Draft version November 3, 2020 Typeset using LATEX twocolumn style in AASTeX63

Tracing Dark Matter Halos with Satellite Kinematics and the Central Stellar Velocity Dispersion of Galaxies Gangil Seo1, Jubee Sohn2, Myung Gyooon Lee1,∗1, 2

11 Astronomy Program, Department of Physics and Astronomy, Seoul National University, Gwanak-gu, Seoul 151-742, Republic of Korea 22 Smithsonian Astrophysical Observatory, 60 Garden Street, Cambridge, MA 02138, USA

ABSTRACT It has been suggested that the central stellar velocity dispersion of galaxies can trace dark matter halo mass directly. We test this hypothesis using a complete spectroscopic sample of isolated galaxies surrounded by faint satellite galaxies from the Sloan Digital Sky Survey Data Release 12. We apply a friends-of-friends algorithm with projected linking length ∆D < 100 kpc and radial velocity linking length ∆V < 1000 km s−1 to construct our sample. Our sample includes 2807 isolated galaxies with 3417 satellite galaxies at 0.01 < z < 0.14. We divide the sample into two groups based on the primary galaxy color: red and blue primary galaxies separated at (g − r)0 = 0.85. The central stellar velocity dispersions of the primary galaxies are proportional to the luminosities and stellar masses of the same galaxies. Stacking the sample based on the central velocity dispersion of the primary galaxies, we derive the velocity dispersions of their satellite galaxies, which trace the dark matter halo mass of the primary galaxies. The system velocity dispersion of the satellite galaxies shows a remarkably tight correlation with the central velocity dispersion of the primary galaxies for both red and blue samples. In particular, the slope of the relation is identical to 1 for red primary systems. This tight relation suggests that the central stellar velocity dispersion of galaxies is indeed an efficient and robust tracer for dark matter halo mass. We provide empirical relations between the central stellar velocity dispersion and the dark matter halo mass.

Keywords: dark matter — galaxies:evolution—galaxies:formation—galaxies:kinematics and dynamics—galaxies: halos

1. INTRODUCTION for mass estimation. In particular, the kinematics of 1.1. Tracing Dark Matter Halo Mass satellite galaxies among them has been a powerful tool to study the dark matter in the Local Group (Lynden- According to the modern galaxy formation scenario Bell, & Frenk 1981; Watkins et al. 2010; McConnachie based on ΛCDM cosmology, galaxies are formed and em- 2012). However, this method can be applied only to a bedded in dark matter halos, so there must be close re- small number of nearby galaxies, and they must have a lations between galaxy properties and dark matter halo reasonable number of satellite galaxies (Watkins et al. properties (Wechsler, & Tinker 2018; Behroozi et al. 2010; Wechsler, & Tinker 2018). 2019; Burkert & Forbes 2020 and references therein). However, it is not easy to derive directly the extent and arXiv:2010.00693v2 [astro-ph.GA] 31 Oct 2020 total mass of dark matter around galaxies. 1.2. Satellite Kinematics Various methods have been applied to estimate the A number of previous studies used statistical (or amount of dark matter in galaxies and groups/clusters stacking) analysis of satellite kinematics for a large num- of galaxies (see reviews like Courteau et al. 2014; Salucci ber of isolated galaxies in order to investigate the prop- 2019 and references therein). Traditionally, dynamical erty of dark matter halos (Zaritsky et al. 1993, 1997; tracers (stars, globular clusters, planetary nebulae, HII McKay et al. 2002; Prada et al. 2003; Brainerd 2005; regions, HI gas, or satellite galaxies) have been used Conroy et al. 2007; Klypin, & Prada 2009; Dutton et al. 2010; More et al. 2011; Wojtak & Mamon 2013; van Uitert et al. 2016; Lange et al. 2019). However, this ap- ∗ corresponding author : [email protected] proach can only be applied to galaxies with satellites, which are only found in the local universe. 2

Therefore, simple proxies that can be applied to a al.(2020) showed that the central stellar velocity disper- large sample of galaxies are often used to estimate dark sion of the Brightest Cluster Galaxies (BCGs) is tightly matter halo mass indirectly. It is known that the lumi- correlated with the system velocity dispersion of cluster nosity or stellar mass of galaxies shows a correlation with galaxies. This tight relation supports the idea that the the dark matter halo mass derived from their satellite central stellar velocity dispersion of the central galaxy observations, but their relations have been controversial can trace the dark matter halo mass. (Norberg et al. 2008; Wojtak & Mamon 2013; van Uitert et al. 2016). 1.4. Goals of This Study In this study, we present an empirical test of the above 1.3. Central Velocity Dispersion as a Proxy for Dark hypothesis that the central stellar velocity dispersion of Matter Halo Mass a galaxy can be a robust proxy of the dark matter halo velocity dispersion and the dark matter halo mass, using In their pioneering paper, Faber & Jackson(1976) a large sample of isolated galaxies with satellite galaxies. found that the central stellar velocity dispersion ob- We construct a complete sample of isolated galaxies sur- tained from 25 elliptical and S0 galaxies increases with rounded by faint satellite galaxies from the Sloan Digital the galaxy B-band luminosity according to L ∝ σ4, the B 0 Sky Survey (SDSS) Data Release 12 (DR12) spectro- Faber-Jackson relation. Later, Whitmore et al.(1979) scopic data. Based on this complete sample, we explore found that the bulges of spiral galaxies follow a simi- the relation between the satellite velocity dispersion and lar relation. These studies and numerous following ones the central stellar velocity dispersion of primary galax- showed that the kinematic properties of the central re- ies. We also study the dependence of the satellite ve- gion of a galaxy are linked with the properties of the en- locity dispersion on other physical properties of their tire galaxy (like luminosity and stellar mass; see Davis primary galaxies (i.e., luminosity and stellar mass). We et al.(2019) and references therein). then derive the empirical relations between the central Recently, it has been suggested that the central stel- stellar velocity dispersion of the primary galaxy and its lar velocity dispersion of a galaxy can also be a robust dark matter halo mass. proxy of the dark matter halo velocity dispersion and This paper is organized as follows. In Section 2 we the dark matter halo mass (Schechter 2015; Zahid et al. describe how we select samples of isolated galaxies with 2016, 2018; Utsumi et al. 2020). Based on the Illustris-1 satellites from the SDSS data. In 3, we present the hydrodynamical cosmological simulations (Vogelsberger § properties of the primary galaxies and estimate the sys- et al. 2014), Zahid et al.(2018) showed that the stellar tem velocity dispersion of satellite galaxies. Then, we velocity dispersion of quiescent galaxies is proportional investigate the relations between the satellite velocity to the dark matter halo velocity dispersion. The tight dispersion and the properties of the primary galaxies relation (∼0.2 dex scatter) indicates that the dark mat- (central stellar velocity dispersion, stellar mass, lumi- ter halo mass can be estimated well from the central nosity, and dark matter halo mass) in 4. In 5, we stellar velocity dispersion of central (primary) galaxies. § § discuss the implications of the main results, and we sum- Several observational studies also suggest that the cen- marize the main results in 6. Throughout the paper, tral stellar velocity dispersion is a good tracer of dark § we adopt a standard ΛCDM cosmology (Ω = 0.3 and matter halo mass (Wake et al. 2012; Zahid et al. 2016; m Ω = 0.7) with H = 70 km s−1 Mpc−1. Sohn et al. 2020). The central velocity dispersion has Λ 0 several advantages in estimating the dark matter halo 2. DATA AND SAMPLE SELECTION mass over luminosity and stellar mass (e.g., Sohn et al. 2017a). The central velocity dispersion is relatively 2.1. Data insensitive to the systematic uncertainties (Sohn et al. We used a catalog of galaxies at 0.01 ≤ z < 0.14 2017a) that can be introduced by photometry and the based on the enhanced SDSS DR12 spectroscopic cata- underlying physical models to derive photometry and log, which was used in the study of compact groups of stellar mass (e.g., Conroy et al. 2007; Bernardi et al. galaxies by Sohn et al.(2016). A detailed description of 2013). Zahid et al.(2016) demonstrated that the stel- the catalog is given in Sohn et al.(2016, 2017a). This lar mass and central stellar velocity dispersion relation catalog includes spectroscopic redshifts obtained from has a similar slope to that of the relation between the the SDSS and the NASA/IPAC Extragalactic Database. dark matter halo mass and dark matter velocity disper- For r-band absolute magnitudes and (g − r)0 colors of sion. This consistency suggests that the central stellar galaxies, we use composite model (cModel) magnitudes. velocity dispersion is directly proportional to the dark Absolute magnitudes of galaxies are corrected for kz=0.0 matter halo velocity dispersion. More recently, Sohn et correction, evolution, and foreground extinction. We de- 3

500 2.2. Identification of Isolated Galaxies with Satellites Median = 1.26 Field galaxies that host many low-mass satellite galax- Mean = 1.45 400 ies are ideal targets to investigate the relation between the kinematics of satellite galaxies and the dark mat- ter halo. However, such systems are rare except for the 300 massive galaxies in the nearby universe (Watkins et al. 2010; Karachentsev et al. 2013; Karachentsev, & Kaisina 2019). Therefore, most previous studies of satellite kine- 200 matics use a method of stacking satellite systems from a large number of isolated galaxies (Zaritsky et al. 1993,

Number of Primaries 1997; McKay et al. 2002; Prada et al. 2003; Brainerd 100 2005; Conroy et al. 2007; Klypin, & Prada 2009; Dutton et al. 2010; More et al. 2011; Wojtak & Mamon 2013; van Uitert et al. 2016; Zahid et al. 2018; Lange et al. 0 0 1 2 3 4 5 2019). We adopt the same method in this study. The physical size of the SDSS fiber radius (kpc) We construct a complete set of isolated galaxies with Figure 1. A histogram of the linear scale of SDSS fiber faint satellite galaxies from the SDSS spectroscopic sam- radius (1.5 arcsec) used for the central region of each primary ple. We use a friends-of-friends (FoF) algorithm (Huchra galaxy. The dashed and dotted lines represent the mean and & Geller 1982) to find systems hosting a dominant pri- median value, respectively. mary galaxy surrounded by faint satellites. Sohn et al. (2016) already identified similar systems based on SDSS rive the stellar masses of galaxies using the LePHARE while they were identifying compact groups of galaxies. spectral energy distribution fitting code (Arnouts et al. However, they applied a strict definition of faint satel- 1999; Ilbert et al. 2006). lites (∆r > 3 mag) and used a tighter projected linking We obtain the central stellar velocity dispersion of length of ∆D ≤ 50h−1 kpc. In this study, we extend galaxies from the Portsmouth reduction (Thomas et al. the identification of such systems in Sohn et al.(2016) 2013). These velocity dispersion measurements are in based on more generous classification criteria to enlarge good agreement with those from the SDSS pipeline. We the sample size. use only the values of the galaxies that show a signal-to- We use a fixed projected linking length of ∆D = 100 noise ratio larger than σ0/δσ0 = 3. The central veloc- kpc and a radial velocity linking length of |∆V | = 1000 ity dispersion we use is measured through a 1.5 00radius km s−1. The radial linking length of |∆V | = 1000 km SDSS fiber. Because our target galaxies are distributed s−1 is consistent with previous studies that identified in a wide redshift range 0.01 < z < 0.14, the physical galaxy groups (Barton et al. 1996; Sohn et al. 2016). We coverage of the SDSS fiber varies with the redshift of the note that more than a half of satellite dwarf galaxies are target galaxies. Figure1 shows the distribution of the found at R < 100 kpc in the Milky Way galaxy (MWG), physical coverage of the SDSS fiber radius on the sample which has a virial radius of about 300 kpc (McConnachie galaxies. The physical coverage of the SDSS fiber radius 2012; Kashibadze, & Karachentsev 2018; Karachentsev, ranges from 0.3 to 5 kpc with a mean value of 1.5 kpc. & Kaisina 2019). We adopt a projected linking length We apply an aperture correction to compute the cen- smaller than the virial radius to reduce any contamina- tral velocity dispersion measured with a fixed aperture. tion due to interlopers in the samples. We calculate the aperture-corrected velocity dispersion We then identify isolated galaxies with faint satellite within 1.5 kpc corresponding to the mean SDSS fiber galaxies. We define faint satellite galaxies as the galax- size of the sample. The aperture correction we apply is ies that are more than 2 mag fainter than their host −0.054±0.005 σ∗/σSDSS = (R0/RSDSS) , where σ∗ is the (i.e., ∆r > 2 mag). We identify the galaxy systems aperture-corrected velocity dispersion, and R0 is the an- that only contain the faint satellite galaxies. In other gular size of 1.5 kpc at given redshifts of galaxies. This words, if an FoF system contains any satellite galaxies aperture correction is similar to that in previous studies with ∆r < 2 mag compared to their host galaxy, we re- (e.g., Cappellari et al. 2006; Zahid et al. 2016). The typ- move this FoF system from our sample. As a result, our ical amount of the aperture correction is only ∼ 2.5%. sample includes the FoF systems consisting of primary Hereafter, we refer to this aperture-corrected velocity galaxies surrounded by only satellites that are at least dispersion σ∗ as the central stellar velocity dispersion. 2 mag fainter than their primaries. Because the limit- ing r−band magnitude of the main SDSS spectroscopic 4

1500 cated at ∆D < 100 kpc, and a small number of them are located out to ∆D ≈ 260 kpc. We plot the histogram 1000 of ∆V for all satellites in the right panel of the figure. Most satellites are concentrated around ∆V = 0, which ] 1 500 indicates that a majority are gravitationally bound to s their primary galaxies.

m 0

k The sharp edge at ∆D ∼ 100 kpc is due to the linking [

v 500 length we adopted. There are a small number of satellite galaxies at ∆D > 100 kpc. These satellites are the

1000 members of the systems with N ≥ 3. These satellites are friends of satellites within the linking length from 1500 the primary galaxies. 0 50 100 150 200 250 250 500 750 The ∆V distribution shows a trumpet-like shape be- D [kpc] N(sat) cause the wing components are higher than the normal Figure 2. (Left) The relative rest-frame radial velocity dif- Gaussian distribution. In Figure3, we display satel- ferences between primary galaxies and their satellite galax- lite velocity distributions as a function of central stellar ies as a function of projected galactocentric distance to the velocity dispersion of primary galaxies (σ∗,prim). The satellite galaxies. (Right) The distribution of radial velocity velocity spread of satellite galaxies increases as σ∗,prim differences. increases. Thus, the trumpet-like shape in Figure2 is mainly due to the superposition of narrow velocity dis- sample is 17.77 mag, we identify primary galaxies with tributions of a large number of low-mass systems and r < 15.77 mag. broad velocity distributions of a smaller number of high- We identify 26,430 FoF systems from the SDSS spec- mass systems, and due to the possibility that the higher troscopic data: 22,288 pairs (N = 2) and 4142 groups velocity dispersion systems (high-mass systems) have (N ≥ 3). Among these systems, there are only 3390 proportionally more satellites at larger radii. (∼ 13%) systems with 4218 faint satellites that satisfy There are a number of satellite galaxies with ∆V > the satellite magnitude selection (∆r > 2 mag): 2842 500 km s−1. Those in low-mass systems are mostly pairs (N = 2) and 548 groups (N ≥ 3). The satellite 3σ outliers, while those in high-mass systems are magnitude selection significantly removes bright pairs not. Those in high-mass systems with σ∗,prim > 200 and groups that are widely used for studying galaxy km s−1 are mostly considered to be satellite members properties in galaxy pairs (e.g., Ellison et al. 2010) or of groups. groups (e.g., Tempel et al. 2014). We additionally check the contamination of central galaxies in dense cluster 2.3. Classification of Primary Galaxies environments. We compile the three largest catalogs of the BCGs (Lauer et al. 2014; Kluge et al. 2020; Sohn et Previous studies that explore the stellar velocity dis- al. 2020); there are 566 BCGs within the redshift range persion of galaxies often select early-type or quiescent 0.00 < z < 0.14. Among the 3390 systems we identify, galaxies (Faber & Jackson 1976; Zahid et al. 2016; Sohn the primary galaxies of 33 systems are known BCGs. To et al. 2017a,b), using colors or central spectral features reduce the contamination of the cluster galaxies, we re- Dn4000 of galaxies. Here we build two subsamples move these systems (33 primaries hosting 89 satellites) with early-type and late-type primary galaxies based in our sample. We then match this list with the catalog on color and Dn4000. Dn4000 is a spectral index for ˚ of the central stellar velocity dispersion with σ /δσ = 3. the strength of the 4000 A break, which is defined 0 0 ˚ ˚ The final catalog we use includes 2807 primary galaxies as Fν (4000 − 4100A)/Fν (3850 − 3950A) (Balogh et al. with central stellar velocity dispersion and 3417 satellite 1999; Fabricant et al. 2008; Geller et al. 2014). Dn4000 galaxies. shows a strong correlation with stellar population age, Figure2 shows the stacked phase-space diagram (i.e., and Dn4000 = 1.6 corresponds approximately to the R − v diagram) of the entire sample. We place the pri- stellar population age of 1 Gyr (Kauffmann et al. 2003). mary galaxies at the center and calculate the projected In the literature, the galaxies with Dn4000 ≥ 1.6 are galactocentric distance of satellites (∆D) and the radial often classified as quiescent galaxies, and those with Dn4000 < 1.6 are classified as star-forming galaxies velocity differences (∆V = c∆z/(1 + zprimary), where (Mignoli et al. 2005; Woods et al. 2010). We note that ∆z is the redshift difference and zprimary is the redshift of the primary galaxy). A majority of satellites are lo- the color of a galaxy represents the mean property of the entire stellar population of the galaxy, while the Dn4000 5

1 1 1 1 1 1 50 < *, Prim 100 (km s ) 100 < *, Prim 150 (km s ) 150 < *, Prim 200 (km s ) 200 < *, Prim 250 (km s ) 250 < *, Prim 300 (km s ) 300 < *, Prim 450 (km s ) 60 10 100 100 75 20 40 50 5 50 50 10 20 25

0 0 0 0 0 0 1 1 1 1 50 < *, Prim 100 (km s ) 100 < *, Prim 150 (km s ) 150 < *, Prim 200 (km s ) 200 < *, Prim 370 (km s ) 100

Number 6 60 30

50 40 20 4

20 10 2

0 0 0 0 1000 0 1000 1000 0 1000 1000 0 1000 1000 0 1000 v (km/s)

Figure 3. The distribution of relative velocity difference (∆V ) of satellite galaxies with respect to their primary galaxies. Top rows show ∆V distributions for red primary galaxies, and bottom rows show those for blue primary galaxies. Open histograms show the ∆V of all satellites, light-shaded histograms represent selected satellite galaxies after 2.7σ clipping, and dark-shaded regions represent the biweight σsat value.

500 rable peaks at Dn4000 ≈ 1.4 and 1.8, with a minimum at Dn4000 ≈ 1.6. This comparison suggests that a quies-

N 250 cent galaxy identification based only on Dn4000 includes a majority of red galaxies, but also includes some blue galaxies. 3.0 Figure5 displays a color-magnitude diagram (CMD) of the selected primary galaxies: quiescent galaxies with Dn4000 ≥ 1.6 and star forming galaxies with Dn4000 < 2.5 1.6. For comparison, we plot the color histograms of qui-

0 escent and star forming galaxies in the top panel. The 0 0

4 2.0 CMD of the quiescent galaxies shows a strong concentra- n

D tion of red galaxies along the red sequence with a peak color at (g − r)0 ≈ 0.95. Star-forming galaxies are much 1.5 bluer than the red sequence and show a much broader color range. It is also noted that a small number of qui- escent galaxies (as well as star forming galaxies) show 1.0 opposite trends. These galaxies must have different star 0.2 0.4 0.6 0.8 1.0 1.2 1.4 100 200 300 400 500 formation histories between the central region and outer (g r) N 0 region in the same galaxy. Figure 4. Dn4000 vs. (g −r)0 color of primary galaxies and We divide the sample of primary galaxies according the distribution of each parameter (upper and right panels). to their colors: red galaxies with (g − r)0 > 0.85 and Blue and red symbols represent blue and red galaxies, re- blue galaxies with (g − r)0 ≤ 0.85. We divided the sam- spectively. The dashed lines mark the boundary used for the ple by color because it is an efficient approximate proxy classification of primary galaxies. for morphology (early types vs. late types), kinemat- ics (pressure dominated versus rotation dominated) and of a galaxy we measure with SDSS fibers represents only star-formation history, as adopted in the previous stud- the stellar population in the central region of the galaxy. ies (e.g., More et al. 2011; Wojtak & Mamon 2013; Lange Figure4 displays Dn4000 versus (g − r)0 of the pri- et al. 2019). In this classification, ∼ 95% of red galax- mary galaxies. Dn4000 shows a strong correlation with ies are quiescent galaxies with Dn4000 ≥ 1.6. Based on (g − r)0 color, but with a significant scatter. The top visual inspection of SDSS images of the primary galax- panel of Figure4 shows the ( g − r)0 color distribution ies, we confirm that the red galaxies are mostly early of the primary galaxies. The right panel of Figure4 types and the blue galaxies are mainly late types. Here- displays the histogram of Dn4000 of the primary galax- after, these red and blue primary galaxies are the main ies. The Dn4000 distribution of red galaxies shows a subsamples in the following analysis. dominant single peak at Dn4000 ≈ 2.0. In contrast, the Dn4000 distribution of blue galaxies shows two compa- 6

500 Table 1. Number of Galaxies in Subsamples

Sample NP rimary NSatellite

N 250 FoF Systems 3390 4218

All Primaries (σ0/δσ0 > 3) 2807 3417 Red Primaries ((g − r)0 > 0.85, σ/δσ > 3) 1938 2461 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Blue Primaries ((g − r)0 ≤ 0.85, σ/δσ > 3) 869 956 Dn4000 1.6 24

22 galaxies (as well as in applying the aperture correction and the inclination correction), because the effect of ro- 20 tation is considered to be minor and because it is diffi- cult to estimate the fraction of the rotation effect in our

r 18 sample from the SDSS fiber spectra. M 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Table1 summarizes the number of each subsample. Dn4000 < 1.6 Main subsamples for the following analysis are 1938 red 24 primaries with 2461 satellites and 869 blue primaries with 956 satellites. The number of quiescent primaries 22 (2142) with satellites (2681) is slightly larger than that of the red ones.

20 2.4. Sample Completeness

18 The galaxy catalog we use is constructed from the 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 SDSS spectroscopic sample, which is slightly incomplete (g r)0 due to fiber collision. Lazo et al.(2018) demonstrated Figure 5. Color-magnitude diagrams (middle and bottom that the incompleteness of the SDSS spectroscopic sam- panels) and color distributions (top panel) of primary galax- ple is ∼ 7% (see also Strauss et al. 2002). This survey ies. Red and blue symbols show quiescent and star-forming incompleteness may impact the identification of satellite populations segregated by Dn4000 = 1.6. The black dashed galaxies around isolated galaxies. lines mark the boundary of blue and red galaxies. We calculate the ratio between the number of spectro- scopic galaxies and the number of photometric galaxies We note that we repeated the following analyses based as a spectroscopic survey completeness. Figure6 com- on the quiescent and star-forming galaxies segregated pares the distributions of satellite galaxy candidates in by Dn4000 = 1.6. The results for quiescent primary the photometric (open histograms) and spectroscopic galaxies are essentially identical to the results for red (hatched histograms) samples with respect to (a) the primary galaxies. Thus, we mention the analyses for r-band magnitude of primary galaxies, (b) angular sep- quiescent galaxies only when needed in the following. aration of satellite candidates from their primary galax- We include the systems with blue primary galaxies. ies, (c) redshift of primary galaxies, and (d) projected Because blue galaxies are mostly disk galaxies, their distances (∆D) of satellites from their primary galaxies. central velocity dispersion can be affected by rotation. The blue symbols show the spectroscopic survey com- However, the central stellar velocity dispersion of these pleteness. Here we compute the spectroscopic survey galaxies measured with an SDSS fiber represents mainly completeness to r = 17.77 mag, the SDSS spectroscopic the bulge kinematics. Thus, it may be much less affected survey limit. The completeness of our sample is larger by the effect of rotation; ? showed that the effect of rota- than 90% and is almost constant for the entire range of tion is minor on the stellar velocity dispersion at R < 1.5 apparent magnitudes, redshifts, and angular separation. kpc in the low-redshift late-type galaxies based on stel- The survey completeness is relatively low (∼ 70%) near lar kinematics derived from the MANGA and CALIFA primary galaxies (∆D < 20 kpc), mainly due to the fiber sample of 2458 galaxies (see their Figures 2 and 3). We collision. However, the number of galaxies with missing did not take the rotation effect into account in using the spectroscopic information near primary galaxies is also measured central velocity dispersion of individual blue small (N = 12). In conclusion, the spectroscopic survey is essentially complete, and it would not significantly 7

1.0 1.0 600 (a) (b) 1500 0.9 0.9 All 500 Completeness Completeness 3 Red 0.8 0.8 400 10 1000 Blue 300 0.7 0.7 Number Number 0.6 0.6 200 500

100 0.5 0.5 102 0 0.4 0 0.4 0 50 100 150 200 250 0 120 240 360 480 600 720 840 960 D [kpc] 10801200 Angular separation [arcsec] 3500 1200 1.0 (c) (d) 1.0 3000 1000 1 0.9 10 0.9 2500 Completeness Completeness

0.8 800 Number of Primaries 2000 0.8 Photometric Galaxy Spectroscopic Galaxy 0.7 600 1500 0.7 Number Number 0.6 400 1000 0.6 100

0.5 200 500 0.5 0 1 2 3 4 5 6 7 8 9 10 11 0 0.4 0 0.4 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 Number of Satellites zprim 16.0 15.5 15.0 14.5 14.0 13.5 13.0 12.5 12.0 11.5 mr(Prim) Figure 7. Distributions of the number of satellite galaxies Figure 6. The number distribution of faint satellite galaxy for each primary galaxy: all primaries (open), red primaries candidates (two or more magnitudes fainter than their pri- (hatched), and blue primaries (filled). mary galaxies) in the photometric (gray open histograms) and spectroscopic (green hatched histograms) samples with Because each primary galaxy in the sample hosts only respect to (a) the r-band magnitudes of primary galaxies, (b) angular separation from primary galaxies, (c) redshifts of pri- a few satellites, estimating the velocity dispersion of an mary galaxies, and (d) projected distances of satellites from individual galaxy system has a significant uncertainty. primary galaxies. Blue symbols represent the completeness Thus, we compute the system velocity dispersion from of the spectroscopic samples. Completeness scale is labeled the stacked galaxy sample. We are interested in the rela- on the right axes. tion between the velocity dispersion of a galaxy system and the central stellar velocity dispersion of the host affect analysis of the galaxy systems with spectroscopi- galaxy in the system. Therefore, we stacked samples cally identified satellites. based on the central stellar velocity dispersion of the host galaxy. We divide the red and blue primary galaxies 3. PHYSICAL PROPERTIES OF GALAXY into several velocity dispersion bins with 50 km s−1 in- SYSTEMS tervals. We also apply 2.7σ clipping to exclude possible 3.1. Satellite Number Distribution intruders in each sample following Mamon et al.(2010) (see also Ferragamo et al.(2020) and references therein). Figure7 displays the distribution of satellite numbers The line-of-sight velocity dispersion profile measured for all (open), red (hatched), and blue (filled) primary with the objects within 2.7σ boundary yields the best fit galaxies. All primary galaxies have fewer than 10 satel- to that expected from the Navarro-Frenk-White density lite galaxies, and the median number of satellites per and velocity anisotropic profile. The clipping is iter- primary is 2. The blue primary galaxies generally host ated until the distribution converges or stops when the fewer satellites than the red primary galaxies because iteration number is 5. In each velocity dispersion bin, the former are generally fainter (and less massive) than there are typically ∼ 480 (median) galaxy systems with the latter. ∼ 590 (median) satellite galaxies. For these samples, we calculate the radial velocity differences of the satellites 3.2. Estimation of Velocity Dispersion of Galaxy with respective to the primary galaxies and use them for Systems computing the system velocity dispersion. The velocity dispersion of relaxed systems is widely We use the biweight scale estimator (Beers et al. 1990) used to estimate the dynamical mass of a system based to calculate the velocity dispersion of galaxy systems. on the virial theorem (Watkins et al. 2010). In gen- The biweight method is widely used for computing the eral, the velocity dispersion of galaxy systems is derived velocity dispersion of galaxy systems (e.g., Zhang et al. from the line-of-sight radial velocity of satellites except 2011; Sohn et al. 2020). It is known to be more robust in some cases like the Local group where there are three- than other methods based on the assumption of Gaus- dimensional velocities of satellites available. 8 sian distribution (e.g., Danese et al. 1980), as described are, on average, larger than those of the blue primary in Beers et al.(1990). We calculate the uncertainties of galaxies for given magnitudes. The relation for the red velocity dispersions using 10,000 times bootstrap resam- primary galaxies shows a break at Mr ≈ −23 mag, and pling. Hereafter, we refer to σsys as the system velocity its slope becomes flatter at the brightest end (Mr ≤ −23 dispersion of the stacked sample. mag). We fit the relations between σ∗,prim and Mr with a linear function. Table2 lists the fitting results. Note 4. RESULTS that RMS values in these results are as small as 0.01 − Based on the spectroscopically identified sample of iso- 0.02 dex. The relation for the red primary galaxies with lated galaxies surrounded by faint satellites, we study Mr > −23 mag is given by the relation between the physical properties of the pri- mary galaxies and the properties of their dark matter log σ∗,prim = (−0.12 ± 0.01)Mr − (0.29 ± 0.23), (1) halo traced by the satellites. We first examine the rela- tion between the central stellar velocity dispersion and while the relation for the brightest galaxies at Mr ≤ −23 other physical properties (e.g. luminosity and stellar mag is almost flat, mass) in Sections 4.1 and 4.2. We then explore the re- log σ = (−0.09 ± 0.02)M + (0.40 ± 0.46). (2) lations between luminosity, stellar mass, and velocity ∗,prim r dispersion of primary galaxies and their system velocity The blue primary galaxies follow a similar relation but dispersion (Sections 4.3 - 4.5). with a steeper slope and a ∼0.2 dex smaller zero point for given magnitudes: 4.1. Central Stellar Velocity Dispersion and log σ = (−0.19 ± 0.01)M − (2.02 ± 0.19). (3) Luminosity of Primary Galaxies ∗,prim r Figure8 (a) shows the central stellar velocity disper- The linear relation of the red primary galaxies suggests 3.3 sion (σ∗,prim) as a function of r-band absolute magni- Lr ∝ σ∗,prim, which is consistent with the known Faber- tude (Mr) of the blue and red primary galaxies. Con- Jackson relation. This result is also similar to the recent tours represent the number density of galaxies in the results for the early-type galaxies (with log σ0 = 1.3 − diagram, and the median values of the central velocity 3.2) in the WINGS survey given by ? (see their Figure dispersion in each bin are marked by the symbols. In 1): log LV ∝ (3.25 ± 0.07) log σ0 with an RMS scatter of general, there is a strong correlation between σ∗,prim 0.35 dex. In contrast, the blue primary galaxies follow 2.1 and Mr. The σ∗,prim values of the red primary galaxies Lr ∝ σ∗,prim with a smaller power-law index.

Table 2. Fitting results between Mr and σ∗,prim for the sample galaxies

log σ∗,prim = α × Mr + β Sample α β Rms [dex]

All Primaries (Mr > −23) −0.18 ± 0.01 −1.65 ± 0.23 0.02 All Primaries (Mr ≤ −23) −0.07 ± 0.03 +0.88 ± 0.69 0.02 Red Primaries (Mr > −23) −0.12 ± 0.01 −0.29 ± 0.23 0.01 Red Primaries (Mr ≤ −23) −0.09 ± 0.02 +0.40 ± 0.46 0.01 Blue Primaries −0.19 ± 0.01 −2.02 ± 0.19 0.02 Satellite −0.25 ± 0.01 −3.01 ± 0.18 0.01

4.2. Central Velocity Dispersion and Stellar Mass of Primary Galaxies

In Figure8 (b), we show σ∗,prim as a function of stel- tively. The σ∗,prim of the blue and red primary galaxies lar mass (M∗) for the blue and red primary galaxies. is tightly correlated with their M∗, but with a different Blue and red symbols show the median σ∗,prim at each slope. Similar to the relations with luminosity, we note log M∗ bin for the blue and red primary galaxies, respec- that the slope of red primary galaxies becomes flatter 9

2.8 Red (a) M31 (b) 2.6 Blue MW ) 1 2.4 s

m 2.2 k / m i

r 2.0 P , * 1.8 ( g o

l 1.6

1.4

19 20 21 22 23 24 9 10 11 12 Mr log(M*/M )

Figure 8. Central velocity dispersion as a function of (a) the r−band absolute magnitude and (b) stellar mass of primary galaxies. Blue and red symbols show the mean values of blue and red galaxies, respectively. The blue and red dashed lines are the best-fit relations. The black solid line in the right panel shows the relation for SDSS quiescent galaxies in Zahid et al.(2016) (their Table 1). The gray plus and cross indicate the MWG and M31, respectively.

Table 3. Power-law fitting results between log(M∗/M ) and log σ∗,prim for the sample galaxies

log σ∗,prim = α × log(M∗/M ) + β Sample α β Rms [dex]

All Primaries (log M∗ ≤ 11.2) 0.44 ± 0.02 −2.50 ± 0.22 0.02 All Primaries (log M∗ > 11.2) 0.14 ± 0.05 +0.86 ± 0.56 0.01 Red Primaries (log M∗ ≤ 11.2) 0.32 ± 0.02 −1.16 ± 0.22 0.01 Red Primaries (log M∗ > 11.2) 0.14 ± 0.02 +0.85 ± 0.22 0.01 Blue Primaries 0.40 ± 0.02 −2.16 ± 0.22 0.04 Satellites 0.46 ± 0.04 −2.59 ± 0.38 0.14

at log(M∗/M ) > 11.2. This break is consistent with the red primary galaxies is the relation of SDSS early-type galaxies (Bernardi et al. 2011, see their Figure 1). Noting that major dry merg- ers can change stellar masses of galaxies significantly, log σ∗,prim = (0.32 ± 0.02) log(M∗/M ) − (1.16 ± 0.22) but they change the central velocity dispersions (and for log(M∗/M ) ≤ 11.2, colors) much less, Bernardi et al.(2011) suggest that this break (which appeared at log(M∗/M ) = 11.3 in their case) can be evidence of major-dry-merger-driven and growth of the massive galaxies. We fit, with a power law, the median σ∗,prim of blue log σ∗,prim = (0.14 ± 0.02) log(M∗/M ) + (0.85 ± 0.22) and red primary galaxies as a function of M∗. Table3 summarizes the fitting results. The best-fit relation for for log(M∗/M ) > 11.2. 10

The blue primary galaxies follow a steeper relation with function. Table4 lists the best fit results to these rela- a ∼0.1 dex smaller zero point for a given stellar mass, tions. The best-fit relations are log σ∗,prim = (0.40 ± 0.02) log(M∗/M ) − (2.16 ± 0.22). log σsys = −(0.16 ± 0.01)Mr − (1.04 ± 0.27) (4) for the red primaries, These relations suggest that the red primary galaxies 3.1 follow M∗ ∝ σ∗,prim, and that the blue primary galaxies and 2.5 follow M∗ ∝ σ∗,prim. We compare the relation for the red primary galax- log σsys = −(0.12 ± 0.03)Mr − (0.50 ± 0.63) ies with a similar relation derived from SDSS quiescent for the blue primaries. galaxies from Zahid et al.(2016) (the black solid line in Figure8 (b)). Zahid et al.(2016) use a sample of qui- Thus, the red and blue primary galaxies show a similar slope in this relation, but the red primary galaxies show, escent galaxies with Dn4000 > 1.5, log(M∗/M ) > 9, and 0.02 < z < 0.20. They also show that the observed on average, ∼ 0.2 dex larger log σsys values than the relations are fitted with a broken power law; the slope blue primary galaxies for given magnitudes. The best- changes at log(M /M ) = 10.26. The slope of this re- fit results suggest that the red primary galaxies follow ∗ 2.5 lation for the same mass range is similar to the slope of Lr ∝ σsys and the blue primary galaxies follow Lr ∝ 3.3 the red primary galaxies we derive. σsys. These power-law indices are slightly different from We also plot the data for the MWG and M31 in Fig- those in the Lr − σ∗,prim relation (see Section 4.1). ure8 (b). We adopt the stellar mass of the MWG 10 11 Table 4. Fitting results between Mr and log σsys for the ((5±1)×10 M ) and M31 (1.0×10 M ) from Bland- Hawthorn & Gerhard(2016) and Sick et al.(2015), re- primary galaxies spectively. We use the σ∗,prim of the MWG (113 ± 3 log σsys = α × Mr + β −1 km s ) measured within the half-mass radius of its Sample α β Rms [dex] bulge (< 1 kpc) from Bland-Hawthorn & Gerhard All Primaries −0.15 ± 0.02 −0.98 ± 0.50 0.04 (2016). The σ of M31 (173 km s−1) is from Whit- ∗,prim Red Primaries −0.16 ± 0.01 −1.04 ± 0.27 0.02 more et al.(1979) (see also a slightly lower value, 166 Blue Primaries −0.12 ± 0.03 −0.50 ± 0.63 0.05 km s−1, in Figure 4 of Saglia et al.(2010)). The MWG and M31 are located close to the relation for the blue primary galaxies. Figure9 displays σ∗,prim versus Mr and M∗ for all primary and satellite galaxies. Fitting results for these 4.4. System Velocity Dispersion and Stellar Mass of data are also listed in Tables2 and3. While the primary Primary Galaxies galaxies show a flatter slope than the satellite galaxies In Figure 11 we plot the relation between σsys and in the σ∗,prim − Mr relation, both galaxy samples show the stellar mass of primary galaxies (M∗). Similar to a similar slope in the σ∗,prim −M∗ relation. For compar- ison we plot the relation for the SDSS quiescent galaxies the relations with luminosity, σsys is proportional to M∗ from Zahid et al.(2016). The relation for the primary for both red and blue primary galaxies. Table5 lists the and satellite galaxies in this study agrees well with the best-fit results for these relations we obtain with a power relation for the quiescent galaxies in Zahid et al.(2016). law. The best-fit relations are The MWG and M31 are located close to the relation for log σsys = (0.33 ± 0.03) log(M∗/M ) − (1.25 ± 0.28) the primary galaxies. for the red primaries, 4.3. System Velocity Dispersion versus Luminosity of Primary Galaxies and

Figure 10 shows the relation between the system ve- log σsys = (0.25 ± 0.01) log(M∗/M ) − (0.49 ± 0.16) locity dispersion (σsys) and r−band absolute magnitude for the blue primaries. of the primary galaxies. Here, we adjusted the bin size along the absolute magnitude so that the number of data Thus, the red primary galaxies show, on average, ∼0.1 in each bin is similar. Then, we calculate the system ve- dex larger log σsys values than the blue primary galax- locity dispersion from the stacked systems. ies for given stellar masses. It is noted that the red For both blue and red primary galaxies, σsys is pro- primary galaxies show a steeper slope than the blue pri- portional to Mr. We fit these relations using a linear mary galaxies in the log σsys − log M∗ relation, while 11

Primary (a) M31 (b) Satellite MW ) 2.5 1 s

2.0 m k / m i 1.5 r P , *

( 1.0 g o l 0.5

17 18 19 20 21 22 23 24 8 9 10 11 12 Mr log(M*/M )

Figure 9. Same as Figure8, but for primary (magenta) and satellite (green) galaxies. The meanings of the symbols are the same as in Figure8.

Red 2.75 2.75 Blue M31 MW ) ) 1 2.50 1 2.50 s s

m m k k

/ /

s s y y s s 2.25 2.25 ( ( g g o o l l

2.00 2.00 Red Blue

20 21 22 23 24 9 10 11 12 Mr log(M*/M )

Figure 10. Velocity dispersion of galaxy systems as a func- Figure 11. Velocity dispersion of galaxy systems as a func- tion of r−band absolute magnitude for blue (blue circles) and tion of the stellar mass of blue and red primary galaxies. The red (red squares) primary galaxies. The blue and red lines lines and symbols are the same as in Figure 10. The gray show the fitting results for blue and red primary galaxies. plus and cross indicate the MWG and M31, respectively.

they show the opposite trend in the log σ∗,prim − log M∗ MWG and M31 using the line-of-sight velocity of their relation (for log M∗ < 11.2). These relations indicate satellite galaxies. From the satellite galaxy catalogs in 0.33 that σsys ∝ M∗ for the red primary galaxies and Kashibadze, & Karachentsev(2018), we select 33 and 0.25 σsys ∝ M∗ for the blue primary galaxies. 40 satellite galaxies within 200 kpc from the MWG and We plot the data for the MWG and M31 in Figure M31, respectively. Based on their line-of-sight veloci- 11. We estimate the system velocity dispersion of the ties, we compute the system velocity dispersion of the 12

−1 MWG (σ = 119±15 km s ) and M31 (σM31 = 110±15 Sim (Zahid+18) km s−1), respectively. The errors of the system veloc- Early Type (Serra+18) ity dispersion were derived by bootstrapping with 1000 Spirals (Davis+19) Red 2.50 times resampling. The MWG shows a slightly lower σsys Blue

) M31 value than the mean relation for the blue primary galax- 1 MW s ies for given stellar mass, but its offset is only at the 2σ m level. M31 shows a larger offset at the 3σ level. The k 2.25

/ line-of-sight velocity distribution of the satellite galax- s y ies of the blue primary systems has a wider range of s (

−1 g

−450 < (vlos/km s ) < 450 than that for the MWG o −1 l 2.00 and M31 (−250 < (vlos/km s ) < 250). The differ- ence in the line-of-sight velocity distribution of satellite galaxies appears to be intrinsic because the same dif- ference is shown when we use the blue primary systems 1.75 identified by a tighter FoF linking length (∆V < 500 1.75 2.00 2.25 2.50 −1 1 km s ). The wider radial velocity distribution results log( *, Prim/ km s ) in the larger velocity dispersion of the blue primary sys- tems than the MWG and M31. To further understand Figure 12. Velocity dispersion of galaxy systems as a func- this issue, we need to identify the MWG and M31 system tion of central stellar velocity dispersion of primary galaxies. The symbols are the same as in Figure 11. The dotted and analogs that have a similar magnitude (mass) distribu- dotted-dashed lines represent the power-law fits for red and tion of the satellite galaxies. This is beyond the scope blue primaries. The black solid line shows the relation for of this paper. quiescent galaxies from the Illustris-1 simulations (Zahid et al. 2018). The cyan band represents the vmax − σ0 relation

Table 5. Fitting results between log(M∗/M ) and for spiral galaxies in Davis et al.(2019). The magenta band log σsys for the primary galaxies is the vcirc − σe relation for early-type galaxies in Serra et al.(2016). The gray line is a one-to-one relation. The gray log σsys = α × log(M∗/M ) + β plus and cross indicate the MWG and M31, respectively. Sample α β Rms [dex]

All Primaries 0.35 ± 0.01 −1.42 ± 0.15 0.02 the σsys and σ∗,prim for the blue primary galaxies are Red Primaries 0.33 ± 0.03 −1.25 ± 0.28 0.02 strongly correlated with each other. Blue Primaries 0.25 ± 0.01 −0.49 ± 0.16 0.01 We compare our results with the relation for the dark matter halos based on hydrodynamic simulations. Based on the Illistris-1 cosmological simulations, Za- hid et al.(2018) derive scaling relations between the dark matter halo mass and the velocity dispersion of 4.5. System Velocity Dispersion and Primary Central quiescent galaxies with star formation rates less than −10 −1 Velocity Dispersion 2 × 10 M yr . Combining their Equations (4) and Figure 12 shows the velocity dispersion of galaxy sys- (7), we obtain a relation between the dark matter halo velocity dispersion (σT,DM ) and the line-of-sight stellar tems (σsys) versus the central stellar velocity dispersion velocity dispersion for the half-mass radius of galaxies (σ∗,prim) of the blue and red primary galaxies. Remark- (σh,∗): log σT,DM = 0.99 log σh,∗ + 0.06 with an RMS ably, σ∗,prim of the red primary galaxies shows a tight of 0.17 dex. Here we assume that σT,DM corresponds correlation with σsys. Table6 summarizes the power-law fitting results. The best-fit result for the red primary to the system velocity dispersion (σsys) and σh,∗ cor- galaxies is responds to the stellar velocity dispersion of primary galaxies (σ∗,prim). log σsys = (1.01 ± 0.13) log σ∗,prim + (0.05 ± 0.31). (5) The black solid line in Figure 12 shows the relation from Zahid et al.(2018). Their relation has a simi- The slope of this relation is essentially identical to one. lar slope to that of the red primary galaxies although We also check the same relation for quiescent primary the zero point is slightly lower. The difference in zero galaxies with Dn4000 ≥ 1.6; their relation is consistent points could have originated from the different defini- with that of the red primary galaxies. The relation for tions of system velocity dispersion in the two studies. the blue primary galaxies is slightly flatter (α = 0.74 ± The relation of the blue primary galaxies is significantly 0.09) than for that for the red primaries. Nevertheless, 13 shallower than the relation from the simulations for qui- with the central stellar velocity dispersion of each of the escent galaxies. same galaxy. We also plot the data for the MWG and M31 in Fig- Estimating the rotational velocity of early-type galax- ure 12. The data for the MWG are consistent with the ies is not straightforward. Serra et al.(2016) measured mean relation for the blue primary galaxies (it is slightly the rotational velocity (vcirc,HI ) of 16 early-type galax- lower but at the 1σ level). However, M31 shows a larger ies hosting a large regular HI disk (or ring). They de- offset at the 3σ level. M31 also shows an offset from the rived a tight relation between vcirc,HI and central stel- simulations (Zahid et al. 2018) at the 3σ level, while the lar velocity dispersion (log σe = 2.0 − 2.4): log vmax ∝ MWG is consistent with the simulations. (0.96 ± 0.11) log σe (their equation (1)). We plot this re- lation with a magenta band in Figure 12. This relation Table 6. Power law fitting results between σ∗,prim and σsys is in good agreement with the σsys − σ∗,prim relation for for the primary galaxies the red primary galaxies we derived.

log σsys = α × log σ∗,prim + β Sample α β RMS [dex] 5.2. Relations between Primary Properties and Dark All Primaries 0.94 ± 0.09 0.22 ± 0.22 0.03 Red Primaries 1.01 ± 0.13 0.05 ± 0.31 0.02 Matter Halo Mass Quiescent Primaries 0.99 ± 0.13 0.10 ± 0.30 0.03 Many previous studies investigate the relation be- Blue Primaries 0.74 ± 0.09 0.62 ± 0.20 0.01 tween the physical properties of primary (central) galax- a Zahid et al.(2018) 0 .99 ± 0.03 0.06 ± 0.06 0.17 ies and their dark matter halo mass. Following this, we aRMS of Zahid et al.(2018) was derived from quadratic summa- also investigate the relation between the stellar mass tion of their equations (3) and (8). and the central velocity dispersion of primary galaxies as a function of the dark matter halo mass inferred from their system velocity dispersions. We convert σsys into MDM based on the two scaling relations from Rines et al.(2016) and Abdullah et al. 5. DISCUSSION (2020). Rines et al.(2016) derive the relation based on the system velocity dispersion of 21 clusters and their 5.1. Comparison of σsys − σ∗,prim and vmax − σ0 mass estimated from Planck Sunyaev-Zel’dovich mea- Relations surements (MSZ , see their equation 5). We also used The rotation velocity of galaxies is another well-known the scaling relation from Abdullah et al.(2020) based on tracer for the dark matter halo of rotating galaxies, GalWcat, which includes ∼18,000 galaxy clusters iden- which is often applied to disk galaxies (Serra et al. 2016; tified from the SDSS spectroscopic sample. Abdullah et Davis et al. 2019; Katz et al. 2019). For spiral galaxies, al.(2020) estimate the σsys and MDM of galaxy clus- the maximum rotational velocity has been used as an ters using a virial mass estimator (Binney & Tremaine accurate tracer of circular velocity at large radii (i.e., 1987; Rines et al. 2013) and derive the best-fit rela- 15 (0.349±0.142) vmax ≈ vcirc). Thus, the maximum rotational velocity is tion σ200 = (946 ± 52) ∗ [h(z)M200/10 M ] , used to estimate the mass of the dark matter halo (Davis where σ200 is the velocity dispersion of galaxies within et al. 2019; Katz et al. 2019). For example, Davis et al. R200 and M200 is the mass within R200 in the unit of −1 (2019) explored a scaling relation between vmax and the [h M ]. The relation explains the observed relation 14 bulge (central) stellar velocity dispersion (σ0) based on well, particularly at ∼ 10 M , where the relations de- a sample of 40 spiral galaxies with log σ0 = 2.0 − 2.35. rived from cosmological simulations show slight offsets They showed that there is a correlation between vmax (e.g., Evrard et al. 2008; Munari et al. 2013; Saro et al. and σ0, but with a large scatter (their equation (6): 2013; Armitage et al. 2018). We simply assume that log vmax ∝ (0.65±0.10) log σ0 with an RMS of 0.07 dex). σsys and MDM correspond to σ200 and M200 (or MSZ ), The cyan band in Figure 12 shows the vmax − σ0 re- respectively. lation for the spiral galaxies from Davis et al.(2019). Figure 13 illustrates the M∗ and MDM relations for The slope of this relation is similar to that of the the isolated galaxy systems. We use the scaling relations σsys −σ∗,prim relation of the blue primary galaxies. The from Rines et al.(2016) (the left panel) and Abdullah et zero-point difference (∼ 0.15 dex) is within in the un- al.(2020) (the right panel), respectively. The relations certainties of the relations. The consistency in these between M∗ and MDM were derived based on the two relations suggest that both vmax for the spiral galaxies scaling relations that show a zero-point offset resulting and σsys for the blue primary galaxies are correlated from the different definitions of MDM : MDM = MSZ in 14

UM (z=0.1, Q) (a) W&M 13 (Red) (b) 14 UM (z=0.1, SF) W&M 13 (Blue) Girelli20 (z = 0) M31

) Red MW Blue M

/ 13 M D M ( g o

l 12

11 9 10 11 12 9 10 11 12

log(M*/M )

Figure 13. Dark matter halo mass as a function of stellar mass of the primary galaxies based on (a) the Rines et al.(2016) relation and (b) the Abdullah et al.(2020) relation. Blue and red symbols and lines are for the blue and red primary systems. The open plus and cross indicate the MWG and M31, respectively, based on the DM halo mass from the literature. The filled plus and cross display the MWG and M31, respectively, based on their DM halo mass we translate from their stellar velocity dispersions using the scaling relation. Black filled and open triangles show the relations for blue and red SDSS galaxies from Wojtak & Mamon(2013). The relations of Wojtak & Mamon(2013) are shifted to be consistent with the results in this study (see Section 5.2.). Yellow solid and blue dashed curves are the relations for quenched and star-formation populations generated with UniverseMachine DR1 (Behroozi et al. 2019). The dotted-dashed line is the model for z = 0 from Girelli et al.(2020).

Rines et al.(2016) and MDM = M200,virial in Abdullah and et al.(2020). There are strong correlations between M∗ and MDM log(MDM /M ) = (0.79±0.05) log(M∗/M )+(4.05±0.49), for both blue and red primary systems, similar to the (7) results from previous observational studies (e.g., Woj- respectively. We also list the best-fit relations between tak & Mamon 2013; Erfanianfar et al. 2019). Table7 M∗ and MDM estimated based on the scaling relation summarizes the fitting results based on the scaling re- from Abdullah et al.(2020) in Table7. The slope for lation from Rines et al.(2016) with the RMS values of the red primary galaxies based on the two scaling rela- 0.04 − 0.05 dex. The best-fit relations between M∗ and tions is essentially identical to one; the slope for the blue MDM for the red and blue primary galaxies are primary galaxies is slightly shallower. Interestingly, the red primary galaxies are located in a more massive (∼0.4 log(MDM /M ) = (1.05±0.08) log(M∗/M )+(1.66±0.88), dex) dark matter halo than the blue primary galaxies at (6) a given M∗ of the primary galaxies. Figure 13 compares the relations we derive with those magnitude relation rather than a simple color selection. from Wojtak & Mamon(2013), also obtained from the They obtained the stellar mass of galaxies from the isolated galaxy systems in the SDSS spectroscopic sam- SDSS MPA/JHU catalog (Salim et al. 2007); these mea- ple. Wojtak & Mamon(2013) identified the isolated surements are systematically larger (∼0.2 dex) than the galaxy systems from SDSS DR7 based on more generous mass estimates we use. Thus, we shifted the relations criteria than our selections: ∆D < 1 Mpc, |∆V | < 1500 from Wojtak & Mamon(2013) by −0.2 dex in the M∗ −1 km s , and ∆Mr > 1.505. Their sample includes 3800 direction for a fair comparison. They compute the MDM red and 1600 blue primary galaxies surrounded by 8800 values using their projected phase-space (PPS) model, and 2600 satellite galaxies, respectively. The red and which yields an MDM distribution consistent with that blue primary galaxies are separated based on the color- from ΛCDM simulations. 15

Table 7. Fitting results between log(MDM /M and log(M∗/M ) for the primary galaxies

log(MDM /M ) = α × log(M∗/M ) + β Sample α β Rms [dex] Ref.a All Primaries 1.09 ± 0.04 1.13 ± 0.46 0.05 Red Primaries 1.05 ± 0.08 1.66 ± 0.88 0.05 Rines et al.(2016) Blue Primaries 0.79 ± 0.05 4.05 ± 0.49 0.04 All Primaries 1.00 ± 0.04 2.41 ± 0.42 0.04 Red Primaries 0.96 ± 0.07 2.90 ± 0.80 0.05 Abdullah et al.(2020) Blue Primaries 0.73 ± 0.04 5.08 ± 0.45 0.03 a References for the satellite velocity dispersion-halo mass relations we used to compute MDM .

The M∗ and MDM relations we derive using the Rines There are interesting differences between the relations et al.(2016) relation are consistent with the results from from observations and the UniverseMachine. The Uni- Wojtak & Mamon(2013), while those we derive using verseMachine models show a break at log(M/M ) ≈ the Abdullah et al.(2020) relation show slight offsets 10.5, which is absent from our results (also from the to the higher MDM . We note that the MDM used in results of Wojtak & Mamon 2013). The UniverseMa- Wojtak & Mamon(2013) is slightly lower than the MDM chine models at log(MDM /M ) > 12 are much steeper used in other previous studies (see discussion in Wojtak than the relations for the red primary galaxies in this & Mamon 2013; van Uitert et al. 2016). For example, study. The relation for z = 0 in Girelli et al.(2020) is van Uitert et al.(2016) discussed that the MDM from even steeper than the UniverseMachine models at the Wojtak & Mamon(2013) is 30% − 40% lower than the high-mass end. In addition, the difference between the mass estimates based on weak lensing analysis. quenched and star-forming models in UniverseMachine For comparison, we show the MWG and M31 in Fig- is much smaller than that between the red and blue pri- ure 13. We derive MDM,MW from σsys values for the mary galaxies in this study. Further studies with inde- MWG and M31 using the scaling relations in Rines et pendent measurements of MDM of the galaxy systems al.(2016) and Abdullah et al.(2020) and plot them with (e.g., based on lensing techniques) are needed to under- filled symbols. We also obtain MDM,MW ((1.4 ± 0.3) × stand these differences with the models. 12 12 10 M ) and MDM,M31((1.4±0.4)×10 M ) estimated Figure 14 shows the MDM we derived as a function of from Watkins et al.(2010), plot them with open sym- σ∗,prim. As expected from the linear relation between bols. The values we derive using the Rines et al.(2016) σ∗,prim and σsys derived in this study, MDM is propor- relation are more similar to those based on Watkins et tional to σ∗,prim. We fit the relation we derive using the al.(2010). The MWG and M31 show offsets to the scaling relation from Rines et al.(2016) and Abdullah et lower value at the 2σ and 3σ levels from the relation al.(2020) with a power law, listing the results in Table for the blue primary systems derived using the Rines 8. The rms values are 0.04–0.10. The best-fit relations et al.(2016) relation, and they show slightly larger off- between σ∗,prim and MDM estimated using the scaling sets from the results based on the Abdullah et al.(2020) relation from Rines et al.(2016) are relation. We also compare the observed relations with similar log(MDM /M ) = (3.17±0.41) log σ∗,prim+(5.74±0.97), relations based on simulations from UniverseMachine (8) DR1 (Behroozi et al. 2019) and on models from Girelli and et al.(2020). The solid and dashed lines in Figure 13 log(MDM /M ) = (2.33±0.29) log σ∗,prim+(7.53±0.61), display the M∗ − MDM relation for quenched and star- (9) forming populations for z = 0.1, respectively, in the for the red and blue primary galaxies, respectively. UniverseMachine. Because UniverseMachine does not These relations are useful in estimating the dark mat- provide models for blue and red galaxies separated by ter halo mass of the galaxies for which the values of the colors, we assume that the models for quenched and star- central velocity dispersion are available. forming populations correspond to red and blue galaxies The relation for the red primary galaxies based on in this study. the Rines et al.(2016) relation agrees very well with the expected relation based on the Illustris-I simulations 16

Sim (Zahid+18) (a) (b) Red 14 Blue ) M31 M

/ MW

M

D 13 M ( g o l 12

1.8 2.0 2.2 2.4 2.6 1.8 2.0 2.2 2.4 2.6 1 log( *, Prim / km s )

Figure 14. Dark matter halo mass (derived from satellite velocity dispersion using the relation in (a) Rines et al.(2016) and (b) Abdullah et al.(2020)) vs. the central velocity dispersion of blue and red primary galaxies. The yellow band represents the relation (M200 vs. σh,∗(= σ∗,prim)) based on Illustris-1 simulations given by Zahid et al.(2018). The meanings of the gray symbols are the same as in Figure 13. (Zahid et al. 2018), as shown by the yellow band. The pared to the relations from the numerical simulations. observed relation based on the Abdullah et al.(2020) The data of the MWG are consistent with the relation scaling relation shows an offset toward high MDM com- for the blue primary galaxies, and M31 shows an offset at the 3σ level to the lower MDM .

a Table 8. Fitting results between σ∗,prim and MDM for the primary samples

log MDM = α × log σ∗,prim + β Sample α β Rms [dex] Ref.a All Primaries 2.94 ± 0.29 6.27 ± 0.68 0.10 Red Primaries 3.17 ± 0.41 5.74 ± 0.97 0.08 Rines et al.(2016) Blue Primaries 2.33 ± 0.29 7.53 ± 0.61 0.04 All Primaries 2.69 ± 0.27 7.11 ± 0.62 0.09 Red Primaries 2.90 ± 0.38 6.63 ± 0.89 0.07 Abdullah et al.(2020) Blue Primaries 2.13 ± 0.27 8.26 ± 0.56 0.09 a References for the M200 − σsys relations we used to compute MDM .

5.3. Comparison with Massive Clusters this relation is an important test for central galaxy for- We demonstrated that there is a clear relation be- mation and structure formation models. Figure 15 compares the σsys −σ∗,prim relations for the tween σ∗,prim and σsys for both blue and red primary systems. Sohn et al.(2020) investigate a similar relation isolated galaxy systems in this study and for the cluster between the stellar velocity dispersion of the BCGs and sample (gray circles) from Sohn et al.(2020). The com- the cluster velocity dispersion derived from a spectro- parison cluster sample includes 225 HeCS-omnibus clus- scopic sample of cluster members. They suggest that ters. The velocity dispersions of the clusters (σcl) are de- 17

Sim (Zahid+18) Early Type (Serra+18) 3.00 Spirals (Davis+19) Red Blue HeCS (Sohn+20)

) 2.75 M31 1 MW s

m 2.50 k

/ s y s

( 2.25 g o l

2.00

1.75 1.75 2.00 2.25 2.50 2.75 1 log( *, Prim/ km s )

Figure 15. Same as Figure 11, but in comparison with the HeCS-omnibus galaxy cluster sample from Sohn et al.(2020). The symbols are the same as in Figure 12. Gray points are the individual galaxy clusters, and the gray dashed line shows the best-fit relation for these galaxy clusters given by Sohn et al.(2020). The gray plus and cross indicate the MWG and M31, respectively. rived from spectroscopically identified cluster members the numerical simulation in Zahid et al.(2018), as Sohn (∼180 for each cluster) using the biweight scale, as used et al.(2020) pointed out. in this study. A majority of the HeCS-omnibus clusters We note that the definitions of σsys in the two studies −1 have 400 < σcl < 1000 km s , much larger than the are different. For isolated galaxy systems, we use satel- σsys for the isolated galaxy systems in this study. The lite galaxies identified by the FoF algorithm with the gray dashed line is the best-fit result from Sohn et al. limited linking length, which may include some intrud- (2020). ers. In the case of galaxy clusters, the cluster members Interestingly, the relation we derived for the isolated are identified based on the caustic technique (Diaferio & galaxy systems in this study differs clearly from the re- Geller 1997). Furthermore, we compute σsys from the lation for the galaxy clusters. Although the BCGs in stacked sample including many isolated systems, while some clusters with small log σcl ∼ 2.5 have similar stel- Sohn et al.(2020) measured σcl for individual galaxy lar velocity dispersions to those of the isolated systems, clusters. Zahid et al.(2018) estimate the stellar velocity a majority of the galaxy clusters show a much steeper dispersions and the dark matter halo velocity dispersions σcl − σ∗,prim relation. The relation for galaxy clusters from the simulations in a similar manner to this study. also has a different slope compared to the relation from 18

This difference in definition would introduce some dif- related with σ∗,prim, but with a shallower slope ferences in the relations. (α = 0.74 ± 0.09). The different relations of the galaxy clusters and the σ is also proportional to the luminosity and stel- isolated galaxy systems indicate that σsyss for these two • sys systems may trace different halos. In other words, the lar mass of the primary galaxies. In general, the more massive systems host red primary galaxies. σsys we derived for the isolated galaxy systems traces the local halo of the isolated primary galaxies, while Because there is a power-law relation between σ the σ for galaxy clusters traces the extended cluster • sys sys and the dark matter halo mass (M ), the phys- halo, much larger than the BCG halo. This separation is DM ical properties of primary galaxies are correlated consistent with the interpretation of cluster simulations. with their dark matter halo masses. We derive Dolag et al.(2010) demonstrate that there are two dy- the relations between σ and M (as well namically well-distinct stellar components in simulated ∗,prim DM as M ) for the further comparison with various galaxy clusters: a component with a small velocity dis- ∗ models. persion traces the local halo of the BCGs (or cD galax- ies) and the other component represents a diffuse stellar • We compare the σsys − σ∗,prim relation of the iso- population governed by the cluster halo. In this point of lated galaxy systems with the same relation for view, the σsys of galaxy clusters may be consistent with galaxy clusters from Sohn et al.(2020). The rela- the σ of the diffuse stellar population, and the σsys of tion for the galaxy cluster shows a much steeper isolated galaxy systems may be consistent with the σ of slope. The different slopes of the relations sug- the primary galaxy halos. gest that the σsys for our target systems and the To test this hypothesis further, observations of other clusters trace different halos, i.e., the local halo tracers for the primary galaxies of clusters and isolated of the primary galaxies and the halo of the entire galaxies are required. Strong lensing observations pro- clusters, respectively. vide excellent constraints on the mass associated with the galaxy and cluster halo, respectively (e.g., Monna In conclusion, the stellar velocity dispersion of a et al. 2017). For local BCGs or primary galaxies, obser- galaxy is an efficient and robust tracer for its dark mat- vations for other dynamical mass tracers including glob- ter halo mass. We highlight that the stellar velocity ular clusters and planetary nebulae enable us to con- dispersion as a robust spectroscopic measure which can straint the mass within the local halo of the primary be measured from the future large spectroscopic surveys galaxies (e.g., Ko et al. 2017; Longobardi et al. 2018a,b). like DESI, 4MOST, and Subaru/PFS. In the near fu- ture, wide-field imaging and spectroscopic observations 6. SUMMARY AND CONCLUSION will enable the exploration of dark matter halo prop- We construct a complete sample of isolated galax- erties based on statistical analysis of lensing observa- ies hosting faint satellites (∆r > 2 mag) by applying tions. Combining this dark matter mass estimates with the FoF algorithm to the SDSS DR12 spectroscopic a large sample of stellar velocity dispersion measure- galaxy catalog. We first divide the sample according ments would be an important test for galaxy and struc- to the color of the primary galaxies: red primary galax- ture formation models. ies with (g − r)0 > 0.85 and blue primary galaxies with (g − r)0 ≤ 0.85. Based on the large sample, we stack the galaxy systems depending on the physical proper- ACKNOWLEDGMENTS ties (i.e., Mr,M∗, σ∗,prim) of their primary galaxies to This work was supported by the National Research derive the system velocity dispersion. Then, we inves- Foundation grant funded by the Korean Government tigated the relation between the system velocity disper- (NRF-2019R1A2C2084019). We thank an anonymous sion and the physical properties of the primary galaxies. referee for useful comments and Margaret Geller for her The main results are summarized as follows. insightful advice on this project. We thank Brian S. Cho for his help in improving the English in the manuscript. • Velocity dispersions of the galaxy systems (σsys) show a strong correlation with the central stel- We also acknowledge Antonaldo Diaferio and Ken Rines lar velocity dispersion of the primary galaxies for helpful discussions. J.S. is supported by the CfA Fel- lowship. (σ∗,prim). In particular, in the case of systems with the red primary galaxies, σsys is directly This research has made use of NASA’s Astrophysics proportional to σ∗,prim with a slope of 1. The Data System Bibliographic Services. Funding for the σsys for the blue primary galaxies is also cor- Sloan Digital Sky Survey IV has been provided by the 19

Alfred P. Sloan Foundation, the US Department of En- Laboratory, Leibniz Institut f¨urAstrophysik Potsdam ergy Office of Science, and the Participating Institutions. (AIP), Max-Planck-Institut f¨urAstronomie (MPIA Hei- SDSS-IV acknowledges support and resources from the delberg), Max-Planck-Institut f¨urAstrophysik (MPA Center for High-Performance Computing at the Univer- Garching), Max-Planck-Institut f¨urExtraterrestrische sity of Utah. The SDSS website is www.sdss.org. Physik (MPE), National Astronomical Observatories of SDSS-IV is managed by the Astrophysical Research China, New Mexico State University, New York Uni- Consortium for the Participating Institutions of the versity, University of Notre Dame, Observat´arioNa- SDSS Collaboration including the Brazilian Partici- cional / MCTI, The Ohio State University, Pennsylva- pation Group, the Carnegie Institution for Science, nia State University, Shanghai Astronomical Observa- Carnegie Mellon University, the Chilean Participa- tory, United Kingdom Participation Group, Universidad tion Group, the French Participation Group, Harvard- Nacional Aut´onomade M´exico,University of Arizona, Smithsonian Center for Astrophysics, Instituto de As- University of Colorado Boulder, University of Oxford, trof´ısica de Canarias, The Johns Hopkins University, University of Portsmouth, University of Utah, Univer- Kavli Institute for the Physics and Mathematics of sity of Virginia, University of Washington, University of the Universe (IPMU)/University of Tokyo, the Ko- Wisconsin, Vanderbilt University, and Yale University. rean Participation Group, Lawrence Berkeley National

REFERENCES

Abdullah, M. H., Wilson, G., Klypin, A., et al. 2020, ApJS, Davis, B. L., Graham, A. W., & Combes, F. 2019, ApJ, 246, 2 877, 64 Arnouts, S., Cristiani, S., Moscardini, L., et al. 1999, Diaferio, A., & Geller, M. J. 1997, ApJ, 481, 633 MNRAS, 310, 540 D’Onofrio, M., Chiosi, C., Sciarratta, M., et al. 2020, A&A, Armitage, T. J., Barnes, D. J., Kay, S. T., et al. 2018, 641, A94 MNRAS, 474, 3746 Dolag, K., Murante, G., & Borgani, S. 2010, MNRAS, 405, Aquino-Ort´ız,E., S´anchez, S. F., Valenzuela, O., et al. 1544 2020, ApJ, 900, 109 Dutton, A. A., Conroy, C., van den Bosch, F. C., et al. Balogh, M. L., Morris, S. L., Yee, H. K. C., et al. 1999, 2010, MNRAS, 407, 2 ApJ, 527, 54 Ellison, S. L., Patton, D. R., Simard, L., et al. 2010, Barton, E., Geller, M., Ramella, M., et al. 1996, AJ, 112, MNRAS, 407, 1514 871 Erfanianfar, G., Finoguenov, A., Furnell, K., et al. 2019, Beers, T. C., Flynn, K., & Gebhardt, K. 1990, AJ, 100, 32 A&A, 631, A175 Behroozi, P., Wechsler, R. H., Hearin, A. P., et al. 2019, Evrard, A. E., Bialek, J., Busha, M., et al. 2008, ApJ, 672, MNRAS, 488, 3143 122 Bernardi, M., Roche, N., Shankar, F., et al. 2011, MNRAS, Faber, S. M., & Jackson, R. E. 1976, ApJ, 204, 668 412, L6 Fabricant, D. G., Kurtz, M. J., Geller, M. J., et al. 2008, Bernardi, M., Meert, A., Sheth, R. K., et al. 2013, MNRAS, PASP, 120, 1222 436, 697 Ferragamo, A., Rubi˜no-Mart´ın,J. A., Betancort-Rijo, J., et Binney, J., & Tremaine, S. 1987, Galactic Dynamics, al. 2020, arXiv:2006.05949 (A&A, 641, A41) Princeton Geller, M. J., Hwang, H. S., Fabricant, D. G., et al. 2014, Bland-Hawthorn, J., & Gerhard, O. 2016, ARA&A, 54, 529 ApJS, 213, 35 Brainerd, T. G. 2005, ApJL, 628, L101 Girelli, G., Pozzetti, L., Bolzonella, M., et al. 2020, A&A, Burkert, A., & Forbes, D. A. 2020, AJ, 159, 56 634, A135 Cappellari, M., Bacon, R., Bureau, M., et al. 2006, Huchra, J. P., & Geller, M. J. 1982, ApJ, 257, 423 MNRAS, 366, 1126 Ilbert, O., Arnouts, S., McCracken, H. J., et al. 2006, A&A, Conroy, C., Prada, F., Newman, J. A., et al. 2007, ApJ, 457, 841 654, 153 Karachentsev, I. D., Makarov, D. I., & Kaisina, E. I. 2013, Courteau, S., Cappellari, M., de Jong, R. S., et al. 2014, AJ, 145, 101 Reviews of Modern Physics, 86, 47 Karachentsev, I. D., & Kaisina, E. I. 2019, Astrophysical Danese, L., de Zotti, G., & di Tullio, G. 1980, A&A, 82, 322 Bulletin, 74, 111 20

Kashibadze, O. G., & Karachentsev, I. D. 2018, A&A, 609, Salucci, P. 2019, A&A Rv, 27, 2 A11 Saro, A., Mohr, J. J., Bazin, G., et al. 2013, ApJ, 772, 47 Katz, H., Desmond, H., McGaugh, S., et al. 2019, MNRAS, Schechter, P. L. 2015, arXiv:1508.02358 483, L98 Serra, P., Oosterloo, T., Cappellari, M., et al. 2016, Kauffmann, G., Heckman, T. M., White, S. D. M., et al. MNRAS, 460, 1382 2003, MNRAS, 341, 33 Sick, J., Courteau, S., Cuillandre, J.-C., et al. 2015, Galaxy Kluge, M., Neureiter, B., Riffeser, A., et al. 2020, ApJS, Masses as Constraints of Formation Models, 82 247, 43 Sohn, J., Geller, M. J., Hwang, H. S., Zahid, H. J., & Lee, Klypin, A., & Prada, F. 2009, ApJ, 690, 1488 M. G. 2016, ApJS, 225, 23 Ko, Y., Hwang, H. S., Lee, M. G., et al. 2017, ApJ, 835, 212 Sohn, J., Geller, M. J., Zahid, H. J., et al. 2017, ApJS, 229, Lange, J. U., van den Bosch, F. C., Zentner, A. R., Wang, 20 K., & Villarreal, A. S. 2019, MNRAS, 487, 3112 Sohn, J., Zahid, H. J., & Geller, M. J. 2017, ApJ, 845, 73 Lauer, T. R., Postman, M., Strauss, M. A., et al. 2014, Sohn, J., Geller, M. J., Diaferio, A., et al. 2020, ApJ, 891, ApJ, 797, 82 129 Lazo, B., Zahid, H. J., Sohn, J., et al. 2018, Research Notes Strauss, M. A., Weinberg, D. H., Lupton, R. H., et al. 2002, of the American Astronomical Society, 2, 234 AJ, 124, 1810 Longobardi, A., Arnaboldi, M., Gerhard, O., et al. 2018, Tempel, E., Tamm, A., Gramann, M., et al. 2014, A&A, A&A, 620, A111 566, A1 Longobardi, A., Peng, E. W., Cˆot´e,P., et al. 2018, ApJ, Thomas, D., Steele, O., Maraston, C., et al. 2013, MNRAS, 864, 36 431, 1383 Lynden-Bell, D., & Frenk, C. S. 1981, The Observatory, Utsumi, Y., Geller, M. J., Zahid, H. J., et al. 2020, ApJ, 101, 200 900, 50 Mamon, G. A., Biviano, A., & Murante, G. 2010, A&A, van Uitert, E., Cacciato, M., Hoekstra, H., et al. 2016, 520, A30 MNRAS, 459, 3251 McConnachie, A. W. 2012, AJ, 144, 4 Vilardell, F., Ribas, I., Jordi, C., et al. 2010, A&A, 509, A70 McKay, T. A., Sheldon, E. S., Johnston, D. et al. 2002, Vogelsberger, M., Genel, S., Springel, V., et al. 2014, ApJ, 571, L85 MNRAS, 444, 1518 Mignoli, M., Cimatti, A., Zamorani, G., et al. 2005, A&A, Wake, D. A., Franx, M., & van Dokkum, P. G. 2012, arXiv 437, 883 e-prints, arXiv:1201.1913 Monna, A., Seitz, S., Geller, M. J., et al. 2017, MNRAS, Watkins, L. L., Evans, N. W., & An, J. H. 2010, MNRAS, 465, 4589 406, 264 More, S., van den Bosch, F. C., Cacciato, M., et al. 2011, Wechsler, R. H., & Tinker, J. L. 2018, ARA&A, 56, 435 MNRAS, 410, 210 Whitmore, B. C., Kirshner, R. P., & Schechter, P. L. 1979, Munari, E., Biviano, A., Borgani, S., et al. 2013, MNRAS, ApJ, 234, 68 430, 2638 Wojtak, R., & Mamon, G. A. 2013, MNRAS, 428, 2407 Norberg, P., Frenk, C. S., & Cole, S. 2008, MNRAS, 383, Woods, D. F., Geller, M. J., Kurtz, M. J., et al. 2010, AJ, 646 139, 1857 Prada, F., Vitvitska, M., Klypin, A., et al. 2003, ApJ, 598, Zahid, H. J., Geller, M. J., Fabricant, D. G., & Hwang, 260 H. S. 2016, ApJ, 832, 203 Rines, K., Geller, M. J., Diaferio, A., et al. 2013, ApJ, 767, Zahid, H. J., Sohn, J., & Geller, M. J. 2018, ApJ, 859, 96 15 Zhang, Y.-Y., Andernach, H., Caretta, C. A., et al. 2011, Rines, K. J., Geller, M. J., Diaferio, A., et al. 2016, ApJ, A&A, 526, A105 819, 63 Zaritsky, D., Smith, R., Frenk, C., & White, S. D. M. 1993, Saglia, R. P., Fabricius, M., Bender, R., et al. 2010, A&A, ApJ, 405, 464 509, A61 Zaritsky, D., Smith, R., Frenk, C. S., & White, S. D. M. Salim, S., Rich, R. M., Charlot, S., et al. 2007, ApJS, 173, 1997, ApJL, 478, L53 267