The SAMI Galaxy Survey: bulge and disk stellar population properties in cluster galaxies
S. Barsanti, M. S. Owers, R. M. McDermid, K. Bekki, J. Bland-Hawthorn, S. Brough, J. J. Bryant, L. Cortese, S. M. Croom, C. Foster, J. S. Lawrence, A. R. López-Sánchez, S. Oh, A. S. G. Robotham, N. Scott, S. M. Sweet, J. van de Sande
DDraft version November 12, 2020
Typeset using L A TEX twocolumn style in AASTeX63
The SAMI Galaxy Survey: bulge and disk stellar population properties in cluster galaxies
S. Barsanti,
1, 2
M. S. Owers,
1, 2
R. M. McDermid,
1, 2
K. Bekki, J. Bland-Hawthorn, S. Brough,
5, 6
J. J. Bryant,
4, 6, 7
L. Cortese,
3, 6
S. M. Croom,
4, 6
C. Foster,
4, 6
J. S. Lawrence, ´A. R. L´opez-S´anchez,
1, 6, 8
S. Oh,
6, 9
A. S. G. Robotham,
3, 6
N. Scott,
4, 6
S. M. Sweet,
6, 10 and J. van de Sande
4, 6 Department of Physics and Astronomy, Macquarie University, NSW 2109, Australia Astronomy, Astrophysics and Astrophotonics Research Centre, Macquarie University, Sydney, NSW 2109, Australia ICRAR, The University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia Sydney Institute for Astronomy (SIfA), School of Physics, University of Sydney, NSW 2006, Australia School of Physics, University of New South Wales, NSW 2052, Australia ARC Centre of Excellence for All Sky Astrophysics in 3 Dimensions (ASTRO 3D), Australia Australian Astronomical Optics, AAO-USydney, School of Physics, University of Sydney, NSW 2006, Australia Australian Astronomical Optics - Macquarie, Macquarie University, NSW 2109, Australia Research School of Astronomy and Astrophysics, Australian National University, Canberra, ACT 2611, Australia School of Mathematics and Physics, University of Queensland, Brisbane, QLD 4072, Australia
Submitted to ApJ. Accepted on November 9, 2020ABSTRACTWe explore stellar population properties separately in the bulge and the disk of double-componentcluster galaxies to shed light on the formation of lenticular galaxies in dense environments. We studyeight low-redshift clusters from the Sydney-AAO Multi-object Integral field (SAMI) Galaxy Survey,using 2D photometric bulge-disk decomposition in the g , r and i -bands to characterize galaxies. For 192double-component galaxies with M ∗ > M (cid:12) we estimate the color, age and metallicity of the bulgeand the disk. The analysis of the g − i colors reveals that bulges are redder than their surrounding diskswith a median offset of 0.12 ± Keywords: surveys – galaxies: clusters – galaxies: evolution – galaxies: fundamental parameters –galaxies: structure INTRODUCTIONLenticular (S0) galaxies were introduced as a separatemorphology for the first time by Hubble (1936). Theyare characterized by a central bulge, similar to ellipticalgalaxies, and a disk, similar to spiral galaxies. However,
Corresponding author: Stefania [email protected] unlike spiral galaxies, the disks of S0 do not containspiral arms and ongoing star formation. Due to thesefeatures S0s are located between ellipticals and spiralsin the Hubble Sequence and are considered a transitionalphase. S0s show heterogeneous physical properties (e.g.Laurikainen et al. 2010; Cappellari et al. 2011; Barwayet al. 2013), causing their origin still to be highly de-bated. Historically, S0s were speculated to be formedfrom faded spirals due to the similarities between their a r X i v : . [ a s t r o - ph . GA ] N ov Barsanti et al. structural and kinematical properties (van den Bergh1976; Bedregal et al. 2006; Moran et al. 2007; ProchaskaChamberlain et al. 2011). However, some works showthat some properties of S0s differ from those of spiralgalaxies (Christlein & Zabludoff 2004; Williams et al.2010; Falc´on-Barroso et al. 2015), indicating that majormergers can be responsible for their formation (Spitzer& Baade 1951; Bekki 1998; Tapia et al. 2017).S0s are mainly found in groups and clusters of galax-ies while spirals populate lower density regions, imply-ing that the environment plays a fundamental role intheir formation (Dressler 1980). This result suggests ascenario where spiral galaxies are transformed into S0swhen they enter the cluster environment, where theirstar formation is quenched (e.g. Dressler et al. 1997;Fasano et al. 2000; Kormendy & Bender 2012). Thefraction of S0 galaxies compared to spirals in clusters isobserved to increase with time, providing evidence forthe transformation of spirals into S0s (Poggianti et al.2009). Groups and pre-processing mechanisms can alsoplay an important role for this transformation, since thechanging fraction of S0s relative to spirals as a functionof time is found to increase for lower halo masses (Pog-gianti et al. 2009). This suggests that mechanisms thatare not simply due to cluster processes might also beinvolved.To explore the possible formation of S0 galaxies, it isfundamental to understand the individual stellar popu-lations of the bulges and the disks of S0 cluster galaxies.The study of the age and metallicity of the separatestellar populations help us to distinguish between sce-narios exclusively caused by cluster processes and thosedue to mechanisms that may also occur in lower den-sity environments. Environment processes in clustersinclude interactions with the intra-cluster medium whenthe galaxy travels through the cluster. In this scenariothe cold gas in the disk is stripped by ram pressure strip-ping (Gunn & Gott 1972), or the hot halo reservoir isremoved by strangulation (Larson et al. 1980). Theseprocesses act on the disks and leave the old stellar popu-lations in the bulge unperturbed (Quilis et al. 2000). Onthe other hand, a strongly concentrated star formationactivity in the galaxy central region is observed whenneighbouring galaxies interact, causing gas stripping bygalaxy harassment in clusters (Moore et al. 1996) orminor mergers in lower density environments (Mihos &Hernquist 1994). These scenarios imprint different fossilrecords in the stellar populations suggesting that theymay be distinguished via analyses of spectra.In recent years, much effort has been devoted to thestudy of S0 galaxies and the properties of the stellarpopulations of their components, such as colors, ages and metallicities. This allows the development of newmethodologies in order to identify S0 galaxies, decom-pose them in their components and analyze their prop-erties both in galaxy clusters (Hudson et al. 2010; John-ston et al. 2014; Head et al. 2014; Johnston et al. 2020)and in the field environment (Fraser-McKelvie et al.2018; Tabor et al. 2019; M´endez-Abreu et al. 2019a,b;Johnston et al. 2020). Differences between the bulge anddisk properties of cluster and field galaxies may indicatedifferent formation scenarios for S0s. In this paper, wefocus on the stellar populations of S0s in clusters in or-der to understand the role that environment plays intheir formation.While previous studies of S0s in clusters have providedimportant clues that help to understand the impact ofthe cluster environment on S0s, they have been limitedby their focus on photometric investigations (Head et al.2014), or by the relatively small number of objects withresolved spectroscopy (Johnston et al. 2014). Head et al.(2014) studied 200 S0 galaxies in the Coma cluster andperformed 2D photometric bulge-disk decomposition inthe u , g and i -bands. They analyzed the colors of thebulges and the disks, observing that the bulges are red-der than the disk components. This color difference isexpressed as linear combinations of stellar populationages and metallicities. They concluded that bulges areeither 2-3 times older or ∼ α -element abundances, observ-ing younger and more metal-rich stellar populations inbulges than those in the disks. They suggested that themost recent episode of star formation in S0s occurs inthe bulge, caused by gas previously enriched in the diskwhich has been dumped in the bulge during the pro-cesses of stripping and quenching of the disk. This pop-ulation of more metal-rich bulges is consistent with theresults of Head et al. (2014), however only a handful ofgalaxies in one cluster were analyzed by Johnston et al.(2014). In order to overcome these limitations, the largestatistical samples available from previous photometricstudies need to be combined with the spectroscopic pre-cision achieved on small samples. AMI: bulge and disk stellar populations g , r and i -bands using the image analysis package ProFound andthe photometric galaxy profile fitting package
ProFit (Robotham et al. 2017, 2018). We test three methodsto estimate the age and metallicity of the bulge and thedisk, combining the 2D photometric bulge-disk decom-position with the stellar population information fromthe SAMI spectra:1. We develop a new method based on galaxy stellarmass weights for the two components. We use the2D photometric bulge-disk decomposition resultsto estimate for each spatial bin the contributionsof the bulge and the disk to the total galaxy stellarmass. These bulge-disk stellar mass fractions areused as weights for the galaxy age and metallicityderived from the SAMI spectra.2. We adapt a method similar to that of M´endez-Abreu et al. (2019a) based on flux weights forthe two components. We perform 2D photometricbulge-disk decomposition on the SAMI datacubesat each SAMI wavelength to obtain separate dat-acubes and 1D spectra for the bulge and the disk.3. We adapt a method similar to that of Fraser-McKelvie et al. (2018) based on radial separation.We consider as the 1D bulge spectrum the galaxyspectrum from the most central bin, while the rep-resentative 1D disk spectrum is the galaxy spec-trum from the outermost spatial bin.This paper is organized as follows. We presentour SAMI cluster sample and the selection of double-component galaxies in Section 2. Section 3 describesthe three different methods to estimate mass-weighted single-age and single-metallicity values of the bulge andthe disk. In Section 4 we present our results, analyzingthe comparison between the stellar population proper-ties of the bulges and the disks from the three meth-ods. In Section 5 we compare our results with previousworks and we discuss the physical implications. Finally,we summarize and conclude in Section 6. Throughoutthis work, the uncertainties on the percentages are esti-mated using the method of Cameron (2011) to measureconfidence intervals on binomial population proportions.The bin sizes of the histograms are estimated accordingto the Scott’s rule of thumb (Scott 2010). We assumeΩ m = 0 .
3, Ω Λ = 0 . H = 70 km s − Mpc − ascosmological parameters. DATA AND SAMPLE SELECTIONIn this Section we describe the details of the SAMIGalaxy Survey and the considered sample of SAMI clus-ter galaxies. Finally, we select the sample of SAMIcluster galaxies modelled by a photometric double-component profile.2.1.
SAMI Galaxy Survey
SAMI was mounted on the 3.9m Anglo-AustralianTelescope (Croom et al. 2012). The instrument is char-acterized by 13 fused optical fibre bundles (hexabun-dles), each one containing 61 fibres of 1.6 (cid:48)(cid:48) diameter re-sulting in each Integral Field Unit (IFU) having a 15 (cid:48)(cid:48) diameter (Bland-Hawthorn et al. 2011; Bryant et al.2014). The SAMI fibres are fed into the two arms ofthe AAOmega spectrograph (Sharp et al. 2006). TheSAMI Galaxy Survey uses the 580V grating in the bluearm resulting in a resolution R=1812 and wavelengthcoverage of 3700-5700 ˚A, and the 1000R grating in thered arm resulting in the higher resolution R=4263 overthe range 6300-7400 ˚A. The median full-width-at-half-maximum values for each arm are FWHM blue =2.65 ˚Aand FWHM red =1.61 ˚A (van de Sande et al. 2017a).The SAMI Galaxy Survey is a spatially-resolvedspectroscopic survey of more than 3000 galaxies col-lected during 2013-2018, with stellar mass rangelog ( M ∗ /M (cid:12) ) = 8 −
12 and redshift range 0 . < z ≤ .
115 (Bryant et al. 2015). The data are reduced usingthe SAMI python package (Allen et al. 2014) which in-cludes the 2d fdr package (AAO Software Team 2015).We make use of the internal data release v0.11 (for acomplete description of the data reduction from rawframes to datacubes see Sharp et al. 2015; Allen et al.2015; Green et al. 2018; Scott et al. 2018). The final dat-acubes are characterized by a grid of 0.5 (cid:48)(cid:48) × . (cid:48)(cid:48) spaxels,where the blue and red spectra have pixel scales of 1.03 ˚Aand 0.56 ˚A, respectively. Barsanti et al.
SAMI cluster sample
The cluster component of the SAMI Galaxy Surveycontains 906 galaxies selected from eight low-redshiftmassive clusters with virial masses log ( M /M (cid:12) ) =14 . − .
19 at 0 . < z < .
058 (for a detailed descrip-tion see Owers et al. 2017). In summary, the galaxiesare selected using: redshift-dependent stellar mass cutswith a lowest limit of log ( M ∗ /M (cid:12) ) = 9 .
5, projectedcluster-centric distance
R < R and peculiar veloc-ity | v pec | < . σ where σ is the cluster velocitydispersion within R .The SAMI cluster sample contains the four clustersAPMCC0917, EDCC0442, A3880 and A4038 selectedfrom the 2dFGRS catalogue (De Propris et al. 2002)with the photometry provided by the VLT Survey Tele-scope’s ATLAS (VST/ATLAS) survey (Shanks et al.2013, 2015), and the four clusters A85, A168, A119 andA2399 in the SDSS regions with the photometric datafrom SDSS DR9 (Ahn et al. 2012). The imaging dataare used in Section 2.3 for galaxy characterization withthe purpose of selecting only double-component clustergalaxies. 2.3. Galaxy characterization
Our primary aim is to understand the bulge and diskstellar populations of S0 galaxies. To do so, we mustfirst identify those SAMI galaxies that can be decom-posed into double-component systems. We use a purelyphotometric approach to separate the bulge and the diskcomponents. This can cause semantic confusion com-pared to the classical “bulge” and “disk” definitions. Inthis study the “disk” is defined as the exponential com-ponent, while the “bulge” corresponds to the S´ersic com-ponent representing the light excess over the exponentialcomponent. We decide to nominate the two componentsas the “bulge” and the “disk” in line with the previousliterature and in order to offer easier comparisons withthe previous works of Hudson et al. (2010); Head et al.(2014); Johnston et al. (2014); Fraser-McKelvie et al.(2018); M´endez-Abreu et al. (2019a).We make use of the two-dimensional (2D) photo-metric bulge-disk decomposition in the g -, r - and i -bands presented by Barsanti et al. in preparation forthe SAMI cluster sample. They use the VST/ATLASimaging for the APMCC0917, EDCC0442, A3880 andA4038 clusters and the SDSS imaging for the A85, A168,A119 and A2399 clusters, as introduced in Section 2.2.More details about the imaging data for the clustersare reported in Owers et al. (2017). The 2D bulge-disk decomposition is performed for 1204 SDSS and591 VST/ATLAS cluster galaxies, for a total of 1795.This sample includes galaxies characterized by stellar mass M ∗ ≥ . M (cid:12) , projected cluster-centric distancewithin 2 . R and both with and without SAMI obser-vations. For source finding and image analysis they usethe ProFound package (Robotham et al. 2018).
Pro-Found offers the key input requirements to
ProFit ,a package for Bayesian 2D photometric galaxy profilemodelling (Robotham et al. 2017). Barsanti et al. inprep. fit single-component S´ersic profile and multipledouble-component bulge+disk profiles for each galaxyin the r -band. The disk profile is fixed to exponential,while the S´ersic bulge profile varies with increasing com-plexity. The results for the structure parameters fromthe r -band are used as inputs for the single/double-component fits in the g - and i -bands. The pipelines ProFound + ProFit work in the r -, g - and i -bandsfor 1730/1795 cluster galaxies. For these galaxies themodel selection is performed in the r -band using theBayes Factor (BF). The double-component sample hasln(BF) >
60. Double-component galaxies for which theseparation of the bulge and disk measurements is unreli-able, i.e. they are dominated by one component havingbulge-to-total flux ratio B/T < . > .
8, orfor which the fits are unphysical, i.e. the fitted parame-ters are pegged at the fitting boundaries, are excluded.The final number of double-component cluster galaxiesis 469, including galaxies with M ∗ ≥ . M (cid:12) , within2 . R and both with and without SAMI observations.From the 469 double-component galaxies, we selectonly the SAMI primary targets, i.e. galaxies with SAMIobservations within R . We consider only galaxieswith stellar mass M ∗ ≥ M (cid:12) ; for galaxies with M ∗ lower than this limit the SAMI data contain only a smallnumber of spaxels with signal-to-noise ratio > σ <
400 km s − as measured by vande Sande et al. (2017a). High velocity dispersions aretypical of the brightest cluster galaxies. Massive andbright ellipticals are excluded since their complex lightprofiles can contaminate the double-component galaxyselection. Applying these selection criteria we find 192SAMI double-component galaxies.Since we want to study S0 galaxies, we consider thespectroscopic, kinematic and morphological propertiesof the 192 double-component galaxies. Using the spec-troscopic classification of Owers et al. (2019), we findthat 170/190 double-component galaxies are in the pas-sive class, while for 2 galaxies the measurements arenot available. van de Sande et al. (2017b) measuredthe kinematic properties for 175/192 double-componentgalaxies, finding that 160/175 galaxies are fast rotatorsaccording to the definition of Cappellari (2016). Usingthe morphology classification of Cortese et al. (2016) AMI: bulge and disk stellar populations M ∗ ≥ M (cid:12) . Accord-ing to the morphological classification of Cortese et al.(2016), the galaxy morphology is represented by: ellip-tical=0, elliptical/S0=0.5, S0=1, S0/early-spiral=1.5,early-spiral=2, early/late-spiral=2.5 and late-spiral=3.The morphological distribution shows a peak for S0galaxies, which is what expected for double-componentgalaxies and for dense environments (Sil’chenko et al.2012). The median values B/T=0.52 and n bulge =3.48highlight the reliability of the model selection, since weexpect B/T not to assume extreme high or low valuesas implemented in the filtering process and n bulge =4 fora typical galaxy with two components. METHODS FOR SEPARATING BULGE ANDDISK STELLAR POPULATION PROPERTIESThe aim of this paper is to analyze the individualstellar population properties of the bulge and the disk.This is a challenging problem, with projection effects,beam smearing, and imperfect modelling all impedingthe separation of bulge and disk light. There have beenrecent attempts to overcome these problems using Inte-gral Field spectroscopy, including those that decomposeIFU spectra wavelength-by-wavelength (Johnston et al.2017; M´endez-Abreu et al. 2019a), and those that com-bine spectra from regions dominated by bulge and disklight (Fraser-McKelvie et al. 2018). Here, we develop anew method based on the combination of stellar massmaps and IFU spectroscopy. We also apply the two pre-viously used methods of M´endez-Abreu et al. (2019a)and Fraser-McKelvie et al. (2018) to compare our re-sults. 3.1.
Method based on M ∗ weights We implement a new mass-weighted method to sepa-rate the bulge and the disk stellar population properties.At each observed position in a SAMI cube, we assumethat the observed spectrum is a result of the combinationof at most two simple stellar populations coming fromeither or both a bulge and disk component. We assumethat the mass-weighted contributions of the two popu-lations to each observed galaxy spectrum add approxi-mately linearly. Under these assumptions, we are able to disentangle the relative contributions from the bulgeand disk stellar populations in a spectrum if we haveknowledge of the fractional contribution of mass eachpopulation makes to that spectrum. The resolved spec-troscopy offers the advantage that at each position thefractional contribution of the bulge and the disk variesin a known way that can be determined from our 2Dbulge-disk decompositions. In this context, the combi-nation of the 2D bulge-disk decomposition and the 2DIFU spectroscopic information allow us to recover thebulge and disk stellar population properties.For each galaxy we estimate the mass-weighted ageand metallicity using each observed spectrum in thegalaxy SAMI datacube. Taking advantage of the 2Dbulge-disk decomposition, we estimate for each positionthe separate contributions to the galaxy stellar massfrom the bulge and the disk. To measure the bulge-disk ages and metallicities, we consider a linear model:
Age [ x, y ] = f bulgeM ∗ [ x, y ] × Age bulge + f diskM ∗ [ x, y ] × Age disk (1)[
M/H ][ x, y ] = f bulgeM ∗ [ x, y ] × [ M/H ] bulge + f diskM ∗ [ x, y ] × [ M/H ] disk (2) where x and y are the spatial positions of the measuredspectrum, Age [ x, y ] / [ M/H ][ x, y ] is the galaxy mass-weighted age/metallicity at this position, f bulgeM ∗ [ x, y ]and f diskM ∗ [ x, y ] are the fractions of stellar mass dueto the bulge and disk components at the sameposition respectively, and Age bulge / [ M/H ] bulge and Age disk / [ M/H ] disk are the fitted parameters represent-ing the single-age/metallicity values of the galaxy com-ponents. We aim to find the combination of bulgeand disk age/metallicity that best reproduce the wholegalaxy age/metallicity spatial maps. We fit for thesingle-age/metallicity values of both galaxy componentsusing the least-squares method. We minimize the sumsof the squared residuals in the parameter space: S ( Age [ x, y ]) = (cid:88) ( Age [ x, y ] e − Age [ x, y ] m ) (3) S ([ M/H ][ x, y ]) (cid:88) ([ M/H ][ x, y ] e − [ M/H ][ x, y ] m ) (4)where Age [ x, y ] e / [ M/H ][ x, y ] e are the estimatedgalaxy age/metallicity from the fitting and Age [ x, y ] m / [ M/H ][ x, y ] m are the measured values.This model is based on several assumptions that maylimit the outcomes. We assume that galaxies are char-acterized by only two components: a central bulge anda surrounding disk. We consider single values for themass-weighted age and metallicity of the bulge and the Barsanti et al.
Figure 1.
Distributions in M ∗ , morphology, bulge-to-total flux ratio and bulge S´ersic index for the 192 SAMI double-componentgalaxies. Stellar mass and morphology histograms are also plotted for the cluster galaxy sample of Barsanti et al. in preparation.The dashed lines represent the median values. The medians at morphology=1=S0, B/T=0.52 and n bulge = 3 .
48 show thereliability of the model selection. disk, without considering radial gradients. Finally, the2D projected mass-weighted quantities are separatedinto linear combinations, as described by Equations (1)and (2). Despite these limitations, this method offers in-sights to disentangle the bulge-disk mass-weighted stel-lar population properties without being time consuming.It is also the starting point to develop more complicatedmodels, e.g. adding radial gradients. A similar methodhas been explored by Hudson et al. (2010) (see their Ap-pendix), where they combined 2D bulge-disk decompo-sition with single-fibre spectra. They measure the bulgestellar population properties using spectral index mea-surements, making some assumptions on the disk prop-erties. Since our data are spatially-resolved and cover alarger fraction of the galaxy light, we can directly fit forthe disk stellar properties.The following Sections describe the building blocks ofthe method: the spatial binning, the kinematic correc-tions applied to the SAMI spectra, determining f bulgeM ∗ ,bin and f diskM ∗ ,bin , determining Age bin and [
M/H ] bin and fi-nally estimating Age bulge/disk and [
M/H ] bulge/disk .3.1.1. Spatial binning of the SAMI spectra
Determining
Age [ x, y ] and [ M/H ][ x, y ] for our se-lected galaxies represents a crucial step toward our goalof measuring separately the stellar population parame-ters for the bulge and disks. They are obtained fromfull-spectrum fitting using the Penalized Pixel-Fittingsoftware (pPXF; Cappellari & Emsellem 2004, Cappel-lari 2017).The SAMI galaxy spectra are spatially binned to im-prove the signal-to-noise ratio (S/N) for the spectralfitting analysis and make more reliable measurements.The S/N is measured taking the median value of theflux divided by the noise in the rest-frame wavelengthrange 4600-4800 ˚A . The noise is defined as the squareroot of the variance at each wavelength from the SAMIdatacube, including contributions from the covariance. AMI: bulge and disk stellar populations g -band isrebinned onto 50 ×
50 pixels matching the SAMI gridin order to generate a SAMI-like image. The modelimage is convolved with the SAMI PSF, which is mod-elled as a Moffat profile (Allen et al. 2015). We choosethe g -band since it is contained within the wavelengthrange covered by the SAMI blue arm spectra. The fluxdistribution of this model image is used to generate an-nular bins for stacking the SAMI spectra. We followthe isophotes of the model and adaptively grow annu-lar bins until the S/N=20, which is the minimum S/Nrequired to reliably measure age and metallicity (see Ap-pendix A). This methodology for spatial binning allowsfor better separation of bulge and disk spectra in thestacking, with respect to using a single ellipticity andposition angle to define annular bins. Figure 2 shows anexample of the binning for the galaxy 9091700038. Inthis case, a fixed elliptical annulus would not follow therelatively round central bulge component.The outer regions of the galaxies have low S/N and itis possible that the outer bin never meets the require-ment of S/N ≥
20. If this is the case, we still consider theouter bin if its S/N >
10 otherwise we exclude it. For thestacking of the SAMI spectra we use the binning codefrom the SAMI software package, which sums the fluxesof the spaxels for each bin and weights them accountingfor covariances (Sharp et al. 2015; Allen et al. 2015).3.1.2.
Kinematic corrections
Large velocity and velocity dispersion fields can pro-duce complicated line profiles after stacking. Thus, foreach spaxel the spectrum is corrected for velocity andvelocity dispersion prior to stacking. Correcting eachspectrum for the velocity field has the advantage ofaligning emission and absorption features prior to stack-ing. This increases the S/N post-stacking comparedwith non-corrected spectra. Finally, the improved S/Nof absorption lines helps with determining stellar popu-lation properties.We use the 2D SAMI velocity V pP XF and velocity dis-persion σ pP XF maps measured by van de Sande et al.(2017a). Following Johnston et al. (2014), the velocitydispersion is corrected bringing each spectrum to thesame maximum value σ max measured within the galaxy.Each spectrum is convolved with the appropriate Gaus- sian having a corrected σ cor due to: σ cor = ∆ σλδ λ × (1 + z ) c (5)where ∆ σ = (cid:113) σ max − σ pP XF , λ is the SAMI blue orred wavelength range, δ λ is the spectral pixel scale of1.03 ˚A or 0.56 ˚A for the SAMI blue or red arm, z is thegalaxy redshift and c is the speed of light. The velocitycorrection is applied by considering the spectrum of eachspaxel at the corresponding wavelengths λ cor = λ/ (1 + V pP XF /c ). Then, we interpolate to a common SAMIwavelength range for each spectrum. To be consistentwe also apply these kinematic corrections to the noise.In this case the corrected sigma for the Gaussian usedin the convolution is σ cor / √
2. The final corrected noiseis then obtained dividing the original noise by 2 √ πσ cor .We do not account for the covariance introduced by thesmoothing.For each galaxy the corrected spectra and associatednoise in each spaxel are used to build the final cor-rected blue and red SAMI datacubes. Finally, the lim-its and the accuracy of these corrections are tested inAppendix B. If the kinematic corrections are appliedcorrectly, then bulges and disks are expected to havematched zero velocity and maximum velocity dispersionmeasured within the galaxy. We perform the fitting forthe kinematic components of the 1D spectra of the bulgeand the disk obtained with the flux weights method inthe following Section 3.2. We find that velocities andvelocity dispersions for bulges and disks follow one-to-one relationships, but there are some offsets due to thelow signal-to-noise ratio of the 1D disk spectra.3.1.3. Determining f bulgeM ∗ ,bin and f diskM ∗ ,bin From the 2D photometric bulge-disk decompositionresults described in Barsanti et al. in preparation, wegenerate projected stellar mass maps for the bulge anddisk components. These stellar mass maps are requiredto determine the f bulgeM ∗ ,bin and f diskM ∗ ,bin fractions in Equa-tions (1) and (2). We use ProFit to reconstruct sepa-rate 2D bulge and disk flux maps in the g - and i -bandswithin the SAMI-like image. The following formula forthe SAMI galaxies is used to find the projected bulge-disk stellar mass maps from the bulge-disk g - and i -bandflux maps (Bryant et al. 2015; Owers et al. 2017):log ( M ∗ /M (cid:12) ) = − . i + 2 log ( D L / − log (1 + z )+(1 . − . z ) + (0 . − . z ) × ( g − i )(6)where z is the galaxy redshift and D L is the luminositydistance measured in parsec using the cluster redshift. Barsanti et al.
Figure 2.
Mass-weighted age ( left upper panels ) and metallicity ( left lower panels ) maps for the 9091700038 galaxy, representingthe data ( left ), model ( middle ) and residuals ( right ). For one spaxel of each spatial bin
Age bin and [
M/H ] bin are linearly fittedfor Age bulge/disk and [
M/H ] bulge/disk weighted for f bulgeM ∗ ,bin and f diskM ∗ ,bin ( right panels ). On average the residuals agree within 2 σ ,but for some bins the agreement is only at 5 σ . This is consistent with radial gradients within the galaxy that cannot be fittedfrom the liner model represented by Equations (1) and (2). The left panel of Figure 3 shows the VST/ATLAS gri color image for the 9091700038 galaxy, with the red cir-cle showing the diameter of a SAMI hexabundle. The right panels show for the bulge and the disk the g -bandand i -band flux maps from the 2D bulge-disk decompo-sition. These flux maps are used in Equation (6) to buildthe projected stellar mass maps ( right-most panels ) forthe two components.The stellar mass maps are spatially binned as outlinedin Section 3.1.1 in order to match the binning performedon the SAMI cubes. Finally, they are used to estimatethe bulge-disk contributions to the galaxy stellar mass.For each bin we estimate f bulgeM ∗ ,bin as the sum of the bulgestellar masses of the spaxels divided by the summed to- tal bulge+disk stellar masses for the same spaxels: f bulgeM ∗ ,bin = (cid:80) i M bulge ∗ ,i (cid:80) i ( M bulge ∗ ,i + M disk ∗ ,i ) (7)while the disk stellar mass fraction is f diskM ∗ ,bin = 1 − f bulgeM ∗ ,bin . The most right panels of Figure 2 show theprojected bulge and disk stellar mass fraction maps forthe 9091700038 galaxy.3.1.4. Determining
Age bin and [ M/H ] bin The SAMI spectra corrected for velocity and veloc-ity dispersion are spatially binned as described in Sec-tion 3.1.1. These spectra are used to determine themass-weighted age and metallicity associated with eachbin,
Age bin and [
M/H ] bin in Equations (1) and (2). We AMI: bulge and disk stellar populations Figure 3.
The left panel shows the VST/ATLAS gri color image for the 9091700038 galaxy. The red cross marks the galaxycentre, the red circle shows the 15 (cid:48)(cid:48) diameter of a SAMI hexabundle and the top right image is the SAMI PSF. The right panels show the g -band flux, i -band flux and stellar mass maps for the bulge ( top ) and the disk ( bottom ). The projected maps arebuilt within 50 ×
50 pixels to match the SAMI grid. fit the corrected SAMI spectra with pPXF, making useof the MILES single stellar population (SSP) templates(Vazdekis et al. 2010). We consider the isochrones devel-oped by the Padova group (Girardi et al. 2000), and theunimodal initial mass function having logarithmic slope1.3. To fit for galaxy age and metallicity, we follow thefirst three steps of Kacharov et al. (2018). Since weare not interested in star-formation history, we do notuse regularization which finds the smoothest solution inagreement with the data (McDermid et al. 2015; Cap-pellari 2017). Regularization might introduce some bias,smoothing the differences between the bulge-disk prop-erties. Woo & Ellison (2019) found that the accuracy ofthe mass-weighted age and metallicity for the regular-ized solution is similar to the unregularized one. Thus,we decide to estimate unregularized solutions, which arealso much faster to compute (Woo & Ellison 2019).The procedure we apply to build the mass-weightedage and metallicity maps for each galaxy is the following:1. We combine the blue and the red spectra of eachspatial bin following the approach of van de Sandeet al. (2017a). The associated blue and red noisesare combined in the same way.2. We fit for the stellar kinematic components. Wede-redshift the spectrum to a rest-frame wave-length using the galaxy redshift estimated by vande Sande et al. (2017a). We fit for the kinematiccomponents even if the spectrum is corrected forvelocity and velocity dispersion because of somelimits in the applied kinematic corrections (seeFigure 22 in Appendix B). We assume a Gauss-Hermite line-of-sight-velocity-distribution, fitting for galaxy velocity, velocity dispersion, skewnessand kurtosis. We use a 12th order additive poly-nomial to minimise template mismatch and to cor-rect for sky subtraction imperfections and scat-tered light. For the inputs of the kinematic com-ponents we use zero for the velocity and the max-imum value for the velocity dispersion.3. We derive the extinction correction curve. Wekeep the stellar kinematic parameters fixed tothose obtained at step (2). We re-fit using pPXFand we adopt the reddening law of Calzetti et al.(2000). For the remaining steps, the observedgalaxy spectrum is corrected for the best-fittingextinction curve.4. We find the best-fit linear weighted combinationof SSP templates for the spectrum. We use a12th order multiplicative polynomial to correct forany mismatch between the templates and spec-trum due to, e.g., data reduction anomalies. Inthe fit we also include gas emission lines whichcan be well constrained since we include the redspectrum. A more reliable estimate of the noise isperformed considering the residuals between thespectrum and the best-fit model at different wave-length ranges. We estimate the dispersion of theresiduals around the mean wavelength value ofeach range and then we interpolate the dispersionsthrough the entire wavelength interval as the newnoise.5. We perform a second fit following the previoussteps (2)-(4) using the new estimated noise. We0
Barsanti et al. determine
Age bin and [
M/H ] bin using the weightsdetermined by pPXF and Equations (1) and (2) ofMcDermid et al. (2015).6. Uncertainties on Age bin and [
M/H ] bin are deter-mined using a Monte Carlo approach. We addnoise to the best-fit model from step (5), usinga random Gaussian with sigma determined fromthe rescaled noise. We run 1000 simulations. Wecorrect the SSP templates for the kinematic com-ponents and for the multiplicative polynomial de-termined in step (2) and (5), respectively. Thus,in each simulation we fit only for the weights ofeach SSP template that is linearly combined toform the best-fit model. Finally, 1001 Age bin and[
M/H ] bin are obtained for the same bin and thestandard deviations ( σ ) of these distributions areestimated for each bin.Figure 4 shows an example of the output plots frompPXF representing the spectra of each spatial bin forgalaxy 9091700038. We display the spatial bins ( left ),the best-fit spectra ( middle ) and the obtained estimatesin the age-metallicity grid ( right panels ). The spectrumassociated to the most external bin is characterized bythe highest systematic noise since it is obtained fromthe combination of low S/N spaxels. This can be seenin the unphysical upturn at the beginning of the bluespectrum. However, the final results do not change ifwe exclude the initial blue upturns in the pPXF fittingprocess.3.1.5. Putting everything together: determining
Age bulge/disk and [ M/H ] bulge/disk With the estimates for f bulgeM ∗ ,bin , f diskM ∗ ,bin , Age bin and[
M/H ] bin in hand, we can now use Equations (1) and (2)to determine which combination of bulge and disk stellarpopulation parameters best reproduce the observations,i.e. the whole galaxy age/metallicity maps. We attemptto reproduce the observed spatial variation in galaxy ageand metallicity using the bulge and disk fraction massmaps by assigning a single spatially constant age andmetallicity to each component. For each bin, we fit forthe best ( Age, [ M/H ]) bulge and ( Age, [ M/H ]) disk that,when combined, most closely reproduced the observedgalaxy Age bin and [
M/H ] bin , determined in the previousSection 3.1.4.We perform the fitting using the R optim func-tion with the “L-BFGS-B” algorithm, minimizing thesquared sum of the difference between the estimatedand the known galaxy age/metallicity maps. Ages andmetallicities of both components are constrained to thelower and upper limits of the SSP templates used in Section 3.1.4: [0.0631, 17.7828] Gyr for ages and [-2.32,0.22] for metallicities. The upper age limit leads to thepossibility of SSP ages that are higher than the age ofthe Universe ∼ . σ map estimated from theMonte Carlo simulations at step (6) of Section 3.1.4.Without any complicated modelling we are able to re-produce in this galaxy the gradients from outer regionsbeing older and more metal-poor to inner regions beingyounger and more metal-rich. On average the residu-als agree within 2 σ , however for few spatial bins theagreement is only at 5 σ . Further discussion about thisagreement is reported in Section 5.4.For the uncertainties on Age bulge/disk and[
M/H ] bulge/disk , we repeat this procedure fitting theremaining 1000 galaxy age/metallicity maps obtainedfrom the Monte Carlo simulations in Section 3.1.4.The 16th and 84th percentiles of the Age bulge/disk and[
M/H ] bulge/disk distributions are taken as the lowerand upper uncertainties. Figure 5 shows an exam-ple of these distributions for the 9091700038 galaxy.The results ( Age, [ M/H ]) bulge =(5.05 Gyr, 0.20) and( Age, [ M/H ]) disk =(15.06 Gyr, − σ for age and within 2 σ for metallicity.All 192 double-component galaxies have been decom-posed with this method. However, 109/192 have at leastone pegged solution to the lower or upper limits for Age bulge/disk and [
M/H ] bulge/disk . The pegged solutionsoccur for the bulge metallicity in 48 galaxies and the diskage in 64 galaxies. Bulges can have [ M/H ] bulge > . f diskM ∗ ,bin , suggestingthat low M ∗ weights affect the fits. The pegged solu- AMI: bulge and disk stellar populations Figure 4.
Output plots from pPXF representing the spectra of the seven spatial bins for the 9091700038 galaxy. From lefttop to bottom the spectra are from the most central to the outermost bin. The middle panels show the input spectrum (black),the best-fit model (red), the best-fitted gas emission lines (orange), the noise (green) and the masked pixels (grey). The rightpanels displays the age-metallicity grid with the corresponding weights of the best linear combination of SSP templates whichresult in the best-fit model (see Equations (1) and (2) of McDermid et al. 2015). Barsanti et al. tions can occur because this model does not account forradial gradients in the age/metallicity pushing the fit-ting results to the limits. Moreover, low M ∗ weightsmake the results harder to constrain. In Appendix Cwe show that the reliability of the results depends onthe bulge-to-total flux ratio B/T of the galaxy and it isaffected by the M ∗ weights. Low f bulge/diskM ∗ ,bin values im-ply that the results are more uncertain and might leadto pegged results at the lower or upper fit limits. Theage/metallicity of the bulge are better constrained forgalaxies with high B/T, while the results for the diskare better constrained for galaxies with low B/T. Theexclusion of the pegged solutions does not change theresults, thus we decide to consider the full sample of 192galaxies. 3.2. Method based on flux weights
We perform a method similar to the methodologypresented by M´endez-Abreu et al. (2019a,b). M´endez-Abreu et al. (2019a) studied three early-type galaxiesfrom the CALIFA survey (S´anchez et al. 2012), perform-ing spectro-photometric multi-component decomposi-tion of IFU datacubes. At each wavelength, the galaxysurface brightness distribution is fitted with a 2D bulge-disk photometric decomposition. This method providesboth a 1D spectrum and a spatially-resolved spectro-scopic datacube for each galaxy component. Johnstonet al. (2017, 2020) introduced a similar approach for 2Dbulge-disk decomposition of IFU datacubes to extractthe spectral properties of bulges and disks.We adapt the M´endez-Abreu et al. (2019a) method inthe following way.1. We slice the blue/red SAMI galaxy datacubes,which have been corrected for velocity and veloc-ity dispersion as described in Section 3.1.2, into2D narrow-band images. For the blue SAMI wave-length range 3650-6000 ˚A we build 7 narrow-bandimages of ∼
300 ˚A each, while for the red SAMIwavelength range 6240-7450 ˚A we build 5 narrow-band images of ∼
200 ˚A each. The images at eachwavelength bin are made from the datacubes usingthe median flux per pixel. Using the same proce-dure we build the 2D noise map associated witheach narrow-band image. We consider only pixelswith S/N > g , r and i -bands described in Barsanti etal. in preparation. The left panel of Figure 3shows for the 9091700038 galaxy the VST/ATLAS gri color image, where the red circle marks thesize of a SAMI hexabundle. In Figure 3 we also show the g -band and i -band flux projected mapsfor the two components from the 2D bulge-diskdecomposition for this galaxy. For the g , r and i -bands SDSS and VST/ATLAS show similar fil-ter + telescope + atmospheric transmissions (seeFigure 2 of Shanks et al. 2015). Thus, for bothSDSS and VST/ATLAS data each g , r and i -bandimage is assumed to be centred at 4770 ˚A, 6231 ˚Aand 7625 ˚A, respectively. The galaxy photometricand shape parameters are found to be dependenton the wavelength (Vika et al. 2013). We use theresults from the g , r , and i -band fits to linearlyinterpolate estimates for the bulge-to-total flux ra-tio, the effective radii of both components and theS´ersic index of the bulge at the wavelengths of thenarrow-band images. The S´ersic index of the diskis fixed to 1 and the angle and axial ratio of bothcomponents are fixed to the r − band results.3. We use ProFit to make a 2D double-componentgalaxy model of each narrow band image basedon its assigned photometric and shape parametersfrom the interpolations at step (2). Each modelis convolved with the single SAMI PSF image forthat galaxy (Allen et al. 2015). Then, it is fittedusing the 2D noise map as sigma map. We fix allthe parameters to the interpolated values, exceptfor the bulge-to-total flux ratio which is used toestimate initial values for the magnitudes of thebulge and the disk.4. We build monochromatic images using theblue/red SAMI step in wavelength of 1.05/0.59 ˚Aand obtaining a total of 4046 images. Eachmonochromatic image has assigned values of thephotometric and shape parameters, obtained fit-ting linearly the results of the 2D bulge-disk de-composition for the narrow band images. We alsoconstruct the 2D noise map associated with eachmonochromatic image. We perform 2D bulge-diskdecomposition on each monochromatic image us-ing
ProFit and fitting only for the magnitude ofboth components, as in step (3).5. Using the results of the 2D bulge-disk decomposi-tion for each monochromatic image, we build theseparate 2D models of the bulge and the disk con-volved with the SAMI PSF at each wavelength.We use the 2D total, bulge and disk flux mapsto estimate the 2D bulge-to-total and disk-to-total maps at each wavelength which act as fluxweights for the two components. We multiply eachmonochromatic image with the corresponding 2Dbulge-to-total and disk-to-total maps to obtain
AMI: bulge and disk stellar populations Bulge Age [Gyr] N u m be r o f s i m u l a t i on s Disk Age [Gyr] N u m be r o f s i m u l a t i on s
14 15 16 17
Bulge [M/H] N u m be r o f s i m u l a t i on s Disk [M/H] N u m be r o f s i m u l a t i on s −0.35 −0.30 −0.25 −0.20 Figure 5.
Age ( top panels ) and metallicity ( bottom panels ) distributions of bulge (red) and disk (blue) of the 9091700038galaxy. The filled black line represents
Age bulge/disk and [
M/H ] bulge/disk , the dashed black lines are the associated uncertaintiesfrom the 16th and 84th percentiles. two separate 2D bulge and disk monochromaticimages. Finally, stacking all the 2D bulge and diskmonochromatic images we obtain two separate 3Dblue/red datacubes for the two components.6. We build the separate 1D bulge/disk spectra foreach galaxy. Using the 3D bulge/disk datacubeswe sum the flux from all the spaxels to obtain theflux as a function of the wavelength. Figure 7shows the 1D spectra of the galaxy, bulge and diskfor the 9091700038 and 9091700076 galaxies.7. We use pPXF to fit the 1D bulge/disk spec-tra separately in order to estimate the mass-weighted single-age/metallicity of the two compo- nents. We follow the same procedure described inSection 3.1.4, using the 1000 Monte Carlo simula-tions to estimate the 16th and 84th percentiles asuncertainties on the ages and metallicities. We ap-proximate the bulge/disk noise performing a firstfit of the bulge/disk spectrum using the galaxynoise. We consider the residuals binned in wave-length and we estimate the dispersion around themean of each wavelength bin. Then, we interpo-late the bulge/disk noise for the whole wavelengthrange.181/192 double-component galaxies have been reliablydecomposed with this method. The results are based on4 Barsanti et al.
Figure 6.
Mass-weighted age ( left upper panels ) and metallicity ( left lower panels ) maps for the 9091700076 galaxy, representingthe data ( left ), model ( middle ) and residuals ( right ). For one spaxel of each spatial bin
Age bin and [
M/H ] bin are linearly fittedfor Age bulge/disk and [
M/H ] bulge/disk weighted for f bulgeM ∗ ,bin and f diskM ∗ ,bin ( right panels ). This galaxy has a bulge older and moremetal-rich than the disk, oppositely to the 909170038 galaxy that shows a bulge younger than the disk in Figure 2. The residualsagree within 3 σ for age and within 2 σ for metallicity. the collapsed information from the 1D SAMI aperturespectra. This method is based on fewer assumptionsthan the one based on M ∗ weights, however it requireshigh signal-to-noise ratio in each wavelength slice andhigh physical spatial resolution which are strictly lim-ited for the SAMI Galaxy Survey (S´anchez et al. 2017;M´endez-Abreu et al. 2019a). Since the SAMI spectra arenot characterized by high signal-to-noise ratio in eachwavelength slice, some systematic noise is generated andit can be seen in Figure 7 as the upturn at the blue endsof the disk spectrum. We limit the systematic effectsby excluding the initial blue upturns in the pPXF fitsat step (7). Figure 8 shows the 1D bulge/disk spectrafitted with pPXF for the 9091700038 galaxy. The ob-tained results are ( Age, [ M/H ]) bulge =(8.71 Gyr, 0.20)and ( Age, [ M/H ]) disk =(14.44 Gyr, − Effect of flux calibration
The visual inspection of the 1D bulge and disk spectrain Figure 7 for the 9091700038 galaxy suggests that theage of the disk is younger compared to the age of thebulge. The disk spectrum appears very blue with thered flux being lower than the blue flux. However, de-spite the shape of the spectra, we measure a disk olderthan the bulge in Figure 8. This might be due to thedifference in normalisation of the blue and red portionsof the spectrum introduced in the flux calibration duringthe SAMI data reduction process.We investigate that the results are not affected by thedifference in normalisation of the blue and red spectralportions. To this end, we perform the pPXF fitting pro-cedure for the flux weights method by using only theblue portion of the SAMI galaxy spectra and exclud-
AMI: bulge and disk stellar populations Wavelength ( Å ) T o t a l F l u x ( e r g s c m Å ) Figure 7.
1D spectra of the whole galaxy (black), the bulge (red) and the disk (blue) for the 9091700038 galaxy ( left panel )and the 9091700076 galaxy ( right panel ), respectively.
Figure 8.
1D bulge/disk spectra for the 9091700038 galaxy built with the method based on flux weights and fitted with pPXF. ing the red portion. Then, we compare these results forthe 181 galaxies with those previously obtained by fit-ting the full blue+red SAMI spectra. This comparisongave us an idea of the systematics due to the wavelengthrange fitted by pPXF.Figure 9 shows the distributions of the differences be-tween the results obtained by fitting only the blue spec-tra and by fitting the full spectra. The distributionsare Gaussian with medians centred at 0 dex, showingthat the fitted wavelength range does not affect the re-sults. The few largest offsets are measured for galaxiesfor which the result from the blue spectrum fitting isclose to the upper/lower fitting boundaries comparedto the one from the full spectrum which is not. Theaverage standard deviations on the age and metallicityare 0.08 dex for the bulges and 0.13 dex for the disks. Since the fittings of the blue and full spectra containsubstantial overlap in the wavelength range, these stan-dard deviations are not purely systematic. They containcontributions from both the random errors on the resultsobtained by the blue spectra and the full spectra. Therandom errors are estimated using the 1000 Monte Carlosimulations in the pPXF fitting process. The averagerandom errors on the bulge and disk results from thefull spectra are 0.03 dex and 0.06 dex, respectively. Theaverage random errors from the blue spectra are 0.06 dexfor bulges and 0.09 dex for disks. Larger errors are mea-sured for the disks with respect to the bulges since thethe disk results are harder to constrain. The random er-rors from the blue spectra are larger compared to thosefrom the full spectra since a narrow wavelength rangeis fitted. The standard deviations of the difference dis-6
Barsanti et al.
Figure 9.
Differences between the bulge/disk ages and metallicities obtained fitting the 1D blue spectra and the 1D full spectrawith the flux weights method. The distributions are Gaussian with medians ∼ tributions are consistent with the random errors. Thisimplies that the systematic effect due to the differentfitted wavelength ranges is negligible.The blue spectrum contains the typical absorptionlines used to estimate the stellar population proper-ties. The inclusion of the red portion is used to ob-tain better fits of the stellar kinematic components, thegas emission lines and the extinction correction curve.In particular, the fitting for the extinction correctioncurve is less constrained without the red arm data (Scottet al. 2017). Overall, the results obtained with the fluxweights method by fitting the full spectra are not af-fected by the chosen wavelength range and by the dif-ference in normalisation of the blue and red portions ofthe spectrum introduced in the flux calibration.3.3. Method based on radial separation
We perform a method similar to that of Fraser-McKelvie et al. (2018) to estimate individual mass-weighted single-age/metallicity of the bulge and thedisk. Fraser-McKelvie et al. (2018) studied 279 S0 fieldgalaxies of the integral-field MaNGA survey (Bundyet al. 2015). They separated the MaNGA datacubeinto a bulge region considering all the spaxels withinone bulge effective radius and a disk region outside twobulge effective radii. They used these spectra to mea-sure light-weighted ages and metallicities derived fromLick indices. We apply the kinematic corrections to the SAMI spec-tra as described in Section 3.1.2. We do not separate thebulge and disk regions according to the bulge effectiveradius like in Fraser-McKelvie et al. (2018), but we spa-tially bin the SAMI spectra as described in Section 3.1.1and we make use of the estimated bulge-disk contribu-tions in mass. For the bulge spectrum, we select thecentral-most annular bin, which has the highest fractionof light coming from the bulge. The representative 1Ddisk spectrum is the one from the outermost spatial binwith the highest stellar mass contribution from the disk.To obtain reliable results and avoid contamination fromthe other component, we consider only representativebins of the bulge (disk) with S/N >
20 and with a con-tribution in galaxy M ∗ from the disk (bulge) of at least60%. Applying these criteria to galaxy 9091700038, the upper panel of Figure 4 represents the 1D bulge spec-trum, while the second to last represents the 1D diskspectrum. We use pPXF and the same procedure de-scribed in Section 3.1.4 to fit the 1D separated spectrafor mass-weighted ages and metallicities with the 16thand 84th percentiles as uncertainties derived from the1000 Monte Carlo simulations.Using this method, the bulge-disk properties of only54/192 galaxies have been reliably estimated, sincefor the remaining 138 galaxies the selection criteria ofbulge/disk bins with f bulge/diskM ∗ ,bin > . >
20 arenot satisfied. This is an empirical method which allows
AMI: bulge and disk stellar populations RESULTSOur aim is to study the stellar population propertiesof the bulge and the disk in order to understand theformation of S0 galaxies in the cluster environment. Wefocus on disentangling the age-metallicity degeneracy in-herent in analyses using only colors. In this Section, wecompare the results from three different methods andwe analyze the differences between the bulge and diskstellar population properties.4.1. g − i colors Since we are interested in studying the stellar popula-tion properties of the bulge and the disk, we explorethe separate g − i colors of the two components forthe 192 SAMI double-component galaxies. The bulgeand disk g − i colors are estimated from the 2D pho-tometric bulge-disk decomposition of Barsanti et al. inpreparation, where both the apparent magnitudes havebeen corrected for the extinction of the Milky Way us-ing the dust maps of Schlegel et al. (1998) and for theK-correction (Blanton & Roweis 2007).We observe that 73 ±
3% of the galaxy sample havebulges on average 1.4 times redder compared to theirsurrounding disks. Figure 10 shows the g − i color dis-tributions for the bulges and the disks. The disk colordistribution in blue is shifted towards bluer colors com-pared to the bulge color distribution. The median g − i offset separating the two distributions is 0.12 ± Comparison between the three methods
We report the one-to-one galaxy comparison for thebulge/disk mass-weighted single-age/metallicity resultsfrom the three different methods discussed in Sec-tions 3.1, 3.2 and 3.3.Figure 11 shows the comparison between the agesand metallicities of the bulges and disks estimated withthe methods based on mass and flux weights describedin Sections 3.1 and 3.2, respectively. 181/192 SAMIdouble-component cluster members are considered since181 galaxies could be decomposed with the methodbased on flux weights. The values are clustered alongthe bisector, with those characterized by a lower contri-bution to the galaxy stellar mass from the components f M ∗ showing the highest deviation since they are harder g i N u m b e r o f g a l a x i e s BulgeDisk
Figure 10.
Histogram in g − i color for the bulges (red)and the disks (blue) of the 192 SAMI cluster galaxies. Thedashed lines represent the median values. Bulges have reddercolors than disks. to constrain. The galaxies with pegged bulge/disk agevalues have also the smallest f bulge/diskM ∗ contributions,in agreement with the outcomes of Appendix C. Thismight suggest that for the purpose of stellar populationstudy a more stringent cut on B/T to select double-component galaxies than 0 . < B/T < . f bulge/diskM ∗ ,bin > . f bulge/diskM ∗ ,bin < .
6. Forgalaxies with f bulge/diskM ∗ ,bin > . ∼ ∼ . f bulge/diskM ∗ ,bin < . ∼ . >
20 and without anycut on f M ∗ . In Figure 13 the values are clustered alongthe bisector indicating agreement between the results.The largest offsets are mainly due to galaxies with low f bulge/diskM ∗ ,bin . In Figure 14 galaxies with f bulge/diskM ∗ ,bin > . ∼ Barsanti et al. and standard deviations are measured for galaxies with f bulge/diskM ∗ ,bin < .
6, indicating poorer agreement. This jus-tifies our choice in Section 3.3 to consider only galaxieswith f bulge/diskM ∗ ,bin > . f bulge/diskM ∗ ,bin > . f bulge/diskM ∗ ,bin < .
6. They show similar trends as in Figure 14 for thecomparison between the mass weights and radial separa-tion method. Largest medians and standard deviationsare measured for galaxies with f bulge/diskM ∗ ,bin < .
6, whichare harder to constrain due to the contamination fromthe bulge/disk flux.Overall, the best agreement is between the mass andflux weights methods, having the lowest medians andstandard deviations for the difference distributions oftheir results. The comparisons between the radial sep-aration method and the other methods is similar, high-lighting the empirical limits of the radial separationmethod where the bulge/disk results are contaminatedby the flux of the other component and are spatiallylimited. For all the comparisons the highest deviationsbetween the results are for galaxies with low f bulge/diskM ∗ ,bin values. This is particularly evident for the radial separa-tion method, justifying our choice to select only galaxieswith f bulge/diskM ∗ ,bin > . Comparing the stellar populations of bulges anddisks
Our aim is to explore the formation of S0 galaxiesin clusters. Different scenarios imprint different fossilrecords in the stellar population properties of the bulgesand the disks. We analyze the the best-fitting mass-weighted ages and metallicities of the bulges and thedisks for the 192 double-component SAMI galaxies ob-tained with the three methods described in Sections 3.1,3.2 and 3.3. Table 1 reports the method, the number ofanalyzed galaxies, the percentages of galaxies with older,younger, more metal-rich, more metal-poor, redder andbluer bulges with respect to the disks. The age, metal-licity and ( g − i ) results for the three methods are in agreement. The analysis of the age reveals that ∼ ∼ M ∗ weights and flux weights methods shows that ∼
30% of the galaxies have a bulge older or more-metal-rich than the disk according to the flux weights method,but they show an opposite result for the M ∗ weightsmethod. However, as shown in Figure 11 the measure-ments are close the one-to-one relationship. The remain-ing ∼
70% shows the same age and metallicity propertiesaccording to both methods. For the comparison betweenthe flux weights and the radial separation methods wefind the same outcome. The number of galaxies thatdiffer is ∼
20% between the M ∗ weights and the radialseparation methods. Thus, on average the percentagesin Table 1 are characterized for the 75% by galaxies thatshow the same bulge and disk stellar population prop-erties according to the three methods.We investigate the significance of the age andmetallicity differences between the two componentswith respect to their random uncertainties. Forthe M ∗ weights method the random errors on the(log( Age ) , [ M/H ]) bulge/disk results are estimated as thestandard deviations of the (log( Age ) , [ M/H ]) bulge/disk distributions using a Monte Carlo approach, as de-scribed in Section 3.1.5. In case of pegged solu-tion and consequent zero standard deviation, we as-sociate to that result the average standard deviationfrom all the measurements. For the flux weightsand radial separation methods the random errors on(log( Age ) , [ M/H ]) bulge/disk are estimated as the stan-dard deviations of 1000 Monte Carlo simulations, as de-scribed in Sections 3.2 and 3.3, respectively. The meanrandom errors for the bulge results are 0.03 dex, 0.03dex and 0.02 dex for the M ∗ weights, flux weights andradial separation methods, respectively. The mean ran-dom errors for the disk results are 0.10 dex, 0.06 dexand 0.04 dex for the M ∗ weights, flux weights and radialseparation methods, respectively. We assess the statisti-cal significance of the differences in age and metallicitybetween bulges and disks by comparing the differenceto the quadrature sum of the measured bulge and diskrandom uncertainties. Results are significantly differentat the 3 σ level.Table 2 lists for each method the percentages ofgalaxies with bulges significantly (at the 3 σ level) AMI: bulge and disk stellar populations Figure 11.
Comparison for the age and metallicity of the bulge ( upper and lower left ) and of the disk ( upper and lower right ).The x and y axis represent the results from the mass and flux weights methods, respectively. Points with partially missingerrors are due to the 16th and 84th percentiles matching with the result. The dotted black line represents the bisector. Pointsare color-coded according to the highest contribution per bin to M ∗ from the bulge and the disk. The results clustered alongthe bisector suggest agreement between the two methods. The largest offsets are for galaxies with f bulge/diskM ∗ ,bin < . Figure 12.
Distributions of the differences for the bulge/disk ages and metallicities estimated with the mass and flux weightsmethods. Galaxies with f bulge/diskM ∗ ,bin > . < f bulge/diskM ∗ ,bin < .
6, driving the largest offsets for the whole galaxy population. Barsanti et al.
Figure 13.
Comparison for the age and metallicity of the bulge ( upper and lower left ) and of the disk ( upper and lower right ).The x and y axis represent the results from the mass wights and radial separation methods, respectively. The dotted black linerepresents the bisector. Points are color-coded according to the contribution in the galaxy stellar mass from the bulge and thedisk and are group along the bisector, indicating agreement between the two methods. The largest offsets are for galaxies with f bulge/diskM ∗ ,bin < . Figure 14.
Distributions of the differences for the bulge/disk ages and metallicities estimated with the methods based on massweights and radial separation. Galaxies with f bulge/diskM ∗ ,bin > . < f bulge/diskM ∗ ,bin < . AMI: bulge and disk stellar populations Figure 15.
Comparison for the age and metallicity of the bulge ( upper and lower left ) and of the disk ( upper and lower right ).The x and y axis represent the results from the flux wights and radial separation methods, respectively. The dotted black linerepresents the bisector. Points are color-coded according to the contribution in the galaxy stellar mass from the bulge and thedisk and are group along the bisector, indicating agreement between the two methods. The largest offsets are for galaxies with f bulge/diskM ∗ ,bin < . Figure 16.
Distributions of the differences for the bulge/disk ages and metallicities estimated with the methods based on fluxweights and radial separation. Galaxies with f bulge/diskM ∗ ,bin > . < f bulge/diskM ∗ ,bin < . Barsanti et al. older/younger/similar in age and significantly moremetal-rich/more metal-poor/similar in metallicity withrespect to the disks. We observe that on average ∼ ∼
34% of galaxies the bulge is significantlyyounger than the disk. For the remaining ∼
43% ofgalaxies the differences in age are not significant withrespect to their random uncertainties. The analysis ofthe metallicity shows that most galaxies, ∼ ∼
7% of the galaxies have a bulge significantly moremetal-poor than the disk. The remaining ∼
31% ofgalaxies show a metallicity for the bulge consistent withthat of the disk. Overall, we find that most bulges aresignificantly more metal-rich than the disks, while theycan be either significantly older or younger than thedisks. According to the methods based on mass andflux weights, the bulges are on average ∼ ∼ g − i )colors estimated using the decomposed photometry, andthose estimated using the closest SSP templates to theage/metallicity determined for the bulge and disk. Inthis paper the ( g − i ) colors are derived from the 2Dbulge-disk decomposition. However, we observe thesame percentages of redder bulges for the three methodsalso when the ( g − i ) colors of the bulges and the disksare estimated using their age and metallicity. For eachgalaxy component we associate its estimated age andmetallicity from Section 3 with the MILES SSP tem-plate characterized by the closest values. Each MILESSSP template has a predicted ( g − i ) color (Vazdekiset al. 1996). We conclude that for the three methodsbulges tend to be redder and more metal-rich comparedto the disks, but we do not see a clear trend with age.4.4. Comparing the stellar population ages of bulgesand disks
In order to disentangle galaxies where the bulge agediffers significantly from the disk age, we analyze theage-metallicity plots for the three methods in Figure 17.Bulges are represented by red dots, the disks by blueellipses and the grey/magenta lines connect bulge anddisk of the same galaxy. The left panels show the galax-ies where the bulge is significantly older than the disk,while the right panels show the galaxies where the bulgeis significantly younger than the disk counterpart. Thepegged solutions at the upper/lower fit limits might becaused by the assumption of single-age/metallicity in-stead of radial gradients as reported in Section 3.1.5.Moreover, as shown in Appendix C, pegged solutions depend on the bulge-to-total flux ratio of the galaxyand they are generated for low M ∗ weights. For boththe galaxy populations with significantly younger/olderbulge than the disk, bulges are mainly more metal-richwith respect to the disks.Table 3 lists the method, the number of galaxies withbulges significantly older/younger than the disks andthe percentages of bulges significantly more metal-rich,more metal-poor, similar in metallicity, redder and bluerwhen compared to the disks. Combining the results ofthe three methods, we find that 46% of the galaxies withbulges that are significantly older than disks also havebulges that are significantly more metal-rich than theirdisks. In these cases, the bulge is on average 2.3 timesmore metal-rich than the disk. For the 43% of thesegalaxies the bulge and the disk have similar metallicity.For 70% of the galaxies the bulge is also redder than thedisk by a factor of ∼ . ∼ ∼ ∼ .
2. Thus, our results indicate that, re-gardless of bulge and disk age, in a majority of casesbulges are both ∼ ∼ ∼
70% of the galaxies showingthe same stellar population properties for the bulges andthe disks for all the methods.We investigate whether galaxies with significantlyolder or younger bulges than the disks trace differentformation scenarios for S0 galaxies. To this end, we ex-plore their dependence on the properties of the galaxy,the galaxy components and the cluster environment.We study the galaxy stellar mass and the B/T distri-butions of the two populations in age. These two pa-rameters trace possible different in situ formation sce-narios. We study the bulge and disk properties, suchas g -band magnitude, S´ersic index, effective radius andstellar mass. These parameters help us to understandif bulges significantly older or younger than the diskshave different features. Finally, we study the projecteddistance from the cluster centre normalized by R andthe galaxy density. Using these environment metrics,we can assess whether the two galaxy populations inage depend on their environment. The galaxy density ismeasured as the fifth nearest neighbour surface densityΣ by Brough et al. (2013). To check for significant dif-ferences between the two galaxy populations we applythe Anderson − Darling test (Stephens 1974). This teststhe null hypothesis that the two galaxy populations are
AMI: bulge and disk stellar populations Table 1.
Differences in mass-weighted ages, metallicities and ( g − i ) colors between bulges and disks from the three methods.Column 1 lists the method, column 2 the number of analyzed galaxies, columns 3/4/5/6/7/8 the percentages of galaxies withbulges older/younger/more metal-rich/more metal-poor/redder/bluer than the disks. The last line lists the average values forthe three methods.Method N g %(Age) B > D %(Age) B < D %[M/H] B > D %[M/H] B < D %( g − i ) B > D %( g − i ) B < D M ∗ weights 192 45 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Table 2.
Significant differences in bulge/disk mass-weighted ages and metallicities for the three methods. Column 1 lists liststhe method, column 2 the number of analyzed galaxies, columns 3/4/5 the percentages of galaxies with bulges significantlyolder/younger/equal than the disks, and columns 6/7/8 the percentages of galaxies with bulges significantly more metal-rich/metal-poor/equal than the disks. The last line lists the average values for the three methods.Method N g %(Age) B (cid:29) D %(Age) B (cid:28)
D %(Age) B ∼ D %[M/H] B (cid:29)
D %[M/H] B (cid:28)
D %[M/H] B ∼ D M ∗ weights 192 26 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± drawn from the same parent distribution, and it probesthe differences in the distribution tails. We consider theresults only for the mass and flux weights methods, sincethe radial separation method is reliable for only a lownumber of galaxies.For both the methods, we observe that there is nota significant difference between the galaxy populationswith bulges older or younger than the disks as a func-tion of the galaxy, bulge/disk and environment proper-ties. Figure 18 shows for the galaxy populations withbulges significantly older or younger than the disks thedistributions in M ∗ , n bulge , R/R and galaxy density.We plot the results only from the mass weights method,since they are in agreement with those from the fluxweights method. If we exclude those galaxies where themetallicity difference between the bulge and the disks isnot significant, we find consistent results.4.5.
Breaking the age-metallicity degeneracy
As shown in Figure 10, the majority ( ∼ ∼ ∼
23% of redder bulges are also significantly older thanthe disks. We conclude that the redder color in bulgesis mainly driven by their metallicity. Consistent resultsare obtained if the ( g − i ) colors are those predictedfor the closest MILES SSP templates associated withthe bulge/disk age and metallicity, instead of the valuesfrom the 2D bulge-disk decomposition.To explore the correlation between colors, age andmetallicity Figure 19 shows for the three methods thedifference in the g − i color ∆( g − i ) = ( g − i ) bulge − ( g − i ) disk as a function of the metallicity difference∆[ M/H ] = [
M/H ] bulge − [ M/H ] disk ( left panel ) and agedifference ∆( Age ) =
Age bulge − Age disk ( right panel ).The points are color-coded according to the age andmetallicity difference, respectively. Redder points corre-spond to bulges that are older or more metal-rich thanthe disks. Larger points correspond to significant differ-ences in age or metallicity. In agreement for the threemethods, galaxies with bulges redder than the disksare mainly more metal-rich since most of them have∆[ M/H ] >
0. For the ∆( g − i ) versus ∆( Age ) plotswe do not observe any particular trend as a function of∆(
Age ), with galaxies assuming either ∆(
Age ) > Age ) < g − i ) versus∆( Age ) plots. If the metallicity is the main driver for4
Barsanti et al.
Figure 17.
Age-metallicity plots for the three methods ( top to bottom panels ) where galaxies have the bulge significantly older( left panels ) and younger ( right panels ) than the disk. The mean errors on the bulge and disk properties are plotted on thebottom left. In both galaxy populations, bulges are mainly more metal-rich compared to the disks, in agreement for the threemethods.
AMI: bulge and disk stellar populations Table 3.
Metallicities and ( g − i ) colors for bulges significantly older or younger than the disks from the three methods. Column1 lists the method, column 2 the population with bulges significantly older/younger than the disks, column 3 the number ofgalaxies, columns 4/5/6 the percentages of bulges significantly more metal-rich/metal-poor/equal than the disks, and columns7/8 the percentages of bulges redder/bluer than the disks.Method %(Age) N g %[M/H] B (cid:29) D %[M/H] B (cid:28)
D %[M/H] B ∼ D %( g − i ) B > D %( g − i ) B < D M ∗ weights B (cid:29) D 50 56 ± ± ± ± ± M ∗ weights B (cid:28) D 69 64 ± ± ± ± ± (cid:29) D 33 39 ± ± ± ± ± (cid:28) D 66 92 ± ± ± ± ± (cid:29) D 14 43 ±
12 7 ±
13 50 ±
13 57 ±
12 43 ± (cid:28) D 16 69 ±
11 0 ± ±
11 69 ±
11 31 ± Figure 18.
Histograms for the galaxies with a bulge significantly older (grey color) or younger (magenta color) than the diskas a function of the galaxy stellar mass ( top left panel ), bulge S´ersic index ( top right panel ), projected cluster-centric distance( bottom left panel ) and galaxy density ( bottom right panel ). The dashed lines represent the median values. According to theAnderson − Darling test, we measure no significant differences. Barsanti et al.
Table 4.
Mass-weighted ages and metallicities for redder bulges from the three methods. Column 1 lists the method, column 2the number of galaxies with redder bulges than the disks, columns 3/4/5/6 the percentages of galaxies with bulges significantlyolder/younger/more metal-rich/more metal-poor than the disks.Method N g %( g − i ) B > D %(Age) B (cid:29)
D %(Age) B (cid:28)
D %[M/H] B (cid:29)
D %[M/H] B (cid:28) D M ∗ weights 141/192 27 ± ± ± ± ± ± ± ± ± ± ± ± bulges being redder than the disks, we expect a positiveslope in the ∆( g − i ) versus ∆[ M/H ] plots. Consideringall the bulges, the slopes of the best-fitting lines gave ushints on what we expect in terms of trends. Especiallyfor the M ∗ weights method, we observe a stronger pos-itive trend for the ∆( g − i ) versus ∆[ M/H ] comparedto the slope for the ∆( g − i ) versus ∆( Age ). However,considering the uncertainties all the best-fitting lines arenot statistically significant and they are consistent withflat trends. DISCUSSIONIn this Section we discuss the limitations and strengthsof the three methods based on M ∗ weights, flux weightsand radial separation implemented in Section 3.1, 3.2and 3.3, respectively. We compare our results with theprevious findings from the literature. We discuss theirinterpretations in light of simulations and physical pro-cesses. Finally, we address the caveats of this study.5.1. Suggestions for the usage of the methods
We explore three methods to estimate mass-weightedsingle-age/metallicity of the bulges and the disks. Themethod based on M ∗ weights has several limitations andrequires high mass weights from both the galaxy com-ponents to constrain their solutions. The mass weightsdepend on the galaxy B/T but also on how well thedata are able to spatially recover the two components.This method allows for more complicated models to betested when compared with the linear model exploredin this paper, adding radial and/or time gradients tothe age/metallicity of the bulge and the disk. Themethod based on flux weights uses fewer assumptions,but it requires high signal-to-noise ratio at each wave-length range to produce spectra not contaminated bysystematic noise. The separation between the bulge anddisk contributions is most contaminated for the methodbased on radial separation, reducing the analysis to bevalid for a small galaxy sample. However, this methodis the most simple method to be implemented and itallows to obtain similar outcomes.In conclusion, the use of the M ∗ and flux weightsmethods is preferable, especially foreseeing large IFUsurveys observed by spectrographs such as Hector (Bryant et al. 2018). These surveys will be characterizedby high spectral resolution in the blue arm and high S/Nat each wavelength, allowing the methods to overcomesome of the previous data-dependent limitations.5.2. Comparison with previous works
The results for the bulge/disk g − i colors in Section 4.1are in agreement with previous studies on the clusterenvironment, finding most bulges being redder than thedisks (Hudson et al. 2010; Head et al. 2014). Head et al.(2014) studied the colors of the bulges and the disks of S0galaxies in the Coma cluster. They observed bulges withredder colors and a median g − i offset of 0.097 ± ∼
30% consideringsignificant differences) of the whole galaxy sample showssuch stellar population properties. For galaxies with abulge younger than the disk there is the possibility thatthe component identified as the bulge using the pho-tometric bulge-disk decomposition, either in 1D like inJohnston et al. (2014) or in 2D like in Sil’chenko et al.
AMI: bulge and disk stellar populations Figure 19. ∆( g − i ) = ( g − i ) bulge − ( g − i ) disk versus ∆[ M/H ] = [
M/H ] bulge − [ M/H ] disk ( left panel ) and ∆( Age ) =
Age bulge − Age disk ( right panel ) for the three methods. Points are color-coded according to the age and metallicity difference,respectively. Larger points correspond to significant differences in age or metallicity. The dashed black line marks ∆[ M/H ] = 0and ∆(
Age ) = 0, respectively. Redder bulges are mainly more metal-rich than the disk, while they cover all the age values. Thebest-fitting line is filled black and estimated with the Hyper-Fit software (Robotham & Obreschkow 2015). The dotted blacklines mark the 1 σ interval. Flat trends are observed for ∆( g − i ) versus ∆[ M/H ] and versus ∆(
Age ). Barsanti et al. (2012) and in this study, is contaminated by the light ofthe central part of the disk characterized by recent starformation. To explore this possibility a more sophisti-cated combined photometric and kinematics bulge-diskdecomposition would reveal more information. For theCALIFA survey, M´endez-Abreu et al. (2019b) foundthat star formation occurs only in the disk, not in thebulge, and that it is not limited to the outer disk regions,but it also occurs in the central regions dominated bythe bulge light. For the ATLAS survey, early-typegalaxies with younger and more metal-rich stellar pop-ulations in the core regions with respect to the outskirthave also been observed (Krajnovi´c et al. 2013; McDer-mid et al. 2015; Krajnovi´c et al. 2020), in agreementwith our outcome.We do not observe a trend with galaxy stellar massin Section 4.4, unlike Fraser-McKelvie et al. (2018) whoobserved two different population for 279 low- and high-mass S0 field galaxies in the MaNGA survey. Fraser-McKelvie et al. (2018) measured light-weighted averagesof Lick indices and translated them into age/metallicityby bi-linearly interpolating the model lines of the stellarpopulation templates. They found that galaxies with M ∗ /M (cid:12) > are characterized by older bulge andyounger disk, while galaxies with M ∗ /M (cid:12) < mainlyhave younger bulge and older disk. In this work we con-sider only galaxies with M ∗ /M (cid:12) > due to signal-to-noise ratio limitations on SAMI data (Owers et al.2019). The percentage of high-mass galaxies with olderbulge is 85% in the field from Fraser-McKelvie et al.(2018). This percentage drops to 23% in the cluster en-vironment from this study, considering statistically sig-nificant results. This comparison suggests that theremay be differences in the formation mechanisms of S0galaxies in the field with respect to the cluster envi-ronment. This conclusion is supported by the work ofCoccato et al. (2019), who compared the kinematics ofcluster and field S0s finding differences between the twopopulations and suggesting different formation processesin different environments.5.3. Physical interpretations
To explain bulges being redder than disks, Head et al.(2014) speculated that bulges need to be either ∼ ∼ ∼ Caveats
We address the three caveats that mainly character-ized this study: (i) the definitions of “bulge” and “disk”,(ii) the estimate of single-ages/metallicities and (iii) theinterpretation of the age/metallicity measurements.We assume that galaxies are characterized by only twocomponents: a central bulge and a surrounding disk.We use a purely photometric approach to separate thecomponents. However, we need to keep in mind thatthese definitions are limited to an extended “disk” anda centrally-concentrated “bulge”, and they might notrespect the entire complexity of the galaxy components(Head et al. 2015; Fischer et al. 2019). In particular,the fit of the exponential component is dominated bythe outer regions of the galaxy and its extrapolationat the centre can only capture a limited amount of thecomplex central structure that might be associated withthat disk (M´endez-Abreu et al. 2019b). The recent workof Breda et al. (2020) shows insights about the invalid-ity of the standard exponential disk profile within thegalaxy central region, observing a down-bending trendbeneath the bulge radius for a large fraction of late-typegalaxies. Assigning all the left-over light to a “bulge”is how we often proceed, but we may need to balance
AMI: bulge and disk stellar populations σ level. This largediscrepancy for some bins is consistent with the expec-tation that radial gradients exist within the galaxy, butthe proposed linear model is only able to reproduce thedata using single Age bulge/disk and [
M/H ] bulge/disk val-ues. For a better agreement, more complicated versionsof the Equations (1) and (2) that take into account ra-dial gradients for the age/metallicity of the bulge/diskshould be fitted.In regards to the interpretation of the age/metallicitymeasurements, we make use of the MILES SSP tem-plates in Section 3.1.4. These SSP models make predic-tions up to ∼ ± >
14 Gyr, finding that they are consistent with theage of the Universe once observational uncertainties inthe data and systematic uncertainties in the models areconsidered (see their Appendix D). Moreover, as statedby Johnston et al. (2014), different SSP models lead todifferent absolute measurements of age/metallicity be-cause of the uncertainties that characterized the stel-lar astrophysics into these models. In this context, theage/metallicity measurements of the bulge and the diskshould be interpreted in a relative sense and consideredas constraints on the different stellar populations ratherthan the absolute results. In our analysis we excludeage/metallicity differences between the galaxy compo-nents that are not statistically significant compared totheir random errors. These differences highlight possiblelimitations in the used data or methods, since we are notable to discern between the age/metallicity of the bulgeand the disk. The use of photometric and spectroscopicdata with higher S/N and spatial resolution as well as a larger galaxy sample might help us to obtain morestatistically significant results. SUMMARY AND CONCLUSIONSWe explore the stellar population properties of thebulge and disk components in cluster galaxies. Specif-ically, we investigate their colors, mass-weighted agesand metallicities separately. We study the SAMI clustersample characterized by eight low-redshift clusters. Thissample allows us to combine photometric with spectro-scopic information and to conduct a statistical study onthe bulge and disk stellar populations.The characterization of the double-component galaxysample is based on the 2D photometric bulge-disk de-composition performed by Barsanti et al. in preparationfor the SAMI clusters. They use a Bayesian approachand the packages
ProFound for source detection and
ProFit for galaxy light modelling. From the double-component sample of 469 galaxies within 2 . R iden-tified by Barsanti et al. in preparation, we select onlyprimary SAMI targets with R < R , M ∗ /M (cid:12) > and σ <
400 km s − . We find 192 double-componentgalaxies, where the bulge + disk components are pref-erentially modelled by the S´ersic + exponential profile.These double-component galaxies are mainly character-ized by fast rotators, passive spectra and S0 morphology.To measure mass-weighted single-age and single-metallicity for the bulges and the disks we investigatethree methods:1. We develop a new method based on stellar massweights. The use of the 2D bulge-disk decompo-sition performed in the r , g and i -bands allows toshift from light to mass. We estimate the spatially-resolved contributions separately from the bulgeand the disk to the galaxy stellar mass. Wespatially bin and kinematically correct the SAMIgalaxy spectra. For each spatial bin the bulge-disk mass fractions are used as weights on themass-weighted age and metallicity of the galaxy,derived using full spectral fitting. Finally, mass-weighted single-age and single-metallicity valuesare obtained for the bulge and the disk of eachgalaxy.2. We test a method similar to that of M´endez-Abreuet al. (2019a) based on flux weights for the twocomponents. Separate 1D SAMI aperture spectraare obtained for the bulge and the disk accordingto their spatially-resolved bulge-to-total and disk-to-total flux ratios.3. We test a method similar to that of Fraser-McKelvie et al. (2018) based on radial separation.0 Barsanti et al.
Separate 1D bulge and disk spectra are obtainedconsidering the galaxy spectra from the most cen-tral and outermost bins, respectively. To avoid thecontamination of the flux from the other compo-nents, representative bins of the bulge (disk) areselected having a contribution in galaxy M ∗ fromthe bulge (disk) of at least 60%.Comparing the three methods, we observe that the M ∗ and flux weights methods show the best agreement. Thehighest deviations between the results are for galaxieswith low f bulge/diskM ∗ ,bin values. This is particularly evidentfor the radial separation method. The implementationof the M ∗ and flux weights methods is preferable, es-pecially foreseeing large IFU surveys characterized byhigher S/N and higher spectral resolution in the bluearm than the SAMI data (e.g., Hector, Bryant et al.2018).Our results can be summarised as follows.1. Bulges are mainly redder than the disks. The me-dian g − i offset separating the color distributionsof the bulges and the disks is 0.12 ± ∼ ∼
34% have significantlyyounger bulges. For the remaining ∼
43% of galax-ies the differences in age are not significant withrespect to their random uncertainties. We find ∼
62% of bulges on average ∼ ∼
7% ofthe galaxies have a bulge significantly more metal-poor than the disk.3. According to all three methods, both galaxy pop-ulations with older or younger bulges compared tothe disks show redder and more metal-rich bulges.The two populations do not show significant differ-ences when comparing galaxy, bulge/disk and en-vironment properties. Limitations of this analysisare: the exclusion of galaxies with M ∗ < M (cid:12) and a sample not large enough to perform a reli-able study as a function of environment metrics.4. According to all three methods, redder bulges aremore metal-rich than the disks, whereas they canbe either younger or older than the disks. Ouranalysis suggests that the redder color in bulges ismainly driven by their metallicity.Further analysis is required to understand the roleplayed by the galaxy stellar mass and the environment in affecting the stellar population properties of the bulgeand the disk. A complete study of high- and low-mass S0galaxies in high- and low-galaxy density environments isnecessary to investigate possible differences. A study ofthe projected-phase space would be ideal to shed light onthe possible environmental processes acting in quench-ing star formation activity within the bulge and the disk,but a larger sample of double-component galaxies is re-quired to extrapolate reliable conclusions.Overall, the three different methods tested to estimateage and metallicity of the bulge and the disk suggest thesame results. Bulges are mainly redder and ∼ AMI: bulge and disk stellar populations
Software: fdr package (AAO Software Team2015), astrolibR (Chakraborty et al. 2014), astropy (Astropy Collaboration et al. 2013, 2018), Hyper-Fit(Robotham & Obreschkow 2015), K-corrections (Blan-ton & Roweis 2007), pPXF (Cappellari & Emsellem 2004;Cappellari 2017), ProFit (Robotham et al. 2017),
Pro-Found (Robotham et al. 2018), SAMI python package(Allen et al. 2014).APPENDIX A. REQUIRED SIGNAL-TO-NOISE RATIO FOR SPATIAL BINFor IFU surveys, the outer spaxels where the galaxies are fainter have lower continuum signal-to-noise ratio (S/N)when compared with the brighter central regions. Improving the S/N of the SAMI spectra requires spatial binningso that reliable stellar population parameters can be estimated. However, there is also a trade-off in the amount ofbinning required and the spatial resolution necessary for disentangling bulge and disk properties. Therefore, we wishto determine the minimum S/N required to produce reliable SSP information without the unnecessary loss of spatialinformation. As defined in Section 3.1, S/N is measured taking the median value of the flux divided by the noise inthe rest-frame wavelength range 4600-4800 ˚A .We investigate the accuracy of stellar ages estimated by pPXF in order to find the minimum S/N required for eachspatial bin. We make use of the MILES single stellar population (SSP) templates (Vazdekis et al. 2010). We consider6 metallicities [M/H]=0.22, 0.00, -0.40, -0.71, -1.31, -1.71 and for each metal value we select 10 ages from 1 to 16 Gyr.We change these templates in order to reproduce simulated spectra representing the features of the observed SAMIspectra. Each simulated spectrum is convolved with the typical velocity dispersion of a galaxy σ gal = 200 km s − .We use the formula of Owers et al. (2019), which considers the convolution with the instrument resolution for theblue arm of the spectrograph σ inst = 1 .
13 ˚A (van de Sande et al. 2017a) and the resolution of the MILES templates σ MILES = 1 .
06 ˚A (Falc´on-Barroso et al. 2011). The final wavelength-dependent velocity dispersion is: σ = (cid:16) λσ gal c (cid:17) + σ inst − σ MILES (A1)where λ is the wavelength in ˚A and c is the speed of light in km s − . Then, to make each simulated spectrum similarto the SAMI spectrum, the simulated spectrum is interpolated onto a grid with the same wavelength spacing as in theblue SAMI spectrum and in the same rest-frame wavelength range 3650-7000 ˚A .The initial noise associated with each simulated spectrum is extracted from the central region of the same galaxy asthe square root of the variance, combining the red and blue SAMI data cubes. For each simulated spectrum we considerthe different S/N values=10, 20, 30, 40, 50 scaling the noise every time. For each S/N value we run 100 iterationsadding noise to the simulated spectrum using a random Gaussian array. We use pPXF to fit the simulated spectrum.First, we fit for the velocity and velocity dispersion using 0 and 200 km s − as inputs, considering the MILES library,excluding those pixels with noise=NaN and using a 12th degree additive polynomial to minimize template mismatch.Then, we fix the kinematic components and we fit for mass-weighted ages without considering any gas emission lineand using no multiplicative polynomial and no regularization.For each S/N value, we estimate the mean log(Age) out and the scatter σ out as standard deviation of the output logage distribution and compare these quantities against their input values. Figure 20 shows for the fixed metallicity[M/H]=0.00 the offset log(Age) out − log(Age) in and the scatter σ out normalized by log(Age) in as a function of S/N inthe left and right panels , respectively. As expected, offset and scatter decrease with increasing S/N. Considering the left panel , there is a sharp decline in the offset from S/N=10 to S/N=20. Beyond S/N=20, the decline is less steep.Considering the right panel , σ out / log(Age) in ∼ − ∼
5% and for young ages of ∼ Barsanti et al.
Figure 20.
For the fixed metallicity [M/H]=0.00, the left panel shows the offset log(Age) out − log(Age) in and the right panel shows the scatter around log(Age) out normalized by log(Age) in . Both panels are as a function of S/N values. The S/N=20 ischosen as the minimum required S/N value for spatially binned spaxels. Figure 21.
Scatter for all the considered ages and metallicities for fixed S/N=20. Bluer and bigger the circle, larger is thescatter. The circle sizes range according σ out from 0.015 to 0.255 yr. Younger ages are characterized by a larger scatter comparedto older ages. Our results are in agreement with Ge et al. (2018); younger ages are characterized by a larger offset and scattercompared to older ages. Young stellar populations have strong Balmer lines through which they can easily identified,however they show a larger stellar mass-light ratio compared to old stars, especially at low S/N, which makes the fitharder to constrain. Figure 21 shows the scatter for all the ages and metallicities for fixed S/N=20, where the highestoffset is ∼ σ out =0.08-0.06 yr over the ages, thus we prefer S/N=20to preserve the spatial resolution. We find larger scatters and offsets compared to Ge et al. (2018) due to the factthat we consider mass-weighted ages, which are less directly linked to the light and therefore harder to constrain thanluminosity-weighted ages (see Figure 11 of Ge et al. 2018). We do not test for the shape of the error spectrum andthe level of dust extinction, since Ge et al. (2018) found that pPXF results are nearly independent of these quantities. AMI: bulge and disk stellar populations B. TEST FOR KINEMATIC CORRECTIONSWe use the method based on flux weights described in Section 3.2 to test the accuracy of the kinematic correctionsapplied to the SAMI spectra in Section 3.1.2. 181/192 double-component galaxies have been reliably decomposedwith this method into separate 1D bulge/disk spectra. We apply pPXF to fit for the kinematic components of the 1Dbulge/disk spectra separately and we run only 100 Monte Carlo simulations to estimate the 16th and 84th percentilesas uncertainties, since the kinematic fits are time consuming to compute due to their non-linear determination. Wecompare the results for the two components and implement the same test of Johnston et al. (2014). If the kinematiccorrections are applied correctly, then the 1D spectra of the bulge and the disk are expected to have matched zerovelocity and maximum velocity dispersion measured within the galaxy.Figure 22 displays the comparison between the velocities of the bulge and the disk ( left panel ), and for the velocitydispersions ( right panel ). For most of the galaxies, the bulge and disk velocities are grouped around the bisector withan offset of ±
200 km s − . The bulge and disk velocity dispersions match along the bisector and correspond to themaximum value. The offset and the outliers are due to the low signal-to-noise ratio of the bulge/disk spectra. Theseplots highlight the reliability of the applied kinematic corrections, but the deviations show the reason we fit for thesecomponents in the step (2) of Section 3.1.4, before to estimate the age/metallicity values of the galaxy components. Figure 22.
Comparison between the velocity of the bulge and that of the disk ( left panel ) and for the velocity dispersion ( rightpanel ). The dotted black line represents the bisector. In the right panel points are color-coded according to the maximumvelocity dispersion within the galaxy. For both plots, the points are grouped along the bisectors with few outliers, confirmingthe accuracy of the applied kinematic corrections.C.
TESTING THE METHOD BASED ON M ∗ WEIGHTSThe method based on M ∗ weights described in Section 3.1 estimates single-age/metallicity values of the bulge andthe disk, according to Equations (1) and (2). We implement simulated SAMI galaxy spectra to test this method,comparing input age/metallicity values for the bulge and the disk with the respective outputs. Specifically, we explorehow the results might be affected by low M ∗ weights from the bulge and the disk, i.e. low f bulge/diskM ∗ ,bin values. We considerthree galaxies having bulge-to-total flux ratio B/T=0.3, 0.5, 0.7 and for each galaxy we consider their projected bulgeand disk stellar mass factions maps representing f bulge/diskM ∗ ,bin (similar to the right panels of Figure 2). For each B/Twe use two MILES SSP templates representing the bulge and disk spectra, respectively, and their inputs for theage/metallicity. For each spatial bin in the simulated galaxies, we generate fake spectra by multiplying the bulge/disk4 Barsanti et al. input spectra with the associated f bulge/diskM ∗ ,bin and summing the flux. We make the simulated galaxy spectra similarto the SAMI spectra and characterized by S/N=20, as described in Appendix A. For the noise associated to eachsimulated spectrum we make use of the blue and red SAMI data cubes of the considered galaxy. We extract the noisefrom a spaxel in the respective bin as the square root of the variance.We apply the pPXF steps to find Age bin and [
M/H ] bin as described in Section 3.1.4. We run 1000 Monte Carlosimulations adding noise to each simulated galaxy spectrum per bin using a random Gaussian array. For each simulationwe fit Equations (1) and (2) as in Section 3.1.5 to find ( Age, [ M/H ]) bulge and ( Age, [ M/H ]) disk . The final outputsfor the age/metallicity of bulge/disk are represented by the mean values of the Age bulge/disk and [
M/H ] bulge/disk distributions obtained from the simulations. The scatter associated with the mean values are obtained as the 16thand 84th percentiles of these distributions. For each B/T value, we consider the same 24 input combinations for theage/metallicity of the bulge/disk.Figure 23 shows the age-metallicity plots for B/T=0.3, 0.5, 0.7 ( top to bottom panels ) separating the inputs withages of the bulge older or younger compared to the disk ( left and right panels , respectively). For increasing B/T values,the measured age/metallicity values more closely match the input values: both the offsets (red lines) and the scatterassociated with the outputs (black lines) decrease. On the other hand, considering the age/metallicity disk results, forincreasing B/T values both the offsets (blue lines) and the scatter associated with the outputs (black lines) increaseand some outputs are pegged to the upper or lower fit limits.Pegged solutions for the bulge/disk are obtained for low/high B/T values, i.e. when the bulge/disk results are harderto constrain, respectively. Table 5 lists for each B/T value the associated maximum values for f bulge/diskM ∗ ,bin and the meanoffsets for the bulges and the disks from the 24 input combinations and the respective mean scatter. The bulge/diskoffsets are estimated as the difference between ([ M/H ] , Age ) input and ([ M/H ] , Age ) output for the bulge/disk results,respectively. The bulge/disk mean scatter is estimated as the mean of the scatter associated with ([ M/H ] , Age ) output for the bulge/disk results, respectively. For increasing B/T values and increasing maximum f bulgeM ∗ ,bin , the mean offsetfor the bulges decreases with the outputs being closer to their inputs, while the mean offset for the disks increases.The same trend is observed for the mean scatter. For increasing B/T, the mean scatter for the bulges decreases withthe outputs being more precise, while the mean scatter for the disks increases meaning more spread output solutions.Low f bulge/diskM ∗ ,bin implies that the results are harder to constrain and might lead to pegged results at the lower or upperfit limits. We conclude that this method depends on the contributions from the bulge and the disk to the galaxystellar mass, which might depend on the galaxy B/T but also on how well the data are able to spatially recover thetwo components. Table 5.
Testing the method based on M ∗ weights. Column 1 lists B/T, columns 2/3 list the maximum values for f bulge/diskM ∗ ,bin ,columns 4/5 and columns 6/7 list the mean offsets and scatters for the bulges and the disks from the 24 input combinations.B/T MAX( f bulgeM ∗ ,bin ) MAX( f diskM ∗ ,bin ) < offset > bulge < offset > disk < scatter > bulge < scatter > disk Gyr Gyr Gyr Gyr0.30 0.58 0.90 1.96 2.50 0.09 0.070.50 0.80 0.65 1.87 2.86 0.09 0.150.70 0.91 0.51 1.28 2.88 0.05 0.21
REFERENCES
AAO Software Team. 2015, 2dfdr: Data reduction software.http://ascl.net/1505.015Ade, P. A. R., Aghanim, N., Alves, M. I. R., et al. 2014,A&A, 571, A1, doi: 10.1051/0004-6361/201321529Ahn, C. P., Alexandroff, R., Allende Prieto, C., et al. 2012,ApJS, 203, 21, doi: 10.1088/0067-0049/203/2/21 Allen, J. T., Green, A. W., Fogarty, L. M. R., et al. 2014,SAMI: Sydney-AAO Multi-object Integral fieldspectrograph pipeline, Astrophysics Source Code Library.http://ascl.net/1407.006Allen, J. T., Croom, S. M., Konstantopoulos, I. S., et al.2015, MNRAS, 446, 1567, doi: 10.1093/mnras/stu2057
AMI: bulge and disk stellar populations Figure 23.
Age-metallicity plots for B/T=0.3, 0.5, 0.7 ( top to bottom panels ) separating the inputs with ages of the bulge olderor younger compared to the disk ( left and right panels , respectively). The bulge/disk inputs are represented by circles/ellipseswith red/blue contours and they are connected by the grey line. The bulge/disk outputs are represented by filled red/bluecircles/ellipses with the associated 16th and 84th percentiles. Bulge/disk inputs and outputs are connected by red/blue linesrepresenting the offsets, respectively. The bulge inputs are more closely recovered for high B/T, while for the disk the resultsare better constrained for low B/T. Barsanti et al.
Astropy Collaboration, Robitaille, T. P., Tollerud, E. J.,et al. 2013, A&A, 558, A33,doi: 10.1051/0004-6361/201322068Astropy Collaboration, Price-Whelan, A. M., Sip˝ocz, B. M.,et al. 2018, AJ, 156, 123, doi: 10.3847/1538-3881/aabc4fBarway, S., Wadadekar, Y., Vaghmare, K., & Kembhavi,A. K. 2013, MNRAS, 432, 430,doi: 10.1093/mnras/stt478Bedregal, A. G., Arag´on-Salamanca, A., & Merrifield,M. R. 2006, MNRAS, 373, 1125,doi: 10.1111/j.1365-2966.2006.11031.xBedregal, A. G., Cardiel, N., Arag´on-Salamanca, A., &Merrifield, M. R. 2011, MNRAS, 415, 2063,doi: 10.1111/j.1365-2966.2011.18752.xBekki, K. 1998, ApJL, 502, L133, doi: 10.1086/311508Bekki, K., & Couch, W. J. 2011, MNRAS, 415, 1783,doi: 10.1111/j.1365-2966.2011.18821.xBland-Hawthorn, J., Bryant, J., Robertson, G., et al. 2011,Optics Express, 19, 2649, doi: 10.1364/OE.19.002649Blanton, M. R., & Roweis, S. 2007, AJ, 133, 734,doi: 10.1086/510127Breda, I., Papaderos, P., & Gomes, J. M. 2020, arXive-prints, arXiv:2006.02307.https://arxiv.org/abs/2006.02307Brough, S., Croom, S., Sharp, R., et al. 2013, MNRAS, 435,2903, doi: 10.1093/mnras/stt1489Bryant, J. J., Bland-Hawthorn, J., Fogarty, L. M. R.,Lawrence, J. S., & Croom, S. M. 2014, MNRAS, 438,869, doi: 10.1093/mnras/stt2254Bryant, J. J., Owers, M. S., Robotham, A. S. G., et al.2015, MNRAS, 447, 2857, doi: 10.1093/mnras/stu2635Bryant, J. J., Bland-Hawthorn, J., Lawrence, J., et al.2018, in Society of Photo-Optical InstrumentationEngineers (SPIE) Conference Series, Vol. 10702,Proc. SPIE, 107021H, doi: 10.1117/12.2309893Bundy, K., Bershady, M. A., Law, D. R., et al. 2015, ApJ,798, 7, doi: 10.1088/0004-637X/798/1/7Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ,533, 682, doi: 10.1086/308692Cameron, E. 2011, PASA, 28, 128, doi: 10.1071/AS10046Cappellari, M. 2016, Annual Review of Astronomy andAstrophysics, 54, 597,doi: 10.1146/annurev-astro-082214-122432—. 2017, MNRAS, 466, 798, doi: 10.1093/mnras/stw3020Cappellari, M., & Emsellem, E. 2004, PASP, 116, 138,doi: 10.1086/381875Cappellari, M., Emsellem, E., Krajnovi´c, D., et al. 2011,MNRAS, 416, 1680,doi: 10.1111/j.1365-2966.2011.18600.x Chakraborty, A., Feigelson, E. D., & Babu, G. J. 2014,astrolabe: Astronomy Users Library for R.http://CRAN.R-project.org/package=astrolibRChristlein, D., & Zabludoff, A. I. 2004, ApJ, 616, 192,doi: 10.1086/424909Coccato, L., Jaff´e, Y. L., Cortesi, A., et al. 2019, MNRAS,3218, doi: 10.1093/mnras/stz3592Cortese, L., Fogarty, L. M. R., Bekki, K., et al. 2016,Monthly Notices of the Royal Astronomical Society, 463,170, doi: 10.1093/mnras/stw1891Croom, S. M., Lawrence, J. S., Bland-Hawthorn, J., et al.2012, MNRAS, 421, 872,doi: 10.1111/j.1365-2966.2011.20365.xDe Propris, R., Couch, W. J., Colless, M., et al. 2002,MNRAS, 329, 87, doi: 10.1046/j.1365-8711.2002.04958.xDressler, A. 1980, ApJ, 236, 351, doi: 10.1086/157753Dressler, A., Oemler, Augustus, J., Couch, W. J., et al.1997, The Astrophysical Journal, 490, 577,doi: 10.1086/304890Falc´on-Barroso, J., Lyubenova, M., & van de Ven, G. 2015,in IAU Symposium, Vol. 311, Galaxy Masses asConstraints of Formation Models, ed. M. Cappellari &S. Courteau, 78–81, doi: 10.1017/S1743921315003439Falc´on-Barroso, J., S´anchez-Bl´azquez, P., Vazdekis, A.,et al. 2011, A&A, 532, A95,doi: 10.1051/0004-6361/201116842Fasano, G., Poggianti, B. M., Couch, W. J., et al. 2000,The Astrophysical Journal, 542, 673, doi: 10.1086/317047Fischer, J. L., Dom´ınguez S´anchez, H., & Bernardi, M.2019, MNRAS, 483, 2057, doi: 10.1093/mnras/sty3135Fraser-McKelvie, A., Arag´on-Salamanca, A., Merrifield, M.,et al. 2018, MNRAS, 481, 5580,doi: 10.1093/mnras/sty2563Ge, J., Yan, R., Cappellari, M., et al. 2018, MNRAS, 478,2633, doi: 10.1093/mnras/sty1245Girardi, L., Bressan, A., Bertelli, G., & Chiosi, C. 2000,A&AS, 141, 371, doi: 10.1051/aas:2000126Green, A. W., Croom, S. M., Scott, N., et al. 2018,MNRAS, 475, 716, doi: 10.1093/mnras/stx3135Gunn, J. E., & Gott, III, J. R. 1972, ApJ, 176, 1,doi: 10.1086/151605Head, J. T. C. G., Lucey, J. R., & Hudson, M. J. 2015,MNRAS, 453, 3729, doi: 10.1093/mnras/stv1662Head, J. T. C. G., Lucey, J. R., Hudson, M. J., & Smith,R. J. 2014, MNRAS, 440, 1690,doi: 10.1093/mnras/stu325Hubble, E. P. 1936, Realm of the NebulaeHudson, M. J., Stevenson, J. B., Smith, R. J., et al. 2010,MNRAS, 409, 405, doi: 10.1111/j.1365-2966.2010.17318.x