Correlation of structure and stellar properties of galaxies in Stripe 82
aa r X i v : . [ a s t r o - ph . GA ] J u l Draft version July 17, 2020
Preprint typeset using L A TEX style emulateapj v. 12/16/11
CORRELATION OF STRUCTURE AND STELLAR PROPERTIES OF GALAXIES IN STRIPE 82
Sonali Sachdeva
Kavli Institute for Astronomy and Astrophysics, Peking University, Beijing 100871, China
Luis C. Ho and Yang Li
Kavli Institute for Astronomy and Astrophysics, Peking University, Beijing 100871, China andDepartment of Astronomy, School of Physics, Peking University, Beijing 100871, China andFrancesco Shankar
Department of Physics and Astronomy, University of Southampton, Southampton SO171BJ, UK
Draft version July 17, 2020
ABSTRACTEstablishing a correlation (or lack thereof) between the bimodal colour distribution of galaxies andtheir structural parameters is crucial to understand the origin of bimodality. To achieve that, we haveperformed 2D mass-based structural decomposition (bulge+disc) of all disc galaxies (total=1263) inthe Herschel imaging area of the Stripe 82 region using K s band images from the VICS82 survey.The scaling relations thus derived are found to reflect the internal kinematics and are employed incombination to select an indubitable set of classical and pseudo bulge hosting disc galaxies. The restof the galaxies ( < M ∗ , sSFR, r − K s ). All pseudobulge disc galaxies are blue and star-forming and all classical bulge disc galaxies are red and quiescentwith less than 5% digressions. Ambiguous bulge disc galaxies are intermittent to pseudo and classicalbulge disc galaxies in the distribution of all structural and stellar parameters. ∆ h µ eb i - based on theplacement of bulges on the Kormendy relation - is found to be the most efficient single structuralindicator of both bulge type and stellar activity. The placement of ambiguous bulge disc galaxies onscaling relations and fundamental plane, in addition to their peculiar stellar properties suggest thatthey are dominantly a part of the green valley. Subject headings: galaxies: bulges — galaxies: structure — galaxies: star formation — galaxies: stellarcontent INTRODUCTION
Over the past two decades, it has been emerging withincreasing clarity that the structure of galaxies is corre-lated with their star formation history and on-going ac-tivity (Kauffmann et al. 2003, 2006; Baldry et al. 2006;Franx et al. 2008). This conjecture is supported bythe observation of statistically large samples that re-veal that while redder galaxies are more early-typeand bulge (or spheroid) dominated, bluer galaxies aremore late-type and disc dominated (Strateva et al. 2001;Brinchmann et al. 2004).Quantitatively, Driver et al. (2006) demonstrated thatS´ersic index of a galaxy - parameter defining the shapeof its intensity profile - is most efficient in separat-ing galaxies based on their colour. In addition to theS´ersic index, concentration and bulge-to-total flux ratiohave been demonstrated to be competitively effective forlarger samples and at higher redshifts (Cameron et al.2009; Wuyts et al. 2011; Bell et al. 2012; Mendel et al.2013; Lang et al. 2014; Bluck et al. 2014). In some stud-ies, central mass density of galaxies within 1 kpc has beenargued to be a more efficient separator of star formingand quiescent galaxies than other morphology indicators(Cheung et al. 2012; Fang et al. 2013; Luo et al. 2020).Recent studies have also argued that stellar kinemat-ics, mainly in terms of central velocity dispersion, area better differentiator of galaxy colour than stellar mass, surface mass density and morphology (Wake et al. 2012;Bluck et al. 2016; van de Sande et al. 2018). Note thata requisite for accurate morphological decomposition isthat it reflects the internal kinematics of the galaxy. Re-cently, Graham et al. (2018) reported that kinematics of2300 galaxies obtained from the latest IFU (MANGA)survey demonstrate a tight correlation with their struc-ture. Thus, all proposed colour differentiators - S´ersic in-dex ( n b ), bulge-to-total flux ratio ( B/T ), concentration( C ), central mass density (Σ ) and velocity dispersion( σ o ) - are structural type indicators, firmly suggestingthat the inherent structure of star forming galaxies dif-fers from their quiescent counterparts.Interestingly, all these structural indicators ( n b , B/T , C , Σ , σ o ) have also been found to be efficient differen-tiators of the bulge-type hosted by the disc galaxy (re-viewed in Fisher & Drory 2016; Kormendy 2016). Thehigher values of these indicators for a disc galaxy suggestthat the central bulge is elliptical like, i.e., “classical”,and lower values suggest that the central bulge is disclike, i.e, “pseudo” (Fisher & Drory 2008; Gadotti 2009;Sachdeva & Saha 2016; Neumann et al. 2017; Gao et al.2020). Involvement of the same set of indicators in theadjudication of both the bulge-type and stellar activityin disc galaxies implies that disc galaxies hosting differ-ent bulge types should exhibit distinct stellar activity.Drory & Fisher (2007) have argued that the underlying Sachdeva et al.correlation of the colour of the galaxy is with its bulgetype and correlation with all other morphological indica-tors is a consequence of that.Thus, separation of disc galaxies on the basis of theirbulge-type should also result in their separation on thebasis of colour, and vice-versa. However, two sets ofstudies have produced contradictory findings. One set ofstudies state that while pseudo bulges are often red, clas-sical bulges are rarely found to be blue (Fisher & Drory2016; Kormendy 2016). Counter to that, other set ofstudies state that while pseudo bulges are rarely foundto be red, classical bulges are often blue (Gadotti 2009;Fang et al. 2013; Luo et al. 2020). Luo et al. (2020)highlighting these differences claim that even if they ap-ply the same colour separation criteria as Fisher & Drory(2016), 42% of their classicals will be marked blue andall pseudo bulges will still be contained in the blue zone.The origin of contradictory results lies in the incon-sistency of criteria applied in these studies for bulgeclassification. For example, while Fisher & Drory (2008)employ bulge S´ersic index ( n b ) to separate bulge-types,other studies have elaborated on the inefficiency of n b to separate bulge-types without severe contamination(Gadotti 2009; Gao et al. 2020). The reason being thatwhile there are bulges which will satisfy multiple kinds ofcriteria to be unambiguously labelled as either pseudo orclassical, there are also a considerable fraction of bulgeswhich do not meet all the chosen criteria (reviewed inKormendy 2016). Classification of such bulges is thensubjective and both pseudo and classical samples becomeprone to contamination.In this work, we attempt to overcome this issue bystudying and applying multiple stringent complimentarycriteria to select an indubitable set of pseudo and classi-cal bulges. In addition to that, we prevent the contami-nation of these bulge sets by separating out those bulgeswhich do not satisfy all the applied stringent criteria.We make a number of other improvisations to extract aclearer picture. Firstly, our sample consists of all local( z < .
3) disc dominated galaxies (total=1263) in theHerschel imaging area of the Stripe 82 region. Since thisarea has coverage from a large number of deep multi-wavelength surveys, high quality and large quantity ofdata in terms of images, spectra and derived parametersis available. This is advantageous in analysing the ef-fect of structural transformation on the whole range ofstellar, gas and dust properties. Crucially, structural pa-rameters for all the galaxies in our sample are available inthe optical bands (ugriz) from the latest decompositionperformed by Bottrell et al. (2019) using deep co-addedSloan images. This has enabled us to perform consis-tency checks for our derived decomposition parametersin the K s band.Secondly, we have performed the decomposition in K s band. This band is the best tracer of stellar mass ingalaxies since it is not biased by the dominating opticalemission from young stars which account for a small frac-tion of galaxy’s mass. It is least affected by dust obscura-tion and accounts for all the light emitted by middle-ageand old stars which form the bulk of galaxy’s baryonicmass (Cowie et al. 1996; Bundy et al. 2006; Bluck et al.2019). Our decomposition also benefits from the factthat the K s band images (VICS82 survey, Geach et al.2017) boast of a high resolution (0.3”/pixel) and depth (21.4 mag). Thirdly, we have extracted all the possiblestructural measures of the galaxies using both paramet-ric and non-parametric techniques. This allows us toassess the performance of all potential indicators, alongwith the kinematic information, to classify bulges in arobust manner. Fourthly, the stellar parameters ( M ∗ ,SFR, sSFR) are from a recent work (Salim et al. 2016,2018) that includes far-IR flux from Herschel along withthe mid-IR flux from WISE to account for the dust at-tenuation affect which is a critical factor in the accurateestimation of stellar activity.This paper is organized in the following manner. InSection 2, we define the sample and elaborate our com-putation of all defining structural parameters using bothparametric and non-parametric techniques. We also de-tail the source and computation of kinematic and stellarparameters employed in this work. In Section 3, we elab-orate on our separation of an indubitable set of pseudoand classical bulges with the identification and applica-tion of multiple stringent complimentary criteria. Fol-lowing that, we compare the distribution of all struc-tural and stellar parameters for the two bulge-types. Wehave also attempted to identify the most effective singlestructural indicator for both bulge-type and stellar activ-ity in disc galaxies. In Section 4, we discuss our findingspertaining to the classification of bulges, bimodality oftheir properties and clues regarding the constitution ofthe green valley. We conform to a flat Λ-dominated Uni-verse with Ω Λ = 0 . m = 0 .
286 and H o = 69 . − Mpc − . All magnitudes are in the AB system. DATA
Bottrell et al. (2019) performed bulge disc decomposi-tion of all galaxies in the Stripe 82 region (16908 in total)in optical (u,g,r,i,z) bands using deep co-added SDSS im-ages from Annis et al. (2014). They demonstrated thatdeep images result in a more accurate determination ofbulge and disc parameters compared to the computationsbased on SDSS Legacy images (Simard et al. 2011). Intheir work, only the decomposition performed in r bandhas all fitting parameters kept free. For rest of the bands( u,g,i,z ) either some or most of the parameters are heldfixed according to their r band values. We, thus, taketheir r band catalogue for our sample selection and laterfor comparison with our mass-based morphological mea-surements. Using their r band catalogue, we first selectthose galaxies which are determined to be 2-component(disc + free-bulge) systems, according to their analysis.We put a further cut on that sample by selectingonly those galaxies which are in the 79 deg contigu-ous imaging field of Herschel Stripe82 Survey (HerS,Viero et al. 2014). This ensures maximum overlap withall major multi-wavelength deep surveys in Stripe 82including BOSS (Eisenstein et al. 2011), VLA-Stripe82(Hodge et al. 2011), HETDEX (Hill et al. 2008), SHELA(Papovich et al. 2012), SpIES (Timlin et al. 2016), HSC(Miyazaki et al. 2012) and VICS82 (Geach et al. 2017).This overlap is essential to perform mass-based morpho-logical decomposition of galaxies and to obtain accurateestimates of their stellar parameters, dynamical param-eters, dust content and gas content, which are crucial totrace the affect of structural transformation of galaxy onits stellar activity.Out of the total sample of 16908 galaxies oforrelation of structure and stellar properties of galaxies 3Bottrell et al. (2019), 1263 galaxies satisfy both the cri-teria, i.e., are determined to be 2-component (disc + free-bulge) systems and are present in the HerS field. We willperform the structural decomposition of this sample of1263 galaxies in K s band. Selection of this band is cru-cial to trace the underlying mass distribution of galaxiesin the most accurate manner. Bluck et al. (2019) using5 X 10 local ( z < .
2) SDSS galaxies, have demon-strated that
B/T computed in all optical bands underes-timates
B/T by mass. For r band, the difference is morethan 0.2, i.e., a component having 40% of galaxy’s fulllight might be designated to be accounting for only 20%of that (see their Figure B1). They report that opticaland mass-based structural fractions coincide closely forthe K s band. This is because K s band accounts for allthe light emitted by young and old stars and does notget affected by the brightness of young stellar popula-tions (Cowie et al. 1996; Bundy et al. 2006; Bluck et al.2019). Since our work is focused on classifying bulgesin an indubitable manner and examine the bimodality oftheir stellar properties, it is paramount to perform a mass(baryonic) based decomposition. In addition to that, our K s band images support a much better resolution (0.3”)than the Sloan images (0.396”) used by (Bottrell et al.2019).The K s band images are obtained from VICS82 sur-vey, i.e., VISTA-CFHT Stripe 82 near-IR survey, whichcovers near-contiguous 150 deg of Stripe 82 to an aver-age depth of 21.9 mag in J band and 21.4 mag in K s band (Geach et al. 2017). This survey has been con-ducted using Canada-France Hawaii Telescope (CFHT)WIRCam instrument and Visible Infrared Survey Tele-scope for Astronomy (VISTA) VIRCAM instrument. Allimages have been processed including dark and flat-fieldcorrection, refined sky subtraction, distortion correction,quality control, astrometric and photometric calibration.Using their K s selected catalogue, we match the RA Decof our sample of 1263 galaxies to 1” difference and ob-tain information regarding the tile and coordinates onwhich each galaxy is present. Using this information, wewrite a code which automatically selects the tile (out of33 VIRCAM and 55 WIRCam tiles in K s band) on whicha galaxy is present and extracts a 180” X 180” cut-outfor that galaxy. This cutout size is reasonably large con-sidering that out of 1263, only 6 sources have total ex-tent (Petrosian radius) more than 30” and none exceeds40”. It is optimal for the running of ellipse , GALFIT and other algorithms followed in this work. In additionto that, large size ensures that sky dominates the fittingregion which is also an important criteria for
GALFIT to create an accurate sigma image. We, thus, obtaincut-outs of all 1263 galaxies in the K s band. Sky variation and masking
Accurate measurement of sky background variation isessential for robust fitting of galaxy images (Peng et al.2010). We have estimated this variation separately foreach of the VICS82 survey tiles from which our samplegalaxies’ cut-outs are obtained. The tiles have alreadybeen through rigorous background modification duringpre-processing, which includes subtraction of the run-ning sky, “destriping” of the images in both directions,removal of background gradients, visual inspection of im-ages to removes defects, patterns, residuals, in addition to fitting and removal of a fifth-degree polynomial surface(Geach et al. 2017).To compute the sky variation, we first run Source Ex-tractor (Bertin & Arnouts 1996) with a low detectionthreshold (0.8 times relative to the background RMS)and broad filter (Gaussian with FWHM of 5 pixels) tomask all possible sources present on the tile along withtheir faint outskirts (Akhlaghi & Ichikawa 2015). Wefurther grow each mask region according to the maskedsource size to rule out the possibility of contamina-tion from extended diffused light. On the non-masked(sky) pixels, we perform 3-sigma clipping using stan-dard
Python algorithm to obtain the statistical measures.Since the background has already been through rigorousprocedures, the mean and median of the sky backgroundare found to be zero across all tiles. The standard devia-tion is found be similar to 2nd decimal place for all tilesof a given telescope (CFHT and VISTA) and band ( K s ),confirming the robustness of computation. For CFHT K s band images, the value ranges from 23.90 to 24.17mag/arcsec and for VISTA K s band images, the valueranges from 23.80 to 23.98 mag/arcsec . These back-ground variation values will be given as an input to GAL-FIT to aid the algorithm to obtain the most accuratebulge and disc parameters.The mask (or segmentation-map) images generated foreach cut-out in this process are also a critical input to thefitting algorithms employed in this work. These imagesare first given as an input to a custom-made code whichunmasks the central source, i.e., the galaxy of interest,in each cut-out. We visually examined all modified maskimages and corrected for a few cases (less than 5%) inwhich some part of the galaxy of interest got maskedor some neighbouring source did not get masked prop-erly. The final mask images are given as an input tothe isophotal fitting algorithm (
IRAF ellipse ) which isused to generate input values for
GALFIT and later to
GALFIT as well.
Initial input parameters
We will perform the bulge-disc decomposition on allgalaxy images using
GALFIT which is an algorithm thatuses the Levenberg-Marquardt (LM) technique on galaxyimages to find the best fitting structural parameters ina flexible and fast manner (Peng et al. 2002, 2010). Toobtain the initial input parameters for
GALFIT , we per-form isophotal fitting on all galaxy images using the el-lipse task of
IRAF . This task finds best fitting isophotesfor a galaxy at successively increasing radii, using LMtechnique, varying the intensity, central coordinates, el-lipticity and position angle (Jedrzejewski 1987). Thecutouts and their corresponding masks are given as aninput. We use the resulting table of isophotal values tocreate surface brightness profile for each galaxy accord-ing to their magnitude zero-point and plate-scale. Theseprofiles are given as an input to a custom written codewhich extracts initial input parameters for
GALFIT , i.e.,total radius, magnitude, half light radius, ellipticity andposition angle, for each of the 1263 galaxies. The totalradius is defined by the outer-most fitting isophote andthe total magnitude is according to the total flux con-tained in this radius. The half light radius is taken asthat radius which contains half of the total flux, whileellipticity and position angle are obtained by computing Sachdeva et al.the mode of all values across the profile. Other than gen-erating the input values, the isophotal profiles will alsobe utilized after running
GALFIT to examine if the out-put parameters generated by
GALFIT are describing thegalaxy well or significant residual is left.
Sigma image and PSF
A vital requirement for
GALFIT ’s fitting process isthe sigma image which informs the algorithm regardingthe standard deviation of flux at each pixel of the inputimage (Peng et al. 2002).
GALFIT can create an accu-rate sigma image internally if information regarding theunits of the input image, flux conversion factor, gain ofthe telescope, exposure time involved, number of framescombined to create the input image and read noise ofthe detector is known to the user. We assimilated allsuch information from survey documentation, headers ofthe large CFHT/VISTA tiles and image analysis. WhileCFHT images are in micro-Janskys, VISTA images arein counts. Using the appropriate flux conversion factor,exposure time and gain of the telescope, we create the“gain” parameter in the header of each cut-out (or inputimage) in such a manner that its multiplication with theimage units will yield the total number of electrons foreach pixel. We also create the “rdnoise” parameter ineach cut-out’s header. These header parameters will beutilized by
GALFIT to generate sigma image.Another important requirement for
GALFIT ’s fittingprocess is the point spread function (PSF) which the al-gorithm convolves with the model to mimic the effectof the telescope and filter on the actual flux distribu-tion. The VICS82 survey provides an average PSF foreach telescope (CFHT and VISTA) and band ( J and K s )by evaluating the FWHM of 10 point sources extractedfrom randomly selected image tiles belonging to eachtelescope and band. In the creation of the median stack,which includes the normalization of each source to itspeak flux, they ensured that all selected 10 point sources( CLASS ST AR > .
95) are bright (14 < K s <
15 mag)and unsaturated. They exhibited the quality of the gen-erated PSF by employing it to derive aperture correctionsfor their photometric analysis. We followed the same pro-cess to create our own PSF, for each CFHT and VISTA K s band tile, selecting some ∼
100 point sources using
DAOPHOT PSF task of IRAF and compared the PSFthus generated with the one provided by the survey. Al-though we selected a far fewer number ( ∼ σ ∼ ± σ ∼ ± K s band. Considering the insen-sitivity of the fitting to the PSF variation, we use thePSF provided by VICS82 survey, throughout this work,to rule-out any biases originating from the creation of aPSF in a non-rigorous and/or non-uniform manner forany tile. Fitting 1 and 2-component models
To run
GALFIT in a batch, i.e., on the full sampleof 1263 galaxies, a customized input file for each galaxy is required. This file stores all information pertaining tothe running of the algorithm on that galaxy, i.e., name ofthe cut-out, its mask file, PSF file, fitting box, convolu-tion box, magnitude zeropoint, plate scale, initial inputparameters, etc. Using the initial input parameters, ob-tained earlier through isophotal fitting, we write a codeto generate such files for all galaxies in an automatedmanner. The dimension of the fitting box is twice thediameter of the central source and that of the convolu-tion box ranges from 40-60 times the PSF-FWHM of theimage. Using the customized files, we first fit all galaxieswith a single S´ersic component, I sersic ( r ) = I e ( r e ) exp[ − b n (( rr e ) /n − , (1)where r e is the half light radius, I e ( r e ) is the intensity atthat radius, n is the S´ersic index and b n is a constant de-pendent on n . Out of 1263 galaxies, while 946 convergedwell (1 < χ ν < GALFIT did not crash butone or more final parameter values were marked withasterisk signifying that those parameter values are nonphysical. We re-fitted those 317 galaxies fixing theirS´ersic index to range of values ( n = 1 . , . , .
0) andselecting the best fit, following which, 287 converged and30 galaxies still did not converge. Examining the imagesof these 30 galaxies, we found that while 10 of them arepoorly imaged, the rest do not have any clear signs thatcould explain their lack of convergence.We consolidate the output parameters, obtained from1-component fitting, of all galaxies and use them to cre-ate input files for 2-component fitting. In 2-componentfitting, S´ersic component (Equation 1) is simultaneouslyfitted with an exponential (or disc) component. The ex-ponential component is the special case of the S´ersic com-ponent with n = 1, such that, I disc ( r ) = I o exp( − rr d ) , (2)where I o is the intensity at the centre of the disc and r d isthe scale length of the disc. Thus, I total = I sersic + I disc is simultaneously fitted for the full sample. Out of 1263galaxies, while 605 converged well (1 < χ ν < n b = 1 . , . , .
0) and selected thebest fit. Out of those 658, 337 converged, however, 321did not converge. We tried to re-fit them by fixing variousparameters including ellipticity and position angle of thedisc (derived from isophotal fitting), however, they stilldid not converge.Thus, for 2-component (bulge+disc) fitting, 942 galax-ies converged. Fig. 1 and Fig. 2 depict the fitting ofa few sources in K s band along with a comparison ofthe best-fit parameters with the observed intensity pro-file for each source. The first three columns exhibit thereal, model and residual images produced by GALFIT for the best fit parameters. The fourth column showsthe bulge, disc and total (bulge+disc) intensity profilesthat were generated using the best-fit model parameters,along with the galaxy’s observed isophotal intensity pro-file. Computation and analysis of the residual of theobserved and model profiles reveals that galaxies fittedwith a free bulge S´ersic index fit significantly better thanorrelation of structure and stellar properties of galaxies 5those for which it was held fixed suggesting that it is acrucial factor in accurate determination of bulge param-eters.
Non-parametric measures
In addition to parametric measures, we compute non-parametric measures for all galaxies, mainly their Pet-rosian radius, Concentration and Asymmetry. Non-parametric measures, by definition, are not constrainedby any functional form and are thus considered leastbiased measures of galaxies’ structure (Conselice 2014).The algorithm for measuring Petrosian radius of a galaxyis based on the computation of the ratio of intensity atsuccessively increasing radii to the intensity inside thoseradii. When the ratio η ( r ) falls to an empirically deter-mined fraction (0.2 in this work), η ( r p ) = I ( r p ) h I ( < r p ) i = 0 . , (3)at some radius r p , then Petrosian radius ( r P ) is givenas 1 . × r p where 1.5 is again an empirically determinedmultiple. By “empirically determined” we mean thatthese values (0.2 and 1.5) were found to provide most ac-curate estimates for a representative sample of galaxies(Bershady et al. 2000; Conselice 2003; Lotz et al. 2004).The algorithm thus works on extrapolation of intensityprofile from the centre to the outskirts to obtain the to-tal extent of the galaxy. For Concentration, first theflux inside the total (Petrosian) radius of the galaxy iscomputed. Then the algorithm finds those radii whichcontain 20% and 80% of the total flux to obtain, C = 5 log ( r r ) , (4)where C is the concentration index of the galaxy and5 is an empirically determined multiple (Bershady et al.2000; Graham et al. 2005). For Asymmetry, galaxy im-age is rotated by 180 o about its “centre of symmetry”(a central point, found iteratively, where asymmetry isminimum) and subtracted from the main image. Fluxthus obtained from the subtracted image is normalizedto obtain the asymmetry index ( A ) of the galaxy. Ourcode for computing the three measures ( r P , C and A )ran smoothly for 1253 out of the total sample of 1263galaxies. For the rest 10 galaxies, iterative algorithm tofind the centre of the galaxy did not converge. Stellar parameters
Stellar parameters, i.e., stellar masses ( M ∗ ) and starformation rates (SFR) are taken from GALEX-SDSS-WISE Legacy Catalogue 2 (Salim et al. 2018). They fol-low a Bayesian approach to SED fitting on the combi-nation of GALEX and SDSS data of all galaxies (0.7million) with z < .
3. Their most important modifica-tion over previous works (Salim et al. 2005, 2007, 2016)is the usage of IR luminosity to set constraints on dustemission which allows dust attenuation curve parametersto be fitted freely while creating models. Computationof “true” IR luminosity is based on far-IR flux from Her-schel (Valiante et al. 2016) in addition to mid-IR fluxfrom WISE (Lang et al. 2016). An accurate estimation of the dust attenuation affect is a critical factor in the de-termination of stellar activity of galaxies. They demon-strate by comparing with other works that if far-IR isnot included in the computation, SFRs get systemati-cally over-estimated especially for quiescent galaxies.Their project involves usage of Herschel ATLAS to cal-ibrate computation of true IR luminosity for the full sam-ple of 0 . M ∗ and SFR estimates for 1205 galaxies. Central velocity dispersion
Stellar velocity dispersions for galaxies in our sam-ple have been obtained from SDSS DR15 spectroscopiccatalogue (Bolton et al. 2012; Aguado et al. 2019). TheSDSS spectra, since DR9, have been obtained usingBOSS spectrograph (Dawson et al. 2013) which has a fi-bre diameter of 2” and wavelength coverage from 365to 1040 nm. Each spectra is reduced by the spectro-scopic pipeline ( idlspec2d ) which is refined with eachdata release. Stellar velocity dispersions are derived fol-lowing a Principal Component Analysis (PCA) methodwhere 24 eigenspectra from ELODIE stellar library(Prugniel & Soubiran 2001) are convolved and binned tomatch the instrumental resolution and constant-velocitypixel scale of reduced SDSS spectra. These templatesets are redshifted, broadened by successively large ve-locity widths and modelled through least square fittingof linear combination of each trial broadening. Best-fitvelocity dispersion value is thus determined through chi-square minimization and error on the value is determinedfrom curvature of the chi-squared curve around globalminimum. Based on the average S/N and instrumen-tal resolution of SDSS spectra, velocity dispersion mea-surements below 70 km/s should be treated with caution(Thomas et al. 2013). In our work, we will use these mea-surements only for bulge classification, where, all galaxieswith velocity dispersion below 90 km/s will be clubbedtogether. Thus even large ( ∼ σ o ),using, σ o = ( r eb / r ap ) − . σ ap (5)where, r eb is the half light radius of the bulge in arc-seconds, r ap is the radius of the aperture (1”), σ ap is the measured stellar velocity dispersion and 0 . < ∼ Consolidation and comparison
Sachdeva et al. −6 −4 −2 0 2 4 6arcsecs−6−4−20246 a r c s e c s I - 292, z - 0.147 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 M - 292 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 R - 292, rChi2 - 1.166−6 −4 −2 0 2 4 6arcsecs−6−4−20246 a r c s e c s I - 625, z - 0.159 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 M - 625 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 R - 625, rChi2 - 1.143−6 −4 −2 0 2 4 6arcsecs−6−4−20246 a r c s e c s I - 1086, z - 0.135 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 M - 1086 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 R - 1086, rChi2 - 1.189−6 −4 −2 0 2 4 6arcsecs−6−4−20246 a r c s e c s I - 1176, z - 0.174 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 M - 1176 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 R - 1176, rChi2 - 1.139−6 −4 −2 0 2 4 6arcsecs−6−4−20246 a r c s e c s I - 1181, z - 0.213 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 M - 1181 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 R - 1181, rChi2 - 1.142
Fig. 1.—
Fitting 2-components:
Each row depicts the fitting of a particular source in K s band. In a row, first three columns show theactual image of the source, its GALFIT model and its residual, respectively. While the redshift of the source has been mentioned at thetop of actual image, the χ ν of the fitting has been mentioned at the top of residual image. The fourth column compares the model image’sprofile (green solid line) with the observed profile of the actual image (black solid points) obtained through isophotal analysis. Profiles ofthe bulge (red solid line) and the disc (blue solid line) component of the model image have also been marked. The difference of the modeland observed profile is shown in the bottom panel of the plot. orrelation of structure and stellar properties of galaxies 7 −15−10−5 0 5 10 15arcsecs−15−10−5051015 a r c s e c s I - 246, z - 0.018 −15−10−5 0 5 10 15arcsecs−15−10−5051015 M - 246 −15−10−5 0 5 10 15arcsecs−15−10−5051015 R - 246, rChi2 - 1.068 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 a r c s e c s I - 449, z - 0.079 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 M - 449 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 R - 449, rChi2 - 1.185−6 −4 −2 0 2 4 6arcsecs−6−4−20246 a r c s e c s I - 455, z - 0.079 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 M - 455 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 R - 455, rChi2 - 1.123−6 −4 −2 0 2 4 6arcsecs−6−4−20246 a r c s e c s I - 587, z - 0.018 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 M - 587 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 R - 587, rChi2 - 1.128−6 −4 −2 0 2 4 6arcsecs−6−4−20246 a r c s e c s I - 909, z - 0.017 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 M - 909 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 R - 909, rChi2 - 1.153−6 −4 −2 0 2 4 6arcsecs−6−4−20246 a r c s e c s I - 1206, z - 0.068 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 M - 1206 −6 −4 −2 0 2 4 6arcsecs−6−4−20246 R - 1206, rChi2 - 1.157
Fig. 2.—
Fitting 2-components:
While the description is same as Fig. 1, this figure is focused on smaller redshift ( z < .
1) sources.
Sachdeva et al.
TABLE 1Bulge-disc decomposition values of all the 1263 sources analysed in K s band ∗ ID m b r eb n b ar b PA b m d r d ar d mag arcsec mag arcsec1 2 3 4 5 6 7 8 9310 13.00( ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ∗ This table presents the values obtained through 2-component (bulge-disc) 2D fitting of galaxies. The first column is the uniqueID we have given to the 1263 objects. The next five columns (2-6) depict the magnitude ( m b ), effective radius ( r eb ), S´ersic index( n b ), axis ratio ( ar b ) and position angle (PA b ) of the bulge component. The last three columns (7-9) are the magnitude ( m d ),scale length ( r d ) and axis ratio ( ar d ) of the disc component. The full table has been made available. TABLE 2S´ersic fitting and non-parametric values of all the 1263 sources analysed in K s band ∗ ID m g r eg n g ar g PA g r P C A mag arcsec arcsec1 2 3 4 5 6 7 8 9310 12.35( ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ∗ This table presents the values obtained through single component (S´ersic) 2D fitting and non-parametric fitting of galaxies. Thefirst column is the unique ID we have given to the 1263 objects. The next five columns (2-6) are the magnitude ( m g ), effectiveradius ( r eg ), S´ersic index ( n g ), axis ratio ( ar g ) and position angle (PA g ) of the galaxy. The last three columns (7-9) are thePetrosian radius ( r P ), concentration ( C ) and asymmetry index ( A ) of the galaxy. The full table has been made available with thiswork. In consolidation, we have obtained both paramet-ric and non-parametric morphological measurements forour sample of 1263 galaxies in K s band. Based onthe cosmological parameters adopted, redshift of thegalaxies and corresponding K-corrections involved, weconvert all measured apparent quantities to intrinsicones in rest-frame K s band (equations illustrated inGraham & Driver 2005; Sachdeva 2013; Sachdeva et al.2015). Thus, for the full galaxy and for the bulge and thedisc component separately, we obtain their absolute mag-nitudes ( M g , M b , M d ), intrinsic characteristic sizes ( R eg , R eb , R d ) in Kpc, S´ersic indices ( n g , n b ) along with otherdefining parameters, i.e., position angle, ellipticity, etc.Note that subscript ‘g’ is for global, ‘b’ is for bulge and‘d’ is for disc. In addition to that, we have acquired thelatest and most accurate values of their stellar ( M ∗ , SFR, sSFR) and kinematic ( σ o ) parameters. In terms of para-metric morphology, we have performed both single com-ponent (S´ersic function) and 2-component (S´ersic func-tion + exponential function) fitting. For single compo-nent fitting, nearly all galaxies (barring 30) converged tophysically meaningful parameters, where, 946 convergedwith free n g and 287 converged when it was held fixed.For 2-component fitting, 942 galaxies (75% of the total)converged to physically meaningful parameters, where,605 converged with free n b and 337 converged when itwas held fixed.Tables (or catalogues) consolidating parametric, non-parametric, kinematic and stellar measures, of all the1263 galaxies in our sample, have been made availablewith this paper. This includes apparent as well as intrin-sic measures in rest-frame K s band, along with errors. Inorrelation of structure and stellar properties of galaxies 9 TABLE 3Intrinsic values of all the 1263 sources analysed in K s band with kinematic and stellar parameters ∗ ID M b R eb h µ eb i M d R d log( M ∗ ) log(SFR) σ o mag kpc mag/arcsec mag kpc [ M ⊙ ] [ M ⊙ /yr ] km/s1 2 3 4 5 6 7 8 9310 -23.32( ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ∗ This table presents the intrinsic (or absolute) values of the bulge-disc component obtained according to their redshift, adoptedcosmology and K-correction. The first column is the unique ID we have given to the 1263 objects. The next three columns (2-4)are the magnitude ( M b ), effective radius ( R eb ) and average surface brightness ( µ eb ) inside that radius for the bulge component.The following two columns (5-6) are the magnitude ( M d ) and scale length ( R d ) of the disc component. The last three columns(7-9) are the total stellar mass ( M ∗ ), star formation rate (SFR) and central velocity dispersion ( σ o ) of the galaxy. The full tablehas been made available with this work. Table 1, 2 and 3, we present a small glimpse of the fullappended version. Each version also includes a detaileddescription of the columns in its beginning.In Fig. 3, we compare the distribution of the param-eters obtained in K s band with those obtained in theoptical ( r band) by Bottrell et al. (2019). Note thatwe compare only that sample which converged in ourfitting process with S´ersic index (whether global n g orbulge n b ) kept free so as to match the assumptions madeby Bottrell et al. (2019) during their fitting of the samesample. We find that there is quite a reasonable matchbetween the two bands confirming the accuracy of thedecomposition carried out using two different methods,i.e., GIM2D and
GALFIT . More than 80% of all radii( r g , r eb , r d ) and most importantly the bulge-to-total ra-tio ( B/T ) are within 0.2 dex for the two bands. Evenfor a sensitive parameter like n b , for more than 70% thedifference is within 0.2 dex. Ellipticity ( e g , e b ) and posi-tion angle (PA) of all ( > n b . This is also necessary because n b is one of the key parameters whose performance asa differentiator of bulge type will be examined in thiswork. Henceforth, all the presented analysis is based onthis sample. RESULTS
Performance of bulge morphology indicators
Over the past few years, Kormendy relation (KR) - aprojection of the fundamental plane exhibited by ellip- tical galaxies (Kormendy 1977) - has established itselfas one of the most efficient classifiers of bulge morphol-ogy (Gadotti 2009; Sachdeva et al. 2017, 2019; Gao et al.2020). This is driven by the fact that classical bulges -being more dominated by dispersion than pseudo bulges- are found to lie within ± σ boundaries of the KR fol-lowed by elliptical galaxies. Pseudo bulges, in contrast,are found to be low surface brightness outliers to KR.To perform an unambiguous classification of bulges inour sample, we will extract a single morphology indica-tor (∆ h µ eb i ) from KR and employ that to investigate theperformance of other morphology indicators. Followingthat, the indicator which best compliments ∆ h µ eb i , willbe used in conjunction with it to select an indubitableclass of classical and pseudo bulges.To obtain KR for elliptical galaxies in rest-frame K s band, we select elliptical galaxies from the full sampleaccording to their global S´ersic index ( n g ). Analysis ofstatistically large samples of galaxies, over the past twodecades, has revealed that n g is an effective separatorof late-type ( n g < . − .
5) and early-type ( n g > . − .
0) galaxies (Shen et al. 2003; Ravindranath et al. 2004;Barden et al. 2005). In addition to that, studies focusingon early-type galaxies have stressed that higher is the n g ,lesser is the contamination from late-type counterparts(Blakeslee et al. 2006; Rettura et al. 2006; van der Wel2008; Huang et al. 2013). Based on these findings, weselect those galaxies from the full sample that have n g >
5. There is a possibility that a high S´ersic index cutmay result in the selection of the most luminous of theellipticals. However, we find that the relation observedby galaxies in the smaller range (3 . < n g < .
0) matcheswith that obtained for our selected sample with n g > h µ eg i = (2 . ± .
13) log( R eg ) + (16 . ± .
11) (6)0 Sachdeva et al.
Fig. 3.—
Band comparison:
The histograms show the distribution of the difference of parameter values computed by us in K s bandand those computed by Bottrell et al. (2019) in the r - band. The first three plots in the first row depict the distribution of the differencein the global effective radius ( r eg ), bulge effective radius ( r eb ) and disc scale length ( r d ). The fourth plot shows the difference in the S´ersicindex values of the two bands, both for the global (solid line) and the bulge (dashed line). In the second row, difference in the bulge-to-totalratio ( B/T ), global ellipticity ( e g ), bulge ellipticity ( e b ) and global position angle (PA g ) for the two bands is shown. Average error-barsare marked in each plot. where R eg is the effective radius of the galaxy (in Kpc)and h µ eg i is the intrinsic average surface brightness in-side it (in mag/arcsec ). The fitting has been performedusing the fit function of Gnuplot which implements non-linear least-square Marquardt-Levenberg algorithm. The1 σ scatter on h µ eg i is ± .
81 mag/arcsec . The scatteris larger than that observed by a recent work studyingthe placement of local ellipticals on the Kormendy planein the optical (Gao et al. 2020). This could be due toseveral factors, including a more stringent selection ofellipticals. To overcome this possibility, we adopt the+1 σ boundary, instead of +2 σ (Neumann et al. 2017) or+3 σ (Gadotti 2009; Gao et al. 2020) boundary, to ensurethat only those bulges are eventually chosen to be clas-sical which are most ‘elliptical like’. Fig. 4 also marksthe ± σ boundaries of the relation. Based on the +1 σ boundary, we define a quantity ∆ h µ eb i which marks therelative distance of the bulge from this boundary i.e.,∆ h µ eb i = h µ eb i − .
98 log( R eb ) − .
68 (7)where R eb is the effective radius of the bulge (in Kpc)and h µ eb i is the intrinsic average surface brightness in-side it (in mag/arcsec ). The quantity ∆ h µ eb i works asa bulge morphology classifier which embodies KR, suchthat for bulges which lie on the +1 σ boundary of KR, thisvalue is zero. More positive is this quantity more ‘non-elliptical like’ (NEL) is the bulge and more negative isthe quantity more ‘elliptical like’ (EL) is the bulge.Now we employ this indicator to investigate the per-formance of other potential bulge morphology classifiers,i.e., bulge S´ersic index ( n b ), bulge-to-total light ratio( B/T ), concentration index ( C ) and central velocity dis- persion ( σ o ). In Fig. 5, for successively increasing valuesof n b , B/T , C and σ o , we trace the increase in the frac-tion of bulges which are EL (i.e., have ∆ h µ eb i <
0) anddecrease in the fraction of bulges which are NEL (i.e.,have ∆ h µ eb i > n b , 80% of the bulges with n b > . n b < . C , 80% of the galaxies with C > . C < . C appears to be a more efficient classifierthan n b since with the increasing values of C , rise in thefraction of EL bulges and decline in the fraction of NELbulges is sharper than n b (Fig. 5). In contrast, B/T does not appear to be a good classifier. The trends ofincrease in EL bulges and decrease in NEL bulges withthe increasing values of
B/T are not as clear as in thecase of n b and C .Amongst the four morphology indicators, σ o turns outto be the most efficient classifier, where the two curvesare most sharp, smooth and stable (Fig. 5). The transi-tion occurs at σ o = 90 km/s, where 90% of the galaxieswith σ o >
90 km/s have EL bulges and 90% of the galax-ies with σ o <
90 km/s have NEL bulges. With increasingvalues of σ o , rise in the fraction of EL bulges and declinein the fraction of NEL bulges is sharper and more stablethan that observed for other morphology indicators.orrelation of structure and stellar properties of galaxies 11 Fig. 4.—
Kormendy relation:
The distribution of all galaxieswith global S´ersic index ( n g ) more than 5.0 is shown on the Ko-rmendy plane. These galaxies have also been confirmed visuallyto be lacking any disc component, i.e., have a high probability ofbeing pure ellipticals. While the solid black line marks the rela-tion followed by these sources, the two dotted lines mark the ± σ boundary of the relation, i.e., confines ∼
66% of the sources. Wewill use the +1 σ (lower dotted line) boundary as a separator of “el-liptical like” (EL) and “non-elliptical like” (NEL) bulges. Averageerror-bar is marked. Selection of pseudo, classical and “ambiguous”bulges
The investigation of the performance of morphologyindicators reveals that σ o best compliments KR in theseparation of EL and NEL bulges. For a sufficiently high(low) value of σ o , all bulges will be EL (NEL) based onKR, however, we also need to ensure that least fractionof bulges are left unclassified. In accordance to that, wefind that more than 60% of our total sample has σ o > ∼ σ o <
90 km/s and nearly all ( ∼ σ o is shown on the Kormendy plane.It is the most efficient of classifiers because it leaves lessthan 20% of the bulges as unclassified.If σ o is not available, then the combination of n b and C works best on KR for the separation. An analysis ofall potentialities reveals that ( C > . n b > .
0) and(
C < . n b < .
0) are the best combinations for theselection of EL and NEL bulges, respectively (Fig. 6).However, note that even the combination of n b and C isnot as effective as σ o .Since in our work, σ o is available for almost the fullsample (593 out of 605), we will employ it to classifybulges in conjunction with the Kormendy relation. Thus,those bulges which have ∆ h µ eb i < σ o > h µ eb i > σ o <
90 km/s are marked to be an indubitableclass of pseudo bulges. Out of the total sample of 593galaxies, 353 galaxies are thus classified to be discs host-ing classical bulges and 130 galaxies are classified to bedisc hosting pseudo bulges. The rest of the bulges (110, <
20% of the total) are marked as “ambiguous”. Fig. 6also shows the placement of classical, pseudo and am-biguous bulges on the Kormendy plane. Note that thiscategory “ambiguous” is not the third bulge type. It onlyindicates that these bulges could not be unambiguouslyclassified as either classical or pseudo. It is certainly possible that some of the EL bulges,possibly having ∆ h µ eb i slightly greater than zero or σ o slightly less than 130 km/s, did not get included in theclassical bulge sample. Similarly, it is also possible thatsome of the NEL bulges have been missed out of thepseudo bulge sample. However, strict criteria ensure thatthe two bulge-type samples have minimal contamination.Most importantly, the bulges which could not be indu-bitably classified into one type are saved from contam-inating the sample of the other type by being groupedinto the separate ambiguous bulge category.In Fig. 7, the placement of disc galaxies with pseudo,classical and ambiguous bulges has been depicted onthe Fundamental (Kormendy 1977) and Faber-Jackson(Faber & Jackson 1976) plane. The relation and thescatter of the Fundamental Plane is as reported byvan den Bosch (2016) for their analysis of BH-host galax-ies in K s band. They found that irrespective of thebulge nature, all galaxies follow the Fundamental Planein K s band because it efficiently reflects the total massinside effective radius. They also noted that domi-nance of the bulge can increase the tightness of the re-lation. Distribution of our sample is consistent withtheir findings (Fig. 7). The Faber-Jackson relation is asfound by Gallazzi et al. (2006), whereas, the scatter linesare those reported by later works (Cortese et al. 2014;Aquino-Ort´ız et al. 2018) for a more representative sam-ple of early type galaxies. It can be seen that classicalbulge hosting disc galaxies are the best adherents of thescaling relations found for elliptical galaxies. Althoughthis is expected because the selection of these bulges isitself based on a scaling relation, this demonstrates theeffectiveness of the Kormendy relation in bulge-type de-termination. Bimodality of structural and stellar properties
In Fig. 8, we analyse the distribution of the two bulgetypes, along with ambiguous bulges, with respect tostructural indicators and stellar parameters. In the caseof the concentration index ( C ), pseudo (PBD) and clas-sical bulge disc (CBD) galaxies exhibit two well sepa-rated peaks. All CBDs ( ∼ C more than 3.0,whereas, more than 85% of PBDs have C less than 3.5.As found earlier, n b and B/T are not as efficient dif-ferentiators of bulge types as C . Although more than80% of PBDs have n b less than 2.5, CBDs are quite uni-formly distributed over the full range. This is consistentwith the earlier and most recent findings (Gadotti 2009;Gao et al. 2020), discussed later.Our focus in this work is to examine the distributionof these bulges in terms of their stellar parameters. Wereport the presence of clear bimodality in the case of totalstellar mass ( M ∗ ), specific star formation rate (sSFR)and global colour ( r − K s ) of the galaxy. In the caseof M ∗ , log( M ∗ /M ⊙ )=10.5 marks the middle point whichcritically separates the two bulge populations. All PBDs( ∼ M ∗ lower than this value and all CBDs( ∼ M ∗ higher than this value. ABDs neatlyoccupy a narrow region in the middle area, indicatingthat as in the case of ∆ h µ eb i and σ o , sufficiently high orlow value of M ∗ can lead to unambiguous classificationof bulge type.In the case of sSFR, all PBDs ( ∼ < -1.4 and CBDs have2 Sachdeva et al. Fig. 5.—
Complementarity to Kormendy relation:
The performance of four morphology indicators ( n b , C , B/T , σ o ), in separating“elliptical like” (EL) and “non-elliptical like” (NEL) bulges (defined according to Kormendy relation), is examined. Each point on the redcurve marks the fraction of those galaxies which host EL bulges and have an indicator value higher than that point. On the contrary, eachpoint on the blue curve marks the fraction of those galaxies which have NEL bulges and have an indicator value lower than that point.The efficiency of the indicator depends on the sharpness, smoothness and stability of the increase (decrease) in galaxy fraction exhibitedby the red (blue) curve with successively increasing indicator value. Average error-bars are marked in each plot. Fig. 6.—
Efficient combinations and selection:
The first plot shows the placement of galaxies with different values of central velocitydispersion ( σ o ) on the Kormendy plane. While σ o best compliments the Kormendy relation in bulge classification, in its absence thecombination of concentration ( C ) and S´ersic index of the bulge ( n b ) turns out to be effective. The second and third plot depict thestrategies for separating “elliptical like” (red) and “non-elliptical like” (blue) bulge disc galaxies from others (green) using combination of C and n b . In the fourth plot, the selection of pseudo (blue) and classical (red) bulges, as carried out in this work, based on the combinationof Kormendy relation and σ o , is shown. The bulges which could not be unambiguously classified, as either pseudo or classical, have beenput in the “ambiguous” (green) category. Average error-bars are marked in each plot. Fig. 7.—
Fundamental and Faber-Jackson plane:
In thefirst plot, the placement of disc galaxies with pseudo (blue), clas-sical (red) and ambiguous (green) bulges is shown on the Funda-mental Plane. The plane, the relation (solid line) and the scat-ter (dotted lines) are from van den Bosch (2016) analysis of BH-host galaxies. In the second plot, their placement is shown onthe mass-based Faber-Jackson plane, where, total stellar mass ofthe galaxy ( M ∗ ) is plotted against its central velocity dispersion( σ o ). The relation (solid line) is from Gallazzi et al. (2006) and thescatter (dotted lines) is as obtained by Cortese et al. (2014) andAquino-Ort´ız et al. (2018) for their sample of early type galaxies.Average error-bars are marked in both plots. log(sSFR/(1/Gyr)) > -1.6. Similarly, in the case ofcolour ( r − K s ), the difference between the peaks of PBDs and CBDs is as large as 0.6 mag. Here, the middlepoint (1.1 mag) reports less than 10% digressions. Ourwork, thus, demonstrates that if bulges are indubitablyclassified to be pseudo and classical, their stellar prop-erties will be markedly distinct , i.e., not only all PBDswill be young and star forming, all CBDs will be oldand quiescent with negligible ( < Most effective single structural indicator
Discs hosting different bulge types have been foundto differ in their stellar properties. However, instead ofdetermining the bulge type, it will be more efficient ifa single morphological indicator can be applied for theseparation of star forming or quiescent populations. Al-though M ∗ (total stellar mass), M ∗ /R eg , M ∗ /R eg , σ o , n b ,Σ (surface mass density within 1 kpc), etc., many indi-cators have been suggested, they have not been proven tobe ultimate demarker. The issue being that substantialamount of dispersion survives and the thresholds deter-mined are at the most necessary, not sufficient, in sepa-rating star forming and quiescent populations.orrelation of structure and stellar properties of galaxies 13 Fig. 8.—
Bulge properties:
The distribution of galaxies with pseudo (blue), classical (red) and ambiguous (green) bulges is shown fordifferent structural and stellar parameters. In the first row, the three parameters are concentration ( C ), S´ersic index of the bulge ( n b ) andrelative average surface mass density within bulge effective radius (∆ h µ eb i ). In the second row, the three parameters are total stellar mass( M ∗ ), specific star formation rate (sSFR) and global colour ( r − K s ) of the galaxy. Average error-bars are marked in each plot. We will examine if our indicator ∆ h µ eb i , based on KR,is successful in solving this issue. Fig. 9 indeed revealsthat total stellar mass ( M ∗ ) of the galaxy has tightestand most well defined correlation with ∆ h µ eb i comparedto other indicators including C , n b , R eb , h µ eb i and σ o . Itdepicts that more is the total stellar mass of the galaxy,more is the stellar mass density within the effective radiusof the central bulge (Fig. 9).∆ h µ eb i , by definition, is a clear demarker of bulge-type of galaxies. We now examine its performance asa predictor of stellar activity in galaxies. In Fig. 10 wefind that the indicator exhibits an “elbow-like” correla-tion with both star formation rate (SFR) and specificSFR (sSFR) of galaxies. Here, the middle region of theelbow is dominated by galaxies with ambiguous bulges(Fig. 10). Thus, as found in the previous section, discswith ambiguous bulges are emerging to be mainly placedin the green valley. The correlation plot of the indicatorwith global colour ( r − K s ) of the galaxy provides furtherevidence in support of this argument. Interestingly, whilethe correlations with SFR and sSFR are elbow shaped,the indicator exhibits a straight correlation with colour.It shows that more is the central stellar mass density ofa galaxy, redder is its colour. Although ambiguous bulgegalaxies were depicting an equivalent or higher SFR thanpseudo bulge galaxies, they are redder in colour , suggest-ing that either they are dust ridden or have compositepopulations. This could have resulted in the difficulty inresolving these bulges into either pseudo or classical. DISCUSSION
In this work, we have explored the connection betweenstructural and stellar properties of local ( z < .
3) galax- ies by comparing the distribution (and correlation) of theparameters describing these properties for discs hostingpseudo and classical bulges. To achieve that, we haveperformed 2D bulge-disc decomposition of 1263 galax-ies in rest-frame K s band. Bulge type classification hasbeen performed based on multiple stringent complimen-tary criteria. The major findings of the study are dis-cussed below. Classification of bulge types
Over the past two decades, it has been establishedthat multiple criteria, based on the structural and kine-matic properties of the bulge, are required to be ap-plied in combination to ascertain its type. A par-ticular type of bulges, whether pseudo or classical,will not satisfy all the properties or criteria associatedwith them (reviewed in Fisher & Drory 2016; Kormendy2016). Kormendy relation - which probes the similar-ity of bulge structure with elliptical galaxies - has beenfound to be the most effective and reliable differentia-tor of bulge type (Gadotti 2009; Neumann et al. 2017;Sachdeva et al. 2019; Gao et al. 2020). Other than thestructure, it is crucial that the classification reflects theinternal kinematics of the system, i.e., classical bulges,unlike pseudo bulges, should have high velocity dis-persion (Kormendy & Kennicutt 2004; Fisher & Drory2016).Interestingly, in our work, we find that the two cri-teria, i.e., Kormendy relation (KR) and central velocitydispersion ( σ o ), are highly complimentary to each other.More than 60% of our sample has σ o >
130 km/s and allof them ( ∼ ± σ boundary of the Kormendy relation for ellipticals. For4 Sachdeva et al. Fig. 9.—
Complimenting mass:
The correlation of all morphological indicators is examined with respect to the total stellar mass( M ∗ ) of the galaxy to select a single morphological indicator which is most adept at separating star forming and quiescent populations. Inthe first row, the correlation is explored with concentration ( C ), bulge S´ersic index ( n b ) and bulge effective radius ( R eb ). In the secondrow, the correlation is explored with average surface mass density within bulge effective radius ( h µ eb i ), central velocity dispersion ( σ o ) andrelative average surface mass density within bulge effective radius (∆ h µ eb i ). Average error-bars are marked in each plot. the rest 40%, a large proportion ( ∼ σ o < ∼ < σ o and the two planes is in-between thetwo bulge types, which is the reason for ambiguity intheir classification. Bimodality of bulge properties
After ensuring that we have a well separated set of in-dubitably pseudo and classical bulges, we have analyzedthe distribution of these two bulges types in terms oftheir structural and stellar parameters.In terms of structural parameters, we have found con-centration C to be a better differentiator of bulge typesthan n b and B/T . This is consistent with the studyof bulges in the CALIFA sample (Neumann et al. 2017)that reported C to be the closest follower of the Kor-mendy relation compared to all other structural indica-tors. We report that both n b and B/T are inefficientclassifiers of bulge type. While pseudo bulges are mainly concentrated below n b ∼ .
0, classical bulges are uni-formly distributed over the full range of n b . This isconsistent with earlier studies, both based on statisti-cally large samples as well as individual decomposition ofsmaller samples Gadotti (2009); Gao et al. (2020). Ourresults for B/T are also consistent with all previous stud-ies (Gadotti 2009; Fisher & Drory 2016; Neumann et al.2017; Gao et al. 2020) reporting the inefficiency of thisindicator due to substantial overlap between the distri-bution of the two bulge types. The argument proposedis that the classical bulge - being a merger remnant- is independent of the host disc and can easily be anon-dominant component of the galaxy (Fisher & Drory2008; Gadotti 2009). We note that all our findingsare consistent with the latest work on classifying bulges(Gao et al. 2019, 2020), where, 320 galaxies of a local,well-resolved sample have been individually decomposedinto multiple components in the optical.Our major focus is to examine if morphology of thebulges is correlated with their stellar activity. Bimodal-ity, with regard to stellar parameters, has not been ob-served by earlier studies, i.e., either a large fractionof pseudo bulges are found in the territory of classi-cal bulges or vice-versa (Gadotti 2009; Fisher & Drory2016; Luo et al. 2020). Crucially, which side contributesto the digression is also not agreed upon. In our study,we find that pseudo and classical bulges are well sep-arated in terms of all stellar parameters. All pseudobulge galaxies have total stellar mass ( M ∗ ) less than10 . M ⊙ and all classical bulge galaxies have M ∗ morethan that, with negligible ( < Fig. 10.—
SFR and colour:
The correlation of the morphologyindicator based on the Kormendy relation (∆ h µ eb i ) is exploredwith stellar parameters which include star formation rate (SFR),specific SFR (sSFR) and global colour ( r − K s ) of the galaxy. Whilein the case of SFR and sSFR, elbow-like pattern appears, there is astraight correlation in the case of colour. The galaxies are markedaccording to the bulge they host, i.e., pseudo (blue), classical (red)and ambiguous (green). Average error-bars are marked in eachplot. − . Gyr − marks the transition point, again with in-significant ( < r − K s ) of pseudo and classical bulge host galax-ies differ by ∼ . . Single structural indicator for quenching
The finding that galaxies with different bulge typesdiffer in their stellar properties, highlights that struc-ture of a galaxy is correlated with its stellar parameters.Thus, there is a search for a single structural indicatorwhich is the neatest separator of star forming and quies-cent galaxy populations. Cheung et al. (2012) exploredall potential structural indicators, including n b , B/T , C , M ∗ /R eb , to report that Σ (stellar mass surface densitywithin 1 kpc) is the best in this regard. Adding to that,recent work has claimed that ∆Σ (Σ with mass trendremoved) is not only the best predictor of quenching butalso the best indicator of galaxy’s bulge type (Luo et al.2020). They find that the reason for the success of ∆Σ lies in its similarity to ∆ h µ eb i which is most suitable.Our results based on ∆ h µ eb i in K s band add clarity tothis picture.As observed in the case of ∆Σ (Luo et al. 2020),∆ h µ eb i also portrays an elbow-like pattern with bothSFR and sSFR of the galaxy. Based on the “elbow-like”structure they argued that while all pseudo bulge galax-ies are star forming, classical bulge galaxies span the fullrange of stellar activity, where, those in the middle re-gion are transiting from pseudo to classical stage. Ourwork reveals that all pseudo bulge galaxies are star form-ing, all classical bulge galaxies are quiescent and galaxieswith ambiguous bulges are mainly in the elbow (or mid-dle) region - possibly the green valley.Most importantly, the elbow-like pattern transformsinto a straight correlation when SFR (or sSFR) is re-placed with the global colour ( r − K s ) of the galaxy.Thus, although ambiguous bulge galaxies have simi-lar SFR as pseudo bulge galaxies, they are redder incolour suggesting that either they are dust ridden orhave composite populations. It is possible that the pres-ence of dust has been the cause for the ambiguity inbulge type determination. Other than that, the place-ment of ambiguous bulge disc galaxies in our work sug-gests that they are dominantly in the green valley, con-stantly found between pseudo bulge hosting (blue) andclassical bulge hosting (red) disc galaxies. Green val-ley galaxies, in addition to being dusty, have been ob-served to be a complex mixture of morphology and stellarpopulations (Schawinski et al. 2014; Kelvin et al. 2018;Phillipps et al. 2019; Angthopo et al. 2020). A more ex-haustive analysis of discs with ambiguous bulges, includ-ing the fitting of multiple components and accounting forthe dust and gas fraction, has the potential to provideinsight regarding the green valley.This work is supported by the National Key R&DProgram of China (2016YFA0400702) and the NationalScience Foundation of China (11721303, 11991052). SSacknowledges support from China Postdoctoral ScienceFoundation fellowship (2019M660299). We are thankfulto the anonymous referee for careful reading and use-ful comments. SS is thankful to James Geach and Yen-Ting Lin for providing access to the data and detailingimage specifications. SS is also thankful to Sandra M.Faber, Yingjie Peng and Hassen M. Yesuf for useful dis-cussions. FS acknowledges partial support from a Lever-hulme trust Research Fellowship. REFERENCESAguado, D. S., Ahumada, R., Almeida, A., et al. 2019, ApJS, 240,23 Akhlaghi, M., & Ichikawa, T. 2015, ApJS, 220, 16 Sachdeva et al.