The nature and origins of the low surface brightness outskirts of massive, central galaxies in Subaru HSC
Thomas M. Jackson, Anna Pasquali, Francesco La Barbera, Surhud More, Eva K. Grebel
MMNRAS , 1–14 (2020) Preprint 5 February 2021 Compiled using MNRAS L A TEX style file v3.0
The nature and origins of the low surface brightness outskirts ofmassive, central galaxies in Subaru HSC
Thomas M. Jackson ★ , Anna Pasquali , Francesco La Barbera , Surhud More , ,Eva K. Grebel Astronomisches Rechen-Institut, Zentrum für Astronomie der Universität Heidelberg, Mönchhofstr. 12-14, 69120 Heidelberg, Germany INAF-Osservatorio Astronomico di Capodimonte, sal. Moiariello 16, Napoli, 80131, Italy Inter University Centre for Astronomy and Astrophysics, Ganeshkhind, Pune 411007, India Kavli Institute for the Physics and Mathematics of the Universe (WPI), 5-1-5 Kashiwanoha, Kashiwa 2778583, Japan
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
We explore the stellar mass density and colour profiles of 118 low redshift, massive, centralgalaxies, selected to have assembled 90 percent of their stellar mass 6 Gyr ago, finding evidenceof the minor merger activity expected to be the driver behind the size growth of quiescentgalaxies. We use imaging data in the 𝑔, 𝑟, 𝑖, 𝑧, 𝑦 bands from the Subaru Hyper Suprime-Camsurvey and perform SED fitting to construct spatially well-resolved radial profiles in colourand stellar mass surface density. Our visual morphological classification reveals that ∼ ∼
43 percent of theremaining sample display a diffuse stellar halo and only ∼
14 percent display no features, downto a limiting 𝜇 𝑟 − band ∼
28 mag arcsec − . We find good agreement between the stacked colourprofiles of our sample to those derived from previous studies and an expected smooth, decliningstellar mass surface density profile in the central regions (< 3 R e ). However, we also see aflattening of the profile ( Σ ∗ ∼ . M (cid:12) kpc − ) in the outskirts (up to 10 R e ), which is revealedby our method of specifically targeting tidal/accretion features. We find similar levels of tidalfeatures and behaviour in the stellar mass surface density profiles in a younger comparisonsample, however a lack of diffuse haloes. We also apply stacking techniques, similar to those inprevious studies, finding such procedures wash out tidal features and thereby produces smoothdeclining profiles. The stellar material in the outskirts contributes on average ∼ M (cid:12) or afew percent of the total stellar mass and has similar colours to SDSS satellites of similar stellarmass. Key words: galaxies: evolution – galaxies: interactions – galaxies: elliptical and lenticular,cD – galaxies: structure
Early studies of high redshift quiescent galaxies yielded the rathersurprising result that massive galaxies at high redshift have signif-icantly smaller effective radii than their low redshift counterparts,despite similar stellar masses (Daddi et al. 2005). Numerous followup studies found similar results (Trujillo et al. 2006, 2007; Zirmet al. 2007; van Dokkum et al. 2008; Cimatti et al. 2008). This sizedifference can amount to a factor of ∼ ★ E-mail: [email protected] © a r X i v : . [ a s t r o - ph . GA ] F e b Thomas M. Jackson et al. as the sole driver of these size growth processes. We note, how-ever, that some studies (e.g. de la Rosa et al. 2016) have shownthat some of these high redshift compact objects evolve differently,meaning that the population of present day ellipticals is likely tohave multiple evolutionary paths.Numerous mechanisms have therefore been suggested to ac-count for this size evolution. Some studies argue that major mergers(stellar mass ratio < 1:4) could increase the size of an object dra-matically (Feldmann et al. 2010). Other studies, however, state thatthis is likely to increase the stellar mass of an individual object toogreatly (van der Wel et al. 2008; Bezanson et al. 2009), especiallyin the central regions of massive galaxies (Naab et al. 2009), result-ing in much greater numbers of extremely high mass objects thanobserved in the local universe.Feedback from Active Galactic Nuclei or AGN is another ar-gument proposed that could cause the size growth of such galaxies(AGN, Fan et al. 2008, 2010). Numerous counter arguments tothis theory, however, have been presented (see Trujillo et al. 2009;Bezanson et al. 2009, and references therein), including that thefeedback would need to be extremely fine tuned to reproduce theproperties of local, massive ellipticals.The favoured theory to explain this size growth behaviour is atwo phase evolutionary scenario (e.g. Naab et al. 2007, 2009; Oseret al. 2012). The first phase is a rapid formation of the bulk ofthe stellar mass. This is followed by the accretion of material fromsatellites via gas-poor or “dry” minor mergers (stellar mass ratio (cid:38) 𝑧 = 𝑧 = 𝑧 (cid:46) 𝜇 𝑔 − band ∼
29 mag arcsec − in individual galaxies (Bílek et al.2020; Duc 2020). Simulations also show that as imaging data godeeper, an increasing number of tidal features and distorted stellarmaterial is revealed (Mancillas et al. 2019). However, a lot of theobservational studies considering larger samples, although findingsignificant signs of tidal material, generally have not attemptedto quantify the physical properties such as colour or stellar massof this material until very recently, namely due to a lack of multi-wavelength data that is deep enough. Properties of the material suchas colour and stellar mass (as well as the prevalence of this mergeractivity) can help us understand its origin and thereby possibleprocesses driving the size evolution of galaxies.In this work we attempt to address this by using deep imaging( 𝜇 𝑟 − band ∼
28 mag arcsec ) from the Hyper Suprime-Cam (HSC)Subaru Strategic Programme (-SSP) of an old, passive galaxy sam-ple in order to search for and quantify signatures of minor mergeractivity that may be driving the size growth of quiescent galaxies.We firstly morphologically classify our objects in order to quantifythe amount of merger activity present and the trends of activity withhalo and stellar mass. We then account for PSF effects and imple-ment Voronoi binning (Cappellari & Copin 2003) on the imagingdata to attain bins of pixels with high enough signal-to-noise ( 𝑆 / 𝑁 )in order to more accurately constrain fitted Spectral Energy Distri-butions (hereafter SEDs) using cigale (Noll et al. 2009; Boquienet al. 2019). We construct spatially resolved radial profiles in colourfrom the image photometry and estimates of the stellar mass sur-face density yielded by the SED fitting in order to investigate andquantify the nature of the tidal features. We repeat these processeson a younger comparison sample. We finally briefly compare ourresults to other techniques from previous studies on the size growthof massive galaxies.In Section 2 we present the observational data used in thispaper. In Section 3 we outline the methods used to analyse the data.We then present our results in Section 4. In Section 5 we discussthese results, before summarising our work in Section 6. Large scale extragalactic surveys such as the Sloan Digital Sky Sur-vey (SDSS, York et al. 2000) have revolutionised our understandingof galaxy evolution by providing a large sample of uniform imagingacross a significant area of the sky. The drawback with these sam-ples, however, is that they usually lack the depth of other surveysfor more detailed studies such as the Mass Assembly of early-TypeGaLAxies with their fine Structures (MATLAS) survey, usually car-ried out with larger aperture telescopes (Duc et al. 2015; Duc 2020).The disadvantage of such deeper observations however is that thecoverage of the sky is usually smaller. The HSC-SSP (Aihara et al.2018) is one survey that is trying to bridge this gap.The HSC-SSP is an ongoing survey designed to image sig-nificant parts of the sky, in 5 different bands ( 𝑔, 𝑟, 𝑖, 𝑧, 𝑦 ). 1400 sqdegrees of the sky are being imaged down to 26.1 mag (wide sur-vey), 27 sq degrees down to a magnitude of 27.1 mag (deep survey)and 3.5 sq degrees down to 27.7 mag (ultra-deep survey) in the 𝑟 -band. The camera has a pixel scale of 0.168” (Miyazaki et al.2018) and the survey has so far experienced a median seeing inthe 𝑖 -band of 0.6” (Aihara et al. 2018). In order to attain a statisti-cally significant sample we used the wide survey of the incremental MNRAS000
28 mag arcsec ) from the Hyper Suprime-Cam (HSC)Subaru Strategic Programme (-SSP) of an old, passive galaxy sam-ple in order to search for and quantify signatures of minor mergeractivity that may be driving the size growth of quiescent galaxies.We firstly morphologically classify our objects in order to quantifythe amount of merger activity present and the trends of activity withhalo and stellar mass. We then account for PSF effects and imple-ment Voronoi binning (Cappellari & Copin 2003) on the imagingdata to attain bins of pixels with high enough signal-to-noise ( 𝑆 / 𝑁 )in order to more accurately constrain fitted Spectral Energy Distri-butions (hereafter SEDs) using cigale (Noll et al. 2009; Boquienet al. 2019). We construct spatially resolved radial profiles in colourfrom the image photometry and estimates of the stellar mass sur-face density yielded by the SED fitting in order to investigate andquantify the nature of the tidal features. We repeat these processeson a younger comparison sample. We finally briefly compare ourresults to other techniques from previous studies on the size growthof massive galaxies.In Section 2 we present the observational data used in thispaper. In Section 3 we outline the methods used to analyse the data.We then present our results in Section 4. In Section 5 we discussthese results, before summarising our work in Section 6. Large scale extragalactic surveys such as the Sloan Digital Sky Sur-vey (SDSS, York et al. 2000) have revolutionised our understandingof galaxy evolution by providing a large sample of uniform imagingacross a significant area of the sky. The drawback with these sam-ples, however, is that they usually lack the depth of other surveysfor more detailed studies such as the Mass Assembly of early-TypeGaLAxies with their fine Structures (MATLAS) survey, usually car-ried out with larger aperture telescopes (Duc et al. 2015; Duc 2020).The disadvantage of such deeper observations however is that thecoverage of the sky is usually smaller. The HSC-SSP (Aihara et al.2018) is one survey that is trying to bridge this gap.The HSC-SSP is an ongoing survey designed to image sig-nificant parts of the sky, in 5 different bands ( 𝑔, 𝑟, 𝑖, 𝑧, 𝑦 ). 1400 sqdegrees of the sky are being imaged down to 26.1 mag (wide sur-vey), 27 sq degrees down to a magnitude of 27.1 mag (deep survey)and 3.5 sq degrees down to 27.7 mag (ultra-deep survey) in the 𝑟 -band. The camera has a pixel scale of 0.168” (Miyazaki et al.2018) and the survey has so far experienced a median seeing inthe 𝑖 -band of 0.6” (Aihara et al. 2018). In order to attain a statisti-cally significant sample we used the wide survey of the incremental MNRAS000 , 1–14 (2020)
SB galaxy outskirts in HSC L o g M H a l o ( M ) Passive galaxy Sample
DiffuseFeaturesFeatureless
Counts
Log M Stellar (M ) C o un t s Redshift L o g M H a l o ( M ) Comparison Sample
DiffuseFeaturesFeatureless
Counts
Log M Stellar (M ) C o un t s Redshift
Figure 1.
The main panel of each sample shows the distribution of our observational sample in stellar and halo mass phase space. Black points give the parentsample of central galaxies as detailed in Jackson et al. (2020). Blue (green, red) points show galaxies with signs of merger activity (a diffuse stellar halo, noactivity) as determined by the visual classification described in Section 3.1. The side panel and bottom panel of each sample show the distributions in stellarand halo mass of each classified sub-sample. The small bottom right panel of each sample shows the distribution of each sub-sample in redshift. data release 2 (300 sq degrees, Aihara et al. 2019). The HSC datareduction pipeline (Bosch et al. 2018) automatically performs skysubtraction, bias, flat field correction and flux calibration. Huanget al. (2018a) injected synthetic galaxies into HSC images, achiev-ing a 13 and 18 percent precision at a depth of 20 and 25 magrespectively in the i -band for extended objects when fitting singleSérsic profiles. The HSC fields have been chosen especially for theirlow levels of extinction along the line of sight, however we still ap-plied a correction for Galactic foreground reddening to each imageaccording to the law of Cardelli et al. (1989), using the extinctionmaps of Schlafly & Finkbeiner (2011).The parent sample is selected from Jackson et al. (2020): Inthis work, we tracked the stellar mass assembly of central galaxiesin the SDSS (York et al. 2000) as a function of their stellar and halomass using group catalogues (Lim et al. 2017). We used the timesat which galaxies assembled 10, 50 and 90 per cent (hereafter 𝑡 , 𝑡 and 𝑡 ), of their stellar mass as determined from the SED fittingtechniques of Pacifici et al. (2012, 2016). These were derived usingsemi-analytic simulations to generate realistic, non-parametric starformation histories, in conjunction with stellar population models.They were then treated with effects such as dust modelling andnebular emission to generate a library of ∼ Halo > M (cid:12) wereapplied in order to select groups and clusters of galaxies.To obtain the galaxy sample for this study, we applied the fol-lowing cuts to the parent sample of ∼ 𝑡 value larger than 6 Gyr (reducing the sample to ∼ < 𝑧 < 𝑔, 𝑟, 𝑖, 𝑧, 𝑦 , reducing the sample to 134 galaxies)and a minimum of 50 bins at a 𝑆 / 𝑁 >
10 in its outskirts (beyondthe SDSS Petrosian radius of ∼ e as given in Pacifici et al. 2016)as computed by the Voronoi binning (see Section 3.2). Selectinggalaxies that have a minimum number of the bins in the outskirtsmay slightly bias our results, whereby galaxies with little activityare likely to have less light/mass excess in their outskirts and aretherefore more likely to be discarded, however we stress that only4 galaxies are discarded from this specific selection process, so wedo not expect our results to be significantly impacted. Two galaxiescontained in the Kauffmann et al. (2003) catalogues defined as AGNor LINER (based on the selection methods of Baldwin et al. 1981)were also removed so as to avoid radiation from AGN biasing ourresults.In the visual classification we also removed galaxies that hadproblems in the data with at least 1 band, such as missing data, over-saturation of pixels or imaging artefacts (10 galaxies). The finalsample contains 118 centrals from the parent sample of ∼ , 𝑡 < MNRAS , 1–14 (2020)
Thomas M. Jackson et al. same scheme for the subplots as for the passive galaxy sample inthe left half.We see that each distribution spans a significant range in halomass ( ∼ ∼ Halo (cid:38) M (cid:12) ) and those that displayno tidal features (Featureless) generally occupy the least massivehaloes (M Halo (cid:46) M (cid:12) ). The galaxies that display tidal features,indicative of merger activity (Features) display a tendency to occupymore massive haloes, however are distributed across the full rangeof halo masses. These trends are similar in stellar mass, in line withfindings from previous studies (Bílek et al. 2020). Bílek et al. (2020)find 1.7 times the amount of tidal features above a stellar mass of10 M (cid:12) than below, which compares to an increase of 1.2 in oursample (a difference of 1 𝜎 ). This difference may be due to slightlydifferent classification schemes or, more likely, the different datasets, whereby the data of Bílek et al. (2020) is deeper than that usedin this study. They also find an increase in some of their featureswith increasing environmental density, similar to our results. In thebottom right panel, we see a fairly even distribution of all threesub-samples across redshift. The implications of this behaviour arediscussed further in Section 5.2.We also visually classified galaxies into early and late type,finding that ∼ ∼
26 per cent for our comparison sample, with ∼
74 per cent of galaxies classified as early type. To reinforce this,we calculated the average Sérsic index of each sample using thecatalogues of Blanton et al. (2005) and the bulge to total mass ratiosusing the catalogues of Mendel et al. (2014). We find that the averageSérsic index of the passive galaxy sample is 5.00 with an averagebulge to total mass ratio of 0.78, compared to an average Sérsicindex of 3.14 for the comparison sample with an average bulge tototal mass ratio of 0.60. This shows that our older, passive galaxysample is dominated by early type morphologies, which have largerSérsic indices than the comparison sample.
Visual morphological classification, although with a reasonable de-gree of subjectivity, has proven to be an extremely useful tool in thestudy of galaxy evolution, with one of the first and the most usedclassification systems first developed by Hubble (1926). Morpho-logical classification can be used as an indicator of the current stageof evolution that a galaxy may be at in its lifetime and morpho-logical features or disruptions can help indicate if a galaxy may beundergoing environmental processes such as merging or stripping.Today, researchers can use a number of methods; they can eithermanually classify their own data set, or if it is too large, publicschemes such as Galaxy Zoo (Lintott et al. 2008) have been set up.Machine learning techniques have also been and are being devel-oped for this application (e.g. Banerji et al. 2010; Hocking et al.2018).As both of our samples are relatively small, each galaxy wasindependently visually inspected and classified by different co-authors, changing the contrast of each image in the i -band to searchfor features, with a high level of agreement of the classifications be-tween the different co-authors. The i -band was chosen as it has the best median seeing ( ∼ . In order to analyse the data while accounting for biases and reducingerrors we implemented the following procedures on the imagingdata.In order to account for the effects of the Point Spread Function(hereafter PSF) over all bands we measured the Full-Width-Half-Maximum (FWHM) for each galaxy in each band as given by thePSF models constructed by the HSC pipeline (see Bosch et al.2018, for more details). We took the largest FWHM in the entiregalaxy sample and convolved each galaxy in each band with thequadrature of the difference between that band and the maximumoverall FWHM in order to account for the worst PSF across theentire sample. We note that this technique has been used in themajority of previous studies investigating similar galaxy properties,but that this technique does not take into account the effects of thewings of the PSF. We address this in Section 4.3.We then assigned a threshold 𝑆 / 𝑁 > photoutils package in astropy inorder to locate all pixels associated with each galaxy, including thetidal features. The routine uses an algorithm based on the numberof neighbouring pixels also associated with the source to find allpixels associated with a local maximum. Examples of this can beseen in the top two rows of Figure 2, where the top rows show theoriginal i -band images from HSC and the second rows show thedetected galaxies in the r -band surface brightness maps.Some pixels in our sources have low signal-to-noise, however, MNRAS000
Visual morphological classification, although with a reasonable de-gree of subjectivity, has proven to be an extremely useful tool in thestudy of galaxy evolution, with one of the first and the most usedclassification systems first developed by Hubble (1926). Morpho-logical classification can be used as an indicator of the current stageof evolution that a galaxy may be at in its lifetime and morpho-logical features or disruptions can help indicate if a galaxy may beundergoing environmental processes such as merging or stripping.Today, researchers can use a number of methods; they can eithermanually classify their own data set, or if it is too large, publicschemes such as Galaxy Zoo (Lintott et al. 2008) have been set up.Machine learning techniques have also been and are being devel-oped for this application (e.g. Banerji et al. 2010; Hocking et al.2018).As both of our samples are relatively small, each galaxy wasindependently visually inspected and classified by different co-authors, changing the contrast of each image in the i -band to searchfor features, with a high level of agreement of the classifications be-tween the different co-authors. The i -band was chosen as it has the best median seeing ( ∼ . In order to analyse the data while accounting for biases and reducingerrors we implemented the following procedures on the imagingdata.In order to account for the effects of the Point Spread Function(hereafter PSF) over all bands we measured the Full-Width-Half-Maximum (FWHM) for each galaxy in each band as given by thePSF models constructed by the HSC pipeline (see Bosch et al.2018, for more details). We took the largest FWHM in the entiregalaxy sample and convolved each galaxy in each band with thequadrature of the difference between that band and the maximumoverall FWHM in order to account for the worst PSF across theentire sample. We note that this technique has been used in themajority of previous studies investigating similar galaxy properties,but that this technique does not take into account the effects of thewings of the PSF. We address this in Section 4.3.We then assigned a threshold 𝑆 / 𝑁 > photoutils package in astropy inorder to locate all pixels associated with each galaxy, including thetidal features. The routine uses an algorithm based on the numberof neighbouring pixels also associated with the source to find allpixels associated with a local maximum. Examples of this can beseen in the top two rows of Figure 2, where the top rows show theoriginal i -band images from HSC and the second rows show thedetected galaxies in the r -band surface brightness maps.Some pixels in our sources have low signal-to-noise, however, MNRAS000 , 1–14 (2020)
SB galaxy outskirts in HSC a r c s e c L o g S r b a n d ( C o un t s ) a r c s e c i b a n d ( m a g / a r c s e c ) a r c s e c g - r ( m a g ) arcsec a r c s e c M * ( M / k p c a r c s e c L o g S r b a n d ( C o un t s ) a r c s e c i b a n d ( m a g / a r c s e c ) a r c s e c g - r ( m a g ) arcsec a r c s e c M * ( M / k p c a r c s e c L o g S r b a n d ( C o un t s ) a r c s e c i b a n d ( m a g / a r c s e c ) a r c s e c g - r ( m a g ) arcsec a r c s e c M * ( M / k p c a r c s e c L o g S r b a n d ( C o un t s ) a r c s e c i b a n d ( m a g / a r c s e c ) a r c s e c g - r ( m a g ) arcsec a r c s e c M * ( M / k p c Figure 2.
From the top to bottom panels; The i -band image of a galaxy, the Voronoi binned r -band surface brightness map, the Voronoi binned g - r colour mapand the stellar mass density map as yielded by the SED fitting on the Voronoi binned maps. From left to right; an example of a galaxy classified as displayingno merger activity, a galaxy classified as displaying a diffuse stellar halo, a galaxy with a stream of material to the top right and a galaxy that displays a slightshell structure to the bottom right. just reaching the threshold 𝑆 / 𝑁 of 3 per pixel. These are usuallythe low surface brightness features we want to quantify. In orderto reduce uncertainties in the estimates yielded by the SED fittingprocess we used Voronoi binning (Cappellari & Copin 2003) tomaximise the signal-to-noise of these features, while still preserv-ing the spatial resolution of higher 𝑆 / 𝑁 areas. We took the SDSSPetrosian apertures ( ∼ e ) for the central areas of our galaxies asthis is the aperture used in the study of Pacifici et al. (2016) fromwhich the parent sample is built in Jackson et al. (2020) and fromwithin which many of the integrated galaxy properties are estimatedsuch as 𝑡 . We then Voronoi binned all pixels inside of this aper-ture to a constant 𝑆 / 𝑁 = 50 per bin, as the signal-to-noise per pixelin these regions is high. For all pixels outside of this aperture, theoutskirts of the galaxy, we defined a constant 𝑆 / 𝑁 = 10 per bin, asthis is a high enough signal-to-noise to reduce uncertainties in thestellar mass estimates while avoiding the washing out of detail thatcan occur when Voronoi binning at a signal-to-noise of 50. Thisallowed us to probe features that have a surface brightness as low as ∼
28 mag arcsec − in the r -band. We also masked bins with contani- mation from significant foreground and background objects in orderto reduce contamination that could bias our results. We note thatthis makes up only 1.4 percent of all bins, and although increaseswith radius, due to less bins at higher radii, never contributes above2.5 percent of all bins, hence we do not expect this to significantlybias our radial profiles. As tidal features are not uniform, we donot attempt to interpolate across neighbouring bins as this methodwould introduce its own biases to our results.Total fluxes were then calculated for each Voronoi bin fromthe imaging data in the following manner: The median flux of eachVoronoi bin was calculated and multiplied by the amount of pixelsin the Voronoi bin, so as to avoid fluctuations due to contaminatingobjects that can bias the mean, especially at low surface brightness.Associated errors were calculated using the variance maps added inquadrature with the flux calibration error. The fluxes and associatederrors were then converted into magnitudes for colour profiles, withan median error of ∼ MNRAS , 1–14 (2020)
Thomas M. Jackson et al. S h e ll s R i n g S t r e a m D i s t o r t i o n U m b r e ll a M i n o r m e r g e r C o un t s Passive galaxy sampleComparison sampleFeatureless14.4%Diffuse 43.2% Features42.4%
Passive Galaxy Sample
Featureless 41.7% Diffuse19.1% Features39.2%
Comparison Sample
Figure 3.
The left hand panel shows a pie chart of the percentage of the passive galaxy sample that was classified as either displaying visual signs of mergeractivity (Features), no activity but a diffuse stellar halo (Diffuse), or no activity at all (Featureless). The central panel shows the same as the left hand panel butfor our younger comparison sample. The right hand panel shows the sub-samples of galaxies that were classified as displaying visual signs of merger activity,split dependent on the type of merger activity. bin, to verify if our adaptive Voronoi binning process introducesbiases to our results, finding minimal differences in the profiles.The results of this process can be seen in Figure 2, where thethird panel shows the corresponding g - r colour map and the bottompanel shows the corresponding stellar mass surface density map asrecovered by the SED fitting process described in Section 3.3. Fromleft to right we see an example of a Featureless galaxy, a Diffusegalaxy, a galaxy that exhibits a stream and a galaxy that displays ashell (both classified in the Features sub-sample). In order to estimate the stellar mass corresponding to each Voronoibin in our data, we used the SED fitting software cigale (Burgarellaet al. 2005; Noll et al. 2009; Boquien et al. 2019). We chose cigaledue to its speed, since due to the resolution and depth of our imag-ing data, the Voronoi binning process yielded over 500,000 SEDsthat needed fitting for each sample. cigale can fit SEDs spanningfrom the Ultraviolet (UV) to the Infrared (IR) range of the elec-tromagnetic spectrum and is based on an energy balance principle,whereby the attenuation due to dust of the UV radiation producedby massive young stars is expected to be re-radiated in the IR part ofthe spectrum. It also has models to account for AGN emission, dustattenuation, multiple star formation histories and stellar populationmodels.As we expected these systems to be dominated by old stellarmaterial, however with the possibility of some recent star forma-tion triggered by interactions, we generated models in cigale basedon double-exponential star formation histories spanning ranges be-tween 2 and 12 Gyr for the main burst. As our passive galaxiesare selected to have formed 90 percent of their stellar mass over 6Gyr ago (this is between 3 and 4 Gyr for our comparison samplein almost all cases), we limit the additional burst to a contributionbetween 0.1 and 5 per cent of the mass with ages ranging between100 Myr and 2 Gyr. The best fit SFH contains a 0.1 percent sec-ondary burst in the overwhelming majority of cases. We then usedthese star formation histories with the synthetic stellar libraries ofBruzual & Charlot (2003), assuming a Chabrier (2003) initial massfunction and metallicites spanning the full range from 1/30 th solarto 4 times solar in order to generate the model SEDs to fit to the HSC photometry. We neglected AGN contributions to the modelling aswe removed AGN from the sample as described in Section 2.We also inputted the g , r , i , z and y HSC filter functions intocigale and fit the SEDs. The average uncertainties yielded by theSED fitting process for the estimations of stellar masses are of theorder of 0.3 dex, which when accounted for in the stellar masssurface densities is, on average, smaller than the scatter in the cor-responding radial profiles of our sample of galaxies, meaning theerrors do not significantly bias our results.
Figure 3 shows the statistics yielded by our morphological classi-fication. The left hand panel shows the split of our passive galaxysample into those centrals that display tidal features (Features),which make up 42.4 per cent of the sample. We note that this per-centage is very similar to previous studies such as Duc et al. (2015).Those that show a diffuse stellar halo (Diffuse) make up 43.2 percent of our sample and those that are Featureless contribute 14.4per cent of our passive galaxy sample.When we compare these statistics to the comparison sampleof younger central galaxies in the central panel we see that the per-centage of centrals classified with Features remains similar at 39.2per cent. The percentage of galaxies classified as Diffuse, however,drops to 19.1 per cent and the percentage of galaxies classified asFeatureless increases to 41.7 per cent. Assuming Poissonian errorson each sub-sample, we can calculate that the differences in theDiffuse and Featureless sub-samples seen between the passive andcomparison samples are significant (> 5 𝜎 ), however are not signif-icant (< 1 𝜎 ) in those galaxies classified as exhibiting features. Wediscuss these differences and possible scenarios for this behaviourin Section 5.2.The right hand panel shows the further morphological classifi-cation we used on the Features sub-sample of galaxies. We note thatsome galaxies may display different kinds of features simultane-ously and therefore may count many times in this specific sub-plot.We see a range of different features, from shells and significant dis-tortions in the light distribution to minor mergers and rings. In our MNRAS000
Figure 3 shows the statistics yielded by our morphological classi-fication. The left hand panel shows the split of our passive galaxysample into those centrals that display tidal features (Features),which make up 42.4 per cent of the sample. We note that this per-centage is very similar to previous studies such as Duc et al. (2015).Those that show a diffuse stellar halo (Diffuse) make up 43.2 percent of our sample and those that are Featureless contribute 14.4per cent of our passive galaxy sample.When we compare these statistics to the comparison sampleof younger central galaxies in the central panel we see that the per-centage of centrals classified with Features remains similar at 39.2per cent. The percentage of galaxies classified as Diffuse, however,drops to 19.1 per cent and the percentage of galaxies classified asFeatureless increases to 41.7 per cent. Assuming Poissonian errorson each sub-sample, we can calculate that the differences in theDiffuse and Featureless sub-samples seen between the passive andcomparison samples are significant (> 5 𝜎 ), however are not signif-icant (< 1 𝜎 ) in those galaxies classified as exhibiting features. Wediscuss these differences and possible scenarios for this behaviourin Section 5.2.The right hand panel shows the further morphological classifi-cation we used on the Features sub-sample of galaxies. We note thatsome galaxies may display different kinds of features simultane-ously and therefore may count many times in this specific sub-plot.We see a range of different features, from shells and significant dis-tortions in the light distribution to minor mergers and rings. In our MNRAS000 , 1–14 (2020)
SB galaxy outskirts in HSC R / R e g - r ( m a g ) Passive galaxy profileLa Barbera et al. 2012
R / R e L o g M S t e ll a r ( M k p c ) Figure 4.
The left hand panel displays the stacked radial profile (in bins of 0.1 R/R e ) of g - r colour for the 118 galaxies in our older ( 𝑡 > th , 16 th , 25 th , 75 th , 84 th and 95 th percentiles of each stacked bin. For comparison, the results from the study of La Barbera et al. (2012) are given by the solid red line and shaded areas. Theright hand panel shows the stacked radial profile (also in bins of 0.1 R/R e ) of the stellar mass surface density with corresponding percentiles, normalised by theintegrated stellar mass within 1 R e . R / R e g - r ( m a g ) Comparison profilePassive galaxy profileLa Barbera et al. 2012
R / R e L o g M S t e ll a r ( M k p c ) Comparison profilePassive galaxy profile
Figure 5.
The left hand panel displays the stacked radial profile (in bins of 0.1 R/R e ) of g - r colour for the 118 younger ( 𝑡 < th , 16 th , 25 th , 75 th , 84 th and 95 th percentilesof each stacked bin. For comparison, the results from the study of La Barbera et al. (2012) are given by the solid red line and shaded areas and the median ofthe passive galaxy sample is given by the black dotted line. The right hand panel shows the stacked radial profile (also in bins of 0.1 R/R e ) of the stellar masssurface density with corresponding percentiles, normalised by the integrated stellar mass within 1 R e . The median of the passive galaxy sample is given by theblack dotted line. passive galaxy sample we see that distortions and shells are the mostprevalent signs of merger activity with respect to the comparisonsample. The comparison sample, on the other hand, shows an in-creased amount of streams compared to our passive galaxy sample,with all three signs of merger activity the most prevalent in bothgalaxy samples. When we calculate the Poissonian errors, we findno significant differences (< 1 𝜎 ) in the number of shells, rings andumbrellas, differences above 1 𝜎 n the number of streams and minormergers and differences above 2 𝜎 in the numbers of distortions.We caution, however, against over interpretation due to relativelysmall number statistics. One of the most effective ways to investigate the star formation andassembly histories of galaxies is to plot radial profiles of various properties. This can help reveal possible evolutionary scenarioswhich a galaxy has undergone. For example metallicity and ageprofiles in the central regions may reveal inside out star formation(e.g. La Barbera et al. 2011). The outskirts, however are dominatedby ex-situ stars according to simulations (Davison et al. 2020), henceprofiles out to large R e reveal the nature of accreted material (LaBarbera et al. 2012; Hirschmann et al. 2015; Davison et al. 2020),which is the aim of this study. We therefore utilise this technique toexplore the nature of the colour and stellar mass surface density ofthe accreted material in both galaxy samples.From the Voronoi binning process we calculated an averagedistance for each bin to the centre of the galaxy. Using this wethen stacked every Voronoi bin from every galaxy in our passivegalaxy sample in bins of 0.1 R/R e (where R e is the effective radiusas measured by SDSS, in order to be able to better compare theseresults to previous studies), and constructed distributions of g - MNRAS , 1–14 (2020)
Thomas M. Jackson et al. r colour and stellar mass surface density. We note that very fewVoronoi bins (< 0.1 percent) contains pixels in multiple radii. Fromthe distribution in each bin we calculate the median of each bin,and the 5 th , 16 th , 25 th , 75 th , 84 th and 95 th percentiles to constructour stacked radial profiles. We also normalise each bin of stellarmass density by the total stellar mass within R e of each galaxyfollowing previous studies (we note that normalising by 2 or 3 R e has a negligible difference on the results).Figure 4 shows the g - r colour profile of our passive galaxysample. The black solid line shows the median value of the distri-bution in each bin of R/R e , the dark grey shading shows the 25 th and 75 th percentiles, the mid grey the 16 th and 84 th percentiles andthe light grey the 5 th and 95 th percentiles. We see a slight decreaseof the order of 0.1 dex in the g - r colour in the inner 2 R e , witha very narrow distribution. As we go beyond this radius towards 6R e , the g - r colour still declines, becoming bluer by another 0.1dex, however the distribution experiences much greater range (upto ∼ e , the stellar material is bluer,reaching a median g - r colour of 0.4 mag. On closer inspection ofthe maps of the galaxies predominantly contributing at these radii,this material is dominated by incoming streams and from satellitegalaxies in the process of minor mergers, however we note that thereare relatively few bins (between 10 and 100), meaning we have tobe careful about how we interpret these results.We also observe similar gradual decline of ∼ g - i , r - i , i - z ) within 6 R e , howeverbeyond this the r - i and i - z colour profiles continue to graduallydecline rather than dropping ∼ . and 10 . M (cid:12) , and contained in the SDSSand UKIRT Infrared Deep Sky Survey (UKIDSS) out to 8 R e . Theirresults are represented by the solid red line for the median and thered shading indicating the 16 th and 84 th percentiles. We find goodagreement between our results and those from La Barbera et al.(2012) as the values are well within the scatter of our distributionat almost all radii.The right hand panel of Figure 4 shows the stellar mass sur-face density profile as yielded by the estimates of stellar mass fromcigale. Once again the median of each distribution in a bin of 0.1R/R e is represented by the black solid line with the 5 th , 16 th , 25 th ,75 th , 84 th and 95 th percentiles given by the various shading lev-els. We see a fairly smoothly declining profile within ∼ e , asexpected, but beyond this radius we observe a flattening of the aver-age stellar mass density profile, with some bumps. This behaviouroccurs not only in the median values of the distribution but also invarious percentiles of the distribution. This indicates that significantstellar material is present in the observed tidal features we see inthe imaging.Figure 5 shows the same radial profiles for our comparisongalaxy sample, with the median (in bins of 0.1 R/R e ) represented bythe solid blue line and the shaded regions indicating the 5 th , 16 th ,25 th , 75 th , 84 th and 95 th percentiles, the same as for our passivegalaxy sample in Figure 4. The median g - r colour for our passivegalaxy sample is given by the dotted black line for comparison. Wesee slightly bluer average colours of the order of ∼ ∼ e . Beyond this point,much of the material is redder than our passive galaxy sample or themeasured profiles of La Barbera et al. (2012, given by the solid red line). We note, however, that the profile beyond this point relies onless than 40 Voronoi bins, which introduces large variations. Thesespikes are predominantly driven by streams of accreted material andminor merger remnants, similar to the passive galaxy sample.The stellar mass density profile of our comparison sample(median represented by the solid blue line with shaded regionsshowing the various percentiles of the distribution) shows similarbehaviour to our passive galaxy sample (median represented by thedotted black line), whereby we observe a steeply declining profilein the inner parts of the galaxy before a flattening of the profile atlarger radii with some bumps. Some previous works have carried out similar studies of the growthprocesses of massive galaxies including the stellar mass content oftheir outskirts such as that of van Dokkum et al. (2010). In theirstudy, they took imaging data from a number of surveys such asSDSS and their own NEWFIRM Medium Band Survey and stackedthe galaxy images in bins of redshift, normalised and constructedradial profiles of the stellar mass density. Their profiles showed anincrease in the average stellar mass density in the outskirts of theirsample with decreasing redshift, which drives the increase in effec-tive radius and thereby the measured size of their galaxy sample.These profiles, however, are smooth in contrast to our passive galaxyand comparison galaxy profiles. This is because averaging, eitherthrough stacking or isophotal analysis washes out or dampens thesignal of specific merger features which generally have an asymetricorientation on the sky compared to the host galaxy. To check thatthis hypothesis holds, we employed similar processes, stacking ourgalaxy images and running source detection, Voronoi binning, andSED fitting, to see if we replicate similar trends to van Dokkumet al. (2010).We took the galaxy images convolved to the worst seeing (asdescribed in Section 3.2 in each band) and magnified each imageto the median redshift of 0.806 using iraf in order to bring eachgalaxy onto the same physical spatial scale while conserving theflux (i.e. the sum of the total fluxes in all of the old and new pixels isequal). We normalised each flux value by the integrated stellar massas in van Dokkum et al. (2010), corrected for redshift dimming andthen stacked all images in each band, taking the median for eachstacked pixel. We then applied the same Voronoi binning procedureas before, measured the g and r colours, and fitted the photometryfor each Voronoi bin with an SED to yield stellar mass estimatesfor the stellar mass surface density profiles. We then constructedsimilar radial profiles as the previous section.In order to compare with our radial profiles in Section 4.2,we also multiplied the effective radii of each galaxy by the samemagnification factors yielding a median angular effective radius of3.4 arcsec. The results are shown in Figure 6, with the solid blackline representing the median with the accompanying grey shadedareas representing the same percentile limits as Figures 4 and 5.As stated earlier, the wings of the PSF may also have an effecton surface brightness profiles, and thereby on the colour and stellarmass surface density profiles. To investigate the effect this may haveon our results, we constructed full PSFs to be convolved with ourstacked galaxy image. For each galaxy in our sample we attemptedto select a field star in the vicinity that was suitably isolated andreasonably bright. Altogether we found 95 suitable field stars from http://ast.noao.edu/data/software MNRAS000
Thomas M. Jackson et al. r colour and stellar mass surface density. We note that very fewVoronoi bins (< 0.1 percent) contains pixels in multiple radii. Fromthe distribution in each bin we calculate the median of each bin,and the 5 th , 16 th , 25 th , 75 th , 84 th and 95 th percentiles to constructour stacked radial profiles. We also normalise each bin of stellarmass density by the total stellar mass within R e of each galaxyfollowing previous studies (we note that normalising by 2 or 3 R e has a negligible difference on the results).Figure 4 shows the g - r colour profile of our passive galaxysample. The black solid line shows the median value of the distri-bution in each bin of R/R e , the dark grey shading shows the 25 th and 75 th percentiles, the mid grey the 16 th and 84 th percentiles andthe light grey the 5 th and 95 th percentiles. We see a slight decreaseof the order of 0.1 dex in the g - r colour in the inner 2 R e , witha very narrow distribution. As we go beyond this radius towards 6R e , the g - r colour still declines, becoming bluer by another 0.1dex, however the distribution experiences much greater range (upto ∼ e , the stellar material is bluer,reaching a median g - r colour of 0.4 mag. On closer inspection ofthe maps of the galaxies predominantly contributing at these radii,this material is dominated by incoming streams and from satellitegalaxies in the process of minor mergers, however we note that thereare relatively few bins (between 10 and 100), meaning we have tobe careful about how we interpret these results.We also observe similar gradual decline of ∼ g - i , r - i , i - z ) within 6 R e , howeverbeyond this the r - i and i - z colour profiles continue to graduallydecline rather than dropping ∼ . and 10 . M (cid:12) , and contained in the SDSSand UKIRT Infrared Deep Sky Survey (UKIDSS) out to 8 R e . Theirresults are represented by the solid red line for the median and thered shading indicating the 16 th and 84 th percentiles. We find goodagreement between our results and those from La Barbera et al.(2012) as the values are well within the scatter of our distributionat almost all radii.The right hand panel of Figure 4 shows the stellar mass sur-face density profile as yielded by the estimates of stellar mass fromcigale. Once again the median of each distribution in a bin of 0.1R/R e is represented by the black solid line with the 5 th , 16 th , 25 th ,75 th , 84 th and 95 th percentiles given by the various shading lev-els. We see a fairly smoothly declining profile within ∼ e , asexpected, but beyond this radius we observe a flattening of the aver-age stellar mass density profile, with some bumps. This behaviouroccurs not only in the median values of the distribution but also invarious percentiles of the distribution. This indicates that significantstellar material is present in the observed tidal features we see inthe imaging.Figure 5 shows the same radial profiles for our comparisongalaxy sample, with the median (in bins of 0.1 R/R e ) represented bythe solid blue line and the shaded regions indicating the 5 th , 16 th ,25 th , 75 th , 84 th and 95 th percentiles, the same as for our passivegalaxy sample in Figure 4. The median g - r colour for our passivegalaxy sample is given by the dotted black line for comparison. Wesee slightly bluer average colours of the order of ∼ ∼ e . Beyond this point,much of the material is redder than our passive galaxy sample or themeasured profiles of La Barbera et al. (2012, given by the solid red line). We note, however, that the profile beyond this point relies onless than 40 Voronoi bins, which introduces large variations. Thesespikes are predominantly driven by streams of accreted material andminor merger remnants, similar to the passive galaxy sample.The stellar mass density profile of our comparison sample(median represented by the solid blue line with shaded regionsshowing the various percentiles of the distribution) shows similarbehaviour to our passive galaxy sample (median represented by thedotted black line), whereby we observe a steeply declining profilein the inner parts of the galaxy before a flattening of the profile atlarger radii with some bumps. Some previous works have carried out similar studies of the growthprocesses of massive galaxies including the stellar mass content oftheir outskirts such as that of van Dokkum et al. (2010). In theirstudy, they took imaging data from a number of surveys such asSDSS and their own NEWFIRM Medium Band Survey and stackedthe galaxy images in bins of redshift, normalised and constructedradial profiles of the stellar mass density. Their profiles showed anincrease in the average stellar mass density in the outskirts of theirsample with decreasing redshift, which drives the increase in effec-tive radius and thereby the measured size of their galaxy sample.These profiles, however, are smooth in contrast to our passive galaxyand comparison galaxy profiles. This is because averaging, eitherthrough stacking or isophotal analysis washes out or dampens thesignal of specific merger features which generally have an asymetricorientation on the sky compared to the host galaxy. To check thatthis hypothesis holds, we employed similar processes, stacking ourgalaxy images and running source detection, Voronoi binning, andSED fitting, to see if we replicate similar trends to van Dokkumet al. (2010).We took the galaxy images convolved to the worst seeing (asdescribed in Section 3.2 in each band) and magnified each imageto the median redshift of 0.806 using iraf in order to bring eachgalaxy onto the same physical spatial scale while conserving theflux (i.e. the sum of the total fluxes in all of the old and new pixels isequal). We normalised each flux value by the integrated stellar massas in van Dokkum et al. (2010), corrected for redshift dimming andthen stacked all images in each band, taking the median for eachstacked pixel. We then applied the same Voronoi binning procedureas before, measured the g and r colours, and fitted the photometryfor each Voronoi bin with an SED to yield stellar mass estimatesfor the stellar mass surface density profiles. We then constructedsimilar radial profiles as the previous section.In order to compare with our radial profiles in Section 4.2,we also multiplied the effective radii of each galaxy by the samemagnification factors yielding a median angular effective radius of3.4 arcsec. The results are shown in Figure 6, with the solid blackline representing the median with the accompanying grey shadedareas representing the same percentile limits as Figures 4 and 5.As stated earlier, the wings of the PSF may also have an effecton surface brightness profiles, and thereby on the colour and stellarmass surface density profiles. To investigate the effect this may haveon our results, we constructed full PSFs to be convolved with ourstacked galaxy image. For each galaxy in our sample we attemptedto select a field star in the vicinity that was suitably isolated andreasonably bright. Altogether we found 95 suitable field stars from http://ast.noao.edu/data/software MNRAS000 , 1–14 (2020) SB galaxy outskirts in HSC R / R e g - r ( m a g ) Passive galaxy medianStacked profilePSF convolved profileLa Barbera et al. 2012
R / R e L o g M S t e ll a r ( M k p c ) Passive galaxy medianStacked profilePSF convolved profile
Figure 6.
The left hand panel shows the g - r colour profile for our stacked galaxy. The solid black line indicates the median of the profile, with the variousshadings representing the 5 th , 16 th , 25 th , 75 th , 84 th and 95 th percentiles of each stacked bin. The golden lines represent the stacked galaxy profile whenconvolved with the constructed PSF. For comparison the results of La Barbera et al. (2012) are indicated by the solid red line and shading and the dotted blackline represents the median profile from our passive galaxies in Figure 4. The right hand panel shows the stellar mass surface density profile for our stackedgalaxy (black line) with respective percentiles and for the accounting of the constructed PSF (golden line). For comparison the observed profile in Figure 4 isshown by the black dotted line.
118 galaxies, choosing at random one of the 95 stars as a replacementfor those galaxies without a star. As the radial profiles extend to ∼
10 R e , we cut out stellar images in each of the HSC bands whichextend to these angular scales ( ∼
30 arcsec radius). We magnifiedeach PSF image in the same way as its corresponding galaxy, thenstacked and normalised the PSF images in each band to produce afinal median PSF in each band.We then took the stacked galaxy images in each band, fittingeach with a two component profile containing a bulge and diskcontribution in two dimensions. This fitting process yielded a bulgeto total luminosity fraction of 0.85, slightly higher than our averagebulge to total mass fraction of 0.78 measured in our sample. Wefind the bulge component has a Sérsic index of 8.0, whereas thedisk component is fixed at an index of 1. Once again, on averageslightly higher than the average values ( 𝑛 = 5.00) retained from thecatalogues. These slight differences are likely due to the differentdepths of the data from the study of (van Dokkum et al. 2010)and this one, the treatment of the PSF and slight differences in thefitting algorithms, even though both fit two dimensional bulge anddisk models.We then convolved each fitted profile with the constructedPSFs in each band, to produce final images of the stacked and PSFconvolved galaxy. We ran the same process of Voronoi binning andSED fitting using cigale in order to estimate stellar mass surfacedensities, and subsequently used these to construct the profiles seenin Figure 6. The solid golden line represents the median in eachbin of 0.1 R/R e and accompanying shaded areas are as in previousfigures.We observe little difference in the g - r colour profiles in boththe stacked galaxy profile (median given by the solid black line andaccompanying grey shaded areas showing the various percentiles)and when convolved with the constructed PSF (median given by thegolden line with corresponding shaded areas showing the variouspercentiles), with differences less than 0.1 mag at almost all radii.The stacked galaxy stellar mass density profile appears as a smoothdecline out to ∼ e , similar to previous works. As may be expecteddue to the smoothing of the PSF wings, we see a deficiency of stellarmass surface density of roughly 0.5 dex in the central regions of the galaxy and some excesses of 0.2 dex in the outskirts beyond 5R e compared to the stacked galaxy profile. When the total stellarmass is computed by integrating the stellar mass density profiles,the differences in the total stellar mass are less than 1 per cent. Wecome to the conclusion that although PSF effects may contributeminorly to the stellar mass density seen in the outskirts of galaxiesusing our techniques from Section 3.2, they are not the main driver,meaning that the bumps seen in the stellar mass density profiles aredriven by the tidal features we specifically detect.Using the median redshift and effective radius of our sample,we can calculate an approximate median physical effective radiusof our stacked profile of 5.9 kpc. We can therefore approximatelycompare our stellar mass surface density profile of the PSF correctedstacked (golden) sample to that of van Dokkum et al. (2010). Wechoose physical radii of 10, 20 and 30 kpc: van Dokkum et al. (2010)find stellar mass densities of ∼ , 10 . and ∼ M (cid:12) kpc − at 10,20 and 30 kpc respectively, that compares to our results of ∼ . ,10 . and 10 . M (cid:12) kpc − at 10, 20 and 30 kpc respectively. Wetherefore see a very similar behaviour in our profiles compared tovan Dokkum et al. (2010), however a systematic offset of ∼ In order to further investigate the nature of the bumps in the passivegalaxy stellar mass density radial profiles, we split the sample by themorphological classifications as described in Section 3.1, namelythe Features, the Diffuse and the Featureless sub-samples outlinedin Section 3.1. We then plot the same radial profiles as in the twoprevious sections.We initially see in Figure 7 that all sub-samples have extremelysimilar g - r colour profiles inside ∼ e , decreasing from ∼ ∼ e . The Featureless galaxies(solid red line for the median, 16 th and 84 th percentiles are shaded) MNRAS , 1–14 (2020) Thomas M. Jackson et al.
R / R e g - r ( m a g ) R / R e L o g M * ( M k p c ) Passive galaxy profileFeaturelessDiffuseFeatures
Figure 7.
Left hand panel: The g - r colour profiles of those galaxies that display signs of merger activity (Features, median represented by the blue solid lineand 16 th and 84 th percentiles represented by the blue shaded region), those that do not display clear signs of merger activity but a diffuse stellar halo (Diffuse,median represented by the green solid line and 16 th and 84 th percentiles represented by the green shaded region) and those that display no merger activity ordiffuse halo (Featureless, median represented by the red solid line and 16 th and 84 th percentiles represented by the red shaded region). Right hand panel: Thestellar mass density profile for each of the three aforementioned sub-samples. do not extend out beyond this radius. The Diffuse galaxies (solidgreen line for the median, 16 th and 84 th percentiles are shaded)increase in g - r colour by 0.1 mag to ∼ th and84 th percentiles are shaded) dominates the sample number counts athigh radii, and follows a similar profile to the overall passive galaxysample (dotted black line for the median).We see that the stellar mass surface density profiles for allsub-samples decrease in the central regions, until ∼ e . For theFeatureless sub-sample, this trend appears to generally decrease, butcuts off at 4 R e , meaning they do not contribute to the flattening ofthe stellar mass surface density profiles we observe. For the Diffusesub-sample, we see a flattening off of the profile beyond 3 R e andthen a significant decrease, possibly due to a truncation, in thestellar mass surface density at 6 R e . The Features sub-sample (solidblue line for the median, 16 th and 84 th percentiles are shaded) showsvery similar behaviour to the Diffuse sub-sample, initially flatteningoff at 3 R e . However, beyond 6 R e , this sub-sample dominates theexcess material we detect, remaining flat with some bumps. Onvisual inspection, these bins are dominated by galaxies that areundergoing minor mergers or have stellar streams. This shows thatthe stellar material in the outskirts of our passive galaxy sampleis dominated by the tidal features seen in these galaxies that areindicative of recent galaxy interactions and/or mergers. One source of bias in our results may come from the SED fittingwe apply to the imaging data. In order to check that our stellarmass estimates are consistent with previous work, we compare totalstellar mass estimates from our SED fitting using cigale on the HSCimaging data to the stellar mass estimates from the parent sample inJackson et al. (2020), which are derived from Pacifici et al. (2016)(see Section 2 or Pacifici et al. 2012, 2016, for more detail). Wemeasure the fluxes within the same apertures sizes (SDSS Petrosianradius) from our HSC imaging data as those used in Pacifici et al.(2016) for consistency.The results of our comparison of the stellar masses are shown L o g M S t e ll a r , P a c i f i c i ( M ) Passive sampleComparison sample10.0 10.5 11.0 11.5 12.0
Log M Stellar, HSC (M ) L o g M ( M ) Figure 8.
Top panel: A comparison of the stellar mass estimates from thestudy of Pacifici et al. (2016) with the estimates yielded from cigale for thecentral regions of the HSC data. Bottom panel: The difference in log spacebetween the two different methods. in Figure 8. In the top panel, we see a tight correlation between theestimates from the HSC data computed by cigale and the routinefrom Pacifici et al. (2016). The bottom panel of Figure 8 showsthat there are no obvious systematic trends or outliers in both ourpassive galaxy sample and the comparison sample, with maximumdifferences in the estimates of ∼ MNRAS000
Top panel: A comparison of the stellar mass estimates from thestudy of Pacifici et al. (2016) with the estimates yielded from cigale for thecentral regions of the HSC data. Bottom panel: The difference in log spacebetween the two different methods. in Figure 8. In the top panel, we see a tight correlation between theestimates from the HSC data computed by cigale and the routinefrom Pacifici et al. (2016). The bottom panel of Figure 8 showsthat there are no obvious systematic trends or outliers in both ourpassive galaxy sample and the comparison sample, with maximumdifferences in the estimates of ∼ MNRAS000 , 1–14 (2020)
SB galaxy outskirts in HSC The star formation history used in generating models for SEDfitting routines is also a potential source of bias. Recent studiessuch as that of Leja et al. (2020) have shown that using different starformation histories, including non-parametric ones, can cause dif-ferences of 0.1 - 1 dex in the stellar mass estimates. In order to checkthis potential source of bias we also ran our SED fitting with cigaleassuming a delayed star formation history on all Voronoi bins inthe passive galaxy sample, including those in the outskirts of thegalaxies. The estimates of stellar mass yielded had a median differ-ence of 0.1 dex with a maximum difference of 0.3 dex, generallywithin the estimated uncertainties of the stellar mass. This meansthat the radial profiles did not change significantly. This combinedwith the agreement between our estimates for the central regions ofthe galaxy samples and those from Pacifici et al. (2016), which usenon-parametric star formation histories, indicates our SED fittingdoes not significantly bias our results.Redshift biases may also be present in our morphological clas-sification. At larger distances or higher redshifts, the sensitivitylevel of the survey means that we do not probe to as deep in surfacebrightness, meaning that some low surface brightness features maybe missed and a galaxy classified as displaying no merger activitywhen there may be activity below the sensitivity level of the survey.To investigate possible biases we plot the redshift distributions ofeach sub-sample in redshift space and compare them in the bottomright panels of Figure 1.We see that both in the passive galaxy sample and the com-parison sample there is significant overlap in redshift between thethree sub-samples and that any differences in the average redshiftare minimal ( Δ 𝑧 < One way to analyse the nature of the material we observe in theoutskirts of these galaxies is to investigate the colours of variouspopulations of satellite galaxies in comparison to the outskirts them-selves.To do this, we take the SDSS catalogues of Lim et al. (2017)and select only satellite galaxies, defined as those galaxies in a halothat are not designated as the central galaxy of that halo, acrossthe same redshift range as our passive galaxy sample (0.05 < 𝑧 <0.1). We then calculate the g - r colour of each satellite from theSDSS photometry and bin in intervals of 0.5 dex in stellar mass.The distributions can be seen in Figure 9. We see the well knowntrend whereby more massive galaxies tend to be on average redderin their g - r colour.The median g - r colour profile in the inner parts of our passivegalaxy sample decreases from 0.9 to 0.8 mag between 0 and 6 R e (indicated by the black dashed lines in Figure 9), with an averagescatter of ∼ e , weobserve a range of median g - r colours, from ∼ g - r colour range ∼ e ) areindicated by the red dashed line in Figure 9. The comparison sample g - r (mag) N o r m a li s e d C o un t s M < 10 < M < 10 < M < 10 < M < 10 M > 10 Figure 9.
The distribution in g - r colour of satellites taken from SDSScatalogues in bins of 0.5 dex in stellar mass. The dashed black and red linesmark the approximate colour limits of the bulk of stellar material in thecentre (< 2 R e ) and outskirts (> 2 R e ) of our passive galaxy radial profilesrespectively. displays similar behaviour in the nature of the stellar material insideof 6 R e , with median g - r colours ranging between 0.9 and 0.7 mag,however with greater scatter ( ∼ g - r colours between 0.2 and 1.2mag, probably because of stochatisticity due to the small sampleand of the merger activity.When we compare to the distribution of SDSS satellites inFigure 9, we see that those satellite galaxies ranging in stellar massM Stellar < 10 . M (cid:12) have the most similar colours to the lowsurface brightness material. Some of the redder material may belinked to higher mass satellites, however this is not the majority ofthe material beyond 6 R e . If we account for a possible contributionfrom redder, in-situ stars with similar colours to the profiles within 6R e , we would expect the accreted material to have even bluer averagecolours. This would shift potential accreted satellite galaxies to evenbluer colours and, on average, smaller stellar masses. In Section 4.2, we showed tidal features due to accreted materialcan be quantified in colour and stellar mass surface density. Thesetidal features are reflected as a flattening in the outskirts of stellarmass surface density profiles ( (cid:38) e ) when trying to specificallydetect and quantify these features. We see from Sections 4.3 and 4.4that PSF effects, although shifting some of the stellar mass surfacedensity from the centres of the radial profiles to the outskirts, cannotaccount for the levels of stellar mass surface density at radii of (cid:38) e . We have also seen that the stellar mass surface densities aredriven predominantly by galaxies that show features, hence signs ofmerger activity, with some minor contributions by galaxies whichdisplay a diffuse stellar halo. We also showed that the material in theoutskirts of our galaxy samples has similar g - r colours to satellitegalaxies ranging in stellar mass M Stellar < 10 . M (cid:12) .In order to further link the stellar mass excesses to possiblesatellite galaxy accretion, we quantify the total stellar mass of thismaterial in both our passive and comparison galaxy samples, as alarge percentage of the stellar mass contained in the outskirts may MNRAS , 1–14 (2020) Thomas M. Jackson et al.
Table 1.
The total stellar mass in bins of 1 R/R e Radius Passive (Features) Passive (Diffuse) Comparison (Features) Comparison (Diffuse)(R e ) Log M (M (cid:12) ) Percent Log M (M (cid:12) ) Percent Log M (M (cid:12) ) Percent Log M (M (cid:12) ) PercentTotal 11.41 - 11.42 - 11.32 - 11.39 -<1 R e e e e e e e e e indicate different origins of this material. We firstly isolate the sub-samples of galaxies that display features in both the passive andcomparison galaxy samples. We then calculate the average stellarmass per galaxy contained in a bin of 1 R/R e from the estimatesof stellar mass yielded by the SED fitting process. We then workout the percentage stellar mass that each bin in 1 R/R e contributesto the total stellar mass contained on average in one galaxy. Theseresults can be seen in Table 1.We firstly see that for both our passive and comparison galaxysamples, those galaxies that display features contribute more stellarmass in their outskirts. The integrated stellar mass in these regions,however, is relatively small; as outside of 2 R e for the passivegalaxy sample this totals a stellar mass of 10 . M (cid:12) (equating to ∼ . M (cid:12) (equating to ∼ e is10 . M (cid:12) (or ∼ ∼ . − . M (cid:12) and convolv-ing them with the standard definition of a minor merger as 1:4 < 𝜇 < 1:100 (as in Hirschmann et al. 2015), we get an average stellarmass of the accreted satellites of ∼ . − . M (cid:12) . Comparing thisfigure to the integrated stellar mass in the outskirts (outside of 2R e ) of 10 . − . M (cid:12) , we see that a minor merger is capable ofdelivering the amount of stellar mass required to explain the stellarmaterial found in the outskirts.The amount of mass required to be accreted is, however, alsolikely to be smaller than this estimate, as not all stars at these radiiare ex-situ. Davison et al. (2020) find that some in-situ stars exist atthese radii using simulations, whereby the ex-situ fraction of starsin galaxies of stellar mass 10 M (cid:12) - 10 M (cid:12) is ∼
25 per cent at 2R e , up to 40 per cent at 4 R e . This is increased to 65 and 70 percent,respectively, for galaxies of stellar masses of 10 M (cid:12) - 10 M (cid:12) .This significantly lowers the required mass to be delivered by thesatellite galaxies by 30 - 60 percent.Another factor to account for in this type of comparison isthe survival time of various merger features. A recent study byMancillas et al. (2019) used N -body simulations to compare theformation of merger features or tidal activity to observational dataand calculated the average survival time of these types of features.They estimated a survival time of ∼ ∼ ∼ − ,we assume these times to be longer than what we would expect with our imaging, which only goes down to 𝜇 𝑟 − band ∼
28 mag arcsec − .As our passive galaxies are selected to have assembled 90 per centof their stellar mass over 6 Gyr ago, this means that the features areextremely unlikely to be due to major mergers, as a major mergerwith a stellar mass ratio of at least 1 to 4 would deliver more than10 percent of the stellar mass observed today. Any activity thereforecan not be older than this time, fitting the picture of these tidalfeatures caused by minor mergers.To summarise: The average stellar mass per galaxy containedin the outskirts of our sample, mainly in the form of tidal andmerger features is of the order of 10 . − . M (cid:12) . The ex-situ masscontained here is likely to be slightly lower as not all stars at theseradii (> 2 R e ) ) are ex-situ. The colour of this material is similar tothe average SDSS satellite population with stellar masses M Stellar <10 . M (cid:12) . This leads to the plausible scenario that this material wasaccreted from the surrounding satellite population. These results,combined with the widespread merger activity we see ( ∼
40 percent in both samples) quantitatively strengthen the scenario whereminor mergers drive the size growth of central galaxies observed inprevious studies (e.g. van Dokkum et al. 2010; van der Wel et al.2014).We can also compare some of our results to those found inprevious studies of massive galaxies in HSC. Huang et al. (2018b)investigated the stellar mass surface density profiles of massive (M> 10 . M (cid:12) ) galaxies by fitting concentric ring models to HSCimaging data and performing SED fitting. Their profiles are fairlysmooth compared to our own, in line with the expectations outlinedin Section 6. They find stellar mass excesses of up to 0.1 - 0.2 dex(corresponding to masses of up to 10 M (cid:12) or roughly 10 percent,however in more massive galaxies than in our sample) due to lowsurface brightness material in the outskirts of galaxies due to lowsurface brightness material when comparing their profiles to cModelphotometry. Their profiles for individual galaxies can extend up to100 kpc, much further than our sample, however we stress thisis probably due to our 𝑆 / 𝑁 cut of 3 per pixel and our lower massgalaxy sample. In Huang et al. (2018c), they also investigate the lightprofiles of these massive galaxies as a function of environment (viathe halo mass), finding that similar stellar mass galaxies in moremassive haloes have shallower and more extended light profiles.This can be compared to our result halo mass appears to be themain driver of the levels of merger activity we observe, wherebymore massive haloes tend to display merger activity or diffuse stellarhaloes.We finally postulate that although the levels of merger activityare fairly constant across our two samples, as our passive galaxysample is older than our comparison sample, these galaxies have MNRAS000
40 percent in both samples) quantitatively strengthen the scenario whereminor mergers drive the size growth of central galaxies observed inprevious studies (e.g. van Dokkum et al. 2010; van der Wel et al.2014).We can also compare some of our results to those found inprevious studies of massive galaxies in HSC. Huang et al. (2018b)investigated the stellar mass surface density profiles of massive (M> 10 . M (cid:12) ) galaxies by fitting concentric ring models to HSCimaging data and performing SED fitting. Their profiles are fairlysmooth compared to our own, in line with the expectations outlinedin Section 6. They find stellar mass excesses of up to 0.1 - 0.2 dex(corresponding to masses of up to 10 M (cid:12) or roughly 10 percent,however in more massive galaxies than in our sample) due to lowsurface brightness material in the outskirts of galaxies due to lowsurface brightness material when comparing their profiles to cModelphotometry. Their profiles for individual galaxies can extend up to100 kpc, much further than our sample, however we stress thisis probably due to our 𝑆 / 𝑁 cut of 3 per pixel and our lower massgalaxy sample. In Huang et al. (2018c), they also investigate the lightprofiles of these massive galaxies as a function of environment (viathe halo mass), finding that similar stellar mass galaxies in moremassive haloes have shallower and more extended light profiles.This can be compared to our result halo mass appears to be themain driver of the levels of merger activity we observe, wherebymore massive haloes tend to display merger activity or diffuse stellarhaloes.We finally postulate that although the levels of merger activityare fairly constant across our two samples, as our passive galaxysample is older than our comparison sample, these galaxies have MNRAS000 , 1–14 (2020)
SB galaxy outskirts in HSC had more time to allow features to settle into a state of equilibriumand hence form a diffuse stellar halo, accounting for the differencein the percentage of Diffuse to Featureless galaxies classified in thepassive galaxy and comparison samples. In this work we have taken deep imaging data (down to surfacebrightness limits of 𝜇 r − band ∼
28 mag arcsec − ) of 118 low red-shift, massive, central galaxies from the Subaru HSC-SSP widesurvey, selected to have assembled 90 per cent of their stellar massover 6 Gyr ago ( 𝑡 > 𝑆 / 𝑁 > 𝑡 < g and r magnitudes, we constructed radial profiles of g - r colour andstellar mass surface density. We find that the colour profiles arein good agreement with the previous work of La Barbera et al.(2012). We also find expected declining stellar mass surface densityprofiles in the inner regions of our sample, but a flattening of thestellar mass surface density profile in the outskirts beyond ∼ e ( Σ ∗ ∼ . M (cid:12) kpc − ), driven by the low surface brightnessfeatures we observe. We find slightly bluer g - r colour profiles forthe younger comparison sample (0.1 mag difference) and similarbehaviour in the stellar mass density surface profiles, namely adeclining profile in the central regions with excesses in the outskirts(also with Σ ∗ ∼ . M (cid:12) kpc − ).In order to compare to previous studies, (cf. van Dokkum et al.2010), we then stacked and normalised all images. We fitted profilesto these resulting combined galaxy images and repeated the aboveprocess. We also convolved them with a constructed PSF to investi-gate whether the PSF wings could make a significant difference toour profile, thereby accounting for the stellar mass surface densitiesat larger effective radii ( (cid:38) e ). We find the colour profiles yieldedby this process in good agreement with previous studies, finding thatthe stellar mass surface density profiles are smooth and declining atall R/R e , and that accounting for the wings of the PSF by manuallyconstructing a full PSF has a minimal impact on the profiles, so thePSF wings are unlikely to drive the stellar mass surface densitiesobserved at large radii in our original profiles.Using visual morphological classification, we also split oursample into three different categories, finding that those that dis-play tidal features make up 42.4 per cent of our sample, similar toprevious studies (e.g. Duc et al. 2015). Those that have no interac-tion signatures but display a diffuse stellar halo make up 43.2 percent of our sample and those that are featureless make up only 14.4per cent of our sample. When we split the profiles by these threemorphological classes, we find that those classified in the Featuressub-sample are the drivers of the stellar mass surface densities atlarge effective radii ( (cid:38) e ) with some contribution from the Dif-fuse sub-sample. We find that our young comparison sample showssimilar levels of merger activity, however many more galaxies arefeatureless and very few display a diffuse halo. We also find that in-creasing the stellar or halo mass increases the abundance of featuresor diffuse haloes, similar to previous studies (Bílek et al. 2020).We find that the material in these outskirts makes up a minor percentage of the total stellar mass of these systems ( ∼ e ), corresponding to ∼ M (cid:12) . This material hassimilar g - r colours to SDSS satellites of M Stellar < 10 . M (cid:12) ,leading to a plausible scenario that this material is accreted fromthe surrounding satellite population.These results show that there is an abundance of minor mergeractivity around central galaxies and that minor mergers can plausiblybe the driver behind the size growth of massive, central galaxies. ACKNOWLEDGEMENTS
Thomas Jackson is a fellow of the International Max Planck Re-search School for Astronomy and Cosmic Physics at the Universityof Heidelberg (IMPRS-HD).The authors thank Dr. Andy Goulding for his advice and inputon the use of the HSC imaging data and Dr. David Rosario for hisadvice on the SED fitting process. AP acknowledges support fromthe Kavli Institute for the Physics and Mathematics of the Universe(Kavli IPMU).The Hyper Suprime-Cam (HSC) collaboration includes the as-tronomical communities of Japan and Taiwan, and Princeton Uni-versity. The HSC instrumentation and software were developedby the National Astronomical Observatory of Japan (NAOJ), theKavli Institute for the Physics and Mathematics of the Universe(Kavli IPMU), the University of Tokyo, the High Energy Acceler-ator Research Organization (KEK), the Academia Sinica Institutefor Astronomy and Astrophysics in Taiwan (ASIAA), and PrincetonUniversity. Funding was contributed by the FIRST program fromJapanese Cabinet Office, the Ministry of Education, Culture, Sports,Science and Technology (MEXT), the Japan Society for the Pro-motion of Science (JSPS), Japan Science and Technology Agency(JST), the Toray Science Foundation, NAOJ, Kavli IPMU, KEK,ASIAA, and Princeton University.This paper makes use of software developed for the LargeSynoptic Survey Telescope. We thank the LSST Project for makingtheir code available as free software at http://dm.lsst.orgThis paper is based [in part] on data collected at the SubaruTelescope and retrieved from the HSC data archive system, whichis operated by Subaru Telescope and Astronomy Data Center atNational Astronomical Observatory of Japan. Data analysis was inpart carried out with the cooperation of Center for ComputationalAstrophysics, National Astronomical Observatory of Japan.
The data underlying this article are available from the public sourcesin the links or references given in the article (or references therein).Estimations from the article may be available on request.
REFERENCES
Aihara H., et al., 2018, PASJ, 70, S8Aihara H., et al., 2019, PASJ, 71, 114Atkinson A. M., Abraham R. G., Ferguson A. M. N., 2013, ApJ, 765, 28Baldwin J. A., Phillips M. M., Terlevich R., 1981, PASP, 93, 5Banerji M., et al., 2010, MNRAS, 406, 342Bezanson R., van Dokkum P. G., Tal T., Marchesini D., Kriek M., Franx M.,Coppi P., 2009, ApJ, 697, 1290Bílek M., et al., 2020, MNRAS, 498, 2138Blanton M. R., et al., 2005, AJ, 129, 2562MNRAS , 1–14 (2020) Thomas M. Jackson et al.
Boquien M., Burgarella D., Roehlly Y., Buat V., Ciesla L., Corre D., InoueA. K., Salas H., 2019, A&A, 622, A103Bosch J., et al., 2018, PASJ, 70, S5Bruzual G., Charlot S., 2003, MNRAS, 344, 1000Burgarella D., Buat V., Iglesias-Páramo J., 2005, MNRAS, 360, 1413Cappellari M., Copin Y., 2003, MNRAS, 342, 345Cardelli J. A., Clayton G. C., Mathis J. S., 1989, ApJ, 345, 245Chabrier G., 2003, PASP, 115, 763Cimatti A., et al., 2008, A&A, 482, 21Daddi E., et al., 2005, ApJ, 626, 680Davison T. A., Norris M. A., Pfeffer J. L., Davies J. J., Crain R. A., 2020,MNRAS, 497, 81Duc P.-A., 2020, arXiv e-prints, p. arXiv:2007.13874Duc P.-A., et al., 2015, MNRAS, 446, 120Fan L., Lapi A., De Zotti G., Danese L., 2008, ApJ, 689, L101Fan L., Lapi A., Bressan A., Bernardi M., De Zotti G., Danese L., 2010,ApJ, 718, 1460Feldmann R., Carollo C. M., Mayer L., Renzini A., Lake G., Quinn T.,Stinson G. S., Yepes G., 2010, ApJ, 709, 218Ferreras I., Lisker T., Pasquali A., Khochfar S., Kaviraj S., 2009, MNRAS,396, 1573Ferreras I., et al., 2014, MNRAS, 444, 906Hirschmann M., Naab T., Ostriker J. P., Forbes D. A., Duc P.-A., Davé R.,Oser L., Karabal E., 2015, MNRAS, 449, 528Hocking A., Geach J. E., Sun Y., Davey N., 2018, MNRAS, 473, 1108Hopkins P. F., Bundy K., Murray N., Quataert E., Lauer T. R., Ma C.-P.,2009, MNRAS, 398, 898Huang S., et al., 2018a, PASJ, 70, S6Huang S., Leauthaud A., Greene J. E., Bundy K., Lin Y.-T., Tanaka M.,Miyazaki S., Komiyama Y., 2018b, MNRAS, 475, 3348Huang S., et al., 2018c, MNRAS, 480, 521Hubble E. P., 1926, ApJ, 64, 321Jackson T. M., Pasquali A., Pacifici C., Engler C., Pillepich A., Grebel E. K.,2020, MNRAS, 497, 4262Kauffmann G., et al., 2003, MNRAS, 346, 1055Koch A., Black C. S., Rich R. M., Longstaff F. A., Collins M. L. M., JanzJ., 2017, Astronomische Nachrichten, 338, 503Kroupa P., 2001, MNRAS, 322, 231La Barbera F., Ferreras I., de Carvalho R. R., Lopes P. A. A., Pasquali A.,de la Rosa I. G., De Lucia G., 2011, ApJ, 740, L41La Barbera F., Ferreras I., de Carvalho R. R., Bruzual G., Charlot S., PasqualiA., Merlin E., 2012, MNRAS, 426, 2300Leja J., Speagle J. S., Johnson B. D., Conroy C., van Dokkum P., Franx M.,2020, ApJ, 893, 111Lim S. H., Mo H. J., Lu Y., Wang H., Yang X., 2017, MNRAS, 470, 2982Lintott C. J., et al., 2008, MNRAS, 389, 1179Ludwig J., Pasquali A., Grebel E. K., Gallagher John S. I., 2012, AJ, 144,190Mancillas B., Duc P.-A., Combes F., Bournaud F., Emsellem E., Martig M.,Michel-Dansac L., 2019, A&A, 632, A122Martínez-Delgado D., et al., 2010, AJ, 140, 962Mendel J. T., Simard L., Palmer M., Ellison S. L., Patton D. R., 2014, ApJS,210, 3Miyazaki S., et al., 2018, PASJ, 70, S1Morales G., Martínez-Delgado D., Grebel E. K., Cooper A. P., JavanmardiB., Miskolczi A., 2018, A&A, 614, A143Naab T., Johansson P. H., Ostriker J. P., Efstathiou G., 2007, ApJ, 658, 710Naab T., Johansson P. H., Ostriker J. P., 2009, ApJ, 699, L178Newman A. B., Ellis R. S., Bundy K., Treu T., 2012, ApJ, 746, 162Noll S., Burgarella D., Giovannoli E., Buat V., Marcillac D., Muñoz-MateosJ. C., 2009, A&A, 507, 1793Oser L., Ostriker J. P., Naab T., Johansson P. H., Burkert A., 2010, ApJ, 725,2312Oser L., Naab T., Ostriker J. P., Johansson P. H., 2012, ApJ, 744, 63Pacifici C., Charlot S., Blaizot J., Brinchmann J., 2012, MNRAS, 421, 2002Pacifici C., Oh S., Oh K., Lee J., Yi S. K., 2016, ApJ, 824, 45Schlafly E. F., Finkbeiner D. P., 2011, ApJ, 737, 103 Taylor E. N., Franx M., Glazebrook K., Brinchmann J., van der Wel A., vanDokkum P. G., 2010, ApJ, 720, 723Thomas D., Maraston C., Bender R., Mendes de Oliveira C., 2005, ApJ,621, 673Thomas D., Maraston C., Schawinski K., Sarzi M., Silk J., 2010, MNRAS,404, 1775Trujillo I., et al., 2006, MNRAS, 373, L36Trujillo I., Conselice C. J., Bundy K., Cooper M. C., Eisenhardt P., EllisR. S., 2007, MNRAS, 382, 109Trujillo I., Cenarro A. J., de Lorenzo-Cáceres A., Vazdekis A., de la RosaI. G., Cava A., 2009, ApJ, 692, L118Trujillo I., Ferreras I., de La Rosa I. G., 2011, MNRAS, 415, 3903Valentinuzzi T., et al., 2010, ApJ, 712, 226York D. G., et al., 2000, AJ, 120, 1579Zirm A. W., et al., 2007, ApJ, 656, 66de la Rosa I. G., La Barbera F., Ferreras I., Sánchez Almeida J., DallaVecchia C., Martínez-Valpuesta I., Stringer M., 2016, MNRAS, 457,1916van Dokkum P. G., Franx M., 2001, ApJ, 553, 90van Dokkum P. G., et al., 2008, ApJ, 677, L5van Dokkum P. G., et al., 2010, ApJ, 709, 1018van der Wel A., Holden B. P., Zirm A. W., Franx M., Rettura A., IllingworthG. D., Ford H. C., 2008, ApJ, 688, 48van der Wel A., et al., 2014, ApJ, 788, 28This paper has been typeset from a TEX/L A TEX file prepared by the author.MNRAS000