On the Origin of the Hubble Sequence: I. Insights on Galaxy Color Migration from Cosmological Simulations
OOn the Origin of the Hubble Sequence: I. Insights on Galaxy ColorMigration from Cosmological Simulations
Renyue Cen ABSTRACT
An analysis of more than 3000 galaxies resolved at better than 114 pc/h at z =0 .
62 in a “
LAOZI ” cosmological adaptive mesh refinement hydrodynamic simulation isperformed and insights gained on star formation quenching and color migration. Thevast majority of red galaxies are found to be within three virial radii of a larger galaxy,at the onset of quenching when the specific star formation rate experiences the sharpestdecline to fall below ∼ − − − Gyr − (depending on the redshift). We shall thus callthis mechanism “environment quenching”, which encompasses satellite quenching. Twophysical processes are largely responsible: ram-pressure stripping first disconnects thegalaxy from the cold gas supply on large scales, followed by a longer period of cold gasstarvation taking place in high velocity dispersion environment, during the early partof which the existing dense cold gas in the central region ( ≤ in situ star formation. Quenching is found to be more efficient (i.e., a larger fractionof galaxies being quenched), but not faster (i.e., duration being weakly dependent onenvironment), on average, in denser environment. Throughout this quenching periodand the ensuing one in the red sequence galaxies follow nearly vertical tracks in thecolor-stellar-mass diagram. In contrast, individual galaxies of all masses grow most oftheir stellar masses in the blue cloud, prior to the onset of quenching, and progressivelymore massive blue galaxies with already relatively older mean stellar ages continue toenter the red sequence. Consequently, correlations among observables of red galaxies -such as the age-mass relation - are largely inherited from their blue progenitors at theonset of quenching. While the color makeup of the entire galaxy population stronglydepends on environment, which is a direct result of environment quenching, physicalproperties of blue galaxies as a sub-population show little dependence on environment.A variety of predictions from the simulation are shown to be in accord with extantobservations. Subject headings:
Methods: numerical, Galaxies: formation, Galaxies: evolution, Galax-ies: interactions, intergalactic medium
1. Introduction
The bimodal distribution of galaxy colors at low redshift is well established (e.g., Stratevaet al. 2001; Blanton et al. 2003a; Kauffmann et al. 2003; Baldry et al. 2004). The “blue cloud”, Princeton University Observatory, Princeton, NJ 08544; [email protected] a r X i v : . [ a s t r o - ph . C O ] J a n z = 0 . − . z ∼ .
08 andfrom ultraviolet imaging with GALEX at z ∼ .
4, find that the fraction of passive galaxies ishigher in groups than the field at both redshifts, with the difference between the group and fieldgrowing with time and larger at low masses. With the NOAO Extremely Wide-Field Infrared Imager(NEWFIRM) Survey of the All-wavelength Extended Groth strip International Survey (AEGIS) andCosmic Evolution Survey (COSMOS) fields, Whitaker et al. (2011) show evidence for a bimodalcolor distribution between quiescent and star-forming galaxies that persists to z ∼
3. Presottoet al. (2012) study the evolution of galaxies located within groups using the group catalog obtainedfrom zCOSMOS spectroscopic data and the complementary photometric data from the COSMOS 3 –survey at z = 0 . − . z = 0 . − .
20 using a spectroscopic sample of797 cluster and field galaxies drawn from the Gemini Cluster Astrophysics Spectroscopic Survey,finding that post starburst galaxies with M ∗ = 10 . − . M (cid:12) are three times more common in high-density regions compared to low-density regions. Based on data from the zCOSMOS survey Tanakaet al. (2012) perform an environment study and find that quiescent galaxies prefer more massivesystems at z = 0 . −
1. Rasmussen et al. (2012), analyzing GALEX imaging of a statisticallyrepresentative sample of 23 galaxy groups at z ∼ .
06, suggest an average quenching timescale of ≥ z = 0 . − ∼
20 per cent of the field in the mass range M star = 10 . − . M (cid:12) but are thedominant component of groups. Using SDSS ( z ∼ .
1) and the All-Wavelength Extended GrothStrip International Survey (AEGIS; z ∼
1) data, Woo et al. (2013) find a strong environmentaldependence of quenching in terms of halo mass and distance to the centrals at both redshifts.The widespread observational evidence of environment quenching is unsurprising theoretically.In regions of overdensity, whether around a large collapsed halo or unvirialized structure (e.g., aZel’dovich pancake or a filament), gas is gravitationally shock heated when converging flows meet.In regions filled with hot shock-heated gas, multiple gasdynamical processes would occur. One ofthe most important gasdynamical processes is ram-pressure stripping of gas, when a galaxy movesthrough the ambient hot gas at a significant speed, which includes, but is not limited to, the infallvelocity of a satellite galaxy. The theoretical basis for the ram-pressure stripping process is laiddown in the seminal work of Gunn & Gott (1972). Recent works with detailed simulations of thiseffect on galaxies (in non-cosmological settings) include those of Mori & Burkert (2000), Quilis et al.(2000), Kronberger et al. (2008), Bekki (2009) and Tonnesen & Bryan (2009).Even in the absence of ram-pressure stripping, ubiquitous supersonic and transonic motions ofgalaxies of complex acceleration patterns through ambient medium (intergalactic or circumgalacticmedium) subject them to the Raleigh-Taylor and Richtmyer-Meshkov instabilities. Large shearvelocities at the interfaces between galaxies and the ambient medium allow the Kelvin-Helmholtz(KH) instability to play an important role. When these processes work in tandem with ram-pressuredisplacements, the disruptive effects are amplified. For example, the KH instability time scale issubstantially shorter for a non-self-gravitating gas cloud (e.g., Murray et al. 1993) than for onesitting inside a virialized dark matter halo (e.g., Cen & Riquelme 2008).Another important process in hot environments is starvation of cold gas that is fuel for SF(e.g., Larson et al. 1980; Balogh et al. 2000; Dekel & Birnboim 2006). In regions with high-temperature and high-entropy cooling of hot gas is an inefficient process for fueling SF, an importantpoint noted long ago to account for the basic properties (mass, size) of galaxies (e.g., Binney1977; Rees & Ostriker 1977; Silk 1977). This phenomenon may be understood by consideringthe dependence of cooling time on the entropy of the gas: the gas cooling time can be written 4 –as t cool ( T, S ) = S / (cid:20) (cid:16) µ e µ (cid:17) k B T / Λ( T ) (cid:21) (Scannapieco & Oh 2004) . It follows that the minimumcooling time of a gas parcel just scales with S / . As a numerical example, for a gas parcel of entropy S = 10 K cm (say, for temperature 10 K and density 10 − cm − ) and metallicity 0 . (cid:12) , its coolingtime is no shorter than the Hubble time at z = 1 hence the gas can no longer cool efficiently to fuelSF. It may be that the combination of cold gas removal and dispersal by ram-pressure stripping,hydrodynamic instabilities, and cold gas starvation, all of which are expected to become increasinglyimportant in more massive environments, plays a primary role in driving the color migration fromthe blue cloud to the red sequence. In dense environments, gravitational tidal (stripping and shock)effects and relatively close fly-bys between galaxies (e.g., Moore et al. 1996) also become important.To understand the overall effect on SF quenching by these external processes in the context of thestandard cold dark matter model, a realistic cosmological setting is imperative, in order to capturecomplex external processes that are likely intertwined with large variations of internal propertiesof galaxies. In this paper we perform ab initio L arge-scale A daptive-mesh-refinement O mniscient Z oom- I n cosmological hydrodynamic simulations, called LAOZI Simulations , to obtain a largesample of galaxies to, for the first time, perform a chronological and statistical investigation on a verylarge scale. The large simulated galaxy sample size and very high resolution of LAOZI simulationsprovide an unprecedented opportunity to undertake the study presented. Our study shares thespirit of the work by Feldmann et al. (2011), who examine the evolution of a dozen galaxies fallingonto a forming group of galaxies, with a substantial improvement in the statistical treatment, thesimulation resolution, the range of environment probed, and the analysis scope. Feedback fromAGN is not included in this simulation, partly because of its large uncertainties and present lack ofdefinitive driving sources and primarily due to our intention to focus on external effects. Internaleffects due to SF are automatically included and we find no evidence that SF or merger triggeredSF plays a primary role in quenching from our study. The outline of this paper is as follows. In § § § § § § §
3, organized in an approximately chronologicalorder, starting with the ram-pressure stripping effects in § § § § z = 0 .
62 in § § where k B is Boltzmann’s constant, T temperature and Λ cooling function, µ = 0 .
62 and µ e = 1 .
18 for ionizedgas that we are concerned with, S is the gas entropy defined as S ≡ Tn / in units of K cm ( n is gas number density).If one conservatively adopts the lowest value of the term inside the bracket at the cooling peak at temperature T min ∼ . K, it follows that the minimum cooling time of a gas parcel just scales with S / .
2. Simulations2.1. Hydrocode and Simulation Parameters
We perform cosmological simulations with the AMR Eulerian hydro code, Enzo (Bryan &Norman 1999; Joung et al. 2009). First we run a low resolution simulation with a periodic boxof 120 h − Mpc (comoving) on a side. We identify a region centered on a cluster of mass of ∼ × M (cid:12) at z = 0. We then resimulate with high resolution of the chosen region embedded inthe outer 120 h − Mpc box to properly take into account the large-scale tidal field and appropriateboundary conditions at the surface of a refined region. The refined region has a comoving size of21 × × h − Mpc and represents a +1 . σ matter density fluctuation on that volume. The darkmatter particle mass in the refined region is 1 . × h − M (cid:12) . The refined region is surrounded bythree layers (each of ∼ h − Mpc) of buffer zones with particle masses successively larger by a factorof 8 for each layer, which then connects with the outer root grid that has a dark matter particle mass8 times that in the refined region. We choose the mesh refinement criterion such that the resolutionis always smaller than 111 h − pc (physical), corresponding to a maximum mesh refinement level of13 at z = 0. An identical comparison run that has four times better resolution of 29pc/h wasalso run down to z = 3 and some relevant comparisons between the two simulations are made tounderstand effects of limited resolution on our results. The simulations include a metagalactic UVbackground (Haardt & Madau 1996), and a model for self-shielding of UV radiation (Cen et al.2005). They include metallicity-dependent radiative cooling (Cen et al. 1995). Our simulations alsosolve relevant gas chemistry chains for molecular hydrogen formation (Abel et al. 1997), molecularformation on dust grains (Joung et al. 2009), and metal cooling extended down to 10 K (Dalgarno& McCray 1972). Star particles are created in cells that satisfy a set of criteria for SF proposedby Cen & Ostriker (1992). Each star particle is tagged with its initial mass, creation time, andmetallicity; star particles typically have masses of ∼ M (cid:12) .Supernova feedback from SF is modeled following Cen et al. (2005). Feedback energy andejected metal-enriched mass are distributed into 27 local gas cells centered at the star particle inquestion, weighted by the specific volume of each cell, which is to mimic the physical process ofsupernova blastwave propagation that tends to channel energy, momentum and mass into the leastdense regions (with the least resistance and cooling). The primary advantages of this supernovaenergy based feedback mechanism are three-fold. First, nature does drive winds in this way andenergy input is realistic. Second, it has only one free parameter e SN , namely, the fraction of the restmass energy of stars formed that is deposited as thermal energy on the cell scale at the location ofsupernovae. Third, the processes are treated physically, obeying their respective conservation laws(where they apply), allowing transport of metals, mass, energy and momentum to be treated self-consistently and taking into account relevant heating/cooling processes at all times. We allow theentire feedback processes to be hydrodynamically coupled to surroundings and subject to relevantphysical processes, such as cooling and heating. The total amount of explosion kinetic energy fromType II supernovae with a Chabrier initial mass function (IMF) is 6 . × − M ∗ c (where c is thespeed of light), for an amount M ∗ of star formed. Taking into account the contribution of promptType I supernovae, we use e SN = 1 × − in our simulations. Observations of local starburst galaxiesindicate that nearly all of the SF produced kinetic energy is used to power galactic superwinds (e.g., 6 –Heckman 2001). Supernova feedback is important primarily for regulating SF and for transportingenergy and metals into the intergalactic medium. The extremely inhomogeneous metal enrichmentprocess demands that both metals and energy (and momentum) are correctly modeled so that theyare transported in a physically sound (albeit still approximate at the current resolution) way.We use the following cosmological parameters that are consistent with the WMAP7-normalized(Komatsu et al. 2010) ΛCDM model: Ω M = 0 .
28, Ω b = 0 . Λ = 0 . σ = 0 . H =100 h km s − Mpc − = 70 km s − Mpc − and n = 0 .
96. These parameters are consistent with thosefrom Planck first-year data (Planck Collaboration et al. 2013) if we average Planck derived H withSN Ia and HST based H .We note that the size of the refined region, 21 × × h − Mpc , is still relatively small andthe region biased. This is, of course, designed on purpose. Because of that, however, we are notable to cover all possible environment, such as the center of a void. Also because of that, wehave avoided addressing any measures that requires a precise characterization of the abundance ofany large galaxy systems, such as the mass function or luminosity function of massive galaxies orgroups. Despite that, measures that are characterized as a function of environment/system massesshould still be valid. Our environment coverage is substantially larger than probed in, for example,Feldmann et al. (2011). In Tonnesen & Cen (2012) we show that the present simulation box (Cbox) (run to z = 0 with a lower resolution previously) spans a wide range in environment fromrich clusters to the field, and there is a substantial overlap in the field environment with anothersimulation centered on a void (V box). It is the density peaks higher than we model here (i.e., moremassive clusters of galaxies) that we fail to probe. As it should be clear later, this shortcomingshould not affect any of our conclusions, which may be appropriately extrapolated. We identify galaxies in our high resolution simulations using the HOP algorithm (Eisenstein &Hu 1999) operating on the stellar particles, which is tested to be robust and insensitive to specificchoices of concerned parameters within reasonable ranges. Satellites within a galaxy down to massof ∼ M (cid:12) are clearly identified separately in most cases. The luminosity of each stellar particlein each of the Sloan Digital Sky Survey (SDSS) five bands is computed using the GISSEL stellarsynthesis code (Bruzual & Charlot 2003), by supplying the formation time, metallicity and stellarmass. Collecting luminosity and other quantities of member stellar particles, gas cells and darkmatter particles yields the following physical parameters for each galaxy: position, velocity, totalmass, stellar mass, gas mass, mean formation time, mean stellar metallicity, mean gas metallicity,SFR, luminosities in five SDSS bands (and various colors) and others. At a spatial resolution of159pc (physical) with thousands of well resolved galaxies at z ∼ . −
6, the simulated galaxy catalogspresent an excellent (by far, the best available) tool to study galaxy formation and evolution. 7 –
When we start the analysis for this paper, the simulation has reached z = 0 .
62. For eachgalaxy at z = 0 .
62 a genealogical line is constructed from z = 0 .
62 to z = 6 by connecting galaxycatalogs at a series of redshifts. Galaxy catalogs are constructed from z = 0 .
62 to z = 1 .
40 at aredshift increment of ∆ z = 0 .
02 and from z = 1 .
40 to z = 6 at a redshift increment of ∆ z = 0 . g − r < .
55 “blue”, those with g − r = 0 . − .
65 “green” and thosewith g − r > .
65 “red”, in accord with the bimodal color distribution that we will show below andwith that of observed galaxies (e.g., Blanton et al. 2003b), where g and r are magnitudes of SDSS g and r bands.In subsequent analysis, we will examine gasdynamic processes, e.g., cold gas loss or lack of coldgas accretion, under the working hypothesis that ram-pressure stripping and gas starvation are theprimary detrimental processes to star formation. We should assume that other processes, such ashydrodynamical instabilities (e.g., RT, KH, tidal shocks, etc), may either be “lumped together”with ram-pressure stripping or play some role to enhance cold gas destruction that is initiated byram-pressure stripping. A word on tidal stripping may be instructive. It is noted that while tidalstripping would affect both stars and gas, ram-pressure operates only on the latter. As one will seelater, in some cases the stellar masses of galaxies decrease with time, which are likely due to tidaleffects. A simple argument suggests that ram-pressure effects are likely to be more far-reachingspatially and are more consistent with the environment effects becoming effective at 2-3 virial radiithan tidal stripping that we will show later. Let us take a specific example to illustrate this.Let us assume that the primary and infalling galaxies have a velocity dispersion of σ and σ ,respectively, and that they both have isothermal sphere density profiles for both dark matter andbaryons. The virial radius is proportional to its velocity dispersion in each case. Under such aconfiguration, we find that the tidal radius for the satellite galaxy at its virial radius is equal tothe virial radius of the primary galaxy. On the other hand, the ram-pressure force on the gas inthe satellite at its virial radius is already equal to the gravitational restoring by the satellite, whenthe satellite is ( σ /σ ) virial radii away from the primary galaxy. In reality, of course, the densityprofiles for dark matter and baryons are different and neither is isothermal, and the gas may displaya varying degree of non-sphericity. But the relative importance of ram-pressure and tidal strippingsis likely to remain the same for relatively diffuse gas. The relative situation is unchanged, if oneallows the gas to cool and condense. As an example, if the gas within the virial radius of the satelliteand the primary galaxies in the above example is allowed to shrink spherically by a factor of 10in radius (we will continue to assume that the velocity dispersion or rotation velocity remains flatand at the same amplitude), we find that the tidal stripping radius is now a factor of 10 smallerthan before (equal to 0.1 times the virial radius of the primary galaxy), while the new ram-pressurestripping radius is σ /σ times the new tidal stripping radius. As a third example, if the gas withinthe virial radius of the satellite galaxy in the above example is allowed to shrink spherically by afactor of 10 in radius but the gas in the primary galaxy does not shrink in size, it can be shownthat in this case the tidal stripping radius is equal to 0.1 times the virial radius of the primary 8 –galaxy, while the new ram-pressure stripping radius is now 0.1 times (sigma1/sigma2) times thevirial radius of the primary galaxy. As the last example, if the gas within the virial radius of thesatellite galaxy in the above example is allowed to shrink by a factor of 10 in radius to become adisk but the gas in the primary galaxy does not shrink in size, it can be shown that in this casethe tidal stripping radius is equal to 0.1 times the virial radius of the primary galaxy. The newram-pressure stripping radius depends on the orientation of the motion vector and the normal ofthe disk: if the motion vector is normal to the disk, the tidal stripping radius is 0.1 times σ /σ times the virial radius of the primary galaxy; if the motion vector is in the plane to the disk, thetidal stripping radius is zero. − − − l og v a r i ab l e quan t i t i e s − − − log ram pressure − − )log halo mass − − − − l og v a r i ab l e quan t i t i e s log SFR (Msun/yr)log cold gas (<10kpc) − − − − − Fig. 1.— four panels show the histories of six variables for four randomly selected red galaxies withstellar mass of ∼ M (cid:12) at z = 0 .
62: log SFR (in M (cid:12) /yr) (solid dots), log ram pressure (in Kelvincm − ) - 2 (stars), log cold gas within 10kpc (in M (cid:12) ) -10 (open circles), log cold gas within 100kpc(in M (cid:12) ) -10 (open squares), log sSFR (in Gyr − ) (down-pointing triangles) and log halo mass (inM (cid:12) ) -10 (solid diamonds); The color of symbols at any given time corresponds the color of thegalaxy at that time. The logarithm of the stellar mass at z = 0 .
62 is indicated in the upper-rightcorner in each panel. The vertical dashed line in each panel indicates the location of t q .We denote a point in time when the galaxy turns from blue to green as t g , a point in timewhen the galaxy turns from green to red as t r . Convention for time is that the Big Bang occursat t = 0. We identify a point in time, searched over the range t g − t g +1Gyr, when thederivative of SFR with respect to time, dSFR / dt, is most negative, as t q (q stands for quenching);in practice, to reduce uncertainties due to temporal fluctuations in SFR, t q is set to equal to t ( n + 1) 9 – − − − l og v a r i ab l e quan t i t i e s − − − log ram pressure − − )log halo mass − − − − − l og v a r i ab l e quan t i t i e s log SFR (Msun/yr)log cold gas (<10kpc) − − − − − − Fig. 2.— four panels show the histories of six variables for four randomly selected red galaxies withstellar mass of ∼ M (cid:12) at z = 0 .
62: log SFR (in M (cid:12) /yr) (solid dots), log ram pressure (in Kelvincm − ) - 2 (stars), log cold gas within 10kpc (in M (cid:12) ) -10 (open circles), log cold gas within 100kpc(in M (cid:12) ) -10 (open squares), log sSFR (in Gyr − ) (down-pointing triangles) and log halo mass (inM (cid:12) ) -10 (solid diamonds); The color of symbols at any given time corresponds the color of thegalaxy at that time. The logarithm of the stellar mass at z = 0 .
62 is indicated in the upper-rightcorner in each panel. The vertical dashed line in each panel indicates the location of t q .when the sliding-window difference (SFR(n + 3) − SFR(n) / (t(n + 3) − t(n)) is most negative, where t (1) , t (2) , ..., t ( n ) , ... are the times of our data outputs, as noted earlier. Galaxies at t q are collectivelycalled SFQs for star formation quenching galaxies. To demonstrate the reliability and accuracy ofidentification of t q we show in Figure 1 the histories for a set of four randomly selected red galaxiesat z = 0 .
62 of stellar mass ∼ M (cid:12) . The vertical dashed line in each panel shows t q , which is thelocation of steepest drop of SFR (solid dots). In all four cases, our method identifies the locationaccurately. Figure 2 is similar to Figure 1 but for galaxies of stellar mass ∼ M (cid:12) , where we seeour method identifies t q with a similar accuracy.Similarly, we identify a point in time in the range t g − t g +1Gyr, when the derivativeof the amount of cold gas, ( M , M , M ) within radial ranges (0 − , − , − t , t , t ), respectively. We define cold gas as gas withtemperature less than 10 K. The exponential decay time scale of SFR at t q is defined by τ q ≡ ( d ln SFR /dt ) − . The exponential decay time scale of ( M , M , M ) at ( t , t , t ) are defined 10 –by [ τ ≡ ( d ln M /dt ) − , τ ≡ ( d ln M /dt ) − , τ ≡ ( d ln M /dt ) − ]. The time interval between t q and t r is denoted as t qr , The time duration that the galaxy spends in the green valley beforeturning red is called t green . The time duration the galaxy has spent in the red sequence by z = 0 . t red .We make a needed simplification by approximating the ram-pressure, denoted as p , byp ≡ ρ (300) T (300), where ρ (300) and T (300), respectively, are the mean density of gas withtemperature ≥ K and T (300) the mean mass-weighted gas temperature within a proper radius of300kpc centered on the galaxy in question. This tradeoff is made thanks chiefly to the difficulty ofdefining precisely the motion of a galaxy relative to its ambient gas environment, where the latteroften has complex density and velocity structures, and the former has complex, generally non-spherical gas distribution geometry. In a gravitationally shock heated medium, this approximationshould be reasonably good, because the ram-pressure is approximately equal to thermal pressure inpost-shock regions. We define a point in time searched over the time interval between t q − t q +1Gyrs, when the derivative of p with respect to time is maximum as t ram , intended to serveas the point in time when ram-pressure has the steepest rise,As stated in the introduction, it is convenient to express gas cooling time that is proportionalto gas entropy to the power 3 / S / . Thus, we approximate gas starvation from large scales bythe value of environmental entropy S , defined to be the average gas entropy within a top-hatsphere of proper radius 300kpc.For convenience, frequently used symbols and their definitions are given in Table 1. The galaxy formation simulation in a cosmological setting used here includes sophisticatedphysical treatment, ultra-high resolution and a very large galaxy sample to statistically addresscosmological and astrophysical questions. While this simulation represents the current state-of-the-art in these respects, feedback from SF is still far from being treated from first principles. Thus, itis necessary that we validate the feedback prescription empirically.In Cen (2012b) we presented an examination of the damped Lyman alpha systems (DLAs) andfound that the simulations, for the first time, are able to match all observed properties of DLAs,including abundance, size, metallicity and kinematics. In particular, the metal distribution in andaround galaxies over a wide range of redshift ( z = 0 −
5) is shown to be in excellent agreementwith observations (Rafelski et al. 2012). The scales probed by DLAs range from stellar disks at lowredshift to about one half of the virial radius at high redshift. In Cen (2012a) we further show thatthe properties of O VI absorption lines at low redshift, including their abundance, Doppler-columndensity distribution, temperature range, metallicity and coincidence between O VII and O VI lines,are all in good agreement with observations (Danforth & Shull 2008; Tripp et al. 2008; Yao et al.2009). The agreement between simulations and observations with respect to O VI lines is recentlyshown to extend to the correlation between galaxies and O VI lines, the relative incidence ratio ofO VI around red to blue galaxies, the amount of oxygen mass around red and blue galaxies as well 11 –Table 1. Definitions of symbols and names symbol/name definition/meaning t g a point in time when galaxy has g − r = 0 . t r a point in time when galaxy has g − r = 0 . M amount of cold gas within a radius of 10kpc M amount of cold gas within a radius of 30kpc M amount of cold gas within a radius of 100kpc t q a point in time of quenching for SFR t a point time of quenching for M t a point time of quenching for M t a point time of quenching for M t ram a point in time of largest first derivative of ram-pressure w.r.t time τ q exponential decay time of SFR at t q τ exponential decay time of M at t τ exponential decay time of M at t τ exponential decay time of M at t ∆ M ∗ stellar mass change between t q and t r ∆ M M mass change between t q and t r ∆ M M mass change between t q and t r ∆ M M mass between t q and t r r SFRe effective radius of young stars formed within the past 100Myr T environmental temperature within physical radius of 300kpc S environmental entropy within physical radius of 300kpc p environmental pressure within physical radius of 300kpc δ environmental overdensity within comoving radius of 2 h − Mpc d/r cv distance to primary galaxy in units of virial radius of primary galaxy t qr time duration from t q to t r t green time spent in green valley t red time spent in red sequence M ch halo mass of primary galaxy M s ∗ /M c ∗ stellar mass ratio of satellite to primary galaxy
12 –as cold gas around red galaxies (Cen 2013).In addition to agreements with observations with respect to circumgalactic and intergalacticmedium, we find that our simulations are able to match the global SFR history (the Madau plot)and galaxy evolution (Cen 2011a), the luminosity function of galaxies at high (Cen 2011b) andlow redshift (Cen 2011a), and the galaxy color distribution (Cen 2011a; Tonnesen & Cen 2012),within observational uncertainties. In Cen (2011a) we show that our simulations reproduce manytrends in the global evolution of galaxies and various manifestations of the cosmic downsizingphenomenon. Specifically, our simulations show that, at any redshift, the specific star formationrate of galaxies, on average, correlates negatively with galaxy stellar mass, which seems to be theprimary physical process for driving the cosmic downsizing phenomena observed. Smoothed particlehydrodynamic (SPH) simulations and semi-analytic methods, in comparison, appear to produce apositive correlation between the specific star formation rate of galaxies and galaxy stellar mass,which is opposite to what we find (e.g., Weinmann et al. 2012). These broad agreements betweenour simulations and observations indicate that, among others, our treatment of feedback processesfrom SF, i.e., the transport of metals and energy from galaxies, from SF sites to megaparsec scale(i.e., from interstellar to intergalactic medium) are realistically modeled as a function of distance andenvironment, at least in a statistical sense, and it is meaningful to employ our simulated galaxies,circumgalactic and intergalactic medium for understanding physical processes and for confrontationswith other, independent observations.In order to determine what galaxies in our simulations to use in our subsequent analysis, wemake an empirical numerical convergence test. Top-right panel in Figure 3 shows comparisonsbetween galaxies of two simulations at z = 3 with different resolutions for the luminosity functionsin rest-frame g and r bands. The fiducial simulation has a resolution of 114pc/h and an identicalcomparison run has four times better resolution of 29pc/h. We are not able to make comparisonsat redshift substantially lower than z = 3 at this time. In any case, we expect that the comparisonat z = 3 is a more stringent test, because the resolution effect is likely more severe at higherredshift than at lower redshift in a hierarchical growth model where galaxies become increasinglylarger with time. The comparisons are best done statistically, because not all individual galaxiescan be identified at a one-to-one basis due to resolution-dependent star formation and merginghistories. Comparisons with respect to other measures, such as stellar mass function, SFR, etc,give comparable convergence. Based on results shown, we decide to place a lower stellar masslimit of 10 . M (cid:12) , which is more than 75% complete for almost all relevant quantities, to the extentthat we are able to make statistical comparisons between these two runs with respect to the globalproperties of galaxies (stellar mass, luminosity, SFR, sSFR, etc).In terms of checking the validity and applicability of the simulations, we also make comparisonsfor the galaxy cumulative mass function at z = 1 with observations in the top-left panel of Figure 3.We see that the simulated galaxies have a higher abundance than observed by a factor of 4 − g band galaxy luminosity 13 – − − − − * (Msun) l og n ( > M * ) ( h M p c − ) CMF@z=1van der Burg 2013 − − − − − − − − − − − g & M r l og n ( < M ) ( h M p c − ) g LF@z=3, 114pc/hg LF@z=3, 29pc/hr LF@z=3, 114pc/hr LF@z=3, 29pc/h − − − − − − − − − − − g & M B l og n ( < M ) ( h M p c − ) g band LF@z=0.62Ramos et al 2011b band LF@z=1Faber et al 2007 − − e (kpc) P D F Fig. 3.— Top-left panel: galaxy stellar mass function (SMF) at z = 1 from the fiducial run withspatial resolution of 114pc/h (solid curve) and observations by van der Burg et al. (2013) at z = 1;the observational points have been shifted upward by 0 . ≥ . M (cid:12) that we apply to simulated galaxies. Top-right panel: galaxyluminosity function (LF) in SDSS g band (green) and r band (red) at z = 3 from the fiducialrun (solid curves) and the higher resolution run (dashed curves) for galaxies with M ∗ ≥ . M (cid:12) .Bottom-left panel: galaxy luminosity function (LF) in restframe g band (green) at z = 0 .
62 andB band (blue) at z = 1 for galaxies with M ∗ ≥ . M (cid:12) , to compared from the correspondingobservations of the same color dashed curves at z = 0 . z = 1(Faber et al.2007), respectively. The observational points have been shifted upward by 0 . M ∗ ≥ . M (cid:12) at z = 3 from the fiducial (black)and the higher resolution run (red). The vertical lines indicate the effective resolutions of the twosimulations, which are found to be 1 .
84 times the maximum refined resolution.function at z = 0 .
62 and B band luminosity function at z = 1 (lower-left panel of Figure 3). Thedifference between simulation and observations is also seen to increase with decreasing redshift,as expected. We note that at the high luminosity end dust effects may become very important,causing the apparent discrepancies between simulations and observations larger than they actuallyare. Overall, our simulation serves our purposes well for two reasons. First, it contains massivegroup/cluster systems as well as field regions that allow us to probe environment effects withenough leverage/range. Second, the faint end slope of the galaxy population matches observationswell, albeit with a larger overall amplitude; as a result, relative, comparative statements that we 14 –will make across the stellar mass range 10 . and 10 M (cid:12) is approximately valid even though theabsolute number of galaxies may be off.As will become clear later, during quenching by ram-pressure stripping and gas starvation, starformation often continues in the central 10kpc regions of galaxies. Therefore, it is useful to have anestimate of the effective resolutions of simulations empirically as follows. Let us denote the actualresolution, given the maximum refinement resolution of ∆ l , as ∆ x = C ∆ l , where C is a contantthat is expected to be larger than but of order unity for a mesh code. We then assume that thecomputed effective stellar radius (enclosing 50% of stellar mass) of any galaxy is r = r + (∆ x ) = r + ( C ∆ l ) , (1)where r c is the actual effective radius from simulation and r t is the true effective radius if one hadinfinite resolution. We have two equations of Eq 1 for the two simulations with different resolutions(∆ l = 114pc/h and ∆ l = 29pc/h) and two unknowns (constant C and vector r t ). We solve for C by requiring the distributions of r t derived from the two simulations have the closest agreement; C = 1 .
84 is found. The bottom-right panel of Figure 3 shows the distributions of derived r t forall galaxies with stellar mass ≥ . M (cid:12) from the two simulations with C = 1 .
84. The effectiveresolutions ( C ∆ l ) for the two runs, 312pc and 78pc, are indicated by the vertical dashed linescorresponding to the histograms with the same colors. As we will show later, the central gas within10kpc is largely retained by the galaxy that is being subject to ram-pressure stripping. Clearly,our resolution is adequate to resolve gas distribution, where it is demanded to address ram-pressurestripping physics. On the other hand, consistent with other panels of Figure 3, we see that atthe stellar mass cut of 10 . M (cid:12) , while the majority of galaxies are effectively resolved with respectto their stellar distribution, a significant fraction of galaxies in our selected sample with stellarmass close to 10 . M (cid:12) is under-resolved in the central regions. We expect that, although oursimulation may overestimate the ram-pressure stripping effects for the most inner regions for someof the smaller galaxies, the numerical effect on overall ram-pressure stripping process due to thislimitation is likely small. This is because the amount of gas within our resolution limit (313pc) issmall compared to the total mass of gas within (say) 10kpc, within which, we will show later, gas islittle affected by ram-pressure stripping. In other words, an under-resolution should not materiallyimpact the overall the amount of ram-pressure stripped gas, since the gas in the central region muchlarger than our resolution is not stripped anyway. z = 0 . g − r color distributions of simulated galaxies in threestellar mass ranges, 3 × − × M (cid:12) (black), 1 × − × M (cid:12) (cyan) and > × M (cid:12) (magenta), at z = 0 .
62. The g-r color distributions show clear bimodalities for all three subsetsof galaxies, with the red peak becoming more prominent for less luminous galaxies at z = 0 . z ∼ .
62) are perhaps less complete than at low redshift, andidentification of low mass (and especially low surface brightness) galaxies, in particular those thatare satellite galaxies and red, may be challenged at present (e.g., Knobel et al. 2013). Our mainpurpose is to make a comparative study of galaxies of different types in the simulation and tounderstand how blue galaxies turn red.It is intriguing to note that there is no lack of red dwarf galaxies. While a direct comparisonto observations with respect to abundant red dwarf galaxies can not be made at z = 0 .
62, futureobservations may be able to check this. Since our simulation does not include AGN mechanicalfeedback, this suggests that the bimodal nature of galaxy colors does not necessarily require AGNfeedback for galaxies in the mass ranges examined. This finding is in agreement with Feldmannet al. (2011), who find that AGN feedback is not an essential ingredient for producing quiescent, redelliptical galaxies in galaxy groups. While SF feedback is included in our simulation, our subsequentanalysis shows that environmental effects play the dominant role in driving galaxy color evolutionand consequently color bimodality. Our results do not, however, exclude the possibility that AGNfeedback may play an important role in regulating larger, central galaxies, such as cD galaxies atthe centers of rich clusters of galaxies, for which we do not have a sufficient sample to make astatistical statement. Our earlier comparison between simulated luminosity functions of galaxies at z = 0 and SDSS observations indicates that some additional feedback, likely in the way of AGN,may be required to suppress star formation in the most massive galaxies (Cen 2011a). − r P D F ( g − r) × − × Msun1 × − × Msun>3 × Msun
Fig. 4.— shows g − r color distributions of simulated galaxies in three stellar mass ranges, 3 × − × M (cid:12) (black), 1 × − × M (cid:12) (cyan) and > × M (cid:12) (magenta), at z = 0 .
62. 16 –
3. Results
Most of our results shown are presented through a variety of comparisons of the dependenciesof galaxies of different types on a set of environmental variables, to learn how galaxies change color.We organize our analysis in an approximately chronological order. In § t q , followed by the ensuing period of gas starvation in hot environmentin § § § z = 0 .
62 in § @t q (keV cm ) l og (cid:111) q ( M y r) @t q (K cm − ) l og (cid:111) q ( M y r) − (cid:98) @t q l og (cid:111) q ( M y r) vc @t q l og (cid:111) q ( M y r) red: satellitesblack: centrals M * = 3 × M * = 10 M * = 10 Fig. 5.— shows τ q , the exponential decay time of SFR, against four environmental variables at t q : ram-pressure p on 300kpc proper scale, environmental entropy S on 300kpc proper scale, distance to primary galaxy d/r cv in units of the primary galaxy’s virial radius and environmental overdensity δ on 2 h − Mpc comoving scale. Themagenta solid dots with dispersions are the means.
Figure 5 shows the quenching time scale τ q (star formation rate exponential decay time) againstfour environmental variables at the quenching time t q : ram-pressure p , environmental entropy S , distance to primary galaxy d/r cv and environmental overdensity δ . It is useful to make clearsome nomenclature here. We have used the distance to the primary galaxy, d/r cv , as an environment 17 –variable, which runs from zero to values significant above unity. This is merely saying that anygalaxy (except the most massive galaxy in the simulation) can find a larger galaxy at some distance,not necessarily at d/r cv ≤
1. The definition of “satellite galaxies” is reserved only for those galaxieswith d/r cv ≤
1, shown clearly as red circles in the low-left panel of Figure 5. The black circles,labeled as “centrals” are galaxies with d/r cv >
1, i.e., those that are not not “satellite galaxies”.The observation that galaxies are being quenched at all radii - d/r cv > d/r cv <
1- indicates that the most likey physical mechanism for the onset of quenching is ram-pressure.Tidal stripping is not expected to be effective at removing gas (or stars) at d/r cv > § τ q decreases with increasing p is self-consistent with ram-pressure beingresponsible for the onset of quenching. The outcome that τ q only very weakly anti-correlates with p indicates that the onset of quenching is some “threshold” event, which presumably occurs whenthe ram-pressure exceeds gravitational restoring force (i.e., the threshold), thus strongly re-enforcingthe observation that ram-pressure is largely responsible for the onset of quenching. A “threshold”type mechanism fits nicely with the fact that the dispersion of τ q at a given p is substantiallylarger than the correlation trend, because galaxies that cross the “threshold” are expected to dependon very inhomogeneous internal properties among galaxies (see Figure 6 below). The weak anti-correlation between τ q and δ stems from a broad positive correlation between p and δ . The factthat there is no discernible correlation between τ q and S indicates that the onset of quenching is not initiated by gas starvation.The most noticeable contrast to the weak trends noted above is the difference between satellitegalaxies (at d/r cv <
1) and central galaxies (at d/r cv > τ q of the former is lower thanthat of the latter by a factor of ∼
2. This is naturally explained as follows. First, at d/r cv < d/r cv > d/r cv > d/r cv < τ q as a function of the stellar surface density Σ e within the effective stellar radius r e .We see a significant positive correlation between τ q and Σ e in the sense that it takes longer to ram-pressure-remove cold gas with higher central surface density (hence higher gravitational restoringforce) galaxies. While this positive correlation between τ q and Σ e is consistent with observationalindications (e.g., Cheung et al. 2012), the underlying physical origin is in a sense subtle. Since ram-pressure stripping is a “threshold” event, as noted earlier, when ram-pressure force just exceeds theinternal gravitational restoring force, one would have expected that a high surface density would 18 –yield a shorter dynamic time hence a shorter τ q . This is in fact an incorrect interpretation. Rather,the gas in the central regions where Σ e is measured is immune to ram-pressure stripping in the vastmajority of cases (see Figure (8) below). Instead, a higher Σ e translates, on average, to a largerscale where gas is removed, which has a longer dynamic time hence a longer τ q . (cid:89) e (Msun/kpc ) l og t q r ( M y r) × l og (cid:111) q ( M y r) red: satellitesblack: centrals Fig. 6.—
Top panel: the exponential decay time scale of SFR, τ q , as a function of the stellar surface density Σ e within the stellar effective radius r e . Bottom panel: the time interval between onset of quenching and the time thegalaxy turns red, t qr as a function of Σ e . The magenta dots are the averages at a given x-axis value. Taken together, we conclude that, while a high ram-pressure provides the conditions for ram-pressure stripping to take effect, the effectiveness or timescale for gas removal by ram-pressurestripping also depend on the internal structure of galaxies. It is very interesting to note that, unlikebetween t q and Σ e , t qr (the time interval between the onset of quenching t q and the time whenthe galaxy turns red) and Σ e shown in the bottom-panel of Figure 6, if anything, is weakly anti-correlated. We attribute this outcome to the phenomenon that galaxies with higher central surface 19 –density have a shorter time scale for consuming the existing cold gas hence, once the overall coldgas reservoir is removed. This explanation will be elaborated more later. − − − − − q − t ram (Gyr) P D F M * >3 × M * >3 × Fig. 7.— the histograms of t q − t ram for red galaxies at z = 0 .
62, where t ram is the point in time when the derivativeof p with respect to time is maximum, and t q is the onset of quenching time for SFR. The vertical thick linesshow the medians of the corresponding histograms of the same colors, and the vertical thin lines for each color arefor 25% and 75% percentiles. We have made the case above that ram-pressure stripping is primarily responsible for the onsetof quenching process based on evidence on the dependence of exponential decay time of SFR at theonset of quenching on environment variables. We now make a direct comparison between t q and t ram . Figure (7) shows the histograms of t q − t ram . We see that the time difference between the twois centered around zero, indicating a casual connection between the onset of SFR quenching andthe rapid rise of ram-pressure. The width of the distribution of a few hundred Myrs reflects thefact that the exact strength of ram-pressure stripping required to dislodge the gas varies greatly,depending on many variables as discussed above. This is as yet the strongest supporting evidencefor ram-pressure stripping being responsible for the onset of quenching, especially considering theconjunctional evidence that the onset of quenching could occur outside the virial radius of a largerneighboring galaxy where tidal stripping is expected to be less effective and yet ram-pressure isexpected to become important.Evidence so far supports the notion that ram-pressure stripping is the initial driver for thedecline of SFR in galaxies that are en route to the red sequence. The immediate question is then:What region in galaxies does ram-pressure stripping affect? To answer this question, we need tocompare the amount of cold gas available at t q with the amount of star formation that ocurrssubsequently. We compute the following ratios: the ratios of the amount of stars formed during thetime interval from t q to the time the galaxy turns red ( t r ) to the difference between the amount of 20 – − − * >3x10 P D F − − − * >3x10 log − (cid:54) M * / (cid:54) M X P D F X=10X=30X=100
Fig. 8.— shows the distribution of − ∆ M ∗ / ∆ M X at three radial ranges, X = (10 , , M ∗ is the amount ofstars formed during the time interval from the onset of quenching t q to the time the galaxy turns red t r , and ∆ M X is the difference of the amount of cold ( T < K) gas within a radius X kpc between t r and t q . The vertical dashedlines show the medians of the corresponding histograms of the same colors. cold ( T < K) gas at t q and t r , denoted as ( − ∆ M ∗ / ∆ M , − ∆ M ∗ / ∆ M , − ∆ M ∗ / ∆ M ) withinthree radii (10 , , M ∗ could be negative. We set a floor value to the above ratios at 10 − . Figure (8)shows the distributions of − ∆ M ∗ / ∆ M X , where X = (10 , , r ≤ − ∆ M ∗ / ∆ M >
1, typicallyin the range of 2 −
10, with the vast majority of cases at >
1. For a larger radius r ≤ − ∆ M ∗ / ∆ M ∼ <
1, while for a still larger radius r ≤ − ∆ M ∗ / ∆ M ≤ . , .
4) for galaxies of stellar mass > (3 × , × ) M (cid:12) , respectively. This is unambiguous evidence that ram-pressure strippingremoves the majority of cold gas on scales ≥ r < r > τ q seen in Figure 5. We thus conclude 21 – − − e (SFR) (kpc) P D F M * =3 × M * >3 × − − − e (SFR)/dlnSFR (kpc) P D F red M * >3 × M * >3 × Fig. 9.—
Left panel shows the distribution of the effective radius of stars formed in the last 100Myrs, r e (SFR),at t q for galaxies in two stellar mass ranges. The vertical dot-dashed line indicates the effective resolution of thesimulation, taken from the bottom-right panel in Figure 3. Right panel shows the distribution of the ratio of thedecline of r e (SFR) with respect to the decline of SFR, d r e (SFR) /d ln SFR at t q . In each panel, we differentiatebetween galaxies in three separate stellar mass ranges. The vertical lines show the medians of the correspondinghistograms of the same colors. that there is continued nuclear SF in the quenching phase. Feldmann et al. (2011), based on asmaller sample of simulated galaxies that form a group of galaxies with a spatial resolution of 300pc(compared to 160pc here), find that in situ star formation is responsible for consuming a substantialfraction of the residual gas on small scales after gas accretion is stopped subsequent to the infall,consistent with our results. This outside-in ram-pressure stripping picture and continuous SF inthe inner region that emerges from the above analysis has important implications and observableconsequences, consistent with the latest observations (e.g., Gavazzi et al. 2013).We quantify how centrally concentrated the star formation is at the outset of SF quenchingin Figure (9), in part to assess our ability to resolve SF during the quenching phase. The leftpanel shows the distribution of the effective radius of stars formed in the last 100Myrs prior to t q ,denoted as r SFR e , for galaxies in two stellar mass ranges. The right panel shows the distributionof the ratio of the decline of r SFR e with respect to the decline of SFR, dr SFR e /d ln SFR at t q . Itis evident from Figure (9) that more massive galaxies tend to have larger r SFR e , as expected. Itis also evident that the recent formation for the vast majority of galaxies occurs within a radiusof a few kiloparsecs. It is noted that ongoing SF in a significant fraction of galaxies with stellarmasses ≤ × M (cid:12) is under-resolved, as indicated by the vertical dot-dashed line in the leftpanel. However, none of our subsequent conclusions would be much altered by this numericaleffect, because (1) all of our conclusions appear to be universal across the stellar mass ranges and(2) the inner region of 10kpc is not much affected by ram-pressure stripping anyway (thus under-resolving a small central fraction within 10kpc does not affect the overall ram-pressure strippingeffects). What is interesting is that more than 50% of galaxies in both stellar mass ranges havenegative values of dr SFR e /d ln SFR at t q , indicating that, when the SFR decreases in the quenching 22 –phase, star formation proceeds at progressively larger radii in the central region. This result, whilemaybe somewhat counter-intuitive, is physically understandable. We attribute this inside-out starformation picture to the star formation rate surface density being a superlinear function of gassurface density in the Kennicutt-Schmidt (Schmidt 1959; Kennicutt 1998) law. The picture goesas follows: when gas supply from large scales ( ∼ ≤ ≥ The previous subsection details some of the effects on galaxies being quenched due to gasremoval by ram-pressure stripping (in conjunction with other hydrodynamical processes) along withconsumption by concurrent SF. Our attention is now turned to the subsequent evolution. Figure 10plots t qr against four environmental variables at t q . From all panels we consistently see the expectedtrends: the time interval t qr from onset of quenching t q to turning red t r , on average, decreaseswith increasing environmental pressure, increasing environmental entropy, increasing environmentaloverdensity and decreasing distance to the primary galaxy. While there is a discernible differencein t qr between satellite galaxies and central galaxies, the difference is substantially smaller thanthat in the initial exponential decay time scale of SFR τ q (see Figure 5). This observation makesit clear that the onset of quenching initiated by ram-pressure stripping does not determine theoverall duration of quenching. Since all the environment variables used tend to broadly correlatewith one another - higher density regions tend to have higher temperatures, higher gas entropyand higher pressure - it is not surprising that we see t qr are correlated with all of them in theexpected sense. Earlier we have shown that t qr is weakly anti-correlated with the stellar surfacedensity at r e , Σ e (see bottom-panel of Figure 6). This suggests that the overall duration from onsetof quenching to turning red is not a matter of a galaxy’s ability to hold on to its existing cold gasbut rather the extent of the external gas supply condition, i.e., environment. This hypothesis issignificantly affirmed by noticing that the strongest anti-correlation is found between t qr and S ,among all environment variables examined. Thus, we conclude, given available evidence, that theeventual “push” of galaxies into the red sequence is not as a spectacular event as the initial onsetof quenching that is triggered by a cutoff of large-scale gas supply due to ram-pressure stripping,and is essentially the process of gas starvation, when the galaxy has entered a low cold gas densityand/or high temperature and/or high velocity dispersion environment.We present distributions of t qr in Figure 11. The top-left panel shows the distribution of t qr for satellite galaxies (those with d/r ch ≤ @t q (keV cm ) l og t q r ( M y r) @t q (K cm − ) l og t q r ( M y r) M * = 3 × M * = 10 M * = 10 − (cid:98) @t q l og t q r ( M y r) red: satellitesblack: centrals1 2 3 4 52.533.5 d/r vc @t q l og t q r ( M y r) Fig. 10.— shows t qr (time interval from the onset of quenching to the time the galaxy turns red) against fourenvironmental variables at t q : ram-pressure p on 300kpc proper scale, environmental entropy S on 300kpcproper scale, distance to primary galaxy d/r cv in units of the primary galaxy’s virial radius and environmentaloverdensity δ on 2 h − Mpc comoving scale. The red dash line in the upper-right panel is intended to indicate avisually noticeable trend. Red circles are satellite galaxies at t q , i.e., within the virial radius of a larger galaxy, andblack circles are for non-satellite galaxies. The size of each circle indicates the stellar mass of a galaxy, as shown inthe legend in the lower-left panel. M ch = 10 − M (cid:12) (black), M ch = 10 − M (cid:12) (green), M ch > M (cid:12) (red); the mediansof the distributions are (1.2,1.3,1.2)Gyr, respectively. The top-right panel shows the distributionof t qr for satellite galaxies grouped into three ranges of the ratio of satellite to cental stellar mass: M s ∗ /M c ∗ = 0 . − M s ∗ /M c ∗ = 0 . − . M s ∗ /M c ∗ = 0 . − .
01 (red); the mediansof the distributions of the three groups are nearly identical at ∼ . t qr for primary galaxies (those with d/r ch > M ch = 10 − M (cid:12) (black), and M ch = 10 − M (cid:12) (green). We see that the medians ofthe distributions are 1 . − − qr (Gyr) P D F satellites M hc =10 − M hc =10 − M hc >10 − − qr (Gyr) P D F satellites M *s /M *c =0.1 − *s /M *c =10 − − − M *s /M *c =10 − − − − − qr (Gyr) P D F centrals M hc =10 − M hc =10 − − − qr (Gyr) P D F all satellitesall centralsall galaxies Fig. 11.—
Top-left panel: shows the distribution of t qr for satellite galaxies at z = 0 .
62, separated into three primaryhalo mass ranges: M ch = 10 − M (cid:12) (black), M ch = 10 − M (cid:12) (green), M ch > M (cid:12) (red). Top-rightpanel: shows the distribution of t qr for satellite galaxies at z = 0 .
62, separated into three ranges of satellite stellarmass to primary stellar mass ratio: M s ∗ /M c ∗ = 0 . − M s ∗ /M c ∗ = 0 . − . M s ∗ /M c ∗ = 0 . − . t qr for primary galaxies at z = 0 .
62, separated into three primaryhalo mass ranges: M ch = 10 − M (cid:12) (black), M ch = 10 − M (cid:12) (green), M ch > M (cid:12) (red). The threevertical dashed lines of order (thin, thick, thin) are the (25%, 50%, 75%) percentiles for the histograms of the samecolor. Bottom right panel: shows the distribution of t qr for all satellite galaxies (blue), all primary galaxies (red)and all galaxies (black) at z = 0 .
62. An eye-balling lognormal fit is shown as the magenta line (see Eq 2).
Figure 12 shows the distributions of d/r cv at t q for satellite (red) and central (black) red galaxiesat z = 0 .
62. While it is not a surprise that the vast majority of the satellite galaxies at z = 0 . d/r cv ≤ t q , it is evident that the same appearsto be true for the central galaxies at z = 0 .
62. This observation supports the picture that bothsatellite and central red galaxies at z = 0 .
62 have been subject to similar environment effects thatturn them red. It is noted again that this statement that red central galaxies have been subjectto similar processes as the red satellite galaxies has been quantitatively confirmed in Figure 11. 25 – vc @t q P D F M * > 3 × satellites at z=0.62M * > 3 × centrals at z=0.62 Fig. 12.— shows the distribution of the relative distance d/r cv of progenitors at t q of red galaxies at z = 0 .
62 fortwo subsets of galaxies: the red histogram for those that are within the virial radius of a larger galaxy (i.e., satellitegalaxies at z = 0 .
62) and the black histogram for those that are not within the virial radius of a larger galaxy at z = 0 .
62. The thick blue vertical dashed lines are 50% percentiles for all galaxies being quenched and the thin bluevertical dashed lines are 25% and 75% percentiles.
The suggestion by Wetzel et al. (2013b) that some central galaxies are ejected satellite galaxies isconsistent with our findings here. Our study thus clearly indicates that one should not confusered central galaxies with their being quenched by processes other than environment. In fact, allavailable evidence suggests that it is environment quenching that plays the dominant role for thevast majority of galaxies that turn red, whether they become satellite galaxies at z = 0 .
62 or not.Feldmann et al. (2011), using a much smaller sample of simulated galaxies that form a group ofgalaxies, find that quenching of gas accretion starts at a few virial radii from the group center,in good agreement with our results. It is seen in Figure 12 that only about 20% of the onset ofgalaxy quenching occurs as satellites, i.e., within the virial radius of a larger galaxy, consistent withconclusion derived by others (e.g., van den Bosch et al. 2008).In the bottom-right panel of Figure 11 we provide an approximate fit to the distribution of t qr for all quenched galaxies normalized to galaxies at z = 0 .
62 as f (log t qr ) = 12 log t med √ π exp (cid:2) − (log t qr / log t med − / (cid:3) , (2)where t qr and t med are in Gyr and log t med = 0 . − . × log((1 + z ) / . t med =0 . − . × log(1 + z ) / .
62 dependence on z is merely an estimate of the time scale, had it scaled 26 –with redshift proportional to the dynamical time of the universe. One is cautioned not to apply thisliterally. Nevertheless, it is likely that the median quenching time at lower redshift is longer than ∼ . z = 0 .
62, perhaps in the range of 2 − z = 0, is consistent with theoretical interpretation of observational data insemi-analytic modeling or N-Body simulations (e.g., Taranu et al. 2012; Wetzel et al. 2013a).In semi-analytic modeling (e.g., Kimm et al. 2009), the quenching time is often taken to be adelta function. In other words, the satellite quenching process is assumed to be uniform, independentof the internal and external properties of the satellites. Our simulation results (see Eq 2) indicatethat such a simplistic approach is not well motivated physically. We suggest that, if a spread inquenching time is introduced in the semi-analytic modeling, an improvement on the agreementbetween predictions based on semi-analytic modeling and observations may result in.In summary, we find that, within the environmental sphere of influence, galaxies are discon-nected with their large-scale cold gas supply by ram-pressure stripping, and subsequently lack of gascooling and/or accretion in high velocity environment ensures a prolonged period of gas starvationthat ultimately turns galaxies red. This applies to satellite galaxies as well as the vast majorityof “apparent” central red galaxies. The dominance of environment quenching that is found in abinitio cosmological simulations here is in accord with observations (e.g., van den Bosch et al. 2008;Peng et al. 2012; Kovac et al. 2013). On its way to the red sequence, a galaxy has to pass through the green valley. Do all galaxiesin the green valley migrate to the red sequence? We examine the entire population of green galaxiesin the redshift range z = 1 − .
5. Tracing these green galaxies to z = 0 .
62, we find that for galaxieswith stellar masses greater than (10 . , 10 , 10 . ) M (cid:12) , respectively, (40%, 40%, 48%) of galaxies inthe green valley at z = 1 − .
5, do not become red galaxies by z = 0 .
62. While this is an importantprediction of our simulations, we do not provide more information on how one might tell apartthese two different population of galaxies in the green valley, except to point out that attempts toidentify galaxies in the green valley as progenitors of red galaxies may generate some confusion.We examine the distributions (not shown) of the time that red galaxies spent in the green valley, t green , en route to the red sequence. The trends with respect to M h and M s ∗ /M c ∗ seen are similarto those seen in Figure 11. No significant differentiation among halo masses of central galaxies isvisible, once again supportive of environment quenching. Overall, one may summarize the resultsin three points. First, t green is almost universal, independent of being satellites or not, the mass,or the ratio of masses. Second, the range t green = 0 . ± . t green ∼ . t qr = 1 . − . − − * (Msun) g − r t(green)=200Myrst(green)=500Myrs Fig. 13.— shows the evolutionary tracks of 30 semi-randomly selected galaxies on the stellar mass M ∗ -g-r colorplane. The 30 galaxies are selected to be clustered around three masses, M ∗ = (10 . , . , ) M (cid:12) . Each trackhas a circle attached at the end of the green period to indicate the time spent in the green valley. We shall call thisdiagram “skyrockets” diagram of galaxy color migration. Figure 13 shows the color-stellar mass diagram for 30 semi-randomly selected red galaxies. It isstriking that the color evolution in the green valley and red sequence is mostly vertical, i.e., notaccompanied by significant change in stellar mass. This means that the stellar mass growth of mostgalaxies must occur in the blue cloud. One can see easily that the blue tracks are mostly movingfrom lower left to upper right with time for g − r ≤ .
3, indicating that galaxies grow when in the bluecloud. In the blue cloud it is seen that there are occasional horizontal tracks, representing mergersthat maintain overall color. These are mergers that do not result in red galaxies. The examplesof these include the two most massive galaxies in the plot with final stellar masses of ∼ . M (cid:12) ,where there is a major binary merger of (10 . + 10 . ) M (cid:12) at g − r = 0 .
26. There are also caseswhere the tracks temporarily go from north-west to south-east, indicating significant/major mergersthat trigger starbursts that render the remnant galaxies bluer. This anecdotal evidence that galaxiesdo not significantly grow mass in the red sequence will be confirmed below quantitatively. Feldmannet al. (2011), using a small sample of simulated galaxies that form a group of galaxies, find thatmergers and significant mass growth in galaxies occur, prior to their entering groupd environment,consistent with the findings here. Thus, this “skyrockets” diagram of color-stellar mass evolution inFigure 13 turns out to be a fair representation of typical tracks of galaxies that become red galaxies.We address the stellar mass growth of red galaxies quantitatively in two different ways. Theleft panel of Figure 14 shows the histogram of the ratio of stellar mass of red galaxies at z = 0 . t q . We see that the overall stellar massgrowth of red galaxies since the onset of quenching is relatively moderate, with the vast majority of 28 – * /M *q P D F >3 × >3 × − − − − * (Msun) l og n ( > M * ) ( h M p c − ) CMF of red galaxies z=0.62CMF of red galaxies z=0.86
Fig. 14.—
Left panel: the histogram of the ratio of stellar mass of red galaxies at z = 0 .
62 to their progenitor’sstellar mass at the onset of quenching t q , for two stellar mass ranges of the red galaxies at z = 0 .
62. Right panel:cumulative stellar mass functions of red galaxies at z = 0 .
62 (blue) and z = 1 (magenta). For the red galaxies at z = 0 .
62 we find that the median value of t q corresponds to redshift z ∼ .
86, thus the choice of z = 0 . galaxies gaining less than 30% of their stellar mass during this period, consistent with observations(e.g., Peng et al. 2010, 2012). There is a non-negligible fraction of galaxies that experience a declineof stellar mass, due to tidal interactions and collisions. There is 5 −
10% of red galaxies that gainmore than ≥
40% of their stellar mass during this period, possibly due to mergers and accretionof satellite galaxies. We do not address red galaxies more massive than 10 M (cid:12) because of lackof a statistically significant red sample. Since these larger galaxies tend to reside at the centers ofgroups and clusters, there is a larger probability that AGN feedback may play a significant rolein them. Empirical evidence suggests that radio jets get extinguished in the near vicinity of thecentral galaxies in groups/clusters (e.g., McNamara & Nulsen 2007), in sharp contrast to AGNs inisolated galaxies where jets, seen as large radio lobes, appear to deposit most of their energy onscales much larger than the star formation regions. Thus, AGN feedback in the central massivegalaxies in clusters/groups may be energetically important to have a major effect on gas coolingand star formation in them (e.g., Omma & Binney 2004). Thus, our neglect of AGN feedback inthe simulation cautions us to not draw any definitive conclusion with respect to this special classof galaxies at this time.The stellar mass growth of individual red galaxies shown in the left panel of Figure 14 containsvery useful information. However, it does not address a related but separate question: How doesthe stellar mass function of red galaxies evolve with redshift? We address this question here. Wecompute the cumulative stellar mass function of red galaxies at z = 0 .
62 and z = 0 .
86, separately,and show them in the right panel of Figure 14. We see that for red galaxies with stellar massesgreater than ∼ × M (cid:12) , when matched in abundance, the stellar masses grow a factor of ∼ . z = 0 .
86 to z = 0 .
62, much larger than 10% (for about 75% of galaxies) seen in the left panelof Figure 14. We refrain from making a direct comparison to observations in this case, because our 29 –limited simulation volume is highly biased with respect to the massive end of the mass function.We strict ourselves to a comparative analysis of galaxies in our simulation volume and ask thequestion of how red galaxies in our simulation volume grow with time. The most important pointto note is that this apparent growth of stellar mass of red galaxies based on abundance matchingcould not be due to growth of individual red galaxies in the red sequence, since the actual stellarmass increase since the onset of quenching is moderate, ≤
10% typically, seen in the left panel ofFigure 14. Physically, this suggests that dry mergers do not play a major role in the “apparent”stellar mass growth of red galaxies, consistent with observations (e.g., Pozzetti et al. 2007). Rather,galaxies grow their stellar mass when they are still in the blue cloud, illustrated in Figure 13.A physical picture of galaxy color migration emerges based on our results.
The migrationfrom the blue cloud to the red sequence proceeds in a staggered fashion: stellar masses of individualgalaxies continuously grow, predominantly in the blue cloud, and blue galaxies over the entire massrange continuously migrate into the red sequence over time.
Galaxies migrate from the blue cloudto the red sequence almost vertically in the usual color-magnitude diagram (see Figure 13). Forsimplicity we will call this type of color migration “Vertical Tracks”, which correspond most closelyto “B tracks” proposed by Faber et al. (2007), with the growth since the onset of quenching beingmoderate ( ≤ The vertical tracks found have many implications on observables. The first question one asksis this: if galaxies follow the vertical tracks, is the galaxy age-mass relation consistent with ob-servations? We address this question in this subsection. Figure 15 shows a scatter plot of redgalaxies in the stellar mass M ∗ -mean galaxy formation time t f plane at z = 0 .
62 (top) and z = 1(bottom), where t f is stellar formation time, not lookback time. The red galaxies are subdividedinto two groups: centrals (black circles) and satellites (red circles). For the purpose of comparisonto observations, we only show galaxies with high surface brightness of µ B <
23 mag arcsec − (e.g.,Impey & Bothun 1997). Several interesting results can be learned.First, no systematic difference between satellite and central galaxies is visible, supportingearlier findings that there is no appreciable differences between satellites and centrals with respectto duration from quenching to turning red t qr (Figure 11). Second, at any given redshift, thebrightest red galaxies are relatively “old” (but not necessarily the oldest), of ages of several billionyears (age = t H − t f and t H = (7 . , . z = (0 . , . − . M (cid:12) red galaxies have a nearly uniform mean age; the agespread at a given stellar mass of ∼ . − . M (cid:12) ;we see that the age difference between the two ends of the mass range is ∼ . . z = 0 .
62 and z = 1, suggesting a steepening with decreasing redshift of the agedifference between galaxies of different masses in the red sequence. Demarco et al. (2010) find anage difference between the faint and bright ends of red sequence galaxies of ∼ z = 0 .
84, in 30 – f (high mass)=3.2 Gyr l og t f ( G y r) red galaxies@z=0.62 centralssatellitesmean
62 (top) and z = 1 (bottom), where t f is stellar formation time, not lookback time. The red galaxies aresubdivided into two groups: centrals (black circles) and satellites (red circles). For the purpose of comparison toobservations, we only show galaxies with high surface brightness of <
23B mag arcsec − (e.g., Impey & Bothun1997). The green horizontal dashed lines indicate the mean formation redshift of the most luminous red galaxies atthe two redshifts. The magenta dots are the averages of t f at the stellar mass bins. excellent agreement with our results. The physical origin for the steepening with decreasing redshiftof the age difference between galaxies of different masses in the red sequence is traceable to thesteepening of specific SFR with stellar mass with decreasing redshift that is, in a fundamental way,related to the cosmic downsizing phenomenon (Cen 2011a).It is interesting to note that, in Figure 15, scatters notwithstanding, there appears to be acritical stellar mass of ∼ . − . M (cid:12) , above which the age (or formation time) of red galaxiesflattens out to a constant value. At least for the redshift range that we have examined, z = 0 . − ∼ . M (cid:12) discovered by Kauffmann et al. (2003) at low redshift,which appears to demarcate a number of interesting trends in galaxy properties. This physicalorigin of this mass is unclear and deferred to a future study.Given the “vertical tracks”, i.e., lack of significant stellar mass growth subsequently to quench- 31 – l og t f ( G y r) red blue@z=0.80 − − red blue@z=0.80 − * (Msun) l og t f ( G y r) blue@z=3, res=114pc/h 9 10 11 121.21.52 z f =4 log M * (Msun)blue@z=3, res=29pc/h Fig. 16.— shows the stellar mass M ∗ -mean galaxy formation time t f scatter plot for blue galaxies at z = 0 . − . z = 0 .
62. Bottom left panel: shows the stellar mass M ∗ -mean galaxy formation time t f scatter plot for bluegalaxies at z = 3 with the fiducial resolution 114pc/h. Bottom right panel: shows the stellar mass M ∗ -mean galaxyformation time t f scatter plot for blue galaxies at z = 3 with the four times better resolution of 29pc/h. The bluedots indicate average values. ing, one may ask this: is the age-mass relation of red galaxies inherited from their blue progenitors?We will now address this question. To select progenitors of red galaxies at z = 0 .
62, we note thatthe majority of galaxies that turn red by z = 0 .
62 have t qr = 1 − . z = 0 . − .
94 (8 snapshots with z = (0 . , . , . , . , . , . , . , . z = 0 .
62 and z = 0 .
80 and z = 0 .
94 are (1 . , . z = 0 .
62 near theonset of quenching. We separate the blue galaxies into two groups: one group contains the blueprogenitors of z = 0 .
62 red galaxies and the other group other blue galaxies that have not turnedinto red galaxies by z = 0 .
62. Figure 16 shows the stellar mass M ∗ -mean galaxy formation time t f scatter plot for blue galaxies at z = 0 . − .
94 that are progenitors of red galaxies at z = 0 .
62 (topleft) and those that do not become red galaxies (top right). Each small group of mostly linearlyaligned circles is one galaxy that appears multiple times (maximum is 8). Within the scatters we seethat the green dashed line, borrowed from Figure 15, provides a good match to the near constant 32 –age at the high mass end for the progenitors of red galaxies. The magenta dots, borrowed fromFigure 15, match well the trend for the blue dots in the mass range 10 . − . M (cid:12) . These resultsare fully consistent with our initial expectation based on the observation (of our simulation) of twophysical processes: (1) that stellar mass growth is moderate during t qr hence evolution during t qr does not significantly alter the mean star formation time of each galaxy, (2) less massive forminggalaxies have higher sSFR than massive galaxies, causing a steepening of the age-mass relation atthe low mass end. This explains the physical origin of the age-mass relation seen in Figure 15. Itis prudent to make sure that these important general trends seen in the simulation are robust. Inthe bottom two panels of Figure 16 we make a comparison between blue galaxies of two simulationswith different resolutions, at z = 3. The bottom-left panel is from the fiducial simulation with aresolution of 114pc/h and the bottom-right panel is from an identical simulation with four timesbetter resolution of 29pc/h. We see that both the age-mass trend at low mass end and the nearconstancy of stellar age at the high mass end are shared by the two simulations, suggesting thatresults from our fiducial simulation are sufficiently converged for the general trends presented atthe level of concerned accuracies.A comparison between the top left and top right panels in Figure 16 makes it clear that theage-mass relation of the blue progenitors of red galaxies at quenching is, to a large degree, sharedby blue galaxies that do not become red galaxies by z = 0 .
62. One subtle difference is that themost massive non-progenitor blue galaxies are slightly younger than the most massive progenitorsof red galaxies, suggesting that the blue progenitors of red galaxies, on “their way” to become redgalaxies, have started to “foreshadow” quenching effects mildly.
At a given redshift the cumulative environmental effects are imprinted on the relative distri-bution of galaxies of different color types and possibly on the properties of galaxies within eachtype. We now present predictions of our simulations with respect to these aspects. Figure 17 showsdistributions of three types of galaxies as a function of distance to the primary galaxy in unitsof the virial radius of the primary galaxy at z = 0 .
62. All galaxies with distance larger than 4virial radii of the primary galaxy are added to the bin with d/r cv = 4 −
5. We use the total galaxypopulation above the respective stellar mass threshold as a reference sample and distributions inthe top (blue galaxies), middle (green galaxies) and bottom (red galaxies) panels are normalizedrelative to reference sample. Comparing the top (blue galaxies), middle (green galaxies) and bottom(red galaxies) panels, we see clear differences of environmental dependencies of the three types ofgalaxies. For blue galaxies, there is a deficit at d/r cv ≤
2, which is compensated by a comparableexcess at d/r cv ≥
3. The range d/r cv = 2 − − . d/r cv ≤ d/r cv ≥
3. This trend is in agreement with observational indications (e.g., Woo et al. 33 – − P D F b / P D F t − blue galaxies>3 × >1 × >3 × − P D F g / P D F t − green galaxies0 1 2 3 4 5 −
101 red galaxiesd/r vc P D F r / P D F t − Fig. 17.—
Top panel shows distribution of the distance to the nearest primary galaxy for blue galaxies at z = 0 . b / PDF t −
1. The middle panel shows the normalized distribution of green galaxies, PDF g / PDF t −
1. Thebottom panel shows the normalized distribution of red galaxies, PDF r / PDF t −
1. All galaxies with distance largerthan 4 virial radii of the primary galaxy are added to the bin with d/r cv = 4 − S . We see that the excess of red galaxies starts at S = 100 keV cm and rises toward higherentropy regions for red galaxies. The trend for blue galaxies is almost an inverted version of that forred galaxies. The trend for green galaxies lie in-between those for blue and red galaxies, as expected.In Cen (2011a) we put forth the notion that a critical entropy S c = 100 keV cm (at z = 0 .
62 andweakly dependent on redshift), marks a transition to a regime of inefficient gas cooling hence coldgas starvation, because above this entropy the gas cooling time exceeds the Hubble time. This isborne out with our more detailed analysis here. We also plot (not shown here) distributions of threetypes of galaxies as a function of the environmental pressure p and environmental overdensity δ , respectively, and find that the trend is broadly similar to that see in Figure 18. Overall, ourresults are in accord with the observed density-morphology relation (e.g., Oemler 1974; Dressler1980; Postman & Geller 1984; Cooper et al. 2006; Tanaka et al. 2007; Bundy et al. 2006; Quadriet al. 2012; Muzzin et al. 2012), and with the general observed trend of galaxy population becoming 34 – − P D F b / P D F t − blue galaxies >3 × >1 × >3 × − P D F g / P D F t − green galaxies1 2 3 − (keV cm ) P D F r / P D F t − red galaxies Fig. 18.—
Top panel shows the normalized environmental entropy distribution of blue galaxies at z = 0 . b / PDF t −
1. The middle panel shows the normalized difference distribution of green and blue galaxies,PDF g / PDF b −
1. The bottom panel shows the normalized difference distribution of red and blue galaxies,PDF r / PDF b − bluer or mean/median specific star formation rate becoming higher towards underdense regions inthe local universe (e.g., Lewis et al. 2002; Goto et al. 2003; G´omez et al. 2003; Tanaka et al. 2004;Rojas et al. 2004).Having examined the dependencies of three types of galaxies on environmental variables, we nowexplore the dependencies on two additional variables: the mass of the halo of the primary galaxy andthe secondary to primary galaxy stellar mass ratio. Figure 19 shows fractions of three populationsof galaxies in terms of color (red, green, blue) as a function of secondary to primary galaxy stellarmass ratio. The (left, middle, right) columns are for primary galaxies of halo masses in threeranges (10 − M (cid:12) ,10 − M (cid:12) ,10 − M (cid:12) ) respectively. The four rows from top to bottom are forsecondaries within four different radial shells centered on the primary galaxy ( ≤ r cv , [1 − r cv , [2 − r cv ,[3 − r cv ). We adopt the following language to make comparative statements: the environmentquenching is important if the fraction of blue galaxies is less than the fraction of red galaxies andvice versa. We see two separate trends in Figure 19. First, more massive environments are moreable to quench star formation; for primary galaxies with halo masses in the range of 10 − M (cid:12) the quenching appears to extend at least to [2 − r cv , whereas for primary galaxies with lower halo 35 – − − hc =10 − − − P D F ( d = [ − ] r vc ) − − − P D F ( d = [ − ] r vc ) log M *s /M *c − − − P D F ( d = [ − ] r vc ) log M *s /M *c − − hc =10 − P D F ( d = [ − ] r vc ) − − P D F ( d = [ − ] r vc ) − − − P D F ( d = [ − ] r vc ) log M *s /M *c − − − P D F ( d = [ − ] r vc ) log M *s /M *c − − hc =10 − P D F ( d < r vc ) − − P D F ( d = [ − ] r vc ) − − − log M *s /M *c P D F ( d = [ − ] r vc ) red galaxiesgreen galaxiesblue galaxies − − − P D F ( d = [ − ] r vc ) log M *s /M *c Fig. 19.— shows fractions of three populations of galaxies in terms of color, (red, green, blue), as a functionof satellite to primary galaxy stellar mass ratio. The (left, middle, right) columns are for primary halo mass of(10 − M (cid:12) ,10 − M (cid:12) ,10 − M (cid:12) ). The four rows from top to bottom are for satellites within four differentradial shells centered on the primary galaxy ( ≤ r cv , [1 − r cv , [2 − r cv , [3 − r cv ). masses the quenching effect is no longer significant at [2 − r cv . Second, for any given primarygalaxy halo mass, the quenching is more effective when the ratio of secondary to primary stellarmass is smaller. That the impact of environmental effects increases with decreasing secondary toprimary galaxy stellar mass ratio is perhaps not surprising and consistent with observations (e.g.,Poggianti et al. 2009; Thomas et al. 2010; Wetzel et al. 2012), but at odds with the assumptionof mass-independent environment quenching in empirical modeling (e.g., Peng et al. 2010). Ourfinding that quenching is still effective down to at least halo mass range 10 − M (cid:12) for theprimary is in agreement with observations (e.g., Wetzel et al. 2012).Mok et al. (2013), using deep GMOS-S spectroscopy for 11 galaxy groups at z = 0 . −
1, showthat the strongest environmental dependence is observed in the fraction of passive galaxies, whichmake up only ∼
20 percent of the field in the mass range M star = 10 . − . M (cid:12) but are thedominant component of groups. If we take the radial range 3 − r cv as the “field”, we see that the 36 –fraction of red galaxies is in the range 20 −
30% (Figure 19), reaching 65 −
90% within 2 r cv , inagreement with Mok et al. (2013). In a large galaxy sample study of a z = 0 .
83 cluster, Patel et al.(2009) find that red galaxy fraction is 93 ±
3% in the central region of the cluster and declines to alevel of 64 ±
3% at a projected cluster-centric radius of ∼ − M (cid:12) , we see that within the virial radius the red fractions arein the range of 60 − − − more efficient in more hostile environment(closer to a larger galaxy, higher density, high pressure, high entropy, etc), but not faster, for themost part .Given the unambiguous environmental effects on the relative distributions of galaxy colors, weask the following reverse question: Can we decipher environmental effects on a set of blue galaxiesalone across different environments at any given epoch? Figure (20) shows the specific SFR (sSFR)distributions of blue galaxies as a function of the four environmental variables at z = 0 .
62. Whileone sees trends of the overall number of galaxies with respect to each of the environmental variables,as already shown in Figures (17,18), there is no strong visible trend of the mean value of sSFR withrespect to environment for blue galaxies. To put it slightly differently, the variations of sSFRamong blue galaxies at any given mass are sufficiently large and the quenching effects on galaxiesat a given sSFR and stellar mass are sufficiently varied that at any environment the scatter insSFR at a given stellar mass is larger than the difference in average sSFR. We attribute this resultphysically to the combined effects of substantial non-uniformity of the properties of blue galaxiesprior to entering the environmental “spheres” of influence and substantial non-uniformity of theinfluence of the environment on galaxies illustrated in Figures (5 and 6). This result - that theintrinsic properties of star-forming galaxies appear independent of environment - is, however, inaccord with observations (e.g., Ideue et al. 2012; Wijesinghe et al. 2012). 37 – − −
10 log S (keV cm ) l og s S F R ( G y r − ) blue galaxies − − − −
10 log p (K cm − ) l og s S F R ( G y r − ) blue galaxies 0 1 2 − −
10 log 1+ (cid:98) l og s S F R ( G y r − ) blue galaxies M * = 3 × M * = 10 M * = 10 − −
10 d/r vc l og s S F R ( G y r − ) blue galaxies Fig. 20.— shows the specific SFR of blue galaxies at z = 0 .
62 as a function of each of the four environmentalvariables. Black circles are for central galaxies and red for satellites. The size of each circle indicates the stellar massof a galaxy, as shown in the legend. The blue dots indicate median values.
4. Conclusions
Utilizing ab initio L arge-scale A daptive-mesh-refinement O mniscient Z oom- I n cosmologicalhydrodynamic simulations ( LAOZI Simulations ) of the standard cold dark matter model, weperform a chronological and statistical study of formation and evolution of (1664, 367, 1296) (blue,green, red) galaxies of stellar mass 10 . − . M (cid:12) at z = 0 .
62. The simulations have an ultra-highresolution of ≤ ∼
10 kpc)region is unaffected by ram-pressure but consumed by diminishing SF with an exponential time ofseveral hundred Myr. Subsequently, gas starvation in the high-velocity environment guards againstfurther significant SF, pushing it through the green valley to the red sequence. The total durationfrom the initial sharp drop of SFR to the time of entry into the red sequence is ∼ . − . z = 0 .
62. We suggestthat adopting a spread in quenching time in the semi-analytic modeling may result in an improvedtreatment in that approach.(2) The radius of a galaxy’s quenching sphere depends on properties of the infalling galaxy beingquenched, causing large variations in the quenching time scales at a given radius and large variationsin the quenching radius at a given mass of infalling galaxy. Nevertheless, the vast majority of redgalaxies are found to be within three virial radii of a larger galaxy, at the onset of quenching whenthe specific star formation rate experiences the sharpest decline to fall below ∼ (10 − − − )Gyr − (depending on the redshift when it occurs). The exponential decay time of SFR at the onset ofquenching is, on average, about a factor of two shorter for events occurring within the virial radiusof the quenching galaxy than for those outside the virial radius of the quenching galaxy, whichmay be evidence of enhanced quenching with the additional aid of tidal stripping when the onset ofquenching takes place within the virial radius. However, it is stressed that the overall duration fromthe onset of quenching to the time entering the sequence does not depend much on environment.(3) Not all galaxies in the green valley migrate to the red sequence; (40%,40%,48%) of greenvalley galaxies of stellar mass > (3 × , , × ) M (cid:12) do not proceed to become red galaxiesat z = 0 .
62 after having turned green for ≥ being more efficient , not faster , on average, in denser environment. Physically,this comes about because most of the time it takes to drive a blue star-forming galaxy to the redsequence is spent during the starvation phase, not the initial gas removal phase by ram-pressurestripping that displays a stronger dependence on environment. 39 –I would like to thank Claire Lackner for providing the SQL based merger tree constructionsoftware, and Drs. John Wise, Matthew Turk and Sam Skillmans for help with analysis andvisualization program yt (Turk et al. 2011). I would like to thank Michael Vogeley for a very carefulreading of the manuscript, many colleagues for useful discussions, and an anonymous referee for acritical and constructive report. Computing resources were in part provided by the NASA High-End Computing (HEC) Program through the NASA Advanced Supercomputing (NAS) Divisionat Ames Research Center. This work is supported in part by grant NASA NNX11AI23G. Thesimulation data are available from the author upon request. REFERENCES
Abel, T., Anninos, P., Zhang, Y., & Norman, M. L. 1997, New Astronomy, 2, 181Aird, J., Coil, A. L., Moustakas, J., Blanton, M. R., Burles, S. M., Cool, R. J., Eisenstein, D. J.,Smith, M. S. M., Wong, K. C., & Zhu, G. 2012, ApJ, 746, 90Baldry, I. K., Glazebrook, K., Brinkmann, J., Ivezi´c, ˇZ., Lupton, R. H., Nichol, R. C., & Szalay,A. S. 2004, ApJ, 600, 681Balogh, M. L., Baldry, I. K., Nichol, R., Miller, C., Bower, R., & Glazebrook, K. 2004, ApJ, 615,L101Balogh, M. L., Navarro, J. F., & Morris, S. L. 2000, ApJ, 540, 113Bekki, K. 2009, MNRAS, 399, 2221Bell, E. F., Wolf, C., Meisenheimer, K., Rix, H., Borch, A., Dye, S., Kleinheinrich, M., Wisotzki,L., & McIntosh, D. H. 2004, ApJ, 608, 752Binney, J. 1977, ApJ, 215, 483Blanton, M. R., Hogg, D. W., Bahcall, N. A., Baldry, I. K., Brinkmann, J., Csabai, I., Eisenstein,D., Fukugita, M., Gunn, J. E., Ivezi´c, ˇZ., Lamb, D. Q., Lupton, R. H., Loveday, J., Munn,J. A., Nichol, R. C., Okamura, S., Schlegel, D. J., Shimasaku, K., Strauss, M. A., Vogeley,M. S., & Weinberg, D. H. 2003a, ApJ, 594, 186Blanton, M. R., Hogg, D. W., Bahcall, N. A., Brinkmann, J., Britton, M., Connolly, A. J., Csabai, I.,Fukugita, M., Loveday, J., Meiksin, A., Munn, J. A., Nichol, R. C., Okamura, S., Quinn, T.,Schneider, D. P., Shimasaku, K., Strauss, M. A., Tegmark, M., Vogeley, M. S., & Weinberg,D. H. 2003b, ApJ, 592, 819Bongiorno, A., Merloni, A., Brusa, M., Magnelli, B., Salvato, M., Mignoli, M., Zamorani, G., Fiore,F., Rosario, D., Mainieri, V., Hao, H., Comastri, A., Vignali, C., Balestra, I., Bardelli, S.,Berta, S., Civano, F., Kampczyk, P., Le Floc’h, E., Lusso, E., Lutz, D., Pozzetti, L., Pozzi,F., Riguccini, L., Shankar, F., & Silverman, J. 2012, MNRAS, 427, 3103Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000 40 –Bryan, G. L., & Norman, M. L. 1999, in Structured Adaptive Mesh Refinement Grid Methods, ed.N. P. C. S. B. Baden (IMA Volumes on Structured Adaptive Mesh Refinement Methods, No.117), 165Bundy, K., Ellis, R. S., Conselice, C. J., Taylor, J. E., Cooper, M. C., Willmer, C. N. A., Weiner,B. J., Coil, A. L., Noeske, K. G., & Eisenhardt, P. R. M. 2006, ApJ, 651, 120Bundy, K., Georgakakis, A., Nandra, K., Ellis, R. S., Conselice, C. J., Laird, E., Coil, A., Cooper,M. C., Faber, S. M., Newman, J. A., Pierce, C. M., Primack, J. R., & Yan, R. 2008, ApJ,681, 931Cen, R. 2011a, ApJ, 741, 99—. 2011b, ApJ, 742, L33—. 2012a, ApJ, 753, 17—. 2012b, ApJ, 748, 121—. 2013, ArXiv e-printsCen, R., Kang, H., Ostriker, J. P., & Ryu, D. 1995, ApJ, 451, 436Cen, R., Nagamine, K., & Ostriker, J. P. 2005, ApJ, 635, 86Cen, R., & Ostriker, J. P. 1992, ApJ, 399, L113Cen, R., & Riquelme, M. A. 2008, ApJ, 674, 644Cheung, E., Faber, S. M., Koo, D. C., Dutton, A. A., Simard, L., McGrath, E. J., Huang, J.-S., Bell,E. F., Dekel, A., Fang, J. J., Salim, S., Barro, G., Bundy, K., Coil, A. L., Cooper, M. C.,Conselice, C. J., Davis, M., Dom´ınguez, A., Kassin, S. A., Kocevski, D. D., Koekemoer,A. M., Lin, L., Lotz, J. M., Newman, J. A., Phillips, A. C., Rosario, D. J., Weiner, B. J., &Willmer, C. N. A. 2012, ApJ, 760, 131Coil, A. L., Newman, J. A., Croton, D., Cooper, M. C., Davis, M., Faber, S. M., Gerke, B. F., Koo,D. C., Padmanabhan, N., Wechsler, R. H., & Weiner, B. J. 2008, ApJ, 672, 153Cooper, M. C., Newman, J. A., Croton, D. J., Weiner, B. J., Willmer, C. N. A., Gerke, B. F.,Madgwick, D. S., Faber, S. M., Davis, M., Coil, A. L., Finkbeiner, D. P., Guhathakurta, P.,& Koo, D. C. 2006, MNRAS, 370, 198Dalgarno, A., & McCray, R. A. 1972, ARA&A, 10, 375Danforth, C. W., & Shull, J. M. 2008, ApJ, 679, 194Davis, M., & Geller, M. J. 1976, ApJ, 208, 13Dekel, A., & Birnboim, Y. 2006, MNRAS, 368, 2 41 –Demarco, R., Gobat, R., Rosati, P., Lidman, C., Rettura, A., Nonino, M., van der Wel, A., Jee,M. J., Blakeslee, J. P., Ford, H. C., & Postman, M. 2010, ApJ, 725, 1252Dressler, A. 1980, ApJ, 236, 351Eisenstein, D., & Hu, P. 1999, ApJ, 511, 5Faber, S. M., Willmer, C. N. A., Wolf, C., Koo, D. C., Weiner, B. J., Newman, J. A., Im, M., Coil,A. L., Conroy, C., Cooper, M. C., Davis, M., Finkbeiner, D. P., Gerke, B. F., Gebhardt,K., Groth, E. J., Guhathakurta, P., Harker, J., Kaiser, N., Kassin, S., Kleinheinrich, M.,Konidaris, N. P., Kron, R. G., Lin, L., Luppino, G., Madgwick, D. S., Meisenheimer, K.,Noeske, K. G., Phillips, A. C., Sarajedini, V. L., Schiavon, R. P., Simard, L., Szalay, A. S.,Vogt, N. P., & Yan, R. 2007, ApJ, 665, 265Feldmann, R., Carollo, C. M., & Mayer, L. 2011, ApJ, 736, 88Gavazzi, G., Savorgnan, G., Fossati, M., Dotti, M., Fumagalli, M., Boselli, A., Guti´errez, L.,Hern´andez Toledo, H., Giovanelli, R., & Haynes, M. P. 2013, A&A, 553, A90G´omez, P. L., Nichol, R. C., Miller, C. J., Balogh, M. L., Goto, T., Zabludoff, A. I., Romer, A. K.,Bernardi, M., Sheth, R., Hopkins, A. M., Castander, F. J., Connolly, A. J., Schneider, D. P.,Brinkmann, J., Lamb, D. Q., SubbaRao, M., & York, D. G. 2003, ApJ, 584, 210Goto, T., Yamauchi, C., Fujita, Y., Okamura, S., Sekiguchi, M., Smail, I., Bernardi, M., & Gomez,P. L. 2003, MNRAS, 346, 601Gunn, J. E., & Gott, J. R. I. 1972, ApJ, 176, 1Haardt, F., & Madau, P. 1996, ApJ, 461, 20Harrison, C. M., Alexander, D. M., Mullaney, J. R., Altieri, B., Coia, D., Charmandaris, V., Daddi,E., Dannerbauer, H., Dasyra, K., Del Moro, A., Dickinson, M., Hickox, R. C., Ivison, R. J.,Kartaltepe, J., Le Floc’h, E., Leiton, R., Magnelli, B., Popesso, P., Rovilos, E., Rosario, D.,& Swinbank, A. M. 2012, ApJ, 760, L15Heckman, T. M. 2001, in Astronomical Society of the Pacific Conference Series, Vol. 240, Gas andGalaxy Evolution, ed. J. E. Hibbard, M. Rupen, & J. H. van Gorkom, 345Hogg, D. W., Blanton, M. R., Eisenstein, D. J., Gunn, J. E., Schlegel, D. J., Zehavi, I., Bahcall,N. A., Brinkmann, J., Csabai, I., Schneider, D. P., Weinberg, D. H., & York, D. G. 2003,ApJ, 585, L5Ideue, Y., Taniguchi, Y., Nagao, T., Shioya, Y., Kajisawa, M., Trump, J. R., Vergani, D., Iovino,A., Koekemoer, A. M., Le F`evre, O., Ilbert, O., & Scoville, N. Z. 2012, ApJ, 747, 42Impey, C., & Bothun, G. 1997, ARA&A, 35, 267Joung, M. R., Cen, R., & Bryan, G. L. 2009, ApJ, 692, L1 42 –Kauffmann, G., Heckman, T. M., White, S. D. M., Charlot, S., Tremonti, C., Peng, E. W., Seibert,M., Brinkmann, J., Nichol, R. C., SubbaRao, M., & York, D. 2003, MNRAS, 341, 54Kauffmann, G., White, S. D. M., Heckman, T. M., M´enard, B., Brinchmann, J., Charlot, S.,Tremonti, C., & Brinkmann, J. 2004, MNRAS, 353, 713Kennicutt, Jr., R. C. 1998, ApJ, 498, 541Kimm, T., Somerville, R. S., Yi, S. K., van den Bosch, F. C., Salim, S., Fontanot, F., Monaco, P.,Mo, H., Pasquali, A., Rich, R. M., & Yang, X. 2009, MNRAS, 394, 1131Knobel, C., Lilly, S. J., Kovaˇc, K., Peng, Y., Bschorr, T. J., Carollo, C. M., Contini, T., Kneib, J.-P.,Le Fevre, O., Mainieri, V., Renzini, A., Scodeggio, M., Zamorani, G., Bardelli, S., Bolzonella,M., Bongiorno, A., Caputi, K., Cucciati, O., de la Torre, S., de Ravel, L., Franzetti, P.,Garilli, B., Iovino, A., Kampczyk, P., Lamareille, F., Le Borgne, J.-F., Le Brun, V., Maier,C., Mignoli, M., Pello, R., Perez Montero, E., Presotto, V., Silverman, J., Tanaka, M.,Tasca, L., Tresse, L., Vergani, D., Zucca, E., Barnes, L., Bordoloi, R., Cappi, A., Cimatti,A., Coppa, G., Koekemoer, A. M., L´opez-Sanjuan, C., McCracken, H. J., Moresco, M., Nair,P., Pozzetti, L., & Welikala, N. 2013, ApJ, 769, 24Komatsu, E., Smith, K. M., Dunkley, J., Bennett, C. L., Gold, B., Hinshaw, G., Jarosik, N., Larson,D., Nolta, M. R., Page, L., Spergel, D. N., Halpern, M., Hill, R. S., Kogut, A., Limon, M.,Meyer, S. S., Odegard, N., Tucker, G. S., Weiland, J. L., Wollack, E., & Wright, E. L. 2010,ArXiv e-printsKormendy, J., Fisher, D. B., Cornell, M. E., & Bender, R. 2009, ApJS, 182, 216Kovac, K., Lilly, S. J., Knobel, C., Bschorr, T. J., Peng, Y., Carollo, C. M., Contini, T., Kneib, J.-P.,Le Fevre, O., Mainieri, V., Renzini, A., Scodeggio, M., Zamorani, G., Bardelli, S., Bolzonella,M., Bongiorno, A., Caputi, K., Cucciati, O., de la Torre, S., de Ravel, L., Franzetti, P.,Garilli, B., Iovino, A., Kampczyk, P., Lamareille, F., Le Borgne, J.-F., Le Brun, V., Maier,C., Mignoli, M., Oesch, P., Pello, R., Perez Montero, E., Presotto, V., Silverman, J., Tanaka,M., Tasca, L., Tresse, L., Vergani, D., Zucca, E., Aussel, H., Koekemoer, A. M., Le Floch,E., Moresco, M., & Pozzetti, L. 2013, ArXiv e-printsKronberger, T., Kapferer, W., Ferrari, C., Unterguggenberger, S., & Schindler, S. 2008, A&A, 481,337Larson, R. B., Tinsley, B. M., & Caldwell, C. N. 1980, ApJ, 237, 692Lewis, I., Balogh, M., De Propris, R., Couch, W., Bower, R., Offer, A., Bland-Hawthorn, J.,Baldry, I. K., Baugh, C., Bridges, T., Cannon, R., Cole, S., Colless, M., Collins, C., Cross,N., Dalton, G., Driver, S. P., Efstathiou, G., Ellis, R. S., Frenk, C. S., Glazebrook, K.,Hawkins, E., Jackson, C., Lahav, O., Lumsden, S., Maddox, S., Madgwick, D., Norberg, P.,Peacock, J. A., Percival, W., Peterson, B. A., Sutherland, W., & Taylor, K. 2002, MNRAS,334, 673 43 –McGee, S. L., Balogh, M. L., Wilman, D. J., Bower, R. G., Mulchaey, J. S., Parker, L. C., &Oemler, A. 2011, MNRAS, 413, 996McNamara, B. R., & Nulsen, P. E. J. 2007, ARA&A, 45, 117Mendel, J. T., Simard, L., Ellison, S. L., & Patton, D. R. 2013, MNRAS, 429, 2212Mendez, A. J., Coil, A. L., Lotz, J., Salim, S., Moustakas, J., & Simard, L. 2011, ArXiv e-printsMok, A., Balogh, M. L., McGee, S. L., Wilman, D. J., Finoguenov, A., Tanaka, M., Giodini, S.,Bower, R. G., Connelly, J. L., Hou, A., Mulchaey, J. S., & Parker, L. C. 2013, ArXiv e-printsMoore, B., Katz, N., Lake, G., Dressler, A., & Oemler, A. 1996, Nature, 379, 613Mori, M., & Burkert, A. 2000, ApJ, 538, 559Murray, S. D., White, S. D. M., Blondin, J. M., & Lin, D. N. C. 1993, ApJ, 407, 588Muzzin, A., Wilson, G., Yee, H. K. C., Gilbank, D., Hoekstra, H., Demarco, R., Balogh, M., vanDokkum, P., Franx, M., Ellingson, E., Hicks, A., Nantais, J., Noble, A., Lacy, M., Lidman,C., Rettura, A., Surace, J., & Webb, T. 2012, ApJ, 746, 188Oemler, Jr., A. 1974, ApJ, 194, 1Omma, H., & Binney, J. 2004, MNRAS, 350, L13Park, C., Choi, Y.-Y., Vogeley, M. S., Gott, III, J. R., Blanton, M. R., & SDSS Collaboration.2007, ApJ, 658, 898Patel, S. G., Kelson, D. D., Holden, B. P., Illingworth, G. D., Franx, M., van der Wel, A., & Ford,H. 2009, ApJ, 694, 1349Peng, Y.-j., Lilly, S. J., Kovaˇc, K., Bolzonella, M., Pozzetti, L., Renzini, A., Zamorani, G., Ilbert,O., Knobel, C., Iovino, A., Maier, C., Cucciati, O., Tasca, L., Carollo, C. M., Silverman, J.,Kampczyk, P., de Ravel, L., Sanders, D., Scoville, N., Contini, T., Mainieri, V., Scodeggio,M., Kneib, J.-P., Le F`evre, O., Bardelli, S., Bongiorno, A., Caputi, K., Coppa, G., de laTorre, S., Franzetti, P., Garilli, B., Lamareille, F., Le Borgne, J.-F., Le Brun, V., Mignoli,M., Perez Montero, E., Pello, R., Ricciardelli, E., Tanaka, M., Tresse, L., Vergani, D.,Welikala, N., Zucca, E., Oesch, P., Abbas, U., Barnes, L., Bordoloi, R., Bottini, D., Cappi,A., Cassata, P., Cimatti, A., Fumana, M., Hasinger, G., Koekemoer, A., Leauthaud, A.,Maccagni, D., Marinoni, C., McCracken, H., Memeo, P., Meneux, B., Nair, P., Porciani, C.,Presotto, V., & Scaramella, R. 2010, ApJ, 721, 193Peng, Y.-j., Lilly, S. J., Renzini, A., & Carollo, M. 2012, ApJ, 757, 4Planck Collaboration, Ade, P. A. R., Aghanim, N., Armitage-Caplan, C., Arnaud, M., Ashdown,M., Atrio-Barandela, F., Aumont, J., Baccigalupi, C., Banday, A. J., & et al. 2013, ArXive-prints 44 –Poggianti, B. M., Arag´on-Salamanca, A., Zaritsky, D., De Lucia, G., Milvang-Jensen, B., Desai, V.,Jablonka, P., Halliday, C., Rudnick, G., Varela, J., Bamford, S., Best, P., Clowe, D., Noll,S., Saglia, R., Pell´o, R., Simard, L., von der Linden, A., & White, S. 2009, ApJ, 693, 112Postman, M., & Geller, M. J. 1984, ApJ, 281, 95Pozzetti, L., Bolzonella, M., Lamareille, F., Zamorani, G., Franzetti, P., Le F`evre, O., Iovino, A.,Temporin, S., Ilbert, O., Arnouts, S., Charlot, S., Brinchmann, J., Zucca, E., Tresse, L.,Scodeggio, M., Guzzo, L., Bottini, D., Garilli, B., Le Brun, V., Maccagni, D., Picat, J. P.,Scaramella, R., Vettolani, G., Zanichelli, A., Adami, C., Bardelli, S., Cappi, A., Ciliegi,P., Contini, T., Foucaud, S., Gavignaud, I., McCracken, H. J., Marano, B., Marinoni, C.,Mazure, A., Meneux, B., Merighi, R., Paltani, S., Pell`o, R., Pollo, A., Radovich, M., Bondi,M., Bongiorno, A., Cucciati, O., de la Torre, S., Gregorini, L., Mellier, Y., Merluzzi, P.,Vergani, D., & Walcher, C. J. 2007, A&A, 474, 443Presotto, V., Iovino, A., Scodeggio, M., Cucciati, O., Knobel, C., Bolzonella, M., Oesch, P.,Finoguenov, A., Tanaka, M., Kovaˇc, K., Peng, Y., Zamorani, G., Bardelli, S., Pozzetti, L.,Kampczyk, P., L´opez-Sanjuan, C., Vergani, D., Zucca, E., Tasca, L. A. M., Carollo, C. M.,Contini, T., Kneib, J.-P., Le F`evre, O., Lilly, S., Mainieri, V., Renzini, A., Bongiorno, A.,Caputi, K., de la Torre, S., de Ravel, L., Franzetti, P., Garilli, B., Lamareille, F., Le Borgne,J.-F., Le Brun, V., Maier, C., Mignoli, M., Pell`o, R., Perez-Montero, E., Ricciardelli, E.,Silverman, J. D., Tresse, L., Barnes, L., Bordoloi, R., Cappi, A., Cimatti, A., Coppa, G.,Koekemoer, A. M., McCracken, H. J., Moresco, M., Nair, P., & Welikala, N. 2012, A&A,539, A55Quadri, R. F., Williams, R. J., Franx, M., & Hildebrandt, H. 2012, ApJ, 744, 88Quilis, V., Moore, B., & Bower, R. 2000, Science, 288, 1617Rafelski, M., Wolfe, A. M., Prochaska, J. X., Neeleman, M., & Mendez, A. J. 2012, ArXiv e-printsRamos, B. H. F., Pellegrini, P. S., Benoist, C., da Costa, L. N., Maia, M. A. G., Makler, M.,Ogando, R. L. C., de Simoni, F., & Mesquita, A. A. 2011, AJ, 142, 41Rasmussen, J., Mulchaey, J. S., Bai, L., Ponman, T. J., Raychaudhury, S., & Dariush, A. 2012,ApJ, 757, 122Rees, M. J., & Ostriker, J. P. 1977, MNRAS, 179, 541Rojas, R. R., Vogeley, M. S., Hoyle, F., & Brinkmann, J. 2004, ApJ, 617, 50Rosario, D. J., Mozena, M., Wuyts, S., Nandra, K., Koekemoer, A., McGrath, E., Hathi, N. P.,Dekel, A., Donley, J., Dunlop, J. S., Faber, S. M., Ferguson, H., Giavalisco, M., Grogin, N.,Guo, Y., Kocevski, D. D., Koo, D. C., Laird, E., Newman, J., Rangel, C., & Somerville, R.2013, ApJ, 763, 59 45 –Salim, S., Rich, R. M., Charlot, S., Brinchmann, J., Johnson, B. D., Schiminovich, D., Seibert, M.,Mallery, R., Heckman, T. M., Forster, K., Friedman, P. G., Martin, D. C., Morrissey, P.,Neff, S. G., Small, T., Wyder, T. K., Bianchi, L., Donas, J., Lee, Y., Madore, B. F., Milliard,B., Szalay, A. S., Welsh, B. Y., & Yi, S. K. 2007, ApJS, 173, 267Santini, P., Rosario, D. J., Shao, L., Lutz, D., Maiolino, R., Alexander, D. M., Altieri, B., Andreani,P., Aussel, H., Bauer, F. E., Berta, S., Bongiovanni, A., Brandt, W. N., Brusa, M., Cepa, J.,Cimatti, A., Daddi, E., Elbaz, D., Fontana, A., F¨orster Schreiber, N. M., Genzel, R., Grazian,A., Le Floc’h, E., Magnelli, B., Mainieri, V., Nordon, R., P´erez Garcia, A. M., Poglitsch,A., Popesso, P., Pozzi, F., Riguccini, L., Rodighiero, G., Salvato, M., Sanchez-Portal, M.,Sturm, E., Tacconi, L. J., Valtchanov, I., & Wuyts, S. 2012, A&A, 540, A109Scannapieco, E., & Oh, S. P. 2004, ApJ, 608, 62Schmidt, M. 1959, ApJ, 129, 243Silk, J. 1977, ApJ, 211, 638Strateva, I., Ivezi´c, ˇZ., Knapp, G. R., Narayanan, V. K., Strauss, M. A., Gunn, J. E., Lupton,R. H., Schlegel, D., Bahcall, N. A., Brinkmann, J., Brunner, R. J., Budav´ari, T., Csabai, I.,Castander, F. J., Doi, M., Fukugita, M., Gy˝ory, Z., Hamabe, M., Hennessy, G., Ichikawa,T., Kunszt, P. Z., Lamb, D. Q., McKay, T. A., Okamura, S., Racusin, J., Sekiguchi, M.,Schneider, D. P., Shimasaku, K., & York, D. 2001, AJ, 122, 1861Swinbank, A. M., Balogh, M. L., Bower, R. G., Zabludoff, A. I., Lucey, J. R., McGee, S. L., Miller,C. J., & Nichol, R. C. 2012, MNRAS, 420, 672Tanaka, M., Finoguenov, A., Lilly, S. J., Bolzonella, M., Carollo, C. M., Contini, T., Iovino, A.,Kneib, J.-P., Lamareille, F., Le Fevre, O., Mainieri, V., Presotto, V., Renzini, A., Scodeggio,M., Silverman, J. D., Zamorani, G., Bardelli, S., Bongiorno, A., Caputi, K., Cucciati, O., dela Torre, S., de Ravel, L., Franzetti, P., Garilli, B., Kampczyk, P., Knobel, C., Kovaˇc, K.,Le Borgne, J.-F., Le Brun, V., L´opez-Sanjuan, C., Maier, C., Mignoli, M., Pello, R., Peng,Y., Perez-Montero, E., Tasca, L., Tresse, L., Vergani, D., Zucca, E., Barnes, L., Bordoloi,R., Cappi, A., Cimatti, A., Coppa, G., Koekemoer, A. M., McCracken, H. J., Moresco, M.,Nair, P., Oesch, P., Pozzetti, L., & Welikala, N. 2012, PASJ, 64, 22Tanaka, M., Goto, T., Okamura, S., Shimasaku, K., & Brinkmann, J. 2004, AJ, 128, 2677Tanaka, M., Kodama, T., Kajisawa, M., Bower, R., Demarco, R., Finoguenov, A., Lidman, C., &Rosati, P. 2007, MNRAS, 377, 1206Taranu, D. S., Hudson, M. J., Balogh, M. L., Smith, R. J., Power, C., & Krane, B. 2012, ArXive-printsThomas, D., Maraston, C., Schawinski, K., Sarzi, M., & Silk, J. 2010, MNRAS, 404, 1775Tonnesen, S., & Bryan, G. L. 2009, ApJ, 694, 789 46 –Tonnesen, S., & Cen, R. 2012, MNRAS, 425, 2313Tripp, T. M., Sembach, K. R., Bowen, D. V., Savage, B. D., Jenkins, E. B., Lehner, N., & Richter,P. 2008, ApJS, 177, 39Turk, M. J., Smith, B. D., Oishi, J. S., Skory, S., Skillman, S. W., Abel, T., & Norman, M. L. 2011,ApJS, 192, 9van den Bosch, F. C., Aquino, D., Yang, X., Mo, H. J., Pasquali, A., McIntosh, D. H., Weinmann,S. M., & Kang, X. 2008, MNRAS, 387, 79van der Burg, R. F. J., Muzzin, A., Hoekstra, H., Lidman, C., Rettura, A., Wilson, G., Yee,H. K. C., Hildebrandt, H., Marchesini, D., Stefanon, M., & Kuijken, K. 2013, ArXiv e-printsWeinmann, S. M., Pasquali, A., Oppenheimer, B. D., Finlator, K., Mendel, J. T., Crain, R. A., &Macci`o, A. V. 2012, MNRAS, 426, 2797Weinmann, S. M., van den Bosch, F. C., Yang, X., & Mo, H. J. 2006, MNRAS, 366, 2Wetzel, A. R., Tinker, J. L., Conroy, C., & van den Bosch, F. C. 2012, ArXiv e-prints—. 2013a, MNRAS, 432, 336—. 2013b, ArXiv e-printsWhitaker, K. E., Labb´e, I., van Dokkum, P. G., Brammer, G., Kriek, M., Marchesini, D., Quadri,R. F., Franx, M., Muzzin, A., Williams, R. J., Bezanson, R., Illingworth, G. D., Lee, K.-S.,Lundgren, B., Nelson, E. J., Rudnick, G., Tal, T., & Wake, D. A. 2011, ApJ, 735, 86Wijesinghe, D. B., Hopkins, A. M., Brough, S., Taylor, E. N., Norberg, P., Bauer, A., Brown,M. J. I., Cameron, E., Conselice, C. J., Croom, S., Driver, S., Grootes, M. W., Jones,D. H., Kelvin, L., Loveday, J., Pimbblet, K. A., Popescu, C. C., Prescott, M., Sharp, R.,Baldry, I., Sadler, E. M., Liske, J., Robotham, A. S. G., Bamford, S., Bland-Hawthorn, J.,Gunawardhana, M., Meyer, M., Parkinson, H., Drinkwater, M. J., Peacock, J., & Tuffs, R.2012, MNRAS, 423, 3679Willmer, C. N. A., Faber, S. M., Koo, D. C., Weiner, B. J., Newman, J. A., Coil, A. L., Connolly,A. J., Conroy, C., Cooper, M. C., Davis, M., Finkbeiner, D. P., Gerke, B. F., Guhathakurta,P., Harker, J., Kaiser, N., Kassin, S., Konidaris, N. P., Lin, L., Luppino, G., Madgwick,D. S., Noeske, K. G., Phillips, A. C., & Yan, R. 2006, ApJ, 647, 853Woo, J., Dekel, A., Faber, S. M., Noeske, K., Koo, D. C., Gerke, B. F., Cooper, M. C., Salim, S.,Dutton, A. A., Newman, J., Weiner, B. J., Bundy, K., Willmer, C. N. A., Davis, M., & Yan,R. 2013, MNRAS, 428, 3306Xue, Y. Q., Brandt, W. N., Luo, B., Rafferty, D. A., Alexander, D. M., Bauer, F. E., Lehmer,B. D., Schneider, D. P., & Silverman, J. D. 2010, ApJ, 720, 368 47 –Yao, Y., Tripp, T. M., Wang, Q. D., Danforth, C. W., Canizares, C. R., Shull, J. M., Marshall,H. L., & Song, L. 2009, ApJ, 697, 1784Zehavi, I., Zheng, Z., Weinberg, D. H., Blanton, M. R., Bahcall, N. A., Berlind, A. A., Brinkmann,J., Frieman, J. A., Gunn, J. E., Lupton, R. H., Nichol, R. C., Percival, W. J., Schneider,D. P., Skibba, R. A., Strauss, M. A., Tegmark, M., & York, D. G. 2011, ApJ, 736, 59Zheng, X. Z., Bell, E. F., Papovich, C., Wolf, C., Meisenheimer, K., Rix, H., Rieke, G. H., &Somerville, R. 2007, ApJ, 661, L41