The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies
E.-D. Paspaliaris, E.M. Xilouris, A. Nersesian, V.A. Masoura, M. Plionis, I. Georgantopoulos, S. Bianchi, S. Katsioli, G. Mountrichas
AAstronomy & Astrophysics manuscript no. main © ESO 2021February 9, 2021
The physical properties of local (U)LIRGs: a comparison withnearby early- and late-type galaxies
E.-D. Paspaliaris , (cid:63) , E.M. Xilouris , A. Nersesian , , V.A. Masoura , , M. Plionis , , I. Georgantopoulos ,S. Bianchi , S. Katsioli , , and G. Mountrichas National Observatory of Athens, Institute for Astronomy, Astrophysics, Space Applications and Remote Sensing, Ioannou Metaxaand Vasileos Pavlou GR-15236, Athens, Greece Department of Astrophysics, Astronomy & Mechanics, School of Physics, Aristotle University of Thessaloniki Sterrenkundig Observatorium, Universiteit Gent, Krijgslaan 281 S9, 9000 Gent, Belgium INAF - Osservatorio Astrofisico di Arcetri, Largo E. Fermi 5, 50125, Florence, Italy Department of Astrophysics, Astronomy & Mechanics, Faculty of Physics, University of Athens, Panepistimiopolis, GR15784Zografos, Athens, Greece Instituto de Fisica de Cantabria (CSIC-Universidad de Cantabria), Avenida de los Castros, 39005 Santander, SpainReceived / Accepted
ABSTRACT
Aims.
In order to pinpoint the place of the (ultra-) luminous infrared galaxies [(U)LIRGs] in the local universe we examine theproperties of a sample of 67 such nearby systems and compare them with those of 268 early- and 542 late-type, well studied, galaxiesfrom the DustPedia database.
Methods.
We make use of multi-wavelength photometric data (from the ultra-violet to the sub-millimetre), culled from the literature,and the CIGALE Spectral Energy Distribution fitting code, to extract the physical parameters of each system. The median SpectralEnergy Distributions as well as the values of the derived parameters were compared to those of the local early- and late-type galaxies.In addition to that, (U)LIRGs were divided into seven classes, according to the merging stage of each system, and variations in thederived parameters were investigated.
Results. (U)LIRGs occupy the ‘high-end’ on the dust mass, stellar mass, and star-formation rate plane in the local Universe withmedian values of 5.2 × M (cid:12) , 6.3 × M (cid:12) and 52 M (cid:12) yr − , respectively. The median value of the dust temperature in (U)LIRGs is32 K, which is higher compared to both the early-type (28 K) and the late-type (22 K) galaxies. The dust emission in PDR regions in(U)LIRGs is 11.7% of the total dust luminosity, significantly higher than early-type (1.6%) and the late-type (5.2%) galaxies. Smalldi ff erences, in the derived parameters, are seen for the seven merging classes of our sample of (U)LIRGs with the most evident onebeing on the star-formation rate, where in systems in late merging stages (‘M3’ and ‘M4’) the median SFR reaches up to 99 M (cid:12) yr − compared to 26 M (cid:12) yr − for the isolated ones. In contrast to the local early- and late-type galaxies where the old stars are the dominantsource of the stellar emission, the young stars in (U)LIRGs contribute with 64% of their luminosity to the total stellar luminosity. Thefraction of the stellar luminosity absorbed by the dust is extremely high in (U)LIRGs (78%) compared to 7% and 25% in early- andlate-type galaxies, respectively. The fraction of the stellar luminosity used to heat up the dust grains is very high in (U)LIRGs, forboth stellar components (92% and 56%, for the young and the old stellar populations, respectively) while 74% of the dust emissioncomes from the young stars. Key words. galaxies: evolution - galaxies: ISM - galaxies: interactions - dust, extinction - galaxies: star-formation - galaxies: stellarcontent
1. Introduction
A multi-wavelength approach is needed for a comprehensivestudy of galaxies. Each region of the electromagnetic spectrumprovides unique information about the di ff erent building blocksof galaxies (stars, dust, and gas) and their physical properties.These building blocks constituting the baryonic matter in galax-ies are not in isolation, but, instead, are constantly interactingwith each other, modifying the stellar content and the inter-stellar medium (ISM). This evolutionary process is imprintedon the galaxy’s spectral energy distribution (SED). Studying theSEDs of galaxies is, thus, a key procedure towards understand-ing galaxy formation and evolution.One property, often related to the degree of the currentstar-formation is the infrared (IR) luminosity ( L IR = L − µ m ), (cid:63) e-mail: [email protected] which, in most of the cases, is dominated by the emission ofdust grains heated by the interstellar radiation field (ISRF).Early-type galaxies (elliptical and lenticular galaxies) exhibitvery low to moderate IR-luminosities ( L IR < L (cid:12) ), whilethe star-forming spiral galaxies are brighter in the infrared(10 < L IR / L (cid:12) < ; Lonsdale et al. 2006). Imaging studies of(Ultra) Luminous Infrared Galaxies [(U)LIRGs; L IR > L (cid:12) ]in the local Universe, indicate a more violent kind of forma-tion of these systems with the majority of them exhibiting signsof galaxy interactions (Sanders & Mirabel 1996; Duc et al.1997). Armus et al. (1987) concluded that more than ∼ ff ∼ Article number, page 1 of 33 a r X i v : . [ a s t r o - ph . GA ] F e b & A proofs: manuscript no. main redshift s m M1 M2 M3 M4 M55101520 merging class log L IR [ L fl ] frac AGN N u m b e r o f g a l a x i e s Fig. 1.
Distributions of the basic properties of our galaxy sample. Theredshift distribution is plotted in the top-left panel and the number ofsources per merging class in the top-right panel. In the bottom-leftpanel the infrared luminosity distribution, based on IRAS observations(Sanders et al. 2003) is shown, while, the bolometric AGN fractionscalculated in Díaz-Santos et al. (2017), are plotted in the bottom-rightpanel. other studies concluded in fractions of such sources lower than70% (Lawrence et al. 1989; Zou et al. 1991; Leech et al. 1994).In all cases, however, it becomes clear that galaxy merging is akey process that is responsible for the high IR-luminosity anddrives the high level of star-formation activity observed in thesesystems, making them a unique class of objects in the local Uni-verse.The mid-infrared (MIR) part of the SED is a very com-plex regime where di ff erent components of the galaxy contributeto the emission. The SED in these wavelengths can be a mix-ture of light originating from old stellar populations, emissionof dust grains (mainly heated by newly formed stars), as wellas radiation emitted by an active galactic nucleus (AGN). In(U)LIRGs, the MIR spectra can probe the conditions in theirstar-forming regions, to which they owe their high luminosi-ties. Although galaxy interactions can trigger and enhance nu-clear activity in local (U)LIRGs, star-formation processes domi-nate the MIR emission in the majority of these systems (Genzelet al. 1998; Petric et al. 2011; Stierwalt et al. 2013). The UVand optical radiation emitted by the newly formed stars, in dustymolecular clouds, is absorbed and then re-emitted by dust in in-frared wavelengths. Dust emission is seen either through emis-sion features (e.g. Polycyclic Aromatic Hydrocarbons; PAHs) orcontinuum emission originating from the photo-dissociation re-gions (PDRs), as well as dust di ff usely distributed throughoutthe galaxy.One way to decompose the SED of a galaxy and extract use-ful information on the di ff erent components contributing theiremission is by using SED-fitting codes. Such codes usually makeuse of templates for the emission of di ff erent components com-bined in such a way so that energy is conserved. Assuming twodi ff erent stellar populations (old and young stars) and a certainvariation of the star-formation activity through cosmic time (i.e.the star-formation history; SFH), several characteristic proper-ties of the galaxies can be determined by fitting their observedSEDs. Such properties are stellar and dust masses, luminosi-ties, and the star-formation rate ( SFR ). In addition to the stel- lar and dust components, an AGN component can be consid-ered when needed. The decomposition of the galactic SEDs haslead to versatile studies of the properties of galaxies (Giovannoliet al. 2011; Małek et al. 2014; Pappalardo et al. 2016; Vika et al.2017), and more specifically, studies about the attenuation pro-cesses (Boquien et al. 2013; Salim et al. 2018) or the contributionof the di ff erent stellar populations in the heating of the dust andthe dependence on the morphological type (Bianchi et al. 2018;Nersesian et al. 2019).In this paper we model the SEDs of 67 IR-luminous localgalaxies, in order to derive their physical properties and to inves-tigate how these properties compare with other types of galaxiesin the local Universe. We explore how several physical proper-ties, such as the SFR , the stellar mass and the dust mass, dusttemperature, and PDR luminosity vary with galaxy type. Divid-ing (U)LIRGs in di ff erent classes, according to the interaction oftheir parent galaxies, we investigate the evolution of all the afore-mentioned parameters with merging stage. Finally, we examinethe relative contributions of the di ff erent stellar populations tothe bolometric luminosity of galaxies and their role in the dustheating. By comparing our results with a recent study on passiveand star-forming galaxies (Nersesian et al. 2019), we determinethe role of each stellar population in the dust heating and for thedi ff erent types of galaxies in the local universe.This paper is structured as follows. The sample analysed inthis work is presented in Sect. 2 followed by a description of theSED fitting method in Sect. 3. The results of our analysis as wellas a comparison of the properties of the (U)LIRGs in our samplewith those of ‘normal’ galaxies in the local universe are exploredin Sec. 4. The evolution of the physical properties of (U)LIRGswith merging stage is investigated in Sect. 5, while the di ff erentstellar populations and their role in the heating of the dust grainsis discussed in Sec. 6. Our findings are summarised in Sec. 7.Finally, a mock analysis, performed by CIGALE, is presented inAppendix A, the best-fit SED models are shown in Appendix B,a comparison with other studies is performed in Appendix C,while the cumulative distributions of various physical parame-ters are presented in Appendix D.
2. The sample
Our main sample consists of 67 local (U)LIRGs (see Ta-ble 1) drawn from the Great Observatories All-sky LIRG Sur-vey (GOALS; Armus et al. 2009). All sources have su ffi cientwavelength coverage with an average of 8 observations in theUV / optical / NIR ( < µ m) part of the spectrum and 16 in theMIR / FIR / submm ( > µ m) part of the spectrum (see Table 1,but also the corresponding SEDs in Fig. B.1). This allows us toconduct a full SED fitting study and derive valuable informationon the intrinsic properties of the sources.All galaxies have been observed by the Herschel Space Ob-servatory (HSO; Pilbratt et al. 2010) with the correspondingphotometry published in Chu et al. (2017). In this study, totalsystem fluxes (if systems of galaxies consist of more than onegalaxy) and component fluxes (where possible) were computedfor all six
Herschel bands [PACS (70, 100, and 160 µ m) andSPIRE (250, 350, and 500 µ m)]. The photometric apertures werecarefully chosen, first by visual inspection, and, subsequently, byplotting the curve of growth and checking that all of the flux wasincluded in.Apart from the Herschel photometry, the rest of the multi-wavelength data were compiled from two literature resources,the photometry presented in U et al. (2012) (for 64 galax-ies) and the photometry presented in Clark et al. (2018) (for 3 Article number, page 2 of 33.-D. Paspaliaris et al.: The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies
Table 1.
Properties of the (U)LIRGs in our sample.
IRAS name alternative name redshift merging class log ( L IR ) [L (cid:12) ] ( a ) f rac AGN ( b ) bands; λ< µ m bands; λ> µ mF00085-1223 NGC 0034 0.0196 M4 11.49 0.04 ± / ± ± / B 0.0201 M3 11.71 0.12 ± ± + ± ± + ± + ± ± + ± + ± + + + ± ± ± + ± + ± + + ± + ± + ± + + / ± + / ± ± + + ± + + ± + + ± ± + / B 0.0340 M1 11.64 0.05 ± + / IC 694 0.0103 M3 11.93 0.04 ± + + ± ± + (cid:63) + ± ± + ± + ± + ± + ± ± ± + ± + / ± + ± + ± ± + ± + ± + + + ± + ± + ± + ± + / ± + + ± ± ± + / IC 5283 0.0163 M2 11.65 0.24 ± + ± ± + ± ± + ± + / ± + ± ( a ) L IR (from Sanders et al. 2003). ( b ) Average bolometric AGN fractions by Díaz-Santos et al. (2017), or (cid:63) in the range 5-1000 µ m , by Fritz et al. (2006) galaxies; F06107 + database). All the details for the data assembly and http: // dustpedia.astro.noa.gr / Article number, page 3 of 33 & A proofs: manuscript no. main treatment are discussed in the relevant papers. In summary, theSpace-based observations include Ultraviolet data from GALEX(Morrissey et al. 2007, FUV; λ e ff = . µ m, and NUV; λ e ff = . µ m), MIR data from Spitzer / IRAC (Werner et al. 2004, 3.6,4.6, 5.4, and 8 µ m) and FIR data from Spitzer / MIPS (Werneret al. 2004, 24, 70, and 160 µ m). IRAS data at 12, 25, 60, and 100 µ m as published in the Revised Bright Galaxy Sample (RBGS;Sanders et al. 2003) have also been incorporated into our anal-ysis. Concerning the ground-based data, the main source in op-tical wavebands are observations taken with the University ofHawaii (UH) 2.2 m Telescope on Mauna Kea with the remainingoptical data compiled from the literature and the NASA / IPACExtragalactic Database (NED; see U et al. 2012, and referencestherein). NIR J, H, K s were extracted from the Two MicronAll Sky Survey (2MASS; Skrutskie et al. 2006). Submillime-ter data at 850 µ m and 450 µ m, obtained using the Submil-limeter Common User Bolometer Array (SCUBA) at the JamesClerk Maxwell Telescope were taken from Dunne et al. (2000)and Dunne & Eales (2001). In both of these studies, a carefulaperture-matched photometry was performed to the datasets, se-curing a consistent photometry among the di ff erent photometricbands. In the case of U et al. (2012) masked photometry has beenextracted from the images taken at e ff ective wavelengths 0.15 µ m < λ e f f < µ m and at MIPS 24 µ m band with the masks de-fined based on isophotes in the median- and boxcar-smoothed I -band images at the surface brightness limit of 24.5 mag arcsec − so that the global flux from tidal debris as well as individualcomponents within these merger systems are encapsulated. Forthe remaining wavelengths, masks would not improve the pre-cision of the total flux, due to lack in resolution, so circular orelliptical apertures were used. In Clark et al. (2018) the photom-etry and uncertainty estimation, in all bands, is done within amaster elliptical aperture which is found after combining the el-liptical apertures fitted in each band, separately, in such a way sothat the di ff erent beam sizes are taken into account.Out of these 67 systems, the vast majority (56) are LIRGs(L IR > L (cid:12) ), whereas 11 are ULIRGs (L IR > L (cid:12) ). Theirredshifts range from 0.003 to 0.083 with a mean value of 0.029,while ∼
80% of the galaxies are in the redshift range between0.01 and 0.04. Their IR luminosities (based on
IRAS photome-try, by Sanders et al. 2003), as well as their redshift distributionare presented in Fig. 1 (see also Table 1).Since merging is a key process of the regulation of the prop-erties of these systems, we also wanted to analyse the proper-ties of our sample considering their di ff erent merging stages. Forthe classification of their stage we have used the categorizationadopted in Larson et al. (2016). In that paper, optical imagingdata from the GOALS HS T sample (PID: 10592, PI: Evans; Kimet al. 2013) as well as optical I -band images obtained with theUH 2.2 m telescope on Mauna Kea by Ishida (2004) were usedto visually classify the galaxies. Several criteria were imposed toconduct the classification, all discussed in detail in Larson et al.(2016). For the sake of completeness we, briefly, repeat here thebroad characteristics of the seven di ff erent merging classes alongwith their abbreviation: s Single galaxies: galaxies that show no current signof an interaction or merger event. m Minor mergers: interacting pairs with estimatedmass ratios > M1 Major merger-stage 1: pairs that appear to be ontheir initial approach and have no prominent tidalfeatures; with low relative velocity and galaxy sep-aration, ∆ V <
250 km s − , n sep <
75 kpc. M2 Major merger-stage 2: interacting galaxy pairs withobvious tidal bridges and tails (Toomre & Toomre1972) or other disturbances consistent with havingalready undergone a first close passage. M3 Major merger-stage 3: merging galaxies with mul-tiple nuclei. These systems have distinct nucleiin disturbed, overlapping disks, along with visibletidal tails. M4 Major merger-stage 4: galaxies with apparent sin-gle nuclei and obvious tidal tails. The galaxy nucleihave n sep ≤ M5 Major merger-stage 5: galaxies which appear to beevolved merger remnants. These galaxies have dif-fuse envelopes which may exhibit shells or otherfine structures (Schweizer & Seitzer 1992) and asingle, possibly o ff -center nucleus. These mergerremnants no longer have bright tidal tails.Examples of galaxies in the di ff erent classes are illustrated inFig. 1 of Larson et al. (2016).Two of the galaxies in our sample (namely, F10173 + + ff use disk and a possible disconnected di ff use tail ( ∼ ff use structureis the remnant of a tidal tail. In order to ease the analysis wehave chosen to keep their most obvious classification in place(i.e. ‘M4’ for F10173 + + ff er-ent properties) are forced to be treated as one entity. In such sys- Article number, page 4 of 33.-D. Paspaliaris et al.: The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies
Table 2.
The grid of input parameters used in CIGALE for the best-fit model computation. The combination of these parameters results to14,515,200 models per redshift bin. With 9 redshift bins used, 130,636,800 models are computed in total. The parameters in the brackets indicatethe exact naming of each module as it is used in CIGALE.
Parameters Values star-formation history delayed + burst + quenching (i) [[sfhdelayedbq]] τ main , e-folding time [Myr] 1000, 3000, 5000, 10000 t gal , galaxy age [Myr] 4500, 7000, 9500, 12000 t bq , quenching or bursting age [Myr] 10, 20, 30, 100 r SFR , ratio of the
SFR after / before quenching or bursting age 1.0, 3.16, 10.0, 20.0, 100.0, 1000.0, 1200.0 stellar population model Bruzual & Charlot (ii) [[bc03]]initial mass function Salpeter (iii) metallicity 0.02 dust attenuation modified Calzetti (iv) [[dustatt_calzleit]]
E(B-V) , colour excess of the young stellar population 0.20, 0.29, 0.44, 0.66, 1.0
E(B-V) old / E(B-V) young , reduction factor for
E(B-V) δ , power law index of the attenuation curve -0.5, -0.25, 0.0 dust grain model THEMIS (v) [[themis]] q hac , fraction of small hydrocarbon solids 0.02, 0.17 U min , minimum radiation field 5, 10, 25, 50, 80 α , power law index of the radiation field 2.0 γ , fraction illuminated from U min to U max active nucleus model Skirtor (vi) [[skirtor2016]] τ . , optical depth at 9.7 µ m 3.0, 7.0pl, torus density radial parameter 1.0q, torus density angular parameter 1.0oa, angle between the equatorial plan and edge of the torus [deg] 40R, ratio of outer to inner radius 20Mcl, fraction of total dust mass inside clumps [%] 97i, inclination (viewing angle) [deg] 30 (type 1), 70 (type 2)AGN fraction 0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.85extinction law of polar dust SMCE(B-V) of polar dust 0, 0.8polar dust temperature [K] 100polar dust emissivity index 1.6References: (i) Ciesla et al. (2016); (ii) Bruzual & Charlot (2003); (iii) Salpeter (1955); (iv) Calzetti et al.(2000); (v) Jones et al. (2017); (vi) Stalevski et al. (2012, 2016). tems we assume that the SED modelling performed, producedan average description of each multiple system.Since nearby (U)LIRGs constitute a separate class of galac-tic systems in the local Universe (which, by definition, showhigh IR luminosities and enhanced SFR ) we need to have a ref-erence sample of ‘normal’ galaxies [in the sense of being lessactive in forming stars; either passive (ETGs) or low level star-forming galaxies (LTGs)] to compare with. The most complete,to date, sample of local galaxies is that of DustPedia (Davieset al. 2017; Clark et al. 2018). This sample includes 875 galaxies(with distances of less than 40 Mpc) all having
Herschel detec-tions, with D > (cid:48) ( D being the major axis isophote at whichoptical surface brightness falls beneath 25 mag arcsec − ). Addi-tionally, the DustPedia sample includes, mostly, isolated galax-ies that have a detected stellar component; WISE observations at3.4 µ m are the deepest all-sky data sensitive to the stellar compo-nent of galaxies, and hence provide the most consistent way ofimplementing this stellar detection requirement. For full sampledetails see Davies et al. (2017), and Clark et al. (2018). Further-more, it contains galaxies of various morphologies parametrizedby the Hubble Stage (T; Makarov et al. 2014) ranging from T = -5 (pure elliptical galaxies) to T =
10 (irregular galaxies). Through-out our analysis we consider as ETGs galaxies with T < ≥
3. SED modelling
In order to exploit the information hidden in the SEDs of thegalaxies we make use of CIGALE (see Boquien et al. 2019, andreferences therein). CIGALE models the SED of each galaxy,by selecting a suitable set of parameters, through a Bayesian ap-proach, that best represents the observed SED. A basic principleof the code is the conservation of the energy between the amountabsorbed by the dust and that re-emitted by the dust grains in in-frared and submm wavelengths (Roehlly et al. 2014). In the cur-rent study we use the version 2020.0 of CIGALE. Beyond many
Article number, page 5 of 33 & A proofs: manuscript no. main
Table 3.
CIGALE-derived physical properties of the (U)LIRGs in our sample id log M star [M (cid:12) ] log M dust [M (cid:12) ] T dust [K] SFR [M (cid:12) yr − ] f rac AGN reduced χ F00085-1223 10.66 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± + ± ± ± ± ± ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± ± ± ± ± ± + ± ± ± ± ± ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± ± ± ± ± ± + ± ± ± ± ± ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± + ± ± ± ± ± improvements and optimisations related to the diminution of thecomputational time and the estimation of the physical proper-ties, the most important new feature is related to the AGN modelused. In addition to the active nucleus model formulated by Fritz et al. (2006), the SKIRTOR module (Stalevski et al. 2012, 2016)has been added. Following recent theoretical and observationalstudies (Nikutta et al. 2009; Ichikawa et al. 2012; Stalevski et al.2012; Tanimoto et al. 2019), SKIRTOR assumes a two-phase Article number, page 6 of 33.-D. Paspaliaris et al.: The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies dusty clumpy torus, where most of the dust has high density andis clumpy, while the rest is smoothly distributed. SKIRTOR isbased on the 3D radiative transfer code SKIRT (Baes et al. 2011;Camps & Baes 2015). In addition, this version of CIGALE in-cludes not only torus obscuration, but also obscuration by dustsettled along polar directions, allowing for a more precise treat-ment of both type-1 and type-2 AGN cases (Bongiorno et al.2012; Elvis et al. 2012; Stalevski et al. 2017, 2019; Lyu & Rieke2018).The stellar radiation field is built based on the Bruzual &Charlot (2003) population synthesis model and a Salpeter (1955)initial mass function (IMF). After the fitting procedure, the stel-lar emission can be decomposed into two distinct populations, anold and a young, depending on the SFH of each galaxy. To ac-count for the starlight attenuation by dust, a modified Calzettiet al. (2000) starburst attenuation law is applied (Noll et al.2009), to the intrinsic spectra of the di ff erent stellar populations.For the dust emission properties we adopted the THEMIS model(Jones et al. 2017). In addition, we use a flexible SFH that getsan analytical expression of a delayed SFH allowing for an instan-taneous burst (or quenching) of the star-formation activity (see,Ciesla et al. 2015 and Nersesian et al. 2019). Since all of thegalaxies in our sample are actively forming stars we only use thebursting mode of the delayed SFH (over the last 10 to 100 Myr)with six cases of instantaneous bursts (at various levels) and onecase with no burst. The nuclear activity of local (U)LIRGs has al-ready been investigated by others (e.g., Nardini et al. 2010; Ricciet al. 2017; U et al. 2019). In a series of two papers [C-GOALSI (Iwasawa et al. 2011) and C-GOALS II (Torres-Albà et al.2018)], a total of 107 (U)LIRGs of the GOALS sample, wereobserved with the Chandra
X-ray Observatory revealing AGNsignatures for 31 ±
5% and 38 ±
7% of the two sub-samples re-spectively. A careful treatment of the AGN component is, there-fore, necessary for the purposes of the current study. To accountfor a realistic description of the AGN emission we make use ofthe SKIRTOR module, that also includes polar-dust extinction.The parameter space of this module follows the parametrizationpresented in Yang et al. (2020), except for the values of polardust colour excess, E(B-V), where only two values were used(0, 0.8). This was indicated by Mountrichas et al. (2020), whofind that CIGALE is not very sensitive in this parameter.Concerning the input parameters used in CIGALE, exceptfrom those defining the AGN module, we adopt a similar param-eter space as the one used in Nersesian et al. (2019), modifiedaccordingly so that it is able to successfully model the SEDs ofthe systems in the current sample. By performing several test-runs with CIGALE, we were able to minimize the number ofvalues that control the SFH and the dust modules, in favour ofthe SKIRTOR module, taking into account the computational de-mands imposed by the total number of parameters. These test-runs indicated values of the parameters that were not used byCIGALE and thus were excluded from the parameter space, re-sulting in the final input parameter grid listed in Table 2. Theattenuation law considered here (Calzetti et al. 2000) was alsotested against that of the widely adopted Charlot & Fall (2000)attenuation law resulting in small di ff erences, and within the un-certainties, on the shape of the SEDs and on the output parame-ters.In order to assess whether or not physical properties can ac-tually be estimated in a reliable way we have performed a mockanalysis by using the relevant feature built in CIGALE. The re-sults of the mock analysis are presented and discussed in Ap-pendix A. The modelled SEDs of all the systems in our sam-ple are presented in Fig. B.1. To further evaluate the CIGALE- derived properties, we perform a comparison with other studiesin Appendix C.
4. The physical properties and SEDs of local(U)LIRGs
The values and associated uncertainties of the physical prop-erties ( M star , M dust , T dust , SFR , f rac AGN ), as derived byCIGALE, for each source in our sample, along with the re-duced χ of the fit, are listed in Table 3. (U)LIRGs are sys-tems of enhanced IR luminosity and, as such, are expectedto show up as the dustiest and most actively star-forminggalaxies in the local Universe (as already shown in previ-ous studies, see, e.g. da Cunha et al. 2010a and U et al.2012). The mean dust mass of the galaxies in our sample is (cid:104) M dust (cid:105) = . × M (cid:12) ranging from M dust = . × M (cid:12) (forF13126 + M dust = . × M (cid:12) (for F15327 + SFR we compute a mean value of (cid:104)
SFR (cid:105) =
81 M (cid:12) yr − ranging from 2.7 M (cid:12) yr − (for F13197-1627) to 410.8 M (cid:12) yr − for (F14348-1447). As a compari-son, the dustier and more actively star-forming ‘normal’ galax-ies in the local Universe (those of morphological classesSb-Sc) show an average dust mass of 2 × M (cid:12) andan average SFR of 2 . (cid:12) yr − (Nersesian et al. 2019).(U)LIRGs are also amongst the most massive galaxies in thelocal Universe with a mean value of their stellar mass of (cid:104) M star (cid:105) = . × M (cid:12) ranging from M star = . × M (cid:12) (forF12224-0624) to M star = . × M (cid:12) (for F14547 + . × M (cid:12) (Ners-esian et al. 2019). The SEDs of galaxies have been proven to be valuable assetsfor characterising their content in stars and dust, as well as theircurrent star-formation activity. Even by visual inspection of theSEDs one can spot the di ff erences between di ff erent populationsand this is what we do next by comparing, in a statistically sig-nificant and systematic way, the SEDs of local ETGs, LTGs, and(U)LIRGs.In our analysis we have modelled the SEDs of the 67 lo-cal (U)LIRGs in our sample, using CIGALE, and compare themwith those of 268 Early-Type Galaxies (ETGs) and 542 Late-Type Galaxies (LTGs), already modelled in Nersesian et al.(2019). This comparison is shown in Fig. 2 with the medianSEDs of ETGs, LTGs, and (U)LIRGs on the top-left, top-right,and bottom-left panels, respectively. In the bottom-right panelwe present the median SEDs of ETGs, LTGs, and (U)LIRGs(red, blue, and yellow curves respectively) so that a direct com-parison can be made. For each of the three classes, both the at-tenuated and the unattenuated SEDs are also shown (thick solidlines and thin dotted lines, respectively, in the UV-optical partof the SED). The most obvious change among them is the dif-ferences in the IR and UV wavelength regimes of the SEDs,and more specifically, at the peak of the dust emission (around100 µ m) and the peak of the stellar emission (around 1 µ m).What is clearly visible is that the dust emission shows a largevariation with the peak being higher by 1.17 dex for LTGs (com-pared to ETGs), and higher by 0.56 dex for (U)LIRGs (comparedto LTGs). This already dictates the di ff erences in dust contentamong the three populations with ETGs being the most devoidand (U)LIRGs the abundant ones. The slight shift in the peak Article number, page 7 of 33 & A proofs: manuscript no. main N o r m a li s e d f l u x ETGs
Median model spectrumMedian stellar youngMedian stellar old Median diffuse dustMedian PDR dustMedian AGN
LTGs Rest-frame wavelength [ m] N o r m a li s e d f l u x (U)LIRGs Rest-frame wavelength [ m]
ETGsLTGs(U)LIRGs
Fig. 2.
Template SEDs of the DustPedia ETGs (top-left panel), LTGs (top-right panel) and the local (U)LIRGs (bottom-left panel), as modelledby CIGALE. The black curves indicate the median SEDs, while the blue and the red curves denote the median unattenuated SEDs of the youngand old stellar populations, respectively. The median distribution of the di ff use dust is shown by the orange curve and the emission from the PDRsis shown by the green curve. The purple curve, in the bottom-left panel, indicates the median AGN distribution. The shaded areas cover the rangebetween the 16 th and 84 th percentiles to the median values. For clarity reasons, since the PDR and the AGN components may vary substantially,we choose not to present the percentile widths of the curves. The median SEDs of the three di ff erent types of local galaxies are plotted in thebottom-right panel. Red curve represents the median, observed, SED of ETGs, blue curve represents the one of LTGs, while the yellow curvecorresponds to the median SED of (U)LIRGs. The thin, dotted, curves show the unattenuated total stellar emission for the di ff erent classes. Thereader is also referred to the corresponding median SEDs in Bianchi et al. (2018) and Nersesian et al. (2019). (in the wavelength axis) seen in the dust SED inclines towardsa variation in dust temperature with LTGs being the cooler onesand (U)LIRGs the warmest ones. The actual wavelength wherethe peak of the FIR emission happens is at 96, 116, and 70 µ m forETGs, LTGs, and (U)LIRGs, respectively. The dust emission, astreated by CIGALE, consists of two components, one account-ing for the di ff use dust and one accounting for the PDRs, mostlyassociated with the star-forming regions (Hollenbach & Tielens1999). From Fig. 2 (green curves) we see that the significanceof the dust in the PDR regions increases from ETGs (havingnegligible, practically non existent emission) to LTGs (with asignificant, but still, less dominant contribution compared to thedi ff use dust) to (U)LIRGs (where the PDR emission is compara-ble to the di ff use dust emission in the 5-30 µ m wavelength range,with a clear e ff ect on the shape of the SED in this region). Theshape of the SED in this wavelength range can also be a ff ected bythe presence of an AGN (purple curve), though it is only a fewgalaxies ( ∼ f rac AGN > + µ m) being the highestfor ETGs, slightly lower (by 0.23 dex) for LTGs and much lower(by 1.95 dex) for (U)LIRGs (compared to ETGs). This happens,of course, when the dust e ff ects are considered, and so stellarlight is re-processed under di ff erent levels of attenuation, de-pending on the dust content in di ff erent galaxy populations. Thispicture changes substantially when the unattenuated stellar lightis considered (see the thin dotted lines in the SEDs in the bottom-right panel in Fig. 2). The most obvious change is seen in theFUV (0.15 µ m) where the (U)LIRGs show the highest emission,followed by the LTGs (lower by 0.37 dex) and with the ETGs Article number, page 8 of 33.-D. Paspaliaris et al.: The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies showing about two orders of magnitude less emission (1.95 dex)compared to (U)LIRGs. The e ff ects of dust attenuation are betterseen in the UV wavelengths, with the ETGs showing only a verysmall change between attenuated and unattenuated curves (a dif-ference of 0.28 dex at 0.15 µ m; a combination of, both, smallamounts of dust and small fraction of young stars). These e ff ectsare more prominent for LTGs where a decent amount of dust ispresent (a change by 0.78 dex), while it is severe for (U)LIRGs (achange of 2.44 dex between attenuated and unattenuated curves)where large amounts of dust dim the stellar light. What is also in-teresting is that, in (U)LIRGs, the attenuation by the dust starts tobecome significant shortwards of ∼ µ m and so both stellar pop-ulations (old and young) are significantly a ff ected (better seenin the bottom-left panel of Fig. 2). For LTGs the dust e ff ects onthe stellar populations are seen shortwards of ∼ µ m, but sig-nificantly a ff ects only the young stars (see the top-right panelof Fig. 2), while for ETGs, the e ff ect is negligible, making itspresence obvious in wavelengths shortwards of ∼ µ m with asmall change in the emission of the young stars (top-left panelof Fig. 2). The existence of a very tight correlation between the
SFR andthe stellar mass for star-forming galaxies (often referred to as the‘main-sequence’) is well established, not only for local galaxies(see, e.g. Elbaz et al. 2007; Whitaker et al. 2012; Davies et al.2019, and references therein) but also for galaxies at high red-shifts (see, e.g. Brinchmann et al. 2004; Elbaz et al. 2007; Wuytset al. 2011b; Magdis et al. 2012; Chang et al. 2015; Pearson et al.2018, and references therein). Local galaxies of various mor-phologies (ranging from pure ellipticals to irregulars) occupydi ff erent loci in the SFR / M star plane with the most prominentones being the two distinct sequences, one for the star-forminggalaxies (with high SFR s) and one for the more relaxed systems(with low
SFR s), with a third population occupying the space inbetween (see, e.g., Davies et al. (2019), and references therein).This is presented in Fig. 3 (top panel) with the blue circles be-ing the LTGs (galaxies with T ≥ < lo g ( SFR [ M (cid:12) y r − ]) = . lo g ( M star [ M (cid:12) ]) − . , with a Spearman’s correlation coe ffi cient (Spearman 1904) of ρ = .
69, indicating a moderate positive correlation, while thesame relation for ETGs (magenta dash-dotted line) is lo g ( SFR [ M (cid:12) y r − ]) = . lo g ( M star [ M (cid:12) ]) − . , with ρ = .
44, indicating a moderate positive correlation, aswell. For comparison we overplot the relation for SDSS star-forming galaxies at 0.015 ≤ z ≤ ff ering slightly (0.77 in Elbaz et al. (2007) com-pared to 0.75 in the current study) and with an o ff set of 0.4 dexat M star = × M (cid:12) , which is to be expected, though, since thegalaxies in Elbaz et al. (2007) extend to larger redshifts. Theevolution with redshift is well studied by several works (e. g.Whitaker et al. 2014; Johnston et al. 2015; Schreiber et al. 2015).Completing the picture in the local Universe, we include thenearest (U)LIRGs (yellow stars). What is immediately evident is that these systems are amongst the most massive (with stel-lar masses above ∼ M (cid:12) ) and the most actively star-formingones, with a clear threshold in SFR above ∼
10 M (cid:12) y r − , cov-ering the high end of the main-sequence. A linear regression tothis population gives lo g ( SFR [ M (cid:12) y r − ]) = . lo g ( M star [ M (cid:12) ]) + . M star and SFR (slope of 0.02)indicating no evident correlation. This is also supported by thevalue of the Spearman’s correlation coe ffi cient being very closeto zero ( ρ = − .
1) indicating a very weak correlation. Such a be-haviour is to be expected since the parameter space for this popu-lation of galaxies is somehow limited with their values boundedby their high-end M star and SFR extremes. Similar results arealso presented in previous studies analysing samples of local lu-minous galaxies, e.g., da Cunha et al. (2010a), and Kilerci Eseret al. (2014)A similar behaviour is seen, even more clearly, by compar-ing the
SFR with the dust mass (middle panel in Fig. 3). Here,the di ff erent symbols are as in the top panel. It is obvious that(U)LIRGs are a separate population of local galaxies with boththe SFR (as mentioned above) and the dust mass occupying thehigh ends of these quantities (with M dust being above ∼ M (cid:12) ).A linear regression to the DustPedia LTGs in the M dust SFR planegives lo g ( SFR [ M (cid:12) y r − ]) = . lo g ( M dust [ M (cid:12) ]) − . , (cyan dotted line) with a Spearman’s correlation coe ffi cient of ρ = .
8, indicating a strong correlation, while this relation forthe ETGs becomes lo g ( SFR [ M (cid:12) y r − ]) = . lo g ( M dust [ M (cid:12) ]) − . , (magenta dash-dotted line) with ρ = .
67 indicating a moder-ate correlation. Our findings are in fair agreement with the re-lation found in da Cunha et al. (2010b) studying a sample ofstar-forming galaxies from the SDSS at z ≤ ff erent regime ofthe SFR - M dust space, which is also indicated by the followingbest-fit relation, lo g ( SFR [ M (cid:12) y r − ]) = . lo g ( M dust [ M (cid:12) ]) − . ffi cient being very low ( ρ = .
22) indicat-ing a weak monotonic increase of
SFR with M dust .Although local (U)LIRGs exhibit increased SFR with respectto the normal star-forming galaxies (by more than an order of amagnitude; see the transition from blue circles to yellow starsin the top and middle panels of Fig. 3) this is not the case forthe ratio of dust luminosity to the bolometric luminosity ( f abs = L dust / L bolo ). This quantity (discussed in detail in Bianchi et al.2018) indicates the significance of the dust in galaxies and thee ff ectiveness of the dust grains in absorbing the stellar radiation(a combination of the total amount of dust, the geometry and thestrength of the ISRF), which, in turn, is re-emitted at FIR / submmwavelengths. This is shown in the bottom panel of Fig. 3 with f abs plotted against the bolometric luminosity ( L bolo ). Here, the Article number, page 9 of 33 & A proofs: manuscript no. main log M star [M ] l o g S F R [ M y r ] Elbaz et al. (2007)fit for ETGsfit for LTGsfit for (U)LIRGs log M dust [M ] l o g S F R [ M y r ] da Cunha et al. (2010b) log L bolo [L ] l o g f a b s [ % ] fit for T 2.5; Bianchi et al. (2018) Fig. 3.
The correlations of
SFR with stellar mass and dust mass (up-per and middle panels respectively), and between f abs and L bolo (bottompanel; see the text for the definition of the parameters). The DustPediaETGs and LTGs are shown with red and blue circles, respectively, whileyellow stars are the local (U)LIRGs in our sample. All values are plot-ted along with their corresponding uncertainties. In all plots, magentadash-dotted, cyan dotted, and yellow dashed lines are the linear fits tothe ETGs, LTGs, and (U)LIRGs respectively. The black solid lines cor-respond to the best fits found in Elbaz et al. (2007) to the 0.015 ≤ z ≤ ≥ di ff erent symbols are as in the top panel. A linear regression tothe data gives lo g ( f abs [%]) = . lo g ( L bolo [ L (cid:12) ]) − . , for LTGs (cyan dotted line) and, lo g ( f abs [%]) = − . lo g ( L bolo [ L (cid:12) ]) + . , for ETGs (magenta dash-dotted line) with Spearman’s corre-lation coe ffi cients of 0.62 and -0.07, respectively, indicating amoderate correlation of f abs increasing with L bolo for LTGs, buta very weak decrease for ETGs. The black solid line is the best-fit linear relation for LTGs with T ≥ ≤ T < f abs values Fig. 4.
Physical properties, as derived by CIGALE, for the di ff erentgalaxy types. From top to the bottom M dust , M star , sSFR , T dust and L PDR are plotted as a function of galaxy type. Each red and blue dot corre-sponds to an individual ETG and LTG galaxy, while yellow dots corre-spond to (U)LIRGs respectively. Black diamonds stand for the medianvalues per galaxy type, while the associated 16 th and 84 th percentileranges are indicated with error-bars. Side plots of the distributions ofeach galaxy type, for all the physical properties are also presented, fol-lowing the same colouring as for the dots. In the bottom panel ( L PDR )the median value of ETGs is dragged down to lower values becauseCIGALE predicts no PDR contribution for some sources (see the textfor more details). show a smooth transition from the respective values of the star-forming galaxies (see the di ff erences among the di ff erent stylesof data-points in the bottom panel of Fig. 3), reaching a plateaufor large values of L bolo . This plateau is expected since the f abs cannot exceed 100% and is also indicated by the slope of thebest-fit to the data (yellow dashed line), lo g ( f abs [%]) = . lo g ( L bolo [ L (cid:12) ]) + . . but also from the Spearman’s coe ffi cient of the data (0.46) in-dicating a moderate correlation of f abs increasing with L bolo . Amore detailed comparison of the values of f abs between the dif-ferent populations of local galaxies will follow (Sect. 6).In Fig. 4 we have plotted several physical properties ofthe three samples of galaxies (ETGs, LTGs, and (U)LIRGs) sothat we can visualise and quantify the di ff erences. In all cases,red, blue, and yellow symbols correspond to ETGs, LTGs, and(U)LIRGs respectively. The right-hand side plots in each panelshow the histograms of the relevant quantity for the three galaxypopulations. In addition, in Appendix D, we have plotted the rel-evant cumulative distributions of the parameters for each galaxypopulation (Fig. D.1) and report the p-values (Tab. D.1) of theKolmogorov-Smirnov (KS hereafter) tests Smirnov (1948). Article number, page 10 of 33.-D. Paspaliaris et al.: The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies
In the top panel we show the variation in dust mass. As al-ready stated earlier in Sect. 4.1, the shape of the SEDs alreadyindicate an increase in the dust mass, with ETGs being the mostdevoid of dust, followed by the LTGs and with a sharp rise of thedust mass for (U)LIRGs. The median values of the three popula-tions are 4 . × M (cid:12) , 5 × M (cid:12) , and 5.2 × M (cid:12) for ETGs,LTGs, and (U)LIRGs respectively. As can also be seen from thehistograms on the side plot, but also from the cumulative dis-tributions in Fig. D.1, M dust shows a very di ff erent distributionamong the three populations occupying di ff erent ranges of theparameter.The second panel, from the top, shows the stellar mass vari-ation among the three populations. It turns out that, althoughthere is a large overlap in masses between ETGs and LTGs,there is a systematic trend with LTGs being the less massiveones and with ETGs and (U)LIRGs the most massive ones. Me-dian values of the stellar mass are 1 . × M (cid:12) , 4 . × M (cid:12) and 6.3 × M (cid:12) for ETGs, LTGs, and (U)LIRGs respectively.The historgams on the side plot, but also from the cumulativedistributions in Fig. D.1, show that M star shows di ff erent dis-tribution among the three populations (especially between the(U)LIRGs and the other two). In interacting galaxies, like mostof the (U)LIRGs in our sample, high values of stellar mass areexpected. This happens not only because of the aggregation ofthe stellar mass from the individual galaxies that undergo merg-ing, but also due to the intense star-formation of these systemslasting hundreds to thousands of Myrs. This leads to the for-mation of stars of tens to hundreds of solar masses per year, asalso predicted by numerical and hydrodynamical simulations (DiMatteo et al. 2008; Wuyts et al. 2009; Hopkins et al. 2013). Com-paring the stellar and dust masses, we find that the dust-to-stellarmass ratio is the lowest in ETGs (a median value of 2 . × − ),while it is very similar in LTGs and (U)LIRGs (median valuesof 8 . × − and 7.5 × − , respectively). This further supportsthe argument that the enhanced star-formation activity in local(U)LIRGs is mainly driven by external processes (interactions)rather than their intrinsic properties.The specific star-formation rate ( sSFR ) is defined as the cur-rent SFR over the stellar mass of the galaxy ( SFR / M star ). A usefulinterpretation of this quantity is to think of M star as the cumula-tive result of the star-formation that occurred in the past. Underthis assumption, sSFR is then a measure of how intensively thegalaxy forms stars now compared to how it used to form stars inthe past. In our analysis we have calculated the sSFR for the threepopulations and presented in the third, from the top, panel. Weclearly see that the three kinds of systems show a very di ff erentbehaviour with ETGs having very low values of sSFR (a medianvalue of 7 × − Gyr − ), followed by the LTGs (with a medianvalue of 0.1 Gyr − ), with (U)LIRGs having very high values (amedian value of 1 Gyr − ). Three distinct distributions are alsoseen in the side histograms but also from the quite deviant cumu-lative distributions in Fig. D.1 This behaviour is mainly drivenby the variation in SFR among the three populations (on aver-age 0.3, 1.2, 81 M (cid:12) yr − for ETGs, LTGs, and (U)LIRGs, re-spectively) but also from the variation in stellar mass discussedabove. This already shows how actively local (U)LIRGs are cur-rently forming stars compared to earlier stages in their lives com-pared to ETGs which show a sSFR of more than three orders ofmagnitude less.As already stated earlier, Sect. 4.1, the shape of the dustemission SED indicates a notable dust temperature variationamong the three di ff erent galaxy populations. Following a more quantitative approach we approximate the dust temperature by: T CIGALEdust = T U (1 / (4 + β ))min , (1)(Aniano et al. 2012; Nersesian et al. 2019) with U min being theminimum level of the interstellar radiation field that is able toheat the dust (in our analysis calculated by CIGALE for eachgalaxy), T the measured, in the solar neighbourhood, dust tem-perature (18.3 K), and β the dust emissivity index [with a valueof 1.79 accounting for the THEMIS dust grain model (Joneset al. 2017), assumed here]. The dust temperature for eachgalaxy, for the three populations, is plotted in the fourth, fromthe top, panel in Fig. 4, with the median values, for each galaxypopulation, indicated by black squares. These median valuesare 28, 22, and 32 K for ETGs, LTGs, and (U)LIRGs respec-tively, confirming the earlier findings (Sect. 4.1) that ETGs and(U)LIRGs can heat the dust into higher temperatures, comparedto LTGs where dust is cooler. It is interesting to notice, though,that the distributions of the dust temperature for ETGs and LTGsare quite similar. This can be confirmed by the histograms onthe side plot in Fig. 4 but also by the cumulative distributionin Fig. D.1. Although the dust in ETGs is, on average, hotterthan in LTGs (median values of 28 and 22 K respectively) theirrange and distributions are very similar, something that may indi-cate a similar mechanism of dust heating. The distribution of thedust temperature in (U)LIRGs, though, is very di ff erent from theother two. Although ETGs and (U)LIRGs may heat up the dustgrains into similar temperature levels (at least comparing theirmedian values), the source of heating is quite di ff erent, with theETGs using, mostly, their intense NIR radiation field of the oldstars while (U)LIRGs use the intense UV radiation field of theyoung stars to heat up the dust. In LTGs the radiation field ismilder heating the di ff use dust into cooler temperatures.As already indicated by the di ff erent SEDs (Fig. 2) the PDRpart of the dust emission becomes an important contributor tothe energy output in (U)LIRGs as compared to LTGs and ETGs.This is what we show in the bottom panel in Fig. 4 with the PDRluminosity for each galaxy, as predicted by CIGALE, plotted forthe three di ff erent galaxy populations. As in the panels above,the black squares indicate the median values in each galaxy pop-ulation. We have to notice here, though, that in the case of ETGs,CIGALE predicts no PDR contribution for 129 sources (out of268), dragging the median value to lower values. Our analysissuggests a median PDR luminosity of 3 . × L (cid:12) , 1 . × L (cid:12) , and 3.5 × L (cid:12) for ETGs, LTGs, and (U)LIRGs respec-tively, which, as expected, shows the lack of the PDR emissionin ETGs and the strength of this component in (U)LIRGs. Thehistorgams on the side plot, but also from the cumulative distri-butions in Fig. D.1, show that L PDR shows di ff erent distributionamong the three populations (especially between the (U)LIRGsand the other two). The importance of this component is mademore clear when comparing the PDR luminosities ( L PDR ) withthe total dust luminosities ( L PDR + L di ff use ). Our analysis suggeststhat the median values of the ratios of the PDR-to-total dust lu-minosity increases from 1.6% for ETGs, to 5.2% for LTGs, to11.7% for (U)LIRGs.As also presented in Appendix D almost all the KS p-valuesare ≤ Article number, page 11 of 33 & A proofs: manuscript no. main common characteristics between these two galaxy populations.This comes in contrast to the behaviour of local (U)LIRGs whichshow very low p-values ( < .
15) when comparing their dust tem-perature with those of the other two populations, probably indi-cating that the mechanism of merging events (which is evidentin most of these systems) may result in a di ff erent e ffi ciency ofthe way that the dust grains are heated.
5. Evolution of the physical properties of local(U)LIRGs with merging stage
Several simulation studies model the interactions between galax-ies and predict the way that
SFR evolves with time (the SFH). InSpringel et al. (2005) the authors perform numerical simulationsto model the feedback from stars and black holes in galaxiesconsisting of gas and stellar disks, central bulges and surroundedby dark matter halos, using a Salpeter type initial mass function(IMF). The derived SFH (see their Fig. 14) predicts that at thepre-merging stage (0.71 Gyr) the
SFR is ∼
10 M (cid:12) yr − but dur-ing coalescence the SFR may raise up to 50 M (cid:12) yr − followed bya decrease. Cox et al. (2006) investigate the influence of severalfeedback parameters and models on the SFR , for Sbc-like gas-rich galaxies. They find a pre-merging
SFR of ∼
30 M (cid:12) yr − ,peaking at ∼
75 M (cid:12) yr − during the coalescence and then de-creases, at the post-merging era, to ∼
10 M (cid:12) yr − . Di Matteoet al. (2008) use data-sets of numerical simulations aiming atstudying starburst episodes by galaxy mergers. They use co-planar Sa and Sb galaxies in direct and retrograde encounters,and with varying parameters, like gas fraction, galaxy relativevelocity and minimum separation. For di ff erent cases the rela-tive SFR during the coalescence is at least 2 times higher thanduring the pre- and the post-merging periods (see their Fig. 4).In a more recent study, Hopkins et al. (2013), a realistic hy-drodynamic simulation investigates the star-formation in galaxymergers with Milky Way-like galaxies at z =
0. The authors find
SFR ∼ (cid:12) yr − before the main merging event peaking at ∼
50 M (cid:12) yr − and finally, dropping to ∼ (cid:12) yr − after coa-lescence. With an infrared luminosity of 6.2 × L (cid:12) the Anten-nae Galaxies (Arp 244) is the nearest IR-bright and perhaps theyoungest prototypical galaxy-galaxy merger system (Gao et al.2001). Even if, strictly speaking, Arp 244 is not a LIRG (unlessa Virgocentric flow distance of 29.5 Mpc is considered, leadingto L IR = . × L (cid:12) ; Gao et al. 2001) it is an interesting systemthat its merging sequence has been studied and modelled in de-tail (using a hydrodynamical simulation; Renaud et al. 2015). Inthis simulation two main bursts of star-formation are predicted,the first starburst occuring at ∼
20 Myr after the first pericen-tric passage while the second episode of star-formation takingplace at ∼
170 Myr after the first pericentric passage. During thefirst starburst the SFR can reach up to ∼
80 M (cid:12) yr − which getshigher, up to ∼
110 M (cid:12) yr − , during the second starburst event.The simulations mentioned above may be derived for spe-cific parameter spaces and specific systems, yet, they are indica-tive of what to expect during a merging event. In our analysis itis not possible to trace the history of the recent star-formation foreach galaxy system but we do see a ‘snapshot’, along this time-line, for each system, depending on how far the interaction hasevolved. The rate that stars are formed during merging events de-pends on the exact content of the interacting galaxies in gas, starsand dust (Larson & Tinsley 1978; Kennicutt et al. 1996; Struck2006), as well as the morphological types (Tinsley 1968), andthe relative inclinations and velocities (Di Matteo et al. 2008) ofthe merging galaxies. The grouping that we have adopted for theseven merging stages (see Sect. 2 and Table 1) allows us to have a broad representation of di ff erent stages along the merging se-quence (from single galaxies, to galaxies in long-separated dis-tances and galaxies in coalescence). We take advantage of thisgrouping and investigate possible di ff erences in the SEDs andphysical properties of the systems in our sample. We caution,however, the reader again, as we have already done in Sect. 2,that some classes (e.g. ‘m’ and ’M5’) are underrepresented andthe robustness on the derived quantities may be questionable.The di ff erent SEDs of the seven merging classes are pre-sented in Fig. 5 with classes ‘s’, ‘m’, and ‘M1’ on the top-rowpanels, classes ‘M2’, ‘M3’, and ‘M4’ on the middle row-panels,and class ‘M5’ on the bottom-left panel. For each sub-class thedi ff erent components are also indicated in the same manner as inFig. 2. The bottom-right panel, on the other hand, compares thetotal SEDs of the di ff erent merging classes. In order to avoid con-fusion in the plot we choose to plot three of the merging classesonly, namely, ‘s’, ‘M1’ and ‘M3’ which give a broader descrip-tion of totally isolated galaxies, galaxies in the first stages of themerging, and galaxies in advanced merging, respectively. Thedi ff erent merging classes are shown with di ff erent colours as in-dicated in the inset of this panel. It is true that the di ff erencesare very small but su ffi cient to explain the variations seen in thephysical parameters (see Figs. 6, 7, and 8 and the discussion re-lated to these figures). Comparing the SEDs in the bottom-rightpanel, one can spot two evident di ff erences, a variation in theFUV to the MIR ( ∼ µ m) wavelength range, and a small shiftof the peak of the FIR emission and the Rayleigh-Jeans tail ofthe dust emission. These di ff erences already indicate di ff erencesin stellar masses, dust masses, dust temperatures and SFR whichare discussed, in detail, below where the variation in the di ff erentparameters with merging stage is investigated.The change in SFR with the merging class is visualised inFig. 6. Here, the yellow-coloured circles are the values for indi-vidual sources in each sub-class with the median value indicatedas black squares. The error-bars bracket the 16 th and 84 th per-centiles from the median. The green line connects the medianvalues and indicates the general trend. The side-plots show thehistograms of the SFR for the three merging classes, ‘s’, ‘M1’,and ‘M3’. We see that, although there is a large scatter in eachsub-class among the di ff erent sources, there is a clear trend withthe maximum median SFR occurring at sources of class ‘M4’(99 M (cid:12) yr − ), followed by class ‘M3’ (93 M (cid:12) yr − ), with thelowest SFR at class ‘s’ (26 M (cid:12) yr − ) and class ‘M2’ (51 M (cid:12) yr − ) with the rest of the classes obtaining intermediate medianvalues (66 M (cid:12) yr − , 54 M (cid:12) yr − , and 71 M (cid:12) yr − , for classes‘m’, ‘M1’, and ‘M5’ respectively). This general behaviour is tobe expected since galaxies in ‘s’ and ‘M2’ classes are more re-laxed systems [either separated (class ‘M2’) or totally isolated(class ‘s’)] with their SFR being driven mainly by internal pro-cesses or by past minor merging events which are not as pow-erful mechanisms as the tidal disruptions that take place duringmajor merger interactions. It should be noted, however, that if thenuclear
SFR is considered, as opposed to the global
SFR treatedhere, a more obvious, increasing, trend with merging stage isevident (U et al. 2019). The change in
SFR is also evident inthe individual SEDs of the di ff erent sub-classes. Looking care-fully at the median young-stellar SEDs in each plot of Fig. 5(the blue curves) we see that there is an obvious enhancementof the young stellar population for class ‘M3’ and ‘M4’ systems(2 nd and 3 rd middle panel, from the left), followed by class ‘M1’systems (top-right panel), compared to class ‘s’ and ‘M5’ sys-tems (upper-left and bottom-left panels) which show the lowestcontent in young stars. This can be appreciated either by com-paring the maximum values of the blue curves, or by comparing Article number, page 12 of 33.-D. Paspaliaris et al.: The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies N o r m a li s e d f l u x s Median model spectrumMedian stellar youngMedian stellar old Median AGNMedian diffuse dustMedian PDR dust m M1 N o r m a li s e d f l u x M2 M3 M4 Rest-frame wavelength [ m] N o r m a li s e d f l u x M5 Rest-frame wavelength [ m] s M1 M3
Fig. 5.
Median SEDs of the local (U)LIRGs, separated in classes based on their merging stage. Classes ‘s’, ‘m’ and ‘M1’ are presented in thethree top panels (from left to right), classes ‘M2’, ‘M3, and ‘M4’ in the three middle panels, while in the bottom-left panel the median SED of the‘M5’ class is plotted. The curves and the shaded areas are coloured in the same manner as in Fig. 2. The bottom-right panel shows the comparisonamong the median SEDs of ‘s’, ‘M1’ and ‘M3’ classes, allowing for the di ff erences to be spotted. the relative contribution of the young and old stellar components(comparison of the blue and red curves, more evident in the MIRwavelength range). This picture is in accordance with previousstudies. Already from the IRAS era, observations suggested thatinteractions in merging systems enhance the rate that stars areformed (Soifer et al. 1987; Kennicutt et al. 1987). In Sanders& Mirabel (1996) the authors find a clear maximum in the in-frared luminosity produced by LIRGs in the stage where theirnuclei merge. Similarly, Haan et al. (2011) report that LIRGs inlate merging stages posses total IR luminosities larger by a fac-tor of two than pre- or non-merging LIRGs. The median SFR per merging class, extracted from our SED modelling, follows asimilar trend as the simulations suggest. Despite the large scatterin the
SFR all the simulations seem to suggest an enhancement in
SFR close to coalescence (our merging classes ‘M3’ and ‘M4’)with lower
SFRs in the other stages of the interaction where theparent galaxies are either isolated ‘s’ or apart from each other(‘M1’ and ‘M2’) or, even, systems that have been evolved to isolated galaxies (class ‘M5’). This trend is also evident whencomparing the p-values of the KS tests for all combinations ofthe merging types (see Table D.1). If we do not consider typesof ‘M5’, we see that ‘s’ types show very di ff erent distributionsfrom all the rest giving low p-values in SFR , while for the restof the combinations give low p-values with ‘M3’ and ‘M4’ typesindicating very di ff erent distributions. The combination of ‘M3’and ‘M4’ though give a p-value of 0.84 suggesting very similardistributions (with a probability that they are coming from thesame parent population of 84%).The strength of the AGN in the systems of our sample, asderived by CIGALE, is presented, by merging class, in Fig. 7.We see in this plot that, although the values of the AGN frac-tion are, in general low ( ≤ .
2) there is a larger scatter of thevalues in ‘M2’ and ‘M3’ types with two of the three strongestAGNs ( f rac
AGN > .
6) appearing at these groups. We are awarethat the AGN fraction range is limited and probably a larger sam-ple is needed. This is also shown with the KS tests performed
Article number, page 13 of 33 & A proofs: manuscript no. main
Fig. 6.
SFR of (U)LIRGs in di ff erent merging stages, as derived fromCIGALE. Each circle corresponds to an individual source. Black dia-monds stand for the median values per merging class, while error barsindicate the range between the 16 th and 84 th percentiles from the me-dian. Side plots of the distributions of class ‘s’, ‘M1’ and ‘M3’ systems,with the corresponding colour, are also presented. where all p-values values in f rac AGN being relatively high (seeTable D.1), indicating that this parameter shows high probabil-ity that the distributions come from the same parent distributionin all combinations of merging stages. Nevertheless, our resultsagree with those presented in Petric et al. (2011), where they re-port no strong trends of the f rac
AGN with merging stage, but theyobserve an increase in the number of AGN dominated sources inthe latest stages. In the current work, we do not find any trendof f rac
AGN with merging stage and systems with strong AGN( f rac
AGN > .
2) lie in stages later than ‘M2’ (with F13197-1627being the only exception, since it is a ‘s’ system). It is also worthnoting that ‘M3’ sources exhibit lower median emission by theAGN component. This is obvious in Fig. 5, where the medianAGN template is absent in this merging class, while there isan indication (given the small number of sources) of a mild dipin the median value of f rac
AGN of the same class of objects inFig. 7. Springel et al. (2005), found that during the main mergingevent starburst and AGN co-exist in interacting systems. Theyalso claim that tidal forces are able not only to trigger a nuclearstarburst but also to fuel rapid growth of the black holes. Thus,although the interacting system is both starburst and AGN, itis likely that the AGN activity would be obscured by gas anddust that surrounds the nucleus. At later stages, when outflowsremove the dense gas layers during the final stages of the coales-cence, the remnant could be visible as an AGN. Our results, andmore specifically the absence of the AGN component in class‘M3’, may indicate a similar scenario.Apart from the current
SFR and the AGN fraction it is in-teresting to investigate how other physical properties vary forgalaxies of di ff erent merging stages. We do that in Fig. 8 wherethe change in M dust , M star , sSFR , T dust , and L PDR with mergingclass is plotted (top to bottom). On the side-plots the histogramsof the parameters for three merging classes, ‘s’, ‘M1’, and ‘M3’,are also presented. The symbols are as in Fig. 6. Looking at themeasurements of the individual systems we see that there is alarge scatter. The median values, though, suggest a few trendswith merging stage which worth investigating further.The median values of M dust suggest that this parameter re-mains practically unchanged with merging stage, with medianvalues of 5.0 × M (cid:12) , 5.1 × M (cid:12) , 7.7 × M (cid:12) , 6.6 × M (cid:12) ,4.9 × M (cid:12) , 7.5 × M (cid:12) and 3.3 × M (cid:12) for merging classes‘s’, ‘m’ and ‘M1’ to ‘M5’ respectively. From the side, histogramplots, but also from the cumulative distributions in Appendix Done can see that all the distributions of M dust are very similar andwell within the scatter of each individual group. This can alsobe confirmed by the respective p-values of the KS tests with allof them being large, for all combinations of merging stages (see Fig. 7.
AGN fraction of (U)LIRGs in di ff erent merging stages, as de-rived from CIGALE, along with distributions for the class ‘s’, ‘M1’ and‘M3’ systems. The style of markers is the same as in Fig. 6 Table D.1), indicating large probability that all the distributionsof the dust mass may originate from the same parent distribution. M star shows a mild change with objects with mergingstages ‘M3’ and ‘M4’ being less massive. The median val-ues are 6.4 × M (cid:12) , 8.1 × M (cid:12) , 9.3 × M (cid:12) , 1.2 × M (cid:12) ,5.3 × M (cid:12) , 5.3 × M (cid:12) and 8.1 × M (cid:12) for mergingclasses ‘s’, ‘m’ and ‘M1’ to ‘M5’. Comparison of the medianSEDs in Fig. 5 suggest that class ‘M3’ and ‘M4’ systems showa deficit in the old stellar population (evident in the NIR wave-length range) compared to the rest of the merging classes (thisdi ff erence is more clearly seen in the bottom-left panel of Fig. 5by comparing the SED of ‘M3’ with the other two SEDs). Thisis better seen (and will be discussed later) in Fig. 10 (left panel)where the histograms of the relative contribution of the old andyoung stellar populations are presented. The deficit of the oldstars in these systems (responsible for the bulk of the stellarmass) results in the slightly lower stellar mass observed. Thiskind of trend is also confirmed by comparing the p-values of theKS tests for M star in Table D.1. We see that all combinations ofmerging stages exhibit relatively large p-values providing highprobabilities that the distributions originate from the same par-ent distribution with the exception being the p-vales of ‘M2’ with‘M3’ and ‘M4’ which show low values indicating very di ff erentdistributions.The clearest, and most significant, change (compared to therest of the parameters examined here) is seen in the sSFR . Themedian values of this parameter are 0.57 Gyr − , 0.57 Gyr − ,0.58 Gyr − , 0.39 Gyr − , 2.35 Gyr − , 1.96 Gyr − and 0.78 Gyr − for merging classes ‘s’, ‘m’ and ‘M1’ to ‘M5’ respectively.The combination of the enhanced SFR (Fig. 6) and deficit instellar mass (Fig. 8) of classes ‘M3’ and ‘M4’ systems makethese merging classes di ff erentiate from the rest. Consideringthat sSFR is a measure of the current over the past SFR sug-gests that these classes of objects (undergoing, or, have gonethrough, a major merging event) show the most active currentstar-formation activity. This e ff ect is clearly seen in the derivedp-values of the KS tests with only the combinations includ-ing classes ‘M3’ and ‘M4’ showing low values indicating thattheir distributions di ff er substantially from the rest. The p-value,on the other hand, of the combination of these two mergingclasses is large (0.87) indicating high confidence that these twodistributions are similar. This is also supported by the dust-to-stellar mass ratio that is comparable in all the merging classes6.7 × − , 8.6 × − , 8.1 × − , 5.9 × − , 1.1 × − , 1.2 × − ,4.0 × − , for class ‘s’, ‘m’ and ‘M1’ to ‘M5’ objects respec-tively, indicating that the variance in the star-formation activityis closer related to the merging stage rather than the stellar anddust content of the galaxies. Article number, page 14 of 33.-D. Paspaliaris et al.: The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies
Fig. 8.
Physical properties, as derived by CIGALE, for the four di ff er-ent merging classes of (U)LIRGs. From top to the bottom M dust , M star , sSFR , T dust and L PDR are plotted as a function of merging class. Eachcircle (following the colour coding of Fig. 6) corresponds to an indi-vidual system. Black diamonds stand for the median values per galaxytype, while the associated 16 th and 84 th percentile ranges indicated witherror-bars. Distributions of class ‘s’, ‘M1’ and ‘M3’ systems, with thecorresponding colour, are also presented as side plots. A mild change in the dust temperature is also obvious withthe more relaxed, isolated, galaxies showing colder dust temper-atures. The median dust temperatures are 27.3 K, 31.6 K, 29.5 K,31.9 K, 32.7 K, 33.1 K and 35.5 K, for each merging class, from‘s’ to ‘m’ and ‘M1’ to ‘M5’, respectively. The colder dust tem-perature that is seen in ‘s’ systems can also be explained from theSEDs. As can be seen from the lower-right panel in Fig. 5, fol-lowing the merging stage evolution from ‘s’ to ‘M1’ and ‘M3’,a shift of the dust peak is obvious towards shorter wavelengthswhich translates to hotter temperatures. The p-values of the KStests for the dust temperature are generally high, indicating largeprobability that the distributions may originate from the sameparent population (Table D.1) with only a few exceptions, mainlyinvolving ‘M3’ and ‘M4’ merging classes.Finally, the PDR luminosity varies slightly for di ff er-ent merging stages with values of 1.9 × L (cid:12) , 3.9 × L (cid:12) ,2.8 × L (cid:12) , 3.5 × L (cid:12) , 4.9 × L (cid:12) , 5.3 × L (cid:12) and4.2 × L (cid:12) , for each merging class from ‘s’ to ‘M5’. A slightenhancement of L PDR is seen in merging classes ‘M3’ and ‘M4’compared to the rest. This di ff erence is more notable when com-paring the distributions of ‘M2’ class systems with ‘M3’ and‘M4’ with the p-values in the KS tests giving very small valuesindicating di ff erent populations.
40 50 60 70 80 90 100
YSP spectra [%] Y S P C I G A L E [ % ] F13428+5608
Fig. 9.
Comparison of the fraction of the young stellar population inseven systems in common between the sample studied here (YSP
CIGALE )and the sample of Rodríguez Zaurín et al. (2009). In Rodríguez Zaurínet al. (2009) the young stellar populations (YSP spectra ) have been de-rived through spectral synthesis population modelling of optical long-slit spectra (see the text for more details).
6. Old and young stellar populations in (U)LIRGsand their role in dust heating
In Nersesian et al. (2019) the authors explored the di ff erent stel-lar populations in local galaxies and their role in dust heat-ing. The SEDs of 814 galaxies of various morphological types[parametrized with their Hubble Stage (T)], ranging from pureellipticals (T = -5) to irregular galaxies (T =
10) were modelledwith CIGALE, in the same way as we do with the current sam-ple. One of the main findings of that study is that the luminosityof ETGs is dominated by the emission of the old stars with onlya small contribution (maximum of ∼
10% at T =
0) from youngstars. For later types (T = >
5. In addition to that, the role of thetwo di ff erent stellar populations (old and young) to the dust heat-ing was investigated for the various morphological types withSb (T =
3) being the most e ffi cient galaxies in the dust heating.In these galaxies, the young stars donate up to ∼
77% of theirluminosity to the dust heating while this fraction is ∼
24% forthe old stars. In what follows, we extend their analysis to local(U)LIRGs, using exactly the same methodology, and comparewith the ‘normal’ local galaxies.Although the use of SED modelling can provide us with use-ful information on the stellar populations in galaxies it is notas robust as the use of optical spectra where the imprints ofthe stellar populations can be recognised in the form of vari-ous emission lines. In Rodríguez Zaurín et al. (2009) the authorsuse long-slit spectroscopy of 36 (U)LIRGs (with z < t YSP ≤ Article number, page 15 of 33 & A proofs: manuscript no. main
Fig. 10.
Left panel: The contribution of the old (red) and young (blue) stellar populations to the unattenuated luminosity, over the bolometricluminosity, per galaxy type and (U)LIRGs merging class. Right panel: The contribution of the old and young stellar populations to the attenuatedluminosity, over the bolometric luminosity, per galaxy type and (U)LIRGs merging class (same colours as in left panel), together with the ratio ofthe dust luminosity to the bolometric luminosity (yellow). stellar populations of the seven sources in common betweenthe two samples (F08572 + + + + + spectra ) and those derived from CIGALE (YSP CIGALE ) arecompared. Since YSP spectra is extracted from several positions ineach system, the mean value is considered, while the minimumand maximum values of YSP spectra define the uncertainty in thisvalue (the error-bars in the plot). From this plot we see that, al-though the scatter is large and also the uncertainty in each sourceis large, there is a clear trend of the sources lining up (within theerrors) the one-to-one line (with the exception of F13428 + spectra of 75%, much closer to the value of 83% derived byCIGALE). This indicates that the values derived from the twomethods are comparable, given the very di ff erent approaches.Furthermore, the median values of the two samples also show tocompare well. For the 36 (U)LIRGs in the sample of RodríguezZaurín et al. (2009) the median value for the young stellar popu-lation is 74 ±
13 %, while, for the 67 sources in our sample thisfraction is 64 ±
18 %.The di ff erent stellar populations (old and young) in our sys-tems are presented, and compared with the local ‘normal’ galax-ies, in Fig. 10. In the left panel of Fig. 10 the histograms ofthe unattenuated luminosities of both stellar components to thebolometric luminosity of each galaxy ( f unattold = L unattold / L bolo , and f unattyoung = L unattyoung / L bolo , where L bolo = L unattold + L unattyoung ) are plotted.In this plot the red and blue histograms indicate the mean val-ues of f unattold and f unattyoung respectively. In the leftmost sub-panel therelative contribution of the young and old stellar populations inETGs, LTGs, and (U)LIRGs are compared, while these valuesfor each of the seven merging subclasses of the (U)LIRGs sam-ple are indicated in the rightmost sub-panel. The exact numbersare presented in Table 4. The most striking feature from this plotis the large increase in f unattyoung for (U)LIRGs compared to localETGs and LTGs. As already stated in Nersesian et al. (2019) theold stars are the prominent luminosity source in ETGs and LTGs(with mean values of f unattold of 96% and 79% respectively) while,this picture now reverses in the case of (U)LIRGs with 64% oftheir bolometric luminosity originating from young stars. Thisincrease (by a factor ∼
3) in the luminosity of the young stellarcomponent is, of course, the result of the intense star-formationthat takes place in these systems, mostly, due to merging events. The fraction of the young stars in all, seven, merging sub-classesstays around the mean value of 64% but it is class ‘M3’ and‘M4’ systems that show the highest fraction (72% and 79% re-spectively) of young stars following the trend in
SFR (Fig. 6).All the related values are given in Table 4.The right panel in Fig. 10 shows the e ff ects of dust inthe stellar populations discussed previously. Here, the ratioof the dust-attenuated luminosity of the old stellar population( f attold = L attold / L bolo ) and of the young stellar population ( f attyoung = L attyoung / L bolo ) to the bolometric luminosity is plotted with red andblue colours respectively, while the fraction of the dust-absorbedluminosity ( f abs ; see also Sect. 4.2) is indicated with yellowcolour. The leftmost sub-panel shows the comparison of thesequantities for ETGs, LTGs, and (U)LIRGs, while the rightmostsub-panel shows the comparison among the di ff erent mergingsub-classes. What is evident from this plot is the large e ff ect thatdust has on the energy budget of (U)LIRGs compared to ETGsand LTGs. f abs changes from very low (7%) for ETGs, to moder-ate (25%) for LTGs, to very high (78%) for (U)LIRGs. This canbe explained, mainly, by the higher dust mass that is detected in(U)LIRGs (upper panel in Fig. 4) and, especially, the dust asso-ciated with the PDR regions (bottom panel in Fig. 4). The dra-matic e ff ect of the dust on the stellar populations is clearly seenby comparing the leftmost sub-panels of each panel in Fig. 10.For (U)LIRGs we see that the fraction of the young stars is ab-sorbed so heavily that it goes from 64% in the unattenuated case,to 5% in the case where absorption by dust is considered. It isalso worth mentioning that the attenuated fraction of the lumi-nosity of the young stars is higher for LTGs (9%) compared to2% and 5% in ETGs and (U)LIRGs. Classes ‘M3’ and ‘M4’ sys-tems show the highest mean f abs values (84% and 87%, respec-tively), with classes ‘s’ and ‘M2’ having the lowest mean f abs values (73%). All the related values are given in Table 4.As was already discussed above, a large fraction of the en-ergy emitted by the stars, in (U)LIRGs, is absorbed by the dustgrains (see the right panel in Fig. 10) resulting in their heatingand the production of large amounts of IR radiation in those sys-tems. With our analysis we can not only quantify the total stellarradiation that is absorbed by the dust, but we can also distinguishbetween the two stellar populations and calculate their e ffi ciencyin heating up the dust grains. The quantity that shows the fractionof each stellar population that is absorbed, by the dust, is the ratioof the absorbed luminosity, of each stellar component, to the re-spective unattenuated stellar component ( F absold = L absold / L unattold and Article number, page 16 of 33.-D. Paspaliaris et al.: The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies E T G s L T G s ( U ) L I R G s s m M M M M M L u m i n o s i t y o v e r un a tt e nu a t e d l u m i n o s i t y F absold F attold E T G s L T G s ( U ) L I R G s s m M M M M M F absyoung F attyoung Fig. 11.
Mean values of the fraction of the luminosity of the old and theyoung stellar populations (left and right panel, respectively) used for thedust heating. The crossed bars show the mean values of the ratio of thedust-absorbed, luminosity, to the unattenuated luminosity of the corre-sponding stellar component, while the solid bars are the ratios of theattenuated luminosity of the specific stellar component to its unattenu-ated luminosity. In each panel the ratios are presented for each of thethree galaxy populations (ETGs, LTGs, and (U)LIRGs) in the leftmostsub-panel, and for the four merging classes in the rightmost sub-panel. F absyoung = L absyoung / L unattyoung , for the old and the young stellar compo-nent respectively). The remaining luminosity (not absorbed bythe dust) is the attenuated luminosity of each stellar component,which, divided by the unattenuated luminosity gives the fractionof the, una ff ected, by the dust, luminosity ( F attold = L attold / L unattold and F attyoung = L attyoung / L unattyoung , for the old and the young stellar compo-nent respectively). For each stellar population, the crossed barsin Fig. 11, show the mean values of the absorbed fraction ofthe luminosity which contributes to the dust heating ( F absold, young ),with the rest, solid bars, being the fraction of the luminosityemitted by stars, without being a ff ected by the dust ( F attold, young ).The left panel shows the di ff erent contributions of the old stars(red colour) and the right panel those of the young stars (bluecolour), for each galaxy population (ETGs, LTGs, and (U)LIRGsin the leftmost sub-panels) and for each merging class (rightmostsub-panels). Looking at the di ff erent galaxy types, it is evidentthat it is the young stellar component that o ff ers the larger por-tion of its total luminosity in the dust heating compared to the oldone. In particular, for ETGs, it is 30% of the young stellar lumi-nosity donated to the dust heating compared to only 6% which isthe case for the old stars, while in LTGs these fractions are 58%and 17%, and get extremely high in (U)LIRGs (92% and 56%for the young and the old stellar populations, respectively). It isnoteworthy that even the old stars in (U)LIRGs have a signifi-cant role in the dust heating contributing with more than half oftheir luminosities. The mean values of the fractions of the dust-absorbed luminosity for each of the two stellar components, havecomparable values for the four merging classes with no signifi-cant deviations (see the rightmost sub-panels in Fig. 11). All therelated values are summarised in Table 4.It is also interesting to investigate the relative contributionof the stellar populations to the dust heating. The parameter thatindicates this contribution is the ratio of the dust-absorbed lu-minosity, for each stellar population, to the total dust luminos-ity ( S absold = L attold / L dust and S absyoung = L attyoung / L dust for the old andthe young stars respectively). The histograms of the mean val-ues of this parameter are presented in Fig. 12 with red and bluecolours representing the contribution of the old and the young ETGs LTGs (U)LIRGs s m M1 M2 M3 M4 M50.00.20.40.60.81.0 L u m i n o s i t y o v e r d u s t l u m i n o s i t y S absyoung S absold Fig. 12.
Mean values of the ratios of the dust-absorbed stellar luminosity(originating from old and young stars) to the dust luminosity (red andblue colours, respectively). The ratios are presented for each of the threegalaxy populations (ETGs, LTGs, and (U)LIRGs) in the leftmost sub-panel, and for the four merging classes in the rightmost sub-panel. stellar components respectively. In this plot, these contributionsin ETGs, LTGs, and (U)LIRGs are plotted in the leftmost part,while the rightmost part shows the relative contribution for thefour merging classes. For ETGs, as already described in Ners-esian et al. (2019), it is mainly the old stars that contribute moreto the dust heating (by 86%) while, in LTGs, both the old and theyoung stars contribute almost equally to the dust heating (52%for the old stellar population). In (U)LIRGs the picture reverseswith the young stars taking over the heating of the dust grainswith 74% of the luminosity of this stellar population absorbedby dust.It is also interesting to investigate the relative contributionof the stellar populations to the dust heating. The parameter thatindicates this contribution is the ratio of the dust-absorbed lu-minosity, for each stellar population, to the total dust luminos-ity ( S absold = L attold / L dust and S absyoung = L attyoung / L dust for the old andthe young stars respectively). The histograms of the mean val-ues of this parameter are presented in Fig. 12 with red and bluecolours representing the contribution of the old and the youngstellar components respectively. In this plot, these contributionsin ETGs, LTGs, and (U)LIRGs are plotted in the leftmost part,while the rightmost part shows the relative contribution for thefour merging classes. For ETGs, as already described in Ners-esian et al. (2019), it is mainly the old stars that contribute moreto the dust heating (by 86%) while, in LTGs, both the old and theyoung stars contribute almost equally to the dust heating (52%for the old stellar population). In (U)LIRGs the picture reverseswith the young stars taking over the heating of the dust grainswith 74% of the luminosity of this stellar population absorbedby dust.Concerning the relative contribution of the two stellar popu-lations to the dust heating of the di ff erent merging classes we seethat it remains close to the mean value of (U)LIRGs (74% for theyoung stars) with only small deviations. The largest deviationsfor the contribution of the young stars to the dust heating arefound between classes ‘s’ and ‘m’ (65% and 67%) and classes‘M3’ and ‘M4’ (80% and 85%) respectively. All the related val-ues are summarised in Table 4. Article number, page 17 of 33 & A proofs: manuscript no. main
Table 4.
Mean values of the ratios of various combinations of the stellar and dust luminosity components extracted using the CIGALE SEDfitting tool. The di ff erent ratios (also presented in Nersesian et al. 2019) are defined as f unattold = L unattold / L bolo , f unattyoung = L unattyoung / L bolo , f attold = L attold / L bolo , f attyoung = L attyoung / L bolo , f abs = L dust / L bolo , F attold = L attold / L unattold , F absold = L absold / L unattold , F attyoung = L attyoung / L unattyoung , F absyoung = L absyoung / L unattyoung , S absold = L attold / L dust , and S absyoung = L attyoung / L dust , where, L unattold and L unattyoung the unattenuated luminosities of the old and the young stars, L bolo is the bolometric luminosity of eachsystem ( L bolo = L unattold + L unattyoung ), L dust the dust luminosity, L attold and L attyoung the attenuated luminosity of the old and the young stars, L absold and L absyoung theluminosity of the old and the young stars absorbed by dust. These ratios are presented for the three general galaxy populations compared in thisstudy (ETGs, LTGs, and (U)LIRGs), as well as for the four merging classes of (U)LIRGs. Galaxy Type f unattold f unattyoung f attold f attyoung f abs F attold F absold F attyoung F absyoung S absold S absyoung ETGs 0.96 0.04 0.91 0.02 0.07 0.94 0.06 0.70 0.30 0.86 0.14LTGs 0.79 0.21 0.66 0.09 0.25 0.83 0.17 0.42 0.58 0.52 0.48(U)LIRGs 0.36 0.64 0.17 0.05 0.78 0.44 0.56 0.08 0.92 0.26 0.74s 0.46 0.54 0.23 0.04 0.73 0.47 0.53 0.08 0.92 0.33 0.67m 0.45 0.55 0.17 0.04 0.79 0.39 0.61 0.07 0.93 0.35 0.65M1 0.40 0.60 0.21 0.04 0.75 0.52 0.48 0.07 0.93 0.27 0.73M2 0.44 0.56 0.22 0.05 0.73 0.49 0.51 0.09 0.91 0.31 0.69M3 0.28 0.72 0.11 0.05 0.84 0.40 0.60 0.07 0.93 0.20 0.80M4 0.21 0.79 0.08 0.05 0.87 0.39 0.61 0.07 0.93 0.15 0.85M5 0.43 0.57 0.14 0.04 0.82 0.33 0.67 0.06 0.94 0.35 0.65
7. Summary
In this study we model the SEDs of 67 local (U)LIRGs, usingthe CIGALE code, to calculate their physical properties. Theirstellar mass, dust mass, star-formation rate, dust temperatureas well as their luminosity arising from PDR regions are de-rived and compared to those of 268 ETGs and 542 LTGs (al-ready derived in a similar way in Nersesian et al. 2019). Fur-thermore, the (U)LIRGs are categorised in seven classes accord-ing to their merging stage (ranging from totally isolated to pre-mergers, mergers, and post-mergers), providing useful informa-tion on how their properties depend on the phase of the interac-tion of the parent galaxies. Finally, the contribution of the twostellar populations (old and young) to the bolometric luminosityof these systems and their role in the dust heating is also ex-plored. Our findings are summarised as follows: – (U)LIRGs occupy the ‘high-end’ on the M dust , M star and SFR plane in the local Universe compared to ETGs and LTG (withthe corresponding parameters already calculated in Ners-esian et al. 2019). Their median dust mass is 5 . × M (cid:12) compared to 4 . × M (cid:12) and 5 × M (cid:12) for ETGs and LTGs,respectively. Their median stellar mass is 6 . × M (cid:12) com-pared to 1 . × M (cid:12) , and 4 . × M (cid:12) for ETGs and LTGs,respectively. SFR in (U)LIRGs gets a much higher medianvalue of 52 . (cid:12) yr − compared to 0.01, and 0 . (cid:12) yr − forETGs and LTGs, respectively. The median values of the spe-cific star-formation rate ( sSFR ), on the other hand, rangesfrom 7 × − and 0.1 Gyr − , for ETGs and LTGs, to 1.0Gyr − for (U)LIRGs, respectively. These di ff erences amongthe three galaxy populations in the local universe can also betraced by carefully examining their median SEDs. – The median SEDs show a slight shift (in the wavelengthaxis) of the dust emission peak indicating that dust temper-ature is cooler in LTGs and warmer in (U)LIRGs with me-dian dust temperatures of 28, 22, and 32 K for ETGs, LTGs,and (U)LIRGs respectively. The SEDs also reveal the signif-icance of the dust emission in PDR regions increasing fromETGs to LTGs to (U)LIRGs with the median values of the ra-tios of the PDR-to-total dust luminosity changing from 1.6%,to 5.2%, to 11.7% respectively. Furthermore, the attenuatione ff ects, caused by the dust, are evident in the median SEDswith the peak of the stellar emission (measured at 1 µ m) be-ing the highest for ETGs, slightly lower (by 0.23 dex) for LTGs and much lower (by 1.95 dex) for (U)LIRGs (com-pared to ETGs). Comparison of the attenuated and unattenu-ated curves of the SEDs indicates that the attenuation by thedust becomes significant shortwards of ∼ . µ m, ∼ µ m,and ∼ µ m for ETGs, LTGs, and (U)LIRGs, respectively. – Small di ff erences, in the derived parameters, are seen forthe seven merging classes of our sample of (U)LIRGs. Dustmass is very similar among di ff erent merging classes (withinthe scatter of the measurements), while a mild deficit is seenin the stellar mass for class ‘M3’ and ‘M4’ objects. The mostevident change is seen in the SFR with the median valuescomputed for class ‘M4’ objects being the highest (99 M (cid:12) yr − ) followed by class ‘M3’ (93 M (cid:12) yr − ), with the low-est SFR occurring at class ‘s’ (26 M (cid:12) yr − ) sources. A mildchange in the dust temperature is found with an increasingtrend of the median value from 27.3 K to 35.5 K from theisolated to the more evolved systems respectively. The PDRluminosity is slightly enhanced for classes ‘M3’ and ‘M4’systems, compared to the rest of the classes, consistent withthe higher SFR observed in those systems. – In contrast to the local ‘normal’ galaxies where the old starsare the dominant source of the stellar emission (with thefraction of their luminosity over the bolometric luminos-ity being 96% and 79% for ETGs and LTGs, respectively)this picture reverses in (U)LIRGs with the young stars be-ing the dominant source of stellar emission with the frac-tion of their luminosity being 64% of the bolometric one.Out of the seven merging classes, classes ‘M4’ and ‘M3’show the highest such contribution (79% and 72%, respec-tively). The e ff ects of dust in (U)LIRGs, parametrized by thedust-absorbed luminosity, to the bolometric luminosity is ex-tremely high (78%) compared to 7% and 25% in ETGs andLTGs, respectively. – The fraction of the stellar luminosity used to heat up the dustgrains is very high in (U)LIRGs, for both stellar components(92% and 56% for young and old, respectively) comparedto 30% and 6% for ETGs, and 58% and 17% for LTGs, re-spectively. In (U)LIRGs 74% of the dust heating comes fromthe young stars, with the old stars being the dominant sourceof dust heating contributing with 86% in ETGs and 52% inLTGs.
Acknowledgements.
We would like to thank the anonymous referee for pro-viding comments and suggestions, which helped to improve the quality of the
Article number, page 18 of 33.-D. Paspaliaris et al.: The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies manuscript. This research is co-financed by Greece and the European Union(European Social Fund-ESF) through the Operational Programme "Human Re-sources Development, Education and Lifelong Learning 2014-2020" in the con-text of the project “Anatomy of galaxies: their stellar and dust content throughcosmic time” (MIS 5052455). GM acknowledges support by the Agencia Estatalde Investigación, Unidad de Excelencia María de Maeztu, ref. MDM-2017-0765.DustPedia is a collaborative focused research project supported by the EuropeanUnion under the Seventh Framework Programme (2007-2013) call (proposal no.606847). The participating institutions are: Cardi ff University, UK; National Ob-servatory of Athens, Greece; Ghent University, Belgium; Université Paris Sud,France; National Institute for Astrophysics, Italy and CEA, France. This researchhas made use of the NASA / IPAC Extragalactic Database (NED), which is oper-ated by the Jet Propulsion Laboratory, California Institute of Technology, undercontract with the National Aeronautics and Space Administration. We have alsomade use of the VizieR catalogue access tool, CDS, Strasbourg, France (DOI: 10.26093 / cds / vizier). The original description of the VizieR service was pub-lished in A&AS 143, 23. References
Aniano, G., Draine, B. T., Calzetti, D., et al. 2012, ApJ, 756, 138Armus, L., Heckman, T., & Miley, G. 1987, AJ, 94, 831Armus, L., Mazzarella, J. M., Evans, A. S., et al. 2009, PASP, 121, 559Baes, M., Verstappen, J., De Looze, I., et al. 2011, ApJS, 196, 22Bianchi, S., De Vis, P., Viaene, S., et al. 2018, A&A, 620, A112Bongiorno, A., Merloni, A., Brusa, M., et al. 2012, MNRAS, 427, 3103Boquien, M., Boselli, A., Buat, V., et al. 2013, A&A, 554, A14Boquien, M., Burgarella, D., Roehlly, Y., et al. 2019, A&A, 622, A103Brinchmann, J., Charlot, S., White, S. D. M., et al. 2004, MNRAS, 351, 1151Bruzual, G. & Charlot, S. 2003, MNRAS, 344, 1000Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ, 533, 682Camps, P. & Baes, M. 2015, Astronomy and Computing, 9, 20Casey, C. M. 2012, MNRAS, 425, 3094Chabrier, G. 2003, PASP, 115, 763Chang, Y.-Y., van der Wel, A., da Cunha, E., & Rix, H.-W. 2015, ApJS, 219, 8Charlot, S. & Fall, S. M. 2000, ApJ, 539, 718Chu, J. K., Sanders, D. B., Larson, K. L., et al. 2017, ApJS, 229, 25Ciesla, L., Boselli, A., Elbaz, D., et al. 2016, A&A, 585, A43Ciesla, L., Charmandaris, V., Georgakakis, A., et al. 2015, A&A, 576, A10Clark, C. J. R., Verstocken, S., Bianchi, S., et al. 2018, A&A, 609, A37Clements, D. L., Sutherland, W. J., McMahon, R. G., & Saunders, W. 1996,MNRAS, 279, 477Cox, T. J., Jonsson, P., Primack, J. R., & Somerville, R. S. 2006, MNRAS, 373,1013da Cunha, E., Charmandaris, V., Díaz-Santos, T., et al. 2010a, A&A, 523, A78da Cunha, E., Eminian, C., Charlot, S., & Blaizot, J. 2010b, MNRAS, 403, 1894Dale, D. A., Helou, G., Magdis, G. E., et al. 2014, ApJ, 784, 83Davies, J. I., Baes, M., Bianchi, S., et al. 2017, PASP, 129, 044102Davies, J. I., Nersesian, A., Baes, M., et al. 2019, A&A, 626, A63Di Matteo, P., Bournaud, F., Martig, M., et al. 2008, A&A, 492, 31Díaz-Santos, T., Armus, L., Charmandaris, V., et al. 2017, ApJ, 846, 32Duc, P. A., Brinks, E., Wink, J. E., & Mirabel, I. F. 1997, A&A, 326, 537Dunne, L., Eales, S., Edmunds, M., et al. 2000, MNRAS, 315, 115Dunne, L. & Eales, S. A. 2001, MNRAS, 327, 697Elbaz, D., Daddi, E., Le Borgne, D., et al. 2007, A&A, 468, 33Elvis, M., Hao, H., Civano, F., et al. 2012, ApJ, 759, 6Fritz, J., Franceschini, A., & Hatziminaoglou, E. 2006, MNRAS, 366, 767Gao, Y., Lo, K. Y., Lee, S. W., & Lee, T. H. 2001, ApJ, 548, 172Genzel, R., Lutz, D., Sturm, E., et al. 1998, ApJ, 498, 579Giovannoli, E., Buat, V., Noll, S., Burgarella, D., & Magnelli, B. 2011, A&A,525, A150Haan, S., Surace, J. A., Armus, L., et al. 2011, AJ, 141, 100Hainline, L. J., Blain, A. W., Smail, I., et al. 2011, ApJ, 740, 96Herrero-Illana, R., Privon, G. C., Evans, A. S., et al. 2019, A&A, 628, A71Hollenbach, D. J. & Tielens, A. G. G. M. 1999, Rev. Mod. Phys., 71, 173Hopkins, P. F., Cox, T. J., Hernquist, L., et al. 2013, MNRAS, 430, 1901Hutchings, J. B. & Ne ff , S. G. 1991, AJ, 101, 434Ichikawa, K., Ueda, Y., Terashima, Y., et al. 2012, ApJ, 754, 45Ishida, C. M. 2004, PhD thesis, UNIVERSITY OF HAWAI’IIwasawa, K., Sanders, D. B., Teng, S. H., et al. 2011, A&A, 529, A106Johnston, R., Vaccari, M., Jarvis, M., et al. 2015, MNRAS, 453, 2540Jones, A. P., Köhler, M., Ysard, N., Bocchio, M., & Verstraete, L. 2017, A&A,602, A46Kennicutt, R. C., J., Schweizer, F., & Barnes, J. E. 1996, Galaxies: Interactionsand Induced Star FormationKennicutt, Robert C., J., Keel, W. C., van der Hulst, J. M., Hummel, E., & Roet-tiger, K. A. 1987, AJ, 93, 1011Kilerci Eser, E., Goto, T., & Doi, Y. 2014, ApJ, 797, 54 Kim, D. C., Evans, A. S., Vavilkin, T., et al. 2013, ApJ, 768, 102Larson, K. L., Sanders, D. B., Barnes, J. E., et al. 2016, ApJ, 825, 128Larson, R. B. & Tinsley, B. M. 1978, ApJ, 219, 46Lawrence, A., Rowan-Robinson, M., Leech, K., Jones, D. H. P., & Wall, J. V.1989, MNRAS, 240, 329Leech, K. J., Rowan-Robinson, M., Lawrence, A., & Hughes, J. D. 1994, MN-RAS, 267, 253Lonsdale, C. J., Farrah, D., & Smith, H. E. 2006, Ultraluminous Infrared Galax-ies, ed. J. W. Mason, 285Lyu, J. & Rieke, G. H. 2018, ApJ, 866, 92Madau, P. & Dickinson, M. 2014, ARA&A, 52, 415Magdis, G. E., Daddi, E., Béthermin, M., et al. 2012, ApJ, 760, 6Makarov, D., Prugniel, P., Terekhova, N., Courtois, H., & Vauglin, I. 2014, A&A,570, A13Małek, K., Pollo, A., Takeuchi, T. T., et al. 2014, A&A, 562, A15Melnick, J. & Mirabel, I. F. 1990, A&A, 231, L19Morrissey, P., Conrow, T., Barlow, T. A., et al. 2007, ApJS, 173, 682Mountrichas, G., Buat, V., Yang, G., et al. 2020, arXiv e-prints,arXiv:2011.09220Nardini, E., Risaliti, G., Watabe, Y., Salvati, M., & Sani, E. 2010, MNRAS, 405,2505Nersesian, A., Xilouris, E. M., Bianchi, S., et al. 2019, A&A, 624, A80Nikutta, R., Elitzur, M., & Lacy, M. 2009, ApJ, 707, 1550Noll, S., Burgarella, D., Giovannoli, E., et al. 2009, A&A, 507, 1793Pappalardo, C., Bizzocchi, L., Fritz, J., et al. 2016, A&A, 589, A11Pearson, W. J., Wang, L., Hurley, P. D., et al. 2018, A&A, 615, A146Petric, A. O., Armus, L., Howell, J., et al. 2011, ApJ, 730, 28Pilbratt, G. L., Riedinger, J. R., Passvogel, T., et al. 2010, A&A, 518, L1Ramos P, A. F., Ashby, M. L. N., Smith, H. A., et al. 2020, MN-RAS[ arXiv:2009.05614 ]Renaud, F., Bournaud, F., & Duc, P.-A. 2015, MNRAS, 446, 2038Ricci, C., Bauer, F. E., Treister, E., et al. 2017, MNRAS, 468, 1273Rodríguez Zaurín, J., Tadhunter, C. N., & González Delgado, R. M. 2009, MN-RAS, 400, 1139Roehlly, Y., Burgarella, D., Buat, V., et al. 2014, in Astronomical Society of thePacific Conference Series, Vol. 485, Astronomical Data Analysis Softwareand Systems XXIII, ed. N. Manset & P. Forshay, 347Salim, S., Boquien, M., & Lee, J. C. 2018, ApJ, 859, 11Salpeter, E. E. 1955, ApJ, 121, 161Sanders, D. B., Mazzarella, J. M., Kim, D. C., Surace, J. A., & Soifer, B. T. 2003,AJ, 126, 1607Sanders, D. B. & Mirabel, I. F. 1996, ARA&A, 34, 749Schreiber, C., Pannella, M., Elbaz, D., et al. 2015, A&A, 575, A74Schweizer, F. & Seitzer, P. 1992, AJ, 104, 1039Skrutskie, M. F., Cutri, R. M., Stiening, R., et al. 2006, AJ, 131, 1163Smirnov, N. 1948, Ann. Math. Statist., 19, 279Soifer, B. T., Neugebauer, G., & Houck, J. R. 1987, ARA&A, 25, 187Spearman, C. 1904, The American Journal of Psychology, 15, 72Springel, V., Di Matteo, T., & Hernquist, L. 2005, MNRAS, 361, 776Stalevski, M., Asmus, D., & Tristram, K. R. W. 2017, MNRAS, 472, 3854Stalevski, M., Fritz, J., Baes, M., Nakos, T., & Popovi´c, L. ˇC. 2012, MNRAS,420, 2756Stalevski, M., Ricci, C., Ueda, Y., et al. 2016, MNRAS, 458, 2288Stalevski, M., Tristram, K. R. W., & Asmus, D. 2019, MNRAS, 484, 3334Stierwalt, S., Armus, L., Surace, J. A., et al. 2013, ApJS, 206, 1Struck, C. 2006, Galaxy Collisions — Dawn of a New Era, ed. J. W. Mason, 115Tanimoto, A., Ueda, Y., Odaka, H., et al. 2019, ApJ, 883, 110Tinsley, B. M. 1968, ApJ, 151, 547Toomre, A. & Toomre, J. 1972, ApJ, 178, 623Torres-Albà, N., Iwasawa, K., Díaz-Santos, T., et al. 2018, A&A, 620, A140U, V., Medling, A. M., Inami, H., et al. 2019, ApJ, 871, 166U, V., Sanders, D. B., Mazzarella, J. M., et al. 2012, ApJS, 203, 9Vika, M., Ciesla, L., Charmandaris, V., Xilouris, E. M., & Lebouteiller, V. 2017,A&A, 597, A51Werner, M. W., Roellig, T. L., Low, F. J., et al. 2004, ApJS, 154, 1Whitaker, K. E., Franx, M., Leja, J., et al. 2014, ApJ, 795, 104Whitaker, K. E., van Dokkum, P. G., Brammer, G., & Franx, M. 2012, ApJ, 754,L29Wuyts, S., Förster Schreiber, N. M., Lutz, D., et al. 2011a, ApJ, 738, 106Wuyts, S., Förster Schreiber, N. M., van der Wel, A., et al. 2011b, ApJ, 742, 96Wuyts, S., Franx, M., Cox, T. J., et al. 2009, ApJ, 696, 348Yang, G., Boquien, M., Buat, V., et al. 2020, MNRAS, 491, 740Zou, Z., Xia, X., Deng, Z., & Su, H. 1991, MNRAS, 252, 593 Article number, page 19 of 33 & A proofs: manuscript no. main
Appendix A: CIGALE validation (mock analysis)
In order to examine how well the derived parameters can be con-strained from the multi-wavelength SED fitting that CIGALEperforms, and to monitor the accuracy and precision expectedfor each parameter, we made use of the CIGALE module thatperforms a mock analysis. This module creates a mock SEDfor each galaxy based on the best fitted parameters, allowingthe fluxes to vary within the uncertainties of the observationsby adding a value taken from a Gaussian distribution with thesame standard deviation as indicated by the observations. Bymodelling these mock SEDs with CIGALE we can then retrievethe best set of the mock fitted parameters and compare them withthose used as an input. This provides us with a direct measure ofhow accurately one can retrieve specific parameters for a specificsample of galaxies.The results of the mock analysis are presented in Fig. A.1with the best fitted values of each parameter (input values; x-axis) compared to the mock values of the each parameter (mockvalues; y-axis). The red circles indicate the strongest AGNs inthe sample with f rac
AGN > .
2. The solid blue line correspondsto the one-to-one relation while the orange dashed line the bestlinear fit to the data. The relevant value of the Spearman’s corre-lation coe ffi cient ( ρ ) is also indicated in each panel.The parameters presented in the mock analysis, are the onesused in this work or have been used for the calculation of otherquantities (e.g., U min for the calculation of T dust ). It is evidentthat all the mock-derived values have a strong correlation withthe input parameters with the Spearman’s correlation coe ffi cientbeing more than ∼ . U min showing some deviant points, es-pecially for galaxies with strong AGNs, but, overall, the inputand mock data are in a good agreement with ρ = . Appendix B: Best-fit SED models
The SEDs of the 67 (U)LIRGs, analysed in this study, are pre-sented here. The goodness of the fit for each source is indi-cated with the reduced χ of the fit listed in Table 3. For afew exceptional cases, e.g., F12224-0624, F15107 + + χ values. This may havesome e ff ect on derived dust parameters for these systems. log M star [M fl ] (input) l o g M s t a r [ M fl ] ( m o c k ) log M dust [M fl ] (input) l o g M du s t [ M fl ] ( m o c k ) log L bol [L fl ] (input) l o g L b o l [ L fl ] ( m o c k ) log L dust [L fl ] (input) l o g L du s t [ L fl ] ( m o c k ) log SFR [M fl yr − ] (input) l o g S F R [ M fl y r − ] ( m o c k ) log U min (input) l o g U m i n ( m o c k ) Fig. A.1.
Best fitted parameters (input values; x-axis) versus the mockparameters (mock values; y-axis) as derived by the mock analysis per-formed with the CIGALE. Each circle corresponds to an individualgalaxy with the red circles indicating the strongest AGNs in the sam-ple with f rac
AGN > .
2. The solid blue line corresponds to the one-to-one relation while the orange dashed line the best linear fit to the data.The relevant value of the Spearman’s correlation coe ffi cient ( ρ ) is alsoindicated in each panel.Article number, page 20 of 33.-D. Paspaliaris et al.: The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies Observed ( m)10 S ( m J y ) F00085-1223
Stellar old unattenuatedStellar young unattenuatedNebular emissionDiffuse dust emissionPDR dust emissionAGN emissionModel spectrumModel fluxesObserved fluxes Observed ( m)10 S ( m J y ) F00402-2349 Observed ( m)10 S ( m J y ) F01076-1707 Observed ( m)10 S ( m J y ) F01364-1042 Observed ( m)10 S ( m J y ) F00163-1039 Observed ( m)10 S ( m J y ) F01053-1746 Observed ( m)10 S ( m J y ) F01173+1405 Observed ( m)10 S ( m J y ) F01417+1651
Article number, page 21 of 33 & A proofs: manuscript no. main Observed ( m)10 S ( m J y ) F01484+2220 Observed ( m)10 S ( m J y ) F02435+1253 Observed ( m)10 S ( m J y ) F03359+1523 Observed ( m)10 S ( m J y ) F04191-1855 Observed ( m)10 S ( m J y ) F02281-0309 Observed ( m)10 S ( m J y ) F02512+1446 Observed ( m)10 S ( m J y ) F04097+0525 Observed ( m)10 S ( m J y ) F04315-0840
Article number, page 22 of 33.-D. Paspaliaris et al.: The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies Observed ( m)10 S ( m J y ) F05189-2524 Observed ( m)10 S ( m J y ) F08354+2555 Observed ( m)10 S ( m J y ) F09126+4432 Observed ( m)10 S ( m J y ) F09333+4841 Observed ( m)10 S ( m J y ) F06107+7822 Observed ( m)10 S ( m J y ) F08572+3915 Observed ( m)10 S ( m J y ) F09320+6134 Observed ( m)10 S ( m J y ) F09437+0317
Article number, page 23 of 33 & A proofs: manuscript no. main Observed ( m)10 S ( m J y ) F10015-0614 Observed ( m)10 S ( m J y ) F10257-4339 Observed ( m)10 S ( m J y ) F11011+4107 Observed ( m)10 S ( m J y ) F11231+1456 Observed ( m)10 S ( m J y ) F10173+0828 Observed ( m)10 S ( m J y ) F10565+2448 Observed ( m)10 S ( m J y ) F11186-0242 Observed ( m)10 S ( m J y ) F11257+5850
Article number, page 24 of 33.-D. Paspaliaris et al.: The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies Observed ( m)10 S ( m J y ) F12112+0305 Observed ( m)10 S ( m J y ) F12540+5708 Observed ( m)10 S ( m J y ) F13001-2339 Observed ( m)10 S ( m J y ) F13136+6223 Observed ( m)10 S ( m J y ) F12224-0624 Observed ( m)10 S ( m J y ) F12590+2934 Observed ( m)10 S ( m J y ) F13126+2453 Observed ( m)10 S ( m J y ) F13182+3424
Article number, page 25 of 33 & A proofs: manuscript no. main Observed ( m)10 S ( m J y ) F13188+0036 Observed ( m)10 S ( m J y ) F13229-2934 Observed ( m)10 S ( m J y ) F13373+0105 Observed ( m)10 S ( m J y ) F14179+4927 Observed ( m)10 S ( m J y ) F13197-1627 Observed ( m)10 S ( m J y ) F13362+4831 Observed ( m)10 S ( m J y ) F13428+5608 Observed ( m)10 S ( m J y ) F14348-1447
Article number, page 26 of 33.-D. Paspaliaris et al.: The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies Observed ( m)10 S ( m J y ) F14547+2449 Observed ( m)10 S ( m J y ) F15163+4255 Observed ( m)10 S ( m J y ) F15327+2340 Observed ( m)10 S ( m J y ) F16284+0411 Observed ( m)10 S ( m J y ) F15107+0724 Observed ( m)10 S ( m J y ) F15250+3608 Observed ( m)10 S ( m J y ) F16104+5235 Observed ( m)10 S ( m J y ) F16577+5900
Article number, page 27 of 33 & A proofs: manuscript no. main Observed ( m)10 S ( m J y ) F17132+5313 Observed ( m)10 S ( m J y ) F22491-1808 Observed ( m)10 S ( m J y ) F23024+1916 Observed ( m)10 S ( m J y ) F23135+2517 Observed ( m)10 S ( m J y ) F22287-1917 Observed ( m)10 S ( m J y ) F23007+0836 Observed ( m)10 S ( m J y ) F23133-4251 Observed ( m)10 S ( m J y ) F23157-0441
Article number, page 28 of 33.-D. Paspaliaris et al.: The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies Observed ( m)10 S ( m J y ) F23254+0830 Observed ( m)10 S ( m J y ) F23488+2018 Observed ( m)10 S ( m J y ) F23488+1949
Fig. B.1.
Best-fit SED models for the 67 galaxies in the current sam-ple containing the unattenuated emission by the old (red) and the young(blue) stellar populations, the nebular line emission (dark-yellow), thedi ff use dust (orange), the emission from the PDR regions (green), aswell as the AGN emission (purple) are also presented. The best-fit SEDis indicated as a black curve while the observations, for each galaxy,along with their uncertainties, are indicated by violet open squares.Light-green dots stand for the best-model flux densities. Article number, page 29 of 33 & A proofs: manuscript no. main
Appendix C: Comparison with other studies
The (U)LIRGs sample under consideration has already beenstudied by others and some of the physical parameters discussedin this paper have been computed using either a similar approachor a totally di ff erent methodology. In what follows we will com-pare the parameters provided in the literature with what is com-puted in the current work and try to explain any di ff erences.The AGN is inevitably a very important contributor to theenergetics of the galaxies which can shape a large part of theirSED (at least for the galaxies that host strong AGNs). The con-tribution of this component can be parametrized by the fractionof its emitted luminosity to the bolometric luminosity emitted.This fraction ( f rac AGN ) can be estimated with various meth-ods exploiting the parameter space of observables like X-rays,emission lines at MIR, or MIR colors, but also through SEDmodelling like the one we are considering in this study. Dueto its multi-component nature the SED modelling is not alwaysa robust way to estimate f rac
AGN . In Ciesla et al. (2015) it isfound that only strong AGNs (with f rac
AGN > .
5) can be wellretrieved. On the other hand, other methods, like, e.g., MIR emis-sion line ratios, the 6.2 µ m PAH EW, the S / S dust continuumslope as well as MIR diagnostic diagrams provide a more robustindication of the strength of the AGN. Such a study is presentedin Díaz-Santos et al. (2017) where the average fractional lumi-nosity contribution of the AGN to the bolometric luminosity ofthe galaxies in the GOALS survey, based on the above methods,is provided, using the Kaplan-Meier (KM) maximum likelihoodestimator. In Fig. C.1 we compare the values of f rac AGN derivedin the current study with CIGALE with those calculated in Díaz-Santos et al. (2017) for the galaxies in common (yellow points).We see that, despite the large scatter of the measurements, es-pecially in the low- f rac
AGN end there is an overall agreementwith stronger AGNs showing higher f rac
AGN with both meth-ods. Some of the scatter seen in this plot arises from the fact thatthe values derived with CIGALE come from a parameter gridwhere discrete values have been pre-selected. In Díaz-Santoset al. (2017), on the other hand, a continuous range for f rac
AGN is available. Apart from Díaz-Santos et al. (2017) we indicatethe resulting parameters (for the common galaxies) from twomore studies which use an approach similar to what we use inour study. These studies are Fritz et al. (2006) (red boxes), andRamos P et al. (2020) (blue "X"s). In Fritz et al. (2006) f rac
AGN is calculated (in the range 5-1000 µ m) by introducing a smoothlydistributed, toroidal-like, dusty structure around the galaxy’s nu-cleus, heated by a central source. In Ramos P et al. (2020) anapproach similar to one presented in our work, is used, withCIGALE SED modelling performed using a di ff erent parame-ter grid, with the most obvious di ff erences being the use of theFritz et al. (2006) AGN module and the Dale et al. (2014) dustmodel. We see that, despite the small number of the galaxies incommon, the findings broadly agree between the two methods(Fritz et al. 2006 and Ramos P et al. 2020).In Fig. C.2 we present the comparison of four basic param-eters for the galaxies (namely, M star , M dust , SFR , and T dust ) be-tween the values derived in U et al. (2012) and the current work(with additional information from Casey 2012 and Herrero-Illana et al. 2019 for the case of T dust ).In the top-left panel of Fig. C.2 we compare the stellarmasses between the two studies. In U et al. (2012) two di ff er-ent IMFs have been considered Salpeter (1955), and Chabrier(2003). We compare with the Salpeter IMF since this is the oneconsidered in the current study. The stellar masses derived in Uet al. (2012) are computed with two methods, by performing an frac AGN literature f r a c A G N C I G A L E Díaz-Santos et al. 2017Fritz et al. 2006Ramos P et al. 2020
Fig. C.1.
CIGALE-derived AGN fractions of this study, compared tothe corresponding fractions calculated in previous works. Yellow circlesdepict the comparison with AGN fractions by Díaz-Santos et al. (2017),red squares to the ones by Fritz et al. (2006), and filled blue "X"s to theones by Ramos P et al. (2020). All the data come with their error-bars(black), while the grey solid line stands for the one-to-one relation. optical-NIR SED fitting, adopting the Bruzual & Charlot (2003)stellar population synthesis model, or by scaling from the H-band luminosity. Since the SED modelling method adopted inU et al. (2012) di ff ers from the one used in the current work,we choose to compare our results with the scaling, from the H-band luminosity, method. The H-band, being largely una ff ectedby dust attenuation but also from contamination by hot dustemission from AGNs (Hainline et al. 2011) is a reliable tracerof the stellar mass. From Fig. C.2 we see an overall good agree-ment between the two studies (we find an average di ff erence of0.096 dex), with only a few exceptions deviating substantiallyfrom the one-to-one relation (the di ff erence becomes 0.077 dexwhen the outliers are not considered). For F09333 + + ffi cult to judge if their fit was successful, or not. Finally,concerning F23488 + ff erences are due to the significant e ff ects of thedust on the stellar mass computation. In this plot we also indicatethe galaxies with the largest AGN fractions ( f rac AGN > .
2; pinkoctagons). It is evident that the majority of the stronger AGNs in
Article number, page 30 of 33.-D. Paspaliaris et al.: The physical properties of local (U)LIRGs: a comparison with nearby early- and late-type galaxies
Fig. C.2.
Comparison between the CIGALE derived properties (this work), M star (top-left panel), M dust (top-right panel), SFR (bottom-left panel),and T dust (bottom-right panel) and the corresponding properties presented in U et al. (2012). Pink octagons indicate sources with strong AGN( f rac AGN > . our sample are hosted in galaxies on the high-end of the stellarmasses.In the top-right panel of Fig. C.2 we present the compar-ison of the dust masses derived in U et al. (2012) and in ourwork. We find that, apart from some outliers, which are going tobe discussed later, the derived masses between the two methodsagree with a systematic o ff set of 0.46 dex (0.45 when the out-liers are not considered). This can be explained with the use ofthe di ff erent dust absorption coe ffi cients ( κ = .
15 m kg − and κ = . g − ; assuming an emissivity index of β =
2) in thecase of U et al. (2012) compared to κ = . g − dictatedby the THEMIS model (used in the current study) which trans-lates to a 0.43 dex di ff erence. Concerning the obvious outliers inthis plot we see that the majority of them host a luminous AGN(with f rac AGN > .
2; pink octagons) making the estimation of thedust mass quite uncertain if no AGN component is considered inthe modelling. In the case of F04097 + µ m observation) while in the current work,the Herschel data fill this gap and a more accurately determina-tion of the dust mass is achieved. A similar case is F14348-1444 with the dust emission being constrained only by the IRAS 60and 100 µ m observation in the case of U et al. (2012) comparedto our analysis where Herschel observations are also available.In the bottom-left panel of Fig. C.2 we present the compari-son of the SFR derived in U et al. (2012) and in our work. In Uet al. (2012) the
SFR is derived by combining the monochromaticUV luminosity at 2800Å and IR luminosity, using the Wuytset al. (2011a) recipe. Since the IMFs used in the two methodsare di ff erent, the values derived with the Chabrier IMF were di-vided by a constant scaling factor of 0.63 (Madau & Dickinson2014). We see that there are some obvious outliers (with F13197-1627 being the most extreme case) though all are amongst thestrongest AGNs in our sample with f rac AGN > . ff ected by thepresence of the AGN (see, e.g., the SED of F13197-1627) it isexpected that the SFR is overestimated when the IR luminosity isused as an
SFR tracer. The di ff erence between the two methodsis 0.12 dex (with 0.08 dex when the outliers are not considered).In the bottom-right panel of Fig. C.2 the comparison amongthe values of T dust derived in this work and the studies of Uet al. (2012), Casey (2012), and Herrero-Illana et al. (2019)(yellow circles, green diamonds, and blue boxes respectively) Article number, page 31 of 33 & A proofs: manuscript no. main is presented. The strongest AGNs in our sample (those with f rac
AGN > .
2) are indicated in pink while the line is the one-to-one relation. In all three literature studies mentioned above, thedust temperature was estimated by fitting a single temperaturemodified black-body to observations landwards of ∼ µ m. Onething to notice is that the resulting values of our study come indiscrete ranges in dust temperature. This is due to the discretenature of the parameter space used by CIGALE and in particularfor the parameters that define U min (see Eq. 1). Even if the scat-ter is large it is evident that the values derived in this work fol-low the general trend observed in other studies. Indicative of thescatter is the average di ff erence and the standard deviation of thedi ff erences between our values and the literature values whichis − . ± . ff erence and the standarddeviation of the di ff erences in dust temperature then drops to0 . ± . f rac AGN = Appendix D: Cumulative distributions
Examining the cumulative distributions of the various physicalparameters is a powerful tool that may indicate if two populationof objects can originate from the same parent population or not.The Kolmogorov-Smirnov (Smirnov (1948)) test is a well knownnon-parametric statistical method that compares distributions bymeasuring a “di ff erence” between the distributions and report ap-value which shows a statistical significance of the result. Non-zero p-value of less than 0.15 means that the null hypothesis thatthe distributions come from the same parent distribution can berejected with 85% probability (see, e.g., Haan et al. (2011)).In Fig. D.1 we present the cumulative distributions of thephysical parameters presented in Sect. 4.2 (see also Fig. 4) forthe three di ff erent populations of ETGs, LTGs, and (U)LIRGs(red, blue, and yellow color respectively). As can be seen fromthe plots the relevant distributions are very di ff erent among thethree galaxy populations with p-values less than 0.15. The onlyexception being the comparison of the temperature distributionsbetween ETGs and LTGs which give a p-value of 0.82. A rel-evant discussion on the results of the KS tests is presented inSect. 4.2.In Fig. D.2 the cumulative distributions of the physical pa-rameters presented in Sect. 5 (see also Figs. 6, 7, 8) for the dif-ferent merging stages. The p-values drawn from the cumulativedistributions of all the combinations of merging stages for eachphysical parameter are given in Table D.1. A relevant discussionon the results of the KS tests is presented in Sect. 5. M dust [M fl ] Galaxy type
ETGsLTGs(U)LIRGs10 M star [M fl ] -4 -3 -2 -1 SFR [M fl yr − ] -4 -3 -2 -1 sSFR [Gyr − ] T dust [K] L PDR [L fl ] C u m u l a t i v e D i s t r i b u t i o n Fig. D.1.
Cumulative distributions of the physical parameters discussedin this study, for each galaxy type. The colouring is identical to Fig. 4.
Table D.1.
Kolmogorov-Smirnov (KS) test p-values of the physicalproperties ( M dust , M star , SFR , sSFR , T dust , L PDR and f rac
AGN ) for all themerging stage combinations. All p-values ≤ M dust M star SFR sSFR T dust L PDR f rac
AGN s-m 0.97 0.92 0.07 0.75 0.54 0.05 0.92s-M1 0.27 0.56 0.11 0.99 0.89 0.11 0.35s-M2 0.64 0.16 0.02 0.92 0.16 0.01 0.64s-M3 0.73 0.39 0.00 0.01 0.00 0.00 0.26s-M4 0.36 0.85 0.00 0.01 0.01 0.00 0.75s-M5 0.35 0.70 0.70 0.97 0.05 0.70 0.70m-M1 0.70 1.00 0.70 0.92 0.55 0.70 0.55m-M2 0.92 0.54 0.41 0.54 0.92 0.75 0.92m-M3 0.80 0.41 0.14 0.11 0.30 0.23 0.23m-M4 0.69 0.76 0.08 0.13 0.92 0.19 0.41m_M5 0.93 1.00 0.93 1.00 0.93 0.93 0.93M1-M2 0.92 0.66 0.81 0.81 0.71 0.56 0.35M1-M3 0.29 0.45 0.22 0.03 0.02 0.25 0.36M1-M4 0.96 0.31 0.09 0.03 0.51 0.25 0.51M1-M5 0.43 0.86 0.86 1.00 0.43 0.86 0.86M2-M3 0.39 0.03 0.21 0.02 0.03 0.07 0.26M2-M4 0.98 0.11 0.11 0.01 0.64 0.03 0.47M2-M5 0.25 0.47 0.87 0.70 0.17 0.87 0.87M3-M4 0.49 0.93 0.84 0.87 0.51 0.84 0.95M3-M5 0.25 0.33 0.63 0.25 0.79 0.63 0.53M4-M5 0.30 0.46 0.76 0.18 0.76 0.76 0.76