Multiwavelength dissection of a massive heavily dust-obscured galaxy and its blue companion at z \sim 2
M. Hamed, L. Ciesla, M. Béthermin, K. Ma?ek, E. Daddi, M. T. Sargent, R. Gobat
AAstronomy & Astrophysics manuscript no. ulirgz2 © ESO 2021January 20, 2021
Multiwavelength dissection of a massive heavily dust-obscuredgalaxy and its blue companion at z ∼ M. Hamed , L. Ciesla , M. Béthermin , K. Małek , , E. Daddi , M. T. Sargent , and R. Gobat National Centre for Nuclear Research, ul. Pasteura 7, 02-093 Warszawa, Poland Aix Marseille Univ. CNRS, CNES, LAM, Marseille, France CEA, Irfu, DAp, AIM, Universitè Paris-Saclay, Universitè de Paris, CNRS, F-91191 Gif-sur-Yvette, France Astronomy Centre, Department of Physics and Astronomy, University of Sussex, Brighton, BN1 9QH, UK Instituto de Fisica, Pontificia Universidad Católica de Valparaíso, Casilla 4059, Valparaiso, ChileReceived 02 October 2020 / Accepted 15 January 2021
ABSTRACT
Aims.
In this work we study a system of two galaxies, namely Astarte and Adonis, at z ∼ Methods.
We use CIGALE - a spectral energy distribution modeling code - that relies on the energetic balance between the UVand the IR, to derive some of the key physical properties of Astarte and Adonis, mainly their star formation rates (SFRs), stellarmasses, and dust luminosities. We inspect the variation of the physical parameters depending on the assumed dust attenuation law.We also estimate the molecular gas mass of Astarte from its CO emission, using di ff erent α CO and transition ratios ( r ) and discussthe implication of the various assumptions on the gas mass derivation. Results.
We find that Astarte exhibits a MS-like star formation activity, while Adonis is undergoing a strong starburst (SB) phase.The molecular gas mass of Astarte is far below the gas fraction of typical star-forming galaxies at z =
2. This low gas content and highSFR, result in a depletion time of 0 . ± .
07 Gyrs, which is slightly shorter than what is expected for a MS galaxy at this redshift.The CO luminosity relative to the total IR luminosity suggests a MS-like activity if we assume a galactic conversion factor and a lowtransition ratio. The SFR of Astarte is of the same order using di ff erent attenuation laws, unlike its stellar mass that increases usingshallow attenuation laws ( ∼ × M (cid:12) assuming a Calzetti relation versus ∼ × M (cid:12) assuming a shallow attenuation law). Wediscuss these properties and suggest that Astarte might be experiencing a recent decrease of star formation activity and is quenchingthrough the MS following a SB epoch. Key words. galaxies: evolution - galaxies: high-redshift - galaxies: star formation - galaxies: starburst - infrared: galaxies - ISM: dust,extinction.
1. Introduction
Studying galaxy evolution throughout the cosmic time is a keyelement of modern astrophysics, and is crucial for our under-standing of the life cycle of the progenitors of passive ellipticalgalaxies that we observe in the local Universe. Evidences sug-gest that the star formation rate (SFR) density has peaked arounda redshift of z ≈ ffl uence of multiwavelength data, especially the far in-frared (FIR) detections from Herschel , played a central rolein understanding how DSFGs evolve as a function of redshift. However, there are still controversies regarding how these galax-ies build up their stellar masses. These controversies arise fromthe systematic uncertainties caused by the heavy dust attenuationin this type of objects (e.g., Hainline et al. 2011; Michałowskiet al. 2012). This is caused by the sensitivity of the stellar massestimate to the type of star formation history (SFH), the choiceof the synthetic stellar population (SSP), and the assumed initialmass function (IMF). The debate about the accuracy of derivedstellar masses of DSFGs was also discussed in details in Caseyet al. (2014).On the other hand, the growing number of ALMA observa-tions in the recent years is providing an unparalleled help in con-straining the evolution of DSFGs. These data are allowing us tobuild a comprehensive view of the role of these giant IR-brightsources by tracing their molecular gas and dust content (e.g.,Donevski et al. 2020). The wealth of multiwavelength data alsocontributed to improve significantly estimating physical proper-ties that govern such galaxies, by modeling their spectral energydistribution (SED, e.g., Burgarella et al. 2005; da Cunha et al.2008; Noll et al. 2009; Conroy 2013; Ciesla et al. 2014).
Article number, page 1 of 10 a r X i v : . [ a s t r o - ph . GA ] J a n & A proofs: manuscript no. ulirgz2
To build an SED, di ff erent aspects of a galaxy must be con-sidered, most importantly the star formation history (SFH), thechange of which has a strong impact on the derived SFR (e.g.Buat et al. 2014; Ciesla et al. 2017), stellar populations of var-ied ages and metallicities, dust emission with di ff erent dust grainsizes and temperatures, nebular and synchrotron emissions, etc.Extinction caused by dust is critically important in any spectrumfitting of a galaxy, since it mutates the shape of the SED the mostby absorbing a significant amount of the UV photons and ther-mally re-emitting them in the IR. This behavior can be modeledby assuming that dust absorbs the shorter wavelength spectrumof galaxies following attenuation laws, which are typically de-scribed by simple power laws with varying complexities, and isable to reproduce the observed extinction in galaxies of di ff er-ent redshifts and types. However, dust attenuation laws are notuniversal (e.g., Wild et al. 2011; Buat et al. 2018; Małek et al.2018; Salim & Narayanan 2020). The need of di ff erent attenu-ation recipes is inevitable in order to reproduce the spectra ofgalaxies of di ff erent masses, IR luminosities and naturally theredshift. This makes it challenging to interpret some of the phys-ical features especially when di ff erent attenuation laws can re-produce a good SED of a galaxy Buat et al. (2019).A non-negligent fraction of galaxies exhibit a non-alignmentand sometimes a total disconnection between the dust contin-uum and the stellar population (Dunlop et al. 2017; Elbaz et al.2018). This directly challenges SED fitting techniques that relyon the energetic balance between the UV and the IR, since thekey assumption for such techniques is that any physical propertyderived from one part of the spectrum should be valid for theentire galaxy. Several approaches were investigated to test thevalidity of this strategy, Buat et al. (2019) suggested the decou-pling of the stellar continuum from the IR emission by modelingtheir fluxes apart to compare the derived parameters such as theSFRs, dust luminosities and stellar masses with the ones derivedusing full SEDs. Statistical samples of such massive and dustygalaxies (e.g., Dunlop et al. 2017; Elbaz et al. 2018; Buat et al.2019; Donevski et al. 2020) o ff er an important insight into theevolution of dust and gas mass through the cosmic time. How-ever, the nature of these giants is not fully understood.The interstellar medium (ISM) is the most important elementin understanding the physical processes of star formation itself,since it contains the building materials for future stars, mostimportantly the hydrogen. Hydrogen’s density was found to betightly correlated with the SFR, as suggested by Schmidt (1959)and investigated by Kennicutt (1998). This correlation is knownas the Schmidt-Kennicutt law, and it takes into account the gas inits molecular and atomic forms, albeit the molecular gas havingthe biggest impact. The mass of this gas can be estimated withthe emission of the easily excited CO molecules (e.g., Carilli& Walter 2013; Weiß et al. 2013b; Decarli et al. 2019; Riech-ers et al. 2020). Tracing the molecular gas with CO emissionrelies entirely on already established abundances in galaxies ofthe local Universe. Large interferometers such as ALMA o ff erunique opportunities to detect these emission lines with an un-precedented accuracy, the luminosity of which can give an esti-mate of the molecular hydrogen mass of a galaxy, typically us-ing a conversion factor. On the other hand, conversion factors inhigh-redshift galaxies are highly debated (see Bolatto et al. 2013for a comprehensive review).Nonetheless, an estimate of the molecular gas reservoir ofgalaxies at the high-mass end of the main sequence is crucialto characterise their star formation activity. For instance, Elbaz et al. (2018) showed that some galaxies exhibit a SB-like gasdepletion time scale despite residing on the MS.Despite the growing number of detection of such heavilydust-obscured ultra-massive objects at high redshift, the progressof SED modeling, and the better comprehension of the high-redshift ISM, we still lack a full picture of how these galaxiesform and quench. Were they always steadily forming stars alongthe MS? Or are they former SBs transiting to the red sequencethrough the main sequence?To answer these questions, it is essential to understand howis the star formation fueled by the gas in massive objects, andwhy does this activity cease. Quenching mechanisms are still notfully understood and they might be caused by AGN feedback oroutflows (e.g. Cattaneo et al. 2009; Dubois et al. 2013; Combes2017) to environmental e ff ects that can lead to gas stripping (e.g.Coil et al. 2008; Mendez et al. 2011).In this paper, and motivated by the aforementioned ques-tions, we analyze and interpret the multi-wavelengths observa-tions of a pair of galaxies at z ∼
2, with original COSMOS2015catalog Laigle et al. (2016) IDs: 647980 and 648299, hereafterAstarte and Adonis. Astarte is an ultra-massive (M (cid:63) > M (cid:12) ),IR-bright galaxy, for which the CO emission is serendipitouslydetected with ALMA. Adonis is a low-mass galaxy bright innear-UV (NUV) and optical bands.The structure of this paper is as follows: in Section 2 we de-scribe the data of the two galaxies analysed in this work. In Sec-tion 3.1 we probe the molecular gas of Astarte using its ALMA-detected CO emission line, and in Section 3.1.5 we investigatethe morphology of this line compared to multiwavelength detec-tion. In Section 3.2 we derive the physical properties of the twogalaxies using SED fitting. the discussion and the conclusion arein Sections 4 and 5 respectively.Throughout this paper, we adopt the stellar initial mass function(IMF) of Chabrier (2003) and a Λ CDM cosmology parameters(WMAP7, Komatsu et al. 2011): H = − Mpc − , Ω M = Ω Λ =
2. Observations
The system of Astarte and Adonis studied in this paper was ini-tially part of a selection of z ∼ (cid:63) > M (cid:12) ) of the main-sequence (MS) of star-forminggalaxies (e.g., Noeske et al. 2007; Daddi et al. 2010; Rodighieroet al. 2011; Schreiber et al. 2015) detected by Herschel / PACSobservations of the COSMOS field (PEP survey, Lutz et al.2011). In the COSMOS2011 catalog where the system was se-lected, the system is not deblended even in the optical and near-IR and appears as a single source. It is probably caused by thefact that this early catalog is mainly built using the i band, whereAstarte is particularly faint. The zCOSMOS survey Lilly et al.(2009) measured the spectroscopic redshift at the position of theHST / ACS source from zCOSMOS and found z spec = . Spitzer / IRAC. Astarte is detectedat 250 and 350 µ m with Herschel / SPIRE using a 24 µ m priorOliver et al. (2012). The aforementioned deblending, coupledwith the FIR detection of Astarte results in a low-mass low-SFRobject (SFR = M (cid:12) yr − with a stellar mass of 9.46 × M (cid:12) ),and a dust-obscured ultra-massive object (SFR = M (cid:12) yr − Article number, page 2 of 10. Hamed et al.: Multiwavelength dissection of a massive heavily dust-obscured galaxy and its blue companion at z ∼ with a stellar mass of 1.41 × M (cid:12) ), as estimated initially us-ing LePhare Arnouts & Ilbert (2011).Astarte and Adonis were observed by ALMA as part of aprogram (2013.1.00914.S, PI: Bethermin) targeting a pilot sam-ple of four massive z ∼ The NUV (rest-frame FUV) detections of our two galaxies areprovided by the Canada France Hawaii Telescope (CFHT) in the u band. Visible and NIR detections (rest-frame mid-UV to NUV)are obtained via the broad band Suprime-Cam of Subaru in the B,V, r, i + bands and the mid-IR data (rest-frame NIR) are from theIRAC camera of Spitzer . The IR-bright Astarte has a MIPS de-tection at 24 µ m with a signal-to-noise ratio (hereafter S / N) > / N ∼
20) in IR detections of
Herschel wherethe beam size is large. The 100 µ m observation from PACS doesnot detect any of the two galaxies, but provides an upper limit,which is taken into account by the SED fitting of Astarte since itconstrains the far-IR part of the spectrum.The radio continuum of Astarte is tentatively detected withthe Karl G. Jansky Very Large Array (VLA) in the S band at ν = σ (S / N = σ upper limit from the standard deviation in thecutout image around our two sources. The beam width of theVLA detection is 0 . (cid:48)(cid:48) , and the continuum is shown in Figure 1.The Jin et al. (2018) catalog provides JCMT’s fluxes at850 µ m for both of our galaxies (2440 ± µ Jy for Astarteand 3910 ± µ Jy for Adonis). We refrain from using thesesuper-deblended fluxes due to the high uncertainties, probablycaused by the degeneracies in the deblending of such a closepair, and since the majority of flux is unexpectedly attributed tothe smaller less IR-bright Adonis. Table 1 presents a summaryof the available photometric data from di ff erent instruments ofthe two galaxies. Astarte was observed at 2.7 mm with ALMA (band 3) with atime-on-source of 45 minutes using 32 antennas on Septemberthe 5 th of 2015, cycle-2 (P.I. M.Béthermin). We use the Com-mon Astronomy Software Applications package and pipeline(CASA) v5.4 (McMullin et al. 2007) for flagging and to re-duce the visibility data. The deconvolution was performed withthe CLEAN algorithm using natural weighting for an optimalS / N. Multi-frequency synthesis mode of the line-free channelsshowed a non-significant continuum emission of the spectrumtherefore its subtraction was not needed. In the deconvolutionprocess, the cell size was set to 0 . (cid:48)(cid:48) . The achieved synthesizedbeam size is 0 . (cid:48)(cid:48) × . (cid:48)(cid:48) , the velocity resolution of the cube is21.36 km s − and the rms is 0.47 mJy beam − km s − per channel. https://casa.nrao.edu/ + + + Adonis + Astarte Fig. 1.
Integrated flux of ALMA-detected CO emission (green con-tours), along with the VLA-detected radio continuum at 3 GHz (ma-genta contours), on the RGB image (VISTA’s Ks, H and J) of Astarteand Adonis. The beam size of ALMA is 0 . × .
50” (lower left beam).The beam FWHM of VLA is 0 . σ significance. The subsequent contoursare in steps of 1 σ with the innermost contour showing 5 σ . The ma-genta contours show 2 and 3 σ significance. The blue cross is centeredon Adonis, the red cross is centered on Astarte. Astarte AdonisCOSMOS15 ID 647980 648299redshift z phot = .
153 z spec = . / Filter λ S ν S ν Instrument ( µ m) ( µ Jy) ( µ Jy)CFHT / u 0.383 0 . ± .
032 0 . ± . / B 0.446 0 . ± .
018 0 . ± . . ± .
033 0 . ± . . ± .
029 0 . ± . + . ± .
035 0 . ± . ++ . ± .
062 1 . ± . / Y 1.02 0 . ± .
155 1 . ± . . ± .
175 2 . ± . . ± .
241 3 . ± . . ± .
351 4 . ± . Spitzer
IRAC1 3.6 18 . ± .
07 3 . ± . . ± .
10 2 . ± . . ± .
00 3 . ± . . ± .
30 -
Spitzer
MIPS1 24 351 ± Herschel
PACS 100 < Herschel
SPIRE 250 17792 ±
744 -SPIRE 350 16058 ± < × . ± . < . Table 1.
Summary of the data of the two sources observed through thedi ff erent instruments. S ν is the flux in ( µ Jy). λ is the center of the spe-cific filter band. Article number, page 3 of 10 & A proofs: manuscript no. ulirgz2
3. Results
In the data cube we find only one significant line and no signif-icant continuum source in the field of view. The line extractionprocedure along with the derivation of the luminosity and the gasmass are described in the following subsections.
The ALMA-detected emission line of Astarte corresponds to theCO(3-2), with a peak at an observed frequency of ν obs = . z CO (3 − = . z phot , Astarte = . + . − . Laigle et al. (2016). Thisvalidates Astarte as the origin of the detected-CO emission. Wedo not detect Astarte in the continuum and measured a 3- σ up-per limit from the map of 0.117 mJy. The expected flux densitiesfrom the SED modeling discussed in Sect.3.2.6 are 0.007 mJyand 0.049 mJy for Adonis and Astarte, respectively. It is thusnot surprising that none of our two sources are detected. Theflux uncertainty was determined by deriving the standard devia-tion in the source-free pixels in the non primary-beam correctedmap, since it has similar noise levels across the emission-freepixels (the noise at the central region is ∼
2% higher than theoutermost region of the map). The achieved S / N is 5.2 for thebrightest channel of the CO(3-2) of Astarte. The emission linewas extracted by fitting a Gaussian over the profile. The good-ness of the Gaussian fit was verified with a χ test, its propertiesare summarized in Figure 2 along with the redshifted CO(3-2)line. The full width at half maximum (FWHM) of the Gaussianis found to be 152.74 ± − . The spectroscopic redshift Fig. 2.
Spectral profile of Astarte with the redshifted CO(3 →
2) line(dashed vertical grey line) and the Gaussian fit (red) with its properties. of the system at the position of HST detection found by by thezCOSMOS survey Lilly et al. (2009) is z spec = . z spec corresponds to that of Adonis since only this UV-bright galaxyis detected with HST / ACS. Taking into account the redshift dif-ference of Astarte and Adonis, the corresponding radial velocitydi ff erence ∆ V is 1335 km s − . This velocity di ff erence is greaterthan what is found in interacting pairs of galaxies, which is typi-cally ∆ V < km s − (Lambas et al. 2003; Alonso et al. 2004).Outflows and absorption in the UV lines could account for fewhundreds of km s − Cassata et al. (2020), or a division of the Hubble flow and peculiar motions, which could account for asignificant velocity contribution if it is along the line of sight.Therefore, this does not eliminate a possible interaction betweenAstarte and Adonis.
The intensity is calculated by integrating over the Gaussian fit ofthe line, which is then converted to the apparent line luminosity(L (cid:48) ), by using the expression from Solomon et al. (1997) thatexpresses L (cid:48) with the integrated source brightness temperaturein units of K km s − pc : L (cid:48) line = . × × I × D L (1 + z ) ν obs , where D L is the luminosity distance in (Mpc), I is the intensityin (Jy km / s), and ν obs is the observed frequency in (GHz). As aconsistency test, we also estimated the integrated flux of the lineusing the moment-zero map, which was obtained by summingthe channels where the emission line is detected. The line fluxis measured in the moment-0 map using a 2-dimension Gaus-sian fit of the source. As shown in Béthermin et al. (2020), thereis no significant di ff erence between this method and a fit in theuv plane for faint compact sources observed by ALMA. The re-sulting flux densities of the two methods are presented in Ta-ble 2. There is 1.2 σ significant di ff erence between the intensitiesderived by each methods. The spectrum is extracted at a singlepoint assuming a point source, while the 2-dimension fit can re-cover the flux from an extended source. This small di ff erence offlux suggests that our source could be marginally resolved. Here-after, we use the flux from the moment-0 map, which takes thisaccount. However, we cannot formally exclude another faint anddi ff use component at larger scale considering the depth of ourdata. Figure 1 shows the flux-integrated moment-0 map of As-tarte represented by confidence levels contours. The size of theCO disk is ∼ kpc . Peak flux I specCO (3 − I momCO (3 − L (cid:48) CO (3 − density (mJy) ( Jy km s − ) (10 K km s − pc )1 . ± .
277 0 . ± .
062 0 . ± .
047 8 . ± . Table 2.
Summary of the CO(3-2) emission line properties of Astarte. I specCO (3 − is achieved by integrating over the Gaussian of the emissionline. I momCO (3 − is the intensity derived from the moment-0 map. To derive the total mass of the molecular gas in a galaxy weassume that the H mass is proportional to the CO(1-0) line lu-minosity which is the commonly used tracer of the cold star-forming molecular clouds, thanks to its small excitation potentialrequirement. The H mass can be derived using a conversion fac-tor α CO (e.g. Downes & Solomon 2003; Greve et al. 2005; Tac-coni et al. 2006; Carilli & Walter 2013; Bothwell et al. 2013b): M H = α CO L (cid:48) CO (1 − where M H is the mass of the molecular hydrogen in M (cid:12) , α CO is the conversion factor and L (cid:48) CO (1 − is the line luminosity in K km s − pc . The practice of H mass derivation with thismethod is very common especially for galaxies at high redshifts Article number, page 4 of 10. Hamed et al.: Multiwavelength dissection of a massive heavily dust-obscured galaxy and its blue companion at z ∼ where information and spatial resolution is often limited. OurCO(3-2) line luminosity has to be converted to CO(1-0) lumi-nosity using a luminosity line ratio r = L (cid:48) CO (3 − / L (cid:48) CO (1 − . Weuse r = . ± .
07, which is the average ratio found for z = . L (cid:48) CO (1 − = (2 . ± . × K km s − pc . To convert this luminosity into hydrogen mass, we use two con-version factors: α CO = . α CO = .
36. The first one manages to recover the moleculargas mass in starbursts (SBs) and submillimeter galaxies (SMGs),where the gas is e ffi ciently heated by dust. The galactic con-version factor is suitable for normal main sequence galaxies(Downes & Solomon 1998; Bolatto et al. 2013; Carilli & Walter2013). For α CO = .
8, the mass of the molecular hydrogen is: M H ( α = . = (1 . ± . × M (cid:12) . Whereas α CO = .
36 results in M H = (8 . ± . × M (cid:12) ,five times larger gas reservoir than the one derived with α CO = . With the velocity FWHM of the CO(3-2) line, we use the methoddescribed in Bothwell et al. (2013a) to estimate the dynamicalmass of Astarte. Assuming that a rotating disk is the origin ofthe detected line, the dynamical mass can be written as in Neriet al. (2003): M dyn ( M (cid:12) ) = × ∆ V R / sin ( i ) , where ∆ V is the FWHM of the line velocity, i is the inclinationangle of the disk and R is the radius of the disk in kpc. For a ran-dom inclination of (cid:104) i (cid:105) = . ◦ (Law et al. 2009), the dynamicalmass is found to be (1.11 ± × M (cid:12) .The gas mass to dynamical mass ratio for a galactic conversionfactor is therefore M H / M dyn = . ± .
33. For α CO = . M H / M dyn = . ± . We investigate the morphology of the CO(3-2) emission line ofAstarte in relation to other wavelength detections of the system,to closely study the association of the CO component with theUV, optical and IR components, as shown in Figure 3.HST’s observation in the I band (mid-UV rest-frame) do notshow Astarte due to its heavy dust obscuration. However, theyoung stellar population of the less-dusty Adonis is visible inHST’s I band and is bright in the u band detection of CFHT (rest-frame far-UV) and in the J band of VISTA (rest-frame NUV).In the Ks bands of CFHT and VISTA, which correspond to rest-frame visible light, Adonis becomes less-bright, and is very faintat higher wavelengths observations of ALMA and VLA.The dusty Astarte is not visible in the u band of CFHT (rest-frame FUV). It is however detected in the Ks bands of VISTAand CFHT showing a bright stellar population in the visible rest-frame wavelengths. A spatial o ff set (of ∼ ff set between the COS-MOS2015 catalogue and the Gaia reference frame of ∆ ( RA ) = − . + . − . milliarcsec and ∆ ( Dec ) = − . + . − . milliarcsec . Thissystematic o ff set cannot explain the visible o ff set between theCO emission and the rest-frame optical counterparts of Astarte.Moreover, we show that the continuum detected by the VLA at3 GHz of Astarte and its CO emission detected by ALMA arealigned, eliminating the possibility of a systematic error due toALMA’s synthesized beam size.Although the original spectroscopic redshift of 2.140 (forboth sources) found by zCOSMOS (Lilly et al. 2009), was de-rived from the visible range of HST’s observation, where As-tarte is not observed, ALMA o ff ers a spectroscopic redshiftfor the latter (z ALMA = We use the SED modeling code CIGALE (Boquien et al. 2019)to derive the physical properties of our sources. The code allowsto model galaxies’ SED from the UV to the radio wavelengths,taking into account the energetic balance between the emissionabsorbed by dust in the UV-visible range and the IR emission.CIGALE o ff ers a variety of modules for each physical process agalaxy may undergo. The modules that we use in our SED fittingprocedures are described below. To model the stellar component of Astarte and Adonis, we usethe stellar population synthesis of Bruzual & Charlot (2003).This stellar library computes the direct stellar contribution to thespectrum (UV-NIR range) by populating the galaxy with youngand old stars of di ff erent masses, as well as the required gas massthat will produce such population. This model was developedbased on observations of nearby stellar populations and it de-scribes well the various stellar emissions that one expects to en-counter in any galaxy. These models depend on the metallicityand the separation age . We use a solar-like metallicity and wetake into account nebular emission since they contribute to thetotal SED model from the UV to NIR.Di ff erent stellar demographics must be modeled with an ap-propriate SFH in any SED modeling, since it is critical to esti-mate the contribution of the young and old stars to the total flux.An appropriate SFH is key to derive the SFR of a galaxy as itstrongly depends on the assumptions made (Ciesla et al. 2017).CIGALE o ff ers di ff erent SFH scenarios varying from the sim-ple delayed SFH to more complex ones containing episodes ofbursts or sudden drops in SFRs. We use the SFH proposed byCiesla et al. (2017) which is a combination of a smooth delayedbuildup of the stellar population to model the long term SFH ofa galaxy, and a recent flexibility in the last few hundred Myrsto allow for recent and drastic SFR variations (burst or quench).This SFH model has been proven to limit biases by decouplingthe estimations of the stellar mass, mainly constrained by rest http://cigale.lam.fr Age of the separation between the young stellar population and theold one. Article number, page 5 of 10 & A proofs: manuscript no. ulirgz2
VISTA - J band (1.252 μ m)VISTA - Ks band (2.147 μ m) D e c ( J ) D e c ( J ) RA (J2000) RA (J2000) RA (J2000)
HST - ACS I band (0.805 μ m) VLA - S band (3 GHz)CFHT - Ks band (2.146 μ m)CFHT - u band (0.383 μ m) RA (J2000) D e c ( J ) D e c ( J ) Fig. 3.
ALMA-detected CO(3-2) emission line contour map (red contours) of Astarte overlaid on detections from di ff erent telescopes / instrumentsat di ff erent bands as specified in every figure. From upper left to lower right: CFHT U band at 0.383 µ m . HST’s I band at 0.805 µ m . VISTA J bandat 1.252 µ m . CFHT Ks band at 2.146 µ m . VISTA Ks band at 2.147 µ m and on the VLA detection at 3GHz. The outermost contour is 3 σ , and thesubsequent contours are in steps of 1 σ with red innermost contour showing 5 σ . The beam size is 0.78" × frame NIR data, from the SFR, constrained by UV and IR data(e.g., Ciesla et al. 2016, 2018; Schreiber et al. 2018a,b). Thiskind of SFH was used in the study of high-redshift ( z <
3) pas-sive galaxies to model their SED (e.g., Schreiber et al. 2018a,b;Merlin et al. 2018). We limit the recent burst / quench episodes tothe last 100 Myrs of the life of our sources. The recent burst ismotivated by the ALMA detection, however, it is important tonote that this burst makes it di ffi cult to constrain the past SFH.The burst part of the SFH is usually responsible for fitting theUV data, whereas the previous SFH (delayed) is driven by forthe older stellar population, manifested in the visible part of theSED. Two prominent attenuation laws are the ones of Calzetti et al.(2000) (hereafter C00) and Charlot & Fall (2000) (hereafterCF00). They are widely used in the literature, and along withtheir alternations can describe the behavior of the extinction inthe UV to NIR caused by dust.C00 and its recipes is in its core equivalent to reducing the short-wavelength flux coming from a stellar population by an opaquescreen, with the opacity being dependant on the total extinctionof the stellar emission at the B and V bands.Another approach is CF00 power-law which is fundamentallydi ff erent from C00: it attributes di ff erent attenuation to the ISM and to the birth clouds (hereafter BC). This makes the dust moree ff ective at absorbing the UV light since the young stellar emis-sion has to pass through the dust in the BC and the ISM. Starsthat are older are attenuated only by the ISM dust. The CF00 ap-proach is slightly more complex and physical than C00 for highredshift ultra dusty galaxies, embodying di ff erent dust distribu-tions and densities throughout a galaxy.C00 and CF00 rely for their e ffi ciency of attenuating the stel-lar population on power-laws for their slopes. The power-lawslopes for BCs and ISM in CF00 were originally fixed at -0.7each. Lo Faro et al. (2017)’s recipe (hereafter LF17) of CF00was tuned by assuming a power-law for the slope of the atten-uation in the ISM equal to -0.48. This recipe provides a steeperattenuation curve at shorter wavelengths. In this work we usethese three attenuation laws and compare their best fits and theire ff ects on deriving the physical properties of our sources.To assess which attenuation laws to use when di ff erent mod-ules can produce good and comparable fits, we employ theBayesian information criterion (BIC), defined as the χ + k ln n ,where χ is the non reduced goodness of the fit, k is the degreesof freedom of the model and n is the total number of photo-metric fluxes used in the fit of the galaxy. We then evaluate thepreference of a model over the other one by calculating the dif-ference between their BICs: ∆ BIC > ff erence between the two laws and the fit with the lowest χ is preferred. This method was used by Ciesla et al. (2018) for Article number, page 6 of 10. Hamed et al.: Multiwavelength dissection of a massive heavily dust-obscured galaxy and its blue companion at z ∼ SED of Astarte SED of Adonis
Fig. 4.
Best fits of the constructed SEDs of Astarte and Adonis along with their relative residuals. The SED of Astarte (left) is produced usingLF17 attenuation law. The SED of Adonis (right) is produced using CF00 attenuation law. The best fit is shown in black. The unattenuated stellaremission is shown with the blue line. The filled region shows the di ff erence between the unattenuated and the attenuated stellar emission, absorbedby dust. Red dots are the best fit values of the observations which are shown with the purple boxes. Upper limits are shown as purple triangles. Parameter ValuesStar formation history Ciesla et al. (2017)Stellar age ( i ) age main ( ii ) τ main / quench episode t f lex
5, 10, 50, 100 MyrSFR ratio after / before r S FR − , 10 − , 0, 10 , 10 , 10 Stellar synthesis population (Bruzual & Charlot 2003)Initial mass function IMF (Chabrier 2003)Metallicity Z ( iii ) f att (Lo Faro et al. 2017) V-band attenuation in the ISM A
IS MV
IS MV / (A BCV + A IS MV ) µ -0.48 Power law slope of the BC -0.7Dust emission model (Draine et al. 2014)Mass fraction of PAH q
PAH min
10, 25, 30, 40Power law slope α / radio correlation coe ffi cient 2.3 - 2.9 by a bin of 0.1Power law slope slope α synchrotron Table 3.
Input parameters used to fit the SEDs of Astarte and Adonis with CIGALE. (i) The stellar age is the age of the main stellar population. (ii)The e-folding time is the time required for the assembly of the majority of the stellar population. (iii) The reduction factor f att is the color excessin old stars relative to the young ones. carefully choosing successful scenarios of SFHs of quenchinggalaxies, and by Buat et al. (2019) for assessing SEDs of z ∼ To model the dust emission we use the Draine et al. (2014) IRemission models, which was calibrated using high resolution ob-servation of the Andromeda galaxy. Draine et al. (2014) consid-ers a variety of dust grains heated by di ff erent intensities comingfrom the stars and the photodissociation regions, and is an im- Article number, page 7 of 10 & A proofs: manuscript no. ulirgz2
Attenuation law χ reduced χ BICC00 43.22 2.12 73.18Astarte CF00 18.34 0.97 51.30LF17 16.06 0.84 49.01C00 16.28 1.10 46.78Adonis CF00 10.51 0.70 41.01LF17 13.99 0.93 44.49
Table 4.
Comparison between the quality of fits of Astarte and Adonisproduced with CIGALE with the three attenuation laws. proved version of the older Draine & Li (2007) model by vary-ing dust opacity across the radius of a galaxy. This IR modelwas successful in reproducing dust emissions of millions of
Her-schel -detected galaxies as a part of the HELP project (Małeket al. 2018).
The VLA detection at 3 GHz of Astarte and Adonis allow us tomodel the synchrotron emission of our objects, taking into ac-count a non-thermal power law of the synchrotron spectrum andthe ratio of the FIR / radio correlation. The di ff erent parametersused to build our SEDs are shown in Table 3. In the case of Astarte, CF00 and LF17 attenuation laws result inbest SED fits over C00. The BIC of every model was calculatedand is shown in Table 4 along with the other quality of fit assess-ments. ∆ BIC ( C , CF = ∆ BIC ( CF , LF = ∆ BIC ( LF , CF = ∆ BIC ( C , CF = IR of Astarte of the order of 10 , qualifies it to be anUltra Luminous IR Galaxy (hereafter ULIRG). While Adonis’spoor dust content is manifested in the weaker IR luminosity andlower dust mass.To closely inspect the visible dissociation of the gas andthe stellar population in Astarte, we follow the method usedin Buat et al. (2019) by dissecting the stellar continuum andthe IR emission apart and comparing their derived propertieswith the ones obtained using full SEDs. Taking into accountthe UV-NIR data (0.3 - 8 µ m ), the best fit for the stellar contin-uum was obtained with the C00 law, with ∆ BIC ( CF , C = . ∆ BIC ( LF , C = .
7. The better quality fit of the stellarcontinuum produced using C00 is expected, since this power-law e ff ectively attenuates the young stellar population, while theother two laws can be equally e ffi cient in attenuating the olderstars, a behavior that translates into a rise in the near IR ab-sorbed light and therefore a rise in the total IR emission. TheIR luminosity derived from the stellar continuum gives L dust = (2 . ± . × L (cid:12) , relatively close to L dust derived from Physical property Astarte Adonisredshift z CO = . z spec = . IR (10 L (cid:12) ) 3 . ± .
06 0 . ± . M (cid:12) / yr ) 395 ±
20 129 ± (cid:63) ( M (cid:12) ) (3 . ± . × (9 . ± . × M dust (10 M (cid:12) ) 1 . ± .
11 0 . ± . Table 5.
Summary of the physical properties obtained for Astarte andAdonis obtained with CIGALE. log(Stellar mass) (M ) l o g ( s S F R ) ( y r )
10 x MS4 x MSMSPHIBBS, 2 Fig. 5. Relative position of the sSFR (SFR / M (cid:63) ) and stellar mass of As-tarte using the three attenuation laws to the main sequence of Schreiberet al. (2015) at z = 2. The yellow star shows the relative position of Ado-nis to the MS. Magenta squares denote PHIBSS CO-detected SFGs atz ≈ < z < MS × MS × the full SED. The stellar mass derived from the stellar emissiongives (1 . ± . × M (cid:12) and the SFR ( UV − NIR ) = M (cid:12) yr − .From the IR data (MIPS - ALMA continua), we get L dust = (3 . ± . × L (cid:12) , consistent with the one derived withthe full SED. This result is in agreement with the results of Buatet al. (2019) where they found consistent dust luminosities de-rived from both the full SED and the IR data, while L dust deducedfrom the stellar continuum was underestimated. 4. Discussion Figure 5 shows the relative position of our galaxies to the MSof Schreiber et al. (2015). Adonis lies on 10 × MS qualifying itto be a strong starburst, despite its relatively low SFR. Whilebeing a starburst, this type of source cannot be detected even bythe deepest 3 mm ALMA survey (expected flux of 7 µ Jy), whichhave a 1- σ noise of 9.7 µ Jy González-López et al. (2020).Astarte is a MS galaxy with all the di ff erent attenuation recipesused. However, there is a clear di ff erence concerning the positionof Astarte relative to the MS as a result of the three attenuationlaws. This is attributed to the significant di ff erence in the derivedstellar masses, with CF00 and LF17 attenuation laws resultingin larger stellar mass than C00 due to the highest attenuation inNIR. This contributes to a lower specific SFR (sSFR = SFR / M (cid:63) )since SFRs do not di ff er significantly with the three laws.Host halo masses of such z ∼ Herschel -detected massive MSgalaxies were investigated in Béthermin et al. (2014) using clus- Article number, page 8 of 10. Hamed et al.: Multiwavelength dissection of a massive heavily dust-obscured galaxy and its blue companion at z ∼ log(sSFR) ( yr ) l o g ( M o l e c u l a r g a s ) ( M ) PHIBBS, 2 Molecular gas masses derived with the CO conversion factorversus the sSFR. The magenta squares are from T13. Turquoise circlesare ULIRGs at 2 < z < α CO = . 8, and the associated error bar shows thevariation of the molecular gas mass using 0.3 < α CO < 1. The blue triangleshows the molecular gas of Astarte for a galactic CO conversion factorof 4.36. L IR (L ) L C O ( ) ( K k m s p c ) SBMS PHIBBS, 2 Correlation between CO(1-0) luminosities and the total IR lu-minosity. The magenta-filled squares are from T13 and turquoise-filledcircles are the sources from G10. The solid black line is the linear re-gression for MS galaxies (Sargent et al. 2014) and the dashed line is thatfor SBs (the regression lines are from the complete sample in Sargentet al. 2014). tering and X-ray stacking and were found to reside in halos of > M (cid:12) . Such halo masses are also expected from the stellarmass to halo mass relation (Behroozi et al. 2013; Durkalec et al.2018; Behroozi et al. 2019). Astarte is ∼ ∼ α CO = α CO = . α CO = . 36 produces a higher molecular gasmass with respect to its sSFR.In Figure 7 we show the correlation between CO luminosi-ties and the total infrared luminosities. IR luminosities were de-rived from the SFRs of all the sources (G10, T13 and Astarte)using the Kennicutt relation (Kennicutt 1998). The initial choiceof r = . ± . 07 (the average in Daddi et al. 2015) places As-tarte on the SB line from Sargent et al. (2014), contradicting itsSED result. We therefore investigate the lowest excitation ratiofrom Daddi et al. (2015) of r = . ± . 07. This lower ratiomoves Astarte closer to the MS within the error bars.Using the total molecular gas mass of Astarte derived withthe least excited CO(3-2) from Daddi et al. (2015) ( r = . ± . f gas = M gas / ( M gas + M (cid:63) ) = (0 . ± . ≈ α CO adapted for SB with a higher r ratioreduces the gas fraction significantly.The gas mass derived with a galactic conversion factor givesAstarte a rather small depletion time of 0 . ± . 07 Gyrs, mak-ing it very e ffi cient at forming stars (for comparison, SBs havea depletion time of the order of ∼ 100 Myrs, (see Fig. 10 inBéthermin et al. 2015). Recently, Elbaz et al. (2018) found thatcompact SFGs on the MS with a relatively low depletion time arenot uncommon. These active ultra-massive objects can be hiddenat the higher end of the tail of the MS. The average depletiontime for Elbaz et al. (2018) galaxies is around 0.25 Gyrs, and al-though we do not detect the continuum of Astarte with ALMA,its CO emission is compact as it is the case for the continuum ofALMA-detected galaxies from Elbaz et al. (2018). This is alsoconfirmed in Puglisi et al. (2019) where compact massive galax-ies at the top of the MS exhibit high SFRs at their cores followingtheir SB epoch. 5. Conclusion In this paper, we analyzed two galaxies, Astarte and Adonis,at the peak of the SFR density using multiwavelength dissec-tion combining ALMA observations with UV-submm SED mod-eling. We investigated the molecular gas content of Astartethrough the ALMA detection of its CO(3-2) emission, relyingon di ff erent excitation ratios of L (cid:48) CO (3 − / L (cid:48) CO (1 − and di ff erentCO conversion factors. A galactic conversion factor when usedalong with the least excitation ratio from Daddi et al. (2015) con-firmed the relative position of Astarte to the MS, as found fromits SED modeling. Although the obtained gas fraction is on thelower limits of that in MS galaxies (Santini et al. 2014; Béther-min et al. 2015), a possible explanation might be that the CO(3-2) instantaneous emission does not fully recover the molecularmass and the dynamics of Astarte, due to its weak excitation(Daddi et al. 2015). Detections of other transition levels of COwould be helpful to better constrain the molecular mass of As-tarte, and therefore its physical characteristics.The physical dissociation of the CO line and the rest-framestellar population in Astarte was also investigated as done inBuat et al. (2019), by deriving physical properties from the stel-lar emission (UV-NIR) and the IR emission apart. As in Buat Article number, page 9 of 10 & A proofs: manuscript no. ulirgz2 et al. (2019), the dust luminosity derived from the full SED is inagreement with the one derived from the the IR emission, while L dust derived from the stellar emission is slightly underestimated.Furthermore, C00 attenuation law was preferred when fitting thestellar continuum only. This is consistent with the results of Buatet al. (2019) for galaxies with the same radii extension of rest-frame stellar emission and ALMA-detected emission. LF17 at-tenuation law, which was tuned for ULIRGs at z ∼ ∼ . ffi cient in forming stars, whose depletion time is an order ofmagnitude lower than what is expected in typical MS galaxies(Béthermin et al. 2015). This SB-like star formation activity onthe MS was found for massive compact SFGs in Puglisi et al.(2019), in their post-SB phase. Low depletion times of MS mas-sive galaxies were also found in Elbaz et al. (2018), confirmingthat Astarte is caught in the middle of quenching following anearlier SB activity.Central galaxies at z ∼ ∼ Acknowledgements. M.H. and K.M have been supported by the NationalScience Centre (UMO-2018 / / E / ST9 / / JAO.ALMA / NRAO and NAOJ. References