Towards a census of super-compact massive galaxies in the Kilo Degree Survey
The abundance of compact, massive, early-type galaxies (ETGs) provides important constraints to galaxy formation scenarios. Thanks to the area covered, depth, excellent spatial resolution and seeing, the ESO Public optical Kilo Degree Survey (KiDS), carried out with the VLT Survey Telescope (VST), offers a unique opportunity to conduct a complete census of the most compact galaxies in the Universe. This paper presents a first census of such systems from the first 156 square degrees of KiDS. Our analysis relies on g-, r-, and i-band effective radii (), derived by fitting galaxy images with PSF-convolved Sérsic models, high-quality photometric redshifts, , estimated from machine learning techniques, and stellar masses, , calculated from KiDS aperture photometry. After massiveness () and compactness ( in -, - and -bands) criteria are applied, a visual inspection of the candidates plus near-infrared photometry from VIKING-DR1 are used to refine our sample. The final catalog, to be spectroscopically confirmed, consists of 92 systems in the redshift range . This sample, which we expect to increase by a factor of ten over the total survey area, represents the first attempt to select massive super-compact ETGs (MSCGs) in KiDS. We investigate the impact of redshift systematics in the selection, finding that this seems to be a major source of contamination in our sample. A preliminary analysis shows that MSCGs exhibit negative internal colour gradients, consistent with a passive evolution of these systems. We find that the number density of MSCGs is only mildly consistent with predictions from simulations at , while no such system is found at .
keywords:galaxies: evolution – galaxies: general – galaxies: elliptical and lenticular, cD – galaxies: structure.
The understanding of the physical processes which drive the galaxy mass built up and size accretion are among the most timely topics in galaxy evolution studies. Massive early-type galaxies (ETGs) are found to be much more compact in the past than in the present Universe (Daddi et al. 2005; Trujillo et al. 2006; Trujillo et al. 2007; van der Wel et al. 2008). At redshifts , while the massive star-forming disks have effective radii of several kpc (Genzel et al. 2008), the quenched spheroids (“red nuggets”) have small effective radii of about 1 kpc. Such red nuggets are thought to form through a chain of different processes: a) accretion-driven violent disc instability, b) dissipative contraction resulting in the formation of compact, star-forming “blue nuggets”, c) quenching of star formation (Dekel & Burkert 2014). After these processes occur, a gradual expansion in size of the red nuggets may take place, leading to the formation of the massive ETGs we observe in the nearby Universe. Theoretical studies point to dry-mergers as the dominant mechanism for the size and stellar mass growth of very dense massive galaxies (Khochfar & Silk 2006). In particular, minor mergers would provide a modest stellar mass accretion, but a strong evolution in galaxy size (van Dokkum et al. 2010; Hilz et al. 2013; Belli et al. 2014; Tortora et al. 2014). Mergers are believed to be common for very massive systems at high redshifts, with major merger rates (mergers per galaxy per Gyr) in the range at and smaller than at (Hopkins et al. 2010). An alternative scenario explains the size evolution as the result of (e.g.) quasar feedback, rather than merging, making galaxies to puff up after a lost of large amounts of (cold) gas (Fan et al. 2008, 2010).
Over cosmic time, one may expect that high-z compact galaxies evolve into present-day, massive, big galaxies. However, a fraction of these objects might survive intact till the present epoch, resulting in compact, relic systems in the nearby Universe characterized by old stellar populations. Recently there have been some efforts to search for massive compact galaxies at low redshifts (Trujillo et al. 2009; Taylor et al. 2010; Valentinuzzi et al. 2010; Shih & Stockton 2011; Trujillo et al. 2012; Poggianti et al. 2013a; Poggianti et al. 2013b; Damjanov et al. 2013, 2014, 2015a; Saulder et al. 2015) and investigate further their dynamical and stellar population properties, as well as the role of environment on their properties (e.g. Ferré-Mateu et al. 2012; Läsker et al. 2013; Ferré-Mateu et al. 2015; Trujillo et al. 2014; Yldrm et al. 2015; Damjanov et al. 2015b; Stringer et al. 2015; Wellons et al. 2015a).
In some theoretical models, that include the effect of galaxy mergers, the fraction of massive objects that would survive without undergoing any significant transformation since to the present could reach a fraction of about (Hopkins et al. 2009; Quilis & Trujillo 2013). At “low” redshifts (), theoretical models predict a density of relic remnants in the range , which means that, in large surveys like the Sloan Digital Sky Survey (SDSS), we might expect to find a few candidates in this redshift range. However, recent observational works have shown the paucity of old, super-compact ( kpc), massive () galaxies in the local Universe (; Trujillo et al. 2009; Taylor et al. 2010; Ferré-Mateu et al. 2012). Indeed, NGC1277, a nearby lenticular galaxy in the Perseus cluster, is actually the only well-characterized, old system at , that might be a true relic galaxy (Trujillo et al. 2014; see also Martín-Navarro et al. 2015). Other candidates have been recently detected by Saulder et al. (2015), although only a few of them fulfill the above (restrictive) size and mass criteria ( kpc and ), none of them being at . Larger numbers of old compact systems have been found at lower masses (), when relaxing the compactness selection criteria (Valentinuzzi et al. 2010; Poggianti et al. 2013a). The (almost) lack of nearby super-compact, relic systems may represent a challenge for the current paradigm of galaxy formation; in particular, we have to understand whether this lack is due to some observational bias, e.g. the limited spatial resolution of photometric data at ; a failure of theoretical predictions; and/or an environmental effect, for which, as suggested by NGC 1277, relic galaxies might be more frequent in high-density cluster regions.
In the intermediate redshift range (), compacts have been recently investigated in detail by Damjanov et al. (2014), who selected massive compacts from a sample of stellar-like objects within the 6373.2 sq. deg. of the BOSS survey; of these galaxies are dominated by old stellar populations, which make them reliable candidates to be the product of the unperturbed evolution of compact high–z systems. However, of these galaxies do not have measured which hampers the selection of such systems as truly compact objects. More recently, Damjanov et al. (2015a), have analyzed F814W HST images for the COSMOS field, providing robust size measurements for a sample of 1599 compact systems in the range . Other studies have performed detailed analysis of stellar populations and morphology of small samples of compact galaxies at these redshifts (Hsu et al. 2014; Stockton et al. 2014). The population of dense passively evolving galaxies in this intermediate redshift range possibly represents a link between compact systems, dominating the massive quiescent galaxy population at high z, and their descendants in the nearby Universe. Indeed, large samples of compacts, with high-quality photometry (to derive reliable structural parameters), and spectroscopic data, are actually necessary to better understand the formation and evolution of these systems.
The Kilo Degree Survey (KiDS; de Jong et al. 2015) is one of the ESO public surveys carried out with the VLT Survey Telescope (VST; Capaccioli & Schipani 2011) equipped with the one square degree field of view and high angular resolution () OmegaCAM camera (Kuijken et al. 2004; Kuijken 2011). KiDS is mainly designed for weak lensing studies, providing deep imaging in four optical bands (), over a 1500 square degree of the sky with excellent seeing (e.g. median FWHM in -band). The high image quality and deep photometry are ideal to investigate massive compact systems.
According to predictions from simulations (Guo et al. 2011a, 2013), we can expect to find relic per square degree, at redshift . This prediction does critically depend on the physical processes shaping size and mass evolution of galaxies, such as the relative importance of major and minor galaxy merging.
Several compact galaxy definitions have been adopted in the literature (Trujillo et al. 2009; Taylor et al. 2010; Poggianti et al. 2013a; Damjanov et al. 2015a). In the present work, we present the properties of a sample of dense massive galaxy candidates in KiDS, defining as massive super-compact galaxies (MSCGs) those early-type systems with and (Trujillo et al. 2009). These selection criteria are rather conservative, providing the ideal benchmark for galaxy evolution theories. The paper is organized as follows. In Sec. 2 we present the KiDS data sample and the selection of our photometrically selected compact galaxies. The main results, and in particular the evolution of number density as a function of redshift, are presented in Sec. 3. A discussion of the results and future prospects are outlined in Sec. 4. We adopt a cosmological model with , where (Komatsu et al. 2011).
2 Sample selection
The galaxy sample presented in this work is based on the data included in the first and second data releases of KiDS presented in de Jong et al. (2015), which we address the interested reader for details. The total dataset includes 156 KiDS pointings (133 from the KiDS data release 2), in which we have identified about 22 million sources including 7 millions which have been classified as high quality extended sources (mostly galaxy-like, see below). A full description of the galaxy sample is given in Napolitano et al. (2016, in prep.). In the following section, we summarise the main steps for the galaxy selection procedure and the determination of galaxy physical quantities as structural parameters, photometric redshifts and stellar masses.
2.1 KiDS high signal-to-noise galaxy sample
We start from the KiDS multi-band source catalogs, where the photometry has been obtained with S-Extractor (Bertin & Arnouts 1996) in dual image mode, using as reference the positions of the sources detected in the -band images. While magnitudes are measured for all of the filters, the star/galaxy separation, positional and shape parameters are based on the -band data. The choice of -band is motivated by the fact that it typically has the best image quality and thus provides the most reliable source positions and shapes. Critical areas as saturated pixels, star spikes, reflection halos, satellite tracks, etc. have been masked using both a dedicated automatic procedure and visual inspection. The total area after removing the borders of the individual KiDS pointings is 163.9 square degrees, while the unmasked effective area adopted is of 105.4 square degrees.
Star/galaxy separation is based on the distribution of the S-Extractor parameters CLASS_STAR and (signal-to- noise ratio) of a number of sure stars (see La Barbera et al. 2008 and de Jong et al. 2015 for further details). We have further retained those sources which were marked as being out of critical area from our masking procedure. From the original catalog, the star/galaxy separation leaves million galaxies, of which millions have high quality photometry being non-deblended sources located out of the masked area.
To perform accurate surface photometry and determine reliable structural parameters, the highest-quality sources have been further selected (La Barbera et al. 2008, 2010). Thus, we have finally gathered those systems with the highest in the -band images, 1/MAGERR_AUTO_r (La Barbera et al. 2008, 2010; Roy et al. 2016, in preparation). This sample consists of galaxies.
Relevant properties for each galaxy are derived as described here below:
Photometry. As a standard KiDS catalogs’ parameters, we have derived S-Extractor aperture photometry in the four bands () within several radii. For our analysis we have adopted aperture magnitudes MAGAP_2, MAGAP_4 and MAGAP_6, measured within circular apertures of , and of diameter, respectively, while a first probe of the total magnitude is provided by the Kron-like MAG_AUTO.
Structural parameters. Surface photometry is performed using the 2DPHOT environment. 2DPHOT is an automatic software designed to obtain both integrated and surface photometry of galaxies in wide-field images. The software first produces a local PSF model from a series of identified sure stars. This is done, for each galaxy, by fitting the four closest stars to that galaxy with a sum of three two-dimensional Moffat functions. Then galaxy snapshots are fitted with PSF-convolved Sérsic models having elliptical isophotes plus a local background value (see La Barbera et al. 2008 for further details). The fit provides the following parameters for the four wavebands: surface brightness , circularized effective radius, , Sérsic index, , total magnitude, , axis ratio, , and position angle. As it is common use in the literature, in the paper we use the circularized effective radius, , defined as , where is the major-axis effective radius.
Photometric redshifts. Redshifts are determined with the Multi Layer Perceptron with Quasi Newton Algorithm (MLPQNA) method (Brescia et al. 2013, 2014), and fully presented in Cavuoti et al. (2015), which we refer for all details. Both apertures of and of diameter are used. In machine learning supervised methods, a knowledge sample is needed to train the neural network performing the mapping between magnitudes and redshift. The knowledge base consisted on spectroscopic redshift from the Sloan Digital Sky Survey data release 9 (SDSS-DR9; Ahn et al. 2012) and Galaxy And Mass Assembly data release 2 (GAMA-DR2; Driver et al. 2011) which together provide redshifts up to . This knowledge base includes objects, 60% of which are used as training sample, and the remaining ones are used for the blind test set. The redshifts in the blind test sample resemble the spectroscopic redshifts quite well. The scatter in the quantity is . After these experiments, we have finally produced the final catalogue of redshifts for our sample. The cut operated in the fitting magnitudes to resemble the luminosity ranges in the knowledge base will impact the completeness in the faint regime. But our high- sample is not affected by this source of incompleteness.
Stellar masses. We have used the software Le Phare (Arnouts et al. 1999; Ilbert et al. 2006), which performs a simple fitting method between the stellar population synthesis (SPS) theoretical models and data. Single burst models from Bruzual & Charlot (2003) and a Chabrier (2001) IMF are used. Models are redshifted using the photometric redshifts. We adopt the observed magnitudes (and related uncertainties) within a aperture of diameter, which are corrected for Galactic extinction using the map in Schlafly & Finkbeiner (2011). Total magnitudes derived from the Sérsic fitting, , are used to correct the outcomes of Le Phare for missing flux. The single burst assumption is suitable to describe the old stellar populations in the compact galaxies we are interested in (Thomas et al. 2005; Tortora et al. 2009). The estimated photometric ages are used to check if galaxies are compatible with being relic remnants of systems formed at . The degeneracy between age and metallicity in the stellar population analysis can be solved only with forthcoming spectroscopic follow-ups.
"Galaxy classification". Using Le Phare, we have also fitted the observed magnitudes MAGAP_6 with a set of empirical spectral templates used in Ilbert et al. (2006). The set is based on the four basic templates (Ell, Sbc, Scd, Irr) described in Coleman et al. (1980), and star burst models from Kinney et al. (1996). GISSEL synthetic models (Bruzual & Charlot, 2003) are used to linearly extrapolate this set of templates into ultraviolet and near-infrared. The final set of templates ( for ellipticals, for Sbc, for Scd, for Im, and for starburst) is obtained by linearly interpolating the original templates, in order to improve the sampling of the colour space. We have selected the ETGs by choosing the galaxies which are best fitted by one of the elliptical templates (see Napolitano et al. 2016, in preparation).
Near Infrared photometry from VIKING-DR1. As a complementary dataset for our selection of compact galaxies, we have used the data from the first release (DR1) of the ESO VISTA Kilo Degree Infrared Galaxy (VIKING) survey. The VIKING survey is the KiDS twin survey, and provides the near-IR coverage of the same sky region in the 5 wavebands Z, Y, J, H and Ks. VIKING-DR1 consists of 108 observed tiles (Edge et al. 2014). We have found that it overlaps with of our 156 sq. deg. in the KiDS survey. For uniformity, we have used photometry within the same aperture as the optical data: APERMAG_3 for VIKING and MAGAP_2 for KiDS, which correspond to an aperture of of diameter. In particular, we have concentrated on J and Ks passbands, which we do not use for SPS fitting, but only to improve the star-galaxy separation criterion (Maddox et al. 2008; Muzzin et al. 2013) as will be discussed in details in the next section.
The sample of high- galaxies is complete down to a magnitude of , which correspond to stellar masses up to redshift (see see Napolitano et al. 2016, in prep., for further details).
2.2 Selection of compact galaxies
MSCGs have been selected using the following criteria:
Massiveness. The most massive galaxies with are taken (Trujillo et al. 2009), reducing the original sample of galaxies to massive galaxies.
Compactness. We select the densest galaxies by following recent literature (Trujillo et al. 2009). We get galaxies with median circularized radius, , among the -, - and -bands, less than . About 1300 compact candidates remain after this selection.
Best-fitted effective radii. The best fitted structural parameters are considered, taking those systems with a reduced from 2DPHOT smaller than in , and filters (La Barbera et al. 2010). To avoid any accidental wrong fits we have also removed galaxies with unreasonable best-fitted parameters, applying a minimum value for the size ( arcsec), the Sérsic index () and the axial ratio () in all the bands. The minimum value in the Sérsic index has also allowed us to possibly remove misclassified stars, which are expected to be fitted by a box-like profile (mimicked by a Sérsic profile with ). These last criteria on the quality of the structural parameters reduce further the sample to the 106 highly reliable candidates.
Eye-ball check. We have made an eyeball inspection of the images and residuals from Sérsic fitting of the candidates, with the aim of removing problematic objects or possible misclassified stars. To reduce subjectivity, three of the authors have independently checked the images and graded them according to the following scheme: grade-2 are sure galaxies, grade-1 are uncertain galaxies lacking a well-defined elliptical shape, and grade-0 are misclassified objects (either stars or corrupted fits). The mean of the three classifications has been adopted as the final grade. In this way, we have retained those galaxies with a grade larger than , to include only systems graded by at least one of the observers as a sure galaxy. The candidates are further reduced to 96, after removing objects with significant contamination from neighbours.
Optical+NIR star-galaxy separation. We have adopted a morphological criterion to perform the star-galaxy classification (Bertin & Arnouts 1996; La Barbera et al. 2008) and used visual inspection as ultimate check of the galaxy classification. However, based on optical data only a star can be still misclassified as a galaxy on the basis of its morphology, and this issue can be highly dramatic for very compact objects (generally with size comparable or smaller than the seeing). In absence of spectroscopic information, optical+NIR colour-colour diagrams can provide a strong constraint on the galaxy nature of the candidates (e.g. Maddox et al. 2008; Muzzin et al. 2013). In particular, , and magnitudes within of diameter are adopted for this purpose on those fields with coverage by the two surveys. The Vega VISTA magnitudes are converted to AB using the conversion formulae from Le Phare, and . Stars and galaxies with the best J and Ks photometry are also considered (). The results are shown in Fig. 1. Stars have blue colours (i.e., , see light blue shaded region in Fig. 1). We have also found four objects with and, indeed, are erroneously classified as galaxies. We take as compact candidates those systems with and (see light-yellow shaded region in Fig. 1). For those galaxies with available VIKING photometry, this selection allows our set of candidates from the previous classification steps to be refined. Out of 46 galaxies with VIKING photometry, we remain with 42 high-confidence candidates.
In Fig. 1 we have also highlighted the locus populated by point sources with red (), but blue () colours, classified as stars by S-Extractor, which are presumably quasars (Maddox et al. 2008, see pink region in Fig. 1). However, the analysis of this class of objects is beyond the scope of the paper.
To perform an homogeneous comparison of the sample of our compact candidates with a sample of "normal" galaxies, using the same original catalog described in Sec. 2, we have created two control samples (C-Samples) of galaxies with similar stellar masses (), the same lower limits on Sérsic index (), axis ratio () and ( arcsec) as those for the compact systems, but without applying any compactness criterion. The first FULL C-Sample consists of all galaxies, with no restriction on the galaxy type, while the ETGs C-Sample consists of ETGs, classified as described in Sec. 2.
3.1 The final sample
We start by summarizing the sample we are left with for our analysis: after the first three criteria we have selected 106 candidates, which are reduced to 96 after the eye-ball check in (iv). The matching with VIKING included 46 candidates and, after applying the criterion (v), we are left with a sample of 42 out 46 galaxy candidates (i.e. 4/46 objects are likely stars according to their optical-NIR colours, corresponding to a contamination). Updating the total number, we are left with 92 candidates, of which further sources from the region that does not overlap with VIKING might be stars. This corresponds to a density of compact galaxies per square degree, cumulatively for , while we do not find objects at . In contrast, Trujillo et al. (2009) find 29 secure MSCGs at within the SDSS area, using our same criteria. However, only one of these is at very low redshift (), and none of them features old stellar populations (Ferré-Mateu et al. 2012), and thus they cannot actually be classified as relics. The absence of compact systems at in our datasample might be related to environmental issues, since in the current KiDS area we are missing nearby clusters of galaxies, where a larger fraction of compact galaxies is predicted (Stringer et al. 2015), and seems to be observed indeed (Poggianti et al. 2013a; Valentinuzzi et al. 2010). A similar environment effect would be present in the Trujillo et al. (2009) sample. Examples of the 2D fitting results for four candidates are presented in Fig. 2, where postage stamps of the galaxy images and the residuals are shown.
Fig. 3 shows the distributions of redshift, axis ratio, Sérsic index, effective radius, stellar mass and surface mass density for our sample of compact candidates. The median redshift of the sample is , with an rms of , larger than the median redshifts of the FULL and ETGs C-Samples ( and , respectively). Compact candidates have a median r-band effective radius of kpc (rms=0.31 kpc), a Sérsic index of (rms=2.3), an axis ratio of 0.40 (rms=0.14) and median stellar mass of dex (rms=0.09). On the contrary, for the FULL C-Sample the median size, Sérsic index, axis ratio and are kpc, 4.7, 0.73 and dex, respectively. Finally, if the ETGs C-Sample is considered, the median size, Sérsic index, axis ratio and are kpc, 5.5, 0.75 and dex, respectively. Thus, smaller sizes translate into shallower light profiles (smaller ) when compacts are compared with normal galaxies (consistently with the galaxy merging framework in Hilz et al. 2013). Compact candidates also present lower axis ratios than normal galaxies. These elongated shapes are common in massive compact galaxies, both at high-z (van der Wel et al. 2011; Buitrago et al. 2013), and at lower redshifts (Trujillo et al. 2012, 2014).
In our compact sample, the average surface mass density within 1 is (rms of ), about 2 orders of magnitude larger than the average surface mass density of the C-Samples. This median density is a bit larger than the range of values of found in 4 compact galaxies at by Trujillo et al. (2012), but the values are consistent within the scatter. Taking the only 3 galaxies from Szomoru et al. (2012), which satisfy the same criteria we are adopting in this paper, we find a median density of , which is fully consistent with our result.
The -band as a function of is shown in Fig. 4. The compact candidates are plotted with a random subset of galaxies from the FULL C-Sample. As the compact candidates are selected to have small sizes, they lie in a region of the size-mass diagram where very few objects are detected, which provides a further evidence in favor of the rarity of these compact objects at (Fig. 4; Trujillo et al. 2009). In Fig. 4 we also overplot the galaxies in Trujillo et al. (2012) and the three galaxies in Szomoru et al. (2012) which fullfil our selection criteria. The evolution with redshift of the sizes, magnitudes and colours are shown in Fig. 5. The s look quite constant as a function of redshift for the compact systems. At all the redshifts, almost all the compacts have faint MAG_AUTO when compared with the control sample population (middle panel in Fig. 5). We find that 75 candidates ( of the whole sample) lie below the r-band magnitude completeness limit of , implying that our sample is complete up to redshift . Finally, most of the galaxies populate the red-sequence, and are the best candidates to be remnants of high-z red nuggets (right panel in Fig. 5).
58 out of 92 galaxies have old ages consistent with a formation redshift , and so could be the remnants of the compact galaxies observed at (Gargiulo et al. 2011, 2012; Szomoru et al. 2012). 82 out of 92 galaxies (i.e. of the whole sample) are best fitted by an elliptical template, and are classified as ETGs (see Sec. 2). The rest of the galaxies have redshifts and have colours which are best-fitted by Sbc models. However, particularly at these redshifts, one should take this colour-based classification with caution, as a spectroscopic follow-up is actually needed to perform an accurate stellar population analysis.
3.2 Systematics from wrong redshifts
We have finally cross-matched our sample of candidates with SDSS and GAMA spectroscopy catalogs, finding spectra for 9 of the selected candidates.
One of the main systematics in our selection of compact galaxies is induced by wrong redshift determinations, which can affect both the (linear) effective radii and stellar masses, moving the compact out of our selection criteria (e.g., see Figs. 4 and 5). 7 out 9 of these systems are included in the SDSS+GAMA datasample used for photometric redshift determinations (see Sec. 2): 5 in the training sample and 2 in the test sample. Although the photometric redshifts are shown to approximate quite well the spectroscopic ones (see Sec. 2 and Cavuoti et al. 2015), we note that almost all of the photometric redshifts are underestimated with respect to the spectroscopic value. The median difference is , which increases to if the galaxies not in the training sample are considered. This systematics can be probably related to the under-sampling of this galaxy population, just 5 galaxies in the training sample, which can fail the optimum training of the network, and the quality of photometric redshift outcomes is degraded.
As first test, we have re-computed the sizes and stellar masses of these 9 objects using the . Only 4 galaxies survive to the size and mass selection criteria, i.e. (1 not in the training sample).
Despite the small sample available for this test, we cannot exclude the presence of a systematics of the photometric redshift determination for compact sources which we might quantify of the order of . Qualitatively, if the spectroscopic redshift is larger than photometric one, then in physical scale gets larger. The effect on is not as simple as the one on . As an exercise, we have investigated the impact of a redshift error of on the 83 galaxies in the sample without spectroscopic redshift. Using re-calculated values for and stellar mass, we find that 26 of the sources (31%) survive the cuts.
Therefore, a spectroscopic follow-up will be needed to finally confirm the nature of these galaxies, allowing us to classify these candidates as compact or very-compact galaxies. Increasing the knowledge base and the population of compact systems with measured spectroscopic redshifts will improve the quality of the photometric redshift estimates.
3.3 Colour gradients
A big improvement of our analysis with respect to previous works on compact galaxies at similar redshift ranges (e.g. Damjanov et al. 2014) is the high- photometry which allows us to derive robust structural parameters and obtain the first determination of colour profiles for these systems. Therefore, following Tortora et al. (2010), we define the colour gradient as the angular coefficient of the relation vs. , , measured in (omitted in the following unless needed for clarity). The fit of each color profile is performed in the range , where the effective radius is measured in the -band. By definition, a positive CG, , means that a galaxy is redder as increases, while it is bluer outward for a negative gradient. PSF-convolved Sérsic profiles in -, - and -band are used and, in particular, , and colours are discussed. We omit detailed analysis in terms of redshift and stellar mass, but we pinpoint what is the range of colour gradients in our galaxy sample, comparing these results with the ETGs C-Sample and some previous literature.
On average, compact population has (rms 0.52), (rms 0.59) and (rms 0.73) which are substantially consistent with the gradients for the control population of ETGs which are (rms 0.33), (rms 0.23) and (rms 0.40). Hence, compact candidates look quite similar to normal ETGs, except for the scatter, which is partially or totally related to the larger uncertainties on the structural parameters in our small objects.
These results agree with previously reported ranges of color gradients of passively evolving massive galaxies at low- or intermediate-z (Tamura et al. 2000; Tamura & Ohta 2000; Wu et al. 2005; Tortora et al. 2010; La Barbera et al. 2012) or at high-z (Guo et al. 2011b; Gargiulo et al. 2011, 2012) and with simulations (e.g., Tortora et al. 2013).
Finally, we find that 30 compact candidates have all negative colour gradients, 9 have positive gradients, while the rest of the sample have at least one of the three gradients with a different sign with respect to the others. This wide range of behaviours demonstrate that these objects can be formed in quite different initial conditions (Gargiulo et al. 2012; Damjanov et al. 2014).
3.4 Abundance vs. redshift
The number density of compact massive galaxies as a function of redshift is an important constraint on models of galaxy assembly. In recent years there have been different efforts to produce a census of such systems in different redshift bins (e.g. Trujillo et al. 2009; Taylor et al. 2010; Cassata et al. 2013; Poggianti et al. 2013b, a; Damjanov et al. 2014, 2015a, Saulder et al. 2015). Sample size is a crucial aspect to increase the constraining power. If the compact galaxies found in the present work are a representative subsample of the whole population of compacts over the whole area of 1500 sq. deg. that will be mapped by the KiDS survey, we expect to increase the present sample by a factor of ten in the next few years.
For what concerns our current sample, we have binned galaxies with respect to redshift and normalized to the comoving volume corresponding to the observed KiDS sky area111This is obtained by multiplying the number of candidates by , where ( sq. deg.) is the full sky area and is the effective KiDS-DR2 area. Then, the density is derived by dividing for the comoving volume corresponding to each redshift bin.. The errors on number counts take into account fluctuations due to Poisson noise, as well as those due to large-scale structure (i.e. the cosmic variance): they are calculated with the online CosmicVarianceCalculator222http://casa.colorado.edu/trenti/CosmicVariance.html tool (Trenti & Stiavelli 2008). The cosmic variance increases the Poissonian error budget by . The total relative error on abundances (i.e. number densities) amounts to .
In Fig. 6 we plot the redshift evolution of the abundance of compact galaxies (left panel) and that for the subsample of systems with old photometric ages (i.e. formation redshift ; right panel). We consider these potentially old MSCGs, as candidate remnants of the compact ETGs found at high redshift by several studies (see Sec. 1). We also re-determine the abundances by accounting for possible systematics in the photometric redshifts (see Sec. 3.2; see grey symbols in Fig. 6). We remind the reader that our sample starts to be incomplete at (see vertical dashed lines).
In the left panel of Fig. 6, we plot number densities for the whole sample of KiDS compacts, independent of the galaxy formation redshift. Our number densities are compared with estimates from Damjanov et al. (2014), for the number density of stellar-like objects having spectroscopic redshifts from BOSS-DR10 (Ahn et al. 2014) and with three different cuts in total dynamical mass, (, , and , respectively; see Fig. 6). Notice that objects in the Damjanov et al. (2014) sample are not classified according to either morphology or galaxy age, nor do they have an accurate estimate of the intrinsic . Hence, their selection may be missing compact systems that are actually spatially resolved in BOSS-DR10. On the other hand, since Damjanov et al. (2014) also include compacts with and are selected with respect to , rather than , it is not surprising, perhaps, that those abundance estimates are larger than ours. In fact, if we consider the abundance estimate of massive BOSS targets with kpc and from Damjanov et al. (2014) (see purple triangle in the left panel of Fig. 6), the number density of compacts in the lowest redshift bin () is , in much better agreement with our density estimate of for the same redshift bin. On the other hand, selecting galaxies with kpc in KiDS, would lead to abundances dex larger than for kpc, still in good agreement with estimates for stellar-like objects from Damjanov et al. (2014). The yellow region in the left panel of Fig. 6 plots number densities for galaxies in the COSMOS survey (Damjanov et al., 2015a)333 These data are kindly computed for us by I. Damjanov (private comunication) by applying the same size and mass selection criteria as in the present work.. Remarkably, no evolution with redshift is found, for both KiDS and COSMOS samples, although, surprisingly, abundances for COSMOS (on average ) are about one order of magnitude larger than our KiDS estimates. Since Damjanov et al. (2014) claim to find consistent density estimates between COSMOS and BOSS (the latter having an area 4000 times larger than COSMOS), cosmic variance seems not to be responsible for the above discrepancy. However, we cannot exclude that galaxy environment, which might be the actual driver of the number density of compact relics at (see, e.g. Trujillo et al. 2014; Poggianti et al. 2013a; Valentinuzzi et al. 2010), may be different for galaxies in the KiDS-DR2 and COSMOS areas – an issue that will be addressed in forthcoming extensions of the present work.
The results for photometrically old MSCGs (right panel of Fig. 6) are first compared with estimates for compact galaxies from Damjanov et al. (2014), who selected samples with two cuts in ( and , respectively; see Fig. 6). As for the left panel, there are differences in the selection criteria that make a direct comparison not trivial. However, the trends for both BOSS and KiDS samples look qualitatively similar, with a very mild evolution with redshift in the range where our KiDS sample is complete. Moreover, as noticed above, for BOSS compacts with , the Damjanov et al. (2014) abundance estimates are fairly consistent with ours, within the uncertainties.
Finally, we compare MSCGs number densities with predictions from semi-analytical models 444We caution the reader that stellar masses and sizes are measured in a different way between simulations and observations, hampering a straightforward comparison of the two. (SAMs). Quilis & Trujillo (2013) have determined the evolution of the abundance of compact galaxies from SAMs based on Millennium N-body simulations (Guo et al. 2011a, 2013), where relic compacts are defined as galaxies which have barely increased their stellar mass between and . Operatively, they selected from the merger tree those objects that have increased their mass since by less than 10% and 30% , respectively, i.e. galaxies whose mass at is larger than 90% and 70% of the mass limit applied to select compacts. Our results are consistent in the lowest redshift bin with Guo et al. (2013) for simulated galaxies which have increased their mass at most by , while we estimate lower densities than in simulations at higher redshift. However, similarly to what discussed for the comparison to BOSS estimates, theoretical predictions should be actually considered as upper limits, as Quilis & Trujillo (2013) did not apply any precise selection in size, since the resolution in the simulations does not allow reliable estimates of galaxy effective radii to be obtained. We also compare our findings to results from the hydrodynamical Illustris simulation of Wellons et al. (2015a) (see also Wellons et al. 2015b). Wellons et al. (2015a) select 35 massive and compact galaxies at and follow their evolution to . Only 1 out of these 35 systems evolves into a galaxy that satisfies our mass and size criteria at . This corresponds to a number density of , which is consistent with the abundances of compact galaxies that have accreted less than 10% of their final mass from the Millenium simulations. As a further caveat here, we point out that in our selection we adopt the same mass cutoff value at all redshifts, while simulations perform the mass selection at , implying that at high redshifts, simulated galaxies with masses smaller than the cutoff value are actually included in the analysis. This is another reason why one may expect that predicted number densities for compacts at high redshift are actually higher than the observed ones.
At redshifts , in the left panel of Fig. 6, we see a lack of candidates. This seems to contrast the results of Trujillo et al. (2009) who found 29 secure MSCGs at fulfilling our same criteria, all of them having young ages (see also Ferré-Mateu et al. 2012). However, one should notice that out of the 29 MSCGs of Trujillo et al. (2009), only one is at redshift , still pointing to the extreme paucity of such systems in the nearby Universe, consistent with our result. Similarly, Taylor et al. (2010) found one possible old MSCG at low redshift, using a more relaxed criterion for the size, than the one we adopt here.
Saulder et al. (2015) have found a sample of 76 compact galaxies from SDSS at , which resemble quiescent galaxies at high-z, i.e. systems with small effective radii and large velocity dispersions. If we consider their compacts with kpc and , 1 galaxy at and 6 galaxies at are left when the effective radius from a de Vaucouleurs profile is used. Instead, these numbers change to 1 galaxy at and only 1 at if a Sérsic profile is fitted. These numbers correspond to abundances of in the redshift range , and and at , if de Vaucouleurs or Sérsic profile are fitted, respectively. As mentioned in Sec. 1, these findings seem to trouble the current hierarchical paradigm of galaxy formation, where some relic systems at are actually expected to be found. In contrast, Poggianti et al. (2013a) have found 4 galaxies fulfilling our same criteria (corresponding to of their sample galaxies with masses larger than ), and all of these galaxies are old, with mass-weighted ages older than . These numbers translate into a very large abundance of . Recently, Trujillo et al. (2014) added a new brick to the story, finding one relic compact in the Perseus cluster, i.e. NGC 1277, reconciling the observations at with predictions from simulations. All these results point to an overabundance of MSCGs in dense cluster regions, that are actually under-represented over the area currently mapped by KiDS.
As discussed in Sec. 3.2, the sample of MSCGs would be reduced significantly in size by systematics in the redshift determination. Fig. 6 “corrects” the abundance estimates for the possible systematics in redshift. The corrected abundances (see gray boxes and lines in Fig. 6) should be seen as our current lower limit on number densities of SMCGs in KiDS. This issue will be addressed by on-going spectroscopic follow-up of MSCGs, and discussed in details in a future paper.
Thanks to the large area covered, high image quality, excellent spatial resolution and seeing, the Kilo Degree Survey (KiDS) provides a unique dataset to study the properties of super-compact massive galaxies (MSCGs) – a family of systems which plays a key role into our understanding of galaxy formation and evolution.
In this paper, we present a sample of candidate MSCGs , based on 156 sq. deg. of KiDS, in four optical bands (, , and ). We start from a sample of million galaxies with high-, measured photometry and structural parameters in all the four bands. For a subsample of these galaxies, we have used the KiDS photometry to estimate: 1) photometric redshifts based on machine learning techniques (Brescia et al. 2014; Cavuoti et al. 2015); 2) structural parameters using the software 2DPHOT (La Barbera et al. 2008); 3) stellar masses, fitting colours with SPS model predictions. The resulting sample is complete down to an -band magnitude , and down to a stellar mass of , up to a redshift . We select the most massive () and most compact () galaxies with (photometric) redshift . We remove star contaminants by performing a visual inspection of the final sample of candidates and then, for galaxies with available near-IR photometry from VIKING-DR1, we combine optical+NIR photometry to reduce the fraction of contaminants.
The final sample consists of 92 compact candidates, with a number density of compact galaxies per square degree, at . Nine candidates have spectroscopic information from SDSS and GAMA surveys, that is used to assess the systematics in the redshift determination of our sample. On average, compact galaxies have negative colour gradients which are similar to the ones for normal passively evolving galaxies (e.g., Tamura et al. 2000; Tamura & Ohta 2000; Tortora et al. 2010; Tortora et al. 2013; La Barbera et al. 2012). However, the variety of gradients, to be confirmed with a spectroscopic follow-up of the present sample, seems to suggest that compact galaxies formed under a wide range of initial conditions (Gargiulo et al. 2012). We also discuss the evolution with redshift of the number density of compact systems, and in particular that for the oldest galaxies, which are possibly remnants of the high-redshift () compact population detected by many studies. Remarkably, we do not find any MSCGs candidate at . This finding, which is consistent with the recent studies, might be related to the effect of galaxy environment on the abundance of compact systems (see Sec. 1). Although observational studies at intermediate- and high-z do not point to a clear picture of how environment affects galaxy sizes (see Damjanov et al. 2015b and references therein), recent cosmological simulations predict a larger fraction of massive compact systems in high- than in low-density regions (Stringer et al. 2015). This prediction is supported by the fact that NGC 1277 – the only well-characterized compact, massive, ETG at – resides in the core region of a nearby rich cluster of galaxies (Trujillo et al. 2014; see also Valentinuzzi et al. 2010; Poggianti et al. 2013a). Thus, the absence of compact galaxies at could be related to the smaller fraction of dense structures in the area currently mapped by KiDS.
We plan to have 10 times more compact candidates at the end of the KiDS survey, when all the 1500 sq. deg. will have been observed. Only nine of our candidates have a spectroscopic coverage at the moment. Thus, a spectroscopic follow-up is actually necessary, to fully validate and characterize our sample, and determine accurate number densities to be compared with theoretical expectations.
Based on data products from observations made with ESO Telescopes at the La Silla Paranal Observatory under programme IDs 177.A-3016, 177.A-3017 and 177.A-3018, and on data products produced by Target/OmegaCEN, INAF-OACN, INAF-OAPD and the KiDS production team, on behalf of the KiDS consortium. OmegaCEN and the KiDS production team acknowledge support by NOVA and NWO-M grants. Members of INAF-OAPD and INAF-OACN also acknowledge the support from the Department of Physics & Astronomy of the University of Padova, and of the Department of Physics of Univ. Federico II (Naples). CT has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement n. 267251 “Astronomy Fellowships in Italy” (AstroFIt). This work was partially funded by the MIUR PRIN Cosmology with Euclid. MB acknowledges the PRIN-INAF 2014 Glittering kaleidoscopes in the sky: the multifaceted nature and role of Galaxy Clusters. We thank the referee for his/her comments, which have helped to improve the paper. We thank I. Damjanov for having provided us with the abundances of compacts for the COSMOS field, applying the same size and mass selections as in the present work. We thank M. Cacciato, R. C. E. Van den Bosch, M. Viola and I. Trujillo for helpful comments, B. Poggianti for comments about their findings at low redshift, V. Quilis who provided us with simulation predictions. We also thank S. Wellons and A. Pillepich for comments about their Illustris simulations.
- Ahn et al. (2014) Ahn C. P. et al., 2014, ApJS, 211, 17
- Ahn et al. (2012) Ahn C. P. et al., 2012, ApJS, 203, 21
- Arnouts et al. (1999) Arnouts S., Cristiani S., Moscardini L., Matarrese S., Lucchin F., Fontana A., Giallongo E., 1999, MNRAS, 310, 540
- Belli et al. (2014) Belli S., Newman A. B., Ellis R. S., 2014, ApJ, 783, 117
- Bertin & Arnouts (1996) Bertin E., Arnouts S., 1996, A&AS, 117, 393
- Brescia et al. (2013) Brescia M., Cavuoti S., D’Abrusco R., Longo G., Mercurio A., 2013, ApJ, 772, 140
- Brescia et al. (2014) Brescia M., Cavuoti S., Longo G., De Stefano V., 2014, A&A, 568, A126
- Bruzual & Charlot (2003) Bruzual G., Charlot S., 2003, MNRAS, 344, 1000
- Buitrago et al. (2013) Buitrago F., Trujillo I., Conselice C. J., Häußler B., 2013, MNRAS, 428, 1460
- Capaccioli & Schipani (2011) Capaccioli M., Schipani P., 2011, The Messenger, 146, 2
- Cassata et al. (2013) Cassata P. et al., 2013, ApJ, 775, 106
- Cavuoti et al. (2015) Cavuoti S. et al., 2015, MNRAS, 452, 3100
- Chabrier (2001) Chabrier G., 2001, ApJ, 554, 1274
- Coleman et al. (1980) Coleman G. D., Wu C.-C., Weedman D. W., 1980, ApJS, 43, 393
- Daddi et al. (2005) Daddi E. et al., 2005, ApJ, 626, 680
- Damjanov et al. (2013) Damjanov I., Chilingarian I., Hwang H. S., Geller M. J., 2013, ApJ, 775, L48
- Damjanov et al. (2015a) Damjanov I., Geller M. J., Zahid H. J., Hwang H. S., 2015a, ApJ, 806, 158
- Damjanov et al. (2014) Damjanov I., Hwang H. S., Geller M. J., Chilingarian I., 2014, ApJ, 793, 39
- Damjanov et al. (2015b) Damjanov I., Zahid H. J., Geller M. J., Hwang H. S., 2015b, ArXiv e-prints
- de Jong et al. (2015) de Jong J. T. A. et al., 2015, A&A, 582, A62
- Dekel & Burkert (2014) Dekel A., Burkert A., 2014, MNRAS, 438, 1870
- Driver et al. (2011) Driver S. P. et al., 2011, MNRAS, 413, 971
- Edge et al. (2014) Edge A., Sutherland W., The Viking Team, 2014, VizieR Online Data Catalog, 2329, 0
- Fan et al. (2010) Fan L., Lapi A., Bressan A., Bernardi M., De Zotti G., Danese L., 2010, ApJ, 718, 1460
- Fan et al. (2008) Fan L., Lapi A., De Zotti G., Danese L., 2008, ApJ, 689, L101
- Ferré-Mateu et al. (2015) Ferré-Mateu A., Mezcua M., Trujillo I., Balcells M., van den Bosch R. C. E., 2015, ApJ, 808, 79
- Ferré-Mateu et al. (2012) Ferré-Mateu A., Vazdekis A., Trujillo I., Sánchez-Blázquez P., Ricciardelli E., de la Rosa I. G., 2012, MNRAS, 423, 632
- Gargiulo et al. (2011) Gargiulo A., Saracco P., Longhetti M., 2011, MNRAS, 412, 1804
- Gargiulo et al. (2012) Gargiulo A., Saracco P., Longhetti M., La Barbera F., Tamburri S., 2012, MNRAS, 425, 2698
- Genzel et al. (2008) Genzel R. et al., 2008, ApJ, 687, 59
- Guo et al. (2013) Guo Q., White S., Angulo R. E., Henriques B., Lemson G., Boylan-Kolchin M., Thomas P., Short C., 2013, MNRAS, 428, 1351
- Guo et al. (2011a) Guo Q. et al., 2011a, MNRAS, 413, 101
- Guo et al. (2011b) Guo Y. et al., 2011b, ApJ, 735, 18
- Hilz et al. (2013) Hilz M., Naab T., Ostriker J. P., 2013, MNRAS, 429, 2924
- Hopkins et al. (2010) Hopkins P. F. et al., 2010, ApJ, 724, 915
- Hopkins et al. (2009) Hopkins P. F., Hernquist L., Cox T. J., Keres D., Wuyts S., 2009, ApJ, 691, 1424
- Hsu et al. (2014) Hsu L.-Y., Stockton A., Shih H.-Y., 2014, ApJ, 796, 92
- Ilbert et al. (2006) Ilbert O. et al., 2006, A&A, 457, 841
- Khochfar & Silk (2006) Khochfar S., Silk J., 2006, ApJ, 648, L21
- Kinney et al. (1996) Kinney A. L., Calzetti D., Bohlin R. C., McQuade K., Storchi-Bergmann T., Schmitt H. R., 1996, ApJ, 467, 38
- Komatsu et al. (2011) Komatsu E. et al., 2011, ApJS, 192, 18
- Kuijken (2011) Kuijken K., 2011, The Messenger, 146, 8
- Kuijken et al. (2004) Kuijken K. et al., 2004, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 5492, Ground-based Instrumentation for Astronomy, Moorwood A. F. M., Iye M., eds., pp. 484–493
- La Barbera et al. (2010) La Barbera F., de Carvalho R. R., de La Rosa I. G., Lopes P. A. A., Kohl-Moreira J. L., Capelato H. V., 2010, MNRAS, 408, 1313
- La Barbera et al. (2008) La Barbera F., de Carvalho R. R., Kohl-Moreira J. L., Gal R. R., Soares-Santos M., Capaccioli M., Santos R., Sant’anna N., 2008, PASP, 120, 681
- La Barbera et al. (2012) La Barbera F., Ferreras I., de Carvalho R. R., Bruzual G., Charlot S., Pasquali A., Merlin E., 2012, MNRAS, 426, 2300
- Läsker et al. (2013) Läsker R., van den Bosch R. C. E., van de Ven G., Ferreras I., La Barbera F., Vazdekis A., Falcón-Barroso J., 2013, MNRAS, 434, L31
- Maddox et al. (2008) Maddox N., Hewett P. C., Warren S. J., Croom S. M., 2008, MNRAS, 386, 1605
- Martín-Navarro et al. (2015) Martín-Navarro I., La Barbera F., Vazdekis A., Ferré-Mateu A., Trujillo I., Beasley M. A., 2015, MNRAS, 451, 1081
- Muzzin et al. (2013) Muzzin A. et al., 2013, ApJS, 206, 8
- Poggianti et al. (2013a) Poggianti B. M. et al., 2013a, ApJ, 762, 77
- Poggianti et al. (2013b) Poggianti B. M., Moretti A., Calvi R., D’Onofrio M., Valentinuzzi T., Fritz J., Renzini A., 2013b, ApJ, 777, 125
- Quilis & Trujillo (2013) Quilis V., Trujillo I., 2013, ApJ, 773, L8
- Saulder et al. (2015) Saulder C., van den Bosch R. C. E., Mieske S., 2015, A&A, 578, A134
- Schlafly & Finkbeiner (2011) Schlafly E. F., Finkbeiner D. P., 2011, ApJ, 737, 103
- Shih & Stockton (2011) Shih H.-Y., Stockton A., 2011, ApJ, 733, 45
- Stockton et al. (2014) Stockton A., Shih H.-Y., Larson K., Mann A. W., 2014, ApJ, 780, 134
- Stringer et al. (2015) Stringer M., Trujillo I., Dalla Vecchia C., Martinez-Valpuesta I., 2015, MNRAS, 449, 2396
- Szomoru et al. (2012) Szomoru D., Franx M., van Dokkum P. G., 2012, ApJ, 749, 121
- Tamura et al. (2000) Tamura N., Kobayashi C., Arimoto N., Kodama T., Ohta K., 2000, AJ, 119, 2134
- Tamura & Ohta (2000) Tamura N., Ohta K., 2000, AJ, 120, 533
- Taylor et al. (2010) Taylor E. N., Franx M., Glazebrook K., Brinchmann J., van der Wel A., van Dokkum P. G., 2010, ApJ, 720, 723
- Thomas et al. (2005) Thomas D., Maraston C., Bender R., Mendes de Oliveira C., 2005, ApJ, 621, 673
- Tortora et al. (2010) Tortora C., Napolitano N. R., Cardone V. F., Capaccioli M., Jetzer P., Molinaro R., 2010, MNRAS, 407, 144
- Tortora et al. (2009) Tortora C., Napolitano N. R., Romanowsky A. J., Capaccioli M., Covone G., 2009, MNRAS, 396, 1132
- Tortora et al. (2014) Tortora C., Napolitano N. R., Saglia R. P., Romanowsky A. J., Covone G., Capaccioli M., 2014, MNRAS, 445, 162
- Tortora et al. (2013) Tortora C., Pipino A., D’Ercole A., Napolitano N. R., Matteucci F., 2013, MNRAS, 435, 786
- Trenti & Stiavelli (2008) Trenti M., Stiavelli M., 2008, ApJ, 676, 767
- Trujillo et al. (2012) Trujillo I., Carrasco E. R., Ferré-Mateu A., 2012, ApJ, 751, 45
- Trujillo et al. (2009) Trujillo I., Cenarro A. J., de Lorenzo-Cáceres A., Vazdekis A., de la Rosa I. G., Cava A., 2009, ApJ, 692, L118
- Trujillo et al. (2007) Trujillo I., Conselice C. J., Bundy K., Cooper M. C., Eisenhardt P., Ellis R. S., 2007, MNRAS, 382, 109
- Trujillo et al. (2014) Trujillo I., Ferré-Mateu A., Balcells M., Vazdekis A., Sánchez-Blázquez P., 2014, ApJ, 780, L20
- Trujillo et al. (2006) Trujillo I. et al., 2006, ApJ, 650, 18
- Valentinuzzi et al. (2010) Valentinuzzi T. et al., 2010, ApJ, 712, 226
- van der Wel et al. (2008) van der Wel A., Holden B. P., Zirm A. W., Franx M., Rettura A., Illingworth G. D., Ford H. C., 2008, ApJ, 688, 48
- van der Wel et al. (2011) van der Wel A. et al., 2011, ApJ, 730, 38
- van Dokkum et al. (2010) van Dokkum P. G. et al., 2010, ApJ, 709, 1018
- Wellons et al. (2015a) Wellons S. et al., 2015a, ArXiv e-prints
- Wellons et al. (2015b) Wellons S. et al., 2015b, MNRAS, 449, 361
- Wu et al. (2005) Wu H., Shao Z., Mo H. J., Xia X., Deng Z., 2005, ApJ, 622, 244
- Yldrm et al. (2015) Yldrm A., van den Bosch R. C. E., van de Ven G., Husemann B., Lyubenova M., Walsh J. L., Gebhardt K., Gültekin K., 2015, MNRAS, 452, 1792