Interacting galaxies in the IllustrisTNG simulations -- II: Star formation in the post-merger stage
Maan H. Hani, Hayman Gosain, Sara L. Ellison, David R. Patton, Paul Torrey
MMNRAS , 1–18 (2020) Preprint 14 February 2020 Compiled using MNRAS L A TEX style file v3.0
Interacting galaxies in the IllustrisTNG simulations – II:Star formation in the post-merger stage.
Maan H. Hani, (cid:63) † Hayman Gosain, , Sara L. Ellison, David R. Patton, Paul Torrey Department of Physics and Astronomy, University of Victoria, Victoria, British Columbia, V8P 1A1, Canada Department of Physical Sciences, Indian Institute of Science Education and Research Mohali, Manauli, Punjab 140306, India Department of Physics and Astronomy, Trent University, 1600 West Bank Drive, Peterborough, ON K9L 0G2, Canada Department of Astronomy, University of Florida, 211 Bryant Space Sciences Center, Gainesville, FL, USA
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
Galaxy mergers are a major evolutionary transformation whose effects are borne outby a plethora of observations and numerical simulations. However, most previous sim-ulations have used idealised, isolated, binary mergers and there has not been signifi-cant progress on studying statistical samples of galaxy mergers in large cosmologicalsimulations. We present a sample of 27,691 post-merger (PM) galaxies ( ≤ z ≤ )identified from IllustrisTNG: a cosmological, large box, magneto-hydrodynamical sim-ulation suite. The PM sample spans a wide range of merger and galaxy properties( M (cid:63) , µ , f gas ). We demonstrate that star forming (SF) PMs exhibit enhanced star for-mation rates (SFRs) on average by a factor of ∼ , while the passive PMs show nostatistical enhancement. We find that the SFR enhancements: (1) show no dependenceon redshift, (2) anti-correlate with the PM’s stellar mass, and (3) correlate with thegas fraction of the PM’s progenitors. However, SF PMs show stronger enhancementswhich may indicate other processes being at play (e.g., gas phase, feedback efficiency).Although the SFR enhancement correlates mildly with the merger mass ratio, themore abundant minor mergers ( . ≤ µ < . ) still contribute ∼ of the total SFRenhancement. By tracing the PM sample forward in time, we find that galaxy mergerscan drive significant SFR enhancements which decay over ∼ . Gyr independent of themerger mass ratio, although the decay timescale is dependent on the simulation reso-lution. The strongest merger-driven starburst galaxies evolve to be passive/quenchedon faster timescales than their controls.
Key words: galaxies: evolution – galaxies: star formation – galaxies: interactions –methods: numerical
Galaxy mergers are at the foundation of the hierarchicalmodel for structure formation (White & Rees 1978; Lacey &Cole 1993). Through mergers, galaxies can grow their stel-lar, gas, and dark matter content. However, the effects ofgalaxy mergers are not limited to the growth of galaxies.There exists a rich body of theoretical predictions and ob-servational evidence linking mergers to triggering and accel-erating gas evolution (i.e., depletion, ejection, enrichment, (cid:63)
E-mail: [email protected] † Vanier Scholar cooling/heating) thus confirming that galaxy mergers are amajor rite of passage in galaxy evolution.Observationally, a profusion of studies report severalwell accepted observational signatures of galaxy mergers. Adilution in the central metallicity has been reported by ob-servational studies of galaxy pairs (e.g., Kewley et al. 2006;Ellison et al. 2008; Scudder et al. 2012; Ellison et al. 2018),and post-mergers (Ellison et al. 2013; Thorp et al. 2019).Additionally, merging galaxies exhibit enhanced morpholog-ical disturbances (Casteels et al. 2014; Patton et al. 2016),increased atomic and molecular gas fractions (e.g., Ellisonet al. 2015, 2018; Pan et al. 2018; Violino et al. 2018; Duttaet al. 2018, 2019), elevated AGN fractions (e.g., Ellison et al.2011; Satyapal et al. 2014; Ellison et al. 2019), and enhanced © a r X i v : . [ a s t r o - ph . GA ] F e b M. H. Hani et al.
SFRs (e.g., Ellison et al. 2008, 2013; Patton et al. 2013;Knapen et al. 2015; Thorp et al. 2019) when compared totheir non-interacting counterparts. Coupled with enhancedstar formation and AGN activity, feedback processes (e.g.,stellar & AGN feedback) play a vital role in the evolutionof mergers. Galactic outflows have been linked to enhancedstar formation (e.g., Martin 2005; Rupke et al. 2005a; Strick-land & Heckman 2009) and AGN activity (e.g., Rupke et al.2005b; Veilleux et al. 2013; Woo et al. 2017).The observational signatures of galaxy interactionsand mergers are complemented by a well-posed theoret-ical framework. Galaxy interactions trigger strong non-axisymmetric features which create tidal torques capableof driving instabilities in the dynamically cool inter-stellarmedium (ISM) thus causing strong inflows of gas into thegalactic centre (e.g., Hernquist 1989; Barnes & Hernquist1991; Mihos & Hernquist 1996; Blumenthal & Barnes 2018).The gravitational instability-driven gas inflow manifests insimulations as a dilution in central metallicities (e.g., Mon-tuori et al. 2010; Torrey et al. 2012; Moreno et al. 2015;Bustamante et al. 2018), an enhancement in central star for-mation rates (SFRs; e.g., Cox et al. 2008; Di Matteo et al.2008; Moreno et al. 2015; Sparre & Springel 2016), and anincrease in accretion onto the central super-massive blackhole thus giving rise to an active galactic nucleus (AGN;e.g., Di Matteo et al. 2005; Hopkins & Quataert 2010). Theassociated feedback processes often manifest (or are imple-mented) as outflows (e.g., Hayward & Hopkins 2017; Morenoet al. 2019) which can shape the ISM and even the circum-galactic medium (e.g., Cox et al. 2004; Hani et al. 2018).Additionally, mergers’ descendants possess unique morpho-logical disturbances such as shells, ripples, tidal tails, andtidal plumes (e.g., Di Matteo et al. 2007; Lotz et al. 2008,2010a,b; Pop et al. 2018).Numerical simulations provide a controlled environmentto experiment and investigate the physical process at playduring a galaxy merger. Such controlled experiments are of-ten carried out in simulations of binary mergers where thesame galaxies that partake in the merger are evolved in iso-lation (e.g., Torrey et al. 2012; Patton et al. 2013; Morenoet al. 2015, 2019). While isolated binary merger suites pro-vide a controlled environment to study the details of galaxymergers, they are limited in their level of realism (non-cosmological environment, lack of circum-galactic gas andfully realistic range of parameters such as galaxy morphol-ogy, gas fraction, and orbital geometry), and therefore donot represent the majority of observed galaxies (see Morenoet al. 2013).Cosmological zoom-in simulations provide a more re-alistic, yet still limited (e.g., galaxy morphology, gas frac-tion, orbital geometry), approach to study galaxy mergerswhere galaxies are evolved from cosmological initial condi-tions (e.g., Sparre & Springel 2016; Bustamante et al. 2018;Hani et al. 2018). One particular drawback of such simula-tions is the limited ability to trace the galaxies for a longtime after the merger. In cosmological zoom-in simulations,the effects of the merger are often compared to a pre-mergerstate (e.g., Sparre & Springel 2016; Hani et al. 2018); how-ever, on long timescales, post-mergers evolve significantlyand therefore should not be compared to the pre-mergerstate.While both binary merger simulations and cosmological zoom-in simulations provide powerful tools to study galaxymergers at high resolution (spatial and temporal), they bothlack the large and diverse galaxy samples of large-box cos-mological simulations. Recent studies have begun to investi-gate the effect of galaxy mergers in large cosmological boxes(e.g., Blumenthal et al. 2019; Patton et al. 2020; Rodr´ıguezMontero et al. 2019).The work presented here leverages the sizeable sampleof galaxy mergers available in large cosmological box hydro-dynamical simulations to investigate the impact of mergerson galactic star formation during the post-merger stage (i.e.,post-coalescence). Using numerical simulations mitigates theunconstrained merger timescales in observational studies,while the large simulation box provides a diverse and repre-sentative galaxy sample therefore alleviating a major chal-lenge for idealised and cosmological zoom-in simulations. Weidentify galaxy mergers in the IllustrisTNG simulation suite(Marinacci et al. 2018; Naiman et al. 2018; Nelson et al.2018; Pillepich et al. 2018b; Springel et al. 2018; Nelsonet al. 2019a) and then analyse our post-merger galaxy sam-ple by implementing a methodology akin to observations. Bycomparing the post-merger star formation rates to controlswhich have not undergone a merger, we are able to isolatethe effects of the merger on star formation rates (e.g., Pattonet al. 2020).The manuscript is structured as follows: We first intro-duce the methodology in Section 2. In Section 3 we presentthe results highlighting the role of galaxy mergers induc-ing star formation and we investigate major drivers to thestrength of the SFR enhancements. Section 4 discusses theeffects of control matching and simulation resolution, andties our results to observational studies. Finally, we sum-marise our conclusions in Section 5.
The work presented here investigates the signatures ofgalaxy mergers during the post-merger stage. We identifygalaxy mergers in the IllustrisTNG simulation suite (Mari-nacci et al. 2018; Naiman et al. 2018; Nelson et al. 2018;Pillepich et al. 2018b; Springel et al. 2018; Nelson et al.2019a) and compare the properties of the mergers’ descen-dants to galaxies which have not undergone a merger in theirrecent evolution. In this section, we describe the numericalsimulations (i.e., IllustrisTNG simulations), the details ofour merger-identification, and the process of generating astatistical control galaxy sample for comparison.
The work presented here is primarily focused on quantify-ing the effects of galaxy mergers beyond coalescence (post-mergers). We employ the IllustrisTNG simulation suite(Marinacci et al. 2018; Naiman et al. 2018; Nelson et al.2018; Pillepich et al. 2018b; Springel et al. 2018; Nelsonet al. 2019a) which consists of several large box cosmo-logical magneto-hydrodynamical simulations. We focus ouranalysis on the highest resolution runs of the largest twovolumes of the IllustrisTNG simulations – TNG100-1 andTNG300-1. Using such large volumes ( . Mpc and MNRAS000
The work presented here is primarily focused on quantify-ing the effects of galaxy mergers beyond coalescence (post-mergers). We employ the IllustrisTNG simulation suite(Marinacci et al. 2018; Naiman et al. 2018; Nelson et al.2018; Pillepich et al. 2018b; Springel et al. 2018; Nelsonet al. 2019a) which consists of several large box cosmo-logical magneto-hydrodynamical simulations. We focus ouranalysis on the highest resolution runs of the largest twovolumes of the IllustrisTNG simulations – TNG100-1 andTNG300-1. Using such large volumes ( . Mpc and MNRAS000 , 1–18 (2020) llustrisTNG post-merger galaxies . Mpc for TNG100-1 and TNG300-1, respectively)ensures a substantial sample of galaxies ideal for a statis-tical study of galaxy mergers. The galaxy sample is sim-ulated in a fully cosmological environment, and it has re-alistic evolutionary histories, and galactic properties (e.g.,Nelson et al. 2018; Pillepich et al. 2018b; Springel et al.2018; Donnari et al. 2019; Huertas-Company et al. 2019;Rodriguez-Gomez et al. 2019; Tacchella et al. 2019; Torreyet al. 2019). While TNG300-1 provides larger galaxy andmerger samples, TNG100-1 allows us to investigate the ef-fects of galaxy mergers at higher mass and spatial resolu-tions. TNG100-1 (TNG300-1) has a dark matter mass res-olution m dm = . × M (cid:12) ( m dm = . × M (cid:12) ), and abaryonic mass resolution m b ∼ M (cid:12) ( m b ∼ M (cid:12) ). Wealso use TNG100-2 to ensure model convergence and investi-gate resolution effects; TNG100-2 is a re-run of TNG100-1 ata roughly equal resolution to TNG300-1. All the simulationshave a temporal resolution ∼ Myr.For most of our analysis we will report results fromTNG300-1 as our fiducial simulation because the large boxsize provides exquisite statistics. Nonetheless, the results arebroadly consistent between TNG300-1 and TNG100-1. Werefer the reader to Section 4.2 where we discuss the effectsof varying the simulation resolution on our conclusions.The IllustrisTNG galaxy formation model which wasintroduced in Weinberger et al. (2017) and Pillepich et al.(2018a) builds on its predecessor, the Illustris model (Vogels-berger et al. 2013; Torrey et al. 2014), with several additionsand modifications to the numerical framework and physicalmodel. The IllustrisTNG physical model includes:(i) Star formation: Star formation occurs in a pressurisedinterstellar medium (ISM) for n H (cid:38) . cm − following theSpringel & Hernquist (2003) formalism. Star particles rep-resent stellar populations with a Chabrier (Chabrier 2003)initial mass function.(ii) Galactic winds: Stellar feedback is implemented usinghydrodynamically decoupled winds which transport mass,momentum, metals, and thermal energy.(iii) Metal enrichment: Metals are returned to the ISM bysupernovae (SN) type Ia, SN type II, and asymptotic giantbranch (AGB) stars.(iv) Gas cooling and heating: Gas can cool through pri-mordial channels as well as metal-line cooling. Gas heating iscalculated assuming a superposition of a redshift-dependent,spatially uniform UV background (Faucher-Gigu`ere et al.2009), and the active galactic nucleus (AGN) radiation field(Vogelsberger et al. 2013).(v) Growth and feedback of black holes (BHs): Accretiononto BH sink particles is described by Bondi-Hoyle accre-tion. The associated feedback from AGN employs kineticfeedback (low accretion rates), and thermal feedback (highaccretion rates).For a detailed description of the simulations, and physi-cal model, we refer the reader to the IllustrisTNG introduc-tion and methods papers: Weinberger et al. (2017); Mari-nacci et al. (2018); Naiman et al. (2018); Nelson et al. (2018);Pillepich et al. (2018a,b); Springel et al. (2018), and refer-ences therein. We identify galaxy-galaxy mergers in IllustrisTNG fromthe publicly available merger trees created using
Sublink (Rodriguez-Gomez et al. 2015). The
Sublink merger treeslink sub-haloes to their progenitors and descendants. Conse-quently, galaxy mergers are defined as nodes within a tree;viz. a merger occurs when a sub-halo has two distinct di-rect progenitors. Note that we use “post-merger” to refer togalaxies immediately after the merger (i.e., first snapshot af-ter the merger) which corresponds to ≤ Myr post-mergerat the snapshot time resolution of IllustrisTNG. In Section3.2, we follow the evolution of post-mergers and hence, inthat section only, “post-merger” refers to galaxies which haveundergone a merger in their recent past. Therefore, post-mergers (merger remnants) are parametrized by: • z : The redshift of the post-merger. We include in ouranalysis post-mergers at z ≤ which allows us to studythe redshift evolution (or lack thereof) of mergers and theirdescendants. • M (cid:63) : The stellar mass of the post-merger. Various stud-ies have noted the spurious assignment of mass caused bysub-haloes in close proximity when using halo finders (i.e.,numerical stripping). For example, the dark matter mass ofsatellites has been shown to correlate with the distance totheir host galaxies (Sales et al. 2007; Wetzel et al. 2009).Numerical stripping has also been demonstrated to affectstellar mass estimates (i.e., Rodriguez-Gomez et al. 2015).We circumvent the subtleties of stellar mass calculationsby adopting the maximum stellar mass over the past . Gyr ( M max (cid:63) ) for all mass ratio calculations following Pat-ton et al. (2020). This approach is similar to Rodriguez-Gomez et al. (2015), with an additional restriction on look-back time; i.e., we minimise the effects of numerical strippingyet we still account for physical stripping by limiting thelookback time. Driven by the simulation resolution, galaxieswith stellar masses log ( M (cid:63) / M (cid:12) ) ≥ are reliably resolved.Therefore, we limit our analysis to post-mergers with stel-lar mass ≤ M (cid:63) / M (cid:12) ≤ and progenitors with stellarmass M (cid:63) ≥ M (cid:12) (see the mass ratio description). • µ : The stellar mass ratio of the progenitors. Note thatwe define the mass ratio to be < µ ≤ . In cases wheremore than two direct progenitors are found, we parametrizethe merger remnant using only the most major merger.The masses used to calculate µ correspond to the stellarmass within twice the stellar half-mass radius. Our analy-sis includes descendants of mergers with µ ≥ . which is arequisite for investigating the varying effects of mass ratioin galaxy mergers. The simulation’s mass resolution limit,coupled with the mass ratio range determines the stellarmass range of the post-merger sample and their progenitors: ≤ M (cid:63) / M (cid:12) ≤ .We are primarily interested in studying the effects ofgalaxy mergers on star formation during the post-mergerstage, therefore we exclude from our sample post-mergersthat are currently undergoing new close interactions buthave not yet fully merged. Particularly, we ignore post-mergers that are overlapping with another galaxy which hin-ders our ability to separate the effects of the current inter-action from those of the merger. Following the methodology MNRAS , 1–18 (2020)
M. H. Hani et al. M ? [M (cid:12) ])0.00.30.60.91.2 pd f µ )0.01.02.03.04.05.0 pd f TNG100TNG300 z )0.00.30.60.91.21.5 pd f Figure 1.
The stellar mass ( M (cid:63) ), mass ratio ( µ ), and redshift ( z )distributions of our post-merger samples selected from TNG100(salmon histogram) and TNG300 (purple histogram). The samplespans a wide range in mass ratio ( µ ≥ . ) and redshift ( z ≤ )which is key when studying the effects of galaxy mergers duringthe post-merger stage. In total, we selected , post-mergersfrom TNG100 and , from TNG300. of Patton et al. (2020), we define: r sep = rR host1 / + R comp1 / (1)where r is the separation of the host (post-merger) from itsnearest neighbour (companion), R host1 / is the stellar half-massradius of the host (post-merger), and R comp1 / is the stellarhalf-mass radius of the nearest neighbour. We exclude post-mergers with r sep ≤ from our sample. In addition, we ignoregalaxies with unresolved SFRs, viz. SFR < − M (cid:12) yr − and SFR < − M (cid:12) yr − in TNG100-1 and TNG300-1,respectively (see Donnari et al. 2019).Our post-merger sample consists of , post-mergersin TNG100-1, and , post-mergers in TNG300-1 withstellar masses ≤ M (cid:63) / M (cid:12) ≤ , z ≤ , and µ ≥ . .The properties of the post-merger sample are summarisedin Figure 1 which shows the distributions of M (cid:63) , µ , and z for the post-merger sample in TNG300-1 and TNG100-1. µ = 0 .
10 kpc µ = 0 .
10 kpc µ = 0 .
10 kpc
10 kpc
10 kpc
10 kpc
Figure 2.
An example of three z = post-mergers (top panels)with different mass ratios ( µ ) along with their respective controlgalaxies (bottom panels) selected from TNG300-1. The panelsshow synthetic stellar composite images using the JWST_f200w,JWST_f115w and
JWST_f070w photometric filters. The
SubfindID is indicated in the lower left corner of each panel, and scale bar(top left) indicates a kpc physical scale. The merger remnantsexhibit evident low surface brightness features and disturbed mor-phologies. The post-mergers in our sample are uniformly distributedacross redshift ≤ z ≤ . While the merger sample in Illus-trisTNG spans a large range of mass ratios, minor mergers( µ < . ) dominate the population of mergers. We note thatthe dearth of post-mergers at M (cid:63) (cid:38) M (cid:12) is driven bythe stellar mass limits applied in the post-merger selection(i.e., progenitor stellar mass M (cid:63) ≥ M (cid:12) ). Figure 2 de-picts selected examples of post-mergers from our sample.The post-mergers exhibit disturbed morphologies with evi-dent low surface brightness features (e.g., shells, tidal tails). We quantify the effects of galaxy mergers by adopting acommonly used control matching approach in observationalstudies of galaxy interactions/mergers (e.g., Ellison et al.2013; Patton et al. 2013). For each post-merger, we iden-tify a control galaxy to which we compare the properties ofthe post-merger. In this section, we describe the process ofcreating the control sample.Large-box cosmological simulations provide a large anddiverse population of galaxies thus allowing us to statisti-cally study the changes in post-mergers compared to similargalaxies which have not undergone a merger. Comparingmerger descendants to galaxies which have not undergonea recent merger allows us to isolate the effects of a mergerwhilst removing biases which may arise from known galaxycorrelations (e.g., environment, stellar mass, redshift).Following Patton et al. (2016) and Patton et al. (2020),we assign a control galaxy to each post-merger in our sam-ple. However, whereas Patton et al. (2020) studied thepair interaction phase we are interested in studying post-mergers. Therefore, we apply some modifications to our con-trol matching process. We first create a galaxy pool whichincludes all galaxies at z ≤ with resolved SFRs, M (cid:63) ≥ M (cid:12) , and r sep > . We exclude galaxies which have undergone MNRAS000
SubfindID is indicated in the lower left corner of each panel, and scale bar(top left) indicates a kpc physical scale. The merger remnantsexhibit evident low surface brightness features and disturbed mor-phologies. The post-mergers in our sample are uniformly distributedacross redshift ≤ z ≤ . While the merger sample in Illus-trisTNG spans a large range of mass ratios, minor mergers( µ < . ) dominate the population of mergers. We note thatthe dearth of post-mergers at M (cid:63) (cid:38) M (cid:12) is driven bythe stellar mass limits applied in the post-merger selection(i.e., progenitor stellar mass M (cid:63) ≥ M (cid:12) ). Figure 2 de-picts selected examples of post-mergers from our sample.The post-mergers exhibit disturbed morphologies with evi-dent low surface brightness features (e.g., shells, tidal tails). We quantify the effects of galaxy mergers by adopting acommonly used control matching approach in observationalstudies of galaxy interactions/mergers (e.g., Ellison et al.2013; Patton et al. 2013). For each post-merger, we iden-tify a control galaxy to which we compare the properties ofthe post-merger. In this section, we describe the process ofcreating the control sample.Large-box cosmological simulations provide a large anddiverse population of galaxies thus allowing us to statisti-cally study the changes in post-mergers compared to similargalaxies which have not undergone a merger. Comparingmerger descendants to galaxies which have not undergonea recent merger allows us to isolate the effects of a mergerwhilst removing biases which may arise from known galaxycorrelations (e.g., environment, stellar mass, redshift).Following Patton et al. (2016) and Patton et al. (2020),we assign a control galaxy to each post-merger in our sam-ple. However, whereas Patton et al. (2020) studied thepair interaction phase we are interested in studying post-mergers. Therefore, we apply some modifications to our con-trol matching process. We first create a galaxy pool whichincludes all galaxies at z ≤ with resolved SFRs, M (cid:63) ≥ M (cid:12) , and r sep > . We exclude galaxies which have undergone MNRAS000 , 1–18 (2020) llustrisTNG post-merger galaxies M ? [M (cid:12) ])0.00.30.60.91.2 pd f N + 1)0.00.30.60.91.2 pd f PMControl ∞ log( r [kpc])0.00.20.40.60.8 pd f Figure 3.
A comparison between the post-merger sample andtheir respective controls in TNG300-1. The grey shaded bin inthe r distribution includes all galaxies with r > Mpc. Thefigure only shows the quantities which are allowed to vary duringthe control matching (i.e., M (cid:63) , N , and r ). The control matchingis equally excellent for the TNG100 sample (not shown). a merger ( µ ≥ . ) within the past Gyr thus generating aninitial pool which includes , , galaxies in TNG300-1( , galaxies in TNG100-1). We then quantify the envi-ronment for each galaxy in our pool by calculating the num-ber of neighbours within a Mpc radius (hereafter N ) andthe distance to the nearest neighbour with M (cid:63) ≥ . × M (cid:63), host (hereafter r ). We also classify control galaxies and post-mergers as star-forming (SF) or passive. The classificationis performed by applying a linear fit to the star-forming mainsequence (SFMS) at stellar masses between − . M (cid:12) which is extrapolated to higher M (cid:63) following the methodol-ogy of Donnari et al. (2019). Galaxies that lie below σ MS from the SFMS are classified as passive . Finally, for eachpost-merger we select the single best statistically matchedcontrol galaxy as follows: We first reduce the control pool σ MS is the standard deviation of the SFMS residuals between − . M (cid:12) ; for TNG300-1 σ MS = . dex to those galaxies that are in the same snapshot and havethe same class (i.e., SF or passive) as the post-merger (i.e.,matching in redshift and class). We stress the importance ofthe SF/passive class match: ignoring the class in the match-ing procedure would yield an unfair comparison where SFgalaxies may have passive controls (and vice versa). Then,we search the snapshot-culled control pool for galaxies with M (cid:63) within a tolerance of . dex, and N and r within 10%of each post-merger. In most cases the initial tolerances yieldat least one match. However, in some cases where no matchesare found within the initial tolerances, we grow the tolerancerange by . dex and 10% and repeat the search until atleast one match is found. If more than one match is found,we select the best match in all three parameters M (cid:63) , N , and r following the weighting scheme of Patton et al. (2016). Weexclude from our analysis post-mergers which require morethan 3 grows to identify a control; four grows corresponds tounacceptably large matching tolerances (i.e., ∆ log M (cid:63) = . ,40% in N , and 40% in r ). Figure 3 compares the distribu-tions of post-mergers and their controls (both selected fromTNG300-1) in the three matching parameters: M (cid:63) , r , and N . The post-mergers are excellently matched by the controlpopulation. The TNG100-1 post-merger sample is matchedto a similar quality as that of TNG300-1.Once the controls have been identified, we statisticallyquantify the enhancement (or suppression) in star formationby comparing post-mergers to their controls as an ensembleusing the following metric: Q ( sSFR ) ≡ (cid:104) sSFR pm (cid:105)(cid:104) sSFR control (cid:105) (2)where (cid:104) sSFR pm (cid:105) is the arithmetic mean of the post-mergers’specific star formation rates (sSFRs), and (cid:104) sSFR control (cid:105) is thearithmetic mean of the controls’ sSFRs. The sSFR is calcu-lated using the SFR within twice the stellar half mass ra-dius and the associated stellar mass. The metric used in thiswork has been extensively used in the literature to demon-strate the effects of galaxy interactions on star formation(e.g., Patton et al. 2013, 2020). We first investigate the effect of mergers on the SFR of theselected post-merger galaxy sample. Figure 4 shows the dis-tribution of post-mergers in the SFR − M (cid:63) plane in TNG300-1, our fiducial simulation. While most of the post-mergerspopulate the star-forming galaxy main sequence ( of thepost-merger sample in TNG300-1), our post-merger samplespans a wide range in stellar mass and SFR including pas-sive galaxies ( of the post-merger sample in TNG300-1). Therefore, the post-merger sample selected from Illus-trisTNG is well-suited for a systematic and statistical studyof the SFRs in post-merger systems.To fairly investigate the effects of galaxy mergers onthe post-mergers’ SFR we compare the SFR of post-merger MNRAS , 1–18 (2020)
M. H. Hani et al. M ? [M (cid:12) ])-3.0-2.0-1.00.01.02.03.0 l og ( S F R [ M (cid:12) / y r ] ) Figure 4.
The distribution of post-mergers in TNG300-1 inSFR − M (cid:63) space. The post-mergers sample we present spans alarge range in SFR and stellar mass including star-forming galax-ies as well as passive galaxies. galaxies to their control counterparts. We remind the readerthat the control galaxies are chosen to have not undergone amerger ( µ ≥ . ) within the past Gyr while being matchedto the post-merger sample in N , r , redshift, and M (cid:63) . Table1 summarises the comparison of the average sSFRs of post-merger and their respective controls, Q ( sSFR ) (defined inequation 2). In the full sample, post-mergers statisticallyhave elevated SFRs with a mean enhancement of Q ( sSFR ) = . ± . .The SFR enhancement (or lack thereof) is more pro-nounced for the star-forming and passive post-merger subsamples, respectively. The SFRs are enhanced in the star-forming sample with Q ( sSFR ) = . ± . . The enhance-ment reported in our SF post-merger sample is consistentwith those reported in observational studies. For exampleEllison et al. (2013) measured a factor of 2.5 increase in theSFR of (star forming) post mergers in the SDSS, very sim-ilar to the factor of two found here. Conversely, the passivepost-merger sample is statistically consistent with no SFRenhancement ( Q ( sSFR ) = . ± . ).In summary, the post-merger sample in IllustrisTNG ex-hibits a statistical enhancement in SFR consistent with ob-servations. The enhancements are more pronounced in star-forming post-mergers. In the following sections, we will dis-sect the dependence of the SFR enhancement/suppressionon galaxy properties (e.g., redshift, stellar mass) and there-fore gain insight on the physical mechanisms driving saidenhancement/suppression. From the perspective of galaxy mergers, the Universe wasmuch sprier at earlier times: several theoretical and obser-vational studies report the decline of the galaxy merger ratewith decreasing redshift (e.g., Lin et al. 2008; de Ravel et al.2009; Lotz et al. 2011; L´opez-Sanjuan et al. 2013; Rodriguez-Gomez et al. 2015). Whilst mergers can efficiently drive SFRenhancements at low redshifts, their contribution to the Uni-verse’s star formation rate density is thought to decrease Q ( sSF R ) σ Q ( sSFR ) / √ N all .
057 0 . star-forming .
073 0 . passive .
983 0 . Table 1.
A comparison of the sSFRs of post-mergers and theircontrols. The table reports Q ( sSF R ) and the associated stan-dard error on the mean for the full post-merger sample, and sub-samples therein. The post-merger sample generally has elevated Q ( sSF R ) values. The signal is dominated by the star-formingpost-mergers while the passive post-mergers exhibit sSFRs whichare, on average, consistent with their controls. with increasing redshifts (e.g., Rodighiero et al. 2011; Kavi-raj et al. 2013; Madau & Dickinson 2014; Lofthouse et al.2017; Wilson et al. 2019). The scenario in which the contri-bution of galaxy mergers to star formation decreases withincreasing redshift is challenged by other studies which re-port that galaxy interactions continue to induce star forma-tion at high redshift (e.g., Lin et al. 2007; Wong et al. 2011).Therefore, to properly quantify the contribution of mergersto the global star formation rate density, one must assess theeffect of mergers on SFR enhancements across cosmic time.For example, Patton et al. (2020) examined the redshift de-pendence of SFR enhancement in galaxy pairs and reportedsmaller mean SFR enhancement with increasing redshift.Patton et al. (2020) demonstrate that the decrease in themean SFR enhancements at higher redshift is offset by theincreased fraction of galaxies undergoing merger-triggeredSFR enhancements leading to higher net SFR enhancements( at z < . to at . < z < ) in their sample.Figure 5 shows the evolution in the mean sSFR acrossredshift for post-mergers and their controls (top panel), andthe associated sSFR enhancement in post-mergers (mid-dle and bottom panels). The mean sSFRs of both post-mergers and controls decline with decreasing redshift, awell-understood consequence of the decreasing accretion rateonto haloes (e.g., Dekel et al. 2009) and therefore the increas-ing quenched fraction (e.g., Donnari et al. 2019). Despitethe evolution of the average sSFR across cosmic time, post-mergers exhibit a steady sSFR enhancement, Q ( sSFR ) ∼ ,at all redshifts ≤ z ≤ . By separating the sample by theirSFRs, the lower panel of Figure 5 shows that star formingpost-mergers drive the aforementioned enhancement, whilepassive post-mergers have sSFRs which are statistically con-sistent with their respective controls. The distinction be-tween the star-forming and passive post-merger sampleshints at galaxy mergers enhancing the pre-existing condi-tions for star formation rather than triggering new processes;possibly emphasising the role of the galactic gas content (see § z = in our sample. This isconsistent with results by Rodr´ıguez Montero et al. (2019),who analysed galaxy mergers with µ ≥ . in the simba simulation (Dav´e et al. 2019) and found an enhancement ofa factor of ∼ − for z ≤ . However, the absence of evolu-tion in Q ( sSFR ) with redshift in our sample is in contrast torecent work by Patton et al. (2020), who studied SFR en-hancement during the pair phase using the same simulation MNRAS000
A comparison of the sSFRs of post-mergers and theircontrols. The table reports Q ( sSF R ) and the associated stan-dard error on the mean for the full post-merger sample, and sub-samples therein. The post-merger sample generally has elevated Q ( sSF R ) values. The signal is dominated by the star-formingpost-mergers while the passive post-mergers exhibit sSFRs whichare, on average, consistent with their controls. with increasing redshifts (e.g., Rodighiero et al. 2011; Kavi-raj et al. 2013; Madau & Dickinson 2014; Lofthouse et al.2017; Wilson et al. 2019). The scenario in which the contri-bution of galaxy mergers to star formation decreases withincreasing redshift is challenged by other studies which re-port that galaxy interactions continue to induce star forma-tion at high redshift (e.g., Lin et al. 2007; Wong et al. 2011).Therefore, to properly quantify the contribution of mergersto the global star formation rate density, one must assess theeffect of mergers on SFR enhancements across cosmic time.For example, Patton et al. (2020) examined the redshift de-pendence of SFR enhancement in galaxy pairs and reportedsmaller mean SFR enhancement with increasing redshift.Patton et al. (2020) demonstrate that the decrease in themean SFR enhancements at higher redshift is offset by theincreased fraction of galaxies undergoing merger-triggeredSFR enhancements leading to higher net SFR enhancements( at z < . to at . < z < ) in their sample.Figure 5 shows the evolution in the mean sSFR acrossredshift for post-mergers and their controls (top panel), andthe associated sSFR enhancement in post-mergers (mid-dle and bottom panels). The mean sSFRs of both post-mergers and controls decline with decreasing redshift, awell-understood consequence of the decreasing accretion rateonto haloes (e.g., Dekel et al. 2009) and therefore the increas-ing quenched fraction (e.g., Donnari et al. 2019). Despitethe evolution of the average sSFR across cosmic time, post-mergers exhibit a steady sSFR enhancement, Q ( sSFR ) ∼ ,at all redshifts ≤ z ≤ . By separating the sample by theirSFRs, the lower panel of Figure 5 shows that star formingpost-mergers drive the aforementioned enhancement, whilepassive post-mergers have sSFRs which are statistically con-sistent with their respective controls. The distinction be-tween the star-forming and passive post-merger sampleshints at galaxy mergers enhancing the pre-existing condi-tions for star formation rather than triggering new processes;possibly emphasising the role of the galactic gas content (see § z = in our sample. This isconsistent with results by Rodr´ıguez Montero et al. (2019),who analysed galaxy mergers with µ ≥ . in the simba simulation (Dav´e et al. 2019) and found an enhancement ofa factor of ∼ − for z ≤ . However, the absence of evolu-tion in Q ( sSFR ) with redshift in our sample is in contrast torecent work by Patton et al. (2020), who studied SFR en-hancement during the pair phase using the same simulation MNRAS000 , 1–18 (2020) llustrisTNG post-merger galaxies -10.5-10.0-9.5-9.0 h s S F R i ControlPM Q ( s S F R ) z )0.01.02.03.0 Q ( s S F R ) Star-formingPassive
Figure 5.
The redshift-evolution of the sSFR enhancement inpost-mergers. The top panel shows the running average sSFR forpost-mergers (dark-purple) and their associated controls (orange)with the shaded region representing twice the standard error onthe mean in bins or redshift. The middle panel shows the depen-dence of Q ( sSF R ) (and the associated uncertainty; see equation2) on redshift for the full post-merger sample while the bottompanel shows Q ( sSF R ) for star-forming (blue) and passive (red)post-mergers. Post-mergers exhibit enhanced sSFRs at all red-shifts ( ≤ z ≤ ). The enhancement is dominated by the star-forming post-mergers ( of the sample) while the sSFRs of pas-sive post-mergers ( of the sample) are consistent with thoseof the controls. suite and with similar methods as those used herein. Pattonet al. (2020) report a mild evolution in Q ( sSFR ) with red-shift: a mean Q ( sSFR ) ∼ . at . < z < increases, with de-creasing redshift, to Q ( sSFR ) ∼ . at z < . . The differentredshift evolution results between our study of post-mergersand Patton et al.’s pairs work may be due to differences inthe galaxy properties of the two samples. Namely, the massdistribution of the post-merger sample presented here lackspost-mergers at M (cid:63) ∼ M (cid:12) while the pairs sample ofPatton et al. (2020) includes galaxies to lower masses. Al-ternatively, the difference in our control matching strategyand that of Patton et al. (2020) may also contribute theaforementioned difference in redshift evolution.In order to more fairly compare our work to that ofPatton et al. (2020) we repeated our analysis using theexact matching strategy used in Patton et al. (2020). Wefind that the sSFR enhancement in the reconstructed post-merger sample evolves with redshift ( Q ( sSFR ) ∼ at z = to Q ( sSFR ) = . at z = ). The redshift-evolution of the merger-induced SFR enhancement in the reconstructed star-forming post-mergers is even stronger, while the passivepost-mergers now show suppressed SFRs (i.e., matched tostar-forming control galaxies). Additionally, we tested theeffects of the sample’s mass distribution on the redshift evo-lution of Q ( sSFR ) . We repeated the analysis of Patton et al.(2020) using our control matching strategy which resultedin no redshift evolution in Q ( sSFR ) . Therefore, the redshiftdependence of Q ( sSFR ) is driven by the control matchingstrategy: i.e., star-forming galaxies being matched to thepassive galaxies which are more abundant at lower redshift. Hydrodynamical simulations provide a solid understandingof the importance of the merger mass ratio on the mea-sured enhancements and timescale on which the effects ofa merger are visible (e.g., Bournaud et al. 2005; Cox et al.2008; Johansson et al. 2009; Lotz et al. 2010a). These previ-ous works have found that minor mergers (small µ ) induceweaker starbursts, and short-lived observable asymmetries.On the contrary, major, gas-rich mergers induce the mostprominent starbursts and the most long-lived asymmetri-cal features. However, previous works have mostly focusedon mergers in idealised settings (i.e., binary disc mergers)where one can perform well targeted experiments to isolatethe effects of galaxy properties (e.g., mass ratio, gas fraction)on the merger outcome (e.g., starburst strength, asymmetrymetrics).In this sub-section, we investigate the relevance of themerger mass ratio ( µ ) on the induced sSFR enhancementin a large-box cosmological simulation. Figure 6 shows themean sSFR at different mass ratios for post-mergers andtheir associated controls (top panel), and the dependenceof the induced sSFR enhancement on the merger mass ra-tio (middle and bottom panels). Mergers with all mass ra-tios drive an enhancement in sSFR in our post-merger sam-ple. The enhancement is more pronounced for major merg-ers ( µ ≥ . ) with Q ( sSFR ) ∼ . , while minor mergers µ ∼ . show more modest (yet still significant) enhance-ments with Q ( sSFR ) ∼ . Our results are consistent withother works exploring mergers in cosmological numericalsimulations (i.e., Rodr´ıguez Montero et al. 2019). Passivepost-mergers exhibit sSFRs which are consistent with theircontrols while star-forming post-mergers dominate the en-hancements shown in the full post-merger sample.The results of the work presented here are qualitativelyconsistent with previous simulations of idealised mergers(e.g., Cox et al. 2008; Johansson et al. 2009). Major mergersinduce aggressive gravitational torques which drive dynam-ical instabilities in the ISM thus enhancing star formation.On the contrary, minor mergers induce more modest gravi-tational torques thus driving smaller enhancements. We notethat the impact of minor mergers may be overestimated forgalaxies which undergo multiple mergers; viz. if a galaxy un-dergoes a merger before the effects of a previous merger havedecayed we would overestimate the boost in star formation.However, such a scenario is rare and the associated effectsshould be small.Comparisons to observational studies of post-mergergalaxies are particularly complicated owing to the difficultyin assessing the mass ratio of a galaxy merger resulting in MNRAS , 1–18 (2020)
M. H. Hani et al. -10.5-10.0-9.5-9.0 h s S F R i ControlPM Q ( s S F R ) µ )0.01.02.03.0 Q ( s S F R ) Star-formingPassive
Figure 6.
The dependence of the enhancement in the post-mergers’ sSFR on the parents’ merger mass ratio ( µ ). The shadedregion represents twice the standard error on the mean in bins of µ . Post-mergers exhibit enhanced sSFR for all mass ratios withan increasing enhancement for major mergers. The enhancementis dominated by the star-forming post-mergers, while the passivepost-mergers sSFR are consistent with their controls. an observed post-merger galaxy. Nonetheless, one can qual-itatively compare this work’s results to the enhancementsdriven by galaxy mergers of varying mass ratios duringthe pair stage. For example, Scudder et al. (2012) studiedgalaxy pairs selected from SDSS and concluded that modestenhancements in SFR ( log ( SFR pm ) − log ( SFR control ) ≤ . dex) can be achieved over a wide range of mass ratios( . ≤ µ ≤ ), whereas the strongest SFR enhancements( log ( SFR pm ) − log ( SFR control ) ≥ . dex) are preferentiallydriven by major mergers ( µ ≥ . ). Figure 7 shows the frac-tion of star-forming descendants of major mergers ( µ ≥ . )with enhancements greater than ∆ SFR relative to that of mi-nor mergers ( . ≤ µ < . ). We define the merger-inducedstar formation, ∆ SFR , for a star-forming post-merger as thevertical offset in SFR from the SFMS: i.e., ∆ SFR = SFR pm − SFR
SFMS . (3)Post-mergers below the SFMS contribute a negative ∆ SFR .Figure 7 shows that strong enhancements of SFR are domi-nated by major mergers. For example, mergers that triggeran SFR that is M (cid:12) /yr above the SFMS are approximatelyfour times as likely to be major, rather than minor, merg-ers. Our results are therefore qualitatively consistent withthe conclusions of Scudder et al. (2012), namely that SFRenhancements are triggered by a wide range of mass ratios -50 -25 0 25 50∆ SFR [M (cid:12) /yr]01234 cc d f m a j o r / cc d f m i n o r Figure 7.
The fraction of star-forming descendants of majormergers ( µ ≥ . ) with SFR enhancement greater than ∆ SFR (complementary cumulative distribution function; ccdf) relativeto that of minor mergers ( . ≤ µ < . ). The definition of ∆ SFR is given in equation 3. Major mergers dominate the high tail ofthe ∆ SFR distribution. (Figure 6), but that the largest starbursts are preferentiallyproduced by major mergers (Figure 7).Although major mergers trigger the strongest star-bursts, minor mergers are far more common (e.g. Figure1). In Figure 8, we therefore quantify the fractional contri-bution to the total merger-induced star formation budget as a function of mass ratio in the star-forming post-mergersample. The contribution to the total merger-induced SFRenhancement budget increases with decreasing mass ratio.Despite their lower average enhancement, the cumulativeenhancement from minor mergers ( . ≤ µ < . ) accountsfor ∼ of the merger-induced SFR budget. Hence, mi-nor mergers are a significant contributor to the merger-induced SFR budget. These results are qualitatively consis-tent with earlier idealised merger simulations (e.g., Scudderet al. 2015). Since the contribution to the merger star forma-tion budget continues to increase towards the smallest massratios in our samples (Figure 8), it would be of great inter-est to extend this study to even lower mass ratios. However,fully investigating the merger-induced SFR budget requiresa much higher mass resolution. Such a study will be possi-ble with the forthcoming IllustrisTNG50 simulation (Nelsonet al. 2019b; Pillepich et al. 2019). In Section 3.1.1 we demonstrated that the enhancement insSFR is more pronounced in the star-forming population ofthe post-merger sample while the passive post-mergers showstatistically consistent sSFRs with their controls. The dis-crepancy in sSFR enhancement between star-forming andpassive post-mergers holds at all redshifts (Figure 5) andmass ratios (Figure 6). Knowing that there is a correlationbetween the fraction of passive galaxies and stellar mass Note that we compute the total merger-induced SFR budgetand not the total cosmic star formation budget.MNRAS000
The fraction of star-forming descendants of majormergers ( µ ≥ . ) with SFR enhancement greater than ∆ SFR (complementary cumulative distribution function; ccdf) relativeto that of minor mergers ( . ≤ µ < . ). The definition of ∆ SFR is given in equation 3. Major mergers dominate the high tail ofthe ∆ SFR distribution. (Figure 6), but that the largest starbursts are preferentiallyproduced by major mergers (Figure 7).Although major mergers trigger the strongest star-bursts, minor mergers are far more common (e.g. Figure1). In Figure 8, we therefore quantify the fractional contri-bution to the total merger-induced star formation budget as a function of mass ratio in the star-forming post-mergersample. The contribution to the total merger-induced SFRenhancement budget increases with decreasing mass ratio.Despite their lower average enhancement, the cumulativeenhancement from minor mergers ( . ≤ µ < . ) accountsfor ∼ of the merger-induced SFR budget. Hence, mi-nor mergers are a significant contributor to the merger-induced SFR budget. These results are qualitatively consis-tent with earlier idealised merger simulations (e.g., Scudderet al. 2015). Since the contribution to the merger star forma-tion budget continues to increase towards the smallest massratios in our samples (Figure 8), it would be of great inter-est to extend this study to even lower mass ratios. However,fully investigating the merger-induced SFR budget requiresa much higher mass resolution. Such a study will be possi-ble with the forthcoming IllustrisTNG50 simulation (Nelsonet al. 2019b; Pillepich et al. 2019). In Section 3.1.1 we demonstrated that the enhancement insSFR is more pronounced in the star-forming population ofthe post-merger sample while the passive post-mergers showstatistically consistent sSFRs with their controls. The dis-crepancy in sSFR enhancement between star-forming andpassive post-mergers holds at all redshifts (Figure 5) andmass ratios (Figure 6). Knowing that there is a correlationbetween the fraction of passive galaxies and stellar mass Note that we compute the total merger-induced SFR budgetand not the total cosmic star formation budget.MNRAS000 , 1–18 (2020) llustrisTNG post-merger galaxies µ )0.00.10.20.30.4 f r a c t i o n a l S F R b oo s t Figure 8.
An accounting of the contribution to the total merger-driven SFR enhancement budget in the star-forming post-mergersample. The merger-induced star formation boost for a given post-merger is defined as the vertical offset from the SFMS (see equa-tion 3). The contribution to the total merger-induced SFR en-hancement increases with decreasing mass ratio thus indicatingthat minor mergers ( . ≤ µ < . ) are significant contributors tothe global merger-induced SFR budget. Although major mergers( µ ≥ . ) induce the strongest bursts, the abundance of minormergers compensates for their modest SFR enhancement whencompared to major mergers. (see Figure 4), we investigate the dependence of the sSFRenhancement in post-mergers on stellar mass. We are inter-ested in disentangling the source of the discrepancy in en-hancement between star-forming and passive post-mergers.Figure 9 shows the mean sSFR of post-mergers and theircontrols (top panel), and the associated enhancement (mid-dle and bottom panels) as a function of post-merger stel-lar mass. The average sSFR of post-mergers and controlsdeclines with stellar mass reconfirming that passive galax-ies dominate the high-mass galaxy population (e.g., Bundyet al. 2010; Bluck et al. 2014). Unlike redshift and massratio, the stellar mass has a significant impact on the mea-sured enhancement in sSFR. As the stellar mass increases,the sSFR enhancement diminishes to be consistent withthe control sample and vanishes for log ( M (cid:63) / M (cid:12) ) > . .Post-mergers with stellar masses log ( M (cid:63) / M (cid:12) ) ≤ . ex-hibit an average enhancement in sSFR with Q ( sSFR ) ∼ .On the contrary, the high stellar mass post-mergers, i.e. log ( M (cid:63) / M (cid:12) ) > . are characterised by normal sSFRs(compared to their controls). The decline in Q ( sSFR ) at log ( M (cid:63) / M (cid:12) ) < . is caused by the minimum stellar masslimit we impose ( log ( M (cid:63) / M (cid:12) ) ≥ ) in order to reliably quan-tify the galaxies’ recent merger histories: i.e., post-mergerswith log ( M (cid:63) / M (cid:12) ) ∼ will be paired with more massive con-trols which, on average, have higher SFRs thus leading to adecline in Q ( sSFR ) at log ( M (cid:63) / M (cid:12) ) ∼ . The reported corre-lation of sSFR enhancement with stellar mass is consistentwith the results of Rodr´ıguez Montero et al. (2019) study-ing mergers in simba , another cosmological hydrodynamicalsimulation (with a different physical model).Dissecting the post-merger sample further is key to un-derstanding the source of the dependence of sSFR enhance-ment on stellar mass. The lower panel of Figure 9 shows -12.0-11.0-10.0-9.0 h s S F R i ControlPM Q ( s S F R ) M ? [M (cid:12) ])0.01.02.03.0 Q ( s S F R ) Star-formingPassive
Figure 9.
The dependence of the enhancement in the post-merger’s sSFR on post-merger stellar mass. The shaded re-gions represent twice the standard error on the mean in bins of log ( M (cid:63) / M (cid:12) ) . Star formation is most enhanced in post-mergerswith . ≤ log ( M (cid:63) / M (cid:12) ) ≤ . with a declining enhancement forlarger M (cid:63) . For log ( M (cid:63) / M (cid:12) ) > . post-mergers form stars at anaverage rate which is consistent with that of the control galax-ies. The star-forming post-mergers show the same behaviour asthe full post-merger sample while passive post-mergers are consis-tent with no enhancement for all post-merger M (cid:63) with a possibleslight enhancement at the largest M (cid:63) . the enhancement at different stellar masses for the star-forming and passive post-merger sub-samples. While thestar-forming post mergers follow the behaviour describedabove, the passive post-mergers exhibit typical sSFRs whencompared to their controls at all stellar masses. Therefore,the stellar mass is not the fundamental driver of the dif-ferences between the star-forming and passive post-mergersamples. This is suggestive of a separate galaxy property,which correlates with stellar mass and sSFR, driving the re-ported disparity between the passive and star-forming post-merger sub-samples (see § The previous sections ( § § § Similar conclusions are supported by Figures 5, and 6: The red-shift and mass ratio are not fundamental drivers of the differencesbetween the star-forming and passive post-merger samples.MNRAS , 1–18 (2020) M. H. Hani et al. . . . . . . f gas pd f Figure 10.
The gas fraction distribution of the post-merger sam-ple in TNG300-1. The gas fraction is defined as the progenitors’gas mass normalised by the total baryonic mass within × R half ,(cid:63) (see Equation 4). The filled purple histogram represents the fullsample while the red and blue histograms represent the pas-sive and star-forming post-mergers, respectively. The star-formingpost-mergers are the descendants of gas-rich mergers when com-pared to their passive counterparts. Note that f gas accounts for allgas phases and therefore should not be compared to the molecu-lar and cold gas fractions reported by observational studies (e.g.,Combes et al. 2013; Ellison et al. 2018; Tacconi et al. 2018). may not be fundamental drivers of the discrepancy in thesSFR enhancement between passive and star-forming post-mergers. In fact, passive post-mergers show indistinguish-able sSFRs from their controls at all redshifts, mass ratios,and stellar masses, while the star-forming post-merger ex-hibit strong enhancements at all z , µ , and low M (cid:63) . It is pos-sible that the gas content, a fundamental contributor to starformation, is indeed driving the observed trends: gas contentcorrelates with stellar mass, and drives star formation.In this section, we explore the effect of the progenitor’sgas content on the measured post-merger sSFR. We define f gas ≡ (cid:213) prog M gas (cid:213) prog M gas + (cid:213) prog M (cid:63) (4)where (cid:205) prog M gas is the total gas mass (within × R half ,(cid:63) )of the progenitors and (cid:205) prog M star is the total stellar mass(within × R half ,(cid:63) ) of the progenitors. Therefore, f gas canbe thought of as the total gas fraction available to formstars during the merger. Note that f gas accounts for all gasphases and therefore should not be compared to the molec-ular and cold gas fractions reported by observational studies(e.g., Combes et al. 2013; Ellison et al. 2018; Tacconi et al.2018). Figure 10 shows the distribution of f gas for the post-mergers in our sample. The gas fraction distribution is bi-modal. While the post-merger sample includes descendantsof both gas-rich and gas-poor mergers, star-forming post-mergers are predominantly the descendants of gas-rich merg-ers. The gas fractions in Figure 10 are significantly higherthan those reported in other studies using the IllustrisTNGsimulations owing to differences in the definition of f gas . Forexample, Pillepich et al. (2019) report lower gas fractions -11.0-10.5-10.0-9.5-9.0 h s S F R i ControlPM Q ( s S F R ) f gas )0.01.02.03.0 Q ( s S F R ) Star-formingPassive
Figure 11.
The effect of the progenitors’ gas content on the ob-served enhancement in the post-merger phase. The shaded regionsrepresent twice the error on the mean in bins of f gas . Gas rich merg-ers (i.e., high f gas ) yield post-mergers with enhanced sSFRs. Alter-natively, mergers between gas-poor galaxies (i.e., low f gas ) developinto post-mergers with suppressed sSFR. The aforementioned cor-relation between Q ( sSF R ) and the parents’ gas content holds forboth star-forming and passive galaxies with the effects being es-pecially pronounced for the star-forming post-merger sample. for star-forming galaxies in the TNG50 box where they nor-malise the gas mass by the total dynamical mass within × R half ,(cid:63) instead of the total baryonic mass as shown inEquation 4. We elect to normalise the gas mass by the bary-onic mass following the commonly employed formalism inobservational studies.Figure 11 shows the average sSFR for post-mergers andcontrols, and the associated enhancement as a function of f gas . As expected, both the sSFRs of post-mergers and con-trols correlate with f gas (see the top panel of Figure 11). Ad-ditionally, the persistent enhancement reported in previoussections correlates strongly with f gas with evident suppres-sion in gas poor systems ( f gas < . ), and elevated enhance-ments up to Q ( sSFR ) ∼ at the highest f gas . The correlationof Q ( sSFR ) with f gas also applies to the passive post-mergersub-sample, albeit to a lesser extent. Qualitatively, Q ( sSFR ) for both passive and star-forming post-mergers show a con-sistent dependence on f gas : i.e., even passive post-mergersdescending from gas rich mergers exhibit enhanced sSFRswhen compared to their controls.We note that the level of enhancement (i.e., Q ( sSFR ) ) ofstar-forming and passive post-mergers, and the critical valueof f gas above which enhancements are measured remain in- MNRAS000
The effect of the progenitors’ gas content on the ob-served enhancement in the post-merger phase. The shaded regionsrepresent twice the error on the mean in bins of f gas . Gas rich merg-ers (i.e., high f gas ) yield post-mergers with enhanced sSFRs. Alter-natively, mergers between gas-poor galaxies (i.e., low f gas ) developinto post-mergers with suppressed sSFR. The aforementioned cor-relation between Q ( sSF R ) and the parents’ gas content holds forboth star-forming and passive galaxies with the effects being es-pecially pronounced for the star-forming post-merger sample. for star-forming galaxies in the TNG50 box where they nor-malise the gas mass by the total dynamical mass within × R half ,(cid:63) instead of the total baryonic mass as shown inEquation 4. We elect to normalise the gas mass by the bary-onic mass following the commonly employed formalism inobservational studies.Figure 11 shows the average sSFR for post-mergers andcontrols, and the associated enhancement as a function of f gas . As expected, both the sSFRs of post-mergers and con-trols correlate with f gas (see the top panel of Figure 11). Ad-ditionally, the persistent enhancement reported in previoussections correlates strongly with f gas with evident suppres-sion in gas poor systems ( f gas < . ), and elevated enhance-ments up to Q ( sSFR ) ∼ at the highest f gas . The correlationof Q ( sSFR ) with f gas also applies to the passive post-mergersub-sample, albeit to a lesser extent. Qualitatively, Q ( sSFR ) for both passive and star-forming post-mergers show a con-sistent dependence on f gas : i.e., even passive post-mergersdescending from gas rich mergers exhibit enhanced sSFRswhen compared to their controls.We note that the level of enhancement (i.e., Q ( sSFR ) ) ofstar-forming and passive post-mergers, and the critical valueof f gas above which enhancements are measured remain in- MNRAS000 , 1–18 (2020) llustrisTNG post-merger galaxies consistent between the star-forming and passive post-mergersample. Even at fixed f gas , the passive post-mergers havelower Q ( sSFR ) (for most values of f gas ), thus suggesting thatthe gas fraction is not the sole driver, albeit a strong driver,of SFR enhancement in galaxy mergers. Such a discrep-ancy could be engendered by various, possibly competing,effects such as: (1) the efficiency of feedback at differentstellar masses may require higher f gas for high M (cid:63) galaxiesin order to achieve an enhancement; and (2) the metric f gas does not differentiate between hot, cold, and molecular gaswhich may affect the quantitative results. In fact, the Il-lustrisTNG physical model treats star-forming gas using aneffective equation of state. The sub-grid multi-phase pres-surised ISM treatment does not follow the detailed proper-ties and physical state of the gas (i.e., atomic and molecularfractions) which poses a limitation to further analysis of thegas content of post-mergers in this work. Particularly, withthe current model, one cannot reliably discern the cause ofthe measured enhancement – i.e., increased star formationefficiency or increased molecular gas content – without fur-ther modelling of the state of the gas (e.g., Lagos et al. 2011,2014; Diemer et al. 2018, 2019) . Nonetheless, we will revisitthe effect of the post-merger gas content in Section 4.1.In addition to the possible effects of feedback efficiencyand the sub-resolution properties of the gas reservoir, galaxymorphology could contribute to the aforementioned discrep-ancy between star-forming and passive galaxies. Mergersof bulge-dominated galaxies have been predicted (e.g., Coxet al. 2008) and observed (e.g., Saintonge et al. 2012) tohave lower SFR enhancements. However, exploring the in-terplay between galaxy morphology and SFR enhancementsin mergers within IllustrisTNG is beyond the scope of thiswork. In Section 3.1 we demonstrated that post-mergers exhibitenhanced star formation rates compared to their controls,and tied the strength of the SFR enhancement to galaxyproperties (e.g., z , µ , M (cid:63) , f gas ). All the results presented thusfar treat the post-mergers immediately following the galaxymerger. In this section, we explore the forward evolution ofthe SFR enhancement (or suppression).We trace the evolution of the post-mergers in two dis-tinct ways.(i) Our first method tracks every post-merger forwardin time until another merger occurs ( µ ≥ . ) or we reach z = . Then, at every snapshot, we identify control galaxiesfor the post-merger’s descendants as described in Section2.3. Regenerating the controls independently for each de-scendant is akin to observational studies where we have noprior knowledge of a post-merger’s parents or their proper-ties.(ii) In contrast, our second method traces both thepost-merger and the associated control (identified immedi-ately after the merger) forward in time (similar to numeri-cal simulations of isolated binary galaxy mergers) until ei-ther the control or the post-merger undergoes a merger with µ ≥ . .We stress the subtle differences between the two meth-ods. Calculating the controls at every snapshot indepen- dently compares the merger descendants to similar galax-ies. Therefore, our first method is insensitive to some of theeffects of galaxy mergers (i.e., transitioning between star-forming and passive, quenching). For example, in our firstmethod post-mergers and their controls belong to the sameclass, hence the effects of post-mergers transitioning betweenclasses are suppressed. Additionally, post-mergers with un-resolved SFRs are removed from our sample, thus our firstmethod does not include the effects of quenching. On theother hand, comparing the merger descendants to the asso-ciated control’s descendants at the same redshift highlightsthe difference in the evolution of post-mergers and the con-trols which evolve secularly. Therefore, our second methodis sensitive to post-mergers changing class or quenching. We first examine the evolution of the SFR enhancement andits dependence on the merger mass ratio. Figure 12 showsthe evolution of sSFR and the sSFR enhancements over Gyr following the merger; T merge is a measure of the timesince the most recent merger with µ ≥ . . The left col-umn of Figure 12 depicts the results of independently re-generating the galaxy controls at each snapshot (method i).The merger-induced sSFR enhancement decays following themerger; after ∼ . Gyr the post-mergers’ sSFRs are consis-tent with those of the controls (left middle panel). The bot-tom left panel shows the evolution of Q ( sSFR ) for differentmass ratios computed using method i. While there is a smalldependence in the average Q ( sSFR ) on µ , the µ -dependentenhancement only persists for (cid:46) Myr post-merger ( ∼ Myr given the time resolution of the simulation). After ∼ Myr, the decay of Q ( sSFR ) to normal values is identical forall mass ratios.The middle column of Figure 12 encapsulates the re-sults of our second method. The controls at T merge = Gyrare traced forward in time and compared to the post-mergerdescendants at their respective redshift. Similar to methodi, the SFR enhancement in post-mergers decays over ∼ . Gyr, and the dependence on the merger mass ratio is shortlived. The upturn in Q ( sSFR ) at large T merge is caused by adeviation in the properties of the post-merger descendantsand control’s descendants which renders the control’s de-scendants inadequate control matches for the post-mergerdescendants. Namely, the post-merger descendants and thecontrol descendants do not belong to the same class (i.e.,star-forming, passive) which leads to an unfair comparisonof galaxies and their SFRs.In order to mitigate the effects of the control qualityon the results, we introduce a refined version of method ii,hereafter method ii x . We remove post-merger descendants(and their associated controls) if the control quality does notadhere to the conditions described in section 2.3. Namely,we remove post-merger descendants (and their associatedcontrols) in cases of unacceptably large matching tolerances(i.e., ∆ log ( M (cid:63) ) > = . , 40% in N and r ), unresolved SFRs,and when the post-mergers and their associated controls donot belong to the same class (i.e., star-forming, passive).The results are shown in the right column of Figure 12.Constraining the quality of control matching (particularlythe control and post-merger class) alleviates the spuriousenhancement at T merge > Gyr while maintaining similar
MNRAS , 1–18 (2020) M. H. Hani et al. -10.5-10.0-9.5-9.0 h s S F R i method i ControlPM Q ( s S F R ) T merge [Gyr]0.01.02.0 Q ( s S F R ) µ < . . ≤ µ < . µ ≥ . method ii ControlPM T merge [Gyr] µ < . . ≤ µ < . µ ≥ . method ii x ControlPM T merge [Gyr] µ < . . ≤ µ < . µ ≥ . Figure 12.
The evolution of star formation during the post-merger phase. The three columns (from left to right) show the results of thetracing/control matching methods: method i, method ii, and method ii x , respectively. The top panels shows the running average sSFRfor post-mergers (dark-purple) and their controls (orange) as a function of time after the merger. The middle panels show the evolutionof Q ( sSF R ) for the full post-merger sample, and the bottom panels show Q ( sSF R ) for a sub-sample of post-mergers selected based onthe parents’ mass ratios. The shaded regions represent twice the standard error on the mean in bins of T merge . All methods show thatthe enhancement in sSFR decays following the merger and vanishes after ∼ Myr. Although mergers with different mass ratios induceenhancements of varying strengths, all enhancements decay similarly after ∼ Myr. The enhancement at later T merge shown in methodii (middle column) is driven by the deviation in the post-merger and control properties (namely, different class for post-mergers andcontrols). Ensuring good control quality (method ii x ; right column) removes the spurious enhancements seen at large T merge in method ii. results to method i (decay timescale of the SFR enhance-ment, and the short-lived dependence on µ ).All our methods show that the SFR enhancement de-cays following the merger and vanishes after ∼ Myr, withlittle dependence on mass ratio. Although the mass ratio hasbeen reported to have a significant impact on the extent andtime-scale of observed morphological disturbances of post-mergers (i.e., Lotz et al. 2010a), our post-merger sampleindicates that the evolution of SFR enhancement beyond ∼ Myr may be independent of mass ratio. Nonetheless,the decay timescales presented in this work are broadly con-sistent with results from high resolution idealised simula-tions (e.g., Moreno et al. 2015). We stress a possible caveatwhich may affect our results: Since T merge measures the timesince the most recent merger, it does not have any memoryof a galaxy’s previous history. Therefore, consecutive merg-ers (although rare) may hinder our ability to discern theeffects of mergers with different mass ratios. We next investigate the effect of merger-induced starburstson the evolution of post-merger galaxies. Figure 13 shows the Comparing the results presented in this work with idealised bi-nary galaxy merger simulations is not wholly fair. Idealised sim-ulations traditionally use gas-rich disc galaxies which does notreflect the nature of the post-mergers in IllustrisTNG. evolution of Q ( sSFR ) for post-mergers with different SFRstrengths at T merge = Gyr: starburst galaxies with
SFR > SFR MS + σ MS , starburst galaxies with SFR > SFR MS + σ MS ,and the rest of the star-forming post-mergers. The top, mid-dle, and bottom panels show the results of method i, methodii, and method ii x , respectively. Independent of the method,the weakest starbursts (i.e., star-forming galaxies, or galax-ies with SFR > SFR MS + σ MS ) never lead to a statisticalreduction in SFR following the merger. However, some merg-ers (with the strongest merger-induced SFRs) can evolve tohave statistically suppressed star formation, although thisis dependent on the control matching method. For example,method i (new controls generated at each snapshot) is insen-sitive to galaxy class transitions or quenching. Therefore, noSFR suppression is evident shortly after the merger. How-ever, at ∼ Gyr, the strongest bursts evolve to have slightlysuppressed SFRs. The suppression is caused by the system-atically lower SFRs (compared to typical star-forming con-trols) of previously passive post-merger descendants evolv-ing to be star-forming, an effect previously reported as ‘re-juvenation’ by Rodr´ıguez Montero et al. (2019). The sup-pression vanishes after ∼ Gyr. In method ii (the samecontrols are traced in time alongside the post-mergers), thestrongest merger-induced starbursts evolve to have statis-tically suppressed SFRs compared to their secularly evolv-ing controls beyond ∼ . Gyr. After ∼ Gyr, the mergerdescendants have SFRs that are consistent with their con-trols. We note that method ii provides an unfair compari-
MNRAS000
MNRAS000 , 1–18 (2020) llustrisTNG post-merger galaxies Q ( s S F R ) σ MS burst1 σ MS burstSF m e t h o d i Q ( s S F R ) m e t h o d ii T merge [Gyr]0.01.02.0 Q ( s S F R ) m e t h o d ii x Figure 13.
The evolution of star formation during the post-merger phase for post-mergers with different merger-induced star-burst strengths. From top to bottom, the panels show Q ( sSF R ) for a sub-sample of post-mergers selected based on the strength ofthe SFR at T merge = Gyr and traced using methods i, ii, and ii x ,respectively. The shaded regions represent twice the standard er-ror on the mean in bins of T merge . Independent of the method, theweakest starbursts (i.e., SF, σ MS burst) do not induce suppres-sion in SFR. However, the strongest merger-induced SFRs canevolve to have statistically suppressed star formation. The leveland existence of SFR suppression depend on the merger/controltracing method. Note: The Q ( sSF R )− axis is truncated for visualpurposes; σ bursts exhibit a peak Q ( sSF R ) (cid:39) while σ burstsshow an enhancement of Q ( sSF R ) = . at T merge = Gyr. son of star-forming galaxies to passive or quenched (unre-solved SFRs) galaxies which causes the spurious enhance-ments at late T merge (see Figure 12). Method ii x alleviatesthe unfair comparison in method ii. Consequently, forcingthe post-merger descendants and the control descendants tobelong to the same class conceals the suppression evidentin method ii. The dependence of the SFR suppression onclass matching suggests that merger descendants exhibitingmerger-induced starbursts evolve not only to have statisti-cally suppressed SFRs but also become passive or quenchedon faster timescales than their controls.Our sample shows that the descendants of galaxy merg-ers can evolve to have modestly suppressed SFRs that maypersist for a Gyr or so. Some post-merger descendants can berejuvenated after a period of SFR suppression. However, thelevel and existence of SFR suppression depends on the con-trol matching method (unlike the SFR enhancements whichare very robust to method). The results presented here per- taining to merger-induced starbursts are consistent with theframework where galaxy mergers may drive SFR suppression(e.g., Hopkins et al. 2008). However, a detailed explorationof the connection between galaxy mergers and SFR suppres-sion (i.e., timescales, physical processes, feedback) is beyondthe scope of this work. By taking the different methods wehave 1) shown that the largest starbursts can eventuallysuppress the SFR where the descendants become passive orquenched faster than their controls, but 2) the techniquesused in observations (equivalent to method i and methodii x ) would be largely insensitive to suppression in SFR. In Section 3.1.5 we showed that the SFR enhancement cor-relates strongly with the progenitors’ gas fraction ( f gas ). Yet,our results do not account for the post-merger’s gas contentwhen searching for a control galaxy. Understanding the ef-fect of controlling for the gas content on the presented resultsis a vital test for observational studies, because the choiceof matching parameters can significantly affect the conclu-sions. For example, Scudder et al. (2015) report different cor-relations between SFR and gas fraction depending whetherthey control for gas fraction; i.e., the merger-induced SFRenhancement correlates with the gas fraction only when thesample is not matched in gas fraction. In Figure 11 we al-ready showed that, in agreement with Scudder et al. (2015),there is a correlation between SFR enhancement and gasfraction when gas fraction is not included in the control pa-rameters. Now, we investigate whether the aforementionedresult changes if gas fraction is controlled for.The control sample is regenerated using the single bestmatch in M gas , M (cid:63) , z , N , and r as described in Section2.3. An initial tolerance of . dex is used for M gas ; if nomatches were found, we grow the tolerance by . dex. Allthe other parameters are treated as before (see § M gas -matched control sample. The results remainqualitatively unchanged although the strength of the en-hancement (or suppression) is reduced when controlling for M gas . The results presented here are consistent with the re-sults of Cao et al. (2016), who report enhanced SFRs in theirstar-forming spiral pairs compared to control galaxies withsimilar gas mass.In terms of a direct comparison with observations, itis worth noting that the IllustrisTNG physical model doesnot explicitly track the atomic and molecular fractions ingas cells during the simulation. Therefore, we are unable todiscern the sub-resolution physical state (i.e., atomic frac-tion, molecular fraction) of the gas reservoirs in the post-mergers and their controls. However, knowing that the Il-lustrisTNG physical model attributes star formation to gascells which meet the sub-resolution star formation criterion( SFR ∝ ρ . ), one can use the SFR as a proxy for the molec-ular gas content of galaxies (see Diemer et al. 2019). Themeasured SFR enhancement, which persists even when con-trolling for gas fraction, is suggestive of an enhanced molec- MNRAS , 1–18 (2020) M. H. Hani et al. z )0.01.02.03.0 Q ( s S F R ) µ ) 0.0 0.1 0.2 0.3 0.4 0.5 0.6 f gas M ? [M (cid:12) ])0.01.02.03.0 Q ( s S F R ) T merge [Gyr] without M gas with M gas Figure 14.
The effect of the controlling for gas mass (i.e., gas fraction) on the measured merger-driven SFR enhancement. The shadedregions represent twice the standard error on the mean. The results of TNG300-1 for two different types of matching, with and withoutmatching on gas mass, are shown in magenta and dark-purple, respectively. Controlling for the gas mass reduces the strength of themeasured SFR enhancement (or suppression). Nonetheless the results are consistent between the two control matching techniques. ular gas content in post-mergers. We will explore gas frac-tion evolution during galaxy mergers more extensively inthe future. Our results are qualitatively consistent with ob-servational studies reporting that post-mergers are home toan enhanced molecular gas content (e.g., Pan et al. 2018;Violino et al. 2018). Additionally, our results are broadlyconsistent with the high resolution simulations of Morenoet al. (2019) who showed that the cold-dense gas content ofgalaxies is enhanced during the pair phase and immediatelybefore the coalescence.
A common effect of numerical resolution on star formationarises from the correlation of star formation with dense gas:i.e., stars are spawned from gas cells above a given densitythreshold (see § > Myr for galaxies inthe same mass range as the post-mergers studied here.
A particular strength of numerical simulations of galaxymergers is the ability to track the evolution of galaxy merg-ers and quantify the strength and temporal extent of SFRenhancement and their dependence on galaxy properties.On the contrary, observational studies are unable to accu-rately follow the evolution of galaxy mergers beyond coales-cence; therefore observational studies of post-mergers group
MNRAS000
MNRAS000 , 1–18 (2020) llustrisTNG post-merger galaxies z )0.01.02.03.0 Q ( s S F R ) µ ) 0.0 0.1 0.2 0.3 0.4 0.5 0.6 f gas M ? [M (cid:12) ])0.01.02.03.0 Q ( s S F R ) T merge [Gyr] Figure 15.
The effect of the simulation resolution on the results presented in this work. The shaded regions represent twice the standarderror on the mean. The results of TNG100-1, TNG100-2, and TNG300-1 are broadly consistent. The effect of resolution are mostprominent in the timescale on which the Q ( sSF R ) decays in the post-merger phase. The enhancement in the lower resolution simulations(i.e., TNG100-2 and TNG300-1) wanes over ∼ Myr vs. ∼ Myr for the higher resolution simulation (i.e., TNG100-1). -10.5-10.0-9.5 h s S F R i ControlPM50100150200250 r [kpc]0.01.02.0 Q ( s S F R ) Pair phase 0.0 0.5 1.0 1.5 2.0 T merge [Gyr]PM phase Figure 16.
Reconciling the observable evolution of mergers during the galaxy pair phase (left panels; Patton et al. 2020) and theevolution of post-merger systems (right panels; see Figure 12) in TNG300-1. The controls are recalculated independently for each mergerremnant as in the left column of Figure 12 (i.e. method i). The shaded regions (although remarkably small) represent twice the standarderror on the mean. As the separation of galaxy pairs ( r ) decreases, the galaxies experience an enhancement in their SFRs. Followingcoalescence, the SFR enhancement decays to be consistent with the controls.MNRAS , 1–18 (2020) M. H. Hani et al. the merger remnants together disregarding their evolution (e.g., Ellison et al. 2015; Thorp et al. 2019). On the otherhand, observational studies use the projected galaxy separa-tion as a proxy for time evolution during the interaction/pairphase of galaxy mergers (e.g., Ellison et al. 2013; Pattonet al. 2016).Patton et al. (2020) studied galaxy pairs in Illustris-TNG applying a methodology akin to observations, and re-ported an enhancement in SFR for close pairs which is inqualitative agreement with observational galaxy pairs stud-ies using SDSS. Knowing that our control matching strategyis different than that used in Patton et al. (2020), we repro-duce the results of Patton et al. (2020) following the controlmatching methods used in this work (i.e., ignoring galax-ies with unresolved SFRs, matching within the same class,limiting r sep > ). Figure 16 combines the results of Pattonet al. (2020) with our results to form a comprehensive pictureof galaxy mergers and their evolution in Illustris-TNG. Asthe separation of galaxy pairs ( r ) decreases (pericentric pas-sage), the galaxies experience an enhancement in their SFRs.The enhancement then decays as the galaxies move to theirrespective apocenters. Following coalescence, the SFR en-hancement decays to be consistent with the controls within ∼ Myr in TNG300-1. We note that the results of thepair phase are possibly underestimated (overestimated) atsmall (large) separations because, unlike the time domain,at a given separation ( r ) we would include the contributionfrom galaxies which haven not (have) undergone their firstpericentric passage. We present a large sample of post-mergers simulated in acosmological environment (selected from the IllustrisTNGsimulation suite) and analysed in an observationally moti-vated scheme (including the use of statistical controls) todemonstrate the effects of galaxy mergers on star formation.We quantify the strength and temporal extent of merger-driven SFR enhancements. Additionally, we tie the enhance-ment to various galactic properties and hence physical mech-anisms. The large sample of post-mergers presented hereprovides a powerful tool to investigate the detailed evolutionof galaxy mergers. We demonstrated that, for TNG300-1: • SFR enhancement:
The SFRs of post-merger galax-ies, shortly following the merger (i.e., at the first snapshotfollowing the merger), are enhanced by, on average, a factorof ∼ compared to their controls. The star-forming post-mergers dominate this enhancement, with a factor of ∼ higher SFRs than their controls. Passive post-mergers ex-hibit statistically consistent SFRs compared to their con-trols. • The redshift evolution of merger driven SFR en-hancements:
Our post-merger sample exhibits a robust en-hancement in SFR (factor of ∼ ) for ≤ z ≤ , with noevident evolution with redshift. The global enhancement isdriven by the star-forming post-merger galaxies, while the Observational studies of post-mergers depend on morphologi-cal selection which are most sensitive to short timescales beyondcoalescence (e.g., Lotz et al. 2008, 2010a,b) passive post-mergers have SFRs which are consistent withthose of their controls. • The effect of the mass ratio:
The merger-drivenSFR enhancement during the post-merger phase, on aver-age, mildly depends on the mass ratio of the parent galax-ies’ merger. Minor mergers ( . ≤ µ < . ) drive modestenhancements (factor of ∼ ) while major mergers ( µ ≥ . )induce stronger SFR enhancements (factor of ∼ . ). • The contribution of minor mergers to themerger-induced SFR budget:
Although major mergers( µ ≥ . ) drive stronger SFR bursts, they are vastly out-numbered by minor mergers ( . ≤ µ < . ) which, in spiteof their weaker SFR enhancement, still contribute ∼ ofthe total merger-driven SFR enhancement in TNG300-1. • The dependence on stellar mass:
The SFR en-hancements in post-merger galaxies show a strong depen-dence on the hosts’ stellar masses. Post-mergers with stel-lar mass log ( M (cid:63) / M (cid:12) ) ≤ . exhibit SFR enhancementsof a factor of ∼ with declining enhancements for largerstellar masses. For post-mergers with stellar masses above log ( M (cid:63) / M (cid:12) ) > . , the dependence on SFR enhancementdisappears. The dependence of SFR enhancement on stel-lar mass is driven by the star-forming post-merger sample;passive post-mergers have SFRs which are, on average, con-sistent with the SFRs of the controls. • The effects of the merger gas content:
Merger in-duced star-formation in our post-merger sample (both star-forming and passive) correlates strongly with the mergergas fraction. The descendants of the mergers with the high-est gas fractions exhibit the strongest enhancements ( ∼ . − . ) while the descendants of the lowest gas fractionmergers show suppression in SFR when compared to theircontrols. • The effect of controlling for gas fraction:
The SFRenhancements in star-forming post-mergers are strongerthan their passive counterparts even at the same gas frac-tion. Moreover, even when constraining the post-merger gasfraction in the control-matching process, SFR enhancementspersist, albeit to a weaker extent. The discrepancy in SFRenhancement between the star-forming and passive post-mergers, and the persistence of the SFR enhancement whencontrolling for the post-merger gas content, are suggestiveof additional processes being at play in driving the SFRenhancement (e.g., gas phase, feedback, star-formation effi-ciency). • The evolution of post-mergers beyond coales-cence:
The SFR enhancements in post-mergers decay on atimescale of ∼ . Gyr. While the strength of the merger-driven SFR enhancement within ∼ − Myr post-coalescence is dependent on the merger mass ratio, the decayin SFR enhancement is independent of mass ratio beyondthis timescale. • The role of galaxy mergers in suppressing starformation:
Although galaxy mergers do not globally sup-press star formation (i.e., quenching), the strongest merger-driven starburst galaxies evolve to be passive/quenched onfaster timescales than their controls. • Effects of simulation resolution:
The results re-ported here are qualitatively consistent for the different Il-lustrisTNG simulation boxes (i.e., resolutions).
MNRAS000
MNRAS000 , 1–18 (2020) llustrisTNG post-merger galaxies ACKNOWLEDGMENTS
The authors thank Connor Bottrell, Jorge Moreno, andJoanna Woo for their insightful comments and helpful dis-cussions. MHH acknowledges the receipt of a Vanier CanadaGraduate Scholarship. SLE and DRP acknowledges the re-ceipt of an NSERC Discovery Grant. HG acknowledges Mi-tacs Globalink. This research was enabled, in part, by thecomputing resources provided by WestGrid and ComputeCanada.
REFERENCES
Barnes J. E., Hernquist L. E., 1991, ApJ, 370, L65Bluck A. F. L., Mendel J. T., Ellison S. L., Moreno J., Simard L.,Patton D. R., Starkenburg E., 2014, MNRAS, 441, 599Blumenthal K. A., Barnes J. E., 2018, MNRAS, 479, 3952Blumenthal K. A., et al., 2019, MNRAS, p. 3093Bournaud F., Jog C. J., Combes F., 2005, A&A, 437, 69Bundy K., et al., 2010, ApJ, 719, 1969Bustamante S., Sparre M., Springel V., Grand R. J. J., 2018,MNRAS, 479, 3381Cao C., et al., 2016, ApJS, 222, 16Casteels K. R. V., et al., 2014, MNRAS, 445, 1157Chabrier G., 2003, PASP, 115, 763Combes F., Garc´ıa-Burillo S., Braine J., Schinnerer E., Walter F.,Colina L., 2013, A&A, 550, A41Cox T. J., Primack J., Jonsson P., Somerville R. S., 2004, ApJ,607, L87Cox T. J., Jonsson P., Somerville R. S., Primack J. R., Dekel A.,2008, MNRAS, 384, 386Dav´e R., Angl´es-Alc´azar D., Narayanan D., Li Q., RafieferantsoaM. H., Appleby S., 2019, MNRAS, 486, 2827Dekel A., Sari R., Ceverino D., 2009, ApJ, 703, 785Di Matteo T., Springel V., Hernquist L., 2005, Nature, 433, 604Di Matteo P., Combes F., Melchior A. L., Semelin B., 2007, A&A,468, 61Di Matteo P., Bournaud F., Martig M., Combes F., MelchiorA. L., Semelin B., 2008, A&A, 492, 31Diemer B., et al., 2018, ApJS, 238, 33Diemer B., et al., 2019, MNRAS, 487, 1529Donnari M., et al., 2019, MNRAS, 485, 4817Dutta R., Srianand R., Gupta N., 2018, MNRAS, 480, 947Dutta R., Srianand R., Gupta N., 2019, MNRAS, 489, 1099Ellison S. L., Patton D. R., Simard L., McConnachie A. W., 2008,AJ, 135, 1877Ellison S. L., Patton D. R., Mendel J. T., Scudder J. M., 2011,MNRAS, 418, 2043Ellison S. L., Mendel J. T., Patton D. R., Scudder J. M., 2013,MNRAS, 435, 3627Ellison S. L., Fertig D., Rosenberg J. L., Nair P., Simard L., Tor-rey P., Patton D. R., 2015, MNRAS, 448, 221Ellison S. L., Catinella B., Cortese L., 2018, MNRAS, 478, 3447Ellison S. L., Viswanathan A., Patton D. R., Bottrell C., Mc-Connachie A. W., Gwyn S., Cuillandre J.-C., 2019, MNRAS,487, 2491Faucher-Gigu`ere C.-A., Lidz A., Zaldarriaga M., Hernquist L.,2009, ApJ, 703, 1416Hani M. H., Sparre M., Ellison S. L., Torrey P., Vogelsberger M.,2018, MNRAS, 475, 1160Hayward C. C., Hopkins P. F., 2017, MNRAS, 465, 1682Hernquist L., 1989, Nature, 340, 687Hopkins P. F., Quataert E., 2010, MNRAS, 407, 1529Hopkins P. F., Hernquist L., Cox T. J., Kereˇs D., 2008, ApJS,175, 356Huertas-Company M., et al., 2019, MNRAS, 489, 1859 Johansson P. H., Naab T., Burkert A., 2009, ApJ, 690, 802Kaviraj S., et al., 2013, MNRAS, 429, L40Kewley L. J., Geller M. J., Barton E. J., 2006, AJ, 131, 2004Knapen J. H., Cisternas M., Querejeta M., 2015, MNRAS, 454,1742Lacey C., Cole S., 1993, MNRAS, 262, 627Lagos C. D. P., Baugh C. M., Lacey C. G., Benson A. J., KimH.-S., Power C., 2011, MNRAS, 418, 1649Lagos C. D. P., Baugh C. M., Zwaan M. A., Lacey C. G.,Gonzalez-Perez V., Power C., Swinbank A. M., van KampenE., 2014, MNRAS, 440, 920Lin L., et al., 2007, ApJ, 660, L51Lin L., et al., 2008, ApJ, 681, 232Lofthouse E. K., Kaviraj S., Conselice C. J., Mortlock A., HartleyW., 2017, MNRAS, 465, 2895L´opez-Sanjuan C., et al., 2013, A&A, 553, A78Lotz J. M., Jonsson P., Cox T. J., Primack J. R., 2008, MNRAS,391, 1137Lotz J. M., Jonsson P., Cox T. J., Primack J. R., 2010a, MNRAS,404, 575Lotz J. M., Jonsson P., Cox T. J., Primack J. R., 2010b, MNRAS,404, 590Lotz J. M., Jonsson P., Cox T. J., Croton D., Primack J. R.,Somerville R. S., Stewart K., 2011, ApJ, 742, 103Madau P., Dickinson M., 2014, ARA&A, 52, 415Marinacci F., et al., 2018, MNRAS, 480, 5113Martin C. L., 2005, ApJ, 621, 227Mihos J. C., Hernquist L., 1996, ApJ, 464, 641Montuori M., Di Matteo P., Lehnert M. D., Combes F., SemelinB., 2010, A&A, 518, A56Moreno J., Bluck A. F. L., Ellison S. L., Patton D. R., Torrey P.,Moster B. P., 2013, MNRAS, 436, 1765Moreno J., Torrey P., Ellison S. L., Patton D. R., Bluck A. F. L.,Bansal G., Hernquist L., 2015, MNRAS, 448, 1107Moreno J., et al., 2019, MNRAS, 485, 1320Naiman J. P., et al., 2018, MNRAS, 477, 1206Nelson D., et al., 2018, MNRAS, 475, 624Nelson D., et al., 2019a, Computational Astrophysics and Cos-mology, 6, 2Nelson D., et al., 2019b, MNRAS, 490, 3234Pan H.-A., et al., 2018, ApJ, 868, 132Patton D. R., Torrey P., Ellison S. L., Mendel J. T., ScudderJ. M., 2013, MNRAS, 433, L59Patton D. R., Qamar F. D., Ellison S. L., Bluck A. F. L., SimardL., Mendel J. T., Moreno J., Torrey P., 2016, MNRAS, 461,2589Patton D. R., et al., 2020, MNRAS, submittedPillepich A., et al., 2018a, MNRAS, 473, 4077Pillepich A., et al., 2018b, MNRAS, 475, 648Pillepich A., et al., 2019, MNRAS, 490, 3196Pop A.-R., Pillepich A., Amorisco N. C., Hernquist L., 2018, MN-RAS, 480, 1715Rodighiero G., et al., 2011, ApJ, 739, L40Rodriguez-Gomez V., et al., 2015, MNRAS, 449, 49Rodriguez-Gomez V., et al., 2019, MNRAS, 483, 4140Rodr´ıguez Montero F., Dav´e R., Wild V., Angl´es-Alc´azar D.,Narayanan D., 2019, MNRAS, 490, 2139Rupke D. S., Veilleux S., Sanders D. B., 2005a, ApJS, 160, 115Rupke D. S., Veilleux S., Sanders D. B., 2005b, ApJ, 632, 751Saintonge A., et al., 2012, ApJ, 758, 73Sales L. V., Navarro J. F., Abadi M. G., Steinmetz M., 2007,MNRAS, 379, 1464Satyapal S., Ellison S. L., McAlpine W., Hickox R. C., PattonD. R., Mendel J. T., 2014, MNRAS, 441, 1297Scudder J. M., Ellison S. L., Torrey P., Patton D. R., MendelJ. T., 2012, MNRAS, 426, 549Scudder J. M., Ellison S. L., Momjian E., Rosenberg J. L., TorreyP., Patton D. R., Fertig D., Mendel J. T., 2015, MNRAS, 449,MNRAS , 1–18 (2020) M. H. Hani et al. A TEX file prepared bythe author. MNRAS000