) = . ± . ( 𝐿 TIR / L (cid:12) ) = . ± .
02. The TIR-based SFR is950 ±
50 M (cid:12) yr − (using the Kennicutt & Evans 2012 SFR calibra-tion, with a Kroupa 2002 IMF). IRX − 𝛽 relation The IRX − 𝛽 relation (Calzetti et al. 1994; Meurer et al. 1999) be-tween the ratio of the FIR and UV luminosity (IRX = L FIR / L )and the spectral slope ( 𝛽 , where 𝑓 𝜆 ∝ 𝜆 𝛽 ) evaluated at 1600 Å is apopular method used to infer SFRs where only rest-frame UV lumi-nosities are available. This appears to work for samples of galaxieswith relatively low dust content (especially at very high redshift).However, individual galaxies show a large amount of scatter aroundthis relation, and it has been shown that this method is not appropri-ate highly star-forming galaxies (e.g. Casey et al. 2014; Chen et al.2017; Narayanan et al. 2018), although it is difficult to identify thesebased on their UV properties.We can derive both IRX and 𝛽 for SHiZELS-14. We use the pub-licly available HST 𝐼 -band image ( 𝜆 mean = 𝜆 mean = 𝜆 mean = 𝜆 mean = 𝛽 . Adopting our de-rived 𝛽 = − . ± .
1, and applying the relation 𝐴 = . + . 𝛽 ,we derive 𝐴 = . ± . 𝐴 estimated fromscaling the 𝐴 𝑉 obtained via SED fitting; see Table 2). Correct-ing the global SFR inferred from the FUV flux accordingly yieldsSFR = + − M (cid:12) yr − . This estimate is approximately two timeslower than the SFR inferred from the SED fitting presented in Sec-tions 3.1 and 3.2, and three times lower than the TIR-derived SFRin Section 3.3. We calculate IRX using the TIR luminosity derivedin Section 3.3, and the rest-frame 1851 Å luminosity. Globally, thegalaxy has log IRX = . ± .
06. In combination with the derived 𝛽 , this places it ∼ . ± . In Table 2 we present global SFR estimates from global measure-ments in different wavebands, using the calibrations of Kennicutt &Evans (2012) and assuming a Kroupa (2002) IMF. It is clear that ap-plying standard SFR calibrations to flux measurements that probe starformation via direct emission at shorter wavelengths predict vastlylower SFRs than the dust-obscured tracers. This suggests that thedeficit in global SFR derived from the dust-sensitive SFR indicators
MNRAS in press, 1–17 (2021)
HiZELS-14: Resolving a dusty, star-forming galaxy at 𝑧 = . Waveband (Instrument) Formula for log ( SFR / M (cid:12) yr − ) SFR / M (cid:12) yr − SFRs from individual tracers
TIR − 𝜇 m (dust SED fit) log ( 𝐿 TIR / erg s − ) − . ± Radio (1.4 GHz, VLA, Bell 2003 conversion) log ( 𝐿 . , rest / erg s − Hz − ) − .
43 1180 ± ( 𝐿 . , rest / erg s − Hz − ) − . ± 𝛼 (SINFONI/VLT) log ( 𝐿 H 𝛼 / erg s − ) − .
27 33 ± 𝐿 H 𝛼 corrected using 1 mag dust extinction 83 ± 𝐿 H 𝛼 corrected using 𝑀 ∗ -dependent dust extinction 180 ± 𝑀 ∗ / M (cid:12) = . 𝐿 H 𝛼 corrected using 𝐴 H 𝛼 = . ± .
1, derived from scaled 𝐴 𝑉 ± 𝐿 H 𝛼 corrected using 𝐴 H 𝛼 = . ± .
1, derived from scaled 𝐴 𝑉 & 1000 ± HST
F606W) log ( 𝜈𝐿 𝜈 / erg s − ) − .
17 13 ±
1” corrected using 𝐴 derived from 𝛽 , with 𝛽 = − . ± . + − ” corrected using 𝐴 UV = . ± .
2, derived from scaled 𝐴 𝑉 ± 𝐿 𝜈, corr = 𝐿 𝜈, obs + . 𝐿 TIR , 𝐿 𝜈 − SFR conversion above 440 ± 𝛼 + TIR 𝐿 H 𝛼, corr = 𝐿 H 𝛼, obs + . 𝐿 TIR , 𝐿 H 𝛼 − SFR conversion above 330 ± 𝐿 FUV , corr = 𝐿 FUV , obs + . × 𝐿 . ± MAGPHYS ± BAGPIPES ± < (cid:12) yr − . These UV or H 𝛼 -inferred SFRs lie well below the valuesderived from the TIR or radio (SFR = − (cid:12) yr − ). Applying either a standard dust correction corresponding to 𝐴 H 𝛼 = 𝛼 -derived SFR (SFR = −
200 M (cid:12) yr − ). Similarly, correctingthe UV-derived SFR using the IRX- 𝛽 relation derived using HST data only raises the SFR to 300 M (cid:12) yr − . We also estimate 𝐴 H 𝛼 and 𝐴 by scaling the 𝐴 𝑉 derived from MAGPHYS according to a Calzetti et al. (2000) law. This correction brings the UV-derived SFR into betteragreement with the
MAGPHYS -derived SFR, however the H 𝛼 -derived SFR remains low, indicating additional extinction of the 𝐴 H 𝛼 line.Including additional extinction of the 𝐴 H 𝛼 line according to Charlot & Fall (2000) brings the H 𝛼 -derived SFR into line with the FIR estimate(SFR ∼ (cid:12) yr − ).is due to the highly dusty nature of this galaxy. In the followingsection, we explore the differences in the spatially-resolved SFRs,derived at different wavelengths. As discussed in Section 3.5, the SFRs derived from the UV, H 𝛼 and FIR differ greatly. In this section, we investigate whether thepresence of an active galactic nucleus (AGN) could be a factor inthis. In this scenario, the extreme dust continuum emission towardsthe centre of the galaxy could be powered by heating from a centralAGN, rather than a compact region of star formation. Since differenttypes of AGN emit in different wavebands (see Heckman & Best2014 for a review, and Garn et al. 2010 for a discussion of AGNwithin the HiZELS sample), identification of AGN requires a multi-wavelength approach. Here, we use some of the key methods forAGN identification to hunt for signs of AGN activity. X-ray emission probes the accretion disk corona very close to asupermassive black hole. The
Chandra
COSMOS-Legacy Survey(Civano et al. 2016) imaged 2 . of the COSMOS field in the wavelength range 0 . −
10 keV. SHiZELS-14 lies in the central regionof the COSMOS-Legacy field, where effective exposure times are ∼
160 ks. The limiting depths are 2 . × − , 1 . × − , and8 . × − erg cm − s − in the bands 0 . −
2, 2 −
10, and 0 . −
10 keV,respectively. At these limiting depths, SHiZELS-14 is undetected.We derive a limit on the rest-frame hard-band 2 −
10 keV luminosityfollowing Alexander et al. (2003): 𝐿 − , lim = 𝜋𝐷 𝐿 𝑓 − , lim ( + 𝑧 ) Γ − , using Γ = . 𝐿 − < . × erg s − .We can predict the X-ray luminosity associated with star formationusing the 𝐿 − -SFR calibration proposed by Kennicutt & Evans(2012) and the SFR measured from the other indicators. Given theSFR derived from the dust SED fit, 950 ±
50 M (cid:12) yr − , we estimate 𝐿 − = ( . ± . ) × erg s − . This is an order of magnitudelower than the limit imposed by the survey depth. Therefore, thelack of an X-ray detection is consistent with SHiZELS-14 being astar-forming galaxy. II ]/ H 𝛼 excess The ratio of [N II ]-to-H 𝛼 line flux reflects the hardness of the ionisingsource driving the nebular emission, and hence can be used to inferthe presence of an AGN, often in combination with other line ratios MNRAS in press, 1–17 (2021) R.K. Cochrane et al. (e.g. Baldwin et al. 1981). For SHiZELS-14, there is no evidencefor a strong excess in [N II ]/H 𝛼 . For the clump nearest the peakof the rest-frame FIR emission, [N II ]/H 𝛼 = .
12 (Swinbank et al.2012b, clump 14a), well within the range expected for star-formingregions (e.g. Kewley et al. 2006). Swinbank et al. (2012a) show thatthe [N II ]/H 𝛼 radial profile of SHiZELS-14 is slightly negative, inline with the rest of the SHiZELS sample. The derived gradientsreflect slightly enhanced metallicity towards the central regions ofthe SHiZELS galaxies, consistent with simulations of star-forminggalaxies of similar mass and redshift (e.g. Ma et al. 2017). Obscured AGN are characterised by a strong mid-infrared (rest-frame ∼ − 𝜇 m) excess, produced by a dusty obscuring torus. Our MAGPHYS fit (Figure 4) shows no sign of such an excess, being well-fitted by an SED constructed without AGN templates. Fitting theSED with
CIGALE , which does allow for the inclusion of emissionfrom AGN, provides no evidence of an AGN ( 𝑓 AGN , best = . ± The ratio of IR to radio luminosity (e.g. Appleton et al. 2004) isfrequently employed to separate radio-loud AGN from star-forminggalaxies. Following Ivison et al. (2010), we use the following equa-tion with the TIR luminosity calculated in Section 3.3: 𝑞 TIR = log (cid:32) 𝐿 TIR . × W (cid:33) − log (cid:32) 𝐿 . WHz − (cid:33) . (3)The rest-frame 1 . 𝐿 . , rest = 𝜋𝐷 𝐿 ( + 𝑧 ) + 𝛼 (cid:32) 𝜈 . 𝜈 obs (cid:33) 𝛼 𝑆 . , obs = . ± . WHz − . (4)We assume a spectral index 𝛼 = − .
77, derived from the VLA 3 GHzand 1 . 𝑞 TIR = . ± .
10. This is broadlyin line with the distribution of 𝑞 TIR values for the 250 𝜇 m-selectedsample of Ivison et al. (2010) (median 𝑞 TIR = . 𝜎 𝑞 = .
24; seealso Algera et al. 2020). The 𝑞 TIR value for SHiZELS-14 is wellwithin 1 𝜎 of the median relation derived for star-forming galaxies.This indicates that the radio continuum emission is not contaminatedby a compact radio core. Overall, we find no evidence that SHiZELS-14 is host to a radio-loud AGN. In Figure 6, we present maps of SFR surface density, derived foreach of the four SFR tracers using the luminosity-SFR calibrationsof Kennicutt & Evans (2012) and Bell (2003). In order to do this,we assume that these global calibrations are also valid on smallerspatial scales, which may not be the case. In reality, gradients in dusttemperature and opacity (e.g. Galametz et al. 2012) may apply to theTIR model, and gradients in the reddening will influence the H 𝛼 andUV maps. The radio emission is sensitive to cosmic ray propagationand starburst age, which likely results in smoother and more extendedemission than the true SFR distribution (Thomson et al. 2019). Mak-ing this assumption and adopting standard SFR calibrations, it isclear from Figure 6 that the SFRs derived from the four indicatorsdiffer across the galaxy. To investigate this more quantitatively, wederive star-formation rate radial profiles by applying Kennicutt &Evans (2012) calibrations to the rest-frame FUV F606W, H 𝛼 , andTIR flux maps (see Figure 7, thick dashed lines). The three profilesare discrepant, with the TIR-based SFR profile increasing sharplytowards the centre, and the FUV-derived profile decreasing at radiismaller than ∼ 𝛼 -derived SFRsare lower than the FIR-derived SFR across the radial extent of thegalaxy. The FUV profile broadly follows the H 𝛼 profile in shape,but with a different normalisation. The FUV is most strongly atten-uated by dust, and yields the lowest dust-uncorrected SFRs acrossthe galaxy. Thus, the discrepancy between the SFRs derived globallycannot be attributed solely to the compact dusty centre of the galaxy,though this is where the measurements are most discrepant. Instead,short-wavelength light is attenuated across the galaxy.We also show the effects of applying a dust correction. 𝐴 H 𝛼 and 𝐴 UV are calculated from the MAGPHYS -derived 𝐴 𝑉 , according tothe Calzetti et al. (2000) law and a Charlot & Fall (2000) birth cloudattenuation. These dust corrections bring the outermost regions ofthe FUV and H 𝛼 profiles further towards agreement at radii greaterthan ∼ 𝛼 -derived SFR estimates are much lower than the TIR-derived estimate in the centre, particularly at radii less than ∼ H 𝛼 and FIR maps In Figure 6, we showed that the SFR surface densities derived in dif-ferent wavebands from dust-uncorrected fluxes of the dust-sensitivetracers are far lower than the TIR measurement. We can use thisto estimate the spatially-resolved dust attenuation. In the left-handpanel of Figure 8, we present the ratio of the H 𝛼 -derived SFR (withno dust correction applied) to the TIR-derived SFR. We can alsouse this ratio of the fluxes to estimate 𝐴 H 𝛼 in a spatially-resolvedway, as follows. Folding in a dust-correction to the H 𝛼 flux, and thenequating the two SFRs:SFR / M (cid:12) yr − = L TIR × − . = L H 𝛼 × − . × . H 𝛼 (5) MNRAS in press, 1–17 (2021)
HiZELS-14: Resolving a dusty, star-forming galaxy at 𝑧 = . D e c UV S F R / M y r k p c H r = 2 kpc, 4 kpc, 6 kpc S F R / M y r k p c h m s s s s RA D e c FIR S F R / M y r k p c h m s s s s RA Radio S F R / M y r k p c Figure 6: Maps of SFR surface density, derived for each of the four SFR indicators using the luminosity-SFR calibrations of Kennicutt &Evans (2012) (upper two and lower left panels) and Bell (2003) (lower right panel). Note that dust corrections were not applied to the UV orH 𝛼 maps. Pixels that have fluxes below the minimum of our Σ SFR scale (0 .
07 M (cid:12) yr − kpc − ) are coloured white to avoid overly noisy images.We plot the maps on the same SFR scale, to compare the SFRs directly, and show the position of the peak of the ALMA emission as a blackcross on each panel. We also overplot three concentric rings, of radii 2 kpc, 4 kpc and 6 kpc. It is clear that the derived SFRs differ across thespatial extent of the galaxy, not only in its dusty centre. The UV map shows a ’hole’ where the rest-frame FIR emission peaks. As shown inFigure 2, the angular resolution of the radio imaging is lower than the other images, which causes the emission to appear more extended.yields an expression for 𝐴 H 𝛼 : 𝐴 H 𝛼 = . (cid:32) L TIR L H 𝛼 (cid:33) − . . (6)Note that this method assumes that H 𝛼 and FIR flux are tracing onlyrecently formed stars, and sensitive to star formation on the sametimescales.We plot the spatially-resolved 𝐴 H 𝛼 in the right-hand panel ofFigure 8. 𝐴 H 𝛼 substantially exceeds 1, the canonical value appliedto global studies, across the spatial extent of the galaxy. The derived 𝐴 H 𝛼 is larger than that derived from scaling 𝐴 𝑉 according to theCalzetti et al. (2000) law and Charlot & Fall (2000) prescription( 𝐴 H 𝛼 = .
7) in the dustiest parts of the galaxy. In the most dustycentral region, it reaches a peak of 𝐴 H 𝛼 ∼
5. In fact, the true valueis likely to be above that due to gradients in the dust temperature andopacity.
While the H 𝛼 emission traces broadly the same spatial extent as therest-frame FIR emission, the rest-frame UV emission is concentratedtowards the North East of the galaxy. Assuming that H 𝛼 and UV areprobing the same star formation, we can predict the observed UVflux, 𝐼 obs , UV , from the observed H 𝛼 flux, 𝐼 obs , H 𝛼 , using the 𝐴 H 𝛼 map shown in Figure 8 and: 𝐼 int , H 𝛼 = 𝐼 obs , H 𝛼 . 𝐴 H 𝛼 = . . × 𝐼 obs , UV . 𝐴 UV . (7)The predicted UV flux map is highly dependent on the assumed re-lation between 𝐴 UV and 𝐴 H 𝛼 ; if we account for extra attenuationtowards birth clouds according to Charlot & Fall (2000), the pre-dicted UV flux is slightly higher than observed, and extends towardsthe South West end of the galaxy. If we use a lower 𝑘 H 𝛼 based on thecontinuum 𝑘 𝜆 , the flux falls below the noise level of the HST imageacross the galaxy’s spatial extent. MNRAS in press, 1–17 (2021) R.K. Cochrane et al. S F R / M y r k p c UVHFIR
Figure 7: Star-formation rate surface density profiles derived usingrest-frame FUV F606W, H 𝛼 , and rest-frame FIR flux map (scaled tothe SFR derived from fits to the dust SED; note that this assumes aconstant dust temperature across the galaxy). The profiles are centredon the peak of the rest-frame FIR emission, shown by a black cross inFigure 6. The thick dashed lines show the surface densities derivedusing Kennicutt & Evans (2012) calibrations, with no dust correc-tions applied. The solid transparent lines show the profiles derivedusing an 𝐴 H 𝛼 = . 𝐴 UV = .
3, derived using 𝐴 𝑉 = . MAGPHYS , the attenuation curve of Calzetti et al. (2000) and thepreferential attenuation towards birth clouds proposed by Charlot &Fall (2000). These corrections can bring the profiles broadly into lineat large radii, but still underestimate the star-formation rate surfacedensity at radii less than ∼ 𝛼 image and the 𝐴 H 𝛼 map. This may imply thatthe recent star-formation (probed by H 𝛼 ) is attenuated so strongly asto be undetectable in our F606W HST image. In this case, the UVflux that we do observe is tracing star formation on slightly longertimescales. This scenario is consistent with the peak of the stellarmass lying towards the North East of the H 𝛼 flux (see the F140Wimage). Indeed, qualitatively, we are broadly able to model the ob-served UV emission by assuming that, before obscuration, the UVlight traces the same region as the optical image. If we then applya dust attenuation map like the one shown in Figure 8, we recovera peak of UV emission in the region that is observed. A detailedquantitative comparison of this would require assumptions about therelation between rest-frame UV and rest-frame optical light, whichis sensitive to the age and metallicity of the starburst. SHiZELS-14 was identified by the HiZELS survey, which uses anH 𝛼 -based selection and largely probes typical star-forming galaxies,assuming typical extinction. However, it displays a number of ex-treme properties including high star-formation rate, dust mass anddust attenuation, and a TIR luminosity that places it in the ULIRGregime. In this section, we examine SHiZELS-14 in the context ofsub-millimeter galaxies at similar redshifts. We compare SHiZELS-14 to galaxies from the ALMA follow-upof the SCUBA-2 Cosmology Legacy Survey’s UKIDSS-UDS field(AS2UDS; Stach et al. 2018, 2019; Dudzeviciute et al. 2020). Thisis drawn from a ∼ SCUBA-2 survey. The ALMA pointingstarget ∼
700 submillimeter-luminous ( 𝑆 (cid:39) 𝑧 phot = . ± .
09 (Dudzeviciute et al. 2020).In Figure 9 (left-hand panel), we show the distribution of dust-to-stellar mass ratio versus total infrared luminosity for the AS2UDSSMGs from the analysis of Dudzeviciute et al. (2020) (red circles).SHiZELS-14 lies above the average of the galaxies in TIR luminos-ity, but has a fairly unremarkable dust mass-to-stellar mass ratio. Inthe right-hand panel of the same figure, we show the effective radiusof the dust continuum emission along the semi-major axis versus thetotal infrared luminosity for a subsample of the AS2UDS sourceswith higher spatial resolution ALMA observations from Gullberget al. (2019), with SHiZELS-14 overplotted on the same axes. In thecontext of these bright SMGs, SHiZELS-14’s TIR luminosity is notexceptional. However, it has a dust continuum size that is larger thanany of the comparison sample.Like other SMGs, SHiZELS-14 displays a compact core of dustcontinuum emission. As seen from Figure 1, it also has substantialextended emission, which we are able to resolve due to our deepALMA imaging. Gullberg et al. (2019) discuss the possibility ofan extended component in the AS2UDS SMGs. For sources withSCUBA-2 flux densities brighter than 4mJy beam − (SHiZELS-14is in this class), the median flux recovery from the ALMA pointingsis 97 + − per cent (Stach et al. 2019). For sources with fainter SCUBA-2 flux densities (2 . ≤ 𝑆 ≤ . − ), the median fluxrecovery of those with ALMA detections is 88 ± ∼ . (cid:48)(cid:48) and ∼ . (cid:48)(cid:48) . The extended componentaccounts for only 13 ± − . (cid:48)(cid:48) component with flux density 0 . . + . − . . We apply this cor-rection to the AS2UDS data, and note that SHiZELS-14 remainsan outlier. The ∼ ∼
10 percent of the total flux for the ASUDS SMGs is the dominant compo-nent for SHiZELS-14, which displays an extended disk-like structurethat is well-fitted by a single two-dimensional Sérsic profile with 𝑅 maj 𝑒 = . 𝑛 = 𝑞 = .
47. SHiZELS-14 is genuinely moreextended than the majority of the AS2UDS SMGs, perhaps due tobeing a mid-stage merger.We also compare SHiZELS-14 to a sample of 𝐾 -band-identified,stellar mass-selected ( 𝑀 ∗ > 𝑀 (cid:12) ), 𝑈𝑉 𝐽 -classified star-forming,intermediate redshift ( 𝑧 = . − .
6) galaxies studied by Tadaki et al.(2020). Unlike the AS2UDS sources, these galaxies were not explic-itly selected to be sub-millimeter bright. For the 69 of their sourcesthat lie in UDS, we make use of the same multi-wavelength parentcatalogues used in Dudzeviciute et al. (2020) for the AS2UDS galax-ies, and repeat the MAGPHYS SED fitting procedure. We also adopt
MNRAS in press, 1–17 (2021)
HiZELS-14: Resolving a dusty, star-forming galaxy at 𝑧 = . h m s s s s RA A H h m s s s s RA D e c SFR
TIR /SFR H Figure 8: Left: the ratio of TIR-derived SFR to H 𝛼 -derived SFR, assuming the luminosity-SFR calibrations of Kennicutt & Evans (2012),without any correction for dust attenuation. The TIR-derived SFR is larger than that derived from H 𝛼 across the full extent of the galaxy, butthe estimates are discrepant by a factor of ∼
50 in the dusty central region. Right: the dust attenuation 𝐴 H 𝛼 derived from this ratio. Where theH 𝛼 flux is below the detection limit, neither ratio nor 𝐴 H 𝛼 value is plotted. 𝐴 H 𝛼 varies across the galaxy, within a broad range 𝐴 H 𝛼 ∼ − 𝐴 H 𝛼 =
1, but the dust attenuation of SHiZELS-14 derived here iswell in excess of this value. ALMA contours are overlaid on both panels in red.the effective radii presented by Tadaki et al. (2020), obtained viafitting in the 𝑢𝑣 plane (with fixed 𝑛 = ( 𝐿 TIR / L (cid:12) ) = . . 𝑇 dust = . . 𝑅 𝑒 versus LIR plane to theAS2UDS galaxies. These less TIR-luminous galaxies tend to havelarger sizes (median 𝑅 majeff , ALMA ( 𝑛 = ) = . 𝑅 majeff , ALMA ( 𝑛 = ) > 𝑅 majeff , ALMA ∼ ( 𝐿 TIR / L (cid:12) ) < ( 𝐿 TIR / L (cid:12) ) (cid:39) .
5) are just as compactas the AS2UDS sources.Here, we relate the observed morphology of the dust continuumemission to the physical processes taking place within star-forminggalaxies around the peak of cosmic star formation. As discussed byCheng et al. (2020), the extended dust continuum emission observedin the less-FIR luminous SHiZELS galaxies suggests a dominantcomponent of extended, disk-wide star formation; in contrast, theemission from sub-millimeter selected galaxies appears to be dom- inated by a compact, nuclear starburst. SHiZELS-14 is an outlierin the sense that it has both a sub-millimeter bright compact coreand very extended emission. Tadaki et al. (2020) show that the mostcompact galaxies in their sample tend to have high gas fractions(derived via 𝑆 𝜇 m ), and argue that this reflects efficient radial gasinflows. Numerical simulations have long shown that galaxy mergersare capable of triggering tidally-driven gas inflows (Hernquist 1989;Barnes & Hernquist 1991), which can cause strong nuclear starbursts(e.g. Mihos & Hernquist 1994, 1996; Hopkins et al. 2013; Morenoet al. 2015). However, observations of local galaxies such as the An-tennae system demonstrate that galaxy interactions can also triggerwidespread star formation that is not limited to a compact, nuclearregion (Wang et al. 2004). More recent, high resolution simulationsshow that these observations can be explained via merger-driven in-jections of turbulence into the ISM: extended compression resultsin fragmentation into dense, star-forming gas, and spatially extendedstarburst activity (Renaud et al. 2014, 2015). Renaud et al. (2015)argue that this process is particularly important in the early and midstages of a galaxy merger: during the first two simulated pericenterpassages, star clusters form kiloparsecs from the galactic nucleus,with the central starburst dominating only from the beginning of thefinal coalescence. This progression of star formation from extendedto compact as the merger unfolds is also consistent with observationsof local galaxies (Pan et al. 2019). The extended star formation ob-served in SHiZELS-14 may therefore suggest that we are viewing theshort-lived mid-stages of a merger; this would be consistent with itscomplex, irregular morphology and dispersion-dominated H 𝛼 kine-matics. The similarly TIR-luminous but more compact sources withinthe AS2UDS samples may comprise galaxies experiencing a widerrange of evolutionary stages, including some later-stage mergers. MNRAS in press, 1–17 (2021) R.K. Cochrane et al. L TIR / L fl -5 -4 -3 -2 -1 M du s t / M ∗ L TIR / L fl R m a j e , A L M A / k p c SHiZELS-14Cheng+20AS2UDSTadaki+20
Figure 9: Left: dust mass-to-stellar mass ratio versus total infrared luminosity, for the AS2UDS sub-millimeter bright galaxies (red circles;Dudzeviciute et al. 2020), the subset of stellar mass-selected galaxies from (Tadaki et al. 2020) that are in UDS and have size measurements,and the three TIR-brightest SHiZELS galaxies from Cheng et al. (2020) (black). SHiZELS-14 (black star) is well within the range of bothparameters derived for the sub-millimeter galaxy population. Dust mass and stellar mass are derived using
MAGPHYS for SHiZELS-14 andthe AS2UDS and Tadaki et al. (2020) samples. Right: effective radius (along the semi-major axis) versus total infrared luminosity for thesubsample of these galaxies targeted at higher spatial resolution with ALMA, presented by Gullberg et al. (2019) 𝑎 , with the small statisticalcorrection derived by Smail et al. (2020) applied to the source sizes. The grey line at 𝑅 𝑒, ALMA = 𝑛 =
1) in the image plane. For the remaining SHiZELSgalaxies, radii were derived using a curve-of-growth analysis (Cheng et al. 2020). For the (Tadaki et al. 2020) sample, radii were derived usingGaussian model fits in the 𝑢𝑣 -plane. SHiZELS-14 has a much larger effective radius (as measured in the rest-frame FIR) than the majority ofthe AS2UDS galaxies, though several less FIR-luminous galaxies in the (Tadaki et al. 2020) sample are similarly extended. The extended dustemission suggests that SHiZELS-14 is caught in the mid-stages of a merger. 𝑎 We have corrected a minor error in the table presented by Gullberg et al. (2019). This error is noted in Smail et al. (2020). The stated 𝑅 𝑒 values for the case ofthe fixed 𝑛 = 𝑅 maj 𝑒, ALMA = 𝑅 𝑒, original / axial ratio. In this paper, we have presented a study of SHiZELS-14, a 𝑧 = .
24 galaxy originally identified by HiZELS via its H 𝛼 emission.SHiZELS-14 was one of the galaxies selected for high spatial res-olution follow-up, due to its proximity to a guide star (for adaptiveoptics observations), rather than any special properties. However,this galaxy has some intriguing features when resolved at high spa-tial resolution, particularly at long wavelengths.The global properties of SHiZELS-14 show that it is highly star-forming. SED fits to photometric data indicate a strong burst ofstar formation within ∼
200 Myr of 𝑧 = .
24 and a stellar mass of10 . ± . M (cid:12) . Fitting the dust SED with modified black body mod-els yields a dust mass of M dust = . ± . M (cid:12) and a TIR luminosityof log ( 𝐿 TIR / L (cid:12) ) = . ± .
02. This bright IR emission placesit in the category of a ULIRG, while its strong submillimeter detec-tion shows it is an SMG. SHiZELS-14 lies on the 𝑧 ∼ 𝛼 , FIR and radio continuum emission are all used to inferSFR, individually and in combination. We investigate the agreementof widely-used SFR calibrations, globally and in a spatially-resolvedmanner. Without any dust corrections, the SFRs inferred from FUVand H 𝛼 are 13 ± (cid:12) yr − and 33 ± (cid:12) yr − , respectively. The SFRinferred from the TIR emission is 950 ±
50 M (cid:12) yr − , and the radio-derived SFR is also in the region ∼ (cid:12) yr − . Thus, SFR inferred from short wavelength light is orders of magnitude lower than thatmeasured at longer wavelengths. This suggests that SHiZELS-14 isaffected by a large degree of dust attenuation, in line with its sub-stantial dust mass and FIR flux, and it shares many properties withthe known population of high redshift SMGs.We present kpc-scale imaging in the rest-frame FUV and optical(from HST ), at FIR-wavelengths (from ALMA), of the H 𝛼 emis-sion line (from SINFONI, on the VLT), and of the radio continuum(from the JVLA). The range of wavelengths probed enables us todetect both unattenuated and dust-reprocessed emission. SHiZELS-14 shows striking, extended emission in both H 𝛼 and the FIR, withH 𝛼 -derived effective radius 4 . ± . . ± . 𝑞 = . − 𝛼 . The irregular, extended structures anddisordered H 𝛼 kinematics, together with the intense burst of dustystar formation observed, likely reflects ongoing (at = .
24) mergeractivity.The high spatial resolution of our data enables us to study emissionon kpc scales, and compare SFRs in a spatially-resolved manner. Weshow that the SFR surface density maps derived from UV, H 𝛼 andTIR are discrepant across the the extent of the galaxy. Comparison of MNRAS in press, 1–17 (2021)
HiZELS-14: Resolving a dusty, star-forming galaxy at 𝑧 = . the H 𝛼 and TIR maps enables us to map the dust attenuation, underthe assumption of minimal gradients in dust temperature and opticaldepth. We find high levels of dust attenuation across the galaxy, with 𝐴 H 𝛼 ∼ − 𝐴 H 𝛼 > ACKNOWLEDGEMENTS
We thank the anonymous reviewer for helpful suggestions that im-proved the paper. RKC acknowledges funding from an STFC stu-dentship, the Institute for Astronomy, University of Edinburgh, andthe John Harvard Distinguished Science Fellowship. PNB is grate-ful for support from STFC via grant ST/R000972/1. AMS and IRSacknowledge support from STFC (ST/T000244/1). EI acknowledgespartial support from FONDECYT through grant N ◦ BAGPIPES
SED fittingtool and Richard Bower for helpful discussions during the PhD viva.We thank Jiasheng Huang for providing us with the CLAUDS imageof SHiZELS-14.This research is based on observations made with the NASA/ESA
Hubble Space Telescope obtained from the Space Telescope Sci-ence Institute, which is operated by the Association of Univer-sities for Research in Astronomy, Inc., under NASA contractNAS 5-26555. These observations are associated with program14919. This paper makes use of the following ALMA data:ADS/JAO.ALMA2015.1.00026.S. ALMA is a partnership of ESO(representing its member states), NSF (USA) and NINS (Japan), to-gether with NRC (Canada), MOST and ASIAA (Taiwan), and KASI(Republic of Korea), in cooperation with the Republic of Chile.The Joint ALMA Observatory is operated by ESO, AUI/NRAO andNAOJ. This paper uses data from SINFONI, based on observationscollected at the European Organisation for Astronomical Researchin the Southern Hemisphere under ESO programme 084.B-0300.This work is based on data products from observations made withESO Telescopes at the La Silla Paranal Observatory under ESO pro-gramme 179.A-2005 and on data products produced by TERAPIXand the Cambridge Astronomy Survey Unit on behalf of the UltraV-ISTA consortium.The HSC collaboration includes the astronomical communities ofJapan and Taiwan, and Princeton University. The HSC instrumen-tation and software were developed by the National AstronomicalObservatory of Japan (NAOJ), the Kavli Institute for the Physicsand Mathematics of the Universe (Kavli IPMU), the University ofTokyo, the High Energy Accelerator Research Organization (KEK),ASIAA, and Princeton University. Funding was contributed by theFIRST program from Japanese Cabinet Office, the Ministry of Edu-cation, Culture, Sports, Science and Technology (MEXT), the JapanSociety for the Promotion of Science (JSPS), Japan Science andTechnology Agency (JST), the Toray Science Foundation, NAOJ,Kavli IPMU, KEK, ASIAA, and Princeton University. This work is based on observations obtained with MegaPrime/MegaCam, a joint project of CFHT and CEA/DAPNIA, at theCFHT which is operated by the National Research Council (NRC) ofCanada, the Institut National des Science de l’Univers of the CentreNational de la Recherche Scientifique (CNRS) of France, and theUniversity of Hawaii. This research uses data obtained through theTelescope Access Program (TAP), which has been funded by the Na-tional Astronomical Observatories, Chinese Academy of Sciences,and the Special Fund for Astronomy from the Ministry of Finance.This work uses data products from TERAPIX and the CanadianAstronomy Data Centre. It was carried out using resources fromCompute Canada and Canadian Advanced Network For Astrophys-ical Research (CAN-FAR) infrastructure. These data were obtainedand processed as part of CLAUDS, which is a collaboration betweenastronomers from Canada, France, and China described in Sawickiet al. (2019).This research has made use of the VizieR catalogue access tool,CDS, Strasbourg, France. The original description of the VizieRservice was published in A&AS 143, 23. The
Herschel
Extragalac-tic Legacy Project (HELP) is a European Commission ResearchExecutive Agency funded project under the SP1-Cooperation, Col-laborative project, Small or medium-scale focused research project,FP7-SPACE-2013-1 scheme, Grant Agreement Number 607254.
DATA AVAILABILITY
The data underlying this article will be shared on reasonable requestto the corresponding author.
REFERENCES
Aihara H., Hattori S., Ota N., Zhang Y.-Y., Akamatsu H., Finoguenov A.,2019, PASJ, 71, 114Alexander D. M., et al., 2003, ApJ, 125, 383Algera H. S. B., et al., 2020, arXiv:2009.06647Appleton P. N., et al., 2004, ApJS, 154, 147Asplund M., Grevesse N., Sauval A. J., Scott P., 2009, ARAA, 47, 481Baldwin J. A., Phillips M. M., Terlevich R., 1981, PASP, 93, 5Barger A. J., Cowie L., Sanders D. B., Fulton E., Taniguchi Y., Sato Y.,Kawara K., Okuda H., 1998, Nature, 394, 248Barnes J. E., Hernquist L. E., 1991, ApJ, 370, L65Barro G., et al., 2016, ApJL, 827, L32Battisti A. J., et al., 2019, ApJ, 882, 61Bell E. F., 2003, ApJ, 586, 794Belli S., Newman A. B., Ellis R. S., 2016, ApJ, 834, 1Bertin E., Arnouts S., 1996, A&AS, 117, 393Blain A. W., Smail I., Ivison R. J., Kneib J. P., Frayer D. T., 2002, Phys. Rep.,369, 111Bouwens R. J., et al., 2010, ApJL, 709, 133Bruzual G., Charlot S., 2003, MNRAS, 344, 1000Byler N., Dalcanton J. J., Conroy C., Johnson B. D., 2017, ApJ, 840, 44Calistro Rivera G., et al., 2018, ApJ, 863, 56Calura F., et al., 2017, MNRAS, 465, 54Calzetti D., Kinney A. L., Storchi-Bergmann T., 1994, ApJ, 429, 582Calzetti D., Armus L., Bohlin R., Kinney A. L., Koornneef J., Storchi-Bergmann T., 2000, ApJ, 533, 682Carnall A. C., McLure R. J., Dunlop J. S., Davé R., 2018, MNRAS, 480,4379Casey C. M., Narayanan D., Cooray A., 2014, Phys. Rep., 541, 45Chabrier G., 2003, PASP, 115, 763Chapman S. C., Blain A. W., Smail I., Ivison R. J., 2005, ApJ, 622, 772Charlot S., Fall S. M., 2000, ApJ, 539, 718 MNRAS in press, 1–17 (2021) R.K. Cochrane et al.
Chen C.-C., Smail I., Swinbank A., Simpson J. M., Ma C.-J., AlexanderD. M., Biggs A. D., Brandt W. N., 2015, ApJ, 799, 194Chen C.-C., et al., 2017, ApJ, 846, 108Cheng C., et al., 2020, arxiv:2011.00686Civano F., et al., 2016, ApJ, 819, 62Cochrane R. K., Best P. N., Sobral D., Smail I., Wake D. A., Stott J. P., GeachE., 2017, MNRAS, 469, 2913Cochrane R. K., Best P. N., Sobral D., Smail I., Geach J. E., Stott J. P., WakeD. A., 2018, MNRAS, 475, 3730Cochrane R. K., et al., 2019, MNRAS, 488, 1779Condon J. J., 1992, ARA&A, 30, 575Da Cunha E., Charlot S., Elbaz D., 2008, MNRAS, 388, 1595Da Cunha E., et al., 2015, ApJ, 806, 110Danielson A., et al., 2017, ApJ, 840, 78Draine B. T., Li A., 2007, ApJ, 657, 810Dudzeviciute U., et al., 2020, MNRAS, 494, 3828Dunlop J. S., et al., 2013, MNRAS, 432, 3520Dunlop J. S., et al., 2017, MNRAS, 466, 861Elmegreen B. G., Elmegreen D. M., Sánchez Almeida J., Muñoz-Tuñón C.,Dewberry J., Putko J., Teich Y., Popinchalk M., 2013, ApJ, 774, 86Faisst A. L., Fudamoto Y., Oesch P. A., Scoville N., Riechers D. A., PavesiR., Capak P., 2020, MNRAS, 498, 4192Falgarone E., et al., 2017, Nature, 548, 430Ferland G. J., et al., 2017, Revista Mexicana de Astronomia y Astrofisica, 53,385Fisher D. B., et al., 2017, MNRAS, 464, 491Foreman-Mackey D., 2016, JOSS, 24Franco M., et al., 2018, A&A, 620, A152Galametz M., et al., 2012, MNRAS, 425, 763Garn T., et al., 2010, MNRAS, 402, 2017Geach J. E., Smail I., Best P. N., Kurk J., Casali M., Ivison R. J., Coppin K.,2008, MNRAS, 388, 1473Genzel R., et al., 2008, ApJ, 687, 59Genzel R., et al., 2013, ApJ, 773, 68Gillman S., et al., 2019, MNRAS, 486, 175Goldader J. D., Meurer G., Heckman T. M., Seibert M., Sanders D. B., CalzettiD., Steidel C. C., 2002, ApJ, 568, 651Gómez-Guijarro C., et al., 2018, ApJ, 856, 121Gullberg B., et al., 2019, MNRAS, 4974, 4956Guo Y., et al., 2015, ApJ, 800, 39Guo Y., et al., 2017, ApJ, 853, 108Heckman T., Best P., 2014, ARAA, 52, 589Hernquist L., 1989, Nature, 340, 687Hodge J. A., da Cunha E., 2020Hodge J. A., Riechers D., Decarli R., Walter F., Carilli C. L., Daddi E.,Dannerbauer H., 2015, ApJL, 798, 1Hodge J. A., et al., 2016, ApJ, 833, 1Hodge J. A., et al., 2019, ApJ, 876, 130Hopkins P. F., Cox T. J., Hernquist L., Narayanan D., Hayward C. C., MurrayN., 2013, MNRAS, 430, 1901Hurley P. D., et al., 2017, MNRAS, 464, 885Illingworth G. D., et al., 2013, ApJS, 209, 6Ivison R. J., et al., 2010, A&A, 518, 31Ivison R. J., Richard J., Biggs A. D., Zwaan M. A., Falgarone E., ArumugamV., Van Der Werf P. P., Rujopakarn W., 2020, MNRAS, 495, L1James A., Dunne L., Eales S., Edmunds M. G., 2002, MNRAS, 335, 753Jiménez-Andrade E. F., et al., 2018, A&A, 615, A25Kennicutt R. C., Evans N. J., 2012, ARA&A, 50, 531Kewley L. J., Groves B., Kauffmann G., Heckman T., 2006, MNRAS, 372,961Koprowski M. P., Dunlop J. S., Michałowski M. J., Cirasuolo M., BowlerR. A., 2014, MNRAS, 444, 117Krogager J. K., Zirm A. W., Toft S., Man A., Brammer G., 2014, ApJ, 797Kroupa P., 2002, Science, 295, 82Laigle C., et al., 2016, ApJS, 224, 1Lang P., et al., 2019, ApJ, 879, 54Lin L., et al., 2017, ApJ, 851, 18 Ma X., Hopkins P. F., Feldmann R., Torrey P., Faucher-Giguère C. A., KerešD., 2017, MNRAS, 466, 4780Madau P., Dickinson M., 2014, ARAA, p. 76Magdis G. E., et al., 2016, MNRAS, 456, 4533Magnelli B., et al., 2014, A&A, 561, A86Magnelli B., et al., 2020, ApJ, 892, 66Martí-Vidal I., Vlemmings W. H., Muller S., Casey S., 2014, A&A, 563McCracken H. J., et al., 2015, MNRAS, 449, 901Meurer G., Heckman T. M., Calzetti D., 1999, ApJ, 521, 64Mihos J. C., Hernquist L., 1994, ApJ, 431, L9Mihos J. C., Hernquist L., 1996, ApJ, 464, 641Molina J., Ibar E., Swinbank A. M., Sobral D., Best P. N., Smail I., EscalaA., Cirasuolo M., 2017, MNRAS, 466, 892Molina J., et al., 2019, MNRAS, 482, 1499Moreno J., Torrey P., Ellison S. L., Patton D. R., Bluck A. F., Bansal G.,Hernquist L., 2015, MNRAS, 448, 1107Nandra K., Pounds K. A., 1994, MNRAS, 268, 405Narayanan D., Davé R., Conroy C., Thompson R., Geach J., Johnson B. D.,2018, MNRAS, 474, 1718Nelson E. J., et al., 2019, ApJ, 870, 130Oesch P. A., et al., 2010, ApJL, 709, L21Oliver S. J., et al., 2012, MNRAS, 424, 1614Onodera M., et al., 2015, ApJ, 808, 161Oteo I., Sobral D., Ivison R. J., Chandra X.-r., 2015, MNRAS, 452, 2018Oteo I., et al., 2017, ApJ, 850, 170Pan H.-A., et al., 2019, ApJ, 881, 119Papovich C., et al., 2016, Nature Astronomy, 1, 3Pearson W. J., Wang L., van der Tak F. F. S., Hurley P. D., Burgarella D.,Oliver S. J., 2017, A&A, 603, A102Reddy N. A., et al., 2015, ApJ, 806, 259Renaud F., Bournaud F., Kraljic K., Duc P. A., 2014, MNRAS, 442, L33Renaud F., Bournaud F., Duc P. A., 2015, MNRAS, 446, 2038Rieke G. H., et al., 2004, ApJS, 154, 25Rujopakarn W., et al., 2019, ApJ, 882, 107Sawicki M., et al., 2019, MNRAS, 489, 5202Schinnerer E., et al., 2007, ApJS, 172, 46Schinnerer E., et al., 2010, ApJS, 188, 384Scoville N., et al., 2007, ApJS, 172, 1Scoville N., et al., 2014, ApJ, 783, 84Simons R. C., et al., 2017, ApJ, 843, 46Simpson J. M., et al., 2014, ApJ, 788Simpson J. M., et al., 2015, ApJ, 799, 81Simpson J. M., et al., 2017, ApJ, 839, 58Simpson J. M., et al., 2019, ApJ, 880, 43Smail I., Ivison R. J., Blain A. W., 1997, ApJ, 490, L5Smail I., et al., 2020, 9, 1Smith D. J., Hayward C. C., Jarvis M. J., Simpson C., 2017, MNRAS, 471,2453Smolčić V., et al., 2017, A&A, 602, A2Sobral D., et al., 2009, MNRAS, 398, 75Sobral D., Best P. N., Geach J. E., Smail I., Cirasuolo M., Garn T., DaltonG. B., Kurk J., 2010, MNRAS, 1563, 1551Sobral D., Smail I., Best P. N., Geach J. E., Matsuda Y., Stott J. P., CirasuoloM., Kurk J., 2013a, MNRAS, 428, 1128Sobral D., et al., 2013b, ApJ, 779, 139Sobral D., Best P. N., Smail I., Mobasher B., Stott J., Nisbet D., 2014,MNRAS, 427, 3516Sobral D., et al., 2015, MNRAS, 451, 2303Soto E., et al., 2017, ApJ, 837, 6Speagle J. S., Steinhardt C. L., Capak P. L., Silverman J. D., 2014, ApJSS,214, 15Stach S. M., et al., 2018, ApJ, 860, 161Stach S. M., et al., 2019, MNRAS, 487, 4648Stott J. P., et al., 2016, MNRAS, 457, 1888Strandet M., et al., 2017, ApJL, 842, L15Swinbank A. M., Sobral D., Smail I., Geach J. E., Best P. N., Mccarthy I. G.,Crain R. A., Theuns T., 2012a, MNRAS, 426, 935MNRAS in press, 1–17 (2021)
HiZELS-14: Resolving a dusty, star-forming galaxy at 𝑧 = . Figure A1: 𝑢𝑣 -amplitude plot for the ALMA 260 GHz data. We fit aGaussian model with varying axis ratio in the 𝑢𝑣 -plane, using CASA ’s uvmodelfit task. The effective radius along the semi-major axis, 𝑅 maj 𝑒 ,4 . ± . uvmultfit tool(Martí-Vidal et al. 2014). We also fit an exponential profile and obtaina best-fitting scale length 𝑟 = . ± . Swinbank A. M., Smail I., Sobral D., Theuns T., Best P. N., Geach J. E.,2012b, ApJ, 760, 130Swinbank A. M., et al., 2014, MNRAS, 438, 1267Tacconi L. J., Genzel R., Neri R., Cox P., Cooper M. C., Shapiro K., BolattoA., 2010, Nature, 463, 781Tacconi L. J., et al., 2013, ApJ, 768, 74Tacconi L. J., et al., 2017, ApJ, 853, 179Tadaki K.-i., et al., 2016, ApJ, 834, 1Tadaki K.-i., et al., 2017, ApJL, 841, L25Tadaki K., et al., 2018, Nature, 560, 613Tadaki K.-i., et al., 2020Taniguchi Y., et al., 2007, ApJS, 172, 9Taniguchi Y., et al., 2015, PASJ, 67, 1Thomson A. P., et al., 2019, ApJ, 883, 204Toft S., et al., 2014, ApJ, 782Wang Z., et al., 2004, ApJS, 154, 193Wang T., et al., 2019, Nature, 572, 211Wuyts S., et al., 2012, ApJ, 753
APPENDIX A: DERIVING A REST-FRAME FIR SIZEFROM THE ALMA DATA
In Figure A1 we show a Gaussian model fit to the 260 GHz ALMAvisibility data. The derived effective radius along the semi-majoraxis, 𝑅 maj 𝑒 = . ± . 𝛼 image (4 . ± . 𝑓 ( 𝑟 ) ∝ 𝑒 − 𝑟 / 𝑟 ) with the uvmultfit tool (Martí-Vidal et al. 2014) and obtain a best-fitting scale length 𝑟 = . ± . APPENDIX B: DETAILED DESCRIPTION OF SEDFITTING WITH BAGPIPES
To assess the sensitivity of our derived physical parameters to ourchoice of SED fitting code, we refit the photometry with anothercode. Our fitting makes use of the 2016 version of the BC03 SSP Figure B1: Data presented in Table B1, fitted with the
BAGPIPES code (Carnall et al. 2018), using a double power law star formationhistory. Error bars are plotted on the data points, but are small. Thefitting yields SFR = ±
60 M (cid:12) / yr and log 𝑀 ∗ / 𝑀 (cid:12) = . ± . MAGPHYS .templates, with a Kroupa (2002) IMF (note that the difference be-tween a Kroupa and Chabrier IMF is negligible). Nebular emissionis computed using the
CLOUDY photoionization code (Ferland et al.2017), following Byler et al. (2017).
CLOUDY is run using eachSSP template as the input spectrum. Dust grains are included using
CLOUDY ’s ‘ISM’ prescription, which implements a grain-size distri-bution and abundance pattern that reproduces the observed extinctionproperties for the ISM of the Milky Way. We select a Calzetti et al.(2000) dust attenuation curve. Dust emission includes both a hot dustcomponent from HII regions and a grey body component from thecold, diffuse dust.We impose a wide dust attenuation prior, 𝐴 𝑣 = [ , ] , which givesthe code the option to fit a high degree of attenuation. FollowingDraine & Li (2007), we fit three parameters that affect the shape ofthe dust SED: 𝑈 min , the lower limit of the starlight intensity; 𝛾 , thefraction of stars at 𝑈 min ; and 𝑞 PAH , the mass fraction of polycyclicaromatic hydrocarbons. Our priors on these parameters are broad, toallow the model the option to fit a hot dusty galaxy: 𝑈 min = [ , ] , 𝛾 = [ , ] , and 𝑞 PAH = [ , ] . We also fit 𝜂 , the multiplicativefactor on 𝐴 𝑉 for stars in birth clouds, using the range 𝜂 = [ , ] .We allow metallicity to vary in the range 𝑍 = [ , . ] 𝑍 (cid:12) , old , where 𝑍 (cid:12) , old denotes solar models prior to Asplund et al. (2009). We fixthe redshift at 𝑧 = . 𝑀 ∗ / M (cid:12) = . ± .
1. Allparametrisations, even those allowing multiple bursts, favour a recent(at 𝑧 = . = ±
60 M (cid:12) yr − , and the estimated specific star-formation rate (sSFR) is log ( sSFR / yr − ) = − . ± .
11. Notethat the SFR is more sensitive than the stellar mass to the parametri-sation of the SFH and the data included in the fit, and averagingover multiple SFH models increases the uncertainty on the SFR to ∼
100 M (cid:12) yr − . The posterior estimate for the dust attenuation in the 𝑉 -band is 𝐴 𝑣 = . ± .
1. All of these derived physical parametersare consistent with the estimates from
MAGPHYS , which indicatesthat our fitting is robust to choice of SED fitting code.
This paper has been typeset from a TEX/L A TEX file prepared by the author.MNRAS in press, 1–17 (2021) R.K. Cochrane et al.
Instrument/Telescope Filter Measurement(Survey) ( 𝜇 Jy)MegaCam/CFHT 𝑢 ∗ . ± . 𝐵 . ± . 𝑉 . ± . 𝑟 . ± . 𝑖 + . ± . 𝑧 + . ± . 𝑧 ++ . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . 𝑌 HSC . ± . 𝑌 . ± . 𝐽 . ± . 𝐻 . ± . 𝐾 𝑠 . ± . 𝐾 sw . ± . 𝐻 w . ± . Spitzer /IRAC 3 . 𝜇 m 35 . ± . . 𝜇 m 44 . ± . . 𝜇 m 48 . ± . 𝜇 m 33 . ± . Spitzer /MIPS 24 𝜇 m 403 ± 𝜇 m 8 . ± . ( mJy ) HELP catalogue values 160 𝜇 m 20 . ± . ( mJy ) 𝜇 m 31 . ± . ( mJy ) 𝜇 m 36 . ± . ( mJy ) 𝜇 m 27 . ± . ( mJy ) ALMA Band 6, this paper 260 GHz 2 . ± . ( mJy ) ALMA Band 7, Scoville+14 350 GHz 4 . ± . ( mJy ) SCUBA-2, Simpson+19 350 GHz 5 . ± . ( mJy ) JVLA, This paper 6 GHz 20 ± ± . ± (cid:48)(cid:48) diameter aperture.diameter aperture.