Deep Extragalactic VIsible Legacy Survey (DEVILS): Stellar Mass Growth by Morphological Type since z = 1
Abdolhosein Hashemizadeh, Simon P. Driver, Luke J. M. Davies, Aaron S. G. Robotham, Sabine Bellstedt, Rogier A. Windhorst, Malcolm Bremer, Steven Phillipps, Matt Jarvis, Benne W. Holwerda, Claudia del P. Lagos, Soheil Koushan, Malgorzata Siudek, Natasha Maddox, Jessica E. Thorne, Pascal Elahi
MMNRAS , 1–24 (2021) Preprint 1 March 2021 Compiled using MNRAS L A TEX style file v3.0
Deep Extragalactic VIsible Legacy Survey (DEVILS):Stellar Mass Growth by Morphological Type since z = 1 Abdolhosein Hashemizadeh (cid:63) , Simon P. Driver , Luke J. M. Davies ,Aaron S. G. Robotham , Sabine Bellstedt , Rogier A. Windhorst , Malcolm Bremer ,Steven Phillipps , Matt Jarvis , , Benne W. Holwerda , Claudia del P. Lagos ,Soheil Koushan , Malgorzata Siudek , , Natasha Maddox , Jessica E. Thorne ,Pascal Elahi ICRAR, The University of Western Australia, 35 Stirling Highway, Crawley WA 6009, Australia School of Earth and Space Exploration, Arizona State University, Tempe, AZ 85287-1404 Astrophysics Group, School of Physics, University of Bristol, Tyndall Avenue, Bristol BS8 1TL, UK Department of Astrophysics, University of Oxford, The Denys Wilkinson Building, Keble Road, Oxford OX1 3RH, UK Department of Physics, University of the Western Cape, Bellville 7535, South Africa Department of Physics and Astronomy, University of Louisville, Natural Science Building 102, 40292 KY Louisville, USA Institut de F´ısica d’Altes Energies (IFAE), The Barcelona Institute of Science and Technology, 08193 Bellaterra (Barcelona), Spain National Centre for Nuclear Research, ul. Hoza 69, 00-681 Warsaw, Poland Faculty of Physics, Ludwig-Maximilians-Universit¨at, Scheinerstr. 1, 81679 Munich, Germany
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
Using high-resolution Hubble Space Telescope imaging data, we perform a visual mor-phological classification of ∼ ,
000 galaxies at z < z > z ∼
1. Double-componentgalaxies dominate the SMD at all epochs and increase in their contribution to the stel-lar mass budget to the present day. Elliptical galaxies are the second most dominantmorphological type and increase their SMD by ∼ . ∼ z <
1, and conclude that double-component galaxies arepredominantly being built by the in-situ evolution in disks (apparent as the growthof the low-mass end with time), while mergers are likely responsible for the growth ofellipticals (apparent as the increase of intermediate/high-mass end).
Key words: galaxies: formation - galaxies: evolution - galaxies: bulges - galaxies: disk- galaxies: elliptical - galaxies: mass function - galaxies: structure - galaxies: general
The galaxy population in the local Universe is observed tobe bimodal. This bimodality manifests in multiple prop- (cid:63)
E-mail: [email protected] erties such as colour, morphology, metallicity, light profileshape and environment (e.g. Kauffmann et al. 2003; Baldryet al. 2004; Brinchmann et al. 2004). This bimodality is alsofound to extend to earlier epochs (see e.g., Strateva et al.2001; Hogg et al. 2002; Bell et al. 2004; Driver et al. 2006;Taylor et al. 2009; Brammer et al. 2009; Williams et al. © a r X i v : . [ a s t r o - ph . GA ] F e b Hashemizadeh et al. . -10 M (cid:12) and a steepening to lower masses (for exam-ple see: Baldry et al. 2008; Peng et al. 2010; Baldry et al.2012; Kelvin et al. 2014; Weigel et al. 2016; Moffett et al.2016; Wright et al. 2017 ). Several studies have also investi-gated the evolution of the SMF at higher redshifts (Pozzettiet al. 2010; Muzzin et al. 2013; Whitaker et al. 2014; Lejaet al. 2015; Mortlock et al. 2015; Wright et al. 2018; Kaw-inwanichakij et al. 2020). A fingerprint of this bimodality isalso observed in the double-component Schechter functionrequired to fit the local SMF (e.g. Baldry et al. 2012; Wrightet al. 2018). At least two distinct galaxy populations corre-sponding to star-forming and passive systems are thoughtto be the origin of this bimodal shape, with star-forming systems dominating the low-mass tail and passive galaxiesdominating the high-mass “hump” of the SMF (Baldry et al.2012; Muzzin et al. 2013; Wright et al. 2018).Many previous studies have separated galaxies into twomain populations of star forming and passive (or a proxythereof), and measured their individual stellar mass assem-blies (e.g. Pozzetti et al. 2010; Tomczak et al. 2014; Lejaet al. 2015; Davidzon et al. 2017). For example, by separat-ing their sample into early- and late-type galaxies based oncolour, Vergani et al. (2008) studied the SMF of the sam-ples and confirmed that ∼
50% of the red sequence galaxieswere already formed by z ∼
1. Pannella et al. (2006) useda sample of ∼ z ∼ z < . ≤ z ≤ ∗ > . , selected from theDEVILS sample and analysis, for which classification is re-liable. This intermediate redshift range is a key phase in theevolution of the Universe where a large fraction ( ∼ MNRAS , 1–24 (2021)
EVILS: Mass Growth of Morphological Types the various systems within which stars are located, andhow they evolve. In this paper, we investigate the evolu-tion of the stellar mass function of different morphologicaltypes and explore the contribution of each morphology tothe global stellar mass build-up of the Universe. Sheddinglight on the redistribution of stellar mass in the Universeand also the transformation and redistribution of the stellarmass between different morphologies likely explains the ori-gin of the bimodality in galaxy populations observed in thelocal Universe.The ultimate goal of this study and its companion pa-pers is to not only probe the evolution of the morphologi-cal types but also study the formation and evolution of thegalaxy structures, including bulges and disks. This will bepresented in a companion paper (Hashemizadeh et al. inprep.), while in this paper we explore the visual morpholog-ical evolution of galaxies and conduct an assessment into thepossibility of the bulge formation scenarios by constructingthe global stellar mass distributions and densities of variousmorphological types.The data that we use in this work are presented in Sec-tion 8. In Section 1.3, we define our sample and sample se-lection. Section 2 presents different methods that we explorefor the morphological classification. In Section 3, we describethe parameterisation of the SMF. We then show the total SMF at low- z in Section 3.1. Section 3.2 describes the ef-fects of the cosmic large scale structure due to the limitingsize of the DEVILS/COSMOS field, and the technique weuse to correct for this. The evolution of the SMF and theSMD and their subdivision by morphological type are dis-cussed in Sections 3.3-4. We finally discuss and summarizeour results in Sections 5-6.Throughout this paper, we use a flat standard ΛCDMcosmology of Ω M = 0 .
3, Ω Λ = 0 . H =70kms − Mpc − . Magnitudes are given in the AB system. The Deep Extragalactic VIsible Legacy Survey (DEVILS)(Davies et al. 2018), is an ongoing magnitude-limited spec-troscopic and multi-wavelength survey. Spectroscopic ob-servations are currently being undertaken at the Anglo-Australian Telescope (AAT), providing spectroscopic red-shift completeness of >
95% to Y-mag < . ∼
25 mag in the D10 region.The objective of the DEVILS campaign is to obtain asample with high spectroscopic completeness extending overintermediate redshifts (0 . < z < . Figure 1.
Background shows the ACS/F814W mosaic image ofthe COSMOS field. The cyan rectangle represents the D10 regionin the DEVILS survey. Gold points are the D10/ACS sources usedin this work consisting of ∼
36k galaxies. See section 1.3 for moredetail.
COSMOS). In this work, we only explore the D10 region asthis overlaps with the HST COSMOS imaging. The spec-troscopic redshifts used in this paper are from the DEVILScombined spectroscopic catalogue which includes all avail-able redshifts in the COSMOS region, including zCOSMOS(Lilly et al. 2007, Davies et al. 2015), hCOSMOS (Dam-janov et al. 2018) and DEVILS (Davies et al. 2018). As theDEVILS survey is ongoing the spectroscopic observations forour full sample are still incomplete (the completeness of theDEVILS combined data is currently ∼
90 per cent to Y-mag= 20). Those objects without spectroscopic redshifts are as-signed photometric redshifts in the DEVILS master redshiftcatalogue (
DEVILS D10MasterRedshiftCat v0.2 catalogue),described in detail in Thorne et al. (2020). In this work,we also use the stellar mass measurements for the D10 re-gion (
DEVILS D10ProSpectCat v0.3 catalogue) reported byThorne et al. (2020). Briefly, to estimate stellar massesthey used the
ProSpect
SED fitting code (Robothamet al. 2020) adopting Bruzual & Charlot (2003) stellar li-braries, the Chabrier (2003) IMF together with Charlot& Fall (2000) to model dust attenuation and Dale et al.(2014) to model dust emission. This study makes use ofthe new multiwavelength photometry catalogue in the D10field (
DEVILS PhotomCat v0.4 ; Davies et al. in prep.) andfinds stellar masses ∼ . MNRAS000
DEVILS PhotomCat v0.4 ; Davies et al. in prep.) andfinds stellar masses ∼ . MNRAS000 , 1–24 (2021)
Hashemizadeh et al.
Figure 2.
A random galaxy in our sample at redshift z ∼ . gri colour image. The image shows a cutoutof 5 × R90 on each side, where R90 = 3 .
23 arcsec measured fromUltraVISTA Y-band (Davies et al. 2018).
The Cosmic Evolution Survey (COSMOS) is one of the mostcomprehensive deep-field surveys to date, covering almost 2contiguous square degrees of sky, designed to explore largescale structures and the evolution of galaxies, AGN and darkmatter (Scoville et al. 2007). The high resolution HubbleSpace Telescope (HST) F814W imaging in COSMOS allowsfor the study of galaxy morphology and structure out tothe detection limits. In total COSMOS detects ∼ <
100 pc (Scoville et al. 2007).The COSMOS region is centred at RA = 150 .
121 (10 :00 : 28 . .
21 (+02 : 12 : 21 .
00) (J2000)and is supplemented by 1.7 square degrees of imaging withthe Advanced Camera for Surveys (ACS ) on HST. This1.7 square degree region was observed during 590 orbitsin the F814W (I-band) filter and also, 0.03 square degreeswith F475W (g-band). In this study we exclusively use theF814W filter, not only providing coverage but also suitablerest-frame wavelength for the study of optical morphologyof galaxies out to z ∼ as fits images. irsa.ipac.caltech.edu/data/COSMOS/images/acs 2.0/I/ As part of the DEVILS survey we have selected an HSTimaging sample with which to perform various morphologi-cal and structural science projects. Figure 1, shows the ACSmosaic imaging of the COSMOS field with the full D10 re-gion overlaid as a cyan rectangle (defined from the UltraVISTA imaging). Our final sample, D10/ACS, is the com-mon subset of sources from ACS and D10. The positionof the D10/ACS sample on the plane of RA and DEC isoverplotted on the same figure as yellow dots. Note that asshown in Figure 1, we exclude objects in the jagged area ofthe ACS imaging leading to a rectangular effective area of1 . z ∼ . ,
264 spectroscopic red-shifts are available in the D10 region (excluding the jaggededges) out of which 2 ,
903 redshifts are observed by DEV-ILS. See the DEVILS website for a full description of thesedata. We select 284 random galaxies drawn from across theentire redshift and stellar mass distribution, and visually in-spect them. These 284 galaxies are shown as circles in Figure3. From our visual inspection, we identify the boundarieswithin which we believe the majority of galaxies are suffi-ciently resolved that morphological classifications should bepossible (i.e., not too small or faint). Our visual assessmentsare indicated by colour in Figure 3 showing two-component(grey), single component (blue), and problematic cases (red;merger, disrupted, and low S/N). Unsurprisingly, in agree-ment with other studies (e.g. Conselice et al. 2005), we findthe fraction of problematic galaxies increases drastically athigh redshifts ( z > . ∼ compact , hence the notion ofgalaxies as predominantly bulge plus disk systems becomesuntenable. Note that the vast majority of galaxies in thisredshift range are disturbed interacting systems, howeverour observation of a fraction of the clumpy galaxies couldbe due to the fact that the bluer rest-frame emission is moresensitive to star forming regions dominating the flux. Theclumps and potential bulges at this epoch, are mostly com-parable or smaller than the HST PSF, hence conventional2D bulge+disk fits are also unlikely to be credible even atHST resolution. https://devilsurvey.org MNRAS , 1–24 (2021) EVILS: Mass Growth of Morphological Types a b c F i g u r e . ( a ) T h e r e l a t i o nb e t w ee n s t e ll a r m a ss a nd l oo k b a c k t i m e ( r e d s h i f t) o f o u r s a m p l e . A ll ga l a x i e s a r e s h o w n i n c y a n i nb a c k g r o und . C i r c l e s r e p r e s e n t l a x i e s t h a t w e r a nd o m l y s a m p l e f o r i n i t i a l v i s u a li n s p ec t i o n . G r e y a ndb l u ec i r c l e ss h o w d o ub l e a nd s i n g l ec o m p o n e n t ga l a x i e s , r e s p ec t i v e l y . R e d s y m b o l s a r ec o m p li c a t e d ga l a x i e s w h i c h c o n s i s t o f m e r g i n g s y s t e m s , p e r t u r b e d ga l a x i e s ,l o w S / N o r h i g h r e d s h i f t c l u m p y ga l a x i e s . T h e d o tt e d r ec t a n g l ec o rr e s p o nd s t oo u r i n i t i a l s a m p l e r e g i o n ( z < . nd l og ( M ∗ / M (cid:12) ) > ) . T h e s o li d r ec t a n g l e s h o w s o u r fin a l s a m p l e r e g i o n w h i c h c o v e r s ga l a x i e s up t o z < . nd l og ( M ∗ / M (cid:12) ) > . . T h e s o li db l a c k li n e r e p r e s e n t s t h e s p li n e fi tt o t h e m o d a l v a l u e o f t h e s t e ll a r m a ss i nb i n s o f l oo k b a c k t i m e i nd i c a t i n g t h e s t e ll a r m a ss c o m p l e t e n e ss . B l a c k d a s h e d li n e s h o w s T h o r n ee t a l. ( ) c o m p l e t e n e ss li m i t . S ee t e x t f o r m o r e d e t a il s . P a n e l s ( b ) a nd ( c ) d i s p l a y t h e d i s t r i bu t i o n o f s t e ll a r m a ss a nd r e d s h i f t o f o u r fin a l s a m p l e ( i. e . w i t h i n t h e s o li d r ec t a n g l e ) . N o t e t h a tt h e P D F s a r e s m oo t h e db y a k e r n e l w i t h s t a nd a r dd e v i a t i o n o f . . MNRAS000
903 redshifts are observed by DEV-ILS. See the DEVILS website for a full description of thesedata. We select 284 random galaxies drawn from across theentire redshift and stellar mass distribution, and visually in-spect them. These 284 galaxies are shown as circles in Figure3. From our visual inspection, we identify the boundarieswithin which we believe the majority of galaxies are suffi-ciently resolved that morphological classifications should bepossible (i.e., not too small or faint). Our visual assessmentsare indicated by colour in Figure 3 showing two-component(grey), single component (blue), and problematic cases (red;merger, disrupted, and low S/N). Unsurprisingly, in agree-ment with other studies (e.g. Conselice et al. 2005), we findthe fraction of problematic galaxies increases drastically athigh redshifts ( z > . ∼ compact , hence the notion ofgalaxies as predominantly bulge plus disk systems becomesuntenable. Note that the vast majority of galaxies in thisredshift range are disturbed interacting systems, howeverour observation of a fraction of the clumpy galaxies couldbe due to the fact that the bluer rest-frame emission is moresensitive to star forming regions dominating the flux. Theclumps and potential bulges at this epoch, are mostly com-parable or smaller than the HST PSF, hence conventional2D bulge+disk fits are also unlikely to be credible even atHST resolution. https://devilsurvey.org MNRAS , 1–24 (2021) EVILS: Mass Growth of Morphological Types a b c F i g u r e . ( a ) T h e r e l a t i o nb e t w ee n s t e ll a r m a ss a nd l oo k b a c k t i m e ( r e d s h i f t) o f o u r s a m p l e . A ll ga l a x i e s a r e s h o w n i n c y a n i nb a c k g r o und . C i r c l e s r e p r e s e n t l a x i e s t h a t w e r a nd o m l y s a m p l e f o r i n i t i a l v i s u a li n s p ec t i o n . G r e y a ndb l u ec i r c l e ss h o w d o ub l e a nd s i n g l ec o m p o n e n t ga l a x i e s , r e s p ec t i v e l y . R e d s y m b o l s a r ec o m p li c a t e d ga l a x i e s w h i c h c o n s i s t o f m e r g i n g s y s t e m s , p e r t u r b e d ga l a x i e s ,l o w S / N o r h i g h r e d s h i f t c l u m p y ga l a x i e s . T h e d o tt e d r ec t a n g l ec o rr e s p o nd s t oo u r i n i t i a l s a m p l e r e g i o n ( z < . nd l og ( M ∗ / M (cid:12) ) > ) . T h e s o li d r ec t a n g l e s h o w s o u r fin a l s a m p l e r e g i o n w h i c h c o v e r s ga l a x i e s up t o z < . nd l og ( M ∗ / M (cid:12) ) > . . T h e s o li db l a c k li n e r e p r e s e n t s t h e s p li n e fi tt o t h e m o d a l v a l u e o f t h e s t e ll a r m a ss i nb i n s o f l oo k b a c k t i m e i nd i c a t i n g t h e s t e ll a r m a ss c o m p l e t e n e ss . B l a c k d a s h e d li n e s h o w s T h o r n ee t a l. ( ) c o m p l e t e n e ss li m i t . S ee t e x t f o r m o r e d e t a il s . P a n e l s ( b ) a nd ( c ) d i s p l a y t h e d i s t r i bu t i o n o f s t e ll a r m a ss a nd r e d s h i f t o f o u r fin a l s a m p l e ( i. e . w i t h i n t h e s o li d r ec t a n g l e ) . N o t e t h a tt h e P D F s a r e s m oo t h e db y a k e r n e l w i t h s t a nd a r dd e v i a t i o n o f . . MNRAS000 , 1–24 (2021)
Hashemizadeh et al.
Note that as can be seen in Figure 3, we also start tosuffer significant mass incompleteness at low redshifts dueto the COSMOS F814W sensitivity limit. The sample com-pleteness limit is shown here as a smooth spline fitted to themodal values of stellar mass in bins of lookback time (shownas a solid line), i.e., peaks of the stellar mass histograms innarrow redshift slices. We also show the completeness limitreported by Thorne et al. (2020) and shown as a dashedline in Figure 3, which is based on a g − i colour analy-sis and is formulated as log(M ∗ ) = 0 . t + 7 .
25, where t islookback time in Gyr. From this initial inspection we defineour provisional window of 105,185 galaxies at z < . ∗ / M (cid:12) ) >
9, shown as a dotted rectangle in Figure 3.To further tune this selection, we then generatedpostage stamps of 4 ,
000 random galaxies (now initial se-lection) using the HST F814W imaging and colour gri in-sets from the Subaru Suprime-Cam data (Taniguchi et al.2007). Figure 2 shows an example of the cutouts we gen-erated for our visual inspection. The postage stamps aregenerated with 5 × R90 on each side, where R90 is the ra-dius enclosing 90% of the total flux in the UltraVISTA Y-band (soon to be presented in DEVILS photometry cata-logue, Davies et al. in prep.). These stamps were indepen-dently reviewed by five authors (AH, SPD, ASGR, LJMD,SB) and classified into single component, double componentand complicated systems (hereafter: hard ). The single com-ponent systems were later subdivided into disk or ellipticalsystems. Note that the hard class consists of asymmetric,merging, clumpy, extremely compact and low-S/N systems,for which 2D structural decomposition would be unlikely toyield meaningful output. Objects with three or more votesin one category were adopted and more disparate outcomesdiscussed and debated until an agreement was obtained. Inthis way, we established a “gold calibration sample” of 4kgalaxies to justify our final redshift and stellar mass range,and for later use as a training sample in our automated-classification process, see Section 2 for more details.Figure 4 shows the fraction of each of the above classifi-cations versus redshift and total stellar mass (dashed lines).As the left panel shows, the fraction of hard galaxies (graydashed line) drastically increases at z >
1. At the highestredshift of our sample, z ∼ .
4, 40 percent of the galaxiesare deemed unfittable, or at least inconsistent with the no-tion of a classical/pseudo-bulge plus disk systems (this isconsistent with Abraham et al. 1996a and Conselice et al.2005). Also see the review by Abraham & van den Bergh(2001). We therefore further restrict our redshift range to z ≤ ∗ / M (cid:12) ) = 9 . M ∗ - z parameter space leads us to a final sample of galax-ies for which we can confidently study their morphology andstructure. We conclude that we can study the structure ofgalaxies up to z ∼ ∗ / M (cid:12) ) ≥ .
5. Withinthis selection, our sample consists of 35 ,
803 galaxies with14 ,
036 available spectroscopic redshifts.
In this section, we initially aim to develop a semi-automaticmethod for morphological classification. Our ultimate goalwould be to reach a fully automatic algorithm for classify-ing galaxies into various morphological classes. While thisultimately proves unsuitable we discuss it here to explainwhy we finally visually inspect all systems. These classes areas mentioned above bulge+disk (BD; double-component), pure-disk (D), elliptical (E) and hard (H) systems. In or-der to overcome this problem we test various methods in-cluding: cross-matching with Galaxy Zoo, Hubble catalogue(GZH) and Zurich Estimator of Structural Type (ZEST)catalogues. In the end, none of the methods proved to berobust and we resort to full visual inspection.
A large fraction of COSMOS galaxies with I F W < . pure-disk galaxy. To distinguishbetween them, we make use of the single S´ersic index (n). Asshown in the flowchart, n > . n ≤ . pure-disk , respectively. The S´ersic indices aretaken from our structural analysis which will be describedin Hashemizadeh et al. (in prep.). In order to capture theS0 galaxies or double-component systems with smooth diskprofiles we add an extra condition as to whether there are atleast 2 Galaxy Zoo votes for a prominent bulge. If so thenit is likely that the galaxy is a double-component system.An advantage of using the GZH decision tree is that it canidentify a vast majority of galaxies with odd features such asmerger-induced asymmetry etc. Table 1 shows a confusionmatrix comparing the GZH predictions with the visual in-spection of our 4k gold sample as the ground truth. Doublecomponent galaxies are predicted by the GZH with maxi-mum accuracy 81%. Single component (i.e. pure-disk and elliptical galaxies) and hard galaxies are predicted at a sig-nificantly lower accuracy with high misclassification rates.We further visually inspected misclassified galaxies and donot find good agreement between our classification and thoseof GZH. As such, we do not trust the GZH classification forour sample. MNRAS , 1–24 (2021)
EVILS: Mass Growth of Morphological Types Figure 4.
The number fraction of each morphological type (BD: bulge+disk , D: pure-disk , E: elliptical , C: compact , H: hard ) in binsof lookback time (redshift) and total stellar mass, left and right panels, respectively. Each colour represents a morphology as indicatedin the inset legend. Dashed lines are the ∼ z < .
4) while solid lines are our full visualinspection of the final sample. Shaded stripes around lines display 95% confidence intervals from beta distribution method as calculatedusing prop.test in R package stats . Another available morphological catalogue for COSMOSgalaxies is the Zurich Estimator of Structural Type (ZEST)(Scarlata et al. 2007a). In ZEST, Scarlata et al. (2007a) usetheir single-S´ersic index (n) as well as five other diagnostics:asymmetry, concentration, Gini coefficient, second-ordermoment of the brightest 20% of galaxy flux, and elliptic-ity. ZEST includes a sample of ∼ ,
000 galaxies withI F W ≤
24. More than 90% of our sample is cross-matchedwith ZEST which we use as a complementary morphologicalclassification. We use the ZEST TYPE flag with four valuesof 1,2,3,9 representing early type galaxy, disk galaxy,irregular galaxy and no classification. For Type 2 (i.e. diskgalaxy) we make use of the additional flag BULG, whichindicates the level of bulge prominence. BULG is flaggedby five integers as follows: 0 = bulge dominated, 1,2 =intermediate-bulge, 3 = pure-disk and 9 = no classification.We present the accuracy of ZEST predictions in a confusionmatrix in Table 2 which shows we do not find an accuratemorphological prediction from ZEST in comparison to ourgold calibration sample. While double-component systemsare classified with an accuracy of 74%, other classes areclassified poorly with high error ratios. We confirm this byvisual inspection of the misclassified objects where we stillfavour our visual classification.
Table 1.
The confusion matrix comparing the morphological pre-dictions of the GZH with our visual inspection of 4k gold calibra-tion sample as the ground truth. For example, 0.81 means 81% ofdouble component galaxies (in our visual classification) are alsocorrectly classified by the GZH.GZH PredGround Truth Double Hard SingleDouble
Overall, by analysing both of the above catalogues wedo not find them to be sufficiently accurate for predictingthe proper morphologies of galaxies when we compare theirestimates with our 4k gold calibration visual classification.
As no automatic classification robustly matches our gold cal-ibration sample we opt to visually inspect all galaxies in ourfull sample. However, we can use the predictions from GZHand ZEST as a pre-classification decision and put galaxiesin master directories according to their prediction. The hard class is adopted from the GZH prediction as it is shown toperform well in detecting galaxies with odd features. In addi-
MNRAS000
MNRAS000 , 1–24 (2021)
Hashemizadeh et al.
Figure 5.
This flowchart shows the decision tree that we adopt to translate Galaxy Zoo: Hubble (GZH) tasks and outputs into thedesired morphologies to be used in this study. The weight of final arrows are proportional to the number of galaxies following thosepaths. In addition, the fraction of galaxies falling into each morphological category is annotated.
Table 2.
The confusion matrix comparing the morphological pre-dictions of the ZEST catalogue with our visual inspection of 4kgold calibration sample as the ground truth.ZEST PredGround Truth Double Hard SingleDouble tion, from analysing the distribution of the half-light radius(R50) of our 4k gold calibration sample, extracted from ourDEVILS photometric analysis, we know that resolving thestructure of galaxies with a spatial size of R50 ≤ .
15 arcsecis nearly impossible. R50 is measured by using the
Pro-Found package (Robotham et al. 2018), a tool written inthe R language for source finding and photometric analysis.So, we put these galaxies into a separate compact class (C).Having pre-classified galaxies, we now assign each classto one of our team members (AH, SPD, ASGR, LJMD,SB) so each of the authors is independently a specialistin, and responsible for, only one morphological class. Ini-tially, we inspect galaxies and relocate incorrectly classifiedgalaxies from our master directories to transition folders for further inspections by the associated responsible person. Inthe second iteration, the incorrectly classified systems willbe reclassified and moved back into master directories. Theleft-over sources in the transition directories are thereforeambiguous. For these galaxies, all classifiers voted and weselected the most voted class as the final morphology. Asa final assessment and quality check three of our authors(SPD, LJMD, SB) independently reviewed the entire clas-sifications ( ∼ k each) and identified ∼ k that were stillfelt to be questionable. These three classifiers then indepen-dently classified these 7k objects. The final classification waseither the majority decision or in the case of a 3-way diver-gence the classification of SPD.Figure 6 shows the stellar mass versus redshift planecolour coded according to the level of agreement betweenour 3 classifiers. Colour indicates the percentage of objectsin each cell consistently classified by at least two classifiers,obtained as follows: Agreement = Number of objects with two or more agreement in the cellTotal number of objects in the cell , (1)This figure implies that we have the highest agreement( ∼ ∼ − MNRAS , 1–24 (2021)
EVILS: Mass Growth of Morphological Types Figure 6.
Stellar mass versus redshift, colour coded by the level of agreement in our final classifications (by three co-authors), i.e. thepercentage of objects in each cell consistently classified by at least two of our classifiers. Note that we increase the resolution of the colourmap of agreement (cid:38) .
92 to highlight the variation of agreement in this level. See the text for more details. ure indicates that, on average, our visual classifications per-formed independently by different team members agree by ∼
95% over the complete sample.We report the number of objects in each morphologicalclass in Table 3. 44 .
4% of our sample (15,931) are classifiedas double-component (BD) systems. We find 13,212 pure-disk (D) galaxies (37%) while only 3,890 (11%) elliptical (E) galaxies. The
Compact (C) and hard (H) systems, intotal, occupy 7 .
6% of our D10/ACS sample.The visual morphological classification is released inthe team-internal DEVILS data release one (DR1) inD10VisualMoprhologyCat data management unit (DMU).
The morphology of a subset of nearby galaxy classes showssome significant dependence on rest-frame wavelengths, es-pecially below the Balmer break and towards the UV (e.g.Hibbard & Vacca 1997; Windhorst et al. 2002; Papovichet al. 2003; Taylor et al. 2005; Taylor-Mager et al. 2007;Huertas-Company et al. 2009; Rawat et al. 2009; Holwerdaet al. 2012; Mager et al. 2018). This “Morphological K-correction” can be quite significant, even between the B- andnear-IR bands (e.g., Knapen & Beckman 1996), and mustbe quantified in order to distinguish genuine evolutionary effects from simple bandpass shifting. Hence, the results offaint galaxy classifications may, to some extent, depend onthe rest-frame wavelength sampled.Our D10/ACS classifications are done in the F814W fil-ter (I-band), and the largest redshift in our sample, z ∼ ∼
412 nm (B-band), so for our particular case, themain question is, to what extent galaxy rest-frame morphol-ogy changes significantly from 412–823 nm across the BVRIfilters. Here we briefly summarize how any such effects mayaffect our classifications.Windhorst et al. (2002) imaged 37 nearby galaxies ofall types with the Hubble Space Telescope (HST), gatheringavailable data mostly at 150, 255, 300, 336, 450, 555, 680,and/or 814 nm, including some ground based images to com-plement filters missing with HST. These nearby galaxies areall large enough that a ground-based V-band image yieldsthe same classification as an HST F555W or F606W im-age. These authors conclude that the change in rest-framemorphology going from the red to the mid–far UV is morepronounced for early type galaxies (as defined at the tradi-tional optical wavelengths or V-band), but not necessarilynegligible for all mid-type spirals or star-forming galaxies.Late-type galaxies generally look more similar in morphol-ogy from the mid-UV to the red due to their more uniformand pervasive SF that shows similar structure for young–oldstars in all filters from the mid-UV through the red. Wind-
MNRAS000
MNRAS000 , 1–24 (2021) Hashemizadeh et al.
Figure 7.
Top panel: the PDF of the specific star formationrate (sSFR = SFR/M ∗ ) of the same morphologies. Bottom panel:the stellar mass probability density function (PDF) of threemorphologies in our sample (all redshifts), i.e. pure-disk (cyan), double-component (green) and elliptical (magenta) galaxies. Notethat for clarity, the PDFs are slightly smoothed by a kernel withstandard deviation of 0.02. horst et al. (2002) conclude qualitatively that in the rest-frame mid-UV, early- to mid-type galaxies are more likelyto be misclassified as later types than late-type galaxies arelikely to be misclassified as earlier types.To quantify these earlier qualitative findings regardingthe morphological K-correction, much larger sample of 199nearby galaxies across all Hubble types (as defined in V-band) was observed by Taylor et al. (2005) and Taylor-Mager et al. (2007), and 2071 nearby galaxies were simi-larly analyzed by Mager et al. (2018). They determined theirSB-profiles, radial light-profiles, color gradients, and CASparameters (Concentration index, Asymmetry, and Clumpi-ness; e.g. Conselice 2004) as a function of rest-frame wave-length from 150-823 nm.Taylor-Mager et al. (2007) and Mager et al. (2018) con-clude that early-type galaxies (E–S0) have CAS parame-ters that appear, within their errors, to be similar at allwavelengths longwards of the Balmer break, but that inthe far-UV, E–S0 galaxies have concentrations and asym-metries that more closely resemble those of spiral and pecu-liar/merging galaxies in the optical. This may be explainedby extended disks containing recent star formation. TheCAS parameters for galaxy types later than S0 show a mildbut significant wavelength dependence, even over the wave-length range 436-823 nm, and a little more significantly sofor the earlier spiral galaxy types (Sa–Sc). These galaxies Table 3.
Final number of objects in each of our morphologicalclasses. Morphology Object Number PercentageDouble 15,931 44.4%Pure Disk 13,212 37%Elliptical 3,890 11%Compact 1,124 3.1%Hard 1,600 4.5% generally become less concentrated and more asymmetricand somewhat more clumpy towards shorter wavelengths.The same is true for mergers when they progress from pre-merger via major-merger to merger-remnant stages.While these trends are mostly small and within the sta-tistical error bars for most galaxy types from 436–823 nm,this is not the case for the Concentration index and Asym-metry of Sa–Sc galaxies. For these galaxies, the Concen-tration index decreases and the Asymmetry increases go-ing from 823 nm to 436 nm (Fig. 17 of Taylor-Mager et al.2007 and Fig. 5 of Mager et al. 2018). Hence, to the extentthat our visual classifications of apparent Sa–Sc galaxies de-pended on their Concentration index and Asymmetry, it ispossible that some of these objects may have been misclas-sified. E-S0 galaxies show no such trend in Concentrationwith wavelength for 436–823 nm, and have Concentrationindices much higher than Sa–Sc galaxies and Asymmetryand Clumpiness parameters generally much lower than Sa–Sc galaxies. Hence, it is not likely that a significant frac-tion of E–S0 galaxies are misclassified as Sa–Sc galaxies at z (cid:46) . z (cid:39) In Figure 4, we show the associated fractions of our finalvisual inspection as solid lines. The global trends are ingood agreement with our initial 4k gold calibration sam-ple. We find that, although our inspection procedure mayhave slightly changed from the 4k gold calibration sampleto the full sample, the outcome classifications are consistentin the three primary classes making up more than 92% ofour sample.Figure 7 shows the probability density function (PDF)of the sSFR (upper panel) and total stellar mass (lowerpanel) for galaxies classified as D, BD and E. We use the
MNRAS , 1–24 (2021)
EVILS: Mass Growth of Morphological Types Figure 8.
Random postage stamps of various morphologies that we generate to perform our visual inspection. Background images areHST ACS/F814W while insets are combined SUBARU gri colour image. Rows represent different morphologies. Galaxies are randomlyselected within different redshift bins shown in columns. Redshifts annotated in the first row are the mean of the associated redshift bins.Cutouts indicate 5 × R90 on each side, where R90 is measured from UltraVISTA Y-band in the DEVILS photometry catalogue (Davieset al. in prep.).
SFR from the
ProSpect
SED fits described in Thorne et al.(2020). These figures indicate that, as one would expect,D galaxies dominate lower stellar mass and higher sSFRregime, opposite to Es which occupy the high stellar massend and low sSFR. BD galaxies are systems located in-between these two classes in terms of both stellar mass andsSFR. These results are as expected and provide some con-fidence that our classifications are sensible.Figure 8 displays a set of random galaxies in each of themorphological classes within different redshift intervals. Inaddition, we present 50 random galaxies of each morphologyin Figures 17-21.Having finalised our morphological classification, wenow investigate the stellar mass functions for different mor-phologies and their evolution from z = 1. For parameterizing the SMF, we assume a fitting func-tion that can describe the galaxy number density, Φ (M).The typical function adopted is that described by Schechter(1976) as: Φ( M )d M = Φ ∗ e − M/M ∗ (cid:18) MM ∗ (cid:19) α d M, (2)where the three key parameters of the function are, α , thepower low slope or the slope of the low-mass end, Φ ∗ , thenormalization, and, M ∗ , the characteristic mass (also knownas the break mass or the knee of the Schechter function).At very low redshifts a number of studies have arguedthat the shape of the SMF is better described by a doubleSchechter function (Baldry et al. 2008; Pozzetti et al. 2010;Baldry et al. 2012), i.e. a combination of two single Schechterfunctions, parameterized by a single break mass ( M ∗ ), andgiven by:Φ( M )d M = e − M/M ∗ (cid:20) Φ ∗ (cid:18) MM ∗ (cid:19) α + Φ ∗ (cid:18) MM ∗ (cid:19) α (cid:21) d MM ∗ , (3)where α < α , indicating that the second term predomi-nantly drives the lower stellar mass range.To fit our SMFs, we use a modified maximum like-lihood (MML) method (i.e., not 1 /V max ) as implemented MNRAS000
SED fits described in Thorne et al.(2020). These figures indicate that, as one would expect,D galaxies dominate lower stellar mass and higher sSFRregime, opposite to Es which occupy the high stellar massend and low sSFR. BD galaxies are systems located in-between these two classes in terms of both stellar mass andsSFR. These results are as expected and provide some con-fidence that our classifications are sensible.Figure 8 displays a set of random galaxies in each of themorphological classes within different redshift intervals. Inaddition, we present 50 random galaxies of each morphologyin Figures 17-21.Having finalised our morphological classification, wenow investigate the stellar mass functions for different mor-phologies and their evolution from z = 1. For parameterizing the SMF, we assume a fitting func-tion that can describe the galaxy number density, Φ (M).The typical function adopted is that described by Schechter(1976) as: Φ( M )d M = Φ ∗ e − M/M ∗ (cid:18) MM ∗ (cid:19) α d M, (2)where the three key parameters of the function are, α , thepower low slope or the slope of the low-mass end, Φ ∗ , thenormalization, and, M ∗ , the characteristic mass (also knownas the break mass or the knee of the Schechter function).At very low redshifts a number of studies have arguedthat the shape of the SMF is better described by a doubleSchechter function (Baldry et al. 2008; Pozzetti et al. 2010;Baldry et al. 2012), i.e. a combination of two single Schechterfunctions, parameterized by a single break mass ( M ∗ ), andgiven by:Φ( M )d M = e − M/M ∗ (cid:20) Φ ∗ (cid:18) MM ∗ (cid:19) α + Φ ∗ (cid:18) MM ∗ (cid:19) α (cid:21) d MM ∗ , (3)where α < α , indicating that the second term predomi-nantly drives the lower stellar mass range.To fit our SMFs, we use a modified maximum like-lihood (MML) method (i.e., not 1 /V max ) as implemented MNRAS000 , 1–24 (2021) Hashemizadeh et al.
Figure 9.
Top panel: Single and double Schechter functions fitted to a low- z sample of D10/ACS (0 . < z < . dftools down to the stellar mass limit of the sample, i.e. 10 . M (cid:12) . Transparent region shows the error range calculated by1000 times sampling of the full posterior probability distribution of the single Schechter fit parameters. Black data points show the binnedgalaxy counts for which the stellar mass range from minimum to maximum (9 . ≤ log(M ∗ / M (cid:12) ) ≤
12) is binned into 25 equal-sized bins.Note that the Schechter functions are not fitted directly to the galaxy counts. Bottom panel: Comparison of our SMF at 0 . < z < . in dftools (Obreschkow et al. 2018). This technique hasmultiple advantages including: it is free of binning, accounts https://github.com/obreschkow/dftools for small statistics and Eddington bias. dftools recoversthe mass function (MF) while simultaneously handling anycomplex selection function, Eddington bias and the cosmiclarge-scale structure (LSS). Eddington bias tends to changethe distribution of galaxies, particularly in the low-mass MNRAS , 1–24 (2021)
EVILS: Mass Growth of Morphological Types regime which is more sensitive to the survey depth and S/N,as well as high-mass regime due to the exponential cut-offwhich is sensitive to the scatters by noise (e.g. Ilbert et al.2013; Caputi et al. 2015). Eddington bias occurs becauseof the observational/photometric errors and often can dom-inate over the shot noise (Obreschkow et al. 2018). Pho-tometric uncertainties, which are introduced in the redshiftestimation, as well as the stellar mass measurements are fun-damentally the source of this bias (Davidzon et al. 2017). Weaccount for Eddington bias in dftools by providing the er-rors on the stellar masses from the ProSpect analysis byThorne et al. (2020).The R language implementation and MML methodmake dftools very fast. In the fitting procedures describedin this work, we use the inbuilt optim function with thedefault optimization algorithm of Nelder & Mead (1965)for maximizing the likelihood function. In order to accountfor the volume corrections, we use the effective unmaskedarea of the D10/ACS region which we calculate to be 1 . dftools and its methodology. In this paper, we fit both single anddouble Schechter functions to the SMF. We examine bothfunctions and will discuss further in Section 3.1. We first validate our total
SMF fitting process at low- z andcompare with the known literature, prior to splitting by red-shift and morphology.To achieve this, we compare the SMF of D10/ACSgalaxies in our lowest- z bin, i.e. 0 < z < .
25, with lit-erature studies. We primarily choose this redshift range tocompare with the SMF Muzzin et al. (2013) and Davidzonet al. (2017) at 0 . < z < . z < .
06 (Baldry et al.2012; Kelvin et al. 2014; Moffett et al. 2016; Wright et al.2017). We fit the SMF within this redshift range using bothsingle and double Schechter functions (Equations 2 and 3,respectively). The upper panel of Figure 9, shows our singleand double Schechter fit to this low- z sample. As annotatedin the figure, we report the best fit single Schechter param-eters of log(M ∗ / M (cid:12) ) = 11 . ± . α = − . ± . ∗ ) = − . ± .
06 and log(M ∗ / M (cid:12) ) = 10 . ± . α = − . ± . α = − . ±
3, log(Φ ∗ ) = − . ± . ∗ ) = − . ± M ∗ /M (cid:12) ) < . Figure 10.
The N( z ) distribution of the D10/ACS sample (solidline) compared with the SHARK semi-analytic model predic-tion (dashed line). SHARK data represent a light-cone covering107.889 square degrees with Y-mag < .
5. Colour bands repre-sent the redshift bins that we consider in this work. Note that thePDFs are smoothed by a kernel with standard deviation of 0.3 soare non-zero beyond the nominal limits.
Despite larger errors on the double Schechter parame-ters, we elect to use this function for our future analysis atall redshifts as it has been shown that a double Schechter canbetter describe the stellar mass distribution even at higher- z (e.g., Wright et al. 2018).In the lower panel of Figure 9, we show that our SMF isin good agreement with Muzzin et al. (2013) and Davidzonet al. (2017) within the quoted errors. Overall, we see a goodagreement with the mass function of galaxies at low- z . Weobserve no significant flattening of the SMF at the interme-diate stellar masses (10 . − . ) as reported in the localUniverse by e.g., Moffett et al. (2016). We are unsurprisedthat we do not see this flattening as the D10/ACS containsonly 3 galaxies in the redshift range comparable to GAMA( z < . < z < .
25 that would explain the slightly higher numberdensity in the intermediate mass range. Our fitted Schechterfunctions however also deviate from the GAMA SMFs in thehigh stellar mass end. We find that this is systematically dueto the Schechter function fitting process. Unlike D10/ACS,as shown in Figure 9, the local literature data extend tolower stellar mass regimes (log(M ∗ / M (cid:12) ) = 8 and 7 . ∗ . In this regime the upturn of the stel-lar mass distribution is remarkably more pronounced. Thisstrong upturn drives the optimization fitting and impactsthe high-mass end. This highlights the difficulty in directlycomparing fitted Schechter values if fitted over different massranges. MNRAS , 1–24 (2021) Hashemizadeh et al.
Figure 11.
Top panel: our measurement of the evolution of the total
SMD (per comoving Mpc ) compared with a compilationof GAMA, COSMOS and 3D-HST by Driver et al. (2018). Theblack curve represents a spline fit to the Driver et al. (2018) data.Middle panel: the large scale structure correction factor appliedto our SMD in order to meet the predictions of the spline fit.Bottom panel: the residual of the SMDs before and after the LSScorrection indicating the correction coefficient we apply to eachredshift bin. All galaxy surveys are to some extent influenced by the cos-mic large scale structure (LSS, Obreschkow et al. 2018).Generally, the LSS produces local over- and under-densitiesof the galaxies at particular redshifts in comparison to themean density of the Universe at that epoch. For example,GAMA regions are ∼
10% underdense compared with SDSS(Driver et al. 2011).We observe this phenomenon in the nonuniformD10/ACS redshift distribution, N( z ). To highlight this, Fig-ure 10 compares the N( z ) distribution of the D10/ACS sam-ple with the prediction of the SHARK semi-analytic model(Lagos et al. 2018; Lagos et al. 2019). SHARK data in thisfigure represent a light-cone covering 107.889 square degreeswith Y-mag < .
5, and because of the much larger simu-lated volume, is less susceptible to LSS. In this figure the red-shift bins we shall use later in our analysis are shown as back-ground colour bands indicating redshift intervals of (0, 0.25,0.45, 0.6, 0.7, 0.8, 0.9 and 1.0). SHARK predicts a nearlyuniform galaxy distribution with no significant over- andunder-density regions, while the empirical D10/ACS sam-ple shows a nonuniform N( z ) with overdensities and under-densities. These density fluctuations can introduce system- atic errors in the construction of the SMF by, for example,overestimating the number density of very low-mass galax-ies which are only detectable at lower redshifts (Obreschkowet al. 2018). In other words, the LSS can artificially changethe shape/normalization of the SMF.Using the distance distribution of galaxies, dftools (Obreschkow et al. 2018) internally accounts for the LSS bymodeling the relative density in the survey volume, g ( r ),i.e. it measures the mean galaxy density of the survey atthe comoving distance r , relative to the mean density of theUniverse. Incorporating this modification into the effectivevolume, the MML formalism works well for a sensitivity-limited sample (see Obreschkow et al. 2018 for details). How-ever, for our volume limited sample, this method is unableto thoroughly model the density fluctuations. Therefore, totake this non-uniformity into account, we perform a manualcorrection as follows.In a comprehensive study of the cosmic star formationhistory, Driver et al. (2018) compiled GAMA, COSMOS and3D-HST data to estimate the total stellar mass density fromhigh redshifts to the local Universe (0 < z < . ρ ∗ . We then introduce a set of correction factors ( β ) to ourempirical measurements of the total SMDs to match the pre-diction of the spline fit (middle panel of Figure 11). We thenapply these correction factors (bottom panel of Figure 11)to our estimations of the morphological SMD values. Thisignores any coupling between morphology and LSS whichwe consider a second order effect. We report the β correc-tion factors in Table 4 and show them in the bottom panelof Figure 11. MNRAS , 1–24 (2021)
EVILS: Mass Growth of Morphological Types F i g u r e . T h e t o t a l a nd m o r ph o l og i c a l S M F s i n e i g h t r e d s h i f t b i n s . T o p r o w h i g h li g h t e db yy e ll o w c o l o u rr e p r e s e n t s t h e G A M A S M F s ( ≤ z ≤ . ) . D a t a p o i n t s a r e ga l a xy c o un t s i n e a c h o f e q u a l - s i ze s t e ll a r m a ss b i n s . W i d t h o f s t e ll a r m a ss b i n s a r e s h o w n a s h o r i z o n t a l b a r s o nd a t a p o i n t s . V e r t i c a l b a r s a r e p o i ss o n e rr o r s . Sh a d e d r e g i o n s a r o und t h e b e s t fi t c u r v e s a r e p e r ce n t c o nfid e n ce r e g i o n s . B l a c k s o li d a ndd a s h e d c u r v e s d e m o n s t r a t e d o ub l e a nd s i n g l e S c h ec h t e r f un c t i o n s o f a ll ga l a x i e s , r e s p ec t i v e l y , w h il e d o tt e d c u r v e s o v e r - p l o tt e d o n h i g h e r - z b i n s a r e t h e G A M A z = S M F s t o h i g h li g h tt h ee v o l u t i o n o f t h e S M F . MNRAS000
EVILS: Mass Growth of Morphological Types F i g u r e . T h e t o t a l a nd m o r ph o l og i c a l S M F s i n e i g h t r e d s h i f t b i n s . T o p r o w h i g h li g h t e db yy e ll o w c o l o u rr e p r e s e n t s t h e G A M A S M F s ( ≤ z ≤ . ) . D a t a p o i n t s a r e ga l a xy c o un t s i n e a c h o f e q u a l - s i ze s t e ll a r m a ss b i n s . W i d t h o f s t e ll a r m a ss b i n s a r e s h o w n a s h o r i z o n t a l b a r s o nd a t a p o i n t s . V e r t i c a l b a r s a r e p o i ss o n e rr o r s . Sh a d e d r e g i o n s a r o und t h e b e s t fi t c u r v e s a r e p e r ce n t c o nfid e n ce r e g i o n s . B l a c k s o li d a ndd a s h e d c u r v e s d e m o n s t r a t e d o ub l e a nd s i n g l e S c h ec h t e r f un c t i o n s o f a ll ga l a x i e s , r e s p ec t i v e l y , w h il e d o tt e d c u r v e s o v e r - p l o tt e d o n h i g h e r - z b i n s a r e t h e G A M A z = S M F s t o h i g h li g h tt h ee v o l u t i o n o f t h e S M F . MNRAS000 , 1–24 (2021) Hashemizadeh et al. z = 1With the LSS correction in place, we now split the sampleinto 7 bins of redshift (0-0.25, 0.25-0.45, 0.45-0.6, 0.6-0.7,0.7-0.8, 0.8-0.9 and 0.9-1.0) so that we have enough numbersof objects in each bin and are not too wide to incorporate asignificant evolution (these redshift ranges are shown by thecolour bands in Figure 10). Each bin contain 1,108, 5,041,4,216, 4,867, 5,277, 7,090, 8,207 galaxies, respectively.Figure 12 shows the SMF in each redshift bin for thefull sample, as well as for different morphological types of: double-component (BD), pure-disk (D), elliptical (E), com-pact (C) and hard (H). These SMF measurements includethe correction for LSS as discussed in Section 3.2. Note thatthe first row of Figure 12 highlighted by yellow shade showsGAMA z = 0 SMFs (Driver et al., in prep.).As shown in Figure 12, we fit the total SMF within all8 redshift bins (including GAMA) by both single and dou-ble Schechter functions (displayed as dashed and solid blackcurves, respectively), while the morphological SMFs are wellfit by single Schechter function at all epochs. The differ-ence between double and single Schechter fits is insignifi-cant, compared to the error on individual points. However,as noted in Section 3.1, for calculating the stellar mass den-sity we will use our double Schechter fits. As we will see later,the effect of this choice on the measurement of our total stel-lar mass density is negligible. In Figure 12, we over-plot theGAMA z = 0 SMFs (Driver et al., in prep.) on higher- z tohighlight the evolution (dotted curves). In the total SMFwe see a growth at the low-mass end and a relative stabil-ity at high-mass end, particularly when comparing high- z with the D10/ACS low- z (0 < z < . z = 1 in-situ star formation and minormergers (low-mass end) play an important role in forming ortransforming mass within galaxies (Robotham et al. 2014).Looking at the SMF of the morphological sub-classes wefind that the BD systems show no significant growth withtime at their high-mass end, while at the low-mass end wesee a noticeable increase in number density. In D galaxies, atboth low- and high-mass ends we find a variation in the num-ber density with the low-mass end increasing and extremesteepening at the lower- z and high-mass end decreasing withtime. Note that we do not rule out the possible effects ofsome degree of incompleteness in this mass regime on theevolution of the low-mass end. For E galaxies, however, wereport a modest growth in their high-mass end (again whencomparing with D10/ACS low- z ) and a significant growth inintermediate- and low-mass regimes from z = 1 to z = 0. Fi-nally, H and C systems become less prominent with decliningredshift as fewer galaxies occupy these classes. The physicalimplications of these trends will be discussed in Section 5.Figure 13 further summarizes the trends in Figure 12,showing the evolution of our best fit single Schechter pa-rameters Φ ∗ , M ∗ and α . We report the best Schechter fitparameters in Table 6. For comparison, we also show a com-pilation of literature values in the right panel. The litera-ture data show only single Schechter parameters of the total SMF. Comparing our total
SMF with other studies, includ-ing Muzzin et al. (2013), we find a good agreement withinthe quoted errors.Note that the three parameters are, of course, highlycorrelated and the mass limits to which they are fitted vary. The Φ ∗ of BD systems has the largest value and almostmimics the trend of the total Φ ∗ which is largely consistentwith no evolution. This is expected as BD galaxies domi-nate the sample at almost all redshifts. We report a slightdecrease in the Φ ∗ value of D galaxies, while almost no evo-lution in E systems. Note that as mentioned earlier, ourlowest redshift bin (0 . < z < .
25) contains only 1 , ∗ values) owing to thefact that the abundance of mergers, clumpy and disruptedsystems dramatically increases at higher redshifts (despitelower resolution). We note that due to our LSS corrections(Section 3.2) the Φ ∗ reported here is not the directly mea-sured parameter but modified according to our LSS correc-tion coefficients (reported in Table 4). This normalizationsmooths out the evolutionary trends, otherwise fluctuatingdue to the large scale structures, but does not impact theirglobal trends.The characteristic mass, M ∗ , of the BD systemspresents a stable trend since z = 1. We interpret thisbehaviour to be a result of the lack of significant evolutionof the massive/intermediate-mass regime of the SMF of thismorphological class. D galaxies also show slight evidence ofevolution with some fluctuations in M ∗ value. E galaxies,likewise, evolve only a small amount. Overall, we observe abehaviour consistent with no evolution in M ∗ for all mor-phological types. Note that large variation of the M ∗ in Hand C types, at low- z in particular, is not physical. This ismassively driven by the lack of H and C galaxies at low- z resulting in an unconstrained turnover and as can be seenin Figure 13 large uncertainties.The low-mass end slope, α , of different morphologicalclasses also shows some evolution. Similar to D systems, BDgalaxies show a marked steepening increase in their slopeat later times, indicating that the SMF steepens at lowerredshifts. The steepest mass function at almost all times isfor the D galaxies. E galaxies occupy the lowest steepeningbut constantly growing from α = 0 .
34 to − .
75 at z ∼ . z = 1In this section, we investigate the evolution of the StellarMass Density (SMD) as a function of morphological types.To determine the SMD, we integrate under the best fitSchechter functions over all stellar masses from 10 . to ∞ .This integral can be expressed as a gamma function: ρ ∗ = (cid:90) ∞ M =10 . M (cid:48) Φ( M (cid:48) ) dM (cid:48) = Φ ∗ M ∗ Γ( α + 2 , . /M ∗ ) , (4)where Φ ∗ , M ∗ and α are the best regressed Schechter pa-rameters. Figure 14 shows the evolution of the distributionof the SMDs, term M (cid:48) Φ( M (cid:48) ), for different morphologies. Wealso illustrate the fitting errors by sampling 1000 times thefull posterior probability distribution of the fit parameters.These are shown in Figure 14 as transparent shaded regionsaround the best fit curves. The standard deviations of the MNRAS , 1–24 (2021)
EVILS: Mass Growth of Morphological Types Figure 13.
Left: Evolution of the single Schechter best fit parameters, Φ ∗ , M ∗ and α . Error bars are the standard deviations on eachof the most likely parameters. Black line is the total SMF while colour coded data represent the best fit parameters of the Schechterfunction of different morphologies. Note that for simplicity we remove compacts as their trends are very noisy and washes out the trendsof other morphological types. Right: Zoomed-in plot showing the evolution of our total
Schechter function parameters compared with acompilation of other studies. Highlighted region shows the epoch covered by the GAMA data. integrated SMD calculated from each of these functions arereported as the fit error on ρ ∗ in Table 4. As can be seen inFigure 14, the distribution of the stellar mass density of allindividual morphological types in almost all redshift bins isbounded, implying that integrating under these curves willcapture the majority of the stellar mass for each class.Figure 15 then shows the evolution of the integratedSMD, ρ ∗ , in the Universe between z = 0 −
1. This includesthe LSS correction by forcing our total
SMD values to matchthe smooth spline fit to the Driver et al. (2018) data, asdescribed in Section 3.2. The uncorrected evolutionary pathof ρ ∗ is shown in the upper panel of Figure 11.We report the empirical LSS corrected ρ ∗ values in Ta-ble 4. This table also provides the LSS correction factor, β ,so one might obtain the original values by ρ Orig ∗ = ρ corr ∗ /β .For completeness, we show the evolution of the SMDsbefore we apply the LSS corrections in Figure 22 of Ap-pendix 11. In this Figure, the trends are not as smooth asone would expect without the LSS correction being applied,given the evident structure in the N ( z ) distribution fromFigure 10. However, even without the LSS corrections, themain trends are still present, albeit not as strong. This high- lights that the LSS correction is an important aspect andalso the need for much wider deep coverage than that cur-rently provided by HST. This will become possible in thecoming Euclid and Roman era.Before we analyse the evolution of the SMDs in Figure15, below we investigate the errors that are involved in thiscalculation. MNRAS000
SMD values to matchthe smooth spline fit to the Driver et al. (2018) data, asdescribed in Section 3.2. The uncorrected evolutionary pathof ρ ∗ is shown in the upper panel of Figure 11.We report the empirical LSS corrected ρ ∗ values in Ta-ble 4. This table also provides the LSS correction factor, β ,so one might obtain the original values by ρ Orig ∗ = ρ corr ∗ /β .For completeness, we show the evolution of the SMDsbefore we apply the LSS corrections in Figure 22 of Ap-pendix 11. In this Figure, the trends are not as smooth asone would expect without the LSS correction being applied,given the evident structure in the N ( z ) distribution fromFigure 10. However, even without the LSS corrections, themain trends are still present, albeit not as strong. This high- lights that the LSS correction is an important aspect andalso the need for much wider deep coverage than that cur-rently provided by HST. This will become possible in thecoming Euclid and Roman era.Before we analyse the evolution of the SMDs in Figure15, below we investigate the errors that are involved in thiscalculation. MNRAS000 , 1–24 (2021) Hashemizadeh et al. F i g u r e . T h e d i s t r i bu t i o n o f t o t a l a nd m o r ph o l og i c a l s t e ll a r m a ss d e n s i t y i nd i ff e r e n t r e d s h i f t s . R e d s h i f t b i n s a r e t h e s a m e a s F i g u r e . P o i n t s a nd li n e s i nd i c a t e M (cid:48) Φ ( M (cid:48) ) i n E q u a t i o n , w h e r e Φ ( M (cid:48) ) i s o u r S c h ec h t e r f un c t i o nfi t . T h e s h a d e d t r a n s p a r e n t r e g i o n s r e p r e s e n tt h ee rr o rr a n g ec a l c u l a t e db y t i m e ss a m p li n go f t h e f u ll p o s t e r i o r p r o b a b ili t y d i s t r i bu t i o n o f t h e fi t p a r a m e t e r s . T h e d i s t r i bu t i o n o f t h e s t e ll a r m a ss d e n s i t y o f a lli nd i v i du a l m o r ph o l og i c a l t y p e s i n a l m o s t a ll r e d s h i f t b i n s i s b o und e d . T o p r o w h i g h li g h t e db yy e ll o w c o l o u rr e p r e s e n t s t h e G A M A d a t a ( ≤ z ≤ . ) . MNRAS , 1–24 (2021)
EVILS: Mass Growth of Morphological Types The error budget on our analysis of the SMD includes cosmicvariance (CV), fit error , classification error and Poissonerror . Cosmic variance and classification are the dominantsources of error.We make use of Driver & Robotham (2010) equation 3to calculate the cosmic variance in the volume encompassedwithin each redshift bin. This calculation is implemented inthe R package: celestial . Note that for our low- z GAMAdata, instead of using Driver & Robotham (2010) equationthat estimates a ∼ .
8% cosmic variance, we empiricallycalculate the CV using the variation of the SMD between 3different GAMA regions of G09, G12, and G15 with a totaleffective area of 169.3 square degrees (G09:54.93; G12: 57.44;G15: 56.93, Bellstedt et al. 2020) and find CV to be ∼ fit error by 1000 times sampling ofthe full posterior probability distribution of the Schechterparameters (shown as shaded error regions in Figure 14) andcalculating the associated ρ ∗ in each iteration. We calculatethe classification error by measuring the stellar mass densityassociated with each of our 3 independent morphologicalclassifications. The range of variation of the SMD betweenclassifiers gives the error of our classification. The Poissonerror is calculated by using the number of objects in eachmorphology per redshift bin.The combination of all the above error sources willprovide us with the total error that is reported in Table 4.As can be seen in Figure 15, the extrapolation of ourD10/ACS ρ ∗ to z = 0 agrees well with the our local GAMAestimations. Note slight difference in D, but consistent inerrors.The total change in the stellar mass is consistent withobserved SFR evolution (e.g., Madau & Dickinson 2014;Driver et al. 2018) as we will discuss more in Section 5.Analysing the evolution of the SMD (Figure 15), we find thatin total (black symbols), 68% of the current stellar mass ingalaxies was in place ∼ z ∼ . ρ ∗ ( ρ ∗ z /ρ ∗ z =0 ),where ρ ∗ z is the SMD at redshift z while ρ ∗ z =0 represents thefinal SMD at z = 0.According to our visual inspections C types are closerto Es than other subcategories. We, therefore, combine Ctypes with E galaxies that shows a smooth growth withtime of a factor of ∼ . z = 0 .
25 and flattens outsince then (0 . < z < . ∼ f s =Ω ∗ / Ω b ) locked in each of our morphological types. We adoptΩ b = 0 . The online version of this cosmology calculator is available at:http://cosmocalc.icrar.org/ verse at the median redshift of GAMA, i.e. z = 0 .
06 to be ρ c = 1 . × M (cid:12) Mpc − in a 737 cosmology.As shown in the bottom panel of Figure 16, at ourD10/ACS lowest redshift bin z ∼ .
18 we find the frac-tion of baryons in stars f s ∼ . ± . z ∼ .
06 we find this fraction tobe f s ∼ . ± . f s , shows that it has increased from (2 . ± . z ∼ . ± . z ∼ ∼ . f s toeach of our morphological subcategories highlighting thatas expected BD systems contribute the most to the stellarbaryon fraction increasing from f s = 0 . ± .
006 to f s =0 . ± .
005 at 0 < z <
1. E+C systems take less percentageof the total f s but increase their contribution from 0 . ± .
001 to 0 . ± .
002 while D galaxies decrease from f s =0 . ± .
002 to f s = 0 . ± .
005 between 0 < z <
1. Wereport our full f s values for all morphological types at allredshifts in Table. 5.In summary, over the last 8 Gyr, double componentgalaxies clearly dominate the overall stellar mass density ofthe Universe at all epochs. The second dominant system is Egalaxies. However the extrapolation of the trends to higherredshifts in Figures 15 and 16 indicates that D systems arelikely to dominate over Es in the very high- z regime ( z > Making use of our morphological classifications, we explorethe evolution of the stellar mass function at 0 ≤ z ≤
1, toassess the physical processes that are likely affecting the in-dividual morphological SMF. In particular, major mergersare thought to primarily occur between comparable masscompanions (1:3) resulting in the fast growth of the SMF(Robotham et al. 2014). Conversely, secular activities, mi-nor mergers and tidal interactions will primarily alter thenumber density at the low-stellar mass end of the SMF(Robotham et al. 2014). Investigating the total
SMF (Fig-ure 12) shows that unlike high-mass end the low-mass endgrows significantly. This suggests that since z ∼
1, thegalaxy population goes through mainly in-situ/secular pro-cesses at the low-mass end.An important caveat, in what follows, is that althoughwe have undertaken multiple tests of our visual classifi-cations, possible uncertainties due to human classificationerror or inconsistencies will be present. Nevertheless someclear trends, which we believe are resilient to the classifica-tions uncertainties are evident. Furthermore, we do not ruleout the effects of dust in distinguishing bulges, particularlyat high- z leading to overestimating the number of pure-disk systems at high redshifts. Although, high level of agreementin our visual classifications (see Figure 6) gives us some con-fidence that our evolutionary trends are unlikely dominated MNRAS000
1, thegalaxy population goes through mainly in-situ/secular pro-cesses at the low-mass end.An important caveat, in what follows, is that althoughwe have undertaken multiple tests of our visual classifi-cations, possible uncertainties due to human classificationerror or inconsistencies will be present. Nevertheless someclear trends, which we believe are resilient to the classifica-tions uncertainties are evident. Furthermore, we do not ruleout the effects of dust in distinguishing bulges, particularlyat high- z leading to overestimating the number of pure-disk systems at high redshifts. Although, high level of agreementin our visual classifications (see Figure 6) gives us some con-fidence that our evolutionary trends are unlikely dominated MNRAS000 , 1–24 (2021) Hashemizadeh et al.
Figure 15.
The evolution of the total and morphological stellar mass density, ρ ∗ , in the last 8 Gyr of the cosmic age. ρ ∗ expresses ourmeasurements of the analytical integration under the best Schechter function fits in 8 redshift bins, i.e. Equation 4. Highlighted regionshows the epoch covered by the GAMA data (0 ≤ z ≤ . by incorrect classifications, it is possible that all classifiersagree on the incorrect classification.Firstly, double component galaxies display a modest de-crease with redshift in the high-stellar mass end of their SMF(Figure 12), whilst their low-stellar mass end steepens sig-nificantly. This could be interpreted as most of the stellarmass in double component systems evolving via lower massweighted star formation or secular activity, rather than ma-jor merging, i.e., BD systems are not merging together toform higher-mass BD systems.Secondly, pure-disk systems show a strong increase atthe low-mass end and a decrease at their high-mass end.The low-mass end evolution suggests in-situ star formationof the disk and/or the formation of new disks. However, thedecrease at the high-mass end is to some extent unphysical, unless these systems are undergoing a transformation fromthe D class to another class. The most likely prospect is thesecular formation of a central bulge component, resulting ina morphological transformation into the BD class. Hence, astime progresses and the second component forms, galaxiesexit the D class leading to a mass-deficit at the high-massend. The new component forming through such an in-situprocess is most likely a pseudo-bulge (pB), resulting in massloss from the high-mass D SMF and a corresponding massgain at comparable mass in the BD class.Finally, elliptical galaxies, generally thought to be inert,show little growth in their high-mass end but a significantgrowth of low- and intermediate mass end presumably dueto mergers. MNRAS , 1–24 (2021)
EVILS: Mass Growth of Morphological Types Table 4.
The total and morphological stellar mass density in different redshift ranges displayed in Figure 15 (left panel). ρ ∗ values arecalculated from the integration under the best Schechter function fits. Note that the ρ ∗ are presented after applying the LSS correction.Columns represent: z − bin: redshift bins, z : mean redshift, LBT: lookback time, T (S-Schechter): total with single Schechter , T (D-Schechter): total with double Schechter , BD: bulge+disk , pD: pure-disk , E: elliptical , C: compact , H: hard , β : the large scale structurecorrection factor. Errors incorporate all error sources including, cosmic variance, fit error, Poisson error and classification error (seeSection 4.1 for details). log10( ρ ∗ /M (cid:12) ) z − bin z LBT (Gyr) T (S-Schechter) T (D-Schechter) BD pD E C H β . ≤ z < .
08 0 .
06 0 .
82 8 . ± .
08 8 . ± .
11 8 . ± .
08 6 . ± .
26 7 . ± . − . ± .
25 1 . . ≤ z < .
25 0 .
18 2 .
27 8 . ± .
14 8 . ± .
14 8 . ± .
14 6 . ± .
53 7 . ± .
20 5 . ± .
83 6 . ± .
38 0 . . ≤ z < .
45 0 .
36 3 .
95 8 . ± .
10 8 . ± .
10 8 . ± .
10 7 . ± .
17 7 . ± .
12 5 . ± .
23 6 . ± .
43 0 . . ≤ z < .
60 0 .
53 5 .
23 8 . ± .
10 8 . ± .
16 8 . ± .
10 7 . ± .
12 7 . ± .
13 5 . ± .
01 6 . ± .
21 1 . . ≤ z < .
70 0 .
65 6 .
05 8 . ± .
12 8 . ± .
21 7 . ± .
12 7 . ± .
13 7 . ± .
13 5 . ± .
60 6 . ± .
25 0 . . ≤ z < .
80 0 .
74 6 .
55 8 . ± .
11 8 . ± .
12 7 . ± .
12 7 . ± .
12 7 . ± .
13 6 . ± .
95 6 . ± .
17 0 . . ≤ z < .
90 0 .
85 7 .
07 8 . ± .
11 8 . ± .
13 7 . ± .
11 7 . ± .
12 7 . ± .
12 6 . ± .
55 6 . ± .
15 0 . . ≤ z ≤ .
00 0 .
95 7 .
51 8 . ± .
11 8 . ± .
13 7 . ± .
11 7 . ± .
12 7 . ± .
12 6 . ± .
43 6 . ± .
14 0 . The evolution of the global and morphological SMDs in-dicates that BD galaxies dominate ( ∼
60% on average) thestellar mass density of the Universe since at least at z < rate of mass growth in the BDsystems also decreases with time. This is reflected in thedecrease of their SMD slope although it is still steeper thanthe total
SMD evolution and consistent with the generaldecline in the cosmic star-formation history. On the otherhand, the E galaxies experience an initial growth until z ∼ . z = 1 − We have presented a visual morphological classification ofa sample of galaxies in the DEVILS/COSMOS survey withHST imaging, from the D10/ACS sample. The quality ofthe imaging data (HST/ACS) provides arguably the bestcurrent insight into galaxy structures and therefore the bestpathway with which to explore morphological evolution.We summarize our results as below: • By visually inspecting galaxies out to z ∼ .
5, we findthat morphological classification becomes far more challeng-ing at z > . > • The SMF of the D10/ACS sample in our lowest redshiftbin ( z < .
25) is consistent with previous measurementsfrom the local Universe. • The evolution of the global SMF shows enhancement inboth low- and slightly in high-mass ends. We interpret this as suggesting that at least two evolutionary pathways are inplay, and that both are significantly impacting the SMF. • Despite a slight decrease in their high-mass end, BD sys-tems demonstrate a non-negligible growth in the low-massend of their SMF. In the D type, we witness a significantvariation in both high- and low-mass ends of the SMF. Weinterpret the high-mass end decrease in D systems, which isat first sight unphysical, as an indication of significant sec-ular mass transfer through the formation of pseudo-bulgesand hence an apparent mass loss as galaxies transit to theBD category. • E galaxies experience a modest growth in their high-massend as well as an enhancement in their low/intermediate-mass end which we interpret as a consequence of majormergers resulting in the relentless stellar mass growth ofthis class. • Despite the above shuffling of mass we find that the bestregressed Schechter function parameters in the total
SMFare observed to be relatively stable from z = 1. This is con-sistent with previous studies (Muzzin et al. 2013; Tomczaket al. 2014; Wright et al. 2018). Conversely, the componentSMFs show significant evolution. This implication is thatwhile stellar-mass growth is slowing, mass-transformationprocesses via merging and in-situ evolution are shufflingmass between types behind the scenes. • We measured the integrated total stellar mass densityand its evolution since z = 1 and find that approximately32% of the current stellar mass in the galaxy population wasformed during the last 8 Gyr. • As shown in Figure 15, we find that the BD populationdominates the SMD of the Universe within 0 ≤ z ≤ ∼ ∼
85% of it’s original value until z ∼ .
2. Onthe contrary, the E population experiences constant growthof a factor of ∼ . z = 1. We observe that the ex-trapolation of the trends of all of our morphological SMDestimations meets GAMA measurements at z = 0 (see Fig-ure 15) except for the pure-disk systems which is likely dueto unbound distribution of their SMD (see Figure 14).One clear outcome of our analysis is that the late Uni- MNRAS , 1–24 (2021) Hashemizadeh et al.
Table 5.
Total and morphological stellar baryon fraction ( f s ) in different times. See the text for details. RedshiftMorphology Type 0 . ≤ z < .
08 0 . ≤ z < .
25 0 . ≤ z < .
45 0 . ≤ z < .
60 0 . ≤ z < .
70 0 . ≤ z < .
80 0 . ≤ z < .
90 0 . ≤ z < . . ± .
006 0 . ± .
009 0 . ± .
006 0 . ± .
006 0 . ± .
006 0 . ± .
006 0 . ± .
005 0 . ± . . ± .
004 0 . ± .
005 0 . ± .
004 0 . ± .
004 0 . ± .
004 0 . ± .
003 0 . ± .
003 0 . ± . . ± .
001 0 . ± .
001 0 . ± .
001 0 . ± .
001 0 . ± .
001 0 . ± .
001 0 . ± .
001 0 . ± . . ± .
002 0 . ± .
004 0 . ± .
002 0 . ± .
002 0 . ± .
002 0 . ± .
002 0 . ± .
001 0 . ± . − . ± .
000 0 . ± .
000 0 . ± .
000 0 . ± .
000 0 . ± .
000 0 . ± .
000 0 . ± . . ± .
002 0 . ± .
004 0 . ± .
002 0 . ± .
002 0 . ± .
002 0 . ± .
002 0 . ± .
001 0 . ± . . ± .
000 0 . ± .
000 0 . ± .
000 0 . ± .
000 0 . ± .
000 0 . ± .
000 0 . ± .
000 0 . ± . Figure 16.
Top panel: the variation of the stellar mass densityshowing the fraction of final stellar mass density assembled orlost by each redshift, i.e. ρ ∗ z /ρ ∗ z =0 . ρ ∗ z =0 is taken from GAMAestimations of the local Universe. Bottom panel: the evolution ofthe baryon fraction in form of stars ( f s ). For simplicity, in thisfigure we only show the main morphological types and removethe hard and compact sub-classes. Highlighted region shows theepoch covered by the GAMA data. verse ( z <
1) appears to be a time of profound spheroid andbulge growth/emergence. To move forward and explore thisfurther we conclude that to move forward it is key to decom-pose the double component morphological type, which sig-nificantly dominates the stellar mass density, into disks, clas-sical bulges, and pseudo-bulges. To do this, requires robuststructural decomposition which we will describe in Hashem-izadeh et al. (in prep.) using our morphological classifica-tions to guide the decomposition process.
DEVILS is an Australian project based around a spec-troscopic campaign using the Anglo-Australian Telescope.The DEVILS input catalogue is generated from data takenas part of the ESO VISTA-VIDEO (Jarvis et al. 2013)and UltraVISTA (McCracken et al. 2012) surveys. DEV-ILS is part funded via Discovery Programs by the Aus-tralian Research Council and the participating institutions.The DEVILS website is https://devilsurvey.org. The DEV-ILS data is hosted and provided by AAO Data Central(https://datacentral.org.au). This work was supported byresources provided by The Pawsey Supercomputing Cen-tre with funding from the Australian Government and theGovernment of Western Australia. We also gratefully ac-knowledge the contribution of the entire COSMOS col-laboration consisting of more than 200 scientists. TheHST COSMOS Treasury program was supported throughNASA grant HST-GO-09822. SB and SPD acknowledge sup-port by the Australian Research Council’s funding schemeDP180103740. MS has been supported by the EuropeanUnion’s Horizon 2020 research and innovation programmeunder the Maria Sk(cid:32)lodowska-Curie (grant agreement No754510), the National Science Centre of Poland (grantUMO-2016/23/N/ST9/02963) and by the Spanish Min-istry of Science and Innovation through Juan de la Cierva-formacion program (reference FJC2018-038792-I). ASGRand LJMD acknowledge support from the Australian Re-search Council’s Future Fellowship scheme (FT200100375and FT200100055, respectively).This work was made possible by the free and openR software (R Core Team 2020). A number of figures inthis paper were generated using the R magicaxis pack-age (Robotham 2016b). This work also makes use ofthe celestial package (Robotham 2016a) and dftools (Obreschkow et al. 2018).
In this work, we draw upon two datasets; the establishedHST imaging of the COSMOS region (Scoville et al. 2007,Koekemoer et al. 2007), and multiple data products pro-duced as part of the DEVILS survey (Davies et al. 2018),consisting of a spectroscopic campaign currently beingconducted on the Anglo-Australian Telescope, photometry(Davies et al. in prep.), and deep redshift catalogues, andstellar mass measurements from Thorne et al. (2020). Thesedatasets are briefly described below.
MNRAS , 1–24 (2021)
EVILS: Mass Growth of Morphological Types REFERENCES
Abraham R. G., van den Bergh S., 2001, Science, 293, 1273Abraham R. G., van den Bergh S., Glazebrook K., Ellis R. S.,Santiago B. X., Surma P., Griffiths R. E., 1996a, ApJS, 107,1Abraham R. G., Tanvir N. R., Santiago B. X., Ellis R. S., Glaze-brook K., van den Bergh S., 1996b, MNRAS, 279, L47Baldry I. K., Glazebrook K., Brinkmann J., Ivezi´c ˇZ., LuptonR. H., Nichol R. C., Szalay A. S., 2004, ApJ, 600, 681Baldry I. K., Balogh M. L., Bower R. G., Glazebrook K., NicholR. C., Bamford S. P., Budavari T., 2006, MNRAS, 373, 469Baldry I. K., Glazebrook K., Driver S. P., 2008, MNRAS, 388,945Baldry I. K., et al., 2012, MNRAS, 421, 621Bell E. F., et al., 2004, ApJ, 608, 752Bellstedt S., et al., 2020, MNRAS, 496, 3235Bernardi M., Shankar F., Hyde J. B., Mei S., Marulli F., ShethR. K., 2010, MNRAS, 404, 2087Brammer G. B., et al., 2009, ApJ, 706, L173Brammer G. B., et al., 2011, ApJ, 739, 24Brinchmann J., Charlot S., White S. D. M., Tremonti C., Kauff-mann G., Heckman T., Brinkmann J., 2004, MNRAS, 351,1151Bruzual G., Charlot S., 2003, MNRAS, 344, 1000Bundy K., Ellis R. S., Conselice C. J., 2005, ApJ, 625, 621Caputi K. I., et al., 2015, ApJ, 810, 73Cassata P., et al., 2008, A&A, 483, L39Chabrier G., 2003, PASP, 115, 763Charlot S., Fall S. M., 2000, ApJ, 539, 718Cirasuolo M., et al., 2007, MNRAS, 380, 585Conselice C. J., 2004, The Galaxy Structure-Redshift Relation-ship. p. 489, doi:10.1007/978-1-4020-2862-5˙44Conselice C. J., Blackburne J. A., Papovich C., 2005, ApJ, 620,564Dale D. A., Helou G., Magdis G. E., Armus L., D´ıaz-Santos T.,Shi Y., 2014, ApJ, 784, 83Damjanov I., Zahid H. J., Geller M. J., Fabricant D. G., HwangH. S., 2018, ApJS, 234, 21Davidzon I., et al., 2017, A&A, 605, A70Davies L. J. M., et al., 2015, MNRAS, 447, 1014Davies L. J. M., et al., 2018, MNRAS, 480, 768De Lucia G., Blaizot J., 2007, MNRAS, 375, 2Driver S. P., Robotham A. S. G., 2010, MNRAS, 407, 2131Driver S. P., et al., 2006, MNRAS, 368, 414Driver S. P., et al., 2011, MNRAS, 413, 971Driver S. P., Robotham A. S. G., Bland-Hawthorn J., Brown M.,Hopkins A., Liske J., Phillipps S., Wilkins S., 2013, MNRAS,430, 2622Driver S. P., et al., 2018, MNRAS, 475, 2891Fontana A., et al., 2004, A&A, 424, 23Fukugita M., et al., 2007, AJ, 134, 579Hibbard J. E., Vacca W. D., 1997, AJ, 114, 1741Hogg D. W., et al., 2002, AJ, 124, 646Holwerda B. W., Dalcanton J. J., Radburn-Smith D., de JongR. S., Guhathakurta P., Koekemoer A., Allen R. J., B¨okerT., 2012, ApJ, 753, 25Huertas-Company M., et al., 2009, A&A, 497, 743Huertas-Company M., et al., 2015, ApJS, 221, 8Ilbert O., et al., 2010, ApJ, 709, 644Ilbert O., et al., 2013, A&A, 556, A55Jarvis M. J., et al., 2013, MNRAS, 428, 1281Kauffmann G., et al., 2003, MNRAS, 341, 33Kawinwanichakij L., et al., 2020, ApJ, 892, 7Kelvin L. S., et al., 2014, MNRAS, 444, 1647Knapen J. H., Beckman J. E., 1996, MNRAS, 283, 251Koekemoer A., Fruchter A., Hack W., 2003, Space Telescope Eu-ropean Coordinating Facility Newsletter, 33, 10 Koekemoer A. M., et al., 2007, ApJS, 172, 196Krywult J., et al., 2017, A&A, 598, A120Lagos C. d. P., Tobar R. J., Robotham A. S. G., ObreschkowD., Mitchell P. D., Power C., Elahi P. J., 2018, MNRAS, 481,3573Lagos C. d. P., et al., 2019, MNRAS, 489, 4196Laigle C., et al., 2016, ApJS, 224, 24Lange R., et al., 2015, MNRAS, 447, 2603Leja J., van Dokkum P. G., Franx M., Whitaker K. E., 2015, ApJ,798, 115Lilly S. J., et al., 2007, ApJS, 172, 70Lintott C., et al., 2011, MNRAS, 410, 166Madau P., Dickinson M., 2014, ARA&A, 52, 415Mager V. A., Conselice C. J., Seibert M., Gusbar C., KatonaA. P., Villari J. M., Madore B. F., Windhorst R. A., 2018,ApJ, 864, 123McCracken H. J., et al., 2012, A&A, 544, A156Moffett A. J., et al., 2016, MNRAS, 457, 1308Mortlock A., et al., 2015, MNRAS, 447, 2Muzzin A., et al., 2013, ApJ, 777, 18Nair P. B., Abraham R. G., 2010, VizieR Online Data Catalog,p. J/ApJS/186/427Nelder J. A., Mead R., 1965, The computer journal, 7, 308Obreschkow D., Murray S. G., Robotham A. S. G., WestmeierT., 2018, MNRAS, 474, 5500Pannella M., Hopp U., Saglia R. P., Bender R., Drory N., SalvatoM., Gabasch A., Feulner G., 2006, ApJ, 639, L1Papovich C., Giavalisco M., Dickinson M., Conselice C. J., Fer-guson H. C., 2003, ApJ, 598, 827Peng Y.-j., et al., 2010, ApJ, 721, 193Planck Collaboration et al., 2020, A&A, 641, A6Pozzetti L., et al., 2010, A&A, 523, A13R Core Team 2020, R: A Language and Environment for Sta-tistical Computing. R Foundation for Statistical Computing,Vienna, Austria,
Rawat A., Wadadekar Y., De Mello D., 2009, ApJ, 695, 1315Renzini A., Peng Y.-j., 2015, ApJ, 801, L29Robotham A. S. G., 2016a, Celestial: Common astronomical con-version routines and functions (ascl:1602.011)Robotham A. S. G., 2016b, magicaxis: Pretty scientific plottingwith minor-tick and log minor-tick support (ascl:1604.004)Robotham A. S. G., et al., 2014, MNRAS, 444, 3986Robotham A. S. G., Davies L. J. M., Driver S. P., Koushan S.,Taranu D. S., Casura S., Liske J., 2018, MNRAS, 476, 3137Robotham A. S. G., Bellstedt S., Lagos C. d. P., Thorne J. E.,Davies L. J., Driver S. P., Bravo M., 2020, MNRAS, 495, 905Sandage A., 2005, ARA&A, 43, 581Scarlata C., et al., 2007a, ApJS, 172, 406Scarlata C., et al., 2007b, ApJS, 172, 494Schechter P., 1976, ApJ, 203, 297Scodeggio M., et al., 2018, A&A, 609, A84Scoville N., et al., 2007, ApJS, 172, 150Siudek M., et al., 2018, A&A, 617, A70Strateva I., et al., 2001, AJ, 122, 1861Taniguchi Y., et al., 2007, ApJS, 172, 9Taylor-Mager V. A., Conselice C. J., Windhorst R. A., JansenR. A., 2007, ApJ, 659, 162Taylor V. A., Jansen R. A., Windhorst R. A., Odewahn S. C.,Hibbard J. E., 2005, ApJ, 630, 784Taylor E. N., et al., 2009, ApJ, 694, 1171Taylor E. N., et al., 2015, MNRAS, 446, 2144Thorne J. E., et al., 2020, arXiv e-prints, p. arXiv:2011.13605Tomczak A. R., et al., 2014, ApJ, 783, 85Trayford J. W., Theuns T., Bower R. G., Crain R. A., Lagos C.d. P., Schaller M., Schaye J., 2016, MNRAS, 460, 3925Vergani D., et al., 2008, A&A, 487, 89Weigel A. K., Schawinski K., Bruderer C., 2016, MNRAS, 459,2150MNRAS , 1–24 (2021) Hashemizadeh et al.
Whitaker K. E., et al., 2014, ApJ, 795, 104Whitaker K. E., et al., 2015, ApJ, 811, L12Willett K. W., et al., 2017, MNRAS, 464, 4176Williams R. J., Quadri R. F., Franx M., van Dokkum P., Labb´eI., 2009, ApJ, 691, 1879Windhorst R. A., et al., 2002, ApJS, 143, 113Wright A. H., et al., 2017, MNRAS, 470, 283Wright A. H., Driver S. P., Robotham A. S. G., 2018, MNRAS,480, 3491van den Bergh S., Abraham R. G., Ellis R. S., Tanvir N. R.,Santiago B. X., Glazebrook K. G., 1996, AJ, 112, 359van der Wel A., et al., 2012, ApJS, 203, 24
Here we show 49 random galaxies in each of our morpholog-ical categories (BD,D,E,H,C). We show our stamps in bothHST/F814W filter and Subaru gri .
10 APPENDIX B: BEST SCHECHTER FITPARAMETERS.
This table shows our best Schechter fit parameters of thetotal and morphological SMFs, shown in Figure 12.
11 THE NON-LSS-CORRECTED EVOLUTIONOF THE SMD
Figure 22 shows the evolution of the integrated stellar massdensity, ρ ∗ before we apply our large scale structure correc-tions (reported in Table 4 and shown in Figure 11). This is tofurther confirm that the corrections do not derive the overalltrends that we find in Figure 15 as explained in Section 4. MNRAS , 1–24 (2021)
EVILS: Mass Growth of Morphological Types Figure 17.
Random sample of double component systems. Left set of panels: ACS/F814W image. Right set of panels: SUBARU gri combined image.
Figure 18.
Random sample of elliptical systems. Left set of panels: ACS/F814W image. Right set of panels: SUBARU gri combinedimage.MNRAS000
Random sample of elliptical systems. Left set of panels: ACS/F814W image. Right set of panels: SUBARU gri combinedimage.MNRAS000 , 1–24 (2021) Hashemizadeh et al.
Figure 19.
Random sample of pure disk systems. Left set of panels: ACS/F814W image. Right set of panels: SUBARU gri combinedimage.
Figure 20.
Random sample of complex systems ( hard ). Left set of panels: ACS/F814W image. Right set of panels: SUBARU gri combined image. MNRAS , 1–24 (2021)
EVILS: Mass Growth of Morphological Types Figure 21.
Random sample of low angular sized systems ( compact ). Left set of panels: ACS/F814W image. Right set of panels: SUBARU gri combined image.MNRAS000
Random sample of low angular sized systems ( compact ). Left set of panels: ACS/F814W image. Right set of panels: SUBARU gri combined image.MNRAS000 , 1–24 (2021) Hashemizadeh et al.
Table 6.
Best Schechter fit parameters of total and different morphological types in different redshift bins. z -bins 0 . ≤ z < .
08 0 . ≤ z < .
25 0 . ≤ z < .
45 0 . ≤ z < .
60 0 . ≤ z < .
70 0 . ≤ z < .
80 0 . ≤ z < .
90 0 . ≤ z ≤ . Total (Double Schechter) log Φ ∗ − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . M ∗ . ± .
04 10 . ± .
06 10 . ± .
03 10 . ± .
06 10 . ± .
05 10 . ± .
04 10 . ± .
04 10 . ± . α − . ± . − . ± . − . ± . − . ± . − . ± .
33 0 . ± . − . ± . − . ± . Φ ∗ − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . α − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . Total (Single Schechter) log Φ ∗ − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . M ∗ . ± .
03 11 . ± .
05 11 . ± .
03 10 . ± .
03 10 . ± .
02 11 . ± .
02 11 . ± .
02 11 . ± . α − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . Double log Φ ∗ − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . M ∗ . ± .
03 10 . ± .
05 10 . ± .
03 10 . ± .
03 10 . ± .
02 10 . ± .
02 10 . ± .
02 10 . ± . α − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . Pure-Disk log Φ ∗ − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . M ∗ . ± .
22 10 . ± .
18 10 . ± .
14 10 . ± .
08 10 . ± .
06 10 . ± .
06 10 . ± .
03 10 . ± . α − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . Elliptical log Φ ∗ − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . M ∗ . ± .
04 11 . ± .
09 11 . ± .
04 10 . ± .
05 10 . ± .
04 10 . ± .
04 10 . ± .
03 10 . ± . α − . ± . − . ± . − . ± . − . ± . − . ± .
07 0 . ± . − . ± . − . ± . Compact log Φ ∗ − − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . M ∗ − . ± .
74 9 . ± .
51 12 . ± .
57 10 . ± .
41 9 . ± .
17 10 . ± .
20 10 . ± . α − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . Hard log Φ ∗ − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . M ∗ . ± .
70 11 . ± .
56 12 . ± .
43 11 . ± .
33 11 . ± .
33 11 . ± .
19 11 . ± .
11 11 . ± . α − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . − . ± . MNRAS , 1–24 (2021)
EVILS: Mass Growth of Morphological Types Figure 22.
The evolution of the stellar mass density (SMD) of total and morphological types before applying the LSS corrections.Highlighted region shows the epoch covered by the GAMA data.MNRAS000