The breakBRD Breakdown: Using IllustrisTNG to Track the Quenching of an Observationally-Motivated Sample of Centrally Star-Forming Galaxies
Claire Kopenhafer, Tjitske K. Starkenburg, Stephanie Tonnesen, Sarah Tuttle
DD raft version O ctober
1, 2020Typeset using L A TEX preprint2 style in AASTeX62
The breakBRD Breakdown: Using IllustrisTNG to Track the Quenching of an Observationally-MotivatedSample of Centrally Star-Forming Galaxies C laire K openhafer , ∗ T jitske K. S tarkenburg ,
2, 3 S tephanie T onnesen , and S arah T uttle Department of Physics and Astronomy and Department of Computational Science, Mathematics, and Engineering,Michigan State University, 567 Wilson Rd, East Lansing MI 48823, USA Flatiron Institute, 162 5th Avenue, New York NY 10010, USA Center for Interdisciplinary Exploration and Research in Astrophysics (CIERA) andDepartment of Physics and Astronomy, Northwestern University, 1800 Sherman Ave, Evanston IL 60201, USA University of Washington, Seattle, 3910 15th Ave NE, Room C319, Seattle, WA, 98195-0002, USA
ABSTRACTThe observed breakBRD (“break bulges in red disks") galaxies are a nearby sample of face-on disk galaxies with particularly centrally concentrated star formation: they have red disksbut recent star formation in their centers as measured by the D n ∼
4% at z = ∼ z = .
5) of galaxies fulfill the breakBRD criteria, in agreement with observations. In com-parison with the mass-weighted parent IllustrisTNG sample, these galaxies tend to consist ofa higher fraction of satellite and splashback galaxies. However, the central, non-splashbackbreakBRD galaxies show similar environments, black hole masses, and merger rates, indicat-ing that there is not a single formation trigger for inner star formation and outer quenching.We determine that breakBRD analogue galaxies as a whole are in the process of quenching.The breakBRD state—with its highly centrally concentrated star formation—is uncommonin the history of either currently quiescent or star-forming galaxies; however, approximately10% of 10 < M ∗ / M (cid:12) < quiescent galaxies at z = ∼ INTRODUCTIONIn the standard picture of galaxy formation,galaxies form in an “inside-out” sense: high densi-ties and rapid gas cooling cause an early formationof the inner regions, while the outer parts formlater due to lower densities and slower gas accre-tion and cooling (e.g. Larson 1976; Matteucci &Francois 1989; Burkert et al. 1992; Chiappini et al.1997; van den Bosch 1998; Kepner 1999). ∗ [email protected] By far, most galaxies in the Local Universe seemto follow this general picture for their formationand growth, which can be traced through metallic-ity gradients and / or stellar age gradients and trendsof scale radii with stellar age (e.g. Pagel & Ed-munds 1981; Shaver et al. 1983; Williams et al.2009; Vila-Costas & Edmunds 1992; Dale et al.2016; see Maiolino & Mannucci 2019, for a re-view), even when taking into account processessuch as stellar radial orbit migration (Magrini et al.2016; Frankel et al. 2019). High resolution andspatially resolved observations largely corroborate a r X i v : . [ a s t r o - ph . GA ] S e p K openhafer et al .the general “inside-out” picture (e.g. Sánchez et al.2014; Belfiore et al. 2017; López Fernández et al.2018; Sánchez-Menguiano et al. 2018; Poetrod-jojo et al. 2018; reviewed by Sanchez 2019), butalso reinforce a number of caveats. For exam-ple, outside-in growth has been observed in dwarfgalaxies (Gallart et al. 2008; Zhang et al. 2012;Pérez et al. 2013; Pan et al. 2015; Ibarra-Medelet al. 2016; Wang et al. 2019), and galaxy interac-tions may a ff ect metallicity gradients through drymergers as well as through triggering gas flows(Mehlert et al. 2003; Sánchez-Blázquez et al. 2007;Queyrel et al. 2012).In more detail, the distribution of star formationwithin galaxies may be related to the global starformation rate. For example, Morselli et al. (2019)found that star formation is centrally enhanced ingalaxies near the star-forming sequence and cen-trally suppressed below it. In agreement with thiswork, several authors have found that galaxies notonly form inside-out but also quench from theinside-out (Ellison et al. 2018; Nelson et al. 2016;Li et al. 2015; Rowlands et al. 2018; Spilker et al.2019). Using the MaNGA survey (Bundy et al.2015; Yan et al. 2016), Lin et al. (2019) find thatthe fraction of galaxies quenching inside-out growswith mass (see also Belfiore et al. 2018). Simi-lar central sSFR suppression in galaxies below thestar forming main sequence was found for both theCALIFA (Sánchez et al. 2012) and SAMI (Croomet al. 2012; Bryant et al. 2015) surveys (GonzálezDelgado et al. 2016; Medling et al. 2018). On theother hand, post-starburst galaxies, and galaxieswith post-starburst or recently quenching regions,show large diversity and irregularity in the distri-bution of their star-forming and quiescent regions(Rowlands et al. 2018; Chen et al. 2019; Quai et al.2019).The presence of a bulge may also a ff ect theglobal star formation rate and the distribution ofstar formation within a galaxy (Genzel et al. 2014;Méndez-Abreu et al. 2019; Martig et al. 2009;Gensior et al. 2020, although see also Martig et al. 2013; Kretschmer & Teyssier 2020; Su et al. 2019).Observations indicate that a large bulge compo-nent is often correlated with galaxy quenching, butthat it is unlikely to cause quenching on its own(e.g. Bundy et al. 2010; Fang et al. 2013; Bellet al. 2012; Whitaker et al. 2015; Bait et al. 2017;Omand et al. 2014; Bluck et al. 2014; GonzálezDelgado et al. 2016), and Lang et al. (2014) andMcPartland et al. (2019) argue that the growth ofthe bulge precedes the global shutdown of star for-mation. However, when we focus on the disk,Abramson et al. (2014) have argued that exclud-ing the bulge leads to a constant sSFR for the starforming disk component of galaxies. In agreement,Medling et al. (2018) find that bulges have little ef-fect on the star formation in the disks of even early-type galaxies.There are, of course, many factors that can a ff ectstar formation and its distribution within a galaxy.Feedback from Active Galactic Nuclei (AGN) maydecrease or even quench the SFR from the centeroutwards. It could play a critical role in the regula-tion of star formation (Prasad et al. 2020; Voit et al.2020) and is often argued to be required to keepgalaxies quenched (e.g. Su et al. 2019). AGN feed-back is also related to bulge mass, illustrated in theM BH -M bulge relation (e.g. Kormendy & Ho 2013;Heckman & Best 2014, and references therein; butsee also Martin et al. 2018; Ding et al. 2020).In large-scale cosmological simulations, the tun-ing of AGN feedback makes it the dominant feed-back mechanism at low redshift: it decreases theSFR of highly star-forming objects (e.g. Katsianiset al. 2017; Davé et al. 2019) and is associated withquenching galaxies (e.g. Weinberger et al. 2018).Major as well as minor mergers can bring in ad-ditional gas and increase star formation (Mihos &Hernquist 1994; Cox et al. 2008; Kaviraj 2014;Willett et al. 2015; Hani et al. 2020) and inter-actions can redistribute angular momentum andfunnel gas toward the central regions (e.g. Mi-hos & Hernquist 1994; Hernquist & Mihos 1995;Naab et al. 2014; Lagos et al. 2018; Blumenthal & he break BRD B reakdown n g − r color). Theparent sample consisted of a bulge / disk decom-posed sample (Lackner & Gunn 2012) based onthe galaxy sample and spectral information fromthe NYU Value Added Galaxy Catalog (VAGC,Blanton et al. 2005) from the the Sloan Digital SkySurvey (SDSS) DR7 (Abazajian et al. 2009), whileusing images with improved sky subtraction fromSDSS DR8 (Aihara et al. 2011).BreakBRDs appear transitional in the opticalbands, while presenting as star-forming in the UVand IR bands. Most of this sample has stellarmasses above 10 M (cid:12) , so on the basis of massthey should present "inside-out" star formation. Inaddition, TT20 found that both the sSFR and thebulge-to-total (B / T) mass ratio in breakBRD galax-ies tended to be larger than that in the star-formingparent sample galaxies (sSFR > − . yr − ), in-dicating that the large B / T ratio is not driving adecrease in sSFR. Finally, breakBRD galaxies arewell-distributed across environmental density in a similar fashion to their parent sample, indicatingthat environmental processes are not driving theirunusual star formation distribution.In this paper we use the large-scale cosmologicalhydrodynamic simulation IllustrisTNG100 (Nel-son et al. 2018; Springel et al. 2018; Pillepichet al. 2018; Marinacci et al. 2018; Naiman et al.2018) to gain insight into this unusual galaxy sam-ple. We note that it is well-known that simulationscannot recreate the observed universe with per-fect accuracy or completeness. For example, thequenched fraction of low-mass satellites is poorlyreproduced in simulations (Donnari et al. 2020;Davé et al. 2017), there are too few low-mass blackholes at z = z = openhafer et al .analogues identified at z = z = z = . METHODSThe IllustrisTNG100 (public data release: Nel-son et al. 2019) is part of a suite of simulationsrun using the AREPO moving mesh code (Springel2010) with upgraded subgrid models compared tothe Illustris simulation (Vogelsberger et al. 2014;Genel et al. 2014); in particular, the upgrades mod-ified the black hole accretion and feedback model(Weinberger et al. 2017), galactic winds (Pillepichet al. 2018), and added magnetohydrodynamics(Pakmor et al. 2011). TNG100 has a volume of110.7 Mpc and a mass resolution of 7 . × M (cid:12) and 1 . × M (cid:12) for dark matter and baryonic ele-ments, respectively. The gravitational softening is0 .
74 kpc at z = .
18 kpc at z = .
03, 0 .
1, and 0 .
5, which were chosen fortheir lookback times of 0.48, 1.3, and 5.2 Gyr re-spectively. These times allow us to assess how longthe breakBRD state lasts and determine their futureevolution. In this paper we primarily focus on red-shifts 0 and 0 .
5, looking at the bBRDa properties Figure 1.
Normalized histograms of stellar mass of thewhole galaxy (left) and the inner 2 kpc (right) for ourparent (solid grey) and breakBRD analogues (green)selection at z = . M ∗ / M (cid:12) ) =
10 is due to our selection criteria. and their histories and futures. The other redshiftsare included to help better understand how long thebreakBRD analogue state might last. The analogueselections were made with a combination of photo-metric and spectral criteria, as described in Section2.2. 2.1.
Parent Sample
We define the “parent” sample of our analogueswith two di ff erent criteria. We first require thatgalaxy stellar mass must lie within 10 < M ∗ < M (cid:12) . The lower mass limit was chosen out ofconcern for mass resolution: since our analysis re-quires that we directly measure properties in thecentral 2 kpc of our galaxies, we must be sure thereis significant mass both in the entire galaxies andin the central regions. Histograms of the total andcentral mass distributions of the parent sample at z = . M ∗ > M (cid:12) , as these have low numbersin TNG100, and can be assumed to be giant ellipti-cals. Moreover, most ( ∼ observed breakBRD sample have M ∗ > M (cid:12) ,and all have M ∗ < M (cid:12) (TT20).We also require galaxies in our parent sample tohave R / > R / is the stellar halfmass radius. This removes galaxies which do not he break BRD B reakdown Table 1.
Number of galaxies in our TNG100 parent sample and its subsamplesName Selection Criteria z = . z = . z = . z = . Parent < log( M ∗ / M (cid:12) ) <
12 and R / > D n ∩ D n < . r < g-r Parent ∩ g − r > .
655 for r > bBRDa D n ∩ g-r
235 (37%) 288 (28%) 247 (29%) 72 (19%)NOTE. - Selection criteria described in Sections 2.1 and 2.2. The “ bBRDa ” designation is short for “breakBRDanalogues”. Numbers in parentheses are the percentage of that subsample that are central galaxies. have a well-defined central region. The first rowin Table 1 lists the size of our parent samples at z = . z = . z = .
1, and z = .
5. Thepercentage of galaxies that are central galaxies isquoted in parentheses.2.2.
BreakBRD Analogue Sample
Our breakBRD analogues are selected from theparent sample using two additional criteria. Thesecriteria are designed to mimic the selection crite-ria of the observational breakBRD sample, and se-lect for galaxies with red disks and outskirts (andtherefore little-to-no total star formation) but re-cent star formation in their centers. We accomplishthis through an SDSS color cut of g − r > .
655 for r > n < . r < (cid:46) r is defined relative tothe particle with the lowest gravitational potentialin the galaxy. We use this instead of the galaxycenter of mass, which may not reflect the galaxy’srotation center due to its sensitivity to structure atlarge radii (Genel et al. 2015). We use 2 kpc as theboundary between our inner and outer regions bothto exceed the minimum spatial resolution of Illus-trisTNG and to approximate the size of the SDSSspectral fiber at 0 . < z < .
05, which is the red-shift range of the breakBRD observational sample.These considerations are again why we restrict theparent to have R / > r < r > ff ective absorption of τ = . λ/ . µ m) − . ,which is reduced by a factor of 3 for stars olderthan 3 × yr. This is the same dust model em-ployed by Torrey et al. (2015) for stellar mocks ofthe original Illustris simulation. We emphasize thatwe only use the photometric colors and D n ff ect on our selection.We generate separate spectra for the inner ( r < r > openhafer et al .(Section 5.4). We convolve the spectra with theSDSS bandpass functions to generate broadbandmagnitudes such as g and r . We also calculatedthe D n n < . ∼ n z = . z = . g − r > .
655 outside 2 kpccomprise our g − r selection (third row in Table 1).There are 2816 such galaxies at z = .
0, and 890 at z = . n g − r se-lections yields our breakBRD analogue galaxies,also referred to as the bBRDa galaxies. We referto them as “analogues” because the observationalsample in TT20 are the “true” breakBRDs. Thereare 235 such galaxies at z = .
0, 269 at z = . z = .
1, and 72 at z = . n g − r criteria. Since galax-ies evolve over time, galaxies that were breakBRDanalogues at one redshift may not qualify at an-other; we discuss this, and what we can infer fromthis about the length of the breakBRD state, in Sec-tion 4.2.In Figure 2 we plot the D n z = . g − r color in the outer region ( r > Figure 2. D n g − r color in the r > z = .
0. The parentsample is shown in grey, while the breakBRD analoguesare green. Dashed black lines show the D n < . g − r > .
655 selection cuts. Note that the histogrambins used for plotting straddle these cuts.
The stellar mass distribution of our analoguesample is shown in Figure 1, along with the stellarmass within 2 kpc. The total (and central) stellarmass of the bBRDa galaxies tends to be lower thanthat of the parent sample. Many of the quantitiesexplored in our analysis are correlated with stel-lar mass; to that end, we weight the parent sam-ple so that its mass distribution matches that ofthe bBRDa galaxies, as we discuss below. Thisweighting is applied from Figure 3 onward. Wenote that for all figures, the breakBRD galaxies areincluded in the parent samples. This does not a ff ectour results. 2.3. Sample Comparison
Many of the properties we examine in Sections3 and 4 are dependent on both galaxy mass andgalaxy classification as a central or satellite. Wetherefore separate centrals and satellites in ouranalysis for both the parent and breakBRD ana- he break
BRD B reakdown Table 2.
Percentage of centrals that are splashbacksDefinition bRRDa Parent | z | After z = . . z = . . z = weighted parent samples. Splashbacks are centrals that “recently”transitioned from being satellites, defined as either af-ter z = .
1, or z = . ff erentat the α = .
05 level if | z | > .
96, and at α = .
01 if | z | > . logue samples. The fraction of central galaxies foreach sample is given in parentheses in Table 1.To minimize the influence of stellar mass whenexamining di ff erences between the parent andbBRDa galaxies (see Figure 1), we apply a weight-ing to the parent central and satellites so that theirstellar mass distributions match their respectiveanalogue sample.As we are treating centrals and satellites sepa-rately, we also make a point to identify those galax-ies that are splashbacks: galaxies that were consid-ered to be satellites of more massive hosts at ear-lier times but have since escaped that halo. Splash-back galaxies (also called backsplash galaxies orejected satellites) are approximately related to aknown caustic in the phase space structure of darkmatter halos beyond the virial radius: the splash-back radius (see e.g. Adhikari et al. 2014; Diemeret al. 2017; Haggar et al. 2020; Diemer 2020a) , In IllustrisTNG the central / satellite definition depends onhalos identified through the Friends-Of-Friends (FOF, Daviset al. 1985) and SUBFIND (Springel et al. 2001) algorithms,and are therefore not identical to central / satellite definitionsbased on halo radii determined through spherical overdensitycriteria or phase space structure-motivated size definitions.More et al. (2011) show that corresponding spherical over-densities for FOF halos with linking length b = . but we note that our definition can also includegalaxies that interacted with the more massive hostin a flyby encounter and were never gravitationallycaptured. Because, according to our definition, asplashback galaxy “recently” made the transitionfrom satellite to central, it may exhibit propertiessimilar to satellites (Diemand et al. 2007; Knebeet al. 2011; Wetzel et al. 2014; Buck et al. 2019).For z = .
0, we use both z = . z = . ff s: if a galaxy has transitioned since thechosen redshift, it is counted as a splashback. Wefound that splashbacks only had an impact on theproperties discussed in Sections 3.2.2 and 3.2.3, aswe discuss in those sections. Table 2 contains thefractions of splashbacks in both the bBRDa and weighted parent samples.Throughout the following sections we make useof three statistical tools to help quantify the dif-ferences and similarities between our various sub-samples. First, in most figures, we have anno-tated the medians and median absolute deviations(MADs) —robust statistical measures that are re-silient to outliers—for relevant distributions. As areference, the MAD ≈ . σ for a Gaussian distri-bution; however, there is no strong indication thatour data follow Gaussian distributions and so wekeep our statistical evaluations distribution-free.Second, we perform two-sample Kolmogorov-Smirnov (KS) tests to measure the “distance” be-tween distributions and determine if it is statisti-cally significant. Our tests take into account theweighting that we apply to the parent distributionsto remove any dependencies on stellar mass. Moresensitive tests, such as Anderson-Darling, exist;but, since we are trying to establish statisticallymeaningful di ff erences rather than similarities, andthe KS test functions best when di ff erences areglobal, we consider the KS test to be the more con-servative choice. (see e.g. Diemer 2020a,b). We conclude that there may betrue splashback galaxies among our centrals. Defined as med( | X i − ˜ X | ) where ˜ X is the median of { X i } K openhafer et al .Finally, there are a few instances in which wecompare proportions instead of distributions. Todetermine if two proportions are statistically dif-ferent from each other, we use a two-proportionz-test. This test functions by comparing the di ff er-ence in proportions to zero.Both the KS test and the z-test have critical val-ues for when to accept or reject the null hypothe-sis. For the two-sample KS test, the null hypoth-esis is that both empirical distributions are drawnfrom the same underlying distribution. For thetwo-proportion z-test, the null hypothesis is thatthe two proportions are equal. We use two com-mon significance levels, α = .
05 and 0 .
01. AKS test result of p < α indicates the null hypothe-sis can be rejected. For the two-proportion z-test,these significance levels correspond to critical val-ues of | z | > .
96 and 2 .
58, respectively. BREAKBRD ANALOGUE GALAXIES INTHE LOCAL UNIVERSEIn this section we focus on the breakBRD ana-logues identified at z =
0. We first compare themto the z = Star Formation and Gas Properties
Here we explore the properties of the z = Global Star Formation
We first consider where breakBRD analogues liein the stellar mass-sSFR plane. Figure 3 com-pares galaxy total stellar mass to total specificSFR (sSFR). Galaxies with no current star forma-tion (SFR =
0) and those with log(sSFR / yr) < −
13 were set to log(sSFR / yr) = −
13. The toprow shows central galaxies, and the bottom shows
Figure 3.
Total specific star formation rate vs total stel-lar mass for central galaxies (top) and satellites (bottom)in our z = . / yr) = −
13, to which galaxies with no starformation were also set. The numbers in the upper rightgive the median and MAD when ignoring these galax-ies. satellites . We remind the reader that in each case,as discussed in Section 2.3, the (central or satellite)parent sample has been weighted to match the totalstellar mass distribution of the bBRDa centrals andsatellites, respectively.To consider this quantitatively, we calculate themedian and median absolute deviations (MAD) ofthese distributions, taking into account the weight-ing, which are annotated on Figure 3. When calcu-lating these values we ignore galaxies that were setto log(sSFR / yr) = −
13. All of the bBRDa galax- For all figures with contours, the background (white) con-tour level contains, on average, 5% of parent galaxies. Thepercentage is at most 11%. he break
BRD B reakdown / yr) > −
13 and so have not beenmodified.The two subsets of bBRDa galaxies coverroughly the same sSFR range, and tend to belower in sSFR than each of their respective parents.Though the centrals’ distributions overlap slightly,the median + MAD of the parent and breakBRDcentrals do not, indicating each subsample is wellclustered around their respective medians. The twobBRDa medians lie almost on top of each other andtheir MADs are of similar magnitude, in contrastto the two parent subsamples. The low sSFR ofboth breakBRD samples may be a signature of ourselection criteria.This shows that the two bBRDa populations aremore similar to each other in sSFR than they areto their respective parent populations, highlight-ing that in particular central breakBRD analoguesare di ff erent from the general central star-forminggalaxy population.3.1.2. The Concentration of Stars and Gas
In Figure 4 we examine the concentration ofstar formation and gas using the ratio of the in-ner 2 kpc of galaxies to their whole. From topto bottom, the three concentration ratios we exam-ine are in star formation rate, “dense” gas mass,and total gas mass. Dense gas is that which isabove 0.1 cm − , the threshold for star formationin IllustrisTNG (Pillepich et al. 2018). Each ratiois plotted against both total stellar mass and totalinstantaneous sSFR, the two quantities from Fig-ure 3. Once again, the parent and breakBRD ana-logues are broken into central and satellite galax-ies. For some parent galaxies, these ratios are un-defined: 1439 parent galaxies (563 centrals) haveno SFR at all, 1385 (517 centrals) have no densegas, and 620 (9 centrals) have no total gas. Noneof these galaxies belong to the breakBRD analoguesubsample. These galaxies are naturally excludedfrom Figure 4 and are ignored in the ensuing dis-cussion. We apply the same weighting to the par-ent sample as used in Figure 3, and even thoughsome galaxies are excluded, the mass distributions of the weighted parents remains similar to those ofthe bBRDa galaxies for all ratios plotted (the low-est p -value from two-sample KS tests is 0.80).The combination of the g − r photometric se-lection and D n + MAD ranges, though as with the SFR, satel-lites exhibit slightly higher concentrations. Thedi ff erence in concentration between the parent andbBRDa galaxies is once again more pronouncedthan the di ff erences between centrals and satellitesin either sample. We also highlight that the densegas concentrations of breakBRD satellites and cen-trals are more alike than those of the parent satel-lites and centrals.0 K openhafer et al . Figure 4.
Concentration of star formation rate (top), mass of “dense” gas (middle), and total gas mass (bottom) withinthe inner 2 kpc of galaxies with nonzero SFR in our z = . Finally, the bottom row of Figure 4 shows thedegree to which all gas in the galaxies is cen-trally concentrated. Both classifications of parentand breakBRD galaxies tend to have concentra-tions near 0. This trend is stronger for the centrals:99.5% of parent centrals have 5% or less of theirtotal gas in their inner 2 kpc, compared to 74.4%of parent satellites. As satellites are expected to ex-perience ram pressure stripping, this di ff erence isnot surprising. For both satellites and centrals, thebreakBRD analogues tend to have slightly higher concentrations than their corresponding parents,but these di ff erences are not as strong as for theother quantities in Figure 4. Because all of thepopulations skew towards zero concentration, weuse a two-sample Kolmogorov-Smirnov (KS) testto further distinguish the parent and bBRDa galax-ies. This KS test accounts for the weighting in theparent (see Section 2.3). Testing centrals againstcentrals and satellites against satellites yields p -values of order 10 − or less, indicating the break- he break BRD B reakdown Figure 5.
Total mass of dense gas (top) and mass of dense gas within the inner 2 kpc (bottom) vs total stellar mass forour selection at z = .
0. “Dense” gas is defined by the SF threshold in IllustrisTNG (Pillepich et al. 2018). Both theweighted parent (contours) and breakBRD analogue samples (points) are split into central galaxies (left) and satellites(right). Guides show where the gas mass is 100% (dashed), 10% (dot-dashed) and 1% (dotted) of the stellar mass.Medians and MAD gas masses are labeled; they change by less than 0 .
2% when splashbacks are removed. See Figure4 for a description of the normalized histograms to the far right.
BRD analogues and parent samples are not drawnfrom the same underlying distribution.For all of the quantities shown in Figure 4, thebreakBRD analogues always have much higherconcentrations than the parent sample. We clearlysee that for the concentrations of dense gas andSFR, the two breakBRD subpopulations are moresimilar than the two parent subsamples. Thesetrends are unchanged when splashbacks are re-moved from the central subsamples.With Figure 5, we re-examine the spatial distri-bution of dense gas in our galaxies in order to de-termine whether this gas is more concentrated inthe center of galaxies because the central regionshave additional gas, or because dense gas is miss-ing from galaxy outskirts. As a function of stel- lar mass, we plot the galaxy’s total dense gas massin the top row, and the mass of dense gas within2 kpc in the bottom. Central galaxies are on theleft and satellite galaxies are on the right. Lines areincluded to guide the eye: they show where the gasmass is equivalent to 1%, 10%, and 100% of thestellar mass. As with Figure 4, we exclude parentgalaxies that do not have any dense gas (1385 total,517 centrals), or lack dense gas in their inner 2 kpc(1610 total, 1199 centrals). Figure 5 again uses thesame weighting as Figure 3. The exclusion of theaforementioned galaxies does alter the mass distri-bution of the weighted parent sample, but KS test p -values remain acceptable ( p ∼ . p ∼ . − . openhafer et al .The top row of Figure 5 shows that both bBRDasubsamples are systematically lower than their re-spective parents in total dense gas mass, and aremore similar to each other in terms of median andMAD than the two parent samples. Specifically,the central bBRDa galaxies are significantly di ff er-ent from the parent central distribution. This resultalso holds when splashbacks are removed from thecentral samples (the medians and MADs change byless than 0 . ff erence in central gas massis more pronounced between satellites and cen-trals than between parents and breakBRDs. Thisis born out when using a two-sample KS test:comparing satellites to satellites and centrals tocentrals results in p values above a threshold of0.05, while comparing the two bBRDa (and par-ent) samples to each other yield p below 0.01. Re-moving splashbacks from the central samples onlystrengthens the similarity between those two dis-tributions. These results suggest that breakBRDgalaxies exhibit an outer deficit of gas and not ancentral enhancement.We note that Figure 5 is qualitatively the samewhen we use all of the bound gas instead of onlythe dense gas in the galaxies. The total gas massesare low in the bBRDa galaxies while the centralgas masses are similar in all subpopulations.In summary, we find that the breakBRD ana-logues have more centrally-concentrated SFRsthan the weighted parent samples. Dense, SF-eligible gas is also more centrally concentratedin the bBRDa galaxies. Being a breakBRD is astronger predictor of concentration for these twoquantities than being a central or satellite. Whenconsidering all of the dense gas, the two breakBRDsubsamples are both lower than their respective parents, but all subpopulations have roughly thesame mass of dense gas in their central 2 kpc. ThebreakBRD analogues therefore seem to be missinggas in their outer regions with respect to the parentgalaxy sample.3.2. Possible Evolutionary Drivers
Thus far we have shown that the galaxies in thebreakBRD analogue sample at z = ff er from the par-ent population.3.2.1. Black Hole Mass
It has often been argued that there is a correla-tion and co-evolution of galaxies, or the central re-gions of galaxies, and their central supermassiveblack holes (see e.g. Heckman & Best 2014). Thus,the centrally-concentrated gas and star formationin breakBRD galaxies could also be reflected intheir black hole mass. We might expect high blackhole mass because high central gas density mayenhance black hole accretion and possibly alsoAGN activity. The causal link may be reversed,however, and strong AGN feedback from massiveblack holes may remove gas from a galaxy’s out-skirts. Indeed, in IllustrisTNG AGN feedback is animportant mechanism quenching star formation ingalaxies (Weinberger et al. 2018).Therefore, we now look at the mass of the centralblack hole in breakBRD analogues as compared tothe mass-weighted parent sample in Figure 6. Thisis plotted against total stellar mass. A line is in-cluded where the black hole mass is 5% of the to-tal stellar mass only to guide the eye. Some par-ent galaxies do not have black holes (their massesare stored as negative infinity). There are 121 suchsatellites, 10 of which are also breakBRD ana- he break
BRD B reakdown Figure 6.
Black hole mass vs total stellar mass for cen-tral galaxies (top) and satellites (bottom) in our z = . logues. There is also one parent central withouta black hole. These have been omitted from Figure6 and our analysis, and do not significantly impactthe weighting of the parent sample.The black hole mass distributions of the centraland satellite breakBRD galaxies match well withtheir parent distributions. As seen in Figure 6, theirmedians and MADs are all very similar. This resultremains when we instead calculate these statisticsfor log( M BH ) / log( M ∗ ), where the statistics are all ∼ . ± . z < .
03 and since z < .
5. Because AGNfeedback is related to black hole accretion rates (Weinberger et al. 2018), if breakBRD galaxies hadexperienced strong AGN feedback, the maximumblack hole accretion rate within their recent his-tory would likely be high as well. Yet the median + MAD accretion rate ranges heavily overlap be-tween the breakBRD analogues and their parents,whether using the rates at z = z = .
5. There isno evidence that the growth of the BHs in break-BRDs di ff ers from that of the parent sample.Using weighted two-sample KS tests to statisti-cally assess the similarity of the distribution shape(Section 2.3), we cannot reject the null hypothesisthat the parent and analogue satellites or centralsare drawn from the same underlying distributions( p > . Environment
As we have discussed in the introduction, satel-lite galaxies are more likely to experience envi-ronmental e ff ects that may quench their outskirtsby removing gas while enhancing star formationin the inner regions. We might expect breakBRDgalaxies to only reside in dense environments; al-though, as we discuss in Section 5.4, the observedsample shows no environmental influence. Thus,in this section we consider several indicators forthe environment of breakBRD analogues and theparent sample.First, we can simply compare the satellite frac-tions in the bBRDa and parent samples using Ta-ble 1. The parent sample has a ∼
37% satellitefraction while the breakBRD analogue sample hasa ∼
63% satellite fraction. This di ff erence is sta-tistically significant according to a two-proportionz-test ( | z | = . | z | > .
58 is significant at α = . openhafer et al . Figure 7.
Number of galaxies with M ∗ > M (cid:12) within 2 Mpc vs total stellar mass, broken into centrals(left) and satellites (right), for our z = . n of neighboring galaxies is plotted as log( n + z = . n +
1) are labeled for the fullsample, including splashbacks. We note that the hostgalaxies of the satellites are not excluded from our en-vironment measurement for either the parent or break-BRD samples. pathways for satellites to become centrally concen-trated, the enhanced satellite fraction among break-BRDs is not a surprise; however, there remains alarge fraction of central bBRDa galaxies.In Section 2.3 we introduced Table 2 and thefraction of splashback galaxies in the analogueand weighted parent samples. In both cases, theproportion of splashbacks in the bBRDa sampleis double that in the parent sample; however, thestatistical significance of this di ff erence depends on the redshift used. The table includes the two-proportion z-test statistic, which tells us whether toreject the null hypothesis that the two proportionsare statistically the same. At the α = .
05 level, thebreakBRD analogues have a significantly higherproportion of splashbacks since z = .
5. The pro-portions of splashbacks since redshift z = . z = .
5) compared to 90% ofthe parent.Within our central and satellite galaxy selectionswe can also measure the environmental density. InFigure 7, we look at the number of M ∗ > M (cid:12) galaxies within 2 Mpc as a measure of the densityof surrounding galaxies. The number of neighbor-ing galaxies that meet this criteria are plotted aslog( n + log ( n +
1) space in thefigure, so are equivalent to 1 galaxy for both cen-tral samples and 7 and 8 galaxies for the bBRDaand parent satellite samples, respectively. There issignificant overlap in the MADs.When we remove either kind of splashback (from z = M ∗ > M (cid:12) within 2 Mpc to 0 galaxies, while the bBRDa me-dian density remains at 1 galaxy. We see that re-moving splashbacks decreases the width of the tailin the centrals’ histogram at higher densities; thisis shown with the dashed lines in the upper right ofFigure 7. This is to be expected: since splashbacksare galaxies that were recently satellites, we shouldfind them in denser environments. However, evenwith splashbacks removed, the MADs overlap forthe parent and breakBRD central samples.For a closer look at the environmental densitydistributions we also performed two-sample KStests. The breakBRDs’ density distributions aresignificantly di ff erent from those of the parents he break BRD B reakdown p = .
011 for the satellites, and the centralshave p = × − ). The p -value for the centralsincreases as we remove galaxies that have beensplashbacks since earlier times, reaching p = . z = .
5. The non-splashback central breakBRD galaxies thus live issimilar environments to the non-splashback centralparent sample.While the two satellite distributions have an in-significant p -value, they cover roughly the samedensity range. Because the median values are closeand the MADs overlap, even though the shapes oftheir distributions are di ff erent, we do not considerthe satellite breakBRDs to be in meaningfully dif-ferent environments than the parent sample satel-lites.Additionally, we check whether these resultschange if we look at the distance and mass of thenearest more massive galaxy, or at the distance andmass of the nearest cluster (where we define clus-ter as a halo with M DM > M (cid:12) ). We find thatthe environment of breakBRD analogues are sim-ilar using both of these measures. This similarityexists in both the shape of the distribution mea-sured using the KS test, and in the close overlap ofthe medians and MADs.Therefore, while the fraction of current or recentsatellite galaxies is higher in the breakBRD thanin the parent sample, the environment of galaxieswithin either the satellite or central samples is quitesimilar to the corresponding parent sample. Thismay indicate that a more subtle environmental ef-fect is responsible for the evolution of breakBRDsand that this avenue is worthy of further investiga-tion. 3.2.3. Merger History
We counted the number of central and satellitegalaxies that experienced at least one merger withmass ratio M ∗ , / M ∗ , ≥ .
01 since z = .
5. Merg-ers were classified by the stellar mass ratio of thegalaxies involved: M ∗ , / M ∗ , ≥ / / / / / ff erent between the breakBRDs andtheir weighted parent samples. This is true eventhough none of the breakBRD galaxies have ex-perienced a major merger, because only ∼
2% ofgalaxies in the weighted central and satellite par-ent samples have. This lack of significance holdswhen we remove centrals that have been splash-backs since redshift 0.1 and 0.5.According to our z-tests, the only significant dif-ference is in the frequency of very minor mergers(mass ratios of 1 / /
10) experienced by thecentrals: 26% of the parent centrals experience atleast one accretion event, versus 10% of breakBRDcentrals. This is significant at α = .
01, even whenremoving splashbacks since redshift 0.1, and is sig-nificant at α = .
05 when removing splashbackssince redshift 0.5.The decreased frequency of low mass accretionevents can be interpreted as a sign of less cos-mic accretion overall (including smooth, whichthis method poorly accounts for), as small satel-lites are often accreted onto more massive halostogether with general accretion along a filament.This makes it unlikely that breakBRD analoguegalaxies are generally formed by gas being broughtin from outside the halo or funneled from the out-skirts to the central region through tidal e ff ects.Both central funneling and the addition of newgas are mechanisms that would increase the cen-tral concentration of gas and star formation rateas seen in Section 3.1.2. However, the dearth ofrecently experienced minor accretion events couldpoint to indirect and subtle starvation-like environ-6 K openhafer et al .mental e ff ects on the evolution of the central break-BRD galaxies.3.3. The History of breakBRDs
In an e ff ort to understand why and how break-BRD analogue galaxies have their unique star-forming and color characteristics, we now tracethe history of the z = . ff er from the parent sam-ple, tracking their evolution allows us to under-stand these di ff erences in more detail. Moreover,as Table 3 shows that none of the z = . z = . z = . z = . ff er-ences in log, or ratio, of the amounts at z = . z =
0) of a number of galaxy propertiessince redshift z = . z = central galaxies: the total gas mass, thegas mass outside the disk, the total stellar mass,the total star formation rate, the total black holemass, and the total halo mass. This is comparedagainst the SFR concentration at the present day.The parent sample is mass-weighted with respectto the mass distribution of the breakBRD galaxiesat z =
0. Overplotted lines indicate the rolling me-dian trends for the parent (dotted) and breakBRDanalogues (dashed) samples. We choose to showthe relative growth with time against the z = .
77 dex, or 68%;see the upper left panel of Figure 8). The most dra-matic gas loss in breakBRD analogue galaxies isfor splashback galaxies, without which the break-BRD analogues lie closer to the parent sample witha median and MAD of − . ± . z = .
5. This trend is stronger for more centrallyconcentrated satellites. This is to be expected ifthese satellites lose their gas through, for example,ram-pressure stripping.When we check the growth of the gas mass in theouter regions of the galaxies, we find that the massin total gas that is lost is predominantly lost fromthe outskirts ( r > R / , see the upper middle panelof Figure 8). While not shown, we have checkedthat the gas evolution in the inner parts is similarto that of the parent sample (akin to the di ff erencesbetween the upper and lower panels of Figure 5).Additionally, we check whether the amount ofgas that is lost is converted into stars, as opposedto being removed from or never accreted by thegalaxy. About half of the central bBRDa galax-ies (including most of the splashbacks and about athird of the non-splashbacks) have lost more gasmass than they have gained in stars that formedover the same time interval. The other half of thecentral breakBRDs have retained more gas thanthey have converted into stars, or even gained asmall amount of gas since z = .
5. So, for thebBRDa galaxies that have lost significant amountsof gas, some of that gas may have been convertedinto stars, but even more gas is still lost. Over-all, the upper right panel of Figure 8 shows that thebBRDas have formed fewer stars since z = . ff er-ence is smaller when comparing to parent galaxieswith similar high SFR concentration.The star formation rate evolution for the cen-tral breakBRDs (lower left of Figure 8) followsa trend similar to that of the total gas mass: thecentral breakBRD analogues experience a higher he break BRD B reakdown Figure 8.
Present day star formation rate concentration against relative growth (log X ( z = . − log X ( z = . z = . reduction of SFR compared to the parent sample.This remains high even compared to the highly-SF-concentrated parent sample. However, the amountof reduction is larger for the breakBRDs: the me-dian value of SFR loss is 0 .
55 dex, which amountsto a reduction in star formation of 72%, with lossesup to 1 .
25 dex (94% SFR reduction). It is notewor-thy that excluding splashback galaxies has less ofan impact on the evolution of the SFR than on thegas content: when removing splashbacks, the me-dian + MAD is − . ± .
22 for the breakBRDs.Studying the evolution of the central black holeis particularly interesting as AGN related feed-back can strongly influence both the central re-gions as well as the more extended gas distribu-tions of galaxies. In Figure 6 we show that thebreakBRD analogue galaxies fall on the black hole mass – stellar mass relation of the parent; in Figure8, we study their black hole growth since z = . . ± .
05 vs. 0 . ± . . ± . openhafer et al .conclusions when comparing the maximum blackhole accretion rates in each galaxy’s recent history(see Section 3.2.1). Black holes are therefore un-likely to be the main driver of the breakBRD state.Lastly, we measure the growth of the dark matterhalos between z = . z = ff ects could have played apart in forming the current state of these galaxies.For the splashback breakBRD galaxies this is verylikely the case. For the non-splashback centralbreakBRD analogues, the environment may haveplayed a more subtle role in their formation. How-ever, as the current environment of the breakBRDsis very similar to that of the parent sample, thismay indicate that the past environment of break-BRDs may have been di ff erent.By following the change in galaxy properties,we see that for the central galaxy samples, boththe gas mass and SFR has decreased more in thebreakBRD sample than in the parent sample, andthe black holes masses, stellar masses, and halomasses have grown less. We see that these dif-ferences are partially ameliorated when restrictingthe parent to high SFR concentration, but that evenwhen the breakBRD centrals are compared only tothis subsample of parent centrals, di ff erences re-main (see e.g. the running medians in Figure 8).The largest di ff erence is in the change in the SFRover time, likely because breakBRDs are selectedto not only have centralized SF, but also to have reddisks.Our results indicate that before reaching thebreakBRD state, the z = z = . z = .
1, we find that whilethe trends are similar, the di ff erences between thebBRDa and the parent samples are much smaller.Moreover, while the breakBRD galaxies have lostgas and SFR since z = . z = . ff erent redshifts (see Table 3). We there-fore conclude that the process involved must havea ff ected breakBRD galaxies over a significant timerange and is likely to have gradually changed thegalaxies to the breakBRD state.We now look more closely at the intrinsic break-BRD sample at z = . BREAKBRD ANALOGUES AT HIGHERREDSHIFTSIn the previous section, we focused on the break-BRD analogue population at z = . z = . ∼ . z = . ∼ . z = . ∼ he break BRD B reakdown Figure 9. D n g − r color inthe r > z = .
5. Thehistograms at the top and right are normalized. The par-ent sample is shown in grey, while the z = . n < . g − r > .
655 selection cuts. Notethat the histogram bins used for plotting straddle thesecut, and the axis limits are di ff erent from Figure 2. discuss both these possibilities when studying thefuture of breakBRD galaxies in Section 4.2.In the following sections, we briefly show thesimilarity of z = . z =
0, and then look at the future of the z = . z = . z = . .
03 bBRDa galaxies,but they follow the same patterns as the redshift 0and 0.5 samples. 4.1.
Properties
Both Table 1 and Figure 9 show that thereare more galaxies at z = . n < .
4, and therefore recent star forma-tion, which is consistent with our expectationsfor increased star formation rates at this redshiftcompared to z =
0. There are fewer galaxies at z = . Figure 10.
Total instantaneous specific star formationrate vs total stellar mass for central galaxies (top) andsatellites (bottom) in our z = . / yr) = − logues at z = . z = .
0. This is not unexpected, as theincreased star formation will make disks tend tobluer. While we compute g-r color magnitudes inthe local frame, our selection cut ( g − r > . ff erences, however,do not a ff ect the overall behavior of the selectioncuts, as we also select galaxies at extreme ends ofthe distributions at z =
0. The fraction of ana-logues that are centrals is also lower, being 19%instead of the 37% at z = .
0. This is despite a rel-atively consistent fraction of centrals in the parentsamples at both redshifts.In order to compensate for correlations with stel-lar mass, we again weight the z = . openhafer et al .logues, just as we did with our z = . z = . z = . / yr) = −
13. We see the same trend hereas we did in Figure 3, where the breakBRD ana-logues tend to be at lower sSFR than their weightedparent samples. For the satellites, this di ff erence iseven more pronounced at z = . z = . z = . ff erence between theblack hole mass distributions, just like at z = . z = . z = . z = .
5. Using the concentrationmeasures introduced in Figure 4, the z = . z = . z =
0, where there was onlya slight enhancement over the parent sample (seeFigure 7). The primary goal of the z = . The Future of breakBRDs
Table 3.
Overlap in breakBRD analogue samplesRedshifts ∆ t (Gyr) Number Percentage0.0 & 0.03 0.48 135 46.90.03 & 0.1 0.86 126 51.00.0 & 0.1 1.3 63 25.50.1 & 0.5 3.9 2 2.80.03 & 0.5 4.7 1 1.40.0 & 0.5 5.2 0 0.0NOTE. - Table is ordererd by the time elapsed betweenthe two redshifts. Percentages are calculated relative tothe higher redshift, using the bBRDa numbers in Table1. Figure 11.
Same as Figure 10 but the parent and break-BRD samples selected at z = . z = .
0. Parent galaxies maintain the weight-ing applied in Figure 10; see text for more details. X’smark breakBRD galaxies that have merged with one ofthe parent sample by z = . In Table 3 we provide the overlap in bBRDa sam-ples, i.e. the number of galaxies that are break-BRDs at two epochs, at our four redshifts: 0, 0.03,0.1, and 0.5. It is out of the scope of this paper he break
BRD B reakdown z = . z = .
03 samplesshows that ∼
50% of the breakBRDs at z = .
03 arestill in that state 0.48 Gyr later, at redshift 0. Thispercentage drops to ∼
25% over the 1.3 Gyr spanbetween z = . ∼ z = . z = . / yr) = − z = . .
5; descendent parents are onlycounted once, and are weighted according to themost massive of their progenitors in the z = . / yr) (cid:38) −
11, whichwe use as our delineation between the star-formingand quiescent galaxy populations (see e.g. Cas-sata et al. 2010; Tamburri et al. 2014; Matthee& Schaye 2019; Katsianis et al. 2019). Three ofthese are centrals. When tracked to redshift 0 (Fig- ure 11), 85.7% of central breakBRD analogues (12out of 14), and 94.8% of satellites (55 out of 58)have log(sSFR / yr) < −
11; therefore, most of thegalaxies that were breakBRDs at z = . z = .
5, weuse the sSFR of the most massive progenitor inthe sample, akin to how we handled the weight-ing. Again, some bBRDa galaxies have mergedwith redshift z = . z = .
5. With weightingapplied, the fraction of parent centrals and satel-lites that quench by redshift z = α = .
01 level.Though not shown, we also track forward theredshift z = . z = .
1, these galaxiesdisplay the same sSFR pattern as seen in Figures3 and 10; yet when tracked to z = .
0, thesegalaxies have quenched to a lesser extent. Ofthe satellites, 51.1% of the bBRDa galaxies havelog(sSFR / yr) < = −
11, compared to 10.7% of theweighted parent. For the centrals, 26.0% of thebBRDa have log(sSFR / yr) < = −
11 versus 4.5% ofthe parent. Both fractions are significantly di ff er-ent at α = .
01. Preferential quenching in satellitesis not unexpected, because in satellites, the break-BRD state may indicate strong ram pressure strip-ping or tidal e ff ects which could rapidly quench thesatellite. Such a straightforward explanation wouldnot apply to the centrals, however.We see a similar set of properties for the break-BRDs at redshifts 0, 0.1, and 0.5, which suggeststhat the breakBRD selection criteria chooses a con-sistent set of galaxies. We therefore feel confi-2 K openhafer et al .dent that the quenching experienced by the z = . DISCUSSIONThe following discussion will use the results putforth in Sections 3 and 4 to answer a series of ques-tions about the formation and evolution of break-BRD analogues. Finally, we will briefly comparethe simulated analogues to the observational sam-ple.5.1.
Central Enhancement vs Outer Deficit
In Figure 4 and the accompanying discussion, wefind that the breakBRD analogues have SFR andgas distributions that are centrally concentrated. Inthe case of SFR and dense, star formation-eligiblegas, these concentrations can be quite high, withat least 40% of each contained within the cen-tral 2 kpc-radius. Our selection criteria were de-signed to choose galaxies with highly concentratedstar formation, and these stars must be formingout of dense gas, so high concentrations are ex-pected. Note, however, that the parent sample alsoincludes non-breakBRD galaxies with highly con-centrated SFR and dense gas: the bBRDa galax-ies di ff er from these because these non-breakBRDSFR-concentrated galaxies also have some star for-mation in their outer regions. We now ask: did aprocess cause breakBRD galaxies to have highercentral concentrations? Or do they appear this waybecause gas was removed from their outskirts?Figure 5 is useful for distinguishing central gasenhancement from outer gas deficit, as the densegas ratio from Figure 4 is separated into its compo-nents. Here we find that the breakBRD analogueshave roughly the same gas mass in their interiors asparent galaxies weighted for the same stellar mass,despite being lower in dense gas mass overall. Weinterpret this as a strong indication that gas hasbeen removed from the outskirts ( r > The Formation of breakBRDs
The selection criteria laid out in Section 2.2 cer-tainly select a unique sample of galaxies, and it isinstructive to investigate how these galaxies form.Indeed, the stripping suggested by Figure 5 (anddiscussed in Section 5.1) motivated Section 3.2,as black holes, environment, and merger historycould all possibly explain a deficiency of outer gas(see Section 1). While we discuss these processesseparately, here we endeavor to synthesize whatthe results of 3.2 tell us about the formation ofbreakBRD analogues.First, we consider black holes. With Figure 6,we find no di ff erence between the black hole massof z = ff er-ence between parent and bBRDa galaxies in termsof their black hole growth: though the bBRDacentrals tend to have black holes that grow moreslowly, the apparent di ff erence is not significant. he break BRD B reakdown ff erencefor the fraction of centrals that have had at least oneminor accretion event (mass ratio of 1 / / ff ect would still notfully explain the breakBRD state, however, giventhat most breakBRD galaxies have lost significantamounts of gas.Finally, satellite galaxies may have outlying gasstripped by environmental processes (either grav-itational or hydrodynamical). We do see a muchhigher satellite fraction and splashback fraction inthe breakBRD analogue sample than the parentsample, indicating that environmental processescould be an important process forming breakBRDgalaxies. Also, when we examined the change ingas mass of central z = z = .
5. They have also experienceda commensurate change in dark matter mass, pos-sibly hinting at subtle environmental e ff ects form-ing even the central breakBRD population.However, when measuring the density of theirpresent day environment (Figure 7), the median + MAD values overlap for the non-splashbackbreakBRD analogues and the non-splashback par-ent sample. Though not shown, we also quanti-fied their environment through the distance to thenearest more massive neighbor and to the nearest cluster. Here we also found significant overlap be-tween the median + MAD of the central break-BRD and parent samples. Therefore, while envi-ronmental processes can explain the properties ofthe satellites and splashback galaxies in the break-BRD sample, if the environment has an importantrole in the formation of central breakBRD ana-logues, this may indicate that the environment ofbreakBRD central galaxies has changed.The strongest di ff erence between breakBRD ana-logues and their TNG parent sample seems tobe environment; however, the di ff erence is nota “smoking gun.” Importantly, we stress thatthroughout our findings, breakBRD analogues aremore like each other (across the satellite-central di-vide) than galaxies in the parent sample. For exam-ple, the median dense gas mass and concentrationof breakBRD satellites and centrals are similar,and both are significantly larger than the concen-trations of the parent subsamples. While environ-ment may be an influencial factor in the formationof breakBRDs, the exact way in which it exerts thisinfluence appears to be subtle and worthy of fur-ther, more detailed investigation beyond the grossstatistics explored herein.5.3. Quenching vs Stochastic Star Formation
One benefit of cosmological simulations is thatwe can answer the question “what will happen tothis galaxy in the future?”. This is exactly the ques-tion we asked of the z = . z = .
0. Though weconclude that the z = . could have ever been in4 K openhafer et al . Figure 12.
Maximum star formation rate concentration,max(SFR / SFR total ), of each galaxy, either since z = . / yr) ≥ −
10 for quenched parent galaxies(bottom) against the stellar mass they had at the time ofpeak SFR concentration. The plot is broken into cen-trals (left) and satellites (right). Overplotted is the z = z = .
5. In contrast to other fig-ures, the breakBRD sample is not included in either thestar-forming or quiescent parent sample. Note that incontrast to most other figures, here we show all parentgalaxies without weighting . the breakBRD state by looking at maximum SFRconcentration each galaxy has reached during itshistory. We first split our z = . / yr) = −
11 as the boundary. For the star-forming sample, the maximum SFR concentrationof the parent sample is determined across all
Illus-trisTNG outputs between z = . z =
0. Red-shift 0.5 is chosen as the endpoint because thereis no overlap in the breakBRD population between z = . z = .
0. For the quiescent sample,we find the maximum SFR concentration since thegalaxy was on the star-forming main sequence (atlog(sSFR / yr) > − z = . Figure 13.
Similar to Figure 12 but with peak SFR con-centration plotted against the sSFR at the time of peakSFR. Moreover, here we show the parent galaxy sam-ple distribution weighted to have the same present-daystellar mass distribution as the z = . / quenched satellites / centralsare all weighted separately. As in Figure 12 the break-BRD sample is not included in either the star-formingor quiescent parent sample. BRD analogues. We note that high SFR concen-tration does not necessarily mean a galaxy would meet the breakBRD selection criteria, as this addi-tionally requires red disk color (see Section 2.2).Figure 12 shows the maximum SFR concentra-tion for the unweighted parent sample in grey,while Figure 13 shows maximum SFR concentra-tion for the weighted parent sample. Note thatthe star-forming / quenched satellites / centrals havetheir weights calculated separately, making theirweightings slightly di ff erent than those used inSection 3. Additionally, the parent sample ex-cludes the breakBRD subsample in both of thesefigures, in contrast to the rest of this work. Thecentral (blue; left) and satellite (orange; right) z = . z = . he break BRD B reakdown total sSFR at that time. This is be-cause in addition to SFR concentration, breakBRDgalaxies tend to cluster in total sSFR space (Fig-ures 3 & 10). The combination of star-formingcenter and red disks tends to make breakBRDs sitsomewhat below the star-forming main sequencewhile still well above the sSFR of quenched galax-ies, so this may give us more discriminating powerin our comparison with the parent sample.We first focus on the star-forming galaxies inthe top panels of Figures 12 and 13. Whether weconsider the weighted (Figure 12) or unweighted(Figures 13) samples, we see that the maximumconcentration of star-forming galaxies extends tomuch lower values than the breakBRD sample.This is also reflected in their median + MAD val-ues. The central and satellite breakBRD MADranges overlap with each other, but generally notwith their respective parent populations. ThebreakBRDs only have overlapping median + MADswith the weighted quenched parent population. Wenote that at the time of maximum concentration,the sSFR of star-forming parent galaxies extendsto higher values than the breakBRD analogue sam-ple. This is very dramatic in the central parentsample, but can be seen even in the star-formingsatellite parent galaxies. We therefore concludethat the breakBRD state is unlikely to be a re-sult of stochastically-distributed star formation ingalaxies on the star-forming main sequence. Thiscomes as no surprise since we found that break-BRD galaxies (identified at z = .
5) are likely toquench.We next focus on the quenched galaxies in thebottom panels of Figures 12 and 13. When we ex-amine the unweighted sample, we clearly see that most quenched galaxies have not passed througha breakBRD state: there is a the large numberof galaxies with low concentrations, especially athigher masses. Focusing on the unweighted par-ent sample in Figure 12 allows us to highlight theunusual mass distribution of breakBRD galaxieswith respect to quenched galaxies—they tend to belower mass. At the mass range typical for break-BRD galaxies, there seems to be a high fraction ofquenched galaxies that have experienced centrallyconcentrated star formation in their past.When looking at the weighted sample in Figure13, which emphasises the breakBRD mass range,we see that nearly all quenched satellites andabout half of the quenched centrals have had highSFR concentrations, indicative of a breakBRD-like state. Looking at central galaxies in detail, wesee that the median is below most of the range ofthe breakBRD sample, but because the distributionis so extended there is significant overlap in theconcentrations. This state is therefore more com-mon among quenching galaxies with lower stel-lar masses, but even among low-mass quenchinggalaxies the breakBRD state is far from ubiquitous.Because both central star formation (measuredusing D n + MAD values in Fig-ure 13 indicate. We can begin to see this whenwe examine the range of sSFR values in the par-ent sample in Figure 13. We see that for galax-ies quenched at z =
0, the sSFR at maximum SFRconcentration can be much lower than that found inthe breakBRD sample, indicating that these galax-ies might not meet the the star-formation thresholdfor D n < . openhafer et al .parent sample is the simulation snapshot with themaximum SFR concentration. There may be othertimes when the SFR concentration is still quitehigh and the total sSFR is more in line with thatfound for breakBRDs. We can only draw conclu-sions about the fraction of galaxies that could reachthe breakBRD state criteria based on this simpleproxy.By searching for the maximum SFR concen-tration of quenched parent galaxies, we concludethat the majority of these galaxies do not expe-rience a breakBRD state. Moreover, when welook at quenched parent galaxies at similar stellarmasses as the breakBRDs, at least half of centralgalaxies never reach high SFR concentrations.Thisdi ff erence in SFR concentrations between high-and low-mass quiescent galaxies does suggest thatthese galaxies can experience di ff erent quenchingpathways in IllustrisTNG. Nevertheless, we con-clude that breakBRD galaxies remain an unusualsample of quenching galaxies, even when attempt-ing to align our samples to the same point in theirevolution.5.4. Comparison with the Observational Sample
In this paper we have specifically chosen tocompare galaxies within the TNG100 simulation,rather than compare simulated galaxies directlyto observed galaxies. This way, we mitigate any“systematic” errors in the simulation, which weknow does not exactly reproduce the observed uni-verse. For instance, for the stellar mass range ofthe breakBRDs, the TNG black hole mass-stellarmass relation tends towards the higher end of whatis observed (Habouzit et al. 2020), and the satel-lite quenched fraction is higher than seen in SDSS(Donnari et al. 2020). That said, because we areusing this work to draw conclusions about the evo-lution of breakBRD galaxies, it is worthwhile tomake some comparisons between the simulatedand observed breakBRD galaxy samples.Looking at the stellar mass distribution of thebreakBRD observed and analogue samples, wefind that they are quite similar. Although we re- quire that IllustrisTNG galaxies have M ∗ > M (cid:12) , about 80% of observed breakBRDs are abovethis mass. The majority of observed breakBRDsare between 10 - 10 . M (cid:12) (see Figure 3 inTT20), roughly agreeing with the tendency ofbreakBRD analogues to have lower masses withinour selected range of 10 - 10 M (cid:12) (Figure 1).Figure 14 compares the sSFR and color distri-butions of the observed and simulated breakBRDsamples, as well as the weighted TNG parent sam-ple. Note that a mass cut of log( M ∗ ) >
10 M (cid:12) has been applied the observed breakBRD sample,in agreement with the mass cut used to select theTNG parent sample (Section 2.1).We first focus on the top panel, which comparesthe sSFR of the aforementioned samples. TheTNG data combines satellites and centrals fromFigure 3. The observed parent is omitted, but itsstar-forming component (log(sSFR / yr) > −
10) hasan sSFR distribution that is very similar to theweighted TNG parent’s. With this in mind, we cansee that the sSFR distribution of observed break-BRDs (green points) is similar to that of its star-forming parent sample, while the simulated break-BRDs (blue points) have sSFRs lower than the me-dian of their parent samples. This may indicatethat the breakBRDs in the simulation have alreadybegun quenching while the observed galaxies arestill on the star-forming sequence. However, theglobal sSFR from Brinchmann et al. (2004) is usedin TT20. This uses the emission lines in the SDSSspectra from the central region (the region coveredby the spectral fiber) as well as the optical colorsfrom the SDSS photometry inside and outside thefiber regions to place galaxies on a global sSFRrelation. Brinchmann et al. (2004) note however,that the likelyhoods of SFR / L i for redder colorsare broader and sometimes multipeaked, and theSFR / L i is therefore less well constrained. Becausethe breakBRD galaxies have unusual central starformation rates for their red disk colors, the globalsSFRs from Brinchmann et al. (2004) may be moreuncertain for these galaxies. he break BRD B reakdown g − r ) and ( u − r ) colors of the sim-ulated and observed galaxies. We see that par-ticularly in the ( g − r ) galaxy color, the observed(green points) and simulated (blue points) break-BRD galaxies are quite similar, although more ofthe observed breakBRDs extend to bluer colors. Inaddition we have shown the observed “green val-ley" with black dashed lines, as fit using color-stellar mass diagrams (Mendel et al. 2013; Schaw-inski et al. 2014). We also have found the mini-mum of the color distributions in the TNG100 par-ent sample as a function of stellar mass. It is nearlyconstant with mass, and we denote the ± g − r ) colors for the simulatedsample are bluer with respect to the parent than theobserved sample. The ( u − r ) colors are similarwith respect to the valley in both the simulated andobserved samples.Finally, we compare the environmental measuresof the observed and simulated breakBRD galaxies.The satellite fraction of observed breakBRD galax-ies seems to be quite similar to the observed par-ent sample (40% and 39%, respectively), while thesatellite fraction of breakBRD analogues is muchlarger than that of the parent sample (see Table1). However, for both the observed and analoguebreakBRD samples, the environmental density issimilar to that of their parent samples, even whenincluding splashback galaxies (which cannot be re-moved from the observed sample).In summary, we find that there is generally goodagreement between the observed and simulatedsamples. However, when compared to the parentsamples, the sSFRs of the simulated breakBRDstend to be lower than the sSFRs of the observedbreakBRDs, and the satellite fraction of breakBRDgalaxies is comparatively higher in the simulationsthan in the observations. Figure 14.
Whole-galaxy sSFR (top), g − r (middle) and u − r (bottom) for the full breakBRD analogue sample(blue solid) and its weighted parent (greyscale). ThesSFRs and colors of the observed breakBRD sample(Tuttle & Tonnesen 2020) with log( M / M (cid:12) ) >
10 areshown as green open circles and thin lines. In the lowertwo plots, black dashed lines show the “green valley”(Mendel et al. 2013; Schawinski et al. 2014), and reddot-dashed lines bracket the “valley” present in the col-ors of the un-weighted TNG parent; see text for moredetails. 6.
CONCLUSIONTuttle & Tonnesen (2020) discovered an unusualsample of galaxies in observations, called break-BRDs, with red disks and recent star formationin their centers. By generating synthetic observa-tions, we applied similar cuts to galaxies within Il-8 K openhafer et al .lustrisTNG to find galaxies in a state analogous tothe observed breakBRDs. These cuts were appliedat multiple redshifts: 0.0, 0.03, 0.1, and 0.5. • IllustrisTNG contains galaxies that are anal-ogous to the observational sample of TT20in terms of their disk colors and D n • Our color- and D n > . − ) gas mass is low andcentrally concentrated in bBRDa galaxies(Figure 5) • The bBRDa populations have a higher satel-lite fraction than in the parent sample (Sec-tion 3.2.2; Table 1). This aligns with ouridea that environmental e ff ects could drivecentrally-concentrated star formation. Wealso find a somewhat higher splashback frac-tion in the central galaxies of the bBRDasample compared to the central parent sam-ple (Section 3.2.2; Table 2). Together thissuggests that environment may have been adriver in forming a significant fraction of ourbreakBRD sample. • Central breakBRD galaxies at z = z = . ff erences between the samples are not sig-nificant. • When we only consider central galaxies thathave not been splashbacks since z = .
5, we find no clear cause for the breakBRD state.The black hole mass distribution and localenvironment are similar between the bBRDaand parent samples (Figures 6 & 7). Mergersare similarly prevalent in both samples, withthe exception that central breakBRD galax-ies have experienced less very minor merg-ers or clumpy cosmological accretion (Sec-tion 3.2.3). We infer that any environmentaldriver must therefore be subtle, and possiblyact over long timescales. • Within IllustrisTNG, the breakBRD-analoguestate is a transient one, lasting between a fewhundred Myr to ∼ • We find that breakBRD analogues at z = . z = • We argue that breakBRD galaxies are aunique population of quenching galaxies.They are of lower stellar mass than mostquenching or quiescent galaxies. By-and-large, both star-forming and quiescent galax-ies do not reach the same high SFR anddense-gas concentrations exhibited by thebreakBRD samples (Figures 12 and 13)and therefore never experience a breakBRDstate. However, present-day low-mass qui-escent galaxies are more likely to have ex-perienced a centrally-concentrated phase intheir past.This paper was motivated by the observationalbreakBRD sample from TT20, in the hopes thatIllustrisTNG could provide an explanation for theappearance of the observed sample. Our findingsplace breakBRDs in a unique space outside of the he break
BRD B reakdown asignificant fraction of quenching, low stellar masscentral galaxies undergo a phase of centrally-concentrated star formation (Figure 12 and 13).Therefore, understanding the breakBRD state inmore detail may be an important step in under-standing the process by which low mass centralgalaxies quench.It is worth noting that breakBRD galaxies are notalone in indicating that galaxies may follow severalpaths to red, quenched, early-type galaxies. For ex-ample, Suh et al. (2010) identify early-type galax-ies in the Sloan Digital Sky Survey DR6, and findthat about 30% of their sample are “blue-cored”.They find that these early-type galaxies tend to belower mass and posit that they may be formed viamergers with gas-rich galaxies, while the majorityof early-type galaxies are formed via equal-mass“dry” mergers. Evans et al. (2018) find a popu-lation of “red misfits” characterised by red opticalcolors and high sSFRs, and conclude that they arelikely to be gradually quenching via internal pro-cesses.Another interesting class of galaxies are passivespirals, first found in clusters (van den Bergh 1976;Poggianti et al. 1999). More recent studies havefound passive spirals across environments (Mas-ters et al. 2010; Bamford et al. 2009). Bundy et al.(2010) argue that up to 60% of spirals may passthrough this phase on the way to the red sequence.Interestingly, the authors find that passive spiralsare bulge-dominated at all masses, indicating that the simple fading of disks is not a viable formationmechanism. Given that we find that nearly halfof quenching central galaxies (and most quench-ing satellite galaxies) in IllustrisTNG pass througha breakBRD-like state (Figure 13), more study of apossible connection between these galaxy classesmay be warranted.Through our analysis of breakBRD galaxies wehave made testable predictions that require moredetailed observations of the gas distribution of ob-served breakBRD galaxies. Specifically, we ar-gue that the observed breakBRD population willshow normal central gas masses and apparent outerdeficits (see Section 5.1). Given that breakBRDanalogues within IllustrisTNG appear to quench, itwill also be useful to search the observed samplefor additional signs of quenching. Both of our pre-dictions can be evaluated by searching for the gassupply of breakBRD galaxies, which should be lowoverall but normal in the galaxy center. ACKNOWLEDGEMENTSThe authors would like to thank the referee forcomments that improved this work. It is a pleasureto thank Shy Genel, Vicente Rodriguez-Gomez,and Brian W. O’Shea for valuable suggestions anddiscussion. The authors would like to thank theIllustrisTNG collaboration for making their datapublic. CK is supported by the Department ofEnergy Computational Science Graduate Fellow-ship program (DE-FG02-97ER25308). This re-search was supported in part through the compu-tational resources and sta ff contributions providedby the Quest high performance computing facilityat Northwestern University, which is jointly sup-ported by the O ffi ce of the Provost, the O ffi ce forResearch, and Northwestern University Informa-tion Technology, and by the Institute for Cyber-Enabled Research at Michigan State University.The data used in this work were, in part, hostedon facilities supported by the Scientific Comput-ing Core at the Flatiron Institute, a division of theSimons Foundation, and the analysis was largelydone using those facilities. This work was initi-0 K openhafer et al .ated as a project for the Kavli Summer Program inAstrophysics held at the Center for ComputationalAstrophysics of the Flatiron Institute in 2018. Theprogram was co-funded by the Kavli Foundationand the Simons Foundation. We thank them fortheir generous support. The Flatiron Institute issupported by the Simons Foundation. Software:
Astropy (Astropy Collaborationet al. 2018), FSPS (Conroy et al. 2009; Conroy &Gunn 2010), python-fsps (Foreman-Mackey et al.2014), and SciPy (esp. NumPy & Matplotlib) (Vir-tanen et al. 2020).REFERENCES
Abazajian, K. N., Adelman-McCarthy, J. K., Agüeros,M. A., et al. 2009, ApJS, 182, 543Abramson, L. E., Kelson, D. D., Dressler, A., et al.2014, ApJL, 785, L36Adhikari, S., Dalal, N., & Chamberlain, R. T. 2014,JCAP, 2014, 019Aihara, H., Allende Prieto, C., An, D., et al. 2011,ApJS, 193, 29Astropy Collaboration, Price-Whelan, A. M., Sip˝ocz,B. M., et al. 2018, The Astronomical Journal, 156,123Bait, O., Barway, S., & Wadadekar, Y. 2017, MNRAS,471, 2687Balogh, M. L., Morris, S. L., Yee, H. K. C., Carlberg,R. G., & Ellingson, E. 1999, ApJ, 527, 54Bamford, S. P., Nichol, R. C., Baldry, I. K., et al. 2009,MNRAS, 393, 1324Belfiore, F., Maiolino, R., Tremonti, C., et al. 2017,MNRAS, 469, 151Belfiore, F., Maiolino, R., Bundy, K., et al. 2018,MNRAS, 477, 3014Bell, E. F., van der Wel, A., Papovich, C., et al. 2012,ApJ, 753, 167Blanton, M. R., Schlegel, D. J., Strauss, M. A., et al.2005, AJ, 129, 2562Bluck, A. F. L., Mendel, J. T., Ellison, S. L., et al.2014, MNRAS, 441, 599Blumenthal, K. A., & Barnes, J. E. 2018, MNRAS,479, 3952Boselli, A., & Gavazzi, G. 2006, PASP, 118, 517Brinchmann, J., Charlot, S., White, S. D. M., et al.2004, MNRAS, 351, 1151Bryan, G. L., & Norman, M. L. 1998, ApJ, 495, 80Bryant, J. J., Owers, M. S., Robotham, A. S. G., et al.2015, MNRAS, 447, 2857Buck, T., Macciò, A. V., Dutton, A. A., Obreja, A., &Frings, J. 2019, MNRAS, 483, 1314Bundy, K., Scarlata, C., Carollo, C. M., et al. 2010,ApJ, 719, 1969 Bundy, K., Bershady, M. A., Law, D. R., et al. 2015,ApJ, 798, 7Burkert, A., Truran, J. W., & Hensler, G. 1992, ApJ,391, 651Byrd, G., & Valtonen, M. 1990, ApJ, 350, 89Cassata, P., Giavalisco, M., Guo, Y., et al. 2010, ApJL,714, L79Chabrier, G. 2003, PASP, 115, 763Charlot, S., & Fall, S. M. 2000, ApJ, 539, 718Chen, Y.-M., Shi, Y., Wild, V., et al. 2019, MNRAS,489, 5709Chiappini, C., Matteucci, F., & Gratton, R. 1997, ApJ,477, 765Choi, J., Dotter, A., Conroy, C., et al. 2016, ApJ, 823,102Conroy, C., & Gunn, J. E. 2010, ApJ, 712, 833Conroy, C., Gunn, J. E., & White, M. 2009, ApJ, 699,486Cox, T. J., Jonsson, P., Somerville, R. S., Primack,J. R., & Dekel, A. 2008, MNRAS, 384, 386Crain, R. A., Schaye, J., Bower, R. G., et al. 2015,MNRAS, 450, 1937Croom, S. M., Lawrence, J. S., Bland-Hawthorn, J.,et al. 2012, MNRAS, 421, 872Dale, D. A., Beltz-Mohrmann, G. D., Egan, A. A.,et al. 2016, AJ, 151, 4Davé, R., Anglés-Alcázar, D., Narayanan, D., et al.2019, MNRAS, 486, 2827Davé, R., Rafieferantsoa, M. H., & Thompson, R. J.2017, MNRAS, 471, 1671Davis, M., Efstathiou, G., Frenk, C. S., & White,S. D. M. 1985, ApJ, 292, 371Diemand, J., Kuhlen, M., & Madau, P. 2007, ApJ, 667,859Diemer, B. 2020a, arXiv e-prints, arXiv:2007.10992—. 2020b, arXiv e-prints, arXiv:2007.09149Diemer, B., Mansfield, P., Kravtsov, A. V., & More, S.2017, ApJ, 843, 140 he break
BRD B reakdown Ding, X., Silverman, J., Treu, T., et al. 2020, ApJ, 888,37Donnari, M., Pillepich, A., Nelson, D., et al. 2020,arXiv e-prints, arXiv:2008.00004Dotter, A. 2016, ApJS, 222, 8Ellison, S. L., Patton, D. R., Mendel, J. T., & Scudder,J. M. 2011, MNRAS, 418, 2043Ellison, S. L., Sánchez, S. F., Ibarra-Medel, H., et al.2018, MNRAS, 474, 2039Ellison, S. L., Viswanathan, A., Patton, D. R., et al.2019, MNRAS, 487, 2491Evans, F. A., Parker, L. C., & Roberts, I. D. 2018,MNRAS, 476, 5284Falcón-Barroso, J., Sánchez-Blázquez, P., Vazdekis,A., et al. 2011, A&A, 532, A95Fang, J. J., Faber, S. M., Koo, D. C., & Dekel, A. 2013,ApJ, 776, 63Foreman-Mackey, D., Sick, J., & Johnson, B. 2014,Python-Fsps: Python Bindings to FSPS (v0.1.1),Zenodo, , , doi:10.5281 / zenodo.12157.https: // zenodo.org / record / // doi.org / / mnras / stz030McAlpine, S., Harrison, C. M., Rosario, D. J., et al.2020, MNRAS, 494, 5713McPartland, C., Sanders, D. B., Kewley, L. J., &Leslie, S. K. 2019, MNRAS, 482, L129Medling, A. M., Cortese, L., Croom, S. M., et al. 2018,MNRAS, 475, 5194Mehlert, D., Thomas, D., Saglia, R. P., Bender, R., &Wegner, G. 2003, A&A, 407, 423 openhafer et al . Mendel, J. T., Simard, L., Ellison, S. L., & Patton,D. R. 2013, MNRAS, 429, 2212Méndez-Abreu, J., Sánchez, S. F., & deLorenzo-Cáceres, A. 2019, MNRAS, 488, L80Mihos, J. C., & Hernquist, L. 1994, ApJL, 425, L13More, S., Kravtsov, A. V., Dalal, N., & Gottlöber, S.2011, ApJS, 195, 4Morselli, L., Popesso, P., Cibinel, A., et al. 2019,A&A, 626, A61Moustakas, J., Kennicutt, Robert C., J., & Tremonti,C. A. 2006, ApJ, 642, 775Naab, T., Oser, L., Emsellem, E., et al. 2014, MNRAS,444, 3357Naiman, J. P., Pillepich, A., Springel, V., et al. 2018,MNRAS, 477, 1206Nelson, D., Pillepich, A., Springel, V., et al. 2018,MNRAS, 475, 624Nelson, D., Springel, V., Pillepich, A., et al. 2019,Computational Astrophysics and Cosmology, 6, 2Nelson, E. J., van Dokkum, P. G., Förster Schreiber,N. M., et al. 2016, ApJ, 828, 27Omand, C. M. B., Balogh, M. L., & Poggianti, B. M.2014, MNRAS, 440, 843Pagel, B. E. J., & Edmunds, M. G. 1981, ARA&A, 19,77Pakmor, R., Bauer, A., & Springel, V. 2011, MNRAS,418, 1392Pan, Z., Li, J., Lin, W., et al. 2015, ApJL, 804, L42Paxton, B., Bildsten, L., Dotter, A., et al. 2011, ApJS,192, 3Paxton, B., Cantiello, M., Arras, P., et al. 2013, ApJS,208, 4Paxton, B., Marchant, P., Schwab, J., et al. 2015, ApJS,220, 15Pérez, E., Cid Fernandes, R., González Delgado,R. M., et al. 2013, ApJL, 764, L1Pillepich, A., Nelson, D., Hernquist, L., et al. 2018,MNRAS, 475, 648Pillepich, A., Springel, V., Nelson, D., et al. 2018,MNRAS, 473, 4077Poetrodjojo, H., Groves, B., Kewley, L. J., et al. 2018,MNRAS, 479, 5235Poggianti, B. M., Smail, I., Dressler, A., et al. 1999,ApJ, 518, 576Prasad, D., Voit, G. M., O’shea, B. W., & Glines, F.2020, arXiv e-prints, arXiv:2006.10809Quai, S., Pozzetti, L., Moresco, M., et al. 2019,MNRAS, 490, 2347Queyrel, J., Contini, T., Kissler-Patig, M., et al. 2012,A&A, 539, A93 Rowlands, K., Heckman, T., Wild, V., et al. 2018,MNRAS, 480, 2544Sanchez, S. F. 2019, arXiv e-prints, arXiv:1911.06925Sánchez, S. F., Kennicutt, R. C., Gil de Paz, A., et al.2012, A&A, 538, A8Sánchez, S. F., Rosales-Ortega, F. F., Iglesias-Páramo,J., et al. 2014, A&A, 563, A49Sánchez-Blázquez, P., Forbes, D. A., Strader, J.,Brodie, J., & Proctor, R. 2007, MNRAS, 377, 759Sánchez-Blázquez, P., Peletier, R. F., Jiménez-Vicente,J., et al. 2006, MNRAS, 371, 703Sánchez-Menguiano, L., Sánchez, S. F., Pérez, I., et al.2018, A&A, 609, A119Schawinski, K., Urry, C. M., Simmons, B. D., et al.2014, MNRAS, 440, 889Shaver, P. A., McGee, R. X., Newton, L. M., Danks,A. C., & Pottasch, S. R. 1983, MNRAS, 204, 53Spilker, J. S., Bezanson, R., Weiner, B. J., Whitaker,K. E., & Williams, C. C. 2019, ApJ, 883, 81Springel, V. 2010, MNRAS, 401, 791Springel, V., White, S. D. M., Tormen, G., &Kau ff mann, G. 2001, MNRAS, 328, 726Springel, V., Pakmor, R., Pillepich, A., et al. 2018,MNRAS, 475, 676Starkenburg, T. K., Tonnesen, S., & Kopenhafer, C.2019, ApJL, 874, L17Su, K.-Y., Hopkins, P. F., Hayward, C. C., et al. 2019,MNRAS, 487, 4393Suh, H., Jeong, H., Oh, K., et al. 2010, ApJS, 187, 374Tamburri, S., Saracco, P., Longhetti, M., et al. 2014,A&A, 570, A102Tonnesen, S., & Bryan, G. L. 2012, MNRAS, 422,1609Torrey, P., Snyder, G. F., Vogelsberger, M., et al. 2015,MNRAS, 447, 2753Tuttle, S. E., & Tonnesen, S. 2020, ApJ, 889, 188van den Bergh, S. 1976, ApJ, 206, 883van den Bosch, F. C. 1998, ApJ, 507, 601Vila-Costas, M. B., & Edmunds, M. G. 1992, MNRAS,259, 121Virtanen, P., Gommers, R., Oliphant, T. E., et al. 2020,Nature Methods, 17, 261Vogelsberger, M., Genel, S., Springel, V., et al. 2014,MNRAS, 444, 1518Voit, G. M., Bryan, G. L., Prasad, D., et al. 2020, arXive-prints, arXiv:2006.09381Wang, X., Jones, T. A., Treu, T., et al. 2019, ApJ, 882,94Weinberger, R., Springel, V., Hernquist, L., et al. 2017,MNRAS, 465, 3291 he break BRD B reakdown33